Staging: android: lowmemorykiller.c: fix it for "oom: move oom_adj value from task_st...
[linux-2.6/mini2440.git] / fs / ext4 / xattr.c
blob62b31c24699418235b98bbc164257db346202bdd
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/mbcache.h>
57 #include <linux/quotaops.h>
58 #include <linux/rwsem.h>
59 #include "ext4_jbd2.h"
60 #include "ext4.h"
61 #include "xattr.h"
62 #include "acl.h"
64 #define BHDR(bh) ((struct ext4_xattr_header *)((bh)->b_data))
65 #define ENTRY(ptr) ((struct ext4_xattr_entry *)(ptr))
66 #define BFIRST(bh) ENTRY(BHDR(bh)+1)
67 #define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
69 #ifdef EXT4_XATTR_DEBUG
70 # define ea_idebug(inode, f...) do { \
71 printk(KERN_DEBUG "inode %s:%lu: ", \
72 inode->i_sb->s_id, inode->i_ino); \
73 printk(f); \
74 printk("\n"); \
75 } while (0)
76 # define ea_bdebug(bh, f...) do { \
77 char b[BDEVNAME_SIZE]; \
78 printk(KERN_DEBUG "block %s:%lu: ", \
79 bdevname(bh->b_bdev, b), \
80 (unsigned long) bh->b_blocknr); \
81 printk(f); \
82 printk("\n"); \
83 } while (0)
84 #else
85 # define ea_idebug(f...)
86 # define ea_bdebug(f...)
87 #endif
89 static void ext4_xattr_cache_insert(struct buffer_head *);
90 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
91 struct ext4_xattr_header *,
92 struct mb_cache_entry **);
93 static void ext4_xattr_rehash(struct ext4_xattr_header *,
94 struct ext4_xattr_entry *);
95 static int ext4_xattr_list(struct inode *inode, char *buffer,
96 size_t buffer_size);
98 static struct mb_cache *ext4_xattr_cache;
100 static struct xattr_handler *ext4_xattr_handler_map[] = {
101 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
102 #ifdef CONFIG_EXT4_FS_POSIX_ACL
103 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext4_xattr_acl_access_handler,
104 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext4_xattr_acl_default_handler,
105 #endif
106 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
107 #ifdef CONFIG_EXT4_FS_SECURITY
108 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
109 #endif
112 struct xattr_handler *ext4_xattr_handlers[] = {
113 &ext4_xattr_user_handler,
114 &ext4_xattr_trusted_handler,
115 #ifdef CONFIG_EXT4_FS_POSIX_ACL
116 &ext4_xattr_acl_access_handler,
117 &ext4_xattr_acl_default_handler,
118 #endif
119 #ifdef CONFIG_EXT4_FS_SECURITY
120 &ext4_xattr_security_handler,
121 #endif
122 NULL
125 static inline struct xattr_handler *
126 ext4_xattr_handler(int name_index)
128 struct xattr_handler *handler = NULL;
130 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
131 handler = ext4_xattr_handler_map[name_index];
132 return handler;
136 * Inode operation listxattr()
138 * dentry->d_inode->i_mutex: don't care
140 ssize_t
141 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
143 return ext4_xattr_list(dentry->d_inode, buffer, size);
146 static int
147 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
149 while (!IS_LAST_ENTRY(entry)) {
150 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
151 if ((void *)next >= end)
152 return -EIO;
153 entry = next;
155 return 0;
158 static inline int
159 ext4_xattr_check_block(struct buffer_head *bh)
161 int error;
163 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
164 BHDR(bh)->h_blocks != cpu_to_le32(1))
165 return -EIO;
166 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
167 return error;
170 static inline int
171 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
173 size_t value_size = le32_to_cpu(entry->e_value_size);
175 if (entry->e_value_block != 0 || value_size > size ||
176 le16_to_cpu(entry->e_value_offs) + value_size > size)
177 return -EIO;
178 return 0;
181 static int
182 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
183 const char *name, size_t size, int sorted)
185 struct ext4_xattr_entry *entry;
186 size_t name_len;
187 int cmp = 1;
189 if (name == NULL)
190 return -EINVAL;
191 name_len = strlen(name);
192 entry = *pentry;
193 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
194 cmp = name_index - entry->e_name_index;
195 if (!cmp)
196 cmp = name_len - entry->e_name_len;
197 if (!cmp)
198 cmp = memcmp(name, entry->e_name, name_len);
199 if (cmp <= 0 && (sorted || cmp == 0))
200 break;
202 *pentry = entry;
203 if (!cmp && ext4_xattr_check_entry(entry, size))
204 return -EIO;
205 return cmp ? -ENODATA : 0;
208 static int
209 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
210 void *buffer, size_t buffer_size)
212 struct buffer_head *bh = NULL;
213 struct ext4_xattr_entry *entry;
214 size_t size;
215 int error;
217 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
218 name_index, name, buffer, (long)buffer_size);
220 error = -ENODATA;
221 if (!EXT4_I(inode)->i_file_acl)
222 goto cleanup;
223 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
224 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
225 if (!bh)
226 goto cleanup;
227 ea_bdebug(bh, "b_count=%d, refcount=%d",
228 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
229 if (ext4_xattr_check_block(bh)) {
230 bad_block: ext4_error(inode->i_sb, __func__,
231 "inode %lu: bad block %llu", inode->i_ino,
232 EXT4_I(inode)->i_file_acl);
233 error = -EIO;
234 goto cleanup;
236 ext4_xattr_cache_insert(bh);
237 entry = BFIRST(bh);
238 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
239 if (error == -EIO)
240 goto bad_block;
241 if (error)
242 goto cleanup;
243 size = le32_to_cpu(entry->e_value_size);
244 if (buffer) {
245 error = -ERANGE;
246 if (size > buffer_size)
247 goto cleanup;
248 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
249 size);
251 error = size;
253 cleanup:
254 brelse(bh);
255 return error;
258 static int
259 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
260 void *buffer, size_t buffer_size)
262 struct ext4_xattr_ibody_header *header;
263 struct ext4_xattr_entry *entry;
264 struct ext4_inode *raw_inode;
265 struct ext4_iloc iloc;
266 size_t size;
267 void *end;
268 int error;
270 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
271 return -ENODATA;
272 error = ext4_get_inode_loc(inode, &iloc);
273 if (error)
274 return error;
275 raw_inode = ext4_raw_inode(&iloc);
276 header = IHDR(inode, raw_inode);
277 entry = IFIRST(header);
278 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
279 error = ext4_xattr_check_names(entry, end);
280 if (error)
281 goto cleanup;
282 error = ext4_xattr_find_entry(&entry, name_index, name,
283 end - (void *)entry, 0);
284 if (error)
285 goto cleanup;
286 size = le32_to_cpu(entry->e_value_size);
287 if (buffer) {
288 error = -ERANGE;
289 if (size > buffer_size)
290 goto cleanup;
291 memcpy(buffer, (void *)IFIRST(header) +
292 le16_to_cpu(entry->e_value_offs), size);
294 error = size;
296 cleanup:
297 brelse(iloc.bh);
298 return error;
302 * ext4_xattr_get()
304 * Copy an extended attribute into the buffer
305 * provided, or compute the buffer size required.
306 * Buffer is NULL to compute the size of the buffer required.
308 * Returns a negative error number on failure, or the number of bytes
309 * used / required on success.
312 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
313 void *buffer, size_t buffer_size)
315 int error;
317 down_read(&EXT4_I(inode)->xattr_sem);
318 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
319 buffer_size);
320 if (error == -ENODATA)
321 error = ext4_xattr_block_get(inode, name_index, name, buffer,
322 buffer_size);
323 up_read(&EXT4_I(inode)->xattr_sem);
324 return error;
327 static int
328 ext4_xattr_list_entries(struct inode *inode, struct ext4_xattr_entry *entry,
329 char *buffer, size_t buffer_size)
331 size_t rest = buffer_size;
333 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
334 struct xattr_handler *handler =
335 ext4_xattr_handler(entry->e_name_index);
337 if (handler) {
338 size_t size = handler->list(inode, buffer, rest,
339 entry->e_name,
340 entry->e_name_len);
341 if (buffer) {
342 if (size > rest)
343 return -ERANGE;
344 buffer += size;
346 rest -= size;
349 return buffer_size - rest;
352 static int
353 ext4_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
355 struct buffer_head *bh = NULL;
356 int error;
358 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
359 buffer, (long)buffer_size);
361 error = 0;
362 if (!EXT4_I(inode)->i_file_acl)
363 goto cleanup;
364 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
365 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
366 error = -EIO;
367 if (!bh)
368 goto cleanup;
369 ea_bdebug(bh, "b_count=%d, refcount=%d",
370 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
371 if (ext4_xattr_check_block(bh)) {
372 ext4_error(inode->i_sb, __func__,
373 "inode %lu: bad block %llu", inode->i_ino,
374 EXT4_I(inode)->i_file_acl);
375 error = -EIO;
376 goto cleanup;
378 ext4_xattr_cache_insert(bh);
379 error = ext4_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
381 cleanup:
382 brelse(bh);
384 return error;
387 static int
388 ext4_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
390 struct ext4_xattr_ibody_header *header;
391 struct ext4_inode *raw_inode;
392 struct ext4_iloc iloc;
393 void *end;
394 int error;
396 if (!(EXT4_I(inode)->i_state & EXT4_STATE_XATTR))
397 return 0;
398 error = ext4_get_inode_loc(inode, &iloc);
399 if (error)
400 return error;
401 raw_inode = ext4_raw_inode(&iloc);
402 header = IHDR(inode, raw_inode);
403 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
404 error = ext4_xattr_check_names(IFIRST(header), end);
405 if (error)
406 goto cleanup;
407 error = ext4_xattr_list_entries(inode, IFIRST(header),
408 buffer, buffer_size);
410 cleanup:
411 brelse(iloc.bh);
412 return error;
416 * ext4_xattr_list()
418 * Copy a list of attribute names into the buffer
419 * provided, or compute the buffer size required.
420 * Buffer is NULL to compute the size of the buffer required.
422 * Returns a negative error number on failure, or the number of bytes
423 * used / required on success.
425 static int
426 ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
428 int i_error, b_error;
430 down_read(&EXT4_I(inode)->xattr_sem);
431 i_error = ext4_xattr_ibody_list(inode, buffer, buffer_size);
432 if (i_error < 0) {
433 b_error = 0;
434 } else {
435 if (buffer) {
436 buffer += i_error;
437 buffer_size -= i_error;
439 b_error = ext4_xattr_block_list(inode, buffer, buffer_size);
440 if (b_error < 0)
441 i_error = 0;
443 up_read(&EXT4_I(inode)->xattr_sem);
444 return i_error + b_error;
448 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
449 * not set, set it.
451 static void ext4_xattr_update_super_block(handle_t *handle,
452 struct super_block *sb)
454 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
455 return;
457 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
458 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
459 sb->s_dirt = 1;
460 ext4_handle_dirty_metadata(handle, NULL, EXT4_SB(sb)->s_sbh);
465 * Release the xattr block BH: If the reference count is > 1, decrement
466 * it; otherwise free the block.
468 static void
469 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
470 struct buffer_head *bh)
472 struct mb_cache_entry *ce = NULL;
473 int error = 0;
475 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
476 error = ext4_journal_get_write_access(handle, bh);
477 if (error)
478 goto out;
480 lock_buffer(bh);
481 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
482 ea_bdebug(bh, "refcount now=0; freeing");
483 if (ce)
484 mb_cache_entry_free(ce);
485 ext4_free_blocks(handle, inode, bh->b_blocknr, 1, 1);
486 get_bh(bh);
487 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
488 } else {
489 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
490 error = ext4_handle_dirty_metadata(handle, inode, bh);
491 if (IS_SYNC(inode))
492 ext4_handle_sync(handle);
493 vfs_dq_free_block(inode, 1);
494 ea_bdebug(bh, "refcount now=%d; releasing",
495 le32_to_cpu(BHDR(bh)->h_refcount));
496 if (ce)
497 mb_cache_entry_release(ce);
499 unlock_buffer(bh);
500 out:
501 ext4_std_error(inode->i_sb, error);
502 return;
506 * Find the available free space for EAs. This also returns the total number of
507 * bytes used by EA entries.
509 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
510 size_t *min_offs, void *base, int *total)
512 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
513 *total += EXT4_XATTR_LEN(last->e_name_len);
514 if (!last->e_value_block && last->e_value_size) {
515 size_t offs = le16_to_cpu(last->e_value_offs);
516 if (offs < *min_offs)
517 *min_offs = offs;
520 return (*min_offs - ((void *)last - base) - sizeof(__u32));
523 struct ext4_xattr_info {
524 int name_index;
525 const char *name;
526 const void *value;
527 size_t value_len;
530 struct ext4_xattr_search {
531 struct ext4_xattr_entry *first;
532 void *base;
533 void *end;
534 struct ext4_xattr_entry *here;
535 int not_found;
538 static int
539 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
541 struct ext4_xattr_entry *last;
542 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
544 /* Compute min_offs and last. */
545 last = s->first;
546 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
547 if (!last->e_value_block && last->e_value_size) {
548 size_t offs = le16_to_cpu(last->e_value_offs);
549 if (offs < min_offs)
550 min_offs = offs;
553 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
554 if (!s->not_found) {
555 if (!s->here->e_value_block && s->here->e_value_size) {
556 size_t size = le32_to_cpu(s->here->e_value_size);
557 free += EXT4_XATTR_SIZE(size);
559 free += EXT4_XATTR_LEN(name_len);
561 if (i->value) {
562 if (free < EXT4_XATTR_SIZE(i->value_len) ||
563 free < EXT4_XATTR_LEN(name_len) +
564 EXT4_XATTR_SIZE(i->value_len))
565 return -ENOSPC;
568 if (i->value && s->not_found) {
569 /* Insert the new name. */
570 size_t size = EXT4_XATTR_LEN(name_len);
571 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
572 memmove((void *)s->here + size, s->here, rest);
573 memset(s->here, 0, size);
574 s->here->e_name_index = i->name_index;
575 s->here->e_name_len = name_len;
576 memcpy(s->here->e_name, i->name, name_len);
577 } else {
578 if (!s->here->e_value_block && s->here->e_value_size) {
579 void *first_val = s->base + min_offs;
580 size_t offs = le16_to_cpu(s->here->e_value_offs);
581 void *val = s->base + offs;
582 size_t size = EXT4_XATTR_SIZE(
583 le32_to_cpu(s->here->e_value_size));
585 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
586 /* The old and the new value have the same
587 size. Just replace. */
588 s->here->e_value_size =
589 cpu_to_le32(i->value_len);
590 memset(val + size - EXT4_XATTR_PAD, 0,
591 EXT4_XATTR_PAD); /* Clear pad bytes. */
592 memcpy(val, i->value, i->value_len);
593 return 0;
596 /* Remove the old value. */
597 memmove(first_val + size, first_val, val - first_val);
598 memset(first_val, 0, size);
599 s->here->e_value_size = 0;
600 s->here->e_value_offs = 0;
601 min_offs += size;
603 /* Adjust all value offsets. */
604 last = s->first;
605 while (!IS_LAST_ENTRY(last)) {
606 size_t o = le16_to_cpu(last->e_value_offs);
607 if (!last->e_value_block &&
608 last->e_value_size && o < offs)
609 last->e_value_offs =
610 cpu_to_le16(o + size);
611 last = EXT4_XATTR_NEXT(last);
614 if (!i->value) {
615 /* Remove the old name. */
616 size_t size = EXT4_XATTR_LEN(name_len);
617 last = ENTRY((void *)last - size);
618 memmove(s->here, (void *)s->here + size,
619 (void *)last - (void *)s->here + sizeof(__u32));
620 memset(last, 0, size);
624 if (i->value) {
625 /* Insert the new value. */
626 s->here->e_value_size = cpu_to_le32(i->value_len);
627 if (i->value_len) {
628 size_t size = EXT4_XATTR_SIZE(i->value_len);
629 void *val = s->base + min_offs - size;
630 s->here->e_value_offs = cpu_to_le16(min_offs - size);
631 memset(val + size - EXT4_XATTR_PAD, 0,
632 EXT4_XATTR_PAD); /* Clear the pad bytes. */
633 memcpy(val, i->value, i->value_len);
636 return 0;
639 struct ext4_xattr_block_find {
640 struct ext4_xattr_search s;
641 struct buffer_head *bh;
644 static int
645 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
646 struct ext4_xattr_block_find *bs)
648 struct super_block *sb = inode->i_sb;
649 int error;
651 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
652 i->name_index, i->name, i->value, (long)i->value_len);
654 if (EXT4_I(inode)->i_file_acl) {
655 /* The inode already has an extended attribute block. */
656 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
657 error = -EIO;
658 if (!bs->bh)
659 goto cleanup;
660 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
661 atomic_read(&(bs->bh->b_count)),
662 le32_to_cpu(BHDR(bs->bh)->h_refcount));
663 if (ext4_xattr_check_block(bs->bh)) {
664 ext4_error(sb, __func__,
665 "inode %lu: bad block %llu", inode->i_ino,
666 EXT4_I(inode)->i_file_acl);
667 error = -EIO;
668 goto cleanup;
670 /* Find the named attribute. */
671 bs->s.base = BHDR(bs->bh);
672 bs->s.first = BFIRST(bs->bh);
673 bs->s.end = bs->bh->b_data + bs->bh->b_size;
674 bs->s.here = bs->s.first;
675 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
676 i->name, bs->bh->b_size, 1);
677 if (error && error != -ENODATA)
678 goto cleanup;
679 bs->s.not_found = error;
681 error = 0;
683 cleanup:
684 return error;
687 static int
688 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
689 struct ext4_xattr_info *i,
690 struct ext4_xattr_block_find *bs)
692 struct super_block *sb = inode->i_sb;
693 struct buffer_head *new_bh = NULL;
694 struct ext4_xattr_search *s = &bs->s;
695 struct mb_cache_entry *ce = NULL;
696 int error = 0;
698 #define header(x) ((struct ext4_xattr_header *)(x))
700 if (i->value && i->value_len > sb->s_blocksize)
701 return -ENOSPC;
702 if (s->base) {
703 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
704 bs->bh->b_blocknr);
705 error = ext4_journal_get_write_access(handle, bs->bh);
706 if (error)
707 goto cleanup;
708 lock_buffer(bs->bh);
710 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
711 if (ce) {
712 mb_cache_entry_free(ce);
713 ce = NULL;
715 ea_bdebug(bs->bh, "modifying in-place");
716 error = ext4_xattr_set_entry(i, s);
717 if (!error) {
718 if (!IS_LAST_ENTRY(s->first))
719 ext4_xattr_rehash(header(s->base),
720 s->here);
721 ext4_xattr_cache_insert(bs->bh);
723 unlock_buffer(bs->bh);
724 if (error == -EIO)
725 goto bad_block;
726 if (!error)
727 error = ext4_handle_dirty_metadata(handle,
728 inode,
729 bs->bh);
730 if (error)
731 goto cleanup;
732 goto inserted;
733 } else {
734 int offset = (char *)s->here - bs->bh->b_data;
736 unlock_buffer(bs->bh);
737 jbd2_journal_release_buffer(handle, bs->bh);
738 if (ce) {
739 mb_cache_entry_release(ce);
740 ce = NULL;
742 ea_bdebug(bs->bh, "cloning");
743 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
744 error = -ENOMEM;
745 if (s->base == NULL)
746 goto cleanup;
747 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
748 s->first = ENTRY(header(s->base)+1);
749 header(s->base)->h_refcount = cpu_to_le32(1);
750 s->here = ENTRY(s->base + offset);
751 s->end = s->base + bs->bh->b_size;
753 } else {
754 /* Allocate a buffer where we construct the new block. */
755 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
756 /* assert(header == s->base) */
757 error = -ENOMEM;
758 if (s->base == NULL)
759 goto cleanup;
760 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
761 header(s->base)->h_blocks = cpu_to_le32(1);
762 header(s->base)->h_refcount = cpu_to_le32(1);
763 s->first = ENTRY(header(s->base)+1);
764 s->here = ENTRY(header(s->base)+1);
765 s->end = s->base + sb->s_blocksize;
768 error = ext4_xattr_set_entry(i, s);
769 if (error == -EIO)
770 goto bad_block;
771 if (error)
772 goto cleanup;
773 if (!IS_LAST_ENTRY(s->first))
774 ext4_xattr_rehash(header(s->base), s->here);
776 inserted:
777 if (!IS_LAST_ENTRY(s->first)) {
778 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
779 if (new_bh) {
780 /* We found an identical block in the cache. */
781 if (new_bh == bs->bh)
782 ea_bdebug(new_bh, "keeping");
783 else {
784 /* The old block is released after updating
785 the inode. */
786 error = -EDQUOT;
787 if (vfs_dq_alloc_block(inode, 1))
788 goto cleanup;
789 error = ext4_journal_get_write_access(handle,
790 new_bh);
791 if (error)
792 goto cleanup_dquot;
793 lock_buffer(new_bh);
794 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
795 ea_bdebug(new_bh, "reusing; refcount now=%d",
796 le32_to_cpu(BHDR(new_bh)->h_refcount));
797 unlock_buffer(new_bh);
798 error = ext4_handle_dirty_metadata(handle,
799 inode,
800 new_bh);
801 if (error)
802 goto cleanup_dquot;
804 mb_cache_entry_release(ce);
805 ce = NULL;
806 } else if (bs->bh && s->base == bs->bh->b_data) {
807 /* We were modifying this block in-place. */
808 ea_bdebug(bs->bh, "keeping this block");
809 new_bh = bs->bh;
810 get_bh(new_bh);
811 } else {
812 /* We need to allocate a new block */
813 ext4_fsblk_t goal = ext4_group_first_block_no(sb,
814 EXT4_I(inode)->i_block_group);
815 ext4_fsblk_t block = ext4_new_meta_blocks(handle, inode,
816 goal, NULL, &error);
817 if (error)
818 goto cleanup;
819 ea_idebug(inode, "creating block %d", block);
821 new_bh = sb_getblk(sb, block);
822 if (!new_bh) {
823 getblk_failed:
824 ext4_free_blocks(handle, inode, block, 1, 1);
825 error = -EIO;
826 goto cleanup;
828 lock_buffer(new_bh);
829 error = ext4_journal_get_create_access(handle, new_bh);
830 if (error) {
831 unlock_buffer(new_bh);
832 goto getblk_failed;
834 memcpy(new_bh->b_data, s->base, new_bh->b_size);
835 set_buffer_uptodate(new_bh);
836 unlock_buffer(new_bh);
837 ext4_xattr_cache_insert(new_bh);
838 error = ext4_handle_dirty_metadata(handle,
839 inode, new_bh);
840 if (error)
841 goto cleanup;
845 /* Update the inode. */
846 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
848 /* Drop the previous xattr block. */
849 if (bs->bh && bs->bh != new_bh)
850 ext4_xattr_release_block(handle, inode, bs->bh);
851 error = 0;
853 cleanup:
854 if (ce)
855 mb_cache_entry_release(ce);
856 brelse(new_bh);
857 if (!(bs->bh && s->base == bs->bh->b_data))
858 kfree(s->base);
860 return error;
862 cleanup_dquot:
863 vfs_dq_free_block(inode, 1);
864 goto cleanup;
866 bad_block:
867 ext4_error(inode->i_sb, __func__,
868 "inode %lu: bad block %llu", inode->i_ino,
869 EXT4_I(inode)->i_file_acl);
870 goto cleanup;
872 #undef header
875 struct ext4_xattr_ibody_find {
876 struct ext4_xattr_search s;
877 struct ext4_iloc iloc;
880 static int
881 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
882 struct ext4_xattr_ibody_find *is)
884 struct ext4_xattr_ibody_header *header;
885 struct ext4_inode *raw_inode;
886 int error;
888 if (EXT4_I(inode)->i_extra_isize == 0)
889 return 0;
890 raw_inode = ext4_raw_inode(&is->iloc);
891 header = IHDR(inode, raw_inode);
892 is->s.base = is->s.first = IFIRST(header);
893 is->s.here = is->s.first;
894 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
895 if (EXT4_I(inode)->i_state & EXT4_STATE_XATTR) {
896 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
897 if (error)
898 return error;
899 /* Find the named attribute. */
900 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
901 i->name, is->s.end -
902 (void *)is->s.base, 0);
903 if (error && error != -ENODATA)
904 return error;
905 is->s.not_found = error;
907 return 0;
910 static int
911 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
912 struct ext4_xattr_info *i,
913 struct ext4_xattr_ibody_find *is)
915 struct ext4_xattr_ibody_header *header;
916 struct ext4_xattr_search *s = &is->s;
917 int error;
919 if (EXT4_I(inode)->i_extra_isize == 0)
920 return -ENOSPC;
921 error = ext4_xattr_set_entry(i, s);
922 if (error)
923 return error;
924 header = IHDR(inode, ext4_raw_inode(&is->iloc));
925 if (!IS_LAST_ENTRY(s->first)) {
926 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
927 EXT4_I(inode)->i_state |= EXT4_STATE_XATTR;
928 } else {
929 header->h_magic = cpu_to_le32(0);
930 EXT4_I(inode)->i_state &= ~EXT4_STATE_XATTR;
932 return 0;
936 * ext4_xattr_set_handle()
938 * Create, replace or remove an extended attribute for this inode. Buffer
939 * is NULL to remove an existing extended attribute, and non-NULL to
940 * either replace an existing extended attribute, or create a new extended
941 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
942 * specify that an extended attribute must exist and must not exist
943 * previous to the call, respectively.
945 * Returns 0, or a negative error number on failure.
948 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
949 const char *name, const void *value, size_t value_len,
950 int flags)
952 struct ext4_xattr_info i = {
953 .name_index = name_index,
954 .name = name,
955 .value = value,
956 .value_len = value_len,
959 struct ext4_xattr_ibody_find is = {
960 .s = { .not_found = -ENODATA, },
962 struct ext4_xattr_block_find bs = {
963 .s = { .not_found = -ENODATA, },
965 unsigned long no_expand;
966 int error;
968 if (!name)
969 return -EINVAL;
970 if (strlen(name) > 255)
971 return -ERANGE;
972 down_write(&EXT4_I(inode)->xattr_sem);
973 no_expand = EXT4_I(inode)->i_state & EXT4_STATE_NO_EXPAND;
974 EXT4_I(inode)->i_state |= EXT4_STATE_NO_EXPAND;
976 error = ext4_get_inode_loc(inode, &is.iloc);
977 if (error)
978 goto cleanup;
980 if (EXT4_I(inode)->i_state & EXT4_STATE_NEW) {
981 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
982 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
983 EXT4_I(inode)->i_state &= ~EXT4_STATE_NEW;
986 error = ext4_xattr_ibody_find(inode, &i, &is);
987 if (error)
988 goto cleanup;
989 if (is.s.not_found)
990 error = ext4_xattr_block_find(inode, &i, &bs);
991 if (error)
992 goto cleanup;
993 if (is.s.not_found && bs.s.not_found) {
994 error = -ENODATA;
995 if (flags & XATTR_REPLACE)
996 goto cleanup;
997 error = 0;
998 if (!value)
999 goto cleanup;
1000 } else {
1001 error = -EEXIST;
1002 if (flags & XATTR_CREATE)
1003 goto cleanup;
1005 error = ext4_journal_get_write_access(handle, is.iloc.bh);
1006 if (error)
1007 goto cleanup;
1008 if (!value) {
1009 if (!is.s.not_found)
1010 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1011 else if (!bs.s.not_found)
1012 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1013 } else {
1014 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1015 if (!error && !bs.s.not_found) {
1016 i.value = NULL;
1017 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1018 } else if (error == -ENOSPC) {
1019 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1020 error = ext4_xattr_block_find(inode, &i, &bs);
1021 if (error)
1022 goto cleanup;
1024 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1025 if (error)
1026 goto cleanup;
1027 if (!is.s.not_found) {
1028 i.value = NULL;
1029 error = ext4_xattr_ibody_set(handle, inode, &i,
1030 &is);
1034 if (!error) {
1035 ext4_xattr_update_super_block(handle, inode->i_sb);
1036 inode->i_ctime = ext4_current_time(inode);
1037 if (!value)
1038 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1039 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1041 * The bh is consumed by ext4_mark_iloc_dirty, even with
1042 * error != 0.
1044 is.iloc.bh = NULL;
1045 if (IS_SYNC(inode))
1046 ext4_handle_sync(handle);
1049 cleanup:
1050 brelse(is.iloc.bh);
1051 brelse(bs.bh);
1052 if (no_expand == 0)
1053 EXT4_I(inode)->i_state &= ~EXT4_STATE_NO_EXPAND;
1054 up_write(&EXT4_I(inode)->xattr_sem);
1055 return error;
1059 * ext4_xattr_set()
1061 * Like ext4_xattr_set_handle, but start from an inode. This extended
1062 * attribute modification is a filesystem transaction by itself.
1064 * Returns 0, or a negative error number on failure.
1067 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1068 const void *value, size_t value_len, int flags)
1070 handle_t *handle;
1071 int error, retries = 0;
1073 retry:
1074 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1075 if (IS_ERR(handle)) {
1076 error = PTR_ERR(handle);
1077 } else {
1078 int error2;
1080 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1081 value, value_len, flags);
1082 error2 = ext4_journal_stop(handle);
1083 if (error == -ENOSPC &&
1084 ext4_should_retry_alloc(inode->i_sb, &retries))
1085 goto retry;
1086 if (error == 0)
1087 error = error2;
1090 return error;
1094 * Shift the EA entries in the inode to create space for the increased
1095 * i_extra_isize.
1097 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1098 int value_offs_shift, void *to,
1099 void *from, size_t n, int blocksize)
1101 struct ext4_xattr_entry *last = entry;
1102 int new_offs;
1104 /* Adjust the value offsets of the entries */
1105 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1106 if (!last->e_value_block && last->e_value_size) {
1107 new_offs = le16_to_cpu(last->e_value_offs) +
1108 value_offs_shift;
1109 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1110 > blocksize);
1111 last->e_value_offs = cpu_to_le16(new_offs);
1114 /* Shift the entries by n bytes */
1115 memmove(to, from, n);
1119 * Expand an inode by new_extra_isize bytes when EAs are present.
1120 * Returns 0 on success or negative error number on failure.
1122 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1123 struct ext4_inode *raw_inode, handle_t *handle)
1125 struct ext4_xattr_ibody_header *header;
1126 struct ext4_xattr_entry *entry, *last, *first;
1127 struct buffer_head *bh = NULL;
1128 struct ext4_xattr_ibody_find *is = NULL;
1129 struct ext4_xattr_block_find *bs = NULL;
1130 char *buffer = NULL, *b_entry_name = NULL;
1131 size_t min_offs, free;
1132 int total_ino, total_blk;
1133 void *base, *start, *end;
1134 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1135 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1137 down_write(&EXT4_I(inode)->xattr_sem);
1138 retry:
1139 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1140 up_write(&EXT4_I(inode)->xattr_sem);
1141 return 0;
1144 header = IHDR(inode, raw_inode);
1145 entry = IFIRST(header);
1148 * Check if enough free space is available in the inode to shift the
1149 * entries ahead by new_extra_isize.
1152 base = start = entry;
1153 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1154 min_offs = end - base;
1155 last = entry;
1156 total_ino = sizeof(struct ext4_xattr_ibody_header);
1158 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1159 if (free >= new_extra_isize) {
1160 entry = IFIRST(header);
1161 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1162 - new_extra_isize, (void *)raw_inode +
1163 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1164 (void *)header, total_ino,
1165 inode->i_sb->s_blocksize);
1166 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1167 error = 0;
1168 goto cleanup;
1172 * Enough free space isn't available in the inode, check if
1173 * EA block can hold new_extra_isize bytes.
1175 if (EXT4_I(inode)->i_file_acl) {
1176 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1177 error = -EIO;
1178 if (!bh)
1179 goto cleanup;
1180 if (ext4_xattr_check_block(bh)) {
1181 ext4_error(inode->i_sb, __func__,
1182 "inode %lu: bad block %llu", inode->i_ino,
1183 EXT4_I(inode)->i_file_acl);
1184 error = -EIO;
1185 goto cleanup;
1187 base = BHDR(bh);
1188 first = BFIRST(bh);
1189 end = bh->b_data + bh->b_size;
1190 min_offs = end - base;
1191 free = ext4_xattr_free_space(first, &min_offs, base,
1192 &total_blk);
1193 if (free < new_extra_isize) {
1194 if (!tried_min_extra_isize && s_min_extra_isize) {
1195 tried_min_extra_isize++;
1196 new_extra_isize = s_min_extra_isize;
1197 brelse(bh);
1198 goto retry;
1200 error = -1;
1201 goto cleanup;
1203 } else {
1204 free = inode->i_sb->s_blocksize;
1207 while (new_extra_isize > 0) {
1208 size_t offs, size, entry_size;
1209 struct ext4_xattr_entry *small_entry = NULL;
1210 struct ext4_xattr_info i = {
1211 .value = NULL,
1212 .value_len = 0,
1214 unsigned int total_size; /* EA entry size + value size */
1215 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1216 unsigned int min_total_size = ~0U;
1218 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1219 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1220 if (!is || !bs) {
1221 error = -ENOMEM;
1222 goto cleanup;
1225 is->s.not_found = -ENODATA;
1226 bs->s.not_found = -ENODATA;
1227 is->iloc.bh = NULL;
1228 bs->bh = NULL;
1230 last = IFIRST(header);
1231 /* Find the entry best suited to be pushed into EA block */
1232 entry = NULL;
1233 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1234 total_size =
1235 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1236 EXT4_XATTR_LEN(last->e_name_len);
1237 if (total_size <= free && total_size < min_total_size) {
1238 if (total_size < new_extra_isize) {
1239 small_entry = last;
1240 } else {
1241 entry = last;
1242 min_total_size = total_size;
1247 if (entry == NULL) {
1248 if (small_entry) {
1249 entry = small_entry;
1250 } else {
1251 if (!tried_min_extra_isize &&
1252 s_min_extra_isize) {
1253 tried_min_extra_isize++;
1254 new_extra_isize = s_min_extra_isize;
1255 goto retry;
1257 error = -1;
1258 goto cleanup;
1261 offs = le16_to_cpu(entry->e_value_offs);
1262 size = le32_to_cpu(entry->e_value_size);
1263 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1264 i.name_index = entry->e_name_index,
1265 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1266 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1267 if (!buffer || !b_entry_name) {
1268 error = -ENOMEM;
1269 goto cleanup;
1271 /* Save the entry name and the entry value */
1272 memcpy(buffer, (void *)IFIRST(header) + offs,
1273 EXT4_XATTR_SIZE(size));
1274 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1275 b_entry_name[entry->e_name_len] = '\0';
1276 i.name = b_entry_name;
1278 error = ext4_get_inode_loc(inode, &is->iloc);
1279 if (error)
1280 goto cleanup;
1282 error = ext4_xattr_ibody_find(inode, &i, is);
1283 if (error)
1284 goto cleanup;
1286 /* Remove the chosen entry from the inode */
1287 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1289 entry = IFIRST(header);
1290 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1291 shift_bytes = new_extra_isize;
1292 else
1293 shift_bytes = entry_size + size;
1294 /* Adjust the offsets and shift the remaining entries ahead */
1295 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1296 shift_bytes, (void *)raw_inode +
1297 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1298 (void *)header, total_ino - entry_size,
1299 inode->i_sb->s_blocksize);
1301 extra_isize += shift_bytes;
1302 new_extra_isize -= shift_bytes;
1303 EXT4_I(inode)->i_extra_isize = extra_isize;
1305 i.name = b_entry_name;
1306 i.value = buffer;
1307 i.value_len = size;
1308 error = ext4_xattr_block_find(inode, &i, bs);
1309 if (error)
1310 goto cleanup;
1312 /* Add entry which was removed from the inode into the block */
1313 error = ext4_xattr_block_set(handle, inode, &i, bs);
1314 if (error)
1315 goto cleanup;
1316 kfree(b_entry_name);
1317 kfree(buffer);
1318 brelse(is->iloc.bh);
1319 kfree(is);
1320 kfree(bs);
1322 brelse(bh);
1323 up_write(&EXT4_I(inode)->xattr_sem);
1324 return 0;
1326 cleanup:
1327 kfree(b_entry_name);
1328 kfree(buffer);
1329 if (is)
1330 brelse(is->iloc.bh);
1331 kfree(is);
1332 kfree(bs);
1333 brelse(bh);
1334 up_write(&EXT4_I(inode)->xattr_sem);
1335 return error;
1341 * ext4_xattr_delete_inode()
1343 * Free extended attribute resources associated with this inode. This
1344 * is called immediately before an inode is freed. We have exclusive
1345 * access to the inode.
1347 void
1348 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1350 struct buffer_head *bh = NULL;
1352 if (!EXT4_I(inode)->i_file_acl)
1353 goto cleanup;
1354 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1355 if (!bh) {
1356 ext4_error(inode->i_sb, __func__,
1357 "inode %lu: block %llu read error", inode->i_ino,
1358 EXT4_I(inode)->i_file_acl);
1359 goto cleanup;
1361 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1362 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1363 ext4_error(inode->i_sb, __func__,
1364 "inode %lu: bad block %llu", inode->i_ino,
1365 EXT4_I(inode)->i_file_acl);
1366 goto cleanup;
1368 ext4_xattr_release_block(handle, inode, bh);
1369 EXT4_I(inode)->i_file_acl = 0;
1371 cleanup:
1372 brelse(bh);
1376 * ext4_xattr_put_super()
1378 * This is called when a file system is unmounted.
1380 void
1381 ext4_xattr_put_super(struct super_block *sb)
1383 mb_cache_shrink(sb->s_bdev);
1387 * ext4_xattr_cache_insert()
1389 * Create a new entry in the extended attribute cache, and insert
1390 * it unless such an entry is already in the cache.
1392 * Returns 0, or a negative error number on failure.
1394 static void
1395 ext4_xattr_cache_insert(struct buffer_head *bh)
1397 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1398 struct mb_cache_entry *ce;
1399 int error;
1401 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1402 if (!ce) {
1403 ea_bdebug(bh, "out of memory");
1404 return;
1406 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1407 if (error) {
1408 mb_cache_entry_free(ce);
1409 if (error == -EBUSY) {
1410 ea_bdebug(bh, "already in cache");
1411 error = 0;
1413 } else {
1414 ea_bdebug(bh, "inserting [%x]", (int)hash);
1415 mb_cache_entry_release(ce);
1420 * ext4_xattr_cmp()
1422 * Compare two extended attribute blocks for equality.
1424 * Returns 0 if the blocks are equal, 1 if they differ, and
1425 * a negative error number on errors.
1427 static int
1428 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1429 struct ext4_xattr_header *header2)
1431 struct ext4_xattr_entry *entry1, *entry2;
1433 entry1 = ENTRY(header1+1);
1434 entry2 = ENTRY(header2+1);
1435 while (!IS_LAST_ENTRY(entry1)) {
1436 if (IS_LAST_ENTRY(entry2))
1437 return 1;
1438 if (entry1->e_hash != entry2->e_hash ||
1439 entry1->e_name_index != entry2->e_name_index ||
1440 entry1->e_name_len != entry2->e_name_len ||
1441 entry1->e_value_size != entry2->e_value_size ||
1442 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1443 return 1;
1444 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1445 return -EIO;
1446 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1447 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1448 le32_to_cpu(entry1->e_value_size)))
1449 return 1;
1451 entry1 = EXT4_XATTR_NEXT(entry1);
1452 entry2 = EXT4_XATTR_NEXT(entry2);
1454 if (!IS_LAST_ENTRY(entry2))
1455 return 1;
1456 return 0;
1460 * ext4_xattr_cache_find()
1462 * Find an identical extended attribute block.
1464 * Returns a pointer to the block found, or NULL if such a block was
1465 * not found or an error occurred.
1467 static struct buffer_head *
1468 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1469 struct mb_cache_entry **pce)
1471 __u32 hash = le32_to_cpu(header->h_hash);
1472 struct mb_cache_entry *ce;
1474 if (!header->h_hash)
1475 return NULL; /* never share */
1476 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1477 again:
1478 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0,
1479 inode->i_sb->s_bdev, hash);
1480 while (ce) {
1481 struct buffer_head *bh;
1483 if (IS_ERR(ce)) {
1484 if (PTR_ERR(ce) == -EAGAIN)
1485 goto again;
1486 break;
1488 bh = sb_bread(inode->i_sb, ce->e_block);
1489 if (!bh) {
1490 ext4_error(inode->i_sb, __func__,
1491 "inode %lu: block %lu read error",
1492 inode->i_ino, (unsigned long) ce->e_block);
1493 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1494 EXT4_XATTR_REFCOUNT_MAX) {
1495 ea_idebug(inode, "block %lu refcount %d>=%d",
1496 (unsigned long) ce->e_block,
1497 le32_to_cpu(BHDR(bh)->h_refcount),
1498 EXT4_XATTR_REFCOUNT_MAX);
1499 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1500 *pce = ce;
1501 return bh;
1503 brelse(bh);
1504 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1506 return NULL;
1509 #define NAME_HASH_SHIFT 5
1510 #define VALUE_HASH_SHIFT 16
1513 * ext4_xattr_hash_entry()
1515 * Compute the hash of an extended attribute.
1517 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1518 struct ext4_xattr_entry *entry)
1520 __u32 hash = 0;
1521 char *name = entry->e_name;
1522 int n;
1524 for (n = 0; n < entry->e_name_len; n++) {
1525 hash = (hash << NAME_HASH_SHIFT) ^
1526 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1527 *name++;
1530 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1531 __le32 *value = (__le32 *)((char *)header +
1532 le16_to_cpu(entry->e_value_offs));
1533 for (n = (le32_to_cpu(entry->e_value_size) +
1534 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1535 hash = (hash << VALUE_HASH_SHIFT) ^
1536 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1537 le32_to_cpu(*value++);
1540 entry->e_hash = cpu_to_le32(hash);
1543 #undef NAME_HASH_SHIFT
1544 #undef VALUE_HASH_SHIFT
1546 #define BLOCK_HASH_SHIFT 16
1549 * ext4_xattr_rehash()
1551 * Re-compute the extended attribute hash value after an entry has changed.
1553 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1554 struct ext4_xattr_entry *entry)
1556 struct ext4_xattr_entry *here;
1557 __u32 hash = 0;
1559 ext4_xattr_hash_entry(header, entry);
1560 here = ENTRY(header+1);
1561 while (!IS_LAST_ENTRY(here)) {
1562 if (!here->e_hash) {
1563 /* Block is not shared if an entry's hash value == 0 */
1564 hash = 0;
1565 break;
1567 hash = (hash << BLOCK_HASH_SHIFT) ^
1568 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1569 le32_to_cpu(here->e_hash);
1570 here = EXT4_XATTR_NEXT(here);
1572 header->h_hash = cpu_to_le32(hash);
1575 #undef BLOCK_HASH_SHIFT
1577 int __init
1578 init_ext4_xattr(void)
1580 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL,
1581 sizeof(struct mb_cache_entry) +
1582 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1583 if (!ext4_xattr_cache)
1584 return -ENOMEM;
1585 return 0;
1588 void
1589 exit_ext4_xattr(void)
1591 if (ext4_xattr_cache)
1592 mb_cache_destroy(ext4_xattr_cache);
1593 ext4_xattr_cache = NULL;