eCryptfs: fix write zeros behavior
[linux-2.6/kmemtrace.git] / fs / ext4 / xattr.c
blobe832e96095b33c177e880db6370e1686153439f6
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;
478 int error = 0;
480 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
481 error = ext4_journal_get_write_access(handle, bh);
482 if (error)
483 goto out;
485 lock_buffer(bh);
486 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
487 ea_bdebug(bh, "refcount now=0; freeing");
488 if (ce)
489 mb_cache_entry_free(ce);
490 ext4_free_blocks(handle, inode, bh->b_blocknr, 1);
491 get_bh(bh);
492 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
493 } else {
494 BHDR(bh)->h_refcount = cpu_to_le32(
495 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
496 error = ext4_journal_dirty_metadata(handle, bh);
497 if (IS_SYNC(inode))
498 handle->h_sync = 1;
499 DQUOT_FREE_BLOCK(inode, 1);
500 ea_bdebug(bh, "refcount now=%d; releasing",
501 le32_to_cpu(BHDR(bh)->h_refcount));
502 if (ce)
503 mb_cache_entry_release(ce);
505 unlock_buffer(bh);
506 out:
507 ext4_std_error(inode->i_sb, error);
508 return;
511 struct ext4_xattr_info {
512 int name_index;
513 const char *name;
514 const void *value;
515 size_t value_len;
518 struct ext4_xattr_search {
519 struct ext4_xattr_entry *first;
520 void *base;
521 void *end;
522 struct ext4_xattr_entry *here;
523 int not_found;
526 static int
527 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
529 struct ext4_xattr_entry *last;
530 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
532 /* Compute min_offs and last. */
533 last = s->first;
534 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
535 if (!last->e_value_block && last->e_value_size) {
536 size_t offs = le16_to_cpu(last->e_value_offs);
537 if (offs < min_offs)
538 min_offs = offs;
541 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
542 if (!s->not_found) {
543 if (!s->here->e_value_block && s->here->e_value_size) {
544 size_t size = le32_to_cpu(s->here->e_value_size);
545 free += EXT4_XATTR_SIZE(size);
547 free += EXT4_XATTR_LEN(name_len);
549 if (i->value) {
550 if (free < EXT4_XATTR_SIZE(i->value_len) ||
551 free < EXT4_XATTR_LEN(name_len) +
552 EXT4_XATTR_SIZE(i->value_len))
553 return -ENOSPC;
556 if (i->value && s->not_found) {
557 /* Insert the new name. */
558 size_t size = EXT4_XATTR_LEN(name_len);
559 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
560 memmove((void *)s->here + size, s->here, rest);
561 memset(s->here, 0, size);
562 s->here->e_name_index = i->name_index;
563 s->here->e_name_len = name_len;
564 memcpy(s->here->e_name, i->name, name_len);
565 } else {
566 if (!s->here->e_value_block && s->here->e_value_size) {
567 void *first_val = s->base + min_offs;
568 size_t offs = le16_to_cpu(s->here->e_value_offs);
569 void *val = s->base + offs;
570 size_t size = EXT4_XATTR_SIZE(
571 le32_to_cpu(s->here->e_value_size));
573 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
574 /* The old and the new value have the same
575 size. Just replace. */
576 s->here->e_value_size =
577 cpu_to_le32(i->value_len);
578 memset(val + size - EXT4_XATTR_PAD, 0,
579 EXT4_XATTR_PAD); /* Clear pad bytes. */
580 memcpy(val, i->value, i->value_len);
581 return 0;
584 /* Remove the old value. */
585 memmove(first_val + size, first_val, val - first_val);
586 memset(first_val, 0, size);
587 s->here->e_value_size = 0;
588 s->here->e_value_offs = 0;
589 min_offs += size;
591 /* Adjust all value offsets. */
592 last = s->first;
593 while (!IS_LAST_ENTRY(last)) {
594 size_t o = le16_to_cpu(last->e_value_offs);
595 if (!last->e_value_block &&
596 last->e_value_size && o < offs)
597 last->e_value_offs =
598 cpu_to_le16(o + size);
599 last = EXT4_XATTR_NEXT(last);
602 if (!i->value) {
603 /* Remove the old name. */
604 size_t size = EXT4_XATTR_LEN(name_len);
605 last = ENTRY((void *)last - size);
606 memmove(s->here, (void *)s->here + size,
607 (void *)last - (void *)s->here + sizeof(__u32));
608 memset(last, 0, size);
612 if (i->value) {
613 /* Insert the new value. */
614 s->here->e_value_size = cpu_to_le32(i->value_len);
615 if (i->value_len) {
616 size_t size = EXT4_XATTR_SIZE(i->value_len);
617 void *val = s->base + min_offs - size;
618 s->here->e_value_offs = cpu_to_le16(min_offs - size);
619 memset(val + size - EXT4_XATTR_PAD, 0,
620 EXT4_XATTR_PAD); /* Clear the pad bytes. */
621 memcpy(val, i->value, i->value_len);
624 return 0;
627 struct ext4_xattr_block_find {
628 struct ext4_xattr_search s;
629 struct buffer_head *bh;
632 static int
633 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
634 struct ext4_xattr_block_find *bs)
636 struct super_block *sb = inode->i_sb;
637 int error;
639 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
640 i->name_index, i->name, i->value, (long)i->value_len);
642 if (EXT4_I(inode)->i_file_acl) {
643 /* The inode already has an extended attribute block. */
644 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
645 error = -EIO;
646 if (!bs->bh)
647 goto cleanup;
648 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
649 atomic_read(&(bs->bh->b_count)),
650 le32_to_cpu(BHDR(bs->bh)->h_refcount));
651 if (ext4_xattr_check_block(bs->bh)) {
652 ext4_error(sb, __FUNCTION__,
653 "inode %lu: bad block %llu", inode->i_ino,
654 EXT4_I(inode)->i_file_acl);
655 error = -EIO;
656 goto cleanup;
658 /* Find the named attribute. */
659 bs->s.base = BHDR(bs->bh);
660 bs->s.first = BFIRST(bs->bh);
661 bs->s.end = bs->bh->b_data + bs->bh->b_size;
662 bs->s.here = bs->s.first;
663 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
664 i->name, bs->bh->b_size, 1);
665 if (error && error != -ENODATA)
666 goto cleanup;
667 bs->s.not_found = error;
669 error = 0;
671 cleanup:
672 return error;
675 static int
676 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
677 struct ext4_xattr_info *i,
678 struct ext4_xattr_block_find *bs)
680 struct super_block *sb = inode->i_sb;
681 struct buffer_head *new_bh = NULL;
682 struct ext4_xattr_search *s = &bs->s;
683 struct mb_cache_entry *ce = NULL;
684 int error = 0;
686 #define header(x) ((struct ext4_xattr_header *)(x))
688 if (i->value && i->value_len > sb->s_blocksize)
689 return -ENOSPC;
690 if (s->base) {
691 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
692 bs->bh->b_blocknr);
693 error = ext4_journal_get_write_access(handle, bs->bh);
694 if (error)
695 goto cleanup;
696 lock_buffer(bs->bh);
698 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
699 if (ce) {
700 mb_cache_entry_free(ce);
701 ce = NULL;
703 ea_bdebug(bs->bh, "modifying in-place");
704 error = ext4_xattr_set_entry(i, s);
705 if (!error) {
706 if (!IS_LAST_ENTRY(s->first))
707 ext4_xattr_rehash(header(s->base),
708 s->here);
709 ext4_xattr_cache_insert(bs->bh);
711 unlock_buffer(bs->bh);
712 if (error == -EIO)
713 goto bad_block;
714 if (!error)
715 error = ext4_journal_dirty_metadata(handle,
716 bs->bh);
717 if (error)
718 goto cleanup;
719 goto inserted;
720 } else {
721 int offset = (char *)s->here - bs->bh->b_data;
723 unlock_buffer(bs->bh);
724 jbd2_journal_release_buffer(handle, bs->bh);
725 if (ce) {
726 mb_cache_entry_release(ce);
727 ce = NULL;
729 ea_bdebug(bs->bh, "cloning");
730 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
731 error = -ENOMEM;
732 if (s->base == NULL)
733 goto cleanup;
734 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
735 s->first = ENTRY(header(s->base)+1);
736 header(s->base)->h_refcount = cpu_to_le32(1);
737 s->here = ENTRY(s->base + offset);
738 s->end = s->base + bs->bh->b_size;
740 } else {
741 /* Allocate a buffer where we construct the new block. */
742 s->base = kmalloc(sb->s_blocksize, GFP_KERNEL);
743 /* assert(header == s->base) */
744 error = -ENOMEM;
745 if (s->base == NULL)
746 goto cleanup;
747 memset(s->base, 0, sb->s_blocksize);
748 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
749 header(s->base)->h_blocks = cpu_to_le32(1);
750 header(s->base)->h_refcount = cpu_to_le32(1);
751 s->first = ENTRY(header(s->base)+1);
752 s->here = ENTRY(header(s->base)+1);
753 s->end = s->base + sb->s_blocksize;
756 error = ext4_xattr_set_entry(i, s);
757 if (error == -EIO)
758 goto bad_block;
759 if (error)
760 goto cleanup;
761 if (!IS_LAST_ENTRY(s->first))
762 ext4_xattr_rehash(header(s->base), s->here);
764 inserted:
765 if (!IS_LAST_ENTRY(s->first)) {
766 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
767 if (new_bh) {
768 /* We found an identical block in the cache. */
769 if (new_bh == bs->bh)
770 ea_bdebug(new_bh, "keeping");
771 else {
772 /* The old block is released after updating
773 the inode. */
774 error = -EDQUOT;
775 if (DQUOT_ALLOC_BLOCK(inode, 1))
776 goto cleanup;
777 error = ext4_journal_get_write_access(handle,
778 new_bh);
779 if (error)
780 goto cleanup_dquot;
781 lock_buffer(new_bh);
782 BHDR(new_bh)->h_refcount = cpu_to_le32(1 +
783 le32_to_cpu(BHDR(new_bh)->h_refcount));
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 = ext4_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 ext4_fsblk_t goal = le32_to_cpu(
802 EXT4_SB(sb)->s_es->s_first_data_block) +
803 (ext4_fsblk_t)EXT4_I(inode)->i_block_group *
804 EXT4_BLOCKS_PER_GROUP(sb);
805 ext4_fsblk_t block = ext4_new_block(handle, inode,
806 goal, &error);
807 if (error)
808 goto cleanup;
809 ea_idebug(inode, "creating block %d", block);
811 new_bh = sb_getblk(sb, block);
812 if (!new_bh) {
813 getblk_failed:
814 ext4_free_blocks(handle, inode, block, 1);
815 error = -EIO;
816 goto cleanup;
818 lock_buffer(new_bh);
819 error = ext4_journal_get_create_access(handle, new_bh);
820 if (error) {
821 unlock_buffer(new_bh);
822 goto getblk_failed;
824 memcpy(new_bh->b_data, s->base, new_bh->b_size);
825 set_buffer_uptodate(new_bh);
826 unlock_buffer(new_bh);
827 ext4_xattr_cache_insert(new_bh);
828 error = ext4_journal_dirty_metadata(handle, new_bh);
829 if (error)
830 goto cleanup;
834 /* Update the inode. */
835 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
837 /* Drop the previous xattr block. */
838 if (bs->bh && bs->bh != new_bh)
839 ext4_xattr_release_block(handle, inode, bs->bh);
840 error = 0;
842 cleanup:
843 if (ce)
844 mb_cache_entry_release(ce);
845 brelse(new_bh);
846 if (!(bs->bh && s->base == bs->bh->b_data))
847 kfree(s->base);
849 return error;
851 cleanup_dquot:
852 DQUOT_FREE_BLOCK(inode, 1);
853 goto cleanup;
855 bad_block:
856 ext4_error(inode->i_sb, __FUNCTION__,
857 "inode %lu: bad block %llu", inode->i_ino,
858 EXT4_I(inode)->i_file_acl);
859 goto cleanup;
861 #undef header
864 struct ext4_xattr_ibody_find {
865 struct ext4_xattr_search s;
866 struct ext4_iloc iloc;
869 static int
870 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
871 struct ext4_xattr_ibody_find *is)
873 struct ext4_xattr_ibody_header *header;
874 struct ext4_inode *raw_inode;
875 int error;
877 if (EXT4_I(inode)->i_extra_isize == 0)
878 return 0;
879 raw_inode = ext4_raw_inode(&is->iloc);
880 header = IHDR(inode, raw_inode);
881 is->s.base = is->s.first = IFIRST(header);
882 is->s.here = is->s.first;
883 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
884 if (EXT4_I(inode)->i_state & EXT4_STATE_XATTR) {
885 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
886 if (error)
887 return error;
888 /* Find the named attribute. */
889 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
890 i->name, is->s.end -
891 (void *)is->s.base, 0);
892 if (error && error != -ENODATA)
893 return error;
894 is->s.not_found = error;
896 return 0;
899 static int
900 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
901 struct ext4_xattr_info *i,
902 struct ext4_xattr_ibody_find *is)
904 struct ext4_xattr_ibody_header *header;
905 struct ext4_xattr_search *s = &is->s;
906 int error;
908 if (EXT4_I(inode)->i_extra_isize == 0)
909 return -ENOSPC;
910 error = ext4_xattr_set_entry(i, s);
911 if (error)
912 return error;
913 header = IHDR(inode, ext4_raw_inode(&is->iloc));
914 if (!IS_LAST_ENTRY(s->first)) {
915 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
916 EXT4_I(inode)->i_state |= EXT4_STATE_XATTR;
917 } else {
918 header->h_magic = cpu_to_le32(0);
919 EXT4_I(inode)->i_state &= ~EXT4_STATE_XATTR;
921 return 0;
925 * ext4_xattr_set_handle()
927 * Create, replace or remove an extended attribute for this inode. Buffer
928 * is NULL to remove an existing extended attribute, and non-NULL to
929 * either replace an existing extended attribute, or create a new extended
930 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
931 * specify that an extended attribute must exist and must not exist
932 * previous to the call, respectively.
934 * Returns 0, or a negative error number on failure.
937 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
938 const char *name, const void *value, size_t value_len,
939 int flags)
941 struct ext4_xattr_info i = {
942 .name_index = name_index,
943 .name = name,
944 .value = value,
945 .value_len = value_len,
948 struct ext4_xattr_ibody_find is = {
949 .s = { .not_found = -ENODATA, },
951 struct ext4_xattr_block_find bs = {
952 .s = { .not_found = -ENODATA, },
954 int error;
956 if (!name)
957 return -EINVAL;
958 if (strlen(name) > 255)
959 return -ERANGE;
960 down_write(&EXT4_I(inode)->xattr_sem);
961 error = ext4_get_inode_loc(inode, &is.iloc);
962 if (error)
963 goto cleanup;
965 if (EXT4_I(inode)->i_state & EXT4_STATE_NEW) {
966 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
967 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
968 EXT4_I(inode)->i_state &= ~EXT4_STATE_NEW;
971 error = ext4_xattr_ibody_find(inode, &i, &is);
972 if (error)
973 goto cleanup;
974 if (is.s.not_found)
975 error = ext4_xattr_block_find(inode, &i, &bs);
976 if (error)
977 goto cleanup;
978 if (is.s.not_found && bs.s.not_found) {
979 error = -ENODATA;
980 if (flags & XATTR_REPLACE)
981 goto cleanup;
982 error = 0;
983 if (!value)
984 goto cleanup;
985 } else {
986 error = -EEXIST;
987 if (flags & XATTR_CREATE)
988 goto cleanup;
990 error = ext4_journal_get_write_access(handle, is.iloc.bh);
991 if (error)
992 goto cleanup;
993 if (!value) {
994 if (!is.s.not_found)
995 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
996 else if (!bs.s.not_found)
997 error = ext4_xattr_block_set(handle, inode, &i, &bs);
998 } else {
999 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1000 if (!error && !bs.s.not_found) {
1001 i.value = NULL;
1002 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1003 } else if (error == -ENOSPC) {
1004 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1005 if (error)
1006 goto cleanup;
1007 if (!is.s.not_found) {
1008 i.value = NULL;
1009 error = ext4_xattr_ibody_set(handle, inode, &i,
1010 &is);
1014 if (!error) {
1015 ext4_xattr_update_super_block(handle, inode->i_sb);
1016 inode->i_ctime = CURRENT_TIME_SEC;
1017 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1019 * The bh is consumed by ext4_mark_iloc_dirty, even with
1020 * error != 0.
1022 is.iloc.bh = NULL;
1023 if (IS_SYNC(inode))
1024 handle->h_sync = 1;
1027 cleanup:
1028 brelse(is.iloc.bh);
1029 brelse(bs.bh);
1030 up_write(&EXT4_I(inode)->xattr_sem);
1031 return error;
1035 * ext4_xattr_set()
1037 * Like ext4_xattr_set_handle, but start from an inode. This extended
1038 * attribute modification is a filesystem transaction by itself.
1040 * Returns 0, or a negative error number on failure.
1043 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1044 const void *value, size_t value_len, int flags)
1046 handle_t *handle;
1047 int error, retries = 0;
1049 retry:
1050 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1051 if (IS_ERR(handle)) {
1052 error = PTR_ERR(handle);
1053 } else {
1054 int error2;
1056 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1057 value, value_len, flags);
1058 error2 = ext4_journal_stop(handle);
1059 if (error == -ENOSPC &&
1060 ext4_should_retry_alloc(inode->i_sb, &retries))
1061 goto retry;
1062 if (error == 0)
1063 error = error2;
1066 return error;
1070 * ext4_xattr_delete_inode()
1072 * Free extended attribute resources associated with this inode. This
1073 * is called immediately before an inode is freed. We have exclusive
1074 * access to the inode.
1076 void
1077 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1079 struct buffer_head *bh = NULL;
1081 if (!EXT4_I(inode)->i_file_acl)
1082 goto cleanup;
1083 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1084 if (!bh) {
1085 ext4_error(inode->i_sb, __FUNCTION__,
1086 "inode %lu: block %llu read error", inode->i_ino,
1087 EXT4_I(inode)->i_file_acl);
1088 goto cleanup;
1090 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1091 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1092 ext4_error(inode->i_sb, __FUNCTION__,
1093 "inode %lu: bad block %llu", inode->i_ino,
1094 EXT4_I(inode)->i_file_acl);
1095 goto cleanup;
1097 ext4_xattr_release_block(handle, inode, bh);
1098 EXT4_I(inode)->i_file_acl = 0;
1100 cleanup:
1101 brelse(bh);
1105 * ext4_xattr_put_super()
1107 * This is called when a file system is unmounted.
1109 void
1110 ext4_xattr_put_super(struct super_block *sb)
1112 mb_cache_shrink(sb->s_bdev);
1116 * ext4_xattr_cache_insert()
1118 * Create a new entry in the extended attribute cache, and insert
1119 * it unless such an entry is already in the cache.
1121 * Returns 0, or a negative error number on failure.
1123 static void
1124 ext4_xattr_cache_insert(struct buffer_head *bh)
1126 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1127 struct mb_cache_entry *ce;
1128 int error;
1130 ce = mb_cache_entry_alloc(ext4_xattr_cache);
1131 if (!ce) {
1132 ea_bdebug(bh, "out of memory");
1133 return;
1135 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1136 if (error) {
1137 mb_cache_entry_free(ce);
1138 if (error == -EBUSY) {
1139 ea_bdebug(bh, "already in cache");
1140 error = 0;
1142 } else {
1143 ea_bdebug(bh, "inserting [%x]", (int)hash);
1144 mb_cache_entry_release(ce);
1149 * ext4_xattr_cmp()
1151 * Compare two extended attribute blocks for equality.
1153 * Returns 0 if the blocks are equal, 1 if they differ, and
1154 * a negative error number on errors.
1156 static int
1157 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1158 struct ext4_xattr_header *header2)
1160 struct ext4_xattr_entry *entry1, *entry2;
1162 entry1 = ENTRY(header1+1);
1163 entry2 = ENTRY(header2+1);
1164 while (!IS_LAST_ENTRY(entry1)) {
1165 if (IS_LAST_ENTRY(entry2))
1166 return 1;
1167 if (entry1->e_hash != entry2->e_hash ||
1168 entry1->e_name_index != entry2->e_name_index ||
1169 entry1->e_name_len != entry2->e_name_len ||
1170 entry1->e_value_size != entry2->e_value_size ||
1171 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1172 return 1;
1173 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1174 return -EIO;
1175 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1176 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1177 le32_to_cpu(entry1->e_value_size)))
1178 return 1;
1180 entry1 = EXT4_XATTR_NEXT(entry1);
1181 entry2 = EXT4_XATTR_NEXT(entry2);
1183 if (!IS_LAST_ENTRY(entry2))
1184 return 1;
1185 return 0;
1189 * ext4_xattr_cache_find()
1191 * Find an identical extended attribute block.
1193 * Returns a pointer to the block found, or NULL if such a block was
1194 * not found or an error occurred.
1196 static struct buffer_head *
1197 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1198 struct mb_cache_entry **pce)
1200 __u32 hash = le32_to_cpu(header->h_hash);
1201 struct mb_cache_entry *ce;
1203 if (!header->h_hash)
1204 return NULL; /* never share */
1205 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1206 again:
1207 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0,
1208 inode->i_sb->s_bdev, hash);
1209 while (ce) {
1210 struct buffer_head *bh;
1212 if (IS_ERR(ce)) {
1213 if (PTR_ERR(ce) == -EAGAIN)
1214 goto again;
1215 break;
1217 bh = sb_bread(inode->i_sb, ce->e_block);
1218 if (!bh) {
1219 ext4_error(inode->i_sb, __FUNCTION__,
1220 "inode %lu: block %lu read error",
1221 inode->i_ino, (unsigned long) ce->e_block);
1222 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1223 EXT4_XATTR_REFCOUNT_MAX) {
1224 ea_idebug(inode, "block %lu refcount %d>=%d",
1225 (unsigned long) ce->e_block,
1226 le32_to_cpu(BHDR(bh)->h_refcount),
1227 EXT4_XATTR_REFCOUNT_MAX);
1228 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1229 *pce = ce;
1230 return bh;
1232 brelse(bh);
1233 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1235 return NULL;
1238 #define NAME_HASH_SHIFT 5
1239 #define VALUE_HASH_SHIFT 16
1242 * ext4_xattr_hash_entry()
1244 * Compute the hash of an extended attribute.
1246 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1247 struct ext4_xattr_entry *entry)
1249 __u32 hash = 0;
1250 char *name = entry->e_name;
1251 int n;
1253 for (n=0; n < entry->e_name_len; n++) {
1254 hash = (hash << NAME_HASH_SHIFT) ^
1255 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1256 *name++;
1259 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1260 __le32 *value = (__le32 *)((char *)header +
1261 le16_to_cpu(entry->e_value_offs));
1262 for (n = (le32_to_cpu(entry->e_value_size) +
1263 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1264 hash = (hash << VALUE_HASH_SHIFT) ^
1265 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1266 le32_to_cpu(*value++);
1269 entry->e_hash = cpu_to_le32(hash);
1272 #undef NAME_HASH_SHIFT
1273 #undef VALUE_HASH_SHIFT
1275 #define BLOCK_HASH_SHIFT 16
1278 * ext4_xattr_rehash()
1280 * Re-compute the extended attribute hash value after an entry has changed.
1282 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1283 struct ext4_xattr_entry *entry)
1285 struct ext4_xattr_entry *here;
1286 __u32 hash = 0;
1288 ext4_xattr_hash_entry(header, entry);
1289 here = ENTRY(header+1);
1290 while (!IS_LAST_ENTRY(here)) {
1291 if (!here->e_hash) {
1292 /* Block is not shared if an entry's hash value == 0 */
1293 hash = 0;
1294 break;
1296 hash = (hash << BLOCK_HASH_SHIFT) ^
1297 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1298 le32_to_cpu(here->e_hash);
1299 here = EXT4_XATTR_NEXT(here);
1301 header->h_hash = cpu_to_le32(hash);
1304 #undef BLOCK_HASH_SHIFT
1306 int __init
1307 init_ext4_xattr(void)
1309 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL,
1310 sizeof(struct mb_cache_entry) +
1311 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1312 if (!ext4_xattr_cache)
1313 return -ENOMEM;
1314 return 0;
1317 void
1318 exit_ext4_xattr(void)
1320 if (ext4_xattr_cache)
1321 mb_cache_destroy(ext4_xattr_cache);
1322 ext4_xattr_cache = NULL;