Commit 7c6d1fb9 authored by Andrew Morton's avatar Andrew Morton Committed by Linus Torvalds

[PATCH] Fix ext3 htree / NFS compatibility problems

Patch from "Theodore Ts'o" <tytso@mit.edu>

The following patch should (in theory) fix the htree/NFS readdir problems
that people have reported.  Specifically, it should fix the NFS looping on
EOF problem with readdir, as well as the problems caused by coverting a
directory to HTREE while an NFS readdir is in progress problem.

I'd appreciate it if people who can easily replicate these NFS/htree problems
could give this patch (against BK-recent / 2.5.63) a whirl.  Thanks!!
parent 96198d06
...@@ -103,7 +103,11 @@ static int ext3_readdir(struct file * filp, ...@@ -103,7 +103,11 @@ static int ext3_readdir(struct file * filp,
sb = inode->i_sb; sb = inode->i_sb;
if (is_dx(inode)) { #ifdef CONFIG_EXT3_INDEX
if (EXT3_HAS_COMPAT_FEATURE(inode->i_sb,
EXT3_FEATURE_COMPAT_DIR_INDEX) &&
((EXT3_I(inode)->i_flags & EXT3_INDEX_FL) ||
((inode->i_size >> sb->s_blocksize_bits) == 1))) {
err = ext3_dx_readdir(filp, dirent, filldir); err = ext3_dx_readdir(filp, dirent, filldir);
if (err != ERR_BAD_DX_DIR) { if (err != ERR_BAD_DX_DIR) {
ret = err; ret = err;
...@@ -115,6 +119,7 @@ static int ext3_readdir(struct file * filp, ...@@ -115,6 +119,7 @@ static int ext3_readdir(struct file * filp,
*/ */
EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL; EXT3_I(filp->f_dentry->d_inode)->i_flags &= ~EXT3_INDEX_FL;
} }
#endif
stored = 0; stored = 0;
bh = NULL; bh = NULL;
offset = filp->f_pos & (sb->s_blocksize - 1); offset = filp->f_pos & (sb->s_blocksize - 1);
...@@ -434,6 +439,9 @@ static int ext3_dx_readdir(struct file * filp, ...@@ -434,6 +439,9 @@ static int ext3_dx_readdir(struct file * filp,
filp->private_data = info; filp->private_data = info;
} }
if (filp->f_pos == -1)
return 0; /* EOF */
/* Some one has messed with f_pos; reset the world */ /* Some one has messed with f_pos; reset the world */
if (info->last_pos != filp->f_pos) { if (info->last_pos != filp->f_pos) {
free_rb_tree_fname(&info->root); free_rb_tree_fname(&info->root);
...@@ -470,8 +478,10 @@ static int ext3_dx_readdir(struct file * filp, ...@@ -470,8 +478,10 @@ static int ext3_dx_readdir(struct file * filp,
&info->next_hash); &info->next_hash);
if (ret < 0) if (ret < 0)
return ret; return ret;
if (ret == 0) if (ret == 0) {
filp->f_pos = -1;
break; break;
}
info->curr_node = rb_first(&info->root); info->curr_node = rb_first(&info->root);
} }
...@@ -483,6 +493,10 @@ static int ext3_dx_readdir(struct file * filp, ...@@ -483,6 +493,10 @@ static int ext3_dx_readdir(struct file * filp,
info->curr_node = rb_next(info->curr_node); info->curr_node = rb_next(info->curr_node);
if (!info->curr_node) { if (!info->curr_node) {
if (info->next_hash == ~0) {
filp->f_pos = -1;
break;
}
info->curr_hash = info->next_hash; info->curr_hash = info->next_hash;
info->curr_minor_hash = 0; info->curr_minor_hash = 0;
} }
......
...@@ -170,7 +170,7 @@ static struct ext3_dir_entry_2* dx_pack_dirents (char *base, int size); ...@@ -170,7 +170,7 @@ 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 void dx_insert_block (struct dx_frame *frame, u32 hash, u32 block);
static int ext3_htree_next_block(struct inode *dir, __u32 hash, static int ext3_htree_next_block(struct inode *dir, __u32 hash,
struct dx_frame *frame, struct dx_frame *frame,
struct dx_frame *frames, int *err, struct dx_frame *frames,
__u32 *start_hash); __u32 *start_hash);
static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry,
struct ext3_dir_entry_2 **res_dir, int *err); struct ext3_dir_entry_2 **res_dir, int *err);
...@@ -239,6 +239,17 @@ static inline unsigned dx_node_limit (struct inode *dir) ...@@ -239,6 +239,17 @@ static inline unsigned dx_node_limit (struct inode *dir)
* Debug * Debug
*/ */
#ifdef DX_DEBUG #ifdef DX_DEBUG
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 struct stats
{ {
unsigned names; unsigned names;
...@@ -447,22 +458,21 @@ static void dx_release (struct dx_frame *frames) ...@@ -447,22 +458,21 @@ static void dx_release (struct dx_frame *frames)
* *
* This function returns 1 if the caller should continue to search, * 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 * or 0 if it should not. If there is an error reading one of the
* index blocks, it will return -1. * index blocks, it will a negative error code.
* *
* If start_hash is non-null, it will be filled in with the starting * If start_hash is non-null, it will be filled in with the starting
* hash of the next page. * hash of the next page.
*/ */
static int ext3_htree_next_block(struct inode *dir, __u32 hash, static int ext3_htree_next_block(struct inode *dir, __u32 hash,
struct dx_frame *frame, struct dx_frame *frame,
struct dx_frame *frames, int *err, struct dx_frame *frames,
__u32 *start_hash) __u32 *start_hash)
{ {
struct dx_frame *p; struct dx_frame *p;
struct buffer_head *bh; struct buffer_head *bh;
int num_frames = 0; int err, num_frames = 0;
__u32 bhash; __u32 bhash;
*err = ENOENT;
p = frame; p = frame;
/* /*
* Find the next leaf page by incrementing the frame pointer. * Find the next leaf page by incrementing the frame pointer.
...@@ -500,8 +510,8 @@ static int ext3_htree_next_block(struct inode *dir, __u32 hash, ...@@ -500,8 +510,8 @@ static int ext3_htree_next_block(struct inode *dir, __u32 hash,
*/ */
while (num_frames--) { while (num_frames--) {
if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at), if (!(bh = ext3_bread(NULL, dir, dx_get_block(p->at),
0, err))) 0, &err)))
return -1; /* Failure */ return err; /* Failure */
p++; p++;
brelse (p->bh); brelse (p->bh);
p->bh = bh; p->bh = bh;
...@@ -519,6 +529,46 @@ static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 * ...@@ -519,6 +529,46 @@ static inline struct ext3_dir_entry_2 *ext3_next_entry(struct ext3_dir_entry_2 *
return (struct ext3_dir_entry_2 *)((char*)p + le16_to_cpu(p->rec_len)); 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 block. It returns the number directory entries loaded
* into the tree. If there is an error it is returned in err.
*/
static int htree_dirblock_to_tree(struct file *dir_file,
struct inode *dir, int block,
struct dx_hash_info *hinfo,
__u32 start_hash, __u32 start_minor_hash)
{
struct buffer_head *bh;
struct ext3_dir_entry_2 *de, *top;
int err, count = 0;
dxtrace(printk("In htree dirblock_to_tree: block %d\n", block));
if (!(bh = ext3_bread (NULL, dir, block, 0, &err)))
return err;
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) {
brelse(bh);
return err;
}
count++;
}
brelse(bh);
return count;
}
/* /*
* This function fills a red-black tree with information from a * This function fills a red-black tree with information from a
* directory. We start scanning the directory in hash order, starting * directory. We start scanning the directory in hash order, starting
...@@ -531,8 +581,7 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ...@@ -531,8 +581,7 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash,
__u32 start_minor_hash, __u32 *next_hash) __u32 start_minor_hash, __u32 *next_hash)
{ {
struct dx_hash_info hinfo; struct dx_hash_info hinfo;
struct buffer_head *bh; struct ext3_dir_entry_2 *de;
struct ext3_dir_entry_2 *de, *top;
struct dx_frame frames[2], *frame; struct dx_frame frames[2], *frame;
struct inode *dir; struct inode *dir;
int block, err; int block, err;
...@@ -543,6 +592,14 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ...@@ -543,6 +592,14 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash,
dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash, dxtrace(printk("In htree_fill_tree, start hash: %x:%x\n", start_hash,
start_minor_hash)); start_minor_hash));
dir = dir_file->f_dentry->d_inode; dir = dir_file->f_dentry->d_inode;
if (!(EXT3_I(dir)->i_flags & EXT3_INDEX_FL)) {
hinfo.hash_version = EXT3_SB(dir->i_sb)->s_def_hash_version;
hinfo.seed = EXT3_SB(dir->i_sb)->s_hash_seed;
count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo,
start_hash, start_minor_hash);
*next_hash = ~0;
return count;
}
hinfo.hash = start_hash; hinfo.hash = start_hash;
hinfo.minor_hash = 0; hinfo.minor_hash = 0;
frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err); frame = dx_probe(0, dir_file->f_dentry->d_inode, &hinfo, frames, &err);
...@@ -562,34 +619,21 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ...@@ -562,34 +619,21 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash,
while (1) { while (1) {
block = dx_get_block(frame->at); block = dx_get_block(frame->at);
dxtrace(printk("Reading block %d\n", block)); ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo,
if (!(bh = ext3_bread (NULL, dir, block, 0, &err))) start_hash, start_minor_hash);
if (ret < 0) {
err = ret;
goto errout; 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) {
brelse(bh);
goto errout;
}
count++;
} }
brelse (bh); count += ret;
hashval = ~1; hashval = ~0;
ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS, ret = ext3_htree_next_block(dir, HASH_NB_ALWAYS,
frame, frames, &err, &hashval); frame, frames, &hashval);
if (next_hash)
*next_hash = hashval; *next_hash = hashval;
if (ret == -1) if (ret < 0) {
err = ret;
goto errout; goto errout;
}
/* /*
* Stop if: (a) there are no more entries, or * Stop if: (a) there are no more entries, or
* (b) we have inserted at least one entry and the * (b) we have inserted at least one entry and the
...@@ -600,7 +644,8 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash, ...@@ -600,7 +644,8 @@ int ext3_htree_fill_tree(struct file *dir_file, __u32 start_hash,
break; break;
} }
dx_release(frames); dx_release(frames);
dxtrace(printk("Fill tree: returned %d entries\n", count)); dxtrace(printk("Fill tree: returned %d entries, next hash: %x\n",
count, *next_hash));
return count; return count;
errout: errout:
dx_release(frames); dx_release(frames);
...@@ -909,11 +954,12 @@ static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry, ...@@ -909,11 +954,12 @@ static struct buffer_head * ext3_dx_find_entry(struct dentry *dentry,
brelse (bh); brelse (bh);
/* Check to see if we should continue to search */ /* Check to see if we should continue to search */
retval = ext3_htree_next_block(dir, hash, frame, retval = ext3_htree_next_block(dir, hash, frame,
frames, err, 0); frames, 0);
if (retval == -1) { if (retval < 0) {
ext3_warning(sb, __FUNCTION__, ext3_warning(sb, __FUNCTION__,
"error reading index page in directory #%lu", "error reading index page in directory #%lu",
dir->i_ino); dir->i_ino);
*err = retval;
goto errout; goto errout;
} }
} while (retval == 1); } while (retval == 1);
......
Markdown is supported
0%
or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment