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 "ext4_jbd2.h"
63 #ifdef EXT4_XATTR_DEBUG
64 # define ea_idebug(inode, fmt, ...) \
65 printk(KERN_DEBUG "inode %s:%lu: " fmt "\n", \
66 inode->i_sb->s_id, inode->i_ino, ##__VA_ARGS__)
67 # define ea_bdebug(bh, fmt, ...) \
68 printk(KERN_DEBUG "block %pg:%lu: " fmt "\n", \
69 bh->b_bdev, (unsigned long)bh->b_blocknr, ##__VA_ARGS__)
71 # define ea_idebug(inode, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
72 # define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
75 static void ext4_xattr_block_cache_insert(struct mb_cache
*,
76 struct buffer_head
*);
77 static struct buffer_head
*
78 ext4_xattr_block_cache_find(struct inode
*, struct ext4_xattr_header
*,
79 struct mb_cache_entry
**);
80 static __le32
ext4_xattr_hash_entry(char *name
, size_t name_len
, __le32
*value
,
82 static void ext4_xattr_rehash(struct ext4_xattr_header
*);
84 static const struct xattr_handler
* const ext4_xattr_handler_map
[] = {
85 [EXT4_XATTR_INDEX_USER
] = &ext4_xattr_user_handler
,
86 #ifdef CONFIG_EXT4_FS_POSIX_ACL
87 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS
] = &posix_acl_access_xattr_handler
,
88 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT
] = &posix_acl_default_xattr_handler
,
90 [EXT4_XATTR_INDEX_TRUSTED
] = &ext4_xattr_trusted_handler
,
91 #ifdef CONFIG_EXT4_FS_SECURITY
92 [EXT4_XATTR_INDEX_SECURITY
] = &ext4_xattr_security_handler
,
96 const struct xattr_handler
*ext4_xattr_handlers
[] = {
97 &ext4_xattr_user_handler
,
98 &ext4_xattr_trusted_handler
,
99 #ifdef CONFIG_EXT4_FS_POSIX_ACL
100 &posix_acl_access_xattr_handler
,
101 &posix_acl_default_xattr_handler
,
103 #ifdef CONFIG_EXT4_FS_SECURITY
104 &ext4_xattr_security_handler
,
109 #define EA_BLOCK_CACHE(inode) (((struct ext4_sb_info *) \
110 inode->i_sb->s_fs_info)->s_ea_block_cache)
112 #define EA_INODE_CACHE(inode) (((struct ext4_sb_info *) \
113 inode->i_sb->s_fs_info)->s_ea_inode_cache)
116 ext4_expand_inode_array(struct ext4_xattr_inode_array
**ea_inode_array
,
117 struct inode
*inode
);
119 #ifdef CONFIG_LOCKDEP
120 void ext4_xattr_inode_set_class(struct inode
*ea_inode
)
122 lockdep_set_subclass(&ea_inode
->i_rwsem
, 1);
126 static __le32
ext4_xattr_block_csum(struct inode
*inode
,
128 struct ext4_xattr_header
*hdr
)
130 struct ext4_sb_info
*sbi
= EXT4_SB(inode
->i_sb
);
132 __le64 dsk_block_nr
= cpu_to_le64(block_nr
);
133 __u32 dummy_csum
= 0;
134 int offset
= offsetof(struct ext4_xattr_header
, h_checksum
);
136 csum
= ext4_chksum(sbi
, sbi
->s_csum_seed
, (__u8
*)&dsk_block_nr
,
137 sizeof(dsk_block_nr
));
138 csum
= ext4_chksum(sbi
, csum
, (__u8
*)hdr
, offset
);
139 csum
= ext4_chksum(sbi
, csum
, (__u8
*)&dummy_csum
, sizeof(dummy_csum
));
140 offset
+= sizeof(dummy_csum
);
141 csum
= ext4_chksum(sbi
, csum
, (__u8
*)hdr
+ offset
,
142 EXT4_BLOCK_SIZE(inode
->i_sb
) - offset
);
144 return cpu_to_le32(csum
);
147 static int ext4_xattr_block_csum_verify(struct inode
*inode
,
148 struct buffer_head
*bh
)
150 struct ext4_xattr_header
*hdr
= BHDR(bh
);
153 if (ext4_has_metadata_csum(inode
->i_sb
)) {
155 ret
= (hdr
->h_checksum
== ext4_xattr_block_csum(inode
,
156 bh
->b_blocknr
, hdr
));
162 static void ext4_xattr_block_csum_set(struct inode
*inode
,
163 struct buffer_head
*bh
)
165 if (ext4_has_metadata_csum(inode
->i_sb
))
166 BHDR(bh
)->h_checksum
= ext4_xattr_block_csum(inode
,
167 bh
->b_blocknr
, BHDR(bh
));
170 static inline const struct xattr_handler
*
171 ext4_xattr_handler(int name_index
)
173 const struct xattr_handler
*handler
= NULL
;
175 if (name_index
> 0 && name_index
< ARRAY_SIZE(ext4_xattr_handler_map
))
176 handler
= ext4_xattr_handler_map
[name_index
];
181 ext4_xattr_check_entries(struct ext4_xattr_entry
*entry
, void *end
,
184 struct ext4_xattr_entry
*e
= entry
;
186 /* Find the end of the names list */
187 while (!IS_LAST_ENTRY(e
)) {
188 struct ext4_xattr_entry
*next
= EXT4_XATTR_NEXT(e
);
189 if ((void *)next
>= end
)
190 return -EFSCORRUPTED
;
194 /* Check the values */
195 while (!IS_LAST_ENTRY(entry
)) {
196 if (entry
->e_value_size
!= 0 &&
197 entry
->e_value_inum
== 0) {
198 u16 offs
= le16_to_cpu(entry
->e_value_offs
);
199 u32 size
= le32_to_cpu(entry
->e_value_size
);
203 * The value cannot overlap the names, and the value
204 * with padding cannot extend beyond 'end'. Check both
205 * the padded and unpadded sizes, since the size may
206 * overflow to 0 when adding padding.
208 if (offs
> end
- value_start
)
209 return -EFSCORRUPTED
;
210 value
= value_start
+ offs
;
211 if (value
< (void *)e
+ sizeof(u32
) ||
212 size
> end
- value
||
213 EXT4_XATTR_SIZE(size
) > end
- value
)
214 return -EFSCORRUPTED
;
216 entry
= EXT4_XATTR_NEXT(entry
);
223 ext4_xattr_check_block(struct inode
*inode
, struct buffer_head
*bh
)
227 if (buffer_verified(bh
))
230 if (BHDR(bh
)->h_magic
!= cpu_to_le32(EXT4_XATTR_MAGIC
) ||
231 BHDR(bh
)->h_blocks
!= cpu_to_le32(1))
232 return -EFSCORRUPTED
;
233 if (!ext4_xattr_block_csum_verify(inode
, bh
))
235 error
= ext4_xattr_check_entries(BFIRST(bh
), bh
->b_data
+ bh
->b_size
,
238 set_buffer_verified(bh
);
243 __xattr_check_inode(struct inode
*inode
, struct ext4_xattr_ibody_header
*header
,
244 void *end
, const char *function
, unsigned int line
)
246 int error
= -EFSCORRUPTED
;
248 if (end
- (void *)header
< sizeof(*header
) + sizeof(u32
) ||
249 (header
->h_magic
!= cpu_to_le32(EXT4_XATTR_MAGIC
)))
251 error
= ext4_xattr_check_entries(IFIRST(header
), end
, IFIRST(header
));
254 __ext4_error_inode(inode
, function
, line
, 0,
255 "corrupted in-inode xattr");
259 #define xattr_check_inode(inode, header, end) \
260 __xattr_check_inode((inode), (header), (end), __func__, __LINE__)
263 ext4_xattr_find_entry(struct ext4_xattr_entry
**pentry
, int name_index
,
264 const char *name
, int sorted
)
266 struct ext4_xattr_entry
*entry
;
272 name_len
= strlen(name
);
274 for (; !IS_LAST_ENTRY(entry
); entry
= EXT4_XATTR_NEXT(entry
)) {
275 cmp
= name_index
- entry
->e_name_index
;
277 cmp
= name_len
- entry
->e_name_len
;
279 cmp
= memcmp(name
, entry
->e_name
, name_len
);
280 if (cmp
<= 0 && (sorted
|| cmp
== 0))
284 return cmp
? -ENODATA
: 0;
288 ext4_xattr_inode_hash(struct ext4_sb_info
*sbi
, const void *buffer
, size_t size
)
290 return ext4_chksum(sbi
, sbi
->s_csum_seed
, buffer
, size
);
293 static u64
ext4_xattr_inode_get_ref(struct inode
*ea_inode
)
295 return ((u64
)ea_inode
->i_ctime
.tv_sec
<< 32) |
296 ((u32
)ea_inode
->i_version
);
299 static void ext4_xattr_inode_set_ref(struct inode
*ea_inode
, u64 ref_count
)
301 ea_inode
->i_ctime
.tv_sec
= (u32
)(ref_count
>> 32);
302 ea_inode
->i_version
= (u32
)ref_count
;
305 static u32
ext4_xattr_inode_get_hash(struct inode
*ea_inode
)
307 return (u32
)ea_inode
->i_atime
.tv_sec
;
310 static void ext4_xattr_inode_set_hash(struct inode
*ea_inode
, u32 hash
)
312 ea_inode
->i_atime
.tv_sec
= hash
;
316 * Read the EA value from an inode.
318 static int ext4_xattr_inode_read(struct inode
*ea_inode
, void *buf
, size_t size
)
320 unsigned long block
= 0;
321 struct buffer_head
*bh
;
322 int blocksize
= ea_inode
->i_sb
->s_blocksize
;
323 size_t csize
, copied
= 0;
324 void *copy_pos
= buf
;
326 while (copied
< size
) {
327 csize
= (size
- copied
) > blocksize
? blocksize
: size
- copied
;
328 bh
= ext4_bread(NULL
, ea_inode
, block
, 0);
332 return -EFSCORRUPTED
;
334 memcpy(copy_pos
, bh
->b_data
, csize
);
344 static int ext4_xattr_inode_iget(struct inode
*parent
, unsigned long ea_ino
,
345 struct inode
**ea_inode
)
350 inode
= ext4_iget(parent
->i_sb
, ea_ino
);
352 err
= PTR_ERR(inode
);
353 ext4_error(parent
->i_sb
,
354 "error while reading EA inode %lu err=%d", ea_ino
,
359 if (is_bad_inode(inode
)) {
360 ext4_error(parent
->i_sb
,
361 "error while reading EA inode %lu is_bad_inode",
367 if (!(EXT4_I(inode
)->i_flags
& EXT4_EA_INODE_FL
)) {
368 ext4_error(parent
->i_sb
,
369 "EA inode %lu does not have EXT4_EA_INODE_FL flag",
383 ext4_xattr_inode_verify_hashes(struct inode
*ea_inode
,
384 struct ext4_xattr_entry
*entry
, void *buffer
,
389 /* Verify stored hash matches calculated hash. */
390 hash
= ext4_xattr_inode_hash(EXT4_SB(ea_inode
->i_sb
), buffer
, size
);
391 if (hash
!= ext4_xattr_inode_get_hash(ea_inode
))
392 return -EFSCORRUPTED
;
395 __le32 e_hash
, tmp_data
;
397 /* Verify entry hash. */
398 tmp_data
= cpu_to_le32(hash
);
399 e_hash
= ext4_xattr_hash_entry(entry
->e_name
, entry
->e_name_len
,
401 if (e_hash
!= entry
->e_hash
)
402 return -EFSCORRUPTED
;
407 #define EXT4_XATTR_INODE_GET_PARENT(inode) ((__u32)(inode)->i_mtime.tv_sec)
410 * Read xattr value from the EA inode.
413 ext4_xattr_inode_get(struct inode
*inode
, struct ext4_xattr_entry
*entry
,
414 void *buffer
, size_t size
)
416 struct mb_cache
*ea_inode_cache
= EA_INODE_CACHE(inode
);
417 struct inode
*ea_inode
;
420 err
= ext4_xattr_inode_iget(inode
, le32_to_cpu(entry
->e_value_inum
),
427 if (i_size_read(ea_inode
) != size
) {
428 ext4_warning_inode(ea_inode
,
429 "ea_inode file size=%llu entry size=%zu",
430 i_size_read(ea_inode
), size
);
435 err
= ext4_xattr_inode_read(ea_inode
, buffer
, size
);
439 err
= ext4_xattr_inode_verify_hashes(ea_inode
, entry
, buffer
, size
);
441 * Compatibility check for old Lustre ea_inode implementation. Old
442 * version does not have hash validation, but it has a backpointer
443 * from ea_inode to the parent inode.
445 if (err
== -EFSCORRUPTED
) {
446 if (EXT4_XATTR_INODE_GET_PARENT(ea_inode
) != inode
->i_ino
||
447 ea_inode
->i_generation
!= inode
->i_generation
) {
448 ext4_warning_inode(ea_inode
,
449 "EA inode hash validation failed");
452 /* Do not add ea_inode to the cache. */
453 ea_inode_cache
= NULL
;
458 mb_cache_entry_create(ea_inode_cache
, GFP_NOFS
,
459 ext4_xattr_inode_get_hash(ea_inode
),
460 ea_inode
->i_ino
, true /* reusable */);
467 ext4_xattr_block_get(struct inode
*inode
, int name_index
, const char *name
,
468 void *buffer
, size_t buffer_size
)
470 struct buffer_head
*bh
= NULL
;
471 struct ext4_xattr_entry
*entry
;
474 struct mb_cache
*ea_block_cache
= EA_BLOCK_CACHE(inode
);
476 ea_idebug(inode
, "name=%d.%s, buffer=%p, buffer_size=%ld",
477 name_index
, name
, buffer
, (long)buffer_size
);
480 if (!EXT4_I(inode
)->i_file_acl
)
482 ea_idebug(inode
, "reading block %llu",
483 (unsigned long long)EXT4_I(inode
)->i_file_acl
);
484 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
487 ea_bdebug(bh
, "b_count=%d, refcount=%d",
488 atomic_read(&(bh
->b_count
)), le32_to_cpu(BHDR(bh
)->h_refcount
));
489 if (ext4_xattr_check_block(inode
, bh
)) {
490 EXT4_ERROR_INODE(inode
, "bad block %llu",
491 EXT4_I(inode
)->i_file_acl
);
492 error
= -EFSCORRUPTED
;
495 ext4_xattr_block_cache_insert(ea_block_cache
, bh
);
497 error
= ext4_xattr_find_entry(&entry
, name_index
, name
, 1);
500 size
= le32_to_cpu(entry
->e_value_size
);
503 if (size
> buffer_size
)
505 if (entry
->e_value_inum
) {
506 error
= ext4_xattr_inode_get(inode
, entry
, buffer
,
511 memcpy(buffer
, bh
->b_data
+
512 le16_to_cpu(entry
->e_value_offs
), size
);
523 ext4_xattr_ibody_get(struct inode
*inode
, int name_index
, const char *name
,
524 void *buffer
, size_t buffer_size
)
526 struct ext4_xattr_ibody_header
*header
;
527 struct ext4_xattr_entry
*entry
;
528 struct ext4_inode
*raw_inode
;
529 struct ext4_iloc iloc
;
534 if (!ext4_test_inode_state(inode
, EXT4_STATE_XATTR
))
536 error
= ext4_get_inode_loc(inode
, &iloc
);
539 raw_inode
= ext4_raw_inode(&iloc
);
540 header
= IHDR(inode
, raw_inode
);
541 end
= (void *)raw_inode
+ EXT4_SB(inode
->i_sb
)->s_inode_size
;
542 error
= xattr_check_inode(inode
, header
, end
);
545 entry
= IFIRST(header
);
546 error
= ext4_xattr_find_entry(&entry
, name_index
, name
, 0);
549 size
= le32_to_cpu(entry
->e_value_size
);
552 if (size
> buffer_size
)
554 if (entry
->e_value_inum
) {
555 error
= ext4_xattr_inode_get(inode
, entry
, buffer
,
560 memcpy(buffer
, (void *)IFIRST(header
) +
561 le16_to_cpu(entry
->e_value_offs
), size
);
574 * Copy an extended attribute into the buffer
575 * provided, or compute the buffer size required.
576 * Buffer is NULL to compute the size of the buffer required.
578 * Returns a negative error number on failure, or the number of bytes
579 * used / required on success.
582 ext4_xattr_get(struct inode
*inode
, int name_index
, const char *name
,
583 void *buffer
, size_t buffer_size
)
587 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode
->i_sb
))))
590 if (strlen(name
) > 255)
593 down_read(&EXT4_I(inode
)->xattr_sem
);
594 error
= ext4_xattr_ibody_get(inode
, name_index
, name
, buffer
,
596 if (error
== -ENODATA
)
597 error
= ext4_xattr_block_get(inode
, name_index
, name
, buffer
,
599 up_read(&EXT4_I(inode
)->xattr_sem
);
604 ext4_xattr_list_entries(struct dentry
*dentry
, struct ext4_xattr_entry
*entry
,
605 char *buffer
, size_t buffer_size
)
607 size_t rest
= buffer_size
;
609 for (; !IS_LAST_ENTRY(entry
); entry
= EXT4_XATTR_NEXT(entry
)) {
610 const struct xattr_handler
*handler
=
611 ext4_xattr_handler(entry
->e_name_index
);
613 if (handler
&& (!handler
->list
|| handler
->list(dentry
))) {
614 const char *prefix
= handler
->prefix
?: handler
->name
;
615 size_t prefix_len
= strlen(prefix
);
616 size_t size
= prefix_len
+ entry
->e_name_len
+ 1;
621 memcpy(buffer
, prefix
, prefix_len
);
622 buffer
+= prefix_len
;
623 memcpy(buffer
, entry
->e_name
, entry
->e_name_len
);
624 buffer
+= entry
->e_name_len
;
630 return buffer_size
- rest
; /* total size */
634 ext4_xattr_block_list(struct dentry
*dentry
, char *buffer
, size_t buffer_size
)
636 struct inode
*inode
= d_inode(dentry
);
637 struct buffer_head
*bh
= NULL
;
640 ea_idebug(inode
, "buffer=%p, buffer_size=%ld",
641 buffer
, (long)buffer_size
);
644 if (!EXT4_I(inode
)->i_file_acl
)
646 ea_idebug(inode
, "reading block %llu",
647 (unsigned long long)EXT4_I(inode
)->i_file_acl
);
648 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
652 ea_bdebug(bh
, "b_count=%d, refcount=%d",
653 atomic_read(&(bh
->b_count
)), le32_to_cpu(BHDR(bh
)->h_refcount
));
654 if (ext4_xattr_check_block(inode
, bh
)) {
655 EXT4_ERROR_INODE(inode
, "bad block %llu",
656 EXT4_I(inode
)->i_file_acl
);
657 error
= -EFSCORRUPTED
;
660 ext4_xattr_block_cache_insert(EA_BLOCK_CACHE(inode
), bh
);
661 error
= ext4_xattr_list_entries(dentry
, BFIRST(bh
), buffer
, buffer_size
);
670 ext4_xattr_ibody_list(struct dentry
*dentry
, char *buffer
, size_t buffer_size
)
672 struct inode
*inode
= d_inode(dentry
);
673 struct ext4_xattr_ibody_header
*header
;
674 struct ext4_inode
*raw_inode
;
675 struct ext4_iloc iloc
;
679 if (!ext4_test_inode_state(inode
, EXT4_STATE_XATTR
))
681 error
= ext4_get_inode_loc(inode
, &iloc
);
684 raw_inode
= ext4_raw_inode(&iloc
);
685 header
= IHDR(inode
, raw_inode
);
686 end
= (void *)raw_inode
+ EXT4_SB(inode
->i_sb
)->s_inode_size
;
687 error
= xattr_check_inode(inode
, header
, end
);
690 error
= ext4_xattr_list_entries(dentry
, IFIRST(header
),
691 buffer
, buffer_size
);
699 * Inode operation listxattr()
701 * d_inode(dentry)->i_rwsem: don't care
703 * Copy a list of attribute names into the buffer
704 * provided, or compute the buffer size required.
705 * Buffer is NULL to compute the size of the buffer required.
707 * Returns a negative error number on failure, or the number of bytes
708 * used / required on success.
711 ext4_listxattr(struct dentry
*dentry
, char *buffer
, size_t buffer_size
)
715 down_read(&EXT4_I(d_inode(dentry
))->xattr_sem
);
716 ret
= ret2
= ext4_xattr_ibody_list(dentry
, buffer
, buffer_size
);
723 ret
= ext4_xattr_block_list(dentry
, buffer
, buffer_size
);
728 up_read(&EXT4_I(d_inode(dentry
))->xattr_sem
);
733 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
736 static void ext4_xattr_update_super_block(handle_t
*handle
,
737 struct super_block
*sb
)
739 if (ext4_has_feature_xattr(sb
))
742 BUFFER_TRACE(EXT4_SB(sb
)->s_sbh
, "get_write_access");
743 if (ext4_journal_get_write_access(handle
, EXT4_SB(sb
)->s_sbh
) == 0) {
744 ext4_set_feature_xattr(sb
);
745 ext4_handle_dirty_super(handle
, sb
);
749 int ext4_get_inode_usage(struct inode
*inode
, qsize_t
*usage
)
751 struct ext4_iloc iloc
= { .bh
= NULL
};
752 struct buffer_head
*bh
= NULL
;
753 struct ext4_inode
*raw_inode
;
754 struct ext4_xattr_ibody_header
*header
;
755 struct ext4_xattr_entry
*entry
;
756 qsize_t ea_inode_refs
= 0;
760 lockdep_assert_held_read(&EXT4_I(inode
)->xattr_sem
);
762 if (ext4_test_inode_state(inode
, EXT4_STATE_XATTR
)) {
763 ret
= ext4_get_inode_loc(inode
, &iloc
);
766 raw_inode
= ext4_raw_inode(&iloc
);
767 header
= IHDR(inode
, raw_inode
);
768 end
= (void *)raw_inode
+ EXT4_SB(inode
->i_sb
)->s_inode_size
;
769 ret
= xattr_check_inode(inode
, header
, end
);
773 for (entry
= IFIRST(header
); !IS_LAST_ENTRY(entry
);
774 entry
= EXT4_XATTR_NEXT(entry
))
775 if (entry
->e_value_inum
)
779 if (EXT4_I(inode
)->i_file_acl
) {
780 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
786 if (ext4_xattr_check_block(inode
, bh
)) {
791 for (entry
= BFIRST(bh
); !IS_LAST_ENTRY(entry
);
792 entry
= EXT4_XATTR_NEXT(entry
))
793 if (entry
->e_value_inum
)
796 *usage
= ea_inode_refs
+ 1;
804 static inline size_t round_up_cluster(struct inode
*inode
, size_t length
)
806 struct super_block
*sb
= inode
->i_sb
;
807 size_t cluster_size
= 1 << (EXT4_SB(sb
)->s_cluster_bits
+
809 size_t mask
= ~(cluster_size
- 1);
811 return (length
+ cluster_size
- 1) & mask
;
814 static int ext4_xattr_inode_alloc_quota(struct inode
*inode
, size_t len
)
818 err
= dquot_alloc_inode(inode
);
821 err
= dquot_alloc_space_nodirty(inode
, round_up_cluster(inode
, len
));
823 dquot_free_inode(inode
);
827 static void ext4_xattr_inode_free_quota(struct inode
*inode
, size_t len
)
829 dquot_free_space_nodirty(inode
, round_up_cluster(inode
, len
));
830 dquot_free_inode(inode
);
833 int __ext4_xattr_set_credits(struct super_block
*sb
, struct inode
*inode
,
834 struct buffer_head
*block_bh
, size_t value_len
,
841 * 1) Owner inode update
842 * 2) Ref count update on old xattr block
844 * 4) block bitmap update for new xattr block
845 * 5) group descriptor for new xattr block
846 * 6) block bitmap update for old xattr block
847 * 7) group descriptor for old block
849 * 6 & 7 can happen if we have two racing threads T_a and T_b
850 * which are each trying to set an xattr on inodes I_a and I_b
851 * which were both initially sharing an xattr block.
856 credits
+= EXT4_MAXQUOTAS_TRANS_BLOCKS(sb
);
859 * In case of inline data, we may push out the data to a block,
860 * so we need to reserve credits for this eventuality
862 if (inode
&& ext4_has_inline_data(inode
))
863 credits
+= ext4_writepage_trans_blocks(inode
) + 1;
865 /* We are done if ea_inode feature is not enabled. */
866 if (!ext4_has_feature_ea_inode(sb
))
869 /* New ea_inode, inode map, block bitmap, group descriptor. */
873 blocks
= (value_len
+ sb
->s_blocksize
- 1) >> sb
->s_blocksize_bits
;
875 /* Indirection block or one level of extent tree. */
878 /* Block bitmap and group descriptor updates for each block. */
879 credits
+= blocks
* 2;
881 /* Blocks themselves. */
885 /* Dereference ea_inode holding old xattr value.
886 * Old ea_inode, inode map, block bitmap, group descriptor.
890 /* Data blocks for old ea_inode. */
891 blocks
= XATTR_SIZE_MAX
>> sb
->s_blocksize_bits
;
893 /* Indirection block or one level of extent tree for old
898 /* Block bitmap and group descriptor updates for each block. */
899 credits
+= blocks
* 2;
902 /* We may need to clone the existing xattr block in which case we need
903 * to increment ref counts for existing ea_inodes referenced by it.
906 struct ext4_xattr_entry
*entry
= BFIRST(block_bh
);
908 for (; !IS_LAST_ENTRY(entry
); entry
= EXT4_XATTR_NEXT(entry
))
909 if (entry
->e_value_inum
)
910 /* Ref count update on ea_inode. */
916 static int ext4_xattr_ensure_credits(handle_t
*handle
, struct inode
*inode
,
917 int credits
, struct buffer_head
*bh
,
918 bool dirty
, bool block_csum
)
922 if (!ext4_handle_valid(handle
))
925 if (handle
->h_buffer_credits
>= credits
)
928 error
= ext4_journal_extend(handle
, credits
- handle
->h_buffer_credits
);
932 ext4_warning(inode
->i_sb
, "Extend journal (error %d)", error
);
938 ext4_xattr_block_csum_set(inode
, bh
);
939 error
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
941 ext4_warning(inode
->i_sb
, "Handle metadata (error %d)",
947 error
= ext4_journal_restart(handle
, credits
);
949 ext4_warning(inode
->i_sb
, "Restart journal (error %d)", error
);
954 error
= ext4_journal_get_write_access(handle
, bh
);
956 ext4_warning(inode
->i_sb
,
957 "Get write access failed (error %d)",
965 static int ext4_xattr_inode_update_ref(handle_t
*handle
, struct inode
*ea_inode
,
968 struct mb_cache
*ea_inode_cache
= EA_INODE_CACHE(ea_inode
);
969 struct ext4_iloc iloc
;
974 inode_lock(ea_inode
);
976 ret
= ext4_reserve_inode_write(handle
, ea_inode
, &iloc
);
982 ref_count
= ext4_xattr_inode_get_ref(ea_inode
);
983 ref_count
+= ref_change
;
984 ext4_xattr_inode_set_ref(ea_inode
, ref_count
);
986 if (ref_change
> 0) {
987 WARN_ONCE(ref_count
<= 0, "EA inode %lu ref_count=%lld",
988 ea_inode
->i_ino
, ref_count
);
990 if (ref_count
== 1) {
991 WARN_ONCE(ea_inode
->i_nlink
, "EA inode %lu i_nlink=%u",
992 ea_inode
->i_ino
, ea_inode
->i_nlink
);
994 set_nlink(ea_inode
, 1);
995 ext4_orphan_del(handle
, ea_inode
);
997 if (ea_inode_cache
) {
998 hash
= ext4_xattr_inode_get_hash(ea_inode
);
999 mb_cache_entry_create(ea_inode_cache
,
1002 true /* reusable */);
1006 WARN_ONCE(ref_count
< 0, "EA inode %lu ref_count=%lld",
1007 ea_inode
->i_ino
, ref_count
);
1009 if (ref_count
== 0) {
1010 WARN_ONCE(ea_inode
->i_nlink
!= 1,
1011 "EA inode %lu i_nlink=%u",
1012 ea_inode
->i_ino
, ea_inode
->i_nlink
);
1014 clear_nlink(ea_inode
);
1015 ext4_orphan_add(handle
, ea_inode
);
1017 if (ea_inode_cache
) {
1018 hash
= ext4_xattr_inode_get_hash(ea_inode
);
1019 mb_cache_entry_delete(ea_inode_cache
, hash
,
1025 ret
= ext4_mark_iloc_dirty(handle
, ea_inode
, &iloc
);
1028 ext4_warning_inode(ea_inode
,
1029 "ext4_mark_iloc_dirty() failed ret=%d", ret
);
1032 inode_unlock(ea_inode
);
1036 static int ext4_xattr_inode_inc_ref(handle_t
*handle
, struct inode
*ea_inode
)
1038 return ext4_xattr_inode_update_ref(handle
, ea_inode
, 1);
1041 static int ext4_xattr_inode_dec_ref(handle_t
*handle
, struct inode
*ea_inode
)
1043 return ext4_xattr_inode_update_ref(handle
, ea_inode
, -1);
1046 static int ext4_xattr_inode_inc_ref_all(handle_t
*handle
, struct inode
*parent
,
1047 struct ext4_xattr_entry
*first
)
1049 struct inode
*ea_inode
;
1050 struct ext4_xattr_entry
*entry
;
1051 struct ext4_xattr_entry
*failed_entry
;
1052 unsigned int ea_ino
;
1055 for (entry
= first
; !IS_LAST_ENTRY(entry
);
1056 entry
= EXT4_XATTR_NEXT(entry
)) {
1057 if (!entry
->e_value_inum
)
1059 ea_ino
= le32_to_cpu(entry
->e_value_inum
);
1060 err
= ext4_xattr_inode_iget(parent
, ea_ino
, &ea_inode
);
1063 err
= ext4_xattr_inode_inc_ref(handle
, ea_inode
);
1065 ext4_warning_inode(ea_inode
, "inc ref error %d", err
);
1075 failed_entry
= entry
;
1077 for (entry
= first
; entry
!= failed_entry
;
1078 entry
= EXT4_XATTR_NEXT(entry
)) {
1079 if (!entry
->e_value_inum
)
1081 ea_ino
= le32_to_cpu(entry
->e_value_inum
);
1082 err
= ext4_xattr_inode_iget(parent
, ea_ino
, &ea_inode
);
1084 ext4_warning(parent
->i_sb
,
1085 "cleanup ea_ino %u iget error %d", ea_ino
,
1089 err
= ext4_xattr_inode_dec_ref(handle
, ea_inode
);
1091 ext4_warning_inode(ea_inode
, "cleanup dec ref error %d",
1099 ext4_xattr_inode_dec_ref_all(handle_t
*handle
, struct inode
*parent
,
1100 struct buffer_head
*bh
,
1101 struct ext4_xattr_entry
*first
, bool block_csum
,
1102 struct ext4_xattr_inode_array
**ea_inode_array
,
1103 int extra_credits
, bool skip_quota
)
1105 struct inode
*ea_inode
;
1106 struct ext4_xattr_entry
*entry
;
1108 unsigned int ea_ino
;
1112 /* One credit for dec ref on ea_inode, one for orphan list addition, */
1113 credits
= 2 + extra_credits
;
1115 for (entry
= first
; !IS_LAST_ENTRY(entry
);
1116 entry
= EXT4_XATTR_NEXT(entry
)) {
1117 if (!entry
->e_value_inum
)
1119 ea_ino
= le32_to_cpu(entry
->e_value_inum
);
1120 err
= ext4_xattr_inode_iget(parent
, ea_ino
, &ea_inode
);
1124 err
= ext4_expand_inode_array(ea_inode_array
, ea_inode
);
1126 ext4_warning_inode(ea_inode
,
1127 "Expand inode array err=%d", err
);
1132 err
= ext4_xattr_ensure_credits(handle
, parent
, credits
, bh
,
1135 ext4_warning_inode(ea_inode
, "Ensure credits err=%d",
1140 err
= ext4_xattr_inode_dec_ref(handle
, ea_inode
);
1142 ext4_warning_inode(ea_inode
, "ea_inode dec ref err=%d",
1148 ext4_xattr_inode_free_quota(parent
,
1149 le32_to_cpu(entry
->e_value_size
));
1152 * Forget about ea_inode within the same transaction that
1153 * decrements the ref count. This avoids duplicate decrements in
1154 * case the rest of the work spills over to subsequent
1157 entry
->e_value_inum
= 0;
1158 entry
->e_value_size
= 0;
1165 * Note that we are deliberately skipping csum calculation for
1166 * the final update because we do not expect any journal
1167 * restarts until xattr block is freed.
1170 err
= ext4_handle_dirty_metadata(handle
, NULL
, bh
);
1172 ext4_warning_inode(parent
,
1173 "handle dirty metadata err=%d", err
);
1178 * Release the xattr block BH: If the reference count is > 1, decrement it;
1179 * otherwise free the block.
1182 ext4_xattr_release_block(handle_t
*handle
, struct inode
*inode
,
1183 struct buffer_head
*bh
,
1184 struct ext4_xattr_inode_array
**ea_inode_array
,
1187 struct mb_cache
*ea_block_cache
= EA_BLOCK_CACHE(inode
);
1191 BUFFER_TRACE(bh
, "get_write_access");
1192 error
= ext4_journal_get_write_access(handle
, bh
);
1197 hash
= le32_to_cpu(BHDR(bh
)->h_hash
);
1198 ref
= le32_to_cpu(BHDR(bh
)->h_refcount
);
1200 ea_bdebug(bh
, "refcount now=0; freeing");
1202 * This must happen under buffer lock for
1203 * ext4_xattr_block_set() to reliably detect freed block
1206 mb_cache_entry_delete(ea_block_cache
, hash
,
1211 if (ext4_has_feature_ea_inode(inode
->i_sb
))
1212 ext4_xattr_inode_dec_ref_all(handle
, inode
, bh
,
1214 true /* block_csum */,
1217 true /* skip_quota */);
1218 ext4_free_blocks(handle
, inode
, bh
, 0, 1,
1219 EXT4_FREE_BLOCKS_METADATA
|
1220 EXT4_FREE_BLOCKS_FORGET
);
1223 BHDR(bh
)->h_refcount
= cpu_to_le32(ref
);
1224 if (ref
== EXT4_XATTR_REFCOUNT_MAX
- 1) {
1225 struct mb_cache_entry
*ce
;
1227 if (ea_block_cache
) {
1228 ce
= mb_cache_entry_get(ea_block_cache
, hash
,
1232 mb_cache_entry_put(ea_block_cache
, ce
);
1237 ext4_xattr_block_csum_set(inode
, bh
);
1239 * Beware of this ugliness: Releasing of xattr block references
1240 * from different inodes can race and so we have to protect
1241 * from a race where someone else frees the block (and releases
1242 * its journal_head) before we are done dirtying the buffer. In
1243 * nojournal mode this race is harmless and we actually cannot
1244 * call ext4_handle_dirty_metadata() with locked buffer as
1245 * that function can call sync_dirty_buffer() so for that case
1246 * we handle the dirtying after unlocking the buffer.
1248 if (ext4_handle_valid(handle
))
1249 error
= ext4_handle_dirty_metadata(handle
, inode
, bh
);
1251 if (!ext4_handle_valid(handle
))
1252 error
= ext4_handle_dirty_metadata(handle
, inode
, bh
);
1254 ext4_handle_sync(handle
);
1255 dquot_free_block(inode
, EXT4_C2B(EXT4_SB(inode
->i_sb
), 1));
1256 ea_bdebug(bh
, "refcount now=%d; releasing",
1257 le32_to_cpu(BHDR(bh
)->h_refcount
));
1260 ext4_std_error(inode
->i_sb
, error
);
1265 * Find the available free space for EAs. This also returns the total number of
1266 * bytes used by EA entries.
1268 static size_t ext4_xattr_free_space(struct ext4_xattr_entry
*last
,
1269 size_t *min_offs
, void *base
, int *total
)
1271 for (; !IS_LAST_ENTRY(last
); last
= EXT4_XATTR_NEXT(last
)) {
1272 if (!last
->e_value_inum
&& last
->e_value_size
) {
1273 size_t offs
= le16_to_cpu(last
->e_value_offs
);
1274 if (offs
< *min_offs
)
1278 *total
+= EXT4_XATTR_LEN(last
->e_name_len
);
1280 return (*min_offs
- ((void *)last
- base
) - sizeof(__u32
));
1284 * Write the value of the EA in an inode.
1286 static int ext4_xattr_inode_write(handle_t
*handle
, struct inode
*ea_inode
,
1287 const void *buf
, int bufsize
)
1289 struct buffer_head
*bh
= NULL
;
1290 unsigned long block
= 0;
1291 int blocksize
= ea_inode
->i_sb
->s_blocksize
;
1292 int max_blocks
= (bufsize
+ blocksize
- 1) >> ea_inode
->i_blkbits
;
1293 int csize
, wsize
= 0;
1298 while (ret
>= 0 && ret
< max_blocks
) {
1299 struct ext4_map_blocks map
;
1300 map
.m_lblk
= block
+= ret
;
1301 map
.m_len
= max_blocks
-= ret
;
1303 ret
= ext4_map_blocks(handle
, ea_inode
, &map
,
1304 EXT4_GET_BLOCKS_CREATE
);
1306 ext4_mark_inode_dirty(handle
, ea_inode
);
1307 if (ret
== -ENOSPC
&&
1308 ext4_should_retry_alloc(ea_inode
->i_sb
, &retries
)) {
1320 while (wsize
< bufsize
) {
1323 csize
= (bufsize
- wsize
) > blocksize
? blocksize
:
1325 bh
= ext4_getblk(handle
, ea_inode
, block
, 0);
1328 ret
= ext4_journal_get_write_access(handle
, bh
);
1332 memcpy(bh
->b_data
, buf
, csize
);
1333 set_buffer_uptodate(bh
);
1334 ext4_handle_dirty_metadata(handle
, ea_inode
, bh
);
1341 inode_lock(ea_inode
);
1342 i_size_write(ea_inode
, wsize
);
1343 ext4_update_i_disksize(ea_inode
, wsize
);
1344 inode_unlock(ea_inode
);
1346 ext4_mark_inode_dirty(handle
, ea_inode
);
1355 * Create an inode to store the value of a large EA.
1357 static struct inode
*ext4_xattr_inode_create(handle_t
*handle
,
1358 struct inode
*inode
, u32 hash
)
1360 struct inode
*ea_inode
= NULL
;
1361 uid_t owner
[2] = { i_uid_read(inode
), i_gid_read(inode
) };
1365 * Let the next inode be the goal, so we try and allocate the EA inode
1366 * in the same group, or nearby one.
1368 ea_inode
= ext4_new_inode(handle
, inode
->i_sb
->s_root
->d_inode
,
1369 S_IFREG
| 0600, NULL
, inode
->i_ino
+ 1, owner
,
1371 if (!IS_ERR(ea_inode
)) {
1372 ea_inode
->i_op
= &ext4_file_inode_operations
;
1373 ea_inode
->i_fop
= &ext4_file_operations
;
1374 ext4_set_aops(ea_inode
);
1375 ext4_xattr_inode_set_class(ea_inode
);
1376 unlock_new_inode(ea_inode
);
1377 ext4_xattr_inode_set_ref(ea_inode
, 1);
1378 ext4_xattr_inode_set_hash(ea_inode
, hash
);
1379 err
= ext4_mark_inode_dirty(handle
, ea_inode
);
1381 err
= ext4_inode_attach_jinode(ea_inode
);
1384 return ERR_PTR(err
);
1388 * Xattr inodes are shared therefore quota charging is performed
1389 * at a higher level.
1391 dquot_free_inode(ea_inode
);
1392 dquot_drop(ea_inode
);
1393 inode_lock(ea_inode
);
1394 ea_inode
->i_flags
|= S_NOQUOTA
;
1395 inode_unlock(ea_inode
);
1401 static struct inode
*
1402 ext4_xattr_inode_cache_find(struct inode
*inode
, const void *value
,
1403 size_t value_len
, u32 hash
)
1405 struct inode
*ea_inode
;
1406 struct mb_cache_entry
*ce
;
1407 struct mb_cache
*ea_inode_cache
= EA_INODE_CACHE(inode
);
1410 if (!ea_inode_cache
)
1413 ce
= mb_cache_entry_find_first(ea_inode_cache
, hash
);
1417 ea_data
= ext4_kvmalloc(value_len
, GFP_NOFS
);
1419 mb_cache_entry_put(ea_inode_cache
, ce
);
1424 ea_inode
= ext4_iget(inode
->i_sb
, ce
->e_value
);
1425 if (!IS_ERR(ea_inode
) &&
1426 !is_bad_inode(ea_inode
) &&
1427 (EXT4_I(ea_inode
)->i_flags
& EXT4_EA_INODE_FL
) &&
1428 i_size_read(ea_inode
) == value_len
&&
1429 !ext4_xattr_inode_read(ea_inode
, ea_data
, value_len
) &&
1430 !ext4_xattr_inode_verify_hashes(ea_inode
, NULL
, ea_data
,
1432 !memcmp(value
, ea_data
, value_len
)) {
1433 mb_cache_entry_touch(ea_inode_cache
, ce
);
1434 mb_cache_entry_put(ea_inode_cache
, ce
);
1439 if (!IS_ERR(ea_inode
))
1441 ce
= mb_cache_entry_find_next(ea_inode_cache
, ce
);
1448 * Add value of the EA in an inode.
1450 static int ext4_xattr_inode_lookup_create(handle_t
*handle
, struct inode
*inode
,
1451 const void *value
, size_t value_len
,
1452 struct inode
**ret_inode
)
1454 struct inode
*ea_inode
;
1458 hash
= ext4_xattr_inode_hash(EXT4_SB(inode
->i_sb
), value
, value_len
);
1459 ea_inode
= ext4_xattr_inode_cache_find(inode
, value
, value_len
, hash
);
1461 err
= ext4_xattr_inode_inc_ref(handle
, ea_inode
);
1467 *ret_inode
= ea_inode
;
1471 /* Create an inode for the EA value */
1472 ea_inode
= ext4_xattr_inode_create(handle
, inode
, hash
);
1473 if (IS_ERR(ea_inode
))
1474 return PTR_ERR(ea_inode
);
1476 err
= ext4_xattr_inode_write(handle
, ea_inode
, value
, value_len
);
1478 ext4_xattr_inode_dec_ref(handle
, ea_inode
);
1483 if (EA_INODE_CACHE(inode
))
1484 mb_cache_entry_create(EA_INODE_CACHE(inode
), GFP_NOFS
, hash
,
1485 ea_inode
->i_ino
, true /* reusable */);
1487 *ret_inode
= ea_inode
;
1492 * Reserve min(block_size/8, 1024) bytes for xattr entries/names if ea_inode
1493 * feature is enabled.
1495 #define EXT4_XATTR_BLOCK_RESERVE(inode) min(i_blocksize(inode)/8, 1024U)
1497 static int ext4_xattr_set_entry(struct ext4_xattr_info
*i
,
1498 struct ext4_xattr_search
*s
,
1499 handle_t
*handle
, struct inode
*inode
,
1502 struct ext4_xattr_entry
*last
;
1503 struct ext4_xattr_entry
*here
= s
->here
;
1504 size_t min_offs
= s
->end
- s
->base
, name_len
= strlen(i
->name
);
1505 int in_inode
= i
->in_inode
;
1506 struct inode
*old_ea_inode
= NULL
;
1507 struct inode
*new_ea_inode
= NULL
;
1508 size_t old_size
, new_size
;
1511 /* Space used by old and new values. */
1512 old_size
= (!s
->not_found
&& !here
->e_value_inum
) ?
1513 EXT4_XATTR_SIZE(le32_to_cpu(here
->e_value_size
)) : 0;
1514 new_size
= (i
->value
&& !in_inode
) ? EXT4_XATTR_SIZE(i
->value_len
) : 0;
1517 * Optimization for the simple case when old and new values have the
1518 * same padded sizes. Not applicable if external inodes are involved.
1520 if (new_size
&& new_size
== old_size
) {
1521 size_t offs
= le16_to_cpu(here
->e_value_offs
);
1522 void *val
= s
->base
+ offs
;
1524 here
->e_value_size
= cpu_to_le32(i
->value_len
);
1525 if (i
->value
== EXT4_ZERO_XATTR_VALUE
) {
1526 memset(val
, 0, new_size
);
1528 memcpy(val
, i
->value
, i
->value_len
);
1529 /* Clear padding bytes. */
1530 memset(val
+ i
->value_len
, 0, new_size
- i
->value_len
);
1535 /* Compute min_offs and last. */
1537 for (; !IS_LAST_ENTRY(last
); last
= EXT4_XATTR_NEXT(last
)) {
1538 if (!last
->e_value_inum
&& last
->e_value_size
) {
1539 size_t offs
= le16_to_cpu(last
->e_value_offs
);
1540 if (offs
< min_offs
)
1545 /* Check whether we have enough space. */
1549 free
= min_offs
- ((void *)last
- s
->base
) - sizeof(__u32
);
1551 free
+= EXT4_XATTR_LEN(name_len
) + old_size
;
1553 if (free
< EXT4_XATTR_LEN(name_len
) + new_size
) {
1559 * If storing the value in an external inode is an option,
1560 * reserve space for xattr entries/names in the external
1561 * attribute block so that a long value does not occupy the
1562 * whole space and prevent futher entries being added.
1564 if (ext4_has_feature_ea_inode(inode
->i_sb
) &&
1565 new_size
&& is_block
&&
1566 (min_offs
+ old_size
- new_size
) <
1567 EXT4_XATTR_BLOCK_RESERVE(inode
)) {
1574 * Getting access to old and new ea inodes is subject to failures.
1575 * Finish that work before doing any modifications to the xattr data.
1577 if (!s
->not_found
&& here
->e_value_inum
) {
1578 ret
= ext4_xattr_inode_iget(inode
,
1579 le32_to_cpu(here
->e_value_inum
),
1582 old_ea_inode
= NULL
;
1586 if (i
->value
&& in_inode
) {
1587 WARN_ON_ONCE(!i
->value_len
);
1589 ret
= ext4_xattr_inode_alloc_quota(inode
, i
->value_len
);
1593 ret
= ext4_xattr_inode_lookup_create(handle
, inode
, i
->value
,
1597 new_ea_inode
= NULL
;
1598 ext4_xattr_inode_free_quota(inode
, i
->value_len
);
1604 /* We are ready to release ref count on the old_ea_inode. */
1605 ret
= ext4_xattr_inode_dec_ref(handle
, old_ea_inode
);
1607 /* Release newly required ref count on new_ea_inode. */
1611 err
= ext4_xattr_inode_dec_ref(handle
,
1614 ext4_warning_inode(new_ea_inode
,
1615 "dec ref new_ea_inode err=%d",
1617 ext4_xattr_inode_free_quota(inode
,
1623 ext4_xattr_inode_free_quota(inode
,
1624 le32_to_cpu(here
->e_value_size
));
1627 /* No failures allowed past this point. */
1629 if (!s
->not_found
&& here
->e_value_offs
) {
1630 /* Remove the old value. */
1631 void *first_val
= s
->base
+ min_offs
;
1632 size_t offs
= le16_to_cpu(here
->e_value_offs
);
1633 void *val
= s
->base
+ offs
;
1635 memmove(first_val
+ old_size
, first_val
, val
- first_val
);
1636 memset(first_val
, 0, old_size
);
1637 min_offs
+= old_size
;
1639 /* Adjust all value offsets. */
1641 while (!IS_LAST_ENTRY(last
)) {
1642 size_t o
= le16_to_cpu(last
->e_value_offs
);
1644 if (!last
->e_value_inum
&&
1645 last
->e_value_size
&& o
< offs
)
1646 last
->e_value_offs
= cpu_to_le16(o
+ old_size
);
1647 last
= EXT4_XATTR_NEXT(last
);
1652 /* Remove old name. */
1653 size_t size
= EXT4_XATTR_LEN(name_len
);
1655 last
= ENTRY((void *)last
- size
);
1656 memmove(here
, (void *)here
+ size
,
1657 (void *)last
- (void *)here
+ sizeof(__u32
));
1658 memset(last
, 0, size
);
1659 } else if (s
->not_found
) {
1660 /* Insert new name. */
1661 size_t size
= EXT4_XATTR_LEN(name_len
);
1662 size_t rest
= (void *)last
- (void *)here
+ sizeof(__u32
);
1664 memmove((void *)here
+ size
, here
, rest
);
1665 memset(here
, 0, size
);
1666 here
->e_name_index
= i
->name_index
;
1667 here
->e_name_len
= name_len
;
1668 memcpy(here
->e_name
, i
->name
, name_len
);
1670 /* This is an update, reset value info. */
1671 here
->e_value_inum
= 0;
1672 here
->e_value_offs
= 0;
1673 here
->e_value_size
= 0;
1677 /* Insert new value. */
1679 here
->e_value_inum
= cpu_to_le32(new_ea_inode
->i_ino
);
1680 } else if (i
->value_len
) {
1681 void *val
= s
->base
+ min_offs
- new_size
;
1683 here
->e_value_offs
= cpu_to_le16(min_offs
- new_size
);
1684 if (i
->value
== EXT4_ZERO_XATTR_VALUE
) {
1685 memset(val
, 0, new_size
);
1687 memcpy(val
, i
->value
, i
->value_len
);
1688 /* Clear padding bytes. */
1689 memset(val
+ i
->value_len
, 0,
1690 new_size
- i
->value_len
);
1693 here
->e_value_size
= cpu_to_le32(i
->value_len
);
1699 /* Entry hash calculation. */
1704 * Feed crc32c hash instead of the raw value for entry
1705 * hash calculation. This is to avoid walking
1706 * potentially long value buffer again.
1708 crc32c_hash
= cpu_to_le32(
1709 ext4_xattr_inode_get_hash(new_ea_inode
));
1710 hash
= ext4_xattr_hash_entry(here
->e_name
,
1713 } else if (is_block
) {
1714 __le32
*value
= s
->base
+ min_offs
- new_size
;
1716 hash
= ext4_xattr_hash_entry(here
->e_name
,
1717 here
->e_name_len
, value
,
1720 here
->e_hash
= hash
;
1724 ext4_xattr_rehash((struct ext4_xattr_header
*)s
->base
);
1733 struct ext4_xattr_block_find
{
1734 struct ext4_xattr_search s
;
1735 struct buffer_head
*bh
;
1739 ext4_xattr_block_find(struct inode
*inode
, struct ext4_xattr_info
*i
,
1740 struct ext4_xattr_block_find
*bs
)
1742 struct super_block
*sb
= inode
->i_sb
;
1745 ea_idebug(inode
, "name=%d.%s, value=%p, value_len=%ld",
1746 i
->name_index
, i
->name
, i
->value
, (long)i
->value_len
);
1748 if (EXT4_I(inode
)->i_file_acl
) {
1749 /* The inode already has an extended attribute block. */
1750 bs
->bh
= sb_bread(sb
, EXT4_I(inode
)->i_file_acl
);
1754 ea_bdebug(bs
->bh
, "b_count=%d, refcount=%d",
1755 atomic_read(&(bs
->bh
->b_count
)),
1756 le32_to_cpu(BHDR(bs
->bh
)->h_refcount
));
1757 if (ext4_xattr_check_block(inode
, bs
->bh
)) {
1758 EXT4_ERROR_INODE(inode
, "bad block %llu",
1759 EXT4_I(inode
)->i_file_acl
);
1760 error
= -EFSCORRUPTED
;
1763 /* Find the named attribute. */
1764 bs
->s
.base
= BHDR(bs
->bh
);
1765 bs
->s
.first
= BFIRST(bs
->bh
);
1766 bs
->s
.end
= bs
->bh
->b_data
+ bs
->bh
->b_size
;
1767 bs
->s
.here
= bs
->s
.first
;
1768 error
= ext4_xattr_find_entry(&bs
->s
.here
, i
->name_index
,
1770 if (error
&& error
!= -ENODATA
)
1772 bs
->s
.not_found
= error
;
1781 ext4_xattr_block_set(handle_t
*handle
, struct inode
*inode
,
1782 struct ext4_xattr_info
*i
,
1783 struct ext4_xattr_block_find
*bs
)
1785 struct super_block
*sb
= inode
->i_sb
;
1786 struct buffer_head
*new_bh
= NULL
;
1787 struct ext4_xattr_search s_copy
= bs
->s
;
1788 struct ext4_xattr_search
*s
= &s_copy
;
1789 struct mb_cache_entry
*ce
= NULL
;
1791 struct mb_cache
*ea_block_cache
= EA_BLOCK_CACHE(inode
);
1792 struct inode
*ea_inode
= NULL
;
1793 size_t old_ea_inode_size
= 0;
1795 #define header(x) ((struct ext4_xattr_header *)(x))
1798 BUFFER_TRACE(bs
->bh
, "get_write_access");
1799 error
= ext4_journal_get_write_access(handle
, bs
->bh
);
1802 lock_buffer(bs
->bh
);
1804 if (header(s
->base
)->h_refcount
== cpu_to_le32(1)) {
1805 __u32 hash
= le32_to_cpu(BHDR(bs
->bh
)->h_hash
);
1808 * This must happen under buffer lock for
1809 * ext4_xattr_block_set() to reliably detect modified
1813 mb_cache_entry_delete(ea_block_cache
, hash
,
1815 ea_bdebug(bs
->bh
, "modifying in-place");
1816 error
= ext4_xattr_set_entry(i
, s
, handle
, inode
,
1817 true /* is_block */);
1819 ext4_xattr_block_cache_insert(ea_block_cache
,
1821 ext4_xattr_block_csum_set(inode
, bs
->bh
);
1822 unlock_buffer(bs
->bh
);
1823 if (error
== -EFSCORRUPTED
)
1826 error
= ext4_handle_dirty_metadata(handle
,
1833 int offset
= (char *)s
->here
- bs
->bh
->b_data
;
1835 unlock_buffer(bs
->bh
);
1836 ea_bdebug(bs
->bh
, "cloning");
1837 s
->base
= kmalloc(bs
->bh
->b_size
, GFP_NOFS
);
1839 if (s
->base
== NULL
)
1841 memcpy(s
->base
, BHDR(bs
->bh
), bs
->bh
->b_size
);
1842 s
->first
= ENTRY(header(s
->base
)+1);
1843 header(s
->base
)->h_refcount
= cpu_to_le32(1);
1844 s
->here
= ENTRY(s
->base
+ offset
);
1845 s
->end
= s
->base
+ bs
->bh
->b_size
;
1848 * If existing entry points to an xattr inode, we need
1849 * to prevent ext4_xattr_set_entry() from decrementing
1850 * ref count on it because the reference belongs to the
1851 * original block. In this case, make the entry look
1852 * like it has an empty value.
1854 if (!s
->not_found
&& s
->here
->e_value_inum
) {
1856 * Defer quota free call for previous inode
1857 * until success is guaranteed.
1859 old_ea_inode_size
= le32_to_cpu(
1860 s
->here
->e_value_size
);
1861 s
->here
->e_value_inum
= 0;
1862 s
->here
->e_value_size
= 0;
1866 /* Allocate a buffer where we construct the new block. */
1867 s
->base
= kzalloc(sb
->s_blocksize
, GFP_NOFS
);
1868 /* assert(header == s->base) */
1870 if (s
->base
== NULL
)
1872 header(s
->base
)->h_magic
= cpu_to_le32(EXT4_XATTR_MAGIC
);
1873 header(s
->base
)->h_blocks
= cpu_to_le32(1);
1874 header(s
->base
)->h_refcount
= cpu_to_le32(1);
1875 s
->first
= ENTRY(header(s
->base
)+1);
1876 s
->here
= ENTRY(header(s
->base
)+1);
1877 s
->end
= s
->base
+ sb
->s_blocksize
;
1880 error
= ext4_xattr_set_entry(i
, s
, handle
, inode
, true /* is_block */);
1881 if (error
== -EFSCORRUPTED
)
1886 if (i
->value
&& s
->here
->e_value_inum
) {
1887 unsigned int ea_ino
;
1890 * A ref count on ea_inode has been taken as part of the call to
1891 * ext4_xattr_set_entry() above. We would like to drop this
1892 * extra ref but we have to wait until the xattr block is
1893 * initialized and has its own ref count on the ea_inode.
1895 ea_ino
= le32_to_cpu(s
->here
->e_value_inum
);
1896 error
= ext4_xattr_inode_iget(inode
, ea_ino
, &ea_inode
);
1904 if (!IS_LAST_ENTRY(s
->first
)) {
1905 new_bh
= ext4_xattr_block_cache_find(inode
, header(s
->base
),
1908 /* We found an identical block in the cache. */
1909 if (new_bh
== bs
->bh
)
1910 ea_bdebug(new_bh
, "keeping");
1914 WARN_ON_ONCE(dquot_initialize_needed(inode
));
1916 /* The old block is released after updating
1918 error
= dquot_alloc_block(inode
,
1919 EXT4_C2B(EXT4_SB(sb
), 1));
1922 BUFFER_TRACE(new_bh
, "get_write_access");
1923 error
= ext4_journal_get_write_access(handle
,
1927 lock_buffer(new_bh
);
1929 * We have to be careful about races with
1930 * freeing, rehashing or adding references to
1931 * xattr block. Once we hold buffer lock xattr
1932 * block's state is stable so we can check
1933 * whether the block got freed / rehashed or
1934 * not. Since we unhash mbcache entry under
1935 * buffer lock when freeing / rehashing xattr
1936 * block, checking whether entry is still
1937 * hashed is reliable. Same rules hold for
1938 * e_reusable handling.
1940 if (hlist_bl_unhashed(&ce
->e_hash_list
) ||
1943 * Undo everything and check mbcache
1946 unlock_buffer(new_bh
);
1947 dquot_free_block(inode
,
1948 EXT4_C2B(EXT4_SB(sb
),
1951 mb_cache_entry_put(ea_block_cache
, ce
);
1956 ref
= le32_to_cpu(BHDR(new_bh
)->h_refcount
) + 1;
1957 BHDR(new_bh
)->h_refcount
= cpu_to_le32(ref
);
1958 if (ref
>= EXT4_XATTR_REFCOUNT_MAX
)
1960 ea_bdebug(new_bh
, "reusing; refcount now=%d",
1962 ext4_xattr_block_csum_set(inode
, new_bh
);
1963 unlock_buffer(new_bh
);
1964 error
= ext4_handle_dirty_metadata(handle
,
1970 mb_cache_entry_touch(ea_block_cache
, ce
);
1971 mb_cache_entry_put(ea_block_cache
, ce
);
1973 } else if (bs
->bh
&& s
->base
== bs
->bh
->b_data
) {
1974 /* We were modifying this block in-place. */
1975 ea_bdebug(bs
->bh
, "keeping this block");
1979 /* We need to allocate a new block */
1980 ext4_fsblk_t goal
, block
;
1982 WARN_ON_ONCE(dquot_initialize_needed(inode
));
1984 goal
= ext4_group_first_block_no(sb
,
1985 EXT4_I(inode
)->i_block_group
);
1987 /* non-extent files can't have physical blocks past 2^32 */
1988 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
1989 goal
= goal
& EXT4_MAX_BLOCK_FILE_PHYS
;
1991 block
= ext4_new_meta_blocks(handle
, inode
, goal
, 0,
1996 if (!(ext4_test_inode_flag(inode
, EXT4_INODE_EXTENTS
)))
1997 BUG_ON(block
> EXT4_MAX_BLOCK_FILE_PHYS
);
1999 ea_idebug(inode
, "creating block %llu",
2000 (unsigned long long)block
);
2002 new_bh
= sb_getblk(sb
, block
);
2003 if (unlikely(!new_bh
)) {
2006 ext4_free_blocks(handle
, inode
, NULL
, block
, 1,
2007 EXT4_FREE_BLOCKS_METADATA
);
2010 error
= ext4_xattr_inode_inc_ref_all(handle
, inode
,
2011 ENTRY(header(s
->base
)+1));
2015 /* Drop the extra ref on ea_inode. */
2016 error
= ext4_xattr_inode_dec_ref(handle
,
2019 ext4_warning_inode(ea_inode
,
2026 lock_buffer(new_bh
);
2027 error
= ext4_journal_get_create_access(handle
, new_bh
);
2029 unlock_buffer(new_bh
);
2033 memcpy(new_bh
->b_data
, s
->base
, new_bh
->b_size
);
2034 ext4_xattr_block_csum_set(inode
, new_bh
);
2035 set_buffer_uptodate(new_bh
);
2036 unlock_buffer(new_bh
);
2037 ext4_xattr_block_cache_insert(ea_block_cache
, new_bh
);
2038 error
= ext4_handle_dirty_metadata(handle
, inode
,
2045 if (old_ea_inode_size
)
2046 ext4_xattr_inode_free_quota(inode
, old_ea_inode_size
);
2048 /* Update the inode. */
2049 EXT4_I(inode
)->i_file_acl
= new_bh
? new_bh
->b_blocknr
: 0;
2051 /* Drop the previous xattr block. */
2052 if (bs
->bh
&& bs
->bh
!= new_bh
) {
2053 struct ext4_xattr_inode_array
*ea_inode_array
= NULL
;
2055 ext4_xattr_release_block(handle
, inode
, bs
->bh
,
2057 0 /* extra_credits */);
2058 ext4_xattr_inode_array_free(ea_inode_array
);
2066 error2
= ext4_xattr_inode_dec_ref(handle
, ea_inode
);
2068 ext4_warning_inode(ea_inode
, "dec ref error=%d",
2071 /* If there was an error, revert the quota charge. */
2073 ext4_xattr_inode_free_quota(inode
,
2074 i_size_read(ea_inode
));
2078 mb_cache_entry_put(ea_block_cache
, ce
);
2080 if (!(bs
->bh
&& s
->base
== bs
->bh
->b_data
))
2086 dquot_free_block(inode
, EXT4_C2B(EXT4_SB(sb
), 1));
2090 EXT4_ERROR_INODE(inode
, "bad block %llu",
2091 EXT4_I(inode
)->i_file_acl
);
2097 int ext4_xattr_ibody_find(struct inode
*inode
, struct ext4_xattr_info
*i
,
2098 struct ext4_xattr_ibody_find
*is
)
2100 struct ext4_xattr_ibody_header
*header
;
2101 struct ext4_inode
*raw_inode
;
2104 if (EXT4_I(inode
)->i_extra_isize
== 0)
2106 raw_inode
= ext4_raw_inode(&is
->iloc
);
2107 header
= IHDR(inode
, raw_inode
);
2108 is
->s
.base
= is
->s
.first
= IFIRST(header
);
2109 is
->s
.here
= is
->s
.first
;
2110 is
->s
.end
= (void *)raw_inode
+ EXT4_SB(inode
->i_sb
)->s_inode_size
;
2111 if (ext4_test_inode_state(inode
, EXT4_STATE_XATTR
)) {
2112 error
= xattr_check_inode(inode
, header
, is
->s
.end
);
2115 /* Find the named attribute. */
2116 error
= ext4_xattr_find_entry(&is
->s
.here
, i
->name_index
,
2118 if (error
&& error
!= -ENODATA
)
2120 is
->s
.not_found
= error
;
2125 int ext4_xattr_ibody_inline_set(handle_t
*handle
, struct inode
*inode
,
2126 struct ext4_xattr_info
*i
,
2127 struct ext4_xattr_ibody_find
*is
)
2129 struct ext4_xattr_ibody_header
*header
;
2130 struct ext4_xattr_search
*s
= &is
->s
;
2133 if (EXT4_I(inode
)->i_extra_isize
== 0)
2135 error
= ext4_xattr_set_entry(i
, s
, handle
, inode
, false /* is_block */);
2137 if (error
== -ENOSPC
&&
2138 ext4_has_inline_data(inode
)) {
2139 error
= ext4_try_to_evict_inline_data(handle
, inode
,
2140 EXT4_XATTR_LEN(strlen(i
->name
) +
2141 EXT4_XATTR_SIZE(i
->value_len
)));
2144 error
= ext4_xattr_ibody_find(inode
, i
, is
);
2147 error
= ext4_xattr_set_entry(i
, s
, handle
, inode
,
2148 false /* is_block */);
2153 header
= IHDR(inode
, ext4_raw_inode(&is
->iloc
));
2154 if (!IS_LAST_ENTRY(s
->first
)) {
2155 header
->h_magic
= cpu_to_le32(EXT4_XATTR_MAGIC
);
2156 ext4_set_inode_state(inode
, EXT4_STATE_XATTR
);
2158 header
->h_magic
= cpu_to_le32(0);
2159 ext4_clear_inode_state(inode
, EXT4_STATE_XATTR
);
2164 static int ext4_xattr_ibody_set(handle_t
*handle
, struct inode
*inode
,
2165 struct ext4_xattr_info
*i
,
2166 struct ext4_xattr_ibody_find
*is
)
2168 struct ext4_xattr_ibody_header
*header
;
2169 struct ext4_xattr_search
*s
= &is
->s
;
2172 if (EXT4_I(inode
)->i_extra_isize
== 0)
2174 error
= ext4_xattr_set_entry(i
, s
, handle
, inode
, false /* is_block */);
2177 header
= IHDR(inode
, ext4_raw_inode(&is
->iloc
));
2178 if (!IS_LAST_ENTRY(s
->first
)) {
2179 header
->h_magic
= cpu_to_le32(EXT4_XATTR_MAGIC
);
2180 ext4_set_inode_state(inode
, EXT4_STATE_XATTR
);
2182 header
->h_magic
= cpu_to_le32(0);
2183 ext4_clear_inode_state(inode
, EXT4_STATE_XATTR
);
2188 static int ext4_xattr_value_same(struct ext4_xattr_search
*s
,
2189 struct ext4_xattr_info
*i
)
2193 /* When e_value_inum is set the value is stored externally. */
2194 if (s
->here
->e_value_inum
)
2196 if (le32_to_cpu(s
->here
->e_value_size
) != i
->value_len
)
2198 value
= ((void *)s
->base
) + le16_to_cpu(s
->here
->e_value_offs
);
2199 return !memcmp(value
, i
->value
, i
->value_len
);
2202 static struct buffer_head
*ext4_xattr_get_block(struct inode
*inode
)
2204 struct buffer_head
*bh
;
2207 if (!EXT4_I(inode
)->i_file_acl
)
2209 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
2211 return ERR_PTR(-EIO
);
2212 error
= ext4_xattr_check_block(inode
, bh
);
2214 return ERR_PTR(error
);
2219 * ext4_xattr_set_handle()
2221 * Create, replace or remove an extended attribute for this inode. Value
2222 * is NULL to remove an existing extended attribute, and non-NULL to
2223 * either replace an existing extended attribute, or create a new extended
2224 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
2225 * specify that an extended attribute must exist and must not exist
2226 * previous to the call, respectively.
2228 * Returns 0, or a negative error number on failure.
2231 ext4_xattr_set_handle(handle_t
*handle
, struct inode
*inode
, int name_index
,
2232 const char *name
, const void *value
, size_t value_len
,
2235 struct ext4_xattr_info i
= {
2236 .name_index
= name_index
,
2239 .value_len
= value_len
,
2242 struct ext4_xattr_ibody_find is
= {
2243 .s
= { .not_found
= -ENODATA
, },
2245 struct ext4_xattr_block_find bs
= {
2246 .s
= { .not_found
= -ENODATA
, },
2253 if (strlen(name
) > 255)
2256 ext4_write_lock_xattr(inode
, &no_expand
);
2258 /* Check journal credits under write lock. */
2259 if (ext4_handle_valid(handle
)) {
2260 struct buffer_head
*bh
;
2263 bh
= ext4_xattr_get_block(inode
);
2265 error
= PTR_ERR(bh
);
2269 credits
= __ext4_xattr_set_credits(inode
->i_sb
, inode
, bh
,
2271 flags
& XATTR_CREATE
);
2274 if (!ext4_handle_has_enough_credits(handle
, credits
)) {
2280 error
= ext4_reserve_inode_write(handle
, inode
, &is
.iloc
);
2284 if (ext4_test_inode_state(inode
, EXT4_STATE_NEW
)) {
2285 struct ext4_inode
*raw_inode
= ext4_raw_inode(&is
.iloc
);
2286 memset(raw_inode
, 0, EXT4_SB(inode
->i_sb
)->s_inode_size
);
2287 ext4_clear_inode_state(inode
, EXT4_STATE_NEW
);
2290 error
= ext4_xattr_ibody_find(inode
, &i
, &is
);
2294 error
= ext4_xattr_block_find(inode
, &i
, &bs
);
2297 if (is
.s
.not_found
&& bs
.s
.not_found
) {
2299 if (flags
& XATTR_REPLACE
)
2306 if (flags
& XATTR_CREATE
)
2311 if (!is
.s
.not_found
)
2312 error
= ext4_xattr_ibody_set(handle
, inode
, &i
, &is
);
2313 else if (!bs
.s
.not_found
)
2314 error
= ext4_xattr_block_set(handle
, inode
, &i
, &bs
);
2317 /* Xattr value did not change? Save us some work and bail out */
2318 if (!is
.s
.not_found
&& ext4_xattr_value_same(&is
.s
, &i
))
2320 if (!bs
.s
.not_found
&& ext4_xattr_value_same(&bs
.s
, &i
))
2323 if (ext4_has_feature_ea_inode(inode
->i_sb
) &&
2324 (EXT4_XATTR_SIZE(i
.value_len
) >
2325 EXT4_XATTR_MIN_LARGE_EA_SIZE(inode
->i_sb
->s_blocksize
)))
2328 error
= ext4_xattr_ibody_set(handle
, inode
, &i
, &is
);
2329 if (!error
&& !bs
.s
.not_found
) {
2331 error
= ext4_xattr_block_set(handle
, inode
, &i
, &bs
);
2332 } else if (error
== -ENOSPC
) {
2333 if (EXT4_I(inode
)->i_file_acl
&& !bs
.s
.base
) {
2334 error
= ext4_xattr_block_find(inode
, &i
, &bs
);
2338 error
= ext4_xattr_block_set(handle
, inode
, &i
, &bs
);
2339 if (!error
&& !is
.s
.not_found
) {
2341 error
= ext4_xattr_ibody_set(handle
, inode
, &i
,
2343 } else if (error
== -ENOSPC
) {
2345 * Xattr does not fit in the block, store at
2346 * external inode if possible.
2348 if (ext4_has_feature_ea_inode(inode
->i_sb
) &&
2357 ext4_xattr_update_super_block(handle
, inode
->i_sb
);
2358 inode
->i_ctime
= current_time(inode
);
2361 error
= ext4_mark_iloc_dirty(handle
, inode
, &is
.iloc
);
2363 * The bh is consumed by ext4_mark_iloc_dirty, even with
2368 ext4_handle_sync(handle
);
2374 ext4_write_unlock_xattr(inode
, &no_expand
);
2378 int ext4_xattr_set_credits(struct inode
*inode
, size_t value_len
,
2379 bool is_create
, int *credits
)
2381 struct buffer_head
*bh
;
2386 if (!EXT4_SB(inode
->i_sb
)->s_journal
)
2389 down_read(&EXT4_I(inode
)->xattr_sem
);
2391 bh
= ext4_xattr_get_block(inode
);
2395 *credits
= __ext4_xattr_set_credits(inode
->i_sb
, inode
, bh
,
2396 value_len
, is_create
);
2401 up_read(&EXT4_I(inode
)->xattr_sem
);
2408 * Like ext4_xattr_set_handle, but start from an inode. This extended
2409 * attribute modification is a filesystem transaction by itself.
2411 * Returns 0, or a negative error number on failure.
2414 ext4_xattr_set(struct inode
*inode
, int name_index
, const char *name
,
2415 const void *value
, size_t value_len
, int flags
)
2418 struct super_block
*sb
= inode
->i_sb
;
2419 int error
, retries
= 0;
2422 error
= dquot_initialize(inode
);
2427 error
= ext4_xattr_set_credits(inode
, value_len
, flags
& XATTR_CREATE
,
2432 handle
= ext4_journal_start(inode
, EXT4_HT_XATTR
, credits
);
2433 if (IS_ERR(handle
)) {
2434 error
= PTR_ERR(handle
);
2438 error
= ext4_xattr_set_handle(handle
, inode
, name_index
, name
,
2439 value
, value_len
, flags
);
2440 error2
= ext4_journal_stop(handle
);
2441 if (error
== -ENOSPC
&&
2442 ext4_should_retry_alloc(sb
, &retries
))
2452 * Shift the EA entries in the inode to create space for the increased
2455 static void ext4_xattr_shift_entries(struct ext4_xattr_entry
*entry
,
2456 int value_offs_shift
, void *to
,
2457 void *from
, size_t n
)
2459 struct ext4_xattr_entry
*last
= entry
;
2462 /* We always shift xattr headers further thus offsets get lower */
2463 BUG_ON(value_offs_shift
> 0);
2465 /* Adjust the value offsets of the entries */
2466 for (; !IS_LAST_ENTRY(last
); last
= EXT4_XATTR_NEXT(last
)) {
2467 if (!last
->e_value_inum
&& last
->e_value_size
) {
2468 new_offs
= le16_to_cpu(last
->e_value_offs
) +
2470 last
->e_value_offs
= cpu_to_le16(new_offs
);
2473 /* Shift the entries by n bytes */
2474 memmove(to
, from
, n
);
2478 * Move xattr pointed to by 'entry' from inode into external xattr block
2480 static int ext4_xattr_move_to_block(handle_t
*handle
, struct inode
*inode
,
2481 struct ext4_inode
*raw_inode
,
2482 struct ext4_xattr_entry
*entry
)
2484 struct ext4_xattr_ibody_find
*is
= NULL
;
2485 struct ext4_xattr_block_find
*bs
= NULL
;
2486 char *buffer
= NULL
, *b_entry_name
= NULL
;
2487 size_t value_size
= le32_to_cpu(entry
->e_value_size
);
2488 struct ext4_xattr_info i
= {
2491 .name_index
= entry
->e_name_index
,
2492 .in_inode
= !!entry
->e_value_inum
,
2494 struct ext4_xattr_ibody_header
*header
= IHDR(inode
, raw_inode
);
2497 is
= kzalloc(sizeof(struct ext4_xattr_ibody_find
), GFP_NOFS
);
2498 bs
= kzalloc(sizeof(struct ext4_xattr_block_find
), GFP_NOFS
);
2499 buffer
= kmalloc(value_size
, GFP_NOFS
);
2500 b_entry_name
= kmalloc(entry
->e_name_len
+ 1, GFP_NOFS
);
2501 if (!is
|| !bs
|| !buffer
|| !b_entry_name
) {
2506 is
->s
.not_found
= -ENODATA
;
2507 bs
->s
.not_found
= -ENODATA
;
2511 /* Save the entry name and the entry value */
2512 if (entry
->e_value_inum
) {
2513 error
= ext4_xattr_inode_get(inode
, entry
, buffer
, value_size
);
2517 size_t value_offs
= le16_to_cpu(entry
->e_value_offs
);
2518 memcpy(buffer
, (void *)IFIRST(header
) + value_offs
, value_size
);
2521 memcpy(b_entry_name
, entry
->e_name
, entry
->e_name_len
);
2522 b_entry_name
[entry
->e_name_len
] = '\0';
2523 i
.name
= b_entry_name
;
2525 error
= ext4_get_inode_loc(inode
, &is
->iloc
);
2529 error
= ext4_xattr_ibody_find(inode
, &i
, is
);
2533 /* Remove the chosen entry from the inode */
2534 error
= ext4_xattr_ibody_set(handle
, inode
, &i
, is
);
2539 i
.value_len
= value_size
;
2540 error
= ext4_xattr_block_find(inode
, &i
, bs
);
2544 /* Add entry which was removed from the inode into the block */
2545 error
= ext4_xattr_block_set(handle
, inode
, &i
, bs
);
2550 kfree(b_entry_name
);
2553 brelse(is
->iloc
.bh
);
2560 static int ext4_xattr_make_inode_space(handle_t
*handle
, struct inode
*inode
,
2561 struct ext4_inode
*raw_inode
,
2562 int isize_diff
, size_t ifree
,
2563 size_t bfree
, int *total_ino
)
2565 struct ext4_xattr_ibody_header
*header
= IHDR(inode
, raw_inode
);
2566 struct ext4_xattr_entry
*small_entry
;
2567 struct ext4_xattr_entry
*entry
;
2568 struct ext4_xattr_entry
*last
;
2569 unsigned int entry_size
; /* EA entry size */
2570 unsigned int total_size
; /* EA entry size + value size */
2571 unsigned int min_total_size
;
2574 while (isize_diff
> ifree
) {
2577 min_total_size
= ~0U;
2578 last
= IFIRST(header
);
2579 /* Find the entry best suited to be pushed into EA block */
2580 for (; !IS_LAST_ENTRY(last
); last
= EXT4_XATTR_NEXT(last
)) {
2581 total_size
= EXT4_XATTR_LEN(last
->e_name_len
);
2582 if (!last
->e_value_inum
)
2583 total_size
+= EXT4_XATTR_SIZE(
2584 le32_to_cpu(last
->e_value_size
));
2585 if (total_size
<= bfree
&&
2586 total_size
< min_total_size
) {
2587 if (total_size
+ ifree
< isize_diff
) {
2591 min_total_size
= total_size
;
2596 if (entry
== NULL
) {
2597 if (small_entry
== NULL
)
2599 entry
= small_entry
;
2602 entry_size
= EXT4_XATTR_LEN(entry
->e_name_len
);
2603 total_size
= entry_size
;
2604 if (!entry
->e_value_inum
)
2605 total_size
+= EXT4_XATTR_SIZE(
2606 le32_to_cpu(entry
->e_value_size
));
2607 error
= ext4_xattr_move_to_block(handle
, inode
, raw_inode
,
2612 *total_ino
-= entry_size
;
2613 ifree
+= total_size
;
2614 bfree
-= total_size
;
2621 * Expand an inode by new_extra_isize bytes when EAs are present.
2622 * Returns 0 on success or negative error number on failure.
2624 int ext4_expand_extra_isize_ea(struct inode
*inode
, int new_extra_isize
,
2625 struct ext4_inode
*raw_inode
, handle_t
*handle
)
2627 struct ext4_xattr_ibody_header
*header
;
2628 struct buffer_head
*bh
= NULL
;
2630 size_t ifree
, bfree
;
2633 int error
= 0, tried_min_extra_isize
= 0;
2634 int s_min_extra_isize
= le16_to_cpu(EXT4_SB(inode
->i_sb
)->s_es
->s_min_extra_isize
);
2635 int isize_diff
; /* How much do we need to grow i_extra_isize */
2638 if (ext4_write_trylock_xattr(inode
, &no_expand
) == 0)
2642 isize_diff
= new_extra_isize
- EXT4_I(inode
)->i_extra_isize
;
2643 if (EXT4_I(inode
)->i_extra_isize
>= new_extra_isize
)
2646 header
= IHDR(inode
, raw_inode
);
2649 * Check if enough free space is available in the inode to shift the
2650 * entries ahead by new_extra_isize.
2653 base
= IFIRST(header
);
2654 end
= (void *)raw_inode
+ EXT4_SB(inode
->i_sb
)->s_inode_size
;
2655 min_offs
= end
- base
;
2656 total_ino
= sizeof(struct ext4_xattr_ibody_header
);
2658 error
= xattr_check_inode(inode
, header
, end
);
2662 ifree
= ext4_xattr_free_space(base
, &min_offs
, base
, &total_ino
);
2663 if (ifree
>= isize_diff
)
2667 * Enough free space isn't available in the inode, check if
2668 * EA block can hold new_extra_isize bytes.
2670 if (EXT4_I(inode
)->i_file_acl
) {
2671 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
2675 if (ext4_xattr_check_block(inode
, bh
)) {
2676 EXT4_ERROR_INODE(inode
, "bad block %llu",
2677 EXT4_I(inode
)->i_file_acl
);
2678 error
= -EFSCORRUPTED
;
2682 end
= bh
->b_data
+ bh
->b_size
;
2683 min_offs
= end
- base
;
2684 bfree
= ext4_xattr_free_space(BFIRST(bh
), &min_offs
, base
,
2686 if (bfree
+ ifree
< isize_diff
) {
2687 if (!tried_min_extra_isize
&& s_min_extra_isize
) {
2688 tried_min_extra_isize
++;
2689 new_extra_isize
= s_min_extra_isize
;
2697 bfree
= inode
->i_sb
->s_blocksize
;
2700 error
= ext4_xattr_make_inode_space(handle
, inode
, raw_inode
,
2701 isize_diff
, ifree
, bfree
,
2704 if (error
== -ENOSPC
&& !tried_min_extra_isize
&&
2705 s_min_extra_isize
) {
2706 tried_min_extra_isize
++;
2707 new_extra_isize
= s_min_extra_isize
;
2714 /* Adjust the offsets and shift the remaining entries ahead */
2715 ext4_xattr_shift_entries(IFIRST(header
), EXT4_I(inode
)->i_extra_isize
2716 - new_extra_isize
, (void *)raw_inode
+
2717 EXT4_GOOD_OLD_INODE_SIZE
+ new_extra_isize
,
2718 (void *)header
, total_ino
);
2719 EXT4_I(inode
)->i_extra_isize
= new_extra_isize
;
2722 ext4_write_unlock_xattr(inode
, &no_expand
);
2728 * Inode size expansion failed; don't try again
2731 ext4_write_unlock_xattr(inode
, &no_expand
);
2735 #define EIA_INCR 16 /* must be 2^n */
2736 #define EIA_MASK (EIA_INCR - 1)
2738 /* Add the large xattr @inode into @ea_inode_array for deferred iput().
2739 * If @ea_inode_array is new or full it will be grown and the old
2740 * contents copied over.
2743 ext4_expand_inode_array(struct ext4_xattr_inode_array
**ea_inode_array
,
2744 struct inode
*inode
)
2746 if (*ea_inode_array
== NULL
) {
2748 * Start with 15 inodes, so it fits into a power-of-two size.
2749 * If *ea_inode_array is NULL, this is essentially offsetof()
2752 kmalloc(offsetof(struct ext4_xattr_inode_array
,
2755 if (*ea_inode_array
== NULL
)
2757 (*ea_inode_array
)->count
= 0;
2758 } else if (((*ea_inode_array
)->count
& EIA_MASK
) == EIA_MASK
) {
2759 /* expand the array once all 15 + n * 16 slots are full */
2760 struct ext4_xattr_inode_array
*new_array
= NULL
;
2761 int count
= (*ea_inode_array
)->count
;
2763 /* if new_array is NULL, this is essentially offsetof() */
2764 new_array
= kmalloc(
2765 offsetof(struct ext4_xattr_inode_array
,
2766 inodes
[count
+ EIA_INCR
]),
2768 if (new_array
== NULL
)
2770 memcpy(new_array
, *ea_inode_array
,
2771 offsetof(struct ext4_xattr_inode_array
, inodes
[count
]));
2772 kfree(*ea_inode_array
);
2773 *ea_inode_array
= new_array
;
2775 (*ea_inode_array
)->inodes
[(*ea_inode_array
)->count
++] = inode
;
2780 * ext4_xattr_delete_inode()
2782 * Free extended attribute resources associated with this inode. Traverse
2783 * all entries and decrement reference on any xattr inodes associated with this
2784 * inode. This is called immediately before an inode is freed. We have exclusive
2785 * access to the inode. If an orphan inode is deleted it will also release its
2786 * references on xattr block and xattr inodes.
2788 int ext4_xattr_delete_inode(handle_t
*handle
, struct inode
*inode
,
2789 struct ext4_xattr_inode_array
**ea_inode_array
,
2792 struct buffer_head
*bh
= NULL
;
2793 struct ext4_xattr_ibody_header
*header
;
2794 struct ext4_iloc iloc
= { .bh
= NULL
};
2795 struct ext4_xattr_entry
*entry
;
2798 error
= ext4_xattr_ensure_credits(handle
, inode
, extra_credits
,
2801 false /* block_csum */);
2803 EXT4_ERROR_INODE(inode
, "ensure credits (error %d)", error
);
2807 if (ext4_has_feature_ea_inode(inode
->i_sb
) &&
2808 ext4_test_inode_state(inode
, EXT4_STATE_XATTR
)) {
2810 error
= ext4_get_inode_loc(inode
, &iloc
);
2812 EXT4_ERROR_INODE(inode
, "inode loc (error %d)", error
);
2816 error
= ext4_journal_get_write_access(handle
, iloc
.bh
);
2818 EXT4_ERROR_INODE(inode
, "write access (error %d)",
2823 header
= IHDR(inode
, ext4_raw_inode(&iloc
));
2824 if (header
->h_magic
== cpu_to_le32(EXT4_XATTR_MAGIC
))
2825 ext4_xattr_inode_dec_ref_all(handle
, inode
, iloc
.bh
,
2827 false /* block_csum */,
2830 false /* skip_quota */);
2833 if (EXT4_I(inode
)->i_file_acl
) {
2834 bh
= sb_bread(inode
->i_sb
, EXT4_I(inode
)->i_file_acl
);
2836 EXT4_ERROR_INODE(inode
, "block %llu read error",
2837 EXT4_I(inode
)->i_file_acl
);
2841 error
= ext4_xattr_check_block(inode
, bh
);
2843 EXT4_ERROR_INODE(inode
, "bad block %llu (error %d)",
2844 EXT4_I(inode
)->i_file_acl
, error
);
2848 if (ext4_has_feature_ea_inode(inode
->i_sb
)) {
2849 for (entry
= BFIRST(bh
); !IS_LAST_ENTRY(entry
);
2850 entry
= EXT4_XATTR_NEXT(entry
))
2851 if (entry
->e_value_inum
)
2852 ext4_xattr_inode_free_quota(inode
,
2853 le32_to_cpu(entry
->e_value_size
));
2857 ext4_xattr_release_block(handle
, inode
, bh
, ea_inode_array
,
2860 * Update i_file_acl value in the same transaction that releases
2863 EXT4_I(inode
)->i_file_acl
= 0;
2864 error
= ext4_mark_inode_dirty(handle
, inode
);
2866 EXT4_ERROR_INODE(inode
, "mark inode dirty (error %d)",
2878 void ext4_xattr_inode_array_free(struct ext4_xattr_inode_array
*ea_inode_array
)
2882 if (ea_inode_array
== NULL
)
2885 for (idx
= 0; idx
< ea_inode_array
->count
; ++idx
)
2886 iput(ea_inode_array
->inodes
[idx
]);
2887 kfree(ea_inode_array
);
2891 * ext4_xattr_block_cache_insert()
2893 * Create a new entry in the extended attribute block cache, and insert
2894 * it unless such an entry is already in the cache.
2896 * Returns 0, or a negative error number on failure.
2899 ext4_xattr_block_cache_insert(struct mb_cache
*ea_block_cache
,
2900 struct buffer_head
*bh
)
2902 struct ext4_xattr_header
*header
= BHDR(bh
);
2903 __u32 hash
= le32_to_cpu(header
->h_hash
);
2904 int reusable
= le32_to_cpu(header
->h_refcount
) <
2905 EXT4_XATTR_REFCOUNT_MAX
;
2908 if (!ea_block_cache
)
2910 error
= mb_cache_entry_create(ea_block_cache
, GFP_NOFS
, hash
,
2911 bh
->b_blocknr
, reusable
);
2913 if (error
== -EBUSY
)
2914 ea_bdebug(bh
, "already in cache");
2916 ea_bdebug(bh
, "inserting [%x]", (int)hash
);
2922 * Compare two extended attribute blocks for equality.
2924 * Returns 0 if the blocks are equal, 1 if they differ, and
2925 * a negative error number on errors.
2928 ext4_xattr_cmp(struct ext4_xattr_header
*header1
,
2929 struct ext4_xattr_header
*header2
)
2931 struct ext4_xattr_entry
*entry1
, *entry2
;
2933 entry1
= ENTRY(header1
+1);
2934 entry2
= ENTRY(header2
+1);
2935 while (!IS_LAST_ENTRY(entry1
)) {
2936 if (IS_LAST_ENTRY(entry2
))
2938 if (entry1
->e_hash
!= entry2
->e_hash
||
2939 entry1
->e_name_index
!= entry2
->e_name_index
||
2940 entry1
->e_name_len
!= entry2
->e_name_len
||
2941 entry1
->e_value_size
!= entry2
->e_value_size
||
2942 entry1
->e_value_inum
!= entry2
->e_value_inum
||
2943 memcmp(entry1
->e_name
, entry2
->e_name
, entry1
->e_name_len
))
2945 if (!entry1
->e_value_inum
&&
2946 memcmp((char *)header1
+ le16_to_cpu(entry1
->e_value_offs
),
2947 (char *)header2
+ le16_to_cpu(entry2
->e_value_offs
),
2948 le32_to_cpu(entry1
->e_value_size
)))
2951 entry1
= EXT4_XATTR_NEXT(entry1
);
2952 entry2
= EXT4_XATTR_NEXT(entry2
);
2954 if (!IS_LAST_ENTRY(entry2
))
2960 * ext4_xattr_block_cache_find()
2962 * Find an identical extended attribute block.
2964 * Returns a pointer to the block found, or NULL if such a block was
2965 * not found or an error occurred.
2967 static struct buffer_head
*
2968 ext4_xattr_block_cache_find(struct inode
*inode
,
2969 struct ext4_xattr_header
*header
,
2970 struct mb_cache_entry
**pce
)
2972 __u32 hash
= le32_to_cpu(header
->h_hash
);
2973 struct mb_cache_entry
*ce
;
2974 struct mb_cache
*ea_block_cache
= EA_BLOCK_CACHE(inode
);
2976 if (!ea_block_cache
)
2978 if (!header
->h_hash
)
2979 return NULL
; /* never share */
2980 ea_idebug(inode
, "looking for cached blocks [%x]", (int)hash
);
2981 ce
= mb_cache_entry_find_first(ea_block_cache
, hash
);
2983 struct buffer_head
*bh
;
2985 bh
= sb_bread(inode
->i_sb
, ce
->e_value
);
2987 EXT4_ERROR_INODE(inode
, "block %lu read error",
2988 (unsigned long)ce
->e_value
);
2989 } else if (ext4_xattr_cmp(header
, BHDR(bh
)) == 0) {
2994 ce
= mb_cache_entry_find_next(ea_block_cache
, ce
);
2999 #define NAME_HASH_SHIFT 5
3000 #define VALUE_HASH_SHIFT 16
3003 * ext4_xattr_hash_entry()
3005 * Compute the hash of an extended attribute.
3007 static __le32
ext4_xattr_hash_entry(char *name
, size_t name_len
, __le32
*value
,
3012 while (name_len
--) {
3013 hash
= (hash
<< NAME_HASH_SHIFT
) ^
3014 (hash
>> (8*sizeof(hash
) - NAME_HASH_SHIFT
)) ^
3017 while (value_count
--) {
3018 hash
= (hash
<< VALUE_HASH_SHIFT
) ^
3019 (hash
>> (8*sizeof(hash
) - VALUE_HASH_SHIFT
)) ^
3020 le32_to_cpu(*value
++);
3022 return cpu_to_le32(hash
);
3025 #undef NAME_HASH_SHIFT
3026 #undef VALUE_HASH_SHIFT
3028 #define BLOCK_HASH_SHIFT 16
3031 * ext4_xattr_rehash()
3033 * Re-compute the extended attribute hash value after an entry has changed.
3035 static void ext4_xattr_rehash(struct ext4_xattr_header
*header
)
3037 struct ext4_xattr_entry
*here
;
3040 here
= ENTRY(header
+1);
3041 while (!IS_LAST_ENTRY(here
)) {
3042 if (!here
->e_hash
) {
3043 /* Block is not shared if an entry's hash value == 0 */
3047 hash
= (hash
<< BLOCK_HASH_SHIFT
) ^
3048 (hash
>> (8*sizeof(hash
) - BLOCK_HASH_SHIFT
)) ^
3049 le32_to_cpu(here
->e_hash
);
3050 here
= EXT4_XATTR_NEXT(here
);
3052 header
->h_hash
= cpu_to_le32(hash
);
3055 #undef BLOCK_HASH_SHIFT
3057 #define HASH_BUCKET_BITS 10
3060 ext4_xattr_create_cache(void)
3062 return mb_cache_create(HASH_BUCKET_BITS
);
3065 void ext4_xattr_destroy_cache(struct mb_cache
*cache
)
3068 mb_cache_destroy(cache
);