# endif
#endif
+#ifndef LP_POISON
+# define LI_POISON ((int)0x5a5a5a5a5a5a5a5a)
+# define LL_POISON ((long)0x5a5a5a5a5a5a5a5a)
+# define LP_POISON ((void *)(long)0x5a5a5a5a5a5a5a5a)
+#endif
+
#if defined(__x86_64__)
# define LPU64 "%Lu"
# define LPD64 "%Ld"
# define LPX64 "%#Lx"
# define LPSZ "%lu"
# define LPSSZ "%ld"
-# define LP_POISON ((void *)0x5a5a5a5a5a5a5a5a)
#elif (BITS_PER_LONG == 32 || __WORDSIZE == 32)
# define LPU64 "%Lu"
# define LPD64 "%Ld"
# define LPX64 "%#Lx"
# define LPSZ "%u"
# define LPSSZ "%d"
-# define LP_POISON ((void *)0x5a5a5a5a)
#elif (BITS_PER_LONG == 64 || __WORDSIZE == 64)
# define LPU64 "%lu"
# define LPD64 "%ld"
# define LPX64 "%#lx"
# define LPSZ "%lu"
# define LPSSZ "%ld"
-# define LP_POISON ((void *)0x5a5a5a5a5a5a5a5a)
#endif
#ifndef LPU64
# error "No word size defined"
- deal with strange write() on x86-64 (3043)
- don't dereference NULL peer_ni in ldlm_handle_ast_error (3258)
- clear page->private before handing to FS (3119)
- - drop scimac NAL
+ - tune the read pipeline (3236)
+ * miscellania
+ - drop scimac NAL (unmaintained)
tbd Cluster File Systems, Inc. <info@clusterfs.com>
* version 1.2.2
- initialize RPC timeout timer earlier for 2.6 (3219)
- don't dereference NULL reply buffer if mdc_close was never sent (2410)
- print nal/nid for unknown nid (3258)
+ - additional checks for oscc recovery before doing precreate (3284)
+ - fix ll_extent_lock() error return code for 64-bit systems (3043)
+ - don't crash in mdc_close for bad permissions on open (3285)
* miscellania
- allow default OST striping configuration per directory (1414)
- increase maximum number of MDS request buffers for large systems
#define PTR_ERR(a) ((long)(a))
#define ERR_PTR(a) ((void*)((long)(a)))
+#define capable(foo) 1
+#define CAP_SYS_ADMIN 1
+
typedef struct {
void *cwd;
}mm_segment_t;
int pid;
int fsuid;
int fsgid;
- int max_groups;
- int ngroups;
- gid_t *groups;
__u32 cap_effective;
-
- struct fs_struct __fs;
};
extern struct task_struct *current;
-int in_group_p(gid_t gid);
-static inline int capable(int cap)
-{
- if (current->cap_effective & (1 << cap))
- return 1;
- else
- return 0;
-}
+
+#define in_group_p(a) 0 /* FIXME */
#define set_current_state(foo) do { current->state = foo; } while (0)
unsigned long now = jiffies;
void *parent_handle = oti ? oti->oti_handle : NULL;
void *handle = obd->obd_fsops->fs_start(inode, op, parent_handle, logs);
- CDEBUG(D_HA, "started handle %p (%p)\n", handle, parent_handle);
+ CDEBUG(D_INFO, "started handle %p (%p)\n", handle, parent_handle);
if (oti != NULL) {
if (parent_handle == NULL) {
{
unsigned long now = jiffies;
int rc = obd->obd_fsops->fs_commit(inode, handle, force_sync);
- CDEBUG(D_HA, "committing handle %p\n", handle);
+ CDEBUG(D_INFO, "committing handle %p\n", handle);
if (time_after(jiffies, now + 15 * HZ))
CERROR("long journal start time %lus\n", (jiffies - now) / HZ);
#include <linux/lustre_idl.h>
#include <linux/lustre_cfg.h>
+#define LI_POISON ((int)0x5a5a5a5a5a5a5a5a)
+#define LP_POISON ((void *)(long)0x5a5a5a5a5a5a5a5a)
+
#ifndef LPU64
/* x86_64 has 64bit longs and defines u64 as long long */
#if BITS_PER_LONG > 32 && !defined(__x86_64__)
#define LPU64 "%lu"
#define LPD64 "%ld"
#define LPX64 "%#lx"
-#define LP_POISON ((void *)0x5a5a5a5a5a5a5a5a)
#else
#define LPU64 "%Lu"
#define LPD64 "%Ld"
#define LPX64 "%#Lx"
-#define LP_POISON ((void *)0x5a5a5a5a)
#endif
#endif
# define PTLRPC_MAX_BRW_PAGES (PTLRPC_MAX_BRW_SIZE / PAGE_SIZE)
#endif
+#if ((PTLRPC_MAX_BRW_PAGES & (PTLRPC_MAX_BRW_PAGES - 1)) != 0)
+#error "PTLRPC_MAX_BRW_PAGES isn't a power of two"
+#endif
/* Size over which to OBD_VMALLOC() rather than OBD_ALLOC() service request
* buffers */
--- /dev/null
+Index: linux-2.4.21/arch/i386/kernel/entry.S
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/entry.S 2004-04-24 02:39:01.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/entry.S 2004-04-24 02:42:58.000000000 -0400
+@@ -45,6 +45,7 @@
+ #include <linux/linkage.h>
+ #include <asm/segment.h>
+ #include <asm/smp.h>
++#include <asm/current.h>
+
+ EBX = 0x00
+ ECX = 0x04
+@@ -130,10 +131,6 @@
+ .long 3b,6b; \
+ .previous
+
+-#define GET_CURRENT(reg) \
+- movl $-8192, reg; \
+- andl %esp, reg
+-
+ ENTRY(lcall7)
+ pushfl # We get a different stack layout with call gates,
+ pushl %eax # which has to be cleaned up later..
+@@ -149,7 +146,7 @@
+ movl %ecx,CS(%esp) #
+ movl %esp,%ebx
+ pushl %ebx
+- andl $-8192,%ebx # GET_CURRENT
++ andl $-THREAD_SIZE,%ebx # GET_CURRENT
+ movl exec_domain(%ebx),%edx # Get the execution domain
+ movl 4(%edx),%edx # Get the lcall7 handler for the domain
+ pushl $0x7
+@@ -173,7 +170,7 @@
+ movl %ecx,CS(%esp) #
+ movl %esp,%ebx
+ pushl %ebx
+- andl $-8192,%ebx # GET_CURRENT
++ andl $-THREAD_SIZE,%ebx # GET_CURRENT
+ movl exec_domain(%ebx),%edx # Get the execution domain
+ movl 4(%edx),%edx # Get the lcall7 handler for the domain
+ pushl $0x27
+Index: linux-2.4.21/arch/i386/kernel/smpboot.c
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/smpboot.c 2004-04-24 02:39:05.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/smpboot.c 2004-04-24 02:42:58.000000000 -0400
+@@ -837,7 +837,7 @@
+
+ /* So we see what's up */
+ printk("Booting processor %d/%d eip %lx\n", cpu, apicid, start_eip);
+- stack_start.esp = (void *) (1024 + PAGE_SIZE + (char *)idle);
++ stack_start.esp = (void *)idle->thread.esp;
+
+ /*
+ * This grunge runs the startup process for
+@@ -918,7 +918,7 @@
+ Dprintk("CPU has booted.\n");
+ } else {
+ boot_error= 1;
+- if (*((volatile unsigned char *)phys_to_virt(8192))
++ if (*((volatile unsigned char *)phys_to_virt(THREAD_SIZE))
+ == 0xA5)
+ /* trampoline started but...? */
+ printk("Stuck ??\n");
+@@ -941,7 +941,7 @@
+ }
+
+ /* mark "stuck" area as not stuck */
+- *((volatile unsigned long *)phys_to_virt(8192)) = 0;
++ *((volatile unsigned long *)phys_to_virt(THREAD_SIZE)) = 0;
+
+ #ifdef CONFIG_ES7000
+ if (!es7000_plat)
+Index: linux-2.4.21/arch/i386/kernel/traps.c
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/traps.c 2004-04-24 02:39:18.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/traps.c 2004-04-24 02:42:58.000000000 -0400
+@@ -304,7 +304,7 @@
+ unsigned long esp = tsk->thread.esp;
+
+ /* User space on another CPU? */
+- if ((esp ^ (unsigned long)tsk) & (PAGE_MASK<<1))
++ if ((esp ^ (unsigned long)tsk) & ~(THREAD_SIZE - 1))
+ return;
+ show_trace((unsigned long *)esp);
+ }
+Index: linux-2.4.21/arch/i386/kernel/head.S
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/head.S 2004-04-24 02:38:42.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/head.S 2004-04-24 02:42:58.000000000 -0400
+@@ -15,6 +15,7 @@
+ #include <asm/page.h>
+ #include <asm/pgtable.h>
+ #include <asm/desc.h>
++#include <asm/current.h>
+
+ #define OLD_CL_MAGIC_ADDR 0x90020
+ #define OLD_CL_MAGIC 0xA33F
+@@ -326,7 +327,7 @@
+ ret
+
+ ENTRY(stack_start)
+- .long SYMBOL_NAME(init_task_union)+8192
++ .long SYMBOL_NAME(init_task_union)+THREAD_SIZE
+ .long __KERNEL_DS
+
+ /* This is the default interrupt "handler" :-) */
+Index: linux-2.4.21/arch/i386/kernel/irq.c
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/irq.c 2004-04-24 02:39:18.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/irq.c 2004-04-24 02:44:26.000000000 -0400
+@@ -602,7 +602,10 @@
+ long esp;
+
+ /* Debugging check for stack overflow: is there less than 1KB free? */
+- __asm__ __volatile__("andl %%esp,%0" : "=r" (esp) : "0" (8191));
++ __asm__ __volatile__(
++ "andl %%esp,%0"
++ : "=r" (esp) : "0" (THREAD_SIZE-1));
++
+ if (unlikely(esp < (sizeof(struct task_struct) + sysctl_stackwarn))) {
+ static unsigned long next_jiffies; /* ratelimiting */
+ static long least_esp = THREAD_SIZE;
+Index: linux-2.4.21/arch/i386/lib/getuser.S
+===================================================================
+--- linux-2.4.21.orig/arch/i386/lib/getuser.S 1998-01-12 16:42:52.000000000 -0500
++++ linux-2.4.21/arch/i386/lib/getuser.S 2004-04-24 02:42:58.000000000 -0400
+@@ -21,6 +21,10 @@
+ * as they get called from within inline assembly.
+ */
+
++/* Duplicated from asm/processor.h */
++#include <asm/current.h>
++#include <linux/config.h>
++
+ addr_limit = 12
+
+ .text
+@@ -28,7 +32,7 @@
+ .globl __get_user_1
+ __get_user_1:
+ movl %esp,%edx
+- andl $0xffffe000,%edx
++ andl $~(THREAD_SIZE - 1),%edx
+ cmpl addr_limit(%edx),%eax
+ jae bad_get_user
+ 1: movzbl (%eax),%edx
+@@ -41,7 +45,7 @@
+ addl $1,%eax
+ movl %esp,%edx
+ jc bad_get_user
+- andl $0xffffe000,%edx
++ andl $~(THREAD_SIZE - 1),%edx
+ cmpl addr_limit(%edx),%eax
+ jae bad_get_user
+ 2: movzwl -1(%eax),%edx
+@@ -54,7 +58,7 @@
+ addl $3,%eax
+ movl %esp,%edx
+ jc bad_get_user
+- andl $0xffffe000,%edx
++ andl $~(THREAD_SIZE - 1),%edx
+ cmpl addr_limit(%edx),%eax
+ jae bad_get_user
+ 3: movl -3(%eax),%edx
+Index: linux-2.4.21/arch/i386/config.in
+===================================================================
+--- linux-2.4.21.orig/arch/i386/config.in 2004-04-24 02:39:21.000000000 -0400
++++ linux-2.4.21/arch/i386/config.in 2004-04-24 02:42:58.000000000 -0400
+@@ -326,6 +326,29 @@
+ if [ "$CONFIG_SMP" = "y" -a "$CONFIG_X86_CMPXCHG" = "y" ]; then
+ define_bool CONFIG_HAVE_DEC_LOCK y
+ fi
++
++choice 'Bigger Stack Size Support' \
++ "off CONFIG_NOBIGSTACK \
++ 16KB CONFIG_STACK_SIZE_16KB \
++ 32KB CONFIG_STACK_SIZE_32KB \
++ 64KB CONFIG_STACK_SIZE_64KB" off
++
++if [ "$CONFIG_NOBIGSTACK" = "y" ]; then
++ define_int CONFIG_STACK_SIZE_SHIFT 1
++else
++ if [ "$CONFIG_STACK_SIZE_16KB" = "y" ]; then
++ define_int CONFIG_STACK_SIZE_SHIFT 2
++ else
++ if [ "$CONFIG_STACK_SIZE_32KB" = "y" ]; then
++ define_int CONFIG_STACK_SIZE_SHIFT 3
++ else
++ if [ "$CONFIG_STACK_SIZE_64KB" = "y" ]; then
++ define_int CONFIG_STACK_SIZE_SHIFT 4
++ fi
++ fi
++ fi
++fi
++
+ endmenu
+
+ mainmenu_option next_comment
+Index: linux-2.4.21/arch/i386/vmlinux.lds.S
+===================================================================
+--- linux-2.4.21.orig/arch/i386/vmlinux.lds.S 2004-04-24 02:38:06.000000000 -0400
++++ linux-2.4.21/arch/i386/vmlinux.lds.S 2004-04-24 02:42:58.000000000 -0400
+@@ -39,7 +39,8 @@
+
+ _edata = .; /* End of data section */
+
+- . = ALIGN(8192); /* init_task */
++/* chose the biggest of the possible stack sizes here? */
++ . = ALIGN(65536); /* init_task */
+ .data.init_task : { *(.data.init_task) }
+
+ . = ALIGN(4096); /* Init code and data */
+Index: linux-2.4.21/include/asm-i386/current.h
+===================================================================
+--- linux-2.4.21.orig/include/asm-i386/current.h 1998-08-14 19:35:22.000000000 -0400
++++ linux-2.4.21/include/asm-i386/current.h 2004-04-24 02:42:58.000000000 -0400
+@@ -1,15 +1,43 @@
+ #ifndef _I386_CURRENT_H
+ #define _I386_CURRENT_H
++#include <asm/page.h>
++
++/*
++ * Configurable page sizes on i386, mainly for debugging purposes.
++ * (c) Balbir Singh
++ */
++
++#ifdef __ASSEMBLY__
++
++#define PAGE_SIZE 4096 /* as cannot handle 1UL << 12 */
++#define THREAD_SIZE ((1 << CONFIG_STACK_SIZE_SHIFT) * PAGE_SIZE)
++
++#define GET_CURRENT(reg) \
++ movl $-THREAD_SIZE, reg; \
++ andl %esp, reg
++
++#else /* __ASSEMBLY__ */
++
++#define THREAD_SIZE ((1 << CONFIG_STACK_SIZE_SHIFT) * PAGE_SIZE)
++#define alloc_task_struct() \
++ ((struct task_struct *) __get_free_pages(GFP_KERNEL,CONFIG_STACK_SIZE_SHIFT))
++
++#define free_task_struct(p) \
++ free_pages((unsigned long) (p), CONFIG_STACK_SIZE_SHIFT)
++
++#define INIT_TASK_SIZE THREAD_SIZE
+
+ struct task_struct;
+
+ static inline struct task_struct * get_current(void)
+ {
+ struct task_struct *current;
+- __asm__("andl %%esp,%0; ":"=r" (current) : "0" (~8191UL));
++ __asm__("andl %%esp,%0; ":"=r" (current) : "0" (~(THREAD_SIZE - 1)));
+ return current;
+ }
+
+ #define current get_current()
+
++#endif /* __ASSEMBLY__ */
++
+ #endif /* !(_I386_CURRENT_H) */
+Index: linux-2.4.21/include/asm-i386/hw_irq.h
+===================================================================
+--- linux-2.4.21.orig/include/asm-i386/hw_irq.h 2004-04-24 02:39:05.000000000 -0400
++++ linux-2.4.21/include/asm-i386/hw_irq.h 2004-04-24 02:42:58.000000000 -0400
+@@ -16,6 +16,7 @@
+ #include <linux/sched.h>
+ #include <asm/atomic.h>
+ #include <asm/irq.h>
++#include <asm/current.h>
+
+ /*
+ * IDT vectors usable for external interrupt sources start
+@@ -120,10 +121,6 @@
+ #define IRQ_NAME2(nr) nr##_interrupt(void)
+ #define IRQ_NAME(nr) IRQ_NAME2(IRQ##nr)
+
+-#define GET_CURRENT \
+- "movl %esp, %ebx\n\t" \
+- "andl $-8192, %ebx\n\t"
+-
+ /*
+ * SMP has a few special interrupts for IPI messages
+ */
+Index: linux-2.4.21/include/asm-i386/processor.h
+===================================================================
+--- linux-2.4.21.orig/include/asm-i386/processor.h 2004-04-24 02:39:21.000000000 -0400
++++ linux-2.4.21/include/asm-i386/processor.h 2004-04-24 02:42:58.000000000 -0400
+@@ -14,6 +14,7 @@
+ #include <asm/types.h>
+ #include <asm/sigcontext.h>
+ #include <asm/cpufeature.h>
++#include <asm/current.h>
+ #include <linux/cache.h>
+ #include <linux/config.h>
+ #include <linux/threads.h>
+@@ -466,9 +467,6 @@
+ #define KSTK_EIP(tsk) (((unsigned long *)(4096+(unsigned long)(tsk)))[1019])
+ #define KSTK_ESP(tsk) (((unsigned long *)(4096+(unsigned long)(tsk)))[1022])
+
+-#define THREAD_SIZE (2*PAGE_SIZE)
+-#define alloc_task_struct() ((struct task_struct *) __get_free_pages(GFP_KERNEL,1))
+-#define free_task_struct(p) free_pages((unsigned long) (p), 1)
+ #define get_task_struct(tsk) atomic_inc(&virt_to_page(tsk)->count)
+
+ #define init_task (init_task_union.task)
+Index: linux-2.4.21/include/linux/sched.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/sched.h 2004-04-24 02:39:20.000000000 -0400
++++ linux-2.4.21/include/linux/sched.h 2004-04-24 02:42:58.000000000 -0400
+@@ -2,6 +2,7 @@
+ #define _LINUX_SCHED_H
+
+ #include <asm/param.h> /* for HZ */
++#include <asm/current.h> /* maybe for INIT_TASK_SIZE */
+
+ extern unsigned long event;
+
+Index: linux-2.4.21/include/asm-x86_64/current.h
+===================================================================
+--- linux-2.4.21.orig/include/asm-x86_64/current.h 2002-11-28 18:53:15.000000000 -0500
++++ linux-2.4.21/include/asm-x86_64/current.h 2004-04-24 02:42:58.000000000 -0400
+@@ -5,6 +5,7 @@
+ struct task_struct;
+
+ #include <asm/pda.h>
++#include <asm/page.h>
+
+ static inline struct task_struct *get_current(void)
+ {
--- /dev/null
+ drivers/block/blkpg.c | 36 ++++++++++++++++++++++++++++++++++++
+ drivers/block/loop.c | 3 +++
+ drivers/ide/ide-disk.c | 4 ++++
+ 3 files changed, 43 insertions(+)
+
+Index: linux-2.4.19/drivers/block/blkpg.c
+===================================================================
+--- linux-2.4.19.orig/drivers/block/blkpg.c 2002-08-02 20:39:43.000000000 -0400
++++ linux-2.4.19/drivers/block/blkpg.c 2004-04-23 18:24:40.000000000 -0400
+@@ -296,3 +296,37 @@
+ }
+
+ EXPORT_SYMBOL(blk_ioctl);
++
++#define NUM_DEV_NO_WRITE 16
++static int dev_no_write[NUM_DEV_NO_WRITE];
++/*
++ * Debug code for turning block devices "read-only" (will discard writes
++ * silently). This is for filesystem crash/recovery testing.
++ */
++void dev_set_rdonly(kdev_t dev, int no_write)
++{
++ if (dev) {
++ printk(KERN_WARNING "Turning device %s read-only\n",
++ bdevname(dev));
++ dev_no_write[no_write] = 0xdead0000 + dev;
++ }
++}
++
++int dev_check_rdonly(kdev_t dev) {
++ int i;
++
++ for (i = 0; i < NUM_DEV_NO_WRITE; i++) {
++ if ((dev_no_write[i] & 0xffff0000) == 0xdead0000 &&
++ dev == (dev_no_write[i] & 0xffff))
++ return 1;
++ }
++ return 0;
++}
++
++void dev_clear_rdonly(int no_write) {
++ dev_no_write[no_write] = 0;
++}
++
++EXPORT_SYMBOL(dev_set_rdonly);
++EXPORT_SYMBOL(dev_check_rdonly);
++EXPORT_SYMBOL(dev_clear_rdonly);
+Index: linux-2.4.19/drivers/block/loop.c
+===================================================================
+--- linux-2.4.19.orig/drivers/block/loop.c 2004-04-23 17:53:56.000000000 -0400
++++ linux-2.4.19/drivers/block/loop.c 2004-04-23 18:23:16.000000000 -0400
+@@ -478,6 +478,9 @@
+ spin_unlock_irq(&lo->lo_lock);
+
+ if (rw == WRITE) {
++ if (dev_check_rdonly(rbh->b_rdev))
++ goto err;
++
+ if (lo->lo_flags & LO_FLAGS_READ_ONLY)
+ goto err;
+ } else if (rw == READA) {
+Index: linux-2.4.19/drivers/ide/ide-disk.c
+===================================================================
+--- linux-2.4.19.orig/drivers/ide/ide-disk.c 2004-04-23 17:53:51.000000000 -0400
++++ linux-2.4.19/drivers/ide/ide-disk.c 2004-04-23 18:23:16.000000000 -0400
+@@ -558,6 +558,10 @@
+ */
+ static ide_startstop_t do_rw_disk (ide_drive_t *drive, struct request *rq, unsigned long block)
+ {
++ if (rq->cmd == WRITE && dev_check_rdonly(rq->rq_dev)) {
++ ide_end_request(1, HWGROUP(drive));
++ return ide_stopped;
++ }
+ if (IDE_CONTROL_REG)
+ OUT_BYTE(drive->ctl,IDE_CONTROL_REG);
+
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
+Index: linux-2.4.19/fs/ext3/Makefile
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/Makefile 2004-04-23 17:53:55.000000000 -0400
++++ linux-2.4.19/fs/ext3/Makefile 2004-04-23 18:25:03.000000000 -0400
@@ -9,6 +9,8 @@
O_TARGET := ext3.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);
+Index: linux-2.4.19/fs/ext3/super.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/super.c 2004-04-23 17:53:55.000000000 -0400
++++ linux-2.4.19/fs/ext3/super.c 2004-04-23 18:25:03.000000000 -0400
+@@ -1821,7 +1821,7 @@
+ exit_ext3_xattr();
}
-EXPORT_NO_SYMBOLS;
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
+Index: linux-2.4.19/include/linux/fs.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/fs.h 2004-04-23 17:54:14.000000000 -0400
++++ linux-2.4.19/include/linux/fs.h 2004-04-23 18:25:27.000000000 -0400
+@@ -1183,6 +1183,7 @@
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(const char *type, int flags, char *name, void *data);
+ extern void umount_tree(struct vfsmount *);
#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);
+Index: linux-2.4.19/kernel/ksyms.c
+===================================================================
+--- linux-2.4.19.orig/kernel/ksyms.c 2004-04-23 17:54:14.000000000 -0400
++++ linux-2.4.19/kernel/ksyms.c 2004-04-23 18:25:03.000000000 -0400
+@@ -330,6 +330,10 @@
EXPORT_SYMBOL(dcache_readdir);
EXPORT_SYMBOL(dcache_dir_ops);
/* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */
EXPORT_SYMBOL(default_llseek);
EXPORT_SYMBOL(dentry_open);
-
-_
--- /dev/null
+ fs/ext3/Makefile | 2
+ fs/ext3/dir.c | 299 +++++++++
+ fs/ext3/file.c | 3
+ fs/ext3/hash.c | 215 ++++++
+ fs/ext3/namei.c | 1388 ++++++++++++++++++++++++++++++++++++++++-----
+ 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, 1887 insertions(+), 160 deletions(-)
+
+Index: linux-2.4.19/fs/ext3/Makefile
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/Makefile 2004-04-23 18:25:03.000000000 -0400
++++ linux-2.4.19/fs/ext3/Makefile 2004-04-23 18:26:27.000000000 -0400
+@@ -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 := $(O_TARGET)
+
+ obj-$(CONFIG_EXT3_FS_XATTR) += xattr.o
+Index: linux-2.4.19/fs/ext3/dir.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/dir.c 2001-11-09 17:25:04.000000000 -0500
++++ linux-2.4.19/fs/ext3/dir.c 2004-04-23 18:26:27.000000000 -0400
+@@ -21,12 +21,16 @@
+ #include <linux/fs.h>
+ #include <linux/jbd.h>
+ #include <linux/ext3_fs.h>
++#include <linux/slab.h>
++#include <linux/rbtree.h>
+
+ 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 +207,269 @@
+ 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
+Index: linux-2.4.19/fs/ext3/file.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/file.c 2004-04-23 17:54:02.000000000 -0400
++++ linux-2.4.19/fs/ext3/file.c 2004-04-23 18:26:27.000000000 -0400
+@@ -38,6 +38,9 @@
+ {
+ 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/fs/ext3/hash.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/hash.c 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.19/fs/ext3/hash.c 2004-04-23 18:26:27.000000000 -0400
+@@ -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 <linux/fs.h>
++#include <linux/jbd.h>
++#include <linux/sched.h>
++#include <linux/ext3_fs.h>
++
++#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/fs/ext3/namei.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/namei.c 2004-04-23 17:53:55.000000000 -0400
++++ linux-2.4.19/fs/ext3/namei.c 2004-04-23 22:24:05.000000000 -0400
+@@ -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 <linux/fs.h>
+@@ -40,6 +46,630 @@
+ #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.
+ *
+@@ -96,6 +726,7 @@
+ return 0;
+ }
+
++
+ /*
+ * ext3_find_entry()
+ *
+@@ -107,6 +738,8 @@
+ * 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)
+ {
+@@ -121,12 +754,32 @@
+ 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;
+@@ -167,7 +820,7 @@
+ 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 {
+@@ -198,6 +851,66 @@
+ 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<<EXT3_BLOCK_SIZE_BITS(sb))
++ +((char *)de - bh->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;
+@@ -214,8 +927,9 @@
+ brelse (bh);
+ inode = iget(dir->i_sb, ino);
+
+- if (!inode)
++ if (!inode) {
+ return ERR_PTR(-EACCES);
++ }
+ }
+ d_add(dentry, inode);
+ return NULL;
+@@ -239,6 +953,301 @@
+ 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()
+ *
+@@ -249,127 +1258,197 @@
+ * 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;
+- }
+-
+- ext3_debug ("creating next block\n");
+-
+- 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 {
+-
+- ext3_debug ("skipping to next block\n");
++ de->inode = 0;
++ de->rec_len = cpu_to_le16(rlen = blocksize);
++ nlen = 0;
++ return add_dirent_to_buf(handle, dentry, inode, de, bh);
++}
+
+- 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;
++#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;
++
++ frame = dx_probe(dentry, 0, &hinfo, frames, &err);
++ if (!frame)
++ return err;
++ entries = frame->entries;
++ at = frame->at;
++
++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err)))
++ goto cleanup;
++
++ 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
+@@ -453,9 +1532,11 @@
+ 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;
+@@ -480,9 +1561,11 @@
+ 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;
+@@ -512,9 +1595,11 @@
+ 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;
+@@ -526,7 +1611,8 @@
+
+ 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) {
+ inode->i_nlink--; /* is this nlink == 0? */
+@@ -555,21 +1641,19 @@
+ brelse (dir_block);
+ 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;
+ }
+
+ /*
+@@ -656,7 +1740,7 @@
+ 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
+@@ -697,7 +1781,7 @@
+ * 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",
+@@ -715,25 +1799,26 @@
+ 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
+@@ -794,8 +1879,9 @@
+ 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);
+@@ -833,7 +1919,7 @@
+ 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:
+@@ -851,8 +1937,9 @@
+ 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;
+@@ -879,7 +1966,7 @@
+ 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)
+@@ -905,9 +1992,11 @@
+ 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;
+@@ -917,7 +2006,7 @@
+ 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 = &ext3_symlink_inode_operations;
+ inode->i_mapping->a_ops = &ext3_aops;
+ /*
+@@ -926,25 +2015,23 @@
+ * 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;
+ ext3_mark_inode_dirty(handle, inode);
+ err = ext3_add_nondir(handle, dentry, 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,
+@@ -957,12 +2044,15 @@
+ 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;
+@@ -996,9 +2086,11 @@
+
+ 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;
+@@ -1078,7 +2170,7 @@
+ 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);
+@@ -1090,7 +2182,7 @@
+ 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/fs/ext3/super.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/super.c 2004-04-23 18:25:03.000000000 -0400
++++ linux-2.4.19/fs/ext3/super.c 2004-04-23 18:26:27.000000000 -0400
+@@ -741,6 +741,7 @@
+ 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
+@@ -751,6 +752,7 @@
+ 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) {
+@@ -925,6 +927,7 @@
+ return res;
+ }
+
++
+ struct super_block * ext3_read_super (struct super_block * sb, void * data,
+ int silent)
+ {
+@@ -1113,6 +1116,9 @@
+ 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
+@@ -1821,6 +1827,7 @@
+ 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");
+Index: linux-2.4.19/include/linux/ext3_fs.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/ext3_fs.h 2004-04-23 17:53:55.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_fs.h 2004-04-23 18:26:27.000000000 -0400
+@@ -40,6 +40,11 @@
+ #define EXT3FS_VERSION "2.4-0.9.18"
+
+ /*
++ * Always enable hashed directories
++ */
++#define CONFIG_EXT3_INDEX
++
++/*
+ * Debug code
+ */
+ #ifdef EXT3FS_DEBUG
+@@ -414,8 +419,11 @@
+ /*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__
+@@ -552,9 +560,46 @@
+ #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
+@@ -564,6 +609,27 @@
+ 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
+ */
+@@ -591,11 +657,20 @@
+
+ /* 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 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 *, struct inode *, int);
+ extern void ext3_free_inode (handle_t *, struct inode *);
+@@ -628,6 +703,8 @@
+ /* 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/include/linux/ext3_fs_sb.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/ext3_fs_sb.h 2004-04-23 17:53:54.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_fs_sb.h 2004-04-23 18:26:27.000000000 -0400
+@@ -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/include/linux/ext3_jbd.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/ext3_jbd.h 2004-04-23 17:53:54.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_jbd.h 2004-04-23 18:26:27.000000000 -0400
+@@ -69,6 +69,8 @@
+
+ #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/include/linux/rbtree.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/rbtree.h 2001-11-22 14:46:18.000000000 -0500
++++ linux-2.4.19/include/linux/rbtree.h 2004-04-23 18:26:27.000000000 -0400
+@@ -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/lib/rbtree.c
+===================================================================
+--- linux-2.4.19.orig/lib/rbtree.c 2002-08-02 20:39:46.000000000 -0400
++++ linux-2.4.19/lib/rbtree.c 2004-04-23 18:26:27.000000000 -0400
+@@ -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 <linux/rbtree.h>
+@@ -294,3 +296,43 @@
+ __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);
++
-Index: linux-2.4.19.SuSE/fs/ext3/namei.c
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/namei.c 2004-04-23 22:36:03.000000000 -0400
++++ linux-2.4.19/fs/ext3/namei.c 2004-04-23 22:37:37.000000000 -0400
@@ -1751,8 +1751,8 @@
struct super_block *sb = inode->i_sb;
struct ext3_iloc iloc;
@@ -1813,20 +1813,19 @@
{
struct list_head *prev;
- struct ext3_inode_info *ei = EXT3_I(inode);
+ 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;
int err = 0;
- lock_super(inode->i_sb);
-+ down(&sbi->s_orphan_lock);
- if (list_empty(&ei->i_orphan)) {
++ down(&sbi->s_orphan_lock);
+ if (list_empty(&ei->i_orphan)) {
- unlock_super(inode->i_sb);
-+ up(&sbi->s_orphan_lock);
++ up(&sbi->s_orphan_lock);
return 0;
}
ino_next = NEXT_ORPHAN(inode);
- prev = ei->i_orphan.prev;
+ prev = ei->i_orphan.prev;
- sbi = EXT3_SB(inode->i_sb);
jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino);
return err;
out_brelse:
-Index: linux-2.4.19.SuSE/fs/ext3/super.c
+Index: linux-2.4.19/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 @@
+--- linux-2.4.19.orig/fs/ext3/super.c 2004-04-23 22:30:41.000000000 -0400
++++ linux-2.4.19/fs/ext3/super.c 2004-04-23 22:36:22.000000000 -0400
+@@ -1179,6 +1179,7 @@
*/
sb->s_op = &ext3_sops;
INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
sb->s_root = 0;
-Index: linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/include/linux/ext3_fs_sb.h 2004-04-23 18:26:27.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_fs_sb.h 2004-04-23 22:36:22.000000000 -0400
@@ -69,6 +69,7 @@
struct inode * s_journal_inode;
struct journal_s * s_journal;
--- /dev/null
+ fs/ext3/xattr.c | 12 +++++-
+ fs/ext3/xattr_trusted.c | 86 +++++++++++++++++++++++++++++++++++++++++++++
+ include/linux/ext3_xattr.h | 6 +++
+ 3 files changed, 102 insertions(+), 2 deletions(-)
+
+Index: linux-2.4.19/fs/ext3/xattr.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/xattr.c 2004-04-23 22:44:57.000000000 -0400
++++ linux-2.4.19/fs/ext3/xattr.c 2004-04-23 22:45:20.000000000 -0400
+@@ -1785,18 +1785,25 @@
+ int __init
+ init_ext3_xattr(void)
+ {
++ int error;
++
+ 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;
++ error = init_ext3_xattr_trusted();
++ if (error)
++ mb_cache_destroy(ext3_xattr_cache);
++
++ return error;
+ }
+
+ void
+ exit_ext3_xattr(void)
+ {
++ exit_ext3_xattr_trusted();
+ if (ext3_xattr_cache)
+ mb_cache_destroy(ext3_xattr_cache);
+ ext3_xattr_cache = NULL;
+@@ -1807,12 +1814,13 @@
+ int __init
+ init_ext3_xattr(void)
+ {
+- return 0;
++ return init_ext3_xattr_trusted();
+ }
+
+ void
+ exit_ext3_xattr(void)
+ {
++ exit_ext3_xattr_trusted();
+ }
+
+ #endif /* CONFIG_EXT3_FS_XATTR_SHARING */
+Index: linux-2.4.19/fs/ext3/xattr_trusted.c
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/xattr_trusted.c 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.19/fs/ext3/xattr_trusted.c 2004-04-23 22:45:20.000000000 -0400
+@@ -0,0 +1,86 @@
++/*
++ * linux/fs/ext3/xattr_trusted.c
++ * Handler for trusted extended attributes.
++ *
++ * Copyright (C) 2003 by Andreas Gruenbacher, <a.gruenbacher@computer.org>
++ */
++
++#include <linux/module.h>
++#include <linux/string.h>
++#include <linux/fs.h>
++#include <linux/ext3_jbd.h>
++#include <linux/ext3_fs.h>
++#include <linux/ext3_xattr.h>
++
++#define XATTR_TRUSTED_PREFIX "trusted."
++
++static size_t
++ext3_xattr_trusted_list(char *list, struct inode *inode,
++ const char *name, int name_len)
++{
++ const int prefix_len = sizeof(XATTR_TRUSTED_PREFIX)-1;
++
++ if (!capable(CAP_SYS_ADMIN))
++ return 0;
++
++ if (list) {
++ memcpy(list, XATTR_TRUSTED_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_trusted_get(struct inode *inode, const char *name,
++ void *buffer, size_t size)
++{
++ if (strcmp(name, "") == 0)
++ return -EINVAL;
++ if (!capable(CAP_SYS_ADMIN))
++ return -EPERM;
++ return ext3_xattr_get(inode, EXT3_XATTR_INDEX_TRUSTED, name,
++ buffer, size);
++}
++
++static int
++ext3_xattr_trusted_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 (!capable(CAP_SYS_ADMIN))
++ return -EPERM;
++ 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_TRUSTED, name,
++ value, size, flags);
++ ext3_journal_stop(handle, inode);
++
++ return error;
++}
++
++struct ext3_xattr_handler ext3_xattr_trusted_handler = {
++ .prefix = XATTR_TRUSTED_PREFIX,
++ .list = ext3_xattr_trusted_list,
++ .get = ext3_xattr_trusted_get,
++ .set = ext3_xattr_trusted_set,
++};
++
++int __init
++init_ext3_xattr_trusted(void)
++{
++ return ext3_xattr_register(EXT3_XATTR_INDEX_TRUSTED,
++ &ext3_xattr_trusted_handler);
++}
++
++void
++exit_ext3_xattr_trusted(void)
++{
++ ext3_xattr_unregister(EXT3_XATTR_INDEX_TRUSTED,
++ &ext3_xattr_trusted_handler);
++}
+Index: linux-2.4.19/fs/ext3/Makefile
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/Makefile 2004-04-23 22:38:38.000000000 -0400
++++ linux-2.4.19/fs/ext3/Makefile 2004-04-23 22:49:23.000000000 -0400
+@@ -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 iopen.o \
+- ioctl.o namei.o super.o symlink.o hash.o
++ ioctl.o namei.o super.o symlink.o hash.o xattr_trusted.o
+ obj-m := $(O_TARGET)
+
+ obj-$(CONFIG_EXT3_FS_XATTR) += xattr.o
+Index: linux-2.4.19/include/linux/ext3_xattr.h
+===================================================================
+--- linux-2.4.19.orig/include/linux/ext3_xattr.h 2004-04-23 17:53:54.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_xattr.h 2004-04-23 22:45:20.000000000 -0400
+@@ -21,6 +21,9 @@
+ #define EXT3_XATTR_INDEX_USER 1
+ #define EXT3_XATTR_INDEX_POSIX_ACL_ACCESS 2
+ #define EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT 3
++#define EXT3_XATTR_INDEX_TRUSTED 4
++#define EXT3_XATTR_INDEX_LUSTRE 5
++#define EXT3_XATTR_INDEX_SECURITY 6
+
+ struct ext3_xattr_header {
+ __u32 h_magic; /* magic number for identification */
+@@ -84,6 +87,9 @@
+ extern int init_ext3_xattr(void) __init;
+ extern void exit_ext3_xattr(void);
+
++extern int init_ext3_xattr_trusted(void) __init;
++extern void exit_ext3_xattr_trusted(void);
++
+ # else /* CONFIG_EXT3_FS_XATTR */
+ # define ext3_setxattr NULL
+ # define ext3_getxattr NULL
./fs/ext3/namei.c | 11 +++++------
1 files changed, 5 insertions(+), 6 deletions(-)
-Index: linux-2.4.19.SuSE/./fs/ext3/namei.c
+Index: linux-2.4.19/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 @@
+--- linux-2.4.19.orig/fs/ext3/namei.c 2004-04-23 22:30:41.000000000 -0400
++++ linux-2.4.19/fs/ext3/namei.c 2004-04-23 22:36:03.000000000 -0400
+@@ -1522,8 +1522,11 @@
{
int err = ext3_add_entry(handle, dentry, inode);
if (!err) {
}
ext3_dec_count(handle, inode);
iput(inode);
-@@ -1560,7 +1563,6 @@
+@@ -1559,7 +1562,6 @@
inode->i_op = &ext3_file_inode_operations;
inode->i_fop = &ext3_file_operations;
inode->i_mapping->a_ops = &ext3_aops;
err = ext3_add_nondir(handle, dentry, inode);
}
ext3_journal_stop(handle, dir);
-@@ -1590,7 +1592,6 @@
+@@ -1589,7 +1591,6 @@
#ifdef CONFIG_EXT3_FS_XATTR
inode->i_op = &ext3_special_inode_operations;
#endif
@@ -2039,7 +2040,6 @@
inode->i_size = l-1;
}
- EXT3_I(inode)->i_disksize = inode->i_size;
+ EXT3_I(inode)->i_disksize = inode->i_size;
- ext3_mark_inode_dirty(handle, inode);
err = ext3_add_nondir(handle, dentry, inode);
out_stop:
--- /dev/null
+
+
+
+ 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-2.4.21/fs/inode.c
+===================================================================
+--- linux-2.4.21.orig/fs/inode.c 2004-04-24 02:38:51.000000000 -0400
++++ linux-2.4.21/fs/inode.c 2004-04-26 19:41:58.000000000 -0400
+@@ -651,7 +651,8 @@
+ /*
+ * 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;
+@@ -676,6 +677,11 @@
+ 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 */
+@@ -694,22 +700,23 @@
+ /**
+ * 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);
+@@ -735,7 +742,7 @@
+ * hold).
+ */
+ shrink_dcache_sb(sb);
+- res = invalidate_inodes(sb);
++ res = invalidate_inodes(sb, 0);
+ drop_super(sb);
+ }
+ invalidate_buffers(dev);
+Index: linux-2.4.21/fs/super.c
+===================================================================
+--- linux-2.4.21.orig/fs/super.c 2004-04-24 02:38:51.000000000 -0400
++++ linux-2.4.21/fs/super.c 2004-04-26 19:41:58.000000000 -0400
+@@ -932,7 +932,7 @@
+ lock_super(sb);
+ lock_kernel();
+ 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);
+@@ -941,7 +941,7 @@
+ }
+
+ /* Forget any remaining inodes */
+- if (invalidate_inodes(sb)) {
++ if (invalidate_inodes(sb, 1)) {
+ printk(KERN_ERR "VFS: Busy inodes after unmount. "
+ "Self-destruct in 5 seconds. Have a nice day...\n");
+ }
+Index: linux-2.4.21/fs/smbfs/inode.c
+===================================================================
+--- linux-2.4.21.orig/fs/smbfs/inode.c 2004-04-24 02:38:44.000000000 -0400
++++ linux-2.4.21/fs/smbfs/inode.c 2004-04-26 19:41:58.000000000 -0400
+@@ -167,7 +167,7 @@
+ {
+ VERBOSE("\n");
+ shrink_dcache_sb(SB_of(server));
+- invalidate_inodes(SB_of(server));
++ invalidate_inodes(SB_of(server), 0);
+ }
+
+ /*
+Index: linux-2.4.21/fs/ntfs/super.c
+===================================================================
+--- linux-2.4.21.orig/fs/ntfs/super.c 2004-04-24 02:38:38.000000000 -0400
++++ linux-2.4.21/fs/ntfs/super.c 2004-04-26 19:42:12.000000000 -0400
+@@ -1604,7 +1604,7 @@
+ * method again... FIXME: Do we need to do this twice now because of
+ * attribute inodes? I think not, so leave as is for now... (AIA)
+ */
+- if (invalidate_inodes(sb)) {
++ if (invalidate_inodes(sb, 0)) {
+ ntfs_error(sb, "Busy inodes left. This is most likely a NTFS "
+ "driver bug.");
+ /* Copied from fs/super.c. I just love this message. (-; */
+Index: linux-2.4.21/include/linux/fs.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/fs.h 2004-04-26 19:06:32.000000000 -0400
++++ linux-2.4.21/include/linux/fs.h 2004-04-26 19:41:58.000000000 -0400
+@@ -1401,7 +1401,7 @@
+ 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 *);
include/linux/ext3_fs.h | 2
8 files changed, 318 insertions(+), 2 deletions(-)
-Index: linux-2.4.19.SuSE/Documentation/filesystems/ext2.txt
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/Documentation/filesystems/ext2.txt 2001-07-11 18:44:45.000000000 -0400
++++ linux-2.4.19/Documentation/filesystems/ext2.txt 2004-04-23 22:37:48.000000000 -0400
@@ -35,6 +35,22 @@
sb=n Use alternate superblock at this location.
grpquota,noquota,quota,usrquota Quota options are silently ignored by ext2.
-Index: linux-2.4.19.SuSE/fs/ext3/Makefile
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/Makefile 2004-04-23 18:26:27.000000000 -0400
++++ linux-2.4.19/fs/ext3/Makefile 2004-04-23 22:38:38.000000000 -0400
@@ -11,7 +11,7 @@
- export-objs := ext3-exports.o
+ export-objs := super.o inode.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
+ ioctl.o namei.o super.o symlink.o hash.o
obj-m := $(O_TARGET)
-Index: linux-2.4.19.SuSE/fs/ext3/inode.c
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/inode.c 2004-04-23 22:37:42.000000000 -0400
++++ linux-2.4.19/fs/ext3/inode.c 2004-04-23 22:37:48.000000000 -0400
@@ -34,6 +34,7 @@
#include <linux/highuid.h>
#include <linux/quotaops.h>
if(ext3_get_inode_loc(inode, &iloc))
goto bad_inode;
bh = iloc.bh;
-Index: linux-2.4.19.SuSE/fs/ext3/iopen.c
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/iopen.c 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.19/fs/ext3/iopen.c 2004-04-23 22:37:48.000000000 -0400
@@ -0,0 +1,258 @@
+/*
+ * linux/fs/ext3/iopen.c
+
+ return 1;
+}
-Index: linux-2.4.19.SuSE/fs/ext3/iopen.h
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/iopen.h 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.19/fs/ext3/iopen.h 2004-04-23 22:37:48.000000000 -0400
@@ -0,0 +1,15 @@
+/*
+ * iopen.h
+extern int ext3_iopen_get_inode(struct inode *inode);
+extern struct dentry *iopen_connect_dentry(struct dentry *de,
+ struct inode *inode);
-Index: linux-2.4.19.SuSE/fs/ext3/namei.c
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/namei.c 2004-04-23 22:37:42.000000000 -0400
++++ linux-2.4.19/fs/ext3/namei.c 2004-04-23 22:37:48.000000000 -0400
@@ -36,7 +36,7 @@
#include <linux/string.h>
#include <linux/locks.h>
/*
* define how far ahead to read directories while searching them.
-@@ -922,10 +922,14 @@
+@@ -928,10 +928,14 @@
struct inode * inode;
struct ext3_dir_entry_2 * de;
struct buffer_head * bh;
bh = ext3_find_entry(dentry, &de);
inode = NULL;
if (bh) {
-@@ -943,7 +948,28 @@
+@@ -943,7 +947,28 @@
return ERR_PTR(-EACCES);
}
}
return NULL;
}
-Index: linux-2.4.19.SuSE/fs/ext3/super.c
+Index: linux-2.4.19/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 +864,18 @@
+--- linux-2.4.19.orig/fs/ext3/super.c 2004-04-23 22:37:42.000000000 -0400
++++ linux-2.4.19/fs/ext3/super.c 2004-04-23 22:37:48.000000000 -0400
+@@ -861,6 +861,18 @@
|| !strcmp (this_char, "quota")
|| !strcmp (this_char, "usrquota"))
/* Don't do anything ;-) */ ;
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
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/include/linux/ext3_fs.h 2004-04-23 22:37:42.000000000 -0400
++++ linux-2.4.19/include/linux/ext3_fs.h 2004-04-23 22:37:48.000000000 -0400
@@ -324,6 +324,8 @@
#define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */
#define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */
--- /dev/null
+ 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-2.4.21/Documentation/filesystems/ext2.txt
+===================================================================
+--- linux-2.4.21.orig/Documentation/filesystems/ext2.txt 2001-07-11 18:44:45.000000000 -0400
++++ linux-2.4.21/Documentation/filesystems/ext2.txt 2004-04-24 02:46:32.000000000 -0400
+@@ -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-2.4.21/fs/ext3/Makefile
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/Makefile 2004-04-24 02:46:18.000000000 -0400
++++ linux-2.4.21/fs/ext3/Makefile 2004-04-24 02:47:02.000000000 -0400
+@@ -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-2.4.21/fs/ext3/inode.c
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/inode.c 2004-04-24 02:46:19.000000000 -0400
++++ linux-2.4.21/fs/ext3/inode.c 2004-04-24 02:46:32.000000000 -0400
+@@ -34,6 +34,7 @@
+ #include <linux/highuid.h>
+ #include <linux/quotaops.h>
+ #include <linux/module.h>
++#include "iopen.h"
+
+ /*
+ * SEARCH_FROM_ZERO forces each block allocation to search from the start
+@@ -2252,6 +2253,9 @@
+ 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-2.4.21/fs/ext3/iopen.c
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/iopen.c 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.21/fs/ext3/iopen.c 2004-04-24 02:46:32.000000000 -0400
+@@ -0,0 +1,258 @@
++/*
++ * 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 <linux/sched.h>
++#include <linux/fs.h>
++#include <linux/locks.h>
++#include <linux/ext3_jbd.h>
++#include <linux/jbd.h>
++#include <linux/ext3_fs.h>
++#include <linux/smp_lock.h>
++#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);
++ }
++
++ assert(list_empty(&dentry->d_alias)); /* d_instantiate */
++ assert(list_empty(&dentry->d_hash)); /* d_rehash */
++
++ /* 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;
++
++ /* d_add(), but don't drop dcache_lock before adding dentry to inode */
++ list_add(&dentry->d_alias, &inode->i_dentry); /* d_instantiate */
++ dentry->d_inode = inode;
++
++ __d_rehash(dentry, 0); /* d_rehash */
++ spin_unlock(&dcache_lock);
++
++ 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.
++ * Caller must hold dcache_lock.
++ */
++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 */
++ 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)
++ 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);
++
++ 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-2.4.21/fs/ext3/iopen.h
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/iopen.h 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.21/fs/ext3/iopen.h 2004-04-24 02:46:32.000000000 -0400
+@@ -0,0 +1,15 @@
++/*
++ * 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);
++extern struct dentry *iopen_connect_dentry(struct dentry *de,
++ struct inode *inode);
+Index: linux-2.4.21/fs/ext3/namei.c
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/namei.c 2004-04-24 02:46:19.000000000 -0400
++++ linux-2.4.21/fs/ext3/namei.c 2004-04-24 02:46:32.000000000 -0400
+@@ -36,7 +36,7 @@
+ #include <linux/string.h>
+ #include <linux/locks.h>
+ #include <linux/quotaops.h>
+-
++#include "iopen.h"
+
+ /*
+ * define how far ahead to read directories while searching them.
+@@ -928,10 +928,14 @@
+ 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) {
+@@ -943,7 +947,28 @@
+ return ERR_PTR(-EACCES);
+ }
+ }
+- d_add(dentry, inode);
++
++ /* verify this dentry is really new */
++ assert(!dentry->d_inode);
++ assert(list_empty(&dentry->d_alias)); /* d_instantiate */
++ assert(list_empty(&dentry->d_hash)); /* d_rehash */
++ assert(list_empty(&dentry->d_subdirs));
++
++ spin_lock(&dcache_lock);
++ if (inode && (alternate = iopen_connect_dentry(dentry, inode))) {
++ spin_unlock(&dcache_lock);
++ iput(inode);
++ return alternate;
++ }
++
++ /* d_add(), but don't drop dcache_lock before adding dentry to inode */
++ if (inode) /* d_instantiate */
++ list_add(&dentry->d_alias, &inode->i_dentry);
++ dentry->d_inode = inode;
++
++ __d_rehash(dentry, 0); /* d_rehash */
++ spin_unlock(&dcache_lock);
++
+ return NULL;
+ }
+
+Index: linux-2.4.21/fs/ext3/super.c
+===================================================================
+--- linux-2.4.21.orig/fs/ext3/super.c 2004-04-24 02:46:19.000000000 -0400
++++ linux-2.4.21/fs/ext3/super.c 2004-04-24 02:46:32.000000000 -0400
+@@ -869,6 +869,18 @@
+ || !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-2.4.21/include/linux/ext3_fs.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/ext3_fs.h 2004-04-24 02:46:19.000000000 -0400
++++ linux-2.4.21/include/linux/ext3_fs.h 2004-04-24 02:46:32.000000000 -0400
+@@ -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
--- /dev/null
+Index: linux-2.4.21/arch/um/kernel/Makefile
+===================================================================
+--- linux-2.4.21.orig/arch/um/kernel/Makefile 2004-04-24 02:37:58.000000000 -0400
++++ linux-2.4.21/arch/um/kernel/Makefile 2004-04-24 02:51:03.000000000 -0400
+@@ -37,7 +37,8 @@
+ export-objs-$(CONFIG_GPROF) += gprof_syms.o
+ export-objs-$(CONFIG_GCOV) += gmon_syms.o
+
+-export-objs = ksyms.o process_kern.o signal_kern.o user_syms.o $(export-objs-y)
++export-objs = ksyms.o process_kern.o signal_kern.o user_syms.o sysrq.o \
++ $(export-objs-y)
+
+ CFLAGS_user_syms.o = -D__AUTOCONF_INCLUDED__ $(DMODULES-y) $(DMODVERSIONS-y) \
+ -I/usr/include -I../include
+Index: linux-2.4.21/arch/um/kernel/sysrq.c
+===================================================================
+--- linux-2.4.21.orig/arch/um/kernel/sysrq.c 2004-04-24 02:37:58.000000000 -0400
++++ linux-2.4.21/arch/um/kernel/sysrq.c 2004-04-24 02:51:03.000000000 -0400
+@@ -86,6 +86,37 @@
+ show_trace((unsigned long *)esp);
+ }
+
++#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;
++}
++
++EXPORT_SYMBOL(is_kernel_text_address);
++
+ /*
+ * Overrides for Emacs so that we follow Linus's tabbing style.
+ * Emacs will notice this stuff at the end of the file and automatically
+Index: linux-2.4.21/arch/i386/kernel/Makefile
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/Makefile 2004-04-24 02:39:05.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/Makefile 2004-04-24 02:51:26.000000000 -0400
+@@ -20,7 +20,7 @@
+
+ O_TARGET := kernel.o
+
+-export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o traps.o dr_alloc.o
++export-objs := mca.o mtrr.o msr.o cpuid.o microcode.o i386_ksyms.o time.o traps.o dr_alloc.o traps.o
+
+ ifdef CONFIG_X86_SPEEDSTEP_ICH
+ export-objs += speedstep-lib.o
+Index: linux-2.4.21/arch/i386/kernel/traps.c
+===================================================================
+--- linux-2.4.21.orig/arch/i386/kernel/traps.c 2004-04-24 02:42:58.000000000 -0400
++++ linux-2.4.21/arch/i386/kernel/traps.c 2004-04-24 02:51:03.000000000 -0400
+@@ -1339,3 +1339,41 @@
+ 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);
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
ext3/ext3-exports.c | 13 +++++++++++++
2 files changed, 14 insertions(+), 2 deletions(-)
-Index: linux-2.4.19.SuSE/fs/ext2/super.c
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext2/super.c 2004-04-23 17:53:55.000000000 -0400
++++ linux-2.4.19/fs/ext2/super.c 2004-04-23 22:30:41.000000000 -0400
@@ -70,6 +70,7 @@
{
va_list args;
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
+Index: linux-2.4.19/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 @@
+--- linux-2.4.19.orig/fs/ext3/super.c 2004-04-23 18:26:27.000000000 -0400
++++ linux-2.4.19/fs/ext3/super.c 2004-04-23 22:30:41.000000000 -0400
+@@ -1827,8 +1827,6 @@
exit_ext3_xattr();
}
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
+Index: linux-2.4.19/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
+--- linux-2.4.19.orig/fs/ext3/ext3-exports.c 2003-01-30 05:24:37.000000000 -0500
++++ linux-2.4.19/fs/ext3/ext3-exports.c 2004-04-23 22:30:41.000000000 -0400
@@ -0,0 +1,13 @@
+#include <linux/config.h>
+#include <linux/module.h>
+EXPORT_SYMBOL(ext3_xattr_get);
+EXPORT_SYMBOL(ext3_xattr_list);
+EXPORT_SYMBOL(ext3_xattr_set);
+Index: linux-2.4.19/fs/ext3/Makefile
+===================================================================
+--- linux-2.4.19.orig/fs/ext3/Makefile 2004-04-23 18:26:27.000000000 -0400
++++ linux-2.4.19/fs/ext3/Makefile 2004-04-23 23:05:20.000000000 -0400
+@@ -9,7 +9,7 @@
+
+ O_TARGET := ext3.o
+
+-export-objs := super.o inode.o
++export-objs := ext3-exports.c
+
+ 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
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
--- /dev/null Fri Aug 30 17:31:37 2002
+++ linux-2.4.18-18.8.0-l12-braam/include/linux/lustre_version.h Thu Feb 13 07:58:33 2003
@@ -0,0 +1 @@
-+#define LUSTRE_KERNEL_VERSION 34
++#define LUSTRE_KERNEL_VERSION 35
_
--- /dev/null
+Index: linux-2.4.20-30.9/scripts/mkdep.c
+===================================================================
+--- linux-2.4.20-30.9.orig/scripts/mkdep.c 2004-02-19 19:40:51.000000000 -0500
++++ linux-2.4.20-30.9/scripts/mkdep.c 2004-04-28 17:24:54.000000000 -0400
+@@ -48,8 +48,6 @@
+ char __depname[512] = "\n\t@touch ";
+ #define depname (__depname+9)
+ int hasdep;
+-char cwd[PATH_MAX];
+-int lcwd;
+
+ struct path_struct {
+ int len;
+@@ -204,22 +202,8 @@
+ memcpy(path->buffer+path->len, name, len);
+ path->buffer[path->len+len] = '\0';
+ if (access(path->buffer, F_OK) == 0) {
+- int l = lcwd + strlen(path->buffer);
+- char name2[l+2], *p;
+- if (path->buffer[0] == '/') {
+- memcpy(name2, path->buffer, l+1);
+- }
+- else {
+- memcpy(name2, cwd, lcwd);
+- name2[lcwd] = '/';
+- memcpy(name2+lcwd+1, path->buffer, path->len+len+1);
+- }
+- while ((p = strstr(name2, "/../"))) {
+- *p = '\0';
+- strcpy(strrchr(name2, '/'), p+3);
+- }
+ do_depname();
+- printf(" \\\n %s", name2);
++ printf(" \\\n %s", path->buffer);
+ return;
+ }
+ }
+@@ -601,12 +585,6 @@
+ return 1;
+ }
+
+- if (!getcwd(cwd, sizeof(cwd))) {
+- fprintf(stderr, "mkdep: getcwd() failed %m\n");
+- return 1;
+- }
+- lcwd = strlen(cwd);
+-
+ add_path("."); /* for #include "..." */
+
+ while (++argv, --argc > 0) {
--- /dev/null
+Index: linux-2.4.21/include/linux/skbuff.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/skbuff.h 2004-04-24 02:38:40.000000000 -0400
++++ linux-2.4.21/include/linux/skbuff.h 2004-04-24 02:47:46.000000000 -0400
+@@ -116,6 +116,30 @@
+ __u16 size;
+ };
+
++/* Support for callback when skb data has been released */
++typedef struct zccd /* Zero Copy Callback Descriptor */
++{ /* (embed as first member of custom struct) */
++ atomic_t zccd_count; /* reference count */
++ void (*zccd_destructor)(struct zccd *); /* callback when refcount reaches zero */
++} zccd_t;
++
++static inline void zccd_init (zccd_t *d, void (*callback)(zccd_t *))
++{
++ atomic_set (&d->zccd_count, 1);
++ d->zccd_destructor = callback;
++}
++
++static inline void zccd_get (zccd_t *d) /* take a reference */
++{
++ atomic_inc (&d->zccd_count);
++}
++
++static inline void zccd_put (zccd_t *d) /* release a reference */
++{
++ if (atomic_dec_and_test (&d->zccd_count))
++ (d->zccd_destructor)(d);
++}
++
+ /* This data is invariant across clones and lives at
+ * the end of the header data, ie. at skb->end.
+ */
+@@ -123,6 +147,12 @@
+ atomic_t dataref;
+ unsigned int nr_frags;
+ struct sk_buff *frag_list;
++ zccd_t *zccd; /* zero copy descriptor */
++ zccd_t *zccd2; /* 2nd zero copy descriptor */
++ /* NB we expect zero-copy data to be at least 1 packet, so
++ * having 2 zccds means we don't unneccessarily split the packet
++ * where consecutive zero-copy sends abutt.
++ */
+ skb_frag_t frags[MAX_SKB_FRAGS];
+ };
+
+Index: linux-2.4.21/include/net/tcp.h
+===================================================================
+--- linux-2.4.21.orig/include/net/tcp.h 2004-04-24 02:39:20.000000000 -0400
++++ linux-2.4.21/include/net/tcp.h 2004-04-24 02:48:27.000000000 -0400
+@@ -646,6 +646,8 @@
+
+ extern int tcp_sendmsg(struct sock *sk, struct msghdr *msg, int size);
+ extern ssize_t tcp_sendpage(struct socket *sock, struct page *page, int offset, size_t size, int flags);
++extern ssize_t tcp_sendpage_zccd(struct socket *sock, struct page *page, int offset, size_t size,
++ int flags, zccd_t *zccd);
+
+ extern int tcp_ioctl(struct sock *sk,
+ int cmd,
+@@ -742,6 +744,10 @@
+ struct msghdr *msg,
+ int len, int nonblock,
+ int flags, int *addr_len);
++extern int tcp_recvpackets(struct sock *sk,
++ struct sk_buff_head *packets,
++ int len, int nonblock);
++
+
+ extern int tcp_listen_start(struct sock *sk);
+
+Index: linux-2.4.21/net/netsyms.c
+===================================================================
+--- linux-2.4.21.orig/net/netsyms.c 2004-04-24 02:39:13.000000000 -0400
++++ linux-2.4.21/net/netsyms.c 2004-04-24 02:47:46.000000000 -0400
+@@ -403,6 +403,8 @@
+
+ #endif
+
++EXPORT_SYMBOL(tcp_sendpage_zccd);
++EXPORT_SYMBOL(tcp_recvpackets);
+ EXPORT_SYMBOL(tcp_read_sock);
+
+ EXPORT_SYMBOL(netlink_set_err);
+Index: linux-2.4.21/net/core/skbuff.c
+===================================================================
+--- linux-2.4.21.orig/net/core/skbuff.c 2004-04-24 02:38:40.000000000 -0400
++++ linux-2.4.21/net/core/skbuff.c 2004-04-24 02:47:46.000000000 -0400
+@@ -208,6 +208,8 @@
+ atomic_set(&(skb_shinfo(skb)->dataref), 1);
+ skb_shinfo(skb)->nr_frags = 0;
+ skb_shinfo(skb)->frag_list = NULL;
++ skb_shinfo(skb)->zccd = NULL; /* skbuffs kick off with NO user zero copy descriptors */
++ skb_shinfo(skb)->zccd2 = NULL;
+ return skb;
+
+ nodata:
+@@ -277,6 +279,10 @@
+ {
+ if (!skb->cloned ||
+ atomic_dec_and_test(&(skb_shinfo(skb)->dataref))) {
++ if (skb_shinfo(skb)->zccd != NULL) /* zero copy callback descriptor? */
++ zccd_put (skb_shinfo(skb)->zccd); /* release hold */
++ if (skb_shinfo(skb)->zccd2 != NULL) /* 2nd zero copy callback descriptor? */
++ zccd_put (skb_shinfo(skb)->zccd2); /* release hold */
+ if (skb_shinfo(skb)->nr_frags) {
+ int i;
+ for (i = 0; i < skb_shinfo(skb)->nr_frags; i++)
+@@ -535,6 +541,8 @@
+ atomic_set(&(skb_shinfo(skb)->dataref), 1);
+ skb_shinfo(skb)->nr_frags = 0;
+ skb_shinfo(skb)->frag_list = NULL;
++ skb_shinfo(skb)->zccd = NULL; /* copied data => no user zero copy descriptor */
++ skb_shinfo(skb)->zccd2 = NULL;
+
+ /* We are no longer a clone, even if we were. */
+ skb->cloned = 0;
+@@ -581,6 +589,14 @@
+ n->data_len = skb->data_len;
+ n->len = skb->len;
+
++ if (skb_shinfo(skb)->zccd != NULL) /* user zero copy descriptor? */
++ zccd_get (skb_shinfo(skb)->zccd); /* 1 more ref (pages are shared) */
++ skb_shinfo(n)->zccd = skb_shinfo(skb)->zccd;
++
++ if (skb_shinfo(skb)->zccd2 != NULL) /* 2nd user zero copy descriptor? */
++ zccd_get (skb_shinfo(skb)->zccd2); /* 1 more ref (pages are shared) */
++ skb_shinfo(n)->zccd2 = skb_shinfo(skb)->zccd2;
++
+ if (skb_shinfo(skb)->nr_frags) {
+ int i;
+
+@@ -623,6 +639,8 @@
+ u8 *data;
+ int size = nhead + (skb->end - skb->head) + ntail;
+ long off;
++ zccd_t *zccd = skb_shinfo(skb)->zccd; /* stash user zero copy descriptor */
++ zccd_t *zccd2 = skb_shinfo(skb)->zccd2; /* stash 2nd user zero copy descriptor */
+
+ if (skb_shared(skb))
+ BUG();
+@@ -644,6 +662,11 @@
+ if (skb_shinfo(skb)->frag_list)
+ skb_clone_fraglist(skb);
+
++ if (zccd != NULL) /* user zero copy descriptor? */
++ zccd_get (zccd); /* extra ref (pages are shared) */
++ if (zccd2 != NULL) /* 2nd user zero copy descriptor? */
++ zccd_get (zccd2); /* extra ref (pages are shared) */
++
+ skb_release_data(skb);
+
+ off = (data+nhead) - skb->head;
+@@ -658,6 +681,8 @@
+ skb->nh.raw += off;
+ skb->cloned = 0;
+ atomic_set(&skb_shinfo(skb)->dataref, 1);
++ skb_shinfo(skb)->zccd = zccd;
++ skb_shinfo(skb)->zccd2 = zccd2;
+ return 0;
+
+ nodata:
+Index: linux-2.4.21/net/ipv4/tcp.c
+===================================================================
+--- linux-2.4.21.orig/net/ipv4/tcp.c 2004-04-24 02:39:21.000000000 -0400
++++ linux-2.4.21/net/ipv4/tcp.c 2004-04-24 02:50:40.000000000 -0400
+@@ -748,7 +748,7 @@
+ goto out;
+ }
+
+-ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset, size_t psize, int flags);
++ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset, size_t psize, int flags, zccd_t *zccd);
+
+ static inline int
+ can_coalesce(struct sk_buff *skb, int i, struct page *page, int off)
+@@ -827,7 +827,7 @@
+ return err;
+ }
+
+-ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset, size_t psize, int flags)
++ssize_t do_tcp_sendpages(struct sock *sk, struct page **pages, int poffset, size_t psize, int flags, zccd_t *zccd)
+ {
+ struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
+ int mss_now;
+@@ -875,6 +875,17 @@
+ copy = size;
+
+ i = skb_shinfo(skb)->nr_frags;
++
++ if (zccd != NULL && /* this is a zcc I/O */
++ skb_shinfo(skb)->zccd != NULL && /* skb is part of a zcc I/O */
++ skb_shinfo(skb)->zccd2 != NULL &&
++ skb_shinfo(skb)->zccd != zccd && /* not the same one */
++ skb_shinfo(skb)->zccd2 != zccd)
++ {
++ tcp_mark_push (tp, skb);
++ goto new_segment;
++ }
++
+ if (can_coalesce(skb, i, page, offset)) {
+ skb_shinfo(skb)->frags[i-1].size += copy;
+ } else if (i < MAX_SKB_FRAGS) {
+@@ -885,6 +896,20 @@
+ goto new_segment;
+ }
+
++ if (zccd != NULL && /* this is a zcc I/O */
++ skb_shinfo(skb)->zccd != zccd && /* not already referencing this zccd */
++ skb_shinfo(skb)->zccd2 != zccd)
++ {
++ zccd_get (zccd); /* bump ref count */
++
++ BUG_TRAP (skb_shinfo(skb)->zccd2 == NULL);
++
++ if (skb_shinfo(skb)->zccd == NULL) /* reference this zccd */
++ skb_shinfo(skb)->zccd = zccd;
++ else
++ skb_shinfo(skb)->zccd2 = zccd;
++ }
++
+ skb->len += copy;
+ skb->data_len += copy;
+ skb->ip_summed = CHECKSUM_HW;
+@@ -948,7 +973,29 @@
+
+ lock_sock(sk);
+ TCP_CHECK_TIMER(sk);
+- res = do_tcp_sendpages(sk, &page, offset, size, flags);
++ res = do_tcp_sendpages(sk, &page, offset, size, flags, NULL);
++ TCP_CHECK_TIMER(sk);
++ release_sock(sk);
++ return res;
++}
++
++ssize_t tcp_sendpage_zccd(struct socket *sock, struct page *page, int offset, size_t size,
++ int flags, zccd_t *zccd)
++{
++ ssize_t res;
++ struct sock *sk = sock->sk;
++
++#define TCP_ZC_CSUM_FLAGS (NETIF_F_IP_CSUM|NETIF_F_NO_CSUM|NETIF_F_HW_CSUM)
++
++ if (!(sk->route_caps & NETIF_F_SG) || /* caller shouldn't waste her time */
++ !(sk->route_caps & TCP_ZC_CSUM_FLAGS)) /* on double mapping */
++ BUG ();
++
++ lock_sock(sk);
++ TCP_CHECK_TIMER(sk);
++
++ res = do_tcp_sendpages(sk, &page, 0, size, flags, zccd);
++
+ TCP_CHECK_TIMER(sk);
+ release_sock(sk);
+ return res;
+@@ -1772,6 +1819,202 @@
+ goto out;
+ }
+
++int tcp_recvpackets (struct sock *sk, struct sk_buff_head *packets,
++ int len, int nonblock)
++{
++ struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp);
++ int copied;
++ long timeo;
++
++ BUG_TRAP (len > 0);
++ /*BUG_TRAP ((flags & (MSG_OOB | MSG_PEEK | MSG_TRUNC)) == 0);*/
++
++ lock_sock(sk);
++
++ TCP_CHECK_TIMER(sk);
++
++ copied = -ENOTCONN;
++ if (sk->state == TCP_LISTEN)
++ goto out;
++
++ copied = 0;
++ timeo = sock_rcvtimeo(sk, nonblock);
++
++ do {
++ struct sk_buff * skb;
++ u32 offset;
++ unsigned long used;
++ int exhausted;
++ int eaten;
++
++ /* Are we at urgent data? Stop if we have read anything. */
++ if (copied && tp->urg_data && tp->urg_seq == tp->copied_seq)
++ break;
++
++ /* We need to check signals first, to get correct SIGURG
++ * handling. FIXME: Need to check this doesnt impact 1003.1g
++ * and move it down to the bottom of the loop
++ */
++ if (signal_pending(current)) {
++ if (copied)
++ break;
++ copied = timeo ? sock_intr_errno(timeo) : -EAGAIN;
++ break;
++ }
++
++ /* Next get a buffer. */
++
++ skb = skb_peek(&sk->receive_queue);
++
++ if (skb == NULL) /* nothing ready */
++ {
++ if (copied) {
++ if (sk->err ||
++ sk->state == TCP_CLOSE ||
++ (sk->shutdown & RCV_SHUTDOWN) ||
++ !timeo ||
++ (0))
++ break;
++ } else {
++ if (sk->done)
++ break;
++
++ if (sk->err) {
++ copied = sock_error(sk);
++ break;
++ }
++
++ if (sk->shutdown & RCV_SHUTDOWN)
++ break;
++
++ if (sk->state == TCP_CLOSE) {
++ if (!sk->done) {
++ /* This occurs when user tries to read
++ * from never connected socket.
++ */
++ copied = -ENOTCONN;
++ break;
++ }
++ break;
++ }
++
++ if (!timeo) {
++ copied = -EAGAIN;
++ break;
++ }
++ }
++
++ cleanup_rbuf(sk, copied);
++ timeo = tcp_data_wait(sk, timeo);
++ continue;
++ }
++
++ BUG_TRAP (atomic_read (&skb->users) == 1);
++
++ exhausted = eaten = 0;
++
++ offset = tp->copied_seq - TCP_SKB_CB(skb)->seq;
++ if (skb->h.th->syn)
++ offset--;
++
++ used = skb->len - offset;
++
++ if (tp->urg_data) {
++ u32 urg_offset = tp->urg_seq - tp->copied_seq;
++ if (urg_offset < used) {
++ if (!urg_offset) { /* at urgent date */
++ if (!sk->urginline) {
++ tp->copied_seq++; /* discard the single byte of urgent data */
++ offset++;
++ used--;
++ }
++ } else /* truncate read */
++ used = urg_offset;
++ }
++ }
++
++ BUG_TRAP (used >= 0);
++ if (len < used)
++ used = len;
++
++ if (used == 0)
++ exhausted = 1;
++ else
++ {
++ if (skb_is_nonlinear (skb))
++ {
++ int rc = skb_linearize (skb, GFP_KERNEL);
++
++ printk ("tcp_recvpackets(): linearising: %d\n", rc);
++
++ if (rc)
++ {
++ if (!copied)
++ copied = rc;
++ break;
++ }
++ }
++
++ if ((offset + used) == skb->len) /* consuming the whole packet */
++ {
++ __skb_unlink (skb, &sk->receive_queue);
++ dst_release (skb->dst);
++ skb_orphan (skb);
++ __skb_pull (skb, offset);
++ __skb_queue_tail (packets, skb);
++ exhausted = eaten = 1;
++ }
++ else /* consuming only part of the packet */
++ {
++ struct sk_buff *skb2 = skb_clone (skb, GFP_KERNEL);
++
++ if (skb2 == NULL)
++ {
++ if (!copied)
++ copied = -ENOMEM;
++ break;
++ }
++
++ dst_release (skb2->dst);
++ __skb_pull (skb2, offset);
++ __skb_trim (skb2, used);
++ __skb_queue_tail (packets, skb2);
++ }
++
++ tp->copied_seq += used;
++ copied += used;
++ len -= used;
++ }
++
++ if (tp->urg_data && after(tp->copied_seq,tp->urg_seq)) {
++ tp->urg_data = 0;
++ tcp_fast_path_check(sk, tp);
++ }
++
++ if (!exhausted)
++ continue;
++
++ if (skb->h.th->fin)
++ {
++ tp->copied_seq++;
++ if (!eaten)
++ tcp_eat_skb (sk, skb);
++ break;
++ }
++
++ if (!eaten)
++ tcp_eat_skb (sk, skb);
++
++ } while (len > 0);
++
++ out:
++ /* Clean up data we have read: This will do ACK frames. */
++ cleanup_rbuf(sk, copied);
++ TCP_CHECK_TIMER(sk);
++ release_sock(sk);
++ return copied;
++}
++
+ /*
+ * State processing on a close. This implements the state shift for
+ * sending our FIN frame. Note that we only send a FIN for some
kernel/ksyms.c | 1
11 files changed, 564 insertions(+), 126 deletions(-)
+Index: linux-2.4.18-p4smp/fs/dcache.c
+===================================================================
+--- linux-2.4.18-p4smp.orig/fs/dcache.c 2004-02-03 01:00:10.000000000 -0500
++++ linux-2.4.18-p4smp/fs/dcache.c 2004-03-19 16:05:42.000000000 -0500
+@@ -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.
+@@ -859,13 +866,19 @@ void d_delete(struct dentry * dentry)
+ * 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-2.4.18-p4smp/fs/exec.c
===================================================================
--- linux-2.4.18-p4smp.orig/fs/exec.c 2004-02-03 01:00:10.000000000 -0500
struct nameidata nd;
int error;
+ struct lookup_intent it = { .it_op = IT_OPEN,
-+ .it_flags = FMODE_READ|FMODE_EXEC };
++ .it_flags = FMODE_READ|FMODE_EXEC };
- error = user_path_walk(library, &nd);
+ error = user_path_walk_it(library, &nd, &it);
error = PTR_ERR(file);
if (IS_ERR(file))
goto out;
-@@ -359,8 +362,9 @@ struct file *open_exec(const char *name)
+@@ -359,8 +362,10 @@ struct file *open_exec(const char *name)
struct inode *inode;
struct file *file;
int err = 0;
--
-- err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+ struct lookup_intent it = { .it_op = IT_OPEN,
-+ .it_flags = FMODE_READ|FMODE_EXEC };
++ .it_flags = FMODE_READ|FMODE_EXEC };
+
+- err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+ err = path_lookup_it(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
file = ERR_PTR(err);
if (!err) {
goto close_fail;
retval = binfmt->core_dump(signr, regs, file);
-Index: linux-2.4.18-p4smp/fs/dcache.c
-===================================================================
---- linux-2.4.18-p4smp.orig/fs/dcache.c 2004-02-03 01:00:10.000000000 -0500
-+++ linux-2.4.18-p4smp/fs/dcache.c 2004-03-19 16:05:42.000000000 -0500
-@@ -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.
-@@ -859,13 +866,19 @@ void d_delete(struct dentry * dentry)
- * 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-2.4.18-p4smp/fs/namespace.c
-===================================================================
---- linux-2.4.18-p4smp.orig/fs/namespace.c 2004-02-03 01:00:10.000000000 -0500
-+++ linux-2.4.18-p4smp/fs/namespace.c 2004-03-19 16:05:42.000000000 -0500
-@@ -99,6 +99,7 @@
- {
- 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);
-@@ -110,6 +111,7 @@ static void attach_mnt(struct vfsmount *
- {
- 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++;
-@@ -485,14 +487,17 @@ static int do_loopback(struct nameidata
- {
- 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;
-- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
-- if (err)
-+ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
-+ if (err) {
-+ intent_release(&it);
- return err;
-+ }
-
- down_write(¤t->namespace->sem);
- err = -EINVAL;
-@@ -515,6 +520,7 @@ static int do_loopback(struct nameidata
- }
-
- up_write(¤t->namespace->sem);
-+ intent_release(&it);
- path_release(&old_nd);
- return err;
- }
-@@ -698,6 +704,7 @@ long do_mount(char * dev_name, char * di
- unsigned long flags, void *data_page)
- {
- struct nameidata nd;
-+ struct lookup_intent it = { .it_op = IT_GETATTR };
- int retval = 0;
- int mnt_flags = 0;
-
-@@ -722,10 +729,11 @@ long do_mount(char * dev_name, char * di
- flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
-
- /* ... and get the mountpoint */
-- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
-- if (retval)
-+ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
-+ if (retval) {
-+ intent_release(&it);
- return retval;
--
-+ }
- if (flags & MS_REMOUNT)
- retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
-@@ -736,6 +744,8 @@ long do_mount(char * dev_name, char * di
- else
- retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- dev_name, data_page);
-+
-+ intent_release(&it);
- path_release(&nd);
- return retval;
- }
-@@ -901,6 +911,8 @@ asmlinkage long sys_pivot_root(const cha
- {
- 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 };
- int error;
-
- if (!capable(CAP_SYS_ADMIN))
-@@ -908,14 +920,14 @@ asmlinkage long sys_pivot_root(const cha
-
- lock_kernel();
-
-- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
-+ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
- if (error)
- goto out0;
- error = -EINVAL;
- if (!check_mnt(new_nd.mnt))
- goto out1;
-
-- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
-+ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
- if (error)
- goto out1;
-
-@@ -970,8 +982,10 @@ out2:
- 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-2.4.18-p4smp/fs/namei.c
===================================================================
--- linux-2.4.18-p4smp.orig/fs/namei.c 2004-02-03 01:00:10.000000000 -0500
;
err = -ENOENT;
-@@ -548,8 +585,8 @@ int link_path_walk(const char * name, st
- if (!inode->i_op)
+@@ -549,7 +586,7 @@ int link_path_walk(const char * name, st
goto out_dput;
-- if (inode->i_op->follow_link) {
+ if (inode->i_op->follow_link) {
- err = do_follow_link(dentry, nd);
-+ if (inode->i_op->follow_link || inode->i_op->follow_link2) {
+ err = do_follow_link(dentry, nd, NULL);
dput(dentry);
if (err)
break;
continue;
/* here ends the main loop */
-@@ -592,22 +629,23 @@ last_component:
+@@ -592,22 +629,22 @@ last_component:
if (err < 0)
break;
}
;
inode = dentry->d_inode;
if ((lookup_flags & LOOKUP_FOLLOW)
-- && inode && inode->i_op && inode->i_op->follow_link) {
+ && inode && inode->i_op && inode->i_op->follow_link) {
- err = do_follow_link(dentry, nd);
-+ && inode && inode->i_op &&
-+ (inode->i_op->follow_link || inode->i_op->follow_link2)) {
+ err = do_follow_link(dentry, nd, it);
dput(dentry);
if (err)
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
if (page) {
kunmap(page);
page_cache_release(page);
+Index: linux-2.4.18-p4smp/fs/namespace.c
+===================================================================
+--- linux-2.4.18-p4smp.orig/fs/namespace.c 2004-02-03 01:00:10.000000000 -0500
++++ linux-2.4.18-p4smp/fs/namespace.c 2004-03-19 16:05:42.000000000 -0500
+@@ -99,6 +99,7 @@
+ {
+ 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);
+@@ -110,6 +111,7 @@ static void attach_mnt(struct vfsmount *
+ {
+ 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++;
+@@ -485,14 +487,17 @@ static int do_loopback(struct nameidata
+ {
+ 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;
+- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
+- if (err)
++ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
++ if (err) {
++ intent_release(&it);
+ return err;
++ }
+
+ down_write(¤t->namespace->sem);
+ err = -EINVAL;
+@@ -515,6 +520,7 @@ static int do_loopback(struct nameidata
+ }
+
+ up_write(¤t->namespace->sem);
++ intent_release(&it);
+ path_release(&old_nd);
+ return err;
+ }
+@@ -698,6 +704,7 @@ long do_mount(char * dev_name, char * di
+ unsigned long flags, void *data_page)
+ {
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+ int retval = 0;
+ int mnt_flags = 0;
+
+@@ -722,9 +729,11 @@ long do_mount(char * dev_name, char * di
+ flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
+
+ /* ... and get the mountpoint */
+- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+- if (retval)
++ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
++ if (retval) {
++ intent_release(&it);
+ return retval;
++ }
+
+ if (flags & MS_REMOUNT)
+ retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
+@@ -736,6 +744,8 @@ long do_mount(char * dev_name, char * di
+ else
+ retval = do_add_mount(&nd, type_page, flags, mnt_flags,
+ dev_name, data_page);
++
++ intent_release(&it);
+ path_release(&nd);
+ return retval;
+ }
+@@ -901,6 +911,8 @@ asmlinkage long sys_pivot_root(const cha
+ {
+ 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 };
+ int error;
+
+ if (!capable(CAP_SYS_ADMIN))
+@@ -908,14 +920,14 @@ asmlinkage long sys_pivot_root(const cha
+
+ lock_kernel();
+
+- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
++ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
+ if (error)
+ goto out0;
+ error = -EINVAL;
+ if (!check_mnt(new_nd.mnt))
+ goto out1;
+
+- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
++ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
+ if (error)
+ goto out1;
+
+@@ -970,8 +982,10 @@ out2:
+ 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-2.4.18-p4smp/fs/open.c
===================================================================
--- linux-2.4.18-p4smp.orig/fs/open.c 2004-02-03 01:00:10.000000000 -0500
path_release(&nd);
}
-@@ -385,8 +430,11 @@ asmlinkage long sys_chdir(const char * f
+@@ -385,8 +430,9 @@ asmlinkage long sys_chdir(const char * f
{
int error;
struct nameidata nd;
+ struct lookup_intent it = { .it_op = IT_GETATTR };
- error = __user_walk(filename,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd);
-+ error = __user_walk_it(filename,
-+ LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,
-+ &nd, &it);
++ error = __user_walk_it(filename,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd, &it);
if (error)
goto out;
+ void (*d_unpin)(struct dentry *, struct vfsmount *, int);
};
-+#define PIN(de,mnt,flag) if (de->d_op && de->d_op->d_pin) \
++#define PIN(de,mnt,flag) if (de && 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) \
++#define UNPIN(de,mnt,flag) if (de && de->d_op && de->d_op->d_unpin) \
+ de->d_op->d_unpin(de, mnt, flag);
+
+
#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 /* called from open path, ie O_TRUNC */
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
/*
* File types
-@@ -900,21 +908,34 @@ struct file_operations {
+@@ -900,21 +908,32 @@ struct file_operations {
struct inode_operations {
int (*create) (struct inode *,struct dentry *,int);
+ int (*rename_raw) (struct nameidata *, struct nameidata *);
int (*readlink) (struct dentry *, char *,int);
int (*follow_link) (struct dentry *, struct nameidata *);
-+ int (*follow_link2) (struct dentry *, struct nameidata *,
-+ struct lookup_intent *it);
void (*truncate) (struct inode *);
int (*permission) (struct inode *, int);
int (*revalidate) (struct dentry *);
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-2.4.18-p4smp/kernel/fork.c
-===================================================================
---- linux-2.4.18-p4smp.orig/kernel/fork.c 2004-02-03 01:00:10.000000000 -0500
-+++ linux-2.4.18-p4smp/kernel/fork.c 2004-03-19 16:05:42.000000000 -0500
-@@ -399,10 +399,13 @@
- 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-2.4.18-p4smp/kernel/exit.c
===================================================================
--- linux-2.4.18-p4smp.orig/kernel/exit.c 2004-02-03 01:00:10.000000000 -0500
dput(fs->altroot);
mntput(fs->altrootmnt);
}
+Index: linux-2.4.18-p4smp/kernel/fork.c
+===================================================================
+--- linux-2.4.18-p4smp.orig/kernel/fork.c 2004-02-03 01:00:10.000000000 -0500
++++ linux-2.4.18-p4smp/kernel/fork.c 2004-03-19 16:05:42.000000000 -0500
+@@ -399,10 +399,13 @@
+ 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-2.4.18-p4smp/kernel/ksyms.c
===================================================================
--- linux-2.4.18-p4smp.orig/kernel/ksyms.c 2004-03-19 16:05:40.000000000 -0500
kernel/ksyms.c | 1
12 files changed, 558 insertions(+), 128 deletions(-)
+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/exec.c
===================================================================
--- linux-2.4.19-pre1.orig/fs/exec.c 2003-11-21 02:41:00.000000000 +0300
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
if (!dentry) {
- dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE);
+ dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE,
-+ NULL);
++ NULL);
err = PTR_ERR(dentry);
if (IS_ERR(dentry))
break;
break;
}
goto return_base;
-@@ -625,21 +663,66 @@
+@@ -625,21 +663,68 @@
nd->last_type = LAST_DOT;
else if (this.len == 2 && this.name[1] == '.')
nd->last_type = LAST_DOTDOT;
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
+ }
+ nd->dentry = new;
+ }
++ if (!nd->dentry->d_inode)
++ goto no_inode;
+ } else
+ if (dentry && dentry->d_op && dentry->d_op->d_revalidate) {
+ err = -ESTALE;
if (IS_ERR(dentry))
goto fail;
if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-@@ -1252,7 +1392,16 @@
+@@ -1252,7 +1392,20 @@
error = path_walk(tmp, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
+
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
path_release(&nd);
out:
putname(tmp);
-@@ -1321,7 +1471,14 @@
+@@ -1321,7 +1471,18 @@
error = path_walk(tmp, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 1);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
error = PTR_ERR(dentry);
if (!IS_ERR(dentry)) {
/* Why not before? Because we want correct error value */
-@@ -1557,15 +1730,23 @@
+@@ -1557,15 +1730,27 @@
error = path_walk(to, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
putname(to);
}
putname(from);
-@@ -1648,7 +1829,14 @@
+@@ -1648,7 +1829,18 @@
error = -EXDEV;
if (old_nd.mnt != nd.mnt)
goto out_release;
- new_dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out_release;
++ }
+ 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);
if (page) {
kunmap(page);
page_cache_release(page);
+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/open.c
===================================================================
--- linux-2.4.19-pre1.orig/fs/open.c 2003-11-21 02:41:00.000000000 +0300
===================================================================
--- 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 @@
+@@ -17,10 +17,12 @@
* Revalidate the inode. This is required for proper NFS attribute caching.
*/
static __inline__ int
{
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)
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
dput(old_pwd);
mntput(old_pwdmnt);
}
-Index: linux-2.4.19-pre1/kernel/ksyms.c
+Index: linux-2.4.19-pre1/kernel/exit.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);
+--- 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);
+ }
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
fs->altrootmnt = mntget(old->altrootmnt);
fs->altroot = dget(old->altroot);
} else {
-Index: linux-2.4.19-pre1/kernel/exit.c
+Index: linux-2.4.19-pre1/kernel/ksyms.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);
- }
+--- 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);
kernel/ksyms.c | 1
12 files changed, 558 insertions(+), 128 deletions(-)
+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/exec.c
===================================================================
--- linux-2.4.19.SuSE.orig/fs/exec.c Mon Jan 27 05:08:35 2003
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
}
/* This does the actual lookups.. */
- dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE);
-+ dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE,NULL);
++ dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE, NULL);
if (!dentry) {
err = -EWOULDBLOCKIO;
if (atomic)
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
+ }
+ nd->dentry = new;
+ }
++ if (!nd->dentry->d_inode)
++ goto no_inode;
+ } else
if (dentry && dentry->d_op && dentry->d_op->d_revalidate) {
err = -ESTALE;
if (IS_ERR(dentry))
goto fail;
if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-@@ -1286,7 +1414,16 @@
+@@ -1286,7 +1414,20 @@
error = path_walk(tmp, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
+
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
path_release(&nd);
out:
putname(tmp);
-@@ -1356,7 +1494,14 @@
+@@ -1356,7 +1494,18 @@
error = path_walk(tmp, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 1);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
error = PTR_ERR(dentry);
if (!IS_ERR(dentry)) {
/* Why not before? Because we want correct error value */
-@@ -1595,15 +1756,23 @@
+@@ -1595,15 +1756,27 @@
error = path_walk(to, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
if (old_nd.mnt != nd.mnt)
goto out_release;
- new_dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out_release;
++ }
+ 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);
if (page) {
kunmap(page);
page_cache_release(page);
+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/open.c
===================================================================
--- linux-2.4.19.SuSE.orig/fs/open.c Mon Jan 27 05:08:00 2003
===================================================================
--- 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 @@
+@@ -17,10 +17,16 @@
* Revalidate the inode. This is required for proper NFS attribute caching.
*/
static __inline__ int
{
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)
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
dput(old_pwd);
mntput(old_pwdmnt);
}
-Index: linux-2.4.19.SuSE/kernel/ksyms.c
+Index: linux-2.4.19.SuSE/kernel/exit.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);
+--- 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);
+ }
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
fs->altrootmnt = mntget(old->altrootmnt);
fs->altroot = dget(old->altroot);
} else {
-Index: linux-2.4.19.SuSE/kernel/exit.c
+Index: linux-2.4.19.SuSE/kernel/ksyms.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);
- }
+--- 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);
kernel/ksyms.c | 1
13 files changed, 591 insertions(+), 133 deletions(-)
+Index: linux/fs/dcache.c
+===================================================================
+--- linux.orig/fs/dcache.c Thu Nov 28 18:53:15 2002
++++ linux/fs/dcache.c Wed Mar 17 13:11:25 2004
+@@ -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.
+@@ -830,13 +837,19 @@
+ * 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/fs/exec.c
===================================================================
--- linux.orig/fs/exec.c Wed Mar 17 13:00:38 2004
goto close_fail;
retval = binfmt->core_dump(signr, regs, file);
-Index: linux/fs/dcache.c
-===================================================================
---- linux.orig/fs/dcache.c Thu Nov 28 18:53:15 2002
-+++ linux/fs/dcache.c Wed Mar 17 13:11:25 2004
-@@ -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.
-@@ -830,13 +837,19 @@
- * 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/fs/namespace.c
-===================================================================
---- linux.orig/fs/namespace.c Thu Nov 28 18:53:15 2002
-+++ linux/fs/namespace.c Wed Mar 17 13:11:25 2004
-@@ -99,6 +99,7 @@
- {
- 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);
-@@ -110,6 +111,7 @@
- {
- 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++;
-@@ -485,14 +487,17 @@
- {
- 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;
-- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
-- if (err)
-+ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
-+ if (err) {
-+ intent_release(&it);
- return err;
-+ }
-
- down_write(¤t->namespace->sem);
- err = -EINVAL;
-@@ -515,6 +520,7 @@
- }
-
- up_write(¤t->namespace->sem);
-+ intent_release(&it);
- path_release(&old_nd);
- return err;
- }
-@@ -698,6 +704,7 @@
- unsigned long flags, void *data_page)
- {
- struct nameidata nd;
-+ struct lookup_intent it = { .it_op = IT_GETATTR };
- int retval = 0;
- int mnt_flags = 0;
-
-@@ -722,10 +729,11 @@
- flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
-
- /* ... and get the mountpoint */
-- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
-- if (retval)
-+ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
-+ if (retval) {
-+ intent_release(&it);
- return retval;
--
-+ }
- if (flags & MS_REMOUNT)
- retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
-@@ -736,6 +744,8 @@
- else
- retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- dev_name, data_page);
-+
-+ intent_release(&it);
- path_release(&nd);
- return retval;
- }
-@@ -901,6 +911,8 @@
- {
- 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 };
- int error;
-
- if (!capable(CAP_SYS_ADMIN))
-@@ -908,14 +920,14 @@
-
- lock_kernel();
-
-- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
-+ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
- if (error)
- goto out0;
- error = -EINVAL;
- if (!check_mnt(new_nd.mnt))
- goto out1;
-
-- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
-+ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
- if (error)
- goto out1;
-
-@@ -970,8 +982,10 @@
- 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/fs/namei.c
===================================================================
--- linux.orig/fs/namei.c Wed Mar 17 13:00:37 2004
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
if (page) {
kunmap(page);
page_cache_release(page);
+Index: linux/fs/namespace.c
+===================================================================
+--- linux.orig/fs/namespace.c Thu Nov 28 18:53:15 2002
++++ linux/fs/namespace.c Wed Mar 17 13:11:25 2004
+@@ -99,6 +99,7 @@
+ {
+ 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);
+@@ -110,6 +111,7 @@
+ {
+ 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++;
+@@ -485,14 +487,17 @@
+ {
+ 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;
+- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
+- if (err)
++ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
++ if (err) {
++ intent_release(&it);
+ return err;
++ }
+
+ down_write(¤t->namespace->sem);
+ err = -EINVAL;
+@@ -515,6 +520,7 @@
+ }
+
+ up_write(¤t->namespace->sem);
++ intent_release(&it);
+ path_release(&old_nd);
+ return err;
+ }
+@@ -698,6 +704,7 @@
+ unsigned long flags, void *data_page)
+ {
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+ int retval = 0;
+ int mnt_flags = 0;
+
+@@ -722,10 +729,11 @@
+ flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
+
+ /* ... and get the mountpoint */
+- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+- if (retval)
++ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
++ if (retval) {
++ intent_release(&it);
+ return retval;
+-
++ }
+ if (flags & MS_REMOUNT)
+ retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
+ data_page);
+@@ -736,6 +744,8 @@
+ else
+ retval = do_add_mount(&nd, type_page, flags, mnt_flags,
+ dev_name, data_page);
++
++ intent_release(&it);
+ path_release(&nd);
+ return retval;
+ }
+@@ -901,6 +911,8 @@
+ {
+ 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 };
+ int error;
+
+ if (!capable(CAP_SYS_ADMIN))
+@@ -908,14 +920,14 @@
+
+ lock_kernel();
+
+- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
++ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
+ if (error)
+ goto out0;
+ error = -EINVAL;
+ if (!check_mnt(new_nd.mnt))
+ goto out1;
+
+- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
++ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
+ if (error)
+ goto out1;
+
+@@ -970,8 +982,10 @@
+ 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/fs/open.c
===================================================================
--- linux.orig/fs/open.c Thu Nov 28 18:53:15 2002
/*
* Find an empty file descriptor entry, and mark it busy.
*/
+Index: linux/fs/proc/base.c
+===================================================================
+--- linux.orig/fs/proc/base.c Wed Mar 17 13:00:35 2004
++++ linux/fs/proc/base.c Wed Mar 17 13:11:25 2004
+@@ -481,6 +481,9 @@
+
+ error = inode->u.proc_i.op.proc_get_link(inode, &nd->dentry, &nd->mnt);
+ nd->last_type = LAST_BIND;
++
++ if (nd->intent != NULL)
++ nd->intent->d.lustre.it_int_flags |= IT_FL_FOLLOWED;
+ out:
+ return error;
+ }
Index: linux/fs/stat.c
===================================================================
--- linux.orig/fs/stat.c Thu Sep 13 19:04:43 2001
if (!err)
err = cp_new_stat64(dentry->d_inode, statbuf);
fput(f);
-Index: linux/fs/proc/base.c
-===================================================================
---- linux.orig/fs/proc/base.c Wed Mar 17 13:00:35 2004
-+++ linux/fs/proc/base.c Wed Mar 17 13:11:25 2004
-@@ -481,6 +481,9 @@
-
- error = inode->u.proc_i.op.proc_get_link(inode, &nd->dentry, &nd->mnt);
- nd->last_type = LAST_BIND;
-+
-+ if (nd->intent != NULL)
-+ nd->intent->d.lustre.it_int_flags |= IT_FL_FOLLOWED;
- out:
- return error;
- }
Index: linux/include/linux/dcache.h
===================================================================
--- linux.orig/include/linux/dcache.h Thu Nov 28 18:53:15 2002
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
dput(old_pwd);
mntput(old_pwdmnt);
}
-Index: linux/kernel/ksyms.c
+Index: linux/kernel/exit.c
===================================================================
---- linux.orig/kernel/ksyms.c Wed Mar 17 13:11:23 2004
-+++ linux/kernel/ksyms.c Wed Mar 17 13:11:25 2004
-@@ -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);
+--- linux.orig/kernel/exit.c Wed Mar 17 13:00:38 2004
++++ linux/kernel/exit.c Wed Mar 17 13:11:25 2004
+@@ -239,11 +239,14 @@
+ {
+ /* 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);
+ }
Index: linux/kernel/fork.c
===================================================================
--- linux.orig/kernel/fork.c Wed Mar 17 13:00:38 2004
fs->altrootmnt = mntget(old->altrootmnt);
fs->altroot = dget(old->altroot);
} else {
-Index: linux/kernel/exit.c
+Index: linux/kernel/ksyms.c
===================================================================
---- linux.orig/kernel/exit.c Wed Mar 17 13:00:38 2004
-+++ linux/kernel/exit.c Wed Mar 17 13:11:25 2004
-@@ -239,11 +239,14 @@
- {
- /* 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);
- }
+--- linux.orig/kernel/ksyms.c Wed Mar 17 13:11:23 2004
++++ linux/kernel/ksyms.c Wed Mar 17 13:11:25 2004
+@@ -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);
{
struct dentry *dentry;
struct inode *inode;
-@@ -526,19 +579,18 @@
+@@ -526,18 +579,18 @@
break;
}
/* This does the actual lookups.. */
}
/* Check mountpoints.. */
- while (d_mountpoint(dentry) && __follow_down(&nd->mnt, &dentry))
-- ;
-+ while (d_mountpoint(dentry) && __follow_down(&nd->mnt, &dentry, NULL));
++ while (d_mountpoint(dentry) && __follow_down(&nd->mnt, &dentry, NULL))
+ ;
err = -ENOENT;
- inode = dentry->d_inode;
@@ -549,7 +601,7 @@
goto out_dput;
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
int retval = 0;
int mnt_flags = 0;
-@@ -722,10 +729,11 @@
+@@ -722,9 +729,11 @@
flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
/* ... and get the mountpoint */
+ if (retval) {
+ intent_release(&it);
return retval;
--
+ }
+
if (flags & MS_REMOUNT)
retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
@@ -736,6 +744,8 @@
else
retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
+ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
-+ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
if (error)
goto out;
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
dput(old_pwd);
mntput(old_pwdmnt);
}
-Index: linux-2.4.20/kernel/ksyms.c
+Index: linux-2.4.20/kernel/exit.c
===================================================================
---- linux-2.4.20.orig/kernel/ksyms.c Wed Mar 17 13:57:11 2004
-+++ linux-2.4.20/kernel/ksyms.c Wed Mar 17 13:57:11 2004
-@@ -297,6 +297,7 @@
- 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);
+--- linux-2.4.20.orig/kernel/exit.c Wed Mar 17 13:57:05 2004
++++ linux-2.4.20/kernel/exit.c Wed Mar 17 13:57:11 2004
+@@ -345,11 +345,14 @@
+ {
+ /* 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);
+ }
Index: linux-2.4.20/kernel/fork.c
===================================================================
--- linux-2.4.20.orig/kernel/fork.c Wed Mar 17 13:57:05 2004
fs->altrootmnt = mntget(old->altrootmnt);
fs->altroot = dget(old->altroot);
} else {
-Index: linux-2.4.20/kernel/exit.c
+Index: linux-2.4.20/kernel/ksyms.c
===================================================================
---- linux-2.4.20.orig/kernel/exit.c Wed Mar 17 13:57:05 2004
-+++ linux-2.4.20/kernel/exit.c Wed Mar 17 13:57:11 2004
-@@ -345,11 +345,14 @@
- {
- /* 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);
- }
+--- linux-2.4.20.orig/kernel/ksyms.c Wed Mar 17 13:57:11 2004
++++ linux-2.4.20/kernel/ksyms.c Wed Mar 17 13:57:11 2004
+@@ -297,6 +297,7 @@
+ 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);
kernel/ksyms.c | 1
12 files changed, 558 insertions(+), 128 deletions(-)
+Index: linux-2.4.24/fs/dcache.c
+===================================================================
+--- linux-2.4.24.orig/fs/dcache.c Fri Jun 13 07:51:37 2003
++++ linux-2.4.24/fs/dcache.c Wed Mar 17 17:36:14 2004
+@@ -181,6 +181,13 @@ int d_invalidate(struct dentry * dentry)
+ 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.
+@@ -830,13 +837,19 @@ void d_delete(struct dentry * dentry)
+ * 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-2.4.24/fs/exec.c
===================================================================
--- linux-2.4.24.orig/fs/exec.c Fri Nov 28 10:26:21 2003
goto close_fail;
retval = binfmt->core_dump(signr, regs, file);
-Index: linux-2.4.24/fs/dcache.c
-===================================================================
---- linux-2.4.24.orig/fs/dcache.c Fri Jun 13 07:51:37 2003
-+++ linux-2.4.24/fs/dcache.c Wed Mar 17 17:36:14 2004
-@@ -181,6 +181,13 @@ int d_invalidate(struct dentry * dentry)
- 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.
-@@ -830,13 +837,19 @@ void d_delete(struct dentry * dentry)
- * 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-2.4.24/fs/namespace.c
-===================================================================
---- linux-2.4.24.orig/fs/namespace.c Fri Nov 28 10:26:21 2003
-+++ linux-2.4.24/fs/namespace.c Wed Mar 17 17:36:14 2004
-@@ -98,6 +98,7 @@ static void detach_mnt(struct vfsmount *
- {
- 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);
-@@ -109,6 +110,7 @@ static void attach_mnt(struct vfsmount *
- {
- 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_tail(&mnt->mnt_child, &nd->mnt->mnt_mounts);
- nd->dentry->d_mounted++;
-@@ -488,14 +490,17 @@ static int do_loopback(struct nameidata
- {
- 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;
-- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
-- if (err)
-+ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
-+ if (err) {
-+ intent_release(&it);
- return err;
-+ }
-
- down_write(¤t->namespace->sem);
- err = -EINVAL;
-@@ -518,6 +523,7 @@ static int do_loopback(struct nameidata
- }
-
- up_write(¤t->namespace->sem);
-+ intent_release(&it);
- path_release(&old_nd);
- return err;
- }
-@@ -701,6 +707,7 @@ long do_mount(char * dev_name, char * di
- unsigned long flags, void *data_page)
- {
- struct nameidata nd;
-+ struct lookup_intent it = { .it_op = IT_GETATTR };
- int retval = 0;
- int mnt_flags = 0;
-
-@@ -725,10 +732,11 @@ long do_mount(char * dev_name, char * di
- flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
-
- /* ... and get the mountpoint */
-- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
-- if (retval)
-+ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
-+ if (retval) {
-+ intent_release(&it);
- return retval;
--
-+ }
- if (flags & MS_REMOUNT)
- retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
-@@ -739,6 +747,8 @@ long do_mount(char * dev_name, char * di
- else
- retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- dev_name, data_page);
-+
-+ intent_release(&it);
- path_release(&nd);
- return retval;
- }
-@@ -904,6 +914,8 @@ asmlinkage long sys_pivot_root(const cha
- {
- 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 };
- int error;
-
- if (!capable(CAP_SYS_ADMIN))
-@@ -911,14 +923,14 @@ asmlinkage long sys_pivot_root(const cha
-
- lock_kernel();
-
-- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
-+ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
- if (error)
- goto out0;
- error = -EINVAL;
- if (!check_mnt(new_nd.mnt))
- goto out1;
-
-- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
-+ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
- if (error)
- goto out1;
-
-@@ -973,8 +985,10 @@ out2:
- 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-2.4.24/fs/namei.c
===================================================================
--- linux-2.4.24.orig/fs/namei.c Mon Aug 25 04:44:43 2003
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
if (page) {
kunmap(page);
page_cache_release(page);
+Index: linux-2.4.24/fs/namespace.c
+===================================================================
+--- linux-2.4.24.orig/fs/namespace.c Fri Nov 28 10:26:21 2003
++++ linux-2.4.24/fs/namespace.c Wed Mar 17 17:36:14 2004
+@@ -98,6 +98,7 @@ static void detach_mnt(struct vfsmount *
+ {
+ 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);
+@@ -109,6 +110,7 @@ static void attach_mnt(struct vfsmount *
+ {
+ 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_tail(&mnt->mnt_child, &nd->mnt->mnt_mounts);
+ nd->dentry->d_mounted++;
+@@ -488,14 +490,17 @@ static int do_loopback(struct nameidata
+ {
+ 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;
+- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
+- if (err)
++ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
++ if (err) {
++ intent_release(&it);
+ return err;
++ }
+
+ down_write(¤t->namespace->sem);
+ err = -EINVAL;
+@@ -518,6 +523,7 @@ static int do_loopback(struct nameidata
+ }
+
+ up_write(¤t->namespace->sem);
++ intent_release(&it);
+ path_release(&old_nd);
+ return err;
+ }
+@@ -701,6 +707,7 @@ long do_mount(char * dev_name, char * di
+ unsigned long flags, void *data_page)
+ {
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+ int retval = 0;
+ int mnt_flags = 0;
+
+@@ -725,9 +732,11 @@ long do_mount(char * dev_name, char * di
+ flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
+
+ /* ... and get the mountpoint */
+- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+- if (retval)
++ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
++ if (retval) {
++ intent_release(&it);
+ return retval;
++ }
+
+ if (flags & MS_REMOUNT)
+ retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
+@@ -739,6 +747,8 @@ long do_mount(char * dev_name, char * di
+ else
+ retval = do_add_mount(&nd, type_page, flags, mnt_flags,
+ dev_name, data_page);
++
++ intent_release(&it);
+ path_release(&nd);
+ return retval;
+ }
+@@ -904,6 +914,8 @@ asmlinkage long sys_pivot_root(const cha
+ {
+ 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 };
+ int error;
+
+ if (!capable(CAP_SYS_ADMIN))
+@@ -911,14 +923,14 @@ asmlinkage long sys_pivot_root(const cha
+
+ lock_kernel();
+
+- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
++ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
+ if (error)
+ goto out0;
+ error = -EINVAL;
+ if (!check_mnt(new_nd.mnt))
+ goto out1;
+
+- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
++ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
+ if (error)
+ goto out1;
+
+@@ -973,8 +985,10 @@ out2:
+ 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-2.4.24/fs/open.c
===================================================================
--- linux-2.4.24.orig/fs/open.c Mon Aug 25 04:44:43 2003
- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
+ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
-+ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
if (error)
goto out;
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
dput(old_pwd);
mntput(old_pwdmnt);
}
-Index: linux-2.4.24/kernel/ksyms.c
+Index: linux-2.4.24/kernel/exit.c
===================================================================
---- linux-2.4.24.orig/kernel/ksyms.c Wed Mar 17 17:36:14 2004
-+++ linux-2.4.24/kernel/ksyms.c Wed Mar 17 17:36:14 2004
-@@ -275,6 +275,7 @@ EXPORT_SYMBOL(set_page_dirty);
- EXPORT_SYMBOL(mark_page_accessed);
- 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);
+--- linux-2.4.24.orig/kernel/exit.c Thu Nov 28 15:53:15 2002
++++ linux-2.4.24/kernel/exit.c Wed Mar 17 17:36:14 2004
+@@ -238,11 +238,14 @@ static inline void __put_fs_struct(struc
+ {
+ /* 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);
+ }
Index: linux-2.4.24/kernel/fork.c
===================================================================
--- linux-2.4.24.orig/kernel/fork.c Fri Nov 28 10:26:21 2003
fs->altrootmnt = mntget(old->altrootmnt);
fs->altroot = dget(old->altroot);
} else {
-Index: linux-2.4.24/kernel/exit.c
+Index: linux-2.4.24/kernel/ksyms.c
===================================================================
---- linux-2.4.24.orig/kernel/exit.c Thu Nov 28 15:53:15 2002
-+++ linux-2.4.24/kernel/exit.c Wed Mar 17 17:36:14 2004
-@@ -238,11 +238,14 @@ static inline void __put_fs_struct(struc
- {
- /* 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);
- }
+--- linux-2.4.24.orig/kernel/ksyms.c Wed Mar 17 17:36:14 2004
++++ linux-2.4.24/kernel/ksyms.c Wed Mar 17 17:36:14 2004
+@@ -275,6 +275,7 @@ EXPORT_SYMBOL(set_page_dirty);
+ EXPORT_SYMBOL(mark_page_accessed);
+ 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);
+ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) {
+ struct dentry *new;
+ err = permission(dentry->d_parent->d_inode,
-+ MAY_EXEC);
++ MAY_EXEC);
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
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, it);
++ error = vfs_create_it(dir->d_inode, dentry, mode, it);
up(&dir->d_inode->i_sem);
dput(nd->dentry);
nd->dentry = dentry;
int retval = 0;
int mnt_flags = 0;
-@@ -725,10 +732,11 @@ long do_mount(char * dev_name, char * di
+@@ -725,9 +732,11 @@ long do_mount(char * dev_name, char * di
flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
/* ... and get the mountpoint */
+ if (retval) {
+ intent_release(&it);
return retval;
--
+ }
+
if (flags & MS_REMOUNT)
retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
@@ -739,6 +747,8 @@ long do_mount(char * dev_name, char * di
else
retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
+ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
-+ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
if (error)
goto out;
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
--- /dev/null
+ 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-2.4.21/fs/dcache.c
+===================================================================
+--- linux-2.4.21.orig/fs/dcache.c 2004-04-24 02:38:00.000000000 -0400
++++ linux-2.4.21/fs/dcache.c 2004-04-26 19:06:31.000000000 -0400
+@@ -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-2.4.21/fs/exec.c
+===================================================================
+--- linux-2.4.21.orig/fs/exec.c 2004-04-24 02:39:01.000000000 -0400
++++ linux-2.4.21/fs/exec.c 2004-04-26 19:06:31.000000000 -0400
+@@ -113,8 +113,10 @@
+ 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;
+
+@@ -126,7 +128,8 @@
+ 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;
+@@ -383,8 +386,10 @@
+ struct inode *inode;
+ struct file *file;
+ int err = 0;
++ struct lookup_intent it = { .it_op = IT_OPEN,
++ .it_flags = FMODE_READ|FMODE_EXEC };
+
+- err = path_lookup(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
++ err = path_lookup_it(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
+ file = ERR_PTR(err);
+ if (!err) {
+ inode = nd.dentry->d_inode;
+@@ -396,7 +401,8 @@
+ 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) {
+@@ -408,6 +414,7 @@
+ return file;
+ }
+ }
++ intent_release(&it);
+ path_release(&nd);
+ }
+ goto out;
+@@ -1147,7 +1154,7 @@
+ 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-2.4.21/fs/namei.c
+===================================================================
+--- linux-2.4.21.orig/fs/namei.c 2004-04-24 02:39:02.000000000 -0400
++++ linux-2.4.21/fs/namei.c 2004-04-26 19:06:38.000000000 -0400
+@@ -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..
+@@ -274,10 +281,19 @@
+ * 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);
+@@ -295,11 +311,15 @@
+ * 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
+@@ -314,6 +334,9 @@
+ 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)
+@@ -335,6 +358,15 @@
+ 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;
+ }
+@@ -346,7 +378,8 @@
+ * 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)
+@@ -360,10 +393,12 @@
+ 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;
+ }
+@@ -462,7 +497,8 @@
+ * We expect 'base' to be positive and a directory.
+ */
+ static inline int __attribute__((always_inline))
+-__link_path_walk(const char * name, struct nameidata *nd)
++__link_path_walk_it(const char * name, struct nameidata *nd,
++ struct lookup_intent *it)
+ {
+ struct dentry *dentry;
+ struct inode *inode;
+@@ -539,12 +575,12 @@
+ 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;
+@@ -562,7 +598,7 @@
+ 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;
+@@ -578,7 +614,7 @@
+ 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 */
+@@ -605,12 +641,12 @@
+ 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;
+@@ -620,7 +656,7 @@
+ 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;
+@@ -634,7 +670,8 @@
+ 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;
+@@ -658,6 +695,27 @@
+ * 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, it);
++ d_invalidate(dentry);
++ dput(dentry);
++ if (IS_ERR(new)) {
++ err = PTR_ERR(new);
++ break;
++ }
++ nd->dentry = new;
++ }
++ if (!nd->dentry->d_inode)
++ goto no_inode;
++ } else
+ if (dentry && dentry->d_op && dentry->d_op->d_revalidate) {
+ err = -ESTALE;
+ if (!dentry->d_op->d_revalidate(dentry, lookup_flags & LOOKUP_PARENT)) {
+@@ -671,6 +729,8 @@
+ dput(dentry);
+ break;
+ }
++ if (err)
++ intent_release(it);
+ path_release(nd);
+ return_err:
+ return err;
+@@ -678,13 +738,13 @@
+
+ int link_path_walk(const char * name, struct nameidata *nd)
+ {
+- return __link_path_walk(name,nd);
++ return __link_path_walk_it(name, nd, NULL);
+ }
+
+ static inline 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);
+ }
+
+ int path_walk(const char * name, struct nameidata *nd)
+@@ -692,6 +752,12 @@
+ return __path_walk(name, nd);
+ }
+
++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);
++}
++
+ /* SMP-safe */
+ /* returns 1 if everything is done */
+ static int __emul_lookup_dentry(const char *name, struct nameidata *nd)
+@@ -774,6 +840,17 @@
+ }
+
+ /* SMP-safe */
++int path_lookup_it(const char *path, unsigned flags, struct nameidata *nd,
++ struct lookup_intent *it)
++{
++ int error = 0;
++ if (path_init(path, flags, nd))
++ error = path_walk_it(path, nd, it);
++ return error;
++}
++
++
++/* SMP-safe */
+ int path_lookup(const char *path, unsigned flags, struct nameidata *nd)
+ {
+ int error = 0;
+@@ -788,6 +865,7 @@
+ {
+ 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);
+@@ -802,7 +880,8 @@
+ * 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;
+@@ -825,13 +904,16 @@
+ 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)
+@@ -843,6 +925,12 @@
+ 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)
+ {
+@@ -864,7 +952,7 @@
+ }
+ this.hash = end_name_hash(hash);
+
+- return lookup_hash(&this, base);
++ return lookup_hash_it(&this, base, NULL);
+ access:
+ return ERR_PTR(-EACCES);
+ }
+@@ -895,6 +983,23 @@
+ 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.
+@@ -992,7 +1097,8 @@
+ 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;
+
+@@ -1005,12 +1111,15 @@
+ 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);
+@@ -1019,6 +1128,11 @@
+ return error;
+ }
+
++int vfs_create(struct inode *dir, struct dentry *dentry, int mode)
++{
++ return vfs_create_it(dir, dentry, mode, NULL);
++}
++
+ /*
+ * open_namei()
+ *
+@@ -1033,7 +1147,8 @@
+ * 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;
+@@ -1043,11 +1158,14 @@
+
+ acc_mode = ACC_MODE(flag);
+
++ if (it)
++ it->it_flags = flag;
++
+ /*
+ * The simplest case - just a plain lookup.
+ */
+ if (!(flag & O_CREAT)) {
+- error = path_lookup(pathname, lookup_flags(flag), nd);
++ error = path_lookup_it(pathname, lookup_flags(flag), nd, it);
+ if (error)
+ return error;
+ dentry = nd->dentry;
+@@ -1057,6 +1175,10 @@
+ /*
+ * Create - we need to know the parent.
+ */
++ if (it) {
++ it->it_create_mode = mode;
++ it->it_op |= IT_CREAT;
++ }
+ error = path_lookup(pathname, LOOKUP_PARENT, nd);
+ if (error)
+ return error;
+@@ -1072,7 +1194,7 @@
+
+ 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);
+@@ -1081,11 +1203,12 @@
+ 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, it);
+ up(&dir->d_inode->i_sem);
+ #ifndef DENTRY_WASTE_RAM
+ if (error)
+@@ -1193,7 +1316,7 @@
+ if (!error) {
+ DQUOT_INIT(inode);
+
+- error = do_truncate(dentry, 0);
++ error = do_truncate(dentry, 0, 1);
+ }
+ put_write_access(inode);
+ if (error)
+@@ -1205,8 +1328,10 @@
+ return 0;
+
+ exit_dput:
++ intent_release(it);
+ dput(dentry);
+ exit:
++ intent_release(it);
+ path_release(nd);
+ return error;
+
+@@ -1225,7 +1350,10 @@
+ * 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;
+@@ -1247,13 +1375,20 @@
+ }
+ 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 */
+-struct dentry *lookup_create(struct nameidata *nd, int is_dir)
++struct dentry *lookup_create(struct nameidata *nd, int is_dir,
++ struct lookup_intent *it)
+ {
+ struct dentry *dentry;
+
+@@ -1261,7 +1396,7 @@
+ 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)
+@@ -1317,7 +1452,20 @@
+ error = path_lookup(tmp, LOOKUP_PARENT, &nd);
+ if (error)
+ goto out;
+- dentry = lookup_create(&nd, 0);
++
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
++ 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))
+@@ -1339,6 +1487,7 @@
+ dput(dentry);
+ }
+ up(&nd.dentry->d_inode->i_sem);
++out2:
+ path_release(&nd);
+ out:
+ putname(tmp);
+@@ -1386,7 +1535,18 @@
+ error = path_lookup(tmp, LOOKUP_PARENT, &nd);
+ if (error)
+ goto out;
+- dentry = lookup_create(&nd, 1);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
++ 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))
+@@ -1395,6 +1555,7 @@
+ dput(dentry);
+ }
+ up(&nd.dentry->d_inode->i_sem);
++out2:
+ path_release(&nd);
+ out:
+ putname(tmp);
+@@ -1495,8 +1656,16 @@
+ 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);
+@@ -1554,8 +1723,15 @@
+ 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 */
+@@ -1622,15 +1798,27 @@
+ error = path_lookup(to, LOOKUP_PARENT, &nd);
+ if (error)
+ goto out;
+- dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
++ 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);
+@@ -1706,7 +1894,18 @@
+ error = -EXDEV;
+ if (old_nd.mnt != nd.mnt)
+ goto out_release;
+- new_dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out_release;
++ }
++ 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);
+@@ -1750,7 +1949,7 @@
+ * 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;
+@@ -1829,7 +2028,7 @@
+ }
+
+ 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;
+
+@@ -1917,9 +2116,18 @@
+ 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;
+@@ -1935,16 +2143,16 @@
+ 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);
+@@ -1995,20 +2203,26 @@
+ }
+
+ static inline int __attribute__((always_inline))
+-__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;
+@@ -2032,7 +2246,13 @@
+
+ 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 */
+@@ -2074,7 +2294,7 @@
+ {
+ 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-2.4.21/fs/namespace.c
+===================================================================
+--- linux-2.4.21.orig/fs/namespace.c 2004-04-24 02:38:41.000000000 -0400
++++ linux-2.4.21/fs/namespace.c 2004-04-26 19:06:32.000000000 -0400
+@@ -98,6 +98,7 @@
+ {
+ 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);
+@@ -109,6 +110,7 @@
+ {
+ 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++;
+@@ -488,14 +490,17 @@
+ {
+ 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;
+- err = path_lookup(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd);
+- if (err)
++ err = path_lookup_it(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd, &it);
++ if (err) {
++ intent_release(&it);
+ return err;
++ }
+
+ down_write(¤t->namespace->sem);
+ err = -EINVAL;
+@@ -518,6 +523,7 @@
+ }
+
+ up_write(¤t->namespace->sem);
++ intent_release(&it);
+ path_release(&old_nd);
+ return err;
+ }
+@@ -701,6 +707,7 @@
+ unsigned long flags, void *data_page)
+ {
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+ int retval = 0;
+ int mnt_flags = 0;
+
+@@ -725,9 +732,11 @@
+ flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
+
+ /* ... and get the mountpoint */
+- retval = path_lookup(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd);
+- if (retval)
++ retval = path_lookup_it(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd, &it);
++ if (retval) {
++ intent_release(&it);
+ return retval;
++ }
+
+ if (flags & MS_REMOUNT)
+ retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
+@@ -739,6 +748,8 @@
+ else
+ retval = do_add_mount(&nd, type_page, flags, mnt_flags,
+ dev_name, data_page);
++
++ intent_release(&it);
+ path_release(&nd);
+ return retval;
+ }
+@@ -904,6 +915,8 @@
+ {
+ 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 };
+ int error;
+
+ if (!capable(CAP_SYS_ADMIN))
+@@ -911,14 +924,14 @@
+
+ lock_kernel();
+
+- error = __user_walk(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd);
++ error = __user_walk_it(new_root, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd, &new_it);
+ if (error)
+ goto out0;
+ error = -EINVAL;
+ if (!check_mnt(new_nd.mnt))
+ goto out1;
+
+- error = __user_walk(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd);
++ error = __user_walk_it(put_old, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd, &old_it);
+ if (error)
+ goto out1;
+
+@@ -973,8 +986,10 @@
+ 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-2.4.21/fs/open.c
+===================================================================
+--- linux-2.4.21.orig/fs/open.c 2004-04-24 02:39:01.000000000 -0400
++++ linux-2.4.21/fs/open.c 2004-04-26 19:06:32.000000000 -0400
+@@ -20,6 +20,8 @@
+ #include <asm/uaccess.h>
+
+ #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)
+ {
+@@ -96,9 +98,10 @@
+ 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;
+
+@@ -110,7 +113,13 @@
+ 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);
+ up_write(&inode->i_alloc_sem);
+ return error;
+@@ -121,12 +130,13 @@
+ 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;
+@@ -166,11 +176,13 @@
+ 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;
+@@ -218,7 +230,7 @@
+
+ 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:
+@@ -263,11 +275,13 @@
+ 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;
+@@ -282,11 +296,25 @@
+ 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);
+@@ -307,12 +335,14 @@
+ 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;
+@@ -327,7 +357,20 @@
+ 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;
+@@ -350,6 +393,7 @@
+ 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;
+@@ -367,13 +411,14 @@
+ 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);
+ }
+
+@@ -388,8 +433,9 @@
+ {
+ int error;
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+
+- error = __user_walk(filename,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd);
++ error = __user_walk_it(filename,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd, &it);
+ if (error)
+ goto out;
+
+@@ -400,6 +446,7 @@
+ set_fs_pwd(current->fs, nd.mnt, nd.dentry);
+
+ dput_and_out:
++ intent_release(&it);
+ path_release(&nd);
+ out:
+ return error;
+@@ -439,9 +486,10 @@
+ {
+ int error;
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_GETATTR };
+
+- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
+- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
++ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
+ if (error)
+ goto out;
+
+@@ -457,39 +505,56 @@
+ 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;
+@@ -498,30 +563,14 @@
+ 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;
+@@ -541,6 +590,20 @@
+ 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;
+@@ -645,6 +708,7 @@
+ {
+ int namei_flags, error;
+ struct nameidata nd;
++ struct lookup_intent it = { .it_op = IT_OPEN };
+
+ namei_flags = flags;
+ if ((namei_flags+1) & O_ACCMODE)
+@@ -652,14 +716,15 @@
+ 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;
+@@ -687,7 +752,9 @@
+ file_move(f, &inode->i_sb->s_files);
+
+ 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;
+ }
+@@ -699,6 +766,7 @@
+ !inode->i_mapping->a_ops->direct_IO))
+ goto cleanup_all;
+
++ intent_release(it);
+ return f;
+
+ cleanup_all:
+@@ -711,11 +779,17 @@
+ 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-2.4.21/fs/stat.c
+===================================================================
+--- linux-2.4.21.orig/fs/stat.c 2004-04-24 02:37:58.000000000 -0400
++++ linux-2.4.21/fs/stat.c 2004-04-26 19:06:32.000000000 -0400
+@@ -17,10 +17,12 @@
+ * 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->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 +143,15 @@
+ 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 +161,15 @@
+ 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 +184,15 @@
+ 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 +203,15 @@
+ 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 +232,7 @@
+ 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 +251,7 @@
+ 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 +273,7 @@
+
+ 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 +349,14 @@
+ {
+ 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 +366,14 @@
+ {
+ 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 +388,7 @@
+ 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-2.4.21/include/linux/dcache.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/dcache.h 2004-04-24 02:37:59.000000000 -0400
++++ linux-2.4.21/include/linux/dcache.h 2004-04-26 19:06:32.000000000 -0400
+@@ -7,6 +7,51 @@
+ #include <linux/gdb.h>
+ #include <linux/mount.h>
+ #include <linux/kernel.h>
++#include <linux/string.h>
++
++#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
+@@ -94,8 +139,22 @@
+ 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 && de->d_op && de->d_op->d_pin) \
++ de->d_op->d_pin(de, mnt, flag);
++#define UNPIN(de,mnt,flag) if (de && 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
+@@ -127,6 +186,7 @@
+ * 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-2.4.21/include/linux/fs.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/fs.h 2004-04-26 18:56:44.000000000 -0400
++++ linux-2.4.21/include/linux/fs.h 2004-04-26 19:06:32.000000000 -0400
+@@ -74,6 +74,7 @@
+
+ #define FMODE_READ 1
+ #define FMODE_WRITE 2
++#define FMODE_EXEC 4
+
+ #define READ 0
+ #define WRITE 1
+@@ -360,6 +361,9 @@
+ #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
+@@ -504,6 +508,7 @@
+ 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 */
+@@ -666,6 +671,7 @@
+
+ /* needed for tty driver, and maybe others */
+ void *private_data;
++ struct lookup_intent *f_it;
+ struct list_head f_ep_links;
+ spinlock_t f_ep_lock;
+ };
+@@ -795,6 +801,7 @@
+ struct qstr last;
+ unsigned int flags;
+ int last_type;
++ struct lookup_intent *intent;
+ };
+
+ /*
+@@ -916,7 +923,8 @@
+ 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
+@@ -991,21 +999,32 @@
+
+ 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);
+@@ -1204,10 +1223,14 @@
+
+ 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 *);
+
+@@ -1503,6 +1526,7 @@
+ 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(path_lookup(const char *, unsigned, struct nameidata *));
+@@ -1515,6 +1539,8 @@
+ 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 inode_init_once(struct inode *);
+ extern void _inode_init_once(struct inode *);
+@@ -1666,6 +1692,8 @@
+
+ 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-2.4.21/include/linux/fs_struct.h
+===================================================================
+--- linux-2.4.21.orig/include/linux/fs_struct.h 2001-07-13 18:10:44.000000000 -0400
++++ linux-2.4.21/include/linux/fs_struct.h 2004-04-26 19:06:32.000000000 -0400
+@@ -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-2.4.21/kernel/exit.c
+===================================================================
+--- linux-2.4.21.orig/kernel/exit.c 2004-04-24 02:39:01.000000000 -0400
++++ linux-2.4.21/kernel/exit.c 2004-04-26 19:06:32.000000000 -0400
+@@ -292,11 +292,14 @@
+ {
+ /* 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);
+ }
+Index: linux-2.4.21/kernel/fork.c
+===================================================================
+--- linux-2.4.21.orig/kernel/fork.c 2004-04-24 02:39:01.000000000 -0400
++++ linux-2.4.21/kernel/fork.c 2004-04-26 19:06:32.000000000 -0400
+@@ -466,10 +466,13 @@
+ 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-2.4.21/kernel/ksyms.c
+===================================================================
+--- linux-2.4.21.orig/kernel/ksyms.c 2004-04-26 18:56:44.000000000 -0400
++++ linux-2.4.21/kernel/ksyms.c 2004-04-26 19:06:32.000000000 -0400
+@@ -329,6 +329,7 @@
+ 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);
+ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) {
+ struct dentry *new;
+ err = permission(dentry->d_parent->d_inode,
-+ MAY_EXEC);
++ MAY_EXEC);
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
dput(dentry);
break;
}
-+ if (err)
-+ intent_release(it);
++ if (err)
++ intent_release(it);
path_release(nd);
return_err:
return err;
int retval = 0;
int mnt_flags = 0;
-@@ -725,10 +732,11 @@
+@@ -725,9 +732,11 @@
flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
/* ... and get the mountpoint */
+ if (retval) {
+ intent_release(&it);
return retval;
--
+ }
+
if (flags & MS_REMOUNT)
retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
@@ -739,6 +747,8 @@
else
retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
+ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
-+ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
if (error)
goto out;
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
+ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) {
+ struct dentry *new;
+ err = permission(dentry->d_parent->d_inode,
-+ MAY_EXEC);
++ MAY_EXEC);
+ if (err)
+ break;
+ new = real_lookup(dentry->d_parent,
-+ &dentry->d_name, 0, NULL);
++ &dentry->d_name, 0, it);
+ d_invalidate(dentry);
+ dput(dentry);
+ if (IS_ERR(new)) {
if (IS_ERR(dentry))
goto fail;
if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode)
-@@ -1278,7 +1416,16 @@ asmlinkage long sys_mknod(const char * f
+@@ -1278,7 +1416,20 @@ asmlinkage long sys_mknod(const char * f
error = path_lookup(tmp, LOOKUP_PARENT, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
+
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
path_release(&nd);
out:
putname(tmp);
-@@ -1346,7 +1494,14 @@ asmlinkage long sys_mkdir(const char * p
+@@ -1346,7 +1494,20 @@ asmlinkage long sys_mkdir(const char * p
error = path_lookup(tmp, LOOKUP_PARENT, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 1);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
error = PTR_ERR(dentry);
if (!IS_ERR(dentry)) {
/* Why not before? Because we want correct error value */
-@@ -1581,15 +1752,23 @@ asmlinkage long sys_symlink(const char *
+@@ -1581,15 +1752,27 @@ asmlinkage long sys_symlink(const char *
error = path_lookup(to, LOOKUP_PARENT, &nd);
if (error)
goto out;
- dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
putname(to);
}
putname(from);
-@@ -1665,7 +1844,14 @@ asmlinkage long sys_link(const char * ol
+@@ -1665,7 +1844,18 @@ asmlinkage long sys_link(const char * ol
error = -EXDEV;
if (old_nd.mnt != nd.mnt)
goto out_release;
- new_dentry = lookup_create(&nd, 0);
++ if (nd.last_type != LAST_NORM) {
++ error = -EEXIST;
++ goto out2;
++ }
+ 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);
int retval = 0;
int mnt_flags = 0;
-@@ -725,10 +732,11 @@ long do_mount(char * dev_name, char * di
+@@ -725,9 +732,11 @@ long do_mount(char * dev_name, char * di
flags &= ~(MS_NOSUID|MS_NOEXEC|MS_NODEV);
/* ... and get the mountpoint */
+ if (retval) {
+ intent_release(&it);
return retval;
--
+ }
+
if (flags & MS_REMOUNT)
retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags,
- data_page);
@@ -739,6 +747,8 @@ long do_mount(char * dev_name, char * di
else
retval = do_add_mount(&nd, type_page, flags, mnt_flags,
- error = __user_walk(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
- LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd);
+ error = __user_walk_it(filename, LOOKUP_POSITIVE | LOOKUP_FOLLOW |
-+ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
++ LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd, &it);
if (error)
goto out;
*/
--- linux-2.4.22-ac1/fs/stat.c~vfs_intent-2.4.22-rh 2003-09-25 14:16:27.000000000 +0400
+++ linux-2.4.22-ac1-alexey/fs/stat.c 2003-09-25 14:42:46.000000000 +0400
-@@ -17,10 +17,14 @@
+@@ -17,10 +17,12 @@
* Revalidate the inode. This is required for proper NFS attribute caching.
*/
static __inline__ int
{
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)
#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
++#define ATTR_CTIME_SET 0x2000
/*
* This is the Inode Attributes structure, used for notify_change(). It
name = getname(pathname);
if(IS_ERR(name))
-@@ -1716,6 +1737,16 @@
+@@ -1716,6 +1737,14 @@
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);
error = PTR_ERR(dentry);
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext2_xattr_cache_insert(new_bh);
++ (void)ext2_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT2_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext2_xattr_cache_insert(new_bh);
+
+ ext2_xattr_update_super_block(sb);
+ }
+ } else if (old_bh && header == HDR(old_bh)) {
+ /* Keep this block. */
+ new_bh = old_bh;
-+ ext3_xattr_cache_insert(new_bh);
++ (void)ext3_xattr_cache_insert(new_bh);
+ } else {
+ /* We need to allocate a new block */
+ int force = EXT3_I(inode)->i_file_acl != 0;
+ 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);
++ (void)ext3_xattr_cache_insert(new_bh);
+
+ ext3_xattr_update_super_block(handle, sb);
+ }
ext3-truncate-buffer-head.patch
inode-max-readahead-2.4.24.patch
dcache_refcount_debug.patch
+mkdep-revert-rh-2.4.patch
nfs_export_kernel-2.4.18.patch
ext3-raw-lookup-pdirops.patch
ext3-truncate-buffer-head.patch
+mkdep-revert-rh-2.4.patch
procfs-ndynamic-2.4.patch
ext3-truncate-buffer-head.patch
inode-max-readahead-2.4.24.patch
+mkdep-revert-rh-2.4.patch
-dev_read_only_hp_2.4.20.patch
+dev_read_only-suse-2.4.19.patch
exports_2.4.19-suse.patch
lustre_version.patch
vfs_intent-2.4.19-suse.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-1-suse-2.4.19.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
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
+ext3-trusted_ea-suse-2.4.19.patch
ext3-truncate-buffer-head.patch
--- /dev/null
+configurable-x86-stack-2.4.21-sles8sp3.patch
+dev_read_only_2.4.20-rh.patch
+exports_2.4.20-rh-hp.patch
+lustre_version.patch
+vfs_intent-2.4.21-sles8sp3.patch
+invalidate_show-2.4.21-sles8sp3.patch
+iod-stock-24-exports_hp.patch
+ext3-htree-2.4.21-chaos.patch
+linux-2.4.21-xattr-0.8.54-suse2.patch
+ext3-orphan_lock-2.4.22-rh.patch
+ext3-noread-2.4.21-suse2.patch
+ext3-delete_thread-2.4.21-chaos.patch
+extN-wantedi-2.4.21-suse2.patch
+ext3-san-2.4.20.patch
+ext3-map_inode_page-2.4.21-suse2.patch
+ext3-error-export.patch
+iopen-2.4.21-sles8sp3.patch
+tcp-zero-copy-2.4.21-sles8sp3.patch
+jbd-dont-account-blocks-twice.patch
+jbd-commit-tricks.patch
+ext3-no-write-super-chaos.patch
+add_page_private.patch
+nfs_export_kernel-2.4.21-suse2.patch
+ext3-raw-lookup.patch
+ext3-ea-in-inode-2.4.21-suse2.patch
+listman-2.4.20.patch
+gfp_memalloc-2.4.24.patch
+ext3-xattr-ptr-arith-fix.patch
+kernel_text_address-2.4.21-sles8sp3.patch
+ext3-truncate-buffer-head.patch
+export-truncate.patch
chaos-2.4.20 linux-chaos-2.4.20 same as rh-2.4.20-8 i386
kgdb-2.5.73 linux-2.5.73 vanilla 2.5.73 with kgdb i386
bproc-2.4.20-hp-pnnl linux-2.4.20-hp4_pnnl9 hp-pnnl + bproc i386
+suse-2.4.19 SUSE ES 8
struct ldlm_resource *ldlm_resource_getref(struct ldlm_resource *res)
{
LASSERT(res != NULL);
- LASSERT(res != (void *)0x5a5a5a5a);
+ LASSERT(res != LP_POISON);
atomic_inc(&res->lr_refcount);
CDEBUG(D_INFO, "getref res: %p count: %d\n", res,
atomic_read(&res->lr_refcount));
CDEBUG(D_INFO, "putref res: %p count: %d\n", res,
atomic_read(&res->lr_refcount) - 1);
LASSERT(atomic_read(&res->lr_refcount) > 0);
- LASSERT(atomic_read(&res->lr_refcount) < 0x5a5a5a5a);
+ LASSERT(atomic_read(&res->lr_refcount) < LI_POISON);
if (atomic_dec_and_test(&res->lr_refcount)) {
struct ldlm_namespace *ns = res->lr_namespace;
# create shared lib lustre
rm -f $CWD/liblustre.so
$LD -shared -o $CWD/liblustre.so -init __liblustre_setup_ -fini __liblustre_cleanup_ \
- $ALL_OBJS -lcap -lpthread
+ $ALL_OBJS -lpthread
#rm -rf $sysio_tmp
#include <signal.h>
#include <sys/types.h>
#include <sys/queue.h>
-#include <sys/capability.h>
#include <netinet/in.h>
#include <sys/socket.h>
return str;
}
-int in_group_p(gid_t gid)
-{
- int i;
-
- if (gid == current->fsgid)
- return 1;
-
- for (i = 0; i < current->ngroups; i++) {
- if (gid == current->groups[i])
- return 1;
- }
-
- return 0;
-}
-
-static void init_capability(int *res)
-{
- cap_value_t cap_types[] = {
- CAP_CHOWN,
- CAP_DAC_OVERRIDE,
- CAP_DAC_READ_SEARCH,
- CAP_FOWNER,
- CAP_FSETID,
- CAP_KILL,
- CAP_SETGID,
- CAP_SETUID,
- /* following are linux specific, we could simply
- * remove them I think */
- CAP_SETPCAP,
- CAP_LINUX_IMMUTABLE,
- CAP_NET_BIND_SERVICE,
- CAP_NET_BROADCAST,
- CAP_NET_ADMIN,
- CAP_NET_RAW,
- CAP_IPC_LOCK,
- CAP_IPC_OWNER,
- CAP_SYS_MODULE,
- CAP_SYS_RAWIO,
- CAP_SYS_CHROOT,
- CAP_SYS_PTRACE,
- CAP_SYS_PACCT,
- CAP_SYS_ADMIN,
- CAP_SYS_BOOT,
- CAP_SYS_NICE,
- CAP_SYS_RESOURCE,
- CAP_SYS_TIME,
- CAP_SYS_TTY_CONFIG,
- CAP_MKNOD,
- CAP_LEASE,
- };
- cap_t syscap;
- cap_flag_value_t capval;
- int i;
-
- *res = 0;
-
- syscap = cap_get_proc();
- if (!syscap) {
- printf("Liblustre: Warning: failed to get system capability, "
- "set to minimal\n");
- return;
- }
-
- for (i = 0; i < sizeof(cap_types)/sizeof(cap_t); i++) {
- LASSERT(cap_types[i] < 32);
- if (!cap_get_flag(syscap, cap_types[i],
- CAP_EFFECTIVE, &capval)) {
- if (capval == CAP_SET) {
- *res |= 1 << cap_types[i];
- }
- }
- }
-}
-
-static int init_current(char *comm)
+void init_current(char *comm)
{
current = malloc(sizeof(*current));
- if (!current) {
- CERROR("Not enough memory\n");
- return -ENOMEM;
- }
- current->fs = ¤t->__fs;
+ current->fs = malloc(sizeof(*current->fs));
current->fs->umask = umask(0777);
umask(current->fs->umask);
-
strncpy(current->comm, comm, sizeof(current->comm));
current->pid = getpid();
- current->fsuid = geteuid();
- current->fsgid = getegid();
+ current->fsuid = 0;
+ current->fsgid = 0;
+ current->cap_effective = -1;
memset(¤t->pending, 0, sizeof(current->pending));
-
- current->max_groups = sysconf(_SC_NGROUPS_MAX);
- current->groups = malloc(sizeof(gid_t) * current->max_groups);
- if (!current->groups) {
- CERROR("Not enough memory\n");
- return -ENOMEM;
- }
- current->ngroups = getgroups(current->max_groups, current->groqps);
- if (current->ngroups < 0) {
- perror("Error getgroups");
- return -EINVAL;
- }
-
- init_capability(¤t->cap_effective);
-
- return 0;
}
/* FIXME */
printf("LibLustre: TCPNAL NID: %016llx\n", tcpnal_mynid);
}
- if (init_current("dummy") ||
- init_obdclass() ||
+ init_current("dummy");
+ if (init_obdclass() ||
init_lib_portals() ||
ptlrpc_init() ||
mdc_init() ||
RETURN(rc);
}
+static void sighandler_USR1(int signum)
+{
+ /* do nothing */
+}
+
/* parse host:/mdsname/profile string */
int ll_parse_mount_target(const char *target, char **mdsnid,
char **mdsname, char **profile)
*/
srand(time(NULL) + getpid());
+ signal(SIGUSR1, sighandler_USR1);
+
lustre_path = getenv(ENV_LUSTRE_MNTPNT);
if (!lustre_path) {
lustre_path = "/mnt/lustre";
/* NB 1 request reference will be taken away by ll_intent_lock()
* when I return
- */
- /* FIXME: for CREAT, libsysio require the inode must be generated here
- * currently here we don't know the whether the create is successful
- * or failed on mds. thus blinded return -EPERM in llu_iget(). need
- * a fix later.
+ * Note: libsysio require the inode must be generated here
*/
if ((it->it_op & IT_CREAT) || !it_disposition(it, DISP_LOOKUP_NEG)) {
struct lustre_md md;
RETURN(rc);
inode = llu_iget(parent->i_fs, &md);
- if (!inode || IS_ERR(inode)) {
+ if (!inode) {
/* free the lsm if we allocated one above */
if (md.lsm != NULL)
obd_free_memmd(sbi->ll_osc_exp, &md.lsm);
- RETURN(inode ? PTR_ERR(inode) : -ENOMEM);
+ RETURN(-ENOMEM);
} else if (md.lsm != NULL &&
llu_i2info(inode)->lli_smd != md.lsm) {
obd_free_memmd(sbi->ll_osc_exp, &md.lsm);
#include <sys/stat.h>
#include <sys/fcntl.h>
#include <sys/queue.h>
-#include <sys/capability.h>
#ifndef __CYGWIN__
# include <sys/statvfs.h>
#else
if ((md->body->valid &
(OBD_MD_FLGENER | OBD_MD_FLID | OBD_MD_FLTYPE)) !=
- (OBD_MD_FLGENER | OBD_MD_FLID | OBD_MD_FLTYPE)) {
- /* FIXME this is workaround for for open(O_CREAT),
- * see lookup_it_finish(). */
- return ERR_PTR(-EPERM);
- }
+ (OBD_MD_FLGENER | OBD_MD_FLID | OBD_MD_FLTYPE))
+ CERROR("invalide fields!\n");
/* try to find existing inode */
fid.id = md->body->ino;
LASSERT(sbi->ll_rootino != 0);
root = llu_iget(fs, &md);
- if (!root || IS_ERR(root)) {
+ if (root == NULL) {
CERROR("fail to generate root inode\n");
GOTO(out_request, err = -EBADF);
}
AM_CFLAGS = $(LLCFLAGS)
LIBS = $(LIBEFENCE) $(LIBREADLINE)
-LLIB_EXEC= ../liblustre.a -lcap -lpthread
+LLIB_EXEC= ../liblustre.a -lpthread
if LIBLUSTRE
noinst_LIBRARIES = libtestcommon.a
echo_test_SOURCES = echo_test.c ../../utils/parser.c ../../utils/obd.c ../../utils/lustre_cfg.c
echo_test_CFLAGS = $(LL_CFLAGS)
-echo_test_LDADD = ../liblsupport.a $(LIBREADLINE) -lcap -lpthread
+echo_test_LDADD = ../liblsupport.a $(LIBREADLINE) -lpthread
echo_test_DEPENDENCIES=$(top_builddir)/liblustre/liblsupport.a
sanity_SOURCES = sanity.c
return 0;
}
-int in_group_p(gid_t gid)
-{
- return 0;
-}
-
int init_current(int argc, char **argv)
{
current = malloc(sizeof(*current));
fail:
SetPageChecked(page);
SetPageError(page);
- LBUG();
}
static struct page *ll_get_dir_page(struct inode *dir, unsigned long n)
struct lov_stripe_md *lsm = lli->lli_smd;
struct lustre_handle lockh = { 0 };
ldlm_policy_data_t policy;
- ldlm_error_t err;
+ int rc;
ssize_t retval;
__u64 kms;
ENTRY;
policy.l_extent.start = *ppos;
policy.l_extent.end = *ppos + count - 1;
- err = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
- (filp->f_flags & O_NONBLOCK)?LDLM_FL_BLOCK_NOWAIT:
- 0);
- if (err != ELDLM_OK)
- RETURN(err);
+ rc = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
+ (filp->f_flags & O_NONBLOCK) ?
+ LDLM_FL_BLOCK_NOWAIT: 0);
+ if (rc != 0)
+ RETURN(rc);
kms = lov_merge_size(lsm, 1);
if (*ppos + count - 1 > kms) {
struct lustre_handle lockh = { 0 };
ldlm_policy_data_t policy;
loff_t maxbytes = ll_file_maxbytes(inode);
- ldlm_error_t err;
ssize_t retval;
- int nonblock = 0;
+ int nonblock = 0, rc;
ENTRY;
if (file->f_flags & O_NONBLOCK)
nonblock = LDLM_FL_BLOCK_NOWAIT;
policy.l_extent.end = *ppos + count - 1;
}
- err = ll_extent_lock(fd, inode, lsm, LCK_PW, &policy, &lockh, nonblock);
- if (err != ELDLM_OK)
- RETURN(err);
+ rc = ll_extent_lock(fd, inode, lsm, LCK_PW, &policy, &lockh, nonblock);
+ if (rc != 0)
+ RETURN(rc);
/* this is ok, g_f_w will overwrite this under i_sem if it races
* with a local truncate, it just makes our maxbyte checking easier */
struct lustre_handle lockh = { 0 };
struct ll_inode_info *lli = ll_i2info(inode);
struct lov_stripe_md *lsm = lli->lli_smd;
- ldlm_error_t err;
- int flags = 0;
+ int flags = 0, rc;
ENTRY;
if (fd->fd_flags & LL_FILE_GROUP_LOCKED) {
if (file->f_flags & O_NONBLOCK)
flags = LDLM_FL_BLOCK_NOWAIT;
- err = ll_extent_lock(fd, inode, lsm, LCK_GROUP, &policy, &lockh, flags);
- if (err)
- RETURN(err);
+ rc = ll_extent_lock(fd, inode, lsm, LCK_GROUP, &policy, &lockh, flags);
+ if (rc != 0)
+ RETURN(rc);
fd->fd_flags |= LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK;
fd->fd_gid = arg;
struct ll_file_data *fd = file->private_data;
struct ll_inode_info *lli = ll_i2info(inode);
struct lov_stripe_md *lsm = lli->lli_smd;
- ldlm_error_t err;
+ int rc;
ENTRY;
if (!(fd->fd_flags & LL_FILE_GROUP_LOCKED)) {
if (fd->fd_gid != arg) /* Ugh? Unlocking with different gid? */
RETURN(-EINVAL);
-
+
fd->fd_flags &= ~(LL_FILE_GROUP_LOCKED|LL_FILE_IGNORE_LOCK);
- err = ll_extent_unlock(fd, inode, lsm, LCK_GROUP, &fd->fd_cwlockh);
- if (err)
- RETURN(err);
+ rc = ll_extent_unlock(fd, inode, lsm, LCK_GROUP, &fd->fd_cwlockh);
+ if (rc)
+ RETURN(rc);
fd->fd_gid = 0;
memset(&fd->fd_cwlockh, 0, sizeof(fd->fd_cwlockh));
RETURN(0);
-}
+}
int ll_file_ioctl(struct inode *inode, struct file *file, unsigned int cmd,
unsigned long arg)
lprocfs_counter_incr(ll_i2sbi(inode)->ll_stats, LPROC_LL_LLSEEK);
if (origin == 2) { /* SEEK_END */
- ldlm_error_t err;
- int nonblock = 0;
+ int nonblock = 0, rc;
ldlm_policy_data_t policy = { .l_extent = {0, OBD_OBJECT_EOF }};
if (file->f_flags & O_NONBLOCK)
nonblock = LDLM_FL_BLOCK_NOWAIT;
- err = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
+ rc = ll_extent_lock(fd, inode, lsm, LCK_PR, &policy, &lockh,
nonblock);
- if (err != ELDLM_OK)
- RETURN(err);
+ if (rc != 0)
+ RETURN(rc);
offset += inode->i_size;
} else if (origin == 1) { /* SEEK_CUR */
#ifndef LLITE_INTERNAL_H
#define LLITE_INTERNAL_H
+/* default to about 40meg of readahead on a given system. That much tied
+ * up in 512k readahead requests serviced at 40ms each is about 1GB/s. */
+#define SBI_DEFAULT_RA_MAX ((40 << 20) >> PAGE_CACHE_SHIFT)
+
struct ll_sb_info {
+ /* this protects pglist and max_r_a_pages. It isn't safe to
+ * grab from interrupt contexts */
+ spinlock_t ll_lock;
struct obd_uuid ll_sb_uuid;
-// struct lustre_handle ll_mdc_conn;
struct obd_export *ll_mdc_exp;
struct obd_export *ll_osc_exp;
struct proc_dir_entry* ll_proc_root;
obd_id ll_rootino; /* number of root inode */
- struct obd_uuid ll_mds_uuid;
- struct obd_uuid ll_mds_peer_uuid;
struct lustre_mount_data *ll_lmd;
char *ll_instance;
int ll_flags;
- wait_queue_head_t ll_commitcbd_waitq;
- wait_queue_head_t ll_commitcbd_ctl_waitq;
- int ll_commitcbd_flags;
- struct task_struct *ll_commitcbd_thread;
- time_t ll_commitcbd_waketime;
- time_t ll_commitcbd_timeout;
- spinlock_t ll_commitcbd_lock;
struct list_head ll_conn_chain; /* per-conn chain of SBs */
struct hlist_head ll_orphan_dentry_list; /*please don't ask -p*/
struct lprocfs_stats *ll_stats; /* lprocfs stats counter */
- spinlock_t ll_pglist_lock;
unsigned long ll_pglist_gen;
struct list_head ll_pglist;
+
+ unsigned long ll_read_ahead_pages;
+ unsigned long ll_max_read_ahead_pages;
+
};
struct ll_readahead_state {
spinlock_t ras_lock;
- unsigned long ras_last, ras_window, ras_next_index;
+ unsigned long ras_last_readpage, ras_consecutive;
+ unsigned long ras_window_start, ras_window_len;
+ unsigned long ras_next_readahead;
+
};
extern kmem_cache_t *ll_file_data_slab;
void ll_inode_fill_obdo(struct inode *inode, int cmd, struct obdo *oa);
void ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc);
void ll_removepage(struct page *page);
-int ll_sync_page(struct page *page);
int ll_readpage(struct file *file, struct page *page);
struct ll_async_page *llap_from_cookie(void *cookie);
struct ll_async_page *llap_from_page(struct page *page);
if (!sbi)
RETURN(NULL);
- spin_lock_init(&sbi->ll_pglist_lock);
+ spin_lock_init(&sbi->ll_lock);
INIT_LIST_HEAD(&sbi->ll_pglist);
sbi->ll_pglist_gen = 0;
+ sbi->ll_max_read_ahead_pages = SBI_DEFAULT_RA_MAX;
INIT_LIST_HEAD(&sbi->ll_conn_chain);
INIT_HLIST_HEAD(&sbi->ll_orphan_dentry_list);
ll_s2sbi(sb) = sbi;
rc = ll_extent_lock(NULL, inode, lsm, LCK_PW, &policy, &lockh,
ast_flags);
down(&inode->i_sem);
- if (rc != ELDLM_OK)
+ if (rc != 0)
RETURN(rc);
rc = vmtruncate(inode, attr->ia_size);
}
-#if 0
-static int ll_rd_path(char *page, char **start, off_t off, int count, int *eof,
- void *data)
-{
- return 0;
-}
-#endif
-
static int ll_rd_fstype(char *page, char **start, off_t off, int count,
int *eof, void *data)
{
RETURN(count);
}
+static int ll_rd_max_read_ahead_mb(char *page, char **start, off_t off,
+ int count, int *eof, void *data)
+{
+ struct super_block *sb = data;
+ struct ll_sb_info *sbi = ll_s2sbi(sb);
+ unsigned val;
+
+ spin_lock(&sbi->ll_lock);
+ val = (sbi->ll_max_read_ahead_pages << PAGE_CACHE_SHIFT) >> 20;
+ spin_unlock(&sbi->ll_lock);
+
+ return snprintf(page, count, "%u\n", val);
+}
+
+static int ll_wr_max_read_ahead_mb(struct file *file, const char *buffer,
+ unsigned long count, void *data)
+{
+ struct super_block *sb = data;
+ struct ll_sb_info *sbi = ll_s2sbi(sb);
+ int val, rc;
+
+ rc = lprocfs_write_helper(buffer, count, &val);
+ if (rc)
+ return rc;
+
+ if (val < 0 || val > (num_physpages << PAGE_SHIFT) >> 20)
+ return -ERANGE;
+
+ spin_lock(&sbi->ll_lock);
+ sbi->ll_max_read_ahead_pages = (val << 20) >> PAGE_CACHE_SHIFT;
+ spin_unlock(&sbi->ll_lock);
+
+ return count;
+}
+
static struct lprocfs_vars lprocfs_obd_vars[] = {
{ "uuid", ll_rd_sb_uuid, 0, 0 },
//{ "mntpt_path", ll_rd_path, 0, 0 },
{ "filesfree", ll_rd_filesfree, 0, 0 },
//{ "filegroups", lprocfs_rd_filegroups, 0, 0 },
{ "read_ahead", ll_rd_read_ahead, ll_wr_read_ahead, 0 },
+ { "max_read_ahead_mb", ll_rd_max_read_ahead_mb,
+ ll_wr_max_read_ahead_mb, 0 },
{ 0 }
};
return 0;
}
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
if (llap != NULL) {
seq_puts(seq, "]\n");
}
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
return 0;
}
/* we've just displayed the llap that is after us in the list.
* we advance to a position beyond it, returning null if there
* isn't another llap in the list beyond that new position. */
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
list_del_init(&dummy_llap->llap_proc_item);
if (llap) {
list_add(&dummy_llap->llap_proc_item, &llap->llap_proc_item);
llap = llite_pglist_next_llap(sbi, &dummy_llap->llap_proc_item);
}
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
++*pos;
if (llap == NULL) {
seq = file->private_data;
seq->private = llap;
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
list_add(&llap->llap_proc_item, &sbi->ll_pglist);
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
return 0;
}
struct ll_async_page *llap = seq->private;
struct ll_sb_info *sbi = llap->llap_cookie;
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
if (!list_empty(&llap->llap_proc_item))
list_del_init(&llap->llap_proc_item);
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
OBD_FREE(llap, sizeof(*llap));
return seq_release(inode, file);
page = llap->llap_page;
- if (cmd == OBD_BRW_READ)
- RETURN(0);
+ LASSERT(cmd != OBD_BRW_READ);
/* we're trying to write, but the page is locked.. come back later */
if (TryLockPage(page))
page->private = (unsigned long)llap;
llap->llap_page = page;
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
sbi->ll_pglist_gen++;
list_add_tail(&llap->llap_proc_item, &sbi->ll_pglist);
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
RETURN(llap);
}
RETURN(rc);
}
+static unsigned long ll_ra_count_get(struct ll_sb_info *sbi, unsigned long len)
+{
+ unsigned long ret;
+ ENTRY;
+
+ spin_lock(&sbi->ll_lock);
+ ret = min(sbi->ll_max_read_ahead_pages - sbi->ll_read_ahead_pages,
+ len);
+ sbi->ll_read_ahead_pages += ret;
+ spin_unlock(&sbi->ll_lock);
+
+ RETURN(ret);
+}
+
+static void ll_ra_count_put(struct ll_sb_info *sbi, unsigned long len)
+{
+ spin_lock(&sbi->ll_lock);
+ LASSERTF(sbi->ll_read_ahead_pages >= len, "r_a_p %lu len %lu\n",
+ sbi->ll_read_ahead_pages, len);
+ sbi->ll_read_ahead_pages -= len;
+ spin_unlock(&sbi->ll_lock);
+}
+
/* called for each page in a completed rpc.*/
void ll_ap_completion(void *data, int cmd, struct obdo *oa, int rc)
{
LL_CDEBUG_PAGE(D_PAGE, page, "completing cmd %d with %d\n", cmd, rc);
+ if (cmd == OBD_BRW_READ && llap->llap_defer_uptodate)
+ ll_ra_count_put(ll_i2sbi(page->mapping->host), 1);
+
if (rc == 0) {
if (cmd == OBD_BRW_READ) {
if (!llap->llap_defer_uptodate)
SetPageError(page);
}
-
unlock_page(page);
if (0 && cmd == OBD_BRW_WRITE) {
* is providing exclusivity to memory pressure/truncate/writeback..*/
page->private = 0;
- spin_lock(&sbi->ll_pglist_lock);
+ spin_lock(&sbi->ll_lock);
if (!list_empty(&llap->llap_proc_item))
list_del_init(&llap->llap_proc_item);
sbi->ll_pglist_gen++;
- spin_unlock(&sbi->ll_pglist_lock);
+ spin_unlock(&sbi->ll_lock);
OBD_FREE(llap, sizeof(*llap));
EXIT;
}
llap->llap_defer_uptodate = defer;
rc = obd_queue_group_io(exp, ll_i2info(page->mapping->host)->lli_smd,
NULL, oig, llap->llap_cookie, OBD_BRW_READ, 0,
- PAGE_SIZE, 0, ASYNC_COUNT_STABLE);
+ PAGE_SIZE, 0, ASYNC_COUNT_STABLE | ASYNC_READY
+ | ASYNC_URGENT);
if (rc) {
LL_CDEBUG_PAGE(D_ERROR, page, "read queue failed: rc %d\n", rc);
page_cache_release(page);
RETURN(rc);
}
-#define LL_RA_MIN(inode) ((unsigned long)PTLRPC_MAX_BRW_PAGES / 2)
-#define LL_RA_MAX(inode) ((ll_i2info(inode)->lli_smd->lsm_xfersize * 3) >> \
- PAGE_CACHE_SHIFT)
+#define RAS_CDEBUG(ras) \
+ CDEBUG(D_READA, "lrp %lu c %lu ws %lu wl %lu nra %lu\n", \
+ ras->ras_last_readpage, ras->ras_consecutive, \
+ ras->ras_window_start, ras->ras_window_len, \
+ ras->ras_next_readahead);
-static void ll_readahead(struct ll_readahead_state *ras,
+static int ll_readahead(struct ll_readahead_state *ras,
struct obd_export *exp, struct address_space *mapping,
struct obd_io_group *oig, int flags)
{
- unsigned long i, start, end;
+ unsigned long i, start = 0, end = 0, reserved;
struct ll_async_page *llap;
struct page *page;
- int rc;
+ int rc, ret = 0;
+ __u64 kms;
+ ENTRY;
- if (mapping->host->i_size == 0)
- return;
+ kms = lov_merge_size(ll_i2info(mapping->host)->lli_smd, 1);
+ if (kms == 0)
+ RETURN(0);
spin_lock(&ras->ras_lock);
- /* make sure to issue a window's worth of read-ahead pages */
- end = ras->ras_last;
- start = end - ras->ras_window;
- if (start > end)
- start = 0;
-
- /* but don't iterate over pages that we've already issued. this
- * will set start to end + 1 if we've already read-ahead up to
- * ras_last sothe for() won't be entered */
- if (ras->ras_next_index > start)
- start = ras->ras_next_index;
- if (end != ~0UL)
- ras->ras_next_index = end + 1;
+ if (ras->ras_window_len) {
+ start = ras->ras_next_readahead;
+ end = ras->ras_window_start + ras->ras_window_len - 1;
+ end = min(end, (unsigned long)(kms >> PAGE_CACHE_SHIFT));
+ ras->ras_next_readahead = max(end, end + 1);
- CDEBUG(D_READA, "ni %lu last %lu win %lu: reading from %lu to %lu\n",
- ras->ras_next_index, ras->ras_last, ras->ras_window,
- start, end);
+ RAS_CDEBUG(ras);
+ }
spin_unlock(&ras->ras_lock);
- /* clamp to filesize */
- i = (mapping->host->i_size - 1) >> PAGE_CACHE_SHIFT;
- end = min(end, i);
+ if (end == 0)
+ RETURN(0);
+
+ reserved = ll_ra_count_get(ll_i2sbi(mapping->host), end - start + 1);
- for (i = start; i <= end; i++) {
- /* grab_cache_page_nowait returns null if this races with
- * truncating the page (page->mapping == NULL) */
+ for (i = start; reserved > 0 && i <= end; i++) {
+ /* skip locked pages from previous readpage calls */
page = grab_cache_page_nowait(mapping, i);
- if (page == NULL)
- break;
+ if (page == NULL) {
+ CDEBUG(D_READA, "g_c_p_n failed\n");
+ continue;
+ }
+
+ /* we do this first so that we can see the page in the /proc
+ * accounting */
+ llap = llap_from_page(page);
+ if (IS_ERR(llap) || llap->llap_defer_uptodate)
+ goto next_page;
- /* the book-keeping above promises that we've tried
- * all the indices from start to end, so we don't
- * stop if anyone returns an error. This may not be good. */
+ /* skip completed pages */
if (Page_Uptodate(page))
goto next_page;
+ /* bail when we hit the end of the lock. */
if ((rc = ll_page_matches(page, flags)) <= 0) {
LL_CDEBUG_PAGE(D_READA | D_PAGE, page,
"lock match failed: rc %d\n", rc);
+ i = end;
goto next_page;
}
- llap = llap_from_page(page);
- if (IS_ERR(llap) || llap->llap_defer_uptodate)
- goto next_page;
-
rc = ll_issue_page_read(exp, llap, oig, 1);
- if (rc == 0)
- LL_CDEBUG_PAGE(D_PAGE, page, "started read-ahead\n");
+ if (rc == 0) {
+ reserved--;
+ ret++;
+ LL_CDEBUG_PAGE(D_READA| D_PAGE, page,
+ "started read-ahead\n");
+ }
if (rc) {
next_page:
- LL_CDEBUG_PAGE(D_PAGE, page, "skipping read-ahead\n");
+ LL_CDEBUG_PAGE(D_READA | D_PAGE, page,
+ "skipping read-ahead\n");
unlock_page(page);
}
page_cache_release(page);
}
+
+ LASSERTF(reserved >= 0, "reserved %lu\n", reserved);
+ if (reserved != 0)
+ ll_ra_count_put(ll_i2sbi(mapping->host), reserved);
+ RETURN(ret);
+}
+
+static void ras_set_start(struct ll_readahead_state *ras,
+ unsigned long index)
+{
+ ras->ras_window_start = index & (~(PTLRPC_MAX_BRW_PAGES - 1));
+ ras->ras_next_readahead = max(ras->ras_window_start,
+ ras->ras_next_readahead);
}
/* called with the ras_lock held or from places where it doesn't matter */
-static void ll_readahead_set(struct inode *inode,
- struct ll_readahead_state *ras,
- unsigned long index)
+static void ras_reset(struct ll_readahead_state *ras,
+ unsigned long index)
{
- ras->ras_next_index = index;
- if (ras->ras_next_index != ~0UL)
- ras->ras_next_index++;
- ras->ras_window = LL_RA_MIN(inode);
- ras->ras_last = ras->ras_next_index + ras->ras_window;
- if (ras->ras_last < ras->ras_next_index)
- ras->ras_last = ~0UL;
- CDEBUG(D_READA, "ni %lu last %lu win %lu: set %lu\n",
- ras->ras_next_index, ras->ras_last, ras->ras_window,
- index);
+ ras->ras_last_readpage = index;
+ ras->ras_consecutive = 1;
+ ras->ras_window_len = 0;
+ ras_set_start(ras, index);
+ ras->ras_next_readahead = ras->ras_window_start;
+
+ RAS_CDEBUG(ras);
}
void ll_readahead_init(struct inode *inode, struct ll_readahead_state *ras)
{
spin_lock_init(&ras->ras_lock);
- ll_readahead_set(inode, ras, 0);
+ ras_reset(ras, 0);
}
-static void ll_readahead_update(struct inode *inode,
- struct ll_readahead_state *ras,
- unsigned long index, int hit)
+static void ras_update(struct ll_readahead_state *ras,
+ unsigned long index, unsigned long max)
{
- unsigned long issued_start, new_last;
+ ENTRY;
spin_lock(&ras->ras_lock);
- /* we're interested in noticing the index's relation to the
- * previously issued read-ahead pages */
- issued_start = ras->ras_next_index - ras->ras_window - 1;
- if (issued_start > ras->ras_next_index)
- issued_start = 0;
-
- CDEBUG(D_READA, "ni %lu last %lu win %lu: %s ind %lu start %lu\n",
- ras->ras_next_index, ras->ras_last, ras->ras_window,
- hit ? "hit" : "miss", index, issued_start);
- if (!hit &&
- index == ras->ras_next_index && index == ras->ras_last + 1) {
- /* special case the kernel's read-ahead running into the
- * page just beyond our read-ahead window as an extension
- * of our read-ahead. sigh. wishing it was easier to
- * turn off 2.4's read-ahead. */
- ras->ras_window = min(LL_RA_MAX(inode), ras->ras_window + 1);
- if (index != ~0UL)
- ras->ras_next_index = index + 1;
- ras->ras_last = index;
- } else if (!hit &&
- (index > issued_start || ras->ras_next_index >= index)) {
- /* deal with a miss way out of the window. we interpret
- * this as a seek and restart the window */
- ll_readahead_set(inode, ras, index);
-
- } else if (!hit &&
- issued_start <= index && index < ras->ras_next_index) {
- /* a miss inside the window? surely its memory pressure
- * evicting our read pages before the app can see them.
- * we shrink the window aggressively */
- unsigned long old_window = ras->ras_window;
-
- ras->ras_window = max(ras->ras_window / 2, LL_RA_MIN(inode));
- ras->ras_last -= old_window - ras->ras_window;
- if (ras->ras_next_index > ras->ras_last)
- ras->ras_next_index = ras->ras_last + 1;
- CDEBUG(D_READA, "ni %lu last %lu win %lu: miss inside\n",
- ras->ras_next_index, ras->ras_last, ras->ras_window);
-
- } else if (hit &&
- issued_start <= index && index < ras->ras_next_index) {
- /* a hit inside the window. grow the window by twice the
- * number of pages that are satisified within the window. */
- ras->ras_window = min(LL_RA_MAX(inode), ras->ras_window + 2);
-
- /* we want the next readahead pass to issue a windows worth
- * beyond where the app currently is */
- new_last = index + ras->ras_window;
- if (new_last > ras->ras_last)
- ras->ras_last = new_last;
-
- CDEBUG(D_READA, "ni %lu last %lu win %lu: extended window/last\n",
- ras->ras_next_index, ras->ras_last, ras->ras_window);
+ if (index != ras->ras_last_readpage + 1) {
+ ras_reset(ras, index);
+ GOTO(out_unlock, 0);
}
+ ras->ras_last_readpage = index;
+ ras->ras_consecutive++;
+ ras_set_start(ras, index);
+
+ if (ras->ras_consecutive == 2) {
+ ras->ras_window_len = PTLRPC_MAX_BRW_PAGES;
+ GOTO(out_unlock, 0);
+ }
+
+ /* we need to increase the window sometimes. we'll arbitrarily
+ * do it half-way through the pages in an rpc */
+ if ((index & (PTLRPC_MAX_BRW_PAGES - 1)) ==
+ (PTLRPC_MAX_BRW_PAGES >> 1)) {
+ ras->ras_window_len += PTLRPC_MAX_BRW_PAGES;
+ ras->ras_window_len = min(ras->ras_window_len, max);
+ }
+
+ EXIT;
+out_unlock:
+ RAS_CDEBUG(ras);
spin_unlock(&ras->ras_lock);
+ return;
}
/*
if (IS_ERR(llap))
GOTO(out, rc = PTR_ERR(llap));
+ if (ll_i2sbi(inode)->ll_flags & LL_SBI_READAHEAD)
+ ras_update(&fd->fd_ras, page->index,
+ ll_i2sbi(inode)->ll_max_read_ahead_pages);
+
if (llap->llap_defer_uptodate) {
- ll_readahead_update(inode, &fd->fd_ras, page->index, 1);
- ll_readahead(&fd->fd_ras, exp, page->mapping, oig,fd->fd_flags);
- obd_trigger_group_io(exp, ll_i2info(inode)->lli_smd, NULL,
- oig);
+ rc = ll_readahead(&fd->fd_ras, exp, page->mapping, oig,
+ fd->fd_flags);
+ if (rc > 0)
+ obd_trigger_group_io(exp, ll_i2info(inode)->lli_smd,
+ NULL, oig);
LL_CDEBUG_PAGE(D_PAGE, page, "marking uptodate from defer\n");
SetPageUptodate(page);
unlock_page(page);
GOTO(out_oig, rc = 0);
}
- ll_readahead_update(inode, &fd->fd_ras, page->index, 0);
-
rc = ll_page_matches(page, fd->fd_flags);
if (rc < 0) {
LL_CDEBUG_PAGE(D_ERROR, page, "lock match failed: rc %d\n", rc);
GOTO(out, rc);
LL_CDEBUG_PAGE(D_PAGE, page, "queued readpage\n");
- if ((ll_i2sbi(inode)->ll_flags & LL_SBI_READAHEAD))
- ll_readahead(&fd->fd_ras, exp, page->mapping, oig,fd->fd_flags);
+ if (ll_i2sbi(inode)->ll_flags & LL_SBI_READAHEAD)
+ ll_readahead(&fd->fd_ras, exp, page->mapping, oig,
+ fd->fd_flags);
rc = obd_trigger_group_io(exp, ll_i2info(inode)->lli_smd, NULL, oig);
oig_release(oig);
RETURN(rc);
}
-
-#if 0
-/* this is for read pages. we issue them as ready but not urgent. when
- * someone waits on them we fire them off, hopefully merged with adjacent
- * reads that were queued by read-ahead. */
-int ll_sync_page(struct page *page)
-{
- struct obd_export *exp;
- struct ll_async_page *llap;
- int rc;
- ENTRY;
-
- /* we're using a low bit flag to signify that a queued read should
- * be issued once someone goes to lock it. it is also cleared
- * as the page is built into an RPC */
- if (!test_and_clear_bit(LL_PRIVBITS_READ, &page->private))
- RETURN(0);
-
- /* careful to only deref page->mapping after checking the bit */
- exp = ll_i2obdexp(page->mapping->host);
- if (exp == NULL)
- RETURN(-EINVAL);
-
- llap = llap_from_page(page);
- if (IS_ERR(llap))
- RETURN(PTR_ERR(llap));
-
- LL_CDEBUG_PAGE(D_PAGE, page, "setting ready|urgent\n");
-
- rc = obd_set_async_flags(exp, ll_i2info(page->mapping->host)->lli_smd,
- NULL, llap->llap_cookie,
- ASYNC_READY|ASYNC_URGENT);
- return rc;
-}
-#endif
spin_unlock(&req->rq_lock);
}
+ DEBUG_REQ(D_RPCTRACE, req, "disposition: %x, status: %d",
+ it->d.lustre.it_disposition, it->d.lustre.it_status);
+
/* We know what to expect, so we do any byte flipping required here */
LASSERT(reply_buffers == 4 || reply_buffers == 3 || reply_buffers == 1);
if (reply_buffers >= 3) {
* we're sure we won't need to fix up the close request in the future),
* but make sure that replay doesn't poke at the och, which is about to
* be freed. */
- LASSERT(mod != (void *)0x5a5a5a5a);
+ LASSERT(mod != LP_POISON);
if (mod != NULL)
mod->mod_och = NULL;
och->och_mod = NULL;
open_req = mod->mod_open_req;
LASSERT(open_req != NULL);
- LASSERT(open_req != (void *)0x5a5a5a5a);
+ LASSERT(open_req != LP_POISON);
+ LASSERT(open_req->rq_type != LI_POISON);
DEBUG_REQ(D_HA, open_req, "open req balanced");
LASSERT(open_req->rq_transno != 0);
mod = och->och_mod;
if (likely(mod != NULL)) {
mod->mod_close_req = req;
+ LASSERT(mod->mod_open_req->rq_type != LI_POISON);
DEBUG_REQ(D_HA, mod->mod_open_req, "matched open req %p",
mod->mod_open_req);
} else {
if (req->rq_repmsg == NULL) {
CDEBUG(D_HA, "request failed to send: %p, %d\n", req,
req->rq_status);
- rc = req->rq_status;
+ rc = req->rq_status ? req->rq_status : -EIO;
} else if (rc == 0) {
rc = req->rq_repmsg->status;
if (req->rq_repmsg->type == PTL_RPC_MSG_ERR) {
GOTO(cleanup, rc = -EISDIR);
}
if (ll_permission(dchild->d_inode, acc_mode, NULL)) {
+ intent_set_disposition(rep, DISP_OPEN_OPEN);
GOTO(cleanup, rc = -EACCES);
}
}
* kernel patch */
#include <linux/lustre_version.h>
#define LUSTRE_MIN_VERSION 28
-#define LUSTRE_MAX_VERSION 34
+#define LUSTRE_MAX_VERSION 35
#if (LUSTRE_KERNEL_VERSION < LUSTRE_MIN_VERSION)
# error Cannot continue: Your Lustre kernel patch is older than the sources
#elif (LUSTRE_KERNEL_VERSION > LUSTRE_MAX_VERSION)
void class_destroy_import(struct obd_import *import)
{
LASSERT(import != NULL);
- LASSERT((unsigned long)import != 0x5a5a5a5a);
+ LASSERT(import != LP_POISON);
class_handle_unhash(&import->imp_handle);
#include <asm/div64.h>
#else
#include <liblustre.h>
-#include <sys/capability.h>
#endif
#include <linux/obd.h>
int rc;
ENTRY;
- CDEBUG(D_INODE, "server last_objid for group "LPU64": "LPU64"\n",
- group, filter->fo_last_objids[group]);
+ CDEBUG(D_INODE, "%s: server last_objid for group "LPU64": "LPU64"\n",
+ obd->obd_name, group, filter->fo_last_objids[group]);
tmp = cpu_to_le64(filter->fo_last_objids[group]);
rc = fsfilt_write_record(obd, filter->fo_last_objid_files[group],
}
filter->fo_last_objids[i] =
le64_to_cpu(filter->fo_last_objids[i]);
- CDEBUG(D_INODE, "%s: server last_objid group %d: "LPU64"\n",
+ CDEBUG(D_HA, "%s: server last_objid group %d: "LPU64"\n",
obd->obd_name, i, filter->fo_last_objids[i]);
}
doa.o_mode = S_IFREG;
last = filter_last_id(filter, &doa);
- CWARN("deleting orphan objects from "LPU64" to "LPU64"\n",
- oa->o_id + 1, last);
+ CWARN("%s: deleting orphan objects from "LPU64" to "LPU64"\n",
+ exp->exp_obd->obd_name, oa->o_id + 1, last);
for (id = oa->o_id + 1; id <= last; id++) {
doa.o_id = id;
filter_destroy(exp, &doa, NULL, NULL);
}
+
+ CDEBUG(D_HA, "%s: after destroy: set last_objids["LPU64"] = "LPU64"\n",
+ exp->exp_obd->obd_name, doa.o_gr, oa->o_id);
+
spin_lock(&filter->fo_objidlock);
filter->fo_last_objids[doa.o_gr] = oa->o_id;
spin_unlock(&filter->fo_objidlock);
recreate_obj = 1;
}
+ CDEBUG(D_HA, "%s: precreating %d objects\n", obd->obd_name, *num);
+
for (i = 0; i < *num && err == 0; i++) {
int cleanup_phase = 0;
* already exists
*/
if (recreate_obj) {
- CERROR("Serious error: recreating obj %*s but "
- "obj already exists \n",
- dchild->d_name.len, dchild->d_name.name);
+ CERROR("%s: Serious error: recreating obj %*s "
+ "but obj already exists \n",
+ obd->obd_name, dchild->d_name.len,
+ dchild->d_name.name);
+ LBUG();
} else {
- CERROR("Serious error: objid %*s already "
+ CERROR("%s: Serious error: objid %*s already "
"exists; is this filesystem corrupt?\n",
- dchild->d_name.len, dchild->d_name.name);
+ obd->obd_name, dchild->d_name.len,
+ dchild->d_name.name);
+ LBUG();
}
GOTO(cleanup, rc = -EEXIST);
}
}
*num = i;
- CDEBUG(D_INFO, "filter_precreate() created %d objects\n", i);
+ CDEBUG(D_HA, "%s: server last_objid for group "LPU64": "LPU64"\n",
+ obd->obd_name, group, filter->fo_last_objids[group]);
+
+ CDEBUG(D_HA, "%s: filter_precreate() created %d objects\n",
+ obd->obd_name, i);
RETURN(rc);
}
i++, lnb++, rnb++) {
/* We still set up for ungranted pages so that granted pages
* can be written to disk as they were promised, and portals
- * needs to keep the pages all aligned properly. */
+ * needs to keep the pages all aligned properly. */
lnb->dentry = dentry;
lnb->offset = rnb->offset;
lnb->len = rnb->len;
{
struct obd_device *dev = data;
struct client_obd *cli = &dev->u.cli;
- int val;
- int rc;
+ unsigned val;
spin_lock(&cli->cl_loi_list_lock);
val = cli->cl_dirty_max >> 20;
- rc = snprintf(page, count, "%d\n", val);
spin_unlock(&cli->cl_loi_list_lock);
- return rc;
+
+ return snprintf(page, count, "%u\n", val);
}
int osc_wr_max_dirty_mb(struct file *file, const char *buffer,
oscc->oscc_flags &= ~OSCC_FLAG_CREATING;
spin_unlock(&oscc->oscc_lock);
- CDEBUG(D_INFO, "preallocated through id "LPU64" (last used "LPU64")\n",
+ CDEBUG(D_HA, "preallocated through id "LPU64" (last used "LPU64")\n",
oscc->oscc_last_id, oscc->oscc_next_id);
wake_up(&oscc->oscc_waitq);
ENTRY;
spin_lock(&oscc->oscc_lock);
- if (oscc->oscc_flags & OSCC_FLAG_CREATING) {
+ if (oscc->oscc_flags & OSCC_FLAG_CREATING ||
+ oscc->oscc_flags & OSCC_FLAG_RECOVERING) {
spin_unlock(&oscc->oscc_lock);
RETURN(0);
}
spin_lock(&oscc->oscc_lock);
body->oa.o_id = oscc->oscc_last_id + oscc->oscc_grow_count;
body->oa.o_valid |= OBD_MD_FLID;
- CDEBUG(D_INFO, "preallocating through id "LPU64" (last used "LPU64")\n",
+ CDEBUG(D_HA, "preallocating through id "LPU64" (last used "LPU64")\n",
body->oa.o_id, oscc->oscc_next_id);
spin_unlock(&oscc->oscc_lock);
oa->o_valid |= OBD_MD_FLID;
oa->o_id = oscc->oscc_next_id - 1;
+ CDEBUG(D_HA, "%s: deleting to next_id: "LPU64"\n",
+ oscc->oscc_obd->u.cli.cl_import->imp_target_uuid.uuid,
+ oa->o_id);
+
rc = osc_real_create(exp, oa, ea, NULL);
spin_lock(&oscc->oscc_lock);
RETURN(rc);
}
- /* If orphans are being recovered, then we must wait until it is
- finished before we can continue with create. */
- if (oscc_recovering(oscc)) {
- struct l_wait_info lwi;
-
- CDEBUG(D_HA, "%p: oscc recovery in progress, waiting\n", oscc);
-
- lwi = LWI_TIMEOUT(MAX(obd_timeout * HZ, 1), NULL, NULL);
- rc = l_wait_event(oscc->oscc_waitq, !oscc_recovering(oscc),
- &lwi);
- LASSERT(rc == 0 || rc == -ETIMEDOUT);
- if (rc == -ETIMEDOUT) {
- CDEBUG(D_HA, "%p: timed out waiting for recovery\n", oscc);
- RETURN(rc);
- }
- CDEBUG(D_HA, "%p: oscc recovery over, waking up\n", oscc);
- }
-
-
while (try_again) {
+ /* If orphans are being recovered, then we must wait until
+ it is finished before we can continue with create. */
+ if (oscc_recovering(oscc)) {
+ struct l_wait_info lwi;
+
+ CDEBUG(D_HA,"%p: oscc recovery in progress, waiting\n",
+ oscc);
+
+ lwi = LWI_TIMEOUT(MAX(obd_timeout * HZ, 1), NULL, NULL);
+ rc = l_wait_event(oscc->oscc_waitq,
+ !oscc_recovering(oscc), &lwi);
+ LASSERT(rc == 0 || rc == -ETIMEDOUT);
+ if (rc == -ETIMEDOUT) {
+ CDEBUG(D_HA, "%p: timed out waiting for "
+ "recovery\n", oscc);
+ RETURN(rc);
+ }
+ CDEBUG(D_HA, "%p: oscc recovery over, waking up\n",
+ oscc);
+ }
+
spin_lock(&oscc->oscc_lock);
if (oscc->oscc_last_id >= oscc->oscc_next_id) {
memcpy(oa, &oscc->oscc_oa, sizeof(*oa));
}
if (rc == 0)
- CDEBUG(D_INFO, "returning objid "LPU64"\n", lsm->lsm_object_id);
+ CDEBUG(D_HA, "%s: returning objid "LPU64"\n",
+ oscc->oscc_obd->u.cli.cl_import->imp_target_uuid.uuid,
+ lsm->lsm_object_id);
else if (*ea == NULL)
obd_free_memmd(exp, &lsm);
RETURN(rc);
if (vallen != sizeof(obd_id))
RETURN(-EINVAL);
obd->u.cli.cl_oscc.oscc_next_id = *((obd_id*)val) + 1;
- CDEBUG(D_INODE, "%s: set oscc_next_id = "LPU64"\n",
+ CDEBUG(D_HA, "%s: set oscc_next_id = "LPU64"\n",
exp->exp_obd->obd_name,
obd->u.cli.cl_oscc.oscc_next_id);
# endif
#endif
+#ifndef LP_POISON
+# define LI_POISON ((int)0x5a5a5a5a5a5a5a5a)
+# define LL_POISON ((long)0x5a5a5a5a5a5a5a5a)
+# define LP_POISON ((void *)(long)0x5a5a5a5a5a5a5a5a)
+#endif
+
#if defined(__x86_64__)
# define LPU64 "%Lu"
# define LPD64 "%Ld"
# define LPX64 "%#Lx"
# define LPSZ "%lu"
# define LPSSZ "%ld"
-# define LP_POISON ((void *)0x5a5a5a5a5a5a5a5a)
#elif (BITS_PER_LONG == 32 || __WORDSIZE == 32)
# define LPU64 "%Lu"
# define LPD64 "%Ld"
# define LPX64 "%#Lx"
# define LPSZ "%u"
# define LPSSZ "%d"
-# define LP_POISON ((void *)0x5a5a5a5a)
#elif (BITS_PER_LONG == 64 || __WORDSIZE == 64)
# define LPU64 "%lu"
# define LPD64 "%ld"
# define LPX64 "%#lx"
# define LPSZ "%lu"
# define LPSSZ "%ld"
-# define LP_POISON ((void *)0x5a5a5a5a5a5a5a5a)
#endif
#ifndef LPU64
# error "No word size defined"
ENTRY;
LASSERT(desc != NULL);
- LASSERT(desc->bd_iov_count != 0x5a5a5a5a); /* not freed already */
+ LASSERT(desc->bd_iov_count != LI_POISON); /* not freed already */
LASSERT(!desc->bd_network_rw); /* network hands off or */
LASSERT((desc->bd_export != NULL) ^ (desc->bd_import != NULL));
if (desc->bd_export)
SIGNAL_MASK_UNLOCK(current, flags);
LASSERTF(strlen(data->name) < sizeof(current->comm),
- "name %d > len %d\n",strlen(data->name),sizeof(current->comm));
+ "name %d > len %d\n",
+ (int)strlen(data->name), (int)sizeof(current->comm));
THREAD_NAME(current->comm, sizeof(current->comm) - 1, "%s", data->name);
unlock_kernel();
/* Wait for recovery to complete and resend. If evicted, then
this request will be errored out later.*/
spin_lock_irqsave(&failed_req->rq_lock, flags);
- if (!failed_req->rq_no_resend)
- failed_req->rq_resend = 1;
+ failed_req->rq_resend = 1;
spin_unlock_irqrestore(&failed_req->rq_lock, flags);
EXIT;
SIGNAL_MASK_UNLOCK(current, flags);
LASSERTF(strlen(data->name) < sizeof(current->comm),
- "name %d > len %d\n",strlen(data->name),sizeof(current->comm));
+ "name %d > len %d\n",
+ (int)strlen(data->name), (int)sizeof(current->comm));
THREAD_NAME(current->comm, sizeof(current->comm) - 1, "%s", data->name);
unlock_kernel();
my $tagfile = new IO::File;
if (!$tagfile->open("CVS/Tag")) {
my $verfile = new IO::File;
- if (!$verfile->open("portals/include/config.h")) {
+ if (!$verfile->open("include/config.h")) {
return "UNKNOWN";
}
while(defined($line = <$verfile>)) {
if [ "$DBENCH" != "no" ]; then
mount | grep $MOUNT || sh llmount.sh
- SPACE=`df $MOUNT | tail -1 | awk '{ print $4 }'`
+ SPACE=`df $MOUNT | tail -n 1 | awk '{ print $4 }'`
DB_THREADS=`expr $SPACE / 50000`
[ $THREADS -lt $DB_THREADS ] && DB_THREADS=$THREADS
fi
if [ "$IOZONE_DIR" != "no" ]; then
mount | grep $MOUNT || sh llmount.sh
- SPACE=`df $MOUNT | tail -1 | awk '{ print $4 }'`
+ SPACE=`df $MOUNT | tail -n 1 | awk '{ print $4 }'`
IOZ_THREADS=`expr $SPACE / \( $SIZE + $SIZE / 512 \)`
[ $THREADS -lt $IOZ_THREADS ] && IOZ_THREADS=$THREADS
NETTYPE=${NETTYPE:-tcp}
TIMEOUT=${TIMEOUT:-30}
-PTLDEBUG=${PTLDEBUG:-0}
-SUBSYSTEM=${SUBSYSTEM:-0}
+PTLDEBUG=${PTLDEBUG:-0x3f0400}
+SUBSYSTEM=${SUBSYSTEM:- 0xffb7e3ff}
MOUNT=${MOUNT:-"/mnt/lustre"}
#CLIENT_UPCALL=${CLIENT_UPCALL:-`pwd`/client-upcall-mdev.sh}
UPCALL=${CLIENT_UPCALL:-`pwd`/replay-single-upcall.sh}
wait_for mds
start mds $MDSLCONFARGS ${REFORMAT}
while ! do_node $CLIENTS "ls -d $LUSTRE" > /dev/null; do sleep 5; done
- zconf_mount $CLIENTS $MOUNT
+ grep " $MOUNT " /proc/mounts || zconf_mount $CLIENTS $MOUNT
}
file=$1
for c in $LIVE_CLIENT $FAIL_CLIENTS; do
if echo $DOWN_CLIENTS | grep -q $c; then continue; fi
- $PDSH $c touch $MOUNT/${c}_$file
+ $PDSH $c touch $MOUNT/${c}_$file || return 1
done
}
mv $TMP/debug $TMP/debug-busy.`date +%s`
exit 255
fi
-LEAK_LUSTRE=`dmesg | tail -30 | grep "obd mem.*leaked"`
-LEAK_PORTALS=`dmesg | tail -20 | grep "Portals memory leaked"`
+LEAK_LUSTRE=`dmesg | tail -n 30 | grep "obd mem.*leaked"`
+LEAK_PORTALS=`dmesg | tail -n 20 | grep "Portals memory leaked"`
if [ "$LEAK_LUSTRE" -o "$LEAK_PORTALS" ]; then
echo "$LEAK_LUSTRE" 1>&2
echo "$LEAK_PORTALS" 1>&2
perror("write");
exit(1);
}
- /* b=3043 write() on Suse x86-64 is returning -errno
- instead of -1, and not setting errno. */
- if (rc < 0) {
- fprintf(stderr, "MULTIOP: broken write() "
- "returned %d, errno %d\n",
- rc, errno);
- exit(1);
- }
break;
}
case 'W':
#lctl clear
#lctl debug_daemon start /r/tmp/debug 1024
-STRIPECOUNT=`cat /proc/fs/lustre/lov/*/activeobd | head -1`
-ORIGFREE=`cat /proc/fs/lustre/llite/*/kbytesavail | head -1`
+STRIPECOUNT=`cat /proc/fs/lustre/lov/*/activeobd | head -n 1`
+ORIGFREE=`cat /proc/fs/lustre/llite/*/kbytesavail | head -n 1`
MAXFREE=${MAXFREE:-$((200000 * $STRIPECOUNT))}
if [ $ORIGFREE -gt $MAXFREE ]; then
echo "skipping out-of-space test on $OSC"
sleep 1 # to ensure we get up-to-date statfs info
-STRIPECOUNT=`cat /proc/fs/lustre/lov/*/activeobd | head -1`
-ORIGFREE=`cat /proc/fs/lustre/llite/*/kbytesavail | head -1`
+STRIPECOUNT=`cat /proc/fs/lustre/lov/*/activeobd | head -n 1`
+ORIGFREE=`cat /proc/fs/lustre/llite/*/kbytesavail | head -n 1`
MAXFREE=${MAXFREE:-$((200000 * $STRIPECOUNT))}
if [ $ORIGFREE -gt $MAXFREE ]; then
echo "skipping out-of-space test on $OSC"
start ost2 --reformat $OSTLCONFARGS
[ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
start mds $MDSLCONFARGS --reformat
- zconf_mount `hostname` $MOUNT
+ grep " $MOUNT " /proc/mounts || zconf_mount `hostname` $MOUNT
}
cleanup() {
do_facet client "cmp /etc/termcap $MOUNT/termcap" && return 1
sysctl -w lustre.fail_loc=0
# give recovery a chance to finish (shouldn't take long)
- sleep 1
+ sleep $TIMEOUT
do_facet client "cmp /etc/termcap $MOUNT/termcap" || return 2
}
run_test 16 "timeout bulk put, evict client (2732)"
start ost2 --reformat $OSTLCONFARGS
[ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
start mds $MDSLCONFARGS --reformat
-zconf_mount `hostname` $MOUNT
-zconf_mount `hostname` $MOUNT2
+grep " $MOUNT " /proc/mounts || zconf_mount `hostname` $MOUNT
+grep " $MOUNT2 " /proc/mounts || zconf_mount `hostname` $MOUNT2
echo $TIMEOUT > /proc/sys/lustre/timeout
echo $UPCALL > /proc/sys/lustre/upcall
start ost --reformat $OSTLCONFARGS
[ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
start mds --reformat $MDSLCONFARGS
- zconf_mount `hostname` $MOUNT
+ grep " $MOUNT " /proc/mounts || zconf_mount `hostname` $MOUNT
}
mkdir -p $DIR
start ost2 --reformat $OSTLCONFARGS
[ "$DAEMONFILE" ] && $LCTL debug_daemon start $DAEMONFILE $DAEMONSIZE
start mds $MDSLCONFARGS --reformat
- zconf_mount `hostname` $MOUNT
- echo 0x3f0410 > /proc/sys/portals/debug
+ grep " $MOUNT " /proc/mounts || zconf_mount `hostname` $MOUNT
}
$SETUP
}
run_test 0 "empty replay"
+test_0b() {
+ # this test attempts to trigger a race in the precreation code,
+ # and must run before any other objects are created on the filesystem
+ fail ost
+ createmany -o $DIR/$tfile 20 || return 1
+ unlinkmany $DIR/$tfile 20 || return 2
+}
+run_test 0b "ensure object created after recover exists. (3284)"
+
test_1() {
replay_barrier mds
mcreate $DIR/$tfile
# test MDS recovery after ost failure
test_42() {
- blocks=`df $MOUNT | tail -1 | awk '{ print $1 }'`
+ blocks=`df $MOUNT | tail -n 1 | awk '{ print $1 }'`
createmany -o $DIR/$tfile-%d 800
replay_barrier ost
unlinkmany $DIR/$tfile-%d 0 400
facet_failover ost
# osc is evicted, fs is smaller
- blocks_after=`df $MOUNT | tail -1 | awk '{ print $1 }'`
+ blocks_after=`df $MOUNT | tail -n 1 | awk '{ print $1 }'`
[ $blocks_after -lt $blocks ] || return 1
echo wait for MDS to timeout and recover
sleep $((TIMEOUT * 2))
PATH=`dirname $0`:`dirname $0`/../utils:$PATH
TMP=${TMP:-/tmp}
-MDS=`ls /proc/fs/lustre/mds | grep -v num_refs | head -1`
+MDS=`ls /proc/fs/lustre/mds | grep -v num_refs | head -n 1`
[ -z "$MDS" ] && echo "no MDS available, skipping llog test" && exit 0
insmod ../obdclass/llog_test.o || exit 1
#include <unistd.h>
#include <string.h>
#include <errno.h>
+#include <ctype.h>
#include <sys/types.h>
#include <grp.h>
#include <sys/wait.h>
#define DEBUG 0
+#ifndef NGROUPS_MAX
+#define NGROUPS_MAX 32
+#endif
+
static const char usage[] =
"Usage: %s -u user_id [-g grp_id ] [ -G ] command\n"
-" -u user_id switch to UID user_id\n"
-" -g grp_id switch to GID grp_id\n"
-" -G clear supplementary groups\n";
+" -u user_id switch to UID user_id\n"
+" -g grp_id switch to GID grp_id\n"
+" -G[gid0,gid1,...] set supplementary groups\n";
void Usage_and_abort(const char *name)
{
int main(int argc, char **argv)
{
- char **my_argv, *name = argv[0];
- int status;
- int c,i;
- int gid_is_set = 0;
- int uid_is_set = 0;
- int clear_supp_groups = 0;
+ char **my_argv, *name = argv[0], *grp;
+ int status, c, i;
+ int gid_is_set = 0, uid_is_set = 0, num_supp = -1;
uid_t user_id = 0;
- gid_t grp_id = 0;
+ gid_t grp_id = 0, supp_groups[NGROUPS_MAX] = { 0 };
if (argc == 1)
Usage_and_abort(name);
// get UID and GID
- while ((c = getopt (argc, argv, "+u:g:hG")) != -1) {
+ while ((c = getopt(argc, argv, "+u:g:hG::")) != -1) {
switch (c) {
case 'u':
user_id = (uid_t)atoi(optarg);
break;
case 'G':
- clear_supp_groups = 1;
+ num_supp = 0;
+ if (optarg == NULL || !isdigit(optarg[0]))
+ break;
+ while ((grp = strsep(&optarg, ",")) != NULL) {
+ printf("adding supp group %d\n", atoi(grp));
+ supp_groups[num_supp++] = atoi(grp);
+ if (num_supp >= NGROUPS_MAX)
+ break;
+ }
break;
default:
exit(-1);
}
- if (clear_supp_groups) {
- status = setgroups(0, NULL);
+ if (num_supp >= 0) {
+ status = setgroups(num_supp, supp_groups);
if (status == -1) {
- perror("clearing supplementary groups");
+ perror("setting supplementary groups");
exit(-1);
}
}
-
+
// set UID
status = setreuid(user_id, user_id );
if(status == -1) {
exit(-1);
}
- fprintf(stderr, "running as UID %d, GID %d%s:", user_id, grp_id,
- clear_supp_groups ? ", cleared groups" : "");
+ fprintf(stderr, "running as UID %d, GID %d", user_id, grp_id);
+ for (i = 0; i < num_supp; i++)
+ fprintf(stderr, ":%d", supp_groups[i]);
+ fprintf(stderr, "\n");
for (i = 0; i < argc - optind; i++)
fprintf(stderr, " [%s]", my_argv[i]);
ENDRUN=endrun-`hostname`
-ECHONAME="`lctl device_list 2> /dev/null | awk '/ echo_client / { print $4 }' | tail -1`"
+ECHONAME="`lctl device_list 2> /dev/null | awk '/ echo_client / { print $4 }' | tail -n 1`"
if [ -z "$ECHONAME" ]; then
echo "$0: needs an ECHO_CLIENT set up first" 1>&2
[ -z "$*" ] && fail "usage: $0 [--reformat] <conf>.xml" 1
-OSCMT="`mount | awk '/ lustre_lite / { print $3 }' | tail -1`"
+OSCMT="`mount | awk '/ lustre_lite / { print $3 }' | tail -n 1`"
if [ -z "$OSCMT" ]; then
$LCONF $@ || exit 1
trap cleanup 0
- OSCMT="`mount | awk '/ lustre_lite / { print $3 }' | tail -1`"
+ OSCMT="`mount | awk '/ lustre_lite / { print $3 }' | tail -n 1`"
[ -z "$OSCMT" ] && fail "no lustre filesystem mounted" 1
fi
done
OSCTMP=`echo $OSCMT | tr "/" "."`
-USED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -1`
+USED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -n 1`
USED=`expr $USED + 16` # Some space for the status file
THREADS=1
rm -f $ENDRUN
-NOWUSED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -1`
+NOWUSED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -n 1`
if [ $NOWUSED -gt $USED ]; then
echo "Space not all freed: now ${NOWUSED}kB, was ${USED}kB." 1>&2
echo "This is normal on BA OSTs, because of subdirectories." 1>&2
ENDRUN=endrun-`hostname`
-ECHONAME="`lctl device_list 2> /dev/null | awk '/ echo_client / { print $4 }' | tail -1`"
+ECHONAME="`lctl device_list 2> /dev/null | awk '/ echo_client / { print $4 }' | tail -n 1`"
if [ -z "$ECHONAME" ]; then
echo "$0: needs an ECHO_CLIENT set up first" 1>&2
shift
done
-MOUNT="`mount | awk '/ lustre_lite / { print $3 }' | tail -1`"
+MOUNT="`mount | awk '/ lustre_lite / { print $3 }' | tail -n 1`"
if [ -z "$MOUNT" ]; then
sh llmount.sh
- MOUNT="`mount | awk '/ lustre_lite / { print $3 }' | tail -1`"
+ MOUNT="`mount | awk '/ lustre_lite / { print $3 }' | tail -n 1`"
[ -z "$MOUNT" ] && fail "no lustre filesystem mounted" 1
I_MOUNTED="yes"
fi
OSCTMP=`echo $MOUNT | tr "/" "."`
-USED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -1`
+USED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -n 1`
USED=`expr $USED + 16` # Some space for the status file
# let's start slowly here...
# ok, that hopefully worked, so let's do a little more, with files that
# haven't changed in the last day (hopefully they don't change during test)
-FILES=`find $SRC -type f -mtime +1 -ctime +1 | head -$COUNT`
+FILES=`find $SRC -type f -mtime +1 -ctime +1 | head -n $COUNT`
log "copying files from $SRC to $DST$SRC"
tar cf - $FILES | tar xvf - -C $DST || fail "copying $SRC" 11
log "done"
-NOWUSED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -1`
+NOWUSED=`df | awk "/$OSCTMP/ { print \\$3 }" | tail -n 1`
if [ `expr $NOWUSED - $USED` -gt 1024 ]; then
echo "Space not all freed: now ${NOWUSED}kB, was ${USED}kB." 1>&2
fi
run_test 36d "non-root OST utime check (open, utime) ==========="
test_36e() {
- [ $RUNAS_ID -eq $UID ] && return
+ [ $RUNAS_ID -eq $UID ] && echo "skipping test 36e" && return
[ ! -d $DIR/d36 ] && mkdir $DIR/d36
touch $DIR/d36/f36e
$RUNAS utime $DIR/d36/f36e && error "utime worked, want failure" || true
# bug 2319 - oig_wait() interrupted causes crash because of invalid waitq.
test_63() {
- MAX_DIRTY_MB=`cat /proc/fs/lustre/osc/*/max_dirty_mb | head -1`
+ MAX_DIRTY_MB=`cat /proc/fs/lustre/osc/*/max_dirty_mb | head -n 1`
for i in /proc/fs/lustre/osc/*/max_dirty_mb ; do
echo 0 > $i
done
}
run_test 66 "update inode blocks count on client ==============="
+test_67() { # bug 3285 - supplementary group fails on MDS, passes on client
+ [ "$RUNAS_ID" = "$UID" ] && echo "skipping test 67" && return
+ mkdir $DIR/d67
+ chmod 771 $DIR/d67
+ chgrp $RUNAS_ID $DIR/d67
+ $RUNAS -g $((RUNAS_ID + 1)) -G1,2,$RUNAS_ID ls $DIR/d67 && error || true
+}
+run_test 67 "supplementary group failure (should return error) ="
+
# on the LLNL clusters, runas will still pick up root's $TMP settings,
# which will not be writable for the runas user, and then you get a CVS
# error message with a corrupt path string (CVS bug) and panic.
echo PASS
}
-export MOUNT1=`mount| awk '/ lustre/ { print $3 }'| head -1`
-export MOUNT2=`mount| awk '/ lustre/ { print $3 }'| tail -1`
+export MOUNT1=`mount| awk '/ lustre/ { print $3 }'| head -n 1`
+export MOUNT2=`mount| awk '/ lustre/ { print $3 }'| tail -n 1`
[ -z "$MOUNT1" ] && error "NAME=$NAME not mounted once"
[ "$MOUNT1" = "$MOUNT2" ] && error "NAME=$NAME not mounted twice"
[ `mount| awk '/ lustre/ { print $3 }'| wc -l` -ne 2 ] && \
init_test_env() {
export LUSTRE=`absolute_path $LUSTRE`
export TESTSUITE=`basename $0 .sh`
- export XMLCONFIG="${TESTSUITE}.xml"
+ export XMLCONFIG=${XMLCONFIG:-${TESTSUITE}.xml}
export LTESTDIR=${LTESTDIR:-$LUSTRE/../ltest}
[ -d /r ] && export ROOT=/r