/*
* rehash.c --- rebuild hash tree directories
- *
+ *
* Copyright (C) 2002 Theodore Ts'o
*
* %Begin-Header%
* This file may be redistributed under the terms of the GNU Public
* License.
* %End-Header%
- *
+ *
* This algorithm is designed for simplicity of implementation and to
* pack the directory as much as possible. It however requires twice
* as much memory as the size of the directory. The maximum size
* --------------------------------------------------------
* ^ ptr ^ptr
* tail new head old
- *
+ *
* This is going to be a pain in the tuckus to implement, and will
* require a lot more disk accesses. So I'm going to skip it for now;
* it's only really going to be an issue for really, really big
* require that e2fsck use VM first.
*/
+#include <string.h>
+#include <ctype.h>
#include <errno.h>
#include "e2fsck.h"
#include "problem.h"
struct hash_entry {
ext2_dirhash_t hash;
ext2_dirhash_t minor_hash;
+ ino_t ino;
struct ext2_dir_entry *dir;
};
static int fill_dir_block(ext2_filsys fs,
blk_t *block_nr,
e2_blkcnt_t blockcnt,
- blk_t ref_block,
- int ref_offset,
+ blk_t ref_block EXT2FS_ATTR((unused)),
+ int ref_offset EXT2FS_ATTR((unused)),
void *priv_data)
{
struct fill_dir_struct *fd = (struct fill_dir_struct *) priv_data;
struct hash_entry *new_array, *ent;
struct ext2_dir_entry *dirent;
char *dir;
- int offset, dir_offset;
-
+ unsigned int offset, dir_offset;
+ int rec_len, hash_alg;
+
if (blockcnt < 0)
return 0;
if (fd->err)
return BLOCK_ABORT;
}
+ hash_alg = fs->super->s_def_hash_version;
+ if ((hash_alg <= EXT2_HASH_TEA) &&
+ (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
+ hash_alg += 3;
/* While the directory block is "hot", index it. */
dir_offset = 0;
while (dir_offset < fs->blocksize) {
dirent = (struct ext2_dir_entry *) (dir + dir_offset);
- if (((dir_offset + dirent->rec_len) > fs->blocksize) ||
- (dirent->rec_len < 8) ||
- ((dirent->rec_len % 4) != 0) ||
- (((dirent->name_len & 0xFF)+8) > dirent->rec_len)) {
+ rec_len = (dirent->rec_len || fs->blocksize < 65536) ?
+ dirent->rec_len : 65536;
+ if (((dir_offset + rec_len) > fs->blocksize) ||
+ (rec_len < 8) ||
+ ((rec_len % 4) != 0) ||
+ (((dirent->name_len & 0xFF)+8) > rec_len)) {
fd->err = EXT2_ET_DIR_CORRUPTED;
return BLOCK_ABORT;
}
- dir_offset += dirent->rec_len;
+ dir_offset += rec_len;
if (dirent->inode == 0)
continue;
if (!fd->compress && ((dirent->name_len&0xFF) == 1) &&
ent = fd->harray + fd->num_array++;
ent->dir = dirent;
fd->dir_size += EXT2_DIR_REC_LEN(dirent->name_len & 0xFF);
+ ent->ino = dirent->inode;
if (fd->compress)
ent->hash = ent->minor_hash = 0;
else {
- fd->err = ext2fs_dirhash(fs->super->s_def_hash_version,
- dirent->name,
+ fd->err = ext2fs_dirhash(hash_alg, dirent->name,
dirent->name_len & 0xFF,
fs->super->s_hash_seed,
&ent->hash, &ent->minor_hash);
return BLOCK_ABORT;
}
}
-
+
return 0;
}
/* Used for sorting the hash entry */
-static EXT2_QSORT_TYPE hash_cmp(const void *a, const void *b)
+static EXT2_QSORT_TYPE ino_cmp(const void *a, const void *b)
{
const struct hash_entry *he_a = (const struct hash_entry *) a;
const struct hash_entry *he_b = (const struct hash_entry *) b;
- int ret;
-
- if (he_a->hash > he_b->hash)
- ret = 1;
- else if (he_a->hash < he_b->hash)
- ret = -1;
- else {
- if (he_a->minor_hash > he_b->minor_hash)
- ret = 1;
- else if (he_a->minor_hash < he_b->minor_hash)
- ret = -1;
- else
- ret = 0;
- }
- return ret;
+
+ return (he_a->ino - he_b->ino);
}
/* Used for sorting the hash entry */
else if (he_a->dir->name_len < he_b->dir->name_len)
ret = -1;
else
- ret = 0;
+ ret = he_b->dir->inode - he_a->dir->inode;
+ }
+ return ret;
+}
+
+/* Used for sorting the hash entry */
+static EXT2_QSORT_TYPE hash_cmp(const void *a, const void *b)
+{
+ const struct hash_entry *he_a = (const struct hash_entry *) a;
+ const struct hash_entry *he_b = (const struct hash_entry *) b;
+ int ret;
+
+ if (he_a->hash > he_b->hash)
+ ret = 1;
+ else if (he_a->hash < he_b->hash)
+ ret = -1;
+ else {
+ if (he_a->minor_hash > he_b->minor_hash)
+ ret = 1;
+ else if (he_a->minor_hash < he_b->minor_hash)
+ ret = -1;
+ else
+ ret = name_cmp(a, b);
}
return ret;
}
-static errcode_t alloc_size_dir(ext2_filsys fs, struct out_dir *outdir,
+static errcode_t alloc_size_dir(ext2_filsys fs, struct out_dir *outdir,
int blocks)
{
void *new_mem;
static void free_out_dir(struct out_dir *outdir)
{
- free(outdir->buf);
- free(outdir->hashes);
+ if (outdir->buf)
+ free(outdir->buf);
+ if (outdir->hashes)
+ free(outdir->hashes);
outdir->max = 0;
outdir->num =0;
}
return 0;
}
+/*
+ * This function is used to make a unique filename. We do this by
+ * appending ~0, and then incrementing the number. However, we cannot
+ * expand the length of the filename beyond the padding available in
+ * the directory entry.
+ */
+static void mutate_name(char *str, __u16 *len)
+{
+ int i;
+ __u16 l = *len & 0xFF, h = *len & 0xff00;
+
+ /*
+ * First check to see if it looks the name has been mutated
+ * already
+ */
+ for (i = l-1; i > 0; i--) {
+ if (!isdigit(str[i]))
+ break;
+ }
+ if ((i == l-1) || (str[i] != '~')) {
+ if (((l-1) & 3) < 2)
+ l += 2;
+ else
+ l = (l+3) & ~3;
+ str[l-2] = '~';
+ str[l-1] = '0';
+ *len = l | h;
+ return;
+ }
+ for (i = l-1; i >= 0; i--) {
+ if (isdigit(str[i])) {
+ if (str[i] == '9')
+ str[i] = '0';
+ else {
+ str[i]++;
+ return;
+ }
+ continue;
+ }
+ if (i == 1) {
+ if (str[0] == 'z')
+ str[0] = 'A';
+ else if (str[0] == 'Z') {
+ str[0] = '~';
+ str[1] = '0';
+ } else
+ str[0]++;
+ } else if (i > 0) {
+ str[i] = '1';
+ str[i-1] = '~';
+ } else {
+ if (str[0] == '~')
+ str[0] = 'a';
+ else
+ str[0]++;
+ }
+ break;
+ }
+}
+
+static int duplicate_search_and_fix(e2fsck_t ctx, ext2_filsys fs,
+ ext2_ino_t ino,
+ struct fill_dir_struct *fd)
+{
+ struct problem_context pctx;
+ struct hash_entry *ent, *prev;
+ int i, j;
+ int fixed = 0;
+ char new_name[256];
+ __u16 new_len;
+ int hash_alg;
+
+ clear_problem_context(&pctx);
+ pctx.ino = ino;
+
+ hash_alg = fs->super->s_def_hash_version;
+ if ((hash_alg <= EXT2_HASH_TEA) &&
+ (fs->super->s_flags & EXT2_FLAGS_UNSIGNED_HASH))
+ hash_alg += 3;
+
+ for (i=1; i < fd->num_array; i++) {
+ ent = fd->harray + i;
+ prev = ent - 1;
+ if (!ent->dir->inode ||
+ ((ent->dir->name_len & 0xFF) !=
+ (prev->dir->name_len & 0xFF)) ||
+ (strncmp(ent->dir->name, prev->dir->name,
+ ent->dir->name_len & 0xFF)))
+ continue;
+ pctx.dirent = ent->dir;
+ if ((ent->dir->inode == prev->dir->inode) &&
+ fix_problem(ctx, PR_2_DUPLICATE_DIRENT, &pctx)) {
+ e2fsck_adjust_inode_count(ctx, ent->dir->inode, -1);
+ ent->dir->inode = 0;
+ fixed++;
+ continue;
+ }
+ memcpy(new_name, ent->dir->name, ent->dir->name_len & 0xFF);
+ new_len = ent->dir->name_len;
+ mutate_name(new_name, &new_len);
+ for (j=0; j < fd->num_array; j++) {
+ if ((i==j) ||
+ ((ent->dir->name_len & 0xFF) !=
+ (fd->harray[j].dir->name_len & 0xFF)) ||
+ (strncmp(new_name, fd->harray[j].dir->name,
+ new_len & 0xFF)))
+ continue;
+ mutate_name(new_name, &new_len);
+
+ j = -1;
+ }
+ new_name[new_len & 0xFF] = 0;
+ pctx.str = new_name;
+ if (fix_problem(ctx, PR_2_NON_UNIQUE_FILE, &pctx)) {
+ memcpy(ent->dir->name, new_name, new_len & 0xFF);
+ ent->dir->name_len = new_len;
+ ext2fs_dirhash(hash_alg, ent->dir->name,
+ ent->dir->name_len & 0xFF,
+ fs->super->s_hash_seed,
+ &ent->hash, &ent->minor_hash);
+ fixed++;
+ }
+ }
+ return fixed;
+}
+
static errcode_t copy_dir_entries(ext2_filsys fs,
struct fill_dir_struct *fd,
int i, rec_len, left;
ext2_dirhash_t prev_hash;
int offset;
-
+
outdir->max = 0;
retval = alloc_size_dir(fs, outdir,
(fd->dir_size / fs->blocksize) + 2);
left = fs->blocksize;
for (i=0; i < fd->num_array; i++) {
ent = fd->harray + i;
+ if (ent->dir->inode == 0)
+ continue;
rec_len = EXT2_DIR_REC_LEN(ent->dir->name_len & 0xFF);
if (rec_len > left) {
if (left)
if (fs->super->s_feature_incompat & EXT2_FEATURE_INCOMPAT_FILETYPE)
filetype = EXT2_FT_DIR << 8;
-
+
memset(buf, 0, fs->blocksize);
dir = (struct ext2_dir_entry *) buf;
dir->inode = ino;
dir->name[1] = '.';
dir->name_len = 2 | filetype;
dir->rec_len = fs->blocksize - 12;
-
+
root = (struct ext2_dx_root_info *) (buf+24);
root->reserved_zero = 0;
root->hash_version = fs->super->s_def_hash_version;
dir = (struct ext2_dir_entry *) buf;
dir->inode = 0;
dir->rec_len = fs->blocksize;
-
+
limits = (struct ext2_dx_countlimit *) (buf+8);
limits->limit = (fs->blocksize - 8) / sizeof(struct ext2_dx_entry);
limits->count = 0;
char * block_start;
int i, c1, c2, nblks;
int limit_offset, root_offset;
-
+
root_info = set_root_node(fs, outdir->buf, ino, parent);
root_offset = limit_offset = ((char *) root_info - outdir->buf) +
root_info->info_length;
return ENOSPC;
if (c2 == 0) {
if (limit)
- limit->limit = limit->count =
+ limit->limit = limit->count =
ext2fs_cpu_to_le16(limit->limit);
root = (struct ext2_dx_entry *)
(outdir->buf + root_offset);
static int write_dir_block(ext2_filsys fs,
blk_t *block_nr,
e2_blkcnt_t blockcnt,
- blk_t ref_block,
- int ref_offset,
+ blk_t ref_block EXT2FS_ATTR((unused)),
+ int ref_offset EXT2FS_ATTR((unused)),
void *priv_data)
{
struct write_dir_struct *wd = (struct write_dir_struct *) priv_data;
else
inode.i_flags |= EXT2_INDEX_FL;
inode.i_size = outdir->num * fs->blocksize;
- inode.i_blocks -= (fs->blocksize / 512) * wd.cleared;
+ ext2fs_iblk_sub_blocks(fs, &inode, wd.cleared);
e2fsck_write_inode(ctx, ino, &inode, "rehash_dir");
return 0;
char *dir_buf = 0;
struct fill_dir_struct fd;
struct out_dir outdir;
-
+
+ outdir.max = outdir.num = 0;
+ outdir.buf = 0;
+ outdir.hashes = 0;
e2fsck_read_inode(ctx, ino, &inode, "rehash_dir");
retval = ENOMEM;
#endif
/* Sort the list */
+resort:
if (fd.compress)
qsort(fd.harray+2, fd.num_array-2,
- sizeof(struct hash_entry), name_cmp);
+ sizeof(struct hash_entry), ino_cmp);
else
qsort(fd.harray, fd.num_array,
sizeof(struct hash_entry), hash_cmp);
/*
+ * Look for duplicates
+ */
+ if (duplicate_search_and_fix(ctx, fs, ino, &fd))
+ goto resort;
+
+ if (ctx->options & E2F_OPT_NO) {
+ retval = 0;
+ goto errout;
+ }
+
+ /*
* Copy the directory entries. In a htree directory these
* will become the leaf nodes.
*/
retval = copy_dir_entries(fs, &fd, &outdir);
if (retval)
goto errout;
-
+
free(dir_buf); dir_buf = 0;
if (!fd.compress) {
if (retval)
goto errout;
}
-
+
retval = write_directory(ctx, fs, &outdir, ino, fd.compress);
if (retval)
goto errout;
#endif
struct dir_info *dir;
ext2_u32_iterate iter;
+ struct dir_info_iter * dirinfo_iter = 0;
ext2_ino_t ino;
errcode_t retval;
- int i, all_dirs, dir_index, first = 1;
+ int cur, max, all_dirs, dir_index, first = 1;
#ifdef RESOURCE_TRACK
- init_resource_track(&rtrack);
+ init_resource_track(&rtrack, ctx->fs->io);
#endif
all_dirs = ctx->options & E2F_OPT_COMPRESS_DIRS;
return;
e2fsck_get_lost_and_found(ctx, 0);
-
+
clear_problem_context(&pctx);
dir_index = ctx->fs->super->s_feature_compat & EXT2_FEATURE_COMPAT_DIR_INDEX;
- if (all_dirs)
- i = 0;
- else {
- retval = ext2fs_u32_list_iterate_begin(ctx->dirs_to_hash,
+ cur = 0;
+ if (all_dirs) {
+ dirinfo_iter = e2fsck_dir_info_iter_begin(ctx);
+ max = e2fsck_get_num_dirinfo(ctx);
+ } else {
+ retval = ext2fs_u32_list_iterate_begin(ctx->dirs_to_hash,
&iter);
if (retval) {
pctx.errcode = retval;
fix_problem(ctx, PR_3A_OPTIMIZE_ITER, &pctx);
return;
}
+ max = ext2fs_u32_list_count(ctx->dirs_to_hash);
}
while (1) {
if (all_dirs) {
- if ((dir = e2fsck_dir_info_iter(ctx, &i)) == 0)
+ if ((dir = e2fsck_dir_info_iter(ctx,
+ dirinfo_iter)) == 0)
break;
ino = dir->ino;
} else {
fix_problem(ctx, PR_3A_PASS_HEADER, &pctx);
first = 0;
}
+#if 0
fix_problem(ctx, PR_3A_OPTIMIZE_DIR, &pctx);
+#endif
pctx.errcode = e2fsck_rehash_dir(ctx, ino);
if (pctx.errcode) {
end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
fix_problem(ctx, PR_3A_OPTIMIZE_DIR_ERR, &pctx);
}
+ if (ctx->progress && !ctx->progress_fd)
+ e2fsck_simple_progress(ctx, "Rebuilding directory",
+ 100.0 * (float) (++cur) / (float) max, ino);
}
end_problem_latch(ctx, PR_LATCH_OPTIMIZE_DIR);
- if (!all_dirs)
+ if (all_dirs)
+ e2fsck_dir_info_iter_end(ctx, dirinfo_iter);
+ else
ext2fs_u32_list_iterate_end(iter);
-
+
if (ctx->dirs_to_hash)
ext2fs_u32_list_free(ctx->dirs_to_hash);
ctx->dirs_to_hash = 0;
#ifdef RESOURCE_TRACK
if (ctx->options & E2F_OPT_TIME2) {
e2fsck_clear_progbar(ctx);
- print_resource_track("Pass 3A", &rtrack);
+ print_resource_track("Pass 3A", &rtrack, ctx->fs->io);
}
#endif
}
git://git.whamcloud.com / tools / e2fsprogs.git / blobdiff