]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blob - fs/ext4/xattr.c
Merge branch 'irq-core-for-linus' of git://git.kernel.org/pub/scm/linux/kernel/git...
[mirror_ubuntu-artful-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 <linux/rwsem.h>
59 #include "ext4_jbd2.h"
60 #include "ext4.h"
61 #include "xattr.h"
62 #include "acl.h"
63
64 #ifdef EXT4_XATTR_DEBUG
65 # define ea_idebug(inode, f...) do { \
66 printk(KERN_DEBUG "inode %s:%lu: ", \
67 inode->i_sb->s_id, inode->i_ino); \
68 printk(f); \
69 printk("\n"); \
70 } while (0)
71 # define ea_bdebug(bh, f...) do { \
72 char b[BDEVNAME_SIZE]; \
73 printk(KERN_DEBUG "block %s:%lu: ", \
74 bdevname(bh->b_bdev, b), \
75 (unsigned long) bh->b_blocknr); \
76 printk(f); \
77 printk("\n"); \
78 } while (0)
79 #else
80 # define ea_idebug(inode, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
81 # define ea_bdebug(bh, fmt, ...) no_printk(fmt, ##__VA_ARGS__)
82 #endif
83
84 static void ext4_xattr_cache_insert(struct mb_cache *, struct buffer_head *);
85 static struct buffer_head *ext4_xattr_cache_find(struct inode *,
86 struct ext4_xattr_header *,
87 struct mb_cache_entry **);
88 static void ext4_xattr_rehash(struct ext4_xattr_header *,
89 struct ext4_xattr_entry *);
90 static int ext4_xattr_list(struct dentry *dentry, char *buffer,
91 size_t buffer_size);
92
93 static const struct xattr_handler *ext4_xattr_handler_map[] = {
94 [EXT4_XATTR_INDEX_USER] = &ext4_xattr_user_handler,
95 #ifdef CONFIG_EXT4_FS_POSIX_ACL
96 [EXT4_XATTR_INDEX_POSIX_ACL_ACCESS] = &posix_acl_access_xattr_handler,
97 [EXT4_XATTR_INDEX_POSIX_ACL_DEFAULT] = &posix_acl_default_xattr_handler,
98 #endif
99 [EXT4_XATTR_INDEX_TRUSTED] = &ext4_xattr_trusted_handler,
100 #ifdef CONFIG_EXT4_FS_SECURITY
101 [EXT4_XATTR_INDEX_SECURITY] = &ext4_xattr_security_handler,
102 #endif
103 };
104
105 const struct xattr_handler *ext4_xattr_handlers[] = {
106 &ext4_xattr_user_handler,
107 &ext4_xattr_trusted_handler,
108 #ifdef CONFIG_EXT4_FS_POSIX_ACL
109 &posix_acl_access_xattr_handler,
110 &posix_acl_default_xattr_handler,
111 #endif
112 #ifdef CONFIG_EXT4_FS_SECURITY
113 &ext4_xattr_security_handler,
114 #endif
115 NULL
116 };
117
118 #define EXT4_GET_MB_CACHE(inode) (((struct ext4_sb_info *) \
119 inode->i_sb->s_fs_info)->s_mb_cache)
120
121 static __le32 ext4_xattr_block_csum(struct inode *inode,
122 sector_t block_nr,
123 struct ext4_xattr_header *hdr)
124 {
125 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
126 __u32 csum;
127 __le32 save_csum;
128 __le64 dsk_block_nr = cpu_to_le64(block_nr);
129
130 save_csum = hdr->h_checksum;
131 hdr->h_checksum = 0;
132 csum = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&dsk_block_nr,
133 sizeof(dsk_block_nr));
134 csum = ext4_chksum(sbi, csum, (__u8 *)hdr,
135 EXT4_BLOCK_SIZE(inode->i_sb));
136
137 hdr->h_checksum = save_csum;
138 return cpu_to_le32(csum);
139 }
140
141 static int ext4_xattr_block_csum_verify(struct inode *inode,
142 sector_t block_nr,
143 struct ext4_xattr_header *hdr)
144 {
145 if (EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
146 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
147 (hdr->h_checksum != ext4_xattr_block_csum(inode, block_nr, hdr)))
148 return 0;
149 return 1;
150 }
151
152 static void ext4_xattr_block_csum_set(struct inode *inode,
153 sector_t block_nr,
154 struct ext4_xattr_header *hdr)
155 {
156 if (!EXT4_HAS_RO_COMPAT_FEATURE(inode->i_sb,
157 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
158 return;
159
160 hdr->h_checksum = ext4_xattr_block_csum(inode, block_nr, hdr);
161 }
162
163 static inline int ext4_handle_dirty_xattr_block(handle_t *handle,
164 struct inode *inode,
165 struct buffer_head *bh)
166 {
167 ext4_xattr_block_csum_set(inode, bh->b_blocknr, BHDR(bh));
168 return ext4_handle_dirty_metadata(handle, inode, bh);
169 }
170
171 static inline const struct xattr_handler *
172 ext4_xattr_handler(int name_index)
173 {
174 const struct xattr_handler *handler = NULL;
175
176 if (name_index > 0 && name_index < ARRAY_SIZE(ext4_xattr_handler_map))
177 handler = ext4_xattr_handler_map[name_index];
178 return handler;
179 }
180
181 /*
182 * Inode operation listxattr()
183 *
184 * dentry->d_inode->i_mutex: don't care
185 */
186 ssize_t
187 ext4_listxattr(struct dentry *dentry, char *buffer, size_t size)
188 {
189 return ext4_xattr_list(dentry, buffer, size);
190 }
191
192 static int
193 ext4_xattr_check_names(struct ext4_xattr_entry *entry, void *end)
194 {
195 while (!IS_LAST_ENTRY(entry)) {
196 struct ext4_xattr_entry *next = EXT4_XATTR_NEXT(entry);
197 if ((void *)next >= end)
198 return -EIO;
199 entry = next;
200 }
201 return 0;
202 }
203
204 static inline int
205 ext4_xattr_check_block(struct inode *inode, struct buffer_head *bh)
206 {
207 int error;
208
209 if (buffer_verified(bh))
210 return 0;
211
212 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
213 BHDR(bh)->h_blocks != cpu_to_le32(1))
214 return -EIO;
215 if (!ext4_xattr_block_csum_verify(inode, bh->b_blocknr, BHDR(bh)))
216 return -EIO;
217 error = ext4_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
218 if (!error)
219 set_buffer_verified(bh);
220 return error;
221 }
222
223 static inline int
224 ext4_xattr_check_entry(struct ext4_xattr_entry *entry, size_t size)
225 {
226 size_t value_size = le32_to_cpu(entry->e_value_size);
227
228 if (entry->e_value_block != 0 || value_size > size ||
229 le16_to_cpu(entry->e_value_offs) + value_size > size)
230 return -EIO;
231 return 0;
232 }
233
234 static int
235 ext4_xattr_find_entry(struct ext4_xattr_entry **pentry, int name_index,
236 const char *name, size_t size, int sorted)
237 {
238 struct ext4_xattr_entry *entry;
239 size_t name_len;
240 int cmp = 1;
241
242 if (name == NULL)
243 return -EINVAL;
244 name_len = strlen(name);
245 entry = *pentry;
246 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
247 cmp = name_index - entry->e_name_index;
248 if (!cmp)
249 cmp = name_len - entry->e_name_len;
250 if (!cmp)
251 cmp = memcmp(name, entry->e_name, name_len);
252 if (cmp <= 0 && (sorted || cmp == 0))
253 break;
254 }
255 *pentry = entry;
256 if (!cmp && ext4_xattr_check_entry(entry, size))
257 return -EIO;
258 return cmp ? -ENODATA : 0;
259 }
260
261 static int
262 ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
263 void *buffer, size_t buffer_size)
264 {
265 struct buffer_head *bh = NULL;
266 struct ext4_xattr_entry *entry;
267 size_t size;
268 int error;
269 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
270
271 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
272 name_index, name, buffer, (long)buffer_size);
273
274 error = -ENODATA;
275 if (!EXT4_I(inode)->i_file_acl)
276 goto cleanup;
277 ea_idebug(inode, "reading block %llu",
278 (unsigned long long)EXT4_I(inode)->i_file_acl);
279 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
280 if (!bh)
281 goto cleanup;
282 ea_bdebug(bh, "b_count=%d, refcount=%d",
283 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
284 if (ext4_xattr_check_block(inode, bh)) {
285 bad_block:
286 EXT4_ERROR_INODE(inode, "bad block %llu",
287 EXT4_I(inode)->i_file_acl);
288 error = -EIO;
289 goto cleanup;
290 }
291 ext4_xattr_cache_insert(ext4_mb_cache, bh);
292 entry = BFIRST(bh);
293 error = ext4_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
294 if (error == -EIO)
295 goto bad_block;
296 if (error)
297 goto cleanup;
298 size = le32_to_cpu(entry->e_value_size);
299 if (buffer) {
300 error = -ERANGE;
301 if (size > buffer_size)
302 goto cleanup;
303 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
304 size);
305 }
306 error = size;
307
308 cleanup:
309 brelse(bh);
310 return error;
311 }
312
313 int
314 ext4_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
315 void *buffer, size_t buffer_size)
316 {
317 struct ext4_xattr_ibody_header *header;
318 struct ext4_xattr_entry *entry;
319 struct ext4_inode *raw_inode;
320 struct ext4_iloc iloc;
321 size_t size;
322 void *end;
323 int error;
324
325 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
326 return -ENODATA;
327 error = ext4_get_inode_loc(inode, &iloc);
328 if (error)
329 return error;
330 raw_inode = ext4_raw_inode(&iloc);
331 header = IHDR(inode, raw_inode);
332 entry = IFIRST(header);
333 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
334 error = ext4_xattr_check_names(entry, end);
335 if (error)
336 goto cleanup;
337 error = ext4_xattr_find_entry(&entry, name_index, name,
338 end - (void *)entry, 0);
339 if (error)
340 goto cleanup;
341 size = le32_to_cpu(entry->e_value_size);
342 if (buffer) {
343 error = -ERANGE;
344 if (size > buffer_size)
345 goto cleanup;
346 memcpy(buffer, (void *)IFIRST(header) +
347 le16_to_cpu(entry->e_value_offs), size);
348 }
349 error = size;
350
351 cleanup:
352 brelse(iloc.bh);
353 return error;
354 }
355
356 /*
357 * ext4_xattr_get()
358 *
359 * Copy an extended attribute into the buffer
360 * provided, or compute the buffer size required.
361 * Buffer is NULL to compute the size of the buffer required.
362 *
363 * Returns a negative error number on failure, or the number of bytes
364 * used / required on success.
365 */
366 int
367 ext4_xattr_get(struct inode *inode, int name_index, const char *name,
368 void *buffer, size_t buffer_size)
369 {
370 int error;
371
372 if (strlen(name) > 255)
373 return -ERANGE;
374
375 down_read(&EXT4_I(inode)->xattr_sem);
376 error = ext4_xattr_ibody_get(inode, name_index, name, buffer,
377 buffer_size);
378 if (error == -ENODATA)
379 error = ext4_xattr_block_get(inode, name_index, name, buffer,
380 buffer_size);
381 up_read(&EXT4_I(inode)->xattr_sem);
382 return error;
383 }
384
385 static int
386 ext4_xattr_list_entries(struct dentry *dentry, struct ext4_xattr_entry *entry,
387 char *buffer, size_t buffer_size)
388 {
389 size_t rest = buffer_size;
390
391 for (; !IS_LAST_ENTRY(entry); entry = EXT4_XATTR_NEXT(entry)) {
392 const struct xattr_handler *handler =
393 ext4_xattr_handler(entry->e_name_index);
394
395 if (handler) {
396 size_t size = handler->list(dentry, buffer, rest,
397 entry->e_name,
398 entry->e_name_len,
399 handler->flags);
400 if (buffer) {
401 if (size > rest)
402 return -ERANGE;
403 buffer += size;
404 }
405 rest -= size;
406 }
407 }
408 return buffer_size - rest;
409 }
410
411 static int
412 ext4_xattr_block_list(struct dentry *dentry, char *buffer, size_t buffer_size)
413 {
414 struct inode *inode = dentry->d_inode;
415 struct buffer_head *bh = NULL;
416 int error;
417 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
418
419 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
420 buffer, (long)buffer_size);
421
422 error = 0;
423 if (!EXT4_I(inode)->i_file_acl)
424 goto cleanup;
425 ea_idebug(inode, "reading block %llu",
426 (unsigned long long)EXT4_I(inode)->i_file_acl);
427 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
428 error = -EIO;
429 if (!bh)
430 goto cleanup;
431 ea_bdebug(bh, "b_count=%d, refcount=%d",
432 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
433 if (ext4_xattr_check_block(inode, bh)) {
434 EXT4_ERROR_INODE(inode, "bad block %llu",
435 EXT4_I(inode)->i_file_acl);
436 error = -EIO;
437 goto cleanup;
438 }
439 ext4_xattr_cache_insert(ext4_mb_cache, bh);
440 error = ext4_xattr_list_entries(dentry, BFIRST(bh), buffer, buffer_size);
441
442 cleanup:
443 brelse(bh);
444
445 return error;
446 }
447
448 static int
449 ext4_xattr_ibody_list(struct dentry *dentry, char *buffer, size_t buffer_size)
450 {
451 struct inode *inode = dentry->d_inode;
452 struct ext4_xattr_ibody_header *header;
453 struct ext4_inode *raw_inode;
454 struct ext4_iloc iloc;
455 void *end;
456 int error;
457
458 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR))
459 return 0;
460 error = ext4_get_inode_loc(inode, &iloc);
461 if (error)
462 return error;
463 raw_inode = ext4_raw_inode(&iloc);
464 header = IHDR(inode, raw_inode);
465 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
466 error = ext4_xattr_check_names(IFIRST(header), end);
467 if (error)
468 goto cleanup;
469 error = ext4_xattr_list_entries(dentry, IFIRST(header),
470 buffer, buffer_size);
471
472 cleanup:
473 brelse(iloc.bh);
474 return error;
475 }
476
477 /*
478 * ext4_xattr_list()
479 *
480 * Copy a list of attribute names into the buffer
481 * provided, or compute the buffer size required.
482 * Buffer is NULL to compute the size of the buffer required.
483 *
484 * Returns a negative error number on failure, or the number of bytes
485 * used / required on success.
486 */
487 static int
488 ext4_xattr_list(struct dentry *dentry, char *buffer, size_t buffer_size)
489 {
490 int ret, ret2;
491
492 down_read(&EXT4_I(dentry->d_inode)->xattr_sem);
493 ret = ret2 = ext4_xattr_ibody_list(dentry, buffer, buffer_size);
494 if (ret < 0)
495 goto errout;
496 if (buffer) {
497 buffer += ret;
498 buffer_size -= ret;
499 }
500 ret = ext4_xattr_block_list(dentry, buffer, buffer_size);
501 if (ret < 0)
502 goto errout;
503 ret += ret2;
504 errout:
505 up_read(&EXT4_I(dentry->d_inode)->xattr_sem);
506 return ret;
507 }
508
509 /*
510 * If the EXT4_FEATURE_COMPAT_EXT_ATTR feature of this file system is
511 * not set, set it.
512 */
513 static void ext4_xattr_update_super_block(handle_t *handle,
514 struct super_block *sb)
515 {
516 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR))
517 return;
518
519 BUFFER_TRACE(EXT4_SB(sb)->s_sbh, "get_write_access");
520 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
521 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
522 ext4_handle_dirty_super(handle, sb);
523 }
524 }
525
526 /*
527 * Release the xattr block BH: If the reference count is > 1, decrement it;
528 * otherwise free the block.
529 */
530 static void
531 ext4_xattr_release_block(handle_t *handle, struct inode *inode,
532 struct buffer_head *bh)
533 {
534 struct mb_cache_entry *ce = NULL;
535 int error = 0;
536 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
537
538 ce = mb_cache_entry_get(ext4_mb_cache, bh->b_bdev, bh->b_blocknr);
539 BUFFER_TRACE(bh, "get_write_access");
540 error = ext4_journal_get_write_access(handle, bh);
541 if (error)
542 goto out;
543
544 lock_buffer(bh);
545 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
546 ea_bdebug(bh, "refcount now=0; freeing");
547 if (ce)
548 mb_cache_entry_free(ce);
549 get_bh(bh);
550 unlock_buffer(bh);
551 ext4_free_blocks(handle, inode, bh, 0, 1,
552 EXT4_FREE_BLOCKS_METADATA |
553 EXT4_FREE_BLOCKS_FORGET);
554 } else {
555 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
556 if (ce)
557 mb_cache_entry_release(ce);
558 /*
559 * Beware of this ugliness: Releasing of xattr block references
560 * from different inodes can race and so we have to protect
561 * from a race where someone else frees the block (and releases
562 * its journal_head) before we are done dirtying the buffer. In
563 * nojournal mode this race is harmless and we actually cannot
564 * call ext4_handle_dirty_xattr_block() with locked buffer as
565 * that function can call sync_dirty_buffer() so for that case
566 * we handle the dirtying after unlocking the buffer.
567 */
568 if (ext4_handle_valid(handle))
569 error = ext4_handle_dirty_xattr_block(handle, inode,
570 bh);
571 unlock_buffer(bh);
572 if (!ext4_handle_valid(handle))
573 error = ext4_handle_dirty_xattr_block(handle, inode,
574 bh);
575 if (IS_SYNC(inode))
576 ext4_handle_sync(handle);
577 dquot_free_block(inode, EXT4_C2B(EXT4_SB(inode->i_sb), 1));
578 ea_bdebug(bh, "refcount now=%d; releasing",
579 le32_to_cpu(BHDR(bh)->h_refcount));
580 }
581 out:
582 ext4_std_error(inode->i_sb, error);
583 return;
584 }
585
586 /*
587 * Find the available free space for EAs. This also returns the total number of
588 * bytes used by EA entries.
589 */
590 static size_t ext4_xattr_free_space(struct ext4_xattr_entry *last,
591 size_t *min_offs, void *base, int *total)
592 {
593 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
594 if (!last->e_value_block && last->e_value_size) {
595 size_t offs = le16_to_cpu(last->e_value_offs);
596 if (offs < *min_offs)
597 *min_offs = offs;
598 }
599 if (total)
600 *total += EXT4_XATTR_LEN(last->e_name_len);
601 }
602 return (*min_offs - ((void *)last - base) - sizeof(__u32));
603 }
604
605 static int
606 ext4_xattr_set_entry(struct ext4_xattr_info *i, struct ext4_xattr_search *s)
607 {
608 struct ext4_xattr_entry *last;
609 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
610
611 /* Compute min_offs and last. */
612 last = s->first;
613 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
614 if (!last->e_value_block && last->e_value_size) {
615 size_t offs = le16_to_cpu(last->e_value_offs);
616 if (offs < min_offs)
617 min_offs = offs;
618 }
619 }
620 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
621 if (!s->not_found) {
622 if (!s->here->e_value_block && s->here->e_value_size) {
623 size_t size = le32_to_cpu(s->here->e_value_size);
624 free += EXT4_XATTR_SIZE(size);
625 }
626 free += EXT4_XATTR_LEN(name_len);
627 }
628 if (i->value) {
629 if (free < EXT4_XATTR_SIZE(i->value_len) ||
630 free < EXT4_XATTR_LEN(name_len) +
631 EXT4_XATTR_SIZE(i->value_len))
632 return -ENOSPC;
633 }
634
635 if (i->value && s->not_found) {
636 /* Insert the new name. */
637 size_t size = EXT4_XATTR_LEN(name_len);
638 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
639 memmove((void *)s->here + size, s->here, rest);
640 memset(s->here, 0, size);
641 s->here->e_name_index = i->name_index;
642 s->here->e_name_len = name_len;
643 memcpy(s->here->e_name, i->name, name_len);
644 } else {
645 if (!s->here->e_value_block && s->here->e_value_size) {
646 void *first_val = s->base + min_offs;
647 size_t offs = le16_to_cpu(s->here->e_value_offs);
648 void *val = s->base + offs;
649 size_t size = EXT4_XATTR_SIZE(
650 le32_to_cpu(s->here->e_value_size));
651
652 if (i->value && size == EXT4_XATTR_SIZE(i->value_len)) {
653 /* The old and the new value have the same
654 size. Just replace. */
655 s->here->e_value_size =
656 cpu_to_le32(i->value_len);
657 if (i->value == EXT4_ZERO_XATTR_VALUE) {
658 memset(val, 0, size);
659 } else {
660 /* Clear pad bytes first. */
661 memset(val + size - EXT4_XATTR_PAD, 0,
662 EXT4_XATTR_PAD);
663 memcpy(val, i->value, i->value_len);
664 }
665 return 0;
666 }
667
668 /* Remove the old value. */
669 memmove(first_val + size, first_val, val - first_val);
670 memset(first_val, 0, size);
671 s->here->e_value_size = 0;
672 s->here->e_value_offs = 0;
673 min_offs += size;
674
675 /* Adjust all value offsets. */
676 last = s->first;
677 while (!IS_LAST_ENTRY(last)) {
678 size_t o = le16_to_cpu(last->e_value_offs);
679 if (!last->e_value_block &&
680 last->e_value_size && o < offs)
681 last->e_value_offs =
682 cpu_to_le16(o + size);
683 last = EXT4_XATTR_NEXT(last);
684 }
685 }
686 if (!i->value) {
687 /* Remove the old name. */
688 size_t size = EXT4_XATTR_LEN(name_len);
689 last = ENTRY((void *)last - size);
690 memmove(s->here, (void *)s->here + size,
691 (void *)last - (void *)s->here + sizeof(__u32));
692 memset(last, 0, size);
693 }
694 }
695
696 if (i->value) {
697 /* Insert the new value. */
698 s->here->e_value_size = cpu_to_le32(i->value_len);
699 if (i->value_len) {
700 size_t size = EXT4_XATTR_SIZE(i->value_len);
701 void *val = s->base + min_offs - size;
702 s->here->e_value_offs = cpu_to_le16(min_offs - size);
703 if (i->value == EXT4_ZERO_XATTR_VALUE) {
704 memset(val, 0, size);
705 } else {
706 /* Clear the pad bytes first. */
707 memset(val + size - EXT4_XATTR_PAD, 0,
708 EXT4_XATTR_PAD);
709 memcpy(val, i->value, i->value_len);
710 }
711 }
712 }
713 return 0;
714 }
715
716 struct ext4_xattr_block_find {
717 struct ext4_xattr_search s;
718 struct buffer_head *bh;
719 };
720
721 static int
722 ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
723 struct ext4_xattr_block_find *bs)
724 {
725 struct super_block *sb = inode->i_sb;
726 int error;
727
728 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
729 i->name_index, i->name, i->value, (long)i->value_len);
730
731 if (EXT4_I(inode)->i_file_acl) {
732 /* The inode already has an extended attribute block. */
733 bs->bh = sb_bread(sb, EXT4_I(inode)->i_file_acl);
734 error = -EIO;
735 if (!bs->bh)
736 goto cleanup;
737 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
738 atomic_read(&(bs->bh->b_count)),
739 le32_to_cpu(BHDR(bs->bh)->h_refcount));
740 if (ext4_xattr_check_block(inode, bs->bh)) {
741 EXT4_ERROR_INODE(inode, "bad block %llu",
742 EXT4_I(inode)->i_file_acl);
743 error = -EIO;
744 goto cleanup;
745 }
746 /* Find the named attribute. */
747 bs->s.base = BHDR(bs->bh);
748 bs->s.first = BFIRST(bs->bh);
749 bs->s.end = bs->bh->b_data + bs->bh->b_size;
750 bs->s.here = bs->s.first;
751 error = ext4_xattr_find_entry(&bs->s.here, i->name_index,
752 i->name, bs->bh->b_size, 1);
753 if (error && error != -ENODATA)
754 goto cleanup;
755 bs->s.not_found = error;
756 }
757 error = 0;
758
759 cleanup:
760 return error;
761 }
762
763 static int
764 ext4_xattr_block_set(handle_t *handle, struct inode *inode,
765 struct ext4_xattr_info *i,
766 struct ext4_xattr_block_find *bs)
767 {
768 struct super_block *sb = inode->i_sb;
769 struct buffer_head *new_bh = NULL;
770 struct ext4_xattr_search *s = &bs->s;
771 struct mb_cache_entry *ce = NULL;
772 int error = 0;
773 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
774
775 #define header(x) ((struct ext4_xattr_header *)(x))
776
777 if (i->value && i->value_len > sb->s_blocksize)
778 return -ENOSPC;
779 if (s->base) {
780 ce = mb_cache_entry_get(ext4_mb_cache, bs->bh->b_bdev,
781 bs->bh->b_blocknr);
782 BUFFER_TRACE(bs->bh, "get_write_access");
783 error = ext4_journal_get_write_access(handle, bs->bh);
784 if (error)
785 goto cleanup;
786 lock_buffer(bs->bh);
787
788 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
789 if (ce) {
790 mb_cache_entry_free(ce);
791 ce = NULL;
792 }
793 ea_bdebug(bs->bh, "modifying in-place");
794 error = ext4_xattr_set_entry(i, s);
795 if (!error) {
796 if (!IS_LAST_ENTRY(s->first))
797 ext4_xattr_rehash(header(s->base),
798 s->here);
799 ext4_xattr_cache_insert(ext4_mb_cache,
800 bs->bh);
801 }
802 unlock_buffer(bs->bh);
803 if (error == -EIO)
804 goto bad_block;
805 if (!error)
806 error = ext4_handle_dirty_xattr_block(handle,
807 inode,
808 bs->bh);
809 if (error)
810 goto cleanup;
811 goto inserted;
812 } else {
813 int offset = (char *)s->here - bs->bh->b_data;
814
815 unlock_buffer(bs->bh);
816 if (ce) {
817 mb_cache_entry_release(ce);
818 ce = NULL;
819 }
820 ea_bdebug(bs->bh, "cloning");
821 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
822 error = -ENOMEM;
823 if (s->base == NULL)
824 goto cleanup;
825 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
826 s->first = ENTRY(header(s->base)+1);
827 header(s->base)->h_refcount = cpu_to_le32(1);
828 s->here = ENTRY(s->base + offset);
829 s->end = s->base + bs->bh->b_size;
830 }
831 } else {
832 /* Allocate a buffer where we construct the new block. */
833 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
834 /* assert(header == s->base) */
835 error = -ENOMEM;
836 if (s->base == NULL)
837 goto cleanup;
838 header(s->base)->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
839 header(s->base)->h_blocks = cpu_to_le32(1);
840 header(s->base)->h_refcount = cpu_to_le32(1);
841 s->first = ENTRY(header(s->base)+1);
842 s->here = ENTRY(header(s->base)+1);
843 s->end = s->base + sb->s_blocksize;
844 }
845
846 error = ext4_xattr_set_entry(i, s);
847 if (error == -EIO)
848 goto bad_block;
849 if (error)
850 goto cleanup;
851 if (!IS_LAST_ENTRY(s->first))
852 ext4_xattr_rehash(header(s->base), s->here);
853
854 inserted:
855 if (!IS_LAST_ENTRY(s->first)) {
856 new_bh = ext4_xattr_cache_find(inode, header(s->base), &ce);
857 if (new_bh) {
858 /* We found an identical block in the cache. */
859 if (new_bh == bs->bh)
860 ea_bdebug(new_bh, "keeping");
861 else {
862 /* The old block is released after updating
863 the inode. */
864 error = dquot_alloc_block(inode,
865 EXT4_C2B(EXT4_SB(sb), 1));
866 if (error)
867 goto cleanup;
868 BUFFER_TRACE(new_bh, "get_write_access");
869 error = ext4_journal_get_write_access(handle,
870 new_bh);
871 if (error)
872 goto cleanup_dquot;
873 lock_buffer(new_bh);
874 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
875 ea_bdebug(new_bh, "reusing; refcount now=%d",
876 le32_to_cpu(BHDR(new_bh)->h_refcount));
877 unlock_buffer(new_bh);
878 error = ext4_handle_dirty_xattr_block(handle,
879 inode,
880 new_bh);
881 if (error)
882 goto cleanup_dquot;
883 }
884 mb_cache_entry_release(ce);
885 ce = NULL;
886 } else if (bs->bh && s->base == bs->bh->b_data) {
887 /* We were modifying this block in-place. */
888 ea_bdebug(bs->bh, "keeping this block");
889 new_bh = bs->bh;
890 get_bh(new_bh);
891 } else {
892 /* We need to allocate a new block */
893 ext4_fsblk_t goal, block;
894
895 goal = ext4_group_first_block_no(sb,
896 EXT4_I(inode)->i_block_group);
897
898 /* non-extent files can't have physical blocks past 2^32 */
899 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
900 goal = goal & EXT4_MAX_BLOCK_FILE_PHYS;
901
902 /*
903 * take i_data_sem because we will test
904 * i_delalloc_reserved_flag in ext4_mb_new_blocks
905 */
906 down_read(&EXT4_I(inode)->i_data_sem);
907 block = ext4_new_meta_blocks(handle, inode, goal, 0,
908 NULL, &error);
909 up_read((&EXT4_I(inode)->i_data_sem));
910 if (error)
911 goto cleanup;
912
913 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
914 BUG_ON(block > EXT4_MAX_BLOCK_FILE_PHYS);
915
916 ea_idebug(inode, "creating block %llu",
917 (unsigned long long)block);
918
919 new_bh = sb_getblk(sb, block);
920 if (unlikely(!new_bh)) {
921 error = -ENOMEM;
922 getblk_failed:
923 ext4_free_blocks(handle, inode, NULL, block, 1,
924 EXT4_FREE_BLOCKS_METADATA);
925 goto cleanup;
926 }
927 lock_buffer(new_bh);
928 error = ext4_journal_get_create_access(handle, new_bh);
929 if (error) {
930 unlock_buffer(new_bh);
931 error = -EIO;
932 goto getblk_failed;
933 }
934 memcpy(new_bh->b_data, s->base, new_bh->b_size);
935 set_buffer_uptodate(new_bh);
936 unlock_buffer(new_bh);
937 ext4_xattr_cache_insert(ext4_mb_cache, new_bh);
938 error = ext4_handle_dirty_xattr_block(handle,
939 inode, new_bh);
940 if (error)
941 goto cleanup;
942 }
943 }
944
945 /* Update the inode. */
946 EXT4_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
947
948 /* Drop the previous xattr block. */
949 if (bs->bh && bs->bh != new_bh)
950 ext4_xattr_release_block(handle, inode, bs->bh);
951 error = 0;
952
953 cleanup:
954 if (ce)
955 mb_cache_entry_release(ce);
956 brelse(new_bh);
957 if (!(bs->bh && s->base == bs->bh->b_data))
958 kfree(s->base);
959
960 return error;
961
962 cleanup_dquot:
963 dquot_free_block(inode, EXT4_C2B(EXT4_SB(sb), 1));
964 goto cleanup;
965
966 bad_block:
967 EXT4_ERROR_INODE(inode, "bad block %llu",
968 EXT4_I(inode)->i_file_acl);
969 goto cleanup;
970
971 #undef header
972 }
973
974 int ext4_xattr_ibody_find(struct inode *inode, struct ext4_xattr_info *i,
975 struct ext4_xattr_ibody_find *is)
976 {
977 struct ext4_xattr_ibody_header *header;
978 struct ext4_inode *raw_inode;
979 int error;
980
981 if (EXT4_I(inode)->i_extra_isize == 0)
982 return 0;
983 raw_inode = ext4_raw_inode(&is->iloc);
984 header = IHDR(inode, raw_inode);
985 is->s.base = is->s.first = IFIRST(header);
986 is->s.here = is->s.first;
987 is->s.end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
988 if (ext4_test_inode_state(inode, EXT4_STATE_XATTR)) {
989 error = ext4_xattr_check_names(IFIRST(header), is->s.end);
990 if (error)
991 return error;
992 /* Find the named attribute. */
993 error = ext4_xattr_find_entry(&is->s.here, i->name_index,
994 i->name, is->s.end -
995 (void *)is->s.base, 0);
996 if (error && error != -ENODATA)
997 return error;
998 is->s.not_found = error;
999 }
1000 return 0;
1001 }
1002
1003 int ext4_xattr_ibody_inline_set(handle_t *handle, struct inode *inode,
1004 struct ext4_xattr_info *i,
1005 struct ext4_xattr_ibody_find *is)
1006 {
1007 struct ext4_xattr_ibody_header *header;
1008 struct ext4_xattr_search *s = &is->s;
1009 int error;
1010
1011 if (EXT4_I(inode)->i_extra_isize == 0)
1012 return -ENOSPC;
1013 error = ext4_xattr_set_entry(i, s);
1014 if (error) {
1015 if (error == -ENOSPC &&
1016 ext4_has_inline_data(inode)) {
1017 error = ext4_try_to_evict_inline_data(handle, inode,
1018 EXT4_XATTR_LEN(strlen(i->name) +
1019 EXT4_XATTR_SIZE(i->value_len)));
1020 if (error)
1021 return error;
1022 error = ext4_xattr_ibody_find(inode, i, is);
1023 if (error)
1024 return error;
1025 error = ext4_xattr_set_entry(i, s);
1026 }
1027 if (error)
1028 return error;
1029 }
1030 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1031 if (!IS_LAST_ENTRY(s->first)) {
1032 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1033 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1034 } else {
1035 header->h_magic = cpu_to_le32(0);
1036 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1037 }
1038 return 0;
1039 }
1040
1041 static int ext4_xattr_ibody_set(handle_t *handle, struct inode *inode,
1042 struct ext4_xattr_info *i,
1043 struct ext4_xattr_ibody_find *is)
1044 {
1045 struct ext4_xattr_ibody_header *header;
1046 struct ext4_xattr_search *s = &is->s;
1047 int error;
1048
1049 if (EXT4_I(inode)->i_extra_isize == 0)
1050 return -ENOSPC;
1051 error = ext4_xattr_set_entry(i, s);
1052 if (error)
1053 return error;
1054 header = IHDR(inode, ext4_raw_inode(&is->iloc));
1055 if (!IS_LAST_ENTRY(s->first)) {
1056 header->h_magic = cpu_to_le32(EXT4_XATTR_MAGIC);
1057 ext4_set_inode_state(inode, EXT4_STATE_XATTR);
1058 } else {
1059 header->h_magic = cpu_to_le32(0);
1060 ext4_clear_inode_state(inode, EXT4_STATE_XATTR);
1061 }
1062 return 0;
1063 }
1064
1065 /*
1066 * ext4_xattr_set_handle()
1067 *
1068 * Create, replace or remove an extended attribute for this inode. Value
1069 * is NULL to remove an existing extended attribute, and non-NULL to
1070 * either replace an existing extended attribute, or create a new extended
1071 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
1072 * specify that an extended attribute must exist and must not exist
1073 * previous to the call, respectively.
1074 *
1075 * Returns 0, or a negative error number on failure.
1076 */
1077 int
1078 ext4_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
1079 const char *name, const void *value, size_t value_len,
1080 int flags)
1081 {
1082 struct ext4_xattr_info i = {
1083 .name_index = name_index,
1084 .name = name,
1085 .value = value,
1086 .value_len = value_len,
1087
1088 };
1089 struct ext4_xattr_ibody_find is = {
1090 .s = { .not_found = -ENODATA, },
1091 };
1092 struct ext4_xattr_block_find bs = {
1093 .s = { .not_found = -ENODATA, },
1094 };
1095 unsigned long no_expand;
1096 int error;
1097
1098 if (!name)
1099 return -EINVAL;
1100 if (strlen(name) > 255)
1101 return -ERANGE;
1102 down_write(&EXT4_I(inode)->xattr_sem);
1103 no_expand = ext4_test_inode_state(inode, EXT4_STATE_NO_EXPAND);
1104 ext4_set_inode_state(inode, EXT4_STATE_NO_EXPAND);
1105
1106 error = ext4_reserve_inode_write(handle, inode, &is.iloc);
1107 if (error)
1108 goto cleanup;
1109
1110 if (ext4_test_inode_state(inode, EXT4_STATE_NEW)) {
1111 struct ext4_inode *raw_inode = ext4_raw_inode(&is.iloc);
1112 memset(raw_inode, 0, EXT4_SB(inode->i_sb)->s_inode_size);
1113 ext4_clear_inode_state(inode, EXT4_STATE_NEW);
1114 }
1115
1116 error = ext4_xattr_ibody_find(inode, &i, &is);
1117 if (error)
1118 goto cleanup;
1119 if (is.s.not_found)
1120 error = ext4_xattr_block_find(inode, &i, &bs);
1121 if (error)
1122 goto cleanup;
1123 if (is.s.not_found && bs.s.not_found) {
1124 error = -ENODATA;
1125 if (flags & XATTR_REPLACE)
1126 goto cleanup;
1127 error = 0;
1128 if (!value)
1129 goto cleanup;
1130 } else {
1131 error = -EEXIST;
1132 if (flags & XATTR_CREATE)
1133 goto cleanup;
1134 }
1135 if (!value) {
1136 if (!is.s.not_found)
1137 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1138 else if (!bs.s.not_found)
1139 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1140 } else {
1141 error = ext4_xattr_ibody_set(handle, inode, &i, &is);
1142 if (!error && !bs.s.not_found) {
1143 i.value = NULL;
1144 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1145 } else if (error == -ENOSPC) {
1146 if (EXT4_I(inode)->i_file_acl && !bs.s.base) {
1147 error = ext4_xattr_block_find(inode, &i, &bs);
1148 if (error)
1149 goto cleanup;
1150 }
1151 error = ext4_xattr_block_set(handle, inode, &i, &bs);
1152 if (error)
1153 goto cleanup;
1154 if (!is.s.not_found) {
1155 i.value = NULL;
1156 error = ext4_xattr_ibody_set(handle, inode, &i,
1157 &is);
1158 }
1159 }
1160 }
1161 if (!error) {
1162 ext4_xattr_update_super_block(handle, inode->i_sb);
1163 inode->i_ctime = ext4_current_time(inode);
1164 if (!value)
1165 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1166 error = ext4_mark_iloc_dirty(handle, inode, &is.iloc);
1167 /*
1168 * The bh is consumed by ext4_mark_iloc_dirty, even with
1169 * error != 0.
1170 */
1171 is.iloc.bh = NULL;
1172 if (IS_SYNC(inode))
1173 ext4_handle_sync(handle);
1174 }
1175
1176 cleanup:
1177 brelse(is.iloc.bh);
1178 brelse(bs.bh);
1179 if (no_expand == 0)
1180 ext4_clear_inode_state(inode, EXT4_STATE_NO_EXPAND);
1181 up_write(&EXT4_I(inode)->xattr_sem);
1182 return error;
1183 }
1184
1185 /*
1186 * ext4_xattr_set()
1187 *
1188 * Like ext4_xattr_set_handle, but start from an inode. This extended
1189 * attribute modification is a filesystem transaction by itself.
1190 *
1191 * Returns 0, or a negative error number on failure.
1192 */
1193 int
1194 ext4_xattr_set(struct inode *inode, int name_index, const char *name,
1195 const void *value, size_t value_len, int flags)
1196 {
1197 handle_t *handle;
1198 int error, retries = 0;
1199 int credits = ext4_jbd2_credits_xattr(inode);
1200
1201 retry:
1202 handle = ext4_journal_start(inode, EXT4_HT_XATTR, credits);
1203 if (IS_ERR(handle)) {
1204 error = PTR_ERR(handle);
1205 } else {
1206 int error2;
1207
1208 error = ext4_xattr_set_handle(handle, inode, name_index, name,
1209 value, value_len, flags);
1210 error2 = ext4_journal_stop(handle);
1211 if (error == -ENOSPC &&
1212 ext4_should_retry_alloc(inode->i_sb, &retries))
1213 goto retry;
1214 if (error == 0)
1215 error = error2;
1216 }
1217
1218 return error;
1219 }
1220
1221 /*
1222 * Shift the EA entries in the inode to create space for the increased
1223 * i_extra_isize.
1224 */
1225 static void ext4_xattr_shift_entries(struct ext4_xattr_entry *entry,
1226 int value_offs_shift, void *to,
1227 void *from, size_t n, int blocksize)
1228 {
1229 struct ext4_xattr_entry *last = entry;
1230 int new_offs;
1231
1232 /* Adjust the value offsets of the entries */
1233 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1234 if (!last->e_value_block && last->e_value_size) {
1235 new_offs = le16_to_cpu(last->e_value_offs) +
1236 value_offs_shift;
1237 BUG_ON(new_offs + le32_to_cpu(last->e_value_size)
1238 > blocksize);
1239 last->e_value_offs = cpu_to_le16(new_offs);
1240 }
1241 }
1242 /* Shift the entries by n bytes */
1243 memmove(to, from, n);
1244 }
1245
1246 /*
1247 * Expand an inode by new_extra_isize bytes when EAs are present.
1248 * Returns 0 on success or negative error number on failure.
1249 */
1250 int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
1251 struct ext4_inode *raw_inode, handle_t *handle)
1252 {
1253 struct ext4_xattr_ibody_header *header;
1254 struct ext4_xattr_entry *entry, *last, *first;
1255 struct buffer_head *bh = NULL;
1256 struct ext4_xattr_ibody_find *is = NULL;
1257 struct ext4_xattr_block_find *bs = NULL;
1258 char *buffer = NULL, *b_entry_name = NULL;
1259 size_t min_offs, free;
1260 int total_ino;
1261 void *base, *start, *end;
1262 int extra_isize = 0, error = 0, tried_min_extra_isize = 0;
1263 int s_min_extra_isize = le16_to_cpu(EXT4_SB(inode->i_sb)->s_es->s_min_extra_isize);
1264
1265 down_write(&EXT4_I(inode)->xattr_sem);
1266 retry:
1267 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) {
1268 up_write(&EXT4_I(inode)->xattr_sem);
1269 return 0;
1270 }
1271
1272 header = IHDR(inode, raw_inode);
1273 entry = IFIRST(header);
1274
1275 /*
1276 * Check if enough free space is available in the inode to shift the
1277 * entries ahead by new_extra_isize.
1278 */
1279
1280 base = start = entry;
1281 end = (void *)raw_inode + EXT4_SB(inode->i_sb)->s_inode_size;
1282 min_offs = end - base;
1283 last = entry;
1284 total_ino = sizeof(struct ext4_xattr_ibody_header);
1285
1286 free = ext4_xattr_free_space(last, &min_offs, base, &total_ino);
1287 if (free >= new_extra_isize) {
1288 entry = IFIRST(header);
1289 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize
1290 - new_extra_isize, (void *)raw_inode +
1291 EXT4_GOOD_OLD_INODE_SIZE + new_extra_isize,
1292 (void *)header, total_ino,
1293 inode->i_sb->s_blocksize);
1294 EXT4_I(inode)->i_extra_isize = new_extra_isize;
1295 error = 0;
1296 goto cleanup;
1297 }
1298
1299 /*
1300 * Enough free space isn't available in the inode, check if
1301 * EA block can hold new_extra_isize bytes.
1302 */
1303 if (EXT4_I(inode)->i_file_acl) {
1304 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1305 error = -EIO;
1306 if (!bh)
1307 goto cleanup;
1308 if (ext4_xattr_check_block(inode, bh)) {
1309 EXT4_ERROR_INODE(inode, "bad block %llu",
1310 EXT4_I(inode)->i_file_acl);
1311 error = -EIO;
1312 goto cleanup;
1313 }
1314 base = BHDR(bh);
1315 first = BFIRST(bh);
1316 end = bh->b_data + bh->b_size;
1317 min_offs = end - base;
1318 free = ext4_xattr_free_space(first, &min_offs, base, NULL);
1319 if (free < new_extra_isize) {
1320 if (!tried_min_extra_isize && s_min_extra_isize) {
1321 tried_min_extra_isize++;
1322 new_extra_isize = s_min_extra_isize;
1323 brelse(bh);
1324 goto retry;
1325 }
1326 error = -1;
1327 goto cleanup;
1328 }
1329 } else {
1330 free = inode->i_sb->s_blocksize;
1331 }
1332
1333 while (new_extra_isize > 0) {
1334 size_t offs, size, entry_size;
1335 struct ext4_xattr_entry *small_entry = NULL;
1336 struct ext4_xattr_info i = {
1337 .value = NULL,
1338 .value_len = 0,
1339 };
1340 unsigned int total_size; /* EA entry size + value size */
1341 unsigned int shift_bytes; /* No. of bytes to shift EAs by? */
1342 unsigned int min_total_size = ~0U;
1343
1344 is = kzalloc(sizeof(struct ext4_xattr_ibody_find), GFP_NOFS);
1345 bs = kzalloc(sizeof(struct ext4_xattr_block_find), GFP_NOFS);
1346 if (!is || !bs) {
1347 error = -ENOMEM;
1348 goto cleanup;
1349 }
1350
1351 is->s.not_found = -ENODATA;
1352 bs->s.not_found = -ENODATA;
1353 is->iloc.bh = NULL;
1354 bs->bh = NULL;
1355
1356 last = IFIRST(header);
1357 /* Find the entry best suited to be pushed into EA block */
1358 entry = NULL;
1359 for (; !IS_LAST_ENTRY(last); last = EXT4_XATTR_NEXT(last)) {
1360 total_size =
1361 EXT4_XATTR_SIZE(le32_to_cpu(last->e_value_size)) +
1362 EXT4_XATTR_LEN(last->e_name_len);
1363 if (total_size <= free && total_size < min_total_size) {
1364 if (total_size < new_extra_isize) {
1365 small_entry = last;
1366 } else {
1367 entry = last;
1368 min_total_size = total_size;
1369 }
1370 }
1371 }
1372
1373 if (entry == NULL) {
1374 if (small_entry) {
1375 entry = small_entry;
1376 } else {
1377 if (!tried_min_extra_isize &&
1378 s_min_extra_isize) {
1379 tried_min_extra_isize++;
1380 new_extra_isize = s_min_extra_isize;
1381 kfree(is); is = NULL;
1382 kfree(bs); bs = NULL;
1383 brelse(bh);
1384 goto retry;
1385 }
1386 error = -1;
1387 goto cleanup;
1388 }
1389 }
1390 offs = le16_to_cpu(entry->e_value_offs);
1391 size = le32_to_cpu(entry->e_value_size);
1392 entry_size = EXT4_XATTR_LEN(entry->e_name_len);
1393 i.name_index = entry->e_name_index,
1394 buffer = kmalloc(EXT4_XATTR_SIZE(size), GFP_NOFS);
1395 b_entry_name = kmalloc(entry->e_name_len + 1, GFP_NOFS);
1396 if (!buffer || !b_entry_name) {
1397 error = -ENOMEM;
1398 goto cleanup;
1399 }
1400 /* Save the entry name and the entry value */
1401 memcpy(buffer, (void *)IFIRST(header) + offs,
1402 EXT4_XATTR_SIZE(size));
1403 memcpy(b_entry_name, entry->e_name, entry->e_name_len);
1404 b_entry_name[entry->e_name_len] = '\0';
1405 i.name = b_entry_name;
1406
1407 error = ext4_get_inode_loc(inode, &is->iloc);
1408 if (error)
1409 goto cleanup;
1410
1411 error = ext4_xattr_ibody_find(inode, &i, is);
1412 if (error)
1413 goto cleanup;
1414
1415 /* Remove the chosen entry from the inode */
1416 error = ext4_xattr_ibody_set(handle, inode, &i, is);
1417 if (error)
1418 goto cleanup;
1419
1420 entry = IFIRST(header);
1421 if (entry_size + EXT4_XATTR_SIZE(size) >= new_extra_isize)
1422 shift_bytes = new_extra_isize;
1423 else
1424 shift_bytes = entry_size + size;
1425 /* Adjust the offsets and shift the remaining entries ahead */
1426 ext4_xattr_shift_entries(entry, EXT4_I(inode)->i_extra_isize -
1427 shift_bytes, (void *)raw_inode +
1428 EXT4_GOOD_OLD_INODE_SIZE + extra_isize + shift_bytes,
1429 (void *)header, total_ino - entry_size,
1430 inode->i_sb->s_blocksize);
1431
1432 extra_isize += shift_bytes;
1433 new_extra_isize -= shift_bytes;
1434 EXT4_I(inode)->i_extra_isize = extra_isize;
1435
1436 i.name = b_entry_name;
1437 i.value = buffer;
1438 i.value_len = size;
1439 error = ext4_xattr_block_find(inode, &i, bs);
1440 if (error)
1441 goto cleanup;
1442
1443 /* Add entry which was removed from the inode into the block */
1444 error = ext4_xattr_block_set(handle, inode, &i, bs);
1445 if (error)
1446 goto cleanup;
1447 kfree(b_entry_name);
1448 kfree(buffer);
1449 b_entry_name = NULL;
1450 buffer = NULL;
1451 brelse(is->iloc.bh);
1452 kfree(is);
1453 kfree(bs);
1454 }
1455 brelse(bh);
1456 up_write(&EXT4_I(inode)->xattr_sem);
1457 return 0;
1458
1459 cleanup:
1460 kfree(b_entry_name);
1461 kfree(buffer);
1462 if (is)
1463 brelse(is->iloc.bh);
1464 kfree(is);
1465 kfree(bs);
1466 brelse(bh);
1467 up_write(&EXT4_I(inode)->xattr_sem);
1468 return error;
1469 }
1470
1471
1472
1473 /*
1474 * ext4_xattr_delete_inode()
1475 *
1476 * Free extended attribute resources associated with this inode. This
1477 * is called immediately before an inode is freed. We have exclusive
1478 * access to the inode.
1479 */
1480 void
1481 ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1482 {
1483 struct buffer_head *bh = NULL;
1484
1485 if (!EXT4_I(inode)->i_file_acl)
1486 goto cleanup;
1487 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1488 if (!bh) {
1489 EXT4_ERROR_INODE(inode, "block %llu read error",
1490 EXT4_I(inode)->i_file_acl);
1491 goto cleanup;
1492 }
1493 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1494 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1495 EXT4_ERROR_INODE(inode, "bad block %llu",
1496 EXT4_I(inode)->i_file_acl);
1497 goto cleanup;
1498 }
1499 ext4_xattr_release_block(handle, inode, bh);
1500 EXT4_I(inode)->i_file_acl = 0;
1501
1502 cleanup:
1503 brelse(bh);
1504 }
1505
1506 /*
1507 * ext4_xattr_put_super()
1508 *
1509 * This is called when a file system is unmounted.
1510 */
1511 void
1512 ext4_xattr_put_super(struct super_block *sb)
1513 {
1514 mb_cache_shrink(sb->s_bdev);
1515 }
1516
1517 /*
1518 * ext4_xattr_cache_insert()
1519 *
1520 * Create a new entry in the extended attribute cache, and insert
1521 * it unless such an entry is already in the cache.
1522 *
1523 * Returns 0, or a negative error number on failure.
1524 */
1525 static void
1526 ext4_xattr_cache_insert(struct mb_cache *ext4_mb_cache, struct buffer_head *bh)
1527 {
1528 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1529 struct mb_cache_entry *ce;
1530 int error;
1531
1532 ce = mb_cache_entry_alloc(ext4_mb_cache, GFP_NOFS);
1533 if (!ce) {
1534 ea_bdebug(bh, "out of memory");
1535 return;
1536 }
1537 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1538 if (error) {
1539 mb_cache_entry_free(ce);
1540 if (error == -EBUSY) {
1541 ea_bdebug(bh, "already in cache");
1542 error = 0;
1543 }
1544 } else {
1545 ea_bdebug(bh, "inserting [%x]", (int)hash);
1546 mb_cache_entry_release(ce);
1547 }
1548 }
1549
1550 /*
1551 * ext4_xattr_cmp()
1552 *
1553 * Compare two extended attribute blocks for equality.
1554 *
1555 * Returns 0 if the blocks are equal, 1 if they differ, and
1556 * a negative error number on errors.
1557 */
1558 static int
1559 ext4_xattr_cmp(struct ext4_xattr_header *header1,
1560 struct ext4_xattr_header *header2)
1561 {
1562 struct ext4_xattr_entry *entry1, *entry2;
1563
1564 entry1 = ENTRY(header1+1);
1565 entry2 = ENTRY(header2+1);
1566 while (!IS_LAST_ENTRY(entry1)) {
1567 if (IS_LAST_ENTRY(entry2))
1568 return 1;
1569 if (entry1->e_hash != entry2->e_hash ||
1570 entry1->e_name_index != entry2->e_name_index ||
1571 entry1->e_name_len != entry2->e_name_len ||
1572 entry1->e_value_size != entry2->e_value_size ||
1573 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1574 return 1;
1575 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1576 return -EIO;
1577 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1578 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1579 le32_to_cpu(entry1->e_value_size)))
1580 return 1;
1581
1582 entry1 = EXT4_XATTR_NEXT(entry1);
1583 entry2 = EXT4_XATTR_NEXT(entry2);
1584 }
1585 if (!IS_LAST_ENTRY(entry2))
1586 return 1;
1587 return 0;
1588 }
1589
1590 /*
1591 * ext4_xattr_cache_find()
1592 *
1593 * Find an identical extended attribute block.
1594 *
1595 * Returns a pointer to the block found, or NULL if such a block was
1596 * not found or an error occurred.
1597 */
1598 static struct buffer_head *
1599 ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1600 struct mb_cache_entry **pce)
1601 {
1602 __u32 hash = le32_to_cpu(header->h_hash);
1603 struct mb_cache_entry *ce;
1604 struct mb_cache *ext4_mb_cache = EXT4_GET_MB_CACHE(inode);
1605
1606 if (!header->h_hash)
1607 return NULL; /* never share */
1608 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1609 again:
1610 ce = mb_cache_entry_find_first(ext4_mb_cache, inode->i_sb->s_bdev,
1611 hash);
1612 while (ce) {
1613 struct buffer_head *bh;
1614
1615 if (IS_ERR(ce)) {
1616 if (PTR_ERR(ce) == -EAGAIN)
1617 goto again;
1618 break;
1619 }
1620 bh = sb_bread(inode->i_sb, ce->e_block);
1621 if (!bh) {
1622 EXT4_ERROR_INODE(inode, "block %lu read error",
1623 (unsigned long) ce->e_block);
1624 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1625 EXT4_XATTR_REFCOUNT_MAX) {
1626 ea_idebug(inode, "block %lu refcount %d>=%d",
1627 (unsigned long) ce->e_block,
1628 le32_to_cpu(BHDR(bh)->h_refcount),
1629 EXT4_XATTR_REFCOUNT_MAX);
1630 } else if (ext4_xattr_cmp(header, BHDR(bh)) == 0) {
1631 *pce = ce;
1632 return bh;
1633 }
1634 brelse(bh);
1635 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1636 }
1637 return NULL;
1638 }
1639
1640 #define NAME_HASH_SHIFT 5
1641 #define VALUE_HASH_SHIFT 16
1642
1643 /*
1644 * ext4_xattr_hash_entry()
1645 *
1646 * Compute the hash of an extended attribute.
1647 */
1648 static inline void ext4_xattr_hash_entry(struct ext4_xattr_header *header,
1649 struct ext4_xattr_entry *entry)
1650 {
1651 __u32 hash = 0;
1652 char *name = entry->e_name;
1653 int n;
1654
1655 for (n = 0; n < entry->e_name_len; n++) {
1656 hash = (hash << NAME_HASH_SHIFT) ^
1657 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1658 *name++;
1659 }
1660
1661 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1662 __le32 *value = (__le32 *)((char *)header +
1663 le16_to_cpu(entry->e_value_offs));
1664 for (n = (le32_to_cpu(entry->e_value_size) +
1665 EXT4_XATTR_ROUND) >> EXT4_XATTR_PAD_BITS; n; n--) {
1666 hash = (hash << VALUE_HASH_SHIFT) ^
1667 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1668 le32_to_cpu(*value++);
1669 }
1670 }
1671 entry->e_hash = cpu_to_le32(hash);
1672 }
1673
1674 #undef NAME_HASH_SHIFT
1675 #undef VALUE_HASH_SHIFT
1676
1677 #define BLOCK_HASH_SHIFT 16
1678
1679 /*
1680 * ext4_xattr_rehash()
1681 *
1682 * Re-compute the extended attribute hash value after an entry has changed.
1683 */
1684 static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1685 struct ext4_xattr_entry *entry)
1686 {
1687 struct ext4_xattr_entry *here;
1688 __u32 hash = 0;
1689
1690 ext4_xattr_hash_entry(header, entry);
1691 here = ENTRY(header+1);
1692 while (!IS_LAST_ENTRY(here)) {
1693 if (!here->e_hash) {
1694 /* Block is not shared if an entry's hash value == 0 */
1695 hash = 0;
1696 break;
1697 }
1698 hash = (hash << BLOCK_HASH_SHIFT) ^
1699 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1700 le32_to_cpu(here->e_hash);
1701 here = EXT4_XATTR_NEXT(here);
1702 }
1703 header->h_hash = cpu_to_le32(hash);
1704 }
1705
1706 #undef BLOCK_HASH_SHIFT
1707
1708 #define HASH_BUCKET_BITS 10
1709
1710 struct mb_cache *
1711 ext4_xattr_create_cache(char *name)
1712 {
1713 return mb_cache_create(name, HASH_BUCKET_BITS);
1714 }
1715
1716 void ext4_xattr_destroy_cache(struct mb_cache *cache)
1717 {
1718 if (cache)
1719 mb_cache_destroy(cache);
1720 }
1721