4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
11 * linux/fs/minix/dir.c
13 * Copyright (C) 1991, 1992 Linus Torvalds
15 * ext4 directory handling functions
17 * Big-endian to little-endian byte-swapping/bitmaps by
18 * David S. Miller (davem@caip.rutgers.edu), 1995
20 * Hash Tree Directory indexing (c) 2001 Daniel Phillips
25 #include <linux/jbd2.h>
26 #include <linux/buffer_head.h>
27 #include <linux/slab.h>
28 #include <linux/rbtree.h>
31 static unsigned char ext4_filetype_table[] = {
32 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
35 static int ext4_readdir(struct file *, void *, filldir_t);
36 static int ext4_dx_readdir(struct file *filp,
37 void *dirent, filldir_t filldir);
38 static int ext4_release_dir(struct inode *inode,
41 const struct file_operations ext4_dir_operations = {
42 .llseek = generic_file_llseek,
43 .read = generic_read_dir,
44 .readdir = ext4_readdir, /* we take BKL. needed?*/
45 .unlocked_ioctl = ext4_ioctl,
47 .compat_ioctl = ext4_compat_ioctl,
49 .fsync = ext4_sync_file,
50 .release = ext4_release_dir,
54 static unsigned char get_dtype(struct super_block *sb, int filetype)
56 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE) ||
57 (filetype >= EXT4_FT_MAX))
60 return (ext4_filetype_table[filetype]);
64 int ext4_check_dir_entry(const char *function, struct inode *dir,
65 struct ext4_dir_entry_2 *de,
66 struct buffer_head *bh,
69 const char *error_msg = NULL;
70 const int rlen = ext4_rec_len_from_disk(de->rec_len);
72 if (rlen < EXT4_DIR_REC_LEN(1))
73 error_msg = "rec_len is smaller than minimal";
74 else if (rlen % 4 != 0)
75 error_msg = "rec_len % 4 != 0";
76 else if (rlen < EXT4_DIR_REC_LEN(de->name_len))
77 error_msg = "rec_len is too small for name_len";
78 else if (((char *) de - bh->b_data) + rlen > dir->i_sb->s_blocksize)
79 error_msg = "directory entry across blocks";
80 else if (le32_to_cpu(de->inode) >
81 le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count))
82 error_msg = "inode out of bounds";
84 if (error_msg != NULL)
85 ext4_error(dir->i_sb, function,
86 "bad entry in directory #%lu: %s - "
87 "offset=%lu, inode=%lu, rec_len=%d, name_len=%d",
88 dir->i_ino, error_msg, offset,
89 (unsigned long) le32_to_cpu(de->inode),
91 return error_msg == NULL ? 1 : 0;
94 static int ext4_readdir(struct file *filp,
95 void *dirent, filldir_t filldir)
100 struct ext4_dir_entry_2 *de;
101 struct super_block *sb;
103 struct inode *inode = filp->f_path.dentry->d_inode;
105 int dir_has_error = 0;
109 if (EXT4_HAS_COMPAT_FEATURE(inode->i_sb,
110 EXT4_FEATURE_COMPAT_DIR_INDEX) &&
111 ((EXT4_I(inode)->i_flags & EXT4_INDEX_FL) ||
112 ((inode->i_size >> sb->s_blocksize_bits) == 1))) {
113 err = ext4_dx_readdir(filp, dirent, filldir);
114 if (err != ERR_BAD_DX_DIR) {
119 * We don't set the inode dirty flag since it's not
120 * critical that it get flushed back to the disk.
122 EXT4_I(filp->f_path.dentry->d_inode)->i_flags &= ~EXT4_INDEX_FL;
125 offset = filp->f_pos & (sb->s_blocksize - 1);
127 while (!error && !stored && filp->f_pos < inode->i_size) {
128 ext4_lblk_t blk = filp->f_pos >> EXT4_BLOCK_SIZE_BITS(sb);
129 struct buffer_head map_bh;
130 struct buffer_head *bh = NULL;
133 err = ext4_get_blocks_wrap(NULL, inode, blk, 1, &map_bh,
136 pgoff_t index = map_bh.b_blocknr >>
137 (PAGE_CACHE_SHIFT - inode->i_blkbits);
138 if (!ra_has_index(&filp->f_ra, index))
139 page_cache_sync_readahead(
140 sb->s_bdev->bd_inode->i_mapping,
143 filp->f_ra.prev_pos = (loff_t)index << PAGE_CACHE_SHIFT;
144 bh = ext4_bread(NULL, inode, blk, 0, &err);
148 * We ignore I/O errors on directories so users have a chance
149 * of recovering data when there's a bad sector
152 if (!dir_has_error) {
153 ext4_error(sb, __func__, "directory #%lu "
154 "contains a hole at offset %Lu",
156 (unsigned long long) filp->f_pos);
159 /* corrupt size? Maybe no more blocks to read */
160 if (filp->f_pos > inode->i_blocks << 9)
162 filp->f_pos += sb->s_blocksize - offset;
167 /* If the dir block has changed since the last call to
168 * readdir(2), then we might be pointing to an invalid
169 * dirent right now. Scan from the start of the block
171 if (filp->f_version != inode->i_version) {
172 for (i = 0; i < sb->s_blocksize && i < offset; ) {
173 de = (struct ext4_dir_entry_2 *)
175 /* It's too expensive to do a full
176 * dirent test each time round this
177 * loop, but we do have to test at
178 * least that it is non-zero. A
179 * failure will be detected in the
180 * dirent test below. */
181 if (ext4_rec_len_from_disk(de->rec_len)
182 < EXT4_DIR_REC_LEN(1))
184 i += ext4_rec_len_from_disk(de->rec_len);
187 filp->f_pos = (filp->f_pos & ~(sb->s_blocksize - 1))
189 filp->f_version = inode->i_version;
192 while (!error && filp->f_pos < inode->i_size
193 && offset < sb->s_blocksize) {
194 de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
195 if (!ext4_check_dir_entry("ext4_readdir", inode, de,
198 * On error, skip the f_pos to the next block
200 filp->f_pos = (filp->f_pos |
201 (sb->s_blocksize - 1)) + 1;
206 offset += ext4_rec_len_from_disk(de->rec_len);
207 if (le32_to_cpu(de->inode)) {
208 /* We might block in the next section
209 * if the data destination is
210 * currently swapped out. So, use a
211 * version stamp to detect whether or
212 * not the directory has been modified
213 * during the copy operation.
215 u64 version = filp->f_version;
217 error = filldir(dirent, de->name,
220 le32_to_cpu(de->inode),
221 get_dtype(sb, de->file_type));
224 if (version != filp->f_version)
228 filp->f_pos += ext4_rec_len_from_disk(de->rec_len);
238 * These functions convert from the major/minor hash to an f_pos
241 * Currently we only use major hash numer. This is unfortunate, but
242 * on 32-bit machines, the same VFS interface is used for lseek and
243 * llseek, so if we use the 64 bit offset, then the 32-bit versions of
244 * lseek/telldir/seekdir will blow out spectacularly, and from within
245 * the ext2 low-level routine, we don't know if we're being called by
246 * a 64-bit version of the system call or the 32-bit version of the
247 * system call. Worse yet, NFSv2 only allows for a 32-bit readdir
250 #define hash2pos(major, minor) (major >> 1)
251 #define pos2maj_hash(pos) ((pos << 1) & 0xffffffff)
252 #define pos2min_hash(pos) (0)
255 * This structure holds the nodes of the red-black tree used to store
256 * the directory entry in hash order.
261 struct rb_node rb_hash;
270 * This functoin implements a non-recursive way of freeing all of the
271 * nodes in the red-black tree.
273 static void free_rb_tree_fname(struct rb_root *root)
275 struct rb_node *n = root->rb_node;
276 struct rb_node *parent;
280 /* Do the node's children first */
290 * The node has no children; free it, and then zero
291 * out parent's link to it. Finally go to the
292 * beginning of the loop and try to free the parent
295 parent = rb_parent(n);
296 fname = rb_entry(n, struct fname, rb_hash);
298 struct fname *old = fname;
303 root->rb_node = NULL;
304 else if (parent->rb_left == n)
305 parent->rb_left = NULL;
306 else if (parent->rb_right == n)
307 parent->rb_right = NULL;
313 static struct dir_private_info *ext4_htree_create_dir_info(loff_t pos)
315 struct dir_private_info *p;
317 p = kzalloc(sizeof(struct dir_private_info), GFP_KERNEL);
320 p->curr_hash = pos2maj_hash(pos);
321 p->curr_minor_hash = pos2min_hash(pos);
325 void ext4_htree_free_dir_info(struct dir_private_info *p)
327 free_rb_tree_fname(&p->root);
332 * Given a directory entry, enter it into the fname rb tree.
334 int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
336 struct ext4_dir_entry_2 *dirent)
338 struct rb_node **p, *parent = NULL;
339 struct fname *fname, *new_fn;
340 struct dir_private_info *info;
343 info = (struct dir_private_info *) dir_file->private_data;
344 p = &info->root.rb_node;
346 /* Create and allocate the fname structure */
347 len = sizeof(struct fname) + dirent->name_len + 1;
348 new_fn = kzalloc(len, GFP_KERNEL);
352 new_fn->minor_hash = minor_hash;
353 new_fn->inode = le32_to_cpu(dirent->inode);
354 new_fn->name_len = dirent->name_len;
355 new_fn->file_type = dirent->file_type;
356 memcpy(new_fn->name, dirent->name, dirent->name_len);
357 new_fn->name[dirent->name_len] = 0;
361 fname = rb_entry(parent, struct fname, rb_hash);
364 * If the hash and minor hash match up, then we put
365 * them on a linked list. This rarely happens...
367 if ((new_fn->hash == fname->hash) &&
368 (new_fn->minor_hash == fname->minor_hash)) {
369 new_fn->next = fname->next;
370 fname->next = new_fn;
374 if (new_fn->hash < fname->hash)
376 else if (new_fn->hash > fname->hash)
378 else if (new_fn->minor_hash < fname->minor_hash)
380 else /* if (new_fn->minor_hash > fname->minor_hash) */
384 rb_link_node(&new_fn->rb_hash, parent, p);
385 rb_insert_color(&new_fn->rb_hash, &info->root);
392 * This is a helper function for ext4_dx_readdir. It calls filldir
393 * for all entres on the fname linked list. (Normally there is only
394 * one entry on the linked list, unless there are 62 bit hash collisions.)
396 static int call_filldir(struct file *filp, void *dirent,
397 filldir_t filldir, struct fname *fname)
399 struct dir_private_info *info = filp->private_data;
401 struct inode *inode = filp->f_path.dentry->d_inode;
402 struct super_block *sb;
408 printk(KERN_ERR "ext4: call_filldir: called with "
412 curr_pos = hash2pos(fname->hash, fname->minor_hash);
414 error = filldir(dirent, fname->name,
415 fname->name_len, curr_pos,
417 get_dtype(sb, fname->file_type));
419 filp->f_pos = curr_pos;
420 info->extra_fname = fname;
428 static int ext4_dx_readdir(struct file *filp,
429 void *dirent, filldir_t filldir)
431 struct dir_private_info *info = filp->private_data;
432 struct inode *inode = filp->f_path.dentry->d_inode;
437 info = ext4_htree_create_dir_info(filp->f_pos);
440 filp->private_data = info;
443 if (filp->f_pos == EXT4_HTREE_EOF)
446 /* Some one has messed with f_pos; reset the world */
447 if (info->last_pos != filp->f_pos) {
448 free_rb_tree_fname(&info->root);
449 info->curr_node = NULL;
450 info->extra_fname = NULL;
451 info->curr_hash = pos2maj_hash(filp->f_pos);
452 info->curr_minor_hash = pos2min_hash(filp->f_pos);
456 * If there are any leftover names on the hash collision
457 * chain, return them first.
459 if (info->extra_fname) {
460 if (call_filldir(filp, dirent, filldir, info->extra_fname))
462 info->extra_fname = NULL;
464 } else if (!info->curr_node)
465 info->curr_node = rb_first(&info->root);
469 * Fill the rbtree if we have no more entries,
470 * or the inode has changed since we last read in the
473 if ((!info->curr_node) ||
474 (filp->f_version != inode->i_version)) {
475 info->curr_node = NULL;
476 free_rb_tree_fname(&info->root);
477 filp->f_version = inode->i_version;
478 ret = ext4_htree_fill_tree(filp, info->curr_hash,
479 info->curr_minor_hash,
484 filp->f_pos = EXT4_HTREE_EOF;
487 info->curr_node = rb_first(&info->root);
490 fname = rb_entry(info->curr_node, struct fname, rb_hash);
491 info->curr_hash = fname->hash;
492 info->curr_minor_hash = fname->minor_hash;
493 if (call_filldir(filp, dirent, filldir, fname))
496 info->curr_node = rb_next(info->curr_node);
497 if (info->curr_node) {
498 fname = rb_entry(info->curr_node, struct fname,
500 info->curr_hash = fname->hash;
501 info->curr_minor_hash = fname->minor_hash;
503 if (info->next_hash == ~0) {
504 filp->f_pos = EXT4_HTREE_EOF;
507 info->curr_hash = info->next_hash;
508 info->curr_minor_hash = 0;
512 info->last_pos = filp->f_pos;
516 static int ext4_release_dir(struct inode *inode, struct file *filp)
518 if (filp->private_data)
519 ext4_htree_free_dir_info(filp->private_data);