blob: 4cbc6d0212d34406ac69ee67bac392bd7a312fea [file] [log] [blame]
Linus Torvalds1da177e2005-04-16 15:20:36 -07001/*
2 * linux/fs/ext3/xattr.c
3 *
4 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
5 *
6 * Fix by Harrison Xing <harrison@mountainviewdata.com>.
7 * Ext3 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 * EXT3_I(inode)->i_file_acl is protected by EXT3_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/ext3_jbd.h>
57#include <linux/ext3_fs.h>
58#include <linux/mbcache.h>
59#include <linux/quotaops.h>
60#include <linux/rwsem.h>
61#include "xattr.h"
62#include "acl.h"
63
64#define BHDR(bh) ((struct ext3_xattr_header *)((bh)->b_data))
65#define ENTRY(ptr) ((struct ext3_xattr_entry *)(ptr))
66#define BFIRST(bh) ENTRY(BHDR(bh)+1)
67#define IS_LAST_ENTRY(entry) (*(__u32 *)(entry) == 0)
68
69#define IHDR(inode, raw_inode) \
70 ((struct ext3_xattr_ibody_header *) \
71 ((void *)raw_inode + \
72 EXT3_GOOD_OLD_INODE_SIZE + \
73 EXT3_I(inode)->i_extra_isize))
74#define IFIRST(hdr) ((struct ext3_xattr_entry *)((hdr)+1))
75
76#ifdef EXT3_XATTR_DEBUG
77# define ea_idebug(inode, f...) do { \
78 printk(KERN_DEBUG "inode %s:%ld: ", \
79 inode->i_sb->s_id, inode->i_ino); \
80 printk(f); \
81 printk("\n"); \
82 } while (0)
83# define ea_bdebug(bh, f...) do { \
84 char b[BDEVNAME_SIZE]; \
85 printk(KERN_DEBUG "block %s:%lu: ", \
86 bdevname(bh->b_bdev, b), \
87 (unsigned long) bh->b_blocknr); \
88 printk(f); \
89 printk("\n"); \
90 } while (0)
91#else
92# define ea_idebug(f...)
93# define ea_bdebug(f...)
94#endif
95
96static void ext3_xattr_cache_insert(struct buffer_head *);
97static struct buffer_head *ext3_xattr_cache_find(struct inode *,
98 struct ext3_xattr_header *,
99 struct mb_cache_entry **);
100static void ext3_xattr_rehash(struct ext3_xattr_header *,
101 struct ext3_xattr_entry *);
102
103static struct mb_cache *ext3_xattr_cache;
104
105static struct xattr_handler *ext3_xattr_handler_map[] = {
106 [EXT3_XATTR_INDEX_USER] = &ext3_xattr_user_handler,
107#ifdef CONFIG_EXT3_FS_POSIX_ACL
108 [EXT3_XATTR_INDEX_POSIX_ACL_ACCESS] = &ext3_xattr_acl_access_handler,
109 [EXT3_XATTR_INDEX_POSIX_ACL_DEFAULT] = &ext3_xattr_acl_default_handler,
110#endif
111 [EXT3_XATTR_INDEX_TRUSTED] = &ext3_xattr_trusted_handler,
112#ifdef CONFIG_EXT3_FS_SECURITY
113 [EXT3_XATTR_INDEX_SECURITY] = &ext3_xattr_security_handler,
114#endif
115};
116
117struct xattr_handler *ext3_xattr_handlers[] = {
118 &ext3_xattr_user_handler,
119 &ext3_xattr_trusted_handler,
120#ifdef CONFIG_EXT3_FS_POSIX_ACL
121 &ext3_xattr_acl_access_handler,
122 &ext3_xattr_acl_default_handler,
123#endif
124#ifdef CONFIG_EXT3_FS_SECURITY
125 &ext3_xattr_security_handler,
126#endif
127 NULL
128};
129
130static inline struct xattr_handler *
131ext3_xattr_handler(int name_index)
132{
133 struct xattr_handler *handler = NULL;
134
135 if (name_index > 0 && name_index < ARRAY_SIZE(ext3_xattr_handler_map))
136 handler = ext3_xattr_handler_map[name_index];
137 return handler;
138}
139
140/*
141 * Inode operation listxattr()
142 *
143 * dentry->d_inode->i_sem: don't care
144 */
145ssize_t
146ext3_listxattr(struct dentry *dentry, char *buffer, size_t size)
147{
148 return ext3_xattr_list(dentry->d_inode, buffer, size);
149}
150
151static int
152ext3_xattr_check_names(struct ext3_xattr_entry *entry, void *end)
153{
154 while (!IS_LAST_ENTRY(entry)) {
155 struct ext3_xattr_entry *next = EXT3_XATTR_NEXT(entry);
156 if ((void *)next >= end)
157 return -EIO;
158 entry = next;
159 }
160 return 0;
161}
162
163static inline int
164ext3_xattr_check_block(struct buffer_head *bh)
165{
166 int error;
167
168 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
169 BHDR(bh)->h_blocks != cpu_to_le32(1))
170 return -EIO;
171 error = ext3_xattr_check_names(BFIRST(bh), bh->b_data + bh->b_size);
172 return error;
173}
174
175static inline int
176ext3_xattr_check_entry(struct ext3_xattr_entry *entry, size_t size)
177{
178 size_t value_size = le32_to_cpu(entry->e_value_size);
179
180 if (entry->e_value_block != 0 || value_size > size ||
181 le16_to_cpu(entry->e_value_offs) + value_size > size)
182 return -EIO;
183 return 0;
184}
185
186static int
187ext3_xattr_find_entry(struct ext3_xattr_entry **pentry, int name_index,
188 const char *name, size_t size, int sorted)
189{
190 struct ext3_xattr_entry *entry;
191 size_t name_len;
192 int cmp = 1;
193
194 if (name == NULL)
195 return -EINVAL;
196 name_len = strlen(name);
197 entry = *pentry;
198 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
199 cmp = name_index - entry->e_name_index;
200 if (!cmp)
201 cmp = name_len - entry->e_name_len;
202 if (!cmp)
203 cmp = memcmp(name, entry->e_name, name_len);
204 if (cmp <= 0 && (sorted || cmp == 0))
205 break;
206 }
207 *pentry = entry;
208 if (!cmp && ext3_xattr_check_entry(entry, size))
209 return -EIO;
210 return cmp ? -ENODATA : 0;
211}
212
213int
214ext3_xattr_block_get(struct inode *inode, int name_index, const char *name,
215 void *buffer, size_t buffer_size)
216{
217 struct buffer_head *bh = NULL;
218 struct ext3_xattr_entry *entry;
219 size_t size;
220 int error;
221
222 ea_idebug(inode, "name=%d.%s, buffer=%p, buffer_size=%ld",
223 name_index, name, buffer, (long)buffer_size);
224
225 error = -ENODATA;
226 if (!EXT3_I(inode)->i_file_acl)
227 goto cleanup;
228 ea_idebug(inode, "reading block %d", EXT3_I(inode)->i_file_acl);
229 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
230 if (!bh)
231 goto cleanup;
232 ea_bdebug(bh, "b_count=%d, refcount=%d",
233 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
234 if (ext3_xattr_check_block(bh)) {
235bad_block: ext3_error(inode->i_sb, __FUNCTION__,
236 "inode %ld: bad block %d", inode->i_ino,
237 EXT3_I(inode)->i_file_acl);
238 error = -EIO;
239 goto cleanup;
240 }
241 ext3_xattr_cache_insert(bh);
242 entry = BFIRST(bh);
243 error = ext3_xattr_find_entry(&entry, name_index, name, bh->b_size, 1);
244 if (error == -EIO)
245 goto bad_block;
246 if (error)
247 goto cleanup;
248 size = le32_to_cpu(entry->e_value_size);
249 if (buffer) {
250 error = -ERANGE;
251 if (size > buffer_size)
252 goto cleanup;
253 memcpy(buffer, bh->b_data + le16_to_cpu(entry->e_value_offs),
254 size);
255 }
256 error = size;
257
258cleanup:
259 brelse(bh);
260 return error;
261}
262
263static int
264ext3_xattr_ibody_get(struct inode *inode, int name_index, const char *name,
265 void *buffer, size_t buffer_size)
266{
267 struct ext3_xattr_ibody_header *header;
268 struct ext3_xattr_entry *entry;
269 struct ext3_inode *raw_inode;
270 struct ext3_iloc iloc;
271 size_t size;
272 void *end;
273 int error;
274
275 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
276 return -ENODATA;
277 error = ext3_get_inode_loc(inode, &iloc);
278 if (error)
279 return error;
280 raw_inode = ext3_raw_inode(&iloc);
281 header = IHDR(inode, raw_inode);
282 entry = IFIRST(header);
283 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
284 error = ext3_xattr_check_names(entry, end);
285 if (error)
286 goto cleanup;
287 error = ext3_xattr_find_entry(&entry, name_index, name,
288 end - (void *)entry, 0);
289 if (error)
290 goto cleanup;
291 size = le32_to_cpu(entry->e_value_size);
292 if (buffer) {
293 error = -ERANGE;
294 if (size > buffer_size)
295 goto cleanup;
296 memcpy(buffer, (void *)IFIRST(header) +
297 le16_to_cpu(entry->e_value_offs), size);
298 }
299 error = size;
300
301cleanup:
302 brelse(iloc.bh);
303 return error;
304}
305
306/*
307 * ext3_xattr_get()
308 *
309 * Copy an extended attribute into the buffer
310 * provided, or compute the buffer size required.
311 * Buffer is NULL to compute the size of the buffer required.
312 *
313 * Returns a negative error number on failure, or the number of bytes
314 * used / required on success.
315 */
316int
317ext3_xattr_get(struct inode *inode, int name_index, const char *name,
318 void *buffer, size_t buffer_size)
319{
320 int error;
321
322 down_read(&EXT3_I(inode)->xattr_sem);
323 error = ext3_xattr_ibody_get(inode, name_index, name, buffer,
324 buffer_size);
325 if (error == -ENODATA)
326 error = ext3_xattr_block_get(inode, name_index, name, buffer,
327 buffer_size);
328 up_read(&EXT3_I(inode)->xattr_sem);
329 return error;
330}
331
332static int
333ext3_xattr_list_entries(struct inode *inode, struct ext3_xattr_entry *entry,
334 char *buffer, size_t buffer_size)
335{
336 size_t rest = buffer_size;
337
338 for (; !IS_LAST_ENTRY(entry); entry = EXT3_XATTR_NEXT(entry)) {
339 struct xattr_handler *handler =
340 ext3_xattr_handler(entry->e_name_index);
341
342 if (handler) {
343 size_t size = handler->list(inode, buffer, rest,
344 entry->e_name,
345 entry->e_name_len);
346 if (buffer) {
347 if (size > rest)
348 return -ERANGE;
349 buffer += size;
350 }
351 rest -= size;
352 }
353 }
354 return buffer_size - rest;
355}
356
357int
358ext3_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
359{
360 struct buffer_head *bh = NULL;
361 int error;
362
363 ea_idebug(inode, "buffer=%p, buffer_size=%ld",
364 buffer, (long)buffer_size);
365
366 error = 0;
367 if (!EXT3_I(inode)->i_file_acl)
368 goto cleanup;
369 ea_idebug(inode, "reading block %d", EXT3_I(inode)->i_file_acl);
370 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
371 error = -EIO;
372 if (!bh)
373 goto cleanup;
374 ea_bdebug(bh, "b_count=%d, refcount=%d",
375 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
376 if (ext3_xattr_check_block(bh)) {
377 ext3_error(inode->i_sb, __FUNCTION__,
378 "inode %ld: bad block %d", inode->i_ino,
379 EXT3_I(inode)->i_file_acl);
380 error = -EIO;
381 goto cleanup;
382 }
383 ext3_xattr_cache_insert(bh);
384 error = ext3_xattr_list_entries(inode, BFIRST(bh), buffer, buffer_size);
385
386cleanup:
387 brelse(bh);
388
389 return error;
390}
391
392static int
393ext3_xattr_ibody_list(struct inode *inode, char *buffer, size_t buffer_size)
394{
395 struct ext3_xattr_ibody_header *header;
396 struct ext3_inode *raw_inode;
397 struct ext3_iloc iloc;
398 void *end;
399 int error;
400
401 if (!(EXT3_I(inode)->i_state & EXT3_STATE_XATTR))
402 return 0;
403 error = ext3_get_inode_loc(inode, &iloc);
404 if (error)
405 return error;
406 raw_inode = ext3_raw_inode(&iloc);
407 header = IHDR(inode, raw_inode);
408 end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
409 error = ext3_xattr_check_names(IFIRST(header), end);
410 if (error)
411 goto cleanup;
412 error = ext3_xattr_list_entries(inode, IFIRST(header),
413 buffer, buffer_size);
414
415cleanup:
416 brelse(iloc.bh);
417 return error;
418}
419
420/*
421 * ext3_xattr_list()
422 *
423 * Copy a list of attribute names into the buffer
424 * provided, or compute the buffer size required.
425 * Buffer is NULL to compute the size of the buffer required.
426 *
427 * Returns a negative error number on failure, or the number of bytes
428 * used / required on success.
429 */
430int
431ext3_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
432{
433 int i_error, b_error;
434
435 down_read(&EXT3_I(inode)->xattr_sem);
436 i_error = ext3_xattr_ibody_list(inode, buffer, buffer_size);
437 if (i_error < 0) {
438 b_error = 0;
439 } else {
440 if (buffer) {
441 buffer += i_error;
442 buffer_size -= i_error;
443 }
444 b_error = ext3_xattr_block_list(inode, buffer, buffer_size);
445 if (b_error < 0)
446 i_error = 0;
447 }
448 up_read(&EXT3_I(inode)->xattr_sem);
449 return i_error + b_error;
450}
451
452/*
453 * If the EXT3_FEATURE_COMPAT_EXT_ATTR feature of this file system is
454 * not set, set it.
455 */
456static void ext3_xattr_update_super_block(handle_t *handle,
457 struct super_block *sb)
458{
459 if (EXT3_HAS_COMPAT_FEATURE(sb, EXT3_FEATURE_COMPAT_EXT_ATTR))
460 return;
461
462 lock_super(sb);
463 if (ext3_journal_get_write_access(handle, EXT3_SB(sb)->s_sbh) == 0) {
464 EXT3_SB(sb)->s_es->s_feature_compat |=
465 cpu_to_le32(EXT3_FEATURE_COMPAT_EXT_ATTR);
466 sb->s_dirt = 1;
467 ext3_journal_dirty_metadata(handle, EXT3_SB(sb)->s_sbh);
468 }
469 unlock_super(sb);
470}
471
472/*
473 * Release the xattr block BH: If the reference count is > 1, decrement
474 * it; otherwise free the block.
475 */
476static void
477ext3_xattr_release_block(handle_t *handle, struct inode *inode,
478 struct buffer_head *bh)
479{
480 struct mb_cache_entry *ce = NULL;
481
482 ce = mb_cache_entry_get(ext3_xattr_cache, bh->b_bdev, bh->b_blocknr);
483 if (BHDR(bh)->h_refcount == cpu_to_le32(1)) {
484 ea_bdebug(bh, "refcount now=0; freeing");
485 if (ce)
486 mb_cache_entry_free(ce);
487 ext3_free_blocks(handle, inode, bh->b_blocknr, 1);
488 get_bh(bh);
489 ext3_forget(handle, 1, inode, bh, bh->b_blocknr);
490 } else {
491 if (ext3_journal_get_write_access(handle, bh) == 0) {
492 lock_buffer(bh);
493 BHDR(bh)->h_refcount = cpu_to_le32(
494 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
495 ext3_journal_dirty_metadata(handle, bh);
496 if (IS_SYNC(inode))
497 handle->h_sync = 1;
498 DQUOT_FREE_BLOCK(inode, 1);
499 unlock_buffer(bh);
500 ea_bdebug(bh, "refcount now=%d; releasing",
501 le32_to_cpu(BHDR(bh)->h_refcount));
502 }
503 if (ce)
504 mb_cache_entry_release(ce);
505 }
506}
507
508struct ext3_xattr_info {
509 int name_index;
510 const char *name;
511 const void *value;
512 size_t value_len;
513};
514
515struct ext3_xattr_search {
516 struct ext3_xattr_entry *first;
517 void *base;
518 void *end;
519 struct ext3_xattr_entry *here;
520 int not_found;
521};
522
523static int
524ext3_xattr_set_entry(struct ext3_xattr_info *i, struct ext3_xattr_search *s)
525{
526 struct ext3_xattr_entry *last;
527 size_t free, min_offs = s->end - s->base, name_len = strlen(i->name);
528
529 /* Compute min_offs and last. */
530 last = s->first;
531 for (; !IS_LAST_ENTRY(last); last = EXT3_XATTR_NEXT(last)) {
532 if (!last->e_value_block && last->e_value_size) {
533 size_t offs = le16_to_cpu(last->e_value_offs);
534 if (offs < min_offs)
535 min_offs = offs;
536 }
537 }
538 free = min_offs - ((void *)last - s->base) - sizeof(__u32);
539 if (!s->not_found) {
540 if (!s->here->e_value_block && s->here->e_value_size) {
541 size_t size = le32_to_cpu(s->here->e_value_size);
542 free += EXT3_XATTR_SIZE(size);
543 }
544 free += EXT3_XATTR_LEN(name_len);
545 }
546 if (i->value) {
547 if (free < EXT3_XATTR_SIZE(i->value_len) ||
548 free < EXT3_XATTR_LEN(name_len) +
549 EXT3_XATTR_SIZE(i->value_len))
550 return -ENOSPC;
551 }
552
553 if (i->value && s->not_found) {
554 /* Insert the new name. */
555 size_t size = EXT3_XATTR_LEN(name_len);
556 size_t rest = (void *)last - (void *)s->here + sizeof(__u32);
557 memmove((void *)s->here + size, s->here, rest);
558 memset(s->here, 0, size);
559 s->here->e_name_index = i->name_index;
560 s->here->e_name_len = name_len;
561 memcpy(s->here->e_name, i->name, name_len);
562 } else {
563 if (!s->here->e_value_block && s->here->e_value_size) {
564 void *first_val = s->base + min_offs;
565 size_t offs = le16_to_cpu(s->here->e_value_offs);
566 void *val = s->base + offs;
567 size_t size = EXT3_XATTR_SIZE(
568 le32_to_cpu(s->here->e_value_size));
569
570 if (i->value && size == EXT3_XATTR_SIZE(i->value_len)) {
571 /* The old and the new value have the same
572 size. Just replace. */
573 s->here->e_value_size =
574 cpu_to_le32(i->value_len);
575 memset(val + size - EXT3_XATTR_PAD, 0,
576 EXT3_XATTR_PAD); /* Clear pad bytes. */
577 memcpy(val, i->value, i->value_len);
578 return 0;
579 }
580
581 /* Remove the old value. */
582 memmove(first_val + size, first_val, val - first_val);
583 memset(first_val, 0, size);
584 s->here->e_value_size = 0;
585 s->here->e_value_offs = 0;
586 min_offs += size;
587
588 /* Adjust all value offsets. */
589 last = s->first;
590 while (!IS_LAST_ENTRY(last)) {
591 size_t o = le16_to_cpu(last->e_value_offs);
592 if (!last->e_value_block &&
593 last->e_value_size && o < offs)
594 last->e_value_offs =
595 cpu_to_le16(o + size);
596 last = EXT3_XATTR_NEXT(last);
597 }
598 }
599 if (!i->value) {
600 /* Remove the old name. */
601 size_t size = EXT3_XATTR_LEN(name_len);
602 last = ENTRY((void *)last - size);
603 memmove(s->here, (void *)s->here + size,
604 (void *)last - (void *)s->here + sizeof(__u32));
605 memset(last, 0, size);
606 }
607 }
608
609 if (i->value) {
610 /* Insert the new value. */
611 s->here->e_value_size = cpu_to_le32(i->value_len);
612 if (i->value_len) {
613 size_t size = EXT3_XATTR_SIZE(i->value_len);
614 void *val = s->base + min_offs - size;
615 s->here->e_value_offs = cpu_to_le16(min_offs - size);
616 memset(val + size - EXT3_XATTR_PAD, 0,
617 EXT3_XATTR_PAD); /* Clear the pad bytes. */
618 memcpy(val, i->value, i->value_len);
619 }
620 }
621 return 0;
622}
623
624struct ext3_xattr_block_find {
625 struct ext3_xattr_search s;
626 struct buffer_head *bh;
627};
628
629int
630ext3_xattr_block_find(struct inode *inode, struct ext3_xattr_info *i,
631 struct ext3_xattr_block_find *bs)
632{
633 struct super_block *sb = inode->i_sb;
634 int error;
635
636 ea_idebug(inode, "name=%d.%s, value=%p, value_len=%ld",
637 i->name_index, i->name, i->value, (long)i->value_len);
638
639 if (EXT3_I(inode)->i_file_acl) {
640 /* The inode already has an extended attribute block. */
641 bs->bh = sb_bread(sb, EXT3_I(inode)->i_file_acl);
642 error = -EIO;
643 if (!bs->bh)
644 goto cleanup;
645 ea_bdebug(bs->bh, "b_count=%d, refcount=%d",
646 atomic_read(&(bs->bh->b_count)),
647 le32_to_cpu(BHDR(bs->bh)->h_refcount));
648 if (ext3_xattr_check_block(bs->bh)) {
649 ext3_error(sb, __FUNCTION__,
650 "inode %ld: bad block %d", inode->i_ino,
651 EXT3_I(inode)->i_file_acl);
652 error = -EIO;
653 goto cleanup;
654 }
655 /* Find the named attribute. */
656 bs->s.base = BHDR(bs->bh);
657 bs->s.first = BFIRST(bs->bh);
658 bs->s.end = bs->bh->b_data + bs->bh->b_size;
659 bs->s.here = bs->s.first;
660 error = ext3_xattr_find_entry(&bs->s.here, i->name_index,
661 i->name, bs->bh->b_size, 1);
662 if (error && error != -ENODATA)
663 goto cleanup;
664 bs->s.not_found = error;
665 }
666 error = 0;
667
668cleanup:
669 return error;
670}
671
672static int
673ext3_xattr_block_set(handle_t *handle, struct inode *inode,
674 struct ext3_xattr_info *i,
675 struct ext3_xattr_block_find *bs)
676{
677 struct super_block *sb = inode->i_sb;
678 struct buffer_head *new_bh = NULL;
679 struct ext3_xattr_search *s = &bs->s;
680 struct mb_cache_entry *ce = NULL;
681 int error;
682
683#define header(x) ((struct ext3_xattr_header *)(x))
684
685 if (i->value && i->value_len > sb->s_blocksize)
686 return -ENOSPC;
687 if (s->base) {
688 ce = mb_cache_entry_get(ext3_xattr_cache, bs->bh->b_bdev,
689 bs->bh->b_blocknr);
690 if (header(s->base)->h_refcount == cpu_to_le32(1)) {
691 if (ce) {
692 mb_cache_entry_free(ce);
693 ce = NULL;
694 }
695 ea_bdebug(bs->bh, "modifying in-place");
696 error = ext3_journal_get_write_access(handle, bs->bh);
697 if (error)
698 goto cleanup;
699 lock_buffer(bs->bh);
700 error = ext3_xattr_set_entry(i, s);
701 if (!error) {
702 if (!IS_LAST_ENTRY(s->first))
703 ext3_xattr_rehash(header(s->base),
704 s->here);
705 ext3_xattr_cache_insert(bs->bh);
706 }
707 unlock_buffer(bs->bh);
708 if (error == -EIO)
709 goto bad_block;
710 if (!error)
711 error = ext3_journal_dirty_metadata(handle,
712 bs->bh);
713 if (error)
714 goto cleanup;
715 goto inserted;
716 } else {
717 int offset = (char *)s->here - bs->bh->b_data;
718
719 if (ce) {
720 mb_cache_entry_release(ce);
721 ce = NULL;
722 }
723 ea_bdebug(bs->bh, "cloning");
724 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL);
725 error = -ENOMEM;
726 if (s->base == NULL)
727 goto cleanup;
728 memcpy(s->base, BHDR(bs->bh), bs->bh->b_size);
729 s->first = ENTRY(header(s->base)+1);
730 header(s->base)->h_refcount = cpu_to_le32(1);
731 s->here = ENTRY(s->base + offset);
732 s->end = s->base + bs->bh->b_size;
733 }
734 } else {
735 /* Allocate a buffer where we construct the new block. */
736 s->base = kmalloc(sb->s_blocksize, GFP_KERNEL);
737 /* assert(header == s->base) */
738 error = -ENOMEM;
739 if (s->base == NULL)
740 goto cleanup;
741 memset(s->base, 0, sb->s_blocksize);
742 header(s->base)->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
743 header(s->base)->h_blocks = cpu_to_le32(1);
744 header(s->base)->h_refcount = cpu_to_le32(1);
745 s->first = ENTRY(header(s->base)+1);
746 s->here = ENTRY(header(s->base)+1);
747 s->end = s->base + sb->s_blocksize;
748 }
749
750 error = ext3_xattr_set_entry(i, s);
751 if (error == -EIO)
752 goto bad_block;
753 if (error)
754 goto cleanup;
755 if (!IS_LAST_ENTRY(s->first))
756 ext3_xattr_rehash(header(s->base), s->here);
757
758inserted:
759 if (!IS_LAST_ENTRY(s->first)) {
760 new_bh = ext3_xattr_cache_find(inode, header(s->base), &ce);
761 if (new_bh) {
762 /* We found an identical block in the cache. */
763 if (new_bh == bs->bh)
764 ea_bdebug(new_bh, "keeping");
765 else {
766 /* The old block is released after updating
767 the inode. */
768 error = -EDQUOT;
769 if (DQUOT_ALLOC_BLOCK(inode, 1))
770 goto cleanup;
771 error = ext3_journal_get_write_access(handle,
772 new_bh);
773 if (error)
774 goto cleanup_dquot;
775 lock_buffer(new_bh);
776 BHDR(new_bh)->h_refcount = cpu_to_le32(1 +
777 le32_to_cpu(BHDR(new_bh)->h_refcount));
778 ea_bdebug(new_bh, "reusing; refcount now=%d",
779 le32_to_cpu(BHDR(new_bh)->h_refcount));
780 unlock_buffer(new_bh);
781 error = ext3_journal_dirty_metadata(handle,
782 new_bh);
783 if (error)
784 goto cleanup_dquot;
785 }
786 mb_cache_entry_release(ce);
787 ce = NULL;
788 } else if (bs->bh && s->base == bs->bh->b_data) {
789 /* We were modifying this block in-place. */
790 ea_bdebug(bs->bh, "keeping this block");
791 new_bh = bs->bh;
792 get_bh(new_bh);
793 } else {
794 /* We need to allocate a new block */
795 int goal = le32_to_cpu(
796 EXT3_SB(sb)->s_es->s_first_data_block) +
797 EXT3_I(inode)->i_block_group *
798 EXT3_BLOCKS_PER_GROUP(sb);
799 int block = ext3_new_block(handle, inode, goal, &error);
800 if (error)
801 goto cleanup;
802 ea_idebug(inode, "creating block %d", block);
803
804 new_bh = sb_getblk(sb, block);
805 if (!new_bh) {
806getblk_failed:
807 ext3_free_blocks(handle, inode, block, 1);
808 error = -EIO;
809 goto cleanup;
810 }
811 lock_buffer(new_bh);
812 error = ext3_journal_get_create_access(handle, new_bh);
813 if (error) {
814 unlock_buffer(new_bh);
815 goto getblk_failed;
816 }
817 memcpy(new_bh->b_data, s->base, new_bh->b_size);
818 set_buffer_uptodate(new_bh);
819 unlock_buffer(new_bh);
820 ext3_xattr_cache_insert(new_bh);
821 error = ext3_journal_dirty_metadata(handle, new_bh);
822 if (error)
823 goto cleanup;
824 }
825 }
826
827 /* Update the inode. */
828 EXT3_I(inode)->i_file_acl = new_bh ? new_bh->b_blocknr : 0;
829
830 /* Drop the previous xattr block. */
831 if (bs->bh && bs->bh != new_bh)
832 ext3_xattr_release_block(handle, inode, bs->bh);
833 error = 0;
834
835cleanup:
836 if (ce)
837 mb_cache_entry_release(ce);
838 brelse(new_bh);
839 if (!(bs->bh && s->base == bs->bh->b_data))
840 kfree(s->base);
841
842 return error;
843
844cleanup_dquot:
845 DQUOT_FREE_BLOCK(inode, 1);
846 goto cleanup;
847
848bad_block:
849 ext3_error(inode->i_sb, __FUNCTION__,
850 "inode %ld: bad block %d", inode->i_ino,
851 EXT3_I(inode)->i_file_acl);
852 goto cleanup;
853
854#undef header
855}
856
857struct ext3_xattr_ibody_find {
858 struct ext3_xattr_search s;
859 struct ext3_iloc iloc;
860};
861
862int
863ext3_xattr_ibody_find(struct inode *inode, struct ext3_xattr_info *i,
864 struct ext3_xattr_ibody_find *is)
865{
866 struct ext3_xattr_ibody_header *header;
867 struct ext3_inode *raw_inode;
868 int error;
869
870 if (EXT3_I(inode)->i_extra_isize == 0)
871 return 0;
872 raw_inode = ext3_raw_inode(&is->iloc);
873 header = IHDR(inode, raw_inode);
874 is->s.base = is->s.first = IFIRST(header);
875 is->s.here = is->s.first;
876 is->s.end = (void *)raw_inode + EXT3_SB(inode->i_sb)->s_inode_size;
877 if (EXT3_I(inode)->i_state & EXT3_STATE_XATTR) {
878 error = ext3_xattr_check_names(IFIRST(header), is->s.end);
879 if (error)
880 return error;
881 /* Find the named attribute. */
882 error = ext3_xattr_find_entry(&is->s.here, i->name_index,
883 i->name, is->s.end -
884 (void *)is->s.base, 0);
885 if (error && error != -ENODATA)
886 return error;
887 is->s.not_found = error;
888 }
889 return 0;
890}
891
892static int
893ext3_xattr_ibody_set(handle_t *handle, struct inode *inode,
894 struct ext3_xattr_info *i,
895 struct ext3_xattr_ibody_find *is)
896{
897 struct ext3_xattr_ibody_header *header;
898 struct ext3_xattr_search *s = &is->s;
899 int error;
900
901 if (EXT3_I(inode)->i_extra_isize == 0)
902 return -ENOSPC;
903 error = ext3_xattr_set_entry(i, s);
904 if (error)
905 return error;
906 header = IHDR(inode, ext3_raw_inode(&is->iloc));
907 if (!IS_LAST_ENTRY(s->first)) {
908 header->h_magic = cpu_to_le32(EXT3_XATTR_MAGIC);
909 EXT3_I(inode)->i_state |= EXT3_STATE_XATTR;
910 } else {
911 header->h_magic = cpu_to_le32(0);
912 EXT3_I(inode)->i_state &= ~EXT3_STATE_XATTR;
913 }
914 return 0;
915}
916
917/*
918 * ext3_xattr_set_handle()
919 *
920 * Create, replace or remove an extended attribute for this inode. Buffer
921 * is NULL to remove an existing extended attribute, and non-NULL to
922 * either replace an existing extended attribute, or create a new extended
923 * attribute. The flags XATTR_REPLACE and XATTR_CREATE
924 * specify that an extended attribute must exist and must not exist
925 * previous to the call, respectively.
926 *
927 * Returns 0, or a negative error number on failure.
928 */
929int
930ext3_xattr_set_handle(handle_t *handle, struct inode *inode, int name_index,
931 const char *name, const void *value, size_t value_len,
932 int flags)
933{
934 struct ext3_xattr_info i = {
935 .name_index = name_index,
936 .name = name,
937 .value = value,
938 .value_len = value_len,
939
940 };
941 struct ext3_xattr_ibody_find is = {
942 .s = { .not_found = -ENODATA, },
943 };
944 struct ext3_xattr_block_find bs = {
945 .s = { .not_found = -ENODATA, },
946 };
947 int error;
948
949 if (IS_RDONLY(inode))
950 return -EROFS;
951 if (IS_IMMUTABLE(inode) || IS_APPEND(inode))
952 return -EPERM;
953 if (!name)
954 return -EINVAL;
955 if (strlen(name) > 255)
956 return -ERANGE;
957 down_write(&EXT3_I(inode)->xattr_sem);
958 error = ext3_get_inode_loc(inode, &is.iloc);
959 if (error)
960 goto cleanup;
961
962 if (EXT3_I(inode)->i_state & EXT3_STATE_NEW) {
963 struct ext3_inode *raw_inode = ext3_raw_inode(&is.iloc);
964 memset(raw_inode, 0, EXT3_SB(inode->i_sb)->s_inode_size);
965 EXT3_I(inode)->i_state &= ~EXT3_STATE_NEW;
966 }
967
968 error = ext3_xattr_ibody_find(inode, &i, &is);
969 if (error)
970 goto cleanup;
971 if (is.s.not_found)
972 error = ext3_xattr_block_find(inode, &i, &bs);
973 if (error)
974 goto cleanup;
975 if (is.s.not_found && bs.s.not_found) {
976 error = -ENODATA;
977 if (flags & XATTR_REPLACE)
978 goto cleanup;
979 error = 0;
980 if (!value)
981 goto cleanup;
982 } else {
983 error = -EEXIST;
984 if (flags & XATTR_CREATE)
985 goto cleanup;
986 }
987 error = ext3_journal_get_write_access(handle, is.iloc.bh);
988 if (error)
989 goto cleanup;
990 if (!value) {
991 if (!is.s.not_found)
992 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
993 else if (!bs.s.not_found)
994 error = ext3_xattr_block_set(handle, inode, &i, &bs);
995 } else {
996 error = ext3_xattr_ibody_set(handle, inode, &i, &is);
997 if (!error && !bs.s.not_found) {
998 i.value = NULL;
999 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1000 } else if (error == -ENOSPC) {
1001 error = ext3_xattr_block_set(handle, inode, &i, &bs);
1002 if (error)
1003 goto cleanup;
1004 if (!is.s.not_found) {
1005 i.value = NULL;
1006 error = ext3_xattr_ibody_set(handle, inode, &i,
1007 &is);
1008 }
1009 }
1010 }
1011 if (!error) {
1012 ext3_xattr_update_super_block(handle, inode->i_sb);
1013 inode->i_ctime = CURRENT_TIME_SEC;
1014 error = ext3_mark_iloc_dirty(handle, inode, &is.iloc);
1015 /*
1016 * The bh is consumed by ext3_mark_iloc_dirty, even with
1017 * error != 0.
1018 */
1019 is.iloc.bh = NULL;
1020 if (IS_SYNC(inode))
1021 handle->h_sync = 1;
1022 }
1023
1024cleanup:
1025 brelse(is.iloc.bh);
1026 brelse(bs.bh);
1027 up_write(&EXT3_I(inode)->xattr_sem);
1028 return error;
1029}
1030
1031/*
1032 * ext3_xattr_set()
1033 *
1034 * Like ext3_xattr_set_handle, but start from an inode. This extended
1035 * attribute modification is a filesystem transaction by itself.
1036 *
1037 * Returns 0, or a negative error number on failure.
1038 */
1039int
1040ext3_xattr_set(struct inode *inode, int name_index, const char *name,
1041 const void *value, size_t value_len, int flags)
1042{
1043 handle_t *handle;
1044 int error, retries = 0;
1045
1046retry:
1047 handle = ext3_journal_start(inode, EXT3_DATA_TRANS_BLOCKS);
1048 if (IS_ERR(handle)) {
1049 error = PTR_ERR(handle);
1050 } else {
1051 int error2;
1052
1053 error = ext3_xattr_set_handle(handle, inode, name_index, name,
1054 value, value_len, flags);
1055 error2 = ext3_journal_stop(handle);
1056 if (error == -ENOSPC &&
1057 ext3_should_retry_alloc(inode->i_sb, &retries))
1058 goto retry;
1059 if (error == 0)
1060 error = error2;
1061 }
1062
1063 return error;
1064}
1065
1066/*
1067 * ext3_xattr_delete_inode()
1068 *
1069 * Free extended attribute resources associated with this inode. This
1070 * is called immediately before an inode is freed. We have exclusive
1071 * access to the inode.
1072 */
1073void
1074ext3_xattr_delete_inode(handle_t *handle, struct inode *inode)
1075{
1076 struct buffer_head *bh = NULL;
1077
1078 if (!EXT3_I(inode)->i_file_acl)
1079 goto cleanup;
1080 bh = sb_bread(inode->i_sb, EXT3_I(inode)->i_file_acl);
1081 if (!bh) {
1082 ext3_error(inode->i_sb, __FUNCTION__,
1083 "inode %ld: block %d read error", inode->i_ino,
1084 EXT3_I(inode)->i_file_acl);
1085 goto cleanup;
1086 }
1087 if (BHDR(bh)->h_magic != cpu_to_le32(EXT3_XATTR_MAGIC) ||
1088 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1089 ext3_error(inode->i_sb, __FUNCTION__,
1090 "inode %ld: bad block %d", inode->i_ino,
1091 EXT3_I(inode)->i_file_acl);
1092 goto cleanup;
1093 }
1094 ext3_xattr_release_block(handle, inode, bh);
1095 EXT3_I(inode)->i_file_acl = 0;
1096
1097cleanup:
1098 brelse(bh);
1099}
1100
1101/*
1102 * ext3_xattr_put_super()
1103 *
1104 * This is called when a file system is unmounted.
1105 */
1106void
1107ext3_xattr_put_super(struct super_block *sb)
1108{
1109 mb_cache_shrink(ext3_xattr_cache, sb->s_bdev);
1110}
1111
1112/*
1113 * ext3_xattr_cache_insert()
1114 *
1115 * Create a new entry in the extended attribute cache, and insert
1116 * it unless such an entry is already in the cache.
1117 *
1118 * Returns 0, or a negative error number on failure.
1119 */
1120static void
1121ext3_xattr_cache_insert(struct buffer_head *bh)
1122{
1123 __u32 hash = le32_to_cpu(BHDR(bh)->h_hash);
1124 struct mb_cache_entry *ce;
1125 int error;
1126
1127 ce = mb_cache_entry_alloc(ext3_xattr_cache);
1128 if (!ce) {
1129 ea_bdebug(bh, "out of memory");
1130 return;
1131 }
1132 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash);
1133 if (error) {
1134 mb_cache_entry_free(ce);
1135 if (error == -EBUSY) {
1136 ea_bdebug(bh, "already in cache");
1137 error = 0;
1138 }
1139 } else {
1140 ea_bdebug(bh, "inserting [%x]", (int)hash);
1141 mb_cache_entry_release(ce);
1142 }
1143}
1144
1145/*
1146 * ext3_xattr_cmp()
1147 *
1148 * Compare two extended attribute blocks for equality.
1149 *
1150 * Returns 0 if the blocks are equal, 1 if they differ, and
1151 * a negative error number on errors.
1152 */
1153static int
1154ext3_xattr_cmp(struct ext3_xattr_header *header1,
1155 struct ext3_xattr_header *header2)
1156{
1157 struct ext3_xattr_entry *entry1, *entry2;
1158
1159 entry1 = ENTRY(header1+1);
1160 entry2 = ENTRY(header2+1);
1161 while (!IS_LAST_ENTRY(entry1)) {
1162 if (IS_LAST_ENTRY(entry2))
1163 return 1;
1164 if (entry1->e_hash != entry2->e_hash ||
1165 entry1->e_name_index != entry2->e_name_index ||
1166 entry1->e_name_len != entry2->e_name_len ||
1167 entry1->e_value_size != entry2->e_value_size ||
1168 memcmp(entry1->e_name, entry2->e_name, entry1->e_name_len))
1169 return 1;
1170 if (entry1->e_value_block != 0 || entry2->e_value_block != 0)
1171 return -EIO;
1172 if (memcmp((char *)header1 + le16_to_cpu(entry1->e_value_offs),
1173 (char *)header2 + le16_to_cpu(entry2->e_value_offs),
1174 le32_to_cpu(entry1->e_value_size)))
1175 return 1;
1176
1177 entry1 = EXT3_XATTR_NEXT(entry1);
1178 entry2 = EXT3_XATTR_NEXT(entry2);
1179 }
1180 if (!IS_LAST_ENTRY(entry2))
1181 return 1;
1182 return 0;
1183}
1184
1185/*
1186 * ext3_xattr_cache_find()
1187 *
1188 * Find an identical extended attribute block.
1189 *
1190 * Returns a pointer to the block found, or NULL if such a block was
1191 * not found or an error occurred.
1192 */
1193static struct buffer_head *
1194ext3_xattr_cache_find(struct inode *inode, struct ext3_xattr_header *header,
1195 struct mb_cache_entry **pce)
1196{
1197 __u32 hash = le32_to_cpu(header->h_hash);
1198 struct mb_cache_entry *ce;
1199
1200 if (!header->h_hash)
1201 return NULL; /* never share */
1202 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1203again:
1204 ce = mb_cache_entry_find_first(ext3_xattr_cache, 0,
1205 inode->i_sb->s_bdev, hash);
1206 while (ce) {
1207 struct buffer_head *bh;
1208
1209 if (IS_ERR(ce)) {
1210 if (PTR_ERR(ce) == -EAGAIN)
1211 goto again;
1212 break;
1213 }
1214 bh = sb_bread(inode->i_sb, ce->e_block);
1215 if (!bh) {
1216 ext3_error(inode->i_sb, __FUNCTION__,
1217 "inode %ld: block %ld read error",
1218 inode->i_ino, (unsigned long) ce->e_block);
1219 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
1220 EXT3_XATTR_REFCOUNT_MAX) {
1221 ea_idebug(inode, "block %ld refcount %d>=%d",
1222 (unsigned long) ce->e_block,
1223 le32_to_cpu(BHDR(bh)->h_refcount),
1224 EXT3_XATTR_REFCOUNT_MAX);
1225 } else if (ext3_xattr_cmp(header, BHDR(bh)) == 0) {
1226 *pce = ce;
1227 return bh;
1228 }
1229 brelse(bh);
1230 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash);
1231 }
1232 return NULL;
1233}
1234
1235#define NAME_HASH_SHIFT 5
1236#define VALUE_HASH_SHIFT 16
1237
1238/*
1239 * ext3_xattr_hash_entry()
1240 *
1241 * Compute the hash of an extended attribute.
1242 */
1243static inline void ext3_xattr_hash_entry(struct ext3_xattr_header *header,
1244 struct ext3_xattr_entry *entry)
1245{
1246 __u32 hash = 0;
1247 char *name = entry->e_name;
1248 int n;
1249
1250 for (n=0; n < entry->e_name_len; n++) {
1251 hash = (hash << NAME_HASH_SHIFT) ^
1252 (hash >> (8*sizeof(hash) - NAME_HASH_SHIFT)) ^
1253 *name++;
1254 }
1255
1256 if (entry->e_value_block == 0 && entry->e_value_size != 0) {
1257 __le32 *value = (__le32 *)((char *)header +
1258 le16_to_cpu(entry->e_value_offs));
1259 for (n = (le32_to_cpu(entry->e_value_size) +
1260 EXT3_XATTR_ROUND) >> EXT3_XATTR_PAD_BITS; n; n--) {
1261 hash = (hash << VALUE_HASH_SHIFT) ^
1262 (hash >> (8*sizeof(hash) - VALUE_HASH_SHIFT)) ^
1263 le32_to_cpu(*value++);
1264 }
1265 }
1266 entry->e_hash = cpu_to_le32(hash);
1267}
1268
1269#undef NAME_HASH_SHIFT
1270#undef VALUE_HASH_SHIFT
1271
1272#define BLOCK_HASH_SHIFT 16
1273
1274/*
1275 * ext3_xattr_rehash()
1276 *
1277 * Re-compute the extended attribute hash value after an entry has changed.
1278 */
1279static void ext3_xattr_rehash(struct ext3_xattr_header *header,
1280 struct ext3_xattr_entry *entry)
1281{
1282 struct ext3_xattr_entry *here;
1283 __u32 hash = 0;
1284
1285 ext3_xattr_hash_entry(header, entry);
1286 here = ENTRY(header+1);
1287 while (!IS_LAST_ENTRY(here)) {
1288 if (!here->e_hash) {
1289 /* Block is not shared if an entry's hash value == 0 */
1290 hash = 0;
1291 break;
1292 }
1293 hash = (hash << BLOCK_HASH_SHIFT) ^
1294 (hash >> (8*sizeof(hash) - BLOCK_HASH_SHIFT)) ^
1295 le32_to_cpu(here->e_hash);
1296 here = EXT3_XATTR_NEXT(here);
1297 }
1298 header->h_hash = cpu_to_le32(hash);
1299}
1300
1301#undef BLOCK_HASH_SHIFT
1302
1303int __init
1304init_ext3_xattr(void)
1305{
1306 ext3_xattr_cache = mb_cache_create("ext3_xattr", NULL,
1307 sizeof(struct mb_cache_entry) +
1308 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1309 if (!ext3_xattr_cache)
1310 return -ENOMEM;
1311 return 0;
1312}
1313
1314void
1315exit_ext3_xattr(void)
1316{
1317 if (ext3_xattr_cache)
1318 mb_cache_destroy(ext3_xattr_cache);
1319 ext3_xattr_cache = NULL;
1320}