From: adilger Date: Sun, 16 Nov 2003 22:08:54 +0000 (+0000) Subject: New patch files for SuSE 2.4.19 ia64 kernel as used by SDSC. Mostly cosmetic X-Git-Tag: v1_7_0_51~2^7~230 X-Git-Url: https://git.whamcloud.com/?a=commitdiff_plain;h=86d9c609239b39d741074b3918a4b4811d503a70;p=fs%2Flustre-release.git New patch files for SuSE 2.4.19 ia64 kernel as used by SDSC. Mostly cosmetic except vfs_intent still used the old-style path_init()/path_walk(), and it already had EA patches applied. --- diff --git a/lustre/kernel_patches/patches/exports_2.4.19-suse.patch b/lustre/kernel_patches/patches/exports_2.4.19-suse.patch new file mode 100644 index 0000000..feaeec6 --- /dev/null +++ b/lustre/kernel_patches/patches/exports_2.4.19-suse.patch @@ -0,0 +1,53 @@ + fs/ext3/Makefile | 2 ++ + fs/ext3/super.c | 2 +- + include/linux/fs.h | 1 + + kernel/ksyms.c | 4 ++++ + 4 files changed, 8 insertions(+), 1 deletion(-) + +--- linux/fs/ext3/Makefile~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/fs/ext3/Makefile Wed Apr 9 10:19:53 2003 +@@ -9,6 +9,8 @@ + + O_TARGET := ext3.o + ++export-objs := super.o inode.o ++ + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) +--- linux/fs/ext3/super.c~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/fs/ext3/super.c Wed Apr 9 10:19:53 2003 +@@ -1769,7 +1769,7 @@ static void __exit exit_ext3_fs(void) + unregister_filesystem(&ext3_fs_type); + } + +-EXPORT_NO_SYMBOLS; ++EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +--- linux/include/linux/fs.h~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/include/linux/fs.h Wed Apr 9 10:19:53 2003 +@@ -1020,6 +1020,7 @@ extern int unregister_filesystem(struct + extern struct vfsmount *kern_mount(struct file_system_type *); + extern int may_umount(struct vfsmount *); + extern long do_mount(char *, char *, char *, unsigned long, void *); ++struct vfsmount *do_kern_mount(const char *type, int flags, char *name, void *data); + + #define kern_umount mntput + +--- linux/kernel/ksyms.c~exports_2.4.20 Wed Apr 9 10:07:14 2003 ++++ linux-mmonroe/kernel/ksyms.c Wed Apr 9 10:19:53 2003 +@@ -308,6 +308,10 @@ EXPORT_SYMBOL(dcache_dir_fsync); + EXPORT_SYMBOL(dcache_readdir); + EXPORT_SYMBOL(dcache_dir_ops); + ++/* lustre */ ++EXPORT_SYMBOL(pagecache_lock_cacheline); ++EXPORT_SYMBOL(do_kern_mount); ++ + /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */ + EXPORT_SYMBOL(default_llseek); + EXPORT_SYMBOL(dentry_open); + +_ diff --git a/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch b/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch new file mode 100644 index 0000000..748671f --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-1-suse.patch @@ -0,0 +1,2528 @@ + fs/ext3/Makefile | 2 + fs/ext3/dir.c | 299 +++++++++ + fs/ext3/file.c | 3 + fs/ext3/hash.c | 215 ++++++ + fs/ext3/namei.c | 1387 ++++++++++++++++++++++++++++++++++++++++----- + fs/ext3/super.c | 7 + include/linux/ext3_fs.h | 85 ++ + include/linux/ext3_fs_sb.h | 2 + include/linux/ext3_jbd.h | 2 + include/linux/rbtree.h | 2 + lib/rbtree.c | 42 + + 11 files changed, 1886 insertions(+), 160 deletions(-) + +--- linux-2.4.20/fs/ext3/Makefile~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/Makefile Sat Apr 5 03:57:05 2003 +@@ -12,7 +12,7 @@ O_TARGET := ext3.o + export-objs := super.o inode.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o ++ ioctl.o namei.o super.o symlink.o hash.o + obj-m := $(O_TARGET) + + include $(TOPDIR)/Rules.make +--- linux-2.4.20/fs/ext3/dir.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/dir.c Sat Apr 5 03:56:31 2003 +@@ -21,12 +21,16 @@ + #include + #include + #include ++#include ++#include + + static unsigned char ext3_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK + }; + + static int ext3_readdir(struct file *, void *, filldir_t); ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir); + + struct file_operations ext3_dir_operations = { + read: generic_read_dir, +@@ -35,6 +39,17 @@ struct file_operations ext3_dir_operatio + fsync: ext3_sync_file, /* BKL held */ + }; + ++ ++static unsigned char get_dtype(struct super_block *sb, int filetype) ++{ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) || ++ (filetype >= EXT3_FT_MAX)) ++ return DT_UNKNOWN; ++ ++ return (ext3_filetype_table[filetype]); ++} ++ ++ + int ext3_check_dir_entry (const char * function, struct inode * dir, + struct ext3_dir_entry_2 * de, + struct buffer_head * bh, +@@ -79,6 +94,16 @@ static int ext3_readdir(struct file * fi + + sb = inode->i_sb; + ++ if (is_dx(inode)) { ++ err = ext3_dx_readdir(filp, dirent, filldir); ++ if (err != ERR_BAD_DX_DIR) ++ return err; ++ /* ++ * We don't set the inode dirty flag since it's not ++ * critical that it get flushed back to the disk. ++ */ ++ EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; ++ } + stored = 0; + bh = NULL; + offset = filp->f_pos & (sb->s_blocksize - 1); +@@ -162,18 +187,12 @@ revalidate: + * during the copy operation. + */ + unsigned long version = filp->f_version; +- unsigned char d_type = DT_UNKNOWN; + +- if (EXT3_HAS_INCOMPAT_FEATURE(sb, +- EXT3_FEATURE_INCOMPAT_FILETYPE) +- && de->file_type < EXT3_FT_MAX) +- d_type = +- ext3_filetype_table[de->file_type]; + error = filldir(dirent, de->name, + de->name_len, + filp->f_pos, + le32_to_cpu(de->inode), +- d_type); ++ get_dtype(sb, de->file_type)); + if (error) + break; + if (version != filp->f_version) +@@ -188,3 +207,269 @@ revalidate: + UPDATE_ATIME(inode); + return 0; + } ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * These functions convert from the major/minor hash to an f_pos ++ * value. ++ * ++ * Currently we only use major hash numer. This is unfortunate, but ++ * on 32-bit machines, the same VFS interface is used for lseek and ++ * llseek, so if we use the 64 bit offset, then the 32-bit versions of ++ * lseek/telldir/seekdir will blow out spectacularly, and from within ++ * the ext2 low-level routine, we don't know if we're being called by ++ * a 64-bit version of the system call or the 32-bit version of the ++ * system call. Worse yet, NFSv2 only allows for a 32-bit readdir ++ * cookie. Sigh. ++ */ ++#define hash2pos(major, minor) (major >> 1) ++#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff) ++#define pos2min_hash(pos) (0) ++ ++/* ++ * This structure holds the nodes of the red-black tree used to store ++ * the directory entry in hash order. ++ */ ++struct fname { ++ __u32 hash; ++ __u32 minor_hash; ++ rb_node_t rb_hash; ++ struct fname *next; ++ __u32 inode; ++ __u8 name_len; ++ __u8 file_type; ++ char name[0]; ++}; ++ ++/* ++ * This functoin implements a non-recursive way of freeing all of the ++ * nodes in the red-black tree. ++ */ ++static void free_rb_tree_fname(rb_root_t *root) ++{ ++ rb_node_t *n = root->rb_node; ++ rb_node_t *parent; ++ struct fname *fname; ++ ++ while (n) { ++ /* Do the node's children first */ ++ if ((n)->rb_left) { ++ n = n->rb_left; ++ continue; ++ } ++ if (n->rb_right) { ++ n = n->rb_right; ++ continue; ++ } ++ /* ++ * The node has no children; free it, and then zero ++ * out parent's link to it. Finally go to the ++ * beginning of the loop and try to free the parent ++ * node. ++ */ ++ parent = n->rb_parent; ++ fname = rb_entry(n, struct fname, rb_hash); ++ kfree(fname); ++ if (!parent) ++ root->rb_node = 0; ++ else if (parent->rb_left == n) ++ parent->rb_left = 0; ++ else if (parent->rb_right == n) ++ parent->rb_right = 0; ++ n = parent; ++ } ++ root->rb_node = 0; ++} ++ ++ ++struct dir_private_info *create_dir_info(loff_t pos) ++{ ++ struct dir_private_info *p; ++ ++ p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL); ++ if (!p) ++ return NULL; ++ p->root.rb_node = 0; ++ p->curr_node = 0; ++ p->extra_fname = 0; ++ p->last_pos = 0; ++ p->curr_hash = pos2maj_hash(pos); ++ p->curr_minor_hash = pos2min_hash(pos); ++ p->next_hash = 0; ++ return p; ++} ++ ++void ext3_htree_free_dir_info(struct dir_private_info *p) ++{ ++ free_rb_tree_fname(&p->root); ++ kfree(p); ++} ++ ++/* ++ * Given a directory entry, enter it into the fname rb tree. ++ */ ++void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent) ++{ ++ rb_node_t **p, *parent = NULL; ++ struct fname * fname, *new_fn; ++ struct dir_private_info *info; ++ int len; ++ ++ info = (struct dir_private_info *) dir_file->private_data; ++ p = &info->root.rb_node; ++ ++ /* Create and allocate the fname structure */ ++ len = sizeof(struct fname) + dirent->name_len + 1; ++ new_fn = kmalloc(len, GFP_KERNEL); ++ memset(new_fn, 0, len); ++ new_fn->hash = hash; ++ new_fn->minor_hash = minor_hash; ++ new_fn->inode = le32_to_cpu(dirent->inode); ++ new_fn->name_len = dirent->name_len; ++ new_fn->file_type = dirent->file_type; ++ memcpy(new_fn->name, dirent->name, dirent->name_len); ++ new_fn->name[dirent->name_len] = 0; ++ ++ while (*p) { ++ parent = *p; ++ fname = rb_entry(parent, struct fname, rb_hash); ++ ++ /* ++ * If the hash and minor hash match up, then we put ++ * them on a linked list. This rarely happens... ++ */ ++ if ((new_fn->hash == fname->hash) && ++ (new_fn->minor_hash == fname->minor_hash)) { ++ new_fn->next = fname->next; ++ fname->next = new_fn; ++ return; ++ } ++ ++ if (new_fn->hash < fname->hash) ++ p = &(*p)->rb_left; ++ else if (new_fn->hash > fname->hash) ++ p = &(*p)->rb_right; ++ else if (new_fn->minor_hash < fname->minor_hash) ++ p = &(*p)->rb_left; ++ else /* if (new_fn->minor_hash > fname->minor_hash) */ ++ p = &(*p)->rb_right; ++ } ++ ++ rb_link_node(&new_fn->rb_hash, parent, p); ++ rb_insert_color(&new_fn->rb_hash, &info->root); ++} ++ ++ ++ ++/* ++ * This is a helper function for ext3_dx_readdir. It calls filldir ++ * for all entres on the fname linked list. (Normally there is only ++ * one entry on the linked list, unless there are 62 bit hash collisions.) ++ */ ++static int call_filldir(struct file * filp, void * dirent, ++ filldir_t filldir, struct fname *fname) ++{ ++ struct dir_private_info *info = filp->private_data; ++ loff_t curr_pos; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct super_block * sb; ++ int error; ++ ++ sb = inode->i_sb; ++ ++ if (!fname) { ++ printk("call_filldir: called with null fname?!?\n"); ++ return 0; ++ } ++ curr_pos = hash2pos(fname->hash, fname->minor_hash); ++ while (fname) { ++ error = filldir(dirent, fname->name, ++ fname->name_len, curr_pos, ++ fname->inode, ++ get_dtype(sb, fname->file_type)); ++ if (error) { ++ filp->f_pos = curr_pos; ++ info->extra_fname = fname->next; ++ return error; ++ } ++ fname = fname->next; ++ } ++ return 0; ++} ++ ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir) ++{ ++ struct dir_private_info *info = filp->private_data; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct fname *fname; ++ int ret; ++ ++ if (!info) { ++ info = create_dir_info(filp->f_pos); ++ if (!info) ++ return -ENOMEM; ++ filp->private_data = info; ++ } ++ ++ /* Some one has messed with f_pos; reset the world */ ++ if (info->last_pos != filp->f_pos) { ++ free_rb_tree_fname(&info->root); ++ info->curr_node = 0; ++ info->extra_fname = 0; ++ info->curr_hash = pos2maj_hash(filp->f_pos); ++ info->curr_minor_hash = pos2min_hash(filp->f_pos); ++ } ++ ++ /* ++ * If there are any leftover names on the hash collision ++ * chain, return them first. ++ */ ++ if (info->extra_fname && ++ call_filldir(filp, dirent, filldir, info->extra_fname)) ++ goto finished; ++ ++ if (!info->curr_node) ++ info->curr_node = rb_get_first(&info->root); ++ ++ while (1) { ++ /* ++ * Fill the rbtree if we have no more entries, ++ * or the inode has changed since we last read in the ++ * cached entries. ++ */ ++ if ((!info->curr_node) || ++ (filp->f_version != inode->i_version)) { ++ info->curr_node = 0; ++ free_rb_tree_fname(&info->root); ++ filp->f_version = inode->i_version; ++ ret = ext3_htree_fill_tree(filp, info->curr_hash, ++ info->curr_minor_hash, ++ &info->next_hash); ++ if (ret < 0) ++ return ret; ++ if (ret == 0) ++ break; ++ info->curr_node = rb_get_first(&info->root); ++ } ++ ++ fname = rb_entry(info->curr_node, struct fname, rb_hash); ++ info->curr_hash = fname->hash; ++ info->curr_minor_hash = fname->minor_hash; ++ if (call_filldir(filp, dirent, filldir, fname)) ++ break; ++ ++ info->curr_node = rb_get_next(info->curr_node); ++ if (!info->curr_node) { ++ info->curr_hash = info->next_hash; ++ info->curr_minor_hash = 0; ++ } ++ } ++finished: ++ info->last_pos = filp->f_pos; ++ UPDATE_ATIME(inode); ++ return 0; ++} ++#endif +--- linux-2.4.20/fs/ext3/file.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/file.c Sat Apr 5 03:56:31 2003 +@@ -35,6 +35,9 @@ static int ext3_release_file (struct ino + { + if (filp->f_mode & FMODE_WRITE) + ext3_discard_prealloc (inode); ++ if (is_dx(inode) && filp->private_data) ++ ext3_htree_free_dir_info(filp->private_data); ++ + return 0; + } + +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext3/hash.c Sat Apr 5 03:56:31 2003 +@@ -0,0 +1,215 @@ ++/* ++ * linux/fs/ext3/hash.c ++ * ++ * Copyright (C) 2002 by Theodore Ts'o ++ * ++ * This file is released under the GPL v2. ++ * ++ * This file may be redistributed under the terms of the GNU Public ++ * License. ++ */ ++ ++#include ++#include ++#include ++#include ++ ++#define DELTA 0x9E3779B9 ++ ++static void TEA_transform(__u32 buf[4], __u32 const in[]) ++{ ++ __u32 sum = 0; ++ __u32 b0 = buf[0], b1 = buf[1]; ++ __u32 a = in[0], b = in[1], c = in[2], d = in[3]; ++ int n = 16; ++ ++ do { ++ sum += DELTA; ++ b0 += ((b1 << 4)+a) ^ (b1+sum) ^ ((b1 >> 5)+b); ++ b1 += ((b0 << 4)+c) ^ (b0+sum) ^ ((b0 >> 5)+d); ++ } while(--n); ++ ++ buf[0] += b0; ++ buf[1] += b1; ++} ++ ++/* F, G and H are basic MD4 functions: selection, majority, parity */ ++#define F(x, y, z) ((z) ^ ((x) & ((y) ^ (z)))) ++#define G(x, y, z) (((x) & (y)) + (((x) ^ (y)) & (z))) ++#define H(x, y, z) ((x) ^ (y) ^ (z)) ++ ++/* ++ * The generic round function. The application is so specific that ++ * we don't bother protecting all the arguments with parens, as is generally ++ * good macro practice, in favor of extra legibility. ++ * Rotation is separate from addition to prevent recomputation ++ */ ++#define ROUND(f, a, b, c, d, x, s) \ ++ (a += f(b, c, d) + x, a = (a << s) | (a >> (32-s))) ++#define K1 0 ++#define K2 013240474631UL ++#define K3 015666365641UL ++ ++/* ++ * Basic cut-down MD4 transform. Returns only 32 bits of result. ++ */ ++static void halfMD4Transform (__u32 buf[4], __u32 const in[]) ++{ ++ __u32 a = buf[0], b = buf[1], c = buf[2], d = buf[3]; ++ ++ /* Round 1 */ ++ ROUND(F, a, b, c, d, in[0] + K1, 3); ++ ROUND(F, d, a, b, c, in[1] + K1, 7); ++ ROUND(F, c, d, a, b, in[2] + K1, 11); ++ ROUND(F, b, c, d, a, in[3] + K1, 19); ++ ROUND(F, a, b, c, d, in[4] + K1, 3); ++ ROUND(F, d, a, b, c, in[5] + K1, 7); ++ ROUND(F, c, d, a, b, in[6] + K1, 11); ++ ROUND(F, b, c, d, a, in[7] + K1, 19); ++ ++ /* Round 2 */ ++ ROUND(G, a, b, c, d, in[1] + K2, 3); ++ ROUND(G, d, a, b, c, in[3] + K2, 5); ++ ROUND(G, c, d, a, b, in[5] + K2, 9); ++ ROUND(G, b, c, d, a, in[7] + K2, 13); ++ ROUND(G, a, b, c, d, in[0] + K2, 3); ++ ROUND(G, d, a, b, c, in[2] + K2, 5); ++ ROUND(G, c, d, a, b, in[4] + K2, 9); ++ ROUND(G, b, c, d, a, in[6] + K2, 13); ++ ++ /* Round 3 */ ++ ROUND(H, a, b, c, d, in[3] + K3, 3); ++ ROUND(H, d, a, b, c, in[7] + K3, 9); ++ ROUND(H, c, d, a, b, in[2] + K3, 11); ++ ROUND(H, b, c, d, a, in[6] + K3, 15); ++ ROUND(H, a, b, c, d, in[1] + K3, 3); ++ ROUND(H, d, a, b, c, in[5] + K3, 9); ++ ROUND(H, c, d, a, b, in[0] + K3, 11); ++ ROUND(H, b, c, d, a, in[4] + K3, 15); ++ ++ buf[0] += a; ++ buf[1] += b; ++ buf[2] += c; ++ buf[3] += d; ++} ++ ++#undef ROUND ++#undef F ++#undef G ++#undef H ++#undef K1 ++#undef K2 ++#undef K3 ++ ++/* The old legacy hash */ ++static __u32 dx_hack_hash (const char *name, int len) ++{ ++ __u32 hash0 = 0x12a3fe2d, hash1 = 0x37abe8f9; ++ while (len--) { ++ __u32 hash = hash1 + (hash0 ^ (*name++ * 7152373)); ++ ++ if (hash & 0x80000000) hash -= 0x7fffffff; ++ hash1 = hash0; ++ hash0 = hash; ++ } ++ return (hash0 << 1); ++} ++ ++static void str2hashbuf(const char *msg, int len, __u32 *buf, int num) ++{ ++ __u32 pad, val; ++ int i; ++ ++ pad = (__u32)len | ((__u32)len << 8); ++ pad |= pad << 16; ++ ++ val = pad; ++ if (len > num*4) ++ len = num * 4; ++ for (i=0; i < len; i++) { ++ if ((i % 4) == 0) ++ val = pad; ++ val = msg[i] + (val << 8); ++ if ((i % 4) == 3) { ++ *buf++ = val; ++ val = pad; ++ num--; ++ } ++ } ++ if (--num >= 0) ++ *buf++ = val; ++ while (--num >= 0) ++ *buf++ = pad; ++} ++ ++/* ++ * Returns the hash of a filename. If len is 0 and name is NULL, then ++ * this function can be used to test whether or not a hash version is ++ * supported. ++ * ++ * The seed is an 4 longword (32 bits) "secret" which can be used to ++ * uniquify a hash. If the seed is all zero's, then some default seed ++ * may be used. ++ * ++ * A particular hash version specifies whether or not the seed is ++ * represented, and whether or not the returned hash is 32 bits or 64 ++ * bits. 32 bit hashes will return 0 for the minor hash. ++ */ ++int ext3fs_dirhash(const char *name, int len, struct dx_hash_info *hinfo) ++{ ++ __u32 hash; ++ __u32 minor_hash = 0; ++ const char *p; ++ int i; ++ __u32 in[8], buf[4]; ++ ++ /* Initialize the default seed for the hash checksum functions */ ++ buf[0] = 0x67452301; ++ buf[1] = 0xefcdab89; ++ buf[2] = 0x98badcfe; ++ buf[3] = 0x10325476; ++ ++ /* Check to see if the seed is all zero's */ ++ if (hinfo->seed) { ++ for (i=0; i < 4; i++) { ++ if (hinfo->seed[i]) ++ break; ++ } ++ if (i < 4) ++ memcpy(buf, hinfo->seed, sizeof(buf)); ++ } ++ ++ switch (hinfo->hash_version) { ++ case DX_HASH_LEGACY: ++ hash = dx_hack_hash(name, len); ++ break; ++ case DX_HASH_HALF_MD4: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 8); ++ halfMD4Transform(buf, in); ++ len -= 32; ++ p += 32; ++ } ++ minor_hash = buf[2]; ++ hash = buf[1]; ++ break; ++ case DX_HASH_TEA: ++ p = name; ++ while (len > 0) { ++ str2hashbuf(p, len, in, 4); ++ TEA_transform(buf, in); ++ len -= 16; ++ p += 16; ++ } ++ hash = buf[0]; ++ minor_hash = buf[1]; ++ break; ++ default: ++ hinfo->hash = 0; ++ return -1; ++ } ++ hinfo->hash = hash & ~1; ++ hinfo->minor_hash = minor_hash; ++ return 0; ++} +--- linux-2.4.20/fs/ext3/namei.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/namei.c Sat Apr 5 03:56:31 2003 +@@ -16,6 +16,12 @@ + * David S. Miller (davem@caip.rutgers.edu), 1995 + * Directory entry file type support and forward compatibility hooks + * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 ++ * Hash Tree Directory indexing (c) ++ * Daniel Phillips, 2001 ++ * Hash Tree Directory indexing porting ++ * Christopher Li, 2002 ++ * Hash Tree Directory indexing cleanup ++ * Theodore Ts'o, 2002 + */ + + #include +@@ -38,6 +44,630 @@ + #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) + #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + ++static struct buffer_head *ext3_append(handle_t *handle, ++ struct inode *inode, ++ u32 *block, int *err) ++{ ++ struct buffer_head *bh; ++ ++ *block = inode->i_size >> inode->i_sb->s_blocksize_bits; ++ ++ if ((bh = ext3_bread(handle, inode, *block, 1, err))) { ++ inode->i_size += inode->i_sb->s_blocksize; ++ EXT3_I(inode)->i_disksize = inode->i_size; ++ ext3_journal_get_write_access(handle,bh); ++ } ++ return bh; ++} ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#ifndef swap ++#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0) ++#endif ++ ++typedef struct { u32 v; } le_u32; ++typedef struct { u16 v; } le_u16; ++ ++#ifdef DX_DEBUG ++#define dxtrace(command) command ++#else ++#define dxtrace(command) ++#endif ++ ++struct fake_dirent ++{ ++ /*le*/u32 inode; ++ /*le*/u16 rec_len; ++ u8 name_len; ++ u8 file_type; ++}; ++ ++struct dx_countlimit ++{ ++ le_u16 limit; ++ le_u16 count; ++}; ++ ++struct dx_entry ++{ ++ le_u32 hash; ++ le_u32 block; ++}; ++ ++/* ++ * dx_root_info is laid out so that if it should somehow get overlaid by a ++ * dirent the two low bits of the hash version will be zero. Therefore, the ++ * hash version mod 4 should never be 0. Sincerely, the paranoia department. ++ */ ++ ++struct dx_root ++{ ++ struct fake_dirent dot; ++ char dot_name[4]; ++ struct fake_dirent dotdot; ++ char dotdot_name[4]; ++ struct dx_root_info ++ { ++ le_u32 reserved_zero; ++ u8 hash_version; ++ u8 info_length; /* 8 */ ++ u8 indirect_levels; ++ u8 unused_flags; ++ } ++ info; ++ struct dx_entry entries[0]; ++}; ++ ++struct dx_node ++{ ++ struct fake_dirent fake; ++ struct dx_entry entries[0]; ++}; ++ ++ ++struct dx_frame ++{ ++ struct buffer_head *bh; ++ struct dx_entry *entries; ++ struct dx_entry *at; ++}; ++ ++struct dx_map_entry ++{ ++ u32 hash; ++ u32 offs; ++}; ++ ++#ifdef CONFIG_EXT3_INDEX ++static inline unsigned dx_get_block (struct dx_entry *entry); ++static void dx_set_block (struct dx_entry *entry, unsigned value); ++static inline unsigned dx_get_hash (struct dx_entry *entry); ++static void dx_set_hash (struct dx_entry *entry, unsigned value); ++static unsigned dx_get_count (struct dx_entry *entries); ++static unsigned dx_get_limit (struct dx_entry *entries); ++static void dx_set_count (struct dx_entry *entries, unsigned value); ++static void dx_set_limit (struct dx_entry *entries, unsigned value); ++static unsigned dx_root_limit (struct inode *dir, unsigned infosize); ++static unsigned dx_node_limit (struct inode *dir); ++static struct dx_frame *dx_probe(struct dentry *dentry, ++ struct inode *dir, ++ struct dx_hash_info *hinfo, ++ struct dx_frame *frame, ++ int *err); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static struct ext3_dir_entry_2 *dx_move_dirents (char *from, char *to, ++ struct dx_map_entry *offsets, int count); ++static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash); ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err); ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode); ++ ++/* ++ * Future: use high four bits of block for coalesce-on-delete flags ++ * Mask them off for now. ++ */ ++ ++static inline unsigned dx_get_block (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->block.v) & 0x00ffffff; ++} ++ ++static inline void dx_set_block (struct dx_entry *entry, unsigned value) ++{ ++ entry->block.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_hash (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->hash.v); ++} ++ ++static inline void dx_set_hash (struct dx_entry *entry, unsigned value) ++{ ++ entry->hash.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_count (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->count.v); ++} ++ ++static inline unsigned dx_get_limit (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->limit.v); ++} ++ ++static inline void dx_set_count (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->count.v = cpu_to_le16(value); ++} ++ ++static inline void dx_set_limit (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->limit.v = cpu_to_le16(value); ++} ++ ++static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(1) - ++ EXT3_DIR_REC_LEN(2) - infosize; ++ return 0? 20: entry_space / sizeof(struct dx_entry); ++} ++ ++static inline unsigned dx_node_limit (struct inode *dir) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(0); ++ return 0? 22: entry_space / sizeof(struct dx_entry); ++} ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext3_dir_entry_2 *de, ++ int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ printk("names: "); ++ while ((char *) de < base + size) ++ { ++ if (de->inode) ++ { ++ if (show_names) ++ { ++ int len = de->name_len; ++ char *name = de->name; ++ while (len--) printk("%c", *name++); ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ printk(":%x.%u ", h.hash, ++ ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, ++ struct dx_entry *entries, int levels) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count = dx_get_count (entries), names = 0, space = 0, i; ++ unsigned bcount = 0; ++ struct buffer_head *bh; ++ int err; ++ printk("%i indexed blocks...\n", count); ++ for (i = 0; i < count; i++, entries++) ++ { ++ u32 block = dx_get_block(entries), hash = i? dx_get_hash(entries): 0; ++ u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; ++ struct stats stats; ++ printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); ++ if (!(bh = ext3_bread (NULL,dir, block, 0,&err))) continue; ++ stats = levels? ++ dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf(hinfo, (struct ext3_dir_entry_2 *) bh->b_data, blocksize, 0); ++ names += stats.names; ++ space += stats.space; ++ bcount += stats.bcount; ++ brelse (bh); ++ } ++ if (bcount) ++ printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", ++ names, space/bcount,(space/bcount)*100/blocksize); ++ return (struct stats) { names, space, bcount}; ++} ++#endif /* DX_DEBUG */ ++ ++/* ++ * Probe for a directory leaf block to search. ++ * ++ * dx_probe can return ERR_BAD_DX_DIR, which means there was a format ++ * error in the directory index, and the caller should fall back to ++ * searching the directory normally. The callers of dx_probe **MUST** ++ * check for this error code, and make sure it never gets reflected ++ * back to userspace. ++ */ ++static struct dx_frame * ++dx_probe(struct dentry *dentry, struct inode *dir, ++ struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) ++{ ++ unsigned count, indirect; ++ struct dx_entry *at, *entries, *p, *q, *m; ++ struct dx_root *root; ++ struct buffer_head *bh; ++ struct dx_frame *frame = frame_in; ++ u32 hash; ++ ++ frame->bh = NULL; ++ if (dentry) ++ dir = dentry->d_parent->d_inode; ++ if (!(bh = ext3_bread (NULL,dir, 0, 0, err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version != DX_HASH_TEA && ++ root->info.hash_version != DX_HASH_HALF_MD4 && ++ root->info.hash_version != DX_HASH_LEGACY) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unrecognised inode hash code %d", ++ root->info.hash_version); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ hinfo->hash_version = root->info.hash_version; ++ hinfo->seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ if (dentry) ++ ext3fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); ++ hash = hinfo->hash; ++ ++ if (root->info.unused_flags & 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash flags: %#06x", ++ root->info.unused_flags); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ if ((indirect = root->info.indirect_levels) > 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash depth: %#06x", ++ root->info.indirect_levels); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ entries = (struct dx_entry *) (((char *)&root->info) + ++ root->info.info_length); ++ assert(dx_get_limit(entries) == dx_root_limit(dir, ++ root->info.info_length)); ++ dxtrace (printk("Look up %x", hash)); ++ while (1) ++ { ++ count = dx_get_count(entries); ++ assert (count && count <= dx_get_limit(entries)); ++ p = entries + 1; ++ q = entries + count - 1; ++ while (p <= q) ++ { ++ m = p + (q - p)/2; ++ dxtrace(printk(".")); ++ if (dx_get_hash(m) > hash) ++ q = m - 1; ++ else ++ p = m + 1; ++ } ++ ++ if (0) // linear search cross check ++ { ++ unsigned n = count - 1; ++ at = entries; ++ while (n--) ++ { ++ dxtrace(printk(",")); ++ if (dx_get_hash(++at) > hash) ++ { ++ at--; ++ break; ++ } ++ } ++ assert (at == p - 1); ++ } ++ ++ at = p - 1; ++ dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); ++ frame->bh = bh; ++ frame->entries = entries; ++ frame->at = at; ++ if (!indirect--) return frame; ++ if (!(bh = ext3_bread (NULL,dir, dx_get_block(at), 0, err))) ++ goto fail2; ++ at = entries = ((struct dx_node *) bh->b_data)->entries; ++ assert (dx_get_limit(entries) == dx_node_limit (dir)); ++ frame++; ++ } ++fail2: ++ while (frame >= frame_in) { ++ brelse(frame->bh); ++ frame--; ++ } ++fail: ++ return NULL; ++} ++ ++static void dx_release (struct dx_frame *frames) ++{ ++ if (frames[0].bh == NULL) ++ return; ++ ++ if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) ++ brelse(frames[1].bh); ++ brelse(frames[0].bh); ++} ++ ++/* ++ * This function increments the frame pointer to search the next leaf ++ * block, and reads in the necessary intervening nodes if the search ++ * should be necessary. Whether or not the search is necessary is ++ * controlled by the hash parameter. If the hash value is even, then ++ * the search is only continued if the next block starts with that ++ * hash value. This is used if we are searching for a specific file. ++ * ++ * If the hash value is HASH_NB_ALWAYS, then always go to the next block. ++ * ++ * This function returns 1 if the caller should continue to search, ++ * or 0 if it should not. If there is an error reading one of the ++ * index blocks, it will return -1. ++ * ++ * If start_hash is non-null, it will be filled in with the starting ++ * hash of the next page. ++ */ ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash) ++{ ++ struct dx_frame *p; ++ struct buffer_head *bh; ++ int num_frames = 0; ++ __u32 bhash; ++ ++ *err = ENOENT; ++ p = frame; ++ /* ++ * Find the next leaf page by incrementing the frame pointer. ++ * If we run out of entries in the interior node, loop around and ++ * increment pointer in the parent node. When we break out of ++ * this loop, num_frames indicates the number of interior ++ * nodes need to be read. ++ */ ++ while (1) { ++ if (++(p->at) < p->entries + dx_get_count(p->entries)) ++ break; ++ if (p == frames) ++ return 0; ++ num_frames++; ++ p--; ++ } ++ ++ /* ++ * If the hash is 1, then continue only if the next page has a ++ * continuation hash of any value. This is used for readdir ++ * handling. Otherwise, check to see if the hash matches the ++ * desired contiuation hash. If it doesn't, return since ++ * there's no point to read in the successive index pages. ++ */ ++ bhash = dx_get_hash(p->at); ++ if (start_hash) ++ *start_hash = bhash; ++ if ((hash & 1) == 0) { ++ if ((bhash & ~1) != hash) ++ return 0; ++ } ++ /* ++ * If the hash is HASH_NB_ALWAYS, we always go to the next ++ * block so no check is necessary ++ */ ++ while (num_frames--) { ++ if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), ++ 0, err))) ++ return -1; /* Failure */ ++ p++; ++ brelse (p->bh); ++ p->bh = bh; ++ p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; ++ } ++ return 1; ++} ++ ++ ++/* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *p) ++{ ++ return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++/* ++ * This function fills a red-black tree with information from a ++ * directory. We start scanning the directory in hash order, starting ++ * at start_hash and start_minor_hash. ++ * ++ * This function returns the number of entries inserted into the tree, ++ * or a negative error code. ++ */ ++int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash) ++{ ++ struct dx_hash_info hinfo; ++ struct buffer_head *bh; ++ struct ext3_dir_entry_2 *de, *top; ++ static struct dx_frame frames[2], *frame; ++ struct inode *dir; ++ int block, err; ++ int count = 0; ++ int ret; ++ __u32 hashval; ++ ++ dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, ++ start_minor_hash)); ++ dir = dir_file->f_dentry->d_inode; ++ hinfo.hash = start_hash; ++ hinfo.minor_hash = 0; ++ frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ ++ while (1) { ++ block = dx_get_block(frame->at); ++ dxtrace(printk("Reading block %d\n", block)); ++ if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) ++ goto errout; ++ ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + dir->i_sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) { ++ ext3fs_dirhash(de->name, de->name_len, &hinfo); ++ if ((hinfo.hash < start_hash) || ++ ((hinfo.hash == start_hash) && ++ (hinfo.minor_hash < start_minor_hash))) ++ continue; ++ ext3_htree_store_dirent(dir_file, hinfo.hash, ++ hinfo.minor_hash, de); ++ count++; ++ } ++ brelse (bh); ++ hashval = ~1; ++ ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ++ frame, frames, &err, &hashval); ++ if (next_hash) ++ *next_hash = hashval; ++ if (ret == -1) ++ goto errout; ++ /* ++ * Stop if: (a) there are no more entries, or ++ * (b) we have inserted at least one entry and the ++ * next hash value is not a continuation ++ */ ++ if ((ret == 0) || ++ (count && ((hashval & 1) == 0))) ++ break; ++ } ++ dx_release(frames); ++ dxtrace(printk("Fill tree: returned %d entries\n", count)); ++ return count; ++errout: ++ dx_release(frames); ++ return (err); ++} ++ ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry *map_tail) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ while ((char *) de < base + size) ++ { ++ if (de->name_len && de->inode) { ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ map_tail--; ++ map_tail->hash = h.hash; ++ map_tail->offs = (u32) ((char *) de - base); ++ count++; ++ } ++ /* XXX: do we need to check rec_len == 0 case? -Chris */ ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ return count; ++} ++ ++static void dx_sort_map (struct dx_map_entry *map, unsigned count) ++{ ++ struct dx_map_entry *p, *q, *top = map + count - 1; ++ int more; ++ /* Combsort until bubble sort doesn't suck */ ++ while (count > 2) ++ { ++ count = count*10/13; ++ if (count - 9 < 2) /* 9, 10 -> 11 */ ++ count = 11; ++ for (p = top, q = p - count; q >= map; p--, q--) ++ if (p->hash < q->hash) ++ swap(*p, *q); ++ } ++ /* Garden variety bubble sort */ ++ do { ++ more = 0; ++ q = top; ++ while (q-- > map) ++ { ++ if (q[1].hash >= q[0].hash) ++ continue; ++ swap(*(q+1), *q); ++ more = 1; ++ } ++ } while(more); ++} ++ ++static void dx_insert_block(struct dx_frame *frame, u32 hash, u32 block) ++{ ++ struct dx_entry *entries = frame->entries; ++ struct dx_entry *old = frame->at, *new = old + 1; ++ int count = dx_get_count(entries); ++ ++ assert(count < dx_get_limit(entries)); ++ assert(old < entries + count); ++ memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); ++ dx_set_hash(new, hash); ++ dx_set_block(new, block); ++ dx_set_count(entries, count + 1); ++} ++#endif ++ ++ ++static void ext3_update_dx_flag(struct inode *inode) ++{ ++ if (!EXT3_HAS_COMPAT_FEATURE(inode->i_sb, ++ EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -94,6 +724,7 @@ static int inline search_dirblock(struct + return 0; + } + ++ + /* + * ext3_find_entry() + * +@@ -105,6 +736,8 @@ static int inline search_dirblock(struct + * The returned buffer_head has ->b_count elevated. The caller is expected + * to brelse() it when appropriate. + */ ++ ++ + static struct buffer_head * ext3_find_entry (struct dentry *dentry, + struct ext3_dir_entry_2 ** res_dir) + { +@@ -119,12 +752,32 @@ static struct buffer_head * ext3_find_en + int num = 0; + int nblocks, i, err; + struct inode *dir = dentry->d_parent->d_inode; ++ int namelen; ++ const u8 *name; ++ unsigned blocksize; + + *res_dir = NULL; + sb = dir->i_sb; +- ++ blocksize = sb->s_blocksize; ++ namelen = dentry->d_name.len; ++ name = dentry->d_name.name; ++ if (namelen > EXT3_NAME_LEN) ++ return NULL; ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ bh = ext3_dx_find_entry(dentry, res_dir, &err); ++ /* ++ * On success, or if the error was file not found, ++ * return. Otherwise, fall back to doing a search the ++ * old fashioned way. ++ */ ++ if (bh || (err != ERR_BAD_DX_DIR)) ++ return bh; ++ dxtrace(printk("ext3_find_entry: dx failed, falling back\n")); ++ } ++#endif + nblocks = dir->i_size >> EXT3_BLOCK_SIZE_BITS(sb); +- start = dir->u.ext3_i.i_dir_start_lookup; ++ start = EXT3_I(dir)->i_dir_start_lookup; + if (start >= nblocks) + start = 0; + block = start; +@@ -165,7 +818,7 @@ restart: + i = search_dirblock(bh, dir, dentry, + block << EXT3_BLOCK_SIZE_BITS(sb), res_dir); + if (i == 1) { +- dir->u.ext3_i.i_dir_start_lookup = block; ++ EXT3_I(dir)->i_dir_start_lookup = block; + ret = bh; + goto cleanup_and_exit; + } else { +@@ -196,6 +849,66 @@ cleanup_and_exit: + return ret; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err) ++{ ++ struct super_block * sb; ++ struct dx_hash_info hinfo; ++ u32 hash; ++ struct dx_frame frames[2], *frame; ++ struct ext3_dir_entry_2 *de, *top; ++ struct buffer_head *bh; ++ unsigned long block; ++ int retval; ++ int namelen = dentry->d_name.len; ++ const u8 *name = dentry->d_name.name; ++ struct inode *dir = dentry->d_parent->d_inode; ++ ++ sb = dir->i_sb; ++ if (!(frame = dx_probe (dentry, 0, &hinfo, frames, err))) ++ return NULL; ++ hash = hinfo.hash; ++ do { ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, err))) ++ goto errout; ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match (namelen, name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto errout; ++ } ++ *res_dir = de; ++ dx_release (frames); ++ return bh; ++ } ++ brelse (bh); ++ /* Check to see if we should continue to search */ ++ retval = ext3_htree_next_block(dir, hash, frame, ++ frames, err, 0); ++ if (retval == -1) { ++ ext3_warning(sb, __FUNCTION__, ++ "error reading index page in directory #%lu", ++ dir->i_ino); ++ goto errout; ++ } ++ } while (retval == 1); ++ ++ *err = -ENOENT; ++errout: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++} ++#endif ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -212,8 +925,9 @@ static struct dentry *ext3_lookup(struct + brelse (bh); + inode = iget(dir->i_sb, ino); + +- if (!inode) ++ if (!inode) { + return ERR_PTR(-EACCES); ++ } + } + d_add(dentry, inode); + return NULL; +@@ -237,6 +951,301 @@ static inline void ext3_set_de_type(stru + de->file_type = ext3_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct ext3_dir_entry_2 * ++dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count) ++{ ++ unsigned rec_len = 0; ++ ++ while (count--) { ++ struct ext3_dir_entry_2 *de = (struct ext3_dir_entry_2 *) (from + map->offs); ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ memcpy (to, de, rec_len); ++ ((struct ext3_dir_entry_2 *) to)->rec_len = rec_len; ++ de->inode = 0; ++ map++; ++ to += rec_len; ++ } ++ return (struct ext3_dir_entry_2 *) (to - rec_len); ++} ++ ++static struct ext3_dir_entry_2* dx_pack_dirents(char *base, int size) ++{ ++ struct ext3_dir_entry_2 *next, *to, *prev, *de = (struct ext3_dir_entry_2 *) base; ++ unsigned rec_len = 0; ++ ++ prev = to = de; ++ while ((char*)de < base + size) { ++ next = (struct ext3_dir_entry_2 *) ((char *) de + ++ le16_to_cpu(de->rec_len)); ++ if (de->inode && de->name_len) { ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ if (de > to) ++ memmove(to, de, rec_len); ++ to->rec_len = rec_len; ++ prev = to; ++ to = (struct ext3_dir_entry_2 *) (((char *) to) + rec_len); ++ } ++ de = next; ++ } ++ return prev; ++} ++ ++static struct ext3_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, ++ struct buffer_head **bh,struct dx_frame *frame, ++ struct dx_hash_info *hinfo, int *error) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count, continued; ++ struct buffer_head *bh2; ++ u32 newblock; ++ u32 hash2; ++ struct dx_map_entry *map; ++ char *data1 = (*bh)->b_data, *data2; ++ unsigned split; ++ struct ext3_dir_entry_2 *de = NULL, *de2; ++ int err; ++ ++ bh2 = ext3_append (handle, dir, &newblock, error); ++ if (!(bh2)) { ++ brelse(*bh); ++ *bh = NULL; ++ goto errout; ++ } ++ ++ BUFFER_TRACE(*bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, *bh); ++ if (err) { ++ journal_error: ++ brelse(*bh); ++ brelse(bh2); ++ *bh = NULL; ++ ext3_std_error(dir->i_sb, err); ++ goto errout; ++ } ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ ++ data2 = bh2->b_data; ++ ++ /* create map in the end of data2 block */ ++ map = (struct dx_map_entry *) (data2 + blocksize); ++ count = dx_make_map ((struct ext3_dir_entry_2 *) data1, ++ blocksize, hinfo, map); ++ map -= count; ++ split = count/2; // need to adjust to actual middle ++ dx_sort_map (map, count); ++ hash2 = map[split].hash; ++ continued = hash2 == map[split - 1].hash; ++ dxtrace(printk("Split block %i at %x, %i/%i\n", ++ dx_get_block(frame->at), hash2, split, count-split)); ++ ++ /* Fancy dance to stay within two buffers */ ++ de2 = dx_move_dirents(data1, data2, map + split, count - split); ++ de = dx_pack_dirents(data1,blocksize); ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ de2->rec_len = cpu_to_le16(data2 + blocksize - (char *) de2); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data1, blocksize, 1)); ++ dxtrace(dx_show_leaf (hinfo, (struct ext3_dir_entry_2 *) data2, blocksize, 1)); ++ ++ /* Which block gets the new entry? */ ++ if (hinfo->hash >= hash2) ++ { ++ swap(*bh, bh2); ++ de = de2; ++ } ++ dx_insert_block (frame, hash2 + continued, newblock); ++ err = ext3_journal_dirty_metadata (handle, bh2); ++ if (err) ++ goto journal_error; ++ err = ext3_journal_dirty_metadata (handle, frame->bh); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ dxtrace(dx_show_index ("frame", frame->entries)); ++errout: ++ return de; ++} ++#endif ++ ++ ++/* ++ * Add a new entry into a directory (leaf) block. If de is non-NULL, ++ * it points to a directory entry which is guaranteed to be large ++ * enough for new directory entry. If de is NULL, then ++ * add_dirent_to_buf will attempt search the directory block for ++ * space. It will return -ENOSPC if no space is available, and -EIO ++ * and -EEXIST if directory entry already exists. ++ * ++ * NOTE! bh is NOT released in the case where ENOSPC is returned. In ++ * all other cases bh is released. ++ */ ++static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct ext3_dir_entry_2 *de, ++ struct buffer_head * bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ unsigned long offset = 0; ++ unsigned short reclen; ++ int nlen, rlen, err; ++ char *top; ++ ++ reclen = EXT3_DIR_REC_LEN(namelen); ++ if (!de) { ++ de = (struct ext3_dir_entry_2 *)bh->b_data; ++ top = bh->b_data + dir->i_sb->s_blocksize - reclen; ++ while ((char *) de <= top) { ++ if (!ext3_check_dir_entry("ext3_add_entry", dir, de, ++ bh, offset)) { ++ brelse (bh); ++ return -EIO; ++ } ++ if (ext3_match (namelen, name, de)) { ++ brelse (bh); ++ return -EEXIST; ++ } ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if ((de->inode? rlen - nlen: rlen) >= reclen) ++ break; ++ de = (struct ext3_dir_entry_2 *)((char *)de + rlen); ++ offset += rlen; ++ } ++ if ((char *) de > top) ++ return -ENOSPC; ++ } ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) { ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return err; ++ } ++ ++ /* By now the buffer is marked for journaling */ ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if (de->inode) { ++ struct ext3_dir_entry_2 *de1 = (struct ext3_dir_entry_2 *)((char *)de + nlen); ++ de1->rec_len = cpu_to_le16(rlen - nlen); ++ de->rec_len = cpu_to_le16(nlen); ++ de = de1; ++ } ++ de->file_type = EXT3_FT_UNKNOWN; ++ if (inode) { ++ de->inode = cpu_to_le32(inode->i_ino); ++ ext3_set_de_type(dir->i_sb, de, inode->i_mode); ++ } else ++ de->inode = 0; ++ de->name_len = namelen; ++ memcpy (de->name, name, namelen); ++ /* ++ * XXX shouldn't update any times until successful ++ * completion of syscall, but too many callers depend ++ * on this. ++ * ++ * XXX similarly, too many callers depend on ++ * ext3_new_inode() setting the times, but error ++ * recovery deletes the inode, so the worst that can ++ * happen is that the times are slightly out of date ++ * and/or different from the directory change time. ++ */ ++ dir->i_mtime = dir->i_ctime = CURRENT_TIME; ++ ext3_update_dx_flag(dir); ++ dir->i_version = ++event; ++ ext3_mark_inode_dirty(handle, dir); ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) ++ ext3_std_error(dir->i_sb, err); ++ brelse(bh); ++ return 0; ++} ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * This converts a one block unindexed directory to a 3 block indexed ++ * directory, and adds the dentry to the indexed directory. ++ */ ++static int make_indexed_dir(handle_t *handle, struct dentry *dentry, ++ struct inode *inode, struct buffer_head *bh) ++{ ++ struct inode *dir = dentry->d_parent->d_inode; ++ const char *name = dentry->d_name.name; ++ int namelen = dentry->d_name.len; ++ struct buffer_head *bh2; ++ struct dx_root *root; ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries; ++ struct ext3_dir_entry_2 *de, *de2; ++ char *data1, *top; ++ unsigned len; ++ int retval; ++ unsigned blocksize; ++ struct dx_hash_info hinfo; ++ u32 block; ++ ++ blocksize = dir->i_sb->s_blocksize; ++ dxtrace(printk("Creating index\n")); ++ retval = ext3_journal_get_write_access(handle, bh); ++ if (retval) { ++ ext3_std_error(dir->i_sb, retval); ++ brelse(bh); ++ return retval; ++ } ++ root = (struct dx_root *) bh->b_data; ++ ++ EXT3_I(dir)->i_flags |= EXT3_INDEX_FL; ++ bh2 = ext3_append (handle, dir, &block, &retval); ++ if (!(bh2)) { ++ brelse(bh); ++ return retval; ++ } ++ data1 = bh2->b_data; ++ ++ /* The 0th block becomes the root, move the dirents out */ ++ de = (struct ext3_dir_entry_2 *) &root->dotdot; ++ de = (struct ext3_dir_entry_2 *) ((char *)de + de->rec_len); ++ len = ((char *) root) + blocksize - (char *) de; ++ memcpy (data1, de, len); ++ de = (struct ext3_dir_entry_2 *) data1; ++ top = data1 + len; ++ while (((char *) de2=(char*)de+le16_to_cpu(de->rec_len)) < top) ++ de = de2; ++ de->rec_len = cpu_to_le16(data1 + blocksize - (char *) de); ++ /* Initialize the root; the dot dirents already exist */ ++ de = (struct ext3_dir_entry_2 *) (&root->dotdot); ++ de->rec_len = cpu_to_le16(blocksize - EXT3_DIR_REC_LEN(2)); ++ memset (&root->info, 0, sizeof(root->info)); ++ root->info.info_length = sizeof(root->info); ++ root->info.hash_version = dir->i_sb->u.ext3_sb.s_def_hash_version; ++ entries = root->entries; ++ dx_set_block (entries, 1); ++ dx_set_count (entries, 1); ++ dx_set_limit (entries, dx_root_limit(dir, sizeof(root->info))); ++ ++ /* Initialize as for dx_probe */ ++ hinfo.hash_version = root->info.hash_version; ++ hinfo.seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ ext3fs_dirhash(name, namelen, &hinfo); ++ frame = frames; ++ frame->entries = entries; ++ frame->at = entries; ++ frame->bh = bh; ++ bh = bh2; ++ de = do_split(handle,dir, &bh, frame, &hinfo, &retval); ++ dx_release (frames); ++ if (!(de)) ++ return retval; ++ ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} ++#endif ++ + /* + * ext3_add_entry() + * +@@ -247,127 +1255,198 @@ static inline void ext3_set_de_type(stru + * may not sleep between calling this and putting something into + * the entry, as someone else might have used it while you slept. + */ +- +-/* +- * AKPM: the journalling code here looks wrong on the error paths +- */ + static int ext3_add_entry (handle_t *handle, struct dentry *dentry, + struct inode *inode) + { + struct inode *dir = dentry->d_parent->d_inode; +- const char *name = dentry->d_name.name; +- int namelen = dentry->d_name.len; + unsigned long offset; +- unsigned short rec_len; + struct buffer_head * bh; +- struct ext3_dir_entry_2 * de, * de1; ++ struct ext3_dir_entry_2 *de; + struct super_block * sb; + int retval; ++#ifdef CONFIG_EXT3_INDEX ++ int dx_fallback=0; ++#endif ++ unsigned blocksize; ++ unsigned nlen, rlen; ++ u32 block, blocks; + + sb = dir->i_sb; +- +- if (!namelen) ++ blocksize = sb->s_blocksize; ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &retval); ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ retval = ext3_dx_add_entry(handle, dentry, inode); ++ if (!retval || (retval != ERR_BAD_DX_DIR)) ++ return retval; ++ EXT3_I(dir)->i_flags &= ~EXT3_INDEX_FL; ++ dx_fallback++; ++ ext3_mark_inode_dirty(handle, dir); ++ } ++#endif ++ blocks = dir->i_size >> sb->s_blocksize_bits; ++ for (block = 0, offset = 0; block < blocks; block++) { ++ bh = ext3_bread(handle, dir, block, 0, &retval); ++ if(!bh) ++ return retval; ++ retval = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (retval != -ENOSPC) ++ return retval; ++ ++#ifdef CONFIG_EXT3_INDEX ++ if (blocks == 1 && !dx_fallback && ++ EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ return make_indexed_dir(handle, dentry, inode, bh); ++#endif ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); + if (!bh) + return retval; +- rec_len = EXT3_DIR_REC_LEN(namelen); +- offset = 0; + de = (struct ext3_dir_entry_2 *) bh->b_data; +- while (1) { +- if ((char *)de >= sb->s_blocksize + bh->b_data) { +- brelse (bh); +- bh = NULL; +- bh = ext3_bread (handle, dir, +- offset >> EXT3_BLOCK_SIZE_BITS(sb), 1, &retval); +- if (!bh) +- return retval; +- if (dir->i_size <= offset) { +- if (dir->i_size == 0) { +- brelse(bh); +- return -ENOENT; +- } ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = blocksize); ++ nlen = 0; ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} + +- ext3_debug ("creating next block\n"); ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * Returns 0 for success, or a negative error value ++ */ ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode) ++{ ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ struct dx_hash_info hinfo; ++ struct buffer_head * bh; ++ struct inode *dir = dentry->d_parent->d_inode; ++ struct super_block * sb = dir->i_sb; ++ struct ext3_dir_entry_2 *de; ++ int err; + +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- } else { ++ frame = dx_probe(dentry, 0, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ entries = frame->entries; ++ at = frame->at; + +- ext3_debug ("skipping to next block\n"); ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err))) ++ goto cleanup; + +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- } +- } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { +- brelse (bh); +- return -EEXIST; ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) ++ goto journal_error; ++ ++ err = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (err != -ENOSPC) { ++ bh = 0; ++ goto cleanup; ++ } ++ ++ /* Block full, should compress but for now just split */ ++ dxtrace(printk("using %u of %u node entries\n", ++ dx_get_count(entries), dx_get_limit(entries))); ++ /* Need to split index? */ ++ if (dx_get_count(entries) == dx_get_limit(entries)) { ++ u32 newblock; ++ unsigned icount = dx_get_count(entries); ++ int levels = frame - frames; ++ struct dx_entry *entries2; ++ struct dx_node *node2; ++ struct buffer_head *bh2; ++ ++ if (levels && (dx_get_count(frames->entries) == ++ dx_get_limit(frames->entries))) { ++ ext3_warning(sb, __FUNCTION__, ++ "Directory index full!\n"); ++ err = -ENOSPC; ++ goto cleanup; + } +- if ((le32_to_cpu(de->inode) == 0 && +- le16_to_cpu(de->rec_len) >= rec_len) || +- (le16_to_cpu(de->rec_len) >= +- EXT3_DIR_REC_LEN(de->name_len) + rec_len)) { +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- /* By now the buffer is marked for journaling */ +- offset += le16_to_cpu(de->rec_len); +- if (le32_to_cpu(de->inode)) { +- de1 = (struct ext3_dir_entry_2 *) ((char *) de + +- EXT3_DIR_REC_LEN(de->name_len)); +- de1->rec_len = +- cpu_to_le16(le16_to_cpu(de->rec_len) - +- EXT3_DIR_REC_LEN(de->name_len)); +- de->rec_len = cpu_to_le16( +- EXT3_DIR_REC_LEN(de->name_len)); +- de = de1; ++ bh2 = ext3_append (handle, dir, &newblock, &err); ++ if (!(bh2)) ++ goto cleanup; ++ node2 = (struct dx_node *)(bh2->b_data); ++ entries2 = node2->entries; ++ node2->fake.rec_len = cpu_to_le16(sb->s_blocksize); ++ node2->fake.inode = 0; ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ if (levels) { ++ unsigned icount1 = icount/2, icount2 = icount - icount1; ++ unsigned hash2 = dx_get_hash(entries + icount1); ++ dxtrace(printk("Split index %i/%i\n", icount1, icount2)); ++ ++ BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ ++ err = ext3_journal_get_write_access(handle, ++ frames[0].bh); ++ if (err) ++ goto journal_error; ++ ++ memcpy ((char *) entries2, (char *) (entries + icount1), ++ icount2 * sizeof(struct dx_entry)); ++ dx_set_count (entries, icount1); ++ dx_set_count (entries2, icount2); ++ dx_set_limit (entries2, dx_node_limit(dir)); ++ ++ /* Which index block gets the new entry? */ ++ if (at - entries >= icount1) { ++ frame->at = at = at - entries - icount1 + entries2; ++ frame->entries = entries = entries2; ++ swap(frame->bh, bh2); + } +- de->file_type = EXT3_FT_UNKNOWN; +- if (inode) { +- de->inode = cpu_to_le32(inode->i_ino); +- ext3_set_de_type(dir->i_sb, de, inode->i_mode); +- } else +- de->inode = 0; +- de->name_len = namelen; +- memcpy (de->name, name, namelen); +- /* +- * XXX shouldn't update any times until successful +- * completion of syscall, but too many callers depend +- * on this. +- * +- * XXX similarly, too many callers depend on +- * ext3_new_inode() setting the times, but error +- * recovery deletes the inode, so the worst that can +- * happen is that the times are slightly out of date +- * and/or different from the directory change time. +- */ +- dir->i_mtime = dir->i_ctime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- dir->i_version = ++event; +- ext3_mark_inode_dirty(handle, dir); +- BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +- ext3_journal_dirty_metadata(handle, bh); +- brelse(bh); +- return 0; ++ dx_insert_block (frames + 0, hash2, newblock); ++ dxtrace(dx_show_index ("node", frames[1].entries)); ++ dxtrace(dx_show_index ("node", ++ ((struct dx_node *) bh2->b_data)->entries)); ++ err = ext3_journal_dirty_metadata(handle, bh2); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ } else { ++ dxtrace(printk("Creating second level index...\n")); ++ memcpy((char *) entries2, (char *) entries, ++ icount * sizeof(struct dx_entry)); ++ dx_set_limit(entries2, dx_node_limit(dir)); ++ ++ /* Set up root */ ++ dx_set_count(entries, 1); ++ dx_set_block(entries + 0, newblock); ++ ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; ++ ++ /* Add new access path frame */ ++ frame = frames + 1; ++ frame->at = at = at - entries + entries2; ++ frame->entries = entries = entries2; ++ frame->bh = bh2; ++ err = ext3_journal_get_write_access(handle, ++ frame->bh); ++ if (err) ++ goto journal_error; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- brelse (bh); +- return -ENOSPC; ++ de = do_split(handle, dir, &bh, frame, &hinfo, &err); ++ if (!de) ++ goto cleanup; ++ err = add_dirent_to_buf(handle, dentry, inode, de, bh); ++ bh = 0; ++ goto cleanup; ++ ++journal_error: ++ ext3_std_error(dir->i_sb, err); ++cleanup: ++ if (bh) ++ brelse(bh); ++ dx_release(frames); ++ return err; + } ++#endif + + /* + * ext3_delete_entry deletes a directory entry by merging it with the +@@ -451,9 +1530,11 @@ static int ext3_create (struct inode * d + struct inode * inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -478,9 +1559,11 @@ static int ext3_mknod (struct inode * di + struct inode *inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -507,9 +1590,11 @@ static int ext3_mkdir(struct inode * dir + if (dir->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -521,7 +1606,7 @@ static int ext3_mkdir(struct inode * dir + + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; +- inode->i_size = inode->u.ext3_i.i_disksize = inode->i_sb->s_blocksize; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { +@@ -554,21 +1639,19 @@ static int ext3_mkdir(struct inode * dir + inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); +- if (err) +- goto out_no_entry; ++ if (err) { ++ inode->i_nlink = 0; ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + dir->i_nlink++; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + d_instantiate(dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- inode->i_nlink = 0; +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + /* +@@ -655,7 +1738,7 @@ int ext3_orphan_add(handle_t *handle, st + int err = 0, rc; + + lock_super(sb); +- if (!list_empty(&inode->u.ext3_i.i_orphan)) ++ if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + + /* Orphan handling is only valid for files with data blocks +@@ -696,7 +1779,7 @@ int ext3_orphan_add(handle_t *handle, st + * This is safe: on error we're going to ignore the orphan list + * anyway on the next recovery. */ + if (!err) +- list_add(&inode->u.ext3_i.i_orphan, &EXT3_SB(sb)->s_orphan); ++ list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); + + jbd_debug(4, "superblock will point to %ld\n", inode->i_ino); + jbd_debug(4, "orphan inode %ld will point to %d\n", +@@ -714,25 +1797,26 @@ out_unlock: + int ext3_orphan_del(handle_t *handle, struct inode *inode) + { + struct list_head *prev; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct ext3_sb_info *sbi; + ino_t ino_next; + struct ext3_iloc iloc; + int err = 0; + + lock_super(inode->i_sb); +- if (list_empty(&inode->u.ext3_i.i_orphan)) { ++ if (list_empty(&ei->i_orphan)) { + unlock_super(inode->i_sb); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); +- prev = inode->u.ext3_i.i_orphan.prev; ++ prev = ei->i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %ld from orphan list\n", inode->i_ino); + +- list_del(&inode->u.ext3_i.i_orphan); +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ list_del(&ei->i_orphan); ++ INIT_LIST_HEAD(&ei->i_orphan); + + /* If we're on an error path, we may not have a valid + * transaction handle with which to update the orphan list on +@@ -793,8 +1877,9 @@ static int ext3_rmdir (struct inode * di + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + retval = -ENOENT; + bh = ext3_find_entry (dentry, &de); +@@ -832,7 +1917,7 @@ static int ext3_rmdir (struct inode * di + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; + ext3_mark_inode_dirty(handle, inode); +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -850,8 +1935,9 @@ static int ext3_unlink(struct inode * di + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -878,7 +1964,7 @@ static int ext3_unlink(struct inode * di + if (retval) + goto end_unlink; + dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + inode->i_nlink--; + if (!inode->i_nlink) +@@ -904,9 +1990,11 @@ static int ext3_symlink (struct inode * + if (l > dir->i_sb->s_blocksize) + return -ENAMETOOLONG; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 5); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -916,7 +2004,7 @@ static int ext3_symlink (struct inode * + if (IS_ERR(inode)) + goto out_stop; + +- if (l > sizeof (inode->u.ext3_i.i_data)) { ++ if (l > sizeof (EXT3_I(inode)->i_data)) { + inode->i_op = &page_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* +@@ -925,25 +2013,23 @@ static int ext3_symlink (struct inode * + * i_size in generic_commit_write(). + */ + err = block_symlink(inode, symname, l); +- if (err) +- goto out_no_entry; ++ if (err) { ++ ext3_dec_count(handle, inode); ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + } else { + inode->i_op = &ext3_fast_symlink_inode_operations; +- memcpy((char*)&inode->u.ext3_i.i_data,symname,l); ++ memcpy((char*)&EXT3_I(inode)->i_data,symname,l); + inode->i_size = l-1; + } +- inode->u.ext3_i.i_disksize = inode->i_size; ++ EXT3_I(inode)->i_disksize = inode->i_size; + err = ext3_add_nondir(handle, dentry, inode); + ext3_mark_inode_dirty(handle, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- ext3_dec_count(handle, inode); +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + static int ext3_link (struct dentry * old_dentry, +@@ -956,12 +2042,15 @@ static int ext3_link (struct dentry * ol + if (S_ISDIR(inode->i_mode)) + return -EPERM; + +- if (inode->i_nlink >= EXT3_LINK_MAX) ++ if (inode->i_nlink >= EXT3_LINK_MAX) { + return -EMLINK; ++ } + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -995,9 +2084,11 @@ static int ext3_rename (struct inode * o + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 2); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(old_dir) || IS_SYNC(new_dir)) + handle->h_sync = 1; +@@ -1077,7 +2168,7 @@ static int ext3_rename (struct inode * o + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +2180,7 @@ static int ext3_rename (struct inode * o + new_inode->i_nlink--; + } else { + new_dir->i_nlink++; +- new_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(new_dir); + ext3_mark_inode_dirty(handle, new_dir); + } + } +--- linux-2.4.20/fs/ext3/super.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/fs/ext3/super.c Sat Apr 5 03:56:31 2003 +@@ -707,6 +707,7 @@ static int ext3_setup_super(struct super + es->s_mtime = cpu_to_le32(CURRENT_TIME); + ext3_update_dynamic_rev(sb); + EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); ++ + ext3_commit_super (sb, es, 1); + if (test_opt (sb, DEBUG)) + printk (KERN_INFO +@@ -717,6 +718,7 @@ static int ext3_setup_super(struct super + EXT3_BLOCKS_PER_GROUP(sb), + EXT3_INODES_PER_GROUP(sb), + sbi->s_mount_opt); ++ + printk(KERN_INFO "EXT3 FS " EXT3FS_VERSION ", " EXT3FS_DATE " on %s, ", + bdevname(sb->s_dev)); + if (EXT3_SB(sb)->s_journal->j_inode == NULL) { +@@ -890,6 +892,7 @@ static loff_t ext3_max_size(int bits) + return res; + } + ++ + struct super_block * ext3_read_super (struct super_block * sb, void * data, + int silent) + { +@@ -1066,6 +1069,9 @@ struct super_block * ext3_read_super (st + sbi->s_mount_state = le16_to_cpu(es->s_state); + sbi->s_addr_per_block_bits = log2(EXT3_ADDR_PER_BLOCK(sb)); + sbi->s_desc_per_block_bits = log2(EXT3_DESC_PER_BLOCK(sb)); ++ for (i=0; i < 4; i++) ++ sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]); ++ sbi->s_def_hash_version = es->s_def_hash_version; + + if (sbi->s_blocks_per_group > blocksize * 8) { + printk (KERN_ERR +@@ -1769,6 +1775,7 @@ static void __exit exit_ext3_fs(void) + unregister_filesystem(&ext3_fs_type); + } + ++EXPORT_SYMBOL(ext3_force_commit); + EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); +--- linux-2.4.20/include/linux/ext3_fs.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_fs.h Sat Apr 5 03:56:31 2003 +@@ -40,6 +40,11 @@ + #define EXT3FS_VERSION "2.4-0.9.19" + + /* ++ * Always enable hashed directories ++ */ ++#define CONFIG_EXT3_INDEX ++ ++/* + * Debug code + */ + #ifdef EXT3FS_DEBUG +@@ -437,8 +442,11 @@ struct ext3_super_block { + /*E0*/ __u32 s_journal_inum; /* inode number of journal file */ + __u32 s_journal_dev; /* device number of journal file */ + __u32 s_last_orphan; /* start of list of inodes to delete */ +- +-/*EC*/ __u32 s_reserved[197]; /* Padding to the end of the block */ ++ __u32 s_hash_seed[4]; /* HTREE hash seed */ ++ __u8 s_def_hash_version; /* Default hash version to use */ ++ __u8 s_reserved_char_pad; ++ __u16 s_reserved_word_pad; ++ __u32 s_reserved[192]; /* Padding to the end of the block */ + }; + + #ifdef __KERNEL__ +@@ -575,9 +583,46 @@ struct ext3_dir_entry_2 { + #define EXT3_DIR_ROUND (EXT3_DIR_PAD - 1) + #define EXT3_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT3_DIR_ROUND) & \ + ~EXT3_DIR_ROUND) ++/* ++ * Hash Tree Directory indexing ++ * (c) Daniel Phillips, 2001 ++ */ ++ ++#ifdef CONFIG_EXT3_INDEX ++ #define is_dx(dir) (EXT3_HAS_COMPAT_FEATURE(dir->i_sb, \ ++ EXT3_FEATURE_COMPAT_DIR_INDEX) && \ ++ (EXT3_I(dir)->i_flags & EXT3_INDEX_FL)) ++#define EXT3_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) ++#else ++ #define is_dx(dir) 0 ++#define EXT3_DIR_LINK_MAX(dir) ((dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2) ++#endif ++ ++/* Legal values for the dx_root hash_version field: */ ++ ++#define DX_HASH_LEGACY 0 ++#define DX_HASH_HALF_MD4 1 ++#define DX_HASH_TEA 2 ++ ++/* hash info structure used by the directory hash */ ++struct dx_hash_info ++{ ++ u32 hash; ++ u32 minor_hash; ++ int hash_version; ++ u32 *seed; ++}; + + #ifdef __KERNEL__ + /* ++ * Control parameters used by ext3_htree_next_block ++ */ ++#define HASH_NB_ALWAYS 1 ++ ++ ++/* + * Describe an inode's exact location on disk and in memory + */ + struct ext3_iloc +@@ -587,6 +632,27 @@ struct ext3_iloc + unsigned long block_group; + }; + ++ ++/* ++ * This structure is stuffed into the struct file's private_data field ++ * for directories. It is where we put information so that we can do ++ * readdir operations in hash tree order. ++ */ ++struct dir_private_info { ++ rb_root_t root; ++ rb_node_t *curr_node; ++ struct fname *extra_fname; ++ loff_t last_pos; ++ __u32 curr_hash; ++ __u32 curr_minor_hash; ++ __u32 next_hash; ++}; ++ ++/* ++ * Special error return code only used by dx_probe() and its callers. ++ */ ++#define ERR_BAD_DX_DIR -75000 ++ + /* + * Function prototypes + */ +@@ -614,11 +680,20 @@ extern struct ext3_group_desc * ext3_get + + /* dir.c */ + extern int ext3_check_dir_entry(const char *, struct inode *, +- struct ext3_dir_entry_2 *, struct buffer_head *, +- unsigned long); ++ struct ext3_dir_entry_2 *, ++ struct buffer_head *, unsigned long); ++extern void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent); ++extern void ext3_htree_free_dir_info(struct dir_private_info *p); ++ + /* fsync.c */ + extern int ext3_sync_file (struct file *, struct dentry *, int); + ++/* hash.c */ ++extern int ext3fs_dirhash(const char *name, int len, struct ++ dx_hash_info *hinfo); ++ + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); +@@ -650,6 +725,8 @@ extern int ext3_ioctl (struct inode *, s + /* namei.c */ + extern int ext3_orphan_add(handle_t *, struct inode *); + extern int ext3_orphan_del(handle_t *, struct inode *); ++extern int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash); + + /* super.c */ + extern void ext3_error (struct super_block *, const char *, const char *, ...) +--- linux-2.4.20/include/linux/ext3_fs_sb.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_fs_sb.h Sat Apr 5 03:56:31 2003 +@@ -62,6 +62,8 @@ struct ext3_sb_info { + int s_inode_size; + int s_first_ino; + u32 s_next_generation; ++ u32 s_hash_seed[4]; ++ int s_def_hash_version; + + /* Journaling */ + struct inode * s_journal_inode; +--- linux-2.4.20/include/linux/ext3_jbd.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/ext3_jbd.h Sat Apr 5 03:56:31 2003 +@@ -63,6 +63,8 @@ extern int ext3_writepage_trans_blocks(s + + #define EXT3_RESERVE_TRANS_BLOCKS 12 + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, +--- linux-2.4.20/include/linux/rbtree.h~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/include/linux/rbtree.h Sat Apr 5 03:56:31 2003 +@@ -120,6 +120,8 @@ rb_root_t; + + extern void rb_insert_color(rb_node_t *, rb_root_t *); + extern void rb_erase(rb_node_t *, rb_root_t *); ++extern rb_node_t *rb_get_first(rb_root_t *root); ++extern rb_node_t *rb_get_next(rb_node_t *n); + + static inline void rb_link_node(rb_node_t * node, rb_node_t * parent, rb_node_t ** rb_link) + { +--- linux-2.4.20/lib/rbtree.c~ext-2.4-patch-1 Sat Apr 5 03:56:31 2003 ++++ linux-2.4.20-braam/lib/rbtree.c Sat Apr 5 03:56:31 2003 +@@ -17,6 +17,8 @@ + Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + + linux/lib/rbtree.c ++ ++ rb_get_first and rb_get_next written by Theodore Ts'o, 9/8/2002 + */ + + #include +@@ -294,3 +296,43 @@ void rb_erase(rb_node_t * node, rb_root_ + __rb_erase_color(child, parent, root); + } + EXPORT_SYMBOL(rb_erase); ++ ++/* ++ * This function returns the first node (in sort order) of the tree. ++ */ ++rb_node_t *rb_get_first(rb_root_t *root) ++{ ++ rb_node_t *n; ++ ++ n = root->rb_node; ++ if (!n) ++ return 0; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++} ++EXPORT_SYMBOL(rb_get_first); ++ ++/* ++ * Given a node, this function will return the next node in the tree. ++ */ ++rb_node_t *rb_get_next(rb_node_t *n) ++{ ++ rb_node_t *parent; ++ ++ if (n->rb_right) { ++ n = n->rb_right; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++ } else { ++ while ((parent = n->rb_parent)) { ++ if (n == parent->rb_left) ++ return parent; ++ n = parent; ++ } ++ return 0; ++ } ++} ++EXPORT_SYMBOL(rb_get_next); ++ + +_ diff --git a/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch b/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch new file mode 100644 index 0000000..4bcefce --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-delete_thread-2.4.19-suse.patch @@ -0,0 +1,481 @@ + fs/ext3/file.c | 4 + fs/ext3/inode.c | 116 ++++++++++++++++++++++ + fs/ext3/super.c | 230 +++++++++++++++++++++++++++++++++++++++++++++ + include/linux/ext3_fs.h | 5 + include/linux/ext3_fs_sb.h | 10 + + 5 files changed, 365 insertions(+) + +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 01:18:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:19:22 2003 +@@ -401,6 +401,220 @@ + } + } + ++#ifdef EXT3_DELETE_THREAD ++/* ++ * Delete inodes in a loop until there are no more to be deleted. ++ * Normally, we run in the background doing the deletes and sleeping again, ++ * and clients just add new inodes to be deleted onto the end of the list. ++ * If someone is concerned about free space (e.g. block allocation or similar) ++ * then they can sleep on s_delete_waiter_queue and be woken up when space ++ * has been freed. ++ */ ++int ext3_delete_thread(void *data) ++{ ++ struct super_block *sb = data; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ struct task_struct *tsk = current; ++ ++ /* Almost like daemonize, but not quite */ ++ exit_mm(current); ++ tsk->session = 1; ++ tsk->pgrp = 1; ++ tsk->tty = NULL; ++ exit_files(current); ++ reparent_to_init(); ++ ++ sprintf(tsk->comm, "kdelext3-%s", kdevname(sb->s_dev)); ++ sigfillset(&tsk->blocked); ++ ++ /*tsk->flags |= PF_KERNTHREAD;*/ ++ ++ INIT_LIST_HEAD(&sbi->s_delete_list); ++ wake_up(&sbi->s_delete_waiter_queue); ++ ext3_debug("delete thread on %s started\n", kdevname(sb->s_dev)); ++ ++ /* main loop */ ++ for (;;) { ++ wait_event_interruptible(sbi->s_delete_thread_queue, ++ !list_empty(&sbi->s_delete_list) || ++ !test_opt(sb, ASYNCDEL)); ++ ext3_debug("%s woken up: %lu inodes, %lu blocks\n", ++ tsk->comm,sbi->s_delete_inodes,sbi->s_delete_blocks); ++ ++ spin_lock(&sbi->s_delete_lock); ++ if (list_empty(&sbi->s_delete_list)) { ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ memset(&sbi->s_delete_list, 0, ++ sizeof(sbi->s_delete_list)); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("delete thread on %s exiting\n", ++ kdevname(sb->s_dev)); ++ wake_up(&sbi->s_delete_waiter_queue); ++ break; ++ } ++ ++ while (!list_empty(&sbi->s_delete_list)) { ++ struct inode *inode=list_entry(sbi->s_delete_list.next, ++ struct inode, i_dentry); ++ unsigned long blocks = inode->i_blocks >> ++ (inode->i_blkbits - 9); ++ ++ list_del_init(&inode->i_dentry); ++ spin_unlock(&sbi->s_delete_lock); ++ ext3_debug("%s delete ino %lu blk %lu\n", ++ tsk->comm, inode->i_ino, blocks); ++ ++ iput(inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ sbi->s_delete_blocks -= blocks; ++ sbi->s_delete_inodes--; ++ } ++ if (sbi->s_delete_blocks != 0 || sbi->s_delete_inodes != 0) { ++ ext3_warning(sb, __FUNCTION__, ++ "%lu blocks, %lu inodes on list?\n", ++ sbi->s_delete_blocks,sbi->s_delete_inodes); ++ sbi->s_delete_blocks = 0; ++ sbi->s_delete_inodes = 0; ++ } ++ spin_unlock(&sbi->s_delete_lock); ++ wake_up(&sbi->s_delete_waiter_queue); ++ } ++ ++ return 0; ++} ++ ++static void ext3_start_delete_thread(struct super_block *sb) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ int rc; ++ ++ spin_lock_init(&sbi->s_delete_lock); ++ init_waitqueue_head(&sbi->s_delete_thread_queue); ++ init_waitqueue_head(&sbi->s_delete_waiter_queue); ++ ++ if (!test_opt(sb, ASYNCDEL)) ++ return; ++ ++ rc = kernel_thread(ext3_delete_thread, sb, CLONE_VM | CLONE_FILES); ++ if (rc < 0) ++ printk(KERN_ERR "EXT3-fs: cannot start delete thread: rc %d\n", ++ rc); ++ else ++ wait_event(sbi->s_delete_waiter_queue, sbi->s_delete_list.next); ++} ++ ++static void ext3_stop_delete_thread(struct ext3_sb_info *sbi) ++{ ++ if (sbi->s_delete_list.next == 0) /* thread never started */ ++ return; ++ ++ clear_opt(sbi->s_mount_opt, ASYNCDEL); ++ wake_up(&sbi->s_delete_thread_queue); ++ wait_event(sbi->s_delete_waiter_queue, list_empty(&sbi->s_delete_list)); ++} ++ ++/* Instead of playing games with the inode flags, destruction, etc we just ++ * create a new inode locally and put it on a list for the truncate thread. ++ * We need large parts of the inode struct in order to complete the ++ * truncate and unlink, so we may as well just have a real inode to do it. ++ * ++ * If we have any problem deferring the delete, just delete it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ */ ++static void ext3_delete_inode_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (is_bad_inode(old_inode)) { ++ clear_inode(old_inode); ++ return; ++ } ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_delete; ++ ++ /* We may want to delete the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS) ++ goto out_delete; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_delete; ++ } ++ ++ /* We can iget this inode again here, because our caller has unhashed ++ * old_inode, so new_inode will be in a different inode struct. ++ * ++ * We need to ensure that the i_orphan pointers in the other inodes ++ * point at the new inode copy instead of the old one so the orphan ++ * list doesn't get corrupted when the old orphan inode is freed. ++ */ ++ down(&sbi->s_orphan_lock); ++ ++ sbi->s_mount_state |= EXT3_ORPHAN_FS; ++ new_inode = iget(old_inode->i_sb, old_inode->i_ino); ++ sbi->s_mount_state &= ~EXT3_ORPHAN_FS; ++ if (is_bad_inode(new_inode)) { ++ printk(KERN_WARNING "read bad inode %lu\n", old_inode->i_ino); ++ iput(new_inode); ++ new_inode = NULL; ++ } ++ if (!new_inode) { ++ up(&sbi->s_orphan_lock); ++ ext3_debug("delete inode %lu directly (bad read)\n", ++ old_inode->i_ino); ++ goto out_delete; ++ } ++ J_ASSERT(new_inode != old_inode); ++ ++ J_ASSERT(!list_empty(&oei->i_orphan)); ++ ++ nei = EXT3_I(new_inode); ++ /* Ugh. We need to insert new_inode into the same spot on the list ++ * as old_inode was, to ensure the in-memory orphan list is still ++ * in the same order as the on-disk orphan list (badness otherwise). ++ */ ++ nei->i_orphan = oei->i_orphan; ++ nei->i_orphan.next->prev = &nei->i_orphan; ++ nei->i_orphan.prev->next = &nei->i_orphan; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up(&sbi->s_orphan_lock); ++ ++ clear_inode(old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_delete: ++ ext3_delete_inode(old_inode); ++} ++#else ++#define ext3_start_delete_thread(sbi) do {} while(0) ++#define ext3_stop_delete_thread(sbi) do {} while(0) ++#endif /* EXT3_DELETE_THREAD */ ++ + void ext3_put_super (struct super_block * sb) + { + struct ext3_sb_info *sbi = EXT3_SB(sb); +@@ -408,6 +622,7 @@ + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ ext3_stop_delete_thread(sbi); + ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { +@@ -476,7 +691,11 @@ + write_inode: ext3_write_inode, /* BKL not held. Don't need */ + dirty_inode: ext3_dirty_inode, /* BKL not held. We take it */ + put_inode: ext3_put_inode, /* BKL not held. Don't need */ ++#ifdef EXT3_DELETE_THREAD ++ delete_inode: ext3_delete_inode_thread,/* BKL not held. We take it */ ++#else + delete_inode: ext3_delete_inode, /* BKL not held. We take it */ ++#endif + put_super: ext3_put_super, /* BKL held */ + write_super: ext3_write_super, /* BKL held */ + sync_fs: ext3_sync_fs, +@@ -553,6 +772,13 @@ + clear_opt (*mount_options, POSIX_ACL); + else + #endif ++#ifdef EXT3_DELETE_THREAD ++ if (!strcmp(this_char, "asyncdel")) ++ set_opt(*mount_options, ASYNCDEL); ++ else if (!strcmp(this_char, "noasyncdel")) ++ clear_opt(*mount_options, ASYNCDEL); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -1254,6 +1480,7 @@ + } + + ext3_setup_super (sb, es, sb->s_flags & MS_RDONLY); ++ ext3_start_delete_thread(sb); + /* + * akpm: core read_super() calls in here with the superblock locked. + * That deadlocks, because orphan cleanup needs to lock the superblock +@@ -1692,6 +1919,9 @@ + if (!parse_options(data, &tmp, sbi, &tmp, 1)) + return -EINVAL; + ++ if (!test_opt(sb, ASYNCDEL) || (*flags & MS_RDONLY)) ++ ext3_stop_delete_thread(sbi); ++ + if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) + ext3_abort(sb, __FUNCTION__, "Abort forced by user"); + +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:02:56 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:19:22 2003 +@@ -2114,6 +2114,118 @@ + ext3_journal_stop(handle, inode); + } + ++#ifdef EXT3_DELETE_THREAD ++/* Move blocks from to-be-truncated inode over to a new inode, and delete ++ * that one from the delete thread instead. This avoids a lot of latency ++ * when truncating large files. ++ * ++ * If we have any problem deferring the truncate, just truncate it right away. ++ * If we defer it, we also mark how many blocks it would free, so that we ++ * can keep the statfs data correct, and we know if we should sleep on the ++ * delete thread when we run out of space. ++ */ ++void ext3_truncate_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ struct ext3_inode_info *nei, *oei = EXT3_I(old_inode); ++ struct inode *new_inode; ++ handle_t *handle; ++ unsigned long blocks = old_inode->i_blocks >> (old_inode->i_blkbits-9); ++ ++ if (!test_opt(old_inode->i_sb, ASYNCDEL) || !sbi->s_delete_list.next) ++ goto out_truncate; ++ ++ /* XXX This is a temporary limitation for code simplicity. ++ * We could truncate to arbitrary sizes at some later time. ++ */ ++ if (old_inode->i_size != 0) ++ goto out_truncate; ++ ++ /* We may want to truncate the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS || ++ old_inode->i_size > oei->i_disksize) ++ goto out_truncate; ++ ++ /* We can't use the delete thread as-is during real orphan recovery, ++ * as we add to the orphan list here, causing ext3_orphan_cleanup() ++ * to loop endlessly. It would be nice to do so, but needs work. ++ */ ++ if (oei->i_state & EXT3_STATE_DELETE || ++ sbi->s_mount_state & EXT3_ORPHAN_FS) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ goto out_truncate; ++ } ++ ++ ext3_discard_prealloc(old_inode); ++ ++ /* old_inode = 1 ++ * new_inode = sb + GDT + ibitmap ++ * orphan list = 1 inode/superblock for add, 2 inodes for del ++ * quota files = 2 * EXT3_SINGLEDATA_TRANS_BLOCKS ++ */ ++ handle = ext3_journal_start(old_inode, 7); ++ if (IS_ERR(handle)) ++ goto out_truncate; ++ ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ if (IS_ERR(new_inode)) { ++ ext3_debug("truncate inode %lu directly (no new inodes)\n", ++ old_inode->i_ino); ++ goto out_journal; ++ } ++ ++ nei = EXT3_I(new_inode); ++ ++ down_write(&oei->truncate_sem); ++ new_inode->i_size = old_inode->i_size; ++ new_inode->i_blocks = old_inode->i_blocks; ++ new_inode->i_uid = old_inode->i_uid; ++ new_inode->i_gid = old_inode->i_gid; ++ new_inode->i_nlink = 0; ++ ++ /* FIXME when we do arbitrary truncates */ ++ old_inode->i_blocks = oei->i_file_acl ? old_inode->i_blksize / 512 : 0; ++ old_inode->i_mtime = old_inode->i_ctime = CURRENT_TIME; ++ ++ memcpy(nei->i_data, oei->i_data, sizeof(nei->i_data)); ++ memset(oei->i_data, 0, sizeof(oei->i_data)); ++ ++ nei->i_disksize = oei->i_disksize; ++ nei->i_state |= EXT3_STATE_DELETE; ++ up_write(&oei->truncate_sem); ++ ++ if (ext3_orphan_add(handle, new_inode) < 0) ++ goto out_journal; ++ ++ if (ext3_orphan_del(handle, old_inode) < 0) { ++ ext3_orphan_del(handle, new_inode); ++ iput(new_inode); ++ goto out_journal; ++ } ++ ++ ext3_journal_stop(handle, old_inode); ++ ++ spin_lock(&sbi->s_delete_lock); ++ J_ASSERT(list_empty(&new_inode->i_dentry)); ++ list_add_tail(&new_inode->i_dentry, &sbi->s_delete_list); ++ sbi->s_delete_blocks += blocks; ++ sbi->s_delete_inodes++; ++ spin_unlock(&sbi->s_delete_lock); ++ ++ ext3_debug("delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++ return; ++ ++out_journal: ++ ext3_journal_stop(handle, old_inode); ++out_truncate: ++ ext3_truncate(old_inode); ++} ++#endif /* EXT3_DELETE_THREAD */ ++ + /* + * ext3_get_inode_loc returns with an extra refcount against the + * inode's underlying buffer_head on success. +Index: linux-2.4.19.SuSE/fs/ext3/file.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/file.c Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/file.c Sun Nov 16 01:19:22 2003 +@@ -132,7 +132,11 @@ + }; + + struct inode_operations ext3_file_inode_operations = { ++#ifdef EXT3_DELETE_THREAD ++ truncate: ext3_truncate_thread, /* BKL held */ ++#else + truncate: ext3_truncate, /* BKL held */ ++#endif + setattr: ext3_setattr, /* BKL held */ + setxattr: ext3_setxattr, /* BKL held */ + getxattr: ext3_getxattr, /* BKL held */ +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:02:51 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:20:06 2003 +@@ -193,6 +193,7 @@ + */ + #define EXT3_STATE_JDATA 0x00000001 /* journaled data exists */ + #define EXT3_STATE_NEW 0x00000002 /* inode is newly created */ ++#define EXT3_STATE_DELETE 0x00000010 /* deferred delete inode */ + + /* + * ioctl commands +@@ -321,6 +322,7 @@ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ + #define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + #define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */ ++#define EXT3_MOUNT_ASYNCDEL 0x20000 /* Delayed deletion */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H +@@ -695,6 +697,9 @@ + extern void ext3_dirty_inode(struct inode *); + extern int ext3_change_inode_journal_flag(struct inode *, int); + extern void ext3_truncate (struct inode *); ++#ifdef EXT3_DELETE_THREAD ++extern void ext3_truncate_thread(struct inode *inode); ++#endif + + /* ioctl.c */ + extern int ext3_ioctl (struct inode *, struct file *, unsigned int, +Index: linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs_sb.h Sun Nov 16 01:18:41 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h Sun Nov 16 01:19:22 2003 +@@ -29,6 +29,8 @@ + + #define EXT3_MAX_GROUP_LOADED 8 + ++#define EXT3_DELETE_THREAD ++ + /* + * third extended-fs super-block data in memory + */ +@@ -75,6 +77,14 @@ + struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */ + wait_queue_head_t ro_wait_queue; /* For people waiting for the fs to go read-only */ + #endif ++#ifdef EXT3_DELETE_THREAD ++ spinlock_t s_delete_lock; ++ struct list_head s_delete_list; ++ unsigned long s_delete_blocks; ++ unsigned long s_delete_inodes; ++ wait_queue_head_t s_delete_thread_queue; ++ wait_queue_head_t s_delete_waiter_queue; ++#endif + }; + + #endif /* _LINUX_EXT3_FS_SB */ diff --git a/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch new file mode 100644 index 0000000..3c6b5e5 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-orphan_lock-2.4.19-suse.patch @@ -0,0 +1,85 @@ +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:14:50 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:18:04 2003 +@@ -1751,8 +1751,8 @@ + struct super_block *sb = inode->i_sb; + struct ext3_iloc iloc; + int err = 0, rc; +- +- lock_super(sb); ++ ++ down(&EXT3_SB(sb)->s_orphan_lock); + if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + +@@ -1800,7 +1800,7 @@ + jbd_debug(4, "orphan inode %ld will point to %d\n", + inode->i_ino, NEXT_ORPHAN(inode)); + out_unlock: +- unlock_super(sb); ++ up(&EXT3_SB(sb)->s_orphan_lock); + ext3_std_error(inode->i_sb, err); + return err; + } +@@ -1813,20 +1813,19 @@ + { + struct list_head *prev; + struct ext3_inode_info *ei = EXT3_I(inode); +- struct ext3_sb_info *sbi; ++ struct ext3_sb_info *sbi = EXT3_SB(inode->i_sb); + unsigned long ino_next; + struct ext3_iloc iloc; + int err = 0; + +- lock_super(inode->i_sb); ++ down(&sbi->s_orphan_lock); + if (list_empty(&ei->i_orphan)) { +- unlock_super(inode->i_sb); ++ up(&sbi->s_orphan_lock); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); + prev = ei->i_orphan.prev; +- sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); + +@@ -1872,10 +1871,10 @@ + if (err) + goto out_brelse; + +-out_err: ++out_err: + ext3_std_error(inode->i_sb, err); + out: +- unlock_super(inode->i_sb); ++ up(&sbi->s_orphan_lock); + return err; + + out_brelse: +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:18:04 2003 +@@ -1182,6 +1182,7 @@ + */ + sb->s_op = &ext3_sops; + INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */ ++ sema_init(&sbi->s_orphan_lock, 1); + + sb->s_root = 0; + +Index: linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs_sb.h Sat Nov 15 23:58:28 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs_sb.h Sun Nov 16 01:18:41 2003 +@@ -69,6 +69,7 @@ + struct inode * s_journal_inode; + struct journal_s * s_journal; + struct list_head s_orphan; ++ struct semaphore s_orphan_lock; + struct block_device *journal_bdev; + #ifdef CONFIG_JBD_DEBUG + struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */ diff --git a/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch b/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch new file mode 100644 index 0000000..1dab6d8 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-use-after-free-suse.patch @@ -0,0 +1,53 @@ + ./fs/ext3/namei.c | 11 +++++------ + 1 files changed, 5 insertions(+), 6 deletions(-) + +Index: linux-2.4.19.SuSE/./fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/./fs/ext3/namei.c Sun Nov 16 01:02:51 2003 ++++ linux-2.4.19.SuSE/./fs/ext3/namei.c Sun Nov 16 01:14:50 2003 +@@ -1523,8 +1523,11 @@ + { + int err = ext3_add_entry(handle, dentry, inode); + if (!err) { +- d_instantiate(dentry, inode); +- return 0; ++ err = ext3_mark_inode_dirty(handle, inode); ++ if (err == 0) { ++ d_instantiate(dentry, inode); ++ return 0; ++ } + } + ext3_dec_count(handle, inode); + iput(inode); +@@ -1560,7 +1563,6 @@ + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -1590,7 +1592,6 @@ + #ifdef CONFIG_EXT3_FS_XATTR + inode->i_op = &ext3_special_inode_operations; + #endif +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + } + ext3_journal_stop(handle, dir); +@@ -2039,7 +2040,6 @@ + inode->i_size = l-1; + } + EXT3_I(inode)->i_disksize = inode->i_size; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); +@@ -2073,7 +2073,6 @@ + ext3_inc_count(handle, inode); + atomic_inc(&inode->i_count); + +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); + ext3_journal_stop(handle, dir); + return err; diff --git a/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch b/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch new file mode 100644 index 0000000..02cfef1 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-wantedi-2.4.19-suse.patch @@ -0,0 +1,226 @@ + fs/ext3/ialloc.c | 40 ++++++++++++++++++++++++++++++++++++++-- + fs/ext3/inode.c | 2 +- + fs/ext3/ioctl.c | 25 +++++++++++++++++++++++++ + fs/ext3/namei.c | 21 +++++++++++++++++---- + include/linux/dcache.h | 5 +++++ + include/linux/ext3_fs.h | 5 ++++- + 6 files changed, 90 insertions(+), 8 deletions(-) + +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:18:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:23:20 2003 +@@ -1534,6 +1534,19 @@ + return err; + } + ++static struct inode * ext3_new_inode_wantedi(handle_t *handle, struct inode *dir, ++ int mode, struct dentry *dentry) ++{ ++ unsigned long inum = 0; ++ ++ if (dentry->d_fsdata != NULL) { ++ struct dentry_params *param = ++ (struct dentry_params *) dentry->d_fsdata; ++ inum = param->p_inum; ++ } ++ return ext3_new_inode(handle, dir, mode, inum); ++} ++ + /* + * By the time this is called, we already have created + * the directory cache entry for the new file, but it +@@ -1557,7 +1570,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + inode->i_op = &ext3_file_inode_operations; +@@ -1585,7 +1598,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode_wantedi (handle, dir, mode, dentry); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, inode->i_mode, rdev); +@@ -1618,7 +1631,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR | mode); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFDIR | mode, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -2013,7 +2026,7 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); ++ inode = ext3_new_inode_wantedi (handle, dir, S_IFLNK|S_IRWXUGO, dentry); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +Index: linux-2.4.19.SuSE/fs/ext3/ialloc.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ialloc.c Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/fs/ext3/ialloc.c Sun Nov 16 01:24:49 2003 +@@ -330,7 +330,8 @@ + * For other inodes, search forward from the parent directory's block + * group to find a free inode. + */ +-struct inode * ext3_new_inode (handle_t *handle, struct inode * dir, int mode) ++struct inode * ext3_new_inode(handle_t *handle, const struct inode * dir, ++ int mode, unsigned long goal) + { + struct super_block * sb; + struct buffer_head * bh; +@@ -355,7 +356,41 @@ + init_rwsem(&inode->u.ext3_i.truncate_sem); + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; ++ ++ if (goal) { ++ i = (goal - 1) / EXT3_INODES_PER_GROUP(sb); ++ j = (goal - 1) % EXT3_INODES_PER_GROUP(sb); ++ gdp = ext3_get_group_desc(sb, i, &bh2); ++ ++ bitmap_nr = load_inode_bitmap (sb, i); ++ if (bitmap_nr < 0) { ++ err = bitmap_nr; ++ goto fail; ++ } ++ ++ bh = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; ++ ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) goto fail; ++ ++ if (ext3_set_bit(j, bh->b_data)) { ++ printk(KERN_ERR "goal inode %lu unavailable\n", goal); ++ /* Oh well, we tried. */ ++ goto repeat; ++ } ++ ++ BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, bh); ++ if (err) goto fail; ++ ++ /* We've shortcircuited the allocation system successfully, ++ * now finish filling in the inode. ++ */ ++ goto have_bit_and_group; ++ } ++ + repeat: + gdp = NULL; + i = 0; +@@ -470,6 +505,7 @@ + } + goto repeat; + } ++ have_bit_and_group: + j += i * EXT3_INODES_PER_GROUP(sb) + 1; + if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:23:20 2003 +@@ -2168,7 +2168,7 @@ + if (IS_ERR(handle)) + goto out_truncate; + +- new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode); ++ new_inode = ext3_new_inode(handle, old_inode, old_inode->i_mode, 0); + if (IS_ERR(new_inode)) { + ext3_debug("truncate inode %lu directly (no new inodes)\n", + old_inode->i_ino); +Index: linux-2.4.19.SuSE/fs/ext3/ioctl.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ioctl.c Fri Nov 9 14:25:04 2001 ++++ linux-2.4.19.SuSE/fs/ext3/ioctl.c Sun Nov 16 01:23:20 2003 +@@ -23,6 +23,31 @@ + ext3_debug ("cmd = %u, arg = %lu\n", cmd, arg); + + switch (cmd) { ++ case EXT3_IOC_CREATE_INUM: { ++ char name[32]; ++ struct dentry *dchild, *dparent; ++ int rc = 0; ++ ++ dparent = list_entry(inode->i_dentry.next, struct dentry, ++ d_alias); ++ snprintf(name, sizeof name, "%lu", arg); ++ dchild = lookup_one_len(name, dparent, strlen(name)); ++ if (dchild->d_inode) { ++ printk(KERN_ERR "%*s/%lu already exists (ino %lu)\n", ++ dparent->d_name.len, dparent->d_name.name, arg, ++ dchild->d_inode->i_ino); ++ rc = -EEXIST; ++ } else { ++ dchild->d_fsdata = (void *)arg; ++ rc = vfs_create(inode, dchild, 0644); ++ if (rc) ++ printk(KERN_ERR "vfs_create: %d\n", rc); ++ else if (dchild->d_inode->i_ino != arg) ++ rc = -EEXIST; ++ } ++ dput(dchild); ++ return rc; ++ } + case EXT3_IOC_GETFLAGS: + flags = inode->u.ext3_i.i_flags & EXT3_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:20:17 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:25:42 2003 +@@ -202,6 +202,7 @@ + #define EXT3_IOC_SETFLAGS _IOW('f', 2, long) + #define EXT3_IOC_GETVERSION _IOR('f', 3, long) + #define EXT3_IOC_SETVERSION _IOW('f', 4, long) ++/* EXT3_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ + #define EXT3_IOC_GETVERSION_OLD _IOR('v', 1, long) + #define EXT3_IOC_SETVERSION_OLD _IOW('v', 2, long) + #ifdef CONFIG_JBD_DEBUG +@@ -674,7 +675,8 @@ + dx_hash_info *hinfo); + + /* ialloc.c */ +-extern struct inode * ext3_new_inode (handle_t *, struct inode *, int); ++extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int, ++ unsigned long); + extern void ext3_free_inode (handle_t *, struct inode *); + extern struct inode * ext3_orphan_get (struct super_block *, unsigned long); + extern unsigned long ext3_count_free_inodes (struct super_block *); +@@ -765,4 +767,5 @@ + + #endif /* __KERNEL__ */ + ++#define EXT3_IOC_CREATE_INUM _IOW('f', 5, long) + #endif /* _LINUX_EXT3_FS_H */ +Index: linux-2.4.19.SuSE/include/linux/dcache.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/dcache.h Sat Nov 15 17:35:46 2003 ++++ linux-2.4.19.SuSE/include/linux/dcache.h Sun Nov 16 01:23:20 2003 +@@ -62,6 +62,11 @@ + + #define IS_ROOT(x) ((x) == (x)->d_parent) + ++struct dentry_params { ++ unsigned long p_inum; ++ void *p_ptr; ++}; ++ + /* + * "quick string" -- eases parameter passing, but more importantly + * saves "metadata" about the string (ie length and the hash). diff --git a/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch b/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch new file mode 100644 index 0000000..2040fcd --- /dev/null +++ b/lustre/kernel_patches/patches/iod-stock-24-exports-2.4.19-suse.patch @@ -0,0 +1,52 @@ + fs/Makefile | 2 +- + fs/inode.c | 4 +++- + mm/page_alloc.c | 1 + + 3 files changed, 5 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/fs/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/inode.c Sat Nov 15 18:02:13 2003 ++++ linux-2.4.19.SuSE/fs/inode.c Sat Nov 15 18:03:04 2003 +@@ -5,6 +5,7 @@ + */ + + #include ++#include + #include + #include + #include +@@ -67,7 +68,8 @@ + * NOTE! You also have to own the lock if you change + * the i_state of an inode while it is in use.. + */ +-static spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++spinlock_t inode_lock = SPIN_LOCK_UNLOCKED; ++EXPORT_SYMBOL(inode_lock); + + /* + * Statistics gathering.. +Index: linux-2.4.19.SuSE/fs/Makefile +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/Makefile Mon Jan 27 05:08:56 2003 ++++ linux-2.4.19.SuSE/fs/Makefile Sat Nov 15 18:03:54 2003 +@@ -7,7 +7,7 @@ + + O_TARGET := fs.o + +-export-objs := filesystems.o open.o dcache.o buffer.o ++export-objs := filesystems.o open.o dcache.o buffer.o inode.o + mod-subdirs := nls + + obj-y := open.o read_write.o devices.o file_table.o buffer.o \ +Index: linux-2.4.19.SuSE/mm/page_alloc.c +=================================================================== +--- linux-2.4.19.SuSE.orig/mm/page_alloc.c Mon Jan 27 05:08:55 2003 ++++ linux-2.4.19.SuSE/mm/page_alloc.c Sat Nov 15 18:03:04 2003 +@@ -32,6 +32,7 @@ + LIST_HEAD(inactive_list); + LIST_HEAD(active_list); + pg_data_t *pgdat_list; ++EXPORT_SYMBOL(pgdat_list); + + /* Used to look up the address of the struct zone encoded in page->zone */ + zone_t *zone_table[MAX_NR_ZONES*MAX_NR_NODES]; diff --git a/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch b/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch new file mode 100644 index 0000000..ab267bb --- /dev/null +++ b/lustre/kernel_patches/patches/iopen-2.4.19-suse.patch @@ -0,0 +1,437 @@ + Documentation/filesystems/ext2.txt | 16 ++ + fs/ext3/Makefile | 2 + fs/ext3/inode.c | 4 + fs/ext3/iopen.c | 259 +++++++++++++++++++++++++++++++++++++ + fs/ext3/iopen.h | 13 + + fs/ext3/namei.c | 13 + + fs/ext3/super.c | 11 + + include/linux/ext3_fs.h | 2 + 8 files changed, 318 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/Documentation/filesystems/ext2.txt +=================================================================== +--- linux-2.4.19.SuSE.orig/Documentation/filesystems/ext2.txt Wed Jul 11 15:44:45 2001 ++++ linux-2.4.19.SuSE/Documentation/filesystems/ext2.txt Sun Nov 16 01:27:31 2003 +@@ -35,6 +35,22 @@ + + sb=n Use alternate superblock at this location. + ++iopen Makes an invisible pseudo-directory called ++ __iopen__ available in the root directory ++ of the filesystem. Allows open-by-inode- ++ number. i.e., inode 3145 can be accessed ++ via /mntpt/__iopen__/3145 ++ ++iopen_nopriv This option makes the iopen directory be ++ world-readable. This may be safer since it ++ allows daemons to run as an unprivileged user, ++ however it significantly changes the security ++ model of a Unix filesystem, since previously ++ all files under a mode 700 directory were not ++ generally avilable even if the ++ permissions on the file itself is ++ world-readable. ++ + grpquota,noquota,quota,usrquota Quota options are silently ignored by ext2. + + +Index: linux-2.4.19.SuSE/fs/ext3/Makefile +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/Makefile Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/fs/ext3/Makefile Sun Nov 16 01:27:31 2003 +@@ -11,7 +11,7 @@ + + export-objs := ext3-exports.o + +-obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ ++obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o iopen.o \ + ioctl.o namei.o super.o symlink.o hash.o ext3-exports.o + obj-m := $(O_TARGET) + +Index: linux-2.4.19.SuSE/fs/ext3/inode.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/inode.c Sun Nov 16 01:26:04 2003 ++++ linux-2.4.19.SuSE/fs/ext3/inode.c Sun Nov 16 01:27:31 2003 +@@ -34,6 +34,7 @@ + #include + #include + #include ++#include "iopen.h" + + /* + * SEARCH_FROM_ZERO forces each block allocation to search from the start +@@ -2350,6 +2351,9 @@ + struct buffer_head *bh; + int block; + ++ if (ext3_iopen_get_inode(inode)) ++ return; ++ + if(ext3_get_inode_loc(inode, &iloc)) + goto bad_inode; + bh = iloc.bh; +Index: linux-2.4.19.SuSE/fs/ext3/iopen.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/iopen.c Sun Nov 16 01:27:31 2003 ++++ linux-2.4.19.SuSE/fs/ext3/iopen.c Sun Nov 16 01:27:31 2003 +@@ -0,0 +1,259 @@ ++/* ++ * linux/fs/ext3/iopen.c ++ * ++ * Special support for open by inode number ++ * ++ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). ++ * ++ * This file may be redistributed under the terms of the GNU General ++ * Public License. ++ * ++ * ++ * Invariants: ++ * - there is only ever a single DCACHE_NFSD_DISCONNECTED dentry alias ++ * for an inode at one time. ++ * - there are never both connected and DCACHE_NFSD_DISCONNECTED dentry ++ * aliases on an inode at the same time. ++ * ++ * If we have any connected dentry aliases for an inode, use one of those ++ * in iopen_lookup(). Otherwise, we instantiate a single NFSD_DISCONNECTED ++ * dentry for this inode, which thereafter will be found by the dcache ++ * when looking up this inode number in __iopen__, so we don't return here ++ * until it is gone. ++ * ++ * If we get an inode via a regular name lookup, then we "rename" the ++ * NFSD_DISCONNECTED dentry to the proper name and parent. This ensures ++ * existing users of the disconnected dentry will continue to use the same ++ * dentry as the connected users, and there will never be both kinds of ++ * dentry aliases at one time. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include "iopen.h" ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#define IOPEN_NAME_LEN 32 ++ ++/* ++ * This implements looking up an inode by number. ++ */ ++static struct dentry *iopen_lookup(struct inode *dir, struct dentry *dentry) ++{ ++ struct inode *inode; ++ unsigned long ino; ++ struct list_head *lp; ++ struct dentry *alternate; ++ char buf[IOPEN_NAME_LEN]; ++ ++ if (dentry->d_name.len >= IOPEN_NAME_LEN) ++ return ERR_PTR(-ENAMETOOLONG); ++ ++ memcpy(buf, dentry->d_name.name, dentry->d_name.len); ++ buf[dentry->d_name.len] = 0; ++ ++ if (strcmp(buf, ".") == 0) ++ ino = dir->i_ino; ++ else if (strcmp(buf, "..") == 0) ++ ino = EXT3_ROOT_INO; ++ else ++ ino = simple_strtoul(buf, 0, 0); ++ ++ if ((ino != EXT3_ROOT_INO && ++ //ino != EXT3_ACL_IDX_INO && ++ //ino != EXT3_ACL_DATA_INO && ++ ino < EXT3_FIRST_INO(dir->i_sb)) || ++ ino > le32_to_cpu(dir->i_sb->u.ext3_sb.s_es->s_inodes_count)) ++ return ERR_PTR(-ENOENT); ++ ++ inode = iget(dir->i_sb, ino); ++ if (!inode) ++ return ERR_PTR(-EACCES); ++ if (is_bad_inode(inode)) { ++ iput(inode); ++ return ERR_PTR(-ENOENT); ++ } ++ ++ /* preferrably return a connected dentry */ ++ spin_lock(&dcache_lock); ++ list_for_each(lp, &inode->i_dentry) { ++ alternate = list_entry(lp, struct dentry, d_alias); ++ assert(!(alternate->d_flags & DCACHE_NFSD_DISCONNECTED)); ++ } ++ ++ if (!list_empty(&inode->i_dentry)) { ++ alternate = list_entry(inode->i_dentry.next, ++ struct dentry, d_alias); ++ dget_locked(alternate); ++ alternate->d_vfs_flags |= DCACHE_REFERENCED; ++ iput(inode); ++ spin_unlock(&dcache_lock); ++ return alternate; ++ } ++ dentry->d_flags |= DCACHE_NFSD_DISCONNECTED; ++ spin_unlock(&dcache_lock); ++ ++ d_add(dentry, inode); ++ return NULL; ++} ++ ++#define do_switch(x,y) do { \ ++ __typeof__ (x) __tmp = x; \ ++ x = y; y = __tmp; } while (0) ++ ++static inline void switch_names(struct dentry *dentry, struct dentry *target) ++{ ++ const unsigned char *old_name, *new_name; ++ ++ memcpy(dentry->d_iname, target->d_iname, DNAME_INLINE_LEN); ++ old_name = target->d_name.name; ++ new_name = dentry->d_name.name; ++ if (old_name == target->d_iname) ++ old_name = dentry->d_iname; ++ if (new_name == dentry->d_iname) ++ new_name = target->d_iname; ++ target->d_name.name = new_name; ++ dentry->d_name.name = old_name; ++} ++ ++/* This function is spliced into ext3_lookup and does the move of a ++ * disconnected dentry (if it exists) to a connected dentry. ++ */ ++struct dentry *iopen_connect_dentry(struct dentry *de, struct inode *inode) ++{ ++ struct dentry *tmp, *goal = NULL; ++ struct list_head *lp; ++ ++ /* preferrably return a connected dentry */ ++ spin_lock(&dcache_lock); ++ /* verify this dentry is really new */ ++ assert(!de->d_inode); ++ assert(list_empty(&de->d_subdirs)); ++ assert(list_empty(&de->d_alias)); ++ ++ ++ list_for_each(lp, &inode->i_dentry) { ++ tmp = list_entry(lp, struct dentry, d_alias); ++ if (tmp->d_flags & DCACHE_NFSD_DISCONNECTED) { ++ assert(tmp->d_alias.next == &inode->i_dentry); ++ assert(tmp->d_alias.prev == &inode->i_dentry); ++ goal = tmp; ++ dget_locked(goal); ++ break; ++ } ++ } ++ ++ if (!goal) { ++ spin_unlock(&dcache_lock); ++ return NULL; ++ } ++ ++ /* Move the goal to the de hash queue - like d_move() */ ++ goal->d_flags &= ~DCACHE_NFSD_DISCONNECTED; ++ list_del_init(&goal->d_hash); ++ ++ list_del(&goal->d_child); ++ list_del(&de->d_child); ++ ++ /* Switch the parents and the names.. */ ++ switch_names(goal, de); ++ do_switch(goal->d_parent, de->d_parent); ++ do_switch(goal->d_name.len, de->d_name.len); ++ do_switch(goal->d_name.hash, de->d_name.hash); ++ ++ /* And add them back to the (new) parent lists */ ++ list_add(&goal->d_child, &goal->d_parent->d_subdirs); ++ list_add(&de->d_child, &de->d_parent->d_subdirs); ++ __d_rehash(goal, 0); ++ spin_unlock(&dcache_lock); ++ ++ return goal; ++} ++ ++/* ++ * These are the special structures for the iopen pseudo directory. ++ */ ++ ++static struct inode_operations iopen_inode_operations = { ++ lookup: iopen_lookup, /* BKL held */ ++}; ++ ++static struct file_operations iopen_file_operations = { ++ read: generic_read_dir, ++}; ++ ++static int match_dentry(struct dentry *dentry, const char *name) ++{ ++ int len; ++ ++ len = strlen(name); ++ if (dentry->d_name.len != len) ++ return 0; ++ if (strncmp(dentry->d_name.name, name, len)) ++ return 0; ++ return 1; ++} ++ ++/* ++ * This function is spliced into ext3_lookup and returns 1 the file ++ * name is __iopen__ and dentry has been filled in appropriately. ++ */ ++int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry) ++{ ++ struct inode *inode; ++ ++ if (dir->i_ino != EXT3_ROOT_INO || ++ !test_opt(dir->i_sb, IOPEN) || ++ !match_dentry(dentry, "__iopen__")) ++ return 0; ++ ++ inode = iget(dir->i_sb, EXT3_BAD_INO); ++ ++ if (!inode) ++ return 0; ++ d_add(dentry, inode); ++ return 1; ++} ++ ++/* ++ * This function is spliced into read_inode; it returns 1 if inode ++ * number is the one for /__iopen__, in which case the inode is filled ++ * in appropriately. Otherwise, this fuction returns 0. ++ */ ++int ext3_iopen_get_inode(struct inode *inode) ++{ ++ if (inode->i_ino != EXT3_BAD_INO) ++ return 0; ++ ++ inode->i_mode = S_IFDIR | S_IRUSR | S_IXUSR; ++ if (test_opt(inode->i_sb, IOPEN_NOPRIV)) ++ inode->i_mode |= 0777; ++ inode->i_uid = 0; ++ inode->i_gid = 0; ++ inode->i_nlink = 1; ++ inode->i_size = 4096; ++ inode->i_atime = CURRENT_TIME; ++ inode->i_ctime = CURRENT_TIME; ++ inode->i_mtime = CURRENT_TIME; ++ inode->u.ext3_i.i_dtime = 0; ++ inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size ++ * (for stat), not the fs block ++ * size */ ++ inode->i_blocks = 0; ++ inode->i_version = 1; ++ inode->i_generation = 0; ++ ++ inode->i_op = &iopen_inode_operations; ++ inode->i_fop = &iopen_file_operations; ++ inode->i_mapping->a_ops = 0; ++ ++ return 1; ++} +Index: linux-2.4.19.SuSE/fs/ext3/iopen.h +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/iopen.h Sun Nov 16 01:27:31 2003 ++++ linux-2.4.19.SuSE/fs/ext3/iopen.h Sun Nov 16 01:27:31 2003 +@@ -0,0 +1,13 @@ ++/* ++ * iopen.h ++ * ++ * Special support for opening files by inode number. ++ * ++ * Copyright (C) 2001 by Theodore Ts'o (tytso@alum.mit.edu). ++ * ++ * This file may be redistributed under the terms of the GNU General ++ * Public License. ++ */ ++ ++extern int ext3_check_for_iopen(struct inode *dir, struct dentry *dentry); ++extern int ext3_iopen_get_inode(struct inode *inode); +Index: linux-2.4.19.SuSE/fs/ext3/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/namei.c Sun Nov 16 01:23:20 2003 ++++ linux-2.4.19.SuSE/fs/ext3/namei.c Sun Nov 16 01:27:31 2003 +@@ -36,7 +36,7 @@ + #include + #include + #include +- ++#include "iopen.h" + + /* + * define how far ahead to read directories while searching them. +@@ -922,16 +922,21 @@ + return NULL; + } + #endif ++struct dentry *iopen_connect_dentry(struct dentry *de, struct inode *inode); + + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; + struct ext3_dir_entry_2 * de; + struct buffer_head * bh; ++ struct dentry *alternate = NULL; + + if (dentry->d_name.len > EXT3_NAME_LEN) + return ERR_PTR(-ENAMETOOLONG); + ++ if (ext3_check_for_iopen(dir, dentry)) ++ return NULL; ++ + bh = ext3_find_entry(dentry, &de); + inode = NULL; + if (bh) { +@@ -943,6 +948,12 @@ + return ERR_PTR(-EACCES); + } + } ++ ++ if (inode && (alternate = iopen_connect_dentry(dentry, inode))) { ++ iput(inode); ++ return alternate; ++ } ++ + d_add(dentry, inode); + return NULL; + } +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Sun Nov 16 01:19:22 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 01:27:31 2003 +@@ -864,6 +864,17 @@ + || !strcmp (this_char, "quota") + || !strcmp (this_char, "usrquota")) + /* Don't do anything ;-) */ ; ++ else if (!strcmp (this_char, "iopen")) { ++ set_opt (sbi->s_mount_opt, IOPEN); ++ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } else if (!strcmp (this_char, "noiopen")) { ++ clear_opt (sbi->s_mount_opt, IOPEN); ++ clear_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } ++ else if (!strcmp (this_char, "iopen_nopriv")) { ++ set_opt (sbi->s_mount_opt, IOPEN); ++ set_opt (sbi->s_mount_opt, IOPEN_NOPRIV); ++ } + else if (!strcmp (this_char, "journal")) { + /* @@@ FIXME */ + /* Eventually we will want to be able to create +Index: linux-2.4.19.SuSE/include/linux/ext3_fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/ext3_fs.h Sun Nov 16 01:25:42 2003 ++++ linux-2.4.19.SuSE/include/linux/ext3_fs.h Sun Nov 16 01:30:05 2003 +@@ -324,6 +324,8 @@ + #define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + #define EXT3_MOUNT_POSIX_ACL 0x8000 /* POSIX Access Control Lists */ + #define EXT3_MOUNT_ASYNCDEL 0x20000 /* Delayed deletion */ ++#define EXT3_MOUNT_IOPEN 0x40000 /* Allow access via iopen */ ++#define EXT3_MOUNT_IOPEN_NOPRIV 0x80000 /* Make iopen world-readable */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H diff --git a/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch b/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch new file mode 100644 index 0000000..8411137 --- /dev/null +++ b/lustre/kernel_patches/patches/jbd-flushtime-2.4.19-suse.patch @@ -0,0 +1,35 @@ +Index: linux-2.4.19.SuSE/fs/jbd/transaction.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/jbd/transaction.c Sun Nov 16 01:38:25 2003 ++++ linux-2.4.19.SuSE/fs/jbd/transaction.c Sun Nov 16 01:44:26 2003 +@@ -1094,7 +1094,6 @@ + + spin_lock(&journal_datalist_lock); + set_bit(BH_JBDDirty, &bh->b_state); +- set_buffer_flushtime(bh); + + J_ASSERT_JH(jh, jh->b_transaction != NULL); + +@@ -1995,6 +1994,13 @@ + spin_unlock(&journal_datalist_lock); + } + ++static void jbd_refile_buffer(struct buffer_head *bh) ++{ ++ if (buffer_dirty(bh) && (bh->b_list != BUF_DIRTY)) ++ set_buffer_flushtime(bh); ++ refile_buffer(bh); ++} ++ + /* + * Remove a buffer from its current buffer list in preparation for + * dropping it from its current transaction entirely. If the buffer has +@@ -2022,7 +2028,7 @@ + J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING); + } else { + /* Onto BUF_DIRTY for writeback */ +- refile_buffer(jh2bh(jh)); ++ jbd_refile_buffer(jh2bh(jh)); + } + } + diff --git a/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch b/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch new file mode 100644 index 0000000..26d3af9 --- /dev/null +++ b/lustre/kernel_patches/patches/linux-2.4.19-xattr-0.8.54-suse.patch @@ -0,0 +1,47 @@ + ext2/super.c | 3 +-- + ext3/ext3-exports.c | 13 +++++++++++++ + 2 files changed, 14 insertions(+), 2 deletions(-) + +Index: linux-2.4.19.SuSE/fs/ext2/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext2/super.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/ext2/super.c Sun Nov 16 00:40:59 2003 +@@ -70,6 +70,7 @@ + { + va_list args; + ++ ext2_xattr_put_super(sb); + if (!(sb->s_flags & MS_RDONLY)) { + sb->u.ext2_sb.s_mount_state |= EXT2_ERROR_FS; + sb->u.ext2_sb.s_es->s_state = +Index: linux-2.4.19.SuSE/fs/ext3/super.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/super.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/ext3/super.c Sun Nov 16 00:40:59 2003 +@@ -1822,8 +1828,6 @@ + exit_ext3_xattr(); + } + +-EXPORT_SYMBOL(ext3_force_commit); +-EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); +Index: linux-2.4.19.SuSE/fs/ext3/ext3-exports.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/ext3/ext3-exports.c Sun Nov 16 00:40:58 2003 ++++ linux-2.4.19.SuSE/fs/ext3/ext3-exports.c Sun Nov 16 00:40:59 2003 +@@ -0,0 +1,13 @@ ++#include ++#include ++#include ++#include ++#include ++ ++EXPORT_SYMBOL(ext3_force_commit); ++EXPORT_SYMBOL(ext3_bread); ++EXPORT_SYMBOL(ext3_xattr_register); ++EXPORT_SYMBOL(ext3_xattr_unregister); ++EXPORT_SYMBOL(ext3_xattr_get); ++EXPORT_SYMBOL(ext3_xattr_list); ++EXPORT_SYMBOL(ext3_xattr_set); diff --git a/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch b/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch new file mode 100644 index 0000000..4602f96 --- /dev/null +++ b/lustre/kernel_patches/patches/removepage-2.4.19-suse.patch @@ -0,0 +1,30 @@ + include/linux/fs.h | 1 + + mm/filemap.c | 3 +++ + 2 files changed, 4 insertions(+) + +Index: linux-2.4.19.SuSE/include/linux/fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs.h Sun Nov 16 00:40:59 2003 ++++ linux-2.4.19.SuSE/include/linux/fs.h Sun Nov 16 01:38:06 2003 +@@ -428,6 +428,7 @@ + int (*releasepage) (struct page *, int); + #define KERNEL_HAS_O_DIRECT /* this is for modules out of the kernel */ + int (*direct_IO)(int, struct file *, struct kiobuf *, unsigned long, int); ++ void (*removepage)(struct page *); /* called when page gets removed from the inode */ + }; + + struct address_space { +Index: linux-2.4.19.SuSE/mm/filemap.c +=================================================================== +--- linux-2.4.19.SuSE.orig/mm/filemap.c Sat Nov 15 18:02:15 2003 ++++ linux-2.4.19.SuSE/mm/filemap.c Sun Nov 16 01:37:11 2003 +@@ -97,6 +97,9 @@ + { + struct address_space * mapping = page->mapping; + ++ if (mapping->a_ops->removepage) ++ mapping->a_ops->removepage(page); ++ + mapping->nrpages--; + list_del(&page->list); + page->mapping = NULL; diff --git a/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch b/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch new file mode 100644 index 0000000..4e37bb8 --- /dev/null +++ b/lustre/kernel_patches/patches/vfs_intent-2.4.19-suse.patch @@ -0,0 +1,1835 @@ + fs/dcache.c | 19 ++ + fs/exec.c | 17 +- + fs/namei.c | 295 +++++++++++++++++++++++++++++++++++++++------- + fs/namespace.c | 28 +++- + fs/open.c | 172 +++++++++++++++++++------- + fs/stat.c | 52 +++++--- + include/linux/dcache.h | 60 +++++++++ + include/linux/fs.h | 32 ++++ + include/linux/fs_struct.h | 4 + kernel/exit.c | 3 + kernel/fork.c | 3 + kernel/ksyms.c | 1 + 12 files changed, 558 insertions(+), 128 deletions(-) + +Index: linux-2.4.19.SuSE/fs/exec.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/exec.c Mon Jan 27 05:08:35 2003 ++++ linux-2.4.19.SuSE/fs/exec.c Sat Nov 15 17:34:06 2003 +@@ -107,8 +107,10 @@ + struct file * file; + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + +- error = user_path_walk(library, &nd); ++ error = user_path_walk_it(library, &nd, &it); + if (error) + goto out; + +@@ -120,7 +122,8 @@ + if (error) + goto exit; + +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + error = PTR_ERR(file); + if (IS_ERR(file)) + goto out; +@@ -346,9 +349,11 @@ + struct inode *inode; + struct file *file; + int err = 0; ++ struct lookup_intent it = { .it_op = IT_OPEN, ++ .it_flags = FMODE_READ|FMODE_EXEC }; + + if (path_init(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- err = path_walk(name, &nd); ++ err = path_walk_it(name, &nd, &it); + file = ERR_PTR(err); + if (!err) { + inode = nd.dentry->d_inode; +@@ -360,7 +365,8 @@ + err = -EACCES; + file = ERR_PTR(err); + if (!err) { +- file = dentry_open(nd.dentry, nd.mnt, O_RDONLY); ++ file = dentry_open_it(nd.dentry, nd.mnt, O_RDONLY, &it); ++ intent_release(&it); + if (!IS_ERR(file)) { + err = deny_write_access(file); + if (err) { +@@ -372,6 +378,7 @@ + return file; + } + } ++ intent_release(&it); + path_release(&nd); + } + goto out; +@@ -981,7 +988,7 @@ + goto close_fail; + if (!file->f_op->write) + goto close_fail; +- if (do_truncate(file->f_dentry, 0) != 0) ++ if (do_truncate(file->f_dentry, 0, 0) != 0) + goto close_fail; + + retval = binfmt->core_dump(signr, regs, file); +Index: linux-2.4.19.SuSE/fs/dcache.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/dcache.c Mon Jan 27 05:08:04 2003 ++++ linux-2.4.19.SuSE/fs/dcache.c Sat Nov 15 17:29:03 2003 +@@ -186,6 +186,13 @@ + spin_unlock(&dcache_lock); + return 0; + } ++ ++ /* network invalidation by Lustre */ ++ if (dentry->d_flags & DCACHE_LUSTRE_INVALID) { ++ spin_unlock(&dcache_lock); ++ return 0; ++ } ++ + /* + * Check whether to do a partial shrink_dcache + * to get rid of unused child entries. +@@ -838,13 +845,19 @@ + * Adds a dentry to the hash according to its name. + */ + +-void d_rehash(struct dentry * entry) ++void __d_rehash(struct dentry * entry, int lock) + { + struct list_head *list = d_hash(entry->d_parent, entry->d_name.hash); + if (!list_empty(&entry->d_hash)) BUG(); +- spin_lock(&dcache_lock); ++ if (lock) spin_lock(&dcache_lock); + list_add(&entry->d_hash, list); +- spin_unlock(&dcache_lock); ++ if (lock) spin_unlock(&dcache_lock); ++} ++EXPORT_SYMBOL(__d_rehash); ++ ++void d_rehash(struct dentry * entry) ++{ ++ __d_rehash(entry, 1); + } + + #define do_switch(x,y) do { \ +Index: linux-2.4.19.SuSE/fs/namespace.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/namespace.c Mon Jan 27 05:08:07 2003 ++++ linux-2.4.19.SuSE/fs/namespace.c Sat Nov 15 17:56:42 2003 +@@ -97,6 +97,7 @@ + { + old_nd->dentry = mnt->mnt_mountpoint; + old_nd->mnt = mnt->mnt_parent; ++ UNPIN(old_nd->dentry, old_nd->mnt, 1); + mnt->mnt_parent = mnt; + mnt->mnt_mountpoint = mnt->mnt_root; + list_del_init(&mnt->mnt_child); +@@ -108,6 +109,7 @@ + { + mnt->mnt_parent = mntget(nd->mnt); + mnt->mnt_mountpoint = dget(nd->dentry); ++ PIN(nd->dentry, nd->mnt, 1); + list_add(&mnt->mnt_hash, mount_hashtable+hash(nd->mnt, nd->dentry)); + list_add(&mnt->mnt_child, &nd->mnt->mnt_mounts); + nd->dentry->d_mounted++; +@@ -491,15 +493,18 @@ + { + struct nameidata old_nd; + struct vfsmount *mnt = NULL; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int err = mount_is_safe(nd); + if (err) + return err; + if (!old_name || !*old_name) + return -EINVAL; + if (path_init(old_name, LOOKUP_POSITIVE|LOOKUP_FOLLOW, &old_nd)) +- err = path_walk(old_name, &old_nd); +- if (err) ++ err = path_walk_it(old_name, &old_nd, &it); ++ if (err) { ++ intent_release(&it); + return err; ++ } + + down_write(¤t->namespace->sem); + err = -EINVAL; +@@ -522,6 +527,7 @@ + } + + up_write(¤t->namespace->sem); ++ intent_release(&it); + path_release(&old_nd); + return err; + } +@@ -725,6 +731,7 @@ + unsigned long flags, void *data_page) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int retval = 0; + int mnt_flags = 0; + +@@ -750,9 +757,11 @@ + + /* ... and get the mountpoint */ + if (path_init(dir_name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, &nd)) +- retval = path_walk(dir_name, &nd); +- if (retval) ++ retval = path_walk_it(dir_name, &nd, &it); ++ if (retval) { ++ intent_release(&it); + return retval; ++ } + + if (flags & MS_REMOUNT) + retval = do_remount(&nd, flags & ~MS_REMOUNT, mnt_flags, +@@ -764,6 +773,8 @@ + else + retval = do_add_mount(&nd, type_page, flags, mnt_flags, + dev_name, data_page); ++ ++ intent_release(&it); + path_release(&nd); + return retval; + } +@@ -929,6 +940,8 @@ + { + struct vfsmount *tmp; + struct nameidata new_nd, old_nd, parent_nd, root_parent, user_nd; ++ struct lookup_intent new_it = { .it_op = IT_GETATTR }; ++ struct lookup_intent old_it = { .it_op = IT_GETATTR }; + char *name; + int error; + +@@ -943,7 +956,7 @@ + goto out0; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &new_nd)) +- error = path_walk(name, &new_nd); ++ error = path_walk_it(name, &new_nd, &new_it); + putname(name); + if (error) + goto out0; +@@ -957,7 +970,7 @@ + goto out1; + error = 0; + if (path_init(name, LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY, &old_nd)) +- error = path_walk(name, &old_nd); ++ error = path_walk_it(name, &old_nd, &old_it); + putname(name); + if (error) + goto out1; +@@ -1013,8 +1026,10 @@ + up(&old_nd.dentry->d_inode->i_zombie); + up_write(¤t->namespace->sem); + path_release(&user_nd); ++ intent_release(&old_it); + path_release(&old_nd); + out1: ++ intent_release(&new_it); + path_release(&new_nd); + out0: + unlock_kernel(); +Index: linux-2.4.19.SuSE/fs/namei.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/namei.c Mon Jan 27 05:08:07 2003 ++++ linux-2.4.19.SuSE/fs/namei.c Sat Nov 15 17:52:03 2003 +@@ -94,6 +94,13 @@ + * XEmacs seems to be relying on it... + */ + ++void intent_release(struct lookup_intent *it) ++{ ++ if (it && it->it_op_release) ++ it->it_op_release(it); ++ ++} ++ + /* In order to reduce some races, while at the same time doing additional + * checking and hopefully speeding things up, we copy filenames to the + * kernel data space before using them.. +@@ -260,10 +267,19 @@ + * Internal lookup() using the new generic dcache. + * SMP-safe + */ +-static struct dentry * cached_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *cached_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * dentry = d_lookup(parent, name); + ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ if (!dentry->d_op->d_revalidate_it(dentry, flags, it) && ++ !d_invalidate(dentry)) { ++ dput(dentry); ++ dentry = NULL; ++ } ++ return dentry; ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + if (!dentry->d_op->d_revalidate(dentry, flags) && !d_invalidate(dentry)) { + dput(dentry); +@@ -281,11 +297,15 @@ + * make sure that nobody added the entry to the dcache in the meantime.. + * SMP-safe + */ +-static struct dentry * real_lookup(struct dentry * parent, struct qstr * name, int flags) ++static struct dentry *real_lookup(struct dentry *parent, struct qstr *name, ++ int flags, struct lookup_intent *it) + { + struct dentry * result; + struct inode *dir = parent->d_inode; ++ int counter = 0; + ++again: ++ counter++; + down(&dir->i_sem); + /* + * First re-do the cached lookup just in case it was created +@@ -300,6 +320,9 @@ + result = ERR_PTR(-ENOMEM); + if (dentry) { + lock_kernel(); ++ if (dir->i_op->lookup_it) ++ result = dir->i_op->lookup_it(dir, dentry, it, flags); ++ else + result = dir->i_op->lookup(dir, dentry); + unlock_kernel(); + if (result) +@@ -321,6 +344,15 @@ + dput(result); + result = ERR_PTR(-ENOENT); + } ++ } else if (result->d_op && result->d_op->d_revalidate_it) { ++ if (!result->d_op->d_revalidate_it(result, flags, it) && ++ !d_invalidate(result)) { ++ dput(result); ++ if (counter > 10) ++ result = ERR_PTR(-ESTALE); ++ if (!IS_ERR(result)) ++ goto again; ++ } + } + return result; + } +@@ -332,7 +364,8 @@ + * Without that kind of total limit, nasty chains of consecutive + * symlinks can cause almost arbitrarily long lookups. + */ +-static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd) ++static inline int do_follow_link(struct dentry *dentry, struct nameidata *nd, ++ struct lookup_intent *it) + { + int err; + if (current->link_count >= 8) +@@ -346,10 +379,12 @@ + current->link_count++; + current->total_link_count++; + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + err = dentry->d_inode->i_op->follow_link(dentry, nd); + current->link_count--; + return err; + loop: ++ intent_release(it); + path_release(nd); + return -ELOOP; + } +@@ -447,7 +482,8 @@ + * + * We expect 'base' to be positive and a directory. + */ +-int link_path_walk(const char * name, struct nameidata *nd) ++int link_path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it) + { + struct dentry *dentry; + struct inode *inode; +@@ -524,12 +560,13 @@ + break; + } + /* This does the actual lookups.. */ +- dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = cached_lookup(nd->dentry, &this, LOOKUP_CONTINUE,NULL); + if (!dentry) { + err = -EWOULDBLOCKIO; + if (atomic) + break; +- dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE); ++ dentry = real_lookup(nd->dentry, &this, LOOKUP_CONTINUE, ++ NULL); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -547,7 +584,7 @@ + goto out_dput; + + if (inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, NULL); + dput(dentry); + if (err) + goto return_err; +@@ -563,7 +600,7 @@ + nd->dentry = dentry; + } + err = -ENOTDIR; +- if (!inode->i_op->lookup) ++ if (!inode->i_op->lookup && !inode->i_op->lookup_it) + break; + continue; + /* here ends the main loop */ +@@ -590,12 +627,12 @@ + if (err < 0) + break; + } +- dentry = cached_lookup(nd->dentry, &this, 0); ++ dentry = cached_lookup(nd->dentry, &this, 0, it); + if (!dentry) { + err = -EWOULDBLOCKIO; + if (atomic) + break; +- dentry = real_lookup(nd->dentry, &this, 0); ++ dentry = real_lookup(nd->dentry, &this, 0, it); + err = PTR_ERR(dentry); + if (IS_ERR(dentry)) + break; +@@ -605,7 +642,7 @@ + inode = dentry->d_inode; + if ((lookup_flags & LOOKUP_FOLLOW) + && inode && inode->i_op && inode->i_op->follow_link) { +- err = do_follow_link(dentry, nd); ++ err = do_follow_link(dentry, nd, it); + dput(dentry); + if (err) + goto return_err; +@@ -619,7 +656,8 @@ + goto no_inode; + if (lookup_flags & LOOKUP_DIRECTORY) { + err = -ENOTDIR; +- if (!inode->i_op || !inode->i_op->lookup) ++ if (!inode->i_op || ++ (!inode->i_op->lookup && !inode->i_op->lookup_it)) + break; + } + goto return_base; +@@ -643,6 +681,25 @@ + * Check the cached dentry for staleness. + */ + dentry = nd->dentry; ++ if (dentry && dentry->d_op && dentry->d_op->d_revalidate_it) { ++ err = -ESTALE; ++ if (!dentry->d_op->d_revalidate_it(dentry, 0, it)) { ++ struct dentry *new; ++ err = permission(dentry->d_parent->d_inode, ++ MAY_EXEC); ++ if (err) ++ break; ++ new = real_lookup(dentry->d_parent, ++ &dentry->d_name, 0, NULL); ++ d_invalidate(dentry); ++ dput(dentry); ++ if (IS_ERR(new)) { ++ err = PTR_ERR(new); ++ break; ++ } ++ nd->dentry = new; ++ } ++ } else + if (dentry && dentry->d_op && dentry->d_op->d_revalidate) { + err = -ESTALE; + if (!dentry->d_op->d_revalidate(dentry, lookup_flags & LOOKUP_PARENT)) { +@@ -656,15 +713,28 @@ + dput(dentry); + break; + } ++ if (err) ++ intent_release(it); + path_release(nd); + return_err: + return err; + } + ++int link_path_walk(const char * name, struct nameidata *nd) ++{ ++ return link_path_walk_it(name, nd, NULL); ++} ++ ++int path_walk_it(const char * name, struct nameidata *nd, struct lookup_intent *it) ++{ ++ current->total_link_count = 0; ++ return link_path_walk_it(name, nd, it); ++} ++ + int path_walk(const char * name, struct nameidata *nd) + { + current->total_link_count = 0; +- return link_path_walk(name, nd); ++ return link_path_walk_it(name, nd, NULL); + } + + /* SMP-safe */ +@@ -753,6 +823,7 @@ + { + nd->last_type = LAST_ROOT; /* if there are only slashes... */ + nd->flags = flags; ++ nd->intent = NULL; + if (*name=='/') + return walk_init_root(name,nd); + read_lock(¤t->fs->lock); +@@ -767,7 +838,8 @@ + * needs parent already locked. Doesn't follow mounts. + * SMP-safe. + */ +-struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++struct dentry * lookup_hash_it(struct qstr *name, struct dentry * base, ++ struct lookup_intent *it) + { + struct dentry * dentry; + struct inode *inode; +@@ -790,13 +862,16 @@ + goto out; + } + +- dentry = cached_lookup(base, name, 0); ++ dentry = cached_lookup(base, name, 0, it); + if (!dentry) { + struct dentry *new = d_alloc(base, name); + dentry = ERR_PTR(-ENOMEM); + if (!new) + goto out; + lock_kernel(); ++ if (inode->i_op->lookup_it) ++ dentry = inode->i_op->lookup_it(inode, new, it, 0); ++ else + dentry = inode->i_op->lookup(inode, new); + unlock_kernel(); + if (!dentry) +@@ -808,6 +883,12 @@ + return dentry; + } + ++struct dentry * lookup_hash(struct qstr *name, struct dentry * base) ++{ ++ return lookup_hash_it(name, base, NULL); ++} ++ ++ + /* SMP-safe */ + struct dentry * lookup_one_len(const char * name, struct dentry * base, int len) + { +@@ -829,7 +910,7 @@ + } + this.hash = end_name_hash(hash); + +- return lookup_hash(&this, base); ++ return lookup_hash_it(&this, base, NULL); + access: + return ERR_PTR(-EACCES); + } +@@ -861,6 +942,23 @@ + return err; + } + ++int __user_walk_it(const char *name, unsigned flags, struct nameidata *nd, ++ struct lookup_intent *it) ++{ ++ char *tmp; ++ int err; ++ ++ tmp = getname(name); ++ err = PTR_ERR(tmp); ++ if (!IS_ERR(tmp)) { ++ err = 0; ++ if (path_init(tmp, flags, nd)) ++ err = path_walk_it(tmp, nd, it); ++ putname(tmp); ++ } ++ return err; ++} ++ + /* + * It's inline, so penalty for filesystems that don't use sticky bit is + * minimal. +@@ -958,7 +1056,8 @@ + return retval; + } + +-int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++static int vfs_create_it(struct inode *dir, struct dentry *dentry, int mode, ++ struct lookup_intent *it) + { + int error; + +@@ -971,12 +1070,15 @@ + goto exit_lock; + + error = -EACCES; /* shouldn't it be ENOSYS? */ +- if (!dir->i_op || !dir->i_op->create) ++ if (!dir->i_op || (!dir->i_op->create && !dir->i_op->create_it)) + goto exit_lock; + + DQUOT_INIT(dir); + lock_kernel(); +- error = dir->i_op->create(dir, dentry, mode); ++ if (dir->i_op->create_it) ++ error = dir->i_op->create_it(dir, dentry, mode, it); ++ else ++ error = dir->i_op->create(dir, dentry, mode); + unlock_kernel(); + exit_lock: + up(&dir->i_zombie); +@@ -985,6 +1087,11 @@ + return error; + } + ++int vfs_create(struct inode *dir, struct dentry *dentry, int mode) ++{ ++ return vfs_create_it(dir, dentry, mode, NULL); ++} ++ + /* + * open_namei() + * +@@ -999,7 +1106,8 @@ + * for symlinks (where the permissions are checked later). + * SMP-safe + */ +-int open_namei(const char * pathname, int flag, int mode, struct nameidata *nd) ++int open_namei_it(const char *pathname, int flag, int mode, ++ struct nameidata *nd, struct lookup_intent *it) + { + int acc_mode, error = 0; + struct inode *inode; +@@ -1009,12 +1117,14 @@ + + acc_mode = ACC_MODE(flag); + ++ if (it) ++ it->it_flags = flag; + /* + * The simplest case - just a plain lookup. + */ + if (!(flag & O_CREAT)) { + if (path_init(pathname, lookup_flags(flag), nd)) +- error = path_walk(pathname, nd); ++ error = path_walk_it(pathname, nd, it); + if (error) + return error; + dentry = nd->dentry; +@@ -1024,6 +1134,10 @@ + /* + * Create - we need to know the parent. + */ ++ if (it) { ++ it->it_create_mode = mode; ++ it->it_op |= IT_CREAT; ++ } + if (path_init(pathname, LOOKUP_PARENT, nd)) + error = path_walk(pathname, nd); + if (error) +@@ -1040,7 +1154,7 @@ + + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + + do_last: + error = PTR_ERR(dentry); +@@ -1049,11 +1163,13 @@ + goto exit; + } + ++ it->it_create_mode = mode; + /* Negative dentry, just create the file */ + if (!dentry->d_inode) { + if (!IS_POSIXACL(dir->d_inode)) + mode &= ~current->fs->umask; +- error = vfs_create(dir->d_inode, dentry, mode); ++ error = vfs_create_it(dir->d_inode, dentry, ++ mode & ~current->fs->umask, it); + up(&dir->d_inode->i_sem); + #ifndef DENTRY_WASTE_RAM + if (error) +@@ -1161,7 +1277,7 @@ + if (!error) { + DQUOT_INIT(inode); + +- error = do_truncate(dentry, 0); ++ error = do_truncate(dentry, 0, 1); + } + put_write_access(inode); + if (error) +@@ -1173,8 +1289,10 @@ + return 0; + + exit_dput: ++ intent_release(it); + dput(dentry); + exit: ++ intent_release(it); + path_release(nd); + return error; + +@@ -1193,7 +1311,10 @@ + * are done. Procfs-like symlinks just set LAST_BIND. + */ + UPDATE_ATIME(dentry->d_inode); ++ nd->intent = it; + error = dentry->d_inode->i_op->follow_link(dentry, nd); ++ if (error) ++ intent_release(it); + dput(dentry); + if (error) + return error; +@@ -1215,13 +1336,20 @@ + } + dir = nd->dentry; + down(&dir->d_inode->i_sem); +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + putname(nd->last.name); + goto do_last; + } + ++int open_namei(const char *pathname, int flag, int mode, struct nameidata *nd) ++{ ++ return open_namei_it(pathname, flag, mode, nd, NULL); ++} ++ ++ + /* SMP-safe */ +-static struct dentry *lookup_create(struct nameidata *nd, int is_dir) ++static struct dentry *lookup_create(struct nameidata *nd, int is_dir, ++ struct lookup_intent *it) + { + struct dentry *dentry; + +@@ -1229,7 +1357,7 @@ + dentry = ERR_PTR(-EEXIST); + if (nd->last_type != LAST_NORM) + goto fail; +- dentry = lookup_hash(&nd->last, nd->dentry); ++ dentry = lookup_hash_it(&nd->last, nd->dentry, it); + if (IS_ERR(dentry)) + goto fail; + if (!is_dir && nd->last.name[nd->last.len] && !dentry->d_inode) +@@ -1286,7 +1414,16 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ ++ if (nd.dentry->d_inode->i_op->mknod_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mknod_raw(&nd, mode, dev); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + + if (!IS_POSIXACL(nd.dentry->d_inode)) +@@ -1308,6 +1445,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1356,7 +1494,14 @@ + error = path_walk(tmp, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 1); ++ if (nd.dentry->d_inode->i_op->mkdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->mkdir_raw(&nd, mode); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 1, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + if (!IS_POSIXACL(nd.dentry->d_inode)) +@@ -1365,6 +1510,7 @@ + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++out2: + path_release(&nd); + out: + putname(tmp); +@@ -1466,8 +1612,16 @@ + error = -EBUSY; + goto exit1; + } ++ if (nd.dentry->d_inode->i_op->rmdir_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ error = op->rmdir_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_rmdir(nd.dentry->d_inode, dentry); +@@ -1526,8 +1680,15 @@ + error = -EISDIR; + if (nd.last_type != LAST_NORM) + goto exit1; ++ if (nd.dentry->d_inode->i_op->unlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->unlink_raw(&nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit1; ++ } + down(&nd.dentry->d_inode->i_sem); +- dentry = lookup_hash(&nd.last, nd.dentry); ++ dentry = lookup_hash_it(&nd.last, nd.dentry, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + /* Why not before? Because we want correct error value */ +@@ -1595,15 +1756,23 @@ + error = path_walk(to, &nd); + if (error) + goto out; +- dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->symlink_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->symlink_raw(&nd, from); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out2; ++ } ++ dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(dentry); + if (!IS_ERR(dentry)) { + error = vfs_symlink(nd.dentry->d_inode, dentry, from); + dput(dentry); + } + up(&nd.dentry->d_inode->i_sem); ++ out2: + path_release(&nd); +-out: ++ out: + putname(to); + } + putname(from); +@@ -1686,7 +1855,14 @@ + error = -EXDEV; + if (old_nd.mnt != nd.mnt) + goto out_release; +- new_dentry = lookup_create(&nd, 0); ++ if (nd.dentry->d_inode->i_op->link_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ error = op->link_raw(&old_nd, &nd); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto out_release; ++ } ++ new_dentry = lookup_create(&nd, 0, NULL); + error = PTR_ERR(new_dentry); + if (!IS_ERR(new_dentry)) { + error = vfs_link(old_nd.dentry, nd.dentry->d_inode, new_dentry); +@@ -1732,7 +1908,7 @@ + * locking]. + */ + int vfs_rename_dir(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + struct inode *target; +@@ -1811,7 +1987,7 @@ + } + + int vfs_rename_other(struct inode *old_dir, struct dentry *old_dentry, +- struct inode *new_dir, struct dentry *new_dentry) ++ struct inode *new_dir, struct dentry *new_dentry) + { + int error; + +@@ -1902,9 +2078,18 @@ + if (newnd.last_type != LAST_NORM) + goto exit2; + ++ if (old_dir->d_inode->i_op->rename_raw) { ++ lock_kernel(); ++ error = old_dir->d_inode->i_op->rename_raw(&oldnd, &newnd); ++ unlock_kernel(); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto exit2; ++ } ++ + double_lock(new_dir, old_dir); + +- old_dentry = lookup_hash(&oldnd.last, old_dir); ++ old_dentry = lookup_hash_it(&oldnd.last, old_dir, NULL); + error = PTR_ERR(old_dentry); + if (IS_ERR(old_dentry)) + goto exit3; +@@ -1920,16 +2105,16 @@ + if (newnd.last.name[newnd.last.len]) + goto exit4; + } +- new_dentry = lookup_hash(&newnd.last, new_dir); ++ new_dentry = lookup_hash_it(&newnd.last, new_dir, NULL); + error = PTR_ERR(new_dentry); + if (IS_ERR(new_dentry)) + goto exit4; + ++ + lock_kernel(); + error = vfs_rename(old_dir->d_inode, old_dentry, + new_dir->d_inode, new_dentry); + unlock_kernel(); +- + dput(new_dentry); + exit4: + dput(old_dentry); +@@ -1980,20 +2165,26 @@ + } + + static inline int +-__vfs_follow_link(struct nameidata *nd, const char *link) ++__vfs_follow_link(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) + { + int res = 0; + char *name; + if (IS_ERR(link)) + goto fail; + ++ if (it == NULL) ++ it = nd->intent; ++ else if (it != nd->intent) ++ printk("it != nd->intent: tell phil@clusterfs.com\n"); ++ + if (*link == '/') { + path_release(nd); + if (!walk_init_root(link, nd)) + /* weird __emul_prefix() stuff did it */ + goto out; + } +- res = link_path_walk(link, nd); ++ res = link_path_walk_it(link, nd, it); + out: + if (current->link_count || res || nd->last_type!=LAST_NORM) + return res; +@@ -2015,7 +2206,13 @@ + + int vfs_follow_link(struct nameidata *nd, const char *link) + { +- return __vfs_follow_link(nd, link); ++ return __vfs_follow_link(nd, link, NULL); ++} ++ ++int vfs_follow_link_it(struct nameidata *nd, const char *link, ++ struct lookup_intent *it) ++{ ++ return __vfs_follow_link(nd, link, it); + } + + /* get the link contents into pagecache */ +@@ -2057,7 +2254,7 @@ + { + struct page *page = NULL; + char *s = page_getlink(dentry, &page); +- int res = __vfs_follow_link(nd, s); ++ int res = __vfs_follow_link(nd, s, NULL); + if (page) { + kunmap(page); + page_cache_release(page); +Index: linux-2.4.19.SuSE/fs/open.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/open.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/open.c Sat Nov 15 17:43:27 2003 +@@ -19,6 +19,8 @@ + #include + + #define special_file(m) (S_ISCHR(m)||S_ISBLK(m)||S_ISFIFO(m)||S_ISSOCK(m)) ++extern int path_walk_it(const char *name, struct nameidata *nd, ++ struct lookup_intent *it); + + int vfs_statfs(struct super_block *sb, struct statfs *buf) + { +@@ -95,9 +97,10 @@ + write_unlock(&files->file_lock); + } + +-int do_truncate(struct dentry *dentry, loff_t length) ++int do_truncate(struct dentry *dentry, loff_t length, int called_from_open) + { + struct inode *inode = dentry->d_inode; ++ struct inode_operations *op = dentry->d_inode->i_op; + int error; + struct iattr newattrs; + +@@ -108,7 +111,13 @@ + down(&inode->i_sem); + newattrs.ia_size = length; + newattrs.ia_valid = ATTR_SIZE | ATTR_CTIME; +- error = notify_change(dentry, &newattrs); ++ if (called_from_open) ++ newattrs.ia_valid |= ATTR_FROM_OPEN; ++ if (op->setattr_raw) { ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ } else ++ error = notify_change(dentry, &newattrs); + up(&inode->i_sem); + return error; + } +@@ -118,12 +127,13 @@ + struct nameidata nd; + struct inode * inode; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + error = -EINVAL; + if (length < 0) /* sorry, but loff_t says... */ + goto out; + +- error = user_path_walk(path, &nd); ++ error = user_path_walk_it(path, &nd, &it); + if (error) + goto out; + inode = nd.dentry->d_inode; +@@ -163,11 +173,13 @@ + error = locks_verify_truncate(inode, NULL, length); + if (!error) { + DQUOT_INIT(inode); +- error = do_truncate(nd.dentry, length); ++ intent_release(&it); ++ error = do_truncate(nd.dentry, length, 0); + } + put_write_access(inode); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -215,7 +227,7 @@ + + error = locks_verify_truncate(inode, file, length); + if (!error) +- error = do_truncate(dentry, length); ++ error = do_truncate(dentry, length, 0); + out_putf: + fput(file); + out: +@@ -260,11 +272,13 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -279,11 +293,25 @@ + goto dput_and_out; + + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!times) { + if (current->fsuid != inode->i_uid && + (error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; + } ++ + error = notify_change(nd.dentry, &newattrs); + dput_and_out: + path_release(&nd); +@@ -304,12 +332,14 @@ + struct inode * inode; + struct iattr newattrs; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, NULL); + + if (error) + goto out; + inode = nd.dentry->d_inode; + ++ /* this is safe without a Lustre lock because it only depends ++ on the super block */ + error = -EROFS; + if (IS_RDONLY(inode)) + goto dput_and_out; +@@ -324,7 +354,20 @@ + newattrs.ia_atime = times[0].tv_sec; + newattrs.ia_mtime = times[1].tv_sec; + newattrs.ia_valid |= ATTR_ATIME_SET | ATTR_MTIME_SET; +- } else { ++ } ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = nd.dentry->d_inode->i_op; ++ ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ goto dput_and_out; ++ } ++ ++ error = -EPERM; ++ if (!utimes) { + if (current->fsuid != inode->i_uid && + (error = permission(inode,MAY_WRITE)) != 0) + goto dput_and_out; +@@ -347,6 +390,7 @@ + int old_fsuid, old_fsgid; + kernel_cap_t old_cap; + int res; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + if (mode & ~S_IRWXO) /* where's F_OK, X_OK, W_OK, R_OK? */ + return -EINVAL; +@@ -364,13 +408,14 @@ + else + current->cap_effective = current->cap_permitted; + +- res = user_path_walk(filename, &nd); ++ res = user_path_walk_it(filename, &nd, &it); + if (!res) { + res = permission(nd.dentry->d_inode, mode); + /* SuS v2 requires we report a read only fs too */ + if(!res && (mode & S_IWOTH) && IS_RDONLY(nd.dentry->d_inode) + && !special_file(nd.dentry->d_inode->i_mode)) + res = -EROFS; ++ intent_release(&it); + path_release(&nd); + } + +@@ -386,6 +431,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -394,7 +440,7 @@ + + error = 0; + if (path_init(name,LOOKUP_POSITIVE|LOOKUP_FOLLOW|LOOKUP_DIRECTORY,&nd)) +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -406,6 +452,7 @@ + set_fs_pwd(current->fs, nd.mnt, nd.dentry); + + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; +@@ -446,6 +493,7 @@ + int error; + struct nameidata nd; + char *name; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + + name = getname(filename); + error = PTR_ERR(name); +@@ -454,7 +502,7 @@ + + path_init(name, LOOKUP_POSITIVE | LOOKUP_FOLLOW | + LOOKUP_DIRECTORY | LOOKUP_NOALT, &nd); +- error = path_walk(name, &nd); ++ error = path_walk_it(name, &nd, &it); + putname(name); + if (error) + goto out; +@@ -471,39 +519,56 @@ + set_fs_altroot(); + error = 0; + dput_and_out: ++ intent_release(&it); + path_release(&nd); + out: + return error; + } + +-asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++int chmod_common(struct dentry *dentry, mode_t mode) + { +- struct inode * inode; +- struct dentry * dentry; +- struct file * file; +- int err = -EBADF; ++ struct inode *inode = dentry->d_inode; + struct iattr newattrs; ++ int err = -EROFS; + +- file = fget(fd); +- if (!file) ++ if (IS_RDONLY(inode)) + goto out; + +- dentry = file->f_dentry; +- inode = dentry->d_inode; ++ if (inode->i_op->setattr_raw) { ++ newattrs.ia_mode = mode; ++ newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ err = inode->i_op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (err != -EOPNOTSUPP) ++ goto out; ++ } + +- err = -EROFS; +- if (IS_RDONLY(inode)) +- goto out_putf; + err = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto out_putf; ++ goto out; ++ + if (mode == (mode_t) -1) + mode = inode->i_mode; + newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); + newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; + err = notify_change(dentry, &newattrs); + +-out_putf: ++out: ++ return err; ++} ++ ++asmlinkage long sys_fchmod(unsigned int fd, mode_t mode) ++{ ++ struct file * file; ++ int err = -EBADF; ++ ++ file = fget(fd); ++ if (!file) ++ goto out; ++ ++ err = chmod_common(file->f_dentry, mode); ++ + fput(file); + out: + return err; +@@ -512,30 +577,14 @@ + asmlinkage long sys_chmod(const char * filename, mode_t mode) + { + struct nameidata nd; +- struct inode * inode; + int error; +- struct iattr newattrs; + + error = user_path_walk(filename, &nd); + if (error) + goto out; +- inode = nd.dentry->d_inode; +- +- error = -EROFS; +- if (IS_RDONLY(inode)) +- goto dput_and_out; + +- error = -EPERM; +- if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) +- goto dput_and_out; ++ error = chmod_common(nd.dentry, mode); + +- if (mode == (mode_t) -1) +- mode = inode->i_mode; +- newattrs.ia_mode = (mode & S_IALLUGO) | (inode->i_mode & ~S_IALLUGO); +- newattrs.ia_valid = ATTR_MODE | ATTR_CTIME; +- error = notify_change(nd.dentry, &newattrs); +- +-dput_and_out: + path_release(&nd); + out: + return error; +@@ -555,6 +604,20 @@ + error = -EROFS; + if (IS_RDONLY(inode)) + goto out; ++ ++ if (inode->i_op->setattr_raw) { ++ struct inode_operations *op = dentry->d_inode->i_op; ++ ++ newattrs.ia_uid = user; ++ newattrs.ia_gid = group; ++ newattrs.ia_valid = ATTR_UID | ATTR_GID | ATTR_CTIME; ++ newattrs.ia_valid |= ATTR_RAW; ++ error = op->setattr_raw(inode, &newattrs); ++ /* the file system wants to use normal vfs path now */ ++ if (error != -EOPNOTSUPP) ++ return error; ++ } ++ + error = -EPERM; + if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) + goto out; +@@ -659,6 +722,7 @@ + { + int namei_flags, error; + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_OPEN }; + + namei_flags = flags; + if ((namei_flags+1) & O_ACCMODE) +@@ -666,14 +730,15 @@ + if (namei_flags & O_TRUNC) + namei_flags |= 2; + +- error = open_namei(filename, namei_flags, mode, &nd); +- if (!error) +- return dentry_open(nd.dentry, nd.mnt, flags); ++ error = open_namei_it(filename, namei_flags, mode, &nd, &it); ++ if (error) ++ return ERR_PTR(error); + +- return ERR_PTR(error); ++ return dentry_open_it(nd.dentry, nd.mnt, flags, &it); + } + +-struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it) + { + struct file * f; + struct inode *inode; +@@ -710,7 +775,9 @@ + } + + if (f->f_op && f->f_op->open) { ++ f->f_it = it; + error = f->f_op->open(inode,f); ++ f->f_it = NULL; + if (error) + goto cleanup_all; + } +@@ -722,6 +789,7 @@ + !inode->i_mapping->a_ops->direct_IO)) + goto cleanup_all; + ++ intent_release(it); + return f; + + cleanup_all: +@@ -736,11 +804,17 @@ + cleanup_file: + put_filp(f); + cleanup_dentry: ++ intent_release(it); + dput(dentry); + mntput(mnt); + return ERR_PTR(error); + } + ++struct file *dentry_open(struct dentry *dentry, struct vfsmount *mnt, int flags) ++{ ++ return dentry_open_it(dentry, mnt, flags, NULL); ++} ++ + /* + * Find an empty file descriptor entry, and mark it busy. + */ +Index: linux-2.4.19.SuSE/fs/stat.c +=================================================================== +--- linux-2.4.19.SuSE.orig/fs/stat.c Mon Jan 27 05:08:00 2003 ++++ linux-2.4.19.SuSE/fs/stat.c Sat Nov 15 17:29:03 2003 +@@ -17,10 +17,14 @@ + * Revalidate the inode. This is required for proper NFS attribute caching. + */ + static __inline__ int +-do_revalidate(struct dentry *dentry) ++do_revalidate(struct dentry *dentry, struct lookup_intent *it) + { + struct inode * inode = dentry->d_inode; +- if (inode->i_op && inode->i_op->revalidate) ++ if (!inode) ++ return -ENOENT; ++ if (inode->i_op && inode->i_op->revalidate_it) ++ return inode->i_op->revalidate_it(dentry, it); ++ else if (inode->i_op && inode->i_op->revalidate) + return inode->i_op->revalidate(dentry); + return 0; + } +@@ -141,13 +145,15 @@ + asmlinkage long sys_stat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -157,13 +163,15 @@ + asmlinkage long sys_newstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -178,13 +186,15 @@ + asmlinkage long sys_lstat(char * filename, struct __old_kernel_stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_old_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -195,13 +205,15 @@ + asmlinkage long sys_newlstat(char * filename, struct stat * statbuf) + { + struct nameidata nd; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + int error; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -222,7 +234,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_old_stat(dentry->d_inode, statbuf); + fput(f); +@@ -241,7 +253,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat(dentry->d_inode, statbuf); + fput(f); +@@ -263,7 +275,7 @@ + + error = -EINVAL; + if (inode->i_op && inode->i_op->readlink && +- !(error = do_revalidate(nd.dentry))) { ++ !(error = do_revalidate(nd.dentry, NULL))) { + UPDATE_ATIME(inode); + error = inode->i_op->readlink(nd.dentry, buf, bufsiz); + } +@@ -339,12 +351,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk(filename, &nd); ++ error = user_path_walk_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -354,12 +368,14 @@ + { + struct nameidata nd; + int error; ++ struct lookup_intent it = { .it_op = IT_GETATTR }; + +- error = user_path_walk_link(filename, &nd); ++ error = user_path_walk_link_it(filename, &nd, &it); + if (!error) { +- error = do_revalidate(nd.dentry); ++ error = do_revalidate(nd.dentry, &it); + if (!error) + error = cp_new_stat64(nd.dentry->d_inode, statbuf); ++ intent_release(&it); + path_release(&nd); + } + return error; +@@ -374,7 +390,7 @@ + if (f) { + struct dentry * dentry = f->f_dentry; + +- err = do_revalidate(dentry); ++ err = do_revalidate(dentry, NULL); + if (!err) + err = cp_new_stat64(dentry->d_inode, statbuf); + fput(f); +Index: linux-2.4.19.SuSE/include/linux/dcache.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/dcache.h Mon Jan 27 05:13:15 2003 ++++ linux-2.4.19.SuSE/include/linux/dcache.h Sat Nov 15 17:35:46 2003 +@@ -5,6 +5,51 @@ + + #include + #include ++#include ++ ++#define IT_OPEN 0x0001 ++#define IT_CREAT 0x0002 ++#define IT_READDIR 0x0004 ++#define IT_GETATTR 0x0008 ++#define IT_LOOKUP 0x0010 ++#define IT_UNLINK 0x0020 ++#define IT_GETXATTR 0x0040 ++#define IT_EXEC 0x0080 ++#define IT_PIN 0x0100 ++ ++#define IT_FL_LOCKED 0x0001 ++#define IT_FL_FOLLOWED 0x0002 /* set by vfs_follow_link */ ++ ++#define INTENT_MAGIC 0x19620323 ++ ++ ++struct lustre_intent_data { ++ int it_disposition; ++ int it_status; ++ __u64 it_lock_handle; ++ void *it_data; ++ int it_lock_mode; ++ int it_int_flags; ++}; ++struct lookup_intent { ++ int it_magic; ++ void (*it_op_release)(struct lookup_intent *); ++ int it_op; ++ int it_flags; ++ int it_create_mode; ++ union { ++ struct lustre_intent_data lustre; ++ } d; ++}; ++ ++static inline void intent_init(struct lookup_intent *it, int op, int flags) ++{ ++ memset(it, 0, sizeof(*it)); ++ it->it_magic = INTENT_MAGIC; ++ it->it_op = op; ++ it->it_flags = flags; ++} ++ + + /* + * linux/include/linux/dcache.h +@@ -92,8 +137,22 @@ + int (*d_delete)(struct dentry *); + void (*d_release)(struct dentry *); + void (*d_iput)(struct dentry *, struct inode *); ++ int (*d_revalidate_it)(struct dentry *, int, struct lookup_intent *); ++ void (*d_pin)(struct dentry *, struct vfsmount * , int); ++ void (*d_unpin)(struct dentry *, struct vfsmount *, int); + }; + ++#define PIN(de,mnt,flag) if (de->d_op && de->d_op->d_pin) \ ++ de->d_op->d_pin(de, mnt, flag); ++#define UNPIN(de,mnt,flag) if (de->d_op && de->d_op->d_unpin) \ ++ de->d_op->d_unpin(de, mnt, flag); ++ ++ ++/* defined in fs/namei.c */ ++extern void intent_release(struct lookup_intent *it); ++/* defined in fs/dcache.c */ ++extern void __d_rehash(struct dentry * entry, int lock); ++ + /* the dentry parameter passed to d_hash and d_compare is the parent + * directory of the entries to be compared. It is used in case these + * functions need any directory specific information for determining +@@ -125,6 +184,7 @@ + * s_nfsd_free_path semaphore will be down + */ + #define DCACHE_REFERENCED 0x0008 /* Recently used, don't discard. */ ++#define DCACHE_LUSTRE_INVALID 0x0010 /* Lustre invalidated */ + + extern spinlock_t dcache_lock; + +Index: linux-2.4.19.SuSE/include/linux/fs.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs.h Sat Nov 15 17:25:06 2003 ++++ linux-2.4.19.SuSE/include/linux/fs.h Sat Nov 15 17:29:03 2003 +@@ -73,6 +73,7 @@ + + #define FMODE_READ 1 + #define FMODE_WRITE 2 ++#define FMODE_EXEC 4 + + #define READ 0 + #define WRITE 1 +@@ -363,6 +364,9 @@ + #define ATTR_MTIME_SET 256 + #define ATTR_FORCE 512 /* Not a change, but a change it */ + #define ATTR_ATTR_FLAG 1024 ++#define ATTR_RAW 0x0800 /* file system, not vfs will massage attrs */ ++#define ATTR_FROM_OPEN 0x1000 /* called from open path, ie O_TRUNC */ ++#define ATTR_CTIME_SET 0x2000 + + /* + * This is the Inode Attributes structure, used for notify_change(). It +@@ -507,6 +511,7 @@ + struct pipe_inode_info *i_pipe; + struct block_device *i_bdev; + struct char_device *i_cdev; ++ void *i_filterdata; + + unsigned long i_dnotify_mask; /* Directory notify events */ + struct dnotify_struct *i_dnotify; /* for directory notifications */ +@@ -669,6 +674,7 @@ + + /* needed for tty driver, and maybe others */ + void *private_data; ++ struct lookup_intent *f_it; + + /* preallocated helper kiobuf to speedup O_DIRECT */ + struct kiobuf *f_iobuf; +@@ -799,6 +805,7 @@ + struct qstr last; + unsigned int flags; + int last_type; ++ struct lookup_intent *intent; + }; + + #define DQUOT_USR_ENABLED 0x01 /* User diskquotas enabled */ +@@ -947,7 +954,8 @@ + extern int __vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_rmdir(struct inode *, struct dentry *); + extern int vfs_unlink(struct inode *, struct dentry *); +-extern int vfs_rename(struct inode *, struct dentry *, struct inode *, struct dentry *); ++int vfs_rename(struct inode *old_dir, struct dentry *old_dentry, ++ struct inode *new_dir, struct dentry *new_dentry); + + /* + * File types +@@ -1020,21 +1028,32 @@ + + struct inode_operations { + int (*create) (struct inode *,struct dentry *,int); ++ int (*create_it) (struct inode *,struct dentry *,int, struct lookup_intent *); + struct dentry * (*lookup) (struct inode *,struct dentry *); ++ struct dentry * (*lookup_it) (struct inode *,struct dentry *, struct lookup_intent *, int flags); + int (*link) (struct dentry *,struct inode *,struct dentry *); ++ int (*link_raw) (struct nameidata *,struct nameidata *); + int (*unlink) (struct inode *,struct dentry *); ++ int (*unlink_raw) (struct nameidata *); + int (*symlink) (struct inode *,struct dentry *,const char *); ++ int (*symlink_raw) (struct nameidata *,const char *); + int (*mkdir) (struct inode *,struct dentry *,int); ++ int (*mkdir_raw) (struct nameidata *,int); + int (*rmdir) (struct inode *,struct dentry *); ++ int (*rmdir_raw) (struct nameidata *); + int (*mknod) (struct inode *,struct dentry *,int,int); ++ int (*mknod_raw) (struct nameidata *,int,dev_t); + int (*rename) (struct inode *, struct dentry *, + struct inode *, struct dentry *); ++ int (*rename_raw) (struct nameidata *, struct nameidata *); + int (*readlink) (struct dentry *, char *,int); + int (*follow_link) (struct dentry *, struct nameidata *); + void (*truncate) (struct inode *); + int (*permission) (struct inode *, int); + int (*revalidate) (struct dentry *); ++ int (*revalidate_it) (struct dentry *, struct lookup_intent *); + int (*setattr) (struct dentry *, struct iattr *); ++ int (*setattr_raw) (struct inode *, struct iattr *); + int (*getattr) (struct dentry *, struct iattr *); + int (*setxattr) (struct dentry *, const char *, const void *, size_t, int); + ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); +@@ -1244,10 +1263,14 @@ + + asmlinkage long sys_open(const char *, int, int); + asmlinkage long sys_close(unsigned int); /* yes, it's really unsigned */ +-extern int do_truncate(struct dentry *, loff_t start); ++extern int do_truncate(struct dentry *, loff_t start, int called_from_open); + + extern struct file *filp_open(const char *, int, int); + extern struct file * dentry_open(struct dentry *, struct vfsmount *, int); ++extern int open_namei_it(const char *filename, int namei_flags, int mode, ++ struct nameidata *nd, struct lookup_intent *it); ++extern struct file *dentry_open_it(struct dentry *dentry, struct vfsmount *mnt, ++ int flags, struct lookup_intent *it); + extern int filp_close(struct file *, fl_owner_t id); + extern char * getname(const char *); + +@@ -1515,6 +1538,7 @@ + extern loff_t default_llseek(struct file *file, loff_t offset, int origin); + + extern int FASTCALL(__user_walk(const char *, unsigned, struct nameidata *)); ++extern int FASTCALL(__user_walk_it(const char *, unsigned, struct nameidata *, struct lookup_intent *it)); + extern int FASTCALL(path_init(const char *, unsigned, struct nameidata *)); + extern int FASTCALL(path_walk(const char *, struct nameidata *)); + extern int FASTCALL(link_path_walk(const char *, struct nameidata *)); +@@ -1526,6 +1550,8 @@ + extern struct dentry * lookup_hash(struct qstr *, struct dentry *); + #define user_path_walk(name,nd) __user_walk(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd) + #define user_path_walk_link(name,nd) __user_walk(name, LOOKUP_POSITIVE, nd) ++#define user_path_walk_it(name,nd,it) __user_walk_it(name, LOOKUP_FOLLOW|LOOKUP_POSITIVE, nd, it) ++#define user_path_walk_link_it(name,nd,it) __user_walk_it(name, LOOKUP_POSITIVE, nd, it) + + extern void iput(struct inode *); + extern void force_delete(struct inode *); +@@ -1646,6 +1672,8 @@ + + extern int vfs_readlink(struct dentry *, char *, int, const char *); + extern int vfs_follow_link(struct nameidata *, const char *); ++extern int vfs_follow_link_it(struct nameidata *, const char *, ++ struct lookup_intent *it); + extern int page_readlink(struct dentry *, char *, int); + extern int page_follow_link(struct dentry *, struct nameidata *); + extern struct inode_operations page_symlink_inode_operations; +Index: linux-2.4.19.SuSE/include/linux/fs_struct.h +=================================================================== +--- linux-2.4.19.SuSE.orig/include/linux/fs_struct.h Fri Jul 13 15:10:44 2001 ++++ linux-2.4.19.SuSE/include/linux/fs_struct.h Sat Nov 15 17:29:03 2003 +@@ -34,10 +34,12 @@ + write_lock(&fs->lock); + old_root = fs->root; + old_rootmnt = fs->rootmnt; ++ PIN(dentry, mnt, 1); + fs->rootmnt = mntget(mnt); + fs->root = dget(dentry); + write_unlock(&fs->lock); + if (old_root) { ++ UNPIN(old_root, old_rootmnt, 1); + dput(old_root); + mntput(old_rootmnt); + } +@@ -57,10 +59,12 @@ + write_lock(&fs->lock); + old_pwd = fs->pwd; + old_pwdmnt = fs->pwdmnt; ++ PIN(dentry, mnt, 0); + fs->pwdmnt = mntget(mnt); + fs->pwd = dget(dentry); + write_unlock(&fs->lock); + if (old_pwd) { ++ UNPIN(old_pwd, old_pwdmnt, 0); + dput(old_pwd); + mntput(old_pwdmnt); + } +Index: linux-2.4.19.SuSE/kernel/ksyms.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/ksyms.c Sat Nov 15 17:24:46 2003 ++++ linux-2.4.19.SuSE/kernel/ksyms.c Sat Nov 15 17:29:03 2003 +@@ -315,6 +315,7 @@ + EXPORT_SYMBOL(set_page_dirty); + EXPORT_SYMBOL(vfs_readlink); + EXPORT_SYMBOL(vfs_follow_link); ++EXPORT_SYMBOL(vfs_follow_link_it); + EXPORT_SYMBOL(page_readlink); + EXPORT_SYMBOL(page_follow_link); + EXPORT_SYMBOL(page_symlink_inode_operations); +Index: linux-2.4.19.SuSE/kernel/fork.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/fork.c Mon Jan 27 05:08:56 2003 ++++ linux-2.4.19.SuSE/kernel/fork.c Sat Nov 15 17:29:03 2003 +@@ -454,10 +454,13 @@ + fs->umask = old->umask; + read_lock(&old->lock); + fs->rootmnt = mntget(old->rootmnt); ++ PIN(old->pwd, old->pwdmnt, 0); ++ PIN(old->root, old->rootmnt, 1); + fs->root = dget(old->root); + fs->pwdmnt = mntget(old->pwdmnt); + fs->pwd = dget(old->pwd); + if (old->altroot) { ++ PIN(old->altroot, old->altrootmnt, 1); + fs->altrootmnt = mntget(old->altrootmnt); + fs->altroot = dget(old->altroot); + } else { +Index: linux-2.4.19.SuSE/kernel/exit.c +=================================================================== +--- linux-2.4.19.SuSE.orig/kernel/exit.c Mon Jan 27 05:08:16 2003 ++++ linux-2.4.19.SuSE/kernel/exit.c Sat Nov 15 17:29:03 2003 +@@ -288,11 +288,14 @@ + { + /* No need to hold fs->lock if we are killing it */ + if (atomic_dec_and_test(&fs->count)) { ++ UNPIN(fs->pwd, fs->pwdmnt, 0); ++ UNPIN(fs->root, fs->rootmnt, 1); + dput(fs->root); + mntput(fs->rootmnt); + dput(fs->pwd); + mntput(fs->pwdmnt); + if (fs->altroot) { ++ UNPIN(fs->altroot, fs->altrootmnt, 1); + dput(fs->altroot); + mntput(fs->altrootmnt); + }