]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/ext4/xattr.c
ext4: give more helpful error message in ext4_ext_rm_leaf()
[mirror_ubuntu-artful-kernel.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 #define BHDR(bh) ((struct ext4_xattr_header *)((bh)->b_data))
65 #define ENTRY(ptr) ((struct ext4_xattr_entry *)(ptr))
66 #define BFIRST(bh) ENTRY(BHDR(bh)+1)
67 #define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
68
69 #ifdef EXT4_XATTR_DEBUG
70 # define ea_idebug(inode, f...) do { \
71 printk(KERN_DEBUG "inode %s:%lu: ", \
72 inode->i_sb->s_id, inode->i_ino); \
73 printk(f); \
74 printk("\n"); \
75 } while (0)
76 # define ea_bdebug(bh, f...) do { \
77 char b[BDEVNAME_SIZE]; \
78 printk(KERN_DEBUG "block %s:%lu: ", \
79 bdevname(bh->b_bdev, b), \
80 (unsigned long) bh->b_blocknr); \
81 printk(f); \
82 printk("\n"); \
83 } while (0)
84 #else
85 # define ea_idebug(f...)
86 # define ea_bdebug(f...)
87 #endif
88
89 static void ext4_xattr_cache_insert(struct buffer_head *);
90 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
91 struct ext4_xattr_header *,
92 struct mb_cache_entry **);
93 static void ext4_xattr_rehash(struct ext4_xattr_header *,
94 struct ext4_xattr_entry *);
95 static int ext4_xattr_list(struct dentry *dentry, char *buffer,
96 size_t buffer_size);
97
98 static struct mb_cache *ext4_xattr_cache;
99
100 static const struct xattr_handler *ext4_xattr_handler_map[] = {
101 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
102 #ifdef CONFIG_EXT4_FS_POSIX_ACL
103 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
104 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
105 #endif
106 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
107 #ifdef CONFIG_EXT4_FS_SECURITY
108 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
109 #endif
110 };
111
112 const struct xattr_handler *ext4_xattr_handlers[] = {
113 &ext4_xattr_user_handler,
114 &ext4_xattr_trusted_handler,
115 #ifdef CONFIG_EXT4_FS_POSIX_ACL
116 &ext4_xattr_acl_access_handler,
117 &ext4_xattr_acl_default_handler,
118 #endif
119 #ifdef CONFIG_EXT4_FS_SECURITY
120 &ext4_xattr_security_handler,
121 #endif
122 NULL
123 };
124
125 static inline const struct xattr_handler *
126 ext4_xattr_handler(int name_index)
127 {
128 const struct xattr_handler *handler = NULL;
129
130 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
131 handler = ext4_xattr_handler_map[name_index];
132 return handler;
133 }
134
135 /*
136 * Inode operation listxattr()
137 *
138 * dentry->d_inode->i_mutex: don't care
139 */
140 ssize_t
141 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
142 {
143 return ext4_xattr_list(dentry, buffer, size);
144 }
145
146 static int
147 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
148 {
149 while (!IS_LAST_ENTRY(entry)) {
150 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
151 if ((void *)next >= end)
152 return -EIO;
153 entry = next;
154 }
155 return 0;
156 }
157
158 static inline int
159 ext4_xattr_check_block(struct buffer_head *bh)
160 {
161 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
162 BHDR(bh)->h_blocks != cpu_to_le32(1))
163 return -EIO;
164 return ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
165 }
166
167 static inline int
168 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
169 {
170 size_t value_size = le32_to_cpu(entry->e_value_size);
171
172 if (entry->e_value_block != 0 || value_size > size ||
173 le16_to_cpu(entry->e_value_offs) + value_size > size)
174 return -EIO;
175 return 0;
176 }
177
178 static int
179 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
180 const char *name, size_t size, int sorted)
181 {
182 struct ext4_xattr_entry *entry;
183 size_t name_len;
184 int cmp = 1;
185
186 if (name == NULL)
187 return -EINVAL;
188 name_len = strlen(name);
189 entry = *pentry;
190 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
191 cmp = name_index - entry->e_name_index;
192 if (!cmp)
193 cmp = name_len - entry->e_name_len;
194 if (!cmp)
195 cmp = memcmp(name, entry->e_name, name_len);
196 if (cmp <= 0 && (sorted || cmp == 0))
197 break;
198 }
199 *pentry = entry;
200 if (!cmp && ext4_xattr_check_entry(entry, size))
201 return -EIO;
202 return cmp ? -ENODATA : 0;
203 }
204
205 static int
206 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
207 void *buffer, size_t buffer_size)
208 {
209 struct buffer_head *bh = NULL;
210 struct ext4_xattr_entry *entry;
211 size_t size;
212 int error;
213
214 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
215 name_index, name, buffer, (long)buffer_size);
216
217 error = -ENODATA;
218 if (!EXT4_I(inode)->i_file_acl)
219 goto cleanup;
220 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
221 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
222 if (!bh)
223 goto cleanup;
224 ea_bdebug(bh, "b_count=%d, refcount=%d",
225 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
226 if (ext4_xattr_check_block(bh)) {
227 bad_block:
228 EXT4_ERROR_INODE(inode, "bad block %llu",
229 EXT4_I(inode)->i_file_acl);
230 error = -EIO;
231 goto cleanup;
232 }
233 ext4_xattr_cache_insert(bh);
234 entry = BFIRST(bh);
235 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
236 if (error == -EIO)
237 goto bad_block;
238 if (error)
239 goto cleanup;
240 size = le32_to_cpu(entry->e_value_size);
241 if (buffer) {
242 error = -ERANGE;
243 if (size > buffer_size)
244 goto cleanup;
245 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
246 size);
247 }
248 error = size;
249
250 cleanup:
251 brelse(bh);
252 return error;
253 }
254
255 static int
256 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
257 void *buffer, size_t buffer_size)
258 {
259 struct ext4_xattr_ibody_header *header;
260 struct ext4_xattr_entry *entry;
261 struct ext4_inode *raw_inode;
262 struct ext4_iloc iloc;
263 size_t size;
264 void *end;
265 int error;
266
267 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
268 return -ENODATA;
269 error = ext4_get_inode_loc(inode, &iloc);
270 if (error)
271 return error;
272 raw_inode = ext4_raw_inode(&iloc);
273 header = IHDR(inode, raw_inode);
274 entry = IFIRST(header);
275 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
276 error = ext4_xattr_check_names(entry, end);
277 if (error)
278 goto cleanup;
279 error = ext4_xattr_find_entry(&entry, name_index, name,
280 end - (void *)entry, 0);
281 if (error)
282 goto cleanup;
283 size = le32_to_cpu(entry->e_value_size);
284 if (buffer) {
285 error = -ERANGE;
286 if (size > buffer_size)
287 goto cleanup;
288 memcpy(buffer, (void *)IFIRST(header) +
289 le16_to_cpu(entry->e_value_offs), size);
290 }
291 error = size;
292
293 cleanup:
294 brelse(iloc.bh);
295 return error;
296 }
297
298 /*
299 * ext4_xattr_get()
300 *
301 * Copy an extended attribute into the buffer
302 * provided, or compute the buffer size required.
303 * Buffer is NULL to compute the size of the buffer required.
304 *
305 * Returns a negative error number on failure, or the number of bytes
306 * used / required on success.
307 */
308 int
309 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
310 void *buffer, size_t buffer_size)
311 {
312 int error;
313
314 down_read(&EXT4_I(inode)->xattr_sem);
315 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
316 buffer_size);
317 if (error == -ENODATA)
318 error = ext4_xattr_block_get(inode, name_index, name, buffer,
319 buffer_size);
320 up_read(&EXT4_I(inode)->xattr_sem);
321 return error;
322 }
323
324 static int
325 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
326 char *buffer, size_t buffer_size)
327 {
328 size_t rest = buffer_size;
329
330 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
331 const struct xattr_handler *handler =
332 ext4_xattr_handler(entry->e_name_index);
333
334 if (handler) {
335 size_t size = handler->list(dentry, buffer, rest,
336 entry->e_name,
337 entry->e_name_len,
338 handler->flags);
339 if (buffer) {
340 if (size > rest)
341 return -ERANGE;
342 buffer += size;
343 }
344 rest -= size;
345 }
346 }
347 return buffer_size - rest;
348 }
349
350 static int
351 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
352 {
353 struct inode *inode = dentry->d_inode;
354 struct buffer_head *bh = NULL;
355 int error;
356
357 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
358 buffer, (long)buffer_size);
359
360 error = 0;
361 if (!EXT4_I(inode)->i_file_acl)
362 goto cleanup;
363 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
364 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
365 error = -EIO;
366 if (!bh)
367 goto cleanup;
368 ea_bdebug(bh, "b_count=%d, refcount=%d",
369 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
370 if (ext4_xattr_check_block(bh)) {
371 EXT4_ERROR_INODE(inode, "bad block %llu",
372 EXT4_I(inode)->i_file_acl);
373 error = -EIO;
374 goto cleanup;
375 }
376 ext4_xattr_cache_insert(bh);
377 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
378
379 cleanup:
380 brelse(bh);
381
382 return error;
383 }
384
385 static int
386 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
387 {
388 struct inode *inode = dentry->d_inode;
389 struct ext4_xattr_ibody_header *header;
390 struct ext4_inode *raw_inode;
391 struct ext4_iloc iloc;
392 void *end;
393 int error;
394
395 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
396 return 0;
397 error = ext4_get_inode_loc(inode, &iloc);
398 if (error)
399 return error;
400 raw_inode = ext4_raw_inode(&iloc);
401 header = IHDR(inode, raw_inode);
402 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
403 error = ext4_xattr_check_names(IFIRST(header), end);
404 if (error)
405 goto cleanup;
406 error = ext4_xattr_list_entries(dentry, IFIRST(header),
407 buffer, buffer_size);
408
409 cleanup:
410 brelse(iloc.bh);
411 return error;
412 }
413
414 /*
415 * ext4_xattr_list()
416 *
417 * Copy a list of attribute names into the buffer
418 * provided, or compute the buffer size required.
419 * Buffer is NULL to compute the size of the buffer required.
420 *
421 * Returns a negative error number on failure, or the number of bytes
422 * used / required on success.
423 */
424 static int
425 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
426 {
427 int ret, ret2;
428
429 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
430 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
431 if (ret < 0)
432 goto errout;
433 if (buffer) {
434 buffer += ret;
435 buffer_size -= ret;
436 }
437 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
438 if (ret < 0)
439 goto errout;
440 ret += ret2;
441 errout:
442 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
443 return ret;
444 }
445
446 /*
447 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
448 * not set, set it.
449 */
450 static void ext4_xattr_update_super_block(handle_t *handle,
451 struct super_block *sb)
452 {
453 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
454 return;
455
456 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
457 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
458 ext4_handle_dirty_super(handle, sb);
459 }
460 }
461
462 /*
463 * Release the xattr block BH: If the reference count is > 1, decrement
464 * it; otherwise free the block.
465 */
466 static void
467 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
468 struct buffer_head *bh)
469 {
470 struct mb_cache_entry *ce = NULL;
471 int error = 0;
472
473 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
474 error = ext4_journal_get_write_access(handle, bh);
475 if (error)
476 goto out;
477
478 lock_buffer(bh);
479 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
480 ea_bdebug(bh, "refcount now=0; freeing");
481 if (ce)
482 mb_cache_entry_free(ce);
483 get_bh(bh);
484 ext4_free_blocks(handle, inode, bh, 0, 1,
485 EXT4_FREE_BLOCKS_METADATA |
486 EXT4_FREE_BLOCKS_FORGET);
487 unlock_buffer(bh);
488 } else {
489 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
490 if (ce)
491 mb_cache_entry_release(ce);
492 unlock_buffer(bh);
493 error = ext4_handle_dirty_metadata(handle, inode, bh);
494 if (IS_SYNC(inode))
495 ext4_handle_sync(handle);
496 dquot_free_block(inode, 1);
497 ea_bdebug(bh, "refcount now=%d; releasing",
498 le32_to_cpu(BHDR(bh)->h_refcount));
499 }
500 out:
501 ext4_std_error(inode->i_sb, error);
502 return;
503 }
504
505 /*
506 * Find the available free space for EAs. This also returns the total number of
507 * bytes used by EA entries.
508 */
509 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
510 size_t *min_offs, void *base, int *total)
511 {
512 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
513 *total += EXT4_XATTR_LEN(last->e_name_len);
514 if (!last->e_value_block && last->e_value_size) {
515 size_t offs = le16_to_cpu(last->e_value_offs);
516 if (offs < *min_offs)
517 *min_offs = offs;
518 }
519 }
520 return (*min_offs - ((void *)last - base) - sizeof(__u32));
521 }
522
523 struct ext4_xattr_info {
524 int name_index;
525 const char *name;
526 const void *value;
527 size_t value_len;
528 };
529
530 struct ext4_xattr_search {
531 struct ext4_xattr_entry *first;
532 void *base;
533 void *end;
534 struct ext4_xattr_entry *here;
535 int not_found;
536 };
537
538 static int
539 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
540 {
541 struct ext4_xattr_entry *last;
542 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
543
544 /* Compute min_offs and last. */
545 last = s->first;
546 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
547 if (!last->e_value_block && last->e_value_size) {
548 size_t offs = le16_to_cpu(last->e_value_offs);
549 if (offs < min_offs)
550 min_offs = offs;
551 }
552 }
553 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
554 if (!s->not_found) {
555 if (!s->here->e_value_block && s->here->e_value_size) {
556 size_t size = le32_to_cpu(s->here->e_value_size);
557 free += EXT4_XATTR_SIZE(size);
558 }
559 free += EXT4_XATTR_LEN(name_len);
560 }
561 if (i->value) {
562 if (free < EXT4_XATTR_SIZE(i->value_len) ||
563 free < EXT4_XATTR_LEN(name_len) +
564 EXT4_XATTR_SIZE(i->value_len))
565 return -ENOSPC;
566 }
567
568 if (i->value && s->not_found) {
569 /* Insert the new name. */
570 size_t size = EXT4_XATTR_LEN(name_len);
571 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
572 memmove((void *)s->here + size, s->here, rest);
573 memset(s->here, 0, size);
574 s->here->e_name_index = i->name_index;
575 s->here->e_name_len = name_len;
576 memcpy(s->here->e_name, i->name, name_len);
577 } else {
578 if (!s->here->e_value_block && s->here->e_value_size) {
579 void *first_val = s->base + min_offs;
580 size_t offs = le16_to_cpu(s->here->e_value_offs);
581 void *val = s->base + offs;
582 size_t size = EXT4_XATTR_SIZE(
583 le32_to_cpu(s->here->e_value_size));
584
585 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
586 /* The old and the new value have the same
587 size. Just replace. */
588 s->here->e_value_size =
589 cpu_to_le32(i->value_len);
590 memset(val + size - EXT4_XATTR_PAD, 0,
591 EXT4_XATTR_PAD); /* Clear pad bytes. */
592 memcpy(val, i->value, i->value_len);
593 return 0;
594 }
595
596 /* Remove the old value. */
597 memmove(first_val + size, first_val, val - first_val);
598 memset(first_val, 0, size);
599 s->here->e_value_size = 0;
600 s->here->e_value_offs = 0;
601 min_offs += size;
602
603 /* Adjust all value offsets. */
604 last = s->first;
605 while (!IS_LAST_ENTRY(last)) {
606 size_t o = le16_to_cpu(last->e_value_offs);
607 if (!last->e_value_block &&
608 last->e_value_size && o < offs)
609 last->e_value_offs =
610 cpu_to_le16(o + size);
611 last = EXT4_XATTR_NEXT(last);
612 }
613 }
614 if (!i->value) {
615 /* Remove the old name. */
616 size_t size = EXT4_XATTR_LEN(name_len);
617 last = ENTRY((void *)last - size);
618 memmove(s->here, (void *)s->here + size,
619 (void *)last - (void *)s->here + sizeof(__u32));
620 memset(last, 0, size);
621 }
622 }
623
624 if (i->value) {
625 /* Insert the new value. */
626 s->here->e_value_size = cpu_to_le32(i->value_len);
627 if (i->value_len) {
628 size_t size = EXT4_XATTR_SIZE(i->value_len);
629 void *val = s->base + min_offs - size;
630 s->here->e_value_offs = cpu_to_le16(min_offs - size);
631 memset(val + size - EXT4_XATTR_PAD, 0,
632 EXT4_XATTR_PAD); /* Clear the pad bytes. */
633 memcpy(val, i->value, i->value_len);
634 }
635 }
636 return 0;
637 }
638
639 struct ext4_xattr_block_find {
640 struct ext4_xattr_search s;
641 struct buffer_head *bh;
642 };
643
644 static int
645 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
646 struct ext4_xattr_block_find *bs)
647 {
648 struct super_block *sb = inode->i_sb;
649 int error;
650
651 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
652 i->name_index, i->name, i->value, (long)i->value_len);
653
654 if (EXT4_I(inode)->i_file_acl) {
655 /* The inode already has an extended attribute block. */
656 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
657 error = -EIO;
658 if (!bs->bh)
659 goto cleanup;
660 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
661 atomic_read(&(bs->bh->b_count)),
662 le32_to_cpu(BHDR(bs->bh)->h_refcount));
663 if (ext4_xattr_check_block(bs->bh)) {
664 EXT4_ERROR_INODE(inode, "bad block %llu",
665 EXT4_I(inode)->i_file_acl);
666 error = -EIO;
667 goto cleanup;
668 }
669 /* Find the named attribute. */
670 bs->s.base = BHDR(bs->bh);
671 bs->s.first = BFIRST(bs->bh);
672 bs->s.end = bs->bh->b_data + bs->bh->b_size;
673 bs->s.here = bs->s.first;
674 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
675 i->name, bs->bh->b_size, 1);
676 if (error && error != -ENODATA)
677 goto cleanup;
678 bs->s.not_found = error;
679 }
680 error = 0;
681
682 cleanup:
683 return error;
684 }
685
686 static int
687 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
688 struct ext4_xattr_info *i,
689 struct ext4_xattr_block_find *bs)
690 {
691 struct super_block *sb = inode->i_sb;
692 struct buffer_head *new_bh = NULL;
693 struct ext4_xattr_search *s = &bs->s;
694 struct mb_cache_entry *ce = NULL;
695 int error = 0;
696
697 #define header(x) ((struct ext4_xattr_header *)(x))
698
699 if (i->value && i->value_len > sb->s_blocksize)
700 return -ENOSPC;
701 if (s->base) {
702 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
703 bs->bh->b_blocknr);
704 error = ext4_journal_get_write_access(handle, bs->bh);
705 if (error)
706 goto cleanup;
707 lock_buffer(bs->bh);
708
709 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
710 if (ce) {
711 mb_cache_entry_free(ce);
712 ce = NULL;
713 }
714 ea_bdebug(bs->bh, "modifying in-place");
715 error = ext4_xattr_set_entry(i, s);
716 if (!error) {
717 if (!IS_LAST_ENTRY(s->first))
718 ext4_xattr_rehash(header(s->base),
719 s->here);
720 ext4_xattr_cache_insert(bs->bh);
721 }
722 unlock_buffer(bs->bh);
723 if (error == -EIO)
724 goto bad_block;
725 if (!error)
726 error = ext4_handle_dirty_metadata(handle,
727 inode,
728 bs->bh);
729 if (error)
730 goto cleanup;
731 goto inserted;
732 } else {
733 int offset = (char *)s->here - bs->bh->b_data;
734
735 unlock_buffer(bs->bh);
736 ext4_handle_release_buffer(handle, bs->bh);
737 if (ce) {
738 mb_cache_entry_release(ce);
739 ce = NULL;
740 }
741 ea_bdebug(bs->bh, "cloning");
742 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
743 error = -ENOMEM;
744 if (s->base == NULL)
745 goto cleanup;
746 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
747 s->first = ENTRY(header(s->base)+1);
748 header(s->base)->h_refcount = cpu_to_le32(1);
749 s->here = ENTRY(s->base + offset);
750 s->end = s->base + bs->bh->b_size;
751 }
752 } else {
753 /* Allocate a buffer where we construct the new block. */
754 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
755 /* assert(header == s->base) */
756 error = -ENOMEM;
757 if (s->base == NULL)
758 goto cleanup;
759 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
760 header(s->base)->h_blocks = cpu_to_le32(1);
761 header(s->base)->h_refcount = cpu_to_le32(1);
762 s->first = ENTRY(header(s->base)+1);
763 s->here = ENTRY(header(s->base)+1);
764 s->end = s->base + sb->s_blocksize;
765 }
766
767 error = ext4_xattr_set_entry(i, s);
768 if (error == -EIO)
769 goto bad_block;
770 if (error)
771 goto cleanup;
772 if (!IS_LAST_ENTRY(s->first))
773 ext4_xattr_rehash(header(s->base), s->here);
774
775 inserted:
776 if (!IS_LAST_ENTRY(s->first)) {
777 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
778 if (new_bh) {
779 /* We found an identical block in the cache. */
780 if (new_bh == bs->bh)
781 ea_bdebug(new_bh, "keeping");
782 else {
783 /* The old block is released after updating
784 the inode. */
785 error = dquot_alloc_block(inode, 1);
786 if (error)
787 goto cleanup;
788 error = ext4_journal_get_write_access(handle,
789 new_bh);
790 if (error)
791 goto cleanup_dquot;
792 lock_buffer(new_bh);
793 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
794 ea_bdebug(new_bh, "reusing; refcount now=%d",
795 le32_to_cpu(BHDR(new_bh)->h_refcount));
796 unlock_buffer(new_bh);
797 error = ext4_handle_dirty_metadata(handle,
798 inode,
799 new_bh);
800 if (error)
801 goto cleanup_dquot;
802 }
803 mb_cache_entry_release(ce);
804 ce = NULL;
805 } else if (bs->bh && s->base == bs->bh->b_data) {
806 /* We were modifying this block in-place. */
807 ea_bdebug(bs->bh, "keeping this block");
808 new_bh = bs->bh;
809 get_bh(new_bh);
810 } else {
811 /* We need to allocate a new block */
812 ext4_fsblk_t goal, block;
813
814 goal = ext4_group_first_block_no(sb,
815 EXT4_I(inode)->i_block_group);
816
817 /* non-extent files can't have physical blocks past 2^32 */
818 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
819 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
820
821 /*
822 * take i_data_sem because we will test
823 * i_delalloc_reserved_flag in ext4_mb_new_blocks
824 */
825 down_read((&EXT4_I(inode)->i_data_sem));
826 block = ext4_new_meta_blocks(handle, inode, goal, 0,
827 NULL, &error);
828 up_read((&EXT4_I(inode)->i_data_sem));
829 if (error)
830 goto cleanup;
831
832 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
833 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
834
835 ea_idebug(inode, "creating block %d", block);
836
837 new_bh = sb_getblk(sb, block);
838 if (!new_bh) {
839 getblk_failed:
840 ext4_free_blocks(handle, inode, NULL, block, 1,
841 EXT4_FREE_BLOCKS_METADATA);
842 error = -EIO;
843 goto cleanup;
844 }
845 lock_buffer(new_bh);
846 error = ext4_journal_get_create_access(handle, new_bh);
847 if (error) {
848 unlock_buffer(new_bh);
849 goto getblk_failed;
850 }
851 memcpy(new_bh->b_data, s->base, new_bh->b_size);
852 set_buffer_uptodate(new_bh);
853 unlock_buffer(new_bh);
854 ext4_xattr_cache_insert(new_bh);
855 error = ext4_handle_dirty_metadata(handle,
856 inode, new_bh);
857 if (error)
858 goto cleanup;
859 }
860 }
861
862 /* Update the inode. */
863 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
864
865 /* Drop the previous xattr block. */
866 if (bs->bh && bs->bh != new_bh)
867 ext4_xattr_release_block(handle, inode, bs->bh);
868 error = 0;
869
870 cleanup:
871 if (ce)
872 mb_cache_entry_release(ce);
873 brelse(new_bh);
874 if (!(bs->bh && s->base == bs->bh->b_data))
875 kfree(s->base);
876
877 return error;
878
879 cleanup_dquot:
880 dquot_free_block(inode, 1);
881 goto cleanup;
882
883 bad_block:
884 EXT4_ERROR_INODE(inode, "bad block %llu",
885 EXT4_I(inode)->i_file_acl);
886 goto cleanup;
887
888 #undef header
889 }
890
891 struct ext4_xattr_ibody_find {
892 struct ext4_xattr_search s;
893 struct ext4_iloc iloc;
894 };
895
896 static int
897 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
898 struct ext4_xattr_ibody_find *is)
899 {
900 struct ext4_xattr_ibody_header *header;
901 struct ext4_inode *raw_inode;
902 int error;
903
904 if (EXT4_I(inode)->i_extra_isize == 0)
905 return 0;
906 raw_inode = ext4_raw_inode(&is->iloc);
907 header = IHDR(inode, raw_inode);
908 is->s.base = is->s.first = IFIRST(header);
909 is->s.here = is->s.first;
910 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
911 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
912 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
913 if (error)
914 return error;
915 /* Find the named attribute. */
916 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
917 i->name, is->s.end -
918 (void *)is->s.base, 0);
919 if (error && error != -ENODATA)
920 return error;
921 is->s.not_found = error;
922 }
923 return 0;
924 }
925
926 static int
927 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
928 struct ext4_xattr_info *i,
929 struct ext4_xattr_ibody_find *is)
930 {
931 struct ext4_xattr_ibody_header *header;
932 struct ext4_xattr_search *s = &is->s;
933 int error;
934
935 if (EXT4_I(inode)->i_extra_isize == 0)
936 return -ENOSPC;
937 error = ext4_xattr_set_entry(i, s);
938 if (error)
939 return error;
940 header = IHDR(inode, ext4_raw_inode(&is->iloc));
941 if (!IS_LAST_ENTRY(s->first)) {
942 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
943 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
944 } else {
945 header->h_magic = cpu_to_le32(0);
946 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
947 }
948 return 0;
949 }
950
951 /*
952 * ext4_xattr_set_handle()
953 *
954 * Create, replace or remove an extended attribute for this inode. Value
955 * is NULL to remove an existing extended attribute, and non-NULL to
956 * either replace an existing extended attribute, or create a new extended
957 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
958 * specify that an extended attribute must exist and must not exist
959 * previous to the call, respectively.
960 *
961 * Returns 0, or a negative error number on failure.
962 */
963 int
964 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
965 const char *name, const void *value, size_t value_len,
966 int flags)
967 {
968 struct ext4_xattr_info i = {
969 .name_index = name_index,
970 .name = name,
971 .value = value,
972 .value_len = value_len,
973
974 };
975 struct ext4_xattr_ibody_find is = {
976 .s = { .not_found = -ENODATA, },
977 };
978 struct ext4_xattr_block_find bs = {
979 .s = { .not_found = -ENODATA, },
980 };
981 unsigned long no_expand;
982 int error;
983
984 if (!name)
985 return -EINVAL;
986 if (strlen(name) > 255)
987 return -ERANGE;
988 down_write(&EXT4_I(inode)->xattr_sem);
989 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
990 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
991
992 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
993 if (error)
994 goto cleanup;
995
996 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
997 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
998 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
999 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1000 }
1001
1002 error = ext4_xattr_ibody_find(inode, &i, &is);
1003 if (error)
1004 goto cleanup;
1005 if (is.s.not_found)
1006 error = ext4_xattr_block_find(inode, &i, &bs);
1007 if (error)
1008 goto cleanup;
1009 if (is.s.not_found && bs.s.not_found) {
1010 error = -ENODATA;
1011 if (flags & XATTR_REPLACE)
1012 goto cleanup;
1013 error = 0;
1014 if (!value)
1015 goto cleanup;
1016 } else {
1017 error = -EEXIST;
1018 if (flags & XATTR_CREATE)
1019 goto cleanup;
1020 }
1021 if (!value) {
1022 if (!is.s.not_found)
1023 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1024 else if (!bs.s.not_found)
1025 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1026 } else {
1027 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1028 if (!error && !bs.s.not_found) {
1029 i.value = NULL;
1030 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1031 } else if (error == -ENOSPC) {
1032 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1033 error = ext4_xattr_block_find(inode, &i, &bs);
1034 if (error)
1035 goto cleanup;
1036 }
1037 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1038 if (error)
1039 goto cleanup;
1040 if (!is.s.not_found) {
1041 i.value = NULL;
1042 error = ext4_xattr_ibody_set(handle, inode, &i,
1043 &is);
1044 }
1045 }
1046 }
1047 if (!error) {
1048 ext4_xattr_update_super_block(handle, inode->i_sb);
1049 inode->i_ctime = ext4_current_time(inode);
1050 if (!value)
1051 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1052 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1053 /*
1054 * The bh is consumed by ext4_mark_iloc_dirty, even with
1055 * error != 0.
1056 */
1057 is.iloc.bh = NULL;
1058 if (IS_SYNC(inode))
1059 ext4_handle_sync(handle);
1060 }
1061
1062 cleanup:
1063 brelse(is.iloc.bh);
1064 brelse(bs.bh);
1065 if (no_expand == 0)
1066 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1067 up_write(&EXT4_I(inode)->xattr_sem);
1068 return error;
1069 }
1070
1071 /*
1072 * ext4_xattr_set()
1073 *
1074 * Like ext4_xattr_set_handle, but start from an inode. This extended
1075 * attribute modification is a filesystem transaction by itself.
1076 *
1077 * Returns 0, or a negative error number on failure.
1078 */
1079 int
1080 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1081 const void *value, size_t value_len, int flags)
1082 {
1083 handle_t *handle;
1084 int error, retries = 0;
1085
1086 retry:
1087 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1088 if (IS_ERR(handle)) {
1089 error = PTR_ERR(handle);
1090 } else {
1091 int error2;
1092
1093 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1094 value, value_len, flags);
1095 error2 = ext4_journal_stop(handle);
1096 if (error == -ENOSPC &&
1097 ext4_should_retry_alloc(inode->i_sb, &retries))
1098 goto retry;
1099 if (error == 0)
1100 error = error2;
1101 }
1102
1103 return error;
1104 }
1105
1106 /*
1107 * Shift the EA entries in the inode to create space for the increased
1108 * i_extra_isize.
1109 */
1110 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1111 int value_offs_shift, void *to,
1112 void *from, size_t n, int blocksize)
1113 {
1114 struct ext4_xattr_entry *last = entry;
1115 int new_offs;
1116
1117 /* Adjust the value offsets of the entries */
1118 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1119 if (!last->e_value_block && last->e_value_size) {
1120 new_offs = le16_to_cpu(last->e_value_offs) +
1121 value_offs_shift;
1122 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1123 > blocksize);
1124 last->e_value_offs = cpu_to_le16(new_offs);
1125 }
1126 }
1127 /* Shift the entries by n bytes */
1128 memmove(to, from, n);
1129 }
1130
1131 /*
1132 * Expand an inode by new_extra_isize bytes when EAs are present.
1133 * Returns 0 on success or negative error number on failure.
1134 */
1135 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1136 struct ext4_inode *raw_inode, handle_t *handle)
1137 {
1138 struct ext4_xattr_ibody_header *header;
1139 struct ext4_xattr_entry *entry, *last, *first;
1140 struct buffer_head *bh = NULL;
1141 struct ext4_xattr_ibody_find *is = NULL;
1142 struct ext4_xattr_block_find *bs = NULL;
1143 char *buffer = NULL, *b_entry_name = NULL;
1144 size_t min_offs, free;
1145 int total_ino, total_blk;
1146 void *base, *start, *end;
1147 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1148 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1149
1150 down_write(&EXT4_I(inode)->xattr_sem);
1151 retry:
1152 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1153 up_write(&EXT4_I(inode)->xattr_sem);
1154 return 0;
1155 }
1156
1157 header = IHDR(inode, raw_inode);
1158 entry = IFIRST(header);
1159
1160 /*
1161 * Check if enough free space is available in the inode to shift the
1162 * entries ahead by new_extra_isize.
1163 */
1164
1165 base = start = entry;
1166 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1167 min_offs = end - base;
1168 last = entry;
1169 total_ino = sizeof(struct ext4_xattr_ibody_header);
1170
1171 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1172 if (free >= new_extra_isize) {
1173 entry = IFIRST(header);
1174 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1175 - new_extra_isize, (void *)raw_inode +
1176 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1177 (void *)header, total_ino,
1178 inode->i_sb->s_blocksize);
1179 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1180 error = 0;
1181 goto cleanup;
1182 }
1183
1184 /*
1185 * Enough free space isn't available in the inode, check if
1186 * EA block can hold new_extra_isize bytes.
1187 */
1188 if (EXT4_I(inode)->i_file_acl) {
1189 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1190 error = -EIO;
1191 if (!bh)
1192 goto cleanup;
1193 if (ext4_xattr_check_block(bh)) {
1194 EXT4_ERROR_INODE(inode, "bad block %llu",
1195 EXT4_I(inode)->i_file_acl);
1196 error = -EIO;
1197 goto cleanup;
1198 }
1199 base = BHDR(bh);
1200 first = BFIRST(bh);
1201 end = bh->b_data + bh->b_size;
1202 min_offs = end - base;
1203 free = ext4_xattr_free_space(first, &min_offs, base,
1204 &total_blk);
1205 if (free < new_extra_isize) {
1206 if (!tried_min_extra_isize && s_min_extra_isize) {
1207 tried_min_extra_isize++;
1208 new_extra_isize = s_min_extra_isize;
1209 brelse(bh);
1210 goto retry;
1211 }
1212 error = -1;
1213 goto cleanup;
1214 }
1215 } else {
1216 free = inode->i_sb->s_blocksize;
1217 }
1218
1219 while (new_extra_isize > 0) {
1220 size_t offs, size, entry_size;
1221 struct ext4_xattr_entry *small_entry = NULL;
1222 struct ext4_xattr_info i = {
1223 .value = NULL,
1224 .value_len = 0,
1225 };
1226 unsigned int total_size; /* EA entry size + value size */
1227 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1228 unsigned int min_total_size = ~0U;
1229
1230 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1231 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1232 if (!is || !bs) {
1233 error = -ENOMEM;
1234 goto cleanup;
1235 }
1236
1237 is->s.not_found = -ENODATA;
1238 bs->s.not_found = -ENODATA;
1239 is->iloc.bh = NULL;
1240 bs->bh = NULL;
1241
1242 last = IFIRST(header);
1243 /* Find the entry best suited to be pushed into EA block */
1244 entry = NULL;
1245 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1246 total_size =
1247 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1248 EXT4_XATTR_LEN(last->e_name_len);
1249 if (total_size <= free && total_size < min_total_size) {
1250 if (total_size < new_extra_isize) {
1251 small_entry = last;
1252 } else {
1253 entry = last;
1254 min_total_size = total_size;
1255 }
1256 }
1257 }
1258
1259 if (entry == NULL) {
1260 if (small_entry) {
1261 entry = small_entry;
1262 } else {
1263 if (!tried_min_extra_isize &&
1264 s_min_extra_isize) {
1265 tried_min_extra_isize++;
1266 new_extra_isize = s_min_extra_isize;
1267 goto retry;
1268 }
1269 error = -1;
1270 goto cleanup;
1271 }
1272 }
1273 offs = le16_to_cpu(entry->e_value_offs);
1274 size = le32_to_cpu(entry->e_value_size);
1275 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1276 i.name_index = entry->e_name_index,
1277 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1278 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1279 if (!buffer || !b_entry_name) {
1280 error = -ENOMEM;
1281 goto cleanup;
1282 }
1283 /* Save the entry name and the entry value */
1284 memcpy(buffer, (void *)IFIRST(header) + offs,
1285 EXT4_XATTR_SIZE(size));
1286 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1287 b_entry_name[entry->e_name_len] = '\0';
1288 i.name = b_entry_name;
1289
1290 error = ext4_get_inode_loc(inode, &is->iloc);
1291 if (error)
1292 goto cleanup;
1293
1294 error = ext4_xattr_ibody_find(inode, &i, is);
1295 if (error)
1296 goto cleanup;
1297
1298 /* Remove the chosen entry from the inode */
1299 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1300 if (error)
1301 goto cleanup;
1302
1303 entry = IFIRST(header);
1304 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1305 shift_bytes = new_extra_isize;
1306 else
1307 shift_bytes = entry_size + size;
1308 /* Adjust the offsets and shift the remaining entries ahead */
1309 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1310 shift_bytes, (void *)raw_inode +
1311 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1312 (void *)header, total_ino - entry_size,
1313 inode->i_sb->s_blocksize);
1314
1315 extra_isize += shift_bytes;
1316 new_extra_isize -= shift_bytes;
1317 EXT4_I(inode)->i_extra_isize = extra_isize;
1318
1319 i.name = b_entry_name;
1320 i.value = buffer;
1321 i.value_len = size;
1322 error = ext4_xattr_block_find(inode, &i, bs);
1323 if (error)
1324 goto cleanup;
1325
1326 /* Add entry which was removed from the inode into the block */
1327 error = ext4_xattr_block_set(handle, inode, &i, bs);
1328 if (error)
1329 goto cleanup;
1330 kfree(b_entry_name);
1331 kfree(buffer);
1332 b_entry_name = NULL;
1333 buffer = NULL;
1334 brelse(is->iloc.bh);
1335 kfree(is);
1336 kfree(bs);
1337 }
1338 brelse(bh);
1339 up_write(&EXT4_I(inode)->xattr_sem);
1340 return 0;
1341
1342 cleanup:
1343 kfree(b_entry_name);
1344 kfree(buffer);
1345 if (is)
1346 brelse(is->iloc.bh);
1347 kfree(is);
1348 kfree(bs);
1349 brelse(bh);
1350 up_write(&EXT4_I(inode)->xattr_sem);
1351 return error;
1352 }
1353
1354
1355
1356 /*
1357 * ext4_xattr_delete_inode()
1358 *
1359 * Free extended attribute resources associated with this inode. This
1360 * is called immediately before an inode is freed. We have exclusive
1361 * access to the inode.
1362 */
1363 void
1364 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1365 {
1366 struct buffer_head *bh = NULL;
1367
1368 if (!EXT4_I(inode)->i_file_acl)
1369 goto cleanup;
1370 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1371 if (!bh) {
1372 EXT4_ERROR_INODE(inode, "block %llu read error",
1373 EXT4_I(inode)->i_file_acl);
1374 goto cleanup;
1375 }
1376 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1377 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1378 EXT4_ERROR_INODE(inode, "bad block %llu",
1379 EXT4_I(inode)->i_file_acl);
1380 goto cleanup;
1381 }
1382 ext4_xattr_release_block(handle, inode, bh);
1383 EXT4_I(inode)->i_file_acl = 0;
1384
1385 cleanup:
1386 brelse(bh);
1387 }
1388
1389 /*
1390 * ext4_xattr_put_super()
1391 *
1392 * This is called when a file system is unmounted.
1393 */
1394 void
1395 ext4_xattr_put_super(struct super_block *sb)
1396 {
1397 mb_cache_shrink(sb->s_bdev);
1398 }
1399
1400 /*
1401 * ext4_xattr_cache_insert()
1402 *
1403 * Create a new entry in the extended attribute cache, and insert
1404 * it unless such an entry is already in the cache.
1405 *
1406 * Returns 0, or a negative error number on failure.
1407 */
1408 static void
1409 ext4_xattr_cache_insert(struct buffer_head *bh)
1410 {
1411 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1412 struct mb_cache_entry *ce;
1413 int error;
1414
1415 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1416 if (!ce) {
1417 ea_bdebug(bh, "out of memory");
1418 return;
1419 }
1420 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1421 if (error) {
1422 mb_cache_entry_free(ce);
1423 if (error == -EBUSY) {
1424 ea_bdebug(bh, "already in cache");
1425 error = 0;
1426 }
1427 } else {
1428 ea_bdebug(bh, "inserting [%x]", (int)hash);
1429 mb_cache_entry_release(ce);
1430 }
1431 }
1432
1433 /*
1434 * ext4_xattr_cmp()
1435 *
1436 * Compare two extended attribute blocks for equality.
1437 *
1438 * Returns 0 if the blocks are equal, 1 if they differ, and
1439 * a negative error number on errors.
1440 */
1441 static int
1442 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1443 struct ext4_xattr_header *header2)
1444 {
1445 struct ext4_xattr_entry *entry1, *entry2;
1446
1447 entry1 = ENTRY(header1+1);
1448 entry2 = ENTRY(header2+1);
1449 while (!IS_LAST_ENTRY(entry1)) {
1450 if (IS_LAST_ENTRY(entry2))
1451 return 1;
1452 if (entry1->e_hash != entry2->e_hash ||
1453 entry1->e_name_index != entry2->e_name_index ||
1454 entry1->e_name_len != entry2->e_name_len ||
1455 entry1->e_value_size != entry2->e_value_size ||
1456 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1457 return 1;
1458 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1459 return -EIO;
1460 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1461 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1462 le32_to_cpu(entry1->e_value_size)))
1463 return 1;
1464
1465 entry1 = EXT4_XATTR_NEXT(entry1);
1466 entry2 = EXT4_XATTR_NEXT(entry2);
1467 }
1468 if (!IS_LAST_ENTRY(entry2))
1469 return 1;
1470 return 0;
1471 }
1472
1473 /*
1474 * ext4_xattr_cache_find()
1475 *
1476 * Find an identical extended attribute block.
1477 *
1478 * Returns a pointer to the block found, or NULL if such a block was
1479 * not found or an error occurred.
1480 */
1481 static struct buffer_head *
1482 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1483 struct mb_cache_entry **pce)
1484 {
1485 __u32 hash = le32_to_cpu(header->h_hash);
1486 struct mb_cache_entry *ce;
1487
1488 if (!header->h_hash)
1489 return NULL; /* never share */
1490 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1491 again:
1492 ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1493 hash);
1494 while (ce) {
1495 struct buffer_head *bh;
1496
1497 if (IS_ERR(ce)) {
1498 if (PTR_ERR(ce) == -EAGAIN)
1499 goto again;
1500 break;
1501 }
1502 bh = sb_bread(inode->i_sb, ce->e_block);
1503 if (!bh) {
1504 EXT4_ERROR_INODE(inode, "block %lu read error",
1505 (unsigned long) ce->e_block);
1506 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1507 EXT4_XATTR_REFCOUNT_MAX) {
1508 ea_idebug(inode, "block %lu refcount %d>=%d",
1509 (unsigned long) ce->e_block,
1510 le32_to_cpu(BHDR(bh)->h_refcount),
1511 EXT4_XATTR_REFCOUNT_MAX);
1512 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1513 *pce = ce;
1514 return bh;
1515 }
1516 brelse(bh);
1517 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1518 }
1519 return NULL;
1520 }
1521
1522 #define NAME_HASH_SHIFT 5
1523 #define VALUE_HASH_SHIFT 16
1524
1525 /*
1526 * ext4_xattr_hash_entry()
1527 *
1528 * Compute the hash of an extended attribute.
1529 */
1530 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1531 struct ext4_xattr_entry *entry)
1532 {
1533 __u32 hash = 0;
1534 char *name = entry->e_name;
1535 int n;
1536
1537 for (n = 0; n < entry->e_name_len; n++) {
1538 hash = (hash << NAME_HASH_SHIFT) ^
1539 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1540 *name++;
1541 }
1542
1543 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1544 __le32 *value = (__le32 *)((char *)header +
1545 le16_to_cpu(entry->e_value_offs));
1546 for (n = (le32_to_cpu(entry->e_value_size) +
1547 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1548 hash = (hash << VALUE_HASH_SHIFT) ^
1549 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1550 le32_to_cpu(*value++);
1551 }
1552 }
1553 entry->e_hash = cpu_to_le32(hash);
1554 }
1555
1556 #undef NAME_HASH_SHIFT
1557 #undef VALUE_HASH_SHIFT
1558
1559 #define BLOCK_HASH_SHIFT 16
1560
1561 /*
1562 * ext4_xattr_rehash()
1563 *
1564 * Re-compute the extended attribute hash value after an entry has changed.
1565 */
1566 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1567 struct ext4_xattr_entry *entry)
1568 {
1569 struct ext4_xattr_entry *here;
1570 __u32 hash = 0;
1571
1572 ext4_xattr_hash_entry(header, entry);
1573 here = ENTRY(header+1);
1574 while (!IS_LAST_ENTRY(here)) {
1575 if (!here->e_hash) {
1576 /* Block is not shared if an entry's hash value == 0 */
1577 hash = 0;
1578 break;
1579 }
1580 hash = (hash << BLOCK_HASH_SHIFT) ^
1581 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1582 le32_to_cpu(here->e_hash);
1583 here = EXT4_XATTR_NEXT(here);
1584 }
1585 header->h_hash = cpu_to_le32(hash);
1586 }
1587
1588 #undef BLOCK_HASH_SHIFT
1589
1590 int __init
1591 ext4_init_xattr(void)
1592 {
1593 ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1594 if (!ext4_xattr_cache)
1595 return -ENOMEM;
1596 return 0;
1597 }
1598
1599 void
1600 ext4_exit_xattr(void)
1601 {
1602 if (ext4_xattr_cache)
1603 mb_cache_destroy(ext4_xattr_cache);
1604 ext4_xattr_cache = NULL;
1605 }