?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.12.28
[linux/fpc-iii.git] / fs / ext4 / xattr.c
blob298e9c8da3648cf04d3f61c3f4de386da7b10b5e
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 it;
521 * 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 unlock_buffer(bh);
542 ext4_free_blocks(handle, inode, bh, 0, 1,
543 EXT4_FREE_BLOCKS_METADATA |
544 EXT4_FREE_BLOCKS_FORGET);
545 } else {
546 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
547 if (ce)
548 mb_cache_entry_release(ce);
549 /*
550 * Beware of this ugliness: Releasing of xattr block references
551 * from different inodes can race and so we have to protect
552 * from a race where someone else frees the block (and releases
553 * its journal_head) before we are done dirtying the buffer. In
554 * nojournal mode this race is harmless and we actually cannot
555 * call ext4_handle_dirty_xattr_block() with locked buffer as
556 * that function can call sync_dirty_buffer() so for that case
557 * we handle the dirtying after unlocking the buffer.
558 */
559 if (ext4_handle_valid(handle))
560 error = ext4_handle_dirty_xattr_block(handle, inode,
561 bh);
562 unlock_buffer(bh);
563 if (!ext4_handle_valid(handle))
564 error = ext4_handle_dirty_xattr_block(handle, inode,
565 bh);
566 if (IS_SYNC(inode))
567 ext4_handle_sync(handle);
568 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
569 ea_bdebug(bh, "refcount now=%d; releasing",
570 le32_to_cpu(BHDR(bh)->h_refcount));
571 }
572 out:
573 ext4_std_error(inode->i_sb, error);
574 return;
575 }
576
577 /*
578 * Find the available free space for EAs. This also returns the total number of
579 * bytes used by EA entries.
580 */
581 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
582 size_t *min_offs, void *base, int *total)
583 {
584 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
585 *total += EXT4_XATTR_LEN(last->e_name_len);
586 if (!last->e_value_block && last->e_value_size) {
587 size_t offs = le16_to_cpu(last->e_value_offs);
588 if (offs < *min_offs)
589 *min_offs = offs;
590 }
591 }
592 return (*min_offs - ((void *)last - base) - sizeof(__u32));
593 }
594
595 static int
596 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
597 {
598 struct ext4_xattr_entry *last;
599 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
600
601 /* Compute min_offs and last. */
602 last = s->first;
603 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
604 if (!last->e_value_block && last->e_value_size) {
605 size_t offs = le16_to_cpu(last->e_value_offs);
606 if (offs < min_offs)
607 min_offs = offs;
608 }
609 }
610 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
611 if (!s->not_found) {
612 if (!s->here->e_value_block && s->here->e_value_size) {
613 size_t size = le32_to_cpu(s->here->e_value_size);
614 free += EXT4_XATTR_SIZE(size);
615 }
616 free += EXT4_XATTR_LEN(name_len);
617 }
618 if (i->value) {
619 if (free < EXT4_XATTR_SIZE(i->value_len) ||
620 free < EXT4_XATTR_LEN(name_len) +
621 EXT4_XATTR_SIZE(i->value_len))
622 return -ENOSPC;
623 }
624
625 if (i->value && s->not_found) {
626 /* Insert the new name. */
627 size_t size = EXT4_XATTR_LEN(name_len);
628 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
629 memmove((void *)s->here + size, s->here, rest);
630 memset(s->here, 0, size);
631 s->here->e_name_index = i->name_index;
632 s->here->e_name_len = name_len;
633 memcpy(s->here->e_name, i->name, name_len);
634 } else {
635 if (!s->here->e_value_block && s->here->e_value_size) {
636 void *first_val = s->base + min_offs;
637 size_t offs = le16_to_cpu(s->here->e_value_offs);
638 void *val = s->base + offs;
639 size_t size = EXT4_XATTR_SIZE(
640 le32_to_cpu(s->here->e_value_size));
641
642 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
643 /* The old and the new value have the same
644 size. Just replace. */
645 s->here->e_value_size =
646 cpu_to_le32(i->value_len);
647 if (i->value == EXT4_ZERO_XATTR_VALUE) {
648 memset(val, 0, size);
649 } else {
650 /* Clear pad bytes first. */
651 memset(val + size - EXT4_XATTR_PAD, 0,
652 EXT4_XATTR_PAD);
653 memcpy(val, i->value, i->value_len);
654 }
655 return 0;
656 }
657
658 /* Remove the old value. */
659 memmove(first_val + size, first_val, val - first_val);
660 memset(first_val, 0, size);
661 s->here->e_value_size = 0;
662 s->here->e_value_offs = 0;
663 min_offs += size;
664
665 /* Adjust all value offsets. */
666 last = s->first;
667 while (!IS_LAST_ENTRY(last)) {
668 size_t o = le16_to_cpu(last->e_value_offs);
669 if (!last->e_value_block &&
670 last->e_value_size && o < offs)
671 last->e_value_offs =
672 cpu_to_le16(o + size);
673 last = EXT4_XATTR_NEXT(last);
674 }
675 }
676 if (!i->value) {
677 /* Remove the old name. */
678 size_t size = EXT4_XATTR_LEN(name_len);
679 last = ENTRY((void *)last - size);
680 memmove(s->here, (void *)s->here + size,
681 (void *)last - (void *)s->here + sizeof(__u32));
682 memset(last, 0, size);
683 }
684 }
685
686 if (i->value) {
687 /* Insert the new value. */
688 s->here->e_value_size = cpu_to_le32(i->value_len);
689 if (i->value_len) {
690 size_t size = EXT4_XATTR_SIZE(i->value_len);
691 void *val = s->base + min_offs - size;
692 s->here->e_value_offs = cpu_to_le16(min_offs - size);
693 if (i->value == EXT4_ZERO_XATTR_VALUE) {
694 memset(val, 0, size);
695 } else {
696 /* Clear the pad bytes first. */
697 memset(val + size - EXT4_XATTR_PAD, 0,
698 EXT4_XATTR_PAD);
699 memcpy(val, i->value, i->value_len);
700 }
701 }
702 }
703 return 0;
704 }
705
706 struct ext4_xattr_block_find {
707 struct ext4_xattr_search s;
708 struct buffer_head *bh;
709 };
710
711 static int
712 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
713 struct ext4_xattr_block_find *bs)
714 {
715 struct super_block *sb = inode->i_sb;
716 int error;
717
718 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
719 i->name_index, i->name, i->value, (long)i->value_len);
720
721 if (EXT4_I(inode)->i_file_acl) {
722 /* The inode already has an extended attribute block. */
723 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
724 error = -EIO;
725 if (!bs->bh)
726 goto cleanup;
727 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
728 atomic_read(&(bs->bh->b_count)),
729 le32_to_cpu(BHDR(bs->bh)->h_refcount));
730 if (ext4_xattr_check_block(inode, bs->bh)) {
731 EXT4_ERROR_INODE(inode, "bad block %llu",
732 EXT4_I(inode)->i_file_acl);
733 error = -EIO;
734 goto cleanup;
735 }
736 /* Find the named attribute. */
737 bs->s.base = BHDR(bs->bh);
738 bs->s.first = BFIRST(bs->bh);
739 bs->s.end = bs->bh->b_data + bs->bh->b_size;
740 bs->s.here = bs->s.first;
741 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
742 i->name, bs->bh->b_size, 1);
743 if (error && error != -ENODATA)
744 goto cleanup;
745 bs->s.not_found = error;
746 }
747 error = 0;
748
749 cleanup:
750 return error;
751 }
752
753 static int
754 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
755 struct ext4_xattr_info *i,
756 struct ext4_xattr_block_find *bs)
757 {
758 struct super_block *sb = inode->i_sb;
759 struct buffer_head *new_bh = NULL;
760 struct ext4_xattr_search *s = &bs->s;
761 struct mb_cache_entry *ce = NULL;
762 int error = 0;
763
764 #define header(x) ((struct ext4_xattr_header *)(x))
765
766 if (i->value && i->value_len > sb->s_blocksize)
767 return -ENOSPC;
768 if (s->base) {
769 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
770 bs->bh->b_blocknr);
771 error = ext4_journal_get_write_access(handle, bs->bh);
772 if (error)
773 goto cleanup;
774 lock_buffer(bs->bh);
775
776 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
777 if (ce) {
778 mb_cache_entry_free(ce);
779 ce = NULL;
780 }
781 ea_bdebug(bs->bh, "modifying in-place");
782 error = ext4_xattr_set_entry(i, s);
783 if (!error) {
784 if (!IS_LAST_ENTRY(s->first))
785 ext4_xattr_rehash(header(s->base),
786 s->here);
787 ext4_xattr_cache_insert(bs->bh);
788 }
789 unlock_buffer(bs->bh);
790 if (error == -EIO)
791 goto bad_block;
792 if (!error)
793 error = ext4_handle_dirty_xattr_block(handle,
794 inode,
795 bs->bh);
796 if (error)
797 goto cleanup;
798 goto inserted;
799 } else {
800 int offset = (char *)s->here - bs->bh->b_data;
801
802 unlock_buffer(bs->bh);
803 if (ce) {
804 mb_cache_entry_release(ce);
805 ce = NULL;
806 }
807 ea_bdebug(bs->bh, "cloning");
808 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
809 error = -ENOMEM;
810 if (s->base == NULL)
811 goto cleanup;
812 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
813 s->first = ENTRY(header(s->base)+1);
814 header(s->base)->h_refcount = cpu_to_le32(1);
815 s->here = ENTRY(s->base + offset);
816 s->end = s->base + bs->bh->b_size;
817 }
818 } else {
819 /* Allocate a buffer where we construct the new block. */
820 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
821 /* assert(header == s->base) */
822 error = -ENOMEM;
823 if (s->base == NULL)
824 goto cleanup;
825 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
826 header(s->base)->h_blocks = cpu_to_le32(1);
827 header(s->base)->h_refcount = cpu_to_le32(1);
828 s->first = ENTRY(header(s->base)+1);
829 s->here = ENTRY(header(s->base)+1);
830 s->end = s->base + sb->s_blocksize;
831 }
832
833 error = ext4_xattr_set_entry(i, s);
834 if (error == -EIO)
835 goto bad_block;
836 if (error)
837 goto cleanup;
838 if (!IS_LAST_ENTRY(s->first))
839 ext4_xattr_rehash(header(s->base), s->here);
840
841 inserted:
842 if (!IS_LAST_ENTRY(s->first)) {
843 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
844 if (new_bh) {
845 /* We found an identical block in the cache. */
846 if (new_bh == bs->bh)
847 ea_bdebug(new_bh, "keeping");
848 else {
849 /* The old block is released after updating
850 the inode. */
851 error = dquot_alloc_block(inode,
852 EXT4_C2B(EXT4_SB(sb), 1));
853 if (error)
854 goto cleanup;
855 error = ext4_journal_get_write_access(handle,
856 new_bh);
857 if (error)
858 goto cleanup_dquot;
859 lock_buffer(new_bh);
860 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
861 ea_bdebug(new_bh, "reusing; refcount now=%d",
862 le32_to_cpu(BHDR(new_bh)->h_refcount));
863 unlock_buffer(new_bh);
864 error = ext4_handle_dirty_xattr_block(handle,
865 inode,
866 new_bh);
867 if (error)
868 goto cleanup_dquot;
869 }
870 mb_cache_entry_release(ce);
871 ce = NULL;
872 } else if (bs->bh && s->base == bs->bh->b_data) {
873 /* We were modifying this block in-place. */
874 ea_bdebug(bs->bh, "keeping this block");
875 new_bh = bs->bh;
876 get_bh(new_bh);
877 } else {
878 /* We need to allocate a new block */
879 ext4_fsblk_t goal, block;
880
881 goal = ext4_group_first_block_no(sb,
882 EXT4_I(inode)->i_block_group);
883
884 /* non-extent files can't have physical blocks past 2^32 */
885 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
886 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
887
888 /*
889 * take i_data_sem because we will test
890 * i_delalloc_reserved_flag in ext4_mb_new_blocks
891 */
892 down_read((&EXT4_I(inode)->i_data_sem));
893 block = ext4_new_meta_blocks(handle, inode, goal, 0,
894 NULL, &error);
895 up_read((&EXT4_I(inode)->i_data_sem));
896 if (error)
897 goto cleanup;
898
899 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
900 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
901
902 ea_idebug(inode, "creating block %llu",
903 (unsigned long long)block);
904
905 new_bh = sb_getblk(sb, block);
906 if (unlikely(!new_bh)) {
907 error = -ENOMEM;
908 getblk_failed:
909 ext4_free_blocks(handle, inode, NULL, block, 1,
910 EXT4_FREE_BLOCKS_METADATA);
911 goto cleanup;
912 }
913 lock_buffer(new_bh);
914 error = ext4_journal_get_create_access(handle, new_bh);
915 if (error) {
916 unlock_buffer(new_bh);
917 error = -EIO;
918 goto getblk_failed;
919 }
920 memcpy(new_bh->b_data, s->base, new_bh->b_size);
921 set_buffer_uptodate(new_bh);
922 unlock_buffer(new_bh);
923 ext4_xattr_cache_insert(new_bh);
924 error = ext4_handle_dirty_xattr_block(handle,
925 inode, new_bh);
926 if (error)
927 goto cleanup;
928 }
929 }
930
931 /* Update the inode. */
932 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
933
934 /* Drop the previous xattr block. */
935 if (bs->bh && bs->bh != new_bh)
936 ext4_xattr_release_block(handle, inode, bs->bh);
937 error = 0;
938
939 cleanup:
940 if (ce)
941 mb_cache_entry_release(ce);
942 brelse(new_bh);
943 if (!(bs->bh && s->base == bs->bh->b_data))
944 kfree(s->base);
945
946 return error;
947
948 cleanup_dquot:
949 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
950 goto cleanup;
951
952 bad_block:
953 EXT4_ERROR_INODE(inode, "bad block %llu",
954 EXT4_I(inode)->i_file_acl);
955 goto cleanup;
956
957 #undef header
958 }
959
960 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
961 struct ext4_xattr_ibody_find *is)
962 {
963 struct ext4_xattr_ibody_header *header;
964 struct ext4_inode *raw_inode;
965 int error;
966
967 if (EXT4_I(inode)->i_extra_isize == 0)
968 return 0;
969 raw_inode = ext4_raw_inode(&is->iloc);
970 header = IHDR(inode, raw_inode);
971 is->s.base = is->s.first = IFIRST(header);
972 is->s.here = is->s.first;
973 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
974 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
975 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
976 if (error)
977 return error;
978 /* Find the named attribute. */
979 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
980 i->name, is->s.end -
981 (void *)is->s.base, 0);
982 if (error && error != -ENODATA)
983 return error;
984 is->s.not_found = error;
985 }
986 return 0;
987 }
988
989 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
990 struct ext4_xattr_info *i,
991 struct ext4_xattr_ibody_find *is)
992 {
993 struct ext4_xattr_ibody_header *header;
994 struct ext4_xattr_search *s = &is->s;
995 int error;
996
997 if (EXT4_I(inode)->i_extra_isize == 0)
998 return -ENOSPC;
999 error = ext4_xattr_set_entry(i, s);
1000 if (error) {
1001 if (error == -ENOSPC &&
1002 ext4_has_inline_data(inode)) {
1003 error = ext4_try_to_evict_inline_data(handle, inode,
1004 EXT4_XATTR_LEN(strlen(i->name) +
1005 EXT4_XATTR_SIZE(i->value_len)));
1006 if (error)
1007 return error;
1008 error = ext4_xattr_ibody_find(inode, i, is);
1009 if (error)
1010 return error;
1011 error = ext4_xattr_set_entry(i, s);
1012 }
1013 if (error)
1014 return error;
1015 }
1016 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1017 if (!IS_LAST_ENTRY(s->first)) {
1018 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1019 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1020 } else {
1021 header->h_magic = cpu_to_le32(0);
1022 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1023 }
1024 return 0;
1025 }
1026
1027 static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
1028 struct ext4_xattr_info *i,
1029 struct ext4_xattr_ibody_find *is)
1030 {
1031 struct ext4_xattr_ibody_header *header;
1032 struct ext4_xattr_search *s = &is->s;
1033 int error;
1034
1035 if (EXT4_I(inode)->i_extra_isize == 0)
1036 return -ENOSPC;
1037 error = ext4_xattr_set_entry(i, s);
1038 if (error)
1039 return error;
1040 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1041 if (!IS_LAST_ENTRY(s->first)) {
1042 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1043 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1044 } else {
1045 header->h_magic = cpu_to_le32(0);
1046 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1047 }
1048 return 0;
1049 }
1050
1051 /*
1052 * ext4_xattr_set_handle()
1053 *
1054 * Create, replace or remove an extended attribute for this inode. Value
1055 * is NULL to remove an existing extended attribute, and non-NULL to
1056 * either replace an existing extended attribute, or create a new extended
1057 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1058 * specify that an extended attribute must exist and must not exist
1059 * previous to the call, respectively.
1060 *
1061 * Returns 0, or a negative error number on failure.
1062 */
1063 int
1064 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1065 const char *name, const void *value, size_t value_len,
1066 int flags)
1067 {
1068 struct ext4_xattr_info i = {
1069 .name_index = name_index,
1070 .name = name,
1071 .value = value,
1072 .value_len = value_len,
1073
1074 };
1075 struct ext4_xattr_ibody_find is = {
1076 .s = { .not_found = -ENODATA, },
1077 };
1078 struct ext4_xattr_block_find bs = {
1079 .s = { .not_found = -ENODATA, },
1080 };
1081 unsigned long no_expand;
1082 int error;
1083
1084 if (!name)
1085 return -EINVAL;
1086 if (strlen(name) > 255)
1087 return -ERANGE;
1088 down_write(&EXT4_I(inode)->xattr_sem);
1089 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
1090 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
1091
1092 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1093 if (error)
1094 goto cleanup;
1095
1096 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1097 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1098 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1099 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1100 }
1101
1102 error = ext4_xattr_ibody_find(inode, &i, &is);
1103 if (error)
1104 goto cleanup;
1105 if (is.s.not_found)
1106 error = ext4_xattr_block_find(inode, &i, &bs);
1107 if (error)
1108 goto cleanup;
1109 if (is.s.not_found && bs.s.not_found) {
1110 error = -ENODATA;
1111 if (flags & XATTR_REPLACE)
1112 goto cleanup;
1113 error = 0;
1114 if (!value)
1115 goto cleanup;
1116 } else {
1117 error = -EEXIST;
1118 if (flags & XATTR_CREATE)
1119 goto cleanup;
1120 }
1121 if (!value) {
1122 if (!is.s.not_found)
1123 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1124 else if (!bs.s.not_found)
1125 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1126 } else {
1127 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1128 if (!error && !bs.s.not_found) {
1129 i.value = NULL;
1130 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1131 } else if (error == -ENOSPC) {
1132 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1133 error = ext4_xattr_block_find(inode, &i, &bs);
1134 if (error)
1135 goto cleanup;
1136 }
1137 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1138 if (error)
1139 goto cleanup;
1140 if (!is.s.not_found) {
1141 i.value = NULL;
1142 error = ext4_xattr_ibody_set(handle, inode, &i,
1143 &is);
1144 }
1145 }
1146 }
1147 if (!error) {
1148 ext4_xattr_update_super_block(handle, inode->i_sb);
1149 inode->i_ctime = ext4_current_time(inode);
1150 if (!value)
1151 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1152 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1153 /*
1154 * The bh is consumed by ext4_mark_iloc_dirty, even with
1155 * error != 0.
1156 */
1157 is.iloc.bh = NULL;
1158 if (IS_SYNC(inode))
1159 ext4_handle_sync(handle);
1160 }
1161
1162 cleanup:
1163 brelse(is.iloc.bh);
1164 brelse(bs.bh);
1165 if (no_expand == 0)
1166 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1167 up_write(&EXT4_I(inode)->xattr_sem);
1168 return error;
1169 }
1170
1171 /*
1172 * ext4_xattr_set()
1173 *
1174 * Like ext4_xattr_set_handle, but start from an inode. This extended
1175 * attribute modification is a filesystem transaction by itself.
1176 *
1177 * Returns 0, or a negative error number on failure.
1178 */
1179 int
1180 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1181 const void *value, size_t value_len, int flags)
1182 {
1183 handle_t *handle;
1184 int error, retries = 0;
1185 int credits = ext4_jbd2_credits_xattr(inode);
1186
1187 retry:
1188 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1189 if (IS_ERR(handle)) {
1190 error = PTR_ERR(handle);
1191 } else {
1192 int error2;
1193
1194 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1195 value, value_len, flags);
1196 error2 = ext4_journal_stop(handle);
1197 if (error == -ENOSPC &&
1198 ext4_should_retry_alloc(inode->i_sb, &retries))
1199 goto retry;
1200 if (error == 0)
1201 error = error2;
1202 }
1203
1204 return error;
1205 }
1206
1207 /*
1208 * Shift the EA entries in the inode to create space for the increased
1209 * i_extra_isize.
1210 */
1211 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1212 int value_offs_shift, void *to,
1213 void *from, size_t n, int blocksize)
1214 {
1215 struct ext4_xattr_entry *last = entry;
1216 int new_offs;
1217
1218 /* Adjust the value offsets of the entries */
1219 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1220 if (!last->e_value_block && last->e_value_size) {
1221 new_offs = le16_to_cpu(last->e_value_offs) +
1222 value_offs_shift;
1223 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1224 > blocksize);
1225 last->e_value_offs = cpu_to_le16(new_offs);
1226 }
1227 }
1228 /* Shift the entries by n bytes */
1229 memmove(to, from, n);
1230 }
1231
1232 /*
1233 * Expand an inode by new_extra_isize bytes when EAs are present.
1234 * Returns 0 on success or negative error number on failure.
1235 */
1236 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1237 struct ext4_inode *raw_inode, handle_t *handle)
1238 {
1239 struct ext4_xattr_ibody_header *header;
1240 struct ext4_xattr_entry *entry, *last, *first;
1241 struct buffer_head *bh = NULL;
1242 struct ext4_xattr_ibody_find *is = NULL;
1243 struct ext4_xattr_block_find *bs = NULL;
1244 char *buffer = NULL, *b_entry_name = NULL;
1245 size_t min_offs, free;
1246 int total_ino, total_blk;
1247 void *base, *start, *end;
1248 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1249 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1250
1251 down_write(&EXT4_I(inode)->xattr_sem);
1252 retry:
1253 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1254 up_write(&EXT4_I(inode)->xattr_sem);
1255 return 0;
1256 }
1257
1258 header = IHDR(inode, raw_inode);
1259 entry = IFIRST(header);
1260
1261 /*
1262 * Check if enough free space is available in the inode to shift the
1263 * entries ahead by new_extra_isize.
1264 */
1265
1266 base = start = entry;
1267 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1268 min_offs = end - base;
1269 last = entry;
1270 total_ino = sizeof(struct ext4_xattr_ibody_header);
1271
1272 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1273 if (free >= new_extra_isize) {
1274 entry = IFIRST(header);
1275 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1276 - new_extra_isize, (void *)raw_inode +
1277 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1278 (void *)header, total_ino,
1279 inode->i_sb->s_blocksize);
1280 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1281 error = 0;
1282 goto cleanup;
1283 }
1284
1285 /*
1286 * Enough free space isn't available in the inode, check if
1287 * EA block can hold new_extra_isize bytes.
1288 */
1289 if (EXT4_I(inode)->i_file_acl) {
1290 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1291 error = -EIO;
1292 if (!bh)
1293 goto cleanup;
1294 if (ext4_xattr_check_block(inode, bh)) {
1295 EXT4_ERROR_INODE(inode, "bad block %llu",
1296 EXT4_I(inode)->i_file_acl);
1297 error = -EIO;
1298 goto cleanup;
1299 }
1300 base = BHDR(bh);
1301 first = BFIRST(bh);
1302 end = bh->b_data + bh->b_size;
1303 min_offs = end - base;
1304 free = ext4_xattr_free_space(first, &min_offs, base,
1305 &total_blk);
1306 if (free < new_extra_isize) {
1307 if (!tried_min_extra_isize && s_min_extra_isize) {
1308 tried_min_extra_isize++;
1309 new_extra_isize = s_min_extra_isize;
1310 brelse(bh);
1311 goto retry;
1312 }
1313 error = -1;
1314 goto cleanup;
1315 }
1316 } else {
1317 free = inode->i_sb->s_blocksize;
1318 }
1319
1320 while (new_extra_isize > 0) {
1321 size_t offs, size, entry_size;
1322 struct ext4_xattr_entry *small_entry = NULL;
1323 struct ext4_xattr_info i = {
1324 .value = NULL,
1325 .value_len = 0,
1326 };
1327 unsigned int total_size; /* EA entry size + value size */
1328 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1329 unsigned int min_total_size = ~0U;
1330
1331 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1332 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1333 if (!is || !bs) {
1334 error = -ENOMEM;
1335 goto cleanup;
1336 }
1337
1338 is->s.not_found = -ENODATA;
1339 bs->s.not_found = -ENODATA;
1340 is->iloc.bh = NULL;
1341 bs->bh = NULL;
1342
1343 last = IFIRST(header);
1344 /* Find the entry best suited to be pushed into EA block */
1345 entry = NULL;
1346 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1347 total_size =
1348 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1349 EXT4_XATTR_LEN(last->e_name_len);
1350 if (total_size <= free && total_size < min_total_size) {
1351 if (total_size < new_extra_isize) {
1352 small_entry = last;
1353 } else {
1354 entry = last;
1355 min_total_size = total_size;
1356 }
1357 }
1358 }
1359
1360 if (entry == NULL) {
1361 if (small_entry) {
1362 entry = small_entry;
1363 } else {
1364 if (!tried_min_extra_isize &&
1365 s_min_extra_isize) {
1366 tried_min_extra_isize++;
1367 new_extra_isize = s_min_extra_isize;
1368 kfree(is); is = NULL;
1369 kfree(bs); bs = NULL;
1370 brelse(bh);
1371 goto retry;
1372 }
1373 error = -1;
1374 goto cleanup;
1375 }
1376 }
1377 offs = le16_to_cpu(entry->e_value_offs);
1378 size = le32_to_cpu(entry->e_value_size);
1379 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1380 i.name_index = entry->e_name_index,
1381 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1382 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1383 if (!buffer || !b_entry_name) {
1384 error = -ENOMEM;
1385 goto cleanup;
1386 }
1387 /* Save the entry name and the entry value */
1388 memcpy(buffer, (void *)IFIRST(header) + offs,
1389 EXT4_XATTR_SIZE(size));
1390 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1391 b_entry_name[entry->e_name_len] = '\0';
1392 i.name = b_entry_name;
1393
1394 error = ext4_get_inode_loc(inode, &is->iloc);
1395 if (error)
1396 goto cleanup;
1397
1398 error = ext4_xattr_ibody_find(inode, &i, is);
1399 if (error)
1400 goto cleanup;
1401
1402 /* Remove the chosen entry from the inode */
1403 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1404 if (error)
1405 goto cleanup;
1406
1407 entry = IFIRST(header);
1408 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1409 shift_bytes = new_extra_isize;
1410 else
1411 shift_bytes = entry_size + size;
1412 /* Adjust the offsets and shift the remaining entries ahead */
1413 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1414 shift_bytes, (void *)raw_inode +
1415 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1416 (void *)header, total_ino - entry_size,
1417 inode->i_sb->s_blocksize);
1418
1419 extra_isize += shift_bytes;
1420 new_extra_isize -= shift_bytes;
1421 EXT4_I(inode)->i_extra_isize = extra_isize;
1422
1423 i.name = b_entry_name;
1424 i.value = buffer;
1425 i.value_len = size;
1426 error = ext4_xattr_block_find(inode, &i, bs);
1427 if (error)
1428 goto cleanup;
1429
1430 /* Add entry which was removed from the inode into the block */
1431 error = ext4_xattr_block_set(handle, inode, &i, bs);
1432 if (error)
1433 goto cleanup;
1434 kfree(b_entry_name);
1435 kfree(buffer);
1436 b_entry_name = NULL;
1437 buffer = NULL;
1438 brelse(is->iloc.bh);
1439 kfree(is);
1440 kfree(bs);
1441 }
1442 brelse(bh);
1443 up_write(&EXT4_I(inode)->xattr_sem);
1444 return 0;
1445
1446 cleanup:
1447 kfree(b_entry_name);
1448 kfree(buffer);
1449 if (is)
1450 brelse(is->iloc.bh);
1451 kfree(is);
1452 kfree(bs);
1453 brelse(bh);
1454 up_write(&EXT4_I(inode)->xattr_sem);
1455 return error;
1456 }
1457
1458
1459
1460 /*
1461 * ext4_xattr_delete_inode()
1462 *
1463 * Free extended attribute resources associated with this inode. This
1464 * is called immediately before an inode is freed. We have exclusive
1465 * access to the inode.
1466 */
1467 void
1468 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1469 {
1470 struct buffer_head *bh = NULL;
1471
1472 if (!EXT4_I(inode)->i_file_acl)
1473 goto cleanup;
1474 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1475 if (!bh) {
1476 EXT4_ERROR_INODE(inode, "block %llu read error",
1477 EXT4_I(inode)->i_file_acl);
1478 goto cleanup;
1479 }
1480 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1481 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1482 EXT4_ERROR_INODE(inode, "bad block %llu",
1483 EXT4_I(inode)->i_file_acl);
1484 goto cleanup;
1485 }
1486 ext4_xattr_release_block(handle, inode, bh);
1487 EXT4_I(inode)->i_file_acl = 0;
1488
1489 cleanup:
1490 brelse(bh);
1491 }
1492
1493 /*
1494 * ext4_xattr_put_super()
1495 *
1496 * This is called when a file system is unmounted.
1497 */
1498 void
1499 ext4_xattr_put_super(struct super_block *sb)
1500 {
1501 mb_cache_shrink(sb->s_bdev);
1502 }
1503
1504 /*
1505 * ext4_xattr_cache_insert()
1506 *
1507 * Create a new entry in the extended attribute cache, and insert
1508 * it unless such an entry is already in the cache.
1509 *
1510 * Returns 0, or a negative error number on failure.
1511 */
1512 static void
1513 ext4_xattr_cache_insert(struct buffer_head *bh)
1514 {
1515 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1516 struct mb_cache_entry *ce;
1517 int error;
1518
1519 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1520 if (!ce) {
1521 ea_bdebug(bh, "out of memory");
1522 return;
1523 }
1524 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1525 if (error) {
1526 mb_cache_entry_free(ce);
1527 if (error == -EBUSY) {
1528 ea_bdebug(bh, "already in cache");
1529 error = 0;
1530 }
1531 } else {
1532 ea_bdebug(bh, "inserting [%x]", (int)hash);
1533 mb_cache_entry_release(ce);
1534 }
1535 }
1536
1537 /*
1538 * ext4_xattr_cmp()
1539 *
1540 * Compare two extended attribute blocks for equality.
1541 *
1542 * Returns 0 if the blocks are equal, 1 if they differ, and
1543 * a negative error number on errors.
1544 */
1545 static int
1546 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1547 struct ext4_xattr_header *header2)
1548 {
1549 struct ext4_xattr_entry *entry1, *entry2;
1550
1551 entry1 = ENTRY(header1+1);
1552 entry2 = ENTRY(header2+1);
1553 while (!IS_LAST_ENTRY(entry1)) {
1554 if (IS_LAST_ENTRY(entry2))
1555 return 1;
1556 if (entry1->e_hash != entry2->e_hash ||
1557 entry1->e_name_index != entry2->e_name_index ||
1558 entry1->e_name_len != entry2->e_name_len ||
1559 entry1->e_value_size != entry2->e_value_size ||
1560 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1561 return 1;
1562 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1563 return -EIO;
1564 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1565 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1566 le32_to_cpu(entry1->e_value_size)))
1567 return 1;
1568
1569 entry1 = EXT4_XATTR_NEXT(entry1);
1570 entry2 = EXT4_XATTR_NEXT(entry2);
1571 }
1572 if (!IS_LAST_ENTRY(entry2))
1573 return 1;
1574 return 0;
1575 }
1576
1577 /*
1578 * ext4_xattr_cache_find()
1579 *
1580 * Find an identical extended attribute block.
1581 *
1582 * Returns a pointer to the block found, or NULL if such a block was
1583 * not found or an error occurred.
1584 */
1585 static struct buffer_head *
1586 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1587 struct mb_cache_entry **pce)
1588 {
1589 __u32 hash = le32_to_cpu(header->h_hash);
1590 struct mb_cache_entry *ce;
1591
1592 if (!header->h_hash)
1593 return NULL; /* never share */
1594 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1595 again:
1596 ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1597 hash);
1598 while (ce) {
1599 struct buffer_head *bh;
1600
1601 if (IS_ERR(ce)) {
1602 if (PTR_ERR(ce) == -EAGAIN)
1603 goto again;
1604 break;
1605 }
1606 bh = sb_bread(inode->i_sb, ce->e_block);
1607 if (!bh) {
1608 EXT4_ERROR_INODE(inode, "block %lu read error",
1609 (unsigned long) ce->e_block);
1610 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1611 EXT4_XATTR_REFCOUNT_MAX) {
1612 ea_idebug(inode, "block %lu refcount %d>=%d",
1613 (unsigned long) ce->e_block,
1614 le32_to_cpu(BHDR(bh)->h_refcount),
1615 EXT4_XATTR_REFCOUNT_MAX);
1616 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1617 *pce = ce;
1618 return bh;
1619 }
1620 brelse(bh);
1621 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1622 }
1623 return NULL;
1624 }
1625
1626 #define NAME_HASH_SHIFT 5
1627 #define VALUE_HASH_SHIFT 16
1628
1629 /*
1630 * ext4_xattr_hash_entry()
1631 *
1632 * Compute the hash of an extended attribute.
1633 */
1634 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1635 struct ext4_xattr_entry *entry)
1636 {
1637 __u32 hash = 0;
1638 char *name = entry->e_name;
1639 int n;
1640
1641 for (n = 0; n < entry->e_name_len; n++) {
1642 hash = (hash << NAME_HASH_SHIFT) ^
1643 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1644 *name++;
1645 }
1646
1647 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1648 __le32 *value = (__le32 *)((char *)header +
1649 le16_to_cpu(entry->e_value_offs));
1650 for (n = (le32_to_cpu(entry->e_value_size) +
1651 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1652 hash = (hash << VALUE_HASH_SHIFT) ^
1653 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1654 le32_to_cpu(*value++);
1655 }
1656 }
1657 entry->e_hash = cpu_to_le32(hash);
1658 }
1659
1660 #undef NAME_HASH_SHIFT
1661 #undef VALUE_HASH_SHIFT
1662
1663 #define BLOCK_HASH_SHIFT 16
1664
1665 /*
1666 * ext4_xattr_rehash()
1667 *
1668 * Re-compute the extended attribute hash value after an entry has changed.
1669 */
1670 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1671 struct ext4_xattr_entry *entry)
1672 {
1673 struct ext4_xattr_entry *here;
1674 __u32 hash = 0;
1675
1676 ext4_xattr_hash_entry(header, entry);
1677 here = ENTRY(header+1);
1678 while (!IS_LAST_ENTRY(here)) {
1679 if (!here->e_hash) {
1680 /* Block is not shared if an entry's hash value == 0 */
1681 hash = 0;
1682 break;
1683 }
1684 hash = (hash << BLOCK_HASH_SHIFT) ^
1685 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1686 le32_to_cpu(here->e_hash);
1687 here = EXT4_XATTR_NEXT(here);
1688 }
1689 header->h_hash = cpu_to_le32(hash);
1690 }
1691
1692 #undef BLOCK_HASH_SHIFT
1693
1694 int __init
1695 ext4_init_xattr(void)
1696 {
1697 ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1698 if (!ext4_xattr_cache)
1699 return -ENOMEM;
1700 return 0;
1701 }
1702
1703 void
1704 ext4_exit_xattr(void)
1705 {
1706 if (ext4_xattr_cache)
1707 mb_cache_destroy(ext4_xattr_cache);
1708 ext4_xattr_cache = NULL;
1709 }
Linux with added FPC-III support