[MIPS] excite: Rename CONFIG option
[linux-2.6/linux-mips/linux-dm7025.git] / fs / ext4 / xattr.c
blobdc969c357aa1d4ef9e207054e506053b8efdc0c6
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 #define IHDR(inode, raw_inode) \
70 ((struct ext4_xattr_ibody_header *) \
71 ((void *)raw_inode + \
72 EXT4_GOOD_OLD_INODE_SIZE + \
73 EXT4_I(inode)->i_extra_isize))
74 #define IFIRST(hdr) ((struct ext4_xattr_entry *)((hdr)+1))
76 #ifdef EXT4_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 ext4_xattr_cache_insert(struct buffer_head *);
97 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
98 struct ext4_xattr_header *,
99 struct mb_cache_entry **);
100 static void ext4_xattr_rehash(struct ext4_xattr_header *,
101 struct ext4_xattr_entry *);
103 static struct mb_cache *ext4_xattr_cache;
105 static struct xattr_handler *ext4_xattr_handler_map[] = {
106 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
107 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
108 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
109 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
110 #endif
111 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
112 #ifdef CONFIG_EXT4DEV_FS_SECURITY
113 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
114 #endif
117 struct xattr_handler *ext4_xattr_handlers[] = {
118 &ext4_xattr_user_handler,
119 &ext4_xattr_trusted_handler,
120 #ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
121 &ext4_xattr_acl_access_handler,
122 &ext4_xattr_acl_default_handler,
123 #endif
124 #ifdef CONFIG_EXT4DEV_FS_SECURITY
125 &ext4_xattr_security_handler,
126 #endif
127 NULL
130 static inline struct xattr_handler *
131 ext4_xattr_handler(int name_index)
133 struct xattr_handler *handler = NULL;
135 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
136 handler = ext4_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 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
148 return ext4_xattr_list(dentry->d_inode, buffer, size);
151 static int
152 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
154 while (!IS_LAST_ENTRY(entry)) {
155 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
156 if ((void *)next >= end)
157 return -EIO;
158 entry = next;
160 return 0;
163 static inline int
164 ext4_xattr_check_block(struct buffer_head *bh)
166 int error;
168 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
169 BHDR(bh)->h_blocks != cpu_to_le32(1))
170 return -EIO;
171 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
172 return error;
175 static inline int
176 ext4_xattr_check_entry(struct ext4_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 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
188 const char *name, size_t size, int sorted)
190 struct ext4_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 = EXT4_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 && ext4_xattr_check_entry(entry, size))
209 return -EIO;
210 return cmp ? -ENODATA : 0;
213 static int
214 ext4_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 ext4_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 (!EXT4_I(inode)->i_file_acl)
227 goto cleanup;
228 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
229 bh = sb_bread(inode->i_sb, EXT4_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 (ext4_xattr_check_block(bh)) {
235 bad_block: ext4_error(inode->i_sb, __FUNCTION__,
236 "inode %lu: bad block %llu", inode->i_ino,
237 EXT4_I(inode)->i_file_acl);
238 error = -EIO;
239 goto cleanup;
241 ext4_xattr_cache_insert(bh);
242 entry = BFIRST(bh);
243 error = ext4_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 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
265 void *buffer, size_t buffer_size)
267 struct ext4_xattr_ibody_header *header;
268 struct ext4_xattr_entry *entry;
269 struct ext4_inode *raw_inode;
270 struct ext4_iloc iloc;
271 size_t size;
272 void *end;
273 int error;
275 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
276 return -ENODATA;
277 error = ext4_get_inode_loc(inode, &iloc);
278 if (error)
279 return error;
280 raw_inode = ext4_raw_inode(&iloc);
281 header = IHDR(inode, raw_inode);
282 entry = IFIRST(header);
283 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
284 error = ext4_xattr_check_names(entry, end);
285 if (error)
286 goto cleanup;
287 error = ext4_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 * ext4_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 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
318 void *buffer, size_t buffer_size)
320 int error;
322 down_read(&EXT4_I(inode)->xattr_sem);
323 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
324 buffer_size);
325 if (error == -ENODATA)
326 error = ext4_xattr_block_get(inode, name_index, name, buffer,
327 buffer_size);
328 up_read(&EXT4_I(inode)->xattr_sem);
329 return error;
332 static int
333 ext4_xattr_list_entries(struct inode *inode, struct ext4_xattr_entry *entry,
334 char *buffer, size_t buffer_size)
336 size_t rest = buffer_size;
338 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
339 struct xattr_handler *handler =
340 ext4_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 ext4_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 (!EXT4_I(inode)->i_file_acl)
368 goto cleanup;
369 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
370 bh = sb_bread(inode->i_sb, EXT4_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 (ext4_xattr_check_block(bh)) {
377 ext4_error(inode->i_sb, __FUNCTION__,
378 "inode %lu: bad block %llu", inode->i_ino,
379 EXT4_I(inode)->i_file_acl);
380 error = -EIO;
381 goto cleanup;
383 ext4_xattr_cache_insert(bh);
384 error = ext4_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
386 cleanup:
387 brelse(bh);
389 return error;
392 static int
393 ext4_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
395 struct ext4_xattr_ibody_header *header;
396 struct ext4_inode *raw_inode;
397 struct ext4_iloc iloc;
398 void *end;
399 int error;
401 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
402 return 0;
403 error = ext4_get_inode_loc(inode, &iloc);
404 if (error)
405 return error;
406 raw_inode = ext4_raw_inode(&iloc);
407 header = IHDR(inode, raw_inode);
408 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
409 error = ext4_xattr_check_names(IFIRST(header), end);
410 if (error)
411 goto cleanup;
412 error = ext4_xattr_list_entries(inode, IFIRST(header),
413 buffer, buffer_size);
415 cleanup:
416 brelse(iloc.bh);
417 return error;
421 * ext4_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 ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
433 int i_error, b_error;
435 down_read(&EXT4_I(inode)->xattr_sem);
436 i_error = ext4_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 = ext4_xattr_block_list(inode, buffer, buffer_size);
445 if (b_error < 0)
446 i_error = 0;
448 up_read(&EXT4_I(inode)->xattr_sem);
449 return i_error + b_error;
453 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
454 * not set, set it.
456 static void ext4_xattr_update_super_block(handle_t *handle,
457 struct super_block *sb)
459 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
460 return;
462 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
463 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
464 sb->s_dirt = 1;
465 ext4_journal_dirty_metadata(handle, EXT4_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 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
475 struct buffer_head *bh)
477 struct mb_cache_entry *ce = NULL;
479 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
480 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
481 ea_bdebug(bh, "refcount now=0; freeing");
482 if (ce)
483 mb_cache_entry_free(ce);
484 ext4_free_blocks(handle, inode, bh->b_blocknr, 1);
485 get_bh(bh);
486 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
487 } else {
488 if (ext4_journal_get_write_access(handle, bh) == 0) {
489 lock_buffer(bh);
490 BHDR(bh)->h_refcount = cpu_to_le32(
491 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
492 ext4_journal_dirty_metadata(handle, bh);
493 if (IS_SYNC(inode))
494 handle->h_sync = 1;
495 DQUOT_FREE_BLOCK(inode, 1);
496 unlock_buffer(bh);
497 ea_bdebug(bh, "refcount now=%d; releasing",
498 le32_to_cpu(BHDR(bh)->h_refcount));
500 if (ce)
501 mb_cache_entry_release(ce);
505 struct ext4_xattr_info {
506 int name_index;
507 const char *name;
508 const void *value;
509 size_t value_len;
512 struct ext4_xattr_search {
513 struct ext4_xattr_entry *first;
514 void *base;
515 void *end;
516 struct ext4_xattr_entry *here;
517 int not_found;
520 static int
521 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
523 struct ext4_xattr_entry *last;
524 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
526 /* Compute min_offs and last. */
527 last = s->first;
528 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
529 if (!last->e_value_block && last->e_value_size) {
530 size_t offs = le16_to_cpu(last->e_value_offs);
531 if (offs < min_offs)
532 min_offs = offs;
535 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
536 if (!s->not_found) {
537 if (!s->here->e_value_block && s->here->e_value_size) {
538 size_t size = le32_to_cpu(s->here->e_value_size);
539 free += EXT4_XATTR_SIZE(size);
541 free += EXT4_XATTR_LEN(name_len);
543 if (i->value) {
544 if (free < EXT4_XATTR_SIZE(i->value_len) ||
545 free < EXT4_XATTR_LEN(name_len) +
546 EXT4_XATTR_SIZE(i->value_len))
547 return -ENOSPC;
550 if (i->value && s->not_found) {
551 /* Insert the new name. */
552 size_t size = EXT4_XATTR_LEN(name_len);
553 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
554 memmove((void *)s->here + size, s->here, rest);
555 memset(s->here, 0, size);
556 s->here->e_name_index = i->name_index;
557 s->here->e_name_len = name_len;
558 memcpy(s->here->e_name, i->name, name_len);
559 } else {
560 if (!s->here->e_value_block && s->here->e_value_size) {
561 void *first_val = s->base + min_offs;
562 size_t offs = le16_to_cpu(s->here->e_value_offs);
563 void *val = s->base + offs;
564 size_t size = EXT4_XATTR_SIZE(
565 le32_to_cpu(s->here->e_value_size));
567 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
568 /* The old and the new value have the same
569 size. Just replace. */
570 s->here->e_value_size =
571 cpu_to_le32(i->value_len);
572 memset(val + size - EXT4_XATTR_PAD, 0,
573 EXT4_XATTR_PAD); /* Clear pad bytes. */
574 memcpy(val, i->value, i->value_len);
575 return 0;
578 /* Remove the old value. */
579 memmove(first_val + size, first_val, val - first_val);
580 memset(first_val, 0, size);
581 s->here->e_value_size = 0;
582 s->here->e_value_offs = 0;
583 min_offs += size;
585 /* Adjust all value offsets. */
586 last = s->first;
587 while (!IS_LAST_ENTRY(last)) {
588 size_t o = le16_to_cpu(last->e_value_offs);
589 if (!last->e_value_block &&
590 last->e_value_size && o < offs)
591 last->e_value_offs =
592 cpu_to_le16(o + size);
593 last = EXT4_XATTR_NEXT(last);
596 if (!i->value) {
597 /* Remove the old name. */
598 size_t size = EXT4_XATTR_LEN(name_len);
599 last = ENTRY((void *)last - size);
600 memmove(s->here, (void *)s->here + size,
601 (void *)last - (void *)s->here + sizeof(__u32));
602 memset(last, 0, size);
606 if (i->value) {
607 /* Insert the new value. */
608 s->here->e_value_size = cpu_to_le32(i->value_len);
609 if (i->value_len) {
610 size_t size = EXT4_XATTR_SIZE(i->value_len);
611 void *val = s->base + min_offs - size;
612 s->here->e_value_offs = cpu_to_le16(min_offs - size);
613 memset(val + size - EXT4_XATTR_PAD, 0,
614 EXT4_XATTR_PAD); /* Clear the pad bytes. */
615 memcpy(val, i->value, i->value_len);
618 return 0;
621 struct ext4_xattr_block_find {
622 struct ext4_xattr_search s;
623 struct buffer_head *bh;
626 static int
627 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
628 struct ext4_xattr_block_find *bs)
630 struct super_block *sb = inode->i_sb;
631 int error;
633 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
634 i->name_index, i->name, i->value, (long)i->value_len);
636 if (EXT4_I(inode)->i_file_acl) {
637 /* The inode already has an extended attribute block. */
638 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
639 error = -EIO;
640 if (!bs->bh)
641 goto cleanup;
642 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
643 atomic_read(&(bs->bh->b_count)),
644 le32_to_cpu(BHDR(bs->bh)->h_refcount));
645 if (ext4_xattr_check_block(bs->bh)) {
646 ext4_error(sb, __FUNCTION__,
647 "inode %lu: bad block %llu", inode->i_ino,
648 EXT4_I(inode)->i_file_acl);
649 error = -EIO;
650 goto cleanup;
652 /* Find the named attribute. */
653 bs->s.base = BHDR(bs->bh);
654 bs->s.first = BFIRST(bs->bh);
655 bs->s.end = bs->bh->b_data + bs->bh->b_size;
656 bs->s.here = bs->s.first;
657 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
658 i->name, bs->bh->b_size, 1);
659 if (error && error != -ENODATA)
660 goto cleanup;
661 bs->s.not_found = error;
663 error = 0;
665 cleanup:
666 return error;
669 static int
670 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
671 struct ext4_xattr_info *i,
672 struct ext4_xattr_block_find *bs)
674 struct super_block *sb = inode->i_sb;
675 struct buffer_head *new_bh = NULL;
676 struct ext4_xattr_search *s = &bs->s;
677 struct mb_cache_entry *ce = NULL;
678 int error;
680 #define header(x) ((struct ext4_xattr_header *)(x))
682 if (i->value && i->value_len > sb->s_blocksize)
683 return -ENOSPC;
684 if (s->base) {
685 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
686 bs->bh->b_blocknr);
687 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
688 if (ce) {
689 mb_cache_entry_free(ce);
690 ce = NULL;
692 ea_bdebug(bs->bh, "modifying in-place");
693 error = ext4_journal_get_write_access(handle, bs->bh);
694 if (error)
695 goto cleanup;
696 lock_buffer(bs->bh);
697 error = ext4_xattr_set_entry(i, s);
698 if (!error) {
699 if (!IS_LAST_ENTRY(s->first))
700 ext4_xattr_rehash(header(s->base),
701 s->here);
702 ext4_xattr_cache_insert(bs->bh);
704 unlock_buffer(bs->bh);
705 if (error == -EIO)
706 goto bad_block;
707 if (!error)
708 error = ext4_journal_dirty_metadata(handle,
709 bs->bh);
710 if (error)
711 goto cleanup;
712 goto inserted;
713 } else {
714 int offset = (char *)s->here - bs->bh->b_data;
716 if (ce) {
717 mb_cache_entry_release(ce);
718 ce = NULL;
720 ea_bdebug(bs->bh, "cloning");
721 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
722 error = -ENOMEM;
723 if (s->base == NULL)
724 goto cleanup;
725 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
726 s->first = ENTRY(header(s->base)+1);
727 header(s->base)->h_refcount = cpu_to_le32(1);
728 s->here = ENTRY(s->base + offset);
729 s->end = s->base + bs->bh->b_size;
731 } else {
732 /* Allocate a buffer where we construct the new block. */
733 s->base = kmalloc(sb->s_blocksize, GFP_KERNEL);
734 /* assert(header == s->base) */
735 error = -ENOMEM;
736 if (s->base == NULL)
737 goto cleanup;
738 memset(s->base, 0, sb->s_blocksize);
739 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
740 header(s->base)->h_blocks = cpu_to_le32(1);
741 header(s->base)->h_refcount = cpu_to_le32(1);
742 s->first = ENTRY(header(s->base)+1);
743 s->here = ENTRY(header(s->base)+1);
744 s->end = s->base + sb->s_blocksize;
747 error = ext4_xattr_set_entry(i, s);
748 if (error == -EIO)
749 goto bad_block;
750 if (error)
751 goto cleanup;
752 if (!IS_LAST_ENTRY(s->first))
753 ext4_xattr_rehash(header(s->base), s->here);
755 inserted:
756 if (!IS_LAST_ENTRY(s->first)) {
757 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
758 if (new_bh) {
759 /* We found an identical block in the cache. */
760 if (new_bh == bs->bh)
761 ea_bdebug(new_bh, "keeping");
762 else {
763 /* The old block is released after updating
764 the inode. */
765 error = -EDQUOT;
766 if (DQUOT_ALLOC_BLOCK(inode, 1))
767 goto cleanup;
768 error = ext4_journal_get_write_access(handle,
769 new_bh);
770 if (error)
771 goto cleanup_dquot;
772 lock_buffer(new_bh);
773 BHDR(new_bh)->h_refcount = cpu_to_le32(1 +
774 le32_to_cpu(BHDR(new_bh)->h_refcount));
775 ea_bdebug(new_bh, "reusing; refcount now=%d",
776 le32_to_cpu(BHDR(new_bh)->h_refcount));
777 unlock_buffer(new_bh);
778 error = ext4_journal_dirty_metadata(handle,
779 new_bh);
780 if (error)
781 goto cleanup_dquot;
783 mb_cache_entry_release(ce);
784 ce = NULL;
785 } else if (bs->bh && s->base == bs->bh->b_data) {
786 /* We were modifying this block in-place. */
787 ea_bdebug(bs->bh, "keeping this block");
788 new_bh = bs->bh;
789 get_bh(new_bh);
790 } else {
791 /* We need to allocate a new block */
792 ext4_fsblk_t goal = le32_to_cpu(
793 EXT4_SB(sb)->s_es->s_first_data_block) +
794 (ext4_fsblk_t)EXT4_I(inode)->i_block_group *
795 EXT4_BLOCKS_PER_GROUP(sb);
796 ext4_fsblk_t block = ext4_new_block(handle, inode,
797 goal, &error);
798 if (error)
799 goto cleanup;
800 ea_idebug(inode, "creating block %d", block);
802 new_bh = sb_getblk(sb, block);
803 if (!new_bh) {
804 getblk_failed:
805 ext4_free_blocks(handle, inode, block, 1);
806 error = -EIO;
807 goto cleanup;
809 lock_buffer(new_bh);
810 error = ext4_journal_get_create_access(handle, new_bh);
811 if (error) {
812 unlock_buffer(new_bh);
813 goto getblk_failed;
815 memcpy(new_bh->b_data, s->base, new_bh->b_size);
816 set_buffer_uptodate(new_bh);
817 unlock_buffer(new_bh);
818 ext4_xattr_cache_insert(new_bh);
819 error = ext4_journal_dirty_metadata(handle, new_bh);
820 if (error)
821 goto cleanup;
825 /* Update the inode. */
826 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
828 /* Drop the previous xattr block. */
829 if (bs->bh && bs->bh != new_bh)
830 ext4_xattr_release_block(handle, inode, bs->bh);
831 error = 0;
833 cleanup:
834 if (ce)
835 mb_cache_entry_release(ce);
836 brelse(new_bh);
837 if (!(bs->bh && s->base == bs->bh->b_data))
838 kfree(s->base);
840 return error;
842 cleanup_dquot:
843 DQUOT_FREE_BLOCK(inode, 1);
844 goto cleanup;
846 bad_block:
847 ext4_error(inode->i_sb, __FUNCTION__,
848 "inode %lu: bad block %llu", inode->i_ino,
849 EXT4_I(inode)->i_file_acl);
850 goto cleanup;
852 #undef header
855 struct ext4_xattr_ibody_find {
856 struct ext4_xattr_search s;
857 struct ext4_iloc iloc;
860 static int
861 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
862 struct ext4_xattr_ibody_find *is)
864 struct ext4_xattr_ibody_header *header;
865 struct ext4_inode *raw_inode;
866 int error;
868 if (EXT4_I(inode)->i_extra_isize == 0)
869 return 0;
870 raw_inode = ext4_raw_inode(&is->iloc);
871 header = IHDR(inode, raw_inode);
872 is->s.base = is->s.first = IFIRST(header);
873 is->s.here = is->s.first;
874 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
875 if (EXT4_I(inode)->i_state & EXT4_STATE_XATTR) {
876 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
877 if (error)
878 return error;
879 /* Find the named attribute. */
880 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
881 i->name, is->s.end -
882 (void *)is->s.base, 0);
883 if (error && error != -ENODATA)
884 return error;
885 is->s.not_found = error;
887 return 0;
890 static int
891 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
892 struct ext4_xattr_info *i,
893 struct ext4_xattr_ibody_find *is)
895 struct ext4_xattr_ibody_header *header;
896 struct ext4_xattr_search *s = &is->s;
897 int error;
899 if (EXT4_I(inode)->i_extra_isize == 0)
900 return -ENOSPC;
901 error = ext4_xattr_set_entry(i, s);
902 if (error)
903 return error;
904 header = IHDR(inode, ext4_raw_inode(&is->iloc));
905 if (!IS_LAST_ENTRY(s->first)) {
906 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
907 EXT4_I(inode)->i_state |= EXT4_STATE_XATTR;
908 } else {
909 header->h_magic = cpu_to_le32(0);
910 EXT4_I(inode)->i_state &= ~EXT4_STATE_XATTR;
912 return 0;
916 * ext4_xattr_set_handle()
918 * Create, replace or remove an extended attribute for this inode. Buffer
919 * is NULL to remove an existing extended attribute, and non-NULL to
920 * either replace an existing extended attribute, or create a new extended
921 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
922 * specify that an extended attribute must exist and must not exist
923 * previous to the call, respectively.
925 * Returns 0, or a negative error number on failure.
928 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
929 const char *name, const void *value, size_t value_len,
930 int flags)
932 struct ext4_xattr_info i = {
933 .name_index = name_index,
934 .name = name,
935 .value = value,
936 .value_len = value_len,
939 struct ext4_xattr_ibody_find is = {
940 .s = { .not_found = -ENODATA, },
942 struct ext4_xattr_block_find bs = {
943 .s = { .not_found = -ENODATA, },
945 int error;
947 if (!name)
948 return -EINVAL;
949 if (strlen(name) > 255)
950 return -ERANGE;
951 down_write(&EXT4_I(inode)->xattr_sem);
952 error = ext4_get_inode_loc(inode, &is.iloc);
953 if (error)
954 goto cleanup;
956 if (EXT4_I(inode)->i_state & EXT4_STATE_NEW) {
957 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
958 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
959 EXT4_I(inode)->i_state &= ~EXT4_STATE_NEW;
962 error = ext4_xattr_ibody_find(inode, &i, &is);
963 if (error)
964 goto cleanup;
965 if (is.s.not_found)
966 error = ext4_xattr_block_find(inode, &i, &bs);
967 if (error)
968 goto cleanup;
969 if (is.s.not_found && bs.s.not_found) {
970 error = -ENODATA;
971 if (flags & XATTR_REPLACE)
972 goto cleanup;
973 error = 0;
974 if (!value)
975 goto cleanup;
976 } else {
977 error = -EEXIST;
978 if (flags & XATTR_CREATE)
979 goto cleanup;
981 error = ext4_journal_get_write_access(handle, is.iloc.bh);
982 if (error)
983 goto cleanup;
984 if (!value) {
985 if (!is.s.not_found)
986 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
987 else if (!bs.s.not_found)
988 error = ext4_xattr_block_set(handle, inode, &i, &bs);
989 } else {
990 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
991 if (!error && !bs.s.not_found) {
992 i.value = NULL;
993 error = ext4_xattr_block_set(handle, inode, &i, &bs);
994 } else if (error == -ENOSPC) {
995 error = ext4_xattr_block_set(handle, inode, &i, &bs);
996 if (error)
997 goto cleanup;
998 if (!is.s.not_found) {
999 i.value = NULL;
1000 error = ext4_xattr_ibody_set(handle, inode, &i,
1001 &is);
1005 if (!error) {
1006 ext4_xattr_update_super_block(handle, inode->i_sb);
1007 inode->i_ctime = CURRENT_TIME_SEC;
1008 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1010 * The bh is consumed by ext4_mark_iloc_dirty, even with
1011 * error != 0.
1013 is.iloc.bh = NULL;
1014 if (IS_SYNC(inode))
1015 handle->h_sync = 1;
1018 cleanup:
1019 brelse(is.iloc.bh);
1020 brelse(bs.bh);
1021 up_write(&EXT4_I(inode)->xattr_sem);
1022 return error;
1026 * ext4_xattr_set()
1028 * Like ext4_xattr_set_handle, but start from an inode. This extended
1029 * attribute modification is a filesystem transaction by itself.
1031 * Returns 0, or a negative error number on failure.
1034 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1035 const void *value, size_t value_len, int flags)
1037 handle_t *handle;
1038 int error, retries = 0;
1040 retry:
1041 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1042 if (IS_ERR(handle)) {
1043 error = PTR_ERR(handle);
1044 } else {
1045 int error2;
1047 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1048 value, value_len, flags);
1049 error2 = ext4_journal_stop(handle);
1050 if (error == -ENOSPC &&
1051 ext4_should_retry_alloc(inode->i_sb, &retries))
1052 goto retry;
1053 if (error == 0)
1054 error = error2;
1057 return error;
1061 * ext4_xattr_delete_inode()
1063 * Free extended attribute resources associated with this inode. This
1064 * is called immediately before an inode is freed. We have exclusive
1065 * access to the inode.
1067 void
1068 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1070 struct buffer_head *bh = NULL;
1072 if (!EXT4_I(inode)->i_file_acl)
1073 goto cleanup;
1074 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1075 if (!bh) {
1076 ext4_error(inode->i_sb, __FUNCTION__,
1077 "inode %lu: block %llu read error", inode->i_ino,
1078 EXT4_I(inode)->i_file_acl);
1079 goto cleanup;
1081 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1082 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1083 ext4_error(inode->i_sb, __FUNCTION__,
1084 "inode %lu: bad block %llu", inode->i_ino,
1085 EXT4_I(inode)->i_file_acl);
1086 goto cleanup;
1088 ext4_xattr_release_block(handle, inode, bh);
1089 EXT4_I(inode)->i_file_acl = 0;
1091 cleanup:
1092 brelse(bh);
1096 * ext4_xattr_put_super()
1098 * This is called when a file system is unmounted.
1100 void
1101 ext4_xattr_put_super(struct super_block *sb)
1103 mb_cache_shrink(sb->s_bdev);
1107 * ext4_xattr_cache_insert()
1109 * Create a new entry in the extended attribute cache, and insert
1110 * it unless such an entry is already in the cache.
1112 * Returns 0, or a negative error number on failure.
1114 static void
1115 ext4_xattr_cache_insert(struct buffer_head *bh)
1117 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1118 struct mb_cache_entry *ce;
1119 int error;
1121 ce = mb_cache_entry_alloc(ext4_xattr_cache);
1122 if (!ce) {
1123 ea_bdebug(bh, "out of memory");
1124 return;
1126 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1127 if (error) {
1128 mb_cache_entry_free(ce);
1129 if (error == -EBUSY) {
1130 ea_bdebug(bh, "already in cache");
1131 error = 0;
1133 } else {
1134 ea_bdebug(bh, "inserting [%x]", (int)hash);
1135 mb_cache_entry_release(ce);
1140 * ext4_xattr_cmp()
1142 * Compare two extended attribute blocks for equality.
1144 * Returns 0 if the blocks are equal, 1 if they differ, and
1145 * a negative error number on errors.
1147 static int
1148 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1149 struct ext4_xattr_header *header2)
1151 struct ext4_xattr_entry *entry1, *entry2;
1153 entry1 = ENTRY(header1+1);
1154 entry2 = ENTRY(header2+1);
1155 while (!IS_LAST_ENTRY(entry1)) {
1156 if (IS_LAST_ENTRY(entry2))
1157 return 1;
1158 if (entry1->e_hash != entry2->e_hash ||
1159 entry1->e_name_index != entry2->e_name_index ||
1160 entry1->e_name_len != entry2->e_name_len ||
1161 entry1->e_value_size != entry2->e_value_size ||
1162 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1163 return 1;
1164 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1165 return -EIO;
1166 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1167 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1168 le32_to_cpu(entry1->e_value_size)))
1169 return 1;
1171 entry1 = EXT4_XATTR_NEXT(entry1);
1172 entry2 = EXT4_XATTR_NEXT(entry2);
1174 if (!IS_LAST_ENTRY(entry2))
1175 return 1;
1176 return 0;
1180 * ext4_xattr_cache_find()
1182 * Find an identical extended attribute block.
1184 * Returns a pointer to the block found, or NULL if such a block was
1185 * not found or an error occurred.
1187 static struct buffer_head *
1188 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1189 struct mb_cache_entry **pce)
1191 __u32 hash = le32_to_cpu(header->h_hash);
1192 struct mb_cache_entry *ce;
1194 if (!header->h_hash)
1195 return NULL; /* never share */
1196 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1197 again:
1198 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0,
1199 inode->i_sb->s_bdev, hash);
1200 while (ce) {
1201 struct buffer_head *bh;
1203 if (IS_ERR(ce)) {
1204 if (PTR_ERR(ce) == -EAGAIN)
1205 goto again;
1206 break;
1208 bh = sb_bread(inode->i_sb, ce->e_block);
1209 if (!bh) {
1210 ext4_error(inode->i_sb, __FUNCTION__,
1211 "inode %lu: block %lu read error",
1212 inode->i_ino, (unsigned long) ce->e_block);
1213 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1214 EXT4_XATTR_REFCOUNT_MAX) {
1215 ea_idebug(inode, "block %lu refcount %d>=%d",
1216 (unsigned long) ce->e_block,
1217 le32_to_cpu(BHDR(bh)->h_refcount),
1218 EXT4_XATTR_REFCOUNT_MAX);
1219 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1220 *pce = ce;
1221 return bh;
1223 brelse(bh);
1224 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1226 return NULL;
1229 #define NAME_HASH_SHIFT 5
1230 #define VALUE_HASH_SHIFT 16
1233 * ext4_xattr_hash_entry()
1235 * Compute the hash of an extended attribute.
1237 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1238 struct ext4_xattr_entry *entry)
1240 __u32 hash = 0;
1241 char *name = entry->e_name;
1242 int n;
1244 for (n=0; n < entry->e_name_len; n++) {
1245 hash = (hash << NAME_HASH_SHIFT) ^
1246 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1247 *name++;
1250 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1251 __le32 *value = (__le32 *)((char *)header +
1252 le16_to_cpu(entry->e_value_offs));
1253 for (n = (le32_to_cpu(entry->e_value_size) +
1254 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1255 hash = (hash << VALUE_HASH_SHIFT) ^
1256 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1257 le32_to_cpu(*value++);
1260 entry->e_hash = cpu_to_le32(hash);
1263 #undef NAME_HASH_SHIFT
1264 #undef VALUE_HASH_SHIFT
1266 #define BLOCK_HASH_SHIFT 16
1269 * ext4_xattr_rehash()
1271 * Re-compute the extended attribute hash value after an entry has changed.
1273 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1274 struct ext4_xattr_entry *entry)
1276 struct ext4_xattr_entry *here;
1277 __u32 hash = 0;
1279 ext4_xattr_hash_entry(header, entry);
1280 here = ENTRY(header+1);
1281 while (!IS_LAST_ENTRY(here)) {
1282 if (!here->e_hash) {
1283 /* Block is not shared if an entry's hash value == 0 */
1284 hash = 0;
1285 break;
1287 hash = (hash << BLOCK_HASH_SHIFT) ^
1288 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1289 le32_to_cpu(here->e_hash);
1290 here = EXT4_XATTR_NEXT(here);
1292 header->h_hash = cpu_to_le32(hash);
1295 #undef BLOCK_HASH_SHIFT
1297 int __init
1298 init_ext4_xattr(void)
1300 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL,
1301 sizeof(struct mb_cache_entry) +
1302 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1303 if (!ext4_xattr_cache)
1304 return -ENOMEM;
1305 return 0;
1308 void
1309 exit_ext4_xattr(void)
1311 if (ext4_xattr_cache)
1312 mb_cache_destroy(ext4_xattr_cache);
1313 ext4_xattr_cache = NULL;