Merge branch 'clockevents/cmt-mtu2-tmu-cleanups' of git://git.linaro.org/people/danie...
[deliverable/linux.git] / fs / ext4 / xattr.c
1 /*
2 * linux/fs/ext4/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 * suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
11 * Red Hat Inc.
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 * and Andreas Gruenbacher <agruen@suse.de>.
14 */
15
16 /*
17 * Extended attributes are stored directly in inodes (on file systems with
18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
19 * field contains the block number if an inode uses an additional block. All
20 * attributes must fit in the inode and one additional block. Blocks that
21 * contain the identical set of attributes may be shared among several inodes.
22 * Identical blocks are detected by keeping a cache of blocks that have
23 * recently been accessed.
24 *
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
27 *
28 * +------------------+
29 * | header |
30 * | entry 1 | |
31 * | entry 2 | | growing downwards
32 * | entry 3 | v
33 * | four null bytes |
34 * | . . . |
35 * | value 1 | ^
36 * | value 3 | | growing upwards
37 * | value 2 | |
38 * +------------------+
39 *
40 * The header is followed by multiple entry descriptors. In disk blocks, the
41 * entry descriptors are kept sorted. In inodes, they are unsorted. The
42 * attribute values are aligned to the end of the block in no specific order.
43 *
44 * Locking strategy
45 * ----------------
46 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
50 * by the buffer lock.
51 */
52
53 #include <linux/init.h>
54 #include <linux/fs.h>
55 #include <linux/slab.h>
56 #include <linux/mbcache.h>
57 #include <linux/quotaops.h>
58 #include <linux/rwsem.h>
59 #include "ext4_jbd2.h"
60 #include "ext4.h"
61 #include "xattr.h"
62 #include "acl.h"
63
64 #ifdef EXT4_XATTR_DEBUG
65 # define ea_idebug(inode, f...) do { \
66 printk(KERN_DEBUG "inode %s:%lu: ", \
67 inode->i_sb->s_id, inode->i_ino); \
68 printk(f); \
69 printk("\n"); \
70 } while (0)
71 # define ea_bdebug(bh, f...) do { \
72 char b[BDEVNAME_SIZE]; \
73 printk(KERN_DEBUG "block %s:%lu: ", \
74 bdevname(bh->b_bdev, b), \
75 (unsigned long) bh->b_blocknr); \
76 printk(f); \
77 printk("\n"); \
78 } while (0)
79 #else
80 # define ea_idebug(inode, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
81 # define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
82 #endif
83
84 static void ext4_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
85 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
86 struct ext4_xattr_header *,
87 struct mb_cache_entry **);
88 static void ext4_xattr_rehash(struct ext4_xattr_header *,
89 struct ext4_xattr_entry *);
90 static int ext4_xattr_list(struct dentry *dentry, char *buffer,
91 size_t buffer_size);
92
93 static const struct xattr_handler *ext4_xattr_handler_map[] = {
94 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
95 #ifdef CONFIG_EXT4_FS_POSIX_ACL
96 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &posix_acl_access_xattr_handler,
97 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &posix_acl_default_xattr_handler,
98 #endif
99 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
100 #ifdef CONFIG_EXT4_FS_SECURITY
101 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
102 #endif
103 };
104
105 const struct xattr_handler *ext4_xattr_handlers[] = {
106 &ext4_xattr_user_handler,
107 &ext4_xattr_trusted_handler,
108 #ifdef CONFIG_EXT4_FS_POSIX_ACL
109 &posix_acl_access_xattr_handler,
110 &posix_acl_default_xattr_handler,
111 #endif
112 #ifdef CONFIG_EXT4_FS_SECURITY
113 &ext4_xattr_security_handler,
114 #endif
115 NULL
116 };
117
118 #define EXT4_GET_MB_CACHE(inode) (((struct ext4_sb_info *) \
119 inode->i_sb->s_fs_info)->s_mb_cache)
120
121 static __le32 ext4_xattr_block_csum(struct inode *inode,
122 sector_t block_nr,
123 struct ext4_xattr_header *hdr)
124 {
125 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
126 __u32 csum;
127 __le32 save_csum;
128 __le64 dsk_block_nr = cpu_to_le64(block_nr);
129
130 save_csum = hdr->h_checksum;
131 hdr->h_checksum = 0;
132 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
133 sizeof(dsk_block_nr));
134 csum = ext4_chksum(sbi, csum, (__u8 *)hdr,
135 EXT4_BLOCK_SIZE(inode->i_sb));
136
137 hdr->h_checksum = save_csum;
138 return cpu_to_le32(csum);
139 }
140
141 static int ext4_xattr_block_csum_verify(struct inode *inode,
142 sector_t block_nr,
143 struct ext4_xattr_header *hdr)
144 {
145 if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
146 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
147 (hdr->h_checksum != ext4_xattr_block_csum(inode, block_nr, hdr)))
148 return 0;
149 return 1;
150 }
151
152 static void ext4_xattr_block_csum_set(struct inode *inode,
153 sector_t block_nr,
154 struct ext4_xattr_header *hdr)
155 {
156 if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
157 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
158 return;
159
160 hdr->h_checksum = ext4_xattr_block_csum(inode, block_nr, hdr);
161 }
162
163 static inline int ext4_handle_dirty_xattr_block(handle_t *handle,
164 struct inode *inode,
165 struct buffer_head *bh)
166 {
167 ext4_xattr_block_csum_set(inode, bh->b_blocknr, BHDR(bh));
168 return ext4_handle_dirty_metadata(handle, inode, bh);
169 }
170
171 static inline const struct xattr_handler *
172 ext4_xattr_handler(int name_index)
173 {
174 const struct xattr_handler *handler = NULL;
175
176 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
177 handler = ext4_xattr_handler_map[name_index];
178 return handler;
179 }
180
181 /*
182 * Inode operation listxattr()
183 *
184 * dentry->d_inode->i_mutex: don't care
185 */
186 ssize_t
187 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
188 {
189 return ext4_xattr_list(dentry, buffer, size);
190 }
191
192 static int
193 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
194 {
195 while (!IS_LAST_ENTRY(entry)) {
196 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
197 if ((void *)next >= end)
198 return -EIO;
199 entry = next;
200 }
201 return 0;
202 }
203
204 static inline int
205 ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
206 {
207 int error;
208
209 if (buffer_verified(bh))
210 return 0;
211
212 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
213 BHDR(bh)->h_blocks != cpu_to_le32(1))
214 return -EIO;
215 if (!ext4_xattr_block_csum_verify(inode, bh->b_blocknr, BHDR(bh)))
216 return -EIO;
217 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
218 if (!error)
219 set_buffer_verified(bh);
220 return error;
221 }
222
223 static inline int
224 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
225 {
226 size_t value_size = le32_to_cpu(entry->e_value_size);
227
228 if (entry->e_value_block != 0 || value_size > size ||
229 le16_to_cpu(entry->e_value_offs) + value_size > size)
230 return -EIO;
231 return 0;
232 }
233
234 static int
235 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
236 const char *name, size_t size, int sorted)
237 {
238 struct ext4_xattr_entry *entry;
239 size_t name_len;
240 int cmp = 1;
241
242 if (name == NULL)
243 return -EINVAL;
244 name_len = strlen(name);
245 entry = *pentry;
246 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
247 cmp = name_index - entry->e_name_index;
248 if (!cmp)
249 cmp = name_len - entry->e_name_len;
250 if (!cmp)
251 cmp = memcmp(name, entry->e_name, name_len);
252 if (cmp <= 0 && (sorted || cmp == 0))
253 break;
254 }
255 *pentry = entry;
256 if (!cmp && ext4_xattr_check_entry(entry, size))
257 return -EIO;
258 return cmp ? -ENODATA : 0;
259 }
260
261 static int
262 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
263 void *buffer, size_t buffer_size)
264 {
265 struct buffer_head *bh = NULL;
266 struct ext4_xattr_entry *entry;
267 size_t size;
268 int error;
269 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
270
271 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
272 name_index, name, buffer, (long)buffer_size);
273
274 error = -ENODATA;
275 if (!EXT4_I(inode)->i_file_acl)
276 goto cleanup;
277 ea_idebug(inode, "reading block %llu",
278 (unsigned long long)EXT4_I(inode)->i_file_acl);
279 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
280 if (!bh)
281 goto cleanup;
282 ea_bdebug(bh, "b_count=%d, refcount=%d",
283 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
284 if (ext4_xattr_check_block(inode, bh)) {
285 bad_block:
286 EXT4_ERROR_INODE(inode, "bad block %llu",
287 EXT4_I(inode)->i_file_acl);
288 error = -EIO;
289 goto cleanup;
290 }
291 ext4_xattr_cache_insert(ext4_mb_cache, bh);
292 entry = BFIRST(bh);
293 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
294 if (error == -EIO)
295 goto bad_block;
296 if (error)
297 goto cleanup;
298 size = le32_to_cpu(entry->e_value_size);
299 if (buffer) {
300 error = -ERANGE;
301 if (size > buffer_size)
302 goto cleanup;
303 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
304 size);
305 }
306 error = size;
307
308 cleanup:
309 brelse(bh);
310 return error;
311 }
312
313 int
314 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
315 void *buffer, size_t buffer_size)
316 {
317 struct ext4_xattr_ibody_header *header;
318 struct ext4_xattr_entry *entry;
319 struct ext4_inode *raw_inode;
320 struct ext4_iloc iloc;
321 size_t size;
322 void *end;
323 int error;
324
325 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
326 return -ENODATA;
327 error = ext4_get_inode_loc(inode, &iloc);
328 if (error)
329 return error;
330 raw_inode = ext4_raw_inode(&iloc);
331 header = IHDR(inode, raw_inode);
332 entry = IFIRST(header);
333 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
334 error = ext4_xattr_check_names(entry, end);
335 if (error)
336 goto cleanup;
337 error = ext4_xattr_find_entry(&entry, name_index, name,
338 end - (void *)entry, 0);
339 if (error)
340 goto cleanup;
341 size = le32_to_cpu(entry->e_value_size);
342 if (buffer) {
343 error = -ERANGE;
344 if (size > buffer_size)
345 goto cleanup;
346 memcpy(buffer, (void *)IFIRST(header) +
347 le16_to_cpu(entry->e_value_offs), size);
348 }
349 error = size;
350
351 cleanup:
352 brelse(iloc.bh);
353 return error;
354 }
355
356 /*
357 * ext4_xattr_get()
358 *
359 * Copy an extended attribute into the buffer
360 * provided, or compute the buffer size required.
361 * Buffer is NULL to compute the size of the buffer required.
362 *
363 * Returns a negative error number on failure, or the number of bytes
364 * used / required on success.
365 */
366 int
367 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
368 void *buffer, size_t buffer_size)
369 {
370 int error;
371
372 down_read(&EXT4_I(inode)->xattr_sem);
373 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
374 buffer_size);
375 if (error == -ENODATA)
376 error = ext4_xattr_block_get(inode, name_index, name, buffer,
377 buffer_size);
378 up_read(&EXT4_I(inode)->xattr_sem);
379 return error;
380 }
381
382 static int
383 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
384 char *buffer, size_t buffer_size)
385 {
386 size_t rest = buffer_size;
387
388 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
389 const struct xattr_handler *handler =
390 ext4_xattr_handler(entry->e_name_index);
391
392 if (handler) {
393 size_t size = handler->list(dentry, buffer, rest,
394 entry->e_name,
395 entry->e_name_len,
396 handler->flags);
397 if (buffer) {
398 if (size > rest)
399 return -ERANGE;
400 buffer += size;
401 }
402 rest -= size;
403 }
404 }
405 return buffer_size - rest;
406 }
407
408 static int
409 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
410 {
411 struct inode *inode = dentry->d_inode;
412 struct buffer_head *bh = NULL;
413 int error;
414 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
415
416 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
417 buffer, (long)buffer_size);
418
419 error = 0;
420 if (!EXT4_I(inode)->i_file_acl)
421 goto cleanup;
422 ea_idebug(inode, "reading block %llu",
423 (unsigned long long)EXT4_I(inode)->i_file_acl);
424 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
425 error = -EIO;
426 if (!bh)
427 goto cleanup;
428 ea_bdebug(bh, "b_count=%d, refcount=%d",
429 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
430 if (ext4_xattr_check_block(inode, bh)) {
431 EXT4_ERROR_INODE(inode, "bad block %llu",
432 EXT4_I(inode)->i_file_acl);
433 error = -EIO;
434 goto cleanup;
435 }
436 ext4_xattr_cache_insert(ext4_mb_cache, bh);
437 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
438
439 cleanup:
440 brelse(bh);
441
442 return error;
443 }
444
445 static int
446 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
447 {
448 struct inode *inode = dentry->d_inode;
449 struct ext4_xattr_ibody_header *header;
450 struct ext4_inode *raw_inode;
451 struct ext4_iloc iloc;
452 void *end;
453 int error;
454
455 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
456 return 0;
457 error = ext4_get_inode_loc(inode, &iloc);
458 if (error)
459 return error;
460 raw_inode = ext4_raw_inode(&iloc);
461 header = IHDR(inode, raw_inode);
462 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
463 error = ext4_xattr_check_names(IFIRST(header), end);
464 if (error)
465 goto cleanup;
466 error = ext4_xattr_list_entries(dentry, IFIRST(header),
467 buffer, buffer_size);
468
469 cleanup:
470 brelse(iloc.bh);
471 return error;
472 }
473
474 /*
475 * ext4_xattr_list()
476 *
477 * Copy a list of attribute names into the buffer
478 * provided, or compute the buffer size required.
479 * Buffer is NULL to compute the size of the buffer required.
480 *
481 * Returns a negative error number on failure, or the number of bytes
482 * used / required on success.
483 */
484 static int
485 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
486 {
487 int ret, ret2;
488
489 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
490 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
491 if (ret < 0)
492 goto errout;
493 if (buffer) {
494 buffer += ret;
495 buffer_size -= ret;
496 }
497 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
498 if (ret < 0)
499 goto errout;
500 ret += ret2;
501 errout:
502 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
503 return ret;
504 }
505
506 /*
507 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
508 * not set, set it.
509 */
510 static void ext4_xattr_update_super_block(handle_t *handle,
511 struct super_block *sb)
512 {
513 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
514 return;
515
516 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
517 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
518 ext4_handle_dirty_super(handle, sb);
519 }
520 }
521
522 /*
523 * Release the xattr block BH: If the reference count is > 1, decrement
524 * it; otherwise free the block.
525 */
526 static void
527 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
528 struct buffer_head *bh)
529 {
530 struct mb_cache_entry *ce = NULL;
531 int error = 0;
532 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
533
534 ce = mb_cache_entry_get(ext4_mb_cache, bh->b_bdev, bh->b_blocknr);
535 error = ext4_journal_get_write_access(handle, bh);
536 if (error)
537 goto out;
538
539 lock_buffer(bh);
540 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
541 ea_bdebug(bh, "refcount now=0; freeing");
542 if (ce)
543 mb_cache_entry_free(ce);
544 get_bh(bh);
545 ext4_free_blocks(handle, inode, bh, 0, 1,
546 EXT4_FREE_BLOCKS_METADATA |
547 EXT4_FREE_BLOCKS_FORGET);
548 unlock_buffer(bh);
549 } else {
550 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
551 if (ce)
552 mb_cache_entry_release(ce);
553 unlock_buffer(bh);
554 error = ext4_handle_dirty_xattr_block(handle, inode, bh);
555 if (IS_SYNC(inode))
556 ext4_handle_sync(handle);
557 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
558 ea_bdebug(bh, "refcount now=%d; releasing",
559 le32_to_cpu(BHDR(bh)->h_refcount));
560 }
561 out:
562 ext4_std_error(inode->i_sb, error);
563 return;
564 }
565
566 /*
567 * Find the available free space for EAs. This also returns the total number of
568 * bytes used by EA entries.
569 */
570 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
571 size_t *min_offs, void *base, int *total)
572 {
573 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
574 if (!last->e_value_block && last->e_value_size) {
575 size_t offs = le16_to_cpu(last->e_value_offs);
576 if (offs < *min_offs)
577 *min_offs = offs;
578 }
579 if (total)
580 *total += EXT4_XATTR_LEN(last->e_name_len);
581 }
582 return (*min_offs - ((void *)last - base) - sizeof(__u32));
583 }
584
585 static int
586 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
587 {
588 struct ext4_xattr_entry *last;
589 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
590
591 /* Compute min_offs and last. */
592 last = s->first;
593 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
594 if (!last->e_value_block && last->e_value_size) {
595 size_t offs = le16_to_cpu(last->e_value_offs);
596 if (offs < min_offs)
597 min_offs = offs;
598 }
599 }
600 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
601 if (!s->not_found) {
602 if (!s->here->e_value_block && s->here->e_value_size) {
603 size_t size = le32_to_cpu(s->here->e_value_size);
604 free += EXT4_XATTR_SIZE(size);
605 }
606 free += EXT4_XATTR_LEN(name_len);
607 }
608 if (i->value) {
609 if (free < EXT4_XATTR_SIZE(i->value_len) ||
610 free < EXT4_XATTR_LEN(name_len) +
611 EXT4_XATTR_SIZE(i->value_len))
612 return -ENOSPC;
613 }
614
615 if (i->value && s->not_found) {
616 /* Insert the new name. */
617 size_t size = EXT4_XATTR_LEN(name_len);
618 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
619 memmove((void *)s->here + size, s->here, rest);
620 memset(s->here, 0, size);
621 s->here->e_name_index = i->name_index;
622 s->here->e_name_len = name_len;
623 memcpy(s->here->e_name, i->name, name_len);
624 } else {
625 if (!s->here->e_value_block && s->here->e_value_size) {
626 void *first_val = s->base + min_offs;
627 size_t offs = le16_to_cpu(s->here->e_value_offs);
628 void *val = s->base + offs;
629 size_t size = EXT4_XATTR_SIZE(
630 le32_to_cpu(s->here->e_value_size));
631
632 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
633 /* The old and the new value have the same
634 size. Just replace. */
635 s->here->e_value_size =
636 cpu_to_le32(i->value_len);
637 if (i->value == EXT4_ZERO_XATTR_VALUE) {
638 memset(val, 0, size);
639 } else {
640 /* Clear pad bytes first. */
641 memset(val + size - EXT4_XATTR_PAD, 0,
642 EXT4_XATTR_PAD);
643 memcpy(val, i->value, i->value_len);
644 }
645 return 0;
646 }
647
648 /* Remove the old value. */
649 memmove(first_val + size, first_val, val - first_val);
650 memset(first_val, 0, size);
651 s->here->e_value_size = 0;
652 s->here->e_value_offs = 0;
653 min_offs += size;
654
655 /* Adjust all value offsets. */
656 last = s->first;
657 while (!IS_LAST_ENTRY(last)) {
658 size_t o = le16_to_cpu(last->e_value_offs);
659 if (!last->e_value_block &&
660 last->e_value_size && o < offs)
661 last->e_value_offs =
662 cpu_to_le16(o + size);
663 last = EXT4_XATTR_NEXT(last);
664 }
665 }
666 if (!i->value) {
667 /* Remove the old name. */
668 size_t size = EXT4_XATTR_LEN(name_len);
669 last = ENTRY((void *)last - size);
670 memmove(s->here, (void *)s->here + size,
671 (void *)last - (void *)s->here + sizeof(__u32));
672 memset(last, 0, size);
673 }
674 }
675
676 if (i->value) {
677 /* Insert the new value. */
678 s->here->e_value_size = cpu_to_le32(i->value_len);
679 if (i->value_len) {
680 size_t size = EXT4_XATTR_SIZE(i->value_len);
681 void *val = s->base + min_offs - size;
682 s->here->e_value_offs = cpu_to_le16(min_offs - size);
683 if (i->value == EXT4_ZERO_XATTR_VALUE) {
684 memset(val, 0, size);
685 } else {
686 /* Clear the pad bytes first. */
687 memset(val + size - EXT4_XATTR_PAD, 0,
688 EXT4_XATTR_PAD);
689 memcpy(val, i->value, i->value_len);
690 }
691 }
692 }
693 return 0;
694 }
695
696 struct ext4_xattr_block_find {
697 struct ext4_xattr_search s;
698 struct buffer_head *bh;
699 };
700
701 static int
702 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
703 struct ext4_xattr_block_find *bs)
704 {
705 struct super_block *sb = inode->i_sb;
706 int error;
707
708 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
709 i->name_index, i->name, i->value, (long)i->value_len);
710
711 if (EXT4_I(inode)->i_file_acl) {
712 /* The inode already has an extended attribute block. */
713 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
714 error = -EIO;
715 if (!bs->bh)
716 goto cleanup;
717 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
718 atomic_read(&(bs->bh->b_count)),
719 le32_to_cpu(BHDR(bs->bh)->h_refcount));
720 if (ext4_xattr_check_block(inode, bs->bh)) {
721 EXT4_ERROR_INODE(inode, "bad block %llu",
722 EXT4_I(inode)->i_file_acl);
723 error = -EIO;
724 goto cleanup;
725 }
726 /* Find the named attribute. */
727 bs->s.base = BHDR(bs->bh);
728 bs->s.first = BFIRST(bs->bh);
729 bs->s.end = bs->bh->b_data + bs->bh->b_size;
730 bs->s.here = bs->s.first;
731 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
732 i->name, bs->bh->b_size, 1);
733 if (error && error != -ENODATA)
734 goto cleanup;
735 bs->s.not_found = error;
736 }
737 error = 0;
738
739 cleanup:
740 return error;
741 }
742
743 static int
744 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
745 struct ext4_xattr_info *i,
746 struct ext4_xattr_block_find *bs)
747 {
748 struct super_block *sb = inode->i_sb;
749 struct buffer_head *new_bh = NULL;
750 struct ext4_xattr_search *s = &bs->s;
751 struct mb_cache_entry *ce = NULL;
752 int error = 0;
753 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
754
755 #define header(x) ((struct ext4_xattr_header *)(x))
756
757 if (i->value && i->value_len > sb->s_blocksize)
758 return -ENOSPC;
759 if (s->base) {
760 ce = mb_cache_entry_get(ext4_mb_cache, bs->bh->b_bdev,
761 bs->bh->b_blocknr);
762 error = ext4_journal_get_write_access(handle, bs->bh);
763 if (error)
764 goto cleanup;
765 lock_buffer(bs->bh);
766
767 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
768 if (ce) {
769 mb_cache_entry_free(ce);
770 ce = NULL;
771 }
772 ea_bdebug(bs->bh, "modifying in-place");
773 error = ext4_xattr_set_entry(i, s);
774 if (!error) {
775 if (!IS_LAST_ENTRY(s->first))
776 ext4_xattr_rehash(header(s->base),
777 s->here);
778 ext4_xattr_cache_insert(ext4_mb_cache,
779 bs->bh);
780 }
781 unlock_buffer(bs->bh);
782 if (error == -EIO)
783 goto bad_block;
784 if (!error)
785 error = ext4_handle_dirty_xattr_block(handle,
786 inode,
787 bs->bh);
788 if (error)
789 goto cleanup;
790 goto inserted;
791 } else {
792 int offset = (char *)s->here - bs->bh->b_data;
793
794 unlock_buffer(bs->bh);
795 if (ce) {
796 mb_cache_entry_release(ce);
797 ce = NULL;
798 }
799 ea_bdebug(bs->bh, "cloning");
800 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
801 error = -ENOMEM;
802 if (s->base == NULL)
803 goto cleanup;
804 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
805 s->first = ENTRY(header(s->base)+1);
806 header(s->base)->h_refcount = cpu_to_le32(1);
807 s->here = ENTRY(s->base + offset);
808 s->end = s->base + bs->bh->b_size;
809 }
810 } else {
811 /* Allocate a buffer where we construct the new block. */
812 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
813 /* assert(header == s->base) */
814 error = -ENOMEM;
815 if (s->base == NULL)
816 goto cleanup;
817 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
818 header(s->base)->h_blocks = cpu_to_le32(1);
819 header(s->base)->h_refcount = cpu_to_le32(1);
820 s->first = ENTRY(header(s->base)+1);
821 s->here = ENTRY(header(s->base)+1);
822 s->end = s->base + sb->s_blocksize;
823 }
824
825 error = ext4_xattr_set_entry(i, s);
826 if (error == -EIO)
827 goto bad_block;
828 if (error)
829 goto cleanup;
830 if (!IS_LAST_ENTRY(s->first))
831 ext4_xattr_rehash(header(s->base), s->here);
832
833 inserted:
834 if (!IS_LAST_ENTRY(s->first)) {
835 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
836 if (new_bh) {
837 /* We found an identical block in the cache. */
838 if (new_bh == bs->bh)
839 ea_bdebug(new_bh, "keeping");
840 else {
841 /* The old block is released after updating
842 the inode. */
843 error = dquot_alloc_block(inode,
844 EXT4_C2B(EXT4_SB(sb), 1));
845 if (error)
846 goto cleanup;
847 error = ext4_journal_get_write_access(handle,
848 new_bh);
849 if (error)
850 goto cleanup_dquot;
851 lock_buffer(new_bh);
852 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
853 ea_bdebug(new_bh, "reusing; refcount now=%d",
854 le32_to_cpu(BHDR(new_bh)->h_refcount));
855 unlock_buffer(new_bh);
856 error = ext4_handle_dirty_xattr_block(handle,
857 inode,
858 new_bh);
859 if (error)
860 goto cleanup_dquot;
861 }
862 mb_cache_entry_release(ce);
863 ce = NULL;
864 } else if (bs->bh && s->base == bs->bh->b_data) {
865 /* We were modifying this block in-place. */
866 ea_bdebug(bs->bh, "keeping this block");
867 new_bh = bs->bh;
868 get_bh(new_bh);
869 } else {
870 /* We need to allocate a new block */
871 ext4_fsblk_t goal, block;
872
873 goal = ext4_group_first_block_no(sb,
874 EXT4_I(inode)->i_block_group);
875
876 /* non-extent files can't have physical blocks past 2^32 */
877 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
878 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
879
880 /*
881 * take i_data_sem because we will test
882 * i_delalloc_reserved_flag in ext4_mb_new_blocks
883 */
884 down_read((&EXT4_I(inode)->i_data_sem));
885 block = ext4_new_meta_blocks(handle, inode, goal, 0,
886 NULL, &error);
887 up_read((&EXT4_I(inode)->i_data_sem));
888 if (error)
889 goto cleanup;
890
891 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
892 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
893
894 ea_idebug(inode, "creating block %llu",
895 (unsigned long long)block);
896
897 new_bh = sb_getblk(sb, block);
898 if (unlikely(!new_bh)) {
899 error = -ENOMEM;
900 getblk_failed:
901 ext4_free_blocks(handle, inode, NULL, block, 1,
902 EXT4_FREE_BLOCKS_METADATA);
903 goto cleanup;
904 }
905 lock_buffer(new_bh);
906 error = ext4_journal_get_create_access(handle, new_bh);
907 if (error) {
908 unlock_buffer(new_bh);
909 error = -EIO;
910 goto getblk_failed;
911 }
912 memcpy(new_bh->b_data, s->base, new_bh->b_size);
913 set_buffer_uptodate(new_bh);
914 unlock_buffer(new_bh);
915 ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
916 error = ext4_handle_dirty_xattr_block(handle,
917 inode, new_bh);
918 if (error)
919 goto cleanup;
920 }
921 }
922
923 /* Update the inode. */
924 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
925
926 /* Drop the previous xattr block. */
927 if (bs->bh && bs->bh != new_bh)
928 ext4_xattr_release_block(handle, inode, bs->bh);
929 error = 0;
930
931 cleanup:
932 if (ce)
933 mb_cache_entry_release(ce);
934 brelse(new_bh);
935 if (!(bs->bh && s->base == bs->bh->b_data))
936 kfree(s->base);
937
938 return error;
939
940 cleanup_dquot:
941 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
942 goto cleanup;
943
944 bad_block:
945 EXT4_ERROR_INODE(inode, "bad block %llu",
946 EXT4_I(inode)->i_file_acl);
947 goto cleanup;
948
949 #undef header
950 }
951
952 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
953 struct ext4_xattr_ibody_find *is)
954 {
955 struct ext4_xattr_ibody_header *header;
956 struct ext4_inode *raw_inode;
957 int error;
958
959 if (EXT4_I(inode)->i_extra_isize == 0)
960 return 0;
961 raw_inode = ext4_raw_inode(&is->iloc);
962 header = IHDR(inode, raw_inode);
963 is->s.base = is->s.first = IFIRST(header);
964 is->s.here = is->s.first;
965 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
966 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
967 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
968 if (error)
969 return error;
970 /* Find the named attribute. */
971 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
972 i->name, is->s.end -
973 (void *)is->s.base, 0);
974 if (error && error != -ENODATA)
975 return error;
976 is->s.not_found = error;
977 }
978 return 0;
979 }
980
981 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
982 struct ext4_xattr_info *i,
983 struct ext4_xattr_ibody_find *is)
984 {
985 struct ext4_xattr_ibody_header *header;
986 struct ext4_xattr_search *s = &is->s;
987 int error;
988
989 if (EXT4_I(inode)->i_extra_isize == 0)
990 return -ENOSPC;
991 error = ext4_xattr_set_entry(i, s);
992 if (error) {
993 if (error == -ENOSPC &&
994 ext4_has_inline_data(inode)) {
995 error = ext4_try_to_evict_inline_data(handle, inode,
996 EXT4_XATTR_LEN(strlen(i->name) +
997 EXT4_XATTR_SIZE(i->value_len)));
998 if (error)
999 return error;
1000 error = ext4_xattr_ibody_find(inode, i, is);
1001 if (error)
1002 return error;
1003 error = ext4_xattr_set_entry(i, s);
1004 }
1005 if (error)
1006 return error;
1007 }
1008 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1009 if (!IS_LAST_ENTRY(s->first)) {
1010 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1011 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1012 } else {
1013 header->h_magic = cpu_to_le32(0);
1014 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1015 }
1016 return 0;
1017 }
1018
1019 static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
1020 struct ext4_xattr_info *i,
1021 struct ext4_xattr_ibody_find *is)
1022 {
1023 struct ext4_xattr_ibody_header *header;
1024 struct ext4_xattr_search *s = &is->s;
1025 int error;
1026
1027 if (EXT4_I(inode)->i_extra_isize == 0)
1028 return -ENOSPC;
1029 error = ext4_xattr_set_entry(i, s);
1030 if (error)
1031 return error;
1032 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1033 if (!IS_LAST_ENTRY(s->first)) {
1034 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1035 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1036 } else {
1037 header->h_magic = cpu_to_le32(0);
1038 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1039 }
1040 return 0;
1041 }
1042
1043 /*
1044 * ext4_xattr_set_handle()
1045 *
1046 * Create, replace or remove an extended attribute for this inode. Value
1047 * is NULL to remove an existing extended attribute, and non-NULL to
1048 * either replace an existing extended attribute, or create a new extended
1049 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1050 * specify that an extended attribute must exist and must not exist
1051 * previous to the call, respectively.
1052 *
1053 * Returns 0, or a negative error number on failure.
1054 */
1055 int
1056 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1057 const char *name, const void *value, size_t value_len,
1058 int flags)
1059 {
1060 struct ext4_xattr_info i = {
1061 .name_index = name_index,
1062 .name = name,
1063 .value = value,
1064 .value_len = value_len,
1065
1066 };
1067 struct ext4_xattr_ibody_find is = {
1068 .s = { .not_found = -ENODATA, },
1069 };
1070 struct ext4_xattr_block_find bs = {
1071 .s = { .not_found = -ENODATA, },
1072 };
1073 unsigned long no_expand;
1074 int error;
1075
1076 if (!name)
1077 return -EINVAL;
1078 if (strlen(name) > 255)
1079 return -ERANGE;
1080 down_write(&EXT4_I(inode)->xattr_sem);
1081 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
1082 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
1083
1084 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1085 if (error)
1086 goto cleanup;
1087
1088 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1089 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1090 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1091 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1092 }
1093
1094 error = ext4_xattr_ibody_find(inode, &i, &is);
1095 if (error)
1096 goto cleanup;
1097 if (is.s.not_found)
1098 error = ext4_xattr_block_find(inode, &i, &bs);
1099 if (error)
1100 goto cleanup;
1101 if (is.s.not_found && bs.s.not_found) {
1102 error = -ENODATA;
1103 if (flags & XATTR_REPLACE)
1104 goto cleanup;
1105 error = 0;
1106 if (!value)
1107 goto cleanup;
1108 } else {
1109 error = -EEXIST;
1110 if (flags & XATTR_CREATE)
1111 goto cleanup;
1112 }
1113 if (!value) {
1114 if (!is.s.not_found)
1115 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1116 else if (!bs.s.not_found)
1117 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1118 } else {
1119 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1120 if (!error && !bs.s.not_found) {
1121 i.value = NULL;
1122 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1123 } else if (error == -ENOSPC) {
1124 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1125 error = ext4_xattr_block_find(inode, &i, &bs);
1126 if (error)
1127 goto cleanup;
1128 }
1129 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1130 if (error)
1131 goto cleanup;
1132 if (!is.s.not_found) {
1133 i.value = NULL;
1134 error = ext4_xattr_ibody_set(handle, inode, &i,
1135 &is);
1136 }
1137 }
1138 }
1139 if (!error) {
1140 ext4_xattr_update_super_block(handle, inode->i_sb);
1141 inode->i_ctime = ext4_current_time(inode);
1142 if (!value)
1143 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1144 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1145 /*
1146 * The bh is consumed by ext4_mark_iloc_dirty, even with
1147 * error != 0.
1148 */
1149 is.iloc.bh = NULL;
1150 if (IS_SYNC(inode))
1151 ext4_handle_sync(handle);
1152 }
1153
1154 cleanup:
1155 brelse(is.iloc.bh);
1156 brelse(bs.bh);
1157 if (no_expand == 0)
1158 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1159 up_write(&EXT4_I(inode)->xattr_sem);
1160 return error;
1161 }
1162
1163 /*
1164 * ext4_xattr_set()
1165 *
1166 * Like ext4_xattr_set_handle, but start from an inode. This extended
1167 * attribute modification is a filesystem transaction by itself.
1168 *
1169 * Returns 0, or a negative error number on failure.
1170 */
1171 int
1172 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1173 const void *value, size_t value_len, int flags)
1174 {
1175 handle_t *handle;
1176 int error, retries = 0;
1177 int credits = ext4_jbd2_credits_xattr(inode);
1178
1179 retry:
1180 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1181 if (IS_ERR(handle)) {
1182 error = PTR_ERR(handle);
1183 } else {
1184 int error2;
1185
1186 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1187 value, value_len, flags);
1188 error2 = ext4_journal_stop(handle);
1189 if (error == -ENOSPC &&
1190 ext4_should_retry_alloc(inode->i_sb, &retries))
1191 goto retry;
1192 if (error == 0)
1193 error = error2;
1194 }
1195
1196 return error;
1197 }
1198
1199 /*
1200 * Shift the EA entries in the inode to create space for the increased
1201 * i_extra_isize.
1202 */
1203 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1204 int value_offs_shift, void *to,
1205 void *from, size_t n, int blocksize)
1206 {
1207 struct ext4_xattr_entry *last = entry;
1208 int new_offs;
1209
1210 /* Adjust the value offsets of the entries */
1211 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1212 if (!last->e_value_block && last->e_value_size) {
1213 new_offs = le16_to_cpu(last->e_value_offs) +
1214 value_offs_shift;
1215 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1216 > blocksize);
1217 last->e_value_offs = cpu_to_le16(new_offs);
1218 }
1219 }
1220 /* Shift the entries by n bytes */
1221 memmove(to, from, n);
1222 }
1223
1224 /*
1225 * Expand an inode by new_extra_isize bytes when EAs are present.
1226 * Returns 0 on success or negative error number on failure.
1227 */
1228 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1229 struct ext4_inode *raw_inode, handle_t *handle)
1230 {
1231 struct ext4_xattr_ibody_header *header;
1232 struct ext4_xattr_entry *entry, *last, *first;
1233 struct buffer_head *bh = NULL;
1234 struct ext4_xattr_ibody_find *is = NULL;
1235 struct ext4_xattr_block_find *bs = NULL;
1236 char *buffer = NULL, *b_entry_name = NULL;
1237 size_t min_offs, free;
1238 int total_ino;
1239 void *base, *start, *end;
1240 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1241 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1242
1243 down_write(&EXT4_I(inode)->xattr_sem);
1244 retry:
1245 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1246 up_write(&EXT4_I(inode)->xattr_sem);
1247 return 0;
1248 }
1249
1250 header = IHDR(inode, raw_inode);
1251 entry = IFIRST(header);
1252
1253 /*
1254 * Check if enough free space is available in the inode to shift the
1255 * entries ahead by new_extra_isize.
1256 */
1257
1258 base = start = entry;
1259 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1260 min_offs = end - base;
1261 last = entry;
1262 total_ino = sizeof(struct ext4_xattr_ibody_header);
1263
1264 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1265 if (free >= new_extra_isize) {
1266 entry = IFIRST(header);
1267 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1268 - new_extra_isize, (void *)raw_inode +
1269 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1270 (void *)header, total_ino,
1271 inode->i_sb->s_blocksize);
1272 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1273 error = 0;
1274 goto cleanup;
1275 }
1276
1277 /*
1278 * Enough free space isn't available in the inode, check if
1279 * EA block can hold new_extra_isize bytes.
1280 */
1281 if (EXT4_I(inode)->i_file_acl) {
1282 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1283 error = -EIO;
1284 if (!bh)
1285 goto cleanup;
1286 if (ext4_xattr_check_block(inode, bh)) {
1287 EXT4_ERROR_INODE(inode, "bad block %llu",
1288 EXT4_I(inode)->i_file_acl);
1289 error = -EIO;
1290 goto cleanup;
1291 }
1292 base = BHDR(bh);
1293 first = BFIRST(bh);
1294 end = bh->b_data + bh->b_size;
1295 min_offs = end - base;
1296 free = ext4_xattr_free_space(first, &min_offs, base, NULL);
1297 if (free < new_extra_isize) {
1298 if (!tried_min_extra_isize && s_min_extra_isize) {
1299 tried_min_extra_isize++;
1300 new_extra_isize = s_min_extra_isize;
1301 brelse(bh);
1302 goto retry;
1303 }
1304 error = -1;
1305 goto cleanup;
1306 }
1307 } else {
1308 free = inode->i_sb->s_blocksize;
1309 }
1310
1311 while (new_extra_isize > 0) {
1312 size_t offs, size, entry_size;
1313 struct ext4_xattr_entry *small_entry = NULL;
1314 struct ext4_xattr_info i = {
1315 .value = NULL,
1316 .value_len = 0,
1317 };
1318 unsigned int total_size; /* EA entry size + value size */
1319 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1320 unsigned int min_total_size = ~0U;
1321
1322 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1323 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1324 if (!is || !bs) {
1325 error = -ENOMEM;
1326 goto cleanup;
1327 }
1328
1329 is->s.not_found = -ENODATA;
1330 bs->s.not_found = -ENODATA;
1331 is->iloc.bh = NULL;
1332 bs->bh = NULL;
1333
1334 last = IFIRST(header);
1335 /* Find the entry best suited to be pushed into EA block */
1336 entry = NULL;
1337 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1338 total_size =
1339 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1340 EXT4_XATTR_LEN(last->e_name_len);
1341 if (total_size <= free && total_size < min_total_size) {
1342 if (total_size < new_extra_isize) {
1343 small_entry = last;
1344 } else {
1345 entry = last;
1346 min_total_size = total_size;
1347 }
1348 }
1349 }
1350
1351 if (entry == NULL) {
1352 if (small_entry) {
1353 entry = small_entry;
1354 } else {
1355 if (!tried_min_extra_isize &&
1356 s_min_extra_isize) {
1357 tried_min_extra_isize++;
1358 new_extra_isize = s_min_extra_isize;
1359 kfree(is); is = NULL;
1360 kfree(bs); bs = NULL;
1361 brelse(bh);
1362 goto retry;
1363 }
1364 error = -1;
1365 goto cleanup;
1366 }
1367 }
1368 offs = le16_to_cpu(entry->e_value_offs);
1369 size = le32_to_cpu(entry->e_value_size);
1370 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1371 i.name_index = entry->e_name_index,
1372 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1373 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1374 if (!buffer || !b_entry_name) {
1375 error = -ENOMEM;
1376 goto cleanup;
1377 }
1378 /* Save the entry name and the entry value */
1379 memcpy(buffer, (void *)IFIRST(header) + offs,
1380 EXT4_XATTR_SIZE(size));
1381 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1382 b_entry_name[entry->e_name_len] = '\0';
1383 i.name = b_entry_name;
1384
1385 error = ext4_get_inode_loc(inode, &is->iloc);
1386 if (error)
1387 goto cleanup;
1388
1389 error = ext4_xattr_ibody_find(inode, &i, is);
1390 if (error)
1391 goto cleanup;
1392
1393 /* Remove the chosen entry from the inode */
1394 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1395 if (error)
1396 goto cleanup;
1397
1398 entry = IFIRST(header);
1399 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1400 shift_bytes = new_extra_isize;
1401 else
1402 shift_bytes = entry_size + size;
1403 /* Adjust the offsets and shift the remaining entries ahead */
1404 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1405 shift_bytes, (void *)raw_inode +
1406 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1407 (void *)header, total_ino - entry_size,
1408 inode->i_sb->s_blocksize);
1409
1410 extra_isize += shift_bytes;
1411 new_extra_isize -= shift_bytes;
1412 EXT4_I(inode)->i_extra_isize = extra_isize;
1413
1414 i.name = b_entry_name;
1415 i.value = buffer;
1416 i.value_len = size;
1417 error = ext4_xattr_block_find(inode, &i, bs);
1418 if (error)
1419 goto cleanup;
1420
1421 /* Add entry which was removed from the inode into the block */
1422 error = ext4_xattr_block_set(handle, inode, &i, bs);
1423 if (error)
1424 goto cleanup;
1425 kfree(b_entry_name);
1426 kfree(buffer);
1427 b_entry_name = NULL;
1428 buffer = NULL;
1429 brelse(is->iloc.bh);
1430 kfree(is);
1431 kfree(bs);
1432 }
1433 brelse(bh);
1434 up_write(&EXT4_I(inode)->xattr_sem);
1435 return 0;
1436
1437 cleanup:
1438 kfree(b_entry_name);
1439 kfree(buffer);
1440 if (is)
1441 brelse(is->iloc.bh);
1442 kfree(is);
1443 kfree(bs);
1444 brelse(bh);
1445 up_write(&EXT4_I(inode)->xattr_sem);
1446 return error;
1447 }
1448
1449
1450
1451 /*
1452 * ext4_xattr_delete_inode()
1453 *
1454 * Free extended attribute resources associated with this inode. This
1455 * is called immediately before an inode is freed. We have exclusive
1456 * access to the inode.
1457 */
1458 void
1459 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1460 {
1461 struct buffer_head *bh = NULL;
1462
1463 if (!EXT4_I(inode)->i_file_acl)
1464 goto cleanup;
1465 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1466 if (!bh) {
1467 EXT4_ERROR_INODE(inode, "block %llu read error",
1468 EXT4_I(inode)->i_file_acl);
1469 goto cleanup;
1470 }
1471 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1472 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1473 EXT4_ERROR_INODE(inode, "bad block %llu",
1474 EXT4_I(inode)->i_file_acl);
1475 goto cleanup;
1476 }
1477 ext4_xattr_release_block(handle, inode, bh);
1478 EXT4_I(inode)->i_file_acl = 0;
1479
1480 cleanup:
1481 brelse(bh);
1482 }
1483
1484 /*
1485 * ext4_xattr_put_super()
1486 *
1487 * This is called when a file system is unmounted.
1488 */
1489 void
1490 ext4_xattr_put_super(struct super_block *sb)
1491 {
1492 mb_cache_shrink(sb->s_bdev);
1493 }
1494
1495 /*
1496 * ext4_xattr_cache_insert()
1497 *
1498 * Create a new entry in the extended attribute cache, and insert
1499 * it unless such an entry is already in the cache.
1500 *
1501 * Returns 0, or a negative error number on failure.
1502 */
1503 static void
1504 ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
1505 {
1506 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1507 struct mb_cache_entry *ce;
1508 int error;
1509
1510 ce = mb_cache_entry_alloc(ext4_mb_cache, GFP_NOFS);
1511 if (!ce) {
1512 ea_bdebug(bh, "out of memory");
1513 return;
1514 }
1515 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1516 if (error) {
1517 mb_cache_entry_free(ce);
1518 if (error == -EBUSY) {
1519 ea_bdebug(bh, "already in cache");
1520 error = 0;
1521 }
1522 } else {
1523 ea_bdebug(bh, "inserting [%x]", (int)hash);
1524 mb_cache_entry_release(ce);
1525 }
1526 }
1527
1528 /*
1529 * ext4_xattr_cmp()
1530 *
1531 * Compare two extended attribute blocks for equality.
1532 *
1533 * Returns 0 if the blocks are equal, 1 if they differ, and
1534 * a negative error number on errors.
1535 */
1536 static int
1537 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1538 struct ext4_xattr_header *header2)
1539 {
1540 struct ext4_xattr_entry *entry1, *entry2;
1541
1542 entry1 = ENTRY(header1+1);
1543 entry2 = ENTRY(header2+1);
1544 while (!IS_LAST_ENTRY(entry1)) {
1545 if (IS_LAST_ENTRY(entry2))
1546 return 1;
1547 if (entry1->e_hash != entry2->e_hash ||
1548 entry1->e_name_index != entry2->e_name_index ||
1549 entry1->e_name_len != entry2->e_name_len ||
1550 entry1->e_value_size != entry2->e_value_size ||
1551 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1552 return 1;
1553 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1554 return -EIO;
1555 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1556 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1557 le32_to_cpu(entry1->e_value_size)))
1558 return 1;
1559
1560 entry1 = EXT4_XATTR_NEXT(entry1);
1561 entry2 = EXT4_XATTR_NEXT(entry2);
1562 }
1563 if (!IS_LAST_ENTRY(entry2))
1564 return 1;
1565 return 0;
1566 }
1567
1568 /*
1569 * ext4_xattr_cache_find()
1570 *
1571 * Find an identical extended attribute block.
1572 *
1573 * Returns a pointer to the block found, or NULL if such a block was
1574 * not found or an error occurred.
1575 */
1576 static struct buffer_head *
1577 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1578 struct mb_cache_entry **pce)
1579 {
1580 __u32 hash = le32_to_cpu(header->h_hash);
1581 struct mb_cache_entry *ce;
1582 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
1583
1584 if (!header->h_hash)
1585 return NULL; /* never share */
1586 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1587 again:
1588 ce = mb_cache_entry_find_first(ext4_mb_cache, inode->i_sb->s_bdev,
1589 hash);
1590 while (ce) {
1591 struct buffer_head *bh;
1592
1593 if (IS_ERR(ce)) {
1594 if (PTR_ERR(ce) == -EAGAIN)
1595 goto again;
1596 break;
1597 }
1598 bh = sb_bread(inode->i_sb, ce->e_block);
1599 if (!bh) {
1600 EXT4_ERROR_INODE(inode, "block %lu read error",
1601 (unsigned long) ce->e_block);
1602 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1603 EXT4_XATTR_REFCOUNT_MAX) {
1604 ea_idebug(inode, "block %lu refcount %d>=%d",
1605 (unsigned long) ce->e_block,
1606 le32_to_cpu(BHDR(bh)->h_refcount),
1607 EXT4_XATTR_REFCOUNT_MAX);
1608 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1609 *pce = ce;
1610 return bh;
1611 }
1612 brelse(bh);
1613 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1614 }
1615 return NULL;
1616 }
1617
1618 #define NAME_HASH_SHIFT 5
1619 #define VALUE_HASH_SHIFT 16
1620
1621 /*
1622 * ext4_xattr_hash_entry()
1623 *
1624 * Compute the hash of an extended attribute.
1625 */
1626 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1627 struct ext4_xattr_entry *entry)
1628 {
1629 __u32 hash = 0;
1630 char *name = entry->e_name;
1631 int n;
1632
1633 for (n = 0; n < entry->e_name_len; n++) {
1634 hash = (hash << NAME_HASH_SHIFT) ^
1635 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1636 *name++;
1637 }
1638
1639 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1640 __le32 *value = (__le32 *)((char *)header +
1641 le16_to_cpu(entry->e_value_offs));
1642 for (n = (le32_to_cpu(entry->e_value_size) +
1643 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1644 hash = (hash << VALUE_HASH_SHIFT) ^
1645 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1646 le32_to_cpu(*value++);
1647 }
1648 }
1649 entry->e_hash = cpu_to_le32(hash);
1650 }
1651
1652 #undef NAME_HASH_SHIFT
1653 #undef VALUE_HASH_SHIFT
1654
1655 #define BLOCK_HASH_SHIFT 16
1656
1657 /*
1658 * ext4_xattr_rehash()
1659 *
1660 * Re-compute the extended attribute hash value after an entry has changed.
1661 */
1662 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1663 struct ext4_xattr_entry *entry)
1664 {
1665 struct ext4_xattr_entry *here;
1666 __u32 hash = 0;
1667
1668 ext4_xattr_hash_entry(header, entry);
1669 here = ENTRY(header+1);
1670 while (!IS_LAST_ENTRY(here)) {
1671 if (!here->e_hash) {
1672 /* Block is not shared if an entry's hash value == 0 */
1673 hash = 0;
1674 break;
1675 }
1676 hash = (hash << BLOCK_HASH_SHIFT) ^
1677 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1678 le32_to_cpu(here->e_hash);
1679 here = EXT4_XATTR_NEXT(here);
1680 }
1681 header->h_hash = cpu_to_le32(hash);
1682 }
1683
1684 #undef BLOCK_HASH_SHIFT
1685
1686 #define HASH_BUCKET_BITS 10
1687
1688 struct mb_cache *
1689 ext4_xattr_create_cache(char *name)
1690 {
1691 return mb_cache_create(name, HASH_BUCKET_BITS);
1692 }
1693
1694 void ext4_xattr_destroy_cache(struct mb_cache *cache)
1695 {
1696 if (cache)
1697 mb_cache_destroy(cache);
1698 }
1699
This page took 0.064729 seconds and 5 git commands to generate.