]> git.proxmox.com Git - mirror_ubuntu-zesty-kernel.git/blob - fs/ext4/xattr.c
Add support for the AudioInjector.net Octo sound card
[mirror_ubuntu-zesty-kernel.git] / fs / ext4 / xattr.c
1 /*
2 * linux/fs/ext4/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext4 code with a lot of help from Eric Jarman <ejarman@acm.org>.
8 * Extended attributes for symlinks and special files added per
9 * suggestion of Luka Renko <luka.renko@hermes.si>.
10 * xattr consolidation Copyright (c) 2004 James Morris <jmorris@redhat.com>,
11 * Red Hat Inc.
12 * ea-in-inode support by Alex Tomas <alex@clusterfs.com> aka bzzz
13 * and Andreas Gruenbacher <agruen@suse.de>.
14 */
15
16 /*
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.
24 *
25 * The attributes in inodes and on blocks have a different header; the entries
26 * are stored in the same format:
27 *
28 * +------------------+
29 * | header |
30 * | entry 1 | |
31 * | entry 2 | | growing downwards
32 * | entry 3 | v
33 * | four null bytes |
34 * | . . . |
35 * | value 1 | ^
36 * | value 3 | | growing upwards
37 * | value 2 | |
38 * +------------------+
39 *
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.
43 *
44 * Locking strategy
45 * ----------------
46 * EXT4_I(inode)->i_file_acl is protected by EXT4_I(inode)->xattr_sem.
47 * EA blocks are only changed if they are exclusive to an inode, so
48 * holding xattr_sem also means that nothing but the EA block's reference
49 * count can change. Multiple writers to the same block are synchronized
50 * by the buffer lock.
51 */
52
53 #include <linux/init.h>
54 #include <linux/fs.h>
55 #include <linux/slab.h>
56 #include <linux/mbcache.h>
57 #include <linux/quotaops.h>
58 #include "ext4_jbd2.h"
59 #include "ext4.h"
60 #include "xattr.h"
61 #include "acl.h"
62
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__)
70 #else
71 # define ea_idebug(inode, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
72 # define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
73 #endif
74
75 static void ext4_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
76 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
77 struct ext4_xattr_header *,
78 struct mb_cache_entry **);
79 static void ext4_xattr_rehash(struct ext4_xattr_header *,
80 struct ext4_xattr_entry *);
81 static int ext4_xattr_list(struct dentry *dentry, char *buffer,
82 size_t buffer_size);
83
84 static const struct xattr_handler *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,
89 #endif
90 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
91 #ifdef CONFIG_EXT4_FS_SECURITY
92 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
93 #endif
94 };
95
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,
102 #endif
103 #ifdef CONFIG_EXT4_FS_SECURITY
104 &ext4_xattr_security_handler,
105 #endif
106 NULL
107 };
108
109 #define EXT4_GET_MB_CACHE(inode) (((struct ext4_sb_info *) \
110 inode->i_sb->s_fs_info)->s_mb_cache)
111
112 static __le32 ext4_xattr_block_csum(struct inode *inode,
113 sector_t block_nr,
114 struct ext4_xattr_header *hdr)
115 {
116 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
117 __u32 csum;
118 __le64 dsk_block_nr = cpu_to_le64(block_nr);
119 __u32 dummy_csum = 0;
120 int offset = offsetof(struct ext4_xattr_header, h_checksum);
121
122 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
123 sizeof(dsk_block_nr));
124 csum = ext4_chksum(sbi, csum, (__u8 *)hdr, offset);
125 csum = ext4_chksum(sbi, csum, (__u8 *)&dummy_csum, sizeof(dummy_csum));
126 offset += sizeof(dummy_csum);
127 csum = ext4_chksum(sbi, csum, (__u8 *)hdr + offset,
128 EXT4_BLOCK_SIZE(inode->i_sb) - offset);
129
130 return cpu_to_le32(csum);
131 }
132
133 static int ext4_xattr_block_csum_verify(struct inode *inode,
134 struct buffer_head *bh)
135 {
136 struct ext4_xattr_header *hdr = BHDR(bh);
137 int ret = 1;
138
139 if (ext4_has_metadata_csum(inode->i_sb)) {
140 lock_buffer(bh);
141 ret = (hdr->h_checksum == ext4_xattr_block_csum(inode,
142 bh->b_blocknr, hdr));
143 unlock_buffer(bh);
144 }
145 return ret;
146 }
147
148 static void ext4_xattr_block_csum_set(struct inode *inode,
149 struct buffer_head *bh)
150 {
151 if (ext4_has_metadata_csum(inode->i_sb))
152 BHDR(bh)->h_checksum = ext4_xattr_block_csum(inode,
153 bh->b_blocknr, BHDR(bh));
154 }
155
156 static inline const struct xattr_handler *
157 ext4_xattr_handler(int name_index)
158 {
159 const struct xattr_handler *handler = NULL;
160
161 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
162 handler = ext4_xattr_handler_map[name_index];
163 return handler;
164 }
165
166 /*
167 * Inode operation listxattr()
168 *
169 * d_inode(dentry)->i_mutex: don't care
170 */
171 ssize_t
172 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
173 {
174 return ext4_xattr_list(dentry, buffer, size);
175 }
176
177 static int
178 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end,
179 void *value_start)
180 {
181 struct ext4_xattr_entry *e = entry;
182
183 /* Find the end of the names list */
184 while (!IS_LAST_ENTRY(e)) {
185 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(e);
186 if ((void *)next >= end)
187 return -EFSCORRUPTED;
188 e = next;
189 }
190
191 /* Check the values */
192 while (!IS_LAST_ENTRY(entry)) {
193 if (entry->e_value_block != 0)
194 return -EFSCORRUPTED;
195 if (entry->e_value_size != 0) {
196 u16 offs = le16_to_cpu(entry->e_value_offs);
197 u32 size = le32_to_cpu(entry->e_value_size);
198 void *value;
199
200 /*
201 * The value cannot overlap the names, and the value
202 * with padding cannot extend beyond 'end'. Check both
203 * the padded and unpadded sizes, since the size may
204 * overflow to 0 when adding padding.
205 */
206 if (offs > end - value_start)
207 return -EFSCORRUPTED;
208 value = value_start + offs;
209 if (value < (void *)e + sizeof(u32) ||
210 size > end - value ||
211 EXT4_XATTR_SIZE(size) > end - value)
212 return -EFSCORRUPTED;
213 }
214 entry = EXT4_XATTR_NEXT(entry);
215 }
216
217 return 0;
218 }
219
220 static inline int
221 ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
222 {
223 int error;
224
225 if (buffer_verified(bh))
226 return 0;
227
228 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
229 BHDR(bh)->h_blocks != cpu_to_le32(1))
230 return -EFSCORRUPTED;
231 if (!ext4_xattr_block_csum_verify(inode, bh))
232 return -EFSBADCRC;
233 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size,
234 bh->b_data);
235 if (!error)
236 set_buffer_verified(bh);
237 return error;
238 }
239
240 static int
241 __xattr_check_inode(struct inode *inode, struct ext4_xattr_ibody_header *header,
242 void *end, const char *function, unsigned int line)
243 {
244 int error = -EFSCORRUPTED;
245
246 if (end - (void *)header < sizeof(*header) + sizeof(u32) ||
247 (header->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC)))
248 goto errout;
249 error = ext4_xattr_check_names(IFIRST(header), end, IFIRST(header));
250 errout:
251 if (error)
252 __ext4_error_inode(inode, function, line, 0,
253 "corrupted in-inode xattr");
254 return error;
255 }
256
257 #define xattr_check_inode(inode, header, end) \
258 __xattr_check_inode((inode), (header), (end), __func__, __LINE__)
259
260 static inline int
261 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
262 {
263 size_t value_size = le32_to_cpu(entry->e_value_size);
264
265 if (entry->e_value_block != 0 || value_size > size ||
266 le16_to_cpu(entry->e_value_offs) + value_size > size)
267 return -EFSCORRUPTED;
268 return 0;
269 }
270
271 static int
272 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
273 const char *name, size_t size, int sorted)
274 {
275 struct ext4_xattr_entry *entry;
276 size_t name_len;
277 int cmp = 1;
278
279 if (name == NULL)
280 return -EINVAL;
281 name_len = strlen(name);
282 entry = *pentry;
283 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
284 cmp = name_index - entry->e_name_index;
285 if (!cmp)
286 cmp = name_len - entry->e_name_len;
287 if (!cmp)
288 cmp = memcmp(name, entry->e_name, name_len);
289 if (cmp <= 0 && (sorted || cmp == 0))
290 break;
291 }
292 *pentry = entry;
293 if (!cmp && ext4_xattr_check_entry(entry, size))
294 return -EFSCORRUPTED;
295 return cmp ? -ENODATA : 0;
296 }
297
298 static int
299 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
300 void *buffer, size_t buffer_size)
301 {
302 struct buffer_head *bh = NULL;
303 struct ext4_xattr_entry *entry;
304 size_t size;
305 int error;
306 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
307
308 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
309 name_index, name, buffer, (long)buffer_size);
310
311 error = -ENODATA;
312 if (!EXT4_I(inode)->i_file_acl)
313 goto cleanup;
314 ea_idebug(inode, "reading block %llu",
315 (unsigned long long)EXT4_I(inode)->i_file_acl);
316 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
317 if (!bh)
318 goto cleanup;
319 ea_bdebug(bh, "b_count=%d, refcount=%d",
320 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
321 if (ext4_xattr_check_block(inode, bh)) {
322 bad_block:
323 EXT4_ERROR_INODE(inode, "bad block %llu",
324 EXT4_I(inode)->i_file_acl);
325 error = -EFSCORRUPTED;
326 goto cleanup;
327 }
328 ext4_xattr_cache_insert(ext4_mb_cache, bh);
329 entry = BFIRST(bh);
330 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
331 if (error == -EFSCORRUPTED)
332 goto bad_block;
333 if (error)
334 goto cleanup;
335 size = le32_to_cpu(entry->e_value_size);
336 if (buffer) {
337 error = -ERANGE;
338 if (size > buffer_size)
339 goto cleanup;
340 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
341 size);
342 }
343 error = size;
344
345 cleanup:
346 brelse(bh);
347 return error;
348 }
349
350 int
351 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
352 void *buffer, size_t buffer_size)
353 {
354 struct ext4_xattr_ibody_header *header;
355 struct ext4_xattr_entry *entry;
356 struct ext4_inode *raw_inode;
357 struct ext4_iloc iloc;
358 size_t size;
359 void *end;
360 int error;
361
362 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
363 return -ENODATA;
364 error = ext4_get_inode_loc(inode, &iloc);
365 if (error)
366 return error;
367 raw_inode = ext4_raw_inode(&iloc);
368 header = IHDR(inode, raw_inode);
369 entry = IFIRST(header);
370 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
371 error = xattr_check_inode(inode, header, end);
372 if (error)
373 goto cleanup;
374 error = ext4_xattr_find_entry(&entry, name_index, name,
375 end - (void *)entry, 0);
376 if (error)
377 goto cleanup;
378 size = le32_to_cpu(entry->e_value_size);
379 if (buffer) {
380 error = -ERANGE;
381 if (size > buffer_size)
382 goto cleanup;
383 memcpy(buffer, (void *)IFIRST(header) +
384 le16_to_cpu(entry->e_value_offs), size);
385 }
386 error = size;
387
388 cleanup:
389 brelse(iloc.bh);
390 return error;
391 }
392
393 /*
394 * ext4_xattr_get()
395 *
396 * Copy an extended attribute into the buffer
397 * provided, or compute the buffer size required.
398 * Buffer is NULL to compute the size of the buffer required.
399 *
400 * Returns a negative error number on failure, or the number of bytes
401 * used / required on success.
402 */
403 int
404 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
405 void *buffer, size_t buffer_size)
406 {
407 int error;
408
409 if (strlen(name) > 255)
410 return -ERANGE;
411
412 down_read(&EXT4_I(inode)->xattr_sem);
413 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
414 buffer_size);
415 if (error == -ENODATA)
416 error = ext4_xattr_block_get(inode, name_index, name, buffer,
417 buffer_size);
418 up_read(&EXT4_I(inode)->xattr_sem);
419 return error;
420 }
421
422 static int
423 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
424 char *buffer, size_t buffer_size)
425 {
426 size_t rest = buffer_size;
427
428 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
429 const struct xattr_handler *handler =
430 ext4_xattr_handler(entry->e_name_index);
431
432 if (handler && (!handler->list || handler->list(dentry))) {
433 const char *prefix = handler->prefix ?: handler->name;
434 size_t prefix_len = strlen(prefix);
435 size_t size = prefix_len + entry->e_name_len + 1;
436
437 if (buffer) {
438 if (size > rest)
439 return -ERANGE;
440 memcpy(buffer, prefix, prefix_len);
441 buffer += prefix_len;
442 memcpy(buffer, entry->e_name, entry->e_name_len);
443 buffer += entry->e_name_len;
444 *buffer++ = 0;
445 }
446 rest -= size;
447 }
448 }
449 return buffer_size - rest; /* total size */
450 }
451
452 static int
453 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
454 {
455 struct inode *inode = d_inode(dentry);
456 struct buffer_head *bh = NULL;
457 int error;
458 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
459
460 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
461 buffer, (long)buffer_size);
462
463 error = 0;
464 if (!EXT4_I(inode)->i_file_acl)
465 goto cleanup;
466 ea_idebug(inode, "reading block %llu",
467 (unsigned long long)EXT4_I(inode)->i_file_acl);
468 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
469 error = -EIO;
470 if (!bh)
471 goto cleanup;
472 ea_bdebug(bh, "b_count=%d, refcount=%d",
473 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
474 if (ext4_xattr_check_block(inode, bh)) {
475 EXT4_ERROR_INODE(inode, "bad block %llu",
476 EXT4_I(inode)->i_file_acl);
477 error = -EFSCORRUPTED;
478 goto cleanup;
479 }
480 ext4_xattr_cache_insert(ext4_mb_cache, bh);
481 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
482
483 cleanup:
484 brelse(bh);
485
486 return error;
487 }
488
489 static int
490 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
491 {
492 struct inode *inode = d_inode(dentry);
493 struct ext4_xattr_ibody_header *header;
494 struct ext4_inode *raw_inode;
495 struct ext4_iloc iloc;
496 void *end;
497 int error;
498
499 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
500 return 0;
501 error = ext4_get_inode_loc(inode, &iloc);
502 if (error)
503 return error;
504 raw_inode = ext4_raw_inode(&iloc);
505 header = IHDR(inode, raw_inode);
506 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
507 error = xattr_check_inode(inode, header, end);
508 if (error)
509 goto cleanup;
510 error = ext4_xattr_list_entries(dentry, IFIRST(header),
511 buffer, buffer_size);
512
513 cleanup:
514 brelse(iloc.bh);
515 return error;
516 }
517
518 /*
519 * ext4_xattr_list()
520 *
521 * Copy a list of attribute names into the buffer
522 * provided, or compute the buffer size required.
523 * Buffer is NULL to compute the size of the buffer required.
524 *
525 * Returns a negative error number on failure, or the number of bytes
526 * used / required on success.
527 */
528 static int
529 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
530 {
531 int ret, ret2;
532
533 down_read(&EXT4_I(d_inode(dentry))->xattr_sem);
534 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
535 if (ret < 0)
536 goto errout;
537 if (buffer) {
538 buffer += ret;
539 buffer_size -= ret;
540 }
541 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
542 if (ret < 0)
543 goto errout;
544 ret += ret2;
545 errout:
546 up_read(&EXT4_I(d_inode(dentry))->xattr_sem);
547 return ret;
548 }
549
550 /*
551 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
552 * not set, set it.
553 */
554 static void ext4_xattr_update_super_block(handle_t *handle,
555 struct super_block *sb)
556 {
557 if (ext4_has_feature_xattr(sb))
558 return;
559
560 BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
561 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
562 ext4_set_feature_xattr(sb);
563 ext4_handle_dirty_super(handle, sb);
564 }
565 }
566
567 /*
568 * Release the xattr block BH: If the reference count is > 1, decrement it;
569 * otherwise free the block.
570 */
571 static void
572 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
573 struct buffer_head *bh)
574 {
575 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
576 u32 hash, ref;
577 int error = 0;
578
579 BUFFER_TRACE(bh, "get_write_access");
580 error = ext4_journal_get_write_access(handle, bh);
581 if (error)
582 goto out;
583
584 lock_buffer(bh);
585 hash = le32_to_cpu(BHDR(bh)->h_hash);
586 ref = le32_to_cpu(BHDR(bh)->h_refcount);
587 if (ref == 1) {
588 ea_bdebug(bh, "refcount now=0; freeing");
589 /*
590 * This must happen under buffer lock for
591 * ext4_xattr_block_set() to reliably detect freed block
592 */
593 mb_cache_entry_delete_block(ext4_mb_cache, hash, bh->b_blocknr);
594 get_bh(bh);
595 unlock_buffer(bh);
596 ext4_free_blocks(handle, inode, bh, 0, 1,
597 EXT4_FREE_BLOCKS_METADATA |
598 EXT4_FREE_BLOCKS_FORGET);
599 } else {
600 ref--;
601 BHDR(bh)->h_refcount = cpu_to_le32(ref);
602 if (ref == EXT4_XATTR_REFCOUNT_MAX - 1) {
603 struct mb_cache_entry *ce;
604
605 ce = mb_cache_entry_get(ext4_mb_cache, hash,
606 bh->b_blocknr);
607 if (ce) {
608 ce->e_reusable = 1;
609 mb_cache_entry_put(ext4_mb_cache, ce);
610 }
611 }
612
613 ext4_xattr_block_csum_set(inode, bh);
614 /*
615 * Beware of this ugliness: Releasing of xattr block references
616 * from different inodes can race and so we have to protect
617 * from a race where someone else frees the block (and releases
618 * its journal_head) before we are done dirtying the buffer. In
619 * nojournal mode this race is harmless and we actually cannot
620 * call ext4_handle_dirty_metadata() with locked buffer as
621 * that function can call sync_dirty_buffer() so for that case
622 * we handle the dirtying after unlocking the buffer.
623 */
624 if (ext4_handle_valid(handle))
625 error = ext4_handle_dirty_metadata(handle, inode, bh);
626 unlock_buffer(bh);
627 if (!ext4_handle_valid(handle))
628 error = ext4_handle_dirty_metadata(handle, inode, bh);
629 if (IS_SYNC(inode))
630 ext4_handle_sync(handle);
631 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
632 ea_bdebug(bh, "refcount now=%d; releasing",
633 le32_to_cpu(BHDR(bh)->h_refcount));
634 }
635 out:
636 ext4_std_error(inode->i_sb, error);
637 return;
638 }
639
640 /*
641 * Find the available free space for EAs. This also returns the total number of
642 * bytes used by EA entries.
643 */
644 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
645 size_t *min_offs, void *base, int *total)
646 {
647 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
648 if (last->e_value_size) {
649 size_t offs = le16_to_cpu(last->e_value_offs);
650 if (offs < *min_offs)
651 *min_offs = offs;
652 }
653 if (total)
654 *total += EXT4_XATTR_LEN(last->e_name_len);
655 }
656 return (*min_offs - ((void *)last - base) - sizeof(__u32));
657 }
658
659 static int
660 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
661 {
662 struct ext4_xattr_entry *last;
663 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
664
665 /* Compute min_offs and last. */
666 last = s->first;
667 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
668 if (last->e_value_size) {
669 size_t offs = le16_to_cpu(last->e_value_offs);
670 if (offs < min_offs)
671 min_offs = offs;
672 }
673 }
674 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
675 if (!s->not_found) {
676 if (s->here->e_value_size) {
677 size_t size = le32_to_cpu(s->here->e_value_size);
678 free += EXT4_XATTR_SIZE(size);
679 }
680 free += EXT4_XATTR_LEN(name_len);
681 }
682 if (i->value) {
683 if (free < EXT4_XATTR_LEN(name_len) +
684 EXT4_XATTR_SIZE(i->value_len))
685 return -ENOSPC;
686 }
687
688 if (i->value && s->not_found) {
689 /* Insert the new name. */
690 size_t size = EXT4_XATTR_LEN(name_len);
691 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
692 memmove((void *)s->here + size, s->here, rest);
693 memset(s->here, 0, size);
694 s->here->e_name_index = i->name_index;
695 s->here->e_name_len = name_len;
696 memcpy(s->here->e_name, i->name, name_len);
697 } else {
698 if (s->here->e_value_size) {
699 void *first_val = s->base + min_offs;
700 size_t offs = le16_to_cpu(s->here->e_value_offs);
701 void *val = s->base + offs;
702 size_t size = EXT4_XATTR_SIZE(
703 le32_to_cpu(s->here->e_value_size));
704
705 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
706 /* The old and the new value have the same
707 size. Just replace. */
708 s->here->e_value_size =
709 cpu_to_le32(i->value_len);
710 if (i->value == EXT4_ZERO_XATTR_VALUE) {
711 memset(val, 0, size);
712 } else {
713 /* Clear pad bytes first. */
714 memset(val + size - EXT4_XATTR_PAD, 0,
715 EXT4_XATTR_PAD);
716 memcpy(val, i->value, i->value_len);
717 }
718 return 0;
719 }
720
721 /* Remove the old value. */
722 memmove(first_val + size, first_val, val - first_val);
723 memset(first_val, 0, size);
724 s->here->e_value_size = 0;
725 s->here->e_value_offs = 0;
726 min_offs += size;
727
728 /* Adjust all value offsets. */
729 last = s->first;
730 while (!IS_LAST_ENTRY(last)) {
731 size_t o = le16_to_cpu(last->e_value_offs);
732 if (last->e_value_size && o < offs)
733 last->e_value_offs =
734 cpu_to_le16(o + size);
735 last = EXT4_XATTR_NEXT(last);
736 }
737 }
738 if (!i->value) {
739 /* Remove the old name. */
740 size_t size = EXT4_XATTR_LEN(name_len);
741 last = ENTRY((void *)last - size);
742 memmove(s->here, (void *)s->here + size,
743 (void *)last - (void *)s->here + sizeof(__u32));
744 memset(last, 0, size);
745 }
746 }
747
748 if (i->value) {
749 /* Insert the new value. */
750 s->here->e_value_size = cpu_to_le32(i->value_len);
751 if (i->value_len) {
752 size_t size = EXT4_XATTR_SIZE(i->value_len);
753 void *val = s->base + min_offs - size;
754 s->here->e_value_offs = cpu_to_le16(min_offs - size);
755 if (i->value == EXT4_ZERO_XATTR_VALUE) {
756 memset(val, 0, size);
757 } else {
758 /* Clear the pad bytes first. */
759 memset(val + size - EXT4_XATTR_PAD, 0,
760 EXT4_XATTR_PAD);
761 memcpy(val, i->value, i->value_len);
762 }
763 }
764 }
765 return 0;
766 }
767
768 struct ext4_xattr_block_find {
769 struct ext4_xattr_search s;
770 struct buffer_head *bh;
771 };
772
773 static int
774 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
775 struct ext4_xattr_block_find *bs)
776 {
777 struct super_block *sb = inode->i_sb;
778 int error;
779
780 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
781 i->name_index, i->name, i->value, (long)i->value_len);
782
783 if (EXT4_I(inode)->i_file_acl) {
784 /* The inode already has an extended attribute block. */
785 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
786 error = -EIO;
787 if (!bs->bh)
788 goto cleanup;
789 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
790 atomic_read(&(bs->bh->b_count)),
791 le32_to_cpu(BHDR(bs->bh)->h_refcount));
792 if (ext4_xattr_check_block(inode, bs->bh)) {
793 EXT4_ERROR_INODE(inode, "bad block %llu",
794 EXT4_I(inode)->i_file_acl);
795 error = -EFSCORRUPTED;
796 goto cleanup;
797 }
798 /* Find the named attribute. */
799 bs->s.base = BHDR(bs->bh);
800 bs->s.first = BFIRST(bs->bh);
801 bs->s.end = bs->bh->b_data + bs->bh->b_size;
802 bs->s.here = bs->s.first;
803 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
804 i->name, bs->bh->b_size, 1);
805 if (error && error != -ENODATA)
806 goto cleanup;
807 bs->s.not_found = error;
808 }
809 error = 0;
810
811 cleanup:
812 return error;
813 }
814
815 static int
816 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
817 struct ext4_xattr_info *i,
818 struct ext4_xattr_block_find *bs)
819 {
820 struct super_block *sb = inode->i_sb;
821 struct buffer_head *new_bh = NULL;
822 struct ext4_xattr_search *s = &bs->s;
823 struct mb_cache_entry *ce = NULL;
824 int error = 0;
825 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
826
827 #define header(x) ((struct ext4_xattr_header *)(x))
828
829 if (i->value && i->value_len > sb->s_blocksize)
830 return -ENOSPC;
831 if (s->base) {
832 BUFFER_TRACE(bs->bh, "get_write_access");
833 error = ext4_journal_get_write_access(handle, bs->bh);
834 if (error)
835 goto cleanup;
836 lock_buffer(bs->bh);
837
838 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
839 __u32 hash = le32_to_cpu(BHDR(bs->bh)->h_hash);
840
841 /*
842 * This must happen under buffer lock for
843 * ext4_xattr_block_set() to reliably detect modified
844 * block
845 */
846 mb_cache_entry_delete_block(ext4_mb_cache, hash,
847 bs->bh->b_blocknr);
848 ea_bdebug(bs->bh, "modifying in-place");
849 error = ext4_xattr_set_entry(i, s);
850 if (!error) {
851 if (!IS_LAST_ENTRY(s->first))
852 ext4_xattr_rehash(header(s->base),
853 s->here);
854 ext4_xattr_cache_insert(ext4_mb_cache,
855 bs->bh);
856 }
857 ext4_xattr_block_csum_set(inode, bs->bh);
858 unlock_buffer(bs->bh);
859 if (error == -EFSCORRUPTED)
860 goto bad_block;
861 if (!error)
862 error = ext4_handle_dirty_metadata(handle,
863 inode,
864 bs->bh);
865 if (error)
866 goto cleanup;
867 goto inserted;
868 } else {
869 int offset = (char *)s->here - bs->bh->b_data;
870
871 unlock_buffer(bs->bh);
872 ea_bdebug(bs->bh, "cloning");
873 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
874 error = -ENOMEM;
875 if (s->base == NULL)
876 goto cleanup;
877 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
878 s->first = ENTRY(header(s->base)+1);
879 header(s->base)->h_refcount = cpu_to_le32(1);
880 s->here = ENTRY(s->base + offset);
881 s->end = s->base + bs->bh->b_size;
882 }
883 } else {
884 /* Allocate a buffer where we construct the new block. */
885 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
886 /* assert(header == s->base) */
887 error = -ENOMEM;
888 if (s->base == NULL)
889 goto cleanup;
890 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
891 header(s->base)->h_blocks = cpu_to_le32(1);
892 header(s->base)->h_refcount = cpu_to_le32(1);
893 s->first = ENTRY(header(s->base)+1);
894 s->here = ENTRY(header(s->base)+1);
895 s->end = s->base + sb->s_blocksize;
896 }
897
898 error = ext4_xattr_set_entry(i, s);
899 if (error == -EFSCORRUPTED)
900 goto bad_block;
901 if (error)
902 goto cleanup;
903 if (!IS_LAST_ENTRY(s->first))
904 ext4_xattr_rehash(header(s->base), s->here);
905
906 inserted:
907 if (!IS_LAST_ENTRY(s->first)) {
908 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
909 if (new_bh) {
910 /* We found an identical block in the cache. */
911 if (new_bh == bs->bh)
912 ea_bdebug(new_bh, "keeping");
913 else {
914 u32 ref;
915
916 /* The old block is released after updating
917 the inode. */
918 error = dquot_alloc_block(inode,
919 EXT4_C2B(EXT4_SB(sb), 1));
920 if (error)
921 goto cleanup;
922 BUFFER_TRACE(new_bh, "get_write_access");
923 error = ext4_journal_get_write_access(handle,
924 new_bh);
925 if (error)
926 goto cleanup_dquot;
927 lock_buffer(new_bh);
928 /*
929 * We have to be careful about races with
930 * freeing, rehashing or adding references to
931 * xattr block. Once we hold buffer lock xattr
932 * block's state is stable so we can check
933 * whether the block got freed / rehashed or
934 * not. Since we unhash mbcache entry under
935 * buffer lock when freeing / rehashing xattr
936 * block, checking whether entry is still
937 * hashed is reliable. Same rules hold for
938 * e_reusable handling.
939 */
940 if (hlist_bl_unhashed(&ce->e_hash_list) ||
941 !ce->e_reusable) {
942 /*
943 * Undo everything and check mbcache
944 * again.
945 */
946 unlock_buffer(new_bh);
947 dquot_free_block(inode,
948 EXT4_C2B(EXT4_SB(sb),
949 1));
950 brelse(new_bh);
951 mb_cache_entry_put(ext4_mb_cache, ce);
952 ce = NULL;
953 new_bh = NULL;
954 goto inserted;
955 }
956 ref = le32_to_cpu(BHDR(new_bh)->h_refcount) + 1;
957 BHDR(new_bh)->h_refcount = cpu_to_le32(ref);
958 if (ref >= EXT4_XATTR_REFCOUNT_MAX)
959 ce->e_reusable = 0;
960 ea_bdebug(new_bh, "reusing; refcount now=%d",
961 ref);
962 ext4_xattr_block_csum_set(inode, new_bh);
963 unlock_buffer(new_bh);
964 error = ext4_handle_dirty_metadata(handle,
965 inode,
966 new_bh);
967 if (error)
968 goto cleanup_dquot;
969 }
970 mb_cache_entry_touch(ext4_mb_cache, ce);
971 mb_cache_entry_put(ext4_mb_cache, ce);
972 ce = NULL;
973 } else if (bs->bh && s->base == bs->bh->b_data) {
974 /* We were modifying this block in-place. */
975 ea_bdebug(bs->bh, "keeping this block");
976 new_bh = bs->bh;
977 get_bh(new_bh);
978 } else {
979 /* We need to allocate a new block */
980 ext4_fsblk_t goal, block;
981
982 goal = ext4_group_first_block_no(sb,
983 EXT4_I(inode)->i_block_group);
984
985 /* non-extent files can't have physical blocks past 2^32 */
986 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
987 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
988
989 block = ext4_new_meta_blocks(handle, inode, goal, 0,
990 NULL, &error);
991 if (error)
992 goto cleanup;
993
994 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
995 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
996
997 ea_idebug(inode, "creating block %llu",
998 (unsigned long long)block);
999
1000 new_bh = sb_getblk(sb, block);
1001 if (unlikely(!new_bh)) {
1002 error = -ENOMEM;
1003 getblk_failed:
1004 ext4_free_blocks(handle, inode, NULL, block, 1,
1005 EXT4_FREE_BLOCKS_METADATA);
1006 goto cleanup;
1007 }
1008 lock_buffer(new_bh);
1009 error = ext4_journal_get_create_access(handle, new_bh);
1010 if (error) {
1011 unlock_buffer(new_bh);
1012 error = -EIO;
1013 goto getblk_failed;
1014 }
1015 memcpy(new_bh->b_data, s->base, new_bh->b_size);
1016 ext4_xattr_block_csum_set(inode, new_bh);
1017 set_buffer_uptodate(new_bh);
1018 unlock_buffer(new_bh);
1019 ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
1020 error = ext4_handle_dirty_metadata(handle, inode,
1021 new_bh);
1022 if (error)
1023 goto cleanup;
1024 }
1025 }
1026
1027 /* Update the inode. */
1028 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
1029
1030 /* Drop the previous xattr block. */
1031 if (bs->bh && bs->bh != new_bh)
1032 ext4_xattr_release_block(handle, inode, bs->bh);
1033 error = 0;
1034
1035 cleanup:
1036 if (ce)
1037 mb_cache_entry_put(ext4_mb_cache, ce);
1038 brelse(new_bh);
1039 if (!(bs->bh && s->base == bs->bh->b_data))
1040 kfree(s->base);
1041
1042 return error;
1043
1044 cleanup_dquot:
1045 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
1046 goto cleanup;
1047
1048 bad_block:
1049 EXT4_ERROR_INODE(inode, "bad block %llu",
1050 EXT4_I(inode)->i_file_acl);
1051 goto cleanup;
1052
1053 #undef header
1054 }
1055
1056 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
1057 struct ext4_xattr_ibody_find *is)
1058 {
1059 struct ext4_xattr_ibody_header *header;
1060 struct ext4_inode *raw_inode;
1061 int error;
1062
1063 if (EXT4_I(inode)->i_extra_isize == 0)
1064 return 0;
1065 raw_inode = ext4_raw_inode(&is->iloc);
1066 header = IHDR(inode, raw_inode);
1067 is->s.base = is->s.first = IFIRST(header);
1068 is->s.here = is->s.first;
1069 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1070 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
1071 error = xattr_check_inode(inode, header, is->s.end);
1072 if (error)
1073 return error;
1074 /* Find the named attribute. */
1075 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
1076 i->name, is->s.end -
1077 (void *)is->s.base, 0);
1078 if (error && error != -ENODATA)
1079 return error;
1080 is->s.not_found = error;
1081 }
1082 return 0;
1083 }
1084
1085 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
1086 struct ext4_xattr_info *i,
1087 struct ext4_xattr_ibody_find *is)
1088 {
1089 struct ext4_xattr_ibody_header *header;
1090 struct ext4_xattr_search *s = &is->s;
1091 int error;
1092
1093 if (EXT4_I(inode)->i_extra_isize == 0)
1094 return -ENOSPC;
1095 error = ext4_xattr_set_entry(i, s);
1096 if (error) {
1097 if (error == -ENOSPC &&
1098 ext4_has_inline_data(inode)) {
1099 error = ext4_try_to_evict_inline_data(handle, inode,
1100 EXT4_XATTR_LEN(strlen(i->name) +
1101 EXT4_XATTR_SIZE(i->value_len)));
1102 if (error)
1103 return error;
1104 error = ext4_xattr_ibody_find(inode, i, is);
1105 if (error)
1106 return error;
1107 error = ext4_xattr_set_entry(i, s);
1108 }
1109 if (error)
1110 return error;
1111 }
1112 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1113 if (!IS_LAST_ENTRY(s->first)) {
1114 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1115 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1116 } else {
1117 header->h_magic = cpu_to_le32(0);
1118 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1119 }
1120 return 0;
1121 }
1122
1123 static int ext4_xattr_ibody_set(struct inode *inode,
1124 struct ext4_xattr_info *i,
1125 struct ext4_xattr_ibody_find *is)
1126 {
1127 struct ext4_xattr_ibody_header *header;
1128 struct ext4_xattr_search *s = &is->s;
1129 int error;
1130
1131 if (EXT4_I(inode)->i_extra_isize == 0)
1132 return -ENOSPC;
1133 error = ext4_xattr_set_entry(i, s);
1134 if (error)
1135 return error;
1136 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1137 if (!IS_LAST_ENTRY(s->first)) {
1138 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1139 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1140 } else {
1141 header->h_magic = cpu_to_le32(0);
1142 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1143 }
1144 return 0;
1145 }
1146
1147 static int ext4_xattr_value_same(struct ext4_xattr_search *s,
1148 struct ext4_xattr_info *i)
1149 {
1150 void *value;
1151
1152 if (le32_to_cpu(s->here->e_value_size) != i->value_len)
1153 return 0;
1154 value = ((void *)s->base) + le16_to_cpu(s->here->e_value_offs);
1155 return !memcmp(value, i->value, i->value_len);
1156 }
1157
1158 /*
1159 * ext4_xattr_set_handle()
1160 *
1161 * Create, replace or remove an extended attribute for this inode. Value
1162 * is NULL to remove an existing extended attribute, and non-NULL to
1163 * either replace an existing extended attribute, or create a new extended
1164 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1165 * specify that an extended attribute must exist and must not exist
1166 * previous to the call, respectively.
1167 *
1168 * Returns 0, or a negative error number on failure.
1169 */
1170 int
1171 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1172 const char *name, const void *value, size_t value_len,
1173 int flags)
1174 {
1175 struct ext4_xattr_info i = {
1176 .name_index = name_index,
1177 .name = name,
1178 .value = value,
1179 .value_len = value_len,
1180
1181 };
1182 struct ext4_xattr_ibody_find is = {
1183 .s = { .not_found = -ENODATA, },
1184 };
1185 struct ext4_xattr_block_find bs = {
1186 .s = { .not_found = -ENODATA, },
1187 };
1188 int no_expand;
1189 int error;
1190
1191 if (!name)
1192 return -EINVAL;
1193 if (strlen(name) > 255)
1194 return -ERANGE;
1195 ext4_write_lock_xattr(inode, &no_expand);
1196
1197 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1198 if (error)
1199 goto cleanup;
1200
1201 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1202 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1203 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1204 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1205 }
1206
1207 error = ext4_xattr_ibody_find(inode, &i, &is);
1208 if (error)
1209 goto cleanup;
1210 if (is.s.not_found)
1211 error = ext4_xattr_block_find(inode, &i, &bs);
1212 if (error)
1213 goto cleanup;
1214 if (is.s.not_found && bs.s.not_found) {
1215 error = -ENODATA;
1216 if (flags & XATTR_REPLACE)
1217 goto cleanup;
1218 error = 0;
1219 if (!value)
1220 goto cleanup;
1221 } else {
1222 error = -EEXIST;
1223 if (flags & XATTR_CREATE)
1224 goto cleanup;
1225 }
1226 if (!value) {
1227 if (!is.s.not_found)
1228 error = ext4_xattr_ibody_set(inode, &i, &is);
1229 else if (!bs.s.not_found)
1230 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1231 } else {
1232 error = 0;
1233 /* Xattr value did not change? Save us some work and bail out */
1234 if (!is.s.not_found && ext4_xattr_value_same(&is.s, &i))
1235 goto cleanup;
1236 if (!bs.s.not_found && ext4_xattr_value_same(&bs.s, &i))
1237 goto cleanup;
1238
1239 error = ext4_xattr_ibody_set(inode, &i, &is);
1240 if (!error && !bs.s.not_found) {
1241 i.value = NULL;
1242 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1243 } else if (error == -ENOSPC) {
1244 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1245 error = ext4_xattr_block_find(inode, &i, &bs);
1246 if (error)
1247 goto cleanup;
1248 }
1249 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1250 if (error)
1251 goto cleanup;
1252 if (!is.s.not_found) {
1253 i.value = NULL;
1254 error = ext4_xattr_ibody_set(inode, &i, &is);
1255 }
1256 }
1257 }
1258 if (!error) {
1259 ext4_xattr_update_super_block(handle, inode->i_sb);
1260 inode->i_ctime = current_time(inode);
1261 if (!value)
1262 no_expand = 0;
1263 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1264 /*
1265 * The bh is consumed by ext4_mark_iloc_dirty, even with
1266 * error != 0.
1267 */
1268 is.iloc.bh = NULL;
1269 if (IS_SYNC(inode))
1270 ext4_handle_sync(handle);
1271 }
1272
1273 cleanup:
1274 brelse(is.iloc.bh);
1275 brelse(bs.bh);
1276 ext4_write_unlock_xattr(inode, &no_expand);
1277 return error;
1278 }
1279
1280 /*
1281 * ext4_xattr_set()
1282 *
1283 * Like ext4_xattr_set_handle, but start from an inode. This extended
1284 * attribute modification is a filesystem transaction by itself.
1285 *
1286 * Returns 0, or a negative error number on failure.
1287 */
1288 int
1289 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1290 const void *value, size_t value_len, int flags)
1291 {
1292 handle_t *handle;
1293 int error, retries = 0;
1294 int credits = ext4_jbd2_credits_xattr(inode);
1295
1296 retry:
1297 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1298 if (IS_ERR(handle)) {
1299 error = PTR_ERR(handle);
1300 } else {
1301 int error2;
1302
1303 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1304 value, value_len, flags);
1305 error2 = ext4_journal_stop(handle);
1306 if (error == -ENOSPC &&
1307 ext4_should_retry_alloc(inode->i_sb, &retries))
1308 goto retry;
1309 if (error == 0)
1310 error = error2;
1311 }
1312
1313 return error;
1314 }
1315
1316 /*
1317 * Shift the EA entries in the inode to create space for the increased
1318 * i_extra_isize.
1319 */
1320 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1321 int value_offs_shift, void *to,
1322 void *from, size_t n)
1323 {
1324 struct ext4_xattr_entry *last = entry;
1325 int new_offs;
1326
1327 /* We always shift xattr headers further thus offsets get lower */
1328 BUG_ON(value_offs_shift > 0);
1329
1330 /* Adjust the value offsets of the entries */
1331 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1332 if (last->e_value_size) {
1333 new_offs = le16_to_cpu(last->e_value_offs) +
1334 value_offs_shift;
1335 last->e_value_offs = cpu_to_le16(new_offs);
1336 }
1337 }
1338 /* Shift the entries by n bytes */
1339 memmove(to, from, n);
1340 }
1341
1342 /*
1343 * Move xattr pointed to by 'entry' from inode into external xattr block
1344 */
1345 static int ext4_xattr_move_to_block(handle_t *handle, struct inode *inode,
1346 struct ext4_inode *raw_inode,
1347 struct ext4_xattr_entry *entry)
1348 {
1349 struct ext4_xattr_ibody_find *is = NULL;
1350 struct ext4_xattr_block_find *bs = NULL;
1351 char *buffer = NULL, *b_entry_name = NULL;
1352 size_t value_offs, value_size;
1353 struct ext4_xattr_info i = {
1354 .value = NULL,
1355 .value_len = 0,
1356 .name_index = entry->e_name_index,
1357 };
1358 struct ext4_xattr_ibody_header *header = IHDR(inode, raw_inode);
1359 int error;
1360
1361 value_offs = le16_to_cpu(entry->e_value_offs);
1362 value_size = le32_to_cpu(entry->e_value_size);
1363
1364 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1365 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1366 buffer = kmalloc(value_size, GFP_NOFS);
1367 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1368 if (!is || !bs || !buffer || !b_entry_name) {
1369 error = -ENOMEM;
1370 goto out;
1371 }
1372
1373 is->s.not_found = -ENODATA;
1374 bs->s.not_found = -ENODATA;
1375 is->iloc.bh = NULL;
1376 bs->bh = NULL;
1377
1378 /* Save the entry name and the entry value */
1379 memcpy(buffer, (void *)IFIRST(header) + value_offs, value_size);
1380 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1381 b_entry_name[entry->e_name_len] = '\0';
1382 i.name = b_entry_name;
1383
1384 error = ext4_get_inode_loc(inode, &is->iloc);
1385 if (error)
1386 goto out;
1387
1388 error = ext4_xattr_ibody_find(inode, &i, is);
1389 if (error)
1390 goto out;
1391
1392 /* Remove the chosen entry from the inode */
1393 error = ext4_xattr_ibody_set(inode, &i, is);
1394 if (error)
1395 goto out;
1396
1397 i.name = b_entry_name;
1398 i.value = buffer;
1399 i.value_len = value_size;
1400 error = ext4_xattr_block_find(inode, &i, bs);
1401 if (error)
1402 goto out;
1403
1404 /* Add entry which was removed from the inode into the block */
1405 error = ext4_xattr_block_set(handle, inode, &i, bs);
1406 if (error)
1407 goto out;
1408 error = 0;
1409 out:
1410 kfree(b_entry_name);
1411 kfree(buffer);
1412 if (is)
1413 brelse(is->iloc.bh);
1414 kfree(is);
1415 kfree(bs);
1416
1417 return error;
1418 }
1419
1420 static int ext4_xattr_make_inode_space(handle_t *handle, struct inode *inode,
1421 struct ext4_inode *raw_inode,
1422 int isize_diff, size_t ifree,
1423 size_t bfree, int *total_ino)
1424 {
1425 struct ext4_xattr_ibody_header *header = IHDR(inode, raw_inode);
1426 struct ext4_xattr_entry *small_entry;
1427 struct ext4_xattr_entry *entry;
1428 struct ext4_xattr_entry *last;
1429 unsigned int entry_size; /* EA entry size */
1430 unsigned int total_size; /* EA entry size + value size */
1431 unsigned int min_total_size;
1432 int error;
1433
1434 while (isize_diff > ifree) {
1435 entry = NULL;
1436 small_entry = NULL;
1437 min_total_size = ~0U;
1438 last = IFIRST(header);
1439 /* Find the entry best suited to be pushed into EA block */
1440 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1441 total_size =
1442 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1443 EXT4_XATTR_LEN(last->e_name_len);
1444 if (total_size <= bfree &&
1445 total_size < min_total_size) {
1446 if (total_size + ifree < isize_diff) {
1447 small_entry = last;
1448 } else {
1449 entry = last;
1450 min_total_size = total_size;
1451 }
1452 }
1453 }
1454
1455 if (entry == NULL) {
1456 if (small_entry == NULL)
1457 return -ENOSPC;
1458 entry = small_entry;
1459 }
1460
1461 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1462 total_size = entry_size +
1463 EXT4_XATTR_SIZE(le32_to_cpu(entry->e_value_size));
1464 error = ext4_xattr_move_to_block(handle, inode, raw_inode,
1465 entry);
1466 if (error)
1467 return error;
1468
1469 *total_ino -= entry_size;
1470 ifree += total_size;
1471 bfree -= total_size;
1472 }
1473
1474 return 0;
1475 }
1476
1477 /*
1478 * Expand an inode by new_extra_isize bytes when EAs are present.
1479 * Returns 0 on success or negative error number on failure.
1480 */
1481 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1482 struct ext4_inode *raw_inode, handle_t *handle)
1483 {
1484 struct ext4_xattr_ibody_header *header;
1485 struct buffer_head *bh = NULL;
1486 size_t min_offs;
1487 size_t ifree, bfree;
1488 int total_ino;
1489 void *base, *end;
1490 int error = 0, tried_min_extra_isize = 0;
1491 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1492 int isize_diff; /* How much do we need to grow i_extra_isize */
1493 int no_expand;
1494
1495 if (ext4_write_trylock_xattr(inode, &no_expand) == 0)
1496 return 0;
1497
1498 retry:
1499 isize_diff = new_extra_isize - EXT4_I(inode)->i_extra_isize;
1500 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
1501 goto out;
1502
1503 header = IHDR(inode, raw_inode);
1504
1505 /*
1506 * Check if enough free space is available in the inode to shift the
1507 * entries ahead by new_extra_isize.
1508 */
1509
1510 base = IFIRST(header);
1511 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1512 min_offs = end - base;
1513 total_ino = sizeof(struct ext4_xattr_ibody_header);
1514
1515 error = xattr_check_inode(inode, header, end);
1516 if (error)
1517 goto cleanup;
1518
1519 ifree = ext4_xattr_free_space(base, &min_offs, base, &total_ino);
1520 if (ifree >= isize_diff)
1521 goto shift;
1522
1523 /*
1524 * Enough free space isn't available in the inode, check if
1525 * EA block can hold new_extra_isize bytes.
1526 */
1527 if (EXT4_I(inode)->i_file_acl) {
1528 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1529 error = -EIO;
1530 if (!bh)
1531 goto cleanup;
1532 if (ext4_xattr_check_block(inode, bh)) {
1533 EXT4_ERROR_INODE(inode, "bad block %llu",
1534 EXT4_I(inode)->i_file_acl);
1535 error = -EFSCORRUPTED;
1536 goto cleanup;
1537 }
1538 base = BHDR(bh);
1539 end = bh->b_data + bh->b_size;
1540 min_offs = end - base;
1541 bfree = ext4_xattr_free_space(BFIRST(bh), &min_offs, base,
1542 NULL);
1543 if (bfree + ifree < isize_diff) {
1544 if (!tried_min_extra_isize && s_min_extra_isize) {
1545 tried_min_extra_isize++;
1546 new_extra_isize = s_min_extra_isize;
1547 brelse(bh);
1548 goto retry;
1549 }
1550 error = -ENOSPC;
1551 goto cleanup;
1552 }
1553 } else {
1554 bfree = inode->i_sb->s_blocksize;
1555 }
1556
1557 error = ext4_xattr_make_inode_space(handle, inode, raw_inode,
1558 isize_diff, ifree, bfree,
1559 &total_ino);
1560 if (error) {
1561 if (error == -ENOSPC && !tried_min_extra_isize &&
1562 s_min_extra_isize) {
1563 tried_min_extra_isize++;
1564 new_extra_isize = s_min_extra_isize;
1565 brelse(bh);
1566 goto retry;
1567 }
1568 goto cleanup;
1569 }
1570 shift:
1571 /* Adjust the offsets and shift the remaining entries ahead */
1572 ext4_xattr_shift_entries(IFIRST(header), EXT4_I(inode)->i_extra_isize
1573 - new_extra_isize, (void *)raw_inode +
1574 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1575 (void *)header, total_ino);
1576 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1577 brelse(bh);
1578 out:
1579 ext4_write_unlock_xattr(inode, &no_expand);
1580 return 0;
1581
1582 cleanup:
1583 brelse(bh);
1584 /*
1585 * Inode size expansion failed; don't try again
1586 */
1587 no_expand = 1;
1588 ext4_write_unlock_xattr(inode, &no_expand);
1589 return error;
1590 }
1591
1592
1593
1594 /*
1595 * ext4_xattr_delete_inode()
1596 *
1597 * Free extended attribute resources associated with this inode. This
1598 * is called immediately before an inode is freed. We have exclusive
1599 * access to the inode.
1600 */
1601 void
1602 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1603 {
1604 struct buffer_head *bh = NULL;
1605
1606 if (!EXT4_I(inode)->i_file_acl)
1607 goto cleanup;
1608 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1609 if (!bh) {
1610 EXT4_ERROR_INODE(inode, "block %llu read error",
1611 EXT4_I(inode)->i_file_acl);
1612 goto cleanup;
1613 }
1614 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1615 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1616 EXT4_ERROR_INODE(inode, "bad block %llu",
1617 EXT4_I(inode)->i_file_acl);
1618 goto cleanup;
1619 }
1620 ext4_xattr_release_block(handle, inode, bh);
1621 EXT4_I(inode)->i_file_acl = 0;
1622
1623 cleanup:
1624 brelse(bh);
1625 }
1626
1627 /*
1628 * ext4_xattr_cache_insert()
1629 *
1630 * Create a new entry in the extended attribute cache, and insert
1631 * it unless such an entry is already in the cache.
1632 *
1633 * Returns 0, or a negative error number on failure.
1634 */
1635 static void
1636 ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
1637 {
1638 struct ext4_xattr_header *header = BHDR(bh);
1639 __u32 hash = le32_to_cpu(header->h_hash);
1640 int reusable = le32_to_cpu(header->h_refcount) <
1641 EXT4_XATTR_REFCOUNT_MAX;
1642 int error;
1643
1644 error = mb_cache_entry_create(ext4_mb_cache, GFP_NOFS, hash,
1645 bh->b_blocknr, reusable);
1646 if (error) {
1647 if (error == -EBUSY)
1648 ea_bdebug(bh, "already in cache");
1649 } else
1650 ea_bdebug(bh, "inserting [%x]", (int)hash);
1651 }
1652
1653 /*
1654 * ext4_xattr_cmp()
1655 *
1656 * Compare two extended attribute blocks for equality.
1657 *
1658 * Returns 0 if the blocks are equal, 1 if they differ, and
1659 * a negative error number on errors.
1660 */
1661 static int
1662 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1663 struct ext4_xattr_header *header2)
1664 {
1665 struct ext4_xattr_entry *entry1, *entry2;
1666
1667 entry1 = ENTRY(header1+1);
1668 entry2 = ENTRY(header2+1);
1669 while (!IS_LAST_ENTRY(entry1)) {
1670 if (IS_LAST_ENTRY(entry2))
1671 return 1;
1672 if (entry1->e_hash != entry2->e_hash ||
1673 entry1->e_name_index != entry2->e_name_index ||
1674 entry1->e_name_len != entry2->e_name_len ||
1675 entry1->e_value_size != entry2->e_value_size ||
1676 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1677 return 1;
1678 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1679 return -EFSCORRUPTED;
1680 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1681 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1682 le32_to_cpu(entry1->e_value_size)))
1683 return 1;
1684
1685 entry1 = EXT4_XATTR_NEXT(entry1);
1686 entry2 = EXT4_XATTR_NEXT(entry2);
1687 }
1688 if (!IS_LAST_ENTRY(entry2))
1689 return 1;
1690 return 0;
1691 }
1692
1693 /*
1694 * ext4_xattr_cache_find()
1695 *
1696 * Find an identical extended attribute block.
1697 *
1698 * Returns a pointer to the block found, or NULL if such a block was
1699 * not found or an error occurred.
1700 */
1701 static struct buffer_head *
1702 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1703 struct mb_cache_entry **pce)
1704 {
1705 __u32 hash = le32_to_cpu(header->h_hash);
1706 struct mb_cache_entry *ce;
1707 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
1708
1709 if (!header->h_hash)
1710 return NULL; /* never share */
1711 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1712 ce = mb_cache_entry_find_first(ext4_mb_cache, hash);
1713 while (ce) {
1714 struct buffer_head *bh;
1715
1716 bh = sb_bread(inode->i_sb, ce->e_block);
1717 if (!bh) {
1718 EXT4_ERROR_INODE(inode, "block %lu read error",
1719 (unsigned long) ce->e_block);
1720 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1721 *pce = ce;
1722 return bh;
1723 }
1724 brelse(bh);
1725 ce = mb_cache_entry_find_next(ext4_mb_cache, ce);
1726 }
1727 return NULL;
1728 }
1729
1730 #define NAME_HASH_SHIFT 5
1731 #define VALUE_HASH_SHIFT 16
1732
1733 /*
1734 * ext4_xattr_hash_entry()
1735 *
1736 * Compute the hash of an extended attribute.
1737 */
1738 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1739 struct ext4_xattr_entry *entry)
1740 {
1741 __u32 hash = 0;
1742 char *name = entry->e_name;
1743 int n;
1744
1745 for (n = 0; n < entry->e_name_len; n++) {
1746 hash = (hash << NAME_HASH_SHIFT) ^
1747 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1748 *name++;
1749 }
1750
1751 if (entry->e_value_size != 0) {
1752 __le32 *value = (__le32 *)((char *)header +
1753 le16_to_cpu(entry->e_value_offs));
1754 for (n = (le32_to_cpu(entry->e_value_size) +
1755 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1756 hash = (hash << VALUE_HASH_SHIFT) ^
1757 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1758 le32_to_cpu(*value++);
1759 }
1760 }
1761 entry->e_hash = cpu_to_le32(hash);
1762 }
1763
1764 #undef NAME_HASH_SHIFT
1765 #undef VALUE_HASH_SHIFT
1766
1767 #define BLOCK_HASH_SHIFT 16
1768
1769 /*
1770 * ext4_xattr_rehash()
1771 *
1772 * Re-compute the extended attribute hash value after an entry has changed.
1773 */
1774 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1775 struct ext4_xattr_entry *entry)
1776 {
1777 struct ext4_xattr_entry *here;
1778 __u32 hash = 0;
1779
1780 ext4_xattr_hash_entry(header, entry);
1781 here = ENTRY(header+1);
1782 while (!IS_LAST_ENTRY(here)) {
1783 if (!here->e_hash) {
1784 /* Block is not shared if an entry's hash value == 0 */
1785 hash = 0;
1786 break;
1787 }
1788 hash = (hash << BLOCK_HASH_SHIFT) ^
1789 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1790 le32_to_cpu(here->e_hash);
1791 here = EXT4_XATTR_NEXT(here);
1792 }
1793 header->h_hash = cpu_to_le32(hash);
1794 }
1795
1796 #undef BLOCK_HASH_SHIFT
1797
1798 #define HASH_BUCKET_BITS 10
1799
1800 struct mb_cache *
1801 ext4_xattr_create_cache(void)
1802 {
1803 return mb_cache_create(HASH_BUCKET_BITS);
1804 }
1805
1806 void ext4_xattr_destroy_cache(struct mb_cache *cache)
1807 {
1808 if (cache)
1809 mb_cache_destroy(cache);
1810 }
1811