From: liangzhen Date: Fri, 25 Jun 2004 09:32:38 +0000 (+0000) Subject: Adapt the 0-copy TCP kernel patch for 2.4.21 Redhat AS Kernel X-Git-Tag: v1_7_0_51~2^7~48 X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=5df71948a039c694115900396e697d065232816f;p=fs%2Flustre-release.git Adapt the 0-copy TCP kernel patch for 2.4.21 Redhat AS Kernel Series is rh-as-2.4.21-4.EL --- diff --git a/lustre/kernel_patches/patches/configurable-x86-stack-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/configurable-x86-stack-2.4.21-rh-as.patch new file mode 100644 index 0000000..de5f146 --- /dev/null +++ b/lustre/kernel_patches/patches/configurable-x86-stack-2.4.21-rh-as.patch @@ -0,0 +1,310 @@ +Index: linux-2.4.21-4.EL/arch/i386/kernel/entry.S +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/kernel/entry.S 2003-10-04 05:28:53.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/kernel/entry.S 2004-06-23 18:16:17.000000000 +0800 +@@ -47,6 +47,7 @@ + #include + #include + #include ++#include + + EBX = 0x00 + ECX = 0x04 +@@ -94,10 +95,6 @@ + + ENOSYS = 38 + +-#define GET_CURRENT(reg) \ +- movl $-8192, reg; \ +- andl %esp, reg +- + #if CONFIG_X86_HIGH_ENTRY + + #define call_SYMBOL_NAME_ABS(X) movl $X, %ebp; call *%ebp +@@ -193,7 +190,7 @@ + GET_CURRENT(%ebx); \ + movl real_stack(%ebx), %edx; \ + movl %esp, %ebx; \ +- andl $0x1fff, %ebx; \ ++ andl $(THREAD_SIZE-1), %ebx; \ + orl %ebx, %edx; \ + movl %edx, %esp; + +@@ -228,7 +225,7 @@ + return_path_start_marker: \ + nop; \ + movl %esp, %ebx; \ +- andl $0x1fff, %ebx; \ ++ andl $(THREAD_SIZE-1), %ebx; \ + orl %ebx, %edx; \ + movl %esp, %eax; \ + movl %edx, %esp; \ +Index: linux-2.4.21-4.EL/arch/i386/kernel/smpboot.c +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/kernel/smpboot.c 2003-10-04 05:28:37.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/kernel/smpboot.c 2004-06-23 18:13:37.000000000 +0800 +@@ -814,7 +814,7 @@ + + /* 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 +@@ -887,7 +887,7 @@ + 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"); +@@ -910,7 +910,7 @@ + } + + /* mark "stuck" area as not stuck */ +- *((volatile unsigned long *)phys_to_virt(8192)) = 0; ++ *((volatile unsigned long *)phys_to_virt(THREAD_SIZE)) = 0; + + if(clustered_apic_mode == CLUSTERED_APIC_NUMAQ) { + printk("Restoring NMI vector\n"); +Index: linux-2.4.21-4.EL/arch/i386/kernel/traps.c +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/kernel/traps.c 2003-10-04 05:28:44.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/kernel/traps.c 2004-06-23 18:13:37.000000000 +0800 +@@ -180,7 +180,7 @@ + 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-2.4.21-4.EL/arch/i386/kernel/head.S +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/kernel/head.S 2003-10-04 05:28:40.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/kernel/head.S 2004-06-23 18:13:37.000000000 +0800 +@@ -15,6 +15,7 @@ + #include + #include + #include ++#include + + #define OLD_CL_MAGIC_ADDR 0x90020 + #define OLD_CL_MAGIC 0xA33F +@@ -315,7 +316,7 @@ + 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-2.4.21-4.EL/arch/i386/kernel/irq.c +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/kernel/irq.c 2003-10-04 05:28:50.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/kernel/irq.c 2004-06-23 18:13:37.000000000 +0800 +@@ -585,7 +585,10 @@ + 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) + 1024))) { + extern void show_stack(unsigned long *); + +Index: linux-2.4.21-4.EL/arch/i386/lib/getuser.S +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/lib/getuser.S 1998-01-13 05:42:52.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/lib/getuser.S 2004-06-23 18:13:37.000000000 +0800 +@@ -21,6 +21,10 @@ + * as they get called from within inline assembly. + */ + ++/* Duplicated from asm/processor.h */ ++#include ++#include ++ + 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 $~(THREAD_SIZE - 1),%edx + cmpl addr_limit(%edx),%eax + jae bad_get_user + 3: movl -3(%eax),%edx +Index: linux-2.4.21-4.EL/arch/i386/config.in +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/config.in 2003-10-04 05:28:53.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/config.in 2004-06-23 18:13:37.000000000 +0800 +@@ -302,6 +302,29 @@ + if [ "$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-2.4.21-4.EL/arch/i386/vmlinux.lds +=================================================================== +--- linux-2.4.21-4.EL.orig/arch/i386/vmlinux.lds 2003-10-04 05:29:44.000000000 +0800 ++++ linux-2.4.21-4.EL/arch/i386/vmlinux.lds 2004-06-23 18:14:05.000000000 +0800 +@@ -38,7 +38,7 @@ + + _edata = .; + +- . = ALIGN(8192); ++ . = ALIGN(65536); + .data.init_task : { *(.data.init_task) } + + entry_tramp_start = .; +Index: linux-2.4.21-4.EL/include/asm-i386/current.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/asm-i386/current.h 1998-08-15 07:35:22.000000000 +0800 ++++ linux-2.4.21-4.EL/include/asm-i386/current.h 2004-06-23 18:13:37.000000000 +0800 +@@ -1,15 +1,43 @@ + #ifndef _I386_CURRENT_H + #define _I386_CURRENT_H ++#include ++ ++/* ++ * 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-2.4.21-4.EL/include/asm-i386/hw_irq.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/asm-i386/hw_irq.h 2003-10-04 05:29:48.000000000 +0800 ++++ linux-2.4.21-4.EL/include/asm-i386/hw_irq.h 2004-06-23 18:13:37.000000000 +0800 +@@ -146,10 +146,6 @@ + #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-2.4.21-4.EL/include/asm-i386/processor.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/asm-i386/processor.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/asm-i386/processor.h 2004-06-23 18:13:37.000000000 +0800 +@@ -14,6 +14,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -494,10 +495,6 @@ + #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) do { BUG_ON((p)->state < TASK_ZOMBIE); free_pages((unsigned long) (p), 1); } while (0) +- + #define init_task (init_task_union.task) + #define init_stack (init_task_union.stack) + +Index: linux-2.4.21-4.EL/include/linux/sched.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/sched.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/sched.h 2004-06-23 18:13:37.000000000 +0800 +@@ -2,6 +2,7 @@ + #define _LINUX_SCHED_H + + #include /* for HZ */ ++#include /* maybe for INIT_TASK_SIZE */ + + extern unsigned long event; + diff --git a/lustre/kernel_patches/patches/dev_read_only_2.4.21-rh-as.patch b/lustre/kernel_patches/patches/dev_read_only_2.4.21-rh-as.patch new file mode 100644 index 0000000..d7e2e03 --- /dev/null +++ b/lustre/kernel_patches/patches/dev_read_only_2.4.21-rh-as.patch @@ -0,0 +1,58 @@ +Index: linux-2.4.21-4.EL/drivers/block/blkpg.c +=================================================================== +--- linux-2.4.21-4.EL.orig/drivers/block/blkpg.c 2003-10-04 05:28:52.000000000 +0800 ++++ linux-2.4.21-4.EL/drivers/block/blkpg.c 2004-06-24 14:07:34.000000000 +0800 +@@ -535,3 +535,37 @@ + + #endif /* CONFIG_IA64 */ + ++#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-2.4.21-4.EL/drivers/block/ll_rw_blk.c +=================================================================== +--- linux-2.4.21-4.EL.orig/drivers/block/ll_rw_blk.c 2003-10-04 05:28:49.000000000 +0800 ++++ linux-2.4.21-4.EL/drivers/block/ll_rw_blk.c 2004-06-24 14:07:34.000000000 +0800 +@@ -1314,7 +1314,10 @@ + buffer_IO_error(bh); + break; + } +- ++ if ((rw & WRITE)&&(dev_check_rdonly(bh->b_rdev))) { ++ bh->b_end_io(bh, 0); ++ break; ++ } + } while (q->make_request_fn(q, rw, bh)); + } + diff --git a/lustre/kernel_patches/patches/dynamic-locks-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/dynamic-locks-2.4.21-rh-as.patch new file mode 100644 index 0000000..e15c6ac --- /dev/null +++ b/lustre/kernel_patches/patches/dynamic-locks-2.4.21-rh-as.patch @@ -0,0 +1,217 @@ + include/linux/dynlocks.h | 33 ++++++++++ + lib/Makefile | 4 - + lib/dynlocks.c | 152 +++++++++++++++++++++++++++++++++++++++++++++++ + 3 files changed, 187 insertions(+), 2 deletions(-) + +Index: linux-2.4.21-4.EL/include/linux/dynlocks.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/dynlocks.h 2003-01-30 18:24:37.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/dynlocks.h 2004-06-25 15:48:28.000000000 +0800 +@@ -0,0 +1,33 @@ ++#ifndef _LINUX_DYNLOCKS_H ++#define _LINUX_DYNLOCKS_H ++ ++#include ++#include ++ ++struct dynlock_member { ++ struct list_head dl_list; ++ unsigned long dl_value; /* lock value */ ++ int dl_refcount; /* number of users */ ++ int dl_readers; ++ int dl_writers; ++ int dl_pid; /* holder of the lock */ ++ wait_queue_head_t dl_wait; ++}; ++ ++/* ++ * lock's namespace: ++ * - list of locks ++ * - lock to protect this list ++ */ ++struct dynlock { ++ struct list_head dl_list; ++ spinlock_t dl_list_lock; ++}; ++ ++void dynlock_init(struct dynlock *dl); ++void *dynlock_lock(struct dynlock *dl, unsigned long value, int rw, int gfp); ++void dynlock_unlock(struct dynlock *dl, void *lock); ++ ++ ++#endif ++ +Index: linux-2.4.21-4.EL/lib/dynlocks.c +=================================================================== +--- linux-2.4.21-4.EL.orig/lib/dynlocks.c 2003-01-30 18:24:37.000000000 +0800 ++++ linux-2.4.21-4.EL/lib/dynlocks.c 2004-06-25 15:48:28.000000000 +0800 +@@ -0,0 +1,152 @@ ++/* ++ * Dynamic Locks ++ * ++ * struct dynlock is lockspace ++ * one may request lock (exclusive or shared) for some value ++ * in that lockspace ++ * ++ */ ++ ++#include ++#include ++#include ++#include ++ ++/* ++ * dynlock_init ++ * ++ * initialize lockspace ++ * ++ */ ++void dynlock_init(struct dynlock *dl) ++{ ++ spin_lock_init(&dl->dl_list_lock); ++ INIT_LIST_HEAD(&dl->dl_list); ++} ++ ++/* ++ * dynlock_lock ++ * ++ * acquires lock (exclusive or shared) in specified lockspace ++ * each lock in lockspace is allocated separately, so user have ++ * to specify GFP flags. ++ * routine returns pointer to lock. this pointer is intended to ++ * be passed to dynlock_unlock ++ * ++ */ ++void *dynlock_lock(struct dynlock *dl, unsigned long value, int rw, int gfp) ++{ ++ struct dynlock_member *nhl = NULL; ++ struct dynlock_member *hl; ++ struct list_head *cur; ++ ++repeat: ++ /* find requested lock in lockspace */ ++ spin_lock(&dl->dl_list_lock); ++ list_for_each(cur, &dl->dl_list) { ++ hl = list_entry(cur, struct dynlock_member, dl_list); ++ if (hl->dl_value == value) { ++ /* lock is found */ ++ if (nhl) { ++ /* someone else just allocated ++ * lock we didn't find and just created ++ * so, we drop our lock ++ */ ++ kfree(nhl); ++ nhl = NULL; ++ } ++ hl->dl_refcount++; ++ goto found; ++ } ++ } ++ /* lock not found */ ++ if (nhl) { ++ /* we already have allocated lock. use it */ ++ hl = nhl; ++ nhl = NULL; ++ list_add(&hl->dl_list, &dl->dl_list); ++ goto found; ++ } ++ spin_unlock(&dl->dl_list_lock); ++ ++ /* lock not found and we haven't allocated lock yet. allocate it */ ++ nhl = kmalloc(sizeof(struct dynlock_member), gfp); ++ if (nhl == NULL) ++ return NULL; ++ nhl->dl_refcount = 1; ++ nhl->dl_value = value; ++ nhl->dl_readers = 0; ++ nhl->dl_writers = 0; ++ init_waitqueue_head(&nhl->dl_wait); ++ ++ /* while lock is being allocated, someone else may allocate it ++ * and put onto to list. check this situation ++ */ ++ goto repeat; ++ ++found: ++ if (rw) { ++ /* exclusive lock: user don't want to share lock at all ++ * NOTE: one process may take the same lock several times ++ * this functionaly is useful for rename operations */ ++ while ((hl->dl_writers && hl->dl_pid != current->pid) || ++ hl->dl_readers) { ++ spin_unlock(&dl->dl_list_lock); ++ wait_event(hl->dl_wait, ++ hl->dl_writers == 0 && hl->dl_readers == 0); ++ spin_lock(&dl->dl_list_lock); ++ } ++ hl->dl_writers++; ++ } else { ++ /* shared lock: user do not want to share lock with writer */ ++ while (hl->dl_writers) { ++ spin_unlock(&dl->dl_list_lock); ++ wait_event(hl->dl_wait, hl->dl_writers == 0); ++ spin_lock(&dl->dl_list_lock); ++ } ++ hl->dl_readers++; ++ } ++ hl->dl_pid = current->pid; ++ spin_unlock(&dl->dl_list_lock); ++ ++ return hl; ++} ++ ++ ++/* ++ * dynlock_unlock ++ * ++ * user have to specify lockspace (dl) and pointer to lock structure ++ * returned by dynlock_lock() ++ * ++ */ ++void dynlock_unlock(struct dynlock *dl, void *lock) ++{ ++ struct dynlock_member *hl = lock; ++ int wakeup = 0; ++ ++ spin_lock(&dl->dl_list_lock); ++ if (hl->dl_writers) { ++ hl->dl_writers--; ++ if (hl->dl_writers == 0) ++ wakeup = 1; ++ } else { ++ hl->dl_readers--; ++ if (hl->dl_readers == 0) ++ wakeup = 1; ++ } ++ if (wakeup) { ++ hl->dl_pid = 0; ++ wake_up(&hl->dl_wait); ++ } ++ if (--(hl->dl_refcount) == 0) ++ list_del(&hl->dl_list); ++ spin_unlock(&dl->dl_list_lock); ++ if (hl->dl_refcount == 0) ++ kfree(hl); ++} ++ ++EXPORT_SYMBOL(dynlock_init); ++EXPORT_SYMBOL(dynlock_lock); ++EXPORT_SYMBOL(dynlock_unlock); ++ +Index: linux-2.4.21-4.EL/lib/Makefile +=================================================================== +--- linux-2.4.21-4.EL.orig/lib/Makefile 2003-10-04 05:28:48.000000000 +0800 ++++ linux-2.4.21-4.EL/lib/Makefile 2004-06-25 15:50:32.000000000 +0800 +@@ -9,10 +9,10 @@ + L_TARGET := lib.a + + export-objs := cmdline.o dec_and_lock.o rwsem-spinlock.o rwsem.o \ +- rbtree.o crc32.o qsort.o ++ rbtree.o crc32.o qsort.o dynlocks.o + + obj-y := errno.o ctype.o string.o vsprintf.o brlock.o cmdline.o \ +- bust_spinlocks.o rbtree.o dump_stack.o ++ bust_spinlocks.o rbtree.o dump_stack.o dynlocks.o + + obj-$(CONFIG_RWSEM_GENERIC_SPINLOCK) += rwsem-spinlock.o + obj-$(CONFIG_RWSEM_XCHGADD_ALGORITHM) += rwsem.o diff --git a/lustre/kernel_patches/patches/exports_2.4.21-rh-as.patch b/lustre/kernel_patches/patches/exports_2.4.21-rh-as.patch new file mode 100644 index 0000000..34a798c --- /dev/null +++ b/lustre/kernel_patches/patches/exports_2.4.21-rh-as.patch @@ -0,0 +1,59 @@ + fs/ext3/Makefile | 2 ++ + fs/ext3/super.c | 2 +- + include/linux/fs.h | 1 + + kernel/ksyms.c | 4 ++++ + 4 files changed, 8 insertions(+), 1 deletion(-) + +Index: linux-2.4.21-4.EL/fs/ext3/Makefile +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/Makefile 2003-10-04 05:28:47.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/Makefile 2004-06-24 15:11:40.000000000 +0800 +@@ -9,6 +9,8 @@ + + O_TARGET := ext3.o + ++export-objs := super.o inode.o ++ + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) +Index: linux-2.4.21-4.EL/fs/ext3/super.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/super.c 2003-10-04 05:28:47.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/super.c 2004-06-24 15:11:40.000000000 +0800 +@@ -1912,7 +1912,7 @@ + unregister_filesystem(&ext3_fs_type); + } + +-EXPORT_NO_SYMBOLS; ++EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +Index: linux-2.4.21-4.EL/include/linux/fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/fs.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/fs.h 2004-06-24 15:11:40.000000000 +0800 +@@ -1066,6 +1066,7 @@ + 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); + + #define kern_umount mntput + +Index: linux-2.4.21-4.EL/kernel/ksyms.c +=================================================================== +--- linux-2.4.21-4.EL.orig/kernel/ksyms.c 2003-10-04 05:28:52.000000000 +0800 ++++ linux-2.4.21-4.EL/kernel/ksyms.c 2004-06-24 15:11:40.000000000 +0800 +@@ -328,6 +328,10 @@ + EXPORT_SYMBOL(dcache_readdir); + EXPORT_SYMBOL(dcache_dir_ops); + ++/* lustre */ ++EXPORT_SYMBOL(pagecache_lock_cacheline); ++EXPORT_SYMBOL(do_kern_mount); ++ + /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */ + EXPORT_SYMBOL(default_llseek); + EXPORT_SYMBOL(dentry_open); diff --git a/lustre/kernel_patches/patches/ext3-ea-in-inode-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/ext3-ea-in-inode-2.4.21-rh-as.patch new file mode 100644 index 0000000..92d9de2 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-ea-in-inode-2.4.21-rh-as.patch @@ -0,0 +1,758 @@ + fs/ext3/ialloc.c | 6 + fs/ext3/inode.c | 12 + fs/ext3/super.c | 6 + fs/ext3/xattr.c | 597 +++++++++++++++++++++++++++++++++++++++++++++- + include/linux/ext3_fs.h | 2 + include/linux/ext3_fs_i.h | 3 + 6 files changed, 615 insertions(+), 11 deletions(-) + +Index: linux-2.4.21-4.EL/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/ialloc.c 2004-06-24 02:40:43.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/ialloc.c 2004-06-24 02:42:52.000000000 +0800 +@@ -577,6 +577,12 @@ + insert_inode_hash(inode); + inode->i_generation = sb->u.ext3_sb.s_next_generation++; + ++ if (EXT3_INODE_SIZE(inode->i_sb) > EXT3_GOOD_OLD_INODE_SIZE) { ++ EXT3_I(inode)->i_extra_isize = sizeof(__u16) /* i_extra_isize */ ++ + sizeof(__u16); /* i_pad1 */ ++ } else ++ EXT3_I(inode)->i_extra_isize = 0; ++ + inode->u.ext3_i.i_state = EXT3_STATE_NEW; + err = ext3_get_inode_loc_new(inode, &iloc, 1); + if (err) goto fail; +Index: linux-2.4.21-4.EL/fs/ext3/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/inode.c 2004-06-24 02:40:48.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/inode.c 2004-06-24 02:41:02.000000000 +0800 +@@ -2386,6 +2386,12 @@ + + brelse (iloc.bh); + ++ if (EXT3_INODE_SIZE(inode->i_sb) > EXT3_GOOD_OLD_INODE_SIZE) ++ EXT3_I(inode)->i_extra_isize = ++ le16_to_cpu(raw_inode->i_extra_isize); ++ else ++ EXT3_I(inode)->i_extra_isize = 0; ++ + if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; +@@ -2446,6 +2452,8 @@ + if (err) + goto out_brelse; + } ++ if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) ++ memset(raw_inode, 0, EXT3_INODE_SIZE(inode->i_sb)); + raw_inode->i_mode = cpu_to_le16(inode->i_mode); + if(!(test_opt(inode->i_sb, NO_UID32))) { + raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid)); +@@ -2529,6 +2537,10 @@ + else for (block = 0; block < EXT3_N_BLOCKS; block++) + raw_inode->i_block[block] = inode->u.ext3_i.i_data[block]; + ++ if (EXT3_INODE_SIZE(inode->i_sb) > EXT3_GOOD_OLD_INODE_SIZE) ++ raw_inode->i_extra_isize = ++ cpu_to_le16(EXT3_I(inode)->i_extra_isize); ++ + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); + rc = ext3_journal_dirty_metadata(handle, bh); + if (!err) +Index: linux-2.4.21-4.EL/fs/ext3/xattr.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/xattr.c 2004-06-24 02:40:39.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/xattr.c 2004-06-24 02:41:02.000000000 +0800 +@@ -88,6 +88,9 @@ + struct buffer_head *, + struct ext3_xattr_header *); + ++int ext3_xattr_block_set(handle_t *, struct inode *, int, const char *, ++ const void *, size_t, int); ++ + #ifdef CONFIG_EXT3_FS_XATTR_SHARING + + static int ext3_xattr_cache_insert(struct buffer_head *); +@@ -256,17 +259,12 @@ + } + + /* +- * ext3_xattr_get() +- * +- * Copy an extended attribute into the buffer +- * provided, or compute the buffer size required. +- * Buffer is NULL to compute the size of the buffer required. ++ * ext3_xattr_block_get() + * +- * Returns a negative error number on failure, or the number of bytes +- * used / required on success. ++ * routine looks for attribute in EA block and returns it's value and size + */ + int +-ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ext3_xattr_block_get(struct inode *inode, int name_index, const char *name, + void *buffer, size_t buffer_size) + { + struct buffer_head *bh = NULL; +@@ -359,6 +357,94 @@ + } + + /* ++ * ext3_xattr_ibode_get() ++ * ++ * routine looks for attribute in inode body and returns it's value and size ++ */ ++int ++ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ int size, name_len = strlen(name), storage_size; ++ struct ext3_xattr_entry *last; ++ struct ext3_inode *raw_inode; ++ struct ext3_iloc iloc; ++ char *start, *end; ++ int ret = -ENOENT; ++ ++ if (EXT3_SB(inode->i_sb)->s_inode_size <= EXT3_GOOD_OLD_INODE_SIZE) ++ return -ENOENT; ++ ++ ret = ext3_get_inode_loc(inode, &iloc); ++ if (ret) ++ return ret; ++ raw_inode = iloc.raw_inode; ++ ++ storage_size = EXT3_SB(inode->i_sb)->s_inode_size - ++ EXT3_GOOD_OLD_INODE_SIZE - ++ EXT3_I(inode)->i_extra_isize - ++ sizeof(__u32); ++ start = (char *) raw_inode + EXT3_GOOD_OLD_INODE_SIZE + ++ EXT3_I(inode)->i_extra_isize; ++ if (le32_to_cpu((*(__u32*) start)) != EXT3_XATTR_MAGIC) { ++ brelse(iloc.bh); ++ return -ENOENT; ++ } ++ start += sizeof(__u32); ++ end = (char *) raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; ++ ++ last = (struct ext3_xattr_entry *) start; ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ if (le32_to_cpu(last->e_value_size) > storage_size || ++ (char *) next >= end) { ++ ext3_error(inode->i_sb, "ext3_xattr_ibody_get", ++ "inode %ld", inode->i_ino); ++ brelse(iloc.bh); ++ return -EIO; ++ } ++ if (name_index == last->e_name_index && ++ name_len == last->e_name_len && ++ !memcmp(name, last->e_name, name_len)) ++ goto found; ++ last = next; ++ } ++ ++ /* can't find EA */ ++ brelse(iloc.bh); ++ return -ENOENT; ++ ++found: ++ size = le32_to_cpu(last->e_value_size); ++ if (buffer) { ++ ret = -ERANGE; ++ if (buffer_size >= size) { ++ memcpy(buffer, start + le16_to_cpu(last->e_value_offs), ++ size); ++ ret = size; ++ } ++ } else ++ ret = size; ++ brelse(iloc.bh); ++ return ret; ++} ++ ++int ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ int err; ++ ++ /* try to find attribute in inode body */ ++ err = ext3_xattr_ibody_get(inode, name_index, name, ++ buffer, buffer_size); ++ if (err < 0) ++ /* search was unsuccessful, try to find EA in dedicated block */ ++ err = ext3_xattr_block_get(inode, name_index, name, ++ buffer, buffer_size); ++ return err; ++} ++ ++/* + * ext3_xattr_list() + * + * Copy a list of attribute names into the buffer +@@ -369,7 +455,7 @@ + * used / required on success. + */ + int +-ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size) + { + struct buffer_head *bh = NULL; + struct ext3_xattr_entry *entry; +@@ -446,6 +532,131 @@ + return error; + } + ++/* ext3_xattr_ibody_list() ++ * ++ * generate list of attributes stored in inode body ++ */ ++int ++ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct ext3_xattr_entry *last; ++ struct ext3_inode *raw_inode; ++ char *start, *end, *buf; ++ struct ext3_iloc iloc; ++ int storage_size; ++ int ret; ++ int size = 0; ++ ++ if (EXT3_SB(inode->i_sb)->s_inode_size <= EXT3_GOOD_OLD_INODE_SIZE) ++ return 0; ++ ++ ret = ext3_get_inode_loc(inode, &iloc); ++ if (ret) ++ return ret; ++ raw_inode = iloc.raw_inode; ++ ++ storage_size = EXT3_SB(inode->i_sb)->s_inode_size - ++ EXT3_GOOD_OLD_INODE_SIZE - ++ EXT3_I(inode)->i_extra_isize - ++ sizeof(__u32); ++ start = (char *) raw_inode + EXT3_GOOD_OLD_INODE_SIZE + ++ EXT3_I(inode)->i_extra_isize; ++ if (le32_to_cpu((*(__u32*) start)) != EXT3_XATTR_MAGIC) { ++ brelse(iloc.bh); ++ return 0; ++ } ++ start += sizeof(__u32); ++ end = (char *) raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; ++ ++ last = (struct ext3_xattr_entry *) start; ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ struct ext3_xattr_handler *handler; ++ if (le32_to_cpu(last->e_value_size) > storage_size || ++ (char *) next >= end) { ++ ext3_error(inode->i_sb, "ext3_xattr_ibody_list", ++ "inode %ld", inode->i_ino); ++ brelse(iloc.bh); ++ return -EIO; ++ } ++ handler = ext3_xattr_handler(last->e_name_index); ++ if (handler) ++ size += handler->list(NULL, inode, last->e_name, ++ last->e_name_len); ++ last = next; ++ } ++ ++ if (!buffer) { ++ ret = size; ++ goto cleanup; ++ } else { ++ ret = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ last = (struct ext3_xattr_entry *) start; ++ buf = buffer; ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ struct ext3_xattr_handler *handler; ++ handler = ext3_xattr_handler(last->e_name_index); ++ if (handler) ++ buf += handler->list(buf, inode, last->e_name, ++ last->e_name_len); ++ last = next; ++ } ++ ret = size; ++cleanup: ++ brelse(iloc.bh); ++ return ret; ++} ++ ++/* ++ * ext3_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ int error; ++ int size = buffer_size; ++ ++ /* get list of attributes stored in inode body */ ++ error = ext3_xattr_ibody_list(inode, buffer, buffer_size); ++ if (error < 0) { ++ /* some error occured while collecting ++ * attributes in inode body */ ++ size = 0; ++ goto cleanup; ++ } ++ size = error; ++ ++ /* get list of attributes stored in dedicated block */ ++ if (buffer) { ++ buffer_size -= error; ++ if (buffer_size <= 0) { ++ buffer = NULL; ++ buffer_size = 0; ++ } else ++ buffer += error; ++ } ++ ++ error = ext3_xattr_block_list(inode, buffer, buffer_size); ++ if (error < 0) ++ /* listing was successful, so we return len */ ++ size = 0; ++ ++cleanup: ++ return error + size; ++} ++ + /* + * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is + * not set, set it. +@@ -480,6 +691,102 @@ + */ + int + ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t value_len, ++ int flags) ++{ ++ struct ext3_xattr_entry entry; ++ int err, where = 0, found = 0, total; ++ int free1 = -1, free2 = -1; ++ int name_len; ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > inode->i_sb->s_blocksize) ++ return -ERANGE; ++ ++ /* try to find attribute in inode body */ ++ err = ext3_xattr_ibody_find(inode, name_index, name, &entry, &free1); ++ if (err == 0) { ++ /* found EA in inode */ ++ found = 1; ++ where = 0; ++ } else if (err == -ENOENT) { ++ /* there is no such attribute in inode body */ ++ /* try to find attribute in dedicated block */ ++ err = ext3_xattr_block_find(inode, name_index, name, ++ &entry, &free2); ++ if (err != 0 && err != -ENOENT) { ++ /* not found EA in block */ ++ goto finish; ++ } else if (err == 0) { ++ /* found EA in block */ ++ where = 1; ++ found = 1; ++ } ++ } else ++ goto finish; ++ ++ /* check flags: may replace? may create ? */ ++ if (found && (flags & XATTR_CREATE)) { ++ err = -EEXIST; ++ goto finish; ++ } else if (!found && (flags & XATTR_REPLACE)) { ++ err = -ENODATA; ++ goto finish; ++ } ++ ++ /* check if we have enough space to store attribute */ ++ total = EXT3_XATTR_LEN(strlen(name)) + value_len; ++ if (free1 >= 0 && total > free1 && free2 >= 0 && total > free2) { ++ /* have no enough space */ ++ err = -ENOSPC; ++ goto finish; ++ } ++ ++ /* time to remove attribute */ ++ if (found) { ++ if (where == 0) { ++ /* EA is stored in inode body */ ++ ext3_xattr_ibody_set(handle, inode, name_index, name, ++ NULL, 0, flags); ++ } else { ++ /* EA is stored in separated block */ ++ ext3_xattr_block_set(handle, inode, name_index, name, ++ NULL, 0, flags); ++ } ++ } ++ ++ /* try to store EA in inode body */ ++ err = ext3_xattr_ibody_set(handle, inode, name_index, name, ++ value, value_len, flags); ++ if (err) { ++ /* can't store EA in inode body */ ++ /* try to store in block */ ++ err = ext3_xattr_block_set(handle, inode, name_index, ++ name, value, value_len, flags); ++ } ++ ++finish: ++ return err; ++} ++ ++/* ++ * ext3_xattr_block_set() ++ * ++ * this routine add/remove/replace attribute in EA block ++ */ ++int ++ext3_xattr_block_set(handle_t *handle, struct inode *inode, int name_index, + const char *name, const void *value, size_t value_len, + int flags) + { +@@ -868,6 +1175,279 @@ + } + + /* ++ * ext3_xattr_ibody_find() ++ * ++ * search attribute and calculate free space in inode body ++ * NOTE: free space includes space our attribute hold ++ */ ++int ++ext3_xattr_ibody_find(struct inode *inode, int name_index, ++ const char *name, struct ext3_xattr_entry *rentry, int *free) ++{ ++ struct ext3_xattr_entry *last; ++ struct ext3_inode *raw_inode; ++ int name_len = strlen(name); ++ int err, storage_size; ++ struct ext3_iloc iloc; ++ char *start, *end; ++ int ret = -ENOENT; ++ ++ if (EXT3_SB(inode->i_sb)->s_inode_size <= EXT3_GOOD_OLD_INODE_SIZE) ++ return ret; ++ ++ err = ext3_get_inode_loc(inode, &iloc); ++ if (err) ++ return -EIO; ++ raw_inode = iloc.raw_inode; ++ ++ storage_size = EXT3_SB(inode->i_sb)->s_inode_size - ++ EXT3_GOOD_OLD_INODE_SIZE - ++ EXT3_I(inode)->i_extra_isize - ++ sizeof(__u32); ++ *free = storage_size - sizeof(__u32); ++ start = (char *) raw_inode + EXT3_GOOD_OLD_INODE_SIZE + ++ EXT3_I(inode)->i_extra_isize; ++ if (le32_to_cpu((*(__u32*) start)) != EXT3_XATTR_MAGIC) { ++ brelse(iloc.bh); ++ return -ENOENT; ++ } ++ start += sizeof(__u32); ++ end = (char *) raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; ++ ++ last = (struct ext3_xattr_entry *) start; ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ if (le32_to_cpu(last->e_value_size) > storage_size || ++ (char *) next >= end) { ++ ext3_error(inode->i_sb, "ext3_xattr_ibody_find", ++ "inode %ld", inode->i_ino); ++ brelse(iloc.bh); ++ return -EIO; ++ } ++ ++ if (name_index == last->e_name_index && ++ name_len == last->e_name_len && ++ !memcmp(name, last->e_name, name_len)) { ++ memcpy(rentry, last, sizeof(struct ext3_xattr_entry)); ++ ret = 0; ++ } else { ++ *free -= EXT3_XATTR_LEN(last->e_name_len); ++ *free -= le32_to_cpu(last->e_value_size); ++ } ++ last = next; ++ } ++ ++ brelse(iloc.bh); ++ return ret; ++} ++ ++/* ++ * ext3_xattr_block_find() ++ * ++ * search attribute and calculate free space in EA block (if it allocated) ++ * NOTE: free space includes space our attribute hold ++ */ ++int ++ext3_xattr_block_find(struct inode *inode, int name_index, const char *name, ++ struct ext3_xattr_entry *rentry, int *free) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ char *end; ++ int name_len, error = -ENOENT; ++ ++ if (!EXT3_I(inode)->i_file_acl) { ++ *free = inode->i_sb->s_blocksize - ++ sizeof(struct ext3_xattr_header) - ++ sizeof(__u32); ++ return -ENOENT; ++ } ++ ea_idebug(inode, "reading block %d", EXT3_I(inode)->i_file_acl); ++ bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, ++ EXT3_I(inode)->i_file_acl); ++ brelse(bh); ++ return -EIO; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ *free = bh->b_size - sizeof(__u32); ++ ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) { ++ memcpy(rentry, entry, sizeof(struct ext3_xattr_entry)); ++ error = 0; ++ } else { ++ *free -= EXT3_XATTR_LEN(entry->e_name_len); ++ *free -= le32_to_cpu(entry->e_value_size); ++ } ++ entry = next; ++ } ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_inode_set() ++ * ++ * this routine add/remove/replace attribute in inode body ++ */ ++int ++ext3_xattr_ibody_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t value_len, ++ int flags) ++{ ++ struct ext3_xattr_entry *last, *next, *here = NULL; ++ struct ext3_inode *raw_inode; ++ int name_len = strlen(name); ++ int esize = EXT3_XATTR_LEN(name_len); ++ struct buffer_head *bh; ++ int err, storage_size; ++ struct ext3_iloc iloc; ++ int free, min_offs; ++ char *start, *end; ++ ++ if (EXT3_SB(inode->i_sb)->s_inode_size <= EXT3_GOOD_OLD_INODE_SIZE) ++ return -ENOSPC; ++ ++ err = ext3_get_inode_loc(inode, &iloc); ++ if (err) ++ return err; ++ raw_inode = iloc.raw_inode; ++ bh = iloc.bh; ++ ++ storage_size = EXT3_SB(inode->i_sb)->s_inode_size - ++ EXT3_GOOD_OLD_INODE_SIZE - ++ EXT3_I(inode)->i_extra_isize - ++ sizeof(__u32); ++ start = (char *) raw_inode + EXT3_GOOD_OLD_INODE_SIZE + ++ EXT3_I(inode)->i_extra_isize; ++ if ((*(__u32*) start) != EXT3_XATTR_MAGIC) { ++ /* inode had no attributes before */ ++ *((__u32*) start) = cpu_to_le32(EXT3_XATTR_MAGIC); ++ } ++ start += sizeof(__u32); ++ end = (char *) raw_inode + EXT3_SB(inode->i_sb)->s_inode_size; ++ min_offs = storage_size; ++ free = storage_size - sizeof(__u32); ++ ++ last = (struct ext3_xattr_entry *) start; ++ while (!IS_LAST_ENTRY(last)) { ++ next = EXT3_XATTR_NEXT(last); ++ if (le32_to_cpu(last->e_value_size) > storage_size || ++ (char *) next >= end) { ++ ext3_error(inode->i_sb, "ext3_xattr_ibody_set", ++ "inode %ld", inode->i_ino); ++ brelse(bh); ++ return -EIO; ++ } ++ ++ if (last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ if (name_index == last->e_name_index && ++ name_len == last->e_name_len && ++ !memcmp(name, last->e_name, name_len)) ++ here = last; ++ else { ++ /* we calculate all but our attribute ++ * because it will be removed before changing */ ++ free -= EXT3_XATTR_LEN(last->e_name_len); ++ free -= le32_to_cpu(last->e_value_size); ++ } ++ last = next; ++ } ++ ++ if (value && (esize + value_len > free)) { ++ brelse(bh); ++ return -ENOSPC; ++ } ++ ++ err = ext3_reserve_inode_write(handle, inode, &iloc); ++ if (err) { ++ brelse(bh); ++ return err; ++ } ++ ++ if (here) { ++ /* time to remove old value */ ++ struct ext3_xattr_entry *e; ++ int size = le32_to_cpu(here->e_value_size); ++ int border = le16_to_cpu(here->e_value_offs); ++ char *src; ++ ++ /* move tail */ ++ memmove(start + min_offs + size, start + min_offs, ++ border - min_offs); ++ ++ /* recalculate offsets */ ++ e = (struct ext3_xattr_entry *) start; ++ while (!IS_LAST_ENTRY(e)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(e); ++ int offs = le16_to_cpu(e->e_value_offs); ++ if (offs < border) ++ e->e_value_offs = ++ cpu_to_le16(offs + size); ++ e = next; ++ } ++ min_offs += size; ++ ++ /* remove entry */ ++ border = EXT3_XATTR_LEN(here->e_name_len); ++ src = (char *) here + EXT3_XATTR_LEN(here->e_name_len); ++ size = (char *) last - src; ++ if ((char *) here + size > end) ++ printk("ALERT at %s:%d: 0x%p + %d > 0x%p\n", ++ __FILE__, __LINE__, here, size, end); ++ memmove(here, src, size); ++ last = (struct ext3_xattr_entry *) ((char *) last - border); ++ *((__u32 *) last) = 0; ++ } ++ ++ if (value) { ++ int offs = min_offs - value_len; ++ /* use last to create new entry */ ++ last->e_name_len = strlen(name); ++ last->e_name_index = name_index; ++ last->e_value_offs = cpu_to_le16(offs); ++ last->e_value_size = cpu_to_le32(value_len); ++ last->e_hash = last->e_value_block = 0; ++ memset(last->e_name, 0, esize); ++ memcpy(last->e_name, name, last->e_name_len); ++ if (start + offs + value_len > end) ++ printk("ALERT at %s:%d: 0x%p + %d + %d > 0x%p\n", ++ __FILE__, __LINE__, start, offs, ++ value_len, end); ++ memcpy(start + offs, value, value_len); ++ last = EXT3_XATTR_NEXT(last); ++ *((__u32 *) last) = 0; ++ } ++ ++ ext3_mark_iloc_dirty(handle, inode, &iloc); ++ brelse(bh); ++ ++ return 0; ++} ++ ++/* + * ext3_xattr_set_trans() + * + * Like ext3_xattr_set_handle, but start from an inode. This extended +Index: linux-2.4.21-4.EL/fs/ext3/super.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/super.c 2004-06-24 02:40:48.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/super.c 2004-06-24 02:41:02.000000000 +0800 +@@ -1332,7 +1332,9 @@ + } else { + sbi->s_inode_size = le16_to_cpu(es->s_inode_size); + sbi->s_first_ino = le32_to_cpu(es->s_first_ino); +- if (sbi->s_inode_size != EXT3_GOOD_OLD_INODE_SIZE) { ++ if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) || ++ (sbi->s_inode_size & (sbi->s_inode_size - 1)) || ++ (sbi->s_inode_size > blocksize)) { + printk (KERN_ERR + "EXT3-fs: unsupported inode size: %d\n", + sbi->s_inode_size); +Index: linux-2.4.21-4.EL/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs.h 2004-06-24 02:40:48.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs.h 2004-06-24 02:41:02.000000000 +0800 +@@ -264,6 +264,8 @@ + __u32 m_i_reserved2[2]; + } masix2; + } osd2; /* OS dependent 2 */ ++ __u16 i_extra_isize; ++ __u16 i_pad1; + }; + + #define i_size_high i_dir_acl +Index: linux-2.4.21-4.EL/include/linux/ext3_fs_i.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs_i.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs_i.h 2004-06-24 02:41:02.000000000 +0800 +@@ -76,6 +76,9 @@ + */ + loff_t i_disksize; + ++ /* on-disk additional length */ ++ __u16 i_extra_isize; ++ + /* + * truncate_sem is for serialising ext3_truncate() against + * ext3_getblock(). In the 2.4 ext2 design, great chunks of inode's diff --git a/lustre/kernel_patches/patches/ext3-htree-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/ext3-htree-2.4.21-rh-as.patch new file mode 100644 index 0000000..afb4bda --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-htree-2.4.21-rh-as.patch @@ -0,0 +1,2593 @@ + fs/ext3/Makefile | 2 + fs/ext3/dir.c | 302 +++++++++ + fs/ext3/file.c | 3 + fs/ext3/hash.c | 215 ++++++ + fs/ext3/namei.c | 1421 ++++++++++++++++++++++++++++++++++++++++----- + 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, 1922 insertions(+), 161 deletions(-) + +Index: linux-2.4.21-4.EL/fs/ext3/dir.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/dir.c 2001-11-10 06:25:04.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/dir.c 2004-06-23 19:02:41.000000000 +0800 +@@ -21,12 +21,16 @@ + #include + #include + #include ++#include ++#include + + static unsigned char ext3_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK + }; + + static int ext3_readdir(struct file *, void *, filldir_t); ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir); + + struct file_operations ext3_dir_operations = { + read: generic_read_dir, +@@ -35,6 +39,17 @@ + fsync: ext3_sync_file, /* BKL held */ + }; + ++ ++static unsigned char get_dtype(struct super_block *sb, int filetype) ++{ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) || ++ (filetype >= EXT3_FT_MAX)) ++ return DT_UNKNOWN; ++ ++ return (ext3_filetype_table[filetype]); ++} ++ ++ + int ext3_check_dir_entry (const char * function, struct inode * dir, + struct ext3_dir_entry_2 * de, + struct buffer_head * bh, +@@ -79,6 +94,16 @@ + + sb = inode->i_sb; + ++ if (is_dx(inode)) { ++ err = ext3_dx_readdir(filp, dirent, filldir); ++ if (err != ERR_BAD_DX_DIR) ++ return err; ++ /* ++ * We don't set the inode dirty flag since it's not ++ * critical that it get flushed back to the disk. ++ */ ++ EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; ++ } + stored = 0; + bh = NULL; + offset = filp->f_pos & (sb->s_blocksize - 1); +@@ -162,18 +187,12 @@ + * during the copy operation. + */ + unsigned long version = filp->f_version; +- unsigned char d_type = DT_UNKNOWN; + +- if (EXT3_HAS_INCOMPAT_FEATURE(sb, +- EXT3_FEATURE_INCOMPAT_FILETYPE) +- && de->file_type < EXT3_FT_MAX) +- d_type = +- ext3_filetype_table[de->file_type]; + error = filldir(dirent, de->name, + de->name_len, + filp->f_pos, + le32_to_cpu(de->inode), +- d_type); ++ get_dtype(sb, de->file_type)); + if (error) + break; + if (version != filp->f_version) +@@ -188,3 +207,272 @@ + UPDATE_ATIME(inode); + return 0; + } ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * These functions convert from the major/minor hash to an f_pos ++ * value. ++ * ++ * Currently we only use major hash numer. This is unfortunate, but ++ * on 32-bit machines, the same VFS interface is used for lseek and ++ * llseek, so if we use the 64 bit offset, then the 32-bit versions of ++ * lseek/telldir/seekdir will blow out spectacularly, and from within ++ * the ext2 low-level routine, we don't know if we're being called by ++ * a 64-bit version of the system call or the 32-bit version of the ++ * system call. Worse yet, NFSv2 only allows for a 32-bit readdir ++ * cookie. Sigh. ++ */ ++#define hash2pos(major, minor) (major >> 1) ++#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff) ++#define pos2min_hash(pos) (0) ++ ++/* ++ * This structure holds the nodes of the red-black tree used to store ++ * the directory entry in hash order. ++ */ ++struct fname { ++ __u32 hash; ++ __u32 minor_hash; ++ rb_node_t rb_hash; ++ struct fname *next; ++ __u32 inode; ++ __u8 name_len; ++ __u8 file_type; ++ char name[0]; ++}; ++ ++/* ++ * This functoin implements a non-recursive way of freeing all of the ++ * nodes in the red-black tree. ++ */ ++static void free_rb_tree_fname(rb_root_t *root) ++{ ++ rb_node_t *n = root->rb_node; ++ rb_node_t *parent; ++ struct fname *fname; ++ ++ while (n) { ++ /* Do the node's children first */ ++ if ((n)->rb_left) { ++ n = n->rb_left; ++ continue; ++ } ++ if (n->rb_right) { ++ n = n->rb_right; ++ continue; ++ } ++ /* ++ * The node has no children; free it, and then zero ++ * out parent's link to it. Finally go to the ++ * beginning of the loop and try to free the parent ++ * node. ++ */ ++ parent = n->rb_parent; ++ fname = rb_entry(n, struct fname, rb_hash); ++ kfree(fname); ++ if (!parent) ++ root->rb_node = 0; ++ else if (parent->rb_left == n) ++ parent->rb_left = 0; ++ else if (parent->rb_right == n) ++ parent->rb_right = 0; ++ n = parent; ++ } ++ root->rb_node = 0; ++} ++ ++ ++struct dir_private_info *create_dir_info(loff_t pos) ++{ ++ struct dir_private_info *p; ++ ++ p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL); ++ if (!p) ++ return NULL; ++ p->root.rb_node = 0; ++ p->curr_node = 0; ++ p->extra_fname = 0; ++ p->last_pos = 0; ++ p->curr_hash = pos2maj_hash(pos); ++ p->curr_minor_hash = pos2min_hash(pos); ++ p->next_hash = 0; ++ return p; ++} ++ ++void ext3_htree_free_dir_info(struct dir_private_info *p) ++{ ++ free_rb_tree_fname(&p->root); ++ kfree(p); ++} ++ ++/* ++ * Given a directory entry, enter it into the fname rb tree. ++ */ ++int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent) ++{ ++ rb_node_t **p, *parent = NULL; ++ struct fname * fname, *new_fn; ++ struct dir_private_info *info; ++ int len; ++ ++ info = (struct dir_private_info *) dir_file->private_data; ++ p = &info->root.rb_node; ++ ++ /* Create and allocate the fname structure */ ++ len = sizeof(struct fname) + dirent->name_len + 1; ++ new_fn = kmalloc(len, GFP_KERNEL); ++ if (!new_fn) ++ return -ENOMEM; ++ memset(new_fn, 0, len); ++ new_fn->hash = hash; ++ new_fn->minor_hash = minor_hash; ++ new_fn->inode = le32_to_cpu(dirent->inode); ++ new_fn->name_len = dirent->name_len; ++ new_fn->file_type = dirent->file_type; ++ memcpy(new_fn->name, dirent->name, dirent->name_len); ++ new_fn->name[dirent->name_len] = 0; ++ ++ while (*p) { ++ parent = *p; ++ fname = rb_entry(parent, struct fname, rb_hash); ++ ++ /* ++ * If the hash and minor hash match up, then we put ++ * them on a linked list. This rarely happens... ++ */ ++ if ((new_fn->hash == fname->hash) && ++ (new_fn->minor_hash == fname->minor_hash)) { ++ new_fn->next = fname->next; ++ fname->next = new_fn; ++ return 0; ++ } ++ ++ if (new_fn->hash < fname->hash) ++ p = &(*p)->rb_left; ++ else if (new_fn->hash > fname->hash) ++ p = &(*p)->rb_right; ++ else if (new_fn->minor_hash < fname->minor_hash) ++ p = &(*p)->rb_left; ++ else /* if (new_fn->minor_hash > fname->minor_hash) */ ++ p = &(*p)->rb_right; ++ } ++ ++ rb_link_node(&new_fn->rb_hash, parent, p); ++ rb_insert_color(&new_fn->rb_hash, &info->root); ++ return 0; ++} ++ ++ ++ ++/* ++ * This is a helper function for ext3_dx_readdir. It calls filldir ++ * for all entres on the fname linked list. (Normally there is only ++ * one entry on the linked list, unless there are 62 bit hash collisions.) ++ */ ++static int call_filldir(struct file * filp, void * dirent, ++ filldir_t filldir, struct fname *fname) ++{ ++ struct dir_private_info *info = filp->private_data; ++ loff_t curr_pos; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct super_block * sb; ++ int error; ++ ++ sb = inode->i_sb; ++ ++ if (!fname) { ++ printk("call_filldir: called with null fname?!?\n"); ++ return 0; ++ } ++ curr_pos = hash2pos(fname->hash, fname->minor_hash); ++ while (fname) { ++ error = filldir(dirent, fname->name, ++ fname->name_len, curr_pos, ++ fname->inode, ++ get_dtype(sb, fname->file_type)); ++ if (error) { ++ filp->f_pos = curr_pos; ++ info->extra_fname = fname->next; ++ return error; ++ } ++ fname = fname->next; ++ } ++ return 0; ++} ++ ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir) ++{ ++ struct dir_private_info *info = filp->private_data; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct fname *fname; ++ int ret; ++ ++ if (!info) { ++ info = create_dir_info(filp->f_pos); ++ if (!info) ++ return -ENOMEM; ++ filp->private_data = info; ++ } ++ ++ /* Some one has messed with f_pos; reset the world */ ++ if (info->last_pos != filp->f_pos) { ++ free_rb_tree_fname(&info->root); ++ info->curr_node = 0; ++ info->extra_fname = 0; ++ info->curr_hash = pos2maj_hash(filp->f_pos); ++ info->curr_minor_hash = pos2min_hash(filp->f_pos); ++ } ++ ++ /* ++ * If there are any leftover names on the hash collision ++ * chain, return them first. ++ */ ++ if (info->extra_fname && ++ call_filldir(filp, dirent, filldir, info->extra_fname)) ++ goto finished; ++ ++ if (!info->curr_node) ++ info->curr_node = rb_get_first(&info->root); ++ ++ while (1) { ++ /* ++ * Fill the rbtree if we have no more entries, ++ * or the inode has changed since we last read in the ++ * cached entries. ++ */ ++ if ((!info->curr_node) || ++ (filp->f_version != inode->i_version)) { ++ info->curr_node = 0; ++ free_rb_tree_fname(&info->root); ++ filp->f_version = inode->i_version; ++ ret = ext3_htree_fill_tree(filp, info->curr_hash, ++ info->curr_minor_hash, ++ &info->next_hash); ++ if (ret < 0) ++ return ret; ++ if (ret == 0) ++ break; ++ info->curr_node = rb_get_first(&info->root); ++ } ++ ++ fname = rb_entry(info->curr_node, struct fname, rb_hash); ++ info->curr_hash = fname->hash; ++ info->curr_minor_hash = fname->minor_hash; ++ if (call_filldir(filp, dirent, filldir, fname)) ++ break; ++ ++ info->curr_node = rb_get_next(info->curr_node); ++ if (!info->curr_node) { ++ info->curr_hash = info->next_hash; ++ info->curr_minor_hash = 0; ++ } ++ } ++finished: ++ info->last_pos = filp->f_pos; ++ UPDATE_ATIME(inode); ++ return 0; ++} ++#endif +Index: linux-2.4.21-4.EL/fs/ext3/file.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/file.c 2003-10-04 05:28:47.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/file.c 2004-06-23 19:02:41.000000000 +0800 +@@ -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-2.4.21-4.EL/fs/ext3/hash.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/hash.c 2003-01-30 18:24:37.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/hash.c 2004-06-23 19:02:41.000000000 +0800 +@@ -0,0 +1,215 @@ ++/* ++ * linux/fs/ext3/hash.c ++ * ++ * Copyright (C) 2002 by Theodore Ts'o ++ * ++ * This file is released under the GPL v2. ++ * ++ * This file may be redistributed under the terms of the GNU Public ++ * License. ++ */ ++ ++#include ++#include ++#include ++#include ++ ++#define DELTA 0x9E3779B9 ++ ++static void TEA_transform(__u32 buf[4], __u32 const in[]) ++{ ++ __u32 sum = 0; ++ __u32 b0 = buf[0], b1 = buf[1]; ++ __u32 a = in[0], b = in[1], c = in[2], d = in[3]; ++ int n = 16; ++ ++ do { ++ sum += DELTA; ++ b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b); ++ b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d); ++ } while(--n); ++ ++ buf[0] += b0; ++ buf[1] += b1; ++} ++ ++/* F, G and H are basic MD4 functions: selection, majority, parity */ ++#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z)))) ++#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z))) ++#define H(x, y, z) ((x) ^ (y) ^ (z)) ++ ++/* ++ * The generic round function. The application is so specific that ++ * we don't bother protecting all the arguments with parens, as is generally ++ * good macro practice, in favor of extra legibility. ++ * Rotation is separate from addition to prevent recomputation ++ */ ++#define ROUND(f, a, b, c, d, x, s) \ ++ (a += f(b, c, d) + x, a = (a << s) | (a >> (32-s))) ++#define K1 0 ++#define K2 013240474631UL ++#define K3 015666365641UL ++ ++/* ++ * Basic cut-down MD4 transform. Returns only 32 bits of result. ++ */ ++static void halfMD4Transform (__u32 buf[4], __u32 const in[]) ++{ ++ __u32 a = buf[0], b = buf[1], c = buf[2], d = buf[3]; ++ ++ /* Round 1 */ ++ ROUND(F, a, b, c, d, in[0] + K1, 3); ++ ROUND(F, d, a, b, c, in[1] + K1, 7); ++ ROUND(F, c, d, a, b, in[2] + K1, 11); ++ ROUND(F, b, c, d, a, in[3] + K1, 19); ++ ROUND(F, a, b, c, d, in[4] + K1, 3); ++ ROUND(F, d, a, b, c, in[5] + K1, 7); ++ ROUND(F, c, d, a, b, in[6] + K1, 11); ++ ROUND(F, b, c, d, a, in[7] + K1, 19); ++ ++ /* Round 2 */ ++ ROUND(G, a, b, c, d, in[1] + K2, 3); ++ ROUND(G, d, a, b, c, in[3] + K2, 5); ++ ROUND(G, c, d, a, b, in[5] + K2, 9); ++ ROUND(G, b, c, d, a, in[7] + K2, 13); ++ ROUND(G, a, b, c, d, in[0] + K2, 3); ++ ROUND(G, d, a, b, c, in[2] + K2, 5); ++ ROUND(G, c, d, a, b, in[4] + K2, 9); ++ ROUND(G, b, c, d, a, in[6] + K2, 13); ++ ++ /* Round 3 */ ++ ROUND(H, a, b, c, d, in[3] + K3, 3); ++ ROUND(H, d, a, b, c, in[7] + K3, 9); ++ ROUND(H, c, d, a, b, in[2] + K3, 11); ++ ROUND(H, b, c, d, a, in[6] + K3, 15); ++ ROUND(H, a, b, c, d, in[1] + K3, 3); ++ ROUND(H, d, a, b, c, in[5] + K3, 9); ++ ROUND(H, c, d, a, b, in[0] + K3, 11); ++ ROUND(H, b, c, d, a, in[4] + K3, 15); ++ ++ buf[0] += a; ++ buf[1] += b; ++ buf[2] += c; ++ buf[3] += d; ++} ++ ++#undef ROUND ++#undef F ++#undef G ++#undef H ++#undef K1 ++#undef K2 ++#undef K3 ++ ++/* The old legacy hash */ ++static __u32 dx_hack_hash (const char *name, int len) ++{ ++ __u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; ++ while (len--) { ++ __u32 hash = hash1 + (hash0 ^ (*name++ * 7152373)); ++ ++ if (hash & 0x80000000) hash -= 0x7fffffff; ++ hash1 = hash0; ++ hash0 = hash; ++ } ++ return (hash0 << 1); ++} ++ ++static void str2hashbuf(const char *msg, int len, __u32 *buf, int num) ++{ ++ __u32 pad, val; ++ int i; ++ ++ pad = (__u32)len | ((__u32)len << 8); ++ pad |= pad << 16; ++ ++ val = pad; ++ if (len > num*4) ++ len = num * 4; ++ for (i=0; i < len; i++) { ++ if ((i % 4) == 0) ++ val = pad; ++ val = msg[i] + (val << 8); ++ if ((i % 4) == 3) { ++ *buf++ = val; ++ val = pad; ++ num--; ++ } ++ } ++ if (--num >= 0) ++ *buf++ = val; ++ while (--num >= 0) ++ *buf++ = pad; ++} ++ ++/* ++ * Returns the hash of a filename. If len is 0 and name is NULL, then ++ * this function can be used to test whether or not a hash version is ++ * supported. ++ * ++ * The seed is an 4 longword (32 bits) "secret" which can be used to ++ * uniquify a hash. If the seed is all zero's, then some default seed ++ * may be used. ++ * ++ * A particular hash version specifies whether or not the seed is ++ * represented, and whether or not the returned hash is 32 bits or 64 ++ * bits. 32 bit hashes will return 0 for the minor hash. ++ */ ++int ext3fs_dirhash(const char *name, int len, struct dx_hash_info *hinfo) ++{ ++ __u32 hash; ++ __u32 minor_hash = 0; ++ const char *p; ++ int i; ++ __u32 in[8], buf[4]; ++ ++ /* Initialize the default seed for the hash checksum functions */ ++ buf[0] = 0x67452301; ++ buf[1] = 0xefcdab89; ++ buf[2] = 0x98badcfe; ++ buf[3] = 0x10325476; ++ ++ /* Check to see if the seed is all zero's */ ++ if (hinfo->seed) { ++ for (i=0; i < 4; i++) { ++ if (hinfo->seed[i]) ++ break; ++ } ++ if (i < 4) ++ memcpy(buf, hinfo->seed, sizeof(buf)); ++ } ++ ++ switch (hinfo->hash_version) { ++ case DX_HASH_LEGACY: ++ hash = dx_hack_hash(name, len); ++ break; ++ case DX_HASH_HALF_MD4: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 8); ++ halfMD4Transform(buf, in); ++ len -= 32; ++ p += 32; ++ } ++ minor_hash = buf[2]; ++ hash = buf[1]; ++ break; ++ case DX_HASH_TEA: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 4); ++ TEA_transform(buf, in); ++ len -= 16; ++ p += 16; ++ } ++ hash = buf[0]; ++ minor_hash = buf[1]; ++ break; ++ default: ++ hinfo->hash = 0; ++ return -1; ++ } ++ hinfo->hash = hash & ~1; ++ hinfo->minor_hash = minor_hash; ++ return 0; ++} +Index: linux-2.4.21-4.EL/fs/ext3/Makefile +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/Makefile 2004-06-23 18:31:07.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/Makefile 2004-06-23 19:02:41.000000000 +0800 +@@ -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) + + export-objs += xattr.o +Index: linux-2.4.21-4.EL/fs/ext3/namei.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/namei.c 2003-10-04 05:28:47.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/namei.c 2004-06-23 19:04:38.000000000 +0800 +@@ -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 +@@ -40,6 +46,642 @@ + #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) + #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + ++static struct buffer_head *ext3_append(handle_t *handle, ++ struct inode *inode, ++ u32 *block, int *err) ++{ ++ struct buffer_head *bh; ++ ++ *block = inode->i_size >> inode->i_sb->s_blocksize_bits; ++ ++ if ((bh = ext3_bread(handle, inode, *block, 1, err))) { ++ inode->i_size += inode->i_sb->s_blocksize; ++ EXT3_I(inode)->i_disksize = inode->i_size; ++ ext3_journal_get_write_access(handle,bh); ++ } ++ return bh; ++} ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#ifndef swap ++#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0) ++#endif ++ ++typedef struct { u32 v; } le_u32; ++typedef struct { u16 v; } le_u16; ++ ++#ifdef DX_DEBUG ++#define dxtrace(command) command ++#else ++#define dxtrace(command) ++#endif ++ ++struct fake_dirent ++{ ++ /*le*/u32 inode; ++ /*le*/u16 rec_len; ++ u8 name_len; ++ u8 file_type; ++}; ++ ++struct dx_countlimit ++{ ++ le_u16 limit; ++ le_u16 count; ++}; ++ ++struct dx_entry ++{ ++ le_u32 hash; ++ le_u32 block; ++}; ++ ++/* ++ * dx_root_info is laid out so that if it should somehow get overlaid by a ++ * dirent the two low bits of the hash version will be zero. Therefore, the ++ * hash version mod 4 should never be 0. Sincerely, the paranoia department. ++ */ ++ ++struct dx_root ++{ ++ struct fake_dirent dot; ++ char dot_name[4]; ++ struct fake_dirent dotdot; ++ char dotdot_name[4]; ++ struct dx_root_info ++ { ++ le_u32 reserved_zero; ++ u8 hash_version; ++ u8 info_length; /* 8 */ ++ u8 indirect_levels; ++ u8 unused_flags; ++ } ++ info; ++ struct dx_entry entries[0]; ++}; ++ ++struct dx_node ++{ ++ struct fake_dirent fake; ++ struct dx_entry entries[0]; ++}; ++ ++ ++struct dx_frame ++{ ++ struct buffer_head *bh; ++ struct dx_entry *entries; ++ struct dx_entry *at; ++}; ++ ++struct dx_map_entry ++{ ++ u32 hash; ++ u32 offs; ++}; ++ ++#ifdef CONFIG_EXT3_INDEX ++static inline unsigned dx_get_block (struct dx_entry *entry); ++static void dx_set_block (struct dx_entry *entry, unsigned value); ++static inline unsigned dx_get_hash (struct dx_entry *entry); ++static void dx_set_hash (struct dx_entry *entry, unsigned value); ++static unsigned dx_get_count (struct dx_entry *entries); ++static unsigned dx_get_limit (struct dx_entry *entries); ++static void dx_set_count (struct dx_entry *entries, unsigned value); ++static void dx_set_limit (struct dx_entry *entries, unsigned value); ++static unsigned dx_root_limit (struct inode *dir, unsigned infosize); ++static unsigned dx_node_limit (struct inode *dir); ++static struct dx_frame *dx_probe(struct dentry *dentry, ++ struct inode *dir, ++ struct dx_hash_info *hinfo, ++ struct dx_frame *frame, ++ int *err); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static struct ext3_dir_entry_2 *dx_move_dirents (char *from, char *to, ++ struct dx_map_entry *offsets, int count); ++static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash); ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err); ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode); ++ ++/* ++ * Future: use high four bits of block for coalesce-on-delete flags ++ * Mask them off for now. ++ */ ++ ++static inline unsigned dx_get_block (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->block.v) & 0x00ffffff; ++} ++ ++static inline void dx_set_block (struct dx_entry *entry, unsigned value) ++{ ++ entry->block.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_hash (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->hash.v); ++} ++ ++static inline void dx_set_hash (struct dx_entry *entry, unsigned value) ++{ ++ entry->hash.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_count (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->count.v); ++} ++ ++static inline unsigned dx_get_limit (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->limit.v); ++} ++ ++static inline void dx_set_count (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->count.v = cpu_to_le16(value); ++} ++ ++static inline void dx_set_limit (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->limit.v = cpu_to_le16(value); ++} ++ ++static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(1) - ++ EXT3_DIR_REC_LEN(2) - infosize; ++ return 0? 20: entry_space / sizeof(struct dx_entry); ++} ++ ++static inline unsigned dx_node_limit (struct inode *dir) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(0); ++ return 0? 22: entry_space / sizeof(struct dx_entry); ++} ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext3_dir_entry_2 *de, ++ int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ printk("names: "); ++ while ((char *) de < base + size) ++ { ++ if (de->inode) ++ { ++ if (show_names) ++ { ++ int len = de->name_len; ++ char *name = de->name; ++ while (len--) printk("%c", *name++); ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ printk(":%x.%u ", h.hash, ++ ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, ++ struct dx_entry *entries, int levels) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count = dx_get_count (entries), names = 0, space = 0, i; ++ unsigned bcount = 0; ++ struct buffer_head *bh; ++ int err; ++ printk("%i indexed blocks...\n", count); ++ for (i = 0; i < count; i++, entries++) ++ { ++ u32 block = dx_get_block(entries), hash = i? dx_get_hash(entries): 0; ++ u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; ++ struct stats stats; ++ printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); ++ if (!(bh = ext3_bread (NULL,dir, block, 0,&err))) continue; ++ stats = levels? ++ dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf(hinfo, (struct ext3_dir_entry_2 *) bh->b_data, blocksize, 0); ++ names += stats.names; ++ space += stats.space; ++ bcount += stats.bcount; ++ brelse (bh); ++ } ++ if (bcount) ++ printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", ++ names, space/bcount,(space/bcount)*100/blocksize); ++ return (struct stats) { names, space, bcount}; ++} ++#endif /* DX_DEBUG */ ++ ++/* ++ * Probe for a directory leaf block to search. ++ * ++ * dx_probe can return ERR_BAD_DX_DIR, which means there was a format ++ * error in the directory index, and the caller should fall back to ++ * searching the directory normally. The callers of dx_probe **MUST** ++ * check for this error code, and make sure it never gets reflected ++ * back to userspace. ++ */ ++static struct dx_frame * ++dx_probe(struct dentry *dentry, struct inode *dir, ++ struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) ++{ ++ unsigned count, indirect; ++ struct dx_entry *at, *entries, *p, *q, *m; ++ struct dx_root *root; ++ struct buffer_head *bh; ++ struct dx_frame *frame = frame_in; ++ u32 hash; ++ ++ frame->bh = NULL; ++ if (dentry) ++ dir = dentry->d_parent->d_inode; ++ if (!(bh = ext3_bread (NULL,dir, 0, 0, err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version != DX_HASH_TEA && ++ root->info.hash_version != DX_HASH_HALF_MD4 && ++ root->info.hash_version != DX_HASH_LEGACY) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unrecognised inode hash code %d", ++ root->info.hash_version); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ hinfo->hash_version = root->info.hash_version; ++ hinfo->seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ if (dentry) ++ ext3fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); ++ hash = hinfo->hash; ++ ++ if (root->info.unused_flags & 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash flags: %#06x", ++ root->info.unused_flags); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ if ((indirect = root->info.indirect_levels) > 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash depth: %#06x", ++ root->info.indirect_levels); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ entries = (struct dx_entry *) (((char *)&root->info) + ++ root->info.info_length); ++ assert(dx_get_limit(entries) == dx_root_limit(dir, ++ root->info.info_length)); ++ dxtrace (printk("Look up %x", hash)); ++ while (1) ++ { ++ count = dx_get_count(entries); ++ assert (count && count <= dx_get_limit(entries)); ++ p = entries + 1; ++ q = entries + count - 1; ++ while (p <= q) ++ { ++ m = p + (q - p)/2; ++ dxtrace(printk(".")); ++ if (dx_get_hash(m) > hash) ++ q = m - 1; ++ else ++ p = m + 1; ++ } ++ ++ if (0) // linear search cross check ++ { ++ unsigned n = count - 1; ++ at = entries; ++ while (n--) ++ { ++ dxtrace(printk(",")); ++ if (dx_get_hash(++at) > hash) ++ { ++ at--; ++ break; ++ } ++ } ++ assert (at == p - 1); ++ } ++ ++ at = p - 1; ++ dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); ++ frame->bh = bh; ++ frame->entries = entries; ++ frame->at = at; ++ if (!indirect--) return frame; ++ if (!(bh = ext3_bread (NULL,dir, dx_get_block(at), 0, err))) ++ goto fail2; ++ at = entries = ((struct dx_node *) bh->b_data)->entries; ++ assert (dx_get_limit(entries) == dx_node_limit (dir)); ++ frame++; ++ } ++fail2: ++ while (frame >= frame_in) { ++ brelse(frame->bh); ++ frame--; ++ } ++fail: ++ return NULL; ++} ++ ++static void dx_release (struct dx_frame *frames) ++{ ++ if (frames[0].bh == NULL) ++ return; ++ ++ if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) ++ brelse(frames[1].bh); ++ brelse(frames[0].bh); ++} ++ ++/* ++ * This function increments the frame pointer to search the next leaf ++ * block, and reads in the necessary intervening nodes if the search ++ * should be necessary. Whether or not the search is necessary is ++ * controlled by the hash parameter. If the hash value is even, then ++ * the search is only continued if the next block starts with that ++ * hash value. This is used if we are searching for a specific file. ++ * ++ * If the hash value is HASH_NB_ALWAYS, then always go to the next block. ++ * ++ * This function returns 1 if the caller should continue to search, ++ * or 0 if it should not. If there is an error reading one of the ++ * index blocks, it will return -1. ++ * ++ * If start_hash is non-null, it will be filled in with the starting ++ * hash of the next page. ++ */ ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash) ++{ ++ struct dx_frame *p; ++ struct buffer_head *bh; ++ int num_frames = 0; ++ __u32 bhash; ++ ++ *err = ENOENT; ++ p = frame; ++ /* ++ * Find the next leaf page by incrementing the frame pointer. ++ * If we run out of entries in the interior node, loop around and ++ * increment pointer in the parent node. When we break out of ++ * this loop, num_frames indicates the number of interior ++ * nodes need to be read. ++ */ ++ while (1) { ++ if (++(p->at) < p->entries + dx_get_count(p->entries)) ++ break; ++ if (p == frames) ++ return 0; ++ num_frames++; ++ p--; ++ } ++ ++ /* ++ * If the hash is 1, then continue only if the next page has a ++ * continuation hash of any value. This is used for readdir ++ * handling. Otherwise, check to see if the hash matches the ++ * desired contiuation hash. If it doesn't, return since ++ * there's no point to read in the successive index pages. ++ */ ++ bhash = dx_get_hash(p->at); ++ if (start_hash) ++ *start_hash = bhash; ++ if ((hash & 1) == 0) { ++ if ((bhash & ~1) != hash) ++ return 0; ++ } ++ /* ++ * If the hash is HASH_NB_ALWAYS, we always go to the next ++ * block so no check is necessary ++ */ ++ while (num_frames--) { ++ if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), ++ 0, err))) ++ return -1; /* Failure */ ++ p++; ++ brelse (p->bh); ++ p->bh = bh; ++ p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; ++ } ++ return 1; ++} ++ ++ ++/* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *p) ++{ ++ return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++/* ++ * This function fills a red-black tree with information from a ++ * directory. We start scanning the directory in hash order, starting ++ * at start_hash and start_minor_hash. ++ * ++ * This function returns the number of entries inserted into the tree, ++ * or a negative error code. ++ */ ++int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash) ++{ ++ struct dx_hash_info hinfo; ++ struct buffer_head *bh; ++ struct ext3_dir_entry_2 *de, *top; ++ static struct dx_frame frames[2], *frame; ++ struct inode *dir; ++ int block, err; ++ int count = 0; ++ int ret; ++ __u32 hashval; ++ ++ dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, ++ start_minor_hash)); ++ dir = dir_file->f_dentry->d_inode; ++ hinfo.hash = start_hash; ++ hinfo.minor_hash = 0; ++ frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ ++ /* Add '.' and '..' from the htree header */ ++ if (!start_hash && !start_minor_hash) { ++ de = (struct ext3_dir_entry_2 *) frames[0].bh->b_data; ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ de = ext3_next_entry(de); ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ count += 2; ++ } ++ ++ while (1) { ++ block = dx_get_block(frame->at); ++ dxtrace(printk("Reading block %d\n", block)); ++ if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) ++ goto errout; ++ ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + dir->i_sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) { ++ ext3fs_dirhash(de->name, de->name_len, &hinfo); ++ if ((hinfo.hash < start_hash) || ++ ((hinfo.hash == start_hash) && ++ (hinfo.minor_hash < start_minor_hash))) ++ continue; ++ if ((err = ext3_htree_store_dirent(dir_file, ++ hinfo.hash, hinfo.minor_hash, de)) != 0) ++ goto errout; ++ count++; ++ } ++ brelse (bh); ++ hashval = ~1; ++ ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ++ frame, frames, &err, &hashval); ++ if (next_hash) ++ *next_hash = hashval; ++ if (ret == -1) ++ goto errout; ++ /* ++ * Stop if: (a) there are no more entries, or ++ * (b) we have inserted at least one entry and the ++ * next hash value is not a continuation ++ */ ++ if ((ret == 0) || ++ (count && ((hashval & 1) == 0))) ++ break; ++ } ++ dx_release(frames); ++ dxtrace(printk("Fill tree: returned %d entries\n", count)); ++ return count; ++errout: ++ dx_release(frames); ++ return (err); ++} ++ ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry *map_tail) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ while ((char *) de < base + size) ++ { ++ if (de->name_len && de->inode) { ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ map_tail--; ++ map_tail->hash = h.hash; ++ map_tail->offs = (u32) ((char *) de - base); ++ count++; ++ } ++ /* XXX: do we need to check rec_len == 0 case? -Chris */ ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ return count; ++} ++ ++static void dx_sort_map (struct dx_map_entry *map, unsigned count) ++{ ++ struct dx_map_entry *p, *q, *top = map + count - 1; ++ int more; ++ /* Combsort until bubble sort doesn't suck */ ++ while (count > 2) ++ { ++ count = count*10/13; ++ if (count - 9 < 2) /* 9, 10 -> 11 */ ++ count = 11; ++ for (p = top, q = p - count; q >= map; p--, q--) ++ if (p->hash < q->hash) ++ swap(*p, *q); ++ } ++ /* Garden variety bubble sort */ ++ do { ++ more = 0; ++ q = top; ++ while (q-- > map) ++ { ++ if (q[1].hash >= q[0].hash) ++ continue; ++ swap(*(q+1), *q); ++ more = 1; ++ } ++ } while(more); ++} ++ ++static void dx_insert_block(struct dx_frame *frame, u32 hash, u32 block) ++{ ++ struct dx_entry *entries = frame->entries; ++ struct dx_entry *old = frame->at, *new = old + 1; ++ int count = dx_get_count(entries); ++ ++ assert(count < dx_get_limit(entries)); ++ assert(old < entries + count); ++ memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); ++ dx_set_hash(new, hash); ++ dx_set_block(new, block); ++ dx_set_count(entries, count + 1); ++} ++#endif ++ ++ ++static void ext3_update_dx_flag(struct inode *inode) ++{ ++ if (!EXT3_HAS_COMPAT_FEATURE(inode->i_sb, ++ EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -96,6 +738,7 @@ + return 0; + } + ++ + /* + * ext3_find_entry() + * +@@ -107,6 +750,8 @@ + * 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 +766,32 @@ + 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 +832,7 @@ + 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 +863,66 @@ + return ret; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err) ++{ ++ struct super_block * sb; ++ struct dx_hash_info hinfo; ++ u32 hash; ++ struct dx_frame frames[2], *frame; ++ struct ext3_dir_entry_2 *de, *top; ++ struct buffer_head *bh; ++ unsigned long block; ++ int retval; ++ int namelen = dentry->d_name.len; ++ const u8 *name = dentry->d_name.name; ++ struct inode *dir = dentry->d_parent->d_inode; ++ ++ sb = dir->i_sb; ++ if (!(frame = dx_probe (dentry, 0, &hinfo, frames, err))) ++ return NULL; ++ hash = hinfo.hash; ++ do { ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, err))) ++ goto errout; ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match (namelen, name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto errout; ++ } ++ *res_dir = de; ++ dx_release (frames); ++ return bh; ++ } ++ brelse (bh); ++ /* Check to see if we should continue to search */ ++ retval = ext3_htree_next_block(dir, hash, frame, ++ frames, err, 0); ++ if (retval == -1) { ++ ext3_warning(sb, __FUNCTION__, ++ "error reading index page in directory #%lu", ++ dir->i_ino); ++ goto errout; ++ } ++ } while (retval == 1); ++ ++ *err = -ENOENT; ++errout: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++} ++#endif ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -214,8 +939,9 @@ + brelse (bh); + inode = iget(dir->i_sb, ino); + +- if (!inode) ++ if (!inode) { + return ERR_PTR(-EACCES); ++ } + } + d_add(dentry, inode); + return NULL; +@@ -239,6 +965,301 @@ + 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 = cpu_to_le16(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 = cpu_to_le16(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 + le16_to_cpu(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 +1270,198 @@ + * may not sleep between calling this and putting something into + * the entry, as someone else might have used it while you slept. + */ +- +-/* +- * AKPM: the journalling code here looks wrong on the error paths +- */ + static int ext3_add_entry (handle_t *handle, struct dentry *dentry, + struct inode *inode) + { + struct inode *dir = dentry->d_parent->d_inode; +- const char *name = dentry->d_name.name; +- int namelen = dentry->d_name.len; + unsigned long offset; +- unsigned short rec_len; + struct buffer_head * bh; +- struct ext3_dir_entry_2 * de, * de1; ++ struct ext3_dir_entry_2 *de; + struct super_block * sb; + int retval; ++#ifdef CONFIG_EXT3_INDEX ++ int dx_fallback=0; ++#endif ++ unsigned blocksize; ++ unsigned nlen, rlen; ++ u32 block, blocks; + + sb = dir->i_sb; +- +- if (!namelen) ++ blocksize = sb->s_blocksize; ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &retval); ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ retval = ext3_dx_add_entry(handle, dentry, inode); ++ if (!retval || (retval != ERR_BAD_DX_DIR)) ++ return retval; ++ EXT3_I(dir)->i_flags &= ~EXT3_INDEX_FL; ++ dx_fallback++; ++ ext3_mark_inode_dirty(handle, dir); ++ } ++#endif ++ blocks = dir->i_size >> sb->s_blocksize_bits; ++ for (block = 0, offset = 0; block < blocks; block++) { ++ bh = ext3_bread(handle, dir, block, 0, &retval); ++ if(!bh) ++ return retval; ++ retval = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (retval != -ENOSPC) ++ return retval; ++ ++#ifdef CONFIG_EXT3_INDEX ++ if (blocks == 1 && !dx_fallback && ++ EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ return make_indexed_dir(handle, dentry, inode, bh); ++#endif ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); + if (!bh) + return retval; +- rec_len = EXT3_DIR_REC_LEN(namelen); +- offset = 0; + de = (struct ext3_dir_entry_2 *) bh->b_data; +- while (1) { +- if ((char *)de >= sb->s_blocksize + bh->b_data) { +- brelse (bh); +- bh = NULL; +- bh = ext3_bread (handle, dir, +- offset >> EXT3_BLOCK_SIZE_BITS(sb), 1, &retval); +- if (!bh) +- return retval; +- if (dir->i_size <= offset) { +- if (dir->i_size == 0) { +- brelse(bh); +- return -ENOENT; +- } ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = blocksize); ++ nlen = 0; ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} + +- ext3_debug ("creating next block\n"); ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * Returns 0 for success, or a negative error value ++ */ ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode) ++{ ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ struct dx_hash_info hinfo; ++ struct buffer_head * bh; ++ struct inode *dir = dentry->d_parent->d_inode; ++ struct super_block * sb = dir->i_sb; ++ struct ext3_dir_entry_2 *de; ++ int err; + +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- } else { ++ frame = dx_probe(dentry, 0, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ entries = frame->entries; ++ at = frame->at; + +- ext3_debug ("skipping to next block\n"); ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err))) ++ goto cleanup; + +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- } +- } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { +- brelse (bh); +- return -EEXIST; ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) ++ goto journal_error; ++ ++ err = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (err != -ENOSPC) { ++ bh = 0; ++ goto cleanup; ++ } ++ ++ /* Block full, should compress but for now just split */ ++ dxtrace(printk("using %u of %u node entries\n", ++ dx_get_count(entries), dx_get_limit(entries))); ++ /* Need to split index? */ ++ if (dx_get_count(entries) == dx_get_limit(entries)) { ++ u32 newblock; ++ unsigned icount = dx_get_count(entries); ++ int levels = frame - frames; ++ struct dx_entry *entries2; ++ struct dx_node *node2; ++ struct buffer_head *bh2; ++ ++ if (levels && (dx_get_count(frames->entries) == ++ dx_get_limit(frames->entries))) { ++ ext3_warning(sb, __FUNCTION__, ++ "Directory index full!\n"); ++ err = -ENOSPC; ++ goto cleanup; + } +- if ((le32_to_cpu(de->inode) == 0 && +- le16_to_cpu(de->rec_len) >= rec_len) || +- (le16_to_cpu(de->rec_len) >= +- EXT3_DIR_REC_LEN(de->name_len) + rec_len)) { +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- /* By now the buffer is marked for journaling */ +- offset += le16_to_cpu(de->rec_len); +- if (le32_to_cpu(de->inode)) { +- de1 = (struct ext3_dir_entry_2 *) ((char *) de + +- EXT3_DIR_REC_LEN(de->name_len)); +- de1->rec_len = +- cpu_to_le16(le16_to_cpu(de->rec_len) - +- EXT3_DIR_REC_LEN(de->name_len)); +- de->rec_len = cpu_to_le16( +- EXT3_DIR_REC_LEN(de->name_len)); +- de = de1; ++ bh2 = ext3_append (handle, dir, &newblock, &err); ++ if (!(bh2)) ++ goto cleanup; ++ node2 = (struct dx_node *)(bh2->b_data); ++ entries2 = node2->entries; ++ node2->fake.rec_len = cpu_to_le16(sb->s_blocksize); ++ node2->fake.inode = 0; ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ if (levels) { ++ unsigned icount1 = icount/2, icount2 = icount - icount1; ++ unsigned hash2 = dx_get_hash(entries + icount1); ++ dxtrace(printk("Split index %i/%i\n", icount1, icount2)); ++ ++ BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ ++ err = ext3_journal_get_write_access(handle, ++ frames[0].bh); ++ if (err) ++ goto journal_error; ++ ++ memcpy ((char *) entries2, (char *) (entries + icount1), ++ icount2 * sizeof(struct dx_entry)); ++ dx_set_count (entries, icount1); ++ dx_set_count (entries2, icount2); ++ dx_set_limit (entries2, dx_node_limit(dir)); ++ ++ /* Which index block gets the new entry? */ ++ if (at - entries >= icount1) { ++ frame->at = at = at - entries - icount1 + entries2; ++ frame->entries = entries = entries2; ++ swap(frame->bh, bh2); + } +- de->file_type = EXT3_FT_UNKNOWN; +- if (inode) { +- de->inode = cpu_to_le32(inode->i_ino); +- ext3_set_de_type(dir->i_sb, de, inode->i_mode); +- } else +- de->inode = 0; +- de->name_len = namelen; +- memcpy (de->name, name, namelen); +- /* +- * XXX shouldn't update any times until successful +- * completion of syscall, but too many callers depend +- * on this. +- * +- * XXX similarly, too many callers depend on +- * ext3_new_inode() setting the times, but error +- * recovery deletes the inode, so the worst that can +- * happen is that the times are slightly out of date +- * and/or different from the directory change time. +- */ +- dir->i_mtime = dir->i_ctime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- dir->i_version = ++event; +- ext3_mark_inode_dirty(handle, dir); +- BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +- ext3_journal_dirty_metadata(handle, bh); +- brelse(bh); +- return 0; ++ dx_insert_block (frames + 0, hash2, newblock); ++ dxtrace(dx_show_index ("node", frames[1].entries)); ++ dxtrace(dx_show_index ("node", ++ ((struct dx_node *) bh2->b_data)->entries)); ++ err = ext3_journal_dirty_metadata(handle, bh2); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ } else { ++ dxtrace(printk("Creating second level index...\n")); ++ memcpy((char *) entries2, (char *) entries, ++ icount * sizeof(struct dx_entry)); ++ dx_set_limit(entries2, dx_node_limit(dir)); ++ ++ /* Set up root */ ++ dx_set_count(entries, 1); ++ dx_set_block(entries + 0, newblock); ++ ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; ++ ++ /* Add new access path frame */ ++ frame = frames + 1; ++ frame->at = at = at - entries + entries2; ++ frame->entries = entries = entries2; ++ frame->bh = bh2; ++ err = ext3_journal_get_write_access(handle, ++ frame->bh); ++ if (err) ++ goto journal_error; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- brelse (bh); +- return -ENOSPC; ++ de = do_split(handle, dir, &bh, frame, &hinfo, &err); ++ if (!de) ++ goto cleanup; ++ err = add_dirent_to_buf(handle, dentry, inode, de, bh); ++ bh = 0; ++ goto cleanup; ++ ++journal_error: ++ ext3_std_error(dir->i_sb, err); ++cleanup: ++ if (bh) ++ brelse(bh); ++ dx_release(frames); ++ return err; + } ++#endif + + /* + * ext3_delete_entry deletes a directory entry by merging it with the +@@ -456,9 +1548,11 @@ + 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; +@@ -482,9 +1576,11 @@ + 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; +@@ -513,9 +1609,11 @@ + 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; +@@ -527,7 +1625,7 @@ + + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; +- inode->i_size = inode->u.ext3_i.i_disksize = inode->i_sb->s_blocksize; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { + inode->i_nlink--; /* is this nlink == 0? */ +@@ -556,21 +1654,19 @@ + 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; + } + + /* +@@ -657,7 +1753,7 @@ + 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 +@@ -698,7 +1794,7 @@ + * 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", +@@ -716,25 +1812,26 @@ + 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; + unsigned long 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 %lu 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 +@@ -795,8 +1892,9 @@ + 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); +@@ -834,7 +1932,7 @@ + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; + ext3_mark_inode_dirty(handle, inode); +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -852,8 +1950,9 @@ + 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; +@@ -880,7 +1979,7 @@ + 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) +@@ -906,9 +2005,11 @@ + 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; +@@ -918,7 +2019,7 @@ + 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; + /* +@@ -927,24 +2028,22 @@ + * i_size in generic_commit_write(). + */ + err = block_symlink(inode, symname, l); +- if (err) +- goto out_no_entry; ++ if (err) { ++ ext3_dec_count(handle, inode); ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + } else { + inode->i_op = &ext3_fast_symlink_inode_operations; +- memcpy((char*)&inode->u.ext3_i.i_data,symname,l); ++ memcpy((char*)&EXT3_I(inode)->i_data,symname,l); + inode->i_size = l-1; + } +- inode->u.ext3_i.i_disksize = inode->i_size; ++ EXT3_I(inode)->i_disksize = inode->i_size; + err = ext3_add_nondir(handle, dentry, inode); + 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 +2056,15 @@ + if (S_ISDIR(inode->i_mode)) + return -EPERM; + +- if (inode->i_nlink >= EXT3_LINK_MAX) ++ if (inode->i_nlink >= EXT3_LINK_MAX) { + return -EMLINK; ++ } + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -995,9 +2097,11 @@ + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 2); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(old_dir) || IS_SYNC(new_dir)) + handle->h_sync = 1; +@@ -1070,14 +2174,37 @@ + /* + * ok, that's it + */ +- ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ if (le32_to_cpu(old_de->inode) != old_inode->i_ino || ++ old_de->name_len != old_dentry->d_name.len || ++ strncmp(old_de->name, old_dentry->d_name.name, old_de->name_len) || ++ (retval = ext3_delete_entry(handle, old_dir, ++ old_de, old_bh)) == -ENOENT) { ++ /* old_de could have moved from under us during htree split, so ++ * make sure that we are deleting the right entry. We might ++ * also be pointing to a stale entry in the unused part of ++ * old_bh so just checking inum and the name isn't enough. */ ++ struct buffer_head *old_bh2; ++ struct ext3_dir_entry_2 *old_de2; ++ ++ old_bh2 = ext3_find_entry(old_dentry, &old_de2); ++ if (old_bh2) { ++ retval = ext3_delete_entry(handle, old_dir, ++ old_de2, old_bh2); ++ brelse(old_bh2); ++ } ++ } ++ if (retval) { ++ ext3_warning(old_dir->i_sb, "ext3_rename", ++ "Deleting old file (%lu), %d, error=%d", ++ old_dir->i_ino, old_dir->i_nlink, retval); ++ } + + if (new_inode) { + new_inode->i_nlink--; + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +2216,7 @@ + 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-2.4.21-4.EL/fs/ext3/super.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/super.c 2004-06-23 18:31:07.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/super.c 2004-06-23 19:02:41.000000000 +0800 +@@ -751,6 +751,7 @@ + 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 +@@ -761,6 +762,7 @@ + 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) { +@@ -934,6 +936,7 @@ + return res; + } + ++ + struct super_block * ext3_read_super (struct super_block * sb, void * data, + int silent) + { +@@ -1120,6 +1123,9 @@ + 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 +@@ -1912,6 +1918,7 @@ + unregister_filesystem(&ext3_fs_type); + } + ++EXPORT_SYMBOL(ext3_force_commit); + EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); +Index: linux-2.4.21-4.EL/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs.h 2004-06-23 19:02:41.000000000 +0800 +@@ -40,6 +40,11 @@ + #define EXT3FS_VERSION "2.4-0.9.19" + + /* ++ * Always enable hashed directories ++ */ ++#define CONFIG_EXT3_INDEX ++ ++/* + * Debug code + */ + #ifdef EXT3FS_DEBUG +@@ -415,8 +420,11 @@ + /*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__ +@@ -553,9 +561,46 @@ + #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 +@@ -565,6 +610,27 @@ + 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 + */ +@@ -592,11 +658,20 @@ + + /* dir.c */ + extern int ext3_check_dir_entry(const char *, struct inode *, +- struct ext3_dir_entry_2 *, struct buffer_head *, +- unsigned long); ++ struct ext3_dir_entry_2 *, ++ struct buffer_head *, unsigned long); ++extern int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent); ++extern void ext3_htree_free_dir_info(struct dir_private_info *p); ++ + /* fsync.c */ + extern int ext3_sync_file (struct file *, struct dentry *, int); + ++/* hash.c */ ++extern int ext3fs_dirhash(const char *name, int len, struct ++ dx_hash_info *hinfo); ++ + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); +@@ -630,6 +705,8 @@ + /* 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-2.4.21-4.EL/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs_sb.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs_sb.h 2004-06-23 19:02:41.000000000 +0800 +@@ -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-2.4.21-4.EL/include/linux/ext3_jbd.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_jbd.h 2003-10-04 05:29:48.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_jbd.h 2004-06-23 19:02:41.000000000 +0800 +@@ -69,6 +69,8 @@ + + #define EXT3_RESERVE_TRANS_BLOCKS 12U + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, +Index: linux-2.4.21-4.EL/include/linux/rbtree.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/rbtree.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/rbtree.h 2004-06-23 19:02:41.000000000 +0800 +@@ -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-2.4.21-4.EL/lib/rbtree.c +=================================================================== +--- linux-2.4.21-4.EL.orig/lib/rbtree.c 2002-08-03 08:39:46.000000000 +0800 ++++ linux-2.4.21-4.EL/lib/rbtree.c 2004-06-23 19:02:41.000000000 +0800 +@@ -17,6 +17,8 @@ + Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + + linux/lib/rbtree.c ++ ++ rb_get_first and rb_get_next written by Theodore Ts'o, 9/8/2002 + */ + + #include +@@ -294,3 +296,43 @@ + __rb_erase_color(child, parent, root); + } + EXPORT_SYMBOL(rb_erase); ++ ++/* ++ * This function returns the first node (in sort order) of the tree. ++ */ ++rb_node_t *rb_get_first(rb_root_t *root) ++{ ++ rb_node_t *n; ++ ++ n = root->rb_node; ++ if (!n) ++ return 0; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++} ++EXPORT_SYMBOL(rb_get_first); ++ ++/* ++ * Given a node, this function will return the next node in the tree. ++ */ ++rb_node_t *rb_get_next(rb_node_t *n) ++{ ++ rb_node_t *parent; ++ ++ if (n->rb_right) { ++ n = n->rb_right; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++ } else { ++ while ((parent = n->rb_parent)) { ++ if (n == parent->rb_left) ++ return parent; ++ n = parent; ++ } ++ return 0; ++ } ++} ++EXPORT_SYMBOL(rb_get_next); ++ diff --git a/lustre/kernel_patches/patches/ext3-noread-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/ext3-noread-2.4.21-rh-as.patch new file mode 100644 index 0000000..271c5c8 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-noread-2.4.21-rh-as.patch @@ -0,0 +1,219 @@ + fs/ext3/ialloc.c | 47 ++++++++++++++++++++++- + fs/ext3/inode.c | 96 +++++++++++++++++++++++++++++++++++++----------- + include/linux/ext3_fs.h | 2 + + 3 files changed, 121 insertions(+), 24 deletions(-) + +Index: linux-2.4.21-4.EL/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/ialloc.c 2003-10-04 05:28:49.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/ialloc.c 2004-06-23 21:31:07.000000000 +0800 +@@ -290,6 +290,37 @@ + } + + /* ++ * @block_group: block group of inode ++ * @offset: relative offset of inode within @block_group ++ * ++ * Check whether any of the inodes in this disk block are in use. ++ * ++ * Caller must be holding superblock lock (group/bitmap read lock in future). ++ */ ++int ext3_itable_block_used(struct super_block *sb, unsigned int block_group, ++ int offset) ++{ ++ int bitmap_nr = load_inode_bitmap(sb, block_group); ++ int inodes_per_block; ++ unsigned long inum, iend; ++ struct buffer_head *ibitmap; ++ ++ if (bitmap_nr < 0) ++ return 1; ++ ++ inodes_per_block = sb->s_blocksize / EXT3_SB(sb)->s_inode_size; ++ inum = offset & ~(inodes_per_block - 1); ++ iend = inum + inodes_per_block; ++ ibitmap = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; ++ for (; inum < iend; inum++) { ++ if (inum != offset && ext3_test_bit(inum, ibitmap->b_data)) ++ return 1; ++ } ++ ++ return 0; ++} ++ ++/* + * There are two policies for allocating an inode. If the new inode is + * a directory, then a forward search is made for a block group with both + * free space and a low directory-to-inode ratio; if that fails, then of +@@ -310,6 +341,7 @@ + struct ext3_group_desc * gdp; + struct ext3_group_desc * tmp; + struct ext3_super_block * es; ++ struct ext3_iloc iloc; + int err = 0; + + /* Cannot create files in a deleted directory */ +@@ -509,9 +541,19 @@ + inode->i_generation = sb->u.ext3_sb.s_next_generation++; + + inode->u.ext3_i.i_state = EXT3_STATE_NEW; +- err = ext3_mark_inode_dirty(handle, inode); ++ err = ext3_get_inode_loc_new(inode, &iloc, 1); ++ if (err) goto fail; ++ BUFFER_TRACE(iloc->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, iloc.bh); ++ if (err) { ++ brelse(iloc.bh); ++ iloc.bh = NULL; ++ goto fail; ++ } ++ err = ext3_mark_iloc_dirty(handle, inode, &iloc); + if (err) goto fail; + ++ + #ifdef CONFIG_EXT3_FS_XATTR + init_rwsem(&inode->u.ext3_i.xattr_sem); + #endif +Index: linux-2.4.21-4.EL/fs/ext3/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/inode.c 2003-10-04 05:28:53.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/inode.c 2004-06-23 21:20:31.000000000 +0800 +@@ -2172,14 +2172,19 @@ + return; /* AKPM: return what? */ + } + +-/* +- * ext3_get_inode_loc returns with an extra refcount against the +- * inode's underlying buffer_head on success. +- */ ++#define NUM_INODE_PREREAD 16 + +-int ext3_get_inode_loc (struct inode *inode, struct ext3_iloc *iloc) ++/* ++ * ext3_get_inode_loc returns with an extra refcount against the inode's ++ * underlying buffer_head on success. If this is for a new inode allocation ++ * (new is non-zero) then we may be able to optimize away the read if there ++ * are no other in-use inodes in this inode table block. If we need to do ++ * a read, then read in a whole chunk of blocks to avoid blocking again soon ++ * if we are doing lots of creates/updates. ++ */ ++int ext3_get_inode_loc_new(struct inode *inode, struct ext3_iloc *iloc, int new) + { +- struct buffer_head *bh = 0; ++ struct buffer_head *bh[NUM_INODE_PREREAD]; + unsigned long block; + unsigned long block_group; + unsigned long group_desc; +@@ -2204,31 +2209,73 @@ + } + group_desc = block_group >> EXT3_DESC_PER_BLOCK_BITS(inode->i_sb); + desc = block_group & (EXT3_DESC_PER_BLOCK(inode->i_sb) - 1); +- bh = inode->i_sb->u.ext3_sb.s_group_desc[group_desc]; +- if (!bh) { ++ if (!(inode->i_sb->u.ext3_sb.s_group_desc[group_desc])) { + ext3_error (inode->i_sb, "ext3_get_inode_loc", + "Descriptor not loaded"); + goto bad_inode; + } + +- gdp = (struct ext3_group_desc *) bh->b_data; ++ gdp = (struct ext3_group_desc *)(inode->i_sb->u.ext3_sb.s_group_desc[group_desc]->b_data); + /* + * Figure out the offset within the block group inode table + */ +- offset = ((inode->i_ino - 1) % EXT3_INODES_PER_GROUP(inode->i_sb)) * +- EXT3_INODE_SIZE(inode->i_sb); ++ offset = ((inode->i_ino - 1) % EXT3_INODES_PER_GROUP(inode->i_sb)); ++ + block = le32_to_cpu(gdp[desc].bg_inode_table) + +- (offset >> EXT3_BLOCK_SIZE_BITS(inode->i_sb)); +- if (!(bh = sb_bread(inode->i_sb, block))) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "unable to read inode block - " +- "inode=%lu, block=%lu", inode->i_ino, block); +- goto bad_inode; ++ (offset * EXT3_INODE_SIZE(inode->i_sb) >> EXT3_BLOCK_SIZE_BITS(inode->i_sb)); ++ ++ bh[0] = sb_getblk(inode->i_sb, block); ++ if (buffer_uptodate(bh[0])) ++ goto done; ++ ++ /* If we don't really need to read this block, and it isn't already ++ * in memory, then we just zero it out. Otherwise, we keep the ++ * current block contents (deleted inode data) for posterity. ++ */ ++ if (new && !ext3_itable_block_used(inode->i_sb, block_group, offset)) { ++ lock_buffer(bh[0]); ++ memset(bh[0]->b_data, 0, bh[0]->b_size); ++ mark_buffer_uptodate(bh[0], 1); ++ unlock_buffer(bh[0]); ++ } else { ++ unsigned long block_end, itable_end; ++ int count = 1; ++ ++ itable_end = le32_to_cpu(gdp[desc].bg_inode_table) + ++ inode->i_sb->u.ext3_sb.s_itb_per_group; ++ block_end = block + NUM_INODE_PREREAD; ++ if (block_end > itable_end) ++ block_end = itable_end; ++ ++ for (++block; block < block_end; block++) { ++ bh[count] = sb_getblk(inode->i_sb, block); ++ if (count && (buffer_uptodate(bh[count]) || ++ buffer_locked(bh[count]))) { ++ __brelse(bh[count]); ++ } else ++ count++; ++ } ++ ++ ll_rw_block(READ, count, bh); ++ ++ /* Release all but the block we actually need (bh[0]) */ ++ while (--count > 0) ++ __brelse(bh[count]); ++ ++ wait_on_buffer(bh[0]); ++ if (!buffer_uptodate(bh[0])) { ++ ext3_error(inode->i_sb, __FUNCTION__, ++ "unable to read inode block - " ++ "inode=%lu, block=%lu", inode->i_ino, ++ bh[0]->b_blocknr); ++ goto bad_inode; ++ } + } +- offset &= (EXT3_BLOCK_SIZE(inode->i_sb) - 1); ++done: ++ offset = (offset * EXT3_INODE_SIZE(inode->i_sb)) & (EXT3_BLOCK_SIZE(inode->i_sb) - 1); + +- iloc->bh = bh; +- iloc->raw_inode = (struct ext3_inode *) (bh->b_data + offset); ++ iloc->bh = bh[0]; ++ iloc->raw_inode = (struct ext3_inode *)(bh[0]->b_data + offset); + iloc->block_group = block_group; + + return 0; +@@ -2253,6 +2300,11 @@ + } + + ++int ext3_get_inode_loc(struct inode *inode, struct ext3_iloc *iloc) ++{ ++ return ext3_get_inode_loc_new(inode, iloc, 0); ++} ++ + void ext3_read_inode(struct inode * inode) + { + struct ext3_iloc iloc; +Index: linux-2.4.21-4.EL/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs.h 2004-06-23 19:09:01.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs.h 2004-06-23 21:20:31.000000000 +0800 +@@ -685,6 +685,8 @@ + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + ++extern int ext3_itable_block_used(struct super_block *sb, unsigned int, int); ++extern int ext3_get_inode_loc_new(struct inode *, struct ext3_iloc *, int); + extern int ext3_get_inode_loc (struct inode *, struct ext3_iloc *); + extern void ext3_read_inode (struct inode *); + extern void ext3_write_inode (struct inode *, int); diff --git a/lustre/kernel_patches/patches/ext3_delete_thread_2.4.21_rh_as.patch b/lustre/kernel_patches/patches/ext3_delete_thread_2.4.21_rh_as.patch new file mode 100644 index 0000000..66b6cfc --- /dev/null +++ b/lustre/kernel_patches/patches/ext3_delete_thread_2.4.21_rh_as.patch @@ -0,0 +1,500 @@ + fs/ext3/file.c | 4 + fs/ext3/inode.c | 116 ++++++++++++++++++++++ + fs/ext3/super.c | 230 +++++++++++++++++++++++++++++++++++++++++++++ + include/linux/ext3_fs.h | 5 + include/linux/ext3_fs_sb.h | 10 + + 5 files changed, 365 insertions(+) + +Index: linux-2.4.21-4.EL/fs/ext3/super.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/super.c 2004-06-23 21:10:01.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/super.c 2004-06-23 21:34:21.000000000 +0800 +@@ -401,6 +401,221 @@ + } + } + ++#ifdef EXT3_DELETE_THREAD ++/* ++ * Delete inodes in a loop until there are no more to be deleted. ++ * Normally, we run in the background doing the deletes and sleeping again, ++ * and clients just add new inodes to be deleted onto the end of the list. ++ * If someone is concerned about free space (e.g. block allocation or similar) ++ * then they can sleep on s_delete_waiter_queue and be woken up when space ++ * has been freed. ++ */ ++int ext3_delete_thread(void *data) ++{ ++ struct super_block *sb = data; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ struct task_struct *tsk = current; ++ ++ /* Almost like daemonize, but not quite */ ++ exit_mm(current); ++ tsk->session = 1; ++ tsk->pgrp = 1; ++ tsk->tty = NULL; ++ exit_files(current); ++ reparent_to_init(); ++ ++ sprintf(tsk->comm, "kdelext3-%s", kdevname(sb->s_dev)); ++ sigfillset(&tsk->blocked); ++ ++ /*tsk->flags |= PF_KERNTHREAD;*/ ++ ++ INIT_LIST_HEAD(&sbi->s_delete_list); ++ wake_up(&sbi->s_delete_waiter_queue); ++ ext3_debug("delete thread on %s started\n", kdevname(sb->s_dev)); ++ ++ /* main loop */ ++ for (;;) { ++ wait_event_interruptible(sbi->s_delete_thread_queue, ++ !list_empty(&sbi->s_delete_list) || ++ !test_opt(sb, ASYNCDEL)); ++ ext3_debug("%s woken up: %lu inodes, %lu blocks\n", ++ tsk->comm,sbi->s_delete_inodes,sbi->s_delete_blocks); ++ ++ spin_lock(&sbi->s_delete_lock); ++ if (list_empty(&sbi->s_delete_list)) { ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ memset(&sbi->s_delete_list, 0, ++ sizeof(sbi->s_delete_list)); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("delete thread on %s exiting\n", ++ kdevname(sb->s_dev)); ++ wake_up(&sbi->s_delete_waiter_queue); ++ break; ++ } ++ ++ while (!list_empty(&sbi->s_delete_list)) { ++ struct inode *inode=list_entry(sbi->s_delete_list.next, ++ struct inode, i_dentry); ++ unsigned long blocks = inode->i_blocks >> ++ (inode->i_blkbits - 9); ++ ++ list_del_init(&inode->i_dentry); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("%s delete ino %lu blk %lu\n", ++ tsk->comm, inode->i_ino, blocks); ++ ++ iput(inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ sbi->s_delete_blocks -= blocks; ++ sbi->s_delete_inodes--; ++ } ++ if (sbi->s_delete_blocks != 0 || sbi->s_delete_inodes != 0) { ++ ext3_warning(sb, __FUNCTION__, ++ "%lu blocks, %lu inodes on list?\n", ++ sbi->s_delete_blocks,sbi->s_delete_inodes); ++ sbi->s_delete_blocks = 0; ++ sbi->s_delete_inodes = 0; ++ } ++ spin_unlock(&sbi->s_delete_lock); ++ wake_up(&sbi->s_delete_waiter_queue); ++ } ++ ++ return 0; ++} ++ ++static void ext3_start_delete_thread(struct super_block *sb) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ int rc; ++ ++ spin_lock_init(&sbi->s_delete_lock); ++ init_waitqueue_head(&sbi->s_delete_thread_queue); ++ init_waitqueue_head(&sbi->s_delete_waiter_queue); ++ ++ if (!test_opt(sb, ASYNCDEL)) ++ return; ++ ++ rc = kernel_thread(ext3_delete_thread, sb, CLONE_VM | CLONE_FILES); ++ if (rc < 0) ++ printk(KERN_ERR "EXT3-fs: cannot start delete thread: rc %d\n", ++ rc); ++ else ++ wait_event(sbi->s_delete_waiter_queue, sbi->s_delete_list.next); ++} ++ ++static void ext3_stop_delete_thread(struct ext3_sb_info *sbi) ++{ ++ if (sbi->s_delete_list.next == 0) /* thread never started */ ++ return; ++ ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ wake_up(&sbi->s_delete_thread_queue); ++ wait_event(sbi->s_delete_waiter_queue, ++ sbi->s_delete_list.next == 0 && sbi->s_delete_inodes == 0); ++} ++ ++/* Instead of playing games with the inode flags, destruction, etc we just ++ * create a new inode locally and put it on a list for the truncate thread. ++ * We need large parts of the inode struct in order to complete the ++ * truncate and unlink, so we may as well just have a real inode to do it. ++ * ++ * If we have any problem deferring the delete, just delete it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ */ ++static void ext3_delete_inode_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (is_bad_inode(old_inode)) { ++ clear_inode(old_inode); ++ return; ++ } ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_delete; ++ ++ /* We may want to delete the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS) ++ goto out_delete; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_delete; ++ } ++ ++ /* We can iget this inode again here, because our caller has unhashed ++ * old_inode, so new_inode will be in a different inode struct. ++ * ++ * We need to ensure that the i_orphan pointers in the other inodes ++ * point at the new inode copy instead of the old one so the orphan ++ * list doesn't get corrupted when the old orphan inode is freed. ++ */ ++ down(&sbi->s_orphan_lock); ++ ++ sbi->s_mount_state |= EXT3_ORPHAN_FS; ++ new_inode = iget(old_inode->i_sb, old_inode->i_ino); ++ sbi->s_mount_state &= ~EXT3_ORPHAN_FS; ++ if (is_bad_inode(new_inode)) { ++ printk(KERN_WARNING "read bad inode %lu\n", old_inode->i_ino); ++ iput(new_inode); ++ new_inode = NULL; ++ } ++ if (!new_inode) { ++ up(&sbi->s_orphan_lock); ++ ext3_debug("delete inode %lu directly (bad read)\n", ++ old_inode->i_ino); ++ goto out_delete; ++ } ++ J_ASSERT(new_inode != old_inode); ++ ++ J_ASSERT(!list_empty(&oei->i_orphan)); ++ ++ nei = EXT3_I(new_inode); ++ /* Ugh. We need to insert new_inode into the same spot on the list ++ * as old_inode was, to ensure the in-memory orphan list is still ++ * in the same order as the on-disk orphan list (badness otherwise). ++ */ ++ nei->i_orphan = oei->i_orphan; ++ nei->i_orphan.next->prev = &nei->i_orphan; ++ nei->i_orphan.prev->next = &nei->i_orphan; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up(&sbi->s_orphan_lock); ++ ++ clear_inode(old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_delete: ++ ext3_delete_inode(old_inode); ++} ++#else ++#define ext3_start_delete_thread(sbi) do {} while(0) ++#define ext3_stop_delete_thread(sbi) do {} while(0) ++#endif /* EXT3_DELETE_THREAD */ ++ + void ext3_put_super (struct super_block * sb) + { + struct ext3_sb_info *sbi = EXT3_SB(sb); +@@ -408,6 +623,7 @@ + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ J_ASSERT(sbi->s_delete_inodes == 0); + ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { +@@ -477,7 +693,11 @@ + write_inode: ext3_write_inode, /* BKL not held. Don't need */ + dirty_inode: ext3_dirty_inode, /* BKL not held. We take it */ + put_inode: ext3_put_inode, /* BKL not held. Don't need */ ++#ifdef EXT3_DELETE_THREAD ++ delete_inode: ext3_delete_inode_thread,/* BKL not held. We take it */ ++#else + delete_inode: ext3_delete_inode, /* BKL not held. We take it */ ++#endif + put_super: ext3_put_super, /* BKL held */ + write_super: ext3_write_super, /* BKL held */ + sync_fs: ext3_sync_fs, +@@ -555,6 +775,13 @@ + *mount_flags &= ~MS_POSIXACL; + else + #endif ++#ifdef EXT3_DELETE_THREAD ++ if (!strcmp(this_char, "asyncdel")) ++ set_opt(*mount_options, ASYNCDEL); ++ else if (!strcmp(this_char, "noasyncdel")) ++ clear_opt(*mount_options, ASYNCDEL); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -1259,6 +1486,7 @@ + } + + ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY); ++ ext3_start_delete_thread(sb); + /* + * akpm: core read_super() calls in here with the superblock locked. + * That deadlocks, because orphan cleanup needs to lock the superblock +@@ -1650,7 +1878,12 @@ + static int ext3_sync_fs(struct super_block *sb) + { + tid_t target; +- ++ ++ if (atomic_read(&sb->s_active) == 0) { ++ /* fs is being umounted: time to stop delete thread */ ++ ext3_stop_delete_thread(EXT3_SB(sb)); ++ } ++ + sb->s_dirt = 0; + target = log_start_commit(EXT3_SB(sb)->s_journal, NULL); + log_wait_commit(EXT3_SB(sb)->s_journal, target); +@@ -1720,6 +1953,9 @@ + if (!parse_options(data, &tmp, sbi, &mount_flags, &tmp, 1)) + return -EINVAL; + ++ if (!test_opt(sb, ASYNCDEL) || (*flags & MS_RDONLY)) ++ ext3_stop_delete_thread(sbi); ++ + if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) + ext3_abort(sb, __FUNCTION__, "Abort forced by user"); + +Index: linux-2.4.21-4.EL/fs/ext3/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/inode.c 2004-06-23 21:31:33.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/inode.c 2004-06-23 21:34:21.000000000 +0800 +@@ -2739,6 +2739,122 @@ + return err; + } + ++#ifdef EXT3_DELETE_THREAD ++/* Move blocks from to-be-truncated inode over to a new inode, and delete ++ * that one from the delete thread instead. This avoids a lot of latency ++ * when truncating large files. ++ * ++ * If we have any problem deferring the truncate, just truncate it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ * ++ * During normal filesystem usage, we are always called here with a ++ * transaction already started. The only time ext3_truncate is called ++ * without a started transaction is from ext3_orphan_cleanup(), and we ++ * currently just do a direct truncate in that case. ++ */ ++void ext3_truncate_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ handle_t *handle; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_truncate; ++ ++ /* XXX This is a temporary limitation for code simplicity. ++ * We could truncate to arbitrary sizes at some later time. ++ */ ++ if (old_inode->i_size != 0) ++ goto out_truncate; ++ ++ /* We may want to truncate the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS || ++ old_inode->i_size > oei->i_disksize) ++ goto out_truncate; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_truncate; ++ } ++ ++ ext3_discard_prealloc(old_inode); ++ ++ /* old_inode = 1 ++ * new_inode = sb + GDT + ibitmap ++ * orphan list = 1 inode/superblock for add, 2 inodes for del ++ * quota files = 2 * EXT3_SINGLEDATA_TRANS_BLOCKS ++ */ ++ handle = ext3_journal_start(old_inode, 7); ++ if (IS_ERR(handle)) ++ goto out_truncate; ++ ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ if (IS_ERR(new_inode)) { ++ ext3_debug("truncate inode %lu directly (no new inodes)\n", ++ old_inode->i_ino); ++ goto out_journal; ++ } ++ ++ if (ext3_orphan_add(handle, new_inode) < 0) ++ goto out_journal; ++ ++ if (ext3_orphan_del(handle, old_inode) < 0) { ++ ext3_orphan_del(handle, new_inode); ++ iput(new_inode); ++ goto out_journal; ++ } ++ ++ nei = EXT3_I(new_inode); ++ ++ down_write(&oei->truncate_sem); ++ new_inode->i_size = old_inode->i_size; ++ new_inode->i_blocks = old_inode->i_blocks; ++ new_inode->i_uid = old_inode->i_uid; ++ new_inode->i_gid = old_inode->i_gid; ++ new_inode->i_nlink = 0; ++ ++ /* FIXME when we do arbitrary truncates */ ++ old_inode->i_blocks = oei->i_file_acl ? old_inode->i_blksize / 512 : 0; ++ ++ memcpy(nei->i_data, oei->i_data, sizeof(nei->i_data)); ++ memset(oei->i_data, 0, sizeof(oei->i_data)); ++ ++ nei->i_disksize = oei->i_disksize; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up_write(&oei->truncate_sem); ++ ++ ext3_journal_stop(handle, old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_journal: ++ ext3_journal_stop(handle, old_inode); ++out_truncate: ++ ext3_truncate(old_inode); ++} ++#endif /* EXT3_DELETE_THREAD */ ++ + /* + * On success, We end up with an outstanding reference count against + * iloc->bh. This _must_ be cleaned up later. +Index: linux-2.4.21-4.EL/fs/ext3/file.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/file.c 2004-06-23 19:09:01.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/file.c 2004-06-23 21:34:21.000000000 +0800 +@@ -131,7 +131,11 @@ + }; + + struct inode_operations ext3_file_inode_operations = { ++#ifdef EXT3_DELETE_THREAD ++ truncate: ext3_truncate_thread, /* BKL held */ ++#else + truncate: ext3_truncate, /* BKL held */ ++#endif + setattr: ext3_setattr, /* BKL held */ + setxattr: ext3_setxattr, /* BKL held */ + getxattr: ext3_getxattr, /* BKL held */ +Index: linux-2.4.21-4.EL/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs.h 2004-06-23 21:31:33.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs.h 2004-06-23 21:35:42.000000000 +0800 +@@ -193,6 +193,7 @@ + */ + #define EXT3_STATE_JDATA 0x00000001 /* journaled data exists */ + #define EXT3_STATE_NEW 0x00000002 /* inode is newly created */ ++#define EXT3_STATE_DELETE 0x00000010 /* deferred delete inode */ + + /* + * ioctl commands +@@ -320,6 +321,7 @@ + #define EXT3_MOUNT_UPDATE_JOURNAL 0x1000 /* Update the journal format */ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ + #define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ ++#define EXT3_MOUNT_ASYNCDEL 0x20000 /* Delayed deletion */ + #define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ +@@ -698,6 +700,9 @@ + extern void ext3_dirty_inode(struct inode *); + extern int ext3_change_inode_journal_flag(struct inode *, int); + extern void ext3_truncate (struct inode *); ++#ifdef EXT3_DELETE_THREAD ++extern void ext3_truncate_thread(struct inode *inode); ++#endif + extern void ext3_set_inode_flags(struct inode *); + + /* ioctl.c */ +Index: linux-2.4.21-4.EL/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs_sb.h 2004-06-23 21:10:01.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs_sb.h 2004-06-23 21:34:21.000000000 +0800 +@@ -29,6 +29,8 @@ + + #define EXT3_MAX_GROUP_LOADED 32 + ++#define EXT3_DELETE_THREAD ++ + /* + * third extended-fs super-block data in memory + */ +@@ -76,6 +78,14 @@ + struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */ + wait_queue_head_t ro_wait_queue; /* For people waiting for the fs to go read-only */ + #endif ++#ifdef EXT3_DELETE_THREAD ++ spinlock_t s_delete_lock; ++ struct list_head s_delete_list; ++ unsigned long s_delete_blocks; ++ unsigned long s_delete_inodes; ++ wait_queue_head_t s_delete_thread_queue; ++ wait_queue_head_t s_delete_waiter_queue; ++#endif + }; + + #endif /* _LINUX_EXT3_FS_SB */ diff --git a/lustre/kernel_patches/patches/extN-wantedi-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/extN-wantedi-2.4.21-rh-as.patch new file mode 100644 index 0000000..9808193 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-wantedi-2.4.21-rh-as.patch @@ -0,0 +1,227 @@ + fs/ext3/ialloc.c | 40 ++++++++++++++++++++++++++++++++++++++-- + fs/ext3/inode.c | 2 +- + fs/ext3/ioctl.c | 25 +++++++++++++++++++++++++ + fs/ext3/namei.c | 21 +++++++++++++++++---- + include/linux/dcache.h | 5 +++++ + include/linux/ext3_fs.h | 5 ++++- + 6 files changed, 90 insertions(+), 8 deletions(-) + +Index: linux-2.4.21-4.EL/fs/ext3/namei.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/namei.c 2004-06-23 21:48:51.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/namei.c 2004-06-23 21:52:58.000000000 +0800 +@@ -1534,6 +1534,19 @@ + return err; + } + ++static struct inode * ext3_new_inode_wantedi(handle_t *handle, struct inode *dir, ++ int mode, struct dentry *dentry) ++{ ++ unsigned long inum = 0; ++ ++ if (dentry->d_fsdata != NULL) { ++ struct dentry_params *param = ++ (struct dentry_params *) dentry->d_fsdata; ++ inum = param->p_inum; ++ } ++ return ext3_new_inode(handle, dir, mode, inum); ++} ++ + /* + * By the time this is called, we already have created + * the directory cache entry for the new file, but it +@@ -1557,7 +1570,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + inode->i_op = &ext3_file_inode_operations; +@@ -1585,7 +1598,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, inode->i_mode, rdev); +@@ -1618,7 +1631,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR | mode); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFDIR | mode, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -2013,7 +2026,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFLNK|S_IRWXUGO, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +Index: linux-2.4.21-4.EL/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/ialloc.c 2004-06-23 21:49:38.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/ialloc.c 2004-06-23 21:54:12.000000000 +0800 +@@ -330,7 +330,9 @@ + * For other inodes, search forward from the parent directory's block + * group to find a free inode. + */ +-struct inode * ext3_new_inode (handle_t *handle, struct inode * dir, int mode) ++struct inode * ext3_new_inode (handle_t *handle, ++ const struct inode * dir, int mode, ++ unsigned long goal) + { + struct super_block * sb; + struct buffer_head * bh; +@@ -355,7 +357,41 @@ + init_rwsem(&inode->u.ext3_i.truncate_sem); + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; ++ ++ if (goal) { ++ i = (goal - 1) / EXT3_INODES_PER_GROUP(sb); ++ j = (goal - 1) % EXT3_INODES_PER_GROUP(sb); ++ gdp = ext3_get_group_desc(sb, i, &bh2); ++ ++ bitmap_nr = load_inode_bitmap (sb, i); ++ if (bitmap_nr < 0) { ++ err = bitmap_nr; ++ goto fail; ++ } ++ ++ bh = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; ++ ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) goto fail; ++ ++ if (ext3_set_bit(j, bh->b_data)) { ++ printk(KERN_ERR "goal inode %lu unavailable\n", goal); ++ /* Oh well, we tried. */ ++ goto repeat; ++ } ++ ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) goto fail; ++ ++ /* We've shortcircuited the allocation system successfully, ++ * now finish filling in the inode. ++ */ ++ goto have_bit_and_group; ++ } ++ + repeat: + gdp = NULL; + i = 0; +@@ -470,6 +506,7 @@ + } + goto repeat; + } ++ have_bit_and_group: + j += i * EXT3_INODES_PER_GROUP(sb) + 1; + if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", +Index: linux-2.4.21-4.EL/fs/ext3/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/inode.c 2004-06-23 21:52:32.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/inode.c 2004-06-23 21:52:58.000000000 +0800 +@@ -2798,7 +2798,7 @@ + if (IS_ERR(handle)) + goto out_truncate; + +- new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode, 0); + if (IS_ERR(new_inode)) { + ext3_debug("truncate inode %lu directly (no new inodes)\n", + old_inode->i_ino); +Index: linux-2.4.21-4.EL/fs/ext3/ioctl.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/ext3/ioctl.c 2003-10-04 05:28:53.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/ext3/ioctl.c 2004-06-23 21:52:58.000000000 +0800 +@@ -23,6 +23,31 @@ + ext3_debug ("cmd = %u, arg = %lu\n", cmd, arg); + + switch (cmd) { ++ case EXT3_IOC_CREATE_INUM: { ++ char name[32]; ++ struct dentry *dchild, *dparent; ++ int rc = 0; ++ ++ dparent = list_entry(inode->i_dentry.next, struct dentry, ++ d_alias); ++ snprintf(name, sizeof name, "%lu", arg); ++ dchild = lookup_one_len(name, dparent, strlen(name)); ++ if (dchild->d_inode) { ++ printk(KERN_ERR "%*s/%lu already exists (ino %lu)\n", ++ dparent->d_name.len, dparent->d_name.name, arg, ++ dchild->d_inode->i_ino); ++ rc = -EEXIST; ++ } else { ++ dchild->d_fsdata = (void *)arg; ++ rc = vfs_create(inode, dchild, 0644); ++ if (rc) ++ printk(KERN_ERR "vfs_create: %d\n", rc); ++ else if (dchild->d_inode->i_ino != arg) ++ rc = -EEXIST; ++ } ++ dput(dchild); ++ return rc; ++ } + case EXT3_IOC_GETFLAGS: + flags = inode->u.ext3_i.i_flags & EXT3_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); +Index: linux-2.4.21-4.EL/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/ext3_fs.h 2004-06-23 21:52:32.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/ext3_fs.h 2004-06-23 21:56:39.000000000 +0800 +@@ -202,6 +202,7 @@ + #define EXT3_IOC_SETFLAGS _IOW('f', 2, long) + #define EXT3_IOC_GETVERSION _IOR('f', 3, long) + #define EXT3_IOC_SETVERSION _IOW('f', 4, long) ++/* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ + #define EXT3_IOC_GETVERSION_OLD _IOR('v', 1, long) + #define EXT3_IOC_SETVERSION_OLD _IOW('v', 2, long) + #ifdef CONFIG_JBD_DEBUG +@@ -675,7 +676,8 @@ + dx_hash_info *hinfo); + + /* ialloc.c */ +-extern struct inode * ext3_new_inode (handle_t *, struct inode *, int); ++extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int, unsigned long); ++ + extern void ext3_free_inode (handle_t *, struct inode *); + extern struct inode * ext3_orphan_get (struct super_block *, unsigned long); + extern unsigned long ext3_count_free_inodes (struct super_block *); +@@ -767,4 +769,5 @@ + + #endif /* __KERNEL__ */ + ++#define EXT3_IOC_CREATE_INUM _IOW('f', 5, long) + #endif /* _LINUX_EXT3_FS_H */ +Index: linux-2.4.21-4.EL/include/linux/dcache.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/dcache.h 2004-06-23 21:45:06.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/dcache.h 2004-06-23 21:52:58.000000000 +0800 +@@ -63,6 +63,11 @@ + + #define IS_ROOT(x) ((x) == (x)->d_parent) + ++struct dentry_params { ++ unsigned long p_inum; ++ void *p_ptr; ++}; ++ + /* + * "quick string" -- eases parameter passing, but more importantly + * saves "metadata" about the string (ie length and the hash). diff --git a/lustre/kernel_patches/patches/gfp_memalloc-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/gfp_memalloc-2.4.21-rh-as.patch new file mode 100644 index 0000000..f4d5848 --- /dev/null +++ b/lustre/kernel_patches/patches/gfp_memalloc-2.4.21-rh-as.patch @@ -0,0 +1,70 @@ +Index: linux-2.4.21-4.EL/include/linux/mm.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/mm.h 2004-06-24 02:40:51.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/mm.h 2004-06-24 02:58:42.000000000 +0800 +@@ -830,6 +830,7 @@ + #define __GFP_IO 0x40 /* Can start low memory physical IO? */ + #define __GFP_HIGHIO 0x80 /* Can start high mem physical IO? */ + #define __GFP_FS 0x100 /* Can call down to low-level FS? */ ++#define __GFP_MEMALLOC 0x200 /* like PF_MEMALLOC: see __alloc_pages */ + + #define GFP_NOHIGHIO (__GFP_HIGH | __GFP_WAIT | __GFP_IO) + #define GFP_NOIO (__GFP_HIGH | __GFP_WAIT) +@@ -840,6 +841,7 @@ + #define GFP_KERNEL (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_NFS (__GFP_HIGH | __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) + #define GFP_KSWAPD ( __GFP_WAIT | __GFP_IO | __GFP_HIGHIO | __GFP_FS) ++#define GFP_MEMALLOC __GFP_MEMALLOC + + /* Flag - indicates that the buffer will be suitable for DMA. Ignored on some + platforms, used as appropriate on others */ +Index: linux-2.4.21-4.EL/include/linux/slab.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/slab.h 2003-10-04 05:29:48.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/slab.h 2004-06-24 02:55:59.000000000 +0800 +@@ -23,6 +23,7 @@ + #define SLAB_KERNEL GFP_KERNEL + #define SLAB_NFS GFP_NFS + #define SLAB_DMA GFP_DMA ++#define SLAB_MEMALLOC GFP_MEMALLOC + + #define SLAB_LEVEL_MASK (__GFP_WAIT|__GFP_HIGH|__GFP_IO|__GFP_HIGHIO|__GFP_FS) + #define SLAB_NO_GROW 0x00001000UL /* don't grow a cache */ +Index: linux-2.4.21-4.EL/mm/page_alloc.c +=================================================================== +--- linux-2.4.21-4.EL.orig/mm/page_alloc.c 2004-06-24 02:40:38.000000000 +0800 ++++ linux-2.4.21-4.EL/mm/page_alloc.c 2004-06-24 02:55:59.000000000 +0800 +@@ -563,7 +563,8 @@ + /* + * Oh well, we didn't succeed. + */ +- if (!(current->flags & (PF_MEMALLOC|PF_MEMDIE))) { ++ if (!(current->flags & (PF_MEMALLOC|PF_MEMDIE)) && ++ !(gfp_mask & __GFP_MEMALLOC)) { + /* + * Are we dealing with a higher order allocation? + * +@@ -636,7 +637,9 @@ + + /* XXX: is pages_min/4 a good amount to reserve for this? */ + min += z->pages_min / 4; +- if (z->free_pages > min || ((current->flags & PF_MEMALLOC) && !in_interrupt())) { ++ if (z->free_pages > min || ++ (((current->flags & PF_MEMALLOC) || (gfp_mask & __GFP_MEMALLOC)) ++ && !in_interrupt())) { + page = rmqueue(z, order); + if (page) + return page; +Index: linux-2.4.21-4.EL/mm/slab.c +=================================================================== +--- linux-2.4.21-4.EL.orig/mm/slab.c 2003-10-04 05:28:49.000000000 +0800 ++++ linux-2.4.21-4.EL/mm/slab.c 2004-06-24 02:55:59.000000000 +0800 +@@ -1118,7 +1118,7 @@ + /* Be lazy and only check for valid flags here, + * keeping it out of the critical path in kmem_cache_alloc(). + */ +- if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW)) ++ if (flags & ~(SLAB_DMA|SLAB_LEVEL_MASK|SLAB_NO_GROW|SLAB_MEMALLOC)) + BUG(); + if (flags & SLAB_NO_GROW) + return 0; diff --git a/lustre/kernel_patches/patches/iod-stock-exports-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/iod-stock-exports-2.4.21-rh-as.patch new file mode 100644 index 0000000..5918434 --- /dev/null +++ b/lustre/kernel_patches/patches/iod-stock-exports-2.4.21-rh-as.patch @@ -0,0 +1,52 @@ + fs/Makefile | 2 +- + fs/inode.c | 4 +++- + mm/page_alloc.c | 1 + + 3 files changed, 5 insertions(+), 2 deletions(-) + +Index: linux-2.4.21-4.EL/fs/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/inode.c 2004-06-23 18:55:10.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/inode.c 2004-06-23 18:57:23.000000000 +0800 +@@ -5,6 +5,7 @@ + */ + + #include ++#include + #include + #include + #include +@@ -69,7 +70,8 @@ + * NOTE! You also have to own the lock if you change + * the i_state of an inode while it is in use.. + */ +-static spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++EXPORT_SYMBOL(inode_lock); + + /* + * Statistics gathering.. +Index: linux-2.4.21-4.EL/fs/Makefile +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/Makefile 2003-10-04 05:28:48.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/Makefile 2004-06-23 18:59:40.000000000 +0800 +@@ -7,7 +7,7 @@ + + O_TARGET := fs.o + +-export-objs := filesystems.o open.o dcache.o buffer.o dquot.o dcookies.o ++export-objs := filesystems.o open.o dcache.o buffer.o dquot.o dcookies.o inode.o + mod-subdirs := nls + + obj-y := open.o read_write.o devices.o file_table.o buffer.o \ +Index: linux-2.4.21-4.EL/mm/page_alloc.c +=================================================================== +--- linux-2.4.21-4.EL.orig/mm/page_alloc.c 2003-10-04 05:28:44.000000000 +0800 ++++ linux-2.4.21-4.EL/mm/page_alloc.c 2004-06-23 18:57:23.000000000 +0800 +@@ -27,6 +27,7 @@ + + int nr_swap_pages; + pg_data_t *pgdat_list; ++EXPORT_SYMBOL(pgdat_list); + + /* + * diff --git a/lustre/kernel_patches/patches/nfs_export_kernel-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/nfs_export_kernel-2.4.21-rh-as.patch new file mode 100644 index 0000000..c2f0e2b --- /dev/null +++ b/lustre/kernel_patches/patches/nfs_export_kernel-2.4.21-rh-as.patch @@ -0,0 +1,746 @@ + fs/Makefile | 3 + fs/file_table.c | 11 ++ + fs/inode.c | 23 ++++- + fs/namei.c | 12 ++ + fs/nfsd/export.c | 5 + + fs/nfsd/nfsfh.c | 65 +++++++++++++- + fs/nfsd/vfs.c | 235 ++++++++++++++++++++++++++++++++++++++++++++++++----- + include/linux/fs.h | 11 ++ + kernel/ksyms.c | 2 + 9 files changed, 333 insertions(+), 34 deletions(-) + +Index: linux-2.4.21-chaos/fs/file_table.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/file_table.c 2003-07-15 04:41:00.000000000 +0400 ++++ linux-2.4.21-chaos/fs/file_table.c 2003-12-12 16:19:25.000000000 +0300 +@@ -82,7 +82,8 @@ + * and call the open function (if any). The caller must verify that + * inode->i_fop is not NULL. + */ +-int init_private_file(struct file *filp, struct dentry *dentry, int mode) ++int init_private_file_it(struct file *filp, struct dentry *dentry, int mode, ++ struct lookup_intent *it) + { + memset(filp, 0, sizeof(*filp)); + filp->f_mode = mode; +@@ -90,12 +91,20 @@ + filp->f_dentry = dentry; + filp->f_uid = current->fsuid; + filp->f_gid = current->fsgid; ++ if (it) ++ filp->f_it = it; + filp->f_op = dentry->d_inode->i_fop; + if (filp->f_op->open) + return filp->f_op->open(dentry->d_inode, filp); + else + return 0; + } ++EXPORT_SYMBOL(init_private_file_it); ++ ++int init_private_file(struct file *filp, struct dentry *dentry, int mode) ++{ ++ return init_private_file_it(filp, dentry, mode, NULL); ++} + + void fput(struct file * file) + { +Index: linux-2.4.21-chaos/fs/inode.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/inode.c 2003-12-12 16:18:15.000000000 +0300 ++++ linux-2.4.21-chaos/fs/inode.c 2003-12-12 16:19:25.000000000 +0300 +@@ -1054,9 +1054,10 @@ + return inode; + } + +-struct inode *iget4_locked(struct super_block *sb, unsigned long ino, find_inode_t find_actor, void *opaque) ++struct inode *ifind(struct super_block *sb, unsigned long ino, ++ struct list_head *head, ++ find_inode_t find_actor, void *opaque) + { +- struct list_head * head = inode_hashtable + hash(sb,ino); + struct inode * inode; + + spin_lock(&inode_lock); +@@ -1069,6 +1070,24 @@ + } + spin_unlock(&inode_lock); + ++ return NULL; ++} ++ ++struct inode *ilookup4(struct super_block *sb, unsigned long ino, ++ find_inode_t find_actor, void *opaque) ++{ ++ struct list_head * head = inode_hashtable + hash(sb,ino); ++ return ifind(sb, ino, head, find_actor, opaque); ++} ++ ++struct inode *iget4_locked(struct super_block *sb, unsigned long ino, ++ find_inode_t find_actor, void *opaque) ++{ ++ struct list_head * head = inode_hashtable + hash(sb,ino); ++ struct inode *inode = ifind(sb, ino, head, find_actor, opaque); ++ if (inode) ++ return inode; ++ + /* + * get_new_inode() will do the right thing, re-trying the search + * in case it had to block at any point. +Index: linux-2.4.21-chaos/fs/Makefile +=================================================================== +--- linux-2.4.21-chaos.orig/fs/Makefile 2003-12-12 16:18:36.000000000 +0300 ++++ linux-2.4.21-chaos/fs/Makefile 2003-12-12 16:19:25.000000000 +0300 +@@ -9,7 +9,8 @@ + + O_TARGET := fs.o + +-export-objs := filesystems.o open.o dcache.o buffer.o dquot.o dcookies.o inode.o ++export-objs := filesystems.o open.o dcache.o buffer.o dquot.o dcookies.o inode.o \ ++ namei.o file_table.o + mod-subdirs := nls + + obj-y := open.o read_write.o devices.o file_table.o buffer.o \ +Index: linux-2.4.21-chaos/fs/namei.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/namei.c 2003-12-12 16:18:06.000000000 +0300 ++++ linux-2.4.21-chaos/fs/namei.c 2003-12-12 16:19:25.000000000 +0300 +@@ -22,6 +22,7 @@ + #include + #include + #include ++#include + + #include + #include +@@ -100,6 +101,7 @@ + it->it_op_release(it); + + } ++EXPORT_SYMBOL(intent_release); + + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the +@@ -910,7 +912,8 @@ + + + /* SMP-safe */ +-struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) ++struct dentry * lookup_one_len_it(const char * name, struct dentry * base, ++ int len, struct lookup_intent *it) + { + unsigned long hash; + struct qstr this; +@@ -930,11 +933,16 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash_it(&this, base, NULL); ++ return lookup_hash_it(&this, base, it); + access: + return ERR_PTR(-EACCES); + } + ++struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) ++{ ++ return lookup_one_len_it(name, base, len, NULL); ++} ++ + /* + * namei() + * +Index: linux-2.4.21-chaos/fs/nfsd/export.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/nfsd/export.c 2003-09-19 03:49:54.000000000 +0400 ++++ linux-2.4.21-chaos/fs/nfsd/export.c 2003-12-12 16:19:25.000000000 +0300 +@@ -223,6 +223,11 @@ + inode = nd.dentry->d_inode; + dev = inode->i_dev; + ino = inode->i_ino; ++ if ((inode->i_sb->s_type->fs_flags & FS_NFSEXP_FSID) && ++ !(nxp->ex_flags & NFSEXP_FSID)) { ++ nxp->ex_dev = inode->i_sb->s_dev; ++ nxp->ex_flags |= NFSEXP_FSID; ++ } + err = -EINVAL; + + exp = exp_get(clp, dev, ino); +Index: linux-2.4.21-chaos/fs/nfsd/nfsfh.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/nfsd/nfsfh.c 2003-09-19 03:49:54.000000000 +0400 ++++ linux-2.4.21-chaos/fs/nfsd/nfsfh.c 2003-12-12 16:19:25.000000000 +0300 +@@ -36,6 +36,15 @@ + int sequence; /* sequence counter */ + }; + ++static struct dentry *lookup_it(struct inode *inode, struct dentry * dentry) ++{ ++ if (inode->i_op->lookup_it) ++ return inode->i_op->lookup_it(inode, dentry, NULL, 0); ++ else ++ return inode->i_op->lookup(inode, dentry); ++ ++} ++ + /* + * A rather strange filldir function to capture + * the name matching the specified inode number. +@@ -75,6 +84,8 @@ + int error; + struct file file; + struct nfsd_getdents_callback buffer; ++ struct lookup_intent it; ++ struct file *filp = NULL; + + error = -ENOTDIR; + if (!dir || !S_ISDIR(dir->i_mode)) +@@ -85,9 +96,37 @@ + /* + * Open the directory ... + */ +- error = init_private_file(&file, dentry, FMODE_READ); +- if (error) ++ if (dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if ((dentry->d_flags & DCACHE_NFSD_DISCONNECTED) && ++ (dentry->d_parent == dentry) ) { ++ it.it_op_release = NULL; ++ /* ++ * XXX Temporary Hack: Simulating init_private_file without ++ * f_op->open for disconnected dentry Since we don't have actual ++ * dentry->d_name to revalidate in revalidate_it() ++ */ ++ filp = &file; ++ memset(filp, 0, sizeof(*filp)); ++ filp->f_mode = FMODE_READ; ++ atomic_set(&filp->f_count, 1); ++ filp->f_dentry = dentry; ++ filp->f_uid = current->fsuid; ++ filp->f_gid = current->fsgid; ++ filp->f_op = dentry->d_inode->i_fop; ++ error = 0; ++ } else { ++ intent_init(&it, IT_OPEN, 0); ++ error = revalidate_it(dentry, &it); ++ if (error) ++ goto out; ++ error = init_private_file_it(&file, dentry, FMODE_READ, &it); ++ } ++ } else { ++ error = init_private_file_it(&file, dentry, FMODE_READ, NULL); ++ } ++ if (error) + goto out; ++ + error = -EINVAL; + if (!file.f_op->readdir) + goto out_close; +@@ -113,9 +152,13 @@ + } + + out_close: +- if (file.f_op->release) ++ if (file.f_op->release && !filp) + file.f_op->release(dir, &file); + out: ++ if (dentry->d_op && ++ dentry->d_op->d_revalidate_it && ++ it.it_op_release && !filp) ++ intent_release(&it); + return error; + } + +@@ -274,7 +317,7 @@ + * it is well connected. But nobody returns different dentrys do they? + */ + down(&child->d_inode->i_sem); +- pdentry = child->d_inode->i_op->lookup(child->d_inode, tdentry); ++ pdentry = lookup_it(child->d_inode, tdentry); + up(&child->d_inode->i_sem); + d_drop(tdentry); /* we never want ".." hashed */ + if (!pdentry && tdentry->d_inode == NULL) { +@@ -307,6 +350,8 @@ + pdentry->d_flags |= DCACHE_NFSD_DISCONNECTED; + pdentry->d_op = child->d_op; + } ++ if (child->d_op && child->d_op->d_revalidate_it) ++ pdentry->d_op = child->d_op; + } + if (pdentry == NULL) + pdentry = ERR_PTR(-ENOMEM); +@@ -464,6 +509,8 @@ + struct dentry *pdentry; + struct inode *parent; + ++ if (result->d_op && result->d_op->d_revalidate_it) ++ dentry->d_op = result->d_op; + pdentry = nfsd_findparent(dentry); + err = PTR_ERR(pdentry); + if (IS_ERR(pdentry)) +@@ -672,6 +719,11 @@ + + inode = dentry->d_inode; + ++ /* cache coherency for non-device filesystems */ ++ if (inode->i_op && inode->i_op->revalidate_it) { ++ inode->i_op->revalidate_it(dentry, NULL); ++ } ++ + /* Type check. The correct error return for type mismatches + * does not seem to be generally agreed upon. SunOS seems to + * use EISDIR if file isn't S_IFREG; a comment in the NFSv3 +@@ -905,8 +957,9 @@ + dentry->d_parent->d_name.name, dentry->d_name.name); + goto out; + out_uptodate: +- printk(KERN_ERR "fh_update: %s/%s already up-to-date!\n", +- dentry->d_parent->d_name.name, dentry->d_name.name); ++ if(!dentry->d_parent->d_inode->i_op->mkdir_raw) ++ printk(KERN_ERR "fh_update: %s/%s already up-to-date!\n", ++ dentry->d_parent->d_name.name, dentry->d_name.name); + goto out; + } + +Index: linux-2.4.21-chaos/fs/nfsd/vfs.c +=================================================================== +--- linux-2.4.21-chaos.orig/fs/nfsd/vfs.c 2003-09-19 03:49:54.000000000 +0400 ++++ linux-2.4.21-chaos/fs/nfsd/vfs.c 2003-12-12 16:19:25.000000000 +0300 +@@ -78,6 +78,128 @@ + static struct raparms * raparml; + static struct raparms * raparm_cache; + ++static int link_raw(struct dentry *dold, struct dentry *ddir, ++ struct dentry *dnew) ++{ ++ int err; ++ ++ struct nameidata old_nd = { .dentry = dold }; ++ struct nameidata nd = { .dentry = ddir, .last = dnew->d_name }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->link_raw(&old_nd, &nd); ++ d_instantiate(dnew, dold->d_inode); ++ if(dold->d_inode->i_op && dold->d_inode->i_op->revalidate_it) ++ dold->d_inode->i_op->revalidate_it(dnew, NULL); ++ ++ return err; ++} ++ ++static int unlink_raw(struct dentry *dentry, char *fname, int flen, ++ struct dentry *rdentry) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->unlink_raw(&nd); ++ if (!err) ++ d_delete(rdentry); ++ ++ return err; ++} ++ ++static int rmdir_raw(struct dentry *dentry, char *fname, int flen, ++ struct dentry *rdentry) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->rmdir_raw(&nd); ++ if(!err) { ++ rdentry->d_inode->i_flags |= S_DEAD; ++ d_delete(rdentry); ++ } ++ ++ return err; ++} ++ ++static int symlink_raw(struct dentry *dentry, char *fname, int flen, ++ char *path) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->symlink_raw(&nd, path); ++ ++ return err; ++} ++ ++static int mkdir_raw(struct dentry *dentry, char *fname, int flen, int mode) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->mkdir_raw(&nd, mode); ++ ++ return err; ++} ++ ++static int mknod_raw(struct dentry *dentry, char *fname, int flen, int mode, ++ dev_t dev) ++{ ++ int err; ++ struct qstr last = { .name = fname, .len = flen }; ++ struct nameidata nd = { .dentry = dentry, .last = last }; ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ err = op->mknod_raw(&nd, mode, dev); ++ ++ return err; ++} ++ ++static int rename_raw(struct dentry *fdentry, struct dentry *tdentry, ++ struct dentry *odentry, struct dentry *ndentry) ++{ ++ int err; ++ ++ struct nameidata old_nd = { .dentry = fdentry, .last = odentry->d_name}; ++ struct nameidata new_nd = { .dentry = tdentry, .last = ndentry->d_name}; ++ struct inode_operations *op = old_nd.dentry->d_inode->i_op; ++ err = op->rename_raw(&old_nd, &new_nd); ++ d_move(odentry, ndentry); ++ ++ return err; ++} ++ ++static int setattr_raw(struct inode *inode, struct iattr *iap) ++{ ++ int err; ++ ++ iap->ia_valid |= ATTR_RAW; ++ err = inode->i_op->setattr_raw(inode, iap); ++ ++ return err; ++} ++ ++int revalidate_it(struct dentry *dentry, struct lookup_intent *it) ++{ ++ int err = 0; ++ ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ err = -EINVAL; ++ dentry = NULL; ++ return err; ++ } ++ } ++ ++ return err; ++} ++ + /* + * Look up one component of a pathname. + * N.B. After this call _both_ fhp and resfh need an fh_put +@@ -303,7 +425,10 @@ + } + err = nfserr_notsync; + if (!check_guard || guardtime == inode->i_ctime) { +- err = notify_change(dentry, iap); ++ if ( dentry->d_inode->i_op && dentry->d_inode->i_op->setattr_raw) ++ err = setattr_raw(dentry->d_inode, iap); ++ else ++ err = notify_change(dentry, iap); + err = nfserrno(err); + } + if (size_change) { +@@ -430,6 +555,7 @@ + { + struct dentry *dentry; + struct inode *inode; ++ struct lookup_intent it; + int err; + + /* If we get here, then the client has already done an "open", and (hopefully) +@@ -476,6 +602,14 @@ + filp->f_mode = FMODE_READ; + } + ++ intent_init(&it, IT_OPEN, (filp->f_flags & ~O_ACCMODE) | filp->f_mode); ++ ++ err = revalidate_it(dentry, &it); ++ if (err) ++ goto out_nfserr; ++ ++ filp->f_it = ⁢ ++ + err = 0; + if (filp->f_op && filp->f_op->open) { + err = filp->f_op->open(inode, filp); +@@ -490,7 +624,11 @@ + atomic_dec(&filp->f_count); + } + } ++ + out_nfserr: ++ if (it.it_op_release) ++ intent_release(&it); ++ + if (err) + err = nfserrno(err); + out: +@@ -821,7 +959,7 @@ + { + struct dentry *dentry, *dchild; + struct inode *dirp; +- int err; ++ int err, error = -EOPNOTSUPP; + + err = nfserr_perm; + if (!flen) +@@ -837,20 +975,44 @@ + dentry = fhp->fh_dentry; + dirp = dentry->d_inode; + ++ switch (type) { ++ case S_IFDIR: ++ if (dirp->i_op->mkdir_raw) ++ error = mkdir_raw(dentry, fname, flen, iap->ia_mode); ++ break; ++ case S_IFCHR: ++ case S_IFBLK: ++ case S_IFIFO: ++ case S_IFSOCK: ++ case S_IFREG: ++ if (dirp->i_op->mknod_raw) { ++ if (type == S_IFREG) ++ rdev = 0; ++ error = mknod_raw(dentry, fname, flen, iap->ia_mode, rdev); ++ } ++ break; ++ default: ++ printk("nfsd: bad file type %o in nfsd_create\n", type); ++ } ++ + err = nfserr_notdir; +- if(!dirp->i_op || !dirp->i_op->lookup) ++ if(!dirp->i_op || !(dirp->i_op->lookup || dirp->i_op->lookup_it)) + goto out; + /* + * Check whether the response file handle has been verified yet. + * If it has, the parent directory should already be locked. + */ +- if (!resfhp->fh_dentry) { +- /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create */ +- fh_lock(fhp); ++ if (!resfhp->fh_dentry || dirp->i_op->lookup_it) { ++ /* called from nfsd_proc_mkdir, or possibly nfsd3_proc_create ++ and nfsd_proc_create in case of lustre ++ */ ++ if (!resfhp->fh_dentry) ++ fh_lock(fhp); + dchild = lookup_one_len(fname, dentry, flen); + err = PTR_ERR(dchild); + if (IS_ERR(dchild)) + goto out_nfserr; ++ resfhp->fh_dentry = NULL; + err = fh_compose(resfhp, fhp->fh_export, dchild, fhp); + if (err) + goto out; +@@ -871,10 +1033,12 @@ + * Make sure the child dentry is still negative ... + */ + err = nfserr_exist; +- if (dchild->d_inode) { +- dprintk("nfsd_create: dentry %s/%s not negative!\n", +- dentry->d_name.name, dchild->d_name.name); +- goto out; ++ if ( error == -EOPNOTSUPP) { ++ if (dchild->d_inode) { ++ dprintk("nfsd_create: dentry %s/%s not negative!\n", ++ dentry->d_name.name, dchild->d_name.name); ++ goto out; ++ } + } + + if (!(iap->ia_valid & ATTR_MODE)) +@@ -887,16 +1051,19 @@ + err = nfserr_perm; + switch (type) { + case S_IFREG: +- err = vfs_create(dirp, dchild, iap->ia_mode); ++ if (error == -EOPNOTSUPP) ++ err = vfs_create(dirp, dchild, iap->ia_mode); + break; + case S_IFDIR: +- err = vfs_mkdir(dirp, dchild, iap->ia_mode); ++ if (error == -EOPNOTSUPP) ++ err = vfs_mkdir(dirp, dchild, iap->ia_mode); + break; + case S_IFCHR: + case S_IFBLK: + case S_IFIFO: + case S_IFSOCK: +- err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); ++ if (error == -EOPNOTSUPP) ++ err = vfs_mknod(dirp, dchild, iap->ia_mode, rdev); + break; + default: + printk("nfsd: bad file type %o in nfsd_create\n", type); +@@ -965,7 +1132,13 @@ + /* Get all the sanity checks out of the way before + * we lock the parent. */ + err = nfserr_notdir; +- if(!dirp->i_op || !dirp->i_op->lookup) ++ if (dirp->i_op->mknod_raw) { ++ err = mknod_raw(dentry, fname, flen, iap->ia_mode, 0); ++ if (err && err != -EOPNOTSUPP) ++ goto out; ++ } ++ ++ if(!dirp->i_op || !(dirp->i_op->lookup || dirp->i_op->lookup_it)) + goto out; + fh_lock(fhp); + +@@ -1016,6 +1189,8 @@ + case NFS3_CREATE_GUARDED: + err = nfserr_exist; + } ++ if(dirp->i_op->mknod_raw) ++ err = 0; + goto out; + } + +@@ -1122,7 +1297,7 @@ + struct iattr *iap) + { + struct dentry *dentry, *dnew; +- int err, cerr; ++ int err, cerr, error = -EOPNOTSUPP; + + err = nfserr_noent; + if (!flen || !plen) +@@ -1136,12 +1311,18 @@ + goto out; + fh_lock(fhp); + dentry = fhp->fh_dentry; ++ ++ if (dentry->d_inode->i_op->symlink_raw) ++ error = symlink_raw(dentry, fname, flen, path); ++ + dnew = lookup_one_len(fname, dentry, flen); + err = PTR_ERR(dnew); + if (IS_ERR(dnew)) + goto out_nfserr; + +- err = vfs_symlink(dentry->d_inode, dnew, path); ++ err = error; ++ if (err == -EOPNOTSUPP || !dentry->d_inode->i_op->symlink_raw) ++ err = vfs_symlink(dentry->d_inode, dnew, path); + if (!err) { + if (EX_ISSYNC(fhp->fh_export)) + nfsd_sync_dir(dentry); +@@ -1211,7 +1392,10 @@ + dold = tfhp->fh_dentry; + dest = dold->d_inode; + +- err = vfs_link(dold, dirp, dnew); ++ if (dirp->i_op->link_raw) ++ err = link_raw(dold, ddir, dnew); ++ else ++ err = vfs_link(dold, dirp, dnew); + if (!err) { + if (EX_ISSYNC(ffhp->fh_export)) { + nfsd_sync_dir(ddir); +@@ -1296,7 +1480,10 @@ + err = nfserr_perm; + } else + #endif +- err = vfs_rename(fdir, odentry, tdir, ndentry); ++ if(fdir->i_op->rename_raw) ++ err = rename_raw(fdentry, tdentry, odentry, ndentry); ++ else ++ err = vfs_rename(fdir, odentry, tdir, ndentry); + if (!err && EX_ISSYNC(tfhp->fh_export)) { + nfsd_sync_dir(tdentry); + nfsd_sync_dir(fdentry); +@@ -1317,7 +1504,7 @@ + fill_post_wcc(tfhp); + double_up(&tdir->i_sem, &fdir->i_sem); + ffhp->fh_locked = tfhp->fh_locked = 0; +- ++ + out: + return err; + } +@@ -1363,9 +1550,15 @@ + err = nfserr_perm; + } else + #endif +- err = vfs_unlink(dirp, rdentry); ++ if (dirp->i_op->unlink_raw) ++ err = unlink_raw(dentry, fname, flen, rdentry); ++ else ++ err = vfs_unlink(dirp, rdentry); + } else { /* It's RMDIR */ +- err = vfs_rmdir(dirp, rdentry); ++ if (dirp->i_op->rmdir_raw) ++ err = rmdir_raw(dentry, fname, flen, rdentry); ++ else ++ err = vfs_rmdir(dirp, rdentry); + } + + dput(rdentry); +Index: linux-2.4.21-chaos/include/linux/fs.h +=================================================================== +--- linux-2.4.21-chaos.orig/include/linux/fs.h 2003-12-12 16:19:23.000000000 +0300 ++++ linux-2.4.21-chaos/include/linux/fs.h 2003-12-12 16:19:25.000000000 +0300 +@@ -93,6 +93,9 @@ + #define FS_SINGLE 8 /* Filesystem that can have only one superblock */ + #define FS_NOMOUNT 16 /* Never mount from userland */ + #define FS_LITTER 32 /* Keeps the tree in dcache */ ++#define FS_NFSEXP_FSID 64 /* Use file system specific fsid for ++ * exporting non device filesystems. ++ */ + #define FS_ODD_RENAME 32768 /* Temporary stuff; will go away as soon + * as nfs_rename() will be cleaned up + */ +@@ -1159,6 +1162,9 @@ + struct nameidata *nd, struct lookup_intent *it); + extern struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, + int flags, struct lookup_intent *it); ++extern int revalidate_it(struct dentry *dentry, struct lookup_intent *it); ++extern int init_private_file_it(struct file *, struct dentry *dentry, int mode, ++ struct lookup_intent *it); + extern int filp_close(struct file *, fl_owner_t id); + extern char * getname(const char *); + +@@ -1458,6 +1464,8 @@ + extern int follow_down(struct vfsmount **, struct dentry **); + extern int follow_up(struct vfsmount **, struct dentry **); + extern struct dentry * lookup_one_len(const char *, struct dentry *, int); ++extern struct dentry * lookup_one_len_it(const char *, struct dentry *, int, ++ struct lookup_intent *); + extern struct dentry * lookup_hash(struct qstr *, struct dentry *); + #define user_path_walk(name,nd) __user_walk(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd) + #define user_path_walk_link(name,nd) __user_walk(name, LOOKUP_POSITIVE, nd) +@@ -1477,7 +1485,8 @@ + + extern struct inode * iget4_locked(struct super_block *, unsigned long, + find_inode_t, void *); +- ++extern struct inode * ilookup4(struct super_block *, unsigned long, ++ find_inode_t, void *); + static inline struct inode *iget4(struct super_block *sb, unsigned long ino, + find_inode_t find_actor, void *opaque) + { +Index: linux-2.4.21-chaos/kernel/ksyms.c +=================================================================== +--- linux-2.4.21-chaos.orig/kernel/ksyms.c 2003-12-12 16:18:36.000000000 +0300 ++++ linux-2.4.21-chaos/kernel/ksyms.c 2003-12-12 16:19:25.000000000 +0300 +@@ -178,6 +178,7 @@ + EXPORT_SYMBOL(igrab); + EXPORT_SYMBOL(iunique); + EXPORT_SYMBOL(iget4_locked); ++EXPORT_SYMBOL(ilookup4); + EXPORT_SYMBOL(unlock_new_inode); + EXPORT_SYMBOL(iput); + EXPORT_SYMBOL(inode_init_once); +@@ -191,6 +192,7 @@ + EXPORT_SYMBOL(path_release); + EXPORT_SYMBOL(__user_walk); + EXPORT_SYMBOL(lookup_one_len); ++EXPORT_SYMBOL(lookup_one_len_it); + EXPORT_SYMBOL(lookup_hash); + + EXPORT_SYMBOL(sys_close); diff --git a/lustre/kernel_patches/patches/tcp-zero-copy-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/tcp-zero-copy-2.4.21-rh-as.patch new file mode 100644 index 0000000..41ef631 --- /dev/null +++ b/lustre/kernel_patches/patches/tcp-zero-copy-2.4.21-rh-as.patch @@ -0,0 +1,477 @@ + include/linux/skbuff.h | 30 +++++ + include/net/tcp.h | 5 + net/core/skbuff.c | 25 ++++ + net/ipv4/tcp.c | 252 ++++++++++++++++++++++++++++++++++++++++++++++++- + net/netsyms.c | 2 + 5 files changed, 311 insertions(+), 3 deletions(-) + +Index: linux-2.4.21-4.EL/include/linux/skbuff.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/skbuff.h 2004-06-24 15:52:05.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/skbuff.h 2004-06-24 16:31:28.000000000 +0800 +@@ -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-2.4.21-4.EL/include/net/tcp.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/net/tcp.h 2004-06-24 15:52:05.000000000 +0800 ++++ linux-2.4.21-4.EL/include/net/tcp.h 2004-06-24 16:32:30.000000000 +0800 +@@ -636,6 +636,8 @@ + + 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, +@@ -730,6 +732,10 @@ + 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_kvec_read(struct sock *sk, kvec_cb_t cb, int len); + extern int tcp_kvec_write(struct sock *sk, kvec_cb_t cb, int len); + +Index: linux-2.4.21-4.EL/net/core/skbuff.c +=================================================================== +--- linux-2.4.21-4.EL.orig/net/core/skbuff.c 2003-10-04 05:28:51.000000000 +0800 ++++ linux-2.4.21-4.EL/net/core/skbuff.c 2004-06-24 16:31:28.000000000 +0800 +@@ -210,6 +210,8 @@ + 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: +@@ -280,6 +282,10 @@ + { + 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++) +@@ -550,6 +556,8 @@ + 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; +@@ -596,6 +604,14 @@ + 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; + +@@ -638,6 +654,8 @@ + 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(); +@@ -659,6 +677,11 @@ + 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; +@@ -673,6 +696,8 @@ + 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-2.4.21-4.EL/net/ipv4/tcp.c +=================================================================== +--- linux-2.4.21-4.EL.orig/net/ipv4/tcp.c 2003-10-04 05:28:43.000000000 +0800 ++++ linux-2.4.21-4.EL/net/ipv4/tcp.c 2004-06-24 16:31:28.000000000 +0800 +@@ -1015,7 +1015,7 @@ + goto out; + } + +-ssize_t do_tcp_sendpages(struct sock *sk, struct kveclet *let, int poffset, size_t psize, int flags); ++ssize_t do_tcp_sendpages(struct sock *sk, struct kveclet *let, 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) +@@ -1094,7 +1094,7 @@ + return err; + } + +-ssize_t do_tcp_sendpages(struct sock *sk, struct kveclet *let, int poffset, size_t psize, int flags) ++ssize_t do_tcp_sendpages(struct sock *sk, struct kveclet *let, int poffset, size_t psize, int flags, zccd_t *zccd) + { + struct tcp_opt *tp = &(sk->tp_pinfo.af_tcp); + int mss_now; +@@ -1147,6 +1147,17 @@ + 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) { +@@ -1157,6 +1168,20 @@ + 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; +@@ -1224,12 +1249,32 @@ + + lock_sock(sk); + TCP_CHECK_TIMER(sk); +- res = do_tcp_sendpages(sk, &let, 0, size, flags); ++ res = do_tcp_sendpages(sk, &let, 0, 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) ++{ ++ struct kveclet let = { page, offset, size }; ++ ssize_t res; ++ struct sock *sk = sock->sk; ++ ++ 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, &let, 0, size, flags, zccd); ++ TCP_CHECK_TIMER(sk); ++ release_sock(sk); ++ return res; ++} ++ + static void tcp_kvec_write_worker(struct tcp_write_async_info *info) + { + struct sock *sk = info->sk; +@@ -1238,7 +1283,7 @@ + !(sk->route_caps & TCP_ZC_CSUM_FLAGS)) + BUG(); + +- res = do_tcp_sendpages(sk, info->cur_let, info->offset, info->len - info->done, MSG_DONTWAIT); ++ res = do_tcp_sendpages(sk, info->cur_let, info->offset, info->len - info->done, MSG_DONTWAIT, NULL); + if (res > 0) + info->done += res; + +@@ -2102,6 +2147,202 @@ + 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 +Index: linux-2.4.21-4.EL/net/netsyms.c +=================================================================== +--- linux-2.4.21-4.EL.orig/net/netsyms.c 2003-10-04 05:28:43.000000000 +0800 ++++ linux-2.4.21-4.EL/net/netsyms.c 2004-06-24 16:31:28.000000000 +0800 +@@ -424,6 +424,8 @@ + EXPORT_SYMBOL(ip_generic_getfrag); + + #endif ++EXPORT_SYMBOL(tcp_sendpage_zccd); ++EXPORT_SYMBOL(tcp_recvpackets); + EXPORT_SYMBOL(tcp_read_sock); + + EXPORT_SYMBOL(netlink_set_err); diff --git a/lustre/kernel_patches/patches/vfs-pdirops-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/vfs-pdirops-2.4.21-rh-as.patch new file mode 100644 index 0000000..ba7dd66 --- /dev/null +++ b/lustre/kernel_patches/patches/vfs-pdirops-2.4.21-rh-as.patch @@ -0,0 +1,269 @@ + fs/inode.c | 1 + fs/namei.c | 66 ++++++++++++++++++++++++++++++++++++++--------------- + include/linux/fs.h | 11 ++++---- + 3 files changed, 54 insertions(+), 24 deletions(-) + +Index: linux-2.4.21-4.EL/fs/namei.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/namei.c 2004-06-24 03:44:43.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/namei.c 2004-06-24 03:47:12.000000000 +0800 +@@ -101,6 +101,36 @@ + + } + ++static void *lock_dir(struct inode *dir, struct qstr *name) ++{ ++ unsigned long hash; ++ ++ if (!IS_PDIROPS(dir)) { ++ down(&dir->i_sem); ++ return 0; ++ } ++ ++ /* OK. fs understands parallel directory operations. ++ * so, we try to acquire lock for hash of requested ++ * filename in order to prevent any operations with ++ * same name in same time -bzzz */ ++ ++ /* calculate name hash */ ++ hash = full_name_hash(name->name, name->len); ++ ++ /* lock this hash */ ++ return dynlock_lock(&dir->i_dcache_lock, hash, 1, GFP_ATOMIC); ++} ++ ++static void unlock_dir(struct inode *dir, void *lock) ++{ ++ if (!IS_PDIROPS(dir)) { ++ up(&dir->i_sem); ++ return; ++ } ++ dynlock_unlock(&dir->i_dcache_lock, lock); ++} ++ + /* 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.. +@@ -303,10 +333,11 @@ + struct dentry * result; + struct inode *dir = parent->d_inode; + int counter = 0; ++ void *lock; + + again: + counter++; +- down(&dir->i_sem); ++ lock = lock_dir(dir, name); + /* + * First re-do the cached lookup just in case it was created + * while we waited for the directory semaphore.. +@@ -330,7 +361,7 @@ + else + result = dentry; + } +- up(&dir->i_sem); ++ unlock_dir(dir, lock); + return result; + } + +@@ -338,7 +369,7 @@ + * Uhhuh! Nasty case: the cache was re-populated while + * we waited on the semaphore. Need to revalidate. + */ +- up(&dir->i_sem); ++ unlock_dir(dir, lock); + if (result->d_op && result->d_op->d_revalidate) { + if (!result->d_op->d_revalidate(result, flags) && !d_invalidate(result)) { + dput(result); +@@ -1171,13 +1202,13 @@ + goto exit; + + dir = nd->dentry; +- down(&dir->d_inode->i_sem); ++ nd->lock = lock_dir(dir->d_inode, &nd->last); + dentry = lookup_hash_it(&nd->last, nd->dentry, it); + + do_last: + error = PTR_ERR(dentry); + if (IS_ERR(dentry)) { +- up(&dir->d_inode->i_sem); ++ unlock_dir(dir->d_inode, nd->lock); + goto exit; + } + +@@ -1187,7 +1218,7 @@ + if (!IS_POSIXACL(dir->d_inode)) + mode &= ~current->fs->umask; + error = vfs_create_it(dir->d_inode, dentry, mode, it); +- up(&dir->d_inode->i_sem); ++ unlock_dir(dir->d_inode, nd->lock); + dput(nd->dentry); + nd->dentry = dentry; + if (error) +@@ -1201,7 +1232,7 @@ + /* + * It already exists. + */ +- up(&dir->d_inode->i_sem); ++ unlock_dir(dir->d_inode, nd->lock); + + error = -EEXIST; + if (flag & O_EXCL) +@@ -1348,7 +1379,7 @@ + goto exit; + } + dir = nd->dentry; +- down(&dir->d_inode->i_sem); ++ nd->lock = lock_dir(dir->d_inode, &nd->last); + dentry = lookup_hash_it(&nd->last, nd->dentry, it); + putname(nd->last.name); + goto do_last; +@@ -1366,7 +1397,7 @@ + { + struct dentry *dentry; + +- down(&nd->dentry->d_inode->i_sem); ++ nd->lock = lock_dir(nd->dentry->d_inode, &nd->last); + dentry = ERR_PTR(-EEXIST); + if (nd->last_type != LAST_NORM) + goto fail; +@@ -1460,7 +1491,7 @@ + } + dput(dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + out2: + path_release(&nd); + out: +@@ -1528,7 +1559,7 @@ + error = vfs_mkdir(nd.dentry->d_inode, dentry, mode); + dput(dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + out2: + path_release(&nd); + out: +@@ -1638,14 +1669,14 @@ + if (error != -EOPNOTSUPP) + goto exit1; + } +- down(&nd.dentry->d_inode->i_sem); ++ nd.lock = lock_dir(nd.dentry->d_inode, &nd.last); + 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); + dput(dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + exit1: + path_release(&nd); + exit: +@@ -1704,7 +1735,7 @@ + if (error != -EOPNOTSUPP) + goto exit1; + } +- down(&nd.dentry->d_inode->i_sem); ++ nd.lock = lock_dir(nd.dentry->d_inode, &nd.last); + dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { +@@ -1715,7 +1746,7 @@ + exit2: + dput(dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + exit1: + path_release(&nd); + exit: +@@ -1789,7 +1820,7 @@ + error = vfs_symlink(nd.dentry->d_inode, dentry, from); + dput(dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + out2: + path_release(&nd); + out: +@@ -1885,7 +1916,7 @@ + error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry); + dput(new_dentry); + } +- up(&nd.dentry->d_inode->i_sem); ++ unlock_dir(nd.dentry->d_inode, nd.lock); + out_release: + path_release(&nd); + out: +Index: linux-2.4.21-4.EL/fs/inode.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/inode.c 2004-06-24 03:45:36.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/inode.c 2004-06-24 03:46:04.000000000 +0800 +@@ -121,6 +121,7 @@ + mapping->host = inode; + mapping->gfp_mask = GFP_HIGHUSER; + inode->i_mapping = mapping; ++ dynlock_init(&inode->i_dcache_lock); + } + return inode; + } +Index: linux-2.4.21-4.EL/include/linux/fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/fs.h 2004-06-24 03:45:33.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/fs.h 2004-06-24 03:46:04.000000000 +0800 +@@ -21,6 +21,7 @@ + #include + #include + #include ++#include + + #include + #include +@@ -137,6 +138,7 @@ + #define S_IMMUTABLE 16 /* Immutable file */ + #define S_DEAD 32 /* removed, but still open directory */ + #define S_NOQUOTA 64 /* Inode is not counted to quota */ ++#define S_PDIROPS 256 /* Parallel directory operations */ + + /* + * Note that nosuid etc flags are inode-specific: setting some file-system +@@ -164,6 +166,7 @@ + #define IS_NOATIME(inode) (__IS_FLG(inode, MS_NOATIME) || ((inode)->i_flags & S_NOATIME)) + #define IS_NODIRATIME(inode) __IS_FLG(inode, MS_NODIRATIME) + #define IS_POSIXACL(inode) __IS_FLG(inode, MS_POSIXACL) ++#define IS_PDIROPS(inode) __IS_FLG(inode, S_PDIROPS) + + #define IS_DEADDIR(inode) ((inode)->i_flags & S_DEAD) + +@@ -513,6 +516,7 @@ + atomic_t i_writecount; + unsigned int i_attr_flags; + __u32 i_generation; ++ struct dynlock i_dcache_lock; /* for parallel directory ops */ + union { + struct minix_inode_info minix_i; + struct ext2_inode_info ext2_i; +@@ -733,6 +737,7 @@ + unsigned int flags; + int last_type; + struct lookup_intent *intent; ++ void *lock; + }; + + /* +@@ -1716,12 +1721,6 @@ + return dget(dentry->d_parent); + } + +-static inline void unlock_dir(struct dentry *dir) +-{ +- up(&dir->d_inode->i_sem); +- dput(dir); +-} +- + /* + * Whee.. Deadlock country. Happily there are only two VFS + * operations that does this.. diff --git a/lustre/kernel_patches/patches/vfs_intent-2.4.21-rh-as.patch b/lustre/kernel_patches/patches/vfs_intent-2.4.21-rh-as.patch new file mode 100644 index 0000000..aab71fc --- /dev/null +++ b/lustre/kernel_patches/patches/vfs_intent-2.4.21-rh-as.patch @@ -0,0 +1,1852 @@ + 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-2.4.21-4.EL/fs/dcache.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/dcache.c 2003-10-04 05:28:50.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/dcache.c 2004-06-24 15:50:29.000000000 +0800 +@@ -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. +@@ -847,13 +854,19 @@ + * 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-2.4.21-4.EL/fs/exec.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/exec.c 2003-10-04 05:28:49.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/exec.c 2004-06-24 15:50:29.000000000 +0800 +@@ -118,8 +118,10 @@ + 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; + +@@ -131,7 +133,8 @@ + 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; +@@ -399,8 +402,10 @@ + 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; +@@ -412,7 +417,8 @@ + 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) { +@@ -424,6 +430,7 @@ + return file; + } + } ++ intent_release(&it); + path_release(&nd); + } + goto out; +@@ -1330,7 +1337,7 @@ + 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-2.4.21-4.EL/fs/namei.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/namei.c 2003-10-04 05:28:46.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/namei.c 2004-06-24 15:50:53.000000000 +0800 +@@ -94,6 +94,13 @@ + * XEmacs seems to be relying on it... + */ + ++void intent_release(struct lookup_intent *it) ++{ ++ if (it && it->it_op_release) ++ it->it_op_release(it); ++ ++} ++ + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the + * kernel data space before using them.. +@@ -260,10 +267,19 @@ + * Internal lookup() using the new generic dcache. + * SMP-safe + */ +-static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *cached_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * dentry = d_lookup(parent, name); + ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, flags, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ dentry = NULL; ++ } ++ return dentry; ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + if (!dentry->d_op->d_revalidate(dentry, flags) && !d_invalidate(dentry)) { + dput(dentry); +@@ -281,11 +297,15 @@ + * make sure that nobody added the entry to the dcache in the meantime.. + * SMP-safe + */ +-static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *real_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * result; + struct inode *dir = parent->d_inode; ++ int counter = 0; + ++again: ++ counter++; + down(&dir->i_sem); + /* + * First re-do the cached lookup just in case it was created +@@ -300,6 +320,9 @@ + result = ERR_PTR(-ENOMEM); + if (dentry) { + lock_kernel(); ++ if (dir->i_op->lookup_it) ++ result = dir->i_op->lookup_it(dir, dentry, it, flags); ++ else + result = dir->i_op->lookup(dir, dentry); + unlock_kernel(); + if (result) +@@ -321,6 +344,15 @@ + dput(result); + result = ERR_PTR(-ENOENT); + } ++ } else if (result->d_op && result->d_op->d_revalidate_it) { ++ if (!result->d_op->d_revalidate_it(result, flags, it) && ++ !d_invalidate(result)) { ++ dput(result); ++ if (counter > 10) ++ result = ERR_PTR(-ESTALE); ++ if (!IS_ERR(result)) ++ goto again; ++ } + } + return result; + } +@@ -332,7 +364,8 @@ + * Without that kind of total limit, nasty chains of consecutive + * symlinks can cause almost arbitrarily long lookups. + */ +-static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd) ++static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd, ++ struct lookup_intent *it) + { + int err; + if (current->link_count >= 8) +@@ -346,10 +379,12 @@ + current->link_count++; + current->total_link_count++; + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + err = dentry->d_inode->i_op->follow_link(dentry, nd); + current->link_count--; + return err; + loop: ++ intent_release(it); + path_release(nd); + return -ELOOP; + } +@@ -447,7 +482,8 @@ + * + * We expect 'base' to be positive and a directory. + */ +-int link_path_walk(const char * name, struct nameidata *nd) ++int link_path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it) + { + struct dentry *dentry; + struct inode *inode; +@@ -524,12 +560,12 @@ + 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; +@@ -547,7 +583,7 @@ + goto out_dput; + + if (inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, NULL); + dput(dentry); + if (err) + goto return_err; +@@ -563,7 +599,7 @@ + nd->dentry = dentry; + } + err = -ENOTDIR; +- if (!inode->i_op->lookup) ++ if (!inode->i_op->lookup && !inode->i_op->lookup_it) + break; + continue; + /* here ends the main loop */ +@@ -590,12 +626,12 @@ + if (err < 0) + break; + } +- dentry = cached_lookup(nd->dentry, &this, 0); ++ dentry = cached_lookup(nd->dentry, &this, 0, it); + if (!dentry) { + err = -EWOULDBLOCKIO; + if (atomic) + break; +- dentry = real_lookup(nd->dentry, &this, 0); ++ dentry = real_lookup(nd->dentry, &this, 0, it); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -605,7 +641,7 @@ + inode = dentry->d_inode; + if ((lookup_flags & LOOKUP_FOLLOW) + && inode && inode->i_op && inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, it); + dput(dentry); + if (err) + goto return_err; +@@ -619,7 +655,8 @@ + goto no_inode; + if (lookup_flags & LOOKUP_DIRECTORY) { + err = -ENOTDIR; +- if (!inode->i_op || !inode->i_op->lookup) ++ if (!inode->i_op || ++ (!inode->i_op->lookup && !inode->i_op->lookup_it)) + break; + } + goto return_base; +@@ -643,6 +680,25 @@ + * Check the cached dentry for staleness. + */ + dentry = nd->dentry; ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ err = -ESTALE; ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) { ++ struct dentry *new; ++ err = permission(dentry->d_parent->d_inode, ++ MAY_EXEC); ++ if (err) ++ break; ++ new = real_lookup(dentry->d_parent, ++ &dentry->d_name, 0, it); ++ if (IS_ERR(new)) { ++ err = PTR_ERR(new); ++ break; ++ } ++ d_invalidate(dentry); ++ dput(dentry); ++ nd->dentry = new; ++ } ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + err = -ESTALE; + if (!dentry->d_op->d_revalidate(dentry, 0)) { +@@ -656,15 +712,28 @@ + dput(dentry); + break; + } ++ if (err) ++ intent_release(it); + path_release(nd); + return_err: + return err; + } + ++int link_path_walk(const char * name, struct nameidata *nd) ++{ ++ return link_path_walk_it(name, nd, NULL); ++} ++ ++int path_walk_it(const char * name, struct nameidata *nd, struct lookup_intent *it) ++{ ++ current->total_link_count = 0; ++ return link_path_walk_it(name, nd, it); ++} ++ + int path_walk(const char * name, struct nameidata *nd) + { + current->total_link_count = 0; +- return link_path_walk(name, nd); ++ return link_path_walk_it(name, nd, NULL); + } + + /* SMP-safe */ +@@ -749,6 +818,17 @@ + } + + /* 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; +@@ -763,6 +843,7 @@ + { + 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); +@@ -777,7 +858,8 @@ + * 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; +@@ -800,13 +882,16 @@ + 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) +@@ -818,6 +903,12 @@ + 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) + { +@@ -839,7 +930,7 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash(&this, base); ++ return lookup_hash_it(&this, base, NULL); + access: + return ERR_PTR(-EACCES); + } +@@ -870,6 +961,23 @@ + 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. +@@ -967,7 +1075,8 @@ + 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; + +@@ -980,12 +1089,15 @@ + 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); +@@ -994,6 +1106,11 @@ + return error; + } + ++int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++{ ++ return vfs_create_it(dir, dentry, mode, NULL); ++} ++ + /* + * open_namei() + * +@@ -1008,7 +1125,8 @@ + * 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; +@@ -1018,11 +1136,14 @@ + + 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; +@@ -1032,6 +1153,10 @@ + /* + * 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; +@@ -1047,7 +1172,7 @@ + + 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); +@@ -1056,11 +1181,12 @@ + 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); + dput(nd->dentry); + nd->dentry = dentry; +@@ -1164,7 +1290,7 @@ + if (!error) { + DQUOT_INIT(inode); + +- error = do_truncate(dentry, 0); ++ error = do_truncate(dentry, 0, 1); + } + put_write_access(inode); + if (error) +@@ -1176,8 +1302,10 @@ + return 0; + + exit_dput: ++ intent_release(it); + dput(dentry); + exit: ++ intent_release(it); + path_release(nd); + return error; + +@@ -1196,7 +1324,10 @@ + * 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; +@@ -1218,13 +1349,20 @@ + } + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + putname(nd->last.name); + goto do_last; + } + ++int open_namei(const char *pathname, int flag, int mode, struct nameidata *nd) ++{ ++ return open_namei_it(pathname, flag, mode, nd, NULL); ++} ++ ++ + /* SMP-safe */ +-static struct dentry *lookup_create(struct nameidata *nd, int is_dir) ++static struct dentry *lookup_create(struct nameidata *nd, int is_dir, ++ struct lookup_intent *it) + { + struct dentry *dentry; + +@@ -1232,7 +1370,7 @@ + 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) +@@ -1288,7 +1426,20 @@ + 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)) +@@ -1310,6 +1461,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1357,7 +1509,18 @@ + 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)) +@@ -1366,6 +1529,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1466,8 +1630,16 @@ + 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); +@@ -1525,8 +1697,15 @@ + 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 */ +@@ -1593,15 +1772,27 @@ + 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); +@@ -1677,7 +1868,18 @@ + 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); +@@ -1721,7 +1923,7 @@ + * 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; +@@ -1800,7 +2002,7 @@ + } + + 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; + +@@ -1888,9 +2090,18 @@ + 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; +@@ -1906,16 +2117,16 @@ + 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); +@@ -1966,20 +2177,26 @@ + } + + static inline int +-__vfs_follow_link(struct nameidata *nd, const char *link) ++__vfs_follow_link(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) + { + int res = 0; + char *name; + if (IS_ERR(link)) + goto fail; + ++ if (it == NULL) ++ it = nd->intent; ++ else if (it != nd->intent) ++ printk("it != nd->intent: tell phil@clusterfs.com\n"); ++ + if (*link == '/') { + path_release(nd); + if (!walk_init_root(link, nd)) + /* weird __emul_prefix() stuff did it */ + goto out; + } +- res = link_path_walk(link, nd); ++ res = link_path_walk_it(link, nd, it); + out: + if (current->link_count || res || nd->last_type!=LAST_NORM) + return res; +@@ -2003,7 +2220,13 @@ + + 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 */ +@@ -2045,7 +2268,7 @@ + { + 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-2.4.21-4.EL/fs/namespace.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/namespace.c 2003-10-04 05:28:36.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/namespace.c 2004-06-24 15:50:29.000000000 +0800 +@@ -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-2.4.21-4.EL/fs/open.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/open.c 2003-10-04 05:28:39.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/open.c 2004-06-24 15:50:29.000000000 +0800 +@@ -19,6 +19,8 @@ + #include + + #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) ++extern int path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it); + + int vfs_statfs(struct super_block *sb, struct statfs *buf) + { +@@ -95,9 +97,10 @@ + write_unlock(&files->file_lock); + } + +-int do_truncate(struct dentry *dentry, loff_t length) ++int do_truncate(struct dentry *dentry, loff_t length, int called_from_open) + { + struct inode *inode = dentry->d_inode; ++ struct inode_operations *op = dentry->d_inode->i_op; + int error; + struct iattr newattrs; + +@@ -109,7 +112,13 @@ + 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; +@@ -120,12 +129,13 @@ + 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; +@@ -165,11 +175,13 @@ + 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; +@@ -217,7 +229,7 @@ + + 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: +@@ -262,11 +274,13 @@ + 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; +@@ -281,11 +295,25 @@ + 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); +@@ -306,12 +334,14 @@ + 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; +@@ -326,7 +356,20 @@ + 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; +@@ -349,6 +392,7 @@ + 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; +@@ -366,13 +410,14 @@ + 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); + } + +@@ -387,8 +432,9 @@ + { + 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; + +@@ -399,6 +445,7 @@ + set_fs_pwd(current->fs, nd.mnt, nd.dentry); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -438,9 +485,10 @@ + { + 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; + +@@ -456,39 +504,56 @@ + 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; +@@ -497,30 +562,14 @@ + 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; +@@ -540,6 +589,20 @@ + 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; +@@ -644,6 +707,7 @@ + { + int namei_flags, error; + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_OPEN }; + + namei_flags = flags; + if ((namei_flags+1) & O_ACCMODE) +@@ -651,14 +715,15 @@ + 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; +@@ -695,7 +760,9 @@ + } + + if (f->f_op && f->f_op->open) { ++ f->f_it = it; + error = f->f_op->open(inode,f); ++ f->f_it = NULL; + if (error) + goto cleanup_all; + } +@@ -707,6 +774,7 @@ + !(inode->i_mapping->a_ops->direct_IO || + inode->i_mapping->a_ops->direct_sector_IO))) + goto cleanup_all; ++ intent_release(it); + + return f; + +@@ -722,11 +790,17 @@ + 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-2.4.21-4.EL/fs/stat.c +=================================================================== +--- linux-2.4.21-4.EL.orig/fs/stat.c 2003-10-04 05:28:49.000000000 +0800 ++++ linux-2.4.21-4.EL/fs/stat.c 2004-06-24 15:50:29.000000000 +0800 +@@ -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; + } +@@ -143,13 +145,15 @@ + asmlinkage long sys_stat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -159,13 +163,15 @@ + asmlinkage long sys_newstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -180,13 +186,15 @@ + asmlinkage long sys_lstat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -197,13 +205,15 @@ + asmlinkage long sys_newlstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -224,7 +234,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_old_stat(dentry->d_inode, statbuf); + fput(f); +@@ -243,7 +253,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat(dentry->d_inode, statbuf); + fput(f); +@@ -265,7 +275,7 @@ + + error = -EINVAL; + if (inode->i_op && inode->i_op->readlink && +- !(error = do_revalidate(nd.dentry))) { ++ !(error = do_revalidate(nd.dentry, NULL))) { + UPDATE_ATIME(inode); + error = inode->i_op->readlink(nd.dentry, buf, bufsiz); + } +@@ -341,12 +351,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -356,12 +368,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -376,7 +390,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat64(dentry->d_inode, statbuf); + fput(f); +Index: linux-2.4.21-4.EL/include/linux/dcache.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/dcache.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/dcache.h 2004-06-24 15:50:29.000000000 +0800 +@@ -6,6 +6,51 @@ + #include + #include + #include ++#include ++ ++#define IT_OPEN 0x0001 ++#define IT_CREAT 0x0002 ++#define IT_READDIR 0x0004 ++#define IT_GETATTR 0x0008 ++#define IT_LOOKUP 0x0010 ++#define IT_UNLINK 0x0020 ++#define IT_GETXATTR 0x0040 ++#define IT_EXEC 0x0080 ++#define IT_PIN 0x0100 ++ ++#define IT_FL_LOCKED 0x0001 ++#define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */ ++ ++#define INTENT_MAGIC 0x19620323 ++ ++ ++struct lustre_intent_data { ++ int it_disposition; ++ int it_status; ++ __u64 it_lock_handle; ++ void *it_data; ++ int it_lock_mode; ++ int it_int_flags; ++}; ++struct lookup_intent { ++ int it_magic; ++ void (*it_op_release)(struct lookup_intent *); ++ int it_op; ++ int it_flags; ++ int it_create_mode; ++ union { ++ struct lustre_intent_data lustre; ++ } d; ++}; ++ ++static inline void intent_init(struct lookup_intent *it, int op, int flags) ++{ ++ memset(it, 0, sizeof(*it)); ++ it->it_magic = INTENT_MAGIC; ++ it->it_op = op; ++ it->it_flags = flags; ++} ++ + + /* + * linux/include/linux/dcache.h +@@ -96,8 +141,22 @@ + 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 +@@ -129,6 +188,7 @@ + * 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-2.4.21-4.EL/include/linux/fs.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/fs.h 2004-06-24 15:44:46.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/fs.h 2004-06-24 15:50:29.000000000 +0800 +@@ -73,6 +73,7 @@ + + #define FMODE_READ 1 + #define FMODE_WRITE 2 ++#define FMODE_EXEC 4 + + #define READ 0 + #define WRITE 1 +@@ -359,6 +360,9 @@ + #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 +@@ -496,6 +500,7 @@ + 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 */ +@@ -598,6 +603,7 @@ + + /* needed for tty driver, and maybe others */ + void *private_data; ++ struct lookup_intent *f_it; + + /* preallocated helper kiobuf to speedup O_DIRECT */ + struct kiobuf *f_iobuf; +@@ -726,6 +732,7 @@ + struct qstr last; + unsigned int flags; + int last_type; ++ struct lookup_intent *intent; + }; + + /* +@@ -846,7 +853,8 @@ + extern int vfs_link(struct dentry *, struct inode *, struct dentry *); + extern int vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_unlink(struct inode *, struct dentry *); +-extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); ++int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, ++ struct inode *new_dir, struct dentry *new_dentry); + + /* + * File types +@@ -920,21 +928,32 @@ + + 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); +@@ -1131,10 +1150,14 @@ + + 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 *); + +@@ -1425,6 +1448,7 @@ + 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 *)); +@@ -1436,6 +1460,8 @@ + 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 iput(struct inode *); +@@ -1599,6 +1625,8 @@ + + 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-2.4.21-4.EL/include/linux/fs_struct.h +=================================================================== +--- linux-2.4.21-4.EL.orig/include/linux/fs_struct.h 2003-10-04 05:29:47.000000000 +0800 ++++ linux-2.4.21-4.EL/include/linux/fs_struct.h 2004-06-24 15:50:29.000000000 +0800 +@@ -37,10 +37,12 @@ + 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); + } +@@ -60,10 +62,12 @@ + 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-2.4.21-4.EL/kernel/exit.c +=================================================================== +--- linux-2.4.21-4.EL.orig/kernel/exit.c 2003-10-04 05:28:50.000000000 +0800 ++++ linux-2.4.21-4.EL/kernel/exit.c 2004-06-24 15:50:29.000000000 +0800 +@@ -346,11 +346,14 @@ + { + /* 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-2.4.21-4.EL/kernel/fork.c +=================================================================== +--- linux-2.4.21-4.EL.orig/kernel/fork.c 2003-10-04 05:28:53.000000000 +0800 ++++ linux-2.4.21-4.EL/kernel/fork.c 2004-06-24 15:50:29.000000000 +0800 +@@ -470,10 +470,13 @@ + 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-2.4.21-4.EL/kernel/ksyms.c +=================================================================== +--- linux-2.4.21-4.EL.orig/kernel/ksyms.c 2004-06-24 15:44:46.000000000 +0800 ++++ linux-2.4.21-4.EL/kernel/ksyms.c 2004-06-24 15:50:29.000000000 +0800 +@@ -312,6 +312,7 @@ + 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); diff --git a/lustre/kernel_patches/series/rh-as-2.4.21-4.EL b/lustre/kernel_patches/series/rh-as-2.4.21-4.EL new file mode 100644 index 0000000..b91f153 --- /dev/null +++ b/lustre/kernel_patches/series/rh-as-2.4.21-4.EL @@ -0,0 +1,37 @@ +configurable-x86-stack-2.4.21-rh-as.patch +dev_read_only_2.4.21-rh-as.patch +exports_2.4.21-rh-as.patch +lustre_version.patch +vfs_intent-2.4.21-rh-as.patch +invalidate_show-2.4.20-rh.patch +export-truncate.patch +iod-stock-exports-2.4.21-rh-as.patch +ext3-htree-2.4.21-rh-as.patch +linux-2.4.21-xattr-0.8.54-chaos.patch +ext3_orphan_lock-2.4.20-rh.patch +ext3-noread-2.4.21-rh-as.patch +ext3_delete_thread_2.4.21_rh_as.patch +extN-wantedi-2.4.21-rh-as.patch +ext3-san-2.4.20.patch +ext3-map_inode_page.patch +ext3-error-export.patch +iopen-2.4.20.patch +tcp-zero-copy-2.4.21-rh-as.patch +jbd-dont-account-blocks-twice.patch +jbd-commit-tricks.patch +ext3-o_direct-2.4.21-chaos.patch +ext3-no-write-super-chaos.patch +dynamic-locks-2.4.21-rh-as.patch +vfs-pdirops-2.4.21-rh-as.patch +ext3-pdirops-2.4.20-rh.patch +add_page_private.patch +nfs_export_kernel-2.4.21-rh-as.patch +ext3-ea-in-inode-2.4.21-rh-as.patch +listman-2.4.20.patch +kernel_text_address-2.4.18-chaos.patch +gfp_memalloc-2.4.21-rh-as.patch +ext3-xattr-ptr-arith-fix.patch +pagecache-lock-2.4.21-chaos.patch +ext3-truncate-buffer-head.patch +inode-max-readahead-2.4.24.patch +dcache_refcount_debug.patch