[NETFILTER]: nf_conntrack: switch protocol registration/unregistration to mutex
[linux-2.6.22.y-op.git] / fs / ext3 / xattr.c
blobf58cbb26323e1ea666d78eba4cb4d3bd4c16490a
1 /*
2 * linux/fs/ext3/xattr.c
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext3 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 * EXT3_I(inode)->i_file_acl is protected by EXT3_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/ext3_jbd.h>
57 #include <linux/ext3_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 ext3_xattr_header *)((bh)->b_data))
65 #define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr))
66 #define BFIRST(bh) ENTRY(BHDR(bh)+1)
67 #define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
69 #define IHDR(inode, raw_inode) \
70 ((struct ext3_xattr_ibody_header *) \
71 ((void *)raw_inode + \
72 EXT3_GOOD_OLD_INODE_SIZE + \
73 EXT3_I(inode)->i_extra_isize))
74 #define IFIRST(hdr) ((struct ext3_xattr_entry *)((hdr)+1))
76 #ifdef EXT3_XATTR_DEBUG
77 # define ea_idebug(inode, f...) do { \
78 printk(KERN_DEBUG "inode %s:%lu: ", \
79 inode->i_sb->s_id, inode->i_ino); \
80 printk(f); \
81 printk("\n"); \
82 } while (0)
83 # define ea_bdebug(bh, f...) do { \
84 char b[BDEVNAME_SIZE]; \
85 printk(KERN_DEBUG "block %s:%lu: ", \
86 bdevname(bh->b_bdev, b), \
87 (unsigned long) bh->b_blocknr); \
88 printk(f); \
89 printk("\n"); \
90 } while (0)
91 #else
92 # define ea_idebug(f...)
93 # define ea_bdebug(f...)
94 #endif
96 static void ext3_xattr_cache_insert(struct buffer_head *);
97 static struct buffer_head *ext3_xattr_cache_find(struct inode *,
98 struct ext3_xattr_header *,
99 struct mb_cache_entry **);
100 static void ext3_xattr_rehash(struct ext3_xattr_header *,
101 struct ext3_xattr_entry *);
103 static struct mb_cache *ext3_xattr_cache;
105 static struct xattr_handler *ext3_xattr_handler_map[] = {
106 [EXT3_XATTR_INDEX_USER] = &ext3_xattr_user_handler,
107 #ifdef CONFIG_EXT3_FS_POSIX_ACL
108 [EXT3_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext3_xattr_acl_access_handler,
109 [EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler,
110 #endif
111 [EXT3_XATTR_INDEX_TRUSTED] = &ext3_xattr_trusted_handler,
112 #ifdef CONFIG_EXT3_FS_SECURITY
113 [EXT3_XATTR_INDEX_SECURITY] = &ext3_xattr_security_handler,
114 #endif
117 struct xattr_handler *ext3_xattr_handlers[] = {
118 &ext3_xattr_user_handler,
119 &ext3_xattr_trusted_handler,
120 #ifdef CONFIG_EXT3_FS_POSIX_ACL
121 &ext3_xattr_acl_access_handler,
122 &ext3_xattr_acl_default_handler,
123 #endif
124 #ifdef CONFIG_EXT3_FS_SECURITY
125 &ext3_xattr_security_handler,
126 #endif
127 NULL
130 static inline struct xattr_handler *
131 ext3_xattr_handler(int name_index)
133 struct xattr_handler *handler = NULL;
135 if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map))
136 handler = ext3_xattr_handler_map[name_index];
137 return handler;
141 * Inode operation listxattr()
143 * dentry->d_inode->i_mutex: don't care
145 ssize_t
146 ext3_listxattr(struct dentry *dentry, char *buffer, size_t size)
148 return ext3_xattr_list(dentry->d_inode, buffer, size);
151 static int
152 ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end)
154 while (!IS_LAST_ENTRY(entry)) {
155 struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry);
156 if ((void *)next >= end)
157 return -EIO;
158 entry = next;
160 return 0;
163 static inline int
164 ext3_xattr_check_block(struct buffer_head *bh)
166 int error;
168 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
169 BHDR(bh)->h_blocks != cpu_to_le32(1))
170 return -EIO;
171 error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
172 return error;
175 static inline int
176 ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size)
178 size_t value_size = le32_to_cpu(entry->e_value_size);
180 if (entry->e_value_block != 0 || value_size > size ||
181 le16_to_cpu(entry->e_value_offs) + value_size > size)
182 return -EIO;
183 return 0;
186 static int
187 ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index,
188 const char *name, size_t size, int sorted)
190 struct ext3_xattr_entry *entry;
191 size_t name_len;
192 int cmp = 1;
194 if (name == NULL)
195 return -EINVAL;
196 name_len = strlen(name);
197 entry = *pentry;
198 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
199 cmp = name_index - entry->e_name_index;
200 if (!cmp)
201 cmp = name_len - entry->e_name_len;
202 if (!cmp)
203 cmp = memcmp(name, entry->e_name, name_len);
204 if (cmp <= 0 && (sorted || cmp == 0))
205 break;
207 *pentry = entry;
208 if (!cmp && ext3_xattr_check_entry(entry, size))
209 return -EIO;
210 return cmp ? -ENODATA : 0;
213 static int
214 ext3_xattr_block_get(struct inode *inode, int name_index, const char *name,
215 void *buffer, size_t buffer_size)
217 struct buffer_head *bh = NULL;
218 struct ext3_xattr_entry *entry;
219 size_t size;
220 int error;
222 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
223 name_index, name, buffer, (long)buffer_size);
225 error = -ENODATA;
226 if (!EXT3_I(inode)->i_file_acl)
227 goto cleanup;
228 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
229 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
230 if (!bh)
231 goto cleanup;
232 ea_bdebug(bh, "b_count=%d, refcount=%d",
233 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
234 if (ext3_xattr_check_block(bh)) {
235 bad_block: ext3_error(inode->i_sb, __FUNCTION__,
236 "inode %lu: bad block "E3FSBLK, inode->i_ino,
237 EXT3_I(inode)->i_file_acl);
238 error = -EIO;
239 goto cleanup;
241 ext3_xattr_cache_insert(bh);
242 entry = BFIRST(bh);
243 error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
244 if (error == -EIO)
245 goto bad_block;
246 if (error)
247 goto cleanup;
248 size = le32_to_cpu(entry->e_value_size);
249 if (buffer) {
250 error = -ERANGE;
251 if (size > buffer_size)
252 goto cleanup;
253 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
254 size);
256 error = size;
258 cleanup:
259 brelse(bh);
260 return error;
263 static int
264 ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
265 void *buffer, size_t buffer_size)
267 struct ext3_xattr_ibody_header *header;
268 struct ext3_xattr_entry *entry;
269 struct ext3_inode *raw_inode;
270 struct ext3_iloc iloc;
271 size_t size;
272 void *end;
273 int error;
275 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
276 return -ENODATA;
277 error = ext3_get_inode_loc(inode, &iloc);
278 if (error)
279 return error;
280 raw_inode = ext3_raw_inode(&iloc);
281 header = IHDR(inode, raw_inode);
282 entry = IFIRST(header);
283 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
284 error = ext3_xattr_check_names(entry, end);
285 if (error)
286 goto cleanup;
287 error = ext3_xattr_find_entry(&entry, name_index, name,
288 end - (void *)entry, 0);
289 if (error)
290 goto cleanup;
291 size = le32_to_cpu(entry->e_value_size);
292 if (buffer) {
293 error = -ERANGE;
294 if (size > buffer_size)
295 goto cleanup;
296 memcpy(buffer, (void *)IFIRST(header) +
297 le16_to_cpu(entry->e_value_offs), size);
299 error = size;
301 cleanup:
302 brelse(iloc.bh);
303 return error;
307 * ext3_xattr_get()
309 * Copy an extended attribute into the buffer
310 * provided, or compute the buffer size required.
311 * Buffer is NULL to compute the size of the buffer required.
313 * Returns a negative error number on failure, or the number of bytes
314 * used / required on success.
317 ext3_xattr_get(struct inode *inode, int name_index, const char *name,
318 void *buffer, size_t buffer_size)
320 int error;
322 down_read(&EXT3_I(inode)->xattr_sem);
323 error = ext3_xattr_ibody_get(inode, name_index, name, buffer,
324 buffer_size);
325 if (error == -ENODATA)
326 error = ext3_xattr_block_get(inode, name_index, name, buffer,
327 buffer_size);
328 up_read(&EXT3_I(inode)->xattr_sem);
329 return error;
332 static int
333 ext3_xattr_list_entries(struct inode *inode, struct ext3_xattr_entry *entry,
334 char *buffer, size_t buffer_size)
336 size_t rest = buffer_size;
338 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
339 struct xattr_handler *handler =
340 ext3_xattr_handler(entry->e_name_index);
342 if (handler) {
343 size_t size = handler->list(inode, buffer, rest,
344 entry->e_name,
345 entry->e_name_len);
346 if (buffer) {
347 if (size > rest)
348 return -ERANGE;
349 buffer += size;
351 rest -= size;
354 return buffer_size - rest;
357 static int
358 ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
360 struct buffer_head *bh = NULL;
361 int error;
363 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
364 buffer, (long)buffer_size);
366 error = 0;
367 if (!EXT3_I(inode)->i_file_acl)
368 goto cleanup;
369 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
370 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
371 error = -EIO;
372 if (!bh)
373 goto cleanup;
374 ea_bdebug(bh, "b_count=%d, refcount=%d",
375 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
376 if (ext3_xattr_check_block(bh)) {
377 ext3_error(inode->i_sb, __FUNCTION__,
378 "inode %lu: bad block "E3FSBLK, inode->i_ino,
379 EXT3_I(inode)->i_file_acl);
380 error = -EIO;
381 goto cleanup;
383 ext3_xattr_cache_insert(bh);
384 error = ext3_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
386 cleanup:
387 brelse(bh);
389 return error;
392 static int
393 ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
395 struct ext3_xattr_ibody_header *header;
396 struct ext3_inode *raw_inode;
397 struct ext3_iloc iloc;
398 void *end;
399 int error;
401 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
402 return 0;
403 error = ext3_get_inode_loc(inode, &iloc);
404 if (error)
405 return error;
406 raw_inode = ext3_raw_inode(&iloc);
407 header = IHDR(inode, raw_inode);
408 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
409 error = ext3_xattr_check_names(IFIRST(header), end);
410 if (error)
411 goto cleanup;
412 error = ext3_xattr_list_entries(inode, IFIRST(header),
413 buffer, buffer_size);
415 cleanup:
416 brelse(iloc.bh);
417 return error;
421 * ext3_xattr_list()
423 * Copy a list of attribute names into the buffer
424 * provided, or compute the buffer size required.
425 * Buffer is NULL to compute the size of the buffer required.
427 * Returns a negative error number on failure, or the number of bytes
428 * used / required on success.
431 ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
433 int i_error, b_error;
435 down_read(&EXT3_I(inode)->xattr_sem);
436 i_error = ext3_xattr_ibody_list(inode, buffer, buffer_size);
437 if (i_error < 0) {
438 b_error = 0;
439 } else {
440 if (buffer) {
441 buffer += i_error;
442 buffer_size -= i_error;
444 b_error = ext3_xattr_block_list(inode, buffer, buffer_size);
445 if (b_error < 0)
446 i_error = 0;
448 up_read(&EXT3_I(inode)->xattr_sem);
449 return i_error + b_error;
453 * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is
454 * not set, set it.
456 static void ext3_xattr_update_super_block(handle_t *handle,
457 struct super_block *sb)
459 if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR))
460 return;
462 if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) {
463 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR);
464 sb->s_dirt = 1;
465 ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh);
470 * Release the xattr block BH: If the reference count is > 1, decrement
471 * it; otherwise free the block.
473 static void
474 ext3_xattr_release_block(handle_t *handle, struct inode *inode,
475 struct buffer_head *bh)
477 struct mb_cache_entry *ce = NULL;
478 int error = 0;
480 ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr);
481 error = ext3_journal_get_write_access(handle, bh);
482 if (error)
483 goto out;
485 lock_buffer(bh);
487 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
488 ea_bdebug(bh, "refcount now=0; freeing");
489 if (ce)
490 mb_cache_entry_free(ce);
491 ext3_free_blocks(handle, inode, bh->b_blocknr, 1);
492 get_bh(bh);
493 ext3_forget(handle, 1, inode, bh, bh->b_blocknr);
494 } else {
495 BHDR(bh)->h_refcount = cpu_to_le32(
496 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
497 error = ext3_journal_dirty_metadata(handle, bh);
498 if (IS_SYNC(inode))
499 handle->h_sync = 1;
500 DQUOT_FREE_BLOCK(inode, 1);
501 ea_bdebug(bh, "refcount now=%d; releasing",
502 le32_to_cpu(BHDR(bh)->h_refcount));
503 if (ce)
504 mb_cache_entry_release(ce);
506 unlock_buffer(bh);
507 out:
508 ext3_std_error(inode->i_sb, error);
509 return;
512 struct ext3_xattr_info {
513 int name_index;
514 const char *name;
515 const void *value;
516 size_t value_len;
519 struct ext3_xattr_search {
520 struct ext3_xattr_entry *first;
521 void *base;
522 void *end;
523 struct ext3_xattr_entry *here;
524 int not_found;
527 static int
528 ext3_xattr_set_entry(struct ext3_xattr_info *i, struct ext3_xattr_search *s)
530 struct ext3_xattr_entry *last;
531 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
533 /* Compute min_offs and last. */
534 last = s->first;
535 for (; !IS_LAST_ENTRY(last); last = EXT3_XATTR_NEXT(last)) {
536 if (!last->e_value_block && last->e_value_size) {
537 size_t offs = le16_to_cpu(last->e_value_offs);
538 if (offs < min_offs)
539 min_offs = offs;
542 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
543 if (!s->not_found) {
544 if (!s->here->e_value_block && s->here->e_value_size) {
545 size_t size = le32_to_cpu(s->here->e_value_size);
546 free += EXT3_XATTR_SIZE(size);
548 free += EXT3_XATTR_LEN(name_len);
550 if (i->value) {
551 if (free < EXT3_XATTR_SIZE(i->value_len) ||
552 free < EXT3_XATTR_LEN(name_len) +
553 EXT3_XATTR_SIZE(i->value_len))
554 return -ENOSPC;
557 if (i->value && s->not_found) {
558 /* Insert the new name. */
559 size_t size = EXT3_XATTR_LEN(name_len);
560 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
561 memmove((void *)s->here + size, s->here, rest);
562 memset(s->here, 0, size);
563 s->here->e_name_index = i->name_index;
564 s->here->e_name_len = name_len;
565 memcpy(s->here->e_name, i->name, name_len);
566 } else {
567 if (!s->here->e_value_block && s->here->e_value_size) {
568 void *first_val = s->base + min_offs;
569 size_t offs = le16_to_cpu(s->here->e_value_offs);
570 void *val = s->base + offs;
571 size_t size = EXT3_XATTR_SIZE(
572 le32_to_cpu(s->here->e_value_size));
574 if (i->value && size == EXT3_XATTR_SIZE(i->value_len)) {
575 /* The old and the new value have the same
576 size. Just replace. */
577 s->here->e_value_size =
578 cpu_to_le32(i->value_len);
579 memset(val + size - EXT3_XATTR_PAD, 0,
580 EXT3_XATTR_PAD); /* Clear pad bytes. */
581 memcpy(val, i->value, i->value_len);
582 return 0;
585 /* Remove the old value. */
586 memmove(first_val + size, first_val, val - first_val);
587 memset(first_val, 0, size);
588 s->here->e_value_size = 0;
589 s->here->e_value_offs = 0;
590 min_offs += size;
592 /* Adjust all value offsets. */
593 last = s->first;
594 while (!IS_LAST_ENTRY(last)) {
595 size_t o = le16_to_cpu(last->e_value_offs);
596 if (!last->e_value_block &&
597 last->e_value_size && o < offs)
598 last->e_value_offs =
599 cpu_to_le16(o + size);
600 last = EXT3_XATTR_NEXT(last);
603 if (!i->value) {
604 /* Remove the old name. */
605 size_t size = EXT3_XATTR_LEN(name_len);
606 last = ENTRY((void *)last - size);
607 memmove(s->here, (void *)s->here + size,
608 (void *)last - (void *)s->here + sizeof(__u32));
609 memset(last, 0, size);
613 if (i->value) {
614 /* Insert the new value. */
615 s->here->e_value_size = cpu_to_le32(i->value_len);
616 if (i->value_len) {
617 size_t size = EXT3_XATTR_SIZE(i->value_len);
618 void *val = s->base + min_offs - size;
619 s->here->e_value_offs = cpu_to_le16(min_offs - size);
620 memset(val + size - EXT3_XATTR_PAD, 0,
621 EXT3_XATTR_PAD); /* Clear the pad bytes. */
622 memcpy(val, i->value, i->value_len);
625 return 0;
628 struct ext3_xattr_block_find {
629 struct ext3_xattr_search s;
630 struct buffer_head *bh;
633 static int
634 ext3_xattr_block_find(struct inode *inode, struct ext3_xattr_info *i,
635 struct ext3_xattr_block_find *bs)
637 struct super_block *sb = inode->i_sb;
638 int error;
640 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
641 i->name_index, i->name, i->value, (long)i->value_len);
643 if (EXT3_I(inode)->i_file_acl) {
644 /* The inode already has an extended attribute block. */
645 bs->bh = sb_bread(sb, EXT3_I(inode)->i_file_acl);
646 error = -EIO;
647 if (!bs->bh)
648 goto cleanup;
649 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
650 atomic_read(&(bs->bh->b_count)),
651 le32_to_cpu(BHDR(bs->bh)->h_refcount));
652 if (ext3_xattr_check_block(bs->bh)) {
653 ext3_error(sb, __FUNCTION__,
654 "inode %lu: bad block "E3FSBLK, inode->i_ino,
655 EXT3_I(inode)->i_file_acl);
656 error = -EIO;
657 goto cleanup;
659 /* Find the named attribute. */
660 bs->s.base = BHDR(bs->bh);
661 bs->s.first = BFIRST(bs->bh);
662 bs->s.end = bs->bh->b_data + bs->bh->b_size;
663 bs->s.here = bs->s.first;
664 error = ext3_xattr_find_entry(&bs->s.here, i->name_index,
665 i->name, bs->bh->b_size, 1);
666 if (error && error != -ENODATA)
667 goto cleanup;
668 bs->s.not_found = error;
670 error = 0;
672 cleanup:
673 return error;
676 static int
677 ext3_xattr_block_set(handle_t *handle, struct inode *inode,
678 struct ext3_xattr_info *i,
679 struct ext3_xattr_block_find *bs)
681 struct super_block *sb = inode->i_sb;
682 struct buffer_head *new_bh = NULL;
683 struct ext3_xattr_search *s = &bs->s;
684 struct mb_cache_entry *ce = NULL;
685 int error = 0;
687 #define header(x) ((struct ext3_xattr_header *)(x))
689 if (i->value && i->value_len > sb->s_blocksize)
690 return -ENOSPC;
691 if (s->base) {
692 ce = mb_cache_entry_get(ext3_xattr_cache, bs->bh->b_bdev,
693 bs->bh->b_blocknr);
694 error = ext3_journal_get_write_access(handle, bs->bh);
695 if (error)
696 goto cleanup;
697 lock_buffer(bs->bh);
699 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
700 if (ce) {
701 mb_cache_entry_free(ce);
702 ce = NULL;
704 ea_bdebug(bs->bh, "modifying in-place");
705 error = ext3_xattr_set_entry(i, s);
706 if (!error) {
707 if (!IS_LAST_ENTRY(s->first))
708 ext3_xattr_rehash(header(s->base),
709 s->here);
710 ext3_xattr_cache_insert(bs->bh);
712 unlock_buffer(bs->bh);
713 if (error == -EIO)
714 goto bad_block;
715 if (!error)
716 error = ext3_journal_dirty_metadata(handle,
717 bs->bh);
718 if (error)
719 goto cleanup;
720 goto inserted;
721 } else {
722 int offset = (char *)s->here - bs->bh->b_data;
724 unlock_buffer(bs->bh);
725 journal_release_buffer(handle, bs->bh);
727 if (ce) {
728 mb_cache_entry_release(ce);
729 ce = NULL;
731 ea_bdebug(bs->bh, "cloning");
732 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
733 error = -ENOMEM;
734 if (s->base == NULL)
735 goto cleanup;
736 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
737 s->first = ENTRY(header(s->base)+1);
738 header(s->base)->h_refcount = cpu_to_le32(1);
739 s->here = ENTRY(s->base + offset);
740 s->end = s->base + bs->bh->b_size;
742 } else {
743 /* Allocate a buffer where we construct the new block. */
744 s->base = kmalloc(sb->s_blocksize, GFP_KERNEL);
745 /* assert(header == s->base) */
746 error = -ENOMEM;
747 if (s->base == NULL)
748 goto cleanup;
749 memset(s->base, 0, sb->s_blocksize);
750 header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
751 header(s->base)->h_blocks = cpu_to_le32(1);
752 header(s->base)->h_refcount = cpu_to_le32(1);
753 s->first = ENTRY(header(s->base)+1);
754 s->here = ENTRY(header(s->base)+1);
755 s->end = s->base + sb->s_blocksize;
758 error = ext3_xattr_set_entry(i, s);
759 if (error == -EIO)
760 goto bad_block;
761 if (error)
762 goto cleanup;
763 if (!IS_LAST_ENTRY(s->first))
764 ext3_xattr_rehash(header(s->base), s->here);
766 inserted:
767 if (!IS_LAST_ENTRY(s->first)) {
768 new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce);
769 if (new_bh) {
770 /* We found an identical block in the cache. */
771 if (new_bh == bs->bh)
772 ea_bdebug(new_bh, "keeping");
773 else {
774 /* The old block is released after updating
775 the inode. */
776 error = -EDQUOT;
777 if (DQUOT_ALLOC_BLOCK(inode, 1))
778 goto cleanup;
779 error = ext3_journal_get_write_access(handle,
780 new_bh);
781 if (error)
782 goto cleanup_dquot;
783 lock_buffer(new_bh);
784 BHDR(new_bh)->h_refcount = cpu_to_le32(1 +
785 le32_to_cpu(BHDR(new_bh)->h_refcount));
786 ea_bdebug(new_bh, "reusing; refcount now=%d",
787 le32_to_cpu(BHDR(new_bh)->h_refcount));
788 unlock_buffer(new_bh);
789 error = ext3_journal_dirty_metadata(handle,
790 new_bh);
791 if (error)
792 goto cleanup_dquot;
794 mb_cache_entry_release(ce);
795 ce = NULL;
796 } else if (bs->bh && s->base == bs->bh->b_data) {
797 /* We were modifying this block in-place. */
798 ea_bdebug(bs->bh, "keeping this block");
799 new_bh = bs->bh;
800 get_bh(new_bh);
801 } else {
802 /* We need to allocate a new block */
803 ext3_fsblk_t goal = le32_to_cpu(
804 EXT3_SB(sb)->s_es->s_first_data_block) +
805 (ext3_fsblk_t)EXT3_I(inode)->i_block_group *
806 EXT3_BLOCKS_PER_GROUP(sb);
807 ext3_fsblk_t block = ext3_new_block(handle, inode,
808 goal, &error);
809 if (error)
810 goto cleanup;
811 ea_idebug(inode, "creating block %d", block);
813 new_bh = sb_getblk(sb, block);
814 if (!new_bh) {
815 getblk_failed:
816 ext3_free_blocks(handle, inode, block, 1);
817 error = -EIO;
818 goto cleanup;
820 lock_buffer(new_bh);
821 error = ext3_journal_get_create_access(handle, new_bh);
822 if (error) {
823 unlock_buffer(new_bh);
824 goto getblk_failed;
826 memcpy(new_bh->b_data, s->base, new_bh->b_size);
827 set_buffer_uptodate(new_bh);
828 unlock_buffer(new_bh);
829 ext3_xattr_cache_insert(new_bh);
830 error = ext3_journal_dirty_metadata(handle, new_bh);
831 if (error)
832 goto cleanup;
836 /* Update the inode. */
837 EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
839 /* Drop the previous xattr block. */
840 if (bs->bh && bs->bh != new_bh)
841 ext3_xattr_release_block(handle, inode, bs->bh);
842 error = 0;
844 cleanup:
845 if (ce)
846 mb_cache_entry_release(ce);
847 brelse(new_bh);
848 if (!(bs->bh && s->base == bs->bh->b_data))
849 kfree(s->base);
851 return error;
853 cleanup_dquot:
854 DQUOT_FREE_BLOCK(inode, 1);
855 goto cleanup;
857 bad_block:
858 ext3_error(inode->i_sb, __FUNCTION__,
859 "inode %lu: bad block "E3FSBLK, inode->i_ino,
860 EXT3_I(inode)->i_file_acl);
861 goto cleanup;
863 #undef header
866 struct ext3_xattr_ibody_find {
867 struct ext3_xattr_search s;
868 struct ext3_iloc iloc;
871 static int
872 ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i,
873 struct ext3_xattr_ibody_find *is)
875 struct ext3_xattr_ibody_header *header;
876 struct ext3_inode *raw_inode;
877 int error;
879 if (EXT3_I(inode)->i_extra_isize == 0)
880 return 0;
881 raw_inode = ext3_raw_inode(&is->iloc);
882 header = IHDR(inode, raw_inode);
883 is->s.base = is->s.first = IFIRST(header);
884 is->s.here = is->s.first;
885 is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
886 if (EXT3_I(inode)->i_state & EXT3_STATE_XATTR) {
887 error = ext3_xattr_check_names(IFIRST(header), is->s.end);
888 if (error)
889 return error;
890 /* Find the named attribute. */
891 error = ext3_xattr_find_entry(&is->s.here, i->name_index,
892 i->name, is->s.end -
893 (void *)is->s.base, 0);
894 if (error && error != -ENODATA)
895 return error;
896 is->s.not_found = error;
898 return 0;
901 static int
902 ext3_xattr_ibody_set(handle_t *handle, struct inode *inode,
903 struct ext3_xattr_info *i,
904 struct ext3_xattr_ibody_find *is)
906 struct ext3_xattr_ibody_header *header;
907 struct ext3_xattr_search *s = &is->s;
908 int error;
910 if (EXT3_I(inode)->i_extra_isize == 0)
911 return -ENOSPC;
912 error = ext3_xattr_set_entry(i, s);
913 if (error)
914 return error;
915 header = IHDR(inode, ext3_raw_inode(&is->iloc));
916 if (!IS_LAST_ENTRY(s->first)) {
917 header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
918 EXT3_I(inode)->i_state |= EXT3_STATE_XATTR;
919 } else {
920 header->h_magic = cpu_to_le32(0);
921 EXT3_I(inode)->i_state &= ~EXT3_STATE_XATTR;
923 return 0;
927 * ext3_xattr_set_handle()
929 * Create, replace or remove an extended attribute for this inode. Buffer
930 * is NULL to remove an existing extended attribute, and non-NULL to
931 * either replace an existing extended attribute, or create a new extended
932 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
933 * specify that an extended attribute must exist and must not exist
934 * previous to the call, respectively.
936 * Returns 0, or a negative error number on failure.
939 ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
940 const char *name, const void *value, size_t value_len,
941 int flags)
943 struct ext3_xattr_info i = {
944 .name_index = name_index,
945 .name = name,
946 .value = value,
947 .value_len = value_len,
950 struct ext3_xattr_ibody_find is = {
951 .s = { .not_found = -ENODATA, },
953 struct ext3_xattr_block_find bs = {
954 .s = { .not_found = -ENODATA, },
956 int error;
958 if (!name)
959 return -EINVAL;
960 if (strlen(name) > 255)
961 return -ERANGE;
962 down_write(&EXT3_I(inode)->xattr_sem);
963 error = ext3_get_inode_loc(inode, &is.iloc);
964 if (error)
965 goto cleanup;
967 if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) {
968 struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc);
969 memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size);
970 EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW;
973 error = ext3_xattr_ibody_find(inode, &i, &is);
974 if (error)
975 goto cleanup;
976 if (is.s.not_found)
977 error = ext3_xattr_block_find(inode, &i, &bs);
978 if (error)
979 goto cleanup;
980 if (is.s.not_found && bs.s.not_found) {
981 error = -ENODATA;
982 if (flags & XATTR_REPLACE)
983 goto cleanup;
984 error = 0;
985 if (!value)
986 goto cleanup;
987 } else {
988 error = -EEXIST;
989 if (flags & XATTR_CREATE)
990 goto cleanup;
992 error = ext3_journal_get_write_access(handle, is.iloc.bh);
993 if (error)
994 goto cleanup;
995 if (!value) {
996 if (!is.s.not_found)
997 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
998 else if (!bs.s.not_found)
999 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1000 } else {
1001 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
1002 if (!error && !bs.s.not_found) {
1003 i.value = NULL;
1004 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1005 } else if (error == -ENOSPC) {
1006 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1007 if (error)
1008 goto cleanup;
1009 if (!is.s.not_found) {
1010 i.value = NULL;
1011 error = ext3_xattr_ibody_set(handle, inode, &i,
1012 &is);
1016 if (!error) {
1017 ext3_xattr_update_super_block(handle, inode->i_sb);
1018 inode->i_ctime = CURRENT_TIME_SEC;
1019 error = ext3_mark_iloc_dirty(handle, inode, &is.iloc);
1021 * The bh is consumed by ext3_mark_iloc_dirty, even with
1022 * error != 0.
1024 is.iloc.bh = NULL;
1025 if (IS_SYNC(inode))
1026 handle->h_sync = 1;
1029 cleanup:
1030 brelse(is.iloc.bh);
1031 brelse(bs.bh);
1032 up_write(&EXT3_I(inode)->xattr_sem);
1033 return error;
1037 * ext3_xattr_set()
1039 * Like ext3_xattr_set_handle, but start from an inode. This extended
1040 * attribute modification is a filesystem transaction by itself.
1042 * Returns 0, or a negative error number on failure.
1045 ext3_xattr_set(struct inode *inode, int name_index, const char *name,
1046 const void *value, size_t value_len, int flags)
1048 handle_t *handle;
1049 int error, retries = 0;
1051 retry:
1052 handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
1053 if (IS_ERR(handle)) {
1054 error = PTR_ERR(handle);
1055 } else {
1056 int error2;
1058 error = ext3_xattr_set_handle(handle, inode, name_index, name,
1059 value, value_len, flags);
1060 error2 = ext3_journal_stop(handle);
1061 if (error == -ENOSPC &&
1062 ext3_should_retry_alloc(inode->i_sb, &retries))
1063 goto retry;
1064 if (error == 0)
1065 error = error2;
1068 return error;
1072 * ext3_xattr_delete_inode()
1074 * Free extended attribute resources associated with this inode. This
1075 * is called immediately before an inode is freed. We have exclusive
1076 * access to the inode.
1078 void
1079 ext3_xattr_delete_inode(handle_t *handle, struct inode *inode)
1081 struct buffer_head *bh = NULL;
1083 if (!EXT3_I(inode)->i_file_acl)
1084 goto cleanup;
1085 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
1086 if (!bh) {
1087 ext3_error(inode->i_sb, __FUNCTION__,
1088 "inode %lu: block "E3FSBLK" read error", inode->i_ino,
1089 EXT3_I(inode)->i_file_acl);
1090 goto cleanup;
1092 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
1093 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1094 ext3_error(inode->i_sb, __FUNCTION__,
1095 "inode %lu: bad block "E3FSBLK, inode->i_ino,
1096 EXT3_I(inode)->i_file_acl);
1097 goto cleanup;
1099 ext3_xattr_release_block(handle, inode, bh);
1100 EXT3_I(inode)->i_file_acl = 0;
1102 cleanup:
1103 brelse(bh);
1107 * ext3_xattr_put_super()
1109 * This is called when a file system is unmounted.
1111 void
1112 ext3_xattr_put_super(struct super_block *sb)
1114 mb_cache_shrink(sb->s_bdev);
1118 * ext3_xattr_cache_insert()
1120 * Create a new entry in the extended attribute cache, and insert
1121 * it unless such an entry is already in the cache.
1123 * Returns 0, or a negative error number on failure.
1125 static void
1126 ext3_xattr_cache_insert(struct buffer_head *bh)
1128 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1129 struct mb_cache_entry *ce;
1130 int error;
1132 ce = mb_cache_entry_alloc(ext3_xattr_cache);
1133 if (!ce) {
1134 ea_bdebug(bh, "out of memory");
1135 return;
1137 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1138 if (error) {
1139 mb_cache_entry_free(ce);
1140 if (error == -EBUSY) {
1141 ea_bdebug(bh, "already in cache");
1142 error = 0;
1144 } else {
1145 ea_bdebug(bh, "inserting [%x]", (int)hash);
1146 mb_cache_entry_release(ce);
1151 * ext3_xattr_cmp()
1153 * Compare two extended attribute blocks for equality.
1155 * Returns 0 if the blocks are equal, 1 if they differ, and
1156 * a negative error number on errors.
1158 static int
1159 ext3_xattr_cmp(struct ext3_xattr_header *header1,
1160 struct ext3_xattr_header *header2)
1162 struct ext3_xattr_entry *entry1, *entry2;
1164 entry1 = ENTRY(header1+1);
1165 entry2 = ENTRY(header2+1);
1166 while (!IS_LAST_ENTRY(entry1)) {
1167 if (IS_LAST_ENTRY(entry2))
1168 return 1;
1169 if (entry1->e_hash != entry2->e_hash ||
1170 entry1->e_name_index != entry2->e_name_index ||
1171 entry1->e_name_len != entry2->e_name_len ||
1172 entry1->e_value_size != entry2->e_value_size ||
1173 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1174 return 1;
1175 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1176 return -EIO;
1177 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1178 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1179 le32_to_cpu(entry1->e_value_size)))
1180 return 1;
1182 entry1 = EXT3_XATTR_NEXT(entry1);
1183 entry2 = EXT3_XATTR_NEXT(entry2);
1185 if (!IS_LAST_ENTRY(entry2))
1186 return 1;
1187 return 0;
1191 * ext3_xattr_cache_find()
1193 * Find an identical extended attribute block.
1195 * Returns a pointer to the block found, or NULL if such a block was
1196 * not found or an error occurred.
1198 static struct buffer_head *
1199 ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header,
1200 struct mb_cache_entry **pce)
1202 __u32 hash = le32_to_cpu(header->h_hash);
1203 struct mb_cache_entry *ce;
1205 if (!header->h_hash)
1206 return NULL; /* never share */
1207 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1208 again:
1209 ce = mb_cache_entry_find_first(ext3_xattr_cache, 0,
1210 inode->i_sb->s_bdev, hash);
1211 while (ce) {
1212 struct buffer_head *bh;
1214 if (IS_ERR(ce)) {
1215 if (PTR_ERR(ce) == -EAGAIN)
1216 goto again;
1217 break;
1219 bh = sb_bread(inode->i_sb, ce->e_block);
1220 if (!bh) {
1221 ext3_error(inode->i_sb, __FUNCTION__,
1222 "inode %lu: block %lu read error",
1223 inode->i_ino, (unsigned long) ce->e_block);
1224 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1225 EXT3_XATTR_REFCOUNT_MAX) {
1226 ea_idebug(inode, "block %lu refcount %d>=%d",
1227 (unsigned long) ce->e_block,
1228 le32_to_cpu(BHDR(bh)->h_refcount),
1229 EXT3_XATTR_REFCOUNT_MAX);
1230 } else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) {
1231 *pce = ce;
1232 return bh;
1234 brelse(bh);
1235 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1237 return NULL;
1240 #define NAME_HASH_SHIFT 5
1241 #define VALUE_HASH_SHIFT 16
1244 * ext3_xattr_hash_entry()
1246 * Compute the hash of an extended attribute.
1248 static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header,
1249 struct ext3_xattr_entry *entry)
1251 __u32 hash = 0;
1252 char *name = entry->e_name;
1253 int n;
1255 for (n=0; n < entry->e_name_len; n++) {
1256 hash = (hash << NAME_HASH_SHIFT) ^
1257 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1258 *name++;
1261 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1262 __le32 *value = (__le32 *)((char *)header +
1263 le16_to_cpu(entry->e_value_offs));
1264 for (n = (le32_to_cpu(entry->e_value_size) +
1265 EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) {
1266 hash = (hash << VALUE_HASH_SHIFT) ^
1267 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1268 le32_to_cpu(*value++);
1271 entry->e_hash = cpu_to_le32(hash);
1274 #undef NAME_HASH_SHIFT
1275 #undef VALUE_HASH_SHIFT
1277 #define BLOCK_HASH_SHIFT 16
1280 * ext3_xattr_rehash()
1282 * Re-compute the extended attribute hash value after an entry has changed.
1284 static void ext3_xattr_rehash(struct ext3_xattr_header *header,
1285 struct ext3_xattr_entry *entry)
1287 struct ext3_xattr_entry *here;
1288 __u32 hash = 0;
1290 ext3_xattr_hash_entry(header, entry);
1291 here = ENTRY(header+1);
1292 while (!IS_LAST_ENTRY(here)) {
1293 if (!here->e_hash) {
1294 /* Block is not shared if an entry's hash value == 0 */
1295 hash = 0;
1296 break;
1298 hash = (hash << BLOCK_HASH_SHIFT) ^
1299 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1300 le32_to_cpu(here->e_hash);
1301 here = EXT3_XATTR_NEXT(here);
1303 header->h_hash = cpu_to_le32(hash);
1306 #undef BLOCK_HASH_SHIFT
1308 int __init
1309 init_ext3_xattr(void)
1311 ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL,
1312 sizeof(struct mb_cache_entry) +
1313 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1314 if (!ext3_xattr_cache)
1315 return -ENOMEM;
1316 return 0;
1319 void
1320 exit_ext3_xattr(void)
1322 if (ext3_xattr_cache)
1323 mb_cache_destroy(ext3_xattr_cache);
1324 ext3_xattr_cache = NULL;