]>
Commit | Line | Data |
---|---|---|
ac27a0ec | 1 | /* |
617ba13b | 2 | * linux/fs/ext4/namei.c |
ac27a0ec DK |
3 | * |
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) | |
8 | * | |
9 | * from | |
10 | * | |
11 | * linux/fs/minix/namei.c | |
12 | * | |
13 | * Copyright (C) 1991, 1992 Linus Torvalds | |
14 | * | |
15 | * Big-endian to little-endian byte-swapping/bitmaps by | |
16 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
17 | * Directory entry file type support and forward compatibility hooks | |
18 | * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 | |
19 | * Hash Tree Directory indexing (c) | |
20 | * Daniel Phillips, 2001 | |
21 | * Hash Tree Directory indexing porting | |
22 | * Christopher Li, 2002 | |
23 | * Hash Tree Directory indexing cleanup | |
24 | * Theodore Ts'o, 2002 | |
25 | */ | |
26 | ||
27 | #include <linux/fs.h> | |
28 | #include <linux/pagemap.h> | |
dab291af | 29 | #include <linux/jbd2.h> |
ac27a0ec | 30 | #include <linux/time.h> |
ac27a0ec DK |
31 | #include <linux/fcntl.h> |
32 | #include <linux/stat.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/quotaops.h> | |
35 | #include <linux/buffer_head.h> | |
36 | #include <linux/bio.h> | |
3dcf5451 CH |
37 | #include "ext4.h" |
38 | #include "ext4_jbd2.h" | |
ac27a0ec | 39 | |
ac27a0ec DK |
40 | #include "xattr.h" |
41 | #include "acl.h" | |
42 | ||
0562e0ba | 43 | #include <trace/events/ext4.h> |
ac27a0ec DK |
44 | /* |
45 | * define how far ahead to read directories while searching them. | |
46 | */ | |
47 | #define NAMEI_RA_CHUNKS 2 | |
48 | #define NAMEI_RA_BLOCKS 4 | |
8c55e204 | 49 | #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) |
ac27a0ec DK |
50 | #define NAMEI_RA_INDEX(c,b) (((c) * NAMEI_RA_BLOCKS) + (b)) |
51 | ||
617ba13b | 52 | static struct buffer_head *ext4_append(handle_t *handle, |
ac27a0ec | 53 | struct inode *inode, |
725d26d3 | 54 | ext4_lblk_t *block, int *err) |
ac27a0ec DK |
55 | { |
56 | struct buffer_head *bh; | |
57 | ||
58 | *block = inode->i_size >> inode->i_sb->s_blocksize_bits; | |
59 | ||
a871611b AM |
60 | bh = ext4_bread(handle, inode, *block, 1, err); |
61 | if (bh) { | |
ac27a0ec | 62 | inode->i_size += inode->i_sb->s_blocksize; |
617ba13b | 63 | EXT4_I(inode)->i_disksize = inode->i_size; |
a871611b AM |
64 | *err = ext4_journal_get_write_access(handle, bh); |
65 | if (*err) { | |
66 | brelse(bh); | |
67 | bh = NULL; | |
68 | } | |
ac27a0ec DK |
69 | } |
70 | return bh; | |
71 | } | |
72 | ||
73 | #ifndef assert | |
74 | #define assert(test) J_ASSERT(test) | |
75 | #endif | |
76 | ||
ac27a0ec DK |
77 | #ifdef DX_DEBUG |
78 | #define dxtrace(command) command | |
79 | #else | |
80 | #define dxtrace(command) | |
81 | #endif | |
82 | ||
83 | struct fake_dirent | |
84 | { | |
85 | __le32 inode; | |
86 | __le16 rec_len; | |
87 | u8 name_len; | |
88 | u8 file_type; | |
89 | }; | |
90 | ||
91 | struct dx_countlimit | |
92 | { | |
93 | __le16 limit; | |
94 | __le16 count; | |
95 | }; | |
96 | ||
97 | struct dx_entry | |
98 | { | |
99 | __le32 hash; | |
100 | __le32 block; | |
101 | }; | |
102 | ||
103 | /* | |
104 | * dx_root_info is laid out so that if it should somehow get overlaid by a | |
105 | * dirent the two low bits of the hash version will be zero. Therefore, the | |
106 | * hash version mod 4 should never be 0. Sincerely, the paranoia department. | |
107 | */ | |
108 | ||
109 | struct dx_root | |
110 | { | |
111 | struct fake_dirent dot; | |
112 | char dot_name[4]; | |
113 | struct fake_dirent dotdot; | |
114 | char dotdot_name[4]; | |
115 | struct dx_root_info | |
116 | { | |
117 | __le32 reserved_zero; | |
118 | u8 hash_version; | |
119 | u8 info_length; /* 8 */ | |
120 | u8 indirect_levels; | |
121 | u8 unused_flags; | |
122 | } | |
123 | info; | |
124 | struct dx_entry entries[0]; | |
125 | }; | |
126 | ||
127 | struct dx_node | |
128 | { | |
129 | struct fake_dirent fake; | |
130 | struct dx_entry entries[0]; | |
131 | }; | |
132 | ||
133 | ||
134 | struct dx_frame | |
135 | { | |
136 | struct buffer_head *bh; | |
137 | struct dx_entry *entries; | |
138 | struct dx_entry *at; | |
139 | }; | |
140 | ||
141 | struct dx_map_entry | |
142 | { | |
143 | u32 hash; | |
ef2b02d3 ES |
144 | u16 offs; |
145 | u16 size; | |
ac27a0ec DK |
146 | }; |
147 | ||
e6153918 DW |
148 | /* |
149 | * This goes at the end of each htree block. | |
150 | */ | |
151 | struct dx_tail { | |
152 | u32 dt_reserved; | |
153 | __le32 dt_checksum; /* crc32c(uuid+inum+dirblock) */ | |
154 | }; | |
155 | ||
725d26d3 AK |
156 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry); |
157 | static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value); | |
af5bc92d TT |
158 | static inline unsigned dx_get_hash(struct dx_entry *entry); |
159 | static void dx_set_hash(struct dx_entry *entry, unsigned value); | |
160 | static unsigned dx_get_count(struct dx_entry *entries); | |
161 | static unsigned dx_get_limit(struct dx_entry *entries); | |
162 | static void dx_set_count(struct dx_entry *entries, unsigned value); | |
163 | static void dx_set_limit(struct dx_entry *entries, unsigned value); | |
164 | static unsigned dx_root_limit(struct inode *dir, unsigned infosize); | |
165 | static unsigned dx_node_limit(struct inode *dir); | |
f702ba0f | 166 | static struct dx_frame *dx_probe(const struct qstr *d_name, |
ac27a0ec DK |
167 | struct inode *dir, |
168 | struct dx_hash_info *hinfo, | |
169 | struct dx_frame *frame, | |
170 | int *err); | |
af5bc92d | 171 | static void dx_release(struct dx_frame *frames); |
8bad4597 | 172 | static int dx_make_map(struct ext4_dir_entry_2 *de, unsigned blocksize, |
af5bc92d | 173 | struct dx_hash_info *hinfo, struct dx_map_entry map[]); |
ac27a0ec | 174 | static void dx_sort_map(struct dx_map_entry *map, unsigned count); |
af5bc92d | 175 | static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to, |
3d0518f4 | 176 | struct dx_map_entry *offsets, int count, unsigned blocksize); |
8bad4597 | 177 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize); |
725d26d3 AK |
178 | static void dx_insert_block(struct dx_frame *frame, |
179 | u32 hash, ext4_lblk_t block); | |
617ba13b | 180 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
181 | struct dx_frame *frame, |
182 | struct dx_frame *frames, | |
183 | __u32 *start_hash); | |
f702ba0f TT |
184 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, |
185 | const struct qstr *d_name, | |
186 | struct ext4_dir_entry_2 **res_dir, | |
187 | int *err); | |
617ba13b | 188 | static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry, |
ac27a0ec DK |
189 | struct inode *inode); |
190 | ||
dbe89444 DW |
191 | /* checksumming functions */ |
192 | static struct dx_countlimit *get_dx_countlimit(struct inode *inode, | |
193 | struct ext4_dir_entry *dirent, | |
194 | int *offset) | |
195 | { | |
196 | struct ext4_dir_entry *dp; | |
197 | struct dx_root_info *root; | |
198 | int count_offset; | |
199 | ||
200 | if (le16_to_cpu(dirent->rec_len) == EXT4_BLOCK_SIZE(inode->i_sb)) | |
201 | count_offset = 8; | |
202 | else if (le16_to_cpu(dirent->rec_len) == 12) { | |
203 | dp = (struct ext4_dir_entry *)(((void *)dirent) + 12); | |
204 | if (le16_to_cpu(dp->rec_len) != | |
205 | EXT4_BLOCK_SIZE(inode->i_sb) - 12) | |
206 | return NULL; | |
207 | root = (struct dx_root_info *)(((void *)dp + 12)); | |
208 | if (root->reserved_zero || | |
209 | root->info_length != sizeof(struct dx_root_info)) | |
210 | return NULL; | |
211 | count_offset = 32; | |
212 | } else | |
213 | return NULL; | |
214 | ||
215 | if (offset) | |
216 | *offset = count_offset; | |
217 | return (struct dx_countlimit *)(((void *)dirent) + count_offset); | |
218 | } | |
219 | ||
220 | static __le32 ext4_dx_csum(struct inode *inode, struct ext4_dir_entry *dirent, | |
221 | int count_offset, int count, struct dx_tail *t) | |
222 | { | |
223 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
224 | struct ext4_inode_info *ei = EXT4_I(inode); | |
225 | __u32 csum, old_csum; | |
226 | int size; | |
227 | ||
228 | size = count_offset + (count * sizeof(struct dx_entry)); | |
229 | old_csum = t->dt_checksum; | |
230 | t->dt_checksum = 0; | |
231 | csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size); | |
232 | csum = ext4_chksum(sbi, csum, (__u8 *)t, sizeof(struct dx_tail)); | |
233 | t->dt_checksum = old_csum; | |
234 | ||
235 | return cpu_to_le32(csum); | |
236 | } | |
237 | ||
238 | static int ext4_dx_csum_verify(struct inode *inode, | |
239 | struct ext4_dir_entry *dirent) | |
240 | { | |
241 | struct dx_countlimit *c; | |
242 | struct dx_tail *t; | |
243 | int count_offset, limit, count; | |
244 | ||
245 | if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb, | |
246 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) | |
247 | return 1; | |
248 | ||
249 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
250 | if (!c) { | |
251 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
252 | return 1; | |
253 | } | |
254 | limit = le16_to_cpu(c->limit); | |
255 | count = le16_to_cpu(c->count); | |
256 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
257 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
258 | EXT4_ERROR_INODE(inode, "metadata_csum set but no space for " | |
259 | "tree checksum found. Run e2fsck -D."); | |
260 | return 1; | |
261 | } | |
262 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
263 | ||
264 | if (t->dt_checksum != ext4_dx_csum(inode, dirent, count_offset, | |
265 | count, t)) | |
266 | return 0; | |
267 | return 1; | |
268 | } | |
269 | ||
270 | static void ext4_dx_csum_set(struct inode *inode, struct ext4_dir_entry *dirent) | |
271 | { | |
272 | struct dx_countlimit *c; | |
273 | struct dx_tail *t; | |
274 | int count_offset, limit, count; | |
275 | ||
276 | if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb, | |
277 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) | |
278 | return; | |
279 | ||
280 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
281 | if (!c) { | |
282 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
283 | return; | |
284 | } | |
285 | limit = le16_to_cpu(c->limit); | |
286 | count = le16_to_cpu(c->count); | |
287 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
288 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
289 | EXT4_ERROR_INODE(inode, "metadata_csum set but no space for " | |
290 | "tree checksum. Run e2fsck -D."); | |
291 | return; | |
292 | } | |
293 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
294 | ||
295 | t->dt_checksum = ext4_dx_csum(inode, dirent, count_offset, count, t); | |
296 | } | |
297 | ||
298 | static inline int ext4_handle_dirty_dx_node(handle_t *handle, | |
299 | struct inode *inode, | |
300 | struct buffer_head *bh) | |
301 | { | |
302 | ext4_dx_csum_set(inode, (struct ext4_dir_entry *)bh->b_data); | |
303 | return ext4_handle_dirty_metadata(handle, inode, bh); | |
304 | } | |
305 | ||
f795e140 LZ |
306 | /* |
307 | * p is at least 6 bytes before the end of page | |
308 | */ | |
309 | static inline struct ext4_dir_entry_2 * | |
3d0518f4 | 310 | ext4_next_entry(struct ext4_dir_entry_2 *p, unsigned long blocksize) |
f795e140 LZ |
311 | { |
312 | return (struct ext4_dir_entry_2 *)((char *)p + | |
3d0518f4 | 313 | ext4_rec_len_from_disk(p->rec_len, blocksize)); |
f795e140 LZ |
314 | } |
315 | ||
ac27a0ec DK |
316 | /* |
317 | * Future: use high four bits of block for coalesce-on-delete flags | |
318 | * Mask them off for now. | |
319 | */ | |
320 | ||
725d26d3 | 321 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry) |
ac27a0ec DK |
322 | { |
323 | return le32_to_cpu(entry->block) & 0x00ffffff; | |
324 | } | |
325 | ||
725d26d3 | 326 | static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value) |
ac27a0ec DK |
327 | { |
328 | entry->block = cpu_to_le32(value); | |
329 | } | |
330 | ||
af5bc92d | 331 | static inline unsigned dx_get_hash(struct dx_entry *entry) |
ac27a0ec DK |
332 | { |
333 | return le32_to_cpu(entry->hash); | |
334 | } | |
335 | ||
af5bc92d | 336 | static inline void dx_set_hash(struct dx_entry *entry, unsigned value) |
ac27a0ec DK |
337 | { |
338 | entry->hash = cpu_to_le32(value); | |
339 | } | |
340 | ||
af5bc92d | 341 | static inline unsigned dx_get_count(struct dx_entry *entries) |
ac27a0ec DK |
342 | { |
343 | return le16_to_cpu(((struct dx_countlimit *) entries)->count); | |
344 | } | |
345 | ||
af5bc92d | 346 | static inline unsigned dx_get_limit(struct dx_entry *entries) |
ac27a0ec DK |
347 | { |
348 | return le16_to_cpu(((struct dx_countlimit *) entries)->limit); | |
349 | } | |
350 | ||
af5bc92d | 351 | static inline void dx_set_count(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
352 | { |
353 | ((struct dx_countlimit *) entries)->count = cpu_to_le16(value); | |
354 | } | |
355 | ||
af5bc92d | 356 | static inline void dx_set_limit(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
357 | { |
358 | ((struct dx_countlimit *) entries)->limit = cpu_to_le16(value); | |
359 | } | |
360 | ||
af5bc92d | 361 | static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize) |
ac27a0ec | 362 | { |
617ba13b MC |
363 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) - |
364 | EXT4_DIR_REC_LEN(2) - infosize; | |
dbe89444 DW |
365 | |
366 | if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb, | |
367 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) | |
368 | entry_space -= sizeof(struct dx_tail); | |
d9c769b7 | 369 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
370 | } |
371 | ||
af5bc92d | 372 | static inline unsigned dx_node_limit(struct inode *dir) |
ac27a0ec | 373 | { |
617ba13b | 374 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0); |
dbe89444 DW |
375 | |
376 | if (EXT4_HAS_RO_COMPAT_FEATURE(dir->i_sb, | |
377 | EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) | |
378 | entry_space -= sizeof(struct dx_tail); | |
d9c769b7 | 379 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
380 | } |
381 | ||
382 | /* | |
383 | * Debug | |
384 | */ | |
385 | #ifdef DX_DEBUG | |
4776004f | 386 | static void dx_show_index(char * label, struct dx_entry *entries) |
ac27a0ec | 387 | { |
63f57933 | 388 | int i, n = dx_get_count (entries); |
4776004f | 389 | printk(KERN_DEBUG "%s index ", label); |
63f57933 | 390 | for (i = 0; i < n; i++) { |
4776004f | 391 | printk("%x->%lu ", i ? dx_get_hash(entries + i) : |
725d26d3 | 392 | 0, (unsigned long)dx_get_block(entries + i)); |
63f57933 AM |
393 | } |
394 | printk("\n"); | |
ac27a0ec DK |
395 | } |
396 | ||
397 | struct stats | |
398 | { | |
399 | unsigned names; | |
400 | unsigned space; | |
401 | unsigned bcount; | |
402 | }; | |
403 | ||
617ba13b | 404 | static struct stats dx_show_leaf(struct dx_hash_info *hinfo, struct ext4_dir_entry_2 *de, |
ac27a0ec DK |
405 | int size, int show_names) |
406 | { | |
407 | unsigned names = 0, space = 0; | |
408 | char *base = (char *) de; | |
409 | struct dx_hash_info h = *hinfo; | |
410 | ||
411 | printk("names: "); | |
412 | while ((char *) de < base + size) | |
413 | { | |
414 | if (de->inode) | |
415 | { | |
416 | if (show_names) | |
417 | { | |
418 | int len = de->name_len; | |
419 | char *name = de->name; | |
420 | while (len--) printk("%c", *name++); | |
617ba13b | 421 | ext4fs_dirhash(de->name, de->name_len, &h); |
ac27a0ec | 422 | printk(":%x.%u ", h.hash, |
265c6a0f | 423 | (unsigned) ((char *) de - base)); |
ac27a0ec | 424 | } |
617ba13b | 425 | space += EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
426 | names++; |
427 | } | |
3d0518f4 | 428 | de = ext4_next_entry(de, size); |
ac27a0ec DK |
429 | } |
430 | printk("(%i)\n", names); | |
431 | return (struct stats) { names, space, 1 }; | |
432 | } | |
433 | ||
434 | struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, | |
435 | struct dx_entry *entries, int levels) | |
436 | { | |
437 | unsigned blocksize = dir->i_sb->s_blocksize; | |
af5bc92d | 438 | unsigned count = dx_get_count(entries), names = 0, space = 0, i; |
ac27a0ec DK |
439 | unsigned bcount = 0; |
440 | struct buffer_head *bh; | |
441 | int err; | |
442 | printk("%i indexed blocks...\n", count); | |
443 | for (i = 0; i < count; i++, entries++) | |
444 | { | |
725d26d3 AK |
445 | ext4_lblk_t block = dx_get_block(entries); |
446 | ext4_lblk_t hash = i ? dx_get_hash(entries): 0; | |
ac27a0ec DK |
447 | u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; |
448 | struct stats stats; | |
449 | printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); | |
617ba13b | 450 | if (!(bh = ext4_bread (NULL,dir, block, 0,&err))) continue; |
ac27a0ec DK |
451 | stats = levels? |
452 | dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): | |
617ba13b | 453 | dx_show_leaf(hinfo, (struct ext4_dir_entry_2 *) bh->b_data, blocksize, 0); |
ac27a0ec DK |
454 | names += stats.names; |
455 | space += stats.space; | |
456 | bcount += stats.bcount; | |
af5bc92d | 457 | brelse(bh); |
ac27a0ec DK |
458 | } |
459 | if (bcount) | |
60e6679e | 460 | printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n", |
4776004f TT |
461 | levels ? "" : " ", names, space/bcount, |
462 | (space/bcount)*100/blocksize); | |
ac27a0ec DK |
463 | return (struct stats) { names, space, bcount}; |
464 | } | |
465 | #endif /* DX_DEBUG */ | |
466 | ||
467 | /* | |
468 | * Probe for a directory leaf block to search. | |
469 | * | |
470 | * dx_probe can return ERR_BAD_DX_DIR, which means there was a format | |
471 | * error in the directory index, and the caller should fall back to | |
472 | * searching the directory normally. The callers of dx_probe **MUST** | |
473 | * check for this error code, and make sure it never gets reflected | |
474 | * back to userspace. | |
475 | */ | |
476 | static struct dx_frame * | |
f702ba0f | 477 | dx_probe(const struct qstr *d_name, struct inode *dir, |
ac27a0ec DK |
478 | struct dx_hash_info *hinfo, struct dx_frame *frame_in, int *err) |
479 | { | |
480 | unsigned count, indirect; | |
481 | struct dx_entry *at, *entries, *p, *q, *m; | |
482 | struct dx_root *root; | |
483 | struct buffer_head *bh; | |
484 | struct dx_frame *frame = frame_in; | |
485 | u32 hash; | |
486 | ||
487 | frame->bh = NULL; | |
617ba13b | 488 | if (!(bh = ext4_bread (NULL,dir, 0, 0, err))) |
ac27a0ec DK |
489 | goto fail; |
490 | root = (struct dx_root *) bh->b_data; | |
491 | if (root->info.hash_version != DX_HASH_TEA && | |
492 | root->info.hash_version != DX_HASH_HALF_MD4 && | |
493 | root->info.hash_version != DX_HASH_LEGACY) { | |
12062ddd | 494 | ext4_warning(dir->i_sb, "Unrecognised inode hash code %d", |
ac27a0ec DK |
495 | root->info.hash_version); |
496 | brelse(bh); | |
497 | *err = ERR_BAD_DX_DIR; | |
498 | goto fail; | |
499 | } | |
500 | hinfo->hash_version = root->info.hash_version; | |
f99b2589 TT |
501 | if (hinfo->hash_version <= DX_HASH_TEA) |
502 | hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 503 | hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
f702ba0f TT |
504 | if (d_name) |
505 | ext4fs_dirhash(d_name->name, d_name->len, hinfo); | |
ac27a0ec DK |
506 | hash = hinfo->hash; |
507 | ||
508 | if (root->info.unused_flags & 1) { | |
12062ddd | 509 | ext4_warning(dir->i_sb, "Unimplemented inode hash flags: %#06x", |
ac27a0ec DK |
510 | root->info.unused_flags); |
511 | brelse(bh); | |
512 | *err = ERR_BAD_DX_DIR; | |
513 | goto fail; | |
514 | } | |
515 | ||
516 | if ((indirect = root->info.indirect_levels) > 1) { | |
12062ddd | 517 | ext4_warning(dir->i_sb, "Unimplemented inode hash depth: %#06x", |
ac27a0ec DK |
518 | root->info.indirect_levels); |
519 | brelse(bh); | |
520 | *err = ERR_BAD_DX_DIR; | |
521 | goto fail; | |
522 | } | |
523 | ||
dbe89444 DW |
524 | if (!buffer_verified(bh) && |
525 | !ext4_dx_csum_verify(dir, (struct ext4_dir_entry *)bh->b_data)) { | |
526 | ext4_warning(dir->i_sb, "Root failed checksum"); | |
527 | brelse(bh); | |
528 | *err = ERR_BAD_DX_DIR; | |
529 | goto fail; | |
530 | } | |
531 | set_buffer_verified(bh); | |
532 | ||
ac27a0ec DK |
533 | entries = (struct dx_entry *) (((char *)&root->info) + |
534 | root->info.info_length); | |
3d82abae ES |
535 | |
536 | if (dx_get_limit(entries) != dx_root_limit(dir, | |
537 | root->info.info_length)) { | |
12062ddd | 538 | ext4_warning(dir->i_sb, "dx entry: limit != root limit"); |
3d82abae ES |
539 | brelse(bh); |
540 | *err = ERR_BAD_DX_DIR; | |
541 | goto fail; | |
542 | } | |
543 | ||
af5bc92d | 544 | dxtrace(printk("Look up %x", hash)); |
ac27a0ec DK |
545 | while (1) |
546 | { | |
547 | count = dx_get_count(entries); | |
3d82abae | 548 | if (!count || count > dx_get_limit(entries)) { |
12062ddd | 549 | ext4_warning(dir->i_sb, |
3d82abae ES |
550 | "dx entry: no count or count > limit"); |
551 | brelse(bh); | |
552 | *err = ERR_BAD_DX_DIR; | |
553 | goto fail2; | |
554 | } | |
555 | ||
ac27a0ec DK |
556 | p = entries + 1; |
557 | q = entries + count - 1; | |
558 | while (p <= q) | |
559 | { | |
560 | m = p + (q - p)/2; | |
561 | dxtrace(printk(".")); | |
562 | if (dx_get_hash(m) > hash) | |
563 | q = m - 1; | |
564 | else | |
565 | p = m + 1; | |
566 | } | |
567 | ||
568 | if (0) // linear search cross check | |
569 | { | |
570 | unsigned n = count - 1; | |
571 | at = entries; | |
572 | while (n--) | |
573 | { | |
574 | dxtrace(printk(",")); | |
575 | if (dx_get_hash(++at) > hash) | |
576 | { | |
577 | at--; | |
578 | break; | |
579 | } | |
580 | } | |
581 | assert (at == p - 1); | |
582 | } | |
583 | ||
584 | at = p - 1; | |
585 | dxtrace(printk(" %x->%u\n", at == entries? 0: dx_get_hash(at), dx_get_block(at))); | |
586 | frame->bh = bh; | |
587 | frame->entries = entries; | |
588 | frame->at = at; | |
589 | if (!indirect--) return frame; | |
617ba13b | 590 | if (!(bh = ext4_bread (NULL,dir, dx_get_block(at), 0, err))) |
ac27a0ec DK |
591 | goto fail2; |
592 | at = entries = ((struct dx_node *) bh->b_data)->entries; | |
dbe89444 DW |
593 | |
594 | if (!buffer_verified(bh) && | |
595 | !ext4_dx_csum_verify(dir, | |
596 | (struct ext4_dir_entry *)bh->b_data)) { | |
597 | ext4_warning(dir->i_sb, "Node failed checksum"); | |
598 | brelse(bh); | |
599 | *err = ERR_BAD_DX_DIR; | |
600 | goto fail; | |
601 | } | |
602 | set_buffer_verified(bh); | |
603 | ||
3d82abae | 604 | if (dx_get_limit(entries) != dx_node_limit (dir)) { |
12062ddd | 605 | ext4_warning(dir->i_sb, |
3d82abae ES |
606 | "dx entry: limit != node limit"); |
607 | brelse(bh); | |
608 | *err = ERR_BAD_DX_DIR; | |
609 | goto fail2; | |
610 | } | |
ac27a0ec | 611 | frame++; |
3d82abae | 612 | frame->bh = NULL; |
ac27a0ec DK |
613 | } |
614 | fail2: | |
615 | while (frame >= frame_in) { | |
616 | brelse(frame->bh); | |
617 | frame--; | |
618 | } | |
619 | fail: | |
3d82abae | 620 | if (*err == ERR_BAD_DX_DIR) |
12062ddd | 621 | ext4_warning(dir->i_sb, |
9ee49302 | 622 | "Corrupt dir inode %lu, running e2fsck is " |
3d82abae | 623 | "recommended.", dir->i_ino); |
ac27a0ec DK |
624 | return NULL; |
625 | } | |
626 | ||
627 | static void dx_release (struct dx_frame *frames) | |
628 | { | |
629 | if (frames[0].bh == NULL) | |
630 | return; | |
631 | ||
632 | if (((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels) | |
633 | brelse(frames[1].bh); | |
634 | brelse(frames[0].bh); | |
635 | } | |
636 | ||
637 | /* | |
638 | * This function increments the frame pointer to search the next leaf | |
639 | * block, and reads in the necessary intervening nodes if the search | |
640 | * should be necessary. Whether or not the search is necessary is | |
641 | * controlled by the hash parameter. If the hash value is even, then | |
642 | * the search is only continued if the next block starts with that | |
643 | * hash value. This is used if we are searching for a specific file. | |
644 | * | |
645 | * If the hash value is HASH_NB_ALWAYS, then always go to the next block. | |
646 | * | |
647 | * This function returns 1 if the caller should continue to search, | |
648 | * or 0 if it should not. If there is an error reading one of the | |
649 | * index blocks, it will a negative error code. | |
650 | * | |
651 | * If start_hash is non-null, it will be filled in with the starting | |
652 | * hash of the next page. | |
653 | */ | |
617ba13b | 654 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
655 | struct dx_frame *frame, |
656 | struct dx_frame *frames, | |
657 | __u32 *start_hash) | |
658 | { | |
659 | struct dx_frame *p; | |
660 | struct buffer_head *bh; | |
661 | int err, num_frames = 0; | |
662 | __u32 bhash; | |
663 | ||
664 | p = frame; | |
665 | /* | |
666 | * Find the next leaf page by incrementing the frame pointer. | |
667 | * If we run out of entries in the interior node, loop around and | |
668 | * increment pointer in the parent node. When we break out of | |
669 | * this loop, num_frames indicates the number of interior | |
670 | * nodes need to be read. | |
671 | */ | |
672 | while (1) { | |
673 | if (++(p->at) < p->entries + dx_get_count(p->entries)) | |
674 | break; | |
675 | if (p == frames) | |
676 | return 0; | |
677 | num_frames++; | |
678 | p--; | |
679 | } | |
680 | ||
681 | /* | |
682 | * If the hash is 1, then continue only if the next page has a | |
683 | * continuation hash of any value. This is used for readdir | |
684 | * handling. Otherwise, check to see if the hash matches the | |
685 | * desired contiuation hash. If it doesn't, return since | |
686 | * there's no point to read in the successive index pages. | |
687 | */ | |
688 | bhash = dx_get_hash(p->at); | |
689 | if (start_hash) | |
690 | *start_hash = bhash; | |
691 | if ((hash & 1) == 0) { | |
692 | if ((bhash & ~1) != hash) | |
693 | return 0; | |
694 | } | |
695 | /* | |
696 | * If the hash is HASH_NB_ALWAYS, we always go to the next | |
697 | * block so no check is necessary | |
698 | */ | |
699 | while (num_frames--) { | |
617ba13b | 700 | if (!(bh = ext4_bread(NULL, dir, dx_get_block(p->at), |
ac27a0ec DK |
701 | 0, &err))) |
702 | return err; /* Failure */ | |
dbe89444 DW |
703 | |
704 | if (!buffer_verified(bh) && | |
705 | !ext4_dx_csum_verify(dir, | |
706 | (struct ext4_dir_entry *)bh->b_data)) { | |
707 | ext4_warning(dir->i_sb, "Node failed checksum"); | |
708 | return -EIO; | |
709 | } | |
710 | set_buffer_verified(bh); | |
711 | ||
ac27a0ec | 712 | p++; |
af5bc92d | 713 | brelse(p->bh); |
ac27a0ec DK |
714 | p->bh = bh; |
715 | p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; | |
716 | } | |
717 | return 1; | |
718 | } | |
719 | ||
720 | ||
ac27a0ec DK |
721 | /* |
722 | * This function fills a red-black tree with information from a | |
723 | * directory block. It returns the number directory entries loaded | |
724 | * into the tree. If there is an error it is returned in err. | |
725 | */ | |
726 | static int htree_dirblock_to_tree(struct file *dir_file, | |
725d26d3 | 727 | struct inode *dir, ext4_lblk_t block, |
ac27a0ec DK |
728 | struct dx_hash_info *hinfo, |
729 | __u32 start_hash, __u32 start_minor_hash) | |
730 | { | |
731 | struct buffer_head *bh; | |
617ba13b | 732 | struct ext4_dir_entry_2 *de, *top; |
ac27a0ec DK |
733 | int err, count = 0; |
734 | ||
725d26d3 AK |
735 | dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n", |
736 | (unsigned long)block)); | |
617ba13b | 737 | if (!(bh = ext4_bread (NULL, dir, block, 0, &err))) |
ac27a0ec DK |
738 | return err; |
739 | ||
617ba13b MC |
740 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
741 | top = (struct ext4_dir_entry_2 *) ((char *) de + | |
ac27a0ec | 742 | dir->i_sb->s_blocksize - |
617ba13b | 743 | EXT4_DIR_REC_LEN(0)); |
3d0518f4 | 744 | for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) { |
f7c21177 | 745 | if (ext4_check_dir_entry(dir, NULL, de, bh, |
cad3f007 TT |
746 | (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb)) |
747 | + ((char *)de - bh->b_data))) { | |
e6c40211 ES |
748 | /* On error, skip the f_pos to the next block. */ |
749 | dir_file->f_pos = (dir_file->f_pos | | |
750 | (dir->i_sb->s_blocksize - 1)) + 1; | |
af5bc92d | 751 | brelse(bh); |
e6c40211 ES |
752 | return count; |
753 | } | |
617ba13b | 754 | ext4fs_dirhash(de->name, de->name_len, hinfo); |
ac27a0ec DK |
755 | if ((hinfo->hash < start_hash) || |
756 | ((hinfo->hash == start_hash) && | |
757 | (hinfo->minor_hash < start_minor_hash))) | |
758 | continue; | |
759 | if (de->inode == 0) | |
760 | continue; | |
617ba13b | 761 | if ((err = ext4_htree_store_dirent(dir_file, |
ac27a0ec DK |
762 | hinfo->hash, hinfo->minor_hash, de)) != 0) { |
763 | brelse(bh); | |
764 | return err; | |
765 | } | |
766 | count++; | |
767 | } | |
768 | brelse(bh); | |
769 | return count; | |
770 | } | |
771 | ||
772 | ||
773 | /* | |
774 | * This function fills a red-black tree with information from a | |
775 | * directory. We start scanning the directory in hash order, starting | |
776 | * at start_hash and start_minor_hash. | |
777 | * | |
778 | * This function returns the number of entries inserted into the tree, | |
779 | * or a negative error code. | |
780 | */ | |
617ba13b | 781 | int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, |
ac27a0ec DK |
782 | __u32 start_minor_hash, __u32 *next_hash) |
783 | { | |
784 | struct dx_hash_info hinfo; | |
617ba13b | 785 | struct ext4_dir_entry_2 *de; |
ac27a0ec DK |
786 | struct dx_frame frames[2], *frame; |
787 | struct inode *dir; | |
725d26d3 | 788 | ext4_lblk_t block; |
ac27a0ec | 789 | int count = 0; |
725d26d3 | 790 | int ret, err; |
ac27a0ec DK |
791 | __u32 hashval; |
792 | ||
60e6679e | 793 | dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n", |
4776004f | 794 | start_hash, start_minor_hash)); |
9d549890 | 795 | dir = dir_file->f_path.dentry->d_inode; |
12e9b892 | 796 | if (!(ext4_test_inode_flag(dir, EXT4_INODE_INDEX))) { |
617ba13b | 797 | hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
f99b2589 TT |
798 | if (hinfo.hash_version <= DX_HASH_TEA) |
799 | hinfo.hash_version += | |
800 | EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 801 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
ac27a0ec DK |
802 | count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo, |
803 | start_hash, start_minor_hash); | |
804 | *next_hash = ~0; | |
805 | return count; | |
806 | } | |
807 | hinfo.hash = start_hash; | |
808 | hinfo.minor_hash = 0; | |
f702ba0f | 809 | frame = dx_probe(NULL, dir, &hinfo, frames, &err); |
ac27a0ec DK |
810 | if (!frame) |
811 | return err; | |
812 | ||
813 | /* Add '.' and '..' from the htree header */ | |
814 | if (!start_hash && !start_minor_hash) { | |
617ba13b MC |
815 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
816 | if ((err = ext4_htree_store_dirent(dir_file, 0, 0, de)) != 0) | |
ac27a0ec DK |
817 | goto errout; |
818 | count++; | |
819 | } | |
820 | if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) { | |
617ba13b | 821 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
3d0518f4 | 822 | de = ext4_next_entry(de, dir->i_sb->s_blocksize); |
617ba13b | 823 | if ((err = ext4_htree_store_dirent(dir_file, 2, 0, de)) != 0) |
ac27a0ec DK |
824 | goto errout; |
825 | count++; | |
826 | } | |
827 | ||
828 | while (1) { | |
829 | block = dx_get_block(frame->at); | |
830 | ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo, | |
831 | start_hash, start_minor_hash); | |
832 | if (ret < 0) { | |
833 | err = ret; | |
834 | goto errout; | |
835 | } | |
836 | count += ret; | |
837 | hashval = ~0; | |
617ba13b | 838 | ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS, |
ac27a0ec DK |
839 | frame, frames, &hashval); |
840 | *next_hash = hashval; | |
841 | if (ret < 0) { | |
842 | err = ret; | |
843 | goto errout; | |
844 | } | |
845 | /* | |
846 | * Stop if: (a) there are no more entries, or | |
847 | * (b) we have inserted at least one entry and the | |
848 | * next hash value is not a continuation | |
849 | */ | |
850 | if ((ret == 0) || | |
851 | (count && ((hashval & 1) == 0))) | |
852 | break; | |
853 | } | |
854 | dx_release(frames); | |
4776004f TT |
855 | dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, " |
856 | "next hash: %x\n", count, *next_hash)); | |
ac27a0ec DK |
857 | return count; |
858 | errout: | |
859 | dx_release(frames); | |
860 | return (err); | |
861 | } | |
862 | ||
863 | ||
864 | /* | |
865 | * Directory block splitting, compacting | |
866 | */ | |
867 | ||
ef2b02d3 ES |
868 | /* |
869 | * Create map of hash values, offsets, and sizes, stored at end of block. | |
870 | * Returns number of entries mapped. | |
871 | */ | |
8bad4597 TT |
872 | static int dx_make_map(struct ext4_dir_entry_2 *de, unsigned blocksize, |
873 | struct dx_hash_info *hinfo, | |
874 | struct dx_map_entry *map_tail) | |
ac27a0ec DK |
875 | { |
876 | int count = 0; | |
877 | char *base = (char *) de; | |
878 | struct dx_hash_info h = *hinfo; | |
879 | ||
8bad4597 | 880 | while ((char *) de < base + blocksize) { |
ac27a0ec | 881 | if (de->name_len && de->inode) { |
617ba13b | 882 | ext4fs_dirhash(de->name, de->name_len, &h); |
ac27a0ec DK |
883 | map_tail--; |
884 | map_tail->hash = h.hash; | |
9aee2286 | 885 | map_tail->offs = ((char *) de - base)>>2; |
ef2b02d3 | 886 | map_tail->size = le16_to_cpu(de->rec_len); |
ac27a0ec DK |
887 | count++; |
888 | cond_resched(); | |
889 | } | |
890 | /* XXX: do we need to check rec_len == 0 case? -Chris */ | |
3d0518f4 | 891 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
892 | } |
893 | return count; | |
894 | } | |
895 | ||
ef2b02d3 | 896 | /* Sort map by hash value */ |
ac27a0ec DK |
897 | static void dx_sort_map (struct dx_map_entry *map, unsigned count) |
898 | { | |
63f57933 AM |
899 | struct dx_map_entry *p, *q, *top = map + count - 1; |
900 | int more; | |
901 | /* Combsort until bubble sort doesn't suck */ | |
902 | while (count > 2) { | |
903 | count = count*10/13; | |
904 | if (count - 9 < 2) /* 9, 10 -> 11 */ | |
905 | count = 11; | |
906 | for (p = top, q = p - count; q >= map; p--, q--) | |
907 | if (p->hash < q->hash) | |
908 | swap(*p, *q); | |
909 | } | |
910 | /* Garden variety bubble sort */ | |
911 | do { | |
912 | more = 0; | |
913 | q = top; | |
914 | while (q-- > map) { | |
915 | if (q[1].hash >= q[0].hash) | |
ac27a0ec | 916 | continue; |
63f57933 AM |
917 | swap(*(q+1), *q); |
918 | more = 1; | |
ac27a0ec DK |
919 | } |
920 | } while(more); | |
921 | } | |
922 | ||
725d26d3 | 923 | static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block) |
ac27a0ec DK |
924 | { |
925 | struct dx_entry *entries = frame->entries; | |
926 | struct dx_entry *old = frame->at, *new = old + 1; | |
927 | int count = dx_get_count(entries); | |
928 | ||
929 | assert(count < dx_get_limit(entries)); | |
930 | assert(old < entries + count); | |
931 | memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); | |
932 | dx_set_hash(new, hash); | |
933 | dx_set_block(new, block); | |
934 | dx_set_count(entries, count + 1); | |
935 | } | |
ac27a0ec | 936 | |
617ba13b | 937 | static void ext4_update_dx_flag(struct inode *inode) |
ac27a0ec | 938 | { |
617ba13b MC |
939 | if (!EXT4_HAS_COMPAT_FEATURE(inode->i_sb, |
940 | EXT4_FEATURE_COMPAT_DIR_INDEX)) | |
12e9b892 | 941 | ext4_clear_inode_flag(inode, EXT4_INODE_INDEX); |
ac27a0ec DK |
942 | } |
943 | ||
944 | /* | |
617ba13b | 945 | * NOTE! unlike strncmp, ext4_match returns 1 for success, 0 for failure. |
ac27a0ec | 946 | * |
617ba13b | 947 | * `len <= EXT4_NAME_LEN' is guaranteed by caller. |
ac27a0ec DK |
948 | * `de != NULL' is guaranteed by caller. |
949 | */ | |
617ba13b MC |
950 | static inline int ext4_match (int len, const char * const name, |
951 | struct ext4_dir_entry_2 * de) | |
ac27a0ec DK |
952 | { |
953 | if (len != de->name_len) | |
954 | return 0; | |
955 | if (!de->inode) | |
956 | return 0; | |
957 | return !memcmp(name, de->name, len); | |
958 | } | |
959 | ||
960 | /* | |
961 | * Returns 0 if not found, -1 on failure, and 1 on success | |
962 | */ | |
af5bc92d | 963 | static inline int search_dirblock(struct buffer_head *bh, |
ac27a0ec | 964 | struct inode *dir, |
f702ba0f | 965 | const struct qstr *d_name, |
498e5f24 | 966 | unsigned int offset, |
617ba13b | 967 | struct ext4_dir_entry_2 ** res_dir) |
ac27a0ec | 968 | { |
617ba13b | 969 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
970 | char * dlimit; |
971 | int de_len; | |
f702ba0f TT |
972 | const char *name = d_name->name; |
973 | int namelen = d_name->len; | |
ac27a0ec | 974 | |
617ba13b | 975 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec DK |
976 | dlimit = bh->b_data + dir->i_sb->s_blocksize; |
977 | while ((char *) de < dlimit) { | |
978 | /* this code is executed quadratically often */ | |
979 | /* do minimal checking `by hand' */ | |
980 | ||
981 | if ((char *) de + namelen <= dlimit && | |
617ba13b | 982 | ext4_match (namelen, name, de)) { |
ac27a0ec | 983 | /* found a match - just to be sure, do a full check */ |
f7c21177 | 984 | if (ext4_check_dir_entry(dir, NULL, de, bh, offset)) |
ac27a0ec DK |
985 | return -1; |
986 | *res_dir = de; | |
987 | return 1; | |
988 | } | |
989 | /* prevent looping on a bad block */ | |
3d0518f4 WY |
990 | de_len = ext4_rec_len_from_disk(de->rec_len, |
991 | dir->i_sb->s_blocksize); | |
ac27a0ec DK |
992 | if (de_len <= 0) |
993 | return -1; | |
994 | offset += de_len; | |
617ba13b | 995 | de = (struct ext4_dir_entry_2 *) ((char *) de + de_len); |
ac27a0ec DK |
996 | } |
997 | return 0; | |
998 | } | |
999 | ||
1000 | ||
1001 | /* | |
617ba13b | 1002 | * ext4_find_entry() |
ac27a0ec DK |
1003 | * |
1004 | * finds an entry in the specified directory with the wanted name. It | |
1005 | * returns the cache buffer in which the entry was found, and the entry | |
1006 | * itself (as a parameter - res_dir). It does NOT read the inode of the | |
1007 | * entry - you'll have to do that yourself if you want to. | |
1008 | * | |
1009 | * The returned buffer_head has ->b_count elevated. The caller is expected | |
1010 | * to brelse() it when appropriate. | |
1011 | */ | |
f702ba0f TT |
1012 | static struct buffer_head * ext4_find_entry (struct inode *dir, |
1013 | const struct qstr *d_name, | |
617ba13b | 1014 | struct ext4_dir_entry_2 ** res_dir) |
ac27a0ec | 1015 | { |
af5bc92d TT |
1016 | struct super_block *sb; |
1017 | struct buffer_head *bh_use[NAMEI_RA_SIZE]; | |
1018 | struct buffer_head *bh, *ret = NULL; | |
725d26d3 | 1019 | ext4_lblk_t start, block, b; |
8941ec8b | 1020 | const u8 *name = d_name->name; |
ac27a0ec DK |
1021 | int ra_max = 0; /* Number of bh's in the readahead |
1022 | buffer, bh_use[] */ | |
1023 | int ra_ptr = 0; /* Current index into readahead | |
1024 | buffer */ | |
1025 | int num = 0; | |
725d26d3 AK |
1026 | ext4_lblk_t nblocks; |
1027 | int i, err; | |
ac27a0ec | 1028 | int namelen; |
ac27a0ec DK |
1029 | |
1030 | *res_dir = NULL; | |
1031 | sb = dir->i_sb; | |
f702ba0f | 1032 | namelen = d_name->len; |
617ba13b | 1033 | if (namelen > EXT4_NAME_LEN) |
ac27a0ec | 1034 | return NULL; |
8941ec8b | 1035 | if ((namelen <= 2) && (name[0] == '.') && |
6d5c3aa8 | 1036 | (name[1] == '.' || name[1] == '\0')) { |
8941ec8b TT |
1037 | /* |
1038 | * "." or ".." will only be in the first block | |
1039 | * NFS may look up ".."; "." should be handled by the VFS | |
1040 | */ | |
1041 | block = start = 0; | |
1042 | nblocks = 1; | |
1043 | goto restart; | |
1044 | } | |
ac27a0ec | 1045 | if (is_dx(dir)) { |
f702ba0f | 1046 | bh = ext4_dx_find_entry(dir, d_name, res_dir, &err); |
ac27a0ec DK |
1047 | /* |
1048 | * On success, or if the error was file not found, | |
1049 | * return. Otherwise, fall back to doing a search the | |
1050 | * old fashioned way. | |
1051 | */ | |
1052 | if (bh || (err != ERR_BAD_DX_DIR)) | |
1053 | return bh; | |
4776004f TT |
1054 | dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, " |
1055 | "falling back\n")); | |
ac27a0ec | 1056 | } |
617ba13b MC |
1057 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
1058 | start = EXT4_I(dir)->i_dir_start_lookup; | |
ac27a0ec DK |
1059 | if (start >= nblocks) |
1060 | start = 0; | |
1061 | block = start; | |
1062 | restart: | |
1063 | do { | |
1064 | /* | |
1065 | * We deal with the read-ahead logic here. | |
1066 | */ | |
1067 | if (ra_ptr >= ra_max) { | |
1068 | /* Refill the readahead buffer */ | |
1069 | ra_ptr = 0; | |
1070 | b = block; | |
1071 | for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) { | |
1072 | /* | |
1073 | * Terminate if we reach the end of the | |
1074 | * directory and must wrap, or if our | |
1075 | * search has finished at this block. | |
1076 | */ | |
1077 | if (b >= nblocks || (num && block == start)) { | |
1078 | bh_use[ra_max] = NULL; | |
1079 | break; | |
1080 | } | |
1081 | num++; | |
617ba13b | 1082 | bh = ext4_getblk(NULL, dir, b++, 0, &err); |
ac27a0ec DK |
1083 | bh_use[ra_max] = bh; |
1084 | if (bh) | |
65299a3b CH |
1085 | ll_rw_block(READ | REQ_META | REQ_PRIO, |
1086 | 1, &bh); | |
ac27a0ec DK |
1087 | } |
1088 | } | |
1089 | if ((bh = bh_use[ra_ptr++]) == NULL) | |
1090 | goto next; | |
1091 | wait_on_buffer(bh); | |
1092 | if (!buffer_uptodate(bh)) { | |
1093 | /* read error, skip block & hope for the best */ | |
24676da4 TT |
1094 | EXT4_ERROR_INODE(dir, "reading directory lblock %lu", |
1095 | (unsigned long) block); | |
ac27a0ec DK |
1096 | brelse(bh); |
1097 | goto next; | |
1098 | } | |
f702ba0f | 1099 | i = search_dirblock(bh, dir, d_name, |
617ba13b | 1100 | block << EXT4_BLOCK_SIZE_BITS(sb), res_dir); |
ac27a0ec | 1101 | if (i == 1) { |
617ba13b | 1102 | EXT4_I(dir)->i_dir_start_lookup = block; |
ac27a0ec DK |
1103 | ret = bh; |
1104 | goto cleanup_and_exit; | |
1105 | } else { | |
1106 | brelse(bh); | |
1107 | if (i < 0) | |
1108 | goto cleanup_and_exit; | |
1109 | } | |
1110 | next: | |
1111 | if (++block >= nblocks) | |
1112 | block = 0; | |
1113 | } while (block != start); | |
1114 | ||
1115 | /* | |
1116 | * If the directory has grown while we were searching, then | |
1117 | * search the last part of the directory before giving up. | |
1118 | */ | |
1119 | block = nblocks; | |
617ba13b | 1120 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
1121 | if (block < nblocks) { |
1122 | start = 0; | |
1123 | goto restart; | |
1124 | } | |
1125 | ||
1126 | cleanup_and_exit: | |
1127 | /* Clean up the read-ahead blocks */ | |
1128 | for (; ra_ptr < ra_max; ra_ptr++) | |
af5bc92d | 1129 | brelse(bh_use[ra_ptr]); |
ac27a0ec DK |
1130 | return ret; |
1131 | } | |
1132 | ||
f702ba0f | 1133 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name, |
617ba13b | 1134 | struct ext4_dir_entry_2 **res_dir, int *err) |
ac27a0ec | 1135 | { |
8941ec8b | 1136 | struct super_block * sb = dir->i_sb; |
ac27a0ec | 1137 | struct dx_hash_info hinfo; |
ac27a0ec | 1138 | struct dx_frame frames[2], *frame; |
ac27a0ec | 1139 | struct buffer_head *bh; |
725d26d3 | 1140 | ext4_lblk_t block; |
ac27a0ec | 1141 | int retval; |
ac27a0ec | 1142 | |
8941ec8b TT |
1143 | if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err))) |
1144 | return NULL; | |
ac27a0ec DK |
1145 | do { |
1146 | block = dx_get_block(frame->at); | |
7845c049 | 1147 | if (!(bh = ext4_bread(NULL, dir, block, 0, err))) |
ac27a0ec | 1148 | goto errout; |
f3b35f06 | 1149 | |
7845c049 TT |
1150 | retval = search_dirblock(bh, dir, d_name, |
1151 | block << EXT4_BLOCK_SIZE_BITS(sb), | |
1152 | res_dir); | |
1153 | if (retval == 1) { /* Success! */ | |
1154 | dx_release(frames); | |
1155 | return bh; | |
ac27a0ec | 1156 | } |
af5bc92d | 1157 | brelse(bh); |
7845c049 TT |
1158 | if (retval == -1) { |
1159 | *err = ERR_BAD_DX_DIR; | |
1160 | goto errout; | |
1161 | } | |
1162 | ||
ac27a0ec | 1163 | /* Check to see if we should continue to search */ |
8941ec8b | 1164 | retval = ext4_htree_next_block(dir, hinfo.hash, frame, |
ac27a0ec DK |
1165 | frames, NULL); |
1166 | if (retval < 0) { | |
12062ddd | 1167 | ext4_warning(sb, |
ac27a0ec DK |
1168 | "error reading index page in directory #%lu", |
1169 | dir->i_ino); | |
1170 | *err = retval; | |
1171 | goto errout; | |
1172 | } | |
1173 | } while (retval == 1); | |
1174 | ||
1175 | *err = -ENOENT; | |
1176 | errout: | |
265c6a0f | 1177 | dxtrace(printk(KERN_DEBUG "%s not found\n", d_name->name)); |
ac27a0ec DK |
1178 | dx_release (frames); |
1179 | return NULL; | |
1180 | } | |
ac27a0ec | 1181 | |
af5bc92d | 1182 | static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, struct nameidata *nd) |
ac27a0ec | 1183 | { |
af5bc92d TT |
1184 | struct inode *inode; |
1185 | struct ext4_dir_entry_2 *de; | |
1186 | struct buffer_head *bh; | |
ac27a0ec | 1187 | |
617ba13b | 1188 | if (dentry->d_name.len > EXT4_NAME_LEN) |
ac27a0ec DK |
1189 | return ERR_PTR(-ENAMETOOLONG); |
1190 | ||
f702ba0f | 1191 | bh = ext4_find_entry(dir, &dentry->d_name, &de); |
ac27a0ec DK |
1192 | inode = NULL; |
1193 | if (bh) { | |
498e5f24 | 1194 | __u32 ino = le32_to_cpu(de->inode); |
af5bc92d | 1195 | brelse(bh); |
617ba13b | 1196 | if (!ext4_valid_inum(dir->i_sb, ino)) { |
24676da4 | 1197 | EXT4_ERROR_INODE(dir, "bad inode number: %u", ino); |
1d1fe1ee | 1198 | return ERR_PTR(-EIO); |
a6c15c2b | 1199 | } |
1d1fe1ee | 1200 | inode = ext4_iget(dir->i_sb, ino); |
a9049376 AV |
1201 | if (inode == ERR_PTR(-ESTALE)) { |
1202 | EXT4_ERROR_INODE(dir, | |
1203 | "deleted inode referenced: %u", | |
1204 | ino); | |
1205 | return ERR_PTR(-EIO); | |
e6f009b0 | 1206 | } |
ac27a0ec DK |
1207 | } |
1208 | return d_splice_alias(inode, dentry); | |
1209 | } | |
1210 | ||
1211 | ||
617ba13b | 1212 | struct dentry *ext4_get_parent(struct dentry *child) |
ac27a0ec | 1213 | { |
498e5f24 | 1214 | __u32 ino; |
f702ba0f TT |
1215 | static const struct qstr dotdot = { |
1216 | .name = "..", | |
1217 | .len = 2, | |
1218 | }; | |
617ba13b | 1219 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
1220 | struct buffer_head *bh; |
1221 | ||
f702ba0f | 1222 | bh = ext4_find_entry(child->d_inode, &dotdot, &de); |
ac27a0ec DK |
1223 | if (!bh) |
1224 | return ERR_PTR(-ENOENT); | |
1225 | ino = le32_to_cpu(de->inode); | |
1226 | brelse(bh); | |
1227 | ||
617ba13b | 1228 | if (!ext4_valid_inum(child->d_inode->i_sb, ino)) { |
24676da4 TT |
1229 | EXT4_ERROR_INODE(child->d_inode, |
1230 | "bad parent inode number: %u", ino); | |
1d1fe1ee | 1231 | return ERR_PTR(-EIO); |
a6c15c2b VA |
1232 | } |
1233 | ||
44003728 | 1234 | return d_obtain_alias(ext4_iget(child->d_inode->i_sb, ino)); |
ac27a0ec DK |
1235 | } |
1236 | ||
1237 | #define S_SHIFT 12 | |
617ba13b MC |
1238 | static unsigned char ext4_type_by_mode[S_IFMT >> S_SHIFT] = { |
1239 | [S_IFREG >> S_SHIFT] = EXT4_FT_REG_FILE, | |
1240 | [S_IFDIR >> S_SHIFT] = EXT4_FT_DIR, | |
1241 | [S_IFCHR >> S_SHIFT] = EXT4_FT_CHRDEV, | |
1242 | [S_IFBLK >> S_SHIFT] = EXT4_FT_BLKDEV, | |
1243 | [S_IFIFO >> S_SHIFT] = EXT4_FT_FIFO, | |
1244 | [S_IFSOCK >> S_SHIFT] = EXT4_FT_SOCK, | |
1245 | [S_IFLNK >> S_SHIFT] = EXT4_FT_SYMLINK, | |
ac27a0ec DK |
1246 | }; |
1247 | ||
617ba13b MC |
1248 | static inline void ext4_set_de_type(struct super_block *sb, |
1249 | struct ext4_dir_entry_2 *de, | |
ac27a0ec | 1250 | umode_t mode) { |
617ba13b MC |
1251 | if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FILETYPE)) |
1252 | de->file_type = ext4_type_by_mode[(mode & S_IFMT)>>S_SHIFT]; | |
ac27a0ec DK |
1253 | } |
1254 | ||
ef2b02d3 ES |
1255 | /* |
1256 | * Move count entries from end of map between two memory locations. | |
1257 | * Returns pointer to last entry moved. | |
1258 | */ | |
617ba13b | 1259 | static struct ext4_dir_entry_2 * |
3d0518f4 WY |
1260 | dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count, |
1261 | unsigned blocksize) | |
ac27a0ec DK |
1262 | { |
1263 | unsigned rec_len = 0; | |
1264 | ||
1265 | while (count--) { | |
60e6679e | 1266 | struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *) |
9aee2286 | 1267 | (from + (map->offs<<2)); |
617ba13b | 1268 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec | 1269 | memcpy (to, de, rec_len); |
617ba13b | 1270 | ((struct ext4_dir_entry_2 *) to)->rec_len = |
3d0518f4 | 1271 | ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec DK |
1272 | de->inode = 0; |
1273 | map++; | |
1274 | to += rec_len; | |
1275 | } | |
617ba13b | 1276 | return (struct ext4_dir_entry_2 *) (to - rec_len); |
ac27a0ec DK |
1277 | } |
1278 | ||
ef2b02d3 ES |
1279 | /* |
1280 | * Compact each dir entry in the range to the minimal rec_len. | |
1281 | * Returns pointer to last entry in range. | |
1282 | */ | |
8bad4597 | 1283 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize) |
ac27a0ec | 1284 | { |
617ba13b | 1285 | struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base; |
ac27a0ec DK |
1286 | unsigned rec_len = 0; |
1287 | ||
1288 | prev = to = de; | |
8bad4597 | 1289 | while ((char*)de < base + blocksize) { |
3d0518f4 | 1290 | next = ext4_next_entry(de, blocksize); |
ac27a0ec | 1291 | if (de->inode && de->name_len) { |
617ba13b | 1292 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
1293 | if (de > to) |
1294 | memmove(to, de, rec_len); | |
3d0518f4 | 1295 | to->rec_len = ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec | 1296 | prev = to; |
617ba13b | 1297 | to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len); |
ac27a0ec DK |
1298 | } |
1299 | de = next; | |
1300 | } | |
1301 | return prev; | |
1302 | } | |
1303 | ||
ef2b02d3 ES |
1304 | /* |
1305 | * Split a full leaf block to make room for a new dir entry. | |
1306 | * Allocate a new block, and move entries so that they are approx. equally full. | |
1307 | * Returns pointer to de in block into which the new entry will be inserted. | |
1308 | */ | |
617ba13b | 1309 | static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, |
ac27a0ec DK |
1310 | struct buffer_head **bh,struct dx_frame *frame, |
1311 | struct dx_hash_info *hinfo, int *error) | |
1312 | { | |
1313 | unsigned blocksize = dir->i_sb->s_blocksize; | |
1314 | unsigned count, continued; | |
1315 | struct buffer_head *bh2; | |
725d26d3 | 1316 | ext4_lblk_t newblock; |
ac27a0ec DK |
1317 | u32 hash2; |
1318 | struct dx_map_entry *map; | |
1319 | char *data1 = (*bh)->b_data, *data2; | |
59e315b4 | 1320 | unsigned split, move, size; |
617ba13b | 1321 | struct ext4_dir_entry_2 *de = NULL, *de2; |
59e315b4 | 1322 | int err = 0, i; |
ac27a0ec | 1323 | |
fedee54d | 1324 | bh2 = ext4_append (handle, dir, &newblock, &err); |
ac27a0ec DK |
1325 | if (!(bh2)) { |
1326 | brelse(*bh); | |
1327 | *bh = NULL; | |
1328 | goto errout; | |
1329 | } | |
1330 | ||
1331 | BUFFER_TRACE(*bh, "get_write_access"); | |
617ba13b | 1332 | err = ext4_journal_get_write_access(handle, *bh); |
fedee54d DM |
1333 | if (err) |
1334 | goto journal_error; | |
1335 | ||
ac27a0ec | 1336 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 1337 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
1338 | if (err) |
1339 | goto journal_error; | |
1340 | ||
1341 | data2 = bh2->b_data; | |
1342 | ||
1343 | /* create map in the end of data2 block */ | |
1344 | map = (struct dx_map_entry *) (data2 + blocksize); | |
af5bc92d | 1345 | count = dx_make_map((struct ext4_dir_entry_2 *) data1, |
ac27a0ec DK |
1346 | blocksize, hinfo, map); |
1347 | map -= count; | |
af5bc92d | 1348 | dx_sort_map(map, count); |
ef2b02d3 ES |
1349 | /* Split the existing block in the middle, size-wise */ |
1350 | size = 0; | |
1351 | move = 0; | |
1352 | for (i = count-1; i >= 0; i--) { | |
1353 | /* is more than half of this entry in 2nd half of the block? */ | |
1354 | if (size + map[i].size/2 > blocksize/2) | |
1355 | break; | |
1356 | size += map[i].size; | |
1357 | move++; | |
1358 | } | |
1359 | /* map index at which we will split */ | |
1360 | split = count - move; | |
ac27a0ec DK |
1361 | hash2 = map[split].hash; |
1362 | continued = hash2 == map[split - 1].hash; | |
725d26d3 AK |
1363 | dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n", |
1364 | (unsigned long)dx_get_block(frame->at), | |
1365 | hash2, split, count-split)); | |
ac27a0ec DK |
1366 | |
1367 | /* Fancy dance to stay within two buffers */ | |
3d0518f4 | 1368 | de2 = dx_move_dirents(data1, data2, map + split, count - split, blocksize); |
af5bc92d | 1369 | de = dx_pack_dirents(data1, blocksize); |
3d0518f4 WY |
1370 | de->rec_len = ext4_rec_len_to_disk(data1 + blocksize - (char *) de, |
1371 | blocksize); | |
1372 | de2->rec_len = ext4_rec_len_to_disk(data2 + blocksize - (char *) de2, | |
1373 | blocksize); | |
617ba13b MC |
1374 | dxtrace(dx_show_leaf (hinfo, (struct ext4_dir_entry_2 *) data1, blocksize, 1)); |
1375 | dxtrace(dx_show_leaf (hinfo, (struct ext4_dir_entry_2 *) data2, blocksize, 1)); | |
ac27a0ec DK |
1376 | |
1377 | /* Which block gets the new entry? */ | |
1378 | if (hinfo->hash >= hash2) | |
1379 | { | |
1380 | swap(*bh, bh2); | |
1381 | de = de2; | |
1382 | } | |
af5bc92d | 1383 | dx_insert_block(frame, hash2 + continued, newblock); |
0390131b | 1384 | err = ext4_handle_dirty_metadata(handle, dir, bh2); |
ac27a0ec DK |
1385 | if (err) |
1386 | goto journal_error; | |
dbe89444 | 1387 | err = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
ac27a0ec DK |
1388 | if (err) |
1389 | goto journal_error; | |
af5bc92d TT |
1390 | brelse(bh2); |
1391 | dxtrace(dx_show_index("frame", frame->entries)); | |
ac27a0ec | 1392 | return de; |
fedee54d DM |
1393 | |
1394 | journal_error: | |
1395 | brelse(*bh); | |
1396 | brelse(bh2); | |
1397 | *bh = NULL; | |
1398 | ext4_std_error(dir->i_sb, err); | |
1399 | errout: | |
1400 | *error = err; | |
1401 | return NULL; | |
ac27a0ec | 1402 | } |
ac27a0ec DK |
1403 | |
1404 | /* | |
1405 | * Add a new entry into a directory (leaf) block. If de is non-NULL, | |
1406 | * it points to a directory entry which is guaranteed to be large | |
1407 | * enough for new directory entry. If de is NULL, then | |
1408 | * add_dirent_to_buf will attempt search the directory block for | |
1409 | * space. It will return -ENOSPC if no space is available, and -EIO | |
1410 | * and -EEXIST if directory entry already exists. | |
ac27a0ec DK |
1411 | */ |
1412 | static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry, | |
617ba13b | 1413 | struct inode *inode, struct ext4_dir_entry_2 *de, |
af5bc92d | 1414 | struct buffer_head *bh) |
ac27a0ec DK |
1415 | { |
1416 | struct inode *dir = dentry->d_parent->d_inode; | |
1417 | const char *name = dentry->d_name.name; | |
1418 | int namelen = dentry->d_name.len; | |
498e5f24 | 1419 | unsigned int offset = 0; |
3d0518f4 | 1420 | unsigned int blocksize = dir->i_sb->s_blocksize; |
ac27a0ec DK |
1421 | unsigned short reclen; |
1422 | int nlen, rlen, err; | |
1423 | char *top; | |
1424 | ||
617ba13b | 1425 | reclen = EXT4_DIR_REC_LEN(namelen); |
ac27a0ec | 1426 | if (!de) { |
617ba13b | 1427 | de = (struct ext4_dir_entry_2 *)bh->b_data; |
3d0518f4 | 1428 | top = bh->b_data + blocksize - reclen; |
ac27a0ec | 1429 | while ((char *) de <= top) { |
f7c21177 | 1430 | if (ext4_check_dir_entry(dir, NULL, de, bh, offset)) |
ac27a0ec | 1431 | return -EIO; |
2de770a4 | 1432 | if (ext4_match(namelen, name, de)) |
ac27a0ec | 1433 | return -EEXIST; |
617ba13b | 1434 | nlen = EXT4_DIR_REC_LEN(de->name_len); |
3d0518f4 | 1435 | rlen = ext4_rec_len_from_disk(de->rec_len, blocksize); |
ac27a0ec DK |
1436 | if ((de->inode? rlen - nlen: rlen) >= reclen) |
1437 | break; | |
617ba13b | 1438 | de = (struct ext4_dir_entry_2 *)((char *)de + rlen); |
ac27a0ec DK |
1439 | offset += rlen; |
1440 | } | |
1441 | if ((char *) de > top) | |
1442 | return -ENOSPC; | |
1443 | } | |
1444 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 1445 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1446 | if (err) { |
617ba13b | 1447 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1448 | return err; |
1449 | } | |
1450 | ||
1451 | /* By now the buffer is marked for journaling */ | |
617ba13b | 1452 | nlen = EXT4_DIR_REC_LEN(de->name_len); |
3d0518f4 | 1453 | rlen = ext4_rec_len_from_disk(de->rec_len, blocksize); |
ac27a0ec | 1454 | if (de->inode) { |
617ba13b | 1455 | struct ext4_dir_entry_2 *de1 = (struct ext4_dir_entry_2 *)((char *)de + nlen); |
3d0518f4 WY |
1456 | de1->rec_len = ext4_rec_len_to_disk(rlen - nlen, blocksize); |
1457 | de->rec_len = ext4_rec_len_to_disk(nlen, blocksize); | |
ac27a0ec DK |
1458 | de = de1; |
1459 | } | |
617ba13b | 1460 | de->file_type = EXT4_FT_UNKNOWN; |
ac27a0ec DK |
1461 | if (inode) { |
1462 | de->inode = cpu_to_le32(inode->i_ino); | |
617ba13b | 1463 | ext4_set_de_type(dir->i_sb, de, inode->i_mode); |
ac27a0ec DK |
1464 | } else |
1465 | de->inode = 0; | |
1466 | de->name_len = namelen; | |
af5bc92d | 1467 | memcpy(de->name, name, namelen); |
ac27a0ec DK |
1468 | /* |
1469 | * XXX shouldn't update any times until successful | |
1470 | * completion of syscall, but too many callers depend | |
1471 | * on this. | |
1472 | * | |
1473 | * XXX similarly, too many callers depend on | |
617ba13b | 1474 | * ext4_new_inode() setting the times, but error |
ac27a0ec DK |
1475 | * recovery deletes the inode, so the worst that can |
1476 | * happen is that the times are slightly out of date | |
1477 | * and/or different from the directory change time. | |
1478 | */ | |
ef7f3835 | 1479 | dir->i_mtime = dir->i_ctime = ext4_current_time(dir); |
617ba13b | 1480 | ext4_update_dx_flag(dir); |
ac27a0ec | 1481 | dir->i_version++; |
617ba13b | 1482 | ext4_mark_inode_dirty(handle, dir); |
0390131b FM |
1483 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); |
1484 | err = ext4_handle_dirty_metadata(handle, dir, bh); | |
ac27a0ec | 1485 | if (err) |
617ba13b | 1486 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1487 | return 0; |
1488 | } | |
1489 | ||
ac27a0ec DK |
1490 | /* |
1491 | * This converts a one block unindexed directory to a 3 block indexed | |
1492 | * directory, and adds the dentry to the indexed directory. | |
1493 | */ | |
1494 | static int make_indexed_dir(handle_t *handle, struct dentry *dentry, | |
1495 | struct inode *inode, struct buffer_head *bh) | |
1496 | { | |
1497 | struct inode *dir = dentry->d_parent->d_inode; | |
1498 | const char *name = dentry->d_name.name; | |
1499 | int namelen = dentry->d_name.len; | |
1500 | struct buffer_head *bh2; | |
1501 | struct dx_root *root; | |
1502 | struct dx_frame frames[2], *frame; | |
1503 | struct dx_entry *entries; | |
617ba13b | 1504 | struct ext4_dir_entry_2 *de, *de2; |
ac27a0ec DK |
1505 | char *data1, *top; |
1506 | unsigned len; | |
1507 | int retval; | |
1508 | unsigned blocksize; | |
1509 | struct dx_hash_info hinfo; | |
725d26d3 | 1510 | ext4_lblk_t block; |
ac27a0ec DK |
1511 | struct fake_dirent *fde; |
1512 | ||
1513 | blocksize = dir->i_sb->s_blocksize; | |
e6b8bc09 | 1514 | dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino)); |
617ba13b | 1515 | retval = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1516 | if (retval) { |
617ba13b | 1517 | ext4_std_error(dir->i_sb, retval); |
ac27a0ec DK |
1518 | brelse(bh); |
1519 | return retval; | |
1520 | } | |
1521 | root = (struct dx_root *) bh->b_data; | |
1522 | ||
e6b8bc09 TT |
1523 | /* The 0th block becomes the root, move the dirents out */ |
1524 | fde = &root->dotdot; | |
1525 | de = (struct ext4_dir_entry_2 *)((char *)fde + | |
3d0518f4 | 1526 | ext4_rec_len_from_disk(fde->rec_len, blocksize)); |
e6b8bc09 | 1527 | if ((char *) de >= (((char *) root) + blocksize)) { |
24676da4 | 1528 | EXT4_ERROR_INODE(dir, "invalid rec_len for '..'"); |
e6b8bc09 TT |
1529 | brelse(bh); |
1530 | return -EIO; | |
1531 | } | |
1532 | len = ((char *) root) + blocksize - (char *) de; | |
1533 | ||
1534 | /* Allocate new block for the 0th block's dirents */ | |
af5bc92d | 1535 | bh2 = ext4_append(handle, dir, &block, &retval); |
ac27a0ec DK |
1536 | if (!(bh2)) { |
1537 | brelse(bh); | |
1538 | return retval; | |
1539 | } | |
12e9b892 | 1540 | ext4_set_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec DK |
1541 | data1 = bh2->b_data; |
1542 | ||
ac27a0ec | 1543 | memcpy (data1, de, len); |
617ba13b | 1544 | de = (struct ext4_dir_entry_2 *) data1; |
ac27a0ec | 1545 | top = data1 + len; |
3d0518f4 | 1546 | while ((char *)(de2 = ext4_next_entry(de, blocksize)) < top) |
ac27a0ec | 1547 | de = de2; |
3d0518f4 WY |
1548 | de->rec_len = ext4_rec_len_to_disk(data1 + blocksize - (char *) de, |
1549 | blocksize); | |
ac27a0ec | 1550 | /* Initialize the root; the dot dirents already exist */ |
617ba13b | 1551 | de = (struct ext4_dir_entry_2 *) (&root->dotdot); |
3d0518f4 WY |
1552 | de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(2), |
1553 | blocksize); | |
ac27a0ec DK |
1554 | memset (&root->info, 0, sizeof(root->info)); |
1555 | root->info.info_length = sizeof(root->info); | |
617ba13b | 1556 | root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
ac27a0ec | 1557 | entries = root->entries; |
af5bc92d TT |
1558 | dx_set_block(entries, 1); |
1559 | dx_set_count(entries, 1); | |
1560 | dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info))); | |
ac27a0ec DK |
1561 | |
1562 | /* Initialize as for dx_probe */ | |
1563 | hinfo.hash_version = root->info.hash_version; | |
f99b2589 TT |
1564 | if (hinfo.hash_version <= DX_HASH_TEA) |
1565 | hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b MC |
1566 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
1567 | ext4fs_dirhash(name, namelen, &hinfo); | |
ac27a0ec DK |
1568 | frame = frames; |
1569 | frame->entries = entries; | |
1570 | frame->at = entries; | |
1571 | frame->bh = bh; | |
1572 | bh = bh2; | |
6976a6f2 | 1573 | |
dbe89444 | 1574 | ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
6976a6f2 AH |
1575 | ext4_handle_dirty_metadata(handle, dir, bh); |
1576 | ||
ac27a0ec | 1577 | de = do_split(handle,dir, &bh, frame, &hinfo, &retval); |
7ad8e4e6 JK |
1578 | if (!de) { |
1579 | /* | |
1580 | * Even if the block split failed, we have to properly write | |
1581 | * out all the changes we did so far. Otherwise we can end up | |
1582 | * with corrupted filesystem. | |
1583 | */ | |
1584 | ext4_mark_inode_dirty(handle, dir); | |
7ad8e4e6 | 1585 | dx_release(frames); |
ac27a0ec | 1586 | return retval; |
7ad8e4e6 JK |
1587 | } |
1588 | dx_release(frames); | |
ac27a0ec | 1589 | |
2de770a4 TT |
1590 | retval = add_dirent_to_buf(handle, dentry, inode, de, bh); |
1591 | brelse(bh); | |
1592 | return retval; | |
ac27a0ec | 1593 | } |
ac27a0ec DK |
1594 | |
1595 | /* | |
617ba13b | 1596 | * ext4_add_entry() |
ac27a0ec DK |
1597 | * |
1598 | * adds a file entry to the specified directory, using the same | |
617ba13b | 1599 | * semantics as ext4_find_entry(). It returns NULL if it failed. |
ac27a0ec DK |
1600 | * |
1601 | * NOTE!! The inode part of 'de' is left at 0 - which means you | |
1602 | * may not sleep between calling this and putting something into | |
1603 | * the entry, as someone else might have used it while you slept. | |
1604 | */ | |
af5bc92d TT |
1605 | static int ext4_add_entry(handle_t *handle, struct dentry *dentry, |
1606 | struct inode *inode) | |
ac27a0ec DK |
1607 | { |
1608 | struct inode *dir = dentry->d_parent->d_inode; | |
af5bc92d | 1609 | struct buffer_head *bh; |
617ba13b | 1610 | struct ext4_dir_entry_2 *de; |
af5bc92d | 1611 | struct super_block *sb; |
ac27a0ec | 1612 | int retval; |
ac27a0ec | 1613 | int dx_fallback=0; |
ac27a0ec | 1614 | unsigned blocksize; |
725d26d3 | 1615 | ext4_lblk_t block, blocks; |
ac27a0ec DK |
1616 | |
1617 | sb = dir->i_sb; | |
1618 | blocksize = sb->s_blocksize; | |
1619 | if (!dentry->d_name.len) | |
1620 | return -EINVAL; | |
ac27a0ec | 1621 | if (is_dx(dir)) { |
617ba13b | 1622 | retval = ext4_dx_add_entry(handle, dentry, inode); |
ac27a0ec DK |
1623 | if (!retval || (retval != ERR_BAD_DX_DIR)) |
1624 | return retval; | |
12e9b892 | 1625 | ext4_clear_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec | 1626 | dx_fallback++; |
617ba13b | 1627 | ext4_mark_inode_dirty(handle, dir); |
ac27a0ec | 1628 | } |
ac27a0ec | 1629 | blocks = dir->i_size >> sb->s_blocksize_bits; |
498e5f24 | 1630 | for (block = 0; block < blocks; block++) { |
617ba13b | 1631 | bh = ext4_bread(handle, dir, block, 0, &retval); |
ac27a0ec DK |
1632 | if(!bh) |
1633 | return retval; | |
1634 | retval = add_dirent_to_buf(handle, dentry, inode, NULL, bh); | |
2de770a4 TT |
1635 | if (retval != -ENOSPC) { |
1636 | brelse(bh); | |
ac27a0ec | 1637 | return retval; |
2de770a4 | 1638 | } |
ac27a0ec | 1639 | |
ac27a0ec | 1640 | if (blocks == 1 && !dx_fallback && |
1e424a34 TT |
1641 | EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) |
1642 | return make_indexed_dir(handle, dentry, inode, bh); | |
ac27a0ec DK |
1643 | brelse(bh); |
1644 | } | |
617ba13b | 1645 | bh = ext4_append(handle, dir, &block, &retval); |
ac27a0ec DK |
1646 | if (!bh) |
1647 | return retval; | |
617ba13b | 1648 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 1649 | de->inode = 0; |
3d0518f4 | 1650 | de->rec_len = ext4_rec_len_to_disk(blocksize, blocksize); |
2de770a4 TT |
1651 | retval = add_dirent_to_buf(handle, dentry, inode, de, bh); |
1652 | brelse(bh); | |
14ece102 FM |
1653 | if (retval == 0) |
1654 | ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY); | |
2de770a4 | 1655 | return retval; |
ac27a0ec DK |
1656 | } |
1657 | ||
ac27a0ec DK |
1658 | /* |
1659 | * Returns 0 for success, or a negative error value | |
1660 | */ | |
617ba13b | 1661 | static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry, |
ac27a0ec DK |
1662 | struct inode *inode) |
1663 | { | |
1664 | struct dx_frame frames[2], *frame; | |
1665 | struct dx_entry *entries, *at; | |
1666 | struct dx_hash_info hinfo; | |
af5bc92d | 1667 | struct buffer_head *bh; |
ac27a0ec | 1668 | struct inode *dir = dentry->d_parent->d_inode; |
af5bc92d | 1669 | struct super_block *sb = dir->i_sb; |
617ba13b | 1670 | struct ext4_dir_entry_2 *de; |
ac27a0ec DK |
1671 | int err; |
1672 | ||
f702ba0f | 1673 | frame = dx_probe(&dentry->d_name, dir, &hinfo, frames, &err); |
ac27a0ec DK |
1674 | if (!frame) |
1675 | return err; | |
1676 | entries = frame->entries; | |
1677 | at = frame->at; | |
1678 | ||
617ba13b | 1679 | if (!(bh = ext4_bread(handle,dir, dx_get_block(frame->at), 0, &err))) |
ac27a0ec DK |
1680 | goto cleanup; |
1681 | ||
1682 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 1683 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec DK |
1684 | if (err) |
1685 | goto journal_error; | |
1686 | ||
1687 | err = add_dirent_to_buf(handle, dentry, inode, NULL, bh); | |
2de770a4 | 1688 | if (err != -ENOSPC) |
ac27a0ec | 1689 | goto cleanup; |
ac27a0ec DK |
1690 | |
1691 | /* Block full, should compress but for now just split */ | |
4776004f | 1692 | dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n", |
ac27a0ec DK |
1693 | dx_get_count(entries), dx_get_limit(entries))); |
1694 | /* Need to split index? */ | |
1695 | if (dx_get_count(entries) == dx_get_limit(entries)) { | |
725d26d3 | 1696 | ext4_lblk_t newblock; |
ac27a0ec DK |
1697 | unsigned icount = dx_get_count(entries); |
1698 | int levels = frame - frames; | |
1699 | struct dx_entry *entries2; | |
1700 | struct dx_node *node2; | |
1701 | struct buffer_head *bh2; | |
1702 | ||
1703 | if (levels && (dx_get_count(frames->entries) == | |
1704 | dx_get_limit(frames->entries))) { | |
12062ddd | 1705 | ext4_warning(sb, "Directory index full!"); |
ac27a0ec DK |
1706 | err = -ENOSPC; |
1707 | goto cleanup; | |
1708 | } | |
617ba13b | 1709 | bh2 = ext4_append (handle, dir, &newblock, &err); |
ac27a0ec DK |
1710 | if (!(bh2)) |
1711 | goto cleanup; | |
1712 | node2 = (struct dx_node *)(bh2->b_data); | |
1713 | entries2 = node2->entries; | |
1f7bebb9 | 1714 | memset(&node2->fake, 0, sizeof(struct fake_dirent)); |
3d0518f4 WY |
1715 | node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize, |
1716 | sb->s_blocksize); | |
ac27a0ec | 1717 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 1718 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
1719 | if (err) |
1720 | goto journal_error; | |
1721 | if (levels) { | |
1722 | unsigned icount1 = icount/2, icount2 = icount - icount1; | |
1723 | unsigned hash2 = dx_get_hash(entries + icount1); | |
4776004f TT |
1724 | dxtrace(printk(KERN_DEBUG "Split index %i/%i\n", |
1725 | icount1, icount2)); | |
ac27a0ec DK |
1726 | |
1727 | BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ | |
617ba13b | 1728 | err = ext4_journal_get_write_access(handle, |
ac27a0ec DK |
1729 | frames[0].bh); |
1730 | if (err) | |
1731 | goto journal_error; | |
1732 | ||
af5bc92d TT |
1733 | memcpy((char *) entries2, (char *) (entries + icount1), |
1734 | icount2 * sizeof(struct dx_entry)); | |
1735 | dx_set_count(entries, icount1); | |
1736 | dx_set_count(entries2, icount2); | |
1737 | dx_set_limit(entries2, dx_node_limit(dir)); | |
ac27a0ec DK |
1738 | |
1739 | /* Which index block gets the new entry? */ | |
1740 | if (at - entries >= icount1) { | |
1741 | frame->at = at = at - entries - icount1 + entries2; | |
1742 | frame->entries = entries = entries2; | |
1743 | swap(frame->bh, bh2); | |
1744 | } | |
af5bc92d TT |
1745 | dx_insert_block(frames + 0, hash2, newblock); |
1746 | dxtrace(dx_show_index("node", frames[1].entries)); | |
1747 | dxtrace(dx_show_index("node", | |
ac27a0ec | 1748 | ((struct dx_node *) bh2->b_data)->entries)); |
dbe89444 | 1749 | err = ext4_handle_dirty_dx_node(handle, dir, bh2); |
ac27a0ec DK |
1750 | if (err) |
1751 | goto journal_error; | |
1752 | brelse (bh2); | |
1753 | } else { | |
4776004f TT |
1754 | dxtrace(printk(KERN_DEBUG |
1755 | "Creating second level index...\n")); | |
ac27a0ec DK |
1756 | memcpy((char *) entries2, (char *) entries, |
1757 | icount * sizeof(struct dx_entry)); | |
1758 | dx_set_limit(entries2, dx_node_limit(dir)); | |
1759 | ||
1760 | /* Set up root */ | |
1761 | dx_set_count(entries, 1); | |
1762 | dx_set_block(entries + 0, newblock); | |
1763 | ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; | |
1764 | ||
1765 | /* Add new access path frame */ | |
1766 | frame = frames + 1; | |
1767 | frame->at = at = at - entries + entries2; | |
1768 | frame->entries = entries = entries2; | |
1769 | frame->bh = bh2; | |
617ba13b | 1770 | err = ext4_journal_get_write_access(handle, |
ac27a0ec DK |
1771 | frame->bh); |
1772 | if (err) | |
1773 | goto journal_error; | |
1774 | } | |
dbe89444 | 1775 | err = ext4_handle_dirty_dx_node(handle, dir, frames[0].bh); |
b4097142 TT |
1776 | if (err) { |
1777 | ext4_std_error(inode->i_sb, err); | |
1778 | goto cleanup; | |
1779 | } | |
ac27a0ec DK |
1780 | } |
1781 | de = do_split(handle, dir, &bh, frame, &hinfo, &err); | |
1782 | if (!de) | |
1783 | goto cleanup; | |
1784 | err = add_dirent_to_buf(handle, dentry, inode, de, bh); | |
ac27a0ec DK |
1785 | goto cleanup; |
1786 | ||
1787 | journal_error: | |
617ba13b | 1788 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1789 | cleanup: |
1790 | if (bh) | |
1791 | brelse(bh); | |
1792 | dx_release(frames); | |
1793 | return err; | |
1794 | } | |
ac27a0ec DK |
1795 | |
1796 | /* | |
617ba13b | 1797 | * ext4_delete_entry deletes a directory entry by merging it with the |
ac27a0ec DK |
1798 | * previous entry |
1799 | */ | |
af5bc92d TT |
1800 | static int ext4_delete_entry(handle_t *handle, |
1801 | struct inode *dir, | |
1802 | struct ext4_dir_entry_2 *de_del, | |
1803 | struct buffer_head *bh) | |
ac27a0ec | 1804 | { |
af5bc92d | 1805 | struct ext4_dir_entry_2 *de, *pde; |
3d0518f4 | 1806 | unsigned int blocksize = dir->i_sb->s_blocksize; |
b4097142 | 1807 | int i, err; |
ac27a0ec DK |
1808 | |
1809 | i = 0; | |
1810 | pde = NULL; | |
617ba13b | 1811 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 1812 | while (i < bh->b_size) { |
f7c21177 | 1813 | if (ext4_check_dir_entry(dir, NULL, de, bh, i)) |
ac27a0ec DK |
1814 | return -EIO; |
1815 | if (de == de_del) { | |
1816 | BUFFER_TRACE(bh, "get_write_access"); | |
b4097142 TT |
1817 | err = ext4_journal_get_write_access(handle, bh); |
1818 | if (unlikely(err)) { | |
1819 | ext4_std_error(dir->i_sb, err); | |
1820 | return err; | |
1821 | } | |
ac27a0ec | 1822 | if (pde) |
a72d7f83 | 1823 | pde->rec_len = ext4_rec_len_to_disk( |
3d0518f4 WY |
1824 | ext4_rec_len_from_disk(pde->rec_len, |
1825 | blocksize) + | |
1826 | ext4_rec_len_from_disk(de->rec_len, | |
1827 | blocksize), | |
1828 | blocksize); | |
ac27a0ec DK |
1829 | else |
1830 | de->inode = 0; | |
1831 | dir->i_version++; | |
0390131b | 1832 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); |
b4097142 TT |
1833 | err = ext4_handle_dirty_metadata(handle, dir, bh); |
1834 | if (unlikely(err)) { | |
1835 | ext4_std_error(dir->i_sb, err); | |
1836 | return err; | |
1837 | } | |
ac27a0ec DK |
1838 | return 0; |
1839 | } | |
3d0518f4 | 1840 | i += ext4_rec_len_from_disk(de->rec_len, blocksize); |
ac27a0ec | 1841 | pde = de; |
3d0518f4 | 1842 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
1843 | } |
1844 | return -ENOENT; | |
1845 | } | |
1846 | ||
f8628a14 AD |
1847 | /* |
1848 | * DIR_NLINK feature is set if 1) nlinks > EXT4_LINK_MAX or 2) nlinks == 2, | |
1849 | * since this indicates that nlinks count was previously 1. | |
1850 | */ | |
1851 | static void ext4_inc_count(handle_t *handle, struct inode *inode) | |
1852 | { | |
1853 | inc_nlink(inode); | |
1854 | if (is_dx(inode) && inode->i_nlink > 1) { | |
1855 | /* limit is 16-bit i_links_count */ | |
1856 | if (inode->i_nlink >= EXT4_LINK_MAX || inode->i_nlink == 2) { | |
bfe86848 | 1857 | set_nlink(inode, 1); |
f8628a14 AD |
1858 | EXT4_SET_RO_COMPAT_FEATURE(inode->i_sb, |
1859 | EXT4_FEATURE_RO_COMPAT_DIR_NLINK); | |
1860 | } | |
1861 | } | |
1862 | } | |
1863 | ||
1864 | /* | |
1865 | * If a directory had nlink == 1, then we should let it be 1. This indicates | |
1866 | * directory has >EXT4_LINK_MAX subdirs. | |
1867 | */ | |
1868 | static void ext4_dec_count(handle_t *handle, struct inode *inode) | |
1869 | { | |
909a4cf1 AD |
1870 | if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2) |
1871 | drop_nlink(inode); | |
f8628a14 AD |
1872 | } |
1873 | ||
1874 | ||
617ba13b | 1875 | static int ext4_add_nondir(handle_t *handle, |
ac27a0ec DK |
1876 | struct dentry *dentry, struct inode *inode) |
1877 | { | |
617ba13b | 1878 | int err = ext4_add_entry(handle, dentry, inode); |
ac27a0ec | 1879 | if (!err) { |
617ba13b | 1880 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec | 1881 | d_instantiate(dentry, inode); |
6b38e842 | 1882 | unlock_new_inode(inode); |
ac27a0ec DK |
1883 | return 0; |
1884 | } | |
731b9a54 | 1885 | drop_nlink(inode); |
6b38e842 | 1886 | unlock_new_inode(inode); |
ac27a0ec DK |
1887 | iput(inode); |
1888 | return err; | |
1889 | } | |
1890 | ||
1891 | /* | |
1892 | * By the time this is called, we already have created | |
1893 | * the directory cache entry for the new file, but it | |
1894 | * is so far negative - it has no inode. | |
1895 | * | |
1896 | * If the create succeeds, we fill in the inode information | |
1897 | * with d_instantiate(). | |
1898 | */ | |
4acdaf27 | 1899 | static int ext4_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
af5bc92d | 1900 | struct nameidata *nd) |
ac27a0ec DK |
1901 | { |
1902 | handle_t *handle; | |
af5bc92d | 1903 | struct inode *inode; |
ac27a0ec DK |
1904 | int err, retries = 0; |
1905 | ||
871a2931 | 1906 | dquot_initialize(dir); |
907f4554 | 1907 | |
ac27a0ec | 1908 | retry: |
617ba13b MC |
1909 | handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
1910 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 + | |
5aca07eb | 1911 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb)); |
ac27a0ec DK |
1912 | if (IS_ERR(handle)) |
1913 | return PTR_ERR(handle); | |
1914 | ||
1915 | if (IS_DIRSYNC(dir)) | |
0390131b | 1916 | ext4_handle_sync(handle); |
ac27a0ec | 1917 | |
5cb81dab | 1918 | inode = ext4_new_inode(handle, dir, mode, &dentry->d_name, 0, NULL); |
ac27a0ec DK |
1919 | err = PTR_ERR(inode); |
1920 | if (!IS_ERR(inode)) { | |
617ba13b MC |
1921 | inode->i_op = &ext4_file_inode_operations; |
1922 | inode->i_fop = &ext4_file_operations; | |
1923 | ext4_set_aops(inode); | |
1924 | err = ext4_add_nondir(handle, dentry, inode); | |
ac27a0ec | 1925 | } |
617ba13b MC |
1926 | ext4_journal_stop(handle); |
1927 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
1928 | goto retry; |
1929 | return err; | |
1930 | } | |
1931 | ||
af5bc92d | 1932 | static int ext4_mknod(struct inode *dir, struct dentry *dentry, |
1a67aafb | 1933 | umode_t mode, dev_t rdev) |
ac27a0ec DK |
1934 | { |
1935 | handle_t *handle; | |
1936 | struct inode *inode; | |
1937 | int err, retries = 0; | |
1938 | ||
1939 | if (!new_valid_dev(rdev)) | |
1940 | return -EINVAL; | |
1941 | ||
871a2931 | 1942 | dquot_initialize(dir); |
907f4554 | 1943 | |
ac27a0ec | 1944 | retry: |
617ba13b MC |
1945 | handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
1946 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 + | |
5aca07eb | 1947 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb)); |
ac27a0ec DK |
1948 | if (IS_ERR(handle)) |
1949 | return PTR_ERR(handle); | |
1950 | ||
1951 | if (IS_DIRSYNC(dir)) | |
0390131b | 1952 | ext4_handle_sync(handle); |
ac27a0ec | 1953 | |
5cb81dab | 1954 | inode = ext4_new_inode(handle, dir, mode, &dentry->d_name, 0, NULL); |
ac27a0ec DK |
1955 | err = PTR_ERR(inode); |
1956 | if (!IS_ERR(inode)) { | |
1957 | init_special_inode(inode, inode->i_mode, rdev); | |
03010a33 | 1958 | #ifdef CONFIG_EXT4_FS_XATTR |
617ba13b | 1959 | inode->i_op = &ext4_special_inode_operations; |
ac27a0ec | 1960 | #endif |
617ba13b | 1961 | err = ext4_add_nondir(handle, dentry, inode); |
ac27a0ec | 1962 | } |
617ba13b MC |
1963 | ext4_journal_stop(handle); |
1964 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
1965 | goto retry; |
1966 | return err; | |
1967 | } | |
1968 | ||
18bb1db3 | 1969 | static int ext4_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) |
ac27a0ec DK |
1970 | { |
1971 | handle_t *handle; | |
af5bc92d | 1972 | struct inode *inode; |
dabd991f | 1973 | struct buffer_head *dir_block = NULL; |
af5bc92d | 1974 | struct ext4_dir_entry_2 *de; |
3d0518f4 | 1975 | unsigned int blocksize = dir->i_sb->s_blocksize; |
ac27a0ec DK |
1976 | int err, retries = 0; |
1977 | ||
f8628a14 | 1978 | if (EXT4_DIR_LINK_MAX(dir)) |
ac27a0ec DK |
1979 | return -EMLINK; |
1980 | ||
871a2931 | 1981 | dquot_initialize(dir); |
907f4554 | 1982 | |
ac27a0ec | 1983 | retry: |
617ba13b MC |
1984 | handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
1985 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 + | |
5aca07eb | 1986 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb)); |
ac27a0ec DK |
1987 | if (IS_ERR(handle)) |
1988 | return PTR_ERR(handle); | |
1989 | ||
1990 | if (IS_DIRSYNC(dir)) | |
0390131b | 1991 | ext4_handle_sync(handle); |
ac27a0ec | 1992 | |
11013911 | 1993 | inode = ext4_new_inode(handle, dir, S_IFDIR | mode, |
5cb81dab | 1994 | &dentry->d_name, 0, NULL); |
ac27a0ec DK |
1995 | err = PTR_ERR(inode); |
1996 | if (IS_ERR(inode)) | |
1997 | goto out_stop; | |
1998 | ||
617ba13b MC |
1999 | inode->i_op = &ext4_dir_inode_operations; |
2000 | inode->i_fop = &ext4_dir_operations; | |
2001 | inode->i_size = EXT4_I(inode)->i_disksize = inode->i_sb->s_blocksize; | |
af5bc92d | 2002 | dir_block = ext4_bread(handle, inode, 0, 1, &err); |
4cdeed86 AK |
2003 | if (!dir_block) |
2004 | goto out_clear_inode; | |
ac27a0ec | 2005 | BUFFER_TRACE(dir_block, "get_write_access"); |
dabd991f NK |
2006 | err = ext4_journal_get_write_access(handle, dir_block); |
2007 | if (err) | |
2008 | goto out_clear_inode; | |
617ba13b | 2009 | de = (struct ext4_dir_entry_2 *) dir_block->b_data; |
ac27a0ec DK |
2010 | de->inode = cpu_to_le32(inode->i_ino); |
2011 | de->name_len = 1; | |
3d0518f4 WY |
2012 | de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len), |
2013 | blocksize); | |
af5bc92d | 2014 | strcpy(de->name, "."); |
617ba13b | 2015 | ext4_set_de_type(dir->i_sb, de, S_IFDIR); |
3d0518f4 | 2016 | de = ext4_next_entry(de, blocksize); |
ac27a0ec | 2017 | de->inode = cpu_to_le32(dir->i_ino); |
3d0518f4 WY |
2018 | de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(1), |
2019 | blocksize); | |
ac27a0ec | 2020 | de->name_len = 2; |
af5bc92d | 2021 | strcpy(de->name, ".."); |
617ba13b | 2022 | ext4_set_de_type(dir->i_sb, de, S_IFDIR); |
bfe86848 | 2023 | set_nlink(inode, 2); |
0390131b | 2024 | BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata"); |
f9287c1f | 2025 | err = ext4_handle_dirty_metadata(handle, inode, dir_block); |
dabd991f NK |
2026 | if (err) |
2027 | goto out_clear_inode; | |
2028 | err = ext4_mark_inode_dirty(handle, inode); | |
2029 | if (!err) | |
2030 | err = ext4_add_entry(handle, dentry, inode); | |
ac27a0ec | 2031 | if (err) { |
4cdeed86 AK |
2032 | out_clear_inode: |
2033 | clear_nlink(inode); | |
6b38e842 | 2034 | unlock_new_inode(inode); |
617ba13b | 2035 | ext4_mark_inode_dirty(handle, inode); |
af5bc92d | 2036 | iput(inode); |
ac27a0ec DK |
2037 | goto out_stop; |
2038 | } | |
f8628a14 | 2039 | ext4_inc_count(handle, dir); |
617ba13b | 2040 | ext4_update_dx_flag(dir); |
dabd991f NK |
2041 | err = ext4_mark_inode_dirty(handle, dir); |
2042 | if (err) | |
2043 | goto out_clear_inode; | |
ac27a0ec | 2044 | d_instantiate(dentry, inode); |
6b38e842 | 2045 | unlock_new_inode(inode); |
ac27a0ec | 2046 | out_stop: |
dabd991f | 2047 | brelse(dir_block); |
617ba13b MC |
2048 | ext4_journal_stop(handle); |
2049 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
2050 | goto retry; |
2051 | return err; | |
2052 | } | |
2053 | ||
2054 | /* | |
2055 | * routine to check that the specified directory is empty (for rmdir) | |
2056 | */ | |
af5bc92d | 2057 | static int empty_dir(struct inode *inode) |
ac27a0ec | 2058 | { |
498e5f24 | 2059 | unsigned int offset; |
af5bc92d TT |
2060 | struct buffer_head *bh; |
2061 | struct ext4_dir_entry_2 *de, *de1; | |
2062 | struct super_block *sb; | |
ac27a0ec DK |
2063 | int err = 0; |
2064 | ||
2065 | sb = inode->i_sb; | |
617ba13b | 2066 | if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) || |
af5bc92d | 2067 | !(bh = ext4_bread(NULL, inode, 0, 0, &err))) { |
ac27a0ec | 2068 | if (err) |
24676da4 TT |
2069 | EXT4_ERROR_INODE(inode, |
2070 | "error %d reading directory lblock 0", err); | |
ac27a0ec | 2071 | else |
12062ddd | 2072 | ext4_warning(inode->i_sb, |
ac27a0ec DK |
2073 | "bad directory (dir #%lu) - no data block", |
2074 | inode->i_ino); | |
2075 | return 1; | |
2076 | } | |
617ba13b | 2077 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
3d0518f4 | 2078 | de1 = ext4_next_entry(de, sb->s_blocksize); |
ac27a0ec DK |
2079 | if (le32_to_cpu(de->inode) != inode->i_ino || |
2080 | !le32_to_cpu(de1->inode) || | |
af5bc92d TT |
2081 | strcmp(".", de->name) || |
2082 | strcmp("..", de1->name)) { | |
12062ddd | 2083 | ext4_warning(inode->i_sb, |
af5bc92d TT |
2084 | "bad directory (dir #%lu) - no `.' or `..'", |
2085 | inode->i_ino); | |
2086 | brelse(bh); | |
ac27a0ec DK |
2087 | return 1; |
2088 | } | |
3d0518f4 WY |
2089 | offset = ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) + |
2090 | ext4_rec_len_from_disk(de1->rec_len, sb->s_blocksize); | |
2091 | de = ext4_next_entry(de1, sb->s_blocksize); | |
af5bc92d | 2092 | while (offset < inode->i_size) { |
ac27a0ec | 2093 | if (!bh || |
24676da4 TT |
2094 | (void *) de >= (void *) (bh->b_data+sb->s_blocksize)) { |
2095 | unsigned int lblock; | |
ac27a0ec | 2096 | err = 0; |
af5bc92d | 2097 | brelse(bh); |
24676da4 TT |
2098 | lblock = offset >> EXT4_BLOCK_SIZE_BITS(sb); |
2099 | bh = ext4_bread(NULL, inode, lblock, 0, &err); | |
ac27a0ec DK |
2100 | if (!bh) { |
2101 | if (err) | |
24676da4 TT |
2102 | EXT4_ERROR_INODE(inode, |
2103 | "error %d reading directory " | |
2104 | "lblock %u", err, lblock); | |
ac27a0ec DK |
2105 | offset += sb->s_blocksize; |
2106 | continue; | |
2107 | } | |
617ba13b | 2108 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 2109 | } |
f7c21177 | 2110 | if (ext4_check_dir_entry(inode, NULL, de, bh, offset)) { |
617ba13b | 2111 | de = (struct ext4_dir_entry_2 *)(bh->b_data + |
ac27a0ec DK |
2112 | sb->s_blocksize); |
2113 | offset = (offset | (sb->s_blocksize - 1)) + 1; | |
2114 | continue; | |
2115 | } | |
2116 | if (le32_to_cpu(de->inode)) { | |
af5bc92d | 2117 | brelse(bh); |
ac27a0ec DK |
2118 | return 0; |
2119 | } | |
3d0518f4 WY |
2120 | offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize); |
2121 | de = ext4_next_entry(de, sb->s_blocksize); | |
ac27a0ec | 2122 | } |
af5bc92d | 2123 | brelse(bh); |
ac27a0ec DK |
2124 | return 1; |
2125 | } | |
2126 | ||
617ba13b | 2127 | /* ext4_orphan_add() links an unlinked or truncated inode into a list of |
ac27a0ec DK |
2128 | * such inodes, starting at the superblock, in case we crash before the |
2129 | * file is closed/deleted, or in case the inode truncate spans multiple | |
2130 | * transactions and the last transaction is not recovered after a crash. | |
2131 | * | |
2132 | * At filesystem recovery time, we walk this list deleting unlinked | |
617ba13b | 2133 | * inodes and truncating linked inodes in ext4_orphan_cleanup(). |
ac27a0ec | 2134 | */ |
617ba13b | 2135 | int ext4_orphan_add(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2136 | { |
2137 | struct super_block *sb = inode->i_sb; | |
617ba13b | 2138 | struct ext4_iloc iloc; |
ac27a0ec DK |
2139 | int err = 0, rc; |
2140 | ||
0390131b FM |
2141 | if (!ext4_handle_valid(handle)) |
2142 | return 0; | |
2143 | ||
3b9d4ed2 | 2144 | mutex_lock(&EXT4_SB(sb)->s_orphan_lock); |
617ba13b | 2145 | if (!list_empty(&EXT4_I(inode)->i_orphan)) |
ac27a0ec DK |
2146 | goto out_unlock; |
2147 | ||
afb86178 LC |
2148 | /* |
2149 | * Orphan handling is only valid for files with data blocks | |
2150 | * being truncated, or files being unlinked. Note that we either | |
2151 | * hold i_mutex, or the inode can not be referenced from outside, | |
2152 | * so i_nlink should not be bumped due to race | |
ac27a0ec | 2153 | */ |
af5bc92d TT |
2154 | J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || |
2155 | S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); | |
ac27a0ec | 2156 | |
617ba13b MC |
2157 | BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access"); |
2158 | err = ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh); | |
ac27a0ec DK |
2159 | if (err) |
2160 | goto out_unlock; | |
2161 | ||
617ba13b | 2162 | err = ext4_reserve_inode_write(handle, inode, &iloc); |
ac27a0ec DK |
2163 | if (err) |
2164 | goto out_unlock; | |
6e3617e5 DM |
2165 | /* |
2166 | * Due to previous errors inode may be already a part of on-disk | |
2167 | * orphan list. If so skip on-disk list modification. | |
2168 | */ | |
2169 | if (NEXT_ORPHAN(inode) && NEXT_ORPHAN(inode) <= | |
2170 | (le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))) | |
2171 | goto mem_insert; | |
ac27a0ec DK |
2172 | |
2173 | /* Insert this inode at the head of the on-disk orphan list... */ | |
617ba13b MC |
2174 | NEXT_ORPHAN(inode) = le32_to_cpu(EXT4_SB(sb)->s_es->s_last_orphan); |
2175 | EXT4_SB(sb)->s_es->s_last_orphan = cpu_to_le32(inode->i_ino); | |
a9c47317 | 2176 | err = ext4_handle_dirty_super_now(handle, sb); |
617ba13b | 2177 | rc = ext4_mark_iloc_dirty(handle, inode, &iloc); |
ac27a0ec DK |
2178 | if (!err) |
2179 | err = rc; | |
2180 | ||
2181 | /* Only add to the head of the in-memory list if all the | |
2182 | * previous operations succeeded. If the orphan_add is going to | |
2183 | * fail (possibly taking the journal offline), we can't risk | |
2184 | * leaving the inode on the orphan list: stray orphan-list | |
2185 | * entries can cause panics at unmount time. | |
2186 | * | |
2187 | * This is safe: on error we're going to ignore the orphan list | |
2188 | * anyway on the next recovery. */ | |
6e3617e5 | 2189 | mem_insert: |
ac27a0ec | 2190 | if (!err) |
617ba13b | 2191 | list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan); |
ac27a0ec DK |
2192 | |
2193 | jbd_debug(4, "superblock will point to %lu\n", inode->i_ino); | |
2194 | jbd_debug(4, "orphan inode %lu will point to %d\n", | |
2195 | inode->i_ino, NEXT_ORPHAN(inode)); | |
2196 | out_unlock: | |
3b9d4ed2 | 2197 | mutex_unlock(&EXT4_SB(sb)->s_orphan_lock); |
617ba13b | 2198 | ext4_std_error(inode->i_sb, err); |
ac27a0ec DK |
2199 | return err; |
2200 | } | |
2201 | ||
2202 | /* | |
617ba13b | 2203 | * ext4_orphan_del() removes an unlinked or truncated inode from the list |
ac27a0ec DK |
2204 | * of such inodes stored on disk, because it is finally being cleaned up. |
2205 | */ | |
617ba13b | 2206 | int ext4_orphan_del(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2207 | { |
2208 | struct list_head *prev; | |
617ba13b MC |
2209 | struct ext4_inode_info *ei = EXT4_I(inode); |
2210 | struct ext4_sb_info *sbi; | |
498e5f24 | 2211 | __u32 ino_next; |
617ba13b | 2212 | struct ext4_iloc iloc; |
ac27a0ec DK |
2213 | int err = 0; |
2214 | ||
d3d1faf6 CW |
2215 | /* ext4_handle_valid() assumes a valid handle_t pointer */ |
2216 | if (handle && !ext4_handle_valid(handle)) | |
0390131b FM |
2217 | return 0; |
2218 | ||
3b9d4ed2 TT |
2219 | mutex_lock(&EXT4_SB(inode->i_sb)->s_orphan_lock); |
2220 | if (list_empty(&ei->i_orphan)) | |
2221 | goto out; | |
ac27a0ec DK |
2222 | |
2223 | ino_next = NEXT_ORPHAN(inode); | |
2224 | prev = ei->i_orphan.prev; | |
617ba13b | 2225 | sbi = EXT4_SB(inode->i_sb); |
ac27a0ec DK |
2226 | |
2227 | jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); | |
2228 | ||
2229 | list_del_init(&ei->i_orphan); | |
2230 | ||
2231 | /* If we're on an error path, we may not have a valid | |
2232 | * transaction handle with which to update the orphan list on | |
2233 | * disk, but we still need to remove the inode from the linked | |
2234 | * list in memory. */ | |
0390131b | 2235 | if (sbi->s_journal && !handle) |
ac27a0ec DK |
2236 | goto out; |
2237 | ||
617ba13b | 2238 | err = ext4_reserve_inode_write(handle, inode, &iloc); |
ac27a0ec DK |
2239 | if (err) |
2240 | goto out_err; | |
2241 | ||
2242 | if (prev == &sbi->s_orphan) { | |
498e5f24 | 2243 | jbd_debug(4, "superblock will point to %u\n", ino_next); |
ac27a0ec | 2244 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
617ba13b | 2245 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
ac27a0ec DK |
2246 | if (err) |
2247 | goto out_brelse; | |
2248 | sbi->s_es->s_last_orphan = cpu_to_le32(ino_next); | |
a9c47317 | 2249 | err = ext4_handle_dirty_super_now(handle, inode->i_sb); |
ac27a0ec | 2250 | } else { |
617ba13b | 2251 | struct ext4_iloc iloc2; |
ac27a0ec | 2252 | struct inode *i_prev = |
617ba13b | 2253 | &list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec | 2254 | |
498e5f24 | 2255 | jbd_debug(4, "orphan inode %lu will point to %u\n", |
ac27a0ec | 2256 | i_prev->i_ino, ino_next); |
617ba13b | 2257 | err = ext4_reserve_inode_write(handle, i_prev, &iloc2); |
ac27a0ec DK |
2258 | if (err) |
2259 | goto out_brelse; | |
2260 | NEXT_ORPHAN(i_prev) = ino_next; | |
617ba13b | 2261 | err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2); |
ac27a0ec DK |
2262 | } |
2263 | if (err) | |
2264 | goto out_brelse; | |
2265 | NEXT_ORPHAN(inode) = 0; | |
617ba13b | 2266 | err = ext4_mark_iloc_dirty(handle, inode, &iloc); |
ac27a0ec DK |
2267 | |
2268 | out_err: | |
617ba13b | 2269 | ext4_std_error(inode->i_sb, err); |
ac27a0ec | 2270 | out: |
3b9d4ed2 | 2271 | mutex_unlock(&EXT4_SB(inode->i_sb)->s_orphan_lock); |
ac27a0ec DK |
2272 | return err; |
2273 | ||
2274 | out_brelse: | |
2275 | brelse(iloc.bh); | |
2276 | goto out_err; | |
2277 | } | |
2278 | ||
af5bc92d | 2279 | static int ext4_rmdir(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2280 | { |
2281 | int retval; | |
af5bc92d TT |
2282 | struct inode *inode; |
2283 | struct buffer_head *bh; | |
2284 | struct ext4_dir_entry_2 *de; | |
ac27a0ec DK |
2285 | handle_t *handle; |
2286 | ||
2287 | /* Initialize quotas before so that eventual writes go in | |
2288 | * separate transaction */ | |
871a2931 CH |
2289 | dquot_initialize(dir); |
2290 | dquot_initialize(dentry->d_inode); | |
907f4554 | 2291 | |
617ba13b | 2292 | handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb)); |
ac27a0ec DK |
2293 | if (IS_ERR(handle)) |
2294 | return PTR_ERR(handle); | |
2295 | ||
2296 | retval = -ENOENT; | |
f702ba0f | 2297 | bh = ext4_find_entry(dir, &dentry->d_name, &de); |
ac27a0ec DK |
2298 | if (!bh) |
2299 | goto end_rmdir; | |
2300 | ||
2301 | if (IS_DIRSYNC(dir)) | |
0390131b | 2302 | ext4_handle_sync(handle); |
ac27a0ec DK |
2303 | |
2304 | inode = dentry->d_inode; | |
2305 | ||
2306 | retval = -EIO; | |
2307 | if (le32_to_cpu(de->inode) != inode->i_ino) | |
2308 | goto end_rmdir; | |
2309 | ||
2310 | retval = -ENOTEMPTY; | |
af5bc92d | 2311 | if (!empty_dir(inode)) |
ac27a0ec DK |
2312 | goto end_rmdir; |
2313 | ||
617ba13b | 2314 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
2315 | if (retval) |
2316 | goto end_rmdir; | |
f8628a14 | 2317 | if (!EXT4_DIR_LINK_EMPTY(inode)) |
12062ddd | 2318 | ext4_warning(inode->i_sb, |
af5bc92d TT |
2319 | "empty directory has too many links (%d)", |
2320 | inode->i_nlink); | |
ac27a0ec DK |
2321 | inode->i_version++; |
2322 | clear_nlink(inode); | |
2323 | /* There's no need to set i_disksize: the fact that i_nlink is | |
2324 | * zero will ensure that the right thing happens during any | |
2325 | * recovery. */ | |
2326 | inode->i_size = 0; | |
617ba13b | 2327 | ext4_orphan_add(handle, inode); |
ef7f3835 | 2328 | inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode); |
617ba13b | 2329 | ext4_mark_inode_dirty(handle, inode); |
f8628a14 | 2330 | ext4_dec_count(handle, dir); |
617ba13b MC |
2331 | ext4_update_dx_flag(dir); |
2332 | ext4_mark_inode_dirty(handle, dir); | |
ac27a0ec DK |
2333 | |
2334 | end_rmdir: | |
617ba13b | 2335 | ext4_journal_stop(handle); |
af5bc92d | 2336 | brelse(bh); |
ac27a0ec DK |
2337 | return retval; |
2338 | } | |
2339 | ||
af5bc92d | 2340 | static int ext4_unlink(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2341 | { |
2342 | int retval; | |
af5bc92d TT |
2343 | struct inode *inode; |
2344 | struct buffer_head *bh; | |
2345 | struct ext4_dir_entry_2 *de; | |
ac27a0ec DK |
2346 | handle_t *handle; |
2347 | ||
0562e0ba | 2348 | trace_ext4_unlink_enter(dir, dentry); |
ac27a0ec DK |
2349 | /* Initialize quotas before so that eventual writes go |
2350 | * in separate transaction */ | |
871a2931 CH |
2351 | dquot_initialize(dir); |
2352 | dquot_initialize(dentry->d_inode); | |
907f4554 | 2353 | |
617ba13b | 2354 | handle = ext4_journal_start(dir, EXT4_DELETE_TRANS_BLOCKS(dir->i_sb)); |
ac27a0ec DK |
2355 | if (IS_ERR(handle)) |
2356 | return PTR_ERR(handle); | |
2357 | ||
2358 | if (IS_DIRSYNC(dir)) | |
0390131b | 2359 | ext4_handle_sync(handle); |
ac27a0ec DK |
2360 | |
2361 | retval = -ENOENT; | |
f702ba0f | 2362 | bh = ext4_find_entry(dir, &dentry->d_name, &de); |
ac27a0ec DK |
2363 | if (!bh) |
2364 | goto end_unlink; | |
2365 | ||
2366 | inode = dentry->d_inode; | |
2367 | ||
2368 | retval = -EIO; | |
2369 | if (le32_to_cpu(de->inode) != inode->i_ino) | |
2370 | goto end_unlink; | |
2371 | ||
2372 | if (!inode->i_nlink) { | |
12062ddd | 2373 | ext4_warning(inode->i_sb, |
af5bc92d TT |
2374 | "Deleting nonexistent file (%lu), %d", |
2375 | inode->i_ino, inode->i_nlink); | |
bfe86848 | 2376 | set_nlink(inode, 1); |
ac27a0ec | 2377 | } |
617ba13b | 2378 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
2379 | if (retval) |
2380 | goto end_unlink; | |
ef7f3835 | 2381 | dir->i_ctime = dir->i_mtime = ext4_current_time(dir); |
617ba13b MC |
2382 | ext4_update_dx_flag(dir); |
2383 | ext4_mark_inode_dirty(handle, dir); | |
825f1481 | 2384 | drop_nlink(inode); |
ac27a0ec | 2385 | if (!inode->i_nlink) |
617ba13b | 2386 | ext4_orphan_add(handle, inode); |
ef7f3835 | 2387 | inode->i_ctime = ext4_current_time(inode); |
617ba13b | 2388 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec DK |
2389 | retval = 0; |
2390 | ||
2391 | end_unlink: | |
617ba13b | 2392 | ext4_journal_stop(handle); |
af5bc92d | 2393 | brelse(bh); |
0562e0ba | 2394 | trace_ext4_unlink_exit(dentry, retval); |
ac27a0ec DK |
2395 | return retval; |
2396 | } | |
2397 | ||
af5bc92d TT |
2398 | static int ext4_symlink(struct inode *dir, |
2399 | struct dentry *dentry, const char *symname) | |
ac27a0ec DK |
2400 | { |
2401 | handle_t *handle; | |
af5bc92d | 2402 | struct inode *inode; |
ac27a0ec | 2403 | int l, err, retries = 0; |
df5e6223 | 2404 | int credits; |
ac27a0ec DK |
2405 | |
2406 | l = strlen(symname)+1; | |
2407 | if (l > dir->i_sb->s_blocksize) | |
2408 | return -ENAMETOOLONG; | |
2409 | ||
871a2931 | 2410 | dquot_initialize(dir); |
907f4554 | 2411 | |
df5e6223 JK |
2412 | if (l > EXT4_N_BLOCKS * 4) { |
2413 | /* | |
2414 | * For non-fast symlinks, we just allocate inode and put it on | |
2415 | * orphan list in the first transaction => we need bitmap, | |
8c208719 ES |
2416 | * group descriptor, sb, inode block, quota blocks, and |
2417 | * possibly selinux xattr blocks. | |
df5e6223 | 2418 | */ |
8c208719 ES |
2419 | credits = 4 + EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) + |
2420 | EXT4_XATTR_TRANS_BLOCKS; | |
df5e6223 JK |
2421 | } else { |
2422 | /* | |
2423 | * Fast symlink. We have to add entry to directory | |
2424 | * (EXT4_DATA_TRANS_BLOCKS + EXT4_INDEX_EXTRA_TRANS_BLOCKS), | |
2425 | * allocate new inode (bitmap, group descriptor, inode block, | |
2426 | * quota blocks, sb is already counted in previous macros). | |
2427 | */ | |
2428 | credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
2429 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3 + | |
2430 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb); | |
2431 | } | |
ac27a0ec | 2432 | retry: |
df5e6223 | 2433 | handle = ext4_journal_start(dir, credits); |
ac27a0ec DK |
2434 | if (IS_ERR(handle)) |
2435 | return PTR_ERR(handle); | |
2436 | ||
2437 | if (IS_DIRSYNC(dir)) | |
0390131b | 2438 | ext4_handle_sync(handle); |
ac27a0ec | 2439 | |
11013911 | 2440 | inode = ext4_new_inode(handle, dir, S_IFLNK|S_IRWXUGO, |
5cb81dab | 2441 | &dentry->d_name, 0, NULL); |
ac27a0ec DK |
2442 | err = PTR_ERR(inode); |
2443 | if (IS_ERR(inode)) | |
2444 | goto out_stop; | |
2445 | ||
df5e6223 | 2446 | if (l > EXT4_N_BLOCKS * 4) { |
617ba13b MC |
2447 | inode->i_op = &ext4_symlink_inode_operations; |
2448 | ext4_set_aops(inode); | |
ac27a0ec | 2449 | /* |
df5e6223 JK |
2450 | * We cannot call page_symlink() with transaction started |
2451 | * because it calls into ext4_write_begin() which can wait | |
2452 | * for transaction commit if we are running out of space | |
2453 | * and thus we deadlock. So we have to stop transaction now | |
2454 | * and restart it when symlink contents is written. | |
2455 | * | |
2456 | * To keep fs consistent in case of crash, we have to put inode | |
2457 | * to orphan list in the mean time. | |
ac27a0ec | 2458 | */ |
df5e6223 JK |
2459 | drop_nlink(inode); |
2460 | err = ext4_orphan_add(handle, inode); | |
2461 | ext4_journal_stop(handle); | |
2462 | if (err) | |
2463 | goto err_drop_inode; | |
54566b2c | 2464 | err = __page_symlink(inode, symname, l, 1); |
df5e6223 JK |
2465 | if (err) |
2466 | goto err_drop_inode; | |
2467 | /* | |
2468 | * Now inode is being linked into dir (EXT4_DATA_TRANS_BLOCKS | |
2469 | * + EXT4_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified | |
2470 | */ | |
2471 | handle = ext4_journal_start(dir, | |
2472 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
2473 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 1); | |
2474 | if (IS_ERR(handle)) { | |
2475 | err = PTR_ERR(handle); | |
2476 | goto err_drop_inode; | |
2477 | } | |
0ce8c010 | 2478 | set_nlink(inode, 1); |
df5e6223 | 2479 | err = ext4_orphan_del(handle, inode); |
ac27a0ec | 2480 | if (err) { |
df5e6223 | 2481 | ext4_journal_stop(handle); |
825f1481 | 2482 | clear_nlink(inode); |
df5e6223 | 2483 | goto err_drop_inode; |
ac27a0ec DK |
2484 | } |
2485 | } else { | |
e65187e6 | 2486 | /* clear the extent format for fast symlink */ |
12e9b892 | 2487 | ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS); |
617ba13b | 2488 | inode->i_op = &ext4_fast_symlink_inode_operations; |
af5bc92d | 2489 | memcpy((char *)&EXT4_I(inode)->i_data, symname, l); |
ac27a0ec DK |
2490 | inode->i_size = l-1; |
2491 | } | |
617ba13b MC |
2492 | EXT4_I(inode)->i_disksize = inode->i_size; |
2493 | err = ext4_add_nondir(handle, dentry, inode); | |
ac27a0ec | 2494 | out_stop: |
617ba13b MC |
2495 | ext4_journal_stop(handle); |
2496 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
2497 | goto retry; |
2498 | return err; | |
df5e6223 JK |
2499 | err_drop_inode: |
2500 | unlock_new_inode(inode); | |
2501 | iput(inode); | |
2502 | return err; | |
ac27a0ec DK |
2503 | } |
2504 | ||
af5bc92d TT |
2505 | static int ext4_link(struct dentry *old_dentry, |
2506 | struct inode *dir, struct dentry *dentry) | |
ac27a0ec DK |
2507 | { |
2508 | handle_t *handle; | |
2509 | struct inode *inode = old_dentry->d_inode; | |
2510 | int err, retries = 0; | |
2511 | ||
b05ab1dc | 2512 | if (inode->i_nlink >= EXT4_LINK_MAX) |
ac27a0ec | 2513 | return -EMLINK; |
f8628a14 | 2514 | |
871a2931 | 2515 | dquot_initialize(dir); |
907f4554 | 2516 | |
ac27a0ec | 2517 | retry: |
617ba13b MC |
2518 | handle = ext4_journal_start(dir, EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
2519 | EXT4_INDEX_EXTRA_TRANS_BLOCKS); | |
ac27a0ec DK |
2520 | if (IS_ERR(handle)) |
2521 | return PTR_ERR(handle); | |
2522 | ||
2523 | if (IS_DIRSYNC(dir)) | |
0390131b | 2524 | ext4_handle_sync(handle); |
ac27a0ec | 2525 | |
ef7f3835 | 2526 | inode->i_ctime = ext4_current_time(inode); |
f8628a14 | 2527 | ext4_inc_count(handle, inode); |
7de9c6ee | 2528 | ihold(inode); |
ac27a0ec | 2529 | |
6b38e842 AV |
2530 | err = ext4_add_entry(handle, dentry, inode); |
2531 | if (!err) { | |
2532 | ext4_mark_inode_dirty(handle, inode); | |
2533 | d_instantiate(dentry, inode); | |
2534 | } else { | |
2535 | drop_nlink(inode); | |
2536 | iput(inode); | |
2537 | } | |
617ba13b MC |
2538 | ext4_journal_stop(handle); |
2539 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
2540 | goto retry; |
2541 | return err; | |
2542 | } | |
2543 | ||
3d0518f4 WY |
2544 | #define PARENT_INO(buffer, size) \ |
2545 | (ext4_next_entry((struct ext4_dir_entry_2 *)(buffer), size)->inode) | |
ac27a0ec DK |
2546 | |
2547 | /* | |
2548 | * Anybody can rename anything with this: the permission checks are left to the | |
2549 | * higher-level routines. | |
2550 | */ | |
af5bc92d TT |
2551 | static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry, |
2552 | struct inode *new_dir, struct dentry *new_dentry) | |
ac27a0ec DK |
2553 | { |
2554 | handle_t *handle; | |
af5bc92d TT |
2555 | struct inode *old_inode, *new_inode; |
2556 | struct buffer_head *old_bh, *new_bh, *dir_bh; | |
2557 | struct ext4_dir_entry_2 *old_de, *new_de; | |
8750c6d5 | 2558 | int retval, force_da_alloc = 0; |
ac27a0ec | 2559 | |
871a2931 CH |
2560 | dquot_initialize(old_dir); |
2561 | dquot_initialize(new_dir); | |
907f4554 | 2562 | |
ac27a0ec DK |
2563 | old_bh = new_bh = dir_bh = NULL; |
2564 | ||
2565 | /* Initialize quotas before so that eventual writes go | |
2566 | * in separate transaction */ | |
2567 | if (new_dentry->d_inode) | |
871a2931 | 2568 | dquot_initialize(new_dentry->d_inode); |
617ba13b MC |
2569 | handle = ext4_journal_start(old_dir, 2 * |
2570 | EXT4_DATA_TRANS_BLOCKS(old_dir->i_sb) + | |
2571 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2); | |
ac27a0ec DK |
2572 | if (IS_ERR(handle)) |
2573 | return PTR_ERR(handle); | |
2574 | ||
2575 | if (IS_DIRSYNC(old_dir) || IS_DIRSYNC(new_dir)) | |
0390131b | 2576 | ext4_handle_sync(handle); |
ac27a0ec | 2577 | |
f702ba0f | 2578 | old_bh = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de); |
ac27a0ec DK |
2579 | /* |
2580 | * Check for inode number is _not_ due to possible IO errors. | |
2581 | * We might rmdir the source, keep it as pwd of some process | |
2582 | * and merrily kill the link to whatever was created under the | |
2583 | * same name. Goodbye sticky bit ;-< | |
2584 | */ | |
2585 | old_inode = old_dentry->d_inode; | |
2586 | retval = -ENOENT; | |
2587 | if (!old_bh || le32_to_cpu(old_de->inode) != old_inode->i_ino) | |
2588 | goto end_rename; | |
2589 | ||
2590 | new_inode = new_dentry->d_inode; | |
f702ba0f | 2591 | new_bh = ext4_find_entry(new_dir, &new_dentry->d_name, &new_de); |
ac27a0ec DK |
2592 | if (new_bh) { |
2593 | if (!new_inode) { | |
af5bc92d | 2594 | brelse(new_bh); |
ac27a0ec DK |
2595 | new_bh = NULL; |
2596 | } | |
2597 | } | |
2598 | if (S_ISDIR(old_inode->i_mode)) { | |
2599 | if (new_inode) { | |
2600 | retval = -ENOTEMPTY; | |
af5bc92d | 2601 | if (!empty_dir(new_inode)) |
ac27a0ec DK |
2602 | goto end_rename; |
2603 | } | |
2604 | retval = -EIO; | |
af5bc92d | 2605 | dir_bh = ext4_bread(handle, old_inode, 0, 0, &retval); |
ac27a0ec DK |
2606 | if (!dir_bh) |
2607 | goto end_rename; | |
3d0518f4 WY |
2608 | if (le32_to_cpu(PARENT_INO(dir_bh->b_data, |
2609 | old_dir->i_sb->s_blocksize)) != old_dir->i_ino) | |
ac27a0ec DK |
2610 | goto end_rename; |
2611 | retval = -EMLINK; | |
af5bc92d | 2612 | if (!new_inode && new_dir != old_dir && |
2c94eb86 | 2613 | EXT4_DIR_LINK_MAX(new_dir)) |
ac27a0ec | 2614 | goto end_rename; |
ef607893 AG |
2615 | BUFFER_TRACE(dir_bh, "get_write_access"); |
2616 | retval = ext4_journal_get_write_access(handle, dir_bh); | |
2617 | if (retval) | |
2618 | goto end_rename; | |
ac27a0ec DK |
2619 | } |
2620 | if (!new_bh) { | |
af5bc92d | 2621 | retval = ext4_add_entry(handle, new_dentry, old_inode); |
ac27a0ec DK |
2622 | if (retval) |
2623 | goto end_rename; | |
2624 | } else { | |
2625 | BUFFER_TRACE(new_bh, "get write access"); | |
ef607893 AG |
2626 | retval = ext4_journal_get_write_access(handle, new_bh); |
2627 | if (retval) | |
2628 | goto end_rename; | |
ac27a0ec | 2629 | new_de->inode = cpu_to_le32(old_inode->i_ino); |
617ba13b MC |
2630 | if (EXT4_HAS_INCOMPAT_FEATURE(new_dir->i_sb, |
2631 | EXT4_FEATURE_INCOMPAT_FILETYPE)) | |
ac27a0ec DK |
2632 | new_de->file_type = old_de->file_type; |
2633 | new_dir->i_version++; | |
53b7e9f6 JK |
2634 | new_dir->i_ctime = new_dir->i_mtime = |
2635 | ext4_current_time(new_dir); | |
2636 | ext4_mark_inode_dirty(handle, new_dir); | |
0390131b | 2637 | BUFFER_TRACE(new_bh, "call ext4_handle_dirty_metadata"); |
b4097142 TT |
2638 | retval = ext4_handle_dirty_metadata(handle, new_dir, new_bh); |
2639 | if (unlikely(retval)) { | |
2640 | ext4_std_error(new_dir->i_sb, retval); | |
2641 | goto end_rename; | |
2642 | } | |
ac27a0ec DK |
2643 | brelse(new_bh); |
2644 | new_bh = NULL; | |
2645 | } | |
2646 | ||
2647 | /* | |
2648 | * Like most other Unix systems, set the ctime for inodes on a | |
2649 | * rename. | |
2650 | */ | |
ef7f3835 | 2651 | old_inode->i_ctime = ext4_current_time(old_inode); |
617ba13b | 2652 | ext4_mark_inode_dirty(handle, old_inode); |
ac27a0ec DK |
2653 | |
2654 | /* | |
2655 | * ok, that's it | |
2656 | */ | |
2657 | if (le32_to_cpu(old_de->inode) != old_inode->i_ino || | |
2658 | old_de->name_len != old_dentry->d_name.len || | |
2659 | strncmp(old_de->name, old_dentry->d_name.name, old_de->name_len) || | |
617ba13b | 2660 | (retval = ext4_delete_entry(handle, old_dir, |
ac27a0ec DK |
2661 | old_de, old_bh)) == -ENOENT) { |
2662 | /* old_de could have moved from under us during htree split, so | |
2663 | * make sure that we are deleting the right entry. We might | |
2664 | * also be pointing to a stale entry in the unused part of | |
2665 | * old_bh so just checking inum and the name isn't enough. */ | |
2666 | struct buffer_head *old_bh2; | |
617ba13b | 2667 | struct ext4_dir_entry_2 *old_de2; |
ac27a0ec | 2668 | |
f702ba0f | 2669 | old_bh2 = ext4_find_entry(old_dir, &old_dentry->d_name, &old_de2); |
ac27a0ec | 2670 | if (old_bh2) { |
617ba13b | 2671 | retval = ext4_delete_entry(handle, old_dir, |
ac27a0ec DK |
2672 | old_de2, old_bh2); |
2673 | brelse(old_bh2); | |
2674 | } | |
2675 | } | |
2676 | if (retval) { | |
12062ddd | 2677 | ext4_warning(old_dir->i_sb, |
ac27a0ec DK |
2678 | "Deleting old file (%lu), %d, error=%d", |
2679 | old_dir->i_ino, old_dir->i_nlink, retval); | |
2680 | } | |
2681 | ||
2682 | if (new_inode) { | |
f8628a14 | 2683 | ext4_dec_count(handle, new_inode); |
ef7f3835 | 2684 | new_inode->i_ctime = ext4_current_time(new_inode); |
ac27a0ec | 2685 | } |
ef7f3835 | 2686 | old_dir->i_ctime = old_dir->i_mtime = ext4_current_time(old_dir); |
617ba13b | 2687 | ext4_update_dx_flag(old_dir); |
ac27a0ec | 2688 | if (dir_bh) { |
3d0518f4 WY |
2689 | PARENT_INO(dir_bh->b_data, new_dir->i_sb->s_blocksize) = |
2690 | cpu_to_le32(new_dir->i_ino); | |
0390131b | 2691 | BUFFER_TRACE(dir_bh, "call ext4_handle_dirty_metadata"); |
bcaa9929 | 2692 | retval = ext4_handle_dirty_metadata(handle, old_inode, dir_bh); |
b4097142 TT |
2693 | if (retval) { |
2694 | ext4_std_error(old_dir->i_sb, retval); | |
2695 | goto end_rename; | |
2696 | } | |
f8628a14 | 2697 | ext4_dec_count(handle, old_dir); |
ac27a0ec | 2698 | if (new_inode) { |
f8628a14 | 2699 | /* checked empty_dir above, can't have another parent, |
825f1481 | 2700 | * ext4_dec_count() won't work for many-linked dirs */ |
6d6b77f1 | 2701 | clear_nlink(new_inode); |
ac27a0ec | 2702 | } else { |
f8628a14 | 2703 | ext4_inc_count(handle, new_dir); |
617ba13b MC |
2704 | ext4_update_dx_flag(new_dir); |
2705 | ext4_mark_inode_dirty(handle, new_dir); | |
ac27a0ec DK |
2706 | } |
2707 | } | |
617ba13b | 2708 | ext4_mark_inode_dirty(handle, old_dir); |
ac27a0ec | 2709 | if (new_inode) { |
617ba13b | 2710 | ext4_mark_inode_dirty(handle, new_inode); |
ac27a0ec | 2711 | if (!new_inode->i_nlink) |
617ba13b | 2712 | ext4_orphan_add(handle, new_inode); |
afd4672d TT |
2713 | if (!test_opt(new_dir->i_sb, NO_AUTO_DA_ALLOC)) |
2714 | force_da_alloc = 1; | |
ac27a0ec DK |
2715 | } |
2716 | retval = 0; | |
2717 | ||
2718 | end_rename: | |
af5bc92d TT |
2719 | brelse(dir_bh); |
2720 | brelse(old_bh); | |
2721 | brelse(new_bh); | |
617ba13b | 2722 | ext4_journal_stop(handle); |
8750c6d5 TT |
2723 | if (retval == 0 && force_da_alloc) |
2724 | ext4_alloc_da_blocks(old_inode); | |
ac27a0ec DK |
2725 | return retval; |
2726 | } | |
2727 | ||
2728 | /* | |
2729 | * directories can handle most operations... | |
2730 | */ | |
754661f1 | 2731 | const struct inode_operations ext4_dir_inode_operations = { |
617ba13b MC |
2732 | .create = ext4_create, |
2733 | .lookup = ext4_lookup, | |
2734 | .link = ext4_link, | |
2735 | .unlink = ext4_unlink, | |
2736 | .symlink = ext4_symlink, | |
2737 | .mkdir = ext4_mkdir, | |
2738 | .rmdir = ext4_rmdir, | |
2739 | .mknod = ext4_mknod, | |
2740 | .rename = ext4_rename, | |
2741 | .setattr = ext4_setattr, | |
03010a33 | 2742 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec DK |
2743 | .setxattr = generic_setxattr, |
2744 | .getxattr = generic_getxattr, | |
617ba13b | 2745 | .listxattr = ext4_listxattr, |
ac27a0ec DK |
2746 | .removexattr = generic_removexattr, |
2747 | #endif | |
4e34e719 | 2748 | .get_acl = ext4_get_acl, |
abc8746e | 2749 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
2750 | }; |
2751 | ||
754661f1 | 2752 | const struct inode_operations ext4_special_inode_operations = { |
617ba13b | 2753 | .setattr = ext4_setattr, |
03010a33 | 2754 | #ifdef CONFIG_EXT4_FS_XATTR |
ac27a0ec DK |
2755 | .setxattr = generic_setxattr, |
2756 | .getxattr = generic_getxattr, | |
617ba13b | 2757 | .listxattr = ext4_listxattr, |
ac27a0ec DK |
2758 | .removexattr = generic_removexattr, |
2759 | #endif | |
4e34e719 | 2760 | .get_acl = ext4_get_acl, |
ac27a0ec | 2761 | }; |