Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ac27a0ec | 2 | /* |
617ba13b | 3 | * linux/fs/ext4/dir.c |
ac27a0ec DK |
4 | * |
5 | * Copyright (C) 1992, 1993, 1994, 1995 | |
6 | * Remy Card (card@masi.ibp.fr) | |
7 | * Laboratoire MASI - Institut Blaise Pascal | |
8 | * Universite Pierre et Marie Curie (Paris VI) | |
9 | * | |
10 | * from | |
11 | * | |
12 | * linux/fs/minix/dir.c | |
13 | * | |
14 | * Copyright (C) 1991, 1992 Linus Torvalds | |
15 | * | |
617ba13b | 16 | * ext4 directory handling functions |
ac27a0ec DK |
17 | * |
18 | * Big-endian to little-endian byte-swapping/bitmaps by | |
19 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
20 | * | |
21 | * Hash Tree Directory indexing (c) 2001 Daniel Phillips | |
22 | * | |
23 | */ | |
24 | ||
25 | #include <linux/fs.h> | |
ac27a0ec | 26 | #include <linux/buffer_head.h> |
ac27a0ec | 27 | #include <linux/slab.h> |
ee73f9a5 | 28 | #include <linux/iversion.h> |
b886ee3e | 29 | #include <linux/unicode.h> |
3dcf5451 | 30 | #include "ext4.h" |
65d165d9 | 31 | #include "xattr.h" |
ac27a0ec | 32 | |
725bebb2 | 33 | static int ext4_dx_readdir(struct file *, struct dir_context *); |
ac27a0ec | 34 | |
d1f5273e | 35 | /** |
c60990b3 TT |
36 | * is_dx_dir() - check if a directory is using htree indexing |
37 | * @inode: directory inode | |
38 | * | |
d1f5273e | 39 | * Check if the given dir-inode refers to an htree-indexed directory |
70261f56 | 40 | * (or a directory which could potentially get converted to use htree |
d1f5273e FY |
41 | * indexing). |
42 | * | |
43 | * Return 1 if it is a dx dir, 0 if not | |
44 | */ | |
45 | static int is_dx_dir(struct inode *inode) | |
46 | { | |
47 | struct super_block *sb = inode->i_sb; | |
48 | ||
e2b911c5 | 49 | if (ext4_has_feature_dir_index(inode->i_sb) && |
d1f5273e | 50 | ((ext4_test_inode_flag(inode, EXT4_INODE_INDEX)) || |
8af0f082 TM |
51 | ((inode->i_size >> sb->s_blocksize_bits) == 1) || |
52 | ext4_has_inline_data(inode))) | |
d1f5273e FY |
53 | return 1; |
54 | ||
55 | return 0; | |
56 | } | |
57 | ||
471fbbea DR |
58 | static bool is_fake_dir_entry(struct ext4_dir_entry_2 *de) |
59 | { | |
60 | /* Check if . or .. , or skip if namelen is 0 */ | |
61 | if ((de->name_len > 0) && (de->name_len <= 2) && (de->name[0] == '.') && | |
62 | (de->name[1] == '.' || de->name[1] == '\0')) | |
63 | return true; | |
64 | /* Check if this is a csum entry */ | |
65 | if (de->file_type == EXT4_FT_DIR_CSUM) | |
66 | return true; | |
67 | return false; | |
68 | } | |
69 | ||
cad3f007 TT |
70 | /* |
71 | * Return 0 if the directory entry is OK, and 1 if there is a problem | |
72 | * | |
73 | * Note: this is the opposite of what ext2 and ext3 historically returned... | |
65d165d9 TM |
74 | * |
75 | * bh passed here can be an inode block or a dir data block, depending | |
76 | * on the inode inline data flag. | |
cad3f007 | 77 | */ |
60fd4da3 | 78 | int __ext4_check_dir_entry(const char *function, unsigned int line, |
f7c21177 | 79 | struct inode *dir, struct file *filp, |
60fd4da3 | 80 | struct ext4_dir_entry_2 *de, |
226ba972 | 81 | struct buffer_head *bh, char *buf, int size, |
60fd4da3 | 82 | unsigned int offset) |
ac27a0ec | 83 | { |
af5bc92d | 84 | const char *error_msg = NULL; |
3d0518f4 WY |
85 | const int rlen = ext4_rec_len_from_disk(de->rec_len, |
86 | dir->i_sb->s_blocksize); | |
707d1a2f | 87 | const int next_offset = ((char *) de - buf) + rlen; |
471fbbea DR |
88 | bool fake = is_fake_dir_entry(de); |
89 | bool has_csum = ext4_has_metadata_csum(dir->i_sb); | |
ac27a0ec | 90 | |
471fbbea | 91 | if (unlikely(rlen < ext4_dir_rec_len(1, fake ? NULL : dir))) |
ac27a0ec | 92 | error_msg = "rec_len is smaller than minimal"; |
cad3f007 | 93 | else if (unlikely(rlen % 4 != 0)) |
ac27a0ec | 94 | error_msg = "rec_len % 4 != 0"; |
471fbbea DR |
95 | else if (unlikely(rlen < ext4_dir_rec_len(de->name_len, |
96 | fake ? NULL : dir))) | |
ac27a0ec | 97 | error_msg = "rec_len is too small for name_len"; |
707d1a2f | 98 | else if (unlikely(next_offset > size)) |
4d982e25 | 99 | error_msg = "directory entry overrun"; |
471fbbea DR |
100 | else if (unlikely(next_offset > size - ext4_dir_rec_len(1, |
101 | has_csum ? NULL : dir) && | |
707d1a2f | 102 | next_offset != size)) |
109ba779 | 103 | error_msg = "directory entry too close to block end"; |
cad3f007 TT |
104 | else if (unlikely(le32_to_cpu(de->inode) > |
105 | le32_to_cpu(EXT4_SB(dir->i_sb)->s_es->s_inodes_count))) | |
ac27a0ec | 106 | error_msg = "inode out of bounds"; |
cad3f007 TT |
107 | else |
108 | return 0; | |
109 | ||
f7c21177 | 110 | if (filp) |
4fda4003 | 111 | ext4_error_file(filp, function, line, bh->b_blocknr, |
4d982e25 | 112 | "bad entry in directory: %s - offset=%u, " |
471fbbea | 113 | "inode=%u, rec_len=%d, size=%d fake=%d", |
4d982e25 | 114 | error_msg, offset, le32_to_cpu(de->inode), |
471fbbea | 115 | rlen, size, fake); |
f7c21177 | 116 | else |
4fda4003 | 117 | ext4_error_inode(dir, function, line, bh->b_blocknr, |
4d982e25 | 118 | "bad entry in directory: %s - offset=%u, " |
471fbbea | 119 | "inode=%u, rec_len=%d, size=%d fake=%d", |
4d982e25 | 120 | error_msg, offset, le32_to_cpu(de->inode), |
471fbbea | 121 | rlen, size, fake); |
f7c21177 | 122 | |
cad3f007 | 123 | return 1; |
ac27a0ec DK |
124 | } |
125 | ||
725bebb2 | 126 | static int ext4_readdir(struct file *file, struct dir_context *ctx) |
ac27a0ec | 127 | { |
498e5f24 | 128 | unsigned int offset; |
aa13d5f6 | 129 | int i; |
617ba13b | 130 | struct ext4_dir_entry_2 *de; |
ac27a0ec | 131 | int err; |
725bebb2 | 132 | struct inode *inode = file_inode(file); |
d1f5273e | 133 | struct super_block *sb = inode->i_sb; |
44614711 | 134 | struct buffer_head *bh = NULL; |
a7550b30 | 135 | struct fscrypt_str fstr = FSTR_INIT(NULL, 0); |
ac27a0ec | 136 | |
ec0caa97 EB |
137 | err = fscrypt_prepare_readdir(inode); |
138 | if (err) | |
139 | return err; | |
28b4c263 | 140 | |
d1f5273e | 141 | if (is_dx_dir(inode)) { |
725bebb2 | 142 | err = ext4_dx_readdir(file, ctx); |
400086d7 | 143 | if (err != ERR_BAD_DX_DIR) |
725bebb2 | 144 | return err; |
400086d7 | 145 | |
48a34311 JK |
146 | /* Can we just clear INDEX flag to ignore htree information? */ |
147 | if (!ext4_has_metadata_csum(sb)) { | |
148 | /* | |
149 | * We don't set the inode dirty flag since it's not | |
150 | * critical that it gets flushed back to the disk. | |
151 | */ | |
152 | ext4_clear_inode_flag(inode, EXT4_INODE_INDEX); | |
153 | } | |
ac27a0ec | 154 | } |
8af0f082 TM |
155 | |
156 | if (ext4_has_inline_data(inode)) { | |
157 | int has_inline_data = 1; | |
44614711 | 158 | err = ext4_read_inline_dir(file, ctx, |
8af0f082 TM |
159 | &has_inline_data); |
160 | if (has_inline_data) | |
44614711 MH |
161 | return err; |
162 | } | |
163 | ||
592ddec7 | 164 | if (IS_ENCRYPTED(inode)) { |
8b10fe68 | 165 | err = fscrypt_fname_alloc_buffer(EXT4_NAME_LEN, &fstr); |
b7236e21 | 166 | if (err < 0) |
44614711 | 167 | return err; |
8af0f082 TM |
168 | } |
169 | ||
725bebb2 | 170 | while (ctx->pos < inode->i_size) { |
2ed88685 | 171 | struct ext4_map_blocks map; |
ac27a0ec | 172 | |
1f60fbe7 TT |
173 | if (fatal_signal_pending(current)) { |
174 | err = -ERESTARTSYS; | |
175 | goto errout; | |
176 | } | |
177 | cond_resched(); | |
4e19d6b6 | 178 | offset = ctx->pos & (sb->s_blocksize - 1); |
725bebb2 | 179 | map.m_lblk = ctx->pos >> EXT4_BLOCK_SIZE_BITS(sb); |
2ed88685 TT |
180 | map.m_len = 1; |
181 | err = ext4_map_blocks(NULL, inode, &map, 0); | |
4e19d6b6 TT |
182 | if (err == 0) { |
183 | /* m_len should never be zero but let's avoid | |
184 | * an infinite loop if it somehow is */ | |
185 | if (map.m_len == 0) | |
186 | map.m_len = 1; | |
187 | ctx->pos += map.m_len * sb->s_blocksize; | |
188 | continue; | |
189 | } | |
ac27a0ec | 190 | if (err > 0) { |
2ed88685 | 191 | pgoff_t index = map.m_pblk >> |
09cbfeaf | 192 | (PAGE_SHIFT - inode->i_blkbits); |
725bebb2 | 193 | if (!ra_has_index(&file->f_ra, index)) |
cf914a7d | 194 | page_cache_sync_readahead( |
dc7868fc | 195 | sb->s_bdev->bd_inode->i_mapping, |
725bebb2 | 196 | &file->f_ra, file, |
cf914a7d | 197 | index, 1); |
09cbfeaf | 198 | file->f_ra.prev_pos = (loff_t)index << PAGE_SHIFT; |
1c215028 | 199 | bh = ext4_bread(NULL, inode, map.m_lblk, 0); |
c906f38e KT |
200 | if (IS_ERR(bh)) { |
201 | err = PTR_ERR(bh); | |
202 | bh = NULL; | |
203 | goto errout; | |
204 | } | |
ac27a0ec DK |
205 | } |
206 | ||
ac27a0ec | 207 | if (!bh) { |
e6c40211 | 208 | /* corrupt size? Maybe no more blocks to read */ |
725bebb2 | 209 | if (ctx->pos > inode->i_blocks << 9) |
e6c40211 | 210 | break; |
725bebb2 | 211 | ctx->pos += sb->s_blocksize - offset; |
ac27a0ec DK |
212 | continue; |
213 | } | |
214 | ||
b0336e8d DW |
215 | /* Check the checksum */ |
216 | if (!buffer_verified(bh) && | |
f036adb3 | 217 | !ext4_dirblock_csum_verify(inode, bh)) { |
725bebb2 | 218 | EXT4_ERROR_FILE(file, 0, "directory fails checksum " |
b0336e8d | 219 | "at offset %llu", |
725bebb2 AV |
220 | (unsigned long long)ctx->pos); |
221 | ctx->pos += sb->s_blocksize - offset; | |
d5ac7773 | 222 | brelse(bh); |
44614711 | 223 | bh = NULL; |
b0336e8d DW |
224 | continue; |
225 | } | |
226 | set_buffer_verified(bh); | |
227 | ||
ac27a0ec DK |
228 | /* If the dir block has changed since the last call to |
229 | * readdir(2), then we might be pointing to an invalid | |
230 | * dirent right now. Scan from the start of the block | |
231 | * to make sure. */ | |
c472c07b | 232 | if (!inode_eq_iversion(inode, file->f_version)) { |
ac27a0ec | 233 | for (i = 0; i < sb->s_blocksize && i < offset; ) { |
617ba13b | 234 | de = (struct ext4_dir_entry_2 *) |
ac27a0ec DK |
235 | (bh->b_data + i); |
236 | /* It's too expensive to do a full | |
237 | * dirent test each time round this | |
238 | * loop, but we do have to test at | |
239 | * least that it is non-zero. A | |
240 | * failure will be detected in the | |
241 | * dirent test below. */ | |
3d0518f4 | 242 | if (ext4_rec_len_from_disk(de->rec_len, |
471fbbea DR |
243 | sb->s_blocksize) < ext4_dir_rec_len(1, |
244 | inode)) | |
ac27a0ec | 245 | break; |
3d0518f4 WY |
246 | i += ext4_rec_len_from_disk(de->rec_len, |
247 | sb->s_blocksize); | |
ac27a0ec DK |
248 | } |
249 | offset = i; | |
725bebb2 | 250 | ctx->pos = (ctx->pos & ~(sb->s_blocksize - 1)) |
ac27a0ec | 251 | | offset; |
ee73f9a5 | 252 | file->f_version = inode_query_iversion(inode); |
ac27a0ec DK |
253 | } |
254 | ||
725bebb2 | 255 | while (ctx->pos < inode->i_size |
ac27a0ec | 256 | && offset < sb->s_blocksize) { |
617ba13b | 257 | de = (struct ext4_dir_entry_2 *) (bh->b_data + offset); |
725bebb2 | 258 | if (ext4_check_dir_entry(inode, file, de, bh, |
226ba972 TM |
259 | bh->b_data, bh->b_size, |
260 | offset)) { | |
63f57933 | 261 | /* |
725bebb2 | 262 | * On error, skip to the next block |
63f57933 | 263 | */ |
725bebb2 | 264 | ctx->pos = (ctx->pos | |
ac27a0ec | 265 | (sb->s_blocksize - 1)) + 1; |
725bebb2 | 266 | break; |
ac27a0ec | 267 | } |
3d0518f4 WY |
268 | offset += ext4_rec_len_from_disk(de->rec_len, |
269 | sb->s_blocksize); | |
ac27a0ec | 270 | if (le32_to_cpu(de->inode)) { |
592ddec7 | 271 | if (!IS_ENCRYPTED(inode)) { |
44614711 MH |
272 | if (!dir_emit(ctx, de->name, |
273 | de->name_len, | |
274 | le32_to_cpu(de->inode), | |
275 | get_dtype(sb, de->file_type))) | |
276 | goto done; | |
277 | } else { | |
a7550b30 JK |
278 | int save_len = fstr.len; |
279 | struct fscrypt_str de_name = | |
280 | FSTR_INIT(de->name, | |
281 | de->name_len); | |
d2299590 | 282 | |
44614711 | 283 | /* Directory is encrypted */ |
a7550b30 | 284 | err = fscrypt_fname_disk_to_usr(inode, |
471fbbea DR |
285 | EXT4_DIRENT_HASH(de), |
286 | EXT4_DIRENT_MINOR_HASH(de), | |
287 | &de_name, &fstr); | |
ef1eb3aa | 288 | de_name = fstr; |
a7550b30 | 289 | fstr.len = save_len; |
ef1eb3aa | 290 | if (err) |
44614711 MH |
291 | goto errout; |
292 | if (!dir_emit(ctx, | |
ef1eb3aa | 293 | de_name.name, de_name.len, |
44614711 MH |
294 | le32_to_cpu(de->inode), |
295 | get_dtype(sb, de->file_type))) | |
296 | goto done; | |
725bebb2 | 297 | } |
ac27a0ec | 298 | } |
725bebb2 | 299 | ctx->pos += ext4_rec_len_from_disk(de->rec_len, |
3d0518f4 | 300 | sb->s_blocksize); |
ac27a0ec | 301 | } |
ae05327a | 302 | if ((ctx->pos < inode->i_size) && !dir_relax_shared(inode)) |
44614711 | 303 | goto done; |
af5bc92d | 304 | brelse(bh); |
44614711 | 305 | bh = NULL; |
ac27a0ec | 306 | } |
44614711 MH |
307 | done: |
308 | err = 0; | |
309 | errout: | |
a7550b30 | 310 | fscrypt_fname_free_buffer(&fstr); |
44614711 MH |
311 | brelse(bh); |
312 | return err; | |
ac27a0ec DK |
313 | } |
314 | ||
d1f5273e FY |
315 | static inline int is_32bit_api(void) |
316 | { | |
317 | #ifdef CONFIG_COMPAT | |
121cef8f | 318 | return in_compat_syscall(); |
d1f5273e FY |
319 | #else |
320 | return (BITS_PER_LONG == 32); | |
321 | #endif | |
322 | } | |
323 | ||
ac27a0ec DK |
324 | /* |
325 | * These functions convert from the major/minor hash to an f_pos | |
d1f5273e FY |
326 | * value for dx directories |
327 | * | |
328 | * Upper layer (for example NFS) should specify FMODE_32BITHASH or | |
329 | * FMODE_64BITHASH explicitly. On the other hand, we allow ext4 to be mounted | |
330 | * directly on both 32-bit and 64-bit nodes, under such case, neither | |
331 | * FMODE_32BITHASH nor FMODE_64BITHASH is specified. | |
332 | */ | |
333 | static inline loff_t hash2pos(struct file *filp, __u32 major, __u32 minor) | |
334 | { | |
335 | if ((filp->f_mode & FMODE_32BITHASH) || | |
336 | (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api())) | |
337 | return major >> 1; | |
338 | else | |
339 | return ((__u64)(major >> 1) << 32) | (__u64)minor; | |
340 | } | |
341 | ||
342 | static inline __u32 pos2maj_hash(struct file *filp, loff_t pos) | |
343 | { | |
344 | if ((filp->f_mode & FMODE_32BITHASH) || | |
345 | (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api())) | |
346 | return (pos << 1) & 0xffffffff; | |
347 | else | |
348 | return ((pos >> 32) << 1) & 0xffffffff; | |
349 | } | |
350 | ||
351 | static inline __u32 pos2min_hash(struct file *filp, loff_t pos) | |
352 | { | |
353 | if ((filp->f_mode & FMODE_32BITHASH) || | |
354 | (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api())) | |
355 | return 0; | |
356 | else | |
357 | return pos & 0xffffffff; | |
358 | } | |
359 | ||
360 | /* | |
361 | * Return 32- or 64-bit end-of-file for dx directories | |
362 | */ | |
363 | static inline loff_t ext4_get_htree_eof(struct file *filp) | |
364 | { | |
365 | if ((filp->f_mode & FMODE_32BITHASH) || | |
366 | (!(filp->f_mode & FMODE_64BITHASH) && is_32bit_api())) | |
367 | return EXT4_HTREE_EOF_32BIT; | |
368 | else | |
369 | return EXT4_HTREE_EOF_64BIT; | |
370 | } | |
371 | ||
372 | ||
373 | /* | |
ec7268ce ES |
374 | * ext4_dir_llseek() calls generic_file_llseek_size to handle htree |
375 | * directories, where the "offset" is in terms of the filename hash | |
376 | * value instead of the byte offset. | |
ac27a0ec | 377 | * |
ec7268ce ES |
378 | * Because we may return a 64-bit hash that is well beyond offset limits, |
379 | * we need to pass the max hash as the maximum allowable offset in | |
380 | * the htree directory case. | |
381 | * | |
382 | * For non-htree, ext4_llseek already chooses the proper max offset. | |
ac27a0ec | 383 | */ |
d4e43954 | 384 | static loff_t ext4_dir_llseek(struct file *file, loff_t offset, int whence) |
d1f5273e FY |
385 | { |
386 | struct inode *inode = file->f_mapping->host; | |
d1f5273e | 387 | int dx_dir = is_dx_dir(inode); |
e40ff213 | 388 | loff_t ret, htree_max = ext4_get_htree_eof(file); |
d1f5273e | 389 | |
ec7268ce | 390 | if (likely(dx_dir)) |
e40ff213 | 391 | ret = generic_file_llseek_size(file, offset, whence, |
ec7268ce ES |
392 | htree_max, htree_max); |
393 | else | |
e40ff213 TT |
394 | ret = ext4_llseek(file, offset, whence); |
395 | file->f_version = inode_peek_iversion(inode) - 1; | |
396 | return ret; | |
d1f5273e | 397 | } |
ac27a0ec DK |
398 | |
399 | /* | |
400 | * This structure holds the nodes of the red-black tree used to store | |
401 | * the directory entry in hash order. | |
402 | */ | |
403 | struct fname { | |
404 | __u32 hash; | |
405 | __u32 minor_hash; | |
406 | struct rb_node rb_hash; | |
407 | struct fname *next; | |
408 | __u32 inode; | |
409 | __u8 name_len; | |
410 | __u8 file_type; | |
e32ac245 | 411 | char name[]; |
ac27a0ec DK |
412 | }; |
413 | ||
414 | /* | |
fac88735 | 415 | * This function implements a non-recursive way of freeing all of the |
ac27a0ec DK |
416 | * nodes in the red-black tree. |
417 | */ | |
418 | static void free_rb_tree_fname(struct rb_root *root) | |
419 | { | |
d1866bd0 CS |
420 | struct fname *fname, *next; |
421 | ||
422 | rbtree_postorder_for_each_entry_safe(fname, next, root, rb_hash) | |
ac27a0ec | 423 | while (fname) { |
af5bc92d | 424 | struct fname *old = fname; |
ac27a0ec | 425 | fname = fname->next; |
af5bc92d | 426 | kfree(old); |
ac27a0ec | 427 | } |
d1866bd0 CS |
428 | |
429 | *root = RB_ROOT; | |
ac27a0ec DK |
430 | } |
431 | ||
432 | ||
d1f5273e FY |
433 | static struct dir_private_info *ext4_htree_create_dir_info(struct file *filp, |
434 | loff_t pos) | |
ac27a0ec DK |
435 | { |
436 | struct dir_private_info *p; | |
437 | ||
d695a1be | 438 | p = kzalloc(sizeof(*p), GFP_KERNEL); |
ac27a0ec DK |
439 | if (!p) |
440 | return NULL; | |
d1f5273e FY |
441 | p->curr_hash = pos2maj_hash(filp, pos); |
442 | p->curr_minor_hash = pos2min_hash(filp, pos); | |
ac27a0ec DK |
443 | return p; |
444 | } | |
445 | ||
617ba13b | 446 | void ext4_htree_free_dir_info(struct dir_private_info *p) |
ac27a0ec DK |
447 | { |
448 | free_rb_tree_fname(&p->root); | |
449 | kfree(p); | |
450 | } | |
451 | ||
452 | /* | |
453 | * Given a directory entry, enter it into the fname rb tree. | |
2f61830a TT |
454 | * |
455 | * When filename encryption is enabled, the dirent will hold the | |
456 | * encrypted filename, while the htree will hold decrypted filename. | |
457 | * The decrypted filename is passed in via ent_name. parameter. | |
ac27a0ec | 458 | */ |
617ba13b | 459 | int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, |
ac27a0ec | 460 | __u32 minor_hash, |
2f61830a | 461 | struct ext4_dir_entry_2 *dirent, |
a7550b30 | 462 | struct fscrypt_str *ent_name) |
ac27a0ec DK |
463 | { |
464 | struct rb_node **p, *parent = NULL; | |
af5bc92d | 465 | struct fname *fname, *new_fn; |
ac27a0ec DK |
466 | struct dir_private_info *info; |
467 | int len; | |
468 | ||
a271fe85 | 469 | info = dir_file->private_data; |
ac27a0ec DK |
470 | p = &info->root.rb_node; |
471 | ||
472 | /* Create and allocate the fname structure */ | |
2f61830a | 473 | len = sizeof(struct fname) + ent_name->len + 1; |
ac27a0ec DK |
474 | new_fn = kzalloc(len, GFP_KERNEL); |
475 | if (!new_fn) | |
476 | return -ENOMEM; | |
477 | new_fn->hash = hash; | |
478 | new_fn->minor_hash = minor_hash; | |
479 | new_fn->inode = le32_to_cpu(dirent->inode); | |
2f61830a | 480 | new_fn->name_len = ent_name->len; |
ac27a0ec | 481 | new_fn->file_type = dirent->file_type; |
2f61830a | 482 | memcpy(new_fn->name, ent_name->name, ent_name->len); |
ac27a0ec DK |
483 | |
484 | while (*p) { | |
485 | parent = *p; | |
486 | fname = rb_entry(parent, struct fname, rb_hash); | |
487 | ||
488 | /* | |
489 | * If the hash and minor hash match up, then we put | |
490 | * them on a linked list. This rarely happens... | |
491 | */ | |
492 | if ((new_fn->hash == fname->hash) && | |
493 | (new_fn->minor_hash == fname->minor_hash)) { | |
494 | new_fn->next = fname->next; | |
495 | fname->next = new_fn; | |
496 | return 0; | |
497 | } | |
498 | ||
499 | if (new_fn->hash < fname->hash) | |
500 | p = &(*p)->rb_left; | |
501 | else if (new_fn->hash > fname->hash) | |
502 | p = &(*p)->rb_right; | |
503 | else if (new_fn->minor_hash < fname->minor_hash) | |
504 | p = &(*p)->rb_left; | |
505 | else /* if (new_fn->minor_hash > fname->minor_hash) */ | |
506 | p = &(*p)->rb_right; | |
507 | } | |
508 | ||
509 | rb_link_node(&new_fn->rb_hash, parent, p); | |
510 | rb_insert_color(&new_fn->rb_hash, &info->root); | |
511 | return 0; | |
512 | } | |
513 | ||
514 | ||
515 | ||
516 | /* | |
617ba13b | 517 | * This is a helper function for ext4_dx_readdir. It calls filldir |
fac88735 | 518 | * for all entries on the fname linked list. (Normally there is only |
ac27a0ec DK |
519 | * one entry on the linked list, unless there are 62 bit hash collisions.) |
520 | */ | |
725bebb2 AV |
521 | static int call_filldir(struct file *file, struct dir_context *ctx, |
522 | struct fname *fname) | |
ac27a0ec | 523 | { |
725bebb2 AV |
524 | struct dir_private_info *info = file->private_data; |
525 | struct inode *inode = file_inode(file); | |
526 | struct super_block *sb = inode->i_sb; | |
ac27a0ec DK |
527 | |
528 | if (!fname) { | |
92b97816 TT |
529 | ext4_msg(sb, KERN_ERR, "%s:%d: inode #%lu: comm %s: " |
530 | "called with null fname?!?", __func__, __LINE__, | |
531 | inode->i_ino, current->comm); | |
ac27a0ec DK |
532 | return 0; |
533 | } | |
725bebb2 | 534 | ctx->pos = hash2pos(file, fname->hash, fname->minor_hash); |
ac27a0ec | 535 | while (fname) { |
725bebb2 AV |
536 | if (!dir_emit(ctx, fname->name, |
537 | fname->name_len, | |
ac27a0ec | 538 | fname->inode, |
725bebb2 | 539 | get_dtype(sb, fname->file_type))) { |
d0156417 | 540 | info->extra_fname = fname; |
725bebb2 | 541 | return 1; |
ac27a0ec DK |
542 | } |
543 | fname = fname->next; | |
544 | } | |
545 | return 0; | |
546 | } | |
547 | ||
725bebb2 | 548 | static int ext4_dx_readdir(struct file *file, struct dir_context *ctx) |
ac27a0ec | 549 | { |
725bebb2 AV |
550 | struct dir_private_info *info = file->private_data; |
551 | struct inode *inode = file_inode(file); | |
ac27a0ec | 552 | struct fname *fname; |
42cb4474 | 553 | int ret = 0; |
ac27a0ec DK |
554 | |
555 | if (!info) { | |
725bebb2 | 556 | info = ext4_htree_create_dir_info(file, ctx->pos); |
ac27a0ec DK |
557 | if (!info) |
558 | return -ENOMEM; | |
725bebb2 | 559 | file->private_data = info; |
ac27a0ec DK |
560 | } |
561 | ||
725bebb2 | 562 | if (ctx->pos == ext4_get_htree_eof(file)) |
ac27a0ec DK |
563 | return 0; /* EOF */ |
564 | ||
565 | /* Some one has messed with f_pos; reset the world */ | |
725bebb2 | 566 | if (info->last_pos != ctx->pos) { |
ac27a0ec DK |
567 | free_rb_tree_fname(&info->root); |
568 | info->curr_node = NULL; | |
569 | info->extra_fname = NULL; | |
725bebb2 AV |
570 | info->curr_hash = pos2maj_hash(file, ctx->pos); |
571 | info->curr_minor_hash = pos2min_hash(file, ctx->pos); | |
ac27a0ec DK |
572 | } |
573 | ||
574 | /* | |
575 | * If there are any leftover names on the hash collision | |
576 | * chain, return them first. | |
577 | */ | |
d0156417 | 578 | if (info->extra_fname) { |
725bebb2 | 579 | if (call_filldir(file, ctx, info->extra_fname)) |
d0156417 | 580 | goto finished; |
d0156417 | 581 | info->extra_fname = NULL; |
3c37fc86 | 582 | goto next_node; |
d0156417 | 583 | } else if (!info->curr_node) |
ac27a0ec DK |
584 | info->curr_node = rb_first(&info->root); |
585 | ||
586 | while (1) { | |
587 | /* | |
588 | * Fill the rbtree if we have no more entries, | |
589 | * or the inode has changed since we last read in the | |
590 | * cached entries. | |
591 | */ | |
592 | if ((!info->curr_node) || | |
c472c07b | 593 | !inode_eq_iversion(inode, file->f_version)) { |
ac27a0ec DK |
594 | info->curr_node = NULL; |
595 | free_rb_tree_fname(&info->root); | |
ee73f9a5 | 596 | file->f_version = inode_query_iversion(inode); |
725bebb2 | 597 | ret = ext4_htree_fill_tree(file, info->curr_hash, |
ac27a0ec DK |
598 | info->curr_minor_hash, |
599 | &info->next_hash); | |
600 | if (ret < 0) | |
42cb4474 | 601 | goto finished; |
ac27a0ec | 602 | if (ret == 0) { |
725bebb2 | 603 | ctx->pos = ext4_get_htree_eof(file); |
ac27a0ec DK |
604 | break; |
605 | } | |
606 | info->curr_node = rb_first(&info->root); | |
607 | } | |
608 | ||
609 | fname = rb_entry(info->curr_node, struct fname, rb_hash); | |
610 | info->curr_hash = fname->hash; | |
611 | info->curr_minor_hash = fname->minor_hash; | |
725bebb2 | 612 | if (call_filldir(file, ctx, fname)) |
ac27a0ec | 613 | break; |
3c37fc86 | 614 | next_node: |
ac27a0ec | 615 | info->curr_node = rb_next(info->curr_node); |
3c37fc86 TT |
616 | if (info->curr_node) { |
617 | fname = rb_entry(info->curr_node, struct fname, | |
618 | rb_hash); | |
619 | info->curr_hash = fname->hash; | |
620 | info->curr_minor_hash = fname->minor_hash; | |
621 | } else { | |
ac27a0ec | 622 | if (info->next_hash == ~0) { |
725bebb2 | 623 | ctx->pos = ext4_get_htree_eof(file); |
ac27a0ec DK |
624 | break; |
625 | } | |
626 | info->curr_hash = info->next_hash; | |
627 | info->curr_minor_hash = 0; | |
628 | } | |
629 | } | |
630 | finished: | |
725bebb2 | 631 | info->last_pos = ctx->pos; |
42cb4474 | 632 | return ret < 0 ? ret : 0; |
ac27a0ec DK |
633 | } |
634 | ||
af5bc92d | 635 | static int ext4_release_dir(struct inode *inode, struct file *filp) |
ac27a0ec | 636 | { |
63f57933 | 637 | if (filp->private_data) |
617ba13b | 638 | ext4_htree_free_dir_info(filp->private_data); |
ac27a0ec DK |
639 | |
640 | return 0; | |
641 | } | |
d1f5273e | 642 | |
40b163f1 DW |
643 | int ext4_check_all_de(struct inode *dir, struct buffer_head *bh, void *buf, |
644 | int buf_size) | |
645 | { | |
646 | struct ext4_dir_entry_2 *de; | |
18017479 | 647 | int rlen; |
40b163f1 DW |
648 | unsigned int offset = 0; |
649 | char *top; | |
650 | ||
c30365b9 | 651 | de = buf; |
40b163f1 DW |
652 | top = buf + buf_size; |
653 | while ((char *) de < top) { | |
654 | if (ext4_check_dir_entry(dir, NULL, de, bh, | |
655 | buf, buf_size, offset)) | |
6a797d27 | 656 | return -EFSCORRUPTED; |
40b163f1 DW |
657 | rlen = ext4_rec_len_from_disk(de->rec_len, buf_size); |
658 | de = (struct ext4_dir_entry_2 *)((char *)de + rlen); | |
659 | offset += rlen; | |
660 | } | |
661 | if ((char *) de > top) | |
6a797d27 | 662 | return -EFSCORRUPTED; |
40b163f1 DW |
663 | |
664 | return 0; | |
665 | } | |
666 | ||
d1f5273e FY |
667 | const struct file_operations ext4_dir_operations = { |
668 | .llseek = ext4_dir_llseek, | |
669 | .read = generic_read_dir, | |
ae05327a | 670 | .iterate_shared = ext4_readdir, |
d1f5273e FY |
671 | .unlocked_ioctl = ext4_ioctl, |
672 | #ifdef CONFIG_COMPAT | |
673 | .compat_ioctl = ext4_compat_ioctl, | |
674 | #endif | |
675 | .fsync = ext4_sync_file, | |
676 | .release = ext4_release_dir, | |
677 | }; |