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>,
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 * +------------------+
31 * | entry 2 | | growing downwards
36 * | value 3 | | growing upwards
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.
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
53 #include <linux/init.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"
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); \
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); \
85 # define ea_idebug(f...)
86 # define ea_bdebug(f...)
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 dentry *dentry, char *buffer,
98 static struct mb_cache *ext4_xattr_cache;
100 static const 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,
106 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
107 #ifdef CONFIG_EXT4_FS_SECURITY
108 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
112 const 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,
119 #ifdef CONFIG_EXT4_FS_SECURITY
120 &ext4_xattr_security_handler,
125 static inline const struct xattr_handler *
126 ext4_xattr_handler(int name_index)
128 const 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];
136 * Inode operation listxattr()
138 * dentry->d_inode->i_mutex: don't care
141 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
143 return ext4_xattr_list(dentry, buffer, size);
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)
159 ext4_xattr_check_block(struct buffer_head *bh)
163 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
164 BHDR(bh)->h_blocks != cpu_to_le32(1))
166 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
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)
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;
191 name_len = strlen(name);
193 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
194 cmp = name_index - entry->e_name_index;
196 cmp = name_len - entry->e_name_len;
198 cmp = memcmp(name, entry->e_name, name_len);
199 if (cmp <= 0 && (sorted || cmp == 0))
203 if (!cmp && ext4_xattr_check_entry(entry, size))
205 return cmp ? -ENODATA : 0;
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;
217 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
218 name_index, name, buffer, (long)buffer_size);
221 if (!EXT4_I(inode)->i_file_acl)
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);
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)) {
231 EXT4_ERROR_INODE(inode, "bad block %llu",
232 EXT4_I(inode)->i_file_acl);
236 ext4_xattr_cache_insert(bh);
238 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
243 size = le32_to_cpu(entry->e_value_size);
246 if (size > buffer_size)
248 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
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;
270 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
272 error = ext4_get_inode_loc(inode, &iloc);
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);
282 error = ext4_xattr_find_entry(&entry, name_index, name,
283 end - (void *)entry, 0);
286 size = le32_to_cpu(entry->e_value_size);
289 if (size > buffer_size)
291 memcpy(buffer, (void *)IFIRST(header) +
292 le16_to_cpu(entry->e_value_offs), size);
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)
317 down_read(&EXT4_I(inode)->xattr_sem);
318 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
320 if (error == -ENODATA)
321 error = ext4_xattr_block_get(inode, name_index, name, buffer,
323 up_read(&EXT4_I(inode)->xattr_sem);
328 ext4_xattr_list_entries(struct dentry *dentry, 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 const struct xattr_handler *handler =
335 ext4_xattr_handler(entry->e_name_index);
338 size_t size = handler->list(dentry, buffer, rest,
350 return buffer_size - rest;
354 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
356 struct inode *inode = dentry->d_inode;
357 struct buffer_head *bh = NULL;
360 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
361 buffer, (long)buffer_size);
364 if (!EXT4_I(inode)->i_file_acl)
366 ea_idebug(inode, "reading block %u", EXT4_I(inode)->i_file_acl);
367 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
371 ea_bdebug(bh, "b_count=%d, refcount=%d",
372 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
373 if (ext4_xattr_check_block(bh)) {
374 EXT4_ERROR_INODE(inode, "bad block %llu",
375 EXT4_I(inode)->i_file_acl);
379 ext4_xattr_cache_insert(bh);
380 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
389 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
391 struct inode *inode = dentry->d_inode;
392 struct ext4_xattr_ibody_header *header;
393 struct ext4_inode *raw_inode;
394 struct ext4_iloc iloc;
398 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
400 error = ext4_get_inode_loc(inode, &iloc);
403 raw_inode = ext4_raw_inode(&iloc);
404 header = IHDR(inode, raw_inode);
405 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
406 error = ext4_xattr_check_names(IFIRST(header), end);
409 error = ext4_xattr_list_entries(dentry, IFIRST(header),
410 buffer, buffer_size);
420 * Copy a list of attribute names into the buffer
421 * provided, or compute the buffer size required.
422 * Buffer is NULL to compute the size of the buffer required.
424 * Returns a negative error number on failure, or the number of bytes
425 * used / required on success.
428 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
432 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
433 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
440 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
445 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
450 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
453 static void ext4_xattr_update_super_block(handle_t *handle,
454 struct super_block *sb)
456 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
459 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
460 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
461 ext4_handle_dirty_super(handle, sb);
466 * Release the xattr block BH: If the reference count is > 1, decrement
467 * it; otherwise free the block.
470 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
471 struct buffer_head *bh)
473 struct mb_cache_entry *ce = NULL;
476 ce = mb_cache_entry_get(ext4_xattr_cache, bh->b_bdev, bh->b_blocknr);
477 error = ext4_journal_get_write_access(handle, bh);
482 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
483 ea_bdebug(bh, "refcount now=0; freeing");
485 mb_cache_entry_free(ce);
487 ext4_free_blocks(handle, inode, bh, 0, 1,
488 EXT4_FREE_BLOCKS_METADATA |
489 EXT4_FREE_BLOCKS_FORGET);
492 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
494 mb_cache_entry_release(ce);
496 error = ext4_handle_dirty_metadata(handle, inode, bh);
498 ext4_handle_sync(handle);
499 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
500 ea_bdebug(bh, "refcount now=%d; releasing",
501 le32_to_cpu(BHDR(bh)->h_refcount));
504 ext4_std_error(inode->i_sb, error);
509 * Find the available free space for EAs. This also returns the total number of
510 * bytes used by EA entries.
512 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
513 size_t *min_offs, void *base, int *total)
515 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
516 *total += EXT4_XATTR_LEN(last->e_name_len);
517 if (!last->e_value_block && last->e_value_size) {
518 size_t offs = le16_to_cpu(last->e_value_offs);
519 if (offs < *min_offs)
523 return (*min_offs - ((void *)last - base) - sizeof(__u32));
526 struct ext4_xattr_info {
533 struct ext4_xattr_search {
534 struct ext4_xattr_entry *first;
537 struct ext4_xattr_entry *here;
542 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
544 struct ext4_xattr_entry *last;
545 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
547 /* Compute min_offs and last. */
549 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
550 if (!last->e_value_block && last->e_value_size) {
551 size_t offs = le16_to_cpu(last->e_value_offs);
556 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
558 if (!s->here->e_value_block && s->here->e_value_size) {
559 size_t size = le32_to_cpu(s->here->e_value_size);
560 free += EXT4_XATTR_SIZE(size);
562 free += EXT4_XATTR_LEN(name_len);
565 if (free < EXT4_XATTR_SIZE(i->value_len) ||
566 free < EXT4_XATTR_LEN(name_len) +
567 EXT4_XATTR_SIZE(i->value_len))
571 if (i->value && s->not_found) {
572 /* Insert the new name. */
573 size_t size = EXT4_XATTR_LEN(name_len);
574 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
575 memmove((void *)s->here + size, s->here, rest);
576 memset(s->here, 0, size);
577 s->here->e_name_index = i->name_index;
578 s->here->e_name_len = name_len;
579 memcpy(s->here->e_name, i->name, name_len);
581 if (!s->here->e_value_block && s->here->e_value_size) {
582 void *first_val = s->base + min_offs;
583 size_t offs = le16_to_cpu(s->here->e_value_offs);
584 void *val = s->base + offs;
585 size_t size = EXT4_XATTR_SIZE(
586 le32_to_cpu(s->here->e_value_size));
588 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
589 /* The old and the new value have the same
590 size. Just replace. */
591 s->here->e_value_size =
592 cpu_to_le32(i->value_len);
593 memset(val + size - EXT4_XATTR_PAD, 0,
594 EXT4_XATTR_PAD); /* Clear pad bytes. */
595 memcpy(val, i->value, i->value_len);
599 /* Remove the old value. */
600 memmove(first_val + size, first_val, val - first_val);
601 memset(first_val, 0, size);
602 s->here->e_value_size = 0;
603 s->here->e_value_offs = 0;
606 /* Adjust all value offsets. */
608 while (!IS_LAST_ENTRY(last)) {
609 size_t o = le16_to_cpu(last->e_value_offs);
610 if (!last->e_value_block &&
611 last->e_value_size && o < offs)
613 cpu_to_le16(o + size);
614 last = EXT4_XATTR_NEXT(last);
618 /* Remove the old name. */
619 size_t size = EXT4_XATTR_LEN(name_len);
620 last = ENTRY((void *)last - size);
621 memmove(s->here, (void *)s->here + size,
622 (void *)last - (void *)s->here + sizeof(__u32));
623 memset(last, 0, size);
628 /* Insert the new value. */
629 s->here->e_value_size = cpu_to_le32(i->value_len);
631 size_t size = EXT4_XATTR_SIZE(i->value_len);
632 void *val = s->base + min_offs - size;
633 s->here->e_value_offs = cpu_to_le16(min_offs - size);
634 memset(val + size - EXT4_XATTR_PAD, 0,
635 EXT4_XATTR_PAD); /* Clear the pad bytes. */
636 memcpy(val, i->value, i->value_len);
642 struct ext4_xattr_block_find {
643 struct ext4_xattr_search s;
644 struct buffer_head *bh;
648 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
649 struct ext4_xattr_block_find *bs)
651 struct super_block *sb = inode->i_sb;
654 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
655 i->name_index, i->name, i->value, (long)i->value_len);
657 if (EXT4_I(inode)->i_file_acl) {
658 /* The inode already has an extended attribute block. */
659 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
663 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
664 atomic_read(&(bs->bh->b_count)),
665 le32_to_cpu(BHDR(bs->bh)->h_refcount));
666 if (ext4_xattr_check_block(bs->bh)) {
667 EXT4_ERROR_INODE(inode, "bad block %llu",
668 EXT4_I(inode)->i_file_acl);
672 /* Find the named attribute. */
673 bs->s.base = BHDR(bs->bh);
674 bs->s.first = BFIRST(bs->bh);
675 bs->s.end = bs->bh->b_data + bs->bh->b_size;
676 bs->s.here = bs->s.first;
677 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
678 i->name, bs->bh->b_size, 1);
679 if (error && error != -ENODATA)
681 bs->s.not_found = error;
690 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
691 struct ext4_xattr_info *i,
692 struct ext4_xattr_block_find *bs)
694 struct super_block *sb = inode->i_sb;
695 struct buffer_head *new_bh = NULL;
696 struct ext4_xattr_search *s = &bs->s;
697 struct mb_cache_entry *ce = NULL;
700 #define header(x) ((struct ext4_xattr_header *)(x))
702 if (i->value && i->value_len > sb->s_blocksize)
705 ce = mb_cache_entry_get(ext4_xattr_cache, bs->bh->b_bdev,
707 error = ext4_journal_get_write_access(handle, bs->bh);
712 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
714 mb_cache_entry_free(ce);
717 ea_bdebug(bs->bh, "modifying in-place");
718 error = ext4_xattr_set_entry(i, s);
720 if (!IS_LAST_ENTRY(s->first))
721 ext4_xattr_rehash(header(s->base),
723 ext4_xattr_cache_insert(bs->bh);
725 unlock_buffer(bs->bh);
729 error = ext4_handle_dirty_metadata(handle,
736 int offset = (char *)s->here - bs->bh->b_data;
738 unlock_buffer(bs->bh);
739 ext4_handle_release_buffer(handle, bs->bh);
741 mb_cache_entry_release(ce);
744 ea_bdebug(bs->bh, "cloning");
745 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
749 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
750 s->first = ENTRY(header(s->base)+1);
751 header(s->base)->h_refcount = cpu_to_le32(1);
752 s->here = ENTRY(s->base + offset);
753 s->end = s->base + bs->bh->b_size;
756 /* Allocate a buffer where we construct the new block. */
757 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
758 /* assert(header == s->base) */
762 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
763 header(s->base)->h_blocks = cpu_to_le32(1);
764 header(s->base)->h_refcount = cpu_to_le32(1);
765 s->first = ENTRY(header(s->base)+1);
766 s->here = ENTRY(header(s->base)+1);
767 s->end = s->base + sb->s_blocksize;
770 error = ext4_xattr_set_entry(i, s);
775 if (!IS_LAST_ENTRY(s->first))
776 ext4_xattr_rehash(header(s->base), s->here);
779 if (!IS_LAST_ENTRY(s->first)) {
780 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
782 /* We found an identical block in the cache. */
783 if (new_bh == bs->bh)
784 ea_bdebug(new_bh, "keeping");
786 /* The old block is released after updating
788 error = dquot_alloc_block(inode,
789 EXT4_C2B(EXT4_SB(sb), 1));
792 error = ext4_journal_get_write_access(handle,
797 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
798 ea_bdebug(new_bh, "reusing; refcount now=%d",
799 le32_to_cpu(BHDR(new_bh)->h_refcount));
800 unlock_buffer(new_bh);
801 error = ext4_handle_dirty_metadata(handle,
807 mb_cache_entry_release(ce);
809 } else if (bs->bh && s->base == bs->bh->b_data) {
810 /* We were modifying this block in-place. */
811 ea_bdebug(bs->bh, "keeping this block");
815 /* We need to allocate a new block */
816 ext4_fsblk_t goal, block;
818 goal = ext4_group_first_block_no(sb,
819 EXT4_I(inode)->i_block_group);
821 /* non-extent files can't have physical blocks past 2^32 */
822 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
823 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
826 * take i_data_sem because we will test
827 * i_delalloc_reserved_flag in ext4_mb_new_blocks
829 down_read((&EXT4_I(inode)->i_data_sem));
830 block = ext4_new_meta_blocks(handle, inode, goal, 0,
832 up_read((&EXT4_I(inode)->i_data_sem));
836 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
837 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
839 ea_idebug(inode, "creating block %d", block);
841 new_bh = sb_getblk(sb, block);
845 ext4_free_blocks(handle, inode, NULL, block, 1,
846 EXT4_FREE_BLOCKS_METADATA);
850 error = ext4_journal_get_create_access(handle, new_bh);
852 unlock_buffer(new_bh);
856 memcpy(new_bh->b_data, s->base, new_bh->b_size);
857 set_buffer_uptodate(new_bh);
858 unlock_buffer(new_bh);
859 ext4_xattr_cache_insert(new_bh);
860 error = ext4_handle_dirty_metadata(handle,
867 /* Update the inode. */
868 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
870 /* Drop the previous xattr block. */
871 if (bs->bh && bs->bh != new_bh)
872 ext4_xattr_release_block(handle, inode, bs->bh);
877 mb_cache_entry_release(ce);
879 if (!(bs->bh && s->base == bs->bh->b_data))
885 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
889 EXT4_ERROR_INODE(inode, "bad block %llu",
890 EXT4_I(inode)->i_file_acl);
896 struct ext4_xattr_ibody_find {
897 struct ext4_xattr_search s;
898 struct ext4_iloc iloc;
902 ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
903 struct ext4_xattr_ibody_find *is)
905 struct ext4_xattr_ibody_header *header;
906 struct ext4_inode *raw_inode;
909 if (EXT4_I(inode)->i_extra_isize == 0)
911 raw_inode = ext4_raw_inode(&is->iloc);
912 header = IHDR(inode, raw_inode);
913 is->s.base = is->s.first = IFIRST(header);
914 is->s.here = is->s.first;
915 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
916 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
917 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
920 /* Find the named attribute. */
921 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
923 (void *)is->s.base, 0);
924 if (error && error != -ENODATA)
926 is->s.not_found = error;
932 ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
933 struct ext4_xattr_info *i,
934 struct ext4_xattr_ibody_find *is)
936 struct ext4_xattr_ibody_header *header;
937 struct ext4_xattr_search *s = &is->s;
940 if (EXT4_I(inode)->i_extra_isize == 0)
942 error = ext4_xattr_set_entry(i, s);
945 header = IHDR(inode, ext4_raw_inode(&is->iloc));
946 if (!IS_LAST_ENTRY(s->first)) {
947 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
948 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
950 header->h_magic = cpu_to_le32(0);
951 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
957 * ext4_xattr_set_handle()
959 * Create, replace or remove an extended attribute for this inode. Value
960 * is NULL to remove an existing extended attribute, and non-NULL to
961 * either replace an existing extended attribute, or create a new extended
962 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
963 * specify that an extended attribute must exist and must not exist
964 * previous to the call, respectively.
966 * Returns 0, or a negative error number on failure.
969 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
970 const char *name, const void *value, size_t value_len,
973 struct ext4_xattr_info i = {
974 .name_index = name_index,
977 .value_len = value_len,
980 struct ext4_xattr_ibody_find is = {
981 .s = { .not_found = -ENODATA, },
983 struct ext4_xattr_block_find bs = {
984 .s = { .not_found = -ENODATA, },
986 unsigned long no_expand;
991 if (strlen(name) > 255)
993 down_write(&EXT4_I(inode)->xattr_sem);
994 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
995 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
997 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1001 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1002 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1003 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1004 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1007 error = ext4_xattr_ibody_find(inode, &i, &is);
1011 error = ext4_xattr_block_find(inode, &i, &bs);
1014 if (is.s.not_found && bs.s.not_found) {
1016 if (flags & XATTR_REPLACE)
1023 if (flags & XATTR_CREATE)
1027 if (!is.s.not_found)
1028 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1029 else if (!bs.s.not_found)
1030 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1032 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1033 if (!error && !bs.s.not_found) {
1035 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1036 } else if (error == -ENOSPC) {
1037 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1038 error = ext4_xattr_block_find(inode, &i, &bs);
1042 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1045 if (!is.s.not_found) {
1047 error = ext4_xattr_ibody_set(handle, inode, &i,
1053 ext4_xattr_update_super_block(handle, inode->i_sb);
1054 inode->i_ctime = ext4_current_time(inode);
1056 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1057 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1059 * The bh is consumed by ext4_mark_iloc_dirty, even with
1064 ext4_handle_sync(handle);
1071 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1072 up_write(&EXT4_I(inode)->xattr_sem);
1079 * Like ext4_xattr_set_handle, but start from an inode. This extended
1080 * attribute modification is a filesystem transaction by itself.
1082 * Returns 0, or a negative error number on failure.
1085 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1086 const void *value, size_t value_len, int flags)
1089 int error, retries = 0;
1092 handle = ext4_journal_start(inode, EXT4_DATA_TRANS_BLOCKS(inode->i_sb));
1093 if (IS_ERR(handle)) {
1094 error = PTR_ERR(handle);
1098 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1099 value, value_len, flags);
1100 error2 = ext4_journal_stop(handle);
1101 if (error == -ENOSPC &&
1102 ext4_should_retry_alloc(inode->i_sb, &retries))
1112 * Shift the EA entries in the inode to create space for the increased
1115 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1116 int value_offs_shift, void *to,
1117 void *from, size_t n, int blocksize)
1119 struct ext4_xattr_entry *last = entry;
1122 /* Adjust the value offsets of the entries */
1123 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1124 if (!last->e_value_block && last->e_value_size) {
1125 new_offs = le16_to_cpu(last->e_value_offs) +
1127 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1129 last->e_value_offs = cpu_to_le16(new_offs);
1132 /* Shift the entries by n bytes */
1133 memmove(to, from, n);
1137 * Expand an inode by new_extra_isize bytes when EAs are present.
1138 * Returns 0 on success or negative error number on failure.
1140 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1141 struct ext4_inode *raw_inode, handle_t *handle)
1143 struct ext4_xattr_ibody_header *header;
1144 struct ext4_xattr_entry *entry, *last, *first;
1145 struct buffer_head *bh = NULL;
1146 struct ext4_xattr_ibody_find *is = NULL;
1147 struct ext4_xattr_block_find *bs = NULL;
1148 char *buffer = NULL, *b_entry_name = NULL;
1149 size_t min_offs, free;
1150 int total_ino, total_blk;
1151 void *base, *start, *end;
1152 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1153 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1155 down_write(&EXT4_I(inode)->xattr_sem);
1157 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1158 up_write(&EXT4_I(inode)->xattr_sem);
1162 header = IHDR(inode, raw_inode);
1163 entry = IFIRST(header);
1166 * Check if enough free space is available in the inode to shift the
1167 * entries ahead by new_extra_isize.
1170 base = start = entry;
1171 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1172 min_offs = end - base;
1174 total_ino = sizeof(struct ext4_xattr_ibody_header);
1176 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1177 if (free >= new_extra_isize) {
1178 entry = IFIRST(header);
1179 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1180 - new_extra_isize, (void *)raw_inode +
1181 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1182 (void *)header, total_ino,
1183 inode->i_sb->s_blocksize);
1184 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1190 * Enough free space isn't available in the inode, check if
1191 * EA block can hold new_extra_isize bytes.
1193 if (EXT4_I(inode)->i_file_acl) {
1194 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1198 if (ext4_xattr_check_block(bh)) {
1199 EXT4_ERROR_INODE(inode, "bad block %llu",
1200 EXT4_I(inode)->i_file_acl);
1206 end = bh->b_data + bh->b_size;
1207 min_offs = end - base;
1208 free = ext4_xattr_free_space(first, &min_offs, base,
1210 if (free < new_extra_isize) {
1211 if (!tried_min_extra_isize && s_min_extra_isize) {
1212 tried_min_extra_isize++;
1213 new_extra_isize = s_min_extra_isize;
1221 free = inode->i_sb->s_blocksize;
1224 while (new_extra_isize > 0) {
1225 size_t offs, size, entry_size;
1226 struct ext4_xattr_entry *small_entry = NULL;
1227 struct ext4_xattr_info i = {
1231 unsigned int total_size; /* EA entry size + value size */
1232 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1233 unsigned int min_total_size = ~0U;
1235 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1236 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1242 is->s.not_found = -ENODATA;
1243 bs->s.not_found = -ENODATA;
1247 last = IFIRST(header);
1248 /* Find the entry best suited to be pushed into EA block */
1250 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1252 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1253 EXT4_XATTR_LEN(last->e_name_len);
1254 if (total_size <= free && total_size < min_total_size) {
1255 if (total_size < new_extra_isize) {
1259 min_total_size = total_size;
1264 if (entry == NULL) {
1266 entry = small_entry;
1268 if (!tried_min_extra_isize &&
1269 s_min_extra_isize) {
1270 tried_min_extra_isize++;
1271 new_extra_isize = s_min_extra_isize;
1272 kfree(is); is = NULL;
1273 kfree(bs); bs = NULL;
1280 offs = le16_to_cpu(entry->e_value_offs);
1281 size = le32_to_cpu(entry->e_value_size);
1282 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1283 i.name_index = entry->e_name_index,
1284 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1285 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1286 if (!buffer || !b_entry_name) {
1290 /* Save the entry name and the entry value */
1291 memcpy(buffer, (void *)IFIRST(header) + offs,
1292 EXT4_XATTR_SIZE(size));
1293 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1294 b_entry_name[entry->e_name_len] = '\0';
1295 i.name = b_entry_name;
1297 error = ext4_get_inode_loc(inode, &is->iloc);
1301 error = ext4_xattr_ibody_find(inode, &i, is);
1305 /* Remove the chosen entry from the inode */
1306 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1310 entry = IFIRST(header);
1311 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1312 shift_bytes = new_extra_isize;
1314 shift_bytes = entry_size + size;
1315 /* Adjust the offsets and shift the remaining entries ahead */
1316 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1317 shift_bytes, (void *)raw_inode +
1318 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1319 (void *)header, total_ino - entry_size,
1320 inode->i_sb->s_blocksize);
1322 extra_isize += shift_bytes;
1323 new_extra_isize -= shift_bytes;
1324 EXT4_I(inode)->i_extra_isize = extra_isize;
1326 i.name = b_entry_name;
1329 error = ext4_xattr_block_find(inode, &i, bs);
1333 /* Add entry which was removed from the inode into the block */
1334 error = ext4_xattr_block_set(handle, inode, &i, bs);
1337 kfree(b_entry_name);
1339 b_entry_name = NULL;
1341 brelse(is->iloc.bh);
1346 up_write(&EXT4_I(inode)->xattr_sem);
1350 kfree(b_entry_name);
1353 brelse(is->iloc.bh);
1357 up_write(&EXT4_I(inode)->xattr_sem);
1364 * ext4_xattr_delete_inode()
1366 * Free extended attribute resources associated with this inode. This
1367 * is called immediately before an inode is freed. We have exclusive
1368 * access to the inode.
1371 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1373 struct buffer_head *bh = NULL;
1375 if (!EXT4_I(inode)->i_file_acl)
1377 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1379 EXT4_ERROR_INODE(inode, "block %llu read error",
1380 EXT4_I(inode)->i_file_acl);
1383 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1384 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1385 EXT4_ERROR_INODE(inode, "bad block %llu",
1386 EXT4_I(inode)->i_file_acl);
1389 ext4_xattr_release_block(handle, inode, bh);
1390 EXT4_I(inode)->i_file_acl = 0;
1397 * ext4_xattr_put_super()
1399 * This is called when a file system is unmounted.
1402 ext4_xattr_put_super(struct super_block *sb)
1404 mb_cache_shrink(sb->s_bdev);
1408 * ext4_xattr_cache_insert()
1410 * Create a new entry in the extended attribute cache, and insert
1411 * it unless such an entry is already in the cache.
1413 * Returns 0, or a negative error number on failure.
1416 ext4_xattr_cache_insert(struct buffer_head *bh)
1418 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1419 struct mb_cache_entry *ce;
1422 ce = mb_cache_entry_alloc(ext4_xattr_cache, GFP_NOFS);
1424 ea_bdebug(bh, "out of memory");
1427 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1429 mb_cache_entry_free(ce);
1430 if (error == -EBUSY) {
1431 ea_bdebug(bh, "already in cache");
1435 ea_bdebug(bh, "inserting [%x]", (int)hash);
1436 mb_cache_entry_release(ce);
1443 * Compare two extended attribute blocks for equality.
1445 * Returns 0 if the blocks are equal, 1 if they differ, and
1446 * a negative error number on errors.
1449 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1450 struct ext4_xattr_header *header2)
1452 struct ext4_xattr_entry *entry1, *entry2;
1454 entry1 = ENTRY(header1+1);
1455 entry2 = ENTRY(header2+1);
1456 while (!IS_LAST_ENTRY(entry1)) {
1457 if (IS_LAST_ENTRY(entry2))
1459 if (entry1->e_hash != entry2->e_hash ||
1460 entry1->e_name_index != entry2->e_name_index ||
1461 entry1->e_name_len != entry2->e_name_len ||
1462 entry1->e_value_size != entry2->e_value_size ||
1463 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1465 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1467 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1468 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1469 le32_to_cpu(entry1->e_value_size)))
1472 entry1 = EXT4_XATTR_NEXT(entry1);
1473 entry2 = EXT4_XATTR_NEXT(entry2);
1475 if (!IS_LAST_ENTRY(entry2))
1481 * ext4_xattr_cache_find()
1483 * Find an identical extended attribute block.
1485 * Returns a pointer to the block found, or NULL if such a block was
1486 * not found or an error occurred.
1488 static struct buffer_head *
1489 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1490 struct mb_cache_entry **pce)
1492 __u32 hash = le32_to_cpu(header->h_hash);
1493 struct mb_cache_entry *ce;
1495 if (!header->h_hash)
1496 return NULL; /* never share */
1497 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1499 ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1502 struct buffer_head *bh;
1505 if (PTR_ERR(ce) == -EAGAIN)
1509 bh = sb_bread(inode->i_sb, ce->e_block);
1511 EXT4_ERROR_INODE(inode, "block %lu read error",
1512 (unsigned long) ce->e_block);
1513 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1514 EXT4_XATTR_REFCOUNT_MAX) {
1515 ea_idebug(inode, "block %lu refcount %d>=%d",
1516 (unsigned long) ce->e_block,
1517 le32_to_cpu(BHDR(bh)->h_refcount),
1518 EXT4_XATTR_REFCOUNT_MAX);
1519 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1524 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1529 #define NAME_HASH_SHIFT 5
1530 #define VALUE_HASH_SHIFT 16
1533 * ext4_xattr_hash_entry()
1535 * Compute the hash of an extended attribute.
1537 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1538 struct ext4_xattr_entry *entry)
1541 char *name = entry->e_name;
1544 for (n = 0; n < entry->e_name_len; n++) {
1545 hash = (hash << NAME_HASH_SHIFT) ^
1546 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1550 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1551 __le32 *value = (__le32 *)((char *)header +
1552 le16_to_cpu(entry->e_value_offs));
1553 for (n = (le32_to_cpu(entry->e_value_size) +
1554 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1555 hash = (hash << VALUE_HASH_SHIFT) ^
1556 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1557 le32_to_cpu(*value++);
1560 entry->e_hash = cpu_to_le32(hash);
1563 #undef NAME_HASH_SHIFT
1564 #undef VALUE_HASH_SHIFT
1566 #define BLOCK_HASH_SHIFT 16
1569 * ext4_xattr_rehash()
1571 * Re-compute the extended attribute hash value after an entry has changed.
1573 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1574 struct ext4_xattr_entry *entry)
1576 struct ext4_xattr_entry *here;
1579 ext4_xattr_hash_entry(header, entry);
1580 here = ENTRY(header+1);
1581 while (!IS_LAST_ENTRY(here)) {
1582 if (!here->e_hash) {
1583 /* Block is not shared if an entry's hash value == 0 */
1587 hash = (hash << BLOCK_HASH_SHIFT) ^
1588 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1589 le32_to_cpu(here->e_hash);
1590 here = EXT4_XATTR_NEXT(here);
1592 header->h_hash = cpu_to_le32(hash);
1595 #undef BLOCK_HASH_SHIFT
1598 ext4_init_xattr(void)
1600 ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1601 if (!ext4_xattr_cache)
1607 ext4_exit_xattr(void)
1609 if (ext4_xattr_cache)
1610 mb_cache_destroy(ext4_xattr_cache);
1611 ext4_xattr_cache = NULL;