?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'fixes-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git/arm...
[linux-2.6.git] / fs / ext4 / xattr.c
blob1423c4816a4783da4208b6d51fc33815b25597f4
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 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 struct mb_cache *ext4_xattr_cache;
94
95 static const struct xattr_handler *ext4_xattr_handler_map[] = {
96 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
97 #ifdef CONFIG_EXT4_FS_POSIX_ACL
98 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
99 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
100 #endif
101 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
102 #ifdef CONFIG_EXT4_FS_SECURITY
103 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
104 #endif
105 };
106
107 const struct xattr_handler *ext4_xattr_handlers[] = {
108 &ext4_xattr_user_handler,
109 &ext4_xattr_trusted_handler,
110 #ifdef CONFIG_EXT4_FS_POSIX_ACL
111 &ext4_xattr_acl_access_handler,
112 &ext4_xattr_acl_default_handler,
113 #endif
114 #ifdef CONFIG_EXT4_FS_SECURITY
115 &ext4_xattr_security_handler,
116 #endif
117 NULL
118 };
119
120 static __le32 ext4_xattr_block_csum(struct inode *inode,
121 sector_t block_nr,
122 struct ext4_xattr_header *hdr)
123 {
124 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
125 __u32 csum;
126 __le32 save_csum;
127 __le64 dsk_block_nr = cpu_to_le64(block_nr);
128
129 save_csum = hdr->h_checksum;
130 hdr->h_checksum = 0;
131 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
132 sizeof(dsk_block_nr));
133 csum = ext4_chksum(sbi, csum, (__u8 *)hdr,
134 EXT4_BLOCK_SIZE(inode->i_sb));
135
136 hdr->h_checksum = save_csum;
137 return cpu_to_le32(csum);
138 }
139
140 static int ext4_xattr_block_csum_verify(struct inode *inode,
141 sector_t block_nr,
142 struct ext4_xattr_header *hdr)
143 {
144 if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
145 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
146 (hdr->h_checksum != ext4_xattr_block_csum(inode, block_nr, hdr)))
147 return 0;
148 return 1;
149 }
150
151 static void ext4_xattr_block_csum_set(struct inode *inode,
152 sector_t block_nr,
153 struct ext4_xattr_header *hdr)
154 {
155 if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
156 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
157 return;
158
159 hdr->h_checksum = ext4_xattr_block_csum(inode, block_nr, hdr);
160 }
161
162 static inline int ext4_handle_dirty_xattr_block(handle_t *handle,
163 struct inode *inode,
164 struct buffer_head *bh)
165 {
166 ext4_xattr_block_csum_set(inode, bh->b_blocknr, BHDR(bh));
167 return ext4_handle_dirty_metadata(handle, inode, bh);
168 }
169
170 static inline const struct xattr_handler *
171 ext4_xattr_handler(int name_index)
172 {
173 const struct xattr_handler *handler = NULL;
174
175 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
176 handler = ext4_xattr_handler_map[name_index];
177 return handler;
178 }
179
180 /*
181 * Inode operation listxattr()
182 *
183 * dentry->d_inode->i_mutex: don't care
184 */
185 ssize_t
186 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
187 {
188 return ext4_xattr_list(dentry, buffer, size);
189 }
190
191 static int
192 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
193 {
194 while (!IS_LAST_ENTRY(entry)) {
195 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
196 if ((void *)next >= end)
197 return -EIO;
198 entry = next;
199 }
200 return 0;
201 }
202
203 static inline int
204 ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
205 {
206 int error;
207
208 if (buffer_verified(bh))
209 return 0;
210
211 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
212 BHDR(bh)->h_blocks != cpu_to_le32(1))
213 return -EIO;
214 if (!ext4_xattr_block_csum_verify(inode, bh->b_blocknr, BHDR(bh)))
215 return -EIO;
216 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
217 if (!error)
218 set_buffer_verified(bh);
219 return error;
220 }
221
222 static inline int
223 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
224 {
225 size_t value_size = le32_to_cpu(entry->e_value_size);
226
227 if (entry->e_value_block != 0 || value_size > size ||
228 le16_to_cpu(entry->e_value_offs) + value_size > size)
229 return -EIO;
230 return 0;
231 }
232
233 static int
234 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
235 const char *name, size_t size, int sorted)
236 {
237 struct ext4_xattr_entry *entry;
238 size_t name_len;
239 int cmp = 1;
240
241 if (name == NULL)
242 return -EINVAL;
243 name_len = strlen(name);
244 entry = *pentry;
245 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
246 cmp = name_index - entry->e_name_index;
247 if (!cmp)
248 cmp = name_len - entry->e_name_len;
249 if (!cmp)
250 cmp = memcmp(name, entry->e_name, name_len);
251 if (cmp <= 0 && (sorted || cmp == 0))
252 break;
253 }
254 *pentry = entry;
255 if (!cmp && ext4_xattr_check_entry(entry, size))
256 return -EIO;
257 return cmp ? -ENODATA : 0;
258 }
259
260 static int
261 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
262 void *buffer, size_t buffer_size)
263 {
264 struct buffer_head *bh = NULL;
265 struct ext4_xattr_entry *entry;
266 size_t size;
267 int error;
268
269 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
270 name_index, name, buffer, (long)buffer_size);
271
272 error = -ENODATA;
273 if (!EXT4_I(inode)->i_file_acl)
274 goto cleanup;
275 ea_idebug(inode, "reading block %llu",
276 (unsigned long long)EXT4_I(inode)->i_file_acl);
277 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
278 if (!bh)
279 goto cleanup;
280 ea_bdebug(bh, "b_count=%d, refcount=%d",
281 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
282 if (ext4_xattr_check_block(inode, bh)) {
283 bad_block:
284 EXT4_ERROR_INODE(inode, "bad block %llu",
285 EXT4_I(inode)->i_file_acl);
286 error = -EIO;
287 goto cleanup;
288 }
289 ext4_xattr_cache_insert(bh);
290 entry = BFIRST(bh);
291 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
292 if (error == -EIO)
293 goto bad_block;
294 if (error)
295 goto cleanup;
296 size = le32_to_cpu(entry->e_value_size);
297 if (buffer) {
298 error = -ERANGE;
299 if (size > buffer_size)
300 goto cleanup;
301 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
302 size);
303 }
304 error = size;
305
306 cleanup:
307 brelse(bh);
308 return error;
309 }
310
311 int
312 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
313 void *buffer, size_t buffer_size)
314 {
315 struct ext4_xattr_ibody_header *header;
316 struct ext4_xattr_entry *entry;
317 struct ext4_inode *raw_inode;
318 struct ext4_iloc iloc;
319 size_t size;
320 void *end;
321 int error;
322
323 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
324 return -ENODATA;
325 error = ext4_get_inode_loc(inode, &iloc);
326 if (error)
327 return error;
328 raw_inode = ext4_raw_inode(&iloc);
329 header = IHDR(inode, raw_inode);
330 entry = IFIRST(header);
331 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
332 error = ext4_xattr_check_names(entry, end);
333 if (error)
334 goto cleanup;
335 error = ext4_xattr_find_entry(&entry, name_index, name,
336 end - (void *)entry, 0);
337 if (error)
338 goto cleanup;
339 size = le32_to_cpu(entry->e_value_size);
340 if (buffer) {
341 error = -ERANGE;
342 if (size > buffer_size)
343 goto cleanup;
344 memcpy(buffer, (void *)IFIRST(header) +
345 le16_to_cpu(entry->e_value_offs), size);
346 }
347 error = size;
348
349 cleanup:
350 brelse(iloc.bh);
351 return error;
352 }
353
354 /*
355 * ext4_xattr_get()
356 *
357 * Copy an extended attribute into the buffer
358 * provided, or compute the buffer size required.
359 * Buffer is NULL to compute the size of the buffer required.
360 *
361 * Returns a negative error number on failure, or the number of bytes
362 * used / required on success.
363 */
364 int
365 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
366 void *buffer, size_t buffer_size)
367 {
368 int error;
369
370 down_read(&EXT4_I(inode)->xattr_sem);
371 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
372 buffer_size);
373 if (error == -ENODATA)
374 error = ext4_xattr_block_get(inode, name_index, name, buffer,
375 buffer_size);
376 up_read(&EXT4_I(inode)->xattr_sem);
377 return error;
378 }
379
380 static int
381 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
382 char *buffer, size_t buffer_size)
383 {
384 size_t rest = buffer_size;
385
386 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
387 const struct xattr_handler *handler =
388 ext4_xattr_handler(entry->e_name_index);
389
390 if (handler) {
391 size_t size = handler->list(dentry, buffer, rest,
392 entry->e_name,
393 entry->e_name_len,
394 handler->flags);
395 if (buffer) {
396 if (size > rest)
397 return -ERANGE;
398 buffer += size;
399 }
400 rest -= size;
401 }
402 }
403 return buffer_size - rest;
404 }
405
406 static int
407 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
408 {
409 struct inode *inode = dentry->d_inode;
410 struct buffer_head *bh = NULL;
411 int error;
412
413 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
414 buffer, (long)buffer_size);
415
416 error = 0;
417 if (!EXT4_I(inode)->i_file_acl)
418 goto cleanup;
419 ea_idebug(inode, "reading block %llu",
420 (unsigned long long)EXT4_I(inode)->i_file_acl);
421 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
422 error = -EIO;
423 if (!bh)
424 goto cleanup;
425 ea_bdebug(bh, "b_count=%d, refcount=%d",
426 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
427 if (ext4_xattr_check_block(inode, bh)) {
428 EXT4_ERROR_INODE(inode, "bad block %llu",
429 EXT4_I(inode)->i_file_acl);
430 error = -EIO;
431 goto cleanup;
432 }
433 ext4_xattr_cache_insert(bh);
434 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
435
436 cleanup:
437 brelse(bh);
438
439 return error;
440 }
441
442 static int
443 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
444 {
445 struct inode *inode = dentry->d_inode;
446 struct ext4_xattr_ibody_header *header;
447 struct ext4_inode *raw_inode;
448 struct ext4_iloc iloc;
449 void *end;
450 int error;
451
452 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
453 return 0;
454 error = ext4_get_inode_loc(inode, &iloc);
455 if (error)
456 return error;
457 raw_inode = ext4_raw_inode(&iloc);
458 header = IHDR(inode, raw_inode);
459 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
460 error = ext4_xattr_check_names(IFIRST(header), end);
461 if (error)
462 goto cleanup;
463 error = ext4_xattr_list_entries(dentry, IFIRST(header),
464 buffer, buffer_size);
465
466 cleanup:
467 brelse(iloc.bh);
468 return error;
469 }
470
471 /*
472 * ext4_xattr_list()
473 *
474 * Copy a list of attribute names into the buffer
475 * provided, or compute the buffer size required.
476 * Buffer is NULL to compute the size of the buffer required.
477 *
478 * Returns a negative error number on failure, or the number of bytes
479 * used / required on success.
480 */
481 static int
482 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
483 {
484 int ret, ret2;
485
486 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
487 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
488 if (ret < 0)
489 goto errout;
490 if (buffer) {
491 buffer += ret;
492 buffer_size -= ret;
493 }
494 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
495 if (ret < 0)
496 goto errout;
497 ret += ret2;
498 errout:
499 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
500 return ret;
501 }
502
503 /*
504 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
505 * not set, set it.
506 */
507 static void ext4_xattr_update_super_block(handle_t *handle,
508 struct super_block *sb)
509 {
510 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
511 return;
512
513 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
514 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
515 ext4_handle_dirty_super(handle, sb);
516 }
517 }
518
519 /*
520 * Release the xattr block BH: If the reference count is > 1, decrement
521 * it; otherwise free the block.
522 */
523 static void
524 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
525 struct buffer_head *bh)
526 {
527 struct mb_cache_entry *ce = NULL;
528 int error = 0;
529
530 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
531 error = ext4_journal_get_write_access(handle, bh);
532 if (error)
533 goto out;
534
535 lock_buffer(bh);
536 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
537 ea_bdebug(bh, "refcount now=0; freeing");
538 if (ce)
539 mb_cache_entry_free(ce);
540 get_bh(bh);
541 ext4_free_blocks(handle, inode, bh, 0, 1,
542 EXT4_FREE_BLOCKS_METADATA |
543 EXT4_FREE_BLOCKS_FORGET);
544 unlock_buffer(bh);
545 } else {
546 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
547 if (ce)
548 mb_cache_entry_release(ce);
549 unlock_buffer(bh);
550 error = ext4_handle_dirty_xattr_block(handle, inode, bh);
551 if (IS_SYNC(inode))
552 ext4_handle_sync(handle);
553 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
554 ea_bdebug(bh, "refcount now=%d; releasing",
555 le32_to_cpu(BHDR(bh)->h_refcount));
556 }
557 out:
558 ext4_std_error(inode->i_sb, error);
559 return;
560 }
561
562 /*
563 * Find the available free space for EAs. This also returns the total number of
564 * bytes used by EA entries.
565 */
566 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
567 size_t *min_offs, void *base, int *total)
568 {
569 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
570 *total += EXT4_XATTR_LEN(last->e_name_len);
571 if (!last->e_value_block && last->e_value_size) {
572 size_t offs = le16_to_cpu(last->e_value_offs);
573 if (offs < *min_offs)
574 *min_offs = offs;
575 }
576 }
577 return (*min_offs - ((void *)last - base) - sizeof(__u32));
578 }
579
580 static int
581 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
582 {
583 struct ext4_xattr_entry *last;
584 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
585
586 /* Compute min_offs and last. */
587 last = s->first;
588 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
589 if (!last->e_value_block && last->e_value_size) {
590 size_t offs = le16_to_cpu(last->e_value_offs);
591 if (offs < min_offs)
592 min_offs = offs;
593 }
594 }
595 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
596 if (!s->not_found) {
597 if (!s->here->e_value_block && s->here->e_value_size) {
598 size_t size = le32_to_cpu(s->here->e_value_size);
599 free += EXT4_XATTR_SIZE(size);
600 }
601 free += EXT4_XATTR_LEN(name_len);
602 }
603 if (i->value) {
604 if (free < EXT4_XATTR_SIZE(i->value_len) ||
605 free < EXT4_XATTR_LEN(name_len) +
606 EXT4_XATTR_SIZE(i->value_len))
607 return -ENOSPC;
608 }
609
610 if (i->value && s->not_found) {
611 /* Insert the new name. */
612 size_t size = EXT4_XATTR_LEN(name_len);
613 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
614 memmove((void *)s->here + size, s->here, rest);
615 memset(s->here, 0, size);
616 s->here->e_name_index = i->name_index;
617 s->here->e_name_len = name_len;
618 memcpy(s->here->e_name, i->name, name_len);
619 } else {
620 if (!s->here->e_value_block && s->here->e_value_size) {
621 void *first_val = s->base + min_offs;
622 size_t offs = le16_to_cpu(s->here->e_value_offs);
623 void *val = s->base + offs;
624 size_t size = EXT4_XATTR_SIZE(
625 le32_to_cpu(s->here->e_value_size));
626
627 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
628 /* The old and the new value have the same
629 size. Just replace. */
630 s->here->e_value_size =
631 cpu_to_le32(i->value_len);
632 if (i->value == EXT4_ZERO_XATTR_VALUE) {
633 memset(val, 0, size);
634 } else {
635 /* Clear pad bytes first. */
636 memset(val + size - EXT4_XATTR_PAD, 0,
637 EXT4_XATTR_PAD);
638 memcpy(val, i->value, i->value_len);
639 }
640 return 0;
641 }
642
643 /* Remove the old value. */
644 memmove(first_val + size, first_val, val - first_val);
645 memset(first_val, 0, size);
646 s->here->e_value_size = 0;
647 s->here->e_value_offs = 0;
648 min_offs += size;
649
650 /* Adjust all value offsets. */
651 last = s->first;
652 while (!IS_LAST_ENTRY(last)) {
653 size_t o = le16_to_cpu(last->e_value_offs);
654 if (!last->e_value_block &&
655 last->e_value_size && o < offs)
656 last->e_value_offs =
657 cpu_to_le16(o + size);
658 last = EXT4_XATTR_NEXT(last);
659 }
660 }
661 if (!i->value) {
662 /* Remove the old name. */
663 size_t size = EXT4_XATTR_LEN(name_len);
664 last = ENTRY((void *)last - size);
665 memmove(s->here, (void *)s->here + size,
666 (void *)last - (void *)s->here + sizeof(__u32));
667 memset(last, 0, size);
668 }
669 }
670
671 if (i->value) {
672 /* Insert the new value. */
673 s->here->e_value_size = cpu_to_le32(i->value_len);
674 if (i->value_len) {
675 size_t size = EXT4_XATTR_SIZE(i->value_len);
676 void *val = s->base + min_offs - size;
677 s->here->e_value_offs = cpu_to_le16(min_offs - size);
678 if (i->value == EXT4_ZERO_XATTR_VALUE) {
679 memset(val, 0, size);
680 } else {
681 /* Clear the pad bytes first. */
682 memset(val + size - EXT4_XATTR_PAD, 0,
683 EXT4_XATTR_PAD);
684 memcpy(val, i->value, i->value_len);
685 }
686 }
687 }
688 return 0;
689 }
690
691 struct ext4_xattr_block_find {
692 struct ext4_xattr_search s;
693 struct buffer_head *bh;
694 };
695
696 static int
697 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
698 struct ext4_xattr_block_find *bs)
699 {
700 struct super_block *sb = inode->i_sb;
701 int error;
702
703 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
704 i->name_index, i->name, i->value, (long)i->value_len);
705
706 if (EXT4_I(inode)->i_file_acl) {
707 /* The inode already has an extended attribute block. */
708 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
709 error = -EIO;
710 if (!bs->bh)
711 goto cleanup;
712 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
713 atomic_read(&(bs->bh->b_count)),
714 le32_to_cpu(BHDR(bs->bh)->h_refcount));
715 if (ext4_xattr_check_block(inode, bs->bh)) {
716 EXT4_ERROR_INODE(inode, "bad block %llu",
717 EXT4_I(inode)->i_file_acl);
718 error = -EIO;
719 goto cleanup;
720 }
721 /* Find the named attribute. */
722 bs->s.base = BHDR(bs->bh);
723 bs->s.first = BFIRST(bs->bh);
724 bs->s.end = bs->bh->b_data + bs->bh->b_size;
725 bs->s.here = bs->s.first;
726 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
727 i->name, bs->bh->b_size, 1);
728 if (error && error != -ENODATA)
729 goto cleanup;
730 bs->s.not_found = error;
731 }
732 error = 0;
733
734 cleanup:
735 return error;
736 }
737
738 static int
739 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
740 struct ext4_xattr_info *i,
741 struct ext4_xattr_block_find *bs)
742 {
743 struct super_block *sb = inode->i_sb;
744 struct buffer_head *new_bh = NULL;
745 struct ext4_xattr_search *s = &bs->s;
746 struct mb_cache_entry *ce = NULL;
747 int error = 0;
748
749 #define header(x) ((struct ext4_xattr_header *)(x))
750
751 if (i->value && i->value_len > sb->s_blocksize)
752 return -ENOSPC;
753 if (s->base) {
754 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
755 bs->bh->b_blocknr);
756 error = ext4_journal_get_write_access(handle, bs->bh);
757 if (error)
758 goto cleanup;
759 lock_buffer(bs->bh);
760
761 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
762 if (ce) {
763 mb_cache_entry_free(ce);
764 ce = NULL;
765 }
766 ea_bdebug(bs->bh, "modifying in-place");
767 error = ext4_xattr_set_entry(i, s);
768 if (!error) {
769 if (!IS_LAST_ENTRY(s->first))
770 ext4_xattr_rehash(header(s->base),
771 s->here);
772 ext4_xattr_cache_insert(bs->bh);
773 }
774 unlock_buffer(bs->bh);
775 if (error == -EIO)
776 goto bad_block;
777 if (!error)
778 error = ext4_handle_dirty_xattr_block(handle,
779 inode,
780 bs->bh);
781 if (error)
782 goto cleanup;
783 goto inserted;
784 } else {
785 int offset = (char *)s->here - bs->bh->b_data;
786
787 unlock_buffer(bs->bh);
788 if (ce) {
789 mb_cache_entry_release(ce);
790 ce = NULL;
791 }
792 ea_bdebug(bs->bh, "cloning");
793 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
794 error = -ENOMEM;
795 if (s->base == NULL)
796 goto cleanup;
797 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
798 s->first = ENTRY(header(s->base)+1);
799 header(s->base)->h_refcount = cpu_to_le32(1);
800 s->here = ENTRY(s->base + offset);
801 s->end = s->base + bs->bh->b_size;
802 }
803 } else {
804 /* Allocate a buffer where we construct the new block. */
805 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
806 /* assert(header == s->base) */
807 error = -ENOMEM;
808 if (s->base == NULL)
809 goto cleanup;
810 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
811 header(s->base)->h_blocks = cpu_to_le32(1);
812 header(s->base)->h_refcount = cpu_to_le32(1);
813 s->first = ENTRY(header(s->base)+1);
814 s->here = ENTRY(header(s->base)+1);
815 s->end = s->base + sb->s_blocksize;
816 }
817
818 error = ext4_xattr_set_entry(i, s);
819 if (error == -EIO)
820 goto bad_block;
821 if (error)
822 goto cleanup;
823 if (!IS_LAST_ENTRY(s->first))
824 ext4_xattr_rehash(header(s->base), s->here);
825
826 inserted:
827 if (!IS_LAST_ENTRY(s->first)) {
828 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
829 if (new_bh) {
830 /* We found an identical block in the cache. */
831 if (new_bh == bs->bh)
832 ea_bdebug(new_bh, "keeping");
833 else {
834 /* The old block is released after updating
835 the inode. */
836 error = dquot_alloc_block(inode,
837 EXT4_C2B(EXT4_SB(sb), 1));
838 if (error)
839 goto cleanup;
840 error = ext4_journal_get_write_access(handle,
841 new_bh);
842 if (error)
843 goto cleanup_dquot;
844 lock_buffer(new_bh);
845 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
846 ea_bdebug(new_bh, "reusing; refcount now=%d",
847 le32_to_cpu(BHDR(new_bh)->h_refcount));
848 unlock_buffer(new_bh);
849 error = ext4_handle_dirty_xattr_block(handle,
850 inode,
851 new_bh);
852 if (error)
853 goto cleanup_dquot;
854 }
855 mb_cache_entry_release(ce);
856 ce = NULL;
857 } else if (bs->bh && s->base == bs->bh->b_data) {
858 /* We were modifying this block in-place. */
859 ea_bdebug(bs->bh, "keeping this block");
860 new_bh = bs->bh;
861 get_bh(new_bh);
862 } else {
863 /* We need to allocate a new block */
864 ext4_fsblk_t goal, block;
865
866 goal = ext4_group_first_block_no(sb,
867 EXT4_I(inode)->i_block_group);
868
869 /* non-extent files can't have physical blocks past 2^32 */
870 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
871 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
872
873 /*
874 * take i_data_sem because we will test
875 * i_delalloc_reserved_flag in ext4_mb_new_blocks
876 */
877 down_read((&EXT4_I(inode)->i_data_sem));
878 block = ext4_new_meta_blocks(handle, inode, goal, 0,
879 NULL, &error);
880 up_read((&EXT4_I(inode)->i_data_sem));
881 if (error)
882 goto cleanup;
883
884 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
885 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
886
887 ea_idebug(inode, "creating block %llu",
888 (unsigned long long)block);
889
890 new_bh = sb_getblk(sb, block);
891 if (unlikely(!new_bh)) {
892 error = -ENOMEM;
893 getblk_failed:
894 ext4_free_blocks(handle, inode, NULL, block, 1,
895 EXT4_FREE_BLOCKS_METADATA);
896 goto cleanup;
897 }
898 lock_buffer(new_bh);
899 error = ext4_journal_get_create_access(handle, new_bh);
900 if (error) {
901 unlock_buffer(new_bh);
902 error = -EIO;
903 goto getblk_failed;
904 }
905 memcpy(new_bh->b_data, s->base, new_bh->b_size);
906 set_buffer_uptodate(new_bh);
907 unlock_buffer(new_bh);
908 ext4_xattr_cache_insert(new_bh);
909 error = ext4_handle_dirty_xattr_block(handle,
910 inode, new_bh);
911 if (error)
912 goto cleanup;
913 }
914 }
915
916 /* Update the inode. */
917 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
918
919 /* Drop the previous xattr block. */
920 if (bs->bh && bs->bh != new_bh)
921 ext4_xattr_release_block(handle, inode, bs->bh);
922 error = 0;
923
924 cleanup:
925 if (ce)
926 mb_cache_entry_release(ce);
927 brelse(new_bh);
928 if (!(bs->bh && s->base == bs->bh->b_data))
929 kfree(s->base);
930
931 return error;
932
933 cleanup_dquot:
934 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
935 goto cleanup;
936
937 bad_block:
938 EXT4_ERROR_INODE(inode, "bad block %llu",
939 EXT4_I(inode)->i_file_acl);
940 goto cleanup;
941
942 #undef header
943 }
944
945 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
946 struct ext4_xattr_ibody_find *is)
947 {
948 struct ext4_xattr_ibody_header *header;
949 struct ext4_inode *raw_inode;
950 int error;
951
952 if (EXT4_I(inode)->i_extra_isize == 0)
953 return 0;
954 raw_inode = ext4_raw_inode(&is->iloc);
955 header = IHDR(inode, raw_inode);
956 is->s.base = is->s.first = IFIRST(header);
957 is->s.here = is->s.first;
958 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
959 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
960 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
961 if (error)
962 return error;
963 /* Find the named attribute. */
964 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
965 i->name, is->s.end -
966 (void *)is->s.base, 0);
967 if (error && error != -ENODATA)
968 return error;
969 is->s.not_found = error;
970 }
971 return 0;
972 }
973
974 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
975 struct ext4_xattr_info *i,
976 struct ext4_xattr_ibody_find *is)
977 {
978 struct ext4_xattr_ibody_header *header;
979 struct ext4_xattr_search *s = &is->s;
980 int error;
981
982 if (EXT4_I(inode)->i_extra_isize == 0)
983 return -ENOSPC;
984 error = ext4_xattr_set_entry(i, s);
985 if (error) {
986 if (error == -ENOSPC &&
987 ext4_has_inline_data(inode)) {
988 error = ext4_try_to_evict_inline_data(handle, inode,
989 EXT4_XATTR_LEN(strlen(i->name) +
990 EXT4_XATTR_SIZE(i->value_len)));
991 if (error)
992 return error;
993 error = ext4_xattr_ibody_find(inode, i, is);
994 if (error)
995 return error;
996 error = ext4_xattr_set_entry(i, s);
997 }
998 if (error)
999 return error;
1000 }
1001 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1002 if (!IS_LAST_ENTRY(s->first)) {
1003 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1004 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1005 } else {
1006 header->h_magic = cpu_to_le32(0);
1007 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1008 }
1009 return 0;
1010 }
1011
1012 static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
1013 struct ext4_xattr_info *i,
1014 struct ext4_xattr_ibody_find *is)
1015 {
1016 struct ext4_xattr_ibody_header *header;
1017 struct ext4_xattr_search *s = &is->s;
1018 int error;
1019
1020 if (EXT4_I(inode)->i_extra_isize == 0)
1021 return -ENOSPC;
1022 error = ext4_xattr_set_entry(i, s);
1023 if (error)
1024 return error;
1025 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1026 if (!IS_LAST_ENTRY(s->first)) {
1027 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1028 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1029 } else {
1030 header->h_magic = cpu_to_le32(0);
1031 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1032 }
1033 return 0;
1034 }
1035
1036 /*
1037 * ext4_xattr_set_handle()
1038 *
1039 * Create, replace or remove an extended attribute for this inode. Value
1040 * is NULL to remove an existing extended attribute, and non-NULL to
1041 * either replace an existing extended attribute, or create a new extended
1042 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1043 * specify that an extended attribute must exist and must not exist
1044 * previous to the call, respectively.
1045 *
1046 * Returns 0, or a negative error number on failure.
1047 */
1048 int
1049 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1050 const char *name, const void *value, size_t value_len,
1051 int flags)
1052 {
1053 struct ext4_xattr_info i = {
1054 .name_index = name_index,
1055 .name = name,
1056 .value = value,
1057 .value_len = value_len,
1058
1059 };
1060 struct ext4_xattr_ibody_find is = {
1061 .s = { .not_found = -ENODATA, },
1062 };
1063 struct ext4_xattr_block_find bs = {
1064 .s = { .not_found = -ENODATA, },
1065 };
1066 unsigned long no_expand;
1067 int error;
1068
1069 if (!name)
1070 return -EINVAL;
1071 if (strlen(name) > 255)
1072 return -ERANGE;
1073 down_write(&EXT4_I(inode)->xattr_sem);
1074 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
1075 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
1076
1077 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1078 if (error)
1079 goto cleanup;
1080
1081 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1082 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1083 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1084 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1085 }
1086
1087 error = ext4_xattr_ibody_find(inode, &i, &is);
1088 if (error)
1089 goto cleanup;
1090 if (is.s.not_found)
1091 error = ext4_xattr_block_find(inode, &i, &bs);
1092 if (error)
1093 goto cleanup;
1094 if (is.s.not_found && bs.s.not_found) {
1095 error = -ENODATA;
1096 if (flags & XATTR_REPLACE)
1097 goto cleanup;
1098 error = 0;
1099 if (!value)
1100 goto cleanup;
1101 } else {
1102 error = -EEXIST;
1103 if (flags & XATTR_CREATE)
1104 goto cleanup;
1105 }
1106 if (!value) {
1107 if (!is.s.not_found)
1108 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1109 else if (!bs.s.not_found)
1110 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1111 } else {
1112 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1113 if (!error && !bs.s.not_found) {
1114 i.value = NULL;
1115 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1116 } else if (error == -ENOSPC) {
1117 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1118 error = ext4_xattr_block_find(inode, &i, &bs);
1119 if (error)
1120 goto cleanup;
1121 }
1122 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1123 if (error)
1124 goto cleanup;
1125 if (!is.s.not_found) {
1126 i.value = NULL;
1127 error = ext4_xattr_ibody_set(handle, inode, &i,
1128 &is);
1129 }
1130 }
1131 }
1132 if (!error) {
1133 ext4_xattr_update_super_block(handle, inode->i_sb);
1134 inode->i_ctime = ext4_current_time(inode);
1135 if (!value)
1136 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1137 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1138 /*
1139 * The bh is consumed by ext4_mark_iloc_dirty, even with
1140 * error != 0.
1141 */
1142 is.iloc.bh = NULL;
1143 if (IS_SYNC(inode))
1144 ext4_handle_sync(handle);
1145 }
1146
1147 cleanup:
1148 brelse(is.iloc.bh);
1149 brelse(bs.bh);
1150 if (no_expand == 0)
1151 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1152 up_write(&EXT4_I(inode)->xattr_sem);
1153 return error;
1154 }
1155
1156 /*
1157 * ext4_xattr_set()
1158 *
1159 * Like ext4_xattr_set_handle, but start from an inode. This extended
1160 * attribute modification is a filesystem transaction by itself.
1161 *
1162 * Returns 0, or a negative error number on failure.
1163 */
1164 int
1165 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1166 const void *value, size_t value_len, int flags)
1167 {
1168 handle_t *handle;
1169 int error, retries = 0;
1170 int credits = ext4_jbd2_credits_xattr(inode);
1171
1172 retry:
1173 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1174 if (IS_ERR(handle)) {
1175 error = PTR_ERR(handle);
1176 } else {
1177 int error2;
1178
1179 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1180 value, value_len, flags);
1181 error2 = ext4_journal_stop(handle);
1182 if (error == -ENOSPC &&
1183 ext4_should_retry_alloc(inode->i_sb, &retries))
1184 goto retry;
1185 if (error == 0)
1186 error = error2;
1187 }
1188
1189 return error;
1190 }
1191
1192 /*
1193 * Shift the EA entries in the inode to create space for the increased
1194 * i_extra_isize.
1195 */
1196 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1197 int value_offs_shift, void *to,
1198 void *from, size_t n, int blocksize)
1199 {
1200 struct ext4_xattr_entry *last = entry;
1201 int new_offs;
1202
1203 /* Adjust the value offsets of the entries */
1204 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1205 if (!last->e_value_block && last->e_value_size) {
1206 new_offs = le16_to_cpu(last->e_value_offs) +
1207 value_offs_shift;
1208 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1209 > blocksize);
1210 last->e_value_offs = cpu_to_le16(new_offs);
1211 }
1212 }
1213 /* Shift the entries by n bytes */
1214 memmove(to, from, n);
1215 }
1216
1217 /*
1218 * Expand an inode by new_extra_isize bytes when EAs are present.
1219 * Returns 0 on success or negative error number on failure.
1220 */
1221 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1222 struct ext4_inode *raw_inode, handle_t *handle)
1223 {
1224 struct ext4_xattr_ibody_header *header;
1225 struct ext4_xattr_entry *entry, *last, *first;
1226 struct buffer_head *bh = NULL;
1227 struct ext4_xattr_ibody_find *is = NULL;
1228 struct ext4_xattr_block_find *bs = NULL;
1229 char *buffer = NULL, *b_entry_name = NULL;
1230 size_t min_offs, free;
1231 int total_ino, total_blk;
1232 void *base, *start, *end;
1233 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1234 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1235
1236 down_write(&EXT4_I(inode)->xattr_sem);
1237 retry:
1238 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1239 up_write(&EXT4_I(inode)->xattr_sem);
1240 return 0;
1241 }
1242
1243 header = IHDR(inode, raw_inode);
1244 entry = IFIRST(header);
1245
1246 /*
1247 * Check if enough free space is available in the inode to shift the
1248 * entries ahead by new_extra_isize.
1249 */
1250
1251 base = start = entry;
1252 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1253 min_offs = end - base;
1254 last = entry;
1255 total_ino = sizeof(struct ext4_xattr_ibody_header);
1256
1257 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1258 if (free >= new_extra_isize) {
1259 entry = IFIRST(header);
1260 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1261 - new_extra_isize, (void *)raw_inode +
1262 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1263 (void *)header, total_ino,
1264 inode->i_sb->s_blocksize);
1265 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1266 error = 0;
1267 goto cleanup;
1268 }
1269
1270 /*
1271 * Enough free space isn't available in the inode, check if
1272 * EA block can hold new_extra_isize bytes.
1273 */
1274 if (EXT4_I(inode)->i_file_acl) {
1275 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1276 error = -EIO;
1277 if (!bh)
1278 goto cleanup;
1279 if (ext4_xattr_check_block(inode, bh)) {
1280 EXT4_ERROR_INODE(inode, "bad block %llu",
1281 EXT4_I(inode)->i_file_acl);
1282 error = -EIO;
1283 goto cleanup;
1284 }
1285 base = BHDR(bh);
1286 first = BFIRST(bh);
1287 end = bh->b_data + bh->b_size;
1288 min_offs = end - base;
1289 free = ext4_xattr_free_space(first, &min_offs, base,
1290 &total_blk);
1291 if (free < new_extra_isize) {
1292 if (!tried_min_extra_isize && s_min_extra_isize) {
1293 tried_min_extra_isize++;
1294 new_extra_isize = s_min_extra_isize;
1295 brelse(bh);
1296 goto retry;
1297 }
1298 error = -1;
1299 goto cleanup;
1300 }
1301 } else {
1302 free = inode->i_sb->s_blocksize;
1303 }
1304
1305 while (new_extra_isize > 0) {
1306 size_t offs, size, entry_size;
1307 struct ext4_xattr_entry *small_entry = NULL;
1308 struct ext4_xattr_info i = {
1309 .value = NULL,
1310 .value_len = 0,
1311 };
1312 unsigned int total_size; /* EA entry size + value size */
1313 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1314 unsigned int min_total_size = ~0U;
1315
1316 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1317 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1318 if (!is || !bs) {
1319 error = -ENOMEM;
1320 goto cleanup;
1321 }
1322
1323 is->s.not_found = -ENODATA;
1324 bs->s.not_found = -ENODATA;
1325 is->iloc.bh = NULL;
1326 bs->bh = NULL;
1327
1328 last = IFIRST(header);
1329 /* Find the entry best suited to be pushed into EA block */
1330 entry = NULL;
1331 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1332 total_size =
1333 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1334 EXT4_XATTR_LEN(last->e_name_len);
1335 if (total_size <= free && total_size < min_total_size) {
1336 if (total_size < new_extra_isize) {
1337 small_entry = last;
1338 } else {
1339 entry = last;
1340 min_total_size = total_size;
1341 }
1342 }
1343 }
1344
1345 if (entry == NULL) {
1346 if (small_entry) {
1347 entry = small_entry;
1348 } else {
1349 if (!tried_min_extra_isize &&
1350 s_min_extra_isize) {
1351 tried_min_extra_isize++;
1352 new_extra_isize = s_min_extra_isize;
1353 kfree(is); is = NULL;
1354 kfree(bs); bs = NULL;
1355 brelse(bh);
1356 goto retry;
1357 }
1358 error = -1;
1359 goto cleanup;
1360 }
1361 }
1362 offs = le16_to_cpu(entry->e_value_offs);
1363 size = le32_to_cpu(entry->e_value_size);
1364 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1365 i.name_index = entry->e_name_index,
1366 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1367 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1368 if (!buffer || !b_entry_name) {
1369 error = -ENOMEM;
1370 goto cleanup;
1371 }
1372 /* Save the entry name and the entry value */
1373 memcpy(buffer, (void *)IFIRST(header) + offs,
1374 EXT4_XATTR_SIZE(size));
1375 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1376 b_entry_name[entry->e_name_len] = '\0';
1377 i.name = b_entry_name;
1378
1379 error = ext4_get_inode_loc(inode, &is->iloc);
1380 if (error)
1381 goto cleanup;
1382
1383 error = ext4_xattr_ibody_find(inode, &i, is);
1384 if (error)
1385 goto cleanup;
1386
1387 /* Remove the chosen entry from the inode */
1388 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1389 if (error)
1390 goto cleanup;
1391
1392 entry = IFIRST(header);
1393 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1394 shift_bytes = new_extra_isize;
1395 else
1396 shift_bytes = entry_size + size;
1397 /* Adjust the offsets and shift the remaining entries ahead */
1398 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1399 shift_bytes, (void *)raw_inode +
1400 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1401 (void *)header, total_ino - entry_size,
1402 inode->i_sb->s_blocksize);
1403
1404 extra_isize += shift_bytes;
1405 new_extra_isize -= shift_bytes;
1406 EXT4_I(inode)->i_extra_isize = extra_isize;
1407
1408 i.name = b_entry_name;
1409 i.value = buffer;
1410 i.value_len = size;
1411 error = ext4_xattr_block_find(inode, &i, bs);
1412 if (error)
1413 goto cleanup;
1414
1415 /* Add entry which was removed from the inode into the block */
1416 error = ext4_xattr_block_set(handle, inode, &i, bs);
1417 if (error)
1418 goto cleanup;
1419 kfree(b_entry_name);
1420 kfree(buffer);
1421 b_entry_name = NULL;
1422 buffer = NULL;
1423 brelse(is->iloc.bh);
1424 kfree(is);
1425 kfree(bs);
1426 }
1427 brelse(bh);
1428 up_write(&EXT4_I(inode)->xattr_sem);
1429 return 0;
1430
1431 cleanup:
1432 kfree(b_entry_name);
1433 kfree(buffer);
1434 if (is)
1435 brelse(is->iloc.bh);
1436 kfree(is);
1437 kfree(bs);
1438 brelse(bh);
1439 up_write(&EXT4_I(inode)->xattr_sem);
1440 return error;
1441 }
1442
1443
1444
1445 /*
1446 * ext4_xattr_delete_inode()
1447 *
1448 * Free extended attribute resources associated with this inode. This
1449 * is called immediately before an inode is freed. We have exclusive
1450 * access to the inode.
1451 */
1452 void
1453 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1454 {
1455 struct buffer_head *bh = NULL;
1456
1457 if (!EXT4_I(inode)->i_file_acl)
1458 goto cleanup;
1459 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1460 if (!bh) {
1461 EXT4_ERROR_INODE(inode, "block %llu read error",
1462 EXT4_I(inode)->i_file_acl);
1463 goto cleanup;
1464 }
1465 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1466 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1467 EXT4_ERROR_INODE(inode, "bad block %llu",
1468 EXT4_I(inode)->i_file_acl);
1469 goto cleanup;
1470 }
1471 ext4_xattr_release_block(handle, inode, bh);
1472 EXT4_I(inode)->i_file_acl = 0;
1473
1474 cleanup:
1475 brelse(bh);
1476 }
1477
1478 /*
1479 * ext4_xattr_put_super()
1480 *
1481 * This is called when a file system is unmounted.
1482 */
1483 void
1484 ext4_xattr_put_super(struct super_block *sb)
1485 {
1486 mb_cache_shrink(sb->s_bdev);
1487 }
1488
1489 /*
1490 * ext4_xattr_cache_insert()
1491 *
1492 * Create a new entry in the extended attribute cache, and insert
1493 * it unless such an entry is already in the cache.
1494 *
1495 * Returns 0, or a negative error number on failure.
1496 */
1497 static void
1498 ext4_xattr_cache_insert(struct buffer_head *bh)
1499 {
1500 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1501 struct mb_cache_entry *ce;
1502 int error;
1503
1504 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1505 if (!ce) {
1506 ea_bdebug(bh, "out of memory");
1507 return;
1508 }
1509 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1510 if (error) {
1511 mb_cache_entry_free(ce);
1512 if (error == -EBUSY) {
1513 ea_bdebug(bh, "already in cache");
1514 error = 0;
1515 }
1516 } else {
1517 ea_bdebug(bh, "inserting [%x]", (int)hash);
1518 mb_cache_entry_release(ce);
1519 }
1520 }
1521
1522 /*
1523 * ext4_xattr_cmp()
1524 *
1525 * Compare two extended attribute blocks for equality.
1526 *
1527 * Returns 0 if the blocks are equal, 1 if they differ, and
1528 * a negative error number on errors.
1529 */
1530 static int
1531 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1532 struct ext4_xattr_header *header2)
1533 {
1534 struct ext4_xattr_entry *entry1, *entry2;
1535
1536 entry1 = ENTRY(header1+1);
1537 entry2 = ENTRY(header2+1);
1538 while (!IS_LAST_ENTRY(entry1)) {
1539 if (IS_LAST_ENTRY(entry2))
1540 return 1;
1541 if (entry1->e_hash != entry2->e_hash ||
1542 entry1->e_name_index != entry2->e_name_index ||
1543 entry1->e_name_len != entry2->e_name_len ||
1544 entry1->e_value_size != entry2->e_value_size ||
1545 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1546 return 1;
1547 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1548 return -EIO;
1549 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1550 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1551 le32_to_cpu(entry1->e_value_size)))
1552 return 1;
1553
1554 entry1 = EXT4_XATTR_NEXT(entry1);
1555 entry2 = EXT4_XATTR_NEXT(entry2);
1556 }
1557 if (!IS_LAST_ENTRY(entry2))
1558 return 1;
1559 return 0;
1560 }
1561
1562 /*
1563 * ext4_xattr_cache_find()
1564 *
1565 * Find an identical extended attribute block.
1566 *
1567 * Returns a pointer to the block found, or NULL if such a block was
1568 * not found or an error occurred.
1569 */
1570 static struct buffer_head *
1571 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1572 struct mb_cache_entry **pce)
1573 {
1574 __u32 hash = le32_to_cpu(header->h_hash);
1575 struct mb_cache_entry *ce;
1576
1577 if (!header->h_hash)
1578 return NULL; /* never share */
1579 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1580 again:
1581 ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1582 hash);
1583 while (ce) {
1584 struct buffer_head *bh;
1585
1586 if (IS_ERR(ce)) {
1587 if (PTR_ERR(ce) == -EAGAIN)
1588 goto again;
1589 break;
1590 }
1591 bh = sb_bread(inode->i_sb, ce->e_block);
1592 if (!bh) {
1593 EXT4_ERROR_INODE(inode, "block %lu read error",
1594 (unsigned long) ce->e_block);
1595 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1596 EXT4_XATTR_REFCOUNT_MAX) {
1597 ea_idebug(inode, "block %lu refcount %d>=%d",
1598 (unsigned long) ce->e_block,
1599 le32_to_cpu(BHDR(bh)->h_refcount),
1600 EXT4_XATTR_REFCOUNT_MAX);
1601 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1602 *pce = ce;
1603 return bh;
1604 }
1605 brelse(bh);
1606 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1607 }
1608 return NULL;
1609 }
1610
1611 #define NAME_HASH_SHIFT 5
1612 #define VALUE_HASH_SHIFT 16
1613
1614 /*
1615 * ext4_xattr_hash_entry()
1616 *
1617 * Compute the hash of an extended attribute.
1618 */
1619 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1620 struct ext4_xattr_entry *entry)
1621 {
1622 __u32 hash = 0;
1623 char *name = entry->e_name;
1624 int n;
1625
1626 for (n = 0; n < entry->e_name_len; n++) {
1627 hash = (hash << NAME_HASH_SHIFT) ^
1628 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1629 *name++;
1630 }
1631
1632 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1633 __le32 *value = (__le32 *)((char *)header +
1634 le16_to_cpu(entry->e_value_offs));
1635 for (n = (le32_to_cpu(entry->e_value_size) +
1636 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1637 hash = (hash << VALUE_HASH_SHIFT) ^
1638 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1639 le32_to_cpu(*value++);
1640 }
1641 }
1642 entry->e_hash = cpu_to_le32(hash);
1643 }
1644
1645 #undef NAME_HASH_SHIFT
1646 #undef VALUE_HASH_SHIFT
1647
1648 #define BLOCK_HASH_SHIFT 16
1649
1650 /*
1651 * ext4_xattr_rehash()
1652 *
1653 * Re-compute the extended attribute hash value after an entry has changed.
1654 */
1655 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1656 struct ext4_xattr_entry *entry)
1657 {
1658 struct ext4_xattr_entry *here;
1659 __u32 hash = 0;
1660
1661 ext4_xattr_hash_entry(header, entry);
1662 here = ENTRY(header+1);
1663 while (!IS_LAST_ENTRY(here)) {
1664 if (!here->e_hash) {
1665 /* Block is not shared if an entry's hash value == 0 */
1666 hash = 0;
1667 break;
1668 }
1669 hash = (hash << BLOCK_HASH_SHIFT) ^
1670 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1671 le32_to_cpu(here->e_hash);
1672 here = EXT4_XATTR_NEXT(here);
1673 }
1674 header->h_hash = cpu_to_le32(hash);
1675 }
1676
1677 #undef BLOCK_HASH_SHIFT
1678
1679 int __init
1680 ext4_init_xattr(void)
1681 {
1682 ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1683 if (!ext4_xattr_cache)
1684 return -ENOMEM;
1685 return 0;
1686 }
1687
1688 void
1689 ext4_exit_xattr(void)
1690 {
1691 if (ext4_xattr_cache)
1692 mb_cache_destroy(ext4_xattr_cache);
1693 ext4_xattr_cache = NULL;
1694 }
Linus' kernel tree