From eb1dbbc8c558758bf7edeee30ede6cb0dd0c4a17 Mon Sep 17 00:00:00 2001 From: braam Date: Sat, 5 Apr 2003 11:15:56 +0000 Subject: [PATCH] - add ext3 patches to the kernel patch instead of extN directory, and associated changes. This branch will require a small amount of debugging, cleanup and some testing before landing. --- lustre/kernel_patches/patches/exports_2.4.20.patch | 7 +- .../kernel_patches/patches/ext-2.4-patch-1.patch | 2527 +++++++++ .../kernel_patches/patches/ext-2.4-patch-2.patch | 34 + .../kernel_patches/patches/ext-2.4-patch-3.patch | 96 + .../kernel_patches/patches/ext-2.4-patch-4.patch | 48 + lustre/kernel_patches/patches/ext3-2.4-ino_t.patch | 138 + .../kernel_patches/patches/ext3-2.4.18-fixes.patch | 353 ++ .../patches/ext3-2.4.18-ino_sb_macro.patch | 1554 ++++++ .../kernel_patches/patches/ext3-2.5-noread.patch | 266 + lustre/kernel_patches/patches/ext3-largefile.patch | 23 + .../kernel_patches/patches/ext3-orphan_lock.patch | 79 + .../patches/ext3-truncate_blocks.patch | 92 + .../kernel_patches/patches/ext3-unmount_sync.patch | 59 + .../patches/ext3-use-after-free.patch | 65 + .../patches/extN-2.4.18-exports.patch | 11 + .../patches/extN-2.4.18-ino_sb_fixup.patch | 33 + .../patches/extN-delete_thread.patch | 274 + .../kernel_patches/patches/extN-iget-debug.patch | 48 + .../kernel_patches/patches/extN-misc-fixup.patch | 23 + lustre/kernel_patches/patches/extN-noread.patch | 233 + lustre/kernel_patches/patches/extN-san.patch | 88 + lustre/kernel_patches/patches/extN-wantedi.diff | 163 + lustre/kernel_patches/patches/extN-wantedi.patch | 169 + .../patches/extfs-update-2.4.20rc1.patch | 4391 +++++++++++++++ .../kernel_patches/patches/htree-ext3-2.4.18.patch | 1201 +++++ .../patches/linux-2.4.18ea-0.8.26.patch | 1753 ++++++ .../patches/linux-2.4.20-xattr-0.8.54.patch | 5578 ++++++++++++++++++++ .../patches/patch-2.4.18-chaos22.patch | 165 + lustre/kernel_patches/pc/ext-2.4-patch-1.pc | 11 + lustre/kernel_patches/pc/ext-2.4-patch-2.pc | 1 + lustre/kernel_patches/pc/ext-2.4-patch-3.pc | 3 + lustre/kernel_patches/pc/ext-2.4-patch-4.pc | 1 + lustre/kernel_patches/pc/ext3-2.4-ino_t.pc | 3 + lustre/kernel_patches/pc/ext3-2.4.18-fixes.pc | 7 + .../kernel_patches/pc/ext3-2.4.18-ino_sb_macro.pc | 10 + lustre/kernel_patches/pc/ext3-2.5-noread.pc | 3 + lustre/kernel_patches/pc/ext3-largefile.pc | 1 + lustre/kernel_patches/pc/ext3-orphan_lock.pc | 3 + lustre/kernel_patches/pc/ext3-truncate_blocks.pc | 1 + lustre/kernel_patches/pc/ext3-unmount_sync.pc | 1 + lustre/kernel_patches/pc/ext3-use-after-free.pc | 1 + lustre/kernel_patches/pc/extN-2.4.18-exports.pc | 1 + .../kernel_patches/pc/extN-2.4.18-ino_sb_fixup.pc | 1 + lustre/kernel_patches/pc/extN-delete_thread.pc | 3 + lustre/kernel_patches/pc/extN-iget-debug.pc | 2 + lustre/kernel_patches/pc/extN-misc-fixup.pc | 1 + lustre/kernel_patches/pc/extN-noread.pc | 3 + lustre/kernel_patches/pc/extN-san.pc | 1 + lustre/kernel_patches/pc/extN-wantedi.pc | 4 + lustre/kernel_patches/pc/htree-ext3-2.4.18.pc | 4 + lustre/kernel_patches/pc/linux-2.4.18ea-0.8.26.pc | 9 + .../kernel_patches/pc/linux-2.4.20-xattr-0.8.54.pc | 63 + lustre/kernel_patches/series/vanilla-2.4.20 | 6 + 53 files changed, 19612 insertions(+), 3 deletions(-) create mode 100644 lustre/kernel_patches/patches/ext-2.4-patch-1.patch create mode 100644 lustre/kernel_patches/patches/ext-2.4-patch-2.patch create mode 100644 lustre/kernel_patches/patches/ext-2.4-patch-3.patch create mode 100644 lustre/kernel_patches/patches/ext-2.4-patch-4.patch create mode 100644 lustre/kernel_patches/patches/ext3-2.4-ino_t.patch create mode 100644 lustre/kernel_patches/patches/ext3-2.4.18-fixes.patch create mode 100644 lustre/kernel_patches/patches/ext3-2.4.18-ino_sb_macro.patch create mode 100644 lustre/kernel_patches/patches/ext3-2.5-noread.patch create mode 100644 lustre/kernel_patches/patches/ext3-largefile.patch create mode 100644 lustre/kernel_patches/patches/ext3-orphan_lock.patch create mode 100644 lustre/kernel_patches/patches/ext3-truncate_blocks.patch create mode 100644 lustre/kernel_patches/patches/ext3-unmount_sync.patch create mode 100644 lustre/kernel_patches/patches/ext3-use-after-free.patch create mode 100644 lustre/kernel_patches/patches/extN-2.4.18-exports.patch create mode 100644 lustre/kernel_patches/patches/extN-2.4.18-ino_sb_fixup.patch create mode 100644 lustre/kernel_patches/patches/extN-delete_thread.patch create mode 100644 lustre/kernel_patches/patches/extN-iget-debug.patch create mode 100644 lustre/kernel_patches/patches/extN-misc-fixup.patch create mode 100644 lustre/kernel_patches/patches/extN-noread.patch create mode 100644 lustre/kernel_patches/patches/extN-san.patch create mode 100644 lustre/kernel_patches/patches/extN-wantedi.diff create mode 100644 lustre/kernel_patches/patches/extN-wantedi.patch create mode 100644 lustre/kernel_patches/patches/extfs-update-2.4.20rc1.patch create mode 100644 lustre/kernel_patches/patches/htree-ext3-2.4.18.patch create mode 100644 lustre/kernel_patches/patches/linux-2.4.18ea-0.8.26.patch create mode 100644 lustre/kernel_patches/patches/linux-2.4.20-xattr-0.8.54.patch create mode 100644 lustre/kernel_patches/patches/patch-2.4.18-chaos22.patch create mode 100644 lustre/kernel_patches/pc/ext-2.4-patch-1.pc create mode 100644 lustre/kernel_patches/pc/ext-2.4-patch-2.pc create mode 100644 lustre/kernel_patches/pc/ext-2.4-patch-3.pc create mode 100644 lustre/kernel_patches/pc/ext-2.4-patch-4.pc create mode 100644 lustre/kernel_patches/pc/ext3-2.4-ino_t.pc create mode 100644 lustre/kernel_patches/pc/ext3-2.4.18-fixes.pc create mode 100644 lustre/kernel_patches/pc/ext3-2.4.18-ino_sb_macro.pc create mode 100644 lustre/kernel_patches/pc/ext3-2.5-noread.pc create mode 100644 lustre/kernel_patches/pc/ext3-largefile.pc create mode 100644 lustre/kernel_patches/pc/ext3-orphan_lock.pc create mode 100644 lustre/kernel_patches/pc/ext3-truncate_blocks.pc create mode 100644 lustre/kernel_patches/pc/ext3-unmount_sync.pc create mode 100644 lustre/kernel_patches/pc/ext3-use-after-free.pc create mode 100644 lustre/kernel_patches/pc/extN-2.4.18-exports.pc create mode 100644 lustre/kernel_patches/pc/extN-2.4.18-ino_sb_fixup.pc create mode 100644 lustre/kernel_patches/pc/extN-delete_thread.pc create mode 100644 lustre/kernel_patches/pc/extN-iget-debug.pc create mode 100644 lustre/kernel_patches/pc/extN-misc-fixup.pc create mode 100644 lustre/kernel_patches/pc/extN-noread.pc create mode 100644 lustre/kernel_patches/pc/extN-san.pc create mode 100644 lustre/kernel_patches/pc/extN-wantedi.pc create mode 100644 lustre/kernel_patches/pc/htree-ext3-2.4.18.pc create mode 100644 lustre/kernel_patches/pc/linux-2.4.18ea-0.8.26.pc create mode 100644 lustre/kernel_patches/pc/linux-2.4.20-xattr-0.8.54.pc diff --git a/lustre/kernel_patches/patches/exports_2.4.20.patch b/lustre/kernel_patches/patches/exports_2.4.20.patch index 5adba64..6332735 100644 --- a/lustre/kernel_patches/patches/exports_2.4.20.patch +++ b/lustre/kernel_patches/patches/exports_2.4.20.patch @@ -20,7 +20,7 @@ obj-m := $(O_TARGET) --- linux-2.4.20-hp4_pnnl1/fs/ext3/super.c~exports_hp Tue Apr 1 20:36:07 2003 +++ linux-2.4.20-hp4_pnnl1-braam/fs/ext3/super.c Tue Apr 1 20:36:07 2003 -@@ -1769,7 +1769,7 @@ static void __exit exit_ext3_fs(void) +@@ -1769,7 +1769,7 @@ unregister_filesystem(&ext3_fs_type); } @@ -31,7 +31,7 @@ MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); --- linux-2.4.20-hp4_pnnl1/include/linux/fs.h~exports_hp Tue Apr 1 20:36:07 2003 +++ linux-2.4.20-hp4_pnnl1-braam/include/linux/fs.h Tue Apr 1 20:36:52 2003 -@@ -1020,6 +1020,7 @@ extern int unregister_filesystem(struct +@@ -1020,6 +1020,7 @@ extern struct vfsmount *kern_mount(struct file_system_type *); extern int may_umount(struct vfsmount *); extern long do_mount(char *, char *, char *, unsigned long, void *); @@ -41,12 +41,13 @@ --- linux-2.4.20-hp4_pnnl1/kernel/ksyms.c~exports_hp Tue Apr 1 20:36:07 2003 +++ linux-2.4.20-hp4_pnnl1-braam/kernel/ksyms.c Tue Apr 1 20:36:07 2003 -@@ -308,6 +308,10 @@ EXPORT_SYMBOL(dcache_dir_fsync); +@@ -308,6 +308,11 @@ EXPORT_SYMBOL(dcache_readdir); EXPORT_SYMBOL(dcache_dir_ops); +/* lustre */ +EXPORT_SYMBOL(pagecache_lock_cacheline); ++EXPORT_SYMBOL(panic_notifier_list); +EXPORT_SYMBOL(do_kern_mount); + /* for stackable file systems (lofs, wrapfs, cryptfs, etc.) */ diff --git a/lustre/kernel_patches/patches/ext-2.4-patch-1.patch b/lustre/kernel_patches/patches/ext-2.4-patch-1.patch new file mode 100644 index 0000000..b1b9048 --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-1.patch @@ -0,0 +1,2527 @@ + 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 +951,300 @@ 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->info; ++ 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/ext-2.4-patch-2.patch b/lustre/kernel_patches/patches/ext-2.4-patch-2.patch new file mode 100644 index 0000000..689d33b --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-2.patch @@ -0,0 +1,34 @@ +# This is a BitKeeper generated patch for the following project: +# Project Name: Linux kernel tree +# +# namei.c | 9 +++++++++ +# 1 files changed, 9 insertions(+) +# +# The following is the BitKeeper ChangeSet Log +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.777 +# Add '.' and '..' entries to be returned by readdir of htree directories +# +# This patch from Chris Li adds '.' and '..' to the rbtree so that they +# are properly returned by readdir. +# -------------------------------------------- +# +diff -Nru a/fs/ext3/namei.c b/fs/ext3/namei.c +--- a/fs/ext3/namei.c Thu Nov 7 10:57:30 2002 ++++ b/fs/ext3/namei.c Thu Nov 7 10:57:30 2002 +@@ -546,6 +546,15 @@ + if (!frame) + return err; + ++ /* Add '.' and '..' from the htree header */ ++ if (!start_hash && !start_minor_hash) { ++ de = (struct ext3_dir_entry_2 *) frames[0].bh->b_data; ++ ext3_htree_store_dirent(dir_file, 0, 0, de); ++ de = ext3_next_entry(de); ++ ext3_htree_store_dirent(dir_file, 0, 0, de); ++ count += 2; ++ } ++ + while (1) { + block = dx_get_block(frame->at); + dxtrace(printk("Reading block %d\n", block)); diff --git a/lustre/kernel_patches/patches/ext-2.4-patch-3.patch b/lustre/kernel_patches/patches/ext-2.4-patch-3.patch new file mode 100644 index 0000000..2600b02 --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-3.patch @@ -0,0 +1,96 @@ +# This is a BitKeeper generated patch for the following project: +# Project Name: Linux kernel tree +# +# fs/ext3/dir.c | 7 +++++-- +# fs/ext3/namei.c | 11 +++++++---- +# include/linux/ext3_fs.h | 2 +- +# 3 files changed, 13 insertions(+), 7 deletions(-) +# +# The following is the BitKeeper ChangeSet Log +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.778 +# Check for failed kmalloc() in ext3_htree_store_dirent() +# +# This patch checks for a failed kmalloc() in ext3_htree_store_dirent(), +# and passes the error up to its caller, ext3_htree_fill_tree(). +# -------------------------------------------- +# +diff -Nru a/fs/ext3/dir.c b/fs/ext3/dir.c +--- a/fs/ext3/dir.c Thu Nov 7 10:57:34 2002 ++++ b/fs/ext3/dir.c Thu Nov 7 10:57:34 2002 +@@ -308,7 +308,7 @@ + /* + * Given a directory entry, enter it into the fname rb tree. + */ +-void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, + __u32 minor_hash, + struct ext3_dir_entry_2 *dirent) + { +@@ -323,6 +323,8 @@ + /* Create and allocate the fname structure */ + len = sizeof(struct fname) + dirent->name_len + 1; + new_fn = kmalloc(len, GFP_KERNEL); ++ if (!new_fn) ++ return -ENOMEM; + memset(new_fn, 0, len); + new_fn->hash = hash; + new_fn->minor_hash = minor_hash; +@@ -344,7 +346,7 @@ + (new_fn->minor_hash == fname->minor_hash)) { + new_fn->next = fname->next; + fname->next = new_fn; +- return; ++ return 0; + } + + if (new_fn->hash < fname->hash) +@@ -359,6 +361,7 @@ + + rb_link_node(&new_fn->rb_hash, parent, p); + rb_insert_color(&new_fn->rb_hash, &info->root); ++ return 0; + } + + +diff -Nru a/fs/ext3/namei.c b/fs/ext3/namei.c +--- a/fs/ext3/namei.c Thu Nov 7 10:57:34 2002 ++++ b/fs/ext3/namei.c Thu Nov 7 10:57:34 2002 +@@ -549,9 +549,11 @@ + /* Add '.' and '..' from the htree header */ + if (!start_hash && !start_minor_hash) { + de = (struct ext3_dir_entry_2 *) frames[0].bh->b_data; +- ext3_htree_store_dirent(dir_file, 0, 0, de); ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; + de = ext3_next_entry(de); +- ext3_htree_store_dirent(dir_file, 0, 0, de); ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; + count += 2; + } + +@@ -570,8 +572,9 @@ + ((hinfo.hash == start_hash) && + (hinfo.minor_hash < start_minor_hash))) + continue; +- ext3_htree_store_dirent(dir_file, hinfo.hash, +- hinfo.minor_hash, de); ++ if ((err = ext3_htree_store_dirent(dir_file, ++ hinfo.hash, hinfo.minor_hash, de)) != 0) ++ goto errout; + count++; + } + brelse (bh); +diff -Nru a/include/linux/ext3_fs.h b/include/linux/ext3_fs.h +--- a/include/linux/ext3_fs.h Thu Nov 7 10:57:34 2002 ++++ b/include/linux/ext3_fs.h Thu Nov 7 10:57:34 2002 +@@ -682,7 +682,7 @@ + extern int ext3_check_dir_entry(const char *, struct inode *, + struct ext3_dir_entry_2 *, + struct buffer_head *, unsigned long); +-extern void ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++extern int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, + __u32 minor_hash, + struct ext3_dir_entry_2 *dirent); + extern void ext3_htree_free_dir_info(struct dir_private_info *p); diff --git a/lustre/kernel_patches/patches/ext-2.4-patch-4.patch b/lustre/kernel_patches/patches/ext-2.4-patch-4.patch new file mode 100644 index 0000000..67f5afa --- /dev/null +++ b/lustre/kernel_patches/patches/ext-2.4-patch-4.patch @@ -0,0 +1,48 @@ +# This is a BitKeeper generated patch for the following project: +# Project Name: Linux kernel tree +# +# namei.c | 21 ++++++++++++++++++++- +# 1 files changed, 20 insertions(+), 1 deletion(-) +# +# The following is the BitKeeper ChangeSet Log +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.779 +# Fix ext3 htree rename bug. +# +# This fixes an ext3 htree bug pointed out by Christopher Li; if +# adding the new name to the directory causes a split, this can cause +# the directory entry containing the old name to move to another +# block, and then the removal of the old name will fail. +# -------------------------------------------- +# +diff -Nru a/fs/ext3/namei.c b/fs/ext3/namei.c +--- a/fs/ext3/namei.c Thu Nov 7 10:57:49 2002 ++++ b/fs/ext3/namei.c Thu Nov 7 10:57:49 2002 +@@ -2173,7 +2173,26 @@ + /* + * ok, that's it + */ +- ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ retval = ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ if (retval == -ENOENT) { ++ /* ++ * old_de could have moved out from under us. ++ */ ++ struct buffer_head *old_bh2; ++ struct ext3_dir_entry_2 *old_de2; ++ ++ old_bh2 = ext3_find_entry(old_dentry, &old_de2); ++ if (old_bh2) { ++ retval = ext3_delete_entry(handle, old_dir, ++ old_de2, old_bh2); ++ brelse(old_bh2); ++ } ++ } ++ if (retval) { ++ ext3_warning(old_dir->i_sb, "ext3_rename", ++ "Deleting old file (%lu), %d, error=%d", ++ old_dir->i_ino, old_dir->i_nlink, retval); ++ } + + if (new_inode) { + new_inode->i_nlink--; diff --git a/lustre/kernel_patches/patches/ext3-2.4-ino_t.patch b/lustre/kernel_patches/patches/ext3-2.4-ino_t.patch new file mode 100644 index 0000000..ce1bd88 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-2.4-ino_t.patch @@ -0,0 +1,138 @@ +--- linux/fs/ext3/ialloc.c.orig Sat Oct 19 11:42:23 2002 ++++ linux/fs/ext3/ialloc.c Sat Jan 4 12:14:18 2003 +@@ -64,8 +64,8 @@ static int read_inode_bitmap (struct sup + if (!bh) { + ext3_error (sb, "read_inode_bitmap", + "Cannot read inode bitmap - " +- "block_group = %lu, inode_bitmap = %lu", +- block_group, (unsigned long) gdp->bg_inode_bitmap); ++ "block_group = %lu, inode_bitmap = %u", ++ block_group, gdp->bg_inode_bitmap); + retval = -EIO; + } + /* +@@ -531,19 +532,19 @@ out: + } + + /* Verify that we are loading a valid orphan from disk */ +-struct inode *ext3_orphan_get (struct super_block * sb, ino_t ino) ++struct inode *ext3_orphan_get(struct super_block *sb, unsigned long ino) + { +- ino_t max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count); ++ unsigned long max_ino = le32_to_cpu(EXT3_SB(sb)->s_es->s_inodes_count); + unsigned long block_group; + int bit; + int bitmap_nr; + struct buffer_head *bh; + struct inode *inode = NULL; +- ++ + /* Error cases - e2fsck has already cleaned up for us */ + if (ino > max_ino) { + ext3_warning(sb, __FUNCTION__, +- "bad orphan ino %ld! e2fsck was run?\n", ino); ++ "bad orphan ino %lu! e2fsck was run?\n", ino); + return NULL; + } + +@@ -552,7 +553,7 @@ struct inode *ext3_orphan_get (struct su + if ((bitmap_nr = load_inode_bitmap(sb, block_group)) < 0 || + !(bh = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr])) { + ext3_warning(sb, __FUNCTION__, +- "inode bitmap error for orphan %ld\n", ino); ++ "inode bitmap error for orphan %lu\n", ino); + return NULL; + } + +@@ -563,7 +564,7 @@ struct inode *ext3_orphan_get (struct su + if (!ext3_test_bit(bit, bh->b_data) || !(inode = iget(sb, ino)) || + is_bad_inode(inode) || NEXT_ORPHAN(inode) > max_ino) { + ext3_warning(sb, __FUNCTION__, +- "bad orphan inode %ld! e2fsck was run?\n", ino); ++ "bad orphan inode %lu! e2fsck was run?\n", ino); + printk(KERN_NOTICE "ext3_test_bit(bit=%d, block=%ld) = %d\n", + bit, bh->b_blocknr, ext3_test_bit(bit, bh->b_data)); + printk(KERN_NOTICE "inode=%p\n", inode); +@@ -570,9 +571,9 @@ struct inode *ext3_orphan_get (struct su + if (inode) { + printk(KERN_NOTICE "is_bad_inode(inode)=%d\n", + is_bad_inode(inode)); +- printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%d\n", ++ printk(KERN_NOTICE "NEXT_ORPHAN(inode)=%u\n", + NEXT_ORPHAN(inode)); +- printk(KERN_NOTICE "max_ino=%ld\n", max_ino); ++ printk(KERN_NOTICE "max_ino=%lu\n", max_ino); + } + /* Avoid freeing blocks if we got a bad deleted inode */ + if (inode && inode->i_nlink == 0) +--- linux/fs/ext3/namei.c.orig Sat Oct 19 11:42:45 2002 ++++ linux/fs/ext3/namei.c Sat Jan 4 12:13:27 2003 +@@ -716,10 +716,10 @@ int ext3_orphan_del(handle_t *handle, st + { + struct list_head *prev; + struct ext3_sb_info *sbi; +- ino_t ino_next; ++ unsigned long ino_next; + struct ext3_iloc iloc; + int err = 0; +- ++ + lock_super(inode->i_sb); + if (list_empty(&inode->u.ext3_i.i_orphan)) { + unlock_super(inode->i_sb); +@@ -730,7 +730,7 @@ int ext3_orphan_del(handle_t *handle, st + prev = inode->u.ext3_i.i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + +- jbd_debug(4, "remove inode %ld from orphan list\n", inode->i_ino); ++ jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); + + list_del(&inode->u.ext3_i.i_orphan); + INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); +@@ -741,13 +741,13 @@ int ext3_orphan_del(handle_t *handle, st + * list in memory. */ + if (!handle) + goto out; +- ++ + err = ext3_reserve_inode_write(handle, inode, &iloc); + if (err) + goto out_err; + + if (prev == &sbi->s_orphan) { +- jbd_debug(4, "superblock will point to %ld\n", ino_next); ++ jbd_debug(4, "superblock will point to %lu\n", ino_next); + BUFFER_TRACE(sbi->s_sbh, "get_write_access"); + err = ext3_journal_get_write_access(handle, sbi->s_sbh); + if (err) +@@ -758,8 +758,8 @@ int ext3_orphan_del(handle_t *handle, st + struct ext3_iloc iloc2; + struct inode *i_prev = + list_entry(prev, struct inode, u.ext3_i.i_orphan); +- +- jbd_debug(4, "orphan inode %ld will point to %ld\n", ++ ++ jbd_debug(4, "orphan inode %lu will point to %lu\n", + i_prev->i_ino, ino_next); + err = ext3_reserve_inode_write(handle, i_prev, &iloc2); + if (err) +@@ -774,7 +774,7 @@ int ext3_orphan_del(handle_t *handle, st + if (err) + goto out_brelse; + +-out_err: ++out_err: + ext3_std_error(inode->i_sb, err); + out: + unlock_super(inode->i_sb); +--- linux/include/linux/ext3_fs.h.orig Thu Jan 2 16:10:24 2003 ++++ linux/include/linux/ext3_fs.h Sat Jan 4 12:25:41 2003 +@@ -622,7 +622,7 @@ extern int ext3_sync_file (struct file * + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); +-extern struct inode * ext3_orphan_get (struct super_block *, ino_t); ++extern struct inode * ext3_orphan_get (struct super_block *, unsigned long); + extern unsigned long ext3_count_free_inodes (struct super_block *); + extern void ext3_check_inodes_bitmap (struct super_block *); + extern unsigned long ext3_count_free (struct buffer_head *, unsigned); diff --git a/lustre/kernel_patches/patches/ext3-2.4.18-fixes.patch b/lustre/kernel_patches/patches/ext3-2.4.18-fixes.patch new file mode 100644 index 0000000..56e841e --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-2.4.18-fixes.patch @@ -0,0 +1,353 @@ +diff -ru lum-2.4.18-um30/fs/ext3/balloc.c uml-2.4.18-12.5/fs/ext3/balloc.c +--- lum-2.4.18-um30/fs/ext3/balloc.c Mon Feb 25 12:38:08 2002 ++++ uml-2.4.18-12.5/fs/ext3/balloc.c Thu Sep 19 13:40:11 2002 +@@ -276,7 +276,8 @@ + } + lock_super (sb); + es = sb->u.ext3_sb.s_es; +- if (block < le32_to_cpu(es->s_first_data_block) || ++ if (block < le32_to_cpu(es->s_first_data_block) || ++ block + count < block || + (block + count) > le32_to_cpu(es->s_blocks_count)) { + ext3_error (sb, "ext3_free_blocks", + "Freeing blocks not in datazone - " +@@ -309,17 +310,6 @@ + if (!gdp) + goto error_return; + +- if (in_range (le32_to_cpu(gdp->bg_block_bitmap), block, count) || +- in_range (le32_to_cpu(gdp->bg_inode_bitmap), block, count) || +- in_range (block, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group) || +- in_range (block + count - 1, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group)) +- ext3_error (sb, "ext3_free_blocks", +- "Freeing blocks in system zones - " +- "Block = %lu, count = %lu", +- block, count); +- + /* + * We are about to start releasing blocks in the bitmap, + * so we need undo access. +@@ -345,14 +335,24 @@ + if (err) + goto error_return; + +- for (i = 0; i < count; i++) { ++ for (i = 0; i < count; i++, block++) { ++ if (block == le32_to_cpu(gdp->bg_block_bitmap) || ++ block == le32_to_cpu(gdp->bg_inode_bitmap) || ++ in_range(block, le32_to_cpu(gdp->bg_inode_table), ++ sb->u.ext2_sb.s_itb_per_group)) { ++ ext3_error(sb, __FUNCTION__, ++ "Freeing block in system zone - block = %lu", ++ block); ++ continue; ++ } ++ + /* + * An HJ special. This is expensive... + */ + #ifdef CONFIG_JBD_DEBUG + { + struct buffer_head *debug_bh; +- debug_bh = sb_get_hash_table(sb, block + i); ++ debug_bh = sb_get_hash_table(sb, block); + if (debug_bh) { + BUFFER_TRACE(debug_bh, "Deleted!"); + if (!bh2jh(bitmap_bh)->b_committed_data) +@@ -365,9 +365,8 @@ + #endif + BUFFER_TRACE(bitmap_bh, "clear bit"); + if (!ext3_clear_bit (bit + i, bitmap_bh->b_data)) { +- ext3_error (sb, __FUNCTION__, +- "bit already cleared for block %lu", +- block + i); ++ ext3_error(sb, __FUNCTION__, ++ "bit already cleared for block %lu", block); + BUFFER_TRACE(bitmap_bh, "bit already cleared"); + } else { + dquot_freed_blocks++; +@@ -415,7 +417,6 @@ + if (!err) err = ret; + + if (overflow && !err) { +- block += count; + count = overflow; + goto do_more; + } +@@ -542,6 +543,7 @@ + int i, j, k, tmp, alloctmp; + int bitmap_nr; + int fatal = 0, err; ++ int performed_allocation = 0; + struct super_block * sb; + struct ext3_group_desc * gdp; + struct ext3_super_block * es; +@@ -575,6 +577,7 @@ + + ext3_debug ("goal=%lu.\n", goal); + ++repeat: + /* + * First, test whether the goal block is free. + */ +@@ -644,8 +647,7 @@ + } + + /* No space left on the device */ +- unlock_super (sb); +- return 0; ++ goto out; + + search_back: + /* +@@ -684,16 +686,28 @@ + if (tmp == le32_to_cpu(gdp->bg_block_bitmap) || + tmp == le32_to_cpu(gdp->bg_inode_bitmap) || + in_range (tmp, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group)) +- ext3_error (sb, "ext3_new_block", +- "Allocating block in system zone - " +- "block = %u", tmp); ++ EXT3_SB(sb)->s_itb_per_group)) { ++ ext3_error(sb, __FUNCTION__, ++ "Allocating block in system zone - block = %u", tmp); ++ ++ /* Note: This will potentially use up one of the handle's ++ * buffer credits. Normally we have way too many credits, ++ * so that is OK. In _very_ rare cases it might not be OK. ++ * We will trigger an assertion if we run out of credits, ++ * and we will have to do a full fsck of the filesystem - ++ * better than randomly corrupting filesystem metadata. ++ */ ++ ext3_set_bit(j, bh->b_data); ++ goto repeat; ++ } ++ + + /* The superblock lock should guard against anybody else beating + * us to this point! */ + J_ASSERT_BH(bh, !ext3_test_bit(j, bh->b_data)); + BUFFER_TRACE(bh, "setting bitmap bit"); + ext3_set_bit(j, bh->b_data); ++ performed_allocation = 1; + + #ifdef CONFIG_JBD_DEBUG + { +@@ -815,6 +829,11 @@ + ext3_std_error(sb, fatal); + } + unlock_super (sb); ++ /* ++ * Undo the block allocation ++ */ ++ if (!performed_allocation) ++ DQUOT_FREE_BLOCK(inode, 1); + return 0; + + } +diff -ru lum-2.4.18-um30/fs/ext3/file.c uml-2.4.18-12.5/fs/ext3/file.c +--- lum-2.4.18-um30/fs/ext3/file.c Thu Nov 15 14:37:55 2001 ++++ uml-2.4.18-12.5/fs/ext3/file.c Thu Sep 19 13:40:11 2002 +@@ -61,19 +61,52 @@ + static ssize_t + ext3_file_write(struct file *file, const char *buf, size_t count, loff_t *ppos) + { ++ int ret, err; + struct inode *inode = file->f_dentry->d_inode; + +- /* +- * Nasty: if the file is subject to synchronous writes then we need +- * to force generic_osync_inode() to call ext3_write_inode(). +- * We do that by marking the inode dirty. This adds much more +- * computational expense than we need, but we're going to sync +- * anyway. +- */ +- if (IS_SYNC(inode) || (file->f_flags & O_SYNC)) +- mark_inode_dirty(inode); ++ ret = generic_file_write(file, buf, count, ppos); + +- return generic_file_write(file, buf, count, ppos); ++ /* Skip file flushing code if there was an error, or if nothing ++ was written. */ ++ if (ret <= 0) ++ return ret; ++ ++ /* If the inode is IS_SYNC, or is O_SYNC and we are doing ++ data-journaling, then we need to make sure that we force the ++ transaction to disk to keep all metadata uptodate ++ synchronously. */ ++ ++ if (file->f_flags & O_SYNC) { ++ /* If we are non-data-journaled, then the dirty data has ++ already been flushed to backing store by ++ generic_osync_inode, and the inode has been flushed ++ too if there have been any modifications other than ++ mere timestamp updates. ++ ++ Open question --- do we care about flushing ++ timestamps too if the inode is IS_SYNC? */ ++ if (!ext3_should_journal_data(inode)) ++ return ret; ++ ++ goto force_commit; ++ } ++ ++ /* So we know that there has been no forced data flush. If the ++ inode is marked IS_SYNC, we need to force one ourselves. */ ++ if (!IS_SYNC(inode)) ++ return ret; ++ ++ /* Open question #2 --- should we force data to disk here too? ++ If we don't, the only impact is that data=writeback ++ filesystems won't flush data to disk automatically on ++ IS_SYNC, only metadata (but historically, that is what ext2 ++ has done.) */ ++ ++force_commit: ++ err = ext3_force_commit(inode->i_sb); ++ if (err) ++ return err; ++ return ret; + } + + struct file_operations ext3_file_operations = { +diff -ru lum-2.4.18-um30/fs/ext3/fsync.c uml-2.4.18-12.5/fs/ext3/fsync.c +--- lum-2.4.18-um30/fs/ext3/fsync.c Tue Nov 20 22:34:13 2001 ++++ uml-2.4.18-12.5/fs/ext3/fsync.c Thu Sep 19 13:40:11 2002 +@@ -62,7 +62,12 @@ + * we'll end up waiting on them in commit. + */ + ret = fsync_inode_buffers(inode); +- ret |= fsync_inode_data_buffers(inode); ++ ++ /* In writeback mode, we need to force out data buffers too. In ++ * the other modes, ext3_force_commit takes care of forcing out ++ * just the right data blocks. */ ++ if (test_opt(inode->i_sb, DATA_FLAGS) == EXT3_MOUNT_WRITEBACK_DATA) ++ ret |= fsync_inode_data_buffers(inode); + + ext3_force_commit(inode->i_sb); + +diff -ru lum-2.4.18-um30/fs/ext3/ialloc.c uml-2.4.18-12.5/fs/ext3/ialloc.c +--- lum-2.4.18-um30/fs/ext3/ialloc.c Mon Feb 25 12:38:08 2002 ++++ uml-2.4.18-12.5/fs/ext3/ialloc.c Thu Sep 19 13:40:11 2002 +@@ -392,7 +392,7 @@ + + err = -ENOSPC; + if (!gdp) +- goto fail; ++ goto out; + + err = -EIO; + bitmap_nr = load_inode_bitmap (sb, i); +@@ -523,9 +523,10 @@ + return inode; + + fail: ++ ext3_std_error(sb, err); ++out: + unlock_super(sb); + iput(inode); +- ext3_std_error(sb, err); + return ERR_PTR(err); + } + +diff -ru lum-2.4.18-um30/fs/ext3/inode.c uml-2.4.18-12.5/fs/ext3/inode.c +--- lum-2.4.18-um30/fs/ext3/inode.c Mon Feb 25 12:38:08 2002 ++++ uml-2.4.18-12.5/fs/ext3/inode.c Thu Sep 19 13:40:11 2002 +@@ -412,6 +412,7 @@ + return NULL; + + changed: ++ brelse(bh); + *err = -EAGAIN; + goto no_block; + failure: +@@ -581,8 +582,6 @@ + + parent = nr; + } +- if (IS_SYNC(inode)) +- handle->h_sync = 1; + } + if (n == num) + return 0; +@@ -1015,8 +1018,8 @@ + unsigned from, unsigned to) + { + struct inode *inode = page->mapping->host; +- handle_t *handle = ext3_journal_current_handle(); + int ret, needed_blocks = ext3_writepage_trans_blocks(inode); ++ handle_t *handle; + + lock_kernel(); + handle = ext3_journal_start(inode, needed_blocks); +diff -ru lum-2.4.18-um30/fs/ext3/namei.c uml-2.4.18-12.5/fs/ext3/namei.c +--- lum-2.4.18-um30/fs/ext3/namei.c Fri Nov 9 15:25:04 2001 ++++ uml-2.4.18-12.5/fs/ext3/namei.c Thu Sep 19 13:40:11 2002 +@@ -354,8 +355,8 @@ + */ + dir->i_mtime = dir->i_ctime = CURRENT_TIME; + dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); + dir->i_version = ++event; ++ ext3_mark_inode_dirty(handle, dir); + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); + ext3_journal_dirty_metadata(handle, bh); + brelse(bh); +@@ -464,8 +465,8 @@ + 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_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -489,8 +490,8 @@ + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -933,8 +934,8 @@ + inode->i_size = l-1; + } + inode->u.ext3_i.i_disksize = inode->i_size; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +@@ -970,8 +971,8 @@ + ext3_inc_count(handle, inode); + atomic_inc(&inode->i_count); + +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + ext3_journal_stop(handle, dir); + return err; + } +diff -ru lum-2.4.18-um30/fs/ext3/super.c uml-2.4.18-12.5/fs/ext3/super.c +--- lum-2.4.18-um30/fs/ext3/super.c Fri Jul 12 17:59:37 2002 ++++ uml-2.4.18-12.5/fs/ext3/super.c Thu Sep 19 13:40:11 2002 +@@ -1589,8 +1589,10 @@ + journal_t *journal = EXT3_SB(sb)->s_journal; + + /* Now we set up the journal barrier. */ ++ unlock_super(sb); + journal_lock_updates(journal); + journal_flush(journal); ++ lock_super(sb); + + /* Journal blocked and flushed, clear needs_recovery flag. */ + EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); diff --git a/lustre/kernel_patches/patches/ext3-2.4.18-ino_sb_macro.patch b/lustre/kernel_patches/patches/ext3-2.4.18-ino_sb_macro.patch new file mode 100644 index 0000000..cc47588 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-2.4.18-ino_sb_macro.patch @@ -0,0 +1,1554 @@ +--- ./fs/ext3/balloc.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/balloc.c Tue May 7 15:35:59 2002 +@@ -46,18 +46,18 @@ struct ext3_group_desc * ext3_get_group_ + unsigned long desc; + struct ext3_group_desc * gdp; + +- if (block_group >= sb->u.ext3_sb.s_groups_count) { ++ if (block_group >= EXT3_SB(sb)->s_groups_count) { + ext3_error (sb, "ext3_get_group_desc", + "block_group >= groups_count - " + "block_group = %d, groups_count = %lu", +- block_group, sb->u.ext3_sb.s_groups_count); ++ block_group, EXT3_SB(sb)->s_groups_count); + + return NULL; + } + + group_desc = block_group / EXT3_DESC_PER_BLOCK(sb); + desc = block_group % EXT3_DESC_PER_BLOCK(sb); +- if (!sb->u.ext3_sb.s_group_desc[group_desc]) { ++ if (!EXT3_SB(sb)->s_group_desc[group_desc]) { + ext3_error (sb, "ext3_get_group_desc", + "Group descriptor not loaded - " + "block_group = %d, group_desc = %lu, desc = %lu", +@@ -66,9 +66,9 @@ struct ext3_group_desc * ext3_get_group_ + } + + gdp = (struct ext3_group_desc *) +- sb->u.ext3_sb.s_group_desc[group_desc]->b_data; ++ EXT3_SB(sb)->s_group_desc[group_desc]->b_data; + if (bh) +- *bh = sb->u.ext3_sb.s_group_desc[group_desc]; ++ *bh = EXT3_SB(sb)->s_group_desc[group_desc]; + return gdp + desc; + } + +@@ -104,8 +104,8 @@ static int read_block_bitmap (struct sup + * this group. The IO will be retried next time. + */ + error_out: +- sb->u.ext3_sb.s_block_bitmap_number[bitmap_nr] = block_group; +- sb->u.ext3_sb.s_block_bitmap[bitmap_nr] = bh; ++ EXT3_SB(sb)->s_block_bitmap_number[bitmap_nr] = block_group; ++ EXT3_SB(sb)->s_block_bitmap[bitmap_nr] = bh; + return retval; + } + +@@ -128,16 +128,17 @@ static int __load_block_bitmap (struct s + int i, j, retval = 0; + unsigned long block_bitmap_number; + struct buffer_head * block_bitmap; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); + +- if (block_group >= sb->u.ext3_sb.s_groups_count) ++ if (block_group >= sbi->s_groups_count) + ext3_panic (sb, "load_block_bitmap", + "block_group >= groups_count - " + "block_group = %d, groups_count = %lu", +- block_group, sb->u.ext3_sb.s_groups_count); ++ block_group, EXT3_SB(sb)->s_groups_count); + +- if (sb->u.ext3_sb.s_groups_count <= EXT3_MAX_GROUP_LOADED) { +- if (sb->u.ext3_sb.s_block_bitmap[block_group]) { +- if (sb->u.ext3_sb.s_block_bitmap_number[block_group] == ++ if (sbi->s_groups_count <= EXT3_MAX_GROUP_LOADED) { ++ if (sbi->s_block_bitmap[block_group]) { ++ if (sbi->s_block_bitmap_number[block_group] == + block_group) + return block_group; + ext3_error (sb, "__load_block_bitmap", +@@ -149,21 +150,20 @@ static int __load_block_bitmap (struct s + return block_group; + } + +- for (i = 0; i < sb->u.ext3_sb.s_loaded_block_bitmaps && +- sb->u.ext3_sb.s_block_bitmap_number[i] != block_group; i++) ++ for (i = 0; i < sbi->s_loaded_block_bitmaps && ++ sbi->s_block_bitmap_number[i] != block_group; i++) + ; +- if (i < sb->u.ext3_sb.s_loaded_block_bitmaps && +- sb->u.ext3_sb.s_block_bitmap_number[i] == block_group) { +- block_bitmap_number = sb->u.ext3_sb.s_block_bitmap_number[i]; +- block_bitmap = sb->u.ext3_sb.s_block_bitmap[i]; ++ if (i < sbi->s_loaded_block_bitmaps && ++ sbi->s_block_bitmap_number[i] == block_group) { ++ block_bitmap_number = sbi->s_block_bitmap_number[i]; ++ block_bitmap = sbi->s_block_bitmap[i]; + for (j = i; j > 0; j--) { +- sb->u.ext3_sb.s_block_bitmap_number[j] = +- sb->u.ext3_sb.s_block_bitmap_number[j - 1]; +- sb->u.ext3_sb.s_block_bitmap[j] = +- sb->u.ext3_sb.s_block_bitmap[j - 1]; ++ sbi->s_block_bitmap_number[j] = ++ sbi->s_block_bitmap_number[j - 1]; ++ sbi->s_block_bitmap[j] = sbi->s_block_bitmap[j - 1]; + } +- sb->u.ext3_sb.s_block_bitmap_number[0] = block_bitmap_number; +- sb->u.ext3_sb.s_block_bitmap[0] = block_bitmap; ++ sbi->s_block_bitmap_number[0] = block_bitmap_number; ++ sbi->s_block_bitmap[0] = block_bitmap; + + /* + * There's still one special case here --- if block_bitmap == 0 +@@ -173,17 +173,14 @@ static int __load_block_bitmap (struct s + if (!block_bitmap) + retval = read_block_bitmap (sb, block_group, 0); + } else { +- if (sb->u.ext3_sb.s_loaded_block_bitmapsu.ext3_sb.s_loaded_block_bitmaps++; ++ if (sbi->s_loaded_block_bitmapss_loaded_block_bitmaps++; + else +- brelse (sb->u.ext3_sb.s_block_bitmap +- [EXT3_MAX_GROUP_LOADED - 1]); +- for (j = sb->u.ext3_sb.s_loaded_block_bitmaps - 1; +- j > 0; j--) { +- sb->u.ext3_sb.s_block_bitmap_number[j] = +- sb->u.ext3_sb.s_block_bitmap_number[j - 1]; +- sb->u.ext3_sb.s_block_bitmap[j] = +- sb->u.ext3_sb.s_block_bitmap[j - 1]; ++ brelse(sbi->s_block_bitmap[EXT3_MAX_GROUP_LOADED - 1]); ++ for (j = sbi->s_loaded_block_bitmaps - 1; j > 0; j--) { ++ sbi->s_block_bitmap_number[j] = ++ sbi->s_block_bitmap_number[j - 1]; ++ sbi->s_block_bitmap[j] = sbi->s_block_bitmap[j - 1]; + } + retval = read_block_bitmap (sb, block_group, 0); + } +@@ -206,24 +203,25 @@ static int __load_block_bitmap (struct s + static inline int load_block_bitmap (struct super_block * sb, + unsigned int block_group) + { ++ struct ext3_sb_info *sbi = EXT3_SB(sb); + int slot; +- ++ + /* + * Do the lookup for the slot. First of all, check if we're asking + * for the same slot as last time, and did we succeed that last time? + */ +- if (sb->u.ext3_sb.s_loaded_block_bitmaps > 0 && +- sb->u.ext3_sb.s_block_bitmap_number[0] == block_group && +- sb->u.ext3_sb.s_block_bitmap[0]) { ++ if (sbi->s_loaded_block_bitmaps > 0 && ++ sbi->s_block_bitmap_number[0] == block_group && ++ sbi->s_block_bitmap[0]) { + return 0; + } + /* + * Or can we do a fast lookup based on a loaded group on a filesystem + * small enough to be mapped directly into the superblock? + */ +- else if (sb->u.ext3_sb.s_groups_count <= EXT3_MAX_GROUP_LOADED && +- sb->u.ext3_sb.s_block_bitmap_number[block_group]==block_group +- && sb->u.ext3_sb.s_block_bitmap[block_group]) { ++ else if (sbi->s_groups_count <= EXT3_MAX_GROUP_LOADED && ++ sbi->s_block_bitmap_number[block_group] == block_group ++ && sbi->s_block_bitmap[block_group]) { + slot = block_group; + } + /* +@@ -243,7 +241,7 @@ static inline int load_block_bitmap (str + * If it's a valid slot, we may still have cached a previous IO error, + * in which case the bh in the superblock cache will be zero. + */ +- if (!sb->u.ext3_sb.s_block_bitmap[slot]) ++ if (!sbi->s_block_bitmap[slot]) + return -EIO; + + /* +@@ -275,7 +273,7 @@ void ext3_free_blocks (handle_t *handle, + return; + } + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; + if (block < le32_to_cpu(es->s_first_data_block) || + block + count < block || + (block + count) > le32_to_cpu(es->s_blocks_count)) { +@@ -304,7 +302,7 @@ do_more: + if (bitmap_nr < 0) + goto error_return; + +- bitmap_bh = sb->u.ext3_sb.s_block_bitmap[bitmap_nr]; ++ bitmap_bh = EXT3_SB(sb)->s_block_bitmap[bitmap_nr]; + gdp = ext3_get_group_desc (sb, block_group, &gd_bh); + if (!gdp) + goto error_return; +@@ -330,8 +328,8 @@ do_more: + if (err) + goto error_return; + +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "get_write_access"); +- err = ext3_journal_get_write_access(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); + if (err) + goto error_return; + +@@ -341,7 +339,7 @@ + if (block == le32_to_cpu(gdp->bg_block_bitmap) || + block == le32_to_cpu(gdp->bg_inode_bitmap) || + in_range(block, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext2_sb.s_itb_per_group)) { ++ EXT3_SB(sb)->s_itb_per_group)) { + ext3_error(sb, __FUNCTION__, + "Freeing block in system zone - block = %lu", + block); +@@ -410,8 +407,8 @@ do_more: + if (!err) err = ret; + + /* And the superblock */ +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "dirtied superblock"); +- ret = ext3_journal_dirty_metadata(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "dirtied superblock"); ++ ret = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); + if (!err) err = ret; + + if (overflow && !err) { +@@ -564,12 +560,12 @@ int ext3_new_block (handle_t *handle, st + } + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; + if (le32_to_cpu(es->s_free_blocks_count) <= + le32_to_cpu(es->s_r_blocks_count) && +- ((sb->u.ext3_sb.s_resuid != current->fsuid) && +- (sb->u.ext3_sb.s_resgid == 0 || +- !in_group_p (sb->u.ext3_sb.s_resgid)) && ++ ((EXT3_SB(sb)->s_resuid != current->fsuid) && ++ (EXT3_SB(sb)->s_resgid == 0 || ++ !in_group_p (EXT3_SB(sb)->s_resgid)) && + !capable(CAP_SYS_RESOURCE))) + goto out; + +@@ -598,7 +595,7 @@ int ext3_new_block (handle_t *handle, st + if (bitmap_nr < 0) + goto io_error; + +- bh = sb->u.ext3_sb.s_block_bitmap[bitmap_nr]; ++ bh = EXT3_SB(sb)->s_block_bitmap[bitmap_nr]; + + ext3_debug ("goal is at %d:%d.\n", i, j); + +@@ -621,9 +618,9 @@ int ext3_new_block (handle_t *handle, st + * Now search the rest of the groups. We assume that + * i and gdp correctly point to the last group visited. + */ +- for (k = 0; k < sb->u.ext3_sb.s_groups_count; k++) { ++ for (k = 0; k < EXT3_SB(sb)->s_groups_count; k++) { + i++; +- if (i >= sb->u.ext3_sb.s_groups_count) ++ if (i >= EXT3_SB(sb)->s_groups_count) + i = 0; + gdp = ext3_get_group_desc (sb, i, &bh2); + if (!gdp) { +@@ -635,7 +632,7 @@ int ext3_new_block (handle_t *handle, st + if (bitmap_nr < 0) + goto io_error; + +- bh = sb->u.ext3_sb.s_block_bitmap[bitmap_nr]; ++ bh = EXT3_SB(sb)->s_block_bitmap[bitmap_nr]; + j = find_next_usable_block(-1, bh, + EXT3_BLOCKS_PER_GROUP(sb)); + if (j >= 0) +@@ -674,8 +671,8 @@ got_block: + fatal = ext3_journal_get_write_access(handle, bh2); + if (fatal) goto out; + +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "get_write_access"); +- fatal = ext3_journal_get_write_access(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "get_write_access"); ++ fatal = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); + if (fatal) goto out; + + tmp = j + i * EXT3_BLOCKS_PER_GROUP(sb) +@@ -796,7 +804,7 @@ got_block: + if (!fatal) fatal = err; + + BUFFER_TRACE(bh, "journal_dirty_metadata for superblock"); +- err = ext3_journal_dirty_metadata(handle, sb->u.ext3_sb.s_sbh); ++ err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); + if (!fatal) fatal = err; + + sb->s_dirt = 1; +@@ -829,11 +837,11 @@ unsigned long ext3_count_free_blocks (st + int i; + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; + desc_count = 0; + bitmap_count = 0; + gdp = NULL; +- for (i = 0; i < sb->u.ext3_sb.s_groups_count; i++) { ++ for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { + gdp = ext3_get_group_desc (sb, i, NULL); + if (!gdp) + continue; +@@ -842,7 +850,7 @@ unsigned long ext3_count_free_blocks (st + if (bitmap_nr < 0) + continue; + +- x = ext3_count_free (sb->u.ext3_sb.s_block_bitmap[bitmap_nr], ++ x = ext3_count_free (EXT3_SB(sb)->s_block_bitmap[bitmap_nr], + sb->s_blocksize); + printk ("group %d: stored = %d, counted = %lu\n", + i, le16_to_cpu(gdp->bg_free_blocks_count), x); +@@ -853,7 +861,7 @@ unsigned long ext3_count_free_blocks (st + unlock_super (sb); + return bitmap_count; + #else +- return le32_to_cpu(sb->u.ext3_sb.s_es->s_free_blocks_count); ++ return le32_to_cpu(EXT3_SB(sb)->s_es->s_free_blocks_count); + #endif + } + +@@ -862,7 +870,7 @@ static inline int block_in_use (unsigned + unsigned char * map) + { + return ext3_test_bit ((block - +- le32_to_cpu(sb->u.ext3_sb.s_es->s_first_data_block)) % ++ le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block)) % + EXT3_BLOCKS_PER_GROUP(sb), map); + } + +@@ -930,11 +938,11 @@ void ext3_check_blocks_bitmap (struct su + struct ext3_group_desc * gdp; + int i; + +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; + desc_count = 0; + bitmap_count = 0; + gdp = NULL; +- for (i = 0; i < sb->u.ext3_sb.s_groups_count; i++) { ++ for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { + gdp = ext3_get_group_desc (sb, i, NULL); + if (!gdp) + continue; +@@ -968,7 +976,7 @@ void ext3_check_blocks_bitmap (struct su + "Inode bitmap for group %d is marked free", + i); + +- for (j = 0; j < sb->u.ext3_sb.s_itb_per_group; j++) ++ for (j = 0; j < EXT3_SB(sb)->s_itb_per_group; j++) + if (!block_in_use (le32_to_cpu(gdp->bg_inode_table) + j, + sb, bh->b_data)) + ext3_error (sb, "ext3_check_blocks_bitmap", +--- ./fs/ext3/dir.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/dir.c Tue May 7 14:54:13 2002 +@@ -52,7 +52,7 @@ int ext3_check_dir_entry (const char * f + else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize) + error_msg = "directory entry across blocks"; + else if (le32_to_cpu(de->inode) > +- le32_to_cpu(dir->i_sb->u.ext3_sb.s_es->s_inodes_count)) ++ le32_to_cpu(EXT3_SB(dir->i_sb)->s_es->s_inodes_count)) + error_msg = "inode out of bounds"; + + if (error_msg != NULL) +--- ./fs/ext3/ialloc.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/ialloc.c Tue May 7 15:39:26 2002 +@@ -73,8 +73,8 @@ static int read_inode_bitmap (struct sup + * this group. The IO will be retried next time. + */ + error_out: +- sb->u.ext3_sb.s_inode_bitmap_number[bitmap_nr] = block_group; +- sb->u.ext3_sb.s_inode_bitmap[bitmap_nr] = bh; ++ EXT3_SB(sb)->s_inode_bitmap_number[bitmap_nr] = block_group; ++ EXT3_SB(sb)->s_inode_bitmap[bitmap_nr] = bh; + return retval; + } + +@@ -225,7 +225,7 @@ void ext3_free_inode (handle_t *handle, + clear_inode (inode); + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = EXT3_SB(sb)->s_es; + if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_free_inode", + "reserved or nonexistent inode %lu", ino); +@@ -237,7 +237,7 @@ void ext3_free_inode (handle_t *handle, + if (bitmap_nr < 0) + goto error_return; + +- bh = sb->u.ext3_sb.s_inode_bitmap[bitmap_nr]; ++ bh = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; + + BUFFER_TRACE(bh, "get_write_access"); + fatal = ext3_journal_get_write_access(handle, bh); +@@ -255,8 +255,8 @@ void ext3_free_inode (handle_t *handle, + fatal = ext3_journal_get_write_access(handle, bh2); + if (fatal) goto error_return; + +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "get write access"); +- fatal = ext3_journal_get_write_access(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "get write access"); ++ fatal = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); + if (fatal) goto error_return; + + if (gdp) { +@@ -271,9 +271,9 @@ void ext3_free_inode (handle_t *handle, + if (!fatal) fatal = err; + es->s_free_inodes_count = + cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) + 1); +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, + "call ext3_journal_dirty_metadata"); +- err = ext3_journal_dirty_metadata(handle, sb->u.ext3_sb.s_sbh); ++ err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); + if (!fatal) fatal = err; + } + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +@@ -305,6 +305,8 @@ struct inode * ext3_new_inode (handle_t + int i, j, avefreei; + struct inode * inode; + int bitmap_nr; ++ struct ext3_inode_info *ei; ++ struct ext3_sb_info *sbi; + struct ext3_group_desc * gdp; + struct ext3_group_desc * tmp; + struct ext3_super_block * es; +@@ -318,19 +320,21 @@ struct inode * ext3_new_inode (handle_t + inode = new_inode(sb); + if (!inode) + return ERR_PTR(-ENOMEM); +- init_rwsem(&inode->u.ext3_i.truncate_sem); ++ sbi = EXT3_SB(sb); ++ ei = EXT3_I(inode); ++ init_rwsem(&ei->truncate_sem); + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; ++ es = sbi->s_es; + repeat: + gdp = NULL; + i = 0; + + if (S_ISDIR(mode)) { + avefreei = le32_to_cpu(es->s_free_inodes_count) / +- sb->u.ext3_sb.s_groups_count; ++ sbi->s_groups_count; + if (!gdp) { +- for (j = 0; j < sb->u.ext3_sb.s_groups_count; j++) { ++ for (j = 0; j < sbi->s_groups_count; j++) { + struct buffer_head *temp_buffer; + tmp = ext3_get_group_desc (sb, j, &temp_buffer); + if (tmp && +@@ -350,7 +354,7 @@ repeat: + /* + * Try to place the inode in its parent directory + */ +- i = dir->u.ext3_i.i_block_group; ++ i = EXT3_I(dir)->i_block_group; + tmp = ext3_get_group_desc (sb, i, &bh2); + if (tmp && le16_to_cpu(tmp->bg_free_inodes_count)) + gdp = tmp; +@@ -360,10 +364,10 @@ repeat: + * Use a quadratic hash to find a group with a + * free inode + */ +- for (j = 1; j < sb->u.ext3_sb.s_groups_count; j <<= 1) { ++ for (j = 1; j < sbi->s_groups_count; j <<= 1) { + i += j; +- if (i >= sb->u.ext3_sb.s_groups_count) +- i -= sb->u.ext3_sb.s_groups_count; ++ if (i >= sbi->s_groups_count) ++ i -= sbi->s_groups_count; + tmp = ext3_get_group_desc (sb, i, &bh2); + if (tmp && + le16_to_cpu(tmp->bg_free_inodes_count)) { +@@ -376,9 +380,9 @@ repeat: + /* + * That failed: try linear search for a free inode + */ +- i = dir->u.ext3_i.i_block_group + 1; +- for (j = 2; j < sb->u.ext3_sb.s_groups_count; j++) { +- if (++i >= sb->u.ext3_sb.s_groups_count) ++ i = EXT3_I(dir)->i_block_group + 1; ++ for (j = 2; j < sbi->s_groups_count; j++) { ++ if (++i >= sbi->s_groups_count) + i = 0; + tmp = ext3_get_group_desc (sb, i, &bh2); + if (tmp && +@@ -399,11 +403,11 @@ repeat: + if (bitmap_nr < 0) + goto fail; + +- bh = sb->u.ext3_sb.s_inode_bitmap[bitmap_nr]; ++ bh = sbi->s_inode_bitmap[bitmap_nr]; + + if ((j = ext3_find_first_zero_bit ((unsigned long *) bh->b_data, +- EXT3_INODES_PER_GROUP(sb))) < +- EXT3_INODES_PER_GROUP(sb)) { ++ sbi->s_inodes_per_group)) < ++ sbi->s_inodes_per_group) { + BUFFER_TRACE(bh, "get_write_access"); + err = ext3_journal_get_write_access(handle, bh); + if (err) goto fail; +@@ -436,8 +440,8 @@ repeat: + } + goto repeat; + } +- j += i * EXT3_INODES_PER_GROUP(sb) + 1; +- if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) { ++ j += i * sbi->s_inodes_per_group + 1; ++ if (j < sbi->s_first_ino || j > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", + "reserved inode or inode > inodes count - " + "block_group = %d,inode=%d", i, j); +@@ -457,13 +461,13 @@ repeat: + err = ext3_journal_dirty_metadata(handle, bh2); + if (err) goto fail; + +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "get_write_access"); +- err = ext3_journal_get_write_access(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(sbi->s_sbh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, sbi->s_sbh); + if (err) goto fail; + es->s_free_inodes_count = + cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1); +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "call ext3_journal_dirty_metadata"); +- err = ext3_journal_dirty_metadata(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(sbi->s_sbh, "call ext3_journal_dirty_metadata"); ++ err = ext3_journal_dirty_metadata(handle, sbi->s_sbh); + sb->s_dirt = 1; + if (err) goto fail; + +@@ -483,31 +487,31 @@ repeat: + inode->i_blksize = PAGE_SIZE; + inode->i_blocks = 0; + inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; +- inode->u.ext3_i.i_flags = dir->u.ext3_i.i_flags & ~EXT3_INDEX_FL; ++ ei->i_flags = EXT3_I(dir)->i_flags & ~EXT3_INDEX_FL; + if (S_ISLNK(mode)) +- inode->u.ext3_i.i_flags &= ~(EXT3_IMMUTABLE_FL|EXT3_APPEND_FL); ++ ei->i_flags &= ~(EXT3_IMMUTABLE_FL|EXT3_APPEND_FL); + #ifdef EXT3_FRAGMENTS +- inode->u.ext3_i.i_faddr = 0; +- inode->u.ext3_i.i_frag_no = 0; +- inode->u.ext3_i.i_frag_size = 0; ++ ei->i_faddr = 0; ++ ei->i_frag_no = 0; ++ ei->i_frag_size = 0; + #endif +- inode->u.ext3_i.i_file_acl = 0; +- inode->u.ext3_i.i_dir_acl = 0; +- inode->u.ext3_i.i_dtime = 0; +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ ei->i_file_acl = 0; ++ ei->i_dir_acl = 0; ++ ei->i_dtime = 0; ++ INIT_LIST_HEAD(&ei->i_orphan); + #ifdef EXT3_PREALLOCATE +- inode->u.ext3_i.i_prealloc_count = 0; ++ ei->i_prealloc_count = 0; + #endif +- inode->u.ext3_i.i_block_group = i; ++ ei->i_block_group = i; + +- if (inode->u.ext3_i.i_flags & EXT3_SYNC_FL) ++ if (ei->i_flags & EXT3_SYNC_FL) + inode->i_flags |= S_SYNC; + if (IS_SYNC(inode)) + handle->h_sync = 1; + insert_inode_hash(inode); +- inode->i_generation = sb->u.ext3_sb.s_next_generation++; ++ inode->i_generation = sbi->s_next_generation++; + +- inode->u.ext3_i.i_state = EXT3_STATE_NEW; ++ ei->i_state = EXT3_STATE_NEW; + err = ext3_mark_inode_dirty(handle, inode); + if (err) goto fail; + +@@ -585,19 +589,19 @@ struct inode *ext3_orphan_get (struct su + + unsigned long ext3_count_free_inodes (struct super_block * sb) + { ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ struct ext3_super_block *es = sbi->s_es; + #ifdef EXT3FS_DEBUG +- struct ext3_super_block * es; + unsigned long desc_count, bitmap_count, x; + int bitmap_nr; + struct ext3_group_desc * gdp; + int i; + + lock_super (sb); +- es = sb->u.ext3_sb.s_es; + desc_count = 0; + bitmap_count = 0; + gdp = NULL; +- for (i = 0; i < sb->u.ext3_sb.s_groups_count; i++) { ++ for (i = 0; i < sbi->s_groups_count; i++) { + gdp = ext3_get_group_desc (sb, i, NULL); + if (!gdp) + continue; +@@ -606,8 +610,8 @@ unsigned long ext3_count_free_inodes (st + if (bitmap_nr < 0) + continue; + +- x = ext3_count_free (sb->u.ext3_sb.s_inode_bitmap[bitmap_nr], +- EXT3_INODES_PER_GROUP(sb) / 8); ++ x = ext3_count_free(sbi->s_inode_bitmap[bitmap_nr], ++ sbi->s_inodes_per_group / 8); + printk ("group %d: stored = %d, counted = %lu\n", + i, le16_to_cpu(gdp->bg_free_inodes_count), x); + bitmap_count += x; +@@ -617,7 +621,7 @@ unsigned long ext3_count_free_inodes (st + unlock_super (sb); + return desc_count; + #else +- return le32_to_cpu(sb->u.ext3_sb.s_es->s_free_inodes_count); ++ return le32_to_cpu(es->s_free_inodes_count); + #endif + } + +@@ -626,16 +630,18 @@ unsigned long ext3_count_free_inodes (st + void ext3_check_inodes_bitmap (struct super_block * sb) + { + struct ext3_super_block * es; ++ struct ext3_sb_info *sbi; + unsigned long desc_count, bitmap_count, x; + int bitmap_nr; + struct ext3_group_desc * gdp; + int i; + +- es = sb->u.ext3_sb.s_es; ++ sbi = EXT3_SB(sb); ++ es = sbi->s_es; + desc_count = 0; + bitmap_count = 0; + gdp = NULL; +- for (i = 0; i < sb->u.ext3_sb.s_groups_count; i++) { ++ for (i = 0; i < sbi->s_groups_count; i++) { + gdp = ext3_get_group_desc (sb, i, NULL); + if (!gdp) + continue; +@@ -644,7 +650,7 @@ void ext3_check_inodes_bitmap (struct su + if (bitmap_nr < 0) + continue; + +- x = ext3_count_free (sb->u.ext3_sb.s_inode_bitmap[bitmap_nr], ++ x = ext3_count_free (sbi->s_inode_bitmap[bitmap_nr], + EXT3_INODES_PER_GROUP(sb) / 8); + if (le16_to_cpu(gdp->bg_free_inodes_count) != x) + ext3_error (sb, "ext3_check_inodes_bitmap", +--- ./fs/ext3/inode.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/inode.c Tue May 7 15:41:23 2002 +@@ -196,7 +196,7 @@ void ext3_delete_inode (struct inode * i + * (Well, we could do this if we need to, but heck - it works) + */ + ext3_orphan_del(handle, inode); +- inode->u.ext3_i.i_dtime = CURRENT_TIME; ++ EXT3_I(inode)->i_dtime = CURRENT_TIME; + + /* + * One subtle ordering requirement: if anything has gone wrong +@@ -220,13 +220,14 @@ no_delete: + void ext3_discard_prealloc (struct inode * inode) + { + #ifdef EXT3_PREALLOCATE ++ struct ext3_inode_info *ei = EXT3_I(inode); + lock_kernel(); + /* Writer: ->i_prealloc* */ +- if (inode->u.ext3_i.i_prealloc_count) { +- unsigned short total = inode->u.ext3_i.i_prealloc_count; +- unsigned long block = inode->u.ext3_i.i_prealloc_block; +- inode->u.ext3_i.i_prealloc_count = 0; +- inode->u.ext3_i.i_prealloc_block = 0; ++ if (ei->i_prealloc_count) { ++ unsigned short total = ei->i_prealloc_count; ++ unsigned long block = ei->i_prealloc_block; ++ ei->i_prealloc_count = 0; ++ ei->i_prealloc_block = 0; + /* Writer: end */ + ext3_free_blocks (inode, block, total); + } +@@ -243,13 +244,15 @@ static int ext3_alloc_block (handle_t *h + unsigned long result; + + #ifdef EXT3_PREALLOCATE ++ struct ext3_inode_info *ei = EXT3_I(inode); ++ + /* Writer: ->i_prealloc* */ +- if (inode->u.ext3_i.i_prealloc_count && +- (goal == inode->u.ext3_i.i_prealloc_block || +- goal + 1 == inode->u.ext3_i.i_prealloc_block)) ++ if (ei->i_prealloc_count && ++ (goal == ei->i_prealloc_block || ++ goal + 1 == ei->i_prealloc_block)) + { +- result = inode->u.ext3_i.i_prealloc_block++; +- inode->u.ext3_i.i_prealloc_count--; ++ result = ei->i_prealloc_block++; ++ ei->i_prealloc_count--; + /* Writer: end */ + ext3_debug ("preallocation hit (%lu/%lu).\n", + ++alloc_hits, ++alloc_attempts); +@@ -259,8 +262,8 @@ static int ext3_alloc_block (handle_t *h + alloc_hits, ++alloc_attempts); + if (S_ISREG(inode->i_mode)) + result = ext3_new_block (inode, goal, +- &inode->u.ext3_i.i_prealloc_count, +- &inode->u.ext3_i.i_prealloc_block, err); ++ &ei->i_prealloc_count, ++ &ei->i_prealloc_block, err); + else + result = ext3_new_block (inode, goal, 0, 0, err); + /* +@@ -394,7 +397,7 @@ static Indirect *ext3_get_branch(struct + + *err = 0; + /* i_data is not going away, no lock needed */ +- add_chain (chain, NULL, inode->u.ext3_i.i_data + *offsets); ++ add_chain (chain, NULL, EXT3_I(inode)->i_data + *offsets); + if (!p->key) + goto no_block; + while (--depth) { +@@ -437,7 +440,8 @@ no_block: + + static inline unsigned long ext3_find_near(struct inode *inode, Indirect *ind) + { +- u32 *start = ind->bh ? (u32*) ind->bh->b_data : inode->u.ext3_i.i_data; ++ struct ext3_inode_info *ei = EXT3_I(inode); ++ u32 *start = ind->bh ? (u32*) ind->bh->b_data : ei->i_data; + u32 *p; + + /* Try to find previous block */ +@@ -453,9 +456,8 @@ static inline unsigned long ext3_find_ne + * It is going to be refered from inode itself? OK, just put it into + * the same cylinder group then. + */ +- return (inode->u.ext3_i.i_block_group * +- EXT3_BLOCKS_PER_GROUP(inode->i_sb)) + +- le32_to_cpu(inode->i_sb->u.ext3_sb.s_es->s_first_data_block); ++ return (ei->i_block_group * EXT3_BLOCKS_PER_GROUP(inode->i_sb)) + ++ le32_to_cpu(EXT3_SB(inode->i_sb)->s_es->s_first_data_block); + } + + /** +@@ -474,14 +477,15 @@ + static int ext3_find_goal(struct inode *inode, long block, Indirect chain[4], + Indirect *partial, unsigned long *goal) + { ++ struct ext3_inode_info *ei = EXT3_I(inode); + /* Writer: ->i_next_alloc* */ +- if (block == inode->u.ext3_i.i_next_alloc_block + 1) { +- inode->u.ext3_i.i_next_alloc_block++; +- inode->u.ext3_i.i_next_alloc_goal++; ++ if (block == ei->i_next_alloc_block + 1) { ++ ei->i_next_alloc_block++; ++ ei->i_next_alloc_goal++; + } + #ifdef SEARCH_FROM_ZERO +- inode->u.ext3_i.i_next_alloc_block = 0; +- inode->u.ext3_i.i_next_alloc_goal = 0; ++ ei->i_next_alloc_block = 0; ++ ei->i_next_alloc_goal = 0; + #endif + /* Writer: end */ + /* Reader: pointers, ->i_next_alloc* */ +@@ -490,8 +493,8 @@ static int ext3_find_goal(struct inode * + * try the heuristic for sequential allocation, + * failing that at least try to get decent locality. + */ +- if (block == inode->u.ext3_i.i_next_alloc_block) +- *goal = inode->u.ext3_i.i_next_alloc_goal; ++ if (block == ei->i_next_alloc_block) ++ *goal = ei->i_next_alloc_goal; + if (!*goal) + *goal = ext3_find_near(inode, partial); + #ifdef SEARCH_FROM_ZERO +@@ -619,6 +621,7 @@ + { + int i; + int err = 0; ++ struct ext3_inode_info *ei = EXT3_I(inode); + + /* + * If we're splicing into a [td]indirect block (as opposed to the +@@ -641,11 +644,11 @@ static int ext3_splice_branch(handle_t * + /* That's it */ + + *where->p = where->key; +- inode->u.ext3_i.i_next_alloc_block = block; +- inode->u.ext3_i.i_next_alloc_goal = le32_to_cpu(where[num-1].key); ++ ei->i_next_alloc_block = block; ++ ei->i_next_alloc_goal = le32_to_cpu(where[num-1].key); + #ifdef SEARCH_FROM_ZERO +- inode->u.ext3_i.i_next_alloc_block = 0; +- inode->u.ext3_i.i_next_alloc_goal = 0; ++ ei->i_next_alloc_block = 0; ++ ei->i_next_alloc_goal = 0; + #endif + /* Writer: end */ + +@@ -729,6 +732,7 @@ + unsigned long goal; + int left; + int depth = ext3_block_to_path(inode, iblock, offsets); ++ struct ext3_inode_info *ei = EXT3_I(inode); + loff_t new_size; + + J_ASSERT(handle != NULL || create == 0); +@@ -782,7 +785,7 @@ out: + /* + * Block out ext3_truncate while we alter the tree + */ +- down_read(&inode->u.ext3_i.truncate_sem); ++ down_read(&ei->truncate_sem); + err = ext3_alloc_branch(handle, inode, left, goal, + offsets+(partial-chain), partial); + +@@ -794,7 +797,7 @@ out: + if (!err) + err = ext3_splice_branch(handle, inode, iblock, chain, + partial, left); +- up_read(&inode->u.ext3_i.truncate_sem); ++ up_read(&ei->truncate_sem); + if (err == -EAGAIN) + goto changed; + if (err) +@@ -807,8 +810,8 @@ out: + * truncate is in progress. It is racy between multiple parallel + * instances of get_block, but we have the BKL. + */ +- if (new_size > inode->u.ext3_i.i_disksize) +- inode->u.ext3_i.i_disksize = new_size; ++ if (new_size > ei->i_disksize) ++ ei->i_disksize = new_size; + + bh_result->b_state |= (1UL << BH_New); + goto got_it; +@@ -921,7 +924,7 @@ struct buffer_head *ext3_bread(handle_t + struct buffer_head *tmp_bh; + + for (i = 1; +- inode->u.ext3_i.i_prealloc_count && ++ EXT3_I(inode)->i_prealloc_count && + i < EXT3_SB(inode->i_sb)->s_es->s_prealloc_dir_blocks; + i++) { + /* +@@ -1131,8 +1134,8 @@ static int ext3_commit_write(struct file + kunmap(page); + } + } +- if (inode->i_size > inode->u.ext3_i.i_disksize) { +- inode->u.ext3_i.i_disksize = inode->i_size; ++ if (inode->i_size > EXT3_I(inode)->i_disksize) { ++ EXT3_I(inode)->i_disksize = inode->i_size; + ret2 = ext3_mark_inode_dirty(handle, inode); + if (!ret) + ret = ret2; +@@ -1832,7 +1835,8 @@ static void ext3_free_branches(handle_t + void ext3_truncate(struct inode * inode) + { + handle_t *handle; +- u32 *i_data = inode->u.ext3_i.i_data; ++ struct ext3_inode_info *ei = EXT3_I(inode); ++ u32 *i_data = EXT3_I(inode)->i_data; + int addr_per_block = EXT3_ADDR_PER_BLOCK(inode->i_sb); + int offsets[4]; + Indirect chain[4]; +@@ -1884,13 +1887,13 @@ void ext3_truncate(struct inode * inode) + * on-disk inode. We do this via i_disksize, which is the value which + * ext3 *really* writes onto the disk inode. + */ +- inode->u.ext3_i.i_disksize = inode->i_size; ++ ei->i_disksize = inode->i_size; + + /* + * From here we block out all ext3_get_block() callers who want to + * modify the block allocation tree. + */ +- down_write(&inode->u.ext3_i.truncate_sem); ++ down_write(&ei->truncate_sem); + + if (n == 1) { /* direct blocks */ + ext3_free_data(handle, inode, NULL, i_data+offsets[0], +@@ -1954,7 +1957,7 @@ do_indirects: + case EXT3_TIND_BLOCK: + ; + } +- up_write(&inode->u.ext3_i.truncate_sem); ++ up_write(&ei->truncate_sem); + inode->i_mtime = inode->i_ctime = CURRENT_TIME; + ext3_mark_inode_dirty(handle, inode); + +@@ -1983,6 +1986,8 @@ out_stop: + + int ext3_get_inode_loc (struct inode *inode, struct ext3_iloc *iloc) + { ++ struct super_block *sb = inode->i_sb; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); + struct buffer_head *bh = 0; + unsigned long block; + unsigned long block_group; +@@ -1997,23 +2010,19 @@ int ext3_get_inode_loc (struct inode *in + inode->i_ino != EXT3_JOURNAL_INO && +- inode->i_ino < EXT3_FIRST_INO(inode->i_sb)) || +- inode->i_ino > le32_to_cpu( +- inode->i_sb->u.ext3_sb.s_es->s_inodes_count)) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "bad inode number: %lu", inode->i_ino); ++ inode->i_ino < EXT3_FIRST_INO(sb)) || ++ inode->i_ino > le32_to_cpu(sbi->s_es->s_inodes_count)) { ++ ext3_error (sb, __FUNCTION__, "bad inode #%lu", inode->i_ino); + goto bad_inode; + } +- block_group = (inode->i_ino - 1) / EXT3_INODES_PER_GROUP(inode->i_sb); +- if (block_group >= inode->i_sb->u.ext3_sb.s_groups_count) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "group >= groups count"); ++ block_group = (inode->i_ino - 1) / sbi->s_inodes_per_group; ++ if (block_group >= sbi->s_groups_count) { ++ ext3_error(sb, __FUNCTION__, "group >= groups count"); + goto bad_inode; + } +- group_desc = block_group >> EXT3_DESC_PER_BLOCK_BITS(inode->i_sb); +- desc = block_group & (EXT3_DESC_PER_BLOCK(inode->i_sb) - 1); +- bh = inode->i_sb->u.ext3_sb.s_group_desc[group_desc]; ++ group_desc = block_group >> sbi->s_desc_per_block_bits; ++ desc = block_group & (sbi->s_desc_per_block - 1); ++ bh = sbi->s_group_desc[group_desc]; + if (!bh) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "Descriptor not loaded"); ++ ext3_error(sb, __FUNCTION__, "Descriptor not loaded"); + goto bad_inode; + } + +@@ -2021,17 +2022,17 @@ int ext3_get_inode_loc (struct inode *in + /* + * Figure out the offset within the block group inode table + */ +- offset = ((inode->i_ino - 1) % EXT3_INODES_PER_GROUP(inode->i_sb)) * +- EXT3_INODE_SIZE(inode->i_sb); ++ offset = ((inode->i_ino - 1) % sbi->s_inodes_per_group) * ++ sbi->s_inode_size; + block = le32_to_cpu(gdp[desc].bg_inode_table) + +- (offset >> EXT3_BLOCK_SIZE_BITS(inode->i_sb)); +- if (!(bh = sb_bread(inode->i_sb, block))) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", ++ (offset >> EXT3_BLOCK_SIZE_BITS(sb)); ++ if (!(bh = sb_bread(sb, block))) { ++ ext3_error (sb, __FUNCTION__, + "unable to read inode block - " + "inode=%lu, block=%lu", inode->i_ino, block); + goto bad_inode; + } +- offset &= (EXT3_BLOCK_SIZE(inode->i_sb) - 1); ++ offset &= (EXT3_BLOCK_SIZE(sb) - 1); + + iloc->bh = bh; + iloc->raw_inode = (struct ext3_inode *) (bh->b_data + offset); +@@ -2047,6 +2048,7 @@ void ext3_read_inode(struct inode * inod + { + struct ext3_iloc iloc; + struct ext3_inode *raw_inode; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct buffer_head *bh; + int block; + +@@ -2054,7 +2056,7 @@ void ext3_read_inode(struct inode * inod + goto bad_inode; + bh = iloc.bh; + raw_inode = iloc.raw_inode; +- init_rwsem(&inode->u.ext3_i.truncate_sem); ++ init_rwsem(&ei->truncate_sem); + inode->i_mode = le16_to_cpu(raw_inode->i_mode); + inode->i_uid = (uid_t)le16_to_cpu(raw_inode->i_uid_low); + inode->i_gid = (gid_t)le16_to_cpu(raw_inode->i_gid_low); +@@ -2067,7 +2069,7 @@ void ext3_read_inode(struct inode * inod + inode->i_atime = le32_to_cpu(raw_inode->i_atime); + inode->i_ctime = le32_to_cpu(raw_inode->i_ctime); + inode->i_mtime = le32_to_cpu(raw_inode->i_mtime); +- inode->u.ext3_i.i_dtime = le32_to_cpu(raw_inode->i_dtime); ++ ei->i_dtime = le32_to_cpu(raw_inode->i_dtime); + /* We now have enough fields to check if the inode was active or not. + * This is needed because nfsd might try to access dead inodes + * the test is that same one that e2fsck uses +@@ -2075,7 +2077,7 @@ void ext3_read_inode(struct inode * inod + */ + if (inode->i_nlink == 0) { + if (inode->i_mode == 0 || +- !(inode->i_sb->u.ext3_sb.s_mount_state & EXT3_ORPHAN_FS)) { ++ !(EXT3_SB(inode->i_sb)->s_mount_state & EXT3_ORPHAN_FS)) { + /* this inode is deleted */ + brelse (bh); + goto bad_inode; +@@ -2090,33 +2092,33 @@ void ext3_read_inode(struct inode * inod + * size */ + inode->i_blocks = le32_to_cpu(raw_inode->i_blocks); + inode->i_version = ++event; +- inode->u.ext3_i.i_flags = le32_to_cpu(raw_inode->i_flags); ++ ei->i_flags = le32_to_cpu(raw_inode->i_flags); + #ifdef EXT3_FRAGMENTS +- inode->u.ext3_i.i_faddr = le32_to_cpu(raw_inode->i_faddr); +- inode->u.ext3_i.i_frag_no = raw_inode->i_frag; +- inode->u.ext3_i.i_frag_size = raw_inode->i_fsize; ++ ei->i_faddr = le32_to_cpu(raw_inode->i_faddr); ++ ei->i_frag_no = raw_inode->i_frag; ++ ei->i_frag_size = raw_inode->i_fsize; + #endif +- inode->u.ext3_i.i_file_acl = le32_to_cpu(raw_inode->i_file_acl); ++ ei->i_file_acl = le32_to_cpu(raw_inode->i_file_acl); + if (!S_ISREG(inode->i_mode)) { +- inode->u.ext3_i.i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl); ++ ei->i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl); + } else { + inode->i_size |= + ((__u64)le32_to_cpu(raw_inode->i_size_high)) << 32; + } +- inode->u.ext3_i.i_disksize = inode->i_size; ++ ei->i_disksize = inode->i_size; + inode->i_generation = le32_to_cpu(raw_inode->i_generation); + #ifdef EXT3_PREALLOCATE +- inode->u.ext3_i.i_prealloc_count = 0; ++ ei->i_prealloc_count = 0; + #endif +- inode->u.ext3_i.i_block_group = iloc.block_group; ++ ei->i_block_group = iloc.block_group; + + /* + * NOTE! The in-memory inode i_data array is in little-endian order + * even on big-endian machines: we do NOT byteswap the block numbers! + */ + for (block = 0; block < EXT3_N_BLOCKS; block++) +- inode->u.ext3_i.i_data[block] = iloc.raw_inode->i_block[block]; +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ ei->i_data[block] = iloc.raw_inode->i_block[block]; ++ INIT_LIST_HEAD(&ei->i_orphan); + + brelse (iloc.bh); + +@@ -2143,17 +2145,17 @@ void ext3_read_inode(struct inode * inod + /* inode->i_attr_flags = 0; unused */ +- if (inode->u.ext3_i.i_flags & EXT3_SYNC_FL) { ++ if (ei->i_flags & EXT3_SYNC_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_SYNCRONOUS; unused */ + inode->i_flags |= S_SYNC; + } +- if (inode->u.ext3_i.i_flags & EXT3_APPEND_FL) { ++ if (ei->i_flags & EXT3_APPEND_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_APPEND; unused */ + inode->i_flags |= S_APPEND; + } +- if (inode->u.ext3_i.i_flags & EXT3_IMMUTABLE_FL) { ++ if (ei->i_flags & EXT3_IMMUTABLE_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_IMMUTABLE; unused */ + inode->i_flags |= S_IMMUTABLE; + } +- if (inode->u.ext3_i.i_flags & EXT3_NOATIME_FL) { ++ if (ei->i_flags & EXT3_NOATIME_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_NOATIME; unused */ + inode->i_flags |= S_NOATIME; + } +@@ -2175,6 +2177,7 @@ static int ext3_do_update_inode(handle_t + struct ext3_iloc *iloc) + { + struct ext3_inode *raw_inode = iloc->raw_inode; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct buffer_head *bh = iloc->bh; + int err = 0, rc, block; + +@@ -2192,7 +2195,7 @@ static int ext3_do_update_inode(handle_t + * Fix up interoperability with old kernels. Otherwise, old inodes get + * re-used with the upper 16 bits of the uid/gid intact + */ +- if(!inode->u.ext3_i.i_dtime) { ++ if(!ei->i_dtime) { + raw_inode->i_uid_high = + cpu_to_le16(high_16_bits(inode->i_uid)); + raw_inode->i_gid_high = +@@ -2210,34 +2213,33 @@ static int ext3_do_update_inode(handle_t + raw_inode->i_gid_high = 0; + } + raw_inode->i_links_count = cpu_to_le16(inode->i_nlink); +- raw_inode->i_size = cpu_to_le32(inode->u.ext3_i.i_disksize); ++ raw_inode->i_size = cpu_to_le32(ei->i_disksize); + raw_inode->i_atime = cpu_to_le32(inode->i_atime); + raw_inode->i_ctime = cpu_to_le32(inode->i_ctime); + raw_inode->i_mtime = cpu_to_le32(inode->i_mtime); + raw_inode->i_blocks = cpu_to_le32(inode->i_blocks); +- raw_inode->i_dtime = cpu_to_le32(inode->u.ext3_i.i_dtime); +- raw_inode->i_flags = cpu_to_le32(inode->u.ext3_i.i_flags); ++ raw_inode->i_dtime = cpu_to_le32(ei->i_dtime); ++ raw_inode->i_flags = cpu_to_le32(ei->i_flags); + #ifdef EXT3_FRAGMENTS +- raw_inode->i_faddr = cpu_to_le32(inode->u.ext3_i.i_faddr); +- raw_inode->i_frag = inode->u.ext3_i.i_frag_no; +- raw_inode->i_fsize = inode->u.ext3_i.i_frag_size; ++ raw_inode->i_faddr = cpu_to_le32(ei->i_faddr); ++ raw_inode->i_frag = ei->i_frag_no; ++ raw_inode->i_fsize = ei->i_frag_size; + #else + /* If we are not tracking these fields in the in-memory inode, + * then preserve them on disk, but still initialise them to zero + * for new inodes. */ +- if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) { ++ if (ei->i_state & EXT3_STATE_NEW) { + raw_inode->i_faddr = 0; + raw_inode->i_frag = 0; + raw_inode->i_fsize = 0; + } + #endif +- raw_inode->i_file_acl = cpu_to_le32(inode->u.ext3_i.i_file_acl); ++ raw_inode->i_file_acl = cpu_to_le32(ei->i_file_acl); + if (!S_ISREG(inode->i_mode)) { +- raw_inode->i_dir_acl = cpu_to_le32(inode->u.ext3_i.i_dir_acl); ++ raw_inode->i_dir_acl = cpu_to_le32(ei->i_dir_acl); + } else { +- raw_inode->i_size_high = +- cpu_to_le32(inode->u.ext3_i.i_disksize >> 32); +- if (inode->u.ext3_i.i_disksize > 0x7fffffffULL) { ++ raw_inode->i_size_high = cpu_to_le32(ei->i_disksize >> 32); ++ if (ei->i_disksize > MAX_NON_LFS) { + struct super_block *sb = inode->i_sb; + if (!EXT3_HAS_RO_COMPAT_FEATURE(sb, + EXT3_FEATURE_RO_COMPAT_LARGE_FILE) || +@@ -2247,7 +2249,7 @@ static int ext3_do_update_inode(handle_t + * created, add a flag to the superblock. + */ + err = ext3_journal_get_write_access(handle, +- sb->u.ext3_sb.s_sbh); ++ EXT3_SB(sb)->s_sbh); + if (err) + goto out_brelse; + ext3_update_dynamic_rev(sb); +@@ -2256,7 +2258,7 @@ static int ext3_do_update_inode(handle_t + sb->s_dirt = 1; + handle->h_sync = 1; + err = ext3_journal_dirty_metadata(handle, +- sb->u.ext3_sb.s_sbh); ++ EXT3_SB(sb)->s_sbh); + } + } + } +@@ -2265,13 +2267,13 @@ static int ext3_do_update_inode(handle_t + raw_inode->i_block[0] = + cpu_to_le32(kdev_t_to_nr(inode->i_rdev)); + else for (block = 0; block < EXT3_N_BLOCKS; block++) +- raw_inode->i_block[block] = inode->u.ext3_i.i_data[block]; ++ raw_inode->i_block[block] = ei->i_data[block]; + + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); + rc = ext3_journal_dirty_metadata(handle, bh); + if (!err) + err = rc; +- EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW; ++ ei->i_state &= ~EXT3_STATE_NEW; + + out_brelse: + brelse (bh); +@@ -2379,7 +2381,7 @@ int ext3_setattr(struct dentry *dentry, + } + + error = ext3_orphan_add(handle, inode); +- inode->u.ext3_i.i_disksize = attr->ia_size; ++ EXT3_I(inode)->i_disksize = attr->ia_size; + rc = ext3_mark_inode_dirty(handle, inode); + if (!error) + error = rc; +@@ -2622,9 +2624,9 @@ int ext3_change_inode_journal_flag(struc + */ + + if (val) +- inode->u.ext3_i.i_flags |= EXT3_JOURNAL_DATA_FL; ++ EXT3_I(inode)->i_flags |= EXT3_JOURNAL_DATA_FL; + else +- inode->u.ext3_i.i_flags &= ~EXT3_JOURNAL_DATA_FL; ++ EXT3_I(inode)->i_flags &= ~EXT3_JOURNAL_DATA_FL; + + journal_unlock_updates(journal); + +--- ./fs/ext3/ioctl.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/ioctl.c Tue May 7 15:20:52 2002 +@@ -18,13 +18,14 @@ + int ext3_ioctl (struct inode * inode, struct file * filp, unsigned int cmd, + unsigned long arg) + { ++ struct ext3_inode_info *ei = EXT3_I(inode); + unsigned int flags; + + ext3_debug ("cmd = %u, arg = %lu\n", cmd, arg); + + switch (cmd) { + case EXT3_IOC_GETFLAGS: +- flags = inode->u.ext3_i.i_flags & EXT3_FL_USER_VISIBLE; ++ flags = ei->i_flags & EXT3_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); + case EXT3_IOC_SETFLAGS: { + handle_t *handle = NULL; +@@ -42,7 +42,7 @@ int ext3_ioctl (struct inode * inode, st + if (get_user(flags, (int *) arg)) + return -EFAULT; + +- oldflags = inode->u.ext3_i.i_flags; ++ oldflags = ei->i_flags; + + /* The JOURNAL_DATA flag is modifiable only by root */ + jflag = flags & EXT3_JOURNAL_DATA_FL; +@@ -79,7 +79,7 @@ int ext3_ioctl (struct inode * inode, st + + flags = flags & EXT3_FL_USER_MODIFIABLE; + flags |= oldflags & ~EXT3_FL_USER_MODIFIABLE; +- inode->u.ext3_i.i_flags = flags; ++ ei->i_flags = flags; + + if (flags & EXT3_SYNC_FL) + inode->i_flags |= S_SYNC; +@@ -155,12 +155,12 @@ flags_err: + int ret = 0; + + set_current_state(TASK_INTERRUPTIBLE); +- add_wait_queue(&sb->u.ext3_sb.ro_wait_queue, &wait); +- if (timer_pending(&sb->u.ext3_sb.turn_ro_timer)) { ++ add_wait_queue(&EXT3_SB(sb)->ro_wait_queue, &wait); ++ if (timer_pending(&EXT3_SB(sb)->turn_ro_timer)) { + schedule(); + ret = 1; + } +- remove_wait_queue(&sb->u.ext3_sb.ro_wait_queue, &wait); ++ remove_wait_queue(&EXT3_SB(sb)->ro_wait_queue, &wait); + return ret; + } + #endif +--- ./fs/ext3/namei.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/namei.c Tue May 7 16:05:51 2002 +@@ -636,7 +636,7 @@ static struct buffer_head * ext3_find_en + } + + 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; +@@ -677,7 +677,7 @@ 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 { +@@ -1419,7 +1419,7 @@ 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 +@@ -1430,8 +1430,8 @@ int ext3_orphan_add(handle_t *handle, st + J_ASSERT ((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); + +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "get_write_access"); +- err = ext3_journal_get_write_access(handle, sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(EXT3_SB(sb)->s_sbh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); + if (err) + goto out_unlock; + +@@ -1442,7 +1442,7 @@ int ext3_orphan_add(handle_t *handle, st + /* Insert this inode at the head of the on-disk orphan list... */ + NEXT_ORPHAN(inode) = le32_to_cpu(EXT3_SB(sb)->s_es->s_last_orphan); + EXT3_SB(sb)->s_es->s_last_orphan = cpu_to_le32(inode->i_ino); +- err = ext3_journal_dirty_metadata(handle, sb->u.ext3_sb.s_sbh); ++ err = ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); + rc = ext3_mark_iloc_dirty(handle, inode, &iloc); + if (!err) + err = rc; +@@ -1456,7 +1456,7 @@ 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 +770,25 @@ + int ext3_orphan_del(handle_t *handle, struct inode *inode) + { + struct list_head *prev; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct ext3_sb_info *sbi; + unsigned long ino_next; + struct ext3_iloc iloc; + int err = 0; + + lock_super(inode->i_sb); +- if (list_empty(&inode->u.ext3_i.i_orphan)) { ++ if (list_empty(&ei->i_orphan)) { + unlock_super(inode->i_sb); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); +- prev = inode->u.ext3_i.i_orphan.prev; ++ prev = ei->i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); + +- list_del(&inode->u.ext3_i.i_orphan); +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ list_del_init(&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 +@@ -1520,8 +1520,7 @@ int ext3_orphan_del(handle_t *handle, st + err = ext3_journal_dirty_metadata(handle, sbi->s_sbh); + } else { + struct ext3_iloc iloc2; +- struct inode *i_prev = +- list_entry(prev, struct inode, u.ext3_i.i_orphan); ++ struct inode *i_prev = orphan_list_entry(prev); + + jbd_debug(4, "orphan inode %lu will point to %lu\n", + i_prev->i_ino, ino_next); +@@ -1695,10 +1695,10 @@ static int ext3_symlink (struct inode * + goto out_no_entry; + } 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: +--- ./fs/ext3/super.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/super.c Tue May 7 16:05:44 2002 +@@ -121,7 +121,7 @@ static int ext3_error_behaviour(struct s + /* If no overrides were specified on the mount, then fall back + * to the default behaviour set in the filesystem's superblock + * on disk. */ +- switch (le16_to_cpu(sb->u.ext3_sb.s_es->s_errors)) { ++ switch (le16_to_cpu(EXT3_SB(sb)->s_es->s_errors)) { + case EXT3_ERRORS_PANIC: + return EXT3_ERRORS_PANIC; + case EXT3_ERRORS_RO: +@@ -269,9 +269,9 @@ void ext3_abort (struct super_block * sb + return; + + printk (KERN_CRIT "Remounting filesystem read-only\n"); +- sb->u.ext3_sb.s_mount_state |= EXT3_ERROR_FS; ++ EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; + sb->s_flags |= MS_RDONLY; +- sb->u.ext3_sb.s_mount_opt |= EXT3_MOUNT_ABORT; ++ EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT; + journal_abort(EXT3_SB(sb)->s_journal, -EIO); + } + +@@ -377,8 +377,6 @@ static int ext3_blkdev_remove(struct ext3 + return ret; + } + +-#define orphan_list_entry(l) list_entry((l), struct inode, u.ext3_i.i_orphan) +- + static void dump_orphan_list(struct super_block *sb, struct ext3_sb_info *sbi) + { + struct list_head *l; +@@ -818,7 +818,7 @@ static void ext3_orphan_cleanup (struct + sb->s_flags &= ~MS_RDONLY; + } + +- if (sb->u.ext3_sb.s_mount_state & EXT3_ERROR_FS) { ++ if (EXT3_SB(sb)->s_mount_state & EXT3_ERROR_FS) { + if (es->s_last_orphan) + jbd_debug(1, "Errors on filesystem, " + "clearing orphan list.\n"); +@@ -1463,12 +1463,14 @@ static void ext3_commit_super (struct su + struct ext3_super_block * es, + int sync) + { ++ struct buffer_head *sbh = EXT3_SB(sb)->s_sbh; ++ + es->s_wtime = cpu_to_le32(CURRENT_TIME); +- BUFFER_TRACE(sb->u.ext3_sb.s_sbh, "marking dirty"); +- mark_buffer_dirty(sb->u.ext3_sb.s_sbh); ++ BUFFER_TRACE(sbh, "marking dirty"); ++ mark_buffer_dirty(sbh); + if (sync) { +- ll_rw_block(WRITE, 1, &sb->u.ext3_sb.s_sbh); +- wait_on_buffer(sb->u.ext3_sb.s_sbh); ++ ll_rw_block(WRITE, 1, &sbh); ++ wait_on_buffer(sbh); + } + } + +@@ -1519,7 +1521,7 @@ static void ext3_clear_journal_err(struc + ext3_warning(sb, __FUNCTION__, "Marking fs in need of " + "filesystem check."); + +- sb->u.ext3_sb.s_mount_state |= EXT3_ERROR_FS; ++ EXT3_SB(sb)->s_mount_state |= EXT3_ERROR_FS; + es->s_state |= cpu_to_le16(EXT3_ERROR_FS); + ext3_commit_super (sb, es, 1); + +--- ./fs/ext3/symlink.c.orig Fri Apr 12 10:27:49 2002 ++++ ./fs/ext3/symlink.c Tue May 7 15:25:39 2002 +@@ -23,13 +23,13 @@ + + static int ext3_readlink(struct dentry *dentry, char *buffer, int buflen) + { +- char *s = (char *)dentry->d_inode->u.ext3_i.i_data; +- return vfs_readlink(dentry, buffer, buflen, s); ++ struct ext3_inode_info *ei = EXT3_I(dentry->d_inode); ++ return vfs_readlink(dentry, buffer, buflen, (char *)ei->i_data); + } + + static int ext3_follow_link(struct dentry *dentry, struct nameidata *nd) + { +- char *s = (char *)dentry->d_inode->u.ext3_i.i_data; +- return vfs_follow_link(nd, s); ++ struct ext3_inode_info *ei = EXT3_I(dentry->d_inode); ++ return vfs_follow_link(nd, (char*)ei->i_data); + } + +--- ./include/linux/ext3_fs.h.orig Tue Apr 16 14:27:25 2002 ++++ ./include/linux/ext3_fs.h Tue May 7 16:47:36 2002 +@@ -84,22 +84,25 @@ + #define EXT3_MIN_BLOCK_SIZE 1024 + #define EXT3_MAX_BLOCK_SIZE 4096 + #define EXT3_MIN_BLOCK_LOG_SIZE 10 ++ + #ifdef __KERNEL__ +-# define EXT3_BLOCK_SIZE(s) ((s)->s_blocksize) +-#else +-# define EXT3_BLOCK_SIZE(s) (EXT3_MIN_BLOCK_SIZE << (s)->s_log_block_size) +-#endif +-#define EXT3_ADDR_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (__u32)) +-#ifdef __KERNEL__ +-# define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +-#else +-# define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10) +-#endif +-#ifdef __KERNEL__ +-#define EXT3_ADDR_PER_BLOCK_BITS(s) ((s)->u.ext3_sb.s_addr_per_block_bits) +-#define EXT3_INODE_SIZE(s) ((s)->u.ext3_sb.s_inode_size) +-#define EXT3_FIRST_INO(s) ((s)->u.ext3_sb.s_first_ino) ++#define EXT3_SB(sb) (&((sb)->u.ext3_sb)) ++#define EXT3_I(inode) (&((inode)->u.ext3_i)) ++ ++#define EXT3_BLOCK_SIZE(s) ((s)->s_blocksize) ++#define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) ++#define EXT3_ADDR_PER_BLOCK_BITS(s) (EXT3_SB(s)->s_addr_per_block_bits) ++#define EXT3_INODE_SIZE(s) (EXT3_SB(s)->s_inode_size) ++#define EXT3_FIRST_INO(s) (EXT3_SB(s)->s_first_ino) + #else ++ ++/* Assume that user mode programs are passing in an ext3fs superblock, not ++ * a kernel struct super_block. This will allow us to call the feature-test ++ * macros from user land. */ ++#define EXT3_SB(sb) (sb) ++ ++#define EXT3_BLOCK_SIZE(s) (EXT3_MIN_BLOCK_SIZE << (s)->s_log_block_size) ++#define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10) + #define EXT3_INODE_SIZE(s) (((s)->s_rev_level == EXT3_GOOD_OLD_REV) ? \ + EXT3_GOOD_OLD_INODE_SIZE : \ + (s)->s_inode_size) +@@ -108,6 +110,7 @@ + EXT3_GOOD_OLD_FIRST_INO : \ + (s)->s_first_ino) + #endif ++#define EXT3_ADDR_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (__u32)) + + /* + * Macro-instructions used to manage fragments +@@ -116,8 +120,8 @@ + #define EXT3_MAX_FRAG_SIZE 4096 + #define EXT3_MIN_FRAG_LOG_SIZE 10 + #ifdef __KERNEL__ +-# define EXT3_FRAG_SIZE(s) ((s)->u.ext3_sb.s_frag_size) +-# define EXT3_FRAGS_PER_BLOCK(s) ((s)->u.ext3_sb.s_frags_per_block) ++# define EXT3_FRAG_SIZE(s) (EXT3_SB(s)->s_frag_size) ++# define EXT3_FRAGS_PER_BLOCK(s) (EXT3_SB(s)->s_frags_per_block) + #else + # define EXT3_FRAG_SIZE(s) (EXT3_MIN_FRAG_SIZE << (s)->s_log_frag_size) + # define EXT3_FRAGS_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / EXT3_FRAG_SIZE(s)) +@@ -163,15 +167,13 @@ + /* + * Macro-instructions used to manage group descriptors + */ ++# define EXT3_BLOCKS_PER_GROUP(s) (EXT3_SB(s)->s_blocks_per_group) ++# define EXT3_INODES_PER_GROUP(s) (EXT3_SB(s)->s_inodes_per_group) + #ifdef __KERNEL__ +-# define EXT3_BLOCKS_PER_GROUP(s) ((s)->u.ext3_sb.s_blocks_per_group) +-# define EXT3_DESC_PER_BLOCK(s) ((s)->u.ext3_sb.s_desc_per_block) +-# define EXT3_INODES_PER_GROUP(s) ((s)->u.ext3_sb.s_inodes_per_group) +-# define EXT3_DESC_PER_BLOCK_BITS(s) ((s)->u.ext3_sb.s_desc_per_block_bits) ++# define EXT3_DESC_PER_BLOCK(s) (EXT3_SB(s)->s_desc_per_block) ++# define EXT3_DESC_PER_BLOCK_BITS(s) (EXT3_SB(s)->s_desc_per_block_bits) + #else +-# define EXT3_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group) + # define EXT3_DESC_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (struct ext3_group_desc)) +-# define EXT3_INODES_PER_GROUP(s) ((s)->s_inodes_per_group) + #endif + + /* +@@ -344,7 +347,7 @@ + #ifndef _LINUX_EXT2_FS_H + #define clear_opt(o, opt) o &= ~EXT3_MOUNT_##opt + #define set_opt(o, opt) o |= EXT3_MOUNT_##opt +-#define test_opt(sb, opt) ((sb)->u.ext3_sb.s_mount_opt & \ ++#define test_opt(sb, opt) (EXT3_SB(sb)->s_mount_opt & \ + EXT3_MOUNT_##opt) + #else + #define EXT2_MOUNT_NOLOAD EXT3_MOUNT_NOLOAD +@@ -441,17 +443,11 @@ + /*EC*/ __u32 s_reserved[197]; /* Padding to the end of the block */ + }; + +-#ifdef __KERNEL__ +-#define EXT3_SB(sb) (&((sb)->u.ext3_sb)) +-#define EXT3_I(inode) (&((inode)->u.ext3_i)) +-#else +-/* Assume that user mode programs are passing in an ext3fs superblock, not +- * a kernel struct super_block. This will allow us to call the feature-test +- * macros from user land. */ +-#define EXT3_SB(sb) (sb) +-#endif +- +-#define NEXT_ORPHAN(inode) (inode)->u.ext3_i.i_dtime ++#define NEXT_ORPHAN(inode) EXT3_I(inode)->i_dtime ++static inline struct inode *orphan_list_entry(struct list_head *l) ++{ ++ return list_entry(l, struct inode, u.ext3_i.i_orphan); ++} + + /* + * Codes for operating systems +--- ./include/linux/ext3_jbd.h.orig Tue May 7 14:44:08 2002 ++++ ./include/linux/ext3_jbd.h Tue May 7 14:44:43 2002 +@@ -291,7 +291,7 @@ + return 1; + if (test_opt(inode->i_sb, DATA_FLAGS) == EXT3_MOUNT_JOURNAL_DATA) + return 1; +- if (inode->u.ext3_i.i_flags & EXT3_JOURNAL_DATA_FL) ++ if (EXT3_I(inode)->i_flags & EXT3_JOURNAL_DATA_FL) + return 1; + return 0; + } diff --git a/lustre/kernel_patches/patches/ext3-2.5-noread.patch b/lustre/kernel_patches/patches/ext3-2.5-noread.patch new file mode 100644 index 0000000..f1c611f --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-2.5-noread.patch @@ -0,0 +1,266 @@ +===== fs/ext3/ialloc.c 1.26 vs edited ===== +--- 1.26/fs/ext3/ialloc.c Fri Feb 14 19:24:09 2003 ++++ edited/fs/ext3/ialloc.c Sat Mar 8 01:20:55 2003 +@@ -195,6 +195,36 @@ + } + + /* ++ * @block_group: block group of inode ++ * @offset: relative offset of inode within @block_group ++ * ++ * Check whether any of the inodes in this disk block are in use. ++ * ++ * Caller must be holding superblock lock (group/bitmap read lock in ++ * future). ++ */ ++int ext3_itable_block_used(struct super_block *sb, unsigned int block_group, ++ int offset) ++{ ++ struct buffer_head *ibitmap = read_inode_bitmap(sb, block_group); ++ int inodes_per_block; ++ unsigned long inum, iend; ++ ++ if (!ibitmap) ++ return 1; ++ ++ inodes_per_block = sb->s_blocksize / EXT3_SB(sb)->s_inode_size; ++ inum = offset & ~(inodes_per_block - 1); ++ iend = inum + inodes_per_block; ++ for (; inum < iend; inum++) { ++ if (inum != offset && ext3_test_bit(inum, ibitmap->b_data)) ++ return 1; ++ } ++ ++ return 0; ++} ++ ++/* + * There are two policies for allocating an inode. If the new inode is + * a directory, then a forward search is made for a block group with both + * free space and a low directory-to-inode ratio; if that fails, then of +@@ -422,8 +452,9 @@ + struct ext3_group_desc * gdp; + struct ext3_super_block * es; + struct ext3_inode_info *ei; +- int err = 0; ++ struct ext3_iloc iloc; + struct inode *ret; ++ int err = 0; + + /* Cannot create files in a deleted directory */ + if (!dir || !dir->i_nlink) +@@ -587,16 +618,23 @@ + goto fail2; + } + err = ext3_init_acl(handle, inode, dir); ++ if (err) ++ goto fail3; ++ ++ err = ext3_get_inode_loc_new(inode, &iloc, 1); ++ if (err) ++ goto fail3; ++ ++ BUFFER_TRACE(iloc->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, iloc.bh); + if (err) { +- DQUOT_FREE_INODE(inode); +- goto fail2; +- } +- err = ext3_mark_inode_dirty(handle, inode); +- if (err) { +- ext3_std_error(sb, err); +- DQUOT_FREE_INODE(inode); +- goto fail2; +- } ++ brelse(iloc.bh); ++ iloc.bh = NULL; ++ goto fail3; ++ } ++ err = ext3_mark_iloc_dirty(handle, inode, &iloc); ++ if (err) ++ goto fail3; + + ext3_debug("allocating inode %lu\n", inode->i_ino); + goto really_out; +@@ -610,6 +648,9 @@ + brelse(bitmap_bh); + return ret; + ++fail3: ++ ext3_std_error(sb, err); ++ DQUOT_FREE_INODE(inode); + fail2: + inode->i_flags |= S_NOQUOTA; + inode->i_nlink = 0; +===== fs/ext3/inode.c 1.62 vs edited ===== +--- 1.62/fs/ext3/inode.c Fri Feb 14 19:24:09 2003 ++++ edited/fs/ext3/inode.c Sat Mar 8 02:10:39 2003 +@@ -2144,69 +2144,118 @@ + unlock_kernel(); + } + +-/* +- * ext3_get_inode_loc returns with an extra refcount against the +- * inode's underlying buffer_head on success. +- */ ++#define NUM_INODE_PREREAD 16 + +-int ext3_get_inode_loc (struct inode *inode, struct ext3_iloc *iloc) ++/* ++ * ext3_get_inode_loc returns with an extra refcount against the inode's ++ * underlying buffer_head on success. If this is for a new inode allocation ++ * (new is non-zero) then we may be able to optimize away the read if there ++ * are no other in-use inodes in this inode table block. If we need to do ++ * a read, then read in a whole chunk of blocks to avoid blocking again soon ++ * if we are doing lots of creates/updates. ++ */ ++int ext3_get_inode_loc_new(struct inode *inode, struct ext3_iloc *iloc, int new) + { +- struct buffer_head *bh = 0; ++ struct buffer_head *bh[NUM_INODE_PREREAD]; ++ struct super_block *sb = inode->i_sb; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ unsigned long ino = inode->i_ino; + unsigned long block; + unsigned long block_group; + unsigned long group_desc; + unsigned long desc; + unsigned long offset; + struct ext3_group_desc * gdp; +- +- if ((inode->i_ino != EXT3_ROOT_INO && +- inode->i_ino != EXT3_JOURNAL_INO && +- inode->i_ino < EXT3_FIRST_INO(inode->i_sb)) || +- inode->i_ino > le32_to_cpu( +- EXT3_SB(inode->i_sb)->s_es->s_inodes_count)) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "bad inode number: %lu", inode->i_ino); ++ ++ if ((ino != EXT3_ROOT_INO && ino != EXT3_JOURNAL_INO && ++ ino < EXT3_FIRST_INO(sb)) || ++ ino > le32_to_cpu(sbi->s_es->s_inodes_count)) { ++ ext3_error(sb, "ext3_get_inode_loc", "bad inode number: %lu", ++ ino); + goto bad_inode; + } +- block_group = (inode->i_ino - 1) / EXT3_INODES_PER_GROUP(inode->i_sb); +- if (block_group >= EXT3_SB(inode->i_sb)->s_groups_count) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "group >= groups count"); ++ block_group = (ino - 1) / EXT3_INODES_PER_GROUP(sb); ++ if (block_group >= EXT3_SB(sb)->s_groups_count) { ++ ext3_error(sb, "ext3_get_inode_loc", "group >= groups count"); + goto bad_inode; + } +- group_desc = block_group >> EXT3_DESC_PER_BLOCK_BITS(inode->i_sb); +- desc = block_group & (EXT3_DESC_PER_BLOCK(inode->i_sb) - 1); +- bh = EXT3_SB(inode->i_sb)->s_group_desc[group_desc]; +- if (!bh) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "Descriptor not loaded"); ++ group_desc = block_group >> EXT3_DESC_PER_BLOCK_BITS(sb); ++ desc = block_group & (EXT3_DESC_PER_BLOCK(sb) - 1); ++ if (!sbi->s_group_desc[group_desc]) { ++ ext3_error(sb, "ext3_get_inode_loc", "Descriptor not loaded"); + goto bad_inode; + } + +- gdp = (struct ext3_group_desc *) bh->b_data; ++ gdp = (struct ext3_group_desc *)(sbi->s_group_desc[group_desc]->b_data); + /* + * Figure out the offset within the block group inode table + */ +- offset = ((inode->i_ino - 1) % EXT3_INODES_PER_GROUP(inode->i_sb)) * +- EXT3_INODE_SIZE(inode->i_sb); ++ offset = ((ino - 1) % EXT3_INODES_PER_GROUP(sb)); + block = le32_to_cpu(gdp[desc].bg_inode_table) + +- (offset >> EXT3_BLOCK_SIZE_BITS(inode->i_sb)); +- if (!(bh = sb_bread(inode->i_sb, block))) { +- ext3_error (inode->i_sb, "ext3_get_inode_loc", +- "unable to read inode block - " +- "inode=%lu, block=%lu", inode->i_ino, block); +- goto bad_inode; ++ (offset * sbi->s_inode_size >> EXT3_BLOCK_SIZE_BITS(sb)); ++ bh[0] = sb_getblk(sb, block); ++ if (buffer_uptodate(bh[0])) ++ goto done; ++ ++ /* If we don't really need to read this block, and it isn't already ++ * in memory, then we just zero it out. Otherwise, we keep the ++ * current block contents (deleted inode data) for posterity. ++ */ ++ if (new && !ext3_itable_block_used(sb, block_group, offset)) { ++ lock_buffer(bh[0]); ++ memset(bh[0]->b_data, 0, bh[0]->b_size); ++ set_buffer_uptodate(bh[0]); ++ unlock_buffer(bh[0]); ++ } else { ++ unsigned long block_end, itable_end; ++ int count = 1; ++ ++ itable_end = le32_to_cpu(gdp[desc].bg_inode_table) + ++ sbi->s_itb_per_group; ++ block_end = block + NUM_INODE_PREREAD; ++ if (block_end > itable_end) ++ block_end = itable_end; ++ ++ for (; block < block_end; block++) { ++ bh[count] = sb_getblk(sb, block); ++ if (count && (buffer_uptodate(bh[count]) || ++ buffer_locked(bh[count]))) { ++ __brelse(bh[count]); ++ } else ++ count++; ++ } ++ ++ ll_rw_block(READ, count, bh); ++ ++ /* Release all but the block we actually need (bh[0]) */ ++ while (--count > 0) ++ __brelse(bh[count]); ++ ++ wait_on_buffer(bh[0]); ++ if (!buffer_uptodate(bh[0])) { ++ ext3_error(sb, __FUNCTION__, ++ "unable to read inode block - " ++ "inode=%lu, block=%llu", ino, ++ (unsigned long long)bh[0]->b_blocknr); ++ goto bad_inode; ++ } + } +- offset &= (EXT3_BLOCK_SIZE(inode->i_sb) - 1); ++done: ++ offset = (offset * sbi->s_inode_size) & (EXT3_BLOCK_SIZE(sb) - 1); + +- iloc->bh = bh; +- iloc->raw_inode = (struct ext3_inode *) (bh->b_data + offset); ++ iloc->bh = bh[0]; ++ iloc->raw_inode = (struct ext3_inode *)(bh[0]->b_data + offset); + iloc->block_group = block_group; +- ++ + return 0; +- ++ + bad_inode: + return -EIO; ++} ++ ++int ext3_get_inode_loc(struct inode *inode, struct ext3_iloc *iloc) ++{ ++ return ext3_get_inode_loc_new(inode, iloc, 0); + } + + void ext3_read_inode(struct inode * inode) +===== include/linux/ext3_fs.h 1.22 vs edited ===== +--- 1.22/include/linux/ext3_fs.h Tue Jan 14 00:56:29 2003 ++++ edited/include/linux/ext3_fs.h Sat Mar 8 01:56:28 2003 +@@ -719,6 +719,8 @@ + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + ++extern int ext3_itable_block_used(struct super_block *, unsigned int, int); ++extern int ext3_get_inode_loc_new(struct inode *, struct ext3_iloc *, int); + extern int ext3_get_inode_loc (struct inode *, struct ext3_iloc *); + extern void ext3_read_inode (struct inode *); + extern void ext3_write_inode (struct inode *, int); diff --git a/lustre/kernel_patches/patches/ext3-largefile.patch b/lustre/kernel_patches/patches/ext3-largefile.patch new file mode 100644 index 0000000..db41aab --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-largefile.patch @@ -0,0 +1,23 @@ +Under rare conditions (filesystem corruption, really) it is possible +for ext3_dirty_inode() to require _two_ blocks for the transaction: one +for the inode and one to update the superblock - to set +EXT3_FEATURE_RO_COMPAT_LARGE_FILE. This causes the filesystem to go +BUG. + +So reserve an additional block for that eventuality. + + + fs/ext3/inode.c | 2 +- + 1 files changed, 1 insertion(+), 1 deletion(-) + +--- 25/fs/ext3/inode.c~ext3-transaction-reserved-blocks Sat Dec 14 18:28:21 2002 ++++ 25-akpm/fs/ext3/inode.c Sat Dec 14 18:28:21 2002 +@@ -2698,7 +2698,7 @@ void ext3_dirty_inode(struct inode *inod + handle_t *handle; + + lock_kernel(); +- handle = ext3_journal_start(inode, 1); ++ handle = ext3_journal_start(inode, 2); + if (IS_ERR(handle)) + goto out; + if (current_handle && diff --git a/lustre/kernel_patches/patches/ext3-orphan_lock.patch b/lustre/kernel_patches/patches/ext3-orphan_lock.patch new file mode 100644 index 0000000..d1e5c8d --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-orphan_lock.patch @@ -0,0 +1,79 @@ +--- linux/fs/ext3/namei.c.orig Fri Mar 14 14:11:58 2003 ++++ linux/fs/ext3/namei.c Fri Mar 14 14:39:48 2003 +@@ -1406,8 +1409,8 @@ + 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; + +@@ -1455,7 +1458,7 @@ + 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; + } +@@ -1468,20 +1471,19 @@ + { + 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); + +@@ -1525,10 +1527,10 @@ + 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: +--- linux/fs/ext3/super.c.orig Fri Mar 14 14:11:58 2003 ++++ linux/fs/ext3/super.c Fri Mar 14 14:36:00 2003 +@@ -1134,6 +1314,7 @@ + */ + 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; + +--- linux/include/linux/ext3_fs_sb.h.orig Tue Feb 11 16:34:33 2003 ++++ linux/include/linux/ext3_fs_sb.h Fri Mar 14 14:30:11 2003 +@@ -67,6 +69,7 @@ + struct inode * s_journal_inode; + struct journal_s * s_journal; + struct list_head s_orphan; ++ struct semaphore s_orphan_lock; + unsigned long s_commit_interval; + struct block_device *journal_bdev; + #ifdef CONFIG_JBD_DEBUG diff --git a/lustre/kernel_patches/patches/ext3-truncate_blocks.patch b/lustre/kernel_patches/patches/ext3-truncate_blocks.patch new file mode 100644 index 0000000..ce3928d --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-truncate_blocks.patch @@ -0,0 +1,92 @@ +--- ./fs/ext3/inode.c.orig Wed Mar 12 02:44:06 2003 ++++ ./fs/ext3/inode.c Wed Mar 12 11:55:20 2003 +@@ -99,7 +99,35 @@ int ext3_forget(handle_t *handle, int is + return err; + } + +-/* ++/* ++ * Work out how many blocks we need to progress with the next chunk of a ++ * truncate transaction. ++ */ ++ ++static unsigned long blocks_for_truncate(struct inode *inode) ++{ ++ unsigned long needed; ++ ++ needed = inode->i_blocks >> (inode->i_sb->s_blocksize_bits - 9); ++ ++ /* Give ourselves just enough room to cope with inodes in which ++ * i_blocks is corrupt: we've seen disk corruptions in the past ++ * which resulted in random data in an inode which looked enough ++ * like a regular file for ext3 to try to delete it. Things ++ * will go a bit crazy if that happens, but at least we should ++ * try not to panic the whole kernel. */ ++ if (needed < 2) ++ needed = 2; ++ ++ /* But we need to bound the transaction so we don't overflow the ++ * journal. */ ++ if (needed > EXT3_MAX_TRANS_DATA) ++ needed = EXT3_MAX_TRANS_DATA; ++ ++ return EXT3_DATA_TRANS_BLOCKS + needed; ++} ++ ++/* + * Truncate transactions can be complex and absolutely huge. So we need to + * be able to restart the transaction at a conventient checkpoint to make + * sure we don't overflow the journal. +@@ -110,19 +138,14 @@ int ext3_forget(handle_t *handle, int is + * transaction in the top-level truncate loop. --sct + */ + +-static handle_t *start_transaction(struct inode *inode) ++static handle_t *start_transaction(struct inode *inode) + { +- long needed; + handle_t *result; +- +- needed = inode->i_blocks; +- if (needed > EXT3_MAX_TRANS_DATA) +- needed = EXT3_MAX_TRANS_DATA; +- +- result = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS + needed); ++ ++ result = ext3_journal_start(inode, blocks_for_truncate(inode)); + if (!IS_ERR(result)) + return result; +- ++ + ext3_std_error(inode->i_sb, PTR_ERR(result)); + return result; + } +@@ -135,14 +158,9 @@ static handle_t *start_transaction(struc + */ + static int try_to_extend_transaction(handle_t *handle, struct inode *inode) + { +- long needed; +- + if (handle->h_buffer_credits > EXT3_RESERVE_TRANS_BLOCKS) + return 0; +- needed = inode->i_blocks; +- if (needed > EXT3_MAX_TRANS_DATA) +- needed = EXT3_MAX_TRANS_DATA; +- if (!ext3_journal_extend(handle, EXT3_RESERVE_TRANS_BLOCKS + needed)) ++ if (!ext3_journal_extend(handle, blocks_for_truncate(inode))) + return 0; + return 1; + } +@@ -154,11 +172,8 @@ static int try_to_extend_transaction(han + */ + static int ext3_journal_test_restart(handle_t *handle, struct inode *inode) + { +- long needed = inode->i_blocks; +- if (needed > EXT3_MAX_TRANS_DATA) +- needed = EXT3_MAX_TRANS_DATA; + jbd_debug(2, "restarting handle %p\n", handle); +- return ext3_journal_restart(handle, EXT3_DATA_TRANS_BLOCKS + needed); ++ return ext3_journal_restart(handle, blocks_for_truncate(inode)); + } + + /* diff --git a/lustre/kernel_patches/patches/ext3-unmount_sync.patch b/lustre/kernel_patches/patches/ext3-unmount_sync.patch new file mode 100644 index 0000000..1f9b796 --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-unmount_sync.patch @@ -0,0 +1,59 @@ +From adilger@clusterfs.com Mon Dec 2 10:26:44 2002 +Date: Mon, 2 Dec 2002 10:26:44 -0700 +From: Andreas Dilger +To: Lustre LLNL Mailing list , + Lustre Development Mailing List +Subject: Re: data corrupting bug in 2.4.20 ext3, data=journal +Message-ID: <20021202102644.H1422@schatzie.adilger.int> +Mail-Followup-To: Lustre LLNL Mailing list , + Lustre Development Mailing List +Mime-Version: 1.0 +Content-Type: text/plain; charset=us-ascii +Content-Disposition: inline +User-Agent: Mutt/1.2.5.1i +X-GPG-Key: 1024D/0D35BED6 +X-GPG-Fingerprint: 7A37 5D79 BF1B CECA D44F 8A29 A488 39F5 0D35 BED6 +Status: RO +Content-Length: 1160 +Lines: 39 + +Here is the new-improved fix for the ext3 discarding data at umount bug +discovered late last week. To be used instead of the previous ext3 fix. + +Sadly, this is completely unrelated to the problems Mike is having with +ext3 under UML, since it is an unmount-time problem. + +----- Forwarded message from "Stephen C. Tweedie" ----- +The attached patch seems to fix things for me. + +Cheers, + Stephen + + +--- linux-2.4-ext3merge/fs/ext3/super.c.=K0027=.orig 2002-12-02 15:35:13.000000000 +0000 ++++ linux-2.4-ext3merge/fs/ext3/super.c 2002-12-02 15:35:14.000000000 +0000 +@@ -1640,7 +1640,12 @@ + sb->s_dirt = 0; + target = log_start_commit(EXT3_SB(sb)->s_journal, NULL); + +- if (do_sync_supers) { ++ /* ++ * Tricky --- if we are unmounting, the write really does need ++ * to be synchronous. We can detect that by looking for NULL in ++ * sb->s_root. ++ */ ++ if (do_sync_supers || !sb->s_root) { + unlock_super(sb); + log_wait_commit(EXT3_SB(sb)->s_journal, target); + lock_super(sb); + + +----- End forwarded message ----- + +Cheers, Andreas +-- +Andreas Dilger +http://sourceforge.net/projects/ext2resize/ +http://www-mddsp.enel.ucalgary.ca/People/adilger/ + + diff --git a/lustre/kernel_patches/patches/ext3-use-after-free.patch b/lustre/kernel_patches/patches/ext3-use-after-free.patch new file mode 100644 index 0000000..8cd673f --- /dev/null +++ b/lustre/kernel_patches/patches/ext3-use-after-free.patch @@ -0,0 +1,65 @@ + + +If ext3_add_nondir() fails it will do an iput() of the inode. But we +continue to run ext3_mark_inode_dirty() against the potentially-freed +inode. This oopses when slab poisoning is enabled. + +Fix it so that we only run ext3_mark_inode_dirty() if the inode was +successfully instantiated. + +This bug was added in 2.4.20-pre9. + + + fs/ext3/namei.c | 11 +++++------ + 1 files changed, 5 insertions(+), 6 deletions(-) + +--- 24/fs/ext3/namei.c~ext3-use-after-free Sun Dec 15 11:27:50 2002 ++++ 24-akpm/fs/ext3/namei.c Sun Dec 15 11:27:50 2002 +@@ -429,8 +429,11 @@ static int ext3_add_nondir(handle_t *han + { + 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); +@@ -465,7 +468,6 @@ static int ext3_create (struct inode * d + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; + err = ext3_add_nondir(handle, dentry, inode); +- ext3_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -490,7 +492,6 @@ static int ext3_mknod (struct inode * di + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); + err = ext3_add_nondir(handle, dentry, inode); +- ext3_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -934,7 +935,6 @@ static int ext3_symlink (struct inode * + } + inode->u.ext3_i.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; +@@ -971,7 +971,6 @@ static int ext3_link (struct dentry * ol + atomic_inc(&inode->i_count); + + err = ext3_add_nondir(handle, dentry, inode); +- ext3_mark_inode_dirty(handle, inode); + ext3_journal_stop(handle, dir); + return err; + } + +_ diff --git a/lustre/kernel_patches/patches/extN-2.4.18-exports.patch b/lustre/kernel_patches/patches/extN-2.4.18-exports.patch new file mode 100644 index 0000000..172d7d2 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-2.4.18-exports.patch @@ -0,0 +1,11 @@ +--- linux-2.4.17/fs/ext3/super.c.orig Fri Dec 21 10:41:55 2001 ++++ linux-2.4.17/fs/ext3/super.c Fri Mar 22 11:00:41 2002 +@@ -1742,7 +1742,7 @@ + unregister_filesystem(&ext3_fs_type); + } + +-EXPORT_NO_SYMBOLS; ++EXPORT_SYMBOL(ext3_bread); + + MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); + MODULE_DESCRIPTION("Second Extended Filesystem with journaling extensions"); diff --git a/lustre/kernel_patches/patches/extN-2.4.18-ino_sb_fixup.patch b/lustre/kernel_patches/patches/extN-2.4.18-ino_sb_fixup.patch new file mode 100644 index 0000000..df46643 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-2.4.18-ino_sb_fixup.patch @@ -0,0 +1,33 @@ +--- ./include/linux/ext3_fs.h.orig Tue May 7 17:06:03 2002 ++++ ./include/linux/ext3_fs.h Tue May 7 17:07:11 2002 +@@ -17,6 +17,8 @@ + #define _LINUX_EXT3_FS_H + + #include ++#include ++#include + + /* + * The second extended filesystem constants/structures +@@ -86,8 +88,8 @@ + #define EXT3_MIN_BLOCK_LOG_SIZE 10 + + #ifdef __KERNEL__ +-#define EXT3_SB(sb) (&((sb)->u.ext3_sb)) +-#define EXT3_I(inode) (&((inode)->u.ext3_i)) ++#define EXT3_SB(sb) ((struct ext3_sb_info *)&((sb)->u.generic_sbp)) ++#define EXT3_I(inode) ((struct ext3_inode_info *)&((inode)->u.generic_ip)) + + #define EXT3_BLOCK_SIZE(s) ((s)->s_blocksize) + #define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -447,7 +447,9 @@ + #define NEXT_ORPHAN(inode) EXT3_I(inode)->i_dtime + static inline struct inode *orphan_list_entry(struct list_head *l) + { +- return list_entry(l, struct inode, u.ext3_i.i_orphan); ++ return ((struct inode *)((char *)l - ++ (unsigned long)(offsetof(struct inode, u.generic_ip) + ++ offsetof(struct ext3_inode_info, i_orphan)))); + } + + /* diff --git a/lustre/kernel_patches/patches/extN-delete_thread.patch b/lustre/kernel_patches/patches/extN-delete_thread.patch new file mode 100644 index 0000000..33f43d6 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-delete_thread.patch @@ -0,0 +1,274 @@ +--- linux/include/linux/ext3_fs.h.orig Fri Mar 14 18:09:02 2003 ++++ linux/include/linux/ext3_fs.h Fri Mar 14 18:10:20 2003 +@@ -190,7 +192,8 @@ + */ + #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 +--- linux/include/linux/ext3_fs_sb.h.orig Tue Feb 11 16:34:33 2003 ++++ linux/include/linux/ext3_fs_sb.h Mon Mar 10 14:42:07 2003 +@@ -29,6 +29,8 @@ + + #define EXT3_MAX_GROUP_LOADED 32 + ++#define EXT3_DELETE_THREAD ++ + /* + * third extended-fs super-block data in memory + */ +@@ -73,7 +75,15 @@ + 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 */ +--- linux/fs/ext3/super.c.orig Wed Mar 12 14:05:30 2003 ++++ linux/fs/ext3/super.c Thu Mar 13 19:05:26 2003 +@@ -396,6 +396,207 @@ + } + } + ++#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); ++ printk(KERN_INFO "EXT3-fs: delete thread on %s started\n", ++ kdevname(sb->s_dev)); ++ ++ /* main loop */ ++ for (;;) { ++ sleep_on(&sbi->s_delete_thread_queue); ++ printk(KERN_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)) { ++ memset(&sbi->s_delete_list, 0, ++ sizeof(sbi->s_delete_list)); ++ spin_unlock(&sbi->s_delete_lock); ++ printk(KERN_DEBUG "ext3 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); ++ printk(KERN_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) ++ printk(KERN_WARNING ++ "%lu blocks and %lu left 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); ++ memset(&sbi->s_delete_list, 0, sizeof(sbi->s_delete_list)); ++ init_waitqueue_head(&sbi->s_delete_thread_queue); ++ init_waitqueue_head(&sbi->s_delete_waiter_queue); ++ sbi->s_delete_blocks = 0; ++ sbi->s_delete_inodes = 0; ++ 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) ++{ ++ 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 ++ * duplicate the inode data 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 copy the whole thing. ++ * ++ * 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 ++ * truncate thread when we run out of space. ++ * ++ * One shouldn't consider this duplicate an "inode", as it isn't really ++ * visible to the VFS, but rather a data struct that holds truncate data. ++ * ++ * In 2.5 this can be done much more cleanly by just registering a "drop" ++ * method in the super_operations struct. ++ */ ++static void ext3_delete_inode_thread(struct inode *old_inode) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(old_inode->i_sb); ++ 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; ++ } ++ ++ /* We may want to delete the inode immediately and not defer it */ ++ if (IS_SYNC(old_inode) || blocks <= EXT3_NDIR_BLOCKS || ++ !sbi->s_delete_list.next) { ++ ext3_delete_inode(old_inode); ++ return; ++ } ++ ++ if (EXT3_I(old_inode)->i_state & EXT3_STATE_DELETE) { ++ ext3_debug("doing deferred inode %lu delete (%lu blocks)\n", ++ old_inode->i_ino, blocks); ++ ext3_delete_inode(old_inode); ++ return; ++ } ++ ++ /* 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); ++ ++ EXT3_SB(old_inode->i_sb)->s_mount_state |= EXT3_ORPHAN_FS; ++ new_inode = iget(old_inode->i_sb, old_inode->i_ino); ++ EXT3_SB(old_inode->i_sb)->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(KERN_DEBUG "delete inode %lu directly (bad read)\n", ++ old_inode->i_ino); ++ ext3_delete_inode(old_inode); ++ return; ++ } ++ J_ASSERT(new_inode != old_inode); ++ ++ J_ASSERT(!list_empty(&EXT3_I(old_inode)->i_orphan)); ++ /* 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 ++ * the same as the on-disk orphan list. ++ */ ++ EXT3_I(new_inode)->i_orphan = EXT3_I(old_inode)->i_orphan; ++ EXT3_I(new_inode)->i_orphan.next->prev = &EXT3_I(new_inode)->i_orphan; ++ EXT3_I(new_inode)->i_orphan.prev->next = &EXT3_I(new_inode)->i_orphan; ++ EXT3_I(new_inode)->i_state |= EXT3_STATE_DELETE; ++ up(&sbi->s_orphan_lock); ++ ++ clear_inode(old_inode); ++ ++ printk(KERN_DEBUG "delete inode %lu (%lu blocks) by thread\n", ++ new_inode->i_ino, blocks); ++ 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); ++ ++ wake_up(&sbi->s_delete_thread_queue); ++} ++#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); +@@ -403,6 +578,7 @@ + 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)) { +@@ -451,7 +627,11 @@ + 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 */ + write_super_lockfs: ext3_write_super_lockfs, /* BKL not held. Take it */ +@@ -1205,6 +1385,7 @@ + } + + 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 diff --git a/lustre/kernel_patches/patches/extN-iget-debug.patch b/lustre/kernel_patches/patches/extN-iget-debug.patch new file mode 100644 index 0000000..dbe90c8 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-iget-debug.patch @@ -0,0 +1,48 @@ +--- linux/fs/ext3/namei.c.orig Thu Jan 30 01:15:13 2003 ++++ linux/fs/ext3/namei.c Sat Feb 1 00:33:46 2003 +@@ -710,6 +710,24 @@ + return ret; + } + ++static int ext3_find_inode(struct inode *inode, unsigned long ino, ++ void *opaque) ++{ ++ const char *name = NULL; ++ int len = 0; ++ ++ if (opaque) { ++ struct dentry *dentry = opaque; ++ name = dentry->d_name.name; ++ len = dentry->d_name.len; ++ } ++ printk(KERN_INFO "finding inode %s:%lu (%p) count %d (%p = %*s)\n", ++ kdevname(inode->i_dev), ino, inode, atomic_read(&inode->i_count), ++ opaque, len, name ? name : ""); ++ ++ return 1; ++} ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -724,8 +742,8 @@ + if (bh) { + unsigned long ino = le32_to_cpu(de->inode); + brelse (bh); +- inode = iget(dir->i_sb, ino); ++ inode = iget4(dir->i_sb, ino, ext3_find_inode, dentry); + + if (!inode) + return ERR_PTR(-EACCES); +--- linux/fs/ext3/inode.c.orig Thu Jan 30 01:15:13 2003 ++++ linux/fs/ext3/inode.c Sat Feb 1 00:34:45 2003 +@@ -166,6 +166,9 @@ + */ + void ext3_put_inode (struct inode * inode) + { ++ printk(KERN_INFO "putting inode %s:%lu (%p) count %d\n", ++ kdevname(inode->i_dev), inode->i_ino, inode, ++ atomic_read(&inode->i_count)); + ext3_discard_prealloc (inode); + } + diff --git a/lustre/kernel_patches/patches/extN-misc-fixup.patch b/lustre/kernel_patches/patches/extN-misc-fixup.patch new file mode 100644 index 0000000..06ea72a --- /dev/null +++ b/lustre/kernel_patches/patches/extN-misc-fixup.patch @@ -0,0 +1,23 @@ +--- linux-2.4.17/fs/ext3/super.c.orig Fri Dec 21 10:41:55 2001 ++++ linux-2.4.17/fs/ext3/super.c Fri Mar 22 11:00:41 2002 +@@ -1344,10 +1342,10 @@ + printk(KERN_ERR "EXT3-fs: I/O error on journal device\n"); + goto out_journal; + } +- if (ntohl(journal->j_superblock->s_nr_users) != 1) { ++ if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) { + printk(KERN_ERR "EXT3-fs: External journal has more than one " + "user (unsupported) - %d\n", +- ntohl(journal->j_superblock->s_nr_users)); ++ be32_to_cpu(journal->j_superblock->s_nr_users)); + goto out_journal; + } + EXT3_SB(sb)->journal_bdev = bdev; +@@ -1560,6 +1560,7 @@ + unlock_kernel(); + return ret; + } ++EXPORT_SYMBOL(ext3_force_commit); /* here to avoid potential patch collisions */ + + /* + * Ext3 always journals updates to the superblock itself, so we don't diff --git a/lustre/kernel_patches/patches/extN-noread.patch b/lustre/kernel_patches/patches/extN-noread.patch new file mode 100644 index 0000000..b104177 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-noread.patch @@ -0,0 +1,233 @@ +diff -ru lustre-head/fs/ext3/ialloc.c lustre/fs/ext3/ialloc.c +--- lustre-head/fs/ext3/ialloc.c Mon Dec 23 10:02:58 2002 ++++ lustre/fs/ext3/ialloc.c Mon Dec 23 09:46:20 2002 +@@ -289,6 +289,37 @@ + } + + /* ++ * @block_group: block group of inode ++ * @offset: relative offset of inode within @block_group ++ * ++ * Check whether any of the inodes in this disk block are in use. ++ * ++ * Caller must be holding superblock lock (group/bitmap read lock in future). ++ */ ++int ext3_itable_block_used(struct super_block *sb, unsigned int block_group, ++ int offset) ++{ ++ int bitmap_nr = load_inode_bitmap(sb, block_group); ++ int inodes_per_block; ++ unsigned long inum, iend; ++ struct buffer_head *ibitmap; ++ ++ if (bitmap_nr < 0) ++ return 1; ++ ++ inodes_per_block = sb->s_blocksize / EXT3_SB(sb)->s_inode_size; ++ inum = offset & ~(inodes_per_block - 1); ++ iend = inum + inodes_per_block; ++ ibitmap = EXT3_SB(sb)->s_inode_bitmap[bitmap_nr]; ++ for (; inum < iend; inum++) { ++ if (inum != offset && ext3_test_bit(inum, ibitmap->b_data)) ++ return 1; ++ } ++ ++ return 0; ++} ++ ++/* + * There are two policies for allocating an inode. If the new inode is + * a directory, then a forward search is made for a block group with both + * free space and a low directory-to-inode ratio; if that fails, then of +@@ -312,6 +343,7 @@ + struct ext3_group_desc * gdp; + struct ext3_group_desc * tmp; + struct ext3_super_block * es; ++ struct ext3_iloc iloc; + int err = 0; + + /* Cannot create files in a deleted directory */ +@@ -505,7 +538,7 @@ + ei->i_prealloc_count = 0; + #endif + ei->i_block_group = i; +- ++ + if (ei->i_flags & EXT3_SYNC_FL) + inode->i_flags |= S_SYNC; + if (IS_SYNC(inode)) +@@ -514,9 +547,18 @@ + inode->i_generation = sbi->s_next_generation++; + + ei->i_state = EXT3_STATE_NEW; +- err = ext3_mark_inode_dirty(handle, inode); ++ err = ext3_get_inode_loc_new(inode, &iloc, 1); + if (err) goto fail; +- ++ BUFFER_TRACE(iloc->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, iloc.bh); ++ if (err) { ++ brelse(iloc.bh); ++ iloc.bh = NULL; ++ goto fail; ++ } ++ err = ext3_mark_iloc_dirty(handle, inode, &iloc); ++ if (err) goto fail; ++ + unlock_super (sb); + if(DQUOT_ALLOC_INODE(inode)) { + DQUOT_DROP(inode); +diff -ru lustre-head/fs/ext3/inode.c lustre/fs/ext3/inode.c +--- lustre-head/fs/ext3/inode.c Mon Dec 23 10:02:58 2002 ++++ lustre/fs/ext3/inode.c Mon Dec 23 09:50:25 2002 +@@ -2011,23 +1994,28 @@ + ext3_journal_stop(handle, inode); + } + +-/* +- * ext3_get_inode_loc returns with an extra refcount against the +- * inode's underlying buffer_head on success. +- */ ++#define NUM_INODE_PREREAD 16 + +-int ext3_get_inode_loc (struct inode *inode, struct ext3_iloc *iloc) ++/* ++ * ext3_get_inode_loc returns with an extra refcount against the inode's ++ * underlying buffer_head on success. If this is for a new inode allocation ++ * (new is non-zero) then we may be able to optimize away the read if there ++ * are no other in-use inodes in this inode table block. If we need to do ++ * a read, then read in a whole chunk of blocks to avoid blocking again soon ++ * if we are doing lots of creates/updates. ++ */ ++int ext3_get_inode_loc_new(struct inode *inode, struct ext3_iloc *iloc, int new) + { + struct super_block *sb = inode->i_sb; + struct ext3_sb_info *sbi = EXT3_SB(sb); +- struct buffer_head *bh = 0; ++ struct buffer_head *bh[NUM_INODE_PREREAD]; + unsigned long block; + unsigned long block_group; + unsigned long group_desc; + unsigned long desc; + unsigned long offset; + struct ext3_group_desc * gdp; +- ++ + if ((inode->i_ino != EXT3_ROOT_INO && + inode->i_ino != EXT3_JOURNAL_INO && + inode->i_ino < EXT3_FIRST_INO(sb)) || +@@ -2042,38 +2034,86 @@ + } + group_desc = block_group >> sbi->s_desc_per_block_bits; + desc = block_group & (sbi->s_desc_per_block - 1); +- bh = sbi->s_group_desc[group_desc]; +- if (!bh) { ++ if (!sbi->s_group_desc[group_desc]) { + ext3_error(sb, __FUNCTION__, "Descriptor not loaded"); + goto bad_inode; + } + +- gdp = (struct ext3_group_desc *) bh->b_data; ++ gdp = (struct ext3_group_desc *)(sbi->s_group_desc[group_desc]->b_data); ++ + /* + * Figure out the offset within the block group inode table + */ +- offset = ((inode->i_ino - 1) % sbi->s_inodes_per_group) * +- sbi->s_inode_size; ++ offset = ((inode->i_ino - 1) % sbi->s_inodes_per_group); ++ + block = le32_to_cpu(gdp[desc].bg_inode_table) + +- (offset >> EXT3_BLOCK_SIZE_BITS(sb)); +- if (!(bh = sb_bread(sb, block))) { +- ext3_error (sb, __FUNCTION__, +- "unable to read inode block - " +- "inode=%lu, block=%lu", inode->i_ino, block); +- goto bad_inode; ++ (offset * sbi->s_inode_size >> EXT3_BLOCK_SIZE_BITS(sb)); ++ ++ bh[0] = sb_getblk(sb, block); ++ if (buffer_uptodate(bh[0])) ++ goto done; ++ ++ /* If we don't really need to read this block, and it isn't already ++ * in memory, then we just zero it out. Otherwise, we keep the ++ * current block contents (deleted inode data) for posterity. ++ */ ++ if (new && !ext3_itable_block_used(sb, block_group, offset)) { ++ lock_buffer(bh[0]); ++ memset(bh[0]->b_data, 0, bh[0]->b_size); ++ mark_buffer_uptodate(bh[0], 1); ++ unlock_buffer(bh[0]); ++ } else { ++ unsigned long block_end, itable_end; ++ int count = 1; ++ ++ itable_end = le32_to_cpu(gdp[desc].bg_inode_table) + ++ sbi->s_itb_per_group; ++ block_end = block + NUM_INODE_PREREAD; ++ if (block_end > itable_end) ++ block_end = itable_end; ++ ++ for (; block < block_end; block++) { ++ bh[count] = sb_getblk(sb, block); ++ if (count && (buffer_uptodate(bh[count]) || ++ buffer_locked(bh[count]))) { ++ __brelse(bh[count]); ++ } else ++ count++; ++ } ++ ++ ll_rw_block(READ, count, bh); ++ ++ /* Release all but the block we actually need (bh[0]) */ ++ while (--count > 0) ++ __brelse(bh[count]); ++ ++ wait_on_buffer(bh[0]); ++ if (!buffer_uptodate(bh[0])) { ++ ext3_error(sb, __FUNCTION__, ++ "unable to read inode block - " ++ "inode=%lu, block=%lu", inode->i_ino, ++ bh[0]->b_blocknr); ++ goto bad_inode; ++ } + } +- offset &= (EXT3_BLOCK_SIZE(sb) - 1); ++ done: ++ offset = (offset * sbi->s_inode_size) & (EXT3_BLOCK_SIZE(sb) - 1); + +- iloc->bh = bh; +- iloc->raw_inode = (struct ext3_inode *) (bh->b_data + offset); ++ iloc->bh = bh[0]; ++ iloc->raw_inode = (struct ext3_inode *)(bh[0]->b_data + offset); + iloc->block_group = block_group; +- ++ + return 0; +- ++ + bad_inode: + return -EIO; + } + ++int ext3_get_inode_loc(struct inode *inode, struct ext3_iloc *iloc) ++{ ++ return ext3_get_inode_loc_new(inode, iloc, 0); ++} ++ + void ext3_read_inode(struct inode * inode) + { + struct ext3_iloc iloc; +diff -ru include/linux/ext3_fs.h.orig include/linux/ext3_fs.h +--- lustre/include/linux/ext3_fs.h.orig Sat Mar 8 01:23:09 2003 ++++ lustre/include/linux/ext3_fs.h Sat Mar 8 01:24:31 2003 +@@ -642,6 +646,8 @@ + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + ++extern int ext3_itable_block_used(struct super_block *sb, unsigned int, int); ++extern int ext3_get_inode_loc_new(struct inode *, struct ext3_iloc *, int); + extern int ext3_get_inode_loc (struct inode *, struct ext3_iloc *); + extern void ext3_read_inode (struct inode *); + extern void ext3_write_inode (struct inode *, int); diff --git a/lustre/kernel_patches/patches/extN-san.patch b/lustre/kernel_patches/patches/extN-san.patch new file mode 100644 index 0000000..8cf1dce --- /dev/null +++ b/lustre/kernel_patches/patches/extN-san.patch @@ -0,0 +1,88 @@ +--- linux/fs/ext3/inode.orig.c 2002-12-29 18:48:56.000000000 +0800 ++++ linux/fs/ext3/inode.c 2002-12-29 19:17:24.000000000 +0800 +@@ -2728,3 +2728,85 @@ + * here, in ext3_aops_journal_start() to ensure that the forthcoming "see if we + * need to extend" test in ext3_prepare_write() succeeds. + */ ++ ++/* for each block: 1 ind + 1 dind + 1 tind ++ * for each block: 3 bitmap blocks ++ * for each block: 3 group descriptor blocks ++ * i inode block ++ * 1 superblock ++ * 2 * EXT3_SINGLEDATA_TRANS_BLOCKS for the quote files ++ * ((1+1+1) * 3 * nblocks) + 1 + 1 + 2 * EXT3_SINGLEDATA_TRANS_BLOCKS ++ * ++ * XXX assuming: ++ * (1) fs logic block size == page size ++ * (2) ext3 in writeback mode ++ */ ++static inline int ext3_san_write_trans_blocks(int nblocks) ++{ ++ int ret; ++ ++ ret = (1 + 1 + 1) * 3 * nblocks + 1 + 1; ++ ++#ifdef CONFIG_QUOTA ++ ret += 2 * EXT3_SINGLEDATA_TRANS_BLOCKS; ++#endif ++ ++ return ret; ++} ++ ++/* Alloc blocks for an inode, while don't create any buffer/page ++ * for data I/O; set the inode size if file is extended. ++ * ++ * @inode: target inode ++ * @blocks: array of logic block number ++ * @nblocks: how many blocks need be alloced ++ * @newsize: new filesize we should set ++ * ++ * return: 0 success, otherwise failed ++ * (*blocks) contains physical block number alloced ++ * ++ * XXX this assume the fs block size == page size ++ */ ++int ext3_prep_san_write(struct inode *inode, long *blocks, ++ int nblocks, loff_t newsize) ++{ ++ handle_t *handle; ++ struct buffer_head bh_tmp; ++ int needed_blocks; ++ int i, ret = 0, ret2; ++ ++ needed_blocks = ext3_san_write_trans_blocks(nblocks); ++ ++ lock_kernel(); ++ handle = ext3_journal_start(inode, needed_blocks); ++ if (IS_ERR(handle)) { ++ unlock_kernel(); ++ return PTR_ERR(handle); ++ } ++ unlock_kernel(); ++ ++ /* alloc blocks one by one */ ++ for (i = 0; i < nblocks; i++) { ++ ret = ext3_get_block_handle(handle, inode, blocks[i], ++ &bh_tmp, 1); ++ if (ret) ++ break; ++ ++ blocks[i] = bh_tmp.b_blocknr; ++ } ++ ++ /* set inode size if needed */ ++ if (!ret && (newsize > inode->i_size)) { ++ inode->i_size = newsize; ++ ext3_mark_inode_dirty(handle, inode); ++ } ++ ++ lock_kernel(); ++ ret2 = ext3_journal_stop(handle, inode); ++ unlock_kernel(); ++ ++ if (!ret) ++ ret = ret2; ++ return ret; ++} ++EXPORT_SYMBOL(ext3_prep_san_write); diff --git a/lustre/kernel_patches/patches/extN-wantedi.diff b/lustre/kernel_patches/patches/extN-wantedi.diff new file mode 100644 index 0000000..a55aec0 --- /dev/null +++ b/lustre/kernel_patches/patches/extN-wantedi.diff @@ -0,0 +1,163 @@ +--- lustre/extN-clean/namei.c 2002-12-30 05:56:09.000000000 -0500 ++++ lustre/extN/namei.c 2002-12-30 06:29:39.000000000 -0500 +@@ -1224,7 +1224,8 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = extN_new_inode (handle, dir, mode); ++ inode = extN_new_inode (handle, dir, mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + inode->i_op = &extN_file_inode_operations; +@@ -1254,7 +1254,8 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = extN_new_inode (handle, dir, mode); ++ inode = extN_new_inode (handle, dir, mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); +@@ -1286,7 +1286,8 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = extN_new_inode (handle, dir, S_IFDIR | mode); ++ inode = extN_new_inode (handle, dir, S_IFDIR | mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -1680,7 +1681,8 @@ + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = extN_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); ++ inode = extN_new_inode (handle, dir, S_IFLNK|S_IRWXUGO, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +--- lustre/extN-clean/ialloc.c 2002-12-28 23:56:42.000000000 -0500 ++++ lustre/extN/ialloc.c 2002-12-30 06:29:39.000000000 -0500 +@@ -329,8 +329,8 @@ + * For other inodes, search forward from the parent directory's block + * group to find a free inode. + */ +-struct inode * extN_new_inode (handle_t *handle, +- const struct inode * dir, int mode) ++struct inode *extN_new_inode(handle_t *handle, const struct inode *dir, ++ int mode, unsigned long goal) + { + struct super_block * sb; + struct buffer_head * bh; +@@ -360,6 +361,38 @@ + + lock_super (sb); + es = sbi->s_es; ++ ++ if (goal) { ++ i = (goal - 1) / EXTN_INODES_PER_GROUP(sb); ++ j = (goal - 1) % EXTN_INODES_PER_GROUP(sb); ++ gdp = extN_get_group_desc(sb, i, &bh2); ++ ++ bitmap_nr = load_inode_bitmap (sb, i); ++ if (bitmap_nr < 0) ++ goto fail; ++ ++ bh = sbi->s_inode_bitmap[bitmap_nr]; ++ ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = extN_journal_get_write_access(handle, bh); ++ if (err) goto fail; ++ ++ if (extN_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 extN_journal_dirty_metadata"); ++ err = extN_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; +@@ -474,6 +509,7 @@ + } + goto repeat; + } ++have_bit_and_group: + j += i * sbi->s_inodes_per_group + 1; + if (j < sbi->s_first_ino || j > le32_to_cpu(es->s_inodes_count)) { + extN_error (sb, "extN_new_inode", +--- lustre/extN-clean/ioctl.c 2002-12-28 23:56:42.000000000 -0500 ++++ lustre/extN/ioctl.c 2002-12-30 06:29:39.000000000 -0500 +@@ -24,6 +24,31 @@ + extN_debug ("cmd = %u, arg = %lu\n", cmd, arg); + + switch (cmd) { ++ case EXTN_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 EXTN_IOC_GETFLAGS: + flags = ei->i_flags & EXTN_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); +--- lustre/include/linux/extN_fs.h~ 2002-12-30 06:01:43.000000000 -0500 ++++ lustre/include/linux/extN_fs.h 2002-12-30 06:02:51.000000000 -0500 +@@ -200,6 +200,7 @@ + #define EXTN_IOC_SETFLAGS _IOW('f', 2, long) + #define EXTN_IOC_GETVERSION _IOR('f', 3, long) + #define EXTN_IOC_SETVERSION _IOW('f', 4, long) ++/* EXTN_IOC_CREATE_INUM at bottom of file (visible to kernel and user). */ + #define EXTN_IOC_GETVERSION_OLD _IOR('v', 1, long) + #define EXTN_IOC_SETVERSION_OLD _IOW('v', 2, long) + #ifdef CONFIG_JBD_DEBUG +@@ -632,7 +633,8 @@ + extern int extN_sync_file (struct file *, struct dentry *, int); + + /* ialloc.c */ +-extern struct inode * extN_new_inode (handle_t *, const struct inode *, int); ++extern struct inode * extN_new_inode (handle_t *, const struct inode *, int, ++ unsigned long); + extern void extN_free_inode (handle_t *, struct inode *); + extern struct inode * extN_orphan_get (struct super_block *, ino_t); + extern unsigned long extN_count_free_inodes (struct super_block *); +@@ -714,4 +716,6 @@ + + #endif /* __KERNEL__ */ + ++#define EXTN_IOC_CREATE_INUM _IOW('f', 5, long) ++ + #endif /* _LINUX_EXTN_FS_H */ diff --git a/lustre/kernel_patches/patches/extN-wantedi.patch b/lustre/kernel_patches/patches/extN-wantedi.patch new file mode 100644 index 0000000..efee46f --- /dev/null +++ b/lustre/kernel_patches/patches/extN-wantedi.patch @@ -0,0 +1,169 @@ + fs/ext3/ialloc.c | 36 +++++++++++++++++++++++++++++++++++- + fs/ext3/ioctl.c | 25 +++++++++++++++++++++++++ + fs/ext3/namei.c | 12 ++++++++---- + include/linux/ext3_fs.h | 5 ++++- + 4 files changed, 72 insertions(+), 6 deletions(-) + +--- linux-2.4.18-18/fs/ext3/namei.c~extN-wantedi Sat Apr 5 02:51:31 2003 ++++ linux-2.4.18-18-braam/fs/ext3/namei.c Sat Apr 5 02:51:31 2003 +@@ -1212,7 +1212,8 @@ static int ext3_create (struct inode * d + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode (handle, dir, mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + inode->i_op = &ext3_file_inode_operations; +@@ -1240,7 +1241,8 @@ static int ext3_mknod (struct inode * di + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, mode); ++ inode = ext3_new_inode (handle, dir, mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); +@@ -1270,7 +1272,8 @@ static int ext3_mkdir(struct inode * dir + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR | mode); ++ inode = ext3_new_inode (handle, dir, S_IFDIR | mode, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -1663,7 +1666,8 @@ static int ext3_symlink (struct inode * + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFLNK|S_IRWXUGO); ++ inode = ext3_new_inode (handle, dir, S_IFLNK|S_IRWXUGO, ++ (unsigned long)dentry->d_fsdata); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +--- linux-2.4.18-18/fs/ext3/ialloc.c~extN-wantedi Sat Apr 5 02:51:31 2003 ++++ linux-2.4.18-18-braam/fs/ext3/ialloc.c Sat Apr 5 02:51:31 2003 +@@ -330,7 +330,8 @@ int ext3_itable_block_used(struct super_ + * group to find a free inode. + */ + struct inode * ext3_new_inode (handle_t *handle, +- const struct inode * dir, int mode) ++ const struct inode * dir, int mode, ++ unsigned long goal) + { + struct super_block * sb; + struct buffer_head * bh; +@@ -360,6 +361,38 @@ struct inode * ext3_new_inode (handle_t + + lock_super (sb); + es = sbi->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) ++ goto fail; ++ ++ bh = sbi->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; +@@ -474,6 +507,7 @@ repeat: + } + goto repeat; + } ++ have_bit_and_group: + j += i * sbi->s_inodes_per_group + 1; + if (j < sbi->s_first_ino || j > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", +--- linux-2.4.18-18/fs/ext3/ioctl.c~extN-wantedi Sat Apr 5 02:51:31 2003 ++++ linux-2.4.18-18-braam/fs/ext3/ioctl.c Sat Apr 5 02:51:31 2003 +@@ -24,6 +24,31 @@ int ext3_ioctl (struct inode * inode, st + 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 = ei->i_flags & EXT3_FL_USER_VISIBLE; + return put_user(flags, (int *) arg); +--- linux-2.4.18-18/include/linux/ext3_fs.h~extN-wantedi Sat Apr 5 02:51:31 2003 ++++ linux-2.4.18-18-braam/include/linux/ext3_fs.h Sat Apr 5 02:52:43 2003 +@@ -198,6 +198,7 @@ struct ext3_group_desc + #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 +@@ -628,7 +629,8 @@ extern int ext3_check_dir_entry(const ch + extern int ext3_sync_file (struct file *, struct dentry *, int); + + /* ialloc.c */ +-extern struct inode * ext3_new_inode (handle_t *, const 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 *); +@@ -712,4 +714,5 @@ extern struct inode_operations ext3_fast + + #endif /* __KERNEL__ */ + ++#define EXT3_IOC_CREATE_INUM _IOW('f', 5, long) + #endif /* _LINUX_EXT3_FS_H */ + +_ diff --git a/lustre/kernel_patches/patches/extfs-update-2.4.20rc1.patch b/lustre/kernel_patches/patches/extfs-update-2.4.20rc1.patch new file mode 100644 index 0000000..de0b8f1 --- /dev/null +++ b/lustre/kernel_patches/patches/extfs-update-2.4.20rc1.patch @@ -0,0 +1,4391 @@ +# This is a BitKeeper generated patch for the following project: +# Project Name: Linux kernel tree +# This patch format is intended for GNU patch command version 2.5 or higher. +# This patch includes the following deltas: +# ChangeSet 1.775 -> 1.783 +# fs/ext3/namei.c 1.2 -> 1.6 +# include/linux/ext3_fs.h 1.4 -> 1.9 +# fs/ext2/inode.c 1.15 -> 1.16 +# lib/rbtree.c 1.2 -> 1.3 +# fs/ext3/Makefile 1.2 -> 1.3 +# include/linux/ext3_jbd.h 1.2 -> 1.3 +# include/linux/ext2_fs.h 1.6 -> 1.9 +# include/linux/ext2_fs_sb.h 1.1 -> 1.2 +# include/linux/rbtree.h 1.1 -> 1.2 +# include/linux/ext2_fs_i.h 1.4 -> 1.5 +# fs/ext3/ialloc.c 1.4 -> 1.5 +# fs/ext2/ialloc.c 1.8 -> 1.10 +# fs/ext3/dir.c 1.1 -> 1.3 +# fs/ext2/super.c 1.9 -> 1.13 +# fs/ext3/inode.c 1.10 -> 1.11 +# fs/ext3/super.c 1.7 -> 1.12 +# include/linux/ext3_fs_sb.h 1.2 -> 1.4 +# fs/ext3/file.c 1.3 -> 1.4 +# (new) -> 1.1 fs/ext3/hash.c +# +# The following is the BitKeeper ChangeSet Log +# -------------------------------------------- +# 02/11/06 tytso@snap.thunk.org 1.776 +# Port patch-ext3-dxdir-2.4.19-4 to 2.4.20-rc1 +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.777 +# Add '.' and '..' entries to be returned by readdir of htree directories +# +# This patch from Chris Li adds '.' and '..' to the rbtree so that they +# are properly returned by readdir. +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.778 +# Check for failed kmalloc() in ext3_htree_store_dirent() +# +# This patch checks for a failed kmalloc() in ext3_htree_store_dirent(), +# and passes the error up to its caller, ext3_htree_fill_tree(). +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.779 +# Fix ext3 htree rename bug. +# +# This fixes an ext3 htree bug pointed out by Christopher Li; if +# adding the new name to the directory causes a split, this can cause +# the directory entry containing the old name to move to another +# block, and then the removal of the old name will fail. +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.780 +# Default mount options from superblock for ext2/3 filesystems +# +# This patch adds support for default mount options to be stored in the +# superblock, so they don't have to be specified on the mount command line +# (or in /etc/fstab). While I was in the code, I also cleaned up the +# handling of how mount options are processed in the ext2 and ext3 +# filesystems. +# +# Most mount options are now processed *after* the superblock has been +# read in. This allows for a much cleaner handling of those default mount +# option parameters that were already stored in the superblock: the +# resuid, resgid, and s_errors fields were handled using some fairly gross +# special cases. Now the only mount option which is processed first is +# the sb option, which specifies the location of the superblock. This +# allows the handling of all of the default mount parameters to be much +# more cleanly and more generally handled. +# +# This does change the behaviour from earlier kernels, in that if the sb +# mount option is specified, it must be specified *first*. However, this +# option is rarely used, and if it is, it generally is specified first, so +# this seems to be a reasonable restriction. +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.781 +# Ext2/3 forward compatibility: on-line resizing +# +# This patch allows forward compatibility with future filesystems which +# are dynamically grown by using an alternate algorithm for storing the +# block group descriptors. It's also a bit more efficient, in that it +# uses just a little bit less disk space. Currently, the ext2 filesystem +# format requires either relocating the inode table, or reserving space in +# before doing the on-line resize. The new scheme, which is documented in +# "Planned Extensions to the Ext2/3 Filesystem", by Stephen Tweedie and I +# (see: http://e2fsprogs.sourceforge.net/extensions-ext23) +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.782 +# Ext2/3 forward compatibility: inode size +# +# This patch allows filesystems with expanded inodes to be mounted. +# (compatibility feature flags will be used to control whether or not the +# filesystem should be mounted in case the new inode fields will result in +# compatibility issues). This allows for future compatibility with newer +# versions of ext2fs. +# -------------------------------------------- +# 02/11/07 tytso@snap.thunk.org 1.783 +# Orlov block allocator for ext2/3 +# +# This is Al's implementation of the Orlov block allocator for ext2/3. +# +# At least doubles the throughput for the traverse-a-kernel-tree +# test and is well tested. +# -------------------------------------------- +# +diff -Nru a/fs/ext2/ialloc.c b/fs/ext2/ialloc.c +--- a/fs/ext2/ialloc.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext2/ialloc.c Thu Nov 7 11:58:05 2002 +@@ -17,6 +17,7 @@ + #include + #include + #include ++#include + + + /* +@@ -228,8 +229,7 @@ + * For other inodes, search forward from the parent directory\'s block + * group to find a free inode. + */ +- +-static int find_group_dir(struct super_block *sb, int parent_group) ++static int find_group_dir(struct super_block *sb, const struct inode *parent) + { + struct ext2_super_block * es = sb->u.ext2_sb.s_es; + int ngroups = sb->u.ext2_sb.s_groups_count; +@@ -262,8 +262,140 @@ + return best_group; + } + +-static int find_group_other(struct super_block *sb, int parent_group) ++/* ++ * Orlov's allocator for directories. ++ * ++ * We always try to spread first-level directories. ++ * ++ * If there are blockgroups with both free inodes and free blocks counts ++ * not worse than average we return one with smallest directory count. ++ * Otherwise we simply return a random group. ++ * ++ * For the rest rules look so: ++ * ++ * It's OK to put directory into a group unless ++ * it has too many directories already (max_dirs) or ++ * it has too few free inodes left (min_inodes) or ++ * it has too few free blocks left (min_blocks) or ++ * it's already running too large debt (max_debt). ++ * Parent's group is prefered, if it doesn't satisfy these ++ * conditions we search cyclically through the rest. If none ++ * of the groups look good we just look for a group with more ++ * free inodes than average (starting at parent's group). ++ * ++ * Debt is incremented each time we allocate a directory and decremented ++ * when we allocate an inode, within 0--255. ++ */ ++ ++#define INODE_COST 64 ++#define BLOCK_COST 256 ++ ++static int find_group_orlov(struct super_block *sb, const struct inode *parent) + { ++ int parent_group = parent->u.ext2_i.i_block_group; ++ struct ext2_sb_info *sbi = EXT2_SB(sb); ++ struct ext2_super_block *es = sbi->s_es; ++ int ngroups = sbi->s_groups_count; ++ int inodes_per_group = EXT2_INODES_PER_GROUP(sb); ++ int avefreei = le32_to_cpu(es->s_free_inodes_count) / ngroups; ++ int avefreeb = le32_to_cpu(es->s_free_blocks_count) / ngroups; ++ int blocks_per_dir; ++ int ndirs = sbi->s_dir_count; ++ int max_debt, max_dirs, min_blocks, min_inodes; ++ int group = -1, i; ++ struct ext2_group_desc *desc; ++ struct buffer_head *bh; ++ ++ if ((parent == sb->s_root->d_inode) || ++ (parent->i_flags & EXT2_TOPDIR_FL)) { ++ struct ext2_group_desc *best_desc = NULL; ++ struct buffer_head *best_bh = NULL; ++ int best_ndir = inodes_per_group; ++ int best_group = -1; ++ ++ get_random_bytes(&group, sizeof(group)); ++ parent_group = (unsigned)group % ngroups; ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext2_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) ++ continue; ++ if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb) ++ continue; ++ best_group = group; ++ best_ndir = le16_to_cpu(desc->bg_used_dirs_count); ++ best_desc = desc; ++ best_bh = bh; ++ } ++ if (best_group >= 0) { ++ desc = best_desc; ++ bh = best_bh; ++ group = best_group; ++ goto found; ++ } ++ goto fallback; ++ } ++ ++ blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - ++ le32_to_cpu(es->s_free_blocks_count)) / ndirs; ++ ++ max_dirs = ndirs / ngroups + inodes_per_group / 16; ++ min_inodes = avefreei - inodes_per_group / 4; ++ min_blocks = avefreeb - EXT2_BLOCKS_PER_GROUP(sb) / 4; ++ ++ max_debt = EXT2_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, BLOCK_COST); ++ if (max_debt * INODE_COST > inodes_per_group) ++ max_debt = inodes_per_group / INODE_COST; ++ if (max_debt > 255) ++ max_debt = 255; ++ if (max_debt == 0) ++ max_debt = 1; ++ ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext2_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (sbi->s_debts[group] >= max_debt) ++ continue; ++ if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes) ++ continue; ++ if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks) ++ continue; ++ goto found; ++ } ++ ++fallback: ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext2_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei) ++ goto found; ++ } ++ ++ return -1; ++ ++found: ++ desc->bg_free_inodes_count = ++ cpu_to_le16(le16_to_cpu(desc->bg_free_inodes_count) - 1); ++ desc->bg_used_dirs_count = ++ cpu_to_le16(le16_to_cpu(desc->bg_used_dirs_count) + 1); ++ sbi->s_dir_count++; ++ mark_buffer_dirty(bh); ++ return group; ++} ++ ++static int find_group_other(struct super_block *sb, const struct inode *parent) ++{ ++ int parent_group = parent->u.ext2_i.i_block_group; + int ngroups = sb->u.ext2_sb.s_groups_count; + struct ext2_group_desc *desc; + struct buffer_head *bh; +@@ -331,10 +463,13 @@ + lock_super (sb); + es = sb->u.ext2_sb.s_es; + repeat: +- if (S_ISDIR(mode)) +- group = find_group_dir(sb, dir->u.ext2_i.i_block_group); +- else +- group = find_group_other(sb, dir->u.ext2_i.i_block_group); ++ if (S_ISDIR(mode)) { ++ if (test_opt (sb, OLDALLOC)) ++ group = find_group_dir(sb, dir); ++ else ++ group = find_group_orlov(sb, dir); ++ } else ++ group = find_group_other(sb, dir); + + err = -ENOSPC; + if (group == -1) +@@ -368,6 +503,15 @@ + + es->s_free_inodes_count = + cpu_to_le32(le32_to_cpu(es->s_free_inodes_count) - 1); ++ ++ if (S_ISDIR(mode)) { ++ if (EXT2_SB(sb)->s_debts[group] < 255) ++ EXT2_SB(sb)->s_debts[group]++; ++ } else { ++ if (EXT2_SB(sb)->s_debts[group]) ++ EXT2_SB(sb)->s_debts[group]--; ++ } ++ + mark_buffer_dirty(sb->u.ext2_sb.s_sbh); + sb->s_dirt = 1; + inode->i_uid = current->fsuid; +@@ -385,7 +529,7 @@ + inode->i_blksize = PAGE_SIZE; /* This is the optimal IO size (for stat), not the fs block size */ + inode->i_blocks = 0; + inode->i_mtime = inode->i_atime = inode->i_ctime = CURRENT_TIME; +- inode->u.ext2_i.i_new_inode = 1; ++ inode->u.ext2_i.i_state = EXT2_STATE_NEW; + inode->u.ext2_i.i_flags = dir->u.ext2_i.i_flags; + if (S_ISLNK(mode)) + inode->u.ext2_i.i_flags &= ~(EXT2_IMMUTABLE_FL|EXT2_APPEND_FL); +@@ -469,6 +613,21 @@ + #else + return le32_to_cpu(sb->u.ext2_sb.s_es->s_free_inodes_count); + #endif ++} ++ ++/* Called at mount-time, super-block is locked */ ++unsigned long ext2_count_dirs (struct super_block * sb) ++{ ++ unsigned long count = 0; ++ int i; ++ ++ for (i = 0; i < EXT2_SB(sb)->s_groups_count; i++) { ++ struct ext2_group_desc *gdp = ext2_get_group_desc (sb, i, NULL); ++ if (!gdp) ++ continue; ++ count += le16_to_cpu(gdp->bg_used_dirs_count); ++ } ++ return count; + } + + #ifdef CONFIG_EXT2_CHECK +diff -Nru a/fs/ext2/inode.c b/fs/ext2/inode.c +--- a/fs/ext2/inode.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext2/inode.c Thu Nov 7 11:58:05 2002 +@@ -955,6 +955,7 @@ + else + inode->u.ext2_i.i_dir_acl = le32_to_cpu(raw_inode->i_dir_acl); + inode->i_generation = le32_to_cpu(raw_inode->i_generation); ++ inode->u.ext2_i.i_state = 0; + inode->u.ext2_i.i_prealloc_count = 0; + inode->u.ext2_i.i_block_group = block_group; + +@@ -1061,6 +1062,11 @@ + offset &= EXT2_BLOCK_SIZE(inode->i_sb) - 1; + raw_inode = (struct ext2_inode *) (bh->b_data + offset); + ++ /* For fields not tracked in the in-memory inode, ++ * initialise them to zero for new inodes. */ ++ if (inode->u.ext2_i.i_state & EXT2_STATE_NEW) ++ memset(raw_inode, 0, EXT2_SB(inode->i_sb)->s_inode_size); ++ + raw_inode->i_mode = cpu_to_le16(inode->i_mode); + if(!(test_opt(inode->i_sb, NO_UID32))) { + raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid)); +@@ -1133,6 +1139,7 @@ + err = -EIO; + } + } ++ inode->u.ext2_i.i_state &= ~EXT2_STATE_NEW; + brelse (bh); + return err; + } +diff -Nru a/fs/ext2/super.c b/fs/ext2/super.c +--- a/fs/ext2/super.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext2/super.c Thu Nov 7 11:58:05 2002 +@@ -48,16 +48,12 @@ + va_start (args, fmt); + vsprintf (error_buf, fmt, args); + va_end (args); +- if (test_opt (sb, ERRORS_PANIC) || +- (le16_to_cpu(sb->u.ext2_sb.s_es->s_errors) == EXT2_ERRORS_PANIC && +- !test_opt (sb, ERRORS_CONT) && !test_opt (sb, ERRORS_RO))) ++ if (test_opt (sb, ERRORS_PANIC)) + panic ("EXT2-fs panic (device %s): %s: %s\n", + bdevname(sb->s_dev), function, error_buf); + printk (KERN_CRIT "EXT2-fs error (device %s): %s: %s\n", + bdevname(sb->s_dev), function, error_buf); +- if (test_opt (sb, ERRORS_RO) || +- (le16_to_cpu(sb->u.ext2_sb.s_es->s_errors) == EXT2_ERRORS_RO && +- !test_opt (sb, ERRORS_CONT) && !test_opt (sb, ERRORS_PANIC))) { ++ if (test_opt (sb, ERRORS_RO)) { + printk ("Remounting filesystem read-only\n"); + sb->s_flags |= MS_RDONLY; + } +@@ -136,6 +132,7 @@ + if (sb->u.ext2_sb.s_group_desc[i]) + brelse (sb->u.ext2_sb.s_group_desc[i]); + kfree(sb->u.ext2_sb.s_group_desc); ++ kfree(sb->u.ext2_sb.s_debts); + for (i = 0; i < EXT2_MAX_GROUP_LOADED; i++) + if (sb->u.ext2_sb.s_inode_bitmap[i]) + brelse (sb->u.ext2_sb.s_inode_bitmap[i]); +@@ -158,12 +155,61 @@ + remount_fs: ext2_remount, + }; + ++static unsigned long get_sb_block(void **data) ++{ ++ unsigned long sb_block; ++ char *options = (char *) *data; ++ ++ if (!options || strncmp(options, "sb=", 3) != 0) ++ return 1; /* Default location */ ++ options += 3; ++ sb_block = simple_strtoul(options, &options, 0); ++ if (*options && *options != ',') { ++ printk("EXT2-fs: Invalid sb specification: %s\n", ++ (char *) *data); ++ return 1; ++ } ++ if (*options == ',') ++ options++; ++ *data = (void *) options; ++ return sb_block; ++} ++ ++static int want_value(char *value, char *option) ++{ ++ if (!value || !*value) { ++ printk(KERN_NOTICE "EXT2-fs: the %s option needs an argument\n", ++ option); ++ return -1; ++ } ++ return 0; ++} ++ ++static int want_null_value(char *value, char *option) ++{ ++ if (*value) { ++ printk(KERN_NOTICE "EXT2-fs: Invalid %s argument: %s\n", ++ option, value); ++ return -1; ++ } ++ return 0; ++} ++ ++static int want_numeric(char *value, char *option, unsigned long *number) ++{ ++ if (want_value(value, option)) ++ return -1; ++ *number = simple_strtoul(value, &value, 0); ++ if (want_null_value(value, option)) ++ return -1; ++ return 0; ++} ++ + /* + * This function has been shamelessly adapted from the msdos fs + */ +-static int parse_options (char * options, unsigned long * sb_block, +- unsigned short *resuid, unsigned short * resgid, +- unsigned long * mount_options) ++static int parse_options (char * options, ++ struct ext2_sb_info *sbi) + { + char * this_char; + char * value; +@@ -176,22 +222,22 @@ + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; + if (!strcmp (this_char, "bsddf")) +- clear_opt (*mount_options, MINIX_DF); ++ clear_opt (sbi->s_mount_opt, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +- set_opt (*mount_options, NO_UID32); ++ set_opt (sbi->s_mount_opt, NO_UID32); + } + else if (!strcmp (this_char, "check")) { + if (!value || !*value || !strcmp (value, "none")) +- clear_opt (*mount_options, CHECK); ++ clear_opt (sbi->s_mount_opt, CHECK); + else + #ifdef CONFIG_EXT2_CHECK +- set_opt (*mount_options, CHECK); ++ set_opt (sbi->s_mount_opt, CHECK); + #else + printk("EXT2 Check option not supported\n"); + #endif + } + else if (!strcmp (this_char, "debug")) +- set_opt (*mount_options, DEBUG); ++ set_opt (sbi->s_mount_opt, DEBUG); + else if (!strcmp (this_char, "errors")) { + if (!value || !*value) { + printk ("EXT2-fs: the errors option requires " +@@ -199,19 +245,19 @@ + return 0; + } + if (!strcmp (value, "continue")) { +- clear_opt (*mount_options, ERRORS_RO); +- clear_opt (*mount_options, ERRORS_PANIC); +- set_opt (*mount_options, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_RO); ++ clear_opt (sbi->s_mount_opt, ERRORS_PANIC); ++ set_opt (sbi->s_mount_opt, ERRORS_CONT); + } + else if (!strcmp (value, "remount-ro")) { +- clear_opt (*mount_options, ERRORS_CONT); +- clear_opt (*mount_options, ERRORS_PANIC); +- set_opt (*mount_options, ERRORS_RO); ++ clear_opt (sbi->s_mount_opt, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_PANIC); ++ set_opt (sbi->s_mount_opt, ERRORS_RO); + } + else if (!strcmp (value, "panic")) { +- clear_opt (*mount_options, ERRORS_CONT); +- clear_opt (*mount_options, ERRORS_RO); +- set_opt (*mount_options, ERRORS_PANIC); ++ clear_opt (sbi->s_mount_opt, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_RO); ++ set_opt (sbi->s_mount_opt, ERRORS_PANIC); + } + else { + printk ("EXT2-fs: Invalid errors option: %s\n", +@@ -221,53 +267,30 @@ + } + else if (!strcmp (this_char, "grpid") || + !strcmp (this_char, "bsdgroups")) +- set_opt (*mount_options, GRPID); ++ set_opt (sbi->s_mount_opt, GRPID); + else if (!strcmp (this_char, "minixdf")) +- set_opt (*mount_options, MINIX_DF); ++ set_opt (sbi->s_mount_opt, MINIX_DF); + else if (!strcmp (this_char, "nocheck")) +- clear_opt (*mount_options, CHECK); ++ clear_opt (sbi->s_mount_opt, CHECK); + else if (!strcmp (this_char, "nogrpid") || + !strcmp (this_char, "sysvgroups")) +- clear_opt (*mount_options, GRPID); ++ clear_opt (sbi->s_mount_opt, GRPID); + else if (!strcmp (this_char, "resgid")) { +- if (!value || !*value) { +- printk ("EXT2-fs: the resgid option requires " +- "an argument\n"); ++ unsigned long v; ++ if (want_numeric(value, "resgid", &v)) + return 0; +- } +- *resgid = simple_strtoul (value, &value, 0); +- if (*value) { +- printk ("EXT2-fs: Invalid resgid option: %s\n", +- value); +- return 0; +- } ++ sbi->s_resgid = v; + } + else if (!strcmp (this_char, "resuid")) { +- if (!value || !*value) { +- printk ("EXT2-fs: the resuid option requires " +- "an argument"); ++ unsigned long v; ++ if (want_numeric(value, "resuid", &v)) + return 0; +- } +- *resuid = simple_strtoul (value, &value, 0); +- if (*value) { +- printk ("EXT2-fs: Invalid resuid option: %s\n", +- value); +- return 0; +- } +- } +- else if (!strcmp (this_char, "sb")) { +- if (!value || !*value) { +- printk ("EXT2-fs: the sb option requires " +- "an argument"); +- return 0; +- } +- *sb_block = simple_strtoul (value, &value, 0); +- if (*value) { +- printk ("EXT2-fs: Invalid sb option: %s\n", +- value); +- return 0; +- } ++ sbi->s_resuid = v; + } ++ else if (!strcmp (this_char, "oldalloc")) ++ set_opt (sbi->s_mount_opt, OLDALLOC); ++ else if (!strcmp (this_char, "orlov")) ++ clear_opt (sbi->s_mount_opt, OLDALLOC); + /* Silently ignore the quota options */ + else if (!strcmp (this_char, "grpquota") + || !strcmp (this_char, "noquota") +@@ -397,17 +420,37 @@ + return res; + } + ++static unsigned long descriptor_loc(struct super_block *sb, ++ unsigned long logic_sb_block, ++ int nr) ++{ ++ struct ext2_sb_info *sbi = EXT2_SB(sb); ++ unsigned long bg, first_data_block, first_meta_bg; ++ int has_super = 0; ++ ++ first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block); ++ first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); ++ ++ if (!EXT2_HAS_INCOMPAT_FEATURE(sb, EXT2_FEATURE_INCOMPAT_META_BG) || ++ nr < first_meta_bg) ++ return (logic_sb_block + nr + 1); ++ bg = sbi->s_desc_per_block * nr; ++ if (ext2_bg_has_super(sb, bg)) ++ has_super = 1; ++ return (first_data_block + has_super + (bg * sbi->s_blocks_per_group)); ++} ++ + struct super_block * ext2_read_super (struct super_block * sb, void * data, + int silent) + { + struct buffer_head * bh; ++ struct ext2_sb_info * sbi = EXT2_SB(sb); + struct ext2_super_block * es; +- unsigned long sb_block = 1; +- unsigned short resuid = EXT2_DEF_RESUID; +- unsigned short resgid = EXT2_DEF_RESGID; +- unsigned long logic_sb_block = 1; +- unsigned long offset = 0; ++ unsigned long sb_block = get_sb_block(&data); ++ unsigned long block, logic_sb_block; ++ unsigned long offset; + kdev_t dev = sb->s_dev; ++ unsigned long def_mount_opts; + int blocksize = BLOCK_SIZE; + int db_count; + int i, j; +@@ -423,12 +466,6 @@ + if(blocksize < BLOCK_SIZE ) + blocksize = BLOCK_SIZE; + +- sb->u.ext2_sb.s_mount_opt = 0; +- if (!parse_options ((char *) data, &sb_block, &resuid, &resgid, +- &sb->u.ext2_sb.s_mount_opt)) { +- return NULL; +- } +- + if (set_blocksize(dev, blocksize) < 0) { + printk ("EXT2-fs: unable to set blocksize %d\n", blocksize); + return NULL; +@@ -436,14 +473,11 @@ + sb->s_blocksize = blocksize; + + /* +- * If the superblock doesn't start on a sector boundary, +- * calculate the offset. FIXME(eric) this doesn't make sense +- * that we would have to do this. ++ * If the superblock doesn't start on a hardware sector boundary, ++ * calculate the offset. + */ +- if (blocksize != BLOCK_SIZE) { +- logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize; +- offset = (sb_block*BLOCK_SIZE) % blocksize; +- } ++ logic_sb_block = (sb_block*BLOCK_SIZE) / blocksize; ++ offset = (sb_block*BLOCK_SIZE) % blocksize; + + if (!(bh = sb_bread(sb, logic_sb_block))) { + printk ("EXT2-fs: unable to read superblock\n"); +@@ -462,6 +496,27 @@ + bdevname(dev)); + goto failed_mount; + } ++ ++ /* Set defaults before we parse the mount options */ ++ def_mount_opts = le32_to_cpu(es->s_default_mount_opts); ++ if (def_mount_opts & EXT2_DEFM_DEBUG) ++ set_opt(sbi->s_mount_opt, DEBUG); ++ if (def_mount_opts & EXT2_DEFM_BSDGROUPS) ++ set_opt(sbi->s_mount_opt, GRPID); ++ if (def_mount_opts & EXT2_DEFM_UID16) ++ set_opt(sbi->s_mount_opt, NO_UID32); ++ ++ if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_PANIC) ++ set_opt(sbi->s_mount_opt, ERRORS_PANIC); ++ else if (le16_to_cpu(sbi->s_es->s_errors) == EXT2_ERRORS_RO) ++ set_opt(sbi->s_mount_opt, ERRORS_RO); ++ ++ sbi->s_resuid = le16_to_cpu(es->s_def_resuid); ++ sbi->s_resgid = le16_to_cpu(es->s_def_resgid); ++ ++ if (!parse_options ((char *) data, sbi)) ++ goto failed_mount; ++ + if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV && + (EXT2_HAS_COMPAT_FEATURE(sb, ~0U) || + EXT2_HAS_RO_COMPAT_FEATURE(sb, ~0U) || +@@ -522,14 +577,16 @@ + } + + if (le32_to_cpu(es->s_rev_level) == EXT2_GOOD_OLD_REV) { +- sb->u.ext2_sb.s_inode_size = EXT2_GOOD_OLD_INODE_SIZE; +- sb->u.ext2_sb.s_first_ino = EXT2_GOOD_OLD_FIRST_INO; ++ sbi->s_inode_size = EXT2_GOOD_OLD_INODE_SIZE; ++ sbi->s_first_ino = EXT2_GOOD_OLD_FIRST_INO; + } else { +- sb->u.ext2_sb.s_inode_size = le16_to_cpu(es->s_inode_size); +- sb->u.ext2_sb.s_first_ino = le32_to_cpu(es->s_first_ino); +- if (sb->u.ext2_sb.s_inode_size != EXT2_GOOD_OLD_INODE_SIZE) { ++ sbi->s_inode_size = le16_to_cpu(es->s_inode_size); ++ sbi->s_first_ino = le32_to_cpu(es->s_first_ino); ++ if ((sbi->s_inode_size < EXT2_GOOD_OLD_INODE_SIZE) || ++ (sbi->s_inode_size & (sbi->s_inode_size - 1)) || ++ (sbi->s_inode_size > blocksize)) { + printk ("EXT2-fs: unsupported inode size: %d\n", +- sb->u.ext2_sb.s_inode_size); ++ sbi->s_inode_size); + goto failed_mount; + } + } +@@ -550,14 +607,6 @@ + sb->u.ext2_sb.s_desc_per_block = sb->s_blocksize / + sizeof (struct ext2_group_desc); + sb->u.ext2_sb.s_sbh = bh; +- if (resuid != EXT2_DEF_RESUID) +- sb->u.ext2_sb.s_resuid = resuid; +- else +- sb->u.ext2_sb.s_resuid = le16_to_cpu(es->s_def_resuid); +- if (resgid != EXT2_DEF_RESGID) +- sb->u.ext2_sb.s_resgid = resgid; +- else +- sb->u.ext2_sb.s_resgid = le16_to_cpu(es->s_def_resgid); + sb->u.ext2_sb.s_mount_state = le16_to_cpu(es->s_state); + sb->u.ext2_sb.s_addr_per_block_bits = + log2 (EXT2_ADDR_PER_BLOCK(sb)); +@@ -610,14 +659,22 @@ + printk ("EXT2-fs: not enough memory\n"); + goto failed_mount; + } ++ sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(*sbi->s_debts), ++ GFP_KERNEL); ++ if (!sbi->s_debts) { ++ printk ("EXT2-fs: not enough memory\n"); ++ goto failed_mount_group_desc; ++ } ++ memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(*sbi->s_debts)); + for (i = 0; i < db_count; i++) { +- sb->u.ext2_sb.s_group_desc[i] = sb_bread(sb, logic_sb_block + i + 1); +- if (!sb->u.ext2_sb.s_group_desc[i]) { ++ block = descriptor_loc(sb, logic_sb_block, i); ++ sbi->s_group_desc[i] = sb_bread(sb, block); ++ if (!sbi->s_group_desc[i]) { + for (j = 0; j < i; j++) +- brelse (sb->u.ext2_sb.s_group_desc[j]); +- kfree(sb->u.ext2_sb.s_group_desc); ++ brelse (sbi->s_group_desc[j]); ++ kfree(sbi->s_group_desc); + printk ("EXT2-fs: unable to read group descriptors\n"); +- goto failed_mount; ++ goto failed_mount_group_desc; + } + } + if (!ext2_check_descriptors (sb)) { +@@ -634,6 +691,7 @@ + sb->u.ext2_sb.s_loaded_inode_bitmaps = 0; + sb->u.ext2_sb.s_loaded_block_bitmaps = 0; + sb->u.ext2_sb.s_gdb_count = db_count; ++ sb->u.ext2_sb.s_dir_count = ext2_count_dirs(sb); + /* + * set up enough so that it can read an inode + */ +@@ -654,7 +712,10 @@ + failed_mount2: + for (i = 0; i < db_count; i++) + brelse(sb->u.ext2_sb.s_group_desc[i]); ++failed_mount_group_desc: + kfree(sb->u.ext2_sb.s_group_desc); ++ if (sb->u.ext2_sb.s_debts) ++ kfree(sb->u.ext2_sb.s_debts); + failed_mount: + brelse(bh); + return NULL; +@@ -709,24 +770,16 @@ + + int ext2_remount (struct super_block * sb, int * flags, char * data) + { ++ struct ext2_sb_info * sbi = EXT2_SB(sb); + struct ext2_super_block * es; +- unsigned short resuid = sb->u.ext2_sb.s_resuid; +- unsigned short resgid = sb->u.ext2_sb.s_resgid; +- unsigned long new_mount_opt; +- unsigned long tmp; + + /* + * Allow the "check" option to be passed as a remount option. + */ +- new_mount_opt = sb->u.ext2_sb.s_mount_opt; +- if (!parse_options (data, &tmp, &resuid, &resgid, +- &new_mount_opt)) ++ if (!parse_options (data, sbi)) + return -EINVAL; + +- sb->u.ext2_sb.s_mount_opt = new_mount_opt; +- sb->u.ext2_sb.s_resuid = resuid; +- sb->u.ext2_sb.s_resgid = resgid; +- es = sb->u.ext2_sb.s_es; ++ es = sbi->s_es; + if ((*flags & MS_RDONLY) == (sb->s_flags & MS_RDONLY)) + return 0; + if (*flags & MS_RDONLY) { +diff -Nru a/fs/ext3/Makefile b/fs/ext3/Makefile +--- a/fs/ext3/Makefile Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/Makefile Thu Nov 7 11:58:05 2002 +@@ -10,7 +10,7 @@ + O_TARGET := ext3.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o ++ ioctl.o namei.o super.o symlink.o hash.o + obj-m := $(O_TARGET) + + include $(TOPDIR)/Rules.make +diff -Nru a/fs/ext3/dir.c b/fs/ext3/dir.c +--- a/fs/ext3/dir.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/dir.c Thu Nov 7 11:58:05 2002 +@@ -21,12 +21,16 @@ + #include + #include + #include ++#include ++#include + + static unsigned char ext3_filetype_table[] = { + DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK + }; + + static int ext3_readdir(struct file *, void *, filldir_t); ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir); + + struct file_operations ext3_dir_operations = { + read: generic_read_dir, +@@ -35,6 +39,17 @@ + fsync: ext3_sync_file, /* BKL held */ + }; + ++ ++static unsigned char get_dtype(struct super_block *sb, int filetype) ++{ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_FILETYPE) || ++ (filetype >= EXT3_FT_MAX)) ++ return DT_UNKNOWN; ++ ++ return (ext3_filetype_table[filetype]); ++} ++ ++ + int ext3_check_dir_entry (const char * function, struct inode * dir, + struct ext3_dir_entry_2 * de, + struct buffer_head * bh, +@@ -79,6 +94,16 @@ + + sb = inode->i_sb; + ++ if (is_dx(inode)) { ++ err = ext3_dx_readdir(filp, dirent, filldir); ++ if (err != ERR_BAD_DX_DIR) ++ return err; ++ /* ++ * We don't set the inode dirty flag since it's not ++ * critical that it get flushed back to the disk. ++ */ ++ EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; ++ } + stored = 0; + bh = NULL; + offset = filp->f_pos & (sb->s_blocksize - 1); +@@ -162,18 +187,12 @@ + * during the copy operation. + */ + unsigned long version = filp->f_version; +- unsigned char d_type = DT_UNKNOWN; + +- if (EXT3_HAS_INCOMPAT_FEATURE(sb, +- EXT3_FEATURE_INCOMPAT_FILETYPE) +- && de->file_type < EXT3_FT_MAX) +- d_type = +- ext3_filetype_table[de->file_type]; + error = filldir(dirent, de->name, + de->name_len, + filp->f_pos, + le32_to_cpu(de->inode), +- d_type); ++ get_dtype(sb, de->file_type)); + if (error) + break; + if (version != filp->f_version) +@@ -188,3 +207,272 @@ + UPDATE_ATIME(inode); + return 0; + } ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * These functions convert from the major/minor hash to an f_pos ++ * value. ++ * ++ * Currently we only use major hash numer. This is unfortunate, but ++ * on 32-bit machines, the same VFS interface is used for lseek and ++ * llseek, so if we use the 64 bit offset, then the 32-bit versions of ++ * lseek/telldir/seekdir will blow out spectacularly, and from within ++ * the ext2 low-level routine, we don't know if we're being called by ++ * a 64-bit version of the system call or the 32-bit version of the ++ * system call. Worse yet, NFSv2 only allows for a 32-bit readdir ++ * cookie. Sigh. ++ */ ++#define hash2pos(major, minor) (major >> 1) ++#define pos2maj_hash(pos) ((pos << 1) & 0xffffffff) ++#define pos2min_hash(pos) (0) ++ ++/* ++ * This structure holds the nodes of the red-black tree used to store ++ * the directory entry in hash order. ++ */ ++struct fname { ++ __u32 hash; ++ __u32 minor_hash; ++ rb_node_t rb_hash; ++ struct fname *next; ++ __u32 inode; ++ __u8 name_len; ++ __u8 file_type; ++ char name[0]; ++}; ++ ++/* ++ * This functoin implements a non-recursive way of freeing all of the ++ * nodes in the red-black tree. ++ */ ++static void free_rb_tree_fname(rb_root_t *root) ++{ ++ rb_node_t *n = root->rb_node; ++ rb_node_t *parent; ++ struct fname *fname; ++ ++ while (n) { ++ /* Do the node's children first */ ++ if ((n)->rb_left) { ++ n = n->rb_left; ++ continue; ++ } ++ if (n->rb_right) { ++ n = n->rb_right; ++ continue; ++ } ++ /* ++ * The node has no children; free it, and then zero ++ * out parent's link to it. Finally go to the ++ * beginning of the loop and try to free the parent ++ * node. ++ */ ++ parent = n->rb_parent; ++ fname = rb_entry(n, struct fname, rb_hash); ++ kfree(fname); ++ if (!parent) ++ root->rb_node = 0; ++ else if (parent->rb_left == n) ++ parent->rb_left = 0; ++ else if (parent->rb_right == n) ++ parent->rb_right = 0; ++ n = parent; ++ } ++ root->rb_node = 0; ++} ++ ++ ++struct dir_private_info *create_dir_info(loff_t pos) ++{ ++ struct dir_private_info *p; ++ ++ p = kmalloc(sizeof(struct dir_private_info), GFP_KERNEL); ++ if (!p) ++ return NULL; ++ p->root.rb_node = 0; ++ p->curr_node = 0; ++ p->extra_fname = 0; ++ p->last_pos = 0; ++ p->curr_hash = pos2maj_hash(pos); ++ p->curr_minor_hash = pos2min_hash(pos); ++ p->next_hash = 0; ++ return p; ++} ++ ++void ext3_htree_free_dir_info(struct dir_private_info *p) ++{ ++ free_rb_tree_fname(&p->root); ++ kfree(p); ++} ++ ++/* ++ * Given a directory entry, enter it into the fname rb tree. ++ */ ++int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent) ++{ ++ rb_node_t **p, *parent = NULL; ++ struct fname * fname, *new_fn; ++ struct dir_private_info *info; ++ int len; ++ ++ info = (struct dir_private_info *) dir_file->private_data; ++ p = &info->root.rb_node; ++ ++ /* Create and allocate the fname structure */ ++ len = sizeof(struct fname) + dirent->name_len + 1; ++ new_fn = kmalloc(len, GFP_KERNEL); ++ if (!new_fn) ++ return -ENOMEM; ++ memset(new_fn, 0, len); ++ new_fn->hash = hash; ++ new_fn->minor_hash = minor_hash; ++ new_fn->inode = le32_to_cpu(dirent->inode); ++ new_fn->name_len = dirent->name_len; ++ new_fn->file_type = dirent->file_type; ++ memcpy(new_fn->name, dirent->name, dirent->name_len); ++ new_fn->name[dirent->name_len] = 0; ++ ++ while (*p) { ++ parent = *p; ++ fname = rb_entry(parent, struct fname, rb_hash); ++ ++ /* ++ * If the hash and minor hash match up, then we put ++ * them on a linked list. This rarely happens... ++ */ ++ if ((new_fn->hash == fname->hash) && ++ (new_fn->minor_hash == fname->minor_hash)) { ++ new_fn->next = fname->next; ++ fname->next = new_fn; ++ return 0; ++ } ++ ++ if (new_fn->hash < fname->hash) ++ p = &(*p)->rb_left; ++ else if (new_fn->hash > fname->hash) ++ p = &(*p)->rb_right; ++ else if (new_fn->minor_hash < fname->minor_hash) ++ p = &(*p)->rb_left; ++ else /* if (new_fn->minor_hash > fname->minor_hash) */ ++ p = &(*p)->rb_right; ++ } ++ ++ rb_link_node(&new_fn->rb_hash, parent, p); ++ rb_insert_color(&new_fn->rb_hash, &info->root); ++ return 0; ++} ++ ++ ++ ++/* ++ * This is a helper function for ext3_dx_readdir. It calls filldir ++ * for all entres on the fname linked list. (Normally there is only ++ * one entry on the linked list, unless there are 62 bit hash collisions.) ++ */ ++static int call_filldir(struct file * filp, void * dirent, ++ filldir_t filldir, struct fname *fname) ++{ ++ struct dir_private_info *info = filp->private_data; ++ loff_t curr_pos; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct super_block * sb; ++ int error; ++ ++ sb = inode->i_sb; ++ ++ if (!fname) { ++ printk("call_filldir: called with null fname?!?\n"); ++ return 0; ++ } ++ curr_pos = hash2pos(fname->hash, fname->minor_hash); ++ while (fname) { ++ error = filldir(dirent, fname->name, ++ fname->name_len, curr_pos, ++ fname->inode, ++ get_dtype(sb, fname->file_type)); ++ if (error) { ++ filp->f_pos = curr_pos; ++ info->extra_fname = fname->next; ++ return error; ++ } ++ fname = fname->next; ++ } ++ return 0; ++} ++ ++static int ext3_dx_readdir(struct file * filp, ++ void * dirent, filldir_t filldir) ++{ ++ struct dir_private_info *info = filp->private_data; ++ struct inode *inode = filp->f_dentry->d_inode; ++ struct fname *fname; ++ int ret; ++ ++ if (!info) { ++ info = create_dir_info(filp->f_pos); ++ if (!info) ++ return -ENOMEM; ++ filp->private_data = info; ++ } ++ ++ /* Some one has messed with f_pos; reset the world */ ++ if (info->last_pos != filp->f_pos) { ++ free_rb_tree_fname(&info->root); ++ info->curr_node = 0; ++ info->extra_fname = 0; ++ info->curr_hash = pos2maj_hash(filp->f_pos); ++ info->curr_minor_hash = pos2min_hash(filp->f_pos); ++ } ++ ++ /* ++ * If there are any leftover names on the hash collision ++ * chain, return them first. ++ */ ++ if (info->extra_fname && ++ call_filldir(filp, dirent, filldir, info->extra_fname)) ++ goto finished; ++ ++ if (!info->curr_node) ++ info->curr_node = rb_get_first(&info->root); ++ ++ while (1) { ++ /* ++ * Fill the rbtree if we have no more entries, ++ * or the inode has changed since we last read in the ++ * cached entries. ++ */ ++ if ((!info->curr_node) || ++ (filp->f_version != inode->i_version)) { ++ info->curr_node = 0; ++ free_rb_tree_fname(&info->root); ++ filp->f_version = inode->i_version; ++ ret = ext3_htree_fill_tree(filp, info->curr_hash, ++ info->curr_minor_hash, ++ &info->next_hash); ++ if (ret < 0) ++ return ret; ++ if (ret == 0) ++ break; ++ info->curr_node = rb_get_first(&info->root); ++ } ++ ++ fname = rb_entry(info->curr_node, struct fname, rb_hash); ++ info->curr_hash = fname->hash; ++ info->curr_minor_hash = fname->minor_hash; ++ if (call_filldir(filp, dirent, filldir, fname)) ++ break; ++ ++ info->curr_node = rb_get_next(info->curr_node); ++ if (!info->curr_node) { ++ info->curr_hash = info->next_hash; ++ info->curr_minor_hash = 0; ++ } ++ } ++finished: ++ info->last_pos = filp->f_pos; ++ UPDATE_ATIME(inode); ++ return 0; ++} ++#endif +diff -Nru a/fs/ext3/file.c b/fs/ext3/file.c +--- a/fs/ext3/file.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/file.c Thu Nov 7 11:58:05 2002 +@@ -35,6 +35,9 @@ + { + if (filp->f_mode & FMODE_WRITE) + ext3_discard_prealloc (inode); ++ if (is_dx(inode) && filp->private_data) ++ ext3_htree_free_dir_info(filp->private_data); ++ + return 0; + } + +diff -Nru a/fs/ext3/hash.c b/fs/ext3/hash.c +--- /dev/null Wed Dec 31 16:00:00 1969 ++++ b/fs/ext3/hash.c Thu Nov 7 11:58:05 2002 +@@ -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; ++} +diff -Nru a/fs/ext3/ialloc.c b/fs/ext3/ialloc.c +--- a/fs/ext3/ialloc.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/ialloc.c Thu Nov 7 11:58:05 2002 +@@ -21,6 +21,7 @@ + #include + #include + #include ++#include + + #include + #include +@@ -293,6 +294,199 @@ + * the groups with above-average free space, that group with the fewest + * directories already is chosen. + * ++ * For other inodes, search forward from the parent directory\'s block ++ * group to find a free inode. ++ */ ++static int find_group_dir(struct super_block *sb, const struct inode *parent) ++{ ++ struct ext3_super_block * es = EXT3_SB(sb)->s_es; ++ int ngroups = EXT3_SB(sb)->s_groups_count; ++ int avefreei = le32_to_cpu(es->s_free_inodes_count) / ngroups; ++ struct ext3_group_desc *desc, *best_desc = NULL; ++ struct buffer_head *bh; ++ int group, best_group = -1; ++ ++ for (group = 0; group < ngroups; group++) { ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) ++ continue; ++ if (!best_desc || ++ (le16_to_cpu(desc->bg_free_blocks_count) > ++ le16_to_cpu(best_desc->bg_free_blocks_count))) { ++ best_group = group; ++ best_desc = desc; ++ } ++ } ++ return best_group; ++} ++ ++/* ++ * Orlov's allocator for directories. ++ * ++ * We always try to spread first-level directories. ++ * ++ * If there are blockgroups with both free inodes and free blocks counts ++ * not worse than average we return one with smallest directory count. ++ * Otherwise we simply return a random group. ++ * ++ * For the rest rules look so: ++ * ++ * It's OK to put directory into a group unless ++ * it has too many directories already (max_dirs) or ++ * it has too few free inodes left (min_inodes) or ++ * it has too few free blocks left (min_blocks) or ++ * it's already running too large debt (max_debt). ++ * Parent's group is prefered, if it doesn't satisfy these ++ * conditions we search cyclically through the rest. If none ++ * of the groups look good we just look for a group with more ++ * free inodes than average (starting at parent's group). ++ * ++ * Debt is incremented each time we allocate a directory and decremented ++ * when we allocate an inode, within 0--255. ++ */ ++ ++#define INODE_COST 64 ++#define BLOCK_COST 256 ++ ++static int find_group_orlov(struct super_block *sb, const struct inode *parent) ++{ ++ int parent_group = EXT3_I(parent)->i_block_group; ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ struct ext3_super_block *es = sbi->s_es; ++ int ngroups = sbi->s_groups_count; ++ int inodes_per_group = EXT3_INODES_PER_GROUP(sb); ++ int avefreei = le32_to_cpu(es->s_free_inodes_count) / ngroups; ++ int avefreeb = le32_to_cpu(es->s_free_blocks_count) / ngroups; ++ int blocks_per_dir; ++ int ndirs = sbi->s_dir_count; ++ int max_debt, max_dirs, min_blocks, min_inodes; ++ int group = -1, i; ++ struct ext3_group_desc *desc; ++ struct buffer_head *bh; ++ ++ if ((parent == sb->s_root->d_inode) || ++ (parent->i_flags & EXT3_TOPDIR_FL)) { ++ int best_ndir = inodes_per_group; ++ int best_group = -1; ++ ++ get_random_bytes(&group, sizeof(group)); ++ parent_group = (unsigned)group % ngroups; ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (le16_to_cpu(desc->bg_used_dirs_count) >= best_ndir) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) < avefreei) ++ continue; ++ if (le16_to_cpu(desc->bg_free_blocks_count) < avefreeb) ++ continue; ++ best_group = group; ++ best_ndir = le16_to_cpu(desc->bg_used_dirs_count); ++ } ++ if (best_group >= 0) ++ return best_group; ++ goto fallback; ++ } ++ ++ blocks_per_dir = (le32_to_cpu(es->s_blocks_count) - ++ le32_to_cpu(es->s_free_blocks_count)) / ndirs; ++ ++ max_dirs = ndirs / ngroups + inodes_per_group / 16; ++ min_inodes = avefreei - inodes_per_group / 4; ++ min_blocks = avefreeb - EXT3_BLOCKS_PER_GROUP(sb) / 4; ++ ++ max_debt = EXT3_BLOCKS_PER_GROUP(sb) / max(blocks_per_dir, BLOCK_COST); ++ if (max_debt * INODE_COST > inodes_per_group) ++ max_debt = inodes_per_group / INODE_COST; ++ if (max_debt > 255) ++ max_debt = 255; ++ if (max_debt == 0) ++ max_debt = 1; ++ ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (sbi->s_debts[group] >= max_debt) ++ continue; ++ if (le16_to_cpu(desc->bg_used_dirs_count) >= max_dirs) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) < min_inodes) ++ continue; ++ if (le16_to_cpu(desc->bg_free_blocks_count) < min_blocks) ++ continue; ++ return group; ++ } ++ ++fallback: ++ for (i = 0; i < ngroups; i++) { ++ group = (parent_group + i) % ngroups; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (!desc || !desc->bg_free_inodes_count) ++ continue; ++ if (le16_to_cpu(desc->bg_free_inodes_count) >= avefreei) ++ return group; ++ } ++ ++ return -1; ++} ++ ++static int find_group_other(struct super_block *sb, const struct inode *parent) ++{ ++ int parent_group = EXT3_I(parent)->i_block_group; ++ int ngroups = EXT3_SB(sb)->s_groups_count; ++ struct ext3_group_desc *desc; ++ struct buffer_head *bh; ++ int group, i; ++ ++ /* ++ * Try to place the inode in its parent directory ++ */ ++ group = parent_group; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (desc && le16_to_cpu(desc->bg_free_inodes_count)) ++ return group; ++ ++ /* ++ * Use a quadratic hash to find a group with a ++ * free inode ++ */ ++ for (i = 1; i < ngroups; i <<= 1) { ++ group += i; ++ if (group >= ngroups) ++ group -= ngroups; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (desc && le16_to_cpu(desc->bg_free_inodes_count)) ++ return group; ++ } ++ ++ /* ++ * That failed: try linear search for a free inode ++ */ ++ group = parent_group + 1; ++ for (i = 2; i < ngroups; i++) { ++ if (++group >= ngroups) ++ group = 0; ++ desc = ext3_get_group_desc (sb, group, &bh); ++ if (desc && le16_to_cpu(desc->bg_free_inodes_count)) ++ return group; ++ } ++ ++ return -1; ++} ++ ++/* ++ * There are two policies for allocating an inode. If the new inode is ++ * a directory, then a forward search is made for a block group with both ++ * free space and a low directory-to-inode ratio; if that fails, then of ++ * the groups with above-average free space, that group with the fewest ++ * directories already is chosen. ++ * + * For other inodes, search forward from the parent directory's block + * group to find a free inode. + */ +@@ -302,11 +496,11 @@ + struct super_block * sb; + struct buffer_head * bh; + struct buffer_head * bh2; +- int i, j, avefreei; ++ int group; ++ ino_t ino; + struct inode * inode; + int bitmap_nr; + struct ext3_group_desc * gdp; +- struct ext3_group_desc * tmp; + struct ext3_super_block * es; + int err = 0; + +@@ -323,94 +517,36 @@ + lock_super (sb); + es = sb->u.ext3_sb.s_es; + repeat: +- gdp = NULL; +- i = 0; +- + if (S_ISDIR(mode)) { +- avefreei = le32_to_cpu(es->s_free_inodes_count) / +- sb->u.ext3_sb.s_groups_count; +- if (!gdp) { +- for (j = 0; j < sb->u.ext3_sb.s_groups_count; j++) { +- struct buffer_head *temp_buffer; +- tmp = ext3_get_group_desc (sb, j, &temp_buffer); +- if (tmp && +- le16_to_cpu(tmp->bg_free_inodes_count) && +- le16_to_cpu(tmp->bg_free_inodes_count) >= +- avefreei) { +- if (!gdp || (le16_to_cpu(tmp->bg_free_blocks_count) > +- le16_to_cpu(gdp->bg_free_blocks_count))) { +- i = j; +- gdp = tmp; +- bh2 = temp_buffer; +- } +- } +- } +- } +- } else { +- /* +- * Try to place the inode in its parent directory +- */ +- i = dir->u.ext3_i.i_block_group; +- tmp = ext3_get_group_desc (sb, i, &bh2); +- if (tmp && le16_to_cpu(tmp->bg_free_inodes_count)) +- gdp = tmp; ++ if (test_opt (sb, OLDALLOC)) ++ group = find_group_dir(sb, dir); + else +- { +- /* +- * Use a quadratic hash to find a group with a +- * free inode +- */ +- for (j = 1; j < sb->u.ext3_sb.s_groups_count; j <<= 1) { +- i += j; +- if (i >= sb->u.ext3_sb.s_groups_count) +- i -= sb->u.ext3_sb.s_groups_count; +- tmp = ext3_get_group_desc (sb, i, &bh2); +- if (tmp && +- le16_to_cpu(tmp->bg_free_inodes_count)) { +- gdp = tmp; +- break; +- } +- } +- } +- if (!gdp) { +- /* +- * That failed: try linear search for a free inode +- */ +- i = dir->u.ext3_i.i_block_group + 1; +- for (j = 2; j < sb->u.ext3_sb.s_groups_count; j++) { +- if (++i >= sb->u.ext3_sb.s_groups_count) +- i = 0; +- tmp = ext3_get_group_desc (sb, i, &bh2); +- if (tmp && +- le16_to_cpu(tmp->bg_free_inodes_count)) { +- gdp = tmp; +- break; +- } +- } +- } +- } +- ++ group = find_group_orlov(sb, dir); ++ } else ++ group = find_group_other(sb, dir); ++ + err = -ENOSPC; +- if (!gdp) ++ if (!group == -1) + goto out; + + err = -EIO; +- bitmap_nr = load_inode_bitmap (sb, i); ++ bitmap_nr = load_inode_bitmap (sb, group); + if (bitmap_nr < 0) + goto fail; + + bh = sb->u.ext3_sb.s_inode_bitmap[bitmap_nr]; ++ gdp = ext3_get_group_desc (sb, group, &bh2); + +- if ((j = ext3_find_first_zero_bit ((unsigned long *) bh->b_data, ++ if ((ino = ext3_find_first_zero_bit ((unsigned long *) bh->b_data, + EXT3_INODES_PER_GROUP(sb))) < + EXT3_INODES_PER_GROUP(sb)) { + 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)) { ++ if (ext3_set_bit (ino, bh->b_data)) { + ext3_error (sb, "ext3_new_inode", +- "bit already set for inode %d", j); ++ "bit already set for inode %lu", ino); + goto repeat; + } + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +@@ -420,7 +556,7 @@ + if (le16_to_cpu(gdp->bg_free_inodes_count) != 0) { + ext3_error (sb, "ext3_new_inode", + "Free inodes count corrupted in group %d", +- i); ++ group); + /* Is it really ENOSPC? */ + err = -ENOSPC; + if (sb->s_flags & MS_RDONLY) +@@ -436,11 +572,11 @@ + } + goto repeat; + } +- j += i * EXT3_INODES_PER_GROUP(sb) + 1; +- if (j < EXT3_FIRST_INO(sb) || j > le32_to_cpu(es->s_inodes_count)) { ++ ino += group * EXT3_INODES_PER_GROUP(sb) + 1; ++ if (ino < EXT3_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { + ext3_error (sb, "ext3_new_inode", + "reserved inode or inode > inodes count - " +- "block_group = %d,inode=%d", i, j); ++ "block_group = %d, inode=%lu", group, ino); + err = -EIO; + goto fail; + } +@@ -450,9 +586,11 @@ + if (err) goto fail; + gdp->bg_free_inodes_count = + cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1); +- if (S_ISDIR(mode)) ++ if (S_ISDIR(mode)) { + gdp->bg_used_dirs_count = + cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1); ++ EXT3_SB(sb)->s_dir_count++; ++ } + BUFFER_TRACE(bh2, "call ext3_journal_dirty_metadata"); + err = ext3_journal_dirty_metadata(handle, bh2); + if (err) goto fail; +@@ -478,7 +616,7 @@ + inode->i_gid = current->fsgid; + inode->i_mode = mode; + +- inode->i_ino = j; ++ inode->i_ino = ino; + /* This is the optimal IO size (for stat), not the fs block size */ + inode->i_blksize = PAGE_SIZE; + inode->i_blocks = 0; +@@ -498,7 +636,7 @@ + #ifdef EXT3_PREALLOCATE + inode->u.ext3_i.i_prealloc_count = 0; + #endif +- inode->u.ext3_i.i_block_group = i; ++ inode->u.ext3_i.i_block_group = group; + + if (inode->u.ext3_i.i_flags & EXT3_SYNC_FL) + inode->i_flags |= S_SYNC; +@@ -620,6 +758,21 @@ + #else + return le32_to_cpu(sb->u.ext3_sb.s_es->s_free_inodes_count); + #endif ++} ++ ++/* Called at mount-time, super-block is locked */ ++unsigned long ext3_count_dirs (struct super_block * sb) ++{ ++ unsigned long count = 0; ++ int i; ++ ++ for (i = 0; i < EXT3_SB(sb)->s_groups_count; i++) { ++ struct ext3_group_desc *gdp = ext3_get_group_desc (sb, i, NULL); ++ if (!gdp) ++ continue; ++ count += le16_to_cpu(gdp->bg_used_dirs_count); ++ } ++ return count; + } + + #ifdef CONFIG_EXT3_CHECK +diff -Nru a/fs/ext3/inode.c b/fs/ext3/inode.c +--- a/fs/ext3/inode.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/inode.c Thu Nov 7 11:58:05 2002 +@@ -2194,6 +2194,11 @@ + if (err) + goto out_brelse; + } ++ /* For fields not not tracking in the in-memory inode, ++ * initialise them to zero for new inodes. */ ++ if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) ++ memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size); ++ + raw_inode->i_mode = cpu_to_le16(inode->i_mode); + if(!(test_opt(inode->i_sb, NO_UID32))) { + raw_inode->i_uid_low = cpu_to_le16(low_16_bits(inode->i_uid)); +@@ -2231,15 +2236,6 @@ + raw_inode->i_faddr = cpu_to_le32(inode->u.ext3_i.i_faddr); + raw_inode->i_frag = inode->u.ext3_i.i_frag_no; + raw_inode->i_fsize = inode->u.ext3_i.i_frag_size; +-#else +- /* If we are not tracking these fields in the in-memory inode, +- * then preserve them on disk, but still initialise them to zero +- * for new inodes. */ +- if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) { +- raw_inode->i_faddr = 0; +- raw_inode->i_frag = 0; +- raw_inode->i_fsize = 0; +- } + #endif + raw_inode->i_file_acl = cpu_to_le32(inode->u.ext3_i.i_file_acl); + if (!S_ISREG(inode->i_mode)) { +diff -Nru a/fs/ext3/namei.c b/fs/ext3/namei.c +--- a/fs/ext3/namei.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/namei.c Thu Nov 7 11:58:05 2002 +@@ -16,6 +16,12 @@ + * David S. Miller (davem@caip.rutgers.edu), 1995 + * Directory entry file type support and forward compatibility hooks + * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 ++ * Hash Tree Directory indexing (c) ++ * Daniel Phillips, 2001 ++ * Hash Tree Directory indexing porting ++ * Christopher Li, 2002 ++ * Hash Tree Directory indexing cleanup ++ * Theodore Ts'o, 2002 + */ + + #include +@@ -38,6 +44,642 @@ + #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) + #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) + ++static struct buffer_head *ext3_append(handle_t *handle, ++ struct inode *inode, ++ u32 *block, int *err) ++{ ++ struct buffer_head *bh; ++ ++ *block = inode->i_size >> inode->i_sb->s_blocksize_bits; ++ ++ if ((bh = ext3_bread(handle, inode, *block, 1, err))) { ++ inode->i_size += inode->i_sb->s_blocksize; ++ EXT3_I(inode)->i_disksize = inode->i_size; ++ ext3_journal_get_write_access(handle,bh); ++ } ++ return bh; ++} ++ ++#ifndef assert ++#define assert(test) J_ASSERT(test) ++#endif ++ ++#ifndef swap ++#define swap(x, y) do { typeof(x) z = x; x = y; y = z; } while (0) ++#endif ++ ++typedef struct { u32 v; } le_u32; ++typedef struct { u16 v; } le_u16; ++ ++#ifdef DX_DEBUG ++#define dxtrace(command) command ++#else ++#define dxtrace(command) ++#endif ++ ++struct fake_dirent ++{ ++ /*le*/u32 inode; ++ /*le*/u16 rec_len; ++ u8 name_len; ++ u8 file_type; ++}; ++ ++struct dx_countlimit ++{ ++ le_u16 limit; ++ le_u16 count; ++}; ++ ++struct dx_entry ++{ ++ le_u32 hash; ++ le_u32 block; ++}; ++ ++/* ++ * dx_root_info is laid out so that if it should somehow get overlaid by a ++ * dirent the two low bits of the hash version will be zero. Therefore, the ++ * hash version mod 4 should never be 0. Sincerely, the paranoia department. ++ */ ++ ++struct dx_root ++{ ++ struct fake_dirent dot; ++ char dot_name[4]; ++ struct fake_dirent dotdot; ++ char dotdot_name[4]; ++ struct dx_root_info ++ { ++ le_u32 reserved_zero; ++ u8 hash_version; ++ u8 info_length; /* 8 */ ++ u8 indirect_levels; ++ u8 unused_flags; ++ } ++ info; ++ struct dx_entry entries[0]; ++}; ++ ++struct dx_node ++{ ++ struct fake_dirent fake; ++ struct dx_entry entries[0]; ++}; ++ ++ ++struct dx_frame ++{ ++ struct buffer_head *bh; ++ struct dx_entry *entries; ++ struct dx_entry *at; ++}; ++ ++struct dx_map_entry ++{ ++ u32 hash; ++ u32 offs; ++}; ++ ++#ifdef CONFIG_EXT3_INDEX ++static inline unsigned dx_get_block (struct dx_entry *entry); ++static void dx_set_block (struct dx_entry *entry, unsigned value); ++static inline unsigned dx_get_hash (struct dx_entry *entry); ++static void dx_set_hash (struct dx_entry *entry, unsigned value); ++static unsigned dx_get_count (struct dx_entry *entries); ++static unsigned dx_get_limit (struct dx_entry *entries); ++static void dx_set_count (struct dx_entry *entries, unsigned value); ++static void dx_set_limit (struct dx_entry *entries, unsigned value); ++static unsigned dx_root_limit (struct inode *dir, unsigned infosize); ++static unsigned dx_node_limit (struct inode *dir); ++static struct dx_frame *dx_probe(struct dentry *dentry, ++ struct inode *dir, ++ struct dx_hash_info *hinfo, ++ struct dx_frame *frame, ++ int *err); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static struct ext3_dir_entry_2 *dx_move_dirents (char *from, char *to, ++ struct dx_map_entry *offsets, int count); ++static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash); ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err); ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode); ++ ++/* ++ * Future: use high four bits of block for coalesce-on-delete flags ++ * Mask them off for now. ++ */ ++ ++static inline unsigned dx_get_block (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->block.v) & 0x00ffffff; ++} ++ ++static inline void dx_set_block (struct dx_entry *entry, unsigned value) ++{ ++ entry->block.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_hash (struct dx_entry *entry) ++{ ++ return le32_to_cpu(entry->hash.v); ++} ++ ++static inline void dx_set_hash (struct dx_entry *entry, unsigned value) ++{ ++ entry->hash.v = cpu_to_le32(value); ++} ++ ++static inline unsigned dx_get_count (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->count.v); ++} ++ ++static inline unsigned dx_get_limit (struct dx_entry *entries) ++{ ++ return le16_to_cpu(((struct dx_countlimit *) entries)->limit.v); ++} ++ ++static inline void dx_set_count (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->count.v = cpu_to_le16(value); ++} ++ ++static inline void dx_set_limit (struct dx_entry *entries, unsigned value) ++{ ++ ((struct dx_countlimit *) entries)->limit.v = cpu_to_le16(value); ++} ++ ++static inline unsigned dx_root_limit (struct inode *dir, unsigned infosize) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(1) - ++ EXT3_DIR_REC_LEN(2) - infosize; ++ return 0? 20: entry_space / sizeof(struct dx_entry); ++} ++ ++static inline unsigned dx_node_limit (struct inode *dir) ++{ ++ unsigned entry_space = dir->i_sb->s_blocksize - EXT3_DIR_REC_LEN(0); ++ return 0? 22: entry_space / sizeof(struct dx_entry); ++} ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext3_dir_entry_2 *de, ++ int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ printk("names: "); ++ while ((char *) de < base + size) ++ { ++ if (de->inode) ++ { ++ if (show_names) ++ { ++ int len = de->name_len; ++ char *name = de->name; ++ while (len--) printk("%c", *name++); ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ printk(":%x.%u ", h.hash, ++ ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, ++ struct dx_entry *entries, int levels) ++{ ++ unsigned blocksize = dir->i_sb->s_blocksize; ++ unsigned count = dx_get_count (entries), names = 0, space = 0, i; ++ unsigned bcount = 0; ++ struct buffer_head *bh; ++ int err; ++ printk("%i indexed blocks...\n", count); ++ for (i = 0; i < count; i++, entries++) ++ { ++ u32 block = dx_get_block(entries), hash = i? dx_get_hash(entries): 0; ++ u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; ++ struct stats stats; ++ printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); ++ if (!(bh = ext3_bread (NULL,dir, block, 0,&err))) continue; ++ stats = levels? ++ dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf(hinfo, (struct ext3_dir_entry_2 *) bh->b_data, blocksize, 0); ++ names += stats.names; ++ space += stats.space; ++ bcount += stats.bcount; ++ brelse (bh); ++ } ++ if (bcount) ++ printk("%snames %u, fullness %u (%u%%)\n", levels?"":" ", ++ names, space/bcount,(space/bcount)*100/blocksize); ++ return (struct stats) { names, space, bcount}; ++} ++#endif /* DX_DEBUG */ ++ ++/* ++ * Probe for a directory leaf block to search. ++ * ++ * dx_probe can return ERR_BAD_DX_DIR, which means there was a format ++ * error in the directory index, and the caller should fall back to ++ * searching the directory normally. The callers of dx_probe **MUST** ++ * check for this error code, and make sure it never gets reflected ++ * back to userspace. ++ */ ++static struct dx_frame * ++dx_probe(struct dentry *dentry, struct inode *dir, ++ struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) ++{ ++ unsigned count, indirect; ++ struct dx_entry *at, *entries, *p, *q, *m; ++ struct dx_root *root; ++ struct buffer_head *bh; ++ struct dx_frame *frame = frame_in; ++ u32 hash; ++ ++ frame->bh = NULL; ++ if (dentry) ++ dir = dentry->d_parent->d_inode; ++ if (!(bh = ext3_bread (NULL,dir, 0, 0, err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version != DX_HASH_TEA && ++ root->info.hash_version != DX_HASH_HALF_MD4 && ++ root->info.hash_version != DX_HASH_LEGACY) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unrecognised inode hash code %d", ++ root->info.hash_version); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ hinfo->hash_version = root->info.hash_version; ++ hinfo->seed = dir->i_sb->u.ext3_sb.s_hash_seed; ++ if (dentry) ++ ext3fs_dirhash(dentry->d_name.name, dentry->d_name.len, hinfo); ++ hash = hinfo->hash; ++ ++ if (root->info.unused_flags & 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash flags: %#06x", ++ root->info.unused_flags); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ if ((indirect = root->info.indirect_levels) > 1) { ++ ext3_warning(dir->i_sb, __FUNCTION__, ++ "Unimplemented inode hash depth: %#06x", ++ root->info.indirect_levels); ++ brelse(bh); ++ *err = ERR_BAD_DX_DIR; ++ goto fail; ++ } ++ ++ entries = (struct dx_entry *) (((char *)&root->info) + ++ root->info.info_length); ++ assert(dx_get_limit(entries) == dx_root_limit(dir, ++ root->info.info_length)); ++ dxtrace (printk("Look up %x", hash)); ++ while (1) ++ { ++ count = dx_get_count(entries); ++ assert (count && count <= dx_get_limit(entries)); ++ p = entries + 1; ++ q = entries + count - 1; ++ while (p <= q) ++ { ++ m = p + (q - p)/2; ++ dxtrace(printk(".")); ++ if (dx_get_hash(m) > hash) ++ q = m - 1; ++ else ++ p = m + 1; ++ } ++ ++ if (0) // linear search cross check ++ { ++ unsigned n = count - 1; ++ at = entries; ++ while (n--) ++ { ++ dxtrace(printk(",")); ++ if (dx_get_hash(++at) > hash) ++ { ++ at--; ++ break; ++ } ++ } ++ assert (at == p - 1); ++ } ++ ++ at = p - 1; ++ dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); ++ frame->bh = bh; ++ frame->entries = entries; ++ frame->at = at; ++ if (!indirect--) return frame; ++ if (!(bh = ext3_bread (NULL,dir, dx_get_block(at), 0, err))) ++ goto fail2; ++ at = entries = ((struct dx_node *) bh->b_data)->entries; ++ assert (dx_get_limit(entries) == dx_node_limit (dir)); ++ frame++; ++ } ++fail2: ++ while (frame >= frame_in) { ++ brelse(frame->bh); ++ frame--; ++ } ++fail: ++ return NULL; ++} ++ ++static void dx_release (struct dx_frame *frames) ++{ ++ if (frames[0].bh == NULL) ++ return; ++ ++ if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) ++ brelse(frames[1].bh); ++ brelse(frames[0].bh); ++} ++ ++/* ++ * This function increments the frame pointer to search the next leaf ++ * block, and reads in the necessary intervening nodes if the search ++ * should be necessary. Whether or not the search is necessary is ++ * controlled by the hash parameter. If the hash value is even, then ++ * the search is only continued if the next block starts with that ++ * hash value. This is used if we are searching for a specific file. ++ * ++ * If the hash value is HASH_NB_ALWAYS, then always go to the next block. ++ * ++ * This function returns 1 if the caller should continue to search, ++ * or 0 if it should not. If there is an error reading one of the ++ * index blocks, it will return -1. ++ * ++ * If start_hash is non-null, it will be filled in with the starting ++ * hash of the next page. ++ */ ++static int ext3_htree_next_block(struct inode *dir, __u32 hash, ++ struct dx_frame *frame, ++ struct dx_frame *frames, int *err, ++ __u32 *start_hash) ++{ ++ struct dx_frame *p; ++ struct buffer_head *bh; ++ int num_frames = 0; ++ __u32 bhash; ++ ++ *err = ENOENT; ++ p = frame; ++ /* ++ * Find the next leaf page by incrementing the frame pointer. ++ * If we run out of entries in the interior node, loop around and ++ * increment pointer in the parent node. When we break out of ++ * this loop, num_frames indicates the number of interior ++ * nodes need to be read. ++ */ ++ while (1) { ++ if (++(p->at) < p->entries + dx_get_count(p->entries)) ++ break; ++ if (p == frames) ++ return 0; ++ num_frames++; ++ p--; ++ } ++ ++ /* ++ * If the hash is 1, then continue only if the next page has a ++ * continuation hash of any value. This is used for readdir ++ * handling. Otherwise, check to see if the hash matches the ++ * desired contiuation hash. If it doesn't, return since ++ * there's no point to read in the successive index pages. ++ */ ++ bhash = dx_get_hash(p->at); ++ if (start_hash) ++ *start_hash = bhash; ++ if ((hash & 1) == 0) { ++ if ((bhash & ~1) != hash) ++ return 0; ++ } ++ /* ++ * If the hash is HASH_NB_ALWAYS, we always go to the next ++ * block so no check is necessary ++ */ ++ while (num_frames--) { ++ if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), ++ 0, err))) ++ return -1; /* Failure */ ++ p++; ++ brelse (p->bh); ++ p->bh = bh; ++ p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; ++ } ++ return 1; ++} ++ ++ ++/* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *p) ++{ ++ return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++/* ++ * This function fills a red-black tree with information from a ++ * directory. We start scanning the directory in hash order, starting ++ * at start_hash and start_minor_hash. ++ * ++ * This function returns the number of entries inserted into the tree, ++ * or a negative error code. ++ */ ++int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash) ++{ ++ struct dx_hash_info hinfo; ++ struct buffer_head *bh; ++ struct ext3_dir_entry_2 *de, *top; ++ static struct dx_frame frames[2], *frame; ++ struct inode *dir; ++ int block, err; ++ int count = 0; ++ int ret; ++ __u32 hashval; ++ ++ dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, ++ start_minor_hash)); ++ dir = dir_file->f_dentry->d_inode; ++ hinfo.hash = start_hash; ++ hinfo.minor_hash = 0; ++ frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ ++ /* Add '.' and '..' from the htree header */ ++ if (!start_hash && !start_minor_hash) { ++ de = (struct ext3_dir_entry_2 *) frames[0].bh->b_data; ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ de = ext3_next_entry(de); ++ if ((err = ext3_htree_store_dirent(dir_file, 0, 0, de)) != 0) ++ goto errout; ++ count += 2; ++ } ++ ++ while (1) { ++ block = dx_get_block(frame->at); ++ dxtrace(printk("Reading block %d\n", block)); ++ if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) ++ goto errout; ++ ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + dir->i_sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) { ++ ext3fs_dirhash(de->name, de->name_len, &hinfo); ++ if ((hinfo.hash < start_hash) || ++ ((hinfo.hash == start_hash) && ++ (hinfo.minor_hash < start_minor_hash))) ++ continue; ++ if ((err = ext3_htree_store_dirent(dir_file, ++ hinfo.hash, hinfo.minor_hash, de)) != 0) ++ goto errout; ++ count++; ++ } ++ brelse (bh); ++ hashval = ~1; ++ ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ++ frame, frames, &err, &hashval); ++ if (next_hash) ++ *next_hash = hashval; ++ if (ret == -1) ++ goto errout; ++ /* ++ * Stop if: (a) there are no more entries, or ++ * (b) we have inserted at least one entry and the ++ * next hash value is not a continuation ++ */ ++ if ((ret == 0) || ++ (count && ((hashval & 1) == 0))) ++ break; ++ } ++ dx_release(frames); ++ dxtrace(printk("Fill tree: returned %d entries\n", count)); ++ return count; ++errout: ++ dx_release(frames); ++ return (err); ++} ++ ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (struct ext3_dir_entry_2 *de, int size, ++ struct dx_hash_info *hinfo, struct dx_map_entry *map_tail) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ struct dx_hash_info h = *hinfo; ++ ++ while ((char *) de < base + size) ++ { ++ if (de->name_len && de->inode) { ++ ext3fs_dirhash(de->name, de->name_len, &h); ++ map_tail--; ++ map_tail->hash = h.hash; ++ map_tail->offs = (u32) ((char *) de - base); ++ count++; ++ } ++ /* XXX: do we need to check rec_len == 0 case? -Chris */ ++ de = (struct ext3_dir_entry_2 *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ return count; ++} ++ ++static void dx_sort_map (struct dx_map_entry *map, unsigned count) ++{ ++ struct dx_map_entry *p, *q, *top = map + count - 1; ++ int more; ++ /* Combsort until bubble sort doesn't suck */ ++ while (count > 2) ++ { ++ count = count*10/13; ++ if (count - 9 < 2) /* 9, 10 -> 11 */ ++ count = 11; ++ for (p = top, q = p - count; q >= map; p--, q--) ++ if (p->hash < q->hash) ++ swap(*p, *q); ++ } ++ /* Garden variety bubble sort */ ++ do { ++ more = 0; ++ q = top; ++ while (q-- > map) ++ { ++ if (q[1].hash >= q[0].hash) ++ continue; ++ swap(*(q+1), *q); ++ more = 1; ++ } ++ } while(more); ++} ++ ++static void dx_insert_block(struct dx_frame *frame, u32 hash, u32 block) ++{ ++ struct dx_entry *entries = frame->entries; ++ struct dx_entry *old = frame->at, *new = old + 1; ++ int count = dx_get_count(entries); ++ ++ assert(count < dx_get_limit(entries)); ++ assert(old < entries + count); ++ memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); ++ dx_set_hash(new, hash); ++ dx_set_block(new, block); ++ dx_set_count(entries, count + 1); ++} ++#endif ++ ++ ++static void ext3_update_dx_flag(struct inode *inode) ++{ ++ if (!EXT3_HAS_COMPAT_FEATURE(inode->i_sb, ++ EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -94,6 +736,7 @@ + return 0; + } + ++ + /* + * ext3_find_entry() + * +@@ -105,6 +748,8 @@ + * The returned buffer_head has ->b_count elevated. The caller is expected + * to brelse() it when appropriate. + */ ++ ++ + static struct buffer_head * ext3_find_entry (struct dentry *dentry, + struct ext3_dir_entry_2 ** res_dir) + { +@@ -119,12 +764,32 @@ + int num = 0; + int nblocks, i, err; + struct inode *dir = dentry->d_parent->d_inode; ++ int namelen; ++ const u8 *name; ++ unsigned blocksize; + + *res_dir = NULL; + sb = dir->i_sb; +- ++ blocksize = sb->s_blocksize; ++ namelen = dentry->d_name.len; ++ name = dentry->d_name.name; ++ if (namelen > EXT3_NAME_LEN) ++ return NULL; ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ bh = ext3_dx_find_entry(dentry, res_dir, &err); ++ /* ++ * On success, or if the error was file not found, ++ * return. Otherwise, fall back to doing a search the ++ * old fashioned way. ++ */ ++ if (bh || (err != ERR_BAD_DX_DIR)) ++ return bh; ++ dxtrace(printk("ext3_find_entry: dx failed, falling back\n")); ++ } ++#endif + nblocks = dir->i_size >> EXT3_BLOCK_SIZE_BITS(sb); +- start = dir->u.ext3_i.i_dir_start_lookup; ++ start = EXT3_I(dir)->i_dir_start_lookup; + if (start >= nblocks) + start = 0; + block = start; +@@ -165,7 +830,7 @@ + i = search_dirblock(bh, dir, dentry, + block << EXT3_BLOCK_SIZE_BITS(sb), res_dir); + if (i == 1) { +- dir->u.ext3_i.i_dir_start_lookup = block; ++ EXT3_I(dir)->i_dir_start_lookup = block; + ret = bh; + goto cleanup_and_exit; + } else { +@@ -196,6 +861,66 @@ + return ret; + } + ++#ifdef CONFIG_EXT3_INDEX ++static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ++ struct ext3_dir_entry_2 **res_dir, int *err) ++{ ++ struct super_block * sb; ++ struct dx_hash_info hinfo; ++ u32 hash; ++ struct dx_frame frames[2], *frame; ++ struct ext3_dir_entry_2 *de, *top; ++ struct buffer_head *bh; ++ unsigned long block; ++ int retval; ++ int namelen = dentry->d_name.len; ++ const u8 *name = dentry->d_name.name; ++ struct inode *dir = dentry->d_parent->d_inode; ++ ++ sb = dir->i_sb; ++ if (!(frame = dx_probe (dentry, 0, &hinfo, frames, err))) ++ return NULL; ++ hash = hinfo.hash; ++ do { ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, err))) ++ goto errout; ++ de = (struct ext3_dir_entry_2 *) bh->b_data; ++ top = (struct ext3_dir_entry_2 *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match (namelen, name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto errout; ++ } ++ *res_dir = de; ++ dx_release (frames); ++ return bh; ++ } ++ brelse (bh); ++ /* Check to see if we should continue to search */ ++ retval = ext3_htree_next_block(dir, hash, frame, ++ frames, err, 0); ++ if (retval == -1) { ++ ext3_warning(sb, __FUNCTION__, ++ "error reading index page in directory #%lu", ++ dir->i_ino); ++ goto errout; ++ } ++ } while (retval == 1); ++ ++ *err = -ENOENT; ++errout: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++} ++#endif ++ + static struct dentry *ext3_lookup(struct inode * dir, struct dentry *dentry) + { + struct inode * inode; +@@ -212,8 +937,9 @@ + brelse (bh); + inode = iget(dir->i_sb, ino); + +- if (!inode) ++ if (!inode) { + return ERR_PTR(-EACCES); ++ } + } + d_add(dentry, inode); + return NULL; +@@ -237,6 +963,300 @@ + 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->info; ++ 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 +1267,198 @@ + * may not sleep between calling this and putting something into + * the entry, as someone else might have used it while you slept. + */ +- +-/* +- * AKPM: the journalling code here looks wrong on the error paths +- */ + static int ext3_add_entry (handle_t *handle, struct dentry *dentry, + struct inode *inode) + { + struct inode *dir = dentry->d_parent->d_inode; +- const char *name = dentry->d_name.name; +- int namelen = dentry->d_name.len; + unsigned long offset; +- unsigned short rec_len; + struct buffer_head * bh; +- struct ext3_dir_entry_2 * de, * de1; ++ struct ext3_dir_entry_2 *de; + struct super_block * sb; + int retval; ++#ifdef CONFIG_EXT3_INDEX ++ int dx_fallback=0; ++#endif ++ unsigned blocksize; ++ unsigned nlen, rlen; ++ u32 block, blocks; + + sb = dir->i_sb; +- +- if (!namelen) ++ blocksize = sb->s_blocksize; ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &retval); ++#ifdef CONFIG_EXT3_INDEX ++ if (is_dx(dir)) { ++ retval = ext3_dx_add_entry(handle, dentry, inode); ++ if (!retval || (retval != ERR_BAD_DX_DIR)) ++ return retval; ++ EXT3_I(dir)->i_flags &= ~EXT3_INDEX_FL; ++ dx_fallback++; ++ ext3_mark_inode_dirty(handle, dir); ++ } ++#endif ++ blocks = dir->i_size >> sb->s_blocksize_bits; ++ for (block = 0, offset = 0; block < blocks; block++) { ++ bh = ext3_bread(handle, dir, block, 0, &retval); ++ if(!bh) ++ return retval; ++ retval = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (retval != -ENOSPC) ++ return retval; ++ ++#ifdef CONFIG_EXT3_INDEX ++ if (blocks == 1 && !dx_fallback && ++ EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ return make_indexed_dir(handle, dentry, inode, bh); ++#endif ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); + if (!bh) + return retval; +- rec_len = EXT3_DIR_REC_LEN(namelen); +- offset = 0; + de = (struct ext3_dir_entry_2 *) bh->b_data; +- while (1) { +- if ((char *)de >= sb->s_blocksize + bh->b_data) { +- brelse (bh); +- bh = NULL; +- bh = ext3_bread (handle, dir, +- offset >> EXT3_BLOCK_SIZE_BITS(sb), 1, &retval); +- if (!bh) +- return retval; +- if (dir->i_size <= offset) { +- if (dir->i_size == 0) { +- brelse(bh); +- return -ENOENT; +- } ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = blocksize); ++ nlen = 0; ++ return add_dirent_to_buf(handle, dentry, inode, de, bh); ++} + +- ext3_debug ("creating next block\n"); ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * Returns 0 for success, or a negative error value ++ */ ++static int ext3_dx_add_entry(handle_t *handle, struct dentry *dentry, ++ struct inode *inode) ++{ ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ struct dx_hash_info hinfo; ++ struct buffer_head * bh; ++ struct inode *dir = dentry->d_parent->d_inode; ++ struct super_block * sb = dir->i_sb; ++ struct ext3_dir_entry_2 *de; ++ int err; + +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); +- } else { ++ frame = dx_probe(dentry, 0, &hinfo, frames, &err); ++ if (!frame) ++ return err; ++ entries = frame->entries; ++ at = frame->at; + +- ext3_debug ("skipping to next block\n"); ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(frame->at), 0, &err))) ++ goto cleanup; + +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- } +- } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { +- brelse (bh); +- return -EEXIST; ++ BUFFER_TRACE(bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, bh); ++ if (err) ++ goto journal_error; ++ ++ err = add_dirent_to_buf(handle, dentry, inode, 0, bh); ++ if (err != -ENOSPC) { ++ bh = 0; ++ goto cleanup; ++ } ++ ++ /* Block full, should compress but for now just split */ ++ dxtrace(printk("using %u of %u node entries\n", ++ dx_get_count(entries), dx_get_limit(entries))); ++ /* Need to split index? */ ++ if (dx_get_count(entries) == dx_get_limit(entries)) { ++ u32 newblock; ++ unsigned icount = dx_get_count(entries); ++ int levels = frame - frames; ++ struct dx_entry *entries2; ++ struct dx_node *node2; ++ struct buffer_head *bh2; ++ ++ if (levels && (dx_get_count(frames->entries) == ++ dx_get_limit(frames->entries))) { ++ ext3_warning(sb, __FUNCTION__, ++ "Directory index full!\n"); ++ err = -ENOSPC; ++ goto cleanup; + } +- if ((le32_to_cpu(de->inode) == 0 && +- le16_to_cpu(de->rec_len) >= rec_len) || +- (le16_to_cpu(de->rec_len) >= +- EXT3_DIR_REC_LEN(de->name_len) + rec_len)) { +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- /* By now the buffer is marked for journaling */ +- offset += le16_to_cpu(de->rec_len); +- if (le32_to_cpu(de->inode)) { +- de1 = (struct ext3_dir_entry_2 *) ((char *) de + +- EXT3_DIR_REC_LEN(de->name_len)); +- de1->rec_len = +- cpu_to_le16(le16_to_cpu(de->rec_len) - +- EXT3_DIR_REC_LEN(de->name_len)); +- de->rec_len = cpu_to_le16( +- EXT3_DIR_REC_LEN(de->name_len)); +- de = de1; ++ bh2 = ext3_append (handle, dir, &newblock, &err); ++ if (!(bh2)) ++ goto cleanup; ++ node2 = (struct dx_node *)(bh2->b_data); ++ entries2 = node2->entries; ++ node2->fake.rec_len = cpu_to_le16(sb->s_blocksize); ++ node2->fake.inode = 0; ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ err = ext3_journal_get_write_access(handle, frame->bh); ++ if (err) ++ goto journal_error; ++ if (levels) { ++ unsigned icount1 = icount/2, icount2 = icount - icount1; ++ unsigned hash2 = dx_get_hash(entries + icount1); ++ dxtrace(printk("Split index %i/%i\n", icount1, icount2)); ++ ++ BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ ++ err = ext3_journal_get_write_access(handle, ++ frames[0].bh); ++ if (err) ++ goto journal_error; ++ ++ memcpy ((char *) entries2, (char *) (entries + icount1), ++ icount2 * sizeof(struct dx_entry)); ++ dx_set_count (entries, icount1); ++ dx_set_count (entries2, icount2); ++ dx_set_limit (entries2, dx_node_limit(dir)); ++ ++ /* Which index block gets the new entry? */ ++ if (at - entries >= icount1) { ++ frame->at = at = at - entries - icount1 + entries2; ++ frame->entries = entries = entries2; ++ swap(frame->bh, bh2); + } +- de->file_type = EXT3_FT_UNKNOWN; +- if (inode) { +- de->inode = cpu_to_le32(inode->i_ino); +- ext3_set_de_type(dir->i_sb, de, inode->i_mode); +- } else +- de->inode = 0; +- de->name_len = namelen; +- memcpy (de->name, name, namelen); +- /* +- * XXX shouldn't update any times until successful +- * completion of syscall, but too many callers depend +- * on this. +- * +- * XXX similarly, too many callers depend on +- * ext3_new_inode() setting the times, but error +- * recovery deletes the inode, so the worst that can +- * happen is that the times are slightly out of date +- * and/or different from the directory change time. +- */ +- dir->i_mtime = dir->i_ctime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- dir->i_version = ++event; +- ext3_mark_inode_dirty(handle, dir); +- BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); +- ext3_journal_dirty_metadata(handle, bh); +- brelse(bh); +- return 0; ++ dx_insert_block (frames + 0, hash2, newblock); ++ dxtrace(dx_show_index ("node", frames[1].entries)); ++ dxtrace(dx_show_index ("node", ++ ((struct dx_node *) bh2->b_data)->entries)); ++ err = ext3_journal_dirty_metadata(handle, bh2); ++ if (err) ++ goto journal_error; ++ brelse (bh2); ++ } else { ++ dxtrace(printk("Creating second level index...\n")); ++ memcpy((char *) entries2, (char *) entries, ++ icount * sizeof(struct dx_entry)); ++ dx_set_limit(entries2, dx_node_limit(dir)); ++ ++ /* Set up root */ ++ dx_set_count(entries, 1); ++ dx_set_block(entries + 0, newblock); ++ ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; ++ ++ /* Add new access path frame */ ++ frame = frames + 1; ++ frame->at = at = at - entries + entries2; ++ frame->entries = entries = entries2; ++ frame->bh = bh2; ++ err = ext3_journal_get_write_access(handle, ++ frame->bh); ++ if (err) ++ goto journal_error; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- brelse (bh); +- return -ENOSPC; ++ de = do_split(handle, dir, &bh, frame, &hinfo, &err); ++ if (!de) ++ goto cleanup; ++ err = add_dirent_to_buf(handle, dentry, inode, de, bh); ++ bh = 0; ++ goto cleanup; ++ ++journal_error: ++ ext3_std_error(dir->i_sb, err); ++cleanup: ++ if (bh) ++ brelse(bh); ++ dx_release(frames); ++ return err; + } ++#endif + + /* + * ext3_delete_entry deletes a directory entry by merging it with the +@@ -451,9 +1542,11 @@ + struct inode * inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -478,9 +1571,11 @@ + struct inode *inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -507,9 +1602,11 @@ + if (dir->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -521,7 +1618,7 @@ + + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; +- inode->i_size = inode->u.ext3_i.i_disksize = inode->i_sb->s_blocksize; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { +@@ -554,21 +1651,19 @@ + inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); +- if (err) +- goto out_no_entry; ++ if (err) { ++ inode->i_nlink = 0; ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + dir->i_nlink++; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + d_instantiate(dentry, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +- +-out_no_entry: +- inode->i_nlink = 0; +- ext3_mark_inode_dirty(handle, inode); +- iput (inode); +- goto out_stop; + } + + /* +@@ -655,7 +1750,7 @@ + int err = 0, rc; + + lock_super(sb); +- if (!list_empty(&inode->u.ext3_i.i_orphan)) ++ if (!list_empty(&EXT3_I(inode)->i_orphan)) + goto out_unlock; + + /* Orphan handling is only valid for files with data blocks +@@ -696,7 +1791,7 @@ + * This is safe: on error we're going to ignore the orphan list + * anyway on the next recovery. */ + if (!err) +- list_add(&inode->u.ext3_i.i_orphan, &EXT3_SB(sb)->s_orphan); ++ list_add(&EXT3_I(inode)->i_orphan, &EXT3_SB(sb)->s_orphan); + + jbd_debug(4, "superblock will point to %ld\n", inode->i_ino); + jbd_debug(4, "orphan inode %ld will point to %d\n", +@@ -714,25 +1809,26 @@ + int ext3_orphan_del(handle_t *handle, struct inode *inode) + { + struct list_head *prev; ++ struct ext3_inode_info *ei = EXT3_I(inode); + struct ext3_sb_info *sbi; + ino_t ino_next; + struct ext3_iloc iloc; + int err = 0; + + lock_super(inode->i_sb); +- if (list_empty(&inode->u.ext3_i.i_orphan)) { ++ if (list_empty(&ei->i_orphan)) { + unlock_super(inode->i_sb); + return 0; + } + + ino_next = NEXT_ORPHAN(inode); +- prev = inode->u.ext3_i.i_orphan.prev; ++ prev = ei->i_orphan.prev; + sbi = EXT3_SB(inode->i_sb); + + jbd_debug(4, "remove inode %ld from orphan list\n", inode->i_ino); + +- list_del(&inode->u.ext3_i.i_orphan); +- INIT_LIST_HEAD(&inode->u.ext3_i.i_orphan); ++ list_del(&ei->i_orphan); ++ INIT_LIST_HEAD(&ei->i_orphan); + + /* If we're on an error path, we may not have a valid + * transaction handle with which to update the orphan list on +@@ -793,8 +1889,9 @@ + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + retval = -ENOENT; + bh = ext3_find_entry (dentry, &de); +@@ -832,7 +1929,7 @@ + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; + ext3_mark_inode_dirty(handle, inode); +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -850,8 +1947,9 @@ + handle_t *handle; + + handle = ext3_journal_start(dir, EXT3_DELETE_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -878,7 +1976,7 @@ + if (retval) + goto end_unlink; + dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + inode->i_nlink--; + if (!inode->i_nlink) +@@ -904,9 +2002,11 @@ + if (l > dir->i_sb->s_blocksize) + return -ENAMETOOLONG; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 5); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -916,7 +2016,7 @@ + if (IS_ERR(inode)) + goto out_stop; + +- if (l > sizeof (inode->u.ext3_i.i_data)) { ++ if (l > sizeof (EXT3_I(inode)->i_data)) { + inode->i_op = &page_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* +@@ -925,25 +2025,23 @@ + * i_size in generic_commit_write(). + */ + err = block_symlink(inode, symname, l); +- if (err) +- goto out_no_entry; ++ if (err) { ++ ext3_dec_count(handle, inode); ++ ext3_mark_inode_dirty(handle, inode); ++ iput (inode); ++ goto out_stop; ++ } + } else { + inode->i_op = &ext3_fast_symlink_inode_operations; +- memcpy((char*)&inode->u.ext3_i.i_data,symname,l); ++ memcpy((char*)&EXT3_I(inode)->i_data,symname,l); + inode->i_size = l-1; + } +- inode->u.ext3_i.i_disksize = inode->i_size; ++ EXT3_I(inode)->i_disksize = inode->i_size; + 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 +2054,15 @@ + if (S_ISDIR(inode->i_mode)) + return -EPERM; + +- if (inode->i_nlink >= EXT3_LINK_MAX) ++ if (inode->i_nlink >= EXT3_LINK_MAX) { + return -EMLINK; ++ } + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(dir)) + handle->h_sync = 1; +@@ -995,9 +2096,11 @@ + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); +- if (IS_ERR(handle)) ++ handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 2); ++ if (IS_ERR(handle)) { + return PTR_ERR(handle); ++ } + + if (IS_SYNC(old_dir) || IS_SYNC(new_dir)) + handle->h_sync = 1; +@@ -1070,14 +2173,33 @@ + /* + * ok, that's it + */ +- ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ retval = ext3_delete_entry(handle, old_dir, old_de, old_bh); ++ if (retval == -ENOENT) { ++ /* ++ * old_de could have moved out from under us. ++ */ ++ struct buffer_head *old_bh2; ++ struct ext3_dir_entry_2 *old_de2; ++ ++ old_bh2 = ext3_find_entry(old_dentry, &old_de2); ++ if (old_bh2) { ++ retval = ext3_delete_entry(handle, old_dir, ++ old_de2, old_bh2); ++ brelse(old_bh2); ++ } ++ } ++ if (retval) { ++ ext3_warning(old_dir->i_sb, "ext3_rename", ++ "Deleting old file (%lu), %d, error=%d", ++ old_dir->i_ino, old_dir->i_nlink, retval); ++ } + + if (new_inode) { + new_inode->i_nlink--; + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +2211,7 @@ + new_inode->i_nlink--; + } else { + new_dir->i_nlink++; +- new_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(new_dir); + ext3_mark_inode_dirty(handle, new_dir); + } + } +diff -Nru a/fs/ext3/super.c b/fs/ext3/super.c +--- a/fs/ext3/super.c Thu Nov 7 11:58:05 2002 ++++ b/fs/ext3/super.c Thu Nov 7 11:58:05 2002 +@@ -106,32 +106,6 @@ + + static char error_buf[1024]; + +-/* Determine the appropriate response to ext3_error on a given filesystem */ +- +-static int ext3_error_behaviour(struct super_block *sb) +-{ +- /* First check for mount-time options */ +- if (test_opt (sb, ERRORS_PANIC)) +- return EXT3_ERRORS_PANIC; +- if (test_opt (sb, ERRORS_RO)) +- return EXT3_ERRORS_RO; +- if (test_opt (sb, ERRORS_CONT)) +- return EXT3_ERRORS_CONTINUE; +- +- /* If no overrides were specified on the mount, then fall back +- * to the default behaviour set in the filesystem's superblock +- * on disk. */ +- switch (le16_to_cpu(sb->u.ext3_sb.s_es->s_errors)) { +- case EXT3_ERRORS_PANIC: +- return EXT3_ERRORS_PANIC; +- case EXT3_ERRORS_RO: +- return EXT3_ERRORS_RO; +- default: +- break; +- } +- return EXT3_ERRORS_CONTINUE; +-} +- + /* Deal with the reporting of failure conditions on a filesystem such as + * inconsistencies detected or read IO failures. + * +@@ -157,18 +131,15 @@ + if (sb->s_flags & MS_RDONLY) + return; + +- if (ext3_error_behaviour(sb) != EXT3_ERRORS_CONTINUE) { +- EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT; +- journal_abort(EXT3_SB(sb)->s_journal, -EIO); +- } +- +- if (ext3_error_behaviour(sb) == EXT3_ERRORS_PANIC) ++ if (test_opt (sb, ERRORS_PANIC)) + panic ("EXT3-fs (device %s): panic forced after error\n", + bdevname(sb->s_dev)); +- +- if (ext3_error_behaviour(sb) == EXT3_ERRORS_RO) { ++ if (test_opt (sb, ERRORS_RO)) { + printk (KERN_CRIT "Remounting filesystem read-only\n"); + sb->s_flags |= MS_RDONLY; ++ } else { ++ EXT3_SB(sb)->s_mount_opt |= EXT3_MOUNT_ABORT; ++ journal_abort(EXT3_SB(sb)->s_journal, -EIO); + } + + ext3_commit_super(sb, es, 1); +@@ -258,7 +229,7 @@ + vsprintf (error_buf, fmt, args); + va_end (args); + +- if (ext3_error_behaviour(sb) == EXT3_ERRORS_PANIC) ++ if (test_opt (sb, ERRORS_PANIC)) + panic ("EXT3-fs panic (device %s): %s: %s\n", + bdevname(sb->s_dev), function, error_buf); + +@@ -416,6 +387,7 @@ + for (i = 0; i < sbi->s_gdb_count; i++) + brelse(sbi->s_group_desc[i]); + kfree(sbi->s_group_desc); ++ kfree(sbi->s_debts); + for (i = 0; i < EXT3_MAX_GROUP_LOADED; i++) + brelse(sbi->s_inode_bitmap[i]); + for (i = 0; i < EXT3_MAX_GROUP_LOADED; i++) +@@ -490,17 +462,32 @@ + return 0; + } + ++static unsigned long get_sb_block(void **data) ++{ ++ unsigned long sb_block; ++ char *options = (char *) *data; ++ ++ if (!options || strncmp(options, "sb=", 3) != 0) ++ return 1; /* Default location */ ++ options += 3; ++ sb_block = simple_strtoul(options, &options, 0); ++ if (*options && *options != ',') { ++ printk("EXT3-fs: Invalid sb specification: %s\n", ++ (char *) *data); ++ return 1; ++ } ++ if (*options == ',') ++ options++; ++ *data = (void *) options; ++ return sb_block; ++} ++ + /* + * This function has been shamelessly adapted from the msdos fs + */ +-static int parse_options (char * options, unsigned long * sb_block, +- struct ext3_sb_info *sbi, +- unsigned long * inum, +- int is_remount) +-{ +- unsigned long *mount_options = &sbi->s_mount_opt; +- uid_t *resuid = &sbi->s_resuid; +- gid_t *resgid = &sbi->s_resgid; ++static int parse_options (char * options, struct ext3_sb_info *sbi, ++ unsigned long * inum, int is_remount) ++{ + char * this_char; + char * value; + +@@ -512,42 +499,42 @@ + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; + if (!strcmp (this_char, "bsddf")) +- clear_opt (*mount_options, MINIX_DF); ++ clear_opt (sbi->s_mount_opt, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +- set_opt (*mount_options, NO_UID32); ++ set_opt (sbi->s_mount_opt, NO_UID32); + } + else if (!strcmp (this_char, "abort")) +- set_opt (*mount_options, ABORT); ++ set_opt (sbi->s_mount_opt, ABORT); + else if (!strcmp (this_char, "check")) { + if (!value || !*value || !strcmp (value, "none")) +- clear_opt (*mount_options, CHECK); ++ clear_opt (sbi->s_mount_opt, CHECK); + else + #ifdef CONFIG_EXT3_CHECK +- set_opt (*mount_options, CHECK); ++ set_opt (sbi->s_mount_opt, CHECK); + #else + printk(KERN_ERR + "EXT3 Check option not supported\n"); + #endif + } + else if (!strcmp (this_char, "debug")) +- set_opt (*mount_options, DEBUG); ++ set_opt (sbi->s_mount_opt, DEBUG); + else if (!strcmp (this_char, "errors")) { + if (want_value(value, "errors")) + return 0; + if (!strcmp (value, "continue")) { +- clear_opt (*mount_options, ERRORS_RO); +- clear_opt (*mount_options, ERRORS_PANIC); +- set_opt (*mount_options, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_RO); ++ clear_opt (sbi->s_mount_opt, ERRORS_PANIC); ++ set_opt (sbi->s_mount_opt, ERRORS_CONT); + } + else if (!strcmp (value, "remount-ro")) { +- clear_opt (*mount_options, ERRORS_CONT); +- clear_opt (*mount_options, ERRORS_PANIC); +- set_opt (*mount_options, ERRORS_RO); ++ clear_opt (sbi->s_mount_opt, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_PANIC); ++ set_opt (sbi->s_mount_opt, ERRORS_RO); + } + else if (!strcmp (value, "panic")) { +- clear_opt (*mount_options, ERRORS_CONT); +- clear_opt (*mount_options, ERRORS_RO); +- set_opt (*mount_options, ERRORS_PANIC); ++ clear_opt (sbi->s_mount_opt, ERRORS_CONT); ++ clear_opt (sbi->s_mount_opt, ERRORS_RO); ++ set_opt (sbi->s_mount_opt, ERRORS_PANIC); + } + else { + printk (KERN_ERR +@@ -558,29 +545,25 @@ + } + else if (!strcmp (this_char, "grpid") || + !strcmp (this_char, "bsdgroups")) +- set_opt (*mount_options, GRPID); ++ set_opt (sbi->s_mount_opt, GRPID); + else if (!strcmp (this_char, "minixdf")) +- set_opt (*mount_options, MINIX_DF); ++ set_opt (sbi->s_mount_opt, MINIX_DF); + else if (!strcmp (this_char, "nocheck")) +- clear_opt (*mount_options, CHECK); ++ clear_opt (sbi->s_mount_opt, CHECK); + else if (!strcmp (this_char, "nogrpid") || + !strcmp (this_char, "sysvgroups")) +- clear_opt (*mount_options, GRPID); ++ clear_opt (sbi->s_mount_opt, GRPID); + else if (!strcmp (this_char, "resgid")) { + unsigned long v; + if (want_numeric(value, "resgid", &v)) + return 0; +- *resgid = v; ++ sbi->s_resgid = v; + } + else if (!strcmp (this_char, "resuid")) { + unsigned long v; + if (want_numeric(value, "resuid", &v)) + return 0; +- *resuid = v; +- } +- else if (!strcmp (this_char, "sb")) { +- if (want_numeric(value, "sb", sb_block)) +- return 0; ++ sbi->s_resuid = v; + } + #ifdef CONFIG_JBD_DEBUG + else if (!strcmp (this_char, "ro-after")) { +@@ -611,12 +594,12 @@ + if (want_value(value, "journal")) + return 0; + if (!strcmp (value, "update")) +- set_opt (*mount_options, UPDATE_JOURNAL); ++ set_opt (sbi->s_mount_opt, UPDATE_JOURNAL); + else if (want_numeric(value, "journal", inum)) + return 0; + } + else if (!strcmp (this_char, "noload")) +- set_opt (*mount_options, NOLOAD); ++ set_opt (sbi->s_mount_opt, NOLOAD); + else if (!strcmp (this_char, "data")) { + int data_opt = 0; + +@@ -635,7 +618,7 @@ + return 0; + } + if (is_remount) { +- if ((*mount_options & EXT3_MOUNT_DATA_FLAGS) != ++ if ((sbi->s_mount_opt & EXT3_MOUNT_DATA_FLAGS) != + data_opt) { + printk(KERN_ERR + "EXT3-fs: cannot change data " +@@ -643,8 +626,8 @@ + return 0; + } + } else { +- *mount_options &= ~EXT3_MOUNT_DATA_FLAGS; +- *mount_options |= data_opt; ++ sbi->s_mount_opt &= ~EXT3_MOUNT_DATA_FLAGS; ++ sbi->s_mount_opt |= data_opt; + } + } else if (!strcmp (this_char, "commit")) { + unsigned long v; +@@ -707,6 +690,7 @@ + es->s_mtime = cpu_to_le32(CURRENT_TIME); + ext3_update_dynamic_rev(sb); + EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); ++ + ext3_commit_super (sb, es, 1); + if (test_opt (sb, DEBUG)) + printk (KERN_INFO +@@ -717,6 +701,7 @@ + EXT3_BLOCKS_PER_GROUP(sb), + EXT3_INODES_PER_GROUP(sb), + sbi->s_mount_opt); ++ + printk(KERN_INFO "EXT3 FS " EXT3FS_VERSION ", " EXT3FS_DATE " on %s, ", + bdevname(sb->s_dev)); + if (EXT3_SB(sb)->s_journal->j_inode == NULL) { +@@ -890,17 +875,39 @@ + return res; + } + ++static unsigned long descriptor_loc(struct super_block *sb, ++ unsigned long logic_sb_block, ++ int nr) ++{ ++ struct ext3_sb_info *sbi = EXT3_SB(sb); ++ unsigned long bg, first_data_block, first_meta_bg; ++ int has_super = 0; ++ ++ first_data_block = le32_to_cpu(sbi->s_es->s_first_data_block); ++ first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg); ++ ++ if (!EXT3_HAS_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_META_BG) || ++ nr < first_meta_bg) ++ return (logic_sb_block + nr + 1); ++ bg = sbi->s_desc_per_block * nr; ++ if (ext3_bg_has_super(sb, bg)) ++ has_super = 1; ++ return (first_data_block + has_super + (bg * sbi->s_blocks_per_group)); ++} ++ ++ + struct super_block * ext3_read_super (struct super_block * sb, void * data, + int silent) + { + struct buffer_head * bh; + struct ext3_super_block *es = 0; + struct ext3_sb_info *sbi = EXT3_SB(sb); +- unsigned long sb_block = 1; +- unsigned long logic_sb_block = 1; +- unsigned long offset = 0; ++ unsigned long sb_block = get_sb_block(&data); ++ unsigned long block, logic_sb_block; ++ unsigned long offset; + unsigned long journal_inum = 0; + kdev_t dev = sb->s_dev; ++ unsigned long def_mount_opts; + int blocksize; + int hblock; + int db_count; +@@ -922,14 +929,6 @@ + if (blocksize < hblock) + blocksize = hblock; + +- sbi->s_mount_opt = 0; +- sbi->s_resuid = EXT3_DEF_RESUID; +- sbi->s_resgid = EXT3_DEF_RESGID; +- if (!parse_options ((char *) data, &sb_block, sbi, &journal_inum, 0)) { +- sb->s_dev = 0; +- goto out_fail; +- } +- + sb->s_blocksize = blocksize; + set_blocksize (dev, blocksize); + +@@ -937,10 +936,8 @@ + * The ext3 superblock will not be buffer aligned for other than 1kB + * block sizes. We need to calculate the offset from buffer start. + */ +- if (blocksize != EXT3_MIN_BLOCK_SIZE) { +- logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; +- offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; +- } ++ logic_sb_block = (sb_block * EXT3_MIN_BLOCK_SIZE) / blocksize; ++ offset = (sb_block * EXT3_MIN_BLOCK_SIZE) % blocksize; + + if (!(bh = sb_bread(sb, logic_sb_block))) { + printk (KERN_ERR "EXT3-fs: unable to read superblock\n"); +@@ -960,6 +957,34 @@ + bdevname(dev)); + goto failed_mount; + } ++ ++ /* Set defaults before we parse the mount options */ ++ def_mount_opts = le32_to_cpu(es->s_default_mount_opts); ++ sbi->s_mount_opt = 0; ++ if (def_mount_opts & EXT3_DEFM_DEBUG) ++ set_opt(sbi->s_mount_opt, DEBUG); ++ if (def_mount_opts & EXT3_DEFM_BSDGROUPS) ++ set_opt(sbi->s_mount_opt, GRPID); ++ if (def_mount_opts & EXT3_DEFM_UID16) ++ set_opt(sbi->s_mount_opt, NO_UID32); ++ if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_DATA) ++ sbi->s_mount_opt |= EXT3_MOUNT_JOURNAL_DATA; ++ else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_ORDERED) ++ sbi->s_mount_opt |= EXT3_MOUNT_ORDERED_DATA; ++ else if ((def_mount_opts & EXT3_DEFM_JMODE) == EXT3_DEFM_JMODE_WBACK) ++ sbi->s_mount_opt |= EXT3_MOUNT_WRITEBACK_DATA; ++ ++ if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_PANIC) ++ set_opt(sbi->s_mount_opt, ERRORS_PANIC); ++ else if (le16_to_cpu(sbi->s_es->s_errors) == EXT3_ERRORS_RO) ++ set_opt(sbi->s_mount_opt, ERRORS_RO); ++ ++ sbi->s_resuid = le16_to_cpu(es->s_def_resuid); ++ sbi->s_resgid = le16_to_cpu(es->s_def_resgid); ++ ++ if (!parse_options ((char *) data, sbi, &journal_inum, 0)) ++ goto failed_mount; ++ + if (le32_to_cpu(es->s_rev_level) == EXT3_GOOD_OLD_REV && + (EXT3_HAS_COMPAT_FEATURE(sb, ~0U) || + EXT3_HAS_RO_COMPAT_FEATURE(sb, ~0U) || +@@ -1036,7 +1061,9 @@ + } else { + sbi->s_inode_size = le16_to_cpu(es->s_inode_size); + sbi->s_first_ino = le32_to_cpu(es->s_first_ino); +- if (sbi->s_inode_size != EXT3_GOOD_OLD_INODE_SIZE) { ++ if ((sbi->s_inode_size < EXT3_GOOD_OLD_INODE_SIZE) || ++ (sbi->s_inode_size & (sbi->s_inode_size - 1)) || ++ (sbi->s_inode_size > blocksize)) { + printk (KERN_ERR + "EXT3-fs: unsupported inode size: %d\n", + sbi->s_inode_size); +@@ -1059,13 +1086,12 @@ + sbi->s_itb_per_group = sbi->s_inodes_per_group /sbi->s_inodes_per_block; + sbi->s_desc_per_block = blocksize / sizeof(struct ext3_group_desc); + sbi->s_sbh = bh; +- if (sbi->s_resuid == EXT3_DEF_RESUID) +- sbi->s_resuid = le16_to_cpu(es->s_def_resuid); +- if (sbi->s_resgid == EXT3_DEF_RESGID) +- sbi->s_resgid = le16_to_cpu(es->s_def_resgid); + 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 +@@ -1098,8 +1124,16 @@ + printk (KERN_ERR "EXT3-fs: not enough memory\n"); + goto failed_mount; + } ++ sbi->s_debts = kmalloc(sbi->s_groups_count * sizeof(*sbi->s_debts), ++ GFP_KERNEL); ++ if (!sbi->s_debts) { ++ printk ("EXT3-fs: not enough memory\n"); ++ goto failed_mount2; ++ } ++ memset(sbi->s_debts, 0, sbi->s_groups_count * sizeof(*sbi->s_debts)); + for (i = 0; i < db_count; i++) { +- sbi->s_group_desc[i] = sb_bread(sb, logic_sb_block + i + 1); ++ block = descriptor_loc(sb, logic_sb_block, i); ++ sbi->s_group_desc[i] = sb_bread(sb, block); + if (!sbi->s_group_desc[i]) { + printk (KERN_ERR "EXT3-fs: " + "can't read group descriptor %d\n", i); +@@ -1120,6 +1154,7 @@ + sbi->s_loaded_inode_bitmaps = 0; + sbi->s_loaded_block_bitmaps = 0; + sbi->s_gdb_count = db_count; ++ sbi->s_dir_count = ext3_count_dirs(sb); + get_random_bytes(&sbi->s_next_generation, sizeof(u32)); + /* + * set up enough so that it can read an inode +@@ -1223,6 +1258,8 @@ + failed_mount3: + journal_destroy(sbi->s_journal); + failed_mount2: ++ if (sbi->s_debts) ++ kfree(sbi->s_debts); + for (i = 0; i < db_count; i++) + brelse(sbi->s_group_desc[i]); + kfree(sbi->s_group_desc); +@@ -1651,7 +1688,7 @@ + /* + * Allow the "check" option to be passed as a remount option. + */ +- if (!parse_options(data, &tmp, sbi, &tmp, 1)) ++ if (!parse_options(data, sbi, &tmp, 1)) + return -EINVAL; + + if (sbi->s_mount_opt & EXT3_MOUNT_ABORT) +diff -Nru a/include/linux/ext2_fs.h b/include/linux/ext2_fs.h +--- a/include/linux/ext2_fs.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext2_fs.h Thu Nov 7 11:58:05 2002 +@@ -198,10 +198,11 @@ + #define EXT2_ECOMPR_FL 0x00000800 /* Compression error */ + /* End compression flags --- maybe not all used */ + #define EXT2_BTREE_FL 0x00001000 /* btree format dir */ ++#define EXT2_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ + #define EXT2_RESERVED_FL 0x80000000 /* reserved for ext2 lib */ + +-#define EXT2_FL_USER_VISIBLE 0x00001FFF /* User visible flags */ +-#define EXT2_FL_USER_MODIFIABLE 0x000000FF /* User modifiable flags */ ++#define EXT2_FL_USER_VISIBLE 0x0003DFFF /* User visible flags */ ++#define EXT2_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */ + + /* + * ioctl commands +@@ -307,6 +308,7 @@ + * Mount flags + */ + #define EXT2_MOUNT_CHECK 0x0001 /* Do mount-time checks */ ++#define EXT2_MOUNT_OLDALLOC 0x0002 /* Don't use the new Orlov allocator */ + #define EXT2_MOUNT_GRPID 0x0004 /* Create files with directory's group */ + #define EXT2_MOUNT_DEBUG 0x0008 /* Some debugging messages */ + #define EXT2_MOUNT_ERRORS_CONT 0x0010 /* Continue on errors */ +@@ -392,7 +394,20 @@ + __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/ + __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */ + __u16 s_padding1; +- __u32 s_reserved[204]; /* Padding to the end of the block */ ++ /* ++ * Journaling support valid if EXT3_FEATURE_COMPAT_HAS_JOURNAL set. ++ */ ++ __u8 s_journal_uuid[16]; /* uuid of journal superblock */ ++ __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 */ ++ __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_default_mount_opts; ++ __u32 s_first_meta_bg; /* First metablock block group */ ++ __u32 s_reserved[190]; /* Padding to the end of the block */ + }; + + #ifdef __KERNEL__ +@@ -464,10 +479,12 @@ + #define EXT2_FEATURE_INCOMPAT_FILETYPE 0x0002 + #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 ++#define EXT2_FEATURE_INCOMPAT_META_BG 0x0010 + #define EXT2_FEATURE_INCOMPAT_ANY 0xffffffff + + #define EXT2_FEATURE_COMPAT_SUPP 0 +-#define EXT2_FEATURE_INCOMPAT_SUPP EXT2_FEATURE_INCOMPAT_FILETYPE ++#define EXT2_FEATURE_INCOMPAT_SUPP (EXT2_FEATURE_INCOMPAT_FILETYPE| \ ++ EXT2_FEATURE_INCOMPAT_META_BG) + #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER| \ + EXT2_FEATURE_RO_COMPAT_LARGE_FILE| \ + EXT2_FEATURE_RO_COMPAT_BTREE_DIR) +@@ -475,10 +492,18 @@ + #define EXT2_FEATURE_INCOMPAT_UNSUPPORTED ~EXT2_FEATURE_INCOMPAT_SUPP + + /* +- * Default values for user and/or group using reserved blocks ++ * Default mount options + */ +-#define EXT2_DEF_RESUID 0 +-#define EXT2_DEF_RESGID 0 ++#define EXT2_DEFM_DEBUG 0x0001 ++#define EXT2_DEFM_BSDGROUPS 0x0002 ++#define EXT2_DEFM_XATTR_USER 0x0004 ++#define EXT2_DEFM_ACL 0x0008 ++#define EXT2_DEFM_UID16 0x0010 ++ /* Not used by ext2, but reserved for use by ext3 */ ++#define EXT3_DEFM_JMODE 0x0060 ++#define EXT3_DEFM_JMODE_DATA 0x0020 ++#define EXT3_DEFM_JMODE_ORDERED 0x0040 ++#define EXT3_DEFM_JMODE_WBACK 0x0060 + + /* + * Structure of a directory entry +@@ -576,6 +601,7 @@ + extern struct inode * ext2_new_inode (const struct inode *, int); + extern void ext2_free_inode (struct inode *); + extern unsigned long ext2_count_free_inodes (struct super_block *); ++extern unsigned long ext2_count_dirs (struct super_block *); + extern void ext2_check_inodes_bitmap (struct super_block *); + extern unsigned long ext2_count_free (struct buffer_head *, unsigned); + +diff -Nru a/include/linux/ext2_fs_i.h b/include/linux/ext2_fs_i.h +--- a/include/linux/ext2_fs_i.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext2_fs_i.h Thu Nov 7 11:58:05 2002 +@@ -25,6 +25,7 @@ + __u32 i_faddr; + __u8 i_frag_no; + __u8 i_frag_size; ++ __u16 i_state; + __u32 i_file_acl; + __u32 i_dir_acl; + __u32 i_dtime; +@@ -34,7 +35,11 @@ + __u32 i_prealloc_block; + __u32 i_prealloc_count; + __u32 i_dir_start_lookup; +- int i_new_inode:1; /* Is a freshly allocated inode */ + }; ++ ++/* ++ * Inode dynamic state flags ++ */ ++#define EXT2_STATE_NEW 0x00000001 /* inode is newly created */ + + #endif /* _LINUX_EXT2_FS_I */ +diff -Nru a/include/linux/ext2_fs_sb.h b/include/linux/ext2_fs_sb.h +--- a/include/linux/ext2_fs_sb.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext2_fs_sb.h Thu Nov 7 11:58:05 2002 +@@ -56,6 +56,8 @@ + int s_desc_per_block_bits; + int s_inode_size; + int s_first_ino; ++ unsigned long s_dir_count; ++ u8 *s_debts; + }; + + #endif /* _LINUX_EXT2_FS_SB */ +diff -Nru a/include/linux/ext3_fs.h b/include/linux/ext3_fs.h +--- a/include/linux/ext3_fs.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext3_fs.h Thu Nov 7 11:58:05 2002 +@@ -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 +@@ -203,10 +208,11 @@ + #define EXT3_INDEX_FL 0x00001000 /* hash-indexed directory */ + #define EXT3_IMAGIC_FL 0x00002000 /* AFS directory */ + #define EXT3_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */ ++#define EXT3_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/ + #define EXT3_RESERVED_FL 0x80000000 /* reserved for ext3 lib */ + +-#define EXT3_FL_USER_VISIBLE 0x00005FFF /* User visible flags */ +-#define EXT3_FL_USER_MODIFIABLE 0x000000FF /* User modifiable flags */ ++#define EXT3_FL_USER_VISIBLE 0x0003DFFF /* User visible flags */ ++#define EXT3_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */ + + /* + * Inode dynamic state flags +@@ -325,6 +331,7 @@ + * Mount flags + */ + #define EXT3_MOUNT_CHECK 0x0001 /* Do mount-time checks */ ++#define EXT3_MOUNT_OLDALLOC 0x0002 /* Don't use the new Orlov allocator */ + #define EXT3_MOUNT_GRPID 0x0004 /* Create files with directory's group */ + #define EXT3_MOUNT_DEBUG 0x0008 /* Some debugging messages */ + #define EXT3_MOUNT_ERRORS_CONT 0x0010 /* Continue on errors */ +@@ -437,8 +444,13 @@ + /*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_default_mount_opts; ++ __u32 s_first_meta_bg; /* First metablock block group */ ++ __u32 s_reserved[190]; /* Padding to the end of the block */ + }; + + #ifdef __KERNEL__ +@@ -511,19 +523,28 @@ + #define EXT3_FEATURE_INCOMPAT_FILETYPE 0x0002 + #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ ++#define EXT3_FEATURE_INCOMPAT_META_BG 0x0010 + + #define EXT3_FEATURE_COMPAT_SUPP 0 + #define EXT3_FEATURE_INCOMPAT_SUPP (EXT3_FEATURE_INCOMPAT_FILETYPE| \ +- EXT3_FEATURE_INCOMPAT_RECOVER) ++ EXT3_FEATURE_INCOMPAT_RECOVER| \ ++ EXT3_FEATURE_INCOMPAT_META_BG) + #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER| \ + EXT3_FEATURE_RO_COMPAT_LARGE_FILE| \ + EXT3_FEATURE_RO_COMPAT_BTREE_DIR) + + /* +- * Default values for user and/or group using reserved blocks ++ * Default mount options + */ +-#define EXT3_DEF_RESUID 0 +-#define EXT3_DEF_RESGID 0 ++#define EXT3_DEFM_DEBUG 0x0001 ++#define EXT3_DEFM_BSDGROUPS 0x0002 ++#define EXT3_DEFM_XATTR_USER 0x0004 ++#define EXT3_DEFM_ACL 0x0008 ++#define EXT3_DEFM_UID16 0x0010 ++#define EXT3_DEFM_JMODE 0x0060 ++#define EXT3_DEFM_JMODE_DATA 0x0020 ++#define EXT3_DEFM_JMODE_ORDERED 0x0040 ++#define EXT3_DEFM_JMODE_WBACK 0x0060 + + /* + * Structure of a directory entry +@@ -575,9 +596,46 @@ + #define EXT3_DIR_ROUND (EXT3_DIR_PAD - 1) + #define EXT3_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT3_DIR_ROUND) & \ + ~EXT3_DIR_ROUND) ++/* ++ * Hash Tree Directory indexing ++ * (c) Daniel Phillips, 2001 ++ */ ++ ++#ifdef CONFIG_EXT3_INDEX ++ #define is_dx(dir) (EXT3_HAS_COMPAT_FEATURE(dir->i_sb, \ ++ EXT3_FEATURE_COMPAT_DIR_INDEX) && \ ++ (EXT3_I(dir)->i_flags & EXT3_INDEX_FL)) ++#define EXT3_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1) ++#else ++ #define is_dx(dir) 0 ++#define EXT3_DIR_LINK_MAX(dir) ((dir)->i_nlink >= EXT3_LINK_MAX) ++#define EXT3_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2) ++#endif ++ ++/* Legal values for the dx_root hash_version field: */ ++ ++#define DX_HASH_LEGACY 0 ++#define DX_HASH_HALF_MD4 1 ++#define DX_HASH_TEA 2 ++ ++/* hash info structure used by the directory hash */ ++struct dx_hash_info ++{ ++ u32 hash; ++ u32 minor_hash; ++ int hash_version; ++ u32 *seed; ++}; + + #ifdef __KERNEL__ + /* ++ * Control parameters used by ext3_htree_next_block ++ */ ++#define HASH_NB_ALWAYS 1 ++ ++ ++/* + * Describe an inode's exact location on disk and in memory + */ + struct ext3_iloc +@@ -587,6 +645,27 @@ + unsigned long block_group; + }; + ++ ++/* ++ * This structure is stuffed into the struct file's private_data field ++ * for directories. It is where we put information so that we can do ++ * readdir operations in hash tree order. ++ */ ++struct dir_private_info { ++ rb_root_t root; ++ rb_node_t *curr_node; ++ struct fname *extra_fname; ++ loff_t last_pos; ++ __u32 curr_hash; ++ __u32 curr_minor_hash; ++ __u32 next_hash; ++}; ++ ++/* ++ * Special error return code only used by dx_probe() and its callers. ++ */ ++#define ERR_BAD_DX_DIR -75000 ++ + /* + * Function prototypes + */ +@@ -614,16 +693,26 @@ + + /* dir.c */ + extern int ext3_check_dir_entry(const char *, struct inode *, +- struct ext3_dir_entry_2 *, struct buffer_head *, +- unsigned long); ++ struct ext3_dir_entry_2 *, ++ struct buffer_head *, unsigned long); ++extern int ext3_htree_store_dirent(struct file *dir_file, __u32 hash, ++ __u32 minor_hash, ++ struct ext3_dir_entry_2 *dirent); ++extern void ext3_htree_free_dir_info(struct dir_private_info *p); ++ + /* fsync.c */ + extern int ext3_sync_file (struct file *, struct dentry *, int); + ++/* hash.c */ ++extern int ext3fs_dirhash(const char *name, int len, struct ++ dx_hash_info *hinfo); ++ + /* ialloc.c */ + extern struct inode * ext3_new_inode (handle_t *, const struct inode *, int); + extern void ext3_free_inode (handle_t *, struct inode *); + extern struct inode * ext3_orphan_get (struct super_block *, ino_t); + extern unsigned long ext3_count_free_inodes (struct super_block *); ++extern unsigned long ext3_count_dirs (struct super_block *); + extern void ext3_check_inodes_bitmap (struct super_block *); + extern unsigned long ext3_count_free (struct buffer_head *, unsigned); + +@@ -650,6 +739,8 @@ + /* namei.c */ + extern int ext3_orphan_add(handle_t *, struct inode *); + extern int ext3_orphan_del(handle_t *, struct inode *); ++extern int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ++ __u32 start_minor_hash, __u32 *next_hash); + + /* super.c */ + extern void ext3_error (struct super_block *, const char *, const char *, ...) +diff -Nru a/include/linux/ext3_fs_sb.h b/include/linux/ext3_fs_sb.h +--- a/include/linux/ext3_fs_sb.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext3_fs_sb.h Thu Nov 7 11:58:05 2002 +@@ -62,6 +62,10 @@ + int s_inode_size; + int s_first_ino; + u32 s_next_generation; ++ u32 s_hash_seed[4]; ++ int s_def_hash_version; ++ unsigned long s_dir_count; ++ u8 *s_debts; + + /* Journaling */ + struct inode * s_journal_inode; +diff -Nru a/include/linux/ext3_jbd.h b/include/linux/ext3_jbd.h +--- a/include/linux/ext3_jbd.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/ext3_jbd.h Thu Nov 7 11:58:05 2002 +@@ -63,6 +63,8 @@ + + #define EXT3_RESERVE_TRANS_BLOCKS 12 + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, +diff -Nru a/include/linux/rbtree.h b/include/linux/rbtree.h +--- a/include/linux/rbtree.h Thu Nov 7 11:58:05 2002 ++++ b/include/linux/rbtree.h Thu Nov 7 11:58:05 2002 +@@ -120,6 +120,8 @@ + + extern void rb_insert_color(rb_node_t *, rb_root_t *); + extern void rb_erase(rb_node_t *, rb_root_t *); ++extern rb_node_t *rb_get_first(rb_root_t *root); ++extern rb_node_t *rb_get_next(rb_node_t *n); + + static inline void rb_link_node(rb_node_t * node, rb_node_t * parent, rb_node_t ** rb_link) + { +diff -Nru a/lib/rbtree.c b/lib/rbtree.c +--- a/lib/rbtree.c Thu Nov 7 11:58:05 2002 ++++ b/lib/rbtree.c Thu Nov 7 11:58:05 2002 +@@ -17,6 +17,8 @@ + Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA + + linux/lib/rbtree.c ++ ++ rb_get_first and rb_get_next written by Theodore Ts'o, 9/8/2002 + */ + + #include +@@ -294,3 +296,43 @@ + __rb_erase_color(child, parent, root); + } + EXPORT_SYMBOL(rb_erase); ++ ++/* ++ * This function returns the first node (in sort order) of the tree. ++ */ ++rb_node_t *rb_get_first(rb_root_t *root) ++{ ++ rb_node_t *n; ++ ++ n = root->rb_node; ++ if (!n) ++ return 0; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++} ++EXPORT_SYMBOL(rb_get_first); ++ ++/* ++ * Given a node, this function will return the next node in the tree. ++ */ ++rb_node_t *rb_get_next(rb_node_t *n) ++{ ++ rb_node_t *parent; ++ ++ if (n->rb_right) { ++ n = n->rb_right; ++ while (n->rb_left) ++ n = n->rb_left; ++ return n; ++ } else { ++ while ((parent = n->rb_parent)) { ++ if (n == parent->rb_left) ++ return parent; ++ n = parent; ++ } ++ return 0; ++ } ++} ++EXPORT_SYMBOL(rb_get_next); ++ diff --git a/lustre/kernel_patches/patches/htree-ext3-2.4.18.patch b/lustre/kernel_patches/patches/htree-ext3-2.4.18.patch new file mode 100644 index 0000000..a54e9ca --- /dev/null +++ b/lustre/kernel_patches/patches/htree-ext3-2.4.18.patch @@ -0,0 +1,1201 @@ +--- ./fs/ext3/super.c 2002/03/05 06:18:59 2.1 ++++ ./fs/ext3/super.c 2002/03/05 06:26:56 +@@ -529,6 +529,12 @@ + "EXT3 Check option not supported\n"); + #endif + } ++ else if (!strcmp (this_char, "index")) ++#ifdef CONFIG_EXT3_INDEX ++ set_opt (*mount_options, INDEX); ++#else ++ printk("EXT3 index option not supported\n"); ++#endif + else if (!strcmp (this_char, "debug")) + set_opt (*mount_options, DEBUG); + else if (!strcmp (this_char, "errors")) { +@@ -702,6 +708,12 @@ + es->s_mtime = cpu_to_le32(CURRENT_TIME); + ext3_update_dynamic_rev(sb); + EXT3_SET_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); ++ ++ if (test_opt(sb, INDEX)) ++ EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX); ++ else if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_DIR_INDEX)) ++ set_opt (EXT3_SB(sb)->s_mount_opt, INDEX); ++ + ext3_commit_super (sb, es, 1); + if (test_opt (sb, DEBUG)) + printk (KERN_INFO +--- ./fs/ext3/namei.c 2002/03/05 06:18:59 2.1 ++++ ./fs/ext3/namei.c 2002/03/06 00:13:18 +@@ -16,6 +16,10 @@ + * 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 + */ + + #include +@@ -33,7 +33,7 @@ + #include + #include + #include +- ++#include + + /* + * define how far ahead to read directories while searching them. +@@ -38,6 +42,437 @@ + #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; ++ ++#define dxtrace_on(command) command ++#define dxtrace_off(command) ++ ++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; /* 0 now, 1 at release */ ++ 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; ++}; ++ ++typedef struct ext3_dir_entry_2 ext3_dirent; ++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 unsigned dx_hack_hash (const u8 *name, int len); ++static struct dx_frame *dx_probe (struct inode *dir, u32 hash, struct dx_frame *frame); ++static void dx_release (struct dx_frame *frames); ++static int dx_make_map (ext3_dirent *de, int size, struct dx_map_entry map[]); ++static void dx_sort_map(struct dx_map_entry *map, unsigned count); ++static ext3_dirent *dx_copy_dirents (char *from, char *to, ++ struct dx_map_entry *map, int count); ++static void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block); ++ ++ ++#ifdef CONFIG_EXT3_INDEX ++/* ++ * 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); ++} ++ ++/* Hash function - not bad, but still looking for an ideal default */ ++ ++static unsigned dx_hack_hash (const u8 *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; ++} ++ ++#define dx_hash(s,n) (dx_hack_hash(s,n) << 1) ++ ++/* ++ * Debug ++ */ ++#ifdef DX_DEBUG ++#define dxtrace dxtrace_on ++static void dx_show_index (char * label, struct dx_entry *entries) ++{ ++ int i, n = dx_get_count (entries); ++ printk("%s index ", label); ++ for (i = 0; i < n; i++) ++ { ++ printk("%x->%u ", i? dx_get_hash(entries + i): 0, dx_get_block(entries + i)); ++ } ++ printk("\n"); ++} ++ ++struct stats ++{ ++ unsigned names; ++ unsigned space; ++ unsigned bcount; ++}; ++ ++static struct stats dx_show_leaf (ext3_dirent *de, int size, int show_names) ++{ ++ unsigned names = 0, space = 0; ++ char *base = (char *) de; ++ 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++); ++ printk(":%x.%u ", dx_hash (de->name, de->name_len), ((char *) de - base)); ++ } ++ space += EXT3_DIR_REC_LEN(de->name_len); ++ names++; ++ } ++ de = (ext3_dirent *) ((char *) de + le16_to_cpu(de->rec_len)); ++ } ++ printk("(%i)\n", names); ++ return (struct stats) { names, space, 1 }; ++} ++ ++struct stats dx_show_entries (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 (dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): ++ dx_show_leaf ((ext3_dirent *) 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}; ++} ++#else ++#define dxtrace dxtrace_off ++#endif ++ ++/* ++ * Probe for a directory leaf block to search ++ */ ++ ++static struct dx_frame * ++dx_probe(struct inode *dir, u32 hash, struct dx_frame *frame_in) ++{ ++ 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; ++ int err; ++ ++ frame->bh = NULL; ++ if (!(bh = ext3_bread(NULL, dir, 0, 0, &err))) ++ goto fail; ++ root = (struct dx_root *) bh->b_data; ++ if (root->info.hash_version > 0 || root->info.unused_flags & 1) { ++ brelse(bh); ++ goto fail; ++ } ++ if ((indirect = root->info.indirect_levels) > 1) { ++ brelse(bh); ++ 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); ++} ++ ++/* ++ * Directory block splitting, compacting ++ */ ++ ++static int dx_make_map (ext3_dirent *de, int size, struct dx_map_entry map[]) ++{ ++ int count = 0; ++ char *base = (char *) de; ++ while ((char *) de < base + size) { ++ if (de->name_len && de->inode) { ++ map[count].hash = dx_hash (de->name, de->name_len); ++ map[count].offs = (u32) ((char *) de - base); ++ count++; ++ } ++ de = (ext3_dirent *) ((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 (!test_opt(inode->i_sb, INDEX)) ++ EXT3_I(inode)->i_flags &= ~EXT3_INDEX_FL; ++} ++ + /* + * NOTE! unlike strncmp, ext3_match returns 1 for success, 0 for failure. + * +@@ -95,6 +529,15 @@ + } + + /* ++ * p is at least 6 bytes before the end of page ++ */ ++static inline ext3_dirent *ext3_next_entry(ext3_dirent *p) ++{ ++ return (ext3_dirent *)((char*)p + le16_to_cpu(p->rec_len)); ++} ++ ++ ++/* + * ext3_find_entry() + * + * finds an entry in the specified directory with the wanted name. It +@@ -105,6 +548,8 @@ + * The returned buffer_head has ->b_count elevated. The caller is expected + * to brelse() it when appropriate. + */ ++ ++ + static struct buffer_head * ext3_find_entry (struct dentry *dentry, + struct ext3_dir_entry_2 ** res_dir) + { +@@ -119,10 +564,70 @@ + int num = 0; + int nblocks, i, err; + struct inode *dir = dentry->d_parent->d_inode; ++ ext3_dirent *de, *top; + + *res_dir = NULL; + sb = dir->i_sb; ++ if (dentry->d_name.len > EXT3_NAME_LEN) ++ return NULL; ++ if (ext3_dx && is_dx(dir)) { ++ u32 hash = dx_hash(dentry->d_name.name, dentry->d_name.len); ++ struct dx_frame frames[2], *frame; ++ if (!(frame = dx_probe (dir, hash, frames))) ++ return NULL; ++dxnext: ++ block = dx_get_block(frame->at); ++ if (!(bh = ext3_bread (NULL,dir, block, 0, &err))) ++ goto dxfail; ++ de = (ext3_dirent *) bh->b_data; ++ top = (ext3_dirent *) ((char *) de + sb->s_blocksize - ++ EXT3_DIR_REC_LEN(0)); ++ for (; de < top; de = ext3_next_entry(de)) ++ if (ext3_match(dentry->d_name.len, dentry->d_name.name, de)) { ++ if (!ext3_check_dir_entry("ext3_find_entry", ++ dir, de, bh, ++ (block<b_data))) { ++ brelse (bh); ++ goto dxfail; ++ } ++ *res_dir = de; ++ goto dxfound; ++ } ++ brelse (bh); ++ /* Same hash continues in next block? Search on. */ ++ if (++(frame->at) == frame->entries + dx_get_count(frame->entries)) ++ { ++ struct buffer_head *bh2; ++ if (frame == frames) ++ goto dxfail; ++ if (++(frames->at) == frames->entries + dx_get_count(frames->entries)) ++ goto dxfail; ++ /* should omit read if not continued */ ++ if (!(bh2 = ext3_bread (NULL, dir, ++ dx_get_block(frames->at), ++ 0, &err))) ++ goto dxfail; ++ brelse (frame->bh); ++ frame->bh = bh2; ++ frame->at = frame->entries = ((struct dx_node *) bh2->b_data)->entries; ++ /* Subtle: the 0th entry has the count, find the hash in frame above */ ++ if ((dx_get_hash(frames->at) & -2) == hash) ++ goto dxnext; ++ goto dxfail; ++ } ++ if ((dx_get_hash(frame->at) & -2) == hash) ++ goto dxnext; ++dxfail: ++ dxtrace(printk("%s not found\n", name)); ++ dx_release (frames); ++ return NULL; ++dxfound: ++ dx_release (frames); ++ return bh; + ++ } ++ + nblocks = dir->i_size >> EXT3_BLOCK_SIZE_BITS(sb); + start = dir->u.ext3_i.i_dir_start_lookup; + if (start >= nblocks) +@@ -237,6 +748,90 @@ + de->file_type = ext3_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; + } + ++static ext3_dirent * ++dx_copy_dirents (char *from, char *to, struct dx_map_entry *map, int count) ++{ ++ unsigned rec_len = 0; ++ ++ while (count--) { ++ ext3_dirent *de = (ext3_dirent *) (from + map->offs); ++ rec_len = EXT3_DIR_REC_LEN(de->name_len); ++ memcpy (to, de, rec_len); ++ ((ext3_dirent *) to)->rec_len = rec_len; ++ to += rec_len; ++ map++; ++ } ++ return (ext3_dirent *) (to - rec_len); ++} ++ ++#ifdef CONFIG_EXT3_INDEX ++static ext3_dirent *do_split(handle_t *handle, struct inode *dir, ++ struct buffer_head **bh,struct dx_frame *frame, ++ u32 hash, int *error) ++{ ++ unsigned count; ++ struct buffer_head *bh2; ++ u32 newblock; ++ u32 hash2; ++ struct dx_map_entry *map; ++ char *data1 = (*bh)->b_data, *data2, *data3; ++ unsigned split; ++ ext3_dirent *de, *de2; ++ ++ bh2 = ext3_append (handle, dir, &newblock, error); ++ if (!(bh2)) ++ { ++ brelse(*bh); ++ *bh = NULL; ++ return (ext3_dirent *)bh2; ++ } ++ ++ BUFFER_TRACE(*bh, "get_write_access"); ++ ext3_journal_get_write_access(handle, *bh); ++ BUFFER_TRACE(frame->bh, "get_write_access"); ++ ext3_journal_get_write_access(handle, frame->bh); ++ ++ data2 = bh2->b_data; ++ ++ map = kmalloc(sizeof(*map) * PAGE_CACHE_SIZE/EXT3_DIR_REC_LEN(1) + 1, ++ GFP_KERNEL); ++ if (!map) ++ panic("no memory for do_split\n"); ++ count = dx_make_map((ext3_dirent *)data1, dir->i_sb->s_blocksize, map); ++ split = count/2; // need to adjust to actual middle ++ dx_sort_map (map, count); ++ hash2 = map[split].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_copy_dirents (data1, data2, map + split, count - split); ++ data3 = (char *) de2 + de2->rec_len; ++ de = dx_copy_dirents (data1, data3, map, split); ++ memcpy(data1, data3, (char *) de + de->rec_len - data3); ++ de = (ext3_dirent *) ((char *) de - data3 + data1); // relocate de ++ de->rec_len = cpu_to_le16(data1 + dir->i_sb->s_blocksize - (char *)de); ++ de2->rec_len = cpu_to_le16(data2 + dir->i_sb->s_blocksize-(char *)de2); ++ dxtrace(dx_show_leaf((ext3_dirent *)data1, dir->i_sb->s_blocksize, 1)); ++ dxtrace(dx_show_leaf((ext3_dirent *)data2, dir->i_sb->s_blocksize, 1)); ++ ++ /* Which block gets the new entry? */ ++ if (hash >= hash2) ++ { ++ swap(*bh, bh2); ++ de = de2; ++ } ++ dx_insert_block(frame, hash2 + (hash2 == map[split-1].hash), newblock); ++ ext3_journal_dirty_metadata (handle, bh2); ++ brelse (bh2); ++ ext3_journal_dirty_metadata (handle, frame->bh); ++ dxtrace(dx_show_index ("frame", frame->entries)); ++ kfree(map); ++ return de; ++} ++#endif ++ ++ + /* + * ext3_add_entry() + * +@@ -255,118 +849,278 @@ + 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 super_block * sb; ++ ext3_dirent *de; ++ struct super_block * sb = dir->i_sb; + int retval; ++ unsigned short reclen = EXT3_DIR_REC_LEN(dentry->d_name.len); + +- sb = dir->i_sb; ++ unsigned nlen, rlen; ++ u32 block, blocks; ++ char *top; + +- if (!namelen) ++ if (!dentry->d_name.len) + return -EINVAL; +- bh = ext3_bread (handle, dir, 0, 0, &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; ++ if (ext3_dx && is_dx(dir)) { ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries, *at; ++ u32 hash; ++ char *data1; ++ ++ hash = dx_hash(dentry->d_name.name, dentry->d_name.len); ++ /* FIXME: do something if dx_probe() fails here */ ++ frame = dx_probe(dir, hash, frames); ++ entries = frame->entries; ++ at = frame->at; ++ ++ if (!(bh = ext3_bread(handle,dir, dx_get_block(at), 0,&retval))) ++ goto dxfail1; ++ ++ BUFFER_TRACE(bh, "get_write_access"); ++ ext3_journal_get_write_access(handle, bh); ++ ++ data1 = bh->b_data; ++ de = (ext3_dirent *) data1; ++ top = data1 + (0? 200: sb->s_blocksize); ++ while ((char *) de < top) ++ { ++ /* FIXME: check EEXIST and dir */ ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if ((de->inode? rlen - nlen: rlen) >= reclen) ++ goto dx_add; ++ de = (ext3_dirent *) ((char *) de + rlen); ++ } ++ /* 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)) ++ goto dxfull; ++ bh2 = ext3_append (handle, dir, &newblock, &retval); ++ if (!(bh2)) ++ goto dxfail2; ++ 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"); ++ ext3_journal_get_write_access(handle, frame->bh); ++ 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 */ ++ ext3_journal_get_write_access(handle, frames[0].bh); ++ ++ 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); + } +- +- ext3_debug ("creating next block\n"); +- +- BUFFER_TRACE(bh, "get_write_access"); +- ext3_journal_get_write_access(handle, bh); +- de = (struct ext3_dir_entry_2 *) bh->b_data; +- de->inode = 0; +- de->rec_len = le16_to_cpu(sb->s_blocksize); +- dir->u.ext3_i.i_disksize = +- dir->i_size = offset + sb->s_blocksize; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); ++ 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)); ++ ext3_journal_dirty_metadata(handle, bh2); ++ brelse (bh2); + } else { +- +- ext3_debug ("skipping to next block\n"); +- +- de = (struct ext3_dir_entry_2 *) bh->b_data; ++ 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; ++ ext3_journal_get_write_access(handle, frame->bh); + } ++ ext3_journal_dirty_metadata(handle, frames[0].bh); + } +- if (!ext3_check_dir_entry ("ext3_add_entry", dir, de, bh, +- offset)) { +- brelse (bh); +- return -ENOENT; +- } +- if (ext3_match (namelen, name, de)) { ++ de = do_split(handle, dir, &bh, frame, hash, &retval); ++ dx_release (frames); ++ if (!(de)) ++ goto fail; ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ goto add; ++ ++dx_add: ++ dx_release (frames); ++ goto add; ++ ++dxfull: ++ ext3_warning(sb, __FUNCTION__, "Directory index full!\n"); ++ retval = -ENOSPC; ++dxfail2: ++ brelse(bh); ++dxfail1: ++ dx_release (frames); ++ goto fail1; ++ } ++ ++ 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; ++ de = (ext3_dirent *)bh->b_data; ++ top = bh->b_data + 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(dentry->d_name.len,dentry->d_name.name,de)) { + brelse (bh); + return -EEXIST; +- } +- 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; + } +- 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); ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ if ((de->inode ? rlen - nlen: rlen) >= reclen) ++ goto add; ++ de = (ext3_dirent *)((char *)de + rlen); ++ offset += rlen; ++ } ++ if (ext3_dx && blocks == 1 && test_opt(sb, INDEX)) ++ goto dx_make_index; ++ brelse(bh); ++ } ++ bh = ext3_append(handle, dir, &block, &retval); ++ if (!bh) ++ return retval; ++ de = (ext3_dirent *) bh->b_data; ++ de->inode = 0; ++ de->rec_len = cpu_to_le16(rlen = sb->s_blocksize); ++ nlen = 0; ++ goto add; ++ ++add: ++ BUFFER_TRACE(bh, "get_write_access"); ++ ext3_journal_get_write_access(handle, bh); ++ /* By now the buffer is marked for journaling */ ++ if (de->inode) { ++ ext3_dirent *de1 = (ext3_dirent *)((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 = dentry->d_name.len; ++ memcpy (de->name, dentry->d_name.name, dentry->d_name.len); ++ /* ++ * 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"); ++ ext3_journal_dirty_metadata(handle, bh); ++ brelse(bh); ++ return 0; ++ ++dx_make_index: ++ { ++ struct buffer_head *bh2; ++ struct dx_root *root; ++ struct dx_frame frames[2], *frame; ++ struct dx_entry *entries; ++ ext3_dirent *de2; ++ char *data1; ++ unsigned len; ++ u32 hash; ++ ++ dxtrace(printk("Creating index\n")); ++ ext3_journal_get_write_access(handle, bh); ++ 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 0; ++ return retval; + } +- offset += le16_to_cpu(de->rec_len); +- de = (struct ext3_dir_entry_2 *) +- ((char *) de + le16_to_cpu(de->rec_len)); ++ data1 = bh2->b_data; ++ ++ /* The 0th block becomes the root, move the dirents out */ ++ de = (ext3_dirent *) &root->info; ++ len = ((char *) root) + sb->s_blocksize - (char *) de; ++ memcpy (data1, de, len); ++ de = (ext3_dirent *) 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 + sb->s_blocksize - (char *)de); ++ /* Initialize the root; the dot dirents already exist */ ++ de = (ext3_dirent *) (&root->dotdot); ++ de->rec_len = cpu_to_le16(sb->s_blocksize-EXT3_DIR_REC_LEN(2)); ++ memset (&root->info, 0, sizeof(root->info)); ++ root->info.info_length = sizeof(root->info); ++ 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 */ ++ hash = dx_hash (dentry->d_name.name, dentry->d_name.len); ++ frame = frames; ++ frame->entries = entries; ++ frame->at = entries; ++ frame->bh = bh; ++ bh = bh2; ++ de = do_split(handle,dir, &bh, frame, hash, &retval); ++ dx_release (frames); ++ if (!(de)) ++ return retval; ++ nlen = EXT3_DIR_REC_LEN(de->name_len); ++ rlen = le16_to_cpu(de->rec_len); ++ goto add; + } +- brelse (bh); +- return -ENOSPC; ++fail1: ++ return retval; ++fail: ++ return -ENOENT; + } + + /* +@@ -451,7 +1212,8 @@ + struct inode * inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); + if (IS_ERR(handle)) + return PTR_ERR(handle); + +@@ -478,7 +1240,8 @@ + struct inode *inode; + int err; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); + if (IS_ERR(handle)) + return PTR_ERR(handle); + +@@ -507,7 +1270,8 @@ + if (dir->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 3); ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 3); + if (IS_ERR(handle)) + return PTR_ERR(handle); + +@@ -550,7 +1320,7 @@ + if (err) + goto out_no_entry; + 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: +@@ -832,7 +1596,7 @@ + ext3_mark_inode_dirty(handle, inode); + dir->i_nlink--; + inode->i_ctime = dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + + end_rmdir: +@@ -878,7 +1642,7 @@ + if (retval) + goto end_unlink; + dir->i_ctime = dir->i_mtime = CURRENT_TIME; +- dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(dir); + ext3_mark_inode_dirty(handle, dir); + inode->i_nlink--; + if (!inode->i_nlink) +@@ -904,7 +1668,8 @@ + if (l > dir->i_sb->s_blocksize) + return -ENAMETOOLONG; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + 5); ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS + 5); + if (IS_ERR(handle)) + return PTR_ERR(handle); + +@@ -959,7 +1724,8 @@ + if (inode->i_nlink >= EXT3_LINK_MAX) + return -EMLINK; + +- handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS); ++ handle = ext3_journal_start(dir, EXT3_DATA_TRANS_BLOCKS + ++ EXT3_INDEX_EXTRA_TRANS_BLOCKS); + if (IS_ERR(handle)) + return PTR_ERR(handle); + +@@ -995,7 +1761,8 @@ + + old_bh = new_bh = dir_bh = NULL; + +- handle = ext3_journal_start(old_dir, 2 * EXT3_DATA_TRANS_BLOCKS + 2); ++ 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); + +@@ -1077,7 +1844,7 @@ + new_inode->i_ctime = CURRENT_TIME; + } + old_dir->i_ctime = old_dir->i_mtime = CURRENT_TIME; +- old_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(old_dir); + if (dir_bh) { + BUFFER_TRACE(dir_bh, "get_write_access"); + ext3_journal_get_write_access(handle, dir_bh); +@@ -1089,7 +1856,7 @@ + new_inode->i_nlink--; + } else { + new_dir->i_nlink++; +- new_dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; ++ ext3_update_dx_flag(new_dir); + ext3_mark_inode_dirty(handle, new_dir); + } + } +--- ./include/linux/ext3_fs.h 2002/03/05 06:18:59 2.1 ++++ ./include/linux/ext3_fs.h 2002/03/05 06:26:56 +@@ -339,6 +339,7 @@ + #define EXT3_MOUNT_WRITEBACK_DATA 0x0C00 /* No data ordering */ + #define EXT3_MOUNT_UPDATE_JOURNAL 0x1000 /* Update the journal format */ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ ++#define EXT3_MOUNT_INDEX 0x4000 /* Enable directory index */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H +@@ -575,6 +576,24 @@ + #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 ++ */ ++ ++#define CONFIG_EXT3_INDEX ++ ++#ifdef CONFIG_EXT3_INDEX ++ enum {ext3_dx = 1}; ++ #define is_dx(dir) (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 ++ enum {ext3_dx = 0}; ++ #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 + + #ifdef __KERNEL__ + /* +--- ./include/linux/ext3_jbd.h 2002/03/05 06:18:59 2.1 ++++ ./include/linux/ext3_jbd.h 2002/03/05 06:33:54 +@@ -63,6 +63,8 @@ + + #define EXT3_RESERVE_TRANS_BLOCKS 12 + ++#define EXT3_INDEX_EXTRA_TRANS_BLOCKS 8 ++ + int + ext3_mark_iloc_dirty(handle_t *handle, + struct inode *inode, diff --git a/lustre/kernel_patches/patches/linux-2.4.18ea-0.8.26.patch b/lustre/kernel_patches/patches/linux-2.4.18ea-0.8.26.patch new file mode 100644 index 0000000..8d52d0c --- /dev/null +++ b/lustre/kernel_patches/patches/linux-2.4.18ea-0.8.26.patch @@ -0,0 +1,1753 @@ + fs/ext3/ialloc.c | 2 + fs/ext3/inode.c | 29 - + fs/ext3/namei.c | 12 + fs/ext3/super.c | 21 + fs/ext3/xattr.c | 1247 +++++++++++++++++++++++++++++++++++++++++++++ + include/linux/ext3_fs.h | 46 - + include/linux/ext3_jbd.h | 8 + include/linux/ext3_xattr.h | 155 +++++ + include/linux/xattr.h | 15 + 9 files changed, 1486 insertions(+), 49 deletions(-) + +--- linux-2.4.18-18/fs/ext3/ialloc.c~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/fs/ext3/ialloc.c Sat Apr 5 03:03:06 2003 +@@ -17,6 +17,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -216,6 +217,7 @@ void ext3_free_inode (handle_t *handle, + * as writing the quota to disk may need the lock as well. + */ + DQUOT_INIT(inode); ++ ext3_xattr_drop_inode(handle, inode); + DQUOT_FREE_INODE(inode); + DQUOT_DROP(inode); + +--- linux-2.4.18-18/fs/ext3/inode.c~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/fs/ext3/inode.c Sat Apr 5 03:03:06 2003 +@@ -39,6 +39,18 @@ + */ + #undef SEARCH_FROM_ZERO + ++/* ++ * Test whether an inode is a fast symlink. ++ */ ++static inline int ext3_inode_is_fast_symlink(struct inode *inode) ++{ ++ int ea_blocks = EXT3_I(inode)->i_file_acl ? ++ (inode->i_sb->s_blocksize >> 9) : 0; ++ ++ return (S_ISLNK(inode->i_mode) && ++ inode->i_blocks - ea_blocks == 0); ++} ++ + /* The ext3 forget function must perform a revoke if we are freeing data + * which has been journaled. Metadata (eg. indirect blocks) must be + * revoked in all cases. +@@ -48,7 +60,7 @@ + * still needs to be revoked. + */ + +-static int ext3_forget(handle_t *handle, int is_metadata, ++int ext3_forget(handle_t *handle, int is_metadata, + struct inode *inode, struct buffer_head *bh, + int blocknr) + { +@@ -164,9 +176,7 @@ void ext3_delete_inode (struct inode * i + { + handle_t *handle; + +- if (is_bad_inode(inode) || +- inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) ++ if (is_bad_inode(inode)) + goto no_delete; + + lock_kernel(); +@@ -1861,6 +1871,8 @@ void ext3_truncate(struct inode * inode) + if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode))) + return; ++ if (ext3_inode_is_fast_symlink(inode)) ++ return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + +@@ -2008,8 +2020,6 @@ int ext3_get_inode_loc (struct inode *in + struct ext3_group_desc * gdp; + + if ((inode->i_ino != EXT3_ROOT_INO && +- inode->i_ino != EXT3_ACL_IDX_INO && +- inode->i_ino != EXT3_ACL_DATA_INO && + inode->i_ino != EXT3_JOURNAL_INO && + inode->i_ino < EXT3_FIRST_INO(inode->i_sb)) || + inode->i_ino > le32_to_cpu( +@@ -2136,10 +2146,7 @@ void ext3_read_inode(struct inode * inod + + brelse (iloc.bh); + +- if (inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) +- /* Nothing to do */ ; +- else if (S_ISREG(inode->i_mode)) { ++ if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +@@ -2147,7 +2154,7 @@ void ext3_read_inode(struct inode * inod + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; + } else if (S_ISLNK(inode->i_mode)) { +- if (!inode->i_blocks) ++ if (ext3_inode_is_fast_symlink(inode)) + inode->i_op = &ext3_fast_symlink_inode_operations; + else { + inode->i_op = &page_symlink_inode_operations; +--- linux-2.4.18-18/fs/ext3/namei.c~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/fs/ext3/namei.c Sat Apr 5 03:03:06 2003 +@@ -27,6 +27,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -1183,6 +1184,7 @@ static int ext3_add_nondir(handle_t *han + d_instantiate(dentry, inode); + return 0; + } ++ ext3_xattr_drop_inode(handle, inode); + ext3_dec_count(handle, inode); + iput(inode); + return err; +@@ -1268,15 +1270,14 @@ static int ext3_mkdir(struct inode * dir + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR); ++ inode = ext3_new_inode (handle, dir, S_IFDIR | mode); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; + + 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_blocks = 0; ++ inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { + inode->i_nlink--; /* is this nlink == 0? */ +@@ -1303,9 +1304,6 @@ static int ext3_mkdir(struct inode * dir + BUFFER_TRACE(dir_block, "call ext3_journal_dirty_metadata"); + ext3_journal_dirty_metadata(handle, dir_block); + brelse (dir_block); +- inode->i_mode = S_IFDIR | mode; +- if (dir->i_mode & S_ISGID) +- inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); + if (err) +@@ -1671,7 +1669,7 @@ 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; + /* +--- linux-2.4.18-18/fs/ext3/super.c~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/fs/ext3/super.c Sat Apr 5 03:03:06 2003 +@@ -24,6 +24,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -404,6 +405,7 @@ void ext3_put_super (struct super_block + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { + EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); +@@ -1748,14 +1750,25 @@ int ext3_statfs (struct super_block * sb + + static DECLARE_FSTYPE_DEV(ext3_fs_type, "ext3", ext3_read_super); + +-static int __init init_ext3_fs(void) ++static void exit_ext3_fs(void) + { +- return register_filesystem(&ext3_fs_type); ++ unregister_filesystem(&ext3_fs_type); ++ exit_ext3_xattr_user(); ++ exit_ext3_xattr(); + } + +-static void __exit exit_ext3_fs(void) ++static int __init init_ext3_fs(void) + { +- unregister_filesystem(&ext3_fs_type); ++ int error = init_ext3_xattr(); ++ if (!error) ++ error = init_ext3_xattr_user(); ++ if (!error) ++ error = register_filesystem(&ext3_fs_type); ++ if (!error) ++ return 0; ++ ++ exit_ext3_fs(); ++ return error; + } + + EXPORT_SYMBOL(ext3_bread); +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.18-18-braam/fs/ext3/xattr.c Sat Apr 5 02:54:56 2003 +@@ -0,0 +1,1247 @@ ++/* ++ * linux/fs/ext3/xattr.c ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ * ++ * Fix by Harrison Xing . ++ * Ext3 code with a lot of help from Eric Jarman . ++ * Extended attributes for symlinks and special files added per ++ * suggestion of Luka Renko . ++ */ ++ ++/* ++ * Extended attributes are stored on disk blocks allocated outside of ++ * any inode. The i_file_acl field is then made to point to this allocated ++ * block. If all extended attributes of an inode are identical, these ++ * inodes may share the same extended attribute block. Such situations ++ * are automatically detected by keeping a cache of recent attribute block ++ * numbers and hashes over the block's contents in memory. ++ * ++ * ++ * Extended attribute block layout: ++ * ++ * +------------------+ ++ * | header | ++ * ¦ entry 1 | | ++ * | entry 2 | | growing downwards ++ * | entry 3 | v ++ * | four null bytes | ++ * | . . . | ++ * | value 1 | ^ ++ * | value 3 | | growing upwards ++ * | value 2 | | ++ * +------------------+ ++ * ++ * The block header is followed by multiple entry descriptors. These entry ++ * descriptors are variable in size, and alligned to EXT3_XATTR_PAD ++ * byte boundaries. The entry descriptors are sorted by attribute name, ++ * so that two extended attribute blocks can be compared efficiently. ++ * ++ * Attribute values are aligned to the end of the block, stored in ++ * no specific order. They are also padded to EXT3_XATTR_PAD byte ++ * boundaries. No additional gaps are left between them. ++ * ++ * Locking strategy ++ * ---------------- ++ * The VFS already holds the BKL and the inode->i_sem semaphore when any of ++ * the xattr inode operations are called, so we are guaranteed that only one ++ * processes accesses extended attributes of an inode at any time. ++ * ++ * For writing we also grab the ext3_xattr_sem semaphore. This ensures that ++ * only a single process is modifying an extended attribute block, even ++ * if the block is shared among inodes. ++ * ++ * Note for porting to 2.5 ++ * ----------------------- ++ * The BKL will no longer be held in the xattr inode operations. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++#include ++#endif ++#include ++#include ++#include ++#include ++ ++/* These symbols may be needed by a module. */ ++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); ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++# define mark_buffer_dirty(bh) mark_buffer_dirty(bh, 1) ++#endif ++ ++#define HDR(bh) ((struct ext3_xattr_header *)((bh)->b_data)) ++#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr)) ++#define FIRST_ENTRY(bh) ENTRY(HDR(bh)+1) ++#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) ++ ++#ifdef EXT3_XATTR_DEBUG ++# define ea_idebug(inode, f...) do { \ ++ printk(KERN_DEBUG "inode %s:%ld: ", \ ++ kdevname(inode->i_dev), inode->i_ino); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++# define ea_bdebug(bh, f...) do { \ ++ printk(KERN_DEBUG "block %s:%ld: ", \ ++ kdevname(bh->b_dev), bh->b_blocknr); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++#else ++# define ea_idebug(f...) ++# define ea_bdebug(f...) ++#endif ++ ++static int ext3_xattr_set2(handle_t *, struct inode *, struct buffer_head *, ++ struct ext3_xattr_header *); ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++static int ext3_xattr_cache_insert(struct buffer_head *); ++static struct buffer_head *ext3_xattr_cache_find(struct inode *, ++ struct ext3_xattr_header *); ++static void ext3_xattr_cache_remove(struct buffer_head *); ++static void ext3_xattr_rehash(struct ext3_xattr_header *, ++ struct ext3_xattr_entry *); ++ ++static struct mb_cache *ext3_xattr_cache; ++ ++#else ++# define ext3_xattr_cache_insert(bh) 0 ++# define ext3_xattr_cache_find(inode, header) NULL ++# define ext3_xattr_cache_remove(bh) do {} while(0) ++# define ext3_xattr_rehash(header, entry) do {} while(0) ++#endif ++ ++/* ++ * If a file system does not share extended attributes among inodes, ++ * we should not need the ext3_xattr_sem semaphore. However, the ++ * filesystem may still contain shared blocks, so we always take ++ * the lock. ++ */ ++ ++DECLARE_MUTEX(ext3_xattr_sem); ++ ++static inline void ++ext3_xattr_lock(void) ++{ ++ down(&ext3_xattr_sem); ++} ++ ++static inline void ++ext3_xattr_unlock(void) ++{ ++ up(&ext3_xattr_sem); ++} ++ ++static inline int ++ext3_xattr_new_block(handle_t *handle, struct inode *inode, ++ int * errp, int force) ++{ ++ struct super_block *sb = inode->i_sb; ++ int goal = le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block) + ++ EXT3_I(inode)->i_block_group * EXT3_BLOCKS_PER_GROUP(sb); ++ ++ /* How can we enforce the allocation? */ ++ int block = ext3_new_block(handle, inode, goal, 0, 0, errp); ++#ifdef OLD_QUOTAS ++ if (!*errp) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#endif ++ return block; ++} ++ ++static inline int ++ext3_xattr_quota_alloc(struct inode *inode, int force) ++{ ++ /* How can we enforce the allocation? */ ++#ifdef OLD_QUOTAS ++ int error = DQUOT_ALLOC_BLOCK(inode->i_sb, inode, 1); ++ if (!error) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#else ++ int error = DQUOT_ALLOC_BLOCK(inode, 1); ++#endif ++ return error; ++} ++ ++#ifdef OLD_QUOTAS ++ ++static inline void ++ext3_xattr_quota_free(struct inode *inode) ++{ ++ DQUOT_FREE_BLOCK(inode->i_sb, inode, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++static inline void ++ext3_xattr_free_block(handle_t *handle, struct inode * inode, ++ unsigned long block) ++{ ++ ext3_free_blocks(handle, inode, block, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++#else ++# define ext3_xattr_quota_free(inode) \ ++ DQUOT_FREE_BLOCK(inode, 1) ++# define ext3_xattr_free_block(handle, inode, block) \ ++ ext3_free_blocks(handle, inode, block, 1) ++#endif ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18) ++ ++static inline struct buffer_head * ++sb_bread(struct super_block *sb, int block) ++{ ++ return bread(sb->s_dev, block, sb->s_blocksize); ++} ++ ++static inline struct buffer_head * ++sb_getblk(struct super_block *sb, int block) ++{ ++ return getblk(sb->s_dev, block, sb->s_blocksize); ++} ++ ++#endif ++ ++struct ext3_xattr_handler *ext3_xattr_handlers[EXT3_XATTR_INDEX_MAX]; ++rwlock_t ext3_handler_lock = RW_LOCK_UNLOCKED; ++ ++int ++ext3_xattr_register(int name_index, struct ext3_xattr_handler *handler) ++{ ++ int error = -EINVAL; ++ ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ if (!ext3_xattr_handlers[name_index-1]) { ++ ext3_xattr_handlers[name_index-1] = handler; ++ error = 0; ++ } ++ write_unlock(&ext3_handler_lock); ++ } ++ return error; ++} ++ ++void ++ext3_xattr_unregister(int name_index, struct ext3_xattr_handler *handler) ++{ ++ if (name_index > 0 || name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ ext3_xattr_handlers[name_index-1] = NULL; ++ write_unlock(&ext3_handler_lock); ++ } ++} ++ ++static inline const char * ++strcmp_prefix(const char *a, const char *a_prefix) ++{ ++ while (*a_prefix && *a == *a_prefix) { ++ a++; ++ a_prefix++; ++ } ++ return *a_prefix ? NULL : a; ++} ++ ++/* ++ * Decode the extended attribute name, and translate it into ++ * the name_index and name suffix. ++ */ ++static inline struct ext3_xattr_handler * ++ext3_xattr_resolve_name(const char **name) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ int i; ++ ++ if (!*name) ++ return NULL; ++ read_lock(&ext3_handler_lock); ++ for (i=0; iprefix); ++ if (n) { ++ handler = ext3_xattr_handlers[i]; ++ *name = n; ++ break; ++ } ++ } ++ } ++ read_unlock(&ext3_handler_lock); ++ return handler; ++} ++ ++static inline struct ext3_xattr_handler * ++ext3_xattr_handler(int name_index) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ read_lock(&ext3_handler_lock); ++ handler = ext3_xattr_handlers[name_index-1]; ++ read_unlock(&ext3_handler_lock); ++ } ++ return handler; ++} ++ ++/* ++ * Inode operation getxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_getxattr(struct dentry *dentry, const char *name, ++ void *buffer, size_t size) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->get(inode, name, buffer, size); ++} ++ ++/* ++ * Inode operation listxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_listxattr(struct dentry *dentry, char *buffer, size_t size) ++{ ++ return ext3_xattr_list(dentry->d_inode, buffer, size); ++} ++ ++/* ++ * Inode operation setxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_setxattr(struct dentry *dentry, const char *name, ++ void *value, size_t size, int flags) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ if (size == 0) ++ value = ""; /* empty EA, do not remove */ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, value, size, flags); ++} ++ ++/* ++ * Inode operation removexattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_removexattr(struct dentry *dentry, const char *name) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, NULL, 0, XATTR_REPLACE); ++} ++ ++/* ++ * ext3_xattr_get() ++ * ++ * Copy an extended attribute into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size; ++ char *end; ++ int name_len, error; ++ ++ ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", ++ name_index, name, buffer, (long)buffer_size); ++ ++ if (name == NULL) ++ return -EINVAL; ++ if (!EXT3_I(inode)->i_file_acl) ++ return -ENOATTR; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ ++ error = -ERANGE; ++ if (name_len > 255) ++ goto cleanup; ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) ++ goto found; ++ entry = next; ++ } ++ /* Check the remaining name entries */ ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ entry = next; ++ } ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ error = -ENOATTR; ++ goto cleanup; ++found: ++ /* check the buffer size */ ++ if (entry->e_value_block != 0) ++ goto bad_block; ++ size = le32_to_cpu(entry->e_value_size); ++ if (size > inode->i_sb->s_blocksize || ++ le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize) ++ goto bad_block; ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (buffer) { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ /* return value of attribute */ ++ memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), ++ size); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size = 0; ++ char *buf, *end; ++ int error; ++ ++ ea_idebug(inode, "buffer=%p, buffer_size=%ld", ++ buffer, (long)buffer_size); ++ ++ if (!EXT3_I(inode)->i_file_acl) ++ return 0; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_list", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* compute the size required for the list of attribute names */ ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) { ++ size += handler->list(NULL, inode, entry->e_name, ++ entry->e_name_len) + 1; ++ } ++ } ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (!buffer) { ++ error = size; ++ goto cleanup; ++ } else { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ /* list the attribute names */ ++ buf = buffer; ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) { ++ buf += handler->list(buf, inode, entry->e_name, ++ entry->e_name_len); ++ *buf++ = '\0'; ++ } ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is ++ * not set, set it. ++ */ ++static void ext3_xattr_update_super_block(handle_t *handle, ++ struct super_block *sb) ++{ ++ if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR)) ++ return; ++ ++ lock_super(sb); ++ ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++ EXT3_SB(sb)->s_feature_compat |= EXT3_FEATURE_COMPAT_EXT_ATTR; ++#endif ++ EXT3_SB(sb)->s_es->s_feature_compat |= ++ cpu_to_le32(EXT3_FEATURE_COMPAT_EXT_ATTR); ++ sb->s_dirt = 1; ++ ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); ++ unlock_super(sb); ++} ++ ++/* ++ * ext3_xattr_set() ++ * ++ * Create, replace or remove an extended attribute for this inode. Buffer ++ * is NULL to remove an existing extended attribute, and non-NULL to ++ * either replace an existing extended attribute, or create a new extended ++ * attribute. The flags XATTR_REPLACE and XATTR_CREATE ++ * specify that an extended attribute must exist and must not exist ++ * previous to the call, respectively. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, void *value, size_t value_len, int flags) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_header *header = NULL; ++ struct ext3_xattr_entry *here, *last; ++ unsigned int name_len; ++ int min_offs = sb->s_blocksize, not_found = 1, free, error; ++ char *end; ++ ++ /* ++ * header -- Points either into bh, or to a temporarily ++ * allocated buffer. ++ * here -- The named entry found, or the place for inserting, within ++ * the block pointed to by header. ++ * last -- Points right after the last named entry within the block ++ * pointed to by header. ++ * min_offs -- The offset of the first value (values are aligned ++ * towards the end of the block). ++ * end -- Points right after the block pointed to by header. ++ */ ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > sb->s_blocksize) ++ return -ERANGE; ++ ext3_xattr_lock(); ++ ++ if (EXT3_I(inode)->i_file_acl) { ++ /* The inode already has an extended attribute block. */ ++ int block = EXT3_I(inode)->i_file_acl; ++ ++ bh = sb_bread(sb, block); ++ error = -EIO; ++ if (!bh) ++ goto cleanup; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), ++ le32_to_cpu(HDR(bh)->h_refcount)); ++ header = HDR(bh); ++ end = bh->b_data + bh->b_size; ++ if (header->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ header->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(sb, "ext3_xattr_set", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* Find the named attribute. */ ++ here = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(here)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(here); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!here->e_value_block && here->e_value_size) { ++ int offs = le16_to_cpu(here->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ not_found = name_index - here->e_name_index; ++ if (!not_found) ++ not_found = name_len - here->e_name_len; ++ if (!not_found) ++ not_found = memcmp(name, here->e_name,name_len); ++ if (not_found <= 0) ++ break; ++ here = next; ++ } ++ last = here; ++ /* We still need to compute min_offs and last. */ ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!last->e_value_block && last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ last = next; ++ } ++ ++ /* Check whether we have enough space left. */ ++ free = min_offs - ((char*)last - (char*)header) - sizeof(__u32); ++ } else { ++ /* We will use a new extended attribute block. */ ++ free = sb->s_blocksize - ++ sizeof(struct ext3_xattr_header) - sizeof(__u32); ++ here = last = NULL; /* avoid gcc uninitialized warning. */ ++ } ++ ++ if (not_found) { ++ /* Request to remove a nonexistent attribute? */ ++ error = -ENOATTR; ++ if (flags & XATTR_REPLACE) ++ goto cleanup; ++ error = 0; ++ if (value == NULL) ++ goto cleanup; ++ else ++ free -= EXT3_XATTR_LEN(name_len); ++ } else { ++ /* Request to create an existing attribute? */ ++ error = -EEXIST; ++ if (flags & XATTR_CREATE) ++ goto cleanup; ++ if (!here->e_value_block && here->e_value_size) { ++ unsigned int size = le32_to_cpu(here->e_value_size); ++ ++ if (le16_to_cpu(here->e_value_offs) + size > ++ sb->s_blocksize || size > sb->s_blocksize) ++ goto bad_block; ++ free += EXT3_XATTR_SIZE(size); ++ } ++ } ++ free -= EXT3_XATTR_SIZE(value_len); ++ error = -ENOSPC; ++ if (free < 0) ++ goto cleanup; ++ ++ /* Here we know that we can set the new attribute. */ ++ ++ if (header) { ++ if (header->h_refcount == cpu_to_le32(1)) { ++ ea_bdebug(bh, "modifying in-place"); ++ ext3_xattr_cache_remove(bh); ++ error = ext3_journal_get_write_access(handle, bh); ++ if (error) ++ goto cleanup; ++ } else { ++ int offset; ++ ++ ea_bdebug(bh, "cloning"); ++ header = kmalloc(bh->b_size, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memcpy(header, HDR(bh), bh->b_size); ++ header->h_refcount = cpu_to_le32(1); ++ offset = (char *)header - bh->b_data; ++ here = ENTRY((char *)here + offset); ++ last = ENTRY((char *)last + offset); ++ } ++ } else { ++ /* Allocate a buffer where we construct the new block. */ ++ header = kmalloc(sb->s_blocksize, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memset(header, 0, sb->s_blocksize); ++ end = (char *)header + sb->s_blocksize; ++ header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC); ++ header->h_blocks = header->h_refcount = cpu_to_le32(1); ++ last = here = ENTRY(header+1); ++ } ++ ++ if (not_found) { ++ /* Insert the new name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ int rest = (char *)last - (char *)here; ++ memmove((char *)here + size, here, rest); ++ memset(here, 0, size); ++ here->e_name_index = name_index; ++ here->e_name_len = name_len; ++ memcpy(here->e_name, name, name_len); ++ } else { ++ /* Remove the old value. */ ++ if (!here->e_value_block && here->e_value_size) { ++ char *first_val = (char *)header + min_offs; ++ int offs = le16_to_cpu(here->e_value_offs); ++ char *val = (char *)header + offs; ++ size_t size = EXT3_XATTR_SIZE( ++ le32_to_cpu(here->e_value_size)); ++ memmove(first_val + size, first_val, val - first_val); ++ memset(first_val, 0, size); ++ here->e_value_offs = 0; ++ min_offs += size; ++ ++ /* Adjust all value offsets. */ ++ last = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(last)) { ++ int o = le16_to_cpu(last->e_value_offs); ++ if (!last->e_value_block && o < offs) ++ last->e_value_offs = ++ cpu_to_le16(o + size); ++ last = EXT3_XATTR_NEXT(last); ++ } ++ } ++ if (value == NULL) { ++ /* Remove this attribute. */ ++ if (EXT3_XATTR_NEXT(ENTRY(header+1)) == last) { ++ /* This block is now empty. */ ++ error = ext3_xattr_set2(handle, inode, bh,NULL); ++ goto cleanup; ++ } else { ++ /* Remove the old name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ last = ENTRY((char *)last - size); ++ memmove(here, (char*)here + size, ++ (char*)last - (char*)here); ++ memset(last, 0, size); ++ } ++ } ++ } ++ ++ if (value != NULL) { ++ /* Insert the new value. */ ++ here->e_value_size = cpu_to_le32(value_len); ++ if (value_len) { ++ size_t size = EXT3_XATTR_SIZE(value_len); ++ char *val = (char *)header + min_offs - size; ++ here->e_value_offs = ++ cpu_to_le16((char *)val - (char *)header); ++ memset(val + size - EXT3_XATTR_PAD, 0, ++ EXT3_XATTR_PAD); /* Clear the pad bytes. */ ++ memcpy(val, value, value_len); ++ } ++ } ++ ext3_xattr_rehash(header, here); ++ ++ error = ext3_xattr_set2(handle, inode, bh, header); ++ ++cleanup: ++ brelse(bh); ++ if (!(bh && header == HDR(bh))) ++ kfree(header); ++ ext3_xattr_unlock(); ++ ++ return error; ++} ++ ++/* ++ * Second half of ext3_xattr_set(): Update the file system. ++ */ ++static int ++ext3_xattr_set2(handle_t *handle, struct inode *inode, ++ struct buffer_head *old_bh, struct ext3_xattr_header *header) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *new_bh = NULL; ++ int error; ++ ++ if (header) { ++ new_bh = ext3_xattr_cache_find(inode, header); ++ if (new_bh) { ++ /* ++ * We found an identical block in the cache. ++ * The old block will be released after updating ++ * the inode. ++ */ ++ ea_bdebug(old_bh, "reusing block %ld", ++ new_bh->b_blocknr); ++ ++ error = -EDQUOT; ++ if (ext3_xattr_quota_alloc(inode, 1)) ++ goto cleanup; ++ ++ error = ext3_journal_get_write_access(handle, new_bh); ++ if (error) ++ goto cleanup; ++ HDR(new_bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(new_bh)->h_refcount) + 1); ++ ea_bdebug(new_bh, "refcount now=%d", ++ le32_to_cpu(HDR(new_bh)->h_refcount)); ++ } else if (old_bh && header == HDR(old_bh)) { ++ /* Keep this block. */ ++ new_bh = old_bh; ++ (void)ext3_xattr_cache_insert(new_bh); ++ } else { ++ /* We need to allocate a new block */ ++ int force = EXT3_I(inode)->i_file_acl != 0; ++ int block = ext3_xattr_new_block(handle, inode, ++ &error, force); ++ if (error) ++ goto cleanup; ++ ea_idebug(inode, "creating block %d", block); ++ ++ new_bh = sb_getblk(sb, block); ++ if (!new_bh) { ++getblk_failed: ext3_xattr_free_block(handle, inode, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ lock_buffer(new_bh); ++ error = ext3_journal_get_create_access(handle, new_bh); ++ if (error) { ++ unlock_buffer(new_bh); ++ goto getblk_failed; ++ } ++ memcpy(new_bh->b_data, header, new_bh->b_size); ++ mark_buffer_uptodate(new_bh, 1); ++ unlock_buffer(new_bh); ++ (void)ext3_xattr_cache_insert(new_bh); ++ ext3_xattr_update_super_block(handle, sb); ++ } ++ error = ext3_journal_dirty_metadata(handle, new_bh); ++ if (error) ++ goto cleanup; ++ } ++ ++ /* Update the inode. */ ++ EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; ++ inode->i_ctime = CURRENT_TIME; ++ ext3_mark_inode_dirty(handle, inode); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ++ error = 0; ++ if (old_bh && old_bh != new_bh) { ++ /* ++ * If there was an old block, and we are not still using it, ++ * we now release the old block. ++ */ ++ unsigned int refcount = le32_to_cpu(HDR(old_bh)->h_refcount); ++ ++ error = ext3_journal_get_write_access(handle, old_bh); ++ if (error) ++ goto cleanup; ++ if (refcount == 1) { ++ /* Free the old block. */ ++ ea_bdebug(old_bh, "freeing"); ++ ext3_xattr_free_block(handle, inode, old_bh->b_blocknr); ++ ++ /* ext3_forget() calls bforget() for us, but we ++ let our caller release old_bh, so we need to ++ duplicate the handle before. */ ++ get_bh(old_bh); ++ ext3_forget(handle, 1, inode, old_bh,old_bh->b_blocknr); ++ } else { ++ /* Decrement the refcount only. */ ++ refcount--; ++ HDR(old_bh)->h_refcount = cpu_to_le32(refcount); ++ ext3_xattr_quota_free(inode); ++ ext3_journal_dirty_metadata(handle, old_bh); ++ ea_bdebug(old_bh, "refcount now=%d", refcount); ++ } ++ } ++ ++cleanup: ++ if (old_bh != new_bh) ++ brelse(new_bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_drop_inode() ++ * ++ * Free extended attribute resources associated with this inode. This ++ * is called immediately before an inode is freed. ++ */ ++void ++ext3_xattr_drop_inode(handle_t *handle, struct inode *inode) ++{ ++ struct buffer_head *bh; ++ unsigned int block = EXT3_I(inode)->i_file_acl; ++ ++ if (!block) ++ return; ++ ext3_xattr_lock(); ++ ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_drop_inode", ++ "inode %ld: block %d read error", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "b_count=%d", atomic_read(&(bh->b_count))); ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++ ext3_error(inode->i_sb, "ext3_xattr_drop_inode", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ goto cleanup; ++ } ++ ext3_journal_get_write_access(handle, bh); ++ ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ if (HDR(bh)->h_refcount == cpu_to_le32(1)) { ++ ext3_xattr_cache_remove(bh); ++ ext3_xattr_free_block(handle, inode, block); ++ ext3_forget(handle, 1, inode, bh, block); ++ bh = NULL; ++ } else { ++ HDR(bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ ext3_journal_dirty_metadata(handle, bh); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ext3_xattr_quota_free(inode); ++ } ++ EXT3_I(inode)->i_file_acl = 0; ++ ++cleanup: ++ brelse(bh); ++ ext3_xattr_unlock(); ++} ++ ++/* ++ * ext3_xattr_put_super() ++ * ++ * This is called when a file system is unmounted. ++ */ ++void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ mb_cache_shrink(ext3_xattr_cache, sb->s_dev); ++#endif ++} ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++/* ++ * ext3_xattr_cache_insert() ++ * ++ * Create a new entry in the extended attribute cache, and insert ++ * it unless such an entry is already in the cache. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++static int ++ext3_xattr_cache_insert(struct buffer_head *bh) ++{ ++ __u32 hash = le32_to_cpu(HDR(bh)->h_hash); ++ struct mb_cache_entry *ce; ++ int error; ++ ++ ce = mb_cache_entry_alloc(ext3_xattr_cache); ++ if (!ce) ++ return -ENOMEM; ++ error = mb_cache_entry_insert(ce, bh->b_dev, bh->b_blocknr, &hash); ++ if (error) { ++ mb_cache_entry_free(ce); ++ if (error == -EBUSY) { ++ ea_bdebug(bh, "already in cache (%d cache entries)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ error = 0; ++ } ++ } else { ++ ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash, ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ mb_cache_entry_release(ce); ++ } ++ return error; ++} ++ ++/* ++ * ext3_xattr_cmp() ++ * ++ * Compare two extended attribute blocks for equality. ++ * ++ * Returns 0 if the blocks are equal, 1 if they differ, and ++ * a negative error number on errors. ++ */ ++static int ++ext3_xattr_cmp(struct ext3_xattr_header *header1, ++ struct ext3_xattr_header *header2) ++{ ++ struct ext3_xattr_entry *entry1, *entry2; ++ ++ entry1 = ENTRY(header1+1); ++ entry2 = ENTRY(header2+1); ++ while (!IS_LAST_ENTRY(entry1)) { ++ if (IS_LAST_ENTRY(entry2)) ++ return 1; ++ if (entry1->e_hash != entry2->e_hash || ++ entry1->e_name_len != entry2->e_name_len || ++ entry1->e_value_size != entry2->e_value_size || ++ memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) ++ return 1; ++ if (entry1->e_value_block != 0 || entry2->e_value_block != 0) ++ return -EIO; ++ if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), ++ (char *)header2 + le16_to_cpu(entry2->e_value_offs), ++ le32_to_cpu(entry1->e_value_size))) ++ return 1; ++ ++ entry1 = EXT3_XATTR_NEXT(entry1); ++ entry2 = EXT3_XATTR_NEXT(entry2); ++ } ++ if (!IS_LAST_ENTRY(entry2)) ++ return 1; ++ return 0; ++} ++ ++/* ++ * ext3_xattr_cache_find() ++ * ++ * Find an identical extended attribute block. ++ * ++ * Returns a pointer to the block found, or NULL if such a block was ++ * not found or an error occurred. ++ */ ++static struct buffer_head * ++ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header) ++{ ++ __u32 hash = le32_to_cpu(header->h_hash); ++ struct mb_cache_entry *ce; ++ ++ if (!header->h_hash) ++ return NULL; /* never share */ ++ ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); ++ ce = mb_cache_entry_find_first(ext3_xattr_cache, 0, inode->i_dev, hash); ++ while (ce) { ++ struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block); ++ ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_cache_find", ++ "inode %ld: block %ld read error", ++ inode->i_ino, ce->e_block); ++ } else if (le32_to_cpu(HDR(bh)->h_refcount) > ++ EXT3_XATTR_REFCOUNT_MAX) { ++ ea_idebug(inode, "block %ld refcount %d>%d",ce->e_block, ++ le32_to_cpu(HDR(bh)->h_refcount), ++ EXT3_XATTR_REFCOUNT_MAX); ++ } else if (!ext3_xattr_cmp(header, HDR(bh))) { ++ ea_bdebug(bh, "b_count=%d",atomic_read(&(bh->b_count))); ++ mb_cache_entry_release(ce); ++ return bh; ++ } ++ brelse(bh); ++ ce = mb_cache_entry_find_next(ce, 0, inode->i_dev, hash); ++ } ++ return NULL; ++} ++ ++/* ++ * ext3_xattr_cache_remove() ++ * ++ * Remove the cache entry of a block from the cache. Called when a ++ * block becomes invalid. ++ */ ++static void ++ext3_xattr_cache_remove(struct buffer_head *bh) ++{ ++ struct mb_cache_entry *ce; ++ ++ ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_dev, bh->b_blocknr); ++ if (ce) { ++ ea_bdebug(bh, "removing (%d cache entries remaining)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)-1); ++ mb_cache_entry_free(ce); ++ } else ++ ea_bdebug(bh, "no cache entry"); ++} ++ ++#define NAME_HASH_SHIFT 5 ++#define VALUE_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_hash_entry() ++ * ++ * Compute the hash of an extended attribute. ++ */ ++static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ __u32 hash = 0; ++ char *name = entry->e_name; ++ int n; ++ ++ for (n=0; n < entry->e_name_len; n++) { ++ hash = (hash << NAME_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ ++ *name++; ++ } ++ ++ if (entry->e_value_block == 0 && entry->e_value_size != 0) { ++ __u32 *value = (__u32 *)((char *)header + ++ le16_to_cpu(entry->e_value_offs)); ++ for (n = (le32_to_cpu(entry->e_value_size) + ++ EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) { ++ hash = (hash << VALUE_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ ++ le32_to_cpu(*value++); ++ } ++ } ++ entry->e_hash = cpu_to_le32(hash); ++} ++ ++#undef NAME_HASH_SHIFT ++#undef VALUE_HASH_SHIFT ++ ++#define BLOCK_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_rehash() ++ * ++ * Re-compute the extended attribute hash value after an entry has changed. ++ */ ++static void ext3_xattr_rehash(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ struct ext3_xattr_entry *here; ++ __u32 hash = 0; ++ ++ ext3_xattr_hash_entry(header, entry); ++ here = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(here)) { ++ if (!here->e_hash) { ++ /* Block is not shared if an entry's hash value == 0 */ ++ hash = 0; ++ break; ++ } ++ hash = (hash << BLOCK_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ ++ le32_to_cpu(here->e_hash); ++ here = EXT3_XATTR_NEXT(here); ++ } ++ header->h_hash = cpu_to_le32(hash); ++} ++ ++#undef BLOCK_HASH_SHIFT ++ ++int __init ++init_ext3_xattr(void) ++{ ++ ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL, ++ sizeof(struct mb_cache_entry) + ++ sizeof(struct mb_cache_entry_index), 1, 61); ++ if (!ext3_xattr_cache) ++ return -ENOMEM; ++ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++ if (ext3_xattr_cache) ++ mb_cache_destroy(ext3_xattr_cache); ++ ext3_xattr_cache = NULL; ++} ++ ++#else /* CONFIG_EXT3_FS_XATTR_SHARING */ ++ ++int __init ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_SHARING */ +--- linux-2.4.18-18/include/linux/ext3_fs.h~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/include/linux/ext3_fs.h Sat Apr 5 03:03:06 2003 +@@ -58,8 +58,6 @@ + */ + #define EXT3_BAD_INO 1 /* Bad blocks inode */ + #define EXT3_ROOT_INO 2 /* Root inode */ +-#define EXT3_ACL_IDX_INO 3 /* ACL inode */ +-#define EXT3_ACL_DATA_INO 4 /* ACL inode */ + #define EXT3_BOOT_LOADER_INO 5 /* Boot loader inode */ + #define EXT3_UNDEL_DIR_INO 6 /* Undelete directory inode */ + #define EXT3_RESIZE_INO 7 /* Reserved group descriptors inode */ +@@ -89,7 +87,6 @@ + #else + # define EXT3_BLOCK_SIZE(s) (EXT3_MIN_BLOCK_SIZE << (s)->s_log_block_size) + #endif +-#define EXT3_ACLE_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (struct ext3_acl_entry)) + #define EXT3_ADDR_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (__u32)) + #ifdef __KERNEL__ + # define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -124,28 +121,6 @@ + #endif + + /* +- * ACL structures +- */ +-struct ext3_acl_header /* Header of Access Control Lists */ +-{ +- __u32 aclh_size; +- __u32 aclh_file_count; +- __u32 aclh_acle_count; +- __u32 aclh_first_acle; +-}; +- +-struct ext3_acl_entry /* Access Control List Entry */ +-{ +- __u32 acle_size; +- __u16 acle_perms; /* Access permissions */ +- __u16 acle_type; /* Type of entry */ +- __u16 acle_tag; /* User or group identity */ +- __u16 acle_pad1; +- __u32 acle_next; /* Pointer on next entry for the */ +- /* same inode or on next free entry */ +-}; +- +-/* + * Structure of a blocks group descriptor + */ + struct ext3_group_desc +@@ -513,7 +488,7 @@ struct ext3_super_block { + #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ + +-#define EXT3_FEATURE_COMPAT_SUPP 0 ++#define EXT3_FEATURE_COMPAT_SUPP EXT3_FEATURE_COMPAT_EXT_ATTR + #define EXT3_FEATURE_INCOMPAT_SUPP (EXT3_FEATURE_INCOMPAT_FILETYPE| \ + EXT3_FEATURE_INCOMPAT_RECOVER) + #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER| \ +@@ -606,6 +581,24 @@ struct ext3_iloc + unsigned long block_group; + }; + ++/* Defined for extended attributes */ ++#define CONFIG_EXT3_FS_XATTR y ++#ifndef ENOATTR ++#define ENOATTR ENODATA /* No such attribute */ ++#endif ++#ifndef ENOTSUP ++#define ENOTSUP EOPNOTSUPP /* Operation not supported */ ++#endif ++#ifndef XATTR_NAME_MAX ++#define XATTR_NAME_MAX 255 /* # chars in an extended attribute name */ ++#define XATTR_SIZE_MAX 65536 /* size of an extended attribute value (64k) */ ++#define XATTR_LIST_MAX 65536 /* size of extended attribute namelist (64k) */ ++#endif ++#ifndef XATTR_CREATE ++#define XATTR_CREATE 1 /* set value, fail if attr already exists */ ++#define XATTR_REPLACE 2 /* set value, fail if attr does not exist */ ++#endif ++ + /* + * Function prototypes + */ +@@ -647,6 +640,7 @@ extern void ext3_check_inodes_bitmap (st + extern unsigned long ext3_count_free (struct buffer_head *, unsigned); + + /* inode.c */ ++extern int ext3_forget(handle_t *, int, struct inode *, struct buffer_head *, int); + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + +--- linux-2.4.18-18/include/linux/ext3_jbd.h~linux-2.4.18ea-0.8.26 Sat Apr 5 02:54:56 2003 ++++ linux-2.4.18-18-braam/include/linux/ext3_jbd.h Sat Apr 5 03:03:06 2003 +@@ -30,13 +30,19 @@ + + #define EXT3_SINGLEDATA_TRANS_BLOCKS 8 + ++/* Extended attributes may touch two data buffers, two bitmap buffers, ++ * and two group and summaries. */ ++ ++#define EXT3_XATTR_TRANS_BLOCKS 8 ++ + /* Define the minimum size for a transaction which modifies data. This + * needs to take into account the fact that we may end up modifying two + * quota files too (one for the group, one for the user quota). The + * superblock only gets updated once, of course, so don't bother + * counting that again for the quota updates. */ + +-#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS - 2) ++#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS + \ ++ EXT3_XATTR_TRANS_BLOCKS - 2) + + extern int ext3_writepage_trans_blocks(struct inode *inode); + +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.18-18-braam/include/linux/ext3_xattr.h Sat Apr 5 02:54:56 2003 +@@ -0,0 +1,155 @@ ++/* ++ File: linux/ext3_xattr.h ++ ++ On-disk format of extended attributes for the ext3 filesystem. ++ ++ (C) 2001 Andreas Gruenbacher, ++*/ ++ ++#include ++#include ++#include ++ ++/* Magic value in attribute blocks */ ++#define EXT3_XATTR_MAGIC 0xEA020000 ++ ++/* Maximum number of references to one attribute block */ ++#define EXT3_XATTR_REFCOUNT_MAX 1024 ++ ++/* Name indexes */ ++#define EXT3_XATTR_INDEX_MAX 10 ++#define EXT3_XATTR_INDEX_USER 1 ++ ++struct ext3_xattr_header { ++ __u32 h_magic; /* magic number for identification */ ++ __u32 h_refcount; /* reference count */ ++ __u32 h_blocks; /* number of disk blocks used */ ++ __u32 h_hash; /* hash value of all attributes */ ++ __u32 h_reserved[4]; /* zero right now */ ++}; ++ ++struct ext3_xattr_entry { ++ __u8 e_name_len; /* length of name */ ++ __u8 e_name_index; /* attribute name index */ ++ __u16 e_value_offs; /* offset in disk block of value */ ++ __u32 e_value_block; /* disk block attribute is stored on (n/i) */ ++ __u32 e_value_size; /* size of attribute value */ ++ __u32 e_hash; /* hash value of name and value */ ++ char e_name[0]; /* attribute name */ ++}; ++ ++#define EXT3_XATTR_PAD_BITS 2 ++#define EXT3_XATTR_PAD (1<e_name_len)) ) ++#define EXT3_XATTR_SIZE(size) \ ++ (((size) + EXT3_XATTR_ROUND) & ~EXT3_XATTR_ROUND) ++ ++#ifdef __KERNEL__ ++ ++# ifdef CONFIG_EXT3_FS_XATTR ++ ++struct ext3_xattr_handler { ++ char *prefix; ++ size_t (*list)(char *list, struct inode *inode, const char *name, ++ int name_len); ++ int (*get)(struct inode *inode, const char *name, void *buffer, ++ size_t size); ++ int (*set)(struct inode *inode, const char *name, void *buffer, ++ size_t size, int flags); ++}; ++ ++extern int ext3_xattr_register(int, struct ext3_xattr_handler *); ++extern void ext3_xattr_unregister(int, struct ext3_xattr_handler *); ++ ++extern int ext3_setxattr(struct dentry *, const char *, void *, size_t, int); ++extern ssize_t ext3_getxattr(struct dentry *, const char *, void *, size_t); ++extern ssize_t ext3_listxattr(struct dentry *, char *, size_t); ++extern int ext3_removexattr(struct dentry *, const char *); ++ ++extern int ext3_xattr_get(struct inode *, int, const char *, void *, size_t); ++extern int ext3_xattr_list(struct inode *, char *, size_t); ++extern int ext3_xattr_set(handle_t *handle, struct inode *, int, const char *, void *, size_t, int); ++ ++extern void ext3_xattr_drop_inode(handle_t *, struct inode *); ++extern void ext3_xattr_put_super(struct super_block *); ++ ++extern int init_ext3_xattr(void) __init; ++extern void exit_ext3_xattr(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR */ ++# define ext3_setxattr NULL ++# define ext3_getxattr NULL ++# define ext3_listxattr NULL ++# define ext3_removexattr NULL ++ ++static inline int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_list(struct inode *inode, void *buffer, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, void *value, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline void ++ext3_xattr_drop_inode(handle_t *handle, struct inode *inode) ++{ ++} ++ ++static inline void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++} ++ ++static inline int ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr(void) ++{ ++} ++ ++# endif /* CONFIG_EXT3_FS_XATTR */ ++ ++# ifdef CONFIG_EXT3_FS_XATTR_USER ++ ++extern int init_ext3_xattr_user(void) __init; ++extern void exit_ext3_xattr_user(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++static inline int ++init_ext3_xattr_user(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr_user(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++#endif /* __KERNEL__ */ ++ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.18-18-braam/include/linux/xattr.h Sat Apr 5 02:54:56 2003 +@@ -0,0 +1,15 @@ ++/* ++ File: linux/xattr.h ++ ++ Extended attributes handling. ++ ++ Copyright (C) 2001 by Andreas Gruenbacher ++ Copyright (C) 2001 SGI - Silicon Graphics, Inc ++*/ ++#ifndef _LINUX_XATTR_H ++#define _LINUX_XATTR_H ++ ++#define XATTR_CREATE 1 /* set value, fail if attr already exists */ ++#define XATTR_REPLACE 2 /* set value, fail if attr does not exist */ ++ ++#endif /* _LINUX_XATTR_H */ + +_ diff --git a/lustre/kernel_patches/patches/linux-2.4.20-xattr-0.8.54.patch b/lustre/kernel_patches/patches/linux-2.4.20-xattr-0.8.54.patch new file mode 100644 index 0000000..cccc52a --- /dev/null +++ b/lustre/kernel_patches/patches/linux-2.4.20-xattr-0.8.54.patch @@ -0,0 +1,5578 @@ + Documentation/Configure.help | 66 ++ + arch/alpha/defconfig | 7 + arch/alpha/kernel/entry.S | 12 + arch/arm/defconfig | 7 + arch/arm/kernel/calls.S | 24 + arch/i386/defconfig | 7 + arch/ia64/defconfig | 7 + arch/ia64/kernel/entry.S | 24 + arch/m68k/defconfig | 7 + arch/mips/defconfig | 7 + arch/mips64/defconfig | 7 + arch/ppc/defconfig | 14 + arch/ppc64/kernel/misc.S | 2 + arch/s390/defconfig | 7 + arch/s390/kernel/entry.S | 24 + arch/s390x/defconfig | 7 + arch/s390x/kernel/entry.S | 24 + arch/s390x/kernel/wrapper32.S | 92 +++ + arch/sparc/defconfig | 7 + arch/sparc/kernel/systbls.S | 10 + arch/sparc64/defconfig | 7 + arch/sparc64/kernel/systbls.S | 20 + fs/Config.in | 14 + fs/Makefile | 3 + fs/ext2/Makefile | 4 + fs/ext2/file.c | 5 + fs/ext2/ialloc.c | 2 + fs/ext2/inode.c | 34 - + fs/ext2/namei.c | 14 + fs/ext2/super.c | 29 + fs/ext2/symlink.c | 14 + fs/ext2/xattr.c | 1212 +++++++++++++++++++++++++++++++++++++++++ + fs/ext2/xattr_user.c | 103 +++ + fs/ext3/Makefile | 10 + fs/ext3/file.c | 5 + fs/ext3/ialloc.c | 2 + fs/ext3/inode.c | 35 - + fs/ext3/namei.c | 21 + fs/ext3/super.c | 33 + + fs/ext3/symlink.c | 14 + fs/ext3/xattr.c | 1224 ++++++++++++++++++++++++++++++++++++++++++ + fs/ext3/xattr_user.c | 111 +++ + fs/jfs/jfs_xattr.h | 6 + fs/jfs/xattr.c | 6 + fs/mbcache.c | 648 ++++++++++++++++++++++ + include/asm-arm/unistd.h | 2 + include/asm-ia64/unistd.h | 13 + include/asm-ppc64/unistd.h | 2 + include/asm-s390/unistd.h | 15 + include/asm-s390x/unistd.h | 15 + include/asm-sparc/unistd.h | 26 + include/asm-sparc64/unistd.h | 24 + include/linux/cache_def.h | 15 + include/linux/errno.h | 4 + include/linux/ext2_fs.h | 31 - + include/linux/ext2_xattr.h | 157 +++++ + include/linux/ext3_fs.h | 31 - + include/linux/ext3_jbd.h | 8 + include/linux/ext3_xattr.h | 157 +++++ + include/linux/fs.h | 2 + include/linux/mbcache.h | 69 ++ + kernel/ksyms.c | 4 + mm/vmscan.c | 35 + + 63 files changed, 4355 insertions(+), 193 deletions(-) + +--- linux-2.4.20/Documentation/Configure.help~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/Documentation/Configure.help Sat Apr 5 03:57:18 2003 +@@ -15242,6 +15242,39 @@ CONFIG_EXT2_FS + be compiled as a module, and so this could be dangerous. Most + everyone wants to say Y here. + ++Ext2 extended attributes ++CONFIG_EXT2_FS_XATTR ++ Extended attributes are name:value pairs associated with inodes by ++ the kernel or by users (see the attr(5) manual page, or visit ++ for details). ++ ++ If unsure, say N. ++ ++Ext2 extended attribute block sharing ++CONFIG_EXT2_FS_XATTR_SHARING ++ This options enables code for sharing identical extended attribute ++ blocks among multiple inodes. ++ ++ Usually, say Y. ++ ++Ext2 extended user attributes ++CONFIG_EXT2_FS_XATTR_USER ++ This option enables extended user attributes on ext2. Processes can ++ associate extended user attributes with inodes to store additional ++ information such as the character encoding of files, etc. (see the ++ attr(5) manual page, or visit for details). ++ ++ If unsure, say N. ++ ++Ext2 trusted extended attributes ++CONFIG_EXT2_FS_XATTR_TRUSTED ++ This option enables extended attributes on ext2 that are accessible ++ (and visible) only to users capable of CAP_SYS_ADMIN. Usually this ++ is only the super user. Trusted extended attributes are meant for ++ implementing system/security services. ++ ++ If unsure, say N. ++ + Ext3 journalling file system support (EXPERIMENTAL) + CONFIG_EXT3_FS + This is the journalling version of the Second extended file system +@@ -15274,6 +15307,39 @@ CONFIG_EXT3_FS + of your root partition (the one containing the directory /) cannot + be compiled as a module, and so this may be dangerous. + ++Ext3 extended attributes ++CONFIG_EXT3_FS_XATTR ++ Extended attributes are name:value pairs associated with inodes by ++ the kernel or by users (see the attr(5) manual page, or visit ++ for details). ++ ++ If unsure, say N. ++ ++Ext3 extended attribute block sharing ++CONFIG_EXT3_FS_XATTR_SHARING ++ This options enables code for sharing identical extended attribute ++ blocks among multiple inodes. ++ ++ Usually, say Y. ++ ++Ext3 extended user attributes ++CONFIG_EXT3_FS_XATTR_USER ++ This option enables extended user attributes on ext3. Processes can ++ associate extended user attributes with inodes to store additional ++ information such as the character encoding of files, etc. (see the ++ attr(5) manual page, or visit for details). ++ ++ If unsure, say N. ++ ++Ext3 trusted extended attributes ++CONFIG_EXT3_FS_XATTR_TRUSTED ++ This option enables extended attributes on ext3 that are accessible ++ (and visible) only to users capable of CAP_SYS_ADMIN. Usually this ++ is only the super user. Trusted extended attributes are meant for ++ implementing system/security services. ++ ++ If unsure, say N. ++ + Journal Block Device support (JBD for ext3) (EXPERIMENTAL) + CONFIG_JBD + This is a generic journalling layer for block devices. It is +--- linux-2.4.20/arch/alpha/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/alpha/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_ALPHA=y + # CONFIG_UID16 is not set + # CONFIG_RWSEM_GENERIC_SPINLOCK is not set +--- linux-2.4.20/arch/alpha/kernel/entry.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/alpha/kernel/entry.S Sat Apr 5 03:57:18 2003 +@@ -1154,6 +1154,18 @@ sys_call_table: + .quad sys_readahead + .quad sys_ni_syscall /* 380, sys_security */ + .quad sys_tkill ++ .quad sys_setxattr ++ .quad sys_lsetxattr ++ .quad sys_fsetxattr ++ .quad sys_getxattr /* 385 */ ++ .quad sys_lgetxattr ++ .quad sys_fgetxattr ++ .quad sys_listxattr ++ .quad sys_llistxattr ++ .quad sys_flistxattr /* 390 */ ++ .quad sys_removexattr ++ .quad sys_lremovexattr ++ .quad sys_fremovexattr + + /* Remember to update everything, kids. */ + .ifne (. - sys_call_table) - (NR_SYSCALLS * 8) +--- linux-2.4.20/arch/arm/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/arm/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_ARM=y + # CONFIG_EISA is not set + # CONFIG_SBUS is not set +--- linux-2.4.20/arch/arm/kernel/calls.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/arm/kernel/calls.S Sat Apr 5 03:57:18 2003 +@@ -240,18 +240,18 @@ __syscall_start: + .long SYMBOL_NAME(sys_ni_syscall) /* Security */ + .long SYMBOL_NAME(sys_gettid) + /* 225 */ .long SYMBOL_NAME(sys_readahead) +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_setxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_lsetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_fsetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_getxattr */ +-/* 230 */ .long SYMBOL_NAME(sys_ni_syscall) /* sys_lgetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_fgetxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_listxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_llistxattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_flistxattr */ +-/* 235 */ .long SYMBOL_NAME(sys_ni_syscall) /* sys_removexattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_lremovexattr */ +- .long SYMBOL_NAME(sys_ni_syscall) /* sys_fremovexattr */ ++ .long SYMBOL_NAME(sys_setxattr) ++ .long SYMBOL_NAME(sys_lsetxattr) ++ .long SYMBOL_NAME(sys_fsetxattr) ++ .long SYMBOL_NAME(sys_getxattr) ++/* 230 */ .long SYMBOL_NAME(sys_lgetxattr) ++ .long SYMBOL_NAME(sys_fgetxattr) ++ .long SYMBOL_NAME(sys_listxattr) ++ .long SYMBOL_NAME(sys_llistxattr) ++ .long SYMBOL_NAME(sys_flistxattr) ++/* 235 */ .long SYMBOL_NAME(sys_removexattr) ++ .long SYMBOL_NAME(sys_lremovexattr) ++ .long SYMBOL_NAME(sys_fremovexattr) + .long SYMBOL_NAME(sys_tkill) + /* + * Please check 2.5 _before_ adding calls here, +--- linux-2.4.20/arch/i386/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/i386/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_X86=y + CONFIG_ISA=y + # CONFIG_SBUS is not set +--- linux-2.4.20/arch/ia64/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/ia64/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + + # + # Code maturity level options +--- linux-2.4.20/arch/ia64/kernel/entry.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/ia64/kernel/entry.S Sat Apr 5 03:57:18 2003 +@@ -1170,18 +1170,18 @@ sys_call_table: + data8 sys_getdents64 + data8 sys_getunwind // 1215 + data8 sys_readahead +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall // 1220 +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall // 1225 +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall +- data8 ia64_ni_syscall ++ data8 sys_setxattr ++ data8 sys_lsetxattr ++ data8 sys_fsetxattr ++ data8 sys_getxattr // 1220 ++ data8 sys_lgetxattr ++ data8 sys_fgetxattr ++ data8 sys_listxattr ++ data8 sys_llistxattr ++ data8 sys_flistxattr // 1225 ++ data8 sys_removexattr ++ data8 sys_lremovexattr ++ data8 sys_fremovexattr + data8 sys_tkill + data8 ia64_ni_syscall // 1230 + data8 ia64_ni_syscall +--- linux-2.4.20/arch/m68k/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/m68k/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_UID16=y + + # +--- linux-2.4.20/arch/mips/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/mips/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_MIPS=y + CONFIG_MIPS32=y + # CONFIG_MIPS64 is not set +--- linux-2.4.20/arch/mips64/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/mips64/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_MIPS=y + # CONFIG_MIPS32 is not set + CONFIG_MIPS64=y +--- linux-2.4.20/arch/ppc/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/ppc/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,20 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_UID16 is not set + # CONFIG_RWSEM_GENERIC_SPINLOCK is not set + CONFIG_RWSEM_XCHGADD_ALGORITHM=y +--- linux-2.4.20/arch/ppc64/kernel/misc.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/ppc64/kernel/misc.S Sat Apr 5 03:57:18 2003 +@@ -731,6 +731,7 @@ _GLOBAL(sys_call_table32) + .llong .sys_gettid /* 207 */ + #if 0 /* Reserved syscalls */ + .llong .sys_tkill /* 208 */ ++#endif + .llong .sys_setxattr + .llong .sys_lsetxattr /* 210 */ + .llong .sys_fsetxattr +@@ -743,6 +744,7 @@ _GLOBAL(sys_call_table32) + .llong .sys_removexattr + .llong .sys_lremovexattr + .llong .sys_fremovexattr /* 220 */ ++#if 0 /* Reserved syscalls */ + .llong .sys_futex + #endif + .llong .sys_perfmonctl /* Put this here for now ... */ +--- linux-2.4.20/arch/s390/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/s390/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_ISA is not set + # CONFIG_EISA is not set + # CONFIG_MCA is not set +--- linux-2.4.20/arch/s390/kernel/entry.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/s390/kernel/entry.S Sat Apr 5 03:57:18 2003 +@@ -558,18 +558,18 @@ sys_call_table: + .long sys_fcntl64 + .long sys_ni_syscall + .long sys_ni_syscall +- .long sys_ni_syscall /* 224 - reserved for setxattr */ +- .long sys_ni_syscall /* 225 - reserved for lsetxattr */ +- .long sys_ni_syscall /* 226 - reserved for fsetxattr */ +- .long sys_ni_syscall /* 227 - reserved for getxattr */ +- .long sys_ni_syscall /* 228 - reserved for lgetxattr */ +- .long sys_ni_syscall /* 229 - reserved for fgetxattr */ +- .long sys_ni_syscall /* 230 - reserved for listxattr */ +- .long sys_ni_syscall /* 231 - reserved for llistxattr */ +- .long sys_ni_syscall /* 232 - reserved for flistxattr */ +- .long sys_ni_syscall /* 233 - reserved for removexattr */ +- .long sys_ni_syscall /* 234 - reserved for lremovexattr */ +- .long sys_ni_syscall /* 235 - reserved for fremovexattr */ ++ .long sys_setxattr ++ .long sys_lsetxattr /* 225 */ ++ .long sys_fsetxattr ++ .long sys_getxattr ++ .long sys_lgetxattr ++ .long sys_fgetxattr ++ .long sys_listxattr /* 230 */ ++ .long sys_llistxattr ++ .long sys_flistxattr ++ .long sys_removexattr ++ .long sys_lremovexattr ++ .long sys_fremovexattr /* 235 */ + .long sys_gettid + .long sys_tkill + .rept 255-237 +--- linux-2.4.20/arch/s390x/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/s390x/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + # CONFIG_ISA is not set + # CONFIG_EISA is not set + # CONFIG_MCA is not set +--- linux-2.4.20/arch/s390x/kernel/entry.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/s390x/kernel/entry.S Sat Apr 5 03:57:18 2003 +@@ -591,18 +591,18 @@ sys_call_table: + .long SYSCALL(sys_ni_syscall,sys32_fcntl64_wrapper) + .long SYSCALL(sys_ni_syscall,sys_ni_syscall) + .long SYSCALL(sys_ni_syscall,sys_ni_syscall) +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 224 - reserved for setxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 225 - reserved for lsetxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 226 - reserved for fsetxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 227 - reserved for getxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 228 - reserved for lgetxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 229 - reserved for fgetxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 230 - reserved for listxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 231 - reserved for llistxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 232 - reserved for flistxattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 233 - reserved for removexattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 234 - reserved for lremovexattr */ +- .long SYSCALL(sys_ni_syscall,sys_ni_syscall) /* 235 - reserved for fremovexattr */ ++ .long SYSCALL(sys_setxattr,sys32_setxattr_wrapper) ++ .long SYSCALL(sys_lsetxattr,sys32_lsetxattr_wrapper) /* 225 */ ++ .long SYSCALL(sys_fsetxattr,sys32_fsetxattr_wrapper) ++ .long SYSCALL(sys_getxattr,sys32_getxattr_wrapper) ++ .long SYSCALL(sys_lgetxattr,sys32_lgetxattr_wrapper) ++ .long SYSCALL(sys_fgetxattr,sys32_fgetxattr_wrapper) ++ .long SYSCALL(sys_listxattr,sys32_listxattr_wrapper) /* 230 */ ++ .long SYSCALL(sys_llistxattr,sys32_llistxattr_wrapper) ++ .long SYSCALL(sys_flistxattr,sys32_flistxattr_wrapper) ++ .long SYSCALL(sys_removexattr,sys32_removexattr_wrapper) ++ .long SYSCALL(sys_lremovexattr,sys32_lremovexattr_wrapper) ++ .long SYSCALL(sys_fremovexattr,sys32_fremovexattr_wrapper)/* 235 */ + .long SYSCALL(sys_gettid,sys_gettid) + .long SYSCALL(sys_tkill,sys_tkill) + .rept 255-237 +--- linux-2.4.20/arch/s390x/kernel/wrapper32.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/s390x/kernel/wrapper32.S Sat Apr 5 03:57:18 2003 +@@ -1091,3 +1091,95 @@ sys32_fstat64_wrapper: + llgtr %r3,%r3 # struct stat64 * + llgfr %r4,%r4 # long + jg sys32_fstat64 # branch to system call ++ ++ .globl sys32_setxattr_wrapper ++sys32_setxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_setxattr ++ ++ .globl sys32_lsetxattr_wrapper ++sys32_lsetxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_lsetxattr ++ ++ .globl sys32_fsetxattr_wrapper ++sys32_fsetxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ lgfr %r6,%r6 # int ++ jg sys_fsetxattr ++ ++ .globl sys32_getxattr_wrapper ++sys32_getxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_getxattr ++ ++ .globl sys32_lgetxattr_wrapper ++sys32_lgetxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_lgetxattr ++ ++ .globl sys32_fgetxattr_wrapper ++sys32_fgetxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgtr %r4,%r4 # void * ++ llgfr %r5,%r5 # size_t ++ jg sys_fgetxattr ++ ++ .globl sys32_listxattr_wrapper ++sys32_listxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_listxattr ++ ++ .globl sys32_llistxattr_wrapper ++sys32_llistxattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_llistxattr ++ ++ .globl sys32_flistxattr_wrapper ++sys32_flistxattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ llgfr %r4,%r4 # size_t ++ jg sys_flistxattr ++ ++ .globl sys32_removexattr_wrapper ++sys32_removexattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ jg sys_removexattr ++ ++ .globl sys32_lremovexattr_wrapper ++sys32_lremovexattr_wrapper: ++ llgtr %r2,%r2 # char * ++ llgtr %r3,%r3 # char * ++ jg sys_lremovexattr ++ ++ .globl sys32_fremovexattr_wrapper ++sys32_fremovexattr_wrapper: ++ lgfr %r2,%r2 # int ++ llgtr %r3,%r3 # char * ++ jg sys_fremovexattr ++ ++ +--- linux-2.4.20/arch/sparc/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/sparc/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + CONFIG_UID16=y + CONFIG_HIGHMEM=y + +--- linux-2.4.20/arch/sparc/kernel/systbls.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/sparc/kernel/systbls.S Sat Apr 5 03:57:18 2003 +@@ -51,11 +51,11 @@ sys_call_table: + /*150*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + /*155*/ .long sys_fcntl64, sys_nis_syscall, sys_statfs, sys_fstatfs, sys_oldumount + /*160*/ .long sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_nis_syscall +-/*165*/ .long sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_nis_syscall +-/*170*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_getdents +-/*175*/ .long sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .long sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_sigpending, sys_query_module +-/*185*/ .long sys_setpgid, sys_nis_syscall, sys_tkill, sys_nis_syscall, sys_newuname ++/*165*/ .long sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_setxattr ++/*170*/ .long sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys_getdents ++/*175*/ .long sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .long sys_flistxattr, sys_removexattr, sys_lremovexattr, sys_sigpending, sys_query_module ++/*185*/ .long sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sys_newuname + /*190*/ .long sys_init_module, sys_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + /*195*/ .long sys_nis_syscall, sys_nis_syscall, sys_getppid, sparc_sigaction, sys_sgetmask + /*200*/ .long sys_ssetmask, sys_sigsuspend, sys_newlstat, sys_uselib, old_readdir +--- linux-2.4.20/arch/sparc64/defconfig~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/sparc64/defconfig Sat Apr 5 03:57:18 2003 +@@ -1,6 +1,13 @@ + # + # Automatically generated make config: don't edit + # ++# CONFIG_EXT3_FS_XATTR is not set ++# CONFIG_EXT3_FS_XATTR_SHARING is not set ++# CONFIG_EXT3_FS_XATTR_USER is not set ++# CONFIG_EXT2_FS_XATTR is not set ++# CONFIG_EXT2_FS_XATTR_SHARING is not set ++# CONFIG_EXT2_FS_XATTR_USER is not set ++# CONFIG_FS_MBCACHE is not set + + # + # Code maturity level options +--- linux-2.4.20/arch/sparc64/kernel/systbls.S~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/arch/sparc64/kernel/systbls.S Sat Apr 5 03:57:18 2003 +@@ -52,11 +52,11 @@ sys_call_table32: + /*150*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + .word sys32_fcntl64, sys_nis_syscall, sys32_statfs, sys32_fstatfs, sys_oldumount + /*160*/ .word sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_nis_syscall +- .word sys32_quotactl, sys_nis_syscall, sys32_mount, sys_ustat, sys_nis_syscall +-/*170*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys32_getdents +- .word sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys32_sigpending, sys32_query_module +- .word sys_setpgid, sys_nis_syscall, sys_tkill, sys_nis_syscall, sparc64_newuname ++ .word sys32_quotactl, sys_nis_syscall, sys32_mount, sys_ustat, sys_setxattr ++/*170*/ .word sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys32_getdents ++ .word sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .word sys_flistxattr, sys_removexattr, sys_lremovexattr, sys32_sigpending, sys32_query_module ++ .word sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sparc64_newuname + /*190*/ .word sys32_init_module, sparc64_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + .word sys_nis_syscall, sys_nis_syscall, sys_getppid, sys32_sigaction, sys_sgetmask + /*200*/ .word sys_ssetmask, sys_sigsuspend, sys32_newlstat, sys_uselib, old32_readdir +@@ -111,11 +111,11 @@ sys_call_table: + /*150*/ .word sys_getsockname, sys_nis_syscall, sys_nis_syscall, sys_poll, sys_getdents64 + .word sys_nis_syscall, sys_nis_syscall, sys_statfs, sys_fstatfs, sys_oldumount + /*160*/ .word sys_nis_syscall, sys_nis_syscall, sys_getdomainname, sys_setdomainname, sys_utrap_install +- .word sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_nis_syscall +-/*170*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_getdents +- .word sys_setsid, sys_fchdir, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall +-/*180*/ .word sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall, sys_query_module +- .word sys_setpgid, sys_nis_syscall, sys_tkill, sys_nis_syscall, sparc64_newuname ++ .word sys_quotactl, sys_nis_syscall, sys_mount, sys_ustat, sys_setxattr ++/*170*/ .word sys_lsetxattr, sys_fsetxattr, sys_getxattr, sys_lgetxattr, sys_getdents ++ .word sys_setsid, sys_fchdir, sys_fgetxattr, sys_listxattr, sys_llistxattr ++/*180*/ .word sys_flistxattr, sys_removexattr, sys_lremovexattr, sys_nis_syscall, sys_query_module ++ .word sys_setpgid, sys_fremovexattr, sys_tkill, sys_nis_syscall, sparc64_newuname + /*190*/ .word sys_init_module, sparc64_personality, sys_nis_syscall, sys_nis_syscall, sys_nis_syscall + .word sys_nis_syscall, sys_nis_syscall, sys_getppid, sys_nis_syscall, sys_sgetmask + /*200*/ .word sys_ssetmask, sys_nis_syscall, sys_newlstat, sys_uselib, sys_nis_syscall +--- linux-2.4.20/fs/Config.in~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/Config.in Sat Apr 5 03:57:18 2003 +@@ -25,6 +25,11 @@ dep_mbool ' Debug Befs' CONFIG_BEFS_DEB + dep_tristate 'BFS file system support (EXPERIMENTAL)' CONFIG_BFS_FS $CONFIG_EXPERIMENTAL + + tristate 'Ext3 journalling file system support' CONFIG_EXT3_FS ++dep_mbool ' Ext3 extended attributes' CONFIG_EXT3_FS_XATTR $CONFIG_EXT3_FS ++dep_bool ' Ext3 extended attribute block sharing' \ ++ CONFIG_EXT3_FS_XATTR_SHARING $CONFIG_EXT3_FS_XATTR ++dep_bool ' Ext3 extended user attributes' \ ++ CONFIG_EXT3_FS_XATTR_USER $CONFIG_EXT3_FS_XATTR + # CONFIG_JBD could be its own option (even modular), but until there are + # other users than ext3, we will simply make it be the same as CONFIG_EXT3_FS + # dep_tristate ' Journal Block Device support (JBD for ext3)' CONFIG_JBD $CONFIG_EXT3_FS +@@ -84,6 +89,11 @@ dep_mbool ' QNX4FS write support (DANGE + tristate 'ROM file system support' CONFIG_ROMFS_FS + + tristate 'Second extended fs support' CONFIG_EXT2_FS ++dep_mbool ' Ext2 extended attributes' CONFIG_EXT2_FS_XATTR $CONFIG_EXT2_FS ++dep_bool ' Ext2 extended attribute block sharing' \ ++ CONFIG_EXT2_FS_XATTR_SHARING $CONFIG_EXT2_FS_XATTR ++dep_bool ' Ext2 extended user attributes' \ ++ CONFIG_EXT2_FS_XATTR_USER $CONFIG_EXT2_FS_XATTR + + tristate 'System V/Xenix/V7/Coherent file system support' CONFIG_SYSV_FS + +@@ -155,6 +165,10 @@ else + define_tristate CONFIG_ZISOFS_FS n + fi + ++# Meta block cache for Extended Attributes (ext2/ext3) ++#tristate 'Meta block cache' CONFIG_FS_MBCACHE ++define_tristate CONFIG_FS_MBCACHE y ++ + mainmenu_option next_comment + comment 'Partition Types' + source fs/partitions/Config.in +--- linux-2.4.20/fs/Makefile~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/Makefile Sat Apr 5 03:57:18 2003 +@@ -79,6 +79,9 @@ obj-y += binfmt_script.o + + obj-$(CONFIG_BINFMT_ELF) += binfmt_elf.o + ++export-objs += mbcache.o ++obj-$(CONFIG_FS_MBCACHE) += mbcache.o ++ + # persistent filesystems + obj-y += $(join $(subdir-y),$(subdir-y:%=/%.o)) + +--- linux-2.4.20/fs/ext2/Makefile~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/Makefile Sat Apr 5 03:57:18 2003 +@@ -13,4 +13,8 @@ obj-y := balloc.o bitmap.o dir.o file + ioctl.o namei.o super.o symlink.o + obj-m := $(O_TARGET) + ++export-objs += xattr.o ++obj-$(CONFIG_EXT2_FS_XATTR) += xattr.o ++obj-$(CONFIG_EXT2_FS_XATTR_USER) += xattr_user.o ++ + include $(TOPDIR)/Rules.make +--- linux-2.4.20/fs/ext2/file.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/file.c Sat Apr 5 03:57:18 2003 +@@ -20,6 +20,7 @@ + + #include + #include ++#include + #include + + /* +@@ -51,4 +52,8 @@ struct file_operations ext2_file_operati + + struct inode_operations ext2_file_inode_operations = { + truncate: ext2_truncate, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +--- linux-2.4.20/fs/ext2/ialloc.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/ialloc.c Sat Apr 5 03:57:18 2003 +@@ -15,6 +15,7 @@ + #include + #include + #include ++#include + #include + #include + +@@ -167,6 +168,7 @@ void ext2_free_inode (struct inode * ino + */ + if (!is_bad_inode(inode)) { + /* Quota is already initialized in iput() */ ++ ext2_xattr_delete_inode(inode); + DQUOT_FREE_INODE(inode); + DQUOT_DROP(inode); + } +--- linux-2.4.20/fs/ext2/inode.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/inode.c Sat Apr 5 03:57:18 2003 +@@ -39,6 +39,18 @@ MODULE_LICENSE("GPL"); + static int ext2_update_inode(struct inode * inode, int do_sync); + + /* ++ * Test whether an inode is a fast symlink. ++ */ ++static inline int ext2_inode_is_fast_symlink(struct inode *inode) ++{ ++ int ea_blocks = inode->u.ext2_i.i_file_acl ? ++ (inode->i_sb->s_blocksize >> 9) : 0; ++ ++ return (S_ISLNK(inode->i_mode) && ++ inode->i_blocks - ea_blocks == 0); ++} ++ ++/* + * Called at each iput() + */ + void ext2_put_inode (struct inode * inode) +@@ -53,9 +65,7 @@ void ext2_delete_inode (struct inode * i + { + lock_kernel(); + +- if (is_bad_inode(inode) || +- inode->i_ino == EXT2_ACL_IDX_INO || +- inode->i_ino == EXT2_ACL_DATA_INO) ++ if (is_bad_inode(inode)) + goto no_delete; + inode->u.ext2_i.i_dtime = CURRENT_TIME; + mark_inode_dirty(inode); +@@ -801,6 +811,8 @@ void ext2_truncate (struct inode * inode + if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode))) + return; ++ if (ext2_inode_is_fast_symlink(inode)) ++ return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + +@@ -888,8 +900,7 @@ void ext2_read_inode (struct inode * ino + unsigned long offset; + struct ext2_group_desc * gdp; + +- if ((inode->i_ino != EXT2_ROOT_INO && inode->i_ino != EXT2_ACL_IDX_INO && +- inode->i_ino != EXT2_ACL_DATA_INO && ++ if ((inode->i_ino != EXT2_ROOT_INO && + inode->i_ino < EXT2_FIRST_INO(inode->i_sb)) || + inode->i_ino > le32_to_cpu(inode->i_sb->u.ext2_sb.s_es->s_inodes_count)) { + ext2_error (inode->i_sb, "ext2_read_inode", +@@ -974,10 +985,7 @@ void ext2_read_inode (struct inode * ino + for (block = 0; block < EXT2_N_BLOCKS; block++) + inode->u.ext2_i.i_data[block] = raw_inode->i_block[block]; + +- if (inode->i_ino == EXT2_ACL_IDX_INO || +- inode->i_ino == EXT2_ACL_DATA_INO) +- /* Nothing to do */ ; +- else if (S_ISREG(inode->i_mode)) { ++ if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext2_file_inode_operations; + inode->i_fop = &ext2_file_operations; + inode->i_mapping->a_ops = &ext2_aops; +@@ -986,15 +994,17 @@ void ext2_read_inode (struct inode * ino + inode->i_fop = &ext2_dir_operations; + inode->i_mapping->a_ops = &ext2_aops; + } else if (S_ISLNK(inode->i_mode)) { +- if (!inode->i_blocks) ++ if (ext2_inode_is_fast_symlink(inode)) + inode->i_op = &ext2_fast_symlink_inode_operations; + else { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext2_symlink_inode_operations; + inode->i_mapping->a_ops = &ext2_aops; + } +- } else ++ } else { ++ inode->i_op = &ext2_special_inode_operations; + init_special_inode(inode, inode->i_mode, + le32_to_cpu(raw_inode->i_block[0])); ++ } + brelse (bh); + inode->i_attr_flags = 0; + if (inode->u.ext2_i.i_flags & EXT2_SYNC_FL) { +--- linux-2.4.20/fs/ext2/namei.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/namei.c Sat Apr 5 03:57:18 2003 +@@ -31,6 +31,7 @@ + + #include + #include ++#include + #include + + /* +@@ -136,7 +137,7 @@ static int ext2_symlink (struct inode * + + if (l > sizeof (inode->u.ext2_i.i_data)) { + /* slow symlink */ +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext2_symlink_inode_operations; + inode->i_mapping->a_ops = &ext2_aops; + err = block_symlink(inode, symname, l); + if (err) +@@ -345,4 +346,15 @@ struct inode_operations ext2_dir_inode_o + rmdir: ext2_rmdir, + mknod: ext2_mknod, + rename: ext2_rename, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, ++}; ++ ++struct inode_operations ext2_special_inode_operations = { ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +--- linux-2.4.20/fs/ext2/super.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/super.c Sat Apr 5 03:57:18 2003 +@@ -21,6 +21,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -125,6 +126,7 @@ void ext2_put_super (struct super_block + int db_count; + int i; + ++ ext2_xattr_put_super(sb); + if (!(sb->s_flags & MS_RDONLY)) { + struct ext2_super_block *es = EXT2_SB(sb)->s_es; + +@@ -175,6 +177,13 @@ static int parse_options (char * options + this_char = strtok (NULL, ",")) { + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; ++#ifdef CONFIG_EXT2_FS_XATTR_USER ++ if (!strcmp (this_char, "user_xattr")) ++ set_opt (*mount_options, XATTR_USER); ++ else if (!strcmp (this_char, "nouser_xattr")) ++ clear_opt (*mount_options, XATTR_USER); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -424,6 +433,9 @@ struct super_block * ext2_read_super (st + blocksize = BLOCK_SIZE; + + sb->u.ext2_sb.s_mount_opt = 0; ++#ifdef CONFIG_EXT2_FS_XATTR_USER ++ /* set_opt (sb->u.ext2_sb.s_mount_opt, XATTR_USER); */ ++#endif + if (!parse_options ((char *) data, &sb_block, &resuid, &resgid, + &sb->u.ext2_sb.s_mount_opt)) { + return NULL; +@@ -813,12 +825,27 @@ static DECLARE_FSTYPE_DEV(ext2_fs_type, + + static int __init init_ext2_fs(void) + { +- return register_filesystem(&ext2_fs_type); ++ int error = init_ext2_xattr(); ++ if (error) ++ return error; ++ error = init_ext2_xattr_user(); ++ if (error) ++ goto fail; ++ error = register_filesystem(&ext2_fs_type); ++ if (!error) ++ return 0; ++ ++ exit_ext2_xattr_user(); ++fail: ++ exit_ext2_xattr(); ++ return error; + } + + static void __exit exit_ext2_fs(void) + { + unregister_filesystem(&ext2_fs_type); ++ exit_ext2_xattr_user(); ++ exit_ext2_xattr(); + } + + EXPORT_NO_SYMBOLS; +--- linux-2.4.20/fs/ext2/symlink.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext2/symlink.c Sat Apr 5 03:57:18 2003 +@@ -19,6 +19,7 @@ + + #include + #include ++#include + + static int ext2_readlink(struct dentry *dentry, char *buffer, int buflen) + { +@@ -32,7 +33,20 @@ static int ext2_follow_link(struct dentr + return vfs_follow_link(nd, s); + } + ++struct inode_operations ext2_symlink_inode_operations = { ++ readlink: page_readlink, ++ follow_link: page_follow_link, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, ++}; ++ + struct inode_operations ext2_fast_symlink_inode_operations = { + readlink: ext2_readlink, + follow_link: ext2_follow_link, ++ setxattr: ext2_setxattr, ++ getxattr: ext2_getxattr, ++ listxattr: ext2_listxattr, ++ removexattr: ext2_removexattr, + }; +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext2/xattr.c Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,1212 @@ ++/* ++ * linux/fs/ext2/xattr.c ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ * ++ * Fix by Harrison Xing . ++ * Extended attributes for symlinks and special files added per ++ * suggestion of Luka Renko . ++ */ ++ ++/* ++ * Extended attributes are stored on disk blocks allocated outside of ++ * any inode. The i_file_acl field is then made to point to this allocated ++ * block. If all extended attributes of an inode are identical, these ++ * inodes may share the same extended attribute block. Such situations ++ * are automatically detected by keeping a cache of recent attribute block ++ * numbers and hashes over the block's contents in memory. ++ * ++ * ++ * Extended attribute block layout: ++ * ++ * +------------------+ ++ * | header | ++ * | entry 1 | | ++ * | entry 2 | | growing downwards ++ * | entry 3 | v ++ * | four null bytes | ++ * | . . . | ++ * | value 1 | ^ ++ * | value 3 | | growing upwards ++ * | value 2 | | ++ * +------------------+ ++ * ++ * The block header is followed by multiple entry descriptors. These entry ++ * descriptors are variable in size, and alligned to EXT2_XATTR_PAD ++ * byte boundaries. The entry descriptors are sorted by attribute name, ++ * so that two extended attribute blocks can be compared efficiently. ++ * ++ * Attribute values are aligned to the end of the block, stored in ++ * no specific order. They are also padded to EXT2_XATTR_PAD byte ++ * boundaries. No additional gaps are left between them. ++ * ++ * Locking strategy ++ * ---------------- ++ * The VFS already holds the BKL and the inode->i_sem semaphore when any of ++ * the xattr inode operations are called, so we are guaranteed that only one ++ * processes accesses extended attributes of an inode at any time. ++ * ++ * For writing we also grab the ext2_xattr_sem semaphore. This ensures that ++ * only a single process is modifying an extended attribute block, even ++ * if the block is shared among inodes. ++ * ++ * Note for porting to 2.5 ++ * ----------------------- ++ * The BKL will no longer be held in the xattr inode operations. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++/* These symbols may be needed by a module. */ ++EXPORT_SYMBOL(ext2_xattr_register); ++EXPORT_SYMBOL(ext2_xattr_unregister); ++EXPORT_SYMBOL(ext2_xattr_get); ++EXPORT_SYMBOL(ext2_xattr_list); ++EXPORT_SYMBOL(ext2_xattr_set); ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++# define mark_buffer_dirty(bh) mark_buffer_dirty(bh, 1) ++#endif ++ ++#define HDR(bh) ((struct ext2_xattr_header *)((bh)->b_data)) ++#define ENTRY(ptr) ((struct ext2_xattr_entry *)(ptr)) ++#define FIRST_ENTRY(bh) ENTRY(HDR(bh)+1) ++#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) ++ ++#ifdef EXT2_XATTR_DEBUG ++# define ea_idebug(inode, f...) do { \ ++ printk(KERN_DEBUG "inode %s:%ld: ", \ ++ kdevname(inode->i_dev), inode->i_ino); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++# define ea_bdebug(bh, f...) do { \ ++ printk(KERN_DEBUG "block %s:%ld: ", \ ++ kdevname(bh->b_dev), bh->b_blocknr); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++#else ++# define ea_idebug(f...) ++# define ea_bdebug(f...) ++#endif ++ ++static int ext2_xattr_set2(struct inode *, struct buffer_head *, ++ struct ext2_xattr_header *); ++ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ ++static int ext2_xattr_cache_insert(struct buffer_head *); ++static struct buffer_head *ext2_xattr_cache_find(struct inode *, ++ struct ext2_xattr_header *); ++static void ext2_xattr_cache_remove(struct buffer_head *); ++static void ext2_xattr_rehash(struct ext2_xattr_header *, ++ struct ext2_xattr_entry *); ++ ++static struct mb_cache *ext2_xattr_cache; ++ ++#else ++# define ext2_xattr_cache_insert(bh) 0 ++# define ext2_xattr_cache_find(inode, header) NULL ++# define ext2_xattr_cache_remove(bh) while(0) {} ++# define ext2_xattr_rehash(header, entry) while(0) {} ++#endif ++ ++/* ++ * If a file system does not share extended attributes among inodes, ++ * we should not need the ext2_xattr_sem semaphore. However, the ++ * filesystem may still contain shared blocks, so we always take ++ * the lock. ++ */ ++ ++DECLARE_MUTEX(ext2_xattr_sem); ++ ++static inline int ++ext2_xattr_new_block(struct inode *inode, int * errp, int force) ++{ ++ struct super_block *sb = inode->i_sb; ++ int goal = le32_to_cpu(EXT2_SB(sb)->s_es->s_first_data_block) + ++ EXT2_I(inode)->i_block_group * EXT2_BLOCKS_PER_GROUP(sb); ++ ++ /* How can we enforce the allocation? */ ++ int block = ext2_new_block(inode, goal, 0, 0, errp); ++#ifdef OLD_QUOTAS ++ if (!*errp) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#endif ++ return block; ++} ++ ++static inline int ++ext2_xattr_quota_alloc(struct inode *inode, int force) ++{ ++ /* How can we enforce the allocation? */ ++#ifdef OLD_QUOTAS ++ int error = DQUOT_ALLOC_BLOCK(inode->i_sb, inode, 1); ++ if (!error) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#else ++ int error = DQUOT_ALLOC_BLOCK(inode, 1); ++#endif ++ return error; ++} ++ ++#ifdef OLD_QUOTAS ++ ++static inline void ++ext2_xattr_quota_free(struct inode *inode) ++{ ++ DQUOT_FREE_BLOCK(inode->i_sb, inode, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++static inline void ++ext2_xattr_free_block(struct inode * inode, unsigned long block) ++{ ++ ext2_free_blocks(inode, block, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++#else ++# define ext2_xattr_quota_free(inode) \ ++ DQUOT_FREE_BLOCK(inode, 1) ++# define ext2_xattr_free_block(inode, block) \ ++ ext2_free_blocks(inode, block, 1) ++#endif ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18) ++ ++static inline struct buffer_head * ++sb_bread(struct super_block *sb, int block) ++{ ++ return bread(sb->s_dev, block, sb->s_blocksize); ++} ++ ++static inline struct buffer_head * ++sb_getblk(struct super_block *sb, int block) ++{ ++ return getblk(sb->s_dev, block, sb->s_blocksize); ++} ++ ++#endif ++ ++struct ext2_xattr_handler *ext2_xattr_handlers[EXT2_XATTR_INDEX_MAX]; ++rwlock_t ext2_handler_lock = RW_LOCK_UNLOCKED; ++ ++int ++ext2_xattr_register(int name_index, struct ext2_xattr_handler *handler) ++{ ++ int error = -EINVAL; ++ ++ if (name_index > 0 && name_index <= EXT2_XATTR_INDEX_MAX) { ++ write_lock(&ext2_handler_lock); ++ if (!ext2_xattr_handlers[name_index-1]) { ++ ext2_xattr_handlers[name_index-1] = handler; ++ error = 0; ++ } ++ write_unlock(&ext2_handler_lock); ++ } ++ return error; ++} ++ ++void ++ext2_xattr_unregister(int name_index, struct ext2_xattr_handler *handler) ++{ ++ if (name_index > 0 || name_index <= EXT2_XATTR_INDEX_MAX) { ++ write_lock(&ext2_handler_lock); ++ ext2_xattr_handlers[name_index-1] = NULL; ++ write_unlock(&ext2_handler_lock); ++ } ++} ++ ++static inline const char * ++strcmp_prefix(const char *a, const char *a_prefix) ++{ ++ while (*a_prefix && *a == *a_prefix) { ++ a++; ++ a_prefix++; ++ } ++ return *a_prefix ? NULL : a; ++} ++ ++/* ++ * Decode the extended attribute name, and translate it into ++ * the name_index and name suffix. ++ */ ++static struct ext2_xattr_handler * ++ext2_xattr_resolve_name(const char **name) ++{ ++ struct ext2_xattr_handler *handler = NULL; ++ int i; ++ ++ if (!*name) ++ return NULL; ++ read_lock(&ext2_handler_lock); ++ for (i=0; iprefix); ++ if (n) { ++ handler = ext2_xattr_handlers[i]; ++ *name = n; ++ break; ++ } ++ } ++ } ++ read_unlock(&ext2_handler_lock); ++ return handler; ++} ++ ++static inline struct ext2_xattr_handler * ++ext2_xattr_handler(int name_index) ++{ ++ struct ext2_xattr_handler *handler = NULL; ++ if (name_index > 0 && name_index <= EXT2_XATTR_INDEX_MAX) { ++ read_lock(&ext2_handler_lock); ++ handler = ext2_xattr_handlers[name_index-1]; ++ read_unlock(&ext2_handler_lock); ++ } ++ return handler; ++} ++ ++/* ++ * Inode operation getxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext2_getxattr(struct dentry *dentry, const char *name, ++ void *buffer, size_t size) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->get(inode, name, buffer, size); ++} ++ ++/* ++ * Inode operation listxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext2_listxattr(struct dentry *dentry, char *buffer, size_t size) ++{ ++ return ext2_xattr_list(dentry->d_inode, buffer, size); ++} ++ ++/* ++ * Inode operation setxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext2_setxattr(struct dentry *dentry, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ if (size == 0) ++ value = ""; /* empty EA, do not remove */ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, value, size, flags); ++} ++ ++/* ++ * Inode operation removexattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext2_removexattr(struct dentry *dentry, const char *name) ++{ ++ struct ext2_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext2_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, NULL, 0, XATTR_REPLACE); ++} ++ ++/* ++ * ext2_xattr_get() ++ * ++ * Copy an extended attribute into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext2_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_entry *entry; ++ unsigned int block, size; ++ char *end; ++ int name_len, error; ++ ++ ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", ++ name_index, name, buffer, (long)buffer_size); ++ ++ if (name == NULL) ++ return -EINVAL; ++ if (!EXT2_I(inode)->i_file_acl) ++ return -ENOATTR; ++ block = EXT2_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(inode->i_sb, "ext2_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ ++ error = -ERANGE; ++ if (name_len > 255) ++ goto cleanup; ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) ++ goto found; ++ entry = next; ++ } ++ /* Check the remaining name entries */ ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ entry = next; ++ } ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ error = -ENOATTR; ++ goto cleanup; ++found: ++ /* check the buffer size */ ++ if (entry->e_value_block != 0) ++ goto bad_block; ++ size = le32_to_cpu(entry->e_value_size); ++ if (size > inode->i_sb->s_blocksize || ++ le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize) ++ goto bad_block; ++ ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (buffer) { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ /* return value of attribute */ ++ memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), ++ size); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext2_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext2_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_entry *entry; ++ unsigned int block, size = 0; ++ char *buf, *end; ++ int error; ++ ++ ea_idebug(inode, "buffer=%p, buffer_size=%ld", ++ buffer, (long)buffer_size); ++ ++ if (!EXT2_I(inode)->i_file_acl) ++ return 0; ++ block = EXT2_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(inode->i_sb, "ext2_xattr_list", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* compute the size required for the list of attribute names */ ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT2_XATTR_NEXT(entry)) { ++ struct ext2_xattr_handler *handler; ++ struct ext2_xattr_entry *next = ++ EXT2_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ ++ handler = ext2_xattr_handler(entry->e_name_index); ++ if (handler) ++ size += handler->list(NULL, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ ++ if (ext2_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (!buffer) { ++ error = size; ++ goto cleanup; ++ } else { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ /* list the attribute names */ ++ buf = buffer; ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT2_XATTR_NEXT(entry)) { ++ struct ext2_xattr_handler *handler; ++ ++ handler = ext2_xattr_handler(entry->e_name_index); ++ if (handler) ++ buf += handler->list(buf, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * If the EXT2_FEATURE_COMPAT_EXT_ATTR feature of this file system is ++ * not set, set it. ++ */ ++static void ext2_xattr_update_super_block(struct super_block *sb) ++{ ++ if (EXT2_HAS_COMPAT_FEATURE(sb, EXT2_FEATURE_COMPAT_EXT_ATTR)) ++ return; ++ ++ lock_super(sb); ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++ EXT2_SB(sb)->s_feature_compat |= EXT2_FEATURE_COMPAT_EXT_ATTR; ++#endif ++ EXT2_SB(sb)->s_es->s_feature_compat |= ++ cpu_to_le32(EXT2_FEATURE_COMPAT_EXT_ATTR); ++ sb->s_dirt = 1; ++ mark_buffer_dirty(EXT2_SB(sb)->s_sbh); ++ unlock_super(sb); ++} ++ ++/* ++ * ext2_xattr_set() ++ * ++ * Create, replace or remove an extended attribute for this inode. Buffer ++ * is NULL to remove an existing extended attribute, and non-NULL to ++ * either replace an existing extended attribute, or create a new extended ++ * attribute. The flags XATTR_REPLACE and XATTR_CREATE ++ * specify that an extended attribute must exist and must not exist ++ * previous to the call, respectively. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++int ++ext2_xattr_set(struct inode *inode, int name_index, const char *name, ++ const void *value, size_t value_len, int flags) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *bh = NULL; ++ struct ext2_xattr_header *header = NULL; ++ struct ext2_xattr_entry *here, *last; ++ unsigned int name_len; ++ int block = EXT2_I(inode)->i_file_acl; ++ int min_offs = sb->s_blocksize, not_found = 1, free, error; ++ char *end; ++ ++ /* ++ * header -- Points either into bh, or to a temporarily ++ * allocated buffer. ++ * here -- The named entry found, or the place for inserting, within ++ * the block pointed to by header. ++ * last -- Points right after the last named entry within the block ++ * pointed to by header. ++ * min_offs -- The offset of the first value (values are aligned ++ * towards the end of the block). ++ * end -- Points right after the block pointed to by header. ++ */ ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > sb->s_blocksize) ++ return -ERANGE; ++ down(&ext2_xattr_sem); ++ ++ if (block) { ++ /* The inode already has an extended attribute block. */ ++ ++ bh = sb_bread(sb, block); ++ error = -EIO; ++ if (!bh) ++ goto cleanup; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), ++ le32_to_cpu(HDR(bh)->h_refcount)); ++ header = HDR(bh); ++ end = bh->b_data + bh->b_size; ++ if (header->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ header->h_blocks != cpu_to_le32(1)) { ++bad_block: ext2_error(sb, "ext2_xattr_set", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* Find the named attribute. */ ++ here = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(here)) { ++ struct ext2_xattr_entry *next = EXT2_XATTR_NEXT(here); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!here->e_value_block && here->e_value_size) { ++ int offs = le16_to_cpu(here->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ not_found = name_index - here->e_name_index; ++ if (!not_found) ++ not_found = name_len - here->e_name_len; ++ if (!not_found) ++ not_found = memcmp(name, here->e_name,name_len); ++ if (not_found <= 0) ++ break; ++ here = next; ++ } ++ last = here; ++ /* We still need to compute min_offs and last. */ ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext2_xattr_entry *next = EXT2_XATTR_NEXT(last); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!last->e_value_block && last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ last = next; ++ } ++ ++ /* Check whether we have enough space left. */ ++ free = min_offs - ((char*)last - (char*)header) - sizeof(__u32); ++ } else { ++ /* We will use a new extended attribute block. */ ++ free = sb->s_blocksize - ++ sizeof(struct ext2_xattr_header) - sizeof(__u32); ++ here = last = NULL; /* avoid gcc uninitialized warning. */ ++ } ++ ++ if (not_found) { ++ /* Request to remove a nonexistent attribute? */ ++ error = -ENOATTR; ++ if (flags & XATTR_REPLACE) ++ goto cleanup; ++ error = 0; ++ if (value == NULL) ++ goto cleanup; ++ else ++ free -= EXT2_XATTR_LEN(name_len); ++ } else { ++ /* Request to create an existing attribute? */ ++ error = -EEXIST; ++ if (flags & XATTR_CREATE) ++ goto cleanup; ++ if (!here->e_value_block && here->e_value_size) { ++ unsigned int size = le32_to_cpu(here->e_value_size); ++ ++ if (le16_to_cpu(here->e_value_offs) + size > ++ sb->s_blocksize || size > sb->s_blocksize) ++ goto bad_block; ++ free += EXT2_XATTR_SIZE(size); ++ } ++ } ++ free -= EXT2_XATTR_SIZE(value_len); ++ error = -ENOSPC; ++ if (free < 0) ++ goto cleanup; ++ ++ /* Here we know that we can set the new attribute. */ ++ ++ if (header) { ++ if (header->h_refcount == cpu_to_le32(1)) { ++ ea_bdebug(bh, "modifying in-place"); ++ ext2_xattr_cache_remove(bh); ++ } else { ++ int offset; ++ ++ ea_bdebug(bh, "cloning"); ++ header = kmalloc(bh->b_size, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memcpy(header, HDR(bh), bh->b_size); ++ header->h_refcount = cpu_to_le32(1); ++ offset = (char *)header - bh->b_data; ++ here = ENTRY((char *)here + offset); ++ last = ENTRY((char *)last + offset); ++ } ++ } else { ++ /* Allocate a buffer where we construct the new block. */ ++ header = kmalloc(sb->s_blocksize, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memset(header, 0, sb->s_blocksize); ++ end = (char *)header + sb->s_blocksize; ++ header->h_magic = cpu_to_le32(EXT2_XATTR_MAGIC); ++ header->h_blocks = header->h_refcount = cpu_to_le32(1); ++ last = here = ENTRY(header+1); ++ } ++ ++ if (not_found) { ++ /* Insert the new name. */ ++ int size = EXT2_XATTR_LEN(name_len); ++ int rest = (char *)last - (char *)here; ++ memmove((char *)here + size, here, rest); ++ memset(here, 0, size); ++ here->e_name_index = name_index; ++ here->e_name_len = name_len; ++ memcpy(here->e_name, name, name_len); ++ } else { ++ /* Remove the old value. */ ++ if (!here->e_value_block && here->e_value_size) { ++ char *first_val = (char *)header + min_offs; ++ int offs = le16_to_cpu(here->e_value_offs); ++ char *val = (char *)header + offs; ++ size_t size = EXT2_XATTR_SIZE( ++ le32_to_cpu(here->e_value_size)); ++ memmove(first_val + size, first_val, val - first_val); ++ memset(first_val, 0, size); ++ here->e_value_offs = 0; ++ min_offs += size; ++ ++ /* Adjust all value offsets. */ ++ last = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(last)) { ++ int o = le16_to_cpu(last->e_value_offs); ++ if (!last->e_value_block && o < offs) ++ last->e_value_offs = ++ cpu_to_le16(o + size); ++ last = EXT2_XATTR_NEXT(last); ++ } ++ } ++ if (value == NULL) { ++ /* Remove this attribute. */ ++ if (EXT2_XATTR_NEXT(ENTRY(header+1)) == last) { ++ /* This block is now empty. */ ++ error = ext2_xattr_set2(inode, bh, NULL); ++ goto cleanup; ++ } else { ++ /* Remove the old name. */ ++ int size = EXT2_XATTR_LEN(name_len); ++ last = ENTRY((char *)last - size); ++ memmove(here, (char*)here + size, ++ (char*)last - (char*)here); ++ memset(last, 0, size); ++ } ++ } ++ } ++ ++ if (value != NULL) { ++ /* Insert the new value. */ ++ here->e_value_size = cpu_to_le32(value_len); ++ if (value_len) { ++ size_t size = EXT2_XATTR_SIZE(value_len); ++ char *val = (char *)header + min_offs - size; ++ here->e_value_offs = ++ cpu_to_le16((char *)val - (char *)header); ++ memset(val + size - EXT2_XATTR_PAD, 0, ++ EXT2_XATTR_PAD); /* Clear the pad bytes. */ ++ memcpy(val, value, value_len); ++ } ++ } ++ ext2_xattr_rehash(header, here); ++ ++ error = ext2_xattr_set2(inode, bh, header); ++ ++cleanup: ++ brelse(bh); ++ if (!(bh && header == HDR(bh))) ++ kfree(header); ++ up(&ext2_xattr_sem); ++ ++ return error; ++} ++ ++/* ++ * Second half of ext2_xattr_set(): Update the file system. ++ */ ++static int ++ext2_xattr_set2(struct inode *inode, struct buffer_head *old_bh, ++ struct ext2_xattr_header *header) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *new_bh = NULL; ++ int error; ++ ++ if (header) { ++ new_bh = ext2_xattr_cache_find(inode, header); ++ if (new_bh) { ++ /* ++ * We found an identical block in the cache. ++ * The old block will be released after updating ++ * the inode. ++ */ ++ ea_bdebug(old_bh, "reusing block %ld", ++ new_bh->b_blocknr); ++ ++ error = -EDQUOT; ++ if (ext2_xattr_quota_alloc(inode, 1)) ++ goto cleanup; ++ ++ HDR(new_bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(new_bh)->h_refcount) + 1); ++ ea_bdebug(new_bh, "refcount now=%d", ++ le32_to_cpu(HDR(new_bh)->h_refcount)); ++ } else if (old_bh && header == HDR(old_bh)) { ++ /* Keep this block. */ ++ new_bh = old_bh; ++ ext2_xattr_cache_insert(new_bh); ++ } else { ++ /* We need to allocate a new block */ ++ int force = EXT2_I(inode)->i_file_acl != 0; ++ int block = ext2_xattr_new_block(inode, &error, force); ++ if (error) ++ goto cleanup; ++ ea_idebug(inode, "creating block %d", block); ++ ++ new_bh = sb_getblk(sb, block); ++ if (!new_bh) { ++ ext2_xattr_free_block(inode, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ lock_buffer(new_bh); ++ memcpy(new_bh->b_data, header, new_bh->b_size); ++ mark_buffer_uptodate(new_bh, 1); ++ unlock_buffer(new_bh); ++ ext2_xattr_cache_insert(new_bh); ++ ++ ext2_xattr_update_super_block(sb); ++ } ++ mark_buffer_dirty(new_bh); ++ if (IS_SYNC(inode)) { ++ ll_rw_block(WRITE, 1, &new_bh); ++ wait_on_buffer(new_bh); ++ error = -EIO; ++ if (buffer_req(new_bh) && !buffer_uptodate(new_bh)) ++ goto cleanup; ++ } ++ } ++ ++ /* Update the inode. */ ++ EXT2_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; ++ inode->i_ctime = CURRENT_TIME; ++ if (IS_SYNC(inode)) { ++ error = ext2_sync_inode (inode); ++ if (error) ++ goto cleanup; ++ } else ++ mark_inode_dirty(inode); ++ ++ error = 0; ++ if (old_bh && old_bh != new_bh) { ++ /* ++ * If there was an old block, and we are not still using it, ++ * we now release the old block. ++ */ ++ unsigned int refcount = le32_to_cpu(HDR(old_bh)->h_refcount); ++ ++ if (refcount == 1) { ++ /* Free the old block. */ ++ ea_bdebug(old_bh, "freeing"); ++ ext2_xattr_free_block(inode, old_bh->b_blocknr); ++ mark_buffer_clean(old_bh); ++ } else { ++ /* Decrement the refcount only. */ ++ refcount--; ++ HDR(old_bh)->h_refcount = cpu_to_le32(refcount); ++ ext2_xattr_quota_free(inode); ++ mark_buffer_dirty(old_bh); ++ ea_bdebug(old_bh, "refcount now=%d", refcount); ++ } ++ } ++ ++cleanup: ++ if (old_bh != new_bh) ++ brelse(new_bh); ++ ++ return error; ++} ++ ++/* ++ * ext2_xattr_delete_inode() ++ * ++ * Free extended attribute resources associated with this inode. This ++ * is called immediately before an inode is freed. ++ */ ++void ++ext2_xattr_delete_inode(struct inode *inode) ++{ ++ struct buffer_head *bh; ++ unsigned int block = EXT2_I(inode)->i_file_acl; ++ ++ if (!block) ++ return; ++ down(&ext2_xattr_sem); ++ ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) { ++ ext2_error(inode->i_sb, "ext2_xattr_delete_inode", ++ "inode %ld: block %d read error", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "b_count=%d", atomic_read(&(bh->b_count))); ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT2_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++ ext2_error(inode->i_sb, "ext2_xattr_delete_inode", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ if (HDR(bh)->h_refcount == cpu_to_le32(1)) { ++ ext2_xattr_cache_remove(bh); ++ ext2_xattr_free_block(inode, block); ++ bforget(bh); ++ bh = NULL; ++ } else { ++ HDR(bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ mark_buffer_dirty(bh); ++ if (IS_SYNC(inode)) { ++ ll_rw_block(WRITE, 1, &bh); ++ wait_on_buffer(bh); ++ } ++ ext2_xattr_quota_free(inode); ++ } ++ EXT2_I(inode)->i_file_acl = 0; ++ ++cleanup: ++ brelse(bh); ++ up(&ext2_xattr_sem); ++} ++ ++/* ++ * ext2_xattr_put_super() ++ * ++ * This is called when a file system is unmounted. ++ */ ++void ++ext2_xattr_put_super(struct super_block *sb) ++{ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ mb_cache_shrink(ext2_xattr_cache, sb->s_dev); ++#endif ++} ++ ++#ifdef CONFIG_EXT2_FS_XATTR_SHARING ++ ++/* ++ * ext2_xattr_cache_insert() ++ * ++ * Create a new entry in the extended attribute cache, and insert ++ * it unless such an entry is already in the cache. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++static int ++ext2_xattr_cache_insert(struct buffer_head *bh) ++{ ++ __u32 hash = le32_to_cpu(HDR(bh)->h_hash); ++ struct mb_cache_entry *ce; ++ int error; ++ ++ ce = mb_cache_entry_alloc(ext2_xattr_cache); ++ if (!ce) ++ return -ENOMEM; ++ error = mb_cache_entry_insert(ce, bh->b_dev, bh->b_blocknr, &hash); ++ if (error) { ++ mb_cache_entry_free(ce); ++ if (error == -EBUSY) { ++ ea_bdebug(bh, "already in cache (%d cache entries)", ++ atomic_read(&ext2_xattr_cache->c_entry_count)); ++ error = 0; ++ } ++ } else { ++ ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash, ++ atomic_read(&ext2_xattr_cache->c_entry_count)); ++ mb_cache_entry_release(ce); ++ } ++ return error; ++} ++ ++/* ++ * ext2_xattr_cmp() ++ * ++ * Compare two extended attribute blocks for equality. ++ * ++ * Returns 0 if the blocks are equal, 1 if they differ, and ++ * a negative error number on errors. ++ */ ++static int ++ext2_xattr_cmp(struct ext2_xattr_header *header1, ++ struct ext2_xattr_header *header2) ++{ ++ struct ext2_xattr_entry *entry1, *entry2; ++ ++ entry1 = ENTRY(header1+1); ++ entry2 = ENTRY(header2+1); ++ while (!IS_LAST_ENTRY(entry1)) { ++ if (IS_LAST_ENTRY(entry2)) ++ return 1; ++ if (entry1->e_hash != entry2->e_hash || ++ entry1->e_name_len != entry2->e_name_len || ++ entry1->e_value_size != entry2->e_value_size || ++ memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) ++ return 1; ++ if (entry1->e_value_block != 0 || entry2->e_value_block != 0) ++ return -EIO; ++ if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), ++ (char *)header2 + le16_to_cpu(entry2->e_value_offs), ++ le32_to_cpu(entry1->e_value_size))) ++ return 1; ++ ++ entry1 = EXT2_XATTR_NEXT(entry1); ++ entry2 = EXT2_XATTR_NEXT(entry2); ++ } ++ if (!IS_LAST_ENTRY(entry2)) ++ return 1; ++ return 0; ++} ++ ++/* ++ * ext2_xattr_cache_find() ++ * ++ * Find an identical extended attribute block. ++ * ++ * Returns a pointer to the block found, or NULL if such a block was ++ * not found or an error occurred. ++ */ ++static struct buffer_head * ++ext2_xattr_cache_find(struct inode *inode, struct ext2_xattr_header *header) ++{ ++ __u32 hash = le32_to_cpu(header->h_hash); ++ struct mb_cache_entry *ce; ++ ++ if (!header->h_hash) ++ return NULL; /* never share */ ++ ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); ++ ce = mb_cache_entry_find_first(ext2_xattr_cache, 0, inode->i_dev, hash); ++ while (ce) { ++ struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block); ++ ++ if (!bh) { ++ ext2_error(inode->i_sb, "ext2_xattr_cache_find", ++ "inode %ld: block %ld read error", ++ inode->i_ino, ce->e_block); ++ } else if (le32_to_cpu(HDR(bh)->h_refcount) > ++ EXT2_XATTR_REFCOUNT_MAX) { ++ ea_idebug(inode, "block %ld refcount %d>%d",ce->e_block, ++ le32_to_cpu(HDR(bh)->h_refcount), ++ EXT2_XATTR_REFCOUNT_MAX); ++ } else if (!ext2_xattr_cmp(header, HDR(bh))) { ++ ea_bdebug(bh, "b_count=%d",atomic_read(&(bh->b_count))); ++ mb_cache_entry_release(ce); ++ return bh; ++ } ++ brelse(bh); ++ ce = mb_cache_entry_find_next(ce, 0, inode->i_dev, hash); ++ } ++ return NULL; ++} ++ ++/* ++ * ext2_xattr_cache_remove() ++ * ++ * Remove the cache entry of a block from the cache. Called when a ++ * block becomes invalid. ++ */ ++static void ++ext2_xattr_cache_remove(struct buffer_head *bh) ++{ ++ struct mb_cache_entry *ce; ++ ++ ce = mb_cache_entry_get(ext2_xattr_cache, bh->b_dev, bh->b_blocknr); ++ if (ce) { ++ ea_bdebug(bh, "removing (%d cache entries remaining)", ++ atomic_read(&ext2_xattr_cache->c_entry_count)-1); ++ mb_cache_entry_free(ce); ++ } else ++ ea_bdebug(bh, "no cache entry"); ++} ++ ++#define NAME_HASH_SHIFT 5 ++#define VALUE_HASH_SHIFT 16 ++ ++/* ++ * ext2_xattr_hash_entry() ++ * ++ * Compute the hash of an extended attribute. ++ */ ++static inline void ext2_xattr_hash_entry(struct ext2_xattr_header *header, ++ struct ext2_xattr_entry *entry) ++{ ++ __u32 hash = 0; ++ char *name = entry->e_name; ++ int n; ++ ++ for (n=0; n < entry->e_name_len; n++) { ++ hash = (hash << NAME_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ ++ *name++; ++ } ++ ++ if (entry->e_value_block == 0 && entry->e_value_size != 0) { ++ __u32 *value = (__u32 *)((char *)header + ++ le16_to_cpu(entry->e_value_offs)); ++ for (n = (le32_to_cpu(entry->e_value_size) + ++ EXT2_XATTR_ROUND) >> EXT2_XATTR_PAD_BITS; n; n--) { ++ hash = (hash << VALUE_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ ++ le32_to_cpu(*value++); ++ } ++ } ++ entry->e_hash = cpu_to_le32(hash); ++} ++ ++#undef NAME_HASH_SHIFT ++#undef VALUE_HASH_SHIFT ++ ++#define BLOCK_HASH_SHIFT 16 ++ ++/* ++ * ext2_xattr_rehash() ++ * ++ * Re-compute the extended attribute hash value after an entry has changed. ++ */ ++static void ext2_xattr_rehash(struct ext2_xattr_header *header, ++ struct ext2_xattr_entry *entry) ++{ ++ struct ext2_xattr_entry *here; ++ __u32 hash = 0; ++ ++ ext2_xattr_hash_entry(header, entry); ++ here = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(here)) { ++ if (!here->e_hash) { ++ /* Block is not shared if an entry's hash value == 0 */ ++ hash = 0; ++ break; ++ } ++ hash = (hash << BLOCK_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ ++ le32_to_cpu(here->e_hash); ++ here = EXT2_XATTR_NEXT(here); ++ } ++ header->h_hash = cpu_to_le32(hash); ++} ++ ++#undef BLOCK_HASH_SHIFT ++ ++int __init ++init_ext2_xattr(void) ++{ ++ ext2_xattr_cache = mb_cache_create("ext2_xattr", NULL, ++ sizeof(struct mb_cache_entry) + ++ sizeof(struct mb_cache_entry_index), 1, 61); ++ if (!ext2_xattr_cache) ++ return -ENOMEM; ++ ++ return 0; ++} ++ ++void ++exit_ext2_xattr(void) ++{ ++ mb_cache_destroy(ext2_xattr_cache); ++} ++ ++#else /* CONFIG_EXT2_FS_XATTR_SHARING */ ++ ++int __init ++init_ext2_xattr(void) ++{ ++ return 0; ++} ++ ++void ++exit_ext2_xattr(void) ++{ ++} ++ ++#endif /* CONFIG_EXT2_FS_XATTR_SHARING */ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext2/xattr_user.c Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,103 @@ ++/* ++ * linux/fs/ext2/xattr_user.c ++ * Handler for extended user attributes. ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++ ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++# include ++#endif ++ ++#define XATTR_USER_PREFIX "user." ++ ++static size_t ++ext2_xattr_user_list(char *list, struct inode *inode, ++ const char *name, int name_len) ++{ ++ const int prefix_len = sizeof(XATTR_USER_PREFIX)-1; ++ ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return 0; ++ ++ if (list) { ++ memcpy(list, XATTR_USER_PREFIX, prefix_len); ++ memcpy(list+prefix_len, name, name_len); ++ list[prefix_len + name_len] = '\0'; ++ } ++ return prefix_len + name_len + 1; ++} ++ ++static int ++ext2_xattr_user_get(struct inode *inode, const char *name, ++ void *buffer, size_t size) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++ error = ext2_permission_locked(inode, MAY_READ); ++#else ++ error = permission(inode, MAY_READ); ++#endif ++ if (error) ++ return error; ++ ++ return ext2_xattr_get(inode, EXT2_XATTR_INDEX_USER, name, ++ buffer, size); ++} ++ ++static int ++ext2_xattr_user_set(struct inode *inode, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++ if ( !S_ISREG(inode->i_mode) && ++ (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX)) ++ return -EPERM; ++#ifdef CONFIG_EXT2_FS_POSIX_ACL ++ error = ext2_permission_locked(inode, MAY_WRITE); ++#else ++ error = permission(inode, MAY_WRITE); ++#endif ++ if (error) ++ return error; ++ ++ return ext2_xattr_set(inode, EXT2_XATTR_INDEX_USER, name, ++ value, size, flags); ++} ++ ++struct ext2_xattr_handler ext2_xattr_user_handler = { ++ prefix: XATTR_USER_PREFIX, ++ list: ext2_xattr_user_list, ++ get: ext2_xattr_user_get, ++ set: ext2_xattr_user_set, ++}; ++ ++int __init ++init_ext2_xattr_user(void) ++{ ++ return ext2_xattr_register(EXT2_XATTR_INDEX_USER, ++ &ext2_xattr_user_handler); ++} ++ ++void ++exit_ext2_xattr_user(void) ++{ ++ ext2_xattr_unregister(EXT2_XATTR_INDEX_USER, ++ &ext2_xattr_user_handler); ++} +--- linux-2.4.20/fs/ext3/Makefile~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/Makefile Sat Apr 5 03:57:36 2003 +@@ -1,5 +1,5 @@ + # +-# Makefile for the linux ext2-filesystem routines. ++# Makefile for the linux ext3-filesystem routines. + # + # Note! Dependencies are done automagically by 'make dep', which also + # removes any old dependencies. DON'T put your own dependencies here +@@ -9,10 +9,14 @@ + + O_TARGET := ext3.o + +-export-objs := super.o inode.o ++export-objs := super.o inode.o xattr.o + + obj-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ +- ioctl.o namei.o super.o symlink.o hash.o ++ ioctl.o namei.o super.o symlink.o hash.o xattr.o + obj-m := $(O_TARGET) + ++export-objs += xattr.o ++obj-$(CONFIG_EXT3_FS_XATTR) += xattr.o ++obj-$(CONFIG_EXT3_FS_XATTR_USER) += xattr_user.o ++ + include $(TOPDIR)/Rules.make +--- linux-2.4.20/fs/ext3/file.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/file.c Sat Apr 5 03:57:18 2003 +@@ -23,6 +23,7 @@ + #include + #include + #include ++#include + #include + #include + +@@ -126,5 +127,9 @@ struct file_operations ext3_file_operati + struct inode_operations ext3_file_inode_operations = { + truncate: ext3_truncate, /* BKL held */ + setattr: ext3_setattr, /* BKL held */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; + +--- linux-2.4.20/fs/ext3/ialloc.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/ialloc.c Sat Apr 5 03:57:18 2003 +@@ -17,6 +17,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -216,6 +217,7 @@ void ext3_free_inode (handle_t *handle, + * as writing the quota to disk may need the lock as well. + */ + DQUOT_INIT(inode); ++ ext3_xattr_delete_inode(handle, inode); + DQUOT_FREE_INODE(inode); + DQUOT_DROP(inode); + +--- linux-2.4.20/fs/ext3/inode.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/inode.c Sat Apr 5 03:57:18 2003 +@@ -39,6 +39,18 @@ + */ + #undef SEARCH_FROM_ZERO + ++/* ++ * Test whether an inode is a fast symlink. ++ */ ++static inline int ext3_inode_is_fast_symlink(struct inode *inode) ++{ ++ int ea_blocks = inode->u.ext3_i.i_file_acl ? ++ (inode->i_sb->s_blocksize >> 9) : 0; ++ ++ return (S_ISLNK(inode->i_mode) && ++ inode->i_blocks - ea_blocks == 0); ++} ++ + /* The ext3 forget function must perform a revoke if we are freeing data + * which has been journaled. Metadata (eg. indirect blocks) must be + * revoked in all cases. +@@ -48,7 +60,7 @@ + * still needs to be revoked. + */ + +-static int ext3_forget(handle_t *handle, int is_metadata, ++int ext3_forget(handle_t *handle, int is_metadata, + struct inode *inode, struct buffer_head *bh, + int blocknr) + { +@@ -164,9 +176,7 @@ void ext3_delete_inode (struct inode * i + { + handle_t *handle; + +- if (is_bad_inode(inode) || +- inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) ++ if (is_bad_inode(inode)) + goto no_delete; + + lock_kernel(); +@@ -1855,6 +1865,8 @@ void ext3_truncate(struct inode * inode) + if (!(S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || + S_ISLNK(inode->i_mode))) + return; ++ if (ext3_inode_is_fast_symlink(inode)) ++ return; + if (IS_APPEND(inode) || IS_IMMUTABLE(inode)) + return; + +@@ -2002,8 +2014,6 @@ int ext3_get_inode_loc (struct inode *in + struct ext3_group_desc * gdp; + + if ((inode->i_ino != EXT3_ROOT_INO && +- inode->i_ino != EXT3_ACL_IDX_INO && +- inode->i_ino != EXT3_ACL_DATA_INO && + inode->i_ino != EXT3_JOURNAL_INO && + inode->i_ino < EXT3_FIRST_INO(inode->i_sb)) || + inode->i_ino > le32_to_cpu( +@@ -2130,10 +2140,7 @@ void ext3_read_inode(struct inode * inod + + brelse (iloc.bh); + +- if (inode->i_ino == EXT3_ACL_IDX_INO || +- inode->i_ino == EXT3_ACL_DATA_INO) +- /* Nothing to do */ ; +- else if (S_ISREG(inode->i_mode)) { ++ if (S_ISREG(inode->i_mode)) { + inode->i_op = &ext3_file_inode_operations; + inode->i_fop = &ext3_file_operations; + inode->i_mapping->a_ops = &ext3_aops; +@@ -2141,15 +2148,17 @@ void ext3_read_inode(struct inode * inod + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; + } else if (S_ISLNK(inode->i_mode)) { +- if (!inode->i_blocks) ++ if (ext3_inode_is_fast_symlink(inode)) + inode->i_op = &ext3_fast_symlink_inode_operations; + else { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext3_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + } +- } else ++ } else { ++ inode->i_op = &ext3_special_inode_operations; + init_special_inode(inode, inode->i_mode, + le32_to_cpu(iloc.raw_inode->i_block[0])); ++ } + /* inode->i_attr_flags = 0; unused */ + if (inode->u.ext3_i.i_flags & EXT3_SYNC_FL) { + /* inode->i_attr_flags |= ATTR_FLAG_SYNCRONOUS; unused */ +--- linux-2.4.20/fs/ext3/namei.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/namei.c Sat Apr 5 03:57:18 2003 +@@ -29,6 +29,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -1611,7 +1612,7 @@ static int ext3_mkdir(struct inode * dir + if (IS_SYNC(dir)) + handle->h_sync = 1; + +- inode = ext3_new_inode (handle, dir, S_IFDIR); ++ inode = ext3_new_inode (handle, dir, S_IFDIR | mode); + err = PTR_ERR(inode); + if (IS_ERR(inode)) + goto out_stop; +@@ -1619,7 +1620,6 @@ static int ext3_mkdir(struct inode * dir + inode->i_op = &ext3_dir_inode_operations; + inode->i_fop = &ext3_dir_operations; + inode->i_size = EXT3_I(inode)->i_disksize = inode->i_sb->s_blocksize; +- inode->i_blocks = 0; + dir_block = ext3_bread (handle, inode, 0, 1, &err); + if (!dir_block) { + inode->i_nlink--; /* is this nlink == 0? */ +@@ -1646,9 +1646,6 @@ static int ext3_mkdir(struct inode * dir + BUFFER_TRACE(dir_block, "call ext3_journal_dirty_metadata"); + ext3_journal_dirty_metadata(handle, dir_block); + brelse (dir_block); +- inode->i_mode = S_IFDIR | mode; +- if (dir->i_mode & S_ISGID) +- inode->i_mode |= S_ISGID; + ext3_mark_inode_dirty(handle, inode); + err = ext3_add_entry (handle, dentry, inode); + if (err) { +@@ -2017,7 +2014,7 @@ static int ext3_symlink (struct inode * + goto out_stop; + + if (l > sizeof (EXT3_I(inode)->i_data)) { +- inode->i_op = &page_symlink_inode_operations; ++ inode->i_op = &ext3_symlink_inode_operations; + inode->i_mapping->a_ops = &ext3_aops; + /* + * block_symlink() calls back into ext3_prepare/commit_write. +@@ -2244,4 +2241,16 @@ struct inode_operations ext3_dir_inode_o + rmdir: ext3_rmdir, /* BKL held */ + mknod: ext3_mknod, /* BKL held */ + rename: ext3_rename, /* BKL held */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; ++ ++struct inode_operations ext3_special_inode_operations = { ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ ++}; ++ +--- linux-2.4.20/fs/ext3/super.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/super.c Sat Apr 5 03:57:18 2003 +@@ -24,6 +24,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -404,6 +405,7 @@ void ext3_put_super (struct super_block + kdev_t j_dev = sbi->s_journal->j_dev; + int i; + ++ ext3_xattr_put_super(sb); + journal_destroy(sbi->s_journal); + if (!(sb->s_flags & MS_RDONLY)) { + EXT3_CLEAR_INCOMPAT_FEATURE(sb, EXT3_FEATURE_INCOMPAT_RECOVER); +@@ -499,6 +501,7 @@ static int parse_options (char * options + int is_remount) + { + unsigned long *mount_options = &sbi->s_mount_opt; ++ + uid_t *resuid = &sbi->s_resuid; + gid_t *resgid = &sbi->s_resgid; + char * this_char; +@@ -511,6 +514,13 @@ static int parse_options (char * options + this_char = strtok (NULL, ",")) { + if ((value = strchr (this_char, '=')) != NULL) + *value++ = 0; ++#ifdef CONFIG_EXT3_FS_XATTR_USER ++ if (!strcmp (this_char, "user_xattr")) ++ set_opt (*mount_options, XATTR_USER); ++ else if (!strcmp (this_char, "nouser_xattr")) ++ clear_opt (*mount_options, XATTR_USER); ++ else ++#endif + if (!strcmp (this_char, "bsddf")) + clear_opt (*mount_options, MINIX_DF); + else if (!strcmp (this_char, "nouid32")) { +@@ -928,6 +938,12 @@ struct super_block * ext3_read_super (st + sbi->s_mount_opt = 0; + sbi->s_resuid = EXT3_DEF_RESUID; + sbi->s_resgid = EXT3_DEF_RESGID; ++ ++ /* Default extended attribute flags */ ++#ifdef CONFIG_EXT3_FS_XATTR_USER ++ /* set_opt(sbi->s_mount_opt, XATTR_USER); */ ++#endif ++ + if (!parse_options ((char *) data, &sb_block, sbi, &journal_inum, 0)) { + sb->s_dev = 0; + goto out_fail; +@@ -1767,12 +1783,27 @@ static DECLARE_FSTYPE_DEV(ext3_fs_type, + + static int __init init_ext3_fs(void) + { +- return register_filesystem(&ext3_fs_type); ++ int error = init_ext3_xattr(); ++ if (error) ++ return error; ++ error = init_ext3_xattr_user(); ++ if (error) ++ goto fail; ++ error = register_filesystem(&ext3_fs_type); ++ if (!error) ++ return 0; ++ ++ exit_ext3_xattr_user(); ++fail: ++ exit_ext3_xattr(); ++ return error; + } + + static void __exit exit_ext3_fs(void) + { + unregister_filesystem(&ext3_fs_type); ++ exit_ext3_xattr_user(); ++ exit_ext3_xattr(); + } + + EXPORT_SYMBOL(ext3_force_commit); +--- linux-2.4.20/fs/ext3/symlink.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/ext3/symlink.c Sat Apr 5 03:57:18 2003 +@@ -20,6 +20,7 @@ + #include + #include + #include ++#include + + static int ext3_readlink(struct dentry *dentry, char *buffer, int buflen) + { +@@ -33,7 +34,20 @@ static int ext3_follow_link(struct dentr + return vfs_follow_link(nd, s); + } + ++struct inode_operations ext3_symlink_inode_operations = { ++ readlink: page_readlink, /* BKL not held. Don't need */ ++ follow_link: page_follow_link, /* BKL not held. Don't need */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ ++}; ++ + struct inode_operations ext3_fast_symlink_inode_operations = { + readlink: ext3_readlink, /* BKL not held. Don't need */ + follow_link: ext3_follow_link, /* BKL not held. Don't need */ ++ setxattr: ext3_setxattr, /* BKL held */ ++ getxattr: ext3_getxattr, /* BKL held */ ++ listxattr: ext3_listxattr, /* BKL held */ ++ removexattr: ext3_removexattr, /* BKL held */ + }; +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext3/xattr.c Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,1224 @@ ++/* ++ * linux/fs/ext3/xattr.c ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ * ++ * Fix by Harrison Xing . ++ * Ext3 code with a lot of help from Eric Jarman . ++ * Extended attributes for symlinks and special files added per ++ * suggestion of Luka Renko . ++ */ ++ ++/* ++ * Extended attributes are stored on disk blocks allocated outside of ++ * any inode. The i_file_acl field is then made to point to this allocated ++ * block. If all extended attributes of an inode are identical, these ++ * inodes may share the same extended attribute block. Such situations ++ * are automatically detected by keeping a cache of recent attribute block ++ * numbers and hashes over the block's contents in memory. ++ * ++ * ++ * Extended attribute block layout: ++ * ++ * +------------------+ ++ * | header | ++ * | entry 1 | | ++ * | entry 2 | | growing downwards ++ * | entry 3 | v ++ * | four null bytes | ++ * | . . . | ++ * | value 1 | ^ ++ * | value 3 | | growing upwards ++ * | value 2 | | ++ * +------------------+ ++ * ++ * The block header is followed by multiple entry descriptors. These entry ++ * descriptors are variable in size, and alligned to EXT3_XATTR_PAD ++ * byte boundaries. The entry descriptors are sorted by attribute name, ++ * so that two extended attribute blocks can be compared efficiently. ++ * ++ * Attribute values are aligned to the end of the block, stored in ++ * no specific order. They are also padded to EXT3_XATTR_PAD byte ++ * boundaries. No additional gaps are left between them. ++ * ++ * Locking strategy ++ * ---------------- ++ * The VFS already holds the BKL and the inode->i_sem semaphore when any of ++ * the xattr inode operations are called, so we are guaranteed that only one ++ * processes accesses extended attributes of an inode at any time. ++ * ++ * For writing we also grab the ext3_xattr_sem semaphore. This ensures that ++ * only a single process is modifying an extended attribute block, even ++ * if the block is shared among inodes. ++ * ++ * Note for porting to 2.5 ++ * ----------------------- ++ * The BKL will no longer be held in the xattr inode operations. ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++#define EXT3_EA_USER "user." ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++# define mark_buffer_dirty(bh) mark_buffer_dirty(bh, 1) ++#endif ++ ++#define HDR(bh) ((struct ext3_xattr_header *)((bh)->b_data)) ++#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr)) ++#define FIRST_ENTRY(bh) ENTRY(HDR(bh)+1) ++#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0) ++ ++#ifdef EXT3_XATTR_DEBUG ++# define ea_idebug(inode, f...) do { \ ++ printk(KERN_DEBUG "inode %s:%ld: ", \ ++ kdevname(inode->i_dev), inode->i_ino); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++# define ea_bdebug(bh, f...) do { \ ++ printk(KERN_DEBUG "block %s:%ld: ", \ ++ kdevname(bh->b_dev), bh->b_blocknr); \ ++ printk(f); \ ++ printk("\n"); \ ++ } while (0) ++#else ++# define ea_idebug(f...) ++# define ea_bdebug(f...) ++#endif ++ ++static int ext3_xattr_set2(handle_t *, struct inode *, struct buffer_head *, ++ struct ext3_xattr_header *); ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++static int ext3_xattr_cache_insert(struct buffer_head *); ++static struct buffer_head *ext3_xattr_cache_find(struct inode *, ++ struct ext3_xattr_header *); ++static void ext3_xattr_cache_remove(struct buffer_head *); ++static void ext3_xattr_rehash(struct ext3_xattr_header *, ++ struct ext3_xattr_entry *); ++ ++static struct mb_cache *ext3_xattr_cache; ++ ++#else ++# define ext3_xattr_cache_insert(bh) 0 ++# define ext3_xattr_cache_find(inode, header) NULL ++# define ext3_xattr_cache_remove(bh) while(0) {} ++# define ext3_xattr_rehash(header, entry) while(0) {} ++#endif ++ ++/* ++ * If a file system does not share extended attributes among inodes, ++ * we should not need the ext3_xattr_sem semaphore. However, the ++ * filesystem may still contain shared blocks, so we always take ++ * the lock. ++ */ ++ ++DECLARE_MUTEX(ext3_xattr_sem); ++ ++static inline int ++ext3_xattr_new_block(handle_t *handle, struct inode *inode, ++ int * errp, int force) ++{ ++ struct super_block *sb = inode->i_sb; ++ int goal = le32_to_cpu(EXT3_SB(sb)->s_es->s_first_data_block) + ++ EXT3_I(inode)->i_block_group * EXT3_BLOCKS_PER_GROUP(sb); ++ ++ /* How can we enforce the allocation? */ ++ int block = ext3_new_block(handle, inode, goal, 0, 0, errp); ++#ifdef OLD_QUOTAS ++ if (!*errp) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#endif ++ return block; ++} ++ ++static inline int ++ext3_xattr_quota_alloc(struct inode *inode, int force) ++{ ++ /* How can we enforce the allocation? */ ++#ifdef OLD_QUOTAS ++ int error = DQUOT_ALLOC_BLOCK(inode->i_sb, inode, 1); ++ if (!error) ++ inode->i_blocks += inode->i_sb->s_blocksize >> 9; ++#else ++ int error = DQUOT_ALLOC_BLOCK(inode, 1); ++#endif ++ return error; ++} ++ ++#ifdef OLD_QUOTAS ++ ++static inline void ++ext3_xattr_quota_free(struct inode *inode) ++{ ++ DQUOT_FREE_BLOCK(inode->i_sb, inode, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++static inline void ++ext3_xattr_free_block(handle_t *handle, struct inode * inode, ++ unsigned long block) ++{ ++ ext3_free_blocks(handle, inode, block, 1); ++ inode->i_blocks -= inode->i_sb->s_blocksize >> 9; ++} ++ ++#else ++# define ext3_xattr_quota_free(inode) \ ++ DQUOT_FREE_BLOCK(inode, 1) ++# define ext3_xattr_free_block(handle, inode, block) \ ++ ext3_free_blocks(handle, inode, block, 1) ++#endif ++ ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,18) ++ ++static inline struct buffer_head * ++sb_bread(struct super_block *sb, int block) ++{ ++ return bread(sb->s_dev, block, sb->s_blocksize); ++} ++ ++static inline struct buffer_head * ++sb_getblk(struct super_block *sb, int block) ++{ ++ return getblk(sb->s_dev, block, sb->s_blocksize); ++} ++ ++#endif ++ ++struct ext3_xattr_handler *ext3_xattr_handlers[EXT3_XATTR_INDEX_MAX]; ++rwlock_t ext3_handler_lock = RW_LOCK_UNLOCKED; ++ ++int ++ext3_xattr_register(int name_index, struct ext3_xattr_handler *handler) ++{ ++ int error = -EINVAL; ++ ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ if (!ext3_xattr_handlers[name_index-1]) { ++ ext3_xattr_handlers[name_index-1] = handler; ++ error = 0; ++ } ++ write_unlock(&ext3_handler_lock); ++ } ++ return error; ++} ++ ++void ++ext3_xattr_unregister(int name_index, struct ext3_xattr_handler *handler) ++{ ++ if (name_index > 0 || name_index <= EXT3_XATTR_INDEX_MAX) { ++ write_lock(&ext3_handler_lock); ++ ext3_xattr_handlers[name_index-1] = NULL; ++ write_unlock(&ext3_handler_lock); ++ } ++} ++ ++static inline const char * ++strcmp_prefix(const char *a, const char *a_prefix) ++{ ++ while (*a_prefix && *a == *a_prefix) { ++ a++; ++ a_prefix++; ++ } ++ return *a_prefix ? NULL : a; ++} ++ ++/* ++ * Decode the extended attribute name, and translate it into ++ * the name_index and name suffix. ++ */ ++static inline struct ext3_xattr_handler * ++ext3_xattr_resolve_name(const char **name) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ int i; ++ ++ if (!*name) ++ return NULL; ++ read_lock(&ext3_handler_lock); ++ for (i=0; iprefix); ++ if (n) { ++ handler = ext3_xattr_handlers[i]; ++ *name = n; ++ break; ++ } ++ } ++ } ++ read_unlock(&ext3_handler_lock); ++ return handler; ++} ++ ++static inline struct ext3_xattr_handler * ++ext3_xattr_handler(int name_index) ++{ ++ struct ext3_xattr_handler *handler = NULL; ++ if (name_index > 0 && name_index <= EXT3_XATTR_INDEX_MAX) { ++ read_lock(&ext3_handler_lock); ++ handler = ext3_xattr_handlers[name_index-1]; ++ read_unlock(&ext3_handler_lock); ++ } ++ return handler; ++} ++ ++/* ++ * Inode operation getxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_getxattr(struct dentry *dentry, const char *name, ++ void *buffer, size_t size) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->get(inode, name, buffer, size); ++} ++ ++/* ++ * Inode operation listxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++ssize_t ++ext3_listxattr(struct dentry *dentry, char *buffer, size_t size) ++{ ++ return ext3_xattr_list(dentry->d_inode, buffer, size); ++} ++ ++/* ++ * Inode operation setxattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_setxattr(struct dentry *dentry, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ if (size == 0) ++ value = ""; /* empty EA, do not remove */ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, value, size, flags); ++} ++ ++/* ++ * Inode operation removexattr() ++ * ++ * dentry->d_inode->i_sem down ++ * BKL held [before 2.5.x] ++ */ ++int ++ext3_removexattr(struct dentry *dentry, const char *name) ++{ ++ struct ext3_xattr_handler *handler; ++ struct inode *inode = dentry->d_inode; ++ ++ handler = ext3_xattr_resolve_name(&name); ++ if (!handler) ++ return -ENOTSUP; ++ return handler->set(inode, name, NULL, 0, XATTR_REPLACE); ++} ++ ++/* ++ * ext3_xattr_get() ++ * ++ * Copy an extended attribute into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size; ++ char *end; ++ int name_len, error; ++ ++ ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld", ++ name_index, name, buffer, (long)buffer_size); ++ ++ if (name == NULL) ++ return -EINVAL; ++ if (!EXT3_I(inode)->i_file_acl) ++ return -ENOATTR; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_get", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* find named attribute */ ++ name_len = strlen(name); ++ ++ error = -ERANGE; ++ if (name_len > 255) ++ goto cleanup; ++ entry = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (name_index == entry->e_name_index && ++ name_len == entry->e_name_len && ++ memcmp(name, entry->e_name, name_len) == 0) ++ goto found; ++ entry = next; ++ } ++ /* Check the remaining name entries */ ++ while (!IS_LAST_ENTRY(entry)) { ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ entry = next; ++ } ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ error = -ENOATTR; ++ goto cleanup; ++found: ++ /* check the buffer size */ ++ if (entry->e_value_block != 0) ++ goto bad_block; ++ size = le32_to_cpu(entry->e_value_size); ++ if (size > inode->i_sb->s_blocksize || ++ le16_to_cpu(entry->e_value_offs) + size > inode->i_sb->s_blocksize) ++ goto bad_block; ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (buffer) { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ /* return value of attribute */ ++ memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs), ++ size); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_list() ++ * ++ * Copy a list of attribute names into the buffer ++ * provided, or compute the buffer size required. ++ * Buffer is NULL to compute the size of the buffer required. ++ * ++ * Returns a negative error number on failure, or the number of bytes ++ * used / required on success. ++ */ ++int ++ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) ++{ ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_entry *entry; ++ unsigned int block, size = 0; ++ char *buf, *end; ++ int error; ++ ++ ea_idebug(inode, "buffer=%p, buffer_size=%ld", ++ buffer, (long)buffer_size); ++ ++ if (!EXT3_I(inode)->i_file_acl) ++ return 0; ++ block = EXT3_I(inode)->i_file_acl; ++ ea_idebug(inode, "reading block %d", block); ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) ++ return -EIO; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), le32_to_cpu(HDR(bh)->h_refcount)); ++ end = bh->b_data + bh->b_size; ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(inode->i_sb, "ext3_xattr_list", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* compute the size required for the list of attribute names */ ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ struct ext3_xattr_entry *next = ++ EXT3_XATTR_NEXT(entry); ++ if ((char *)next >= end) ++ goto bad_block; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) ++ size += handler->list(NULL, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ ++ if (ext3_xattr_cache_insert(bh)) ++ ea_idebug(inode, "cache insert failed"); ++ if (!buffer) { ++ error = size; ++ goto cleanup; ++ } else { ++ error = -ERANGE; ++ if (size > buffer_size) ++ goto cleanup; ++ } ++ ++ /* list the attribute names */ ++ buf = buffer; ++ for (entry = FIRST_ENTRY(bh); !IS_LAST_ENTRY(entry); ++ entry = EXT3_XATTR_NEXT(entry)) { ++ struct ext3_xattr_handler *handler; ++ ++ handler = ext3_xattr_handler(entry->e_name_index); ++ if (handler) ++ buf += handler->list(buf, inode, entry->e_name, ++ entry->e_name_len); ++ } ++ error = size; ++ ++cleanup: ++ brelse(bh); ++ ++ return error; ++} ++ ++/* ++ * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is ++ * not set, set it. ++ */ ++static void ext3_xattr_update_super_block(handle_t *handle, ++ struct super_block *sb) ++{ ++ if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR)) ++ return; ++ ++ lock_super(sb); ++ ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh); ++#if LINUX_VERSION_CODE < KERNEL_VERSION(2,4,0) ++ EXT3_SB(sb)->s_feature_compat |= EXT3_FEATURE_COMPAT_EXT_ATTR; ++#endif ++ EXT3_SB(sb)->s_es->s_feature_compat |= ++ cpu_to_le32(EXT3_FEATURE_COMPAT_EXT_ATTR); ++ sb->s_dirt = 1; ++ ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh); ++ unlock_super(sb); ++} ++ ++/* ++ * ext3_xattr_set() ++ * ++ * Create, replace or remove an extended attribute for this inode. Buffer ++ * is NULL to remove an existing extended attribute, and non-NULL to ++ * either replace an existing extended attribute, or create a new extended ++ * attribute. The flags XATTR_REPLACE and XATTR_CREATE ++ * specify that an extended attribute must exist and must not exist ++ * previous to the call, respectively. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t value_len, int flags) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *bh = NULL; ++ struct ext3_xattr_header *header = NULL; ++ struct ext3_xattr_entry *here, *last; ++ unsigned int name_len; ++ int block = EXT3_I(inode)->i_file_acl; ++ int min_offs = sb->s_blocksize, not_found = 1, free, error; ++ char *end; ++ ++ /* ++ * header -- Points either into bh, or to a temporarily ++ * allocated buffer. ++ * here -- The named entry found, or the place for inserting, within ++ * the block pointed to by header. ++ * last -- Points right after the last named entry within the block ++ * pointed to by header. ++ * min_offs -- The offset of the first value (values are aligned ++ * towards the end of the block). ++ * end -- Points right after the block pointed to by header. ++ */ ++ ++ ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld", ++ name_index, name, value, (long)value_len); ++ ++ if (IS_RDONLY(inode)) ++ return -EROFS; ++ if (IS_IMMUTABLE(inode) || IS_APPEND(inode)) ++ return -EPERM; ++ if (value == NULL) ++ value_len = 0; ++ if (name == NULL) ++ return -EINVAL; ++ name_len = strlen(name); ++ if (name_len > 255 || value_len > sb->s_blocksize) ++ return -ERANGE; ++ down(&ext3_xattr_sem); ++ ++ if (block) { ++ /* The inode already has an extended attribute block. */ ++ bh = sb_bread(sb, block); ++ error = -EIO; ++ if (!bh) ++ goto cleanup; ++ ea_bdebug(bh, "b_count=%d, refcount=%d", ++ atomic_read(&(bh->b_count)), ++ le32_to_cpu(HDR(bh)->h_refcount)); ++ header = HDR(bh); ++ end = bh->b_data + bh->b_size; ++ if (header->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ header->h_blocks != cpu_to_le32(1)) { ++bad_block: ext3_error(sb, "ext3_xattr_set", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ /* Find the named attribute. */ ++ here = FIRST_ENTRY(bh); ++ while (!IS_LAST_ENTRY(here)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(here); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!here->e_value_block && here->e_value_size) { ++ int offs = le16_to_cpu(here->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ not_found = name_index - here->e_name_index; ++ if (!not_found) ++ not_found = name_len - here->e_name_len; ++ if (!not_found) ++ not_found = memcmp(name, here->e_name,name_len); ++ if (not_found <= 0) ++ break; ++ here = next; ++ } ++ last = here; ++ /* We still need to compute min_offs and last. */ ++ while (!IS_LAST_ENTRY(last)) { ++ struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(last); ++ if ((char *)next >= end) ++ goto bad_block; ++ if (!last->e_value_block && last->e_value_size) { ++ int offs = le16_to_cpu(last->e_value_offs); ++ if (offs < min_offs) ++ min_offs = offs; ++ } ++ last = next; ++ } ++ ++ /* Check whether we have enough space left. */ ++ free = min_offs - ((char*)last - (char*)header) - sizeof(__u32); ++ } else { ++ /* We will use a new extended attribute block. */ ++ free = sb->s_blocksize - ++ sizeof(struct ext3_xattr_header) - sizeof(__u32); ++ here = last = NULL; /* avoid gcc uninitialized warning. */ ++ } ++ ++ if (not_found) { ++ /* Request to remove a nonexistent attribute? */ ++ error = -ENOATTR; ++ if (flags & XATTR_REPLACE) ++ goto cleanup; ++ error = 0; ++ if (value == NULL) ++ goto cleanup; ++ else ++ free -= EXT3_XATTR_LEN(name_len); ++ } else { ++ /* Request to create an existing attribute? */ ++ error = -EEXIST; ++ if (flags & XATTR_CREATE) ++ goto cleanup; ++ if (!here->e_value_block && here->e_value_size) { ++ unsigned int size = le32_to_cpu(here->e_value_size); ++ ++ if (le16_to_cpu(here->e_value_offs) + size > ++ sb->s_blocksize || size > sb->s_blocksize) ++ goto bad_block; ++ free += EXT3_XATTR_SIZE(size); ++ } ++ } ++ free -= EXT3_XATTR_SIZE(value_len); ++ error = -ENOSPC; ++ if (free < 0) ++ goto cleanup; ++ ++ /* Here we know that we can set the new attribute. */ ++ ++ if (header) { ++ if (header->h_refcount == cpu_to_le32(1)) { ++ ea_bdebug(bh, "modifying in-place"); ++ ext3_xattr_cache_remove(bh); ++ error = ext3_journal_get_write_access(handle, bh); ++ if (error) ++ goto cleanup; ++ } else { ++ int offset; ++ ++ ea_bdebug(bh, "cloning"); ++ header = kmalloc(bh->b_size, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memcpy(header, HDR(bh), bh->b_size); ++ header->h_refcount = cpu_to_le32(1); ++ offset = (char *)header - bh->b_data; ++ here = ENTRY((char *)here + offset); ++ last = ENTRY((char *)last + offset); ++ } ++ } else { ++ /* Allocate a buffer where we construct the new block. */ ++ header = kmalloc(sb->s_blocksize, GFP_KERNEL); ++ error = -ENOMEM; ++ if (header == NULL) ++ goto cleanup; ++ memset(header, 0, sb->s_blocksize); ++ end = (char *)header + sb->s_blocksize; ++ header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC); ++ header->h_blocks = header->h_refcount = cpu_to_le32(1); ++ last = here = ENTRY(header+1); ++ } ++ ++ if (not_found) { ++ /* Insert the new name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ int rest = (char *)last - (char *)here; ++ memmove((char *)here + size, here, rest); ++ memset(here, 0, size); ++ here->e_name_index = name_index; ++ here->e_name_len = name_len; ++ memcpy(here->e_name, name, name_len); ++ } else { ++ /* Remove the old value. */ ++ if (!here->e_value_block && here->e_value_size) { ++ char *first_val = (char *)header + min_offs; ++ int offs = le16_to_cpu(here->e_value_offs); ++ char *val = (char *)header + offs; ++ size_t size = EXT3_XATTR_SIZE( ++ le32_to_cpu(here->e_value_size)); ++ memmove(first_val + size, first_val, val - first_val); ++ memset(first_val, 0, size); ++ here->e_value_offs = 0; ++ min_offs += size; ++ ++ /* Adjust all value offsets. */ ++ last = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(last)) { ++ int o = le16_to_cpu(last->e_value_offs); ++ if (!last->e_value_block && o < offs) ++ last->e_value_offs = ++ cpu_to_le16(o + size); ++ last = EXT3_XATTR_NEXT(last); ++ } ++ } ++ if (value == NULL) { ++ /* Remove this attribute. */ ++ if (EXT3_XATTR_NEXT(ENTRY(header+1)) == last) { ++ /* This block is now empty. */ ++ error = ext3_xattr_set2(handle, inode, bh,NULL); ++ goto cleanup; ++ } else { ++ /* Remove the old name. */ ++ int size = EXT3_XATTR_LEN(name_len); ++ last = ENTRY((char *)last - size); ++ memmove(here, (char*)here + size, ++ (char*)last - (char*)here); ++ memset(last, 0, size); ++ } ++ } ++ } ++ ++ if (value != NULL) { ++ /* Insert the new value. */ ++ here->e_value_size = cpu_to_le32(value_len); ++ if (value_len) { ++ size_t size = EXT3_XATTR_SIZE(value_len); ++ char *val = (char *)header + min_offs - size; ++ here->e_value_offs = ++ cpu_to_le16((char *)val - (char *)header); ++ memset(val + size - EXT3_XATTR_PAD, 0, ++ EXT3_XATTR_PAD); /* Clear the pad bytes. */ ++ memcpy(val, value, value_len); ++ } ++ } ++ ext3_xattr_rehash(header, here); ++ ++ error = ext3_xattr_set2(handle, inode, bh, header); ++ ++cleanup: ++ brelse(bh); ++ if (!(bh && header == HDR(bh))) ++ kfree(header); ++ up(&ext3_xattr_sem); ++ ++ return error; ++} ++ ++/* ++ * Second half of ext3_xattr_set(): Update the file system. ++ */ ++static int ++ext3_xattr_set2(handle_t *handle, struct inode *inode, ++ struct buffer_head *old_bh, struct ext3_xattr_header *header) ++{ ++ struct super_block *sb = inode->i_sb; ++ struct buffer_head *new_bh = NULL; ++ int error; ++ ++ if (header) { ++ new_bh = ext3_xattr_cache_find(inode, header); ++ if (new_bh) { ++ /* ++ * We found an identical block in the cache. ++ * The old block will be released after updating ++ * the inode. ++ */ ++ ea_bdebug(old_bh, "reusing block %ld", ++ new_bh->b_blocknr); ++ ++ error = -EDQUOT; ++ if (ext3_xattr_quota_alloc(inode, 1)) ++ goto cleanup; ++ ++ error = ext3_journal_get_write_access(handle, new_bh); ++ if (error) ++ goto cleanup; ++ HDR(new_bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(new_bh)->h_refcount) + 1); ++ ea_bdebug(new_bh, "refcount now=%d", ++ le32_to_cpu(HDR(new_bh)->h_refcount)); ++ } else if (old_bh && header == HDR(old_bh)) { ++ /* Keep this block. */ ++ new_bh = old_bh; ++ ext3_xattr_cache_insert(new_bh); ++ } else { ++ /* We need to allocate a new block */ ++ int force = EXT3_I(inode)->i_file_acl != 0; ++ int block = ext3_xattr_new_block(handle, inode, ++ &error, force); ++ if (error) ++ goto cleanup; ++ ea_idebug(inode, "creating block %d", block); ++ ++ new_bh = sb_getblk(sb, block); ++ if (!new_bh) { ++getblk_failed: ext3_xattr_free_block(handle, inode, block); ++ error = -EIO; ++ goto cleanup; ++ } ++ lock_buffer(new_bh); ++ error = ext3_journal_get_create_access(handle, new_bh); ++ if (error) { ++ unlock_buffer(new_bh); ++ goto getblk_failed; ++ } ++ memcpy(new_bh->b_data, header, new_bh->b_size); ++ mark_buffer_uptodate(new_bh, 1); ++ unlock_buffer(new_bh); ++ ext3_xattr_cache_insert(new_bh); ++ ++ ext3_xattr_update_super_block(handle, sb); ++ } ++ error = ext3_journal_dirty_metadata(handle, new_bh); ++ if (error) ++ goto cleanup; ++ } ++ ++ /* Update the inode. */ ++ EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0; ++ inode->i_ctime = CURRENT_TIME; ++ ext3_mark_inode_dirty(handle, inode); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ++ error = 0; ++ if (old_bh && old_bh != new_bh) { ++ /* ++ * If there was an old block, and we are not still using it, ++ * we now release the old block. ++ */ ++ unsigned int refcount = le32_to_cpu(HDR(old_bh)->h_refcount); ++ ++ error = ext3_journal_get_write_access(handle, old_bh); ++ if (error) ++ goto cleanup; ++ if (refcount == 1) { ++ /* Free the old block. */ ++ ea_bdebug(old_bh, "freeing"); ++ ext3_xattr_free_block(handle, inode, old_bh->b_blocknr); ++ ++ /* ext3_forget() calls bforget() for us, but we ++ let our caller release old_bh, so we need to ++ duplicate the handle before. */ ++ get_bh(old_bh); ++ ext3_forget(handle, 1, inode, old_bh,old_bh->b_blocknr); ++ } else { ++ /* Decrement the refcount only. */ ++ refcount--; ++ HDR(old_bh)->h_refcount = cpu_to_le32(refcount); ++ ext3_xattr_quota_free(inode); ++ ext3_journal_dirty_metadata(handle, old_bh); ++ ea_bdebug(old_bh, "refcount now=%d", refcount); ++ } ++ } ++ ++cleanup: ++ if (old_bh != new_bh) ++ brelse(new_bh); ++ ++ return error; ++} ++ ++/* ++ * ext3_xattr_delete_inode() ++ * ++ * Free extended attribute resources associated with this inode. This ++ * is called immediately before an inode is freed. ++ */ ++void ++ext3_xattr_delete_inode(handle_t *handle, struct inode *inode) ++{ ++ struct buffer_head *bh; ++ unsigned int block = EXT3_I(inode)->i_file_acl; ++ ++ if (!block) ++ return; ++ down(&ext3_xattr_sem); ++ ++ bh = sb_bread(inode->i_sb, block); ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_delete_inode", ++ "inode %ld: block %d read error", inode->i_ino, block); ++ goto cleanup; ++ } ++ ea_bdebug(bh, "b_count=%d", atomic_read(&(bh->b_count))); ++ if (HDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) || ++ HDR(bh)->h_blocks != cpu_to_le32(1)) { ++ ext3_error(inode->i_sb, "ext3_xattr_delete_inode", ++ "inode %ld: bad block %d", inode->i_ino, block); ++ goto cleanup; ++ } ++ ext3_journal_get_write_access(handle, bh); ++ ea_bdebug(bh, "refcount now=%d", le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ if (HDR(bh)->h_refcount == cpu_to_le32(1)) { ++ ext3_xattr_cache_remove(bh); ++ ext3_xattr_free_block(handle, inode, block); ++ ext3_forget(handle, 1, inode, bh, block); ++ bh = NULL; ++ } else { ++ HDR(bh)->h_refcount = cpu_to_le32( ++ le32_to_cpu(HDR(bh)->h_refcount) - 1); ++ ext3_journal_dirty_metadata(handle, bh); ++ if (IS_SYNC(inode)) ++ handle->h_sync = 1; ++ ext3_xattr_quota_free(inode); ++ } ++ EXT3_I(inode)->i_file_acl = 0; ++ ++cleanup: ++ brelse(bh); ++ up(&ext3_xattr_sem); ++} ++ ++/* ++ * ext3_xattr_put_super() ++ * ++ * This is called when a file system is unmounted. ++ */ ++void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ mb_cache_shrink(ext3_xattr_cache, sb->s_dev); ++#endif ++} ++ ++#ifdef CONFIG_EXT3_FS_XATTR_SHARING ++ ++/* ++ * ext3_xattr_cache_insert() ++ * ++ * Create a new entry in the extended attribute cache, and insert ++ * it unless such an entry is already in the cache. ++ * ++ * Returns 0, or a negative error number on failure. ++ */ ++static int ++ext3_xattr_cache_insert(struct buffer_head *bh) ++{ ++ __u32 hash = le32_to_cpu(HDR(bh)->h_hash); ++ struct mb_cache_entry *ce; ++ int error; ++ ++ ce = mb_cache_entry_alloc(ext3_xattr_cache); ++ if (!ce) ++ return -ENOMEM; ++ error = mb_cache_entry_insert(ce, bh->b_dev, bh->b_blocknr, &hash); ++ if (error) { ++ mb_cache_entry_free(ce); ++ if (error == -EBUSY) { ++ ea_bdebug(bh, "already in cache (%d cache entries)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ error = 0; ++ } ++ } else { ++ ea_bdebug(bh, "inserting [%x] (%d cache entries)", (int)hash, ++ atomic_read(&ext3_xattr_cache->c_entry_count)); ++ mb_cache_entry_release(ce); ++ } ++ return error; ++} ++ ++/* ++ * ext3_xattr_cmp() ++ * ++ * Compare two extended attribute blocks for equality. ++ * ++ * Returns 0 if the blocks are equal, 1 if they differ, and ++ * a negative error number on errors. ++ */ ++static int ++ext3_xattr_cmp(struct ext3_xattr_header *header1, ++ struct ext3_xattr_header *header2) ++{ ++ struct ext3_xattr_entry *entry1, *entry2; ++ ++ entry1 = ENTRY(header1+1); ++ entry2 = ENTRY(header2+1); ++ while (!IS_LAST_ENTRY(entry1)) { ++ if (IS_LAST_ENTRY(entry2)) ++ return 1; ++ if (entry1->e_hash != entry2->e_hash || ++ entry1->e_name_len != entry2->e_name_len || ++ entry1->e_value_size != entry2->e_value_size || ++ memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len)) ++ return 1; ++ if (entry1->e_value_block != 0 || entry2->e_value_block != 0) ++ return -EIO; ++ if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs), ++ (char *)header2 + le16_to_cpu(entry2->e_value_offs), ++ le32_to_cpu(entry1->e_value_size))) ++ return 1; ++ ++ entry1 = EXT3_XATTR_NEXT(entry1); ++ entry2 = EXT3_XATTR_NEXT(entry2); ++ } ++ if (!IS_LAST_ENTRY(entry2)) ++ return 1; ++ return 0; ++} ++ ++/* ++ * ext3_xattr_cache_find() ++ * ++ * Find an identical extended attribute block. ++ * ++ * Returns a pointer to the block found, or NULL if such a block was ++ * not found or an error occurred. ++ */ ++static struct buffer_head * ++ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header) ++{ ++ __u32 hash = le32_to_cpu(header->h_hash); ++ struct mb_cache_entry *ce; ++ ++ if (!header->h_hash) ++ return NULL; /* never share */ ++ ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); ++ ce = mb_cache_entry_find_first(ext3_xattr_cache, 0, inode->i_dev, hash); ++ while (ce) { ++ struct buffer_head *bh = sb_bread(inode->i_sb, ce->e_block); ++ ++ if (!bh) { ++ ext3_error(inode->i_sb, "ext3_xattr_cache_find", ++ "inode %ld: block %ld read error", ++ inode->i_ino, ce->e_block); ++ } else if (le32_to_cpu(HDR(bh)->h_refcount) > ++ EXT3_XATTR_REFCOUNT_MAX) { ++ ea_idebug(inode, "block %ld refcount %d>%d",ce->e_block, ++ le32_to_cpu(HDR(bh)->h_refcount), ++ EXT3_XATTR_REFCOUNT_MAX); ++ } else if (!ext3_xattr_cmp(header, HDR(bh))) { ++ ea_bdebug(bh, "b_count=%d",atomic_read(&(bh->b_count))); ++ mb_cache_entry_release(ce); ++ return bh; ++ } ++ brelse(bh); ++ ce = mb_cache_entry_find_next(ce, 0, inode->i_dev, hash); ++ } ++ return NULL; ++} ++ ++/* ++ * ext3_xattr_cache_remove() ++ * ++ * Remove the cache entry of a block from the cache. Called when a ++ * block becomes invalid. ++ */ ++static void ++ext3_xattr_cache_remove(struct buffer_head *bh) ++{ ++ struct mb_cache_entry *ce; ++ ++ ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_dev, bh->b_blocknr); ++ if (ce) { ++ ea_bdebug(bh, "removing (%d cache entries remaining)", ++ atomic_read(&ext3_xattr_cache->c_entry_count)-1); ++ mb_cache_entry_free(ce); ++ } else ++ ea_bdebug(bh, "no cache entry"); ++} ++ ++#define NAME_HASH_SHIFT 5 ++#define VALUE_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_hash_entry() ++ * ++ * Compute the hash of an extended attribute. ++ */ ++static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ __u32 hash = 0; ++ char *name = entry->e_name; ++ int n; ++ ++ for (n=0; n < entry->e_name_len; n++) { ++ hash = (hash << NAME_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^ ++ *name++; ++ } ++ ++ if (entry->e_value_block == 0 && entry->e_value_size != 0) { ++ __u32 *value = (__u32 *)((char *)header + ++ le16_to_cpu(entry->e_value_offs)); ++ for (n = (le32_to_cpu(entry->e_value_size) + ++ EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) { ++ hash = (hash << VALUE_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^ ++ le32_to_cpu(*value++); ++ } ++ } ++ entry->e_hash = cpu_to_le32(hash); ++} ++ ++#undef NAME_HASH_SHIFT ++#undef VALUE_HASH_SHIFT ++ ++#define BLOCK_HASH_SHIFT 16 ++ ++/* ++ * ext3_xattr_rehash() ++ * ++ * Re-compute the extended attribute hash value after an entry has changed. ++ */ ++static void ext3_xattr_rehash(struct ext3_xattr_header *header, ++ struct ext3_xattr_entry *entry) ++{ ++ struct ext3_xattr_entry *here; ++ __u32 hash = 0; ++ ++ ext3_xattr_hash_entry(header, entry); ++ here = ENTRY(header+1); ++ while (!IS_LAST_ENTRY(here)) { ++ if (!here->e_hash) { ++ /* Block is not shared if an entry's hash value == 0 */ ++ hash = 0; ++ break; ++ } ++ hash = (hash << BLOCK_HASH_SHIFT) ^ ++ (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^ ++ le32_to_cpu(here->e_hash); ++ here = EXT3_XATTR_NEXT(here); ++ } ++ header->h_hash = cpu_to_le32(hash); ++} ++ ++#undef BLOCK_HASH_SHIFT ++ ++int __init ++init_ext3_xattr(void) ++{ ++ ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL, ++ sizeof(struct mb_cache_entry) + ++ sizeof(struct mb_cache_entry_index), 1, 61); ++ if (!ext3_xattr_cache) ++ return -ENOMEM; ++ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++ if (ext3_xattr_cache) ++ mb_cache_destroy(ext3_xattr_cache); ++ ext3_xattr_cache = NULL; ++} ++ ++#else /* CONFIG_EXT3_FS_XATTR_SHARING */ ++ ++int __init ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++void ++exit_ext3_xattr(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_SHARING */ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/ext3/xattr_user.c Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,111 @@ ++/* ++ * linux/fs/ext3/xattr_user.c ++ * Handler for extended user attributes. ++ * ++ * Copyright (C) 2001 by Andreas Gruenbacher, ++ */ ++ ++#include ++#include ++#include ++#include ++#include ++#include ++ ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++# include ++#endif ++ ++#define XATTR_USER_PREFIX "user." ++ ++static size_t ++ext3_xattr_user_list(char *list, struct inode *inode, ++ const char *name, int name_len) ++{ ++ const int prefix_len = sizeof(XATTR_USER_PREFIX)-1; ++ ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return 0; ++ ++ if (list) { ++ memcpy(list, XATTR_USER_PREFIX, prefix_len); ++ memcpy(list+prefix_len, name, name_len); ++ list[prefix_len + name_len] = '\0'; ++ } ++ return prefix_len + name_len + 1; ++} ++ ++static int ++ext3_xattr_user_get(struct inode *inode, const char *name, ++ void *buffer, size_t size) ++{ ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++ error = ext3_permission_locked(inode, MAY_READ); ++#else ++ error = permission(inode, MAY_READ); ++#endif ++ if (error) ++ return error; ++ ++ return ext3_xattr_get(inode, EXT3_XATTR_INDEX_USER, name, ++ buffer, size); ++} ++ ++static int ++ext3_xattr_user_set(struct inode *inode, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ handle_t *handle; ++ int error; ++ ++ if (strcmp(name, "") == 0) ++ return -EINVAL; ++ if (!test_opt(inode->i_sb, XATTR_USER)) ++ return -ENOTSUP; ++ if ( !S_ISREG(inode->i_mode) && ++ (!S_ISDIR(inode->i_mode) || inode->i_mode & S_ISVTX)) ++ return -EPERM; ++#ifdef CONFIG_EXT3_FS_POSIX_ACL ++ error = ext3_permission_locked(inode, MAY_WRITE); ++#else ++ error = permission(inode, MAY_WRITE); ++#endif ++ if (error) ++ return error; ++ ++ handle = ext3_journal_start(inode, EXT3_XATTR_TRANS_BLOCKS); ++ if (IS_ERR(handle)) ++ return PTR_ERR(handle); ++ error = ext3_xattr_set(handle, inode, EXT3_XATTR_INDEX_USER, name, ++ value, size, flags); ++ ext3_journal_stop(handle, inode); ++ ++ return error; ++} ++ ++struct ext3_xattr_handler ext3_xattr_user_handler = { ++ prefix: XATTR_USER_PREFIX, ++ list: ext3_xattr_user_list, ++ get: ext3_xattr_user_get, ++ set: ext3_xattr_user_set, ++}; ++ ++int __init ++init_ext3_xattr_user(void) ++{ ++ return ext3_xattr_register(EXT3_XATTR_INDEX_USER, ++ &ext3_xattr_user_handler); ++} ++ ++void ++exit_ext3_xattr_user(void) ++{ ++ ext3_xattr_unregister(EXT3_XATTR_INDEX_USER, ++ &ext3_xattr_user_handler); ++} +--- linux-2.4.20/fs/jfs/jfs_xattr.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/jfs/jfs_xattr.h Sat Apr 5 03:57:18 2003 +@@ -52,8 +52,10 @@ struct jfs_ea_list { + #define END_EALIST(ealist) \ + ((struct jfs_ea *) (((char *) (ealist)) + EALIST_SIZE(ealist))) + +-extern int __jfs_setxattr(struct inode *, const char *, void *, size_t, int); +-extern int jfs_setxattr(struct dentry *, const char *, void *, size_t, int); ++extern int __jfs_setxattr(struct inode *, const char *, const void *, size_t, ++ int); ++extern int jfs_setxattr(struct dentry *, const char *, const void *, size_t, ++ int); + extern ssize_t __jfs_getxattr(struct inode *, const char *, void *, size_t); + extern ssize_t jfs_getxattr(struct dentry *, const char *, void *, size_t); + extern ssize_t jfs_listxattr(struct dentry *, char *, size_t); +--- linux-2.4.20/fs/jfs/xattr.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/fs/jfs/xattr.c Sat Apr 5 03:57:18 2003 +@@ -641,7 +641,7 @@ static int ea_put(struct inode *inode, s + } + + static int can_set_xattr(struct inode *inode, const char *name, +- void *value, size_t value_len) ++ const void *value, size_t value_len) + { + if (IS_RDONLY(inode)) + return -EROFS; +@@ -660,7 +660,7 @@ static int can_set_xattr(struct inode *i + return permission(inode, MAY_WRITE); + } + +-int __jfs_setxattr(struct inode *inode, const char *name, void *value, ++int __jfs_setxattr(struct inode *inode, const char *name, const void *value, + size_t value_len, int flags) + { + struct jfs_ea_list *ealist; +@@ -799,7 +799,7 @@ int __jfs_setxattr(struct inode *inode, + return rc; + } + +-int jfs_setxattr(struct dentry *dentry, const char *name, void *value, ++int jfs_setxattr(struct dentry *dentry, const char *name, const void *value, + size_t value_len, int flags) + { + if (value == NULL) { /* empty EA, do not remove */ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/fs/mbcache.c Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,648 @@ ++/* ++ * linux/fs/mbcache.c ++ * (C) 2001-2002 Andreas Gruenbacher, ++ */ ++ ++/* ++ * Filesystem Meta Information Block Cache (mbcache) ++ * ++ * The mbcache caches blocks of block devices that need to be located ++ * by their device/block number, as well as by other criteria (such ++ * as the block's contents). ++ * ++ * There can only be one cache entry in a cache per device and block number. ++ * Additional indexes need not be unique in this sense. The number of ++ * additional indexes (=other criteria) can be hardwired at compile time ++ * or specified at cache create time. ++ * ++ * Each cache entry is of fixed size. An entry may be `valid' or `invalid' ++ * in the cache. A valid entry is in the main hash tables of the cache, ++ * and may also be in the lru list. An invalid entry is not in any hashes ++ * or lists. ++ * ++ * A valid cache entry is only in the lru list if no handles refer to it. ++ * Invalid cache entries will be freed when the last handle to the cache ++ * entry is released. Entries that cannot be freed immediately are put ++ * back on the lru list. ++ */ ++ ++#include ++#include ++ ++#include ++#include ++#include ++#include ++#include ++#include ++#include ++ ++ ++#ifdef MB_CACHE_DEBUG ++# define mb_debug(f...) do { \ ++ printk(KERN_DEBUG f); \ ++ printk("\n"); \ ++ } while (0) ++#define mb_assert(c) do { if (!(c)) \ ++ printk(KERN_ERR "assertion " #c " failed\n"); \ ++ } while(0) ++#else ++# define mb_debug(f...) do { } while(0) ++# define mb_assert(c) do { } while(0) ++#endif ++#define mb_error(f...) do { \ ++ printk(KERN_ERR f); \ ++ printk("\n"); \ ++ } while(0) ++ ++MODULE_AUTHOR("Andreas Gruenbacher "); ++MODULE_DESCRIPTION("Meta block cache (for extended attributes)"); ++#if LINUX_VERSION_CODE >= KERNEL_VERSION(2,4,0) ++MODULE_LICENSE("GPL"); ++#endif ++ ++EXPORT_SYMBOL(mb_cache_create); ++EXPORT_SYMBOL(mb_cache_shrink); ++EXPORT_SYMBOL(mb_cache_destroy); ++EXPORT_SYMBOL(mb_cache_entry_alloc); ++EXPORT_SYMBOL(mb_cache_entry_insert); ++EXPORT_SYMBOL(mb_cache_entry_release); ++EXPORT_SYMBOL(mb_cache_entry_takeout); ++EXPORT_SYMBOL(mb_cache_entry_free); ++EXPORT_SYMBOL(mb_cache_entry_dup); ++EXPORT_SYMBOL(mb_cache_entry_get); ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++EXPORT_SYMBOL(mb_cache_entry_find_first); ++EXPORT_SYMBOL(mb_cache_entry_find_next); ++#endif ++ ++ ++/* ++ * Global data: list of all mbcache's, lru list, and a spinlock for ++ * accessing cache data structures on SMP machines. The lru list is ++ * global across all mbcaches. ++ */ ++ ++static LIST_HEAD(mb_cache_list); ++static LIST_HEAD(mb_cache_lru_list); ++static spinlock_t mb_cache_spinlock = SPIN_LOCK_UNLOCKED; ++ ++static inline int ++mb_cache_indexes(struct mb_cache *cache) ++{ ++#ifdef MB_CACHE_INDEXES_COUNT ++ return MB_CACHE_INDEXES_COUNT; ++#else ++ return cache->c_indexes_count; ++#endif ++} ++ ++/* ++ * What the mbcache registers as to get shrunk dynamically. ++ */ ++ ++static void ++mb_cache_memory_pressure(int priority, unsigned int gfp_mask); ++ ++static struct cache_definition mb_cache_definition = { ++ "mb_cache", ++ mb_cache_memory_pressure ++}; ++ ++ ++static inline int ++__mb_cache_entry_is_hashed(struct mb_cache_entry *ce) ++{ ++ return !list_empty(&ce->e_block_list); ++} ++ ++ ++static inline void ++__mb_cache_entry_unhash(struct mb_cache_entry *ce) ++{ ++ int n; ++ ++ if (__mb_cache_entry_is_hashed(ce)) { ++ list_del_init(&ce->e_block_list); ++ for (n=0; ne_cache); n++) ++ list_del(&ce->e_indexes[n].o_list); ++ } ++} ++ ++ ++static inline void ++__mb_cache_entry_forget(struct mb_cache_entry *ce, int gfp_mask) ++{ ++ struct mb_cache *cache = ce->e_cache; ++ ++ mb_assert(atomic_read(&ce->e_used) == 0); ++ if (cache->c_op.free && cache->c_op.free(ce, gfp_mask)) { ++ /* free failed -- put back on the lru list ++ for freeing later. */ ++ spin_lock(&mb_cache_spinlock); ++ list_add(&ce->e_lru_list, &mb_cache_lru_list); ++ spin_unlock(&mb_cache_spinlock); ++ } else { ++ kmem_cache_free(cache->c_entry_cache, ce); ++ atomic_dec(&cache->c_entry_count); ++ } ++} ++ ++ ++static inline void ++__mb_cache_entry_release_unlock(struct mb_cache_entry *ce) ++{ ++ if (atomic_dec_and_test(&ce->e_used)) { ++ if (__mb_cache_entry_is_hashed(ce)) ++ list_add_tail(&ce->e_lru_list, &mb_cache_lru_list); ++ else { ++ spin_unlock(&mb_cache_spinlock); ++ __mb_cache_entry_forget(ce, GFP_KERNEL); ++ return; ++ } ++ } ++ spin_unlock(&mb_cache_spinlock); ++} ++ ++ ++/* ++ * mb_cache_memory_pressure() memory pressure callback ++ * ++ * This function is called by the kernel memory management when memory ++ * gets low. ++ * ++ * @priority: Amount by which to shrink the cache (0 = highes priority) ++ * @gfp_mask: (ignored) ++ */ ++static void ++mb_cache_memory_pressure(int priority, unsigned int gfp_mask) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ int count = 0; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &mb_cache_list) { ++ struct mb_cache *cache = ++ list_entry(l, struct mb_cache, c_cache_list); ++ mb_debug("cache %s (%d)", cache->c_name, ++ atomic_read(&cache->c_entry_count)); ++ count += atomic_read(&cache->c_entry_count); ++ } ++ mb_debug("trying to free %d of %d entries", ++ count / (priority ? priority : 1), count); ++ if (priority) ++ count /= priority; ++ while (count-- && !list_empty(&mb_cache_lru_list)) { ++ struct mb_cache_entry *ce = ++ list_entry(mb_cache_lru_list.next, ++ struct mb_cache_entry, e_lru_list); ++ list_del(&ce->e_lru_list); ++ __mb_cache_entry_unhash(ce); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ } ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), gfp_mask); ++ } ++} ++ ++ ++/* ++ * mb_cache_create() create a new cache ++ * ++ * All entries in one cache are equal size. Cache entries may be from ++ * multiple devices. If this is the first mbcache created, registers ++ * the cache with kernel memory management. Returns NULL if no more ++ * memory was available. ++ * ++ * @name: name of the cache (informal) ++ * @cache_op: contains the callback called when freeing a cache entry ++ * @entry_size: The size of a cache entry, including ++ * struct mb_cache_entry ++ * @indexes_count: number of additional indexes in the cache. Must equal ++ * MB_CACHE_INDEXES_COUNT if the number of indexes is ++ * hardwired. ++ * @bucket_count: number of hash buckets ++ */ ++struct mb_cache * ++mb_cache_create(const char *name, struct mb_cache_op *cache_op, ++ size_t entry_size, int indexes_count, int bucket_count) ++{ ++ int m=0, n; ++ struct mb_cache *cache = NULL; ++ ++ if(entry_size < sizeof(struct mb_cache_entry) + ++ indexes_count * sizeof(struct mb_cache_entry_index)) ++ return NULL; ++ ++ MOD_INC_USE_COUNT; ++ cache = kmalloc(sizeof(struct mb_cache) + ++ indexes_count * sizeof(struct list_head), GFP_KERNEL); ++ if (!cache) ++ goto fail; ++ cache->c_name = name; ++ cache->c_op.free = NULL; ++ if (cache_op) ++ cache->c_op.free = cache_op->free; ++ atomic_set(&cache->c_entry_count, 0); ++ cache->c_bucket_count = bucket_count; ++#ifdef MB_CACHE_INDEXES_COUNT ++ mb_assert(indexes_count == MB_CACHE_INDEXES_COUNT); ++#else ++ cache->c_indexes_count = indexes_count; ++#endif ++ cache->c_block_hash = kmalloc(bucket_count * sizeof(struct list_head), ++ GFP_KERNEL); ++ if (!cache->c_block_hash) ++ goto fail; ++ for (n=0; nc_block_hash[n]); ++ for (m=0; mc_indexes_hash[m] = kmalloc(bucket_count * ++ sizeof(struct list_head), ++ GFP_KERNEL); ++ if (!cache->c_indexes_hash[m]) ++ goto fail; ++ for (n=0; nc_indexes_hash[m][n]); ++ } ++ cache->c_entry_cache = kmem_cache_create(name, entry_size, 0, ++ 0 /*SLAB_POISON | SLAB_RED_ZONE*/, NULL, NULL); ++ if (!cache->c_entry_cache) ++ goto fail; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_add(&cache->c_cache_list, &mb_cache_list); ++ spin_unlock(&mb_cache_spinlock); ++ return cache; ++ ++fail: ++ if (cache) { ++ while (--m >= 0) ++ kfree(cache->c_indexes_hash[m]); ++ if (cache->c_block_hash) ++ kfree(cache->c_block_hash); ++ kfree(cache); ++ } ++ MOD_DEC_USE_COUNT; ++ return NULL; ++} ++ ++ ++/* ++ * mb_cache_shrink() ++ * ++ * Removes all cache entires of a device from the cache. All cache entries ++ * currently in use cannot be freed, and thus remain in the cache. ++ * ++ * @cache: which cache to shrink ++ * @dev: which device's cache entries to shrink ++ */ ++void ++mb_cache_shrink(struct mb_cache *cache, kdev_t dev) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &mb_cache_lru_list) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_lru_list); ++ if (ce->e_dev == dev) { ++ list_del(&ce->e_lru_list); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ __mb_cache_entry_unhash(ce); ++ } ++ } ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), GFP_KERNEL); ++ } ++} ++ ++ ++/* ++ * mb_cache_destroy() ++ * ++ * Shrinks the cache to its minimum possible size (hopefully 0 entries), ++ * and then destroys it. If this was the last mbcache, un-registers the ++ * mbcache from kernel memory management. ++ */ ++void ++mb_cache_destroy(struct mb_cache *cache) ++{ ++ LIST_HEAD(free_list); ++ struct list_head *l, *ltmp; ++ int n; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &mb_cache_lru_list) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_lru_list); ++ if (ce->e_cache == cache) { ++ list_del(&ce->e_lru_list); ++ list_add_tail(&ce->e_lru_list, &free_list); ++ __mb_cache_entry_unhash(ce); ++ } ++ } ++ list_del(&cache->c_cache_list); ++ spin_unlock(&mb_cache_spinlock); ++ list_for_each_safe(l, ltmp, &free_list) { ++ __mb_cache_entry_forget(list_entry(l, struct mb_cache_entry, ++ e_lru_list), GFP_KERNEL); ++ } ++ ++ if (atomic_read(&cache->c_entry_count) > 0) { ++ mb_error("cache %s: %d orphaned entries", ++ cache->c_name, ++ atomic_read(&cache->c_entry_count)); ++ } ++ ++#if (LINUX_VERSION_CODE < KERNEL_VERSION(2,3,0)) ++ /* We don't have kmem_cache_destroy() in 2.2.x */ ++ kmem_cache_shrink(cache->c_entry_cache); ++#else ++ kmem_cache_destroy(cache->c_entry_cache); ++#endif ++ for (n=0; n < mb_cache_indexes(cache); n++) ++ kfree(cache->c_indexes_hash[n]); ++ kfree(cache->c_block_hash); ++ kfree(cache); ++ ++ MOD_DEC_USE_COUNT; ++} ++ ++ ++/* ++ * mb_cache_entry_alloc() ++ * ++ * Allocates a new cache entry. The new entry will not be valid initially, ++ * and thus cannot be looked up yet. It should be filled with data, and ++ * then inserted into the cache using mb_cache_entry_insert(). Returns NULL ++ * if no more memory was available. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_alloc(struct mb_cache *cache) ++{ ++ struct mb_cache_entry *ce; ++ ++ atomic_inc(&cache->c_entry_count); ++ ce = kmem_cache_alloc(cache->c_entry_cache, GFP_KERNEL); ++ if (ce) { ++ INIT_LIST_HEAD(&ce->e_lru_list); ++ INIT_LIST_HEAD(&ce->e_block_list); ++ ce->e_cache = cache; ++ atomic_set(&ce->e_used, 1); ++ } ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_insert() ++ * ++ * Inserts an entry that was allocated using mb_cache_entry_alloc() into ++ * the cache. After this, the cache entry can be looked up, but is not yet ++ * in the lru list as the caller still holds a handle to it. Returns 0 on ++ * success, or -EBUSY if a cache entry for that device + inode exists ++ * already (this may happen after a failed lookup, if another process has ++ * inserted the same cache entry in the meantime). ++ * ++ * @dev: device the cache entry belongs to ++ * @block: block number ++ * @keys: array of additional keys. There must be indexes_count entries ++ * in the array (as specified when creating the cache). ++ */ ++int ++mb_cache_entry_insert(struct mb_cache_entry *ce, kdev_t dev, ++ unsigned long block, unsigned int keys[]) ++{ ++ struct mb_cache *cache = ce->e_cache; ++ unsigned int bucket = (HASHDEV(dev) + block) % cache->c_bucket_count; ++ struct list_head *l; ++ int error = -EBUSY, n; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &cache->c_block_hash[bucket]) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, e_block_list); ++ if (ce->e_dev == dev && ce->e_block == block) ++ goto out; ++ } ++ __mb_cache_entry_unhash(ce); ++ ce->e_dev = dev; ++ ce->e_block = block; ++ list_add(&ce->e_block_list, &cache->c_block_hash[bucket]); ++ for (n=0; ne_indexes[n].o_key = keys[n]; ++ bucket = keys[n] % cache->c_bucket_count; ++ list_add(&ce->e_indexes[n].o_list, ++ &cache->c_indexes_hash[n][bucket]); ++ } ++out: ++ spin_unlock(&mb_cache_spinlock); ++ return error; ++} ++ ++ ++/* ++ * mb_cache_entry_release() ++ * ++ * Release a handle to a cache entry. When the last handle to a cache entry ++ * is released it is either freed (if it is invalid) or otherwise inserted ++ * in to the lru list. ++ */ ++void ++mb_cache_entry_release(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ __mb_cache_entry_release_unlock(ce); ++} ++ ++ ++/* ++ * mb_cache_entry_takeout() ++ * ++ * Take a cache entry out of the cache, making it invalid. The entry can later ++ * be re-inserted using mb_cache_entry_insert(), or released using ++ * mb_cache_entry_release(). ++ */ ++void ++mb_cache_entry_takeout(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ mb_assert(list_empty(&ce->e_lru_list)); ++ __mb_cache_entry_unhash(ce); ++ spin_unlock(&mb_cache_spinlock); ++} ++ ++ ++/* ++ * mb_cache_entry_free() ++ * ++ * This is equivalent to the sequence mb_cache_entry_takeout() -- ++ * mb_cache_entry_release(). ++ */ ++void ++mb_cache_entry_free(struct mb_cache_entry *ce) ++{ ++ spin_lock(&mb_cache_spinlock); ++ mb_assert(list_empty(&ce->e_lru_list)); ++ __mb_cache_entry_unhash(ce); ++ __mb_cache_entry_release_unlock(ce); ++} ++ ++ ++/* ++ * mb_cache_entry_dup() ++ * ++ * Duplicate a handle to a cache entry (does not duplicate the cache entry ++ * itself). After the call, both the old and the new handle must be released. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_dup(struct mb_cache_entry *ce) ++{ ++ atomic_inc(&ce->e_used); ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_get() ++ * ++ * Get a cache entry by device / block number. (There can only be one entry ++ * in the cache per device and block.) Returns NULL if no such cache entry ++ * exists. ++ */ ++struct mb_cache_entry * ++mb_cache_entry_get(struct mb_cache *cache, kdev_t dev, unsigned long block) ++{ ++ unsigned int bucket = (HASHDEV(dev) + block) % cache->c_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ spin_lock(&mb_cache_spinlock); ++ list_for_each(l, &cache->c_block_hash[bucket]) { ++ ce = list_entry(l, struct mb_cache_entry, e_block_list); ++ if (ce->e_dev == dev && ce->e_block == block) { ++ if (!list_empty(&ce->e_lru_list)) ++ list_del_init(&ce->e_lru_list); ++ atomic_inc(&ce->e_used); ++ goto cleanup; ++ } ++ } ++ ce = NULL; ++ ++cleanup: ++ spin_unlock(&mb_cache_spinlock); ++ return ce; ++} ++ ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++ ++static struct mb_cache_entry * ++__mb_cache_entry_find(struct list_head *l, struct list_head *head, ++ int index, kdev_t dev, unsigned int key) ++{ ++ while (l != head) { ++ struct mb_cache_entry *ce = ++ list_entry(l, struct mb_cache_entry, ++ e_indexes[index].o_list); ++ if (ce->e_dev == dev && ce->e_indexes[index].o_key == key) { ++ if (!list_empty(&ce->e_lru_list)) ++ list_del_init(&ce->e_lru_list); ++ atomic_inc(&ce->e_used); ++ return ce; ++ } ++ l = l->next; ++ } ++ return NULL; ++} ++ ++ ++/* ++ * mb_cache_entry_find_first() ++ * ++ * Find the first cache entry on a given device with a certain key in ++ * an additional index. Additonal matches can be found with ++ * mb_cache_entry_find_next(). Returns NULL if no match was found. ++ * ++ * @cache: the cache to search ++ * @index: the number of the additonal index to search (0<=indexc_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ mb_assert(index < mb_cache_indexes(cache)); ++ spin_lock(&mb_cache_spinlock); ++ l = cache->c_indexes_hash[index][bucket].next; ++ ce = __mb_cache_entry_find(l, &cache->c_indexes_hash[index][bucket], ++ index, dev, key); ++ spin_unlock(&mb_cache_spinlock); ++ return ce; ++} ++ ++ ++/* ++ * mb_cache_entry_find_next() ++ * ++ * Find the next cache entry on a given device with a certain key in an ++ * additional index. Returns NULL if no match could be found. The previous ++ * entry is atomatically released, so that mb_cache_entry_find_next() can ++ * be called like this: ++ * ++ * entry = mb_cache_entry_find_first(); ++ * while (entry) { ++ * ... ++ * entry = mb_cache_entry_find_next(entry, ...); ++ * } ++ * ++ * @prev: The previous match ++ * @index: the number of the additonal index to search (0<=indexe_cache; ++ unsigned int bucket = key % cache->c_bucket_count; ++ struct list_head *l; ++ struct mb_cache_entry *ce; ++ ++ mb_assert(index < mb_cache_indexes(cache)); ++ spin_lock(&mb_cache_spinlock); ++ l = prev->e_indexes[index].o_list.next; ++ ce = __mb_cache_entry_find(l, &cache->c_indexes_hash[index][bucket], ++ index, dev, key); ++ __mb_cache_entry_release_unlock(prev); ++ return ce; ++} ++ ++#endif /* !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) */ ++ ++static int __init init_mbcache(void) ++{ ++ register_cache(&mb_cache_definition); ++ return 0; ++} ++ ++static void __exit exit_mbcache(void) ++{ ++ unregister_cache(&mb_cache_definition); ++} ++ ++module_init(init_mbcache) ++module_exit(exit_mbcache) ++ +--- linux-2.4.20/include/asm-arm/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-arm/unistd.h Sat Apr 5 03:57:18 2003 +@@ -244,7 +244,6 @@ + #define __NR_security (__NR_SYSCALL_BASE+223) + #define __NR_gettid (__NR_SYSCALL_BASE+224) + #define __NR_readahead (__NR_SYSCALL_BASE+225) +-#if 0 /* allocated in 2.5 */ + #define __NR_setxattr (__NR_SYSCALL_BASE+226) + #define __NR_lsetxattr (__NR_SYSCALL_BASE+227) + #define __NR_fsetxattr (__NR_SYSCALL_BASE+228) +@@ -257,7 +256,6 @@ + #define __NR_removexattr (__NR_SYSCALL_BASE+235) + #define __NR_lremovexattr (__NR_SYSCALL_BASE+236) + #define __NR_fremovexattr (__NR_SYSCALL_BASE+237) +-#endif + #define __NR_tkill (__NR_SYSCALL_BASE+238) + /* + * Please check 2.5 _before_ adding calls here, +--- linux-2.4.20/include/asm-ia64/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-ia64/unistd.h Sat Apr 5 03:57:18 2003 +@@ -206,8 +206,19 @@ + #define __NR_getdents64 1214 + #define __NR_getunwind 1215 + #define __NR_readahead 1216 ++#define __NR_setxattr 1217 ++#define __NR_lsetxattr 1218 ++#define __NR_fsetxattr 1219 ++#define __NR_getxattr 1220 ++#define __NR_lgetxattr 1221 ++#define __NR_fgetxattr 1222 ++#define __NR_listxattr 1223 ++#define __NR_llistxattr 1224 ++#define __NR_flistxattr 1225 ++#define __NR_removexattr 1226 ++#define __NR_lremovexattr 1227 ++#define __NR_fremovexattr 1228 + /* +- * 1217-1228: reserved for xattr + * 1230-1232: reserved for futex and sched_[sg]etaffinity. + */ + #define __NR_tkill 1229 +--- linux-2.4.20/include/asm-ppc64/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-ppc64/unistd.h Sat Apr 5 03:57:18 2003 +@@ -218,6 +218,7 @@ + #define __NR_gettid 207 + #if 0 /* Reserved syscalls */ + #define __NR_tkill 208 ++#endif + #define __NR_setxattr 209 + #define __NR_lsetxattr 210 + #define __NR_fsetxattr 211 +@@ -230,6 +231,7 @@ + #define __NR_removexattr 218 + #define __NR_lremovexattr 219 + #define __NR_fremovexattr 220 ++#if 0 /* Reserved syscalls */ + #define __NR_futex 221 + #endif + +--- linux-2.4.20/include/asm-s390/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-s390/unistd.h Sat Apr 5 03:57:18 2003 +@@ -212,9 +212,18 @@ + #define __NR_madvise 219 + #define __NR_getdents64 220 + #define __NR_fcntl64 221 +-/* +- * Numbers 224-235 are reserved for posix acl +- */ ++#define __NR_setxattr 224 ++#define __NR_lsetxattr 225 ++#define __NR_fsetxattr 226 ++#define __NR_getxattr 227 ++#define __NR_lgetxattr 228 ++#define __NR_fgetxattr 229 ++#define __NR_listxattr 230 ++#define __NR_llistxattr 231 ++#define __NR_flistxattr 232 ++#define __NR_removexattr 233 ++#define __NR_lremovexattr 234 ++#define __NR_fremovexattr 235 + #define __NR_gettid 236 + #define __NR_tkill 237 + +--- linux-2.4.20/include/asm-s390x/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-s390x/unistd.h Sat Apr 5 03:57:18 2003 +@@ -180,9 +180,18 @@ + #define __NR_pivot_root 217 + #define __NR_mincore 218 + #define __NR_madvise 219 +-/* +- * Numbers 224-235 are reserved for posix acl +- */ ++#define __NR_setxattr 224 ++#define __NR_lsetxattr 225 ++#define __NR_fsetxattr 226 ++#define __NR_getxattr 227 ++#define __NR_lgetxattr 228 ++#define __NR_fgetxattr 229 ++#define __NR_listxattr 230 ++#define __NR_llistxattr 231 ++#define __NR_flistxattr 232 ++#define __NR_removexattr 233 ++#define __NR_lremovexattr 234 ++#define __NR_fremovexattr 235 + #define __NR_gettid 236 + #define __NR_tkill 237 + +--- linux-2.4.20/include/asm-sparc/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-sparc/unistd.h Sat Apr 5 03:57:18 2003 +@@ -1,4 +1,4 @@ +-/* $Id: linux-2.4.20-xattr-0.8.54.patch,v 1.1.2.1 2003/04/05 11:14:59 braam Exp $ */ ++/* $Id: linux-2.4.20-xattr-0.8.54.patch,v 1.1.2.1 2003/04/05 11:14:59 braam Exp $ */ + #ifndef _SPARC_UNISTD_H + #define _SPARC_UNISTD_H + +@@ -184,24 +184,24 @@ + /* #define __NR_exportfs 166 SunOS Specific */ + #define __NR_mount 167 /* Common */ + #define __NR_ustat 168 /* Common */ +-/* #define __NR_semsys 169 SunOS Specific */ +-/* #define __NR_msgsys 170 SunOS Specific */ +-/* #define __NR_shmsys 171 SunOS Specific */ +-/* #define __NR_auditsys 172 SunOS Specific */ +-/* #define __NR_rfssys 173 SunOS Specific */ ++#define __NR_setxattr 169 /* SunOS: semsys */ ++#define __NR_lsetxattr 170 /* SunOS: msgsys */ ++#define __NR_fsetxattr 171 /* SunOS: shmsys */ ++#define __NR_getxattr 172 /* SunOS: auditsys */ ++#define __NR_lgetxattr 173 /* SunOS: rfssys */ + #define __NR_getdents 174 /* Common */ + #define __NR_setsid 175 /* Common */ + #define __NR_fchdir 176 /* Common */ +-/* #define __NR_fchroot 177 SunOS Specific */ +-/* #define __NR_vpixsys 178 SunOS Specific */ +-/* #define __NR_aioread 179 SunOS Specific */ +-/* #define __NR_aiowrite 180 SunOS Specific */ +-/* #define __NR_aiowait 181 SunOS Specific */ +-/* #define __NR_aiocancel 182 SunOS Specific */ ++#define __NR_fgetxattr 177 /* SunOS: fchroot */ ++#define __NR_listxattr 178 /* SunOS: vpixsys */ ++#define __NR_llistxattr 179 /* SunOS: aioread */ ++#define __NR_flistxattr 180 /* SunOS: aiowrite */ ++#define __NR_removexattr 181 /* SunOS: aiowait */ ++#define __NR_lremovexattr 182 /* SunOS: aiocancel */ + #define __NR_sigpending 183 /* Common */ + #define __NR_query_module 184 /* Linux Specific */ + #define __NR_setpgid 185 /* Common */ +-/* #define __NR_pathconf 186 SunOS Specific */ ++#define __NR_fremovexattr 186 /* SunOS: pathconf */ + #define __NR_tkill 187 /* SunOS: fpathconf */ + /* #define __NR_sysconf 188 SunOS Specific */ + #define __NR_uname 189 /* Linux Specific */ +--- linux-2.4.20/include/asm-sparc64/unistd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/asm-sparc64/unistd.h Sat Apr 5 03:57:18 2003 +@@ -184,24 +184,24 @@ + /* #define __NR_exportfs 166 SunOS Specific */ + #define __NR_mount 167 /* Common */ + #define __NR_ustat 168 /* Common */ +-/* #define __NR_semsys 169 SunOS Specific */ +-/* #define __NR_msgsys 170 SunOS Specific */ +-/* #define __NR_shmsys 171 SunOS Specific */ +-/* #define __NR_auditsys 172 SunOS Specific */ +-/* #define __NR_rfssys 173 SunOS Specific */ ++#define __NR_setxattr 169 /* SunOS: semsys */ ++#define __NR_lsetxattr 170 /* SunOS: msgsys */ ++#define __NR_fsetxattr 171 /* SunOS: shmsys */ ++#define __NR_getxattr 172 /* SunOS: auditsys */ ++#define __NR_lgetxattr 173 /* SunOS: rfssys */ + #define __NR_getdents 174 /* Common */ + #define __NR_setsid 175 /* Common */ + #define __NR_fchdir 176 /* Common */ +-/* #define __NR_fchroot 177 SunOS Specific */ +-/* #define __NR_vpixsys 178 SunOS Specific */ +-/* #define __NR_aioread 179 SunOS Specific */ +-/* #define __NR_aiowrite 180 SunOS Specific */ +-/* #define __NR_aiowait 181 SunOS Specific */ +-/* #define __NR_aiocancel 182 SunOS Specific */ ++#define __NR_fgetxattr 177 /* SunOS: fchroot */ ++#define __NR_listxattr 178 /* SunOS: vpixsys */ ++#define __NR_llistxattr 179 /* SunOS: aioread */ ++#define __NR_flistxattr 180 /* SunOS: aiowrite */ ++#define __NR_removexattr 181 /* SunOS: aiowait */ ++#define __NR_lremovexattr 182 /* SunOS: aiocancel */ + #define __NR_sigpending 183 /* Common */ + #define __NR_query_module 184 /* Linux Specific */ + #define __NR_setpgid 185 /* Common */ +-/* #define __NR_pathconf 186 SunOS Specific */ ++#define __NR_fremovexattr 186 /* SunOS: pathconf */ + #define __NR_tkill 187 /* SunOS: fpathconf */ + /* #define __NR_sysconf 188 SunOS Specific */ + #define __NR_uname 189 /* Linux Specific */ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/include/linux/cache_def.h Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,15 @@ ++/* ++ * linux/cache_def.h ++ * Handling of caches defined in drivers, filesystems, ... ++ * ++ * Copyright (C) 2002 by Andreas Gruenbacher, ++ */ ++ ++struct cache_definition { ++ const char *name; ++ void (*shrink)(int, unsigned int); ++ struct list_head link; ++}; ++ ++extern void register_cache(struct cache_definition *); ++extern void unregister_cache(struct cache_definition *); +--- linux-2.4.20/include/linux/errno.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/linux/errno.h Sat Apr 5 03:57:18 2003 +@@ -23,4 +23,8 @@ + + #endif + ++/* Defined for extended attributes */ ++#define ENOATTR ENODATA /* No such attribute */ ++#define ENOTSUP EOPNOTSUPP /* Operation not supported */ ++ + #endif +--- linux-2.4.20/include/linux/ext2_fs.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/linux/ext2_fs.h Sat Apr 5 03:57:18 2003 +@@ -57,8 +57,6 @@ + */ + #define EXT2_BAD_INO 1 /* Bad blocks inode */ + #define EXT2_ROOT_INO 2 /* Root inode */ +-#define EXT2_ACL_IDX_INO 3 /* ACL inode */ +-#define EXT2_ACL_DATA_INO 4 /* ACL inode */ + #define EXT2_BOOT_LOADER_INO 5 /* Boot loader inode */ + #define EXT2_UNDEL_DIR_INO 6 /* Undelete directory inode */ + +@@ -86,7 +84,6 @@ + #else + # define EXT2_BLOCK_SIZE(s) (EXT2_MIN_BLOCK_SIZE << (s)->s_log_block_size) + #endif +-#define EXT2_ACLE_PER_BLOCK(s) (EXT2_BLOCK_SIZE(s) / sizeof (struct ext2_acl_entry)) + #define EXT2_ADDR_PER_BLOCK(s) (EXT2_BLOCK_SIZE(s) / sizeof (__u32)) + #ifdef __KERNEL__ + # define EXT2_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -121,28 +118,6 @@ + #endif + + /* +- * ACL structures +- */ +-struct ext2_acl_header /* Header of Access Control Lists */ +-{ +- __u32 aclh_size; +- __u32 aclh_file_count; +- __u32 aclh_acle_count; +- __u32 aclh_first_acle; +-}; +- +-struct ext2_acl_entry /* Access Control List Entry */ +-{ +- __u32 acle_size; +- __u16 acle_perms; /* Access permissions */ +- __u16 acle_type; /* Type of entry */ +- __u16 acle_tag; /* User or group identity */ +- __u16 acle_pad1; +- __u32 acle_next; /* Pointer on next entry for the */ +- /* same inode or on next free entry */ +-}; +- +-/* + * Structure of a blocks group descriptor + */ + struct ext2_group_desc +@@ -314,6 +289,7 @@ struct ext2_inode { + #define EXT2_MOUNT_ERRORS_PANIC 0x0040 /* Panic on errors */ + #define EXT2_MOUNT_MINIX_DF 0x0080 /* Mimics the Minix statfs */ + #define EXT2_MOUNT_NO_UID32 0x0200 /* Disable 32-bit UIDs */ ++#define EXT2_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + + #define clear_opt(o, opt) o &= ~EXT2_MOUNT_##opt + #define set_opt(o, opt) o |= EXT2_MOUNT_##opt +@@ -397,6 +373,7 @@ struct ext2_super_block { + + #ifdef __KERNEL__ + #define EXT2_SB(sb) (&((sb)->u.ext2_sb)) ++#define EXT2_I(inode) (&((inode)->u.ext2_i)) + #else + /* Assume that user mode programs are passing in an ext2fs superblock, not + * a kernel struct super_block. This will allow us to call the feature-test +@@ -466,7 +443,7 @@ struct ext2_super_block { + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 + #define EXT2_FEATURE_INCOMPAT_ANY 0xffffffff + +-#define EXT2_FEATURE_COMPAT_SUPP 0 ++#define EXT2_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR + #define EXT2_FEATURE_INCOMPAT_SUPP EXT2_FEATURE_INCOMPAT_FILETYPE + #define EXT2_FEATURE_RO_COMPAT_SUPP (EXT2_FEATURE_RO_COMPAT_SPARSE_SUPER| \ + EXT2_FEATURE_RO_COMPAT_LARGE_FILE| \ +@@ -623,8 +600,10 @@ extern struct address_space_operations e + + /* namei.c */ + extern struct inode_operations ext2_dir_inode_operations; ++extern struct inode_operations ext2_special_inode_operations; + + /* symlink.c */ ++extern struct inode_operations ext2_symlink_inode_operations; + extern struct inode_operations ext2_fast_symlink_inode_operations; + + #endif /* __KERNEL__ */ +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/include/linux/ext2_xattr.h Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,157 @@ ++/* ++ File: linux/ext2_xattr.h ++ ++ On-disk format of extended attributes for the ext2 filesystem. ++ ++ (C) 2001 Andreas Gruenbacher, ++*/ ++ ++#include ++#include ++#include ++ ++/* Magic value in attribute blocks */ ++#define EXT2_XATTR_MAGIC 0xEA020000 ++ ++/* Maximum number of references to one attribute block */ ++#define EXT2_XATTR_REFCOUNT_MAX 1024 ++ ++/* Name indexes */ ++#define EXT2_XATTR_INDEX_MAX 10 ++#define EXT2_XATTR_INDEX_USER 1 ++#define EXT2_XATTR_INDEX_POSIX_ACL_ACCESS 2 ++#define EXT2_XATTR_INDEX_POSIX_ACL_DEFAULT 3 ++ ++struct ext2_xattr_header { ++ __u32 h_magic; /* magic number for identification */ ++ __u32 h_refcount; /* reference count */ ++ __u32 h_blocks; /* number of disk blocks used */ ++ __u32 h_hash; /* hash value of all attributes */ ++ __u32 h_reserved[4]; /* zero right now */ ++}; ++ ++struct ext2_xattr_entry { ++ __u8 e_name_len; /* length of name */ ++ __u8 e_name_index; /* attribute name index */ ++ __u16 e_value_offs; /* offset in disk block of value */ ++ __u32 e_value_block; /* disk block attribute is stored on (n/i) */ ++ __u32 e_value_size; /* size of attribute value */ ++ __u32 e_hash; /* hash value of name and value */ ++ char e_name[0]; /* attribute name */ ++}; ++ ++#define EXT2_XATTR_PAD_BITS 2 ++#define EXT2_XATTR_PAD (1<e_name_len)) ) ++#define EXT2_XATTR_SIZE(size) \ ++ (((size) + EXT2_XATTR_ROUND) & ~EXT2_XATTR_ROUND) ++ ++#ifdef __KERNEL__ ++ ++# ifdef CONFIG_EXT2_FS_XATTR ++ ++struct ext2_xattr_handler { ++ char *prefix; ++ size_t (*list)(char *list, struct inode *inode, const char *name, ++ int name_len); ++ int (*get)(struct inode *inode, const char *name, void *buffer, ++ size_t size); ++ int (*set)(struct inode *inode, const char *name, const void *buffer, ++ size_t size, int flags); ++}; ++ ++extern int ext2_xattr_register(int, struct ext2_xattr_handler *); ++extern void ext2_xattr_unregister(int, struct ext2_xattr_handler *); ++ ++extern int ext2_setxattr(struct dentry *, const char *, const void *, size_t, int); ++extern ssize_t ext2_getxattr(struct dentry *, const char *, void *, size_t); ++extern ssize_t ext2_listxattr(struct dentry *, char *, size_t); ++extern int ext2_removexattr(struct dentry *, const char *); ++ ++extern int ext2_xattr_get(struct inode *, int, const char *, void *, size_t); ++extern int ext2_xattr_list(struct inode *, char *, size_t); ++extern int ext2_xattr_set(struct inode *, int, const char *, const void *, size_t, int); ++ ++extern void ext2_xattr_delete_inode(struct inode *); ++extern void ext2_xattr_put_super(struct super_block *); ++ ++extern int init_ext2_xattr(void) __init; ++extern void exit_ext2_xattr(void); ++ ++# else /* CONFIG_EXT2_FS_XATTR */ ++# define ext2_setxattr NULL ++# define ext2_getxattr NULL ++# define ext2_listxattr NULL ++# define ext2_removexattr NULL ++ ++static inline int ++ext2_xattr_get(struct inode *inode, int name_index, ++ const char *name, void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext2_xattr_list(struct inode *inode, char *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext2_xattr_set(struct inode *inode, int name_index, const char *name, ++ const void *value, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline void ++ext2_xattr_delete_inode(struct inode *inode) ++{ ++} ++ ++static inline void ++ext2_xattr_put_super(struct super_block *sb) ++{ ++} ++ ++static inline int ++init_ext2_xattr(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext2_xattr(void) ++{ ++} ++ ++# endif /* CONFIG_EXT2_FS_XATTR */ ++ ++# ifdef CONFIG_EXT2_FS_XATTR_USER ++ ++extern int init_ext2_xattr_user(void) __init; ++extern void exit_ext2_xattr_user(void); ++ ++# else /* CONFIG_EXT2_FS_XATTR_USER */ ++ ++static inline int ++init_ext2_xattr_user(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext2_xattr_user(void) ++{ ++} ++ ++# endif /* CONFIG_EXT2_FS_XATTR_USER */ ++ ++#endif /* __KERNEL__ */ ++ +--- linux-2.4.20/include/linux/ext3_fs.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/linux/ext3_fs.h Sat Apr 5 03:57:18 2003 +@@ -63,8 +63,6 @@ + */ + #define EXT3_BAD_INO 1 /* Bad blocks inode */ + #define EXT3_ROOT_INO 2 /* Root inode */ +-#define EXT3_ACL_IDX_INO 3 /* ACL inode */ +-#define EXT3_ACL_DATA_INO 4 /* ACL inode */ + #define EXT3_BOOT_LOADER_INO 5 /* Boot loader inode */ + #define EXT3_UNDEL_DIR_INO 6 /* Undelete directory inode */ + #define EXT3_RESIZE_INO 7 /* Reserved group descriptors inode */ +@@ -94,7 +92,6 @@ + #else + # define EXT3_BLOCK_SIZE(s) (EXT3_MIN_BLOCK_SIZE << (s)->s_log_block_size) + #endif +-#define EXT3_ACLE_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (struct ext3_acl_entry)) + #define EXT3_ADDR_PER_BLOCK(s) (EXT3_BLOCK_SIZE(s) / sizeof (__u32)) + #ifdef __KERNEL__ + # define EXT3_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits) +@@ -129,28 +126,6 @@ + #endif + + /* +- * ACL structures +- */ +-struct ext3_acl_header /* Header of Access Control Lists */ +-{ +- __u32 aclh_size; +- __u32 aclh_file_count; +- __u32 aclh_acle_count; +- __u32 aclh_first_acle; +-}; +- +-struct ext3_acl_entry /* Access Control List Entry */ +-{ +- __u32 acle_size; +- __u16 acle_perms; /* Access permissions */ +- __u16 acle_type; /* Type of entry */ +- __u16 acle_tag; /* User or group identity */ +- __u16 acle_pad1; +- __u32 acle_next; /* Pointer on next entry for the */ +- /* same inode or on next free entry */ +-}; +- +-/* + * Structure of a blocks group descriptor + */ + struct ext3_group_desc +@@ -344,6 +319,7 @@ struct ext3_inode { + #define EXT3_MOUNT_WRITEBACK_DATA 0x0C00 /* No data ordering */ + #define EXT3_MOUNT_UPDATE_JOURNAL 0x1000 /* Update the journal format */ + #define EXT3_MOUNT_NO_UID32 0x2000 /* Disable 32-bit UIDs */ ++#define EXT3_MOUNT_XATTR_USER 0x4000 /* Extended user attributes */ + + /* Compatibility, for having both ext2_fs.h and ext3_fs.h included at once */ + #ifndef _LINUX_EXT2_FS_H +@@ -520,7 +496,7 @@ struct ext3_super_block { + #define EXT3_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */ + #define EXT3_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */ + +-#define EXT3_FEATURE_COMPAT_SUPP 0 ++#define EXT3_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR + #define EXT3_FEATURE_INCOMPAT_SUPP (EXT3_FEATURE_INCOMPAT_FILETYPE| \ + EXT3_FEATURE_INCOMPAT_RECOVER) + #define EXT3_FEATURE_RO_COMPAT_SUPP (EXT3_FEATURE_RO_COMPAT_SPARSE_SUPER| \ +@@ -703,6 +679,7 @@ extern void ext3_check_inodes_bitmap (st + extern unsigned long ext3_count_free (struct buffer_head *, unsigned); + + /* inode.c */ ++extern int ext3_forget(handle_t *, int, struct inode *, struct buffer_head *, int); + extern struct buffer_head * ext3_getblk (handle_t *, struct inode *, long, int, int *); + extern struct buffer_head * ext3_bread (handle_t *, struct inode *, int, int, int *); + +@@ -771,8 +748,10 @@ extern struct address_space_operations e + + /* namei.c */ + extern struct inode_operations ext3_dir_inode_operations; ++extern struct inode_operations ext3_special_inode_operations; + + /* symlink.c */ ++extern struct inode_operations ext3_symlink_inode_operations; + extern struct inode_operations ext3_fast_symlink_inode_operations; + + +--- linux-2.4.20/include/linux/ext3_jbd.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/linux/ext3_jbd.h Sat Apr 5 03:57:18 2003 +@@ -30,13 +30,19 @@ + + #define EXT3_SINGLEDATA_TRANS_BLOCKS 8 + ++/* Extended attributes may touch two data buffers, two bitmap buffers, ++ * and two group and summaries. */ ++ ++#define EXT3_XATTR_TRANS_BLOCKS 8 ++ + /* Define the minimum size for a transaction which modifies data. This + * needs to take into account the fact that we may end up modifying two + * quota files too (one for the group, one for the user quota). The + * superblock only gets updated once, of course, so don't bother + * counting that again for the quota updates. */ + +-#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS - 2) ++#define EXT3_DATA_TRANS_BLOCKS (3 * EXT3_SINGLEDATA_TRANS_BLOCKS + \ ++ EXT3_XATTR_TRANS_BLOCKS - 2) + + extern int ext3_writepage_trans_blocks(struct inode *inode); + +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/include/linux/ext3_xattr.h Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,157 @@ ++/* ++ File: linux/ext3_xattr.h ++ ++ On-disk format of extended attributes for the ext3 filesystem. ++ ++ (C) 2001 Andreas Gruenbacher, ++*/ ++ ++#include ++#include ++#include ++ ++/* Magic value in attribute blocks */ ++#define EXT3_XATTR_MAGIC 0xEA020000 ++ ++/* Maximum number of references to one attribute block */ ++#define EXT3_XATTR_REFCOUNT_MAX 1024 ++ ++/* Name indexes */ ++#define EXT3_XATTR_INDEX_MAX 10 ++#define EXT3_XATTR_INDEX_USER 1 ++#define EXT3_XATTR_INDEX_POSIX_ACL_ACCESS 2 ++#define EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT 3 ++ ++struct ext3_xattr_header { ++ __u32 h_magic; /* magic number for identification */ ++ __u32 h_refcount; /* reference count */ ++ __u32 h_blocks; /* number of disk blocks used */ ++ __u32 h_hash; /* hash value of all attributes */ ++ __u32 h_reserved[4]; /* zero right now */ ++}; ++ ++struct ext3_xattr_entry { ++ __u8 e_name_len; /* length of name */ ++ __u8 e_name_index; /* attribute name index */ ++ __u16 e_value_offs; /* offset in disk block of value */ ++ __u32 e_value_block; /* disk block attribute is stored on (n/i) */ ++ __u32 e_value_size; /* size of attribute value */ ++ __u32 e_hash; /* hash value of name and value */ ++ char e_name[0]; /* attribute name */ ++}; ++ ++#define EXT3_XATTR_PAD_BITS 2 ++#define EXT3_XATTR_PAD (1<e_name_len)) ) ++#define EXT3_XATTR_SIZE(size) \ ++ (((size) + EXT3_XATTR_ROUND) & ~EXT3_XATTR_ROUND) ++ ++#ifdef __KERNEL__ ++ ++# ifdef CONFIG_EXT3_FS_XATTR ++ ++struct ext3_xattr_handler { ++ char *prefix; ++ size_t (*list)(char *list, struct inode *inode, const char *name, ++ int name_len); ++ int (*get)(struct inode *inode, const char *name, void *buffer, ++ size_t size); ++ int (*set)(struct inode *inode, const char *name, const void *buffer, ++ size_t size, int flags); ++}; ++ ++extern int ext3_xattr_register(int, struct ext3_xattr_handler *); ++extern void ext3_xattr_unregister(int, struct ext3_xattr_handler *); ++ ++extern int ext3_setxattr(struct dentry *, const char *, const void *, size_t, int); ++extern ssize_t ext3_getxattr(struct dentry *, const char *, void *, size_t); ++extern ssize_t ext3_listxattr(struct dentry *, char *, size_t); ++extern int ext3_removexattr(struct dentry *, const char *); ++ ++extern int ext3_xattr_get(struct inode *, int, const char *, void *, size_t); ++extern int ext3_xattr_list(struct inode *, char *, size_t); ++extern int ext3_xattr_set(handle_t *handle, struct inode *, int, const char *, const void *, size_t, int); ++ ++extern void ext3_xattr_delete_inode(handle_t *, struct inode *); ++extern void ext3_xattr_put_super(struct super_block *); ++ ++extern int init_ext3_xattr(void) __init; ++extern void exit_ext3_xattr(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR */ ++# define ext3_setxattr NULL ++# define ext3_getxattr NULL ++# define ext3_listxattr NULL ++# define ext3_removexattr NULL ++ ++static inline int ++ext3_xattr_get(struct inode *inode, int name_index, const char *name, ++ void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_list(struct inode *inode, void *buffer, size_t size) ++{ ++ return -ENOTSUP; ++} ++ ++static inline int ++ext3_xattr_set(handle_t *handle, struct inode *inode, int name_index, ++ const char *name, const void *value, size_t size, int flags) ++{ ++ return -ENOTSUP; ++} ++ ++static inline void ++ext3_xattr_delete_inode(handle_t *handle, struct inode *inode) ++{ ++} ++ ++static inline void ++ext3_xattr_put_super(struct super_block *sb) ++{ ++} ++ ++static inline int ++init_ext3_xattr(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr(void) ++{ ++} ++ ++# endif /* CONFIG_EXT3_FS_XATTR */ ++ ++# ifdef CONFIG_EXT3_FS_XATTR_USER ++ ++extern int init_ext3_xattr_user(void) __init; ++extern void exit_ext3_xattr_user(void); ++ ++# else /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++static inline int ++init_ext3_xattr_user(void) ++{ ++ return 0; ++} ++ ++static inline void ++exit_ext3_xattr_user(void) ++{ ++} ++ ++#endif /* CONFIG_EXT3_FS_XATTR_USER */ ++ ++#endif /* __KERNEL__ */ ++ +--- linux-2.4.20/include/linux/fs.h~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/include/linux/fs.h Sat Apr 5 03:57:18 2003 +@@ -888,7 +888,7 @@ struct inode_operations { + 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 *, void *, size_t, int); ++ int (*setxattr) (struct dentry *, const char *, const void *, size_t, int); + ssize_t (*getxattr) (struct dentry *, const char *, void *, size_t); + ssize_t (*listxattr) (struct dentry *, char *, size_t); + int (*removexattr) (struct dentry *, const char *); +--- /dev/null Fri Aug 30 17:31:37 2002 ++++ linux-2.4.20-braam/include/linux/mbcache.h Sat Apr 5 03:57:18 2003 +@@ -0,0 +1,69 @@ ++/* ++ File: linux/mbcache.h ++ ++ (C) 2001 by Andreas Gruenbacher, ++*/ ++ ++/* Hardwire the number of additional indexes */ ++#define MB_CACHE_INDEXES_COUNT 1 ++ ++struct mb_cache_entry; ++ ++struct mb_cache_op { ++ int (*free)(struct mb_cache_entry *, int); ++}; ++ ++struct mb_cache { ++ struct list_head c_cache_list; ++ const char *c_name; ++ struct mb_cache_op c_op; ++ atomic_t c_entry_count; ++ int c_bucket_count; ++#ifndef MB_CACHE_INDEXES_COUNT ++ int c_indexes_count; ++#endif ++ kmem_cache_t *c_entry_cache; ++ struct list_head *c_block_hash; ++ struct list_head *c_indexes_hash[0]; ++}; ++ ++struct mb_cache_entry_index { ++ struct list_head o_list; ++ unsigned int o_key; ++}; ++ ++struct mb_cache_entry { ++ struct list_head e_lru_list; ++ struct mb_cache *e_cache; ++ atomic_t e_used; ++ kdev_t e_dev; ++ unsigned long e_block; ++ struct list_head e_block_list; ++ struct mb_cache_entry_index e_indexes[0]; ++}; ++ ++/* Functions on caches */ ++ ++struct mb_cache * mb_cache_create(const char *, struct mb_cache_op *, size_t, ++ int, int); ++void mb_cache_shrink(struct mb_cache *, kdev_t); ++void mb_cache_destroy(struct mb_cache *); ++ ++/* Functions on cache entries */ ++ ++struct mb_cache_entry *mb_cache_entry_alloc(struct mb_cache *); ++int mb_cache_entry_insert(struct mb_cache_entry *, kdev_t, unsigned long, ++ unsigned int[]); ++void mb_cache_entry_rehash(struct mb_cache_entry *, unsigned int[]); ++void mb_cache_entry_release(struct mb_cache_entry *); ++void mb_cache_entry_takeout(struct mb_cache_entry *); ++void mb_cache_entry_free(struct mb_cache_entry *); ++struct mb_cache_entry *mb_cache_entry_dup(struct mb_cache_entry *); ++struct mb_cache_entry *mb_cache_entry_get(struct mb_cache *, kdev_t, ++ unsigned long); ++#if !defined(MB_CACHE_INDEXES_COUNT) || (MB_CACHE_INDEXES_COUNT > 0) ++struct mb_cache_entry *mb_cache_entry_find_first(struct mb_cache *cache, int, ++ kdev_t, unsigned int); ++struct mb_cache_entry *mb_cache_entry_find_next(struct mb_cache_entry *, int, ++ kdev_t, unsigned int); ++#endif +--- linux-2.4.20/kernel/ksyms.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/kernel/ksyms.c Sat Apr 5 03:57:18 2003 +@@ -11,6 +11,7 @@ + + #include + #include ++#include + #include + #include + #include +@@ -89,6 +90,7 @@ EXPORT_SYMBOL(exit_mm); + EXPORT_SYMBOL(exit_files); + EXPORT_SYMBOL(exit_fs); + EXPORT_SYMBOL(exit_sighand); ++EXPORT_SYMBOL(copy_fs_struct); + + /* internal kernel memory management */ + EXPORT_SYMBOL(_alloc_pages); +@@ -107,6 +109,8 @@ EXPORT_SYMBOL(kmem_cache_validate); + EXPORT_SYMBOL(kmem_cache_alloc); + EXPORT_SYMBOL(kmem_cache_free); + EXPORT_SYMBOL(kmem_cache_size); ++EXPORT_SYMBOL(register_cache); ++EXPORT_SYMBOL(unregister_cache); + EXPORT_SYMBOL(kmalloc); + EXPORT_SYMBOL(kfree); + EXPORT_SYMBOL(vfree); +--- linux-2.4.20/mm/vmscan.c~linux-2.4.20-xattr-0.8.54 Sat Apr 5 03:57:18 2003 ++++ linux-2.4.20-braam/mm/vmscan.c Sat Apr 5 03:57:18 2003 +@@ -18,6 +18,7 @@ + #include + #include + #include ++#include + #include + #include + #include +@@ -34,6 +35,39 @@ + */ + #define DEF_PRIORITY (6) + ++static DECLARE_MUTEX(other_caches_sem); ++static LIST_HEAD(cache_definitions); ++ ++void register_cache(struct cache_definition *cache) ++{ ++ down(&other_caches_sem); ++ list_add(&cache->link, &cache_definitions); ++ up(&other_caches_sem); ++} ++ ++void unregister_cache(struct cache_definition *cache) ++{ ++ down(&other_caches_sem); ++ list_del(&cache->link); ++ up(&other_caches_sem); ++} ++ ++static void shrink_other_caches(unsigned int priority, int gfp_mask) ++{ ++ struct list_head *p; ++ ++ if (down_trylock(&other_caches_sem)) ++ return; ++ ++ list_for_each_prev(p, &cache_definitions) { ++ struct cache_definition *cache = ++ list_entry(p, struct cache_definition, link); ++ ++ cache->shrink(priority, gfp_mask); ++ } ++ up(&other_caches_sem); ++} ++ + /* + * The swap-out function returns 1 if it successfully + * scanned all the pages it was asked to (`count'). +@@ -577,6 +611,7 @@ static int shrink_caches(zone_t * classz + + shrink_dcache_memory(priority, gfp_mask); + shrink_icache_memory(priority, gfp_mask); ++ shrink_other_caches(priority, gfp_mask); + #ifdef CONFIG_QUOTA + shrink_dqcache_memory(DEF_PRIORITY, gfp_mask); + #endif + +_ diff --git a/lustre/kernel_patches/patches/patch-2.4.18-chaos22.patch b/lustre/kernel_patches/patches/patch-2.4.18-chaos22.patch new file mode 100644 index 0000000..c40d4ea --- /dev/null +++ b/lustre/kernel_patches/patches/patch-2.4.18-chaos22.patch @@ -0,0 +1,165 @@ +diff -ru lum-2.4.18-um30/fs/ext3/balloc.c uml-2.4.18-12.5/fs/ext3/balloc.c +--- lum-2.4.18-um30/fs/ext3/balloc.c Mon Feb 25 12:38:08 2002 ++++ uml-2.4.18-12.5/fs/ext3/balloc.c Thu Sep 19 13:40:11 2002 +@@ -276,7 +276,8 @@ + } + lock_super (sb); + es = sb->u.ext3_sb.s_es; +- if (block < le32_to_cpu(es->s_first_data_block) || ++ if (block < le32_to_cpu(es->s_first_data_block) || ++ block + count < block || + (block + count) > le32_to_cpu(es->s_blocks_count)) { + ext3_error (sb, "ext3_free_blocks", + "Freeing blocks not in datazone - " +@@ -309,17 +310,6 @@ + if (!gdp) + goto error_return; + +- if (in_range (le32_to_cpu(gdp->bg_block_bitmap), block, count) || +- in_range (le32_to_cpu(gdp->bg_inode_bitmap), block, count) || +- in_range (block, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group) || +- in_range (block + count - 1, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group)) +- ext3_error (sb, "ext3_free_blocks", +- "Freeing blocks in system zones - " +- "Block = %lu, count = %lu", +- block, count); +- + /* + * We are about to start releasing blocks in the bitmap, + * so we need undo access. +@@ -345,14 +335,24 @@ + if (err) + goto error_return; + +- for (i = 0; i < count; i++) { ++ for (i = 0; i < count; i++, block++) { ++ if (block == le32_to_cpu(gdp->bg_block_bitmap) || ++ block == le32_to_cpu(gdp->bg_inode_bitmap) || ++ in_range(block, le32_to_cpu(gdp->bg_inode_table), ++ sb->u.ext2_sb.s_itb_per_group)) { ++ ext3_error(sb, __FUNCTION__, ++ "Freeing block in system zone - block = %lu", ++ block); ++ continue; ++ } ++ + /* + * An HJ special. This is expensive... + */ + #ifdef CONFIG_JBD_DEBUG + { + struct buffer_head *debug_bh; +- debug_bh = sb_get_hash_table(sb, block + i); ++ debug_bh = sb_get_hash_table(sb, block); + if (debug_bh) { + BUFFER_TRACE(debug_bh, "Deleted!"); + if (!bh2jh(bitmap_bh)->b_committed_data) +@@ -365,9 +365,8 @@ + #endif + BUFFER_TRACE(bitmap_bh, "clear bit"); + if (!ext3_clear_bit (bit + i, bitmap_bh->b_data)) { +- ext3_error (sb, __FUNCTION__, +- "bit already cleared for block %lu", +- block + i); ++ ext3_error(sb, __FUNCTION__, ++ "bit already cleared for block %lu", block); + BUFFER_TRACE(bitmap_bh, "bit already cleared"); + } else { + dquot_freed_blocks++; +@@ -415,7 +417,6 @@ + if (!err) err = ret; + + if (overflow && !err) { +- block += count; + count = overflow; + goto do_more; + } +@@ -575,6 +577,7 @@ + + ext3_debug ("goal=%lu.\n", goal); + ++repeat: + /* + * First, test whether the goal block is free. + */ +@@ -684,10 +686,21 @@ + if (tmp == le32_to_cpu(gdp->bg_block_bitmap) || + tmp == le32_to_cpu(gdp->bg_inode_bitmap) || + in_range (tmp, le32_to_cpu(gdp->bg_inode_table), +- sb->u.ext3_sb.s_itb_per_group)) +- ext3_error (sb, "ext3_new_block", +- "Allocating block in system zone - " +- "block = %u", tmp); ++ EXT3_SB(sb)->s_itb_per_group)) { ++ ext3_error(sb, __FUNCTION__, ++ "Allocating block in system zone - block = %u", tmp); ++ ++ /* Note: This will potentially use up one of the handle's ++ * buffer credits. Normally we have way too many credits, ++ * so that is OK. In _very_ rare cases it might not be OK. ++ * We will trigger an assertion if we run out of credits, ++ * and we will have to do a full fsck of the filesystem - ++ * better than randomly corrupting filesystem metadata. ++ */ ++ ext3_set_bit(j, bh->b_data); ++ goto repeat; ++ } ++ + + /* The superblock lock should guard against anybody else beating + * us to this point! */ +diff -ru lum-2.4.18-um30/fs/ext3/namei.c uml-2.4.18-12.5/fs/ext3/namei.c +--- lum-2.4.18-um30/fs/ext3/namei.c Fri Nov 9 15:25:04 2001 ++++ uml-2.4.18-12.5/fs/ext3/namei.c Thu Sep 19 13:40:11 2002 +@@ -354,8 +355,8 @@ + */ + dir->i_mtime = dir->i_ctime = CURRENT_TIME; + dir->u.ext3_i.i_flags &= ~EXT3_INDEX_FL; +- ext3_mark_inode_dirty(handle, dir); + dir->i_version = ++event; ++ ext3_mark_inode_dirty(handle, dir); + BUFFER_TRACE(bh, "call ext3_journal_dirty_metadata"); + ext3_journal_dirty_metadata(handle, bh); + brelse(bh); +@@ -464,8 +465,8 @@ + 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_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -489,8 +490,8 @@ + err = PTR_ERR(inode); + if (!IS_ERR(inode)) { + init_special_inode(inode, mode, rdev); +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + } + ext3_journal_stop(handle, dir); + return err; +@@ -933,8 +934,8 @@ + inode->i_size = l-1; + } + inode->u.ext3_i.i_disksize = inode->i_size; +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + out_stop: + ext3_journal_stop(handle, dir); + return err; +@@ -970,8 +971,8 @@ + ext3_inc_count(handle, inode); + atomic_inc(&inode->i_count); + +- ext3_mark_inode_dirty(handle, inode); + err = ext3_add_nondir(handle, dentry, inode); ++ ext3_mark_inode_dirty(handle, inode); + ext3_journal_stop(handle, dir); + return err; + } diff --git a/lustre/kernel_patches/pc/ext-2.4-patch-1.pc b/lustre/kernel_patches/pc/ext-2.4-patch-1.pc new file mode 100644 index 0000000..634b944 --- /dev/null +++ b/lustre/kernel_patches/pc/ext-2.4-patch-1.pc @@ -0,0 +1,11 @@ +fs/ext3/Makefile +fs/ext3/dir.c +fs/ext3/file.c +fs/ext3/hash.c +fs/ext3/namei.c +fs/ext3/super.c +include/linux/ext3_fs.h +include/linux/ext3_fs_sb.h +include/linux/ext3_jbd.h +include/linux/rbtree.h +lib/rbtree.c diff --git a/lustre/kernel_patches/pc/ext-2.4-patch-2.pc b/lustre/kernel_patches/pc/ext-2.4-patch-2.pc new file mode 100644 index 0000000..9b16759 --- /dev/null +++ b/lustre/kernel_patches/pc/ext-2.4-patch-2.pc @@ -0,0 +1 @@ +fs/ext3/namei.c diff --git a/lustre/kernel_patches/pc/ext-2.4-patch-3.pc b/lustre/kernel_patches/pc/ext-2.4-patch-3.pc new file mode 100644 index 0000000..65d4845 --- /dev/null +++ b/lustre/kernel_patches/pc/ext-2.4-patch-3.pc @@ -0,0 +1,3 @@ +fs/ext3/dir.c +fs/ext3/namei.c +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/ext-2.4-patch-4.pc b/lustre/kernel_patches/pc/ext-2.4-patch-4.pc new file mode 100644 index 0000000..9b16759 --- /dev/null +++ b/lustre/kernel_patches/pc/ext-2.4-patch-4.pc @@ -0,0 +1 @@ +fs/ext3/namei.c diff --git a/lustre/kernel_patches/pc/ext3-2.4-ino_t.pc b/lustre/kernel_patches/pc/ext3-2.4-ino_t.pc new file mode 100644 index 0000000..4cef979 --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-2.4-ino_t.pc @@ -0,0 +1,3 @@ +fs/ext3/ialloc.c +fs/ext3/namei.c +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/ext3-2.4.18-fixes.pc b/lustre/kernel_patches/pc/ext3-2.4.18-fixes.pc new file mode 100644 index 0000000..0822c5e --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-2.4.18-fixes.pc @@ -0,0 +1,7 @@ +fs/ext3/balloc.c +fs/ext3/file.c +fs/ext3/fsync.c +fs/ext3/ialloc.c +fs/ext3/inode.c +fs/ext3/namei.c +fs/ext3/super.c diff --git a/lustre/kernel_patches/pc/ext3-2.4.18-ino_sb_macro.pc b/lustre/kernel_patches/pc/ext3-2.4.18-ino_sb_macro.pc new file mode 100644 index 0000000..cd21583 --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-2.4.18-ino_sb_macro.pc @@ -0,0 +1,10 @@ +fs/ext3/balloc.c +fs/ext3/dir.c +fs/ext3/ialloc.c +fs/ext3/inode.c +fs/ext3/ioctl.c +fs/ext3/namei.c +fs/ext3/super.c +fs/ext3/symlink.c +include/linux/ext3_fs.h +include/linux/ext3_jbd.h diff --git a/lustre/kernel_patches/pc/ext3-2.5-noread.pc b/lustre/kernel_patches/pc/ext3-2.5-noread.pc new file mode 100644 index 0000000..9c3cea8 --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-2.5-noread.pc @@ -0,0 +1,3 @@ +fs/ext3/ialloc.c +fs/ext3/inode.c +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/ext3-largefile.pc b/lustre/kernel_patches/pc/ext3-largefile.pc new file mode 100644 index 0000000..76d683f --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-largefile.pc @@ -0,0 +1 @@ +fs/ext3/inode.c diff --git a/lustre/kernel_patches/pc/ext3-orphan_lock.pc b/lustre/kernel_patches/pc/ext3-orphan_lock.pc new file mode 100644 index 0000000..98aebb0 --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-orphan_lock.pc @@ -0,0 +1,3 @@ +fs/ext3/namei.c +fs/ext3/super.c +include/linux/ext3_fs_sb.h diff --git a/lustre/kernel_patches/pc/ext3-truncate_blocks.pc b/lustre/kernel_patches/pc/ext3-truncate_blocks.pc new file mode 100644 index 0000000..76d683f --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-truncate_blocks.pc @@ -0,0 +1 @@ +fs/ext3/inode.c diff --git a/lustre/kernel_patches/pc/ext3-unmount_sync.pc b/lustre/kernel_patches/pc/ext3-unmount_sync.pc new file mode 100644 index 0000000..08795de --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-unmount_sync.pc @@ -0,0 +1 @@ +fs/ext3/super.c diff --git a/lustre/kernel_patches/pc/ext3-use-after-free.pc b/lustre/kernel_patches/pc/ext3-use-after-free.pc new file mode 100644 index 0000000..9b16759 --- /dev/null +++ b/lustre/kernel_patches/pc/ext3-use-after-free.pc @@ -0,0 +1 @@ +fs/ext3/namei.c diff --git a/lustre/kernel_patches/pc/extN-2.4.18-exports.pc b/lustre/kernel_patches/pc/extN-2.4.18-exports.pc new file mode 100644 index 0000000..08795de --- /dev/null +++ b/lustre/kernel_patches/pc/extN-2.4.18-exports.pc @@ -0,0 +1 @@ +fs/ext3/super.c diff --git a/lustre/kernel_patches/pc/extN-2.4.18-ino_sb_fixup.pc b/lustre/kernel_patches/pc/extN-2.4.18-ino_sb_fixup.pc new file mode 100644 index 0000000..7191405 --- /dev/null +++ b/lustre/kernel_patches/pc/extN-2.4.18-ino_sb_fixup.pc @@ -0,0 +1 @@ +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/extN-delete_thread.pc b/lustre/kernel_patches/pc/extN-delete_thread.pc new file mode 100644 index 0000000..bc81732 --- /dev/null +++ b/lustre/kernel_patches/pc/extN-delete_thread.pc @@ -0,0 +1,3 @@ +include/linux/ext3_fs.h +include/linux/ext3_fs_sb.h +fs/ext3/super.c diff --git a/lustre/kernel_patches/pc/extN-iget-debug.pc b/lustre/kernel_patches/pc/extN-iget-debug.pc new file mode 100644 index 0000000..e9fe01e --- /dev/null +++ b/lustre/kernel_patches/pc/extN-iget-debug.pc @@ -0,0 +1,2 @@ +fs/ext3/namei.c +fs/ext3/inode.c diff --git a/lustre/kernel_patches/pc/extN-misc-fixup.pc b/lustre/kernel_patches/pc/extN-misc-fixup.pc new file mode 100644 index 0000000..08795de --- /dev/null +++ b/lustre/kernel_patches/pc/extN-misc-fixup.pc @@ -0,0 +1 @@ +fs/ext3/super.c diff --git a/lustre/kernel_patches/pc/extN-noread.pc b/lustre/kernel_patches/pc/extN-noread.pc new file mode 100644 index 0000000..9c3cea8 --- /dev/null +++ b/lustre/kernel_patches/pc/extN-noread.pc @@ -0,0 +1,3 @@ +fs/ext3/ialloc.c +fs/ext3/inode.c +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/extN-san.pc b/lustre/kernel_patches/pc/extN-san.pc new file mode 100644 index 0000000..76d683f --- /dev/null +++ b/lustre/kernel_patches/pc/extN-san.pc @@ -0,0 +1 @@ +fs/ext3/inode.c diff --git a/lustre/kernel_patches/pc/extN-wantedi.pc b/lustre/kernel_patches/pc/extN-wantedi.pc new file mode 100644 index 0000000..31901ee --- /dev/null +++ b/lustre/kernel_patches/pc/extN-wantedi.pc @@ -0,0 +1,4 @@ +fs/ext3/namei.c +fs/ext3/ialloc.c +fs/ext3/ioctl.c +include/linux/ext3_fs.h diff --git a/lustre/kernel_patches/pc/htree-ext3-2.4.18.pc b/lustre/kernel_patches/pc/htree-ext3-2.4.18.pc new file mode 100644 index 0000000..6499778 --- /dev/null +++ b/lustre/kernel_patches/pc/htree-ext3-2.4.18.pc @@ -0,0 +1,4 @@ +fs/ext3/super.c +fs/ext3/namei.c +include/linux/ext3_fs.h +include/linux/ext3_jbd.h diff --git a/lustre/kernel_patches/pc/linux-2.4.18ea-0.8.26.pc b/lustre/kernel_patches/pc/linux-2.4.18ea-0.8.26.pc new file mode 100644 index 0000000..e00b968 --- /dev/null +++ b/lustre/kernel_patches/pc/linux-2.4.18ea-0.8.26.pc @@ -0,0 +1,9 @@ +fs/ext3/ialloc.c +fs/ext3/inode.c +fs/ext3/namei.c +fs/ext3/super.c +fs/ext3/xattr.c +include/linux/ext3_fs.h +include/linux/ext3_jbd.h +include/linux/ext3_xattr.h +include/linux/xattr.h diff --git a/lustre/kernel_patches/pc/linux-2.4.20-xattr-0.8.54.pc b/lustre/kernel_patches/pc/linux-2.4.20-xattr-0.8.54.pc new file mode 100644 index 0000000..32344f6 --- /dev/null +++ b/lustre/kernel_patches/pc/linux-2.4.20-xattr-0.8.54.pc @@ -0,0 +1,63 @@ +Documentation/Configure.help +arch/alpha/defconfig +arch/alpha/kernel/entry.S +arch/arm/defconfig +arch/arm/kernel/calls.S +arch/i386/defconfig +arch/ia64/defconfig +arch/ia64/kernel/entry.S +arch/m68k/defconfig +arch/mips/defconfig +arch/mips64/defconfig +arch/ppc/defconfig +arch/ppc64/kernel/misc.S +arch/s390/defconfig +arch/s390/kernel/entry.S +arch/s390x/defconfig +arch/s390x/kernel/entry.S +arch/s390x/kernel/wrapper32.S +arch/sparc/defconfig +arch/sparc/kernel/systbls.S +arch/sparc64/defconfig +arch/sparc64/kernel/systbls.S +fs/Config.in +fs/Makefile +fs/ext2/Makefile +fs/ext2/file.c +fs/ext2/ialloc.c +fs/ext2/inode.c +fs/ext2/namei.c +fs/ext2/super.c +fs/ext2/symlink.c +fs/ext2/xattr.c +fs/ext2/xattr_user.c +fs/ext3/Makefile +fs/ext3/file.c +fs/ext3/ialloc.c +fs/ext3/inode.c +fs/ext3/namei.c +fs/ext3/super.c +fs/ext3/symlink.c +fs/ext3/xattr.c +fs/ext3/xattr_user.c +fs/jfs/jfs_xattr.h +fs/jfs/xattr.c +fs/mbcache.c +include/asm-arm/unistd.h +include/asm-ia64/unistd.h +include/asm-ppc64/unistd.h +include/asm-s390/unistd.h +include/asm-s390x/unistd.h +include/asm-sparc/unistd.h +include/asm-sparc64/unistd.h +include/linux/cache_def.h +include/linux/errno.h +include/linux/ext2_fs.h +include/linux/ext2_xattr.h +include/linux/ext3_fs.h +include/linux/ext3_jbd.h +include/linux/ext3_xattr.h +include/linux/fs.h +include/linux/mbcache.h +kernel/ksyms.c +mm/vmscan.c diff --git a/lustre/kernel_patches/series/vanilla-2.4.20 b/lustre/kernel_patches/series/vanilla-2.4.20 index e59cec0..24d4d53 100644 --- a/lustre/kernel_patches/series/vanilla-2.4.20 +++ b/lustre/kernel_patches/series/vanilla-2.4.20 @@ -8,3 +8,9 @@ export-truncate.patch iod-stock-24-exports.patch uml_check_get_page.patch uml_no_panic.patch +ext-2.4-patch-1.patch +ext-2.4-patch-2.patch +ext-2.4-patch-3.patch +ext-2.4-patch-4.patch +linux-2.4.20-xattr-0.8.54.patch +extN-san.patch -- 1.8.3.1