Commit | Line | Data |
---|---|---|
b2441318 | 1 | // SPDX-License-Identifier: GPL-2.0 |
ac27a0ec | 2 | /* |
617ba13b | 3 | * linux/fs/ext4/namei.c |
ac27a0ec DK |
4 | * |
5 | * Copyright (C) 1992, 1993, 1994, 1995 | |
6 | * Remy Card (card@masi.ibp.fr) | |
7 | * Laboratoire MASI - Institut Blaise Pascal | |
8 | * Universite Pierre et Marie Curie (Paris VI) | |
9 | * | |
10 | * from | |
11 | * | |
12 | * linux/fs/minix/namei.c | |
13 | * | |
14 | * Copyright (C) 1991, 1992 Linus Torvalds | |
15 | * | |
16 | * Big-endian to little-endian byte-swapping/bitmaps by | |
17 | * David S. Miller (davem@caip.rutgers.edu), 1995 | |
18 | * Directory entry file type support and forward compatibility hooks | |
19 | * for B-tree directories by Theodore Ts'o (tytso@mit.edu), 1998 | |
20 | * Hash Tree Directory indexing (c) | |
21 | * Daniel Phillips, 2001 | |
22 | * Hash Tree Directory indexing porting | |
23 | * Christopher Li, 2002 | |
24 | * Hash Tree Directory indexing cleanup | |
25 | * Theodore Ts'o, 2002 | |
26 | */ | |
27 | ||
28 | #include <linux/fs.h> | |
29 | #include <linux/pagemap.h> | |
ac27a0ec | 30 | #include <linux/time.h> |
ac27a0ec DK |
31 | #include <linux/fcntl.h> |
32 | #include <linux/stat.h> | |
33 | #include <linux/string.h> | |
34 | #include <linux/quotaops.h> | |
35 | #include <linux/buffer_head.h> | |
36 | #include <linux/bio.h> | |
ae5e165d | 37 | #include <linux/iversion.h> |
3dcf5451 CH |
38 | #include "ext4.h" |
39 | #include "ext4_jbd2.h" | |
ac27a0ec | 40 | |
ac27a0ec DK |
41 | #include "xattr.h" |
42 | #include "acl.h" | |
43 | ||
0562e0ba | 44 | #include <trace/events/ext4.h> |
ac27a0ec DK |
45 | /* |
46 | * define how far ahead to read directories while searching them. | |
47 | */ | |
48 | #define NAMEI_RA_CHUNKS 2 | |
49 | #define NAMEI_RA_BLOCKS 4 | |
8c55e204 | 50 | #define NAMEI_RA_SIZE (NAMEI_RA_CHUNKS * NAMEI_RA_BLOCKS) |
ac27a0ec | 51 | |
617ba13b | 52 | static struct buffer_head *ext4_append(handle_t *handle, |
ac27a0ec | 53 | struct inode *inode, |
0f70b406 | 54 | ext4_lblk_t *block) |
ac27a0ec DK |
55 | { |
56 | struct buffer_head *bh; | |
1c215028 | 57 | int err; |
ac27a0ec | 58 | |
df981d03 TT |
59 | if (unlikely(EXT4_SB(inode->i_sb)->s_max_dir_size_kb && |
60 | ((inode->i_size >> 10) >= | |
0f70b406 TT |
61 | EXT4_SB(inode->i_sb)->s_max_dir_size_kb))) |
62 | return ERR_PTR(-ENOSPC); | |
df981d03 | 63 | |
ac27a0ec DK |
64 | *block = inode->i_size >> inode->i_sb->s_blocksize_bits; |
65 | ||
c5e298ae | 66 | bh = ext4_bread(handle, inode, *block, EXT4_GET_BLOCKS_CREATE); |
1c215028 TT |
67 | if (IS_ERR(bh)) |
68 | return bh; | |
0f70b406 TT |
69 | inode->i_size += inode->i_sb->s_blocksize; |
70 | EXT4_I(inode)->i_disksize = inode->i_size; | |
5d601255 | 71 | BUFFER_TRACE(bh, "get_write_access"); |
0f70b406 TT |
72 | err = ext4_journal_get_write_access(handle, bh); |
73 | if (err) { | |
74 | brelse(bh); | |
75 | ext4_std_error(inode->i_sb, err); | |
76 | return ERR_PTR(err); | |
6d1ab10e | 77 | } |
ac27a0ec DK |
78 | return bh; |
79 | } | |
80 | ||
dc6982ff TT |
81 | static int ext4_dx_csum_verify(struct inode *inode, |
82 | struct ext4_dir_entry *dirent); | |
83 | ||
84 | typedef enum { | |
85 | EITHER, INDEX, DIRENT | |
86 | } dirblock_type_t; | |
87 | ||
88 | #define ext4_read_dirblock(inode, block, type) \ | |
b03a2f7e | 89 | __ext4_read_dirblock((inode), (block), (type), __func__, __LINE__) |
dc6982ff TT |
90 | |
91 | static struct buffer_head *__ext4_read_dirblock(struct inode *inode, | |
b03a2f7e AD |
92 | ext4_lblk_t block, |
93 | dirblock_type_t type, | |
94 | const char *func, | |
95 | unsigned int line) | |
dc6982ff TT |
96 | { |
97 | struct buffer_head *bh; | |
98 | struct ext4_dir_entry *dirent; | |
1c215028 | 99 | int is_dx_block = 0; |
dc6982ff | 100 | |
1c215028 TT |
101 | bh = ext4_bread(NULL, inode, block, 0); |
102 | if (IS_ERR(bh)) { | |
b03a2f7e AD |
103 | __ext4_warning(inode->i_sb, func, line, |
104 | "inode #%lu: lblock %lu: comm %s: " | |
105 | "error %ld reading directory block", | |
106 | inode->i_ino, (unsigned long)block, | |
107 | current->comm, PTR_ERR(bh)); | |
1c215028 TT |
108 | |
109 | return bh; | |
110 | } | |
111 | if (!bh) { | |
b03a2f7e AD |
112 | ext4_error_inode(inode, func, line, block, |
113 | "Directory hole found"); | |
6a797d27 | 114 | return ERR_PTR(-EFSCORRUPTED); |
dc6982ff TT |
115 | } |
116 | dirent = (struct ext4_dir_entry *) bh->b_data; | |
117 | /* Determine whether or not we have an index block */ | |
118 | if (is_dx(inode)) { | |
119 | if (block == 0) | |
120 | is_dx_block = 1; | |
121 | else if (ext4_rec_len_from_disk(dirent->rec_len, | |
122 | inode->i_sb->s_blocksize) == | |
123 | inode->i_sb->s_blocksize) | |
124 | is_dx_block = 1; | |
125 | } | |
126 | if (!is_dx_block && type == INDEX) { | |
b03a2f7e | 127 | ext4_error_inode(inode, func, line, block, |
dc6982ff | 128 | "directory leaf block found instead of index block"); |
de59fae0 | 129 | brelse(bh); |
6a797d27 | 130 | return ERR_PTR(-EFSCORRUPTED); |
dc6982ff | 131 | } |
9aa5d32b | 132 | if (!ext4_has_metadata_csum(inode->i_sb) || |
dc6982ff TT |
133 | buffer_verified(bh)) |
134 | return bh; | |
135 | ||
136 | /* | |
137 | * An empty leaf block can get mistaken for a index block; for | |
138 | * this reason, we can only check the index checksum when the | |
139 | * caller is sure it should be an index block. | |
140 | */ | |
141 | if (is_dx_block && type == INDEX) { | |
142 | if (ext4_dx_csum_verify(inode, dirent)) | |
143 | set_buffer_verified(bh); | |
144 | else { | |
b03a2f7e AD |
145 | ext4_error_inode(inode, func, line, block, |
146 | "Directory index failed checksum"); | |
a871611b | 147 | brelse(bh); |
6a797d27 | 148 | return ERR_PTR(-EFSBADCRC); |
a871611b | 149 | } |
ac27a0ec | 150 | } |
dc6982ff TT |
151 | if (!is_dx_block) { |
152 | if (ext4_dirent_csum_verify(inode, dirent)) | |
153 | set_buffer_verified(bh); | |
154 | else { | |
b03a2f7e AD |
155 | ext4_error_inode(inode, func, line, block, |
156 | "Directory block failed checksum"); | |
dc6982ff | 157 | brelse(bh); |
6a797d27 | 158 | return ERR_PTR(-EFSBADCRC); |
dc6982ff | 159 | } |
6d1ab10e | 160 | } |
ac27a0ec DK |
161 | return bh; |
162 | } | |
163 | ||
164 | #ifndef assert | |
165 | #define assert(test) J_ASSERT(test) | |
166 | #endif | |
167 | ||
ac27a0ec DK |
168 | #ifdef DX_DEBUG |
169 | #define dxtrace(command) command | |
170 | #else | |
171 | #define dxtrace(command) | |
172 | #endif | |
173 | ||
174 | struct fake_dirent | |
175 | { | |
176 | __le32 inode; | |
177 | __le16 rec_len; | |
178 | u8 name_len; | |
179 | u8 file_type; | |
180 | }; | |
181 | ||
182 | struct dx_countlimit | |
183 | { | |
184 | __le16 limit; | |
185 | __le16 count; | |
186 | }; | |
187 | ||
188 | struct dx_entry | |
189 | { | |
190 | __le32 hash; | |
191 | __le32 block; | |
192 | }; | |
193 | ||
194 | /* | |
195 | * dx_root_info is laid out so that if it should somehow get overlaid by a | |
196 | * dirent the two low bits of the hash version will be zero. Therefore, the | |
197 | * hash version mod 4 should never be 0. Sincerely, the paranoia department. | |
198 | */ | |
199 | ||
200 | struct dx_root | |
201 | { | |
202 | struct fake_dirent dot; | |
203 | char dot_name[4]; | |
204 | struct fake_dirent dotdot; | |
205 | char dotdot_name[4]; | |
206 | struct dx_root_info | |
207 | { | |
208 | __le32 reserved_zero; | |
209 | u8 hash_version; | |
210 | u8 info_length; /* 8 */ | |
211 | u8 indirect_levels; | |
212 | u8 unused_flags; | |
213 | } | |
214 | info; | |
215 | struct dx_entry entries[0]; | |
216 | }; | |
217 | ||
218 | struct dx_node | |
219 | { | |
220 | struct fake_dirent fake; | |
221 | struct dx_entry entries[0]; | |
222 | }; | |
223 | ||
224 | ||
225 | struct dx_frame | |
226 | { | |
227 | struct buffer_head *bh; | |
228 | struct dx_entry *entries; | |
229 | struct dx_entry *at; | |
230 | }; | |
231 | ||
232 | struct dx_map_entry | |
233 | { | |
234 | u32 hash; | |
ef2b02d3 ES |
235 | u16 offs; |
236 | u16 size; | |
ac27a0ec DK |
237 | }; |
238 | ||
e6153918 DW |
239 | /* |
240 | * This goes at the end of each htree block. | |
241 | */ | |
242 | struct dx_tail { | |
243 | u32 dt_reserved; | |
244 | __le32 dt_checksum; /* crc32c(uuid+inum+dirblock) */ | |
245 | }; | |
246 | ||
725d26d3 AK |
247 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry); |
248 | static void dx_set_block(struct dx_entry *entry, ext4_lblk_t value); | |
af5bc92d TT |
249 | static inline unsigned dx_get_hash(struct dx_entry *entry); |
250 | static void dx_set_hash(struct dx_entry *entry, unsigned value); | |
251 | static unsigned dx_get_count(struct dx_entry *entries); | |
252 | static unsigned dx_get_limit(struct dx_entry *entries); | |
253 | static void dx_set_count(struct dx_entry *entries, unsigned value); | |
254 | static void dx_set_limit(struct dx_entry *entries, unsigned value); | |
255 | static unsigned dx_root_limit(struct inode *dir, unsigned infosize); | |
256 | static unsigned dx_node_limit(struct inode *dir); | |
5b643f9c | 257 | static struct dx_frame *dx_probe(struct ext4_filename *fname, |
ac27a0ec DK |
258 | struct inode *dir, |
259 | struct dx_hash_info *hinfo, | |
dd73b5d5 | 260 | struct dx_frame *frame); |
af5bc92d | 261 | static void dx_release(struct dx_frame *frames); |
1f3862b5 MH |
262 | static int dx_make_map(struct inode *dir, struct ext4_dir_entry_2 *de, |
263 | unsigned blocksize, struct dx_hash_info *hinfo, | |
264 | struct dx_map_entry map[]); | |
ac27a0ec | 265 | static void dx_sort_map(struct dx_map_entry *map, unsigned count); |
af5bc92d | 266 | static struct ext4_dir_entry_2 *dx_move_dirents(char *from, char *to, |
3d0518f4 | 267 | struct dx_map_entry *offsets, int count, unsigned blocksize); |
8bad4597 | 268 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize); |
725d26d3 AK |
269 | static void dx_insert_block(struct dx_frame *frame, |
270 | u32 hash, ext4_lblk_t block); | |
617ba13b | 271 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
272 | struct dx_frame *frame, |
273 | struct dx_frame *frames, | |
274 | __u32 *start_hash); | |
f702ba0f | 275 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, |
5b643f9c | 276 | struct ext4_filename *fname, |
537d8f93 | 277 | struct ext4_dir_entry_2 **res_dir); |
5b643f9c | 278 | static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 279 | struct inode *dir, struct inode *inode); |
ac27a0ec | 280 | |
dbe89444 | 281 | /* checksumming functions */ |
3c47d541 TM |
282 | void initialize_dirent_tail(struct ext4_dir_entry_tail *t, |
283 | unsigned int blocksize) | |
b0336e8d DW |
284 | { |
285 | memset(t, 0, sizeof(struct ext4_dir_entry_tail)); | |
286 | t->det_rec_len = ext4_rec_len_to_disk( | |
287 | sizeof(struct ext4_dir_entry_tail), blocksize); | |
288 | t->det_reserved_ft = EXT4_FT_DIR_CSUM; | |
289 | } | |
290 | ||
291 | /* Walk through a dirent block to find a checksum "dirent" at the tail */ | |
292 | static struct ext4_dir_entry_tail *get_dirent_tail(struct inode *inode, | |
293 | struct ext4_dir_entry *de) | |
294 | { | |
295 | struct ext4_dir_entry_tail *t; | |
296 | ||
297 | #ifdef PARANOID | |
298 | struct ext4_dir_entry *d, *top; | |
299 | ||
300 | d = de; | |
301 | top = (struct ext4_dir_entry *)(((void *)de) + | |
302 | (EXT4_BLOCK_SIZE(inode->i_sb) - | |
303 | sizeof(struct ext4_dir_entry_tail))); | |
304 | while (d < top && d->rec_len) | |
305 | d = (struct ext4_dir_entry *)(((void *)d) + | |
306 | le16_to_cpu(d->rec_len)); | |
307 | ||
308 | if (d != top) | |
309 | return NULL; | |
310 | ||
311 | t = (struct ext4_dir_entry_tail *)d; | |
312 | #else | |
313 | t = EXT4_DIRENT_TAIL(de, EXT4_BLOCK_SIZE(inode->i_sb)); | |
314 | #endif | |
315 | ||
316 | if (t->det_reserved_zero1 || | |
317 | le16_to_cpu(t->det_rec_len) != sizeof(struct ext4_dir_entry_tail) || | |
318 | t->det_reserved_zero2 || | |
319 | t->det_reserved_ft != EXT4_FT_DIR_CSUM) | |
320 | return NULL; | |
321 | ||
322 | return t; | |
323 | } | |
324 | ||
325 | static __le32 ext4_dirent_csum(struct inode *inode, | |
326 | struct ext4_dir_entry *dirent, int size) | |
327 | { | |
328 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
329 | struct ext4_inode_info *ei = EXT4_I(inode); | |
330 | __u32 csum; | |
331 | ||
332 | csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size); | |
333 | return cpu_to_le32(csum); | |
334 | } | |
335 | ||
b03a2f7e AD |
336 | #define warn_no_space_for_csum(inode) \ |
337 | __warn_no_space_for_csum((inode), __func__, __LINE__) | |
338 | ||
339 | static void __warn_no_space_for_csum(struct inode *inode, const char *func, | |
340 | unsigned int line) | |
dffe9d8d | 341 | { |
b03a2f7e AD |
342 | __ext4_warning_inode(inode, func, line, |
343 | "No space for directory leaf checksum. Please run e2fsck -D."); | |
dffe9d8d TT |
344 | } |
345 | ||
b0336e8d DW |
346 | int ext4_dirent_csum_verify(struct inode *inode, struct ext4_dir_entry *dirent) |
347 | { | |
348 | struct ext4_dir_entry_tail *t; | |
349 | ||
9aa5d32b | 350 | if (!ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d DW |
351 | return 1; |
352 | ||
353 | t = get_dirent_tail(inode, dirent); | |
354 | if (!t) { | |
dffe9d8d | 355 | warn_no_space_for_csum(inode); |
b0336e8d DW |
356 | return 0; |
357 | } | |
358 | ||
359 | if (t->det_checksum != ext4_dirent_csum(inode, dirent, | |
360 | (void *)t - (void *)dirent)) | |
361 | return 0; | |
362 | ||
363 | return 1; | |
364 | } | |
365 | ||
366 | static void ext4_dirent_csum_set(struct inode *inode, | |
367 | struct ext4_dir_entry *dirent) | |
368 | { | |
369 | struct ext4_dir_entry_tail *t; | |
370 | ||
9aa5d32b | 371 | if (!ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d DW |
372 | return; |
373 | ||
374 | t = get_dirent_tail(inode, dirent); | |
375 | if (!t) { | |
dffe9d8d | 376 | warn_no_space_for_csum(inode); |
b0336e8d DW |
377 | return; |
378 | } | |
379 | ||
380 | t->det_checksum = ext4_dirent_csum(inode, dirent, | |
381 | (void *)t - (void *)dirent); | |
382 | } | |
383 | ||
3c47d541 TM |
384 | int ext4_handle_dirty_dirent_node(handle_t *handle, |
385 | struct inode *inode, | |
386 | struct buffer_head *bh) | |
b0336e8d DW |
387 | { |
388 | ext4_dirent_csum_set(inode, (struct ext4_dir_entry *)bh->b_data); | |
389 | return ext4_handle_dirty_metadata(handle, inode, bh); | |
390 | } | |
391 | ||
dbe89444 DW |
392 | static struct dx_countlimit *get_dx_countlimit(struct inode *inode, |
393 | struct ext4_dir_entry *dirent, | |
394 | int *offset) | |
395 | { | |
396 | struct ext4_dir_entry *dp; | |
397 | struct dx_root_info *root; | |
398 | int count_offset; | |
399 | ||
400 | if (le16_to_cpu(dirent->rec_len) == EXT4_BLOCK_SIZE(inode->i_sb)) | |
401 | count_offset = 8; | |
402 | else if (le16_to_cpu(dirent->rec_len) == 12) { | |
403 | dp = (struct ext4_dir_entry *)(((void *)dirent) + 12); | |
404 | if (le16_to_cpu(dp->rec_len) != | |
405 | EXT4_BLOCK_SIZE(inode->i_sb) - 12) | |
406 | return NULL; | |
407 | root = (struct dx_root_info *)(((void *)dp + 12)); | |
408 | if (root->reserved_zero || | |
409 | root->info_length != sizeof(struct dx_root_info)) | |
410 | return NULL; | |
411 | count_offset = 32; | |
412 | } else | |
413 | return NULL; | |
414 | ||
415 | if (offset) | |
416 | *offset = count_offset; | |
417 | return (struct dx_countlimit *)(((void *)dirent) + count_offset); | |
418 | } | |
419 | ||
420 | static __le32 ext4_dx_csum(struct inode *inode, struct ext4_dir_entry *dirent, | |
421 | int count_offset, int count, struct dx_tail *t) | |
422 | { | |
423 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); | |
424 | struct ext4_inode_info *ei = EXT4_I(inode); | |
d6a77105 | 425 | __u32 csum; |
dbe89444 | 426 | int size; |
b47820ed DJ |
427 | __u32 dummy_csum = 0; |
428 | int offset = offsetof(struct dx_tail, dt_checksum); | |
dbe89444 DW |
429 | |
430 | size = count_offset + (count * sizeof(struct dx_entry)); | |
dbe89444 | 431 | csum = ext4_chksum(sbi, ei->i_csum_seed, (__u8 *)dirent, size); |
b47820ed DJ |
432 | csum = ext4_chksum(sbi, csum, (__u8 *)t, offset); |
433 | csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum, sizeof(dummy_csum)); | |
dbe89444 DW |
434 | |
435 | return cpu_to_le32(csum); | |
436 | } | |
437 | ||
438 | static int ext4_dx_csum_verify(struct inode *inode, | |
439 | struct ext4_dir_entry *dirent) | |
440 | { | |
441 | struct dx_countlimit *c; | |
442 | struct dx_tail *t; | |
443 | int count_offset, limit, count; | |
444 | ||
9aa5d32b | 445 | if (!ext4_has_metadata_csum(inode->i_sb)) |
dbe89444 DW |
446 | return 1; |
447 | ||
448 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
449 | if (!c) { | |
450 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
fa964540 | 451 | return 0; |
dbe89444 DW |
452 | } |
453 | limit = le16_to_cpu(c->limit); | |
454 | count = le16_to_cpu(c->count); | |
455 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
456 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
dffe9d8d | 457 | warn_no_space_for_csum(inode); |
fa964540 | 458 | return 0; |
dbe89444 DW |
459 | } |
460 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
461 | ||
462 | if (t->dt_checksum != ext4_dx_csum(inode, dirent, count_offset, | |
463 | count, t)) | |
464 | return 0; | |
465 | return 1; | |
466 | } | |
467 | ||
468 | static void ext4_dx_csum_set(struct inode *inode, struct ext4_dir_entry *dirent) | |
469 | { | |
470 | struct dx_countlimit *c; | |
471 | struct dx_tail *t; | |
472 | int count_offset, limit, count; | |
473 | ||
9aa5d32b | 474 | if (!ext4_has_metadata_csum(inode->i_sb)) |
dbe89444 DW |
475 | return; |
476 | ||
477 | c = get_dx_countlimit(inode, dirent, &count_offset); | |
478 | if (!c) { | |
479 | EXT4_ERROR_INODE(inode, "dir seems corrupt? Run e2fsck -D."); | |
480 | return; | |
481 | } | |
482 | limit = le16_to_cpu(c->limit); | |
483 | count = le16_to_cpu(c->count); | |
484 | if (count_offset + (limit * sizeof(struct dx_entry)) > | |
485 | EXT4_BLOCK_SIZE(inode->i_sb) - sizeof(struct dx_tail)) { | |
dffe9d8d | 486 | warn_no_space_for_csum(inode); |
dbe89444 DW |
487 | return; |
488 | } | |
489 | t = (struct dx_tail *)(((struct dx_entry *)c) + limit); | |
490 | ||
491 | t->dt_checksum = ext4_dx_csum(inode, dirent, count_offset, count, t); | |
492 | } | |
493 | ||
494 | static inline int ext4_handle_dirty_dx_node(handle_t *handle, | |
495 | struct inode *inode, | |
496 | struct buffer_head *bh) | |
497 | { | |
498 | ext4_dx_csum_set(inode, (struct ext4_dir_entry *)bh->b_data); | |
499 | return ext4_handle_dirty_metadata(handle, inode, bh); | |
500 | } | |
501 | ||
f795e140 LZ |
502 | /* |
503 | * p is at least 6 bytes before the end of page | |
504 | */ | |
505 | static inline struct ext4_dir_entry_2 * | |
3d0518f4 | 506 | ext4_next_entry(struct ext4_dir_entry_2 *p, unsigned long blocksize) |
f795e140 LZ |
507 | { |
508 | return (struct ext4_dir_entry_2 *)((char *)p + | |
3d0518f4 | 509 | ext4_rec_len_from_disk(p->rec_len, blocksize)); |
f795e140 LZ |
510 | } |
511 | ||
ac27a0ec DK |
512 | /* |
513 | * Future: use high four bits of block for coalesce-on-delete flags | |
514 | * Mask them off for now. | |
515 | */ | |
516 | ||
725d26d3 | 517 | static inline ext4_lblk_t dx_get_block(struct dx_entry *entry) |
ac27a0ec | 518 | { |
e08ac99f | 519 | return le32_to_cpu(entry->block) & 0x0fffffff; |
ac27a0ec DK |
520 | } |
521 | ||
725d26d3 | 522 | static inline void dx_set_block(struct dx_entry *entry, ext4_lblk_t value) |
ac27a0ec DK |
523 | { |
524 | entry->block = cpu_to_le32(value); | |
525 | } | |
526 | ||
af5bc92d | 527 | static inline unsigned dx_get_hash(struct dx_entry *entry) |
ac27a0ec DK |
528 | { |
529 | return le32_to_cpu(entry->hash); | |
530 | } | |
531 | ||
af5bc92d | 532 | static inline void dx_set_hash(struct dx_entry *entry, unsigned value) |
ac27a0ec DK |
533 | { |
534 | entry->hash = cpu_to_le32(value); | |
535 | } | |
536 | ||
af5bc92d | 537 | static inline unsigned dx_get_count(struct dx_entry *entries) |
ac27a0ec DK |
538 | { |
539 | return le16_to_cpu(((struct dx_countlimit *) entries)->count); | |
540 | } | |
541 | ||
af5bc92d | 542 | static inline unsigned dx_get_limit(struct dx_entry *entries) |
ac27a0ec DK |
543 | { |
544 | return le16_to_cpu(((struct dx_countlimit *) entries)->limit); | |
545 | } | |
546 | ||
af5bc92d | 547 | static inline void dx_set_count(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
548 | { |
549 | ((struct dx_countlimit *) entries)->count = cpu_to_le16(value); | |
550 | } | |
551 | ||
af5bc92d | 552 | static inline void dx_set_limit(struct dx_entry *entries, unsigned value) |
ac27a0ec DK |
553 | { |
554 | ((struct dx_countlimit *) entries)->limit = cpu_to_le16(value); | |
555 | } | |
556 | ||
af5bc92d | 557 | static inline unsigned dx_root_limit(struct inode *dir, unsigned infosize) |
ac27a0ec | 558 | { |
617ba13b MC |
559 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(1) - |
560 | EXT4_DIR_REC_LEN(2) - infosize; | |
dbe89444 | 561 | |
9aa5d32b | 562 | if (ext4_has_metadata_csum(dir->i_sb)) |
dbe89444 | 563 | entry_space -= sizeof(struct dx_tail); |
d9c769b7 | 564 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
565 | } |
566 | ||
af5bc92d | 567 | static inline unsigned dx_node_limit(struct inode *dir) |
ac27a0ec | 568 | { |
617ba13b | 569 | unsigned entry_space = dir->i_sb->s_blocksize - EXT4_DIR_REC_LEN(0); |
dbe89444 | 570 | |
9aa5d32b | 571 | if (ext4_has_metadata_csum(dir->i_sb)) |
dbe89444 | 572 | entry_space -= sizeof(struct dx_tail); |
d9c769b7 | 573 | return entry_space / sizeof(struct dx_entry); |
ac27a0ec DK |
574 | } |
575 | ||
576 | /* | |
577 | * Debug | |
578 | */ | |
579 | #ifdef DX_DEBUG | |
4776004f | 580 | static void dx_show_index(char * label, struct dx_entry *entries) |
ac27a0ec | 581 | { |
63f57933 | 582 | int i, n = dx_get_count (entries); |
d74f3d25 | 583 | printk(KERN_DEBUG "%s index", label); |
63f57933 | 584 | for (i = 0; i < n; i++) { |
d74f3d25 JP |
585 | printk(KERN_CONT " %x->%lu", |
586 | i ? dx_get_hash(entries + i) : 0, | |
587 | (unsigned long)dx_get_block(entries + i)); | |
63f57933 | 588 | } |
d74f3d25 | 589 | printk(KERN_CONT "\n"); |
ac27a0ec DK |
590 | } |
591 | ||
592 | struct stats | |
593 | { | |
594 | unsigned names; | |
595 | unsigned space; | |
596 | unsigned bcount; | |
597 | }; | |
598 | ||
b3098486 MH |
599 | static struct stats dx_show_leaf(struct inode *dir, |
600 | struct dx_hash_info *hinfo, | |
601 | struct ext4_dir_entry_2 *de, | |
602 | int size, int show_names) | |
ac27a0ec DK |
603 | { |
604 | unsigned names = 0, space = 0; | |
605 | char *base = (char *) de; | |
606 | struct dx_hash_info h = *hinfo; | |
607 | ||
608 | printk("names: "); | |
609 | while ((char *) de < base + size) | |
610 | { | |
611 | if (de->inode) | |
612 | { | |
613 | if (show_names) | |
614 | { | |
643fa961 | 615 | #ifdef CONFIG_FS_ENCRYPTION |
b3098486 MH |
616 | int len; |
617 | char *name; | |
a7550b30 JK |
618 | struct fscrypt_str fname_crypto_str = |
619 | FSTR_INIT(NULL, 0); | |
c936e1ec | 620 | int res = 0; |
b3098486 MH |
621 | |
622 | name = de->name; | |
623 | len = de->name_len; | |
592ddec7 | 624 | if (IS_ENCRYPTED(dir)) |
a7550b30 | 625 | res = fscrypt_get_encryption_info(dir); |
b7236e21 TT |
626 | if (res) { |
627 | printk(KERN_WARNING "Error setting up" | |
628 | " fname crypto: %d\n", res); | |
b3098486 | 629 | } |
a7550b30 | 630 | if (!fscrypt_has_encryption_key(dir)) { |
b3098486 MH |
631 | /* Directory is not encrypted */ |
632 | ext4fs_dirhash(de->name, | |
633 | de->name_len, &h); | |
634 | printk("%*.s:(U)%x.%u ", len, | |
635 | name, h.hash, | |
636 | (unsigned) ((char *) de | |
637 | - base)); | |
638 | } else { | |
a7550b30 JK |
639 | struct fscrypt_str de_name = |
640 | FSTR_INIT(name, len); | |
641 | ||
b3098486 | 642 | /* Directory is encrypted */ |
a7550b30 JK |
643 | res = fscrypt_fname_alloc_buffer( |
644 | dir, len, | |
b3098486 | 645 | &fname_crypto_str); |
ef1eb3aa | 646 | if (res) |
b3098486 MH |
647 | printk(KERN_WARNING "Error " |
648 | "allocating crypto " | |
649 | "buffer--skipping " | |
650 | "crypto\n"); | |
a7550b30 JK |
651 | res = fscrypt_fname_disk_to_usr(dir, |
652 | 0, 0, &de_name, | |
653 | &fname_crypto_str); | |
ef1eb3aa | 654 | if (res) { |
b3098486 MH |
655 | printk(KERN_WARNING "Error " |
656 | "converting filename " | |
657 | "from disk to usr" | |
658 | "\n"); | |
659 | name = "??"; | |
660 | len = 2; | |
661 | } else { | |
662 | name = fname_crypto_str.name; | |
663 | len = fname_crypto_str.len; | |
664 | } | |
5de0b4d0 TT |
665 | ext4fs_dirhash(de->name, de->name_len, |
666 | &h); | |
b3098486 MH |
667 | printk("%*.s:(E)%x.%u ", len, name, |
668 | h.hash, (unsigned) ((char *) de | |
669 | - base)); | |
a7550b30 JK |
670 | fscrypt_fname_free_buffer( |
671 | &fname_crypto_str); | |
b3098486 MH |
672 | } |
673 | #else | |
ac27a0ec DK |
674 | int len = de->name_len; |
675 | char *name = de->name; | |
617ba13b | 676 | ext4fs_dirhash(de->name, de->name_len, &h); |
b3098486 | 677 | printk("%*.s:%x.%u ", len, name, h.hash, |
265c6a0f | 678 | (unsigned) ((char *) de - base)); |
b3098486 | 679 | #endif |
ac27a0ec | 680 | } |
617ba13b | 681 | space += EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
682 | names++; |
683 | } | |
3d0518f4 | 684 | de = ext4_next_entry(de, size); |
ac27a0ec | 685 | } |
d74f3d25 | 686 | printk(KERN_CONT "(%i)\n", names); |
ac27a0ec DK |
687 | return (struct stats) { names, space, 1 }; |
688 | } | |
689 | ||
690 | struct stats dx_show_entries(struct dx_hash_info *hinfo, struct inode *dir, | |
691 | struct dx_entry *entries, int levels) | |
692 | { | |
693 | unsigned blocksize = dir->i_sb->s_blocksize; | |
af5bc92d | 694 | unsigned count = dx_get_count(entries), names = 0, space = 0, i; |
ac27a0ec DK |
695 | unsigned bcount = 0; |
696 | struct buffer_head *bh; | |
ac27a0ec DK |
697 | printk("%i indexed blocks...\n", count); |
698 | for (i = 0; i < count; i++, entries++) | |
699 | { | |
725d26d3 AK |
700 | ext4_lblk_t block = dx_get_block(entries); |
701 | ext4_lblk_t hash = i ? dx_get_hash(entries): 0; | |
ac27a0ec DK |
702 | u32 range = i < count - 1? (dx_get_hash(entries + 1) - hash): ~hash; |
703 | struct stats stats; | |
704 | printk("%s%3u:%03u hash %8x/%8x ",levels?"":" ", i, block, hash, range); | |
1c215028 TT |
705 | bh = ext4_bread(NULL,dir, block, 0); |
706 | if (!bh || IS_ERR(bh)) | |
707 | continue; | |
ac27a0ec DK |
708 | stats = levels? |
709 | dx_show_entries(hinfo, dir, ((struct dx_node *) bh->b_data)->entries, levels - 1): | |
b3098486 MH |
710 | dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) |
711 | bh->b_data, blocksize, 0); | |
ac27a0ec DK |
712 | names += stats.names; |
713 | space += stats.space; | |
714 | bcount += stats.bcount; | |
af5bc92d | 715 | brelse(bh); |
ac27a0ec DK |
716 | } |
717 | if (bcount) | |
60e6679e | 718 | printk(KERN_DEBUG "%snames %u, fullness %u (%u%%)\n", |
4776004f TT |
719 | levels ? "" : " ", names, space/bcount, |
720 | (space/bcount)*100/blocksize); | |
ac27a0ec DK |
721 | return (struct stats) { names, space, bcount}; |
722 | } | |
723 | #endif /* DX_DEBUG */ | |
724 | ||
725 | /* | |
726 | * Probe for a directory leaf block to search. | |
727 | * | |
728 | * dx_probe can return ERR_BAD_DX_DIR, which means there was a format | |
729 | * error in the directory index, and the caller should fall back to | |
730 | * searching the directory normally. The callers of dx_probe **MUST** | |
731 | * check for this error code, and make sure it never gets reflected | |
732 | * back to userspace. | |
733 | */ | |
734 | static struct dx_frame * | |
5b643f9c | 735 | dx_probe(struct ext4_filename *fname, struct inode *dir, |
dd73b5d5 | 736 | struct dx_hash_info *hinfo, struct dx_frame *frame_in) |
ac27a0ec DK |
737 | { |
738 | unsigned count, indirect; | |
739 | struct dx_entry *at, *entries, *p, *q, *m; | |
740 | struct dx_root *root; | |
ac27a0ec | 741 | struct dx_frame *frame = frame_in; |
dd73b5d5 | 742 | struct dx_frame *ret_err = ERR_PTR(ERR_BAD_DX_DIR); |
ac27a0ec DK |
743 | u32 hash; |
744 | ||
e08ac99f | 745 | memset(frame_in, 0, EXT4_HTREE_LEVEL * sizeof(frame_in[0])); |
dd73b5d5 TT |
746 | frame->bh = ext4_read_dirblock(dir, 0, INDEX); |
747 | if (IS_ERR(frame->bh)) | |
748 | return (struct dx_frame *) frame->bh; | |
749 | ||
750 | root = (struct dx_root *) frame->bh->b_data; | |
ac27a0ec DK |
751 | if (root->info.hash_version != DX_HASH_TEA && |
752 | root->info.hash_version != DX_HASH_HALF_MD4 && | |
753 | root->info.hash_version != DX_HASH_LEGACY) { | |
b03a2f7e AD |
754 | ext4_warning_inode(dir, "Unrecognised inode hash code %u", |
755 | root->info.hash_version); | |
ac27a0ec DK |
756 | goto fail; |
757 | } | |
5b643f9c TT |
758 | if (fname) |
759 | hinfo = &fname->hinfo; | |
ac27a0ec | 760 | hinfo->hash_version = root->info.hash_version; |
f99b2589 TT |
761 | if (hinfo->hash_version <= DX_HASH_TEA) |
762 | hinfo->hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 763 | hinfo->seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
5b643f9c TT |
764 | if (fname && fname_name(fname)) |
765 | ext4fs_dirhash(fname_name(fname), fname_len(fname), hinfo); | |
ac27a0ec DK |
766 | hash = hinfo->hash; |
767 | ||
768 | if (root->info.unused_flags & 1) { | |
b03a2f7e AD |
769 | ext4_warning_inode(dir, "Unimplemented hash flags: %#06x", |
770 | root->info.unused_flags); | |
ac27a0ec DK |
771 | goto fail; |
772 | } | |
773 | ||
b03a2f7e | 774 | indirect = root->info.indirect_levels; |
e08ac99f AB |
775 | if (indirect >= ext4_dir_htree_level(dir->i_sb)) { |
776 | ext4_warning(dir->i_sb, | |
777 | "Directory (ino: %lu) htree depth %#06x exceed" | |
778 | "supported value", dir->i_ino, | |
779 | ext4_dir_htree_level(dir->i_sb)); | |
780 | if (ext4_dir_htree_level(dir->i_sb) < EXT4_HTREE_LEVEL) { | |
781 | ext4_warning(dir->i_sb, "Enable large directory " | |
782 | "feature to access it"); | |
783 | } | |
ac27a0ec DK |
784 | goto fail; |
785 | } | |
786 | ||
b03a2f7e AD |
787 | entries = (struct dx_entry *)(((char *)&root->info) + |
788 | root->info.info_length); | |
3d82abae ES |
789 | |
790 | if (dx_get_limit(entries) != dx_root_limit(dir, | |
791 | root->info.info_length)) { | |
b03a2f7e AD |
792 | ext4_warning_inode(dir, "dx entry: limit %u != root limit %u", |
793 | dx_get_limit(entries), | |
794 | dx_root_limit(dir, root->info.info_length)); | |
3d82abae ES |
795 | goto fail; |
796 | } | |
797 | ||
af5bc92d | 798 | dxtrace(printk("Look up %x", hash)); |
dd73b5d5 | 799 | while (1) { |
ac27a0ec | 800 | count = dx_get_count(entries); |
3d82abae | 801 | if (!count || count > dx_get_limit(entries)) { |
b03a2f7e AD |
802 | ext4_warning_inode(dir, |
803 | "dx entry: count %u beyond limit %u", | |
804 | count, dx_get_limit(entries)); | |
dd73b5d5 | 805 | goto fail; |
3d82abae ES |
806 | } |
807 | ||
ac27a0ec DK |
808 | p = entries + 1; |
809 | q = entries + count - 1; | |
dd73b5d5 | 810 | while (p <= q) { |
b03a2f7e | 811 | m = p + (q - p) / 2; |
d74f3d25 | 812 | dxtrace(printk(KERN_CONT ".")); |
ac27a0ec DK |
813 | if (dx_get_hash(m) > hash) |
814 | q = m - 1; | |
815 | else | |
816 | p = m + 1; | |
817 | } | |
818 | ||
dd73b5d5 | 819 | if (0) { // linear search cross check |
ac27a0ec DK |
820 | unsigned n = count - 1; |
821 | at = entries; | |
822 | while (n--) | |
823 | { | |
d74f3d25 | 824 | dxtrace(printk(KERN_CONT ",")); |
ac27a0ec DK |
825 | if (dx_get_hash(++at) > hash) |
826 | { | |
827 | at--; | |
828 | break; | |
829 | } | |
830 | } | |
831 | assert (at == p - 1); | |
832 | } | |
833 | ||
834 | at = p - 1; | |
d74f3d25 JP |
835 | dxtrace(printk(KERN_CONT " %x->%u\n", |
836 | at == entries ? 0 : dx_get_hash(at), | |
b03a2f7e | 837 | dx_get_block(at))); |
ac27a0ec DK |
838 | frame->entries = entries; |
839 | frame->at = at; | |
dd73b5d5 TT |
840 | if (!indirect--) |
841 | return frame; | |
842 | frame++; | |
843 | frame->bh = ext4_read_dirblock(dir, dx_get_block(at), INDEX); | |
844 | if (IS_ERR(frame->bh)) { | |
845 | ret_err = (struct dx_frame *) frame->bh; | |
846 | frame->bh = NULL; | |
847 | goto fail; | |
dbe89444 | 848 | } |
dd73b5d5 | 849 | entries = ((struct dx_node *) frame->bh->b_data)->entries; |
dbe89444 | 850 | |
b03a2f7e AD |
851 | if (dx_get_limit(entries) != dx_node_limit(dir)) { |
852 | ext4_warning_inode(dir, | |
853 | "dx entry: limit %u != node limit %u", | |
854 | dx_get_limit(entries), dx_node_limit(dir)); | |
dd73b5d5 | 855 | goto fail; |
3d82abae | 856 | } |
ac27a0ec | 857 | } |
dd73b5d5 | 858 | fail: |
ac27a0ec DK |
859 | while (frame >= frame_in) { |
860 | brelse(frame->bh); | |
861 | frame--; | |
862 | } | |
b3098486 | 863 | |
dd73b5d5 | 864 | if (ret_err == ERR_PTR(ERR_BAD_DX_DIR)) |
b03a2f7e AD |
865 | ext4_warning_inode(dir, |
866 | "Corrupt directory, running e2fsck is recommended"); | |
dd73b5d5 | 867 | return ret_err; |
ac27a0ec DK |
868 | } |
869 | ||
b03a2f7e | 870 | static void dx_release(struct dx_frame *frames) |
ac27a0ec | 871 | { |
e08ac99f AB |
872 | struct dx_root_info *info; |
873 | int i; | |
874 | ||
ac27a0ec DK |
875 | if (frames[0].bh == NULL) |
876 | return; | |
877 | ||
e08ac99f AB |
878 | info = &((struct dx_root *)frames[0].bh->b_data)->info; |
879 | for (i = 0; i <= info->indirect_levels; i++) { | |
880 | if (frames[i].bh == NULL) | |
881 | break; | |
882 | brelse(frames[i].bh); | |
883 | frames[i].bh = NULL; | |
884 | } | |
ac27a0ec DK |
885 | } |
886 | ||
887 | /* | |
888 | * This function increments the frame pointer to search the next leaf | |
889 | * block, and reads in the necessary intervening nodes if the search | |
890 | * should be necessary. Whether or not the search is necessary is | |
891 | * controlled by the hash parameter. If the hash value is even, then | |
892 | * the search is only continued if the next block starts with that | |
893 | * hash value. This is used if we are searching for a specific file. | |
894 | * | |
895 | * If the hash value is HASH_NB_ALWAYS, then always go to the next block. | |
896 | * | |
897 | * This function returns 1 if the caller should continue to search, | |
898 | * or 0 if it should not. If there is an error reading one of the | |
899 | * index blocks, it will a negative error code. | |
900 | * | |
901 | * If start_hash is non-null, it will be filled in with the starting | |
902 | * hash of the next page. | |
903 | */ | |
617ba13b | 904 | static int ext4_htree_next_block(struct inode *dir, __u32 hash, |
ac27a0ec DK |
905 | struct dx_frame *frame, |
906 | struct dx_frame *frames, | |
907 | __u32 *start_hash) | |
908 | { | |
909 | struct dx_frame *p; | |
910 | struct buffer_head *bh; | |
dc6982ff | 911 | int num_frames = 0; |
ac27a0ec DK |
912 | __u32 bhash; |
913 | ||
914 | p = frame; | |
915 | /* | |
916 | * Find the next leaf page by incrementing the frame pointer. | |
917 | * If we run out of entries in the interior node, loop around and | |
918 | * increment pointer in the parent node. When we break out of | |
919 | * this loop, num_frames indicates the number of interior | |
920 | * nodes need to be read. | |
921 | */ | |
922 | while (1) { | |
923 | if (++(p->at) < p->entries + dx_get_count(p->entries)) | |
924 | break; | |
925 | if (p == frames) | |
926 | return 0; | |
927 | num_frames++; | |
928 | p--; | |
929 | } | |
930 | ||
931 | /* | |
932 | * If the hash is 1, then continue only if the next page has a | |
933 | * continuation hash of any value. This is used for readdir | |
934 | * handling. Otherwise, check to see if the hash matches the | |
935 | * desired contiuation hash. If it doesn't, return since | |
936 | * there's no point to read in the successive index pages. | |
937 | */ | |
938 | bhash = dx_get_hash(p->at); | |
939 | if (start_hash) | |
940 | *start_hash = bhash; | |
941 | if ((hash & 1) == 0) { | |
942 | if ((bhash & ~1) != hash) | |
943 | return 0; | |
944 | } | |
945 | /* | |
946 | * If the hash is HASH_NB_ALWAYS, we always go to the next | |
947 | * block so no check is necessary | |
948 | */ | |
949 | while (num_frames--) { | |
dc6982ff TT |
950 | bh = ext4_read_dirblock(dir, dx_get_block(p->at), INDEX); |
951 | if (IS_ERR(bh)) | |
952 | return PTR_ERR(bh); | |
ac27a0ec | 953 | p++; |
af5bc92d | 954 | brelse(p->bh); |
ac27a0ec DK |
955 | p->bh = bh; |
956 | p->at = p->entries = ((struct dx_node *) bh->b_data)->entries; | |
957 | } | |
958 | return 1; | |
959 | } | |
960 | ||
961 | ||
ac27a0ec DK |
962 | /* |
963 | * This function fills a red-black tree with information from a | |
964 | * directory block. It returns the number directory entries loaded | |
965 | * into the tree. If there is an error it is returned in err. | |
966 | */ | |
967 | static int htree_dirblock_to_tree(struct file *dir_file, | |
725d26d3 | 968 | struct inode *dir, ext4_lblk_t block, |
ac27a0ec DK |
969 | struct dx_hash_info *hinfo, |
970 | __u32 start_hash, __u32 start_minor_hash) | |
971 | { | |
972 | struct buffer_head *bh; | |
617ba13b | 973 | struct ext4_dir_entry_2 *de, *top; |
90b0a973 | 974 | int err = 0, count = 0; |
a7550b30 | 975 | struct fscrypt_str fname_crypto_str = FSTR_INIT(NULL, 0), tmp_str; |
ac27a0ec | 976 | |
725d26d3 AK |
977 | dxtrace(printk(KERN_INFO "In htree dirblock_to_tree: block %lu\n", |
978 | (unsigned long)block)); | |
dc6982ff TT |
979 | bh = ext4_read_dirblock(dir, block, DIRENT); |
980 | if (IS_ERR(bh)) | |
981 | return PTR_ERR(bh); | |
b0336e8d | 982 | |
617ba13b MC |
983 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
984 | top = (struct ext4_dir_entry_2 *) ((char *) de + | |
ac27a0ec | 985 | dir->i_sb->s_blocksize - |
617ba13b | 986 | EXT4_DIR_REC_LEN(0)); |
643fa961 | 987 | #ifdef CONFIG_FS_ENCRYPTION |
1f3862b5 | 988 | /* Check if the directory is encrypted */ |
592ddec7 | 989 | if (IS_ENCRYPTED(dir)) { |
a7550b30 | 990 | err = fscrypt_get_encryption_info(dir); |
c936e1ec TT |
991 | if (err < 0) { |
992 | brelse(bh); | |
993 | return err; | |
994 | } | |
a7550b30 | 995 | err = fscrypt_fname_alloc_buffer(dir, EXT4_NAME_LEN, |
1f3862b5 MH |
996 | &fname_crypto_str); |
997 | if (err < 0) { | |
1f3862b5 MH |
998 | brelse(bh); |
999 | return err; | |
1000 | } | |
1001 | } | |
1002 | #endif | |
3d0518f4 | 1003 | for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) { |
f7c21177 | 1004 | if (ext4_check_dir_entry(dir, NULL, de, bh, |
226ba972 | 1005 | bh->b_data, bh->b_size, |
cad3f007 TT |
1006 | (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb)) |
1007 | + ((char *)de - bh->b_data))) { | |
64cb9273 AV |
1008 | /* silently ignore the rest of the block */ |
1009 | break; | |
e6c40211 | 1010 | } |
617ba13b | 1011 | ext4fs_dirhash(de->name, de->name_len, hinfo); |
ac27a0ec DK |
1012 | if ((hinfo->hash < start_hash) || |
1013 | ((hinfo->hash == start_hash) && | |
1014 | (hinfo->minor_hash < start_minor_hash))) | |
1015 | continue; | |
1016 | if (de->inode == 0) | |
1017 | continue; | |
592ddec7 | 1018 | if (!IS_ENCRYPTED(dir)) { |
1f3862b5 MH |
1019 | tmp_str.name = de->name; |
1020 | tmp_str.len = de->name_len; | |
1021 | err = ext4_htree_store_dirent(dir_file, | |
1022 | hinfo->hash, hinfo->minor_hash, de, | |
1023 | &tmp_str); | |
1024 | } else { | |
d2299590 | 1025 | int save_len = fname_crypto_str.len; |
a7550b30 JK |
1026 | struct fscrypt_str de_name = FSTR_INIT(de->name, |
1027 | de->name_len); | |
d2299590 | 1028 | |
1f3862b5 | 1029 | /* Directory is encrypted */ |
a7550b30 JK |
1030 | err = fscrypt_fname_disk_to_usr(dir, hinfo->hash, |
1031 | hinfo->minor_hash, &de_name, | |
1032 | &fname_crypto_str); | |
ef1eb3aa | 1033 | if (err) { |
1f3862b5 MH |
1034 | count = err; |
1035 | goto errout; | |
1036 | } | |
1037 | err = ext4_htree_store_dirent(dir_file, | |
1038 | hinfo->hash, hinfo->minor_hash, de, | |
1039 | &fname_crypto_str); | |
d2299590 | 1040 | fname_crypto_str.len = save_len; |
1f3862b5 | 1041 | } |
2f61830a | 1042 | if (err != 0) { |
1f3862b5 MH |
1043 | count = err; |
1044 | goto errout; | |
ac27a0ec DK |
1045 | } |
1046 | count++; | |
1047 | } | |
1f3862b5 | 1048 | errout: |
ac27a0ec | 1049 | brelse(bh); |
643fa961 | 1050 | #ifdef CONFIG_FS_ENCRYPTION |
a7550b30 | 1051 | fscrypt_fname_free_buffer(&fname_crypto_str); |
1f3862b5 | 1052 | #endif |
ac27a0ec DK |
1053 | return count; |
1054 | } | |
1055 | ||
1056 | ||
1057 | /* | |
1058 | * This function fills a red-black tree with information from a | |
1059 | * directory. We start scanning the directory in hash order, starting | |
1060 | * at start_hash and start_minor_hash. | |
1061 | * | |
1062 | * This function returns the number of entries inserted into the tree, | |
1063 | * or a negative error code. | |
1064 | */ | |
617ba13b | 1065 | int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, |
ac27a0ec DK |
1066 | __u32 start_minor_hash, __u32 *next_hash) |
1067 | { | |
1068 | struct dx_hash_info hinfo; | |
617ba13b | 1069 | struct ext4_dir_entry_2 *de; |
e08ac99f | 1070 | struct dx_frame frames[EXT4_HTREE_LEVEL], *frame; |
ac27a0ec | 1071 | struct inode *dir; |
725d26d3 | 1072 | ext4_lblk_t block; |
ac27a0ec | 1073 | int count = 0; |
725d26d3 | 1074 | int ret, err; |
ac27a0ec | 1075 | __u32 hashval; |
a7550b30 | 1076 | struct fscrypt_str tmp_str; |
ac27a0ec | 1077 | |
60e6679e | 1078 | dxtrace(printk(KERN_DEBUG "In htree_fill_tree, start hash: %x:%x\n", |
4776004f | 1079 | start_hash, start_minor_hash)); |
496ad9aa | 1080 | dir = file_inode(dir_file); |
12e9b892 | 1081 | if (!(ext4_test_inode_flag(dir, EXT4_INODE_INDEX))) { |
617ba13b | 1082 | hinfo.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
f99b2589 TT |
1083 | if (hinfo.hash_version <= DX_HASH_TEA) |
1084 | hinfo.hash_version += | |
1085 | EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
617ba13b | 1086 | hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; |
8af0f082 TM |
1087 | if (ext4_has_inline_data(dir)) { |
1088 | int has_inline_data = 1; | |
1089 | count = htree_inlinedir_to_tree(dir_file, dir, 0, | |
1090 | &hinfo, start_hash, | |
1091 | start_minor_hash, | |
1092 | &has_inline_data); | |
1093 | if (has_inline_data) { | |
1094 | *next_hash = ~0; | |
1095 | return count; | |
1096 | } | |
1097 | } | |
ac27a0ec DK |
1098 | count = htree_dirblock_to_tree(dir_file, dir, 0, &hinfo, |
1099 | start_hash, start_minor_hash); | |
1100 | *next_hash = ~0; | |
1101 | return count; | |
1102 | } | |
1103 | hinfo.hash = start_hash; | |
1104 | hinfo.minor_hash = 0; | |
dd73b5d5 TT |
1105 | frame = dx_probe(NULL, dir, &hinfo, frames); |
1106 | if (IS_ERR(frame)) | |
1107 | return PTR_ERR(frame); | |
ac27a0ec DK |
1108 | |
1109 | /* Add '.' and '..' from the htree header */ | |
1110 | if (!start_hash && !start_minor_hash) { | |
617ba13b | 1111 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
2f61830a TT |
1112 | tmp_str.name = de->name; |
1113 | tmp_str.len = de->name_len; | |
1114 | err = ext4_htree_store_dirent(dir_file, 0, 0, | |
1115 | de, &tmp_str); | |
1116 | if (err != 0) | |
ac27a0ec DK |
1117 | goto errout; |
1118 | count++; | |
1119 | } | |
1120 | if (start_hash < 2 || (start_hash ==2 && start_minor_hash==0)) { | |
617ba13b | 1121 | de = (struct ext4_dir_entry_2 *) frames[0].bh->b_data; |
3d0518f4 | 1122 | de = ext4_next_entry(de, dir->i_sb->s_blocksize); |
2f61830a TT |
1123 | tmp_str.name = de->name; |
1124 | tmp_str.len = de->name_len; | |
1125 | err = ext4_htree_store_dirent(dir_file, 2, 0, | |
1126 | de, &tmp_str); | |
1127 | if (err != 0) | |
ac27a0ec DK |
1128 | goto errout; |
1129 | count++; | |
1130 | } | |
1131 | ||
1132 | while (1) { | |
1f60fbe7 TT |
1133 | if (fatal_signal_pending(current)) { |
1134 | err = -ERESTARTSYS; | |
1135 | goto errout; | |
1136 | } | |
1137 | cond_resched(); | |
ac27a0ec DK |
1138 | block = dx_get_block(frame->at); |
1139 | ret = htree_dirblock_to_tree(dir_file, dir, block, &hinfo, | |
1140 | start_hash, start_minor_hash); | |
1141 | if (ret < 0) { | |
1142 | err = ret; | |
1143 | goto errout; | |
1144 | } | |
1145 | count += ret; | |
1146 | hashval = ~0; | |
617ba13b | 1147 | ret = ext4_htree_next_block(dir, HASH_NB_ALWAYS, |
ac27a0ec DK |
1148 | frame, frames, &hashval); |
1149 | *next_hash = hashval; | |
1150 | if (ret < 0) { | |
1151 | err = ret; | |
1152 | goto errout; | |
1153 | } | |
1154 | /* | |
1155 | * Stop if: (a) there are no more entries, or | |
1156 | * (b) we have inserted at least one entry and the | |
1157 | * next hash value is not a continuation | |
1158 | */ | |
1159 | if ((ret == 0) || | |
1160 | (count && ((hashval & 1) == 0))) | |
1161 | break; | |
1162 | } | |
1163 | dx_release(frames); | |
4776004f TT |
1164 | dxtrace(printk(KERN_DEBUG "Fill tree: returned %d entries, " |
1165 | "next hash: %x\n", count, *next_hash)); | |
ac27a0ec DK |
1166 | return count; |
1167 | errout: | |
1168 | dx_release(frames); | |
1169 | return (err); | |
1170 | } | |
1171 | ||
7335cd3b TM |
1172 | static inline int search_dirblock(struct buffer_head *bh, |
1173 | struct inode *dir, | |
5b643f9c | 1174 | struct ext4_filename *fname, |
7335cd3b TM |
1175 | unsigned int offset, |
1176 | struct ext4_dir_entry_2 **res_dir) | |
1177 | { | |
5b643f9c | 1178 | return ext4_search_dir(bh, bh->b_data, dir->i_sb->s_blocksize, dir, |
d6b97550 | 1179 | fname, offset, res_dir); |
7335cd3b TM |
1180 | } |
1181 | ||
ac27a0ec DK |
1182 | /* |
1183 | * Directory block splitting, compacting | |
1184 | */ | |
1185 | ||
ef2b02d3 ES |
1186 | /* |
1187 | * Create map of hash values, offsets, and sizes, stored at end of block. | |
1188 | * Returns number of entries mapped. | |
1189 | */ | |
1f3862b5 MH |
1190 | static int dx_make_map(struct inode *dir, struct ext4_dir_entry_2 *de, |
1191 | unsigned blocksize, struct dx_hash_info *hinfo, | |
8bad4597 | 1192 | struct dx_map_entry *map_tail) |
ac27a0ec DK |
1193 | { |
1194 | int count = 0; | |
1195 | char *base = (char *) de; | |
1196 | struct dx_hash_info h = *hinfo; | |
1197 | ||
8bad4597 | 1198 | while ((char *) de < base + blocksize) { |
ac27a0ec | 1199 | if (de->name_len && de->inode) { |
617ba13b | 1200 | ext4fs_dirhash(de->name, de->name_len, &h); |
ac27a0ec DK |
1201 | map_tail--; |
1202 | map_tail->hash = h.hash; | |
9aee2286 | 1203 | map_tail->offs = ((char *) de - base)>>2; |
ef2b02d3 | 1204 | map_tail->size = le16_to_cpu(de->rec_len); |
ac27a0ec DK |
1205 | count++; |
1206 | cond_resched(); | |
1207 | } | |
1208 | /* XXX: do we need to check rec_len == 0 case? -Chris */ | |
3d0518f4 | 1209 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
1210 | } |
1211 | return count; | |
1212 | } | |
1213 | ||
ef2b02d3 | 1214 | /* Sort map by hash value */ |
ac27a0ec DK |
1215 | static void dx_sort_map (struct dx_map_entry *map, unsigned count) |
1216 | { | |
63f57933 AM |
1217 | struct dx_map_entry *p, *q, *top = map + count - 1; |
1218 | int more; | |
1219 | /* Combsort until bubble sort doesn't suck */ | |
1220 | while (count > 2) { | |
1221 | count = count*10/13; | |
1222 | if (count - 9 < 2) /* 9, 10 -> 11 */ | |
1223 | count = 11; | |
1224 | for (p = top, q = p - count; q >= map; p--, q--) | |
1225 | if (p->hash < q->hash) | |
1226 | swap(*p, *q); | |
1227 | } | |
1228 | /* Garden variety bubble sort */ | |
1229 | do { | |
1230 | more = 0; | |
1231 | q = top; | |
1232 | while (q-- > map) { | |
1233 | if (q[1].hash >= q[0].hash) | |
ac27a0ec | 1234 | continue; |
63f57933 AM |
1235 | swap(*(q+1), *q); |
1236 | more = 1; | |
ac27a0ec DK |
1237 | } |
1238 | } while(more); | |
1239 | } | |
1240 | ||
725d26d3 | 1241 | static void dx_insert_block(struct dx_frame *frame, u32 hash, ext4_lblk_t block) |
ac27a0ec DK |
1242 | { |
1243 | struct dx_entry *entries = frame->entries; | |
1244 | struct dx_entry *old = frame->at, *new = old + 1; | |
1245 | int count = dx_get_count(entries); | |
1246 | ||
1247 | assert(count < dx_get_limit(entries)); | |
1248 | assert(old < entries + count); | |
1249 | memmove(new + 1, new, (char *)(entries + count) - (char *)(new)); | |
1250 | dx_set_hash(new, hash); | |
1251 | dx_set_block(new, block); | |
1252 | dx_set_count(entries, count + 1); | |
1253 | } | |
ac27a0ec | 1254 | |
ac27a0ec | 1255 | /* |
d9b9f8d5 | 1256 | * Test whether a directory entry matches the filename being searched for. |
ac27a0ec | 1257 | * |
d9b9f8d5 | 1258 | * Return: %true if the directory entry matches, otherwise %false. |
ac27a0ec | 1259 | */ |
d9b9f8d5 EB |
1260 | static inline bool ext4_match(const struct ext4_filename *fname, |
1261 | const struct ext4_dir_entry_2 *de) | |
ac27a0ec | 1262 | { |
067d1023 | 1263 | struct fscrypt_name f; |
1f3862b5 | 1264 | |
ac27a0ec | 1265 | if (!de->inode) |
d9b9f8d5 | 1266 | return false; |
1f3862b5 | 1267 | |
067d1023 EB |
1268 | f.usr_fname = fname->usr_fname; |
1269 | f.disk_name = fname->disk_name; | |
643fa961 | 1270 | #ifdef CONFIG_FS_ENCRYPTION |
067d1023 | 1271 | f.crypto_buf = fname->crypto_buf; |
1f3862b5 | 1272 | #endif |
067d1023 | 1273 | return fscrypt_match_name(&f, de->name, de->name_len); |
ac27a0ec DK |
1274 | } |
1275 | ||
1276 | /* | |
1277 | * Returns 0 if not found, -1 on failure, and 1 on success | |
1278 | */ | |
5b643f9c TT |
1279 | int ext4_search_dir(struct buffer_head *bh, char *search_buf, int buf_size, |
1280 | struct inode *dir, struct ext4_filename *fname, | |
5b643f9c | 1281 | unsigned int offset, struct ext4_dir_entry_2 **res_dir) |
ac27a0ec | 1282 | { |
617ba13b | 1283 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
1284 | char * dlimit; |
1285 | int de_len; | |
1f3862b5 | 1286 | |
7335cd3b TM |
1287 | de = (struct ext4_dir_entry_2 *)search_buf; |
1288 | dlimit = search_buf + buf_size; | |
ac27a0ec DK |
1289 | while ((char *) de < dlimit) { |
1290 | /* this code is executed quadratically often */ | |
1291 | /* do minimal checking `by hand' */ | |
d9b9f8d5 EB |
1292 | if ((char *) de + de->name_len <= dlimit && |
1293 | ext4_match(fname, de)) { | |
1294 | /* found a match - just to be sure, do | |
1295 | * a full check */ | |
1296 | if (ext4_check_dir_entry(dir, NULL, de, bh, bh->b_data, | |
1297 | bh->b_size, offset)) | |
1298 | return -1; | |
1299 | *res_dir = de; | |
1300 | return 1; | |
ac27a0ec DK |
1301 | } |
1302 | /* prevent looping on a bad block */ | |
3d0518f4 WY |
1303 | de_len = ext4_rec_len_from_disk(de->rec_len, |
1304 | dir->i_sb->s_blocksize); | |
d9b9f8d5 EB |
1305 | if (de_len <= 0) |
1306 | return -1; | |
ac27a0ec | 1307 | offset += de_len; |
617ba13b | 1308 | de = (struct ext4_dir_entry_2 *) ((char *) de + de_len); |
ac27a0ec | 1309 | } |
d9b9f8d5 | 1310 | return 0; |
ac27a0ec DK |
1311 | } |
1312 | ||
c6af8803 DW |
1313 | static int is_dx_internal_node(struct inode *dir, ext4_lblk_t block, |
1314 | struct ext4_dir_entry *de) | |
1315 | { | |
1316 | struct super_block *sb = dir->i_sb; | |
1317 | ||
1318 | if (!is_dx(dir)) | |
1319 | return 0; | |
1320 | if (block == 0) | |
1321 | return 1; | |
1322 | if (de->inode == 0 && | |
1323 | ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) == | |
1324 | sb->s_blocksize) | |
1325 | return 1; | |
1326 | return 0; | |
1327 | } | |
ac27a0ec DK |
1328 | |
1329 | /* | |
617ba13b | 1330 | * ext4_find_entry() |
ac27a0ec DK |
1331 | * |
1332 | * finds an entry in the specified directory with the wanted name. It | |
1333 | * returns the cache buffer in which the entry was found, and the entry | |
1334 | * itself (as a parameter - res_dir). It does NOT read the inode of the | |
1335 | * entry - you'll have to do that yourself if you want to. | |
1336 | * | |
1337 | * The returned buffer_head has ->b_count elevated. The caller is expected | |
1338 | * to brelse() it when appropriate. | |
1339 | */ | |
f702ba0f TT |
1340 | static struct buffer_head * ext4_find_entry (struct inode *dir, |
1341 | const struct qstr *d_name, | |
32f7f22c TM |
1342 | struct ext4_dir_entry_2 **res_dir, |
1343 | int *inlined) | |
ac27a0ec | 1344 | { |
af5bc92d TT |
1345 | struct super_block *sb; |
1346 | struct buffer_head *bh_use[NAMEI_RA_SIZE]; | |
1347 | struct buffer_head *bh, *ret = NULL; | |
9699d4f9 | 1348 | ext4_lblk_t start, block; |
8941ec8b | 1349 | const u8 *name = d_name->name; |
9699d4f9 | 1350 | size_t ra_max = 0; /* Number of bh's in the readahead |
ac27a0ec | 1351 | buffer, bh_use[] */ |
9699d4f9 | 1352 | size_t ra_ptr = 0; /* Current index into readahead |
ac27a0ec | 1353 | buffer */ |
725d26d3 | 1354 | ext4_lblk_t nblocks; |
5b643f9c TT |
1355 | int i, namelen, retval; |
1356 | struct ext4_filename fname; | |
ac27a0ec DK |
1357 | |
1358 | *res_dir = NULL; | |
1359 | sb = dir->i_sb; | |
f702ba0f | 1360 | namelen = d_name->len; |
617ba13b | 1361 | if (namelen > EXT4_NAME_LEN) |
ac27a0ec | 1362 | return NULL; |
e8e948e7 | 1363 | |
5b643f9c | 1364 | retval = ext4_fname_setup_filename(dir, d_name, 1, &fname); |
54475f53 EB |
1365 | if (retval == -ENOENT) |
1366 | return NULL; | |
5b643f9c TT |
1367 | if (retval) |
1368 | return ERR_PTR(retval); | |
1369 | ||
e8e948e7 TM |
1370 | if (ext4_has_inline_data(dir)) { |
1371 | int has_inline_data = 1; | |
d6b97550 | 1372 | ret = ext4_find_inline_entry(dir, &fname, res_dir, |
e8e948e7 | 1373 | &has_inline_data); |
32f7f22c TM |
1374 | if (has_inline_data) { |
1375 | if (inlined) | |
1376 | *inlined = 1; | |
5b643f9c | 1377 | goto cleanup_and_exit; |
32f7f22c | 1378 | } |
e8e948e7 TM |
1379 | } |
1380 | ||
8941ec8b | 1381 | if ((namelen <= 2) && (name[0] == '.') && |
6d5c3aa8 | 1382 | (name[1] == '.' || name[1] == '\0')) { |
8941ec8b TT |
1383 | /* |
1384 | * "." or ".." will only be in the first block | |
1385 | * NFS may look up ".."; "." should be handled by the VFS | |
1386 | */ | |
1387 | block = start = 0; | |
1388 | nblocks = 1; | |
1389 | goto restart; | |
1390 | } | |
ac27a0ec | 1391 | if (is_dx(dir)) { |
5b643f9c | 1392 | ret = ext4_dx_find_entry(dir, &fname, res_dir); |
ac27a0ec DK |
1393 | /* |
1394 | * On success, or if the error was file not found, | |
1395 | * return. Otherwise, fall back to doing a search the | |
1396 | * old fashioned way. | |
1397 | */ | |
5b643f9c TT |
1398 | if (!IS_ERR(ret) || PTR_ERR(ret) != ERR_BAD_DX_DIR) |
1399 | goto cleanup_and_exit; | |
4776004f TT |
1400 | dxtrace(printk(KERN_DEBUG "ext4_find_entry: dx failed, " |
1401 | "falling back\n")); | |
f39b3f45 | 1402 | ret = NULL; |
ac27a0ec | 1403 | } |
617ba13b | 1404 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
9d5afec6 CR |
1405 | if (!nblocks) { |
1406 | ret = NULL; | |
1407 | goto cleanup_and_exit; | |
1408 | } | |
617ba13b | 1409 | start = EXT4_I(dir)->i_dir_start_lookup; |
ac27a0ec DK |
1410 | if (start >= nblocks) |
1411 | start = 0; | |
1412 | block = start; | |
1413 | restart: | |
1414 | do { | |
1415 | /* | |
1416 | * We deal with the read-ahead logic here. | |
1417 | */ | |
1418 | if (ra_ptr >= ra_max) { | |
1419 | /* Refill the readahead buffer */ | |
1420 | ra_ptr = 0; | |
9699d4f9 TE |
1421 | if (block < start) |
1422 | ra_max = start - block; | |
1423 | else | |
1424 | ra_max = nblocks - block; | |
1425 | ra_max = min(ra_max, ARRAY_SIZE(bh_use)); | |
1426 | retval = ext4_bread_batch(dir, block, ra_max, | |
1427 | false /* wait */, bh_use); | |
1428 | if (retval) { | |
1429 | ret = ERR_PTR(retval); | |
1430 | ra_max = 0; | |
1431 | goto cleanup_and_exit; | |
ac27a0ec DK |
1432 | } |
1433 | } | |
1434 | if ((bh = bh_use[ra_ptr++]) == NULL) | |
1435 | goto next; | |
1436 | wait_on_buffer(bh); | |
1437 | if (!buffer_uptodate(bh)) { | |
24676da4 TT |
1438 | EXT4_ERROR_INODE(dir, "reading directory lblock %lu", |
1439 | (unsigned long) block); | |
ac27a0ec | 1440 | brelse(bh); |
6febe6f2 KK |
1441 | ret = ERR_PTR(-EIO); |
1442 | goto cleanup_and_exit; | |
ac27a0ec | 1443 | } |
b0336e8d | 1444 | if (!buffer_verified(bh) && |
c6af8803 DW |
1445 | !is_dx_internal_node(dir, block, |
1446 | (struct ext4_dir_entry *)bh->b_data) && | |
b0336e8d DW |
1447 | !ext4_dirent_csum_verify(dir, |
1448 | (struct ext4_dir_entry *)bh->b_data)) { | |
1449 | EXT4_ERROR_INODE(dir, "checksumming directory " | |
1450 | "block %lu", (unsigned long)block); | |
1451 | brelse(bh); | |
bdddf342 TT |
1452 | ret = ERR_PTR(-EFSBADCRC); |
1453 | goto cleanup_and_exit; | |
b0336e8d DW |
1454 | } |
1455 | set_buffer_verified(bh); | |
d6b97550 | 1456 | i = search_dirblock(bh, dir, &fname, |
617ba13b | 1457 | block << EXT4_BLOCK_SIZE_BITS(sb), res_dir); |
ac27a0ec | 1458 | if (i == 1) { |
617ba13b | 1459 | EXT4_I(dir)->i_dir_start_lookup = block; |
ac27a0ec DK |
1460 | ret = bh; |
1461 | goto cleanup_and_exit; | |
1462 | } else { | |
1463 | brelse(bh); | |
1464 | if (i < 0) | |
1465 | goto cleanup_and_exit; | |
1466 | } | |
1467 | next: | |
1468 | if (++block >= nblocks) | |
1469 | block = 0; | |
1470 | } while (block != start); | |
1471 | ||
1472 | /* | |
1473 | * If the directory has grown while we were searching, then | |
1474 | * search the last part of the directory before giving up. | |
1475 | */ | |
1476 | block = nblocks; | |
617ba13b | 1477 | nblocks = dir->i_size >> EXT4_BLOCK_SIZE_BITS(sb); |
ac27a0ec DK |
1478 | if (block < nblocks) { |
1479 | start = 0; | |
1480 | goto restart; | |
1481 | } | |
1482 | ||
1483 | cleanup_and_exit: | |
1484 | /* Clean up the read-ahead blocks */ | |
1485 | for (; ra_ptr < ra_max; ra_ptr++) | |
af5bc92d | 1486 | brelse(bh_use[ra_ptr]); |
5b643f9c | 1487 | ext4_fname_free_filename(&fname); |
ac27a0ec DK |
1488 | return ret; |
1489 | } | |
1490 | ||
5b643f9c TT |
1491 | static struct buffer_head * ext4_dx_find_entry(struct inode *dir, |
1492 | struct ext4_filename *fname, | |
1493 | struct ext4_dir_entry_2 **res_dir) | |
ac27a0ec | 1494 | { |
8941ec8b | 1495 | struct super_block * sb = dir->i_sb; |
e08ac99f | 1496 | struct dx_frame frames[EXT4_HTREE_LEVEL], *frame; |
ac27a0ec | 1497 | struct buffer_head *bh; |
725d26d3 | 1498 | ext4_lblk_t block; |
ac27a0ec | 1499 | int retval; |
ac27a0ec | 1500 | |
643fa961 | 1501 | #ifdef CONFIG_FS_ENCRYPTION |
1f3862b5 MH |
1502 | *res_dir = NULL; |
1503 | #endif | |
5b643f9c | 1504 | frame = dx_probe(fname, dir, NULL, frames); |
dd73b5d5 TT |
1505 | if (IS_ERR(frame)) |
1506 | return (struct buffer_head *) frame; | |
ac27a0ec DK |
1507 | do { |
1508 | block = dx_get_block(frame->at); | |
dc6982ff | 1509 | bh = ext4_read_dirblock(dir, block, DIRENT); |
537d8f93 | 1510 | if (IS_ERR(bh)) |
b0336e8d | 1511 | goto errout; |
537d8f93 | 1512 | |
d6b97550 | 1513 | retval = search_dirblock(bh, dir, fname, |
7845c049 TT |
1514 | block << EXT4_BLOCK_SIZE_BITS(sb), |
1515 | res_dir); | |
537d8f93 TT |
1516 | if (retval == 1) |
1517 | goto success; | |
af5bc92d | 1518 | brelse(bh); |
7845c049 | 1519 | if (retval == -1) { |
537d8f93 | 1520 | bh = ERR_PTR(ERR_BAD_DX_DIR); |
7845c049 TT |
1521 | goto errout; |
1522 | } | |
1523 | ||
ac27a0ec | 1524 | /* Check to see if we should continue to search */ |
5b643f9c | 1525 | retval = ext4_htree_next_block(dir, fname->hinfo.hash, frame, |
ac27a0ec DK |
1526 | frames, NULL); |
1527 | if (retval < 0) { | |
b03a2f7e AD |
1528 | ext4_warning_inode(dir, |
1529 | "error %d reading directory index block", | |
1530 | retval); | |
537d8f93 | 1531 | bh = ERR_PTR(retval); |
ac27a0ec DK |
1532 | goto errout; |
1533 | } | |
1534 | } while (retval == 1); | |
1535 | ||
537d8f93 | 1536 | bh = NULL; |
ac27a0ec | 1537 | errout: |
d6b97550 | 1538 | dxtrace(printk(KERN_DEBUG "%s not found\n", fname->usr_fname->name)); |
537d8f93 TT |
1539 | success: |
1540 | dx_release(frames); | |
1541 | return bh; | |
ac27a0ec | 1542 | } |
ac27a0ec | 1543 | |
00cd8dd3 | 1544 | static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, unsigned int flags) |
ac27a0ec | 1545 | { |
af5bc92d TT |
1546 | struct inode *inode; |
1547 | struct ext4_dir_entry_2 *de; | |
1548 | struct buffer_head *bh; | |
89904275 | 1549 | int err; |
ac27a0ec | 1550 | |
89904275 EB |
1551 | err = fscrypt_prepare_lookup(dir, dentry, flags); |
1552 | if (err) | |
1553 | return ERR_PTR(err); | |
28b4c263 | 1554 | |
89904275 EB |
1555 | if (dentry->d_name.len > EXT4_NAME_LEN) |
1556 | return ERR_PTR(-ENAMETOOLONG); | |
ac27a0ec | 1557 | |
32f7f22c | 1558 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 | 1559 | if (IS_ERR(bh)) |
e884bce1 | 1560 | return ERR_CAST(bh); |
ac27a0ec DK |
1561 | inode = NULL; |
1562 | if (bh) { | |
498e5f24 | 1563 | __u32 ino = le32_to_cpu(de->inode); |
af5bc92d | 1564 | brelse(bh); |
617ba13b | 1565 | if (!ext4_valid_inum(dir->i_sb, ino)) { |
24676da4 | 1566 | EXT4_ERROR_INODE(dir, "bad inode number: %u", ino); |
6a797d27 | 1567 | return ERR_PTR(-EFSCORRUPTED); |
a6c15c2b | 1568 | } |
7e936b73 | 1569 | if (unlikely(ino == dir->i_ino)) { |
a34e15cc DH |
1570 | EXT4_ERROR_INODE(dir, "'%pd' linked to parent dir", |
1571 | dentry); | |
6a797d27 | 1572 | return ERR_PTR(-EFSCORRUPTED); |
7e936b73 | 1573 | } |
8a363970 | 1574 | inode = ext4_iget(dir->i_sb, ino, EXT4_IGET_NORMAL); |
a9049376 AV |
1575 | if (inode == ERR_PTR(-ESTALE)) { |
1576 | EXT4_ERROR_INODE(dir, | |
1577 | "deleted inode referenced: %u", | |
1578 | ino); | |
6a797d27 | 1579 | return ERR_PTR(-EFSCORRUPTED); |
e6f009b0 | 1580 | } |
592ddec7 | 1581 | if (!IS_ERR(inode) && IS_ENCRYPTED(dir) && |
ff978b09 | 1582 | (S_ISDIR(inode->i_mode) || S_ISLNK(inode->i_mode)) && |
a7550b30 | 1583 | !fscrypt_has_permitted_context(dir, inode)) { |
d9cdc903 | 1584 | ext4_warning(inode->i_sb, |
8d2ae1cb | 1585 | "Inconsistent encryption contexts: %lu/%lu", |
8c68084b | 1586 | dir->i_ino, inode->i_ino); |
dd01b690 | 1587 | iput(inode); |
d9cdc903 TT |
1588 | return ERR_PTR(-EPERM); |
1589 | } | |
ac27a0ec DK |
1590 | } |
1591 | return d_splice_alias(inode, dentry); | |
1592 | } | |
1593 | ||
1594 | ||
617ba13b | 1595 | struct dentry *ext4_get_parent(struct dentry *child) |
ac27a0ec | 1596 | { |
498e5f24 | 1597 | __u32 ino; |
26fe5750 | 1598 | static const struct qstr dotdot = QSTR_INIT("..", 2); |
617ba13b | 1599 | struct ext4_dir_entry_2 * de; |
ac27a0ec DK |
1600 | struct buffer_head *bh; |
1601 | ||
2b0143b5 | 1602 | bh = ext4_find_entry(d_inode(child), &dotdot, &de, NULL); |
36de9286 | 1603 | if (IS_ERR(bh)) |
e884bce1 | 1604 | return ERR_CAST(bh); |
ac27a0ec DK |
1605 | if (!bh) |
1606 | return ERR_PTR(-ENOENT); | |
1607 | ino = le32_to_cpu(de->inode); | |
1608 | brelse(bh); | |
1609 | ||
fc64005c | 1610 | if (!ext4_valid_inum(child->d_sb, ino)) { |
2b0143b5 | 1611 | EXT4_ERROR_INODE(d_inode(child), |
24676da4 | 1612 | "bad parent inode number: %u", ino); |
6a797d27 | 1613 | return ERR_PTR(-EFSCORRUPTED); |
a6c15c2b VA |
1614 | } |
1615 | ||
8a363970 | 1616 | return d_obtain_alias(ext4_iget(child->d_sb, ino, EXT4_IGET_NORMAL)); |
ac27a0ec DK |
1617 | } |
1618 | ||
ef2b02d3 ES |
1619 | /* |
1620 | * Move count entries from end of map between two memory locations. | |
1621 | * Returns pointer to last entry moved. | |
1622 | */ | |
617ba13b | 1623 | static struct ext4_dir_entry_2 * |
3d0518f4 WY |
1624 | dx_move_dirents(char *from, char *to, struct dx_map_entry *map, int count, |
1625 | unsigned blocksize) | |
ac27a0ec DK |
1626 | { |
1627 | unsigned rec_len = 0; | |
1628 | ||
1629 | while (count--) { | |
60e6679e | 1630 | struct ext4_dir_entry_2 *de = (struct ext4_dir_entry_2 *) |
9aee2286 | 1631 | (from + (map->offs<<2)); |
617ba13b | 1632 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec | 1633 | memcpy (to, de, rec_len); |
617ba13b | 1634 | ((struct ext4_dir_entry_2 *) to)->rec_len = |
3d0518f4 | 1635 | ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec DK |
1636 | de->inode = 0; |
1637 | map++; | |
1638 | to += rec_len; | |
1639 | } | |
617ba13b | 1640 | return (struct ext4_dir_entry_2 *) (to - rec_len); |
ac27a0ec DK |
1641 | } |
1642 | ||
ef2b02d3 ES |
1643 | /* |
1644 | * Compact each dir entry in the range to the minimal rec_len. | |
1645 | * Returns pointer to last entry in range. | |
1646 | */ | |
8bad4597 | 1647 | static struct ext4_dir_entry_2* dx_pack_dirents(char *base, unsigned blocksize) |
ac27a0ec | 1648 | { |
617ba13b | 1649 | struct ext4_dir_entry_2 *next, *to, *prev, *de = (struct ext4_dir_entry_2 *) base; |
ac27a0ec DK |
1650 | unsigned rec_len = 0; |
1651 | ||
1652 | prev = to = de; | |
8bad4597 | 1653 | while ((char*)de < base + blocksize) { |
3d0518f4 | 1654 | next = ext4_next_entry(de, blocksize); |
ac27a0ec | 1655 | if (de->inode && de->name_len) { |
617ba13b | 1656 | rec_len = EXT4_DIR_REC_LEN(de->name_len); |
ac27a0ec DK |
1657 | if (de > to) |
1658 | memmove(to, de, rec_len); | |
3d0518f4 | 1659 | to->rec_len = ext4_rec_len_to_disk(rec_len, blocksize); |
ac27a0ec | 1660 | prev = to; |
617ba13b | 1661 | to = (struct ext4_dir_entry_2 *) (((char *) to) + rec_len); |
ac27a0ec DK |
1662 | } |
1663 | de = next; | |
1664 | } | |
1665 | return prev; | |
1666 | } | |
1667 | ||
ef2b02d3 ES |
1668 | /* |
1669 | * Split a full leaf block to make room for a new dir entry. | |
1670 | * Allocate a new block, and move entries so that they are approx. equally full. | |
1671 | * Returns pointer to de in block into which the new entry will be inserted. | |
1672 | */ | |
617ba13b | 1673 | static struct ext4_dir_entry_2 *do_split(handle_t *handle, struct inode *dir, |
ac27a0ec | 1674 | struct buffer_head **bh,struct dx_frame *frame, |
f8b3b59d | 1675 | struct dx_hash_info *hinfo) |
ac27a0ec DK |
1676 | { |
1677 | unsigned blocksize = dir->i_sb->s_blocksize; | |
1678 | unsigned count, continued; | |
1679 | struct buffer_head *bh2; | |
725d26d3 | 1680 | ext4_lblk_t newblock; |
ac27a0ec DK |
1681 | u32 hash2; |
1682 | struct dx_map_entry *map; | |
1683 | char *data1 = (*bh)->b_data, *data2; | |
59e315b4 | 1684 | unsigned split, move, size; |
617ba13b | 1685 | struct ext4_dir_entry_2 *de = NULL, *de2; |
b0336e8d DW |
1686 | struct ext4_dir_entry_tail *t; |
1687 | int csum_size = 0; | |
59e315b4 | 1688 | int err = 0, i; |
ac27a0ec | 1689 | |
9aa5d32b | 1690 | if (ext4_has_metadata_csum(dir->i_sb)) |
b0336e8d DW |
1691 | csum_size = sizeof(struct ext4_dir_entry_tail); |
1692 | ||
0f70b406 TT |
1693 | bh2 = ext4_append(handle, dir, &newblock); |
1694 | if (IS_ERR(bh2)) { | |
ac27a0ec DK |
1695 | brelse(*bh); |
1696 | *bh = NULL; | |
f8b3b59d | 1697 | return (struct ext4_dir_entry_2 *) bh2; |
ac27a0ec DK |
1698 | } |
1699 | ||
1700 | BUFFER_TRACE(*bh, "get_write_access"); | |
617ba13b | 1701 | err = ext4_journal_get_write_access(handle, *bh); |
fedee54d DM |
1702 | if (err) |
1703 | goto journal_error; | |
1704 | ||
ac27a0ec | 1705 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 1706 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
1707 | if (err) |
1708 | goto journal_error; | |
1709 | ||
1710 | data2 = bh2->b_data; | |
1711 | ||
1712 | /* create map in the end of data2 block */ | |
1713 | map = (struct dx_map_entry *) (data2 + blocksize); | |
1f3862b5 | 1714 | count = dx_make_map(dir, (struct ext4_dir_entry_2 *) data1, |
ac27a0ec DK |
1715 | blocksize, hinfo, map); |
1716 | map -= count; | |
af5bc92d | 1717 | dx_sort_map(map, count); |
ef2b02d3 ES |
1718 | /* Split the existing block in the middle, size-wise */ |
1719 | size = 0; | |
1720 | move = 0; | |
1721 | for (i = count-1; i >= 0; i--) { | |
1722 | /* is more than half of this entry in 2nd half of the block? */ | |
1723 | if (size + map[i].size/2 > blocksize/2) | |
1724 | break; | |
1725 | size += map[i].size; | |
1726 | move++; | |
1727 | } | |
1728 | /* map index at which we will split */ | |
1729 | split = count - move; | |
ac27a0ec DK |
1730 | hash2 = map[split].hash; |
1731 | continued = hash2 == map[split - 1].hash; | |
725d26d3 AK |
1732 | dxtrace(printk(KERN_INFO "Split block %lu at %x, %i/%i\n", |
1733 | (unsigned long)dx_get_block(frame->at), | |
1734 | hash2, split, count-split)); | |
ac27a0ec DK |
1735 | |
1736 | /* Fancy dance to stay within two buffers */ | |
1f3862b5 MH |
1737 | de2 = dx_move_dirents(data1, data2, map + split, count - split, |
1738 | blocksize); | |
af5bc92d | 1739 | de = dx_pack_dirents(data1, blocksize); |
b0336e8d DW |
1740 | de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) - |
1741 | (char *) de, | |
3d0518f4 | 1742 | blocksize); |
b0336e8d DW |
1743 | de2->rec_len = ext4_rec_len_to_disk(data2 + (blocksize - csum_size) - |
1744 | (char *) de2, | |
3d0518f4 | 1745 | blocksize); |
b0336e8d DW |
1746 | if (csum_size) { |
1747 | t = EXT4_DIRENT_TAIL(data2, blocksize); | |
1748 | initialize_dirent_tail(t, blocksize); | |
1749 | ||
1750 | t = EXT4_DIRENT_TAIL(data1, blocksize); | |
1751 | initialize_dirent_tail(t, blocksize); | |
1752 | } | |
1753 | ||
b3098486 MH |
1754 | dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data1, |
1755 | blocksize, 1)); | |
1756 | dxtrace(dx_show_leaf(dir, hinfo, (struct ext4_dir_entry_2 *) data2, | |
1757 | blocksize, 1)); | |
ac27a0ec DK |
1758 | |
1759 | /* Which block gets the new entry? */ | |
f8b3b59d | 1760 | if (hinfo->hash >= hash2) { |
ac27a0ec DK |
1761 | swap(*bh, bh2); |
1762 | de = de2; | |
1763 | } | |
af5bc92d | 1764 | dx_insert_block(frame, hash2 + continued, newblock); |
b0336e8d | 1765 | err = ext4_handle_dirty_dirent_node(handle, dir, bh2); |
ac27a0ec DK |
1766 | if (err) |
1767 | goto journal_error; | |
dbe89444 | 1768 | err = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
ac27a0ec DK |
1769 | if (err) |
1770 | goto journal_error; | |
af5bc92d TT |
1771 | brelse(bh2); |
1772 | dxtrace(dx_show_index("frame", frame->entries)); | |
ac27a0ec | 1773 | return de; |
fedee54d DM |
1774 | |
1775 | journal_error: | |
1776 | brelse(*bh); | |
1777 | brelse(bh2); | |
1778 | *bh = NULL; | |
1779 | ext4_std_error(dir->i_sb, err); | |
f8b3b59d | 1780 | return ERR_PTR(err); |
ac27a0ec | 1781 | } |
ac27a0ec | 1782 | |
978fef91 TM |
1783 | int ext4_find_dest_de(struct inode *dir, struct inode *inode, |
1784 | struct buffer_head *bh, | |
1785 | void *buf, int buf_size, | |
5b643f9c | 1786 | struct ext4_filename *fname, |
978fef91 TM |
1787 | struct ext4_dir_entry_2 **dest_de) |
1788 | { | |
1789 | struct ext4_dir_entry_2 *de; | |
5b643f9c | 1790 | unsigned short reclen = EXT4_DIR_REC_LEN(fname_len(fname)); |
978fef91 TM |
1791 | int nlen, rlen; |
1792 | unsigned int offset = 0; | |
1793 | char *top; | |
1f3862b5 | 1794 | |
978fef91 TM |
1795 | de = (struct ext4_dir_entry_2 *)buf; |
1796 | top = buf + buf_size - reclen; | |
1797 | while ((char *) de <= top) { | |
1798 | if (ext4_check_dir_entry(dir, NULL, de, bh, | |
d9b9f8d5 EB |
1799 | buf, buf_size, offset)) |
1800 | return -EFSCORRUPTED; | |
1801 | if (ext4_match(fname, de)) | |
1802 | return -EEXIST; | |
978fef91 TM |
1803 | nlen = EXT4_DIR_REC_LEN(de->name_len); |
1804 | rlen = ext4_rec_len_from_disk(de->rec_len, buf_size); | |
1805 | if ((de->inode ? rlen - nlen : rlen) >= reclen) | |
1806 | break; | |
1807 | de = (struct ext4_dir_entry_2 *)((char *)de + rlen); | |
1808 | offset += rlen; | |
1809 | } | |
1f3862b5 | 1810 | if ((char *) de > top) |
d9b9f8d5 EB |
1811 | return -ENOSPC; |
1812 | ||
1813 | *dest_de = de; | |
1814 | return 0; | |
978fef91 TM |
1815 | } |
1816 | ||
1bc0af60 EB |
1817 | void ext4_insert_dentry(struct inode *inode, |
1818 | struct ext4_dir_entry_2 *de, | |
1819 | int buf_size, | |
1820 | struct ext4_filename *fname) | |
978fef91 TM |
1821 | { |
1822 | ||
1823 | int nlen, rlen; | |
1824 | ||
1825 | nlen = EXT4_DIR_REC_LEN(de->name_len); | |
1826 | rlen = ext4_rec_len_from_disk(de->rec_len, buf_size); | |
1827 | if (de->inode) { | |
1828 | struct ext4_dir_entry_2 *de1 = | |
4bdfc873 | 1829 | (struct ext4_dir_entry_2 *)((char *)de + nlen); |
978fef91 TM |
1830 | de1->rec_len = ext4_rec_len_to_disk(rlen - nlen, buf_size); |
1831 | de->rec_len = ext4_rec_len_to_disk(nlen, buf_size); | |
1832 | de = de1; | |
1833 | } | |
1834 | de->file_type = EXT4_FT_UNKNOWN; | |
1835 | de->inode = cpu_to_le32(inode->i_ino); | |
1836 | ext4_set_de_type(inode->i_sb, de, inode->i_mode); | |
5b643f9c TT |
1837 | de->name_len = fname_len(fname); |
1838 | memcpy(de->name, fname_name(fname), fname_len(fname)); | |
978fef91 | 1839 | } |
4bdfc873 | 1840 | |
ac27a0ec DK |
1841 | /* |
1842 | * Add a new entry into a directory (leaf) block. If de is non-NULL, | |
1843 | * it points to a directory entry which is guaranteed to be large | |
1844 | * enough for new directory entry. If de is NULL, then | |
1845 | * add_dirent_to_buf will attempt search the directory block for | |
1846 | * space. It will return -ENOSPC if no space is available, and -EIO | |
1847 | * and -EEXIST if directory entry already exists. | |
ac27a0ec | 1848 | */ |
5b643f9c TT |
1849 | static int add_dirent_to_buf(handle_t *handle, struct ext4_filename *fname, |
1850 | struct inode *dir, | |
617ba13b | 1851 | struct inode *inode, struct ext4_dir_entry_2 *de, |
af5bc92d | 1852 | struct buffer_head *bh) |
ac27a0ec | 1853 | { |
3d0518f4 | 1854 | unsigned int blocksize = dir->i_sb->s_blocksize; |
b0336e8d | 1855 | int csum_size = 0; |
978fef91 | 1856 | int err; |
b0336e8d | 1857 | |
9aa5d32b | 1858 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 1859 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec | 1860 | |
ac27a0ec | 1861 | if (!de) { |
5b643f9c TT |
1862 | err = ext4_find_dest_de(dir, inode, bh, bh->b_data, |
1863 | blocksize - csum_size, fname, &de); | |
978fef91 TM |
1864 | if (err) |
1865 | return err; | |
ac27a0ec DK |
1866 | } |
1867 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 1868 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1869 | if (err) { |
617ba13b | 1870 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1871 | return err; |
1872 | } | |
1873 | ||
1bc0af60 EB |
1874 | /* By now the buffer is marked for journaling */ |
1875 | ext4_insert_dentry(inode, de, blocksize, fname); | |
978fef91 | 1876 | |
ac27a0ec DK |
1877 | /* |
1878 | * XXX shouldn't update any times until successful | |
1879 | * completion of syscall, but too many callers depend | |
1880 | * on this. | |
1881 | * | |
1882 | * XXX similarly, too many callers depend on | |
617ba13b | 1883 | * ext4_new_inode() setting the times, but error |
ac27a0ec DK |
1884 | * recovery deletes the inode, so the worst that can |
1885 | * happen is that the times are slightly out of date | |
1886 | * and/or different from the directory change time. | |
1887 | */ | |
eeca7ea1 | 1888 | dir->i_mtime = dir->i_ctime = current_time(dir); |
617ba13b | 1889 | ext4_update_dx_flag(dir); |
e08ac99f | 1890 | inode_inc_iversion(dir); |
617ba13b | 1891 | ext4_mark_inode_dirty(handle, dir); |
0390131b | 1892 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); |
b0336e8d | 1893 | err = ext4_handle_dirty_dirent_node(handle, dir, bh); |
ac27a0ec | 1894 | if (err) |
617ba13b | 1895 | ext4_std_error(dir->i_sb, err); |
ac27a0ec DK |
1896 | return 0; |
1897 | } | |
1898 | ||
ac27a0ec DK |
1899 | /* |
1900 | * This converts a one block unindexed directory to a 3 block indexed | |
1901 | * directory, and adds the dentry to the indexed directory. | |
1902 | */ | |
5b643f9c | 1903 | static int make_indexed_dir(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 1904 | struct inode *dir, |
ac27a0ec DK |
1905 | struct inode *inode, struct buffer_head *bh) |
1906 | { | |
ac27a0ec DK |
1907 | struct buffer_head *bh2; |
1908 | struct dx_root *root; | |
e08ac99f | 1909 | struct dx_frame frames[EXT4_HTREE_LEVEL], *frame; |
ac27a0ec | 1910 | struct dx_entry *entries; |
617ba13b | 1911 | struct ext4_dir_entry_2 *de, *de2; |
b0336e8d | 1912 | struct ext4_dir_entry_tail *t; |
ac27a0ec DK |
1913 | char *data1, *top; |
1914 | unsigned len; | |
1915 | int retval; | |
1916 | unsigned blocksize; | |
725d26d3 | 1917 | ext4_lblk_t block; |
ac27a0ec | 1918 | struct fake_dirent *fde; |
4bdfc873 MH |
1919 | int csum_size = 0; |
1920 | ||
9aa5d32b | 1921 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 1922 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec DK |
1923 | |
1924 | blocksize = dir->i_sb->s_blocksize; | |
e6b8bc09 | 1925 | dxtrace(printk(KERN_DEBUG "Creating index: inode %lu\n", dir->i_ino)); |
5d601255 | 1926 | BUFFER_TRACE(bh, "get_write_access"); |
617ba13b | 1927 | retval = ext4_journal_get_write_access(handle, bh); |
ac27a0ec | 1928 | if (retval) { |
617ba13b | 1929 | ext4_std_error(dir->i_sb, retval); |
ac27a0ec DK |
1930 | brelse(bh); |
1931 | return retval; | |
1932 | } | |
1933 | root = (struct dx_root *) bh->b_data; | |
1934 | ||
e6b8bc09 TT |
1935 | /* The 0th block becomes the root, move the dirents out */ |
1936 | fde = &root->dotdot; | |
1937 | de = (struct ext4_dir_entry_2 *)((char *)fde + | |
3d0518f4 | 1938 | ext4_rec_len_from_disk(fde->rec_len, blocksize)); |
e6b8bc09 | 1939 | if ((char *) de >= (((char *) root) + blocksize)) { |
24676da4 | 1940 | EXT4_ERROR_INODE(dir, "invalid rec_len for '..'"); |
e6b8bc09 | 1941 | brelse(bh); |
6a797d27 | 1942 | return -EFSCORRUPTED; |
e6b8bc09 | 1943 | } |
b0336e8d | 1944 | len = ((char *) root) + (blocksize - csum_size) - (char *) de; |
e6b8bc09 TT |
1945 | |
1946 | /* Allocate new block for the 0th block's dirents */ | |
0f70b406 TT |
1947 | bh2 = ext4_append(handle, dir, &block); |
1948 | if (IS_ERR(bh2)) { | |
ac27a0ec | 1949 | brelse(bh); |
0f70b406 | 1950 | return PTR_ERR(bh2); |
ac27a0ec | 1951 | } |
12e9b892 | 1952 | ext4_set_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec DK |
1953 | data1 = bh2->b_data; |
1954 | ||
ac27a0ec | 1955 | memcpy (data1, de, len); |
617ba13b | 1956 | de = (struct ext4_dir_entry_2 *) data1; |
ac27a0ec | 1957 | top = data1 + len; |
3d0518f4 | 1958 | while ((char *)(de2 = ext4_next_entry(de, blocksize)) < top) |
ac27a0ec | 1959 | de = de2; |
b0336e8d DW |
1960 | de->rec_len = ext4_rec_len_to_disk(data1 + (blocksize - csum_size) - |
1961 | (char *) de, | |
3d0518f4 | 1962 | blocksize); |
b0336e8d DW |
1963 | |
1964 | if (csum_size) { | |
1965 | t = EXT4_DIRENT_TAIL(data1, blocksize); | |
1966 | initialize_dirent_tail(t, blocksize); | |
1967 | } | |
1968 | ||
ac27a0ec | 1969 | /* Initialize the root; the dot dirents already exist */ |
617ba13b | 1970 | de = (struct ext4_dir_entry_2 *) (&root->dotdot); |
3d0518f4 WY |
1971 | de->rec_len = ext4_rec_len_to_disk(blocksize - EXT4_DIR_REC_LEN(2), |
1972 | blocksize); | |
ac27a0ec DK |
1973 | memset (&root->info, 0, sizeof(root->info)); |
1974 | root->info.info_length = sizeof(root->info); | |
617ba13b | 1975 | root->info.hash_version = EXT4_SB(dir->i_sb)->s_def_hash_version; |
ac27a0ec | 1976 | entries = root->entries; |
af5bc92d TT |
1977 | dx_set_block(entries, 1); |
1978 | dx_set_count(entries, 1); | |
1979 | dx_set_limit(entries, dx_root_limit(dir, sizeof(root->info))); | |
ac27a0ec DK |
1980 | |
1981 | /* Initialize as for dx_probe */ | |
5b643f9c TT |
1982 | fname->hinfo.hash_version = root->info.hash_version; |
1983 | if (fname->hinfo.hash_version <= DX_HASH_TEA) | |
1984 | fname->hinfo.hash_version += EXT4_SB(dir->i_sb)->s_hash_unsigned; | |
1985 | fname->hinfo.seed = EXT4_SB(dir->i_sb)->s_hash_seed; | |
1986 | ext4fs_dirhash(fname_name(fname), fname_len(fname), &fname->hinfo); | |
1987 | ||
6050d47a | 1988 | memset(frames, 0, sizeof(frames)); |
ac27a0ec DK |
1989 | frame = frames; |
1990 | frame->entries = entries; | |
1991 | frame->at = entries; | |
1992 | frame->bh = bh; | |
6976a6f2 | 1993 | |
6050d47a JK |
1994 | retval = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
1995 | if (retval) | |
1996 | goto out_frames; | |
e81d4477 | 1997 | retval = ext4_handle_dirty_dirent_node(handle, dir, bh2); |
6050d47a JK |
1998 | if (retval) |
1999 | goto out_frames; | |
6976a6f2 | 2000 | |
e81d4477 | 2001 | de = do_split(handle,dir, &bh2, frame, &fname->hinfo); |
f8b3b59d | 2002 | if (IS_ERR(de)) { |
6050d47a JK |
2003 | retval = PTR_ERR(de); |
2004 | goto out_frames; | |
7ad8e4e6 | 2005 | } |
ac27a0ec | 2006 | |
e81d4477 | 2007 | retval = add_dirent_to_buf(handle, fname, dir, inode, de, bh2); |
6050d47a JK |
2008 | out_frames: |
2009 | /* | |
2010 | * Even if the block split failed, we have to properly write | |
2011 | * out all the changes we did so far. Otherwise we can end up | |
2012 | * with corrupted filesystem. | |
2013 | */ | |
e81d4477 | 2014 | if (retval) |
2015 | ext4_mark_inode_dirty(handle, dir); | |
6050d47a | 2016 | dx_release(frames); |
e81d4477 | 2017 | brelse(bh2); |
6050d47a | 2018 | return retval; |
ac27a0ec | 2019 | } |
ac27a0ec DK |
2020 | |
2021 | /* | |
617ba13b | 2022 | * ext4_add_entry() |
ac27a0ec DK |
2023 | * |
2024 | * adds a file entry to the specified directory, using the same | |
617ba13b | 2025 | * semantics as ext4_find_entry(). It returns NULL if it failed. |
ac27a0ec DK |
2026 | * |
2027 | * NOTE!! The inode part of 'de' is left at 0 - which means you | |
2028 | * may not sleep between calling this and putting something into | |
2029 | * the entry, as someone else might have used it while you slept. | |
2030 | */ | |
af5bc92d TT |
2031 | static int ext4_add_entry(handle_t *handle, struct dentry *dentry, |
2032 | struct inode *inode) | |
ac27a0ec | 2033 | { |
2b0143b5 | 2034 | struct inode *dir = d_inode(dentry->d_parent); |
e12fb972 | 2035 | struct buffer_head *bh = NULL; |
617ba13b | 2036 | struct ext4_dir_entry_2 *de; |
b0336e8d | 2037 | struct ext4_dir_entry_tail *t; |
af5bc92d | 2038 | struct super_block *sb; |
5b643f9c | 2039 | struct ext4_filename fname; |
ac27a0ec | 2040 | int retval; |
ac27a0ec | 2041 | int dx_fallback=0; |
ac27a0ec | 2042 | unsigned blocksize; |
725d26d3 | 2043 | ext4_lblk_t block, blocks; |
b0336e8d DW |
2044 | int csum_size = 0; |
2045 | ||
9aa5d32b | 2046 | if (ext4_has_metadata_csum(inode->i_sb)) |
b0336e8d | 2047 | csum_size = sizeof(struct ext4_dir_entry_tail); |
ac27a0ec DK |
2048 | |
2049 | sb = dir->i_sb; | |
2050 | blocksize = sb->s_blocksize; | |
2051 | if (!dentry->d_name.len) | |
2052 | return -EINVAL; | |
3c47d541 | 2053 | |
5b643f9c TT |
2054 | retval = ext4_fname_setup_filename(dir, &dentry->d_name, 0, &fname); |
2055 | if (retval) | |
2056 | return retval; | |
2057 | ||
3c47d541 | 2058 | if (ext4_has_inline_data(dir)) { |
56a04915 | 2059 | retval = ext4_try_add_inline_entry(handle, &fname, dir, inode); |
3c47d541 | 2060 | if (retval < 0) |
5b643f9c | 2061 | goto out; |
3c47d541 TM |
2062 | if (retval == 1) { |
2063 | retval = 0; | |
e12fb972 | 2064 | goto out; |
3c47d541 TM |
2065 | } |
2066 | } | |
2067 | ||
ac27a0ec | 2068 | if (is_dx(dir)) { |
56a04915 | 2069 | retval = ext4_dx_add_entry(handle, &fname, dir, inode); |
ac27a0ec | 2070 | if (!retval || (retval != ERR_BAD_DX_DIR)) |
e12fb972 | 2071 | goto out; |
12e9b892 | 2072 | ext4_clear_inode_flag(dir, EXT4_INODE_INDEX); |
ac27a0ec | 2073 | dx_fallback++; |
617ba13b | 2074 | ext4_mark_inode_dirty(handle, dir); |
ac27a0ec | 2075 | } |
ac27a0ec | 2076 | blocks = dir->i_size >> sb->s_blocksize_bits; |
498e5f24 | 2077 | for (block = 0; block < blocks; block++) { |
dc6982ff | 2078 | bh = ext4_read_dirblock(dir, block, DIRENT); |
5b643f9c TT |
2079 | if (IS_ERR(bh)) { |
2080 | retval = PTR_ERR(bh); | |
2081 | bh = NULL; | |
2082 | goto out; | |
2083 | } | |
2084 | retval = add_dirent_to_buf(handle, &fname, dir, inode, | |
2085 | NULL, bh); | |
e12fb972 LC |
2086 | if (retval != -ENOSPC) |
2087 | goto out; | |
ac27a0ec | 2088 | |
ac27a0ec | 2089 | if (blocks == 1 && !dx_fallback && |
e2b911c5 | 2090 | ext4_has_feature_dir_index(sb)) { |
56a04915 | 2091 | retval = make_indexed_dir(handle, &fname, dir, |
5b643f9c | 2092 | inode, bh); |
e12fb972 LC |
2093 | bh = NULL; /* make_indexed_dir releases bh */ |
2094 | goto out; | |
2095 | } | |
ac27a0ec DK |
2096 | brelse(bh); |
2097 | } | |
0f70b406 | 2098 | bh = ext4_append(handle, dir, &block); |
5b643f9c TT |
2099 | if (IS_ERR(bh)) { |
2100 | retval = PTR_ERR(bh); | |
2101 | bh = NULL; | |
2102 | goto out; | |
2103 | } | |
617ba13b | 2104 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 2105 | de->inode = 0; |
b0336e8d DW |
2106 | de->rec_len = ext4_rec_len_to_disk(blocksize - csum_size, blocksize); |
2107 | ||
2108 | if (csum_size) { | |
2109 | t = EXT4_DIRENT_TAIL(bh->b_data, blocksize); | |
2110 | initialize_dirent_tail(t, blocksize); | |
2111 | } | |
2112 | ||
5b643f9c | 2113 | retval = add_dirent_to_buf(handle, &fname, dir, inode, de, bh); |
e12fb972 | 2114 | out: |
5b643f9c | 2115 | ext4_fname_free_filename(&fname); |
2de770a4 | 2116 | brelse(bh); |
14ece102 FM |
2117 | if (retval == 0) |
2118 | ext4_set_inode_state(inode, EXT4_STATE_NEWENTRY); | |
2de770a4 | 2119 | return retval; |
ac27a0ec DK |
2120 | } |
2121 | ||
ac27a0ec DK |
2122 | /* |
2123 | * Returns 0 for success, or a negative error value | |
2124 | */ | |
5b643f9c | 2125 | static int ext4_dx_add_entry(handle_t *handle, struct ext4_filename *fname, |
56a04915 | 2126 | struct inode *dir, struct inode *inode) |
ac27a0ec | 2127 | { |
e08ac99f | 2128 | struct dx_frame frames[EXT4_HTREE_LEVEL], *frame; |
ac27a0ec | 2129 | struct dx_entry *entries, *at; |
af5bc92d | 2130 | struct buffer_head *bh; |
af5bc92d | 2131 | struct super_block *sb = dir->i_sb; |
617ba13b | 2132 | struct ext4_dir_entry_2 *de; |
e08ac99f | 2133 | int restart; |
ac27a0ec DK |
2134 | int err; |
2135 | ||
e08ac99f AB |
2136 | again: |
2137 | restart = 0; | |
5b643f9c | 2138 | frame = dx_probe(fname, dir, NULL, frames); |
dd73b5d5 TT |
2139 | if (IS_ERR(frame)) |
2140 | return PTR_ERR(frame); | |
ac27a0ec DK |
2141 | entries = frame->entries; |
2142 | at = frame->at; | |
dc6982ff TT |
2143 | bh = ext4_read_dirblock(dir, dx_get_block(frame->at), DIRENT); |
2144 | if (IS_ERR(bh)) { | |
2145 | err = PTR_ERR(bh); | |
2146 | bh = NULL; | |
ac27a0ec | 2147 | goto cleanup; |
6d1ab10e | 2148 | } |
ac27a0ec DK |
2149 | |
2150 | BUFFER_TRACE(bh, "get_write_access"); | |
617ba13b | 2151 | err = ext4_journal_get_write_access(handle, bh); |
ac27a0ec DK |
2152 | if (err) |
2153 | goto journal_error; | |
2154 | ||
5b643f9c | 2155 | err = add_dirent_to_buf(handle, fname, dir, inode, NULL, bh); |
2de770a4 | 2156 | if (err != -ENOSPC) |
ac27a0ec | 2157 | goto cleanup; |
ac27a0ec | 2158 | |
e08ac99f | 2159 | err = 0; |
ac27a0ec | 2160 | /* Block full, should compress but for now just split */ |
4776004f | 2161 | dxtrace(printk(KERN_DEBUG "using %u of %u node entries\n", |
ac27a0ec DK |
2162 | dx_get_count(entries), dx_get_limit(entries))); |
2163 | /* Need to split index? */ | |
2164 | if (dx_get_count(entries) == dx_get_limit(entries)) { | |
725d26d3 | 2165 | ext4_lblk_t newblock; |
e08ac99f AB |
2166 | int levels = frame - frames + 1; |
2167 | unsigned int icount; | |
2168 | int add_level = 1; | |
ac27a0ec DK |
2169 | struct dx_entry *entries2; |
2170 | struct dx_node *node2; | |
2171 | struct buffer_head *bh2; | |
2172 | ||
e08ac99f AB |
2173 | while (frame > frames) { |
2174 | if (dx_get_count((frame - 1)->entries) < | |
2175 | dx_get_limit((frame - 1)->entries)) { | |
2176 | add_level = 0; | |
2177 | break; | |
2178 | } | |
2179 | frame--; /* split higher index block */ | |
2180 | at = frame->at; | |
2181 | entries = frame->entries; | |
2182 | restart = 1; | |
2183 | } | |
2184 | if (add_level && levels == ext4_dir_htree_level(sb)) { | |
2185 | ext4_warning(sb, "Directory (ino: %lu) index full, " | |
2186 | "reach max htree level :%d", | |
2187 | dir->i_ino, levels); | |
2188 | if (ext4_dir_htree_level(sb) < EXT4_HTREE_LEVEL) { | |
2189 | ext4_warning(sb, "Large directory feature is " | |
2190 | "not enabled on this " | |
2191 | "filesystem"); | |
2192 | } | |
ac27a0ec DK |
2193 | err = -ENOSPC; |
2194 | goto cleanup; | |
2195 | } | |
e08ac99f | 2196 | icount = dx_get_count(entries); |
0f70b406 TT |
2197 | bh2 = ext4_append(handle, dir, &newblock); |
2198 | if (IS_ERR(bh2)) { | |
2199 | err = PTR_ERR(bh2); | |
ac27a0ec | 2200 | goto cleanup; |
0f70b406 | 2201 | } |
ac27a0ec DK |
2202 | node2 = (struct dx_node *)(bh2->b_data); |
2203 | entries2 = node2->entries; | |
1f7bebb9 | 2204 | memset(&node2->fake, 0, sizeof(struct fake_dirent)); |
3d0518f4 WY |
2205 | node2->fake.rec_len = ext4_rec_len_to_disk(sb->s_blocksize, |
2206 | sb->s_blocksize); | |
ac27a0ec | 2207 | BUFFER_TRACE(frame->bh, "get_write_access"); |
617ba13b | 2208 | err = ext4_journal_get_write_access(handle, frame->bh); |
ac27a0ec DK |
2209 | if (err) |
2210 | goto journal_error; | |
e08ac99f | 2211 | if (!add_level) { |
ac27a0ec DK |
2212 | unsigned icount1 = icount/2, icount2 = icount - icount1; |
2213 | unsigned hash2 = dx_get_hash(entries + icount1); | |
4776004f TT |
2214 | dxtrace(printk(KERN_DEBUG "Split index %i/%i\n", |
2215 | icount1, icount2)); | |
ac27a0ec DK |
2216 | |
2217 | BUFFER_TRACE(frame->bh, "get_write_access"); /* index root */ | |
617ba13b | 2218 | err = ext4_journal_get_write_access(handle, |
e08ac99f | 2219 | (frame - 1)->bh); |
ac27a0ec DK |
2220 | if (err) |
2221 | goto journal_error; | |
2222 | ||
af5bc92d TT |
2223 | memcpy((char *) entries2, (char *) (entries + icount1), |
2224 | icount2 * sizeof(struct dx_entry)); | |
2225 | dx_set_count(entries, icount1); | |
2226 | dx_set_count(entries2, icount2); | |
2227 | dx_set_limit(entries2, dx_node_limit(dir)); | |
ac27a0ec DK |
2228 | |
2229 | /* Which index block gets the new entry? */ | |
2230 | if (at - entries >= icount1) { | |
2231 | frame->at = at = at - entries - icount1 + entries2; | |
2232 | frame->entries = entries = entries2; | |
2233 | swap(frame->bh, bh2); | |
2234 | } | |
e08ac99f AB |
2235 | dx_insert_block((frame - 1), hash2, newblock); |
2236 | dxtrace(dx_show_index("node", frame->entries)); | |
af5bc92d | 2237 | dxtrace(dx_show_index("node", |
ac27a0ec | 2238 | ((struct dx_node *) bh2->b_data)->entries)); |
dbe89444 | 2239 | err = ext4_handle_dirty_dx_node(handle, dir, bh2); |
ac27a0ec DK |
2240 | if (err) |
2241 | goto journal_error; | |
2242 | brelse (bh2); | |
e08ac99f AB |
2243 | err = ext4_handle_dirty_dx_node(handle, dir, |
2244 | (frame - 1)->bh); | |
2245 | if (err) | |
2246 | goto journal_error; | |
2247 | if (restart) { | |
2248 | err = ext4_handle_dirty_dx_node(handle, dir, | |
2249 | frame->bh); | |
2250 | goto journal_error; | |
2251 | } | |
ac27a0ec | 2252 | } else { |
e08ac99f | 2253 | struct dx_root *dxroot; |
ac27a0ec DK |
2254 | memcpy((char *) entries2, (char *) entries, |
2255 | icount * sizeof(struct dx_entry)); | |
2256 | dx_set_limit(entries2, dx_node_limit(dir)); | |
2257 | ||
2258 | /* Set up root */ | |
2259 | dx_set_count(entries, 1); | |
2260 | dx_set_block(entries + 0, newblock); | |
e08ac99f AB |
2261 | dxroot = (struct dx_root *)frames[0].bh->b_data; |
2262 | dxroot->info.indirect_levels += 1; | |
2263 | dxtrace(printk(KERN_DEBUG | |
2264 | "Creating %d level index...\n", | |
799578ab | 2265 | dxroot->info.indirect_levels)); |
e08ac99f | 2266 | err = ext4_handle_dirty_dx_node(handle, dir, frame->bh); |
ac27a0ec DK |
2267 | if (err) |
2268 | goto journal_error; | |
e08ac99f AB |
2269 | err = ext4_handle_dirty_dx_node(handle, dir, bh2); |
2270 | brelse(bh2); | |
2271 | restart = 1; | |
2272 | goto journal_error; | |
b4097142 | 2273 | } |
ac27a0ec | 2274 | } |
5b643f9c | 2275 | de = do_split(handle, dir, &bh, frame, &fname->hinfo); |
f8b3b59d TT |
2276 | if (IS_ERR(de)) { |
2277 | err = PTR_ERR(de); | |
ac27a0ec | 2278 | goto cleanup; |
f8b3b59d | 2279 | } |
5b643f9c | 2280 | err = add_dirent_to_buf(handle, fname, dir, inode, de, bh); |
ac27a0ec DK |
2281 | goto cleanup; |
2282 | ||
2283 | journal_error: | |
e08ac99f | 2284 | ext4_std_error(dir->i_sb, err); /* this is a no-op if err == 0 */ |
ac27a0ec | 2285 | cleanup: |
b1deefc9 | 2286 | brelse(bh); |
ac27a0ec | 2287 | dx_release(frames); |
e08ac99f AB |
2288 | /* @restart is true means htree-path has been changed, we need to |
2289 | * repeat dx_probe() to find out valid htree-path | |
2290 | */ | |
2291 | if (restart && err == 0) | |
2292 | goto again; | |
ac27a0ec DK |
2293 | return err; |
2294 | } | |
ac27a0ec DK |
2295 | |
2296 | /* | |
05019a9e TM |
2297 | * ext4_generic_delete_entry deletes a directory entry by merging it |
2298 | * with the previous entry | |
ac27a0ec | 2299 | */ |
05019a9e TM |
2300 | int ext4_generic_delete_entry(handle_t *handle, |
2301 | struct inode *dir, | |
2302 | struct ext4_dir_entry_2 *de_del, | |
2303 | struct buffer_head *bh, | |
2304 | void *entry_buf, | |
2305 | int buf_size, | |
2306 | int csum_size) | |
ac27a0ec | 2307 | { |
af5bc92d | 2308 | struct ext4_dir_entry_2 *de, *pde; |
3d0518f4 | 2309 | unsigned int blocksize = dir->i_sb->s_blocksize; |
05019a9e | 2310 | int i; |
b0336e8d | 2311 | |
ac27a0ec DK |
2312 | i = 0; |
2313 | pde = NULL; | |
05019a9e TM |
2314 | de = (struct ext4_dir_entry_2 *)entry_buf; |
2315 | while (i < buf_size - csum_size) { | |
226ba972 TM |
2316 | if (ext4_check_dir_entry(dir, NULL, de, bh, |
2317 | bh->b_data, bh->b_size, i)) | |
6a797d27 | 2318 | return -EFSCORRUPTED; |
ac27a0ec | 2319 | if (de == de_del) { |
ac27a0ec | 2320 | if (pde) |
a72d7f83 | 2321 | pde->rec_len = ext4_rec_len_to_disk( |
3d0518f4 WY |
2322 | ext4_rec_len_from_disk(pde->rec_len, |
2323 | blocksize) + | |
2324 | ext4_rec_len_from_disk(de->rec_len, | |
2325 | blocksize), | |
2326 | blocksize); | |
ac27a0ec DK |
2327 | else |
2328 | de->inode = 0; | |
e08ac99f | 2329 | inode_inc_iversion(dir); |
ac27a0ec DK |
2330 | return 0; |
2331 | } | |
3d0518f4 | 2332 | i += ext4_rec_len_from_disk(de->rec_len, blocksize); |
ac27a0ec | 2333 | pde = de; |
3d0518f4 | 2334 | de = ext4_next_entry(de, blocksize); |
ac27a0ec DK |
2335 | } |
2336 | return -ENOENT; | |
2337 | } | |
2338 | ||
05019a9e TM |
2339 | static int ext4_delete_entry(handle_t *handle, |
2340 | struct inode *dir, | |
2341 | struct ext4_dir_entry_2 *de_del, | |
2342 | struct buffer_head *bh) | |
2343 | { | |
2344 | int err, csum_size = 0; | |
2345 | ||
9f40fe54 TM |
2346 | if (ext4_has_inline_data(dir)) { |
2347 | int has_inline_data = 1; | |
2348 | err = ext4_delete_inline_entry(handle, dir, de_del, bh, | |
2349 | &has_inline_data); | |
2350 | if (has_inline_data) | |
2351 | return err; | |
2352 | } | |
2353 | ||
9aa5d32b | 2354 | if (ext4_has_metadata_csum(dir->i_sb)) |
05019a9e TM |
2355 | csum_size = sizeof(struct ext4_dir_entry_tail); |
2356 | ||
2357 | BUFFER_TRACE(bh, "get_write_access"); | |
2358 | err = ext4_journal_get_write_access(handle, bh); | |
2359 | if (unlikely(err)) | |
2360 | goto out; | |
2361 | ||
2362 | err = ext4_generic_delete_entry(handle, dir, de_del, | |
2363 | bh, bh->b_data, | |
2364 | dir->i_sb->s_blocksize, csum_size); | |
2365 | if (err) | |
2366 | goto out; | |
2367 | ||
2368 | BUFFER_TRACE(bh, "call ext4_handle_dirty_metadata"); | |
2369 | err = ext4_handle_dirty_dirent_node(handle, dir, bh); | |
2370 | if (unlikely(err)) | |
2371 | goto out; | |
2372 | ||
2373 | return 0; | |
2374 | out: | |
2375 | if (err != -ENOENT) | |
2376 | ext4_std_error(dir->i_sb, err); | |
2377 | return err; | |
2378 | } | |
2379 | ||
f8628a14 | 2380 | /* |
c7414892 AD |
2381 | * Set directory link count to 1 if nlinks > EXT4_LINK_MAX, or if nlinks == 2 |
2382 | * since this indicates that nlinks count was previously 1 to avoid overflowing | |
2383 | * the 16-bit i_links_count field on disk. Directories with i_nlink == 1 mean | |
2384 | * that subdirectory link counts are not being maintained accurately. | |
2385 | * | |
2386 | * The caller has already checked for i_nlink overflow in case the DIR_LINK | |
2387 | * feature is not enabled and returned -EMLINK. The is_dx() check is a proxy | |
2388 | * for checking S_ISDIR(inode) (since the INODE_INDEX feature will not be set | |
2389 | * on regular files) and to avoid creating huge/slow non-HTREE directories. | |
f8628a14 AD |
2390 | */ |
2391 | static void ext4_inc_count(handle_t *handle, struct inode *inode) | |
2392 | { | |
2393 | inc_nlink(inode); | |
c7414892 AD |
2394 | if (is_dx(inode) && |
2395 | (inode->i_nlink > EXT4_LINK_MAX || inode->i_nlink == 2)) | |
2396 | set_nlink(inode, 1); | |
f8628a14 AD |
2397 | } |
2398 | ||
2399 | /* | |
2400 | * If a directory had nlink == 1, then we should let it be 1. This indicates | |
2401 | * directory has >EXT4_LINK_MAX subdirs. | |
2402 | */ | |
2403 | static void ext4_dec_count(handle_t *handle, struct inode *inode) | |
2404 | { | |
909a4cf1 AD |
2405 | if (!S_ISDIR(inode->i_mode) || inode->i_nlink > 2) |
2406 | drop_nlink(inode); | |
f8628a14 AD |
2407 | } |
2408 | ||
2409 | ||
617ba13b | 2410 | static int ext4_add_nondir(handle_t *handle, |
ac27a0ec DK |
2411 | struct dentry *dentry, struct inode *inode) |
2412 | { | |
617ba13b | 2413 | int err = ext4_add_entry(handle, dentry, inode); |
ac27a0ec | 2414 | if (!err) { |
617ba13b | 2415 | ext4_mark_inode_dirty(handle, inode); |
1e2e547a | 2416 | d_instantiate_new(dentry, inode); |
ac27a0ec DK |
2417 | return 0; |
2418 | } | |
731b9a54 | 2419 | drop_nlink(inode); |
6b38e842 | 2420 | unlock_new_inode(inode); |
ac27a0ec DK |
2421 | iput(inode); |
2422 | return err; | |
2423 | } | |
2424 | ||
2425 | /* | |
2426 | * By the time this is called, we already have created | |
2427 | * the directory cache entry for the new file, but it | |
2428 | * is so far negative - it has no inode. | |
2429 | * | |
2430 | * If the create succeeds, we fill in the inode information | |
2431 | * with d_instantiate(). | |
2432 | */ | |
4acdaf27 | 2433 | static int ext4_create(struct inode *dir, struct dentry *dentry, umode_t mode, |
ebfc3b49 | 2434 | bool excl) |
ac27a0ec DK |
2435 | { |
2436 | handle_t *handle; | |
af5bc92d | 2437 | struct inode *inode; |
1139575a | 2438 | int err, credits, retries = 0; |
ac27a0ec | 2439 | |
a7cdadee JK |
2440 | err = dquot_initialize(dir); |
2441 | if (err) | |
2442 | return err; | |
907f4554 | 2443 | |
1139575a | 2444 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2445 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2446 | retry: |
1139575a TT |
2447 | inode = ext4_new_inode_start_handle(dir, mode, &dentry->d_name, 0, |
2448 | NULL, EXT4_HT_DIR, credits); | |
2449 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2450 | err = PTR_ERR(inode); |
2451 | if (!IS_ERR(inode)) { | |
617ba13b | 2452 | inode->i_op = &ext4_file_inode_operations; |
be64f884 | 2453 | inode->i_fop = &ext4_file_operations; |
617ba13b | 2454 | ext4_set_aops(inode); |
e709e9df | 2455 | err = ext4_add_nondir(handle, dentry, inode); |
1139575a TT |
2456 | if (!err && IS_DIRSYNC(dir)) |
2457 | ext4_handle_sync(handle); | |
ac27a0ec | 2458 | } |
1139575a TT |
2459 | if (handle) |
2460 | ext4_journal_stop(handle); | |
617ba13b | 2461 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2462 | goto retry; |
2463 | return err; | |
2464 | } | |
2465 | ||
af5bc92d | 2466 | static int ext4_mknod(struct inode *dir, struct dentry *dentry, |
1a67aafb | 2467 | umode_t mode, dev_t rdev) |
ac27a0ec DK |
2468 | { |
2469 | handle_t *handle; | |
2470 | struct inode *inode; | |
1139575a | 2471 | int err, credits, retries = 0; |
ac27a0ec | 2472 | |
a7cdadee JK |
2473 | err = dquot_initialize(dir); |
2474 | if (err) | |
2475 | return err; | |
907f4554 | 2476 | |
1139575a | 2477 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2478 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2479 | retry: |
1139575a TT |
2480 | inode = ext4_new_inode_start_handle(dir, mode, &dentry->d_name, 0, |
2481 | NULL, EXT4_HT_DIR, credits); | |
2482 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2483 | err = PTR_ERR(inode); |
2484 | if (!IS_ERR(inode)) { | |
2485 | init_special_inode(inode, inode->i_mode, rdev); | |
617ba13b | 2486 | inode->i_op = &ext4_special_inode_operations; |
617ba13b | 2487 | err = ext4_add_nondir(handle, dentry, inode); |
1139575a TT |
2488 | if (!err && IS_DIRSYNC(dir)) |
2489 | ext4_handle_sync(handle); | |
ac27a0ec | 2490 | } |
1139575a TT |
2491 | if (handle) |
2492 | ext4_journal_stop(handle); | |
617ba13b | 2493 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2494 | goto retry; |
2495 | return err; | |
2496 | } | |
2497 | ||
af51a2ac AV |
2498 | static int ext4_tmpfile(struct inode *dir, struct dentry *dentry, umode_t mode) |
2499 | { | |
2500 | handle_t *handle; | |
2501 | struct inode *inode; | |
2502 | int err, retries = 0; | |
2503 | ||
a7cdadee JK |
2504 | err = dquot_initialize(dir); |
2505 | if (err) | |
2506 | return err; | |
af51a2ac AV |
2507 | |
2508 | retry: | |
2509 | inode = ext4_new_inode_start_handle(dir, mode, | |
2510 | NULL, 0, NULL, | |
2511 | EXT4_HT_DIR, | |
2512 | EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) + | |
2513 | 4 + EXT4_XATTR_TRANS_BLOCKS); | |
2514 | handle = ext4_journal_current_handle(); | |
2515 | err = PTR_ERR(inode); | |
2516 | if (!IS_ERR(inode)) { | |
2517 | inode->i_op = &ext4_file_inode_operations; | |
be64f884 | 2518 | inode->i_fop = &ext4_file_operations; |
af51a2ac | 2519 | ext4_set_aops(inode); |
e94bd349 | 2520 | d_tmpfile(dentry, inode); |
af51a2ac AV |
2521 | err = ext4_orphan_add(handle, inode); |
2522 | if (err) | |
43ae9e3f | 2523 | goto err_unlock_inode; |
af51a2ac | 2524 | mark_inode_dirty(inode); |
af51a2ac AV |
2525 | unlock_new_inode(inode); |
2526 | } | |
2527 | if (handle) | |
2528 | ext4_journal_stop(handle); | |
2529 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
2530 | goto retry; | |
2531 | return err; | |
43ae9e3f | 2532 | err_unlock_inode: |
af51a2ac AV |
2533 | ext4_journal_stop(handle); |
2534 | unlock_new_inode(inode); | |
af51a2ac AV |
2535 | return err; |
2536 | } | |
2537 | ||
a774f9c2 TM |
2538 | struct ext4_dir_entry_2 *ext4_init_dot_dotdot(struct inode *inode, |
2539 | struct ext4_dir_entry_2 *de, | |
2540 | int blocksize, int csum_size, | |
2541 | unsigned int parent_ino, int dotdot_real_len) | |
2542 | { | |
2543 | de->inode = cpu_to_le32(inode->i_ino); | |
2544 | de->name_len = 1; | |
2545 | de->rec_len = ext4_rec_len_to_disk(EXT4_DIR_REC_LEN(de->name_len), | |
2546 | blocksize); | |
2547 | strcpy(de->name, "."); | |
2548 | ext4_set_de_type(inode->i_sb, de, S_IFDIR); | |
2549 | ||
2550 | de = ext4_next_entry(de, blocksize); | |
2551 | de->inode = cpu_to_le32(parent_ino); | |
2552 | de->name_len = 2; | |
2553 | if (!dotdot_real_len) | |
2554 | de->rec_len = ext4_rec_len_to_disk(blocksize - | |
2555 | (csum_size + EXT4_DIR_REC_LEN(1)), | |
2556 | blocksize); | |
2557 | else | |
2558 | de->rec_len = ext4_rec_len_to_disk( | |
2559 | EXT4_DIR_REC_LEN(de->name_len), blocksize); | |
2560 | strcpy(de->name, ".."); | |
2561 | ext4_set_de_type(inode->i_sb, de, S_IFDIR); | |
2562 | ||
2563 | return ext4_next_entry(de, blocksize); | |
2564 | } | |
2565 | ||
2566 | static int ext4_init_new_dir(handle_t *handle, struct inode *dir, | |
2567 | struct inode *inode) | |
ac27a0ec | 2568 | { |
dabd991f | 2569 | struct buffer_head *dir_block = NULL; |
af5bc92d | 2570 | struct ext4_dir_entry_2 *de; |
b0336e8d | 2571 | struct ext4_dir_entry_tail *t; |
dc6982ff | 2572 | ext4_lblk_t block = 0; |
3d0518f4 | 2573 | unsigned int blocksize = dir->i_sb->s_blocksize; |
b0336e8d | 2574 | int csum_size = 0; |
a774f9c2 | 2575 | int err; |
ac27a0ec | 2576 | |
9aa5d32b | 2577 | if (ext4_has_metadata_csum(dir->i_sb)) |
b0336e8d DW |
2578 | csum_size = sizeof(struct ext4_dir_entry_tail); |
2579 | ||
3c47d541 TM |
2580 | if (ext4_test_inode_state(inode, EXT4_STATE_MAY_INLINE_DATA)) { |
2581 | err = ext4_try_create_inline_dir(handle, dir, inode); | |
2582 | if (err < 0 && err != -ENOSPC) | |
2583 | goto out; | |
2584 | if (!err) | |
2585 | goto out; | |
2586 | } | |
2587 | ||
dc6982ff | 2588 | inode->i_size = 0; |
0f70b406 TT |
2589 | dir_block = ext4_append(handle, inode, &block); |
2590 | if (IS_ERR(dir_block)) | |
2591 | return PTR_ERR(dir_block); | |
a774f9c2 TM |
2592 | de = (struct ext4_dir_entry_2 *)dir_block->b_data; |
2593 | ext4_init_dot_dotdot(inode, de, blocksize, csum_size, dir->i_ino, 0); | |
2594 | set_nlink(inode, 2); | |
2595 | if (csum_size) { | |
2596 | t = EXT4_DIRENT_TAIL(dir_block->b_data, blocksize); | |
2597 | initialize_dirent_tail(t, blocksize); | |
2598 | } | |
2599 | ||
2600 | BUFFER_TRACE(dir_block, "call ext4_handle_dirty_metadata"); | |
2601 | err = ext4_handle_dirty_dirent_node(handle, inode, dir_block); | |
2602 | if (err) | |
2603 | goto out; | |
2604 | set_buffer_verified(dir_block); | |
2605 | out: | |
2606 | brelse(dir_block); | |
2607 | return err; | |
2608 | } | |
2609 | ||
2610 | static int ext4_mkdir(struct inode *dir, struct dentry *dentry, umode_t mode) | |
2611 | { | |
2612 | handle_t *handle; | |
2613 | struct inode *inode; | |
1139575a | 2614 | int err, credits, retries = 0; |
a774f9c2 | 2615 | |
f8628a14 | 2616 | if (EXT4_DIR_LINK_MAX(dir)) |
ac27a0ec DK |
2617 | return -EMLINK; |
2618 | ||
a7cdadee JK |
2619 | err = dquot_initialize(dir); |
2620 | if (err) | |
2621 | return err; | |
907f4554 | 2622 | |
1139575a | 2623 | credits = (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
eb9cc7e1 | 2624 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3); |
ac27a0ec | 2625 | retry: |
1139575a TT |
2626 | inode = ext4_new_inode_start_handle(dir, S_IFDIR | mode, |
2627 | &dentry->d_name, | |
2628 | 0, NULL, EXT4_HT_DIR, credits); | |
2629 | handle = ext4_journal_current_handle(); | |
ac27a0ec DK |
2630 | err = PTR_ERR(inode); |
2631 | if (IS_ERR(inode)) | |
2632 | goto out_stop; | |
2633 | ||
617ba13b MC |
2634 | inode->i_op = &ext4_dir_inode_operations; |
2635 | inode->i_fop = &ext4_dir_operations; | |
a774f9c2 | 2636 | err = ext4_init_new_dir(handle, dir, inode); |
dabd991f NK |
2637 | if (err) |
2638 | goto out_clear_inode; | |
2639 | err = ext4_mark_inode_dirty(handle, inode); | |
2640 | if (!err) | |
2641 | err = ext4_add_entry(handle, dentry, inode); | |
ac27a0ec | 2642 | if (err) { |
4cdeed86 AK |
2643 | out_clear_inode: |
2644 | clear_nlink(inode); | |
6b38e842 | 2645 | unlock_new_inode(inode); |
617ba13b | 2646 | ext4_mark_inode_dirty(handle, inode); |
af5bc92d | 2647 | iput(inode); |
ac27a0ec DK |
2648 | goto out_stop; |
2649 | } | |
f8628a14 | 2650 | ext4_inc_count(handle, dir); |
617ba13b | 2651 | ext4_update_dx_flag(dir); |
dabd991f NK |
2652 | err = ext4_mark_inode_dirty(handle, dir); |
2653 | if (err) | |
2654 | goto out_clear_inode; | |
1e2e547a | 2655 | d_instantiate_new(dentry, inode); |
1139575a TT |
2656 | if (IS_DIRSYNC(dir)) |
2657 | ext4_handle_sync(handle); | |
2658 | ||
ac27a0ec | 2659 | out_stop: |
1139575a TT |
2660 | if (handle) |
2661 | ext4_journal_stop(handle); | |
617ba13b | 2662 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) |
ac27a0ec DK |
2663 | goto retry; |
2664 | return err; | |
2665 | } | |
2666 | ||
2667 | /* | |
2668 | * routine to check that the specified directory is empty (for rmdir) | |
2669 | */ | |
a7550b30 | 2670 | bool ext4_empty_dir(struct inode *inode) |
ac27a0ec | 2671 | { |
498e5f24 | 2672 | unsigned int offset; |
af5bc92d TT |
2673 | struct buffer_head *bh; |
2674 | struct ext4_dir_entry_2 *de, *de1; | |
2675 | struct super_block *sb; | |
ac27a0ec | 2676 | |
61f86638 TM |
2677 | if (ext4_has_inline_data(inode)) { |
2678 | int has_inline_data = 1; | |
a7550b30 | 2679 | int ret; |
61f86638 | 2680 | |
a7550b30 | 2681 | ret = empty_inline_dir(inode, &has_inline_data); |
61f86638 | 2682 | if (has_inline_data) |
a7550b30 | 2683 | return ret; |
61f86638 TM |
2684 | } |
2685 | ||
ac27a0ec | 2686 | sb = inode->i_sb; |
dc6982ff TT |
2687 | if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2)) { |
2688 | EXT4_ERROR_INODE(inode, "invalid size"); | |
a7550b30 | 2689 | return true; |
ac27a0ec | 2690 | } |
dc6982ff TT |
2691 | bh = ext4_read_dirblock(inode, 0, EITHER); |
2692 | if (IS_ERR(bh)) | |
a7550b30 | 2693 | return true; |
dc6982ff | 2694 | |
617ba13b | 2695 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
3d0518f4 | 2696 | de1 = ext4_next_entry(de, sb->s_blocksize); |
ac27a0ec | 2697 | if (le32_to_cpu(de->inode) != inode->i_ino || |
b03a2f7e AD |
2698 | le32_to_cpu(de1->inode) == 0 || |
2699 | strcmp(".", de->name) || strcmp("..", de1->name)) { | |
2700 | ext4_warning_inode(inode, "directory missing '.' and/or '..'"); | |
af5bc92d | 2701 | brelse(bh); |
a7550b30 | 2702 | return true; |
ac27a0ec | 2703 | } |
3d0518f4 WY |
2704 | offset = ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize) + |
2705 | ext4_rec_len_from_disk(de1->rec_len, sb->s_blocksize); | |
2706 | de = ext4_next_entry(de1, sb->s_blocksize); | |
af5bc92d | 2707 | while (offset < inode->i_size) { |
236f5ecb | 2708 | if ((void *) de >= (void *) (bh->b_data+sb->s_blocksize)) { |
24676da4 | 2709 | unsigned int lblock; |
af5bc92d | 2710 | brelse(bh); |
24676da4 | 2711 | lblock = offset >> EXT4_BLOCK_SIZE_BITS(sb); |
dc6982ff TT |
2712 | bh = ext4_read_dirblock(inode, lblock, EITHER); |
2713 | if (IS_ERR(bh)) | |
a7550b30 | 2714 | return true; |
617ba13b | 2715 | de = (struct ext4_dir_entry_2 *) bh->b_data; |
ac27a0ec | 2716 | } |
226ba972 TM |
2717 | if (ext4_check_dir_entry(inode, NULL, de, bh, |
2718 | bh->b_data, bh->b_size, offset)) { | |
617ba13b | 2719 | de = (struct ext4_dir_entry_2 *)(bh->b_data + |
ac27a0ec DK |
2720 | sb->s_blocksize); |
2721 | offset = (offset | (sb->s_blocksize - 1)) + 1; | |
2722 | continue; | |
2723 | } | |
2724 | if (le32_to_cpu(de->inode)) { | |
af5bc92d | 2725 | brelse(bh); |
a7550b30 | 2726 | return false; |
ac27a0ec | 2727 | } |
3d0518f4 WY |
2728 | offset += ext4_rec_len_from_disk(de->rec_len, sb->s_blocksize); |
2729 | de = ext4_next_entry(de, sb->s_blocksize); | |
ac27a0ec | 2730 | } |
af5bc92d | 2731 | brelse(bh); |
a7550b30 | 2732 | return true; |
ac27a0ec DK |
2733 | } |
2734 | ||
d745a8c2 JK |
2735 | /* |
2736 | * ext4_orphan_add() links an unlinked or truncated inode into a list of | |
ac27a0ec DK |
2737 | * such inodes, starting at the superblock, in case we crash before the |
2738 | * file is closed/deleted, or in case the inode truncate spans multiple | |
2739 | * transactions and the last transaction is not recovered after a crash. | |
2740 | * | |
2741 | * At filesystem recovery time, we walk this list deleting unlinked | |
617ba13b | 2742 | * inodes and truncating linked inodes in ext4_orphan_cleanup(). |
d745a8c2 JK |
2743 | * |
2744 | * Orphan list manipulation functions must be called under i_mutex unless | |
2745 | * we are just creating the inode or deleting it. | |
ac27a0ec | 2746 | */ |
617ba13b | 2747 | int ext4_orphan_add(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2748 | { |
2749 | struct super_block *sb = inode->i_sb; | |
cd2c080c | 2750 | struct ext4_sb_info *sbi = EXT4_SB(sb); |
617ba13b | 2751 | struct ext4_iloc iloc; |
ac27a0ec | 2752 | int err = 0, rc; |
d745a8c2 | 2753 | bool dirty = false; |
ac27a0ec | 2754 | |
e2bfb088 | 2755 | if (!sbi->s_journal || is_bad_inode(inode)) |
0390131b FM |
2756 | return 0; |
2757 | ||
d745a8c2 | 2758 | WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) && |
5955102c | 2759 | !inode_is_locked(inode)); |
d745a8c2 JK |
2760 | /* |
2761 | * Exit early if inode already is on orphan list. This is a big speedup | |
2762 | * since we don't have to contend on the global s_orphan_lock. | |
2763 | */ | |
617ba13b | 2764 | if (!list_empty(&EXT4_I(inode)->i_orphan)) |
d745a8c2 | 2765 | return 0; |
ac27a0ec | 2766 | |
afb86178 LC |
2767 | /* |
2768 | * Orphan handling is only valid for files with data blocks | |
2769 | * being truncated, or files being unlinked. Note that we either | |
2770 | * hold i_mutex, or the inode can not be referenced from outside, | |
2771 | * so i_nlink should not be bumped due to race | |
ac27a0ec | 2772 | */ |
af5bc92d TT |
2773 | J_ASSERT((S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || |
2774 | S_ISLNK(inode->i_mode)) || inode->i_nlink == 0); | |
ac27a0ec | 2775 | |
cd2c080c JK |
2776 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
2777 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); | |
ac27a0ec | 2778 | if (err) |
d745a8c2 | 2779 | goto out; |
ac27a0ec | 2780 | |
617ba13b | 2781 | err = ext4_reserve_inode_write(handle, inode, &iloc); |
ac27a0ec | 2782 | if (err) |
d745a8c2 JK |
2783 | goto out; |
2784 | ||
2785 | mutex_lock(&sbi->s_orphan_lock); | |
6e3617e5 DM |
2786 | /* |
2787 | * Due to previous errors inode may be already a part of on-disk | |
2788 | * orphan list. If so skip on-disk list modification. | |
2789 | */ | |
d745a8c2 JK |
2790 | if (!NEXT_ORPHAN(inode) || NEXT_ORPHAN(inode) > |
2791 | (le32_to_cpu(sbi->s_es->s_inodes_count))) { | |
2792 | /* Insert this inode at the head of the on-disk orphan list */ | |
2793 | NEXT_ORPHAN(inode) = le32_to_cpu(sbi->s_es->s_last_orphan); | |
2794 | sbi->s_es->s_last_orphan = cpu_to_le32(inode->i_ino); | |
2795 | dirty = true; | |
2796 | } | |
2797 | list_add(&EXT4_I(inode)->i_orphan, &sbi->s_orphan); | |
2798 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2799 | |
d745a8c2 JK |
2800 | if (dirty) { |
2801 | err = ext4_handle_dirty_super(handle, sb); | |
2802 | rc = ext4_mark_iloc_dirty(handle, inode, &iloc); | |
2803 | if (!err) | |
2804 | err = rc; | |
2805 | if (err) { | |
2806 | /* | |
2807 | * We have to remove inode from in-memory list if | |
2808 | * addition to on disk orphan list failed. Stray orphan | |
2809 | * list entries can cause panics at unmount time. | |
2810 | */ | |
2811 | mutex_lock(&sbi->s_orphan_lock); | |
74177f55 | 2812 | list_del_init(&EXT4_I(inode)->i_orphan); |
d745a8c2 JK |
2813 | mutex_unlock(&sbi->s_orphan_lock); |
2814 | } | |
feaf264c VA |
2815 | } else |
2816 | brelse(iloc.bh); | |
2817 | ||
ac27a0ec DK |
2818 | jbd_debug(4, "superblock will point to %lu\n", inode->i_ino); |
2819 | jbd_debug(4, "orphan inode %lu will point to %d\n", | |
2820 | inode->i_ino, NEXT_ORPHAN(inode)); | |
d745a8c2 | 2821 | out: |
cd2c080c | 2822 | ext4_std_error(sb, err); |
ac27a0ec DK |
2823 | return err; |
2824 | } | |
2825 | ||
2826 | /* | |
617ba13b | 2827 | * ext4_orphan_del() removes an unlinked or truncated inode from the list |
ac27a0ec DK |
2828 | * of such inodes stored on disk, because it is finally being cleaned up. |
2829 | */ | |
617ba13b | 2830 | int ext4_orphan_del(handle_t *handle, struct inode *inode) |
ac27a0ec DK |
2831 | { |
2832 | struct list_head *prev; | |
617ba13b | 2833 | struct ext4_inode_info *ei = EXT4_I(inode); |
cd2c080c | 2834 | struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); |
498e5f24 | 2835 | __u32 ino_next; |
617ba13b | 2836 | struct ext4_iloc iloc; |
ac27a0ec DK |
2837 | int err = 0; |
2838 | ||
cd2c080c | 2839 | if (!sbi->s_journal && !(sbi->s_mount_state & EXT4_ORPHAN_FS)) |
0390131b FM |
2840 | return 0; |
2841 | ||
d745a8c2 | 2842 | WARN_ON_ONCE(!(inode->i_state & (I_NEW | I_FREEING)) && |
5955102c | 2843 | !inode_is_locked(inode)); |
d745a8c2 | 2844 | /* Do this quick check before taking global s_orphan_lock. */ |
3b9d4ed2 | 2845 | if (list_empty(&ei->i_orphan)) |
d745a8c2 | 2846 | return 0; |
ac27a0ec | 2847 | |
d745a8c2 JK |
2848 | if (handle) { |
2849 | /* Grab inode buffer early before taking global s_orphan_lock */ | |
2850 | err = ext4_reserve_inode_write(handle, inode, &iloc); | |
2851 | } | |
ac27a0ec | 2852 | |
d745a8c2 | 2853 | mutex_lock(&sbi->s_orphan_lock); |
ac27a0ec DK |
2854 | jbd_debug(4, "remove inode %lu from orphan list\n", inode->i_ino); |
2855 | ||
d745a8c2 | 2856 | prev = ei->i_orphan.prev; |
ac27a0ec DK |
2857 | list_del_init(&ei->i_orphan); |
2858 | ||
2859 | /* If we're on an error path, we may not have a valid | |
2860 | * transaction handle with which to update the orphan list on | |
2861 | * disk, but we still need to remove the inode from the linked | |
2862 | * list in memory. */ | |
d745a8c2 JK |
2863 | if (!handle || err) { |
2864 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2865 | goto out_err; |
d745a8c2 | 2866 | } |
ac27a0ec | 2867 | |
d745a8c2 | 2868 | ino_next = NEXT_ORPHAN(inode); |
ac27a0ec | 2869 | if (prev == &sbi->s_orphan) { |
498e5f24 | 2870 | jbd_debug(4, "superblock will point to %u\n", ino_next); |
ac27a0ec | 2871 | BUFFER_TRACE(sbi->s_sbh, "get_write_access"); |
617ba13b | 2872 | err = ext4_journal_get_write_access(handle, sbi->s_sbh); |
d745a8c2 JK |
2873 | if (err) { |
2874 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2875 | goto out_brelse; |
d745a8c2 | 2876 | } |
ac27a0ec | 2877 | sbi->s_es->s_last_orphan = cpu_to_le32(ino_next); |
d745a8c2 | 2878 | mutex_unlock(&sbi->s_orphan_lock); |
b50924c2 | 2879 | err = ext4_handle_dirty_super(handle, inode->i_sb); |
ac27a0ec | 2880 | } else { |
617ba13b | 2881 | struct ext4_iloc iloc2; |
ac27a0ec | 2882 | struct inode *i_prev = |
617ba13b | 2883 | &list_entry(prev, struct ext4_inode_info, i_orphan)->vfs_inode; |
ac27a0ec | 2884 | |
498e5f24 | 2885 | jbd_debug(4, "orphan inode %lu will point to %u\n", |
ac27a0ec | 2886 | i_prev->i_ino, ino_next); |
617ba13b | 2887 | err = ext4_reserve_inode_write(handle, i_prev, &iloc2); |
d745a8c2 JK |
2888 | if (err) { |
2889 | mutex_unlock(&sbi->s_orphan_lock); | |
ac27a0ec | 2890 | goto out_brelse; |
d745a8c2 | 2891 | } |
ac27a0ec | 2892 | NEXT_ORPHAN(i_prev) = ino_next; |
617ba13b | 2893 | err = ext4_mark_iloc_dirty(handle, i_prev, &iloc2); |
d745a8c2 | 2894 | mutex_unlock(&sbi->s_orphan_lock); |
ac27a0ec DK |
2895 | } |
2896 | if (err) | |
2897 | goto out_brelse; | |
2898 | NEXT_ORPHAN(inode) = 0; | |
617ba13b | 2899 | err = ext4_mark_iloc_dirty(handle, inode, &iloc); |
ac27a0ec | 2900 | out_err: |
617ba13b | 2901 | ext4_std_error(inode->i_sb, err); |
ac27a0ec DK |
2902 | return err; |
2903 | ||
2904 | out_brelse: | |
2905 | brelse(iloc.bh); | |
2906 | goto out_err; | |
2907 | } | |
2908 | ||
af5bc92d | 2909 | static int ext4_rmdir(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2910 | { |
2911 | int retval; | |
af5bc92d TT |
2912 | struct inode *inode; |
2913 | struct buffer_head *bh; | |
2914 | struct ext4_dir_entry_2 *de; | |
8dcfaad2 | 2915 | handle_t *handle = NULL; |
ac27a0ec | 2916 | |
0db1ff22 TT |
2917 | if (unlikely(ext4_forced_shutdown(EXT4_SB(dir->i_sb)))) |
2918 | return -EIO; | |
2919 | ||
ac27a0ec DK |
2920 | /* Initialize quotas before so that eventual writes go in |
2921 | * separate transaction */ | |
a7cdadee JK |
2922 | retval = dquot_initialize(dir); |
2923 | if (retval) | |
2924 | return retval; | |
2925 | retval = dquot_initialize(d_inode(dentry)); | |
2926 | if (retval) | |
2927 | return retval; | |
907f4554 | 2928 | |
ac27a0ec | 2929 | retval = -ENOENT; |
32f7f22c | 2930 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 TT |
2931 | if (IS_ERR(bh)) |
2932 | return PTR_ERR(bh); | |
ac27a0ec DK |
2933 | if (!bh) |
2934 | goto end_rmdir; | |
2935 | ||
2b0143b5 | 2936 | inode = d_inode(dentry); |
ac27a0ec | 2937 | |
6a797d27 | 2938 | retval = -EFSCORRUPTED; |
ac27a0ec DK |
2939 | if (le32_to_cpu(de->inode) != inode->i_ino) |
2940 | goto end_rmdir; | |
2941 | ||
2942 | retval = -ENOTEMPTY; | |
e875a2dd | 2943 | if (!ext4_empty_dir(inode)) |
ac27a0ec DK |
2944 | goto end_rmdir; |
2945 | ||
8dcfaad2 | 2946 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
64044abf | 2947 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb)); |
8dcfaad2 TT |
2948 | if (IS_ERR(handle)) { |
2949 | retval = PTR_ERR(handle); | |
2950 | handle = NULL; | |
2951 | goto end_rmdir; | |
2952 | } | |
2953 | ||
2954 | if (IS_DIRSYNC(dir)) | |
2955 | ext4_handle_sync(handle); | |
2956 | ||
617ba13b | 2957 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
2958 | if (retval) |
2959 | goto end_rmdir; | |
f8628a14 | 2960 | if (!EXT4_DIR_LINK_EMPTY(inode)) |
b03a2f7e AD |
2961 | ext4_warning_inode(inode, |
2962 | "empty directory '%.*s' has too many links (%u)", | |
2963 | dentry->d_name.len, dentry->d_name.name, | |
af5bc92d | 2964 | inode->i_nlink); |
ee73f9a5 | 2965 | inode_inc_iversion(inode); |
ac27a0ec DK |
2966 | clear_nlink(inode); |
2967 | /* There's no need to set i_disksize: the fact that i_nlink is | |
2968 | * zero will ensure that the right thing happens during any | |
2969 | * recovery. */ | |
2970 | inode->i_size = 0; | |
617ba13b | 2971 | ext4_orphan_add(handle, inode); |
eeca7ea1 | 2972 | inode->i_ctime = dir->i_ctime = dir->i_mtime = current_time(inode); |
617ba13b | 2973 | ext4_mark_inode_dirty(handle, inode); |
f8628a14 | 2974 | ext4_dec_count(handle, dir); |
617ba13b MC |
2975 | ext4_update_dx_flag(dir); |
2976 | ext4_mark_inode_dirty(handle, dir); | |
ac27a0ec DK |
2977 | |
2978 | end_rmdir: | |
af5bc92d | 2979 | brelse(bh); |
8dcfaad2 TT |
2980 | if (handle) |
2981 | ext4_journal_stop(handle); | |
ac27a0ec DK |
2982 | return retval; |
2983 | } | |
2984 | ||
af5bc92d | 2985 | static int ext4_unlink(struct inode *dir, struct dentry *dentry) |
ac27a0ec DK |
2986 | { |
2987 | int retval; | |
af5bc92d TT |
2988 | struct inode *inode; |
2989 | struct buffer_head *bh; | |
2990 | struct ext4_dir_entry_2 *de; | |
931b6864 | 2991 | handle_t *handle = NULL; |
ac27a0ec | 2992 | |
0db1ff22 TT |
2993 | if (unlikely(ext4_forced_shutdown(EXT4_SB(dir->i_sb)))) |
2994 | return -EIO; | |
2995 | ||
0562e0ba | 2996 | trace_ext4_unlink_enter(dir, dentry); |
ac27a0ec DK |
2997 | /* Initialize quotas before so that eventual writes go |
2998 | * in separate transaction */ | |
a7cdadee JK |
2999 | retval = dquot_initialize(dir); |
3000 | if (retval) | |
3001 | return retval; | |
3002 | retval = dquot_initialize(d_inode(dentry)); | |
3003 | if (retval) | |
3004 | return retval; | |
907f4554 | 3005 | |
ac27a0ec | 3006 | retval = -ENOENT; |
32f7f22c | 3007 | bh = ext4_find_entry(dir, &dentry->d_name, &de, NULL); |
36de9286 TT |
3008 | if (IS_ERR(bh)) |
3009 | return PTR_ERR(bh); | |
ac27a0ec DK |
3010 | if (!bh) |
3011 | goto end_unlink; | |
3012 | ||
2b0143b5 | 3013 | inode = d_inode(dentry); |
ac27a0ec | 3014 | |
6a797d27 | 3015 | retval = -EFSCORRUPTED; |
ac27a0ec DK |
3016 | if (le32_to_cpu(de->inode) != inode->i_ino) |
3017 | goto end_unlink; | |
3018 | ||
931b6864 | 3019 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
64044abf | 3020 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb)); |
931b6864 TT |
3021 | if (IS_ERR(handle)) { |
3022 | retval = PTR_ERR(handle); | |
3023 | handle = NULL; | |
3024 | goto end_unlink; | |
3025 | } | |
3026 | ||
3027 | if (IS_DIRSYNC(dir)) | |
3028 | ext4_handle_sync(handle); | |
3029 | ||
b03a2f7e AD |
3030 | if (inode->i_nlink == 0) { |
3031 | ext4_warning_inode(inode, "Deleting file '%.*s' with no links", | |
3032 | dentry->d_name.len, dentry->d_name.name); | |
bfe86848 | 3033 | set_nlink(inode, 1); |
ac27a0ec | 3034 | } |
617ba13b | 3035 | retval = ext4_delete_entry(handle, dir, de, bh); |
ac27a0ec DK |
3036 | if (retval) |
3037 | goto end_unlink; | |
eeca7ea1 | 3038 | dir->i_ctime = dir->i_mtime = current_time(dir); |
617ba13b MC |
3039 | ext4_update_dx_flag(dir); |
3040 | ext4_mark_inode_dirty(handle, dir); | |
825f1481 | 3041 | drop_nlink(inode); |
ac27a0ec | 3042 | if (!inode->i_nlink) |
617ba13b | 3043 | ext4_orphan_add(handle, inode); |
eeca7ea1 | 3044 | inode->i_ctime = current_time(inode); |
617ba13b | 3045 | ext4_mark_inode_dirty(handle, inode); |
ac27a0ec DK |
3046 | |
3047 | end_unlink: | |
af5bc92d | 3048 | brelse(bh); |
931b6864 TT |
3049 | if (handle) |
3050 | ext4_journal_stop(handle); | |
0562e0ba | 3051 | trace_ext4_unlink_exit(dentry, retval); |
ac27a0ec DK |
3052 | return retval; |
3053 | } | |
3054 | ||
af5bc92d TT |
3055 | static int ext4_symlink(struct inode *dir, |
3056 | struct dentry *dentry, const char *symname) | |
ac27a0ec DK |
3057 | { |
3058 | handle_t *handle; | |
af5bc92d | 3059 | struct inode *inode; |
f348c252 | 3060 | int err, len = strlen(symname); |
df5e6223 | 3061 | int credits; |
a7550b30 | 3062 | struct fscrypt_str disk_link; |
ac27a0ec | 3063 | |
0db1ff22 TT |
3064 | if (unlikely(ext4_forced_shutdown(EXT4_SB(dir->i_sb)))) |
3065 | return -EIO; | |
3066 | ||
78e1060c EB |
3067 | err = fscrypt_prepare_symlink(dir, symname, len, dir->i_sb->s_blocksize, |
3068 | &disk_link); | |
3069 | if (err) | |
3070 | return err; | |
ac27a0ec | 3071 | |
a7cdadee JK |
3072 | err = dquot_initialize(dir); |
3073 | if (err) | |
78e1060c | 3074 | return err; |
907f4554 | 3075 | |
f348c252 | 3076 | if ((disk_link.len > EXT4_N_BLOCKS * 4)) { |
df5e6223 JK |
3077 | /* |
3078 | * For non-fast symlinks, we just allocate inode and put it on | |
3079 | * orphan list in the first transaction => we need bitmap, | |
8c208719 ES |
3080 | * group descriptor, sb, inode block, quota blocks, and |
3081 | * possibly selinux xattr blocks. | |
df5e6223 | 3082 | */ |
8c208719 ES |
3083 | credits = 4 + EXT4_MAXQUOTAS_INIT_BLOCKS(dir->i_sb) + |
3084 | EXT4_XATTR_TRANS_BLOCKS; | |
df5e6223 JK |
3085 | } else { |
3086 | /* | |
3087 | * Fast symlink. We have to add entry to directory | |
3088 | * (EXT4_DATA_TRANS_BLOCKS + EXT4_INDEX_EXTRA_TRANS_BLOCKS), | |
3089 | * allocate new inode (bitmap, group descriptor, inode block, | |
3090 | * quota blocks, sb is already counted in previous macros). | |
3091 | */ | |
3092 | credits = EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
eb9cc7e1 | 3093 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 3; |
df5e6223 | 3094 | } |
f348c252 | 3095 | |
1139575a TT |
3096 | inode = ext4_new_inode_start_handle(dir, S_IFLNK|S_IRWXUGO, |
3097 | &dentry->d_name, 0, NULL, | |
3098 | EXT4_HT_DIR, credits); | |
3099 | handle = ext4_journal_current_handle(); | |
f348c252 TT |
3100 | if (IS_ERR(inode)) { |
3101 | if (handle) | |
3102 | ext4_journal_stop(handle); | |
78e1060c | 3103 | return PTR_ERR(inode); |
f348c252 TT |
3104 | } |
3105 | ||
78e1060c EB |
3106 | if (IS_ENCRYPTED(inode)) { |
3107 | err = fscrypt_encrypt_symlink(inode, symname, len, &disk_link); | |
ef1eb3aa | 3108 | if (err) |
f348c252 | 3109 | goto err_drop_inode; |
a7a67e8a | 3110 | inode->i_op = &ext4_encrypted_symlink_inode_operations; |
f348c252 | 3111 | } |
ac27a0ec | 3112 | |
f348c252 | 3113 | if ((disk_link.len > EXT4_N_BLOCKS * 4)) { |
78e1060c | 3114 | if (!IS_ENCRYPTED(inode)) |
a7a67e8a | 3115 | inode->i_op = &ext4_symlink_inode_operations; |
21fc61c7 | 3116 | inode_nohighmem(inode); |
617ba13b | 3117 | ext4_set_aops(inode); |
ac27a0ec | 3118 | /* |
df5e6223 JK |
3119 | * We cannot call page_symlink() with transaction started |
3120 | * because it calls into ext4_write_begin() which can wait | |
3121 | * for transaction commit if we are running out of space | |
3122 | * and thus we deadlock. So we have to stop transaction now | |
3123 | * and restart it when symlink contents is written. | |
3124 | * | |
3125 | * To keep fs consistent in case of crash, we have to put inode | |
3126 | * to orphan list in the mean time. | |
ac27a0ec | 3127 | */ |
df5e6223 JK |
3128 | drop_nlink(inode); |
3129 | err = ext4_orphan_add(handle, inode); | |
3130 | ext4_journal_stop(handle); | |
f348c252 | 3131 | handle = NULL; |
df5e6223 JK |
3132 | if (err) |
3133 | goto err_drop_inode; | |
f348c252 | 3134 | err = __page_symlink(inode, disk_link.name, disk_link.len, 1); |
df5e6223 JK |
3135 | if (err) |
3136 | goto err_drop_inode; | |
3137 | /* | |
3138 | * Now inode is being linked into dir (EXT4_DATA_TRANS_BLOCKS | |
3139 | * + EXT4_INDEX_EXTRA_TRANS_BLOCKS), inode is also modified | |
3140 | */ | |
9924a92a | 3141 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
df5e6223 JK |
3142 | EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + |
3143 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 1); | |
3144 | if (IS_ERR(handle)) { | |
3145 | err = PTR_ERR(handle); | |
f348c252 | 3146 | handle = NULL; |
df5e6223 JK |
3147 | goto err_drop_inode; |
3148 | } | |
0ce8c010 | 3149 | set_nlink(inode, 1); |
df5e6223 | 3150 | err = ext4_orphan_del(handle, inode); |
f348c252 | 3151 | if (err) |
df5e6223 | 3152 | goto err_drop_inode; |
ac27a0ec | 3153 | } else { |
e65187e6 | 3154 | /* clear the extent format for fast symlink */ |
12e9b892 | 3155 | ext4_clear_inode_flag(inode, EXT4_INODE_EXTENTS); |
78e1060c | 3156 | if (!IS_ENCRYPTED(inode)) { |
a7a67e8a | 3157 | inode->i_op = &ext4_fast_symlink_inode_operations; |
75e7566b AV |
3158 | inode->i_link = (char *)&EXT4_I(inode)->i_data; |
3159 | } | |
f348c252 TT |
3160 | memcpy((char *)&EXT4_I(inode)->i_data, disk_link.name, |
3161 | disk_link.len); | |
3162 | inode->i_size = disk_link.len - 1; | |
ac27a0ec | 3163 | } |
617ba13b MC |
3164 | EXT4_I(inode)->i_disksize = inode->i_size; |
3165 | err = ext4_add_nondir(handle, dentry, inode); | |
1139575a TT |
3166 | if (!err && IS_DIRSYNC(dir)) |
3167 | ext4_handle_sync(handle); | |
3168 | ||
1139575a TT |
3169 | if (handle) |
3170 | ext4_journal_stop(handle); | |
78e1060c EB |
3171 | goto out_free_encrypted_link; |
3172 | ||
df5e6223 | 3173 | err_drop_inode: |
f348c252 TT |
3174 | if (handle) |
3175 | ext4_journal_stop(handle); | |
f348c252 | 3176 | clear_nlink(inode); |
df5e6223 JK |
3177 | unlock_new_inode(inode); |
3178 | iput(inode); | |
78e1060c EB |
3179 | out_free_encrypted_link: |
3180 | if (disk_link.name != (unsigned char *)symname) | |
3181 | kfree(disk_link.name); | |
df5e6223 | 3182 | return err; |
ac27a0ec DK |
3183 | } |
3184 | ||
af5bc92d TT |
3185 | static int ext4_link(struct dentry *old_dentry, |
3186 | struct inode *dir, struct dentry *dentry) | |
ac27a0ec DK |
3187 | { |
3188 | handle_t *handle; | |
2b0143b5 | 3189 | struct inode *inode = d_inode(old_dentry); |
ac27a0ec DK |
3190 | int err, retries = 0; |
3191 | ||
b05ab1dc | 3192 | if (inode->i_nlink >= EXT4_LINK_MAX) |
ac27a0ec | 3193 | return -EMLINK; |
69725181 EB |
3194 | |
3195 | err = fscrypt_prepare_link(old_dentry, dir, dentry); | |
3196 | if (err) | |
3197 | return err; | |
040cb378 | 3198 | |
a794df0e CIK |
3199 | if ((ext4_test_inode_flag(dir, EXT4_INODE_PROJINHERIT)) && |
3200 | (!projid_eq(EXT4_I(dir)->i_projid, | |
3201 | EXT4_I(old_dentry->d_inode)->i_projid))) | |
040cb378 LX |
3202 | return -EXDEV; |
3203 | ||
a7cdadee JK |
3204 | err = dquot_initialize(dir); |
3205 | if (err) | |
3206 | return err; | |
907f4554 | 3207 | |
ac27a0ec | 3208 | retry: |
9924a92a TT |
3209 | handle = ext4_journal_start(dir, EXT4_HT_DIR, |
3210 | (EXT4_DATA_TRANS_BLOCKS(dir->i_sb) + | |
af51a2ac | 3211 | EXT4_INDEX_EXTRA_TRANS_BLOCKS) + 1); |
ac27a0ec DK |
3212 | if (IS_ERR(handle)) |
3213 | return PTR_ERR(handle); | |
3214 | ||
3215 | if (IS_DIRSYNC(dir)) | |
0390131b | 3216 | ext4_handle_sync(handle); |
ac27a0ec | 3217 | |
eeca7ea1 | 3218 | inode->i_ctime = current_time(inode); |
f8628a14 | 3219 | ext4_inc_count(handle, inode); |
7de9c6ee | 3220 | ihold(inode); |
ac27a0ec | 3221 | |
6b38e842 AV |
3222 | err = ext4_add_entry(handle, dentry, inode); |
3223 | if (!err) { | |
3224 | ext4_mark_inode_dirty(handle, inode); | |
af51a2ac AV |
3225 | /* this can happen only for tmpfile being |
3226 | * linked the first time | |
3227 | */ | |
3228 | if (inode->i_nlink == 1) | |
3229 | ext4_orphan_del(handle, inode); | |
6b38e842 AV |
3230 | d_instantiate(dentry, inode); |
3231 | } else { | |
3232 | drop_nlink(inode); | |
3233 | iput(inode); | |
3234 | } | |
617ba13b MC |
3235 | ext4_journal_stop(handle); |
3236 | if (err == -ENOSPC && ext4_should_retry_alloc(dir->i_sb, &retries)) | |
ac27a0ec DK |
3237 | goto retry; |
3238 | return err; | |
3239 | } | |
3240 | ||
32f7f22c TM |
3241 | |
3242 | /* | |
3243 | * Try to find buffer head where contains the parent block. | |
3244 | * It should be the inode block if it is inlined or the 1st block | |
3245 | * if it is a normal dir. | |
3246 | */ | |
3247 | static struct buffer_head *ext4_get_first_dir_block(handle_t *handle, | |
3248 | struct inode *inode, | |
3249 | int *retval, | |
3250 | struct ext4_dir_entry_2 **parent_de, | |
3251 | int *inlined) | |
3252 | { | |
3253 | struct buffer_head *bh; | |
3254 | ||
3255 | if (!ext4_has_inline_data(inode)) { | |
dc6982ff TT |
3256 | bh = ext4_read_dirblock(inode, 0, EITHER); |
3257 | if (IS_ERR(bh)) { | |
3258 | *retval = PTR_ERR(bh); | |
32f7f22c TM |
3259 | return NULL; |
3260 | } | |
3261 | *parent_de = ext4_next_entry( | |
3262 | (struct ext4_dir_entry_2 *)bh->b_data, | |
3263 | inode->i_sb->s_blocksize); | |
3264 | return bh; | |
3265 | } | |
3266 | ||
3267 | *inlined = 1; | |
3268 | return ext4_get_first_inline_block(inode, parent_de, retval); | |
3269 | } | |
ac27a0ec | 3270 | |
c0d268c3 MS |
3271 | struct ext4_renament { |
3272 | struct inode *dir; | |
3273 | struct dentry *dentry; | |
3274 | struct inode *inode; | |
bd42998a MS |
3275 | bool is_dir; |
3276 | int dir_nlink_delta; | |
c0d268c3 MS |
3277 | |
3278 | /* entry for "dentry" */ | |
3279 | struct buffer_head *bh; | |
3280 | struct ext4_dir_entry_2 *de; | |
3281 | int inlined; | |
3282 | ||
3283 | /* entry for ".." in inode if it's a directory */ | |
3284 | struct buffer_head *dir_bh; | |
3285 | struct ext4_dir_entry_2 *parent_de; | |
3286 | int dir_inlined; | |
3287 | }; | |
3288 | ||
bd1af145 MS |
3289 | static int ext4_rename_dir_prepare(handle_t *handle, struct ext4_renament *ent) |
3290 | { | |
3291 | int retval; | |
3292 | ||
3293 | ent->dir_bh = ext4_get_first_dir_block(handle, ent->inode, | |
3294 | &retval, &ent->parent_de, | |
3295 | &ent->dir_inlined); | |
3296 | if (!ent->dir_bh) | |
3297 | return retval; | |
3298 | if (le32_to_cpu(ent->parent_de->inode) != ent->dir->i_ino) | |
6a797d27 | 3299 | return -EFSCORRUPTED; |
bd1af145 MS |
3300 | BUFFER_TRACE(ent->dir_bh, "get_write_access"); |
3301 | return ext4_journal_get_write_access(handle, ent->dir_bh); | |
3302 | } | |
3303 | ||
3304 | static int ext4_rename_dir_finish(handle_t *handle, struct ext4_renament *ent, | |
3305 | unsigned dir_ino) | |
3306 | { | |
3307 | int retval; | |
3308 | ||
3309 | ent->parent_de->inode = cpu_to_le32(dir_ino); | |
3310 | BUFFER_TRACE(ent->dir_bh, "call ext4_handle_dirty_metadata"); | |
3311 | if (!ent->dir_inlined) { | |
3312 | if (is_dx(ent->inode)) { | |
3313 | retval = ext4_handle_dirty_dx_node(handle, | |
3314 | ent->inode, | |
3315 | ent->dir_bh); | |
3316 | } else { | |
3317 | retval = ext4_handle_dirty_dirent_node(handle, | |
3318 | ent->inode, | |
3319 | ent->dir_bh); | |
3320 | } | |
3321 | } else { | |
3322 | retval = ext4_mark_inode_dirty(handle, ent->inode); | |
3323 | } | |
3324 | if (retval) { | |
3325 | ext4_std_error(ent->dir->i_sb, retval); | |
3326 | return retval; | |
3327 | } | |
3328 | return 0; | |
3329 | } | |
3330 | ||
3331 | static int ext4_setent(handle_t *handle, struct ext4_renament *ent, | |
3332 | unsigned ino, unsigned file_type) | |
3333 | { | |
3334 | int retval; | |
3335 | ||
3336 | BUFFER_TRACE(ent->bh, "get write access"); | |
3337 | retval = ext4_journal_get_write_access(handle, ent->bh); | |
3338 | if (retval) | |
3339 | return retval; | |
3340 | ent->de->inode = cpu_to_le32(ino); | |
e2b911c5 | 3341 | if (ext4_has_feature_filetype(ent->dir->i_sb)) |
bd1af145 | 3342 | ent->de->file_type = file_type; |
ee73f9a5 | 3343 | inode_inc_iversion(ent->dir); |
bd1af145 | 3344 | ent->dir->i_ctime = ent->dir->i_mtime = |
eeca7ea1 | 3345 | current_time(ent->dir); |
bd1af145 MS |
3346 | ext4_mark_inode_dirty(handle, ent->dir); |
3347 | BUFFER_TRACE(ent->bh, "call ext4_handle_dirty_metadata"); | |
3348 | if (!ent->inlined) { | |
3349 | retval = ext4_handle_dirty_dirent_node(handle, | |
3350 | ent->dir, ent->bh); | |
3351 | if (unlikely(retval)) { | |
3352 | ext4_std_error(ent->dir->i_sb, retval); | |
3353 | return retval; | |
3354 | } | |
3355 | } | |
3356 | brelse(ent->bh); | |
3357 | ent->bh = NULL; | |
3358 | ||
3359 | return 0; | |
3360 | } | |
3361 | ||
3362 | static int ext4_find_delete_entry(handle_t *handle, struct inode *dir, | |
3363 | const struct qstr *d_name) | |
3364 | { | |
3365 | int retval = -ENOENT; | |
3366 | struct buffer_head *bh; | |
3367 | struct ext4_dir_entry_2 *de; | |
3368 | ||
3369 | bh = ext4_find_entry(dir, d_name, &de, NULL); | |
36de9286 TT |
3370 | if (IS_ERR(bh)) |
3371 | return PTR_ERR(bh); | |
bd1af145 MS |
3372 | if (bh) { |
3373 | retval = ext4_delete_entry(handle, dir, de, bh); | |
3374 | brelse(bh); | |
3375 | } | |
3376 | return retval; | |
3377 | } | |
3378 | ||
d80d448c DW |
3379 | static void ext4_rename_delete(handle_t *handle, struct ext4_renament *ent, |
3380 | int force_reread) | |
bd1af145 MS |
3381 | { |
3382 | int retval; | |
3383 | /* | |
3384 | * ent->de could have moved from under us during htree split, so make | |
3385 | * sure that we are deleting the right entry. We might also be pointing | |
3386 | * to a stale entry in the unused part of ent->bh so just checking inum | |
3387 | * and the name isn't enough. | |
3388 | */ | |
3389 | if (le32_to_cpu(ent->de->inode) != ent->inode->i_ino || | |
3390 | ent->de->name_len != ent->dentry->d_name.len || | |
3391 | strncmp(ent->de->name, ent->dentry->d_name.name, | |
d80d448c DW |
3392 | ent->de->name_len) || |
3393 | force_reread) { | |
bd1af145 MS |
3394 | retval = ext4_find_delete_entry(handle, ent->dir, |
3395 | &ent->dentry->d_name); | |
3396 | } else { | |
3397 | retval = ext4_delete_entry(handle, ent->dir, ent->de, ent->bh); | |
3398 | if (retval == -ENOENT) { | |
3399 | retval = ext4_find_delete_entry(handle, ent->dir, | |
3400 | &ent->dentry->d_name); | |
3401 | } | |
3402 | } | |
3403 | ||
3404 | if (retval) { | |
b03a2f7e AD |
3405 | ext4_warning_inode(ent->dir, |
3406 | "Deleting old file: nlink %d, error=%d", | |
3407 | ent->dir->i_nlink, retval); | |
bd1af145 MS |
3408 | } |
3409 | } | |
3410 | ||
bd42998a MS |
3411 | static void ext4_update_dir_count(handle_t *handle, struct ext4_renament *ent) |
3412 | { | |
3413 | if (ent->dir_nlink_delta) { | |
3414 | if (ent->dir_nlink_delta == -1) | |
3415 | ext4_dec_count(handle, ent->dir); | |
3416 | else | |
3417 | ext4_inc_count(handle, ent->dir); | |
3418 | ext4_mark_inode_dirty(handle, ent->dir); | |
3419 | } | |
3420 | } | |
3421 | ||
cd808dec MS |
3422 | static struct inode *ext4_whiteout_for_rename(struct ext4_renament *ent, |
3423 | int credits, handle_t **h) | |
3424 | { | |
3425 | struct inode *wh; | |
3426 | handle_t *handle; | |
3427 | int retries = 0; | |
3428 | ||
3429 | /* | |
3430 | * for inode block, sb block, group summaries, | |
3431 | * and inode bitmap | |
3432 | */ | |
3433 | credits += (EXT4_MAXQUOTAS_TRANS_BLOCKS(ent->dir->i_sb) + | |
3434 | EXT4_XATTR_TRANS_BLOCKS + 4); | |
3435 | retry: | |
3436 | wh = ext4_new_inode_start_handle(ent->dir, S_IFCHR | WHITEOUT_MODE, | |
3437 | &ent->dentry->d_name, 0, NULL, | |
3438 | EXT4_HT_DIR, credits); | |
3439 | ||
3440 | handle = ext4_journal_current_handle(); | |
3441 | if (IS_ERR(wh)) { | |
3442 | if (handle) | |
3443 | ext4_journal_stop(handle); | |
3444 | if (PTR_ERR(wh) == -ENOSPC && | |
3445 | ext4_should_retry_alloc(ent->dir->i_sb, &retries)) | |
3446 | goto retry; | |
3447 | } else { | |
3448 | *h = handle; | |
3449 | init_special_inode(wh, wh->i_mode, WHITEOUT_DEV); | |
3450 | wh->i_op = &ext4_special_inode_operations; | |
3451 | } | |
3452 | return wh; | |
3453 | } | |
3454 | ||
ac27a0ec DK |
3455 | /* |
3456 | * Anybody can rename anything with this: the permission checks are left to the | |
3457 | * higher-level routines. | |
0e202704 TT |
3458 | * |
3459 | * n.b. old_{dentry,inode) refers to the source dentry/inode | |
3460 | * while new_{dentry,inode) refers to the destination dentry/inode | |
3461 | * This comes from rename(const char *oldpath, const char *newpath) | |
ac27a0ec | 3462 | */ |
af5bc92d | 3463 | static int ext4_rename(struct inode *old_dir, struct dentry *old_dentry, |
cd808dec MS |
3464 | struct inode *new_dir, struct dentry *new_dentry, |
3465 | unsigned int flags) | |
ac27a0ec | 3466 | { |
5b61de75 | 3467 | handle_t *handle = NULL; |
c0d268c3 MS |
3468 | struct ext4_renament old = { |
3469 | .dir = old_dir, | |
3470 | .dentry = old_dentry, | |
2b0143b5 | 3471 | .inode = d_inode(old_dentry), |
c0d268c3 MS |
3472 | }; |
3473 | struct ext4_renament new = { | |
3474 | .dir = new_dir, | |
3475 | .dentry = new_dentry, | |
2b0143b5 | 3476 | .inode = d_inode(new_dentry), |
c0d268c3 | 3477 | }; |
d80d448c | 3478 | int force_reread; |
0e202704 | 3479 | int retval; |
cd808dec MS |
3480 | struct inode *whiteout = NULL; |
3481 | int credits; | |
3482 | u8 old_file_type; | |
907f4554 | 3483 | |
b50282f3 TT |
3484 | if (new.inode && new.inode->i_nlink == 0) { |
3485 | EXT4_ERROR_INODE(new.inode, | |
3486 | "target of rename is already freed"); | |
3487 | return -EFSCORRUPTED; | |
3488 | } | |
3489 | ||
040cb378 LX |
3490 | if ((ext4_test_inode_flag(new_dir, EXT4_INODE_PROJINHERIT)) && |
3491 | (!projid_eq(EXT4_I(new_dir)->i_projid, | |
3492 | EXT4_I(old_dentry->d_inode)->i_projid))) | |
3493 | return -EXDEV; | |
3494 | ||
a7cdadee JK |
3495 | retval = dquot_initialize(old.dir); |
3496 | if (retval) | |
3497 | return retval; | |
3498 | retval = dquot_initialize(new.dir); | |
3499 | if (retval) | |
3500 | return retval; | |
ac27a0ec DK |
3501 | |
3502 | /* Initialize quotas before so that eventual writes go | |
3503 | * in separate transaction */ | |
a7cdadee JK |
3504 | if (new.inode) { |
3505 | retval = dquot_initialize(new.inode); | |
3506 | if (retval) | |
3507 | return retval; | |
3508 | } | |
ac27a0ec | 3509 | |
c0d268c3 | 3510 | old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, &old.de, NULL); |
36de9286 TT |
3511 | if (IS_ERR(old.bh)) |
3512 | return PTR_ERR(old.bh); | |
ac27a0ec DK |
3513 | /* |
3514 | * Check for inode number is _not_ due to possible IO errors. | |
3515 | * We might rmdir the source, keep it as pwd of some process | |
3516 | * and merrily kill the link to whatever was created under the | |
3517 | * same name. Goodbye sticky bit ;-< | |
3518 | */ | |
ac27a0ec | 3519 | retval = -ENOENT; |
c0d268c3 | 3520 | if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino) |
ac27a0ec DK |
3521 | goto end_rename; |
3522 | ||
c0d268c3 MS |
3523 | new.bh = ext4_find_entry(new.dir, &new.dentry->d_name, |
3524 | &new.de, &new.inlined); | |
36de9286 TT |
3525 | if (IS_ERR(new.bh)) { |
3526 | retval = PTR_ERR(new.bh); | |
a9cfcd63 | 3527 | new.bh = NULL; |
36de9286 TT |
3528 | goto end_rename; |
3529 | } | |
c0d268c3 MS |
3530 | if (new.bh) { |
3531 | if (!new.inode) { | |
3532 | brelse(new.bh); | |
3533 | new.bh = NULL; | |
ac27a0ec DK |
3534 | } |
3535 | } | |
c0d268c3 MS |
3536 | if (new.inode && !test_opt(new.dir->i_sb, NO_AUTO_DA_ALLOC)) |
3537 | ext4_alloc_da_blocks(old.inode); | |
5b61de75 | 3538 | |
cd808dec MS |
3539 | credits = (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) + |
3540 | EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2); | |
3541 | if (!(flags & RENAME_WHITEOUT)) { | |
3542 | handle = ext4_journal_start(old.dir, EXT4_HT_DIR, credits); | |
7071b715 KK |
3543 | if (IS_ERR(handle)) { |
3544 | retval = PTR_ERR(handle); | |
3545 | handle = NULL; | |
3546 | goto end_rename; | |
3547 | } | |
cd808dec MS |
3548 | } else { |
3549 | whiteout = ext4_whiteout_for_rename(&old, credits, &handle); | |
7071b715 KK |
3550 | if (IS_ERR(whiteout)) { |
3551 | retval = PTR_ERR(whiteout); | |
3552 | whiteout = NULL; | |
3553 | goto end_rename; | |
3554 | } | |
cd808dec | 3555 | } |
5b61de75 | 3556 | |
c0d268c3 | 3557 | if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir)) |
5b61de75 TT |
3558 | ext4_handle_sync(handle); |
3559 | ||
c0d268c3 MS |
3560 | if (S_ISDIR(old.inode->i_mode)) { |
3561 | if (new.inode) { | |
ac27a0ec | 3562 | retval = -ENOTEMPTY; |
e875a2dd | 3563 | if (!ext4_empty_dir(new.inode)) |
ac27a0ec | 3564 | goto end_rename; |
0d7d5d67 MS |
3565 | } else { |
3566 | retval = -EMLINK; | |
3567 | if (new.dir != old.dir && EXT4_DIR_LINK_MAX(new.dir)) | |
3568 | goto end_rename; | |
ac27a0ec | 3569 | } |
bd1af145 | 3570 | retval = ext4_rename_dir_prepare(handle, &old); |
ef607893 AG |
3571 | if (retval) |
3572 | goto end_rename; | |
ac27a0ec | 3573 | } |
d80d448c DW |
3574 | /* |
3575 | * If we're renaming a file within an inline_data dir and adding or | |
3576 | * setting the new dirent causes a conversion from inline_data to | |
3577 | * extents/blockmap, we need to force the dirent delete code to | |
3578 | * re-read the directory, or else we end up trying to delete a dirent | |
3579 | * from what is now the extent tree root (or a block map). | |
3580 | */ | |
3581 | force_reread = (new.dir->i_ino == old.dir->i_ino && | |
3582 | ext4_test_inode_flag(new.dir, EXT4_INODE_INLINE_DATA)); | |
cd808dec MS |
3583 | |
3584 | old_file_type = old.de->file_type; | |
3585 | if (whiteout) { | |
3586 | /* | |
3587 | * Do this before adding a new entry, so the old entry is sure | |
3588 | * to be still pointing to the valid old entry. | |
3589 | */ | |
3590 | retval = ext4_setent(handle, &old, whiteout->i_ino, | |
3591 | EXT4_FT_CHRDEV); | |
3592 | if (retval) | |
3593 | goto end_rename; | |
3594 | ext4_mark_inode_dirty(handle, whiteout); | |
3595 | } | |
c0d268c3 MS |
3596 | if (!new.bh) { |
3597 | retval = ext4_add_entry(handle, new.dentry, old.inode); | |
ac27a0ec DK |
3598 | if (retval) |
3599 | goto end_rename; | |
3600 | } else { | |
bd1af145 | 3601 | retval = ext4_setent(handle, &new, |
cd808dec | 3602 | old.inode->i_ino, old_file_type); |
ef607893 AG |
3603 | if (retval) |
3604 | goto end_rename; | |
ac27a0ec | 3605 | } |
d80d448c DW |
3606 | if (force_reread) |
3607 | force_reread = !ext4_test_inode_flag(new.dir, | |
3608 | EXT4_INODE_INLINE_DATA); | |
ac27a0ec DK |
3609 | |
3610 | /* | |
3611 | * Like most other Unix systems, set the ctime for inodes on a | |
3612 | * rename. | |
3613 | */ | |
eeca7ea1 | 3614 | old.inode->i_ctime = current_time(old.inode); |
c0d268c3 | 3615 | ext4_mark_inode_dirty(handle, old.inode); |
ac27a0ec | 3616 | |
cd808dec MS |
3617 | if (!whiteout) { |
3618 | /* | |
3619 | * ok, that's it | |
3620 | */ | |
3621 | ext4_rename_delete(handle, &old, force_reread); | |
3622 | } | |
ac27a0ec | 3623 | |
c0d268c3 MS |
3624 | if (new.inode) { |
3625 | ext4_dec_count(handle, new.inode); | |
eeca7ea1 | 3626 | new.inode->i_ctime = current_time(new.inode); |
ac27a0ec | 3627 | } |
eeca7ea1 | 3628 | old.dir->i_ctime = old.dir->i_mtime = current_time(old.dir); |
c0d268c3 MS |
3629 | ext4_update_dx_flag(old.dir); |
3630 | if (old.dir_bh) { | |
bd1af145 MS |
3631 | retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino); |
3632 | if (retval) | |
b4097142 | 3633 | goto end_rename; |
bd1af145 | 3634 | |
c0d268c3 MS |
3635 | ext4_dec_count(handle, old.dir); |
3636 | if (new.inode) { | |
e875a2dd MH |
3637 | /* checked ext4_empty_dir above, can't have another |
3638 | * parent, ext4_dec_count() won't work for many-linked | |
3639 | * dirs */ | |
c0d268c3 | 3640 | clear_nlink(new.inode); |
ac27a0ec | 3641 | } else { |
c0d268c3 MS |
3642 | ext4_inc_count(handle, new.dir); |
3643 | ext4_update_dx_flag(new.dir); | |
3644 | ext4_mark_inode_dirty(handle, new.dir); | |
ac27a0ec DK |
3645 | } |
3646 | } | |
c0d268c3 MS |
3647 | ext4_mark_inode_dirty(handle, old.dir); |
3648 | if (new.inode) { | |
3649 | ext4_mark_inode_dirty(handle, new.inode); | |
3650 | if (!new.inode->i_nlink) | |
3651 | ext4_orphan_add(handle, new.inode); | |
ac27a0ec DK |
3652 | } |
3653 | retval = 0; | |
3654 | ||
3655 | end_rename: | |
c0d268c3 MS |
3656 | brelse(old.dir_bh); |
3657 | brelse(old.bh); | |
3658 | brelse(new.bh); | |
cd808dec MS |
3659 | if (whiteout) { |
3660 | if (retval) | |
3661 | drop_nlink(whiteout); | |
3662 | unlock_new_inode(whiteout); | |
3663 | iput(whiteout); | |
3664 | } | |
5b61de75 TT |
3665 | if (handle) |
3666 | ext4_journal_stop(handle); | |
ac27a0ec DK |
3667 | return retval; |
3668 | } | |
3669 | ||
bd42998a MS |
3670 | static int ext4_cross_rename(struct inode *old_dir, struct dentry *old_dentry, |
3671 | struct inode *new_dir, struct dentry *new_dentry) | |
3672 | { | |
3673 | handle_t *handle = NULL; | |
3674 | struct ext4_renament old = { | |
3675 | .dir = old_dir, | |
3676 | .dentry = old_dentry, | |
2b0143b5 | 3677 | .inode = d_inode(old_dentry), |
bd42998a MS |
3678 | }; |
3679 | struct ext4_renament new = { | |
3680 | .dir = new_dir, | |
3681 | .dentry = new_dentry, | |
2b0143b5 | 3682 | .inode = d_inode(new_dentry), |
bd42998a MS |
3683 | }; |
3684 | u8 new_file_type; | |
3685 | int retval; | |
95582b00 | 3686 | struct timespec64 ctime; |
bd42998a | 3687 | |
040cb378 LX |
3688 | if ((ext4_test_inode_flag(new_dir, EXT4_INODE_PROJINHERIT) && |
3689 | !projid_eq(EXT4_I(new_dir)->i_projid, | |
3690 | EXT4_I(old_dentry->d_inode)->i_projid)) || | |
3691 | (ext4_test_inode_flag(old_dir, EXT4_INODE_PROJINHERIT) && | |
3692 | !projid_eq(EXT4_I(old_dir)->i_projid, | |
3693 | EXT4_I(new_dentry->d_inode)->i_projid))) | |
3694 | return -EXDEV; | |
3695 | ||
a7cdadee JK |
3696 | retval = dquot_initialize(old.dir); |
3697 | if (retval) | |
3698 | return retval; | |
3699 | retval = dquot_initialize(new.dir); | |
3700 | if (retval) | |
3701 | return retval; | |
bd42998a MS |
3702 | |
3703 | old.bh = ext4_find_entry(old.dir, &old.dentry->d_name, | |
3704 | &old.de, &old.inlined); | |
36de9286 TT |
3705 | if (IS_ERR(old.bh)) |
3706 | return PTR_ERR(old.bh); | |
bd42998a MS |
3707 | /* |
3708 | * Check for inode number is _not_ due to possible IO errors. | |
3709 | * We might rmdir the source, keep it as pwd of some process | |
3710 | * and merrily kill the link to whatever was created under the | |
3711 | * same name. Goodbye sticky bit ;-< | |
3712 | */ | |
3713 | retval = -ENOENT; | |
3714 | if (!old.bh || le32_to_cpu(old.de->inode) != old.inode->i_ino) | |
3715 | goto end_rename; | |
3716 | ||
3717 | new.bh = ext4_find_entry(new.dir, &new.dentry->d_name, | |
3718 | &new.de, &new.inlined); | |
36de9286 TT |
3719 | if (IS_ERR(new.bh)) { |
3720 | retval = PTR_ERR(new.bh); | |
a9cfcd63 | 3721 | new.bh = NULL; |
36de9286 TT |
3722 | goto end_rename; |
3723 | } | |
bd42998a MS |
3724 | |
3725 | /* RENAME_EXCHANGE case: old *and* new must both exist */ | |
3726 | if (!new.bh || le32_to_cpu(new.de->inode) != new.inode->i_ino) | |
3727 | goto end_rename; | |
3728 | ||
3729 | handle = ext4_journal_start(old.dir, EXT4_HT_DIR, | |
3730 | (2 * EXT4_DATA_TRANS_BLOCKS(old.dir->i_sb) + | |
3731 | 2 * EXT4_INDEX_EXTRA_TRANS_BLOCKS + 2)); | |
7071b715 KK |
3732 | if (IS_ERR(handle)) { |
3733 | retval = PTR_ERR(handle); | |
3734 | handle = NULL; | |
3735 | goto end_rename; | |
3736 | } | |
bd42998a MS |
3737 | |
3738 | if (IS_DIRSYNC(old.dir) || IS_DIRSYNC(new.dir)) | |
3739 | ext4_handle_sync(handle); | |
3740 | ||
3741 | if (S_ISDIR(old.inode->i_mode)) { | |
3742 | old.is_dir = true; | |
3743 | retval = ext4_rename_dir_prepare(handle, &old); | |
3744 | if (retval) | |
3745 | goto end_rename; | |
3746 | } | |
3747 | if (S_ISDIR(new.inode->i_mode)) { | |
3748 | new.is_dir = true; | |
3749 | retval = ext4_rename_dir_prepare(handle, &new); | |
3750 | if (retval) | |
3751 | goto end_rename; | |
3752 | } | |
3753 | ||
3754 | /* | |
3755 | * Other than the special case of overwriting a directory, parents' | |
3756 | * nlink only needs to be modified if this is a cross directory rename. | |
3757 | */ | |
3758 | if (old.dir != new.dir && old.is_dir != new.is_dir) { | |
3759 | old.dir_nlink_delta = old.is_dir ? -1 : 1; | |
3760 | new.dir_nlink_delta = -old.dir_nlink_delta; | |
3761 | retval = -EMLINK; | |
3762 | if ((old.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(old.dir)) || | |
3763 | (new.dir_nlink_delta > 0 && EXT4_DIR_LINK_MAX(new.dir))) | |
3764 | goto end_rename; | |
3765 | } | |
3766 | ||
3767 | new_file_type = new.de->file_type; | |
3768 | retval = ext4_setent(handle, &new, old.inode->i_ino, old.de->file_type); | |
3769 | if (retval) | |
3770 | goto end_rename; | |
3771 | ||
3772 | retval = ext4_setent(handle, &old, new.inode->i_ino, new_file_type); | |
3773 | if (retval) | |
3774 | goto end_rename; | |
3775 | ||
3776 | /* | |
3777 | * Like most other Unix systems, set the ctime for inodes on a | |
3778 | * rename. | |
3779 | */ | |
eeca7ea1 DD |
3780 | ctime = current_time(old.inode); |
3781 | old.inode->i_ctime = ctime; | |
3782 | new.inode->i_ctime = ctime; | |
bd42998a MS |
3783 | ext4_mark_inode_dirty(handle, old.inode); |
3784 | ext4_mark_inode_dirty(handle, new.inode); | |
3785 | ||
3786 | if (old.dir_bh) { | |
3787 | retval = ext4_rename_dir_finish(handle, &old, new.dir->i_ino); | |
3788 | if (retval) | |
3789 | goto end_rename; | |
3790 | } | |
3791 | if (new.dir_bh) { | |
3792 | retval = ext4_rename_dir_finish(handle, &new, old.dir->i_ino); | |
3793 | if (retval) | |
3794 | goto end_rename; | |
3795 | } | |
3796 | ext4_update_dir_count(handle, &old); | |
3797 | ext4_update_dir_count(handle, &new); | |
3798 | retval = 0; | |
3799 | ||
3800 | end_rename: | |
3801 | brelse(old.dir_bh); | |
3802 | brelse(new.dir_bh); | |
3803 | brelse(old.bh); | |
3804 | brelse(new.bh); | |
3805 | if (handle) | |
3806 | ext4_journal_stop(handle); | |
3807 | return retval; | |
3808 | } | |
3809 | ||
0a7c3937 MS |
3810 | static int ext4_rename2(struct inode *old_dir, struct dentry *old_dentry, |
3811 | struct inode *new_dir, struct dentry *new_dentry, | |
3812 | unsigned int flags) | |
3813 | { | |
07543d16 EB |
3814 | int err; |
3815 | ||
0db1ff22 TT |
3816 | if (unlikely(ext4_forced_shutdown(EXT4_SB(old_dir->i_sb)))) |
3817 | return -EIO; | |
3818 | ||
cd808dec | 3819 | if (flags & ~(RENAME_NOREPLACE | RENAME_EXCHANGE | RENAME_WHITEOUT)) |
0a7c3937 MS |
3820 | return -EINVAL; |
3821 | ||
07543d16 EB |
3822 | err = fscrypt_prepare_rename(old_dir, old_dentry, new_dir, new_dentry, |
3823 | flags); | |
3824 | if (err) | |
3825 | return err; | |
3826 | ||
bd42998a MS |
3827 | if (flags & RENAME_EXCHANGE) { |
3828 | return ext4_cross_rename(old_dir, old_dentry, | |
3829 | new_dir, new_dentry); | |
3830 | } | |
cd808dec MS |
3831 | |
3832 | return ext4_rename(old_dir, old_dentry, new_dir, new_dentry, flags); | |
0a7c3937 MS |
3833 | } |
3834 | ||
ac27a0ec DK |
3835 | /* |
3836 | * directories can handle most operations... | |
3837 | */ | |
754661f1 | 3838 | const struct inode_operations ext4_dir_inode_operations = { |
617ba13b MC |
3839 | .create = ext4_create, |
3840 | .lookup = ext4_lookup, | |
3841 | .link = ext4_link, | |
3842 | .unlink = ext4_unlink, | |
3843 | .symlink = ext4_symlink, | |
3844 | .mkdir = ext4_mkdir, | |
3845 | .rmdir = ext4_rmdir, | |
3846 | .mknod = ext4_mknod, | |
af51a2ac | 3847 | .tmpfile = ext4_tmpfile, |
2773bf00 | 3848 | .rename = ext4_rename2, |
617ba13b | 3849 | .setattr = ext4_setattr, |
99652ea5 | 3850 | .getattr = ext4_getattr, |
617ba13b | 3851 | .listxattr = ext4_listxattr, |
4e34e719 | 3852 | .get_acl = ext4_get_acl, |
64e178a7 | 3853 | .set_acl = ext4_set_acl, |
abc8746e | 3854 | .fiemap = ext4_fiemap, |
ac27a0ec DK |
3855 | }; |
3856 | ||
754661f1 | 3857 | const struct inode_operations ext4_special_inode_operations = { |
617ba13b | 3858 | .setattr = ext4_setattr, |
99652ea5 | 3859 | .getattr = ext4_getattr, |
617ba13b | 3860 | .listxattr = ext4_listxattr, |
4e34e719 | 3861 | .get_acl = ext4_get_acl, |
64e178a7 | 3862 | .set_acl = ext4_set_acl, |
ac27a0ec | 3863 | }; |