?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
ocfs2/xattr: Restore not_found in xis
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ocfs2 / xattr.c
blob9cb71e1c7c60fcd8fb6a4a4bd6e7b4aba2ebb93f
1 /* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * xattr.c
5 *
6 * Copyright (C) 2004, 2008 Oracle. All rights reserved.
7 *
8 * CREDITS:
9 * Lots of code in this file is copy from linux/fs/ext3/xattr.c.
10 * Copyright (C) 2001-2003 Andreas Gruenbacher, <agruen@suse.de>
11 *
12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public
14 * License version 2 as published by the Free Software Foundation.
15 *
16 * This program is distributed in the hope that it will be useful,
17 * but WITHOUT ANY WARRANTY; without even the implied warranty of
18 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
19 * General Public License for more details.
20 */
21
22 #include <linux/capability.h>
23 #include <linux/fs.h>
24 #include <linux/types.h>
25 #include <linux/slab.h>
26 #include <linux/highmem.h>
27 #include <linux/pagemap.h>
28 #include <linux/uio.h>
29 #include <linux/sched.h>
30 #include <linux/splice.h>
31 #include <linux/mount.h>
32 #include <linux/writeback.h>
33 #include <linux/falloc.h>
34 #include <linux/sort.h>
35 #include <linux/init.h>
36 #include <linux/module.h>
37 #include <linux/string.h>
38 #include <linux/security.h>
39
40 #define MLOG_MASK_PREFIX ML_XATTR
41 #include <cluster/masklog.h>
42
43 #include "ocfs2.h"
44 #include "alloc.h"
45 #include "dlmglue.h"
46 #include "file.h"
47 #include "symlink.h"
48 #include "sysfile.h"
49 #include "inode.h"
50 #include "journal.h"
51 #include "ocfs2_fs.h"
52 #include "suballoc.h"
53 #include "uptodate.h"
54 #include "buffer_head_io.h"
55 #include "super.h"
56 #include "xattr.h"
57
58
59 struct ocfs2_xattr_def_value_root {
60 struct ocfs2_xattr_value_root xv;
61 struct ocfs2_extent_rec er;
62 };
63
64 struct ocfs2_xattr_bucket {
65 /* The inode these xattrs are associated with */
66 struct inode *bu_inode;
67
68 /* The actual buffers that make up the bucket */
69 struct buffer_head *bu_bhs[OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET];
70
71 /* How many blocks make up one bucket for this filesystem */
72 int bu_blocks;
73 };
74
75 struct ocfs2_xattr_set_ctxt {
76 handle_t *handle;
77 struct ocfs2_alloc_context *meta_ac;
78 struct ocfs2_alloc_context *data_ac;
79 struct ocfs2_cached_dealloc_ctxt dealloc;
80 };
81
82 #define OCFS2_XATTR_ROOT_SIZE (sizeof(struct ocfs2_xattr_def_value_root))
83 #define OCFS2_XATTR_INLINE_SIZE 80
84 #define OCFS2_XATTR_FREE_IN_IBODY (OCFS2_MIN_XATTR_INLINE_SIZE \
85 - sizeof(struct ocfs2_xattr_header) \
86 - sizeof(__u32))
87 #define OCFS2_XATTR_FREE_IN_BLOCK(ptr) ((ptr)->i_sb->s_blocksize \
88 - sizeof(struct ocfs2_xattr_block) \
89 - sizeof(struct ocfs2_xattr_header) \
90 - sizeof(__u32))
91
92 static struct ocfs2_xattr_def_value_root def_xv = {
93 .xv.xr_list.l_count = cpu_to_le16(1),
94 };
95
96 struct xattr_handler *ocfs2_xattr_handlers[] = {
97 &ocfs2_xattr_user_handler,
98 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
99 &ocfs2_xattr_acl_access_handler,
100 &ocfs2_xattr_acl_default_handler,
101 #endif
102 &ocfs2_xattr_trusted_handler,
103 &ocfs2_xattr_security_handler,
104 NULL
105 };
106
107 static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
108 [OCFS2_XATTR_INDEX_USER] = &ocfs2_xattr_user_handler,
109 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
110 [OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS]
111 = &ocfs2_xattr_acl_access_handler,
112 [OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT]
113 = &ocfs2_xattr_acl_default_handler,
114 #endif
115 [OCFS2_XATTR_INDEX_TRUSTED] = &ocfs2_xattr_trusted_handler,
116 [OCFS2_XATTR_INDEX_SECURITY] = &ocfs2_xattr_security_handler,
117 };
118
119 struct ocfs2_xattr_info {
120 int name_index;
121 const char *name;
122 const void *value;
123 size_t value_len;
124 };
125
126 struct ocfs2_xattr_search {
127 struct buffer_head *inode_bh;
128 /*
129 * xattr_bh point to the block buffer head which has extended attribute
130 * when extended attribute in inode, xattr_bh is equal to inode_bh.
131 */
132 struct buffer_head *xattr_bh;
133 struct ocfs2_xattr_header *header;
134 struct ocfs2_xattr_bucket *bucket;
135 void *base;
136 void *end;
137 struct ocfs2_xattr_entry *here;
138 int not_found;
139 };
140
141 static int ocfs2_xattr_bucket_get_name_value(struct inode *inode,
142 struct ocfs2_xattr_header *xh,
143 int index,
144 int *block_off,
145 int *new_offset);
146
147 static int ocfs2_xattr_block_find(struct inode *inode,
148 int name_index,
149 const char *name,
150 struct ocfs2_xattr_search *xs);
151 static int ocfs2_xattr_index_block_find(struct inode *inode,
152 struct buffer_head *root_bh,
153 int name_index,
154 const char *name,
155 struct ocfs2_xattr_search *xs);
156
157 static int ocfs2_xattr_tree_list_index_block(struct inode *inode,
158 struct ocfs2_xattr_tree_root *xt,
159 char *buffer,
160 size_t buffer_size);
161
162 static int ocfs2_xattr_create_index_block(struct inode *inode,
163 struct ocfs2_xattr_search *xs,
164 struct ocfs2_xattr_set_ctxt *ctxt);
165
166 static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
167 struct ocfs2_xattr_info *xi,
168 struct ocfs2_xattr_search *xs,
169 struct ocfs2_xattr_set_ctxt *ctxt);
170
171 static int ocfs2_delete_xattr_index_block(struct inode *inode,
172 struct buffer_head *xb_bh);
173
174 static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb)
175 {
176 return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE;
177 }
178
179 static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb)
180 {
181 return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits);
182 }
183
184 static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb)
185 {
186 u16 len = sb->s_blocksize -
187 offsetof(struct ocfs2_xattr_header, xh_entries);
188
189 return len / sizeof(struct ocfs2_xattr_entry);
190 }
191
192 #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr)
193 #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data)
194 #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0))
195
196 static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode)
197 {
198 struct ocfs2_xattr_bucket *bucket;
199 int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
200
201 BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET);
202
203 bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS);
204 if (bucket) {
205 bucket->bu_inode = inode;
206 bucket->bu_blocks = blks;
207 }
208
209 return bucket;
210 }
211
212 static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket)
213 {
214 int i;
215
216 for (i = 0; i < bucket->bu_blocks; i++) {
217 brelse(bucket->bu_bhs[i]);
218 bucket->bu_bhs[i] = NULL;
219 }
220 }
221
222 static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket)
223 {
224 if (bucket) {
225 ocfs2_xattr_bucket_relse(bucket);
226 bucket->bu_inode = NULL;
227 kfree(bucket);
228 }
229 }
230
231 /*
232 * A bucket that has never been written to disk doesn't need to be
233 * read. We just need the buffer_heads. Don't call this for
234 * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes
235 * them fully.
236 */
237 static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
238 u64 xb_blkno)
239 {
240 int i, rc = 0;
241
242 for (i = 0; i < bucket->bu_blocks; i++) {
243 bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb,
244 xb_blkno + i);
245 if (!bucket->bu_bhs[i]) {
246 rc = -EIO;
247 mlog_errno(rc);
248 break;
249 }
250
251 if (!ocfs2_buffer_uptodate(bucket->bu_inode,
252 bucket->bu_bhs[i]))
253 ocfs2_set_new_buffer_uptodate(bucket->bu_inode,
254 bucket->bu_bhs[i]);
255 }
256
257 if (rc)
258 ocfs2_xattr_bucket_relse(bucket);
259 return rc;
260 }
261
262 /* Read the xattr bucket at xb_blkno */
263 static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
264 u64 xb_blkno)
265 {
266 int rc;
267
268 rc = ocfs2_read_blocks(bucket->bu_inode, xb_blkno,
269 bucket->bu_blocks, bucket->bu_bhs, 0,
270 NULL);
271 if (rc)
272 ocfs2_xattr_bucket_relse(bucket);
273 return rc;
274 }
275
276 static int ocfs2_xattr_bucket_journal_access(handle_t *handle,
277 struct ocfs2_xattr_bucket *bucket,
278 int type)
279 {
280 int i, rc = 0;
281
282 for (i = 0; i < bucket->bu_blocks; i++) {
283 rc = ocfs2_journal_access(handle, bucket->bu_inode,
284 bucket->bu_bhs[i], type);
285 if (rc) {
286 mlog_errno(rc);
287 break;
288 }
289 }
290
291 return rc;
292 }
293
294 static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle,
295 struct ocfs2_xattr_bucket *bucket)
296 {
297 int i;
298
299 for (i = 0; i < bucket->bu_blocks; i++)
300 ocfs2_journal_dirty(handle, bucket->bu_bhs[i]);
301 }
302
303 static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest,
304 struct ocfs2_xattr_bucket *src)
305 {
306 int i;
307 int blocksize = src->bu_inode->i_sb->s_blocksize;
308
309 BUG_ON(dest->bu_blocks != src->bu_blocks);
310 BUG_ON(dest->bu_inode != src->bu_inode);
311
312 for (i = 0; i < src->bu_blocks; i++) {
313 memcpy(bucket_block(dest, i), bucket_block(src, i),
314 blocksize);
315 }
316 }
317
318 static int ocfs2_validate_xattr_block(struct super_block *sb,
319 struct buffer_head *bh)
320 {
321 struct ocfs2_xattr_block *xb =
322 (struct ocfs2_xattr_block *)bh->b_data;
323
324 mlog(0, "Validating xattr block %llu\n",
325 (unsigned long long)bh->b_blocknr);
326
327 if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) {
328 ocfs2_error(sb,
329 "Extended attribute block #%llu has bad "
330 "signature %.*s",
331 (unsigned long long)bh->b_blocknr, 7,
332 xb->xb_signature);
333 return -EINVAL;
334 }
335
336 if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) {
337 ocfs2_error(sb,
338 "Extended attribute block #%llu has an "
339 "invalid xb_blkno of %llu",
340 (unsigned long long)bh->b_blocknr,
341 (unsigned long long)le64_to_cpu(xb->xb_blkno));
342 return -EINVAL;
343 }
344
345 if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) {
346 ocfs2_error(sb,
347 "Extended attribute block #%llu has an invalid "
348 "xb_fs_generation of #%u",
349 (unsigned long long)bh->b_blocknr,
350 le32_to_cpu(xb->xb_fs_generation));
351 return -EINVAL;
352 }
353
354 return 0;
355 }
356
357 static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno,
358 struct buffer_head **bh)
359 {
360 int rc;
361 struct buffer_head *tmp = *bh;
362
363 rc = ocfs2_read_block(inode, xb_blkno, &tmp,
364 ocfs2_validate_xattr_block);
365
366 /* If ocfs2_read_block() got us a new bh, pass it up. */
367 if (!rc && !*bh)
368 *bh = tmp;
369
370 return rc;
371 }
372
373 static inline const char *ocfs2_xattr_prefix(int name_index)
374 {
375 struct xattr_handler *handler = NULL;
376
377 if (name_index > 0 && name_index < OCFS2_XATTR_MAX)
378 handler = ocfs2_xattr_handler_map[name_index];
379
380 return handler ? handler->prefix : NULL;
381 }
382
383 static u32 ocfs2_xattr_name_hash(struct inode *inode,
384 const char *name,
385 int name_len)
386 {
387 /* Get hash value of uuid from super block */
388 u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash;
389 int i;
390
391 /* hash extended attribute name */
392 for (i = 0; i < name_len; i++) {
393 hash = (hash << OCFS2_HASH_SHIFT) ^
394 (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^
395 *name++;
396 }
397
398 return hash;
399 }
400
401 /*
402 * ocfs2_xattr_hash_entry()
403 *
404 * Compute the hash of an extended attribute.
405 */
406 static void ocfs2_xattr_hash_entry(struct inode *inode,
407 struct ocfs2_xattr_header *header,
408 struct ocfs2_xattr_entry *entry)
409 {
410 u32 hash = 0;
411 char *name = (char *)header + le16_to_cpu(entry->xe_name_offset);
412
413 hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len);
414 entry->xe_name_hash = cpu_to_le32(hash);
415
416 return;
417 }
418
419 static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len)
420 {
421 int size = 0;
422
423 if (value_len <= OCFS2_XATTR_INLINE_SIZE)
424 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
425 else
426 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
427 size += sizeof(struct ocfs2_xattr_entry);
428
429 return size;
430 }
431
432 int ocfs2_calc_security_init(struct inode *dir,
433 struct ocfs2_security_xattr_info *si,
434 int *want_clusters,
435 int *xattr_credits,
436 struct ocfs2_alloc_context **xattr_ac)
437 {
438 int ret = 0;
439 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
440 int s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
441 si->value_len);
442
443 /*
444 * The max space of security xattr taken inline is
445 * 256(name) + 80(value) + 16(entry) = 352 bytes,
446 * So reserve one metadata block for it is ok.
447 */
448 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
449 s_size > OCFS2_XATTR_FREE_IN_IBODY) {
450 ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac);
451 if (ret) {
452 mlog_errno(ret);
453 return ret;
454 }
455 *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
456 }
457
458 /* reserve clusters for xattr value which will be set in B tree*/
459 if (si->value_len > OCFS2_XATTR_INLINE_SIZE)
460 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
461 si->value_len);
462 return ret;
463 }
464
465 int ocfs2_calc_xattr_init(struct inode *dir,
466 struct buffer_head *dir_bh,
467 int mode,
468 struct ocfs2_security_xattr_info *si,
469 int *want_clusters,
470 int *xattr_credits,
471 struct ocfs2_alloc_context **xattr_ac)
472 {
473 int ret = 0;
474 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
475 int s_size = 0;
476 int a_size = 0;
477 int acl_len = 0;
478
479 if (si->enable)
480 s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
481 si->value_len);
482
483 if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) {
484 acl_len = ocfs2_xattr_get_nolock(dir, dir_bh,
485 OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT,
486 "", NULL, 0);
487 if (acl_len > 0) {
488 a_size = ocfs2_xattr_entry_real_size(0, acl_len);
489 if (S_ISDIR(mode))
490 a_size <<= 1;
491 } else if (acl_len != 0 && acl_len != -ENODATA) {
492 mlog_errno(ret);
493 return ret;
494 }
495 }
496
497 if (!(s_size + a_size))
498 return ret;
499
500 /*
501 * The max space of security xattr taken inline is
502 * 256(name) + 80(value) + 16(entry) = 352 bytes,
503 * The max space of acl xattr taken inline is
504 * 80(value) + 16(entry) * 2(if directory) = 192 bytes,
505 * when blocksize = 512, may reserve one more cluser for
506 * xattr bucket, otherwise reserve one metadata block
507 * for them is ok.
508 */
509 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
510 (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) {
511 ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac);
512 if (ret) {
513 mlog_errno(ret);
514 return ret;
515 }
516 *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
517 }
518
519 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE &&
520 (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) {
521 *want_clusters += 1;
522 *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb);
523 }
524
525 /* reserve clusters for xattr value which will be set in B tree*/
526 if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE)
527 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
528 si->value_len);
529 if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL &&
530 acl_len > OCFS2_XATTR_INLINE_SIZE) {
531 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb, acl_len);
532 if (S_ISDIR(mode))
533 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
534 acl_len);
535 }
536
537 return ret;
538 }
539
540 static int ocfs2_xattr_extend_allocation(struct inode *inode,
541 u32 clusters_to_add,
542 struct buffer_head *xattr_bh,
543 struct ocfs2_xattr_value_root *xv,
544 struct ocfs2_xattr_set_ctxt *ctxt)
545 {
546 int status = 0;
547 handle_t *handle = ctxt->handle;
548 enum ocfs2_alloc_restarted why;
549 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
550 u32 prev_clusters, logical_start = le32_to_cpu(xv->xr_clusters);
551 struct ocfs2_extent_tree et;
552
553 mlog(0, "(clusters_to_add for xattr= %u)\n", clusters_to_add);
554
555 ocfs2_init_xattr_value_extent_tree(&et, inode, xattr_bh, xv);
556
557 status = ocfs2_journal_access(handle, inode, xattr_bh,
558 OCFS2_JOURNAL_ACCESS_WRITE);
559 if (status < 0) {
560 mlog_errno(status);
561 goto leave;
562 }
563
564 prev_clusters = le32_to_cpu(xv->xr_clusters);
565 status = ocfs2_add_clusters_in_btree(osb,
566 inode,
567 &logical_start,
568 clusters_to_add,
569 0,
570 &et,
571 handle,
572 ctxt->data_ac,
573 ctxt->meta_ac,
574 &why);
575 if (status < 0) {
576 mlog_errno(status);
577 goto leave;
578 }
579
580 status = ocfs2_journal_dirty(handle, xattr_bh);
581 if (status < 0) {
582 mlog_errno(status);
583 goto leave;
584 }
585
586 clusters_to_add -= le32_to_cpu(xv->xr_clusters) - prev_clusters;
587
588 /*
589 * We should have already allocated enough space before the transaction,
590 * so no need to restart.
591 */
592 BUG_ON(why != RESTART_NONE || clusters_to_add);
593
594 leave:
595
596 return status;
597 }
598
599 static int __ocfs2_remove_xattr_range(struct inode *inode,
600 struct buffer_head *root_bh,
601 struct ocfs2_xattr_value_root *xv,
602 u32 cpos, u32 phys_cpos, u32 len,
603 struct ocfs2_xattr_set_ctxt *ctxt)
604 {
605 int ret;
606 u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
607 handle_t *handle = ctxt->handle;
608 struct ocfs2_extent_tree et;
609
610 ocfs2_init_xattr_value_extent_tree(&et, inode, root_bh, xv);
611
612 ret = ocfs2_journal_access(handle, inode, root_bh,
613 OCFS2_JOURNAL_ACCESS_WRITE);
614 if (ret) {
615 mlog_errno(ret);
616 goto out;
617 }
618
619 ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, ctxt->meta_ac,
620 &ctxt->dealloc);
621 if (ret) {
622 mlog_errno(ret);
623 goto out;
624 }
625
626 le32_add_cpu(&xv->xr_clusters, -len);
627
628 ret = ocfs2_journal_dirty(handle, root_bh);
629 if (ret) {
630 mlog_errno(ret);
631 goto out;
632 }
633
634 ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, phys_blkno, len);
635 if (ret)
636 mlog_errno(ret);
637
638 out:
639 return ret;
640 }
641
642 static int ocfs2_xattr_shrink_size(struct inode *inode,
643 u32 old_clusters,
644 u32 new_clusters,
645 struct buffer_head *root_bh,
646 struct ocfs2_xattr_value_root *xv,
647 struct ocfs2_xattr_set_ctxt *ctxt)
648 {
649 int ret = 0;
650 u32 trunc_len, cpos, phys_cpos, alloc_size;
651 u64 block;
652
653 if (old_clusters <= new_clusters)
654 return 0;
655
656 cpos = new_clusters;
657 trunc_len = old_clusters - new_clusters;
658 while (trunc_len) {
659 ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos,
660 &alloc_size, &xv->xr_list);
661 if (ret) {
662 mlog_errno(ret);
663 goto out;
664 }
665
666 if (alloc_size > trunc_len)
667 alloc_size = trunc_len;
668
669 ret = __ocfs2_remove_xattr_range(inode, root_bh, xv, cpos,
670 phys_cpos, alloc_size,
671 ctxt);
672 if (ret) {
673 mlog_errno(ret);
674 goto out;
675 }
676
677 block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
678 ocfs2_remove_xattr_clusters_from_cache(inode, block,
679 alloc_size);
680 cpos += alloc_size;
681 trunc_len -= alloc_size;
682 }
683
684 out:
685 return ret;
686 }
687
688 static int ocfs2_xattr_value_truncate(struct inode *inode,
689 struct buffer_head *root_bh,
690 struct ocfs2_xattr_value_root *xv,
691 int len,
692 struct ocfs2_xattr_set_ctxt *ctxt)
693 {
694 int ret;
695 u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len);
696 u32 old_clusters = le32_to_cpu(xv->xr_clusters);
697
698 if (new_clusters == old_clusters)
699 return 0;
700
701 if (new_clusters > old_clusters)
702 ret = ocfs2_xattr_extend_allocation(inode,
703 new_clusters - old_clusters,
704 root_bh, xv, ctxt);
705 else
706 ret = ocfs2_xattr_shrink_size(inode,
707 old_clusters, new_clusters,
708 root_bh, xv, ctxt);
709
710 return ret;
711 }
712
713 static int ocfs2_xattr_list_entry(char *buffer, size_t size,
714 size_t *result, const char *prefix,
715 const char *name, int name_len)
716 {
717 char *p = buffer + *result;
718 int prefix_len = strlen(prefix);
719 int total_len = prefix_len + name_len + 1;
720
721 *result += total_len;
722
723 /* we are just looking for how big our buffer needs to be */
724 if (!size)
725 return 0;
726
727 if (*result > size)
728 return -ERANGE;
729
730 memcpy(p, prefix, prefix_len);
731 memcpy(p + prefix_len, name, name_len);
732 p[prefix_len + name_len] = '\0';
733
734 return 0;
735 }
736
737 static int ocfs2_xattr_list_entries(struct inode *inode,
738 struct ocfs2_xattr_header *header,
739 char *buffer, size_t buffer_size)
740 {
741 size_t result = 0;
742 int i, type, ret;
743 const char *prefix, *name;
744
745 for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) {
746 struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
747 type = ocfs2_xattr_get_type(entry);
748 prefix = ocfs2_xattr_prefix(type);
749
750 if (prefix) {
751 name = (const char *)header +
752 le16_to_cpu(entry->xe_name_offset);
753
754 ret = ocfs2_xattr_list_entry(buffer, buffer_size,
755 &result, prefix, name,
756 entry->xe_name_len);
757 if (ret)
758 return ret;
759 }
760 }
761
762 return result;
763 }
764
765 static int ocfs2_xattr_ibody_list(struct inode *inode,
766 struct ocfs2_dinode *di,
767 char *buffer,
768 size_t buffer_size)
769 {
770 struct ocfs2_xattr_header *header = NULL;
771 struct ocfs2_inode_info *oi = OCFS2_I(inode);
772 int ret = 0;
773
774 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
775 return ret;
776
777 header = (struct ocfs2_xattr_header *)
778 ((void *)di + inode->i_sb->s_blocksize -
779 le16_to_cpu(di->i_xattr_inline_size));
780
781 ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size);
782
783 return ret;
784 }
785
786 static int ocfs2_xattr_block_list(struct inode *inode,
787 struct ocfs2_dinode *di,
788 char *buffer,
789 size_t buffer_size)
790 {
791 struct buffer_head *blk_bh = NULL;
792 struct ocfs2_xattr_block *xb;
793 int ret = 0;
794
795 if (!di->i_xattr_loc)
796 return ret;
797
798 ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
799 &blk_bh);
800 if (ret < 0) {
801 mlog_errno(ret);
802 return ret;
803 }
804
805 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
806 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
807 struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header;
808 ret = ocfs2_xattr_list_entries(inode, header,
809 buffer, buffer_size);
810 } else {
811 struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root;
812 ret = ocfs2_xattr_tree_list_index_block(inode, xt,
813 buffer, buffer_size);
814 }
815
816 brelse(blk_bh);
817
818 return ret;
819 }
820
821 ssize_t ocfs2_listxattr(struct dentry *dentry,
822 char *buffer,
823 size_t size)
824 {
825 int ret = 0, i_ret = 0, b_ret = 0;
826 struct buffer_head *di_bh = NULL;
827 struct ocfs2_dinode *di = NULL;
828 struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode);
829
830 if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb)))
831 return -EOPNOTSUPP;
832
833 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
834 return ret;
835
836 ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0);
837 if (ret < 0) {
838 mlog_errno(ret);
839 return ret;
840 }
841
842 di = (struct ocfs2_dinode *)di_bh->b_data;
843
844 down_read(&oi->ip_xattr_sem);
845 i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size);
846 if (i_ret < 0)
847 b_ret = 0;
848 else {
849 if (buffer) {
850 buffer += i_ret;
851 size -= i_ret;
852 }
853 b_ret = ocfs2_xattr_block_list(dentry->d_inode, di,
854 buffer, size);
855 if (b_ret < 0)
856 i_ret = 0;
857 }
858 up_read(&oi->ip_xattr_sem);
859 ocfs2_inode_unlock(dentry->d_inode, 0);
860
861 brelse(di_bh);
862
863 return i_ret + b_ret;
864 }
865
866 static int ocfs2_xattr_find_entry(int name_index,
867 const char *name,
868 struct ocfs2_xattr_search *xs)
869 {
870 struct ocfs2_xattr_entry *entry;
871 size_t name_len;
872 int i, cmp = 1;
873
874 if (name == NULL)
875 return -EINVAL;
876
877 name_len = strlen(name);
878 entry = xs->here;
879 for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
880 cmp = name_index - ocfs2_xattr_get_type(entry);
881 if (!cmp)
882 cmp = name_len - entry->xe_name_len;
883 if (!cmp)
884 cmp = memcmp(name, (xs->base +
885 le16_to_cpu(entry->xe_name_offset)),
886 name_len);
887 if (cmp == 0)
888 break;
889 entry += 1;
890 }
891 xs->here = entry;
892
893 return cmp ? -ENODATA : 0;
894 }
895
896 static int ocfs2_xattr_get_value_outside(struct inode *inode,
897 struct ocfs2_xattr_value_root *xv,
898 void *buffer,
899 size_t len)
900 {
901 u32 cpos, p_cluster, num_clusters, bpc, clusters;
902 u64 blkno;
903 int i, ret = 0;
904 size_t cplen, blocksize;
905 struct buffer_head *bh = NULL;
906 struct ocfs2_extent_list *el;
907
908 el = &xv->xr_list;
909 clusters = le32_to_cpu(xv->xr_clusters);
910 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
911 blocksize = inode->i_sb->s_blocksize;
912
913 cpos = 0;
914 while (cpos < clusters) {
915 ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
916 &num_clusters, el);
917 if (ret) {
918 mlog_errno(ret);
919 goto out;
920 }
921
922 blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
923 /* Copy ocfs2_xattr_value */
924 for (i = 0; i < num_clusters * bpc; i++, blkno++) {
925 ret = ocfs2_read_block(inode, blkno, &bh, NULL);
926 if (ret) {
927 mlog_errno(ret);
928 goto out;
929 }
930
931 cplen = len >= blocksize ? blocksize : len;
932 memcpy(buffer, bh->b_data, cplen);
933 len -= cplen;
934 buffer += cplen;
935
936 brelse(bh);
937 bh = NULL;
938 if (len == 0)
939 break;
940 }
941 cpos += num_clusters;
942 }
943 out:
944 return ret;
945 }
946
947 static int ocfs2_xattr_ibody_get(struct inode *inode,
948 int name_index,
949 const char *name,
950 void *buffer,
951 size_t buffer_size,
952 struct ocfs2_xattr_search *xs)
953 {
954 struct ocfs2_inode_info *oi = OCFS2_I(inode);
955 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
956 struct ocfs2_xattr_value_root *xv;
957 size_t size;
958 int ret = 0;
959
960 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
961 return -ENODATA;
962
963 xs->end = (void *)di + inode->i_sb->s_blocksize;
964 xs->header = (struct ocfs2_xattr_header *)
965 (xs->end - le16_to_cpu(di->i_xattr_inline_size));
966 xs->base = (void *)xs->header;
967 xs->here = xs->header->xh_entries;
968
969 ret = ocfs2_xattr_find_entry(name_index, name, xs);
970 if (ret)
971 return ret;
972 size = le64_to_cpu(xs->here->xe_value_size);
973 if (buffer) {
974 if (size > buffer_size)
975 return -ERANGE;
976 if (ocfs2_xattr_is_local(xs->here)) {
977 memcpy(buffer, (void *)xs->base +
978 le16_to_cpu(xs->here->xe_name_offset) +
979 OCFS2_XATTR_SIZE(xs->here->xe_name_len), size);
980 } else {
981 xv = (struct ocfs2_xattr_value_root *)
982 (xs->base + le16_to_cpu(
983 xs->here->xe_name_offset) +
984 OCFS2_XATTR_SIZE(xs->here->xe_name_len));
985 ret = ocfs2_xattr_get_value_outside(inode, xv,
986 buffer, size);
987 if (ret < 0) {
988 mlog_errno(ret);
989 return ret;
990 }
991 }
992 }
993
994 return size;
995 }
996
997 static int ocfs2_xattr_block_get(struct inode *inode,
998 int name_index,
999 const char *name,
1000 void *buffer,
1001 size_t buffer_size,
1002 struct ocfs2_xattr_search *xs)
1003 {
1004 struct ocfs2_xattr_block *xb;
1005 struct ocfs2_xattr_value_root *xv;
1006 size_t size;
1007 int ret = -ENODATA, name_offset, name_len, block_off, i;
1008
1009 xs->bucket = ocfs2_xattr_bucket_new(inode);
1010 if (!xs->bucket) {
1011 ret = -ENOMEM;
1012 mlog_errno(ret);
1013 goto cleanup;
1014 }
1015
1016 ret = ocfs2_xattr_block_find(inode, name_index, name, xs);
1017 if (ret) {
1018 mlog_errno(ret);
1019 goto cleanup;
1020 }
1021
1022 if (xs->not_found) {
1023 ret = -ENODATA;
1024 goto cleanup;
1025 }
1026
1027 xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
1028 size = le64_to_cpu(xs->here->xe_value_size);
1029 if (buffer) {
1030 ret = -ERANGE;
1031 if (size > buffer_size)
1032 goto cleanup;
1033
1034 name_offset = le16_to_cpu(xs->here->xe_name_offset);
1035 name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len);
1036 i = xs->here - xs->header->xh_entries;
1037
1038 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
1039 ret = ocfs2_xattr_bucket_get_name_value(inode,
1040 bucket_xh(xs->bucket),
1041 i,
1042 &block_off,
1043 &name_offset);
1044 xs->base = bucket_block(xs->bucket, block_off);
1045 }
1046 if (ocfs2_xattr_is_local(xs->here)) {
1047 memcpy(buffer, (void *)xs->base +
1048 name_offset + name_len, size);
1049 } else {
1050 xv = (struct ocfs2_xattr_value_root *)
1051 (xs->base + name_offset + name_len);
1052 ret = ocfs2_xattr_get_value_outside(inode, xv,
1053 buffer, size);
1054 if (ret < 0) {
1055 mlog_errno(ret);
1056 goto cleanup;
1057 }
1058 }
1059 }
1060 ret = size;
1061 cleanup:
1062 ocfs2_xattr_bucket_free(xs->bucket);
1063
1064 brelse(xs->xattr_bh);
1065 xs->xattr_bh = NULL;
1066 return ret;
1067 }
1068
1069 int ocfs2_xattr_get_nolock(struct inode *inode,
1070 struct buffer_head *di_bh,
1071 int name_index,
1072 const char *name,
1073 void *buffer,
1074 size_t buffer_size)
1075 {
1076 int ret;
1077 struct ocfs2_dinode *di = NULL;
1078 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1079 struct ocfs2_xattr_search xis = {
1080 .not_found = -ENODATA,
1081 };
1082 struct ocfs2_xattr_search xbs = {
1083 .not_found = -ENODATA,
1084 };
1085
1086 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
1087 return -EOPNOTSUPP;
1088
1089 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
1090 ret = -ENODATA;
1091
1092 xis.inode_bh = xbs.inode_bh = di_bh;
1093 di = (struct ocfs2_dinode *)di_bh->b_data;
1094
1095 down_read(&oi->ip_xattr_sem);
1096 ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer,
1097 buffer_size, &xis);
1098 if (ret == -ENODATA && di->i_xattr_loc)
1099 ret = ocfs2_xattr_block_get(inode, name_index, name, buffer,
1100 buffer_size, &xbs);
1101 up_read(&oi->ip_xattr_sem);
1102
1103 return ret;
1104 }
1105
1106 /* ocfs2_xattr_get()
1107 *
1108 * Copy an extended attribute into the buffer provided.
1109 * Buffer is NULL to compute the size of buffer required.
1110 */
1111 static int ocfs2_xattr_get(struct inode *inode,
1112 int name_index,
1113 const char *name,
1114 void *buffer,
1115 size_t buffer_size)
1116 {
1117 int ret;
1118 struct buffer_head *di_bh = NULL;
1119
1120 ret = ocfs2_inode_lock(inode, &di_bh, 0);
1121 if (ret < 0) {
1122 mlog_errno(ret);
1123 return ret;
1124 }
1125 ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index,
1126 name, buffer, buffer_size);
1127
1128 ocfs2_inode_unlock(inode, 0);
1129
1130 brelse(di_bh);
1131
1132 return ret;
1133 }
1134
1135 static int __ocfs2_xattr_set_value_outside(struct inode *inode,
1136 handle_t *handle,
1137 struct ocfs2_xattr_value_root *xv,
1138 const void *value,
1139 int value_len)
1140 {
1141 int ret = 0, i, cp_len, credits;
1142 u16 blocksize = inode->i_sb->s_blocksize;
1143 u32 p_cluster, num_clusters;
1144 u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
1145 u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len);
1146 u64 blkno;
1147 struct buffer_head *bh = NULL;
1148
1149 BUG_ON(clusters > le32_to_cpu(xv->xr_clusters));
1150
1151 /*
1152 * In __ocfs2_xattr_set_value_outside has already been dirtied,
1153 * so we don't need to worry about whether ocfs2_extend_trans
1154 * will create a new transactio for us or not.
1155 */
1156 credits = clusters * bpc;
1157 ret = ocfs2_extend_trans(handle, credits);
1158 if (ret) {
1159 mlog_errno(ret);
1160 goto out;
1161 }
1162
1163 while (cpos < clusters) {
1164 ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
1165 &num_clusters, &xv->xr_list);
1166 if (ret) {
1167 mlog_errno(ret);
1168 goto out;
1169 }
1170
1171 blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
1172
1173 for (i = 0; i < num_clusters * bpc; i++, blkno++) {
1174 ret = ocfs2_read_block(inode, blkno, &bh, NULL);
1175 if (ret) {
1176 mlog_errno(ret);
1177 goto out;
1178 }
1179
1180 ret = ocfs2_journal_access(handle,
1181 inode,
1182 bh,
1183 OCFS2_JOURNAL_ACCESS_WRITE);
1184 if (ret < 0) {
1185 mlog_errno(ret);
1186 goto out;
1187 }
1188
1189 cp_len = value_len > blocksize ? blocksize : value_len;
1190 memcpy(bh->b_data, value, cp_len);
1191 value_len -= cp_len;
1192 value += cp_len;
1193 if (cp_len < blocksize)
1194 memset(bh->b_data + cp_len, 0,
1195 blocksize - cp_len);
1196
1197 ret = ocfs2_journal_dirty(handle, bh);
1198 if (ret < 0) {
1199 mlog_errno(ret);
1200 goto out;
1201 }
1202 brelse(bh);
1203 bh = NULL;
1204
1205 /*
1206 * XXX: do we need to empty all the following
1207 * blocks in this cluster?
1208 */
1209 if (!value_len)
1210 break;
1211 }
1212 cpos += num_clusters;
1213 }
1214 out:
1215 brelse(bh);
1216
1217 return ret;
1218 }
1219
1220 static int ocfs2_xattr_cleanup(struct inode *inode,
1221 handle_t *handle,
1222 struct ocfs2_xattr_info *xi,
1223 struct ocfs2_xattr_search *xs,
1224 size_t offs)
1225 {
1226 int ret = 0;
1227 size_t name_len = strlen(xi->name);
1228 void *val = xs->base + offs;
1229 size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1230
1231 ret = ocfs2_journal_access(handle, inode, xs->xattr_bh,
1232 OCFS2_JOURNAL_ACCESS_WRITE);
1233 if (ret) {
1234 mlog_errno(ret);
1235 goto out;
1236 }
1237 /* Decrease xattr count */
1238 le16_add_cpu(&xs->header->xh_count, -1);
1239 /* Remove the xattr entry and tree root which has already be set*/
1240 memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry));
1241 memset(val, 0, size);
1242
1243 ret = ocfs2_journal_dirty(handle, xs->xattr_bh);
1244 if (ret < 0)
1245 mlog_errno(ret);
1246 out:
1247 return ret;
1248 }
1249
1250 static int ocfs2_xattr_update_entry(struct inode *inode,
1251 handle_t *handle,
1252 struct ocfs2_xattr_info *xi,
1253 struct ocfs2_xattr_search *xs,
1254 size_t offs)
1255 {
1256 int ret;
1257
1258 ret = ocfs2_journal_access(handle, inode, xs->xattr_bh,
1259 OCFS2_JOURNAL_ACCESS_WRITE);
1260 if (ret) {
1261 mlog_errno(ret);
1262 goto out;
1263 }
1264
1265 xs->here->xe_name_offset = cpu_to_le16(offs);
1266 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1267 if (xi->value_len <= OCFS2_XATTR_INLINE_SIZE)
1268 ocfs2_xattr_set_local(xs->here, 1);
1269 else
1270 ocfs2_xattr_set_local(xs->here, 0);
1271 ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
1272
1273 ret = ocfs2_journal_dirty(handle, xs->xattr_bh);
1274 if (ret < 0)
1275 mlog_errno(ret);
1276 out:
1277 return ret;
1278 }
1279
1280 /*
1281 * ocfs2_xattr_set_value_outside()
1282 *
1283 * Set large size value in B tree.
1284 */
1285 static int ocfs2_xattr_set_value_outside(struct inode *inode,
1286 struct ocfs2_xattr_info *xi,
1287 struct ocfs2_xattr_search *xs,
1288 struct ocfs2_xattr_set_ctxt *ctxt,
1289 size_t offs)
1290 {
1291 size_t name_len = strlen(xi->name);
1292 void *val = xs->base + offs;
1293 struct ocfs2_xattr_value_root *xv = NULL;
1294 size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1295 int ret = 0;
1296
1297 memset(val, 0, size);
1298 memcpy(val, xi->name, name_len);
1299 xv = (struct ocfs2_xattr_value_root *)
1300 (val + OCFS2_XATTR_SIZE(name_len));
1301 xv->xr_clusters = 0;
1302 xv->xr_last_eb_blk = 0;
1303 xv->xr_list.l_tree_depth = 0;
1304 xv->xr_list.l_count = cpu_to_le16(1);
1305 xv->xr_list.l_next_free_rec = 0;
1306
1307 ret = ocfs2_xattr_value_truncate(inode, xs->xattr_bh, xv,
1308 xi->value_len, ctxt);
1309 if (ret < 0) {
1310 mlog_errno(ret);
1311 return ret;
1312 }
1313 ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, offs);
1314 if (ret < 0) {
1315 mlog_errno(ret);
1316 return ret;
1317 }
1318 ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, xv,
1319 xi->value, xi->value_len);
1320 if (ret < 0)
1321 mlog_errno(ret);
1322
1323 return ret;
1324 }
1325
1326 /*
1327 * ocfs2_xattr_set_entry_local()
1328 *
1329 * Set, replace or remove extended attribute in local.
1330 */
1331 static void ocfs2_xattr_set_entry_local(struct inode *inode,
1332 struct ocfs2_xattr_info *xi,
1333 struct ocfs2_xattr_search *xs,
1334 struct ocfs2_xattr_entry *last,
1335 size_t min_offs)
1336 {
1337 size_t name_len = strlen(xi->name);
1338 int i;
1339
1340 if (xi->value && xs->not_found) {
1341 /* Insert the new xattr entry. */
1342 le16_add_cpu(&xs->header->xh_count, 1);
1343 ocfs2_xattr_set_type(last, xi->name_index);
1344 ocfs2_xattr_set_local(last, 1);
1345 last->xe_name_len = name_len;
1346 } else {
1347 void *first_val;
1348 void *val;
1349 size_t offs, size;
1350
1351 first_val = xs->base + min_offs;
1352 offs = le16_to_cpu(xs->here->xe_name_offset);
1353 val = xs->base + offs;
1354
1355 if (le64_to_cpu(xs->here->xe_value_size) >
1356 OCFS2_XATTR_INLINE_SIZE)
1357 size = OCFS2_XATTR_SIZE(name_len) +
1358 OCFS2_XATTR_ROOT_SIZE;
1359 else
1360 size = OCFS2_XATTR_SIZE(name_len) +
1361 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1362
1363 if (xi->value && size == OCFS2_XATTR_SIZE(name_len) +
1364 OCFS2_XATTR_SIZE(xi->value_len)) {
1365 /* The old and the new value have the
1366 same size. Just replace the value. */
1367 ocfs2_xattr_set_local(xs->here, 1);
1368 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1369 /* Clear value bytes. */
1370 memset(val + OCFS2_XATTR_SIZE(name_len),
1371 0,
1372 OCFS2_XATTR_SIZE(xi->value_len));
1373 memcpy(val + OCFS2_XATTR_SIZE(name_len),
1374 xi->value,
1375 xi->value_len);
1376 return;
1377 }
1378 /* Remove the old name+value. */
1379 memmove(first_val + size, first_val, val - first_val);
1380 memset(first_val, 0, size);
1381 xs->here->xe_name_hash = 0;
1382 xs->here->xe_name_offset = 0;
1383 ocfs2_xattr_set_local(xs->here, 1);
1384 xs->here->xe_value_size = 0;
1385
1386 min_offs += size;
1387
1388 /* Adjust all value offsets. */
1389 last = xs->header->xh_entries;
1390 for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
1391 size_t o = le16_to_cpu(last->xe_name_offset);
1392
1393 if (o < offs)
1394 last->xe_name_offset = cpu_to_le16(o + size);
1395 last += 1;
1396 }
1397
1398 if (!xi->value) {
1399 /* Remove the old entry. */
1400 last -= 1;
1401 memmove(xs->here, xs->here + 1,
1402 (void *)last - (void *)xs->here);
1403 memset(last, 0, sizeof(struct ocfs2_xattr_entry));
1404 le16_add_cpu(&xs->header->xh_count, -1);
1405 }
1406 }
1407 if (xi->value) {
1408 /* Insert the new name+value. */
1409 size_t size = OCFS2_XATTR_SIZE(name_len) +
1410 OCFS2_XATTR_SIZE(xi->value_len);
1411 void *val = xs->base + min_offs - size;
1412
1413 xs->here->xe_name_offset = cpu_to_le16(min_offs - size);
1414 memset(val, 0, size);
1415 memcpy(val, xi->name, name_len);
1416 memcpy(val + OCFS2_XATTR_SIZE(name_len),
1417 xi->value,
1418 xi->value_len);
1419 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1420 ocfs2_xattr_set_local(xs->here, 1);
1421 ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
1422 }
1423
1424 return;
1425 }
1426
1427 /*
1428 * ocfs2_xattr_set_entry()
1429 *
1430 * Set extended attribute entry into inode or block.
1431 *
1432 * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE,
1433 * We first insert tree root(ocfs2_xattr_value_root) with set_entry_local(),
1434 * then set value in B tree with set_value_outside().
1435 */
1436 static int ocfs2_xattr_set_entry(struct inode *inode,
1437 struct ocfs2_xattr_info *xi,
1438 struct ocfs2_xattr_search *xs,
1439 struct ocfs2_xattr_set_ctxt *ctxt,
1440 int flag)
1441 {
1442 struct ocfs2_xattr_entry *last;
1443 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1444 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1445 size_t min_offs = xs->end - xs->base, name_len = strlen(xi->name);
1446 size_t size_l = 0;
1447 handle_t *handle = ctxt->handle;
1448 int free, i, ret;
1449 struct ocfs2_xattr_info xi_l = {
1450 .name_index = xi->name_index,
1451 .name = xi->name,
1452 .value = xi->value,
1453 .value_len = xi->value_len,
1454 };
1455
1456 /* Compute min_offs, last and free space. */
1457 last = xs->header->xh_entries;
1458
1459 for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
1460 size_t offs = le16_to_cpu(last->xe_name_offset);
1461 if (offs < min_offs)
1462 min_offs = offs;
1463 last += 1;
1464 }
1465
1466 free = min_offs - ((void *)last - xs->base) - sizeof(__u32);
1467 if (free < 0)
1468 return -EIO;
1469
1470 if (!xs->not_found) {
1471 size_t size = 0;
1472 if (ocfs2_xattr_is_local(xs->here))
1473 size = OCFS2_XATTR_SIZE(name_len) +
1474 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1475 else
1476 size = OCFS2_XATTR_SIZE(name_len) +
1477 OCFS2_XATTR_ROOT_SIZE;
1478 free += (size + sizeof(struct ocfs2_xattr_entry));
1479 }
1480 /* Check free space in inode or block */
1481 if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1482 if (free < sizeof(struct ocfs2_xattr_entry) +
1483 OCFS2_XATTR_SIZE(name_len) +
1484 OCFS2_XATTR_ROOT_SIZE) {
1485 ret = -ENOSPC;
1486 goto out;
1487 }
1488 size_l = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1489 xi_l.value = (void *)&def_xv;
1490 xi_l.value_len = OCFS2_XATTR_ROOT_SIZE;
1491 } else if (xi->value) {
1492 if (free < sizeof(struct ocfs2_xattr_entry) +
1493 OCFS2_XATTR_SIZE(name_len) +
1494 OCFS2_XATTR_SIZE(xi->value_len)) {
1495 ret = -ENOSPC;
1496 goto out;
1497 }
1498 }
1499
1500 if (!xs->not_found) {
1501 /* For existing extended attribute */
1502 size_t size = OCFS2_XATTR_SIZE(name_len) +
1503 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1504 size_t offs = le16_to_cpu(xs->here->xe_name_offset);
1505 void *val = xs->base + offs;
1506
1507 if (ocfs2_xattr_is_local(xs->here) && size == size_l) {
1508 /* Replace existing local xattr with tree root */
1509 ret = ocfs2_xattr_set_value_outside(inode, xi, xs,
1510 ctxt, offs);
1511 if (ret < 0)
1512 mlog_errno(ret);
1513 goto out;
1514 } else if (!ocfs2_xattr_is_local(xs->here)) {
1515 /* For existing xattr which has value outside */
1516 struct ocfs2_xattr_value_root *xv = NULL;
1517 xv = (struct ocfs2_xattr_value_root *)(val +
1518 OCFS2_XATTR_SIZE(name_len));
1519
1520 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1521 /*
1522 * If new value need set outside also,
1523 * first truncate old value to new value,
1524 * then set new value with set_value_outside().
1525 */
1526 ret = ocfs2_xattr_value_truncate(inode,
1527 xs->xattr_bh,
1528 xv,
1529 xi->value_len,
1530 ctxt);
1531 if (ret < 0) {
1532 mlog_errno(ret);
1533 goto out;
1534 }
1535
1536 ret = ocfs2_xattr_update_entry(inode,
1537 handle,
1538 xi,
1539 xs,
1540 offs);
1541 if (ret < 0) {
1542 mlog_errno(ret);
1543 goto out;
1544 }
1545
1546 ret = __ocfs2_xattr_set_value_outside(inode,
1547 handle,
1548 xv,
1549 xi->value,
1550 xi->value_len);
1551 if (ret < 0)
1552 mlog_errno(ret);
1553 goto out;
1554 } else {
1555 /*
1556 * If new value need set in local,
1557 * just trucate old value to zero.
1558 */
1559 ret = ocfs2_xattr_value_truncate(inode,
1560 xs->xattr_bh,
1561 xv,
1562 0,
1563 ctxt);
1564 if (ret < 0)
1565 mlog_errno(ret);
1566 }
1567 }
1568 }
1569
1570 ret = ocfs2_journal_access(handle, inode, xs->inode_bh,
1571 OCFS2_JOURNAL_ACCESS_WRITE);
1572 if (ret) {
1573 mlog_errno(ret);
1574 goto out;
1575 }
1576
1577 if (!(flag & OCFS2_INLINE_XATTR_FL)) {
1578 ret = ocfs2_journal_access(handle, inode, xs->xattr_bh,
1579 OCFS2_JOURNAL_ACCESS_WRITE);
1580 if (ret) {
1581 mlog_errno(ret);
1582 goto out;
1583 }
1584 }
1585
1586 /*
1587 * Set value in local, include set tree root in local.
1588 * This is the first step for value size >INLINE_SIZE.
1589 */
1590 ocfs2_xattr_set_entry_local(inode, &xi_l, xs, last, min_offs);
1591
1592 if (!(flag & OCFS2_INLINE_XATTR_FL)) {
1593 ret = ocfs2_journal_dirty(handle, xs->xattr_bh);
1594 if (ret < 0) {
1595 mlog_errno(ret);
1596 goto out;
1597 }
1598 }
1599
1600 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) &&
1601 (flag & OCFS2_INLINE_XATTR_FL)) {
1602 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1603 unsigned int xattrsize = osb->s_xattr_inline_size;
1604
1605 /*
1606 * Adjust extent record count or inline data size
1607 * to reserve space for extended attribute.
1608 */
1609 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1610 struct ocfs2_inline_data *idata = &di->id2.i_data;
1611 le16_add_cpu(&idata->id_count, -xattrsize);
1612 } else if (!(ocfs2_inode_is_fast_symlink(inode))) {
1613 struct ocfs2_extent_list *el = &di->id2.i_list;
1614 le16_add_cpu(&el->l_count, -(xattrsize /
1615 sizeof(struct ocfs2_extent_rec)));
1616 }
1617 di->i_xattr_inline_size = cpu_to_le16(xattrsize);
1618 }
1619 /* Update xattr flag */
1620 spin_lock(&oi->ip_lock);
1621 oi->ip_dyn_features |= flag;
1622 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
1623 spin_unlock(&oi->ip_lock);
1624 /* Update inode ctime */
1625 inode->i_ctime = CURRENT_TIME;
1626 di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
1627 di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
1628
1629 ret = ocfs2_journal_dirty(handle, xs->inode_bh);
1630 if (ret < 0)
1631 mlog_errno(ret);
1632
1633 if (!ret && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1634 /*
1635 * Set value outside in B tree.
1636 * This is the second step for value size > INLINE_SIZE.
1637 */
1638 size_t offs = le16_to_cpu(xs->here->xe_name_offset);
1639 ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt, offs);
1640 if (ret < 0) {
1641 int ret2;
1642
1643 mlog_errno(ret);
1644 /*
1645 * If set value outside failed, we have to clean
1646 * the junk tree root we have already set in local.
1647 */
1648 ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle,
1649 xi, xs, offs);
1650 if (ret2 < 0)
1651 mlog_errno(ret2);
1652 }
1653 }
1654 out:
1655 return ret;
1656 }
1657
1658 static int ocfs2_remove_value_outside(struct inode*inode,
1659 struct buffer_head *bh,
1660 struct ocfs2_xattr_header *header)
1661 {
1662 int ret = 0, i;
1663 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1664 struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
1665
1666 ocfs2_init_dealloc_ctxt(&ctxt.dealloc);
1667
1668 ctxt.handle = ocfs2_start_trans(osb, OCFS2_REMOVE_EXTENT_CREDITS);
1669 if (IS_ERR(ctxt.handle)) {
1670 ret = PTR_ERR(ctxt.handle);
1671 mlog_errno(ret);
1672 goto out;
1673 }
1674
1675 for (i = 0; i < le16_to_cpu(header->xh_count); i++) {
1676 struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
1677
1678 if (!ocfs2_xattr_is_local(entry)) {
1679 struct ocfs2_xattr_value_root *xv;
1680 void *val;
1681
1682 val = (void *)header +
1683 le16_to_cpu(entry->xe_name_offset);
1684 xv = (struct ocfs2_xattr_value_root *)
1685 (val + OCFS2_XATTR_SIZE(entry->xe_name_len));
1686 ret = ocfs2_xattr_value_truncate(inode, bh, xv,
1687 0, &ctxt);
1688 if (ret < 0) {
1689 mlog_errno(ret);
1690 break;
1691 }
1692 }
1693 }
1694
1695 ocfs2_commit_trans(osb, ctxt.handle);
1696 ocfs2_schedule_truncate_log_flush(osb, 1);
1697 ocfs2_run_deallocs(osb, &ctxt.dealloc);
1698 out:
1699 return ret;
1700 }
1701
1702 static int ocfs2_xattr_ibody_remove(struct inode *inode,
1703 struct buffer_head *di_bh)
1704 {
1705
1706 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1707 struct ocfs2_xattr_header *header;
1708 int ret;
1709
1710 header = (struct ocfs2_xattr_header *)
1711 ((void *)di + inode->i_sb->s_blocksize -
1712 le16_to_cpu(di->i_xattr_inline_size));
1713
1714 ret = ocfs2_remove_value_outside(inode, di_bh, header);
1715
1716 return ret;
1717 }
1718
1719 static int ocfs2_xattr_block_remove(struct inode *inode,
1720 struct buffer_head *blk_bh)
1721 {
1722 struct ocfs2_xattr_block *xb;
1723 int ret = 0;
1724
1725 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
1726 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
1727 struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header);
1728 ret = ocfs2_remove_value_outside(inode, blk_bh, header);
1729 } else
1730 ret = ocfs2_delete_xattr_index_block(inode, blk_bh);
1731
1732 return ret;
1733 }
1734
1735 static int ocfs2_xattr_free_block(struct inode *inode,
1736 u64 block)
1737 {
1738 struct inode *xb_alloc_inode;
1739 struct buffer_head *xb_alloc_bh = NULL;
1740 struct buffer_head *blk_bh = NULL;
1741 struct ocfs2_xattr_block *xb;
1742 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1743 handle_t *handle;
1744 int ret = 0;
1745 u64 blk, bg_blkno;
1746 u16 bit;
1747
1748 ret = ocfs2_read_xattr_block(inode, block, &blk_bh);
1749 if (ret < 0) {
1750 mlog_errno(ret);
1751 goto out;
1752 }
1753
1754 ret = ocfs2_xattr_block_remove(inode, blk_bh);
1755 if (ret < 0) {
1756 mlog_errno(ret);
1757 goto out;
1758 }
1759
1760 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
1761 blk = le64_to_cpu(xb->xb_blkno);
1762 bit = le16_to_cpu(xb->xb_suballoc_bit);
1763 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
1764
1765 xb_alloc_inode = ocfs2_get_system_file_inode(osb,
1766 EXTENT_ALLOC_SYSTEM_INODE,
1767 le16_to_cpu(xb->xb_suballoc_slot));
1768 if (!xb_alloc_inode) {
1769 ret = -ENOMEM;
1770 mlog_errno(ret);
1771 goto out;
1772 }
1773 mutex_lock(&xb_alloc_inode->i_mutex);
1774
1775 ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1);
1776 if (ret < 0) {
1777 mlog_errno(ret);
1778 goto out_mutex;
1779 }
1780
1781 handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE);
1782 if (IS_ERR(handle)) {
1783 ret = PTR_ERR(handle);
1784 mlog_errno(ret);
1785 goto out_unlock;
1786 }
1787
1788 ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh,
1789 bit, bg_blkno, 1);
1790 if (ret < 0)
1791 mlog_errno(ret);
1792
1793 ocfs2_commit_trans(osb, handle);
1794 out_unlock:
1795 ocfs2_inode_unlock(xb_alloc_inode, 1);
1796 brelse(xb_alloc_bh);
1797 out_mutex:
1798 mutex_unlock(&xb_alloc_inode->i_mutex);
1799 iput(xb_alloc_inode);
1800 out:
1801 brelse(blk_bh);
1802 return ret;
1803 }
1804
1805 /*
1806 * ocfs2_xattr_remove()
1807 *
1808 * Free extended attribute resources associated with this inode.
1809 */
1810 int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh)
1811 {
1812 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1813 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1814 handle_t *handle;
1815 int ret;
1816
1817 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
1818 return 0;
1819
1820 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
1821 return 0;
1822
1823 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
1824 ret = ocfs2_xattr_ibody_remove(inode, di_bh);
1825 if (ret < 0) {
1826 mlog_errno(ret);
1827 goto out;
1828 }
1829 }
1830
1831 if (di->i_xattr_loc) {
1832 ret = ocfs2_xattr_free_block(inode,
1833 le64_to_cpu(di->i_xattr_loc));
1834 if (ret < 0) {
1835 mlog_errno(ret);
1836 goto out;
1837 }
1838 }
1839
1840 handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)),
1841 OCFS2_INODE_UPDATE_CREDITS);
1842 if (IS_ERR(handle)) {
1843 ret = PTR_ERR(handle);
1844 mlog_errno(ret);
1845 goto out;
1846 }
1847 ret = ocfs2_journal_access(handle, inode, di_bh,
1848 OCFS2_JOURNAL_ACCESS_WRITE);
1849 if (ret) {
1850 mlog_errno(ret);
1851 goto out_commit;
1852 }
1853
1854 di->i_xattr_loc = 0;
1855
1856 spin_lock(&oi->ip_lock);
1857 oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL);
1858 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
1859 spin_unlock(&oi->ip_lock);
1860
1861 ret = ocfs2_journal_dirty(handle, di_bh);
1862 if (ret < 0)
1863 mlog_errno(ret);
1864 out_commit:
1865 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
1866 out:
1867 return ret;
1868 }
1869
1870 static int ocfs2_xattr_has_space_inline(struct inode *inode,
1871 struct ocfs2_dinode *di)
1872 {
1873 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1874 unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size;
1875 int free;
1876
1877 if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE)
1878 return 0;
1879
1880 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1881 struct ocfs2_inline_data *idata = &di->id2.i_data;
1882 free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size);
1883 } else if (ocfs2_inode_is_fast_symlink(inode)) {
1884 free = ocfs2_fast_symlink_chars(inode->i_sb) -
1885 le64_to_cpu(di->i_size);
1886 } else {
1887 struct ocfs2_extent_list *el = &di->id2.i_list;
1888 free = (le16_to_cpu(el->l_count) -
1889 le16_to_cpu(el->l_next_free_rec)) *
1890 sizeof(struct ocfs2_extent_rec);
1891 }
1892 if (free >= xattrsize)
1893 return 1;
1894
1895 return 0;
1896 }
1897
1898 /*
1899 * ocfs2_xattr_ibody_find()
1900 *
1901 * Find extended attribute in inode block and
1902 * fill search info into struct ocfs2_xattr_search.
1903 */
1904 static int ocfs2_xattr_ibody_find(struct inode *inode,
1905 int name_index,
1906 const char *name,
1907 struct ocfs2_xattr_search *xs)
1908 {
1909 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1910 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1911 int ret;
1912 int has_space = 0;
1913
1914 if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE)
1915 return 0;
1916
1917 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) {
1918 down_read(&oi->ip_alloc_sem);
1919 has_space = ocfs2_xattr_has_space_inline(inode, di);
1920 up_read(&oi->ip_alloc_sem);
1921 if (!has_space)
1922 return 0;
1923 }
1924
1925 xs->xattr_bh = xs->inode_bh;
1926 xs->end = (void *)di + inode->i_sb->s_blocksize;
1927 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)
1928 xs->header = (struct ocfs2_xattr_header *)
1929 (xs->end - le16_to_cpu(di->i_xattr_inline_size));
1930 else
1931 xs->header = (struct ocfs2_xattr_header *)
1932 (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size);
1933 xs->base = (void *)xs->header;
1934 xs->here = xs->header->xh_entries;
1935
1936 /* Find the named attribute. */
1937 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
1938 ret = ocfs2_xattr_find_entry(name_index, name, xs);
1939 if (ret && ret != -ENODATA)
1940 return ret;
1941 xs->not_found = ret;
1942 }
1943
1944 return 0;
1945 }
1946
1947 /*
1948 * ocfs2_xattr_ibody_set()
1949 *
1950 * Set, replace or remove an extended attribute into inode block.
1951 *
1952 */
1953 static int ocfs2_xattr_ibody_set(struct inode *inode,
1954 struct ocfs2_xattr_info *xi,
1955 struct ocfs2_xattr_search *xs,
1956 struct ocfs2_xattr_set_ctxt *ctxt)
1957 {
1958 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1959 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1960 int ret;
1961
1962 if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE)
1963 return -ENOSPC;
1964
1965 down_write(&oi->ip_alloc_sem);
1966 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) {
1967 if (!ocfs2_xattr_has_space_inline(inode, di)) {
1968 ret = -ENOSPC;
1969 goto out;
1970 }
1971 }
1972
1973 ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
1974 (OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL));
1975 out:
1976 up_write(&oi->ip_alloc_sem);
1977
1978 return ret;
1979 }
1980
1981 /*
1982 * ocfs2_xattr_block_find()
1983 *
1984 * Find extended attribute in external block and
1985 * fill search info into struct ocfs2_xattr_search.
1986 */
1987 static int ocfs2_xattr_block_find(struct inode *inode,
1988 int name_index,
1989 const char *name,
1990 struct ocfs2_xattr_search *xs)
1991 {
1992 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1993 struct buffer_head *blk_bh = NULL;
1994 struct ocfs2_xattr_block *xb;
1995 int ret = 0;
1996
1997 if (!di->i_xattr_loc)
1998 return ret;
1999
2000 ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
2001 &blk_bh);
2002 if (ret < 0) {
2003 mlog_errno(ret);
2004 return ret;
2005 }
2006
2007 xs->xattr_bh = blk_bh;
2008 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
2009
2010 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
2011 xs->header = &xb->xb_attrs.xb_header;
2012 xs->base = (void *)xs->header;
2013 xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size;
2014 xs->here = xs->header->xh_entries;
2015
2016 ret = ocfs2_xattr_find_entry(name_index, name, xs);
2017 } else
2018 ret = ocfs2_xattr_index_block_find(inode, blk_bh,
2019 name_index,
2020 name, xs);
2021
2022 if (ret && ret != -ENODATA) {
2023 xs->xattr_bh = NULL;
2024 goto cleanup;
2025 }
2026 xs->not_found = ret;
2027 return 0;
2028 cleanup:
2029 brelse(blk_bh);
2030
2031 return ret;
2032 }
2033
2034 /*
2035 * ocfs2_xattr_block_set()
2036 *
2037 * Set, replace or remove an extended attribute into external block.
2038 *
2039 */
2040 static int ocfs2_xattr_block_set(struct inode *inode,
2041 struct ocfs2_xattr_info *xi,
2042 struct ocfs2_xattr_search *xs,
2043 struct ocfs2_xattr_set_ctxt *ctxt)
2044 {
2045 struct buffer_head *new_bh = NULL;
2046 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2047 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
2048 handle_t *handle = ctxt->handle;
2049 struct ocfs2_xattr_block *xblk = NULL;
2050 u16 suballoc_bit_start;
2051 u32 num_got;
2052 u64 first_blkno;
2053 int ret;
2054
2055 if (!xs->xattr_bh) {
2056 ret = ocfs2_journal_access(handle, inode, xs->inode_bh,
2057 OCFS2_JOURNAL_ACCESS_CREATE);
2058 if (ret < 0) {
2059 mlog_errno(ret);
2060 goto end;
2061 }
2062
2063 ret = ocfs2_claim_metadata(osb, handle, ctxt->meta_ac, 1,
2064 &suballoc_bit_start, &num_got,
2065 &first_blkno);
2066 if (ret < 0) {
2067 mlog_errno(ret);
2068 goto end;
2069 }
2070
2071 new_bh = sb_getblk(inode->i_sb, first_blkno);
2072 ocfs2_set_new_buffer_uptodate(inode, new_bh);
2073
2074 ret = ocfs2_journal_access(handle, inode, new_bh,
2075 OCFS2_JOURNAL_ACCESS_CREATE);
2076 if (ret < 0) {
2077 mlog_errno(ret);
2078 goto end;
2079 }
2080
2081 /* Initialize ocfs2_xattr_block */
2082 xs->xattr_bh = new_bh;
2083 xblk = (struct ocfs2_xattr_block *)new_bh->b_data;
2084 memset(xblk, 0, inode->i_sb->s_blocksize);
2085 strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
2086 xblk->xb_suballoc_slot = cpu_to_le16(osb->slot_num);
2087 xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
2088 xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation);
2089 xblk->xb_blkno = cpu_to_le64(first_blkno);
2090
2091 xs->header = &xblk->xb_attrs.xb_header;
2092 xs->base = (void *)xs->header;
2093 xs->end = (void *)xblk + inode->i_sb->s_blocksize;
2094 xs->here = xs->header->xh_entries;
2095
2096 ret = ocfs2_journal_dirty(handle, new_bh);
2097 if (ret < 0) {
2098 mlog_errno(ret);
2099 goto end;
2100 }
2101 di->i_xattr_loc = cpu_to_le64(first_blkno);
2102 ocfs2_journal_dirty(handle, xs->inode_bh);
2103 } else
2104 xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
2105
2106 if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) {
2107 /* Set extended attribute into external block */
2108 ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
2109 OCFS2_HAS_XATTR_FL);
2110 if (!ret || ret != -ENOSPC)
2111 goto end;
2112
2113 ret = ocfs2_xattr_create_index_block(inode, xs, ctxt);
2114 if (ret)
2115 goto end;
2116 }
2117
2118 ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt);
2119
2120 end:
2121
2122 return ret;
2123 }
2124
2125 /* Check whether the new xattr can be inserted into the inode. */
2126 static int ocfs2_xattr_can_be_in_inode(struct inode *inode,
2127 struct ocfs2_xattr_info *xi,
2128 struct ocfs2_xattr_search *xs)
2129 {
2130 u64 value_size;
2131 struct ocfs2_xattr_entry *last;
2132 int free, i;
2133 size_t min_offs = xs->end - xs->base;
2134
2135 if (!xs->header)
2136 return 0;
2137
2138 last = xs->header->xh_entries;
2139
2140 for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
2141 size_t offs = le16_to_cpu(last->xe_name_offset);
2142 if (offs < min_offs)
2143 min_offs = offs;
2144 last += 1;
2145 }
2146
2147 free = min_offs - ((void *)last - xs->base) - sizeof(__u32);
2148 if (free < 0)
2149 return 0;
2150
2151 BUG_ON(!xs->not_found);
2152
2153 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
2154 value_size = OCFS2_XATTR_ROOT_SIZE;
2155 else
2156 value_size = OCFS2_XATTR_SIZE(xi->value_len);
2157
2158 if (free >= sizeof(struct ocfs2_xattr_entry) +
2159 OCFS2_XATTR_SIZE(strlen(xi->name)) + value_size)
2160 return 1;
2161
2162 return 0;
2163 }
2164
2165 static int ocfs2_calc_xattr_set_need(struct inode *inode,
2166 struct ocfs2_dinode *di,
2167 struct ocfs2_xattr_info *xi,
2168 struct ocfs2_xattr_search *xis,
2169 struct ocfs2_xattr_search *xbs,
2170 int *clusters_need,
2171 int *meta_need,
2172 int *credits_need)
2173 {
2174 int ret = 0, old_in_xb = 0;
2175 int clusters_add = 0, meta_add = 0, credits = 0;
2176 struct buffer_head *bh = NULL;
2177 struct ocfs2_xattr_block *xb = NULL;
2178 struct ocfs2_xattr_entry *xe = NULL;
2179 struct ocfs2_xattr_value_root *xv = NULL;
2180 char *base = NULL;
2181 int name_offset, name_len = 0;
2182 u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
2183 xi->value_len);
2184 u64 value_size;
2185
2186 if (xis->not_found && xbs->not_found) {
2187 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2188
2189 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
2190 clusters_add += new_clusters;
2191 credits += ocfs2_calc_extend_credits(inode->i_sb,
2192 &def_xv.xv.xr_list,
2193 new_clusters);
2194 }
2195
2196 goto meta_guess;
2197 }
2198
2199 if (!xis->not_found) {
2200 xe = xis->here;
2201 name_offset = le16_to_cpu(xe->xe_name_offset);
2202 name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
2203 base = xis->base;
2204 credits += OCFS2_INODE_UPDATE_CREDITS;
2205 } else {
2206 int i, block_off = 0;
2207 xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
2208 xe = xbs->here;
2209 name_offset = le16_to_cpu(xe->xe_name_offset);
2210 name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
2211 i = xbs->here - xbs->header->xh_entries;
2212 old_in_xb = 1;
2213
2214 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
2215 ret = ocfs2_xattr_bucket_get_name_value(inode,
2216 bucket_xh(xbs->bucket),
2217 i, &block_off,
2218 &name_offset);
2219 base = bucket_block(xbs->bucket, block_off);
2220 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2221 } else {
2222 base = xbs->base;
2223 credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS;
2224 }
2225 }
2226
2227 /*
2228 * delete a xattr doesn't need metadata and cluster allocation.
2229 * so just calculate the credits and return.
2230 *
2231 * The credits for removing the value tree will be extended
2232 * by ocfs2_remove_extent itself.
2233 */
2234 if (!xi->value) {
2235 if (!ocfs2_xattr_is_local(xe))
2236 credits += OCFS2_REMOVE_EXTENT_CREDITS;
2237
2238 goto out;
2239 }
2240
2241 /* do cluster allocation guess first. */
2242 value_size = le64_to_cpu(xe->xe_value_size);
2243
2244 if (old_in_xb) {
2245 /*
2246 * In xattr set, we always try to set the xe in inode first,
2247 * so if it can be inserted into inode successfully, the old
2248 * one will be removed from the xattr block, and this xattr
2249 * will be inserted into inode as a new xattr in inode.
2250 */
2251 if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) {
2252 clusters_add += new_clusters;
2253 credits += OCFS2_REMOVE_EXTENT_CREDITS +
2254 OCFS2_INODE_UPDATE_CREDITS;
2255 if (!ocfs2_xattr_is_local(xe))
2256 credits += ocfs2_calc_extend_credits(
2257 inode->i_sb,
2258 &def_xv.xv.xr_list,
2259 new_clusters);
2260 goto out;
2261 }
2262 }
2263
2264 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
2265 /* the new values will be stored outside. */
2266 u32 old_clusters = 0;
2267
2268 if (!ocfs2_xattr_is_local(xe)) {
2269 old_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
2270 value_size);
2271 xv = (struct ocfs2_xattr_value_root *)
2272 (base + name_offset + name_len);
2273 value_size = OCFS2_XATTR_ROOT_SIZE;
2274 } else
2275 xv = &def_xv.xv;
2276
2277 if (old_clusters >= new_clusters) {
2278 credits += OCFS2_REMOVE_EXTENT_CREDITS;
2279 goto out;
2280 } else {
2281 meta_add += ocfs2_extend_meta_needed(&xv->xr_list);
2282 clusters_add += new_clusters - old_clusters;
2283 credits += ocfs2_calc_extend_credits(inode->i_sb,
2284 &xv->xr_list,
2285 new_clusters -
2286 old_clusters);
2287 if (value_size >= OCFS2_XATTR_ROOT_SIZE)
2288 goto out;
2289 }
2290 } else {
2291 /*
2292 * Now the new value will be stored inside. So if the new
2293 * value is smaller than the size of value root or the old
2294 * value, we don't need any allocation, otherwise we have
2295 * to guess metadata allocation.
2296 */
2297 if ((ocfs2_xattr_is_local(xe) && value_size >= xi->value_len) ||
2298 (!ocfs2_xattr_is_local(xe) &&
2299 OCFS2_XATTR_ROOT_SIZE >= xi->value_len))
2300 goto out;
2301 }
2302
2303 meta_guess:
2304 /* calculate metadata allocation. */
2305 if (di->i_xattr_loc) {
2306 if (!xbs->xattr_bh) {
2307 ret = ocfs2_read_xattr_block(inode,
2308 le64_to_cpu(di->i_xattr_loc),
2309 &bh);
2310 if (ret) {
2311 mlog_errno(ret);
2312 goto out;
2313 }
2314
2315 xb = (struct ocfs2_xattr_block *)bh->b_data;
2316 } else
2317 xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
2318
2319 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
2320 struct ocfs2_extent_list *el =
2321 &xb->xb_attrs.xb_root.xt_list;
2322 meta_add += ocfs2_extend_meta_needed(el);
2323 credits += ocfs2_calc_extend_credits(inode->i_sb,
2324 el, 1);
2325 }
2326
2327 /*
2328 * This cluster will be used either for new bucket or for
2329 * new xattr block.
2330 * If the cluster size is the same as the bucket size, one
2331 * more is needed since we may need to extend the bucket
2332 * also.
2333 */
2334 clusters_add += 1;
2335 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2336 if (OCFS2_XATTR_BUCKET_SIZE ==
2337 OCFS2_SB(inode->i_sb)->s_clustersize) {
2338 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2339 clusters_add += 1;
2340 }
2341 } else {
2342 meta_add += 1;
2343 credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
2344 }
2345 out:
2346 if (clusters_need)
2347 *clusters_need = clusters_add;
2348 if (meta_need)
2349 *meta_need = meta_add;
2350 if (credits_need)
2351 *credits_need = credits;
2352 brelse(bh);
2353 return ret;
2354 }
2355
2356 static int ocfs2_init_xattr_set_ctxt(struct inode *inode,
2357 struct ocfs2_dinode *di,
2358 struct ocfs2_xattr_info *xi,
2359 struct ocfs2_xattr_search *xis,
2360 struct ocfs2_xattr_search *xbs,
2361 struct ocfs2_xattr_set_ctxt *ctxt,
2362 int *credits)
2363 {
2364 int clusters_add, meta_add, ret;
2365 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2366
2367 memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt));
2368
2369 ocfs2_init_dealloc_ctxt(&ctxt->dealloc);
2370
2371 ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs,
2372 &clusters_add, &meta_add, credits);
2373 if (ret) {
2374 mlog_errno(ret);
2375 return ret;
2376 }
2377
2378 mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, "
2379 "credits = %d\n", xi->name, meta_add, clusters_add, *credits);
2380
2381 if (meta_add) {
2382 ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add,
2383 &ctxt->meta_ac);
2384 if (ret) {
2385 mlog_errno(ret);
2386 goto out;
2387 }
2388 }
2389
2390 if (clusters_add) {
2391 ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac);
2392 if (ret)
2393 mlog_errno(ret);
2394 }
2395 out:
2396 if (ret) {
2397 if (ctxt->meta_ac) {
2398 ocfs2_free_alloc_context(ctxt->meta_ac);
2399 ctxt->meta_ac = NULL;
2400 }
2401
2402 /*
2403 * We cannot have an error and a non null ctxt->data_ac.
2404 */
2405 }
2406
2407 return ret;
2408 }
2409
2410 static int __ocfs2_xattr_set_handle(struct inode *inode,
2411 struct ocfs2_dinode *di,
2412 struct ocfs2_xattr_info *xi,
2413 struct ocfs2_xattr_search *xis,
2414 struct ocfs2_xattr_search *xbs,
2415 struct ocfs2_xattr_set_ctxt *ctxt)
2416 {
2417 int ret = 0, credits, old_found;
2418
2419 if (!xi->value) {
2420 /* Remove existing extended attribute */
2421 if (!xis->not_found)
2422 ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt);
2423 else if (!xbs->not_found)
2424 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2425 } else {
2426 /* We always try to set extended attribute into inode first*/
2427 ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt);
2428 if (!ret && !xbs->not_found) {
2429 /*
2430 * If succeed and that extended attribute existing in
2431 * external block, then we will remove it.
2432 */
2433 xi->value = NULL;
2434 xi->value_len = 0;
2435
2436 old_found = xis->not_found;
2437 xis->not_found = -ENODATA;
2438 ret = ocfs2_calc_xattr_set_need(inode,
2439 di,
2440 xi,
2441 xis,
2442 xbs,
2443 NULL,
2444 NULL,
2445 &credits);
2446 xis->not_found = old_found;
2447 if (ret) {
2448 mlog_errno(ret);
2449 goto out;
2450 }
2451
2452 ret = ocfs2_extend_trans(ctxt->handle, credits +
2453 ctxt->handle->h_buffer_credits);
2454 if (ret) {
2455 mlog_errno(ret);
2456 goto out;
2457 }
2458 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2459 } else if (ret == -ENOSPC) {
2460 if (di->i_xattr_loc && !xbs->xattr_bh) {
2461 ret = ocfs2_xattr_block_find(inode,
2462 xi->name_index,
2463 xi->name, xbs);
2464 if (ret)
2465 goto out;
2466
2467 old_found = xis->not_found;
2468 xis->not_found = -ENODATA;
2469 ret = ocfs2_calc_xattr_set_need(inode,
2470 di,
2471 xi,
2472 xis,
2473 xbs,
2474 NULL,
2475 NULL,
2476 &credits);
2477 xis->not_found = old_found;
2478 if (ret) {
2479 mlog_errno(ret);
2480 goto out;
2481 }
2482
2483 ret = ocfs2_extend_trans(ctxt->handle, credits +
2484 ctxt->handle->h_buffer_credits);
2485 if (ret) {
2486 mlog_errno(ret);
2487 goto out;
2488 }
2489 }
2490 /*
2491 * If no space in inode, we will set extended attribute
2492 * into external block.
2493 */
2494 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2495 if (ret)
2496 goto out;
2497 if (!xis->not_found) {
2498 /*
2499 * If succeed and that extended attribute
2500 * existing in inode, we will remove it.
2501 */
2502 xi->value = NULL;
2503 xi->value_len = 0;
2504 xbs->not_found = -ENODATA;
2505 ret = ocfs2_calc_xattr_set_need(inode,
2506 di,
2507 xi,
2508 xis,
2509 xbs,
2510 NULL,
2511 NULL,
2512 &credits);
2513 if (ret) {
2514 mlog_errno(ret);
2515 goto out;
2516 }
2517
2518 ret = ocfs2_extend_trans(ctxt->handle, credits +
2519 ctxt->handle->h_buffer_credits);
2520 if (ret) {
2521 mlog_errno(ret);
2522 goto out;
2523 }
2524 ret = ocfs2_xattr_ibody_set(inode, xi,
2525 xis, ctxt);
2526 }
2527 }
2528 }
2529
2530 out:
2531 return ret;
2532 }
2533
2534 /*
2535 * This function only called duing creating inode
2536 * for init security/acl xattrs of the new inode.
2537 * The xattrs could be put into ibody or extent block,
2538 * xattr bucket would not be use in this case.
2539 * transanction credits also be reserved in here.
2540 */
2541 int ocfs2_xattr_set_handle(handle_t *handle,
2542 struct inode *inode,
2543 struct buffer_head *di_bh,
2544 int name_index,
2545 const char *name,
2546 const void *value,
2547 size_t value_len,
2548 int flags,
2549 struct ocfs2_alloc_context *meta_ac,
2550 struct ocfs2_alloc_context *data_ac)
2551 {
2552 struct ocfs2_dinode *di;
2553 int ret;
2554
2555 struct ocfs2_xattr_info xi = {
2556 .name_index = name_index,
2557 .name = name,
2558 .value = value,
2559 .value_len = value_len,
2560 };
2561
2562 struct ocfs2_xattr_search xis = {
2563 .not_found = -ENODATA,
2564 };
2565
2566 struct ocfs2_xattr_search xbs = {
2567 .not_found = -ENODATA,
2568 };
2569
2570 struct ocfs2_xattr_set_ctxt ctxt = {
2571 .handle = handle,
2572 .meta_ac = meta_ac,
2573 .data_ac = data_ac,
2574 };
2575
2576 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
2577 return -EOPNOTSUPP;
2578
2579 xis.inode_bh = xbs.inode_bh = di_bh;
2580 di = (struct ocfs2_dinode *)di_bh->b_data;
2581
2582 down_write(&OCFS2_I(inode)->ip_xattr_sem);
2583
2584 ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis);
2585 if (ret)
2586 goto cleanup;
2587 if (xis.not_found) {
2588 ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs);
2589 if (ret)
2590 goto cleanup;
2591 }
2592
2593 ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
2594
2595 cleanup:
2596 up_write(&OCFS2_I(inode)->ip_xattr_sem);
2597 brelse(xbs.xattr_bh);
2598
2599 return ret;
2600 }
2601
2602 /*
2603 * ocfs2_xattr_set()
2604 *
2605 * Set, replace or remove an extended attribute for this inode.
2606 * value is NULL to remove an existing extended attribute, else either
2607 * create or replace an extended attribute.
2608 */
2609 int ocfs2_xattr_set(struct inode *inode,
2610 int name_index,
2611 const char *name,
2612 const void *value,
2613 size_t value_len,
2614 int flags)
2615 {
2616 struct buffer_head *di_bh = NULL;
2617 struct ocfs2_dinode *di;
2618 int ret, credits;
2619 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2620 struct inode *tl_inode = osb->osb_tl_inode;
2621 struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
2622
2623 struct ocfs2_xattr_info xi = {
2624 .name_index = name_index,
2625 .name = name,
2626 .value = value,
2627 .value_len = value_len,
2628 };
2629
2630 struct ocfs2_xattr_search xis = {
2631 .not_found = -ENODATA,
2632 };
2633
2634 struct ocfs2_xattr_search xbs = {
2635 .not_found = -ENODATA,
2636 };
2637
2638 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
2639 return -EOPNOTSUPP;
2640
2641 /*
2642 * Only xbs will be used on indexed trees. xis doesn't need a
2643 * bucket.
2644 */
2645 xbs.bucket = ocfs2_xattr_bucket_new(inode);
2646 if (!xbs.bucket) {
2647 mlog_errno(-ENOMEM);
2648 return -ENOMEM;
2649 }
2650
2651 ret = ocfs2_inode_lock(inode, &di_bh, 1);
2652 if (ret < 0) {
2653 mlog_errno(ret);
2654 goto cleanup_nolock;
2655 }
2656 xis.inode_bh = xbs.inode_bh = di_bh;
2657 di = (struct ocfs2_dinode *)di_bh->b_data;
2658
2659 down_write(&OCFS2_I(inode)->ip_xattr_sem);
2660 /*
2661 * Scan inode and external block to find the same name
2662 * extended attribute and collect search infomation.
2663 */
2664 ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis);
2665 if (ret)
2666 goto cleanup;
2667 if (xis.not_found) {
2668 ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs);
2669 if (ret)
2670 goto cleanup;
2671 }
2672
2673 if (xis.not_found && xbs.not_found) {
2674 ret = -ENODATA;
2675 if (flags & XATTR_REPLACE)
2676 goto cleanup;
2677 ret = 0;
2678 if (!value)
2679 goto cleanup;
2680 } else {
2681 ret = -EEXIST;
2682 if (flags & XATTR_CREATE)
2683 goto cleanup;
2684 }
2685
2686
2687 mutex_lock(&tl_inode->i_mutex);
2688
2689 if (ocfs2_truncate_log_needs_flush(osb)) {
2690 ret = __ocfs2_flush_truncate_log(osb);
2691 if (ret < 0) {
2692 mutex_unlock(&tl_inode->i_mutex);
2693 mlog_errno(ret);
2694 goto cleanup;
2695 }
2696 }
2697 mutex_unlock(&tl_inode->i_mutex);
2698
2699 ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis,
2700 &xbs, &ctxt, &credits);
2701 if (ret) {
2702 mlog_errno(ret);
2703 goto cleanup;
2704 }
2705
2706 ctxt.handle = ocfs2_start_trans(osb, credits);
2707 if (IS_ERR(ctxt.handle)) {
2708 ret = PTR_ERR(ctxt.handle);
2709 mlog_errno(ret);
2710 goto cleanup;
2711 }
2712
2713 ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
2714
2715 ocfs2_commit_trans(osb, ctxt.handle);
2716
2717 if (ctxt.data_ac)
2718 ocfs2_free_alloc_context(ctxt.data_ac);
2719 if (ctxt.meta_ac)
2720 ocfs2_free_alloc_context(ctxt.meta_ac);
2721 if (ocfs2_dealloc_has_cluster(&ctxt.dealloc))
2722 ocfs2_schedule_truncate_log_flush(osb, 1);
2723 ocfs2_run_deallocs(osb, &ctxt.dealloc);
2724 cleanup:
2725 up_write(&OCFS2_I(inode)->ip_xattr_sem);
2726 ocfs2_inode_unlock(inode, 1);
2727 cleanup_nolock:
2728 brelse(di_bh);
2729 brelse(xbs.xattr_bh);
2730 ocfs2_xattr_bucket_free(xbs.bucket);
2731
2732 return ret;
2733 }
2734
2735 /*
2736 * Find the xattr extent rec which may contains name_hash.
2737 * e_cpos will be the first name hash of the xattr rec.
2738 * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list.
2739 */
2740 static int ocfs2_xattr_get_rec(struct inode *inode,
2741 u32 name_hash,
2742 u64 *p_blkno,
2743 u32 *e_cpos,
2744 u32 *num_clusters,
2745 struct ocfs2_extent_list *el)
2746 {
2747 int ret = 0, i;
2748 struct buffer_head *eb_bh = NULL;
2749 struct ocfs2_extent_block *eb;
2750 struct ocfs2_extent_rec *rec = NULL;
2751 u64 e_blkno = 0;
2752
2753 if (el->l_tree_depth) {
2754 ret = ocfs2_find_leaf(inode, el, name_hash, &eb_bh);
2755 if (ret) {
2756 mlog_errno(ret);
2757 goto out;
2758 }
2759
2760 eb = (struct ocfs2_extent_block *) eb_bh->b_data;
2761 el = &eb->h_list;
2762
2763 if (el->l_tree_depth) {
2764 ocfs2_error(inode->i_sb,
2765 "Inode %lu has non zero tree depth in "
2766 "xattr tree block %llu\n", inode->i_ino,
2767 (unsigned long long)eb_bh->b_blocknr);
2768 ret = -EROFS;
2769 goto out;
2770 }
2771 }
2772
2773 for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
2774 rec = &el->l_recs[i];
2775
2776 if (le32_to_cpu(rec->e_cpos) <= name_hash) {
2777 e_blkno = le64_to_cpu(rec->e_blkno);
2778 break;
2779 }
2780 }
2781
2782 if (!e_blkno) {
2783 ocfs2_error(inode->i_sb, "Inode %lu has bad extent "
2784 "record (%u, %u, 0) in xattr", inode->i_ino,
2785 le32_to_cpu(rec->e_cpos),
2786 ocfs2_rec_clusters(el, rec));
2787 ret = -EROFS;
2788 goto out;
2789 }
2790
2791 *p_blkno = le64_to_cpu(rec->e_blkno);
2792 *num_clusters = le16_to_cpu(rec->e_leaf_clusters);
2793 if (e_cpos)
2794 *e_cpos = le32_to_cpu(rec->e_cpos);
2795 out:
2796 brelse(eb_bh);
2797 return ret;
2798 }
2799
2800 typedef int (xattr_bucket_func)(struct inode *inode,
2801 struct ocfs2_xattr_bucket *bucket,
2802 void *para);
2803
2804 static int ocfs2_find_xe_in_bucket(struct inode *inode,
2805 struct ocfs2_xattr_bucket *bucket,
2806 int name_index,
2807 const char *name,
2808 u32 name_hash,
2809 u16 *xe_index,
2810 int *found)
2811 {
2812 int i, ret = 0, cmp = 1, block_off, new_offset;
2813 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
2814 size_t name_len = strlen(name);
2815 struct ocfs2_xattr_entry *xe = NULL;
2816 char *xe_name;
2817
2818 /*
2819 * We don't use binary search in the bucket because there
2820 * may be multiple entries with the same name hash.
2821 */
2822 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
2823 xe = &xh->xh_entries[i];
2824
2825 if (name_hash > le32_to_cpu(xe->xe_name_hash))
2826 continue;
2827 else if (name_hash < le32_to_cpu(xe->xe_name_hash))
2828 break;
2829
2830 cmp = name_index - ocfs2_xattr_get_type(xe);
2831 if (!cmp)
2832 cmp = name_len - xe->xe_name_len;
2833 if (cmp)
2834 continue;
2835
2836 ret = ocfs2_xattr_bucket_get_name_value(inode,
2837 xh,
2838 i,
2839 &block_off,
2840 &new_offset);
2841 if (ret) {
2842 mlog_errno(ret);
2843 break;
2844 }
2845
2846
2847 xe_name = bucket_block(bucket, block_off) + new_offset;
2848 if (!memcmp(name, xe_name, name_len)) {
2849 *xe_index = i;
2850 *found = 1;
2851 ret = 0;
2852 break;
2853 }
2854 }
2855
2856 return ret;
2857 }
2858
2859 /*
2860 * Find the specified xattr entry in a series of buckets.
2861 * This series start from p_blkno and last for num_clusters.
2862 * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains
2863 * the num of the valid buckets.
2864 *
2865 * Return the buffer_head this xattr should reside in. And if the xattr's
2866 * hash is in the gap of 2 buckets, return the lower bucket.
2867 */
2868 static int ocfs2_xattr_bucket_find(struct inode *inode,
2869 int name_index,
2870 const char *name,
2871 u32 name_hash,
2872 u64 p_blkno,
2873 u32 first_hash,
2874 u32 num_clusters,
2875 struct ocfs2_xattr_search *xs)
2876 {
2877 int ret, found = 0;
2878 struct ocfs2_xattr_header *xh = NULL;
2879 struct ocfs2_xattr_entry *xe = NULL;
2880 u16 index = 0;
2881 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2882 int low_bucket = 0, bucket, high_bucket;
2883 struct ocfs2_xattr_bucket *search;
2884 u32 last_hash;
2885 u64 blkno, lower_blkno = 0;
2886
2887 search = ocfs2_xattr_bucket_new(inode);
2888 if (!search) {
2889 ret = -ENOMEM;
2890 mlog_errno(ret);
2891 goto out;
2892 }
2893
2894 ret = ocfs2_read_xattr_bucket(search, p_blkno);
2895 if (ret) {
2896 mlog_errno(ret);
2897 goto out;
2898 }
2899
2900 xh = bucket_xh(search);
2901 high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1;
2902 while (low_bucket <= high_bucket) {
2903 ocfs2_xattr_bucket_relse(search);
2904
2905 bucket = (low_bucket + high_bucket) / 2;
2906 blkno = p_blkno + bucket * blk_per_bucket;
2907 ret = ocfs2_read_xattr_bucket(search, blkno);
2908 if (ret) {
2909 mlog_errno(ret);
2910 goto out;
2911 }
2912
2913 xh = bucket_xh(search);
2914 xe = &xh->xh_entries[0];
2915 if (name_hash < le32_to_cpu(xe->xe_name_hash)) {
2916 high_bucket = bucket - 1;
2917 continue;
2918 }
2919
2920 /*
2921 * Check whether the hash of the last entry in our
2922 * bucket is larger than the search one. for an empty
2923 * bucket, the last one is also the first one.
2924 */
2925 if (xh->xh_count)
2926 xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1];
2927
2928 last_hash = le32_to_cpu(xe->xe_name_hash);
2929
2930 /* record lower_blkno which may be the insert place. */
2931 lower_blkno = blkno;
2932
2933 if (name_hash > le32_to_cpu(xe->xe_name_hash)) {
2934 low_bucket = bucket + 1;
2935 continue;
2936 }
2937
2938 /* the searched xattr should reside in this bucket if exists. */
2939 ret = ocfs2_find_xe_in_bucket(inode, search,
2940 name_index, name, name_hash,
2941 &index, &found);
2942 if (ret) {
2943 mlog_errno(ret);
2944 goto out;
2945 }
2946 break;
2947 }
2948
2949 /*
2950 * Record the bucket we have found.
2951 * When the xattr's hash value is in the gap of 2 buckets, we will
2952 * always set it to the previous bucket.
2953 */
2954 if (!lower_blkno)
2955 lower_blkno = p_blkno;
2956
2957 /* This should be in cache - we just read it during the search */
2958 ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno);
2959 if (ret) {
2960 mlog_errno(ret);
2961 goto out;
2962 }
2963
2964 xs->header = bucket_xh(xs->bucket);
2965 xs->base = bucket_block(xs->bucket, 0);
2966 xs->end = xs->base + inode->i_sb->s_blocksize;
2967
2968 if (found) {
2969 xs->here = &xs->header->xh_entries[index];
2970 mlog(0, "find xattr %s in bucket %llu, entry = %u\n", name,
2971 (unsigned long long)bucket_blkno(xs->bucket), index);
2972 } else
2973 ret = -ENODATA;
2974
2975 out:
2976 ocfs2_xattr_bucket_free(search);
2977 return ret;
2978 }
2979
2980 static int ocfs2_xattr_index_block_find(struct inode *inode,
2981 struct buffer_head *root_bh,
2982 int name_index,
2983 const char *name,
2984 struct ocfs2_xattr_search *xs)
2985 {
2986 int ret;
2987 struct ocfs2_xattr_block *xb =
2988 (struct ocfs2_xattr_block *)root_bh->b_data;
2989 struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root;
2990 struct ocfs2_extent_list *el = &xb_root->xt_list;
2991 u64 p_blkno = 0;
2992 u32 first_hash, num_clusters = 0;
2993 u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name));
2994
2995 if (le16_to_cpu(el->l_next_free_rec) == 0)
2996 return -ENODATA;
2997
2998 mlog(0, "find xattr %s, hash = %u, index = %d in xattr tree\n",
2999 name, name_hash, name_index);
3000
3001 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash,
3002 &num_clusters, el);
3003 if (ret) {
3004 mlog_errno(ret);
3005 goto out;
3006 }
3007
3008 BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash);
3009
3010 mlog(0, "find xattr extent rec %u clusters from %llu, the first hash "
3011 "in the rec is %u\n", num_clusters, (unsigned long long)p_blkno,
3012 first_hash);
3013
3014 ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash,
3015 p_blkno, first_hash, num_clusters, xs);
3016
3017 out:
3018 return ret;
3019 }
3020
3021 static int ocfs2_iterate_xattr_buckets(struct inode *inode,
3022 u64 blkno,
3023 u32 clusters,
3024 xattr_bucket_func *func,
3025 void *para)
3026 {
3027 int i, ret = 0;
3028 u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb));
3029 u32 num_buckets = clusters * bpc;
3030 struct ocfs2_xattr_bucket *bucket;
3031
3032 bucket = ocfs2_xattr_bucket_new(inode);
3033 if (!bucket) {
3034 mlog_errno(-ENOMEM);
3035 return -ENOMEM;
3036 }
3037
3038 mlog(0, "iterating xattr buckets in %u clusters starting from %llu\n",
3039 clusters, (unsigned long long)blkno);
3040
3041 for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) {
3042 ret = ocfs2_read_xattr_bucket(bucket, blkno);
3043 if (ret) {
3044 mlog_errno(ret);
3045 break;
3046 }
3047
3048 /*
3049 * The real bucket num in this series of blocks is stored
3050 * in the 1st bucket.
3051 */
3052 if (i == 0)
3053 num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets);
3054
3055 mlog(0, "iterating xattr bucket %llu, first hash %u\n",
3056 (unsigned long long)blkno,
3057 le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash));
3058 if (func) {
3059 ret = func(inode, bucket, para);
3060 if (ret)
3061 mlog_errno(ret);
3062 /* Fall through to bucket_relse() */
3063 }
3064
3065 ocfs2_xattr_bucket_relse(bucket);
3066 if (ret)
3067 break;
3068 }
3069
3070 ocfs2_xattr_bucket_free(bucket);
3071 return ret;
3072 }
3073
3074 struct ocfs2_xattr_tree_list {
3075 char *buffer;
3076 size_t buffer_size;
3077 size_t result;
3078 };
3079
3080 static int ocfs2_xattr_bucket_get_name_value(struct inode *inode,
3081 struct ocfs2_xattr_header *xh,
3082 int index,
3083 int *block_off,
3084 int *new_offset)
3085 {
3086 u16 name_offset;
3087
3088 if (index < 0 || index >= le16_to_cpu(xh->xh_count))
3089 return -EINVAL;
3090
3091 name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset);
3092
3093 *block_off = name_offset >> inode->i_sb->s_blocksize_bits;
3094 *new_offset = name_offset % inode->i_sb->s_blocksize;
3095
3096 return 0;
3097 }
3098
3099 static int ocfs2_list_xattr_bucket(struct inode *inode,
3100 struct ocfs2_xattr_bucket *bucket,
3101 void *para)
3102 {
3103 int ret = 0, type;
3104 struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para;
3105 int i, block_off, new_offset;
3106 const char *prefix, *name;
3107
3108 for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) {
3109 struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i];
3110 type = ocfs2_xattr_get_type(entry);
3111 prefix = ocfs2_xattr_prefix(type);
3112
3113 if (prefix) {
3114 ret = ocfs2_xattr_bucket_get_name_value(inode,
3115 bucket_xh(bucket),
3116 i,
3117 &block_off,
3118 &new_offset);
3119 if (ret)
3120 break;
3121
3122 name = (const char *)bucket_block(bucket, block_off) +
3123 new_offset;
3124 ret = ocfs2_xattr_list_entry(xl->buffer,
3125 xl->buffer_size,
3126 &xl->result,
3127 prefix, name,
3128 entry->xe_name_len);
3129 if (ret)
3130 break;
3131 }
3132 }
3133
3134 return ret;
3135 }
3136
3137 static int ocfs2_xattr_tree_list_index_block(struct inode *inode,
3138 struct ocfs2_xattr_tree_root *xt,
3139 char *buffer,
3140 size_t buffer_size)
3141 {
3142 struct ocfs2_extent_list *el = &xt->xt_list;
3143 int ret = 0;
3144 u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0;
3145 u64 p_blkno = 0;
3146 struct ocfs2_xattr_tree_list xl = {
3147 .buffer = buffer,
3148 .buffer_size = buffer_size,
3149 .result = 0,
3150 };
3151
3152 if (le16_to_cpu(el->l_next_free_rec) == 0)
3153 return 0;
3154
3155 while (name_hash > 0) {
3156 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno,
3157 &e_cpos, &num_clusters, el);
3158 if (ret) {
3159 mlog_errno(ret);
3160 goto out;
3161 }
3162
3163 ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters,
3164 ocfs2_list_xattr_bucket,
3165 &xl);
3166 if (ret) {
3167 mlog_errno(ret);
3168 goto out;
3169 }
3170
3171 if (e_cpos == 0)
3172 break;
3173
3174 name_hash = e_cpos - 1;
3175 }
3176
3177 ret = xl.result;
3178 out:
3179 return ret;
3180 }
3181
3182 static int cmp_xe(const void *a, const void *b)
3183 {
3184 const struct ocfs2_xattr_entry *l = a, *r = b;
3185 u32 l_hash = le32_to_cpu(l->xe_name_hash);
3186 u32 r_hash = le32_to_cpu(r->xe_name_hash);
3187
3188 if (l_hash > r_hash)
3189 return 1;
3190 if (l_hash < r_hash)
3191 return -1;
3192 return 0;
3193 }
3194
3195 static void swap_xe(void *a, void *b, int size)
3196 {
3197 struct ocfs2_xattr_entry *l = a, *r = b, tmp;
3198
3199 tmp = *l;
3200 memcpy(l, r, sizeof(struct ocfs2_xattr_entry));
3201 memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry));
3202 }
3203
3204 /*
3205 * When the ocfs2_xattr_block is filled up, new bucket will be created
3206 * and all the xattr entries will be moved to the new bucket.
3207 * The header goes at the start of the bucket, and the names+values are
3208 * filled from the end. This is why *target starts as the last buffer.
3209 * Note: we need to sort the entries since they are not saved in order
3210 * in the ocfs2_xattr_block.
3211 */
3212 static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode,
3213 struct buffer_head *xb_bh,
3214 struct ocfs2_xattr_bucket *bucket)
3215 {
3216 int i, blocksize = inode->i_sb->s_blocksize;
3217 int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
3218 u16 offset, size, off_change;
3219 struct ocfs2_xattr_entry *xe;
3220 struct ocfs2_xattr_block *xb =
3221 (struct ocfs2_xattr_block *)xb_bh->b_data;
3222 struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header;
3223 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
3224 u16 count = le16_to_cpu(xb_xh->xh_count);
3225 char *src = xb_bh->b_data;
3226 char *target = bucket_block(bucket, blks - 1);
3227
3228 mlog(0, "cp xattr from block %llu to bucket %llu\n",
3229 (unsigned long long)xb_bh->b_blocknr,
3230 (unsigned long long)bucket_blkno(bucket));
3231
3232 for (i = 0; i < blks; i++)
3233 memset(bucket_block(bucket, i), 0, blocksize);
3234
3235 /*
3236 * Since the xe_name_offset is based on ocfs2_xattr_header,
3237 * there is a offset change corresponding to the change of
3238 * ocfs2_xattr_header's position.
3239 */
3240 off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header);
3241 xe = &xb_xh->xh_entries[count - 1];
3242 offset = le16_to_cpu(xe->xe_name_offset) + off_change;
3243 size = blocksize - offset;
3244
3245 /* copy all the names and values. */
3246 memcpy(target + offset, src + offset, size);
3247
3248 /* Init new header now. */
3249 xh->xh_count = xb_xh->xh_count;
3250 xh->xh_num_buckets = cpu_to_le16(1);
3251 xh->xh_name_value_len = cpu_to_le16(size);
3252 xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size);
3253
3254 /* copy all the entries. */
3255 target = bucket_block(bucket, 0);
3256 offset = offsetof(struct ocfs2_xattr_header, xh_entries);
3257 size = count * sizeof(struct ocfs2_xattr_entry);
3258 memcpy(target + offset, (char *)xb_xh + offset, size);
3259
3260 /* Change the xe offset for all the xe because of the move. */
3261 off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize +
3262 offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header);
3263 for (i = 0; i < count; i++)
3264 le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change);
3265
3266 mlog(0, "copy entry: start = %u, size = %u, offset_change = %u\n",
3267 offset, size, off_change);
3268
3269 sort(target + offset, count, sizeof(struct ocfs2_xattr_entry),
3270 cmp_xe, swap_xe);
3271 }
3272
3273 /*
3274 * After we move xattr from block to index btree, we have to
3275 * update ocfs2_xattr_search to the new xe and base.
3276 *
3277 * When the entry is in xattr block, xattr_bh indicates the storage place.
3278 * While if the entry is in index b-tree, "bucket" indicates the
3279 * real place of the xattr.
3280 */
3281 static void ocfs2_xattr_update_xattr_search(struct inode *inode,
3282 struct ocfs2_xattr_search *xs,
3283 struct buffer_head *old_bh)
3284 {
3285 char *buf = old_bh->b_data;
3286 struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf;
3287 struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header;
3288 int i;
3289
3290 xs->header = bucket_xh(xs->bucket);
3291 xs->base = bucket_block(xs->bucket, 0);
3292 xs->end = xs->base + inode->i_sb->s_blocksize;
3293
3294 if (xs->not_found)
3295 return;
3296
3297 i = xs->here - old_xh->xh_entries;
3298 xs->here = &xs->header->xh_entries[i];
3299 }
3300
3301 static int ocfs2_xattr_create_index_block(struct inode *inode,
3302 struct ocfs2_xattr_search *xs,
3303 struct ocfs2_xattr_set_ctxt *ctxt)
3304 {
3305 int ret;
3306 u32 bit_off, len;
3307 u64 blkno;
3308 handle_t *handle = ctxt->handle;
3309 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
3310 struct ocfs2_inode_info *oi = OCFS2_I(inode);
3311 struct buffer_head *xb_bh = xs->xattr_bh;
3312 struct ocfs2_xattr_block *xb =
3313 (struct ocfs2_xattr_block *)xb_bh->b_data;
3314 struct ocfs2_xattr_tree_root *xr;
3315 u16 xb_flags = le16_to_cpu(xb->xb_flags);
3316
3317 mlog(0, "create xattr index block for %llu\n",
3318 (unsigned long long)xb_bh->b_blocknr);
3319
3320 BUG_ON(xb_flags & OCFS2_XATTR_INDEXED);
3321 BUG_ON(!xs->bucket);
3322
3323 /*
3324 * XXX:
3325 * We can use this lock for now, and maybe move to a dedicated mutex
3326 * if performance becomes a problem later.
3327 */
3328 down_write(&oi->ip_alloc_sem);
3329
3330 ret = ocfs2_journal_access(handle, inode, xb_bh,
3331 OCFS2_JOURNAL_ACCESS_WRITE);
3332 if (ret) {
3333 mlog_errno(ret);
3334 goto out;
3335 }
3336
3337 ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac,
3338 1, 1, &bit_off, &len);
3339 if (ret) {
3340 mlog_errno(ret);
3341 goto out;
3342 }
3343
3344 /*
3345 * The bucket may spread in many blocks, and
3346 * we will only touch the 1st block and the last block
3347 * in the whole bucket(one for entry and one for data).
3348 */
3349 blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off);
3350
3351 mlog(0, "allocate 1 cluster from %llu to xattr block\n",
3352 (unsigned long long)blkno);
3353
3354 ret = ocfs2_init_xattr_bucket(xs->bucket, blkno);
3355 if (ret) {
3356 mlog_errno(ret);
3357 goto out;
3358 }
3359
3360 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
3361 OCFS2_JOURNAL_ACCESS_CREATE);
3362 if (ret) {
3363 mlog_errno(ret);
3364 goto out;
3365 }
3366
3367 ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket);
3368 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
3369
3370 ocfs2_xattr_update_xattr_search(inode, xs, xb_bh);
3371
3372 /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */
3373 memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize -
3374 offsetof(struct ocfs2_xattr_block, xb_attrs));
3375
3376 xr = &xb->xb_attrs.xb_root;
3377 xr->xt_clusters = cpu_to_le32(1);
3378 xr->xt_last_eb_blk = 0;
3379 xr->xt_list.l_tree_depth = 0;
3380 xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb));
3381 xr->xt_list.l_next_free_rec = cpu_to_le16(1);
3382
3383 xr->xt_list.l_recs[0].e_cpos = 0;
3384 xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno);
3385 xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1);
3386
3387 xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED);
3388
3389 ocfs2_journal_dirty(handle, xb_bh);
3390
3391 out:
3392 up_write(&oi->ip_alloc_sem);
3393
3394 return ret;
3395 }
3396
3397 static int cmp_xe_offset(const void *a, const void *b)
3398 {
3399 const struct ocfs2_xattr_entry *l = a, *r = b;
3400 u32 l_name_offset = le16_to_cpu(l->xe_name_offset);
3401 u32 r_name_offset = le16_to_cpu(r->xe_name_offset);
3402
3403 if (l_name_offset < r_name_offset)
3404 return 1;
3405 if (l_name_offset > r_name_offset)
3406 return -1;
3407 return 0;
3408 }
3409
3410 /*
3411 * defrag a xattr bucket if we find that the bucket has some
3412 * holes beteen name/value pairs.
3413 * We will move all the name/value pairs to the end of the bucket
3414 * so that we can spare some space for insertion.
3415 */
3416 static int ocfs2_defrag_xattr_bucket(struct inode *inode,
3417 handle_t *handle,
3418 struct ocfs2_xattr_bucket *bucket)
3419 {
3420 int ret, i;
3421 size_t end, offset, len, value_len;
3422 struct ocfs2_xattr_header *xh;
3423 char *entries, *buf, *bucket_buf = NULL;
3424 u64 blkno = bucket_blkno(bucket);
3425 u16 xh_free_start;
3426 size_t blocksize = inode->i_sb->s_blocksize;
3427 struct ocfs2_xattr_entry *xe;
3428
3429 /*
3430 * In order to make the operation more efficient and generic,
3431 * we copy all the blocks into a contiguous memory and do the
3432 * defragment there, so if anything is error, we will not touch
3433 * the real block.
3434 */
3435 bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS);
3436 if (!bucket_buf) {
3437 ret = -EIO;
3438 goto out;
3439 }
3440
3441 buf = bucket_buf;
3442 for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
3443 memcpy(buf, bucket_block(bucket, i), blocksize);
3444
3445 ret = ocfs2_xattr_bucket_journal_access(handle, bucket,
3446 OCFS2_JOURNAL_ACCESS_WRITE);
3447 if (ret < 0) {
3448 mlog_errno(ret);
3449 goto out;
3450 }
3451
3452 xh = (struct ocfs2_xattr_header *)bucket_buf;
3453 entries = (char *)xh->xh_entries;
3454 xh_free_start = le16_to_cpu(xh->xh_free_start);
3455
3456 mlog(0, "adjust xattr bucket in %llu, count = %u, "
3457 "xh_free_start = %u, xh_name_value_len = %u.\n",
3458 (unsigned long long)blkno, le16_to_cpu(xh->xh_count),
3459 xh_free_start, le16_to_cpu(xh->xh_name_value_len));
3460
3461 /*
3462 * sort all the entries by their offset.
3463 * the largest will be the first, so that we can
3464 * move them to the end one by one.
3465 */
3466 sort(entries, le16_to_cpu(xh->xh_count),
3467 sizeof(struct ocfs2_xattr_entry),
3468 cmp_xe_offset, swap_xe);
3469
3470 /* Move all name/values to the end of the bucket. */
3471 xe = xh->xh_entries;
3472 end = OCFS2_XATTR_BUCKET_SIZE;
3473 for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) {
3474 offset = le16_to_cpu(xe->xe_name_offset);
3475 if (ocfs2_xattr_is_local(xe))
3476 value_len = OCFS2_XATTR_SIZE(
3477 le64_to_cpu(xe->xe_value_size));
3478 else
3479 value_len = OCFS2_XATTR_ROOT_SIZE;
3480 len = OCFS2_XATTR_SIZE(xe->xe_name_len) + value_len;
3481
3482 /*
3483 * We must make sure that the name/value pair
3484 * exist in the same block. So adjust end to
3485 * the previous block end if needed.
3486 */
3487 if (((end - len) / blocksize !=
3488 (end - 1) / blocksize))
3489 end = end - end % blocksize;
3490
3491 if (end > offset + len) {
3492 memmove(bucket_buf + end - len,
3493 bucket_buf + offset, len);
3494 xe->xe_name_offset = cpu_to_le16(end - len);
3495 }
3496
3497 mlog_bug_on_msg(end < offset + len, "Defrag check failed for "
3498 "bucket %llu\n", (unsigned long long)blkno);
3499
3500 end -= len;
3501 }
3502
3503 mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for "
3504 "bucket %llu\n", (unsigned long long)blkno);
3505
3506 if (xh_free_start == end)
3507 goto out;
3508
3509 memset(bucket_buf + xh_free_start, 0, end - xh_free_start);
3510 xh->xh_free_start = cpu_to_le16(end);
3511
3512 /* sort the entries by their name_hash. */
3513 sort(entries, le16_to_cpu(xh->xh_count),
3514 sizeof(struct ocfs2_xattr_entry),
3515 cmp_xe, swap_xe);
3516
3517 buf = bucket_buf;
3518 for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
3519 memcpy(bucket_block(bucket, i), buf, blocksize);
3520 ocfs2_xattr_bucket_journal_dirty(handle, bucket);
3521
3522 out:
3523 kfree(bucket_buf);
3524 return ret;
3525 }
3526
3527 /*
3528 * Move half nums of the xattr bucket in the previous cluster to this new
3529 * cluster. We only touch the last cluster of the previous extend record.
3530 *
3531 * first_bh is the first buffer_head of a series of bucket in the same
3532 * extent rec and header_bh is the header of one bucket in this cluster.
3533 * They will be updated if we move the data header_bh contains to the new
3534 * cluster. first_hash will be set as the 1st xe's name_hash of the new cluster.
3535 */
3536 static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode,
3537 handle_t *handle,
3538 struct buffer_head **first_bh,
3539 struct buffer_head **header_bh,
3540 u64 new_blkno,
3541 u64 prev_blkno,
3542 u32 num_clusters,
3543 u32 *first_hash)
3544 {
3545 int i, ret, credits;
3546 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
3547 int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
3548 int num_buckets = ocfs2_xattr_buckets_per_cluster(osb);
3549 int blocksize = inode->i_sb->s_blocksize;
3550 struct buffer_head *old_bh, *new_bh, *prev_bh, *new_first_bh = NULL;
3551 struct ocfs2_xattr_header *new_xh;
3552 struct ocfs2_xattr_header *xh =
3553 (struct ocfs2_xattr_header *)((*first_bh)->b_data);
3554
3555 BUG_ON(le16_to_cpu(xh->xh_num_buckets) < num_buckets);
3556 BUG_ON(OCFS2_XATTR_BUCKET_SIZE == osb->s_clustersize);
3557
3558 prev_bh = *first_bh;
3559 get_bh(prev_bh);
3560 xh = (struct ocfs2_xattr_header *)prev_bh->b_data;
3561
3562 prev_blkno += (num_clusters - 1) * bpc + bpc / 2;
3563
3564 mlog(0, "move half of xattrs in cluster %llu to %llu\n",
3565 (unsigned long long)prev_blkno, (unsigned long long)new_blkno);
3566
3567 /*
3568 * We need to update the 1st half of the new cluster and
3569 * 1 more for the update of the 1st bucket of the previous
3570 * extent record.
3571 */
3572 credits = bpc / 2 + 1 + handle->h_buffer_credits;
3573 ret = ocfs2_extend_trans(handle, credits);
3574 if (ret) {
3575 mlog_errno(ret);
3576 goto out;
3577 }
3578
3579 ret = ocfs2_journal_access(handle, inode, prev_bh,
3580 OCFS2_JOURNAL_ACCESS_WRITE);
3581 if (ret) {
3582 mlog_errno(ret);
3583 goto out;
3584 }
3585
3586 for (i = 0; i < bpc / 2; i++, prev_blkno++, new_blkno++) {
3587 old_bh = new_bh = NULL;
3588 new_bh = sb_getblk(inode->i_sb, new_blkno);
3589 if (!new_bh) {
3590 ret = -EIO;
3591 mlog_errno(ret);
3592 goto out;
3593 }
3594
3595 ocfs2_set_new_buffer_uptodate(inode, new_bh);
3596
3597 ret = ocfs2_journal_access(handle, inode, new_bh,
3598 OCFS2_JOURNAL_ACCESS_CREATE);
3599 if (ret < 0) {
3600 mlog_errno(ret);
3601 brelse(new_bh);
3602 goto out;
3603 }
3604
3605 ret = ocfs2_read_block(inode, prev_blkno, &old_bh, NULL);
3606 if (ret < 0) {
3607 mlog_errno(ret);
3608 brelse(new_bh);
3609 goto out;
3610 }
3611
3612 memcpy(new_bh->b_data, old_bh->b_data, blocksize);
3613
3614 if (i == 0) {
3615 new_xh = (struct ocfs2_xattr_header *)new_bh->b_data;
3616 new_xh->xh_num_buckets = cpu_to_le16(num_buckets / 2);
3617
3618 if (first_hash)
3619 *first_hash = le32_to_cpu(
3620 new_xh->xh_entries[0].xe_name_hash);
3621 new_first_bh = new_bh;
3622 get_bh(new_first_bh);
3623 }
3624
3625 ocfs2_journal_dirty(handle, new_bh);
3626
3627 if (*header_bh == old_bh) {
3628 brelse(*header_bh);
3629 *header_bh = new_bh;
3630 get_bh(*header_bh);
3631
3632 brelse(*first_bh);
3633 *first_bh = new_first_bh;
3634 get_bh(*first_bh);
3635 }
3636 brelse(new_bh);
3637 brelse(old_bh);
3638 }
3639
3640 le16_add_cpu(&xh->xh_num_buckets, -(num_buckets / 2));
3641
3642 ocfs2_journal_dirty(handle, prev_bh);
3643 out:
3644 brelse(prev_bh);
3645 brelse(new_first_bh);
3646 return ret;
3647 }
3648
3649 /*
3650 * Find the suitable pos when we divide a bucket into 2.
3651 * We have to make sure the xattrs with the same hash value exist
3652 * in the same bucket.
3653 *
3654 * If this ocfs2_xattr_header covers more than one hash value, find a
3655 * place where the hash value changes. Try to find the most even split.
3656 * The most common case is that all entries have different hash values,
3657 * and the first check we make will find a place to split.
3658 */
3659 static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh)
3660 {
3661 struct ocfs2_xattr_entry *entries = xh->xh_entries;
3662 int count = le16_to_cpu(xh->xh_count);
3663 int delta, middle = count / 2;
3664
3665 /*
3666 * We start at the middle. Each step gets farther away in both
3667 * directions. We therefore hit the change in hash value
3668 * nearest to the middle. Note that this loop does not execute for
3669 * count < 2.
3670 */
3671 for (delta = 0; delta < middle; delta++) {
3672 /* Let's check delta earlier than middle */
3673 if (cmp_xe(&entries[middle - delta - 1],
3674 &entries[middle - delta]))
3675 return middle - delta;
3676
3677 /* For even counts, don't walk off the end */
3678 if ((middle + delta + 1) == count)
3679 continue;
3680
3681 /* Now try delta past middle */
3682 if (cmp_xe(&entries[middle + delta],
3683 &entries[middle + delta + 1]))
3684 return middle + delta + 1;
3685 }
3686
3687 /* Every entry had the same hash */
3688 return count;
3689 }
3690
3691 /*
3692 * Move some xattrs in old bucket(blk) to new bucket(new_blk).
3693 * first_hash will record the 1st hash of the new bucket.
3694 *
3695 * Normally half of the xattrs will be moved. But we have to make
3696 * sure that the xattrs with the same hash value are stored in the
3697 * same bucket. If all the xattrs in this bucket have the same hash
3698 * value, the new bucket will be initialized as an empty one and the
3699 * first_hash will be initialized as (hash_value+1).
3700 */
3701 static int ocfs2_divide_xattr_bucket(struct inode *inode,
3702 handle_t *handle,
3703 u64 blk,
3704 u64 new_blk,
3705 u32 *first_hash,
3706 int new_bucket_head)
3707 {
3708 int ret, i;
3709 int count, start, len, name_value_len = 0, xe_len, name_offset = 0;
3710 struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
3711 struct ocfs2_xattr_header *xh;
3712 struct ocfs2_xattr_entry *xe;
3713 int blocksize = inode->i_sb->s_blocksize;
3714
3715 mlog(0, "move some of xattrs from bucket %llu to %llu\n",
3716 (unsigned long long)blk, (unsigned long long)new_blk);
3717
3718 s_bucket = ocfs2_xattr_bucket_new(inode);
3719 t_bucket = ocfs2_xattr_bucket_new(inode);
3720 if (!s_bucket || !t_bucket) {
3721 ret = -ENOMEM;
3722 mlog_errno(ret);
3723 goto out;
3724 }
3725
3726 ret = ocfs2_read_xattr_bucket(s_bucket, blk);
3727 if (ret) {
3728 mlog_errno(ret);
3729 goto out;
3730 }
3731
3732 ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket,
3733 OCFS2_JOURNAL_ACCESS_WRITE);
3734 if (ret) {
3735 mlog_errno(ret);
3736 goto out;
3737 }
3738
3739 /*
3740 * Even if !new_bucket_head, we're overwriting t_bucket. Thus,
3741 * there's no need to read it.
3742 */
3743 ret = ocfs2_init_xattr_bucket(t_bucket, new_blk);
3744 if (ret) {
3745 mlog_errno(ret);
3746 goto out;
3747 }
3748
3749 ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
3750 new_bucket_head ?
3751 OCFS2_JOURNAL_ACCESS_CREATE :
3752 OCFS2_JOURNAL_ACCESS_WRITE);
3753 if (ret) {
3754 mlog_errno(ret);
3755 goto out;
3756 }
3757
3758 xh = bucket_xh(s_bucket);
3759 count = le16_to_cpu(xh->xh_count);
3760 start = ocfs2_xattr_find_divide_pos(xh);
3761
3762 if (start == count) {
3763 xe = &xh->xh_entries[start-1];
3764
3765 /*
3766 * initialized a new empty bucket here.
3767 * The hash value is set as one larger than
3768 * that of the last entry in the previous bucket.
3769 */
3770 for (i = 0; i < t_bucket->bu_blocks; i++)
3771 memset(bucket_block(t_bucket, i), 0, blocksize);
3772
3773 xh = bucket_xh(t_bucket);
3774 xh->xh_free_start = cpu_to_le16(blocksize);
3775 xh->xh_entries[0].xe_name_hash = xe->xe_name_hash;
3776 le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1);
3777
3778 goto set_num_buckets;
3779 }
3780
3781 /* copy the whole bucket to the new first. */
3782 ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket);
3783
3784 /* update the new bucket. */
3785 xh = bucket_xh(t_bucket);
3786
3787 /*
3788 * Calculate the total name/value len and xh_free_start for
3789 * the old bucket first.
3790 */
3791 name_offset = OCFS2_XATTR_BUCKET_SIZE;
3792 name_value_len = 0;
3793 for (i = 0; i < start; i++) {
3794 xe = &xh->xh_entries[i];
3795 xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
3796 if (ocfs2_xattr_is_local(xe))
3797 xe_len +=
3798 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
3799 else
3800 xe_len += OCFS2_XATTR_ROOT_SIZE;
3801 name_value_len += xe_len;
3802 if (le16_to_cpu(xe->xe_name_offset) < name_offset)
3803 name_offset = le16_to_cpu(xe->xe_name_offset);
3804 }
3805
3806 /*
3807 * Now begin the modification to the new bucket.
3808 *
3809 * In the new bucket, We just move the xattr entry to the beginning
3810 * and don't touch the name/value. So there will be some holes in the
3811 * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is
3812 * called.
3813 */
3814 xe = &xh->xh_entries[start];
3815 len = sizeof(struct ocfs2_xattr_entry) * (count - start);
3816 mlog(0, "mv xattr entry len %d from %d to %d\n", len,
3817 (int)((char *)xe - (char *)xh),
3818 (int)((char *)xh->xh_entries - (char *)xh));
3819 memmove((char *)xh->xh_entries, (char *)xe, len);
3820 xe = &xh->xh_entries[count - start];
3821 len = sizeof(struct ocfs2_xattr_entry) * start;
3822 memset((char *)xe, 0, len);
3823
3824 le16_add_cpu(&xh->xh_count, -start);
3825 le16_add_cpu(&xh->xh_name_value_len, -name_value_len);
3826
3827 /* Calculate xh_free_start for the new bucket. */
3828 xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
3829 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
3830 xe = &xh->xh_entries[i];
3831 xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
3832 if (ocfs2_xattr_is_local(xe))
3833 xe_len +=
3834 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
3835 else
3836 xe_len += OCFS2_XATTR_ROOT_SIZE;
3837 if (le16_to_cpu(xe->xe_name_offset) <
3838 le16_to_cpu(xh->xh_free_start))
3839 xh->xh_free_start = xe->xe_name_offset;
3840 }
3841
3842 set_num_buckets:
3843 /* set xh->xh_num_buckets for the new xh. */
3844 if (new_bucket_head)
3845 xh->xh_num_buckets = cpu_to_le16(1);
3846 else
3847 xh->xh_num_buckets = 0;
3848
3849 ocfs2_xattr_bucket_journal_dirty(handle, t_bucket);
3850
3851 /* store the first_hash of the new bucket. */
3852 if (first_hash)
3853 *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash);
3854
3855 /*
3856 * Now only update the 1st block of the old bucket. If we
3857 * just added a new empty bucket, there is no need to modify
3858 * it.
3859 */
3860 if (start == count)
3861 goto out;
3862
3863 xh = bucket_xh(s_bucket);
3864 memset(&xh->xh_entries[start], 0,
3865 sizeof(struct ocfs2_xattr_entry) * (count - start));
3866 xh->xh_count = cpu_to_le16(start);
3867 xh->xh_free_start = cpu_to_le16(name_offset);
3868 xh->xh_name_value_len = cpu_to_le16(name_value_len);
3869
3870 ocfs2_xattr_bucket_journal_dirty(handle, s_bucket);
3871
3872 out:
3873 ocfs2_xattr_bucket_free(s_bucket);
3874 ocfs2_xattr_bucket_free(t_bucket);
3875
3876 return ret;
3877 }
3878
3879 /*
3880 * Copy xattr from one bucket to another bucket.
3881 *
3882 * The caller must make sure that the journal transaction
3883 * has enough space for journaling.
3884 */
3885 static int ocfs2_cp_xattr_bucket(struct inode *inode,
3886 handle_t *handle,
3887 u64 s_blkno,
3888 u64 t_blkno,
3889 int t_is_new)
3890 {
3891 int ret;
3892 struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
3893
3894 BUG_ON(s_blkno == t_blkno);
3895
3896 mlog(0, "cp bucket %llu to %llu, target is %d\n",
3897 (unsigned long long)s_blkno, (unsigned long long)t_blkno,
3898 t_is_new);
3899
3900 s_bucket = ocfs2_xattr_bucket_new(inode);
3901 t_bucket = ocfs2_xattr_bucket_new(inode);
3902 if (!s_bucket || !t_bucket) {
3903 ret = -ENOMEM;
3904 mlog_errno(ret);
3905 goto out;
3906 }
3907
3908 ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno);
3909 if (ret)
3910 goto out;
3911
3912 /*
3913 * Even if !t_is_new, we're overwriting t_bucket. Thus,
3914 * there's no need to read it.
3915 */
3916 ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno);
3917 if (ret)
3918 goto out;
3919
3920 ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
3921 t_is_new ?
3922 OCFS2_JOURNAL_ACCESS_CREATE :
3923 OCFS2_JOURNAL_ACCESS_WRITE);
3924 if (ret)
3925 goto out;
3926
3927 ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket);
3928 ocfs2_xattr_bucket_journal_dirty(handle, t_bucket);
3929
3930 out:
3931 ocfs2_xattr_bucket_free(t_bucket);
3932 ocfs2_xattr_bucket_free(s_bucket);
3933
3934 return ret;
3935 }
3936
3937 /*
3938 * Copy one xattr cluster from src_blk to to_blk.
3939 * The to_blk will become the first bucket header of the cluster, so its
3940 * xh_num_buckets will be initialized as the bucket num in the cluster.
3941 */
3942 static int ocfs2_cp_xattr_cluster(struct inode *inode,
3943 handle_t *handle,
3944 struct buffer_head *first_bh,
3945 u64 src_blk,
3946 u64 to_blk,
3947 u32 *first_hash)
3948 {
3949 int i, ret, credits;
3950 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
3951 int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
3952 int num_buckets = ocfs2_xattr_buckets_per_cluster(osb);
3953 struct buffer_head *bh = NULL;
3954 struct ocfs2_xattr_header *xh;
3955 u64 to_blk_start = to_blk;
3956
3957 mlog(0, "cp xattrs from cluster %llu to %llu\n",
3958 (unsigned long long)src_blk, (unsigned long long)to_blk);
3959
3960 /*
3961 * We need to update the new cluster and 1 more for the update of
3962 * the 1st bucket of the previous extent rec.
3963 */
3964 credits = bpc + 1 + handle->h_buffer_credits;
3965 ret = ocfs2_extend_trans(handle, credits);
3966 if (ret) {
3967 mlog_errno(ret);
3968 goto out;
3969 }
3970
3971 ret = ocfs2_journal_access(handle, inode, first_bh,
3972 OCFS2_JOURNAL_ACCESS_WRITE);
3973 if (ret) {
3974 mlog_errno(ret);
3975 goto out;
3976 }
3977
3978 for (i = 0; i < num_buckets; i++) {
3979 ret = ocfs2_cp_xattr_bucket(inode, handle,
3980 src_blk, to_blk, 1);
3981 if (ret) {
3982 mlog_errno(ret);
3983 goto out;
3984 }
3985
3986 src_blk += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
3987 to_blk += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
3988 }
3989
3990 /* update the old bucket header. */
3991 xh = (struct ocfs2_xattr_header *)first_bh->b_data;
3992 le16_add_cpu(&xh->xh_num_buckets, -num_buckets);
3993
3994 ocfs2_journal_dirty(handle, first_bh);
3995
3996 /* update the new bucket header. */
3997 ret = ocfs2_read_block(inode, to_blk_start, &bh, NULL);
3998 if (ret < 0) {
3999 mlog_errno(ret);
4000 goto out;
4001 }
4002
4003 ret = ocfs2_journal_access(handle, inode, bh,
4004 OCFS2_JOURNAL_ACCESS_WRITE);
4005 if (ret) {
4006 mlog_errno(ret);
4007 goto out;
4008 }
4009
4010 xh = (struct ocfs2_xattr_header *)bh->b_data;
4011 xh->xh_num_buckets = cpu_to_le16(num_buckets);
4012
4013 ocfs2_journal_dirty(handle, bh);
4014
4015 if (first_hash)
4016 *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash);
4017 out:
4018 brelse(bh);
4019 return ret;
4020 }
4021
4022 /*
4023 * Move some xattrs in this cluster to the new cluster.
4024 * This function should only be called when bucket size == cluster size.
4025 * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead.
4026 */
4027 static int ocfs2_divide_xattr_cluster(struct inode *inode,
4028 handle_t *handle,
4029 u64 prev_blk,
4030 u64 new_blk,
4031 u32 *first_hash)
4032 {
4033 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
4034 int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits;
4035
4036 BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize);
4037
4038 ret = ocfs2_extend_trans(handle, credits);
4039 if (ret) {
4040 mlog_errno(ret);
4041 return ret;
4042 }
4043
4044 /* Move half of the xattr in start_blk to the next bucket. */
4045 return ocfs2_divide_xattr_bucket(inode, handle, prev_blk,
4046 new_blk, first_hash, 1);
4047 }
4048
4049 /*
4050 * Move some xattrs from the old cluster to the new one since they are not
4051 * contiguous in ocfs2 xattr tree.
4052 *
4053 * new_blk starts a new separate cluster, and we will move some xattrs from
4054 * prev_blk to it. v_start will be set as the first name hash value in this
4055 * new cluster so that it can be used as e_cpos during tree insertion and
4056 * don't collide with our original b-tree operations. first_bh and header_bh
4057 * will also be updated since they will be used in ocfs2_extend_xattr_bucket
4058 * to extend the insert bucket.
4059 *
4060 * The problem is how much xattr should we move to the new one and when should
4061 * we update first_bh and header_bh?
4062 * 1. If cluster size > bucket size, that means the previous cluster has more
4063 * than 1 bucket, so just move half nums of bucket into the new cluster and
4064 * update the first_bh and header_bh if the insert bucket has been moved
4065 * to the new cluster.
4066 * 2. If cluster_size == bucket_size:
4067 * a) If the previous extent rec has more than one cluster and the insert
4068 * place isn't in the last cluster, copy the entire last cluster to the
4069 * new one. This time, we don't need to upate the first_bh and header_bh
4070 * since they will not be moved into the new cluster.
4071 * b) Otherwise, move the bottom half of the xattrs in the last cluster into
4072 * the new one. And we set the extend flag to zero if the insert place is
4073 * moved into the new allocated cluster since no extend is needed.
4074 */
4075 static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode,
4076 handle_t *handle,
4077 struct buffer_head **first_bh,
4078 struct buffer_head **header_bh,
4079 u64 new_blk,
4080 u64 prev_blk,
4081 u32 prev_clusters,
4082 u32 *v_start,
4083 int *extend)
4084 {
4085 int ret = 0;
4086 int bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
4087
4088 mlog(0, "adjust xattrs from cluster %llu len %u to %llu\n",
4089 (unsigned long long)prev_blk, prev_clusters,
4090 (unsigned long long)new_blk);
4091
4092 if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1)
4093 ret = ocfs2_mv_xattr_bucket_cross_cluster(inode,
4094 handle,
4095 first_bh,
4096 header_bh,
4097 new_blk,
4098 prev_blk,
4099 prev_clusters,
4100 v_start);
4101 else {
4102 u64 last_blk = prev_blk + bpc * (prev_clusters - 1);
4103
4104 if (prev_clusters > 1 && (*header_bh)->b_blocknr != last_blk)
4105 ret = ocfs2_cp_xattr_cluster(inode, handle, *first_bh,
4106 last_blk, new_blk,
4107 v_start);
4108 else {
4109 ret = ocfs2_divide_xattr_cluster(inode, handle,
4110 last_blk, new_blk,
4111 v_start);
4112
4113 if ((*header_bh)->b_blocknr == last_blk && extend)
4114 *extend = 0;
4115 }
4116 }
4117
4118 return ret;
4119 }
4120
4121 /*
4122 * Add a new cluster for xattr storage.
4123 *
4124 * If the new cluster is contiguous with the previous one, it will be
4125 * appended to the same extent record, and num_clusters will be updated.
4126 * If not, we will insert a new extent for it and move some xattrs in
4127 * the last cluster into the new allocated one.
4128 * We also need to limit the maximum size of a btree leaf, otherwise we'll
4129 * lose the benefits of hashing because we'll have to search large leaves.
4130 * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize,
4131 * if it's bigger).
4132 *
4133 * first_bh is the first block of the previous extent rec and header_bh
4134 * indicates the bucket we will insert the new xattrs. They will be updated
4135 * when the header_bh is moved into the new cluster.
4136 */
4137 static int ocfs2_add_new_xattr_cluster(struct inode *inode,
4138 struct buffer_head *root_bh,
4139 struct buffer_head **first_bh,
4140 struct buffer_head **header_bh,
4141 u32 *num_clusters,
4142 u32 prev_cpos,
4143 u64 prev_blkno,
4144 int *extend,
4145 struct ocfs2_xattr_set_ctxt *ctxt)
4146 {
4147 int ret;
4148 u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
4149 u32 prev_clusters = *num_clusters;
4150 u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0;
4151 u64 block;
4152 handle_t *handle = ctxt->handle;
4153 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
4154 struct ocfs2_extent_tree et;
4155
4156 mlog(0, "Add new xattr cluster for %llu, previous xattr hash = %u, "
4157 "previous xattr blkno = %llu\n",
4158 (unsigned long long)OCFS2_I(inode)->ip_blkno,
4159 prev_cpos, (unsigned long long)prev_blkno);
4160
4161 ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh);
4162
4163 ret = ocfs2_journal_access(handle, inode, root_bh,
4164 OCFS2_JOURNAL_ACCESS_WRITE);
4165 if (ret < 0) {
4166 mlog_errno(ret);
4167 goto leave;
4168 }
4169
4170 ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1,
4171 clusters_to_add, &bit_off, &num_bits);
4172 if (ret < 0) {
4173 if (ret != -ENOSPC)
4174 mlog_errno(ret);
4175 goto leave;
4176 }
4177
4178 BUG_ON(num_bits > clusters_to_add);
4179
4180 block = ocfs2_clusters_to_blocks(osb->sb, bit_off);
4181 mlog(0, "Allocating %u clusters at block %u for xattr in inode %llu\n",
4182 num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno);
4183
4184 if (prev_blkno + prev_clusters * bpc == block &&
4185 (prev_clusters + num_bits) << osb->s_clustersize_bits <=
4186 OCFS2_MAX_XATTR_TREE_LEAF_SIZE) {
4187 /*
4188 * If this cluster is contiguous with the old one and
4189 * adding this new cluster, we don't surpass the limit of
4190 * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be
4191 * initialized and used like other buckets in the previous
4192 * cluster.
4193 * So add it as a contiguous one. The caller will handle
4194 * its init process.
4195 */
4196 v_start = prev_cpos + prev_clusters;
4197 *num_clusters = prev_clusters + num_bits;
4198 mlog(0, "Add contiguous %u clusters to previous extent rec.\n",
4199 num_bits);
4200 } else {
4201 ret = ocfs2_adjust_xattr_cross_cluster(inode,
4202 handle,
4203 first_bh,
4204 header_bh,
4205 block,
4206 prev_blkno,
4207 prev_clusters,
4208 &v_start,
4209 extend);
4210 if (ret) {
4211 mlog_errno(ret);
4212 goto leave;
4213 }
4214 }
4215
4216 mlog(0, "Insert %u clusters at block %llu for xattr at %u\n",
4217 num_bits, (unsigned long long)block, v_start);
4218 ret = ocfs2_insert_extent(osb, handle, inode, &et, v_start, block,
4219 num_bits, 0, ctxt->meta_ac);
4220 if (ret < 0) {
4221 mlog_errno(ret);
4222 goto leave;
4223 }
4224
4225 ret = ocfs2_journal_dirty(handle, root_bh);
4226 if (ret < 0)
4227 mlog_errno(ret);
4228
4229 leave:
4230 return ret;
4231 }
4232
4233 /*
4234 * Extend a new xattr bucket and move xattrs to the end one by one until
4235 * We meet with start_bh. Only move half of the xattrs to the bucket after it.
4236 */
4237 static int ocfs2_extend_xattr_bucket(struct inode *inode,
4238 handle_t *handle,
4239 struct buffer_head *first_bh,
4240 struct buffer_head *start_bh,
4241 u32 num_clusters)
4242 {
4243 int ret, credits;
4244 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
4245 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
4246 u64 start_blk = start_bh->b_blocknr, end_blk;
4247 u32 num_buckets = num_clusters * ocfs2_xattr_buckets_per_cluster(osb);
4248 struct ocfs2_xattr_header *first_xh =
4249 (struct ocfs2_xattr_header *)first_bh->b_data;
4250 u16 bucket = le16_to_cpu(first_xh->xh_num_buckets);
4251
4252 mlog(0, "extend xattr bucket in %llu, xattr extend rec starting "
4253 "from %llu, len = %u\n", (unsigned long long)start_blk,
4254 (unsigned long long)first_bh->b_blocknr, num_clusters);
4255
4256 BUG_ON(bucket >= num_buckets);
4257
4258 end_blk = first_bh->b_blocknr + (bucket - 1) * blk_per_bucket;
4259
4260 /*
4261 * We will touch all the buckets after the start_bh(include it).
4262 * Then we add one more bucket.
4263 */
4264 credits = end_blk - start_blk + 3 * blk_per_bucket + 1 +
4265 handle->h_buffer_credits;
4266 ret = ocfs2_extend_trans(handle, credits);
4267 if (ret) {
4268 mlog_errno(ret);
4269 goto out;
4270 }
4271
4272 ret = ocfs2_journal_access(handle, inode, first_bh,
4273 OCFS2_JOURNAL_ACCESS_WRITE);
4274 if (ret) {
4275 mlog_errno(ret);
4276 goto out;
4277 }
4278
4279 while (end_blk != start_blk) {
4280 ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk,
4281 end_blk + blk_per_bucket, 0);
4282 if (ret)
4283 goto out;
4284 end_blk -= blk_per_bucket;
4285 }
4286
4287 /* Move half of the xattr in start_blk to the next bucket. */
4288 ret = ocfs2_divide_xattr_bucket(inode, handle, start_blk,
4289 start_blk + blk_per_bucket, NULL, 0);
4290
4291 le16_add_cpu(&first_xh->xh_num_buckets, 1);
4292 ocfs2_journal_dirty(handle, first_bh);
4293
4294 out:
4295 return ret;
4296 }
4297
4298 /*
4299 * Add new xattr bucket in an extent record and adjust the buckets accordingly.
4300 * xb_bh is the ocfs2_xattr_block.
4301 * We will move all the buckets starting from header_bh to the next place. As
4302 * for this one, half num of its xattrs will be moved to the next one.
4303 *
4304 * We will allocate a new cluster if current cluster is full and adjust
4305 * header_bh and first_bh if the insert place is moved to the new cluster.
4306 */
4307 static int ocfs2_add_new_xattr_bucket(struct inode *inode,
4308 struct buffer_head *xb_bh,
4309 struct buffer_head *header_bh,
4310 struct ocfs2_xattr_set_ctxt *ctxt)
4311 {
4312 struct ocfs2_xattr_header *first_xh = NULL;
4313 struct buffer_head *first_bh = NULL;
4314 struct ocfs2_xattr_block *xb =
4315 (struct ocfs2_xattr_block *)xb_bh->b_data;
4316 struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root;
4317 struct ocfs2_extent_list *el = &xb_root->xt_list;
4318 struct ocfs2_xattr_header *xh =
4319 (struct ocfs2_xattr_header *)header_bh->b_data;
4320 u32 name_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash);
4321 struct super_block *sb = inode->i_sb;
4322 struct ocfs2_super *osb = OCFS2_SB(sb);
4323 int ret, num_buckets, extend = 1;
4324 u64 p_blkno;
4325 u32 e_cpos, num_clusters;
4326
4327 mlog(0, "Add new xattr bucket starting form %llu\n",
4328 (unsigned long long)header_bh->b_blocknr);
4329
4330 /*
4331 * Add refrence for header_bh here because it may be
4332 * changed in ocfs2_add_new_xattr_cluster and we need
4333 * to free it in the end.
4334 */
4335 get_bh(header_bh);
4336
4337 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos,
4338 &num_clusters, el);
4339 if (ret) {
4340 mlog_errno(ret);
4341 goto out;
4342 }
4343
4344 ret = ocfs2_read_block(inode, p_blkno, &first_bh, NULL);
4345 if (ret) {
4346 mlog_errno(ret);
4347 goto out;
4348 }
4349
4350 num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters;
4351 first_xh = (struct ocfs2_xattr_header *)first_bh->b_data;
4352
4353 if (num_buckets == le16_to_cpu(first_xh->xh_num_buckets)) {
4354 ret = ocfs2_add_new_xattr_cluster(inode,
4355 xb_bh,
4356 &first_bh,
4357 &header_bh,
4358 &num_clusters,
4359 e_cpos,
4360 p_blkno,
4361 &extend,
4362 ctxt);
4363 if (ret) {
4364 mlog_errno(ret);
4365 goto out;
4366 }
4367 }
4368
4369 if (extend)
4370 ret = ocfs2_extend_xattr_bucket(inode,
4371 ctxt->handle,
4372 first_bh,
4373 header_bh,
4374 num_clusters);
4375 if (ret)
4376 mlog_errno(ret);
4377 out:
4378 brelse(first_bh);
4379 brelse(header_bh);
4380 return ret;
4381 }
4382
4383 static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode,
4384 struct ocfs2_xattr_bucket *bucket,
4385 int offs)
4386 {
4387 int block_off = offs >> inode->i_sb->s_blocksize_bits;
4388
4389 offs = offs % inode->i_sb->s_blocksize;
4390 return bucket_block(bucket, block_off) + offs;
4391 }
4392
4393 /*
4394 * Handle the normal xattr set, including replace, delete and new.
4395 *
4396 * Note: "local" indicates the real data's locality. So we can't
4397 * just its bucket locality by its length.
4398 */
4399 static void ocfs2_xattr_set_entry_normal(struct inode *inode,
4400 struct ocfs2_xattr_info *xi,
4401 struct ocfs2_xattr_search *xs,
4402 u32 name_hash,
4403 int local)
4404 {
4405 struct ocfs2_xattr_entry *last, *xe;
4406 int name_len = strlen(xi->name);
4407 struct ocfs2_xattr_header *xh = xs->header;
4408 u16 count = le16_to_cpu(xh->xh_count), start;
4409 size_t blocksize = inode->i_sb->s_blocksize;
4410 char *val;
4411 size_t offs, size, new_size;
4412
4413 last = &xh->xh_entries[count];
4414 if (!xs->not_found) {
4415 xe = xs->here;
4416 offs = le16_to_cpu(xe->xe_name_offset);
4417 if (ocfs2_xattr_is_local(xe))
4418 size = OCFS2_XATTR_SIZE(name_len) +
4419 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
4420 else
4421 size = OCFS2_XATTR_SIZE(name_len) +
4422 OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
4423
4424 /*
4425 * If the new value will be stored outside, xi->value has been
4426 * initalized as an empty ocfs2_xattr_value_root, and the same
4427 * goes with xi->value_len, so we can set new_size safely here.
4428 * See ocfs2_xattr_set_in_bucket.
4429 */
4430 new_size = OCFS2_XATTR_SIZE(name_len) +
4431 OCFS2_XATTR_SIZE(xi->value_len);
4432
4433 le16_add_cpu(&xh->xh_name_value_len, -size);
4434 if (xi->value) {
4435 if (new_size > size)
4436 goto set_new_name_value;
4437
4438 /* Now replace the old value with new one. */
4439 if (local)
4440 xe->xe_value_size = cpu_to_le64(xi->value_len);
4441 else
4442 xe->xe_value_size = 0;
4443
4444 val = ocfs2_xattr_bucket_get_val(inode,
4445 xs->bucket, offs);
4446 memset(val + OCFS2_XATTR_SIZE(name_len), 0,
4447 size - OCFS2_XATTR_SIZE(name_len));
4448 if (OCFS2_XATTR_SIZE(xi->value_len) > 0)
4449 memcpy(val + OCFS2_XATTR_SIZE(name_len),
4450 xi->value, xi->value_len);
4451
4452 le16_add_cpu(&xh->xh_name_value_len, new_size);
4453 ocfs2_xattr_set_local(xe, local);
4454 return;
4455 } else {
4456 /*
4457 * Remove the old entry if there is more than one.
4458 * We don't remove the last entry so that we can
4459 * use it to indicate the hash value of the empty
4460 * bucket.
4461 */
4462 last -= 1;
4463 le16_add_cpu(&xh->xh_count, -1);
4464 if (xh->xh_count) {
4465 memmove(xe, xe + 1,
4466 (void *)last - (void *)xe);
4467 memset(last, 0,
4468 sizeof(struct ocfs2_xattr_entry));
4469 } else
4470 xh->xh_free_start =
4471 cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
4472
4473 return;
4474 }
4475 } else {
4476 /* find a new entry for insert. */
4477 int low = 0, high = count - 1, tmp;
4478 struct ocfs2_xattr_entry *tmp_xe;
4479
4480 while (low <= high && count) {
4481 tmp = (low + high) / 2;
4482 tmp_xe = &xh->xh_entries[tmp];
4483
4484 if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash))
4485 low = tmp + 1;
4486 else if (name_hash <
4487 le32_to_cpu(tmp_xe->xe_name_hash))
4488 high = tmp - 1;
4489 else {
4490 low = tmp;
4491 break;
4492 }
4493 }
4494
4495 xe = &xh->xh_entries[low];
4496 if (low != count)
4497 memmove(xe + 1, xe, (void *)last - (void *)xe);
4498
4499 le16_add_cpu(&xh->xh_count, 1);
4500 memset(xe, 0, sizeof(struct ocfs2_xattr_entry));
4501 xe->xe_name_hash = cpu_to_le32(name_hash);
4502 xe->xe_name_len = name_len;
4503 ocfs2_xattr_set_type(xe, xi->name_index);
4504 }
4505
4506 set_new_name_value:
4507 /* Insert the new name+value. */
4508 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(xi->value_len);
4509
4510 /*
4511 * We must make sure that the name/value pair
4512 * exists in the same block.
4513 */
4514 offs = le16_to_cpu(xh->xh_free_start);
4515 start = offs - size;
4516
4517 if (start >> inode->i_sb->s_blocksize_bits !=
4518 (offs - 1) >> inode->i_sb->s_blocksize_bits) {
4519 offs = offs - offs % blocksize;
4520 xh->xh_free_start = cpu_to_le16(offs);
4521 }
4522
4523 val = ocfs2_xattr_bucket_get_val(inode, xs->bucket, offs - size);
4524 xe->xe_name_offset = cpu_to_le16(offs - size);
4525
4526 memset(val, 0, size);
4527 memcpy(val, xi->name, name_len);
4528 memcpy(val + OCFS2_XATTR_SIZE(name_len), xi->value, xi->value_len);
4529
4530 xe->xe_value_size = cpu_to_le64(xi->value_len);
4531 ocfs2_xattr_set_local(xe, local);
4532 xs->here = xe;
4533 le16_add_cpu(&xh->xh_free_start, -size);
4534 le16_add_cpu(&xh->xh_name_value_len, size);
4535
4536 return;
4537 }
4538
4539 /*
4540 * Set the xattr entry in the specified bucket.
4541 * The bucket is indicated by xs->bucket and it should have the enough
4542 * space for the xattr insertion.
4543 */
4544 static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode,
4545 handle_t *handle,
4546 struct ocfs2_xattr_info *xi,
4547 struct ocfs2_xattr_search *xs,
4548 u32 name_hash,
4549 int local)
4550 {
4551 int ret;
4552 u64 blkno;
4553
4554 mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n",
4555 (unsigned long)xi->value_len, xi->name_index,
4556 (unsigned long long)bucket_blkno(xs->bucket));
4557
4558 if (!xs->bucket->bu_bhs[1]) {
4559 blkno = bucket_blkno(xs->bucket);
4560 ocfs2_xattr_bucket_relse(xs->bucket);
4561 ret = ocfs2_read_xattr_bucket(xs->bucket, blkno);
4562 if (ret) {
4563 mlog_errno(ret);
4564 goto out;
4565 }
4566 }
4567
4568 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
4569 OCFS2_JOURNAL_ACCESS_WRITE);
4570 if (ret < 0) {
4571 mlog_errno(ret);
4572 goto out;
4573 }
4574
4575 ocfs2_xattr_set_entry_normal(inode, xi, xs, name_hash, local);
4576 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
4577
4578 out:
4579 return ret;
4580 }
4581
4582 static int ocfs2_xattr_value_update_size(struct inode *inode,
4583 handle_t *handle,
4584 struct buffer_head *xe_bh,
4585 struct ocfs2_xattr_entry *xe,
4586 u64 new_size)
4587 {
4588 int ret;
4589
4590 ret = ocfs2_journal_access(handle, inode, xe_bh,
4591 OCFS2_JOURNAL_ACCESS_WRITE);
4592 if (ret < 0) {
4593 mlog_errno(ret);
4594 goto out;
4595 }
4596
4597 xe->xe_value_size = cpu_to_le64(new_size);
4598
4599 ret = ocfs2_journal_dirty(handle, xe_bh);
4600 if (ret < 0)
4601 mlog_errno(ret);
4602
4603 out:
4604 return ret;
4605 }
4606
4607 /*
4608 * Truncate the specified xe_off entry in xattr bucket.
4609 * bucket is indicated by header_bh and len is the new length.
4610 * Both the ocfs2_xattr_value_root and the entry will be updated here.
4611 *
4612 * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed.
4613 */
4614 static int ocfs2_xattr_bucket_value_truncate(struct inode *inode,
4615 struct buffer_head *header_bh,
4616 int xe_off,
4617 int len,
4618 struct ocfs2_xattr_set_ctxt *ctxt)
4619 {
4620 int ret, offset;
4621 u64 value_blk;
4622 struct buffer_head *value_bh = NULL;
4623 struct ocfs2_xattr_value_root *xv;
4624 struct ocfs2_xattr_entry *xe;
4625 struct ocfs2_xattr_header *xh =
4626 (struct ocfs2_xattr_header *)header_bh->b_data;
4627 size_t blocksize = inode->i_sb->s_blocksize;
4628
4629 xe = &xh->xh_entries[xe_off];
4630
4631 BUG_ON(!xe || ocfs2_xattr_is_local(xe));
4632
4633 offset = le16_to_cpu(xe->xe_name_offset) +
4634 OCFS2_XATTR_SIZE(xe->xe_name_len);
4635
4636 value_blk = offset / blocksize;
4637
4638 /* We don't allow ocfs2_xattr_value to be stored in different block. */
4639 BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize);
4640 value_blk += header_bh->b_blocknr;
4641
4642 ret = ocfs2_read_block(inode, value_blk, &value_bh, NULL);
4643 if (ret) {
4644 mlog_errno(ret);
4645 goto out;
4646 }
4647
4648 xv = (struct ocfs2_xattr_value_root *)
4649 (value_bh->b_data + offset % blocksize);
4650
4651 mlog(0, "truncate %u in xattr bucket %llu to %d bytes.\n",
4652 xe_off, (unsigned long long)header_bh->b_blocknr, len);
4653 ret = ocfs2_xattr_value_truncate(inode, value_bh, xv, len, ctxt);
4654 if (ret) {
4655 mlog_errno(ret);
4656 goto out;
4657 }
4658
4659 ret = ocfs2_xattr_value_update_size(inode, ctxt->handle,
4660 header_bh, xe, len);
4661 if (ret) {
4662 mlog_errno(ret);
4663 goto out;
4664 }
4665
4666 out:
4667 brelse(value_bh);
4668 return ret;
4669 }
4670
4671 static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode,
4672 struct ocfs2_xattr_search *xs,
4673 int len,
4674 struct ocfs2_xattr_set_ctxt *ctxt)
4675 {
4676 int ret, offset;
4677 struct ocfs2_xattr_entry *xe = xs->here;
4678 struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base;
4679
4680 BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe));
4681
4682 offset = xe - xh->xh_entries;
4683 ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket->bu_bhs[0],
4684 offset, len, ctxt);
4685 if (ret)
4686 mlog_errno(ret);
4687
4688 return ret;
4689 }
4690
4691 static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode,
4692 handle_t *handle,
4693 struct ocfs2_xattr_search *xs,
4694 char *val,
4695 int value_len)
4696 {
4697 int offset;
4698 struct ocfs2_xattr_value_root *xv;
4699 struct ocfs2_xattr_entry *xe = xs->here;
4700
4701 BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe));
4702
4703 offset = le16_to_cpu(xe->xe_name_offset) +
4704 OCFS2_XATTR_SIZE(xe->xe_name_len);
4705
4706 xv = (struct ocfs2_xattr_value_root *)(xs->base + offset);
4707
4708 return __ocfs2_xattr_set_value_outside(inode, handle,
4709 xv, val, value_len);
4710 }
4711
4712 static int ocfs2_rm_xattr_cluster(struct inode *inode,
4713 struct buffer_head *root_bh,
4714 u64 blkno,
4715 u32 cpos,
4716 u32 len)
4717 {
4718 int ret;
4719 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
4720 struct inode *tl_inode = osb->osb_tl_inode;
4721 handle_t *handle;
4722 struct ocfs2_xattr_block *xb =
4723 (struct ocfs2_xattr_block *)root_bh->b_data;
4724 struct ocfs2_alloc_context *meta_ac = NULL;
4725 struct ocfs2_cached_dealloc_ctxt dealloc;
4726 struct ocfs2_extent_tree et;
4727
4728 ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh);
4729
4730 ocfs2_init_dealloc_ctxt(&dealloc);
4731
4732 mlog(0, "rm xattr extent rec at %u len = %u, start from %llu\n",
4733 cpos, len, (unsigned long long)blkno);
4734
4735 ocfs2_remove_xattr_clusters_from_cache(inode, blkno, len);
4736
4737 ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac);
4738 if (ret) {
4739 mlog_errno(ret);
4740 return ret;
4741 }
4742
4743 mutex_lock(&tl_inode->i_mutex);
4744
4745 if (ocfs2_truncate_log_needs_flush(osb)) {
4746 ret = __ocfs2_flush_truncate_log(osb);
4747 if (ret < 0) {
4748 mlog_errno(ret);
4749 goto out;
4750 }
4751 }
4752
4753 handle = ocfs2_start_trans(osb, OCFS2_REMOVE_EXTENT_CREDITS);
4754 if (IS_ERR(handle)) {
4755 ret = -ENOMEM;
4756 mlog_errno(ret);
4757 goto out;
4758 }
4759
4760 ret = ocfs2_journal_access(handle, inode, root_bh,
4761 OCFS2_JOURNAL_ACCESS_WRITE);
4762 if (ret) {
4763 mlog_errno(ret);
4764 goto out_commit;
4765 }
4766
4767 ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, meta_ac,
4768 &dealloc);
4769 if (ret) {
4770 mlog_errno(ret);
4771 goto out_commit;
4772 }
4773
4774 le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len);
4775
4776 ret = ocfs2_journal_dirty(handle, root_bh);
4777 if (ret) {
4778 mlog_errno(ret);
4779 goto out_commit;
4780 }
4781
4782 ret = ocfs2_truncate_log_append(osb, handle, blkno, len);
4783 if (ret)
4784 mlog_errno(ret);
4785
4786 out_commit:
4787 ocfs2_commit_trans(osb, handle);
4788 out:
4789 ocfs2_schedule_truncate_log_flush(osb, 1);
4790
4791 mutex_unlock(&tl_inode->i_mutex);
4792
4793 if (meta_ac)
4794 ocfs2_free_alloc_context(meta_ac);
4795
4796 ocfs2_run_deallocs(osb, &dealloc);
4797
4798 return ret;
4799 }
4800
4801 static void ocfs2_xattr_bucket_remove_xs(struct inode *inode,
4802 handle_t *handle,
4803 struct ocfs2_xattr_search *xs)
4804 {
4805 struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket);
4806 struct ocfs2_xattr_entry *last = &xh->xh_entries[
4807 le16_to_cpu(xh->xh_count) - 1];
4808 int ret = 0;
4809
4810 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
4811 OCFS2_JOURNAL_ACCESS_WRITE);
4812 if (ret) {
4813 mlog_errno(ret);
4814 return;
4815 }
4816
4817 /* Remove the old entry. */
4818 memmove(xs->here, xs->here + 1,
4819 (void *)last - (void *)xs->here);
4820 memset(last, 0, sizeof(struct ocfs2_xattr_entry));
4821 le16_add_cpu(&xh->xh_count, -1);
4822
4823 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
4824 }
4825
4826 /*
4827 * Set the xattr name/value in the bucket specified in xs.
4828 *
4829 * As the new value in xi may be stored in the bucket or in an outside cluster,
4830 * we divide the whole process into 3 steps:
4831 * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket)
4832 * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs)
4833 * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside)
4834 * 4. If the clusters for the new outside value can't be allocated, we need
4835 * to free the xattr we allocated in set.
4836 */
4837 static int ocfs2_xattr_set_in_bucket(struct inode *inode,
4838 struct ocfs2_xattr_info *xi,
4839 struct ocfs2_xattr_search *xs,
4840 struct ocfs2_xattr_set_ctxt *ctxt)
4841 {
4842 int ret, local = 1;
4843 size_t value_len;
4844 char *val = (char *)xi->value;
4845 struct ocfs2_xattr_entry *xe = xs->here;
4846 u32 name_hash = ocfs2_xattr_name_hash(inode, xi->name,
4847 strlen(xi->name));
4848
4849 if (!xs->not_found && !ocfs2_xattr_is_local(xe)) {
4850 /*
4851 * We need to truncate the xattr storage first.
4852 *
4853 * If both the old and new value are stored to
4854 * outside block, we only need to truncate
4855 * the storage and then set the value outside.
4856 *
4857 * If the new value should be stored within block,
4858 * we should free all the outside block first and
4859 * the modification to the xattr block will be done
4860 * by following steps.
4861 */
4862 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
4863 value_len = xi->value_len;
4864 else
4865 value_len = 0;
4866
4867 ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
4868 value_len,
4869 ctxt);
4870 if (ret)
4871 goto out;
4872
4873 if (value_len)
4874 goto set_value_outside;
4875 }
4876
4877 value_len = xi->value_len;
4878 /* So we have to handle the inside block change now. */
4879 if (value_len > OCFS2_XATTR_INLINE_SIZE) {
4880 /*
4881 * If the new value will be stored outside of block,
4882 * initalize a new empty value root and insert it first.
4883 */
4884 local = 0;
4885 xi->value = &def_xv;
4886 xi->value_len = OCFS2_XATTR_ROOT_SIZE;
4887 }
4888
4889 ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs,
4890 name_hash, local);
4891 if (ret) {
4892 mlog_errno(ret);
4893 goto out;
4894 }
4895
4896 if (value_len <= OCFS2_XATTR_INLINE_SIZE)
4897 goto out;
4898
4899 /* allocate the space now for the outside block storage. */
4900 ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
4901 value_len, ctxt);
4902 if (ret) {
4903 mlog_errno(ret);
4904
4905 if (xs->not_found) {
4906 /*
4907 * We can't allocate enough clusters for outside
4908 * storage and we have allocated xattr already,
4909 * so need to remove it.
4910 */
4911 ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs);
4912 }
4913 goto out;
4914 }
4915
4916 set_value_outside:
4917 ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle,
4918 xs, val, value_len);
4919 out:
4920 return ret;
4921 }
4922
4923 /*
4924 * check whether the xattr bucket is filled up with the same hash value.
4925 * If we want to insert the xattr with the same hash, return -ENOSPC.
4926 * If we want to insert a xattr with different hash value, go ahead
4927 * and ocfs2_divide_xattr_bucket will handle this.
4928 */
4929 static int ocfs2_check_xattr_bucket_collision(struct inode *inode,
4930 struct ocfs2_xattr_bucket *bucket,
4931 const char *name)
4932 {
4933 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
4934 u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name));
4935
4936 if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash))
4937 return 0;
4938
4939 if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash ==
4940 xh->xh_entries[0].xe_name_hash) {
4941 mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, "
4942 "hash = %u\n",
4943 (unsigned long long)bucket_blkno(bucket),
4944 le32_to_cpu(xh->xh_entries[0].xe_name_hash));
4945 return -ENOSPC;
4946 }
4947
4948 return 0;
4949 }
4950
4951 static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
4952 struct ocfs2_xattr_info *xi,
4953 struct ocfs2_xattr_search *xs,
4954 struct ocfs2_xattr_set_ctxt *ctxt)
4955 {
4956 struct ocfs2_xattr_header *xh;
4957 struct ocfs2_xattr_entry *xe;
4958 u16 count, header_size, xh_free_start;
4959 int free, max_free, need, old;
4960 size_t value_size = 0, name_len = strlen(xi->name);
4961 size_t blocksize = inode->i_sb->s_blocksize;
4962 int ret, allocation = 0;
4963
4964 mlog_entry("Set xattr %s in xattr index block\n", xi->name);
4965
4966 try_again:
4967 xh = xs->header;
4968 count = le16_to_cpu(xh->xh_count);
4969 xh_free_start = le16_to_cpu(xh->xh_free_start);
4970 header_size = sizeof(struct ocfs2_xattr_header) +
4971 count * sizeof(struct ocfs2_xattr_entry);
4972 max_free = OCFS2_XATTR_BUCKET_SIZE -
4973 le16_to_cpu(xh->xh_name_value_len) - header_size;
4974
4975 mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size "
4976 "of %u which exceed block size\n",
4977 (unsigned long long)bucket_blkno(xs->bucket),
4978 header_size);
4979
4980 if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE)
4981 value_size = OCFS2_XATTR_ROOT_SIZE;
4982 else if (xi->value)
4983 value_size = OCFS2_XATTR_SIZE(xi->value_len);
4984
4985 if (xs->not_found)
4986 need = sizeof(struct ocfs2_xattr_entry) +
4987 OCFS2_XATTR_SIZE(name_len) + value_size;
4988 else {
4989 need = value_size + OCFS2_XATTR_SIZE(name_len);
4990
4991 /*
4992 * We only replace the old value if the new length is smaller
4993 * than the old one. Otherwise we will allocate new space in the
4994 * bucket to store it.
4995 */
4996 xe = xs->here;
4997 if (ocfs2_xattr_is_local(xe))
4998 old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
4999 else
5000 old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
5001
5002 if (old >= value_size)
5003 need = 0;
5004 }
5005
5006 free = xh_free_start - header_size;
5007 /*
5008 * We need to make sure the new name/value pair
5009 * can exist in the same block.
5010 */
5011 if (xh_free_start % blocksize < need)
5012 free -= xh_free_start % blocksize;
5013
5014 mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, "
5015 "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len ="
5016 " %u\n", xs->not_found,
5017 (unsigned long long)bucket_blkno(xs->bucket),
5018 free, need, max_free, le16_to_cpu(xh->xh_free_start),
5019 le16_to_cpu(xh->xh_name_value_len));
5020
5021 if (free < need ||
5022 (xs->not_found &&
5023 count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) {
5024 if (need <= max_free &&
5025 count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) {
5026 /*
5027 * We can create the space by defragment. Since only the
5028 * name/value will be moved, the xe shouldn't be changed
5029 * in xs.
5030 */
5031 ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle,
5032 xs->bucket);
5033 if (ret) {
5034 mlog_errno(ret);
5035 goto out;
5036 }
5037
5038 xh_free_start = le16_to_cpu(xh->xh_free_start);
5039 free = xh_free_start - header_size;
5040 if (xh_free_start % blocksize < need)
5041 free -= xh_free_start % blocksize;
5042
5043 if (free >= need)
5044 goto xattr_set;
5045
5046 mlog(0, "Can't get enough space for xattr insert by "
5047 "defragment. Need %u bytes, but we have %d, so "
5048 "allocate new bucket for it.\n", need, free);
5049 }
5050
5051 /*
5052 * We have to add new buckets or clusters and one
5053 * allocation should leave us enough space for insert.
5054 */
5055 BUG_ON(allocation);
5056
5057 /*
5058 * We do not allow for overlapping ranges between buckets. And
5059 * the maximum number of collisions we will allow for then is
5060 * one bucket's worth, so check it here whether we need to
5061 * add a new bucket for the insert.
5062 */
5063 ret = ocfs2_check_xattr_bucket_collision(inode,
5064 xs->bucket,
5065 xi->name);
5066 if (ret) {
5067 mlog_errno(ret);
5068 goto out;
5069 }
5070
5071 ret = ocfs2_add_new_xattr_bucket(inode,
5072 xs->xattr_bh,
5073 xs->bucket->bu_bhs[0],
5074 ctxt);
5075 if (ret) {
5076 mlog_errno(ret);
5077 goto out;
5078 }
5079
5080 ocfs2_xattr_bucket_relse(xs->bucket);
5081
5082 ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh,
5083 xi->name_index,
5084 xi->name, xs);
5085 if (ret && ret != -ENODATA)
5086 goto out;
5087 xs->not_found = ret;
5088 allocation = 1;
5089 goto try_again;
5090 }
5091
5092 xattr_set:
5093 ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt);
5094 out:
5095 mlog_exit(ret);
5096 return ret;
5097 }
5098
5099 static int ocfs2_delete_xattr_in_bucket(struct inode *inode,
5100 struct ocfs2_xattr_bucket *bucket,
5101 void *para)
5102 {
5103 int ret = 0;
5104 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
5105 u16 i;
5106 struct ocfs2_xattr_entry *xe;
5107 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5108 struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,};
5109
5110 ocfs2_init_dealloc_ctxt(&ctxt.dealloc);
5111
5112 ctxt.handle = ocfs2_start_trans(osb, OCFS2_REMOVE_EXTENT_CREDITS);
5113 if (IS_ERR(ctxt.handle)) {
5114 ret = PTR_ERR(ctxt.handle);
5115 mlog_errno(ret);
5116 goto out;
5117 }
5118
5119 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
5120 xe = &xh->xh_entries[i];
5121 if (ocfs2_xattr_is_local(xe))
5122 continue;
5123
5124 ret = ocfs2_xattr_bucket_value_truncate(inode,
5125 bucket->bu_bhs[0],
5126 i, 0, &ctxt);
5127 if (ret) {
5128 mlog_errno(ret);
5129 break;
5130 }
5131 }
5132
5133 ret = ocfs2_commit_trans(osb, ctxt.handle);
5134 ocfs2_schedule_truncate_log_flush(osb, 1);
5135 ocfs2_run_deallocs(osb, &ctxt.dealloc);
5136 out:
5137 return ret;
5138 }
5139
5140 static int ocfs2_delete_xattr_index_block(struct inode *inode,
5141 struct buffer_head *xb_bh)
5142 {
5143 struct ocfs2_xattr_block *xb =
5144 (struct ocfs2_xattr_block *)xb_bh->b_data;
5145 struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list;
5146 int ret = 0;
5147 u32 name_hash = UINT_MAX, e_cpos, num_clusters;
5148 u64 p_blkno;
5149
5150 if (le16_to_cpu(el->l_next_free_rec) == 0)
5151 return 0;
5152
5153 while (name_hash > 0) {
5154 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno,
5155 &e_cpos, &num_clusters, el);
5156 if (ret) {
5157 mlog_errno(ret);
5158 goto out;
5159 }
5160
5161 ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters,
5162 ocfs2_delete_xattr_in_bucket,
5163 NULL);
5164 if (ret) {
5165 mlog_errno(ret);
5166 goto out;
5167 }
5168
5169 ret = ocfs2_rm_xattr_cluster(inode, xb_bh,
5170 p_blkno, e_cpos, num_clusters);
5171 if (ret) {
5172 mlog_errno(ret);
5173 break;
5174 }
5175
5176 if (e_cpos == 0)
5177 break;
5178
5179 name_hash = e_cpos - 1;
5180 }
5181
5182 out:
5183 return ret;
5184 }
5185
5186 /*
5187 * 'security' attributes support
5188 */
5189 static size_t ocfs2_xattr_security_list(struct inode *inode, char *list,
5190 size_t list_size, const char *name,
5191 size_t name_len)
5192 {
5193 const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN;
5194 const size_t total_len = prefix_len + name_len + 1;
5195
5196 if (list && total_len <= list_size) {
5197 memcpy(list, XATTR_SECURITY_PREFIX, prefix_len);
5198 memcpy(list + prefix_len, name, name_len);
5199 list[prefix_len + name_len] = '\0';
5200 }
5201 return total_len;
5202 }
5203
5204 static int ocfs2_xattr_security_get(struct inode *inode, const char *name,
5205 void *buffer, size_t size)
5206 {
5207 if (strcmp(name, "") == 0)
5208 return -EINVAL;
5209 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_SECURITY, name,
5210 buffer, size);
5211 }
5212
5213 static int ocfs2_xattr_security_set(struct inode *inode, const char *name,
5214 const void *value, size_t size, int flags)
5215 {
5216 if (strcmp(name, "") == 0)
5217 return -EINVAL;
5218
5219 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, name, value,
5220 size, flags);
5221 }
5222
5223 int ocfs2_init_security_get(struct inode *inode,
5224 struct inode *dir,
5225 struct ocfs2_security_xattr_info *si)
5226 {
5227 return security_inode_init_security(inode, dir, &si->name, &si->value,
5228 &si->value_len);
5229 }
5230
5231 int ocfs2_init_security_set(handle_t *handle,
5232 struct inode *inode,
5233 struct buffer_head *di_bh,
5234 struct ocfs2_security_xattr_info *si,
5235 struct ocfs2_alloc_context *xattr_ac,
5236 struct ocfs2_alloc_context *data_ac)
5237 {
5238 return ocfs2_xattr_set_handle(handle, inode, di_bh,
5239 OCFS2_XATTR_INDEX_SECURITY,
5240 si->name, si->value, si->value_len, 0,
5241 xattr_ac, data_ac);
5242 }
5243
5244 struct xattr_handler ocfs2_xattr_security_handler = {
5245 .prefix = XATTR_SECURITY_PREFIX,
5246 .list = ocfs2_xattr_security_list,
5247 .get = ocfs2_xattr_security_get,
5248 .set = ocfs2_xattr_security_set,
5249 };
5250
5251 /*
5252 * 'trusted' attributes support
5253 */
5254 static size_t ocfs2_xattr_trusted_list(struct inode *inode, char *list,
5255 size_t list_size, const char *name,
5256 size_t name_len)
5257 {
5258 const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN;
5259 const size_t total_len = prefix_len + name_len + 1;
5260
5261 if (list && total_len <= list_size) {
5262 memcpy(list, XATTR_TRUSTED_PREFIX, prefix_len);
5263 memcpy(list + prefix_len, name, name_len);
5264 list[prefix_len + name_len] = '\0';
5265 }
5266 return total_len;
5267 }
5268
5269 static int ocfs2_xattr_trusted_get(struct inode *inode, const char *name,
5270 void *buffer, size_t size)
5271 {
5272 if (strcmp(name, "") == 0)
5273 return -EINVAL;
5274 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_TRUSTED, name,
5275 buffer, size);
5276 }
5277
5278 static int ocfs2_xattr_trusted_set(struct inode *inode, const char *name,
5279 const void *value, size_t size, int flags)
5280 {
5281 if (strcmp(name, "") == 0)
5282 return -EINVAL;
5283
5284 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_TRUSTED, name, value,
5285 size, flags);
5286 }
5287
5288 struct xattr_handler ocfs2_xattr_trusted_handler = {
5289 .prefix = XATTR_TRUSTED_PREFIX,
5290 .list = ocfs2_xattr_trusted_list,
5291 .get = ocfs2_xattr_trusted_get,
5292 .set = ocfs2_xattr_trusted_set,
5293 };
5294
5295 /*
5296 * 'user' attributes support
5297 */
5298 static size_t ocfs2_xattr_user_list(struct inode *inode, char *list,
5299 size_t list_size, const char *name,
5300 size_t name_len)
5301 {
5302 const size_t prefix_len = XATTR_USER_PREFIX_LEN;
5303 const size_t total_len = prefix_len + name_len + 1;
5304 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5305
5306 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5307 return 0;
5308
5309 if (list && total_len <= list_size) {
5310 memcpy(list, XATTR_USER_PREFIX, prefix_len);
5311 memcpy(list + prefix_len, name, name_len);
5312 list[prefix_len + name_len] = '\0';
5313 }
5314 return total_len;
5315 }
5316
5317 static int ocfs2_xattr_user_get(struct inode *inode, const char *name,
5318 void *buffer, size_t size)
5319 {
5320 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5321
5322 if (strcmp(name, "") == 0)
5323 return -EINVAL;
5324 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5325 return -EOPNOTSUPP;
5326 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_USER, name,
5327 buffer, size);
5328 }
5329
5330 static int ocfs2_xattr_user_set(struct inode *inode, const char *name,
5331 const void *value, size_t size, int flags)
5332 {
5333 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5334
5335 if (strcmp(name, "") == 0)
5336 return -EINVAL;
5337 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5338 return -EOPNOTSUPP;
5339
5340 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_USER, name, value,
5341 size, flags);
5342 }
5343
5344 struct xattr_handler ocfs2_xattr_user_handler = {
5345 .prefix = XATTR_USER_PREFIX,
5346 .list = ocfs2_xattr_user_list,
5347 .get = ocfs2_xattr_user_get,
5348 .set = ocfs2_xattr_user_set,
5349 };
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree