]>
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> | |
ac27a0ec | 29 | #include <linux/time.h> |
ac27a0ec DK |
30 | #include <linux/fcntl.h> |
31 | #include <linux/stat.h> | |
32 | #include <linux/string.h> | |
33 | #include <linux/quotaops.h> | |
34 | #include <linux/buffer_head.h> | |
35 | #include <linux/bio.h> | |
3dcf5451 CH |
36 | #include "ext4.h" |
37 | #include "ext4_jbd2.h" | |
ac27a0ec | 38 | |
ac27a0ec DK |
39 | #include "xattr.h" |
40 | #include "acl.h" | |
41 | ||
0562e0ba | 42 | #include <trace/events/ext4.h> |
ac27a0ec DK |
43 | /* |
44 | * define how far ahead to read directories while searching them. | |
45 | */ | |
46 | #define NAMEI_RA_CHUNKS 2 | |
47 | #define NAMEI_RA_BLOCKS 4 | |
8c55e204 | 48 | #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) |
ac27a0ec | 49 | |
617ba13b | 50 | static struct buffer_head *ext4_append(handle_t *handle, |
ac27a0ec | 51 | struct inode *inode, |
0f70b406 | 52 | ext4_lblk_t *block) |
ac27a0ec DK |
53 | { |
54 | struct buffer_head *bh; | |
1c215028 | 55 | int err; |
ac27a0ec | 56 | |
df981d03 TT |
57 | if (unlikely(EXT4_SB(inode->i_sb)->s_max_dir_size_kb && |
58 | ((inode->i_size >> 10) >= | |
0f70b406 TT |
59 | EXT4_SB(inode->i_sb)->s_max_dir_size_kb))) |
60 | return ERR_PTR(-ENOSPC); | |
df981d03 | 61 | |
ac27a0ec DK |
62 | *block = inode->i_size >> inode->i_sb->s_blocksize_bits; |
63 | ||
c5e298ae | 64 | bh = ext4_bread(handle, inode, *block, EXT4_GET_BLOCKS_CREATE); |
1c215028 TT |
65 | if (IS_ERR(bh)) |
66 | return bh; | |
0f70b406 TT |
67 | inode->i_size += inode->i_sb->s_blocksize; |
68 | EXT4_I(inode)->i_disksize = inode->i_size; | |
5d601255 | 69 | BUFFER_TRACE(bh, "get_write_access"); |
0f70b406 TT |
70 | err = ext4_journal_get_write_access(handle, bh); |
71 | if (err) { | |
72 | brelse(bh); | |
73 | ext4_std_error(inode->i_sb, err); | |
74 | return ERR_PTR(err); | |
6d1ab10e | 75 | } |
ac27a0ec DK |
76 | return bh; |
77 | } | |
78 | ||
dc6982ff TT |
79 | static int ext4_dx_csum_verify(struct inode *inode, |
80 | struct ext4_dir_entry *dirent); | |
81 | ||
82 | typedef enum { | |
83 | EITHER, INDEX, DIRENT | |
84 | } dirblock_type_t; | |
85 | ||
86 | #define ext4_read_dirblock(inode, block, type) \ | |
b03a2f7e | 87 | __ext4_read_dirblock((inode), (block), (type), __func__, __LINE__) |
dc6982ff TT |
88 | |
89 | static struct buffer_head *__ext4_read_dirblock(struct inode *inode, | |
b03a2f7e AD |
90 | ext4_lblk_t block, |
91 | dirblock_type_t type, | |
92 | const char *func, | |
93 | unsigned int line) | |
dc6982ff TT |
94 | { |
95 | struct buffer_head *bh; | |
96 | struct ext4_dir_entry *dirent; | |
1c215028 | 97 | int is_dx_block = 0; |
dc6982ff | 98 | |
1c215028 TT |
99 | bh = ext4_bread(NULL, inode, block, 0); |
100 | if (IS_ERR(bh)) { | |
b03a2f7e AD |
101 | __ext4_warning(inode->i_sb, func, line, |
102 | "inode #%lu: lblock %lu: comm %s: " | |
103 | "error %ld reading directory block", | |
104 | inode->i_ino, (unsigned long)block, | |
105 | current->comm, PTR_ERR(bh)); | |
1c215028 TT |
106 | |
107 | return bh; | |
108 | } | |
109 | if (!bh) { | |
b03a2f7e AD |
110 | ext4_error_inode(inode, func, line, block, |
111 | "Directory hole found"); | |
6a797d27 | 112 | return ERR_PTR(-EFSCORRUPTED); |
dc6982ff TT |
113 | } |
114 | dirent = (struct ext4_dir_entry *) bh->b_data; | |
115 | /* Determine whether or not we have an index block */ | |
116 | if (is_dx(inode)) { | |
117 | if (block == 0) | |
118 | is_dx_block = 1; | |
119 | else if (ext4_rec_len_from_disk(dirent->rec_len, | |
120 | inode->i_sb->s_blocksize) == | |
121 | inode->i_sb->s_blocksize) | |
122 | is_dx_block = 1; | |
123 | } | |
124 | if (!is_dx_block && type == INDEX) { | |
b03a2f7e | 125 | ext4_error_inode(inode, func, line, block, |
dc6982ff | 126 | "directory leaf block found instead of index block"); |
6a797d27 | 127 | return ERR_PTR(-EFSCORRUPTED); |
dc6982ff | 128 | } |
9aa5d32b | 129 | if (!ext4_has_metadata_csum(inode->i_sb) || |
dc6982ff TT |
130 | buffer_verified(bh)) |
131 | return bh; | |
132 | ||
133 | /* | |
134 | * An empty leaf block can get mistaken for a index block; for | |
135 | * this reason, we can only check the index checksum when the | |
136 | * caller is sure it should be an index block. | |
137 | */ | |
138 | if (is_dx_block && type == INDEX) { | |
139 | if (ext4_dx_csum_verify(inode, dirent)) | |
140 | set_buffer_verified(bh); | |
141 | else { | |
b03a2f7e AD |
142 | ext4_error_inode(inode, func, line, block, |
143 | "Directory index failed checksum"); | |
a871611b | 144 | brelse(bh); |
6a797d27 | 145 | return ERR_PTR(-EFSBADCRC); |
a871611b | 146 | } |
ac27a0ec | 147 | } |
dc6982ff TT |
148 | if (!is_dx_block) { |
149 | if (ext4_dirent_csum_verify(inode, dirent)) | |
150 | set_buffer_verified(bh); | |
151 | else { | |
b03a2f7e AD |
152 | ext4_error_inode(inode, func, line, block, |
153 | "Directory block failed checksum"); | |
dc6982ff | 154 | brelse(bh); |
6a797d27 | 155 | return ERR_PTR(-EFSBADCRC); |
dc6982ff | 156 | } |
6d1ab10e | 157 | } |
ac27a0ec DK |
158 | return bh; |
159 | } | |
160 | ||
161 | #ifndef assert | |
162 | #define assert(test) J_ASSERT(test) | |
163 | #endif | |
164 | ||
ac27a0ec DK |
165 | #ifdef DX_DEBUG |
166 | #define dxtrace(command) command | |
167 | #else | |
168 | #define dxtrace(command) | |
169 | #endif | |
170 | ||
171 | struct fake_dirent | |
172 | { | |
173 | __le32 inode; | |
174 | __le16 rec_len; | |
175 | u8 name_len; | |
176 | u8 file_type; | |
177 | }; | |
178 | ||
179 | struct dx_countlimit | |
180 | { | |
181 | __le16 limit; | |
182 | __le16 count; | |
183 | }; | |
184 | ||
185 | struct dx_entry | |
186 | { | |
187 | __le32 hash; | |
188 | __le32 block; | |
189 | }; | |
190 | ||
191 | /* | |
192 | * dx_root_info is laid out so that if it should somehow get overlaid by a | |
193 | * dirent the two low bits of the hash version will be zero. Therefore, the | |
194 | * hash version mod 4 should never be 0. Sincerely, the paranoia department. | |
195 | */ | |
196 | ||
197 | struct dx_root | |
198 | { | |
199 | struct fake_dirent dot; | |
200 | char dot_name[4]; | |
201 | struct fake_dirent dotdot; | |
202 | char dotdot_name[4]; | |
203 | struct dx_root_info | |
204 | { | |
205 | __le32 reserved_zero; | |
206 | u8 hash_version; | |
207 | u8 info_length; /* 8 */ | |
208 | u8 indirect_levels; | |
209 | u8 unused_flags; | |
210 | } | |
211 | info; | |
212 | struct dx_entry entries[0]; | |
213 | }; | |
214 | ||
215 | struct dx_node | |
216 | { | |
217 | struct fake_dirent fake; | |
218 | struct dx_entry entries[0]; | |
219 | }; | |
220 | ||
221 | ||
222 | struct dx_frame | |
223 | { | |
224 | struct buffer_head *bh; | |
225 | struct dx_entry *entries; | |
226 | struct dx_entry *at; | |
227 | }; | |
228 | ||
229 | struct dx_map_entry | |
230 | { | |
231 | u32 hash; | |
ef2b02d3 ES |
232 | u16 offs; |
233 | u16 size; | |
ac27a0ec DK |
234 | }; |
235 | ||
e6153918 DW |
236 | /* |
237 | * This goes at the end of each htree block. | |
238 | */ | |
239 | struct dx_tail { | |
240 | u32 dt_reserved; | |
241 | __le32 dt_checksum; /* crc32c(uuid+inum+dirblock) */ | |
242 | }; | |
243 | ||
725d26d3 AK |
244 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry); |
245 | static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value); | |
af5bc92d TT |
246 | static inline unsigned dx_get_hash(struct dx_entry *entry); |
247 | static void dx_set_hash(struct dx_entry *entry, unsigned value); | |
248 | static unsigned dx_get_count(struct dx_entry *entries); | |
249 | static unsigned dx_get_limit(struct dx_entry *entries); | |
250 | static void dx_set_count(struct dx_entry *entries, unsigned value); | |
251 | static void dx_set_limit(struct dx_entry *entries, unsigned value); | |
252 | static unsigned dx_root_limit(struct inode *dir, unsigned infosize); | |
253 | static unsigned dx_node_limit(struct inode *dir); | |
5b643f9c | 254 | static struct dx_frame *dx_probe(struct ext4_filename *fname, |
ac27a0ec DK |
255 | struct inode *dir, |
256 | struct dx_hash_info *hinfo, | |
dd73b5d5 | 257 | struct dx_frame *frame); |
af5bc92d | 258 | static void dx_release(struct dx_frame *frames); |
1f3862b5 MH |
259 | static int dx_make_map(struct inode *dir, struct ext4_dir_entry_2 *de, |
260 | unsigned blocksize, struct dx_hash_info *hinfo, | |
261 | struct dx_map_entry map[]); | |
ac27a0ec | 262 | static void dx_sort_map(struct dx_map_entry *map, unsigned count); |
af5bc92d | 263 | static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to, |
3d0518f4 | 264 | struct dx_map_entry *offsets, int count, unsigned blocksize); |
8bad4597 | 265 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize); |
725d26d3 AK |
266 | static void dx_insert_block(struct dx_frame *frame, |
267 | u32 hash, ext4_lblk_t block); | |
617ba13b | 268 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
269 | struct dx_frame *frame, |
270 | struct dx_frame *frames, | |
271 | __u32 *start_hash); | |
f702ba0f | 272 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, |
5b643f9c | 273 | struct ext4_filename *fname, |
537d8f93 | 274 | struct ext4_dir_entry_2 **res_dir); |
5b643f9c | 275 | static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 276 | struct inode *dir, struct inode *inode); |
ac27a0ec | 277 | |
dbe89444 | 278 | /* checksumming functions */ |
3c47d541 TM |
279 | void initialize_dirent_tail(struct ext4_dir_entry_tail *t, |
280 | unsigned int blocksize) | |
b0336e8d DW |
281 | { |
282 | memset(t, 0, sizeof(struct ext4_dir_entry_tail)); | |
283 | t->det_rec_len = ext4_rec_len_to_disk( | |
284 | sizeof(struct ext4_dir_entry_tail), blocksize); | |
285 | t->det_reserved_ft = EXT4_FT_DIR_CSUM; | |
286 | } | |
287 | ||
288 | /* Walk through a dirent block to find a checksum "dirent" at the tail */ | |
289 | static struct ext4_dir_entry_tail *get_dirent_tail(struct inode *inode, | |
290 | struct ext4_dir_entry *de) | |
291 | { | |
292 | struct ext4_dir_entry_tail *t; | |
293 | ||
294 | #ifdef PARANOID | |
295 | struct ext4_dir_entry *d, *top; | |
296 | ||
297 | d = de; | |
298 | top = (struct ext4_dir_entry *)(((void *)de) + | |
299 | (EXT4_BLOCK_SIZE(inode->i_sb) - | |
300 | sizeof(struct ext4_dir_entry_tail))); | |
301 | while (d < top && d->rec_len) | |
302 | d = (struct ext4_dir_entry *)(((void *)d) + | |
303 | le16_to_cpu(d->rec_len)); | |
304 | ||
305 | if (d != top) | |
306 | return NULL; | |
307 | ||
308 | t = (struct ext4_dir_entry_tail *)d; | |
309 | #else | |
310 | t = EXT4_DIRENT_TAIL(de, EXT4_BLOCK_SIZE(inode->i_sb)); | |
311 | #endif | |
312 | ||
313 | if (t->det_reserved_zero1 || | |
314 | le16_to_cpu(t->det_rec_len) != sizeof(struct ext4_dir_entry_tail) || | |
315 | t->det_reserved_zero2 || | |
316 | t->det_reserved_ft != EXT4_FT_DIR_CSUM) | |
317 | return NULL; | |
318 | ||
319 | return t; | |
320 | } | |
321 | ||
322 | static __le32 ext4_dirent_csum(struct inode *inode, | |
323 | struct ext4_dir_entry *dirent, int size) | |
324 | { | |
325 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
326 | struct ext4_inode_info *ei = EXT4_I(inode); | |
327 | __u32 csum; | |
328 | ||
329 | csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size); | |
330 | return cpu_to_le32(csum); | |
331 | } | |
332 | ||
b03a2f7e AD |
333 | #define warn_no_space_for_csum(inode) \ |
334 | __warn_no_space_for_csum((inode), __func__, __LINE__) | |
335 | ||
336 | static void __warn_no_space_for_csum(struct inode *inode, const char *func, | |
337 | unsigned int line) | |
dffe9d8d | 338 | { |
b03a2f7e AD |
339 | __ext4_warning_inode(inode, func, line, |
340 | "No space for directory leaf checksum. Please run e2fsck -D."); | |
dffe9d8d TT |
341 | } |
342 | ||
b0336e8d DW |
343 | int ext4_dirent_csum_verify(struct inode *inode, struct ext4_dir_entry *dirent) |
344 | { | |
345 | struct ext4_dir_entry_tail *t; | |
346 | ||
9aa5d32b | 347 | if (!ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d DW |
348 | return 1; |
349 | ||
350 | t = get_dirent_tail(inode, dirent); | |
351 | if (!t) { | |
dffe9d8d | 352 | warn_no_space_for_csum(inode); |
b0336e8d DW |
353 | return 0; |
354 | } | |
355 | ||
356 | if (t->det_checksum != ext4_dirent_csum(inode, dirent, | |
357 | (void *)t - (void *)dirent)) | |
358 | return 0; | |
359 | ||
360 | return 1; | |
361 | } | |
362 | ||
363 | static void ext4_dirent_csum_set(struct inode *inode, | |
364 | struct ext4_dir_entry *dirent) | |
365 | { | |
366 | struct ext4_dir_entry_tail *t; | |
367 | ||
9aa5d32b | 368 | if (!ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d DW |
369 | return; |
370 | ||
371 | t = get_dirent_tail(inode, dirent); | |
372 | if (!t) { | |
dffe9d8d | 373 | warn_no_space_for_csum(inode); |
b0336e8d DW |
374 | return; |
375 | } | |
376 | ||
377 | t->det_checksum = ext4_dirent_csum(inode, dirent, | |
378 | (void *)t - (void *)dirent); | |
379 | } | |
380 | ||
3c47d541 TM |
381 | int ext4_handle_dirty_dirent_node(handle_t *handle, |
382 | struct inode *inode, | |
383 | struct buffer_head *bh) | |
b0336e8d DW |
384 | { |
385 | ext4_dirent_csum_set(inode, (struct ext4_dir_entry *)bh->b_data); | |
386 | return ext4_handle_dirty_metadata(handle, inode, bh); | |
387 | } | |
388 | ||
dbe89444 DW |
389 | static struct dx_countlimit *get_dx_countlimit(struct inode *inode, |
390 | struct ext4_dir_entry *dirent, | |
391 | int *offset) | |
392 | { | |
393 | struct ext4_dir_entry *dp; | |
394 | struct dx_root_info *root; | |
395 | int count_offset; | |
396 | ||
397 | if (le16_to_cpu(dirent->rec_len) == EXT4_BLOCK_SIZE(inode->i_sb)) | |
398 | count_offset = 8; | |
399 | else if (le16_to_cpu(dirent->rec_len) == 12) { | |
400 | dp = (struct ext4_dir_entry *)(((void *)dirent) + 12); | |
401 | if (le16_to_cpu(dp->rec_len) != | |
402 | EXT4_BLOCK_SIZE(inode->i_sb) - 12) | |
403 | return NULL; | |
404 | root = (struct dx_root_info *)(((void *)dp + 12)); | |
405 | if (root->reserved_zero || | |
406 | root->info_length != sizeof(struct dx_root_info)) | |
407 | return NULL; | |
408 | count_offset = 32; | |
409 | } else | |
410 | return NULL; | |
411 | ||
412 | if (offset) | |
413 | *offset = count_offset; | |
414 | return (struct dx_countlimit *)(((void *)dirent) + count_offset); | |
415 | } | |
416 | ||
417 | static __le32 ext4_dx_csum(struct inode *inode, struct ext4_dir_entry *dirent, | |
418 | int count_offset, int count, struct dx_tail *t) | |
419 | { | |
420 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
421 | struct ext4_inode_info *ei = EXT4_I(inode); | |
d6a77105 TT |
422 | __u32 csum; |
423 | __le32 save_csum; | |
dbe89444 DW |
424 | int size; |
425 | ||
426 | size = count_offset + (count * sizeof(struct dx_entry)); | |
d6a77105 | 427 | save_csum = t->dt_checksum; |
dbe89444 DW |
428 | t->dt_checksum = 0; |
429 | csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size); | |
430 | csum = ext4_chksum(sbi, csum, (__u8 *)t, sizeof(struct dx_tail)); | |
d6a77105 | 431 | t->dt_checksum = save_csum; |
dbe89444 DW |
432 | |
433 | return cpu_to_le32(csum); | |
434 | } | |
435 | ||
436 | static int ext4_dx_csum_verify(struct inode *inode, | |
437 | struct ext4_dir_entry *dirent) | |
438 | { | |
439 | struct dx_countlimit *c; | |
440 | struct dx_tail *t; | |
441 | int count_offset, limit, count; | |
442 | ||
9aa5d32b | 443 | if (!ext4_has_metadata_csum(inode->i_sb)) |
dbe89444 DW |
444 | return 1; |
445 | ||
446 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
447 | if (!c) { | |
448 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
449 | return 1; | |
450 | } | |
451 | limit = le16_to_cpu(c->limit); | |
452 | count = le16_to_cpu(c->count); | |
453 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
454 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
dffe9d8d | 455 | warn_no_space_for_csum(inode); |
dbe89444 DW |
456 | return 1; |
457 | } | |
458 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
459 | ||
460 | if (t->dt_checksum != ext4_dx_csum(inode, dirent, count_offset, | |
461 | count, t)) | |
462 | return 0; | |
463 | return 1; | |
464 | } | |
465 | ||
466 | static void ext4_dx_csum_set(struct inode *inode, struct ext4_dir_entry *dirent) | |
467 | { | |
468 | struct dx_countlimit *c; | |
469 | struct dx_tail *t; | |
470 | int count_offset, limit, count; | |
471 | ||
9aa5d32b | 472 | if (!ext4_has_metadata_csum(inode->i_sb)) |
dbe89444 DW |
473 | return; |
474 | ||
475 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
476 | if (!c) { | |
477 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
478 | return; | |
479 | } | |
480 | limit = le16_to_cpu(c->limit); | |
481 | count = le16_to_cpu(c->count); | |
482 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
483 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
dffe9d8d | 484 | warn_no_space_for_csum(inode); |
dbe89444 DW |
485 | return; |
486 | } | |
487 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
488 | ||
489 | t->dt_checksum = ext4_dx_csum(inode, dirent, count_offset, count, t); | |
490 | } | |
491 | ||
492 | static inline int ext4_handle_dirty_dx_node(handle_t *handle, | |
493 | struct inode *inode, | |
494 | struct buffer_head *bh) | |
495 | { | |
496 | ext4_dx_csum_set(inode, (struct ext4_dir_entry *)bh->b_data); | |
497 | return ext4_handle_dirty_metadata(handle, inode, bh); | |
498 | } | |
499 | ||
f795e140 LZ |
500 | /* |
501 | * p is at least 6 bytes before the end of page | |
502 | */ | |
503 | static inline struct ext4_dir_entry_2 * | |
3d0518f4 | 504 | ext4_next_entry(struct ext4_dir_entry_2 *p, unsigned long blocksize) |
f795e140 LZ |
505 | { |
506 | return (struct ext4_dir_entry_2 *)((char *)p + | |
3d0518f4 | 507 | ext4_rec_len_from_disk(p->rec_len, blocksize)); |
f795e140 LZ |
508 | } |
509 | ||
ac27a0ec DK |
510 | /* |
511 | * Future: use high four bits of block for coalesce-on-delete flags | |
512 | * Mask them off for now. | |
513 | */ | |
514 | ||
725d26d3 | 515 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry) |
ac27a0ec DK |
516 | { |
517 | return le32_to_cpu(entry->block) & 0x00ffffff; | |
518 | } | |
519 | ||
725d26d3 | 520 | static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value) |
ac27a0ec DK |
521 | { |
522 | entry->block = cpu_to_le32(value); | |
523 | } | |
524 | ||
af5bc92d | 525 | static inline unsigned dx_get_hash(struct dx_entry *entry) |
ac27a0ec DK |
526 | { |
527 | return le32_to_cpu(entry->hash); | |
528 | } | |
529 | ||
af5bc92d | 530 | static inline void dx_set_hash(struct dx_entry *entry, unsigned value) |
ac27a0ec DK |
531 | { |
532 | entry->hash = cpu_to_le32(value); | |
533 | } | |
534 | ||
af5bc92d | 535 | static inline unsigned dx_get_count(struct dx_entry *entries) |
ac27a0ec DK |
536 | { |
537 | return le16_to_cpu(((struct dx_countlimit *) entries)->count); | |
538 | } | |
539 | ||
af5bc92d | 540 | static inline unsigned dx_get_limit(struct dx_entry *entries) |
ac27a0ec DK |
541 | { |
542 | return le16_to_cpu(((struct dx_countlimit *) entries)->limit); | |
543 | } | |
544 | ||
af5bc92d | 545 | static inline void dx_set_count(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
546 | { |
547 | ((struct dx_countlimit *) entries)->count = cpu_to_le16(value); | |
548 | } | |
549 | ||
af5bc92d | 550 | static inline void dx_set_limit(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
551 | { |
552 | ((struct dx_countlimit *) entries)->limit = cpu_to_le16(value); | |
553 | } | |
554 | ||
af5bc92d | 555 | static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize) |
ac27a0ec | 556 | { |
617ba13b MC |
557 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) - |
558 | EXT4_DIR_REC_LEN(2) - infosize; | |
dbe89444 | 559 | |
9aa5d32b | 560 | if (ext4_has_metadata_csum(dir->i_sb)) |
dbe89444 | 561 | entry_space -= sizeof(struct dx_tail); |
d9c769b7 | 562 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
563 | } |
564 | ||
af5bc92d | 565 | static inline unsigned dx_node_limit(struct inode *dir) |
ac27a0ec | 566 | { |
617ba13b | 567 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0); |
dbe89444 | 568 | |
9aa5d32b | 569 | if (ext4_has_metadata_csum(dir->i_sb)) |
dbe89444 | 570 | entry_space -= sizeof(struct dx_tail); |
d9c769b7 | 571 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
572 | } |
573 | ||
574 | /* | |
575 | * Debug | |
576 | */ | |
577 | #ifdef DX_DEBUG | |
4776004f | 578 | static void dx_show_index(char * label, struct dx_entry *entries) |
ac27a0ec | 579 | { |
63f57933 | 580 | int i, n = dx_get_count (entries); |
4776004f | 581 | printk(KERN_DEBUG "%s index ", label); |
63f57933 | 582 | for (i = 0; i < n; i++) { |
4776004f | 583 | printk("%x->%lu ", i ? dx_get_hash(entries + i) : |
725d26d3 | 584 | 0, (unsigned long)dx_get_block(entries + i)); |
63f57933 AM |
585 | } |
586 | printk("\n"); | |
ac27a0ec DK |
587 | } |
588 | ||
589 | struct stats | |
590 | { | |
591 | unsigned names; | |
592 | unsigned space; | |
593 | unsigned bcount; | |
594 | }; | |
595 | ||
b3098486 MH |
596 | static struct stats dx_show_leaf(struct inode *dir, |
597 | struct dx_hash_info *hinfo, | |
598 | struct ext4_dir_entry_2 *de, | |
599 | int size, int show_names) | |
ac27a0ec DK |
600 | { |
601 | unsigned names = 0, space = 0; | |
602 | char *base = (char *) de; | |
603 | struct dx_hash_info h = *hinfo; | |
604 | ||
605 | printk("names: "); | |
606 | while ((char *) de < base + size) | |
607 | { | |
608 | if (de->inode) | |
609 | { | |
610 | if (show_names) | |
611 | { | |
b3098486 MH |
612 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
613 | int len; | |
614 | char *name; | |
615 | struct ext4_str fname_crypto_str | |
616 | = {.name = NULL, .len = 0}; | |
c936e1ec | 617 | int res = 0; |
b3098486 MH |
618 | |
619 | name = de->name; | |
620 | len = de->name_len; | |
c936e1ec TT |
621 | if (ext4_encrypted_inode(inode)) |
622 | res = ext4_get_encryption_info(dir); | |
b7236e21 TT |
623 | if (res) { |
624 | printk(KERN_WARNING "Error setting up" | |
625 | " fname crypto: %d\n", res); | |
b3098486 MH |
626 | } |
627 | if (ctx == NULL) { | |
628 | /* Directory is not encrypted */ | |
629 | ext4fs_dirhash(de->name, | |
630 | de->name_len, &h); | |
631 | printk("%*.s:(U)%x.%u ", len, | |
632 | name, h.hash, | |
633 | (unsigned) ((char *) de | |
634 | - base)); | |
635 | } else { | |
636 | /* Directory is encrypted */ | |
637 | res = ext4_fname_crypto_alloc_buffer( | |
638 | ctx, de->name_len, | |
639 | &fname_crypto_str); | |
640 | if (res < 0) { | |
641 | printk(KERN_WARNING "Error " | |
642 | "allocating crypto " | |
643 | "buffer--skipping " | |
644 | "crypto\n"); | |
b3098486 MH |
645 | ctx = NULL; |
646 | } | |
5de0b4d0 | 647 | res = ext4_fname_disk_to_usr(ctx, NULL, de, |
b3098486 MH |
648 | &fname_crypto_str); |
649 | if (res < 0) { | |
650 | printk(KERN_WARNING "Error " | |
651 | "converting filename " | |
652 | "from disk to usr" | |
653 | "\n"); | |
654 | name = "??"; | |
655 | len = 2; | |
656 | } else { | |
657 | name = fname_crypto_str.name; | |
658 | len = fname_crypto_str.len; | |
659 | } | |
5de0b4d0 TT |
660 | ext4fs_dirhash(de->name, de->name_len, |
661 | &h); | |
b3098486 MH |
662 | printk("%*.s:(E)%x.%u ", len, name, |
663 | h.hash, (unsigned) ((char *) de | |
664 | - base)); | |
b3098486 MH |
665 | ext4_fname_crypto_free_buffer( |
666 | &fname_crypto_str); | |
667 | } | |
668 | #else | |
ac27a0ec DK |
669 | int len = de->name_len; |
670 | char *name = de->name; | |
617ba13b | 671 | ext4fs_dirhash(de->name, de->name_len, &h); |
b3098486 | 672 | printk("%*.s:%x.%u ", len, name, h.hash, |
265c6a0f | 673 | (unsigned) ((char *) de - base)); |
b3098486 | 674 | #endif |
ac27a0ec | 675 | } |
617ba13b | 676 | space += EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
677 | names++; |
678 | } | |
3d0518f4 | 679 | de = ext4_next_entry(de, size); |
ac27a0ec DK |
680 | } |
681 | printk("(%i)\n", names); | |
682 | return (struct stats) { names, space, 1 }; | |
683 | } | |
684 | ||
685 | struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, | |
686 | struct dx_entry *entries, int levels) | |
687 | { | |
688 | unsigned blocksize = dir->i_sb->s_blocksize; | |
af5bc92d | 689 | unsigned count = dx_get_count(entries), names = 0, space = 0, i; |
ac27a0ec DK |
690 | unsigned bcount = 0; |
691 | struct buffer_head *bh; | |
ac27a0ec DK |
692 | printk("%i indexed blocks...\n", count); |
693 | for (i = 0; i < count; i++, entries++) | |
694 | { | |
725d26d3 AK |
695 | ext4_lblk_t block = dx_get_block(entries); |
696 | ext4_lblk_t hash = i ? dx_get_hash(entries): 0; | |
ac27a0ec DK |
697 | u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; |
698 | struct stats stats; | |
699 | printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); | |
1c215028 TT |
700 | bh = ext4_bread(NULL,dir, block, 0); |
701 | if (!bh || IS_ERR(bh)) | |
702 | continue; | |
ac27a0ec DK |
703 | stats = levels? |
704 | dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): | |
b3098486 MH |
705 | dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) |
706 | bh->b_data, blocksize, 0); | |
ac27a0ec DK |
707 | names += stats.names; |
708 | space += stats.space; | |
709 | bcount += stats.bcount; | |
af5bc92d | 710 | brelse(bh); |
ac27a0ec DK |
711 | } |
712 | if (bcount) | |
60e6679e | 713 | printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n", |
4776004f TT |
714 | levels ? "" : " ", names, space/bcount, |
715 | (space/bcount)*100/blocksize); | |
ac27a0ec DK |
716 | return (struct stats) { names, space, bcount}; |
717 | } | |
718 | #endif /* DX_DEBUG */ | |
719 | ||
720 | /* | |
721 | * Probe for a directory leaf block to search. | |
722 | * | |
723 | * dx_probe can return ERR_BAD_DX_DIR, which means there was a format | |
724 | * error in the directory index, and the caller should fall back to | |
725 | * searching the directory normally. The callers of dx_probe **MUST** | |
726 | * check for this error code, and make sure it never gets reflected | |
727 | * back to userspace. | |
728 | */ | |
729 | static struct dx_frame * | |
5b643f9c | 730 | dx_probe(struct ext4_filename *fname, struct inode *dir, |
dd73b5d5 | 731 | struct dx_hash_info *hinfo, struct dx_frame *frame_in) |
ac27a0ec DK |
732 | { |
733 | unsigned count, indirect; | |
734 | struct dx_entry *at, *entries, *p, *q, *m; | |
735 | struct dx_root *root; | |
ac27a0ec | 736 | struct dx_frame *frame = frame_in; |
dd73b5d5 | 737 | struct dx_frame *ret_err = ERR_PTR(ERR_BAD_DX_DIR); |
ac27a0ec DK |
738 | u32 hash; |
739 | ||
dd73b5d5 TT |
740 | frame->bh = ext4_read_dirblock(dir, 0, INDEX); |
741 | if (IS_ERR(frame->bh)) | |
742 | return (struct dx_frame *) frame->bh; | |
743 | ||
744 | root = (struct dx_root *) frame->bh->b_data; | |
ac27a0ec DK |
745 | if (root->info.hash_version != DX_HASH_TEA && |
746 | root->info.hash_version != DX_HASH_HALF_MD4 && | |
747 | root->info.hash_version != DX_HASH_LEGACY) { | |
b03a2f7e AD |
748 | ext4_warning_inode(dir, "Unrecognised inode hash code %u", |
749 | root->info.hash_version); | |
ac27a0ec DK |
750 | goto fail; |
751 | } | |
5b643f9c TT |
752 | if (fname) |
753 | hinfo = &fname->hinfo; | |
ac27a0ec | 754 | hinfo->hash_version = root->info.hash_version; |
f99b2589 TT |
755 | if (hinfo->hash_version <= DX_HASH_TEA) |
756 | hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 757 | hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
5b643f9c TT |
758 | if (fname && fname_name(fname)) |
759 | ext4fs_dirhash(fname_name(fname), fname_len(fname), hinfo); | |
ac27a0ec DK |
760 | hash = hinfo->hash; |
761 | ||
762 | if (root->info.unused_flags & 1) { | |
b03a2f7e AD |
763 | ext4_warning_inode(dir, "Unimplemented hash flags: %#06x", |
764 | root->info.unused_flags); | |
ac27a0ec DK |
765 | goto fail; |
766 | } | |
767 | ||
b03a2f7e AD |
768 | indirect = root->info.indirect_levels; |
769 | if (indirect > 1) { | |
770 | ext4_warning_inode(dir, "Unimplemented hash depth: %#06x", | |
771 | root->info.indirect_levels); | |
ac27a0ec DK |
772 | goto fail; |
773 | } | |
774 | ||
b03a2f7e AD |
775 | entries = (struct dx_entry *)(((char *)&root->info) + |
776 | root->info.info_length); | |
3d82abae ES |
777 | |
778 | if (dx_get_limit(entries) != dx_root_limit(dir, | |
779 | root->info.info_length)) { | |
b03a2f7e AD |
780 | ext4_warning_inode(dir, "dx entry: limit %u != root limit %u", |
781 | dx_get_limit(entries), | |
782 | dx_root_limit(dir, root->info.info_length)); | |
3d82abae ES |
783 | goto fail; |
784 | } | |
785 | ||
af5bc92d | 786 | dxtrace(printk("Look up %x", hash)); |
dd73b5d5 | 787 | while (1) { |
ac27a0ec | 788 | count = dx_get_count(entries); |
3d82abae | 789 | if (!count || count > dx_get_limit(entries)) { |
b03a2f7e AD |
790 | ext4_warning_inode(dir, |
791 | "dx entry: count %u beyond limit %u", | |
792 | count, dx_get_limit(entries)); | |
dd73b5d5 | 793 | goto fail; |
3d82abae ES |
794 | } |
795 | ||
ac27a0ec DK |
796 | p = entries + 1; |
797 | q = entries + count - 1; | |
dd73b5d5 | 798 | while (p <= q) { |
b03a2f7e | 799 | m = p + (q - p) / 2; |
ac27a0ec DK |
800 | dxtrace(printk(".")); |
801 | if (dx_get_hash(m) > hash) | |
802 | q = m - 1; | |
803 | else | |
804 | p = m + 1; | |
805 | } | |
806 | ||
dd73b5d5 | 807 | if (0) { // linear search cross check |
ac27a0ec DK |
808 | unsigned n = count - 1; |
809 | at = entries; | |
810 | while (n--) | |
811 | { | |
812 | dxtrace(printk(",")); | |
813 | if (dx_get_hash(++at) > hash) | |
814 | { | |
815 | at--; | |
816 | break; | |
817 | } | |
818 | } | |
819 | assert (at == p - 1); | |
820 | } | |
821 | ||
822 | at = p - 1; | |
b03a2f7e AD |
823 | dxtrace(printk(" %x->%u\n", at == entries ? 0 : dx_get_hash(at), |
824 | dx_get_block(at))); | |
ac27a0ec DK |
825 | frame->entries = entries; |
826 | frame->at = at; | |
dd73b5d5 TT |
827 | if (!indirect--) |
828 | return frame; | |
829 | frame++; | |
830 | frame->bh = ext4_read_dirblock(dir, dx_get_block(at), INDEX); | |
831 | if (IS_ERR(frame->bh)) { | |
832 | ret_err = (struct dx_frame *) frame->bh; | |
833 | frame->bh = NULL; | |
834 | goto fail; | |
dbe89444 | 835 | } |
dd73b5d5 | 836 | entries = ((struct dx_node *) frame->bh->b_data)->entries; |
dbe89444 | 837 | |
b03a2f7e AD |
838 | if (dx_get_limit(entries) != dx_node_limit(dir)) { |
839 | ext4_warning_inode(dir, | |
840 | "dx entry: limit %u != node limit %u", | |
841 | dx_get_limit(entries), dx_node_limit(dir)); | |
dd73b5d5 | 842 | goto fail; |
3d82abae | 843 | } |
ac27a0ec | 844 | } |
dd73b5d5 | 845 | fail: |
ac27a0ec DK |
846 | while (frame >= frame_in) { |
847 | brelse(frame->bh); | |
848 | frame--; | |
849 | } | |
b3098486 | 850 | |
dd73b5d5 | 851 | if (ret_err == ERR_PTR(ERR_BAD_DX_DIR)) |
b03a2f7e AD |
852 | ext4_warning_inode(dir, |
853 | "Corrupt directory, running e2fsck is recommended"); | |
dd73b5d5 | 854 | return ret_err; |
ac27a0ec DK |
855 | } |
856 | ||
b03a2f7e | 857 | static void dx_release(struct dx_frame *frames) |
ac27a0ec DK |
858 | { |
859 | if (frames[0].bh == NULL) | |
860 | return; | |
861 | ||
b03a2f7e | 862 | if (((struct dx_root *)frames[0].bh->b_data)->info.indirect_levels) |
ac27a0ec DK |
863 | brelse(frames[1].bh); |
864 | brelse(frames[0].bh); | |
865 | } | |
866 | ||
867 | /* | |
868 | * This function increments the frame pointer to search the next leaf | |
869 | * block, and reads in the necessary intervening nodes if the search | |
870 | * should be necessary. Whether or not the search is necessary is | |
871 | * controlled by the hash parameter. If the hash value is even, then | |
872 | * the search is only continued if the next block starts with that | |
873 | * hash value. This is used if we are searching for a specific file. | |
874 | * | |
875 | * If the hash value is HASH_NB_ALWAYS, then always go to the next block. | |
876 | * | |
877 | * This function returns 1 if the caller should continue to search, | |
878 | * or 0 if it should not. If there is an error reading one of the | |
879 | * index blocks, it will a negative error code. | |
880 | * | |
881 | * If start_hash is non-null, it will be filled in with the starting | |
882 | * hash of the next page. | |
883 | */ | |
617ba13b | 884 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
885 | struct dx_frame *frame, |
886 | struct dx_frame *frames, | |
887 | __u32 *start_hash) | |
888 | { | |
889 | struct dx_frame *p; | |
890 | struct buffer_head *bh; | |
dc6982ff | 891 | int num_frames = 0; |
ac27a0ec DK |
892 | __u32 bhash; |
893 | ||
894 | p = frame; | |
895 | /* | |
896 | * Find the next leaf page by incrementing the frame pointer. | |
897 | * If we run out of entries in the interior node, loop around and | |
898 | * increment pointer in the parent node. When we break out of | |
899 | * this loop, num_frames indicates the number of interior | |
900 | * nodes need to be read. | |
901 | */ | |
902 | while (1) { | |
903 | if (++(p->at) < p->entries + dx_get_count(p->entries)) | |
904 | break; | |
905 | if (p == frames) | |
906 | return 0; | |
907 | num_frames++; | |
908 | p--; | |
909 | } | |
910 | ||
911 | /* | |
912 | * If the hash is 1, then continue only if the next page has a | |
913 | * continuation hash of any value. This is used for readdir | |
914 | * handling. Otherwise, check to see if the hash matches the | |
915 | * desired contiuation hash. If it doesn't, return since | |
916 | * there's no point to read in the successive index pages. | |
917 | */ | |
918 | bhash = dx_get_hash(p->at); | |
919 | if (start_hash) | |
920 | *start_hash = bhash; | |
921 | if ((hash & 1) == 0) { | |
922 | if ((bhash & ~1) != hash) | |
923 | return 0; | |
924 | } | |
925 | /* | |
926 | * If the hash is HASH_NB_ALWAYS, we always go to the next | |
927 | * block so no check is necessary | |
928 | */ | |
929 | while (num_frames--) { | |
dc6982ff TT |
930 | bh = ext4_read_dirblock(dir, dx_get_block(p->at), INDEX); |
931 | if (IS_ERR(bh)) | |
932 | return PTR_ERR(bh); | |
ac27a0ec | 933 | p++; |
af5bc92d | 934 | brelse(p->bh); |
ac27a0ec DK |
935 | p->bh = bh; |
936 | p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; | |
937 | } | |
938 | return 1; | |
939 | } | |
940 | ||
941 | ||
ac27a0ec DK |
942 | /* |
943 | * This function fills a red-black tree with information from a | |
944 | * directory block. It returns the number directory entries loaded | |
945 | * into the tree. If there is an error it is returned in err. | |
946 | */ | |
947 | static int htree_dirblock_to_tree(struct file *dir_file, | |
725d26d3 | 948 | struct inode *dir, ext4_lblk_t block, |
ac27a0ec DK |
949 | struct dx_hash_info *hinfo, |
950 | __u32 start_hash, __u32 start_minor_hash) | |
951 | { | |
952 | struct buffer_head *bh; | |
617ba13b | 953 | struct ext4_dir_entry_2 *de, *top; |
90b0a973 | 954 | int err = 0, count = 0; |
1f3862b5 | 955 | struct ext4_str fname_crypto_str = {.name = NULL, .len = 0}, tmp_str; |
ac27a0ec | 956 | |
725d26d3 AK |
957 | dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n", |
958 | (unsigned long)block)); | |
dc6982ff TT |
959 | bh = ext4_read_dirblock(dir, block, DIRENT); |
960 | if (IS_ERR(bh)) | |
961 | return PTR_ERR(bh); | |
b0336e8d | 962 | |
617ba13b MC |
963 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
964 | top = (struct ext4_dir_entry_2 *) ((char *) de + | |
ac27a0ec | 965 | dir->i_sb->s_blocksize - |
617ba13b | 966 | EXT4_DIR_REC_LEN(0)); |
1f3862b5 MH |
967 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
968 | /* Check if the directory is encrypted */ | |
b7236e21 | 969 | if (ext4_encrypted_inode(dir)) { |
c936e1ec TT |
970 | err = ext4_get_encryption_info(dir); |
971 | if (err < 0) { | |
972 | brelse(bh); | |
973 | return err; | |
974 | } | |
b7236e21 | 975 | err = ext4_fname_crypto_alloc_buffer(dir, EXT4_NAME_LEN, |
1f3862b5 MH |
976 | &fname_crypto_str); |
977 | if (err < 0) { | |
1f3862b5 MH |
978 | brelse(bh); |
979 | return err; | |
980 | } | |
981 | } | |
982 | #endif | |
3d0518f4 | 983 | for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) { |
f7c21177 | 984 | if (ext4_check_dir_entry(dir, NULL, de, bh, |
226ba972 | 985 | bh->b_data, bh->b_size, |
cad3f007 TT |
986 | (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb)) |
987 | + ((char *)de - bh->b_data))) { | |
64cb9273 AV |
988 | /* silently ignore the rest of the block */ |
989 | break; | |
e6c40211 | 990 | } |
617ba13b | 991 | ext4fs_dirhash(de->name, de->name_len, hinfo); |
ac27a0ec DK |
992 | if ((hinfo->hash < start_hash) || |
993 | ((hinfo->hash == start_hash) && | |
994 | (hinfo->minor_hash < start_minor_hash))) | |
995 | continue; | |
996 | if (de->inode == 0) | |
997 | continue; | |
b7236e21 | 998 | if (!ext4_encrypted_inode(dir)) { |
1f3862b5 MH |
999 | tmp_str.name = de->name; |
1000 | tmp_str.len = de->name_len; | |
1001 | err = ext4_htree_store_dirent(dir_file, | |
1002 | hinfo->hash, hinfo->minor_hash, de, | |
1003 | &tmp_str); | |
1004 | } else { | |
d2299590 TT |
1005 | int save_len = fname_crypto_str.len; |
1006 | ||
1f3862b5 | 1007 | /* Directory is encrypted */ |
b7236e21 | 1008 | err = ext4_fname_disk_to_usr(dir, hinfo, de, |
1f3862b5 MH |
1009 | &fname_crypto_str); |
1010 | if (err < 0) { | |
1011 | count = err; | |
1012 | goto errout; | |
1013 | } | |
1014 | err = ext4_htree_store_dirent(dir_file, | |
1015 | hinfo->hash, hinfo->minor_hash, de, | |
1016 | &fname_crypto_str); | |
d2299590 | 1017 | fname_crypto_str.len = save_len; |
1f3862b5 | 1018 | } |
2f61830a | 1019 | if (err != 0) { |
1f3862b5 MH |
1020 | count = err; |
1021 | goto errout; | |
ac27a0ec DK |
1022 | } |
1023 | count++; | |
1024 | } | |
1f3862b5 | 1025 | errout: |
ac27a0ec | 1026 | brelse(bh); |
1f3862b5 | 1027 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
1f3862b5 MH |
1028 | ext4_fname_crypto_free_buffer(&fname_crypto_str); |
1029 | #endif | |
ac27a0ec DK |
1030 | return count; |
1031 | } | |
1032 | ||
1033 | ||
1034 | /* | |
1035 | * This function fills a red-black tree with information from a | |
1036 | * directory. We start scanning the directory in hash order, starting | |
1037 | * at start_hash and start_minor_hash. | |
1038 | * | |
1039 | * This function returns the number of entries inserted into the tree, | |
1040 | * or a negative error code. | |
1041 | */ | |
617ba13b | 1042 | int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, |
ac27a0ec DK |
1043 | __u32 start_minor_hash, __u32 *next_hash) |
1044 | { | |
1045 | struct dx_hash_info hinfo; | |
617ba13b | 1046 | struct ext4_dir_entry_2 *de; |
ac27a0ec DK |
1047 | struct dx_frame frames[2], *frame; |
1048 | struct inode *dir; | |
725d26d3 | 1049 | ext4_lblk_t block; |
ac27a0ec | 1050 | int count = 0; |
725d26d3 | 1051 | int ret, err; |
ac27a0ec | 1052 | __u32 hashval; |
2f61830a | 1053 | struct ext4_str tmp_str; |
ac27a0ec | 1054 | |
60e6679e | 1055 | dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n", |
4776004f | 1056 | start_hash, start_minor_hash)); |
496ad9aa | 1057 | dir = file_inode(dir_file); |
12e9b892 | 1058 | if (!(ext4_test_inode_flag(dir, EXT4_INODE_INDEX))) { |
617ba13b | 1059 | hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
f99b2589 TT |
1060 | if (hinfo.hash_version <= DX_HASH_TEA) |
1061 | hinfo.hash_version += | |
1062 | EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 1063 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
8af0f082 TM |
1064 | if (ext4_has_inline_data(dir)) { |
1065 | int has_inline_data = 1; | |
1066 | count = htree_inlinedir_to_tree(dir_file, dir, 0, | |
1067 | &hinfo, start_hash, | |
1068 | start_minor_hash, | |
1069 | &has_inline_data); | |
1070 | if (has_inline_data) { | |
1071 | *next_hash = ~0; | |
1072 | return count; | |
1073 | } | |
1074 | } | |
ac27a0ec DK |
1075 | count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo, |
1076 | start_hash, start_minor_hash); | |
1077 | *next_hash = ~0; | |
1078 | return count; | |
1079 | } | |
1080 | hinfo.hash = start_hash; | |
1081 | hinfo.minor_hash = 0; | |
dd73b5d5 TT |
1082 | frame = dx_probe(NULL, dir, &hinfo, frames); |
1083 | if (IS_ERR(frame)) | |
1084 | return PTR_ERR(frame); | |
ac27a0ec DK |
1085 | |
1086 | /* Add '.' and '..' from the htree header */ | |
1087 | if (!start_hash && !start_minor_hash) { | |
617ba13b | 1088 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
2f61830a TT |
1089 | tmp_str.name = de->name; |
1090 | tmp_str.len = de->name_len; | |
1091 | err = ext4_htree_store_dirent(dir_file, 0, 0, | |
1092 | de, &tmp_str); | |
1093 | if (err != 0) | |
ac27a0ec DK |
1094 | goto errout; |
1095 | count++; | |
1096 | } | |
1097 | if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) { | |
617ba13b | 1098 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
3d0518f4 | 1099 | de = ext4_next_entry(de, dir->i_sb->s_blocksize); |
2f61830a TT |
1100 | tmp_str.name = de->name; |
1101 | tmp_str.len = de->name_len; | |
1102 | err = ext4_htree_store_dirent(dir_file, 2, 0, | |
1103 | de, &tmp_str); | |
1104 | if (err != 0) | |
ac27a0ec DK |
1105 | goto errout; |
1106 | count++; | |
1107 | } | |
1108 | ||
1109 | while (1) { | |
1110 | block = dx_get_block(frame->at); | |
1111 | ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo, | |
1112 | start_hash, start_minor_hash); | |
1113 | if (ret < 0) { | |
1114 | err = ret; | |
1115 | goto errout; | |
1116 | } | |
1117 | count += ret; | |
1118 | hashval = ~0; | |
617ba13b | 1119 | ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS, |
ac27a0ec DK |
1120 | frame, frames, &hashval); |
1121 | *next_hash = hashval; | |
1122 | if (ret < 0) { | |
1123 | err = ret; | |
1124 | goto errout; | |
1125 | } | |
1126 | /* | |
1127 | * Stop if: (a) there are no more entries, or | |
1128 | * (b) we have inserted at least one entry and the | |
1129 | * next hash value is not a continuation | |
1130 | */ | |
1131 | if ((ret == 0) || | |
1132 | (count && ((hashval & 1) == 0))) | |
1133 | break; | |
1134 | } | |
1135 | dx_release(frames); | |
4776004f TT |
1136 | dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, " |
1137 | "next hash: %x\n", count, *next_hash)); | |
ac27a0ec DK |
1138 | return count; |
1139 | errout: | |
1140 | dx_release(frames); | |
1141 | return (err); | |
1142 | } | |
1143 | ||
7335cd3b TM |
1144 | static inline int search_dirblock(struct buffer_head *bh, |
1145 | struct inode *dir, | |
5b643f9c | 1146 | struct ext4_filename *fname, |
7335cd3b TM |
1147 | const struct qstr *d_name, |
1148 | unsigned int offset, | |
1149 | struct ext4_dir_entry_2 **res_dir) | |
1150 | { | |
5b643f9c TT |
1151 | return ext4_search_dir(bh, bh->b_data, dir->i_sb->s_blocksize, dir, |
1152 | fname, d_name, offset, res_dir); | |
7335cd3b TM |
1153 | } |
1154 | ||
ac27a0ec DK |
1155 | /* |
1156 | * Directory block splitting, compacting | |
1157 | */ | |
1158 | ||
ef2b02d3 ES |
1159 | /* |
1160 | * Create map of hash values, offsets, and sizes, stored at end of block. | |
1161 | * Returns number of entries mapped. | |
1162 | */ | |
1f3862b5 MH |
1163 | static int dx_make_map(struct inode *dir, struct ext4_dir_entry_2 *de, |
1164 | unsigned blocksize, struct dx_hash_info *hinfo, | |
8bad4597 | 1165 | struct dx_map_entry *map_tail) |
ac27a0ec DK |
1166 | { |
1167 | int count = 0; | |
1168 | char *base = (char *) de; | |
1169 | struct dx_hash_info h = *hinfo; | |
1170 | ||
8bad4597 | 1171 | while ((char *) de < base + blocksize) { |
ac27a0ec | 1172 | if (de->name_len && de->inode) { |
617ba13b | 1173 | ext4fs_dirhash(de->name, de->name_len, &h); |
ac27a0ec DK |
1174 | map_tail--; |
1175 | map_tail->hash = h.hash; | |
9aee2286 | 1176 | map_tail->offs = ((char *) de - base)>>2; |
ef2b02d3 | 1177 | map_tail->size = le16_to_cpu(de->rec_len); |
ac27a0ec DK |
1178 | count++; |
1179 | cond_resched(); | |
1180 | } | |
1181 | /* XXX: do we need to check rec_len == 0 case? -Chris */ | |
3d0518f4 | 1182 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
1183 | } |
1184 | return count; | |
1185 | } | |
1186 | ||
ef2b02d3 | 1187 | /* Sort map by hash value */ |
ac27a0ec DK |
1188 | static void dx_sort_map (struct dx_map_entry *map, unsigned count) |
1189 | { | |
63f57933 AM |
1190 | struct dx_map_entry *p, *q, *top = map + count - 1; |
1191 | int more; | |
1192 | /* Combsort until bubble sort doesn't suck */ | |
1193 | while (count > 2) { | |
1194 | count = count*10/13; | |
1195 | if (count - 9 < 2) /* 9, 10 -> 11 */ | |
1196 | count = 11; | |
1197 | for (p = top, q = p - count; q >= map; p--, q--) | |
1198 | if (p->hash < q->hash) | |
1199 | swap(*p, *q); | |
1200 | } | |
1201 | /* Garden variety bubble sort */ | |
1202 | do { | |
1203 | more = 0; | |
1204 | q = top; | |
1205 | while (q-- > map) { | |
1206 | if (q[1].hash >= q[0].hash) | |
ac27a0ec | 1207 | continue; |
63f57933 AM |
1208 | swap(*(q+1), *q); |
1209 | more = 1; | |
ac27a0ec DK |
1210 | } |
1211 | } while(more); | |
1212 | } | |
1213 | ||
725d26d3 | 1214 | static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block) |
ac27a0ec DK |
1215 | { |
1216 | struct dx_entry *entries = frame->entries; | |
1217 | struct dx_entry *old = frame->at, *new = old + 1; | |
1218 | int count = dx_get_count(entries); | |
1219 | ||
1220 | assert(count < dx_get_limit(entries)); | |
1221 | assert(old < entries + count); | |
1222 | memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); | |
1223 | dx_set_hash(new, hash); | |
1224 | dx_set_block(new, block); | |
1225 | dx_set_count(entries, count + 1); | |
1226 | } | |
ac27a0ec | 1227 | |
ac27a0ec | 1228 | /* |
617ba13b | 1229 | * NOTE! unlike strncmp, ext4_match returns 1 for success, 0 for failure. |
ac27a0ec | 1230 | * |
617ba13b | 1231 | * `len <= EXT4_NAME_LEN' is guaranteed by caller. |
ac27a0ec DK |
1232 | * `de != NULL' is guaranteed by caller. |
1233 | */ | |
5b643f9c | 1234 | static inline int ext4_match(struct ext4_filename *fname, |
1f3862b5 | 1235 | struct ext4_dir_entry_2 *de) |
ac27a0ec | 1236 | { |
5b643f9c TT |
1237 | const void *name = fname_name(fname); |
1238 | u32 len = fname_len(fname); | |
1f3862b5 | 1239 | |
ac27a0ec DK |
1240 | if (!de->inode) |
1241 | return 0; | |
1f3862b5 MH |
1242 | |
1243 | #ifdef CONFIG_EXT4_FS_ENCRYPTION | |
5b643f9c TT |
1244 | if (unlikely(!name)) { |
1245 | if (fname->usr_fname->name[0] == '_') { | |
1246 | int ret; | |
1247 | if (de->name_len < 16) | |
1248 | return 0; | |
1249 | ret = memcmp(de->name + de->name_len - 16, | |
1250 | fname->crypto_buf.name + 8, 16); | |
1251 | return (ret == 0) ? 1 : 0; | |
1252 | } | |
1253 | name = fname->crypto_buf.name; | |
1254 | len = fname->crypto_buf.len; | |
1255 | } | |
1f3862b5 | 1256 | #endif |
5b643f9c | 1257 | if (de->name_len != len) |
1f3862b5 | 1258 | return 0; |
5b643f9c | 1259 | return (memcmp(de->name, name, len) == 0) ? 1 : 0; |
ac27a0ec DK |
1260 | } |
1261 | ||
1262 | /* | |
1263 | * Returns 0 if not found, -1 on failure, and 1 on success | |
1264 | */ | |
5b643f9c TT |
1265 | int ext4_search_dir(struct buffer_head *bh, char *search_buf, int buf_size, |
1266 | struct inode *dir, struct ext4_filename *fname, | |
1267 | const struct qstr *d_name, | |
1268 | unsigned int offset, struct ext4_dir_entry_2 **res_dir) | |
ac27a0ec | 1269 | { |
617ba13b | 1270 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
1271 | char * dlimit; |
1272 | int de_len; | |
1f3862b5 MH |
1273 | int res; |
1274 | ||
7335cd3b TM |
1275 | de = (struct ext4_dir_entry_2 *)search_buf; |
1276 | dlimit = search_buf + buf_size; | |
ac27a0ec DK |
1277 | while ((char *) de < dlimit) { |
1278 | /* this code is executed quadratically often */ | |
1279 | /* do minimal checking `by hand' */ | |
1f3862b5 | 1280 | if ((char *) de + de->name_len <= dlimit) { |
5b643f9c | 1281 | res = ext4_match(fname, de); |
1f3862b5 MH |
1282 | if (res < 0) { |
1283 | res = -1; | |
1284 | goto return_result; | |
1285 | } | |
1286 | if (res > 0) { | |
1287 | /* found a match - just to be sure, do | |
1288 | * a full check */ | |
1289 | if (ext4_check_dir_entry(dir, NULL, de, bh, | |
1290 | bh->b_data, | |
1291 | bh->b_size, offset)) { | |
1292 | res = -1; | |
1293 | goto return_result; | |
1294 | } | |
1295 | *res_dir = de; | |
1296 | res = 1; | |
1297 | goto return_result; | |
1298 | } | |
ac27a0ec | 1299 | |
ac27a0ec DK |
1300 | } |
1301 | /* prevent looping on a bad block */ | |
3d0518f4 WY |
1302 | de_len = ext4_rec_len_from_disk(de->rec_len, |
1303 | dir->i_sb->s_blocksize); | |
1f3862b5 MH |
1304 | if (de_len <= 0) { |
1305 | res = -1; | |
1306 | goto return_result; | |
1307 | } | |
ac27a0ec | 1308 | offset += de_len; |
617ba13b | 1309 | de = (struct ext4_dir_entry_2 *) ((char *) de + de_len); |
ac27a0ec | 1310 | } |
1f3862b5 MH |
1311 | |
1312 | res = 0; | |
1313 | return_result: | |
1f3862b5 | 1314 | return res; |
ac27a0ec DK |
1315 | } |
1316 | ||
c6af8803 DW |
1317 | static int is_dx_internal_node(struct inode *dir, ext4_lblk_t block, |
1318 | struct ext4_dir_entry *de) | |
1319 | { | |
1320 | struct super_block *sb = dir->i_sb; | |
1321 | ||
1322 | if (!is_dx(dir)) | |
1323 | return 0; | |
1324 | if (block == 0) | |
1325 | return 1; | |
1326 | if (de->inode == 0 && | |
1327 | ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) == | |
1328 | sb->s_blocksize) | |
1329 | return 1; | |
1330 | return 0; | |
1331 | } | |
ac27a0ec DK |
1332 | |
1333 | /* | |
617ba13b | 1334 | * ext4_find_entry() |
ac27a0ec DK |
1335 | * |
1336 | * finds an entry in the specified directory with the wanted name. It | |
1337 | * returns the cache buffer in which the entry was found, and the entry | |
1338 | * itself (as a parameter - res_dir). It does NOT read the inode of the | |
1339 | * entry - you'll have to do that yourself if you want to. | |
1340 | * | |
1341 | * The returned buffer_head has ->b_count elevated. The caller is expected | |
1342 | * to brelse() it when appropriate. | |
1343 | */ | |
f702ba0f TT |
1344 | static struct buffer_head * ext4_find_entry (struct inode *dir, |
1345 | const struct qstr *d_name, | |
32f7f22c TM |
1346 | struct ext4_dir_entry_2 **res_dir, |
1347 | int *inlined) | |
ac27a0ec | 1348 | { |
af5bc92d TT |
1349 | struct super_block *sb; |
1350 | struct buffer_head *bh_use[NAMEI_RA_SIZE]; | |
1351 | struct buffer_head *bh, *ret = NULL; | |
725d26d3 | 1352 | ext4_lblk_t start, block, b; |
8941ec8b | 1353 | const u8 *name = d_name->name; |
ac27a0ec DK |
1354 | int ra_max = 0; /* Number of bh's in the readahead |
1355 | buffer, bh_use[] */ | |
1356 | int ra_ptr = 0; /* Current index into readahead | |
1357 | buffer */ | |
1358 | int num = 0; | |
725d26d3 | 1359 | ext4_lblk_t nblocks; |
5b643f9c TT |
1360 | int i, namelen, retval; |
1361 | struct ext4_filename fname; | |
ac27a0ec DK |
1362 | |
1363 | *res_dir = NULL; | |
1364 | sb = dir->i_sb; | |
f702ba0f | 1365 | namelen = d_name->len; |
617ba13b | 1366 | if (namelen > EXT4_NAME_LEN) |
ac27a0ec | 1367 | return NULL; |
e8e948e7 | 1368 | |
5b643f9c TT |
1369 | retval = ext4_fname_setup_filename(dir, d_name, 1, &fname); |
1370 | if (retval) | |
1371 | return ERR_PTR(retval); | |
1372 | ||
e8e948e7 TM |
1373 | if (ext4_has_inline_data(dir)) { |
1374 | int has_inline_data = 1; | |
5b643f9c | 1375 | ret = ext4_find_inline_entry(dir, &fname, d_name, res_dir, |
e8e948e7 | 1376 | &has_inline_data); |
32f7f22c TM |
1377 | if (has_inline_data) { |
1378 | if (inlined) | |
1379 | *inlined = 1; | |
5b643f9c | 1380 | goto cleanup_and_exit; |
32f7f22c | 1381 | } |
e8e948e7 TM |
1382 | } |
1383 | ||
8941ec8b | 1384 | if ((namelen <= 2) && (name[0] == '.') && |
6d5c3aa8 | 1385 | (name[1] == '.' || name[1] == '\0')) { |
8941ec8b TT |
1386 | /* |
1387 | * "." or ".." will only be in the first block | |
1388 | * NFS may look up ".."; "." should be handled by the VFS | |
1389 | */ | |
1390 | block = start = 0; | |
1391 | nblocks = 1; | |
1392 | goto restart; | |
1393 | } | |
ac27a0ec | 1394 | if (is_dx(dir)) { |
5b643f9c | 1395 | ret = ext4_dx_find_entry(dir, &fname, res_dir); |
ac27a0ec DK |
1396 | /* |
1397 | * On success, or if the error was file not found, | |
1398 | * return. Otherwise, fall back to doing a search the | |
1399 | * old fashioned way. | |
1400 | */ | |
5b643f9c TT |
1401 | if (!IS_ERR(ret) || PTR_ERR(ret) != ERR_BAD_DX_DIR) |
1402 | goto cleanup_and_exit; | |
4776004f TT |
1403 | dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, " |
1404 | "falling back\n")); | |
ac27a0ec | 1405 | } |
617ba13b MC |
1406 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
1407 | start = EXT4_I(dir)->i_dir_start_lookup; | |
ac27a0ec DK |
1408 | if (start >= nblocks) |
1409 | start = 0; | |
1410 | block = start; | |
1411 | restart: | |
1412 | do { | |
1413 | /* | |
1414 | * We deal with the read-ahead logic here. | |
1415 | */ | |
1416 | if (ra_ptr >= ra_max) { | |
1417 | /* Refill the readahead buffer */ | |
1418 | ra_ptr = 0; | |
1419 | b = block; | |
1420 | for (ra_max = 0; ra_max < NAMEI_RA_SIZE; ra_max++) { | |
1421 | /* | |
1422 | * Terminate if we reach the end of the | |
1423 | * directory and must wrap, or if our | |
1424 | * search has finished at this block. | |
1425 | */ | |
1426 | if (b >= nblocks || (num && block == start)) { | |
1427 | bh_use[ra_max] = NULL; | |
1428 | break; | |
1429 | } | |
1430 | num++; | |
10560082 | 1431 | bh = ext4_getblk(NULL, dir, b++, 0); |
a1c83681 | 1432 | if (IS_ERR(bh)) { |
5b643f9c TT |
1433 | if (ra_max == 0) { |
1434 | ret = bh; | |
1435 | goto cleanup_and_exit; | |
1436 | } | |
36de9286 TT |
1437 | break; |
1438 | } | |
ac27a0ec DK |
1439 | bh_use[ra_max] = bh; |
1440 | if (bh) | |
65299a3b CH |
1441 | ll_rw_block(READ | REQ_META | REQ_PRIO, |
1442 | 1, &bh); | |
ac27a0ec DK |
1443 | } |
1444 | } | |
1445 | if ((bh = bh_use[ra_ptr++]) == NULL) | |
1446 | goto next; | |
1447 | wait_on_buffer(bh); | |
1448 | if (!buffer_uptodate(bh)) { | |
1449 | /* read error, skip block & hope for the best */ | |
24676da4 TT |
1450 | EXT4_ERROR_INODE(dir, "reading directory lblock %lu", |
1451 | (unsigned long) block); | |
ac27a0ec DK |
1452 | brelse(bh); |
1453 | goto next; | |
1454 | } | |
b0336e8d | 1455 | if (!buffer_verified(bh) && |
c6af8803 DW |
1456 | !is_dx_internal_node(dir, block, |
1457 | (struct ext4_dir_entry *)bh->b_data) && | |
b0336e8d DW |
1458 | !ext4_dirent_csum_verify(dir, |
1459 | (struct ext4_dir_entry *)bh->b_data)) { | |
1460 | EXT4_ERROR_INODE(dir, "checksumming directory " | |
1461 | "block %lu", (unsigned long)block); | |
1462 | brelse(bh); | |
1463 | goto next; | |
1464 | } | |
1465 | set_buffer_verified(bh); | |
5b643f9c | 1466 | i = search_dirblock(bh, dir, &fname, d_name, |
617ba13b | 1467 | block << EXT4_BLOCK_SIZE_BITS(sb), res_dir); |
ac27a0ec | 1468 | if (i == 1) { |
617ba13b | 1469 | EXT4_I(dir)->i_dir_start_lookup = block; |
ac27a0ec DK |
1470 | ret = bh; |
1471 | goto cleanup_and_exit; | |
1472 | } else { | |
1473 | brelse(bh); | |
1474 | if (i < 0) | |
1475 | goto cleanup_and_exit; | |
1476 | } | |
1477 | next: | |
1478 | if (++block >= nblocks) | |
1479 | block = 0; | |
1480 | } while (block != start); | |
1481 | ||
1482 | /* | |
1483 | * If the directory has grown while we were searching, then | |
1484 | * search the last part of the directory before giving up. | |
1485 | */ | |
1486 | block = nblocks; | |
617ba13b | 1487 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
1488 | if (block < nblocks) { |
1489 | start = 0; | |
1490 | goto restart; | |
1491 | } | |
1492 | ||
1493 | cleanup_and_exit: | |
1494 | /* Clean up the read-ahead blocks */ | |
1495 | for (; ra_ptr < ra_max; ra_ptr++) | |
af5bc92d | 1496 | brelse(bh_use[ra_ptr]); |
5b643f9c | 1497 | ext4_fname_free_filename(&fname); |
ac27a0ec DK |
1498 | return ret; |
1499 | } | |
1500 | ||
5b643f9c TT |
1501 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, |
1502 | struct ext4_filename *fname, | |
1503 | struct ext4_dir_entry_2 **res_dir) | |
ac27a0ec | 1504 | { |
8941ec8b | 1505 | struct super_block * sb = dir->i_sb; |
ac27a0ec | 1506 | struct dx_frame frames[2], *frame; |
5b643f9c | 1507 | const struct qstr *d_name = fname->usr_fname; |
ac27a0ec | 1508 | struct buffer_head *bh; |
725d26d3 | 1509 | ext4_lblk_t block; |
ac27a0ec | 1510 | int retval; |
ac27a0ec | 1511 | |
1f3862b5 MH |
1512 | #ifdef CONFIG_EXT4_FS_ENCRYPTION |
1513 | *res_dir = NULL; | |
1514 | #endif | |
5b643f9c | 1515 | frame = dx_probe(fname, dir, NULL, frames); |
dd73b5d5 TT |
1516 | if (IS_ERR(frame)) |
1517 | return (struct buffer_head *) frame; | |
ac27a0ec DK |
1518 | do { |
1519 | block = dx_get_block(frame->at); | |
dc6982ff | 1520 | bh = ext4_read_dirblock(dir, block, DIRENT); |
537d8f93 | 1521 | if (IS_ERR(bh)) |
b0336e8d | 1522 | goto errout; |
537d8f93 | 1523 | |
5b643f9c | 1524 | retval = search_dirblock(bh, dir, fname, d_name, |
7845c049 TT |
1525 | block << EXT4_BLOCK_SIZE_BITS(sb), |
1526 | res_dir); | |
537d8f93 TT |
1527 | if (retval == 1) |
1528 | goto success; | |
af5bc92d | 1529 | brelse(bh); |
7845c049 | 1530 | if (retval == -1) { |
537d8f93 | 1531 | bh = ERR_PTR(ERR_BAD_DX_DIR); |
7845c049 TT |
1532 | goto errout; |
1533 | } | |
1534 | ||
ac27a0ec | 1535 | /* Check to see if we should continue to search */ |
5b643f9c | 1536 | retval = ext4_htree_next_block(dir, fname->hinfo.hash, frame, |
ac27a0ec DK |
1537 | frames, NULL); |
1538 | if (retval < 0) { | |
b03a2f7e AD |
1539 | ext4_warning_inode(dir, |
1540 | "error %d reading directory index block", | |
1541 | retval); | |
537d8f93 | 1542 | bh = ERR_PTR(retval); |
ac27a0ec DK |
1543 | goto errout; |
1544 | } | |
1545 | } while (retval == 1); | |
1546 | ||
537d8f93 | 1547 | bh = NULL; |
ac27a0ec | 1548 | errout: |
265c6a0f | 1549 | dxtrace(printk(KERN_DEBUG "%s not found\n", d_name->name)); |
537d8f93 TT |
1550 | success: |
1551 | dx_release(frames); | |
1552 | return bh; | |
ac27a0ec | 1553 | } |
ac27a0ec | 1554 | |
00cd8dd3 | 1555 | static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
ac27a0ec | 1556 | { |
af5bc92d TT |
1557 | struct inode *inode; |
1558 | struct ext4_dir_entry_2 *de; | |
1559 | struct buffer_head *bh; | |
ac27a0ec | 1560 | |
617ba13b | 1561 | if (dentry->d_name.len > EXT4_NAME_LEN) |
ac27a0ec DK |
1562 | return ERR_PTR(-ENAMETOOLONG); |
1563 | ||
32f7f22c | 1564 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 TT |
1565 | if (IS_ERR(bh)) |
1566 | return (struct dentry *) bh; | |
ac27a0ec DK |
1567 | inode = NULL; |
1568 | if (bh) { | |
498e5f24 | 1569 | __u32 ino = le32_to_cpu(de->inode); |
af5bc92d | 1570 | brelse(bh); |
617ba13b | 1571 | if (!ext4_valid_inum(dir->i_sb, ino)) { |
24676da4 | 1572 | EXT4_ERROR_INODE(dir, "bad inode number: %u", ino); |
6a797d27 | 1573 | return ERR_PTR(-EFSCORRUPTED); |
a6c15c2b | 1574 | } |
7e936b73 | 1575 | if (unlikely(ino == dir->i_ino)) { |
a34e15cc DH |
1576 | EXT4_ERROR_INODE(dir, "'%pd' linked to parent dir", |
1577 | dentry); | |
6a797d27 | 1578 | return ERR_PTR(-EFSCORRUPTED); |
7e936b73 | 1579 | } |
f4bb2981 | 1580 | inode = ext4_iget_normal(dir->i_sb, ino); |
a9049376 AV |
1581 | if (inode == ERR_PTR(-ESTALE)) { |
1582 | EXT4_ERROR_INODE(dir, | |
1583 | "deleted inode referenced: %u", | |
1584 | ino); | |
6a797d27 | 1585 | return ERR_PTR(-EFSCORRUPTED); |
e6f009b0 | 1586 | } |
d9cdc903 TT |
1587 | if (!IS_ERR(inode) && ext4_encrypted_inode(dir) && |
1588 | (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || | |
1589 | S_ISLNK(inode->i_mode)) && | |
1590 | !ext4_is_child_context_consistent_with_parent(dir, | |
1591 | inode)) { | |
1592 | iput(inode); | |
1593 | ext4_warning(inode->i_sb, | |
1594 | "Inconsistent encryption contexts: %lu/%lu\n", | |
1595 | (unsigned long) dir->i_ino, | |
1596 | (unsigned long) inode->i_ino); | |
1597 | return ERR_PTR(-EPERM); | |
1598 | } | |
ac27a0ec DK |
1599 | } |
1600 | return d_splice_alias(inode, dentry); | |
1601 | } | |
1602 | ||
1603 | ||
617ba13b | 1604 | struct dentry *ext4_get_parent(struct dentry *child) |
ac27a0ec | 1605 | { |
498e5f24 | 1606 | __u32 ino; |
26fe5750 | 1607 | static const struct qstr dotdot = QSTR_INIT("..", 2); |
617ba13b | 1608 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
1609 | struct buffer_head *bh; |
1610 | ||
2b0143b5 | 1611 | bh = ext4_find_entry(d_inode(child), &dotdot, &de, NULL); |
36de9286 TT |
1612 | if (IS_ERR(bh)) |
1613 | return (struct dentry *) bh; | |
ac27a0ec DK |
1614 | if (!bh) |
1615 | return ERR_PTR(-ENOENT); | |
1616 | ino = le32_to_cpu(de->inode); | |
1617 | brelse(bh); | |
1618 | ||
2b0143b5 DH |
1619 | if (!ext4_valid_inum(d_inode(child)->i_sb, ino)) { |
1620 | EXT4_ERROR_INODE(d_inode(child), | |
24676da4 | 1621 | "bad parent inode number: %u", ino); |
6a797d27 | 1622 | return ERR_PTR(-EFSCORRUPTED); |
a6c15c2b VA |
1623 | } |
1624 | ||
2b0143b5 | 1625 | return d_obtain_alias(ext4_iget_normal(d_inode(child)->i_sb, ino)); |
ac27a0ec DK |
1626 | } |
1627 | ||
ef2b02d3 ES |
1628 | /* |
1629 | * Move count entries from end of map between two memory locations. | |
1630 | * Returns pointer to last entry moved. | |
1631 | */ | |
617ba13b | 1632 | static struct ext4_dir_entry_2 * |
3d0518f4 WY |
1633 | dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count, |
1634 | unsigned blocksize) | |
ac27a0ec DK |
1635 | { |
1636 | unsigned rec_len = 0; | |
1637 | ||
1638 | while (count--) { | |
60e6679e | 1639 | struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *) |
9aee2286 | 1640 | (from + (map->offs<<2)); |
617ba13b | 1641 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec | 1642 | memcpy (to, de, rec_len); |
617ba13b | 1643 | ((struct ext4_dir_entry_2 *) to)->rec_len = |
3d0518f4 | 1644 | ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec DK |
1645 | de->inode = 0; |
1646 | map++; | |
1647 | to += rec_len; | |
1648 | } | |
617ba13b | 1649 | return (struct ext4_dir_entry_2 *) (to - rec_len); |
ac27a0ec DK |
1650 | } |
1651 | ||
ef2b02d3 ES |
1652 | /* |
1653 | * Compact each dir entry in the range to the minimal rec_len. | |
1654 | * Returns pointer to last entry in range. | |
1655 | */ | |
8bad4597 | 1656 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize) |
ac27a0ec | 1657 | { |
617ba13b | 1658 | struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base; |
ac27a0ec DK |
1659 | unsigned rec_len = 0; |
1660 | ||
1661 | prev = to = de; | |
8bad4597 | 1662 | while ((char*)de < base + blocksize) { |
3d0518f4 | 1663 | next = ext4_next_entry(de, blocksize); |
ac27a0ec | 1664 | if (de->inode && de->name_len) { |
617ba13b | 1665 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
1666 | if (de > to) |
1667 | memmove(to, de, rec_len); | |
3d0518f4 | 1668 | to->rec_len = ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec | 1669 | prev = to; |
617ba13b | 1670 | to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len); |
ac27a0ec DK |
1671 | } |
1672 | de = next; | |
1673 | } | |
1674 | return prev; | |
1675 | } | |
1676 | ||
ef2b02d3 ES |
1677 | /* |
1678 | * Split a full leaf block to make room for a new dir entry. | |
1679 | * Allocate a new block, and move entries so that they are approx. equally full. | |
1680 | * Returns pointer to de in block into which the new entry will be inserted. | |
1681 | */ | |
617ba13b | 1682 | static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, |
ac27a0ec | 1683 | struct buffer_head **bh,struct dx_frame *frame, |
f8b3b59d | 1684 | struct dx_hash_info *hinfo) |
ac27a0ec DK |
1685 | { |
1686 | unsigned blocksize = dir->i_sb->s_blocksize; | |
1687 | unsigned count, continued; | |
1688 | struct buffer_head *bh2; | |
725d26d3 | 1689 | ext4_lblk_t newblock; |
ac27a0ec DK |
1690 | u32 hash2; |
1691 | struct dx_map_entry *map; | |
1692 | char *data1 = (*bh)->b_data, *data2; | |
59e315b4 | 1693 | unsigned split, move, size; |
617ba13b | 1694 | struct ext4_dir_entry_2 *de = NULL, *de2; |
b0336e8d DW |
1695 | struct ext4_dir_entry_tail *t; |
1696 | int csum_size = 0; | |
59e315b4 | 1697 | int err = 0, i; |
ac27a0ec | 1698 | |
9aa5d32b | 1699 | if (ext4_has_metadata_csum(dir->i_sb)) |
b0336e8d DW |
1700 | csum_size = sizeof(struct ext4_dir_entry_tail); |
1701 | ||
0f70b406 TT |
1702 | bh2 = ext4_append(handle, dir, &newblock); |
1703 | if (IS_ERR(bh2)) { | |
ac27a0ec DK |
1704 | brelse(*bh); |
1705 | *bh = NULL; | |
f8b3b59d | 1706 | return (struct ext4_dir_entry_2 *) bh2; |
ac27a0ec DK |
1707 | } |
1708 | ||
1709 | BUFFER_TRACE(*bh, "get_write_access"); | |
617ba13b | 1710 | err = ext4_journal_get_write_access(handle, *bh); |
fedee54d DM |
1711 | if (err) |
1712 | goto journal_error; | |
1713 | ||
ac27a0ec | 1714 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 1715 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
1716 | if (err) |
1717 | goto journal_error; | |
1718 | ||
1719 | data2 = bh2->b_data; | |
1720 | ||
1721 | /* create map in the end of data2 block */ | |
1722 | map = (struct dx_map_entry *) (data2 + blocksize); | |
1f3862b5 | 1723 | count = dx_make_map(dir, (struct ext4_dir_entry_2 *) data1, |
ac27a0ec DK |
1724 | blocksize, hinfo, map); |
1725 | map -= count; | |
af5bc92d | 1726 | dx_sort_map(map, count); |
ef2b02d3 ES |
1727 | /* Split the existing block in the middle, size-wise */ |
1728 | size = 0; | |
1729 | move = 0; | |
1730 | for (i = count-1; i >= 0; i--) { | |
1731 | /* is more than half of this entry in 2nd half of the block? */ | |
1732 | if (size + map[i].size/2 > blocksize/2) | |
1733 | break; | |
1734 | size += map[i].size; | |
1735 | move++; | |
1736 | } | |
1737 | /* map index at which we will split */ | |
1738 | split = count - move; | |
ac27a0ec DK |
1739 | hash2 = map[split].hash; |
1740 | continued = hash2 == map[split - 1].hash; | |
725d26d3 AK |
1741 | dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n", |
1742 | (unsigned long)dx_get_block(frame->at), | |
1743 | hash2, split, count-split)); | |
ac27a0ec DK |
1744 | |
1745 | /* Fancy dance to stay within two buffers */ | |
1f3862b5 MH |
1746 | de2 = dx_move_dirents(data1, data2, map + split, count - split, |
1747 | blocksize); | |
af5bc92d | 1748 | de = dx_pack_dirents(data1, blocksize); |
b0336e8d DW |
1749 | de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) - |
1750 | (char *) de, | |
3d0518f4 | 1751 | blocksize); |
b0336e8d DW |
1752 | de2->rec_len = ext4_rec_len_to_disk(data2 + (blocksize - csum_size) - |
1753 | (char *) de2, | |
3d0518f4 | 1754 | blocksize); |
b0336e8d DW |
1755 | if (csum_size) { |
1756 | t = EXT4_DIRENT_TAIL(data2, blocksize); | |
1757 | initialize_dirent_tail(t, blocksize); | |
1758 | ||
1759 | t = EXT4_DIRENT_TAIL(data1, blocksize); | |
1760 | initialize_dirent_tail(t, blocksize); | |
1761 | } | |
1762 | ||
b3098486 MH |
1763 | dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data1, |
1764 | blocksize, 1)); | |
1765 | dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data2, | |
1766 | blocksize, 1)); | |
ac27a0ec DK |
1767 | |
1768 | /* Which block gets the new entry? */ | |
f8b3b59d | 1769 | if (hinfo->hash >= hash2) { |
ac27a0ec DK |
1770 | swap(*bh, bh2); |
1771 | de = de2; | |
1772 | } | |
af5bc92d | 1773 | dx_insert_block(frame, hash2 + continued, newblock); |
b0336e8d | 1774 | err = ext4_handle_dirty_dirent_node(handle, dir, bh2); |
ac27a0ec DK |
1775 | if (err) |
1776 | goto journal_error; | |
dbe89444 | 1777 | err = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
ac27a0ec DK |
1778 | if (err) |
1779 | goto journal_error; | |
af5bc92d TT |
1780 | brelse(bh2); |
1781 | dxtrace(dx_show_index("frame", frame->entries)); | |
ac27a0ec | 1782 | return de; |
fedee54d DM |
1783 | |
1784 | journal_error: | |
1785 | brelse(*bh); | |
1786 | brelse(bh2); | |
1787 | *bh = NULL; | |
1788 | ext4_std_error(dir->i_sb, err); | |
f8b3b59d | 1789 | return ERR_PTR(err); |
ac27a0ec | 1790 | } |
ac27a0ec | 1791 | |
978fef91 TM |
1792 | int ext4_find_dest_de(struct inode *dir, struct inode *inode, |
1793 | struct buffer_head *bh, | |
1794 | void *buf, int buf_size, | |
5b643f9c | 1795 | struct ext4_filename *fname, |
978fef91 TM |
1796 | struct ext4_dir_entry_2 **dest_de) |
1797 | { | |
1798 | struct ext4_dir_entry_2 *de; | |
5b643f9c | 1799 | unsigned short reclen = EXT4_DIR_REC_LEN(fname_len(fname)); |
978fef91 TM |
1800 | int nlen, rlen; |
1801 | unsigned int offset = 0; | |
1802 | char *top; | |
1f3862b5 MH |
1803 | int res; |
1804 | ||
978fef91 TM |
1805 | de = (struct ext4_dir_entry_2 *)buf; |
1806 | top = buf + buf_size - reclen; | |
1807 | while ((char *) de <= top) { | |
1808 | if (ext4_check_dir_entry(dir, NULL, de, bh, | |
1f3862b5 | 1809 | buf, buf_size, offset)) { |
6a797d27 | 1810 | res = -EFSCORRUPTED; |
1f3862b5 MH |
1811 | goto return_result; |
1812 | } | |
1813 | /* Provide crypto context and crypto buffer to ext4 match */ | |
5b643f9c | 1814 | res = ext4_match(fname, de); |
1f3862b5 MH |
1815 | if (res < 0) |
1816 | goto return_result; | |
1817 | if (res > 0) { | |
1818 | res = -EEXIST; | |
1819 | goto return_result; | |
1820 | } | |
978fef91 TM |
1821 | nlen = EXT4_DIR_REC_LEN(de->name_len); |
1822 | rlen = ext4_rec_len_from_disk(de->rec_len, buf_size); | |
1823 | if ((de->inode ? rlen - nlen : rlen) >= reclen) | |
1824 | break; | |
1825 | de = (struct ext4_dir_entry_2 *)((char *)de + rlen); | |
1826 | offset += rlen; | |
1827 | } | |
978fef91 | 1828 | |
1f3862b5 MH |
1829 | if ((char *) de > top) |
1830 | res = -ENOSPC; | |
1831 | else { | |
1832 | *dest_de = de; | |
1833 | res = 0; | |
1834 | } | |
1835 | return_result: | |
1f3862b5 | 1836 | return res; |
978fef91 TM |
1837 | } |
1838 | ||
4bdfc873 MH |
1839 | int ext4_insert_dentry(struct inode *dir, |
1840 | struct inode *inode, | |
1841 | struct ext4_dir_entry_2 *de, | |
1842 | int buf_size, | |
5b643f9c | 1843 | struct ext4_filename *fname) |
978fef91 TM |
1844 | { |
1845 | ||
1846 | int nlen, rlen; | |
1847 | ||
1848 | nlen = EXT4_DIR_REC_LEN(de->name_len); | |
1849 | rlen = ext4_rec_len_from_disk(de->rec_len, buf_size); | |
1850 | if (de->inode) { | |
1851 | struct ext4_dir_entry_2 *de1 = | |
4bdfc873 | 1852 | (struct ext4_dir_entry_2 *)((char *)de + nlen); |
978fef91 TM |
1853 | de1->rec_len = ext4_rec_len_to_disk(rlen - nlen, buf_size); |
1854 | de->rec_len = ext4_rec_len_to_disk(nlen, buf_size); | |
1855 | de = de1; | |
1856 | } | |
1857 | de->file_type = EXT4_FT_UNKNOWN; | |
1858 | de->inode = cpu_to_le32(inode->i_ino); | |
1859 | ext4_set_de_type(inode->i_sb, de, inode->i_mode); | |
5b643f9c TT |
1860 | de->name_len = fname_len(fname); |
1861 | memcpy(de->name, fname_name(fname), fname_len(fname)); | |
4bdfc873 | 1862 | return 0; |
978fef91 | 1863 | } |
4bdfc873 | 1864 | |
ac27a0ec DK |
1865 | /* |
1866 | * Add a new entry into a directory (leaf) block. If de is non-NULL, | |
1867 | * it points to a directory entry which is guaranteed to be large | |
1868 | * enough for new directory entry. If de is NULL, then | |
1869 | * add_dirent_to_buf will attempt search the directory block for | |
1870 | * space. It will return -ENOSPC if no space is available, and -EIO | |
1871 | * and -EEXIST if directory entry already exists. | |
ac27a0ec | 1872 | */ |
5b643f9c TT |
1873 | static int add_dirent_to_buf(handle_t *handle, struct ext4_filename *fname, |
1874 | struct inode *dir, | |
617ba13b | 1875 | struct inode *inode, struct ext4_dir_entry_2 *de, |
af5bc92d | 1876 | struct buffer_head *bh) |
ac27a0ec | 1877 | { |
3d0518f4 | 1878 | unsigned int blocksize = dir->i_sb->s_blocksize; |
b0336e8d | 1879 | int csum_size = 0; |
978fef91 | 1880 | int err; |
b0336e8d | 1881 | |
9aa5d32b | 1882 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 1883 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec | 1884 | |
ac27a0ec | 1885 | if (!de) { |
5b643f9c TT |
1886 | err = ext4_find_dest_de(dir, inode, bh, bh->b_data, |
1887 | blocksize - csum_size, fname, &de); | |
978fef91 TM |
1888 | if (err) |
1889 | return err; | |
ac27a0ec DK |
1890 | } |
1891 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 1892 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1893 | if (err) { |
617ba13b | 1894 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1895 | return err; |
1896 | } | |
1897 | ||
4bdfc873 MH |
1898 | /* By now the buffer is marked for journaling. Due to crypto operations, |
1899 | * the following function call may fail */ | |
5b643f9c | 1900 | err = ext4_insert_dentry(dir, inode, de, blocksize, fname); |
4bdfc873 MH |
1901 | if (err < 0) |
1902 | return err; | |
978fef91 | 1903 | |
ac27a0ec DK |
1904 | /* |
1905 | * XXX shouldn't update any times until successful | |
1906 | * completion of syscall, but too many callers depend | |
1907 | * on this. | |
1908 | * | |
1909 | * XXX similarly, too many callers depend on | |
617ba13b | 1910 | * ext4_new_inode() setting the times, but error |
ac27a0ec DK |
1911 | * recovery deletes the inode, so the worst that can |
1912 | * happen is that the times are slightly out of date | |
1913 | * and/or different from the directory change time. | |
1914 | */ | |
ef7f3835 | 1915 | dir->i_mtime = dir->i_ctime = ext4_current_time(dir); |
617ba13b | 1916 | ext4_update_dx_flag(dir); |
ac27a0ec | 1917 | dir->i_version++; |
617ba13b | 1918 | ext4_mark_inode_dirty(handle, dir); |
0390131b | 1919 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); |
b0336e8d | 1920 | err = ext4_handle_dirty_dirent_node(handle, dir, bh); |
ac27a0ec | 1921 | if (err) |
617ba13b | 1922 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1923 | return 0; |
1924 | } | |
1925 | ||
ac27a0ec DK |
1926 | /* |
1927 | * This converts a one block unindexed directory to a 3 block indexed | |
1928 | * directory, and adds the dentry to the indexed directory. | |
1929 | */ | |
5b643f9c | 1930 | static int make_indexed_dir(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 1931 | struct inode *dir, |
ac27a0ec DK |
1932 | struct inode *inode, struct buffer_head *bh) |
1933 | { | |
ac27a0ec DK |
1934 | struct buffer_head *bh2; |
1935 | struct dx_root *root; | |
1936 | struct dx_frame frames[2], *frame; | |
1937 | struct dx_entry *entries; | |
617ba13b | 1938 | struct ext4_dir_entry_2 *de, *de2; |
b0336e8d | 1939 | struct ext4_dir_entry_tail *t; |
ac27a0ec DK |
1940 | char *data1, *top; |
1941 | unsigned len; | |
1942 | int retval; | |
1943 | unsigned blocksize; | |
725d26d3 | 1944 | ext4_lblk_t block; |
ac27a0ec | 1945 | struct fake_dirent *fde; |
4bdfc873 MH |
1946 | int csum_size = 0; |
1947 | ||
9aa5d32b | 1948 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 1949 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec DK |
1950 | |
1951 | blocksize = dir->i_sb->s_blocksize; | |
e6b8bc09 | 1952 | dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino)); |
5d601255 | 1953 | BUFFER_TRACE(bh, "get_write_access"); |
617ba13b | 1954 | retval = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1955 | if (retval) { |
617ba13b | 1956 | ext4_std_error(dir->i_sb, retval); |
ac27a0ec DK |
1957 | brelse(bh); |
1958 | return retval; | |
1959 | } | |
1960 | root = (struct dx_root *) bh->b_data; | |
1961 | ||
e6b8bc09 TT |
1962 | /* The 0th block becomes the root, move the dirents out */ |
1963 | fde = &root->dotdot; | |
1964 | de = (struct ext4_dir_entry_2 *)((char *)fde + | |
3d0518f4 | 1965 | ext4_rec_len_from_disk(fde->rec_len, blocksize)); |
e6b8bc09 | 1966 | if ((char *) de >= (((char *) root) + blocksize)) { |
24676da4 | 1967 | EXT4_ERROR_INODE(dir, "invalid rec_len for '..'"); |
e6b8bc09 | 1968 | brelse(bh); |
6a797d27 | 1969 | return -EFSCORRUPTED; |
e6b8bc09 | 1970 | } |
b0336e8d | 1971 | len = ((char *) root) + (blocksize - csum_size) - (char *) de; |
e6b8bc09 TT |
1972 | |
1973 | /* Allocate new block for the 0th block's dirents */ | |
0f70b406 TT |
1974 | bh2 = ext4_append(handle, dir, &block); |
1975 | if (IS_ERR(bh2)) { | |
ac27a0ec | 1976 | brelse(bh); |
0f70b406 | 1977 | return PTR_ERR(bh2); |
ac27a0ec | 1978 | } |
12e9b892 | 1979 | ext4_set_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec DK |
1980 | data1 = bh2->b_data; |
1981 | ||
ac27a0ec | 1982 | memcpy (data1, de, len); |
617ba13b | 1983 | de = (struct ext4_dir_entry_2 *) data1; |
ac27a0ec | 1984 | top = data1 + len; |
3d0518f4 | 1985 | while ((char *)(de2 = ext4_next_entry(de, blocksize)) < top) |
ac27a0ec | 1986 | de = de2; |
b0336e8d DW |
1987 | de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) - |
1988 | (char *) de, | |
3d0518f4 | 1989 | blocksize); |
b0336e8d DW |
1990 | |
1991 | if (csum_size) { | |
1992 | t = EXT4_DIRENT_TAIL(data1, blocksize); | |
1993 | initialize_dirent_tail(t, blocksize); | |
1994 | } | |
1995 | ||
ac27a0ec | 1996 | /* Initialize the root; the dot dirents already exist */ |
617ba13b | 1997 | de = (struct ext4_dir_entry_2 *) (&root->dotdot); |
3d0518f4 WY |
1998 | de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(2), |
1999 | blocksize); | |
ac27a0ec DK |
2000 | memset (&root->info, 0, sizeof(root->info)); |
2001 | root->info.info_length = sizeof(root->info); | |
617ba13b | 2002 | root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
ac27a0ec | 2003 | entries = root->entries; |
af5bc92d TT |
2004 | dx_set_block(entries, 1); |
2005 | dx_set_count(entries, 1); | |
2006 | dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info))); | |
ac27a0ec DK |
2007 | |
2008 | /* Initialize as for dx_probe */ | |
5b643f9c TT |
2009 | fname->hinfo.hash_version = root->info.hash_version; |
2010 | if (fname->hinfo.hash_version <= DX_HASH_TEA) | |
2011 | fname->hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
2012 | fname->hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; | |
2013 | ext4fs_dirhash(fname_name(fname), fname_len(fname), &fname->hinfo); | |
2014 | ||
6050d47a | 2015 | memset(frames, 0, sizeof(frames)); |
ac27a0ec DK |
2016 | frame = frames; |
2017 | frame->entries = entries; | |
2018 | frame->at = entries; | |
2019 | frame->bh = bh; | |
2020 | bh = bh2; | |
6976a6f2 | 2021 | |
6050d47a JK |
2022 | retval = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
2023 | if (retval) | |
2024 | goto out_frames; | |
2025 | retval = ext4_handle_dirty_dirent_node(handle, dir, bh); | |
2026 | if (retval) | |
2027 | goto out_frames; | |
6976a6f2 | 2028 | |
5b643f9c | 2029 | de = do_split(handle,dir, &bh, frame, &fname->hinfo); |
f8b3b59d | 2030 | if (IS_ERR(de)) { |
6050d47a JK |
2031 | retval = PTR_ERR(de); |
2032 | goto out_frames; | |
7ad8e4e6 JK |
2033 | } |
2034 | dx_release(frames); | |
ac27a0ec | 2035 | |
5b643f9c | 2036 | retval = add_dirent_to_buf(handle, fname, dir, inode, de, bh); |
2de770a4 TT |
2037 | brelse(bh); |
2038 | return retval; | |
6050d47a JK |
2039 | out_frames: |
2040 | /* | |
2041 | * Even if the block split failed, we have to properly write | |
2042 | * out all the changes we did so far. Otherwise we can end up | |
2043 | * with corrupted filesystem. | |
2044 | */ | |
2045 | ext4_mark_inode_dirty(handle, dir); | |
2046 | dx_release(frames); | |
2047 | return retval; | |
ac27a0ec | 2048 | } |
ac27a0ec DK |
2049 | |
2050 | /* | |
617ba13b | 2051 | * ext4_add_entry() |
ac27a0ec DK |
2052 | * |
2053 | * adds a file entry to the specified directory, using the same | |
617ba13b | 2054 | * semantics as ext4_find_entry(). It returns NULL if it failed. |
ac27a0ec DK |
2055 | * |
2056 | * NOTE!! The inode part of 'de' is left at 0 - which means you | |
2057 | * may not sleep between calling this and putting something into | |
2058 | * the entry, as someone else might have used it while you slept. | |
2059 | */ | |
af5bc92d TT |
2060 | static int ext4_add_entry(handle_t *handle, struct dentry *dentry, |
2061 | struct inode *inode) | |
ac27a0ec | 2062 | { |
2b0143b5 | 2063 | struct inode *dir = d_inode(dentry->d_parent); |
e12fb972 | 2064 | struct buffer_head *bh = NULL; |
617ba13b | 2065 | struct ext4_dir_entry_2 *de; |
b0336e8d | 2066 | struct ext4_dir_entry_tail *t; |
af5bc92d | 2067 | struct super_block *sb; |
5b643f9c | 2068 | struct ext4_filename fname; |
ac27a0ec | 2069 | int retval; |
ac27a0ec | 2070 | int dx_fallback=0; |
ac27a0ec | 2071 | unsigned blocksize; |
725d26d3 | 2072 | ext4_lblk_t block, blocks; |
b0336e8d DW |
2073 | int csum_size = 0; |
2074 | ||
9aa5d32b | 2075 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 2076 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec DK |
2077 | |
2078 | sb = dir->i_sb; | |
2079 | blocksize = sb->s_blocksize; | |
2080 | if (!dentry->d_name.len) | |
2081 | return -EINVAL; | |
3c47d541 | 2082 | |
5b643f9c TT |
2083 | retval = ext4_fname_setup_filename(dir, &dentry->d_name, 0, &fname); |
2084 | if (retval) | |
2085 | return retval; | |
2086 | ||
3c47d541 | 2087 | if (ext4_has_inline_data(dir)) { |
56a04915 | 2088 | retval = ext4_try_add_inline_entry(handle, &fname, dir, inode); |
3c47d541 | 2089 | if (retval < 0) |
5b643f9c | 2090 | goto out; |
3c47d541 TM |
2091 | if (retval == 1) { |
2092 | retval = 0; | |
e12fb972 | 2093 | goto out; |
3c47d541 TM |
2094 | } |
2095 | } | |
2096 | ||
ac27a0ec | 2097 | if (is_dx(dir)) { |
56a04915 | 2098 | retval = ext4_dx_add_entry(handle, &fname, dir, inode); |
ac27a0ec | 2099 | if (!retval || (retval != ERR_BAD_DX_DIR)) |
e12fb972 | 2100 | goto out; |
12e9b892 | 2101 | ext4_clear_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec | 2102 | dx_fallback++; |
617ba13b | 2103 | ext4_mark_inode_dirty(handle, dir); |
ac27a0ec | 2104 | } |
ac27a0ec | 2105 | blocks = dir->i_size >> sb->s_blocksize_bits; |
498e5f24 | 2106 | for (block = 0; block < blocks; block++) { |
dc6982ff | 2107 | bh = ext4_read_dirblock(dir, block, DIRENT); |
5b643f9c TT |
2108 | if (IS_ERR(bh)) { |
2109 | retval = PTR_ERR(bh); | |
2110 | bh = NULL; | |
2111 | goto out; | |
2112 | } | |
2113 | retval = add_dirent_to_buf(handle, &fname, dir, inode, | |
2114 | NULL, bh); | |
e12fb972 LC |
2115 | if (retval != -ENOSPC) |
2116 | goto out; | |
ac27a0ec | 2117 | |
ac27a0ec | 2118 | if (blocks == 1 && !dx_fallback && |
e2b911c5 | 2119 | ext4_has_feature_dir_index(sb)) { |
56a04915 | 2120 | retval = make_indexed_dir(handle, &fname, dir, |
5b643f9c | 2121 | inode, bh); |
e12fb972 LC |
2122 | bh = NULL; /* make_indexed_dir releases bh */ |
2123 | goto out; | |
2124 | } | |
ac27a0ec DK |
2125 | brelse(bh); |
2126 | } | |
0f70b406 | 2127 | bh = ext4_append(handle, dir, &block); |
5b643f9c TT |
2128 | if (IS_ERR(bh)) { |
2129 | retval = PTR_ERR(bh); | |
2130 | bh = NULL; | |
2131 | goto out; | |
2132 | } | |
617ba13b | 2133 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 2134 | de->inode = 0; |
b0336e8d DW |
2135 | de->rec_len = ext4_rec_len_to_disk(blocksize - csum_size, blocksize); |
2136 | ||
2137 | if (csum_size) { | |
2138 | t = EXT4_DIRENT_TAIL(bh->b_data, blocksize); | |
2139 | initialize_dirent_tail(t, blocksize); | |
2140 | } | |
2141 | ||
5b643f9c | 2142 | retval = add_dirent_to_buf(handle, &fname, dir, inode, de, bh); |
e12fb972 | 2143 | out: |
5b643f9c | 2144 | ext4_fname_free_filename(&fname); |
2de770a4 | 2145 | brelse(bh); |
14ece102 FM |
2146 | if (retval == 0) |
2147 | ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY); | |
2de770a4 | 2148 | return retval; |
ac27a0ec DK |
2149 | } |
2150 | ||
ac27a0ec DK |
2151 | /* |
2152 | * Returns 0 for success, or a negative error value | |
2153 | */ | |
5b643f9c | 2154 | static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 2155 | struct inode *dir, struct inode *inode) |
ac27a0ec DK |
2156 | { |
2157 | struct dx_frame frames[2], *frame; | |
2158 | struct dx_entry *entries, *at; | |
af5bc92d | 2159 | struct buffer_head *bh; |
af5bc92d | 2160 | struct super_block *sb = dir->i_sb; |
617ba13b | 2161 | struct ext4_dir_entry_2 *de; |
ac27a0ec DK |
2162 | int err; |
2163 | ||
5b643f9c | 2164 | frame = dx_probe(fname, dir, NULL, frames); |
dd73b5d5 TT |
2165 | if (IS_ERR(frame)) |
2166 | return PTR_ERR(frame); | |
ac27a0ec DK |
2167 | entries = frame->entries; |
2168 | at = frame->at; | |
dc6982ff TT |
2169 | bh = ext4_read_dirblock(dir, dx_get_block(frame->at), DIRENT); |
2170 | if (IS_ERR(bh)) { | |
2171 | err = PTR_ERR(bh); | |
2172 | bh = NULL; | |
ac27a0ec | 2173 | goto cleanup; |
6d1ab10e | 2174 | } |
ac27a0ec DK |
2175 | |
2176 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 2177 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec DK |
2178 | if (err) |
2179 | goto journal_error; | |
2180 | ||
5b643f9c | 2181 | err = add_dirent_to_buf(handle, fname, dir, inode, NULL, bh); |
2de770a4 | 2182 | if (err != -ENOSPC) |
ac27a0ec | 2183 | goto cleanup; |
ac27a0ec DK |
2184 | |
2185 | /* Block full, should compress but for now just split */ | |
4776004f | 2186 | dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n", |
ac27a0ec DK |
2187 | dx_get_count(entries), dx_get_limit(entries))); |
2188 | /* Need to split index? */ | |
2189 | if (dx_get_count(entries) == dx_get_limit(entries)) { | |
725d26d3 | 2190 | ext4_lblk_t newblock; |
ac27a0ec DK |
2191 | unsigned icount = dx_get_count(entries); |
2192 | int levels = frame - frames; | |
2193 | struct dx_entry *entries2; | |
2194 | struct dx_node *node2; | |
2195 | struct buffer_head *bh2; | |
2196 | ||
2197 | if (levels && (dx_get_count(frames->entries) == | |
2198 | dx_get_limit(frames->entries))) { | |
b03a2f7e | 2199 | ext4_warning_inode(dir, "Directory index full!"); |
ac27a0ec DK |
2200 | err = -ENOSPC; |
2201 | goto cleanup; | |
2202 | } | |
0f70b406 TT |
2203 | bh2 = ext4_append(handle, dir, &newblock); |
2204 | if (IS_ERR(bh2)) { | |
2205 | err = PTR_ERR(bh2); | |
ac27a0ec | 2206 | goto cleanup; |
0f70b406 | 2207 | } |
ac27a0ec DK |
2208 | node2 = (struct dx_node *)(bh2->b_data); |
2209 | entries2 = node2->entries; | |
1f7bebb9 | 2210 | memset(&node2->fake, 0, sizeof(struct fake_dirent)); |
3d0518f4 WY |
2211 | node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize, |
2212 | sb->s_blocksize); | |
ac27a0ec | 2213 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 2214 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
2215 | if (err) |
2216 | goto journal_error; | |
2217 | if (levels) { | |
2218 | unsigned icount1 = icount/2, icount2 = icount - icount1; | |
2219 | unsigned hash2 = dx_get_hash(entries + icount1); | |
4776004f TT |
2220 | dxtrace(printk(KERN_DEBUG "Split index %i/%i\n", |
2221 | icount1, icount2)); | |
ac27a0ec DK |
2222 | |
2223 | BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ | |
617ba13b | 2224 | err = ext4_journal_get_write_access(handle, |
ac27a0ec DK |
2225 | frames[0].bh); |
2226 | if (err) | |
2227 | goto journal_error; | |
2228 | ||
af5bc92d TT |
2229 | memcpy((char *) entries2, (char *) (entries + icount1), |
2230 | icount2 * sizeof(struct dx_entry)); | |
2231 | dx_set_count(entries, icount1); | |
2232 | dx_set_count(entries2, icount2); | |
2233 | dx_set_limit(entries2, dx_node_limit(dir)); | |
ac27a0ec DK |
2234 | |
2235 | /* Which index block gets the new entry? */ | |
2236 | if (at - entries >= icount1) { | |
2237 | frame->at = at = at - entries - icount1 + entries2; | |
2238 | frame->entries = entries = entries2; | |
2239 | swap(frame->bh, bh2); | |
2240 | } | |
af5bc92d TT |
2241 | dx_insert_block(frames + 0, hash2, newblock); |
2242 | dxtrace(dx_show_index("node", frames[1].entries)); | |
2243 | dxtrace(dx_show_index("node", | |
ac27a0ec | 2244 | ((struct dx_node *) bh2->b_data)->entries)); |
dbe89444 | 2245 | err = ext4_handle_dirty_dx_node(handle, dir, bh2); |
ac27a0ec DK |
2246 | if (err) |
2247 | goto journal_error; | |
2248 | brelse (bh2); | |
2249 | } else { | |
4776004f TT |
2250 | dxtrace(printk(KERN_DEBUG |
2251 | "Creating second level index...\n")); | |
ac27a0ec DK |
2252 | memcpy((char *) entries2, (char *) entries, |
2253 | icount * sizeof(struct dx_entry)); | |
2254 | dx_set_limit(entries2, dx_node_limit(dir)); | |
2255 | ||
2256 | /* Set up root */ | |
2257 | dx_set_count(entries, 1); | |
2258 | dx_set_block(entries + 0, newblock); | |
2259 | ((struct dx_root *) frames[0].bh->b_data)->info.indirect_levels = 1; | |
2260 | ||
2261 | /* Add new access path frame */ | |
2262 | frame = frames + 1; | |
2263 | frame->at = at = at - entries + entries2; | |
2264 | frame->entries = entries = entries2; | |
2265 | frame->bh = bh2; | |
617ba13b | 2266 | err = ext4_journal_get_write_access(handle, |
ac27a0ec DK |
2267 | frame->bh); |
2268 | if (err) | |
2269 | goto journal_error; | |
2270 | } | |
dbe89444 | 2271 | err = ext4_handle_dirty_dx_node(handle, dir, frames[0].bh); |
b4097142 TT |
2272 | if (err) { |
2273 | ext4_std_error(inode->i_sb, err); | |
2274 | goto cleanup; | |
2275 | } | |
ac27a0ec | 2276 | } |
5b643f9c | 2277 | de = do_split(handle, dir, &bh, frame, &fname->hinfo); |
f8b3b59d TT |
2278 | if (IS_ERR(de)) { |
2279 | err = PTR_ERR(de); | |
ac27a0ec | 2280 | goto cleanup; |
f8b3b59d | 2281 | } |
5b643f9c | 2282 | err = add_dirent_to_buf(handle, fname, dir, inode, de, bh); |
ac27a0ec DK |
2283 | goto cleanup; |
2284 | ||
2285 | journal_error: | |
617ba13b | 2286 | ext4_std_error(dir->i_sb, err); |
ac27a0ec | 2287 | cleanup: |
b1deefc9 | 2288 | brelse(bh); |
ac27a0ec DK |
2289 | dx_release(frames); |
2290 | return err; | |
2291 | } | |
ac27a0ec DK |
2292 | |
2293 | /* | |
05019a9e TM |
2294 | * ext4_generic_delete_entry deletes a directory entry by merging it |
2295 | * with the previous entry | |
ac27a0ec | 2296 | */ |
05019a9e TM |
2297 | int ext4_generic_delete_entry(handle_t *handle, |
2298 | struct inode *dir, | |
2299 | struct ext4_dir_entry_2 *de_del, | |
2300 | struct buffer_head *bh, | |
2301 | void *entry_buf, | |
2302 | int buf_size, | |
2303 | int csum_size) | |
ac27a0ec | 2304 | { |
af5bc92d | 2305 | struct ext4_dir_entry_2 *de, *pde; |
3d0518f4 | 2306 | unsigned int blocksize = dir->i_sb->s_blocksize; |
05019a9e | 2307 | int i; |
b0336e8d | 2308 | |
ac27a0ec DK |
2309 | i = 0; |
2310 | pde = NULL; | |
05019a9e TM |
2311 | de = (struct ext4_dir_entry_2 *)entry_buf; |
2312 | while (i < buf_size - csum_size) { | |
226ba972 TM |
2313 | if (ext4_check_dir_entry(dir, NULL, de, bh, |
2314 | bh->b_data, bh->b_size, i)) | |
6a797d27 | 2315 | return -EFSCORRUPTED; |
ac27a0ec | 2316 | if (de == de_del) { |
ac27a0ec | 2317 | if (pde) |
a72d7f83 | 2318 | pde->rec_len = ext4_rec_len_to_disk( |
3d0518f4 WY |
2319 | ext4_rec_len_from_disk(pde->rec_len, |
2320 | blocksize) + | |
2321 | ext4_rec_len_from_disk(de->rec_len, | |
2322 | blocksize), | |
2323 | blocksize); | |
ac27a0ec DK |
2324 | else |
2325 | de->inode = 0; | |
2326 | dir->i_version++; | |
ac27a0ec DK |
2327 | return 0; |
2328 | } | |
3d0518f4 | 2329 | i += ext4_rec_len_from_disk(de->rec_len, blocksize); |
ac27a0ec | 2330 | pde = de; |
3d0518f4 | 2331 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
2332 | } |
2333 | return -ENOENT; | |
2334 | } | |
2335 | ||
05019a9e TM |
2336 | static int ext4_delete_entry(handle_t *handle, |
2337 | struct inode *dir, | |
2338 | struct ext4_dir_entry_2 *de_del, | |
2339 | struct buffer_head *bh) | |
2340 | { | |
2341 | int err, csum_size = 0; | |
2342 | ||
9f40fe54 TM |
2343 | if (ext4_has_inline_data(dir)) { |
2344 | int has_inline_data = 1; | |
2345 | err = ext4_delete_inline_entry(handle, dir, de_del, bh, | |
2346 | &has_inline_data); | |
2347 | if (has_inline_data) | |
2348 | return err; | |
2349 | } | |
2350 | ||
9aa5d32b | 2351 | if (ext4_has_metadata_csum(dir->i_sb)) |
05019a9e TM |
2352 | csum_size = sizeof(struct ext4_dir_entry_tail); |
2353 | ||
2354 | BUFFER_TRACE(bh, "get_write_access"); | |
2355 | err = ext4_journal_get_write_access(handle, bh); | |
2356 | if (unlikely(err)) | |
2357 | goto out; | |
2358 | ||
2359 | err = ext4_generic_delete_entry(handle, dir, de_del, | |
2360 | bh, bh->b_data, | |
2361 | dir->i_sb->s_blocksize, csum_size); | |
2362 | if (err) | |
2363 | goto out; | |
2364 | ||
2365 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); | |
2366 | err = ext4_handle_dirty_dirent_node(handle, dir, bh); | |
2367 | if (unlikely(err)) | |
2368 | goto out; | |
2369 | ||
2370 | return 0; | |
2371 | out: | |
2372 | if (err != -ENOENT) | |
2373 | ext4_std_error(dir->i_sb, err); | |
2374 | return err; | |
2375 | } | |
2376 | ||
f8628a14 AD |
2377 | /* |
2378 | * DIR_NLINK feature is set if 1) nlinks > EXT4_LINK_MAX or 2) nlinks == 2, | |
2379 | * since this indicates that nlinks count was previously 1. | |
2380 | */ | |
2381 | static void ext4_inc_count(handle_t *handle, struct inode *inode) | |
2382 | { | |
2383 | inc_nlink(inode); | |
2384 | if (is_dx(inode) && inode->i_nlink > 1) { | |
2385 | /* limit is 16-bit i_links_count */ | |
2386 | if (inode->i_nlink >= EXT4_LINK_MAX || inode->i_nlink == 2) { | |
bfe86848 | 2387 | set_nlink(inode, 1); |
e2b911c5 | 2388 | ext4_set_feature_dir_nlink(inode->i_sb); |
f8628a14 AD |
2389 | } |
2390 | } | |
2391 | } | |
2392 | ||
2393 | /* | |
2394 | * If a directory had nlink == 1, then we should let it be 1. This indicates | |
2395 | * directory has >EXT4_LINK_MAX subdirs. | |
2396 | */ | |
2397 | static void ext4_dec_count(handle_t *handle, struct inode *inode) | |
2398 | { | |
909a4cf1 AD |
2399 | if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2) |
2400 | drop_nlink(inode); | |
f8628a14 AD |
2401 | } |
2402 | ||
2403 | ||
617ba13b | 2404 | static int ext4_add_nondir(handle_t *handle, |
ac27a0ec DK |
2405 | struct dentry *dentry, struct inode *inode) |
2406 | { | |
617ba13b | 2407 | int err = ext4_add_entry(handle, dentry, inode); |
ac27a0ec | 2408 | if (!err) { |
617ba13b | 2409 | ext4_mark_inode_dirty(handle, inode); |
6b38e842 | 2410 | unlock_new_inode(inode); |
8fc37ec5 | 2411 | d_instantiate(dentry, inode); |
ac27a0ec DK |
2412 | return 0; |
2413 | } | |
731b9a54 | 2414 | drop_nlink(inode); |
6b38e842 | 2415 | unlock_new_inode(inode); |
ac27a0ec DK |
2416 | iput(inode); |
2417 | return err; | |
2418 | } | |
2419 | ||
2420 | /* | |
2421 | * By the time this is called, we already have created | |
2422 | * the directory cache entry for the new file, but it | |
2423 | * is so far negative - it has no inode. | |
2424 | * | |
2425 | * If the create succeeds, we fill in the inode information | |
2426 | * with d_instantiate(). | |
2427 | */ | |
4acdaf27 | 2428 | static int ext4_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
ebfc3b49 | 2429 | bool excl) |
ac27a0ec DK |
2430 | { |
2431 | handle_t *handle; | |
af5bc92d | 2432 | struct inode *inode; |
1139575a | 2433 | int err, credits, retries = 0; |
ac27a0ec | 2434 | |
a7cdadee JK |
2435 | err = dquot_initialize(dir); |
2436 | if (err) | |
2437 | return err; | |
907f4554 | 2438 | |
1139575a | 2439 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2440 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2441 | retry: |
1139575a TT |
2442 | inode = ext4_new_inode_start_handle(dir, mode, &dentry->d_name, 0, |
2443 | NULL, EXT4_HT_DIR, credits); | |
2444 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2445 | err = PTR_ERR(inode); |
2446 | if (!IS_ERR(inode)) { | |
617ba13b | 2447 | inode->i_op = &ext4_file_inode_operations; |
be64f884 | 2448 | inode->i_fop = &ext4_file_operations; |
617ba13b | 2449 | ext4_set_aops(inode); |
e709e9df | 2450 | err = ext4_add_nondir(handle, dentry, inode); |
1139575a TT |
2451 | if (!err && IS_DIRSYNC(dir)) |
2452 | ext4_handle_sync(handle); | |
ac27a0ec | 2453 | } |
1139575a TT |
2454 | if (handle) |
2455 | ext4_journal_stop(handle); | |
617ba13b | 2456 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2457 | goto retry; |
2458 | return err; | |
2459 | } | |
2460 | ||
af5bc92d | 2461 | static int ext4_mknod(struct inode *dir, struct dentry *dentry, |
1a67aafb | 2462 | umode_t mode, dev_t rdev) |
ac27a0ec DK |
2463 | { |
2464 | handle_t *handle; | |
2465 | struct inode *inode; | |
1139575a | 2466 | int err, credits, retries = 0; |
ac27a0ec | 2467 | |
a7cdadee JK |
2468 | err = dquot_initialize(dir); |
2469 | if (err) | |
2470 | return err; | |
907f4554 | 2471 | |
1139575a | 2472 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2473 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2474 | retry: |
1139575a TT |
2475 | inode = ext4_new_inode_start_handle(dir, mode, &dentry->d_name, 0, |
2476 | NULL, EXT4_HT_DIR, credits); | |
2477 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2478 | err = PTR_ERR(inode); |
2479 | if (!IS_ERR(inode)) { | |
2480 | init_special_inode(inode, inode->i_mode, rdev); | |
617ba13b | 2481 | inode->i_op = &ext4_special_inode_operations; |
617ba13b | 2482 | err = ext4_add_nondir(handle, dentry, inode); |
1139575a TT |
2483 | if (!err && IS_DIRSYNC(dir)) |
2484 | ext4_handle_sync(handle); | |
ac27a0ec | 2485 | } |
1139575a TT |
2486 | if (handle) |
2487 | ext4_journal_stop(handle); | |
617ba13b | 2488 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2489 | goto retry; |
2490 | return err; | |
2491 | } | |
2492 | ||
af51a2ac AV |
2493 | static int ext4_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode) |
2494 | { | |
2495 | handle_t *handle; | |
2496 | struct inode *inode; | |
2497 | int err, retries = 0; | |
2498 | ||
a7cdadee JK |
2499 | err = dquot_initialize(dir); |
2500 | if (err) | |
2501 | return err; | |
af51a2ac AV |
2502 | |
2503 | retry: | |
2504 | inode = ext4_new_inode_start_handle(dir, mode, | |
2505 | NULL, 0, NULL, | |
2506 | EXT4_HT_DIR, | |
2507 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) + | |
2508 | 4 + EXT4_XATTR_TRANS_BLOCKS); | |
2509 | handle = ext4_journal_current_handle(); | |
2510 | err = PTR_ERR(inode); | |
2511 | if (!IS_ERR(inode)) { | |
2512 | inode->i_op = &ext4_file_inode_operations; | |
be64f884 | 2513 | inode->i_fop = &ext4_file_operations; |
af51a2ac | 2514 | ext4_set_aops(inode); |
e94bd349 | 2515 | d_tmpfile(dentry, inode); |
af51a2ac AV |
2516 | err = ext4_orphan_add(handle, inode); |
2517 | if (err) | |
43ae9e3f | 2518 | goto err_unlock_inode; |
af51a2ac | 2519 | mark_inode_dirty(inode); |
af51a2ac AV |
2520 | unlock_new_inode(inode); |
2521 | } | |
2522 | if (handle) | |
2523 | ext4_journal_stop(handle); | |
2524 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
2525 | goto retry; | |
2526 | return err; | |
43ae9e3f | 2527 | err_unlock_inode: |
af51a2ac AV |
2528 | ext4_journal_stop(handle); |
2529 | unlock_new_inode(inode); | |
af51a2ac AV |
2530 | return err; |
2531 | } | |
2532 | ||
a774f9c2 TM |
2533 | struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode, |
2534 | struct ext4_dir_entry_2 *de, | |
2535 | int blocksize, int csum_size, | |
2536 | unsigned int parent_ino, int dotdot_real_len) | |
2537 | { | |
2538 | de->inode = cpu_to_le32(inode->i_ino); | |
2539 | de->name_len = 1; | |
2540 | de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len), | |
2541 | blocksize); | |
2542 | strcpy(de->name, "."); | |
2543 | ext4_set_de_type(inode->i_sb, de, S_IFDIR); | |
2544 | ||
2545 | de = ext4_next_entry(de, blocksize); | |
2546 | de->inode = cpu_to_le32(parent_ino); | |
2547 | de->name_len = 2; | |
2548 | if (!dotdot_real_len) | |
2549 | de->rec_len = ext4_rec_len_to_disk(blocksize - | |
2550 | (csum_size + EXT4_DIR_REC_LEN(1)), | |
2551 | blocksize); | |
2552 | else | |
2553 | de->rec_len = ext4_rec_len_to_disk( | |
2554 | EXT4_DIR_REC_LEN(de->name_len), blocksize); | |
2555 | strcpy(de->name, ".."); | |
2556 | ext4_set_de_type(inode->i_sb, de, S_IFDIR); | |
2557 | ||
2558 | return ext4_next_entry(de, blocksize); | |
2559 | } | |
2560 | ||
2561 | static int ext4_init_new_dir(handle_t *handle, struct inode *dir, | |
2562 | struct inode *inode) | |
ac27a0ec | 2563 | { |
dabd991f | 2564 | struct buffer_head *dir_block = NULL; |
af5bc92d | 2565 | struct ext4_dir_entry_2 *de; |
b0336e8d | 2566 | struct ext4_dir_entry_tail *t; |
dc6982ff | 2567 | ext4_lblk_t block = 0; |
3d0518f4 | 2568 | unsigned int blocksize = dir->i_sb->s_blocksize; |
b0336e8d | 2569 | int csum_size = 0; |
a774f9c2 | 2570 | int err; |
ac27a0ec | 2571 | |
9aa5d32b | 2572 | if (ext4_has_metadata_csum(dir->i_sb)) |
b0336e8d DW |
2573 | csum_size = sizeof(struct ext4_dir_entry_tail); |
2574 | ||
3c47d541 TM |
2575 | if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) { |
2576 | err = ext4_try_create_inline_dir(handle, dir, inode); | |
2577 | if (err < 0 && err != -ENOSPC) | |
2578 | goto out; | |
2579 | if (!err) | |
2580 | goto out; | |
2581 | } | |
2582 | ||
dc6982ff | 2583 | inode->i_size = 0; |
0f70b406 TT |
2584 | dir_block = ext4_append(handle, inode, &block); |
2585 | if (IS_ERR(dir_block)) | |
2586 | return PTR_ERR(dir_block); | |
a774f9c2 TM |
2587 | de = (struct ext4_dir_entry_2 *)dir_block->b_data; |
2588 | ext4_init_dot_dotdot(inode, de, blocksize, csum_size, dir->i_ino, 0); | |
2589 | set_nlink(inode, 2); | |
2590 | if (csum_size) { | |
2591 | t = EXT4_DIRENT_TAIL(dir_block->b_data, blocksize); | |
2592 | initialize_dirent_tail(t, blocksize); | |
2593 | } | |
2594 | ||
2595 | BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata"); | |
2596 | err = ext4_handle_dirty_dirent_node(handle, inode, dir_block); | |
2597 | if (err) | |
2598 | goto out; | |
2599 | set_buffer_verified(dir_block); | |
2600 | out: | |
2601 | brelse(dir_block); | |
2602 | return err; | |
2603 | } | |
2604 | ||
2605 | static int ext4_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) | |
2606 | { | |
2607 | handle_t *handle; | |
2608 | struct inode *inode; | |
1139575a | 2609 | int err, credits, retries = 0; |
a774f9c2 | 2610 | |
f8628a14 | 2611 | if (EXT4_DIR_LINK_MAX(dir)) |
ac27a0ec DK |
2612 | return -EMLINK; |
2613 | ||
a7cdadee JK |
2614 | err = dquot_initialize(dir); |
2615 | if (err) | |
2616 | return err; | |
907f4554 | 2617 | |
1139575a | 2618 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2619 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2620 | retry: |
1139575a TT |
2621 | inode = ext4_new_inode_start_handle(dir, S_IFDIR | mode, |
2622 | &dentry->d_name, | |
2623 | 0, NULL, EXT4_HT_DIR, credits); | |
2624 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2625 | err = PTR_ERR(inode); |
2626 | if (IS_ERR(inode)) | |
2627 | goto out_stop; | |
2628 | ||
617ba13b MC |
2629 | inode->i_op = &ext4_dir_inode_operations; |
2630 | inode->i_fop = &ext4_dir_operations; | |
a774f9c2 | 2631 | err = ext4_init_new_dir(handle, dir, inode); |
dabd991f NK |
2632 | if (err) |
2633 | goto out_clear_inode; | |
2634 | err = ext4_mark_inode_dirty(handle, inode); | |
2635 | if (!err) | |
2636 | err = ext4_add_entry(handle, dentry, inode); | |
ac27a0ec | 2637 | if (err) { |
4cdeed86 AK |
2638 | out_clear_inode: |
2639 | clear_nlink(inode); | |
6b38e842 | 2640 | unlock_new_inode(inode); |
617ba13b | 2641 | ext4_mark_inode_dirty(handle, inode); |
af5bc92d | 2642 | iput(inode); |
ac27a0ec DK |
2643 | goto out_stop; |
2644 | } | |
f8628a14 | 2645 | ext4_inc_count(handle, dir); |
617ba13b | 2646 | ext4_update_dx_flag(dir); |
dabd991f NK |
2647 | err = ext4_mark_inode_dirty(handle, dir); |
2648 | if (err) | |
2649 | goto out_clear_inode; | |
6b38e842 | 2650 | unlock_new_inode(inode); |
8fc37ec5 | 2651 | d_instantiate(dentry, inode); |
1139575a TT |
2652 | if (IS_DIRSYNC(dir)) |
2653 | ext4_handle_sync(handle); | |
2654 | ||
ac27a0ec | 2655 | out_stop: |
1139575a TT |
2656 | if (handle) |
2657 | ext4_journal_stop(handle); | |
617ba13b | 2658 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2659 | goto retry; |
2660 | return err; | |
2661 | } | |
2662 | ||
2663 | /* | |
2664 | * routine to check that the specified directory is empty (for rmdir) | |
2665 | */ | |
e875a2dd | 2666 | int ext4_empty_dir(struct inode *inode) |
ac27a0ec | 2667 | { |
498e5f24 | 2668 | unsigned int offset; |
af5bc92d TT |
2669 | struct buffer_head *bh; |
2670 | struct ext4_dir_entry_2 *de, *de1; | |
2671 | struct super_block *sb; | |
ac27a0ec DK |
2672 | int err = 0; |
2673 | ||
61f86638 TM |
2674 | if (ext4_has_inline_data(inode)) { |
2675 | int has_inline_data = 1; | |
2676 | ||
2677 | err = empty_inline_dir(inode, &has_inline_data); | |
2678 | if (has_inline_data) | |
2679 | return err; | |
2680 | } | |
2681 | ||
ac27a0ec | 2682 | sb = inode->i_sb; |
dc6982ff TT |
2683 | if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2)) { |
2684 | EXT4_ERROR_INODE(inode, "invalid size"); | |
ac27a0ec DK |
2685 | return 1; |
2686 | } | |
dc6982ff TT |
2687 | bh = ext4_read_dirblock(inode, 0, EITHER); |
2688 | if (IS_ERR(bh)) | |
2689 | return 1; | |
2690 | ||
617ba13b | 2691 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
3d0518f4 | 2692 | de1 = ext4_next_entry(de, sb->s_blocksize); |
ac27a0ec | 2693 | if (le32_to_cpu(de->inode) != inode->i_ino || |
b03a2f7e AD |
2694 | le32_to_cpu(de1->inode) == 0 || |
2695 | strcmp(".", de->name) || strcmp("..", de1->name)) { | |
2696 | ext4_warning_inode(inode, "directory missing '.' and/or '..'"); | |
af5bc92d | 2697 | brelse(bh); |
ac27a0ec DK |
2698 | return 1; |
2699 | } | |
3d0518f4 WY |
2700 | offset = ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) + |
2701 | ext4_rec_len_from_disk(de1->rec_len, sb->s_blocksize); | |
2702 | de = ext4_next_entry(de1, sb->s_blocksize); | |
af5bc92d | 2703 | while (offset < inode->i_size) { |
236f5ecb | 2704 | if ((void *) de >= (void *) (bh->b_data+sb->s_blocksize)) { |
24676da4 | 2705 | unsigned int lblock; |
ac27a0ec | 2706 | err = 0; |
af5bc92d | 2707 | brelse(bh); |
24676da4 | 2708 | lblock = offset >> EXT4_BLOCK_SIZE_BITS(sb); |
dc6982ff TT |
2709 | bh = ext4_read_dirblock(inode, lblock, EITHER); |
2710 | if (IS_ERR(bh)) | |
2711 | return 1; | |
617ba13b | 2712 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 2713 | } |
226ba972 TM |
2714 | if (ext4_check_dir_entry(inode, NULL, de, bh, |
2715 | bh->b_data, bh->b_size, offset)) { | |
617ba13b | 2716 | de = (struct ext4_dir_entry_2 *)(bh->b_data + |
ac27a0ec DK |
2717 | sb->s_blocksize); |
2718 | offset = (offset | (sb->s_blocksize - 1)) + 1; | |
2719 | continue; | |
2720 | } | |
2721 | if (le32_to_cpu(de->inode)) { | |
af5bc92d | 2722 | brelse(bh); |
ac27a0ec DK |
2723 | return 0; |
2724 | } | |
3d0518f4 WY |
2725 | offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize); |
2726 | de = ext4_next_entry(de, sb->s_blocksize); | |
ac27a0ec | 2727 | } |
af5bc92d | 2728 | brelse(bh); |
ac27a0ec DK |
2729 | return 1; |
2730 | } | |
2731 | ||
d745a8c2 JK |
2732 | /* |
2733 | * ext4_orphan_add() links an unlinked or truncated inode into a list of | |
ac27a0ec DK |
2734 | * such inodes, starting at the superblock, in case we crash before the |
2735 | * file is closed/deleted, or in case the inode truncate spans multiple | |
2736 | * transactions and the last transaction is not recovered after a crash. | |
2737 | * | |
2738 | * At filesystem recovery time, we walk this list deleting unlinked | |
617ba13b | 2739 | * inodes and truncating linked inodes in ext4_orphan_cleanup(). |
d745a8c2 JK |
2740 | * |
2741 | * Orphan list manipulation functions must be called under i_mutex unless | |
2742 | * we are just creating the inode or deleting it. | |
ac27a0ec | 2743 | */ |
617ba13b | 2744 | int ext4_orphan_add(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2745 | { |
2746 | struct super_block *sb = inode->i_sb; | |
cd2c080c | 2747 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
617ba13b | 2748 | struct ext4_iloc iloc; |
ac27a0ec | 2749 | int err = 0, rc; |
d745a8c2 | 2750 | bool dirty = false; |
ac27a0ec | 2751 | |
e2bfb088 | 2752 | if (!sbi->s_journal || is_bad_inode(inode)) |
0390131b FM |
2753 | return 0; |
2754 | ||
d745a8c2 JK |
2755 | WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) && |
2756 | !mutex_is_locked(&inode->i_mutex)); | |
2757 | /* | |
2758 | * Exit early if inode already is on orphan list. This is a big speedup | |
2759 | * since we don't have to contend on the global s_orphan_lock. | |
2760 | */ | |
617ba13b | 2761 | if (!list_empty(&EXT4_I(inode)->i_orphan)) |
d745a8c2 | 2762 | return 0; |
ac27a0ec | 2763 | |
afb86178 LC |
2764 | /* |
2765 | * Orphan handling is only valid for files with data blocks | |
2766 | * being truncated, or files being unlinked. Note that we either | |
2767 | * hold i_mutex, or the inode can not be referenced from outside, | |
2768 | * so i_nlink should not be bumped due to race | |
ac27a0ec | 2769 | */ |
af5bc92d TT |
2770 | J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || |
2771 | S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); | |
ac27a0ec | 2772 | |
cd2c080c JK |
2773 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
2774 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); | |
ac27a0ec | 2775 | if (err) |
d745a8c2 | 2776 | goto out; |
ac27a0ec | 2777 | |
617ba13b | 2778 | err = ext4_reserve_inode_write(handle, inode, &iloc); |
ac27a0ec | 2779 | if (err) |
d745a8c2 JK |
2780 | goto out; |
2781 | ||
2782 | mutex_lock(&sbi->s_orphan_lock); | |
6e3617e5 DM |
2783 | /* |
2784 | * Due to previous errors inode may be already a part of on-disk | |
2785 | * orphan list. If so skip on-disk list modification. | |
2786 | */ | |
d745a8c2 JK |
2787 | if (!NEXT_ORPHAN(inode) || NEXT_ORPHAN(inode) > |
2788 | (le32_to_cpu(sbi->s_es->s_inodes_count))) { | |
2789 | /* Insert this inode at the head of the on-disk orphan list */ | |
2790 | NEXT_ORPHAN(inode) = le32_to_cpu(sbi->s_es->s_last_orphan); | |
2791 | sbi->s_es->s_last_orphan = cpu_to_le32(inode->i_ino); | |
2792 | dirty = true; | |
2793 | } | |
2794 | list_add(&EXT4_I(inode)->i_orphan, &sbi->s_orphan); | |
2795 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2796 | |
d745a8c2 JK |
2797 | if (dirty) { |
2798 | err = ext4_handle_dirty_super(handle, sb); | |
2799 | rc = ext4_mark_iloc_dirty(handle, inode, &iloc); | |
2800 | if (!err) | |
2801 | err = rc; | |
2802 | if (err) { | |
2803 | /* | |
2804 | * We have to remove inode from in-memory list if | |
2805 | * addition to on disk orphan list failed. Stray orphan | |
2806 | * list entries can cause panics at unmount time. | |
2807 | */ | |
2808 | mutex_lock(&sbi->s_orphan_lock); | |
2809 | list_del(&EXT4_I(inode)->i_orphan); | |
2810 | mutex_unlock(&sbi->s_orphan_lock); | |
2811 | } | |
2812 | } | |
ac27a0ec DK |
2813 | jbd_debug(4, "superblock will point to %lu\n", inode->i_ino); |
2814 | jbd_debug(4, "orphan inode %lu will point to %d\n", | |
2815 | inode->i_ino, NEXT_ORPHAN(inode)); | |
d745a8c2 | 2816 | out: |
cd2c080c | 2817 | ext4_std_error(sb, err); |
ac27a0ec DK |
2818 | return err; |
2819 | } | |
2820 | ||
2821 | /* | |
617ba13b | 2822 | * ext4_orphan_del() removes an unlinked or truncated inode from the list |
ac27a0ec DK |
2823 | * of such inodes stored on disk, because it is finally being cleaned up. |
2824 | */ | |
617ba13b | 2825 | int ext4_orphan_del(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2826 | { |
2827 | struct list_head *prev; | |
617ba13b | 2828 | struct ext4_inode_info *ei = EXT4_I(inode); |
cd2c080c | 2829 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
498e5f24 | 2830 | __u32 ino_next; |
617ba13b | 2831 | struct ext4_iloc iloc; |
ac27a0ec DK |
2832 | int err = 0; |
2833 | ||
cd2c080c | 2834 | if (!sbi->s_journal && !(sbi->s_mount_state & EXT4_ORPHAN_FS)) |
0390131b FM |
2835 | return 0; |
2836 | ||
d745a8c2 JK |
2837 | WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) && |
2838 | !mutex_is_locked(&inode->i_mutex)); | |
2839 | /* Do this quick check before taking global s_orphan_lock. */ | |
3b9d4ed2 | 2840 | if (list_empty(&ei->i_orphan)) |
d745a8c2 | 2841 | return 0; |
ac27a0ec | 2842 | |
d745a8c2 JK |
2843 | if (handle) { |
2844 | /* Grab inode buffer early before taking global s_orphan_lock */ | |
2845 | err = ext4_reserve_inode_write(handle, inode, &iloc); | |
2846 | } | |
ac27a0ec | 2847 | |
d745a8c2 | 2848 | mutex_lock(&sbi->s_orphan_lock); |
ac27a0ec DK |
2849 | jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); |
2850 | ||
d745a8c2 | 2851 | prev = ei->i_orphan.prev; |
ac27a0ec DK |
2852 | list_del_init(&ei->i_orphan); |
2853 | ||
2854 | /* If we're on an error path, we may not have a valid | |
2855 | * transaction handle with which to update the orphan list on | |
2856 | * disk, but we still need to remove the inode from the linked | |
2857 | * list in memory. */ | |
d745a8c2 JK |
2858 | if (!handle || err) { |
2859 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2860 | goto out_err; |
d745a8c2 | 2861 | } |
ac27a0ec | 2862 | |
d745a8c2 | 2863 | ino_next = NEXT_ORPHAN(inode); |
ac27a0ec | 2864 | if (prev == &sbi->s_orphan) { |
498e5f24 | 2865 | jbd_debug(4, "superblock will point to %u\n", ino_next); |
ac27a0ec | 2866 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
617ba13b | 2867 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
d745a8c2 JK |
2868 | if (err) { |
2869 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2870 | goto out_brelse; |
d745a8c2 | 2871 | } |
ac27a0ec | 2872 | sbi->s_es->s_last_orphan = cpu_to_le32(ino_next); |
d745a8c2 | 2873 | mutex_unlock(&sbi->s_orphan_lock); |
b50924c2 | 2874 | err = ext4_handle_dirty_super(handle, inode->i_sb); |
ac27a0ec | 2875 | } else { |
617ba13b | 2876 | struct ext4_iloc iloc2; |
ac27a0ec | 2877 | struct inode *i_prev = |
617ba13b | 2878 | &list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec | 2879 | |
498e5f24 | 2880 | jbd_debug(4, "orphan inode %lu will point to %u\n", |
ac27a0ec | 2881 | i_prev->i_ino, ino_next); |
617ba13b | 2882 | err = ext4_reserve_inode_write(handle, i_prev, &iloc2); |
d745a8c2 JK |
2883 | if (err) { |
2884 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2885 | goto out_brelse; |
d745a8c2 | 2886 | } |
ac27a0ec | 2887 | NEXT_ORPHAN(i_prev) = ino_next; |
617ba13b | 2888 | err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2); |
d745a8c2 | 2889 | mutex_unlock(&sbi->s_orphan_lock); |
ac27a0ec DK |
2890 | } |
2891 | if (err) | |
2892 | goto out_brelse; | |
2893 | NEXT_ORPHAN(inode) = 0; | |
617ba13b | 2894 | err = ext4_mark_iloc_dirty(handle, inode, &iloc); |
ac27a0ec | 2895 | out_err: |
617ba13b | 2896 | ext4_std_error(inode->i_sb, err); |
ac27a0ec DK |
2897 | return err; |
2898 | ||
2899 | out_brelse: | |
2900 | brelse(iloc.bh); | |
2901 | goto out_err; | |
2902 | } | |
2903 | ||
af5bc92d | 2904 | static int ext4_rmdir(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2905 | { |
2906 | int retval; | |
af5bc92d TT |
2907 | struct inode *inode; |
2908 | struct buffer_head *bh; | |
2909 | struct ext4_dir_entry_2 *de; | |
8dcfaad2 | 2910 | handle_t *handle = NULL; |
ac27a0ec DK |
2911 | |
2912 | /* Initialize quotas before so that eventual writes go in | |
2913 | * separate transaction */ | |
a7cdadee JK |
2914 | retval = dquot_initialize(dir); |
2915 | if (retval) | |
2916 | return retval; | |
2917 | retval = dquot_initialize(d_inode(dentry)); | |
2918 | if (retval) | |
2919 | return retval; | |
907f4554 | 2920 | |
ac27a0ec | 2921 | retval = -ENOENT; |
32f7f22c | 2922 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 TT |
2923 | if (IS_ERR(bh)) |
2924 | return PTR_ERR(bh); | |
ac27a0ec DK |
2925 | if (!bh) |
2926 | goto end_rmdir; | |
2927 | ||
2b0143b5 | 2928 | inode = d_inode(dentry); |
ac27a0ec | 2929 | |
6a797d27 | 2930 | retval = -EFSCORRUPTED; |
ac27a0ec DK |
2931 | if (le32_to_cpu(de->inode) != inode->i_ino) |
2932 | goto end_rmdir; | |
2933 | ||
2934 | retval = -ENOTEMPTY; | |
e875a2dd | 2935 | if (!ext4_empty_dir(inode)) |
ac27a0ec DK |
2936 | goto end_rmdir; |
2937 | ||
8dcfaad2 | 2938 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
64044abf | 2939 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb)); |
8dcfaad2 TT |
2940 | if (IS_ERR(handle)) { |
2941 | retval = PTR_ERR(handle); | |
2942 | handle = NULL; | |
2943 | goto end_rmdir; | |
2944 | } | |
2945 | ||
2946 | if (IS_DIRSYNC(dir)) | |
2947 | ext4_handle_sync(handle); | |
2948 | ||
617ba13b | 2949 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
2950 | if (retval) |
2951 | goto end_rmdir; | |
f8628a14 | 2952 | if (!EXT4_DIR_LINK_EMPTY(inode)) |
b03a2f7e AD |
2953 | ext4_warning_inode(inode, |
2954 | "empty directory '%.*s' has too many links (%u)", | |
2955 | dentry->d_name.len, dentry->d_name.name, | |
af5bc92d | 2956 | inode->i_nlink); |
ac27a0ec DK |
2957 | inode->i_version++; |
2958 | clear_nlink(inode); | |
2959 | /* There's no need to set i_disksize: the fact that i_nlink is | |
2960 | * zero will ensure that the right thing happens during any | |
2961 | * recovery. */ | |
2962 | inode->i_size = 0; | |
617ba13b | 2963 | ext4_orphan_add(handle, inode); |
ef7f3835 | 2964 | inode->i_ctime = dir->i_ctime = dir->i_mtime = ext4_current_time(inode); |
617ba13b | 2965 | ext4_mark_inode_dirty(handle, inode); |
f8628a14 | 2966 | ext4_dec_count(handle, dir); |
617ba13b MC |
2967 | ext4_update_dx_flag(dir); |
2968 | ext4_mark_inode_dirty(handle, dir); | |
ac27a0ec DK |
2969 | |
2970 | end_rmdir: | |
af5bc92d | 2971 | brelse(bh); |
8dcfaad2 TT |
2972 | if (handle) |
2973 | ext4_journal_stop(handle); | |
ac27a0ec DK |
2974 | return retval; |
2975 | } | |
2976 | ||
af5bc92d | 2977 | static int ext4_unlink(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2978 | { |
2979 | int retval; | |
af5bc92d TT |
2980 | struct inode *inode; |
2981 | struct buffer_head *bh; | |
2982 | struct ext4_dir_entry_2 *de; | |
931b6864 | 2983 | handle_t *handle = NULL; |
ac27a0ec | 2984 | |
0562e0ba | 2985 | trace_ext4_unlink_enter(dir, dentry); |
ac27a0ec DK |
2986 | /* Initialize quotas before so that eventual writes go |
2987 | * in separate transaction */ | |
a7cdadee JK |
2988 | retval = dquot_initialize(dir); |
2989 | if (retval) | |
2990 | return retval; | |
2991 | retval = dquot_initialize(d_inode(dentry)); | |
2992 | if (retval) | |
2993 | return retval; | |
907f4554 | 2994 | |
ac27a0ec | 2995 | retval = -ENOENT; |
32f7f22c | 2996 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 TT |
2997 | if (IS_ERR(bh)) |
2998 | return PTR_ERR(bh); | |
ac27a0ec DK |
2999 | if (!bh) |
3000 | goto end_unlink; | |
3001 | ||
2b0143b5 | 3002 | inode = d_inode(dentry); |
ac27a0ec | 3003 | |
6a797d27 | 3004 | retval = -EFSCORRUPTED; |
ac27a0ec DK |
3005 | if (le32_to_cpu(de->inode) != inode->i_ino) |
3006 | goto end_unlink; | |
3007 | ||
931b6864 | 3008 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
64044abf | 3009 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb)); |
931b6864 TT |
3010 | if (IS_ERR(handle)) { |
3011 | retval = PTR_ERR(handle); | |
3012 | handle = NULL; | |
3013 | goto end_unlink; | |
3014 | } | |
3015 | ||
3016 | if (IS_DIRSYNC(dir)) | |
3017 | ext4_handle_sync(handle); | |
3018 | ||
b03a2f7e AD |
3019 | if (inode->i_nlink == 0) { |
3020 | ext4_warning_inode(inode, "Deleting file '%.*s' with no links", | |
3021 | dentry->d_name.len, dentry->d_name.name); | |
bfe86848 | 3022 | set_nlink(inode, 1); |
ac27a0ec | 3023 | } |
617ba13b | 3024 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
3025 | if (retval) |
3026 | goto end_unlink; | |
ef7f3835 | 3027 | dir->i_ctime = dir->i_mtime = ext4_current_time(dir); |
617ba13b MC |
3028 | ext4_update_dx_flag(dir); |
3029 | ext4_mark_inode_dirty(handle, dir); | |
825f1481 | 3030 | drop_nlink(inode); |
ac27a0ec | 3031 | if (!inode->i_nlink) |
617ba13b | 3032 | ext4_orphan_add(handle, inode); |
ef7f3835 | 3033 | inode->i_ctime = ext4_current_time(inode); |
617ba13b | 3034 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec DK |
3035 | |
3036 | end_unlink: | |
af5bc92d | 3037 | brelse(bh); |
931b6864 TT |
3038 | if (handle) |
3039 | ext4_journal_stop(handle); | |
0562e0ba | 3040 | trace_ext4_unlink_exit(dentry, retval); |
ac27a0ec DK |
3041 | return retval; |
3042 | } | |
3043 | ||
af5bc92d TT |
3044 | static int ext4_symlink(struct inode *dir, |
3045 | struct dentry *dentry, const char *symname) | |
ac27a0ec DK |
3046 | { |
3047 | handle_t *handle; | |
af5bc92d | 3048 | struct inode *inode; |
f348c252 | 3049 | int err, len = strlen(symname); |
df5e6223 | 3050 | int credits; |
f348c252 TT |
3051 | bool encryption_required; |
3052 | struct ext4_str disk_link; | |
3053 | struct ext4_encrypted_symlink_data *sd = NULL; | |
ac27a0ec | 3054 | |
f348c252 TT |
3055 | disk_link.len = len + 1; |
3056 | disk_link.name = (char *) symname; | |
3057 | ||
6ddb2447 TT |
3058 | encryption_required = (ext4_encrypted_inode(dir) || |
3059 | DUMMY_ENCRYPTION_ENABLED(EXT4_SB(dir->i_sb))); | |
4d3c4e5b TT |
3060 | if (encryption_required) { |
3061 | err = ext4_get_encryption_info(dir); | |
3062 | if (err) | |
3063 | return err; | |
3064 | if (ext4_encryption_info(dir) == NULL) | |
3065 | return -EPERM; | |
3066 | disk_link.len = (ext4_fname_encrypted_size(dir, len) + | |
3067 | sizeof(struct ext4_encrypted_symlink_data)); | |
3068 | sd = kzalloc(disk_link.len, GFP_KERNEL); | |
3069 | if (!sd) | |
3070 | return -ENOMEM; | |
3071 | } | |
3072 | ||
3073 | if (disk_link.len > dir->i_sb->s_blocksize) { | |
3074 | err = -ENAMETOOLONG; | |
3075 | goto err_free_sd; | |
3076 | } | |
ac27a0ec | 3077 | |
a7cdadee JK |
3078 | err = dquot_initialize(dir); |
3079 | if (err) | |
926631c2 | 3080 | goto err_free_sd; |
907f4554 | 3081 | |
f348c252 | 3082 | if ((disk_link.len > EXT4_N_BLOCKS * 4)) { |
df5e6223 JK |
3083 | /* |
3084 | * For non-fast symlinks, we just allocate inode and put it on | |
3085 | * orphan list in the first transaction => we need bitmap, | |
8c208719 ES |
3086 | * group descriptor, sb, inode block, quota blocks, and |
3087 | * possibly selinux xattr blocks. | |
df5e6223 | 3088 | */ |
8c208719 ES |
3089 | credits = 4 + EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) + |
3090 | EXT4_XATTR_TRANS_BLOCKS; | |
df5e6223 JK |
3091 | } else { |
3092 | /* | |
3093 | * Fast symlink. We have to add entry to directory | |
3094 | * (EXT4_DATA_TRANS_BLOCKS + EXT4_INDEX_EXTRA_TRANS_BLOCKS), | |
3095 | * allocate new inode (bitmap, group descriptor, inode block, | |
3096 | * quota blocks, sb is already counted in previous macros). | |
3097 | */ | |
3098 | credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
eb9cc7e1 | 3099 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3; |
df5e6223 | 3100 | } |
f348c252 | 3101 | |
1139575a TT |
3102 | inode = ext4_new_inode_start_handle(dir, S_IFLNK|S_IRWXUGO, |
3103 | &dentry->d_name, 0, NULL, | |
3104 | EXT4_HT_DIR, credits); | |
3105 | handle = ext4_journal_current_handle(); | |
f348c252 TT |
3106 | if (IS_ERR(inode)) { |
3107 | if (handle) | |
3108 | ext4_journal_stop(handle); | |
4d3c4e5b TT |
3109 | err = PTR_ERR(inode); |
3110 | goto err_free_sd; | |
f348c252 TT |
3111 | } |
3112 | ||
3113 | if (encryption_required) { | |
f348c252 TT |
3114 | struct qstr istr; |
3115 | struct ext4_str ostr; | |
3116 | ||
f348c252 TT |
3117 | istr.name = (const unsigned char *) symname; |
3118 | istr.len = len; | |
3119 | ostr.name = sd->encrypted_path; | |
d2299590 | 3120 | ostr.len = disk_link.len; |
b7236e21 | 3121 | err = ext4_fname_usr_to_disk(inode, &istr, &ostr); |
f348c252 TT |
3122 | if (err < 0) |
3123 | goto err_drop_inode; | |
3124 | sd->len = cpu_to_le16(ostr.len); | |
3125 | disk_link.name = (char *) sd; | |
a7a67e8a | 3126 | inode->i_op = &ext4_encrypted_symlink_inode_operations; |
f348c252 | 3127 | } |
ac27a0ec | 3128 | |
f348c252 | 3129 | if ((disk_link.len > EXT4_N_BLOCKS * 4)) { |
a7a67e8a AV |
3130 | if (!encryption_required) |
3131 | inode->i_op = &ext4_symlink_inode_operations; | |
617ba13b | 3132 | ext4_set_aops(inode); |
ac27a0ec | 3133 | /* |
df5e6223 JK |
3134 | * We cannot call page_symlink() with transaction started |
3135 | * because it calls into ext4_write_begin() which can wait | |
3136 | * for transaction commit if we are running out of space | |
3137 | * and thus we deadlock. So we have to stop transaction now | |
3138 | * and restart it when symlink contents is written. | |
3139 | * | |
3140 | * To keep fs consistent in case of crash, we have to put inode | |
3141 | * to orphan list in the mean time. | |
ac27a0ec | 3142 | */ |
df5e6223 JK |
3143 | drop_nlink(inode); |
3144 | err = ext4_orphan_add(handle, inode); | |
3145 | ext4_journal_stop(handle); | |
f348c252 | 3146 | handle = NULL; |
df5e6223 JK |
3147 | if (err) |
3148 | goto err_drop_inode; | |
f348c252 | 3149 | err = __page_symlink(inode, disk_link.name, disk_link.len, 1); |
df5e6223 JK |
3150 | if (err) |
3151 | goto err_drop_inode; | |
3152 | /* | |
3153 | * Now inode is being linked into dir (EXT4_DATA_TRANS_BLOCKS | |
3154 | * + EXT4_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified | |
3155 | */ | |
9924a92a | 3156 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
df5e6223 JK |
3157 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
3158 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 1); | |
3159 | if (IS_ERR(handle)) { | |
3160 | err = PTR_ERR(handle); | |
f348c252 | 3161 | handle = NULL; |
df5e6223 JK |
3162 | goto err_drop_inode; |
3163 | } | |
0ce8c010 | 3164 | set_nlink(inode, 1); |
df5e6223 | 3165 | err = ext4_orphan_del(handle, inode); |
f348c252 | 3166 | if (err) |
df5e6223 | 3167 | goto err_drop_inode; |
ac27a0ec | 3168 | } else { |
e65187e6 | 3169 | /* clear the extent format for fast symlink */ |
12e9b892 | 3170 | ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS); |
75e7566b | 3171 | if (!encryption_required) { |
a7a67e8a | 3172 | inode->i_op = &ext4_fast_symlink_inode_operations; |
75e7566b AV |
3173 | inode->i_link = (char *)&EXT4_I(inode)->i_data; |
3174 | } | |
f348c252 TT |
3175 | memcpy((char *)&EXT4_I(inode)->i_data, disk_link.name, |
3176 | disk_link.len); | |
3177 | inode->i_size = disk_link.len - 1; | |
ac27a0ec | 3178 | } |
617ba13b MC |
3179 | EXT4_I(inode)->i_disksize = inode->i_size; |
3180 | err = ext4_add_nondir(handle, dentry, inode); | |
1139575a TT |
3181 | if (!err && IS_DIRSYNC(dir)) |
3182 | ext4_handle_sync(handle); | |
3183 | ||
1139575a TT |
3184 | if (handle) |
3185 | ext4_journal_stop(handle); | |
f348c252 | 3186 | kfree(sd); |
ac27a0ec | 3187 | return err; |
df5e6223 | 3188 | err_drop_inode: |
f348c252 TT |
3189 | if (handle) |
3190 | ext4_journal_stop(handle); | |
f348c252 | 3191 | clear_nlink(inode); |
df5e6223 JK |
3192 | unlock_new_inode(inode); |
3193 | iput(inode); | |
4d3c4e5b TT |
3194 | err_free_sd: |
3195 | kfree(sd); | |
df5e6223 | 3196 | return err; |
ac27a0ec DK |
3197 | } |
3198 | ||
af5bc92d TT |
3199 | static int ext4_link(struct dentry *old_dentry, |
3200 | struct inode *dir, struct dentry *dentry) | |
ac27a0ec DK |
3201 | { |
3202 | handle_t *handle; | |
2b0143b5 | 3203 | struct inode *inode = d_inode(old_dentry); |
ac27a0ec DK |
3204 | int err, retries = 0; |
3205 | ||
b05ab1dc | 3206 | if (inode->i_nlink >= EXT4_LINK_MAX) |
ac27a0ec | 3207 | return -EMLINK; |
d9cdc903 TT |
3208 | if (ext4_encrypted_inode(dir) && |
3209 | !ext4_is_child_context_consistent_with_parent(dir, inode)) | |
3210 | return -EPERM; | |
a7cdadee JK |
3211 | err = dquot_initialize(dir); |
3212 | if (err) | |
3213 | return err; | |
907f4554 | 3214 | |
ac27a0ec | 3215 | retry: |
9924a92a TT |
3216 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
3217 | (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
af51a2ac | 3218 | EXT4_INDEX_EXTRA_TRANS_BLOCKS) + 1); |
ac27a0ec DK |
3219 | if (IS_ERR(handle)) |
3220 | return PTR_ERR(handle); | |
3221 | ||
3222 | if (IS_DIRSYNC(dir)) | |
0390131b | 3223 | ext4_handle_sync(handle); |
ac27a0ec | 3224 | |
ef7f3835 | 3225 | inode->i_ctime = ext4_current_time(inode); |
f8628a14 | 3226 | ext4_inc_count(handle, inode); |
7de9c6ee | 3227 | ihold(inode); |
ac27a0ec | 3228 | |
6b38e842 AV |
3229 | err = ext4_add_entry(handle, dentry, inode); |
3230 | if (!err) { | |
3231 | ext4_mark_inode_dirty(handle, inode); | |
af51a2ac AV |
3232 | /* this can happen only for tmpfile being |
3233 | * linked the first time | |
3234 | */ | |
3235 | if (inode->i_nlink == 1) | |
3236 | ext4_orphan_del(handle, inode); | |
6b38e842 AV |
3237 | d_instantiate(dentry, inode); |
3238 | } else { | |
3239 | drop_nlink(inode); | |
3240 | iput(inode); | |
3241 | } | |
617ba13b MC |
3242 | ext4_journal_stop(handle); |
3243 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
3244 | goto retry; |
3245 | return err; | |
3246 | } | |
3247 | ||
32f7f22c TM |
3248 | |
3249 | /* | |
3250 | * Try to find buffer head where contains the parent block. | |
3251 | * It should be the inode block if it is inlined or the 1st block | |
3252 | * if it is a normal dir. | |
3253 | */ | |
3254 | static struct buffer_head *ext4_get_first_dir_block(handle_t *handle, | |
3255 | struct inode *inode, | |
3256 | int *retval, | |
3257 | struct ext4_dir_entry_2 **parent_de, | |
3258 | int *inlined) | |
3259 | { | |
3260 | struct buffer_head *bh; | |
3261 | ||
3262 | if (!ext4_has_inline_data(inode)) { | |
dc6982ff TT |
3263 | bh = ext4_read_dirblock(inode, 0, EITHER); |
3264 | if (IS_ERR(bh)) { | |
3265 | *retval = PTR_ERR(bh); | |
32f7f22c TM |
3266 | return NULL; |
3267 | } | |
3268 | *parent_de = ext4_next_entry( | |
3269 | (struct ext4_dir_entry_2 *)bh->b_data, | |
3270 | inode->i_sb->s_blocksize); | |
3271 | return bh; | |
3272 | } | |
3273 | ||
3274 | *inlined = 1; | |
3275 | return ext4_get_first_inline_block(inode, parent_de, retval); | |
3276 | } | |
ac27a0ec | 3277 | |
c0d268c3 MS |
3278 | struct ext4_renament { |
3279 | struct inode *dir; | |
3280 | struct dentry *dentry; | |
3281 | struct inode *inode; | |
bd42998a MS |
3282 | bool is_dir; |
3283 | int dir_nlink_delta; | |
c0d268c3 MS |
3284 | |
3285 | /* entry for "dentry" */ | |
3286 | struct buffer_head *bh; | |
3287 | struct ext4_dir_entry_2 *de; | |
3288 | int inlined; | |
3289 | ||
3290 | /* entry for ".." in inode if it's a directory */ | |
3291 | struct buffer_head *dir_bh; | |
3292 | struct ext4_dir_entry_2 *parent_de; | |
3293 | int dir_inlined; | |
3294 | }; | |
3295 | ||
bd1af145 MS |
3296 | static int ext4_rename_dir_prepare(handle_t *handle, struct ext4_renament *ent) |
3297 | { | |
3298 | int retval; | |
3299 | ||
3300 | ent->dir_bh = ext4_get_first_dir_block(handle, ent->inode, | |
3301 | &retval, &ent->parent_de, | |
3302 | &ent->dir_inlined); | |
3303 | if (!ent->dir_bh) | |
3304 | return retval; | |
3305 | if (le32_to_cpu(ent->parent_de->inode) != ent->dir->i_ino) | |
6a797d27 | 3306 | return -EFSCORRUPTED; |
bd1af145 MS |
3307 | BUFFER_TRACE(ent->dir_bh, "get_write_access"); |
3308 | return ext4_journal_get_write_access(handle, ent->dir_bh); | |
3309 | } | |
3310 | ||
3311 | static int ext4_rename_dir_finish(handle_t *handle, struct ext4_renament *ent, | |
3312 | unsigned dir_ino) | |
3313 | { | |
3314 | int retval; | |
3315 | ||
3316 | ent->parent_de->inode = cpu_to_le32(dir_ino); | |
3317 | BUFFER_TRACE(ent->dir_bh, "call ext4_handle_dirty_metadata"); | |
3318 | if (!ent->dir_inlined) { | |
3319 | if (is_dx(ent->inode)) { | |
3320 | retval = ext4_handle_dirty_dx_node(handle, | |
3321 | ent->inode, | |
3322 | ent->dir_bh); | |
3323 | } else { | |
3324 | retval = ext4_handle_dirty_dirent_node(handle, | |
3325 | ent->inode, | |
3326 | ent->dir_bh); | |
3327 | } | |
3328 | } else { | |
3329 | retval = ext4_mark_inode_dirty(handle, ent->inode); | |
3330 | } | |
3331 | if (retval) { | |
3332 | ext4_std_error(ent->dir->i_sb, retval); | |
3333 | return retval; | |
3334 | } | |
3335 | return 0; | |
3336 | } | |
3337 | ||
3338 | static int ext4_setent(handle_t *handle, struct ext4_renament *ent, | |
3339 | unsigned ino, unsigned file_type) | |
3340 | { | |
3341 | int retval; | |
3342 | ||
3343 | BUFFER_TRACE(ent->bh, "get write access"); | |
3344 | retval = ext4_journal_get_write_access(handle, ent->bh); | |
3345 | if (retval) | |
3346 | return retval; | |
3347 | ent->de->inode = cpu_to_le32(ino); | |
e2b911c5 | 3348 | if (ext4_has_feature_filetype(ent->dir->i_sb)) |
bd1af145 MS |
3349 | ent->de->file_type = file_type; |
3350 | ent->dir->i_version++; | |
3351 | ent->dir->i_ctime = ent->dir->i_mtime = | |
3352 | ext4_current_time(ent->dir); | |
3353 | ext4_mark_inode_dirty(handle, ent->dir); | |
3354 | BUFFER_TRACE(ent->bh, "call ext4_handle_dirty_metadata"); | |
3355 | if (!ent->inlined) { | |
3356 | retval = ext4_handle_dirty_dirent_node(handle, | |
3357 | ent->dir, ent->bh); | |
3358 | if (unlikely(retval)) { | |
3359 | ext4_std_error(ent->dir->i_sb, retval); | |
3360 | return retval; | |
3361 | } | |
3362 | } | |
3363 | brelse(ent->bh); | |
3364 | ent->bh = NULL; | |
3365 | ||
3366 | return 0; | |
3367 | } | |
3368 | ||
3369 | static int ext4_find_delete_entry(handle_t *handle, struct inode *dir, | |
3370 | const struct qstr *d_name) | |
3371 | { | |
3372 | int retval = -ENOENT; | |
3373 | struct buffer_head *bh; | |
3374 | struct ext4_dir_entry_2 *de; | |
3375 | ||
3376 | bh = ext4_find_entry(dir, d_name, &de, NULL); | |
36de9286 TT |
3377 | if (IS_ERR(bh)) |
3378 | return PTR_ERR(bh); | |
bd1af145 MS |
3379 | if (bh) { |
3380 | retval = ext4_delete_entry(handle, dir, de, bh); | |
3381 | brelse(bh); | |
3382 | } | |
3383 | return retval; | |
3384 | } | |
3385 | ||
d80d448c DW |
3386 | static void ext4_rename_delete(handle_t *handle, struct ext4_renament *ent, |
3387 | int force_reread) | |
bd1af145 MS |
3388 | { |
3389 | int retval; | |
3390 | /* | |
3391 | * ent->de could have moved from under us during htree split, so make | |
3392 | * sure that we are deleting the right entry. We might also be pointing | |
3393 | * to a stale entry in the unused part of ent->bh so just checking inum | |
3394 | * and the name isn't enough. | |
3395 | */ | |
3396 | if (le32_to_cpu(ent->de->inode) != ent->inode->i_ino || | |
3397 | ent->de->name_len != ent->dentry->d_name.len || | |
3398 | strncmp(ent->de->name, ent->dentry->d_name.name, | |
d80d448c DW |
3399 | ent->de->name_len) || |
3400 | force_reread) { | |
bd1af145 MS |
3401 | retval = ext4_find_delete_entry(handle, ent->dir, |
3402 | &ent->dentry->d_name); | |
3403 | } else { | |
3404 | retval = ext4_delete_entry(handle, ent->dir, ent->de, ent->bh); | |
3405 | if (retval == -ENOENT) { | |
3406 | retval = ext4_find_delete_entry(handle, ent->dir, | |
3407 | &ent->dentry->d_name); | |
3408 | } | |
3409 | } | |
3410 | ||
3411 | if (retval) { | |
b03a2f7e AD |
3412 | ext4_warning_inode(ent->dir, |
3413 | "Deleting old file: nlink %d, error=%d", | |
3414 | ent->dir->i_nlink, retval); | |
bd1af145 MS |
3415 | } |
3416 | } | |
3417 | ||
bd42998a MS |
3418 | static void ext4_update_dir_count(handle_t *handle, struct ext4_renament *ent) |
3419 | { | |
3420 | if (ent->dir_nlink_delta) { | |
3421 | if (ent->dir_nlink_delta == -1) | |
3422 | ext4_dec_count(handle, ent->dir); | |
3423 | else | |
3424 | ext4_inc_count(handle, ent->dir); | |
3425 | ext4_mark_inode_dirty(handle, ent->dir); | |
3426 | } | |
3427 | } | |
3428 | ||
cd808dec MS |
3429 | static struct inode *ext4_whiteout_for_rename(struct ext4_renament *ent, |
3430 | int credits, handle_t **h) | |
3431 | { | |
3432 | struct inode *wh; | |
3433 | handle_t *handle; | |
3434 | int retries = 0; | |
3435 | ||
3436 | /* | |
3437 | * for inode block, sb block, group summaries, | |
3438 | * and inode bitmap | |
3439 | */ | |
3440 | credits += (EXT4_MAXQUOTAS_TRANS_BLOCKS(ent->dir->i_sb) + | |
3441 | EXT4_XATTR_TRANS_BLOCKS + 4); | |
3442 | retry: | |
3443 | wh = ext4_new_inode_start_handle(ent->dir, S_IFCHR | WHITEOUT_MODE, | |
3444 | &ent->dentry->d_name, 0, NULL, | |
3445 | EXT4_HT_DIR, credits); | |
3446 | ||
3447 | handle = ext4_journal_current_handle(); | |
3448 | if (IS_ERR(wh)) { | |
3449 | if (handle) | |
3450 | ext4_journal_stop(handle); | |
3451 | if (PTR_ERR(wh) == -ENOSPC && | |
3452 | ext4_should_retry_alloc(ent->dir->i_sb, &retries)) | |
3453 | goto retry; | |
3454 | } else { | |
3455 | *h = handle; | |
3456 | init_special_inode(wh, wh->i_mode, WHITEOUT_DEV); | |
3457 | wh->i_op = &ext4_special_inode_operations; | |
3458 | } | |
3459 | return wh; | |
3460 | } | |
3461 | ||
ac27a0ec DK |
3462 | /* |
3463 | * Anybody can rename anything with this: the permission checks are left to the | |
3464 | * higher-level routines. | |
0e202704 TT |
3465 | * |
3466 | * n.b. old_{dentry,inode) refers to the source dentry/inode | |
3467 | * while new_{dentry,inode) refers to the destination dentry/inode | |
3468 | * This comes from rename(const char *oldpath, const char *newpath) | |
ac27a0ec | 3469 | */ |
af5bc92d | 3470 | static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry, |
cd808dec MS |
3471 | struct inode *new_dir, struct dentry *new_dentry, |
3472 | unsigned int flags) | |
ac27a0ec | 3473 | { |
5b61de75 | 3474 | handle_t *handle = NULL; |
c0d268c3 MS |
3475 | struct ext4_renament old = { |
3476 | .dir = old_dir, | |
3477 | .dentry = old_dentry, | |
2b0143b5 | 3478 | .inode = d_inode(old_dentry), |
c0d268c3 MS |
3479 | }; |
3480 | struct ext4_renament new = { | |
3481 | .dir = new_dir, | |
3482 | .dentry = new_dentry, | |
2b0143b5 | 3483 | .inode = d_inode(new_dentry), |
c0d268c3 | 3484 | }; |
d80d448c | 3485 | int force_reread; |
0e202704 | 3486 | int retval; |
cd808dec MS |
3487 | struct inode *whiteout = NULL; |
3488 | int credits; | |
3489 | u8 old_file_type; | |
907f4554 | 3490 | |
a7cdadee JK |
3491 | retval = dquot_initialize(old.dir); |
3492 | if (retval) | |
3493 | return retval; | |
3494 | retval = dquot_initialize(new.dir); | |
3495 | if (retval) | |
3496 | return retval; | |
ac27a0ec DK |
3497 | |
3498 | /* Initialize quotas before so that eventual writes go | |
3499 | * in separate transaction */ | |
a7cdadee JK |
3500 | if (new.inode) { |
3501 | retval = dquot_initialize(new.inode); | |
3502 | if (retval) | |
3503 | return retval; | |
3504 | } | |
ac27a0ec | 3505 | |
c0d268c3 | 3506 | old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, &old.de, NULL); |
36de9286 TT |
3507 | if (IS_ERR(old.bh)) |
3508 | return PTR_ERR(old.bh); | |
ac27a0ec DK |
3509 | /* |
3510 | * Check for inode number is _not_ due to possible IO errors. | |
3511 | * We might rmdir the source, keep it as pwd of some process | |
3512 | * and merrily kill the link to whatever was created under the | |
3513 | * same name. Goodbye sticky bit ;-< | |
3514 | */ | |
ac27a0ec | 3515 | retval = -ENOENT; |
c0d268c3 | 3516 | if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino) |
ac27a0ec DK |
3517 | goto end_rename; |
3518 | ||
d9cdc903 TT |
3519 | if ((old.dir != new.dir) && |
3520 | ext4_encrypted_inode(new.dir) && | |
3521 | !ext4_is_child_context_consistent_with_parent(new.dir, | |
3522 | old.inode)) { | |
3523 | retval = -EPERM; | |
3524 | goto end_rename; | |
3525 | } | |
3526 | ||
c0d268c3 MS |
3527 | new.bh = ext4_find_entry(new.dir, &new.dentry->d_name, |
3528 | &new.de, &new.inlined); | |
36de9286 TT |
3529 | if (IS_ERR(new.bh)) { |
3530 | retval = PTR_ERR(new.bh); | |
a9cfcd63 | 3531 | new.bh = NULL; |
36de9286 TT |
3532 | goto end_rename; |
3533 | } | |
c0d268c3 MS |
3534 | if (new.bh) { |
3535 | if (!new.inode) { | |
3536 | brelse(new.bh); | |
3537 | new.bh = NULL; | |
ac27a0ec DK |
3538 | } |
3539 | } | |
c0d268c3 MS |
3540 | if (new.inode && !test_opt(new.dir->i_sb, NO_AUTO_DA_ALLOC)) |
3541 | ext4_alloc_da_blocks(old.inode); | |
5b61de75 | 3542 | |
cd808dec MS |
3543 | credits = (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) + |
3544 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2); | |
3545 | if (!(flags & RENAME_WHITEOUT)) { | |
3546 | handle = ext4_journal_start(old.dir, EXT4_HT_DIR, credits); | |
7071b715 KK |
3547 | if (IS_ERR(handle)) { |
3548 | retval = PTR_ERR(handle); | |
3549 | handle = NULL; | |
3550 | goto end_rename; | |
3551 | } | |
cd808dec MS |
3552 | } else { |
3553 | whiteout = ext4_whiteout_for_rename(&old, credits, &handle); | |
7071b715 KK |
3554 | if (IS_ERR(whiteout)) { |
3555 | retval = PTR_ERR(whiteout); | |
3556 | whiteout = NULL; | |
3557 | goto end_rename; | |
3558 | } | |
cd808dec | 3559 | } |
5b61de75 | 3560 | |
c0d268c3 | 3561 | if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir)) |
5b61de75 TT |
3562 | ext4_handle_sync(handle); |
3563 | ||
c0d268c3 MS |
3564 | if (S_ISDIR(old.inode->i_mode)) { |
3565 | if (new.inode) { | |
ac27a0ec | 3566 | retval = -ENOTEMPTY; |
e875a2dd | 3567 | if (!ext4_empty_dir(new.inode)) |
ac27a0ec | 3568 | goto end_rename; |
0d7d5d67 MS |
3569 | } else { |
3570 | retval = -EMLINK; | |
3571 | if (new.dir != old.dir && EXT4_DIR_LINK_MAX(new.dir)) | |
3572 | goto end_rename; | |
ac27a0ec | 3573 | } |
bd1af145 | 3574 | retval = ext4_rename_dir_prepare(handle, &old); |
ef607893 AG |
3575 | if (retval) |
3576 | goto end_rename; | |
ac27a0ec | 3577 | } |
d80d448c DW |
3578 | /* |
3579 | * If we're renaming a file within an inline_data dir and adding or | |
3580 | * setting the new dirent causes a conversion from inline_data to | |
3581 | * extents/blockmap, we need to force the dirent delete code to | |
3582 | * re-read the directory, or else we end up trying to delete a dirent | |
3583 | * from what is now the extent tree root (or a block map). | |
3584 | */ | |
3585 | force_reread = (new.dir->i_ino == old.dir->i_ino && | |
3586 | ext4_test_inode_flag(new.dir, EXT4_INODE_INLINE_DATA)); | |
cd808dec MS |
3587 | |
3588 | old_file_type = old.de->file_type; | |
3589 | if (whiteout) { | |
3590 | /* | |
3591 | * Do this before adding a new entry, so the old entry is sure | |
3592 | * to be still pointing to the valid old entry. | |
3593 | */ | |
3594 | retval = ext4_setent(handle, &old, whiteout->i_ino, | |
3595 | EXT4_FT_CHRDEV); | |
3596 | if (retval) | |
3597 | goto end_rename; | |
3598 | ext4_mark_inode_dirty(handle, whiteout); | |
3599 | } | |
c0d268c3 MS |
3600 | if (!new.bh) { |
3601 | retval = ext4_add_entry(handle, new.dentry, old.inode); | |
ac27a0ec DK |
3602 | if (retval) |
3603 | goto end_rename; | |
3604 | } else { | |
bd1af145 | 3605 | retval = ext4_setent(handle, &new, |
cd808dec | 3606 | old.inode->i_ino, old_file_type); |
ef607893 AG |
3607 | if (retval) |
3608 | goto end_rename; | |
ac27a0ec | 3609 | } |
d80d448c DW |
3610 | if (force_reread) |
3611 | force_reread = !ext4_test_inode_flag(new.dir, | |
3612 | EXT4_INODE_INLINE_DATA); | |
ac27a0ec DK |
3613 | |
3614 | /* | |
3615 | * Like most other Unix systems, set the ctime for inodes on a | |
3616 | * rename. | |
3617 | */ | |
c0d268c3 MS |
3618 | old.inode->i_ctime = ext4_current_time(old.inode); |
3619 | ext4_mark_inode_dirty(handle, old.inode); | |
ac27a0ec | 3620 | |
cd808dec MS |
3621 | if (!whiteout) { |
3622 | /* | |
3623 | * ok, that's it | |
3624 | */ | |
3625 | ext4_rename_delete(handle, &old, force_reread); | |
3626 | } | |
ac27a0ec | 3627 | |
c0d268c3 MS |
3628 | if (new.inode) { |
3629 | ext4_dec_count(handle, new.inode); | |
3630 | new.inode->i_ctime = ext4_current_time(new.inode); | |
ac27a0ec | 3631 | } |
c0d268c3 MS |
3632 | old.dir->i_ctime = old.dir->i_mtime = ext4_current_time(old.dir); |
3633 | ext4_update_dx_flag(old.dir); | |
3634 | if (old.dir_bh) { | |
bd1af145 MS |
3635 | retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino); |
3636 | if (retval) | |
b4097142 | 3637 | goto end_rename; |
bd1af145 | 3638 | |
c0d268c3 MS |
3639 | ext4_dec_count(handle, old.dir); |
3640 | if (new.inode) { | |
e875a2dd MH |
3641 | /* checked ext4_empty_dir above, can't have another |
3642 | * parent, ext4_dec_count() won't work for many-linked | |
3643 | * dirs */ | |
c0d268c3 | 3644 | clear_nlink(new.inode); |
ac27a0ec | 3645 | } else { |
c0d268c3 MS |
3646 | ext4_inc_count(handle, new.dir); |
3647 | ext4_update_dx_flag(new.dir); | |
3648 | ext4_mark_inode_dirty(handle, new.dir); | |
ac27a0ec DK |
3649 | } |
3650 | } | |
c0d268c3 MS |
3651 | ext4_mark_inode_dirty(handle, old.dir); |
3652 | if (new.inode) { | |
3653 | ext4_mark_inode_dirty(handle, new.inode); | |
3654 | if (!new.inode->i_nlink) | |
3655 | ext4_orphan_add(handle, new.inode); | |
ac27a0ec DK |
3656 | } |
3657 | retval = 0; | |
3658 | ||
3659 | end_rename: | |
c0d268c3 MS |
3660 | brelse(old.dir_bh); |
3661 | brelse(old.bh); | |
3662 | brelse(new.bh); | |
cd808dec MS |
3663 | if (whiteout) { |
3664 | if (retval) | |
3665 | drop_nlink(whiteout); | |
3666 | unlock_new_inode(whiteout); | |
3667 | iput(whiteout); | |
3668 | } | |
5b61de75 TT |
3669 | if (handle) |
3670 | ext4_journal_stop(handle); | |
ac27a0ec DK |
3671 | return retval; |
3672 | } | |
3673 | ||
bd42998a MS |
3674 | static int ext4_cross_rename(struct inode *old_dir, struct dentry *old_dentry, |
3675 | struct inode *new_dir, struct dentry *new_dentry) | |
3676 | { | |
3677 | handle_t *handle = NULL; | |
3678 | struct ext4_renament old = { | |
3679 | .dir = old_dir, | |
3680 | .dentry = old_dentry, | |
2b0143b5 | 3681 | .inode = d_inode(old_dentry), |
bd42998a MS |
3682 | }; |
3683 | struct ext4_renament new = { | |
3684 | .dir = new_dir, | |
3685 | .dentry = new_dentry, | |
2b0143b5 | 3686 | .inode = d_inode(new_dentry), |
bd42998a MS |
3687 | }; |
3688 | u8 new_file_type; | |
3689 | int retval; | |
3690 | ||
c2faccaf TT |
3691 | if ((ext4_encrypted_inode(old_dir) || |
3692 | ext4_encrypted_inode(new_dir)) && | |
3693 | (old_dir != new_dir) && | |
3694 | (!ext4_is_child_context_consistent_with_parent(new_dir, | |
3695 | old.inode) || | |
3696 | !ext4_is_child_context_consistent_with_parent(old_dir, | |
3697 | new.inode))) | |
3698 | return -EPERM; | |
3699 | ||
a7cdadee JK |
3700 | retval = dquot_initialize(old.dir); |
3701 | if (retval) | |
3702 | return retval; | |
3703 | retval = dquot_initialize(new.dir); | |
3704 | if (retval) | |
3705 | return retval; | |
bd42998a MS |
3706 | |
3707 | old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, | |
3708 | &old.de, &old.inlined); | |
36de9286 TT |
3709 | if (IS_ERR(old.bh)) |
3710 | return PTR_ERR(old.bh); | |
bd42998a MS |
3711 | /* |
3712 | * Check for inode number is _not_ due to possible IO errors. | |
3713 | * We might rmdir the source, keep it as pwd of some process | |
3714 | * and merrily kill the link to whatever was created under the | |
3715 | * same name. Goodbye sticky bit ;-< | |
3716 | */ | |
3717 | retval = -ENOENT; | |
3718 | if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino) | |
3719 | goto end_rename; | |
3720 | ||
3721 | new.bh = ext4_find_entry(new.dir, &new.dentry->d_name, | |
3722 | &new.de, &new.inlined); | |
36de9286 TT |
3723 | if (IS_ERR(new.bh)) { |
3724 | retval = PTR_ERR(new.bh); | |
a9cfcd63 | 3725 | new.bh = NULL; |
36de9286 TT |
3726 | goto end_rename; |
3727 | } | |
bd42998a MS |
3728 | |
3729 | /* RENAME_EXCHANGE case: old *and* new must both exist */ | |
3730 | if (!new.bh || le32_to_cpu(new.de->inode) != new.inode->i_ino) | |
3731 | goto end_rename; | |
3732 | ||
3733 | handle = ext4_journal_start(old.dir, EXT4_HT_DIR, | |
3734 | (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) + | |
3735 | 2 * EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2)); | |
7071b715 KK |
3736 | if (IS_ERR(handle)) { |
3737 | retval = PTR_ERR(handle); | |
3738 | handle = NULL; | |
3739 | goto end_rename; | |
3740 | } | |
bd42998a MS |
3741 | |
3742 | if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir)) | |
3743 | ext4_handle_sync(handle); | |
3744 | ||
3745 | if (S_ISDIR(old.inode->i_mode)) { | |
3746 | old.is_dir = true; | |
3747 | retval = ext4_rename_dir_prepare(handle, &old); | |
3748 | if (retval) | |
3749 | goto end_rename; | |
3750 | } | |
3751 | if (S_ISDIR(new.inode->i_mode)) { | |
3752 | new.is_dir = true; | |
3753 | retval = ext4_rename_dir_prepare(handle, &new); | |
3754 | if (retval) | |
3755 | goto end_rename; | |
3756 | } | |
3757 | ||
3758 | /* | |
3759 | * Other than the special case of overwriting a directory, parents' | |
3760 | * nlink only needs to be modified if this is a cross directory rename. | |
3761 | */ | |
3762 | if (old.dir != new.dir && old.is_dir != new.is_dir) { | |
3763 | old.dir_nlink_delta = old.is_dir ? -1 : 1; | |
3764 | new.dir_nlink_delta = -old.dir_nlink_delta; | |
3765 | retval = -EMLINK; | |
3766 | if ((old.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(old.dir)) || | |
3767 | (new.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(new.dir))) | |
3768 | goto end_rename; | |
3769 | } | |
3770 | ||
3771 | new_file_type = new.de->file_type; | |
3772 | retval = ext4_setent(handle, &new, old.inode->i_ino, old.de->file_type); | |
3773 | if (retval) | |
3774 | goto end_rename; | |
3775 | ||
3776 | retval = ext4_setent(handle, &old, new.inode->i_ino, new_file_type); | |
3777 | if (retval) | |
3778 | goto end_rename; | |
3779 | ||
3780 | /* | |
3781 | * Like most other Unix systems, set the ctime for inodes on a | |
3782 | * rename. | |
3783 | */ | |
3784 | old.inode->i_ctime = ext4_current_time(old.inode); | |
3785 | new.inode->i_ctime = ext4_current_time(new.inode); | |
3786 | ext4_mark_inode_dirty(handle, old.inode); | |
3787 | ext4_mark_inode_dirty(handle, new.inode); | |
3788 | ||
3789 | if (old.dir_bh) { | |
3790 | retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino); | |
3791 | if (retval) | |
3792 | goto end_rename; | |
3793 | } | |
3794 | if (new.dir_bh) { | |
3795 | retval = ext4_rename_dir_finish(handle, &new, old.dir->i_ino); | |
3796 | if (retval) | |
3797 | goto end_rename; | |
3798 | } | |
3799 | ext4_update_dir_count(handle, &old); | |
3800 | ext4_update_dir_count(handle, &new); | |
3801 | retval = 0; | |
3802 | ||
3803 | end_rename: | |
3804 | brelse(old.dir_bh); | |
3805 | brelse(new.dir_bh); | |
3806 | brelse(old.bh); | |
3807 | brelse(new.bh); | |
3808 | if (handle) | |
3809 | ext4_journal_stop(handle); | |
3810 | return retval; | |
3811 | } | |
3812 | ||
0a7c3937 MS |
3813 | static int ext4_rename2(struct inode *old_dir, struct dentry *old_dentry, |
3814 | struct inode *new_dir, struct dentry *new_dentry, | |
3815 | unsigned int flags) | |
3816 | { | |
cd808dec | 3817 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
0a7c3937 MS |
3818 | return -EINVAL; |
3819 | ||
bd42998a MS |
3820 | if (flags & RENAME_EXCHANGE) { |
3821 | return ext4_cross_rename(old_dir, old_dentry, | |
3822 | new_dir, new_dentry); | |
3823 | } | |
cd808dec MS |
3824 | |
3825 | return ext4_rename(old_dir, old_dentry, new_dir, new_dentry, flags); | |
0a7c3937 MS |
3826 | } |
3827 | ||
ac27a0ec DK |
3828 | /* |
3829 | * directories can handle most operations... | |
3830 | */ | |
754661f1 | 3831 | const struct inode_operations ext4_dir_inode_operations = { |
617ba13b MC |
3832 | .create = ext4_create, |
3833 | .lookup = ext4_lookup, | |
3834 | .link = ext4_link, | |
3835 | .unlink = ext4_unlink, | |
3836 | .symlink = ext4_symlink, | |
3837 | .mkdir = ext4_mkdir, | |
3838 | .rmdir = ext4_rmdir, | |
3839 | .mknod = ext4_mknod, | |
af51a2ac | 3840 | .tmpfile = ext4_tmpfile, |
0a7c3937 | 3841 | .rename2 = ext4_rename2, |
617ba13b | 3842 | .setattr = ext4_setattr, |
ac27a0ec DK |
3843 | .setxattr = generic_setxattr, |
3844 | .getxattr = generic_getxattr, | |
617ba13b | 3845 | .listxattr = ext4_listxattr, |
ac27a0ec | 3846 | .removexattr = generic_removexattr, |
4e34e719 | 3847 | .get_acl = ext4_get_acl, |
64e178a7 | 3848 | .set_acl = ext4_set_acl, |
abc8746e | 3849 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
3850 | }; |
3851 | ||
754661f1 | 3852 | const struct inode_operations ext4_special_inode_operations = { |
617ba13b | 3853 | .setattr = ext4_setattr, |
ac27a0ec DK |
3854 | .setxattr = generic_setxattr, |
3855 | .getxattr = generic_getxattr, | |
617ba13b | 3856 | .listxattr = ext4_listxattr, |
ac27a0ec | 3857 | .removexattr = generic_removexattr, |
4e34e719 | 3858 | .get_acl = ext4_get_acl, |
64e178a7 | 3859 | .set_acl = ext4_set_acl, |
ac27a0ec | 3860 | }; |