Linux 2.6.25.20
[linux/fpc-iii.git] / fs / ext4 / xattr.c
blob70d856badabf55168c76641b124f1001c3b9865a
1 /*
2 * linux/fs/ext4/xattr.c
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
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>.
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.
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
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 * +------------------+
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.
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.
53 #include <linux/init.h>
54 #include <linux/fs.h>
55 #include <linux/slab.h>
56 #include <linux/ext4_jbd2.h>
57 #include <linux/ext4_fs.h>
58 #include <linux/mbcache.h>
59 #include <linux/quotaops.h>
60 #include <linux/rwsem.h>
61 #include "xattr.h"
62 #include "acl.h"
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)
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
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 *);
96 static struct mb_cache *ext4_xattr_cache;
98 static struct xattr_handler *ext4_xattr_handler_map[] = {
99 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
100 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
101 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
102 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
103 #endif
104 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
105 #ifdef CONFIG_EXT4DEV_FS_SECURITY
106 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
107 #endif
110 struct xattr_handler *ext4_xattr_handlers[] = {
111 &ext4_xattr_user_handler,
112 &ext4_xattr_trusted_handler,
113 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
114 &ext4_xattr_acl_access_handler,
115 &ext4_xattr_acl_default_handler,
116 #endif
117 #ifdef CONFIG_EXT4DEV_FS_SECURITY
118 &ext4_xattr_security_handler,
119 #endif
120 NULL
123 static inline struct xattr_handler *
124 ext4_xattr_handler(int name_index)
126 struct xattr_handler *handler = NULL;
128 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
129 handler = ext4_xattr_handler_map[name_index];
130 return handler;
134 * Inode operation listxattr()
136 * dentry->d_inode->i_mutex: don't care
138 ssize_t
139 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
141 return ext4_xattr_list(dentry->d_inode, buffer, size);
144 static int
145 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
147 while (!IS_LAST_ENTRY(entry)) {
148 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
149 if ((void *)next >= end)
150 return -EIO;
151 entry = next;
153 return 0;
156 static inline int
157 ext4_xattr_check_block(struct buffer_head *bh)
159 int error;
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 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
165 return error;
168 static inline int
169 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
171 size_t value_size = le32_to_cpu(entry->e_value_size);
173 if (entry->e_value_block != 0 || value_size > size ||
174 le16_to_cpu(entry->e_value_offs) + value_size > size)
175 return -EIO;
176 return 0;
179 static int
180 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
181 const char *name, size_t size, int sorted)
183 struct ext4_xattr_entry *entry;
184 size_t name_len;
185 int cmp = 1;
187 if (name == NULL)
188 return -EINVAL;
189 name_len = strlen(name);
190 entry = *pentry;
191 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
192 cmp = name_index - entry->e_name_index;
193 if (!cmp)
194 cmp = name_len - entry->e_name_len;
195 if (!cmp)
196 cmp = memcmp(name, entry->e_name, name_len);
197 if (cmp <= 0 && (sorted || cmp == 0))
198 break;
200 *pentry = entry;
201 if (!cmp && ext4_xattr_check_entry(entry, size))
202 return -EIO;
203 return cmp ? -ENODATA : 0;
206 static int
207 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
208 void *buffer, size_t buffer_size)
210 struct buffer_head *bh = NULL;
211 struct ext4_xattr_entry *entry;
212 size_t size;
213 int error;
215 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
216 name_index, name, buffer, (long)buffer_size);
218 error = -ENODATA;
219 if (!EXT4_I(inode)->i_file_acl)
220 goto cleanup;
221 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
222 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
223 if (!bh)
224 goto cleanup;
225 ea_bdebug(bh, "b_count=%d, refcount=%d",
226 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
227 if (ext4_xattr_check_block(bh)) {
228 bad_block: ext4_error(inode->i_sb, __FUNCTION__,
229 "inode %lu: bad block %llu", inode->i_ino,
230 EXT4_I(inode)->i_file_acl);
231 error = -EIO;
232 goto cleanup;
234 ext4_xattr_cache_insert(bh);
235 entry = BFIRST(bh);
236 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
237 if (error == -EIO)
238 goto bad_block;
239 if (error)
240 goto cleanup;
241 size = le32_to_cpu(entry->e_value_size);
242 if (buffer) {
243 error = -ERANGE;
244 if (size > buffer_size)
245 goto cleanup;
246 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
247 size);
249 error = size;
251 cleanup:
252 brelse(bh);
253 return error;
256 static int
257 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
258 void *buffer, size_t buffer_size)
260 struct ext4_xattr_ibody_header *header;
261 struct ext4_xattr_entry *entry;
262 struct ext4_inode *raw_inode;
263 struct ext4_iloc iloc;
264 size_t size;
265 void *end;
266 int error;
268 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
269 return -ENODATA;
270 error = ext4_get_inode_loc(inode, &iloc);
271 if (error)
272 return error;
273 raw_inode = ext4_raw_inode(&iloc);
274 header = IHDR(inode, raw_inode);
275 entry = IFIRST(header);
276 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
277 error = ext4_xattr_check_names(entry, end);
278 if (error)
279 goto cleanup;
280 error = ext4_xattr_find_entry(&entry, name_index, name,
281 end - (void *)entry, 0);
282 if (error)
283 goto cleanup;
284 size = le32_to_cpu(entry->e_value_size);
285 if (buffer) {
286 error = -ERANGE;
287 if (size > buffer_size)
288 goto cleanup;
289 memcpy(buffer, (void *)IFIRST(header) +
290 le16_to_cpu(entry->e_value_offs), size);
292 error = size;
294 cleanup:
295 brelse(iloc.bh);
296 return error;
300 * ext4_xattr_get()
302 * Copy an extended attribute into the buffer
303 * provided, or compute the buffer size required.
304 * Buffer is NULL to compute the size of the buffer required.
306 * Returns a negative error number on failure, or the number of bytes
307 * used / required on success.
310 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
311 void *buffer, size_t buffer_size)
313 int error;
315 down_read(&EXT4_I(inode)->xattr_sem);
316 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
317 buffer_size);
318 if (error == -ENODATA)
319 error = ext4_xattr_block_get(inode, name_index, name, buffer,
320 buffer_size);
321 up_read(&EXT4_I(inode)->xattr_sem);
322 return error;
325 static int
326 ext4_xattr_list_entries(struct inode *inode, struct ext4_xattr_entry *entry,
327 char *buffer, size_t buffer_size)
329 size_t rest = buffer_size;
331 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
332 struct xattr_handler *handler =
333 ext4_xattr_handler(entry->e_name_index);
335 if (handler) {
336 size_t size = handler->list(inode, buffer, rest,
337 entry->e_name,
338 entry->e_name_len);
339 if (buffer) {
340 if (size > rest)
341 return -ERANGE;
342 buffer += size;
344 rest -= size;
347 return buffer_size - rest;
350 static int
351 ext4_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
353 struct buffer_head *bh = NULL;
354 int error;
356 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
357 buffer, (long)buffer_size);
359 error = 0;
360 if (!EXT4_I(inode)->i_file_acl)
361 goto cleanup;
362 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
363 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
364 error = -EIO;
365 if (!bh)
366 goto cleanup;
367 ea_bdebug(bh, "b_count=%d, refcount=%d",
368 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
369 if (ext4_xattr_check_block(bh)) {
370 ext4_error(inode->i_sb, __FUNCTION__,
371 "inode %lu: bad block %llu", inode->i_ino,
372 EXT4_I(inode)->i_file_acl);
373 error = -EIO;
374 goto cleanup;
376 ext4_xattr_cache_insert(bh);
377 error = ext4_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
379 cleanup:
380 brelse(bh);
382 return error;
385 static int
386 ext4_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
388 struct ext4_xattr_ibody_header *header;
389 struct ext4_inode *raw_inode;
390 struct ext4_iloc iloc;
391 void *end;
392 int error;
394 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
395 return 0;
396 error = ext4_get_inode_loc(inode, &iloc);
397 if (error)
398 return error;
399 raw_inode = ext4_raw_inode(&iloc);
400 header = IHDR(inode, raw_inode);
401 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
402 error = ext4_xattr_check_names(IFIRST(header), end);
403 if (error)
404 goto cleanup;
405 error = ext4_xattr_list_entries(inode, IFIRST(header),
406 buffer, buffer_size);
408 cleanup:
409 brelse(iloc.bh);
410 return error;
414 * ext4_xattr_list()
416 * Copy a list of attribute names into the buffer
417 * provided, or compute the buffer size required.
418 * Buffer is NULL to compute the size of the buffer required.
420 * Returns a negative error number on failure, or the number of bytes
421 * used / required on success.
424 ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
426 int i_error, b_error;
428 down_read(&EXT4_I(inode)->xattr_sem);
429 i_error = ext4_xattr_ibody_list(inode, buffer, buffer_size);
430 if (i_error < 0) {
431 b_error = 0;
432 } else {
433 if (buffer) {
434 buffer += i_error;
435 buffer_size -= i_error;
437 b_error = ext4_xattr_block_list(inode, buffer, buffer_size);
438 if (b_error < 0)
439 i_error = 0;
441 up_read(&EXT4_I(inode)->xattr_sem);
442 return i_error + b_error;
446 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
447 * not set, set it.
449 static void ext4_xattr_update_super_block(handle_t *handle,
450 struct super_block *sb)
452 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
453 return;
455 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
456 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
457 sb->s_dirt = 1;
458 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
463 * Release the xattr block BH: If the reference count is > 1, decrement
464 * it; otherwise free the block.
466 static void
467 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
468 struct buffer_head *bh)
470 struct mb_cache_entry *ce = NULL;
471 int error = 0;
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;
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 ext4_free_blocks(handle, inode, bh->b_blocknr, 1, 1);
484 get_bh(bh);
485 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
486 } else {
487 BHDR(bh)->h_refcount = cpu_to_le32(
488 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
489 error = ext4_journal_dirty_metadata(handle, bh);
490 if (IS_SYNC(inode))
491 handle->h_sync = 1;
492 DQUOT_FREE_BLOCK(inode, 1);
493 ea_bdebug(bh, "refcount now=%d; releasing",
494 le32_to_cpu(BHDR(bh)->h_refcount));
495 if (ce)
496 mb_cache_entry_release(ce);
498 unlock_buffer(bh);
499 out:
500 ext4_std_error(inode->i_sb, error);
501 return;
505 * Find the available free space for EAs. This also returns the total number of
506 * bytes used by EA entries.
508 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
509 size_t *min_offs, void *base, int *total)
511 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
512 *total += EXT4_XATTR_LEN(last->e_name_len);
513 if (!last->e_value_block && last->e_value_size) {
514 size_t offs = le16_to_cpu(last->e_value_offs);
515 if (offs < *min_offs)
516 *min_offs = offs;
519 return (*min_offs - ((void *)last - base) - sizeof(__u32));
522 struct ext4_xattr_info {
523 int name_index;
524 const char *name;
525 const void *value;
526 size_t value_len;
529 struct ext4_xattr_search {
530 struct ext4_xattr_entry *first;
531 void *base;
532 void *end;
533 struct ext4_xattr_entry *here;
534 int not_found;
537 static int
538 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
540 struct ext4_xattr_entry *last;
541 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
543 /* Compute min_offs and last. */
544 last = s->first;
545 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
546 if (!last->e_value_block && last->e_value_size) {
547 size_t offs = le16_to_cpu(last->e_value_offs);
548 if (offs < min_offs)
549 min_offs = offs;
552 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
553 if (!s->not_found) {
554 if (!s->here->e_value_block && s->here->e_value_size) {
555 size_t size = le32_to_cpu(s->here->e_value_size);
556 free += EXT4_XATTR_SIZE(size);
558 free += EXT4_XATTR_LEN(name_len);
560 if (i->value) {
561 if (free < EXT4_XATTR_SIZE(i->value_len) ||
562 free < EXT4_XATTR_LEN(name_len) +
563 EXT4_XATTR_SIZE(i->value_len))
564 return -ENOSPC;
567 if (i->value && s->not_found) {
568 /* Insert the new name. */
569 size_t size = EXT4_XATTR_LEN(name_len);
570 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
571 memmove((void *)s->here + size, s->here, rest);
572 memset(s->here, 0, size);
573 s->here->e_name_index = i->name_index;
574 s->here->e_name_len = name_len;
575 memcpy(s->here->e_name, i->name, name_len);
576 } else {
577 if (!s->here->e_value_block && s->here->e_value_size) {
578 void *first_val = s->base + min_offs;
579 size_t offs = le16_to_cpu(s->here->e_value_offs);
580 void *val = s->base + offs;
581 size_t size = EXT4_XATTR_SIZE(
582 le32_to_cpu(s->here->e_value_size));
584 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
585 /* The old and the new value have the same
586 size. Just replace. */
587 s->here->e_value_size =
588 cpu_to_le32(i->value_len);
589 memset(val + size - EXT4_XATTR_PAD, 0,
590 EXT4_XATTR_PAD); /* Clear pad bytes. */
591 memcpy(val, i->value, i->value_len);
592 return 0;
595 /* Remove the old value. */
596 memmove(first_val + size, first_val, val - first_val);
597 memset(first_val, 0, size);
598 s->here->e_value_size = 0;
599 s->here->e_value_offs = 0;
600 min_offs += size;
602 /* Adjust all value offsets. */
603 last = s->first;
604 while (!IS_LAST_ENTRY(last)) {
605 size_t o = le16_to_cpu(last->e_value_offs);
606 if (!last->e_value_block &&
607 last->e_value_size && o < offs)
608 last->e_value_offs =
609 cpu_to_le16(o + size);
610 last = EXT4_XATTR_NEXT(last);
613 if (!i->value) {
614 /* Remove the old name. */
615 size_t size = EXT4_XATTR_LEN(name_len);
616 last = ENTRY((void *)last - size);
617 memmove(s->here, (void *)s->here + size,
618 (void *)last - (void *)s->here + sizeof(__u32));
619 memset(last, 0, size);
623 if (i->value) {
624 /* Insert the new value. */
625 s->here->e_value_size = cpu_to_le32(i->value_len);
626 if (i->value_len) {
627 size_t size = EXT4_XATTR_SIZE(i->value_len);
628 void *val = s->base + min_offs - size;
629 s->here->e_value_offs = cpu_to_le16(min_offs - size);
630 memset(val + size - EXT4_XATTR_PAD, 0,
631 EXT4_XATTR_PAD); /* Clear the pad bytes. */
632 memcpy(val, i->value, i->value_len);
635 return 0;
638 struct ext4_xattr_block_find {
639 struct ext4_xattr_search s;
640 struct buffer_head *bh;
643 static int
644 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
645 struct ext4_xattr_block_find *bs)
647 struct super_block *sb = inode->i_sb;
648 int error;
650 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
651 i->name_index, i->name, i->value, (long)i->value_len);
653 if (EXT4_I(inode)->i_file_acl) {
654 /* The inode already has an extended attribute block. */
655 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
656 error = -EIO;
657 if (!bs->bh)
658 goto cleanup;
659 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
660 atomic_read(&(bs->bh->b_count)),
661 le32_to_cpu(BHDR(bs->bh)->h_refcount));
662 if (ext4_xattr_check_block(bs->bh)) {
663 ext4_error(sb, __FUNCTION__,
664 "inode %lu: bad block %llu", inode->i_ino,
665 EXT4_I(inode)->i_file_acl);
666 error = -EIO;
667 goto cleanup;
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;
680 error = 0;
682 cleanup:
683 return error;
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)
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;
697 #define header(x) ((struct ext4_xattr_header *)(x))
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);
709 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
710 if (ce) {
711 mb_cache_entry_free(ce);
712 ce = NULL;
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);
722 unlock_buffer(bs->bh);
723 if (error == -EIO)
724 goto bad_block;
725 if (!error)
726 error = ext4_journal_dirty_metadata(handle,
727 bs->bh);
728 if (error)
729 goto cleanup;
730 goto inserted;
731 } else {
732 int offset = (char *)s->here - bs->bh->b_data;
734 unlock_buffer(bs->bh);
735 jbd2_journal_release_buffer(handle, bs->bh);
736 if (ce) {
737 mb_cache_entry_release(ce);
738 ce = NULL;
740 ea_bdebug(bs->bh, "cloning");
741 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
742 error = -ENOMEM;
743 if (s->base == NULL)
744 goto cleanup;
745 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
746 s->first = ENTRY(header(s->base)+1);
747 header(s->base)->h_refcount = cpu_to_le32(1);
748 s->here = ENTRY(s->base + offset);
749 s->end = s->base + bs->bh->b_size;
751 } else {
752 /* Allocate a buffer where we construct the new block. */
753 s->base = kzalloc(sb->s_blocksize, GFP_KERNEL);
754 /* assert(header == s->base) */
755 error = -ENOMEM;
756 if (s->base == NULL)
757 goto cleanup;
758 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
759 header(s->base)->h_blocks = cpu_to_le32(1);
760 header(s->base)->h_refcount = cpu_to_le32(1);
761 s->first = ENTRY(header(s->base)+1);
762 s->here = ENTRY(header(s->base)+1);
763 s->end = s->base + sb->s_blocksize;
766 error = ext4_xattr_set_entry(i, s);
767 if (error == -EIO)
768 goto bad_block;
769 if (error)
770 goto cleanup;
771 if (!IS_LAST_ENTRY(s->first))
772 ext4_xattr_rehash(header(s->base), s->here);
774 inserted:
775 if (!IS_LAST_ENTRY(s->first)) {
776 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
777 if (new_bh) {
778 /* We found an identical block in the cache. */
779 if (new_bh == bs->bh)
780 ea_bdebug(new_bh, "keeping");
781 else {
782 /* The old block is released after updating
783 the inode. */
784 error = -EDQUOT;
785 if (DQUOT_ALLOC_BLOCK(inode, 1))
786 goto cleanup;
787 error = ext4_journal_get_write_access(handle,
788 new_bh);
789 if (error)
790 goto cleanup_dquot;
791 lock_buffer(new_bh);
792 BHDR(new_bh)->h_refcount = cpu_to_le32(1 +
793 le32_to_cpu(BHDR(new_bh)->h_refcount));
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_journal_dirty_metadata(handle,
798 new_bh);
799 if (error)
800 goto cleanup_dquot;
802 mb_cache_entry_release(ce);
803 ce = NULL;
804 } else if (bs->bh && s->base == bs->bh->b_data) {
805 /* We were modifying this block in-place. */
806 ea_bdebug(bs->bh, "keeping this block");
807 new_bh = bs->bh;
808 get_bh(new_bh);
809 } else {
810 /* We need to allocate a new block */
811 ext4_fsblk_t goal = le32_to_cpu(
812 EXT4_SB(sb)->s_es->s_first_data_block) +
813 (ext4_fsblk_t)EXT4_I(inode)->i_block_group *
814 EXT4_BLOCKS_PER_GROUP(sb);
815 ext4_fsblk_t block = ext4_new_block(handle, inode,
816 goal, &error);
817 if (error)
818 goto cleanup;
819 ea_idebug(inode, "creating block %d", block);
821 new_bh = sb_getblk(sb, block);
822 if (!new_bh) {
823 getblk_failed:
824 ext4_free_blocks(handle, inode, block, 1, 1);
825 error = -EIO;
826 goto cleanup;
828 lock_buffer(new_bh);
829 error = ext4_journal_get_create_access(handle, new_bh);
830 if (error) {
831 unlock_buffer(new_bh);
832 goto getblk_failed;
834 memcpy(new_bh->b_data, s->base, new_bh->b_size);
835 set_buffer_uptodate(new_bh);
836 unlock_buffer(new_bh);
837 ext4_xattr_cache_insert(new_bh);
838 error = ext4_journal_dirty_metadata(handle, new_bh);
839 if (error)
840 goto cleanup;
844 /* Update the inode. */
845 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
847 /* Drop the previous xattr block. */
848 if (bs->bh && bs->bh != new_bh)
849 ext4_xattr_release_block(handle, inode, bs->bh);
850 error = 0;
852 cleanup:
853 if (ce)
854 mb_cache_entry_release(ce);
855 brelse(new_bh);
856 if (!(bs->bh && s->base == bs->bh->b_data))
857 kfree(s->base);
859 return error;
861 cleanup_dquot:
862 DQUOT_FREE_BLOCK(inode, 1);
863 goto cleanup;
865 bad_block:
866 ext4_error(inode->i_sb, __FUNCTION__,
867 "inode %lu: bad block %llu", inode->i_ino,
868 EXT4_I(inode)->i_file_acl);
869 goto cleanup;
871 #undef header
874 struct ext4_xattr_ibody_find {
875 struct ext4_xattr_search s;
876 struct ext4_iloc iloc;
879 static int
880 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
881 struct ext4_xattr_ibody_find *is)
883 struct ext4_xattr_ibody_header *header;
884 struct ext4_inode *raw_inode;
885 int error;
887 if (EXT4_I(inode)->i_extra_isize == 0)
888 return 0;
889 raw_inode = ext4_raw_inode(&is->iloc);
890 header = IHDR(inode, raw_inode);
891 is->s.base = is->s.first = IFIRST(header);
892 is->s.here = is->s.first;
893 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
894 if (EXT4_I(inode)->i_state & EXT4_STATE_XATTR) {
895 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
896 if (error)
897 return error;
898 /* Find the named attribute. */
899 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
900 i->name, is->s.end -
901 (void *)is->s.base, 0);
902 if (error && error != -ENODATA)
903 return error;
904 is->s.not_found = error;
906 return 0;
909 static int
910 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
911 struct ext4_xattr_info *i,
912 struct ext4_xattr_ibody_find *is)
914 struct ext4_xattr_ibody_header *header;
915 struct ext4_xattr_search *s = &is->s;
916 int error;
918 if (EXT4_I(inode)->i_extra_isize == 0)
919 return -ENOSPC;
920 error = ext4_xattr_set_entry(i, s);
921 if (error)
922 return error;
923 header = IHDR(inode, ext4_raw_inode(&is->iloc));
924 if (!IS_LAST_ENTRY(s->first)) {
925 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
926 EXT4_I(inode)->i_state |= EXT4_STATE_XATTR;
927 } else {
928 header->h_magic = cpu_to_le32(0);
929 EXT4_I(inode)->i_state &= ~EXT4_STATE_XATTR;
931 return 0;
935 * ext4_xattr_set_handle()
937 * Create, replace or remove an extended attribute for this inode. Buffer
938 * is NULL to remove an existing extended attribute, and non-NULL to
939 * either replace an existing extended attribute, or create a new extended
940 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
941 * specify that an extended attribute must exist and must not exist
942 * previous to the call, respectively.
944 * Returns 0, or a negative error number on failure.
947 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
948 const char *name, const void *value, size_t value_len,
949 int flags)
951 struct ext4_xattr_info i = {
952 .name_index = name_index,
953 .name = name,
954 .value = value,
955 .value_len = value_len,
958 struct ext4_xattr_ibody_find is = {
959 .s = { .not_found = -ENODATA, },
961 struct ext4_xattr_block_find bs = {
962 .s = { .not_found = -ENODATA, },
964 int error;
966 if (!name)
967 return -EINVAL;
968 if (strlen(name) > 255)
969 return -ERANGE;
970 down_write(&EXT4_I(inode)->xattr_sem);
971 error = ext4_get_inode_loc(inode, &is.iloc);
972 if (error)
973 goto cleanup;
975 if (EXT4_I(inode)->i_state & EXT4_STATE_NEW) {
976 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
977 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
978 EXT4_I(inode)->i_state &= ~EXT4_STATE_NEW;
981 error = ext4_xattr_ibody_find(inode, &i, &is);
982 if (error)
983 goto cleanup;
984 if (is.s.not_found)
985 error = ext4_xattr_block_find(inode, &i, &bs);
986 if (error)
987 goto cleanup;
988 if (is.s.not_found && bs.s.not_found) {
989 error = -ENODATA;
990 if (flags & XATTR_REPLACE)
991 goto cleanup;
992 error = 0;
993 if (!value)
994 goto cleanup;
995 } else {
996 error = -EEXIST;
997 if (flags & XATTR_CREATE)
998 goto cleanup;
1000 error = ext4_journal_get_write_access(handle, is.iloc.bh);
1001 if (error)
1002 goto cleanup;
1003 if (!value) {
1004 if (!is.s.not_found)
1005 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1006 else if (!bs.s.not_found)
1007 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1008 } else {
1009 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1010 if (!error && !bs.s.not_found) {
1011 i.value = NULL;
1012 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1013 } else if (error == -ENOSPC) {
1014 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1015 error = ext4_xattr_block_find(inode, &i, &bs);
1016 if (error)
1017 goto cleanup;
1019 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1020 if (error)
1021 goto cleanup;
1022 if (!is.s.not_found) {
1023 i.value = NULL;
1024 error = ext4_xattr_ibody_set(handle, inode, &i,
1025 &is);
1029 if (!error) {
1030 ext4_xattr_update_super_block(handle, inode->i_sb);
1031 inode->i_ctime = ext4_current_time(inode);
1032 if (!value)
1033 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1034 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1036 * The bh is consumed by ext4_mark_iloc_dirty, even with
1037 * error != 0.
1039 is.iloc.bh = NULL;
1040 if (IS_SYNC(inode))
1041 handle->h_sync = 1;
1044 cleanup:
1045 brelse(is.iloc.bh);
1046 brelse(bs.bh);
1047 up_write(&EXT4_I(inode)->xattr_sem);
1048 return error;
1052 * ext4_xattr_set()
1054 * Like ext4_xattr_set_handle, but start from an inode. This extended
1055 * attribute modification is a filesystem transaction by itself.
1057 * Returns 0, or a negative error number on failure.
1060 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1061 const void *value, size_t value_len, int flags)
1063 handle_t *handle;
1064 int error, retries = 0;
1066 retry:
1067 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1068 if (IS_ERR(handle)) {
1069 error = PTR_ERR(handle);
1070 } else {
1071 int error2;
1073 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1074 value, value_len, flags);
1075 error2 = ext4_journal_stop(handle);
1076 if (error == -ENOSPC &&
1077 ext4_should_retry_alloc(inode->i_sb, &retries))
1078 goto retry;
1079 if (error == 0)
1080 error = error2;
1083 return error;
1087 * Shift the EA entries in the inode to create space for the increased
1088 * i_extra_isize.
1090 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1091 int value_offs_shift, void *to,
1092 void *from, size_t n, int blocksize)
1094 struct ext4_xattr_entry *last = entry;
1095 int new_offs;
1097 /* Adjust the value offsets of the entries */
1098 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1099 if (!last->e_value_block && last->e_value_size) {
1100 new_offs = le16_to_cpu(last->e_value_offs) +
1101 value_offs_shift;
1102 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1103 > blocksize);
1104 last->e_value_offs = cpu_to_le16(new_offs);
1107 /* Shift the entries by n bytes */
1108 memmove(to, from, n);
1112 * Expand an inode by new_extra_isize bytes when EAs are present.
1113 * Returns 0 on success or negative error number on failure.
1115 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1116 struct ext4_inode *raw_inode, handle_t *handle)
1118 struct ext4_xattr_ibody_header *header;
1119 struct ext4_xattr_entry *entry, *last, *first;
1120 struct buffer_head *bh = NULL;
1121 struct ext4_xattr_ibody_find *is = NULL;
1122 struct ext4_xattr_block_find *bs = NULL;
1123 char *buffer = NULL, *b_entry_name = NULL;
1124 size_t min_offs, free;
1125 int total_ino, total_blk;
1126 void *base, *start, *end;
1127 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1128 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1130 down_write(&EXT4_I(inode)->xattr_sem);
1131 retry:
1132 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1133 up_write(&EXT4_I(inode)->xattr_sem);
1134 return 0;
1137 header = IHDR(inode, raw_inode);
1138 entry = IFIRST(header);
1141 * Check if enough free space is available in the inode to shift the
1142 * entries ahead by new_extra_isize.
1145 base = start = entry;
1146 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1147 min_offs = end - base;
1148 last = entry;
1149 total_ino = sizeof(struct ext4_xattr_ibody_header);
1151 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1152 if (free >= new_extra_isize) {
1153 entry = IFIRST(header);
1154 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1155 - new_extra_isize, (void *)raw_inode +
1156 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1157 (void *)header, total_ino,
1158 inode->i_sb->s_blocksize);
1159 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1160 error = 0;
1161 goto cleanup;
1165 * Enough free space isn't available in the inode, check if
1166 * EA block can hold new_extra_isize bytes.
1168 if (EXT4_I(inode)->i_file_acl) {
1169 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1170 error = -EIO;
1171 if (!bh)
1172 goto cleanup;
1173 if (ext4_xattr_check_block(bh)) {
1174 ext4_error(inode->i_sb, __FUNCTION__,
1175 "inode %lu: bad block %llu", inode->i_ino,
1176 EXT4_I(inode)->i_file_acl);
1177 error = -EIO;
1178 goto cleanup;
1180 base = BHDR(bh);
1181 first = BFIRST(bh);
1182 end = bh->b_data + bh->b_size;
1183 min_offs = end - base;
1184 free = ext4_xattr_free_space(first, &min_offs, base,
1185 &total_blk);
1186 if (free < new_extra_isize) {
1187 if (!tried_min_extra_isize && s_min_extra_isize) {
1188 tried_min_extra_isize++;
1189 new_extra_isize = s_min_extra_isize;
1190 brelse(bh);
1191 goto retry;
1193 error = -1;
1194 goto cleanup;
1196 } else {
1197 free = inode->i_sb->s_blocksize;
1200 while (new_extra_isize > 0) {
1201 size_t offs, size, entry_size;
1202 struct ext4_xattr_entry *small_entry = NULL;
1203 struct ext4_xattr_info i = {
1204 .value = NULL,
1205 .value_len = 0,
1207 unsigned int total_size; /* EA entry size + value size */
1208 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1209 unsigned int min_total_size = ~0U;
1211 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1212 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1213 if (!is || !bs) {
1214 error = -ENOMEM;
1215 goto cleanup;
1218 is->s.not_found = -ENODATA;
1219 bs->s.not_found = -ENODATA;
1220 is->iloc.bh = NULL;
1221 bs->bh = NULL;
1223 last = IFIRST(header);
1224 /* Find the entry best suited to be pushed into EA block */
1225 entry = NULL;
1226 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1227 total_size =
1228 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1229 EXT4_XATTR_LEN(last->e_name_len);
1230 if (total_size <= free && total_size < min_total_size) {
1231 if (total_size < new_extra_isize) {
1232 small_entry = last;
1233 } else {
1234 entry = last;
1235 min_total_size = total_size;
1240 if (entry == NULL) {
1241 if (small_entry) {
1242 entry = small_entry;
1243 } else {
1244 if (!tried_min_extra_isize &&
1245 s_min_extra_isize) {
1246 tried_min_extra_isize++;
1247 new_extra_isize = s_min_extra_isize;
1248 goto retry;
1250 error = -1;
1251 goto cleanup;
1254 offs = le16_to_cpu(entry->e_value_offs);
1255 size = le32_to_cpu(entry->e_value_size);
1256 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1257 i.name_index = entry->e_name_index,
1258 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1259 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1260 if (!buffer || !b_entry_name) {
1261 error = -ENOMEM;
1262 goto cleanup;
1264 /* Save the entry name and the entry value */
1265 memcpy(buffer, (void *)IFIRST(header) + offs,
1266 EXT4_XATTR_SIZE(size));
1267 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1268 b_entry_name[entry->e_name_len] = '\0';
1269 i.name = b_entry_name;
1271 error = ext4_get_inode_loc(inode, &is->iloc);
1272 if (error)
1273 goto cleanup;
1275 error = ext4_xattr_ibody_find(inode, &i, is);
1276 if (error)
1277 goto cleanup;
1279 /* Remove the chosen entry from the inode */
1280 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1282 entry = IFIRST(header);
1283 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1284 shift_bytes = new_extra_isize;
1285 else
1286 shift_bytes = entry_size + size;
1287 /* Adjust the offsets and shift the remaining entries ahead */
1288 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1289 shift_bytes, (void *)raw_inode +
1290 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1291 (void *)header, total_ino - entry_size,
1292 inode->i_sb->s_blocksize);
1294 extra_isize += shift_bytes;
1295 new_extra_isize -= shift_bytes;
1296 EXT4_I(inode)->i_extra_isize = extra_isize;
1298 i.name = b_entry_name;
1299 i.value = buffer;
1300 i.value_len = size;
1301 error = ext4_xattr_block_find(inode, &i, bs);
1302 if (error)
1303 goto cleanup;
1305 /* Add entry which was removed from the inode into the block */
1306 error = ext4_xattr_block_set(handle, inode, &i, bs);
1307 if (error)
1308 goto cleanup;
1309 kfree(b_entry_name);
1310 kfree(buffer);
1311 brelse(is->iloc.bh);
1312 kfree(is);
1313 kfree(bs);
1315 brelse(bh);
1316 up_write(&EXT4_I(inode)->xattr_sem);
1317 return 0;
1319 cleanup:
1320 kfree(b_entry_name);
1321 kfree(buffer);
1322 if (is)
1323 brelse(is->iloc.bh);
1324 kfree(is);
1325 kfree(bs);
1326 brelse(bh);
1327 up_write(&EXT4_I(inode)->xattr_sem);
1328 return error;
1334 * ext4_xattr_delete_inode()
1336 * Free extended attribute resources associated with this inode. This
1337 * is called immediately before an inode is freed. We have exclusive
1338 * access to the inode.
1340 void
1341 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1343 struct buffer_head *bh = NULL;
1345 if (!EXT4_I(inode)->i_file_acl)
1346 goto cleanup;
1347 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1348 if (!bh) {
1349 ext4_error(inode->i_sb, __FUNCTION__,
1350 "inode %lu: block %llu read error", inode->i_ino,
1351 EXT4_I(inode)->i_file_acl);
1352 goto cleanup;
1354 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1355 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1356 ext4_error(inode->i_sb, __FUNCTION__,
1357 "inode %lu: bad block %llu", inode->i_ino,
1358 EXT4_I(inode)->i_file_acl);
1359 goto cleanup;
1361 ext4_xattr_release_block(handle, inode, bh);
1362 EXT4_I(inode)->i_file_acl = 0;
1364 cleanup:
1365 brelse(bh);
1369 * ext4_xattr_put_super()
1371 * This is called when a file system is unmounted.
1373 void
1374 ext4_xattr_put_super(struct super_block *sb)
1376 mb_cache_shrink(sb->s_bdev);
1380 * ext4_xattr_cache_insert()
1382 * Create a new entry in the extended attribute cache, and insert
1383 * it unless such an entry is already in the cache.
1385 * Returns 0, or a negative error number on failure.
1387 static void
1388 ext4_xattr_cache_insert(struct buffer_head *bh)
1390 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1391 struct mb_cache_entry *ce;
1392 int error;
1394 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1395 if (!ce) {
1396 ea_bdebug(bh, "out of memory");
1397 return;
1399 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1400 if (error) {
1401 mb_cache_entry_free(ce);
1402 if (error == -EBUSY) {
1403 ea_bdebug(bh, "already in cache");
1404 error = 0;
1406 } else {
1407 ea_bdebug(bh, "inserting [%x]", (int)hash);
1408 mb_cache_entry_release(ce);
1413 * ext4_xattr_cmp()
1415 * Compare two extended attribute blocks for equality.
1417 * Returns 0 if the blocks are equal, 1 if they differ, and
1418 * a negative error number on errors.
1420 static int
1421 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1422 struct ext4_xattr_header *header2)
1424 struct ext4_xattr_entry *entry1, *entry2;
1426 entry1 = ENTRY(header1+1);
1427 entry2 = ENTRY(header2+1);
1428 while (!IS_LAST_ENTRY(entry1)) {
1429 if (IS_LAST_ENTRY(entry2))
1430 return 1;
1431 if (entry1->e_hash != entry2->e_hash ||
1432 entry1->e_name_index != entry2->e_name_index ||
1433 entry1->e_name_len != entry2->e_name_len ||
1434 entry1->e_value_size != entry2->e_value_size ||
1435 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1436 return 1;
1437 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1438 return -EIO;
1439 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1440 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1441 le32_to_cpu(entry1->e_value_size)))
1442 return 1;
1444 entry1 = EXT4_XATTR_NEXT(entry1);
1445 entry2 = EXT4_XATTR_NEXT(entry2);
1447 if (!IS_LAST_ENTRY(entry2))
1448 return 1;
1449 return 0;
1453 * ext4_xattr_cache_find()
1455 * Find an identical extended attribute block.
1457 * Returns a pointer to the block found, or NULL if such a block was
1458 * not found or an error occurred.
1460 static struct buffer_head *
1461 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1462 struct mb_cache_entry **pce)
1464 __u32 hash = le32_to_cpu(header->h_hash);
1465 struct mb_cache_entry *ce;
1467 if (!header->h_hash)
1468 return NULL; /* never share */
1469 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1470 again:
1471 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0,
1472 inode->i_sb->s_bdev, hash);
1473 while (ce) {
1474 struct buffer_head *bh;
1476 if (IS_ERR(ce)) {
1477 if (PTR_ERR(ce) == -EAGAIN)
1478 goto again;
1479 break;
1481 bh = sb_bread(inode->i_sb, ce->e_block);
1482 if (!bh) {
1483 ext4_error(inode->i_sb, __FUNCTION__,
1484 "inode %lu: block %lu read error",
1485 inode->i_ino, (unsigned long) ce->e_block);
1486 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1487 EXT4_XATTR_REFCOUNT_MAX) {
1488 ea_idebug(inode, "block %lu refcount %d>=%d",
1489 (unsigned long) ce->e_block,
1490 le32_to_cpu(BHDR(bh)->h_refcount),
1491 EXT4_XATTR_REFCOUNT_MAX);
1492 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1493 *pce = ce;
1494 return bh;
1496 brelse(bh);
1497 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1499 return NULL;
1502 #define NAME_HASH_SHIFT 5
1503 #define VALUE_HASH_SHIFT 16
1506 * ext4_xattr_hash_entry()
1508 * Compute the hash of an extended attribute.
1510 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1511 struct ext4_xattr_entry *entry)
1513 __u32 hash = 0;
1514 char *name = entry->e_name;
1515 int n;
1517 for (n=0; n < entry->e_name_len; n++) {
1518 hash = (hash << NAME_HASH_SHIFT) ^
1519 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1520 *name++;
1523 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1524 __le32 *value = (__le32 *)((char *)header +
1525 le16_to_cpu(entry->e_value_offs));
1526 for (n = (le32_to_cpu(entry->e_value_size) +
1527 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1528 hash = (hash << VALUE_HASH_SHIFT) ^
1529 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1530 le32_to_cpu(*value++);
1533 entry->e_hash = cpu_to_le32(hash);
1536 #undef NAME_HASH_SHIFT
1537 #undef VALUE_HASH_SHIFT
1539 #define BLOCK_HASH_SHIFT 16
1542 * ext4_xattr_rehash()
1544 * Re-compute the extended attribute hash value after an entry has changed.
1546 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1547 struct ext4_xattr_entry *entry)
1549 struct ext4_xattr_entry *here;
1550 __u32 hash = 0;
1552 ext4_xattr_hash_entry(header, entry);
1553 here = ENTRY(header+1);
1554 while (!IS_LAST_ENTRY(here)) {
1555 if (!here->e_hash) {
1556 /* Block is not shared if an entry's hash value == 0 */
1557 hash = 0;
1558 break;
1560 hash = (hash << BLOCK_HASH_SHIFT) ^
1561 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1562 le32_to_cpu(here->e_hash);
1563 here = EXT4_XATTR_NEXT(here);
1565 header->h_hash = cpu_to_le32(hash);
1568 #undef BLOCK_HASH_SHIFT
1570 int __init
1571 init_ext4_xattr(void)
1573 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL,
1574 sizeof(struct mb_cache_entry) +
1575 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1576 if (!ext4_xattr_cache)
1577 return -ENOMEM;
1578 return 0;
1581 void
1582 exit_ext4_xattr(void)
1584 if (ext4_xattr_cache)
1585 mb_cache_destroy(ext4_xattr_cache);
1586 ext4_xattr_cache = NULL;