x86/amd-iommu: Add per IOMMU reference counting
[linux/fpc-iii.git] / fs / ext3 / xattr.c
blob545e37c4b91eb6387ddc015e17e000ad1ccbaaa3
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 *);
102 static int ext3_xattr_list(struct inode *inode, char *buffer,
103 size_t buffer_size);
105 static struct mb_cache *ext3_xattr_cache;
107 static struct xattr_handler *ext3_xattr_handler_map[] = {
108 [EXT3_XATTR_INDEX_USER] = &ext3_xattr_user_handler,
109 #ifdef CONFIG_EXT3_FS_POSIX_ACL
110 [EXT3_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext3_xattr_acl_access_handler,
111 [EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler,
112 #endif
113 [EXT3_XATTR_INDEX_TRUSTED] = &ext3_xattr_trusted_handler,
114 #ifdef CONFIG_EXT3_FS_SECURITY
115 [EXT3_XATTR_INDEX_SECURITY] = &ext3_xattr_security_handler,
116 #endif
119 struct xattr_handler *ext3_xattr_handlers[] = {
120 &ext3_xattr_user_handler,
121 &ext3_xattr_trusted_handler,
122 #ifdef CONFIG_EXT3_FS_POSIX_ACL
123 &ext3_xattr_acl_access_handler,
124 &ext3_xattr_acl_default_handler,
125 #endif
126 #ifdef CONFIG_EXT3_FS_SECURITY
127 &ext3_xattr_security_handler,
128 #endif
129 NULL
132 static inline struct xattr_handler *
133 ext3_xattr_handler(int name_index)
135 struct xattr_handler *handler = NULL;
137 if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map))
138 handler = ext3_xattr_handler_map[name_index];
139 return handler;
143 * Inode operation listxattr()
145 * dentry->d_inode->i_mutex: don't care
147 ssize_t
148 ext3_listxattr(struct dentry *dentry, char *buffer, size_t size)
150 return ext3_xattr_list(dentry->d_inode, buffer, size);
153 static int
154 ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end)
156 while (!IS_LAST_ENTRY(entry)) {
157 struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry);
158 if ((void *)next >= end)
159 return -EIO;
160 entry = next;
162 return 0;
165 static inline int
166 ext3_xattr_check_block(struct buffer_head *bh)
168 int error;
170 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
171 BHDR(bh)->h_blocks != cpu_to_le32(1))
172 return -EIO;
173 error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
174 return error;
177 static inline int
178 ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size)
180 size_t value_size = le32_to_cpu(entry->e_value_size);
182 if (entry->e_value_block != 0 || value_size > size ||
183 le16_to_cpu(entry->e_value_offs) + value_size > size)
184 return -EIO;
185 return 0;
188 static int
189 ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index,
190 const char *name, size_t size, int sorted)
192 struct ext3_xattr_entry *entry;
193 size_t name_len;
194 int cmp = 1;
196 if (name == NULL)
197 return -EINVAL;
198 name_len = strlen(name);
199 entry = *pentry;
200 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
201 cmp = name_index - entry->e_name_index;
202 if (!cmp)
203 cmp = name_len - entry->e_name_len;
204 if (!cmp)
205 cmp = memcmp(name, entry->e_name, name_len);
206 if (cmp <= 0 && (sorted || cmp == 0))
207 break;
209 *pentry = entry;
210 if (!cmp && ext3_xattr_check_entry(entry, size))
211 return -EIO;
212 return cmp ? -ENODATA : 0;
215 static int
216 ext3_xattr_block_get(struct inode *inode, int name_index, const char *name,
217 void *buffer, size_t buffer_size)
219 struct buffer_head *bh = NULL;
220 struct ext3_xattr_entry *entry;
221 size_t size;
222 int error;
224 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
225 name_index, name, buffer, (long)buffer_size);
227 error = -ENODATA;
228 if (!EXT3_I(inode)->i_file_acl)
229 goto cleanup;
230 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
231 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
232 if (!bh)
233 goto cleanup;
234 ea_bdebug(bh, "b_count=%d, refcount=%d",
235 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
236 if (ext3_xattr_check_block(bh)) {
237 bad_block: ext3_error(inode->i_sb, __func__,
238 "inode %lu: bad block "E3FSBLK, inode->i_ino,
239 EXT3_I(inode)->i_file_acl);
240 error = -EIO;
241 goto cleanup;
243 ext3_xattr_cache_insert(bh);
244 entry = BFIRST(bh);
245 error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
246 if (error == -EIO)
247 goto bad_block;
248 if (error)
249 goto cleanup;
250 size = le32_to_cpu(entry->e_value_size);
251 if (buffer) {
252 error = -ERANGE;
253 if (size > buffer_size)
254 goto cleanup;
255 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
256 size);
258 error = size;
260 cleanup:
261 brelse(bh);
262 return error;
265 static int
266 ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
267 void *buffer, size_t buffer_size)
269 struct ext3_xattr_ibody_header *header;
270 struct ext3_xattr_entry *entry;
271 struct ext3_inode *raw_inode;
272 struct ext3_iloc iloc;
273 size_t size;
274 void *end;
275 int error;
277 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
278 return -ENODATA;
279 error = ext3_get_inode_loc(inode, &iloc);
280 if (error)
281 return error;
282 raw_inode = ext3_raw_inode(&iloc);
283 header = IHDR(inode, raw_inode);
284 entry = IFIRST(header);
285 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
286 error = ext3_xattr_check_names(entry, end);
287 if (error)
288 goto cleanup;
289 error = ext3_xattr_find_entry(&entry, name_index, name,
290 end - (void *)entry, 0);
291 if (error)
292 goto cleanup;
293 size = le32_to_cpu(entry->e_value_size);
294 if (buffer) {
295 error = -ERANGE;
296 if (size > buffer_size)
297 goto cleanup;
298 memcpy(buffer, (void *)IFIRST(header) +
299 le16_to_cpu(entry->e_value_offs), size);
301 error = size;
303 cleanup:
304 brelse(iloc.bh);
305 return error;
309 * ext3_xattr_get()
311 * Copy an extended attribute into the buffer
312 * provided, or compute the buffer size required.
313 * Buffer is NULL to compute the size of the buffer required.
315 * Returns a negative error number on failure, or the number of bytes
316 * used / required on success.
319 ext3_xattr_get(struct inode *inode, int name_index, const char *name,
320 void *buffer, size_t buffer_size)
322 int error;
324 down_read(&EXT3_I(inode)->xattr_sem);
325 error = ext3_xattr_ibody_get(inode, name_index, name, buffer,
326 buffer_size);
327 if (error == -ENODATA)
328 error = ext3_xattr_block_get(inode, name_index, name, buffer,
329 buffer_size);
330 up_read(&EXT3_I(inode)->xattr_sem);
331 return error;
334 static int
335 ext3_xattr_list_entries(struct inode *inode, struct ext3_xattr_entry *entry,
336 char *buffer, size_t buffer_size)
338 size_t rest = buffer_size;
340 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
341 struct xattr_handler *handler =
342 ext3_xattr_handler(entry->e_name_index);
344 if (handler) {
345 size_t size = handler->list(inode, buffer, rest,
346 entry->e_name,
347 entry->e_name_len);
348 if (buffer) {
349 if (size > rest)
350 return -ERANGE;
351 buffer += size;
353 rest -= size;
356 return buffer_size - rest;
359 static int
360 ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
362 struct buffer_head *bh = NULL;
363 int error;
365 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
366 buffer, (long)buffer_size);
368 error = 0;
369 if (!EXT3_I(inode)->i_file_acl)
370 goto cleanup;
371 ea_idebug(inode, "reading block %u", EXT3_I(inode)->i_file_acl);
372 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
373 error = -EIO;
374 if (!bh)
375 goto cleanup;
376 ea_bdebug(bh, "b_count=%d, refcount=%d",
377 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
378 if (ext3_xattr_check_block(bh)) {
379 ext3_error(inode->i_sb, __func__,
380 "inode %lu: bad block "E3FSBLK, inode->i_ino,
381 EXT3_I(inode)->i_file_acl);
382 error = -EIO;
383 goto cleanup;
385 ext3_xattr_cache_insert(bh);
386 error = ext3_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
388 cleanup:
389 brelse(bh);
391 return error;
394 static int
395 ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
397 struct ext3_xattr_ibody_header *header;
398 struct ext3_inode *raw_inode;
399 struct ext3_iloc iloc;
400 void *end;
401 int error;
403 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
404 return 0;
405 error = ext3_get_inode_loc(inode, &iloc);
406 if (error)
407 return error;
408 raw_inode = ext3_raw_inode(&iloc);
409 header = IHDR(inode, raw_inode);
410 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
411 error = ext3_xattr_check_names(IFIRST(header), end);
412 if (error)
413 goto cleanup;
414 error = ext3_xattr_list_entries(inode, IFIRST(header),
415 buffer, buffer_size);
417 cleanup:
418 brelse(iloc.bh);
419 return error;
423 * ext3_xattr_list()
425 * Copy a list of attribute names into the buffer
426 * provided, or compute the buffer size required.
427 * Buffer is NULL to compute the size of the buffer required.
429 * Returns a negative error number on failure, or the number of bytes
430 * used / required on success.
432 static int
433 ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
435 int i_error, b_error;
437 down_read(&EXT3_I(inode)->xattr_sem);
438 i_error = ext3_xattr_ibody_list(inode, buffer, buffer_size);
439 if (i_error < 0) {
440 b_error = 0;
441 } else {
442 if (buffer) {
443 buffer += i_error;
444 buffer_size -= i_error;
446 b_error = ext3_xattr_block_list(inode, buffer, buffer_size);
447 if (b_error < 0)
448 i_error = 0;
450 up_read(&EXT3_I(inode)->xattr_sem);
451 return i_error + b_error;
455 * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is
456 * not set, set it.
458 static void ext3_xattr_update_super_block(handle_t *handle,
459 struct super_block *sb)
461 if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR))
462 return;
464 if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) {
465 EXT3_SET_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR);
466 ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh);
471 * Release the xattr block BH: If the reference count is > 1, decrement
472 * it; otherwise free the block.
474 static void
475 ext3_xattr_release_block(handle_t *handle, struct inode *inode,
476 struct buffer_head *bh)
478 struct mb_cache_entry *ce = NULL;
479 int error = 0;
481 ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr);
482 error = ext3_journal_get_write_access(handle, bh);
483 if (error)
484 goto out;
486 lock_buffer(bh);
488 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
489 ea_bdebug(bh, "refcount now=0; freeing");
490 if (ce)
491 mb_cache_entry_free(ce);
492 ext3_free_blocks(handle, inode, bh->b_blocknr, 1);
493 get_bh(bh);
494 ext3_forget(handle, 1, inode, bh, bh->b_blocknr);
495 } else {
496 le32_add_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 vfs_dq_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, __func__,
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_NOFS);
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 = kzalloc(sb->s_blocksize, GFP_NOFS);
745 /* assert(header == s->base) */
746 error = -ENOMEM;
747 if (s->base == NULL)
748 goto cleanup;
749 header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
750 header(s->base)->h_blocks = cpu_to_le32(1);
751 header(s->base)->h_refcount = cpu_to_le32(1);
752 s->first = ENTRY(header(s->base)+1);
753 s->here = ENTRY(header(s->base)+1);
754 s->end = s->base + sb->s_blocksize;
757 error = ext3_xattr_set_entry(i, s);
758 if (error == -EIO)
759 goto bad_block;
760 if (error)
761 goto cleanup;
762 if (!IS_LAST_ENTRY(s->first))
763 ext3_xattr_rehash(header(s->base), s->here);
765 inserted:
766 if (!IS_LAST_ENTRY(s->first)) {
767 new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce);
768 if (new_bh) {
769 /* We found an identical block in the cache. */
770 if (new_bh == bs->bh)
771 ea_bdebug(new_bh, "keeping");
772 else {
773 /* The old block is released after updating
774 the inode. */
775 error = -EDQUOT;
776 if (vfs_dq_alloc_block(inode, 1))
777 goto cleanup;
778 error = ext3_journal_get_write_access(handle,
779 new_bh);
780 if (error)
781 goto cleanup_dquot;
782 lock_buffer(new_bh);
783 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
784 ea_bdebug(new_bh, "reusing; refcount now=%d",
785 le32_to_cpu(BHDR(new_bh)->h_refcount));
786 unlock_buffer(new_bh);
787 error = ext3_journal_dirty_metadata(handle,
788 new_bh);
789 if (error)
790 goto cleanup_dquot;
792 mb_cache_entry_release(ce);
793 ce = NULL;
794 } else if (bs->bh && s->base == bs->bh->b_data) {
795 /* We were modifying this block in-place. */
796 ea_bdebug(bs->bh, "keeping this block");
797 new_bh = bs->bh;
798 get_bh(new_bh);
799 } else {
800 /* We need to allocate a new block */
801 ext3_fsblk_t goal = ext3_group_first_block_no(sb,
802 EXT3_I(inode)->i_block_group);
803 ext3_fsblk_t block = ext3_new_block(handle, inode,
804 goal, &error);
805 if (error)
806 goto cleanup;
807 ea_idebug(inode, "creating block %d", block);
809 new_bh = sb_getblk(sb, block);
810 if (!new_bh) {
811 getblk_failed:
812 ext3_free_blocks(handle, inode, block, 1);
813 error = -EIO;
814 goto cleanup;
816 lock_buffer(new_bh);
817 error = ext3_journal_get_create_access(handle, new_bh);
818 if (error) {
819 unlock_buffer(new_bh);
820 goto getblk_failed;
822 memcpy(new_bh->b_data, s->base, new_bh->b_size);
823 set_buffer_uptodate(new_bh);
824 unlock_buffer(new_bh);
825 ext3_xattr_cache_insert(new_bh);
826 error = ext3_journal_dirty_metadata(handle, new_bh);
827 if (error)
828 goto cleanup;
832 /* Update the inode. */
833 EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
835 /* Drop the previous xattr block. */
836 if (bs->bh && bs->bh != new_bh)
837 ext3_xattr_release_block(handle, inode, bs->bh);
838 error = 0;
840 cleanup:
841 if (ce)
842 mb_cache_entry_release(ce);
843 brelse(new_bh);
844 if (!(bs->bh && s->base == bs->bh->b_data))
845 kfree(s->base);
847 return error;
849 cleanup_dquot:
850 vfs_dq_free_block(inode, 1);
851 goto cleanup;
853 bad_block:
854 ext3_error(inode->i_sb, __func__,
855 "inode %lu: bad block "E3FSBLK, inode->i_ino,
856 EXT3_I(inode)->i_file_acl);
857 goto cleanup;
859 #undef header
862 struct ext3_xattr_ibody_find {
863 struct ext3_xattr_search s;
864 struct ext3_iloc iloc;
867 static int
868 ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i,
869 struct ext3_xattr_ibody_find *is)
871 struct ext3_xattr_ibody_header *header;
872 struct ext3_inode *raw_inode;
873 int error;
875 if (EXT3_I(inode)->i_extra_isize == 0)
876 return 0;
877 raw_inode = ext3_raw_inode(&is->iloc);
878 header = IHDR(inode, raw_inode);
879 is->s.base = is->s.first = IFIRST(header);
880 is->s.here = is->s.first;
881 is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
882 if (EXT3_I(inode)->i_state & EXT3_STATE_XATTR) {
883 error = ext3_xattr_check_names(IFIRST(header), is->s.end);
884 if (error)
885 return error;
886 /* Find the named attribute. */
887 error = ext3_xattr_find_entry(&is->s.here, i->name_index,
888 i->name, is->s.end -
889 (void *)is->s.base, 0);
890 if (error && error != -ENODATA)
891 return error;
892 is->s.not_found = error;
894 return 0;
897 static int
898 ext3_xattr_ibody_set(handle_t *handle, struct inode *inode,
899 struct ext3_xattr_info *i,
900 struct ext3_xattr_ibody_find *is)
902 struct ext3_xattr_ibody_header *header;
903 struct ext3_xattr_search *s = &is->s;
904 int error;
906 if (EXT3_I(inode)->i_extra_isize == 0)
907 return -ENOSPC;
908 error = ext3_xattr_set_entry(i, s);
909 if (error)
910 return error;
911 header = IHDR(inode, ext3_raw_inode(&is->iloc));
912 if (!IS_LAST_ENTRY(s->first)) {
913 header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
914 EXT3_I(inode)->i_state |= EXT3_STATE_XATTR;
915 } else {
916 header->h_magic = cpu_to_le32(0);
917 EXT3_I(inode)->i_state &= ~EXT3_STATE_XATTR;
919 return 0;
923 * ext3_xattr_set_handle()
925 * Create, replace or remove an extended attribute for this inode. Buffer
926 * is NULL to remove an existing extended attribute, and non-NULL to
927 * either replace an existing extended attribute, or create a new extended
928 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
929 * specify that an extended attribute must exist and must not exist
930 * previous to the call, respectively.
932 * Returns 0, or a negative error number on failure.
935 ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
936 const char *name, const void *value, size_t value_len,
937 int flags)
939 struct ext3_xattr_info i = {
940 .name_index = name_index,
941 .name = name,
942 .value = value,
943 .value_len = value_len,
946 struct ext3_xattr_ibody_find is = {
947 .s = { .not_found = -ENODATA, },
949 struct ext3_xattr_block_find bs = {
950 .s = { .not_found = -ENODATA, },
952 int error;
954 if (!name)
955 return -EINVAL;
956 if (strlen(name) > 255)
957 return -ERANGE;
958 down_write(&EXT3_I(inode)->xattr_sem);
959 error = ext3_get_inode_loc(inode, &is.iloc);
960 if (error)
961 goto cleanup;
963 if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) {
964 struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc);
965 memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size);
966 EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW;
969 error = ext3_xattr_ibody_find(inode, &i, &is);
970 if (error)
971 goto cleanup;
972 if (is.s.not_found)
973 error = ext3_xattr_block_find(inode, &i, &bs);
974 if (error)
975 goto cleanup;
976 if (is.s.not_found && bs.s.not_found) {
977 error = -ENODATA;
978 if (flags & XATTR_REPLACE)
979 goto cleanup;
980 error = 0;
981 if (!value)
982 goto cleanup;
983 } else {
984 error = -EEXIST;
985 if (flags & XATTR_CREATE)
986 goto cleanup;
988 error = ext3_journal_get_write_access(handle, is.iloc.bh);
989 if (error)
990 goto cleanup;
991 if (!value) {
992 if (!is.s.not_found)
993 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
994 else if (!bs.s.not_found)
995 error = ext3_xattr_block_set(handle, inode, &i, &bs);
996 } else {
997 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
998 if (!error && !bs.s.not_found) {
999 i.value = NULL;
1000 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1001 } else if (error == -ENOSPC) {
1002 if (EXT3_I(inode)->i_file_acl && !bs.s.base) {
1003 error = ext3_xattr_block_find(inode, &i, &bs);
1004 if (error)
1005 goto cleanup;
1007 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1008 if (error)
1009 goto cleanup;
1010 if (!is.s.not_found) {
1011 i.value = NULL;
1012 error = ext3_xattr_ibody_set(handle, inode, &i,
1013 &is);
1017 if (!error) {
1018 ext3_xattr_update_super_block(handle, inode->i_sb);
1019 inode->i_ctime = CURRENT_TIME_SEC;
1020 error = ext3_mark_iloc_dirty(handle, inode, &is.iloc);
1022 * The bh is consumed by ext3_mark_iloc_dirty, even with
1023 * error != 0.
1025 is.iloc.bh = NULL;
1026 if (IS_SYNC(inode))
1027 handle->h_sync = 1;
1030 cleanup:
1031 brelse(is.iloc.bh);
1032 brelse(bs.bh);
1033 up_write(&EXT3_I(inode)->xattr_sem);
1034 return error;
1038 * ext3_xattr_set()
1040 * Like ext3_xattr_set_handle, but start from an inode. This extended
1041 * attribute modification is a filesystem transaction by itself.
1043 * Returns 0, or a negative error number on failure.
1046 ext3_xattr_set(struct inode *inode, int name_index, const char *name,
1047 const void *value, size_t value_len, int flags)
1049 handle_t *handle;
1050 int error, retries = 0;
1052 retry:
1053 handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS(inode->i_sb));
1054 if (IS_ERR(handle)) {
1055 error = PTR_ERR(handle);
1056 } else {
1057 int error2;
1059 error = ext3_xattr_set_handle(handle, inode, name_index, name,
1060 value, value_len, flags);
1061 error2 = ext3_journal_stop(handle);
1062 if (error == -ENOSPC &&
1063 ext3_should_retry_alloc(inode->i_sb, &retries))
1064 goto retry;
1065 if (error == 0)
1066 error = error2;
1069 return error;
1073 * ext3_xattr_delete_inode()
1075 * Free extended attribute resources associated with this inode. This
1076 * is called immediately before an inode is freed. We have exclusive
1077 * access to the inode.
1079 void
1080 ext3_xattr_delete_inode(handle_t *handle, struct inode *inode)
1082 struct buffer_head *bh = NULL;
1084 if (!EXT3_I(inode)->i_file_acl)
1085 goto cleanup;
1086 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
1087 if (!bh) {
1088 ext3_error(inode->i_sb, __func__,
1089 "inode %lu: block "E3FSBLK" read error", inode->i_ino,
1090 EXT3_I(inode)->i_file_acl);
1091 goto cleanup;
1093 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
1094 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1095 ext3_error(inode->i_sb, __func__,
1096 "inode %lu: bad block "E3FSBLK, inode->i_ino,
1097 EXT3_I(inode)->i_file_acl);
1098 goto cleanup;
1100 ext3_xattr_release_block(handle, inode, bh);
1101 EXT3_I(inode)->i_file_acl = 0;
1103 cleanup:
1104 brelse(bh);
1108 * ext3_xattr_put_super()
1110 * This is called when a file system is unmounted.
1112 void
1113 ext3_xattr_put_super(struct super_block *sb)
1115 mb_cache_shrink(sb->s_bdev);
1119 * ext3_xattr_cache_insert()
1121 * Create a new entry in the extended attribute cache, and insert
1122 * it unless such an entry is already in the cache.
1124 * Returns 0, or a negative error number on failure.
1126 static void
1127 ext3_xattr_cache_insert(struct buffer_head *bh)
1129 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1130 struct mb_cache_entry *ce;
1131 int error;
1133 ce = mb_cache_entry_alloc(ext3_xattr_cache, GFP_NOFS);
1134 if (!ce) {
1135 ea_bdebug(bh, "out of memory");
1136 return;
1138 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1139 if (error) {
1140 mb_cache_entry_free(ce);
1141 if (error == -EBUSY) {
1142 ea_bdebug(bh, "already in cache");
1143 error = 0;
1145 } else {
1146 ea_bdebug(bh, "inserting [%x]", (int)hash);
1147 mb_cache_entry_release(ce);
1152 * ext3_xattr_cmp()
1154 * Compare two extended attribute blocks for equality.
1156 * Returns 0 if the blocks are equal, 1 if they differ, and
1157 * a negative error number on errors.
1159 static int
1160 ext3_xattr_cmp(struct ext3_xattr_header *header1,
1161 struct ext3_xattr_header *header2)
1163 struct ext3_xattr_entry *entry1, *entry2;
1165 entry1 = ENTRY(header1+1);
1166 entry2 = ENTRY(header2+1);
1167 while (!IS_LAST_ENTRY(entry1)) {
1168 if (IS_LAST_ENTRY(entry2))
1169 return 1;
1170 if (entry1->e_hash != entry2->e_hash ||
1171 entry1->e_name_index != entry2->e_name_index ||
1172 entry1->e_name_len != entry2->e_name_len ||
1173 entry1->e_value_size != entry2->e_value_size ||
1174 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1175 return 1;
1176 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1177 return -EIO;
1178 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1179 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1180 le32_to_cpu(entry1->e_value_size)))
1181 return 1;
1183 entry1 = EXT3_XATTR_NEXT(entry1);
1184 entry2 = EXT3_XATTR_NEXT(entry2);
1186 if (!IS_LAST_ENTRY(entry2))
1187 return 1;
1188 return 0;
1192 * ext3_xattr_cache_find()
1194 * Find an identical extended attribute block.
1196 * Returns a pointer to the block found, or NULL if such a block was
1197 * not found or an error occurred.
1199 static struct buffer_head *
1200 ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header,
1201 struct mb_cache_entry **pce)
1203 __u32 hash = le32_to_cpu(header->h_hash);
1204 struct mb_cache_entry *ce;
1206 if (!header->h_hash)
1207 return NULL; /* never share */
1208 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1209 again:
1210 ce = mb_cache_entry_find_first(ext3_xattr_cache, 0,
1211 inode->i_sb->s_bdev, hash);
1212 while (ce) {
1213 struct buffer_head *bh;
1215 if (IS_ERR(ce)) {
1216 if (PTR_ERR(ce) == -EAGAIN)
1217 goto again;
1218 break;
1220 bh = sb_bread(inode->i_sb, ce->e_block);
1221 if (!bh) {
1222 ext3_error(inode->i_sb, __func__,
1223 "inode %lu: block %lu read error",
1224 inode->i_ino, (unsigned long) ce->e_block);
1225 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1226 EXT3_XATTR_REFCOUNT_MAX) {
1227 ea_idebug(inode, "block %lu refcount %d>=%d",
1228 (unsigned long) ce->e_block,
1229 le32_to_cpu(BHDR(bh)->h_refcount),
1230 EXT3_XATTR_REFCOUNT_MAX);
1231 } else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) {
1232 *pce = ce;
1233 return bh;
1235 brelse(bh);
1236 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1238 return NULL;
1241 #define NAME_HASH_SHIFT 5
1242 #define VALUE_HASH_SHIFT 16
1245 * ext3_xattr_hash_entry()
1247 * Compute the hash of an extended attribute.
1249 static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header,
1250 struct ext3_xattr_entry *entry)
1252 __u32 hash = 0;
1253 char *name = entry->e_name;
1254 int n;
1256 for (n=0; n < entry->e_name_len; n++) {
1257 hash = (hash << NAME_HASH_SHIFT) ^
1258 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1259 *name++;
1262 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1263 __le32 *value = (__le32 *)((char *)header +
1264 le16_to_cpu(entry->e_value_offs));
1265 for (n = (le32_to_cpu(entry->e_value_size) +
1266 EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) {
1267 hash = (hash << VALUE_HASH_SHIFT) ^
1268 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1269 le32_to_cpu(*value++);
1272 entry->e_hash = cpu_to_le32(hash);
1275 #undef NAME_HASH_SHIFT
1276 #undef VALUE_HASH_SHIFT
1278 #define BLOCK_HASH_SHIFT 16
1281 * ext3_xattr_rehash()
1283 * Re-compute the extended attribute hash value after an entry has changed.
1285 static void ext3_xattr_rehash(struct ext3_xattr_header *header,
1286 struct ext3_xattr_entry *entry)
1288 struct ext3_xattr_entry *here;
1289 __u32 hash = 0;
1291 ext3_xattr_hash_entry(header, entry);
1292 here = ENTRY(header+1);
1293 while (!IS_LAST_ENTRY(here)) {
1294 if (!here->e_hash) {
1295 /* Block is not shared if an entry's hash value == 0 */
1296 hash = 0;
1297 break;
1299 hash = (hash << BLOCK_HASH_SHIFT) ^
1300 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1301 le32_to_cpu(here->e_hash);
1302 here = EXT3_XATTR_NEXT(here);
1304 header->h_hash = cpu_to_le32(hash);
1307 #undef BLOCK_HASH_SHIFT
1309 int __init
1310 init_ext3_xattr(void)
1312 ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL,
1313 sizeof(struct mb_cache_entry) +
1314 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1315 if (!ext3_xattr_cache)
1316 return -ENOMEM;
1317 return 0;
1320 void
1321 exit_ext3_xattr(void)
1323 if (ext3_xattr_cache)
1324 mb_cache_destroy(ext3_xattr_cache);
1325 ext3_xattr_cache = NULL;