Merge branch 'master' of git://git.kernel.org/pub/scm/linux/kernel/git/klassert/ipsec...
[deliverable/linux.git] / fs / ext4 / xattr.c
1 /*
2 * linux/fs/ext4/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 * suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
11 * Red Hat Inc.
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 * and Andreas Gruenbacher <agruen@suse.de>.
14 */
15
16 /*
17 * Extended attributes are stored directly in inodes (on file systems with
18 * inodes bigger than 128 bytes) and on additional disk blocks. The i_file_acl
19 * field contains the block number if an inode uses an additional block. All
20 * attributes must fit in the inode and one additional block. Blocks that
21 * contain the identical set of attributes may be shared among several inodes.
22 * Identical blocks are detected by keeping a cache of blocks that have
23 * recently been accessed.
24 *
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
27 *
28 * +------------------+
29 * | header |
30 * | entry 1 | |
31 * | entry 2 | | growing downwards
32 * | entry 3 | v
33 * | four null bytes |
34 * | . . . |
35 * | value 1 | ^
36 * | value 3 | | growing upwards
37 * | value 2 | |
38 * +------------------+
39 *
40 * The header is followed by multiple entry descriptors. In disk blocks, the
41 * entry descriptors are kept sorted. In inodes, they are unsorted. The
42 * attribute values are aligned to the end of the block in no specific order.
43 *
44 * Locking strategy
45 * ----------------
46 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
50 * by the buffer lock.
51 */
52
53 #include <linux/init.h>
54 #include <linux/fs.h>
55 #include <linux/slab.h>
56 #include <linux/mbcache.h>
57 #include <linux/quotaops.h>
58 #include <linux/rwsem.h>
59 #include "ext4_jbd2.h"
60 #include "ext4.h"
61 #include "xattr.h"
62 #include "acl.h"
63
64 #ifdef EXT4_XATTR_DEBUG
65 # define ea_idebug(inode, f...) do { \
66 printk(KERN_DEBUG "inode %s:%lu: ", \
67 inode->i_sb->s_id, inode->i_ino); \
68 printk(f); \
69 printk("\n"); \
70 } while (0)
71 # define ea_bdebug(bh, f...) do { \
72 char b[BDEVNAME_SIZE]; \
73 printk(KERN_DEBUG "block %s:%lu: ", \
74 bdevname(bh->b_bdev, b), \
75 (unsigned long) bh->b_blocknr); \
76 printk(f); \
77 printk("\n"); \
78 } while (0)
79 #else
80 # define ea_idebug(inode, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
81 # define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
82 #endif
83
84 static void ext4_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
85 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
86 struct ext4_xattr_header *,
87 struct mb_cache_entry **);
88 static void ext4_xattr_rehash(struct ext4_xattr_header *,
89 struct ext4_xattr_entry *);
90 static int ext4_xattr_list(struct dentry *dentry, char *buffer,
91 size_t buffer_size);
92
93 static const struct xattr_handler *ext4_xattr_handler_map[] = {
94 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
95 #ifdef CONFIG_EXT4_FS_POSIX_ACL
96 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &posix_acl_access_xattr_handler,
97 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &posix_acl_default_xattr_handler,
98 #endif
99 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
100 #ifdef CONFIG_EXT4_FS_SECURITY
101 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
102 #endif
103 };
104
105 const struct xattr_handler *ext4_xattr_handlers[] = {
106 &ext4_xattr_user_handler,
107 &ext4_xattr_trusted_handler,
108 #ifdef CONFIG_EXT4_FS_POSIX_ACL
109 &posix_acl_access_xattr_handler,
110 &posix_acl_default_xattr_handler,
111 #endif
112 #ifdef CONFIG_EXT4_FS_SECURITY
113 &ext4_xattr_security_handler,
114 #endif
115 NULL
116 };
117
118 #define EXT4_GET_MB_CACHE(inode) (((struct ext4_sb_info *) \
119 inode->i_sb->s_fs_info)->s_mb_cache)
120
121 static __le32 ext4_xattr_block_csum(struct inode *inode,
122 sector_t block_nr,
123 struct ext4_xattr_header *hdr)
124 {
125 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
126 __u32 csum;
127 __le32 save_csum;
128 __le64 dsk_block_nr = cpu_to_le64(block_nr);
129
130 save_csum = hdr->h_checksum;
131 hdr->h_checksum = 0;
132 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
133 sizeof(dsk_block_nr));
134 csum = ext4_chksum(sbi, csum, (__u8 *)hdr,
135 EXT4_BLOCK_SIZE(inode->i_sb));
136
137 hdr->h_checksum = save_csum;
138 return cpu_to_le32(csum);
139 }
140
141 static int ext4_xattr_block_csum_verify(struct inode *inode,
142 sector_t block_nr,
143 struct ext4_xattr_header *hdr)
144 {
145 if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
146 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
147 (hdr->h_checksum != ext4_xattr_block_csum(inode, block_nr, hdr)))
148 return 0;
149 return 1;
150 }
151
152 static void ext4_xattr_block_csum_set(struct inode *inode,
153 sector_t block_nr,
154 struct ext4_xattr_header *hdr)
155 {
156 if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
157 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
158 return;
159
160 hdr->h_checksum = ext4_xattr_block_csum(inode, block_nr, hdr);
161 }
162
163 static inline int ext4_handle_dirty_xattr_block(handle_t *handle,
164 struct inode *inode,
165 struct buffer_head *bh)
166 {
167 ext4_xattr_block_csum_set(inode, bh->b_blocknr, BHDR(bh));
168 return ext4_handle_dirty_metadata(handle, inode, bh);
169 }
170
171 static inline const struct xattr_handler *
172 ext4_xattr_handler(int name_index)
173 {
174 const struct xattr_handler *handler = NULL;
175
176 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
177 handler = ext4_xattr_handler_map[name_index];
178 return handler;
179 }
180
181 /*
182 * Inode operation listxattr()
183 *
184 * dentry->d_inode->i_mutex: don't care
185 */
186 ssize_t
187 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
188 {
189 return ext4_xattr_list(dentry, buffer, size);
190 }
191
192 static int
193 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
194 {
195 while (!IS_LAST_ENTRY(entry)) {
196 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
197 if ((void *)next >= end)
198 return -EIO;
199 entry = next;
200 }
201 return 0;
202 }
203
204 static inline int
205 ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
206 {
207 int error;
208
209 if (buffer_verified(bh))
210 return 0;
211
212 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
213 BHDR(bh)->h_blocks != cpu_to_le32(1))
214 return -EIO;
215 if (!ext4_xattr_block_csum_verify(inode, bh->b_blocknr, BHDR(bh)))
216 return -EIO;
217 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
218 if (!error)
219 set_buffer_verified(bh);
220 return error;
221 }
222
223 static inline int
224 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
225 {
226 size_t value_size = le32_to_cpu(entry->e_value_size);
227
228 if (entry->e_value_block != 0 || value_size > size ||
229 le16_to_cpu(entry->e_value_offs) + value_size > size)
230 return -EIO;
231 return 0;
232 }
233
234 static int
235 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
236 const char *name, size_t size, int sorted)
237 {
238 struct ext4_xattr_entry *entry;
239 size_t name_len;
240 int cmp = 1;
241
242 if (name == NULL)
243 return -EINVAL;
244 name_len = strlen(name);
245 entry = *pentry;
246 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
247 cmp = name_index - entry->e_name_index;
248 if (!cmp)
249 cmp = name_len - entry->e_name_len;
250 if (!cmp)
251 cmp = memcmp(name, entry->e_name, name_len);
252 if (cmp <= 0 && (sorted || cmp == 0))
253 break;
254 }
255 *pentry = entry;
256 if (!cmp && ext4_xattr_check_entry(entry, size))
257 return -EIO;
258 return cmp ? -ENODATA : 0;
259 }
260
261 static int
262 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
263 void *buffer, size_t buffer_size)
264 {
265 struct buffer_head *bh = NULL;
266 struct ext4_xattr_entry *entry;
267 size_t size;
268 int error;
269 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
270
271 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
272 name_index, name, buffer, (long)buffer_size);
273
274 error = -ENODATA;
275 if (!EXT4_I(inode)->i_file_acl)
276 goto cleanup;
277 ea_idebug(inode, "reading block %llu",
278 (unsigned long long)EXT4_I(inode)->i_file_acl);
279 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
280 if (!bh)
281 goto cleanup;
282 ea_bdebug(bh, "b_count=%d, refcount=%d",
283 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
284 if (ext4_xattr_check_block(inode, bh)) {
285 bad_block:
286 EXT4_ERROR_INODE(inode, "bad block %llu",
287 EXT4_I(inode)->i_file_acl);
288 error = -EIO;
289 goto cleanup;
290 }
291 ext4_xattr_cache_insert(ext4_mb_cache, bh);
292 entry = BFIRST(bh);
293 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
294 if (error == -EIO)
295 goto bad_block;
296 if (error)
297 goto cleanup;
298 size = le32_to_cpu(entry->e_value_size);
299 if (buffer) {
300 error = -ERANGE;
301 if (size > buffer_size)
302 goto cleanup;
303 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
304 size);
305 }
306 error = size;
307
308 cleanup:
309 brelse(bh);
310 return error;
311 }
312
313 int
314 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
315 void *buffer, size_t buffer_size)
316 {
317 struct ext4_xattr_ibody_header *header;
318 struct ext4_xattr_entry *entry;
319 struct ext4_inode *raw_inode;
320 struct ext4_iloc iloc;
321 size_t size;
322 void *end;
323 int error;
324
325 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
326 return -ENODATA;
327 error = ext4_get_inode_loc(inode, &iloc);
328 if (error)
329 return error;
330 raw_inode = ext4_raw_inode(&iloc);
331 header = IHDR(inode, raw_inode);
332 entry = IFIRST(header);
333 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
334 error = ext4_xattr_check_names(entry, end);
335 if (error)
336 goto cleanup;
337 error = ext4_xattr_find_entry(&entry, name_index, name,
338 end - (void *)entry, 0);
339 if (error)
340 goto cleanup;
341 size = le32_to_cpu(entry->e_value_size);
342 if (buffer) {
343 error = -ERANGE;
344 if (size > buffer_size)
345 goto cleanup;
346 memcpy(buffer, (void *)IFIRST(header) +
347 le16_to_cpu(entry->e_value_offs), size);
348 }
349 error = size;
350
351 cleanup:
352 brelse(iloc.bh);
353 return error;
354 }
355
356 /*
357 * ext4_xattr_get()
358 *
359 * Copy an extended attribute into the buffer
360 * provided, or compute the buffer size required.
361 * Buffer is NULL to compute the size of the buffer required.
362 *
363 * Returns a negative error number on failure, or the number of bytes
364 * used / required on success.
365 */
366 int
367 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
368 void *buffer, size_t buffer_size)
369 {
370 int error;
371
372 down_read(&EXT4_I(inode)->xattr_sem);
373 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
374 buffer_size);
375 if (error == -ENODATA)
376 error = ext4_xattr_block_get(inode, name_index, name, buffer,
377 buffer_size);
378 up_read(&EXT4_I(inode)->xattr_sem);
379 return error;
380 }
381
382 static int
383 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
384 char *buffer, size_t buffer_size)
385 {
386 size_t rest = buffer_size;
387
388 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
389 const struct xattr_handler *handler =
390 ext4_xattr_handler(entry->e_name_index);
391
392 if (handler) {
393 size_t size = handler->list(dentry, buffer, rest,
394 entry->e_name,
395 entry->e_name_len,
396 handler->flags);
397 if (buffer) {
398 if (size > rest)
399 return -ERANGE;
400 buffer += size;
401 }
402 rest -= size;
403 }
404 }
405 return buffer_size - rest;
406 }
407
408 static int
409 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
410 {
411 struct inode *inode = dentry->d_inode;
412 struct buffer_head *bh = NULL;
413 int error;
414 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
415
416 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
417 buffer, (long)buffer_size);
418
419 error = 0;
420 if (!EXT4_I(inode)->i_file_acl)
421 goto cleanup;
422 ea_idebug(inode, "reading block %llu",
423 (unsigned long long)EXT4_I(inode)->i_file_acl);
424 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
425 error = -EIO;
426 if (!bh)
427 goto cleanup;
428 ea_bdebug(bh, "b_count=%d, refcount=%d",
429 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
430 if (ext4_xattr_check_block(inode, bh)) {
431 EXT4_ERROR_INODE(inode, "bad block %llu",
432 EXT4_I(inode)->i_file_acl);
433 error = -EIO;
434 goto cleanup;
435 }
436 ext4_xattr_cache_insert(ext4_mb_cache, bh);
437 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
438
439 cleanup:
440 brelse(bh);
441
442 return error;
443 }
444
445 static int
446 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
447 {
448 struct inode *inode = dentry->d_inode;
449 struct ext4_xattr_ibody_header *header;
450 struct ext4_inode *raw_inode;
451 struct ext4_iloc iloc;
452 void *end;
453 int error;
454
455 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
456 return 0;
457 error = ext4_get_inode_loc(inode, &iloc);
458 if (error)
459 return error;
460 raw_inode = ext4_raw_inode(&iloc);
461 header = IHDR(inode, raw_inode);
462 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
463 error = ext4_xattr_check_names(IFIRST(header), end);
464 if (error)
465 goto cleanup;
466 error = ext4_xattr_list_entries(dentry, IFIRST(header),
467 buffer, buffer_size);
468
469 cleanup:
470 brelse(iloc.bh);
471 return error;
472 }
473
474 /*
475 * ext4_xattr_list()
476 *
477 * Copy a list of attribute names into the buffer
478 * provided, or compute the buffer size required.
479 * Buffer is NULL to compute the size of the buffer required.
480 *
481 * Returns a negative error number on failure, or the number of bytes
482 * used / required on success.
483 */
484 static int
485 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
486 {
487 int ret, ret2;
488
489 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
490 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
491 if (ret < 0)
492 goto errout;
493 if (buffer) {
494 buffer += ret;
495 buffer_size -= ret;
496 }
497 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
498 if (ret < 0)
499 goto errout;
500 ret += ret2;
501 errout:
502 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
503 return ret;
504 }
505
506 /*
507 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
508 * not set, set it.
509 */
510 static void ext4_xattr_update_super_block(handle_t *handle,
511 struct super_block *sb)
512 {
513 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
514 return;
515
516 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
517 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
518 ext4_handle_dirty_super(handle, sb);
519 }
520 }
521
522 /*
523 * Release the xattr block BH: If the reference count is > 1, decrement it;
524 * otherwise free the block.
525 */
526 static void
527 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
528 struct buffer_head *bh)
529 {
530 struct mb_cache_entry *ce = NULL;
531 int error = 0;
532 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
533
534 ce = mb_cache_entry_get(ext4_mb_cache, bh->b_bdev, bh->b_blocknr);
535 error = ext4_journal_get_write_access(handle, bh);
536 if (error)
537 goto out;
538
539 lock_buffer(bh);
540 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
541 ea_bdebug(bh, "refcount now=0; freeing");
542 if (ce)
543 mb_cache_entry_free(ce);
544 get_bh(bh);
545 unlock_buffer(bh);
546 ext4_free_blocks(handle, inode, bh, 0, 1,
547 EXT4_FREE_BLOCKS_METADATA |
548 EXT4_FREE_BLOCKS_FORGET);
549 } else {
550 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
551 if (ce)
552 mb_cache_entry_release(ce);
553 /*
554 * Beware of this ugliness: Releasing of xattr block references
555 * from different inodes can race and so we have to protect
556 * from a race where someone else frees the block (and releases
557 * its journal_head) before we are done dirtying the buffer. In
558 * nojournal mode this race is harmless and we actually cannot
559 * call ext4_handle_dirty_xattr_block() with locked buffer as
560 * that function can call sync_dirty_buffer() so for that case
561 * we handle the dirtying after unlocking the buffer.
562 */
563 if (ext4_handle_valid(handle))
564 error = ext4_handle_dirty_xattr_block(handle, inode,
565 bh);
566 unlock_buffer(bh);
567 if (!ext4_handle_valid(handle))
568 error = ext4_handle_dirty_xattr_block(handle, inode,
569 bh);
570 if (IS_SYNC(inode))
571 ext4_handle_sync(handle);
572 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
573 ea_bdebug(bh, "refcount now=%d; releasing",
574 le32_to_cpu(BHDR(bh)->h_refcount));
575 }
576 out:
577 ext4_std_error(inode->i_sb, error);
578 return;
579 }
580
581 /*
582 * Find the available free space for EAs. This also returns the total number of
583 * bytes used by EA entries.
584 */
585 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
586 size_t *min_offs, void *base, int *total)
587 {
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 if (total)
595 *total += EXT4_XATTR_LEN(last->e_name_len);
596 }
597 return (*min_offs - ((void *)last - base) - sizeof(__u32));
598 }
599
600 static int
601 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
602 {
603 struct ext4_xattr_entry *last;
604 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
605
606 /* Compute min_offs and last. */
607 last = s->first;
608 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
609 if (!last->e_value_block && last->e_value_size) {
610 size_t offs = le16_to_cpu(last->e_value_offs);
611 if (offs < min_offs)
612 min_offs = offs;
613 }
614 }
615 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
616 if (!s->not_found) {
617 if (!s->here->e_value_block && s->here->e_value_size) {
618 size_t size = le32_to_cpu(s->here->e_value_size);
619 free += EXT4_XATTR_SIZE(size);
620 }
621 free += EXT4_XATTR_LEN(name_len);
622 }
623 if (i->value) {
624 if (free < EXT4_XATTR_SIZE(i->value_len) ||
625 free < EXT4_XATTR_LEN(name_len) +
626 EXT4_XATTR_SIZE(i->value_len))
627 return -ENOSPC;
628 }
629
630 if (i->value && s->not_found) {
631 /* Insert the new name. */
632 size_t size = EXT4_XATTR_LEN(name_len);
633 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
634 memmove((void *)s->here + size, s->here, rest);
635 memset(s->here, 0, size);
636 s->here->e_name_index = i->name_index;
637 s->here->e_name_len = name_len;
638 memcpy(s->here->e_name, i->name, name_len);
639 } else {
640 if (!s->here->e_value_block && s->here->e_value_size) {
641 void *first_val = s->base + min_offs;
642 size_t offs = le16_to_cpu(s->here->e_value_offs);
643 void *val = s->base + offs;
644 size_t size = EXT4_XATTR_SIZE(
645 le32_to_cpu(s->here->e_value_size));
646
647 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
648 /* The old and the new value have the same
649 size. Just replace. */
650 s->here->e_value_size =
651 cpu_to_le32(i->value_len);
652 if (i->value == EXT4_ZERO_XATTR_VALUE) {
653 memset(val, 0, size);
654 } else {
655 /* Clear pad bytes first. */
656 memset(val + size - EXT4_XATTR_PAD, 0,
657 EXT4_XATTR_PAD);
658 memcpy(val, i->value, i->value_len);
659 }
660 return 0;
661 }
662
663 /* Remove the old value. */
664 memmove(first_val + size, first_val, val - first_val);
665 memset(first_val, 0, size);
666 s->here->e_value_size = 0;
667 s->here->e_value_offs = 0;
668 min_offs += size;
669
670 /* Adjust all value offsets. */
671 last = s->first;
672 while (!IS_LAST_ENTRY(last)) {
673 size_t o = le16_to_cpu(last->e_value_offs);
674 if (!last->e_value_block &&
675 last->e_value_size && o < offs)
676 last->e_value_offs =
677 cpu_to_le16(o + size);
678 last = EXT4_XATTR_NEXT(last);
679 }
680 }
681 if (!i->value) {
682 /* Remove the old name. */
683 size_t size = EXT4_XATTR_LEN(name_len);
684 last = ENTRY((void *)last - size);
685 memmove(s->here, (void *)s->here + size,
686 (void *)last - (void *)s->here + sizeof(__u32));
687 memset(last, 0, size);
688 }
689 }
690
691 if (i->value) {
692 /* Insert the new value. */
693 s->here->e_value_size = cpu_to_le32(i->value_len);
694 if (i->value_len) {
695 size_t size = EXT4_XATTR_SIZE(i->value_len);
696 void *val = s->base + min_offs - size;
697 s->here->e_value_offs = cpu_to_le16(min_offs - size);
698 if (i->value == EXT4_ZERO_XATTR_VALUE) {
699 memset(val, 0, size);
700 } else {
701 /* Clear the pad bytes first. */
702 memset(val + size - EXT4_XATTR_PAD, 0,
703 EXT4_XATTR_PAD);
704 memcpy(val, i->value, i->value_len);
705 }
706 }
707 }
708 return 0;
709 }
710
711 struct ext4_xattr_block_find {
712 struct ext4_xattr_search s;
713 struct buffer_head *bh;
714 };
715
716 static int
717 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
718 struct ext4_xattr_block_find *bs)
719 {
720 struct super_block *sb = inode->i_sb;
721 int error;
722
723 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
724 i->name_index, i->name, i->value, (long)i->value_len);
725
726 if (EXT4_I(inode)->i_file_acl) {
727 /* The inode already has an extended attribute block. */
728 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
729 error = -EIO;
730 if (!bs->bh)
731 goto cleanup;
732 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
733 atomic_read(&(bs->bh->b_count)),
734 le32_to_cpu(BHDR(bs->bh)->h_refcount));
735 if (ext4_xattr_check_block(inode, bs->bh)) {
736 EXT4_ERROR_INODE(inode, "bad block %llu",
737 EXT4_I(inode)->i_file_acl);
738 error = -EIO;
739 goto cleanup;
740 }
741 /* Find the named attribute. */
742 bs->s.base = BHDR(bs->bh);
743 bs->s.first = BFIRST(bs->bh);
744 bs->s.end = bs->bh->b_data + bs->bh->b_size;
745 bs->s.here = bs->s.first;
746 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
747 i->name, bs->bh->b_size, 1);
748 if (error && error != -ENODATA)
749 goto cleanup;
750 bs->s.not_found = error;
751 }
752 error = 0;
753
754 cleanup:
755 return error;
756 }
757
758 static int
759 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
760 struct ext4_xattr_info *i,
761 struct ext4_xattr_block_find *bs)
762 {
763 struct super_block *sb = inode->i_sb;
764 struct buffer_head *new_bh = NULL;
765 struct ext4_xattr_search *s = &bs->s;
766 struct mb_cache_entry *ce = NULL;
767 int error = 0;
768 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
769
770 #define header(x) ((struct ext4_xattr_header *)(x))
771
772 if (i->value && i->value_len > sb->s_blocksize)
773 return -ENOSPC;
774 if (s->base) {
775 ce = mb_cache_entry_get(ext4_mb_cache, bs->bh->b_bdev,
776 bs->bh->b_blocknr);
777 error = ext4_journal_get_write_access(handle, bs->bh);
778 if (error)
779 goto cleanup;
780 lock_buffer(bs->bh);
781
782 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
783 if (ce) {
784 mb_cache_entry_free(ce);
785 ce = NULL;
786 }
787 ea_bdebug(bs->bh, "modifying in-place");
788 error = ext4_xattr_set_entry(i, s);
789 if (!error) {
790 if (!IS_LAST_ENTRY(s->first))
791 ext4_xattr_rehash(header(s->base),
792 s->here);
793 ext4_xattr_cache_insert(ext4_mb_cache,
794 bs->bh);
795 }
796 unlock_buffer(bs->bh);
797 if (error == -EIO)
798 goto bad_block;
799 if (!error)
800 error = ext4_handle_dirty_xattr_block(handle,
801 inode,
802 bs->bh);
803 if (error)
804 goto cleanup;
805 goto inserted;
806 } else {
807 int offset = (char *)s->here - bs->bh->b_data;
808
809 unlock_buffer(bs->bh);
810 if (ce) {
811 mb_cache_entry_release(ce);
812 ce = NULL;
813 }
814 ea_bdebug(bs->bh, "cloning");
815 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
816 error = -ENOMEM;
817 if (s->base == NULL)
818 goto cleanup;
819 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
820 s->first = ENTRY(header(s->base)+1);
821 header(s->base)->h_refcount = cpu_to_le32(1);
822 s->here = ENTRY(s->base + offset);
823 s->end = s->base + bs->bh->b_size;
824 }
825 } else {
826 /* Allocate a buffer where we construct the new block. */
827 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
828 /* assert(header == s->base) */
829 error = -ENOMEM;
830 if (s->base == NULL)
831 goto cleanup;
832 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
833 header(s->base)->h_blocks = cpu_to_le32(1);
834 header(s->base)->h_refcount = cpu_to_le32(1);
835 s->first = ENTRY(header(s->base)+1);
836 s->here = ENTRY(header(s->base)+1);
837 s->end = s->base + sb->s_blocksize;
838 }
839
840 error = ext4_xattr_set_entry(i, s);
841 if (error == -EIO)
842 goto bad_block;
843 if (error)
844 goto cleanup;
845 if (!IS_LAST_ENTRY(s->first))
846 ext4_xattr_rehash(header(s->base), s->here);
847
848 inserted:
849 if (!IS_LAST_ENTRY(s->first)) {
850 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
851 if (new_bh) {
852 /* We found an identical block in the cache. */
853 if (new_bh == bs->bh)
854 ea_bdebug(new_bh, "keeping");
855 else {
856 /* The old block is released after updating
857 the inode. */
858 error = dquot_alloc_block(inode,
859 EXT4_C2B(EXT4_SB(sb), 1));
860 if (error)
861 goto cleanup;
862 error = ext4_journal_get_write_access(handle,
863 new_bh);
864 if (error)
865 goto cleanup_dquot;
866 lock_buffer(new_bh);
867 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
868 ea_bdebug(new_bh, "reusing; refcount now=%d",
869 le32_to_cpu(BHDR(new_bh)->h_refcount));
870 unlock_buffer(new_bh);
871 error = ext4_handle_dirty_xattr_block(handle,
872 inode,
873 new_bh);
874 if (error)
875 goto cleanup_dquot;
876 }
877 mb_cache_entry_release(ce);
878 ce = NULL;
879 } else if (bs->bh && s->base == bs->bh->b_data) {
880 /* We were modifying this block in-place. */
881 ea_bdebug(bs->bh, "keeping this block");
882 new_bh = bs->bh;
883 get_bh(new_bh);
884 } else {
885 /* We need to allocate a new block */
886 ext4_fsblk_t goal, block;
887
888 goal = ext4_group_first_block_no(sb,
889 EXT4_I(inode)->i_block_group);
890
891 /* non-extent files can't have physical blocks past 2^32 */
892 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
893 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
894
895 /*
896 * take i_data_sem because we will test
897 * i_delalloc_reserved_flag in ext4_mb_new_blocks
898 */
899 down_read((&EXT4_I(inode)->i_data_sem));
900 block = ext4_new_meta_blocks(handle, inode, goal, 0,
901 NULL, &error);
902 up_read((&EXT4_I(inode)->i_data_sem));
903 if (error)
904 goto cleanup;
905
906 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
907 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
908
909 ea_idebug(inode, "creating block %llu",
910 (unsigned long long)block);
911
912 new_bh = sb_getblk(sb, block);
913 if (unlikely(!new_bh)) {
914 error = -ENOMEM;
915 getblk_failed:
916 ext4_free_blocks(handle, inode, NULL, block, 1,
917 EXT4_FREE_BLOCKS_METADATA);
918 goto cleanup;
919 }
920 lock_buffer(new_bh);
921 error = ext4_journal_get_create_access(handle, new_bh);
922 if (error) {
923 unlock_buffer(new_bh);
924 error = -EIO;
925 goto getblk_failed;
926 }
927 memcpy(new_bh->b_data, s->base, new_bh->b_size);
928 set_buffer_uptodate(new_bh);
929 unlock_buffer(new_bh);
930 ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
931 error = ext4_handle_dirty_xattr_block(handle,
932 inode, new_bh);
933 if (error)
934 goto cleanup;
935 }
936 }
937
938 /* Update the inode. */
939 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
940
941 /* Drop the previous xattr block. */
942 if (bs->bh && bs->bh != new_bh)
943 ext4_xattr_release_block(handle, inode, bs->bh);
944 error = 0;
945
946 cleanup:
947 if (ce)
948 mb_cache_entry_release(ce);
949 brelse(new_bh);
950 if (!(bs->bh && s->base == bs->bh->b_data))
951 kfree(s->base);
952
953 return error;
954
955 cleanup_dquot:
956 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
957 goto cleanup;
958
959 bad_block:
960 EXT4_ERROR_INODE(inode, "bad block %llu",
961 EXT4_I(inode)->i_file_acl);
962 goto cleanup;
963
964 #undef header
965 }
966
967 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
968 struct ext4_xattr_ibody_find *is)
969 {
970 struct ext4_xattr_ibody_header *header;
971 struct ext4_inode *raw_inode;
972 int error;
973
974 if (EXT4_I(inode)->i_extra_isize == 0)
975 return 0;
976 raw_inode = ext4_raw_inode(&is->iloc);
977 header = IHDR(inode, raw_inode);
978 is->s.base = is->s.first = IFIRST(header);
979 is->s.here = is->s.first;
980 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
981 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
982 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
983 if (error)
984 return error;
985 /* Find the named attribute. */
986 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
987 i->name, is->s.end -
988 (void *)is->s.base, 0);
989 if (error && error != -ENODATA)
990 return error;
991 is->s.not_found = error;
992 }
993 return 0;
994 }
995
996 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
997 struct ext4_xattr_info *i,
998 struct ext4_xattr_ibody_find *is)
999 {
1000 struct ext4_xattr_ibody_header *header;
1001 struct ext4_xattr_search *s = &is->s;
1002 int error;
1003
1004 if (EXT4_I(inode)->i_extra_isize == 0)
1005 return -ENOSPC;
1006 error = ext4_xattr_set_entry(i, s);
1007 if (error) {
1008 if (error == -ENOSPC &&
1009 ext4_has_inline_data(inode)) {
1010 error = ext4_try_to_evict_inline_data(handle, inode,
1011 EXT4_XATTR_LEN(strlen(i->name) +
1012 EXT4_XATTR_SIZE(i->value_len)));
1013 if (error)
1014 return error;
1015 error = ext4_xattr_ibody_find(inode, i, is);
1016 if (error)
1017 return error;
1018 error = ext4_xattr_set_entry(i, s);
1019 }
1020 if (error)
1021 return error;
1022 }
1023 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1024 if (!IS_LAST_ENTRY(s->first)) {
1025 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1026 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1027 } else {
1028 header->h_magic = cpu_to_le32(0);
1029 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1030 }
1031 return 0;
1032 }
1033
1034 static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
1035 struct ext4_xattr_info *i,
1036 struct ext4_xattr_ibody_find *is)
1037 {
1038 struct ext4_xattr_ibody_header *header;
1039 struct ext4_xattr_search *s = &is->s;
1040 int error;
1041
1042 if (EXT4_I(inode)->i_extra_isize == 0)
1043 return -ENOSPC;
1044 error = ext4_xattr_set_entry(i, s);
1045 if (error)
1046 return error;
1047 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1048 if (!IS_LAST_ENTRY(s->first)) {
1049 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1050 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1051 } else {
1052 header->h_magic = cpu_to_le32(0);
1053 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1054 }
1055 return 0;
1056 }
1057
1058 /*
1059 * ext4_xattr_set_handle()
1060 *
1061 * Create, replace or remove an extended attribute for this inode. Value
1062 * is NULL to remove an existing extended attribute, and non-NULL to
1063 * either replace an existing extended attribute, or create a new extended
1064 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1065 * specify that an extended attribute must exist and must not exist
1066 * previous to the call, respectively.
1067 *
1068 * Returns 0, or a negative error number on failure.
1069 */
1070 int
1071 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1072 const char *name, const void *value, size_t value_len,
1073 int flags)
1074 {
1075 struct ext4_xattr_info i = {
1076 .name_index = name_index,
1077 .name = name,
1078 .value = value,
1079 .value_len = value_len,
1080
1081 };
1082 struct ext4_xattr_ibody_find is = {
1083 .s = { .not_found = -ENODATA, },
1084 };
1085 struct ext4_xattr_block_find bs = {
1086 .s = { .not_found = -ENODATA, },
1087 };
1088 unsigned long no_expand;
1089 int error;
1090
1091 if (!name)
1092 return -EINVAL;
1093 if (strlen(name) > 255)
1094 return -ERANGE;
1095 down_write(&EXT4_I(inode)->xattr_sem);
1096 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
1097 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
1098
1099 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1100 if (error)
1101 goto cleanup;
1102
1103 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1104 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1105 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1106 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1107 }
1108
1109 error = ext4_xattr_ibody_find(inode, &i, &is);
1110 if (error)
1111 goto cleanup;
1112 if (is.s.not_found)
1113 error = ext4_xattr_block_find(inode, &i, &bs);
1114 if (error)
1115 goto cleanup;
1116 if (is.s.not_found && bs.s.not_found) {
1117 error = -ENODATA;
1118 if (flags & XATTR_REPLACE)
1119 goto cleanup;
1120 error = 0;
1121 if (!value)
1122 goto cleanup;
1123 } else {
1124 error = -EEXIST;
1125 if (flags & XATTR_CREATE)
1126 goto cleanup;
1127 }
1128 if (!value) {
1129 if (!is.s.not_found)
1130 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1131 else if (!bs.s.not_found)
1132 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1133 } else {
1134 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1135 if (!error && !bs.s.not_found) {
1136 i.value = NULL;
1137 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1138 } else if (error == -ENOSPC) {
1139 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1140 error = ext4_xattr_block_find(inode, &i, &bs);
1141 if (error)
1142 goto cleanup;
1143 }
1144 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1145 if (error)
1146 goto cleanup;
1147 if (!is.s.not_found) {
1148 i.value = NULL;
1149 error = ext4_xattr_ibody_set(handle, inode, &i,
1150 &is);
1151 }
1152 }
1153 }
1154 if (!error) {
1155 ext4_xattr_update_super_block(handle, inode->i_sb);
1156 inode->i_ctime = ext4_current_time(inode);
1157 if (!value)
1158 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1159 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1160 /*
1161 * The bh is consumed by ext4_mark_iloc_dirty, even with
1162 * error != 0.
1163 */
1164 is.iloc.bh = NULL;
1165 if (IS_SYNC(inode))
1166 ext4_handle_sync(handle);
1167 }
1168
1169 cleanup:
1170 brelse(is.iloc.bh);
1171 brelse(bs.bh);
1172 if (no_expand == 0)
1173 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1174 up_write(&EXT4_I(inode)->xattr_sem);
1175 return error;
1176 }
1177
1178 /*
1179 * ext4_xattr_set()
1180 *
1181 * Like ext4_xattr_set_handle, but start from an inode. This extended
1182 * attribute modification is a filesystem transaction by itself.
1183 *
1184 * Returns 0, or a negative error number on failure.
1185 */
1186 int
1187 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1188 const void *value, size_t value_len, int flags)
1189 {
1190 handle_t *handle;
1191 int error, retries = 0;
1192 int credits = ext4_jbd2_credits_xattr(inode);
1193
1194 retry:
1195 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1196 if (IS_ERR(handle)) {
1197 error = PTR_ERR(handle);
1198 } else {
1199 int error2;
1200
1201 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1202 value, value_len, flags);
1203 error2 = ext4_journal_stop(handle);
1204 if (error == -ENOSPC &&
1205 ext4_should_retry_alloc(inode->i_sb, &retries))
1206 goto retry;
1207 if (error == 0)
1208 error = error2;
1209 }
1210
1211 return error;
1212 }
1213
1214 /*
1215 * Shift the EA entries in the inode to create space for the increased
1216 * i_extra_isize.
1217 */
1218 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1219 int value_offs_shift, void *to,
1220 void *from, size_t n, int blocksize)
1221 {
1222 struct ext4_xattr_entry *last = entry;
1223 int new_offs;
1224
1225 /* Adjust the value offsets of the entries */
1226 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1227 if (!last->e_value_block && last->e_value_size) {
1228 new_offs = le16_to_cpu(last->e_value_offs) +
1229 value_offs_shift;
1230 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1231 > blocksize);
1232 last->e_value_offs = cpu_to_le16(new_offs);
1233 }
1234 }
1235 /* Shift the entries by n bytes */
1236 memmove(to, from, n);
1237 }
1238
1239 /*
1240 * Expand an inode by new_extra_isize bytes when EAs are present.
1241 * Returns 0 on success or negative error number on failure.
1242 */
1243 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1244 struct ext4_inode *raw_inode, handle_t *handle)
1245 {
1246 struct ext4_xattr_ibody_header *header;
1247 struct ext4_xattr_entry *entry, *last, *first;
1248 struct buffer_head *bh = NULL;
1249 struct ext4_xattr_ibody_find *is = NULL;
1250 struct ext4_xattr_block_find *bs = NULL;
1251 char *buffer = NULL, *b_entry_name = NULL;
1252 size_t min_offs, free;
1253 int total_ino;
1254 void *base, *start, *end;
1255 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1256 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1257
1258 down_write(&EXT4_I(inode)->xattr_sem);
1259 retry:
1260 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1261 up_write(&EXT4_I(inode)->xattr_sem);
1262 return 0;
1263 }
1264
1265 header = IHDR(inode, raw_inode);
1266 entry = IFIRST(header);
1267
1268 /*
1269 * Check if enough free space is available in the inode to shift the
1270 * entries ahead by new_extra_isize.
1271 */
1272
1273 base = start = entry;
1274 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1275 min_offs = end - base;
1276 last = entry;
1277 total_ino = sizeof(struct ext4_xattr_ibody_header);
1278
1279 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1280 if (free >= new_extra_isize) {
1281 entry = IFIRST(header);
1282 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1283 - new_extra_isize, (void *)raw_inode +
1284 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1285 (void *)header, total_ino,
1286 inode->i_sb->s_blocksize);
1287 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1288 error = 0;
1289 goto cleanup;
1290 }
1291
1292 /*
1293 * Enough free space isn't available in the inode, check if
1294 * EA block can hold new_extra_isize bytes.
1295 */
1296 if (EXT4_I(inode)->i_file_acl) {
1297 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1298 error = -EIO;
1299 if (!bh)
1300 goto cleanup;
1301 if (ext4_xattr_check_block(inode, bh)) {
1302 EXT4_ERROR_INODE(inode, "bad block %llu",
1303 EXT4_I(inode)->i_file_acl);
1304 error = -EIO;
1305 goto cleanup;
1306 }
1307 base = BHDR(bh);
1308 first = BFIRST(bh);
1309 end = bh->b_data + bh->b_size;
1310 min_offs = end - base;
1311 free = ext4_xattr_free_space(first, &min_offs, base, NULL);
1312 if (free < new_extra_isize) {
1313 if (!tried_min_extra_isize && s_min_extra_isize) {
1314 tried_min_extra_isize++;
1315 new_extra_isize = s_min_extra_isize;
1316 brelse(bh);
1317 goto retry;
1318 }
1319 error = -1;
1320 goto cleanup;
1321 }
1322 } else {
1323 free = inode->i_sb->s_blocksize;
1324 }
1325
1326 while (new_extra_isize > 0) {
1327 size_t offs, size, entry_size;
1328 struct ext4_xattr_entry *small_entry = NULL;
1329 struct ext4_xattr_info i = {
1330 .value = NULL,
1331 .value_len = 0,
1332 };
1333 unsigned int total_size; /* EA entry size + value size */
1334 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1335 unsigned int min_total_size = ~0U;
1336
1337 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1338 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1339 if (!is || !bs) {
1340 error = -ENOMEM;
1341 goto cleanup;
1342 }
1343
1344 is->s.not_found = -ENODATA;
1345 bs->s.not_found = -ENODATA;
1346 is->iloc.bh = NULL;
1347 bs->bh = NULL;
1348
1349 last = IFIRST(header);
1350 /* Find the entry best suited to be pushed into EA block */
1351 entry = NULL;
1352 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1353 total_size =
1354 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1355 EXT4_XATTR_LEN(last->e_name_len);
1356 if (total_size <= free && total_size < min_total_size) {
1357 if (total_size < new_extra_isize) {
1358 small_entry = last;
1359 } else {
1360 entry = last;
1361 min_total_size = total_size;
1362 }
1363 }
1364 }
1365
1366 if (entry == NULL) {
1367 if (small_entry) {
1368 entry = small_entry;
1369 } else {
1370 if (!tried_min_extra_isize &&
1371 s_min_extra_isize) {
1372 tried_min_extra_isize++;
1373 new_extra_isize = s_min_extra_isize;
1374 kfree(is); is = NULL;
1375 kfree(bs); bs = NULL;
1376 brelse(bh);
1377 goto retry;
1378 }
1379 error = -1;
1380 goto cleanup;
1381 }
1382 }
1383 offs = le16_to_cpu(entry->e_value_offs);
1384 size = le32_to_cpu(entry->e_value_size);
1385 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1386 i.name_index = entry->e_name_index,
1387 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1388 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1389 if (!buffer || !b_entry_name) {
1390 error = -ENOMEM;
1391 goto cleanup;
1392 }
1393 /* Save the entry name and the entry value */
1394 memcpy(buffer, (void *)IFIRST(header) + offs,
1395 EXT4_XATTR_SIZE(size));
1396 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1397 b_entry_name[entry->e_name_len] = '\0';
1398 i.name = b_entry_name;
1399
1400 error = ext4_get_inode_loc(inode, &is->iloc);
1401 if (error)
1402 goto cleanup;
1403
1404 error = ext4_xattr_ibody_find(inode, &i, is);
1405 if (error)
1406 goto cleanup;
1407
1408 /* Remove the chosen entry from the inode */
1409 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1410 if (error)
1411 goto cleanup;
1412
1413 entry = IFIRST(header);
1414 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1415 shift_bytes = new_extra_isize;
1416 else
1417 shift_bytes = entry_size + size;
1418 /* Adjust the offsets and shift the remaining entries ahead */
1419 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1420 shift_bytes, (void *)raw_inode +
1421 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1422 (void *)header, total_ino - entry_size,
1423 inode->i_sb->s_blocksize);
1424
1425 extra_isize += shift_bytes;
1426 new_extra_isize -= shift_bytes;
1427 EXT4_I(inode)->i_extra_isize = extra_isize;
1428
1429 i.name = b_entry_name;
1430 i.value = buffer;
1431 i.value_len = size;
1432 error = ext4_xattr_block_find(inode, &i, bs);
1433 if (error)
1434 goto cleanup;
1435
1436 /* Add entry which was removed from the inode into the block */
1437 error = ext4_xattr_block_set(handle, inode, &i, bs);
1438 if (error)
1439 goto cleanup;
1440 kfree(b_entry_name);
1441 kfree(buffer);
1442 b_entry_name = NULL;
1443 buffer = NULL;
1444 brelse(is->iloc.bh);
1445 kfree(is);
1446 kfree(bs);
1447 }
1448 brelse(bh);
1449 up_write(&EXT4_I(inode)->xattr_sem);
1450 return 0;
1451
1452 cleanup:
1453 kfree(b_entry_name);
1454 kfree(buffer);
1455 if (is)
1456 brelse(is->iloc.bh);
1457 kfree(is);
1458 kfree(bs);
1459 brelse(bh);
1460 up_write(&EXT4_I(inode)->xattr_sem);
1461 return error;
1462 }
1463
1464
1465
1466 /*
1467 * ext4_xattr_delete_inode()
1468 *
1469 * Free extended attribute resources associated with this inode. This
1470 * is called immediately before an inode is freed. We have exclusive
1471 * access to the inode.
1472 */
1473 void
1474 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1475 {
1476 struct buffer_head *bh = NULL;
1477
1478 if (!EXT4_I(inode)->i_file_acl)
1479 goto cleanup;
1480 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1481 if (!bh) {
1482 EXT4_ERROR_INODE(inode, "block %llu read error",
1483 EXT4_I(inode)->i_file_acl);
1484 goto cleanup;
1485 }
1486 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1487 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1488 EXT4_ERROR_INODE(inode, "bad block %llu",
1489 EXT4_I(inode)->i_file_acl);
1490 goto cleanup;
1491 }
1492 ext4_xattr_release_block(handle, inode, bh);
1493 EXT4_I(inode)->i_file_acl = 0;
1494
1495 cleanup:
1496 brelse(bh);
1497 }
1498
1499 /*
1500 * ext4_xattr_put_super()
1501 *
1502 * This is called when a file system is unmounted.
1503 */
1504 void
1505 ext4_xattr_put_super(struct super_block *sb)
1506 {
1507 mb_cache_shrink(sb->s_bdev);
1508 }
1509
1510 /*
1511 * ext4_xattr_cache_insert()
1512 *
1513 * Create a new entry in the extended attribute cache, and insert
1514 * it unless such an entry is already in the cache.
1515 *
1516 * Returns 0, or a negative error number on failure.
1517 */
1518 static void
1519 ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
1520 {
1521 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1522 struct mb_cache_entry *ce;
1523 int error;
1524
1525 ce = mb_cache_entry_alloc(ext4_mb_cache, GFP_NOFS);
1526 if (!ce) {
1527 ea_bdebug(bh, "out of memory");
1528 return;
1529 }
1530 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1531 if (error) {
1532 mb_cache_entry_free(ce);
1533 if (error == -EBUSY) {
1534 ea_bdebug(bh, "already in cache");
1535 error = 0;
1536 }
1537 } else {
1538 ea_bdebug(bh, "inserting [%x]", (int)hash);
1539 mb_cache_entry_release(ce);
1540 }
1541 }
1542
1543 /*
1544 * ext4_xattr_cmp()
1545 *
1546 * Compare two extended attribute blocks for equality.
1547 *
1548 * Returns 0 if the blocks are equal, 1 if they differ, and
1549 * a negative error number on errors.
1550 */
1551 static int
1552 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1553 struct ext4_xattr_header *header2)
1554 {
1555 struct ext4_xattr_entry *entry1, *entry2;
1556
1557 entry1 = ENTRY(header1+1);
1558 entry2 = ENTRY(header2+1);
1559 while (!IS_LAST_ENTRY(entry1)) {
1560 if (IS_LAST_ENTRY(entry2))
1561 return 1;
1562 if (entry1->e_hash != entry2->e_hash ||
1563 entry1->e_name_index != entry2->e_name_index ||
1564 entry1->e_name_len != entry2->e_name_len ||
1565 entry1->e_value_size != entry2->e_value_size ||
1566 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1567 return 1;
1568 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1569 return -EIO;
1570 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1571 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1572 le32_to_cpu(entry1->e_value_size)))
1573 return 1;
1574
1575 entry1 = EXT4_XATTR_NEXT(entry1);
1576 entry2 = EXT4_XATTR_NEXT(entry2);
1577 }
1578 if (!IS_LAST_ENTRY(entry2))
1579 return 1;
1580 return 0;
1581 }
1582
1583 /*
1584 * ext4_xattr_cache_find()
1585 *
1586 * Find an identical extended attribute block.
1587 *
1588 * Returns a pointer to the block found, or NULL if such a block was
1589 * not found or an error occurred.
1590 */
1591 static struct buffer_head *
1592 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1593 struct mb_cache_entry **pce)
1594 {
1595 __u32 hash = le32_to_cpu(header->h_hash);
1596 struct mb_cache_entry *ce;
1597 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
1598
1599 if (!header->h_hash)
1600 return NULL; /* never share */
1601 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1602 again:
1603 ce = mb_cache_entry_find_first(ext4_mb_cache, inode->i_sb->s_bdev,
1604 hash);
1605 while (ce) {
1606 struct buffer_head *bh;
1607
1608 if (IS_ERR(ce)) {
1609 if (PTR_ERR(ce) == -EAGAIN)
1610 goto again;
1611 break;
1612 }
1613 bh = sb_bread(inode->i_sb, ce->e_block);
1614 if (!bh) {
1615 EXT4_ERROR_INODE(inode, "block %lu read error",
1616 (unsigned long) ce->e_block);
1617 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1618 EXT4_XATTR_REFCOUNT_MAX) {
1619 ea_idebug(inode, "block %lu refcount %d>=%d",
1620 (unsigned long) ce->e_block,
1621 le32_to_cpu(BHDR(bh)->h_refcount),
1622 EXT4_XATTR_REFCOUNT_MAX);
1623 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1624 *pce = ce;
1625 return bh;
1626 }
1627 brelse(bh);
1628 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1629 }
1630 return NULL;
1631 }
1632
1633 #define NAME_HASH_SHIFT 5
1634 #define VALUE_HASH_SHIFT 16
1635
1636 /*
1637 * ext4_xattr_hash_entry()
1638 *
1639 * Compute the hash of an extended attribute.
1640 */
1641 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1642 struct ext4_xattr_entry *entry)
1643 {
1644 __u32 hash = 0;
1645 char *name = entry->e_name;
1646 int n;
1647
1648 for (n = 0; n < entry->e_name_len; n++) {
1649 hash = (hash << NAME_HASH_SHIFT) ^
1650 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1651 *name++;
1652 }
1653
1654 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1655 __le32 *value = (__le32 *)((char *)header +
1656 le16_to_cpu(entry->e_value_offs));
1657 for (n = (le32_to_cpu(entry->e_value_size) +
1658 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1659 hash = (hash << VALUE_HASH_SHIFT) ^
1660 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1661 le32_to_cpu(*value++);
1662 }
1663 }
1664 entry->e_hash = cpu_to_le32(hash);
1665 }
1666
1667 #undef NAME_HASH_SHIFT
1668 #undef VALUE_HASH_SHIFT
1669
1670 #define BLOCK_HASH_SHIFT 16
1671
1672 /*
1673 * ext4_xattr_rehash()
1674 *
1675 * Re-compute the extended attribute hash value after an entry has changed.
1676 */
1677 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1678 struct ext4_xattr_entry *entry)
1679 {
1680 struct ext4_xattr_entry *here;
1681 __u32 hash = 0;
1682
1683 ext4_xattr_hash_entry(header, entry);
1684 here = ENTRY(header+1);
1685 while (!IS_LAST_ENTRY(here)) {
1686 if (!here->e_hash) {
1687 /* Block is not shared if an entry's hash value == 0 */
1688 hash = 0;
1689 break;
1690 }
1691 hash = (hash << BLOCK_HASH_SHIFT) ^
1692 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1693 le32_to_cpu(here->e_hash);
1694 here = EXT4_XATTR_NEXT(here);
1695 }
1696 header->h_hash = cpu_to_le32(hash);
1697 }
1698
1699 #undef BLOCK_HASH_SHIFT
1700
1701 #define HASH_BUCKET_BITS 10
1702
1703 struct mb_cache *
1704 ext4_xattr_create_cache(char *name)
1705 {
1706 return mb_cache_create(name, HASH_BUCKET_BITS);
1707 }
1708
1709 void ext4_xattr_destroy_cache(struct mb_cache *cache)
1710 {
1711 if (cache)
1712 mb_cache_destroy(cache);
1713 }
1714
This page took 0.06497 seconds and 5 git commands to generate.