From: ericm Date: Tue, 25 Nov 2003 11:34:23 +0000 (+0000) Subject: merge b_devel to b_eq: 20031125. kernel part only. X-Git-Tag: v1_7_0_51~2^9~136 X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=adc507078813b916b246c0464252944be038dfff;p=fs%2Flustre-release.git merge b_devel to b_eq: 20031125. kernel part only. --- diff --git a/lnet/include/linux/kp30.h b/lnet/include/linux/kp30.h index 0acec81..3e6d5e3 100644 --- a/lnet/include/linux/kp30.h +++ b/lnet/include/linux/kp30.h @@ -74,8 +74,10 @@ extern unsigned int portal_cerror; #ifdef __KERNEL__ # include /* THREAD_SIZE */ -#else -# define THREAD_SIZE 8192 +#else +# ifndef THREAD_SIZE /* x86_64 has THREAD_SIZE in userspace */ +# define THREAD_SIZE 8192 +# endif #endif #define LUSTRE_TRACE_SIZE (THREAD_SIZE >> 5) @@ -277,13 +279,17 @@ do { \ #define PORTAL_VMALLOC_SIZE 16384 +#ifndef GFP_MEMALLOC +#define GFP_MEMALLOC 0 +#endif + #define PORTAL_ALLOC(ptr, size) \ do { \ LASSERT (!in_interrupt()); \ if ((size) > PORTAL_VMALLOC_SIZE) \ (ptr) = vmalloc(size); \ else \ - (ptr) = kmalloc((size), GFP_NOFS); \ + (ptr) = kmalloc((size), (GFP_KERNEL | GFP_MEMALLOC)); \ if ((ptr) == NULL) \ CERROR("PORTALS: out of memory at %s:%d (tried to alloc '"\ #ptr "' = %d)\n", __FILE__, __LINE__, (int)(size));\ @@ -312,10 +318,14 @@ do { \ s, (ptr), atomic_read(&portal_kmemory)); \ } while (0) +#ifndef SLAB_MEMALLOC +#define SLAB_MEMALLOC 0 +#endif + #define PORTAL_SLAB_ALLOC(ptr, slab, size) \ do { \ LASSERT(!in_interrupt()); \ - (ptr) = kmem_cache_alloc((slab), SLAB_KERNEL); \ + (ptr) = kmem_cache_alloc((slab), (SLAB_KERNEL | SLAB_MEMALLOC)); \ if ((ptr) == NULL) { \ CERROR("PORTALS: out of memory at %s:%d (tried to alloc" \ " '" #ptr "' from slab '" #slab "')\n", __FILE__, \ @@ -690,6 +700,9 @@ typedef struct { long lwte_p2; long lwte_p3; long lwte_p4; +#if BITS_PER_LONG > 32 + long lwte_pad; +#endif } lwt_event_t; #if LWT_SUPPORT @@ -1112,14 +1125,19 @@ void kportal_put_ni (int nal); # endif #endif -#if (BITS_PER_LONG == 32 || __WORDSIZE == 32) +#if defined(__x86_64__) +# define LPU64 "%Lu" +# define LPD64 "%Ld" +# define LPX64 "%#Lx" +# define LPSZ "%lu" +# define LPSSZ "%ld" +#elif (BITS_PER_LONG == 32 || __WORDSIZE == 32) # define LPU64 "%Lu" # define LPD64 "%Ld" # define LPX64 "%#Lx" # define LPSZ "%u" # define LPSSZ "%d" -#endif -#if (BITS_PER_LONG == 64 || __WORDSIZE == 64) +#elif (BITS_PER_LONG == 64 || __WORDSIZE == 64) # define LPU64 "%lu" # define LPD64 "%ld" # define LPX64 "%#lx" diff --git a/lnet/klnds/socklnd/socklnd_cb.c b/lnet/klnds/socklnd/socklnd_cb.c index 6cd4e73..b6fdc18 100644 --- a/lnet/klnds/socklnd/socklnd_cb.c +++ b/lnet/klnds/socklnd/socklnd_cb.c @@ -891,6 +891,11 @@ ksocknal_queue_tx_locked (ksock_tx_t *tx, ksock_conn_t *conn) spin_lock_irqsave (&sched->kss_lock, flags); + conn->ksnc_tx_deadline = jiffies + + ksocknal_data.ksnd_io_timeout * HZ; + mb(); + /* Extend deadline BEFORE tx is enqueued */ + list_add_tail (&tx->tx_list, &conn->ksnc_tx_queue); if (conn->ksnc_tx_ready && /* able to send */ @@ -2118,28 +2123,30 @@ ksocknal_hello (struct socket *sock, ptl_nid_t *nid, int *type, __u64 *incarnati } int -ksocknal_setup_sock (struct socket *sock) +ksocknal_setup_sock (struct socket *sock) { mm_segment_t oldmm = get_fs (); int rc; int option; struct linger linger; + sock->sk->allocation = GFP_NOFS; + /* Ensure this socket aborts active sends immediately when we close * it. */ - + linger.l_onoff = 0; linger.l_linger = 0; set_fs (KERNEL_DS); - rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER, + rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER, (char *)&linger, sizeof (linger)); set_fs (oldmm); if (rc != 0) { CERROR ("Can't set SO_LINGER: %d\n", rc); return (rc); } - + option = -1; set_fs (KERNEL_DS); rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_LINGER2, diff --git a/lnet/klnds/toelnd/toenal.c b/lnet/klnds/toelnd/toenal.c index a85b281..07b93cb 100644 --- a/lnet/klnds/toelnd/toenal.c +++ b/lnet/klnds/toelnd/toenal.c @@ -210,7 +210,7 @@ error: int ktoenal_close_sock(ptl_nid_t nid) { - long flags; + unsigned long flags; ksock_conn_t *conn; LIST_HEAD (death_row); struct list_head *tmp; diff --git a/lnet/klnds/toelnd/toenal_cb.c b/lnet/klnds/toelnd/toenal_cb.c index f7fa794..3af9e33 100644 --- a/lnet/klnds/toelnd/toenal_cb.c +++ b/lnet/klnds/toelnd/toenal_cb.c @@ -140,7 +140,7 @@ ktoenal_dist(nal_cb_t *nal, ptl_nid_t nid, unsigned long *dist) ksock_ltx_t * ktoenal_get_ltx (int may_block) { - long flags; + unsigned long flags; ksock_ltx_t *ltx = NULL; for (;;) @@ -268,7 +268,7 @@ ktoenal_recvmsg(struct file *sock, struct iovec *iov, int niov, int toread) } void -ktoenal_process_transmit (ksock_conn_t *conn, long *irq_flags) +ktoenal_process_transmit (ksock_conn_t *conn, unsigned long *irq_flags) { ksock_tx_t *tx = list_entry (conn->ksnc_tx_queue.next, ksock_tx_t, tx_list); int rc; @@ -358,7 +358,7 @@ ktoenal_process_transmit (ksock_conn_t *conn, long *irq_flags) void ktoenal_launch_packet (ksock_conn_t *conn, ksock_tx_t *tx) { - long flags; + unsigned long flags; int nob = tx->tx_nob; struct iovec *iov = tx->tx_iov; int niov = 1; @@ -541,7 +541,7 @@ ktoenal_fmb_callback (void *arg, int error) ksock_fmb_t *fmb = (ksock_fmb_t *)arg; ptl_hdr_t *hdr = (ptl_hdr_t *) page_address(fmb->fmb_pages[0]); ksock_conn_t *conn; - long flags; + unsigned long flags; CDEBUG (D_NET, "routed packet from "LPX64" to "LPX64": %d\n", hdr->src_nid, hdr->dest_nid, error); @@ -790,7 +790,7 @@ ktoenal_new_packet (ksock_conn_t *conn, int nob_to_skip) } void -ktoenal_process_receive (ksock_conn_t *conn, long *irq_flags) +ktoenal_process_receive (ksock_conn_t *conn, unsigned long *irq_flags) { ksock_fmb_t *fmb; int len; diff --git a/lnet/libcfs/debug.c b/lnet/libcfs/debug.c index f27924e..6e2c1ca 100644 --- a/lnet/libcfs/debug.c +++ b/lnet/libcfs/debug.c @@ -775,7 +775,6 @@ portals_debug_msg(int subsys, int mask, char *file, const char *fn, /* If this is the first time, leave a marker in the * output */ debug_daemon_state.overlapped = 1; - ap = NULL; format = "DEBUG MARKER: Debug buffer overlapped\n"; printk(KERN_ERR "LustreError: debug daemon buffer " "overlapped\n"); @@ -982,7 +981,11 @@ extern int lookup_symbol(unsigned long address, char *buf, int buflen); char *portals_debug_dumpstack(void) { +#if defined(__x86_64__) + unsigned long esp = current->thread.rsp; +#else unsigned long esp = current->thread.esp; +#endif unsigned long *stack = (unsigned long *)&esp; int size; unsigned long addr; @@ -1012,7 +1015,7 @@ char *portals_debug_dumpstack(void) /* fix length + sizeof('\0') */ <= pbuf + strlen(buffer) + 28 + 1) break; - size = sprintf(pbuf, "([<%08lx>] %s (0x%x)) ", + size = sprintf(pbuf, "([<%08lx>] %s (0x%p)) ", addr, buffer, stack-1); } pbuf += size; diff --git a/lnet/libcfs/module.c b/lnet/libcfs/module.c index 6432070..7c0cafc 100644 --- a/lnet/libcfs/module.c +++ b/lnet/libcfs/module.c @@ -233,7 +233,7 @@ kportal_router_cmd(struct portals_cfg *pcfg, void * private) CDEBUG (D_IOCTL, "Removing routes via [%d] "LPU64" : "LPU64" - "LPU64"\n", pcfg->pcfg_gw_nal, pcfg->pcfg_nid, pcfg->pcfg_nid2, pcfg->pcfg_nid3); - err = kportal_del_route (pcfg->pcfg_nal, pcfg->pcfg_nid, + err = kportal_del_route (pcfg->pcfg_gw_nal, pcfg->pcfg_nid, pcfg->pcfg_nid2, pcfg->pcfg_nid3); break; diff --git a/lnet/ulnds/procapi.c b/lnet/ulnds/procapi.c index 206aea0..2a3fbd8 100644 --- a/lnet/ulnds/procapi.c +++ b/lnet/ulnds/procapi.c @@ -48,7 +48,7 @@ * forwards a packaged api call from the 'api' side to the 'library' * side, and collects the result */ -static int procbridge_forward(nal_t *n, int id, void *args, ptl_size_t args_len, +static int procbridge_forward(nal_t *n, int id, void *args, size_t args_len, void *ret, ptl_size_t ret_len) { bridge b = (bridge) n->nal_data; @@ -98,7 +98,7 @@ static int procbridge_shutdown(nal_t *n, int ni) /* Function: validate * useless stub */ -static int procbridge_validate(nal_t *nal, void *base, ptl_size_t extent) +static int procbridge_validate(nal_t *nal, void *base, size_t extent) { return(0); } diff --git a/lnet/ulnds/proclib.c b/lnet/ulnds/proclib.c index 2424145..2627253 100644 --- a/lnet/ulnds/proclib.c +++ b/lnet/ulnds/proclib.c @@ -47,7 +47,7 @@ static int nal_write(nal_cb_t *nal, void *private, user_ptr dst_addr, void *src_addr, - ptl_size_t len) + size_t len) { memcpy(dst_addr, src_addr, len); return 0; @@ -64,7 +64,7 @@ static int nal_read(nal_cb_t * nal, } static void *nal_malloc(nal_cb_t *nal, - ptl_size_t len) + size_t len) { void *buf = malloc(len); return buf; @@ -72,7 +72,7 @@ static void *nal_malloc(nal_cb_t *nal, static void nal_free(nal_cb_t *nal, void *buf, - ptl_size_t len) + size_t len) { free(buf); } diff --git a/lnet/ulnds/socklnd/procapi.c b/lnet/ulnds/socklnd/procapi.c index 206aea0..2a3fbd8 100644 --- a/lnet/ulnds/socklnd/procapi.c +++ b/lnet/ulnds/socklnd/procapi.c @@ -48,7 +48,7 @@ * forwards a packaged api call from the 'api' side to the 'library' * side, and collects the result */ -static int procbridge_forward(nal_t *n, int id, void *args, ptl_size_t args_len, +static int procbridge_forward(nal_t *n, int id, void *args, size_t args_len, void *ret, ptl_size_t ret_len) { bridge b = (bridge) n->nal_data; @@ -98,7 +98,7 @@ static int procbridge_shutdown(nal_t *n, int ni) /* Function: validate * useless stub */ -static int procbridge_validate(nal_t *nal, void *base, ptl_size_t extent) +static int procbridge_validate(nal_t *nal, void *base, size_t extent) { return(0); } diff --git a/lnet/ulnds/socklnd/proclib.c b/lnet/ulnds/socklnd/proclib.c index 2424145..2627253 100644 --- a/lnet/ulnds/socklnd/proclib.c +++ b/lnet/ulnds/socklnd/proclib.c @@ -47,7 +47,7 @@ static int nal_write(nal_cb_t *nal, void *private, user_ptr dst_addr, void *src_addr, - ptl_size_t len) + size_t len) { memcpy(dst_addr, src_addr, len); return 0; @@ -64,7 +64,7 @@ static int nal_read(nal_cb_t * nal, } static void *nal_malloc(nal_cb_t *nal, - ptl_size_t len) + size_t len) { void *buf = malloc(len); return buf; @@ -72,7 +72,7 @@ static void *nal_malloc(nal_cb_t *nal, static void nal_free(nal_cb_t *nal, void *buf, - ptl_size_t len) + size_t len) { free(buf); } diff --git a/lnet/ulnds/socklnd/tcplnd.c b/lnet/ulnds/socklnd/tcplnd.c index 3e9b92f..012447f 100644 --- a/lnet/ulnds/socklnd/tcplnd.c +++ b/lnet/ulnds/socklnd/tcplnd.c @@ -130,8 +130,8 @@ int tcpnal_recv(nal_cb_t *n, lib_msg_t *cookie, unsigned int niov, struct iovec *iov, - ptl_size_t mlen, - ptl_size_t rlen) + size_t mlen, + size_t rlen) { int i; diff --git a/lnet/ulnds/tcplnd.c b/lnet/ulnds/tcplnd.c index 3e9b92f..012447f 100644 --- a/lnet/ulnds/tcplnd.c +++ b/lnet/ulnds/tcplnd.c @@ -130,8 +130,8 @@ int tcpnal_recv(nal_cb_t *n, lib_msg_t *cookie, unsigned int niov, struct iovec *iov, - ptl_size_t mlen, - ptl_size_t rlen) + size_t mlen, + size_t rlen) { int i; diff --git a/lustre/kernel_patches/patches/add_page_private-2.4.19-pre1.patch b/lustre/kernel_patches/patches/add_page_private-2.4.19-pre1.patch new file mode 100644 index 0000000..a7e087d --- /dev/null +++ b/lustre/kernel_patches/patches/add_page_private-2.4.19-pre1.patch @@ -0,0 +1,15 @@ + include/linux/mm.h | 1 + + 1 files changed, 1 insertion(+) + +Index: linux-2.4.19-pre1/include/linux/mm.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/mm.h 2003-11-20 19:24:06.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/mm.h 2003-11-21 02:00:13.000000000 +0300 +@@ -162,6 +162,7 @@ + wait_queue_head_t wait; /* Page locked? Stand in line... */ + struct page **pprev_hash; /* Complement to *next_hash. */ + struct buffer_head * buffers; /* Buffer maps us to a disk block. */ ++ unsigned long private; + void *virtual; /* Kernel virtual address (NULL if + not kmapped, ie. highmem) */ + struct zone_struct *zone; /* Memory zone we are in. */ diff --git a/lustre/kernel_patches/patches/exports_2.4.19-pre1.patch b/lustre/kernel_patches/patches/exports_2.4.19-pre1.patch new file mode 100644 index 0000000..9c37bbc --- /dev/null +++ b/lustre/kernel_patches/patches/exports_2.4.19-pre1.patch @@ -0,0 +1,62 @@ + + + + fs/ext3/Makefile | 4 +++- + fs/ext3/super.c | 2 +- + include/linux/fs.h | 1 + + kernel/ksyms.c | 5 +++++ + 4 files changed, 10 insertions(+), 2 deletions(-) + +Index: linux-2.4.19-pre1/fs/ext3/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/Makefile 2003-11-21 02:39:49.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/Makefile 2003-11-21 02:41:04.000000000 +0300 +@@ -9,6 +9,8 @@ + + O_TARGET := ext3.o + ++export-objs := super.o inode.o ++ + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) +Index: linux-2.4.19-pre1/fs/ext3/super.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/super.c 2003-11-21 02:39:49.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/super.c 2003-11-21 02:41:04.000000000 +0300 +@@ -1744,7 +1744,7 @@ + unregister_filesystem(&ext3_fs_type); + } + +-EXPORT_NO_SYMBOLS; ++EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +Index: linux-2.4.19-pre1/include/linux/fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs.h 2003-11-21 02:39:49.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs.h 2003-11-21 02:41:41.000000000 +0300 +@@ -984,6 +984,7 @@ + extern struct vfsmount *kern_mount(struct file_system_type *); + extern int may_umount(struct vfsmount *); + extern long do_mount(char *, char *, char *, unsigned long, void *); ++struct vfsmount *do_kern_mount(char *type, int flags, char *name, void *data); + + #define kern_umount mntput + +Index: linux-2.4.19-pre1/kernel/ksyms.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/ksyms.c 2003-11-21 02:39:49.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/ksyms.c 2003-11-21 02:41:04.000000000 +0300 +@@ -271,6 +271,10 @@ + EXPORT_SYMBOL(lock_may_write); + EXPORT_SYMBOL(dcache_readdir); + ++/* lustre */ ++EXPORT_SYMBOL(panic_notifier_list); ++EXPORT_SYMBOL(do_kern_mount); ++ + /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */ + EXPORT_SYMBOL(default_llseek); + EXPORT_SYMBOL(dentry_open); diff --git a/lustre/kernel_patches/patches/exports_2.4.19-suse.patch b/lustre/kernel_patches/patches/exports_2.4.19-suse.patch new file mode 100644 index 0000000..feaeec6 --- /dev/null +++ b/lustre/kernel_patches/patches/exports_2.4.19-suse.patch @@ -0,0 +1,53 @@ + fs/ext3/Makefile | 2 ++ + fs/ext3/super.c | 2 +- + include/linux/fs.h | 1 + + kernel/ksyms.c | 4 ++++ + 4 files changed, 8 insertions(+), 1 deletion(-) + +--- linux/fs/ext3/Makefile~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/fs/ext3/Makefile Wed Apr 9 10:19:53 2003 +@@ -9,6 +9,8 @@ + + O_TARGET := ext3.o + ++export-objs := super.o inode.o ++ + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) +--- linux/fs/ext3/super.c~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/fs/ext3/super.c Wed Apr 9 10:19:53 2003 +@@ -1769,7 +1769,7 @@ static void __exit exit_ext3_fs(void) + unregister_filesystem(&ext3_fs_type); + } + +-EXPORT_NO_SYMBOLS; ++EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +--- linux/include/linux/fs.h~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/include/linux/fs.h Wed Apr 9 10:19:53 2003 +@@ -1020,6 +1020,7 @@ extern int unregister_filesystem(struct + extern struct vfsmount *kern_mount(struct file_system_type *); + extern int may_umount(struct vfsmount *); + extern long do_mount(char *, char *, char *, unsigned long, void *); ++struct vfsmount *do_kern_mount(const char *type, int flags, char *name, void *data); + + #define kern_umount mntput + +--- linux/kernel/ksyms.c~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/kernel/ksyms.c Wed Apr 9 10:19:53 2003 +@@ -308,6 +308,10 @@ EXPORT_SYMBOL(dcache_dir_fsync); + EXPORT_SYMBOL(dcache_readdir); + EXPORT_SYMBOL(dcache_dir_ops); + ++/* lustre */ ++EXPORT_SYMBOL(pagecache_lock_cacheline); ++EXPORT_SYMBOL(do_kern_mount); ++ + /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */ + EXPORT_SYMBOL(default_llseek); + EXPORT_SYMBOL(dentry_open); + +_ diff --git a/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch b/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch new file mode 100644 index 0000000..748671f --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch @@ -0,0 +1,2528 @@ + fs/ext3/Makefile | 2 + fs/ext3/dir.c | 299 +++++++++ + fs/ext3/file.c | 3 + fs/ext3/hash.c | 215 ++++++ + fs/ext3/namei.c | 1387 ++++++++++++++++++++++++++++++++++++++++----- + fs/ext3/super.c | 7 + include/linux/ext3_fs.h | 85 ++ + include/linux/ext3_fs_sb.h | 2 + include/linux/ext3_jbd.h | 2 + include/linux/rbtree.h | 2 + lib/rbtree.c | 42 + + 11 files changed, 1886 insertions(+), 160 deletions(-) + +--- linux-2.4.20/fs/ext3/Makefile~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/Makefile Sat Apr 5 03:57:05 2003 +@@ -12,7 +12,7 @@ O_TARGET := ext3.o + export-objs := super.o inode.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o ++ ioctl.o namei.o super.o symlink.o hash.o + obj-m := $(O_TARGET) + + include $(TOPDIR)/Rules.make +--- linux-2.4.20/fs/ext3/dir.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/dir.c Sat Apr 5 03:56:31 2003 +@@ -21,12 +21,16 @@ + #include + #include + #include ++#include ++#include + + static unsigned char ext3_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK + }; + + static int ext3_readdir(struct file *, void *, filldir_t); ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir); + + struct file_operations ext3_dir_operations = { + read: generic_read_dir, +@@ -35,6 +39,17 @@ struct file_operations ext3_dir_operatio + fsync: ext3_sync_file, /* BKL held */ + }; + ++ ++static unsigned char get_dtype(struct super_block *sb, int filetype) ++{ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) || ++ (filetype >= EXT3_FT_MAX)) ++ return DT_UNKNOWN; ++ ++ return (ext3_filetype_table[filetype]); ++} ++ ++ + int ext3_check_dir_entry (const char * function, struct inode * dir, + struct ext3_dir_entry_2 * de, + struct buffer_head * bh, +@@ -79,6 +94,16 @@ static int ext3_readdir(struct file * fi + + sb = inode->i_sb; + ++ if (is_dx(inode)) { ++ err = ext3_dx_readdir(filp, dirent, filldir); ++ if (err != ERR_BAD_DX_DIR) ++ return err; ++ /* ++ * We don't set the inode dirty flag since it's not ++ * critical that it get flushed back to the disk. ++ */ ++ EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; ++ } + stored = 0; + bh = NULL; + offset = filp->f_pos & (sb->s_blocksize - 1); +@@ -162,18 +187,12 @@ revalidate: + * during the copy operation. + */ + unsigned long version = filp->f_version; +- unsigned char d_type = DT_UNKNOWN; + +- if (EXT3_HAS_INCOMPAT_FEATURE(sb, +- EXT3_FEATURE_INCOMPAT_FILETYPE) +- && de->file_type < EXT3_FT_MAX) +- d_type = +- ext3_filetype_table[de->file_type]; + error = filldir(dirent, de->name, + de->name_len, + filp->f_pos, + le32_to_cpu(de->inode), +- d_type); ++ get_dtype(sb, de->file_type)); + if (error) + break; + if (version != filp->f_version) +@@ -188,3 +207,269 @@ revalidate: + UPDATE_ATIME(inode); + return 0; + } ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * These functions convert from the major/minor hash to an f_pos ++ * value. ++ * ++ * Currently we only use major hash numer. This is unfortunate, but ++ * on 32-bit machines, the same VFS interface is used for lseek and ++ * llseek, so if we use the 64 bit offset, then the 32-bit versions of ++ * lseek/telldir/seekdir will blow out spectacularly, and from within ++ * the ext2 low-level routine, we don't know if we're being called by ++ * a 64-bit version of the system call or the 32-bit version of the ++ * system call. Worse yet, NFSv2 only allows for a 32-bit readdir ++ * cookie. Sigh. ++ */ ++#define hash2pos(major, minor) (major >> 1) ++#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff) ++#define pos2min_hash(pos) (0) ++ ++/* ++ * This structure holds the nodes of the red-black tree used to store ++ * the directory entry in hash order. ++ */ ++struct fname { ++ __u32 hash; ++ __u32 minor_hash; ++ rb_node_t rb_hash; ++ struct fname *next; ++ __u32 inode; ++ __u8 name_len; ++ __u8 file_type; ++ char name[0]; ++}; ++ ++/* ++ * This functoin implements a non-recursive way of freeing all of the ++ * nodes in the red-black tree. ++ */ ++static void free_rb_tree_fname(rb_root_t *root) ++{ ++ rb_node_t *n = root->rb_node; ++ rb_node_t *parent; ++ struct fname *fname; ++ ++ while (n) { ++ /* Do the node's children first */ ++ if ((n)->rb_left) { ++ n = n->rb_left; ++ continue; ++ } ++ if (n->rb_right) { ++ n = n->rb_right; ++ continue; ++ } ++ /* ++ * The node has no children; free it, and then zero ++ * out parent's link to it. Finally go to the ++ * beginning of the loop and try to free the parent ++ * node. ++ */ ++ parent = n->rb_parent; ++ fname = rb_entry(n, struct fname, rb_hash); ++ kfree(fname); ++ if (!parent) ++ root->rb_node = 0; ++ else if (parent->rb_left == n) ++ parent->rb_left = 0; ++ else if (parent->rb_right == n) ++ parent->rb_right = 0; ++ n = parent; ++ } ++ root->rb_node = 0; ++} ++ ++ ++struct dir_private_info *create_dir_info(loff_t pos) ++{ ++ struct dir_private_info *p; ++ ++ p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL); ++ if (!p) ++ return NULL; ++ p->root.rb_node = 0; ++ p->curr_node = 0; ++ p->extra_fname = 0; ++ p->last_pos = 0; ++ p->curr_hash = pos2maj_hash(pos); ++ p->curr_minor_hash = pos2min_hash(pos); ++ p->next_hash = 0; ++ return p; ++} ++ ++void ext3_htree_free_dir_info(struct dir_private_info *p) ++{ ++ free_rb_tree_fname(&p->root); ++ kfree(p); ++} ++ ++/* ++ * Given a directory entry, enter it into the fname rb tree. ++ */ ++void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent) ++{ ++ rb_node_t **p, *parent = NULL; ++ struct fname * fname, *new_fn; ++ struct dir_private_info *info; ++ int len; ++ ++ info = (struct dir_private_info *) dir_file->private_data; ++ p = &info->root.rb_node; ++ ++ /* Create and allocate the fname structure */ ++ len = sizeof(struct fname) + dirent->name_len + 1; ++ new_fn = kmalloc(len, GFP_KERNEL); ++ memset(new_fn, 0, len); ++ new_fn->hash = hash; ++ new_fn->minor_hash = minor_hash; ++ new_fn->inode = le32_to_cpu(dirent->inode); ++ new_fn->name_len = dirent->name_len; ++ new_fn->file_type = dirent->file_type; ++ memcpy(new_fn->name, dirent->name, dirent->name_len); ++ new_fn->name[dirent->name_len] = 0; ++ ++ while (*p) { ++ parent = *p; ++ fname = rb_entry(parent, struct fname, rb_hash); ++ ++ /* ++ * If the hash and minor hash match up, then we put ++ * them on a linked list. This rarely happens... ++ */ ++ if ((new_fn->hash == fname->hash) && ++ (new_fn->minor_hash == fname->minor_hash)) { ++ new_fn->next = fname->next; ++ fname->next = new_fn; ++ return; ++ } ++ ++ if (new_fn->hash < fname->hash) ++ p = &(*p)->rb_left; ++ else if (new_fn->hash > fname->hash) ++ p = &(*p)->rb_right; ++ else if (new_fn->minor_hash < fname->minor_hash) ++ p = &(*p)->rb_left; ++ else /* if (new_fn->minor_hash > fname->minor_hash) */ ++ p = &(*p)->rb_right; ++ } ++ ++ rb_link_node(&new_fn->rb_hash, parent, p); ++ rb_insert_color(&new_fn->rb_hash, &info->root); ++} ++ ++ ++ ++/* ++ * This is a helper function for ext3_dx_readdir. It calls filldir ++ * for all entres on the fname linked list. (Normally there is only ++ * one entry on the linked list, unless there are 62 bit hash collisions.) ++ */ ++static int call_filldir(struct file * filp, void * dirent, ++ filldir_t filldir, struct fname *fname) ++{ ++ struct dir_private_info *info = filp->private_data; ++ loff_t curr_pos; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct super_block * sb; ++ int error; ++ ++ sb = inode->i_sb; ++ ++ if (!fname) { ++ printk("call_filldir: called with null fname?!?\n"); ++ return 0; ++ } ++ curr_pos = hash2pos(fname->hash, fname->minor_hash); ++ while (fname) { ++ error = filldir(dirent, fname->name, ++ fname->name_len, curr_pos, ++ fname->inode, ++ get_dtype(sb, fname->file_type)); ++ if (error) { ++ filp->f_pos = curr_pos; ++ info->extra_fname = fname->next; ++ return error; ++ } ++ fname = fname->next; ++ } ++ return 0; ++} ++ ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir) ++{ ++ struct dir_private_info *info = filp->private_data; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct fname *fname; ++ int ret; ++ ++ if (!info) { ++ info = create_dir_info(filp->f_pos); ++ if (!info) ++ return -ENOMEM; ++ filp->private_data = info; ++ } ++ ++ /* Some one has messed with f_pos; reset the world */ ++ if (info->last_pos != filp->f_pos) { ++ free_rb_tree_fname(&info->root); ++ info->curr_node = 0; ++ info->extra_fname = 0; ++ info->curr_hash = pos2maj_hash(filp->f_pos); ++ info->curr_minor_hash = pos2min_hash(filp->f_pos); ++ } ++ ++ /* ++ * If there are any leftover names on the hash collision ++ * chain, return them first. ++ */ ++ if (info->extra_fname && ++ call_filldir(filp, dirent, filldir, info->extra_fname)) ++ goto finished; ++ ++ if (!info->curr_node) ++ info->curr_node = rb_get_first(&info->root); ++ ++ while (1) { ++ /* ++ * Fill the rbtree if we have no more entries, ++ * or the inode has changed since we last read in the ++ * cached entries. ++ */ ++ if ((!info->curr_node) || ++ (filp->f_version != inode->i_version)) { ++ info->curr_node = 0; ++ free_rb_tree_fname(&info->root); ++ filp->f_version = inode->i_version; ++ ret = ext3_htree_fill_tree(filp, info->curr_hash, ++ info->curr_minor_hash, ++ &info->next_hash); ++ if (ret < 0) ++ return ret; ++ if (ret == 0) ++ break; ++ info->curr_node = rb_get_first(&info->root); ++ } ++ ++ fname = rb_entry(info->curr_node, struct fname, rb_hash); ++ info->curr_hash = fname->hash; ++ info->curr_minor_hash = fname->minor_hash; ++ if (call_filldir(filp, dirent, filldir, fname)) ++ break; ++ ++ info->curr_node = rb_get_next(info->curr_node); ++ if (!info->curr_node) { ++ info->curr_hash = info->next_hash; ++ info->curr_minor_hash = 0; ++ } ++ } ++finished: ++ info->last_pos = filp->f_pos; ++ UPDATE_ATIME(inode); ++ return 0; ++} ++#endif +--- linux-2.4.20/fs/ext3/file.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/file.c Sat Apr 5 03:56:31 2003 +@@ -35,6 +35,9 @@ static int ext3_release_file (struct ino + { + if (filp->f_mode & FMODE_WRITE) + ext3_discard_prealloc (inode); ++ if (is_dx(inode) && filp->private_data) ++ ext3_htree_free_dir_info(filp->private_data); ++ + return 0; + } + +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext3/hash.c Sat Apr 5 03:56:31 2003 +@@ -0,0 +1,215 @@ ++/* ++ * linux/fs/ext3/hash.c ++ * ++ * Copyright (C) 2002 by Theodore Ts'o ++ * ++ * This file is released under the GPL v2. ++ * ++ * This file may be redistributed under the terms of the GNU Public ++ * License. ++ */ ++ ++#include ++#include ++#include ++#include ++ ++#define DELTA 0x9E3779B9 ++ ++static void TEA_transform(__u32 buf[4], __u32 const in[]) ++{ ++ __u32 sum = 0; ++ __u32 b0 = buf[0], b1 = buf[1]; ++ __u32 a = in[0], b = in[1], c = in[2], d = in[3]; ++ int n = 16; ++ ++ do { ++ sum += DELTA; ++ b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b); ++ b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d); ++ } while(--n); ++ ++ buf[0] += b0; ++ buf[1] += b1; ++} ++ ++/* F, G and H are basic MD4 functions: selection, majority, parity */ ++#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z)))) ++#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z))) ++#define H(x, y, z) ((x) ^ (y) ^ (z)) ++ ++/* ++ * The generic round function. The application is so specific that ++ * we don't bother protecting all the arguments with parens, as is generally ++ * good macro practice, in favor of extra legibility. ++ * Rotation is separate from addition to prevent recomputation ++ */ ++#define ROUND(f, a, b, c, d, x, s) \ ++ (a += f(b, c, d) + x, a = (a << s) | (a >> (32-s))) ++#define K1 0 ++#define K2 013240474631UL ++#define K3 015666365641UL ++ ++/* ++ * Basic cut-down MD4 transform. Returns only 32 bits of result. ++ */ ++static void halfMD4Transform (__u32 buf[4], __u32 const in[]) ++{ ++ __u32 a = buf[0], b = buf[1], c = buf[2], d = buf[3]; ++ ++ /* Round 1 */ ++ ROUND(F, a, b, c, d, in[0] + K1, 3); ++ ROUND(F, d, a, b, c, in[1] + K1, 7); ++ ROUND(F, c, d, a, b, in[2] + K1, 11); ++ ROUND(F, b, c, d, a, in[3] + K1, 19); ++ ROUND(F, a, b, c, d, in[4] + K1, 3); ++ ROUND(F, d, a, b, c, in[5] + K1, 7); ++ ROUND(F, c, d, a, b, in[6] + K1, 11); ++ ROUND(F, b, c, d, a, in[7] + K1, 19); ++ ++ /* Round 2 */ ++ ROUND(G, a, b, c, d, in[1] + K2, 3); ++ ROUND(G, d, a, b, c, in[3] + K2, 5); ++ ROUND(G, c, d, a, b, in[5] + K2, 9); ++ ROUND(G, b, c, d, a, in[7] + K2, 13); ++ ROUND(G, a, b, c, d, in[0] + K2, 3); ++ ROUND(G, d, a, b, c, in[2] + K2, 5); ++ ROUND(G, c, d, a, b, in[4] + K2, 9); ++ ROUND(G, b, c, d, a, in[6] + K2, 13); ++ ++ /* Round 3 */ ++ ROUND(H, a, b, c, d, in[3] + K3, 3); ++ ROUND(H, d, a, b, c, in[7] + K3, 9); ++ ROUND(H, c, d, a, b, in[2] + K3, 11); ++ ROUND(H, b, c, d, a, in[6] + K3, 15); ++ ROUND(H, a, b, c, d, in[1] + K3, 3); ++ ROUND(H, d, a, b, c, in[5] + K3, 9); ++ ROUND(H, c, d, a, b, in[0] + K3, 11); ++ ROUND(H, b, c, d, a, in[4] + K3, 15); ++ ++ buf[0] += a; ++ buf[1] += b; ++ buf[2] += c; ++ buf[3] += d; ++} ++ ++#undef ROUND ++#undef F ++#undef G ++#undef H ++#undef K1 ++#undef K2 ++#undef K3 ++ ++/* The old legacy hash */ ++static __u32 dx_hack_hash (const char *name, int len) ++{ ++ __u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; ++ while (len--) { ++ __u32 hash = hash1 + (hash0 ^ (*name++ * 7152373)); ++ ++ if (hash & 0x80000000) hash -= 0x7fffffff; ++ hash1 = hash0; ++ hash0 = hash; ++ } ++ return (hash0 << 1); ++} ++ ++static void str2hashbuf(const char *msg, int len, __u32 *buf, int num) ++{ ++ __u32 pad, val; ++ int i; ++ ++ pad = (__u32)len | ((__u32)len << 8); ++ pad |= pad << 16; ++ ++ val = pad; ++ if (len > num*4) ++ len = num * 4; ++ for (i=0; i < len; i++) { ++ if ((i % 4) == 0) ++ val = pad; ++ val = msg[i] + (val << 8); ++ if ((i % 4) == 3) { ++ *buf++ = val; ++ val = pad; ++ num--; ++ } ++ } ++ if (--num >= 0) ++ *buf++ = val; ++ while (--num >= 0) ++ *buf++ = pad; ++} ++ ++/* ++ * Returns the hash of a filename. If len is 0 and name is NULL, then ++ * this function can be used to test whether or not a hash version is ++ * supported. ++ * ++ * The seed is an 4 longword (32 bits) "secret" which can be used to ++ * uniquify a hash. If the seed is all zero's, then some default seed ++ * may be used. ++ * ++ * A particular hash version specifies whether or not the seed is ++ * represented, and whether or not the returned hash is 32 bits or 64 ++ * bits. 32 bit hashes will return 0 for the minor hash. ++ */ ++int ext3fs_dirhash(const char *name, int len, struct dx_hash_info *hinfo) ++{ ++ __u32 hash; ++ __u32 minor_hash = 0; ++ const char *p; ++ int i; ++ __u32 in[8], buf[4]; ++ ++ /* Initialize the default seed for the hash checksum functions */ ++ buf[0] = 0x67452301; ++ buf[1] = 0xefcdab89; ++ buf[2] = 0x98badcfe; ++ buf[3] = 0x10325476; ++ ++ /* Check to see if the seed is all zero's */ ++ if (hinfo->seed) { ++ for (i=0; i < 4; i++) { ++ if (hinfo->seed[i]) ++ break; ++ } ++ if (i < 4) ++ memcpy(buf, hinfo->seed, sizeof(buf)); ++ } ++ ++ switch (hinfo->hash_version) { ++ case DX_HASH_LEGACY: ++ hash = dx_hack_hash(name, len); ++ break; ++ case DX_HASH_HALF_MD4: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 8); ++ halfMD4Transform(buf, in); ++ len -= 32; ++ p += 32; ++ } ++ minor_hash = buf[2]; ++ hash = buf[1]; ++ break; ++ case DX_HASH_TEA: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 4); ++ TEA_transform(buf, in); ++ len -= 16; ++ p += 16; ++ } ++ hash = buf[0]; ++ minor_hash = buf[1]; ++ break; ++ default: ++ hinfo->hash = 0; ++ return -1; ++ } ++ hinfo->hash = hash & ~1; ++ hinfo->minor_hash = minor_hash; ++ return 0; ++} +--- linux-2.4.20/fs/ext3/namei.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/namei.c Sat Apr 5 03:56:31 2003 +@@ -16,6 +16,12 @@ + * David S. Miller (davem@caip.rutgers.edu), 1995 + * Directory entry file type support and forward compatibility hooks + * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 ++ * Hash Tree Directory indexing (c) ++ * Daniel Phillips, 2001 ++ * Hash Tree Directory indexing porting ++ * Christopher Li, 2002 ++ * Hash Tree Directory indexing cleanup ++ * Theodore Ts'o, 2002 + */ + + #include +@@ -38,6 +44,630 @@ + #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) + #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + ++static struct buffer_head *ext3_append(handle_t *handle, ++ struct inode *inode, ++ u32 *block, int *err) ++{ ++ struct buffer_head *bh; ++ ++ *block = inode->i_size >> inode->i_sb->s_blocksize_bits; ++ ++ if ((bh = ext3_bread(handle, inode, *block, 1, err))) { ++ inode->i_size += inode->i_sb->s_blocksize; ++ EXT3_I(inode)->i_disksize = inode->i_size; ++ ext3_journal_get_write_access(handle,bh); ++ } ++ return bh; ++} ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#ifndef swap ++#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0) ++#endif ++ ++typedef struct { u32 v; } le_u32; ++typedef struct { u16 v; } le_u16; ++ ++#ifdef DX_DEBUG ++#define dxtrace(command) command ++#else ++#define dxtrace(command) ++#endif ++ ++struct fake_dirent ++{ ++ /*le*/u32 inode; ++ /*le*/u16 rec_len; ++ u8 name_len; ++ u8 file_type; ++}; ++ ++struct dx_countlimit ++{ ++ le_u16 limit; ++ le_u16 count; ++}; ++ ++struct dx_entry ++{ ++ le_u32 hash; ++ le_u32 block; ++}; ++ ++/* ++ * dx_root_info is laid out so that if it should somehow get overlaid by a ++ * dirent the two low bits of the hash version will be zero. Therefore, the ++ * hash version mod 4 should never be 0. Sincerely, the paranoia department. ++ */ ++ ++struct dx_root ++{ ++ struct fake_dirent dot; ++ char dot_name[4]; ++ struct fake_dirent dotdot; ++ char dotdot_name[4]; ++ struct dx_root_info ++ { ++ le_u32 reserved_zero; ++ u8 hash_version; ++ u8 info_length; /* 8 */ ++ u8 indirect_levels; ++ u8 unused_flags; ++ } ++ info; ++ struct dx_entry entries[0]; ++}; ++ ++struct dx_node ++{ ++ struct fake_dirent fake; ++ struct dx_entry entries[0]; ++}; ++ ++ ++struct dx_frame ++{ ++ struct buffer_head *bh; ++ struct dx_entry *entries; ++ struct dx_entry *at; ++}; ++ ++struct dx_map_entry ++{ ++ u32 hash; ++ u32 offs; ++}; ++ ++#ifdef CONFIG_EXT3_INDEX ++static inline unsigned dx_get_block (struct dx_entry *entry); ++static void dx_set_block (struct dx_entry *entry, unsigned value); ++static inline unsigned dx_get_hash (struct dx_entry *entry); ++static void dx_set_hash (struct dx_entry *entry, unsigned value); ++static unsigned dx_get_count (struct dx_entry *entries); ++static unsigned dx_get_limit (struct dx_entry *entries); ++static void dx_set_count (struct dx_entry *entries, unsigned value); ++static void dx_set_limit (struct dx_entry *entries, unsigned value); ++static unsigned dx_root_limit (struct inode *dir, unsigned infosize); ++static unsigned dx_node_limit (struct inode *dir); ++static struct dx_frame *dx_probe(struct dentry *dentry, ++ struct inode *dir, ++ struct dx_hash_info *hinfo, ++ struct dx_frame *frame, ++ int *err); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static struct ext3_dir_entry_2 *dx_move_dirents (char *from, char *to, ++ struct dx_map_entry *offsets, int count); ++static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash); ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err); ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode); ++ ++/* ++ * Future: use high four bits of block for coalesce-on-delete flags ++ * Mask them off for now. ++ */ ++ ++static inline unsigned dx_get_block (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->block.v) & 0x00ffffff; ++} ++ ++static inline void dx_set_block (struct dx_entry *entry, unsigned value) ++{ ++ entry->block.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_hash (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->hash.v); ++} ++ ++static inline void dx_set_hash (struct dx_entry *entry, unsigned value) ++{ ++ entry->hash.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_count (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->count.v); ++} ++ ++static inline unsigned dx_get_limit (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->limit.v); ++} ++ ++static inline void dx_set_count (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->count.v = cpu_to_le16(value); ++} ++ ++static inline void dx_set_limit (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->limit.v = cpu_to_le16(value); ++} ++ ++static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(1) - ++ EXT3_DIR_REC_LEN(2) - infosize; ++ return 0? 20: entry_space / sizeof(struct dx_entry); ++} ++ ++static inline unsigned dx_node_limit (struct inode *dir) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(0); ++ return 0? 22: entry_space / sizeof(struct dx_entry); ++} ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext3_dir_entry_2 *de, ++ int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ printk("names: "); ++ while ((char *) de < base + size) ++ { ++ if (de->inode) ++ { ++ if (show_names) ++ { ++ int len = de->name_len; ++ char *name = de->name; ++ while (len--) printk("%c", *name++); ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ printk(":%x.%u ", h.hash, ++ ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, ++ struct dx_entry *entries, int levels) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count = dx_get_count (entries), names = 0, space = 0, i; ++ unsigned bcount = 0; ++ struct buffer_head *bh; ++ int err; ++ printk("%i indexed blocks...\n", count); ++ for (i = 0; i < count; i++, entries++) ++ { ++ u32 block = dx_get_block(entries), hash = i? dx_get_hash(entries): 0; ++ u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; ++ struct stats stats; ++ printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); ++ if (!(bh = ext3_bread (NULL,dir, block, 0,&err))) continue; ++ stats = levels? ++ dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf(hinfo, (struct ext3_dir_entry_2 *) bh->b_data, blocksize, 0); ++ names += stats.names; ++ space += stats.space; ++ bcount += stats.bcount; ++ brelse (bh); ++ } ++ if (bcount) ++ printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", ++ names, space/bcount,(space/bcount)*100/blocksize); ++ return (struct stats) { names, space, bcount}; ++} ++#endif /* DX_DEBUG */ ++ ++/* ++ * Probe for a directory leaf block to search. ++ * ++ * dx_probe can return ERR_BAD_DX_DIR, which means there was a format ++ * error in the directory index, and the caller should fall back to ++ * searching the directory normally. The callers of dx_probe **MUST** ++ * check for this error code, and make sure it never gets reflected ++ * back to userspace. ++ */ ++static struct dx_frame * ++dx_probe(struct dentry *dentry, struct inode *dir, ++ struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) ++{ ++ unsigned count, indirect; ++ struct dx_entry *at, *entries, *p, *q, *m; ++ struct dx_root *root; ++ struct buffer_head *bh; ++ struct dx_frame *frame = frame_in; ++ u32 hash; ++ ++ frame->bh = NULL; ++ if (dentry) ++ dir = dentry->d_parent->d_inode; ++ if (!(bh = ext3_bread (NULL,dir, 0, 0, err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version != DX_HASH_TEA && ++ root->info.hash_version != DX_HASH_HALF_MD4 && ++ root->info.hash_version != DX_HASH_LEGACY) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unrecognised inode hash code %d", ++ root->info.hash_version); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ hinfo->hash_version = root->info.hash_version; ++ hinfo->seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ if (dentry) ++ ext3fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); ++ hash = hinfo->hash; ++ ++ if (root->info.unused_flags & 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash flags: %#06x", ++ root->info.unused_flags); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ if ((indirect = root->info.indirect_levels) > 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash depth: %#06x", ++ root->info.indirect_levels); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ entries = (struct dx_entry *) (((char *)&root->info) + ++ root->info.info_length); ++ assert(dx_get_limit(entries) == dx_root_limit(dir, ++ root->info.info_length)); ++ dxtrace (printk("Look up %x", hash)); ++ while (1) ++ { ++ count = dx_get_count(entries); ++ assert (count && count <= dx_get_limit(entries)); ++ p = entries + 1; ++ q = entries + count - 1; ++ while (p <= q) ++ { ++ m = p + (q - p)/2; ++ dxtrace(printk(".")); ++ if (dx_get_hash(m) > hash) ++ q = m - 1; ++ else ++ p = m + 1; ++ } ++ ++ if (0) // linear search cross check ++ { ++ unsigned n = count - 1; ++ at = entries; ++ while (n--) ++ { ++ dxtrace(printk(",")); ++ if (dx_get_hash(++at) > hash) ++ { ++ at--; ++ break; ++ } ++ } ++ assert (at == p - 1); ++ } ++ ++ at = p - 1; ++ dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); ++ frame->bh = bh; ++ frame->entries = entries; ++ frame->at = at; ++ if (!indirect--) return frame; ++ if (!(bh = ext3_bread (NULL,dir, dx_get_block(at), 0, err))) ++ goto fail2; ++ at = entries = ((struct dx_node *) bh->b_data)->entries; ++ assert (dx_get_limit(entries) == dx_node_limit (dir)); ++ frame++; ++ } ++fail2: ++ while (frame >= frame_in) { ++ brelse(frame->bh); ++ frame--; ++ } ++fail: ++ return NULL; ++} ++ ++static void dx_release (struct dx_frame *frames) ++{ ++ if (frames[0].bh == NULL) ++ return; ++ ++ if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) ++ brelse(frames[1].bh); ++ brelse(frames[0].bh); ++} ++ ++/* ++ * This function increments the frame pointer to search the next leaf ++ * block, and reads in the necessary intervening nodes if the search ++ * should be necessary. Whether or not the search is necessary is ++ * controlled by the hash parameter. If the hash value is even, then ++ * the search is only continued if the next block starts with that ++ * hash value. This is used if we are searching for a specific file. ++ * ++ * If the hash value is HASH_NB_ALWAYS, then always go to the next block. ++ * ++ * This function returns 1 if the caller should continue to search, ++ * or 0 if it should not. If there is an error reading one of the ++ * index blocks, it will return -1. ++ * ++ * If start_hash is non-null, it will be filled in with the starting ++ * hash of the next page. ++ */ ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash) ++{ ++ struct dx_frame *p; ++ struct buffer_head *bh; ++ int num_frames = 0; ++ __u32 bhash; ++ ++ *err = ENOENT; ++ p = frame; ++ /* ++ * Find the next leaf page by incrementing the frame pointer. ++ * If we run out of entries in the interior node, loop around and ++ * increment pointer in the parent node. When we break out of ++ * this loop, num_frames indicates the number of interior ++ * nodes need to be read. ++ */ ++ while (1) { ++ if (++(p->at) < p->entries + dx_get_count(p->entries)) ++ break; ++ if (p == frames) ++ return 0; ++ num_frames++; ++ p--; ++ } ++ ++ /* ++ * If the hash is 1, then continue only if the next page has a ++ * continuation hash of any value. This is used for readdir ++ * handling. Otherwise, check to see if the hash matches the ++ * desired contiuation hash. If it doesn't, return since ++ * there's no point to read in the successive index pages. ++ */ ++ bhash = dx_get_hash(p->at); ++ if (start_hash) ++ *start_hash = bhash; ++ if ((hash & 1) == 0) { ++ if ((bhash & ~1) != hash) ++ return 0; ++ } ++ /* ++ * If the hash is HASH_NB_ALWAYS, we always go to the next ++ * block so no check is necessary ++ */ ++ while (num_frames--) { ++ if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), ++ 0, err))) ++ return -1; /* Failure */ ++ p++; ++ brelse (p->bh); ++ p->bh = bh; ++ p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; ++ } ++ return 1; ++} ++ ++ ++/* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *p) ++{ ++ return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++/* ++ * This function fills a red-black tree with information from a ++ * directory. We start scanning the directory in hash order, starting ++ * at start_hash and start_minor_hash. ++ * ++ * This function returns the number of entries inserted into the tree, ++ * or a negative error code. ++ */ ++int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash) ++{ ++ struct dx_hash_info hinfo; ++ struct buffer_head *bh; ++ struct ext3_dir_entry_2 *de, *top; ++ static struct dx_frame frames[2], *frame; ++ struct inode *dir; ++ int block, err; ++ int count = 0; ++ int ret; ++ __u32 hashval; ++ ++ dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, ++ start_minor_hash)); ++ dir = dir_file->f_dentry->d_inode; ++ hinfo.hash = start_hash; ++ hinfo.minor_hash = 0; ++ frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ ++ while (1) { ++ block = dx_get_block(frame->at); ++ dxtrace(printk("Reading block %d\n", block)); ++ if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) ++ goto errout; ++ ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + dir->i_sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) { ++ ext3fs_dirhash(de->name, de->name_len, &hinfo); ++ if ((hinfo.hash < start_hash) || ++ ((hinfo.hash == start_hash) && ++ (hinfo.minor_hash < start_minor_hash))) ++ continue; ++ ext3_htree_store_dirent(dir_file, hinfo.hash, ++ hinfo.minor_hash, de); ++ count++; ++ } ++ brelse (bh); ++ hashval = ~1; ++ ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ++ frame, frames, &err, &hashval); ++ if (next_hash) ++ *next_hash = hashval; ++ if (ret == -1) ++ goto errout; ++ /* ++ * Stop if: (a) there are no more entries, or ++ * (b) we have inserted at least one entry and the ++ * next hash value is not a continuation ++ */ ++ if ((ret == 0) || ++ (count && ((hashval & 1) == 0))) ++ break; ++ } ++ dx_release(frames); ++ dxtrace(printk("Fill tree: returned %d entries\n", count)); ++ return count; ++errout: ++ dx_release(frames); ++ return (err); ++} ++ ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry *map_tail) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ while ((char *) de < base + size) ++ { ++ if (de->name_len && de->inode) { ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ map_tail--; ++ map_tail->hash = h.hash; ++ map_tail->offs = (u32) ((char *) de - base); ++ count++; ++ } ++ /* XXX: do we need to check rec_len == 0 case? -Chris */ ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ return count; ++} ++ ++static void dx_sort_map (struct dx_map_entry *map, unsigned count) ++{ ++ struct dx_map_entry *p, *q, *top = map + count - 1; ++ int more; ++ /* Combsort until bubble sort doesn't suck */ ++ while (count > 2) ++ { ++ count = count*10/13; ++ if (count - 9 < 2) /* 9, 10 -> 11 */ ++ count = 11; ++ for (p = top, q = p - count; q >= map; p--, q--) ++ if (p->hash < q->hash) ++ swap(*p, *q); ++ } ++ /* Garden variety bubble sort */ ++ do { ++ more = 0; ++ q = top; ++ while (q-- > map) ++ { ++ if (q[1].hash >= q[0].hash) ++ continue; ++ swap(*(q+1), *q); ++ more = 1; ++ } ++ } while(more); ++} ++ ++static void dx_insert_block(struct dx_frame *frame, u32 hash, u32 block) ++{ ++ struct dx_entry *entries = frame->entries; ++ struct dx_entry *old = frame->at, *new = old + 1; ++ int count = dx_get_count(entries); ++ ++ assert(count < dx_get_limit(entries)); ++ assert(old < entries + count); ++ memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); ++ dx_set_hash(new, hash); ++ dx_set_block(new, block); ++ dx_set_count(entries, count + 1); ++} ++#endif ++ ++ ++static void ext3_update_dx_flag(struct inode *inode) ++{ ++ if (!EXT3_HAS_COMPAT_FEATURE(inode->i_sb, ++ EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -94,6 +724,7 @@ static int inline search_dirblock(struct + return 0; + } + ++ + /* + * ext3_find_entry() + * +@@ -105,6 +736,8 @@ static int inline search_dirblock(struct + * The returned buffer_head has ->b_count elevated. The caller is expected + * to brelse() it when appropriate. + */ ++ ++ + static struct buffer_head * ext3_find_entry (struct dentry *dentry, + struct ext3_dir_entry_2 ** res_dir) + { +@@ -119,12 +752,32 @@ static struct buffer_head * ext3_find_en + int num = 0; + int nblocks, i, err; + struct inode *dir = dentry->d_parent->d_inode; ++ int namelen; ++ const u8 *name; ++ unsigned blocksize; + + *res_dir = NULL; + sb = dir->i_sb; +- ++ blocksize = sb->s_blocksize; ++ namelen = dentry->d_name.len; ++ name = dentry->d_name.name; ++ if (namelen > EXT3_NAME_LEN) ++ return NULL; ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ bh = ext3_dx_find_entry(dentry, res_dir, &err); ++ /* ++ * On success, or if the error was file not found, ++ * return. Otherwise, fall back to doing a search the ++ * old fashioned way. ++ */ ++ if (bh || (err != ERR_BAD_DX_DIR)) ++ return bh; ++ dxtrace(printk("ext3_find_entry: dx failed, falling back\n")); ++ } ++#endif + nblocks = dir->i_size >> EXT3_BLOCK_SIZE_BITS(sb); +- start = dir->u.ext3_i.i_dir_start_lookup; ++ start = EXT3_I(dir)->i_dir_start_lookup; + if (start >= nblocks) + start = 0; + block = start; +@@ -165,7 +818,7 @@ restart: + i = search_dirblock(bh, dir, dentry, + block << EXT3_BLOCK_SIZE_BITS(sb), res_dir); + if (i == 1) { +- dir->u.ext3_i.i_dir_start_lookup = block; ++ EXT3_I(dir)->i_dir_start_lookup = block; + ret = bh; + goto cleanup_and_exit; + } else { +@@ -196,6 +849,66 @@ cleanup_and_exit: + return ret; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err) ++{ ++ struct super_block * sb; ++ struct dx_hash_info hinfo; ++ u32 hash; ++ struct dx_frame frames[2], *frame; ++ struct ext3_dir_entry_2 *de, *top; ++ struct buffer_head *bh; ++ unsigned long block; ++ int retval; ++ int namelen = dentry->d_name.len; ++ const u8 *name = dentry->d_name.name; ++ struct inode *dir = dentry->d_parent->d_inode; ++ ++ sb = dir->i_sb; ++ if (!(frame = dx_probe (dentry, 0, &hinfo, frames, err))) ++ return NULL; ++ hash = hinfo.hash; ++ do { ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, err))) ++ goto errout; ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match (namelen, name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto errout; ++ } ++ *res_dir = de; ++ dx_release (frames); ++ return bh; ++ } ++ brelse (bh); ++ /* Check to see if we should continue to search */ ++ retval = ext3_htree_next_block(dir, hash, frame, ++ frames, err, 0); ++ if (retval == -1) { ++ ext3_warning(sb, __FUNCTION__, ++ "error reading index page in directory #%lu", ++ dir->i_ino); ++ goto errout; ++ } ++ } while (retval == 1); ++ ++ *err = -ENOENT; ++errout: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++} ++#endif ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -212,8 +925,9 @@ static struct dentry *ext3_lookup(struct + brelse (bh); + inode = iget(dir->i_sb, ino); + +- if (!inode) ++ if (!inode) { + return ERR_PTR(-EACCES); ++ } + } + d_add(dentry, inode); + return NULL; +@@ -237,6 +951,301 @@ static inline void ext3_set_de_type(stru + de->file_type = ext3_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct ext3_dir_entry_2 * ++dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count) ++{ ++ unsigned rec_len = 0; ++ ++ while (count--) { ++ struct ext3_dir_entry_2 *de = (struct ext3_dir_entry_2 *) (from + map->offs); ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ memcpy (to, de, rec_len); ++ ((struct ext3_dir_entry_2 *) to)->rec_len = rec_len; ++ de->inode = 0; ++ map++; ++ to += rec_len; ++ } ++ return (struct ext3_dir_entry_2 *) (to - rec_len); ++} ++ ++static struct ext3_dir_entry_2* dx_pack_dirents(char *base, int size) ++{ ++ struct ext3_dir_entry_2 *next, *to, *prev, *de = (struct ext3_dir_entry_2 *) base; ++ unsigned rec_len = 0; ++ ++ prev = to = de; ++ while ((char*)de < base + size) { ++ next = (struct ext3_dir_entry_2 *) ((char *) de + ++ le16_to_cpu(de->rec_len)); ++ if (de->inode && de->name_len) { ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ if (de > to) ++ memmove(to, de, rec_len); ++ to->rec_len = rec_len; ++ prev = to; ++ to = (struct ext3_dir_entry_2 *) (((char *) to) + rec_len); ++ } ++ de = next; ++ } ++ return prev; ++} ++ ++static struct ext3_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, ++ struct buffer_head **bh,struct dx_frame *frame, ++ struct dx_hash_info *hinfo, int *error) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count, continued; ++ struct buffer_head *bh2; ++ u32 newblock; ++ u32 hash2; ++ struct dx_map_entry *map; ++ char *data1 = (*bh)->b_data, *data2; ++ unsigned split; ++ struct ext3_dir_entry_2 *de = NULL, *de2; ++ int err; ++ ++ bh2 = ext3_append (handle, dir, &newblock, error); ++ if (!(bh2)) { ++ brelse(*bh); ++ *bh = NULL; ++ goto errout; ++ } ++ ++ BUFFER_TRACE(*bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, *bh); ++ if (err) { ++ journal_error: ++ brelse(*bh); ++ brelse(bh2); ++ *bh = NULL; ++ ext3_std_error(dir->i_sb, err); ++ goto errout; ++ } ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ ++ data2 = bh2->b_data; ++ ++ /* create map in the end of data2 block */ ++ map = (struct dx_map_entry *) (data2 + blocksize); ++ count = dx_make_map ((struct ext3_dir_entry_2 *) data1, ++ blocksize, hinfo, map); ++ map -= count; ++ split = count/2; // need to adjust to actual middle ++ dx_sort_map (map, count); ++ hash2 = map[split].hash; ++ continued = hash2 == map[split - 1].hash; ++ dxtrace(printk("Split block %i at %x, %i/%i\n", ++ dx_get_block(frame->at), hash2, split, count-split)); ++ ++ /* Fancy dance to stay within two buffers */ ++ de2 = dx_move_dirents(data1, data2, map + split, count - split); ++ de = dx_pack_dirents(data1,blocksize); ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ de2->rec_len = cpu_to_le16(data2 + blocksize - (char *) de2); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data1, blocksize, 1)); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data2, blocksize, 1)); ++ ++ /* Which block gets the new entry? */ ++ if (hinfo->hash >= hash2) ++ { ++ swap(*bh, bh2); ++ de = de2; ++ } ++ dx_insert_block (frame, hash2 + continued, newblock); ++ err = ext3_journal_dirty_metadata (handle, bh2); ++ if (err) ++ goto journal_error; ++ err = ext3_journal_dirty_metadata (handle, frame->bh); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ dxtrace(dx_show_index ("frame", frame->entries)); ++errout: ++ return de; ++} ++#endif ++ ++ ++/* ++ * Add a new entry into a directory (leaf) block. If de is non-NULL, ++ * it points to a directory entry which is guaranteed to be large ++ * enough for new directory entry. If de is NULL, then ++ * add_dirent_to_buf will attempt search the directory block for ++ * space. It will return -ENOSPC if no space is available, and -EIO ++ * and -EEXIST if directory entry already exists. ++ * ++ * NOTE! bh is NOT released in the case where ENOSPC is returned. In ++ * all other cases bh is released. ++ */ ++static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct ext3_dir_entry_2 *de, ++ struct buffer_head * bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ unsigned long offset = 0; ++ unsigned short reclen; ++ int nlen, rlen, err; ++ char *top; ++ ++ reclen = EXT3_DIR_REC_LEN(namelen); ++ if (!de) { ++ de = (struct ext3_dir_entry_2 *)bh->b_data; ++ top = bh->b_data + dir->i_sb->s_blocksize - reclen; ++ while ((char *) de <= top) { ++ if (!ext3_check_dir_entry("ext3_add_entry", dir, de, ++ bh, offset)) { ++ brelse (bh); ++ return -EIO; ++ } ++ if (ext3_match (namelen, name, de)) { ++ brelse (bh); ++ return -EEXIST; ++ } ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if ((de->inode? rlen - nlen: rlen) >= reclen) ++ break; ++ de = (struct ext3_dir_entry_2 *)((char *)de + rlen); ++ offset += rlen; ++ } ++ if ((char *) de > top) ++ return -ENOSPC; ++ } ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) { ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return err; ++ } ++ ++ /* By now the buffer is marked for journaling */ ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if (de->inode) { ++ struct ext3_dir_entry_2 *de1 = (struct ext3_dir_entry_2 *)((char *)de + nlen); ++ de1->rec_len = cpu_to_le16(rlen - nlen); ++ de->rec_len = cpu_to_le16(nlen); ++ de = de1; ++ } ++ de->file_type = EXT3_FT_UNKNOWN; ++ if (inode) { ++ de->inode = cpu_to_le32(inode->i_ino); ++ ext3_set_de_type(dir->i_sb, de, inode->i_mode); ++ } else ++ de->inode = 0; ++ de->name_len = namelen; ++ memcpy (de->name, name, namelen); ++ /* ++ * XXX shouldn't update any times until successful ++ * completion of syscall, but too many callers depend ++ * on this. ++ * ++ * XXX similarly, too many callers depend on ++ * ext3_new_inode() setting the times, but error ++ * recovery deletes the inode, so the worst that can ++ * happen is that the times are slightly out of date ++ * and/or different from the directory change time. ++ */ ++ dir->i_mtime = dir->i_ctime = CURRENT_TIME; ++ ext3_update_dx_flag(dir); ++ dir->i_version = ++event; ++ ext3_mark_inode_dirty(handle, dir); ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return 0; ++} ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * This converts a one block unindexed directory to a 3 block indexed ++ * directory, and adds the dentry to the indexed directory. ++ */ ++static int make_indexed_dir(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct buffer_head *bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ struct buffer_head *bh2; ++ struct dx_root *root; ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries; ++ struct ext3_dir_entry_2 *de, *de2; ++ char *data1, *top; ++ unsigned len; ++ int retval; ++ unsigned blocksize; ++ struct dx_hash_info hinfo; ++ u32 block; ++ ++ blocksize = dir->i_sb->s_blocksize; ++ dxtrace(printk("Creating index\n")); ++ retval = ext3_journal_get_write_access(handle, bh); ++ if (retval) { ++ ext3_std_error(dir->i_sb, retval); ++ brelse(bh); ++ return retval; ++ } ++ root = (struct dx_root *) bh->b_data; ++ ++ EXT3_I(dir)->i_flags |= EXT3_INDEX_FL; ++ bh2 = ext3_append (handle, dir, &block, &retval); ++ if (!(bh2)) { ++ brelse(bh); ++ return retval; ++ } ++ data1 = bh2->b_data; ++ ++ /* The 0th block becomes the root, move the dirents out */ ++ de = (struct ext3_dir_entry_2 *) &root->dotdot; ++ de = (struct ext3_dir_entry_2 *) ((char *)de + de->rec_len); ++ len = ((char *) root) + blocksize - (char *) de; ++ memcpy (data1, de, len); ++ de = (struct ext3_dir_entry_2 *) data1; ++ top = data1 + len; ++ while (((char *) de2=(char*)de+le16_to_cpu(de->rec_len)) < top) ++ de = de2; ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ /* Initialize the root; the dot dirents already exist */ ++ de = (struct ext3_dir_entry_2 *) (&root->dotdot); ++ de->rec_len = cpu_to_le16(blocksize - EXT3_DIR_REC_LEN(2)); ++ memset (&root->info, 0, sizeof(root->info)); ++ root->info.info_length = sizeof(root->info); ++ root->info.hash_version = dir->i_sb->u.ext3_sb.s_def_hash_version; ++ entries = root->entries; ++ dx_set_block (entries, 1); ++ dx_set_count (entries, 1); ++ dx_set_limit (entries, dx_root_limit(dir, sizeof(root->info))); ++ ++ /* Initialize as for dx_probe */ ++ hinfo.hash_version = root->info.hash_version; ++ hinfo.seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ ext3fs_dirhash(name, namelen, &hinfo); ++ frame = frames; ++ frame->entries = entries; ++ frame->at = entries; ++ frame->bh = bh; ++ bh = bh2; ++ de = do_split(handle,dir, &bh, frame, &hinfo, &retval); ++ dx_release (frames); ++ if (!(de)) ++ return retval; ++ ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} ++#endif ++ + /* + * ext3_add_entry() + * +@@ -247,127 +1255,198 @@ static inline void ext3_set_de_type(stru + * may not sleep between calling this and putting something into + * the entry, as someone else might have used it while you slept. + */ +- +-/* +- * AKPM: the journalling code here looks wrong on the error paths +- */ + static int ext3_add_entry (handle_t *handle, struct dentry *dentry, + struct inode *inode) + { + struct inode *dir = dentry->d_parent->d_inode; +- const char *name = dentry->d_name.name; +- int namelen = dentry->d_name.len; + unsigned long offset; +- unsigned short rec_len; + struct buffer_head * bh; +- struct ext3_dir_entry_2 * de, * de1; ++ struct ext3_dir_entry_2 *de; + struct super_block * sb; + int retval; ++#ifdef CONFIG_EXT3_INDEX ++ int dx_fallback=0; ++#endif ++ unsigned blocksize; ++ unsigned nlen, rlen; ++ u32 block, blocks; + + sb = dir->i_sb; +- +- if (!namelen) ++ blocksize = sb->s_blocksize; ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &retval); ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ retval = ext3_dx_add_entry(handle, dentry, inode); ++ if (!retval || (retval != ERR_BAD_DX_DIR)) ++ return retval; ++ EXT3_I(dir)->i_flags &= ~EXT3_INDEX_FL; ++ dx_fallback++; ++ ext3_mark_inode_dirty(handle, dir); ++ } ++#endif ++ blocks = dir->i_size >> sb->s_blocksize_bits; ++ for (block = 0, offset = 0; block < blocks; block++) { ++ bh = ext3_bread(handle, dir, block, 0, &retval); ++ if(!bh) ++ return retval; ++ retval = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (retval != -ENOSPC) ++ return retval; ++ ++#ifdef CONFIG_EXT3_INDEX ++ if (blocks == 1 && !dx_fallback && ++ EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ return make_indexed_dir(handle, dentry, inode, bh); ++#endif ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); + if (!bh) + return retval; +- rec_len = EXT3_DIR_REC_LEN(namelen); +- offset = 0; + de = (struct ext3_dir_entry_2 *) bh->b_data; +- while (1) { +- if ((char *)de >= sb->s_blocksize + bh->b_data) { +- brelse (bh); +- bh = NULL; +- bh = ext3_bread (handle, dir, +- offset >> EXT3_BLOCK_SIZE_BITS(sb), 1, &retval); +- if (!bh) +- return retval; +- if (dir->i_size <= offset) { +- if (dir->i_size == 0) { +- brelse(bh); +- return -ENOENT; +- } ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = blocksize); ++ nlen = 0; ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} + +- ext3_debug ("creating next block\n"); ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * Returns 0 for success, or a negative error value ++ */ ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode) ++{ ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ struct dx_hash_info hinfo; ++ struct buffer_head * bh; ++ struct inode *dir = dentry->d_parent->d_inode; ++ struct super_block * sb = dir->i_sb; ++ struct ext3_dir_entry_2 *de; ++ int err; + +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- } else { ++ frame = dx_probe(dentry, 0, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ entries = frame->entries; ++ at = frame->at; + +- ext3_debug ("skipping to next block\n"); ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err))) ++ goto cleanup; + +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- } +- } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { +- brelse (bh); +- return -EEXIST; ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) ++ goto journal_error; ++ ++ err = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (err != -ENOSPC) { ++ bh = 0; ++ goto cleanup; ++ } ++ ++ /* Block full, should compress but for now just split */ ++ dxtrace(printk("using %u of %u node entries\n", ++ dx_get_count(entries), dx_get_limit(entries))); ++ /* Need to split index? */ ++ if (dx_get_count(entries) == dx_get_limit(entries)) { ++ u32 newblock; ++ unsigned icount = dx_get_count(entries); ++ int levels = frame - frames; ++ struct dx_entry *entries2; ++ struct dx_node *node2; ++ struct buffer_head *bh2; ++ ++ if (levels && (dx_get_count(frames->entries) == ++ dx_get_limit(frames->entries))) { ++ ext3_warning(sb, __FUNCTION__, ++ "Directory index full!\n"); ++ err = -ENOSPC; ++ goto cleanup; + } +- if ((le32_to_cpu(de->inode) == 0 && +- le16_to_cpu(de->rec_len) >= rec_len) || +- (le16_to_cpu(de->rec_len) >= +- EXT3_DIR_REC_LEN(de->name_len) + rec_len)) { +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- /* By now the buffer is marked for journaling */ +- offset += le16_to_cpu(de->rec_len); +- if (le32_to_cpu(de->inode)) { +- de1 = (struct ext3_dir_entry_2 *) ((char *) de + +- EXT3_DIR_REC_LEN(de->name_len)); +- de1->rec_len = +- cpu_to_le16(le16_to_cpu(de->rec_len) - +- EXT3_DIR_REC_LEN(de->name_len)); +- de->rec_len = cpu_to_le16( +- EXT3_DIR_REC_LEN(de->name_len)); +- de = de1; ++ bh2 = ext3_append (handle, dir, &newblock, &err); ++ if (!(bh2)) ++ goto cleanup; ++ node2 = (struct dx_node *)(bh2->b_data); ++ entries2 = node2->entries; ++ node2->fake.rec_len = cpu_to_le16(sb->s_blocksize); ++ node2->fake.inode = 0; ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ if (levels) { ++ unsigned icount1 = icount/2, icount2 = icount - icount1; ++ unsigned hash2 = dx_get_hash(entries + icount1); ++ dxtrace(printk("Split index %i/%i\n", icount1, icount2)); ++ ++ BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ ++ err = ext3_journal_get_write_access(handle, ++ frames[0].bh); ++ if (err) ++ goto journal_error; ++ ++ memcpy ((char *) entries2, (char *) (entries + icount1), ++ icount2 * sizeof(struct dx_entry)); ++ dx_set_count (entries, icount1); ++ dx_set_count (entries2, icount2); ++ dx_set_limit (entries2, dx_node_limit(dir)); ++ ++ /* Which index block gets the new entry? */ ++ if (at - entries >= icount1) { ++ frame->at = at = at - entries - icount1 + entries2; ++ frame->entries = entries = entries2; ++ swap(frame->bh, bh2); + } +- de->file_type = EXT3_FT_UNKNOWN; +- if (inode) { +- de->inode = cpu_to_le32(inode->i_ino); +- ext3_set_de_type(dir->i_sb, de, inode->i_mode); +- } else +- de->inode = 0; +- de->name_len = namelen; +- memcpy (de->name, name, namelen); +- /* +- * XXX shouldn't update any times until successful +- * completion of syscall, but too many callers depend +- * on this. +- * +- * XXX similarly, too many callers depend on +- * ext3_new_inode() setting the times, but error +- * recovery deletes the inode, so the worst that can +- * happen is that the times are slightly out of date +- * and/or different from the directory change time. +- */ +- dir->i_mtime = dir->i_ctime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- dir->i_version = ++event; +- ext3_mark_inode_dirty(handle, dir); +- BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +- ext3_journal_dirty_metadata(handle, bh); +- brelse(bh); +- return 0; ++ dx_insert_block (frames + 0, hash2, newblock); ++ dxtrace(dx_show_index ("node", frames[1].entries)); ++ dxtrace(dx_show_index ("node", ++ ((struct dx_node *) bh2->b_data)->entries)); ++ err = ext3_journal_dirty_metadata(handle, bh2); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ } else { ++ dxtrace(printk("Creating second level index...\n")); ++ memcpy((char *) entries2, (char *) entries, ++ icount * sizeof(struct dx_entry)); ++ dx_set_limit(entries2, dx_node_limit(dir)); ++ ++ /* Set up root */ ++ dx_set_count(entries, 1); ++ dx_set_block(entries + 0, newblock); ++ ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; ++ ++ /* Add new access path frame */ ++ frame = frames + 1; ++ frame->at = at = at - entries + entries2; ++ frame->entries = entries = entries2; ++ frame->bh = bh2; ++ err = ext3_journal_get_write_access(handle, ++ frame->bh); ++ if (err) ++ goto journal_error; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- brelse (bh); +- return -ENOSPC; ++ de = do_split(handle, dir, &bh, frame, &hinfo, &err); ++ if (!de) ++ goto cleanup; ++ err = add_dirent_to_buf(handle, dentry, inode, de, bh); ++ bh = 0; ++ goto cleanup; ++ ++journal_error: ++ ext3_std_error(dir->i_sb, err); ++cleanup: ++ if (bh) ++ brelse(bh); ++ dx_release(frames); ++ return err; + } ++#endif + + /* + * ext3_delete_entry deletes a directory entry by merging it with the +@@ -451,9 +1530,11 @@ static int ext3_create (struct inode * d + struct inode * inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -478,9 +1559,11 @@ static int ext3_mknod (struct inode * di + struct inode *inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -507,9 +1590,11 @@ static int ext3_mkdir(struct inode * dir + if (dir->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -521,7 +1606,7 @@ static int ext3_mkdir(struct inode * dir + + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; +- inode->i_size = inode->u.ext3_i.i_disksize = inode->i_sb->s_blocksize; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { +@@ -554,21 +1639,19 @@ static int ext3_mkdir(struct inode * dir + inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); +- if (err) +- goto out_no_entry; ++ if (err) { ++ inode->i_nlink = 0; ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + dir->i_nlink++; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + d_instantiate(dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- inode->i_nlink = 0; +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + /* +@@ -655,7 +1738,7 @@ int ext3_orphan_add(handle_t *handle, st + int err = 0, rc; + + lock_super(sb); +- if (!list_empty(&inode->u.ext3_i.i_orphan)) ++ if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + + /* Orphan handling is only valid for files with data blocks +@@ -696,7 +1779,7 @@ int ext3_orphan_add(handle_t *handle, st + * This is safe: on error we're going to ignore the orphan list + * anyway on the next recovery. */ + if (!err) +- list_add(&inode->u.ext3_i.i_orphan, &EXT3_SB(sb)->s_orphan); ++ list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); + + jbd_debug(4, "superblock will point to %ld\n", inode->i_ino); + jbd_debug(4, "orphan inode %ld will point to %d\n", +@@ -714,25 +1797,26 @@ out_unlock: + int ext3_orphan_del(handle_t *handle, struct inode *inode) + { + struct list_head *prev; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct ext3_sb_info *sbi; + ino_t ino_next; + struct ext3_iloc iloc; + int err = 0; + + lock_super(inode->i_sb); +- if (list_empty(&inode->u.ext3_i.i_orphan)) { ++ if (list_empty(&ei->i_orphan)) { + unlock_super(inode->i_sb); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); +- prev = inode->u.ext3_i.i_orphan.prev; ++ prev = ei->i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %ld from orphan list\n", inode->i_ino); + +- list_del(&inode->u.ext3_i.i_orphan); +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ list_del(&ei->i_orphan); ++ INIT_LIST_HEAD(&ei->i_orphan); + + /* If we're on an error path, we may not have a valid + * transaction handle with which to update the orphan list on +@@ -793,8 +1877,9 @@ static int ext3_rmdir (struct inode * di + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + retval = -ENOENT; + bh = ext3_find_entry (dentry, &de); +@@ -832,7 +1917,7 @@ static int ext3_rmdir (struct inode * di + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; + ext3_mark_inode_dirty(handle, inode); +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -850,8 +1935,9 @@ static int ext3_unlink(struct inode * di + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -878,7 +1964,7 @@ static int ext3_unlink(struct inode * di + if (retval) + goto end_unlink; + dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + inode->i_nlink--; + if (!inode->i_nlink) +@@ -904,9 +1990,11 @@ static int ext3_symlink (struct inode * + if (l > dir->i_sb->s_blocksize) + return -ENAMETOOLONG; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 5); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -916,7 +2004,7 @@ static int ext3_symlink (struct inode * + if (IS_ERR(inode)) + goto out_stop; + +- if (l > sizeof (inode->u.ext3_i.i_data)) { ++ if (l > sizeof (EXT3_I(inode)->i_data)) { + inode->i_op = &page_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* +@@ -925,25 +2013,23 @@ static int ext3_symlink (struct inode * + * i_size in generic_commit_write(). + */ + err = block_symlink(inode, symname, l); +- if (err) +- goto out_no_entry; ++ if (err) { ++ ext3_dec_count(handle, inode); ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + } else { + inode->i_op = &ext3_fast_symlink_inode_operations; +- memcpy((char*)&inode->u.ext3_i.i_data,symname,l); ++ memcpy((char*)&EXT3_I(inode)->i_data,symname,l); + inode->i_size = l-1; + } +- inode->u.ext3_i.i_disksize = inode->i_size; ++ EXT3_I(inode)->i_disksize = inode->i_size; + err = ext3_add_nondir(handle, dentry, inode); + ext3_mark_inode_dirty(handle, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- ext3_dec_count(handle, inode); +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + static int ext3_link (struct dentry * old_dentry, +@@ -956,12 +2042,15 @@ static int ext3_link (struct dentry * ol + if (S_ISDIR(inode->i_mode)) + return -EPERM; + +- if (inode->i_nlink >= EXT3_LINK_MAX) ++ if (inode->i_nlink >= EXT3_LINK_MAX) { + return -EMLINK; ++ } + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -995,9 +2084,11 @@ static int ext3_rename (struct inode * o + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 2); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(old_dir) || IS_SYNC(new_dir)) + handle->h_sync = 1; +@@ -1077,7 +2168,7 @@ static int ext3_rename (struct inode * o + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +2180,7 @@ static int ext3_rename (struct inode * o + new_inode->i_nlink--; + } else { + new_dir->i_nlink++; +- new_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(new_dir); + ext3_mark_inode_dirty(handle, new_dir); + } + } +--- linux-2.4.20/fs/ext3/super.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/super.c Sat Apr 5 03:56:31 2003 +@@ -707,6 +707,7 @@ static int ext3_setup_super(struct super + es->s_mtime = cpu_to_le32(CURRENT_TIME); + ext3_update_dynamic_rev(sb); + EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); ++ + ext3_commit_super (sb, es, 1); + if (test_opt (sb, DEBUG)) + printk (KERN_INFO +@@ -717,6 +718,7 @@ static int ext3_setup_super(struct super + EXT3_BLOCKS_PER_GROUP(sb), + EXT3_INODES_PER_GROUP(sb), + sbi->s_mount_opt); ++ + printk(KERN_INFO "EXT3 FS " EXT3FS_VERSION ", " EXT3FS_DATE " on %s, ", + bdevname(sb->s_dev)); + if (EXT3_SB(sb)->s_journal->j_inode == NULL) { +@@ -890,6 +892,7 @@ static loff_t ext3_max_size(int bits) + return res; + } + ++ + struct super_block * ext3_read_super (struct super_block * sb, void * data, + int silent) + { +@@ -1066,6 +1069,9 @@ struct super_block * ext3_read_super (st + sbi->s_mount_state = le16_to_cpu(es->s_state); + sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb)); + sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb)); ++ for (i=0; i < 4; i++) ++ sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); ++ sbi->s_def_hash_version = es->s_def_hash_version; + + if (sbi->s_blocks_per_group > blocksize * 8) { + printk (KERN_ERR +@@ -1769,6 +1775,7 @@ static void __exit exit_ext3_fs(void) + unregister_filesystem(&ext3_fs_type); + } + ++EXPORT_SYMBOL(ext3_force_commit); + EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); +--- linux-2.4.20/include/linux/ext3_fs.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_fs.h Sat Apr 5 03:56:31 2003 +@@ -40,6 +40,11 @@ + #define EXT3FS_VERSION "2.4-0.9.19" + + /* ++ * Always enable hashed directories ++ */ ++#define CONFIG_EXT3_INDEX ++ ++/* + * Debug code + */ + #ifdef EXT3FS_DEBUG +@@ -437,8 +442,11 @@ struct ext3_super_block { + /*E0*/ __u32 s_journal_inum; /* inode number of journal file */ + __u32 s_journal_dev; /* device number of journal file */ + __u32 s_last_orphan; /* start of list of inodes to delete */ +- +-/*EC*/ __u32 s_reserved[197]; /* Padding to the end of the block */ ++ __u32 s_hash_seed[4]; /* HTREE hash seed */ ++ __u8 s_def_hash_version; /* Default hash version to use */ ++ __u8 s_reserved_char_pad; ++ __u16 s_reserved_word_pad; ++ __u32 s_reserved[192]; /* Padding to the end of the block */ + }; + + #ifdef __KERNEL__ +@@ -575,9 +583,46 @@ struct ext3_dir_entry_2 { + #define EXT3_DIR_ROUND (EXT3_DIR_PAD - 1) + #define EXT3_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT3_DIR_ROUND) & \ + ~EXT3_DIR_ROUND) ++/* ++ * Hash Tree Directory indexing ++ * (c) Daniel Phillips, 2001 ++ */ ++ ++#ifdef CONFIG_EXT3_INDEX ++ #define is_dx(dir) (EXT3_HAS_COMPAT_FEATURE(dir->i_sb, \ ++ EXT3_FEATURE_COMPAT_DIR_INDEX) && \ ++ (EXT3_I(dir)->i_flags & EXT3_INDEX_FL)) ++#define EXT3_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) ++#else ++ #define is_dx(dir) 0 ++#define EXT3_DIR_LINK_MAX(dir) ((dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2) ++#endif ++ ++/* Legal values for the dx_root hash_version field: */ ++ ++#define DX_HASH_LEGACY 0 ++#define DX_HASH_HALF_MD4 1 ++#define DX_HASH_TEA 2 ++ ++/* hash info structure used by the directory hash */ ++struct dx_hash_info ++{ ++ u32 hash; ++ u32 minor_hash; ++ int hash_version; ++ u32 *seed; ++}; + + #ifdef __KERNEL__ + /* ++ * Control parameters used by ext3_htree_next_block ++ */ ++#define HASH_NB_ALWAYS 1 ++ ++ ++/* + * Describe an inode's exact location on disk and in memory + */ + struct ext3_iloc +@@ -587,6 +632,27 @@ struct ext3_iloc + unsigned long block_group; + }; + ++ ++/* ++ * This structure is stuffed into the struct file's private_data field ++ * for directories. It is where we put information so that we can do ++ * readdir operations in hash tree order. ++ */ ++struct dir_private_info { ++ rb_root_t root; ++ rb_node_t *curr_node; ++ struct fname *extra_fname; ++ loff_t last_pos; ++ __u32 curr_hash; ++ __u32 curr_minor_hash; ++ __u32 next_hash; ++}; ++ ++/* ++ * Special error return code only used by dx_probe() and its callers. ++ */ ++#define ERR_BAD_DX_DIR -75000 ++ + /* + * Function prototypes + */ +@@ -614,11 +680,20 @@ extern struct ext3_group_desc * ext3_get + + /* dir.c */ + extern int ext3_check_dir_entry(const char *, struct inode *, +- struct ext3_dir_entry_2 *, struct buffer_head *, +- unsigned long); ++ struct ext3_dir_entry_2 *, ++ struct buffer_head *, unsigned long); ++extern void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent); ++extern void ext3_htree_free_dir_info(struct dir_private_info *p); ++ + /* fsync.c */ + extern int ext3_sync_file (struct file *, struct dentry *, int); + ++/* hash.c */ ++extern int ext3fs_dirhash(const char *name, int len, struct ++ dx_hash_info *hinfo); ++ + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); +@@ -650,6 +725,8 @@ extern int ext3_ioctl (struct inode *, s + /* namei.c */ + extern int ext3_orphan_add(handle_t *, struct inode *); + extern int ext3_orphan_del(handle_t *, struct inode *); ++extern int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash); + + /* super.c */ + extern void ext3_error (struct super_block *, const char *, const char *, ...) +--- linux-2.4.20/include/linux/ext3_fs_sb.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_fs_sb.h Sat Apr 5 03:56:31 2003 +@@ -62,6 +62,8 @@ struct ext3_sb_info { + int s_inode_size; + int s_first_ino; + u32 s_next_generation; ++ u32 s_hash_seed[4]; ++ int s_def_hash_version; + + /* Journaling */ + struct inode * s_journal_inode; +--- linux-2.4.20/include/linux/ext3_jbd.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_jbd.h Sat Apr 5 03:56:31 2003 +@@ -63,6 +63,8 @@ extern int ext3_writepage_trans_blocks(s + + #define EXT3_RESERVE_TRANS_BLOCKS 12 + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, +--- linux-2.4.20/include/linux/rbtree.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/rbtree.h Sat Apr 5 03:56:31 2003 +@@ -120,6 +120,8 @@ rb_root_t; + + extern void rb_insert_color(rb_node_t *, rb_root_t *); + extern void rb_erase(rb_node_t *, rb_root_t *); ++extern rb_node_t *rb_get_first(rb_root_t *root); ++extern rb_node_t *rb_get_next(rb_node_t *n); + + static inline void rb_link_node(rb_node_t * node, rb_node_t * parent, rb_node_t ** rb_link) + { +--- linux-2.4.20/lib/rbtree.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/lib/rbtree.c Sat Apr 5 03:56:31 2003 +@@ -17,6 +17,8 @@ + Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + + linux/lib/rbtree.c ++ ++ rb_get_first and rb_get_next written by Theodore Ts'o, 9/8/2002 + */ + + #include +@@ -294,3 +296,43 @@ void rb_erase(rb_node_t * node, rb_root_ + __rb_erase_color(child, parent, root); + } + EXPORT_SYMBOL(rb_erase); ++ ++/* ++ * This function returns the first node (in sort order) of the tree. ++ */ ++rb_node_t *rb_get_first(rb_root_t *root) ++{ ++ rb_node_t *n; ++ ++ n = root->rb_node; ++ if (!n) ++ return 0; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++} ++EXPORT_SYMBOL(rb_get_first); ++ ++/* ++ * Given a node, this function will return the next node in the tree. ++ */ ++rb_node_t *rb_get_next(rb_node_t *n) ++{ ++ rb_node_t *parent; ++ ++ if (n->rb_right) { ++ n = n->rb_right; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++ } else { ++ while ((parent = n->rb_parent)) { ++ if (n == parent->rb_left) ++ return parent; ++ n = parent; ++ } ++ return 0; ++ } ++} ++EXPORT_SYMBOL(rb_get_next); ++ + +_ diff --git a/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch b/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch new file mode 100644 index 0000000..4bcefce --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch @@ -0,0 +1,481 @@ + fs/ext3/file.c | 4 + fs/ext3/inode.c | 116 ++++++++++++++++++++++ + fs/ext3/super.c | 230 +++++++++++++++++++++++++++++++++++++++++++++ + include/linux/ext3_fs.h | 5 + include/linux/ext3_fs_sb.h | 10 + + 5 files changed, 365 insertions(+) + +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 01:18:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:19:22 2003 +@@ -401,6 +401,220 @@ + } + } + ++#ifdef EXT3_DELETE_THREAD ++/* ++ * Delete inodes in a loop until there are no more to be deleted. ++ * Normally, we run in the background doing the deletes and sleeping again, ++ * and clients just add new inodes to be deleted onto the end of the list. ++ * If someone is concerned about free space (e.g. block allocation or similar) ++ * then they can sleep on s_delete_waiter_queue and be woken up when space ++ * has been freed. ++ */ ++int ext3_delete_thread(void *data) ++{ ++ struct super_block *sb = data; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ struct task_struct *tsk = current; ++ ++ /* Almost like daemonize, but not quite */ ++ exit_mm(current); ++ tsk->session = 1; ++ tsk->pgrp = 1; ++ tsk->tty = NULL; ++ exit_files(current); ++ reparent_to_init(); ++ ++ sprintf(tsk->comm, "kdelext3-%s", kdevname(sb->s_dev)); ++ sigfillset(&tsk->blocked); ++ ++ /*tsk->flags |= PF_KERNTHREAD;*/ ++ ++ INIT_LIST_HEAD(&sbi->s_delete_list); ++ wake_up(&sbi->s_delete_waiter_queue); ++ ext3_debug("delete thread on %s started\n", kdevname(sb->s_dev)); ++ ++ /* main loop */ ++ for (;;) { ++ wait_event_interruptible(sbi->s_delete_thread_queue, ++ !list_empty(&sbi->s_delete_list) || ++ !test_opt(sb, ASYNCDEL)); ++ ext3_debug("%s woken up: %lu inodes, %lu blocks\n", ++ tsk->comm,sbi->s_delete_inodes,sbi->s_delete_blocks); ++ ++ spin_lock(&sbi->s_delete_lock); ++ if (list_empty(&sbi->s_delete_list)) { ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ memset(&sbi->s_delete_list, 0, ++ sizeof(sbi->s_delete_list)); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("delete thread on %s exiting\n", ++ kdevname(sb->s_dev)); ++ wake_up(&sbi->s_delete_waiter_queue); ++ break; ++ } ++ ++ while (!list_empty(&sbi->s_delete_list)) { ++ struct inode *inode=list_entry(sbi->s_delete_list.next, ++ struct inode, i_dentry); ++ unsigned long blocks = inode->i_blocks >> ++ (inode->i_blkbits - 9); ++ ++ list_del_init(&inode->i_dentry); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("%s delete ino %lu blk %lu\n", ++ tsk->comm, inode->i_ino, blocks); ++ ++ iput(inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ sbi->s_delete_blocks -= blocks; ++ sbi->s_delete_inodes--; ++ } ++ if (sbi->s_delete_blocks != 0 || sbi->s_delete_inodes != 0) { ++ ext3_warning(sb, __FUNCTION__, ++ "%lu blocks, %lu inodes on list?\n", ++ sbi->s_delete_blocks,sbi->s_delete_inodes); ++ sbi->s_delete_blocks = 0; ++ sbi->s_delete_inodes = 0; ++ } ++ spin_unlock(&sbi->s_delete_lock); ++ wake_up(&sbi->s_delete_waiter_queue); ++ } ++ ++ return 0; ++} ++ ++static void ext3_start_delete_thread(struct super_block *sb) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ int rc; ++ ++ spin_lock_init(&sbi->s_delete_lock); ++ init_waitqueue_head(&sbi->s_delete_thread_queue); ++ init_waitqueue_head(&sbi->s_delete_waiter_queue); ++ ++ if (!test_opt(sb, ASYNCDEL)) ++ return; ++ ++ rc = kernel_thread(ext3_delete_thread, sb, CLONE_VM | CLONE_FILES); ++ if (rc < 0) ++ printk(KERN_ERR "EXT3-fs: cannot start delete thread: rc %d\n", ++ rc); ++ else ++ wait_event(sbi->s_delete_waiter_queue, sbi->s_delete_list.next); ++} ++ ++static void ext3_stop_delete_thread(struct ext3_sb_info *sbi) ++{ ++ if (sbi->s_delete_list.next == 0) /* thread never started */ ++ return; ++ ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ wake_up(&sbi->s_delete_thread_queue); ++ wait_event(sbi->s_delete_waiter_queue, list_empty(&sbi->s_delete_list)); ++} ++ ++/* Instead of playing games with the inode flags, destruction, etc we just ++ * create a new inode locally and put it on a list for the truncate thread. ++ * We need large parts of the inode struct in order to complete the ++ * truncate and unlink, so we may as well just have a real inode to do it. ++ * ++ * If we have any problem deferring the delete, just delete it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ */ ++static void ext3_delete_inode_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (is_bad_inode(old_inode)) { ++ clear_inode(old_inode); ++ return; ++ } ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_delete; ++ ++ /* We may want to delete the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS) ++ goto out_delete; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_delete; ++ } ++ ++ /* We can iget this inode again here, because our caller has unhashed ++ * old_inode, so new_inode will be in a different inode struct. ++ * ++ * We need to ensure that the i_orphan pointers in the other inodes ++ * point at the new inode copy instead of the old one so the orphan ++ * list doesn't get corrupted when the old orphan inode is freed. ++ */ ++ down(&sbi->s_orphan_lock); ++ ++ sbi->s_mount_state |= EXT3_ORPHAN_FS; ++ new_inode = iget(old_inode->i_sb, old_inode->i_ino); ++ sbi->s_mount_state &= ~EXT3_ORPHAN_FS; ++ if (is_bad_inode(new_inode)) { ++ printk(KERN_WARNING "read bad inode %lu\n", old_inode->i_ino); ++ iput(new_inode); ++ new_inode = NULL; ++ } ++ if (!new_inode) { ++ up(&sbi->s_orphan_lock); ++ ext3_debug("delete inode %lu directly (bad read)\n", ++ old_inode->i_ino); ++ goto out_delete; ++ } ++ J_ASSERT(new_inode != old_inode); ++ ++ J_ASSERT(!list_empty(&oei->i_orphan)); ++ ++ nei = EXT3_I(new_inode); ++ /* Ugh. We need to insert new_inode into the same spot on the list ++ * as old_inode was, to ensure the in-memory orphan list is still ++ * in the same order as the on-disk orphan list (badness otherwise). ++ */ ++ nei->i_orphan = oei->i_orphan; ++ nei->i_orphan.next->prev = &nei->i_orphan; ++ nei->i_orphan.prev->next = &nei->i_orphan; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up(&sbi->s_orphan_lock); ++ ++ clear_inode(old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_delete: ++ ext3_delete_inode(old_inode); ++} ++#else ++#define ext3_start_delete_thread(sbi) do {} while(0) ++#define ext3_stop_delete_thread(sbi) do {} while(0) ++#endif /* EXT3_DELETE_THREAD */ ++ + void ext3_put_super (struct super_block * sb) + { + struct ext3_sb_info *sbi = EXT3_SB(sb); +@@ -408,6 +622,7 @@ + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ ext3_stop_delete_thread(sbi); + ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { +@@ -476,7 +691,11 @@ + write_inode: ext3_write_inode, /* BKL not held. Don't need */ + dirty_inode: ext3_dirty_inode, /* BKL not held. We take it */ + put_inode: ext3_put_inode, /* BKL not held. Don't need */ ++#ifdef EXT3_DELETE_THREAD ++ delete_inode: ext3_delete_inode_thread,/* BKL not held. We take it */ ++#else + delete_inode: ext3_delete_inode, /* BKL not held. We take it */ ++#endif + put_super: ext3_put_super, /* BKL held */ + write_super: ext3_write_super, /* BKL held */ + sync_fs: ext3_sync_fs, +@@ -553,6 +772,13 @@ + clear_opt (*mount_options, POSIX_ACL); + else + #endif ++#ifdef EXT3_DELETE_THREAD ++ if (!strcmp(this_char, "asyncdel")) ++ set_opt(*mount_options, ASYNCDEL); ++ else if (!strcmp(this_char, "noasyncdel")) ++ clear_opt(*mount_options, ASYNCDEL); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -1254,6 +1480,7 @@ + } + + ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY); ++ ext3_start_delete_thread(sb); + /* + * akpm: core read_super() calls in here with the superblock locked. + * That deadlocks, because orphan cleanup needs to lock the superblock +@@ -1692,6 +1919,9 @@ + if (!parse_options(data, &tmp, sbi, &tmp, 1)) + return -EINVAL; + ++ if (!test_opt(sb, ASYNCDEL) || (*flags & MS_RDONLY)) ++ ext3_stop_delete_thread(sbi); ++ + if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) + ext3_abort(sb, __FUNCTION__, "Abort forced by user"); + +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:02:56 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:19:22 2003 +@@ -2114,6 +2114,118 @@ + ext3_journal_stop(handle, inode); + } + ++#ifdef EXT3_DELETE_THREAD ++/* Move blocks from to-be-truncated inode over to a new inode, and delete ++ * that one from the delete thread instead. This avoids a lot of latency ++ * when truncating large files. ++ * ++ * If we have any problem deferring the truncate, just truncate it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ */ ++void ext3_truncate_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ handle_t *handle; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_truncate; ++ ++ /* XXX This is a temporary limitation for code simplicity. ++ * We could truncate to arbitrary sizes at some later time. ++ */ ++ if (old_inode->i_size != 0) ++ goto out_truncate; ++ ++ /* We may want to truncate the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS || ++ old_inode->i_size > oei->i_disksize) ++ goto out_truncate; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_truncate; ++ } ++ ++ ext3_discard_prealloc(old_inode); ++ ++ /* old_inode = 1 ++ * new_inode = sb + GDT + ibitmap ++ * orphan list = 1 inode/superblock for add, 2 inodes for del ++ * quota files = 2 * EXT3_SINGLEDATA_TRANS_BLOCKS ++ */ ++ handle = ext3_journal_start(old_inode, 7); ++ if (IS_ERR(handle)) ++ goto out_truncate; ++ ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ if (IS_ERR(new_inode)) { ++ ext3_debug("truncate inode %lu directly (no new inodes)\n", ++ old_inode->i_ino); ++ goto out_journal; ++ } ++ ++ nei = EXT3_I(new_inode); ++ ++ down_write(&oei->truncate_sem); ++ new_inode->i_size = old_inode->i_size; ++ new_inode->i_blocks = old_inode->i_blocks; ++ new_inode->i_uid = old_inode->i_uid; ++ new_inode->i_gid = old_inode->i_gid; ++ new_inode->i_nlink = 0; ++ ++ /* FIXME when we do arbitrary truncates */ ++ old_inode->i_blocks = oei->i_file_acl ? old_inode->i_blksize / 512 : 0; ++ old_inode->i_mtime = old_inode->i_ctime = CURRENT_TIME; ++ ++ memcpy(nei->i_data, oei->i_data, sizeof(nei->i_data)); ++ memset(oei->i_data, 0, sizeof(oei->i_data)); ++ ++ nei->i_disksize = oei->i_disksize; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up_write(&oei->truncate_sem); ++ ++ if (ext3_orphan_add(handle, new_inode) < 0) ++ goto out_journal; ++ ++ if (ext3_orphan_del(handle, old_inode) < 0) { ++ ext3_orphan_del(handle, new_inode); ++ iput(new_inode); ++ goto out_journal; ++ } ++ ++ ext3_journal_stop(handle, old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_journal: ++ ext3_journal_stop(handle, old_inode); ++out_truncate: ++ ext3_truncate(old_inode); ++} ++#endif /* EXT3_DELETE_THREAD */ ++ + /* + * ext3_get_inode_loc returns with an extra refcount against the + * inode's underlying buffer_head on success. +Index: linux-2.4.19.SuSE/fs/ext3/file.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/file.c Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/file.c Sun Nov 16 01:19:22 2003 +@@ -132,7 +132,11 @@ + }; + + struct inode_operations ext3_file_inode_operations = { ++#ifdef EXT3_DELETE_THREAD ++ truncate: ext3_truncate_thread, /* BKL held */ ++#else + truncate: ext3_truncate, /* BKL held */ ++#endif + setattr: ext3_setattr, /* BKL held */ + setxattr: ext3_setxattr, /* BKL held */ + getxattr: ext3_getxattr, /* BKL held */ +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:02:51 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:20:06 2003 +@@ -193,6 +193,7 @@ + */ + #define EXT3_STATE_JDATA 0x00000001 /* journaled data exists */ + #define EXT3_STATE_NEW 0x00000002 /* inode is newly created */ ++#define EXT3_STATE_DELETE 0x00000010 /* deferred delete inode */ + + /* + * ioctl commands +@@ -321,6 +322,7 @@ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ + #define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + #define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */ ++#define EXT3_MOUNT_ASYNCDEL 0x20000 /* Delayed deletion */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H +@@ -695,6 +697,9 @@ + extern void ext3_dirty_inode(struct inode *); + extern int ext3_change_inode_journal_flag(struct inode *, int); + extern void ext3_truncate (struct inode *); ++#ifdef EXT3_DELETE_THREAD ++extern void ext3_truncate_thread(struct inode *inode); ++#endif + + /* ioctl.c */ + extern int ext3_ioctl (struct inode *, struct file *, unsigned int, +Index: linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs_sb.h Sun Nov 16 01:18:41 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h Sun Nov 16 01:19:22 2003 +@@ -29,6 +29,8 @@ + + #define EXT3_MAX_GROUP_LOADED 8 + ++#define EXT3_DELETE_THREAD ++ + /* + * third extended-fs super-block data in memory + */ +@@ -75,6 +77,14 @@ + struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */ + wait_queue_head_t ro_wait_queue; /* For people waiting for the fs to go read-only */ + #endif ++#ifdef EXT3_DELETE_THREAD ++ spinlock_t s_delete_lock; ++ struct list_head s_delete_list; ++ unsigned long s_delete_blocks; ++ unsigned long s_delete_inodes; ++ wait_queue_head_t s_delete_thread_queue; ++ wait_queue_head_t s_delete_waiter_queue; ++#endif + }; + + #endif /* _LINUX_EXT3_FS_SB */ diff --git a/lustre/kernel_patches/patches/ext3-htree-2.4.19-pre1.patch b/lustre/kernel_patches/patches/ext3-htree-2.4.19-pre1.patch new file mode 100644 index 0000000..711e1f2 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-htree-2.4.19-pre1.patch @@ -0,0 +1,2584 @@ + fs/ext3/Makefile | 2 + fs/ext3/dir.c | 302 +++++++++ + fs/ext3/file.c | 3 + fs/ext3/hash.c | 215 ++++++ + fs/ext3/namei.c | 1420 ++++++++++++++++++++++++++++++++++++++++----- + fs/ext3/super.c | 7 + include/linux/ext3_fs.h | 85 ++ + include/linux/ext3_fs_sb.h | 2 + include/linux/ext3_jbd.h | 2 + include/linux/rbtree.h | 2 + lib/rbtree.c | 42 + + 11 files changed, 1921 insertions(+), 161 deletions(-) + +Index: linux-2.4.19-pre1/fs/ext3/dir.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/dir.c 2001-11-10 01:25:04.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/dir.c 2003-11-20 23:05:26.000000000 +0300 +@@ -21,12 +21,16 @@ + #include + #include + #include ++#include ++#include + + static unsigned char ext3_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK + }; + + static int ext3_readdir(struct file *, void *, filldir_t); ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir); + + struct file_operations ext3_dir_operations = { + read: generic_read_dir, +@@ -35,6 +39,17 @@ + fsync: ext3_sync_file, /* BKL held */ + }; + ++ ++static unsigned char get_dtype(struct super_block *sb, int filetype) ++{ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) || ++ (filetype >= EXT3_FT_MAX)) ++ return DT_UNKNOWN; ++ ++ return (ext3_filetype_table[filetype]); ++} ++ ++ + int ext3_check_dir_entry (const char * function, struct inode * dir, + struct ext3_dir_entry_2 * de, + struct buffer_head * bh, +@@ -79,6 +94,16 @@ + + sb = inode->i_sb; + ++ if (is_dx(inode)) { ++ err = ext3_dx_readdir(filp, dirent, filldir); ++ if (err != ERR_BAD_DX_DIR) ++ return err; ++ /* ++ * We don't set the inode dirty flag since it's not ++ * critical that it get flushed back to the disk. ++ */ ++ EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; ++ } + stored = 0; + bh = NULL; + offset = filp->f_pos & (sb->s_blocksize - 1); +@@ -162,18 +187,12 @@ + * during the copy operation. + */ + unsigned long version = filp->f_version; +- unsigned char d_type = DT_UNKNOWN; + +- if (EXT3_HAS_INCOMPAT_FEATURE(sb, +- EXT3_FEATURE_INCOMPAT_FILETYPE) +- && de->file_type < EXT3_FT_MAX) +- d_type = +- ext3_filetype_table[de->file_type]; + error = filldir(dirent, de->name, + de->name_len, + filp->f_pos, + le32_to_cpu(de->inode), +- d_type); ++ get_dtype(sb, de->file_type)); + if (error) + break; + if (version != filp->f_version) +@@ -188,3 +207,272 @@ + UPDATE_ATIME(inode); + return 0; + } ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * These functions convert from the major/minor hash to an f_pos ++ * value. ++ * ++ * Currently we only use major hash numer. This is unfortunate, but ++ * on 32-bit machines, the same VFS interface is used for lseek and ++ * llseek, so if we use the 64 bit offset, then the 32-bit versions of ++ * lseek/telldir/seekdir will blow out spectacularly, and from within ++ * the ext2 low-level routine, we don't know if we're being called by ++ * a 64-bit version of the system call or the 32-bit version of the ++ * system call. Worse yet, NFSv2 only allows for a 32-bit readdir ++ * cookie. Sigh. ++ */ ++#define hash2pos(major, minor) (major >> 1) ++#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff) ++#define pos2min_hash(pos) (0) ++ ++/* ++ * This structure holds the nodes of the red-black tree used to store ++ * the directory entry in hash order. ++ */ ++struct fname { ++ __u32 hash; ++ __u32 minor_hash; ++ rb_node_t rb_hash; ++ struct fname *next; ++ __u32 inode; ++ __u8 name_len; ++ __u8 file_type; ++ char name[0]; ++}; ++ ++/* ++ * This functoin implements a non-recursive way of freeing all of the ++ * nodes in the red-black tree. ++ */ ++static void free_rb_tree_fname(rb_root_t *root) ++{ ++ rb_node_t *n = root->rb_node; ++ rb_node_t *parent; ++ struct fname *fname; ++ ++ while (n) { ++ /* Do the node's children first */ ++ if ((n)->rb_left) { ++ n = n->rb_left; ++ continue; ++ } ++ if (n->rb_right) { ++ n = n->rb_right; ++ continue; ++ } ++ /* ++ * The node has no children; free it, and then zero ++ * out parent's link to it. Finally go to the ++ * beginning of the loop and try to free the parent ++ * node. ++ */ ++ parent = n->rb_parent; ++ fname = rb_entry(n, struct fname, rb_hash); ++ kfree(fname); ++ if (!parent) ++ root->rb_node = 0; ++ else if (parent->rb_left == n) ++ parent->rb_left = 0; ++ else if (parent->rb_right == n) ++ parent->rb_right = 0; ++ n = parent; ++ } ++ root->rb_node = 0; ++} ++ ++ ++struct dir_private_info *create_dir_info(loff_t pos) ++{ ++ struct dir_private_info *p; ++ ++ p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL); ++ if (!p) ++ return NULL; ++ p->root.rb_node = 0; ++ p->curr_node = 0; ++ p->extra_fname = 0; ++ p->last_pos = 0; ++ p->curr_hash = pos2maj_hash(pos); ++ p->curr_minor_hash = pos2min_hash(pos); ++ p->next_hash = 0; ++ return p; ++} ++ ++void ext3_htree_free_dir_info(struct dir_private_info *p) ++{ ++ free_rb_tree_fname(&p->root); ++ kfree(p); ++} ++ ++/* ++ * Given a directory entry, enter it into the fname rb tree. ++ */ ++int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent) ++{ ++ rb_node_t **p, *parent = NULL; ++ struct fname * fname, *new_fn; ++ struct dir_private_info *info; ++ int len; ++ ++ info = (struct dir_private_info *) dir_file->private_data; ++ p = &info->root.rb_node; ++ ++ /* Create and allocate the fname structure */ ++ len = sizeof(struct fname) + dirent->name_len + 1; ++ new_fn = kmalloc(len, GFP_KERNEL); ++ if (!new_fn) ++ return -ENOMEM; ++ memset(new_fn, 0, len); ++ new_fn->hash = hash; ++ new_fn->minor_hash = minor_hash; ++ new_fn->inode = le32_to_cpu(dirent->inode); ++ new_fn->name_len = dirent->name_len; ++ new_fn->file_type = dirent->file_type; ++ memcpy(new_fn->name, dirent->name, dirent->name_len); ++ new_fn->name[dirent->name_len] = 0; ++ ++ while (*p) { ++ parent = *p; ++ fname = rb_entry(parent, struct fname, rb_hash); ++ ++ /* ++ * If the hash and minor hash match up, then we put ++ * them on a linked list. This rarely happens... ++ */ ++ if ((new_fn->hash == fname->hash) && ++ (new_fn->minor_hash == fname->minor_hash)) { ++ new_fn->next = fname->next; ++ fname->next = new_fn; ++ return 0; ++ } ++ ++ if (new_fn->hash < fname->hash) ++ p = &(*p)->rb_left; ++ else if (new_fn->hash > fname->hash) ++ p = &(*p)->rb_right; ++ else if (new_fn->minor_hash < fname->minor_hash) ++ p = &(*p)->rb_left; ++ else /* if (new_fn->minor_hash > fname->minor_hash) */ ++ p = &(*p)->rb_right; ++ } ++ ++ rb_link_node(&new_fn->rb_hash, parent, p); ++ rb_insert_color(&new_fn->rb_hash, &info->root); ++ return 0; ++} ++ ++ ++ ++/* ++ * This is a helper function for ext3_dx_readdir. It calls filldir ++ * for all entres on the fname linked list. (Normally there is only ++ * one entry on the linked list, unless there are 62 bit hash collisions.) ++ */ ++static int call_filldir(struct file * filp, void * dirent, ++ filldir_t filldir, struct fname *fname) ++{ ++ struct dir_private_info *info = filp->private_data; ++ loff_t curr_pos; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct super_block * sb; ++ int error; ++ ++ sb = inode->i_sb; ++ ++ if (!fname) { ++ printk("call_filldir: called with null fname?!?\n"); ++ return 0; ++ } ++ curr_pos = hash2pos(fname->hash, fname->minor_hash); ++ while (fname) { ++ error = filldir(dirent, fname->name, ++ fname->name_len, curr_pos, ++ fname->inode, ++ get_dtype(sb, fname->file_type)); ++ if (error) { ++ filp->f_pos = curr_pos; ++ info->extra_fname = fname->next; ++ return error; ++ } ++ fname = fname->next; ++ } ++ return 0; ++} ++ ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir) ++{ ++ struct dir_private_info *info = filp->private_data; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct fname *fname; ++ int ret; ++ ++ if (!info) { ++ info = create_dir_info(filp->f_pos); ++ if (!info) ++ return -ENOMEM; ++ filp->private_data = info; ++ } ++ ++ /* Some one has messed with f_pos; reset the world */ ++ if (info->last_pos != filp->f_pos) { ++ free_rb_tree_fname(&info->root); ++ info->curr_node = 0; ++ info->extra_fname = 0; ++ info->curr_hash = pos2maj_hash(filp->f_pos); ++ info->curr_minor_hash = pos2min_hash(filp->f_pos); ++ } ++ ++ /* ++ * If there are any leftover names on the hash collision ++ * chain, return them first. ++ */ ++ if (info->extra_fname && ++ call_filldir(filp, dirent, filldir, info->extra_fname)) ++ goto finished; ++ ++ if (!info->curr_node) ++ info->curr_node = rb_get_first(&info->root); ++ ++ while (1) { ++ /* ++ * Fill the rbtree if we have no more entries, ++ * or the inode has changed since we last read in the ++ * cached entries. ++ */ ++ if ((!info->curr_node) || ++ (filp->f_version != inode->i_version)) { ++ info->curr_node = 0; ++ free_rb_tree_fname(&info->root); ++ filp->f_version = inode->i_version; ++ ret = ext3_htree_fill_tree(filp, info->curr_hash, ++ info->curr_minor_hash, ++ &info->next_hash); ++ if (ret < 0) ++ return ret; ++ if (ret == 0) ++ break; ++ info->curr_node = rb_get_first(&info->root); ++ } ++ ++ fname = rb_entry(info->curr_node, struct fname, rb_hash); ++ info->curr_hash = fname->hash; ++ info->curr_minor_hash = fname->minor_hash; ++ if (call_filldir(filp, dirent, filldir, fname)) ++ break; ++ ++ info->curr_node = rb_get_next(info->curr_node); ++ if (!info->curr_node) { ++ info->curr_hash = info->next_hash; ++ info->curr_minor_hash = 0; ++ } ++ } ++finished: ++ info->last_pos = filp->f_pos; ++ UPDATE_ATIME(inode); ++ return 0; ++} ++#endif +Index: linux-2.4.19-pre1/fs/ext3/file.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/file.c 2001-11-16 00:37:55.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/file.c 2003-11-20 23:05:26.000000000 +0300 +@@ -35,6 +35,9 @@ + { + if (filp->f_mode & FMODE_WRITE) + ext3_discard_prealloc (inode); ++ if (is_dx(inode) && filp->private_data) ++ ext3_htree_free_dir_info(filp->private_data); ++ + return 0; + } + +Index: linux-2.4.19-pre1/fs/ext3/hash.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/hash.c 2003-11-20 23:05:26.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/hash.c 2003-11-20 23:05:26.000000000 +0300 +@@ -0,0 +1,215 @@ ++/* ++ * linux/fs/ext3/hash.c ++ * ++ * Copyright (C) 2002 by Theodore Ts'o ++ * ++ * This file is released under the GPL v2. ++ * ++ * This file may be redistributed under the terms of the GNU Public ++ * License. ++ */ ++ ++#include ++#include ++#include ++#include ++ ++#define DELTA 0x9E3779B9 ++ ++static void TEA_transform(__u32 buf[4], __u32 const in[]) ++{ ++ __u32 sum = 0; ++ __u32 b0 = buf[0], b1 = buf[1]; ++ __u32 a = in[0], b = in[1], c = in[2], d = in[3]; ++ int n = 16; ++ ++ do { ++ sum += DELTA; ++ b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b); ++ b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d); ++ } while(--n); ++ ++ buf[0] += b0; ++ buf[1] += b1; ++} ++ ++/* F, G and H are basic MD4 functions: selection, majority, parity */ ++#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z)))) ++#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z))) ++#define H(x, y, z) ((x) ^ (y) ^ (z)) ++ ++/* ++ * The generic round function. The application is so specific that ++ * we don't bother protecting all the arguments with parens, as is generally ++ * good macro practice, in favor of extra legibility. ++ * Rotation is separate from addition to prevent recomputation ++ */ ++#define ROUND(f, a, b, c, d, x, s) \ ++ (a += f(b, c, d) + x, a = (a << s) | (a >> (32-s))) ++#define K1 0 ++#define K2 013240474631UL ++#define K3 015666365641UL ++ ++/* ++ * Basic cut-down MD4 transform. Returns only 32 bits of result. ++ */ ++static void halfMD4Transform (__u32 buf[4], __u32 const in[]) ++{ ++ __u32 a = buf[0], b = buf[1], c = buf[2], d = buf[3]; ++ ++ /* Round 1 */ ++ ROUND(F, a, b, c, d, in[0] + K1, 3); ++ ROUND(F, d, a, b, c, in[1] + K1, 7); ++ ROUND(F, c, d, a, b, in[2] + K1, 11); ++ ROUND(F, b, c, d, a, in[3] + K1, 19); ++ ROUND(F, a, b, c, d, in[4] + K1, 3); ++ ROUND(F, d, a, b, c, in[5] + K1, 7); ++ ROUND(F, c, d, a, b, in[6] + K1, 11); ++ ROUND(F, b, c, d, a, in[7] + K1, 19); ++ ++ /* Round 2 */ ++ ROUND(G, a, b, c, d, in[1] + K2, 3); ++ ROUND(G, d, a, b, c, in[3] + K2, 5); ++ ROUND(G, c, d, a, b, in[5] + K2, 9); ++ ROUND(G, b, c, d, a, in[7] + K2, 13); ++ ROUND(G, a, b, c, d, in[0] + K2, 3); ++ ROUND(G, d, a, b, c, in[2] + K2, 5); ++ ROUND(G, c, d, a, b, in[4] + K2, 9); ++ ROUND(G, b, c, d, a, in[6] + K2, 13); ++ ++ /* Round 3 */ ++ ROUND(H, a, b, c, d, in[3] + K3, 3); ++ ROUND(H, d, a, b, c, in[7] + K3, 9); ++ ROUND(H, c, d, a, b, in[2] + K3, 11); ++ ROUND(H, b, c, d, a, in[6] + K3, 15); ++ ROUND(H, a, b, c, d, in[1] + K3, 3); ++ ROUND(H, d, a, b, c, in[5] + K3, 9); ++ ROUND(H, c, d, a, b, in[0] + K3, 11); ++ ROUND(H, b, c, d, a, in[4] + K3, 15); ++ ++ buf[0] += a; ++ buf[1] += b; ++ buf[2] += c; ++ buf[3] += d; ++} ++ ++#undef ROUND ++#undef F ++#undef G ++#undef H ++#undef K1 ++#undef K2 ++#undef K3 ++ ++/* The old legacy hash */ ++static __u32 dx_hack_hash (const char *name, int len) ++{ ++ __u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; ++ while (len--) { ++ __u32 hash = hash1 + (hash0 ^ (*name++ * 7152373)); ++ ++ if (hash & 0x80000000) hash -= 0x7fffffff; ++ hash1 = hash0; ++ hash0 = hash; ++ } ++ return (hash0 << 1); ++} ++ ++static void str2hashbuf(const char *msg, int len, __u32 *buf, int num) ++{ ++ __u32 pad, val; ++ int i; ++ ++ pad = (__u32)len | ((__u32)len << 8); ++ pad |= pad << 16; ++ ++ val = pad; ++ if (len > num*4) ++ len = num * 4; ++ for (i=0; i < len; i++) { ++ if ((i % 4) == 0) ++ val = pad; ++ val = msg[i] + (val << 8); ++ if ((i % 4) == 3) { ++ *buf++ = val; ++ val = pad; ++ num--; ++ } ++ } ++ if (--num >= 0) ++ *buf++ = val; ++ while (--num >= 0) ++ *buf++ = pad; ++} ++ ++/* ++ * Returns the hash of a filename. If len is 0 and name is NULL, then ++ * this function can be used to test whether or not a hash version is ++ * supported. ++ * ++ * The seed is an 4 longword (32 bits) "secret" which can be used to ++ * uniquify a hash. If the seed is all zero's, then some default seed ++ * may be used. ++ * ++ * A particular hash version specifies whether or not the seed is ++ * represented, and whether or not the returned hash is 32 bits or 64 ++ * bits. 32 bit hashes will return 0 for the minor hash. ++ */ ++int ext3fs_dirhash(const char *name, int len, struct dx_hash_info *hinfo) ++{ ++ __u32 hash; ++ __u32 minor_hash = 0; ++ const char *p; ++ int i; ++ __u32 in[8], buf[4]; ++ ++ /* Initialize the default seed for the hash checksum functions */ ++ buf[0] = 0x67452301; ++ buf[1] = 0xefcdab89; ++ buf[2] = 0x98badcfe; ++ buf[3] = 0x10325476; ++ ++ /* Check to see if the seed is all zero's */ ++ if (hinfo->seed) { ++ for (i=0; i < 4; i++) { ++ if (hinfo->seed[i]) ++ break; ++ } ++ if (i < 4) ++ memcpy(buf, hinfo->seed, sizeof(buf)); ++ } ++ ++ switch (hinfo->hash_version) { ++ case DX_HASH_LEGACY: ++ hash = dx_hack_hash(name, len); ++ break; ++ case DX_HASH_HALF_MD4: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 8); ++ halfMD4Transform(buf, in); ++ len -= 32; ++ p += 32; ++ } ++ minor_hash = buf[2]; ++ hash = buf[1]; ++ break; ++ case DX_HASH_TEA: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 4); ++ TEA_transform(buf, in); ++ len -= 16; ++ p += 16; ++ } ++ hash = buf[0]; ++ minor_hash = buf[1]; ++ break; ++ default: ++ hinfo->hash = 0; ++ return -1; ++ } ++ hinfo->hash = hash & ~1; ++ hinfo->minor_hash = minor_hash; ++ return 0; ++} +Index: linux-2.4.19-pre1/fs/ext3/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/Makefile 2003-11-20 19:01:58.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/Makefile 2003-11-20 23:05:26.000000000 +0300 +@@ -12,7 +12,7 @@ + export-objs := super.o inode.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o ++ ioctl.o namei.o super.o symlink.o hash.o + obj-m := $(O_TARGET) + + include $(TOPDIR)/Rules.make +Index: linux-2.4.19-pre1/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/namei.c 2001-11-10 01:25:04.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/namei.c 2003-11-20 23:22:54.000000000 +0300 +@@ -16,6 +16,12 @@ + * David S. Miller (davem@caip.rutgers.edu), 1995 + * Directory entry file type support and forward compatibility hooks + * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 ++ * Hash Tree Directory indexing (c) ++ * Daniel Phillips, 2001 ++ * Hash Tree Directory indexing porting ++ * Christopher Li, 2002 ++ * Hash Tree Directory indexing cleanup ++ * Theodore Ts'o, 2002 + */ + + #include +@@ -38,6 +44,642 @@ + #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) + #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + ++static struct buffer_head *ext3_append(handle_t *handle, ++ struct inode *inode, ++ u32 *block, int *err) ++{ ++ struct buffer_head *bh; ++ ++ *block = inode->i_size >> inode->i_sb->s_blocksize_bits; ++ ++ if ((bh = ext3_bread(handle, inode, *block, 1, err))) { ++ inode->i_size += inode->i_sb->s_blocksize; ++ EXT3_I(inode)->i_disksize = inode->i_size; ++ ext3_journal_get_write_access(handle,bh); ++ } ++ return bh; ++} ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#ifndef swap ++#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0) ++#endif ++ ++typedef struct { u32 v; } le_u32; ++typedef struct { u16 v; } le_u16; ++ ++#ifdef DX_DEBUG ++#define dxtrace(command) command ++#else ++#define dxtrace(command) ++#endif ++ ++struct fake_dirent ++{ ++ /*le*/u32 inode; ++ /*le*/u16 rec_len; ++ u8 name_len; ++ u8 file_type; ++}; ++ ++struct dx_countlimit ++{ ++ le_u16 limit; ++ le_u16 count; ++}; ++ ++struct dx_entry ++{ ++ le_u32 hash; ++ le_u32 block; ++}; ++ ++/* ++ * dx_root_info is laid out so that if it should somehow get overlaid by a ++ * dirent the two low bits of the hash version will be zero. Therefore, the ++ * hash version mod 4 should never be 0. Sincerely, the paranoia department. ++ */ ++ ++struct dx_root ++{ ++ struct fake_dirent dot; ++ char dot_name[4]; ++ struct fake_dirent dotdot; ++ char dotdot_name[4]; ++ struct dx_root_info ++ { ++ le_u32 reserved_zero; ++ u8 hash_version; ++ u8 info_length; /* 8 */ ++ u8 indirect_levels; ++ u8 unused_flags; ++ } ++ info; ++ struct dx_entry entries[0]; ++}; ++ ++struct dx_node ++{ ++ struct fake_dirent fake; ++ struct dx_entry entries[0]; ++}; ++ ++ ++struct dx_frame ++{ ++ struct buffer_head *bh; ++ struct dx_entry *entries; ++ struct dx_entry *at; ++}; ++ ++struct dx_map_entry ++{ ++ u32 hash; ++ u32 offs; ++}; ++ ++#ifdef CONFIG_EXT3_INDEX ++static inline unsigned dx_get_block (struct dx_entry *entry); ++static void dx_set_block (struct dx_entry *entry, unsigned value); ++static inline unsigned dx_get_hash (struct dx_entry *entry); ++static void dx_set_hash (struct dx_entry *entry, unsigned value); ++static unsigned dx_get_count (struct dx_entry *entries); ++static unsigned dx_get_limit (struct dx_entry *entries); ++static void dx_set_count (struct dx_entry *entries, unsigned value); ++static void dx_set_limit (struct dx_entry *entries, unsigned value); ++static unsigned dx_root_limit (struct inode *dir, unsigned infosize); ++static unsigned dx_node_limit (struct inode *dir); ++static struct dx_frame *dx_probe(struct dentry *dentry, ++ struct inode *dir, ++ struct dx_hash_info *hinfo, ++ struct dx_frame *frame, ++ int *err); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static struct ext3_dir_entry_2 *dx_move_dirents (char *from, char *to, ++ struct dx_map_entry *offsets, int count); ++static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash); ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err); ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode); ++ ++/* ++ * Future: use high four bits of block for coalesce-on-delete flags ++ * Mask them off for now. ++ */ ++ ++static inline unsigned dx_get_block (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->block.v) & 0x00ffffff; ++} ++ ++static inline void dx_set_block (struct dx_entry *entry, unsigned value) ++{ ++ entry->block.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_hash (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->hash.v); ++} ++ ++static inline void dx_set_hash (struct dx_entry *entry, unsigned value) ++{ ++ entry->hash.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_count (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->count.v); ++} ++ ++static inline unsigned dx_get_limit (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->limit.v); ++} ++ ++static inline void dx_set_count (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->count.v = cpu_to_le16(value); ++} ++ ++static inline void dx_set_limit (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->limit.v = cpu_to_le16(value); ++} ++ ++static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(1) - ++ EXT3_DIR_REC_LEN(2) - infosize; ++ return 0? 20: entry_space / sizeof(struct dx_entry); ++} ++ ++static inline unsigned dx_node_limit (struct inode *dir) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(0); ++ return 0? 22: entry_space / sizeof(struct dx_entry); ++} ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext3_dir_entry_2 *de, ++ int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ printk("names: "); ++ while ((char *) de < base + size) ++ { ++ if (de->inode) ++ { ++ if (show_names) ++ { ++ int len = de->name_len; ++ char *name = de->name; ++ while (len--) printk("%c", *name++); ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ printk(":%x.%u ", h.hash, ++ ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, ++ struct dx_entry *entries, int levels) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count = dx_get_count (entries), names = 0, space = 0, i; ++ unsigned bcount = 0; ++ struct buffer_head *bh; ++ int err; ++ printk("%i indexed blocks...\n", count); ++ for (i = 0; i < count; i++, entries++) ++ { ++ u32 block = dx_get_block(entries), hash = i? dx_get_hash(entries): 0; ++ u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; ++ struct stats stats; ++ printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); ++ if (!(bh = ext3_bread (NULL,dir, block, 0,&err))) continue; ++ stats = levels? ++ dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf(hinfo, (struct ext3_dir_entry_2 *) bh->b_data, blocksize, 0); ++ names += stats.names; ++ space += stats.space; ++ bcount += stats.bcount; ++ brelse (bh); ++ } ++ if (bcount) ++ printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", ++ names, space/bcount,(space/bcount)*100/blocksize); ++ return (struct stats) { names, space, bcount}; ++} ++#endif /* DX_DEBUG */ ++ ++/* ++ * Probe for a directory leaf block to search. ++ * ++ * dx_probe can return ERR_BAD_DX_DIR, which means there was a format ++ * error in the directory index, and the caller should fall back to ++ * searching the directory normally. The callers of dx_probe **MUST** ++ * check for this error code, and make sure it never gets reflected ++ * back to userspace. ++ */ ++static struct dx_frame * ++dx_probe(struct dentry *dentry, struct inode *dir, ++ struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) ++{ ++ unsigned count, indirect; ++ struct dx_entry *at, *entries, *p, *q, *m; ++ struct dx_root *root; ++ struct buffer_head *bh; ++ struct dx_frame *frame = frame_in; ++ u32 hash; ++ ++ frame->bh = NULL; ++ if (dentry) ++ dir = dentry->d_parent->d_inode; ++ if (!(bh = ext3_bread (NULL,dir, 0, 0, err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version != DX_HASH_TEA && ++ root->info.hash_version != DX_HASH_HALF_MD4 && ++ root->info.hash_version != DX_HASH_LEGACY) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unrecognised inode hash code %d", ++ root->info.hash_version); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ hinfo->hash_version = root->info.hash_version; ++ hinfo->seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ if (dentry) ++ ext3fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); ++ hash = hinfo->hash; ++ ++ if (root->info.unused_flags & 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash flags: %#06x", ++ root->info.unused_flags); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ if ((indirect = root->info.indirect_levels) > 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash depth: %#06x", ++ root->info.indirect_levels); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ entries = (struct dx_entry *) (((char *)&root->info) + ++ root->info.info_length); ++ assert(dx_get_limit(entries) == dx_root_limit(dir, ++ root->info.info_length)); ++ dxtrace (printk("Look up %x", hash)); ++ while (1) ++ { ++ count = dx_get_count(entries); ++ assert (count && count <= dx_get_limit(entries)); ++ p = entries + 1; ++ q = entries + count - 1; ++ while (p <= q) ++ { ++ m = p + (q - p)/2; ++ dxtrace(printk(".")); ++ if (dx_get_hash(m) > hash) ++ q = m - 1; ++ else ++ p = m + 1; ++ } ++ ++ if (0) // linear search cross check ++ { ++ unsigned n = count - 1; ++ at = entries; ++ while (n--) ++ { ++ dxtrace(printk(",")); ++ if (dx_get_hash(++at) > hash) ++ { ++ at--; ++ break; ++ } ++ } ++ assert (at == p - 1); ++ } ++ ++ at = p - 1; ++ dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); ++ frame->bh = bh; ++ frame->entries = entries; ++ frame->at = at; ++ if (!indirect--) return frame; ++ if (!(bh = ext3_bread (NULL,dir, dx_get_block(at), 0, err))) ++ goto fail2; ++ at = entries = ((struct dx_node *) bh->b_data)->entries; ++ assert (dx_get_limit(entries) == dx_node_limit (dir)); ++ frame++; ++ } ++fail2: ++ while (frame >= frame_in) { ++ brelse(frame->bh); ++ frame--; ++ } ++fail: ++ return NULL; ++} ++ ++static void dx_release (struct dx_frame *frames) ++{ ++ if (frames[0].bh == NULL) ++ return; ++ ++ if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) ++ brelse(frames[1].bh); ++ brelse(frames[0].bh); ++} ++ ++/* ++ * This function increments the frame pointer to search the next leaf ++ * block, and reads in the necessary intervening nodes if the search ++ * should be necessary. Whether or not the search is necessary is ++ * controlled by the hash parameter. If the hash value is even, then ++ * the search is only continued if the next block starts with that ++ * hash value. This is used if we are searching for a specific file. ++ * ++ * If the hash value is HASH_NB_ALWAYS, then always go to the next block. ++ * ++ * This function returns 1 if the caller should continue to search, ++ * or 0 if it should not. If there is an error reading one of the ++ * index blocks, it will return -1. ++ * ++ * If start_hash is non-null, it will be filled in with the starting ++ * hash of the next page. ++ */ ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash) ++{ ++ struct dx_frame *p; ++ struct buffer_head *bh; ++ int num_frames = 0; ++ __u32 bhash; ++ ++ *err = ENOENT; ++ p = frame; ++ /* ++ * Find the next leaf page by incrementing the frame pointer. ++ * If we run out of entries in the interior node, loop around and ++ * increment pointer in the parent node. When we break out of ++ * this loop, num_frames indicates the number of interior ++ * nodes need to be read. ++ */ ++ while (1) { ++ if (++(p->at) < p->entries + dx_get_count(p->entries)) ++ break; ++ if (p == frames) ++ return 0; ++ num_frames++; ++ p--; ++ } ++ ++ /* ++ * If the hash is 1, then continue only if the next page has a ++ * continuation hash of any value. This is used for readdir ++ * handling. Otherwise, check to see if the hash matches the ++ * desired contiuation hash. If it doesn't, return since ++ * there's no point to read in the successive index pages. ++ */ ++ bhash = dx_get_hash(p->at); ++ if (start_hash) ++ *start_hash = bhash; ++ if ((hash & 1) == 0) { ++ if ((bhash & ~1) != hash) ++ return 0; ++ } ++ /* ++ * If the hash is HASH_NB_ALWAYS, we always go to the next ++ * block so no check is necessary ++ */ ++ while (num_frames--) { ++ if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), ++ 0, err))) ++ return -1; /* Failure */ ++ p++; ++ brelse (p->bh); ++ p->bh = bh; ++ p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; ++ } ++ return 1; ++} ++ ++ ++/* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *p) ++{ ++ return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++/* ++ * This function fills a red-black tree with information from a ++ * directory. We start scanning the directory in hash order, starting ++ * at start_hash and start_minor_hash. ++ * ++ * This function returns the number of entries inserted into the tree, ++ * or a negative error code. ++ */ ++int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash) ++{ ++ struct dx_hash_info hinfo; ++ struct buffer_head *bh; ++ struct ext3_dir_entry_2 *de, *top; ++ static struct dx_frame frames[2], *frame; ++ struct inode *dir; ++ int block, err; ++ int count = 0; ++ int ret; ++ __u32 hashval; ++ ++ dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, ++ start_minor_hash)); ++ dir = dir_file->f_dentry->d_inode; ++ hinfo.hash = start_hash; ++ hinfo.minor_hash = 0; ++ frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ ++ /* Add '.' and '..' from the htree header */ ++ if (!start_hash && !start_minor_hash) { ++ de = (struct ext3_dir_entry_2 *) frames[0].bh->b_data; ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ de = ext3_next_entry(de); ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ count += 2; ++ } ++ ++ while (1) { ++ block = dx_get_block(frame->at); ++ dxtrace(printk("Reading block %d\n", block)); ++ if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) ++ goto errout; ++ ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + dir->i_sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) { ++ ext3fs_dirhash(de->name, de->name_len, &hinfo); ++ if ((hinfo.hash < start_hash) || ++ ((hinfo.hash == start_hash) && ++ (hinfo.minor_hash < start_minor_hash))) ++ continue; ++ if ((err = ext3_htree_store_dirent(dir_file, ++ hinfo.hash, hinfo.minor_hash, de)) != 0) ++ goto errout; ++ count++; ++ } ++ brelse (bh); ++ hashval = ~1; ++ ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ++ frame, frames, &err, &hashval); ++ if (next_hash) ++ *next_hash = hashval; ++ if (ret == -1) ++ goto errout; ++ /* ++ * Stop if: (a) there are no more entries, or ++ * (b) we have inserted at least one entry and the ++ * next hash value is not a continuation ++ */ ++ if ((ret == 0) || ++ (count && ((hashval & 1) == 0))) ++ break; ++ } ++ dx_release(frames); ++ dxtrace(printk("Fill tree: returned %d entries\n", count)); ++ return count; ++errout: ++ dx_release(frames); ++ return (err); ++} ++ ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry *map_tail) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ while ((char *) de < base + size) ++ { ++ if (de->name_len && de->inode) { ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ map_tail--; ++ map_tail->hash = h.hash; ++ map_tail->offs = (u32) ((char *) de - base); ++ count++; ++ } ++ /* XXX: do we need to check rec_len == 0 case? -Chris */ ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ return count; ++} ++ ++static void dx_sort_map (struct dx_map_entry *map, unsigned count) ++{ ++ struct dx_map_entry *p, *q, *top = map + count - 1; ++ int more; ++ /* Combsort until bubble sort doesn't suck */ ++ while (count > 2) ++ { ++ count = count*10/13; ++ if (count - 9 < 2) /* 9, 10 -> 11 */ ++ count = 11; ++ for (p = top, q = p - count; q >= map; p--, q--) ++ if (p->hash < q->hash) ++ swap(*p, *q); ++ } ++ /* Garden variety bubble sort */ ++ do { ++ more = 0; ++ q = top; ++ while (q-- > map) ++ { ++ if (q[1].hash >= q[0].hash) ++ continue; ++ swap(*(q+1), *q); ++ more = 1; ++ } ++ } while(more); ++} ++ ++static void dx_insert_block(struct dx_frame *frame, u32 hash, u32 block) ++{ ++ struct dx_entry *entries = frame->entries; ++ struct dx_entry *old = frame->at, *new = old + 1; ++ int count = dx_get_count(entries); ++ ++ assert(count < dx_get_limit(entries)); ++ assert(old < entries + count); ++ memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); ++ dx_set_hash(new, hash); ++ dx_set_block(new, block); ++ dx_set_count(entries, count + 1); ++} ++#endif ++ ++ ++static void ext3_update_dx_flag(struct inode *inode) ++{ ++ if (!EXT3_HAS_COMPAT_FEATURE(inode->i_sb, ++ EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -94,6 +736,7 @@ + return 0; + } + ++ + /* + * ext3_find_entry() + * +@@ -105,6 +748,8 @@ + * The returned buffer_head has ->b_count elevated. The caller is expected + * to brelse() it when appropriate. + */ ++ ++ + static struct buffer_head * ext3_find_entry (struct dentry *dentry, + struct ext3_dir_entry_2 ** res_dir) + { +@@ -119,12 +764,32 @@ + int num = 0; + int nblocks, i, err; + struct inode *dir = dentry->d_parent->d_inode; ++ int namelen; ++ const u8 *name; ++ unsigned blocksize; + + *res_dir = NULL; + sb = dir->i_sb; +- ++ blocksize = sb->s_blocksize; ++ namelen = dentry->d_name.len; ++ name = dentry->d_name.name; ++ if (namelen > EXT3_NAME_LEN) ++ return NULL; ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ bh = ext3_dx_find_entry(dentry, res_dir, &err); ++ /* ++ * On success, or if the error was file not found, ++ * return. Otherwise, fall back to doing a search the ++ * old fashioned way. ++ */ ++ if (bh || (err != ERR_BAD_DX_DIR)) ++ return bh; ++ dxtrace(printk("ext3_find_entry: dx failed, falling back\n")); ++ } ++#endif + nblocks = dir->i_size >> EXT3_BLOCK_SIZE_BITS(sb); +- start = dir->u.ext3_i.i_dir_start_lookup; ++ start = EXT3_I(dir)->i_dir_start_lookup; + if (start >= nblocks) + start = 0; + block = start; +@@ -165,7 +830,7 @@ + i = search_dirblock(bh, dir, dentry, + block << EXT3_BLOCK_SIZE_BITS(sb), res_dir); + if (i == 1) { +- dir->u.ext3_i.i_dir_start_lookup = block; ++ EXT3_I(dir)->i_dir_start_lookup = block; + ret = bh; + goto cleanup_and_exit; + } else { +@@ -196,6 +861,66 @@ + return ret; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err) ++{ ++ struct super_block * sb; ++ struct dx_hash_info hinfo; ++ u32 hash; ++ struct dx_frame frames[2], *frame; ++ struct ext3_dir_entry_2 *de, *top; ++ struct buffer_head *bh; ++ unsigned long block; ++ int retval; ++ int namelen = dentry->d_name.len; ++ const u8 *name = dentry->d_name.name; ++ struct inode *dir = dentry->d_parent->d_inode; ++ ++ sb = dir->i_sb; ++ if (!(frame = dx_probe (dentry, 0, &hinfo, frames, err))) ++ return NULL; ++ hash = hinfo.hash; ++ do { ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, err))) ++ goto errout; ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match (namelen, name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto errout; ++ } ++ *res_dir = de; ++ dx_release (frames); ++ return bh; ++ } ++ brelse (bh); ++ /* Check to see if we should continue to search */ ++ retval = ext3_htree_next_block(dir, hash, frame, ++ frames, err, 0); ++ if (retval == -1) { ++ ext3_warning(sb, __FUNCTION__, ++ "error reading index page in directory #%lu", ++ dir->i_ino); ++ goto errout; ++ } ++ } while (retval == 1); ++ ++ *err = -ENOENT; ++errout: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++} ++#endif ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -212,8 +937,9 @@ + brelse (bh); + inode = iget(dir->i_sb, ino); + +- if (!inode) ++ if (!inode) { + return ERR_PTR(-EACCES); ++ } + } + d_add(dentry, inode); + return NULL; +@@ -237,6 +963,301 @@ + de->file_type = ext3_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct ext3_dir_entry_2 * ++dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count) ++{ ++ unsigned rec_len = 0; ++ ++ while (count--) { ++ struct ext3_dir_entry_2 *de = (struct ext3_dir_entry_2 *) (from + map->offs); ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ memcpy (to, de, rec_len); ++ ((struct ext3_dir_entry_2 *) to)->rec_len = rec_len; ++ de->inode = 0; ++ map++; ++ to += rec_len; ++ } ++ return (struct ext3_dir_entry_2 *) (to - rec_len); ++} ++ ++static struct ext3_dir_entry_2* dx_pack_dirents(char *base, int size) ++{ ++ struct ext3_dir_entry_2 *next, *to, *prev, *de = (struct ext3_dir_entry_2 *) base; ++ unsigned rec_len = 0; ++ ++ prev = to = de; ++ while ((char*)de < base + size) { ++ next = (struct ext3_dir_entry_2 *) ((char *) de + ++ le16_to_cpu(de->rec_len)); ++ if (de->inode && de->name_len) { ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ if (de > to) ++ memmove(to, de, rec_len); ++ to->rec_len = rec_len; ++ prev = to; ++ to = (struct ext3_dir_entry_2 *) (((char *) to) + rec_len); ++ } ++ de = next; ++ } ++ return prev; ++} ++ ++static struct ext3_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, ++ struct buffer_head **bh,struct dx_frame *frame, ++ struct dx_hash_info *hinfo, int *error) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count, continued; ++ struct buffer_head *bh2; ++ u32 newblock; ++ u32 hash2; ++ struct dx_map_entry *map; ++ char *data1 = (*bh)->b_data, *data2; ++ unsigned split; ++ struct ext3_dir_entry_2 *de = NULL, *de2; ++ int err; ++ ++ bh2 = ext3_append (handle, dir, &newblock, error); ++ if (!(bh2)) { ++ brelse(*bh); ++ *bh = NULL; ++ goto errout; ++ } ++ ++ BUFFER_TRACE(*bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, *bh); ++ if (err) { ++ journal_error: ++ brelse(*bh); ++ brelse(bh2); ++ *bh = NULL; ++ ext3_std_error(dir->i_sb, err); ++ goto errout; ++ } ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ ++ data2 = bh2->b_data; ++ ++ /* create map in the end of data2 block */ ++ map = (struct dx_map_entry *) (data2 + blocksize); ++ count = dx_make_map ((struct ext3_dir_entry_2 *) data1, ++ blocksize, hinfo, map); ++ map -= count; ++ split = count/2; // need to adjust to actual middle ++ dx_sort_map (map, count); ++ hash2 = map[split].hash; ++ continued = hash2 == map[split - 1].hash; ++ dxtrace(printk("Split block %i at %x, %i/%i\n", ++ dx_get_block(frame->at), hash2, split, count-split)); ++ ++ /* Fancy dance to stay within two buffers */ ++ de2 = dx_move_dirents(data1, data2, map + split, count - split); ++ de = dx_pack_dirents(data1,blocksize); ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ de2->rec_len = cpu_to_le16(data2 + blocksize - (char *) de2); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data1, blocksize, 1)); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data2, blocksize, 1)); ++ ++ /* Which block gets the new entry? */ ++ if (hinfo->hash >= hash2) ++ { ++ swap(*bh, bh2); ++ de = de2; ++ } ++ dx_insert_block (frame, hash2 + continued, newblock); ++ err = ext3_journal_dirty_metadata (handle, bh2); ++ if (err) ++ goto journal_error; ++ err = ext3_journal_dirty_metadata (handle, frame->bh); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ dxtrace(dx_show_index ("frame", frame->entries)); ++errout: ++ return de; ++} ++#endif ++ ++ ++/* ++ * Add a new entry into a directory (leaf) block. If de is non-NULL, ++ * it points to a directory entry which is guaranteed to be large ++ * enough for new directory entry. If de is NULL, then ++ * add_dirent_to_buf will attempt search the directory block for ++ * space. It will return -ENOSPC if no space is available, and -EIO ++ * and -EEXIST if directory entry already exists. ++ * ++ * NOTE! bh is NOT released in the case where ENOSPC is returned. In ++ * all other cases bh is released. ++ */ ++static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct ext3_dir_entry_2 *de, ++ struct buffer_head * bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ unsigned long offset = 0; ++ unsigned short reclen; ++ int nlen, rlen, err; ++ char *top; ++ ++ reclen = EXT3_DIR_REC_LEN(namelen); ++ if (!de) { ++ de = (struct ext3_dir_entry_2 *)bh->b_data; ++ top = bh->b_data + dir->i_sb->s_blocksize - reclen; ++ while ((char *) de <= top) { ++ if (!ext3_check_dir_entry("ext3_add_entry", dir, de, ++ bh, offset)) { ++ brelse (bh); ++ return -EIO; ++ } ++ if (ext3_match (namelen, name, de)) { ++ brelse (bh); ++ return -EEXIST; ++ } ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if ((de->inode? rlen - nlen: rlen) >= reclen) ++ break; ++ de = (struct ext3_dir_entry_2 *)((char *)de + rlen); ++ offset += rlen; ++ } ++ if ((char *) de > top) ++ return -ENOSPC; ++ } ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) { ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return err; ++ } ++ ++ /* By now the buffer is marked for journaling */ ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if (de->inode) { ++ struct ext3_dir_entry_2 *de1 = (struct ext3_dir_entry_2 *)((char *)de + nlen); ++ de1->rec_len = cpu_to_le16(rlen - nlen); ++ de->rec_len = cpu_to_le16(nlen); ++ de = de1; ++ } ++ de->file_type = EXT3_FT_UNKNOWN; ++ if (inode) { ++ de->inode = cpu_to_le32(inode->i_ino); ++ ext3_set_de_type(dir->i_sb, de, inode->i_mode); ++ } else ++ de->inode = 0; ++ de->name_len = namelen; ++ memcpy (de->name, name, namelen); ++ /* ++ * XXX shouldn't update any times until successful ++ * completion of syscall, but too many callers depend ++ * on this. ++ * ++ * XXX similarly, too many callers depend on ++ * ext3_new_inode() setting the times, but error ++ * recovery deletes the inode, so the worst that can ++ * happen is that the times are slightly out of date ++ * and/or different from the directory change time. ++ */ ++ dir->i_mtime = dir->i_ctime = CURRENT_TIME; ++ ext3_update_dx_flag(dir); ++ dir->i_version = ++event; ++ ext3_mark_inode_dirty(handle, dir); ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return 0; ++} ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * This converts a one block unindexed directory to a 3 block indexed ++ * directory, and adds the dentry to the indexed directory. ++ */ ++static int make_indexed_dir(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct buffer_head *bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ struct buffer_head *bh2; ++ struct dx_root *root; ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries; ++ struct ext3_dir_entry_2 *de, *de2; ++ char *data1, *top; ++ unsigned len; ++ int retval; ++ unsigned blocksize; ++ struct dx_hash_info hinfo; ++ u32 block; ++ ++ blocksize = dir->i_sb->s_blocksize; ++ dxtrace(printk("Creating index\n")); ++ retval = ext3_journal_get_write_access(handle, bh); ++ if (retval) { ++ ext3_std_error(dir->i_sb, retval); ++ brelse(bh); ++ return retval; ++ } ++ root = (struct dx_root *) bh->b_data; ++ ++ EXT3_I(dir)->i_flags |= EXT3_INDEX_FL; ++ bh2 = ext3_append (handle, dir, &block, &retval); ++ if (!(bh2)) { ++ brelse(bh); ++ return retval; ++ } ++ data1 = bh2->b_data; ++ ++ /* The 0th block becomes the root, move the dirents out */ ++ de = (struct ext3_dir_entry_2 *) &root->dotdot; ++ de = (struct ext3_dir_entry_2 *) ((char *)de + de->rec_len); ++ len = ((char *) root) + blocksize - (char *) de; ++ memcpy (data1, de, len); ++ de = (struct ext3_dir_entry_2 *) data1; ++ top = data1 + len; ++ while (((char *) de2=(char*)de+le16_to_cpu(de->rec_len)) < top) ++ de = de2; ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ /* Initialize the root; the dot dirents already exist */ ++ de = (struct ext3_dir_entry_2 *) (&root->dotdot); ++ de->rec_len = cpu_to_le16(blocksize - EXT3_DIR_REC_LEN(2)); ++ memset (&root->info, 0, sizeof(root->info)); ++ root->info.info_length = sizeof(root->info); ++ root->info.hash_version = dir->i_sb->u.ext3_sb.s_def_hash_version; ++ entries = root->entries; ++ dx_set_block (entries, 1); ++ dx_set_count (entries, 1); ++ dx_set_limit (entries, dx_root_limit(dir, sizeof(root->info))); ++ ++ /* Initialize as for dx_probe */ ++ hinfo.hash_version = root->info.hash_version; ++ hinfo.seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ ext3fs_dirhash(name, namelen, &hinfo); ++ frame = frames; ++ frame->entries = entries; ++ frame->at = entries; ++ frame->bh = bh; ++ bh = bh2; ++ de = do_split(handle,dir, &bh, frame, &hinfo, &retval); ++ dx_release (frames); ++ if (!(de)) ++ return retval; ++ ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} ++#endif ++ + /* + * ext3_add_entry() + * +@@ -247,127 +1268,198 @@ + * may not sleep between calling this and putting something into + * the entry, as someone else might have used it while you slept. + */ +- +-/* +- * AKPM: the journalling code here looks wrong on the error paths +- */ + static int ext3_add_entry (handle_t *handle, struct dentry *dentry, + struct inode *inode) + { + struct inode *dir = dentry->d_parent->d_inode; +- const char *name = dentry->d_name.name; +- int namelen = dentry->d_name.len; + unsigned long offset; +- unsigned short rec_len; + struct buffer_head * bh; +- struct ext3_dir_entry_2 * de, * de1; ++ struct ext3_dir_entry_2 *de; + struct super_block * sb; + int retval; ++#ifdef CONFIG_EXT3_INDEX ++ int dx_fallback=0; ++#endif ++ unsigned blocksize; ++ unsigned nlen, rlen; ++ u32 block, blocks; + + sb = dir->i_sb; +- +- if (!namelen) ++ blocksize = sb->s_blocksize; ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &retval); ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ retval = ext3_dx_add_entry(handle, dentry, inode); ++ if (!retval || (retval != ERR_BAD_DX_DIR)) ++ return retval; ++ EXT3_I(dir)->i_flags &= ~EXT3_INDEX_FL; ++ dx_fallback++; ++ ext3_mark_inode_dirty(handle, dir); ++ } ++#endif ++ blocks = dir->i_size >> sb->s_blocksize_bits; ++ for (block = 0, offset = 0; block < blocks; block++) { ++ bh = ext3_bread(handle, dir, block, 0, &retval); ++ if(!bh) ++ return retval; ++ retval = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (retval != -ENOSPC) ++ return retval; ++ ++#ifdef CONFIG_EXT3_INDEX ++ if (blocks == 1 && !dx_fallback && ++ EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ return make_indexed_dir(handle, dentry, inode, bh); ++#endif ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); + if (!bh) + return retval; +- rec_len = EXT3_DIR_REC_LEN(namelen); +- offset = 0; + de = (struct ext3_dir_entry_2 *) bh->b_data; +- while (1) { +- if ((char *)de >= sb->s_blocksize + bh->b_data) { +- brelse (bh); +- bh = NULL; +- bh = ext3_bread (handle, dir, +- offset >> EXT3_BLOCK_SIZE_BITS(sb), 1, &retval); +- if (!bh) +- return retval; +- if (dir->i_size <= offset) { +- if (dir->i_size == 0) { +- brelse(bh); +- return -ENOENT; +- } ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = blocksize); ++ nlen = 0; ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} + +- ext3_debug ("creating next block\n"); ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * Returns 0 for success, or a negative error value ++ */ ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode) ++{ ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ struct dx_hash_info hinfo; ++ struct buffer_head * bh; ++ struct inode *dir = dentry->d_parent->d_inode; ++ struct super_block * sb = dir->i_sb; ++ struct ext3_dir_entry_2 *de; ++ int err; + +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- } else { ++ frame = dx_probe(dentry, 0, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ entries = frame->entries; ++ at = frame->at; + +- ext3_debug ("skipping to next block\n"); ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err))) ++ goto cleanup; + +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- } +- } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { +- brelse (bh); +- return -EEXIST; ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) ++ goto journal_error; ++ ++ err = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (err != -ENOSPC) { ++ bh = 0; ++ goto cleanup; ++ } ++ ++ /* Block full, should compress but for now just split */ ++ dxtrace(printk("using %u of %u node entries\n", ++ dx_get_count(entries), dx_get_limit(entries))); ++ /* Need to split index? */ ++ if (dx_get_count(entries) == dx_get_limit(entries)) { ++ u32 newblock; ++ unsigned icount = dx_get_count(entries); ++ int levels = frame - frames; ++ struct dx_entry *entries2; ++ struct dx_node *node2; ++ struct buffer_head *bh2; ++ ++ if (levels && (dx_get_count(frames->entries) == ++ dx_get_limit(frames->entries))) { ++ ext3_warning(sb, __FUNCTION__, ++ "Directory index full!\n"); ++ err = -ENOSPC; ++ goto cleanup; + } +- if ((le32_to_cpu(de->inode) == 0 && +- le16_to_cpu(de->rec_len) >= rec_len) || +- (le16_to_cpu(de->rec_len) >= +- EXT3_DIR_REC_LEN(de->name_len) + rec_len)) { +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- /* By now the buffer is marked for journaling */ +- offset += le16_to_cpu(de->rec_len); +- if (le32_to_cpu(de->inode)) { +- de1 = (struct ext3_dir_entry_2 *) ((char *) de + +- EXT3_DIR_REC_LEN(de->name_len)); +- de1->rec_len = +- cpu_to_le16(le16_to_cpu(de->rec_len) - +- EXT3_DIR_REC_LEN(de->name_len)); +- de->rec_len = cpu_to_le16( +- EXT3_DIR_REC_LEN(de->name_len)); +- de = de1; ++ bh2 = ext3_append (handle, dir, &newblock, &err); ++ if (!(bh2)) ++ goto cleanup; ++ node2 = (struct dx_node *)(bh2->b_data); ++ entries2 = node2->entries; ++ node2->fake.rec_len = cpu_to_le16(sb->s_blocksize); ++ node2->fake.inode = 0; ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ if (levels) { ++ unsigned icount1 = icount/2, icount2 = icount - icount1; ++ unsigned hash2 = dx_get_hash(entries + icount1); ++ dxtrace(printk("Split index %i/%i\n", icount1, icount2)); ++ ++ BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ ++ err = ext3_journal_get_write_access(handle, ++ frames[0].bh); ++ if (err) ++ goto journal_error; ++ ++ memcpy ((char *) entries2, (char *) (entries + icount1), ++ icount2 * sizeof(struct dx_entry)); ++ dx_set_count (entries, icount1); ++ dx_set_count (entries2, icount2); ++ dx_set_limit (entries2, dx_node_limit(dir)); ++ ++ /* Which index block gets the new entry? */ ++ if (at - entries >= icount1) { ++ frame->at = at = at - entries - icount1 + entries2; ++ frame->entries = entries = entries2; ++ swap(frame->bh, bh2); + } +- de->file_type = EXT3_FT_UNKNOWN; +- if (inode) { +- de->inode = cpu_to_le32(inode->i_ino); +- ext3_set_de_type(dir->i_sb, de, inode->i_mode); +- } else +- de->inode = 0; +- de->name_len = namelen; +- memcpy (de->name, name, namelen); +- /* +- * XXX shouldn't update any times until successful +- * completion of syscall, but too many callers depend +- * on this. +- * +- * XXX similarly, too many callers depend on +- * ext3_new_inode() setting the times, but error +- * recovery deletes the inode, so the worst that can +- * happen is that the times are slightly out of date +- * and/or different from the directory change time. +- */ +- dir->i_mtime = dir->i_ctime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- dir->i_version = ++event; +- BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +- ext3_journal_dirty_metadata(handle, bh); +- brelse(bh); +- return 0; ++ dx_insert_block (frames + 0, hash2, newblock); ++ dxtrace(dx_show_index ("node", frames[1].entries)); ++ dxtrace(dx_show_index ("node", ++ ((struct dx_node *) bh2->b_data)->entries)); ++ err = ext3_journal_dirty_metadata(handle, bh2); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ } else { ++ dxtrace(printk("Creating second level index...\n")); ++ memcpy((char *) entries2, (char *) entries, ++ icount * sizeof(struct dx_entry)); ++ dx_set_limit(entries2, dx_node_limit(dir)); ++ ++ /* Set up root */ ++ dx_set_count(entries, 1); ++ dx_set_block(entries + 0, newblock); ++ ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; ++ ++ /* Add new access path frame */ ++ frame = frames + 1; ++ frame->at = at = at - entries + entries2; ++ frame->entries = entries = entries2; ++ frame->bh = bh2; ++ err = ext3_journal_get_write_access(handle, ++ frame->bh); ++ if (err) ++ goto journal_error; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- brelse (bh); +- return -ENOSPC; ++ de = do_split(handle, dir, &bh, frame, &hinfo, &err); ++ if (!de) ++ goto cleanup; ++ err = add_dirent_to_buf(handle, dentry, inode, de, bh); ++ bh = 0; ++ goto cleanup; ++ ++journal_error: ++ ext3_std_error(dir->i_sb, err); ++cleanup: ++ if (bh) ++ brelse(bh); ++ dx_release(frames); ++ return err; + } ++#endif + + /* + * ext3_delete_entry deletes a directory entry by merging it with the +@@ -451,9 +1543,11 @@ + struct inode * inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -478,9 +1572,11 @@ + struct inode *inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -507,9 +1603,11 @@ + if (dir->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -521,7 +1619,7 @@ + + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; +- inode->i_size = inode->u.ext3_i.i_disksize = inode->i_sb->s_blocksize; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { +@@ -554,21 +1652,19 @@ + inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); +- if (err) +- goto out_no_entry; ++ if (err) { ++ inode->i_nlink = 0; ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + dir->i_nlink++; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + d_instantiate(dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- inode->i_nlink = 0; +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + /* +@@ -655,7 +1751,7 @@ + int err = 0, rc; + + lock_super(sb); +- if (!list_empty(&inode->u.ext3_i.i_orphan)) ++ if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + + /* Orphan handling is only valid for files with data blocks +@@ -696,7 +1792,7 @@ + * This is safe: on error we're going to ignore the orphan list + * anyway on the next recovery. */ + if (!err) +- list_add(&inode->u.ext3_i.i_orphan, &EXT3_SB(sb)->s_orphan); ++ list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); + + jbd_debug(4, "superblock will point to %ld\n", inode->i_ino); + jbd_debug(4, "orphan inode %ld will point to %d\n", +@@ -714,25 +1810,26 @@ + int ext3_orphan_del(handle_t *handle, struct inode *inode) + { + struct list_head *prev; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct ext3_sb_info *sbi; + ino_t ino_next; + struct ext3_iloc iloc; + int err = 0; + + lock_super(inode->i_sb); +- if (list_empty(&inode->u.ext3_i.i_orphan)) { ++ if (list_empty(&ei->i_orphan)) { + unlock_super(inode->i_sb); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); +- prev = inode->u.ext3_i.i_orphan.prev; ++ prev = ei->i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %ld from orphan list\n", inode->i_ino); + +- list_del(&inode->u.ext3_i.i_orphan); +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ list_del(&ei->i_orphan); ++ INIT_LIST_HEAD(&ei->i_orphan); + + /* If we're on an error path, we may not have a valid + * transaction handle with which to update the orphan list on +@@ -793,8 +1890,9 @@ + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + retval = -ENOENT; + bh = ext3_find_entry (dentry, &de); +@@ -832,7 +1930,7 @@ + ext3_mark_inode_dirty(handle, inode); + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -850,8 +1948,9 @@ + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -878,7 +1977,7 @@ + if (retval) + goto end_unlink; + dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + inode->i_nlink--; + if (!inode->i_nlink) +@@ -904,9 +2003,11 @@ + if (l > dir->i_sb->s_blocksize) + return -ENAMETOOLONG; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 5); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -916,7 +2017,7 @@ + if (IS_ERR(inode)) + goto out_stop; + +- if (l > sizeof (inode->u.ext3_i.i_data)) { ++ if (l > sizeof (EXT3_I(inode)->i_data)) { + inode->i_op = &page_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* +@@ -925,8 +2026,12 @@ + * i_size in generic_commit_write(). + */ + err = block_symlink(inode, symname, l); +- if (err) +- goto out_no_entry; ++ if (err) { ++ ext3_dec_count(handle, inode); ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + } else { + inode->i_op = &ext3_fast_symlink_inode_operations; + memcpy((char*)&inode->u.ext3_i.i_data,symname,l); +@@ -938,12 +2043,6 @@ + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- ext3_dec_count(handle, inode); +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + static int ext3_link (struct dentry * old_dentry, +@@ -956,12 +2055,15 @@ + if (S_ISDIR(inode->i_mode)) + return -EPERM; + +- if (inode->i_nlink >= EXT3_LINK_MAX) ++ if (inode->i_nlink >= EXT3_LINK_MAX) { + return -EMLINK; ++ } + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -995,9 +2097,11 @@ + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 2); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(old_dir) || IS_SYNC(new_dir)) + handle->h_sync = 1; +@@ -1070,14 +2174,33 @@ + /* + * ok, that's it + */ +- ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ retval = ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ if (retval == -ENOENT) { ++ /* ++ * old_de could have moved out from under us. ++ */ ++ struct buffer_head *old_bh2; ++ struct ext3_dir_entry_2 *old_de2; ++ ++ old_bh2 = ext3_find_entry(old_dentry, &old_de2); ++ if (old_bh2) { ++ retval = ext3_delete_entry(handle, old_dir, ++ old_de2, old_bh2); ++ brelse(old_bh2); ++ } ++ } ++ if (retval) { ++ ext3_warning(old_dir->i_sb, "ext3_rename", ++ "Deleting old file (%lu), %d, error=%d", ++ old_dir->i_ino, old_dir->i_nlink, retval); ++ } + + if (new_inode) { + new_inode->i_nlink--; + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +2212,7 @@ + new_inode->i_nlink--; + } else { + new_dir->i_nlink++; +- new_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(new_dir); + ext3_mark_inode_dirty(handle, new_dir); + } + } +Index: linux-2.4.19-pre1/fs/ext3/super.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/super.c 2003-11-20 19:01:58.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/super.c 2003-11-20 23:05:26.000000000 +0300 +@@ -702,6 +702,7 @@ + es->s_mtime = cpu_to_le32(CURRENT_TIME); + ext3_update_dynamic_rev(sb); + EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); ++ + ext3_commit_super (sb, es, 1); + if (test_opt (sb, DEBUG)) + printk (KERN_INFO +@@ -712,6 +713,7 @@ + EXT3_BLOCKS_PER_GROUP(sb), + EXT3_INODES_PER_GROUP(sb), + sbi->s_mount_opt); ++ + printk(KERN_INFO "EXT3 FS " EXT3FS_VERSION ", " EXT3FS_DATE " on %s, ", + bdevname(sb->s_dev)); + if (EXT3_SB(sb)->s_journal->j_inode == NULL) { +@@ -886,6 +888,7 @@ + return res; + } + ++ + struct super_block * ext3_read_super (struct super_block * sb, void * data, + int silent) + { +@@ -1062,6 +1065,9 @@ + sbi->s_mount_state = le16_to_cpu(es->s_state); + sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb)); + sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb)); ++ for (i=0; i < 4; i++) ++ sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); ++ sbi->s_def_hash_version = es->s_def_hash_version; + + if (sbi->s_blocks_per_group > blocksize * 8) { + printk (KERN_ERR +@@ -1744,6 +1750,7 @@ + unregister_filesystem(&ext3_fs_type); + } + ++EXPORT_SYMBOL(ext3_force_commit); + EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); +Index: linux-2.4.19-pre1/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_fs.h 2003-11-20 19:01:38.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_fs.h 2003-11-20 23:05:26.000000000 +0300 +@@ -40,6 +40,11 @@ + #define EXT3FS_VERSION "2.4-0.9.17" + + /* ++ * Always enable hashed directories ++ */ ++#define CONFIG_EXT3_INDEX ++ ++/* + * Debug code + */ + #ifdef EXT3FS_DEBUG +@@ -437,8 +442,11 @@ + /*E0*/ __u32 s_journal_inum; /* inode number of journal file */ + __u32 s_journal_dev; /* device number of journal file */ + __u32 s_last_orphan; /* start of list of inodes to delete */ +- +-/*EC*/ __u32 s_reserved[197]; /* Padding to the end of the block */ ++ __u32 s_hash_seed[4]; /* HTREE hash seed */ ++ __u8 s_def_hash_version; /* Default hash version to use */ ++ __u8 s_reserved_char_pad; ++ __u16 s_reserved_word_pad; ++ __u32 s_reserved[192]; /* Padding to the end of the block */ + }; + + #ifdef __KERNEL__ +@@ -575,9 +583,46 @@ + #define EXT3_DIR_ROUND (EXT3_DIR_PAD - 1) + #define EXT3_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT3_DIR_ROUND) & \ + ~EXT3_DIR_ROUND) ++/* ++ * Hash Tree Directory indexing ++ * (c) Daniel Phillips, 2001 ++ */ ++ ++#ifdef CONFIG_EXT3_INDEX ++ #define is_dx(dir) (EXT3_HAS_COMPAT_FEATURE(dir->i_sb, \ ++ EXT3_FEATURE_COMPAT_DIR_INDEX) && \ ++ (EXT3_I(dir)->i_flags & EXT3_INDEX_FL)) ++#define EXT3_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) ++#else ++ #define is_dx(dir) 0 ++#define EXT3_DIR_LINK_MAX(dir) ((dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2) ++#endif ++ ++/* Legal values for the dx_root hash_version field: */ ++ ++#define DX_HASH_LEGACY 0 ++#define DX_HASH_HALF_MD4 1 ++#define DX_HASH_TEA 2 ++ ++/* hash info structure used by the directory hash */ ++struct dx_hash_info ++{ ++ u32 hash; ++ u32 minor_hash; ++ int hash_version; ++ u32 *seed; ++}; + + #ifdef __KERNEL__ + /* ++ * Control parameters used by ext3_htree_next_block ++ */ ++#define HASH_NB_ALWAYS 1 ++ ++ ++/* + * Describe an inode's exact location on disk and in memory + */ + struct ext3_iloc +@@ -587,6 +632,27 @@ + unsigned long block_group; + }; + ++ ++/* ++ * This structure is stuffed into the struct file's private_data field ++ * for directories. It is where we put information so that we can do ++ * readdir operations in hash tree order. ++ */ ++struct dir_private_info { ++ rb_root_t root; ++ rb_node_t *curr_node; ++ struct fname *extra_fname; ++ loff_t last_pos; ++ __u32 curr_hash; ++ __u32 curr_minor_hash; ++ __u32 next_hash; ++}; ++ ++/* ++ * Special error return code only used by dx_probe() and its callers. ++ */ ++#define ERR_BAD_DX_DIR -75000 ++ + /* + * Function prototypes + */ +@@ -614,11 +680,20 @@ + + /* dir.c */ + extern int ext3_check_dir_entry(const char *, struct inode *, +- struct ext3_dir_entry_2 *, struct buffer_head *, +- unsigned long); ++ struct ext3_dir_entry_2 *, ++ struct buffer_head *, unsigned long); ++extern int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent); ++extern void ext3_htree_free_dir_info(struct dir_private_info *p); ++ + /* fsync.c */ + extern int ext3_sync_file (struct file *, struct dentry *, int); + ++/* hash.c */ ++extern int ext3fs_dirhash(const char *name, int len, struct ++ dx_hash_info *hinfo); ++ + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); +@@ -650,6 +725,8 @@ + /* namei.c */ + extern int ext3_orphan_add(handle_t *, struct inode *); + extern int ext3_orphan_del(handle_t *, struct inode *); ++extern int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash); + + /* super.c */ + extern void ext3_error (struct super_block *, const char *, const char *, ...) +Index: linux-2.4.19-pre1/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_fs_sb.h 2003-11-20 19:01:38.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_fs_sb.h 2003-11-20 23:05:26.000000000 +0300 +@@ -62,6 +62,8 @@ + int s_inode_size; + int s_first_ino; + u32 s_next_generation; ++ u32 s_hash_seed[4]; ++ int s_def_hash_version; + + /* Journaling */ + struct inode * s_journal_inode; +Index: linux-2.4.19-pre1/include/linux/ext3_jbd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_jbd.h 2001-12-21 20:42:03.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_jbd.h 2003-11-20 23:05:26.000000000 +0300 +@@ -63,6 +63,8 @@ + + #define EXT3_RESERVE_TRANS_BLOCKS 12 + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, +Index: linux-2.4.19-pre1/include/linux/rbtree.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/rbtree.h 2001-11-22 22:46:18.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/rbtree.h 2003-11-20 23:05:26.000000000 +0300 +@@ -120,6 +120,8 @@ + + extern void rb_insert_color(rb_node_t *, rb_root_t *); + extern void rb_erase(rb_node_t *, rb_root_t *); ++extern rb_node_t *rb_get_first(rb_root_t *root); ++extern rb_node_t *rb_get_next(rb_node_t *n); + + static inline void rb_link_node(rb_node_t * node, rb_node_t * parent, rb_node_t ** rb_link) + { +Index: linux-2.4.19-pre1/lib/rbtree.c +=================================================================== +--- linux-2.4.19-pre1.orig/lib/rbtree.c 2001-09-18 02:30:23.000000000 +0400 ++++ linux-2.4.19-pre1/lib/rbtree.c 2003-11-20 23:25:57.000000000 +0300 +@@ -17,6 +17,8 @@ + Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + + linux/lib/rbtree.c ++ ++ rb_get_first and rb_get_next written by Theodore Ts'o, 9/8/2002 + */ + + #include +@@ -291,3 +293,43 @@ + if (color == RB_BLACK) + __rb_erase_color(child, parent, root); + } ++ ++/* ++ * This function returns the first node (in sort order) of the tree. ++ */ ++rb_node_t *rb_get_first(rb_root_t *root) ++{ ++ rb_node_t *n; ++ ++ n = root->rb_node; ++ if (!n) ++ return 0; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++} ++EXPORT_SYMBOL(rb_get_first); ++ ++/* ++ * Given a node, this function will return the next node in the tree. ++ */ ++rb_node_t *rb_get_next(rb_node_t *n) ++{ ++ rb_node_t *parent; ++ ++ if (n->rb_right) { ++ n = n->rb_right; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++ } else { ++ while ((parent = n->rb_parent)) { ++ if (n == parent->rb_left) ++ return parent; ++ n = parent; ++ } ++ return 0; ++ } ++} ++EXPORT_SYMBOL(rb_get_next); ++ diff --git a/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch new file mode 100644 index 0000000..3c6b5e5 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch @@ -0,0 +1,85 @@ +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:14:50 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:18:04 2003 +@@ -1751,8 +1751,8 @@ + struct super_block *sb = inode->i_sb; + struct ext3_iloc iloc; + int err = 0, rc; +- +- lock_super(sb); ++ ++ down(&EXT3_SB(sb)->s_orphan_lock); + if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + +@@ -1800,7 +1800,7 @@ + jbd_debug(4, "orphan inode %ld will point to %d\n", + inode->i_ino, NEXT_ORPHAN(inode)); + out_unlock: +- unlock_super(sb); ++ up(&EXT3_SB(sb)->s_orphan_lock); + ext3_std_error(inode->i_sb, err); + return err; + } +@@ -1813,20 +1813,19 @@ + { + struct list_head *prev; + struct ext3_inode_info *ei = EXT3_I(inode); +- struct ext3_sb_info *sbi; ++ struct ext3_sb_info *sbi = EXT3_SB(inode->i_sb); + unsigned long ino_next; + struct ext3_iloc iloc; + int err = 0; + +- lock_super(inode->i_sb); ++ down(&sbi->s_orphan_lock); + if (list_empty(&ei->i_orphan)) { +- unlock_super(inode->i_sb); ++ up(&sbi->s_orphan_lock); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); + prev = ei->i_orphan.prev; +- sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); + +@@ -1872,10 +1871,10 @@ + if (err) + goto out_brelse; + +-out_err: ++out_err: + ext3_std_error(inode->i_sb, err); + out: +- unlock_super(inode->i_sb); ++ up(&sbi->s_orphan_lock); + return err; + + out_brelse: +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:18:04 2003 +@@ -1182,6 +1182,7 @@ + */ + sb->s_op = &ext3_sops; + INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ ++ sema_init(&sbi->s_orphan_lock, 1); + + sb->s_root = 0; + +Index: linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs_sb.h Sat Nov 15 23:58:28 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h Sun Nov 16 01:18:41 2003 +@@ -69,6 +69,7 @@ + struct inode * s_journal_inode; + struct journal_s * s_journal; + struct list_head s_orphan; ++ struct semaphore s_orphan_lock; + struct block_device *journal_bdev; + #ifdef CONFIG_JBD_DEBUG + struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */ diff --git a/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.22-rh.patch b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.22-rh.patch index 46fc3d5..f2d4889 100644 --- a/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.22-rh.patch +++ b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.22-rh.patch @@ -60,69 +60,6 @@ return err; out_brelse: ---- linux-2.4.22-ac1/fs/ext3/namei.c.orig~ext3-orphan_lock-2.4.22-rh 2003-09-25 14:58:37.000000000 +0400 -+++ linux-2.4.22-ac1-alexey/fs/ext3/namei.c.orig 2003-09-26 00:24:09.000000000 +0400 -@@ -29,6 +29,7 @@ - #include - #include - #include -+#include - #include - #include - #include -@@ -1614,7 +1615,7 @@ static int ext3_mkdir(struct inode * dir - if (IS_SYNC(dir)) - handle->h_sync = 1; - -- inode = ext3_new_inode (handle, dir, S_IFDIR); -+ inode = ext3_new_inode (handle, dir, S_IFDIR | mode); - err = PTR_ERR(inode); - if (IS_ERR(inode)) - goto out_stop; -@@ -1622,7 +1623,6 @@ static int ext3_mkdir(struct inode * dir - inode->i_op = &ext3_dir_inode_operations; - inode->i_fop = &ext3_dir_operations; - inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; -- inode->i_blocks = 0; - dir_block = ext3_bread (handle, inode, 0, 1, &err); - if (!dir_block) { - inode->i_nlink--; /* is this nlink == 0? */ -@@ -1649,9 +1649,6 @@ static int ext3_mkdir(struct inode * dir - BUFFER_TRACE(dir_block, "call ext3_journal_dirty_metadata"); - ext3_journal_dirty_metadata(handle, dir_block); - brelse (dir_block); -- inode->i_mode = S_IFDIR | mode; -- if (dir->i_mode & S_ISGID) -- inode->i_mode |= S_ISGID; - ext3_mark_inode_dirty(handle, inode); - err = ext3_add_entry (handle, dentry, inode); - if (err) { -@@ -2020,7 +2017,7 @@ static int ext3_symlink (struct inode * - goto out_stop; - - if (l > sizeof (EXT3_I(inode)->i_data)) { -- inode->i_op = &page_symlink_inode_operations; -+ inode->i_op = &ext3_symlink_inode_operations; - inode->i_mapping->a_ops = &ext3_aops; - /* - * block_symlink() calls back into ext3_prepare/commit_write. -@@ -2245,4 +2242,16 @@ struct inode_operations ext3_dir_inode_o - rmdir: ext3_rmdir, /* BKL held */ - mknod: ext3_mknod, /* BKL held */ - rename: ext3_rename, /* BKL held */ -+ setxattr: ext3_setxattr, /* BKL held */ -+ getxattr: ext3_getxattr, /* BKL held */ -+ listxattr: ext3_listxattr, /* BKL held */ -+ removexattr: ext3_removexattr, /* BKL held */ - }; -+ -+struct inode_operations ext3_special_inode_operations = { -+ setxattr: ext3_setxattr, /* BKL held */ -+ getxattr: ext3_getxattr, /* BKL held */ -+ listxattr: ext3_listxattr, /* BKL held */ -+ removexattr: ext3_removexattr, /* BKL held */ -+}; -+ --- linux-2.4.22-ac1/fs/ext3/super.c~ext3-orphan_lock-2.4.22-rh 2003-09-26 00:24:09.000000000 +0400 +++ linux-2.4.22-ac1-alexey/fs/ext3/super.c 2003-09-26 00:25:22.000000000 +0400 @@ -1164,6 +1164,7 @@ struct super_block * ext3_read_super (st @@ -143,16 +80,3 @@ unsigned long s_commit_interval; struct block_device *journal_bdev; #ifdef CONFIG_JBD_DEBUG ---- linux-2.4.22-ac1/include/linux/ext3_fs_sb.h.orig~ext3-orphan_lock-2.4.22-rh 2003-09-25 14:16:34.000000000 +0400 -+++ linux-2.4.22-ac1-alexey/include/linux/ext3_fs_sb.h.orig 2003-09-26 00:24:08.000000000 +0400 -@@ -62,6 +62,8 @@ struct ext3_sb_info { - int s_inode_size; - int s_first_ino; - u32 s_next_generation; -+ u32 s_hash_seed[4]; -+ int s_def_hash_version; - - unsigned long s_dir_count; - u8 *s_debts; - -_ diff --git a/lustre/kernel_patches/patches/ext3-use-after-free-2.4.19-pre1.patch b/lustre/kernel_patches/patches/ext3-use-after-free-2.4.19-pre1.patch new file mode 100644 index 0000000..595db54 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-use-after-free-2.4.19-pre1.patch @@ -0,0 +1,53 @@ + ./fs/ext3/namei.c | 11 +++++------ + 1 files changed, 5 insertions(+), 6 deletions(-) + +Index: linux-2.4.19-pre1/./fs/ext3/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/./fs/ext3/namei.c 2003-11-21 01:52:06.000000000 +0300 ++++ linux-2.4.19-pre1/./fs/ext3/namei.c 2003-11-21 01:58:15.000000000 +0300 +@@ -1522,8 +1522,11 @@ + { + int err = ext3_add_entry(handle, dentry, inode); + if (!err) { +- d_instantiate(dentry, inode); +- return 0; ++ err = ext3_mark_inode_dirty(handle, inode); ++ if (err == 0) { ++ d_instantiate(dentry, inode); ++ return 0; ++ } + } + ext3_dec_count(handle, inode); + iput(inode); +@@ -1559,7 +1562,6 @@ + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -1586,7 +1588,6 @@ + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -2035,7 +2036,6 @@ + inode->i_size = l-1; + } + inode->u.ext3_i.i_disksize = inode->i_size; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); +@@ -2069,7 +2069,6 @@ + ext3_inc_count(handle, inode); + atomic_inc(&inode->i_count); + +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + ext3_journal_stop(handle, dir); + return err; diff --git a/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch b/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch new file mode 100644 index 0000000..1dab6d8 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch @@ -0,0 +1,53 @@ + ./fs/ext3/namei.c | 11 +++++------ + 1 files changed, 5 insertions(+), 6 deletions(-) + +Index: linux-2.4.19.SuSE/./fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/./fs/ext3/namei.c Sun Nov 16 01:02:51 2003 ++++ linux-2.4.19.SuSE/./fs/ext3/namei.c Sun Nov 16 01:14:50 2003 +@@ -1523,8 +1523,11 @@ + { + int err = ext3_add_entry(handle, dentry, inode); + if (!err) { +- d_instantiate(dentry, inode); +- return 0; ++ err = ext3_mark_inode_dirty(handle, inode); ++ if (err == 0) { ++ d_instantiate(dentry, inode); ++ return 0; ++ } + } + ext3_dec_count(handle, inode); + iput(inode); +@@ -1560,7 +1563,6 @@ + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -1590,7 +1592,6 @@ + #ifdef CONFIG_EXT3_FS_XATTR + inode->i_op = &ext3_special_inode_operations; + #endif +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -2039,7 +2040,6 @@ + inode->i_size = l-1; + } + EXT3_I(inode)->i_disksize = inode->i_size; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); +@@ -2073,7 +2073,6 @@ + ext3_inc_count(handle, inode); + atomic_inc(&inode->i_count); + +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + ext3_journal_stop(handle, dir); + return err; diff --git a/lustre/kernel_patches/patches/ext3-xattr-ptr-arith-fix.patch b/lustre/kernel_patches/patches/ext3-xattr-ptr-arith-fix.patch new file mode 100644 index 0000000..05fcf61 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-xattr-ptr-arith-fix.patch @@ -0,0 +1,18 @@ +Index: linux-2.4.20/fs/ext3/xattr.c +=================================================================== +--- linux-2.4.20.orig/fs/ext3/xattr.c 2003-11-13 17:14:52.000000000 +0300 ++++ linux-2.4.20/fs/ext3/xattr.c 2003-11-21 16:43:48.000000000 +0300 +@@ -1293,9 +1293,10 @@ + goto cleanup; + memcpy(header, HDR(bh), bh->b_size); + header->h_refcount = cpu_to_le32(1); +- offset = (char *)header - bh->b_data; +- here = ENTRY((char *)here + offset); +- last = ENTRY((char *)last + offset); ++ offset = (char *)here - bh->b_data; ++ here = ENTRY((char *)header + offset); ++ offset = (char *)last - bh->b_data; ++ last = ENTRY((char *)header + offset); + } + } else { + /* Allocate a buffer where we construct the new block. */ diff --git a/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch b/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch new file mode 100644 index 0000000..02cfef1 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch @@ -0,0 +1,226 @@ + fs/ext3/ialloc.c | 40 ++++++++++++++++++++++++++++++++++++++-- + fs/ext3/inode.c | 2 +- + fs/ext3/ioctl.c | 25 +++++++++++++++++++++++++ + fs/ext3/namei.c | 21 +++++++++++++++++---- + include/linux/dcache.h | 5 +++++ + include/linux/ext3_fs.h | 5 ++++- + 6 files changed, 90 insertions(+), 8 deletions(-) + +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:18:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:23:20 2003 +@@ -1534,6 +1534,19 @@ + return err; + } + ++static struct inode * ext3_new_inode_wantedi(handle_t *handle, struct inode *dir, ++ int mode, struct dentry *dentry) ++{ ++ unsigned long inum = 0; ++ ++ if (dentry->d_fsdata != NULL) { ++ struct dentry_params *param = ++ (struct dentry_params *) dentry->d_fsdata; ++ inum = param->p_inum; ++ } ++ return ext3_new_inode(handle, dir, mode, inum); ++} ++ + /* + * By the time this is called, we already have created + * the directory cache entry for the new file, but it +@@ -1557,7 +1570,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + inode->i_op = &ext3_file_inode_operations; +@@ -1585,7 +1598,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, inode->i_mode, rdev); +@@ -1618,7 +1631,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR | mode); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFDIR | mode, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -2013,7 +2026,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFLNK|S_IRWXUGO, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +Index: linux-2.4.19.SuSE/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ialloc.c Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/fs/ext3/ialloc.c Sun Nov 16 01:24:49 2003 +@@ -330,7 +330,8 @@ + * For other inodes, search forward from the parent directory's block + * group to find a free inode. + */ +-struct inode * ext3_new_inode (handle_t *handle, struct inode * dir, int mode) ++struct inode * ext3_new_inode(handle_t *handle, const struct inode * dir, ++ int mode, unsigned long goal) + { + struct super_block * sb; + struct buffer_head * bh; +@@ -355,7 +356,41 @@ + init_rwsem(&inode->u.ext3_i.truncate_sem); + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; ++ ++ if (goal) { ++ i = (goal - 1) / EXT3_INODES_PER_GROUP(sb); ++ j = (goal - 1) % EXT3_INODES_PER_GROUP(sb); ++ gdp = ext3_get_group_desc(sb, i, &bh2); ++ ++ bitmap_nr = load_inode_bitmap (sb, i); ++ if (bitmap_nr < 0) { ++ err = bitmap_nr; ++ goto fail; ++ } ++ ++ bh = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; ++ ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) goto fail; ++ ++ if (ext3_set_bit(j, bh->b_data)) { ++ printk(KERN_ERR "goal inode %lu unavailable\n", goal); ++ /* Oh well, we tried. */ ++ goto repeat; ++ } ++ ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) goto fail; ++ ++ /* We've shortcircuited the allocation system successfully, ++ * now finish filling in the inode. ++ */ ++ goto have_bit_and_group; ++ } ++ + repeat: + gdp = NULL; + i = 0; +@@ -470,6 +505,7 @@ + } + goto repeat; + } ++ have_bit_and_group: + j += i * EXT3_INODES_PER_GROUP(sb) + 1; + if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:23:20 2003 +@@ -2168,7 +2168,7 @@ + if (IS_ERR(handle)) + goto out_truncate; + +- new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode, 0); + if (IS_ERR(new_inode)) { + ext3_debug("truncate inode %lu directly (no new inodes)\n", + old_inode->i_ino); +Index: linux-2.4.19.SuSE/fs/ext3/ioctl.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ioctl.c Fri Nov 9 14:25:04 2001 ++++ linux-2.4.19.SuSE/fs/ext3/ioctl.c Sun Nov 16 01:23:20 2003 +@@ -23,6 +23,31 @@ + ext3_debug ("cmd = %u, arg = %lu\n", cmd, arg); + + switch (cmd) { ++ case EXT3_IOC_CREATE_INUM: { ++ char name[32]; ++ struct dentry *dchild, *dparent; ++ int rc = 0; ++ ++ dparent = list_entry(inode->i_dentry.next, struct dentry, ++ d_alias); ++ snprintf(name, sizeof name, "%lu", arg); ++ dchild = lookup_one_len(name, dparent, strlen(name)); ++ if (dchild->d_inode) { ++ printk(KERN_ERR "%*s/%lu already exists (ino %lu)\n", ++ dparent->d_name.len, dparent->d_name.name, arg, ++ dchild->d_inode->i_ino); ++ rc = -EEXIST; ++ } else { ++ dchild->d_fsdata = (void *)arg; ++ rc = vfs_create(inode, dchild, 0644); ++ if (rc) ++ printk(KERN_ERR "vfs_create: %d\n", rc); ++ else if (dchild->d_inode->i_ino != arg) ++ rc = -EEXIST; ++ } ++ dput(dchild); ++ return rc; ++ } + case EXT3_IOC_GETFLAGS: + flags = inode->u.ext3_i.i_flags & EXT3_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:25:42 2003 +@@ -202,6 +202,7 @@ + #define EXT3_IOC_SETFLAGS _IOW('f', 2, long) + #define EXT3_IOC_GETVERSION _IOR('f', 3, long) + #define EXT3_IOC_SETVERSION _IOW('f', 4, long) ++/* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ + #define EXT3_IOC_GETVERSION_OLD _IOR('v', 1, long) + #define EXT3_IOC_SETVERSION_OLD _IOW('v', 2, long) + #ifdef CONFIG_JBD_DEBUG +@@ -674,7 +675,8 @@ + dx_hash_info *hinfo); + + /* ialloc.c */ +-extern struct inode * ext3_new_inode (handle_t *, struct inode *, int); ++extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int, ++ unsigned long); + extern void ext3_free_inode (handle_t *, struct inode *); + extern struct inode * ext3_orphan_get (struct super_block *, unsigned long); + extern unsigned long ext3_count_free_inodes (struct super_block *); +@@ -765,4 +767,5 @@ + + #endif /* __KERNEL__ */ + ++#define EXT3_IOC_CREATE_INUM _IOW('f', 5, long) + #endif /* _LINUX_EXT3_FS_H */ +Index: linux-2.4.19.SuSE/include/linux/dcache.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/dcache.h Sat Nov 15 17:35:46 2003 ++++ linux-2.4.19.SuSE/include/linux/dcache.h Sun Nov 16 01:23:20 2003 +@@ -62,6 +62,11 @@ + + #define IS_ROOT(x) ((x) == (x)->d_parent) + ++struct dentry_params { ++ unsigned long p_inum; ++ void *p_ptr; ++}; ++ + /* + * "quick string" -- eases parameter passing, but more importantly + * saves "metadata" about the string (ie length and the hash). diff --git a/lustre/kernel_patches/patches/gfp_memalloc-2.4.18-chaos.patch b/lustre/kernel_patches/patches/gfp_memalloc-2.4.18-chaos.patch new file mode 100644 index 0000000..a8489e6 --- /dev/null +++ b/lustre/kernel_patches/patches/gfp_memalloc-2.4.18-chaos.patch @@ -0,0 +1,58 @@ +Index: linux-2.4.18-chaos/include/linux/mm.h +=================================================================== +--- linux-2.4.18-chaos.orig/include/linux/mm.h 2003-11-13 17:06:48.000000000 +0300 ++++ linux-2.4.18-chaos/include/linux/mm.h 2003-11-17 15:46:32.000000000 +0300 +@@ -677,6 +677,7 @@ + #define __GFP_IO 0x40 /* Can start low memory physical IO? */ + #define __GFP_HIGHIO 0x80 /* Can start high mem physical IO? */ + #define __GFP_FS 0x100 /* Can call down to low-level FS? */ ++#define __GFP_MEMALLOC 0x200 /* like PF_MEMALLOC: see __alloc_pages */ + + #define GFP_NOHIGHIO (__GFP_HIGH | __GFP_WAIT | __GFP_IO) + #define GFP_NOIO (__GFP_HIGH | __GFP_WAIT) +@@ -687,6 +688,7 @@ + #define GFP_KERNEL (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_NFS (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_KSWAPD ( __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) ++#define GFP_MEMALLOC __GFP_MEMALLOC + + /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some + platforms, used as appropriate on others */ +Index: linux-2.4.18-chaos/mm/page_alloc.c +=================================================================== +--- linux-2.4.18-chaos.orig/mm/page_alloc.c 2003-11-13 17:06:47.000000000 +0300 ++++ linux-2.4.18-chaos/mm/page_alloc.c 2003-11-17 15:49:11.000000000 +0300 +@@ -554,7 +554,7 @@ + /* + * Oh well, we didn't succeed. + */ +- if (!(current->flags & PF_MEMALLOC)) { ++ if (!(current->flags & PF_MEMALLOC) && !(gfp_mask & __GFP_MEMALLOC)) { + /* + * Are we dealing with a higher order allocation? + * +Index: linux-2.4.18-chaos/include/linux/slab.h +=================================================================== +--- linux-2.4.18-chaos.orig/include/linux/slab.h 2003-07-28 17:52:18.000000000 +0400 ++++ linux-2.4.18-chaos/include/linux/slab.h 2003-11-17 15:46:32.000000000 +0300 +@@ -23,6 +23,7 @@ + #define SLAB_KERNEL GFP_KERNEL + #define SLAB_NFS GFP_NFS + #define SLAB_DMA GFP_DMA ++#define SLAB_MEMALLOC GFP_MEMALLOC + + #define SLAB_LEVEL_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_HIGHIO|__GFP_FS) + #define SLAB_NO_GROW 0x00001000UL /* don't grow a cache */ +Index: linux-2.4.18-chaos/mm/slab.c +=================================================================== +--- linux-2.4.18-chaos.orig/mm/slab.c 2003-07-28 17:52:20.000000000 +0400 ++++ linux-2.4.18-chaos/mm/slab.c 2003-11-17 15:46:32.000000000 +0300 +@@ -1116,7 +1116,7 @@ + /* Be lazy and only check for valid flags here, + * keeping it out of the critical path in kmem_cache_alloc(). + */ +- if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW)) ++ if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW|SLAB_MEMALLOC)) + BUG(); + if (flags & SLAB_NO_GROW) + return 0; diff --git a/lustre/kernel_patches/patches/gfp_memalloc-2.4.20-rh.patch b/lustre/kernel_patches/patches/gfp_memalloc-2.4.20-rh.patch new file mode 100644 index 0000000..3f37e44 --- /dev/null +++ b/lustre/kernel_patches/patches/gfp_memalloc-2.4.20-rh.patch @@ -0,0 +1,59 @@ +Index: linux-2.4.20-rh-20.9/include/linux/mm.h +=================================================================== +--- linux-2.4.20-rh-20.9.orig/include/linux/mm.h 2003-11-13 17:35:48.000000000 +0300 ++++ linux-2.4.20-rh-20.9/include/linux/mm.h 2003-11-17 15:50:22.000000000 +0300 +@@ -713,6 +713,7 @@ + #define __GFP_IO 0x40 /* Can start low memory physical IO? */ + #define __GFP_HIGHIO 0x80 /* Can start high mem physical IO? */ + #define __GFP_FS 0x100 /* Can call down to low-level FS? */ ++#define __GFP_MEMALLOC 0x200 /* like PF_MEMALLOC: see __alloc_pages */ + + #define GFP_NOHIGHIO (__GFP_HIGH | __GFP_WAIT | __GFP_IO) + #define GFP_NOIO (__GFP_HIGH | __GFP_WAIT) +@@ -723,6 +724,7 @@ + #define GFP_KERNEL (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_NFS (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_KSWAPD ( __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) ++#define GFP_MEMALLOC __GFP_MEMALLOC + + /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some + platforms, used as appropriate on others */ +Index: linux-2.4.20-rh-20.9/mm/page_alloc.c +=================================================================== +--- linux-2.4.20-rh-20.9.orig/mm/page_alloc.c 2003-11-13 17:20:37.000000000 +0300 ++++ linux-2.4.20-rh-20.9/mm/page_alloc.c 2003-11-17 15:51:05.000000000 +0300 +@@ -509,7 +509,8 @@ + /* + * Oh well, we didn't succeed. + */ +- if (!(current->flags & (PF_MEMALLOC|PF_MEMDIE))) { ++ if (!(current->flags & (PF_MEMALLOC|PF_MEMDIE)) && ++ !(gfp_mask & __GFP_MEMALLOC)) { + /* + * Are we dealing with a higher order allocation? + * +Index: linux-2.4.20-rh-20.9/include/linux/slab.h +=================================================================== +--- linux-2.4.20-rh-20.9.orig/include/linux/slab.h 2003-11-13 17:35:48.000000000 +0300 ++++ linux-2.4.20-rh-20.9/include/linux/slab.h 2003-11-17 15:50:22.000000000 +0300 +@@ -23,6 +23,7 @@ + #define SLAB_KERNEL GFP_KERNEL + #define SLAB_NFS GFP_NFS + #define SLAB_DMA GFP_DMA ++#define SLAB_MEMALLOC GFP_MEMALLOC + + #define SLAB_LEVEL_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_HIGHIO|__GFP_FS) + #define SLAB_NO_GROW 0x00001000UL /* don't grow a cache */ +Index: linux-2.4.20-rh-20.9/mm/slab.c +=================================================================== +--- linux-2.4.20-rh-20.9.orig/mm/slab.c 2003-09-13 19:34:24.000000000 +0400 ++++ linux-2.4.20-rh-20.9/mm/slab.c 2003-11-17 15:50:22.000000000 +0300 +@@ -1116,7 +1116,7 @@ + /* Be lazy and only check for valid flags here, + * keeping it out of the critical path in kmem_cache_alloc(). + */ +- if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW)) ++ if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW|SLAB_MEMALLOC)) + BUG(); + if (flags & SLAB_NO_GROW) + return 0; diff --git a/lustre/kernel_patches/patches/gfp_memalloc-2.4.22.patch b/lustre/kernel_patches/patches/gfp_memalloc-2.4.22.patch new file mode 100644 index 0000000..015bfc8 --- /dev/null +++ b/lustre/kernel_patches/patches/gfp_memalloc-2.4.22.patch @@ -0,0 +1,59 @@ +Index: linux-2.4.22-vanilla/include/linux/mm.h +=================================================================== +--- linux-2.4.22-vanilla.orig/include/linux/mm.h 2003-11-17 15:26:32.000000000 +0300 ++++ linux-2.4.22-vanilla/include/linux/mm.h 2003-11-17 15:40:32.000000000 +0300 +@@ -612,6 +612,7 @@ + #define __GFP_IO 0x40 /* Can start low memory physical IO? */ + #define __GFP_HIGHIO 0x80 /* Can start high mem physical IO? */ + #define __GFP_FS 0x100 /* Can call down to low-level FS? */ ++#define __GFP_MEMALLOC 0x200 /* like PF_MEMALLOC: see __alloc_pages */ + + #define GFP_NOHIGHIO (__GFP_HIGH | __GFP_WAIT | __GFP_IO) + #define GFP_NOIO (__GFP_HIGH | __GFP_WAIT) +@@ -622,6 +623,7 @@ + #define GFP_KERNEL (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_NFS (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_KSWAPD ( __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) ++#define GFP_MEMALLOC __GFP_MEMALLOC + + /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some + platforms, used as appropriate on others */ +Index: linux-2.4.22-vanilla/mm/page_alloc.c +=================================================================== +--- linux-2.4.22-vanilla.orig/mm/page_alloc.c 2003-11-13 18:19:51.000000000 +0300 ++++ linux-2.4.22-vanilla/mm/page_alloc.c 2003-11-17 15:40:32.000000000 +0300 +@@ -377,7 +377,8 @@ + /* here we're in the low on memory slow path */ + + rebalance: +- if (current->flags & (PF_MEMALLOC | PF_MEMDIE)) { ++ if (current->flags & (PF_MEMALLOC | PF_MEMDIE) || ++ gfp_mask & __GFP_MEMALLOC) { + zone = zonelist->zones; + for (;;) { + zone_t *z = *(zone++); +Index: linux-2.4.22-vanilla/include/linux/slab.h +=================================================================== +--- linux-2.4.22-vanilla.orig/include/linux/slab.h 2003-11-17 14:58:37.000000000 +0300 ++++ linux-2.4.22-vanilla/include/linux/slab.h 2003-11-17 15:42:13.000000000 +0300 +@@ -23,6 +23,7 @@ + #define SLAB_KERNEL GFP_KERNEL + #define SLAB_NFS GFP_NFS + #define SLAB_DMA GFP_DMA ++#define SLAB_MEMALLOC GFP_MEMALLOC + + #define SLAB_LEVEL_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_HIGHIO|__GFP_FS) + #define SLAB_NO_GROW 0x00001000UL /* don't grow a cache */ +Index: linux-2.4.22-vanilla/mm/slab.c +=================================================================== +--- linux-2.4.22-vanilla.orig/mm/slab.c 2003-11-13 17:39:29.000000000 +0300 ++++ linux-2.4.22-vanilla/mm/slab.c 2003-11-17 15:42:13.000000000 +0300 +@@ -1115,7 +1115,7 @@ + /* Be lazy and only check for valid flags here, + * keeping it out of the critical path in kmem_cache_alloc(). + */ +- if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW)) ++ if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW|SLAB_MEMALLOC)) + BUG(); + if (flags & SLAB_NO_GROW) + return 0; diff --git a/lustre/kernel_patches/patches/invalidate_show-2.4.19-pre1.patch b/lustre/kernel_patches/patches/invalidate_show-2.4.19-pre1.patch new file mode 100644 index 0000000..376558b --- /dev/null +++ b/lustre/kernel_patches/patches/invalidate_show-2.4.19-pre1.patch @@ -0,0 +1,121 @@ + + + + fs/inode.c | 21 ++++++++++++++------- + fs/smbfs/inode.c | 2 +- + fs/super.c | 4 ++-- + include/linux/fs.h | 2 +- + 4 files changed, 18 insertions(+), 11 deletions(-) + +Index: linux-2.4.19-pre1/fs/inode.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/inode.c 2003-11-21 02:39:48.000000000 +0300 ++++ linux-2.4.19-pre1/fs/inode.c 2003-11-21 02:39:54.000000000 +0300 +@@ -553,7 +553,8 @@ + /* + * Invalidate all inodes for a device. + */ +-static int invalidate_list(struct list_head *head, struct super_block * sb, struct list_head * dispose) ++static int invalidate_list(struct list_head *head, struct super_block * sb, ++ struct list_head * dispose, int show) + { + struct list_head *next; + int busy = 0, count = 0; +@@ -578,6 +579,11 @@ + count++; + continue; + } ++ if (show) ++ printk(KERN_ERR ++ "inode busy: dev %s:%lu (%p) mode %o count %u\n", ++ kdevname(sb->s_dev), inode->i_ino, inode, ++ inode->i_mode, atomic_read(&inode->i_count)); + busy = 1; + } + /* only unused inodes may be cached with i_count zero */ +@@ -596,22 +602,23 @@ + /** + * invalidate_inodes - discard the inodes on a device + * @sb: superblock ++ * @show: whether we should display any busy inodes found + * + * Discard all of the inodes for a given superblock. If the discard + * fails because there are busy inodes then a non zero value is returned. + * If the discard is successful all the inodes have been discarded. + */ + +-int invalidate_inodes(struct super_block * sb) ++int invalidate_inodes(struct super_block * sb, int show) + { + int busy; + LIST_HEAD(throw_away); + + spin_lock(&inode_lock); +- busy = invalidate_list(&inode_in_use, sb, &throw_away); +- busy |= invalidate_list(&inode_unused, sb, &throw_away); +- busy |= invalidate_list(&sb->s_dirty, sb, &throw_away); +- busy |= invalidate_list(&sb->s_locked_inodes, sb, &throw_away); ++ busy = invalidate_list(&inode_in_use, sb, &throw_away, show); ++ busy |= invalidate_list(&inode_unused, sb, &throw_away, show); ++ busy |= invalidate_list(&sb->s_dirty, sb, &throw_away, show); ++ busy |= invalidate_list(&sb->s_locked_inodes, sb, &throw_away, show); + spin_unlock(&inode_lock); + + dispose_list(&throw_away); +@@ -637,7 +644,7 @@ + * hold). + */ + shrink_dcache_sb(sb); +- res = invalidate_inodes(sb); ++ res = invalidate_inodes(sb, 0); + drop_super(sb); + } + invalidate_buffers(dev); +Index: linux-2.4.19-pre1/fs/super.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/super.c 2003-11-21 02:39:48.000000000 +0300 ++++ linux-2.4.19-pre1/fs/super.c 2003-11-21 02:40:09.000000000 +0300 +@@ -790,7 +790,7 @@ + fsync_super(sb); + lock_super(sb); + sb->s_flags &= ~MS_ACTIVE; +- invalidate_inodes(sb); /* bad name - it should be evict_inodes() */ ++ invalidate_inodes(sb, 0); /* bad name - it should be evict_inodes() */ + if (sop) { + if (sop->write_super && sb->s_dirt) + sop->write_super(sb); +@@ -799,7 +799,7 @@ + } + + /* Forget any remaining inodes */ +- if (invalidate_inodes(sb)) { ++ if (invalidate_inodes(sb, 1)) { + printk("VFS: Busy inodes after unmount. " + "Self-destruct in 5 seconds. Have a nice day...\n"); + } +Index: linux-2.4.19-pre1/include/linux/fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs.h 2003-11-21 02:39:53.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs.h 2003-11-21 02:39:54.000000000 +0300 +@@ -1214,7 +1214,7 @@ + extern void set_buffer_flushtime(struct buffer_head *); + extern void balance_dirty(void); + extern int check_disk_change(kdev_t); +-extern int invalidate_inodes(struct super_block *); ++extern int invalidate_inodes(struct super_block *, int); + extern int invalidate_device(kdev_t, int); + extern void invalidate_inode_pages(struct inode *); + extern void invalidate_inode_pages2(struct address_space *); +Index: linux-2.4.19-pre1/fs/smbfs/inode.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/smbfs/inode.c 2003-11-21 02:39:48.000000000 +0300 ++++ linux-2.4.19-pre1/fs/smbfs/inode.c 2003-11-21 02:39:54.000000000 +0300 +@@ -166,7 +166,7 @@ + { + VERBOSE("\n"); + shrink_dcache_sb(SB_of(server)); +- invalidate_inodes(SB_of(server)); ++ invalidate_inodes(SB_of(server), 0); + } + + /* diff --git a/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch b/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch new file mode 100644 index 0000000..2040fcd --- /dev/null +++ b/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch @@ -0,0 +1,52 @@ + fs/Makefile | 2 +- + fs/inode.c | 4 +++- + mm/page_alloc.c | 1 + + 3 files changed, 5 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/fs/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/inode.c Sat Nov 15 18:02:13 2003 ++++ linux-2.4.19.SuSE/fs/inode.c Sat Nov 15 18:03:04 2003 +@@ -5,6 +5,7 @@ + */ + + #include ++#include + #include + #include + #include +@@ -67,7 +68,8 @@ + * NOTE! You also have to own the lock if you change + * the i_state of an inode while it is in use.. + */ +-static spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++EXPORT_SYMBOL(inode_lock); + + /* + * Statistics gathering.. +Index: linux-2.4.19.SuSE/fs/Makefile +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/Makefile Mon Jan 27 05:08:56 2003 ++++ linux-2.4.19.SuSE/fs/Makefile Sat Nov 15 18:03:54 2003 +@@ -7,7 +7,7 @@ + + O_TARGET := fs.o + +-export-objs := filesystems.o open.o dcache.o buffer.o ++export-objs := filesystems.o open.o dcache.o buffer.o inode.o + mod-subdirs := nls + + obj-y := open.o read_write.o devices.o file_table.o buffer.o \ +Index: linux-2.4.19.SuSE/mm/page_alloc.c +=================================================================== +--- linux-2.4.19.SuSE.orig/mm/page_alloc.c Mon Jan 27 05:08:55 2003 ++++ linux-2.4.19.SuSE/mm/page_alloc.c Sat Nov 15 18:03:04 2003 +@@ -32,6 +32,7 @@ + LIST_HEAD(inactive_list); + LIST_HEAD(active_list); + pg_data_t *pgdat_list; ++EXPORT_SYMBOL(pgdat_list); + + /* Used to look up the address of the struct zone encoded in page->zone */ + zone_t *zone_table[MAX_NR_ZONES*MAX_NR_NODES]; diff --git a/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch b/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch new file mode 100644 index 0000000..ab267bb --- /dev/null +++ b/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch @@ -0,0 +1,437 @@ + Documentation/filesystems/ext2.txt | 16 ++ + fs/ext3/Makefile | 2 + fs/ext3/inode.c | 4 + fs/ext3/iopen.c | 259 +++++++++++++++++++++++++++++++++++++ + fs/ext3/iopen.h | 13 + + fs/ext3/namei.c | 13 + + fs/ext3/super.c | 11 + + include/linux/ext3_fs.h | 2 + 8 files changed, 318 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/Documentation/filesystems/ext2.txt +=================================================================== +--- linux-2.4.19.SuSE.orig/Documentation/filesystems/ext2.txt Wed Jul 11 15:44:45 2001 ++++ linux-2.4.19.SuSE/Documentation/filesystems/ext2.txt Sun Nov 16 01:27:31 2003 +@@ -35,6 +35,22 @@ + + sb=n Use alternate superblock at this location. + ++iopen Makes an invisible pseudo-directory called ++ __iopen__ available in the root directory ++ of the filesystem. Allows open-by-inode- ++ number. i.e., inode 3145 can be accessed ++ via /mntpt/__iopen__/3145 ++ ++iopen_nopriv This option makes the iopen directory be ++ world-readable. This may be safer since it ++ allows daemons to run as an unprivileged user, ++ however it significantly changes the security ++ model of a Unix filesystem, since previously ++ all files under a mode 700 directory were not ++ generally avilable even if the ++ permissions on the file itself is ++ world-readable. ++ + grpquota,noquota,quota,usrquota Quota options are silently ignored by ext2. + + +Index: linux-2.4.19.SuSE/fs/ext3/Makefile +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/Makefile Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/Makefile Sun Nov 16 01:27:31 2003 +@@ -11,7 +11,7 @@ + + export-objs := ext3-exports.o + +-obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ ++obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ + ioctl.o namei.o super.o symlink.o hash.o ext3-exports.o + obj-m := $(O_TARGET) + +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:26:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:27:31 2003 +@@ -34,6 +34,7 @@ + #include + #include + #include ++#include "iopen.h" + + /* + * SEARCH_FROM_ZERO forces each block allocation to search from the start +@@ -2350,6 +2351,9 @@ + struct buffer_head *bh; + int block; + ++ if (ext3_iopen_get_inode(inode)) ++ return; ++ + if(ext3_get_inode_loc(inode, &iloc)) + goto bad_inode; + bh = iloc.bh; +Index: linux-2.4.19.SuSE/fs/ext3/iopen.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/iopen.c Sun Nov 16 01:27:31 2003 ++++ linux-2.4.19.SuSE/fs/ext3/iopen.c Sun Nov 16 01:27:31 2003 +@@ -0,0 +1,259 @@ ++/* ++ * linux/fs/ext3/iopen.c ++ * ++ * Special support for open by inode number ++ * ++ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). ++ * ++ * This file may be redistributed under the terms of the GNU General ++ * Public License. ++ * ++ * ++ * Invariants: ++ * - there is only ever a single DCACHE_NFSD_DISCONNECTED dentry alias ++ * for an inode at one time. ++ * - there are never both connected and DCACHE_NFSD_DISCONNECTED dentry ++ * aliases on an inode at the same time. ++ * ++ * If we have any connected dentry aliases for an inode, use one of those ++ * in iopen_lookup(). Otherwise, we instantiate a single NFSD_DISCONNECTED ++ * dentry for this inode, which thereafter will be found by the dcache ++ * when looking up this inode number in __iopen__, so we don't return here ++ * until it is gone. ++ * ++ * If we get an inode via a regular name lookup, then we "rename" the ++ * NFSD_DISCONNECTED dentry to the proper name and parent. This ensures ++ * existing users of the disconnected dentry will continue to use the same ++ * dentry as the connected users, and there will never be both kinds of ++ * dentry aliases at one time. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include "iopen.h" ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#define IOPEN_NAME_LEN 32 ++ ++/* ++ * This implements looking up an inode by number. ++ */ ++static struct dentry *iopen_lookup(struct inode *dir, struct dentry *dentry) ++{ ++ struct inode *inode; ++ unsigned long ino; ++ struct list_head *lp; ++ struct dentry *alternate; ++ char buf[IOPEN_NAME_LEN]; ++ ++ if (dentry->d_name.len >= IOPEN_NAME_LEN) ++ return ERR_PTR(-ENAMETOOLONG); ++ ++ memcpy(buf, dentry->d_name.name, dentry->d_name.len); ++ buf[dentry->d_name.len] = 0; ++ ++ if (strcmp(buf, ".") == 0) ++ ino = dir->i_ino; ++ else if (strcmp(buf, "..") == 0) ++ ino = EXT3_ROOT_INO; ++ else ++ ino = simple_strtoul(buf, 0, 0); ++ ++ if ((ino != EXT3_ROOT_INO && ++ //ino != EXT3_ACL_IDX_INO && ++ //ino != EXT3_ACL_DATA_INO && ++ ino < EXT3_FIRST_INO(dir->i_sb)) || ++ ino > le32_to_cpu(dir->i_sb->u.ext3_sb.s_es->s_inodes_count)) ++ return ERR_PTR(-ENOENT); ++ ++ inode = iget(dir->i_sb, ino); ++ if (!inode) ++ return ERR_PTR(-EACCES); ++ if (is_bad_inode(inode)) { ++ iput(inode); ++ return ERR_PTR(-ENOENT); ++ } ++ ++ /* preferrably return a connected dentry */ ++ spin_lock(&dcache_lock); ++ list_for_each(lp, &inode->i_dentry) { ++ alternate = list_entry(lp, struct dentry, d_alias); ++ assert(!(alternate->d_flags & DCACHE_NFSD_DISCONNECTED)); ++ } ++ ++ if (!list_empty(&inode->i_dentry)) { ++ alternate = list_entry(inode->i_dentry.next, ++ struct dentry, d_alias); ++ dget_locked(alternate); ++ alternate->d_vfs_flags |= DCACHE_REFERENCED; ++ iput(inode); ++ spin_unlock(&dcache_lock); ++ return alternate; ++ } ++ dentry->d_flags |= DCACHE_NFSD_DISCONNECTED; ++ spin_unlock(&dcache_lock); ++ ++ d_add(dentry, inode); ++ return NULL; ++} ++ ++#define do_switch(x,y) do { \ ++ __typeof__ (x) __tmp = x; \ ++ x = y; y = __tmp; } while (0) ++ ++static inline void switch_names(struct dentry *dentry, struct dentry *target) ++{ ++ const unsigned char *old_name, *new_name; ++ ++ memcpy(dentry->d_iname, target->d_iname, DNAME_INLINE_LEN); ++ old_name = target->d_name.name; ++ new_name = dentry->d_name.name; ++ if (old_name == target->d_iname) ++ old_name = dentry->d_iname; ++ if (new_name == dentry->d_iname) ++ new_name = target->d_iname; ++ target->d_name.name = new_name; ++ dentry->d_name.name = old_name; ++} ++ ++/* This function is spliced into ext3_lookup and does the move of a ++ * disconnected dentry (if it exists) to a connected dentry. ++ */ ++struct dentry *iopen_connect_dentry(struct dentry *de, struct inode *inode) ++{ ++ struct dentry *tmp, *goal = NULL; ++ struct list_head *lp; ++ ++ /* preferrably return a connected dentry */ ++ spin_lock(&dcache_lock); ++ /* verify this dentry is really new */ ++ assert(!de->d_inode); ++ assert(list_empty(&de->d_subdirs)); ++ assert(list_empty(&de->d_alias)); ++ ++ ++ list_for_each(lp, &inode->i_dentry) { ++ tmp = list_entry(lp, struct dentry, d_alias); ++ if (tmp->d_flags & DCACHE_NFSD_DISCONNECTED) { ++ assert(tmp->d_alias.next == &inode->i_dentry); ++ assert(tmp->d_alias.prev == &inode->i_dentry); ++ goal = tmp; ++ dget_locked(goal); ++ break; ++ } ++ } ++ ++ if (!goal) { ++ spin_unlock(&dcache_lock); ++ return NULL; ++ } ++ ++ /* Move the goal to the de hash queue - like d_move() */ ++ goal->d_flags &= ~DCACHE_NFSD_DISCONNECTED; ++ list_del_init(&goal->d_hash); ++ ++ list_del(&goal->d_child); ++ list_del(&de->d_child); ++ ++ /* Switch the parents and the names.. */ ++ switch_names(goal, de); ++ do_switch(goal->d_parent, de->d_parent); ++ do_switch(goal->d_name.len, de->d_name.len); ++ do_switch(goal->d_name.hash, de->d_name.hash); ++ ++ /* And add them back to the (new) parent lists */ ++ list_add(&goal->d_child, &goal->d_parent->d_subdirs); ++ list_add(&de->d_child, &de->d_parent->d_subdirs); ++ __d_rehash(goal, 0); ++ spin_unlock(&dcache_lock); ++ ++ return goal; ++} ++ ++/* ++ * These are the special structures for the iopen pseudo directory. ++ */ ++ ++static struct inode_operations iopen_inode_operations = { ++ lookup: iopen_lookup, /* BKL held */ ++}; ++ ++static struct file_operations iopen_file_operations = { ++ read: generic_read_dir, ++}; ++ ++static int match_dentry(struct dentry *dentry, const char *name) ++{ ++ int len; ++ ++ len = strlen(name); ++ if (dentry->d_name.len != len) ++ return 0; ++ if (strncmp(dentry->d_name.name, name, len)) ++ return 0; ++ return 1; ++} ++ ++/* ++ * This function is spliced into ext3_lookup and returns 1 the file ++ * name is __iopen__ and dentry has been filled in appropriately. ++ */ ++int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry) ++{ ++ struct inode *inode; ++ ++ if (dir->i_ino != EXT3_ROOT_INO || ++ !test_opt(dir->i_sb, IOPEN) || ++ !match_dentry(dentry, "__iopen__")) ++ return 0; ++ ++ inode = iget(dir->i_sb, EXT3_BAD_INO); ++ ++ if (!inode) ++ return 0; ++ d_add(dentry, inode); ++ return 1; ++} ++ ++/* ++ * This function is spliced into read_inode; it returns 1 if inode ++ * number is the one for /__iopen__, in which case the inode is filled ++ * in appropriately. Otherwise, this fuction returns 0. ++ */ ++int ext3_iopen_get_inode(struct inode *inode) ++{ ++ if (inode->i_ino != EXT3_BAD_INO) ++ return 0; ++ ++ inode->i_mode = S_IFDIR | S_IRUSR | S_IXUSR; ++ if (test_opt(inode->i_sb, IOPEN_NOPRIV)) ++ inode->i_mode |= 0777; ++ inode->i_uid = 0; ++ inode->i_gid = 0; ++ inode->i_nlink = 1; ++ inode->i_size = 4096; ++ inode->i_atime = CURRENT_TIME; ++ inode->i_ctime = CURRENT_TIME; ++ inode->i_mtime = CURRENT_TIME; ++ inode->u.ext3_i.i_dtime = 0; ++ inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size ++ * (for stat), not the fs block ++ * size */ ++ inode->i_blocks = 0; ++ inode->i_version = 1; ++ inode->i_generation = 0; ++ ++ inode->i_op = &iopen_inode_operations; ++ inode->i_fop = &iopen_file_operations; ++ inode->i_mapping->a_ops = 0; ++ ++ return 1; ++} +Index: linux-2.4.19.SuSE/fs/ext3/iopen.h +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/iopen.h Sun Nov 16 01:27:31 2003 ++++ linux-2.4.19.SuSE/fs/ext3/iopen.h Sun Nov 16 01:27:31 2003 +@@ -0,0 +1,13 @@ ++/* ++ * iopen.h ++ * ++ * Special support for opening files by inode number. ++ * ++ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). ++ * ++ * This file may be redistributed under the terms of the GNU General ++ * Public License. ++ */ ++ ++extern int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry); ++extern int ext3_iopen_get_inode(struct inode *inode); +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:23:20 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:27:31 2003 +@@ -36,7 +36,7 @@ + #include + #include + #include +- ++#include "iopen.h" + + /* + * define how far ahead to read directories while searching them. +@@ -922,16 +922,21 @@ + return NULL; + } + #endif ++struct dentry *iopen_connect_dentry(struct dentry *de, struct inode *inode); + + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; + struct ext3_dir_entry_2 * de; + struct buffer_head * bh; ++ struct dentry *alternate = NULL; + + if (dentry->d_name.len > EXT3_NAME_LEN) + return ERR_PTR(-ENAMETOOLONG); + ++ if (ext3_check_for_iopen(dir, dentry)) ++ return NULL; ++ + bh = ext3_find_entry(dentry, &de); + inode = NULL; + if (bh) { +@@ -943,6 +948,12 @@ + return ERR_PTR(-EACCES); + } + } ++ ++ if (inode && (alternate = iopen_connect_dentry(dentry, inode))) { ++ iput(inode); ++ return alternate; ++ } ++ + d_add(dentry, inode); + return NULL; + } +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 01:19:22 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:27:31 2003 +@@ -864,6 +864,17 @@ + || !strcmp (this_char, "quota") + || !strcmp (this_char, "usrquota")) + /* Don't do anything ;-) */ ; ++ else if (!strcmp (this_char, "iopen")) { ++ set_opt (sbi->s_mount_opt, IOPEN); ++ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } else if (!strcmp (this_char, "noiopen")) { ++ clear_opt (sbi->s_mount_opt, IOPEN); ++ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } ++ else if (!strcmp (this_char, "iopen_nopriv")) { ++ set_opt (sbi->s_mount_opt, IOPEN); ++ set_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } + else if (!strcmp (this_char, "journal")) { + /* @@@ FIXME */ + /* Eventually we will want to be able to create +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:25:42 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:30:05 2003 +@@ -324,6 +324,8 @@ + #define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + #define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */ + #define EXT3_MOUNT_ASYNCDEL 0x20000 /* Delayed deletion */ ++#define EXT3_MOUNT_IOPEN 0x40000 /* Allow access via iopen */ ++#define EXT3_MOUNT_IOPEN_NOPRIV 0x80000 /* Make iopen world-readable */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H diff --git a/lustre/kernel_patches/patches/jbd-2.4.18-jcberr.patch b/lustre/kernel_patches/patches/jbd-2.4.18-jcberr.patch new file mode 100644 index 0000000..81b4136 --- /dev/null +++ b/lustre/kernel_patches/patches/jbd-2.4.18-jcberr.patch @@ -0,0 +1,274 @@ +Index: linux-2.4.19.SuSE/include/linux/jbd.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/jbd.h Sun Nov 16 13:51:03 2003 ++++ linux-2.4.19.SuSE/include/linux/jbd.h Sun Nov 16 15:10:48 2003 +@@ -283,6 +283,13 @@ + return bh->b_private; + } + ++#define HAVE_JOURNAL_CALLBACK_STATUS ++struct journal_callback { ++ struct list_head jcb_list; ++ void (*jcb_func)(struct journal_callback *jcb, int error); ++ /* user data goes here */ ++}; ++ + struct jbd_revoke_table_s; + + /* The handle_t type represents a single atomic update being performed +@@ -313,6 +320,12 @@ + operations */ + int h_err; + ++ /* List of application registered callbacks for this handle. ++ * The function(s) will be called after the transaction that ++ * this handle is part of has been committed to disk. ++ */ ++ struct list_head h_jcb; ++ + /* Flags */ + unsigned int h_sync: 1; /* sync-on-close */ + unsigned int h_jdata: 1; /* force data journaling */ +@@ -432,6 +445,10 @@ + + /* How many handles used this transaction? */ + int t_handle_count; ++ ++ /* List of registered callback functions for this transaction. ++ * Called when the transaction is committed. */ ++ struct list_head t_jcb; + }; + + +@@ -676,6 +693,9 @@ + extern int journal_try_to_free_buffers(journal_t *, struct page *, int); + extern int journal_stop(handle_t *); + extern int journal_flush (journal_t *); ++extern void journal_callback_set(handle_t *handle, ++ void (*fn)(struct journal_callback *,int), ++ struct journal_callback *jcb); + + extern void journal_lock_updates (journal_t *); + extern void journal_unlock_updates (journal_t *); +Index: linux-2.4.19.SuSE/fs/jbd/checkpoint.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/checkpoint.c Mon Feb 25 11:38:08 2002 ++++ linux-2.4.19.SuSE/fs/jbd/checkpoint.c Sun Nov 16 15:10:48 2003 +@@ -594,7 +594,8 @@ + J_ASSERT (transaction->t_log_list == NULL); + J_ASSERT (transaction->t_checkpoint_list == NULL); + J_ASSERT (transaction->t_updates == 0); +- ++ J_ASSERT (list_empty(&transaction->t_jcb)); ++ + J_ASSERT (transaction->t_journal->j_committing_transaction != + transaction); + +Index: linux-2.4.19.SuSE/fs/jbd/commit.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/commit.c Mon Jan 27 05:08:04 2003 ++++ linux-2.4.19.SuSE/fs/jbd/commit.c Sun Nov 16 15:13:53 2003 +@@ -485,7 +485,7 @@ + transaction's t_log_list queue, and metadata buffers are on + the t_iobuf_list queue. + +- Wait for the transactions in reverse order. That way we are ++ Wait for the buffers in reverse order. That way we are + less likely to be woken up until all IOs have completed, and + so we incur less scheduling load. + */ +@@ -576,8 +576,10 @@ + + jbd_debug(3, "JBD: commit phase 6\n"); + +- if (is_journal_aborted(journal)) ++ if (is_journal_aborted(journal)) { ++ unlock_journal(journal); + goto skip_commit; ++ } + + /* Done it all: now write the commit record. We should have + * cleaned up our previous buffers by now, so if we are in abort +@@ -587,9 +589,10 @@ + descriptor = journal_get_descriptor_buffer(journal); + if (!descriptor) { + __journal_abort_hard(journal); ++ unlock_journal(journal); + goto skip_commit; + } +- ++ + /* AKPM: buglet - add `i' to tmp! */ + for (i = 0; i < jh2bh(descriptor)->b_size; i += 512) { + journal_header_t *tmp = +@@ -610,14 +614,32 @@ + put_bh(bh); /* One for getblk() */ + journal_unlock_journal_head(descriptor); + } +- lock_journal(journal); + + /* End of a transaction! Finally, we can do checkpoint + processing: any buffers committed as a result of this + transaction can be removed from any checkpoint list it was on + before. */ + +-skip_commit: ++skip_commit: /* The journal should be unlocked by now. */ ++ ++ /* Call any callbacks that had been registered for handles in this ++ * transaction. It is up to the callback to free any allocated ++ * memory. ++ */ ++ if (!list_empty(&commit_transaction->t_jcb)) { ++ struct list_head *p, *n; ++ int error = is_journal_aborted(journal); ++ ++ list_for_each_safe(p, n, &commit_transaction->t_jcb) { ++ struct journal_callback *jcb; ++ ++ jcb = list_entry(p, struct journal_callback, jcb_list); ++ list_del(p); ++ jcb->jcb_func(jcb, error); ++ } ++ } ++ ++ lock_journal(journal); + + jbd_debug(3, "JBD: commit phase 7\n"); + +Index: linux-2.4.19.SuSE/fs/jbd/journal.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/journal.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/jbd/journal.c Sun Nov 16 15:10:48 2003 +@@ -59,6 +59,7 @@ + #endif + EXPORT_SYMBOL(journal_flush); + EXPORT_SYMBOL(journal_revoke); ++EXPORT_SYMBOL(journal_callback_set); + + EXPORT_SYMBOL(journal_init_dev); + EXPORT_SYMBOL(journal_init_inode); +Index: linux-2.4.19.SuSE/fs/jbd/transaction.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/transaction.c Sun Nov 16 01:45:26 2003 ++++ linux-2.4.19.SuSE/fs/jbd/transaction.c Sun Nov 16 15:15:34 2003 +@@ -58,6 +58,7 @@ + transaction->t_state = T_RUNNING; + transaction->t_tid = journal->j_transaction_sequence++; + transaction->t_expires = jiffies + bdflush_interval(); ++ INIT_LIST_HEAD(&transaction->t_jcb); + + /* Set up the commit timer for the new transaction. */ + J_ASSERT (!journal->j_commit_timer_active); +@@ -91,7 +92,14 @@ + transaction_t *transaction; + int needed; + int nblocks = handle->h_buffer_credits; +- ++ ++ if (nblocks > journal->j_max_transaction_buffers) { ++ jbd_debug(1, "JBD: %s wants too many credits (%d > %d)\n", ++ current->comm, nblocks, ++ journal->j_max_transaction_buffers); ++ return -ENOSPC; ++ } ++ + jbd_debug(3, "New handle %p going live.\n", handle); + + repeat: +@@ -202,6 +210,20 @@ + return 0; + } + ++/* Allocate a new handle. This should probably be in a slab... */ ++static handle_t *new_handle(int nblocks) ++{ ++ handle_t *handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ if (!handle) ++ return NULL; ++ memset(handle, 0, sizeof (handle_t)); ++ handle->h_buffer_credits = nblocks; ++ handle->h_ref = 1; ++ INIT_LIST_HEAD(&handle->h_jcb); ++ ++ return handle; ++} ++ + /* + * Obtain a new handle. + * +@@ -228,14 +250,11 @@ + handle->h_ref++; + return handle; + } +- +- handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ ++ handle = new_handle(nblocks); + if (!handle) + return ERR_PTR(-ENOMEM); +- memset (handle, 0, sizeof (handle_t)); + +- handle->h_buffer_credits = nblocks; +- handle->h_ref = 1; + current->journal_info = handle; + + err = start_this_handle(journal, handle); +@@ -334,14 +353,11 @@ + + if (is_journal_aborted(journal)) + return ERR_PTR(-EIO); +- +- handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ ++ handle = new_handle(nblocks); + if (!handle) + return ERR_PTR(-ENOMEM); +- memset (handle, 0, sizeof (handle_t)); + +- handle->h_buffer_credits = nblocks; +- handle->h_ref = 1; + current->journal_info = handle; + + err = try_start_this_handle(journal, handle); +@@ -1321,6 +1337,28 @@ + #endif + + /* ++ * Register a callback function for this handle. The function will be ++ * called when the transaction that this handle is part of has been ++ * committed to disk with the original callback data struct and the ++ * error status of the journal as parameters. There is no guarantee of ++ * ordering between handles within a single transaction, nor between ++ * callbacks registered on the same handle. ++ * ++ * The caller is responsible for allocating the journal_callback struct. ++ * This is to allow the caller to add as much extra data to the callback ++ * as needed, but reduce the overhead of multiple allocations. The caller ++ * allocated struct must start with a struct journal_callback at offset 0, ++ * and has the caller-specific data afterwards. ++ */ ++void journal_callback_set(handle_t *handle, ++ void (*func)(struct journal_callback *jcb, int error), ++ struct journal_callback *jcb) ++{ ++ list_add_tail(&jcb->jcb_list, &handle->h_jcb); ++ jcb->jcb_func = func; ++} ++ ++/* + * All done for a particular handle. + * + * There is not much action needed here. We just return any remaining +@@ -1385,7 +1423,10 @@ + wake_up(&journal->j_wait_transaction_locked); + } + +- /* ++ /* Move callbacks from the handle to the transaction. */ ++ list_splice(&handle->h_jcb, &transaction->t_jcb); ++ ++ /* + * If the handle is marked SYNC, we need to set another commit + * going! We also want to force a commit if the current + * transaction is occupying too much of the log, or if the diff --git a/lustre/kernel_patches/patches/jbd-2.4.19-pre1-jcberr.patch b/lustre/kernel_patches/patches/jbd-2.4.19-pre1-jcberr.patch new file mode 100644 index 0000000..bbbf613 --- /dev/null +++ b/lustre/kernel_patches/patches/jbd-2.4.19-pre1-jcberr.patch @@ -0,0 +1,274 @@ +Index: linux-2.4.19-pre1/include/linux/jbd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/jbd.h 2003-11-21 03:00:11.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/jbd.h 2003-11-21 03:04:47.000000000 +0300 +@@ -275,6 +275,13 @@ + return bh->b_private; + } + ++#define HAVE_JOURNAL_CALLBACK_STATUS ++struct journal_callback { ++ struct list_head jcb_list; ++ void (*jcb_func)(struct journal_callback *jcb, int error); ++ /* user data goes here */ ++}; ++ + struct jbd_revoke_table_s; + + /* The handle_t type represents a single atomic update being performed +@@ -305,6 +312,12 @@ + operations */ + int h_err; + ++ /* List of application registered callbacks for this handle. ++ * The function(s) will be called after the transaction that ++ * this handle is part of has been committed to disk. ++ */ ++ struct list_head h_jcb; ++ + /* Flags */ + unsigned int h_sync: 1; /* sync-on-close */ + unsigned int h_jdata: 1; /* force data journaling */ +@@ -424,6 +437,10 @@ + + /* How many handles used this transaction? */ + int t_handle_count; ++ ++ /* List of registered callback functions for this transaction. ++ * Called when the transaction is committed. */ ++ struct list_head t_jcb; + }; + + +@@ -672,6 +689,9 @@ + extern int journal_try_to_free_buffers(journal_t *, struct page *, int); + extern int journal_stop(handle_t *); + extern int journal_flush (journal_t *); ++extern void journal_callback_set(handle_t *handle, ++ void (*fn)(struct journal_callback *,int), ++ struct journal_callback *jcb); + + extern void journal_lock_updates (journal_t *); + extern void journal_unlock_updates (journal_t *); +Index: linux-2.4.19-pre1/fs/jbd/checkpoint.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/jbd/checkpoint.c 2003-11-21 02:53:20.000000000 +0300 ++++ linux-2.4.19-pre1/fs/jbd/checkpoint.c 2003-11-21 03:04:47.000000000 +0300 +@@ -601,7 +601,8 @@ + J_ASSERT (transaction->t_log_list == NULL); + J_ASSERT (transaction->t_checkpoint_list == NULL); + J_ASSERT (transaction->t_updates == 0); +- ++ J_ASSERT (list_empty(&transaction->t_jcb)); ++ + J_ASSERT (transaction->t_journal->j_committing_transaction != + transaction); + +Index: linux-2.4.19-pre1/fs/jbd/commit.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/jbd/commit.c 2003-11-21 02:53:20.000000000 +0300 ++++ linux-2.4.19-pre1/fs/jbd/commit.c 2003-11-21 03:04:47.000000000 +0300 +@@ -480,7 +480,7 @@ + transaction's t_log_list queue, and metadata buffers are on + the t_iobuf_list queue. + +- Wait for the transactions in reverse order. That way we are ++ Wait for the buffers in reverse order. That way we are + less likely to be woken up until all IOs have completed, and + so we incur less scheduling load. + */ +@@ -571,8 +571,10 @@ + + jbd_debug(3, "JBD: commit phase 6\n"); + +- if (is_journal_aborted(journal)) ++ if (is_journal_aborted(journal)) { ++ unlock_journal(journal); + goto skip_commit; ++ } + + /* Done it all: now write the commit record. We should have + * cleaned up our previous buffers by now, so if we are in abort +@@ -582,9 +584,10 @@ + descriptor = journal_get_descriptor_buffer(journal); + if (!descriptor) { + __journal_abort_hard(journal); ++ unlock_journal(journal); + goto skip_commit; + } +- ++ + /* AKPM: buglet - add `i' to tmp! */ + for (i = 0; i < jh2bh(descriptor)->b_size; i += 512) { + journal_header_t *tmp = +@@ -605,14 +608,32 @@ + put_bh(bh); /* One for getblk() */ + journal_unlock_journal_head(descriptor); + } +- lock_journal(journal); + + /* End of a transaction! Finally, we can do checkpoint + processing: any buffers committed as a result of this + transaction can be removed from any checkpoint list it was on + before. */ + +-skip_commit: ++skip_commit: /* The journal should be unlocked by now. */ ++ ++ /* Call any callbacks that had been registered for handles in this ++ * transaction. It is up to the callback to free any allocated ++ * memory. ++ */ ++ if (!list_empty(&commit_transaction->t_jcb)) { ++ struct list_head *p, *n; ++ int error = is_journal_aborted(journal); ++ ++ list_for_each_safe(p, n, &commit_transaction->t_jcb) { ++ struct journal_callback *jcb; ++ ++ jcb = list_entry(p, struct journal_callback, jcb_list); ++ list_del(p); ++ jcb->jcb_func(jcb, error); ++ } ++ } ++ ++ lock_journal(journal); + + jbd_debug(3, "JBD: commit phase 7\n"); + +Index: linux-2.4.19-pre1/fs/jbd/journal.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/jbd/journal.c 2003-11-21 02:53:20.000000000 +0300 ++++ linux-2.4.19-pre1/fs/jbd/journal.c 2003-11-21 03:04:47.000000000 +0300 +@@ -58,6 +58,7 @@ + #endif + EXPORT_SYMBOL(journal_flush); + EXPORT_SYMBOL(journal_revoke); ++EXPORT_SYMBOL(journal_callback_set); + + EXPORT_SYMBOL(journal_init_dev); + EXPORT_SYMBOL(journal_init_inode); +Index: linux-2.4.19-pre1/fs/jbd/transaction.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/jbd/transaction.c 2003-11-21 02:53:20.000000000 +0300 ++++ linux-2.4.19-pre1/fs/jbd/transaction.c 2003-11-21 03:05:14.000000000 +0300 +@@ -57,6 +57,7 @@ + transaction->t_state = T_RUNNING; + transaction->t_tid = journal->j_transaction_sequence++; + transaction->t_expires = jiffies + journal->j_commit_interval; ++ INIT_LIST_HEAD(&transaction->t_jcb); + + /* Set up the commit timer for the new transaction. */ + J_ASSERT (!journal->j_commit_timer_active); +@@ -90,7 +91,14 @@ + transaction_t *transaction; + int needed; + int nblocks = handle->h_buffer_credits; +- ++ ++ if (nblocks > journal->j_max_transaction_buffers) { ++ jbd_debug(1, "JBD: %s wants too many credits (%d > %d)\n", ++ current->comm, nblocks, ++ journal->j_max_transaction_buffers); ++ return -ENOSPC; ++ } ++ + jbd_debug(3, "New handle %p going live.\n", handle); + + repeat: +@@ -196,6 +204,20 @@ + return 0; + } + ++/* Allocate a new handle. This should probably be in a slab... */ ++static handle_t *new_handle(int nblocks) ++{ ++ handle_t *handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ if (!handle) ++ return NULL; ++ memset(handle, 0, sizeof (handle_t)); ++ handle->h_buffer_credits = nblocks; ++ handle->h_ref = 1; ++ INIT_LIST_HEAD(&handle->h_jcb); ++ ++ return handle; ++} ++ + /* + * Obtain a new handle. + * +@@ -222,14 +244,11 @@ + handle->h_ref++; + return handle; + } +- +- handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ ++ handle = new_handle(nblocks); + if (!handle) + return ERR_PTR(-ENOMEM); +- memset (handle, 0, sizeof (handle_t)); + +- handle->h_buffer_credits = nblocks; +- handle->h_ref = 1; + current->journal_info = handle; + + err = start_this_handle(journal, handle); +@@ -328,14 +347,11 @@ + + if (is_journal_aborted(journal)) + return ERR_PTR(-EIO); +- +- handle = jbd_kmalloc(sizeof (handle_t), GFP_NOFS); ++ ++ handle = new_handle(nblocks); + if (!handle) + return ERR_PTR(-ENOMEM); +- memset (handle, 0, sizeof (handle_t)); + +- handle->h_buffer_credits = nblocks; +- handle->h_ref = 1; + current->journal_info = handle; + + err = try_start_this_handle(journal, handle); +@@ -1324,6 +1340,28 @@ + #endif + + /* ++ * Register a callback function for this handle. The function will be ++ * called when the transaction that this handle is part of has been ++ * committed to disk with the original callback data struct and the ++ * error status of the journal as parameters. There is no guarantee of ++ * ordering between handles within a single transaction, nor between ++ * callbacks registered on the same handle. ++ * ++ * The caller is responsible for allocating the journal_callback struct. ++ * This is to allow the caller to add as much extra data to the callback ++ * as needed, but reduce the overhead of multiple allocations. The caller ++ * allocated struct must start with a struct journal_callback at offset 0, ++ * and has the caller-specific data afterwards. ++ */ ++void journal_callback_set(handle_t *handle, ++ void (*func)(struct journal_callback *jcb, int error), ++ struct journal_callback *jcb) ++{ ++ list_add_tail(&jcb->jcb_list, &handle->h_jcb); ++ jcb->jcb_func = func; ++} ++ ++/* + * All done for a particular handle. + * + * There is not much action needed here. We just return any remaining +@@ -1389,7 +1427,10 @@ + wake_up(&journal->j_wait_transaction_locked); + } + +- /* ++ /* Move callbacks from the handle to the transaction. */ ++ list_splice(&handle->h_jcb, &transaction->t_jcb); ++ ++ /* + * If the handle is marked SYNC, we need to set another commit + * going! We also want to force a commit if the current + * transaction is occupying too much of the log, or if the diff --git a/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch b/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch new file mode 100644 index 0000000..8411137 --- /dev/null +++ b/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch @@ -0,0 +1,35 @@ +Index: linux-2.4.19.SuSE/fs/jbd/transaction.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/transaction.c Sun Nov 16 01:38:25 2003 ++++ linux-2.4.19.SuSE/fs/jbd/transaction.c Sun Nov 16 01:44:26 2003 +@@ -1094,7 +1094,6 @@ + + spin_lock(&journal_datalist_lock); + set_bit(BH_JBDDirty, &bh->b_state); +- set_buffer_flushtime(bh); + + J_ASSERT_JH(jh, jh->b_transaction != NULL); + +@@ -1995,6 +1994,13 @@ + spin_unlock(&journal_datalist_lock); + } + ++static void jbd_refile_buffer(struct buffer_head *bh) ++{ ++ if (buffer_dirty(bh) && (bh->b_list != BUF_DIRTY)) ++ set_buffer_flushtime(bh); ++ refile_buffer(bh); ++} ++ + /* + * Remove a buffer from its current buffer list in preparation for + * dropping it from its current transaction entirely. If the buffer has +@@ -2022,7 +2028,7 @@ + J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING); + } else { + /* Onto BUF_DIRTY for writeback */ +- refile_buffer(jh2bh(jh)); ++ jbd_refile_buffer(jh2bh(jh)); + } + } + diff --git a/lustre/kernel_patches/patches/kdev-2.4.19-pre1.patch b/lustre/kernel_patches/patches/kdev-2.4.19-pre1.patch new file mode 100644 index 0000000..5499df3 --- /dev/null +++ b/lustre/kernel_patches/patches/kdev-2.4.19-pre1.patch @@ -0,0 +1,12 @@ +Index: linux-2.4.19-pre1/include/linux/kdev_t.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/kdev_t.h 2003-11-20 19:01:38.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/kdev_t.h 2003-11-21 04:02:03.000000000 +0300 +@@ -82,6 +82,7 @@ + #define minor(d) MINOR(d) + #define kdev_same(a,b) (a==b) + #define kdev_none(d) (!(d)) ++#define val_to_kdev(d) ((kdev_t)(d)) + + /* + As long as device numbers in the outside world have 16 bits only, diff --git a/lustre/kernel_patches/patches/kernel_text_address-2.4.19-pre1.patch b/lustre/kernel_patches/patches/kernel_text_address-2.4.19-pre1.patch new file mode 100644 index 0000000..df37f7a --- /dev/null +++ b/lustre/kernel_patches/patches/kernel_text_address-2.4.19-pre1.patch @@ -0,0 +1,58 @@ +Index: linux-2.4.19-pre1/arch/i386/kernel/traps.c +=================================================================== +--- linux-2.4.19-pre1.orig/arch/i386/kernel/traps.c 2001-09-30 23:26:08.000000000 +0400 ++++ linux-2.4.19-pre1/arch/i386/kernel/traps.c 2003-11-21 02:31:28.000000000 +0300 +@@ -961,3 +961,40 @@ + cobalt_init(); + #endif + } ++#ifdef CONFIG_MODULES ++extern struct module *module_list; ++extern struct module kernel_module; ++#endif ++ ++int is_kernel_text_address(unsigned long addr) ++{ ++ int retval = 0; ++#ifdef CONFIG_MODULES ++ struct module *mod; ++#endif ++ if (addr >= (unsigned long) &_stext && ++ addr <= (unsigned long) &_etext); ++ return 1; ++ ++#ifdef CONFIG_MODULES ++ for (mod = module_list; mod != &kernel_module; mod = mod->next) { ++ /* mod_bound tests for addr being inside the vmalloc'ed ++ * module area. Of course it'd be better to test only ++ * for the .text subset... */ ++ if (mod_bound(addr, 0, mod)) { ++ retval = 1; ++ break; ++ } ++ } ++#endif ++ ++ return retval; ++} ++ ++int lookup_symbol(unsigned long address, char *buf, int buflen) ++{ ++ return -ENOSYS; ++} ++ ++EXPORT_SYMBOL_GPL(is_kernel_text_address); ++EXPORT_SYMBOL_GPL(lookup_symbol); +Index: linux-2.4.19-pre1/arch/i386/kernel/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/arch/i386/kernel/Makefile 2001-11-10 01:21:21.000000000 +0300 ++++ linux-2.4.19-pre1/arch/i386/kernel/Makefile 2003-11-21 02:31:46.000000000 +0300 +@@ -14,7 +14,7 @@ + + O_TARGET := kernel.o + +-export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o ++export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o traps.o + + obj-y := process.o semaphore.o signal.o entry.o traps.o irq.o vm86.o \ + ptrace.o i8259.o ioport.o ldt.o setup.o time.o sys_i386.o \ diff --git a/lustre/kernel_patches/patches/kernel_text_address-2.4.22-vanilla.patch b/lustre/kernel_patches/patches/kernel_text_address-2.4.22-vanilla.patch new file mode 100644 index 0000000..6e9cd93 --- /dev/null +++ b/lustre/kernel_patches/patches/kernel_text_address-2.4.22-vanilla.patch @@ -0,0 +1,59 @@ +Index: linux-2.4.22-vanilla/arch/i386/kernel/traps.c +=================================================================== +--- linux-2.4.22-vanilla.orig/arch/i386/kernel/traps.c 2003-05-16 05:28:59.000000000 +0400 ++++ linux-2.4.22-vanilla/arch/i386/kernel/traps.c 2003-11-14 00:35:36.000000000 +0300 +@@ -1004,3 +1004,40 @@ + cobalt_init(); + #endif + } ++#ifdef CONFIG_MODULES ++extern struct module *module_list; ++extern struct module kernel_module; ++#endif ++ ++int is_kernel_text_address(unsigned long addr) ++{ ++ int retval = 0; ++#ifdef CONFIG_MODULES ++ struct module *mod; ++#endif ++ if (addr >= (unsigned long) &_stext && ++ addr <= (unsigned long) &_etext); ++ return 1; ++ ++#ifdef CONFIG_MODULES ++ for (mod = module_list; mod != &kernel_module; mod = mod->next) { ++ /* mod_bound tests for addr being inside the vmalloc'ed ++ * module area. Of course it'd be better to test only ++ * for the .text subset... */ ++ if (mod_bound(addr, 0, mod)) { ++ retval = 1; ++ break; ++ } ++ } ++#endif ++ ++ return retval; ++} ++ ++int lookup_symbol(unsigned long address, char *buf, int buflen) ++{ ++ return -ENOSYS; ++} ++ ++EXPORT_SYMBOL_GPL(is_kernel_text_address); ++EXPORT_SYMBOL_GPL(lookup_symbol); +Index: linux-2.4.22-vanilla/arch/i386/kernel/Makefile +=================================================================== +--- linux-2.4.22-vanilla.orig/arch/i386/kernel/Makefile 2003-11-03 23:22:06.000000000 +0300 ++++ linux-2.4.22-vanilla/arch/i386/kernel/Makefile 2003-11-14 00:30:28.000000000 +0300 +@@ -14,7 +14,8 @@ + + O_TARGET := kernel.o + +-export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o setup.o ++export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o \ ++ setup.o traps.o + + obj-y := process.o semaphore.o signal.o entry.o traps.o irq.o vm86.o \ + ptrace.o i8259.o ioport.o ldt.o setup.o time.o sys_i386.o \ diff --git a/lustre/kernel_patches/patches/linux-2.4.19-pre1-xattr-0.8.54.patch b/lustre/kernel_patches/patches/linux-2.4.19-pre1-xattr-0.8.54.patch new file mode 100644 index 0000000..d5d0030 --- /dev/null +++ b/lustre/kernel_patches/patches/linux-2.4.19-pre1-xattr-0.8.54.patch @@ -0,0 +1,6038 @@ + Documentation/Configure.help | 66 ++ + arch/alpha/defconfig | 7 + arch/alpha/kernel/entry.S | 12 + arch/arm/defconfig | 7 + arch/arm/kernel/calls.S | 24 + arch/i386/defconfig | 7 + arch/ia64/defconfig | 7 + arch/ia64/kernel/entry.S | 24 + arch/m68k/defconfig | 7 + arch/mips/defconfig | 7 + arch/mips64/defconfig | 7 + arch/ppc/defconfig | 14 + arch/ppc64/kernel/misc.S | 2 + arch/s390/defconfig | 7 + arch/s390/kernel/entry.S | 24 + arch/s390x/defconfig | 7 + arch/s390x/kernel/entry.S | 24 + arch/s390x/kernel/wrapper32.S | 92 +++ + arch/sparc/defconfig | 7 + arch/sparc/kernel/systbls.S | 10 + arch/sparc64/defconfig | 7 + arch/sparc64/kernel/systbls.S | 20 + fs/Config.in | 14 + fs/Makefile | 3 + fs/ext2/Makefile | 4 + fs/ext2/file.c | 5 + fs/ext2/ialloc.c | 2 + fs/ext2/inode.c | 34 - + fs/ext2/namei.c | 14 + fs/ext2/super.c | 29 + fs/ext2/symlink.c | 14 + fs/ext2/xattr.c | 1212 +++++++++++++++++++++++++++++++++++++++++ + fs/ext2/xattr_user.c | 103 +++ + fs/ext3/Makefile | 10 + fs/ext3/file.c | 5 + fs/ext3/ialloc.c | 2 + fs/ext3/inode.c | 35 - + fs/ext3/namei.c | 21 + fs/ext3/super.c | 36 + + fs/ext3/symlink.c | 14 + fs/ext3/xattr.c | 1225 ++++++++++++++++++++++++++++++++++++++++++ + fs/ext3/xattr_user.c | 111 +++ + fs/jfs/jfs_xattr.h | 6 + fs/jfs/xattr.c | 6 + fs/mbcache.c | 648 ++++++++++++++++++++++ + include/asm-arm/unistd.h | 2 + include/asm-ia64/unistd.h | 13 + include/asm-ppc64/unistd.h | 2 + include/asm-s390/unistd.h | 15 + include/asm-s390x/unistd.h | 15 + include/asm-sparc/unistd.h | 24 + include/asm-sparc64/unistd.h | 24 + include/linux/cache_def.h | 15 + include/linux/errno.h | 4 + include/linux/ext2_fs.h | 31 - + include/linux/ext2_xattr.h | 157 +++++ + include/linux/ext3_fs.h | 31 - + include/linux/ext3_jbd.h | 8 + include/linux/ext3_xattr.h | 157 +++++ + include/linux/fs.h | 2 + include/linux/mbcache.h | 69 ++ + kernel/ksyms.c | 4 + mm/vmscan.c | 35 + + fs/ext3/ext3-exports.c | 14 + + 64 files changed, 4355 insertions(+), 195 deletions(-) + +Index: linux-2.4.19-pre1/Documentation/Configure.help +=================================================================== +--- linux-2.4.19-pre1.orig/Documentation/Configure.help 2003-11-20 19:01:44.000000000 +0300 ++++ linux-2.4.19-pre1/Documentation/Configure.help 2003-11-21 03:51:05.000000000 +0300 +@@ -14035,6 +14035,39 @@ + be compiled as a module, and so this could be dangerous. Most + everyone wants to say Y here. + ++Ext2 extended attributes ++CONFIG_EXT2_FS_XATTR ++ Extended attributes are name:value pairs associated with inodes by ++ the kernel or by users (see the attr(5) manual page, or visit ++ for details). ++ ++ If unsure, say N. ++ ++Ext2 extended attribute block sharing ++CONFIG_EXT2_FS_XATTR_SHARING ++ This options enables code for sharing identical extended attribute ++ blocks among multiple inodes. ++ ++ Usually, say Y. ++ ++Ext2 extended user attributes ++CONFIG_EXT2_FS_XATTR_USER ++ This option enables extended user attributes on ext2. Processes can ++ associate extended user attributes with inodes to store additional ++ information such as the character encoding of files, etc. (see the ++ attr(5) manual page, or visit for details). ++ ++ If unsure, say N. ++ ++Ext2 trusted extended attributes ++CONFIG_EXT2_FS_XATTR_TRUSTED ++ This option enables extended attributes on ext2 that are accessible ++ (and visible) only to users capable of CAP_SYS_ADMIN. Usually this ++ is only the super user. Trusted extended attributes are meant for ++ implementing system/security services. ++ ++ If unsure, say N. ++ + Ext3 journalling file system support (EXPERIMENTAL) + CONFIG_EXT3_FS + This is the journalling version of the Second extended file system +@@ -14067,6 +14100,39 @@ + of your root partition (the one containing the directory /) cannot + be compiled as a module, and so this may be dangerous. + ++Ext3 extended attributes ++CONFIG_EXT3_FS_XATTR ++ Extended attributes are name:value pairs associated with inodes by ++ the kernel or by users (see the attr(5) manual page, or visit ++ for details). ++ ++ If unsure, say N. ++ ++Ext3 extended attribute block sharing ++CONFIG_EXT3_FS_XATTR_SHARING ++ This options enables code for sharing identical extended attribute ++ blocks among multiple inodes. ++ ++ Usually, say Y. ++ ++Ext3 extended user attributes ++CONFIG_EXT3_FS_XATTR_USER ++ This option enables extended user attributes on ext3. Processes can ++ associate extended user attributes with inodes to store additional ++ information such as the character encoding of files, etc. (see the ++ attr(5) manual page, or visit for details). ++ ++ If unsure, say N. ++ ++Ext3 trusted extended attributes ++CONFIG_EXT3_FS_XATTR_TRUSTED ++ This option enables extended attributes on ext3 that are accessible ++ (and visible) only to users capable of CAP_SYS_ADMIN. Usually this ++ is only the super user. Trusted extended attributes are meant for ++ implementing system/security services. ++ ++ If unsure, say N. ++ + Journal Block Device support (JBD for ext3) (EXPERIMENTAL) + CONFIG_JBD + This is a generic journalling layer for block devices. It is +Index: linux-2.4.19-pre1/arch/alpha/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/alpha/defconfig 2001-11-20 02:19:42.000000000 +0300 ++++ linux-2.4.19-pre1/arch/alpha/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_ALPHA=y + # CONFIG_UID16 is not set + # CONFIG_RWSEM_GENERIC_SPINLOCK is not set +Index: linux-2.4.19-pre1/arch/alpha/kernel/entry.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/alpha/kernel/entry.S 2001-11-10 00:45:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/alpha/kernel/entry.S 2003-11-21 03:51:05.000000000 +0300 +@@ -1148,3 +1148,16 @@ + .quad sys_gettid + .quad sys_readahead + .quad sys_ni_syscall /* 380, sys_security */ ++ .quad sys_ni_syscall /* 381, sys_tkill */ ++ .quad sys_setxattr ++ .quad sys_lsetxattr ++ .quad sys_fsetxattr ++ .quad sys_getxattr /* 385 */ ++ .quad sys_lgetxattr ++ .quad sys_fgetxattr ++ .quad sys_listxattr ++ .quad sys_llistxattr ++ .quad sys_flistxattr /* 390 */ ++ .quad sys_removexattr ++ .quad sys_lremovexattr ++ .quad sys_fremovexattr +Index: linux-2.4.19-pre1/arch/arm/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/arm/defconfig 2001-05-20 04:43:05.000000000 +0400 ++++ linux-2.4.19-pre1/arch/arm/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_ARM=y + # CONFIG_EISA is not set + # CONFIG_SBUS is not set +Index: linux-2.4.19-pre1/arch/arm/kernel/calls.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/arm/kernel/calls.S 2001-10-08 21:39:18.000000000 +0400 ++++ linux-2.4.19-pre1/arch/arm/kernel/calls.S 2003-11-21 03:51:05.000000000 +0300 +@@ -236,6 +236,22 @@ + .long SYMBOL_NAME(sys_mincore) + /* 220 */ .long SYMBOL_NAME(sys_madvise) + .long SYMBOL_NAME(sys_fcntl64) ++ .long SYMBOL_NAME(sys_ni_syscall) /* TUX */ ++ .long SYMBOL_NAME(sys_ni_syscall) /* Security */ ++ .long SYMBOL_NAME(sys_gettid) ++ /* 225 */ .long SYMBOL_NAME(sys_readahead) ++ .long SYMBOL_NAME(sys_setxattr) ++ .long SYMBOL_NAME(sys_lsetxattr) ++ .long SYMBOL_NAME(sys_fsetxattr) ++ .long SYMBOL_NAME(sys_getxattr) ++/* 230 */ .long SYMBOL_NAME(sys_lgetxattr) ++ .long SYMBOL_NAME(sys_fgetxattr) ++ .long SYMBOL_NAME(sys_listxattr) ++ .long SYMBOL_NAME(sys_llistxattr) ++ .long SYMBOL_NAME(sys_flistxattr) ++/* 235 */ .long SYMBOL_NAME(sys_removexattr) ++ .long SYMBOL_NAME(sys_lremovexattr) ++ .long SYMBOL_NAME(sys_fremovexattr) + __syscall_end: + + .rept NR_syscalls - (__syscall_end - __syscall_start) / 4 +Index: linux-2.4.19-pre1/arch/i386/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/i386/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/i386/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_X86=y + CONFIG_ISA=y + # CONFIG_SBUS is not set +Index: linux-2.4.19-pre1/arch/ia64/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/ia64/defconfig 2001-11-10 01:26:17.000000000 +0300 ++++ linux-2.4.19-pre1/arch/ia64/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + + # + # Code maturity level options +Index: linux-2.4.19-pre1/arch/ia64/kernel/entry.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/ia64/kernel/entry.S 2001-11-10 01:26:17.000000000 +0300 ++++ linux-2.4.19-pre1/arch/ia64/kernel/entry.S 2003-11-21 03:51:05.000000000 +0300 +@@ -1130,18 +1130,18 @@ + data8 sys_getdents64 + data8 sys_getunwind // 1215 + data8 sys_readahead +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall // 1220 +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall // 1225 +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall ++ data8 sys_setxattr ++ data8 sys_lsetxattr ++ data8 sys_fsetxattr ++ data8 sys_getxattr // 1220 ++ data8 sys_lgetxattr ++ data8 sys_fgetxattr ++ data8 sys_listxattr ++ data8 sys_llistxattr ++ data8 sys_flistxattr // 1225 ++ data8 sys_removexattr ++ data8 sys_lremovexattr ++ data8 sys_fremovexattr + data8 ia64_ni_syscall + data8 ia64_ni_syscall // 1230 + data8 ia64_ni_syscall +Index: linux-2.4.19-pre1/arch/m68k/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/m68k/defconfig 2000-06-19 23:56:08.000000000 +0400 ++++ linux-2.4.19-pre1/arch/m68k/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_UID16=y + + # +Index: linux-2.4.19-pre1/arch/mips/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/mips/defconfig 2001-09-09 21:43:02.000000000 +0400 ++++ linux-2.4.19-pre1/arch/mips/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_MIPS=y + # CONFIG_SMP is not set + +Index: linux-2.4.19-pre1/arch/mips64/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/mips64/defconfig 2001-09-09 21:43:02.000000000 +0400 ++++ linux-2.4.19-pre1/arch/mips64/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + + # + # Code maturity level options +Index: linux-2.4.19-pre1/arch/ppc/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/ppc/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/ppc/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,20 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_UID16 is not set + # CONFIG_RWSEM_GENERIC_SPINLOCK is not set + CONFIG_RWSEM_XCHGADD_ALGORITHM=y +Index: linux-2.4.19-pre1/arch/s390/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/s390/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/s390/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_ISA is not set + # CONFIG_EISA is not set + # CONFIG_MCA is not set +Index: linux-2.4.19-pre1/arch/s390/kernel/entry.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/s390/kernel/entry.S 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/s390/kernel/entry.S 2003-11-21 03:51:05.000000000 +0300 +@@ -599,8 +599,19 @@ + .long sys_fcntl64 + .long sys_ni_syscall /* 222 - reserved for posix_acl */ + .long sys_ni_syscall /* 223 - reserved for posix_acl */ +- .long sys_ni_syscall /* 224 - reserved for posix_acl */ +- .rept 255-224 ++ .long sys_setxattr ++ .long sys_lsetxattr /* 225 */ ++ .long sys_fsetxattr ++ .long sys_getxattr ++ .long sys_lgetxattr ++ .long sys_fgetxattr ++ .long sys_listxattr /* 230 */ ++ .long sys_llistxattr ++ .long sys_flistxattr ++ .long sys_removexattr ++ .long sys_lremovexattr ++ .long sys_fremovexattr /* 235 */ ++ .rept 255-235 + .long sys_ni_syscall + .endr + +Index: linux-2.4.19-pre1/arch/s390x/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/s390x/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/s390x/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_ISA is not set + # CONFIG_EISA is not set + # CONFIG_MCA is not set +Index: linux-2.4.19-pre1/arch/s390x/kernel/entry.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/s390x/kernel/entry.S 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/s390x/kernel/entry.S 2003-11-21 03:51:05.000000000 +0300 +@@ -632,8 +632,19 @@ + .long SYSCALL(sys_ni_syscall,sys32_fcntl64_wrapper) + .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 222 - reserved for posix_acl */ + .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 223 - reserved for posix_acl */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 224 - reserved for posix_acl */ +- .rept 255-224 ++ .long SYSCALL(sys_setxattr,sys32_setxattr_wrapper) ++ .long SYSCALL(sys_lsetxattr,sys32_lsetxattr_wrapper) /* 225 */ ++ .long SYSCALL(sys_fsetxattr,sys32_fsetxattr_wrapper) ++ .long SYSCALL(sys_getxattr,sys32_getxattr_wrapper) ++ .long SYSCALL(sys_lgetxattr,sys32_lgetxattr_wrapper) ++ .long SYSCALL(sys_fgetxattr,sys32_fgetxattr_wrapper) ++ .long SYSCALL(sys_listxattr,sys32_listxattr_wrapper) /* 230 */ ++ .long SYSCALL(sys_llistxattr,sys32_llistxattr_wrapper) ++ .long SYSCALL(sys_flistxattr,sys32_flistxattr_wrapper) ++ .long SYSCALL(sys_removexattr,sys32_removexattr_wrapper) ++ .long SYSCALL(sys_lremovexattr,sys32_lremovexattr_wrapper) ++ .long SYSCALL(sys_fremovexattr,sys32_fremovexattr_wrapper)/* 235 */ ++ .rept 255-235 + .long SYSCALL(sys_ni_syscall,sys_ni_syscall) + .endr + +Index: linux-2.4.19-pre1/arch/s390x/kernel/wrapper32.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/s390x/kernel/wrapper32.S 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/s390x/kernel/wrapper32.S 2003-11-21 03:51:05.000000000 +0300 +@@ -1091,3 +1091,95 @@ + llgtr %r3,%r3 # struct stat64 * + llgfr %r4,%r4 # long + jg sys32_fstat64 # branch to system call ++ ++ .globl sys32_setxattr_wrapper ++sys32_setxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_setxattr ++ ++ .globl sys32_lsetxattr_wrapper ++sys32_lsetxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_lsetxattr ++ ++ .globl sys32_fsetxattr_wrapper ++sys32_fsetxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_fsetxattr ++ ++ .globl sys32_getxattr_wrapper ++sys32_getxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_getxattr ++ ++ .globl sys32_lgetxattr_wrapper ++sys32_lgetxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_lgetxattr ++ ++ .globl sys32_fgetxattr_wrapper ++sys32_fgetxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_fgetxattr ++ ++ .globl sys32_listxattr_wrapper ++sys32_listxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_listxattr ++ ++ .globl sys32_llistxattr_wrapper ++sys32_llistxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_llistxattr ++ ++ .globl sys32_flistxattr_wrapper ++sys32_flistxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_flistxattr ++ ++ .globl sys32_removexattr_wrapper ++sys32_removexattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ jg sys_removexattr ++ ++ .globl sys32_lremovexattr_wrapper ++sys32_lremovexattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ jg sys_lremovexattr ++ ++ .globl sys32_fremovexattr_wrapper ++sys32_fremovexattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ jg sys_fremovexattr ++ ++ +Index: linux-2.4.19-pre1/arch/sparc/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/sparc/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/sparc/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_UID16=y + CONFIG_HIGHMEM=y + +Index: linux-2.4.19-pre1/arch/sparc/kernel/systbls.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/sparc/kernel/systbls.S 2001-10-21 21:36:54.000000000 +0400 ++++ linux-2.4.19-pre1/arch/sparc/kernel/systbls.S 2003-11-21 03:51:05.000000000 +0300 +@@ -51,11 +51,11 @@ + /*150*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + /*155*/ .long sys_fcntl64, sys_nis_syscall, sys_statfs, sys_fstatfs, sys_oldumount + /*160*/ .long sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_nis_syscall +-/*165*/ .long sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_nis_syscall +-/*170*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_getdents +-/*175*/ .long sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_sigpending, sys_query_module +-/*185*/ .long sys_setpgid, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_newuname ++/*165*/ .long sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_setxattr ++/*170*/ .long sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys_getdents ++/*175*/ .long sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .long sys_flistxattr, sys_removexattr, sys_lremovexattr, sys_sigpending, sys_query_module ++/*185*/ .long sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sys_newuname + /*190*/ .long sys_init_module, sys_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + /*195*/ .long sys_nis_syscall, sys_nis_syscall, sys_getppid, sparc_sigaction, sys_sgetmask + /*200*/ .long sys_ssetmask, sys_sigsuspend, sys_newlstat, sys_uselib, old_readdir +Index: linux-2.4.19-pre1/arch/sparc64/defconfig +=================================================================== +--- linux-2.4.19-pre1.orig/arch/sparc64/defconfig 2003-11-20 19:01:35.000000000 +0300 ++++ linux-2.4.19-pre1/arch/sparc64/defconfig 2003-11-21 03:51:05.000000000 +0300 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + + # + # Code maturity level options +Index: linux-2.4.19-pre1/arch/sparc64/kernel/systbls.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/sparc64/kernel/systbls.S 2001-10-21 21:36:54.000000000 +0400 ++++ linux-2.4.19-pre1/arch/sparc64/kernel/systbls.S 2003-11-21 03:51:05.000000000 +0300 +@@ -52,11 +52,11 @@ + /*150*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + .word sys32_fcntl64, sys_nis_syscall, sys32_statfs, sys32_fstatfs, sys_oldumount + /*160*/ .word sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_nis_syscall +- .word sys32_quotactl, sys_nis_syscall, sys32_mount, sys_ustat, sys_nis_syscall +-/*170*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys32_getdents +- .word sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys32_sigpending, sys32_query_module +- .word sys_setpgid, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sparc64_newuname ++ .word sys32_quotactl, sys_nis_syscall, sys32_mount, sys_ustat, sys_setxattr ++/*170*/ .word sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys32_getdents ++ .word sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .word sys_flistxattr, sys_removexattr, sys_lremovexattr, sys32_sigpending, sys32_query_module ++ .word sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sparc64_newuname + /*190*/ .word sys32_init_module, sparc64_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + .word sys_nis_syscall, sys_nis_syscall, sys_getppid, sys32_sigaction, sys_sgetmask + /*200*/ .word sys_ssetmask, sys_sigsuspend, sys32_newlstat, sys_uselib, old32_readdir +@@ -111,11 +111,11 @@ + /*150*/ .word sys_getsockname, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + .word sys_nis_syscall, sys_nis_syscall, sys_statfs, sys_fstatfs, sys_oldumount + /*160*/ .word sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_utrap_install +- .word sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_nis_syscall +-/*170*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_getdents +- .word sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_query_module +- .word sys_setpgid, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sparc64_newuname ++ .word sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_setxattr ++/*170*/ .word sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys_getdents ++ .word sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .word sys_flistxattr, sys_removexattr, sys_lremovexattr, sys_nis_syscall, sys_query_module ++ .word sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sparc64_newuname + /*190*/ .word sys_init_module, sparc64_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + .word sys_nis_syscall, sys_nis_syscall, sys_getppid, sys_nis_syscall, sys_sgetmask + /*200*/ .word sys_ssetmask, sys_nis_syscall, sys_newlstat, sys_uselib, sys_nis_syscall +Index: linux-2.4.19-pre1/fs/Config.in +=================================================================== +--- linux-2.4.19-pre1.orig/fs/Config.in 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/Config.in 2003-11-21 03:51:05.000000000 +0300 +@@ -22,6 +22,11 @@ + dep_tristate 'BFS file system support (EXPERIMENTAL)' CONFIG_BFS_FS $CONFIG_EXPERIMENTAL + + tristate 'Ext3 journalling file system support (EXPERIMENTAL)' CONFIG_EXT3_FS ++dep_mbool ' Ext3 extended attributes' CONFIG_EXT3_FS_XATTR $CONFIG_EXT3_FS ++dep_bool ' Ext3 extended attribute block sharing' \ ++ CONFIG_EXT3_FS_XATTR_SHARING $CONFIG_EXT3_FS_XATTR ++dep_bool ' Ext3 extended user attributes' \ ++ CONFIG_EXT3_FS_XATTR_USER $CONFIG_EXT3_FS_XATTR + # CONFIG_JBD could be its own option (even modular), but until there are + # other users than ext3, we will simply make it be the same as CONFIG_EXT3_FS + # dep_tristate ' Journal Block Device support (JBD for ext3)' CONFIG_JBD $CONFIG_EXT3_FS +@@ -77,6 +82,11 @@ + tristate 'ROM file system support' CONFIG_ROMFS_FS + + tristate 'Second extended fs support' CONFIG_EXT2_FS ++dep_mbool ' Ext2 extended attributes' CONFIG_EXT2_FS_XATTR $CONFIG_EXT2_FS ++dep_bool ' Ext2 extended attribute block sharing' \ ++ CONFIG_EXT2_FS_XATTR_SHARING $CONFIG_EXT2_FS_XATTR ++dep_bool ' Ext2 extended user attributes' \ ++ CONFIG_EXT2_FS_XATTR_USER $CONFIG_EXT2_FS_XATTR + + tristate 'System V/Xenix/V7/Coherent file system support' CONFIG_SYSV_FS + +@@ -156,6 +166,10 @@ + fi + fi + ++# Meta block cache for Extended Attributes (ext2/ext3) ++#tristate 'Meta block cache' CONFIG_FS_MBCACHE ++define_tristate CONFIG_FS_MBCACHE y ++ + mainmenu_option next_comment + comment 'Partition Types' + source fs/partitions/Config.in +Index: linux-2.4.19-pre1/fs/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/Makefile 2003-11-21 03:51:01.000000000 +0300 ++++ linux-2.4.19-pre1/fs/Makefile 2003-11-21 03:51:05.000000000 +0300 +@@ -14,7 +14,7 @@ + super.o block_dev.o char_dev.o stat.o exec.o pipe.o namei.o \ + fcntl.o ioctl.o readdir.o select.o fifo.o locks.o \ + dcache.o inode.o attr.o bad_inode.o file.o iobuf.o dnotify.o \ +- filesystems.o namespace.o seq_file.o ++ filesystems.o namespace.o seq_file.o xattr.o + + ifeq ($(CONFIG_QUOTA),y) + obj-y += dquot.o +@@ -78,6 +78,9 @@ + + obj-$(CONFIG_BINFMT_ELF) += binfmt_elf.o + ++export-objs += mbcache.o ++obj-$(CONFIG_FS_MBCACHE) += mbcache.o ++ + # persistent filesystems + obj-y += $(join $(subdir-y),$(subdir-y:%=/%.o)) + +Index: linux-2.4.19-pre1/fs/ext2/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/Makefile 2001-10-11 19:05:18.000000000 +0400 ++++ linux-2.4.19-pre1/fs/ext2/Makefile 2003-11-21 03:51:05.000000000 +0300 +@@ -13,4 +13,8 @@ + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) + ++export-objs += xattr.o ++obj-$(CONFIG_EXT2_FS_XATTR) += xattr.o ++obj-$(CONFIG_EXT2_FS_XATTR_USER) += xattr_user.o ++ + include $(TOPDIR)/Rules.make +Index: linux-2.4.19-pre1/fs/ext2/file.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/file.c 2001-10-11 19:05:18.000000000 +0400 ++++ linux-2.4.19-pre1/fs/ext2/file.c 2003-11-21 03:51:05.000000000 +0300 +@@ -20,6 +20,7 @@ + + #include + #include ++#include + #include + + /* +@@ -51,4 +52,8 @@ + + struct inode_operations ext2_file_inode_operations = { + truncate: ext2_truncate, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +Index: linux-2.4.19-pre1/fs/ext2/ialloc.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/ialloc.c 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext2/ialloc.c 2003-11-21 03:51:05.000000000 +0300 +@@ -15,6 +15,7 @@ + #include + #include + #include ++#include + #include + #include + +@@ -167,6 +168,7 @@ + */ + if (!is_bad_inode(inode)) { + /* Quota is already initialized in iput() */ ++ ext2_xattr_delete_inode(inode); + DQUOT_FREE_INODE(inode); + DQUOT_DROP(inode); + } +Index: linux-2.4.19-pre1/fs/ext2/inode.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/inode.c 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext2/inode.c 2003-11-21 03:51:05.000000000 +0300 +@@ -39,6 +39,18 @@ + static int ext2_update_inode(struct inode * inode, int do_sync); + + /* ++ * Test whether an inode is a fast symlink. ++ */ ++static inline int ext2_inode_is_fast_symlink(struct inode *inode) ++{ ++ int ea_blocks = inode->u.ext2_i.i_file_acl ? ++ (inode->i_sb->s_blocksize >> 9) : 0; ++ ++ return (S_ISLNK(inode->i_mode) && ++ inode->i_blocks - ea_blocks == 0); ++} ++ ++/* + * Called at each iput() + */ + void ext2_put_inode (struct inode * inode) +@@ -53,9 +65,7 @@ + { + lock_kernel(); + +- if (is_bad_inode(inode) || +- inode->i_ino == EXT2_ACL_IDX_INO || +- inode->i_ino == EXT2_ACL_DATA_INO) ++ if (is_bad_inode(inode)) + goto no_delete; + inode->u.ext2_i.i_dtime = CURRENT_TIME; + mark_inode_dirty(inode); +@@ -800,6 +810,8 @@ + if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode))) + return; ++ if (ext2_inode_is_fast_symlink(inode)) ++ return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + +@@ -889,8 +901,7 @@ + unsigned long offset; + struct ext2_group_desc * gdp; + +- if ((inode->i_ino != EXT2_ROOT_INO && inode->i_ino != EXT2_ACL_IDX_INO && +- inode->i_ino != EXT2_ACL_DATA_INO && ++ if ((inode->i_ino != EXT2_ROOT_INO && + inode->i_ino < EXT2_FIRST_INO(inode->i_sb)) || + inode->i_ino > le32_to_cpu(inode->i_sb->u.ext2_sb.s_es->s_inodes_count)) { + ext2_error (inode->i_sb, "ext2_read_inode", +@@ -975,10 +986,7 @@ + for (block = 0; block < EXT2_N_BLOCKS; block++) + inode->u.ext2_i.i_data[block] = raw_inode->i_block[block]; + +- if (inode->i_ino == EXT2_ACL_IDX_INO || +- inode->i_ino == EXT2_ACL_DATA_INO) +- /* Nothing to do */ ; +- else if (S_ISREG(inode->i_mode)) { ++ if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext2_file_inode_operations; + inode->i_fop = &ext2_file_operations; + inode->i_mapping->a_ops = &ext2_aops; +@@ -987,15 +995,17 @@ + inode->i_fop = &ext2_dir_operations; + inode->i_mapping->a_ops = &ext2_aops; + } else if (S_ISLNK(inode->i_mode)) { +- if (!inode->i_blocks) ++ if (ext2_inode_is_fast_symlink(inode)) + inode->i_op = &ext2_fast_symlink_inode_operations; + else { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext2_symlink_inode_operations; + inode->i_mapping->a_ops = &ext2_aops; + } +- } else ++ } else { ++ inode->i_op = &ext2_special_inode_operations; + init_special_inode(inode, inode->i_mode, + le32_to_cpu(raw_inode->i_block[0])); ++ } + brelse (bh); + inode->i_attr_flags = 0; + if (inode->u.ext2_i.i_flags & EXT2_SYNC_FL) { +Index: linux-2.4.19-pre1/fs/ext2/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/namei.c 2001-10-04 09:57:36.000000000 +0400 ++++ linux-2.4.19-pre1/fs/ext2/namei.c 2003-11-21 03:51:05.000000000 +0300 +@@ -31,6 +31,7 @@ + + #include + #include ++#include + #include + + /* +@@ -136,7 +137,7 @@ + + if (l > sizeof (inode->u.ext2_i.i_data)) { + /* slow symlink */ +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext2_symlink_inode_operations; + inode->i_mapping->a_ops = &ext2_aops; + err = block_symlink(inode, symname, l); + if (err) +@@ -345,4 +346,15 @@ + rmdir: ext2_rmdir, + mknod: ext2_mknod, + rename: ext2_rename, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, ++}; ++ ++struct inode_operations ext2_special_inode_operations = { ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +Index: linux-2.4.19-pre1/fs/ext2/super.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/super.c 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext2/super.c 2003-11-21 03:51:05.000000000 +0300 +@@ -21,6 +21,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -125,6 +126,7 @@ + int db_count; + int i; + ++ ext2_xattr_put_super(sb); + if (!(sb->s_flags & MS_RDONLY)) { + struct ext2_super_block *es = EXT2_SB(sb)->s_es; + +@@ -175,6 +177,13 @@ + this_char = strtok (NULL, ",")) { + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; ++#ifdef CONFIG_EXT2_FS_XATTR_USER ++ if (!strcmp (this_char, "user_xattr")) ++ set_opt (*mount_options, XATTR_USER); ++ else if (!strcmp (this_char, "nouser_xattr")) ++ clear_opt (*mount_options, XATTR_USER); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -424,6 +433,9 @@ + blocksize = BLOCK_SIZE; + + sb->u.ext2_sb.s_mount_opt = 0; ++#ifdef CONFIG_EXT2_FS_XATTR_USER ++ /* set_opt (sb->u.ext2_sb.s_mount_opt, XATTR_USER); */ ++#endif + if (!parse_options ((char *) data, &sb_block, &resuid, &resgid, + &sb->u.ext2_sb.s_mount_opt)) { + return NULL; +@@ -810,12 +822,27 @@ + + static int __init init_ext2_fs(void) + { +- return register_filesystem(&ext2_fs_type); ++ int error = init_ext2_xattr(); ++ if (error) ++ return error; ++ error = init_ext2_xattr_user(); ++ if (error) ++ goto fail; ++ error = register_filesystem(&ext2_fs_type); ++ if (!error) ++ return 0; ++ ++ exit_ext2_xattr_user(); ++fail: ++ exit_ext2_xattr(); ++ return error; + } + + static void __exit exit_ext2_fs(void) + { + unregister_filesystem(&ext2_fs_type); ++ exit_ext2_xattr_user(); ++ exit_ext2_xattr(); + } + + EXPORT_NO_SYMBOLS; +Index: linux-2.4.19-pre1/fs/ext2/symlink.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/symlink.c 2000-09-28 00:41:33.000000000 +0400 ++++ linux-2.4.19-pre1/fs/ext2/symlink.c 2003-11-21 03:51:05.000000000 +0300 +@@ -19,6 +19,7 @@ + + #include + #include ++#include + + static int ext2_readlink(struct dentry *dentry, char *buffer, int buflen) + { +@@ -32,7 +33,20 @@ + return vfs_follow_link(nd, s); + } + ++struct inode_operations ext2_symlink_inode_operations = { ++ readlink: page_readlink, ++ follow_link: page_follow_link, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, ++}; ++ + struct inode_operations ext2_fast_symlink_inode_operations = { + readlink: ext2_readlink, + follow_link: ext2_follow_link, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +Index: linux-2.4.19-pre1/fs/ext2/xattr.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/xattr.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext2/xattr.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,1212 @@ ++/* ++ * linux/fs/ext2/xattr.c ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ * ++ * Fix by Harrison Xing . ++ * Extended attributes for symlinks and special files added per ++ * suggestion of Luka Renko . ++ */ ++ ++/* ++ * Extended attributes are stored on disk blocks allocated outside of ++ * any inode. The i_file_acl field is then made to point to this allocated ++ * block. If all extended attributes of an inode are identical, these ++ * inodes may share the same extended attribute block. Such situations ++ * are automatically detected by keeping a cache of recent attribute block ++ * numbers and hashes over the block's contents in memory. ++ * ++ * ++ * Extended attribute block layout: ++ * ++ * +------------------+ ++ * | header | ++ * | entry 1 | | ++ * | entry 2 | | growing downwards ++ * | entry 3 | v ++ * | four null bytes | ++ * | . . . | ++ * | value 1 | ^ ++ * | value 3 | | growing upwards ++ * | value 2 | | ++ * +------------------+ ++ * ++ * The block header is followed by multiple entry descriptors. These entry ++ * descriptors are variable in size, and alligned to EXT2_XATTR_PAD ++ * byte boundaries. The entry descriptors are sorted by attribute name, ++ * so that two extended attribute blocks can be compared efficiently. ++ * ++ * Attribute values are aligned to the end of the block, stored in ++ * no specific order. They are also padded to EXT2_XATTR_PAD byte ++ * boundaries. No additional gaps are left between them. ++ * ++ * Locking strategy ++ * ---------------- ++ * The VFS already holds the BKL and the inode->i_sem semaphore when any of ++ * the xattr inode operations are called, so we are guaranteed that only one ++ * processes accesses extended attributes of an inode at any time. ++ * ++ * For writing we also grab the ext2_xattr_sem semaphore. This ensures that ++ * only a single process is modifying an extended attribute block, even ++ * if the block is shared among inodes. ++ * ++ * Note for porting to 2.5 ++ * ----------------------- ++ * The BKL will no longer be held in the xattr inode operations. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++/* These symbols may be needed by a module. */ ++EXPORT_SYMBOL(ext2_xattr_register); ++EXPORT_SYMBOL(ext2_xattr_unregister); ++EXPORT_SYMBOL(ext2_xattr_get); ++EXPORT_SYMBOL(ext2_xattr_list); ++EXPORT_SYMBOL(ext2_xattr_set); ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++# define mark_buffer_dirty(bh) mark_buffer_dirty(bh, 1) ++#endif ++ ++#define HDR(bh) ((struct ext2_xattr_header *)((bh)->b_data)) ++#define ENTRY(ptr) ((struct ext2_xattr_entry *)(ptr)) ++#define FIRST_ENTRY(bh) ENTRY(HDR(bh)+1) ++#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) ++ ++#ifdef EXT2_XATTR_DEBUG ++# define ea_idebug(inode, f...) do { \ ++ printk(KERN_DEBUG "inode %s:%ld: ", \ ++ kdevname(inode->i_dev), inode->i_ino); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++# define ea_bdebug(bh, f...) do { \ ++ printk(KERN_DEBUG "block %s:%ld: ", \ ++ kdevname(bh->b_dev), bh->b_blocknr); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++#else ++# define ea_idebug(f...) ++# define ea_bdebug(f...) ++#endif ++ ++static int ext2_xattr_set2(struct inode *, struct buffer_head *, ++ struct ext2_xattr_header *); ++ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ ++static int ext2_xattr_cache_insert(struct buffer_head *); ++static struct buffer_head *ext2_xattr_cache_find(struct inode *, ++ struct ext2_xattr_header *); ++static void ext2_xattr_cache_remove(struct buffer_head *); ++static void ext2_xattr_rehash(struct ext2_xattr_header *, ++ struct ext2_xattr_entry *); ++ ++static struct mb_cache *ext2_xattr_cache; ++ ++#else ++# define ext2_xattr_cache_insert(bh) 0 ++# define ext2_xattr_cache_find(inode, header) NULL ++# define ext2_xattr_cache_remove(bh) while(0) {} ++# define ext2_xattr_rehash(header, entry) while(0) {} ++#endif ++ ++/* ++ * If a file system does not share extended attributes among inodes, ++ * we should not need the ext2_xattr_sem semaphore. However, the ++ * filesystem may still contain shared blocks, so we always take ++ * the lock. ++ */ ++ ++DECLARE_MUTEX(ext2_xattr_sem); ++ ++static inline int ++ext2_xattr_new_block(struct inode *inode, int * errp, int force) ++{ ++ struct super_block *sb = inode->i_sb; ++ int goal = le32_to_cpu(EXT2_SB(sb)->s_es->s_first_data_block) + ++ EXT2_I(inode)->i_block_group * EXT2_BLOCKS_PER_GROUP(sb); ++ ++ /* How can we enforce the allocation? */ ++ int block = ext2_new_block(inode, goal, 0, 0, errp); ++#ifdef OLD_QUOTAS ++ if (!*errp) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#endif ++ return block; ++} ++ ++static inline int ++ext2_xattr_quota_alloc(struct inode *inode, int force) ++{ ++ /* How can we enforce the allocation? */ ++#ifdef OLD_QUOTAS ++ int error = DQUOT_ALLOC_BLOCK(inode->i_sb, inode, 1); ++ if (!error) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#else ++ int error = DQUOT_ALLOC_BLOCK(inode, 1); ++#endif ++ return error; ++} ++ ++#ifdef OLD_QUOTAS ++ ++static inline void ++ext2_xattr_quota_free(struct inode *inode) ++{ ++ DQUOT_FREE_BLOCK(inode->i_sb, inode, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++static inline void ++ext2_xattr_free_block(struct inode * inode, unsigned long block) ++{ ++ ext2_free_blocks(inode, block, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++#else ++# define ext2_xattr_quota_free(inode) \ ++ DQUOT_FREE_BLOCK(inode, 1) ++# define ext2_xattr_free_block(inode, block) \ ++ ext2_free_blocks(inode, block, 1) ++#endif ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18) ++ ++static inline struct buffer_head * ++sb_bread(struct super_block *sb, int block) ++{ ++ return bread(sb->s_dev, block, sb->s_blocksize); ++} ++ ++static inline struct buffer_head * ++sb_getblk(struct super_block *sb, int block) ++{ ++ return getblk(sb->s_dev, block, sb->s_blocksize); ++} ++ ++#endif ++ ++struct ext2_xattr_handler *ext2_xattr_handlers[EXT2_XATTR_INDEX_MAX]; ++rwlock_t ext2_handler_lock = RW_LOCK_UNLOCKED; ++ ++int ++ext2_xattr_register(int name_index, struct ext2_xattr_handler *handler) ++{ ++ int error = -EINVAL; ++ ++ if (name_index > 0 && name_index <= EXT2_XATTR_INDEX_MAX) { ++ write_lock(&ext2_handler_lock); ++ if (!ext2_xattr_handlers[name_index-1]) { ++ ext2_xattr_handlers[name_index-1] = handler; ++ error = 0; ++ } ++ write_unlock(&ext2_handler_lock); ++ } ++ return error; ++} ++ ++void ++ext2_xattr_unregister(int name_index, struct ext2_xattr_handler *handler) ++{ ++ if (name_index > 0 || name_index <= EXT2_XATTR_INDEX_MAX) { ++ write_lock(&ext2_handler_lock); ++ ext2_xattr_handlers[name_index-1] = NULL; ++ write_unlock(&ext2_handler_lock); ++ } ++} ++ ++static inline const char * ++strcmp_prefix(const char *a, const char *a_prefix) ++{ ++ while (*a_prefix && *a == *a_prefix) { ++ a++; ++ a_prefix++; ++ } ++ return *a_prefix ? NULL : a; ++} ++ ++/* ++ * Decode the extended attribute name, and translate it into ++ * the name_index and name suffix. ++ */ ++static struct ext2_xattr_handler * ++ext2_xattr_resolve_name(const char **name) ++{ ++ struct ext2_xattr_handler *handler = NULL; ++ int i; ++ ++ if (!*name) ++ return NULL; ++ read_lock(&ext2_handler_lock); ++ for (i=0; iprefix); ++ if (n) { ++ handler = ext2_xattr_handlers[i]; ++ *name = n; ++ break; ++ } ++ } ++ } ++ read_unlock(&ext2_handler_lock); ++ return handler; ++} ++ ++static inline struct ext2_xattr_handler * ++ext2_xattr_handler(int name_index) ++{ ++ struct ext2_xattr_handler *handler = NULL; ++ if (name_index > 0 && name_index <= EXT2_XATTR_INDEX_MAX) { ++ read_lock(&ext2_handler_lock); ++ handler = ext2_xattr_handlers[name_index-1]; ++ read_unlock(&ext2_handler_lock); ++ } ++ return handler; ++} ++ ++/* ++ * Inode operation getxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext2_getxattr(struct dentry *dentry, const char *name, ++ void *buffer, size_t size) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->get(inode, name, buffer, size); ++} ++ ++/* ++ * Inode operation listxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext2_listxattr(struct dentry *dentry, char *buffer, size_t size) ++{ ++ return ext2_xattr_list(dentry->d_inode, buffer, size); ++} ++ ++/* ++ * Inode operation setxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext2_setxattr(struct dentry *dentry, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ if (size == 0) ++ value = ""; /* empty EA, do not remove */ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, value, size, flags); ++} ++ ++/* ++ * Inode operation removexattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext2_removexattr(struct dentry *dentry, const char *name) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, NULL, 0, XATTR_REPLACE); ++} ++ ++/* ++ * ext2_xattr_get() ++ * ++ * Copy an extended attribute into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext2_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_entry *entry; ++ unsigned int block, size; ++ char *end; ++ int name_len, error; ++ ++ ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", ++ name_index, name, buffer, (long)buffer_size); ++ ++ if (name == NULL) ++ return -EINVAL; ++ if (!EXT2_I(inode)->i_file_acl) ++ return -ENOATTR; ++ block = EXT2_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(inode->i_sb, "ext2_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ ++ error = -ERANGE; ++ if (name_len > 255) ++ goto cleanup; ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) ++ goto found; ++ entry = next; ++ } ++ /* Check the remaining name entries */ ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ entry = next; ++ } ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ error = -ENOATTR; ++ goto cleanup; ++found: ++ /* check the buffer size */ ++ if (entry->e_value_block != 0) ++ goto bad_block; ++ size = le32_to_cpu(entry->e_value_size); ++ if (size > inode->i_sb->s_blocksize || ++ le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize) ++ goto bad_block; ++ ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (buffer) { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ /* return value of attribute */ ++ memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), ++ size); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext2_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext2_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_entry *entry; ++ unsigned int block, size = 0; ++ char *buf, *end; ++ int error; ++ ++ ea_idebug(inode, "buffer=%p, buffer_size=%ld", ++ buffer, (long)buffer_size); ++ ++ if (!EXT2_I(inode)->i_file_acl) ++ return 0; ++ block = EXT2_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(inode->i_sb, "ext2_xattr_list", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* compute the size required for the list of attribute names */ ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT2_XATTR_NEXT(entry)) { ++ struct ext2_xattr_handler *handler; ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ ++ handler = ext2_xattr_handler(entry->e_name_index); ++ if (handler) ++ size += handler->list(NULL, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (!buffer) { ++ error = size; ++ goto cleanup; ++ } else { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ /* list the attribute names */ ++ buf = buffer; ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT2_XATTR_NEXT(entry)) { ++ struct ext2_xattr_handler *handler; ++ ++ handler = ext2_xattr_handler(entry->e_name_index); ++ if (handler) ++ buf += handler->list(buf, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * If the EXT2_FEATURE_COMPAT_EXT_ATTR feature of this file system is ++ * not set, set it. ++ */ ++static void ext2_xattr_update_super_block(struct super_block *sb) ++{ ++ if (EXT2_HAS_COMPAT_FEATURE(sb, EXT2_FEATURE_COMPAT_EXT_ATTR)) ++ return; ++ ++ lock_super(sb); ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++ EXT2_SB(sb)->s_feature_compat |= EXT2_FEATURE_COMPAT_EXT_ATTR; ++#endif ++ EXT2_SB(sb)->s_es->s_feature_compat |= ++ cpu_to_le32(EXT2_FEATURE_COMPAT_EXT_ATTR); ++ sb->s_dirt = 1; ++ mark_buffer_dirty(EXT2_SB(sb)->s_sbh); ++ unlock_super(sb); ++} ++ ++/* ++ * ext2_xattr_set() ++ * ++ * Create, replace or remove an extended attribute for this inode. Buffer ++ * is NULL to remove an existing extended attribute, and non-NULL to ++ * either replace an existing extended attribute, or create a new extended ++ * attribute. The flags XATTR_REPLACE and XATTR_CREATE ++ * specify that an extended attribute must exist and must not exist ++ * previous to the call, respectively. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++int ++ext2_xattr_set(struct inode *inode, int name_index, const char *name, ++ const void *value, size_t value_len, int flags) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_header *header = NULL; ++ struct ext2_xattr_entry *here, *last; ++ unsigned int name_len; ++ int block = EXT2_I(inode)->i_file_acl; ++ int min_offs = sb->s_blocksize, not_found = 1, free, error; ++ char *end; ++ ++ /* ++ * header -- Points either into bh, or to a temporarily ++ * allocated buffer. ++ * here -- The named entry found, or the place for inserting, within ++ * the block pointed to by header. ++ * last -- Points right after the last named entry within the block ++ * pointed to by header. ++ * min_offs -- The offset of the first value (values are aligned ++ * towards the end of the block). ++ * end -- Points right after the block pointed to by header. ++ */ ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > sb->s_blocksize) ++ return -ERANGE; ++ down(&ext2_xattr_sem); ++ ++ if (block) { ++ /* The inode already has an extended attribute block. */ ++ ++ bh = sb_bread(sb, block); ++ error = -EIO; ++ if (!bh) ++ goto cleanup; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), ++ le32_to_cpu(HDR(bh)->h_refcount)); ++ header = HDR(bh); ++ end = bh->b_data + bh->b_size; ++ if (header->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ header->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(sb, "ext2_xattr_set", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* Find the named attribute. */ ++ here = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(here)) { ++ struct ext2_xattr_entry *next = EXT2_XATTR_NEXT(here); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!here->e_value_block && here->e_value_size) { ++ int offs = le16_to_cpu(here->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ not_found = name_index - here->e_name_index; ++ if (!not_found) ++ not_found = name_len - here->e_name_len; ++ if (!not_found) ++ not_found = memcmp(name, here->e_name,name_len); ++ if (not_found <= 0) ++ break; ++ here = next; ++ } ++ last = here; ++ /* We still need to compute min_offs and last. */ ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext2_xattr_entry *next = EXT2_XATTR_NEXT(last); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!last->e_value_block && last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ last = next; ++ } ++ ++ /* Check whether we have enough space left. */ ++ free = min_offs - ((char*)last - (char*)header) - sizeof(__u32); ++ } else { ++ /* We will use a new extended attribute block. */ ++ free = sb->s_blocksize - ++ sizeof(struct ext2_xattr_header) - sizeof(__u32); ++ here = last = NULL; /* avoid gcc uninitialized warning. */ ++ } ++ ++ if (not_found) { ++ /* Request to remove a nonexistent attribute? */ ++ error = -ENOATTR; ++ if (flags & XATTR_REPLACE) ++ goto cleanup; ++ error = 0; ++ if (value == NULL) ++ goto cleanup; ++ else ++ free -= EXT2_XATTR_LEN(name_len); ++ } else { ++ /* Request to create an existing attribute? */ ++ error = -EEXIST; ++ if (flags & XATTR_CREATE) ++ goto cleanup; ++ if (!here->e_value_block && here->e_value_size) { ++ unsigned int size = le32_to_cpu(here->e_value_size); ++ ++ if (le16_to_cpu(here->e_value_offs) + size > ++ sb->s_blocksize || size > sb->s_blocksize) ++ goto bad_block; ++ free += EXT2_XATTR_SIZE(size); ++ } ++ } ++ free -= EXT2_XATTR_SIZE(value_len); ++ error = -ENOSPC; ++ if (free < 0) ++ goto cleanup; ++ ++ /* Here we know that we can set the new attribute. */ ++ ++ if (header) { ++ if (header->h_refcount == cpu_to_le32(1)) { ++ ea_bdebug(bh, "modifying in-place"); ++ ext2_xattr_cache_remove(bh); ++ } else { ++ int offset; ++ ++ ea_bdebug(bh, "cloning"); ++ header = kmalloc(bh->b_size, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memcpy(header, HDR(bh), bh->b_size); ++ header->h_refcount = cpu_to_le32(1); ++ offset = (char *)header - bh->b_data; ++ here = ENTRY((char *)here + offset); ++ last = ENTRY((char *)last + offset); ++ } ++ } else { ++ /* Allocate a buffer where we construct the new block. */ ++ header = kmalloc(sb->s_blocksize, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memset(header, 0, sb->s_blocksize); ++ end = (char *)header + sb->s_blocksize; ++ header->h_magic = cpu_to_le32(EXT2_XATTR_MAGIC); ++ header->h_blocks = header->h_refcount = cpu_to_le32(1); ++ last = here = ENTRY(header+1); ++ } ++ ++ if (not_found) { ++ /* Insert the new name. */ ++ int size = EXT2_XATTR_LEN(name_len); ++ int rest = (char *)last - (char *)here; ++ memmove((char *)here + size, here, rest); ++ memset(here, 0, size); ++ here->e_name_index = name_index; ++ here->e_name_len = name_len; ++ memcpy(here->e_name, name, name_len); ++ } else { ++ /* Remove the old value. */ ++ if (!here->e_value_block && here->e_value_size) { ++ char *first_val = (char *)header + min_offs; ++ int offs = le16_to_cpu(here->e_value_offs); ++ char *val = (char *)header + offs; ++ size_t size = EXT2_XATTR_SIZE( ++ le32_to_cpu(here->e_value_size)); ++ memmove(first_val + size, first_val, val - first_val); ++ memset(first_val, 0, size); ++ here->e_value_offs = 0; ++ min_offs += size; ++ ++ /* Adjust all value offsets. */ ++ last = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(last)) { ++ int o = le16_to_cpu(last->e_value_offs); ++ if (!last->e_value_block && o < offs) ++ last->e_value_offs = ++ cpu_to_le16(o + size); ++ last = EXT2_XATTR_NEXT(last); ++ } ++ } ++ if (value == NULL) { ++ /* Remove this attribute. */ ++ if (EXT2_XATTR_NEXT(ENTRY(header+1)) == last) { ++ /* This block is now empty. */ ++ error = ext2_xattr_set2(inode, bh, NULL); ++ goto cleanup; ++ } else { ++ /* Remove the old name. */ ++ int size = EXT2_XATTR_LEN(name_len); ++ last = ENTRY((char *)last - size); ++ memmove(here, (char*)here + size, ++ (char*)last - (char*)here); ++ memset(last, 0, size); ++ } ++ } ++ } ++ ++ if (value != NULL) { ++ /* Insert the new value. */ ++ here->e_value_size = cpu_to_le32(value_len); ++ if (value_len) { ++ size_t size = EXT2_XATTR_SIZE(value_len); ++ char *val = (char *)header + min_offs - size; ++ here->e_value_offs = ++ cpu_to_le16((char *)val - (char *)header); ++ memset(val + size - EXT2_XATTR_PAD, 0, ++ EXT2_XATTR_PAD); /* Clear the pad bytes. */ ++ memcpy(val, value, value_len); ++ } ++ } ++ ext2_xattr_rehash(header, here); ++ ++ error = ext2_xattr_set2(inode, bh, header); ++ ++cleanup: ++ brelse(bh); ++ if (!(bh && header == HDR(bh))) ++ kfree(header); ++ up(&ext2_xattr_sem); ++ ++ return error; ++} ++ ++/* ++ * Second half of ext2_xattr_set(): Update the file system. ++ */ ++static int ++ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh, ++ struct ext2_xattr_header *header) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *new_bh = NULL; ++ int error; ++ ++ if (header) { ++ new_bh = ext2_xattr_cache_find(inode, header); ++ if (new_bh) { ++ /* ++ * We found an identical block in the cache. ++ * The old block will be released after updating ++ * the inode. ++ */ ++ ea_bdebug(old_bh, "reusing block %ld", ++ new_bh->b_blocknr); ++ ++ error = -EDQUOT; ++ if (ext2_xattr_quota_alloc(inode, 1)) ++ goto cleanup; ++ ++ HDR(new_bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(new_bh)->h_refcount) + 1); ++ ea_bdebug(new_bh, "refcount now=%d", ++ le32_to_cpu(HDR(new_bh)->h_refcount)); ++ } else if (old_bh && header == HDR(old_bh)) { ++ /* Keep this block. */ ++ new_bh = old_bh; ++ ext2_xattr_cache_insert(new_bh); ++ } else { ++ /* We need to allocate a new block */ ++ int force = EXT2_I(inode)->i_file_acl != 0; ++ int block = ext2_xattr_new_block(inode, &error, force); ++ if (error) ++ goto cleanup; ++ ea_idebug(inode, "creating block %d", block); ++ ++ new_bh = sb_getblk(sb, block); ++ if (!new_bh) { ++ ext2_xattr_free_block(inode, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ lock_buffer(new_bh); ++ memcpy(new_bh->b_data, header, new_bh->b_size); ++ mark_buffer_uptodate(new_bh, 1); ++ unlock_buffer(new_bh); ++ ext2_xattr_cache_insert(new_bh); ++ ++ ext2_xattr_update_super_block(sb); ++ } ++ mark_buffer_dirty(new_bh); ++ if (IS_SYNC(inode)) { ++ ll_rw_block(WRITE, 1, &new_bh); ++ wait_on_buffer(new_bh); ++ error = -EIO; ++ if (buffer_req(new_bh) && !buffer_uptodate(new_bh)) ++ goto cleanup; ++ } ++ } ++ ++ /* Update the inode. */ ++ EXT2_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; ++ inode->i_ctime = CURRENT_TIME; ++ if (IS_SYNC(inode)) { ++ error = ext2_sync_inode (inode); ++ if (error) ++ goto cleanup; ++ } else ++ mark_inode_dirty(inode); ++ ++ error = 0; ++ if (old_bh && old_bh != new_bh) { ++ /* ++ * If there was an old block, and we are not still using it, ++ * we now release the old block. ++ */ ++ unsigned int refcount = le32_to_cpu(HDR(old_bh)->h_refcount); ++ ++ if (refcount == 1) { ++ /* Free the old block. */ ++ ea_bdebug(old_bh, "freeing"); ++ ext2_xattr_free_block(inode, old_bh->b_blocknr); ++ mark_buffer_clean(old_bh); ++ } else { ++ /* Decrement the refcount only. */ ++ refcount--; ++ HDR(old_bh)->h_refcount = cpu_to_le32(refcount); ++ ext2_xattr_quota_free(inode); ++ mark_buffer_dirty(old_bh); ++ ea_bdebug(old_bh, "refcount now=%d", refcount); ++ } ++ } ++ ++cleanup: ++ if (old_bh != new_bh) ++ brelse(new_bh); ++ ++ return error; ++} ++ ++/* ++ * ext2_xattr_delete_inode() ++ * ++ * Free extended attribute resources associated with this inode. This ++ * is called immediately before an inode is freed. ++ */ ++void ++ext2_xattr_delete_inode(struct inode *inode) ++{ ++ struct buffer_head *bh; ++ unsigned int block = EXT2_I(inode)->i_file_acl; ++ ++ if (!block) ++ return; ++ down(&ext2_xattr_sem); ++ ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) { ++ ext2_error(inode->i_sb, "ext2_xattr_delete_inode", ++ "inode %ld: block %d read error", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "b_count=%d", atomic_read(&(bh->b_count))); ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++ ext2_error(inode->i_sb, "ext2_xattr_delete_inode", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ if (HDR(bh)->h_refcount == cpu_to_le32(1)) { ++ ext2_xattr_cache_remove(bh); ++ ext2_xattr_free_block(inode, block); ++ bforget(bh); ++ bh = NULL; ++ } else { ++ HDR(bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ mark_buffer_dirty(bh); ++ if (IS_SYNC(inode)) { ++ ll_rw_block(WRITE, 1, &bh); ++ wait_on_buffer(bh); ++ } ++ ext2_xattr_quota_free(inode); ++ } ++ EXT2_I(inode)->i_file_acl = 0; ++ ++cleanup: ++ brelse(bh); ++ up(&ext2_xattr_sem); ++} ++ ++/* ++ * ext2_xattr_put_super() ++ * ++ * This is called when a file system is unmounted. ++ */ ++void ++ext2_xattr_put_super(struct super_block *sb) ++{ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ mb_cache_shrink(ext2_xattr_cache, sb->s_dev); ++#endif ++} ++ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ ++/* ++ * ext2_xattr_cache_insert() ++ * ++ * Create a new entry in the extended attribute cache, and insert ++ * it unless such an entry is already in the cache. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++static int ++ext2_xattr_cache_insert(struct buffer_head *bh) ++{ ++ __u32 hash = le32_to_cpu(HDR(bh)->h_hash); ++ struct mb_cache_entry *ce; ++ int error; ++ ++ ce = mb_cache_entry_alloc(ext2_xattr_cache); ++ if (!ce) ++ return -ENOMEM; ++ error = mb_cache_entry_insert(ce, bh->b_dev, bh->b_blocknr, &hash); ++ if (error) { ++ mb_cache_entry_free(ce); ++ if (error == -EBUSY) { ++ ea_bdebug(bh, "already in cache (%d cache entries)", ++ atomic_read(&ext2_xattr_cache->c_entry_count)); ++ error = 0; ++ } ++ } else { ++ ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash, ++ atomic_read(&ext2_xattr_cache->c_entry_count)); ++ mb_cache_entry_release(ce); ++ } ++ return error; ++} ++ ++/* ++ * ext2_xattr_cmp() ++ * ++ * Compare two extended attribute blocks for equality. ++ * ++ * Returns 0 if the blocks are equal, 1 if they differ, and ++ * a negative error number on errors. ++ */ ++static int ++ext2_xattr_cmp(struct ext2_xattr_header *header1, ++ struct ext2_xattr_header *header2) ++{ ++ struct ext2_xattr_entry *entry1, *entry2; ++ ++ entry1 = ENTRY(header1+1); ++ entry2 = ENTRY(header2+1); ++ while (!IS_LAST_ENTRY(entry1)) { ++ if (IS_LAST_ENTRY(entry2)) ++ return 1; ++ if (entry1->e_hash != entry2->e_hash || ++ entry1->e_name_len != entry2->e_name_len || ++ entry1->e_value_size != entry2->e_value_size || ++ memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) ++ return 1; ++ if (entry1->e_value_block != 0 || entry2->e_value_block != 0) ++ return -EIO; ++ if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), ++ (char *)header2 + le16_to_cpu(entry2->e_value_offs), ++ le32_to_cpu(entry1->e_value_size))) ++ return 1; ++ ++ entry1 = EXT2_XATTR_NEXT(entry1); ++ entry2 = EXT2_XATTR_NEXT(entry2); ++ } ++ if (!IS_LAST_ENTRY(entry2)) ++ return 1; ++ return 0; ++} ++ ++/* ++ * ext2_xattr_cache_find() ++ * ++ * Find an identical extended attribute block. ++ * ++ * Returns a pointer to the block found, or NULL if such a block was ++ * not found or an error occurred. ++ */ ++static struct buffer_head * ++ext2_xattr_cache_find(struct inode *inode, struct ext2_xattr_header *header) ++{ ++ __u32 hash = le32_to_cpu(header->h_hash); ++ struct mb_cache_entry *ce; ++ ++ if (!header->h_hash) ++ return NULL; /* never share */ ++ ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); ++ ce = mb_cache_entry_find_first(ext2_xattr_cache, 0, inode->i_dev, hash); ++ while (ce) { ++ struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block); ++ ++ if (!bh) { ++ ext2_error(inode->i_sb, "ext2_xattr_cache_find", ++ "inode %ld: block %ld read error", ++ inode->i_ino, ce->e_block); ++ } else if (le32_to_cpu(HDR(bh)->h_refcount) > ++ EXT2_XATTR_REFCOUNT_MAX) { ++ ea_idebug(inode, "block %ld refcount %d>%d",ce->e_block, ++ le32_to_cpu(HDR(bh)->h_refcount), ++ EXT2_XATTR_REFCOUNT_MAX); ++ } else if (!ext2_xattr_cmp(header, HDR(bh))) { ++ ea_bdebug(bh, "b_count=%d",atomic_read(&(bh->b_count))); ++ mb_cache_entry_release(ce); ++ return bh; ++ } ++ brelse(bh); ++ ce = mb_cache_entry_find_next(ce, 0, inode->i_dev, hash); ++ } ++ return NULL; ++} ++ ++/* ++ * ext2_xattr_cache_remove() ++ * ++ * Remove the cache entry of a block from the cache. Called when a ++ * block becomes invalid. ++ */ ++static void ++ext2_xattr_cache_remove(struct buffer_head *bh) ++{ ++ struct mb_cache_entry *ce; ++ ++ ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_dev, bh->b_blocknr); ++ if (ce) { ++ ea_bdebug(bh, "removing (%d cache entries remaining)", ++ atomic_read(&ext2_xattr_cache->c_entry_count)-1); ++ mb_cache_entry_free(ce); ++ } else ++ ea_bdebug(bh, "no cache entry"); ++} ++ ++#define NAME_HASH_SHIFT 5 ++#define VALUE_HASH_SHIFT 16 ++ ++/* ++ * ext2_xattr_hash_entry() ++ * ++ * Compute the hash of an extended attribute. ++ */ ++static inline void ext2_xattr_hash_entry(struct ext2_xattr_header *header, ++ struct ext2_xattr_entry *entry) ++{ ++ __u32 hash = 0; ++ char *name = entry->e_name; ++ int n; ++ ++ for (n=0; n < entry->e_name_len; n++) { ++ hash = (hash << NAME_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ ++ *name++; ++ } ++ ++ if (entry->e_value_block == 0 && entry->e_value_size != 0) { ++ __u32 *value = (__u32 *)((char *)header + ++ le16_to_cpu(entry->e_value_offs)); ++ for (n = (le32_to_cpu(entry->e_value_size) + ++ EXT2_XATTR_ROUND) >> EXT2_XATTR_PAD_BITS; n; n--) { ++ hash = (hash << VALUE_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ ++ le32_to_cpu(*value++); ++ } ++ } ++ entry->e_hash = cpu_to_le32(hash); ++} ++ ++#undef NAME_HASH_SHIFT ++#undef VALUE_HASH_SHIFT ++ ++#define BLOCK_HASH_SHIFT 16 ++ ++/* ++ * ext2_xattr_rehash() ++ * ++ * Re-compute the extended attribute hash value after an entry has changed. ++ */ ++static void ext2_xattr_rehash(struct ext2_xattr_header *header, ++ struct ext2_xattr_entry *entry) ++{ ++ struct ext2_xattr_entry *here; ++ __u32 hash = 0; ++ ++ ext2_xattr_hash_entry(header, entry); ++ here = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(here)) { ++ if (!here->e_hash) { ++ /* Block is not shared if an entry's hash value == 0 */ ++ hash = 0; ++ break; ++ } ++ hash = (hash << BLOCK_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ ++ le32_to_cpu(here->e_hash); ++ here = EXT2_XATTR_NEXT(here); ++ } ++ header->h_hash = cpu_to_le32(hash); ++} ++ ++#undef BLOCK_HASH_SHIFT ++ ++int __init ++init_ext2_xattr(void) ++{ ++ ext2_xattr_cache = mb_cache_create("ext2_xattr", NULL, ++ sizeof(struct mb_cache_entry) + ++ sizeof(struct mb_cache_entry_index), 1, 61); ++ if (!ext2_xattr_cache) ++ return -ENOMEM; ++ ++ return 0; ++} ++ ++void ++exit_ext2_xattr(void) ++{ ++ mb_cache_destroy(ext2_xattr_cache); ++} ++ ++#else /* CONFIG_EXT2_FS_XATTR_SHARING */ ++ ++int __init ++init_ext2_xattr(void) ++{ ++ return 0; ++} ++ ++void ++exit_ext2_xattr(void) ++{ ++} ++ ++#endif /* CONFIG_EXT2_FS_XATTR_SHARING */ +Index: linux-2.4.19-pre1/fs/ext2/xattr_user.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext2/xattr_user.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext2/xattr_user.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,103 @@ ++/* ++ * linux/fs/ext2/xattr_user.c ++ * Handler for extended user attributes. ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++ ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++# include ++#endif ++ ++#define XATTR_USER_PREFIX "user." ++ ++static size_t ++ext2_xattr_user_list(char *list, struct inode *inode, ++ const char *name, int name_len) ++{ ++ const int prefix_len = sizeof(XATTR_USER_PREFIX)-1; ++ ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return 0; ++ ++ if (list) { ++ memcpy(list, XATTR_USER_PREFIX, prefix_len); ++ memcpy(list+prefix_len, name, name_len); ++ list[prefix_len + name_len] = '\0'; ++ } ++ return prefix_len + name_len + 1; ++} ++ ++static int ++ext2_xattr_user_get(struct inode *inode, const char *name, ++ void *buffer, size_t size) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++ error = ext2_permission_locked(inode, MAY_READ); ++#else ++ error = permission(inode, MAY_READ); ++#endif ++ if (error) ++ return error; ++ ++ return ext2_xattr_get(inode, EXT2_XATTR_INDEX_USER, name, ++ buffer, size); ++} ++ ++static int ++ext2_xattr_user_set(struct inode *inode, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++ if ( !S_ISREG(inode->i_mode) && ++ (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX)) ++ return -EPERM; ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++ error = ext2_permission_locked(inode, MAY_WRITE); ++#else ++ error = permission(inode, MAY_WRITE); ++#endif ++ if (error) ++ return error; ++ ++ return ext2_xattr_set(inode, EXT2_XATTR_INDEX_USER, name, ++ value, size, flags); ++} ++ ++struct ext2_xattr_handler ext2_xattr_user_handler = { ++ prefix: XATTR_USER_PREFIX, ++ list: ext2_xattr_user_list, ++ get: ext2_xattr_user_get, ++ set: ext2_xattr_user_set, ++}; ++ ++int __init ++init_ext2_xattr_user(void) ++{ ++ return ext2_xattr_register(EXT2_XATTR_INDEX_USER, ++ &ext2_xattr_user_handler); ++} ++ ++void ++exit_ext2_xattr_user(void) ++{ ++ ext2_xattr_unregister(EXT2_XATTR_INDEX_USER, ++ &ext2_xattr_user_handler); ++} +Index: linux-2.4.19-pre1/fs/ext3/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/Makefile 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/Makefile 2003-11-21 03:51:05.000000000 +0300 +@@ -1,5 +1,5 @@ + # +-# Makefile for the linux ext2-filesystem routines. ++# Makefile for the linux ext3-filesystem routines. + # + # Note! Dependencies are done automagically by 'make dep', which also + # removes any old dependencies. DON'T put your own dependencies here +@@ -9,10 +9,14 @@ + + O_TARGET := ext3.o + +-export-objs := super.o inode.o ++export-objs := ext3-exports.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o hash.o ++ ioctl.o namei.o super.o symlink.o hash.o ext3-exports.o + obj-m := $(O_TARGET) + ++export-objs += xattr.o ++obj-$(CONFIG_EXT3_FS_XATTR) += xattr.o ++obj-$(CONFIG_EXT3_FS_XATTR_USER) += xattr_user.o ++ + include $(TOPDIR)/Rules.make +Index: linux-2.4.19-pre1/fs/ext3/file.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/file.c 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/file.c 2003-11-21 03:51:05.000000000 +0300 +@@ -23,6 +23,7 @@ + #include + #include + #include ++#include + #include + #include + +@@ -93,5 +94,9 @@ + struct inode_operations ext3_file_inode_operations = { + truncate: ext3_truncate, /* BKL held */ + setattr: ext3_setattr, /* BKL held */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; + +Index: linux-2.4.19-pre1/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/ialloc.c 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/ialloc.c 2003-11-21 03:51:05.000000000 +0300 +@@ -17,6 +17,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -216,6 +217,7 @@ + * as writing the quota to disk may need the lock as well. + */ + DQUOT_INIT(inode); ++ ext3_xattr_delete_inode(handle, inode); + DQUOT_FREE_INODE(inode); + DQUOT_DROP(inode); + +Index: linux-2.4.19-pre1/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/inode.c 2003-11-20 19:01:36.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/inode.c 2003-11-21 03:51:05.000000000 +0300 +@@ -39,6 +39,18 @@ + */ + #undef SEARCH_FROM_ZERO + ++/* ++ * Test whether an inode is a fast symlink. ++ */ ++static inline int ext3_inode_is_fast_symlink(struct inode *inode) ++{ ++ int ea_blocks = inode->u.ext3_i.i_file_acl ? ++ (inode->i_sb->s_blocksize >> 9) : 0; ++ ++ return (S_ISLNK(inode->i_mode) && ++ inode->i_blocks - ea_blocks == 0); ++} ++ + /* The ext3 forget function must perform a revoke if we are freeing data + * which has been journaled. Metadata (eg. indirect blocks) must be + * revoked in all cases. +@@ -48,7 +60,7 @@ + * still needs to be revoked. + */ + +-static int ext3_forget(handle_t *handle, int is_metadata, ++int ext3_forget(handle_t *handle, int is_metadata, + struct inode *inode, struct buffer_head *bh, + int blocknr) + { +@@ -164,9 +176,7 @@ + { + handle_t *handle; + +- if (is_bad_inode(inode) || +- inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) ++ if (is_bad_inode(inode)) + goto no_delete; + + lock_kernel(); +@@ -1845,6 +1855,8 @@ + if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode))) + return; ++ if (ext3_inode_is_fast_symlink(inode)) ++ return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + +@@ -1992,8 +2004,6 @@ + struct ext3_group_desc * gdp; + + if ((inode->i_ino != EXT3_ROOT_INO && +- inode->i_ino != EXT3_ACL_IDX_INO && +- inode->i_ino != EXT3_ACL_DATA_INO && + inode->i_ino != EXT3_JOURNAL_INO && + inode->i_ino < EXT3_FIRST_INO(inode->i_sb)) || + inode->i_ino > le32_to_cpu( +@@ -2120,10 +2130,7 @@ + + brelse (iloc.bh); + +- if (inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) +- /* Nothing to do */ ; +- else if (S_ISREG(inode->i_mode)) { ++ if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +@@ -2131,15 +2138,17 @@ + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; + } else if (S_ISLNK(inode->i_mode)) { +- if (!inode->i_blocks) ++ if (ext3_inode_is_fast_symlink(inode)) + inode->i_op = &ext3_fast_symlink_inode_operations; + else { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext3_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + } +- } else ++ } else { ++ inode->i_op = &ext3_special_inode_operations; + init_special_inode(inode, inode->i_mode, + le32_to_cpu(iloc.raw_inode->i_block[0])); ++ } + /* inode->i_attr_flags = 0; unused */ + if (inode->u.ext3_i.i_flags & EXT3_SYNC_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_SYNCRONOUS; unused */ +Index: linux-2.4.19-pre1/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/namei.c 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/namei.c 2003-11-21 03:51:05.000000000 +0300 +@@ -29,6 +29,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -1612,7 +1613,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR); ++ inode = ext3_new_inode (handle, dir, S_IFDIR | mode); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -1620,7 +1621,6 @@ + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; + inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; +- inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { + inode->i_nlink--; /* is this nlink == 0? */ +@@ -1647,9 +1647,6 @@ + BUFFER_TRACE(dir_block, "call ext3_journal_dirty_metadata"); + ext3_journal_dirty_metadata(handle, dir_block); + brelse (dir_block); +- inode->i_mode = S_IFDIR | mode; +- if (dir->i_mode & S_ISGID) +- inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); + if (err) { +@@ -2018,7 +2015,7 @@ + goto out_stop; + + if (l > sizeof (EXT3_I(inode)->i_data)) { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext3_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* + * block_symlink() calls back into ext3_prepare/commit_write. +@@ -2245,4 +2242,16 @@ + rmdir: ext3_rmdir, /* BKL held */ + mknod: ext3_mknod, /* BKL held */ + rename: ext3_rename, /* BKL held */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; ++ ++struct inode_operations ext3_special_inode_operations = { ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ ++}; ++ +Index: linux-2.4.19-pre1/fs/ext3/super.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/super.c 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/super.c 2003-11-21 03:51:05.000000000 +0300 +@@ -24,6 +24,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -404,6 +405,7 @@ + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { + EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); +@@ -499,6 +501,7 @@ + int is_remount) + { + unsigned long *mount_options = &sbi->s_mount_opt; ++ + uid_t *resuid = &sbi->s_resuid; + gid_t *resgid = &sbi->s_resgid; + char * this_char; +@@ -511,6 +514,13 @@ + this_char = strtok (NULL, ",")) { + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; ++#ifdef CONFIG_EXT3_FS_XATTR_USER ++ if (!strcmp (this_char, "user_xattr")) ++ set_opt (*mount_options, XATTR_USER); ++ else if (!strcmp (this_char, "nouser_xattr")) ++ clear_opt (*mount_options, XATTR_USER); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -924,6 +934,12 @@ + sbi->s_mount_opt = 0; + sbi->s_resuid = EXT3_DEF_RESUID; + sbi->s_resgid = EXT3_DEF_RESGID; ++ ++ /* Default extended attribute flags */ ++#ifdef CONFIG_EXT3_FS_XATTR_USER ++ /* set_opt(sbi->s_mount_opt, XATTR_USER); */ ++#endif ++ + if (!parse_options ((char *) data, &sb_block, sbi, &journal_inum, 0)) { + sb->s_dev = 0; + goto out_fail; +@@ -1742,17 +1758,29 @@ + + static int __init init_ext3_fs(void) + { +- return register_filesystem(&ext3_fs_type); ++ int error = init_ext3_xattr(); ++ if (error) ++ return error; ++ error = init_ext3_xattr_user(); ++ if (error) ++ goto fail; ++ error = register_filesystem(&ext3_fs_type); ++ if (!error) ++ return 0; ++ ++ exit_ext3_xattr_user(); ++fail: ++ exit_ext3_xattr(); ++ return error; + } + + static void __exit exit_ext3_fs(void) + { + unregister_filesystem(&ext3_fs_type); ++ exit_ext3_xattr_user(); ++ exit_ext3_xattr(); + } + +-EXPORT_SYMBOL(ext3_force_commit); +-EXPORT_SYMBOL(ext3_bread); +- + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); + MODULE_LICENSE("GPL"); +Index: linux-2.4.19-pre1/fs/ext3/symlink.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/symlink.c 2001-11-10 01:25:04.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/symlink.c 2003-11-21 03:51:05.000000000 +0300 +@@ -20,6 +20,7 @@ + #include + #include + #include ++#include + + static int ext3_readlink(struct dentry *dentry, char *buffer, int buflen) + { +@@ -33,7 +34,20 @@ + return vfs_follow_link(nd, s); + } + ++struct inode_operations ext3_symlink_inode_operations = { ++ readlink: page_readlink, /* BKL not held. Don't need */ ++ follow_link: page_follow_link, /* BKL not held. Don't need */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ ++}; ++ + struct inode_operations ext3_fast_symlink_inode_operations = { + readlink: ext3_readlink, /* BKL not held. Don't need */ + follow_link: ext3_follow_link, /* BKL not held. Don't need */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; +Index: linux-2.4.19-pre1/fs/ext3/xattr.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/xattr.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/xattr.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,1225 @@ ++/* ++ * linux/fs/ext3/xattr.c ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ * ++ * Fix by Harrison Xing . ++ * Ext3 code with a lot of help from Eric Jarman . ++ * Extended attributes for symlinks and special files added per ++ * suggestion of Luka Renko . ++ */ ++ ++/* ++ * Extended attributes are stored on disk blocks allocated outside of ++ * any inode. The i_file_acl field is then made to point to this allocated ++ * block. If all extended attributes of an inode are identical, these ++ * inodes may share the same extended attribute block. Such situations ++ * are automatically detected by keeping a cache of recent attribute block ++ * numbers and hashes over the block's contents in memory. ++ * ++ * ++ * Extended attribute block layout: ++ * ++ * +------------------+ ++ * | header | ++ * | entry 1 | | ++ * | entry 2 | | growing downwards ++ * | entry 3 | v ++ * | four null bytes | ++ * | . . . | ++ * | value 1 | ^ ++ * | value 3 | | growing upwards ++ * | value 2 | | ++ * +------------------+ ++ * ++ * The block header is followed by multiple entry descriptors. These entry ++ * descriptors are variable in size, and alligned to EXT3_XATTR_PAD ++ * byte boundaries. The entry descriptors are sorted by attribute name, ++ * so that two extended attribute blocks can be compared efficiently. ++ * ++ * Attribute values are aligned to the end of the block, stored in ++ * no specific order. They are also padded to EXT3_XATTR_PAD byte ++ * boundaries. No additional gaps are left between them. ++ * ++ * Locking strategy ++ * ---------------- ++ * The VFS already holds the BKL and the inode->i_sem semaphore when any of ++ * the xattr inode operations are called, so we are guaranteed that only one ++ * processes accesses extended attributes of an inode at any time. ++ * ++ * For writing we also grab the ext3_xattr_sem semaphore. This ensures that ++ * only a single process is modifying an extended attribute block, even ++ * if the block is shared among inodes. ++ * ++ * Note for porting to 2.5 ++ * ----------------------- ++ * The BKL will no longer be held in the xattr inode operations. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++#define EXT3_EA_USER "user." ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++# define mark_buffer_dirty(bh) mark_buffer_dirty(bh, 1) ++#endif ++ ++#define HDR(bh) ((struct ext3_xattr_header *)((bh)->b_data)) ++#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr)) ++#define FIRST_ENTRY(bh) ENTRY(HDR(bh)+1) ++#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) ++ ++#ifdef EXT3_XATTR_DEBUG ++# define ea_idebug(inode, f...) do { \ ++ printk(KERN_DEBUG "inode %s:%ld: ", \ ++ kdevname(inode->i_dev), inode->i_ino); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++# define ea_bdebug(bh, f...) do { \ ++ printk(KERN_DEBUG "block %s:%ld: ", \ ++ kdevname(bh->b_dev), bh->b_blocknr); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++#else ++# define ea_idebug(f...) ++# define ea_bdebug(f...) ++#endif ++ ++static int ext3_xattr_set2(handle_t *, struct inode *, struct buffer_head *, ++ struct ext3_xattr_header *); ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++static int ext3_xattr_cache_insert(struct buffer_head *); ++static struct buffer_head *ext3_xattr_cache_find(struct inode *, ++ struct ext3_xattr_header *); ++static void ext3_xattr_cache_remove(struct buffer_head *); ++static void ext3_xattr_rehash(struct ext3_xattr_header *, ++ struct ext3_xattr_entry *); ++ ++static struct mb_cache *ext3_xattr_cache; ++ ++#else ++# define ext3_xattr_cache_insert(bh) 0 ++# define ext3_xattr_cache_find(inode, header) NULL ++# define ext3_xattr_cache_remove(bh) while(0) {} ++# define ext3_xattr_rehash(header, entry) while(0) {} ++#endif ++ ++/* ++ * If a file system does not share extended attributes among inodes, ++ * we should not need the ext3_xattr_sem semaphore. However, the ++ * filesystem may still contain shared blocks, so we always take ++ * the lock. ++ */ ++ ++DECLARE_MUTEX(ext3_xattr_sem); ++ ++static inline int ++ext3_xattr_new_block(handle_t *handle, struct inode *inode, ++ int * errp, int force) ++{ ++ struct super_block *sb = inode->i_sb; ++ int goal = le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block) + ++ EXT3_I(inode)->i_block_group * EXT3_BLOCKS_PER_GROUP(sb); ++ ++ /* How can we enforce the allocation? */ ++ int block = ext3_new_block(handle, inode, goal, 0, 0, errp); ++#ifdef OLD_QUOTAS ++ if (!*errp) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#endif ++ return block; ++} ++ ++static inline int ++ext3_xattr_quota_alloc(struct inode *inode, int force) ++{ ++ /* How can we enforce the allocation? */ ++#ifdef OLD_QUOTAS ++ int error = DQUOT_ALLOC_BLOCK(inode->i_sb, inode, 1); ++ if (!error) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#else ++ int error = DQUOT_ALLOC_BLOCK(inode, 1); ++#endif ++ return error; ++} ++ ++#ifdef OLD_QUOTAS ++ ++static inline void ++ext3_xattr_quota_free(struct inode *inode) ++{ ++ DQUOT_FREE_BLOCK(inode->i_sb, inode, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++static inline void ++ext3_xattr_free_block(handle_t *handle, struct inode * inode, ++ unsigned long block) ++{ ++ ext3_free_blocks(handle, inode, block, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++#else ++# define ext3_xattr_quota_free(inode) \ ++ DQUOT_FREE_BLOCK(inode, 1) ++# define ext3_xattr_free_block(handle, inode, block) \ ++ ext3_free_blocks(handle, inode, block, 1) ++#endif ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18) ++ ++static inline struct buffer_head * ++sb_bread(struct super_block *sb, int block) ++{ ++ return bread(sb->s_dev, block, sb->s_blocksize); ++} ++ ++static inline struct buffer_head * ++sb_getblk(struct super_block *sb, int block) ++{ ++ return getblk(sb->s_dev, block, sb->s_blocksize); ++} ++ ++#endif ++ ++struct ext3_xattr_handler *ext3_xattr_handlers[EXT3_XATTR_INDEX_MAX]; ++rwlock_t ext3_handler_lock = RW_LOCK_UNLOCKED; ++ ++int ++ext3_xattr_register(int name_index, struct ext3_xattr_handler *handler) ++{ ++ int error = -EINVAL; ++ ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ if (!ext3_xattr_handlers[name_index-1]) { ++ ext3_xattr_handlers[name_index-1] = handler; ++ error = 0; ++ } ++ write_unlock(&ext3_handler_lock); ++ } ++ return error; ++} ++ ++void ++ext3_xattr_unregister(int name_index, struct ext3_xattr_handler *handler) ++{ ++ if (name_index > 0 || name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ ext3_xattr_handlers[name_index-1] = NULL; ++ write_unlock(&ext3_handler_lock); ++ } ++} ++ ++static inline const char * ++strcmp_prefix(const char *a, const char *a_prefix) ++{ ++ while (*a_prefix && *a == *a_prefix) { ++ a++; ++ a_prefix++; ++ } ++ return *a_prefix ? NULL : a; ++} ++ ++/* ++ * Decode the extended attribute name, and translate it into ++ * the name_index and name suffix. ++ */ ++static inline struct ext3_xattr_handler * ++ext3_xattr_resolve_name(const char **name) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ int i; ++ ++ if (!*name) ++ return NULL; ++ read_lock(&ext3_handler_lock); ++ for (i=0; iprefix); ++ if (n) { ++ handler = ext3_xattr_handlers[i]; ++ *name = n; ++ break; ++ } ++ } ++ } ++ read_unlock(&ext3_handler_lock); ++ return handler; ++} ++ ++static inline struct ext3_xattr_handler * ++ext3_xattr_handler(int name_index) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ read_lock(&ext3_handler_lock); ++ handler = ext3_xattr_handlers[name_index-1]; ++ read_unlock(&ext3_handler_lock); ++ } ++ return handler; ++} ++ ++/* ++ * Inode operation getxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_getxattr(struct dentry *dentry, const char *name, ++ void *buffer, size_t size) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->get(inode, name, buffer, size); ++} ++ ++/* ++ * Inode operation listxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_listxattr(struct dentry *dentry, char *buffer, size_t size) ++{ ++ return ext3_xattr_list(dentry->d_inode, buffer, size); ++} ++ ++/* ++ * Inode operation setxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_setxattr(struct dentry *dentry, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ if (size == 0) ++ value = ""; /* empty EA, do not remove */ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, value, size, flags); ++} ++ ++/* ++ * Inode operation removexattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_removexattr(struct dentry *dentry, const char *name) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, NULL, 0, XATTR_REPLACE); ++} ++ ++/* ++ * ext3_xattr_get() ++ * ++ * Copy an extended attribute into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size; ++ char *end; ++ int name_len, error; ++ ++ ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", ++ name_index, name, buffer, (long)buffer_size); ++ ++ if (name == NULL) ++ return -EINVAL; ++ if (!EXT3_I(inode)->i_file_acl) ++ return -ENOATTR; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ ++ error = -ERANGE; ++ if (name_len > 255) ++ goto cleanup; ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) ++ goto found; ++ entry = next; ++ } ++ /* Check the remaining name entries */ ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ entry = next; ++ } ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ error = -ENOATTR; ++ goto cleanup; ++found: ++ /* check the buffer size */ ++ if (entry->e_value_block != 0) ++ goto bad_block; ++ size = le32_to_cpu(entry->e_value_size); ++ if (size > inode->i_sb->s_blocksize || ++ le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize) ++ goto bad_block; ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (buffer) { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ /* return value of attribute */ ++ memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), ++ size); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size = 0; ++ char *buf, *end; ++ int error; ++ ++ ea_idebug(inode, "buffer=%p, buffer_size=%ld", ++ buffer, (long)buffer_size); ++ ++ if (!EXT3_I(inode)->i_file_acl) ++ return 0; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_list", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* compute the size required for the list of attribute names */ ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) ++ size += handler->list(NULL, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (!buffer) { ++ error = size; ++ goto cleanup; ++ } else { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ /* list the attribute names */ ++ buf = buffer; ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) ++ buf += handler->list(buf, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is ++ * not set, set it. ++ */ ++static void ext3_xattr_update_super_block(handle_t *handle, ++ struct super_block *sb) ++{ ++ if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR)) ++ return; ++ ++ lock_super(sb); ++ ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++ EXT3_SB(sb)->s_feature_compat |= EXT3_FEATURE_COMPAT_EXT_ATTR; ++#endif ++ EXT3_SB(sb)->s_es->s_feature_compat |= ++ cpu_to_le32(EXT3_FEATURE_COMPAT_EXT_ATTR); ++ sb->s_dirt = 1; ++ ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); ++ unlock_super(sb); ++} ++ ++/* ++ * ext3_xattr_set() ++ * ++ * Create, replace or remove an extended attribute for this inode. Buffer ++ * is NULL to remove an existing extended attribute, and non-NULL to ++ * either replace an existing extended attribute, or create a new extended ++ * attribute. The flags XATTR_REPLACE and XATTR_CREATE ++ * specify that an extended attribute must exist and must not exist ++ * previous to the call, respectively. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t value_len, int flags) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_header *header = NULL; ++ struct ext3_xattr_entry *here, *last; ++ unsigned int name_len; ++ int block = EXT3_I(inode)->i_file_acl; ++ int min_offs = sb->s_blocksize, not_found = 1, free, error; ++ char *end; ++ ++ /* ++ * header -- Points either into bh, or to a temporarily ++ * allocated buffer. ++ * here -- The named entry found, or the place for inserting, within ++ * the block pointed to by header. ++ * last -- Points right after the last named entry within the block ++ * pointed to by header. ++ * min_offs -- The offset of the first value (values are aligned ++ * towards the end of the block). ++ * end -- Points right after the block pointed to by header. ++ */ ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > sb->s_blocksize) ++ return -ERANGE; ++ down(&ext3_xattr_sem); ++ ++ if (block) { ++ /* The inode already has an extended attribute block. */ ++ bh = sb_bread(sb, block); ++ error = -EIO; ++ if (!bh) ++ goto cleanup; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), ++ le32_to_cpu(HDR(bh)->h_refcount)); ++ header = HDR(bh); ++ end = bh->b_data + bh->b_size; ++ if (header->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ header->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(sb, "ext3_xattr_set", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* Find the named attribute. */ ++ here = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(here)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(here); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!here->e_value_block && here->e_value_size) { ++ int offs = le16_to_cpu(here->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ not_found = name_index - here->e_name_index; ++ if (!not_found) ++ not_found = name_len - here->e_name_len; ++ if (!not_found) ++ not_found = memcmp(name, here->e_name,name_len); ++ if (not_found <= 0) ++ break; ++ here = next; ++ } ++ last = here; ++ /* We still need to compute min_offs and last. */ ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!last->e_value_block && last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ last = next; ++ } ++ ++ /* Check whether we have enough space left. */ ++ free = min_offs - ((char*)last - (char*)header) - sizeof(__u32); ++ } else { ++ /* We will use a new extended attribute block. */ ++ free = sb->s_blocksize - ++ sizeof(struct ext3_xattr_header) - sizeof(__u32); ++ here = last = NULL; /* avoid gcc uninitialized warning. */ ++ } ++ ++ if (not_found) { ++ /* Request to remove a nonexistent attribute? */ ++ error = -ENOATTR; ++ if (flags & XATTR_REPLACE) ++ goto cleanup; ++ error = 0; ++ if (value == NULL) ++ goto cleanup; ++ else ++ free -= EXT3_XATTR_LEN(name_len); ++ } else { ++ /* Request to create an existing attribute? */ ++ error = -EEXIST; ++ if (flags & XATTR_CREATE) ++ goto cleanup; ++ if (!here->e_value_block && here->e_value_size) { ++ unsigned int size = le32_to_cpu(here->e_value_size); ++ ++ if (le16_to_cpu(here->e_value_offs) + size > ++ sb->s_blocksize || size > sb->s_blocksize) ++ goto bad_block; ++ free += EXT3_XATTR_SIZE(size); ++ } ++ } ++ free -= EXT3_XATTR_SIZE(value_len); ++ error = -ENOSPC; ++ if (free < 0) ++ goto cleanup; ++ ++ /* Here we know that we can set the new attribute. */ ++ ++ if (header) { ++ if (header->h_refcount == cpu_to_le32(1)) { ++ ea_bdebug(bh, "modifying in-place"); ++ ext3_xattr_cache_remove(bh); ++ error = ext3_journal_get_write_access(handle, bh); ++ if (error) ++ goto cleanup; ++ } else { ++ int offset; ++ ++ ea_bdebug(bh, "cloning"); ++ header = kmalloc(bh->b_size, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memcpy(header, HDR(bh), bh->b_size); ++ header->h_refcount = cpu_to_le32(1); ++ offset = (char *)header - bh->b_data; ++ here = ENTRY((char *)here + offset); ++ last = ENTRY((char *)last + offset); ++ } ++ } else { ++ /* Allocate a buffer where we construct the new block. */ ++ header = kmalloc(sb->s_blocksize, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memset(header, 0, sb->s_blocksize); ++ end = (char *)header + sb->s_blocksize; ++ header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC); ++ header->h_blocks = header->h_refcount = cpu_to_le32(1); ++ last = here = ENTRY(header+1); ++ } ++ ++ if (not_found) { ++ /* Insert the new name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ int rest = (char *)last - (char *)here; ++ memmove((char *)here + size, here, rest); ++ memset(here, 0, size); ++ here->e_name_index = name_index; ++ here->e_name_len = name_len; ++ memcpy(here->e_name, name, name_len); ++ } else { ++ /* Remove the old value. */ ++ if (!here->e_value_block && here->e_value_size) { ++ char *first_val = (char *)header + min_offs; ++ int offs = le16_to_cpu(here->e_value_offs); ++ char *val = (char *)header + offs; ++ size_t size = EXT3_XATTR_SIZE( ++ le32_to_cpu(here->e_value_size)); ++ memmove(first_val + size, first_val, val - first_val); ++ memset(first_val, 0, size); ++ here->e_value_offs = 0; ++ min_offs += size; ++ ++ /* Adjust all value offsets. */ ++ last = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(last)) { ++ int o = le16_to_cpu(last->e_value_offs); ++ if (!last->e_value_block && o < offs) ++ last->e_value_offs = ++ cpu_to_le16(o + size); ++ last = EXT3_XATTR_NEXT(last); ++ } ++ } ++ if (value == NULL) { ++ /* Remove this attribute. */ ++ if (EXT3_XATTR_NEXT(ENTRY(header+1)) == last) { ++ /* This block is now empty. */ ++ error = ext3_xattr_set2(handle, inode, bh,NULL); ++ goto cleanup; ++ } else { ++ /* Remove the old name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ last = ENTRY((char *)last - size); ++ memmove(here, (char*)here + size, ++ (char*)last - (char*)here); ++ memset(last, 0, size); ++ } ++ } ++ } ++ ++ if (value != NULL) { ++ /* Insert the new value. */ ++ here->e_value_size = cpu_to_le32(value_len); ++ if (value_len) { ++ size_t size = EXT3_XATTR_SIZE(value_len); ++ char *val = (char *)header + min_offs - size; ++ here->e_value_offs = ++ cpu_to_le16((char *)val - (char *)header); ++ memset(val + size - EXT3_XATTR_PAD, 0, ++ EXT3_XATTR_PAD); /* Clear the pad bytes. */ ++ memcpy(val, value, value_len); ++ } ++ } ++ ext3_xattr_rehash(header, here); ++ ++ error = ext3_xattr_set2(handle, inode, bh, header); ++ ++cleanup: ++ brelse(bh); ++ if (!(bh && header == HDR(bh))) ++ kfree(header); ++ up(&ext3_xattr_sem); ++ ++ return error; ++} ++ ++/* ++ * Second half of ext3_xattr_set(): Update the file system. ++ */ ++static int ++ext3_xattr_set2(handle_t *handle, struct inode *inode, ++ struct buffer_head *old_bh, struct ext3_xattr_header *header) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *new_bh = NULL; ++ int error; ++ ++ if (header) { ++ new_bh = ext3_xattr_cache_find(inode, header); ++ if (new_bh) { ++ /* ++ * We found an identical block in the cache. ++ * The old block will be released after updating ++ * the inode. ++ */ ++ ea_bdebug(old_bh, "reusing block %ld", ++ new_bh->b_blocknr); ++ ++ error = -EDQUOT; ++ if (ext3_xattr_quota_alloc(inode, 1)) ++ goto cleanup; ++ ++ error = ext3_journal_get_write_access(handle, new_bh); ++ if (error) ++ goto cleanup; ++ HDR(new_bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(new_bh)->h_refcount) + 1); ++ ea_bdebug(new_bh, "refcount now=%d", ++ le32_to_cpu(HDR(new_bh)->h_refcount)); ++ } else if (old_bh && header == HDR(old_bh)) { ++ /* Keep this block. */ ++ new_bh = old_bh; ++ ext3_xattr_cache_insert(new_bh); ++ } else { ++ /* We need to allocate a new block */ ++ int force = EXT3_I(inode)->i_file_acl != 0; ++ int block = ext3_xattr_new_block(handle, inode, ++ &error, force); ++ if (error) ++ goto cleanup; ++ ea_idebug(inode, "creating block %d", block); ++ ++ new_bh = sb_getblk(sb, block); ++ if (!new_bh) { ++getblk_failed: ext3_xattr_free_block(handle, inode, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ lock_buffer(new_bh); ++ error = ext3_journal_get_create_access(handle, new_bh); ++ if (error) { ++ unlock_buffer(new_bh); ++ goto getblk_failed; ++ } ++ memcpy(new_bh->b_data, header, new_bh->b_size); ++ mark_buffer_uptodate(new_bh, 1); ++ unlock_buffer(new_bh); ++ ext3_xattr_cache_insert(new_bh); ++ ++ ext3_xattr_update_super_block(handle, sb); ++ } ++ error = ext3_journal_dirty_metadata(handle, new_bh); ++ if (error) ++ goto cleanup; ++ } ++ ++ /* Update the inode. */ ++ EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; ++ inode->i_ctime = CURRENT_TIME; ++ ext3_mark_inode_dirty(handle, inode); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ++ error = 0; ++ if (old_bh && old_bh != new_bh) { ++ /* ++ * If there was an old block, and we are not still using it, ++ * we now release the old block. ++ */ ++ unsigned int refcount = le32_to_cpu(HDR(old_bh)->h_refcount); ++ ++ error = ext3_journal_get_write_access(handle, old_bh); ++ if (error) ++ goto cleanup; ++ if (refcount == 1) { ++ /* Free the old block. */ ++ ea_bdebug(old_bh, "freeing"); ++ ext3_xattr_free_block(handle, inode, old_bh->b_blocknr); ++ ++ /* ext3_forget() calls bforget() for us, but we ++ let our caller release old_bh, so we need to ++ duplicate the handle before. */ ++ get_bh(old_bh); ++ ext3_forget(handle, 1, inode, old_bh,old_bh->b_blocknr); ++ } else { ++ /* Decrement the refcount only. */ ++ refcount--; ++ HDR(old_bh)->h_refcount = cpu_to_le32(refcount); ++ ext3_xattr_quota_free(inode); ++ ext3_journal_dirty_metadata(handle, old_bh); ++ ea_bdebug(old_bh, "refcount now=%d", refcount); ++ } ++ } ++ ++cleanup: ++ if (old_bh != new_bh) ++ brelse(new_bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_delete_inode() ++ * ++ * Free extended attribute resources associated with this inode. This ++ * is called immediately before an inode is freed. ++ */ ++void ++ext3_xattr_delete_inode(handle_t *handle, struct inode *inode) ++{ ++ struct buffer_head *bh; ++ unsigned int block = EXT3_I(inode)->i_file_acl; ++ ++ if (!block) ++ return; ++ down(&ext3_xattr_sem); ++ ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_delete_inode", ++ "inode %ld: block %d read error", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "b_count=%d", atomic_read(&(bh->b_count))); ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++ ext3_error(inode->i_sb, "ext3_xattr_delete_inode", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ goto cleanup; ++ } ++ ext3_journal_get_write_access(handle, bh); ++ ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ if (HDR(bh)->h_refcount == cpu_to_le32(1)) { ++ ext3_xattr_cache_remove(bh); ++ ext3_xattr_free_block(handle, inode, block); ++ ext3_forget(handle, 1, inode, bh, block); ++ bh = NULL; ++ } else { ++ HDR(bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ ext3_journal_dirty_metadata(handle, bh); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ext3_xattr_quota_free(inode); ++ } ++ EXT3_I(inode)->i_file_acl = 0; ++ ++cleanup: ++ brelse(bh); ++ up(&ext3_xattr_sem); ++} ++ ++/* ++ * ext3_xattr_put_super() ++ * ++ * This is called when a file system is unmounted. ++ */ ++void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ mb_cache_shrink(ext3_xattr_cache, sb->s_dev); ++#endif ++} ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++/* ++ * ext3_xattr_cache_insert() ++ * ++ * Create a new entry in the extended attribute cache, and insert ++ * it unless such an entry is already in the cache. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++static int ++ext3_xattr_cache_insert(struct buffer_head *bh) ++{ ++ __u32 hash = le32_to_cpu(HDR(bh)->h_hash); ++ struct mb_cache_entry *ce; ++ int error; ++ ++ ce = mb_cache_entry_alloc(ext3_xattr_cache); ++ if (!ce) ++ return -ENOMEM; ++ error = mb_cache_entry_insert(ce, bh->b_dev, bh->b_blocknr, &hash); ++ if (error) { ++ mb_cache_entry_free(ce); ++ if (error == -EBUSY) { ++ ea_bdebug(bh, "already in cache (%d cache entries)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ error = 0; ++ } ++ } else { ++ ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash, ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ mb_cache_entry_release(ce); ++ } ++ return error; ++} ++ ++/* ++ * ext3_xattr_cmp() ++ * ++ * Compare two extended attribute blocks for equality. ++ * ++ * Returns 0 if the blocks are equal, 1 if they differ, and ++ * a negative error number on errors. ++ */ ++static int ++ext3_xattr_cmp(struct ext3_xattr_header *header1, ++ struct ext3_xattr_header *header2) ++{ ++ struct ext3_xattr_entry *entry1, *entry2; ++ ++ entry1 = ENTRY(header1+1); ++ entry2 = ENTRY(header2+1); ++ while (!IS_LAST_ENTRY(entry1)) { ++ if (IS_LAST_ENTRY(entry2)) ++ return 1; ++ if (entry1->e_hash != entry2->e_hash || ++ entry1->e_name_len != entry2->e_name_len || ++ entry1->e_value_size != entry2->e_value_size || ++ memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) ++ return 1; ++ if (entry1->e_value_block != 0 || entry2->e_value_block != 0) ++ return -EIO; ++ if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), ++ (char *)header2 + le16_to_cpu(entry2->e_value_offs), ++ le32_to_cpu(entry1->e_value_size))) ++ return 1; ++ ++ entry1 = EXT3_XATTR_NEXT(entry1); ++ entry2 = EXT3_XATTR_NEXT(entry2); ++ } ++ if (!IS_LAST_ENTRY(entry2)) ++ return 1; ++ return 0; ++} ++ ++/* ++ * ext3_xattr_cache_find() ++ * ++ * Find an identical extended attribute block. ++ * ++ * Returns a pointer to the block found, or NULL if such a block was ++ * not found or an error occurred. ++ */ ++static struct buffer_head * ++ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header) ++{ ++ __u32 hash = le32_to_cpu(header->h_hash); ++ struct mb_cache_entry *ce; ++ ++ if (!header->h_hash) ++ return NULL; /* never share */ ++ ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); ++ ce = mb_cache_entry_find_first(ext3_xattr_cache, 0, inode->i_dev, hash); ++ while (ce) { ++ struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block); ++ ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_cache_find", ++ "inode %ld: block %ld read error", ++ inode->i_ino, ce->e_block); ++ } else if (le32_to_cpu(HDR(bh)->h_refcount) > ++ EXT3_XATTR_REFCOUNT_MAX) { ++ ea_idebug(inode, "block %ld refcount %d>%d",ce->e_block, ++ le32_to_cpu(HDR(bh)->h_refcount), ++ EXT3_XATTR_REFCOUNT_MAX); ++ } else if (!ext3_xattr_cmp(header, HDR(bh))) { ++ ea_bdebug(bh, "b_count=%d",atomic_read(&(bh->b_count))); ++ mb_cache_entry_release(ce); ++ return bh; ++ } ++ brelse(bh); ++ ce = mb_cache_entry_find_next(ce, 0, inode->i_dev, hash); ++ } ++ return NULL; ++} ++ ++/* ++ * ext3_xattr_cache_remove() ++ * ++ * Remove the cache entry of a block from the cache. Called when a ++ * block becomes invalid. ++ */ ++static void ++ext3_xattr_cache_remove(struct buffer_head *bh) ++{ ++ struct mb_cache_entry *ce; ++ ++ ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_dev, bh->b_blocknr); ++ if (ce) { ++ ea_bdebug(bh, "removing (%d cache entries remaining)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)-1); ++ mb_cache_entry_free(ce); ++ } else ++ ea_bdebug(bh, "no cache entry"); ++} ++ ++#define NAME_HASH_SHIFT 5 ++#define VALUE_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_hash_entry() ++ * ++ * Compute the hash of an extended attribute. ++ */ ++static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ __u32 hash = 0; ++ char *name = entry->e_name; ++ int n; ++ ++ for (n=0; n < entry->e_name_len; n++) { ++ hash = (hash << NAME_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ ++ *name++; ++ } ++ ++ if (entry->e_value_block == 0 && entry->e_value_size != 0) { ++ __u32 *value = (__u32 *)((char *)header + ++ le16_to_cpu(entry->e_value_offs)); ++ for (n = (le32_to_cpu(entry->e_value_size) + ++ EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) { ++ hash = (hash << VALUE_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ ++ le32_to_cpu(*value++); ++ } ++ } ++ entry->e_hash = cpu_to_le32(hash); ++} ++ ++#undef NAME_HASH_SHIFT ++#undef VALUE_HASH_SHIFT ++ ++#define BLOCK_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_rehash() ++ * ++ * Re-compute the extended attribute hash value after an entry has changed. ++ */ ++static void ext3_xattr_rehash(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ struct ext3_xattr_entry *here; ++ __u32 hash = 0; ++ ++ ext3_xattr_hash_entry(header, entry); ++ here = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(here)) { ++ if (!here->e_hash) { ++ /* Block is not shared if an entry's hash value == 0 */ ++ hash = 0; ++ break; ++ } ++ hash = (hash << BLOCK_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ ++ le32_to_cpu(here->e_hash); ++ here = EXT3_XATTR_NEXT(here); ++ } ++ header->h_hash = cpu_to_le32(hash); ++} ++ ++#undef BLOCK_HASH_SHIFT ++ ++int __init ++init_ext3_xattr(void) ++{ ++ ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL, ++ sizeof(struct mb_cache_entry) + ++ sizeof(struct mb_cache_entry_index), 1, 61); ++ if (!ext3_xattr_cache) ++ return -ENOMEM; ++ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++ if (ext3_xattr_cache) ++ mb_cache_destroy(ext3_xattr_cache); ++ ext3_xattr_cache = NULL; ++} ++ ++#else /* CONFIG_EXT3_FS_XATTR_SHARING */ ++ ++int __init ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_SHARING */ +Index: linux-2.4.19-pre1/fs/ext3/xattr_user.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/xattr_user.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/xattr_user.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,111 @@ ++/* ++ * linux/fs/ext3/xattr_user.c ++ * Handler for extended user attributes. ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++ ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++# include ++#endif ++ ++#define XATTR_USER_PREFIX "user." ++ ++static size_t ++ext3_xattr_user_list(char *list, struct inode *inode, ++ const char *name, int name_len) ++{ ++ const int prefix_len = sizeof(XATTR_USER_PREFIX)-1; ++ ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return 0; ++ ++ if (list) { ++ memcpy(list, XATTR_USER_PREFIX, prefix_len); ++ memcpy(list+prefix_len, name, name_len); ++ list[prefix_len + name_len] = '\0'; ++ } ++ return prefix_len + name_len + 1; ++} ++ ++static int ++ext3_xattr_user_get(struct inode *inode, const char *name, ++ void *buffer, size_t size) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++ error = ext3_permission_locked(inode, MAY_READ); ++#else ++ error = permission(inode, MAY_READ); ++#endif ++ if (error) ++ return error; ++ ++ return ext3_xattr_get(inode, EXT3_XATTR_INDEX_USER, name, ++ buffer, size); ++} ++ ++static int ++ext3_xattr_user_set(struct inode *inode, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ handle_t *handle; ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++ if ( !S_ISREG(inode->i_mode) && ++ (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX)) ++ return -EPERM; ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++ error = ext3_permission_locked(inode, MAY_WRITE); ++#else ++ error = permission(inode, MAY_WRITE); ++#endif ++ if (error) ++ return error; ++ ++ handle = ext3_journal_start(inode, EXT3_XATTR_TRANS_BLOCKS); ++ if (IS_ERR(handle)) ++ return PTR_ERR(handle); ++ error = ext3_xattr_set(handle, inode, EXT3_XATTR_INDEX_USER, name, ++ value, size, flags); ++ ext3_journal_stop(handle, inode); ++ ++ return error; ++} ++ ++struct ext3_xattr_handler ext3_xattr_user_handler = { ++ prefix: XATTR_USER_PREFIX, ++ list: ext3_xattr_user_list, ++ get: ext3_xattr_user_get, ++ set: ext3_xattr_user_set, ++}; ++ ++int __init ++init_ext3_xattr_user(void) ++{ ++ return ext3_xattr_register(EXT3_XATTR_INDEX_USER, ++ &ext3_xattr_user_handler); ++} ++ ++void ++exit_ext3_xattr_user(void) ++{ ++ ext3_xattr_unregister(EXT3_XATTR_INDEX_USER, ++ &ext3_xattr_user_handler); ++} +Index: linux-2.4.19-pre1/fs/mbcache.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/mbcache.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/mbcache.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,648 @@ ++/* ++ * linux/fs/mbcache.c ++ * (C) 2001-2002 Andreas Gruenbacher, ++ */ ++ ++/* ++ * Filesystem Meta Information Block Cache (mbcache) ++ * ++ * The mbcache caches blocks of block devices that need to be located ++ * by their device/block number, as well as by other criteria (such ++ * as the block's contents). ++ * ++ * There can only be one cache entry in a cache per device and block number. ++ * Additional indexes need not be unique in this sense. The number of ++ * additional indexes (=other criteria) can be hardwired at compile time ++ * or specified at cache create time. ++ * ++ * Each cache entry is of fixed size. An entry may be `valid' or `invalid' ++ * in the cache. A valid entry is in the main hash tables of the cache, ++ * and may also be in the lru list. An invalid entry is not in any hashes ++ * or lists. ++ * ++ * A valid cache entry is only in the lru list if no handles refer to it. ++ * Invalid cache entries will be freed when the last handle to the cache ++ * entry is released. Entries that cannot be freed immediately are put ++ * back on the lru list. ++ */ ++ ++#include ++#include ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++ ++#ifdef MB_CACHE_DEBUG ++# define mb_debug(f...) do { \ ++ printk(KERN_DEBUG f); \ ++ printk("\n"); \ ++ } while (0) ++#define mb_assert(c) do { if (!(c)) \ ++ printk(KERN_ERR "assertion " #c " failed\n"); \ ++ } while(0) ++#else ++# define mb_debug(f...) do { } while(0) ++# define mb_assert(c) do { } while(0) ++#endif ++#define mb_error(f...) do { \ ++ printk(KERN_ERR f); \ ++ printk("\n"); \ ++ } while(0) ++ ++MODULE_AUTHOR("Andreas Gruenbacher "); ++MODULE_DESCRIPTION("Meta block cache (for extended attributes)"); ++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) ++MODULE_LICENSE("GPL"); ++#endif ++ ++EXPORT_SYMBOL(mb_cache_create); ++EXPORT_SYMBOL(mb_cache_shrink); ++EXPORT_SYMBOL(mb_cache_destroy); ++EXPORT_SYMBOL(mb_cache_entry_alloc); ++EXPORT_SYMBOL(mb_cache_entry_insert); ++EXPORT_SYMBOL(mb_cache_entry_release); ++EXPORT_SYMBOL(mb_cache_entry_takeout); ++EXPORT_SYMBOL(mb_cache_entry_free); ++EXPORT_SYMBOL(mb_cache_entry_dup); ++EXPORT_SYMBOL(mb_cache_entry_get); ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++EXPORT_SYMBOL(mb_cache_entry_find_first); ++EXPORT_SYMBOL(mb_cache_entry_find_next); ++#endif ++ ++ ++/* ++ * Global data: list of all mbcache's, lru list, and a spinlock for ++ * accessing cache data structures on SMP machines. The lru list is ++ * global across all mbcaches. ++ */ ++ ++static LIST_HEAD(mb_cache_list); ++static LIST_HEAD(mb_cache_lru_list); ++static spinlock_t mb_cache_spinlock = SPIN_LOCK_UNLOCKED; ++ ++static inline int ++mb_cache_indexes(struct mb_cache *cache) ++{ ++#ifdef MB_CACHE_INDEXES_COUNT ++ return MB_CACHE_INDEXES_COUNT; ++#else ++ return cache->c_indexes_count; ++#endif ++} ++ ++/* ++ * What the mbcache registers as to get shrunk dynamically. ++ */ ++ ++static void ++mb_cache_memory_pressure(int priority, unsigned int gfp_mask); ++ ++static struct cache_definition mb_cache_definition = { ++ "mb_cache", ++ mb_cache_memory_pressure ++}; ++ ++ ++static inline int ++__mb_cache_entry_is_hashed(struct mb_cache_entry *ce) ++{ ++ return !list_empty(&ce->e_block_list); ++} ++ ++ ++static inline void ++__mb_cache_entry_unhash(struct mb_cache_entry *ce) ++{ ++ int n; ++ ++ if (__mb_cache_entry_is_hashed(ce)) { ++ list_del_init(&ce->e_block_list); ++ for (n=0; ne_cache); n++) ++ list_del(&ce->e_indexes[n].o_list); ++ } ++} ++ ++ ++static inline void ++__mb_cache_entry_forget(struct mb_cache_entry *ce, int gfp_mask) ++{ ++ struct mb_cache *cache = ce->e_cache; ++ ++ mb_assert(atomic_read(&ce->e_used) == 0); ++ if (cache->c_op.free && cache->c_op.free(ce, gfp_mask)) { ++ /* free failed -- put back on the lru list ++ for freeing later. */ ++ spin_lock(&mb_cache_spinlock); ++ list_add(&ce->e_lru_list, &mb_cache_lru_list); ++ spin_unlock(&mb_cache_spinlock); ++ } else { ++ kmem_cache_free(cache->c_entry_cache, ce); ++ atomic_dec(&cache->c_entry_count); ++ } ++} ++ ++ ++static inline void ++__mb_cache_entry_release_unlock(struct mb_cache_entry *ce) ++{ ++ if (atomic_dec_and_test(&ce->e_used)) { ++ if (__mb_cache_entry_is_hashed(ce)) ++ list_add_tail(&ce->e_lru_list, &mb_cache_lru_list); ++ else { ++ spin_unlock(&mb_cache_spinlock); ++ __mb_cache_entry_forget(ce, GFP_KERNEL); ++ return; ++ } ++ } ++ spin_unlock(&mb_cache_spinlock); ++} ++ ++ ++/* ++ * mb_cache_memory_pressure() memory pressure callback ++ * ++ * This function is called by the kernel memory management when memory ++ * gets low. ++ * ++ * @priority: Amount by which to shrink the cache (0 = highes priority) ++ * @gfp_mask: (ignored) ++ */ ++static void ++mb_cache_memory_pressure(int priority, unsigned int gfp_mask) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ int count = 0; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &mb_cache_list) { ++ struct mb_cache *cache = ++ list_entry(l, struct mb_cache, c_cache_list); ++ mb_debug("cache %s (%d)", cache->c_name, ++ atomic_read(&cache->c_entry_count)); ++ count += atomic_read(&cache->c_entry_count); ++ } ++ mb_debug("trying to free %d of %d entries", ++ count / (priority ? priority : 1), count); ++ if (priority) ++ count /= priority; ++ while (count-- && !list_empty(&mb_cache_lru_list)) { ++ struct mb_cache_entry *ce = ++ list_entry(mb_cache_lru_list.next, ++ struct mb_cache_entry, e_lru_list); ++ list_del(&ce->e_lru_list); ++ __mb_cache_entry_unhash(ce); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ } ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), gfp_mask); ++ } ++} ++ ++ ++/* ++ * mb_cache_create() create a new cache ++ * ++ * All entries in one cache are equal size. Cache entries may be from ++ * multiple devices. If this is the first mbcache created, registers ++ * the cache with kernel memory management. Returns NULL if no more ++ * memory was available. ++ * ++ * @name: name of the cache (informal) ++ * @cache_op: contains the callback called when freeing a cache entry ++ * @entry_size: The size of a cache entry, including ++ * struct mb_cache_entry ++ * @indexes_count: number of additional indexes in the cache. Must equal ++ * MB_CACHE_INDEXES_COUNT if the number of indexes is ++ * hardwired. ++ * @bucket_count: number of hash buckets ++ */ ++struct mb_cache * ++mb_cache_create(const char *name, struct mb_cache_op *cache_op, ++ size_t entry_size, int indexes_count, int bucket_count) ++{ ++ int m=0, n; ++ struct mb_cache *cache = NULL; ++ ++ if(entry_size < sizeof(struct mb_cache_entry) + ++ indexes_count * sizeof(struct mb_cache_entry_index)) ++ return NULL; ++ ++ MOD_INC_USE_COUNT; ++ cache = kmalloc(sizeof(struct mb_cache) + ++ indexes_count * sizeof(struct list_head), GFP_KERNEL); ++ if (!cache) ++ goto fail; ++ cache->c_name = name; ++ cache->c_op.free = NULL; ++ if (cache_op) ++ cache->c_op.free = cache_op->free; ++ atomic_set(&cache->c_entry_count, 0); ++ cache->c_bucket_count = bucket_count; ++#ifdef MB_CACHE_INDEXES_COUNT ++ mb_assert(indexes_count == MB_CACHE_INDEXES_COUNT); ++#else ++ cache->c_indexes_count = indexes_count; ++#endif ++ cache->c_block_hash = kmalloc(bucket_count * sizeof(struct list_head), ++ GFP_KERNEL); ++ if (!cache->c_block_hash) ++ goto fail; ++ for (n=0; nc_block_hash[n]); ++ for (m=0; mc_indexes_hash[m] = kmalloc(bucket_count * ++ sizeof(struct list_head), ++ GFP_KERNEL); ++ if (!cache->c_indexes_hash[m]) ++ goto fail; ++ for (n=0; nc_indexes_hash[m][n]); ++ } ++ cache->c_entry_cache = kmem_cache_create(name, entry_size, 0, ++ 0 /*SLAB_POISON | SLAB_RED_ZONE*/, NULL, NULL); ++ if (!cache->c_entry_cache) ++ goto fail; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_add(&cache->c_cache_list, &mb_cache_list); ++ spin_unlock(&mb_cache_spinlock); ++ return cache; ++ ++fail: ++ if (cache) { ++ while (--m >= 0) ++ kfree(cache->c_indexes_hash[m]); ++ if (cache->c_block_hash) ++ kfree(cache->c_block_hash); ++ kfree(cache); ++ } ++ MOD_DEC_USE_COUNT; ++ return NULL; ++} ++ ++ ++/* ++ * mb_cache_shrink() ++ * ++ * Removes all cache entires of a device from the cache. All cache entries ++ * currently in use cannot be freed, and thus remain in the cache. ++ * ++ * @cache: which cache to shrink ++ * @dev: which device's cache entries to shrink ++ */ ++void ++mb_cache_shrink(struct mb_cache *cache, kdev_t dev) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &mb_cache_lru_list) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_lru_list); ++ if (ce->e_dev == dev) { ++ list_del(&ce->e_lru_list); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ __mb_cache_entry_unhash(ce); ++ } ++ } ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), GFP_KERNEL); ++ } ++} ++ ++ ++/* ++ * mb_cache_destroy() ++ * ++ * Shrinks the cache to its minimum possible size (hopefully 0 entries), ++ * and then destroys it. If this was the last mbcache, un-registers the ++ * mbcache from kernel memory management. ++ */ ++void ++mb_cache_destroy(struct mb_cache *cache) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ int n; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &mb_cache_lru_list) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_lru_list); ++ if (ce->e_cache == cache) { ++ list_del(&ce->e_lru_list); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ __mb_cache_entry_unhash(ce); ++ } ++ } ++ list_del(&cache->c_cache_list); ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), GFP_KERNEL); ++ } ++ ++ if (atomic_read(&cache->c_entry_count) > 0) { ++ mb_error("cache %s: %d orphaned entries", ++ cache->c_name, ++ atomic_read(&cache->c_entry_count)); ++ } ++ ++#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,3,0)) ++ /* We don't have kmem_cache_destroy() in 2.2.x */ ++ kmem_cache_shrink(cache->c_entry_cache); ++#else ++ kmem_cache_destroy(cache->c_entry_cache); ++#endif ++ for (n=0; n < mb_cache_indexes(cache); n++) ++ kfree(cache->c_indexes_hash[n]); ++ kfree(cache->c_block_hash); ++ kfree(cache); ++ ++ MOD_DEC_USE_COUNT; ++} ++ ++ ++/* ++ * mb_cache_entry_alloc() ++ * ++ * Allocates a new cache entry. The new entry will not be valid initially, ++ * and thus cannot be looked up yet. It should be filled with data, and ++ * then inserted into the cache using mb_cache_entry_insert(). Returns NULL ++ * if no more memory was available. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_alloc(struct mb_cache *cache) ++{ ++ struct mb_cache_entry *ce; ++ ++ atomic_inc(&cache->c_entry_count); ++ ce = kmem_cache_alloc(cache->c_entry_cache, GFP_KERNEL); ++ if (ce) { ++ INIT_LIST_HEAD(&ce->e_lru_list); ++ INIT_LIST_HEAD(&ce->e_block_list); ++ ce->e_cache = cache; ++ atomic_set(&ce->e_used, 1); ++ } ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_insert() ++ * ++ * Inserts an entry that was allocated using mb_cache_entry_alloc() into ++ * the cache. After this, the cache entry can be looked up, but is not yet ++ * in the lru list as the caller still holds a handle to it. Returns 0 on ++ * success, or -EBUSY if a cache entry for that device + inode exists ++ * already (this may happen after a failed lookup, if another process has ++ * inserted the same cache entry in the meantime). ++ * ++ * @dev: device the cache entry belongs to ++ * @block: block number ++ * @keys: array of additional keys. There must be indexes_count entries ++ * in the array (as specified when creating the cache). ++ */ ++int ++mb_cache_entry_insert(struct mb_cache_entry *ce, kdev_t dev, ++ unsigned long block, unsigned int keys[]) ++{ ++ struct mb_cache *cache = ce->e_cache; ++ unsigned int bucket = (HASHDEV(dev) + block) % cache->c_bucket_count; ++ struct list_head *l; ++ int error = -EBUSY, n; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &cache->c_block_hash[bucket]) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_block_list); ++ if (ce->e_dev == dev && ce->e_block == block) ++ goto out; ++ } ++ __mb_cache_entry_unhash(ce); ++ ce->e_dev = dev; ++ ce->e_block = block; ++ list_add(&ce->e_block_list, &cache->c_block_hash[bucket]); ++ for (n=0; ne_indexes[n].o_key = keys[n]; ++ bucket = keys[n] % cache->c_bucket_count; ++ list_add(&ce->e_indexes[n].o_list, ++ &cache->c_indexes_hash[n][bucket]); ++ } ++out: ++ spin_unlock(&mb_cache_spinlock); ++ return error; ++} ++ ++ ++/* ++ * mb_cache_entry_release() ++ * ++ * Release a handle to a cache entry. When the last handle to a cache entry ++ * is released it is either freed (if it is invalid) or otherwise inserted ++ * in to the lru list. ++ */ ++void ++mb_cache_entry_release(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ __mb_cache_entry_release_unlock(ce); ++} ++ ++ ++/* ++ * mb_cache_entry_takeout() ++ * ++ * Take a cache entry out of the cache, making it invalid. The entry can later ++ * be re-inserted using mb_cache_entry_insert(), or released using ++ * mb_cache_entry_release(). ++ */ ++void ++mb_cache_entry_takeout(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ mb_assert(list_empty(&ce->e_lru_list)); ++ __mb_cache_entry_unhash(ce); ++ spin_unlock(&mb_cache_spinlock); ++} ++ ++ ++/* ++ * mb_cache_entry_free() ++ * ++ * This is equivalent to the sequence mb_cache_entry_takeout() -- ++ * mb_cache_entry_release(). ++ */ ++void ++mb_cache_entry_free(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ mb_assert(list_empty(&ce->e_lru_list)); ++ __mb_cache_entry_unhash(ce); ++ __mb_cache_entry_release_unlock(ce); ++} ++ ++ ++/* ++ * mb_cache_entry_dup() ++ * ++ * Duplicate a handle to a cache entry (does not duplicate the cache entry ++ * itself). After the call, both the old and the new handle must be released. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_dup(struct mb_cache_entry *ce) ++{ ++ atomic_inc(&ce->e_used); ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_get() ++ * ++ * Get a cache entry by device / block number. (There can only be one entry ++ * in the cache per device and block.) Returns NULL if no such cache entry ++ * exists. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_get(struct mb_cache *cache, kdev_t dev, unsigned long block) ++{ ++ unsigned int bucket = (HASHDEV(dev) + block) % cache->c_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &cache->c_block_hash[bucket]) { ++ ce = list_entry(l, struct mb_cache_entry, e_block_list); ++ if (ce->e_dev == dev && ce->e_block == block) { ++ if (!list_empty(&ce->e_lru_list)) ++ list_del_init(&ce->e_lru_list); ++ atomic_inc(&ce->e_used); ++ goto cleanup; ++ } ++ } ++ ce = NULL; ++ ++cleanup: ++ spin_unlock(&mb_cache_spinlock); ++ return ce; ++} ++ ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++ ++static struct mb_cache_entry * ++__mb_cache_entry_find(struct list_head *l, struct list_head *head, ++ int index, kdev_t dev, unsigned int key) ++{ ++ while (l != head) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, ++ e_indexes[index].o_list); ++ if (ce->e_dev == dev && ce->e_indexes[index].o_key == key) { ++ if (!list_empty(&ce->e_lru_list)) ++ list_del_init(&ce->e_lru_list); ++ atomic_inc(&ce->e_used); ++ return ce; ++ } ++ l = l->next; ++ } ++ return NULL; ++} ++ ++ ++/* ++ * mb_cache_entry_find_first() ++ * ++ * Find the first cache entry on a given device with a certain key in ++ * an additional index. Additonal matches can be found with ++ * mb_cache_entry_find_next(). Returns NULL if no match was found. ++ * ++ * @cache: the cache to search ++ * @index: the number of the additonal index to search (0<=indexc_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ mb_assert(index < mb_cache_indexes(cache)); ++ spin_lock(&mb_cache_spinlock); ++ l = cache->c_indexes_hash[index][bucket].next; ++ ce = __mb_cache_entry_find(l, &cache->c_indexes_hash[index][bucket], ++ index, dev, key); ++ spin_unlock(&mb_cache_spinlock); ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_find_next() ++ * ++ * Find the next cache entry on a given device with a certain key in an ++ * additional index. Returns NULL if no match could be found. The previous ++ * entry is atomatically released, so that mb_cache_entry_find_next() can ++ * be called like this: ++ * ++ * entry = mb_cache_entry_find_first(); ++ * while (entry) { ++ * ... ++ * entry = mb_cache_entry_find_next(entry, ...); ++ * } ++ * ++ * @prev: The previous match ++ * @index: the number of the additonal index to search (0<=indexe_cache; ++ unsigned int bucket = key % cache->c_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ mb_assert(index < mb_cache_indexes(cache)); ++ spin_lock(&mb_cache_spinlock); ++ l = prev->e_indexes[index].o_list.next; ++ ce = __mb_cache_entry_find(l, &cache->c_indexes_hash[index][bucket], ++ index, dev, key); ++ __mb_cache_entry_release_unlock(prev); ++ return ce; ++} ++ ++#endif /* !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) */ ++ ++static int __init init_mbcache(void) ++{ ++ register_cache(&mb_cache_definition); ++ return 0; ++} ++ ++static void __exit exit_mbcache(void) ++{ ++ unregister_cache(&mb_cache_definition); ++} ++ ++module_init(init_mbcache) ++module_exit(exit_mbcache) ++ +Index: linux-2.4.19-pre1/include/asm-arm/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-arm/unistd.h 2001-08-12 22:14:00.000000000 +0400 ++++ linux-2.4.19-pre1/include/asm-arm/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -240,6 +240,18 @@ + #define __NR_mincore (__NR_SYSCALL_BASE+219) + #define __NR_madvise (__NR_SYSCALL_BASE+220) + #define __NR_fcntl64 (__NR_SYSCALL_BASE+221) ++#define __NR_setxattr (__NR_SYSCALL_BASE+226) ++#define __NR_lsetxattr (__NR_SYSCALL_BASE+227) ++#define __NR_fsetxattr (__NR_SYSCALL_BASE+228) ++#define __NR_getxattr (__NR_SYSCALL_BASE+229) ++#define __NR_lgetxattr (__NR_SYSCALL_BASE+230) ++#define __NR_fgetxattr (__NR_SYSCALL_BASE+231) ++#define __NR_listxattr (__NR_SYSCALL_BASE+232) ++#define __NR_llistxattr (__NR_SYSCALL_BASE+233) ++#define __NR_flistxattr (__NR_SYSCALL_BASE+234) ++#define __NR_removexattr (__NR_SYSCALL_BASE+235) ++#define __NR_lremovexattr (__NR_SYSCALL_BASE+236) ++#define __NR_fremovexattr (__NR_SYSCALL_BASE+237) + + /* + * The following SWIs are ARM private. +Index: linux-2.4.19-pre1/include/asm-ia64/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-ia64/unistd.h 2001-11-10 01:26:17.000000000 +0300 ++++ linux-2.4.19-pre1/include/asm-ia64/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -206,6 +206,18 @@ + #define __NR_getdents64 1214 + #define __NR_getunwind 1215 + #define __NR_readahead 1216 ++#define __NR_setxattr 1217 ++#define __NR_lsetxattr 1218 ++#define __NR_fsetxattr 1219 ++#define __NR_getxattr 1220 ++#define __NR_lgetxattr 1221 ++#define __NR_fgetxattr 1222 ++#define __NR_listxattr 1223 ++#define __NR_llistxattr 1224 ++#define __NR_flistxattr 1225 ++#define __NR_removexattr 1226 ++#define __NR_lremovexattr 1227 ++#define __NR_fremovexattr 1228 + + #if !defined(__ASSEMBLY__) && !defined(ASSEMBLER) + +Index: linux-2.4.19-pre1/include/asm-s390/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-s390/unistd.h 2001-10-11 20:43:38.000000000 +0400 ++++ linux-2.4.19-pre1/include/asm-s390/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -211,6 +211,18 @@ + #define __NR_mincore 218 + #define __NR_madvise 219 + #define __NR_getdents64 220 ++#define __NR_setxattr 224 ++#define __NR_lsetxattr 225 ++#define __NR_fsetxattr 226 ++#define __NR_getxattr 227 ++#define __NR_lgetxattr 228 ++#define __NR_fgetxattr 229 ++#define __NR_listxattr 230 ++#define __NR_llistxattr 231 ++#define __NR_flistxattr 232 ++#define __NR_removexattr 233 ++#define __NR_lremovexattr 234 ++#define __NR_fremovexattr 235 + + + /* user-visible error numbers are in the range -1 - -122: see */ +Index: linux-2.4.19-pre1/include/asm-s390x/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-s390x/unistd.h 2001-10-11 20:43:38.000000000 +0400 ++++ linux-2.4.19-pre1/include/asm-s390x/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -181,6 +181,18 @@ + #define __NR_mincore 218 + #define __NR_madvise 219 + #define __NR_getdents64 220 ++#define __NR_setxattr 224 ++#define __NR_lsetxattr 225 ++#define __NR_fsetxattr 226 ++#define __NR_getxattr 227 ++#define __NR_lgetxattr 228 ++#define __NR_fgetxattr 229 ++#define __NR_listxattr 230 ++#define __NR_llistxattr 231 ++#define __NR_flistxattr 232 ++#define __NR_removexattr 233 ++#define __NR_lremovexattr 234 ++#define __NR_fremovexattr 235 + + + /* user-visible error numbers are in the range -1 - -122: see */ +Index: linux-2.4.19-pre1/include/asm-sparc/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-sparc/unistd.h 2001-10-21 21:36:54.000000000 +0400 ++++ linux-2.4.19-pre1/include/asm-sparc/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -184,24 +184,24 @@ + /* #define __NR_exportfs 166 SunOS Specific */ + #define __NR_mount 167 /* Common */ + #define __NR_ustat 168 /* Common */ +-/* #define __NR_semsys 169 SunOS Specific */ +-/* #define __NR_msgsys 170 SunOS Specific */ +-/* #define __NR_shmsys 171 SunOS Specific */ +-/* #define __NR_auditsys 172 SunOS Specific */ +-/* #define __NR_rfssys 173 SunOS Specific */ ++#define __NR_setxattr 169 /* SunOS: semsys */ ++#define __NR_lsetxattr 170 /* SunOS: msgsys */ ++#define __NR_fsetxattr 171 /* SunOS: shmsys */ ++#define __NR_getxattr 172 /* SunOS: auditsys */ ++#define __NR_lgetxattr 173 /* SunOS: rfssys */ + #define __NR_getdents 174 /* Common */ + #define __NR_setsid 175 /* Common */ + #define __NR_fchdir 176 /* Common */ +-/* #define __NR_fchroot 177 SunOS Specific */ +-/* #define __NR_vpixsys 178 SunOS Specific */ +-/* #define __NR_aioread 179 SunOS Specific */ +-/* #define __NR_aiowrite 180 SunOS Specific */ +-/* #define __NR_aiowait 181 SunOS Specific */ +-/* #define __NR_aiocancel 182 SunOS Specific */ ++#define __NR_fgetxattr 177 /* SunOS: fchroot */ ++#define __NR_listxattr 178 /* SunOS: vpixsys */ ++#define __NR_llistxattr 179 /* SunOS: aioread */ ++#define __NR_flistxattr 180 /* SunOS: aiowrite */ ++#define __NR_removexattr 181 /* SunOS: aiowait */ ++#define __NR_lremovexattr 182 /* SunOS: aiocancel */ + #define __NR_sigpending 183 /* Common */ + #define __NR_query_module 184 /* Linux Specific */ + #define __NR_setpgid 185 /* Common */ +-/* #define __NR_pathconf 186 SunOS Specific */ ++#define __NR_fremovexattr 186 /* SunOS: pathconf */ + /* #define __NR_fpathconf 187 SunOS Specific */ + /* #define __NR_sysconf 188 SunOS Specific */ + #define __NR_uname 189 /* Linux Specific */ +Index: linux-2.4.19-pre1/include/asm-sparc64/unistd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/asm-sparc64/unistd.h 2001-10-21 21:36:54.000000000 +0400 ++++ linux-2.4.19-pre1/include/asm-sparc64/unistd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -184,24 +184,24 @@ + /* #define __NR_exportfs 166 SunOS Specific */ + #define __NR_mount 167 /* Common */ + #define __NR_ustat 168 /* Common */ +-/* #define __NR_semsys 169 SunOS Specific */ +-/* #define __NR_msgsys 170 SunOS Specific */ +-/* #define __NR_shmsys 171 SunOS Specific */ +-/* #define __NR_auditsys 172 SunOS Specific */ +-/* #define __NR_rfssys 173 SunOS Specific */ ++#define __NR_setxattr 169 /* SunOS: semsys */ ++#define __NR_lsetxattr 170 /* SunOS: msgsys */ ++#define __NR_fsetxattr 171 /* SunOS: shmsys */ ++#define __NR_getxattr 172 /* SunOS: auditsys */ ++#define __NR_lgetxattr 173 /* SunOS: rfssys */ + #define __NR_getdents 174 /* Common */ + #define __NR_setsid 175 /* Common */ + #define __NR_fchdir 176 /* Common */ +-/* #define __NR_fchroot 177 SunOS Specific */ +-/* #define __NR_vpixsys 178 SunOS Specific */ +-/* #define __NR_aioread 179 SunOS Specific */ +-/* #define __NR_aiowrite 180 SunOS Specific */ +-/* #define __NR_aiowait 181 SunOS Specific */ +-/* #define __NR_aiocancel 182 SunOS Specific */ ++#define __NR_fgetxattr 177 /* SunOS: fchroot */ ++#define __NR_listxattr 178 /* SunOS: vpixsys */ ++#define __NR_llistxattr 179 /* SunOS: aioread */ ++#define __NR_flistxattr 180 /* SunOS: aiowrite */ ++#define __NR_removexattr 181 /* SunOS: aiowait */ ++#define __NR_lremovexattr 182 /* SunOS: aiocancel */ + #define __NR_sigpending 183 /* Common */ + #define __NR_query_module 184 /* Linux Specific */ + #define __NR_setpgid 185 /* Common */ +-/* #define __NR_pathconf 186 SunOS Specific */ ++#define __NR_fremovexattr 186 /* SunOS: pathconf */ + /* #define __NR_fpathconf 187 SunOS Specific */ + /* #define __NR_sysconf 188 SunOS Specific */ + #define __NR_uname 189 /* Linux Specific */ +Index: linux-2.4.19-pre1/include/linux/cache_def.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/cache_def.h 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/cache_def.h 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,15 @@ ++/* ++ * linux/cache_def.h ++ * Handling of caches defined in drivers, filesystems, ... ++ * ++ * Copyright (C) 2002 by Andreas Gruenbacher, ++ */ ++ ++struct cache_definition { ++ const char *name; ++ void (*shrink)(int, unsigned int); ++ struct list_head link; ++}; ++ ++extern void register_cache(struct cache_definition *); ++extern void unregister_cache(struct cache_definition *); +Index: linux-2.4.19-pre1/include/linux/errno.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/errno.h 2001-02-10 01:46:13.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/errno.h 2003-11-21 03:51:05.000000000 +0300 +@@ -23,4 +23,8 @@ + + #endif + ++/* Defined for extended attributes */ ++#define ENOATTR ENODATA /* No such attribute */ ++#define ENOTSUP EOPNOTSUPP /* Operation not supported */ ++ + #endif +Index: linux-2.4.19-pre1/include/linux/ext2_fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext2_fs.h 2001-11-22 22:46:52.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext2_fs.h 2003-11-21 03:51:05.000000000 +0300 +@@ -57,8 +57,6 @@ + */ + #define EXT2_BAD_INO 1 /* Bad blocks inode */ + #define EXT2_ROOT_INO 2 /* Root inode */ +-#define EXT2_ACL_IDX_INO 3 /* ACL inode */ +-#define EXT2_ACL_DATA_INO 4 /* ACL inode */ + #define EXT2_BOOT_LOADER_INO 5 /* Boot loader inode */ + #define EXT2_UNDEL_DIR_INO 6 /* Undelete directory inode */ + +@@ -86,7 +84,6 @@ + #else + # define EXT2_BLOCK_SIZE(s) (EXT2_MIN_BLOCK_SIZE << (s)->s_log_block_size) + #endif +-#define EXT2_ACLE_PER_BLOCK(s) (EXT2_BLOCK_SIZE(s) / sizeof (struct ext2_acl_entry)) + #define EXT2_ADDR_PER_BLOCK(s) (EXT2_BLOCK_SIZE(s) / sizeof (__u32)) + #ifdef __KERNEL__ + # define EXT2_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -121,28 +118,6 @@ + #endif + + /* +- * ACL structures +- */ +-struct ext2_acl_header /* Header of Access Control Lists */ +-{ +- __u32 aclh_size; +- __u32 aclh_file_count; +- __u32 aclh_acle_count; +- __u32 aclh_first_acle; +-}; +- +-struct ext2_acl_entry /* Access Control List Entry */ +-{ +- __u32 acle_size; +- __u16 acle_perms; /* Access permissions */ +- __u16 acle_type; /* Type of entry */ +- __u16 acle_tag; /* User or group identity */ +- __u16 acle_pad1; +- __u32 acle_next; /* Pointer on next entry for the */ +- /* same inode or on next free entry */ +-}; +- +-/* + * Structure of a blocks group descriptor + */ + struct ext2_group_desc +@@ -314,6 +289,7 @@ + #define EXT2_MOUNT_ERRORS_PANIC 0x0040 /* Panic on errors */ + #define EXT2_MOUNT_MINIX_DF 0x0080 /* Mimics the Minix statfs */ + #define EXT2_MOUNT_NO_UID32 0x0200 /* Disable 32-bit UIDs */ ++#define EXT2_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + + #define clear_opt(o, opt) o &= ~EXT2_MOUNT_##opt + #define set_opt(o, opt) o |= EXT2_MOUNT_##opt +@@ -397,6 +373,7 @@ + + #ifdef __KERNEL__ + #define EXT2_SB(sb) (&((sb)->u.ext2_sb)) ++#define EXT2_I(inode) (&((inode)->u.ext2_i)) + #else + /* Assume that user mode programs are passing in an ext2fs superblock, not + * a kernel struct super_block. This will allow us to call the feature-test +@@ -466,7 +443,7 @@ + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 + #define EXT2_FEATURE_INCOMPAT_ANY 0xffffffff + +-#define EXT2_FEATURE_COMPAT_SUPP 0 ++#define EXT2_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR + #define EXT2_FEATURE_INCOMPAT_SUPP EXT2_FEATURE_INCOMPAT_FILETYPE + #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER| \ + EXT2_FEATURE_RO_COMPAT_LARGE_FILE| \ +@@ -623,8 +600,10 @@ + + /* namei.c */ + extern struct inode_operations ext2_dir_inode_operations; ++extern struct inode_operations ext2_special_inode_operations; + + /* symlink.c */ ++extern struct inode_operations ext2_symlink_inode_operations; + extern struct inode_operations ext2_fast_symlink_inode_operations; + + #endif /* __KERNEL__ */ +Index: linux-2.4.19-pre1/include/linux/ext2_xattr.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext2_xattr.h 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext2_xattr.h 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,157 @@ ++/* ++ File: linux/ext2_xattr.h ++ ++ On-disk format of extended attributes for the ext2 filesystem. ++ ++ (C) 2001 Andreas Gruenbacher, ++*/ ++ ++#include ++#include ++#include ++ ++/* Magic value in attribute blocks */ ++#define EXT2_XATTR_MAGIC 0xEA020000 ++ ++/* Maximum number of references to one attribute block */ ++#define EXT2_XATTR_REFCOUNT_MAX 1024 ++ ++/* Name indexes */ ++#define EXT2_XATTR_INDEX_MAX 10 ++#define EXT2_XATTR_INDEX_USER 1 ++#define EXT2_XATTR_INDEX_POSIX_ACL_ACCESS 2 ++#define EXT2_XATTR_INDEX_POSIX_ACL_DEFAULT 3 ++ ++struct ext2_xattr_header { ++ __u32 h_magic; /* magic number for identification */ ++ __u32 h_refcount; /* reference count */ ++ __u32 h_blocks; /* number of disk blocks used */ ++ __u32 h_hash; /* hash value of all attributes */ ++ __u32 h_reserved[4]; /* zero right now */ ++}; ++ ++struct ext2_xattr_entry { ++ __u8 e_name_len; /* length of name */ ++ __u8 e_name_index; /* attribute name index */ ++ __u16 e_value_offs; /* offset in disk block of value */ ++ __u32 e_value_block; /* disk block attribute is stored on (n/i) */ ++ __u32 e_value_size; /* size of attribute value */ ++ __u32 e_hash; /* hash value of name and value */ ++ char e_name[0]; /* attribute name */ ++}; ++ ++#define EXT2_XATTR_PAD_BITS 2 ++#define EXT2_XATTR_PAD (1<e_name_len)) ) ++#define EXT2_XATTR_SIZE(size) \ ++ (((size) + EXT2_XATTR_ROUND) & ~EXT2_XATTR_ROUND) ++ ++#ifdef __KERNEL__ ++ ++# ifdef CONFIG_EXT2_FS_XATTR ++ ++struct ext2_xattr_handler { ++ char *prefix; ++ size_t (*list)(char *list, struct inode *inode, const char *name, ++ int name_len); ++ int (*get)(struct inode *inode, const char *name, void *buffer, ++ size_t size); ++ int (*set)(struct inode *inode, const char *name, const void *buffer, ++ size_t size, int flags); ++}; ++ ++extern int ext2_xattr_register(int, struct ext2_xattr_handler *); ++extern void ext2_xattr_unregister(int, struct ext2_xattr_handler *); ++ ++extern int ext2_setxattr(struct dentry *, const char *, const void *, size_t, int); ++extern ssize_t ext2_getxattr(struct dentry *, const char *, void *, size_t); ++extern ssize_t ext2_listxattr(struct dentry *, char *, size_t); ++extern int ext2_removexattr(struct dentry *, const char *); ++ ++extern int ext2_xattr_get(struct inode *, int, const char *, void *, size_t); ++extern int ext2_xattr_list(struct inode *, char *, size_t); ++extern int ext2_xattr_set(struct inode *, int, const char *, const void *, size_t, int); ++ ++extern void ext2_xattr_delete_inode(struct inode *); ++extern void ext2_xattr_put_super(struct super_block *); ++ ++extern int init_ext2_xattr(void) __init; ++extern void exit_ext2_xattr(void); ++ ++# else /* CONFIG_EXT2_FS_XATTR */ ++# define ext2_setxattr NULL ++# define ext2_getxattr NULL ++# define ext2_listxattr NULL ++# define ext2_removexattr NULL ++ ++static inline int ++ext2_xattr_get(struct inode *inode, int name_index, ++ const char *name, void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext2_xattr_list(struct inode *inode, char *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext2_xattr_set(struct inode *inode, int name_index, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline void ++ext2_xattr_delete_inode(struct inode *inode) ++{ ++} ++ ++static inline void ++ext2_xattr_put_super(struct super_block *sb) ++{ ++} ++ ++static inline int ++init_ext2_xattr(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext2_xattr(void) ++{ ++} ++ ++# endif /* CONFIG_EXT2_FS_XATTR */ ++ ++# ifdef CONFIG_EXT2_FS_XATTR_USER ++ ++extern int init_ext2_xattr_user(void) __init; ++extern void exit_ext2_xattr_user(void); ++ ++# else /* CONFIG_EXT2_FS_XATTR_USER */ ++ ++static inline int ++init_ext2_xattr_user(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext2_xattr_user(void) ++{ ++} ++ ++# endif /* CONFIG_EXT2_FS_XATTR_USER */ ++ ++#endif /* __KERNEL__ */ ++ +Index: linux-2.4.19-pre1/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_fs.h 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_fs.h 2003-11-21 03:51:05.000000000 +0300 +@@ -63,8 +63,6 @@ + */ + #define EXT3_BAD_INO 1 /* Bad blocks inode */ + #define EXT3_ROOT_INO 2 /* Root inode */ +-#define EXT3_ACL_IDX_INO 3 /* ACL inode */ +-#define EXT3_ACL_DATA_INO 4 /* ACL inode */ + #define EXT3_BOOT_LOADER_INO 5 /* Boot loader inode */ + #define EXT3_UNDEL_DIR_INO 6 /* Undelete directory inode */ + #define EXT3_RESIZE_INO 7 /* Reserved group descriptors inode */ +@@ -94,7 +92,6 @@ + #else + # define EXT3_BLOCK_SIZE(s) (EXT3_MIN_BLOCK_SIZE << (s)->s_log_block_size) + #endif +-#define EXT3_ACLE_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (struct ext3_acl_entry)) + #define EXT3_ADDR_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (__u32)) + #ifdef __KERNEL__ + # define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -129,28 +126,6 @@ + #endif + + /* +- * ACL structures +- */ +-struct ext3_acl_header /* Header of Access Control Lists */ +-{ +- __u32 aclh_size; +- __u32 aclh_file_count; +- __u32 aclh_acle_count; +- __u32 aclh_first_acle; +-}; +- +-struct ext3_acl_entry /* Access Control List Entry */ +-{ +- __u32 acle_size; +- __u16 acle_perms; /* Access permissions */ +- __u16 acle_type; /* Type of entry */ +- __u16 acle_tag; /* User or group identity */ +- __u16 acle_pad1; +- __u32 acle_next; /* Pointer on next entry for the */ +- /* same inode or on next free entry */ +-}; +- +-/* + * Structure of a blocks group descriptor + */ + struct ext3_group_desc +@@ -344,6 +319,7 @@ + #define EXT3_MOUNT_WRITEBACK_DATA 0x0C00 /* No data ordering */ + #define EXT3_MOUNT_UPDATE_JOURNAL 0x1000 /* Update the journal format */ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ ++#define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H +@@ -520,7 +496,7 @@ + #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ + +-#define EXT3_FEATURE_COMPAT_SUPP 0 ++#define EXT3_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR + #define EXT3_FEATURE_INCOMPAT_SUPP (EXT3_FEATURE_INCOMPAT_FILETYPE| \ + EXT3_FEATURE_INCOMPAT_RECOVER) + #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER| \ +@@ -703,6 +679,7 @@ + extern unsigned long ext3_count_free (struct buffer_head *, unsigned); + + /* inode.c */ ++extern int ext3_forget(handle_t *, int, struct inode *, struct buffer_head *, int); + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + +@@ -771,8 +748,10 @@ + + /* namei.c */ + extern struct inode_operations ext3_dir_inode_operations; ++extern struct inode_operations ext3_special_inode_operations; + + /* symlink.c */ ++extern struct inode_operations ext3_symlink_inode_operations; + extern struct inode_operations ext3_fast_symlink_inode_operations; + + +Index: linux-2.4.19-pre1/include/linux/ext3_jbd.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_jbd.h 2003-11-21 03:51:02.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_jbd.h 2003-11-21 03:51:05.000000000 +0300 +@@ -30,13 +30,19 @@ + + #define EXT3_SINGLEDATA_TRANS_BLOCKS 8 + ++/* Extended attributes may touch two data buffers, two bitmap buffers, ++ * and two group and summaries. */ ++ ++#define EXT3_XATTR_TRANS_BLOCKS 8 ++ + /* Define the minimum size for a transaction which modifies data. This + * needs to take into account the fact that we may end up modifying two + * quota files too (one for the group, one for the user quota). The + * superblock only gets updated once, of course, so don't bother + * counting that again for the quota updates. */ + +-#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS - 2) ++#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS + \ ++ EXT3_XATTR_TRANS_BLOCKS - 2) + + extern int ext3_writepage_trans_blocks(struct inode *inode); + +Index: linux-2.4.19-pre1/include/linux/ext3_xattr.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/ext3_xattr.h 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/ext3_xattr.h 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,157 @@ ++/* ++ File: linux/ext3_xattr.h ++ ++ On-disk format of extended attributes for the ext3 filesystem. ++ ++ (C) 2001 Andreas Gruenbacher, ++*/ ++ ++#include ++#include ++#include ++ ++/* Magic value in attribute blocks */ ++#define EXT3_XATTR_MAGIC 0xEA020000 ++ ++/* Maximum number of references to one attribute block */ ++#define EXT3_XATTR_REFCOUNT_MAX 1024 ++ ++/* Name indexes */ ++#define EXT3_XATTR_INDEX_MAX 10 ++#define EXT3_XATTR_INDEX_USER 1 ++#define EXT3_XATTR_INDEX_POSIX_ACL_ACCESS 2 ++#define EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT 3 ++ ++struct ext3_xattr_header { ++ __u32 h_magic; /* magic number for identification */ ++ __u32 h_refcount; /* reference count */ ++ __u32 h_blocks; /* number of disk blocks used */ ++ __u32 h_hash; /* hash value of all attributes */ ++ __u32 h_reserved[4]; /* zero right now */ ++}; ++ ++struct ext3_xattr_entry { ++ __u8 e_name_len; /* length of name */ ++ __u8 e_name_index; /* attribute name index */ ++ __u16 e_value_offs; /* offset in disk block of value */ ++ __u32 e_value_block; /* disk block attribute is stored on (n/i) */ ++ __u32 e_value_size; /* size of attribute value */ ++ __u32 e_hash; /* hash value of name and value */ ++ char e_name[0]; /* attribute name */ ++}; ++ ++#define EXT3_XATTR_PAD_BITS 2 ++#define EXT3_XATTR_PAD (1<e_name_len)) ) ++#define EXT3_XATTR_SIZE(size) \ ++ (((size) + EXT3_XATTR_ROUND) & ~EXT3_XATTR_ROUND) ++ ++#ifdef __KERNEL__ ++ ++# ifdef CONFIG_EXT3_FS_XATTR ++ ++struct ext3_xattr_handler { ++ char *prefix; ++ size_t (*list)(char *list, struct inode *inode, const char *name, ++ int name_len); ++ int (*get)(struct inode *inode, const char *name, void *buffer, ++ size_t size); ++ int (*set)(struct inode *inode, const char *name, const void *buffer, ++ size_t size, int flags); ++}; ++ ++extern int ext3_xattr_register(int, struct ext3_xattr_handler *); ++extern void ext3_xattr_unregister(int, struct ext3_xattr_handler *); ++ ++extern int ext3_setxattr(struct dentry *, const char *, const void *, size_t, int); ++extern ssize_t ext3_getxattr(struct dentry *, const char *, void *, size_t); ++extern ssize_t ext3_listxattr(struct dentry *, char *, size_t); ++extern int ext3_removexattr(struct dentry *, const char *); ++ ++extern int ext3_xattr_get(struct inode *, int, const char *, void *, size_t); ++extern int ext3_xattr_list(struct inode *, char *, size_t); ++extern int ext3_xattr_set(handle_t *handle, struct inode *, int, const char *, const void *, size_t, int); ++ ++extern void ext3_xattr_delete_inode(handle_t *, struct inode *); ++extern void ext3_xattr_put_super(struct super_block *); ++ ++extern int init_ext3_xattr(void) __init; ++extern void exit_ext3_xattr(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR */ ++# define ext3_setxattr NULL ++# define ext3_getxattr NULL ++# define ext3_listxattr NULL ++# define ext3_removexattr NULL ++ ++static inline int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_list(struct inode *inode, void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline void ++ext3_xattr_delete_inode(handle_t *handle, struct inode *inode) ++{ ++} ++ ++static inline void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++} ++ ++static inline int ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr(void) ++{ ++} ++ ++# endif /* CONFIG_EXT3_FS_XATTR */ ++ ++# ifdef CONFIG_EXT3_FS_XATTR_USER ++ ++extern int init_ext3_xattr_user(void) __init; ++extern void exit_ext3_xattr_user(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++static inline int ++init_ext3_xattr_user(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr_user(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++#endif /* __KERNEL__ */ ++ +Index: linux-2.4.19-pre1/include/linux/fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs.h 2003-11-21 03:51:00.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs.h 2003-11-21 03:51:05.000000000 +0300 +@@ -872,6 +872,10 @@ + int (*setattr) (struct dentry *, struct iattr *); + int (*setattr_raw) (struct inode *, struct iattr *); + int (*getattr) (struct dentry *, struct iattr *); ++ int (*setxattr) (struct dentry *, const char *, const void *, size_t, int); ++ ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); ++ ssize_t (*listxattr) (struct dentry *, char *, size_t); ++ int (*removexattr) (struct dentry *, const char *); + }; + + struct seq_file; +Index: linux-2.4.19-pre1/include/linux/mbcache.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/mbcache.h 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/mbcache.h 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,69 @@ ++/* ++ File: linux/mbcache.h ++ ++ (C) 2001 by Andreas Gruenbacher, ++*/ ++ ++/* Hardwire the number of additional indexes */ ++#define MB_CACHE_INDEXES_COUNT 1 ++ ++struct mb_cache_entry; ++ ++struct mb_cache_op { ++ int (*free)(struct mb_cache_entry *, int); ++}; ++ ++struct mb_cache { ++ struct list_head c_cache_list; ++ const char *c_name; ++ struct mb_cache_op c_op; ++ atomic_t c_entry_count; ++ int c_bucket_count; ++#ifndef MB_CACHE_INDEXES_COUNT ++ int c_indexes_count; ++#endif ++ kmem_cache_t *c_entry_cache; ++ struct list_head *c_block_hash; ++ struct list_head *c_indexes_hash[0]; ++}; ++ ++struct mb_cache_entry_index { ++ struct list_head o_list; ++ unsigned int o_key; ++}; ++ ++struct mb_cache_entry { ++ struct list_head e_lru_list; ++ struct mb_cache *e_cache; ++ atomic_t e_used; ++ kdev_t e_dev; ++ unsigned long e_block; ++ struct list_head e_block_list; ++ struct mb_cache_entry_index e_indexes[0]; ++}; ++ ++/* Functions on caches */ ++ ++struct mb_cache * mb_cache_create(const char *, struct mb_cache_op *, size_t, ++ int, int); ++void mb_cache_shrink(struct mb_cache *, kdev_t); ++void mb_cache_destroy(struct mb_cache *); ++ ++/* Functions on cache entries */ ++ ++struct mb_cache_entry *mb_cache_entry_alloc(struct mb_cache *); ++int mb_cache_entry_insert(struct mb_cache_entry *, kdev_t, unsigned long, ++ unsigned int[]); ++void mb_cache_entry_rehash(struct mb_cache_entry *, unsigned int[]); ++void mb_cache_entry_release(struct mb_cache_entry *); ++void mb_cache_entry_takeout(struct mb_cache_entry *); ++void mb_cache_entry_free(struct mb_cache_entry *); ++struct mb_cache_entry *mb_cache_entry_dup(struct mb_cache_entry *); ++struct mb_cache_entry *mb_cache_entry_get(struct mb_cache *, kdev_t, ++ unsigned long); ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++struct mb_cache_entry *mb_cache_entry_find_first(struct mb_cache *cache, int, ++ kdev_t, unsigned int); ++struct mb_cache_entry *mb_cache_entry_find_next(struct mb_cache_entry *, int, ++ kdev_t, unsigned int); ++#endif +Index: linux-2.4.19-pre1/kernel/ksyms.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/ksyms.c 2003-11-21 03:50:59.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/ksyms.c 2003-11-21 03:51:05.000000000 +0300 +@@ -11,6 +11,7 @@ + + #include + #include ++#include + #include + #include + #include +@@ -88,6 +89,7 @@ + EXPORT_SYMBOL(exit_files); + EXPORT_SYMBOL(exit_fs); + EXPORT_SYMBOL(exit_sighand); ++EXPORT_SYMBOL(copy_fs_struct); + + /* internal kernel memory management */ + EXPORT_SYMBOL(_alloc_pages); +@@ -104,6 +106,8 @@ + EXPORT_SYMBOL(kmem_cache_shrink); + EXPORT_SYMBOL(kmem_cache_alloc); + EXPORT_SYMBOL(kmem_cache_free); ++EXPORT_SYMBOL(register_cache); ++EXPORT_SYMBOL(unregister_cache); + EXPORT_SYMBOL(kmalloc); + EXPORT_SYMBOL(kfree); + EXPORT_SYMBOL(vfree); +Index: linux-2.4.19-pre1/mm/vmscan.c +=================================================================== +--- linux-2.4.19-pre1.orig/mm/vmscan.c 2003-11-20 19:01:38.000000000 +0300 ++++ linux-2.4.19-pre1/mm/vmscan.c 2003-11-21 03:51:05.000000000 +0300 +@@ -15,6 +15,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -32,6 +33,39 @@ + */ + #define DEF_PRIORITY (6) + ++static DECLARE_MUTEX(other_caches_sem); ++static LIST_HEAD(cache_definitions); ++ ++void register_cache(struct cache_definition *cache) ++{ ++ down(&other_caches_sem); ++ list_add(&cache->link, &cache_definitions); ++ up(&other_caches_sem); ++} ++ ++void unregister_cache(struct cache_definition *cache) ++{ ++ down(&other_caches_sem); ++ list_del(&cache->link); ++ up(&other_caches_sem); ++} ++ ++static void shrink_other_caches(unsigned int priority, int gfp_mask) ++{ ++ struct list_head *p; ++ ++ if (down_trylock(&other_caches_sem)) ++ return; ++ ++ list_for_each_prev(p, &cache_definitions) { ++ struct cache_definition *cache = ++ list_entry(p, struct cache_definition, link); ++ ++ cache->shrink(priority, gfp_mask); ++ } ++ up(&other_caches_sem); ++} ++ + /* + * The swap-out function returns 1 if it successfully + * scanned all the pages it was asked to (`count'). +@@ -578,6 +612,7 @@ + + shrink_dcache_memory(priority, gfp_mask); + shrink_icache_memory(priority, gfp_mask); ++ shrink_other_caches(priority, gfp_mask); + #ifdef CONFIG_QUOTA + shrink_dqcache_memory(DEF_PRIORITY, gfp_mask); + #endif +Index: linux-2.4.19-pre1/fs/ext3/ext3-exports.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/ext3/ext3-exports.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/ext3/ext3-exports.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,13 @@ ++#include ++#include ++#include ++#include ++#include ++ ++EXPORT_SYMBOL(ext3_force_commit); ++EXPORT_SYMBOL(ext3_bread); ++EXPORT_SYMBOL(ext3_xattr_register); ++EXPORT_SYMBOL(ext3_xattr_unregister); ++EXPORT_SYMBOL(ext3_xattr_get); ++EXPORT_SYMBOL(ext3_xattr_list); ++EXPORT_SYMBOL(ext3_xattr_set); +Index: linux-2.4.19-pre1/include/linux/xattr.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/xattr.h 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/xattr.h 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,15 @@ ++/* ++ File: linux/xattr.h ++ ++ Extended attributes handling. ++ ++ Copyright (C) 2001 by Andreas Gruenbacher ++ Copyright (c) 2001-2002 Silicon Graphics, Inc. All Rights Reserved. ++*/ ++#ifndef _LINUX_XATTR_H ++#define _LINUX_XATTR_H ++ ++#define XATTR_CREATE 0x1 /* set the value, fail if attr already exists */ ++#define XATTR_REPLACE 0x2 /* set the value, fail if attr does not exist */ ++ ++#endif /* _LINUX_XATTR_H */ +Index: linux-2.4.19-pre1/arch/i386/kernel/entry.S +=================================================================== +--- linux-2.4.19-pre1.orig/arch/i386/kernel/entry.S 2003-11-21 03:38:55.000000000 +0300 ++++ linux-2.4.19-pre1/arch/i386/kernel/entry.S 2003-11-21 03:51:05.000000000 +0300 +@@ -622,18 +622,18 @@ + .long SYMBOL_NAME(sys_ni_syscall) /* Reserved for Security */ + .long SYMBOL_NAME(sys_gettid) + .long SYMBOL_NAME(sys_readahead) /* 225 */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for setxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for lsetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for fsetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for getxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* 230 reserved for lgetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for fgetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for listxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for llistxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for flistxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* 235 reserved for removexattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for lremovexattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* reserved for fremovexattr */ ++ .long SYMBOL_NAME(sys_setxattr) ++ .long SYMBOL_NAME(sys_lsetxattr) ++ .long SYMBOL_NAME(sys_fsetxattr) ++ .long SYMBOL_NAME(sys_getxattr) ++ .long SYMBOL_NAME(sys_lgetxattr) /* 230 */ ++ .long SYMBOL_NAME(sys_fgetxattr) ++ .long SYMBOL_NAME(sys_listxattr) ++ .long SYMBOL_NAME(sys_llistxattr) ++ .long SYMBOL_NAME(sys_flistxattr) ++ .long SYMBOL_NAME(sys_removexattr) /* 235 */ ++ .long SYMBOL_NAME(sys_lremovexattr) ++ .long SYMBOL_NAME(sys_fremovexattr) + + .rept NR_syscalls-(.-sys_call_table)/4 + .long SYMBOL_NAME(sys_ni_syscall) +Index: linux-2.4.19-pre1/fs/xattr.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/xattr.c 2003-11-21 03:51:05.000000000 +0300 ++++ linux-2.4.19-pre1/fs/xattr.c 2003-11-21 03:51:05.000000000 +0300 +@@ -0,0 +1,355 @@ ++/* ++ File: fs/xattr.c ++ ++ Extended attribute handling. ++ ++ Copyright (C) 2001 by Andreas Gruenbacher ++ Copyright (C) 2001 SGI - Silicon Graphics, Inc ++ */ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++/* ++ * Extended attribute memory allocation wrappers, originally ++ * based on the Intermezzo PRESTO_ALLOC/PRESTO_FREE macros. ++ * The vmalloc use here is very uncommon - extended attributes ++ * are supposed to be small chunks of metadata, and it is quite ++ * unusual to have very many extended attributes, so lists tend ++ * to be quite short as well. The 64K upper limit is derived ++ * from the extended attribute size limit used by XFS. ++ * Intentionally allow zero @size for value/list size requests. ++ */ ++static void * ++xattr_alloc(size_t size, size_t limit) ++{ ++ void *ptr; ++ ++ if (size > limit) ++ return ERR_PTR(-E2BIG); ++ ++ if (!size) /* size request, no buffer is needed */ ++ return NULL; ++ else if (size <= PAGE_SIZE) ++ ptr = kmalloc((unsigned long) size, GFP_KERNEL); ++ else ++ ptr = vmalloc((unsigned long) size); ++ if (!ptr) ++ return ERR_PTR(-ENOMEM); ++ return ptr; ++} ++ ++static void ++xattr_free(void *ptr, size_t size) ++{ ++ if (!size) /* size request, no buffer was needed */ ++ return; ++ else if (size <= PAGE_SIZE) ++ kfree(ptr); ++ else ++ vfree(ptr); ++} ++ ++/* ++ * Extended attribute SET operations ++ */ ++static long ++setxattr(struct dentry *d, char *name, void *value, size_t size, int flags) ++{ ++ int error; ++ void *kvalue; ++ char kname[XATTR_NAME_MAX + 1]; ++ ++ if (flags & ~(XATTR_CREATE|XATTR_REPLACE)) ++ return -EINVAL; ++ ++ error = strncpy_from_user(kname, name, sizeof(kname)); ++ if (error == 0 || error == sizeof(kname)) ++ error = -ERANGE; ++ if (error < 0) ++ return error; ++ ++ kvalue = xattr_alloc(size, XATTR_SIZE_MAX); ++ if (IS_ERR(kvalue)) ++ return PTR_ERR(kvalue); ++ ++ if (size > 0 && copy_from_user(kvalue, value, size)) { ++ xattr_free(kvalue, size); ++ return -EFAULT; ++ } ++ ++ error = -EOPNOTSUPP; ++ if (d->d_inode->i_op && d->d_inode->i_op->setxattr) { ++ down(&d->d_inode->i_sem); ++ lock_kernel(); ++ error = d->d_inode->i_op->setxattr(d, kname, kvalue, size, flags); ++ unlock_kernel(); ++ up(&d->d_inode->i_sem); ++ } ++ ++ xattr_free(kvalue, size); ++ return error; ++} ++ ++asmlinkage long ++sys_setxattr(char *path, char *name, void *value, size_t size, int flags) ++{ ++ struct nameidata nd; ++ int error; ++ ++ error = user_path_walk(path, &nd); ++ if (error) ++ return error; ++ error = setxattr(nd.dentry, name, value, size, flags); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage long ++sys_lsetxattr(char *path, char *name, void *value, size_t size, int flags) ++{ ++ struct nameidata nd; ++ int error; ++ ++ error = user_path_walk_link(path, &nd); ++ if (error) ++ return error; ++ error = setxattr(nd.dentry, name, value, size, flags); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage long ++sys_fsetxattr(int fd, char *name, void *value, size_t size, int flags) ++{ ++ struct file *f; ++ int error = -EBADF; ++ ++ f = fget(fd); ++ if (!f) ++ return error; ++ error = setxattr(f->f_dentry, name, value, size, flags); ++ fput(f); ++ return error; ++} ++ ++/* ++ * Extended attribute GET operations ++ */ ++static ssize_t ++getxattr(struct dentry *d, char *name, void *value, size_t size) ++{ ++ ssize_t error; ++ void *kvalue; ++ char kname[XATTR_NAME_MAX + 1]; ++ ++ error = strncpy_from_user(kname, name, sizeof(kname)); ++ if (error == 0 || error == sizeof(kname)) ++ error = -ERANGE; ++ if (error < 0) ++ return error; ++ ++ kvalue = xattr_alloc(size, XATTR_SIZE_MAX); ++ if (IS_ERR(kvalue)) ++ return PTR_ERR(kvalue); ++ ++ error = -EOPNOTSUPP; ++ if (d->d_inode->i_op && d->d_inode->i_op->getxattr) { ++ down(&d->d_inode->i_sem); ++ lock_kernel(); ++ error = d->d_inode->i_op->getxattr(d, kname, kvalue, size); ++ unlock_kernel(); ++ up(&d->d_inode->i_sem); ++ } ++ ++ if (kvalue && error > 0) ++ if (copy_to_user(value, kvalue, error)) ++ error = -EFAULT; ++ xattr_free(kvalue, size); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_getxattr(char *path, char *name, void *value, size_t size) ++{ ++ struct nameidata nd; ++ ssize_t error; ++ ++ error = user_path_walk(path, &nd); ++ if (error) ++ return error; ++ error = getxattr(nd.dentry, name, value, size); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_lgetxattr(char *path, char *name, void *value, size_t size) ++{ ++ struct nameidata nd; ++ ssize_t error; ++ ++ error = user_path_walk_link(path, &nd); ++ if (error) ++ return error; ++ error = getxattr(nd.dentry, name, value, size); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_fgetxattr(int fd, char *name, void *value, size_t size) ++{ ++ struct file *f; ++ ssize_t error = -EBADF; ++ ++ f = fget(fd); ++ if (!f) ++ return error; ++ error = getxattr(f->f_dentry, name, value, size); ++ fput(f); ++ return error; ++} ++ ++/* ++ * Extended attribute LIST operations ++ */ ++static ssize_t ++listxattr(struct dentry *d, char *list, size_t size) ++{ ++ ssize_t error; ++ char *klist; ++ ++ klist = (char *)xattr_alloc(size, XATTR_LIST_MAX); ++ if (IS_ERR(klist)) ++ return PTR_ERR(klist); ++ ++ error = -EOPNOTSUPP; ++ if (d->d_inode->i_op && d->d_inode->i_op->listxattr) { ++ down(&d->d_inode->i_sem); ++ lock_kernel(); ++ error = d->d_inode->i_op->listxattr(d, klist, size); ++ unlock_kernel(); ++ up(&d->d_inode->i_sem); ++ } ++ ++ if (klist && error > 0) ++ if (copy_to_user(list, klist, error)) ++ error = -EFAULT; ++ xattr_free(klist, size); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_listxattr(char *path, char *list, size_t size) ++{ ++ struct nameidata nd; ++ ssize_t error; ++ ++ error = user_path_walk(path, &nd); ++ if (error) ++ return error; ++ error = listxattr(nd.dentry, list, size); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_llistxattr(char *path, char *list, size_t size) ++{ ++ struct nameidata nd; ++ ssize_t error; ++ ++ error = user_path_walk_link(path, &nd); ++ if (error) ++ return error; ++ error = listxattr(nd.dentry, list, size); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage ssize_t ++sys_flistxattr(int fd, char *list, size_t size) ++{ ++ struct file *f; ++ ssize_t error = -EBADF; ++ ++ f = fget(fd); ++ if (!f) ++ return error; ++ error = listxattr(f->f_dentry, list, size); ++ fput(f); ++ return error; ++} ++ ++/* ++ * Extended attribute REMOVE operations ++ */ ++static long ++removexattr(struct dentry *d, char *name) ++{ ++ int error; ++ char kname[XATTR_NAME_MAX + 1]; ++ ++ error = strncpy_from_user(kname, name, sizeof(kname)); ++ if (error == 0 || error == sizeof(kname)) ++ error = -ERANGE; ++ if (error < 0) ++ return error; ++ ++ error = -EOPNOTSUPP; ++ if (d->d_inode->i_op && d->d_inode->i_op->removexattr) { ++ down(&d->d_inode->i_sem); ++ lock_kernel(); ++ error = d->d_inode->i_op->removexattr(d, kname); ++ unlock_kernel(); ++ up(&d->d_inode->i_sem); ++ } ++ return error; ++} ++ ++asmlinkage long ++sys_removexattr(char *path, char *name) ++{ ++ struct nameidata nd; ++ int error; ++ ++ error = user_path_walk(path, &nd); ++ if (error) ++ return error; ++ error = removexattr(nd.dentry, name); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage long ++sys_lremovexattr(char *path, char *name) ++{ ++ struct nameidata nd; ++ int error; ++ ++ error = user_path_walk_link(path, &nd); ++ if (error) ++ return error; ++ error = removexattr(nd.dentry, name); ++ path_release(&nd); ++ return error; ++} ++ ++asmlinkage long ++sys_fremovexattr(int fd, char *name) ++{ ++ struct file *f; ++ int error = -EBADF; ++ ++ f = fget(fd); ++ if (!f) ++ return error; ++ error = removexattr(f->f_dentry, name); ++ fput(f); ++ return error; ++} +Index: linux-2.4.19-pre1/include/linux/kernel.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/kernel.h 2003-11-21 02:25:34.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/kernel.h 2003-11-21 03:51:05.000000000 +0300 +@@ -11,6 +11,7 @@ + #include + #include + #include ++#include + + /* Optimization barrier */ + /* The "volatile" is due to gcc bugs */ diff --git a/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch b/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch new file mode 100644 index 0000000..26d3af9 --- /dev/null +++ b/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch @@ -0,0 +1,47 @@ + ext2/super.c | 3 +-- + ext3/ext3-exports.c | 13 +++++++++++++ + 2 files changed, 14 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/fs/ext2/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext2/super.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/ext2/super.c Sun Nov 16 00:40:59 2003 +@@ -70,6 +70,7 @@ + { + va_list args; + ++ ext2_xattr_put_super(sb); + if (!(sb->s_flags & MS_RDONLY)) { + sb->u.ext2_sb.s_mount_state |= EXT2_ERROR_FS; + sb->u.ext2_sb.s_es->s_state = +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 00:40:59 2003 +@@ -1822,8 +1828,6 @@ + exit_ext3_xattr(); + } + +-EXPORT_SYMBOL(ext3_force_commit); +-EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +Index: linux-2.4.19.SuSE/fs/ext3/ext3-exports.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ext3-exports.c Sun Nov 16 00:40:58 2003 ++++ linux-2.4.19.SuSE/fs/ext3/ext3-exports.c Sun Nov 16 00:40:59 2003 +@@ -0,0 +1,13 @@ ++#include ++#include ++#include ++#include ++#include ++ ++EXPORT_SYMBOL(ext3_force_commit); ++EXPORT_SYMBOL(ext3_bread); ++EXPORT_SYMBOL(ext3_xattr_register); ++EXPORT_SYMBOL(ext3_xattr_unregister); ++EXPORT_SYMBOL(ext3_xattr_get); ++EXPORT_SYMBOL(ext3_xattr_list); ++EXPORT_SYMBOL(ext3_xattr_set); diff --git a/lustre/kernel_patches/patches/nfs_export_kernel-2.4.19-pre1.patch b/lustre/kernel_patches/patches/nfs_export_kernel-2.4.19-pre1.patch new file mode 100644 index 0000000..7278b23 --- /dev/null +++ b/lustre/kernel_patches/patches/nfs_export_kernel-2.4.19-pre1.patch @@ -0,0 +1,757 @@ + fs/Makefile | 3 + fs/file_table.c | 11 ++ + fs/inode.c | 23 ++++- + fs/namei.c | 12 ++ + fs/nfsd/export.c | 5 + + fs/nfsd/nfsfh.c | 65 +++++++++++++- + fs/nfsd/vfs.c | 240 ++++++++++++++++++++++++++++++++++++++++++++++++----- + include/linux/fs.h | 10 ++ + kernel/ksyms.c | 2 + 9 files changed, 337 insertions(+), 34 deletions(-) + +Index: linux-2.4.19-pre1/fs/file_table.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/file_table.c 2001-09-18 00:16:30.000000000 +0400 ++++ linux-2.4.19-pre1/fs/file_table.c 2003-11-21 02:02:35.000000000 +0300 +@@ -82,7 +82,8 @@ + * and call the open function (if any). The caller must verify that + * inode->i_fop is not NULL. + */ +-int init_private_file(struct file *filp, struct dentry *dentry, int mode) ++int init_private_file_it(struct file *filp, struct dentry *dentry, int mode, ++ struct lookup_intent *it) + { + memset(filp, 0, sizeof(*filp)); + filp->f_mode = mode; +@@ -90,12 +91,20 @@ + filp->f_dentry = dentry; + filp->f_uid = current->fsuid; + filp->f_gid = current->fsgid; ++ if (it) ++ filp->f_it = it; + filp->f_op = dentry->d_inode->i_fop; + if (filp->f_op->open) + return filp->f_op->open(dentry->d_inode, filp); + else + return 0; + } ++EXPORT_SYMBOL(init_private_file_it); ++ ++int init_private_file(struct file *filp, struct dentry *dentry, int mode) ++{ ++ return init_private_file_it(filp, dentry, mode, NULL); ++} + + void fput(struct file * file) + { +Index: linux-2.4.19-pre1/fs/inode.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/inode.c 2003-11-20 19:24:08.000000000 +0300 ++++ linux-2.4.19-pre1/fs/inode.c 2003-11-21 02:02:35.000000000 +0300 +@@ -971,9 +971,10 @@ + } + + +-struct inode *iget4(struct super_block *sb, unsigned long ino, find_inode_t find_actor, void *opaque) ++static inline struct inode *ifind(struct super_block *sb, unsigned long ino, ++ struct list_head *head, ++ find_inode_t find_actor, void *opaque) + { +- struct list_head * head = inode_hashtable + hash(sb,ino); + struct inode * inode; + + spin_lock(&inode_lock); +@@ -986,6 +987,24 @@ + } + spin_unlock(&inode_lock); + ++ return NULL; ++} ++ ++struct inode *ilookup4(struct super_block *sb, unsigned long ino, ++ find_inode_t find_actor, void *opaque) ++{ ++ struct list_head * head = inode_hashtable + hash(sb,ino); ++ return ifind(sb, ino, head, find_actor, opaque); ++} ++ ++struct inode *iget4(struct super_block *sb, unsigned long ino, ++ find_inode_t find_actor, void *opaque) ++{ ++ struct list_head * head = inode_hashtable + hash(sb,ino); ++ struct inode *inode = ifind(sb, ino, head, find_actor, opaque); ++ if (inode) ++ return inode; ++ + /* + * get_new_inode() will do the right thing, re-trying the search + * in case it had to block at any point. +Index: linux-2.4.19-pre1/fs/Makefile +=================================================================== +--- linux-2.4.19-pre1.orig/fs/Makefile 2003-11-21 01:51:50.000000000 +0300 ++++ linux-2.4.19-pre1/fs/Makefile 2003-11-21 02:02:35.000000000 +0300 +@@ -7,7 +7,8 @@ + + O_TARGET := fs.o + +-export-objs := filesystems.o open.o dcache.o buffer.o inode.o ++export-objs := filesystems.o open.o dcache.o buffer.o inode.o namei.o \ ++ file_table.o + mod-subdirs := nls + + obj-y := open.o read_write.o devices.o file_table.o buffer.o \ +Index: linux-2.4.19-pre1/fs/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/namei.c 2003-11-20 19:16:40.000000000 +0300 ++++ linux-2.4.19-pre1/fs/namei.c 2003-11-21 02:02:35.000000000 +0300 +@@ -22,6 +22,7 @@ + #include + #include + #include ++#include + + #include + #include +@@ -100,6 +101,7 @@ + it->it_op_release(it); + + } ++EXPORT_SYMBOL(intent_release); + + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the +@@ -878,7 +880,8 @@ + + + /* SMP-safe */ +-struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) ++struct dentry * lookup_one_len_it(const char * name, struct dentry * base, ++ int len, struct lookup_intent *it) + { + unsigned long hash; + struct qstr this; +@@ -898,11 +901,16 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash_it(&this, base, NULL); ++ return lookup_hash_it(&this, base, it); + access: + return ERR_PTR(-EACCES); + } + ++struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) ++{ ++ return lookup_one_len_it(name, base, len, NULL); ++} ++ + /* + * namei() + * +Index: linux-2.4.19-pre1/fs/nfsd/export.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/nfsd/export.c 2001-10-04 09:57:36.000000000 +0400 ++++ linux-2.4.19-pre1/fs/nfsd/export.c 2003-11-21 02:13:10.000000000 +0300 +@@ -196,6 +196,11 @@ + goto out_unlock; + + inode = nd.dentry->d_inode; ++ if ((inode->i_sb->s_type->fs_flags & FS_NFSEXP_FSID) && ++ !(nxp->ex_flags & NFSEXP_FSID)) { ++ nxp->ex_dev = inode->i_sb->s_dev; ++ nxp->ex_flags |= NFSEXP_FSID; ++ } + err = -EINVAL; + if (inode->i_dev != dev || inode->i_ino != nxp->ex_ino) { + printk(KERN_DEBUG "exp_export: i_dev = %x, dev = %x\n", +Index: linux-2.4.19-pre1/fs/nfsd/nfsfh.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/nfsd/nfsfh.c 2001-10-04 09:59:22.000000000 +0400 ++++ linux-2.4.19-pre1/fs/nfsd/nfsfh.c 2003-11-21 02:14:53.000000000 +0300 +@@ -36,6 +36,15 @@ + int sequence; /* sequence counter */ + }; + ++static struct dentry *lookup_it(struct inode *inode, struct dentry * dentry) ++{ ++ if (inode->i_op->lookup_it) ++ return inode->i_op->lookup_it(inode, dentry, NULL, 0); ++ else ++ return inode->i_op->lookup(inode, dentry); ++ ++} ++ + /* + * A rather strange filldir function to capture + * the name matching the specified inode number. +@@ -75,6 +84,8 @@ + int error; + struct file file; + struct nfsd_getdents_callback buffer; ++ struct lookup_intent it; ++ struct file *filp = NULL; + + error = -ENOTDIR; + if (!dir || !S_ISDIR(dir->i_mode)) +@@ -85,9 +96,37 @@ + /* + * Open the directory ... + */ +- error = init_private_file(&file, dentry, FMODE_READ); +- if (error) ++ if (dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if ((dentry->d_flags & DCACHE_NFSD_DISCONNECTED) && ++ (dentry->d_parent == dentry) ) { ++ it.it_op_release = NULL; ++ /* ++ * XXX Temporary Hack: Simulating init_private_file without ++ * f_op->open for disconnected dentry Since we don't have actual ++ * dentry->d_name to revalidate in revalidate_it() ++ */ ++ filp = &file; ++ memset(filp, 0, sizeof(*filp)); ++ filp->f_mode = FMODE_READ; ++ atomic_set(&filp->f_count, 1); ++ filp->f_dentry = dentry; ++ filp->f_uid = current->fsuid; ++ filp->f_gid = current->fsgid; ++ filp->f_op = dentry->d_inode->i_fop; ++ error = 0; ++ } else { ++ intent_init(&it, IT_OPEN, 0); ++ error = revalidate_it(dentry, &it); ++ if (error) ++ goto out; ++ error = init_private_file_it(&file, dentry, FMODE_READ, &it); ++ } ++ } else { ++ error = init_private_file_it(&file, dentry, FMODE_READ, NULL); ++ } ++ if (error) + goto out; ++ + error = -EINVAL; + if (!file.f_op->readdir) + goto out_close; +@@ -113,9 +152,13 @@ + } + + out_close: +- if (file.f_op->release) ++ if (file.f_op->release && !filp) + file.f_op->release(dir, &file); + out: ++ if (dentry->d_op && ++ dentry->d_op->d_revalidate_it && ++ it.it_op_release && !filp) ++ intent_release(&it); + return error; + } + +@@ -273,7 +316,7 @@ + /* I'm going to assume that if the returned dentry is different, then + * it is well connected. But nobody returns different dentrys do they? + */ +- pdentry = child->d_inode->i_op->lookup(child->d_inode, tdentry); ++ pdentry = lookup_it(child->d_inode, tdentry); + d_drop(tdentry); /* we never want ".." hashed */ + if (!pdentry && tdentry->d_inode == NULL) { + /* File system cannot find ".." ... sad but possible */ +@@ -304,6 +347,8 @@ + igrab(tdentry->d_inode); + pdentry->d_flags |= DCACHE_NFSD_DISCONNECTED; + } ++ if (child->d_op && child->d_op->d_revalidate_it) ++ pdentry->d_op = child->d_op; + } + if (pdentry == NULL) + pdentry = ERR_PTR(-ENOMEM); +@@ -461,6 +506,8 @@ + struct dentry *pdentry; + struct inode *parent; + ++ if (result->d_op && result->d_op->d_revalidate_it) ++ dentry->d_op = result->d_op; + pdentry = nfsd_findparent(dentry); + err = PTR_ERR(pdentry); + if (IS_ERR(pdentry)) +@@ -650,6 +697,11 @@ + + inode = dentry->d_inode; + ++ /* cache coherency for non-device filesystems */ ++ if (inode->i_op && inode->i_op->revalidate_it) { ++ inode->i_op->revalidate_it(dentry, NULL); ++ } ++ + /* Type check. The correct error return for type mismatches + * does not seem to be generally agreed upon. SunOS seems to + * use EISDIR if file isn't S_IFREG; a comment in the NFSv3 +@@ -880,8 +932,9 @@ + dentry->d_parent->d_name.name, dentry->d_name.name); + goto out; + out_uptodate: +- printk(KERN_ERR "fh_update: %s/%s already up-to-date!\n", +- dentry->d_parent->d_name.name, dentry->d_name.name); ++ if(!dentry->d_parent->d_inode->i_op->mkdir_raw) ++ printk(KERN_ERR "fh_update: %s/%s already up-to-date!\n", ++ dentry->d_parent->d_name.name, dentry->d_name.name); + goto out; + } + +Index: linux-2.4.19-pre1/fs/nfsd/vfs.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/nfsd/vfs.c 2001-12-21 20:41:55.000000000 +0300 ++++ linux-2.4.19-pre1/fs/nfsd/vfs.c 2003-11-21 02:15:47.000000000 +0300 +@@ -77,6 +77,128 @@ + static struct raparms * raparml; + static struct raparms * raparm_cache; + ++static int link_raw(struct dentry *dold, struct dentry *ddir, ++ struct dentry *dnew) ++{ ++ int err; ++ ++ struct nameidata old_nd = { .dentry = dold }; ++ struct nameidata nd = { .dentry = ddir, .last = dnew->d_name }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->link_raw(&old_nd, &nd); ++ d_instantiate(dnew, dold->d_inode); ++ if(dold->d_inode->i_op && dold->d_inode->i_op->revalidate_it) ++ dold->d_inode->i_op->revalidate_it(dnew, NULL); ++ ++ return err; ++} ++ ++static int unlink_raw(struct dentry *dentry, char *fname, int flen, ++ struct dentry *rdentry) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->unlink_raw(&nd); ++ if (!err) ++ d_delete(rdentry); ++ ++ return err; ++} ++ ++static int rmdir_raw(struct dentry *dentry, char *fname, int flen, ++ struct dentry *rdentry) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->rmdir_raw(&nd); ++ if(!err) { ++ rdentry->d_inode->i_flags |= S_DEAD; ++ d_delete(rdentry); ++ } ++ ++ return err; ++} ++ ++static int symlink_raw(struct dentry *dentry, char *fname, int flen, ++ char *path) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->symlink_raw(&nd, path); ++ ++ return err; ++} ++ ++static int mkdir_raw(struct dentry *dentry, char *fname, int flen, int mode) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->mkdir_raw(&nd, mode); ++ ++ return err; ++} ++ ++static int mknod_raw(struct dentry *dentry, char *fname, int flen, int mode, ++ dev_t dev) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->mknod_raw(&nd, mode, dev); ++ ++ return err; ++} ++ ++static int rename_raw(struct dentry *fdentry, struct dentry *tdentry, ++ struct dentry *odentry, struct dentry *ndentry) ++{ ++ int err; ++ ++ struct nameidata old_nd = { .dentry = fdentry, .last = odentry->d_name}; ++ struct nameidata new_nd = { .dentry = tdentry, .last = ndentry->d_name}; ++ struct inode_operations *op = old_nd.dentry->d_inode->i_op; ++ err = op->rename_raw(&old_nd, &new_nd); ++ d_move(odentry, ndentry); ++ ++ return err; ++} ++ ++static int setattr_raw(struct inode *inode, struct iattr *iap) ++{ ++ int err; ++ ++ iap->ia_valid |= ATTR_RAW; ++ err = inode->i_op->setattr_raw(inode, iap); ++ ++ return err; ++} ++ ++int revalidate_it(struct dentry *dentry, struct lookup_intent *it) ++{ ++ int err = 0; ++ ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ err = -EINVAL; ++ dentry = NULL; ++ return err; ++ } ++ } ++ ++ return err; ++} ++ + /* + * Look up one component of a pathname. + * N.B. After this call _both_ fhp and resfh need an fh_put +@@ -303,7 +425,10 @@ + } + err = nfserr_notsync; + if (!check_guard || guardtime == inode->i_ctime) { +- err = notify_change(dentry, iap); ++ if ( dentry->d_inode->i_op && dentry->d_inode->i_op->setattr_raw) ++ err = setattr_raw(dentry->d_inode, iap); ++ else ++ err = notify_change(dentry, iap); + err = nfserrno(err); + } + if (size_change) { +@@ -430,6 +555,7 @@ + { + struct dentry *dentry; + struct inode *inode; ++ struct lookup_intent it; + int err; + + /* If we get here, then the client has already done an "open", and (hopefully) +@@ -476,6 +602,14 @@ + filp->f_mode = FMODE_READ; + } + ++ intent_init(&it, IT_OPEN, (filp->f_flags & ~O_ACCMODE) | filp->f_mode); ++ ++ err = revalidate_it(dentry, &it); ++ if (err) ++ goto out_nfserr; ++ ++ filp->f_it = ⁢ ++ + err = 0; + if (filp->f_op && filp->f_op->open) { + err = filp->f_op->open(inode, filp); +@@ -490,7 +624,11 @@ + atomic_dec(&filp->f_count); + } + } ++ + out_nfserr: ++ if (it.it_op_release) ++ intent_release(&it); ++ + if (err) + err = nfserrno(err); + out: +@@ -812,7 +950,7 @@ + { + struct dentry *dentry, *dchild; + struct inode *dirp; +- int err; ++ int err, error = -EOPNOTSUPP; + + err = nfserr_perm; + if (!flen) +@@ -828,20 +966,44 @@ + dentry = fhp->fh_dentry; + dirp = dentry->d_inode; + ++ switch (type) { ++ case S_IFDIR: ++ if (dirp->i_op->mkdir_raw) ++ error = mkdir_raw(dentry, fname, flen, iap->ia_mode); ++ break; ++ case S_IFCHR: ++ case S_IFBLK: ++ case S_IFIFO: ++ case S_IFSOCK: ++ case S_IFREG: ++ if (dirp->i_op->mknod_raw) { ++ if (type == S_IFREG) ++ rdev = 0; ++ error = mknod_raw(dentry, fname, flen, iap->ia_mode, rdev); ++ } ++ break; ++ default: ++ printk("nfsd: bad file type %o in nfsd_create\n", type); ++ } ++ + err = nfserr_notdir; +- if(!dirp->i_op || !dirp->i_op->lookup) ++ if(!dirp->i_op || !(dirp->i_op->lookup || dirp->i_op->lookup_it)) + goto out; + /* + * Check whether the response file handle has been verified yet. + * If it has, the parent directory should already be locked. + */ +- if (!resfhp->fh_dentry) { +- /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ +- fh_lock(fhp); ++ if (!resfhp->fh_dentry || dirp->i_op->lookup_it) { ++ /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create ++ and nfsd_proc_create in case of lustre ++ */ ++ if (!resfhp->fh_dentry) ++ fh_lock(fhp); + dchild = lookup_one_len(fname, dentry, flen); + err = PTR_ERR(dchild); + if (IS_ERR(dchild)) + goto out_nfserr; ++ resfhp->fh_dentry = NULL; + err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); + if (err) + goto out; +@@ -862,10 +1024,12 @@ + * Make sure the child dentry is still negative ... + */ + err = nfserr_exist; +- if (dchild->d_inode) { +- dprintk("nfsd_create: dentry %s/%s not negative!\n", +- dentry->d_name.name, dchild->d_name.name); +- goto out; ++ if ( error == -EOPNOTSUPP) { ++ if (dchild->d_inode) { ++ dprintk("nfsd_create: dentry %s/%s not negative!\n", ++ dentry->d_name.name, dchild->d_name.name); ++ goto out; ++ } + } + + if (!(iap->ia_valid & ATTR_MODE)) +@@ -878,16 +1042,19 @@ + err = nfserr_perm; + switch (type) { + case S_IFREG: +- err = vfs_create(dirp, dchild, iap->ia_mode); ++ if (error == -EOPNOTSUPP) ++ err = vfs_create(dirp, dchild, iap->ia_mode); + break; + case S_IFDIR: +- err = vfs_mkdir(dirp, dchild, iap->ia_mode); ++ if (error == -EOPNOTSUPP) ++ err = vfs_mkdir(dirp, dchild, iap->ia_mode); + break; + case S_IFCHR: + case S_IFBLK: + case S_IFIFO: + case S_IFSOCK: +- err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); ++ if (error == -EOPNOTSUPP) ++ err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); + break; + default: + printk("nfsd: bad file type %o in nfsd_create\n", type); +@@ -956,7 +1123,13 @@ + /* Get all the sanity checks out of the way before + * we lock the parent. */ + err = nfserr_notdir; +- if(!dirp->i_op || !dirp->i_op->lookup) ++ if (dirp->i_op->mknod_raw) { ++ err = mknod_raw(dentry, fname, flen, iap->ia_mode, 0); ++ if (err && err != -EOPNOTSUPP) ++ goto out; ++ } ++ ++ if(!dirp->i_op || !(dirp->i_op->lookup || dirp->i_op->lookup_it)) + goto out; + fh_lock(fhp); + +@@ -1007,6 +1180,8 @@ + case NFS3_CREATE_GUARDED: + err = nfserr_exist; + } ++ if(dirp->i_op->mknod_raw) ++ err = 0; + goto out; + } + +@@ -1113,7 +1288,7 @@ + struct iattr *iap) + { + struct dentry *dentry, *dnew; +- int err, cerr; ++ int err, cerr, error = -EOPNOTSUPP; + + err = nfserr_noent; + if (!flen || !plen) +@@ -1127,12 +1302,18 @@ + goto out; + fh_lock(fhp); + dentry = fhp->fh_dentry; ++ ++ if (dentry->d_inode->i_op->symlink_raw) ++ error = symlink_raw(dentry, fname, flen, path); ++ + dnew = lookup_one_len(fname, dentry, flen); + err = PTR_ERR(dnew); + if (IS_ERR(dnew)) + goto out_nfserr; + +- err = vfs_symlink(dentry->d_inode, dnew, path); ++ err = error; ++ if (err == -EOPNOTSUPP || !dentry->d_inode->i_op->symlink_raw) ++ err = vfs_symlink(dentry->d_inode, dnew, path); + if (!err) { + if (EX_ISSYNC(fhp->fh_export)) + nfsd_sync_dir(dentry); +@@ -1142,7 +1323,10 @@ + iap->ia_valid |= ATTR_CTIME; + iap->ia_mode = (iap->ia_mode&S_IALLUGO) + | S_IFLNK; +- err = notify_change(dnew, iap); ++ if (dnew->d_inode->i_op && dnew->d_inode->i_op->setattr_raw) ++ err = setattr_raw(dnew->d_inode, iap); ++ else ++ err = notify_change(dnew, iap); + if (!err && EX_ISSYNC(fhp->fh_export)) + write_inode_now(dentry->d_inode, 1); + } +@@ -1200,7 +1384,10 @@ + dold = tfhp->fh_dentry; + dest = dold->d_inode; + +- err = vfs_link(dold, dirp, dnew); ++ if (dirp->i_op->link_raw) ++ err = link_raw(dold, ddir, dnew); ++ else ++ err = vfs_link(dold, dirp, dnew); + if (!err) { + if (EX_ISSYNC(ffhp->fh_export)) { + nfsd_sync_dir(ddir); +@@ -1285,7 +1472,10 @@ + err = nfserr_perm; + } else + #endif +- err = vfs_rename(fdir, odentry, tdir, ndentry); ++ if(fdir->i_op->rename_raw) ++ err = rename_raw(fdentry, tdentry, odentry, ndentry); ++ else ++ err = vfs_rename(fdir, odentry, tdir, ndentry); + if (!err && EX_ISSYNC(tfhp->fh_export)) { + nfsd_sync_dir(tdentry); + nfsd_sync_dir(fdentry); +@@ -1306,7 +1496,7 @@ + fill_post_wcc(tfhp); + double_up(&tdir->i_sem, &fdir->i_sem); + ffhp->fh_locked = tfhp->fh_locked = 0; +- ++ + out: + return err; + } +@@ -1352,9 +1542,15 @@ + err = nfserr_perm; + } else + #endif +- err = vfs_unlink(dirp, rdentry); ++ if (dirp->i_op->unlink_raw) ++ err = unlink_raw(dentry, fname, flen, rdentry); ++ else ++ err = vfs_unlink(dirp, rdentry); + } else { /* It's RMDIR */ +- err = vfs_rmdir(dirp, rdentry); ++ if (dirp->i_op->rmdir_raw) ++ err = rmdir_raw(dentry, fname, flen, rdentry); ++ else ++ err = vfs_rmdir(dirp, rdentry); + } + + dput(rdentry); +Index: linux-2.4.19-pre1/include/linux/fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs.h 2003-11-21 02:00:20.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs.h 2003-11-21 02:02:36.000000000 +0300 +@@ -93,6 +93,9 @@ + #define FS_SINGLE 8 /* Filesystem that can have only one superblock */ + #define FS_NOMOUNT 16 /* Never mount from userland */ + #define FS_LITTER 32 /* Keeps the tree in dcache */ ++#define FS_NFSEXP_FSID 64 /* Use file system specific fsid for ++ * exporting non device filesystems. ++ */ + #define FS_ODD_RENAME 32768 /* Temporary stuff; will go away as soon + * as nfs_rename() will be cleaned up + */ +@@ -1081,6 +1084,9 @@ + struct nameidata *nd, struct lookup_intent *it); + extern struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, + int flags, struct lookup_intent *it); ++extern int revalidate_it(struct dentry *dentry, struct lookup_intent *it); ++extern int init_private_file_it(struct file *, struct dentry *dentry, int mode, ++ struct lookup_intent *it); + extern int filp_close(struct file *, fl_owner_t id); + extern char * getname(const char *); + +@@ -1342,6 +1348,8 @@ + extern int follow_down(struct vfsmount **, struct dentry **); + extern int follow_up(struct vfsmount **, struct dentry **); + extern struct dentry * lookup_one_len(const char *, struct dentry *, int); ++extern struct dentry * lookup_one_len_it(const char *, struct dentry *, int, ++ struct lookup_intent *); + extern struct dentry * lookup_hash(struct qstr *, struct dentry *); + #define user_path_walk(name,nd) __user_walk(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd) + #define user_path_walk_link(name,nd) __user_walk(name, LOOKUP_POSITIVE, nd) +@@ -1355,6 +1363,8 @@ + + typedef int (*find_inode_t)(struct inode *, unsigned long, void *); + extern struct inode * iget4(struct super_block *, unsigned long, find_inode_t, void *); ++extern struct inode * ilookup4(struct super_block *, unsigned long, ++ find_inode_t, void *); + static inline struct inode *iget(struct super_block *sb, unsigned long ino) + { + return iget4(sb, ino, NULL, NULL); +Index: linux-2.4.19-pre1/kernel/ksyms.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/ksyms.c 2003-11-21 01:51:50.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/ksyms.c 2003-11-21 02:02:36.000000000 +0300 +@@ -140,6 +140,7 @@ + EXPORT_SYMBOL(igrab); + EXPORT_SYMBOL(iunique); + EXPORT_SYMBOL(iget4); ++EXPORT_SYMBOL(ilookup4); + EXPORT_SYMBOL(iput); + EXPORT_SYMBOL(force_delete); + EXPORT_SYMBOL(follow_up); +@@ -150,6 +151,7 @@ + EXPORT_SYMBOL(path_release); + EXPORT_SYMBOL(__user_walk); + EXPORT_SYMBOL(lookup_one_len); ++EXPORT_SYMBOL(lookup_one_len_it); + EXPORT_SYMBOL(lookup_hash); + EXPORT_SYMBOL(sys_close); + EXPORT_SYMBOL(dcache_lock); diff --git a/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch b/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch new file mode 100644 index 0000000..4602f96 --- /dev/null +++ b/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch @@ -0,0 +1,30 @@ + include/linux/fs.h | 1 + + mm/filemap.c | 3 +++ + 2 files changed, 4 insertions(+) + +Index: linux-2.4.19.SuSE/include/linux/fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs.h Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/include/linux/fs.h Sun Nov 16 01:38:06 2003 +@@ -428,6 +428,7 @@ + int (*releasepage) (struct page *, int); + #define KERNEL_HAS_O_DIRECT /* this is for modules out of the kernel */ + int (*direct_IO)(int, struct file *, struct kiobuf *, unsigned long, int); ++ void (*removepage)(struct page *); /* called when page gets removed from the inode */ + }; + + struct address_space { +Index: linux-2.4.19.SuSE/mm/filemap.c +=================================================================== +--- linux-2.4.19.SuSE.orig/mm/filemap.c Sat Nov 15 18:02:15 2003 ++++ linux-2.4.19.SuSE/mm/filemap.c Sun Nov 16 01:37:11 2003 +@@ -97,6 +97,9 @@ + { + struct address_space * mapping = page->mapping; + ++ if (mapping->a_ops->removepage) ++ mapping->a_ops->removepage(page); ++ + mapping->nrpages--; + list_del(&page->list); + page->mapping = NULL; diff --git a/lustre/kernel_patches/patches/resched-2.4.19-pre1.patch b/lustre/kernel_patches/patches/resched-2.4.19-pre1.patch new file mode 100644 index 0000000..567e1e8 --- /dev/null +++ b/lustre/kernel_patches/patches/resched-2.4.19-pre1.patch @@ -0,0 +1,16 @@ +Index: linux-2.4.19-pre1/include/linux/sched.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/sched.h 2003-11-21 04:05:05.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/sched.h 2003-11-21 04:10:29.000000000 +0300 +@@ -927,6 +927,11 @@ + return res; + } + ++static inline int need_resched(void) ++{ ++ return (unlikely(current->need_resched)); ++} ++ + #endif /* __KERNEL__ */ + + #endif diff --git a/lustre/kernel_patches/patches/seq-private-2.4.19-pre1.patch b/lustre/kernel_patches/patches/seq-private-2.4.19-pre1.patch new file mode 100644 index 0000000..241bf36 --- /dev/null +++ b/lustre/kernel_patches/patches/seq-private-2.4.19-pre1.patch @@ -0,0 +1,12 @@ +Index: linux-2.4.19-pre1/include/linux/seq_file.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/seq_file.h 2001-11-11 22:23:14.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/seq_file.h 2003-11-21 03:09:39.000000000 +0300 +@@ -12,6 +12,7 @@ + loff_t index; + struct semaphore sem; + struct seq_operations *op; ++ void *private; + }; + + struct seq_operations { diff --git a/lustre/kernel_patches/patches/slab-use-after-free-debug-2.4.22.patch b/lustre/kernel_patches/patches/slab-use-after-free-debug-2.4.22.patch new file mode 100644 index 0000000..ab5afe8 --- /dev/null +++ b/lustre/kernel_patches/patches/slab-use-after-free-debug-2.4.22.patch @@ -0,0 +1,717 @@ +%patch +Index: linux-2.4.22-vanilla/mm/slab.c +=================================================================== +--- linux-2.4.22-vanilla.orig/mm/slab.c 2003-11-17 15:42:13.000000000 +0300 ++++ linux-2.4.22-vanilla/mm/slab.c 2003-11-18 01:15:35.000000000 +0300 +@@ -97,6 +97,8 @@ + #define FORCED_DEBUG 0 + #endif + ++#include ++ + /* + * Parameters for kmem_cache_reap + */ +@@ -825,6 +827,12 @@ + return cachep; + } + ++#ifdef CONFIG_DEBUG_UAF ++void * uaf_alloc(kmem_cache_t *, int gfp_mask); ++int uaf_cache_free(kmem_cache_t *, void *addr); ++int uaf_free(void *addr); ++struct page *uaf_vaddr_to_page(void *obj); ++#endif + + #if DEBUG + /* +@@ -1340,6 +1348,20 @@ + unsigned long save_flags; + void* objp; + ++#ifdef CONFIG_DEBUG_UAF ++ /* try to use uaf-allocator first */ ++ objp = uaf_alloc(cachep, flags); ++ if (objp) { ++ if (cachep->ctor) { ++ unsigned long ctor_flags; ++ ctor_flags = SLAB_CTOR_CONSTRUCTOR; ++ if (!(flags & __GFP_WAIT)) ++ ctor_flags |= SLAB_CTOR_ATOMIC; ++ cachep->ctor(objp, cachep, ctor_flags); ++ } ++ return objp; ++ } ++#endif + kmem_cache_alloc_head(cachep, flags); + try_again: + local_irq_save(save_flags); +@@ -1576,6 +1598,10 @@ + void kmem_cache_free (kmem_cache_t *cachep, void *objp) + { + unsigned long flags; ++#ifdef CONFIG_DEBUG_UAF ++ if (uaf_cache_free(cachep, objp)) ++ return; ++#endif + #if DEBUG + CHECK_PAGE(virt_to_page(objp)); + if (cachep != GET_PAGE_CACHE(virt_to_page(objp))) +@@ -1601,6 +1627,10 @@ + + if (!objp) + return; ++#ifdef CONFIG_DEBUG_UAF ++ if (uaf_free((void *) objp)) ++ return; ++#endif + local_irq_save(flags); + CHECK_PAGE(virt_to_page(objp)); + c = GET_PAGE_CACHE(virt_to_page(objp)); +@@ -2075,3 +2105,460 @@ + #endif + } + #endif ++ ++ ++ ++#ifdef CONFIG_DEBUG_UAF ++ ++#define MAX_UAF_OBJ_SIZE 8 /* in pages */ ++#define UAF_ASSERT(xxx) if (!(xxx)) BUG(); ++#define UAF_DEBUG__ ++#ifdef UAF_DEBUG ++#define uaf_printk(fmt,a...) printk(fmt, ##a) ++#else ++#define uaf_printk(a,...) ++#endif ++ ++struct uaf_stats { ++ atomic_t uaf_allocated; ++ atomic_t uaf_allocations; ++ atomic_t uaf_failed; ++}; ++ ++static int uaf_max = 32768; ++static void *uaf_bitmap = NULL; ++static spinlock_t uaf_lock; ++static int uaf_last_found = 0; ++static int uaf_used = 0; ++static struct vm_struct *uaf_area = NULL; ++static struct uaf_stats uaf_stats[MAX_UAF_OBJ_SIZE + 1]; ++ ++static int __init uaf_setup(char *str) ++{ ++ uaf_max = simple_strtoul(str, NULL, 0); ++ return 1; ++} ++ ++__setup("uaf=", uaf_setup); ++ ++ ++void uaf_init(void) ++{ ++ int size; ++ ++ printk("UAF: total vmalloc-space - %lu\n", ++ VMALLOC_END - VMALLOC_START); ++ ++ uaf_area = get_vm_area(PAGE_SIZE * uaf_max, VM_ALLOC); ++ if (!uaf_area) { ++ printk(KERN_ALERT "UAF: can't reserve %lu bytes in KVA\n", ++ PAGE_SIZE * uaf_max); ++ return; ++ } ++ ++ printk("UAF: reserved %lu bytes in KVA at 0x%p\n", ++ PAGE_SIZE * uaf_max, uaf_area->addr); ++ ++ /* how many bytes we need to track space usage? */ ++ size = uaf_max / 8 + 8; ++ ++ uaf_bitmap = vmalloc(size); ++ if (!uaf_bitmap) { ++ printk(KERN_ALERT ++ "UAF: can't allocate %d bytes for bitmap\n", size); ++ return; ++ } ++ memset(uaf_bitmap, 0, size); ++ spin_lock_init(&uaf_lock); ++ memset(uaf_stats, 0, sizeof(uaf_stats)); ++ ++ printk("UAF: allocated %d for bitmap\n", size); ++} ++ ++static int uaf_find(int len) ++{ ++ int new_last_found = -1; ++ int loop = 0; ++ int i, j; ++ ++ j = uaf_last_found; ++ ++ do { ++ i = find_next_zero_bit(uaf_bitmap, uaf_max, j); ++ if (i >= uaf_max) { ++ /* repeat from 0 */ ++ if (++loop > 1) { ++ /* this is 2nd loop and it's useless */ ++ return -1; ++ } ++ ++ i = find_next_zero_bit(uaf_bitmap, uaf_max, 0); ++ if (i >= uaf_max) ++ return -1; ++ ++ /* save found num for subsequent searches */ ++ if (new_last_found == -1) ++ new_last_found = uaf_last_found = i; ++ UAF_ASSERT(new_last_found < uaf_max); ++ } ++ ++ /* ++ * OK. found first zero bit. ++ * now, try to find requested cont. zero-space ++ */ ++ ++ /* FIXME: implmement multipage allocation! */ ++ break; ++ ++ /* ++ j = find_next_bit(uaf_bitmap, uaf_max, i); ++ if (++loop2 > 10000) { ++ printk("ALERT: loop2=%d\n", loop2); ++ return -1; ++ } ++ */ ++ } while (j - i < len); ++ ++ /* found! */ ++ if (new_last_found == -1) ++ uaf_last_found = i + 1; ++ if (uaf_last_found >= uaf_max) ++ uaf_last_found = 0; ++ return i; ++} ++ ++extern int __vmalloc_area_pages (unsigned long address, unsigned long size, ++ int gfp_mask, pgprot_t prot, ++ struct page ***pages); ++void * uaf_alloc(kmem_cache_t *cachep, int gfp_mask) ++{ ++ struct page *ptrs[MAX_UAF_OBJ_SIZE]; ++ int size = cachep->objsize; ++ struct page **pages; ++ unsigned long flags; ++ unsigned long addr; ++ int i, j, err = -2000; ++ ++ if (uaf_bitmap == NULL) ++ return NULL; ++ ++ if (!(cachep->flags & SLAB_USE_UAF)) ++ return NULL; ++ ++ pages = (struct page **) ptrs; ++ size = (size + (PAGE_SIZE - 1)) / PAGE_SIZE; ++ /* FIXME: implement multipage allocation! */ ++ if (size > 1) ++ return NULL; ++ if (size > MAX_UAF_OBJ_SIZE) { ++ printk(KERN_ALERT "size is too big: %d\n", size); ++ return NULL; ++ } ++ ++ if (uaf_used == uaf_max) { ++ uaf_printk("UAF: space exhausted!\n"); ++ atomic_inc(&uaf_stats[size].uaf_failed); ++ return NULL; ++ } ++ ++ ++ spin_lock_irqsave(&uaf_lock, flags); ++ i = uaf_find(size); ++ if (i < 0) { ++ spin_unlock_irqrestore(&uaf_lock, flags); ++ atomic_inc(&uaf_stats[size].uaf_failed); ++ return NULL; ++ } ++ for (j = 0; j < size; j++) { ++ UAF_ASSERT(!test_bit(i + j, uaf_bitmap)); ++ set_bit(i + j, uaf_bitmap); ++ uaf_used++; ++ } ++ spin_unlock_irqrestore(&uaf_lock, flags); ++ ++ addr = ((unsigned long) uaf_area->addr) + (PAGE_SIZE * i); ++ uaf_printk("UAF: found %d/%d, base 0x%p, map at 0x%lx: ", i, ++ size, uaf_area->addr, addr); ++ ++ /* OK. we've found free space, let's allocate pages */ ++ memset(pages, 0, sizeof(struct page *) * MAX_UAF_OBJ_SIZE); ++ for (j = 0; j < size; j++) { ++ pages[j] = alloc_page(gfp_mask); ++ if (pages[j] == NULL) ++ goto nomem; ++ uaf_printk("0x%p ", pages[j]); ++ } ++ ++ /* time to map just allocated pages */ ++ err = __vmalloc_area_pages(addr, PAGE_SIZE * size, gfp_mask, ++ PAGE_KERNEL, &pages); ++ pages = (struct page **) ptrs; ++ if (err == 0) { ++ /* put slab cache pointer in first page */ ++ ptrs[0]->list.next = (void *) cachep; ++ uaf_printk(" -> 0x%lx\n", addr); ++ atomic_inc(&uaf_stats[size].uaf_allocated); ++ atomic_inc(&uaf_stats[size].uaf_allocations); ++ if (!in_interrupt() && !in_softirq()) ++ flush_tlb_all(); ++ else ++ local_flush_tlb(); ++ //printk("UAF: found %d/%d, base 0x%p, map at 0x%lx\n", ++ // i, cachep->objsize, uaf_area->addr, addr); ++ return (void *) addr; ++ } ++ ++nomem: ++ printk(KERN_ALERT "can't map pages: %d\n", err); ++ for (j = 0; j < size; j++) ++ if (pages[j]) ++ __free_page(pages[j]); ++ ++ /* can't find free pages */ ++ spin_lock_irqsave(&uaf_lock, flags); ++ for (j = 0; j < size; j++) { ++ clear_bit(i + j, uaf_bitmap); ++ uaf_used--; ++ } ++ spin_unlock_irqrestore(&uaf_lock, flags); ++ atomic_inc(&uaf_stats[size].uaf_failed); ++ ++ return NULL; ++} ++ ++extern void free_area_pmd(pgd_t *dir, unsigned long address, ++ unsigned long size); ++static void uaf_unmap(unsigned long address, unsigned long size) ++{ ++ unsigned long end = (address + size); ++ pgd_t *dir; ++ ++ dir = pgd_offset_k(address); ++ flush_cache_all(); ++ do { ++ free_area_pmd(dir, address, end - address); ++ address = (address + PGDIR_SIZE) & PGDIR_MASK; ++ dir++; ++ } while (address && (address < end)); ++ ++ /* ++ * we must not call smp_call_function() with interrtups disabled ++ * otherwise we can get into deadlock ++ */ ++ if (!in_interrupt() && !in_softirq()) ++ flush_tlb_all(); ++ else ++ local_flush_tlb(); ++} ++ ++/* ++ * returns 1 if free was successfull ++ */ ++int uaf_cache_free(kmem_cache_t *cachep, void *addr) ++{ ++ struct page *pages[MAX_UAF_OBJ_SIZE]; ++ int size = cachep->objsize; ++ unsigned long flags; ++ int i, j; ++ ++ uaf_printk("UAF: to free 0x%p/%d\n", addr, size); ++ ++ size = (size + (PAGE_SIZE - 1)) / PAGE_SIZE; ++ if (size > MAX_UAF_OBJ_SIZE) ++ return 0; ++ ++ if (uaf_bitmap == NULL) ++ return 0; ++ ++ /* first, check is address is in UAF space */ ++ if ((unsigned) addr < (unsigned) uaf_area->addr || ++ (unsigned) addr >= (unsigned) uaf_area->addr + uaf_area->size) ++ return 0; ++ ++ UAF_ASSERT(((unsigned long) addr & ~PAGE_MASK) == 0UL); ++ ++ /* calculate placement in bitmap */ ++ i = (unsigned) addr - (unsigned) uaf_area->addr; ++ UAF_ASSERT(i >= 0); ++ i = i / PAGE_SIZE; ++ ++ /* collect all the pages */ ++ uaf_printk("free/unmap %d pages: ", size); ++ /* NOTE: we need not page_table_lock here. bits in bitmap ++ * protect those pte's from to be reused */ ++ for (j = 0; j < size; j++) { ++ unsigned long address; ++ address = ((unsigned long) addr) + (PAGE_SIZE * j); ++ pages[j] = vmalloc_to_page((void *) address); ++ uaf_printk("0x%lx->0x%p ", address, pages[j]); ++ } ++ uaf_printk("\n"); ++ ++ uaf_unmap((unsigned long) addr, PAGE_SIZE * size); ++ /* free all the pages */ ++ for (j = 0; j < size; j++) ++ __free_page(pages[j]); ++ ++ spin_lock_irqsave(&uaf_lock, flags); ++ for (j = 0; j < size; j++) { ++ /* now check is correspondend bit set */ ++ UAF_ASSERT(i+j >= 0 && i+j < uaf_max); ++ UAF_ASSERT(test_bit(i+j, uaf_bitmap)); ++ ++ /* now free space in UAF */ ++ clear_bit(i+j, uaf_bitmap); ++ uaf_used--; ++ } ++ spin_unlock_irqrestore(&uaf_lock, flags); ++ ++ atomic_dec(&uaf_stats[size].uaf_allocated); ++ ++ uaf_printk("UAF: freed %d/%d at 0x%p\n", i, size, addr); ++ //printk("UAF: freed %d/%d at 0x%p\n", i, size, addr); ++ ++ return 1; ++} ++ ++struct page *uaf_vaddr_to_page(void *obj) ++{ ++ if (uaf_bitmap == NULL) ++ return NULL; ++ ++ /* first, check is address is in UAF space */ ++ if ((unsigned) obj < (unsigned) uaf_area->addr || ++ (unsigned) obj >= (unsigned) uaf_area->addr + uaf_area->size) ++ return NULL; ++ ++ return vmalloc_to_page(obj); ++} ++ ++int uaf_free(void *obj) ++{ ++ struct page *page = uaf_vaddr_to_page((void *) obj); ++ kmem_cache_t *c; ++ ++ if (!page) ++ return 0; ++ ++ c = GET_PAGE_CACHE(page); ++ return uaf_cache_free(c, (void *) obj); ++} ++ ++int uaf_is_allocated(void *obj) ++{ ++ unsigned long addr = (unsigned long) obj; ++ int i; ++ ++ if (uaf_bitmap == NULL) ++ return 0; ++ ++ addr &= PAGE_MASK; ++ /* first, check is address is in UAF space */ ++ if (addr < (unsigned long) uaf_area->addr || ++ addr >= (unsigned long) uaf_area->addr + uaf_area->size) ++ return 0; ++ ++ /* calculate placement in bitmap */ ++ i = (unsigned) addr - (unsigned) uaf_area->addr; ++ i = i / PAGE_SIZE; ++ return test_bit(i, uaf_bitmap); ++} ++ ++static void *uaf_s_start(struct seq_file *m, loff_t *pos) ++{ ++ loff_t n = *pos; ++ ++ if (!n) ++ seq_printf(m, "size(pgs) allocated failed allocations. " ++ "%d reserved, %d in use, %d last\n", ++ uaf_max, uaf_used, uaf_last_found); ++ else if (n > MAX_UAF_OBJ_SIZE) ++ return NULL; ++ ++ *pos = 1; ++ return (void *) 1; ++} ++ ++static void *uaf_s_next(struct seq_file *m, void *p, loff_t *pos) ++{ ++ unsigned long n = *pos; ++ ++*pos; ++ if (n + 1 > MAX_UAF_OBJ_SIZE) ++ return NULL; ++ return (void *) (n + 1); ++} ++ ++static void uaf_s_stop(struct seq_file *m, void *p) ++{ ++} ++ ++static int uaf_s_show(struct seq_file *m, void *p) ++{ ++ int n = (int) p; ++ ++ if (n > MAX_UAF_OBJ_SIZE) ++ return 0; ++ seq_printf(m, "%d %d %d %d\n", n, ++ atomic_read(&uaf_stats[n].uaf_allocated), ++ atomic_read(&uaf_stats[n].uaf_failed), ++ atomic_read(&uaf_stats[n].uaf_allocations)); ++ return 0; ++} ++ ++struct seq_operations uafinfo_op = { ++ .start = uaf_s_start, ++ .next = uaf_s_next, ++ .stop = uaf_s_stop, ++ .show = uaf_s_show, ++}; ++ ++ssize_t uafinfo_write(struct file *file, const char *buffer, ++ size_t count, loff_t *ppos) ++{ ++ char kbuf[MAX_SLABINFO_WRITE+1], *tmp; ++ char *key, *name; ++ int res; ++ struct list_head *p; ++ ++ if (count > MAX_SLABINFO_WRITE) ++ return -EINVAL; ++ if (copy_from_user(&kbuf, buffer, count)) ++ return -EFAULT; ++ kbuf[MAX_SLABINFO_WRITE] = '\0'; ++ ++ tmp = kbuf; ++ key = strsep(&tmp, " \t\n"); ++ if (!key) ++ return -EINVAL; ++ if (!strcmp(key, "on")) ++ res = 1; ++ else if (!strcmp(key, "off")) ++ res = 0; ++ else ++ return -EINVAL; ++ ++ name = strsep(&tmp, " \t\n"); ++ if (!name) ++ return -EINVAL; ++ ++ /* Find the cache in the chain of caches. */ ++ down(&cache_chain_sem); ++ list_for_each(p,&cache_chain) { ++ kmem_cache_t *cachep = list_entry(p, kmem_cache_t, next); ++ ++ if (!strcmp(cachep->name, name)) { ++ if (res) { ++ printk("UAF: use on %s\n", cachep->name); ++ cachep->flags |= SLAB_USE_UAF; ++ } else { ++ printk("UAF: dont use on %s\n", cachep->name); ++ cachep->flags &= ~SLAB_USE_UAF; ++ } ++ break; ++ } ++ } ++ up(&cache_chain_sem); ++ return count; ++} ++#endif ++ +Index: linux-2.4.22-vanilla/init/main.c +=================================================================== +--- linux-2.4.22-vanilla.orig/init/main.c 2003-11-03 23:22:13.000000000 +0300 ++++ linux-2.4.22-vanilla/init/main.c 2003-11-18 01:06:45.000000000 +0300 +@@ -436,6 +436,9 @@ + * make syscalls (and thus be locked). + */ + smp_init(); ++#ifdef CONFIG_DEBUG_UAF ++ uaf_init(); ++#endif + rest_init(); + } + +Index: linux-2.4.22-vanilla/fs/proc/proc_misc.c +=================================================================== +--- linux-2.4.22-vanilla.orig/fs/proc/proc_misc.c 2003-11-03 23:22:11.000000000 +0300 ++++ linux-2.4.22-vanilla/fs/proc/proc_misc.c 2003-11-18 01:06:45.000000000 +0300 +@@ -301,6 +301,22 @@ + release: seq_release, + }; + ++#ifdef CONFIG_DEBUG_UAF ++extern struct seq_operations uafinfo_op; ++extern ssize_t uafinfo_write(struct file *, const char *, size_t, loff_t *); ++static int uafinfo_open(struct inode *inode, struct file *file) ++{ ++ return seq_open(file, &uafinfo_op); ++} ++static struct file_operations proc_uafinfo_operations = { ++ .open = uafinfo_open, ++ .read = seq_read, ++ .write = uafinfo_write, ++ .llseek = seq_lseek, ++ .release = seq_release, ++}; ++#endif ++ + static int kstat_read_proc(char *page, char **start, off_t off, + int count, int *eof, void *data) + { +@@ -616,6 +632,9 @@ + create_seq_entry("cpuinfo", 0, &proc_cpuinfo_operations); + create_seq_entry("partitions", 0, &proc_partitions_operations); + create_seq_entry("slabinfo",S_IWUSR|S_IRUGO,&proc_slabinfo_operations); ++#ifdef CONFIG_DEBUG_UAF ++ create_seq_entry("uafinfo",S_IWUSR|S_IRUGO,&proc_uafinfo_operations); ++#endif + #ifdef CONFIG_MODULES + create_seq_entry("ksyms", 0, &proc_ksyms_operations); + #endif +Index: linux-2.4.22-vanilla/include/linux/slab.h +=================================================================== +--- linux-2.4.22-vanilla.orig/include/linux/slab.h 2003-11-17 15:42:13.000000000 +0300 ++++ linux-2.4.22-vanilla/include/linux/slab.h 2003-11-18 02:14:40.000000000 +0300 +@@ -40,6 +40,7 @@ + #define SLAB_HWCACHE_ALIGN 0x00002000UL /* align objs on a h/w cache lines */ + #define SLAB_CACHE_DMA 0x00004000UL /* use GFP_DMA memory */ + #define SLAB_MUST_HWCACHE_ALIGN 0x00008000UL /* force alignment */ ++#define SLAB_USE_UAF 0x00040000UL /* use UAF allocator */ + + /* flags passed to a constructor func */ + #define SLAB_CTOR_CONSTRUCTOR 0x001UL /* if not set, then deconstructor */ +Index: linux-2.4.22-vanilla/arch/i386/config.in +=================================================================== +--- linux-2.4.22-vanilla.orig/arch/i386/config.in 2003-11-03 23:22:06.000000000 +0300 ++++ linux-2.4.22-vanilla/arch/i386/config.in 2003-11-18 01:06:45.000000000 +0300 +@@ -470,6 +470,9 @@ + bool ' Check for stack overflows' CONFIG_DEBUG_STACKOVERFLOW + bool ' Debug high memory support' CONFIG_DEBUG_HIGHMEM + bool ' Debug memory allocations' CONFIG_DEBUG_SLAB ++ if [ "$CONFIG_DEBUG_SLAB" != "n" ]; then ++ bool ' Debug memory allocations (use-after-free via vmalloced space)' CONFIG_DEBUG_UAF ++ fi + bool ' Memory mapped I/O debugging' CONFIG_DEBUG_IOVIRT + bool ' Magic SysRq key' CONFIG_MAGIC_SYSRQ + bool ' Spinlock debugging' CONFIG_DEBUG_SPINLOCK +Index: linux-2.4.22-vanilla/mm/vmalloc.c +=================================================================== +--- linux-2.4.22-vanilla.orig/mm/vmalloc.c 2003-11-03 23:22:13.000000000 +0300 ++++ linux-2.4.22-vanilla/mm/vmalloc.c 2003-11-18 01:06:45.000000000 +0300 +@@ -53,7 +53,7 @@ + } while (address < end); + } + +-static inline void free_area_pmd(pgd_t * dir, unsigned long address, unsigned long size) ++void free_area_pmd(pgd_t * dir, unsigned long address, unsigned long size) + { + pmd_t * pmd; + unsigned long end; +@@ -152,7 +152,7 @@ + return 0; + } + +-static inline int __vmalloc_area_pages (unsigned long address, ++int __vmalloc_area_pages (unsigned long address, + unsigned long size, + int gfp_mask, + pgprot_t prot, +Index: linux-2.4.22-vanilla/include/asm-i386/io.h +=================================================================== +--- linux-2.4.22-vanilla.orig/include/asm-i386/io.h 2003-11-17 14:58:37.000000000 +0300 ++++ linux-2.4.22-vanilla/include/asm-i386/io.h 2003-11-18 02:26:42.000000000 +0300 +@@ -75,6 +75,16 @@ + + static inline unsigned long virt_to_phys(volatile void * address) + { ++#ifdef CONFIG_DEBUG_UAF ++ unsigned long addr = (unsigned long) address; ++ if (vmlist && addr >= VMALLOC_START && addr < VMALLOC_END) { ++ struct page *page = vmalloc_to_page((void *) address); ++ if (page) { ++ unsigned long offset = addr & ~PAGE_MASK; ++ address = page_address(page) + offset; ++ } ++ } ++#endif + return __pa(address); + } + +Index: linux-2.4.22-vanilla/include/asm-i386/page.h +=================================================================== +--- linux-2.4.22-vanilla.orig/include/asm-i386/page.h 2003-11-03 23:51:46.000000000 +0300 ++++ linux-2.4.22-vanilla/include/asm-i386/page.h 2003-11-18 02:14:38.000000000 +0300 +@@ -129,9 +129,49 @@ + #define VMALLOC_RESERVE ((unsigned long)__VMALLOC_RESERVE) + #define __MAXMEM (-__PAGE_OFFSET-__VMALLOC_RESERVE) + #define MAXMEM ((unsigned long)(-PAGE_OFFSET-VMALLOC_RESERVE)) ++ ++#ifndef CONFIG_DEBUG_UAF + #define __pa(x) ((unsigned long)(x)-PAGE_OFFSET) + #define __va(x) ((void *)((unsigned long)(x)+PAGE_OFFSET)) + #define virt_to_page(kaddr) (mem_map + (__pa(kaddr) >> PAGE_SHIFT)) ++#else ++#define __pa(x) ({ \ ++ unsigned long __pn, __fr; \ ++ __pn = (unsigned long)(x)-PAGE_OFFSET; \ ++ __fr = __pn >> PAGE_SHIFT; \ ++ if (jiffies > HZ*3 && __fr >= max_mapnr) { \ ++ printk("invalid arg __pa(0x%x)" \ ++ " at %s:%d\n", (unsigned) (x), \ ++ __FILE__, __LINE__); \ ++ dump_stack(); \ ++ } \ ++ __pn; \ ++ }) ++ ++#define __va(x) ({ \ ++ unsigned long __pn; \ ++ __pn = (unsigned long) (x) >> PAGE_SHIFT; \ ++ if (jiffies > HZ*3 && __pn >= max_mapnr) { \ ++ printk("invalid arg __va(0x%x)" \ ++ " at %s:%d\n", (unsigned) (x), \ ++ __FILE__, __LINE__); \ ++ dump_stack(); \ ++ } \ ++ ((void *)((unsigned long)(x) + PAGE_OFFSET)); \ ++ }) ++ ++#define virt_to_page(ka) ({ \ ++ struct page *_p; \ ++ if ((unsigned long)(ka) >= VMALLOC_START) { \ ++ _p = vmalloc_to_page((void *)(ka)); \ ++ BUG_ON(!_p); \ ++ } else \ ++ _p = mem_map+(__pa(ka) >> PAGE_SHIFT); \ ++ (_p); \ ++ }) ++#endif ++ ++ + #define VALID_PAGE(page) ((page - mem_map) < max_mapnr) + + #define VM_DATA_DEFAULT_FLAGS (VM_READ | VM_WRITE | VM_EXEC | \ + +%diffstat + arch/i386/config.in | 3 + fs/proc/proc_misc.c | 19 + + include/asm-i386/io.h | 10 + include/asm-i386/page.h | 40 +++ + include/linux/slab.h | 1 + init/main.c | 3 + mm/slab.c | 487 ++++++++++++++++++++++++++++++++++++++++++++++++ + mm/vmalloc.c | 4 + 8 files changed, 565 insertions(+), 2 deletions(-) + diff --git a/lustre/kernel_patches/patches/vfs_intent-2.4.19-pre1.patch b/lustre/kernel_patches/patches/vfs_intent-2.4.19-pre1.patch new file mode 100644 index 0000000..43abf92 --- /dev/null +++ b/lustre/kernel_patches/patches/vfs_intent-2.4.19-pre1.patch @@ -0,0 +1,1855 @@ + fs/dcache.c | 19 ++ + fs/exec.c | 17 +- + fs/namei.c | 295 +++++++++++++++++++++++++++++++++++++++------- + fs/namespace.c | 28 +++- + fs/open.c | 172 +++++++++++++++++++------- + fs/stat.c | 52 +++++--- + include/linux/dcache.h | 60 +++++++++ + include/linux/fs.h | 32 ++++ + include/linux/fs_struct.h | 4 + kernel/exit.c | 3 + kernel/fork.c | 3 + kernel/ksyms.c | 1 + 12 files changed, 558 insertions(+), 128 deletions(-) + +Index: linux-2.4.19-pre1/fs/exec.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/exec.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/exec.c 2003-11-21 02:51:38.000000000 +0300 +@@ -107,8 +107,10 @@ + struct file * file; + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + +- error = user_path_walk(library, &nd); ++ error = user_path_walk_it(library, &nd, &it); + if (error) + goto out; + +@@ -120,7 +122,8 @@ + if (error) + goto exit; + +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + error = PTR_ERR(file); + if (IS_ERR(file)) + goto out; +@@ -342,9 +345,11 @@ + struct inode *inode; + struct file *file; + int err = 0; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + + if (path_init(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- err = path_walk(name, &nd); ++ err = path_walk_it(name, &nd, &it); + file = ERR_PTR(err); + if (!err) { + inode = nd.dentry->d_inode; +@@ -356,7 +361,8 @@ + err = -EACCES; + file = ERR_PTR(err); + if (!err) { +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + if (!IS_ERR(file)) { + err = deny_write_access(file); + if (err) { +@@ -368,6 +374,7 @@ + return file; + } + } ++ intent_release(&it); + path_release(&nd); + } + goto out; +@@ -970,7 +977,7 @@ + goto close_fail; + if (!file->f_op->write) + goto close_fail; +- if (do_truncate(file->f_dentry, 0) != 0) ++ if (do_truncate(file->f_dentry, 0, 0) != 0) + goto close_fail; + + retval = binfmt->core_dump(signr, regs, file); +Index: linux-2.4.19-pre1/fs/dcache.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/dcache.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/dcache.c 2003-11-21 02:51:38.000000000 +0300 +@@ -181,6 +181,13 @@ + spin_unlock(&dcache_lock); + return 0; + } ++ ++ /* network invalidation by Lustre */ ++ if (dentry->d_flags & DCACHE_LUSTRE_INVALID) { ++ spin_unlock(&dcache_lock); ++ return 0; ++ } ++ + /* + * Check whether to do a partial shrink_dcache + * to get rid of unused child entries. +@@ -831,13 +838,19 @@ + * Adds a dentry to the hash according to its name. + */ + +-void d_rehash(struct dentry * entry) ++void __d_rehash(struct dentry * entry, int lock) + { + struct list_head *list = d_hash(entry->d_parent, entry->d_name.hash); + if (!list_empty(&entry->d_hash)) BUG(); +- spin_lock(&dcache_lock); ++ if (lock) spin_lock(&dcache_lock); + list_add(&entry->d_hash, list); +- spin_unlock(&dcache_lock); ++ if (lock) spin_unlock(&dcache_lock); ++} ++EXPORT_SYMBOL(__d_rehash); ++ ++void d_rehash(struct dentry * entry) ++{ ++ __d_rehash(entry, 1); + } + + #define do_switch(x,y) do { \ +Index: linux-2.4.19-pre1/fs/namespace.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/namespace.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/namespace.c 2003-11-21 02:51:38.000000000 +0300 +@@ -107,6 +107,7 @@ + { + old_nd->dentry = mnt->mnt_mountpoint; + old_nd->mnt = mnt->mnt_parent; ++ UNPIN(old_nd->dentry, old_nd->mnt, 1); + mnt->mnt_parent = mnt; + mnt->mnt_mountpoint = mnt->mnt_root; + list_del_init(&mnt->mnt_child); +@@ -118,6 +119,7 @@ + { + mnt->mnt_parent = mntget(nd->mnt); + mnt->mnt_mountpoint = dget(nd->dentry); ++ PIN(nd->dentry, nd->mnt, 1); + list_add(&mnt->mnt_hash, mount_hashtable+hash(nd->mnt, nd->dentry)); + list_add(&mnt->mnt_child, &nd->mnt->mnt_mounts); + nd->dentry->d_mounted++; +@@ -500,15 +502,18 @@ + { + struct nameidata old_nd; + struct vfsmount *mnt = NULL; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int err = mount_is_safe(nd); + if (err) + return err; + if (!old_name || !*old_name) + return -EINVAL; + if (path_init(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd)) +- err = path_walk(old_name, &old_nd); +- if (err) ++ err = path_walk_it(old_name, &old_nd, &it); ++ if (err) { ++ intent_release(&it); + return err; ++ } + + down(&mount_sem); + err = -EINVAL; +@@ -531,6 +536,7 @@ + } + + up(&mount_sem); ++ intent_release(&it); + path_release(&old_nd); + return err; + } +@@ -706,6 +712,7 @@ + unsigned long flags, void *data_page) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int retval = 0; + int mnt_flags = 0; + +@@ -731,9 +738,11 @@ + + /* ... and get the mountpoint */ + if (path_init(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- retval = path_walk(dir_name, &nd); +- if (retval) ++ retval = path_walk_it(dir_name, &nd, &it); ++ if (retval) { ++ intent_release(&it); + return retval; ++ } + + if (flags & MS_REMOUNT) + retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags, +@@ -745,6 +754,8 @@ + else + retval = do_add_mount(&nd, type_page, flags, mnt_flags, + dev_name, data_page); ++ ++ intent_release(&it); + path_release(&nd); + return retval; + } +@@ -830,6 +841,8 @@ + { + struct vfsmount *tmp; + struct nameidata new_nd, old_nd, parent_nd, root_parent, user_nd; ++ struct lookup_intent new_it = { .it_op = IT_GETATTR }; ++ struct lookup_intent old_it = { .it_op = IT_GETATTR }; + char *name; + int error; + +@@ -844,7 +857,7 @@ + goto out0; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd)) +- error = path_walk(name, &new_nd); ++ error = path_walk_it(name, &new_nd, &new_it); + putname(name); + if (error) + goto out0; +@@ -858,7 +871,7 @@ + goto out1; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd)) +- error = path_walk(name, &old_nd); ++ error = path_walk_it(name, &old_nd, &old_it); + putname(name); + if (error) + goto out1; +@@ -914,8 +927,10 @@ + up(&old_nd.dentry->d_inode->i_zombie); + up(&mount_sem); + path_release(&user_nd); ++ intent_release(&old_it); + path_release(&old_nd); + out1: ++ intent_release(&new_it); + path_release(&new_nd); + out0: + unlock_kernel(); +Index: linux-2.4.19-pre1/fs/namei.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/namei.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/namei.c 2003-11-21 02:52:01.000000000 +0300 +@@ -94,6 +94,13 @@ + * XEmacs seems to be relying on it... + */ + ++void intent_release(struct lookup_intent *it) ++{ ++ if (it && it->it_op_release) ++ it->it_op_release(it); ++ ++} ++ + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the + * kernel data space before using them.. +@@ -260,10 +267,19 @@ + * Internal lookup() using the new generic dcache. + * SMP-safe + */ +-static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *cached_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * dentry = d_lookup(parent, name); + ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, flags, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ dentry = NULL; ++ } ++ return dentry; ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + if (!dentry->d_op->d_revalidate(dentry, flags) && !d_invalidate(dentry)) { + dput(dentry); +@@ -281,11 +297,15 @@ + * make sure that nobody added the entry to the dcache in the meantime.. + * SMP-safe + */ +-static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *real_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * result; + struct inode *dir = parent->d_inode; ++ int counter = 0; + ++again: ++ counter++; + down(&dir->i_sem); + /* + * First re-do the cached lookup just in case it was created +@@ -300,6 +320,9 @@ + result = ERR_PTR(-ENOMEM); + if (dentry) { + lock_kernel(); ++ if (dir->i_op->lookup_it) ++ result = dir->i_op->lookup_it(dir, dentry, it, flags); ++ else + result = dir->i_op->lookup(dir, dentry); + unlock_kernel(); + if (result) +@@ -321,6 +344,15 @@ + dput(result); + result = ERR_PTR(-ENOENT); + } ++ } else if (result->d_op && result->d_op->d_revalidate_it) { ++ if (!result->d_op->d_revalidate_it(result, flags, it) && ++ !d_invalidate(result)) { ++ dput(result); ++ if (counter > 10) ++ result = ERR_PTR(-ESTALE); ++ if (!IS_ERR(result)) ++ goto again; ++ } + } + return result; + } +@@ -332,7 +364,8 @@ + * Without that kind of total limit, nasty chains of consecutive + * symlinks can cause almost arbitrarily long lookups. + */ +-static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd) ++static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd, ++ struct lookup_intent *it) + { + int err; + if (current->link_count >= 5) +@@ -346,10 +379,12 @@ + current->link_count++; + current->total_link_count++; + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + err = dentry->d_inode->i_op->follow_link(dentry, nd); + current->link_count--; + return err; + loop: ++ intent_release(it); + path_release(nd); + return -ELOOP; + } +@@ -445,7 +480,8 @@ + * + * We expect 'base' to be positive and a directory. + */ +-int link_path_walk(const char * name, struct nameidata *nd) ++int link_path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it) + { + struct dentry *dentry; + struct inode *inode; +@@ -455,7 +491,7 @@ + while (*name=='/') + name++; + if (!*name) +- goto return_base; ++ goto return_reval; + + inode = nd->dentry->d_inode; + if (current->link_count) +@@ -518,9 +554,10 @@ + break; + } + /* This does the actual lookups.. */ +- dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE, NULL); + if (!dentry) { +- dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE, ++ NULL); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -538,7 +575,7 @@ + goto out_dput; + + if (inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, NULL); + dput(dentry); + if (err) + goto return_err; +@@ -554,7 +591,7 @@ + nd->dentry = dentry; + } + err = -ENOTDIR; +- if (!inode->i_op->lookup) ++ if (!inode->i_op->lookup && !inode->i_op->lookup_it) + break; + continue; + /* here ends the main loop */ +@@ -574,16 +611,16 @@ + inode = nd->dentry->d_inode; + /* fallthrough */ + case 1: +- goto return_base; ++ goto return_reval; + } + if (nd->dentry->d_op && nd->dentry->d_op->d_hash) { + err = nd->dentry->d_op->d_hash(nd->dentry, &this); + if (err < 0) + break; + } +- dentry = cached_lookup(nd->dentry, &this, 0); ++ dentry = cached_lookup(nd->dentry, &this, 0, it); + if (!dentry) { +- dentry = real_lookup(nd->dentry, &this, 0); ++ dentry = real_lookup(nd->dentry, &this, 0, it); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -593,7 +630,7 @@ + inode = dentry->d_inode; + if ((lookup_flags & LOOKUP_FOLLOW) + && inode && inode->i_op && inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, it); + dput(dentry); + if (err) + goto return_err; +@@ -607,7 +644,8 @@ + goto no_inode; + if (lookup_flags & LOOKUP_DIRECTORY) { + err = -ENOTDIR; +- if (!inode->i_op || !inode->i_op->lookup) ++ if (!inode->i_op || ++ (!inode->i_op->lookup && !inode->i_op->lookup_it)) + break; + } + goto return_base; +@@ -625,21 +663,66 @@ + nd->last_type = LAST_DOT; + else if (this.len == 2 && this.name[1] == '.') + nd->last_type = LAST_DOTDOT; ++return_reval: ++ /* ++ * We bypassed the ordinary revalidation routines. ++ * Check the cached dentry for staleness. ++ */ ++ dentry = nd->dentry; ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ err = -ESTALE; ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) { ++ struct dentry *new; ++ err = permission(dentry->d_parent->d_inode, ++ MAY_EXEC); ++ if (err) ++ break; ++ new = real_lookup(dentry->d_parent, ++ &dentry->d_name, 0, NULL); ++ d_invalidate(dentry); ++ dput(dentry); ++ if (IS_ERR(new)) { ++ err = PTR_ERR(new); ++ break; ++ } ++ nd->dentry = new; ++ } ++ } else ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { ++ err = -ESTALE; ++ if (!dentry->d_op->d_revalidate(dentry, 0)) { ++ d_invalidate(dentry); ++ break; ++ } ++ } + return_base: + return 0; + out_dput: + dput(dentry); + break; + } ++ if (err) ++ intent_release(it); + path_release(nd); + return_err: + return err; + } + ++int link_path_walk(const char * name, struct nameidata *nd) ++{ ++ return link_path_walk_it(name, nd, NULL); ++} ++ ++int path_walk_it(const char * name, struct nameidata *nd, struct lookup_intent *it) ++{ ++ current->total_link_count = 0; ++ return link_path_walk_it(name, nd, it); ++} ++ + int path_walk(const char * name, struct nameidata *nd) + { + current->total_link_count = 0; +- return link_path_walk(name, nd); ++ return link_path_walk_it(name, nd, NULL); + } + + /* SMP-safe */ +@@ -728,6 +811,7 @@ + { + nd->last_type = LAST_ROOT; /* if there are only slashes... */ + nd->flags = flags; ++ nd->intent = NULL; + if (*name=='/') + return walk_init_root(name,nd); + read_lock(¤t->fs->lock); +@@ -742,7 +826,8 @@ + * needs parent already locked. Doesn't follow mounts. + * SMP-safe. + */ +-struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++struct dentry * lookup_hash_it(struct qstr *name, struct dentry * base, ++ struct lookup_intent *it) + { + struct dentry * dentry; + struct inode *inode; +@@ -765,13 +850,16 @@ + goto out; + } + +- dentry = cached_lookup(base, name, 0); ++ dentry = cached_lookup(base, name, 0, it); + if (!dentry) { + struct dentry *new = d_alloc(base, name); + dentry = ERR_PTR(-ENOMEM); + if (!new) + goto out; + lock_kernel(); ++ if (inode->i_op->lookup_it) ++ dentry = inode->i_op->lookup_it(inode, new, it, 0); ++ else + dentry = inode->i_op->lookup(inode, new); + unlock_kernel(); + if (!dentry) +@@ -783,6 +871,12 @@ + return dentry; + } + ++struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++{ ++ return lookup_hash_it(name, base, NULL); ++} ++ ++ + /* SMP-safe */ + struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) + { +@@ -804,7 +898,7 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash(&this, base); ++ return lookup_hash_it(&this, base, NULL); + access: + return ERR_PTR(-EACCES); + } +@@ -836,6 +930,23 @@ + return err; + } + ++int __user_walk_it(const char *name, unsigned flags, struct nameidata *nd, ++ struct lookup_intent *it) ++{ ++ char *tmp; ++ int err; ++ ++ tmp = getname(name); ++ err = PTR_ERR(tmp); ++ if (!IS_ERR(tmp)) { ++ err = 0; ++ if (path_init(tmp, flags, nd)) ++ err = path_walk_it(tmp, nd, it); ++ putname(tmp); ++ } ++ return err; ++} ++ + /* + * It's inline, so penalty for filesystems that don't use sticky bit is + * minimal. +@@ -929,7 +1040,8 @@ + return retval; + } + +-int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++static int vfs_create_it(struct inode *dir, struct dentry *dentry, int mode, ++ struct lookup_intent *it) + { + int error; + +@@ -942,12 +1054,15 @@ + goto exit_lock; + + error = -EACCES; /* shouldn't it be ENOSYS? */ +- if (!dir->i_op || !dir->i_op->create) ++ if (!dir->i_op || (!dir->i_op->create && !dir->i_op->create_it)) + goto exit_lock; + + DQUOT_INIT(dir); + lock_kernel(); +- error = dir->i_op->create(dir, dentry, mode); ++ if (dir->i_op->create_it) ++ error = dir->i_op->create_it(dir, dentry, mode, it); ++ else ++ error = dir->i_op->create(dir, dentry, mode); + unlock_kernel(); + exit_lock: + up(&dir->i_zombie); +@@ -956,6 +1071,11 @@ + return error; + } + ++int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++{ ++ return vfs_create_it(dir, dentry, mode, NULL); ++} ++ + /* + * open_namei() + * +@@ -970,7 +1090,8 @@ + * for symlinks (where the permissions are checked later). + * SMP-safe + */ +-int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) ++int open_namei_it(const char *pathname, int flag, int mode, ++ struct nameidata *nd, struct lookup_intent *it) + { + int acc_mode, error = 0; + struct inode *inode; +@@ -980,12 +1101,14 @@ + + acc_mode = ACC_MODE(flag); + ++ if (it) ++ it->it_flags = flag; + /* + * The simplest case - just a plain lookup. + */ + if (!(flag & O_CREAT)) { + if (path_init(pathname, lookup_flags(flag), nd)) +- error = path_walk(pathname, nd); ++ error = path_walk_it(pathname, nd, it); + if (error) + return error; + dentry = nd->dentry; +@@ -995,6 +1118,10 @@ + /* + * Create - we need to know the parent. + */ ++ if (it) { ++ it->it_create_mode = mode; ++ it->it_op |= IT_CREAT; ++ } + if (path_init(pathname, LOOKUP_PARENT, nd)) + error = path_walk(pathname, nd); + if (error) +@@ -1011,7 +1138,7 @@ + + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + + do_last: + error = PTR_ERR(dentry); +@@ -1020,10 +1147,11 @@ + goto exit; + } + ++ it->it_create_mode = mode; + /* Negative dentry, just create the file */ + if (!dentry->d_inode) { +- error = vfs_create(dir->d_inode, dentry, +- mode & ~current->fs->umask); ++ error = vfs_create_it(dir->d_inode, dentry, ++ mode & ~current->fs->umask, it); + up(&dir->d_inode->i_sem); + dput(nd->dentry); + nd->dentry = dentry; +@@ -1127,7 +1255,7 @@ + if (!error) { + DQUOT_INIT(inode); + +- error = do_truncate(dentry, 0); ++ error = do_truncate(dentry, 0, 1); + } + put_write_access(inode); + if (error) +@@ -1139,8 +1267,10 @@ + return 0; + + exit_dput: ++ intent_release(it); + dput(dentry); + exit: ++ intent_release(it); + path_release(nd); + return error; + +@@ -1159,7 +1289,10 @@ + * are done. Procfs-like symlinks just set LAST_BIND. + */ + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + error = dentry->d_inode->i_op->follow_link(dentry, nd); ++ if (error) ++ intent_release(it); + dput(dentry); + if (error) + return error; +@@ -1181,13 +1314,20 @@ + } + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + putname(nd->last.name); + goto do_last; + } + ++int open_namei(const char *pathname, int flag, int mode, struct nameidata *nd) ++{ ++ return open_namei_it(pathname, flag, mode, nd, NULL); ++} ++ ++ + /* SMP-safe */ +-static struct dentry *lookup_create(struct nameidata *nd, int is_dir) ++static struct dentry *lookup_create(struct nameidata *nd, int is_dir, ++ struct lookup_intent *it) + { + struct dentry *dentry; + +@@ -1195,7 +1335,7 @@ + dentry = ERR_PTR(-EEXIST); + if (nd->last_type != LAST_NORM) + goto fail; +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + if (IS_ERR(dentry)) + goto fail; + if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) +@@ -1252,7 +1392,16 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ ++ if (nd.dentry->d_inode->i_op->mknod_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mknod_raw(&nd, mode, dev); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + + mode &= ~current->fs->umask; +@@ -1273,6 +1422,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1321,7 +1471,14 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 1); ++ if (nd.dentry->d_inode->i_op->mkdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mkdir_raw(&nd, mode); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 1, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_mkdir(nd.dentry->d_inode, dentry, +@@ -1329,6 +1486,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1428,8 +1586,16 @@ + error = -EBUSY; + goto exit1; + } ++ if (nd.dentry->d_inode->i_op->rmdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ error = op->rmdir_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_rmdir(nd.dentry->d_inode, dentry); +@@ -1488,8 +1654,15 @@ + error = -EISDIR; + if (nd.last_type != LAST_NORM) + goto exit1; ++ if (nd.dentry->d_inode->i_op->unlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->unlink_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + /* Why not before? Because we want correct error value */ +@@ -1557,15 +1730,23 @@ + error = path_walk(to, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->symlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->symlink_raw(&nd, from); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_symlink(nd.dentry->d_inode, dentry, from); + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++ out2: + path_release(&nd); +-out: ++ out: + putname(to); + } + putname(from); +@@ -1648,7 +1829,14 @@ + error = -EXDEV; + if (old_nd.mnt != nd.mnt) + goto out_release; +- new_dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->link_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->link_raw(&old_nd, &nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out_release; ++ } ++ new_dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(new_dentry); + if (!IS_ERR(new_dentry)) { + error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry); +@@ -1694,7 +1882,7 @@ + * locking]. + */ + int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + struct inode *target; +@@ -1775,7 +1963,7 @@ + } + + int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + +@@ -1866,9 +2054,18 @@ + if (newnd.last_type != LAST_NORM) + goto exit2; + ++ if (old_dir->d_inode->i_op->rename_raw) { ++ lock_kernel(); ++ error = old_dir->d_inode->i_op->rename_raw(&oldnd, &newnd); ++ unlock_kernel(); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit2; ++ } ++ + double_lock(new_dir, old_dir); + +- old_dentry = lookup_hash(&oldnd.last, old_dir); ++ old_dentry = lookup_hash_it(&oldnd.last, old_dir, NULL); + error = PTR_ERR(old_dentry); + if (IS_ERR(old_dentry)) + goto exit3; +@@ -1884,16 +2081,16 @@ + if (newnd.last.name[newnd.last.len]) + goto exit4; + } +- new_dentry = lookup_hash(&newnd.last, new_dir); ++ new_dentry = lookup_hash_it(&newnd.last, new_dir, NULL); + error = PTR_ERR(new_dentry); + if (IS_ERR(new_dentry)) + goto exit4; + ++ + lock_kernel(); + error = vfs_rename(old_dir->d_inode, old_dentry, + new_dir->d_inode, new_dentry); + unlock_kernel(); +- + dput(new_dentry); + exit4: + dput(old_dentry); +@@ -1944,20 +2141,26 @@ + } + + static inline int +-__vfs_follow_link(struct nameidata *nd, const char *link) ++__vfs_follow_link(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) + { + int res = 0; + char *name; + if (IS_ERR(link)) + goto fail; + ++ if (it == NULL) ++ it = nd->intent; ++ else if (it != nd->intent) ++ printk("it != nd->intent: tell phil@clusterfs.com\n"); ++ + if (*link == '/') { + path_release(nd); + if (!walk_init_root(link, nd)) + /* weird __emul_prefix() stuff did it */ + goto out; + } +- res = link_path_walk(link, nd); ++ res = link_path_walk_it(link, nd, it); + out: + if (current->link_count || res || nd->last_type!=LAST_NORM) + return res; +@@ -1979,7 +2182,13 @@ + + int vfs_follow_link(struct nameidata *nd, const char *link) + { +- return __vfs_follow_link(nd, link); ++ return __vfs_follow_link(nd, link, NULL); ++} ++ ++int vfs_follow_link_it(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) ++{ ++ return __vfs_follow_link(nd, link, it); + } + + /* get the link contents into pagecache */ +@@ -2021,7 +2230,7 @@ + { + struct page *page = NULL; + char *s = page_getlink(dentry, &page); +- int res = __vfs_follow_link(nd, s); ++ int res = __vfs_follow_link(nd, s, NULL); + if (page) { + kunmap(page); + page_cache_release(page); +Index: linux-2.4.19-pre1/fs/open.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/open.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/open.c 2003-11-21 02:51:38.000000000 +0300 +@@ -19,6 +19,8 @@ + #include + + #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) ++extern int path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it); + + int vfs_statfs(struct super_block *sb, struct statfs *buf) + { +@@ -71,9 +73,10 @@ + return error; + } + +-int do_truncate(struct dentry *dentry, loff_t length) ++int do_truncate(struct dentry *dentry, loff_t length, int called_from_open) + { + struct inode *inode = dentry->d_inode; ++ struct inode_operations *op = dentry->d_inode->i_op; + int error; + struct iattr newattrs; + +@@ -84,7 +87,13 @@ + down(&inode->i_sem); + newattrs.ia_size = length; + newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME; +- error = notify_change(dentry, &newattrs); ++ if (called_from_open) ++ newattrs.ia_valid |= ATTR_FROM_OPEN; ++ if (op->setattr_raw) { ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ } else ++ error = notify_change(dentry, &newattrs); + up(&inode->i_sem); + return error; + } +@@ -94,12 +103,13 @@ + struct nameidata nd; + struct inode * inode; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + error = -EINVAL; + if (length < 0) /* sorry, but loff_t says... */ + goto out; + +- error = user_path_walk(path, &nd); ++ error = user_path_walk_it(path, &nd, &it); + if (error) + goto out; + inode = nd.dentry->d_inode; +@@ -139,11 +149,13 @@ + error = locks_verify_truncate(inode, NULL, length); + if (!error) { + DQUOT_INIT(inode); +- error = do_truncate(nd.dentry, length); ++ intent_release(&it); ++ error = do_truncate(nd.dentry, length, 0); + } + put_write_access(inode); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -191,7 +203,7 @@ + + error = locks_verify_truncate(inode, file, length); + if (!error) +- error = do_truncate(dentry, length); ++ error = do_truncate(dentry, length, 0); + out_putf: + fput(file); + out: +@@ -236,11 +248,13 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -255,11 +269,25 @@ + goto dput_and_out; + + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!times) { + if (current->fsuid != inode->i_uid && + (error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; + } ++ + error = notify_change(nd.dentry, &newattrs); + dput_and_out: + path_release(&nd); +@@ -280,12 +308,14 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -300,7 +330,20 @@ + newattrs.ia_atime = times[0].tv_sec; + newattrs.ia_mtime = times[1].tv_sec; + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!utimes) { + if ((error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; + } +@@ -322,6 +365,7 @@ + int old_fsuid, old_fsgid; + kernel_cap_t old_cap; + int res; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ + return -EINVAL; +@@ -339,13 +383,14 @@ + else + current->cap_effective = current->cap_permitted; + +- res = user_path_walk(filename, &nd); ++ res = user_path_walk_it(filename, &nd, &it); + if (!res) { + res = permission(nd.dentry->d_inode, mode); + /* SuS v2 requires we report a read only fs too */ + if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode) + && !special_file(nd.dentry->d_inode->i_mode)) + res = -EROFS; ++ intent_release(&it); + path_release(&nd); + } + +@@ -361,6 +406,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -369,7 +415,7 @@ + + error = 0; + if (path_init(name,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd)) +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -381,6 +427,7 @@ + set_fs_pwd(current->fs, nd.mnt, nd.dentry); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -421,6 +468,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -429,7 +477,7 @@ + + path_init(name, LOOKUP_POSITIVE | LOOKUP_FOLLOW | + LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd); +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -446,39 +494,56 @@ + set_fs_altroot(); + error = 0; + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; + } + +-asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++int chmod_common(struct dentry *dentry, mode_t mode) + { +- struct inode * inode; +- struct dentry * dentry; +- struct file * file; +- int err = -EBADF; ++ struct inode *inode = dentry->d_inode; + struct iattr newattrs; ++ int err = -EROFS; + +- file = fget(fd); +- if (!file) ++ if (IS_RDONLY(inode)) + goto out; + +- dentry = file->f_dentry; +- inode = dentry->d_inode; ++ if (inode->i_op->setattr_raw) { ++ newattrs.ia_mode = mode; ++ newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ err = inode->i_op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (err != -EOPNOTSUPP) ++ goto out; ++ } + +- err = -EROFS; +- if (IS_RDONLY(inode)) +- goto out_putf; + err = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto out_putf; ++ goto out; ++ + if (mode == (mode_t) -1) + mode = inode->i_mode; + newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); + newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; + err = notify_change(dentry, &newattrs); + +-out_putf: ++out: ++ return err; ++} ++ ++asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++{ ++ struct file * file; ++ int err = -EBADF; ++ ++ file = fget(fd); ++ if (!file) ++ goto out; ++ ++ err = chmod_common(file->f_dentry, mode); ++ + fput(file); + out: + return err; +@@ -487,30 +552,14 @@ + asmlinkage long sys_chmod(const char * filename, mode_t mode) + { + struct nameidata nd; +- struct inode * inode; + int error; +- struct iattr newattrs; + + error = user_path_walk(filename, &nd); + if (error) + goto out; +- inode = nd.dentry->d_inode; +- +- error = -EROFS; +- if (IS_RDONLY(inode)) +- goto dput_and_out; + +- error = -EPERM; +- if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto dput_and_out; ++ error = chmod_common(nd.dentry, mode); + +- if (mode == (mode_t) -1) +- mode = inode->i_mode; +- newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); +- newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; +- error = notify_change(nd.dentry, &newattrs); +- +-dput_and_out: + path_release(&nd); + out: + return error; +@@ -530,6 +579,20 @@ + error = -EROFS; + if (IS_RDONLY(inode)) + goto out; ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = dentry->d_inode->i_op; ++ ++ newattrs.ia_uid = user; ++ newattrs.ia_gid = group; ++ newattrs.ia_valid = ATTR_UID | ATTR_GID | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ return error; ++ } ++ + error = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) + goto out; +@@ -634,6 +697,7 @@ + { + int namei_flags, error; + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_OPEN }; + + namei_flags = flags; + if ((namei_flags+1) & O_ACCMODE) +@@ -641,14 +705,15 @@ + if (namei_flags & O_TRUNC) + namei_flags |= 2; + +- error = open_namei(filename, namei_flags, mode, &nd); +- if (!error) +- return dentry_open(nd.dentry, nd.mnt, flags); ++ error = open_namei_it(filename, namei_flags, mode, &nd, &it); ++ if (error) ++ return ERR_PTR(error); + +- return ERR_PTR(error); ++ return dentry_open_it(nd.dentry, nd.mnt, flags, &it); + } + +-struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it) + { + struct file * f; + struct inode *inode; +@@ -685,12 +750,15 @@ + } + + if (f->f_op && f->f_op->open) { ++ f->f_it = it; + error = f->f_op->open(inode,f); ++ f->f_it = NULL; + if (error) + goto cleanup_all; + } + f->f_flags &= ~(O_CREAT | O_EXCL | O_NOCTTY | O_TRUNC); + ++ intent_release(it); + return f; + + cleanup_all: +@@ -705,11 +773,17 @@ + cleanup_file: + put_filp(f); + cleanup_dentry: ++ intent_release(it); + dput(dentry); + mntput(mnt); + return ERR_PTR(error); + } + ++struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++{ ++ return dentry_open_it(dentry, mnt, flags, NULL); ++} ++ + /* + * Find an empty file descriptor entry, and mark it busy. + */ +Index: linux-2.4.19-pre1/fs/stat.c +=================================================================== +--- linux-2.4.19-pre1.orig/fs/stat.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/fs/stat.c 2003-11-21 02:51:38.000000000 +0300 +@@ -17,10 +17,14 @@ + * Revalidate the inode. This is required for proper NFS attribute caching. + */ + static __inline__ int +-do_revalidate(struct dentry *dentry) ++do_revalidate(struct dentry *dentry, struct lookup_intent *it) + { + struct inode * inode = dentry->d_inode; +- if (inode->i_op && inode->i_op->revalidate) ++ if (!inode) ++ return -ENOENT; ++ if (inode->i_op && inode->i_op->revalidate_it) ++ return inode->i_op->revalidate_it(dentry, it); ++ else if (inode->i_op && inode->i_op->revalidate) + return inode->i_op->revalidate(dentry); + return 0; + } +@@ -135,13 +139,15 @@ + asmlinkage long sys_stat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -151,13 +157,15 @@ + asmlinkage long sys_newstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -172,13 +180,15 @@ + asmlinkage long sys_lstat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -189,13 +199,15 @@ + asmlinkage long sys_newlstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -216,7 +228,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_old_stat(dentry->d_inode, statbuf); + fput(f); +@@ -235,7 +247,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat(dentry->d_inode, statbuf); + fput(f); +@@ -257,7 +269,7 @@ + + error = -EINVAL; + if (inode->i_op && inode->i_op->readlink && +- !(error = do_revalidate(nd.dentry))) { ++ !(error = do_revalidate(nd.dentry, NULL))) { + UPDATE_ATIME(inode); + error = inode->i_op->readlink(nd.dentry, buf, bufsiz); + } +@@ -333,12 +345,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -348,12 +362,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -368,7 +384,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat64(dentry->d_inode, statbuf); + fput(f); +Index: linux-2.4.19-pre1/include/linux/dcache.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/dcache.h 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/dcache.h 2003-11-21 02:51:38.000000000 +0300 +@@ -5,6 +5,51 @@ + + #include + #include ++#include ++ ++#define IT_OPEN 0x0001 ++#define IT_CREAT 0x0002 ++#define IT_READDIR 0x0004 ++#define IT_GETATTR 0x0008 ++#define IT_LOOKUP 0x0010 ++#define IT_UNLINK 0x0020 ++#define IT_GETXATTR 0x0040 ++#define IT_EXEC 0x0080 ++#define IT_PIN 0x0100 ++ ++#define IT_FL_LOCKED 0x0001 ++#define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */ ++ ++#define INTENT_MAGIC 0x19620323 ++ ++ ++struct lustre_intent_data { ++ int it_disposition; ++ int it_status; ++ __u64 it_lock_handle; ++ void *it_data; ++ int it_lock_mode; ++ int it_int_flags; ++}; ++struct lookup_intent { ++ int it_magic; ++ void (*it_op_release)(struct lookup_intent *); ++ int it_op; ++ int it_flags; ++ int it_create_mode; ++ union { ++ struct lustre_intent_data lustre; ++ } d; ++}; ++ ++static inline void intent_init(struct lookup_intent *it, int op, int flags) ++{ ++ memset(it, 0, sizeof(*it)); ++ it->it_magic = INTENT_MAGIC; ++ it->it_op = op; ++ it->it_flags = flags; ++} ++ + + /* + * linux/include/linux/dcache.h +@@ -90,8 +135,22 @@ + int (*d_delete)(struct dentry *); + void (*d_release)(struct dentry *); + void (*d_iput)(struct dentry *, struct inode *); ++ int (*d_revalidate_it)(struct dentry *, int, struct lookup_intent *); ++ void (*d_pin)(struct dentry *, struct vfsmount * , int); ++ void (*d_unpin)(struct dentry *, struct vfsmount *, int); + }; + ++#define PIN(de,mnt,flag) if (de->d_op && de->d_op->d_pin) \ ++ de->d_op->d_pin(de, mnt, flag); ++#define UNPIN(de,mnt,flag) if (de->d_op && de->d_op->d_unpin) \ ++ de->d_op->d_unpin(de, mnt, flag); ++ ++ ++/* defined in fs/namei.c */ ++extern void intent_release(struct lookup_intent *it); ++/* defined in fs/dcache.c */ ++extern void __d_rehash(struct dentry * entry, int lock); ++ + /* the dentry parameter passed to d_hash and d_compare is the parent + * directory of the entries to be compared. It is used in case these + * functions need any directory specific information for determining +@@ -123,6 +182,7 @@ + * s_nfsd_free_path semaphore will be down + */ + #define DCACHE_REFERENCED 0x0008 /* Recently used, don't discard. */ ++#define DCACHE_LUSTRE_INVALID 0x0010 /* Lustre invalidated */ + + extern spinlock_t dcache_lock; + +Index: linux-2.4.19-pre1/include/linux/fs.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs.h 2003-11-21 02:51:37.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs.h 2003-11-21 02:51:38.000000000 +0300 +@@ -73,6 +73,7 @@ + + #define FMODE_READ 1 + #define FMODE_WRITE 2 ++#define FMODE_EXEC 4 + + #define READ 0 + #define WRITE 1 +@@ -332,6 +333,9 @@ + #define ATTR_MTIME_SET 256 + #define ATTR_FORCE 512 /* Not a change, but a change it */ + #define ATTR_ATTR_FLAG 1024 ++#define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */ ++#define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */ ++#define ATTR_CTIME_SET 0x2000 + + /* + * This is the Inode Attributes structure, used for notify_change(). It +@@ -465,6 +469,7 @@ + struct pipe_inode_info *i_pipe; + struct block_device *i_bdev; + struct char_device *i_cdev; ++ void *i_filterdata; + + unsigned long i_dnotify_mask; /* Directory notify events */ + struct dnotify_struct *i_dnotify; /* for directory notifications */ +@@ -534,6 +539,7 @@ + + /* needed for tty driver, and maybe others */ + void *private_data; ++ struct lookup_intent *f_it; + + /* preallocated helper kiobuf to speedup O_DIRECT */ + struct kiobuf *f_iobuf; +@@ -644,6 +650,7 @@ + struct qstr last; + unsigned int flags; + int last_type; ++ struct lookup_intent *intent; + }; + + #define DQUOT_USR_ENABLED 0x01 /* User diskquotas enabled */ +@@ -777,7 +784,8 @@ + extern int vfs_link(struct dentry *, struct inode *, struct dentry *); + extern int vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_unlink(struct inode *, struct dentry *); +-extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); ++int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, ++ struct inode *new_dir, struct dentry *new_dentry); + + /* + * File types +@@ -837,21 +845,32 @@ + + struct inode_operations { + int (*create) (struct inode *,struct dentry *,int); ++ int (*create_it) (struct inode *,struct dentry *,int, struct lookup_intent *); + struct dentry * (*lookup) (struct inode *,struct dentry *); ++ struct dentry * (*lookup_it) (struct inode *,struct dentry *, struct lookup_intent *, int flags); + int (*link) (struct dentry *,struct inode *,struct dentry *); ++ int (*link_raw) (struct nameidata *,struct nameidata *); + int (*unlink) (struct inode *,struct dentry *); ++ int (*unlink_raw) (struct nameidata *); + int (*symlink) (struct inode *,struct dentry *,const char *); ++ int (*symlink_raw) (struct nameidata *,const char *); + int (*mkdir) (struct inode *,struct dentry *,int); ++ int (*mkdir_raw) (struct nameidata *,int); + int (*rmdir) (struct inode *,struct dentry *); ++ int (*rmdir_raw) (struct nameidata *); + int (*mknod) (struct inode *,struct dentry *,int,int); ++ int (*mknod_raw) (struct nameidata *,int,dev_t); + int (*rename) (struct inode *, struct dentry *, + struct inode *, struct dentry *); ++ int (*rename_raw) (struct nameidata *, struct nameidata *); + int (*readlink) (struct dentry *, char *,int); + int (*follow_link) (struct dentry *, struct nameidata *); + void (*truncate) (struct inode *); + int (*permission) (struct inode *, int); + int (*revalidate) (struct dentry *); ++ int (*revalidate_it) (struct dentry *, struct lookup_intent *); + int (*setattr) (struct dentry *, struct iattr *); ++ int (*setattr_raw) (struct inode *, struct iattr *); + int (*getattr) (struct dentry *, struct iattr *); + }; + +@@ -1049,10 +1068,14 @@ + + asmlinkage long sys_open(const char *, int, int); + asmlinkage long sys_close(unsigned int); /* yes, it's really unsigned */ +-extern int do_truncate(struct dentry *, loff_t start); ++extern int do_truncate(struct dentry *, loff_t start, int called_from_open); + + extern struct file *filp_open(const char *, int, int); + extern struct file * dentry_open(struct dentry *, struct vfsmount *, int); ++extern int open_namei_it(const char *filename, int namei_flags, int mode, ++ struct nameidata *nd, struct lookup_intent *it); ++extern struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it); + extern int filp_close(struct file *, fl_owner_t id); + extern char * getname(const char *); + +@@ -1306,6 +1329,7 @@ + extern loff_t default_llseek(struct file *file, loff_t offset, int origin); + + extern int FASTCALL(__user_walk(const char *, unsigned, struct nameidata *)); ++extern int FASTCALL(__user_walk_it(const char *, unsigned, struct nameidata *, struct lookup_intent *it)); + extern int FASTCALL(path_init(const char *, unsigned, struct nameidata *)); + extern int FASTCALL(path_walk(const char *, struct nameidata *)); + extern int FASTCALL(link_path_walk(const char *, struct nameidata *)); +@@ -1316,6 +1340,8 @@ + extern struct dentry * lookup_hash(struct qstr *, struct dentry *); + #define user_path_walk(name,nd) __user_walk(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd) + #define user_path_walk_link(name,nd) __user_walk(name, LOOKUP_POSITIVE, nd) ++#define user_path_walk_it(name,nd,it) __user_walk_it(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd, it) ++#define user_path_walk_link_it(name,nd,it) __user_walk_it(name, LOOKUP_POSITIVE, nd, it) + + extern void iput(struct inode *); + extern void force_delete(struct inode *); +@@ -1423,6 +1449,8 @@ + + extern int vfs_readlink(struct dentry *, char *, int, const char *); + extern int vfs_follow_link(struct nameidata *, const char *); ++extern int vfs_follow_link_it(struct nameidata *, const char *, ++ struct lookup_intent *it); + extern int page_readlink(struct dentry *, char *, int); + extern int page_follow_link(struct dentry *, struct nameidata *); + extern struct inode_operations page_symlink_inode_operations; +Index: linux-2.4.19-pre1/include/linux/fs_struct.h +=================================================================== +--- linux-2.4.19-pre1.orig/include/linux/fs_struct.h 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/include/linux/fs_struct.h 2003-11-21 02:51:38.000000000 +0300 +@@ -34,10 +34,12 @@ + write_lock(&fs->lock); + old_root = fs->root; + old_rootmnt = fs->rootmnt; ++ PIN(dentry, mnt, 1); + fs->rootmnt = mntget(mnt); + fs->root = dget(dentry); + write_unlock(&fs->lock); + if (old_root) { ++ UNPIN(old_root, old_rootmnt, 1); + dput(old_root); + mntput(old_rootmnt); + } +@@ -57,10 +59,12 @@ + write_lock(&fs->lock); + old_pwd = fs->pwd; + old_pwdmnt = fs->pwdmnt; ++ PIN(dentry, mnt, 0); + fs->pwdmnt = mntget(mnt); + fs->pwd = dget(dentry); + write_unlock(&fs->lock); + if (old_pwd) { ++ UNPIN(old_pwd, old_pwdmnt, 0); + dput(old_pwd); + mntput(old_pwdmnt); + } +Index: linux-2.4.19-pre1/kernel/ksyms.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/ksyms.c 2003-11-21 02:51:37.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/ksyms.c 2003-11-21 02:51:38.000000000 +0300 +@@ -260,6 +260,7 @@ + EXPORT_SYMBOL(set_page_dirty); + EXPORT_SYMBOL(vfs_readlink); + EXPORT_SYMBOL(vfs_follow_link); ++EXPORT_SYMBOL(vfs_follow_link_it); + EXPORT_SYMBOL(page_readlink); + EXPORT_SYMBOL(page_follow_link); + EXPORT_SYMBOL(page_symlink_inode_operations); +Index: linux-2.4.19-pre1/kernel/fork.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/fork.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/fork.c 2003-11-21 02:51:38.000000000 +0300 +@@ -372,10 +372,13 @@ + fs->umask = old->umask; + read_lock(&old->lock); + fs->rootmnt = mntget(old->rootmnt); ++ PIN(old->pwd, old->pwdmnt, 0); ++ PIN(old->root, old->rootmnt, 1); + fs->root = dget(old->root); + fs->pwdmnt = mntget(old->pwdmnt); + fs->pwd = dget(old->pwd); + if (old->altroot) { ++ PIN(old->altroot, old->altrootmnt, 1); + fs->altrootmnt = mntget(old->altrootmnt); + fs->altroot = dget(old->altroot); + } else { +Index: linux-2.4.19-pre1/kernel/exit.c +=================================================================== +--- linux-2.4.19-pre1.orig/kernel/exit.c 2003-11-21 02:41:00.000000000 +0300 ++++ linux-2.4.19-pre1/kernel/exit.c 2003-11-21 02:51:38.000000000 +0300 +@@ -245,11 +245,14 @@ + { + /* No need to hold fs->lock if we are killing it */ + if (atomic_dec_and_test(&fs->count)) { ++ UNPIN(fs->pwd, fs->pwdmnt, 0); ++ UNPIN(fs->root, fs->rootmnt, 1); + dput(fs->root); + mntput(fs->rootmnt); + dput(fs->pwd); + mntput(fs->pwdmnt); + if (fs->altroot) { ++ UNPIN(fs->altroot, fs->altrootmnt, 1); + dput(fs->altroot); + mntput(fs->altrootmnt); + } diff --git a/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch b/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch new file mode 100644 index 0000000..4e37bb8 --- /dev/null +++ b/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch @@ -0,0 +1,1835 @@ + fs/dcache.c | 19 ++ + fs/exec.c | 17 +- + fs/namei.c | 295 +++++++++++++++++++++++++++++++++++++++------- + fs/namespace.c | 28 +++- + fs/open.c | 172 +++++++++++++++++++------- + fs/stat.c | 52 +++++--- + include/linux/dcache.h | 60 +++++++++ + include/linux/fs.h | 32 ++++ + include/linux/fs_struct.h | 4 + kernel/exit.c | 3 + kernel/fork.c | 3 + kernel/ksyms.c | 1 + 12 files changed, 558 insertions(+), 128 deletions(-) + +Index: linux-2.4.19.SuSE/fs/exec.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/exec.c Mon Jan 27 05:08:35 2003 ++++ linux-2.4.19.SuSE/fs/exec.c Sat Nov 15 17:34:06 2003 +@@ -107,8 +107,10 @@ + struct file * file; + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + +- error = user_path_walk(library, &nd); ++ error = user_path_walk_it(library, &nd, &it); + if (error) + goto out; + +@@ -120,7 +122,8 @@ + if (error) + goto exit; + +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + error = PTR_ERR(file); + if (IS_ERR(file)) + goto out; +@@ -346,9 +349,11 @@ + struct inode *inode; + struct file *file; + int err = 0; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + + if (path_init(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- err = path_walk(name, &nd); ++ err = path_walk_it(name, &nd, &it); + file = ERR_PTR(err); + if (!err) { + inode = nd.dentry->d_inode; +@@ -360,7 +365,8 @@ + err = -EACCES; + file = ERR_PTR(err); + if (!err) { +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + if (!IS_ERR(file)) { + err = deny_write_access(file); + if (err) { +@@ -372,6 +378,7 @@ + return file; + } + } ++ intent_release(&it); + path_release(&nd); + } + goto out; +@@ -981,7 +988,7 @@ + goto close_fail; + if (!file->f_op->write) + goto close_fail; +- if (do_truncate(file->f_dentry, 0) != 0) ++ if (do_truncate(file->f_dentry, 0, 0) != 0) + goto close_fail; + + retval = binfmt->core_dump(signr, regs, file); +Index: linux-2.4.19.SuSE/fs/dcache.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/dcache.c Mon Jan 27 05:08:04 2003 ++++ linux-2.4.19.SuSE/fs/dcache.c Sat Nov 15 17:29:03 2003 +@@ -186,6 +186,13 @@ + spin_unlock(&dcache_lock); + return 0; + } ++ ++ /* network invalidation by Lustre */ ++ if (dentry->d_flags & DCACHE_LUSTRE_INVALID) { ++ spin_unlock(&dcache_lock); ++ return 0; ++ } ++ + /* + * Check whether to do a partial shrink_dcache + * to get rid of unused child entries. +@@ -838,13 +845,19 @@ + * Adds a dentry to the hash according to its name. + */ + +-void d_rehash(struct dentry * entry) ++void __d_rehash(struct dentry * entry, int lock) + { + struct list_head *list = d_hash(entry->d_parent, entry->d_name.hash); + if (!list_empty(&entry->d_hash)) BUG(); +- spin_lock(&dcache_lock); ++ if (lock) spin_lock(&dcache_lock); + list_add(&entry->d_hash, list); +- spin_unlock(&dcache_lock); ++ if (lock) spin_unlock(&dcache_lock); ++} ++EXPORT_SYMBOL(__d_rehash); ++ ++void d_rehash(struct dentry * entry) ++{ ++ __d_rehash(entry, 1); + } + + #define do_switch(x,y) do { \ +Index: linux-2.4.19.SuSE/fs/namespace.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/namespace.c Mon Jan 27 05:08:07 2003 ++++ linux-2.4.19.SuSE/fs/namespace.c Sat Nov 15 17:56:42 2003 +@@ -97,6 +97,7 @@ + { + old_nd->dentry = mnt->mnt_mountpoint; + old_nd->mnt = mnt->mnt_parent; ++ UNPIN(old_nd->dentry, old_nd->mnt, 1); + mnt->mnt_parent = mnt; + mnt->mnt_mountpoint = mnt->mnt_root; + list_del_init(&mnt->mnt_child); +@@ -108,6 +109,7 @@ + { + mnt->mnt_parent = mntget(nd->mnt); + mnt->mnt_mountpoint = dget(nd->dentry); ++ PIN(nd->dentry, nd->mnt, 1); + list_add(&mnt->mnt_hash, mount_hashtable+hash(nd->mnt, nd->dentry)); + list_add(&mnt->mnt_child, &nd->mnt->mnt_mounts); + nd->dentry->d_mounted++; +@@ -491,15 +493,18 @@ + { + struct nameidata old_nd; + struct vfsmount *mnt = NULL; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int err = mount_is_safe(nd); + if (err) + return err; + if (!old_name || !*old_name) + return -EINVAL; + if (path_init(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd)) +- err = path_walk(old_name, &old_nd); +- if (err) ++ err = path_walk_it(old_name, &old_nd, &it); ++ if (err) { ++ intent_release(&it); + return err; ++ } + + down_write(¤t->namespace->sem); + err = -EINVAL; +@@ -522,6 +527,7 @@ + } + + up_write(¤t->namespace->sem); ++ intent_release(&it); + path_release(&old_nd); + return err; + } +@@ -725,6 +731,7 @@ + unsigned long flags, void *data_page) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int retval = 0; + int mnt_flags = 0; + +@@ -750,9 +757,11 @@ + + /* ... and get the mountpoint */ + if (path_init(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- retval = path_walk(dir_name, &nd); +- if (retval) ++ retval = path_walk_it(dir_name, &nd, &it); ++ if (retval) { ++ intent_release(&it); + return retval; ++ } + + if (flags & MS_REMOUNT) + retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags, +@@ -764,6 +773,8 @@ + else + retval = do_add_mount(&nd, type_page, flags, mnt_flags, + dev_name, data_page); ++ ++ intent_release(&it); + path_release(&nd); + return retval; + } +@@ -929,6 +940,8 @@ + { + struct vfsmount *tmp; + struct nameidata new_nd, old_nd, parent_nd, root_parent, user_nd; ++ struct lookup_intent new_it = { .it_op = IT_GETATTR }; ++ struct lookup_intent old_it = { .it_op = IT_GETATTR }; + char *name; + int error; + +@@ -943,7 +956,7 @@ + goto out0; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd)) +- error = path_walk(name, &new_nd); ++ error = path_walk_it(name, &new_nd, &new_it); + putname(name); + if (error) + goto out0; +@@ -957,7 +970,7 @@ + goto out1; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd)) +- error = path_walk(name, &old_nd); ++ error = path_walk_it(name, &old_nd, &old_it); + putname(name); + if (error) + goto out1; +@@ -1013,8 +1026,10 @@ + up(&old_nd.dentry->d_inode->i_zombie); + up_write(¤t->namespace->sem); + path_release(&user_nd); ++ intent_release(&old_it); + path_release(&old_nd); + out1: ++ intent_release(&new_it); + path_release(&new_nd); + out0: + unlock_kernel(); +Index: linux-2.4.19.SuSE/fs/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/namei.c Mon Jan 27 05:08:07 2003 ++++ linux-2.4.19.SuSE/fs/namei.c Sat Nov 15 17:52:03 2003 +@@ -94,6 +94,13 @@ + * XEmacs seems to be relying on it... + */ + ++void intent_release(struct lookup_intent *it) ++{ ++ if (it && it->it_op_release) ++ it->it_op_release(it); ++ ++} ++ + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the + * kernel data space before using them.. +@@ -260,10 +267,19 @@ + * Internal lookup() using the new generic dcache. + * SMP-safe + */ +-static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *cached_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * dentry = d_lookup(parent, name); + ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, flags, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ dentry = NULL; ++ } ++ return dentry; ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + if (!dentry->d_op->d_revalidate(dentry, flags) && !d_invalidate(dentry)) { + dput(dentry); +@@ -281,11 +297,15 @@ + * make sure that nobody added the entry to the dcache in the meantime.. + * SMP-safe + */ +-static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *real_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * result; + struct inode *dir = parent->d_inode; ++ int counter = 0; + ++again: ++ counter++; + down(&dir->i_sem); + /* + * First re-do the cached lookup just in case it was created +@@ -300,6 +320,9 @@ + result = ERR_PTR(-ENOMEM); + if (dentry) { + lock_kernel(); ++ if (dir->i_op->lookup_it) ++ result = dir->i_op->lookup_it(dir, dentry, it, flags); ++ else + result = dir->i_op->lookup(dir, dentry); + unlock_kernel(); + if (result) +@@ -321,6 +344,15 @@ + dput(result); + result = ERR_PTR(-ENOENT); + } ++ } else if (result->d_op && result->d_op->d_revalidate_it) { ++ if (!result->d_op->d_revalidate_it(result, flags, it) && ++ !d_invalidate(result)) { ++ dput(result); ++ if (counter > 10) ++ result = ERR_PTR(-ESTALE); ++ if (!IS_ERR(result)) ++ goto again; ++ } + } + return result; + } +@@ -332,7 +364,8 @@ + * Without that kind of total limit, nasty chains of consecutive + * symlinks can cause almost arbitrarily long lookups. + */ +-static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd) ++static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd, ++ struct lookup_intent *it) + { + int err; + if (current->link_count >= 8) +@@ -346,10 +379,12 @@ + current->link_count++; + current->total_link_count++; + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + err = dentry->d_inode->i_op->follow_link(dentry, nd); + current->link_count--; + return err; + loop: ++ intent_release(it); + path_release(nd); + return -ELOOP; + } +@@ -447,7 +482,8 @@ + * + * We expect 'base' to be positive and a directory. + */ +-int link_path_walk(const char * name, struct nameidata *nd) ++int link_path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it) + { + struct dentry *dentry; + struct inode *inode; +@@ -524,12 +560,13 @@ + break; + } + /* This does the actual lookups.. */ +- dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE,NULL); + if (!dentry) { + err = -EWOULDBLOCKIO; + if (atomic) + break; +- dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE, ++ NULL); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -547,7 +584,7 @@ + goto out_dput; + + if (inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, NULL); + dput(dentry); + if (err) + goto return_err; +@@ -563,7 +600,7 @@ + nd->dentry = dentry; + } + err = -ENOTDIR; +- if (!inode->i_op->lookup) ++ if (!inode->i_op->lookup && !inode->i_op->lookup_it) + break; + continue; + /* here ends the main loop */ +@@ -590,12 +627,12 @@ + if (err < 0) + break; + } +- dentry = cached_lookup(nd->dentry, &this, 0); ++ dentry = cached_lookup(nd->dentry, &this, 0, it); + if (!dentry) { + err = -EWOULDBLOCKIO; + if (atomic) + break; +- dentry = real_lookup(nd->dentry, &this, 0); ++ dentry = real_lookup(nd->dentry, &this, 0, it); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -605,7 +642,7 @@ + inode = dentry->d_inode; + if ((lookup_flags & LOOKUP_FOLLOW) + && inode && inode->i_op && inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, it); + dput(dentry); + if (err) + goto return_err; +@@ -619,7 +656,8 @@ + goto no_inode; + if (lookup_flags & LOOKUP_DIRECTORY) { + err = -ENOTDIR; +- if (!inode->i_op || !inode->i_op->lookup) ++ if (!inode->i_op || ++ (!inode->i_op->lookup && !inode->i_op->lookup_it)) + break; + } + goto return_base; +@@ -643,6 +681,25 @@ + * Check the cached dentry for staleness. + */ + dentry = nd->dentry; ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ err = -ESTALE; ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) { ++ struct dentry *new; ++ err = permission(dentry->d_parent->d_inode, ++ MAY_EXEC); ++ if (err) ++ break; ++ new = real_lookup(dentry->d_parent, ++ &dentry->d_name, 0, NULL); ++ d_invalidate(dentry); ++ dput(dentry); ++ if (IS_ERR(new)) { ++ err = PTR_ERR(new); ++ break; ++ } ++ nd->dentry = new; ++ } ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + err = -ESTALE; + if (!dentry->d_op->d_revalidate(dentry, lookup_flags & LOOKUP_PARENT)) { +@@ -656,15 +713,28 @@ + dput(dentry); + break; + } ++ if (err) ++ intent_release(it); + path_release(nd); + return_err: + return err; + } + ++int link_path_walk(const char * name, struct nameidata *nd) ++{ ++ return link_path_walk_it(name, nd, NULL); ++} ++ ++int path_walk_it(const char * name, struct nameidata *nd, struct lookup_intent *it) ++{ ++ current->total_link_count = 0; ++ return link_path_walk_it(name, nd, it); ++} ++ + int path_walk(const char * name, struct nameidata *nd) + { + current->total_link_count = 0; +- return link_path_walk(name, nd); ++ return link_path_walk_it(name, nd, NULL); + } + + /* SMP-safe */ +@@ -753,6 +823,7 @@ + { + nd->last_type = LAST_ROOT; /* if there are only slashes... */ + nd->flags = flags; ++ nd->intent = NULL; + if (*name=='/') + return walk_init_root(name,nd); + read_lock(¤t->fs->lock); +@@ -767,7 +838,8 @@ + * needs parent already locked. Doesn't follow mounts. + * SMP-safe. + */ +-struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++struct dentry * lookup_hash_it(struct qstr *name, struct dentry * base, ++ struct lookup_intent *it) + { + struct dentry * dentry; + struct inode *inode; +@@ -790,13 +862,16 @@ + goto out; + } + +- dentry = cached_lookup(base, name, 0); ++ dentry = cached_lookup(base, name, 0, it); + if (!dentry) { + struct dentry *new = d_alloc(base, name); + dentry = ERR_PTR(-ENOMEM); + if (!new) + goto out; + lock_kernel(); ++ if (inode->i_op->lookup_it) ++ dentry = inode->i_op->lookup_it(inode, new, it, 0); ++ else + dentry = inode->i_op->lookup(inode, new); + unlock_kernel(); + if (!dentry) +@@ -808,6 +883,12 @@ + return dentry; + } + ++struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++{ ++ return lookup_hash_it(name, base, NULL); ++} ++ ++ + /* SMP-safe */ + struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) + { +@@ -829,7 +910,7 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash(&this, base); ++ return lookup_hash_it(&this, base, NULL); + access: + return ERR_PTR(-EACCES); + } +@@ -861,6 +942,23 @@ + return err; + } + ++int __user_walk_it(const char *name, unsigned flags, struct nameidata *nd, ++ struct lookup_intent *it) ++{ ++ char *tmp; ++ int err; ++ ++ tmp = getname(name); ++ err = PTR_ERR(tmp); ++ if (!IS_ERR(tmp)) { ++ err = 0; ++ if (path_init(tmp, flags, nd)) ++ err = path_walk_it(tmp, nd, it); ++ putname(tmp); ++ } ++ return err; ++} ++ + /* + * It's inline, so penalty for filesystems that don't use sticky bit is + * minimal. +@@ -958,7 +1056,8 @@ + return retval; + } + +-int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++static int vfs_create_it(struct inode *dir, struct dentry *dentry, int mode, ++ struct lookup_intent *it) + { + int error; + +@@ -971,12 +1070,15 @@ + goto exit_lock; + + error = -EACCES; /* shouldn't it be ENOSYS? */ +- if (!dir->i_op || !dir->i_op->create) ++ if (!dir->i_op || (!dir->i_op->create && !dir->i_op->create_it)) + goto exit_lock; + + DQUOT_INIT(dir); + lock_kernel(); +- error = dir->i_op->create(dir, dentry, mode); ++ if (dir->i_op->create_it) ++ error = dir->i_op->create_it(dir, dentry, mode, it); ++ else ++ error = dir->i_op->create(dir, dentry, mode); + unlock_kernel(); + exit_lock: + up(&dir->i_zombie); +@@ -985,6 +1087,11 @@ + return error; + } + ++int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++{ ++ return vfs_create_it(dir, dentry, mode, NULL); ++} ++ + /* + * open_namei() + * +@@ -999,7 +1106,8 @@ + * for symlinks (where the permissions are checked later). + * SMP-safe + */ +-int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) ++int open_namei_it(const char *pathname, int flag, int mode, ++ struct nameidata *nd, struct lookup_intent *it) + { + int acc_mode, error = 0; + struct inode *inode; +@@ -1009,12 +1117,14 @@ + + acc_mode = ACC_MODE(flag); + ++ if (it) ++ it->it_flags = flag; + /* + * The simplest case - just a plain lookup. + */ + if (!(flag & O_CREAT)) { + if (path_init(pathname, lookup_flags(flag), nd)) +- error = path_walk(pathname, nd); ++ error = path_walk_it(pathname, nd, it); + if (error) + return error; + dentry = nd->dentry; +@@ -1024,6 +1134,10 @@ + /* + * Create - we need to know the parent. + */ ++ if (it) { ++ it->it_create_mode = mode; ++ it->it_op |= IT_CREAT; ++ } + if (path_init(pathname, LOOKUP_PARENT, nd)) + error = path_walk(pathname, nd); + if (error) +@@ -1040,7 +1154,7 @@ + + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + + do_last: + error = PTR_ERR(dentry); +@@ -1049,11 +1163,13 @@ + goto exit; + } + ++ it->it_create_mode = mode; + /* Negative dentry, just create the file */ + if (!dentry->d_inode) { + if (!IS_POSIXACL(dir->d_inode)) + mode &= ~current->fs->umask; +- error = vfs_create(dir->d_inode, dentry, mode); ++ error = vfs_create_it(dir->d_inode, dentry, ++ mode & ~current->fs->umask, it); + up(&dir->d_inode->i_sem); + #ifndef DENTRY_WASTE_RAM + if (error) +@@ -1161,7 +1277,7 @@ + if (!error) { + DQUOT_INIT(inode); + +- error = do_truncate(dentry, 0); ++ error = do_truncate(dentry, 0, 1); + } + put_write_access(inode); + if (error) +@@ -1173,8 +1289,10 @@ + return 0; + + exit_dput: ++ intent_release(it); + dput(dentry); + exit: ++ intent_release(it); + path_release(nd); + return error; + +@@ -1193,7 +1311,10 @@ + * are done. Procfs-like symlinks just set LAST_BIND. + */ + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + error = dentry->d_inode->i_op->follow_link(dentry, nd); ++ if (error) ++ intent_release(it); + dput(dentry); + if (error) + return error; +@@ -1215,13 +1336,20 @@ + } + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + putname(nd->last.name); + goto do_last; + } + ++int open_namei(const char *pathname, int flag, int mode, struct nameidata *nd) ++{ ++ return open_namei_it(pathname, flag, mode, nd, NULL); ++} ++ ++ + /* SMP-safe */ +-static struct dentry *lookup_create(struct nameidata *nd, int is_dir) ++static struct dentry *lookup_create(struct nameidata *nd, int is_dir, ++ struct lookup_intent *it) + { + struct dentry *dentry; + +@@ -1229,7 +1357,7 @@ + dentry = ERR_PTR(-EEXIST); + if (nd->last_type != LAST_NORM) + goto fail; +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + if (IS_ERR(dentry)) + goto fail; + if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) +@@ -1286,7 +1414,16 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ ++ if (nd.dentry->d_inode->i_op->mknod_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mknod_raw(&nd, mode, dev); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + + if (!IS_POSIXACL(nd.dentry->d_inode)) +@@ -1308,6 +1445,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1356,7 +1494,14 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 1); ++ if (nd.dentry->d_inode->i_op->mkdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mkdir_raw(&nd, mode); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 1, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + if (!IS_POSIXACL(nd.dentry->d_inode)) +@@ -1365,6 +1510,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1466,8 +1612,16 @@ + error = -EBUSY; + goto exit1; + } ++ if (nd.dentry->d_inode->i_op->rmdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ error = op->rmdir_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_rmdir(nd.dentry->d_inode, dentry); +@@ -1526,8 +1680,15 @@ + error = -EISDIR; + if (nd.last_type != LAST_NORM) + goto exit1; ++ if (nd.dentry->d_inode->i_op->unlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->unlink_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + /* Why not before? Because we want correct error value */ +@@ -1595,15 +1756,23 @@ + error = path_walk(to, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->symlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->symlink_raw(&nd, from); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_symlink(nd.dentry->d_inode, dentry, from); + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++ out2: + path_release(&nd); +-out: ++ out: + putname(to); + } + putname(from); +@@ -1686,7 +1855,14 @@ + error = -EXDEV; + if (old_nd.mnt != nd.mnt) + goto out_release; +- new_dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->link_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->link_raw(&old_nd, &nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out_release; ++ } ++ new_dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(new_dentry); + if (!IS_ERR(new_dentry)) { + error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry); +@@ -1732,7 +1908,7 @@ + * locking]. + */ + int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + struct inode *target; +@@ -1811,7 +1987,7 @@ + } + + int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + +@@ -1902,9 +2078,18 @@ + if (newnd.last_type != LAST_NORM) + goto exit2; + ++ if (old_dir->d_inode->i_op->rename_raw) { ++ lock_kernel(); ++ error = old_dir->d_inode->i_op->rename_raw(&oldnd, &newnd); ++ unlock_kernel(); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit2; ++ } ++ + double_lock(new_dir, old_dir); + +- old_dentry = lookup_hash(&oldnd.last, old_dir); ++ old_dentry = lookup_hash_it(&oldnd.last, old_dir, NULL); + error = PTR_ERR(old_dentry); + if (IS_ERR(old_dentry)) + goto exit3; +@@ -1920,16 +2105,16 @@ + if (newnd.last.name[newnd.last.len]) + goto exit4; + } +- new_dentry = lookup_hash(&newnd.last, new_dir); ++ new_dentry = lookup_hash_it(&newnd.last, new_dir, NULL); + error = PTR_ERR(new_dentry); + if (IS_ERR(new_dentry)) + goto exit4; + ++ + lock_kernel(); + error = vfs_rename(old_dir->d_inode, old_dentry, + new_dir->d_inode, new_dentry); + unlock_kernel(); +- + dput(new_dentry); + exit4: + dput(old_dentry); +@@ -1980,20 +2165,26 @@ + } + + static inline int +-__vfs_follow_link(struct nameidata *nd, const char *link) ++__vfs_follow_link(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) + { + int res = 0; + char *name; + if (IS_ERR(link)) + goto fail; + ++ if (it == NULL) ++ it = nd->intent; ++ else if (it != nd->intent) ++ printk("it != nd->intent: tell phil@clusterfs.com\n"); ++ + if (*link == '/') { + path_release(nd); + if (!walk_init_root(link, nd)) + /* weird __emul_prefix() stuff did it */ + goto out; + } +- res = link_path_walk(link, nd); ++ res = link_path_walk_it(link, nd, it); + out: + if (current->link_count || res || nd->last_type!=LAST_NORM) + return res; +@@ -2015,7 +2206,13 @@ + + int vfs_follow_link(struct nameidata *nd, const char *link) + { +- return __vfs_follow_link(nd, link); ++ return __vfs_follow_link(nd, link, NULL); ++} ++ ++int vfs_follow_link_it(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) ++{ ++ return __vfs_follow_link(nd, link, it); + } + + /* get the link contents into pagecache */ +@@ -2057,7 +2254,7 @@ + { + struct page *page = NULL; + char *s = page_getlink(dentry, &page); +- int res = __vfs_follow_link(nd, s); ++ int res = __vfs_follow_link(nd, s, NULL); + if (page) { + kunmap(page); + page_cache_release(page); +Index: linux-2.4.19.SuSE/fs/open.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/open.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/open.c Sat Nov 15 17:43:27 2003 +@@ -19,6 +19,8 @@ + #include + + #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) ++extern int path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it); + + int vfs_statfs(struct super_block *sb, struct statfs *buf) + { +@@ -95,9 +97,10 @@ + write_unlock(&files->file_lock); + } + +-int do_truncate(struct dentry *dentry, loff_t length) ++int do_truncate(struct dentry *dentry, loff_t length, int called_from_open) + { + struct inode *inode = dentry->d_inode; ++ struct inode_operations *op = dentry->d_inode->i_op; + int error; + struct iattr newattrs; + +@@ -108,7 +111,13 @@ + down(&inode->i_sem); + newattrs.ia_size = length; + newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME; +- error = notify_change(dentry, &newattrs); ++ if (called_from_open) ++ newattrs.ia_valid |= ATTR_FROM_OPEN; ++ if (op->setattr_raw) { ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ } else ++ error = notify_change(dentry, &newattrs); + up(&inode->i_sem); + return error; + } +@@ -118,12 +127,13 @@ + struct nameidata nd; + struct inode * inode; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + error = -EINVAL; + if (length < 0) /* sorry, but loff_t says... */ + goto out; + +- error = user_path_walk(path, &nd); ++ error = user_path_walk_it(path, &nd, &it); + if (error) + goto out; + inode = nd.dentry->d_inode; +@@ -163,11 +173,13 @@ + error = locks_verify_truncate(inode, NULL, length); + if (!error) { + DQUOT_INIT(inode); +- error = do_truncate(nd.dentry, length); ++ intent_release(&it); ++ error = do_truncate(nd.dentry, length, 0); + } + put_write_access(inode); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -215,7 +227,7 @@ + + error = locks_verify_truncate(inode, file, length); + if (!error) +- error = do_truncate(dentry, length); ++ error = do_truncate(dentry, length, 0); + out_putf: + fput(file); + out: +@@ -260,11 +272,13 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -279,11 +293,25 @@ + goto dput_and_out; + + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!times) { + if (current->fsuid != inode->i_uid && + (error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; + } ++ + error = notify_change(nd.dentry, &newattrs); + dput_and_out: + path_release(&nd); +@@ -304,12 +332,14 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -324,7 +354,20 @@ + newattrs.ia_atime = times[0].tv_sec; + newattrs.ia_mtime = times[1].tv_sec; + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!utimes) { + if (current->fsuid != inode->i_uid && + (error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; +@@ -347,6 +390,7 @@ + int old_fsuid, old_fsgid; + kernel_cap_t old_cap; + int res; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ + return -EINVAL; +@@ -364,13 +408,14 @@ + else + current->cap_effective = current->cap_permitted; + +- res = user_path_walk(filename, &nd); ++ res = user_path_walk_it(filename, &nd, &it); + if (!res) { + res = permission(nd.dentry->d_inode, mode); + /* SuS v2 requires we report a read only fs too */ + if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode) + && !special_file(nd.dentry->d_inode->i_mode)) + res = -EROFS; ++ intent_release(&it); + path_release(&nd); + } + +@@ -386,6 +431,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -394,7 +440,7 @@ + + error = 0; + if (path_init(name,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd)) +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -406,6 +452,7 @@ + set_fs_pwd(current->fs, nd.mnt, nd.dentry); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -446,6 +493,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -454,7 +502,7 @@ + + path_init(name, LOOKUP_POSITIVE | LOOKUP_FOLLOW | + LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd); +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -471,39 +519,56 @@ + set_fs_altroot(); + error = 0; + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; + } + +-asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++int chmod_common(struct dentry *dentry, mode_t mode) + { +- struct inode * inode; +- struct dentry * dentry; +- struct file * file; +- int err = -EBADF; ++ struct inode *inode = dentry->d_inode; + struct iattr newattrs; ++ int err = -EROFS; + +- file = fget(fd); +- if (!file) ++ if (IS_RDONLY(inode)) + goto out; + +- dentry = file->f_dentry; +- inode = dentry->d_inode; ++ if (inode->i_op->setattr_raw) { ++ newattrs.ia_mode = mode; ++ newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ err = inode->i_op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (err != -EOPNOTSUPP) ++ goto out; ++ } + +- err = -EROFS; +- if (IS_RDONLY(inode)) +- goto out_putf; + err = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto out_putf; ++ goto out; ++ + if (mode == (mode_t) -1) + mode = inode->i_mode; + newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); + newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; + err = notify_change(dentry, &newattrs); + +-out_putf: ++out: ++ return err; ++} ++ ++asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++{ ++ struct file * file; ++ int err = -EBADF; ++ ++ file = fget(fd); ++ if (!file) ++ goto out; ++ ++ err = chmod_common(file->f_dentry, mode); ++ + fput(file); + out: + return err; +@@ -512,30 +577,14 @@ + asmlinkage long sys_chmod(const char * filename, mode_t mode) + { + struct nameidata nd; +- struct inode * inode; + int error; +- struct iattr newattrs; + + error = user_path_walk(filename, &nd); + if (error) + goto out; +- inode = nd.dentry->d_inode; +- +- error = -EROFS; +- if (IS_RDONLY(inode)) +- goto dput_and_out; + +- error = -EPERM; +- if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto dput_and_out; ++ error = chmod_common(nd.dentry, mode); + +- if (mode == (mode_t) -1) +- mode = inode->i_mode; +- newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); +- newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; +- error = notify_change(nd.dentry, &newattrs); +- +-dput_and_out: + path_release(&nd); + out: + return error; +@@ -555,6 +604,20 @@ + error = -EROFS; + if (IS_RDONLY(inode)) + goto out; ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = dentry->d_inode->i_op; ++ ++ newattrs.ia_uid = user; ++ newattrs.ia_gid = group; ++ newattrs.ia_valid = ATTR_UID | ATTR_GID | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ return error; ++ } ++ + error = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) + goto out; +@@ -659,6 +722,7 @@ + { + int namei_flags, error; + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_OPEN }; + + namei_flags = flags; + if ((namei_flags+1) & O_ACCMODE) +@@ -666,14 +730,15 @@ + if (namei_flags & O_TRUNC) + namei_flags |= 2; + +- error = open_namei(filename, namei_flags, mode, &nd); +- if (!error) +- return dentry_open(nd.dentry, nd.mnt, flags); ++ error = open_namei_it(filename, namei_flags, mode, &nd, &it); ++ if (error) ++ return ERR_PTR(error); + +- return ERR_PTR(error); ++ return dentry_open_it(nd.dentry, nd.mnt, flags, &it); + } + +-struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it) + { + struct file * f; + struct inode *inode; +@@ -710,7 +775,9 @@ + } + + if (f->f_op && f->f_op->open) { ++ f->f_it = it; + error = f->f_op->open(inode,f); ++ f->f_it = NULL; + if (error) + goto cleanup_all; + } +@@ -722,6 +789,7 @@ + !inode->i_mapping->a_ops->direct_IO)) + goto cleanup_all; + ++ intent_release(it); + return f; + + cleanup_all: +@@ -736,11 +804,17 @@ + cleanup_file: + put_filp(f); + cleanup_dentry: ++ intent_release(it); + dput(dentry); + mntput(mnt); + return ERR_PTR(error); + } + ++struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++{ ++ return dentry_open_it(dentry, mnt, flags, NULL); ++} ++ + /* + * Find an empty file descriptor entry, and mark it busy. + */ +Index: linux-2.4.19.SuSE/fs/stat.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/stat.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/stat.c Sat Nov 15 17:29:03 2003 +@@ -17,10 +17,14 @@ + * Revalidate the inode. This is required for proper NFS attribute caching. + */ + static __inline__ int +-do_revalidate(struct dentry *dentry) ++do_revalidate(struct dentry *dentry, struct lookup_intent *it) + { + struct inode * inode = dentry->d_inode; +- if (inode->i_op && inode->i_op->revalidate) ++ if (!inode) ++ return -ENOENT; ++ if (inode->i_op && inode->i_op->revalidate_it) ++ return inode->i_op->revalidate_it(dentry, it); ++ else if (inode->i_op && inode->i_op->revalidate) + return inode->i_op->revalidate(dentry); + return 0; + } +@@ -141,13 +145,15 @@ + asmlinkage long sys_stat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -157,13 +163,15 @@ + asmlinkage long sys_newstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -178,13 +186,15 @@ + asmlinkage long sys_lstat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -195,13 +205,15 @@ + asmlinkage long sys_newlstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -222,7 +234,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_old_stat(dentry->d_inode, statbuf); + fput(f); +@@ -241,7 +253,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat(dentry->d_inode, statbuf); + fput(f); +@@ -263,7 +275,7 @@ + + error = -EINVAL; + if (inode->i_op && inode->i_op->readlink && +- !(error = do_revalidate(nd.dentry))) { ++ !(error = do_revalidate(nd.dentry, NULL))) { + UPDATE_ATIME(inode); + error = inode->i_op->readlink(nd.dentry, buf, bufsiz); + } +@@ -339,12 +351,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -354,12 +368,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -374,7 +390,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat64(dentry->d_inode, statbuf); + fput(f); +Index: linux-2.4.19.SuSE/include/linux/dcache.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/dcache.h Mon Jan 27 05:13:15 2003 ++++ linux-2.4.19.SuSE/include/linux/dcache.h Sat Nov 15 17:35:46 2003 +@@ -5,6 +5,51 @@ + + #include + #include ++#include ++ ++#define IT_OPEN 0x0001 ++#define IT_CREAT 0x0002 ++#define IT_READDIR 0x0004 ++#define IT_GETATTR 0x0008 ++#define IT_LOOKUP 0x0010 ++#define IT_UNLINK 0x0020 ++#define IT_GETXATTR 0x0040 ++#define IT_EXEC 0x0080 ++#define IT_PIN 0x0100 ++ ++#define IT_FL_LOCKED 0x0001 ++#define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */ ++ ++#define INTENT_MAGIC 0x19620323 ++ ++ ++struct lustre_intent_data { ++ int it_disposition; ++ int it_status; ++ __u64 it_lock_handle; ++ void *it_data; ++ int it_lock_mode; ++ int it_int_flags; ++}; ++struct lookup_intent { ++ int it_magic; ++ void (*it_op_release)(struct lookup_intent *); ++ int it_op; ++ int it_flags; ++ int it_create_mode; ++ union { ++ struct lustre_intent_data lustre; ++ } d; ++}; ++ ++static inline void intent_init(struct lookup_intent *it, int op, int flags) ++{ ++ memset(it, 0, sizeof(*it)); ++ it->it_magic = INTENT_MAGIC; ++ it->it_op = op; ++ it->it_flags = flags; ++} ++ + + /* + * linux/include/linux/dcache.h +@@ -92,8 +137,22 @@ + int (*d_delete)(struct dentry *); + void (*d_release)(struct dentry *); + void (*d_iput)(struct dentry *, struct inode *); ++ int (*d_revalidate_it)(struct dentry *, int, struct lookup_intent *); ++ void (*d_pin)(struct dentry *, struct vfsmount * , int); ++ void (*d_unpin)(struct dentry *, struct vfsmount *, int); + }; + ++#define PIN(de,mnt,flag) if (de->d_op && de->d_op->d_pin) \ ++ de->d_op->d_pin(de, mnt, flag); ++#define UNPIN(de,mnt,flag) if (de->d_op && de->d_op->d_unpin) \ ++ de->d_op->d_unpin(de, mnt, flag); ++ ++ ++/* defined in fs/namei.c */ ++extern void intent_release(struct lookup_intent *it); ++/* defined in fs/dcache.c */ ++extern void __d_rehash(struct dentry * entry, int lock); ++ + /* the dentry parameter passed to d_hash and d_compare is the parent + * directory of the entries to be compared. It is used in case these + * functions need any directory specific information for determining +@@ -125,6 +184,7 @@ + * s_nfsd_free_path semaphore will be down + */ + #define DCACHE_REFERENCED 0x0008 /* Recently used, don't discard. */ ++#define DCACHE_LUSTRE_INVALID 0x0010 /* Lustre invalidated */ + + extern spinlock_t dcache_lock; + +Index: linux-2.4.19.SuSE/include/linux/fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs.h Sat Nov 15 17:25:06 2003 ++++ linux-2.4.19.SuSE/include/linux/fs.h Sat Nov 15 17:29:03 2003 +@@ -73,6 +73,7 @@ + + #define FMODE_READ 1 + #define FMODE_WRITE 2 ++#define FMODE_EXEC 4 + + #define READ 0 + #define WRITE 1 +@@ -363,6 +364,9 @@ + #define ATTR_MTIME_SET 256 + #define ATTR_FORCE 512 /* Not a change, but a change it */ + #define ATTR_ATTR_FLAG 1024 ++#define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */ ++#define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */ ++#define ATTR_CTIME_SET 0x2000 + + /* + * This is the Inode Attributes structure, used for notify_change(). It +@@ -507,6 +511,7 @@ + struct pipe_inode_info *i_pipe; + struct block_device *i_bdev; + struct char_device *i_cdev; ++ void *i_filterdata; + + unsigned long i_dnotify_mask; /* Directory notify events */ + struct dnotify_struct *i_dnotify; /* for directory notifications */ +@@ -669,6 +674,7 @@ + + /* needed for tty driver, and maybe others */ + void *private_data; ++ struct lookup_intent *f_it; + + /* preallocated helper kiobuf to speedup O_DIRECT */ + struct kiobuf *f_iobuf; +@@ -799,6 +805,7 @@ + struct qstr last; + unsigned int flags; + int last_type; ++ struct lookup_intent *intent; + }; + + #define DQUOT_USR_ENABLED 0x01 /* User diskquotas enabled */ +@@ -947,7 +954,8 @@ + extern int __vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_unlink(struct inode *, struct dentry *); +-extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); ++int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, ++ struct inode *new_dir, struct dentry *new_dentry); + + /* + * File types +@@ -1020,21 +1028,32 @@ + + struct inode_operations { + int (*create) (struct inode *,struct dentry *,int); ++ int (*create_it) (struct inode *,struct dentry *,int, struct lookup_intent *); + struct dentry * (*lookup) (struct inode *,struct dentry *); ++ struct dentry * (*lookup_it) (struct inode *,struct dentry *, struct lookup_intent *, int flags); + int (*link) (struct dentry *,struct inode *,struct dentry *); ++ int (*link_raw) (struct nameidata *,struct nameidata *); + int (*unlink) (struct inode *,struct dentry *); ++ int (*unlink_raw) (struct nameidata *); + int (*symlink) (struct inode *,struct dentry *,const char *); ++ int (*symlink_raw) (struct nameidata *,const char *); + int (*mkdir) (struct inode *,struct dentry *,int); ++ int (*mkdir_raw) (struct nameidata *,int); + int (*rmdir) (struct inode *,struct dentry *); ++ int (*rmdir_raw) (struct nameidata *); + int (*mknod) (struct inode *,struct dentry *,int,int); ++ int (*mknod_raw) (struct nameidata *,int,dev_t); + int (*rename) (struct inode *, struct dentry *, + struct inode *, struct dentry *); ++ int (*rename_raw) (struct nameidata *, struct nameidata *); + int (*readlink) (struct dentry *, char *,int); + int (*follow_link) (struct dentry *, struct nameidata *); + void (*truncate) (struct inode *); + int (*permission) (struct inode *, int); + int (*revalidate) (struct dentry *); ++ int (*revalidate_it) (struct dentry *, struct lookup_intent *); + int (*setattr) (struct dentry *, struct iattr *); ++ int (*setattr_raw) (struct inode *, struct iattr *); + int (*getattr) (struct dentry *, struct iattr *); + int (*setxattr) (struct dentry *, const char *, const void *, size_t, int); + ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); +@@ -1244,10 +1263,14 @@ + + asmlinkage long sys_open(const char *, int, int); + asmlinkage long sys_close(unsigned int); /* yes, it's really unsigned */ +-extern int do_truncate(struct dentry *, loff_t start); ++extern int do_truncate(struct dentry *, loff_t start, int called_from_open); + + extern struct file *filp_open(const char *, int, int); + extern struct file * dentry_open(struct dentry *, struct vfsmount *, int); ++extern int open_namei_it(const char *filename, int namei_flags, int mode, ++ struct nameidata *nd, struct lookup_intent *it); ++extern struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it); + extern int filp_close(struct file *, fl_owner_t id); + extern char * getname(const char *); + +@@ -1515,6 +1538,7 @@ + extern loff_t default_llseek(struct file *file, loff_t offset, int origin); + + extern int FASTCALL(__user_walk(const char *, unsigned, struct nameidata *)); ++extern int FASTCALL(__user_walk_it(const char *, unsigned, struct nameidata *, struct lookup_intent *it)); + extern int FASTCALL(path_init(const char *, unsigned, struct nameidata *)); + extern int FASTCALL(path_walk(const char *, struct nameidata *)); + extern int FASTCALL(link_path_walk(const char *, struct nameidata *)); +@@ -1526,6 +1550,8 @@ + extern struct dentry * lookup_hash(struct qstr *, struct dentry *); + #define user_path_walk(name,nd) __user_walk(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd) + #define user_path_walk_link(name,nd) __user_walk(name, LOOKUP_POSITIVE, nd) ++#define user_path_walk_it(name,nd,it) __user_walk_it(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd, it) ++#define user_path_walk_link_it(name,nd,it) __user_walk_it(name, LOOKUP_POSITIVE, nd, it) + + extern void iput(struct inode *); + extern void force_delete(struct inode *); +@@ -1646,6 +1672,8 @@ + + extern int vfs_readlink(struct dentry *, char *, int, const char *); + extern int vfs_follow_link(struct nameidata *, const char *); ++extern int vfs_follow_link_it(struct nameidata *, const char *, ++ struct lookup_intent *it); + extern int page_readlink(struct dentry *, char *, int); + extern int page_follow_link(struct dentry *, struct nameidata *); + extern struct inode_operations page_symlink_inode_operations; +Index: linux-2.4.19.SuSE/include/linux/fs_struct.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs_struct.h Fri Jul 13 15:10:44 2001 ++++ linux-2.4.19.SuSE/include/linux/fs_struct.h Sat Nov 15 17:29:03 2003 +@@ -34,10 +34,12 @@ + write_lock(&fs->lock); + old_root = fs->root; + old_rootmnt = fs->rootmnt; ++ PIN(dentry, mnt, 1); + fs->rootmnt = mntget(mnt); + fs->root = dget(dentry); + write_unlock(&fs->lock); + if (old_root) { ++ UNPIN(old_root, old_rootmnt, 1); + dput(old_root); + mntput(old_rootmnt); + } +@@ -57,10 +59,12 @@ + write_lock(&fs->lock); + old_pwd = fs->pwd; + old_pwdmnt = fs->pwdmnt; ++ PIN(dentry, mnt, 0); + fs->pwdmnt = mntget(mnt); + fs->pwd = dget(dentry); + write_unlock(&fs->lock); + if (old_pwd) { ++ UNPIN(old_pwd, old_pwdmnt, 0); + dput(old_pwd); + mntput(old_pwdmnt); + } +Index: linux-2.4.19.SuSE/kernel/ksyms.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/ksyms.c Sat Nov 15 17:24:46 2003 ++++ linux-2.4.19.SuSE/kernel/ksyms.c Sat Nov 15 17:29:03 2003 +@@ -315,6 +315,7 @@ + EXPORT_SYMBOL(set_page_dirty); + EXPORT_SYMBOL(vfs_readlink); + EXPORT_SYMBOL(vfs_follow_link); ++EXPORT_SYMBOL(vfs_follow_link_it); + EXPORT_SYMBOL(page_readlink); + EXPORT_SYMBOL(page_follow_link); + EXPORT_SYMBOL(page_symlink_inode_operations); +Index: linux-2.4.19.SuSE/kernel/fork.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/fork.c Mon Jan 27 05:08:56 2003 ++++ linux-2.4.19.SuSE/kernel/fork.c Sat Nov 15 17:29:03 2003 +@@ -454,10 +454,13 @@ + fs->umask = old->umask; + read_lock(&old->lock); + fs->rootmnt = mntget(old->rootmnt); ++ PIN(old->pwd, old->pwdmnt, 0); ++ PIN(old->root, old->rootmnt, 1); + fs->root = dget(old->root); + fs->pwdmnt = mntget(old->pwdmnt); + fs->pwd = dget(old->pwd); + if (old->altroot) { ++ PIN(old->altroot, old->altrootmnt, 1); + fs->altrootmnt = mntget(old->altrootmnt); + fs->altroot = dget(old->altroot); + } else { +Index: linux-2.4.19.SuSE/kernel/exit.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/exit.c Mon Jan 27 05:08:16 2003 ++++ linux-2.4.19.SuSE/kernel/exit.c Sat Nov 15 17:29:03 2003 +@@ -288,11 +288,14 @@ + { + /* No need to hold fs->lock if we are killing it */ + if (atomic_dec_and_test(&fs->count)) { ++ UNPIN(fs->pwd, fs->pwdmnt, 0); ++ UNPIN(fs->root, fs->rootmnt, 1); + dput(fs->root); + mntput(fs->rootmnt); + dput(fs->pwd); + mntput(fs->pwdmnt); + if (fs->altroot) { ++ UNPIN(fs->altroot, fs->altrootmnt, 1); + dput(fs->altroot); + mntput(fs->altrootmnt); + } diff --git a/lustre/kernel_patches/series/rh-2.4.22 b/lustre/kernel_patches/series/rh-2.4.22 index d1b13dd..a3bd2b9 100644 --- a/lustre/kernel_patches/series/rh-2.4.22 +++ b/lustre/kernel_patches/series/rh-2.4.22 @@ -1,6 +1,5 @@ dev_read_only_2.4.20-rh.patch exports_2.4.20-rh-hp.patch -kmem_cache_validate_2.4.20.patch lustre_version.patch vfs_intent-2.4.22-rh.patch invalidate_show-2.4.20-rh.patch @@ -24,4 +23,6 @@ socket-exports-2.4.22-rh.patch nfs_export_kernel-2.4.22-rh.patch ext3-ea-in-inode-2.4.22-rh.patch listman-2.4.20.patch -ext3-trusted_ea-2.4.20.patch +ext3-trusted_ea-2.4.20.patch +gfp_memalloc-2.4.22.patch +ext3-xattr-ptr-arith-fix.patch diff --git a/lustre/kernel_patches/series/suse-2.4.19 b/lustre/kernel_patches/series/suse-2.4.19 new file mode 100644 index 0000000..4365c39 --- /dev/null +++ b/lustre/kernel_patches/series/suse-2.4.19 @@ -0,0 +1,35 @@ +dev_read_only_hp_2.4.20.patch +exports_2.4.19-suse.patch +lustre_version.patch +vfs_intent-2.4.19-suse.patch +invalidate_show.patch +export-truncate.patch +iod-stock-24-exports-2.4.19-suse.patch +jbd-2.4.18-jcberr.patch +ext-2.4-patch-1-chaos.patch +ext-2.4-patch-2.patch +ext-2.4-patch-3.patch +ext-2.4-patch-4.patch +linux-2.4.20-xattr-0.8.54-hp.patch +linux-2.4.19-xattr-0.8.54-suse.patch +ext3-2.4-ino_t.patch +ext3-largefile.patch +ext3-truncate_blocks.patch +ext3-use-after-free-suse.patch +ext3-orphan_lock-2.4.19-suse.patch +ext3-delete_thread-2.4.19-suse.patch +ext3-noread-2.4.20.patch +extN-wantedi-2.4.19-suse.patch +ext3-san-2.4.20.patch +ext3-map_inode_page.patch +ext3-error-export.patch +iopen-2.4.19-suse.patch +tcp-zero-copy.patch +add_page_private.patch +removepage-2.4.19-suse.patch +jbd-ctx_switch.patch +jbd-flushtime-2.4.19-suse.patch +jbd-get_write_access.patch +ext3-ea-in-inode-2.4.20.patch +listman-2.4.20.patch +ext3-trusted_ea-2.4.20.patch diff --git a/lustre/kernel_patches/series/vanilla-2.4.19-pre1 b/lustre/kernel_patches/series/vanilla-2.4.19-pre1 new file mode 100644 index 0000000..b6ded90 --- /dev/null +++ b/lustre/kernel_patches/series/vanilla-2.4.19-pre1 @@ -0,0 +1,45 @@ +dev_read_only_2.4.20.patch +exports_2.4.19-pre1.patch +lustre_version.patch +vfs_intent-2.4.19-pre1.patch +invalidate_show-2.4.19-pre1.patch +export-truncate.patch +iod-stock-24-exports.patch +ext3-htree-2.4.19-pre1.patch +linux-2.4.19-pre1-xattr-0.8.54.patch +ext3-2.4.20-fixes.patch +ext3-2.4-ino_t.patch +ext3-largefile.patch +ext3-truncate_blocks.patch +ext3-unmount_sync.patch +ext3-use-after-free-2.4.19-pre1.patch +ext3-orphan_lock.patch +ext3-noread-2.4.20.patch +ext3-delete_thread-2.4.20.patch +extN-wantedi.patch +ext3-san-2.4.20.patch +ext3-map_inode_page.patch +ext3-error-export.patch +iopen-2.4.20.patch +tcp-zero-copy.patch +jbd-dont-account-blocks-twice.patch +jbd-commit-tricks.patch +ext3-no-write-super.patch +add_page_private-2.4.19-pre1.patch +socket-exports-vanilla.patch +removepage-2.4.20.patch +jbd-ctx_switch.patch +jbd-flushtime-2.4.19-suse.patch +jbd-get_write_access.patch +nfs_export_kernel-2.4.19-pre1.patch +ext3-raw-lookup.patch +ext3-ea-in-inode-2.4.20.patch +listman-2.4.20.patch +ext3-trusted_ea-2.4.20.patch +kernel_text_address-2.4.19-pre1.patch +jbd-2.4.19-pre1-jcberr.patch +seq-private-2.4.19-pre1.patch +kdev-2.4.19-pre1.patch +resched-2.4.19-pre1.patch +ext3-xattr-ptr-arith-fix.patch +gfp_memalloc-2.4.22.patch diff --git a/lustre/kernel_patches/series/vanilla-2.4.22 b/lustre/kernel_patches/series/vanilla-2.4.22 index a2726cf..d3b7123 100644 --- a/lustre/kernel_patches/series/vanilla-2.4.22 +++ b/lustre/kernel_patches/series/vanilla-2.4.22 @@ -1,13 +1,12 @@ dev_read_only_2.4.20-rh.patch -exports_2.4.20.patch -kmem_cache_validate_2.4.20.patch +exports_2.4.20-rh-hp.patch lustre_version.patch vfs_intent-2.4.20-vanilla.patch invalidate_show.patch export-truncate.patch -iod-stock-exports-2.4.22.patch +iod-stock-exports-2.4.22.patch ext3-htree-2.4.22-rh.patch -linux-2.4.21-xattr-0.8.54-suse.patch +linux-2.4.22-xattr-0.8.54.patch ext3-orphan_lock-2.4.22-rh.patch ext3-noread-2.4.20.patch ext3-delete_thread-suse.patch @@ -21,10 +20,12 @@ jbd-dont-account-blocks-twice.patch jbd-commit-tricks.patch ext3-no-write-super-chaos.patch add_page_private.patch -socket-exports-vanilla.patch -nfs_export_kernel-2.4.20.patch +socket-exports-2.4.22-rh.patch +nfs_export_kernel-2.4.22.patch ext3-raw-lookup.patch -ext3-ea-in-inode-2.4.20.patch +ext3-ea-in-inode-2.4.22-rh.patch listman-2.4.20.patch ext3-trusted_ea-2.4.20.patch -ext3-inode-reuse-2.4.20.patch +kernel_text_address-2.4.22-vanilla.patch +gfp_memalloc-2.4.22.patch +ext3-xattr-ptr-arith-fix.patch diff --git a/lustre/llite/llite_close.c b/lustre/llite/llite_close.c new file mode 100644 index 0000000..bf064c0 --- /dev/null +++ b/lustre/llite/llite_close.c @@ -0,0 +1,263 @@ +/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- + * vim:expandtab:shiftwidth=8:tabstop=8: + * + * Lustre Lite routines to issue a secondary close after writeback + * + * Copyright (c) 2001-2003 Cluster File Systems, Inc. + * + * This file is part of Lustre, http://www.lustre.org. + * + * Lustre is free software; you can redistribute it and/or + * modify it under the terms of version 2 of the GNU General Public + * License as published by the Free Software Foundation. + * + * Lustre is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with Lustre; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#include + +#define DEBUG_SUBSYSTEM S_LLITE + +#include +#include +#include "llite_internal.h" + +/* record that a write is in flight */ +void llap_write_pending(struct inode *inode, struct ll_async_page *llap) +{ + struct ll_inode_info *lli = ll_i2info(inode); + spin_lock(&lli->lli_lock); + list_add(&llap->llap_pending_write, &lli->lli_pending_write_llaps); + spin_unlock(&lli->lli_lock); +} + +/* record that a write has completed */ +void llap_write_complete(struct inode *inode, struct ll_async_page *llap) +{ + struct ll_inode_info *lli = ll_i2info(inode); + spin_lock(&lli->lli_lock); + if (!list_empty(&llap->llap_pending_write)) + list_del_init(&llap->llap_pending_write); + spin_unlock(&lli->lli_lock); +} + +void ll_open_complete(struct inode *inode) +{ + struct ll_inode_info *lli = ll_i2info(inode); + spin_lock(&lli->lli_lock); + lli->lli_send_done_writing = 0; + spin_unlock(&lli->lli_lock); +} + +/* if we close with writes in flight then we want the completion or cancelation + * of those writes to send a DONE_WRITING rpc to the MDS */ +int ll_is_inode_dirty(struct inode *inode) +{ + struct ll_inode_info *lli = ll_i2info(inode); + int rc = 0; + ENTRY; + + spin_lock(&lli->lli_lock); + if (!list_empty(&lli->lli_pending_write_llaps)) + rc = 1; + spin_unlock(&lli->lli_lock); + RETURN(rc); +} + +void ll_try_done_writing(struct inode *inode) +{ + struct ll_inode_info *lli = ll_i2info(inode); + struct ll_close_queue *lcq = ll_i2sbi(inode)->ll_lcq; + + spin_lock(&lli->lli_lock); + + if (lli->lli_send_done_writing && + list_empty(&lli->lli_pending_write_llaps)) { + + spin_lock(&lcq->lcq_lock); + if (list_empty(&lli->lli_close_item)) { + CDEBUG(D_INODE, "adding inode %lu/%u to close list\n", + inode->i_ino, inode->i_generation); + LASSERT(igrab(inode) == inode); + list_add_tail(&lli->lli_close_item, &lcq->lcq_list); + wake_up(&lcq->lcq_waitq); + } + spin_unlock(&lcq->lcq_lock); + } + + spin_unlock(&lli->lli_lock); +} + +/* The MDS needs us to get the real file attributes, then send a DONE_WRITING */ +void ll_queue_done_writing(struct inode *inode) +{ + struct ll_inode_info *lli = ll_i2info(inode); + ENTRY; + + spin_lock(&lli->lli_lock); + lli->lli_send_done_writing = 1; + spin_unlock(&lli->lli_lock); + + ll_try_done_writing(inode); + EXIT; +} + +/* If we know the file size and have the cookies: + * - send a DONE_WRITING rpc + * + * Otherwise: + * - get a whole-file lock + * - get the authoritative size and all cookies with GETATTRs + * - send a DONE_WRITING rpc + */ +static void ll_close_done_writing(struct inode *inode) +{ + struct ll_inode_info *lli = ll_i2info(inode); + struct ldlm_extent extent = { .start = 0, .end = OBD_OBJECT_EOF }; + struct lustre_handle lockh = { 0 }; + struct obdo obdo; + obd_flag valid; + int rc, ast_flags = 0; + ENTRY; + + memset(&obdo, 0, sizeof(obdo)); + if (test_bit(LLI_F_HAVE_OST_SIZE_LOCK, &lli->lli_flags)) + goto rpc; + + rc = ll_extent_lock_no_validate(NULL, inode, lli->lli_smd, LCK_PW, + &extent, &lockh, ast_flags); + if (rc != ELDLM_OK) { + CERROR("lock acquisition failed (%d): unable to send " + "DONE_WRITING for inode %lu/%u\n", rc, inode->i_ino, + inode->i_generation); + GOTO(out, rc); + } + + rc = ll_lsm_getattr(ll_i2obdexp(inode), lli->lli_smd, &obdo); + if (rc) { + CERROR("inode_getattr failed (%d): unable to send DONE_WRITING " + "for inode %lu/%u\n", rc, inode->i_ino, + inode->i_generation); + ll_extent_unlock(NULL, inode, lli->lli_smd, LCK_PW, &lockh); + GOTO(out, rc); + } + + obdo_refresh_inode(inode, &obdo, valid); + + CDEBUG(D_INODE, "objid "LPX64" size %Lu, blocks %lu, blksize %lu\n", + lli->lli_smd->lsm_object_id, inode->i_size, inode->i_blocks, + inode->i_blksize); + + set_bit(LLI_F_HAVE_OST_SIZE_LOCK, &lli->lli_flags); + + rc = ll_extent_unlock(NULL, inode, lli->lli_smd, LCK_PW, &lockh); + if (rc != ELDLM_OK) + CERROR("unlock failed (%d)? proceeding anyways...\n", rc); + + rpc: + obdo.o_id = inode->i_ino; + obdo.o_size = inode->i_size; + obdo.o_blocks = inode->i_blocks; + obdo.o_valid = OBD_MD_FLID | OBD_MD_FLSIZE | OBD_MD_FLBLOCKS; + + rc = mdc_done_writing(ll_i2sbi(inode)->ll_mdc_exp, &obdo); + out: + iput(inode); +} + +static struct ll_inode_info *ll_close_next_lli(struct ll_close_queue *lcq) +{ + struct ll_inode_info *lli = NULL; + + spin_lock(&lcq->lcq_lock); + + if (lcq->lcq_list.next == NULL) + lli = ERR_PTR(-1); + else if (!list_empty(&lcq->lcq_list)) { + lli = list_entry(lcq->lcq_list.next, struct ll_inode_info, + lli_close_item); + list_del(&lli->lli_close_item); + } + + spin_unlock(&lcq->lcq_lock); + return lli; +} + +static int ll_close_thread(void *arg) +{ + struct ll_close_queue *lcq = arg; + ENTRY; + + /* XXX boiler-plate */ + { + char name[sizeof(current->comm)]; + unsigned long flags; + snprintf(name, sizeof(name) - 1, "ll_close"); + kportal_daemonize(name); + SIGNAL_MASK_LOCK(current, flags); + sigfillset(¤t->blocked); + RECALC_SIGPENDING; + SIGNAL_MASK_UNLOCK(current, flags); + } + + complete(&lcq->lcq_comp); + + while (1) { + struct l_wait_info lwi = { 0 }; + struct ll_inode_info *lli; + struct inode *inode; + + l_wait_event_exclusive(lcq->lcq_waitq, + (lli = ll_close_next_lli(lcq)) != NULL, + &lwi); + if (IS_ERR(lli)) + break; + + inode = ll_info2i(lli); + ll_close_done_writing(inode); + } + + complete(&lcq->lcq_comp); + RETURN(0); +} + +int ll_close_thread_start(struct ll_close_queue **lcq_ret) +{ + struct ll_close_queue *lcq; + pid_t pid; + + OBD_ALLOC(lcq, sizeof(*lcq)); + if (lcq == NULL) + return -ENOMEM; + + spin_lock_init(&lcq->lcq_lock); + INIT_LIST_HEAD(&lcq->lcq_list); + init_waitqueue_head(&lcq->lcq_waitq); + init_completion(&lcq->lcq_comp); + + pid = kernel_thread(ll_close_thread, lcq, 0); + if (pid < 0) { + OBD_FREE(lcq, sizeof(*lcq)); + return pid; + } + + wait_for_completion(&lcq->lcq_comp); + *lcq_ret = lcq; + return 0; +} + +void ll_close_thread_shutdown(struct ll_close_queue *lcq) +{ + init_completion(&lcq->lcq_comp); + lcq->lcq_list.next = NULL; + wake_up(&lcq->lcq_waitq); + wait_for_completion(&lcq->lcq_comp); + OBD_FREE(lcq, sizeof(*lcq)); +} diff --git a/lustre/llite/special.c b/lustre/llite/special.c new file mode 100644 index 0000000..aeb281b --- /dev/null +++ b/lustre/llite/special.c @@ -0,0 +1,365 @@ +/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- + * vim:expandtab:shiftwidth=8:tabstop=8: + * + * Copyright (c) 2002, 2003 Cluster File Systems, Inc. + * Author: Peter Braam + * Author: Phil Schwan + * Author: Andreas Dilger + * + * This file is part of Lustre, http://www.lustre.org. + * + * Lustre is free software; you can redistribute it and/or + * modify it under the terms of version 2 of the GNU General Public + * License as published by the Free Software Foundation. + * + * Lustre is distributed in the hope that it will be useful, + * but WITHOUT ANY WARRANTY; without even the implied warranty of + * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the + * GNU General Public License for more details. + * + * You should have received a copy of the GNU General Public License + * along with Lustre; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#define DEBUG_SUBSYSTEM S_LLITE +#include +#include +#include +#include "llite_internal.h" + +#define FILE_OPS 0 +#define INODE_OPS 1 + +static inline struct file_operations** +get_save_fops(struct file* filp, int mode) +{ + struct inode *inode = filp->f_dentry->d_inode; + struct ll_inode_info *lli = ll_i2info(inode); + if (mode == FILE_OPS){ + if (S_ISFIFO(inode->i_mode)){ + switch (filp->f_mode) { + case 1: /*O_RDONLY*/ + return &(lli->ll_save_ffop); + case 2: /*O_WRONLY*/ + return &(lli->ll_save_wfop); + case 3: /* O_RDWR */ + return &(lli->ll_save_wrfop); + default: + return NULL; + } + } + return &(lli->ll_save_ffop); + } else + return &(lli->ll_save_ifop); + +} + + +static inline void save_fops(struct file *filp, struct inode *inode, + struct file_operations *sfops) +{ + struct ll_inode_info *lli = ll_i2info(inode); + + if (sfops != filp->f_op) { + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + + *pfop = filp->f_op; + if (S_ISCHR(inode->i_mode)) { + filp->f_op = &ll_special_chr_file_fops; + }else if (S_ISFIFO(inode->i_mode)){ + filp->f_op = &ll_special_fifo_file_fops; + } + filp->f_op->owner = lli->ll_save_ffop->owner; + } +} + +static ssize_t ll_special_file_read(struct file *filp, char *buf, + size_t count, loff_t *ppos) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->read) + rc = (*pfop)->read(filp, buf, count, ppos); + + RETURN(rc); +} + +static ssize_t ll_special_file_write(struct file *filp, const char *buf, + size_t count, loff_t *ppos) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->write) + rc = (*pfop)->write(filp, buf, count, ppos); + + RETURN(rc); +} +static int ll_special_file_ioctl(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = -ENOTTY; + + if (pfop && *pfop && (*pfop)->ioctl) { + struct file_operations *sfops = filp->f_op; + + rc = (*pfop)->ioctl(inode, filp, cmd, arg); + save_fops(filp, inode, sfops); + } + RETURN(rc); +} + +static loff_t ll_special_file_seek(struct file *filp, loff_t offset, int origin) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = 0; + + if (pfop && *pfop && (*pfop)->llseek) + rc = (*pfop)->llseek(filp, offset, origin); + else + rc = default_llseek(filp, offset, origin); + + RETURN(rc); +} + + +#define DEFAULT_POLLMASK (POLLIN | POLLOUT | POLLRDNORM | POLLWRNORM) + +static unsigned int +ll_special_file_poll(struct file *filp, struct poll_table_struct *poll_table) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = DEFAULT_POLLMASK; + + if (pfop && *pfop && (*pfop)->poll) + rc = (*pfop)->poll(filp, poll_table); + + RETURN(rc); +} + +static int ll_special_file_open(struct inode *inode, struct file *filp) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->open) + rc = (*pfop)->open(inode, filp); + + RETURN(rc); +} + +static ssize_t ll_special_read(struct file *filp, char *buf, + size_t count, loff_t *ppos) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->read) + rc = (*pfop)->read(filp, buf, count, ppos); + + RETURN(rc); +} + +static ssize_t ll_special_write(struct file *filp, const char *buf, + size_t count, loff_t *ppos) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->write) + rc = (*pfop)->write(filp, buf, count, ppos); + + RETURN(rc); +} + +static int ll_special_ioctl(struct inode *inode, struct file *filp, + unsigned int cmd, unsigned long arg) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = -ENOTTY; + + if (pfop && *pfop && (*pfop)->ioctl) { + struct file_operations *sfops = filp->f_op; + + rc = (*pfop)->ioctl(inode, filp, cmd, arg); + /* sometimes, file_operations will be changed in ioctl */ + save_fops(filp, inode, sfops); + } + + RETURN(rc); +} + +static int ll_special_mmap(struct file * filp, struct vm_area_struct * vma) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = -ENODEV; + + if (pfop && *pfop && (*pfop)->mmap) + rc = (*pfop)->mmap(filp, vma); + + RETURN(rc); +} + +static loff_t ll_special_seek(struct file *filp, loff_t offset, int origin) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = 0; + + if (pfop && *pfop && (*pfop)->llseek) + rc = (*pfop)->llseek(filp, offset, origin); + else + rc = default_llseek(filp, offset, origin); + + RETURN(rc); +} + +static int ll_special_fsync(struct file *filp, struct dentry *dentry, int data) +{ + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->fsync) + rc = (*pfop)->fsync(filp, dentry, data); + + RETURN(rc); +} + +static int ll_special_file_fasync(int fd, struct file *filp, int on) +{ + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = -EINVAL; + + if (pfop && *pfop && (*pfop)->fasync) + rc = (*pfop)->fasync(fd, filp, on); + + RETURN(rc); +} + +static int ll_special_open(struct inode *inode, struct file *filp) +{ + struct ptlrpc_request *req; + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + struct lookup_intent *it; + int rc = -EINVAL; + ENTRY; + + if (pfop && *pfop && (*pfop)->open) { + struct file_operations *sfops = filp->f_op; + + rc = (*pfop)->open(inode, filp); + /* sometimes the file_operations will be changed in open */ + save_fops(filp, inode, sfops); + } + + lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_OPEN); + + it = filp->f_it; + + rc = ll_local_open(filp, it); + if (rc) + RETURN(rc); + req = it->d.lustre.it_data; + if (req) + ptlrpc_req_finished(req); + + RETURN(rc); +} + +static int ll_special_release(struct inode *inode, struct file *filp) +{ + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct file_operations** pfop = get_save_fops (filp, INODE_OPS); + int rc = 0, rc2 = 0; + ENTRY; + + if (pfop && *pfop && (*pfop)->release) { + rc = (*pfop)->release(inode, filp); + } + lprocfs_counter_incr(sbi->ll_stats, LPROC_LL_RELEASE); + + rc2 = ll_mdc_close(sbi->ll_mdc_exp, inode, filp); + + if (rc2 && !rc) + rc = rc2; + + RETURN(rc); +} + +static int ll_special_file_release(struct inode *inode, struct file *filp) +{ + struct ll_sb_info *sbi = ll_i2sbi(inode); + struct file_operations** pfop = get_save_fops (filp, FILE_OPS); + int rc = 0, rc2 = 0; + ENTRY; + + if (pfop && *pfop && (*pfop)->release) { + rc = (*pfop)->release(inode, filp); + } + lprocfs_counter_incr(sbi->ll_stats, LPROC_LL_RELEASE); + + rc2 = ll_mdc_close(sbi->ll_mdc_exp, inode, filp); + + if (rc2 && !rc) + rc = rc2; + + RETURN(rc); + +} + +struct inode_operations ll_special_inode_operations = { + setattr_raw: ll_setattr_raw, + setattr: ll_setattr, +#if (LINUX_VERSION_CODE > KERNEL_VERSION(2,5,0)) + getattr_it: ll_getattr, +#else + revalidate_it: ll_inode_revalidate_it, +#endif +}; + +struct file_operations ll_special_chr_inode_fops = { + open: ll_special_open, +}; + +struct file_operations ll_special_blk_inode_fops = { + read: ll_special_read, + write: ll_special_write, + ioctl: ll_special_ioctl, + open: ll_special_open, + release: ll_special_release, + mmap: ll_special_mmap, + llseek: ll_special_seek, + fsync: ll_special_fsync, +}; + +struct file_operations ll_special_fifo_inode_fops = { + open: ll_special_open, +}; + +struct file_operations ll_special_sock_inode_fops = { + open: ll_special_open +}; + +struct file_operations ll_special_chr_file_fops = { + llseek: ll_special_file_seek, + read: ll_special_file_read, + write: ll_special_file_write, + poll: ll_special_file_poll, + ioctl: ll_special_file_ioctl, + open: ll_special_file_open, + release: ll_special_file_release, + fasync: ll_special_file_fasync, +}; + +struct file_operations ll_special_fifo_file_fops = { + llseek: ll_special_file_seek, + read: ll_special_file_read, + write: ll_special_file_write, + poll: ll_special_file_poll, + ioctl: ll_special_file_ioctl, + open: ll_special_file_open, + release: ll_special_file_release, +}; + diff --git a/lustre/obdclass/llog_internal.h b/lustre/obdclass/llog_internal.h index c364f3f..8674351 100644 --- a/lustre/obdclass/llog_internal.h +++ b/lustre/obdclass/llog_internal.h @@ -5,4 +5,6 @@ int llog_get_cat_list(struct obd_device *obd, struct obd_device *disk_obd, char *name, int count, struct llog_logid *idarray); int llog_put_cat_list(struct obd_device *obd, struct obd_device *disk_obd, char *name, int count, struct llog_logid *); +int llog_cat_id2handle(struct llog_handle *cathandle, struct llog_handle **res, + struct llog_logid *logid); #endif diff --git a/lustre/obdclass/llog_ioctl.c b/lustre/obdclass/llog_ioctl.c new file mode 100644 index 0000000..03279f8 --- /dev/null +++ b/lustre/obdclass/llog_ioctl.c @@ -0,0 +1,294 @@ +/* -*- mode: c; c-basic-offset: 8; indent-tabs-mode: nil; -*- + * vim:expandtab:shiftwidth=8:tabstop=8: + * + * You should have received a copy of the GNU General Public License + * along with Lustre; if not, write to the Free Software + * Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. + */ + +#define DEBUG_SUBSYSTEM S_LOG + +#ifndef EXPORT_SYMTAB +#define EXPORT_SYMTAB +#endif + +#include +#include +#include +#include +#include "llog_internal.h" + +static int str2logid(struct llog_logid *logid, char *str, int len) +{ + char *start, *end, *endp; + + start = str; + if (*start != '#') + RETURN(-EINVAL); + + start++; + if (start - str >= len - 1) + RETURN(-EINVAL); + end = strchr(start, '#'); + if (end == NULL || end == start) + RETURN(-EINVAL); + + *end = '\0'; + logid->lgl_oid = simple_strtoull(start, &endp, 16); + if (endp != end) + RETURN(-EINVAL); + + start = ++end; + if (start - str >= len - 1) + RETURN(-EINVAL); + end = strchr(start, '#'); + if (end == NULL || end == start) + RETURN(-EINVAL); + + *end = '\0'; + logid->lgl_ogr = simple_strtoull(start, &endp, 16); + if (endp != end) + RETURN(-EINVAL); + + start = ++end; + if (start - str >= len - 1) + RETURN(-EINVAL); + logid->lgl_ogen = simple_strtoul(start, &endp, 16); + if (*endp != '\0') + RETURN(-EINVAL); + + RETURN(0); +} + +static int llog_print_cb(struct llog_handle *handle, struct llog_rec_hdr *rec, + void *data) +{ + struct obd_ioctl_data *ioc_data = (struct obd_ioctl_data *)data; + static int l, remains, from, to; + static char *out; + char *endp; + int cur_index; + + if (ioc_data->ioc_inllen1) { + l = 0; + remains = ioc_data->ioc_inllen4 + + size_round(ioc_data->ioc_inllen1) + + size_round(ioc_data->ioc_inllen2) + + size_round(ioc_data->ioc_inllen3); + from = simple_strtol(ioc_data->ioc_inlbuf2, &endp, 0); + if (*endp != '\0') + RETURN(-EINVAL); + to = simple_strtol(ioc_data->ioc_inlbuf3, &endp, 0); + if (*endp != '\0') + RETURN(-EINVAL); + out = ioc_data->ioc_bulk; + ioc_data->ioc_inllen1 = 0; + } + + cur_index = le32_to_cpu(rec->lrh_index); + if (cur_index < from) + RETURN(0); + if (to > 0 && cur_index > to) + RETURN(-LLOG_EEMPTY); + + if (le32_to_cpu(handle->lgh_hdr->llh_flags) & LLOG_F_IS_CAT) { + struct llog_logid_rec *lir = (struct llog_logid_rec *)rec; + if (le32_to_cpu(rec->lrh_type) != LLOG_LOGID_MAGIC) { + CERROR("invalid record in catalog\n"); + RETURN(-EINVAL); + } + + l = snprintf(out, remains, + "[index]: %05d [logid]: #%llx#%llx#%08x\n", + cur_index, lir->lid_id.lgl_oid, + lir->lid_id.lgl_ogr, lir->lid_id.lgl_ogen); + } else { + l = snprintf(out, remains, + "[index]: %05d [type]: %02x [len]: %04d\n", + cur_index, le32_to_cpu(rec->lrh_type), + le32_to_cpu(rec->lrh_len)); + } + out += l; + remains -= l; + if (remains <= 0) { + CERROR("not enough space for print log records\n"); + RETURN(-LLOG_EEMPTY); + } + + RETURN(0); +} + +static int llog_remove_log(struct llog_handle *cat, struct llog_logid *logid) +{ + struct llog_handle *log; + int rc, index = 0; + + down_write(&cat->lgh_lock); + rc = llog_cat_id2handle(cat, &log, logid); + if (rc) { + CDEBUG(D_IOCTL, "cannot find log #%0llx#%0llx#%08x\n", + logid->lgl_oid, logid->lgl_ogr, logid->lgl_ogen); + GOTO(out, rc = -ENOENT); + } + + index = log->u.phd.phd_cookie.lgc_index; + LASSERT(index); + rc = llog_destroy(log); + if (rc) { + CDEBUG(D_IOCTL, "cannot destroy log\n"); + GOTO(out, rc); + } + rc = llog_cancel_rec(cat, index); +out: + up_write(&cat->lgh_lock); + RETURN(rc); + +} + +int llog_ioctl(struct llog_ctxt *ctxt, int cmd, struct obd_ioctl_data *data) +{ + struct llog_logid logid; + int err = 0; + struct llog_handle *handle = NULL; + + if (*data->ioc_inlbuf1 == '#') { + err = str2logid(&logid, data->ioc_inlbuf1, data->ioc_inllen1); + if (err) + GOTO(out, err); + err = llog_create(ctxt, &handle, &logid, NULL); + if (err) + GOTO(out, err); + } else if (*data->ioc_inlbuf1 == '$') { + char *name = data->ioc_inlbuf1 + 1; + err = llog_create(ctxt, &handle, NULL, name); + if (err) + GOTO(out, err); + } else { + GOTO(out, err = -EINVAL); + } + + err = llog_init_handle(handle, 0, NULL); + if (err) + GOTO(out_close, err = -ENOENT); + + switch (cmd) { + case OBD_IOC_LLOG_INFO: { + int l; + int remains = data->ioc_inllen2 + + size_round(data->ioc_inllen1); + char *out = data->ioc_bulk; + + l = snprintf(out, remains, + "logid: #%llx#%llx#%08x\n" + "flags: %x (%s)\n" + "records count: %d\n" + "last index: %d\n", + handle->lgh_id.lgl_oid, handle->lgh_id.lgl_ogr, + handle->lgh_id.lgl_ogen, + le32_to_cpu(handle->lgh_hdr->llh_flags), + le32_to_cpu(handle->lgh_hdr->llh_flags) & + LLOG_F_IS_CAT ? "cat" : "plain", + le32_to_cpu(handle->lgh_hdr->llh_count), + handle->lgh_last_idx); + out += l; + remains -= l; + if (remains <= 0) + CERROR("not enough space for log header info\n"); + + GOTO(out_close, err); + } + case OBD_IOC_LLOG_PRINT: { + LASSERT(data->ioc_inllen1); + err = llog_process(handle, llog_print_cb, data); + if (err == -LLOG_EEMPTY) + err = 0; + + GOTO(out_close, err); + } + case OBD_IOC_LLOG_CANCEL: { + struct llog_cookie cookie; + struct llog_logid plain; + char *endp; + + if (!le32_to_cpu(handle->lgh_hdr->llh_flags) & LLOG_F_IS_CAT) + GOTO(out_close, err = -EINVAL); + + err = str2logid(&plain, data->ioc_inlbuf2, data->ioc_inllen2); + if (err) + GOTO(out_close, err); + cookie.lgc_lgl = plain; + cookie.lgc_index = simple_strtoul(data->ioc_inlbuf3, + &endp, 0); + if (*endp != '\0') + GOTO(out_close, err = -EINVAL); + + err = llog_cat_cancel_records(handle, 1, &cookie); + GOTO(out_close, err); + } + case OBD_IOC_LLOG_REMOVE: { + struct llog_logid plain; + + if (!le32_to_cpu(handle->lgh_hdr->llh_flags) & LLOG_F_IS_CAT) + GOTO(out_close, err = -EINVAL); + + err = str2logid(&plain, data->ioc_inlbuf2, data->ioc_inllen2); + if (err) + GOTO(out_close, err); + err = llog_remove_log(handle, &plain); + GOTO(out_close, err); + } + } + +out_close: + if (handle->lgh_hdr && + le32_to_cpu(handle->lgh_hdr->llh_flags) & LLOG_F_IS_CAT) + llog_cat_put(handle); + else + llog_close(handle); +out: + RETURN(err); +} +EXPORT_SYMBOL(llog_ioctl); + +int llog_catlog_list(struct obd_device *obd, int count, + struct obd_ioctl_data *data) +{ + int size, i; + struct llog_logid *idarray, *id; + char name[32] = "CATLIST"; + char *out; + int l, remains, rc = 0; + + size = sizeof(*idarray) * count; + + OBD_ALLOC(idarray, size); + if (!idarray) + RETURN(-ENOMEM); + memset(idarray, 0, size); + + rc = llog_get_cat_list(obd, obd, name, count, idarray); + if (rc) { + OBD_FREE(idarray, size); + RETURN(rc); + } + + out = data->ioc_bulk; + remains = data->ioc_inllen1; + id = idarray; + for (i = 0; i < count; i++) { + l = snprintf(out, remains, + "catalog log: #%llx#%llx#%08x\n", + id->lgl_oid, id->lgl_ogr, id->lgl_ogen); + id++; + out += l; + remains -= l; + if (remains <= 0) { + CWARN("not enough memory for catlog list\n"); + break; + } + } + OBD_FREE(idarray, size); + RETURN(0); + +} +EXPORT_SYMBOL(llog_catlog_list); diff --git a/lustre/portals/include/linux/kp30.h b/lustre/portals/include/linux/kp30.h index 0acec81..3e6d5e3 100644 --- a/lustre/portals/include/linux/kp30.h +++ b/lustre/portals/include/linux/kp30.h @@ -74,8 +74,10 @@ extern unsigned int portal_cerror; #ifdef __KERNEL__ # include /* THREAD_SIZE */ -#else -# define THREAD_SIZE 8192 +#else +# ifndef THREAD_SIZE /* x86_64 has THREAD_SIZE in userspace */ +# define THREAD_SIZE 8192 +# endif #endif #define LUSTRE_TRACE_SIZE (THREAD_SIZE >> 5) @@ -277,13 +279,17 @@ do { \ #define PORTAL_VMALLOC_SIZE 16384 +#ifndef GFP_MEMALLOC +#define GFP_MEMALLOC 0 +#endif + #define PORTAL_ALLOC(ptr, size) \ do { \ LASSERT (!in_interrupt()); \ if ((size) > PORTAL_VMALLOC_SIZE) \ (ptr) = vmalloc(size); \ else \ - (ptr) = kmalloc((size), GFP_NOFS); \ + (ptr) = kmalloc((size), (GFP_KERNEL | GFP_MEMALLOC)); \ if ((ptr) == NULL) \ CERROR("PORTALS: out of memory at %s:%d (tried to alloc '"\ #ptr "' = %d)\n", __FILE__, __LINE__, (int)(size));\ @@ -312,10 +318,14 @@ do { \ s, (ptr), atomic_read(&portal_kmemory)); \ } while (0) +#ifndef SLAB_MEMALLOC +#define SLAB_MEMALLOC 0 +#endif + #define PORTAL_SLAB_ALLOC(ptr, slab, size) \ do { \ LASSERT(!in_interrupt()); \ - (ptr) = kmem_cache_alloc((slab), SLAB_KERNEL); \ + (ptr) = kmem_cache_alloc((slab), (SLAB_KERNEL | SLAB_MEMALLOC)); \ if ((ptr) == NULL) { \ CERROR("PORTALS: out of memory at %s:%d (tried to alloc" \ " '" #ptr "' from slab '" #slab "')\n", __FILE__, \ @@ -690,6 +700,9 @@ typedef struct { long lwte_p2; long lwte_p3; long lwte_p4; +#if BITS_PER_LONG > 32 + long lwte_pad; +#endif } lwt_event_t; #if LWT_SUPPORT @@ -1112,14 +1125,19 @@ void kportal_put_ni (int nal); # endif #endif -#if (BITS_PER_LONG == 32 || __WORDSIZE == 32) +#if defined(__x86_64__) +# define LPU64 "%Lu" +# define LPD64 "%Ld" +# define LPX64 "%#Lx" +# define LPSZ "%lu" +# define LPSSZ "%ld" +#elif (BITS_PER_LONG == 32 || __WORDSIZE == 32) # define LPU64 "%Lu" # define LPD64 "%Ld" # define LPX64 "%#Lx" # define LPSZ "%u" # define LPSSZ "%d" -#endif -#if (BITS_PER_LONG == 64 || __WORDSIZE == 64) +#elif (BITS_PER_LONG == 64 || __WORDSIZE == 64) # define LPU64 "%lu" # define LPD64 "%ld" # define LPX64 "%#lx" diff --git a/lustre/portals/knals/socknal/socknal_cb.c b/lustre/portals/knals/socknal/socknal_cb.c index 6cd4e73..b6fdc18 100644 --- a/lustre/portals/knals/socknal/socknal_cb.c +++ b/lustre/portals/knals/socknal/socknal_cb.c @@ -891,6 +891,11 @@ ksocknal_queue_tx_locked (ksock_tx_t *tx, ksock_conn_t *conn) spin_lock_irqsave (&sched->kss_lock, flags); + conn->ksnc_tx_deadline = jiffies + + ksocknal_data.ksnd_io_timeout * HZ; + mb(); + /* Extend deadline BEFORE tx is enqueued */ + list_add_tail (&tx->tx_list, &conn->ksnc_tx_queue); if (conn->ksnc_tx_ready && /* able to send */ @@ -2118,28 +2123,30 @@ ksocknal_hello (struct socket *sock, ptl_nid_t *nid, int *type, __u64 *incarnati } int -ksocknal_setup_sock (struct socket *sock) +ksocknal_setup_sock (struct socket *sock) { mm_segment_t oldmm = get_fs (); int rc; int option; struct linger linger; + sock->sk->allocation = GFP_NOFS; + /* Ensure this socket aborts active sends immediately when we close * it. */ - + linger.l_onoff = 0; linger.l_linger = 0; set_fs (KERNEL_DS); - rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER, + rc = sock_setsockopt (sock, SOL_SOCKET, SO_LINGER, (char *)&linger, sizeof (linger)); set_fs (oldmm); if (rc != 0) { CERROR ("Can't set SO_LINGER: %d\n", rc); return (rc); } - + option = -1; set_fs (KERNEL_DS); rc = sock->ops->setsockopt (sock, SOL_TCP, TCP_LINGER2, diff --git a/lustre/portals/knals/toenal/toenal.c b/lustre/portals/knals/toenal/toenal.c index a85b281..07b93cb 100644 --- a/lustre/portals/knals/toenal/toenal.c +++ b/lustre/portals/knals/toenal/toenal.c @@ -210,7 +210,7 @@ error: int ktoenal_close_sock(ptl_nid_t nid) { - long flags; + unsigned long flags; ksock_conn_t *conn; LIST_HEAD (death_row); struct list_head *tmp; diff --git a/lustre/portals/knals/toenal/toenal_cb.c b/lustre/portals/knals/toenal/toenal_cb.c index f7fa794..3af9e33 100644 --- a/lustre/portals/knals/toenal/toenal_cb.c +++ b/lustre/portals/knals/toenal/toenal_cb.c @@ -140,7 +140,7 @@ ktoenal_dist(nal_cb_t *nal, ptl_nid_t nid, unsigned long *dist) ksock_ltx_t * ktoenal_get_ltx (int may_block) { - long flags; + unsigned long flags; ksock_ltx_t *ltx = NULL; for (;;) @@ -268,7 +268,7 @@ ktoenal_recvmsg(struct file *sock, struct iovec *iov, int niov, int toread) } void -ktoenal_process_transmit (ksock_conn_t *conn, long *irq_flags) +ktoenal_process_transmit (ksock_conn_t *conn, unsigned long *irq_flags) { ksock_tx_t *tx = list_entry (conn->ksnc_tx_queue.next, ksock_tx_t, tx_list); int rc; @@ -358,7 +358,7 @@ ktoenal_process_transmit (ksock_conn_t *conn, long *irq_flags) void ktoenal_launch_packet (ksock_conn_t *conn, ksock_tx_t *tx) { - long flags; + unsigned long flags; int nob = tx->tx_nob; struct iovec *iov = tx->tx_iov; int niov = 1; @@ -541,7 +541,7 @@ ktoenal_fmb_callback (void *arg, int error) ksock_fmb_t *fmb = (ksock_fmb_t *)arg; ptl_hdr_t *hdr = (ptl_hdr_t *) page_address(fmb->fmb_pages[0]); ksock_conn_t *conn; - long flags; + unsigned long flags; CDEBUG (D_NET, "routed packet from "LPX64" to "LPX64": %d\n", hdr->src_nid, hdr->dest_nid, error); @@ -790,7 +790,7 @@ ktoenal_new_packet (ksock_conn_t *conn, int nob_to_skip) } void -ktoenal_process_receive (ksock_conn_t *conn, long *irq_flags) +ktoenal_process_receive (ksock_conn_t *conn, unsigned long *irq_flags) { ksock_fmb_t *fmb; int len; diff --git a/lustre/portals/libcfs/debug.c b/lustre/portals/libcfs/debug.c index f27924e..6e2c1ca 100644 --- a/lustre/portals/libcfs/debug.c +++ b/lustre/portals/libcfs/debug.c @@ -775,7 +775,6 @@ portals_debug_msg(int subsys, int mask, char *file, const char *fn, /* If this is the first time, leave a marker in the * output */ debug_daemon_state.overlapped = 1; - ap = NULL; format = "DEBUG MARKER: Debug buffer overlapped\n"; printk(KERN_ERR "LustreError: debug daemon buffer " "overlapped\n"); @@ -982,7 +981,11 @@ extern int lookup_symbol(unsigned long address, char *buf, int buflen); char *portals_debug_dumpstack(void) { +#if defined(__x86_64__) + unsigned long esp = current->thread.rsp; +#else unsigned long esp = current->thread.esp; +#endif unsigned long *stack = (unsigned long *)&esp; int size; unsigned long addr; @@ -1012,7 +1015,7 @@ char *portals_debug_dumpstack(void) /* fix length + sizeof('\0') */ <= pbuf + strlen(buffer) + 28 + 1) break; - size = sprintf(pbuf, "([<%08lx>] %s (0x%x)) ", + size = sprintf(pbuf, "([<%08lx>] %s (0x%p)) ", addr, buffer, stack-1); } pbuf += size; diff --git a/lustre/portals/libcfs/module.c b/lustre/portals/libcfs/module.c index 6432070..7c0cafc 100644 --- a/lustre/portals/libcfs/module.c +++ b/lustre/portals/libcfs/module.c @@ -233,7 +233,7 @@ kportal_router_cmd(struct portals_cfg *pcfg, void * private) CDEBUG (D_IOCTL, "Removing routes via [%d] "LPU64" : "LPU64" - "LPU64"\n", pcfg->pcfg_gw_nal, pcfg->pcfg_nid, pcfg->pcfg_nid2, pcfg->pcfg_nid3); - err = kportal_del_route (pcfg->pcfg_nal, pcfg->pcfg_nid, + err = kportal_del_route (pcfg->pcfg_gw_nal, pcfg->pcfg_nid, pcfg->pcfg_nid2, pcfg->pcfg_nid3); break; diff --git a/lustre/portals/unals/procapi.c b/lustre/portals/unals/procapi.c index 206aea0..2a3fbd8 100644 --- a/lustre/portals/unals/procapi.c +++ b/lustre/portals/unals/procapi.c @@ -48,7 +48,7 @@ * forwards a packaged api call from the 'api' side to the 'library' * side, and collects the result */ -static int procbridge_forward(nal_t *n, int id, void *args, ptl_size_t args_len, +static int procbridge_forward(nal_t *n, int id, void *args, size_t args_len, void *ret, ptl_size_t ret_len) { bridge b = (bridge) n->nal_data; @@ -98,7 +98,7 @@ static int procbridge_shutdown(nal_t *n, int ni) /* Function: validate * useless stub */ -static int procbridge_validate(nal_t *nal, void *base, ptl_size_t extent) +static int procbridge_validate(nal_t *nal, void *base, size_t extent) { return(0); } diff --git a/lustre/portals/unals/proclib.c b/lustre/portals/unals/proclib.c index 2424145..2627253 100644 --- a/lustre/portals/unals/proclib.c +++ b/lustre/portals/unals/proclib.c @@ -47,7 +47,7 @@ static int nal_write(nal_cb_t *nal, void *private, user_ptr dst_addr, void *src_addr, - ptl_size_t len) + size_t len) { memcpy(dst_addr, src_addr, len); return 0; @@ -64,7 +64,7 @@ static int nal_read(nal_cb_t * nal, } static void *nal_malloc(nal_cb_t *nal, - ptl_size_t len) + size_t len) { void *buf = malloc(len); return buf; @@ -72,7 +72,7 @@ static void *nal_malloc(nal_cb_t *nal, static void nal_free(nal_cb_t *nal, void *buf, - ptl_size_t len) + size_t len) { free(buf); } diff --git a/lustre/portals/unals/tcpnal.c b/lustre/portals/unals/tcpnal.c index 3e9b92f..012447f 100644 --- a/lustre/portals/unals/tcpnal.c +++ b/lustre/portals/unals/tcpnal.c @@ -130,8 +130,8 @@ int tcpnal_recv(nal_cb_t *n, lib_msg_t *cookie, unsigned int niov, struct iovec *iov, - ptl_size_t mlen, - ptl_size_t rlen) + size_t mlen, + size_t rlen) { int i; diff --git a/lustre/ptlrpc/import.c b/lustre/ptlrpc/import.c index 650b65a..80742c8 100644 --- a/lustre/ptlrpc/import.c +++ b/lustre/ptlrpc/import.c @@ -192,6 +192,12 @@ int ptlrpc_disconnect_import(struct obd_import *imp) request = ptlrpc_prep_req(imp, rq_opc, 0, NULL, NULL); if (request) { + /* For non-replayable connections, don't attempt + reconnect if this fails */ + if (!imp->imp_obd->obd_replayable) { + imp->imp_state = LUSTRE_IMP_DISCON; + request->rq_send_state = LUSTRE_IMP_DISCON; + } request->rq_replen = lustre_msg_size(0, NULL); rc = ptlrpc_queue_wait(request); ptlrpc_req_finished(request); diff --git a/lustre/ptlrpc/llog_client.c b/lustre/ptlrpc/llog_client.c index 1ddda62..5524843 100644 --- a/lustre/ptlrpc/llog_client.c +++ b/lustre/ptlrpc/llog_client.c @@ -45,7 +45,7 @@ static int llog_client_create(struct llog_ctxt *ctxt, struct llog_handle **res, struct llog_logid *logid, char *name) { - struct obd_import *imp; + struct obd_import *imp; struct llogd_body req_body; struct llogd_body *body; struct llog_handle *handle; @@ -66,17 +66,17 @@ static int llog_client_create(struct llog_ctxt *ctxt, struct llog_handle **res, *res = handle; memset(&req_body, 0, sizeof(req_body)); - if (logid) + if (logid) req_body.lgd_logid = *logid; req_body.lgd_ctxt_idx = ctxt->loc_idx - 1; - + if (name) { size[bufcount] = strlen(name) + 1; tmp[bufcount] = name; bufcount++; } - req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_CREATE, bufcount, size, tmp); + req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_CREATE,bufcount,size,tmp); if (!req) GOTO(err_free, rc = -ENOMEM); @@ -84,7 +84,7 @@ static int llog_client_create(struct llog_ctxt *ctxt, struct llog_handle **res, rc = ptlrpc_queue_wait(req); if (rc) GOTO(err_free, rc); - + body = lustre_swab_repbuf(req, 0, sizeof(*body), lustre_swab_llogd_body); if (body == NULL) { @@ -106,9 +106,9 @@ err_free: } -static int llog_client_next_block(struct llog_handle *loghandle, - int *cur_idx, int next_idx, - __u64 *cur_offset, void *buf, int len) +static int llog_client_next_block(struct llog_handle *loghandle, + int *cur_idx, int next_idx, + __u64 *cur_offset, void *buf, int len) { struct obd_import *imp = loghandle->lgh_ctxt->loc_imp; struct ptlrpc_request *req = NULL; @@ -119,32 +119,32 @@ static int llog_client_next_block(struct llog_handle *loghandle, int rc; ENTRY; - req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_NEXT_BLOCK, 1, &size, NULL); + req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_NEXT_BLOCK, 1,&size,NULL); if (!req) GOTO(out, rc = -ENOMEM); - + body = lustre_msg_buf(req->rq_reqmsg, 0, sizeof (*body)); body->lgd_logid = loghandle->lgh_id; body->lgd_ctxt_idx = loghandle->lgh_ctxt->loc_idx - 1; body->lgd_llh_flags = loghandle->lgh_hdr->llh_flags; - body->lgd_cur_offset = *cur_offset; body->lgd_index = next_idx; body->lgd_saved_index = *cur_idx; body->lgd_len = len; + body->lgd_cur_offset = *cur_offset; repsize[1] = len; req->rq_replen = lustre_msg_size(2, repsize); rc = ptlrpc_queue_wait(req); if (rc) GOTO(out, rc); - + body = lustre_swab_repbuf(req, 0, sizeof(*body), lustre_swab_llogd_body); if (body == NULL) { CERROR ("Can't unpack llogd_body\n"); GOTO(out, rc =-EFAULT); } - + ptr = lustre_msg_buf(req->rq_repmsg, 1, len); if (ptr == NULL) { CERROR ("Can't unpack bitmap\n"); @@ -153,7 +153,7 @@ static int llog_client_next_block(struct llog_handle *loghandle, *cur_idx = body->lgd_saved_index; *cur_offset = body->lgd_cur_offset; - + memcpy(buf, ptr, len); out: @@ -174,10 +174,10 @@ static int llog_client_read_header(struct llog_handle *handle) int rc; ENTRY; - req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_READ_HEADER, 1, &size, NULL); + req = ptlrpc_prep_req(imp, LLOG_ORIGIN_HANDLE_READ_HEADER,1,&size,NULL); if (!req) GOTO(out, rc = -ENOMEM); - + body = lustre_msg_buf(req->rq_reqmsg, 0, sizeof (*body)); body->lgd_logid = handle->lgh_id; body->lgd_ctxt_idx = handle->lgh_ctxt->loc_idx - 1; @@ -187,9 +187,8 @@ static int llog_client_read_header(struct llog_handle *handle) rc = ptlrpc_queue_wait(req); if (rc) GOTO(out, rc); - - hdr = lustre_swab_repbuf(req, 0, sizeof(*hdr), - lustre_swab_llog_hdr); + + hdr = lustre_swab_repbuf(req, 0, sizeof(*hdr), lustre_swab_llog_hdr); if (hdr == NULL) { CERROR ("Can't unpack llog_hdr\n"); GOTO(out, rc =-EFAULT); diff --git a/lustre/ptlrpc/llog_server.c b/lustre/ptlrpc/llog_server.c index aeddb5b..b2c087f 100644 --- a/lustre/ptlrpc/llog_server.c +++ b/lustre/ptlrpc/llog_server.c @@ -137,15 +137,15 @@ int llog_origin_handle_next_block(struct ptlrpc_request *req) GOTO(out_close, rc); memset(buf, 0, LLOG_CHUNK_SIZE); - rc = llog_next_block(loghandle, &body->lgd_saved_index, - body->lgd_index, + rc = llog_next_block(loghandle, &body->lgd_saved_index, + body->lgd_index, &body->lgd_cur_offset, buf, LLOG_CHUNK_SIZE); if (rc) GOTO(out_close, rc); rc = lustre_pack_reply(req, 2, size, NULL); - if (rc) + if (rc) GOTO(out_close, rc = -ENOMEM); ptr = lustre_msg_buf(req->rq_repmsg, 0, sizeof (body)); @@ -210,7 +210,7 @@ int llog_origin_handle_read_header(struct ptlrpc_request *req) rc = lustre_pack_reply(req, 1, size, NULL); - if (rc) + if (rc) GOTO(out_close, rc = -ENOMEM); hdr = lustre_msg_buf(req->rq_repmsg, 0, sizeof (*hdr)); diff --git a/lustre/scripts/lbuild b/lustre/scripts/lbuild new file mode 100755 index 0000000..327ae91 --- /dev/null +++ b/lustre/scripts/lbuild @@ -0,0 +1,437 @@ +#!/bin/sh + +TOPDIR=$PWD + +# CVSROOT is inherited from the environment +KERNELDIR= +LUSTRE= +PUBLISH=0 +RELEASE=0 +DO_SRC=1 +TAG= +TARGET= +TARGET_ARCHS= +CONFIGURE_FLAGS= +EXTRA_VERSION= + +# from target file +KERNEL= +SERIES= +CONFIG= +VERSION= + +BASE_ARCH= +BIGMEM_ARCHS= +BOOT_ARCHS= +JENSEN_ARCHS= +SMP_ARCHS= +UP_ARCHS= + +DATE=$(date) + +cleanup() +{ + true +} + +fatal() +{ + cleanup + [ "$2" ] && echo + [ "$2" ] && echo "${0##*/}: $2" + exit $1 +} + +publishing() +{ + (( $PUBLISH )) || return 0 +} + +is_release() +{ + (( $RELEASE )) || return 0 +} + +list_targets() +{ + echo -n "Available targets:" + for target in $TOPDIR/lustre/kernel_patches/targets/*.target ; do + target_file=${target##*/} + echo -n " ${target_file%%.target}" + done + echo +} + +usage() +{ + cat <] + + -d CVSROOT + Specifies the CVS Root to use when pulling files from CVS. The + environment variable \$CVSROOT is used if this option is not + present. + + --extraversion=EXTRAVERSION + Text to use for the rpm release and kernel extraversion. + + --kerneldir=KERNELDIR + Directory containing Linux source tarballs referenced by target + files. + + --lustre=LUSTRE + Path to an existing lustre source tarball to use instead of + pulling from CVS. + + --nosrc + Do not build a .src.rpm, a full kernel patch, or a patched kernel + tarball. + + --publish + Publish the packages, patches, and tarballs on the ftp server. + + --release + Specifies that the files generated do not include timestamps, and + that this is an official release. + + --tag=TAG + A CVS branch/tag name to build from when pulling from CVS. + + --target=TARGET + The name of the target to build. The available targets are listed + below. + + --target-archs=TARGET_ARCHS + A (space delimited) list of architectures to build. By default, + all of the archs supported by the TARGET will be built, in + addition to a .src.rpm. This option can limit those, for machines + that can only build certain archs or if you only want a certain + arch built (for testing, or a one-off kernel). + + Also note that by using a non-"base" arch (eg, i386) only kernels + will be built - there will be no lustre-lite-utils package. + +EOF + +# list_targets + + fatal "$1" "$2" +} + +check_options() +{ + if [ "$LUSTRE" ] ; then + [ -r "$LUSTRE" ] || \ + usage 1 "Could not find Lustre source tarball '$LUSTRE'." + else + [ "$CVSROOT" ] || \ + usage 1 "Either specify a CVS Root with -d, or a Lustre source tarball with --lustre." + [ "$TAG" ] || \ + usage 1 "A branch/tag name must be specified with --tag when not building from a tarball." + fi + + [ "$KERNELDIR" ] || \ + usage 1 "A kernel directory must be specified with --kerneldir." + + [ -d "$KERNELDIR" ] || \ + usage 1 "$KERNELDIR is not a directory." + + if [ "$RELEASE" = "no" ] ; then + [ "$TAG" ] || \ + usage 1 "When building a snapshot, a tag name must be used." + fi + + TIMESTAMP=$(date -d "$DATE" "+%Y%m%d%H%M") + + [ "$TARGET" ] || usage 1 "A target must be specified with --target." +# TARGET_FILE="$TOPDIR/kernel_patches/targets/$TARGET.target" +# [ -r "$TARGET_FILE" ] || \ +# usage 1 "Target '$TARGET' was not found." +} + +uniqify() +{ + echo $(echo "$*" | xargs -n 1 | sort -u) +} + +load_target() +{ + EXTRA_VERSION_save="$EXTRA_VERSION" + TARGET_FILE="$TOPDIR/lustre/kernel_patches/targets/$TARGET.target" + [ -r "$TARGET_FILE" ] || \ + fatal 1 "Target $TARGET was not found." + + echo "Loading target config file $TARGET.target..." + + . "$TARGET_FILE" + + [ "$KERNEL" ] || fatal 1 "Target $TARGET did not specify a kernel." + [ "$SERIES" ] || fatal 1 "Target $TARGET did not specify a kernel patch series." +# [ "$CONFIG" ] || fatal 1 "Target $TARGET did not specify a kernel config." + [ "$VERSION" ] || fatal 1 "Target $TARGET did not specify a kernel version." + + if [ "$KERNELDIR" ] ; then + KERNEL_FILE="$KERNELDIR/$KERNEL" + [ -r "$KERNELDIR/$KERNEL" ] || \ + fatal 1 "Target $TARGET's kernel file $KERNEL not found in kernel directory $KERNELDIR." + fi + + SERIES_FILE="$TOPDIR/lustre/kernel_patches/series/$SERIES" + [ -r "$SERIES_FILE" ] || \ + fatal 1 "Target $TARGET's series $SERIES missing from $TOPDIR/lustre/kernel_patches/series." + + CONFIG_FILE="$TOPDIR/lustre/kernel_patches/kernel_configs/$CONFIG" + [ -r "$CONFIG_FILE" ] || \ + fatal 1 "Target $TARGET's config file $CONFIG missing from $TOPDIR/lustre/kernel_patches/kernel_configs/configs." + + if [ "$EXTRA_VERSION_save" ] ; then + EXTRA_VERSION="$EXTRA_VERSION_save" + else + EXTRA_VERSION="${EXTRA_VERSION}_${TAG//_/}.${TIMESTAMP}" + fi + EXTRA_VERSION=${EXTRA_VERSION//-/_/} + + ALL_ARCHS="$BASE_ARCHS $BIGMEM_ARCHS $BOOT_ARCHS $JENSEN_ARCHS $SMP_ARCHS $UP_ARCHS" + + BUILD_ARCHS= + for arch in $(uniqify "$ALL_ARCHS") ; do + if [ -z "$TARGET_ARCHS" ] || echo "$TARGET_ARCHS" | grep -s "$arch" ; then + BUILD_ARCHS="$BUILD_ARCHS $arch" + fi + done + [ "$BUILD_ARCHS" ] || usage 1 "No available target archs to build." + echo "Building for: $BUILD_ARCHS" +} + +tarflags() +{ + case "$1" in + '') + fatal 1 "tarflags(): File name argument missing." + ;; + *.tar.gz) + echo 'zxf' + ;; + *.tar.bz2) + echo 'jxf' + ;; + *) + fatal 1 "tarflags(): Unrecognized tar extension in file: $1" + ;; + esac +} + +untar() +{ + echo "Untarring ${1##*/}..." + tar $(tarflags "$1") "$1" +} + +unpack_lustre() +{ + DIRNAME="lustre-$TAG-$TIMESTAMP" + if [ "$LUSTRE" ] ; then + untar "$LUSTRE" + [ -d lustre ] || ln -sf lustre* lustre + else + cvs -d "$CVSROOT" -qz3 co -D "$DATE" "-r$TAG" -d "$DIRNAME" lustre || \ + fatal 1 "There was an error checking out Lustre from CVS." + echo "Creating lustre tarball..." + tar zcf "$DIRNAME.tar.gz" "$DIRNAME" \ + --exclude "CVS" --exclude "*~" --exclude ".cvsignore" || \ + fatal 1 "Could not create Lustre tarball." + LUSTRE="$PWD/$DIRNAME.tar.gz" + ln -sf "$DIRNAME" lustre + fi +} + +unpack_linux() +{ + untar "$KERNEL_FILE" + [ -d linux ] || ln -sf linux* linux +} + +patch_linux() +{ + FULL_PATCH="$PWD/lustre-kernel-${target}-${EXTRA_VERSION}.patch" + [ -f "$FULL_PATCH" ] && rm -f "$FULL_PATCH" + pushd linux >/dev/null + echo -n "Applying patches:" + for patch in $(<"$SERIES_FILE") ; do + echo -n " $patch" + PATCH_FILE="$TOPDIR/lustre/kernel_patches/patches/$patch" + [ -r "$PATCH_FILE" ] || \ + fatal 1 "Patch $patch does not exist in Lustre tree." + cat "$PATCH_FILE" >> "$FULL_PATCH" || \ + fatal 1 "Error adding patch $patch to full patch." + patch -s -p1 < "$PATCH_FILE" || fatal 1 "Error applying patch $patch." + done + echo + popd >/dev/null + echo "Full patch has been saved in ${FULL_PATCH##*/}." + echo "Replacing .config files..." + rm -f linux/configs/* + cp -v lustre/kernel_patches/kernel_configs/kernel-${VERSION}-${TARGET}*.config linux/configs/ +} + +pack_linux() +{ + TARBALL="$(readlink linux)-$EXTRA_VERSION.tar.gz" + echo "Creating patched linux tarball $TARBALL..." + tar zcf "$TARBALL" "$(readlink linux)" \ + --exclude "CVS" --exclude ".cvsignore" || \ + --exclude "*.orig" --exclude "*~" --exclude "*.rej" || \ + fatal 1 "Error creating patched Linux tarball." +} + +clean_linux() +{ + [ -d linux ] || return 0 + echo "Cleaning linux..." + [ -L linux ] && rm -rf $(readlink linux) + rm -rf linux +} + +prep_build() +{ + # make .spec file + sed -e s/@KERNEL_VERSION@/$VERSION/g \ + -e s/@KERNEL_RELEASE@/$EXTRA_VERSION/g \ + -e s/@KERNEL_SOURCE@/$KERNEL/g \ + -e s/@LUSTRE_SOURCE@/${LUSTRE##*/}/g \ + -e s/@LUSTRE_TARGET@/$TARGET/g \ + -e s/@CONFIGURE_FLAGS@/$CONFIGURE_FLAGS/g \ + -e s/@BASE_ARCHS@/$BASE_ARCHS/g \ + -e s/@BIGMEM_ARCHS@/$BIGMEM_ARCHS/g \ + -e s/@BOOT_ARCHS@/$BOOT_ARCHS/g \ + -e s/@JENSEN_ARCHS@/$BOOT_ARCHS/g \ + -e s/@SMP_ARCHS@/$SMP_ARCHS/g \ + -e s/@UP_ARCHS@/$UP_ARCHS/g \ + < $TOPDIR/lustre/scripts/lustre-kernel-2.4.spec.in \ + > lustre-kernel-2.4.spec + [ -d SRPMS ] || mkdir SRPMS + [ -d RPMS ] || mkdir RPMS + [ -d BUILD ] || mkdir BUILD + [ -d SOURCES ] || mkdir SOURCES + cp $TOPDIR/lustre/scripts/linux-merge-config.awk SOURCES + cp $TOPDIR/lustre/scripts/linux-merge-modules.awk SOURCES + cp "$LUSTRE" "$KERNEL_FILE" SOURCES +} + +clean_lustre() +{ + [ -d lustre ] || return 0 + echo "Cleaning lustre..." + [ -L lustre ] && rm -rf $(readlink lustre) + rm -rf lustre +} + +build() +{ + echo "Building rpms for: $BUILD_ARCHS..." + targets= + for arch in $BUILD_ARCHS ; do + targets="--target $arch $targets" + done + + rpmbuild $targets -bb lustre-kernel-2.4.spec \ + --define "_topdir $TOPDIR" \ + fatal 1 "Error building rpms for $arch." + + (( $DO_SRC )) && rpmbuild -bs lustre-kernel-2.4.spec \ + --define "_topdir $TOPDIR" \ + fatal 1 "Error building .src.rpm." +} + +publish() +{ + publishing || return 0 +} + +[ -r ~/.lbuildrc ] && . ~/.lbuildrc + +options=$(getopt -o d:D:h -l extraversion:,kerneldir:,lustre:,nosrc,publish,release,tag:,target:,target-archs: -- "$@") + +eval set -- "$options" + +while [ "$1" ] ; do + case "$1" in + '') + usage 1 + ;; + -d) + CVSROOT=$2 + shift 2 + ;; + -D) + DATE=$2 + shift 2 + ;; + --extraversion) + EXTRA_VERSION=$2 + shift 2 + ;; + --help | -h) + usage 0 + ;; + --kerneldir) + KERNELDIR=$2 + shift 2 + ;; + --lustre) + LUSTRE=$2 + shift 2 + ;; + --nosrc) + DO_SRC=0 + shift 1 + ;; + --publish) + PUBLISH=1 + shift + ;; + --release) + RELEASE=1 + shift + ;; + --tag) + TAG=$2 + shift 2 + ;; + --target) + TARGET=$2 + shift 2 + ;; + --target-archs) + TARGET_ARCHS=$2 + shift 2 + ;; + --) + shift + CONFIGURE_FLAGS=$@ + break + ;; + *) + usage 1 "Unrecognized option: $1" + ;; + esac +done + +check_options + +unpack_lustre +load_target + +if (( $DO_SRC )) ; then + unpack_linux + patch_linux + pack_linux + clean_linux +fi + +# prep_build needs the .spec.in from the lustre source +prep_build +clean_lustre + +build +publish diff --git a/lustre/tests/cfg/insanity-local.sh b/lustre/tests/cfg/insanity-local.sh new file mode 100644 index 0000000..4ce742f --- /dev/null +++ b/lustre/tests/cfg/insanity-local.sh @@ -0,0 +1,30 @@ +mds_HOST=${mds_HOST:-`hostname`} +mdsfailover_HOST=${mdsfailover_HOST:-""} +ost1_HOST=${ost1_HOST:-"`hostname`"} +ost2_HOST=${ost2_HOST:-"`hostname`"} +client_HOST="'*'" +LIVE_CLIENT=${LIVE_CLIENT:-"`hostname`"} +# This should always be a list, not a regexp +FAIL_CLIENTS=${FAIL_CLIENTS:-""} + +NETTYPE=${NETTYPE:-tcp} +TIMEOUT=${TIMEOUT:-30} +PTLDEBUG=${PTLDEBUG:-'"ha|info|ioctl|malloc"'} +MOUNT=${MOUNT:-"/mnt/lustre"} +#CLIENT_UPCALL=${CLIENT_UPCALL:-`pwd`/client-upcall-mdev.sh} +UPCALL=${CLIENT_UPCALL:-`pwd`/replay-single-upcall.sh} + +MDSDEV=${MDSDEV:-$ROOT/tmp/mds-`hostname`} +MDSSIZE=${MDSSIZE:-10000} #50000000 + +OSTDEV=${OSTDEV:-$ROOT/tmp/ost-`hostname`} +OSTSIZE=${OSTSIZE:=10000} #50000000 +FSTYPE=${FSTYPE:-ext3} +STRIPE_BYTES=${STRIPE_BYTES:-65536} #1048576 +STRIPES_PER_OBJ=${STRIPES_PER_OBJ:-0} + +FAILURE_MODE=${FAILURE_MODE:-SOFT} # or HARD +POWER_DOWN=${POWER_DOWN:-"powerman --off"} +POWER_UP=${POWER_UP:-"powerman --on"} + +PDSH=no_dsh diff --git a/lustre/tests/cfg/insanity-mdev.sh b/lustre/tests/cfg/insanity-mdev.sh new file mode 100644 index 0000000..5e69356 --- /dev/null +++ b/lustre/tests/cfg/insanity-mdev.sh @@ -0,0 +1,31 @@ +mds_HOST=${mds_HOST:-mdev4} +mdsfailover_HOST=${mdsfailover_HOST:-mdev5} +ost1_HOST=${ost1_HOST:-mdev2} +ost2_HOST=${ost2_HOST:-mdev3} +client_HOST=client +LIVE_CLIENT=${LIVE_CLIENT:-mdev6} +# This should always be a list, not a regexp +FAIL_CLIENTS=${FAIL_CLIENTS:-mdev7} + +NETTYPE=${NETTYPE:-tcp} + +TIMEOUT=${TIMEOUT:-30} +#PTLDEBUG=${PTLDEBUG:-'"ha|info|ioctl|malloc"'} +PTLDEBUG=${PTLDEBUG:-0} +MOUNT=${MOUNT:-"/mnt/lustre"} +UPCALL=${CLIENT_UPCALL:-`pwd`/replay-single-upcall.sh} + +MDSDEV=${MDSDEV:-/dev/sda1} +MDSSIZE=${MDSSIZE:-50000} + +OSTDEV=${OSTDEV:-/tmp/ost-`hostname`} +OSTSIZE=${OSTSIZE:=50000} +FSTYPE=${FSTYPE:-ext3} +STRIPE_BYTES=${STRIPE_BYTES:-1048576} +STRIPES_PER_OBJ=${STRIPES_PER_OBJ:-0} + +FAILURE_MODE=${FAILURE_MODE:-HARD} # or HARD +POWER_DOWN=${POWER_DOWN:-"powerman --off"} +POWER_UP=${POWER_UP:-"powerman --on"} + +PDSH="pdsh -S -w " diff --git a/lustre/tests/cfg/local.sh b/lustre/tests/cfg/local.sh new file mode 100644 index 0000000..f9c5198 --- /dev/null +++ b/lustre/tests/cfg/local.sh @@ -0,0 +1,33 @@ +# oldstyle +MDSNODE=${MDSNODE:-`hostname`} +OSTNODE=${OSTNODE:-`hostname`} + +mds_HOST=${mds_HOST:-$MDSNODE} +mdsfailover_HOST=${mdsfailover_HOST} +ost_HOST=${ost_HOST:-$OSTNODE} +ost2_HOST=${ost2_HOST:-$ost_HOST} +client_HOST=${client_HOST:-"'*'"} +NETTYPE=${NETTYPE:-tcp} + +MOUNT=${MOUNT:-"/mnt/lustre"} +MOUNT1=${MOUNT1:-$MOUNT} +MOUNT2=${MOUNT2:-"/mnt/lustre2"} +DIR=${DIR:-$MOUNT} +DIR2=${DIR2:-$MOUNT1} +PTLDEBUG=${PTLDEBUG:-0} +PDSH=${PDSH:-no_dsh} + +MDSDEV=${MDSDEV:-$ROOT/tmp/mds-`hostname`} +MDSSIZE=${MDSSIZE:-10000} +OSTDEV=${OSTDEV:-$ROOT/tmp/ost-`hostname`} +OSTSIZE=${OSTSIZE:-10000} +FSTYPE=${FSTYPE:-ext3} +TIMEOUT=${TIMEOUT:-10} +UPCALL=${UPCALL:-$PWD/replay-single-upcall.sh} + +STRIPE_BYTES=${STRIPE_BYTES:-65536} +STRIPES_PER_OBJ=${STRIPES_PER_OBJ:-0} + +FAILURE_MODE=${FAILURE_MODE:-SOFT} # or HARD +POWER_DOWN=${POWER_DOWN:-"powerman --off"} +POWER_UP=${POWER_UP:-"powerman --on"} diff --git a/lustre/tests/cfg/mdev.sh b/lustre/tests/cfg/mdev.sh new file mode 100644 index 0000000..ec8edf2 --- /dev/null +++ b/lustre/tests/cfg/mdev.sh @@ -0,0 +1,30 @@ + +mds_HOST=${mds_HOST:-mdev4} +mdsfailover_HOST=${mdsfailover_HOST:-mdev5} +ost_HOST=${ost_HOST:-mdev2} +ost2_HOST=${ost2_HOST:-mdev3} +client_HOST=${client_HOST:-client} +NETTYPE=${NETTYPE:-tcp} + +MOUNT=${MOUNT:-"/mnt/lustre"} +MOUNT1=${MOUNT1:-$MOUNT} +MOUNT2=${MOUNT2:-"/mnt/lustre2"} +DIR=${DIR:-$MOUNT} +DIR2=${DIR2:-$MOUNT1} +PTLDEBUG=${PTLDEBUG:-0} +PDSH=${PDSH:-pdsh -S -w} + +MDSDEV=${MDSDEV:-/dev/sda1} +MDSSIZE=${MDSSIZE:-50000} +OSTDEV=${OSTDEV:-/tmp/ost-`hostname`} +OSTSIZE=${OSTSIZE:-20000} +FSTYPE=${FSTYPE:-ext3} +TIMEOUT=${TIMEOUT:-10} +UPCALL=${UPCALL:-$PWD/replay-single-upcall.sh} + +STRIPE_BYTES=${STRIPE_BYTES:-65536} +STRIPES_PER_OBJ=${STRIPES_PER_OBJ:-0} + +FAILURE_MODE=${FAILURE_MODE:-SOFT} # or HARD +POWER_DOWN=${POWER_DOWN:-"powerman --off"} +POWER_UP=${POWER_UP:-"powerman --on"} diff --git a/lustre/tests/conf-sanity.sh b/lustre/tests/conf-sanity.sh new file mode 100644 index 0000000..f71e39e --- /dev/null +++ b/lustre/tests/conf-sanity.sh @@ -0,0 +1,216 @@ +#!/bin/bash +# requirement: +# add uml1 uml2 uml3 in your /etc/hosts + +set -e + +SRCDIR=`dirname $0` +PATH=$PWD/$SRCDIR:$SRCDIR:$SRCDIR/../utils:$PATH + +LUSTRE=${LUSTRE:-`dirname $0`/..} +RLUSTRE=${RLUSTRE:-$LUSTRE} + +. $LUSTRE/tests/test-framework.sh + +init_test_env $@ + +. ${CONFIG:=$LUSTRE/tests/cfg/local.sh} + +FORCE=${FORCE:-" --force"} + +gen_config() { + rm -f $XMLCONFIG + + add_mds mds --dev $MDSDEV --size $MDSSIZE + add_lov lov1 mds --stripe_sz $STRIPE_BYTES\ + --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0 + add_ost ost --lov lov1 --dev $OSTDEV --size $OSTSIZE + add_client client mds --lov lov1 --path $MOUNT +} + +gen_second_config() { + rm -f $XMLCONFIG + + add_mds mds2 --dev $MDSDEV --size $MDSSIZE + add_lov lov2 mds2 --stripe_sz $STRIPE_BYTES\ + --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0 + add_ost ost2 --lov lov2 --dev $OSTDEV --size $OSTSIZE + add_client client mds2 --lov lov2 --path $MOUNT2 +} + +start_mds() { + echo "start mds service on `facet_active_host mds`" + start mds --reformat $MDSLCONFARGS > /dev/null || return 94 +} +stop_mds() { + echo "stop mds service on `facet_active_host mds`" + stop mds $@ > /dev/null || return 97 +} + +start_ost() { + echo "start ost service on `facet_active_host ost`" + start ost --reformat $OSTLCONFARGS > /dev/null || return 95 +} + +stop_ost() { + echo "stop ost service on `facet_active_host ost`" + stop ost $@ > /dev/null || return 98 +} + +mount_client() { + local MOUNTPATH=$1 + echo "mount lustre on ${MOUNTPATH}....." + zconf_mount $MOUNTPATH > /dev/null || return 96 +} + +umount_client() { + local MOUNTPATH=$1 + echo "umount lustre on ${MOUNTPATH}....." + zconf_umount $MOUNTPATH > /dev/null || return 97 +} + +manual_umount_client(){ + echo "manual umount lustre on ${MOUNTPATH}...." + do_facet client "umount $MOUNT" +} + +setup() { + start_ost + start_mds + mount_client $MOUNT +} + +cleanup() { + umount_client $MOUNT || return -200 + stop_mds || return -201 + stop_ost || return -202 +} + +check_mount() { + do_facet client "touch $DIR/a" || return 71 + do_facet client "rm $DIR/a" || return 72 + echo "setup single mount lustre success" +} + +check_mount2() { + do_facet client "touch $DIR/a" || return 71 + do_facet client "rm $DIR/a" || return 72 + do_facet client "touch $DIR2/a" || return 73 + do_facet client "rm $DIR2/a" || return 74 + echo "setup double mount lustre success" +} + +build_test_filter + +#create single point mountpoint + +gen_config + + +test_0() { + start_ost + start_mds + mount_client $MOUNT + check_mount || return 41 + cleanup +} +run_test 0 "single mount setup" + +test_1() { + start_ost + echo "start ost second time..." + start ost --reformat $OSTLCONFARGS > /dev/null + start_mds + mount_client $MOUNT + check_mount || return 42 + cleanup +} +run_test 1 "start up ost twice" + +test_2() { + start_ost + start_mds + echo "start mds second time.." + start mds --reformat $MDSLCONFARGS > /dev/null + + mount_client $MOUNT + check_mount || return 43 + cleanup +} +run_test 2 "start up mds twice" + +test_3() { + setup + mount_client $MOUNT + + check_mount || return 44 + + umount_client $MOUNT + cleanup +} +run_test 3 "mount client twice" + +test_4() { + setup + touch $DIR/$tfile || return 85 + stop_ost ${FORCE} + + # cleanup may return an error from the failed + # disconnects; for now I'll consider this successful + # if all the modules have unloaded. + if ! cleanup ; then + lsmod | grep -q portals && return 1 + fi + return 0 +} +run_test 4 "force cleanup ost, then cleanup" + +test_5() { + setup + touch $DIR/$tfile || return 86 + stop_mds ${FORCE} || return 98 + + # cleanup may return an error from the failed + # disconnects; for now I'll consider this successful + # if all the modules have unloaded. + if ! cleanup ; then + lsmod | grep -q portals && return 1 + fi + return 0 +} +run_test 5 "force cleanup mds, then cleanup" + +test_6() { + setup + manual_umount_client + mount_client ${MOUNT} || return 87 + touch $DIR/a || return 86 + cleanup +} +run_test 6 "manual umount, then mount again" + +test_7() { + setup + manual_umount_client + cleanup +} +run_test 7 "manual umount, then cleanup" + +test_8() { + start_ost + start_mds + + mount_client $MOUNT + mount_client $MOUNT2 + + check_mount2 || return 45 + umount $MOUNT + umount_client $MOUNT2 + + stop_mds + stop_ost +} +run_test 8 "double mount setup" + + +equals_msg "Done" diff --git a/lustre/tests/insanity.sh b/lustre/tests/insanity.sh new file mode 100755 index 0000000..7a9b58f --- /dev/null +++ b/lustre/tests/insanity.sh @@ -0,0 +1,541 @@ +#!/bin/sh +# Test multiple failures, AKA Test 17 + +set -e + +LUSTRE=${LUSTRE:-`dirname $0`/..} +. $LUSTRE/tests/test-framework.sh + +init_test_env $@ + +. ${CONFIG:=$LUSTRE/tests/cfg/insanity-local.sh} + +ALWAYS_EXCEPT="" + +build_test_filter + +assert_env mds_HOST ost1_HOST ost2_HOST client_HOST LIVE_CLIENT + +# This can be a regexp, to allow more clients +CLIENTS=${CLIENTS:-"`comma_list $LIVE_CLIENT $FAIL_CLIENTS`"} + +CLIENTLIST="$LIVE_CLIENT $FAIL_CLIENTS" + +DIR=${DIR:-$MOUNT} + +##### +# fail clients round robin + +# list of failable clients +FAIL_LIST=($FAIL_CLIENTS) +FAIL_NUM=${#FAIL_LIST[*]} +FAIL_NEXT=0 +DOWN_NUM=0 # number of nodes currently down + +# return next client to fail +fail_client() { + ret=${FAIL_LIST[$FAIL_NEXT]} + FAIL_NEXT=$(( (FAIL_NEXT+1) % FAIL_NUM )) + echo $ret +} + +shutdown_client() { + client=$1 + if [ "$FAILURE_MODE" = HARD ]; then + $POWER_DOWN $client + elif [ "$FAILURE_MODE" = SOFT ]; then + $PDSH $client $LCONF --clenaup --force --nomod $XMLCONFIG + fi +} + +reboot_node() { + NODE=$1 + if [ "$FAILURE_MODE" = HARD ]; then + $POWER_UP $NODE + fi +} + +fail_clients() { + num=$1 + if [ -z "$num" -o $num -gt $((FAIL_NUM - DOWN_NUM)) ]; then + num=$((FAIL_NUM - DOWN_NUM)) + fi + + if [ -z "$num" -o $num -le 0 ]; then + return + fi + + for i in `seq $num`; do + client=`fail_client` + DOWN_CLIENTS="$DOWN_CLIENTS $client" + client_mkdirs + shutdown_client $client + done + + for client in $DOWN_CLIENTS; do + reboot_node $client + done + DOWN_NUM=`echo $DOWN_CLIENTS | wc -w` + $PDSH $LIVE_CLIENT "cd $MOUNT && rmdir $CLIENTLIST" +} + +reintegrate_clients() { + for client in $DOWN_CLIENTS; do + wait_for_host $client + $PDSH $client "$LCONF --node client --select mds_svc=`facet_active mds` $CLIENTOPTS $XMLCONFIG" + done + DOWN_CLIENTS="" + DOWN_NUM=0 +} + +gen_config() { + rm -f $XMLCONFIG + add_mds mds --dev $MDSDEV --size $MDSSIZE + + if [ ! -z "$mdsfailover_HOST" ]; then + add_mdsfailover mds --dev $MDSDEV --size $MDSSIZE + fi + + add_lov lov1 mds --stripe_sz $STRIPE_BYTES\ + --stripe_cnt $STRIPES_PER_OBJ --stripe_pattern 0 + add_ost ost1 --lov lov1 --dev $OSTDEV --size $OSTSIZE + add_ost ost2 --lov lov1 --dev ${OSTDEV}-2 --size $OSTSIZE + add_client client mds --lov lov1 --path $MOUNT +} + +setup() { + wait_for ost1 + start ost1 ${REFORMAT} $OSTLCONFARGS + wait_for ost2 + start ost2 ${REFORMAT} $OSTLCONFARGS + [ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE + wait_for mds + start mds $MDSLCONFARGS ${REFORMAT} + while ! $PDSH $HOST "ls -ld $LUSTRE"; do sleep 5; done + do_node $CLIENTS lconf --node client_facet \ + --select mds_service=$ACTIVEMDS $XMLCONFIG +} + +cleanup() { + # make sure we are using the primary MDS, so the config log will + # be able to clean up properly. + activemds=`facet_active mds` +# if [ $activemds != "mds" ]; then +# fail mds +# fi + for node in $CLIENTS; do + do_node $node lconf ${FORCE} --select mds_svc=${activemds}_facet --cleanup --node client_facet $XMLCONFIG || true + done + + stop mds ${FORCE} $MDSLCONFARGS + stop ost1 ${FORCE} + stop ost2 ${FORCE} --dump cleanup.log +} + +trap exit INT + +client_mkdirs() { + $PDSH $CLIENTS "mkdir $MOUNT/\`hostname\`; ls $MOUNT/\`hostname\` > /dev/null" +} + +clients_recover_osts() { + facet=$1 + $PDSH $CLIENTS "$LCTL "'--device %OSC_`hostname`_OST_'"${facet}_svc_MNT_client recover" +} + +if [ "$ONLY" == "cleanup" ]; then + cleanup + exit +fi + +gen_config +setup + +if [ "$ONLY" == "setup" ]; then + exit 0 +fi + +# 9 Different Failure Modes Combinations +echo "Starting Test 17 at `date`" + +test_0() { + echo "Failover MDS" + facet_failover mds + wait $DFPID || return 1 + + echo "Failing OST1" + facet_failover ost1 + wait $DFPID || return 2 + + echo "Failing OST2" + facet_failover ost2 + wait $DFPID || return 3 + return 0 +} +run_test 0 "Fail all nodes, independently" + +############### First Failure Mode ############### +test_1() { +echo "Don't do a MDS - MDS Failure Case" +echo "This makes no sense" +# FIXME every test makes sense +} +run_test 1 "MDS/MDS failure" +################################################### + +############### Second Failure Mode ############### +test_2() { + echo "Verify Lustre filesystem is up and running" + client_df + + echo "Failing MDS" + shutdown_facet mds + reboot_facet mds + + # prepare for MDS failover + change_active mds + reboot_facet mds + + client_df & + DFPID=$! + sleep 5 + + echo "Failing OST" + shutdown_facet ost1 + + echo "Reintegrating OST" + reboot_facet ost1 + wait_for ost1 + start ost1 + + echo "Failover MDS" + wait_for mds + start mds + + #Check FS + wait $DFPID + clients_recover_osts ost1 + echo "Verify reintegration" + client_df + +} +run_test 2 "Second Failure Mode: MDS/OST `date`" +################################################### + + +############### Third Failure Mode ############### +test_3() { + #Create files + echo "Verify Lustre filesystem is up and running" + + #MDS Portion + facet_failover mds + wait $DFPID || echo df failed: $? + #Check FS + + echo "Test Lustre stability after MDS failover" + client_df + + #CLIENT Portion + echo "Failing 2 CLIENTS" + fail_clients 2 + + #Check FS + echo "Test Lustre stability after CLIENT failure" + client_df + + #Reintegration + echo "Reintegrating CLIENTS" + reintegrate_clients + + client_df +} +run_test 3 "Thirdb Failure Mode: MDS/CLIENT `date`" +################################################### + +############### Fourth Failure Mode ############### +test_4() { + echo "Fourth Failure Mode: OST/MDS `date`" + + #OST Portion + echo "Failing OST ost1" + shutdown_facet ost1 + + #Check FS + echo "Test Lustre stability after OST failure" + client_df + + #MDS Portion + echo "Failing MDS" + shutdown_facet mds + reboot_facet mds + + # prepare for MDS failover + change_active mds + reboot_facet mds + + client_df & + DFPID=$! + sleep 5 + + #Reintegration + echo "Reintegrating OST" + reboot_facet ost1 + wait_for ost1 + start ost1 + + echo "Failover MDS" + wait_for mds + start mds + #Check FS + + wait $DFPID + clients_recover_osts ost1 + echo "Test Lustre stability after MDS failover" + client_df +} +run_test 4 "Fourth Failure Mode: OST/MDS `date`" +################################################### + +############### Fifth Failure Mode ############### +test_5() { + echo "Fifth Failure Mode: OST/OST `date`" + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + + #OST Portion + echo "Failing OST" + shutdown_facet ost1 + reboot_facet ost1 + + #Check FS + echo "Test Lustre stability after OST failure" + client_df + + #OST Portion + echo "Failing OST" + shutdown_node ost2 + reboot_facet ost2 + + #Check FS + echo "Test Lustre stability after OST failure" + client_df + + #Reintegration + echo "Reintegrating OSTs" + wait_for ost1 + wait_for ost1 + start ost1 + start ost2 + + clients_recover_osts ost1 + clients_recover_osts ost2 + client_df +} +run_test 5 "Fifth Failure Mode: OST/OST `date`" +################################################### + +############### Sixth Failure Mode ############### +test_6() { + echo "Sixth Failure Mode: OST/CLIENT `date`" + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #OST Portion + echo "Failing OST" + shutdown_node ost1 + reboot_facet ost1 + + #Check FS + echo "Test Lustre stability after OST failure" + client_df + + #CLIENT Portion + echo "Failing CLIENTs" + fail_clients + + #Check FS + echo "Test Lustre stability after CLIENTs failure" + client_df + + #Reintegration + echo "Reintegrating OST/CLIENTs" + wait_for ost1 + start ost1 + reintegrate_clients + + echo "Verifying mount" + client_df +} +run_test 6 "Sixth Failure Mode: OST/CLIENT `date`" +################################################### + + +############### Seventh Failure Mode ############### +test_7() { + echo "Seventh Failure Mode: CLIENT/MDS `date`" + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #CLIENT Portion + echo "Part 1: Failing CLIENT" + fail_clients 2 + + #Check FS + echo "Test Lustre stability after CLIENTs failure" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Sleep + echo "Wait 1 minutes" + sleep 60 + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #MDS Portion + echo "Failing MDS" + facet_failover mds + + #Check FS + echo "Test Lustre stability after MDS failover" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Reintegration + echo "Reintegrating CLIENTs" + reintegrate_clients + client_df + + #Sleep + echo "wait 1 minutes" + sleep 60 +} +run_test 7 "Seventh Failure Mode: CLIENT/MDS `date`" +################################################### + + +############### Eighth Failure Mode ############### +test_8() { + echo "Eighth Failure Mode: CLIENT/OST `date`" + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #CLIENT Portion + echo "Failing CLIENTs" + fail_clients 2 + + #Check FS + echo "Test Lustre stability after CLIENTs failure" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Sleep + echo "Wait 1 minutes" + sleep 60 + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #OST Portion + echo "Failing OST" + shutdown_node ost1 + reboot_facet ost1 + + #Check FS + echo "Test Lustre stability after OST failure" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Reintegration + echo "Reintegrating CLIENTs/OST" + reintegrate_clients + start ost1 + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/CLIENT_OST_2\`hostname\`_testfile" + + #Sleep + echo "Wait 1 minutes" + sleep 60 +} +run_test 8 "Eighth Failure Mode: CLIENT/OST `date`" +################################################### + + +############### Ninth Failure Mode ############### +test_9() { + echo + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #CLIENT Portion + echo "Failing CLIENTs" + fail_clients 2 + + #Check FS + echo "Test Lustre stability after CLIENTs failure" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Sleep + echo "Wait 1 minutes" + sleep 60 + + #Create files + echo "Verify Lustre filesystem is up and running" + client_df + $PDSH $CLIENTS "/bin/touch $MOUNT/\`hostname\`_testfile" + + #CLIENT Portion + echo "Failing CLIENTs" + fail_clients 2 + + #Check FS + echo "Test Lustre stability after CLIENTs failure" + client_df + $PDSH $LIVE_CLIENT "ls -l $MOUNT" + $PDSH $LIVE_CLIENT "rm -f $MOUNT/*_testfile" + + #Reintegration + echo "Reintegrating CLIENTs/CLIENTs" + reintegrate_clients + client_df + + #Sleep + echo "Wait 1 minutes" + sleep 60 +} +run_test 9 "Ninth Failure Mode: CLIENT/CLIENT `date`" +################################################### + +test_10() { + #Run availability after all failures + ./availability.sh 21600 +} +run_test 10 "Running Availability for 6 hours..." + +equals_msg "Done, cleaning up" +cleanup diff --git a/lustre/utils/lfs.c b/lustre/utils/lfs.c index b22745a..7339242 100644 --- a/lustre/utils/lfs.c +++ b/lustre/utils/lfs.c @@ -28,21 +28,28 @@ #include #include #include +#include +#include +#include #include #include #include "parser.h" +#include "obdctl.h" extern int op_create_file(char *name, long stripe_size, int stripe_offset, int stripe_count); extern int op_find(char *path, struct obd_uuid *obduuid, int recursive, int verbose, int quiet); +extern int op_check(int type_num, char **obd_type_p, char *dir); /* all functions */ static int lfs_setstripe(int argc, char **argv); static int lfs_find(int argc, char **argv); static int lfs_getstripe(int argc, char **argv); +static int lfs_osts(int argc, char **argv); +static int lfs_check(int argc, char **argv); /* all avaialable commands */ command_t cmdlist[] = { @@ -58,6 +65,10 @@ command_t cmdlist[] = { {"getstripe", lfs_getstripe, 0, "blah...\n" "usage:getstripe "}, + {"check", lfs_check, 0, + "blah...\n" + "usage: check "}, + {"osts", lfs_osts, 0, "osts"}, {"help", Parser_help, 0, "help"}, {"exit", Parser_quit, 0, "quit"}, {"quit", Parser_quit, 0, "quit"}, @@ -187,12 +198,104 @@ static int lfs_getstripe(int argc, char **argv) return rc; } +static int lfs_osts(int argc, char **argv) +{ + FILE *fp; + struct mntent *mnt = NULL; + struct obd_uuid *obduuid = NULL; + int rc=0; + + if (argc != 1) + return CMD_HELP; + + fp = setmntent(MOUNTED, "r"); + + if (fp == NULL) { + fprintf(stderr, "setmntent(%s): %s:", MOUNTED, + strerror (errno)); + } else { + mnt = getmntent(fp); + while (feof(fp) == 0 && ferror(fp) ==0) { + if (strcmp(mnt->mnt_type, "lustre_lite") == 0) { + rc = op_find(mnt->mnt_dir, obduuid, 0, 0, 0); + if (rc) + fprintf(stderr, "error: lfs osts failed for %s\n", + mnt->mnt_dir); + } + mnt = getmntent(fp); + } + endmntent(fp); + } + + return rc; +} + +static int lfs_check(int argc, char **argv) +{ + int rc; + FILE *fp; + struct mntent *mnt = NULL; + int type_num = 1; + char *obd_type_p[2]; + char obd_type1[4]; + char obd_type2[4]; + + if (argc != 2) + return CMD_HELP; + + obd_type_p[1]=obd_type1; + obd_type_p[2]=obd_type2; + + if (strcmp(argv[1],"osts")==0) { + strcpy(obd_type_p[0],"osc"); + } else if (strcmp(argv[1],"mds")==0) { + strcpy(obd_type_p[0],"mdc"); + } else if (strcmp(argv[1],"servers")==0) { + type_num=2; + strcpy(obd_type_p[0],"osc"); + strcpy(obd_type_p[1],"mdc"); + } else { + fprintf(stderr, "error: %s: option '%s' unrecognized\n", + argv[0], argv[1]); + return CMD_HELP; + } + + fp = setmntent(MOUNTED, "r"); + if (fp == NULL) { + fprintf(stderr, "setmntent(%s): %s:", MOUNTED, + strerror (errno)); + } else { + mnt = getmntent(fp); + while (feof(fp) == 0 && ferror(fp) ==0) { + if (strcmp(mnt->mnt_type, "lustre_lite") == 0) + break; + mnt = getmntent(fp); + } + endmntent(fp); + } + + rc = op_check(type_num,obd_type_p,mnt->mnt_dir); + + if (rc) + fprintf(stderr, "error: %s: %s status failed\n", + argv[0],argv[1]); + + return rc; + +} + int main(int argc, char **argv) { int rc; setlinebuf(stdout); - + + ptl_initialize(argc, argv); + if (obd_initialize(argc, argv) < 0) + exit(2); + if (dbg_initialize(argc, argv) < 0) + exit(3); + Parser_init("lfs > ", cmdlist); if (argc > 1) { @@ -201,5 +304,6 @@ int main(int argc, char **argv) rc = Parser_commands(); } + obd_cleanup(argc, argv); return rc; } diff --git a/lustre/utils/liblustreapi.c b/lustre/utils/liblustreapi.c index 2ae9eb8..7ea801c 100644 --- a/lustre/utils/liblustreapi.c +++ b/lustre/utils/liblustreapi.c @@ -29,6 +29,8 @@ #include #include +#include +#include #include #include #include @@ -46,6 +48,8 @@ #include #include +#include + static void err_msg(char *fmt, ...) { va_list args; @@ -448,3 +452,82 @@ out: return ret; } +#define MAX_STRING_SIZE 128 + +int op_check(int type_num, char **obd_type, char *dir) +{ + int rc=0; + int i=0,j=0,k; + char buf[OBD_MAX_IOCTL_BUFFER]; + char *buf2; + struct obd_ioctl_data *data = (struct obd_ioctl_data *)buf; + + memset(buf, 0, sizeof(buf)); + data->ioc_version = OBD_IOCTL_VERSION; + data->ioc_inllen1 = sizeof(buf) - size_round(sizeof(*data)); + data->ioc_len = obd_ioctl_packlen(data); + + rc = l_ioctl(OBD_DEV_ID, OBD_IOC_LIST, data); + + buf2 = data->ioc_bulk; + + if (!data->ioc_inlbuf1) { + err_msg("No buffer passed!\n"); + rc = errno; + } + + do { + char status[3]; + char obd_type_name[sizeof(struct obd_type)]; + char obd_name[MAX_STRING_SIZE]; + char obd_uuid[sizeof(struct obd_uuid)]; + int obd_type_refcnt; + + char rawbuf[OBD_MAX_IOCTL_BUFFER]; + char *bufl = rawbuf; + int max = sizeof(rawbuf); + struct obd_ioctl_data datal; + struct obd_statfs osfs_buffer; + + memset (&osfs_buffer, 0, sizeof (osfs_buffer)); + + memset(bufl, 0, sizeof(rawbuf)); + datal.ioc_pbuf1 = (char *)&osfs_buffer; + datal.ioc_plen1 = sizeof (osfs_buffer); + + j = sscanf(buf2,"%d %s %s %s %s %d",&j, + status,obd_type_name, + obd_name, obd_uuid, + &obd_type_refcnt); + + if (j != 6) break; + + for (k=0;k> "$LMC_BATCH_FILE" } # following user input to create xml config file create_config() { - local config_file=$1 - - for device in $DEVICE_LIST ; do - get_name_in_list $device - echo -n " $DEVICE_NAME" - case $DEVICE_NAME in - mds*) - add_node "$HOST_NAME" "$config_file" || return 1 - $LMC -m "$config_file" --add mds --node "$HOST_NAME" \ - --mds "$DEVICE_NAME" \ - --nid "$HOST_NAME" --fstype "$DEFAULT_FSTYPE" \ - --dev "$DEVICE" || return 1 - ;; - lov*) - $LMC -m "$config_file" --add lov \ - --lov "$DEVICE_NAME" \ - --mds "$DEVICE_MDS" \ - --stripe_sz "$STRIPE_SIZE" \ - --stripe_cnt "$STRIPE_CNT" \ - --stripe_pattern "$STRIPE_PATTERN" || return 1 - ;; - ost*) - add_node "$HOST_NAME" "$config_file" || return 1 - $LMC -m "$config_file" --add ost --node "$HOST_NAME" \ - --ost "$DEVICE_NAME" \ - --lov "$DEVICE_LOV" --fstype "$DEFAULT_FSTYPE" \ - --dev "$DEVICE" || return 1 - ;; - client*) - add_node "$DEVICE_NAME" "$config_file" || return 1 - $LMC -m "$config_file" --add mtpt \ - --node "$DEVICE_NAME" \ - --mds "$DEVICE_MDS" \ - --lov "$DEVICE_LOV" --path "$DEVICE" || return 1 - ;; - esac - done - echo - return 0 + for device in $DEVICE_LIST ; do + get_name_in_list $device + echo -n " $DEVICE_NAME" + case $DEVICE_NAME in + mds*) + add_node "$HOST_NAME" + run_lmc --add mds \ + --node "$HOST_NAME" \ + --mds "$DEVICE_NAME" \ + --fstype "$DEFAULT_FSTYPE" \ + --dev "$DEVICE" \ + --size "$DEVICE_SIZE" + ;; + lov*) + run_lmc --add lov \ + --lov "$DEVICE_NAME" \ + --mds "$DEVICE_MDS" \ + --stripe_sz "$STRIPE_SIZE" \ + --stripe_cnt "$STRIPE_CNT" \ + --stripe_pattern "$STRIPE_PATTERN" + ;; + ost*) + add_node "$HOST_NAME" + run_lmc --add ost \ + --node "$HOST_NAME" \ + --lov "$DEVICE_LOV" \ + --fstype "$DEFAULT_FSTYPE" \ + --dev "$DEVICE" \ + --size "$DEVICE_SIZE" + ;; + client*) + add_node "$DEVICE_NAME" + run_lmc --add mtpt \ + --node "$DEVICE_NAME" \ + --mds "$DEVICE_MDS" \ + --lov "$DEVICE_LOV" \ + --path "$DEVICE" + ;; + esac + done + echo + return 0 +} + +maybe_clean() +{ + [ -f "$1" ] || return 0 + if ! (( $FORCE )) ; then + echo -n "${0##*/}: overwrite existing $2 \"$1\"? " + read answer + if ! [ "${answer:0:1}" = "y" -o "${answer:0:1}" = "Y" ] ; then + echo "(${0##*/}: (Exiting.)" + exit 0 + fi + fi + rm -f "$1" } -#parse options +# parse options get_option "$@" -#some default definitions +# some default definitions LMC=${LMC:-"/usr/sbin/lmc"} -CONFIG_FILE=${CONFIG_FILE:-"config.xml"} -TMP_CONFIG_FILE=${TMP_CONFIG_FILE:-".config.xml.tmp"} + +CONFIG_FILE=${CONFIG_FILE:-"lwizard.xml"} + +# Remove exiting files. + +maybe_clean "$CONFIG_FILE" "Lustre configuration file" +if [ "$LMC_BATCH_FILE" ] ; then + maybe_clean "$LMC_BATCH_FILE" "lmc batch file" +else + LMC_BATCH_FILE=$(mktemp -q "/tmp/${CONFIG_FILE##*/}.XXXXXX") + [ $? -eq 0 ] || fatal 1 "Couldn't create temporary batch file." +fi + DEFAULT_FSTYPE=${DEFAULT_FSTYPE:-"ext3"} DEFAULT_NETTYPE=${DEFAULT_NETTYPE:-"tcp"} DEFAULT_MNTPT=${DEFAULT_MNTPT:-"/mnt/lustre"} -STRIPE_SIZE=${STRIPE_SIZE:-65536} + +STRIPE_SIZE=${STRIPE_SIZE:-$((1 * 1024 * 1024))} STRIPE_CNT=${STRIPE_CNT:-1} STRIPE_PATTERN=${STRIPE_PATTERN:-0} -ANSWER="yes no" - -#print program information -cat <