projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2/xattr: fix credits calculation during index create
[linux-block.git] / fs / ocfs2 / xattr.c
CommitLineData
f56654c4
TM		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * xattr.c
5 *
c3cb6827 6 * Copyright (C) 2004, 2008 Oracle. All rights reserved.
f56654c4 7 *
cf1d6c76 8 * CREDITS:
c3cb6827
TY		 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>
cf1d6c76 11 *
f56654c4
TM		 12 * This program is free software; you can redistribute it and/or
13 * modify it under the terms of the GNU General Public
c3cb6827 14 * License version 2 as published by the Free Software Foundation.
f56654c4
TM		 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.
f56654c4
TM		 20 */
21
cf1d6c76
TY		 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>
01225596 34#include <linux/sort.h>
99219aea
MF		 35#include <linux/init.h>
36#include <linux/module.h>
37#include <linux/string.h>
923f7f31 38#include <linux/security.h>
cf1d6c76 39
f56654c4
TM		 40#define MLOG_MASK_PREFIX ML_XATTR
41#include <cluster/masklog.h>
42
43#include "ocfs2.h"
44#include "alloc.h"
d6b32bbb 45#include "blockcheck.h"
f56654c4
TM		 46#include "dlmglue.h"
47#include "file.h"
cf1d6c76
TY		 48#include "symlink.h"
49#include "sysfile.h"
f56654c4
TM		 50#include "inode.h"
51#include "journal.h"
52#include "ocfs2_fs.h"
53#include "suballoc.h"
54#include "uptodate.h"
55#include "buffer_head_io.h"
0c044f0b 56#include "super.h"
cf1d6c76
TY		 57#include "xattr.h"
58
59
60struct ocfs2_xattr_def_value_root {
61 struct ocfs2_xattr_value_root xv;
62 struct ocfs2_extent_rec er;
63};
64
0c044f0b 65struct ocfs2_xattr_bucket {
ba937127
JB		 66 /* The inode these xattrs are associated with */
67 struct inode *bu_inode;
68
69 /* The actual buffers that make up the bucket */
4ac6032d 70 struct buffer_head *bu_bhs[OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET];
ba937127
JB		 71
72 /* How many blocks make up one bucket for this filesystem */
73 int bu_blocks;
0c044f0b
TM		 74};
75
78f30c31 76struct ocfs2_xattr_set_ctxt {
85db90e7 77 handle_t *handle;
78f30c31
TM		 78 struct ocfs2_alloc_context *meta_ac;
79 struct ocfs2_alloc_context *data_ac;
80 struct ocfs2_cached_dealloc_ctxt dealloc;
81};
82
cf1d6c76
TY		 83#define OCFS2_XATTR_ROOT_SIZE (sizeof(struct ocfs2_xattr_def_value_root))
84#define OCFS2_XATTR_INLINE_SIZE 80
534eaddd
TY		 85#define OCFS2_XATTR_FREE_IN_IBODY (OCFS2_MIN_XATTR_INLINE_SIZE \
86 - sizeof(struct ocfs2_xattr_header) \
87 - sizeof(__u32))
89c38bd0
TY		 88#define OCFS2_XATTR_FREE_IN_BLOCK(ptr) ((ptr)->i_sb->s_blocksize \
89 - sizeof(struct ocfs2_xattr_block) \
90 - sizeof(struct ocfs2_xattr_header) \
91 - sizeof(__u32))
cf1d6c76
TY		 92
93static struct ocfs2_xattr_def_value_root def_xv = {
94 .xv.xr_list.l_count = cpu_to_le16(1),
95};
96
97struct xattr_handler *ocfs2_xattr_handlers[] = {
98 &ocfs2_xattr_user_handler,
929fb014
TY		 99#ifdef CONFIG_OCFS2_FS_POSIX_ACL
100 &ocfs2_xattr_acl_access_handler,
101 &ocfs2_xattr_acl_default_handler,
102#endif
cf1d6c76 103 &ocfs2_xattr_trusted_handler,
923f7f31 104 &ocfs2_xattr_security_handler,
cf1d6c76
TY		 105 NULL
106};
107
c988fd04 108static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = {
cf1d6c76 109 [OCFS2_XATTR_INDEX_USER] = &ocfs2_xattr_user_handler,
929fb014
TY		 110#ifdef CONFIG_OCFS2_FS_POSIX_ACL
111 [OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS]
112 = &ocfs2_xattr_acl_access_handler,
113 [OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT]
114 = &ocfs2_xattr_acl_default_handler,
115#endif
cf1d6c76 116 [OCFS2_XATTR_INDEX_TRUSTED] = &ocfs2_xattr_trusted_handler,
923f7f31 117 [OCFS2_XATTR_INDEX_SECURITY] = &ocfs2_xattr_security_handler,
cf1d6c76
TY		 118};
119
120struct ocfs2_xattr_info {
121 int name_index;
122 const char *name;
123 const void *value;
124 size_t value_len;
125};
126
127struct ocfs2_xattr_search {
128 struct buffer_head *inode_bh;
129 /*
130 * xattr_bh point to the block buffer head which has extended attribute
131 * when extended attribute in inode, xattr_bh is equal to inode_bh.
132 */
133 struct buffer_head *xattr_bh;
134 struct ocfs2_xattr_header *header;
ba937127 135 struct ocfs2_xattr_bucket *bucket;
cf1d6c76
TY		 136 void *base;
137 void *end;
138 struct ocfs2_xattr_entry *here;
139 int not_found;
140};
141
589dc260
TM		 142static int ocfs2_xattr_bucket_get_name_value(struct inode *inode,
143 struct ocfs2_xattr_header *xh,
144 int index,
145 int *block_off,
146 int *new_offset);
147
54f443f4
JB		 148static int ocfs2_xattr_block_find(struct inode *inode,
149 int name_index,
150 const char *name,
151 struct ocfs2_xattr_search *xs);
589dc260
TM		 152static int ocfs2_xattr_index_block_find(struct inode *inode,
153 struct buffer_head *root_bh,
154 int name_index,
155 const char *name,
156 struct ocfs2_xattr_search *xs);
157
0c044f0b
TM		 158static int ocfs2_xattr_tree_list_index_block(struct inode *inode,
159 struct ocfs2_xattr_tree_root *xt,
160 char *buffer,
161 size_t buffer_size);
162
01225596 163static int ocfs2_xattr_create_index_block(struct inode *inode,
78f30c31
TM		 164 struct ocfs2_xattr_search *xs,
165 struct ocfs2_xattr_set_ctxt *ctxt);
01225596
TM		 166
167static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
168 struct ocfs2_xattr_info *xi,
78f30c31
TM		 169 struct ocfs2_xattr_search *xs,
170 struct ocfs2_xattr_set_ctxt *ctxt);
01225596 171
a3944256
TM		 172static int ocfs2_delete_xattr_index_block(struct inode *inode,
173 struct buffer_head *xb_bh);
c58b6032
JB		 174static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
175 u64 src_blk, u64 last_blk, u64 to_blk,
176 unsigned int start_bucket,
177 u32 *first_hash);
a3944256 178
0030e001
TY		 179static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb)
180{
181 return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE;
182}
183
184static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb)
185{
186 return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits);
187}
188
189static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb)
190{
191 u16 len = sb->s_blocksize -
192 offsetof(struct ocfs2_xattr_header, xh_entries);
193
194 return len / sizeof(struct ocfs2_xattr_entry);
195}
196
9c7759aa 197#define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr)
51def39f 198#define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data)
3e632946 199#define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0))
9c7759aa 200
ba937127 201static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode)
6dde41d9 202{
ba937127
JB		 203 struct ocfs2_xattr_bucket *bucket;
204 int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
6dde41d9 205
ba937127
JB		 206 BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET);
207
208 bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS);
209 if (bucket) {
210 bucket->bu_inode = inode;
211 bucket->bu_blocks = blks;
212 }
213
214 return bucket;
215}
216
217static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket)
218{
219 int i;
220
221 for (i = 0; i < bucket->bu_blocks; i++) {
6dde41d9
JB		 222 brelse(bucket->bu_bhs[i]);
223 bucket->bu_bhs[i] = NULL;
224 }
225}
226
ba937127
JB		 227static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket)
228{
229 if (bucket) {
230 ocfs2_xattr_bucket_relse(bucket);
231 bucket->bu_inode = NULL;
232 kfree(bucket);
233 }
234}
235
784b816a
JB		 236/*
237 * A bucket that has never been written to disk doesn't need to be
238 * read. We just need the buffer_heads. Don't call this for
239 * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes
240 * them fully.
241 */
ba937127 242static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
784b816a
JB		 243 u64 xb_blkno)
244{
245 int i, rc = 0;
784b816a 246
ba937127
JB		 247 for (i = 0; i < bucket->bu_blocks; i++) {
248 bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb,
249 xb_blkno + i);
784b816a
JB		 250 if (!bucket->bu_bhs[i]) {
251 rc = -EIO;
252 mlog_errno(rc);
253 break;
254 }
255
757055ad
TM		 256 if (!ocfs2_buffer_uptodate(bucket->bu_inode,
257 bucket->bu_bhs[i]))
258 ocfs2_set_new_buffer_uptodate(bucket->bu_inode,
259 bucket->bu_bhs[i]);
784b816a
JB		 260 }
261
262 if (rc)
ba937127 263 ocfs2_xattr_bucket_relse(bucket);
784b816a
JB		 264 return rc;
265}
266
267/* Read the xattr bucket at xb_blkno */
ba937127 268static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket,
784b816a
JB		 269 u64 xb_blkno)
270{
ba937127 271 int rc;
784b816a 272
ba937127 273 rc = ocfs2_read_blocks(bucket->bu_inode, xb_blkno,
970e4936
JB		 274 bucket->bu_blocks, bucket->bu_bhs, 0,
275 NULL);
4d0e214e
JB		 276 if (!rc) {
277 rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb,
278 bucket->bu_bhs,
279 bucket->bu_blocks,
280 &bucket_xh(bucket)->xh_check);
281 if (rc)
282 mlog_errno(rc);
283 }
284
784b816a 285 if (rc)
ba937127 286 ocfs2_xattr_bucket_relse(bucket);
784b816a
JB		 287 return rc;
288}
289
1224be02 290static int ocfs2_xattr_bucket_journal_access(handle_t *handle,
1224be02
JB		 291 struct ocfs2_xattr_bucket *bucket,
292 int type)
293{
294 int i, rc = 0;
1224be02 295
ba937127
JB		 296 for (i = 0; i < bucket->bu_blocks; i++) {
297 rc = ocfs2_journal_access(handle, bucket->bu_inode,
1224be02
JB		 298 bucket->bu_bhs[i], type);
299 if (rc) {
300 mlog_errno(rc);
301 break;
302 }
303 }
304
305 return rc;
306}
307
308static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle,
1224be02
JB		 309 struct ocfs2_xattr_bucket *bucket)
310{
ba937127 311 int i;
1224be02 312
4d0e214e
JB		 313 ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb,
314 bucket->bu_bhs, bucket->bu_blocks,
315 &bucket_xh(bucket)->xh_check);
316
ba937127 317 for (i = 0; i < bucket->bu_blocks; i++)
1224be02
JB		 318 ocfs2_journal_dirty(handle, bucket->bu_bhs[i]);
319}
320
ba937127 321static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest,
4980c6da
JB		 322 struct ocfs2_xattr_bucket *src)
323{
324 int i;
ba937127
JB		 325 int blocksize = src->bu_inode->i_sb->s_blocksize;
326
327 BUG_ON(dest->bu_blocks != src->bu_blocks);
328 BUG_ON(dest->bu_inode != src->bu_inode);
4980c6da 329
ba937127 330 for (i = 0; i < src->bu_blocks; i++) {
4980c6da
JB		 331 memcpy(bucket_block(dest, i), bucket_block(src, i),
332 blocksize);
333 }
334}
1224be02 335
4ae1d69b
JB		 336static int ocfs2_validate_xattr_block(struct super_block *sb,
337 struct buffer_head *bh)
338{
d6b32bbb 339 int rc;
4ae1d69b
JB		 340 struct ocfs2_xattr_block *xb =
341 (struct ocfs2_xattr_block *)bh->b_data;
342
343 mlog(0, "Validating xattr block %llu\n",
344 (unsigned long long)bh->b_blocknr);
345
d6b32bbb
JB		 346 BUG_ON(!buffer_uptodate(bh));
347
348 /*
349 * If the ecc fails, we return the error but otherwise
350 * leave the filesystem running. We know any error is
351 * local to this block.
352 */
353 rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check);
354 if (rc)
355 return rc;
356
357 /*
358 * Errors after here are fatal
359 */
360
4ae1d69b
JB		 361 if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) {
362 ocfs2_error(sb,
363 "Extended attribute block #%llu has bad "
364 "signature %.*s",
365 (unsigned long long)bh->b_blocknr, 7,
366 xb->xb_signature);
367 return -EINVAL;
368 }
369
370 if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) {
371 ocfs2_error(sb,
372 "Extended attribute block #%llu has an "
373 "invalid xb_blkno of %llu",
374 (unsigned long long)bh->b_blocknr,
375 (unsigned long long)le64_to_cpu(xb->xb_blkno));
376 return -EINVAL;
377 }
378
379 if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) {
380 ocfs2_error(sb,
381 "Extended attribute block #%llu has an invalid "
382 "xb_fs_generation of #%u",
383 (unsigned long long)bh->b_blocknr,
384 le32_to_cpu(xb->xb_fs_generation));
385 return -EINVAL;
386 }
387
388 return 0;
389}
390
391static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno,
392 struct buffer_head **bh)
393{
394 int rc;
395 struct buffer_head *tmp = *bh;
396
970e4936
JB		 397 rc = ocfs2_read_block(inode, xb_blkno, &tmp,
398 ocfs2_validate_xattr_block);
4ae1d69b
JB		 399
400 /* If ocfs2_read_block() got us a new bh, pass it up. */
401 if (!rc && !*bh)
402 *bh = tmp;
403
404 return rc;
405}
406
936b8834 407static inline const char *ocfs2_xattr_prefix(int name_index)
cf1d6c76
TY		 408{
409 struct xattr_handler *handler = NULL;
410
411 if (name_index > 0 && name_index < OCFS2_XATTR_MAX)
412 handler = ocfs2_xattr_handler_map[name_index];
413
936b8834 414 return handler ? handler->prefix : NULL;
cf1d6c76
TY		 415}
416
40daa16a 417static u32 ocfs2_xattr_name_hash(struct inode *inode,
2057e5c6 418 const char *name,
40daa16a 419 int name_len)
cf1d6c76
TY		 420{
421 /* Get hash value of uuid from super block */
422 u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash;
423 int i;
424
cf1d6c76
TY		 425 /* hash extended attribute name */
426 for (i = 0; i < name_len; i++) {
427 hash = (hash << OCFS2_HASH_SHIFT) ^
428 (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^
429 *name++;
430 }
431
432 return hash;
433}
434
435/*
436 * ocfs2_xattr_hash_entry()
437 *
438 * Compute the hash of an extended attribute.
439 */
440static void ocfs2_xattr_hash_entry(struct inode *inode,
441 struct ocfs2_xattr_header *header,
442 struct ocfs2_xattr_entry *entry)
443{
444 u32 hash = 0;
cf1d6c76 445 char *name = (char *)header + le16_to_cpu(entry->xe_name_offset);
cf1d6c76 446
2057e5c6 447 hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len);
cf1d6c76
TY		 448 entry->xe_name_hash = cpu_to_le32(hash);
449
450 return;
451}
f56654c4 452
534eaddd
TY		 453static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len)
454{
455 int size = 0;
456
457 if (value_len <= OCFS2_XATTR_INLINE_SIZE)
458 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len);
459 else
460 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
461 size += sizeof(struct ocfs2_xattr_entry);
462
463 return size;
464}
465
466int ocfs2_calc_security_init(struct inode *dir,
467 struct ocfs2_security_xattr_info *si,
468 int *want_clusters,
469 int *xattr_credits,
470 struct ocfs2_alloc_context **xattr_ac)
471{
472 int ret = 0;
473 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
474 int s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
475 si->value_len);
476
477 /*
478 * The max space of security xattr taken inline is
479 * 256(name) + 80(value) + 16(entry) = 352 bytes,
480 * So reserve one metadata block for it is ok.
481 */
482 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
483 s_size > OCFS2_XATTR_FREE_IN_IBODY) {
484 ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac);
485 if (ret) {
486 mlog_errno(ret);
487 return ret;
488 }
489 *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
490 }
491
492 /* reserve clusters for xattr value which will be set in B tree*/
493 if (si->value_len > OCFS2_XATTR_INLINE_SIZE)
494 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
495 si->value_len);
496 return ret;
497}
498
89c38bd0
TY		 499int ocfs2_calc_xattr_init(struct inode *dir,
500 struct buffer_head *dir_bh,
501 int mode,
502 struct ocfs2_security_xattr_info *si,
503 int *want_clusters,
504 int *xattr_credits,
505 struct ocfs2_alloc_context **xattr_ac)
506{
507 int ret = 0;
508 struct ocfs2_super *osb = OCFS2_SB(dir->i_sb);
509 int s_size = 0;
510 int a_size = 0;
511 int acl_len = 0;
512
513 if (si->enable)
514 s_size = ocfs2_xattr_entry_real_size(strlen(si->name),
515 si->value_len);
516
517 if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) {
518 acl_len = ocfs2_xattr_get_nolock(dir, dir_bh,
519 OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT,
520 "", NULL, 0);
521 if (acl_len > 0) {
522 a_size = ocfs2_xattr_entry_real_size(0, acl_len);
523 if (S_ISDIR(mode))
524 a_size <<= 1;
525 } else if (acl_len != 0 && acl_len != -ENODATA) {
526 mlog_errno(ret);
527 return ret;
528 }
529 }
530
531 if (!(s_size + a_size))
532 return ret;
533
534 /*
535 * The max space of security xattr taken inline is
536 * 256(name) + 80(value) + 16(entry) = 352 bytes,
537 * The max space of acl xattr taken inline is
538 * 80(value) + 16(entry) * 2(if directory) = 192 bytes,
539 * when blocksize = 512, may reserve one more cluser for
540 * xattr bucket, otherwise reserve one metadata block
541 * for them is ok.
542 */
543 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE ||
544 (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) {
545 ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac);
546 if (ret) {
547 mlog_errno(ret);
548 return ret;
549 }
550 *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
551 }
552
553 if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE &&
554 (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) {
555 *want_clusters += 1;
556 *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb);
557 }
558
559 /* reserve clusters for xattr value which will be set in B tree*/
560 if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE)
561 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
562 si->value_len);
563 if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL &&
564 acl_len > OCFS2_XATTR_INLINE_SIZE) {
565 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb, acl_len);
566 if (S_ISDIR(mode))
567 *want_clusters += ocfs2_clusters_for_bytes(dir->i_sb,
568 acl_len);
569 }
570
571 return ret;
572}
573
f56654c4
TM		 574static int ocfs2_xattr_extend_allocation(struct inode *inode,
575 u32 clusters_to_add,
19b801f4 576 struct ocfs2_xattr_value_buf *vb,
78f30c31 577 struct ocfs2_xattr_set_ctxt *ctxt)
f56654c4
TM		 578{
579 int status = 0;
85db90e7 580 handle_t *handle = ctxt->handle;
f56654c4
TM		 581 enum ocfs2_alloc_restarted why;
582 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
19b801f4 583 u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters);
f99b9b7c 584 struct ocfs2_extent_tree et;
f56654c4
TM		 585
586 mlog(0, "(clusters_to_add for xattr= %u)\n", clusters_to_add);
587
19b801f4 588 ocfs2_init_xattr_value_extent_tree(&et, inode, vb);
f99b9b7c 589
19b801f4 590 status = vb->vb_access(handle, inode, vb->vb_bh,
2a50a743 591 OCFS2_JOURNAL_ACCESS_WRITE);
f56654c4
TM		 592 if (status < 0) {
593 mlog_errno(status);
594 goto leave;
595 }
596
19b801f4 597 prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);
f56654c4
TM		 598 status = ocfs2_add_clusters_in_btree(osb,
599 inode,
600 &logical_start,
601 clusters_to_add,
602 0,
f99b9b7c 603 &et,
f56654c4 604 handle,
78f30c31
TM		 605 ctxt->data_ac,
606 ctxt->meta_ac,
f99b9b7c 607 &why);
85db90e7
TM		 608 if (status < 0) {
609 mlog_errno(status);
f56654c4
TM		 610 goto leave;
611 }
612
19b801f4 613 status = ocfs2_journal_dirty(handle, vb->vb_bh);
f56654c4
TM		 614 if (status < 0) {
615 mlog_errno(status);
616 goto leave;
617 }
618
19b801f4 619 clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters;
f56654c4 620
85db90e7
TM		 621 /*
622 * We should have already allocated enough space before the transaction,
623 * so no need to restart.
624 */
625 BUG_ON(why != RESTART_NONE || clusters_to_add);
f56654c4
TM		 626
627leave:
f56654c4
TM		 628
629 return status;
630}
631
632static int __ocfs2_remove_xattr_range(struct inode *inode,
d72cc72d 633 struct ocfs2_xattr_value_buf *vb,
f56654c4 634 u32 cpos, u32 phys_cpos, u32 len,
78f30c31 635 struct ocfs2_xattr_set_ctxt *ctxt)
f56654c4
TM		 636{
637 int ret;
638 u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
85db90e7 639 handle_t *handle = ctxt->handle;
f99b9b7c
JB		 640 struct ocfs2_extent_tree et;
641
d72cc72d 642 ocfs2_init_xattr_value_extent_tree(&et, inode, vb);
f56654c4 643
d72cc72d
JB		 644 ret = vb->vb_access(handle, inode, vb->vb_bh,
645 OCFS2_JOURNAL_ACCESS_WRITE);
f56654c4
TM		 646 if (ret) {
647 mlog_errno(ret);
85db90e7 648 goto out;
f56654c4
TM		 649 }
650
78f30c31
TM		 651 ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, ctxt->meta_ac,
652 &ctxt->dealloc);
f56654c4
TM		 653 if (ret) {
654 mlog_errno(ret);
85db90e7 655 goto out;
f56654c4
TM		 656 }
657
d72cc72d 658 le32_add_cpu(&vb->vb_xv->xr_clusters, -len);
f56654c4 659
d72cc72d 660 ret = ocfs2_journal_dirty(handle, vb->vb_bh);
f56654c4
TM		 661 if (ret) {
662 mlog_errno(ret);
85db90e7 663 goto out;
f56654c4
TM		 664 }
665
78f30c31 666 ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, phys_blkno, len);
f56654c4
TM		 667 if (ret)
668 mlog_errno(ret);
669
f56654c4 670out:
f56654c4
TM		 671 return ret;
672}
673
674static int ocfs2_xattr_shrink_size(struct inode *inode,
675 u32 old_clusters,
676 u32 new_clusters,
19b801f4 677 struct ocfs2_xattr_value_buf *vb,
78f30c31 678 struct ocfs2_xattr_set_ctxt *ctxt)
f56654c4
TM		 679{
680 int ret = 0;
681 u32 trunc_len, cpos, phys_cpos, alloc_size;
682 u64 block;
f56654c4
TM		 683
684 if (old_clusters <= new_clusters)
685 return 0;
686
687 cpos = new_clusters;
688 trunc_len = old_clusters - new_clusters;
689 while (trunc_len) {
690 ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos,
d72cc72d 691 &alloc_size,
19b801f4 692 &vb->vb_xv->xr_list);
f56654c4
TM		 693 if (ret) {
694 mlog_errno(ret);
695 goto out;
696 }
697
698 if (alloc_size > trunc_len)
699 alloc_size = trunc_len;
700
19b801f4 701 ret = __ocfs2_remove_xattr_range(inode, vb, cpos,
f56654c4 702 phys_cpos, alloc_size,
78f30c31 703 ctxt);
f56654c4
TM		 704 if (ret) {
705 mlog_errno(ret);
706 goto out;
707 }
708
709 block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos);
710 ocfs2_remove_xattr_clusters_from_cache(inode, block,
711 alloc_size);
712 cpos += alloc_size;
713 trunc_len -= alloc_size;
714 }
715
716out:
f56654c4
TM		 717 return ret;
718}
719
720static int ocfs2_xattr_value_truncate(struct inode *inode,
b3e5d379 721 struct ocfs2_xattr_value_buf *vb,
78f30c31
TM		 722 int len,
723 struct ocfs2_xattr_set_ctxt *ctxt)
f56654c4
TM		 724{
725 int ret;
726 u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len);
b3e5d379 727 u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters);
f56654c4
TM		 728
729 if (new_clusters == old_clusters)
730 return 0;
731
732 if (new_clusters > old_clusters)
733 ret = ocfs2_xattr_extend_allocation(inode,
734 new_clusters - old_clusters,
b3e5d379 735 vb, ctxt);
f56654c4
TM		 736 else
737 ret = ocfs2_xattr_shrink_size(inode,
738 old_clusters, new_clusters,
b3e5d379 739 vb, ctxt);
f56654c4
TM		 740
741 return ret;
742}
cf1d6c76 743
936b8834
TM		 744static int ocfs2_xattr_list_entry(char *buffer, size_t size,
745 size_t *result, const char *prefix,
746 const char *name, int name_len)
747{
748 char *p = buffer + *result;
749 int prefix_len = strlen(prefix);
750 int total_len = prefix_len + name_len + 1;
751
752 *result += total_len;
753
754 /* we are just looking for how big our buffer needs to be */
755 if (!size)
756 return 0;
757
758 if (*result > size)
759 return -ERANGE;
760
761 memcpy(p, prefix, prefix_len);
762 memcpy(p + prefix_len, name, name_len);
763 p[prefix_len + name_len] = '\0';
764
765 return 0;
766}
767
cf1d6c76
TY		 768static int ocfs2_xattr_list_entries(struct inode *inode,
769 struct ocfs2_xattr_header *header,
770 char *buffer, size_t buffer_size)
771{
936b8834
TM		 772 size_t result = 0;
773 int i, type, ret;
774 const char *prefix, *name;
cf1d6c76
TY		 775
776 for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) {
777 struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
936b8834
TM		 778 type = ocfs2_xattr_get_type(entry);
779 prefix = ocfs2_xattr_prefix(type);
780
781 if (prefix) {
782 name = (const char *)header +
783 le16_to_cpu(entry->xe_name_offset);
784
785 ret = ocfs2_xattr_list_entry(buffer, buffer_size,
786 &result, prefix, name,
787 entry->xe_name_len);
788 if (ret)
789 return ret;
cf1d6c76
TY		 790 }
791 }
792
936b8834 793 return result;
cf1d6c76
TY		 794}
795
796static int ocfs2_xattr_ibody_list(struct inode *inode,
797 struct ocfs2_dinode *di,
798 char *buffer,
799 size_t buffer_size)
800{
801 struct ocfs2_xattr_header *header = NULL;
802 struct ocfs2_inode_info *oi = OCFS2_I(inode);
803 int ret = 0;
804
805 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
806 return ret;
807
808 header = (struct ocfs2_xattr_header *)
809 ((void *)di + inode->i_sb->s_blocksize -
810 le16_to_cpu(di->i_xattr_inline_size));
811
812 ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size);
813
814 return ret;
815}
816
817static int ocfs2_xattr_block_list(struct inode *inode,
818 struct ocfs2_dinode *di,
819 char *buffer,
820 size_t buffer_size)
821{
822 struct buffer_head *blk_bh = NULL;
0c044f0b 823 struct ocfs2_xattr_block *xb;
cf1d6c76
TY		 824 int ret = 0;
825
826 if (!di->i_xattr_loc)
827 return ret;
828
4ae1d69b
JB		 829 ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
830 &blk_bh);
cf1d6c76
TY		 831 if (ret < 0) {
832 mlog_errno(ret);
833 return ret;
834 }
cf1d6c76 835
0c044f0b 836 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
0c044f0b
TM		 837 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
838 struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header;
839 ret = ocfs2_xattr_list_entries(inode, header,
840 buffer, buffer_size);
841 } else {
842 struct ocfs2_xattr_tree_root *xt = &xb->xb_attrs.xb_root;
843 ret = ocfs2_xattr_tree_list_index_block(inode, xt,
844 buffer, buffer_size);
845 }
4ae1d69b 846
cf1d6c76
TY		 847 brelse(blk_bh);
848
849 return ret;
850}
851
852ssize_t ocfs2_listxattr(struct dentry *dentry,
853 char *buffer,
854 size_t size)
855{
856 int ret = 0, i_ret = 0, b_ret = 0;
857 struct buffer_head *di_bh = NULL;
858 struct ocfs2_dinode *di = NULL;
859 struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode);
860
8154da3d
TY		 861 if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb)))
862 return -EOPNOTSUPP;
863
cf1d6c76
TY		 864 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
865 return ret;
866
867 ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0);
868 if (ret < 0) {
869 mlog_errno(ret);
870 return ret;
871 }
872
873 di = (struct ocfs2_dinode *)di_bh->b_data;
874
875 down_read(&oi->ip_xattr_sem);
876 i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size);
877 if (i_ret < 0)
878 b_ret = 0;
879 else {
880 if (buffer) {
881 buffer += i_ret;
882 size -= i_ret;
883 }
884 b_ret = ocfs2_xattr_block_list(dentry->d_inode, di,
885 buffer, size);
886 if (b_ret < 0)
887 i_ret = 0;
888 }
889 up_read(&oi->ip_xattr_sem);
890 ocfs2_inode_unlock(dentry->d_inode, 0);
891
892 brelse(di_bh);
893
894 return i_ret + b_ret;
895}
896
897static int ocfs2_xattr_find_entry(int name_index,
898 const char *name,
899 struct ocfs2_xattr_search *xs)
900{
901 struct ocfs2_xattr_entry *entry;
902 size_t name_len;
903 int i, cmp = 1;
904
905 if (name == NULL)
906 return -EINVAL;
907
908 name_len = strlen(name);
909 entry = xs->here;
910 for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
911 cmp = name_index - ocfs2_xattr_get_type(entry);
912 if (!cmp)
913 cmp = name_len - entry->xe_name_len;
914 if (!cmp)
915 cmp = memcmp(name, (xs->base +
916 le16_to_cpu(entry->xe_name_offset)),
917 name_len);
918 if (cmp == 0)
919 break;
920 entry += 1;
921 }
922 xs->here = entry;
923
924 return cmp ? -ENODATA : 0;
925}
926
927static int ocfs2_xattr_get_value_outside(struct inode *inode,
589dc260 928 struct ocfs2_xattr_value_root *xv,
cf1d6c76
TY		 929 void *buffer,
930 size_t len)
931{
932 u32 cpos, p_cluster, num_clusters, bpc, clusters;
933 u64 blkno;
934 int i, ret = 0;
935 size_t cplen, blocksize;
936 struct buffer_head *bh = NULL;
cf1d6c76
TY		 937 struct ocfs2_extent_list *el;
938
cf1d6c76
TY		 939 el = &xv->xr_list;
940 clusters = le32_to_cpu(xv->xr_clusters);
941 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
942 blocksize = inode->i_sb->s_blocksize;
943
944 cpos = 0;
945 while (cpos < clusters) {
946 ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
947 &num_clusters, el);
948 if (ret) {
949 mlog_errno(ret);
950 goto out;
951 }
952
953 blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
954 /* Copy ocfs2_xattr_value */
955 for (i = 0; i < num_clusters * bpc; i++, blkno++) {
970e4936 956 ret = ocfs2_read_block(inode, blkno, &bh, NULL);
cf1d6c76
TY		 957 if (ret) {
958 mlog_errno(ret);
959 goto out;
960 }
961
962 cplen = len >= blocksize ? blocksize : len;
963 memcpy(buffer, bh->b_data, cplen);
964 len -= cplen;
965 buffer += cplen;
966
967 brelse(bh);
968 bh = NULL;
969 if (len == 0)
970 break;
971 }
972 cpos += num_clusters;
973 }
974out:
975 return ret;
976}
977
978static int ocfs2_xattr_ibody_get(struct inode *inode,
979 int name_index,
980 const char *name,
981 void *buffer,
982 size_t buffer_size,
983 struct ocfs2_xattr_search *xs)
984{
985 struct ocfs2_inode_info *oi = OCFS2_I(inode);
986 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
589dc260 987 struct ocfs2_xattr_value_root *xv;
cf1d6c76
TY		 988 size_t size;
989 int ret = 0;
990
991 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL))
992 return -ENODATA;
993
994 xs->end = (void *)di + inode->i_sb->s_blocksize;
995 xs->header = (struct ocfs2_xattr_header *)
996 (xs->end - le16_to_cpu(di->i_xattr_inline_size));
997 xs->base = (void *)xs->header;
998 xs->here = xs->header->xh_entries;
999
1000 ret = ocfs2_xattr_find_entry(name_index, name, xs);
1001 if (ret)
1002 return ret;
1003 size = le64_to_cpu(xs->here->xe_value_size);
1004 if (buffer) {
1005 if (size > buffer_size)
1006 return -ERANGE;
1007 if (ocfs2_xattr_is_local(xs->here)) {
1008 memcpy(buffer, (void *)xs->base +
1009 le16_to_cpu(xs->here->xe_name_offset) +
1010 OCFS2_XATTR_SIZE(xs->here->xe_name_len), size);
1011 } else {
589dc260
TM		 1012 xv = (struct ocfs2_xattr_value_root *)
1013 (xs->base + le16_to_cpu(
1014 xs->here->xe_name_offset) +
1015 OCFS2_XATTR_SIZE(xs->here->xe_name_len));
1016 ret = ocfs2_xattr_get_value_outside(inode, xv,
cf1d6c76
TY		 1017 buffer, size);
1018 if (ret < 0) {
1019 mlog_errno(ret);
1020 return ret;
1021 }
1022 }
1023 }
1024
1025 return size;
1026}
1027
1028static int ocfs2_xattr_block_get(struct inode *inode,
1029 int name_index,
1030 const char *name,
1031 void *buffer,
1032 size_t buffer_size,
1033 struct ocfs2_xattr_search *xs)
1034{
cf1d6c76 1035 struct ocfs2_xattr_block *xb;
589dc260 1036 struct ocfs2_xattr_value_root *xv;
cf1d6c76 1037 size_t size;
589dc260 1038 int ret = -ENODATA, name_offset, name_len, block_off, i;
cf1d6c76 1039
ba937127
JB		 1040 xs->bucket = ocfs2_xattr_bucket_new(inode);
1041 if (!xs->bucket) {
1042 ret = -ENOMEM;
1043 mlog_errno(ret);
1044 goto cleanup;
1045 }
589dc260 1046
54f443f4
JB		 1047 ret = ocfs2_xattr_block_find(inode, name_index, name, xs);
1048 if (ret) {
cf1d6c76 1049 mlog_errno(ret);
cf1d6c76
TY		 1050 goto cleanup;
1051 }
1052
6c1e183e
TY		 1053 if (xs->not_found) {
1054 ret = -ENODATA;
1055 goto cleanup;
1056 }
1057
54f443f4 1058 xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
cf1d6c76
TY		 1059 size = le64_to_cpu(xs->here->xe_value_size);
1060 if (buffer) {
1061 ret = -ERANGE;
1062 if (size > buffer_size)
1063 goto cleanup;
589dc260
TM		 1064
1065 name_offset = le16_to_cpu(xs->here->xe_name_offset);
1066 name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len);
1067 i = xs->here - xs->header->xh_entries;
1068
1069 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
1070 ret = ocfs2_xattr_bucket_get_name_value(inode,
ba937127 1071 bucket_xh(xs->bucket),
589dc260
TM		 1072 i,
1073 &block_off,
1074 &name_offset);
ba937127 1075 xs->base = bucket_block(xs->bucket, block_off);
589dc260 1076 }
cf1d6c76
TY		 1077 if (ocfs2_xattr_is_local(xs->here)) {
1078 memcpy(buffer, (void *)xs->base +
589dc260 1079 name_offset + name_len, size);
cf1d6c76 1080 } else {
589dc260
TM		 1081 xv = (struct ocfs2_xattr_value_root *)
1082 (xs->base + name_offset + name_len);
1083 ret = ocfs2_xattr_get_value_outside(inode, xv,
cf1d6c76
TY		 1084 buffer, size);
1085 if (ret < 0) {
1086 mlog_errno(ret);
1087 goto cleanup;
1088 }
1089 }
1090 }
1091 ret = size;
1092cleanup:
ba937127 1093 ocfs2_xattr_bucket_free(xs->bucket);
cf1d6c76 1094
54f443f4
JB		 1095 brelse(xs->xattr_bh);
1096 xs->xattr_bh = NULL;
cf1d6c76
TY		 1097 return ret;
1098}
1099
4e3e9d02
TY		 1100int ocfs2_xattr_get_nolock(struct inode *inode,
1101 struct buffer_head *di_bh,
0030e001
TY		 1102 int name_index,
1103 const char *name,
1104 void *buffer,
1105 size_t buffer_size)
cf1d6c76
TY		 1106{
1107 int ret;
1108 struct ocfs2_dinode *di = NULL;
cf1d6c76
TY		 1109 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1110 struct ocfs2_xattr_search xis = {
1111 .not_found = -ENODATA,
1112 };
1113 struct ocfs2_xattr_search xbs = {
1114 .not_found = -ENODATA,
1115 };
1116
8154da3d
TY		 1117 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
1118 return -EOPNOTSUPP;
1119
cf1d6c76
TY		 1120 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
1121 ret = -ENODATA;
1122
cf1d6c76
TY		 1123 xis.inode_bh = xbs.inode_bh = di_bh;
1124 di = (struct ocfs2_dinode *)di_bh->b_data;
1125
1126 down_read(&oi->ip_xattr_sem);
1127 ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer,
1128 buffer_size, &xis);
6c1e183e 1129 if (ret == -ENODATA && di->i_xattr_loc)
cf1d6c76
TY		 1130 ret = ocfs2_xattr_block_get(inode, name_index, name, buffer,
1131 buffer_size, &xbs);
1132 up_read(&oi->ip_xattr_sem);
4e3e9d02
TY		 1133
1134 return ret;
1135}
1136
1137/* ocfs2_xattr_get()
1138 *
1139 * Copy an extended attribute into the buffer provided.
1140 * Buffer is NULL to compute the size of buffer required.
1141 */
1142static int ocfs2_xattr_get(struct inode *inode,
1143 int name_index,
1144 const char *name,
1145 void *buffer,
1146 size_t buffer_size)
1147{
1148 int ret;
1149 struct buffer_head *di_bh = NULL;
1150
1151 ret = ocfs2_inode_lock(inode, &di_bh, 0);
1152 if (ret < 0) {
1153 mlog_errno(ret);
1154 return ret;
1155 }
1156 ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index,
1157 name, buffer, buffer_size);
1158
cf1d6c76
TY		 1159 ocfs2_inode_unlock(inode, 0);
1160
1161 brelse(di_bh);
1162
1163 return ret;
1164}
1165
1166static int __ocfs2_xattr_set_value_outside(struct inode *inode,
85db90e7 1167 handle_t *handle,
cf1d6c76
TY		 1168 struct ocfs2_xattr_value_root *xv,
1169 const void *value,
1170 int value_len)
1171{
71d548a6 1172 int ret = 0, i, cp_len;
cf1d6c76
TY		 1173 u16 blocksize = inode->i_sb->s_blocksize;
1174 u32 p_cluster, num_clusters;
1175 u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
1176 u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len);
1177 u64 blkno;
1178 struct buffer_head *bh = NULL;
cf1d6c76
TY		 1179
1180 BUG_ON(clusters > le32_to_cpu(xv->xr_clusters));
1181
cf1d6c76
TY		 1182 while (cpos < clusters) {
1183 ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster,
1184 &num_clusters, &xv->xr_list);
1185 if (ret) {
1186 mlog_errno(ret);
85db90e7 1187 goto out;
cf1d6c76
TY		 1188 }
1189
1190 blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster);
1191
1192 for (i = 0; i < num_clusters * bpc; i++, blkno++) {
970e4936 1193 ret = ocfs2_read_block(inode, blkno, &bh, NULL);
cf1d6c76
TY		 1194 if (ret) {
1195 mlog_errno(ret);
85db90e7 1196 goto out;
cf1d6c76
TY		 1197 }
1198
1199 ret = ocfs2_journal_access(handle,
1200 inode,
1201 bh,
1202 OCFS2_JOURNAL_ACCESS_WRITE);
1203 if (ret < 0) {
1204 mlog_errno(ret);
85db90e7 1205 goto out;
cf1d6c76
TY		 1206 }
1207
1208 cp_len = value_len > blocksize ? blocksize : value_len;
1209 memcpy(bh->b_data, value, cp_len);
1210 value_len -= cp_len;
1211 value += cp_len;
1212 if (cp_len < blocksize)
1213 memset(bh->b_data + cp_len, 0,
1214 blocksize - cp_len);
1215
1216 ret = ocfs2_journal_dirty(handle, bh);
1217 if (ret < 0) {
1218 mlog_errno(ret);
85db90e7 1219 goto out;
cf1d6c76
TY		 1220 }
1221 brelse(bh);
1222 bh = NULL;
1223
1224 /*
1225 * XXX: do we need to empty all the following
1226 * blocks in this cluster?
1227 */
1228 if (!value_len)
1229 break;
1230 }
1231 cpos += num_clusters;
1232 }
cf1d6c76
TY		 1233out:
1234 brelse(bh);
1235
1236 return ret;
1237}
1238
1239static int ocfs2_xattr_cleanup(struct inode *inode,
85db90e7 1240 handle_t *handle,
cf1d6c76
TY		 1241 struct ocfs2_xattr_info *xi,
1242 struct ocfs2_xattr_search *xs,
512620f4 1243 struct ocfs2_xattr_value_buf *vb,
cf1d6c76
TY		 1244 size_t offs)
1245{
cf1d6c76
TY		 1246 int ret = 0;
1247 size_t name_len = strlen(xi->name);
1248 void *val = xs->base + offs;
1249 size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1250
512620f4
JB		 1251 ret = vb->vb_access(handle, inode, vb->vb_bh,
1252 OCFS2_JOURNAL_ACCESS_WRITE);
cf1d6c76
TY		 1253 if (ret) {
1254 mlog_errno(ret);
85db90e7 1255 goto out;
cf1d6c76
TY		 1256 }
1257 /* Decrease xattr count */
1258 le16_add_cpu(&xs->header->xh_count, -1);
1259 /* Remove the xattr entry and tree root which has already be set*/
1260 memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry));
1261 memset(val, 0, size);
1262
512620f4 1263 ret = ocfs2_journal_dirty(handle, vb->vb_bh);
cf1d6c76
TY		 1264 if (ret < 0)
1265 mlog_errno(ret);
cf1d6c76
TY		 1266out:
1267 return ret;
1268}
1269
1270static int ocfs2_xattr_update_entry(struct inode *inode,
85db90e7 1271 handle_t *handle,
cf1d6c76
TY		 1272 struct ocfs2_xattr_info *xi,
1273 struct ocfs2_xattr_search *xs,
0c748e95 1274 struct ocfs2_xattr_value_buf *vb,
cf1d6c76
TY		 1275 size_t offs)
1276{
85db90e7 1277 int ret;
cf1d6c76 1278
0c748e95
JB		 1279 ret = vb->vb_access(handle, inode, vb->vb_bh,
1280 OCFS2_JOURNAL_ACCESS_WRITE);
cf1d6c76
TY		 1281 if (ret) {
1282 mlog_errno(ret);
85db90e7 1283 goto out;
cf1d6c76
TY		 1284 }
1285
1286 xs->here->xe_name_offset = cpu_to_le16(offs);
1287 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1288 if (xi->value_len <= OCFS2_XATTR_INLINE_SIZE)
1289 ocfs2_xattr_set_local(xs->here, 1);
1290 else
1291 ocfs2_xattr_set_local(xs->here, 0);
1292 ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
1293
0c748e95 1294 ret = ocfs2_journal_dirty(handle, vb->vb_bh);
cf1d6c76
TY		 1295 if (ret < 0)
1296 mlog_errno(ret);
cf1d6c76
TY		 1297out:
1298 return ret;
1299}
1300
1301/*
1302 * ocfs2_xattr_set_value_outside()
1303 *
1304 * Set large size value in B tree.
1305 */
1306static int ocfs2_xattr_set_value_outside(struct inode *inode,
1307 struct ocfs2_xattr_info *xi,
1308 struct ocfs2_xattr_search *xs,
78f30c31 1309 struct ocfs2_xattr_set_ctxt *ctxt,
512620f4 1310 struct ocfs2_xattr_value_buf *vb,
cf1d6c76
TY		 1311 size_t offs)
1312{
1313 size_t name_len = strlen(xi->name);
1314 void *val = xs->base + offs;
1315 struct ocfs2_xattr_value_root *xv = NULL;
1316 size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1317 int ret = 0;
1318
1319 memset(val, 0, size);
1320 memcpy(val, xi->name, name_len);
1321 xv = (struct ocfs2_xattr_value_root *)
1322 (val + OCFS2_XATTR_SIZE(name_len));
1323 xv->xr_clusters = 0;
1324 xv->xr_last_eb_blk = 0;
1325 xv->xr_list.l_tree_depth = 0;
1326 xv->xr_list.l_count = cpu_to_le16(1);
1327 xv->xr_list.l_next_free_rec = 0;
512620f4 1328 vb->vb_xv = xv;
cf1d6c76 1329
512620f4 1330 ret = ocfs2_xattr_value_truncate(inode, vb, xi->value_len, ctxt);
cf1d6c76
TY		 1331 if (ret < 0) {
1332 mlog_errno(ret);
1333 return ret;
1334 }
512620f4 1335 ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, vb, offs);
cf1d6c76
TY		 1336 if (ret < 0) {
1337 mlog_errno(ret);
1338 return ret;
1339 }
512620f4 1340 ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, vb->vb_xv,
85db90e7 1341 xi->value, xi->value_len);
cf1d6c76
TY		 1342 if (ret < 0)
1343 mlog_errno(ret);
1344
1345 return ret;
1346}
1347
1348/*
1349 * ocfs2_xattr_set_entry_local()
1350 *
1351 * Set, replace or remove extended attribute in local.
1352 */
1353static void ocfs2_xattr_set_entry_local(struct inode *inode,
1354 struct ocfs2_xattr_info *xi,
1355 struct ocfs2_xattr_search *xs,
1356 struct ocfs2_xattr_entry *last,
1357 size_t min_offs)
1358{
1359 size_t name_len = strlen(xi->name);
1360 int i;
1361
1362 if (xi->value && xs->not_found) {
1363 /* Insert the new xattr entry. */
1364 le16_add_cpu(&xs->header->xh_count, 1);
1365 ocfs2_xattr_set_type(last, xi->name_index);
1366 ocfs2_xattr_set_local(last, 1);
1367 last->xe_name_len = name_len;
1368 } else {
1369 void *first_val;
1370 void *val;
1371 size_t offs, size;
1372
1373 first_val = xs->base + min_offs;
1374 offs = le16_to_cpu(xs->here->xe_name_offset);
1375 val = xs->base + offs;
1376
1377 if (le64_to_cpu(xs->here->xe_value_size) >
1378 OCFS2_XATTR_INLINE_SIZE)
1379 size = OCFS2_XATTR_SIZE(name_len) +
1380 OCFS2_XATTR_ROOT_SIZE;
1381 else
1382 size = OCFS2_XATTR_SIZE(name_len) +
1383 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1384
1385 if (xi->value && size == OCFS2_XATTR_SIZE(name_len) +
1386 OCFS2_XATTR_SIZE(xi->value_len)) {
1387 /* The old and the new value have the
1388 same size. Just replace the value. */
1389 ocfs2_xattr_set_local(xs->here, 1);
1390 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1391 /* Clear value bytes. */
1392 memset(val + OCFS2_XATTR_SIZE(name_len),
1393 0,
1394 OCFS2_XATTR_SIZE(xi->value_len));
1395 memcpy(val + OCFS2_XATTR_SIZE(name_len),
1396 xi->value,
1397 xi->value_len);
1398 return;
1399 }
1400 /* Remove the old name+value. */
1401 memmove(first_val + size, first_val, val - first_val);
1402 memset(first_val, 0, size);
1403 xs->here->xe_name_hash = 0;
1404 xs->here->xe_name_offset = 0;
1405 ocfs2_xattr_set_local(xs->here, 1);
1406 xs->here->xe_value_size = 0;
1407
1408 min_offs += size;
1409
1410 /* Adjust all value offsets. */
1411 last = xs->header->xh_entries;
1412 for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
1413 size_t o = le16_to_cpu(last->xe_name_offset);
1414
1415 if (o < offs)
1416 last->xe_name_offset = cpu_to_le16(o + size);
1417 last += 1;
1418 }
1419
1420 if (!xi->value) {
1421 /* Remove the old entry. */
1422 last -= 1;
1423 memmove(xs->here, xs->here + 1,
1424 (void *)last - (void *)xs->here);
1425 memset(last, 0, sizeof(struct ocfs2_xattr_entry));
1426 le16_add_cpu(&xs->header->xh_count, -1);
1427 }
1428 }
1429 if (xi->value) {
1430 /* Insert the new name+value. */
1431 size_t size = OCFS2_XATTR_SIZE(name_len) +
1432 OCFS2_XATTR_SIZE(xi->value_len);
1433 void *val = xs->base + min_offs - size;
1434
1435 xs->here->xe_name_offset = cpu_to_le16(min_offs - size);
1436 memset(val, 0, size);
1437 memcpy(val, xi->name, name_len);
1438 memcpy(val + OCFS2_XATTR_SIZE(name_len),
1439 xi->value,
1440 xi->value_len);
1441 xs->here->xe_value_size = cpu_to_le64(xi->value_len);
1442 ocfs2_xattr_set_local(xs->here, 1);
1443 ocfs2_xattr_hash_entry(inode, xs->header, xs->here);
1444 }
1445
1446 return;
1447}
1448
1449/*
1450 * ocfs2_xattr_set_entry()
1451 *
1452 * Set extended attribute entry into inode or block.
1453 *
1454 * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE,
1455 * We first insert tree root(ocfs2_xattr_value_root) with set_entry_local(),
1456 * then set value in B tree with set_value_outside().
1457 */
1458static int ocfs2_xattr_set_entry(struct inode *inode,
1459 struct ocfs2_xattr_info *xi,
1460 struct ocfs2_xattr_search *xs,
78f30c31 1461 struct ocfs2_xattr_set_ctxt *ctxt,
cf1d6c76
TY		 1462 int flag)
1463{
1464 struct ocfs2_xattr_entry *last;
1465 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1466 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1467 size_t min_offs = xs->end - xs->base, name_len = strlen(xi->name);
1468 size_t size_l = 0;
85db90e7 1469 handle_t *handle = ctxt->handle;
cf1d6c76
TY		 1470 int free, i, ret;
1471 struct ocfs2_xattr_info xi_l = {
1472 .name_index = xi->name_index,
1473 .name = xi->name,
1474 .value = xi->value,
1475 .value_len = xi->value_len,
1476 };
512620f4
JB		 1477 struct ocfs2_xattr_value_buf vb = {
1478 .vb_bh = xs->xattr_bh,
1479 .vb_access = ocfs2_journal_access_di,
1480 };
1481
1482 if (!(flag & OCFS2_INLINE_XATTR_FL)) {
1483 BUG_ON(xs->xattr_bh == xs->inode_bh);
1484 vb.vb_access = ocfs2_journal_access_xb;
1485 } else
1486 BUG_ON(xs->xattr_bh != xs->inode_bh);
cf1d6c76
TY		 1487
1488 /* Compute min_offs, last and free space. */
1489 last = xs->header->xh_entries;
1490
1491 for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) {
1492 size_t offs = le16_to_cpu(last->xe_name_offset);
1493 if (offs < min_offs)
1494 min_offs = offs;
1495 last += 1;
1496 }
1497
1498 free = min_offs - ((void *)last - xs->base) - sizeof(__u32);
1499 if (free < 0)
b37c4d84 1500 return -EIO;
cf1d6c76
TY		 1501
1502 if (!xs->not_found) {
1503 size_t size = 0;
1504 if (ocfs2_xattr_is_local(xs->here))
1505 size = OCFS2_XATTR_SIZE(name_len) +
1506 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1507 else
1508 size = OCFS2_XATTR_SIZE(name_len) +
1509 OCFS2_XATTR_ROOT_SIZE;
1510 free += (size + sizeof(struct ocfs2_xattr_entry));
1511 }
1512 /* Check free space in inode or block */
1513 if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1514 if (free < sizeof(struct ocfs2_xattr_entry) +
1515 OCFS2_XATTR_SIZE(name_len) +
1516 OCFS2_XATTR_ROOT_SIZE) {
1517 ret = -ENOSPC;
1518 goto out;
1519 }
1520 size_l = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE;
1521 xi_l.value = (void *)&def_xv;
1522 xi_l.value_len = OCFS2_XATTR_ROOT_SIZE;
1523 } else if (xi->value) {
1524 if (free < sizeof(struct ocfs2_xattr_entry) +
1525 OCFS2_XATTR_SIZE(name_len) +
1526 OCFS2_XATTR_SIZE(xi->value_len)) {
1527 ret = -ENOSPC;
1528 goto out;
1529 }
1530 }
1531
1532 if (!xs->not_found) {
1533 /* For existing extended attribute */
1534 size_t size = OCFS2_XATTR_SIZE(name_len) +
1535 OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size));
1536 size_t offs = le16_to_cpu(xs->here->xe_name_offset);
1537 void *val = xs->base + offs;
1538
1539 if (ocfs2_xattr_is_local(xs->here) && size == size_l) {
1540 /* Replace existing local xattr with tree root */
1541 ret = ocfs2_xattr_set_value_outside(inode, xi, xs,
512620f4 1542 ctxt, &vb, offs);
cf1d6c76
TY		 1543 if (ret < 0)
1544 mlog_errno(ret);
1545 goto out;
1546 } else if (!ocfs2_xattr_is_local(xs->here)) {
1547 /* For existing xattr which has value outside */
512620f4
JB		 1548 vb.vb_xv = (struct ocfs2_xattr_value_root *)
1549 (val + OCFS2_XATTR_SIZE(name_len));
cf1d6c76
TY		 1550
1551 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1552 /*
1553 * If new value need set outside also,
1554 * first truncate old value to new value,
1555 * then set new value with set_value_outside().
1556 */
1557 ret = ocfs2_xattr_value_truncate(inode,
b3e5d379 1558 &vb,
78f30c31
TM		 1559 xi->value_len,
1560 ctxt);
cf1d6c76
TY		 1561 if (ret < 0) {
1562 mlog_errno(ret);
1563 goto out;
1564 }
1565
85db90e7
TM		 1566 ret = ocfs2_xattr_update_entry(inode,
1567 handle,
1568 xi,
1569 xs,
0c748e95 1570 &vb,
85db90e7 1571 offs);
cf1d6c76
TY		 1572 if (ret < 0) {
1573 mlog_errno(ret);
1574 goto out;
1575 }
1576
85db90e7
TM		 1577 ret = __ocfs2_xattr_set_value_outside(inode,
1578 handle,
b3e5d379 1579 vb.vb_xv,
85db90e7
TM		 1580 xi->value,
1581 xi->value_len);
cf1d6c76
TY		 1582 if (ret < 0)
1583 mlog_errno(ret);
1584 goto out;
1585 } else {
1586 /*
1587 * If new value need set in local,
1588 * just trucate old value to zero.
1589 */
1590 ret = ocfs2_xattr_value_truncate(inode,
b3e5d379 1591 &vb,
85db90e7
TM		 1592 0,
1593 ctxt);
cf1d6c76
TY		 1594 if (ret < 0)
1595 mlog_errno(ret);
1596 }
1597 }
1598 }
1599
512620f4
JB		 1600 ret = ocfs2_journal_access_di(handle, inode, xs->inode_bh,
1601 OCFS2_JOURNAL_ACCESS_WRITE);
cf1d6c76
TY		 1602 if (ret) {
1603 mlog_errno(ret);
85db90e7 1604 goto out;
cf1d6c76
TY		 1605 }
1606
1607 if (!(flag & OCFS2_INLINE_XATTR_FL)) {
512620f4
JB		 1608 ret = vb.vb_access(handle, inode, vb.vb_bh,
1609 OCFS2_JOURNAL_ACCESS_WRITE);
cf1d6c76
TY		 1610 if (ret) {
1611 mlog_errno(ret);
85db90e7 1612 goto out;
cf1d6c76
TY		 1613 }
1614 }
1615
1616 /*
1617 * Set value in local, include set tree root in local.
1618 * This is the first step for value size >INLINE_SIZE.
1619 */
1620 ocfs2_xattr_set_entry_local(inode, &xi_l, xs, last, min_offs);
1621
1622 if (!(flag & OCFS2_INLINE_XATTR_FL)) {
1623 ret = ocfs2_journal_dirty(handle, xs->xattr_bh);
1624 if (ret < 0) {
1625 mlog_errno(ret);
85db90e7 1626 goto out;
cf1d6c76
TY		 1627 }
1628 }
1629
1630 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) &&
1631 (flag & OCFS2_INLINE_XATTR_FL)) {
1632 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1633 unsigned int xattrsize = osb->s_xattr_inline_size;
1634
1635 /*
1636 * Adjust extent record count or inline data size
1637 * to reserve space for extended attribute.
1638 */
1639 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1640 struct ocfs2_inline_data *idata = &di->id2.i_data;
1641 le16_add_cpu(&idata->id_count, -xattrsize);
1642 } else if (!(ocfs2_inode_is_fast_symlink(inode))) {
1643 struct ocfs2_extent_list *el = &di->id2.i_list;
1644 le16_add_cpu(&el->l_count, -(xattrsize /
1645 sizeof(struct ocfs2_extent_rec)));
1646 }
1647 di->i_xattr_inline_size = cpu_to_le16(xattrsize);
1648 }
1649 /* Update xattr flag */
1650 spin_lock(&oi->ip_lock);
1651 oi->ip_dyn_features |= flag;
1652 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
1653 spin_unlock(&oi->ip_lock);
cf1d6c76
TY		 1654
1655 ret = ocfs2_journal_dirty(handle, xs->inode_bh);
1656 if (ret < 0)
1657 mlog_errno(ret);
1658
cf1d6c76
TY		 1659 if (!ret && xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
1660 /*
1661 * Set value outside in B tree.
1662 * This is the second step for value size > INLINE_SIZE.
1663 */
1664 size_t offs = le16_to_cpu(xs->here->xe_name_offset);
512620f4
JB		 1665 ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt,
1666 &vb, offs);
cf1d6c76
TY		 1667 if (ret < 0) {
1668 int ret2;
1669
1670 mlog_errno(ret);
1671 /*
1672 * If set value outside failed, we have to clean
1673 * the junk tree root we have already set in local.
1674 */
85db90e7 1675 ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle,
512620f4 1676 xi, xs, &vb, offs);
cf1d6c76
TY		 1677 if (ret2 < 0)
1678 mlog_errno(ret2);
1679 }
1680 }
1681out:
1682 return ret;
cf1d6c76
TY		 1683}
1684
cf1d6c76 1685static int ocfs2_remove_value_outside(struct inode*inode,
4311901d 1686 struct ocfs2_xattr_value_buf *vb,
cf1d6c76
TY		 1687 struct ocfs2_xattr_header *header)
1688{
1689 int ret = 0, i;
78f30c31
TM		 1690 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1691 struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
1692
1693 ocfs2_init_dealloc_ctxt(&ctxt.dealloc);
cf1d6c76 1694
a90714c1
JK		 1695 ctxt.handle = ocfs2_start_trans(osb,
1696 ocfs2_remove_extent_credits(osb->sb));
85db90e7
TM		 1697 if (IS_ERR(ctxt.handle)) {
1698 ret = PTR_ERR(ctxt.handle);
1699 mlog_errno(ret);
1700 goto out;
1701 }
1702
cf1d6c76
TY		 1703 for (i = 0; i < le16_to_cpu(header->xh_count); i++) {
1704 struct ocfs2_xattr_entry *entry = &header->xh_entries[i];
1705
1706 if (!ocfs2_xattr_is_local(entry)) {
cf1d6c76
TY		 1707 void *val;
1708
1709 val = (void *)header +
1710 le16_to_cpu(entry->xe_name_offset);
4311901d 1711 vb->vb_xv = (struct ocfs2_xattr_value_root *)
cf1d6c76 1712 (val + OCFS2_XATTR_SIZE(entry->xe_name_len));
4311901d 1713 ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt);
cf1d6c76
TY		 1714 if (ret < 0) {
1715 mlog_errno(ret);
78f30c31 1716 break;
cf1d6c76
TY		 1717 }
1718 }
1719 }
1720
85db90e7 1721 ocfs2_commit_trans(osb, ctxt.handle);
78f30c31
TM		 1722 ocfs2_schedule_truncate_log_flush(osb, 1);
1723 ocfs2_run_deallocs(osb, &ctxt.dealloc);
85db90e7 1724out:
cf1d6c76
TY		 1725 return ret;
1726}
1727
1728static int ocfs2_xattr_ibody_remove(struct inode *inode,
1729 struct buffer_head *di_bh)
1730{
1731
1732 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1733 struct ocfs2_xattr_header *header;
1734 int ret;
4311901d
JB		 1735 struct ocfs2_xattr_value_buf vb = {
1736 .vb_bh = di_bh,
1737 .vb_access = ocfs2_journal_access_di,
1738 };
cf1d6c76
TY		 1739
1740 header = (struct ocfs2_xattr_header *)
1741 ((void *)di + inode->i_sb->s_blocksize -
1742 le16_to_cpu(di->i_xattr_inline_size));
1743
4311901d 1744 ret = ocfs2_remove_value_outside(inode, &vb, header);
cf1d6c76
TY		 1745
1746 return ret;
1747}
1748
1749static int ocfs2_xattr_block_remove(struct inode *inode,
1750 struct buffer_head *blk_bh)
1751{
1752 struct ocfs2_xattr_block *xb;
cf1d6c76 1753 int ret = 0;
4311901d
JB		 1754 struct ocfs2_xattr_value_buf vb = {
1755 .vb_bh = blk_bh,
1756 .vb_access = ocfs2_journal_access_xb,
1757 };
cf1d6c76
TY		 1758
1759 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
a3944256
TM		 1760 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
1761 struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header);
4311901d 1762 ret = ocfs2_remove_value_outside(inode, &vb, header);
a3944256
TM		 1763 } else
1764 ret = ocfs2_delete_xattr_index_block(inode, blk_bh);
cf1d6c76
TY		 1765
1766 return ret;
1767}
1768
08413899
TM		 1769static int ocfs2_xattr_free_block(struct inode *inode,
1770 u64 block)
1771{
1772 struct inode *xb_alloc_inode;
1773 struct buffer_head *xb_alloc_bh = NULL;
1774 struct buffer_head *blk_bh = NULL;
1775 struct ocfs2_xattr_block *xb;
1776 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
1777 handle_t *handle;
1778 int ret = 0;
1779 u64 blk, bg_blkno;
1780 u16 bit;
1781
4ae1d69b 1782 ret = ocfs2_read_xattr_block(inode, block, &blk_bh);
08413899
TM		 1783 if (ret < 0) {
1784 mlog_errno(ret);
1785 goto out;
1786 }
1787
08413899
TM		 1788 ret = ocfs2_xattr_block_remove(inode, blk_bh);
1789 if (ret < 0) {
1790 mlog_errno(ret);
1791 goto out;
1792 }
1793
4ae1d69b 1794 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
08413899
TM		 1795 blk = le64_to_cpu(xb->xb_blkno);
1796 bit = le16_to_cpu(xb->xb_suballoc_bit);
1797 bg_blkno = ocfs2_which_suballoc_group(blk, bit);
1798
1799 xb_alloc_inode = ocfs2_get_system_file_inode(osb,
1800 EXTENT_ALLOC_SYSTEM_INODE,
1801 le16_to_cpu(xb->xb_suballoc_slot));
1802 if (!xb_alloc_inode) {
1803 ret = -ENOMEM;
1804 mlog_errno(ret);
1805 goto out;
1806 }
1807 mutex_lock(&xb_alloc_inode->i_mutex);
1808
1809 ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1);
1810 if (ret < 0) {
1811 mlog_errno(ret);
1812 goto out_mutex;
1813 }
1814
1815 handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE);
1816 if (IS_ERR(handle)) {
1817 ret = PTR_ERR(handle);
1818 mlog_errno(ret);
1819 goto out_unlock;
1820 }
1821
1822 ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh,
1823 bit, bg_blkno, 1);
1824 if (ret < 0)
1825 mlog_errno(ret);
1826
1827 ocfs2_commit_trans(osb, handle);
1828out_unlock:
1829 ocfs2_inode_unlock(xb_alloc_inode, 1);
1830 brelse(xb_alloc_bh);
1831out_mutex:
1832 mutex_unlock(&xb_alloc_inode->i_mutex);
1833 iput(xb_alloc_inode);
1834out:
1835 brelse(blk_bh);
1836 return ret;
1837}
1838
cf1d6c76
TY		 1839/*
1840 * ocfs2_xattr_remove()
1841 *
1842 * Free extended attribute resources associated with this inode.
1843 */
1844int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh)
1845{
cf1d6c76
TY		 1846 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1847 struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data;
1848 handle_t *handle;
1849 int ret;
1850
8154da3d
TY		 1851 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
1852 return 0;
1853
cf1d6c76
TY		 1854 if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL))
1855 return 0;
1856
1857 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
1858 ret = ocfs2_xattr_ibody_remove(inode, di_bh);
1859 if (ret < 0) {
1860 mlog_errno(ret);
1861 goto out;
1862 }
1863 }
cf1d6c76 1864
08413899
TM		 1865 if (di->i_xattr_loc) {
1866 ret = ocfs2_xattr_free_block(inode,
1867 le64_to_cpu(di->i_xattr_loc));
cf1d6c76
TY		 1868 if (ret < 0) {
1869 mlog_errno(ret);
1870 goto out;
1871 }
1872 }
1873
1874 handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)),
1875 OCFS2_INODE_UPDATE_CREDITS);
1876 if (IS_ERR(handle)) {
1877 ret = PTR_ERR(handle);
1878 mlog_errno(ret);
1879 goto out;
1880 }
84008972
JB		 1881 ret = ocfs2_journal_access_di(handle, inode, di_bh,
1882 OCFS2_JOURNAL_ACCESS_WRITE);
cf1d6c76
TY		 1883 if (ret) {
1884 mlog_errno(ret);
1885 goto out_commit;
1886 }
1887
08413899 1888 di->i_xattr_loc = 0;
cf1d6c76
TY		 1889
1890 spin_lock(&oi->ip_lock);
1891 oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL);
1892 di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features);
1893 spin_unlock(&oi->ip_lock);
1894
1895 ret = ocfs2_journal_dirty(handle, di_bh);
1896 if (ret < 0)
1897 mlog_errno(ret);
1898out_commit:
1899 ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle);
1900out:
cf1d6c76
TY		 1901 return ret;
1902}
1903
1904static int ocfs2_xattr_has_space_inline(struct inode *inode,
1905 struct ocfs2_dinode *di)
1906{
1907 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1908 unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size;
1909 int free;
1910
1911 if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE)
1912 return 0;
1913
1914 if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) {
1915 struct ocfs2_inline_data *idata = &di->id2.i_data;
1916 free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size);
1917 } else if (ocfs2_inode_is_fast_symlink(inode)) {
1918 free = ocfs2_fast_symlink_chars(inode->i_sb) -
1919 le64_to_cpu(di->i_size);
1920 } else {
1921 struct ocfs2_extent_list *el = &di->id2.i_list;
1922 free = (le16_to_cpu(el->l_count) -
1923 le16_to_cpu(el->l_next_free_rec)) *
1924 sizeof(struct ocfs2_extent_rec);
1925 }
1926 if (free >= xattrsize)
1927 return 1;
1928
1929 return 0;
1930}
1931
1932/*
1933 * ocfs2_xattr_ibody_find()
1934 *
1935 * Find extended attribute in inode block and
1936 * fill search info into struct ocfs2_xattr_search.
1937 */
1938static int ocfs2_xattr_ibody_find(struct inode *inode,
1939 int name_index,
1940 const char *name,
1941 struct ocfs2_xattr_search *xs)
1942{
1943 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1944 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1945 int ret;
1946 int has_space = 0;
1947
1948 if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE)
1949 return 0;
1950
1951 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) {
1952 down_read(&oi->ip_alloc_sem);
1953 has_space = ocfs2_xattr_has_space_inline(inode, di);
1954 up_read(&oi->ip_alloc_sem);
1955 if (!has_space)
1956 return 0;
1957 }
1958
1959 xs->xattr_bh = xs->inode_bh;
1960 xs->end = (void *)di + inode->i_sb->s_blocksize;
1961 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)
1962 xs->header = (struct ocfs2_xattr_header *)
1963 (xs->end - le16_to_cpu(di->i_xattr_inline_size));
1964 else
1965 xs->header = (struct ocfs2_xattr_header *)
1966 (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size);
1967 xs->base = (void *)xs->header;
1968 xs->here = xs->header->xh_entries;
1969
1970 /* Find the named attribute. */
1971 if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) {
1972 ret = ocfs2_xattr_find_entry(name_index, name, xs);
1973 if (ret && ret != -ENODATA)
1974 return ret;
1975 xs->not_found = ret;
1976 }
1977
1978 return 0;
1979}
1980
1981/*
1982 * ocfs2_xattr_ibody_set()
1983 *
1984 * Set, replace or remove an extended attribute into inode block.
1985 *
1986 */
1987static int ocfs2_xattr_ibody_set(struct inode *inode,
1988 struct ocfs2_xattr_info *xi,
78f30c31
TM		 1989 struct ocfs2_xattr_search *xs,
1990 struct ocfs2_xattr_set_ctxt *ctxt)
cf1d6c76
TY		 1991{
1992 struct ocfs2_inode_info *oi = OCFS2_I(inode);
1993 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
1994 int ret;
1995
1996 if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE)
1997 return -ENOSPC;
1998
1999 down_write(&oi->ip_alloc_sem);
2000 if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) {
2001 if (!ocfs2_xattr_has_space_inline(inode, di)) {
2002 ret = -ENOSPC;
2003 goto out;
2004 }
2005 }
2006
78f30c31 2007 ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
cf1d6c76
TY		 2008 (OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL));
2009out:
2010 up_write(&oi->ip_alloc_sem);
2011
2012 return ret;
2013}
2014
2015/*
2016 * ocfs2_xattr_block_find()
2017 *
2018 * Find extended attribute in external block and
2019 * fill search info into struct ocfs2_xattr_search.
2020 */
2021static int ocfs2_xattr_block_find(struct inode *inode,
2022 int name_index,
2023 const char *name,
2024 struct ocfs2_xattr_search *xs)
2025{
2026 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
2027 struct buffer_head *blk_bh = NULL;
589dc260 2028 struct ocfs2_xattr_block *xb;
cf1d6c76
TY		 2029 int ret = 0;
2030
2031 if (!di->i_xattr_loc)
2032 return ret;
2033
4ae1d69b
JB		 2034 ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc),
2035 &blk_bh);
cf1d6c76
TY		 2036 if (ret < 0) {
2037 mlog_errno(ret);
2038 return ret;
2039 }
f6087fb7 2040
cf1d6c76 2041 xs->xattr_bh = blk_bh;
4ae1d69b 2042 xb = (struct ocfs2_xattr_block *)blk_bh->b_data;
589dc260
TM		 2043
2044 if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) {
2045 xs->header = &xb->xb_attrs.xb_header;
2046 xs->base = (void *)xs->header;
2047 xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size;
2048 xs->here = xs->header->xh_entries;
2049
2050 ret = ocfs2_xattr_find_entry(name_index, name, xs);
2051 } else
2052 ret = ocfs2_xattr_index_block_find(inode, blk_bh,
2053 name_index,
2054 name, xs);
cf1d6c76 2055
cf1d6c76
TY		 2056 if (ret && ret != -ENODATA) {
2057 xs->xattr_bh = NULL;
2058 goto cleanup;
2059 }
2060 xs->not_found = ret;
2061 return 0;
cf1d6c76
TY		 2062cleanup:
2063 brelse(blk_bh);
2064
2065 return ret;
2066}
2067
2068/*
2069 * ocfs2_xattr_block_set()
2070 *
2071 * Set, replace or remove an extended attribute into external block.
2072 *
2073 */
2074static int ocfs2_xattr_block_set(struct inode *inode,
2075 struct ocfs2_xattr_info *xi,
78f30c31
TM		 2076 struct ocfs2_xattr_search *xs,
2077 struct ocfs2_xattr_set_ctxt *ctxt)
cf1d6c76
TY		 2078{
2079 struct buffer_head *new_bh = NULL;
2080 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2081 struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data;
85db90e7 2082 handle_t *handle = ctxt->handle;
cf1d6c76
TY		 2083 struct ocfs2_xattr_block *xblk = NULL;
2084 u16 suballoc_bit_start;
2085 u32 num_got;
2086 u64 first_blkno;
2087 int ret;
2088
2089 if (!xs->xattr_bh) {
84008972
JB		 2090 ret = ocfs2_journal_access_di(handle, inode, xs->inode_bh,
2091 OCFS2_JOURNAL_ACCESS_CREATE);
cf1d6c76
TY		 2092 if (ret < 0) {
2093 mlog_errno(ret);
85db90e7 2094 goto end;
cf1d6c76
TY		 2095 }
2096
78f30c31 2097 ret = ocfs2_claim_metadata(osb, handle, ctxt->meta_ac, 1,
cf1d6c76
TY		 2098 &suballoc_bit_start, &num_got,
2099 &first_blkno);
2100 if (ret < 0) {
2101 mlog_errno(ret);
85db90e7 2102 goto end;
cf1d6c76
TY		 2103 }
2104
2105 new_bh = sb_getblk(inode->i_sb, first_blkno);
2106 ocfs2_set_new_buffer_uptodate(inode, new_bh);
2107
84008972
JB		 2108 ret = ocfs2_journal_access_xb(handle, inode, new_bh,
2109 OCFS2_JOURNAL_ACCESS_CREATE);
cf1d6c76
TY		 2110 if (ret < 0) {
2111 mlog_errno(ret);
85db90e7 2112 goto end;
cf1d6c76
TY		 2113 }
2114
2115 /* Initialize ocfs2_xattr_block */
2116 xs->xattr_bh = new_bh;
2117 xblk = (struct ocfs2_xattr_block *)new_bh->b_data;
2118 memset(xblk, 0, inode->i_sb->s_blocksize);
2119 strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE);
2120 xblk->xb_suballoc_slot = cpu_to_le16(osb->slot_num);
2121 xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start);
2122 xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation);
2123 xblk->xb_blkno = cpu_to_le64(first_blkno);
2124
2125 xs->header = &xblk->xb_attrs.xb_header;
2126 xs->base = (void *)xs->header;
2127 xs->end = (void *)xblk + inode->i_sb->s_blocksize;
2128 xs->here = xs->header->xh_entries;
2129
cf1d6c76
TY		 2130 ret = ocfs2_journal_dirty(handle, new_bh);
2131 if (ret < 0) {
2132 mlog_errno(ret);
85db90e7 2133 goto end;
cf1d6c76
TY		 2134 }
2135 di->i_xattr_loc = cpu_to_le64(first_blkno);
85db90e7 2136 ocfs2_journal_dirty(handle, xs->inode_bh);
01225596
TM		 2137 } else
2138 xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data;
2139
2140 if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) {
2141 /* Set extended attribute into external block */
78f30c31
TM		 2142 ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt,
2143 OCFS2_HAS_XATTR_FL);
01225596
TM		 2144 if (!ret || ret != -ENOSPC)
2145 goto end;
2146
78f30c31 2147 ret = ocfs2_xattr_create_index_block(inode, xs, ctxt);
01225596
TM		 2148 if (ret)
2149 goto end;
cf1d6c76
TY		 2150 }
2151
78f30c31 2152 ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt);
01225596
TM		 2153
2154end:
cf1d6c76
TY		 2155
2156 return ret;
2157}
2158
78f30c31
TM		 2159/* Check whether the new xattr can be inserted into the inode. */
2160static int ocfs2_xattr_can_be_in_inode(struct inode *inode,
2161 struct ocfs2_xattr_info *xi,
2162 struct ocfs2_xattr_search *xs)
2163{
2164 u64 value_size;
2165 struct ocfs2_xattr_entry *last;
2166 int free, i;
2167 size_t min_offs = xs->end - xs->base;
2168
2169 if (!xs->header)
2170 return 0;
2171
2172 last = xs->header->xh_entries;
2173
2174 for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) {
2175 size_t offs = le16_to_cpu(last->xe_name_offset);
2176 if (offs < min_offs)
2177 min_offs = offs;
2178 last += 1;
2179 }
2180
2181 free = min_offs - ((void *)last - xs->base) - sizeof(__u32);
2182 if (free < 0)
2183 return 0;
2184
2185 BUG_ON(!xs->not_found);
2186
2187 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
2188 value_size = OCFS2_XATTR_ROOT_SIZE;
2189 else
2190 value_size = OCFS2_XATTR_SIZE(xi->value_len);
2191
2192 if (free >= sizeof(struct ocfs2_xattr_entry) +
2193 OCFS2_XATTR_SIZE(strlen(xi->name)) + value_size)
2194 return 1;
2195
2196 return 0;
2197}
2198
2199static int ocfs2_calc_xattr_set_need(struct inode *inode,
2200 struct ocfs2_dinode *di,
2201 struct ocfs2_xattr_info *xi,
2202 struct ocfs2_xattr_search *xis,
2203 struct ocfs2_xattr_search *xbs,
2204 int *clusters_need,
85db90e7
TM		 2205 int *meta_need,
2206 int *credits_need)
78f30c31
TM		 2207{
2208 int ret = 0, old_in_xb = 0;
85db90e7 2209 int clusters_add = 0, meta_add = 0, credits = 0;
78f30c31
TM		 2210 struct buffer_head *bh = NULL;
2211 struct ocfs2_xattr_block *xb = NULL;
2212 struct ocfs2_xattr_entry *xe = NULL;
2213 struct ocfs2_xattr_value_root *xv = NULL;
2214 char *base = NULL;
2215 int name_offset, name_len = 0;
2216 u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
2217 xi->value_len);
2218 u64 value_size;
2219
71d548a6
TM		 2220 /*
2221 * Calculate the clusters we need to write.
2222 * No matter whether we replace an old one or add a new one,
2223 * we need this for writing.
2224 */
2225 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
2226 credits += new_clusters *
2227 ocfs2_clusters_to_blocks(inode->i_sb, 1);
2228
78f30c31 2229 if (xis->not_found && xbs->not_found) {
85db90e7
TM		 2230 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2231
2232 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
78f30c31 2233 clusters_add += new_clusters;
85db90e7
TM		 2234 credits += ocfs2_calc_extend_credits(inode->i_sb,
2235 &def_xv.xv.xr_list,
2236 new_clusters);
2237 }
78f30c31
TM		 2238
2239 goto meta_guess;
2240 }
2241
2242 if (!xis->not_found) {
2243 xe = xis->here;
2244 name_offset = le16_to_cpu(xe->xe_name_offset);
2245 name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
2246 base = xis->base;
85db90e7 2247 credits += OCFS2_INODE_UPDATE_CREDITS;
78f30c31 2248 } else {
970e4936 2249 int i, block_off = 0;
78f30c31
TM		 2250 xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
2251 xe = xbs->here;
2252 name_offset = le16_to_cpu(xe->xe_name_offset);
2253 name_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
2254 i = xbs->here - xbs->header->xh_entries;
2255 old_in_xb = 1;
2256
2257 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
2258 ret = ocfs2_xattr_bucket_get_name_value(inode,
2259 bucket_xh(xbs->bucket),
2260 i, &block_off,
2261 &name_offset);
2262 base = bucket_block(xbs->bucket, block_off);
85db90e7
TM		 2263 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2264 } else {
78f30c31 2265 base = xbs->base;
85db90e7
TM		 2266 credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS;
2267 }
2268 }
2269
2270 /*
2271 * delete a xattr doesn't need metadata and cluster allocation.
2272 * so just calculate the credits and return.
2273 *
2274 * The credits for removing the value tree will be extended
2275 * by ocfs2_remove_extent itself.
2276 */
2277 if (!xi->value) {
2278 if (!ocfs2_xattr_is_local(xe))
a90714c1 2279 credits += ocfs2_remove_extent_credits(inode->i_sb);
85db90e7
TM		 2280
2281 goto out;
78f30c31
TM		 2282 }
2283
2284 /* do cluster allocation guess first. */
2285 value_size = le64_to_cpu(xe->xe_value_size);
2286
2287 if (old_in_xb) {
2288 /*
2289 * In xattr set, we always try to set the xe in inode first,
2290 * so if it can be inserted into inode successfully, the old
2291 * one will be removed from the xattr block, and this xattr
2292 * will be inserted into inode as a new xattr in inode.
2293 */
2294 if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) {
2295 clusters_add += new_clusters;
a90714c1 2296 credits += ocfs2_remove_extent_credits(inode->i_sb) +
85db90e7
TM		 2297 OCFS2_INODE_UPDATE_CREDITS;
2298 if (!ocfs2_xattr_is_local(xe))
2299 credits += ocfs2_calc_extend_credits(
2300 inode->i_sb,
2301 &def_xv.xv.xr_list,
2302 new_clusters);
78f30c31
TM		 2303 goto out;
2304 }
2305 }
2306
2307 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) {
2308 /* the new values will be stored outside. */
2309 u32 old_clusters = 0;
2310
2311 if (!ocfs2_xattr_is_local(xe)) {
2312 old_clusters = ocfs2_clusters_for_bytes(inode->i_sb,
2313 value_size);
2314 xv = (struct ocfs2_xattr_value_root *)
2315 (base + name_offset + name_len);
97aff52a 2316 value_size = OCFS2_XATTR_ROOT_SIZE;
78f30c31
TM		 2317 } else
2318 xv = &def_xv.xv;
2319
85db90e7 2320 if (old_clusters >= new_clusters) {
a90714c1 2321 credits += ocfs2_remove_extent_credits(inode->i_sb);
78f30c31 2322 goto out;
85db90e7 2323 } else {
78f30c31
TM		 2324 meta_add += ocfs2_extend_meta_needed(&xv->xr_list);
2325 clusters_add += new_clusters - old_clusters;
85db90e7
TM		 2326 credits += ocfs2_calc_extend_credits(inode->i_sb,
2327 &xv->xr_list,
2328 new_clusters -
2329 old_clusters);
97aff52a
TM		 2330 if (value_size >= OCFS2_XATTR_ROOT_SIZE)
2331 goto out;
78f30c31
TM		 2332 }
2333 } else {
2334 /*
2335 * Now the new value will be stored inside. So if the new
2336 * value is smaller than the size of value root or the old
2337 * value, we don't need any allocation, otherwise we have
2338 * to guess metadata allocation.
2339 */
2340 if ((ocfs2_xattr_is_local(xe) && value_size >= xi->value_len) ||
2341 (!ocfs2_xattr_is_local(xe) &&
2342 OCFS2_XATTR_ROOT_SIZE >= xi->value_len))
2343 goto out;
2344 }
2345
2346meta_guess:
2347 /* calculate metadata allocation. */
2348 if (di->i_xattr_loc) {
2349 if (!xbs->xattr_bh) {
4ae1d69b
JB		 2350 ret = ocfs2_read_xattr_block(inode,
2351 le64_to_cpu(di->i_xattr_loc),
2352 &bh);
78f30c31
TM		 2353 if (ret) {
2354 mlog_errno(ret);
2355 goto out;
2356 }
2357
2358 xb = (struct ocfs2_xattr_block *)bh->b_data;
2359 } else
2360 xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data;
2361
90cb546c
TM		 2362 /*
2363 * If there is already an xattr tree, good, we can calculate
2364 * like other b-trees. Otherwise we may have the chance of
2365 * create a tree, the credit calculation is borrowed from
2366 * ocfs2_calc_extend_credits with root_el = NULL. And the
2367 * new tree will be cluster based, so no meta is needed.
2368 */
78f30c31
TM		 2369 if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) {
2370 struct ocfs2_extent_list *el =
2371 &xb->xb_attrs.xb_root.xt_list;
2372 meta_add += ocfs2_extend_meta_needed(el);
85db90e7
TM		 2373 credits += ocfs2_calc_extend_credits(inode->i_sb,
2374 el, 1);
90cb546c
TM		 2375 } else
2376 credits += OCFS2_SUBALLOC_ALLOC + 1;
78f30c31
TM		 2377
2378 /*
2379 * This cluster will be used either for new bucket or for
2380 * new xattr block.
2381 * If the cluster size is the same as the bucket size, one
2382 * more is needed since we may need to extend the bucket
2383 * also.
2384 */
2385 clusters_add += 1;
85db90e7 2386 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
78f30c31 2387 if (OCFS2_XATTR_BUCKET_SIZE ==
85db90e7
TM		 2388 OCFS2_SB(inode->i_sb)->s_clustersize) {
2389 credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb);
78f30c31 2390 clusters_add += 1;
85db90e7
TM		 2391 }
2392 } else {
78f30c31 2393 meta_add += 1;
85db90e7
TM		 2394 credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS;
2395 }
78f30c31
TM		 2396out:
2397 if (clusters_need)
2398 *clusters_need = clusters_add;
2399 if (meta_need)
2400 *meta_need = meta_add;
85db90e7
TM		 2401 if (credits_need)
2402 *credits_need = credits;
78f30c31
TM		 2403 brelse(bh);
2404 return ret;
2405}
2406
2407static int ocfs2_init_xattr_set_ctxt(struct inode *inode,
2408 struct ocfs2_dinode *di,
2409 struct ocfs2_xattr_info *xi,
2410 struct ocfs2_xattr_search *xis,
2411 struct ocfs2_xattr_search *xbs,
85db90e7
TM		 2412 struct ocfs2_xattr_set_ctxt *ctxt,
2413 int *credits)
78f30c31
TM		 2414{
2415 int clusters_add, meta_add, ret;
2416 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
2417
2418 memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt));
2419
2420 ocfs2_init_dealloc_ctxt(&ctxt->dealloc);
2421
2422 ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs,
85db90e7 2423 &clusters_add, &meta_add, credits);
78f30c31
TM		 2424 if (ret) {
2425 mlog_errno(ret);
2426 return ret;
2427 }
2428
85db90e7
TM		 2429 mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, "
2430 "credits = %d\n", xi->name, meta_add, clusters_add, *credits);
78f30c31
TM		 2431
2432 if (meta_add) {
2433 ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add,
2434 &ctxt->meta_ac);
2435 if (ret) {
2436 mlog_errno(ret);
2437 goto out;
2438 }
2439 }
2440
2441 if (clusters_add) {
2442 ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac);
2443 if (ret)
2444 mlog_errno(ret);
2445 }
2446out:
2447 if (ret) {
2448 if (ctxt->meta_ac) {
2449 ocfs2_free_alloc_context(ctxt->meta_ac);
2450 ctxt->meta_ac = NULL;
2451 }
2452
2453 /*
2454 * We cannot have an error and a non null ctxt->data_ac.
2455 */
2456 }
2457
2458 return ret;
2459}
2460
85db90e7
TM		 2461static int __ocfs2_xattr_set_handle(struct inode *inode,
2462 struct ocfs2_dinode *di,
2463 struct ocfs2_xattr_info *xi,
2464 struct ocfs2_xattr_search *xis,
2465 struct ocfs2_xattr_search *xbs,
2466 struct ocfs2_xattr_set_ctxt *ctxt)
2467{
9f868f16 2468 int ret = 0, credits, old_found;
85db90e7
TM		 2469
2470 if (!xi->value) {
2471 /* Remove existing extended attribute */
2472 if (!xis->not_found)
2473 ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt);
2474 else if (!xbs->not_found)
2475 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2476 } else {
2477 /* We always try to set extended attribute into inode first*/
2478 ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt);
2479 if (!ret && !xbs->not_found) {
2480 /*
2481 * If succeed and that extended attribute existing in
2482 * external block, then we will remove it.
2483 */
2484 xi->value = NULL;
2485 xi->value_len = 0;
2486
9f868f16 2487 old_found = xis->not_found;
85db90e7
TM		 2488 xis->not_found = -ENODATA;
2489 ret = ocfs2_calc_xattr_set_need(inode,
2490 di,
2491 xi,
2492 xis,
2493 xbs,
2494 NULL,
2495 NULL,
2496 &credits);
9f868f16 2497 xis->not_found = old_found;
85db90e7
TM		 2498 if (ret) {
2499 mlog_errno(ret);
2500 goto out;
2501 }
2502
2503 ret = ocfs2_extend_trans(ctxt->handle, credits +
2504 ctxt->handle->h_buffer_credits);
2505 if (ret) {
2506 mlog_errno(ret);
2507 goto out;
2508 }
2509 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2510 } else if (ret == -ENOSPC) {
2511 if (di->i_xattr_loc && !xbs->xattr_bh) {
2512 ret = ocfs2_xattr_block_find(inode,
2513 xi->name_index,
2514 xi->name, xbs);
2515 if (ret)
2516 goto out;
2517
9f868f16 2518 old_found = xis->not_found;
85db90e7
TM		 2519 xis->not_found = -ENODATA;
2520 ret = ocfs2_calc_xattr_set_need(inode,
2521 di,
2522 xi,
2523 xis,
2524 xbs,
2525 NULL,
2526 NULL,
2527 &credits);
9f868f16 2528 xis->not_found = old_found;
85db90e7
TM		 2529 if (ret) {
2530 mlog_errno(ret);
2531 goto out;
2532 }
2533
2534 ret = ocfs2_extend_trans(ctxt->handle, credits +
2535 ctxt->handle->h_buffer_credits);
2536 if (ret) {
2537 mlog_errno(ret);
2538 goto out;
2539 }
2540 }
2541 /*
2542 * If no space in inode, we will set extended attribute
2543 * into external block.
2544 */
2545 ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt);
2546 if (ret)
2547 goto out;
2548 if (!xis->not_found) {
2549 /*
2550 * If succeed and that extended attribute
2551 * existing in inode, we will remove it.
2552 */
2553 xi->value = NULL;
2554 xi->value_len = 0;
2555 xbs->not_found = -ENODATA;
2556 ret = ocfs2_calc_xattr_set_need(inode,
2557 di,
2558 xi,
2559 xis,
2560 xbs,
2561 NULL,
2562 NULL,
2563 &credits);
2564 if (ret) {
2565 mlog_errno(ret);
2566 goto out;
2567 }
2568
2569 ret = ocfs2_extend_trans(ctxt->handle, credits +
2570 ctxt->handle->h_buffer_credits);
2571 if (ret) {
2572 mlog_errno(ret);
2573 goto out;
2574 }
2575 ret = ocfs2_xattr_ibody_set(inode, xi,
2576 xis, ctxt);
2577 }
2578 }
2579 }
2580
4b3f6209
TM		 2581 if (!ret) {
2582 /* Update inode ctime. */
2583 ret = ocfs2_journal_access(ctxt->handle, inode, xis->inode_bh,
2584 OCFS2_JOURNAL_ACCESS_WRITE);
2585 if (ret) {
2586 mlog_errno(ret);
2587 goto out;
2588 }
2589
2590 inode->i_ctime = CURRENT_TIME;
2591 di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec);
2592 di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec);
2593 ocfs2_journal_dirty(ctxt->handle, xis->inode_bh);
2594 }
85db90e7
TM		 2595out:
2596 return ret;
2597}
2598
6c3faba4
TY		 2599/*
2600 * This function only called duing creating inode
2601 * for init security/acl xattrs of the new inode.
2602 * The xattrs could be put into ibody or extent block,
2603 * xattr bucket would not be use in this case.
2604 * transanction credits also be reserved in here.
2605 */
2606int ocfs2_xattr_set_handle(handle_t *handle,
2607 struct inode *inode,
2608 struct buffer_head *di_bh,
2609 int name_index,
2610 const char *name,
2611 const void *value,
2612 size_t value_len,
2613 int flags,
2614 struct ocfs2_alloc_context *meta_ac,
2615 struct ocfs2_alloc_context *data_ac)
2616{
2617 struct ocfs2_dinode *di;
2618 int ret;
2619
2620 struct ocfs2_xattr_info xi = {
2621 .name_index = name_index,
2622 .name = name,
2623 .value = value,
2624 .value_len = value_len,
2625 };
2626
2627 struct ocfs2_xattr_search xis = {
2628 .not_found = -ENODATA,
2629 };
2630
2631 struct ocfs2_xattr_search xbs = {
2632 .not_found = -ENODATA,
2633 };
2634
2635 struct ocfs2_xattr_set_ctxt ctxt = {
2636 .handle = handle,
2637 .meta_ac = meta_ac,
2638 .data_ac = data_ac,
2639 };
2640
2641 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
2642 return -EOPNOTSUPP;
2643
2644 xis.inode_bh = xbs.inode_bh = di_bh;
2645 di = (struct ocfs2_dinode *)di_bh->b_data;
2646
2647 down_write(&OCFS2_I(inode)->ip_xattr_sem);
2648
2649 ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis);
2650 if (ret)
2651 goto cleanup;
2652 if (xis.not_found) {
2653 ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs);
2654 if (ret)
2655 goto cleanup;
2656 }
2657
2658 ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
2659
2660cleanup:
2661 up_write(&OCFS2_I(inode)->ip_xattr_sem);
2662 brelse(xbs.xattr_bh);
2663
2664 return ret;
2665}
2666
cf1d6c76
TY		 2667/*
2668 * ocfs2_xattr_set()
2669 *
2670 * Set, replace or remove an extended attribute for this inode.
2671 * value is NULL to remove an existing extended attribute, else either
2672 * create or replace an extended attribute.
2673 */
2674int ocfs2_xattr_set(struct inode *inode,
2675 int name_index,
2676 const char *name,
2677 const void *value,
2678 size_t value_len,
2679 int flags)
2680{
2681 struct buffer_head *di_bh = NULL;
2682 struct ocfs2_dinode *di;
85db90e7 2683 int ret, credits;
78f30c31 2684 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
85db90e7 2685 struct inode *tl_inode = osb->osb_tl_inode;
78f30c31 2686 struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, };
cf1d6c76
TY		 2687
2688 struct ocfs2_xattr_info xi = {
2689 .name_index = name_index,
2690 .name = name,
2691 .value = value,
2692 .value_len = value_len,
2693 };
2694
2695 struct ocfs2_xattr_search xis = {
2696 .not_found = -ENODATA,
2697 };
2698
2699 struct ocfs2_xattr_search xbs = {
2700 .not_found = -ENODATA,
2701 };
2702
8154da3d
TY		 2703 if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb)))
2704 return -EOPNOTSUPP;
2705
ba937127
JB		 2706 /*
2707 * Only xbs will be used on indexed trees. xis doesn't need a
2708 * bucket.
2709 */
2710 xbs.bucket = ocfs2_xattr_bucket_new(inode);
2711 if (!xbs.bucket) {
2712 mlog_errno(-ENOMEM);
2713 return -ENOMEM;
2714 }
2715
cf1d6c76
TY		 2716 ret = ocfs2_inode_lock(inode, &di_bh, 1);
2717 if (ret < 0) {
2718 mlog_errno(ret);
ba937127 2719 goto cleanup_nolock;
cf1d6c76
TY		 2720 }
2721 xis.inode_bh = xbs.inode_bh = di_bh;
2722 di = (struct ocfs2_dinode *)di_bh->b_data;
2723
2724 down_write(&OCFS2_I(inode)->ip_xattr_sem);
2725 /*
2726 * Scan inode and external block to find the same name
2727 * extended attribute and collect search infomation.
2728 */
2729 ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis);
2730 if (ret)
2731 goto cleanup;
2732 if (xis.not_found) {
2733 ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs);
2734 if (ret)
2735 goto cleanup;
2736 }
2737
2738 if (xis.not_found && xbs.not_found) {
2739 ret = -ENODATA;
2740 if (flags & XATTR_REPLACE)
2741 goto cleanup;
2742 ret = 0;
2743 if (!value)
2744 goto cleanup;
2745 } else {
2746 ret = -EEXIST;
2747 if (flags & XATTR_CREATE)
2748 goto cleanup;
2749 }
2750
85db90e7
TM		 2751
2752 mutex_lock(&tl_inode->i_mutex);
2753
2754 if (ocfs2_truncate_log_needs_flush(osb)) {
2755 ret = __ocfs2_flush_truncate_log(osb);
2756 if (ret < 0) {
2757 mutex_unlock(&tl_inode->i_mutex);
2758 mlog_errno(ret);
2759 goto cleanup;
2760 }
2761 }
2762 mutex_unlock(&tl_inode->i_mutex);
2763
2764 ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis,
2765 &xbs, &ctxt, &credits);
78f30c31
TM		 2766 if (ret) {
2767 mlog_errno(ret);
2768 goto cleanup;
2769 }
2770
4b3f6209
TM		 2771 /* we need to update inode's ctime field, so add credit for it. */
2772 credits += OCFS2_INODE_UPDATE_CREDITS;
85db90e7
TM		 2773 ctxt.handle = ocfs2_start_trans(osb, credits);
2774 if (IS_ERR(ctxt.handle)) {
2775 ret = PTR_ERR(ctxt.handle);
2776 mlog_errno(ret);
2777 goto cleanup;
cf1d6c76 2778 }
85db90e7
TM		 2779
2780 ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt);
2781
2782 ocfs2_commit_trans(osb, ctxt.handle);
2783
78f30c31
TM		 2784 if (ctxt.data_ac)
2785 ocfs2_free_alloc_context(ctxt.data_ac);
2786 if (ctxt.meta_ac)
2787 ocfs2_free_alloc_context(ctxt.meta_ac);
2788 if (ocfs2_dealloc_has_cluster(&ctxt.dealloc))
2789 ocfs2_schedule_truncate_log_flush(osb, 1);
2790 ocfs2_run_deallocs(osb, &ctxt.dealloc);
cf1d6c76
TY		 2791cleanup:
2792 up_write(&OCFS2_I(inode)->ip_xattr_sem);
2793 ocfs2_inode_unlock(inode, 1);
ba937127 2794cleanup_nolock:
cf1d6c76
TY		 2795 brelse(di_bh);
2796 brelse(xbs.xattr_bh);
ba937127 2797 ocfs2_xattr_bucket_free(xbs.bucket);
cf1d6c76
TY		 2798
2799 return ret;
2800}
2801
0c044f0b
TM		 2802/*
2803 * Find the xattr extent rec which may contains name_hash.
2804 * e_cpos will be the first name hash of the xattr rec.
2805 * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list.
2806 */
2807static int ocfs2_xattr_get_rec(struct inode *inode,
2808 u32 name_hash,
2809 u64 *p_blkno,
2810 u32 *e_cpos,
2811 u32 *num_clusters,
2812 struct ocfs2_extent_list *el)
2813{
2814 int ret = 0, i;
2815 struct buffer_head *eb_bh = NULL;
2816 struct ocfs2_extent_block *eb;
2817 struct ocfs2_extent_rec *rec = NULL;
2818 u64 e_blkno = 0;
2819
2820 if (el->l_tree_depth) {
2821 ret = ocfs2_find_leaf(inode, el, name_hash, &eb_bh);
2822 if (ret) {
2823 mlog_errno(ret);
2824 goto out;
2825 }
2826
2827 eb = (struct ocfs2_extent_block *) eb_bh->b_data;
2828 el = &eb->h_list;
2829
2830 if (el->l_tree_depth) {
2831 ocfs2_error(inode->i_sb,
2832 "Inode %lu has non zero tree depth in "
2833 "xattr tree block %llu\n", inode->i_ino,
2834 (unsigned long long)eb_bh->b_blocknr);
2835 ret = -EROFS;
2836 goto out;
2837 }
2838 }
2839
2840 for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) {
2841 rec = &el->l_recs[i];
2842
2843 if (le32_to_cpu(rec->e_cpos) <= name_hash) {
2844 e_blkno = le64_to_cpu(rec->e_blkno);
2845 break;
2846 }
2847 }
2848
2849 if (!e_blkno) {
2850 ocfs2_error(inode->i_sb, "Inode %lu has bad extent "
2851 "record (%u, %u, 0) in xattr", inode->i_ino,
2852 le32_to_cpu(rec->e_cpos),
2853 ocfs2_rec_clusters(el, rec));
2854 ret = -EROFS;
2855 goto out;
2856 }
2857
2858 *p_blkno = le64_to_cpu(rec->e_blkno);
2859 *num_clusters = le16_to_cpu(rec->e_leaf_clusters);
2860 if (e_cpos)
2861 *e_cpos = le32_to_cpu(rec->e_cpos);
2862out:
2863 brelse(eb_bh);
2864 return ret;
2865}
2866
2867typedef int (xattr_bucket_func)(struct inode *inode,
2868 struct ocfs2_xattr_bucket *bucket,
2869 void *para);
2870
589dc260 2871static int ocfs2_find_xe_in_bucket(struct inode *inode,
e2356a3f 2872 struct ocfs2_xattr_bucket *bucket,
589dc260
TM		 2873 int name_index,
2874 const char *name,
2875 u32 name_hash,
2876 u16 *xe_index,
2877 int *found)
2878{
2879 int i, ret = 0, cmp = 1, block_off, new_offset;
e2356a3f 2880 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
589dc260
TM		 2881 size_t name_len = strlen(name);
2882 struct ocfs2_xattr_entry *xe = NULL;
589dc260
TM		 2883 char *xe_name;
2884
2885 /*
2886 * We don't use binary search in the bucket because there
2887 * may be multiple entries with the same name hash.
2888 */
2889 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
2890 xe = &xh->xh_entries[i];
2891
2892 if (name_hash > le32_to_cpu(xe->xe_name_hash))
2893 continue;
2894 else if (name_hash < le32_to_cpu(xe->xe_name_hash))
2895 break;
2896
2897 cmp = name_index - ocfs2_xattr_get_type(xe);
2898 if (!cmp)
2899 cmp = name_len - xe->xe_name_len;
2900 if (cmp)
2901 continue;
2902
2903 ret = ocfs2_xattr_bucket_get_name_value(inode,
2904 xh,
2905 i,
2906 &block_off,
2907 &new_offset);
2908 if (ret) {
2909 mlog_errno(ret);
2910 break;
2911 }
2912
970e4936 2913
e2356a3f
JB		 2914 xe_name = bucket_block(bucket, block_off) + new_offset;
2915 if (!memcmp(name, xe_name, name_len)) {
589dc260
TM		 2916 *xe_index = i;
2917 *found = 1;
2918 ret = 0;
2919 break;
2920 }
2921 }
2922
2923 return ret;
2924}
2925
2926/*
2927 * Find the specified xattr entry in a series of buckets.
2928 * This series start from p_blkno and last for num_clusters.
2929 * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains
2930 * the num of the valid buckets.
2931 *
2932 * Return the buffer_head this xattr should reside in. And if the xattr's
2933 * hash is in the gap of 2 buckets, return the lower bucket.
2934 */
2935static int ocfs2_xattr_bucket_find(struct inode *inode,
2936 int name_index,
2937 const char *name,
2938 u32 name_hash,
2939 u64 p_blkno,
2940 u32 first_hash,
2941 u32 num_clusters,
2942 struct ocfs2_xattr_search *xs)
2943{
2944 int ret, found = 0;
589dc260
TM		 2945 struct ocfs2_xattr_header *xh = NULL;
2946 struct ocfs2_xattr_entry *xe = NULL;
2947 u16 index = 0;
2948 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
2949 int low_bucket = 0, bucket, high_bucket;
e2356a3f 2950 struct ocfs2_xattr_bucket *search;
589dc260 2951 u32 last_hash;
e2356a3f 2952 u64 blkno, lower_blkno = 0;
589dc260 2953
e2356a3f
JB		 2954 search = ocfs2_xattr_bucket_new(inode);
2955 if (!search) {
2956 ret = -ENOMEM;
2957 mlog_errno(ret);
2958 goto out;
2959 }
2960
2961 ret = ocfs2_read_xattr_bucket(search, p_blkno);
589dc260
TM		 2962 if (ret) {
2963 mlog_errno(ret);
2964 goto out;
2965 }
2966
e2356a3f 2967 xh = bucket_xh(search);
589dc260 2968 high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1;
589dc260 2969 while (low_bucket <= high_bucket) {
e2356a3f 2970 ocfs2_xattr_bucket_relse(search);
589dc260 2971
e2356a3f 2972 bucket = (low_bucket + high_bucket) / 2;
589dc260 2973 blkno = p_blkno + bucket * blk_per_bucket;
e2356a3f 2974 ret = ocfs2_read_xattr_bucket(search, blkno);
589dc260
TM		 2975 if (ret) {
2976 mlog_errno(ret);
2977 goto out;
2978 }
2979
e2356a3f 2980 xh = bucket_xh(search);
589dc260
TM		 2981 xe = &xh->xh_entries[0];
2982 if (name_hash < le32_to_cpu(xe->xe_name_hash)) {
2983 high_bucket = bucket - 1;
2984 continue;
2985 }
2986
2987 /*
2988 * Check whether the hash of the last entry in our
5a095611
TM		 2989 * bucket is larger than the search one. for an empty
2990 * bucket, the last one is also the first one.
589dc260 2991 */
5a095611
TM		 2992 if (xh->xh_count)
2993 xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1];
2994
589dc260
TM		 2995 last_hash = le32_to_cpu(xe->xe_name_hash);
2996
e2356a3f
JB		 2997 /* record lower_blkno which may be the insert place. */
2998 lower_blkno = blkno;
589dc260
TM		 2999
3000 if (name_hash > le32_to_cpu(xe->xe_name_hash)) {
3001 low_bucket = bucket + 1;
3002 continue;
3003 }
3004
3005 /* the searched xattr should reside in this bucket if exists. */
e2356a3f 3006 ret = ocfs2_find_xe_in_bucket(inode, search,
589dc260
TM		 3007 name_index, name, name_hash,
3008 &index, &found);
3009 if (ret) {
3010 mlog_errno(ret);
3011 goto out;
3012 }
3013 break;
3014 }
3015
3016 /*
3017 * Record the bucket we have found.
3018 * When the xattr's hash value is in the gap of 2 buckets, we will
3019 * always set it to the previous bucket.
3020 */
e2356a3f
JB		 3021 if (!lower_blkno)
3022 lower_blkno = p_blkno;
3023
3024 /* This should be in cache - we just read it during the search */
3025 ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno);
3026 if (ret) {
3027 mlog_errno(ret);
3028 goto out;
589dc260 3029 }
589dc260 3030
ba937127
JB		 3031 xs->header = bucket_xh(xs->bucket);
3032 xs->base = bucket_block(xs->bucket, 0);
589dc260
TM		 3033 xs->end = xs->base + inode->i_sb->s_blocksize;
3034
3035 if (found) {
589dc260
TM		 3036 xs->here = &xs->header->xh_entries[index];
3037 mlog(0, "find xattr %s in bucket %llu, entry = %u\n", name,
ba937127 3038 (unsigned long long)bucket_blkno(xs->bucket), index);
589dc260
TM		 3039 } else
3040 ret = -ENODATA;
3041
3042out:
e2356a3f 3043 ocfs2_xattr_bucket_free(search);
589dc260
TM		 3044 return ret;
3045}
3046
3047static int ocfs2_xattr_index_block_find(struct inode *inode,
3048 struct buffer_head *root_bh,
3049 int name_index,
3050 const char *name,
3051 struct ocfs2_xattr_search *xs)
3052{
3053 int ret;
3054 struct ocfs2_xattr_block *xb =
3055 (struct ocfs2_xattr_block *)root_bh->b_data;
3056 struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root;
3057 struct ocfs2_extent_list *el = &xb_root->xt_list;
3058 u64 p_blkno = 0;
3059 u32 first_hash, num_clusters = 0;
2057e5c6 3060 u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name));
589dc260
TM		 3061
3062 if (le16_to_cpu(el->l_next_free_rec) == 0)
3063 return -ENODATA;
3064
3065 mlog(0, "find xattr %s, hash = %u, index = %d in xattr tree\n",
3066 name, name_hash, name_index);
3067
3068 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash,
3069 &num_clusters, el);
3070 if (ret) {
3071 mlog_errno(ret);
3072 goto out;
3073 }
3074
3075 BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash);
3076
3077 mlog(0, "find xattr extent rec %u clusters from %llu, the first hash "
de29c085
MF		 3078 "in the rec is %u\n", num_clusters, (unsigned long long)p_blkno,
3079 first_hash);
589dc260
TM		 3080
3081 ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash,
3082 p_blkno, first_hash, num_clusters, xs);
3083
3084out:
3085 return ret;
3086}
3087
0c044f0b
TM		 3088static int ocfs2_iterate_xattr_buckets(struct inode *inode,
3089 u64 blkno,
3090 u32 clusters,
3091 xattr_bucket_func *func,
3092 void *para)
3093{
6dde41d9 3094 int i, ret = 0;
0c044f0b
TM		 3095 u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb));
3096 u32 num_buckets = clusters * bpc;
ba937127 3097 struct ocfs2_xattr_bucket *bucket;
0c044f0b 3098
ba937127
JB		 3099 bucket = ocfs2_xattr_bucket_new(inode);
3100 if (!bucket) {
3101 mlog_errno(-ENOMEM);
3102 return -ENOMEM;
3103 }
0c044f0b
TM		 3104
3105 mlog(0, "iterating xattr buckets in %u clusters starting from %llu\n",
de29c085 3106 clusters, (unsigned long long)blkno);
0c044f0b 3107
ba937127
JB		 3108 for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) {
3109 ret = ocfs2_read_xattr_bucket(bucket, blkno);
0c044f0b
TM		 3110 if (ret) {
3111 mlog_errno(ret);
ba937127 3112 break;
0c044f0b
TM		 3113 }
3114
0c044f0b
TM		 3115 /*
3116 * The real bucket num in this series of blocks is stored
3117 * in the 1st bucket.
3118 */
3119 if (i == 0)
ba937127 3120 num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets);
0c044f0b 3121
de29c085
MF		 3122 mlog(0, "iterating xattr bucket %llu, first hash %u\n",
3123 (unsigned long long)blkno,
ba937127 3124 le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash));
0c044f0b 3125 if (func) {
ba937127
JB		 3126 ret = func(inode, bucket, para);
3127 if (ret)
0c044f0b 3128 mlog_errno(ret);
ba937127 3129 /* Fall through to bucket_relse() */
0c044f0b
TM		 3130 }
3131
ba937127
JB		 3132 ocfs2_xattr_bucket_relse(bucket);
3133 if (ret)
3134 break;
0c044f0b
TM		 3135 }
3136
ba937127 3137 ocfs2_xattr_bucket_free(bucket);
0c044f0b
TM		 3138 return ret;
3139}
3140
3141struct ocfs2_xattr_tree_list {
3142 char *buffer;
3143 size_t buffer_size;
936b8834 3144 size_t result;
0c044f0b
TM		 3145};
3146
3147static int ocfs2_xattr_bucket_get_name_value(struct inode *inode,
3148 struct ocfs2_xattr_header *xh,
3149 int index,
3150 int *block_off,
3151 int *new_offset)
3152{
3153 u16 name_offset;
3154
3155 if (index < 0 || index >= le16_to_cpu(xh->xh_count))
3156 return -EINVAL;
3157
3158 name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset);
3159
3160 *block_off = name_offset >> inode->i_sb->s_blocksize_bits;
3161 *new_offset = name_offset % inode->i_sb->s_blocksize;
3162
3163 return 0;
3164}
3165
3166static int ocfs2_list_xattr_bucket(struct inode *inode,
3167 struct ocfs2_xattr_bucket *bucket,
3168 void *para)
3169{
936b8834 3170 int ret = 0, type;
0c044f0b 3171 struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para;
0c044f0b 3172 int i, block_off, new_offset;
936b8834 3173 const char *prefix, *name;
0c044f0b 3174
3e632946
JB		 3175 for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) {
3176 struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i];
936b8834
TM		 3177 type = ocfs2_xattr_get_type(entry);
3178 prefix = ocfs2_xattr_prefix(type);
0c044f0b 3179
936b8834 3180 if (prefix) {
0c044f0b 3181 ret = ocfs2_xattr_bucket_get_name_value(inode,
3e632946 3182 bucket_xh(bucket),
0c044f0b
TM		 3183 i,
3184 &block_off,
3185 &new_offset);
3186 if (ret)
3187 break;
936b8834 3188
51def39f 3189 name = (const char *)bucket_block(bucket, block_off) +
936b8834
TM		 3190 new_offset;
3191 ret = ocfs2_xattr_list_entry(xl->buffer,
3192 xl->buffer_size,
3193 &xl->result,
3194 prefix, name,
3195 entry->xe_name_len);
3196 if (ret)
3197 break;
0c044f0b
TM		 3198 }
3199 }
3200
3201 return ret;
3202}
3203
3204static int ocfs2_xattr_tree_list_index_block(struct inode *inode,
3205 struct ocfs2_xattr_tree_root *xt,
3206 char *buffer,
3207 size_t buffer_size)
3208{
3209 struct ocfs2_extent_list *el = &xt->xt_list;
3210 int ret = 0;
3211 u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0;
3212 u64 p_blkno = 0;
3213 struct ocfs2_xattr_tree_list xl = {
3214 .buffer = buffer,
3215 .buffer_size = buffer_size,
936b8834 3216 .result = 0,
0c044f0b
TM		 3217 };
3218
3219 if (le16_to_cpu(el->l_next_free_rec) == 0)
3220 return 0;
3221
3222 while (name_hash > 0) {
3223 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno,
3224 &e_cpos, &num_clusters, el);
3225 if (ret) {
3226 mlog_errno(ret);
3227 goto out;
3228 }
3229
3230 ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters,
3231 ocfs2_list_xattr_bucket,
3232 &xl);
3233 if (ret) {
3234 mlog_errno(ret);
3235 goto out;
3236 }
3237
3238 if (e_cpos == 0)
3239 break;
3240
3241 name_hash = e_cpos - 1;
3242 }
3243
936b8834 3244 ret = xl.result;
0c044f0b
TM		 3245out:
3246 return ret;
3247}
01225596
TM		 3248
3249static int cmp_xe(const void *a, const void *b)
3250{
3251 const struct ocfs2_xattr_entry *l = a, *r = b;
3252 u32 l_hash = le32_to_cpu(l->xe_name_hash);
3253 u32 r_hash = le32_to_cpu(r->xe_name_hash);
3254
3255 if (l_hash > r_hash)
3256 return 1;
3257 if (l_hash < r_hash)
3258 return -1;
3259 return 0;
3260}
3261
3262static void swap_xe(void *a, void *b, int size)
3263{
3264 struct ocfs2_xattr_entry *l = a, *r = b, tmp;
3265
3266 tmp = *l;
3267 memcpy(l, r, sizeof(struct ocfs2_xattr_entry));
3268 memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry));
3269}
3270
3271/*
3272 * When the ocfs2_xattr_block is filled up, new bucket will be created
3273 * and all the xattr entries will be moved to the new bucket.
178eeac3
JB		 3274 * The header goes at the start of the bucket, and the names+values are
3275 * filled from the end. This is why *target starts as the last buffer.
01225596
TM		 3276 * Note: we need to sort the entries since they are not saved in order
3277 * in the ocfs2_xattr_block.
3278 */
3279static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode,
3280 struct buffer_head *xb_bh,
178eeac3 3281 struct ocfs2_xattr_bucket *bucket)
01225596
TM		 3282{
3283 int i, blocksize = inode->i_sb->s_blocksize;
178eeac3 3284 int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
01225596
TM		 3285 u16 offset, size, off_change;
3286 struct ocfs2_xattr_entry *xe;
3287 struct ocfs2_xattr_block *xb =
3288 (struct ocfs2_xattr_block *)xb_bh->b_data;
3289 struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header;
178eeac3 3290 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
01225596 3291 u16 count = le16_to_cpu(xb_xh->xh_count);
178eeac3
JB		 3292 char *src = xb_bh->b_data;
3293 char *target = bucket_block(bucket, blks - 1);
01225596
TM		 3294
3295 mlog(0, "cp xattr from block %llu to bucket %llu\n",
3296 (unsigned long long)xb_bh->b_blocknr,
178eeac3
JB		 3297 (unsigned long long)bucket_blkno(bucket));
3298
3299 for (i = 0; i < blks; i++)
3300 memset(bucket_block(bucket, i), 0, blocksize);
01225596 3301
01225596
TM		 3302 /*
3303 * Since the xe_name_offset is based on ocfs2_xattr_header,
3304 * there is a offset change corresponding to the change of
3305 * ocfs2_xattr_header's position.
3306 */
3307 off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header);
3308 xe = &xb_xh->xh_entries[count - 1];
3309 offset = le16_to_cpu(xe->xe_name_offset) + off_change;
3310 size = blocksize - offset;
3311
3312 /* copy all the names and values. */
01225596
TM		 3313 memcpy(target + offset, src + offset, size);
3314
3315 /* Init new header now. */
3316 xh->xh_count = xb_xh->xh_count;
3317 xh->xh_num_buckets = cpu_to_le16(1);
3318 xh->xh_name_value_len = cpu_to_le16(size);
3319 xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size);
3320
3321 /* copy all the entries. */
178eeac3 3322 target = bucket_block(bucket, 0);
01225596
TM		 3323 offset = offsetof(struct ocfs2_xattr_header, xh_entries);
3324 size = count * sizeof(struct ocfs2_xattr_entry);
3325 memcpy(target + offset, (char *)xb_xh + offset, size);
3326
3327 /* Change the xe offset for all the xe because of the move. */
3328 off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize +
3329 offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header);
3330 for (i = 0; i < count; i++)
3331 le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change);
3332
3333 mlog(0, "copy entry: start = %u, size = %u, offset_change = %u\n",
3334 offset, size, off_change);
3335
3336 sort(target + offset, count, sizeof(struct ocfs2_xattr_entry),
3337 cmp_xe, swap_xe);
3338}
3339
3340/*
3341 * After we move xattr from block to index btree, we have to
3342 * update ocfs2_xattr_search to the new xe and base.
3343 *
3344 * When the entry is in xattr block, xattr_bh indicates the storage place.
3345 * While if the entry is in index b-tree, "bucket" indicates the
3346 * real place of the xattr.
3347 */
178eeac3
JB		 3348static void ocfs2_xattr_update_xattr_search(struct inode *inode,
3349 struct ocfs2_xattr_search *xs,
3350 struct buffer_head *old_bh)
01225596 3351{
01225596
TM		 3352 char *buf = old_bh->b_data;
3353 struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf;
3354 struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header;
178eeac3 3355 int i;
01225596 3356
ba937127 3357 xs->header = bucket_xh(xs->bucket);
178eeac3 3358 xs->base = bucket_block(xs->bucket, 0);
01225596
TM		 3359 xs->end = xs->base + inode->i_sb->s_blocksize;
3360
178eeac3
JB		 3361 if (xs->not_found)
3362 return;
01225596 3363
178eeac3
JB		 3364 i = xs->here - old_xh->xh_entries;
3365 xs->here = &xs->header->xh_entries[i];
01225596
TM		 3366}
3367
3368static int ocfs2_xattr_create_index_block(struct inode *inode,
78f30c31
TM		 3369 struct ocfs2_xattr_search *xs,
3370 struct ocfs2_xattr_set_ctxt *ctxt)
01225596 3371{
85db90e7 3372 int ret;
01225596
TM		 3373 u32 bit_off, len;
3374 u64 blkno;
85db90e7 3375 handle_t *handle = ctxt->handle;
01225596
TM		 3376 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
3377 struct ocfs2_inode_info *oi = OCFS2_I(inode);
01225596
TM		 3378 struct buffer_head *xb_bh = xs->xattr_bh;
3379 struct ocfs2_xattr_block *xb =
3380 (struct ocfs2_xattr_block *)xb_bh->b_data;
3381 struct ocfs2_xattr_tree_root *xr;
3382 u16 xb_flags = le16_to_cpu(xb->xb_flags);
01225596
TM		 3383
3384 mlog(0, "create xattr index block for %llu\n",
3385 (unsigned long long)xb_bh->b_blocknr);
3386
3387 BUG_ON(xb_flags & OCFS2_XATTR_INDEXED);
178eeac3 3388 BUG_ON(!xs->bucket);
01225596 3389
01225596
TM		 3390 /*
3391 * XXX:
3392 * We can use this lock for now, and maybe move to a dedicated mutex
3393 * if performance becomes a problem later.
3394 */
3395 down_write(&oi->ip_alloc_sem);
3396
84008972
JB		 3397 ret = ocfs2_journal_access_xb(handle, inode, xb_bh,
3398 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 3399 if (ret) {
3400 mlog_errno(ret);
85db90e7 3401 goto out;
01225596
TM		 3402 }
3403
78f30c31
TM		 3404 ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac,
3405 1, 1, &bit_off, &len);
01225596
TM		 3406 if (ret) {
3407 mlog_errno(ret);
85db90e7 3408 goto out;
01225596
TM		 3409 }
3410
3411 /*
3412 * The bucket may spread in many blocks, and
3413 * we will only touch the 1st block and the last block
3414 * in the whole bucket(one for entry and one for data).
3415 */
3416 blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off);
3417
de29c085
MF		 3418 mlog(0, "allocate 1 cluster from %llu to xattr block\n",
3419 (unsigned long long)blkno);
01225596 3420
178eeac3
JB		 3421 ret = ocfs2_init_xattr_bucket(xs->bucket, blkno);
3422 if (ret) {
01225596 3423 mlog_errno(ret);
85db90e7 3424 goto out;
01225596
TM		 3425 }
3426
178eeac3
JB		 3427 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
3428 OCFS2_JOURNAL_ACCESS_CREATE);
01225596
TM		 3429 if (ret) {
3430 mlog_errno(ret);
85db90e7 3431 goto out;
01225596
TM		 3432 }
3433
178eeac3
JB		 3434 ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket);
3435 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
01225596 3436
178eeac3 3437 ocfs2_xattr_update_xattr_search(inode, xs, xb_bh);
01225596
TM		 3438
3439 /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */
3440 memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize -
3441 offsetof(struct ocfs2_xattr_block, xb_attrs));
3442
3443 xr = &xb->xb_attrs.xb_root;
3444 xr->xt_clusters = cpu_to_le32(1);
3445 xr->xt_last_eb_blk = 0;
3446 xr->xt_list.l_tree_depth = 0;
3447 xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb));
3448 xr->xt_list.l_next_free_rec = cpu_to_le16(1);
3449
3450 xr->xt_list.l_recs[0].e_cpos = 0;
3451 xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno);
3452 xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1);
3453
3454 xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED);
3455
85db90e7 3456 ocfs2_journal_dirty(handle, xb_bh);
01225596 3457
85db90e7 3458out:
01225596
TM		 3459 up_write(&oi->ip_alloc_sem);
3460
01225596
TM		 3461 return ret;
3462}
3463
3464static int cmp_xe_offset(const void *a, const void *b)
3465{
3466 const struct ocfs2_xattr_entry *l = a, *r = b;
3467 u32 l_name_offset = le16_to_cpu(l->xe_name_offset);
3468 u32 r_name_offset = le16_to_cpu(r->xe_name_offset);
3469
3470 if (l_name_offset < r_name_offset)
3471 return 1;
3472 if (l_name_offset > r_name_offset)
3473 return -1;
3474 return 0;
3475}
3476
3477/*
3478 * defrag a xattr bucket if we find that the bucket has some
3479 * holes beteen name/value pairs.
3480 * We will move all the name/value pairs to the end of the bucket
3481 * so that we can spare some space for insertion.
3482 */
3483static int ocfs2_defrag_xattr_bucket(struct inode *inode,
85db90e7 3484 handle_t *handle,
01225596
TM		 3485 struct ocfs2_xattr_bucket *bucket)
3486{
3487 int ret, i;
3488 size_t end, offset, len, value_len;
3489 struct ocfs2_xattr_header *xh;
3490 char *entries, *buf, *bucket_buf = NULL;
9c7759aa 3491 u64 blkno = bucket_blkno(bucket);
01225596 3492 u16 xh_free_start;
01225596 3493 size_t blocksize = inode->i_sb->s_blocksize;
01225596 3494 struct ocfs2_xattr_entry *xe;
01225596
TM		 3495
3496 /*
3497 * In order to make the operation more efficient and generic,
3498 * we copy all the blocks into a contiguous memory and do the
3499 * defragment there, so if anything is error, we will not touch
3500 * the real block.
3501 */
3502 bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS);
3503 if (!bucket_buf) {
3504 ret = -EIO;
3505 goto out;
3506 }
3507
3508 buf = bucket_buf;
1c32a2fd
TM		 3509 for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
3510 memcpy(buf, bucket_block(bucket, i), blocksize);
01225596 3511
1c32a2fd 3512 ret = ocfs2_xattr_bucket_journal_access(handle, bucket,
161d6f30
JB		 3513 OCFS2_JOURNAL_ACCESS_WRITE);
3514 if (ret < 0) {
3515 mlog_errno(ret);
85db90e7 3516 goto out;
01225596
TM		 3517 }
3518
3519 xh = (struct ocfs2_xattr_header *)bucket_buf;
3520 entries = (char *)xh->xh_entries;
3521 xh_free_start = le16_to_cpu(xh->xh_free_start);
3522
3523 mlog(0, "adjust xattr bucket in %llu, count = %u, "
3524 "xh_free_start = %u, xh_name_value_len = %u.\n",
de29c085
MF		 3525 (unsigned long long)blkno, le16_to_cpu(xh->xh_count),
3526 xh_free_start, le16_to_cpu(xh->xh_name_value_len));
01225596
TM		 3527
3528 /*
3529 * sort all the entries by their offset.
3530 * the largest will be the first, so that we can
3531 * move them to the end one by one.
3532 */
3533 sort(entries, le16_to_cpu(xh->xh_count),
3534 sizeof(struct ocfs2_xattr_entry),
3535 cmp_xe_offset, swap_xe);
3536
3537 /* Move all name/values to the end of the bucket. */
3538 xe = xh->xh_entries;
3539 end = OCFS2_XATTR_BUCKET_SIZE;
3540 for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) {
3541 offset = le16_to_cpu(xe->xe_name_offset);
3542 if (ocfs2_xattr_is_local(xe))
3543 value_len = OCFS2_XATTR_SIZE(
3544 le64_to_cpu(xe->xe_value_size));
3545 else
3546 value_len = OCFS2_XATTR_ROOT_SIZE;
3547 len = OCFS2_XATTR_SIZE(xe->xe_name_len) + value_len;
3548
3549 /*
3550 * We must make sure that the name/value pair
3551 * exist in the same block. So adjust end to
3552 * the previous block end if needed.
3553 */
3554 if (((end - len) / blocksize !=
3555 (end - 1) / blocksize))
3556 end = end - end % blocksize;
3557
3558 if (end > offset + len) {
3559 memmove(bucket_buf + end - len,
3560 bucket_buf + offset, len);
3561 xe->xe_name_offset = cpu_to_le16(end - len);
3562 }
3563
3564 mlog_bug_on_msg(end < offset + len, "Defrag check failed for "
3565 "bucket %llu\n", (unsigned long long)blkno);
3566
3567 end -= len;
3568 }
3569
3570 mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for "
3571 "bucket %llu\n", (unsigned long long)blkno);
3572
3573 if (xh_free_start == end)
85db90e7 3574 goto out;
01225596
TM		 3575
3576 memset(bucket_buf + xh_free_start, 0, end - xh_free_start);
3577 xh->xh_free_start = cpu_to_le16(end);
3578
3579 /* sort the entries by their name_hash. */
3580 sort(entries, le16_to_cpu(xh->xh_count),
3581 sizeof(struct ocfs2_xattr_entry),
3582 cmp_xe, swap_xe);
3583
3584 buf = bucket_buf;
1c32a2fd
TM		 3585 for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize)
3586 memcpy(bucket_block(bucket, i), buf, blocksize);
3587 ocfs2_xattr_bucket_journal_dirty(handle, bucket);
01225596 3588
01225596 3589out:
01225596
TM		 3590 kfree(bucket_buf);
3591 return ret;
3592}
3593
3594/*
b5c03e74
JB		 3595 * prev_blkno points to the start of an existing extent. new_blkno
3596 * points to a newly allocated extent. Because we know each of our
3597 * clusters contains more than bucket, we can easily split one cluster
3598 * at a bucket boundary. So we take the last cluster of the existing
3599 * extent and split it down the middle. We move the last half of the
3600 * buckets in the last cluster of the existing extent over to the new
3601 * extent.
01225596 3602 *
b5c03e74
JB		 3603 * first_bh is the buffer at prev_blkno so we can update the existing
3604 * extent's bucket count. header_bh is the bucket were we were hoping
3605 * to insert our xattr. If the bucket move places the target in the new
3606 * extent, we'll update first_bh and header_bh after modifying the old
3607 * extent.
3608 *
3609 * first_hash will be set as the 1st xe's name_hash in the new extent.
01225596
TM		 3610 */
3611static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode,
3612 handle_t *handle,
41cb8148
JB		 3613 struct ocfs2_xattr_bucket *first,
3614 struct ocfs2_xattr_bucket *target,
01225596 3615 u64 new_blkno,
01225596
TM		 3616 u32 num_clusters,
3617 u32 *first_hash)
3618{
c58b6032 3619 int ret;
41cb8148
JB		 3620 struct super_block *sb = inode->i_sb;
3621 int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb);
3622 int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb));
b5c03e74 3623 int to_move = num_buckets / 2;
c58b6032 3624 u64 src_blkno;
41cb8148
JB		 3625 u64 last_cluster_blkno = bucket_blkno(first) +
3626 ((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1));
01225596 3627
41cb8148
JB		 3628 BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets);
3629 BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize);
01225596 3630
01225596 3631 mlog(0, "move half of xattrs in cluster %llu to %llu\n",
c58b6032 3632 (unsigned long long)last_cluster_blkno, (unsigned long long)new_blkno);
01225596 3633
41cb8148 3634 ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first),
c58b6032
JB		 3635 last_cluster_blkno, new_blkno,
3636 to_move, first_hash);
01225596
TM		 3637 if (ret) {
3638 mlog_errno(ret);
3639 goto out;
3640 }
3641
c58b6032
JB		 3642 /* This is the first bucket that got moved */
3643 src_blkno = last_cluster_blkno + (to_move * blks_per_bucket);
01225596 3644
b5c03e74 3645 /*
c58b6032 3646 * If the target bucket was part of the moved buckets, we need to
41cb8148 3647 * update first and target.
b5c03e74 3648 */
41cb8148 3649 if (bucket_blkno(target) >= src_blkno) {
b5c03e74
JB		 3650 /* Find the block for the new target bucket */
3651 src_blkno = new_blkno +
41cb8148
JB		 3652 (bucket_blkno(target) - src_blkno);
3653
3654 ocfs2_xattr_bucket_relse(first);
3655 ocfs2_xattr_bucket_relse(target);
01225596 3656
b5c03e74 3657 /*
c58b6032 3658 * These shouldn't fail - the buffers are in the
b5c03e74
JB		 3659 * journal from ocfs2_cp_xattr_bucket().
3660 */
41cb8148 3661 ret = ocfs2_read_xattr_bucket(first, new_blkno);
c58b6032
JB		 3662 if (ret) {
3663 mlog_errno(ret);
3664 goto out;
3665 }
41cb8148
JB		 3666 ret = ocfs2_read_xattr_bucket(target, src_blkno);
3667 if (ret)
b5c03e74 3668 mlog_errno(ret);
01225596 3669
01225596
TM		 3670 }
3671
01225596 3672out:
01225596
TM		 3673 return ret;
3674}
3675
01225596 3676/*
80bcaf34
TM		 3677 * Find the suitable pos when we divide a bucket into 2.
3678 * We have to make sure the xattrs with the same hash value exist
3679 * in the same bucket.
3680 *
3681 * If this ocfs2_xattr_header covers more than one hash value, find a
3682 * place where the hash value changes. Try to find the most even split.
3683 * The most common case is that all entries have different hash values,
3684 * and the first check we make will find a place to split.
3685 */
3686static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh)
3687{
3688 struct ocfs2_xattr_entry *entries = xh->xh_entries;
3689 int count = le16_to_cpu(xh->xh_count);
3690 int delta, middle = count / 2;
3691
3692 /*
3693 * We start at the middle. Each step gets farther away in both
3694 * directions. We therefore hit the change in hash value
3695 * nearest to the middle. Note that this loop does not execute for
3696 * count < 2.
3697 */
3698 for (delta = 0; delta < middle; delta++) {
3699 /* Let's check delta earlier than middle */
3700 if (cmp_xe(&entries[middle - delta - 1],
3701 &entries[middle - delta]))
3702 return middle - delta;
3703
3704 /* For even counts, don't walk off the end */
3705 if ((middle + delta + 1) == count)
3706 continue;
3707
3708 /* Now try delta past middle */
3709 if (cmp_xe(&entries[middle + delta],
3710 &entries[middle + delta + 1]))
3711 return middle + delta + 1;
3712 }
3713
3714 /* Every entry had the same hash */
3715 return count;
3716}
3717
3718/*
3719 * Move some xattrs in old bucket(blk) to new bucket(new_blk).
01225596 3720 * first_hash will record the 1st hash of the new bucket.
80bcaf34
TM		 3721 *
3722 * Normally half of the xattrs will be moved. But we have to make
3723 * sure that the xattrs with the same hash value are stored in the
3724 * same bucket. If all the xattrs in this bucket have the same hash
3725 * value, the new bucket will be initialized as an empty one and the
3726 * first_hash will be initialized as (hash_value+1).
01225596 3727 */
80bcaf34
TM		 3728static int ocfs2_divide_xattr_bucket(struct inode *inode,
3729 handle_t *handle,
3730 u64 blk,
3731 u64 new_blk,
3732 u32 *first_hash,
3733 int new_bucket_head)
01225596
TM		 3734{
3735 int ret, i;
80bcaf34 3736 int count, start, len, name_value_len = 0, xe_len, name_offset = 0;
ba937127 3737 struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
01225596
TM		 3738 struct ocfs2_xattr_header *xh;
3739 struct ocfs2_xattr_entry *xe;
3740 int blocksize = inode->i_sb->s_blocksize;
3741
80bcaf34 3742 mlog(0, "move some of xattrs from bucket %llu to %llu\n",
de29c085 3743 (unsigned long long)blk, (unsigned long long)new_blk);
01225596 3744
ba937127
JB		 3745 s_bucket = ocfs2_xattr_bucket_new(inode);
3746 t_bucket = ocfs2_xattr_bucket_new(inode);
3747 if (!s_bucket || !t_bucket) {
3748 ret = -ENOMEM;
3749 mlog_errno(ret);
3750 goto out;
3751 }
01225596 3752
ba937127 3753 ret = ocfs2_read_xattr_bucket(s_bucket, blk);
01225596
TM		 3754 if (ret) {
3755 mlog_errno(ret);
3756 goto out;
3757 }
3758
ba937127 3759 ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket,
1224be02 3760 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 3761 if (ret) {
3762 mlog_errno(ret);
3763 goto out;
3764 }
3765
784b816a
JB		 3766 /*
3767 * Even if !new_bucket_head, we're overwriting t_bucket. Thus,
3768 * there's no need to read it.
3769 */
ba937127 3770 ret = ocfs2_init_xattr_bucket(t_bucket, new_blk);
01225596
TM		 3771 if (ret) {
3772 mlog_errno(ret);
3773 goto out;
3774 }
3775
2b656c1d
JB		 3776 /*
3777 * Hey, if we're overwriting t_bucket, what difference does
3778 * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the
3779 * same part of ocfs2_cp_xattr_bucket().
3780 */
ba937127 3781 ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
1224be02
JB		 3782 new_bucket_head ?
3783 OCFS2_JOURNAL_ACCESS_CREATE :
3784 OCFS2_JOURNAL_ACCESS_WRITE);
3785 if (ret) {
3786 mlog_errno(ret);
3787 goto out;
01225596
TM		 3788 }
3789
ba937127 3790 xh = bucket_xh(s_bucket);
80bcaf34
TM		 3791 count = le16_to_cpu(xh->xh_count);
3792 start = ocfs2_xattr_find_divide_pos(xh);
3793
3794 if (start == count) {
3795 xe = &xh->xh_entries[start-1];
3796
3797 /*
3798 * initialized a new empty bucket here.
3799 * The hash value is set as one larger than
3800 * that of the last entry in the previous bucket.
3801 */
ba937127
JB		 3802 for (i = 0; i < t_bucket->bu_blocks; i++)
3803 memset(bucket_block(t_bucket, i), 0, blocksize);
80bcaf34 3804
ba937127 3805 xh = bucket_xh(t_bucket);
80bcaf34
TM		 3806 xh->xh_free_start = cpu_to_le16(blocksize);
3807 xh->xh_entries[0].xe_name_hash = xe->xe_name_hash;
3808 le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1);
3809
3810 goto set_num_buckets;
3811 }
3812
01225596 3813 /* copy the whole bucket to the new first. */
ba937127 3814 ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket);
01225596
TM		 3815
3816 /* update the new bucket. */
ba937127 3817 xh = bucket_xh(t_bucket);
01225596
TM		 3818
3819 /*
3820 * Calculate the total name/value len and xh_free_start for
3821 * the old bucket first.
3822 */
3823 name_offset = OCFS2_XATTR_BUCKET_SIZE;
3824 name_value_len = 0;
3825 for (i = 0; i < start; i++) {
3826 xe = &xh->xh_entries[i];
3827 xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
3828 if (ocfs2_xattr_is_local(xe))
3829 xe_len +=
3830 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
3831 else
3832 xe_len += OCFS2_XATTR_ROOT_SIZE;
3833 name_value_len += xe_len;
3834 if (le16_to_cpu(xe->xe_name_offset) < name_offset)
3835 name_offset = le16_to_cpu(xe->xe_name_offset);
3836 }
3837
3838 /*
3839 * Now begin the modification to the new bucket.
3840 *
3841 * In the new bucket, We just move the xattr entry to the beginning
3842 * and don't touch the name/value. So there will be some holes in the
3843 * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is
3844 * called.
3845 */
3846 xe = &xh->xh_entries[start];
3847 len = sizeof(struct ocfs2_xattr_entry) * (count - start);
3848 mlog(0, "mv xattr entry len %d from %d to %d\n", len,
ff1ec20e
MF		 3849 (int)((char *)xe - (char *)xh),
3850 (int)((char *)xh->xh_entries - (char *)xh));
01225596
TM		 3851 memmove((char *)xh->xh_entries, (char *)xe, len);
3852 xe = &xh->xh_entries[count - start];
3853 len = sizeof(struct ocfs2_xattr_entry) * start;
3854 memset((char *)xe, 0, len);
3855
3856 le16_add_cpu(&xh->xh_count, -start);
3857 le16_add_cpu(&xh->xh_name_value_len, -name_value_len);
3858
3859 /* Calculate xh_free_start for the new bucket. */
3860 xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
3861 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
3862 xe = &xh->xh_entries[i];
3863 xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len);
3864 if (ocfs2_xattr_is_local(xe))
3865 xe_len +=
3866 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
3867 else
3868 xe_len += OCFS2_XATTR_ROOT_SIZE;
3869 if (le16_to_cpu(xe->xe_name_offset) <
3870 le16_to_cpu(xh->xh_free_start))
3871 xh->xh_free_start = xe->xe_name_offset;
3872 }
3873
80bcaf34 3874set_num_buckets:
01225596
TM		 3875 /* set xh->xh_num_buckets for the new xh. */
3876 if (new_bucket_head)
3877 xh->xh_num_buckets = cpu_to_le16(1);
3878 else
3879 xh->xh_num_buckets = 0;
3880
ba937127 3881 ocfs2_xattr_bucket_journal_dirty(handle, t_bucket);
01225596
TM		 3882
3883 /* store the first_hash of the new bucket. */
3884 if (first_hash)
3885 *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash);
3886
3887 /*
80bcaf34
TM		 3888 * Now only update the 1st block of the old bucket. If we
3889 * just added a new empty bucket, there is no need to modify
3890 * it.
01225596 3891 */
80bcaf34
TM		 3892 if (start == count)
3893 goto out;
3894
ba937127 3895 xh = bucket_xh(s_bucket);
01225596
TM		 3896 memset(&xh->xh_entries[start], 0,
3897 sizeof(struct ocfs2_xattr_entry) * (count - start));
3898 xh->xh_count = cpu_to_le16(start);
3899 xh->xh_free_start = cpu_to_le16(name_offset);
3900 xh->xh_name_value_len = cpu_to_le16(name_value_len);
3901
ba937127 3902 ocfs2_xattr_bucket_journal_dirty(handle, s_bucket);
01225596
TM		 3903
3904out:
ba937127
JB		 3905 ocfs2_xattr_bucket_free(s_bucket);
3906 ocfs2_xattr_bucket_free(t_bucket);
01225596
TM		 3907
3908 return ret;
3909}
3910
3911/*
3912 * Copy xattr from one bucket to another bucket.
3913 *
3914 * The caller must make sure that the journal transaction
3915 * has enough space for journaling.
3916 */
3917static int ocfs2_cp_xattr_bucket(struct inode *inode,
3918 handle_t *handle,
3919 u64 s_blkno,
3920 u64 t_blkno,
3921 int t_is_new)
3922{
4980c6da 3923 int ret;
ba937127 3924 struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL;
01225596
TM		 3925
3926 BUG_ON(s_blkno == t_blkno);
3927
3928 mlog(0, "cp bucket %llu to %llu, target is %d\n",
de29c085
MF		 3929 (unsigned long long)s_blkno, (unsigned long long)t_blkno,
3930 t_is_new);
01225596 3931
ba937127
JB		 3932 s_bucket = ocfs2_xattr_bucket_new(inode);
3933 t_bucket = ocfs2_xattr_bucket_new(inode);
3934 if (!s_bucket || !t_bucket) {
3935 ret = -ENOMEM;
3936 mlog_errno(ret);
3937 goto out;
3938 }
92de109a 3939
ba937127 3940 ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno);
01225596
TM		 3941 if (ret)
3942 goto out;
3943
784b816a
JB		 3944 /*
3945 * Even if !t_is_new, we're overwriting t_bucket. Thus,
3946 * there's no need to read it.
3947 */
ba937127 3948 ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno);
01225596
TM		 3949 if (ret)
3950 goto out;
3951
2b656c1d
JB		 3952 /*
3953 * Hey, if we're overwriting t_bucket, what difference does
3954 * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new
874d65af
JB		 3955 * cluster to fill, we came here from
3956 * ocfs2_mv_xattr_buckets(), and it is really new -
3957 * ACCESS_CREATE is required. But we also might have moved data
3958 * out of t_bucket before extending back into it.
3959 * ocfs2_add_new_xattr_bucket() can do this - its call to
3960 * ocfs2_add_new_xattr_cluster() may have created a new extent
2b656c1d
JB		 3961 * and copied out the end of the old extent. Then it re-extends
3962 * the old extent back to create space for new xattrs. That's
3963 * how we get here, and the bucket isn't really new.
3964 */
ba937127 3965 ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket,
1224be02
JB		 3966 t_is_new ?
3967 OCFS2_JOURNAL_ACCESS_CREATE :
3968 OCFS2_JOURNAL_ACCESS_WRITE);
3969 if (ret)
3970 goto out;
01225596 3971
ba937127
JB		 3972 ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket);
3973 ocfs2_xattr_bucket_journal_dirty(handle, t_bucket);
01225596
TM		 3974
3975out:
ba937127
JB		 3976 ocfs2_xattr_bucket_free(t_bucket);
3977 ocfs2_xattr_bucket_free(s_bucket);
01225596
TM		 3978
3979 return ret;
3980}
3981
3982/*
874d65af
JB		 3983 * src_blk points to the start of an existing extent. last_blk points to
3984 * last cluster in that extent. to_blk points to a newly allocated
54ecb6b6
JB		 3985 * extent. We copy the buckets from the cluster at last_blk to the new
3986 * extent. If start_bucket is non-zero, we skip that many buckets before
3987 * we start copying. The new extent's xh_num_buckets gets set to the
3988 * number of buckets we copied. The old extent's xh_num_buckets shrinks
3989 * by the same amount.
01225596 3990 */
54ecb6b6
JB		 3991static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle,
3992 u64 src_blk, u64 last_blk, u64 to_blk,
3993 unsigned int start_bucket,
3994 u32 *first_hash)
01225596
TM		 3995{
3996 int i, ret, credits;
3997 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
15d60929 3998 int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
01225596 3999 int num_buckets = ocfs2_xattr_buckets_per_cluster(osb);
15d60929 4000 struct ocfs2_xattr_bucket *old_first, *new_first;
01225596 4001
874d65af
JB		 4002 mlog(0, "mv xattrs from cluster %llu to %llu\n",
4003 (unsigned long long)last_blk, (unsigned long long)to_blk);
01225596 4004
54ecb6b6
JB		 4005 BUG_ON(start_bucket >= num_buckets);
4006 if (start_bucket) {
4007 num_buckets -= start_bucket;
4008 last_blk += (start_bucket * blks_per_bucket);
4009 }
4010
15d60929
JB		 4011 /* The first bucket of the original extent */
4012 old_first = ocfs2_xattr_bucket_new(inode);
4013 /* The first bucket of the new extent */
4014 new_first = ocfs2_xattr_bucket_new(inode);
4015 if (!old_first || !new_first) {
4016 ret = -ENOMEM;
4017 mlog_errno(ret);
4018 goto out;
4019 }
4020
874d65af 4021 ret = ocfs2_read_xattr_bucket(old_first, src_blk);
15d60929
JB		 4022 if (ret) {
4023 mlog_errno(ret);
4024 goto out;
4025 }
4026
01225596 4027 /*
54ecb6b6
JB		 4028 * We need to update the first bucket of the old extent and all
4029 * the buckets going to the new extent.
01225596 4030 */
54ecb6b6
JB		 4031 credits = ((num_buckets + 1) * blks_per_bucket) +
4032 handle->h_buffer_credits;
01225596
TM		 4033 ret = ocfs2_extend_trans(handle, credits);
4034 if (ret) {
4035 mlog_errno(ret);
4036 goto out;
4037 }
4038
15d60929
JB		 4039 ret = ocfs2_xattr_bucket_journal_access(handle, old_first,
4040 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4041 if (ret) {
4042 mlog_errno(ret);
4043 goto out;
4044 }
4045
4046 for (i = 0; i < num_buckets; i++) {
4047 ret = ocfs2_cp_xattr_bucket(inode, handle,
874d65af 4048 last_blk + (i * blks_per_bucket),
15d60929
JB		 4049 to_blk + (i * blks_per_bucket),
4050 1);
01225596
TM		 4051 if (ret) {
4052 mlog_errno(ret);
4053 goto out;
4054 }
01225596
TM		 4055 }
4056
15d60929
JB		 4057 /*
4058 * Get the new bucket ready before we dirty anything
4059 * (This actually shouldn't fail, because we already dirtied
4060 * it once in ocfs2_cp_xattr_bucket()).
4061 */
4062 ret = ocfs2_read_xattr_bucket(new_first, to_blk);
4063 if (ret) {
01225596
TM		 4064 mlog_errno(ret);
4065 goto out;
4066 }
15d60929
JB		 4067 ret = ocfs2_xattr_bucket_journal_access(handle, new_first,
4068 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4069 if (ret) {
4070 mlog_errno(ret);
4071 goto out;
4072 }
4073
15d60929
JB		 4074 /* Now update the headers */
4075 le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets);
4076 ocfs2_xattr_bucket_journal_dirty(handle, old_first);
01225596 4077
15d60929
JB		 4078 bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(num_buckets);
4079 ocfs2_xattr_bucket_journal_dirty(handle, new_first);
01225596
TM		 4080
4081 if (first_hash)
15d60929
JB		 4082 *first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash);
4083
01225596 4084out:
15d60929
JB		 4085 ocfs2_xattr_bucket_free(new_first);
4086 ocfs2_xattr_bucket_free(old_first);
01225596
TM		 4087 return ret;
4088}
4089
4090/*
80bcaf34 4091 * Move some xattrs in this cluster to the new cluster.
01225596
TM		 4092 * This function should only be called when bucket size == cluster size.
4093 * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead.
4094 */
80bcaf34
TM		 4095static int ocfs2_divide_xattr_cluster(struct inode *inode,
4096 handle_t *handle,
4097 u64 prev_blk,
4098 u64 new_blk,
4099 u32 *first_hash)
01225596
TM		 4100{
4101 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
85db90e7 4102 int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits;
01225596
TM		 4103
4104 BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize);
4105
4106 ret = ocfs2_extend_trans(handle, credits);
4107 if (ret) {
4108 mlog_errno(ret);
4109 return ret;
4110 }
4111
4112 /* Move half of the xattr in start_blk to the next bucket. */
80bcaf34
TM		 4113 return ocfs2_divide_xattr_bucket(inode, handle, prev_blk,
4114 new_blk, first_hash, 1);
01225596
TM		 4115}
4116
4117/*
4118 * Move some xattrs from the old cluster to the new one since they are not
4119 * contiguous in ocfs2 xattr tree.
4120 *
4121 * new_blk starts a new separate cluster, and we will move some xattrs from
4122 * prev_blk to it. v_start will be set as the first name hash value in this
4123 * new cluster so that it can be used as e_cpos during tree insertion and
4124 * don't collide with our original b-tree operations. first_bh and header_bh
4125 * will also be updated since they will be used in ocfs2_extend_xattr_bucket
4126 * to extend the insert bucket.
4127 *
4128 * The problem is how much xattr should we move to the new one and when should
4129 * we update first_bh and header_bh?
4130 * 1. If cluster size > bucket size, that means the previous cluster has more
4131 * than 1 bucket, so just move half nums of bucket into the new cluster and
4132 * update the first_bh and header_bh if the insert bucket has been moved
4133 * to the new cluster.
4134 * 2. If cluster_size == bucket_size:
4135 * a) If the previous extent rec has more than one cluster and the insert
4136 * place isn't in the last cluster, copy the entire last cluster to the
4137 * new one. This time, we don't need to upate the first_bh and header_bh
4138 * since they will not be moved into the new cluster.
4139 * b) Otherwise, move the bottom half of the xattrs in the last cluster into
4140 * the new one. And we set the extend flag to zero if the insert place is
4141 * moved into the new allocated cluster since no extend is needed.
4142 */
4143static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode,
4144 handle_t *handle,
012ee910
JB		 4145 struct ocfs2_xattr_bucket *first,
4146 struct ocfs2_xattr_bucket *target,
01225596 4147 u64 new_blk,
01225596
TM		 4148 u32 prev_clusters,
4149 u32 *v_start,
4150 int *extend)
4151{
92cf3adf 4152 int ret;
01225596
TM		 4153
4154 mlog(0, "adjust xattrs from cluster %llu len %u to %llu\n",
012ee910 4155 (unsigned long long)bucket_blkno(first), prev_clusters,
de29c085 4156 (unsigned long long)new_blk);
01225596 4157
41cb8148 4158 if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) {
01225596
TM		 4159 ret = ocfs2_mv_xattr_bucket_cross_cluster(inode,
4160 handle,
41cb8148 4161 first, target,
01225596 4162 new_blk,
01225596
TM		 4163 prev_clusters,
4164 v_start);
012ee910 4165 if (ret)
41cb8148 4166 mlog_errno(ret);
41cb8148 4167 } else {
92cf3adf
JB		 4168 /* The start of the last cluster in the first extent */
4169 u64 last_blk = bucket_blkno(first) +
4170 ((prev_clusters - 1) *
4171 ocfs2_clusters_to_blocks(inode->i_sb, 1));
01225596 4172
012ee910 4173 if (prev_clusters > 1 && bucket_blkno(target) != last_blk) {
874d65af 4174 ret = ocfs2_mv_xattr_buckets(inode, handle,
92cf3adf 4175 bucket_blkno(first),
54ecb6b6 4176 last_blk, new_blk, 0,
01225596 4177 v_start);
012ee910
JB		 4178 if (ret)
4179 mlog_errno(ret);
4180 } else {
80bcaf34
TM		 4181 ret = ocfs2_divide_xattr_cluster(inode, handle,
4182 last_blk, new_blk,
4183 v_start);
012ee910
JB		 4184 if (ret)
4185 mlog_errno(ret);
01225596 4186
92cf3adf 4187 if ((bucket_blkno(target) == last_blk) && extend)
01225596
TM		 4188 *extend = 0;
4189 }
4190 }
4191
4192 return ret;
4193}
4194
4195/*
4196 * Add a new cluster for xattr storage.
4197 *
4198 * If the new cluster is contiguous with the previous one, it will be
4199 * appended to the same extent record, and num_clusters will be updated.
4200 * If not, we will insert a new extent for it and move some xattrs in
4201 * the last cluster into the new allocated one.
4202 * We also need to limit the maximum size of a btree leaf, otherwise we'll
4203 * lose the benefits of hashing because we'll have to search large leaves.
4204 * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize,
4205 * if it's bigger).
4206 *
4207 * first_bh is the first block of the previous extent rec and header_bh
4208 * indicates the bucket we will insert the new xattrs. They will be updated
4209 * when the header_bh is moved into the new cluster.
4210 */
4211static int ocfs2_add_new_xattr_cluster(struct inode *inode,
4212 struct buffer_head *root_bh,
ed29c0ca
JB		 4213 struct ocfs2_xattr_bucket *first,
4214 struct ocfs2_xattr_bucket *target,
01225596
TM		 4215 u32 *num_clusters,
4216 u32 prev_cpos,
78f30c31
TM		 4217 int *extend,
4218 struct ocfs2_xattr_set_ctxt *ctxt)
01225596 4219{
85db90e7 4220 int ret;
01225596
TM		 4221 u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1);
4222 u32 prev_clusters = *num_clusters;
4223 u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0;
4224 u64 block;
85db90e7 4225 handle_t *handle = ctxt->handle;
01225596 4226 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
f99b9b7c 4227 struct ocfs2_extent_tree et;
01225596
TM		 4228
4229 mlog(0, "Add new xattr cluster for %llu, previous xattr hash = %u, "
4230 "previous xattr blkno = %llu\n",
4231 (unsigned long long)OCFS2_I(inode)->ip_blkno,
ed29c0ca 4232 prev_cpos, (unsigned long long)bucket_blkno(first));
01225596 4233
8d6220d6 4234 ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh);
f99b9b7c 4235
84008972
JB		 4236 ret = ocfs2_journal_access_xb(handle, inode, root_bh,
4237 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4238 if (ret < 0) {
4239 mlog_errno(ret);
4240 goto leave;
4241 }
4242
78f30c31 4243 ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1,
01225596
TM		 4244 clusters_to_add, &bit_off, &num_bits);
4245 if (ret < 0) {
4246 if (ret != -ENOSPC)
4247 mlog_errno(ret);
4248 goto leave;
4249 }
4250
4251 BUG_ON(num_bits > clusters_to_add);
4252
4253 block = ocfs2_clusters_to_blocks(osb->sb, bit_off);
4254 mlog(0, "Allocating %u clusters at block %u for xattr in inode %llu\n",
4255 num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno);
4256
ed29c0ca 4257 if (bucket_blkno(first) + (prev_clusters * bpc) == block &&
01225596
TM		 4258 (prev_clusters + num_bits) << osb->s_clustersize_bits <=
4259 OCFS2_MAX_XATTR_TREE_LEAF_SIZE) {
4260 /*
4261 * If this cluster is contiguous with the old one and
4262 * adding this new cluster, we don't surpass the limit of
4263 * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be
4264 * initialized and used like other buckets in the previous
4265 * cluster.
4266 * So add it as a contiguous one. The caller will handle
4267 * its init process.
4268 */
4269 v_start = prev_cpos + prev_clusters;
4270 *num_clusters = prev_clusters + num_bits;
4271 mlog(0, "Add contiguous %u clusters to previous extent rec.\n",
4272 num_bits);
4273 } else {
4274 ret = ocfs2_adjust_xattr_cross_cluster(inode,
4275 handle,
012ee910
JB		 4276 first,
4277 target,
01225596 4278 block,
01225596
TM		 4279 prev_clusters,
4280 &v_start,
4281 extend);
4282 if (ret) {
4283 mlog_errno(ret);
4284 goto leave;
4285 }
4286 }
4287
4288 mlog(0, "Insert %u clusters at block %llu for xattr at %u\n",
de29c085 4289 num_bits, (unsigned long long)block, v_start);
f99b9b7c 4290 ret = ocfs2_insert_extent(osb, handle, inode, &et, v_start, block,
78f30c31 4291 num_bits, 0, ctxt->meta_ac);
01225596
TM		 4292 if (ret < 0) {
4293 mlog_errno(ret);
4294 goto leave;
4295 }
4296
4297 ret = ocfs2_journal_dirty(handle, root_bh);
85db90e7 4298 if (ret < 0)
01225596 4299 mlog_errno(ret);
01225596
TM		 4300
4301leave:
01225596
TM		 4302 return ret;
4303}
4304
4305/*
92de109a
JB		 4306 * We are given an extent. 'first' is the bucket at the very front of
4307 * the extent. The extent has space for an additional bucket past
4308 * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number
4309 * of the target bucket. We wish to shift every bucket past the target
4310 * down one, filling in that additional space. When we get back to the
4311 * target, we split the target between itself and the now-empty bucket
4312 * at target+1 (aka, target_blkno + blks_per_bucket).
01225596
TM		 4313 */
4314static int ocfs2_extend_xattr_bucket(struct inode *inode,
85db90e7 4315 handle_t *handle,
92de109a
JB		 4316 struct ocfs2_xattr_bucket *first,
4317 u64 target_blk,
01225596
TM		 4318 u32 num_clusters)
4319{
4320 int ret, credits;
4321 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
4322 u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb);
92de109a
JB		 4323 u64 end_blk;
4324 u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets);
01225596
TM		 4325
4326 mlog(0, "extend xattr bucket in %llu, xattr extend rec starting "
92de109a
JB		 4327 "from %llu, len = %u\n", (unsigned long long)target_blk,
4328 (unsigned long long)bucket_blkno(first), num_clusters);
01225596 4329
92de109a
JB		 4330 /* The extent must have room for an additional bucket */
4331 BUG_ON(new_bucket >=
4332 (num_clusters * ocfs2_xattr_buckets_per_cluster(osb)));
01225596 4333
92de109a
JB		 4334 /* end_blk points to the last existing bucket */
4335 end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket);
01225596
TM		 4336
4337 /*
92de109a
JB		 4338 * end_blk is the start of the last existing bucket.
4339 * Thus, (end_blk - target_blk) covers the target bucket and
4340 * every bucket after it up to, but not including, the last
4341 * existing bucket. Then we add the last existing bucket, the
4342 * new bucket, and the first bucket (3 * blk_per_bucket).
01225596 4343 */
92de109a 4344 credits = (end_blk - target_blk) + (3 * blk_per_bucket) +
85db90e7
TM		 4345 handle->h_buffer_credits;
4346 ret = ocfs2_extend_trans(handle, credits);
4347 if (ret) {
01225596
TM		 4348 mlog_errno(ret);
4349 goto out;
4350 }
4351
92de109a
JB		 4352 ret = ocfs2_xattr_bucket_journal_access(handle, first,
4353 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4354 if (ret) {
4355 mlog_errno(ret);
85db90e7 4356 goto out;
01225596
TM		 4357 }
4358
92de109a 4359 while (end_blk != target_blk) {
01225596
TM		 4360 ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk,
4361 end_blk + blk_per_bucket, 0);
4362 if (ret)
85db90e7 4363 goto out;
01225596
TM		 4364 end_blk -= blk_per_bucket;
4365 }
4366
92de109a
JB		 4367 /* Move half of the xattr in target_blkno to the next bucket. */
4368 ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk,
4369 target_blk + blk_per_bucket, NULL, 0);
01225596 4370
92de109a
JB		 4371 le16_add_cpu(&bucket_xh(first)->xh_num_buckets, 1);
4372 ocfs2_xattr_bucket_journal_dirty(handle, first);
01225596 4373
01225596
TM		 4374out:
4375 return ret;
4376}
4377
4378/*
91f2033f
JB		 4379 * Add new xattr bucket in an extent record and adjust the buckets
4380 * accordingly. xb_bh is the ocfs2_xattr_block, and target is the
4381 * bucket we want to insert into.
01225596 4382 *
91f2033f
JB		 4383 * In the easy case, we will move all the buckets after target down by
4384 * one. Half of target's xattrs will be moved to the next bucket.
4385 *
4386 * If current cluster is full, we'll allocate a new one. This may not
4387 * be contiguous. The underlying calls will make sure that there is
4388 * space for the insert, shifting buckets around if necessary.
4389 * 'target' may be moved by those calls.
01225596
TM		 4390 */
4391static int ocfs2_add_new_xattr_bucket(struct inode *inode,
4392 struct buffer_head *xb_bh,
91f2033f 4393 struct ocfs2_xattr_bucket *target,
78f30c31 4394 struct ocfs2_xattr_set_ctxt *ctxt)
01225596 4395{
01225596
TM		 4396 struct ocfs2_xattr_block *xb =
4397 (struct ocfs2_xattr_block *)xb_bh->b_data;
4398 struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root;
4399 struct ocfs2_extent_list *el = &xb_root->xt_list;
91f2033f
JB		 4400 u32 name_hash =
4401 le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash);
ed29c0ca 4402 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
01225596
TM		 4403 int ret, num_buckets, extend = 1;
4404 u64 p_blkno;
4405 u32 e_cpos, num_clusters;
92de109a 4406 /* The bucket at the front of the extent */
91f2033f 4407 struct ocfs2_xattr_bucket *first;
01225596 4408
91f2033f
JB		 4409 mlog(0, "Add new xattr bucket starting from %llu\n",
4410 (unsigned long long)bucket_blkno(target));
01225596 4411
ed29c0ca 4412 /* The first bucket of the original extent */
92de109a 4413 first = ocfs2_xattr_bucket_new(inode);
91f2033f 4414 if (!first) {
92de109a
JB		 4415 ret = -ENOMEM;
4416 mlog_errno(ret);
4417 goto out;
4418 }
4419
01225596
TM		 4420 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos,
4421 &num_clusters, el);
4422 if (ret) {
4423 mlog_errno(ret);
4424 goto out;
4425 }
4426
ed29c0ca 4427 ret = ocfs2_read_xattr_bucket(first, p_blkno);
01225596
TM		 4428 if (ret) {
4429 mlog_errno(ret);
4430 goto out;
4431 }
4432
ed29c0ca
JB		 4433 num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters;
4434 if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) {
4435 /*
4436 * This can move first+target if the target bucket moves
4437 * to the new extent.
4438 */
01225596
TM		 4439 ret = ocfs2_add_new_xattr_cluster(inode,
4440 xb_bh,
ed29c0ca
JB		 4441 first,
4442 target,
01225596
TM		 4443 &num_clusters,
4444 e_cpos,
78f30c31
TM		 4445 &extend,
4446 ctxt);
01225596
TM		 4447 if (ret) {
4448 mlog_errno(ret);
4449 goto out;
4450 }
4451 }
4452
92de109a 4453 if (extend) {
01225596 4454 ret = ocfs2_extend_xattr_bucket(inode,
85db90e7 4455 ctxt->handle,
ed29c0ca
JB		 4456 first,
4457 bucket_blkno(target),
01225596 4458 num_clusters);
92de109a
JB		 4459 if (ret)
4460 mlog_errno(ret);
4461 }
4462
01225596 4463out:
92de109a 4464 ocfs2_xattr_bucket_free(first);
ed29c0ca 4465
01225596
TM		 4466 return ret;
4467}
4468
4469static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode,
4470 struct ocfs2_xattr_bucket *bucket,
4471 int offs)
4472{
4473 int block_off = offs >> inode->i_sb->s_blocksize_bits;
4474
4475 offs = offs % inode->i_sb->s_blocksize;
51def39f 4476 return bucket_block(bucket, block_off) + offs;
01225596
TM		 4477}
4478
4479/*
4480 * Handle the normal xattr set, including replace, delete and new.
01225596
TM		 4481 *
4482 * Note: "local" indicates the real data's locality. So we can't
4483 * just its bucket locality by its length.
4484 */
4485static void ocfs2_xattr_set_entry_normal(struct inode *inode,
4486 struct ocfs2_xattr_info *xi,
4487 struct ocfs2_xattr_search *xs,
4488 u32 name_hash,
5a095611 4489 int local)
01225596
TM		 4490{
4491 struct ocfs2_xattr_entry *last, *xe;
4492 int name_len = strlen(xi->name);
4493 struct ocfs2_xattr_header *xh = xs->header;
4494 u16 count = le16_to_cpu(xh->xh_count), start;
4495 size_t blocksize = inode->i_sb->s_blocksize;
4496 char *val;
4497 size_t offs, size, new_size;
4498
4499 last = &xh->xh_entries[count];
4500 if (!xs->not_found) {
4501 xe = xs->here;
4502 offs = le16_to_cpu(xe->xe_name_offset);
4503 if (ocfs2_xattr_is_local(xe))
4504 size = OCFS2_XATTR_SIZE(name_len) +
4505 OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
4506 else
4507 size = OCFS2_XATTR_SIZE(name_len) +
4508 OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
4509
4510 /*
4511 * If the new value will be stored outside, xi->value has been
4512 * initalized as an empty ocfs2_xattr_value_root, and the same
4513 * goes with xi->value_len, so we can set new_size safely here.
4514 * See ocfs2_xattr_set_in_bucket.
4515 */
4516 new_size = OCFS2_XATTR_SIZE(name_len) +
4517 OCFS2_XATTR_SIZE(xi->value_len);
4518
4519 le16_add_cpu(&xh->xh_name_value_len, -size);
4520 if (xi->value) {
4521 if (new_size > size)
4522 goto set_new_name_value;
4523
4524 /* Now replace the old value with new one. */
4525 if (local)
4526 xe->xe_value_size = cpu_to_le64(xi->value_len);
4527 else
4528 xe->xe_value_size = 0;
4529
4530 val = ocfs2_xattr_bucket_get_val(inode,
ba937127 4531 xs->bucket, offs);
01225596
TM		 4532 memset(val + OCFS2_XATTR_SIZE(name_len), 0,
4533 size - OCFS2_XATTR_SIZE(name_len));
4534 if (OCFS2_XATTR_SIZE(xi->value_len) > 0)
4535 memcpy(val + OCFS2_XATTR_SIZE(name_len),
4536 xi->value, xi->value_len);
4537
4538 le16_add_cpu(&xh->xh_name_value_len, new_size);
4539 ocfs2_xattr_set_local(xe, local);
4540 return;
4541 } else {
5a095611
TM		 4542 /*
4543 * Remove the old entry if there is more than one.
4544 * We don't remove the last entry so that we can
4545 * use it to indicate the hash value of the empty
4546 * bucket.
4547 */
01225596 4548 last -= 1;
01225596 4549 le16_add_cpu(&xh->xh_count, -1);
5a095611
TM		 4550 if (xh->xh_count) {
4551 memmove(xe, xe + 1,
4552 (void *)last - (void *)xe);
4553 memset(last, 0,
4554 sizeof(struct ocfs2_xattr_entry));
4555 } else
4556 xh->xh_free_start =
4557 cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE);
4558
01225596
TM		 4559 return;
4560 }
4561 } else {
4562 /* find a new entry for insert. */
4563 int low = 0, high = count - 1, tmp;
4564 struct ocfs2_xattr_entry *tmp_xe;
4565
5a095611 4566 while (low <= high && count) {
01225596
TM		 4567 tmp = (low + high) / 2;
4568 tmp_xe = &xh->xh_entries[tmp];
4569
4570 if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash))
4571 low = tmp + 1;
4572 else if (name_hash <
4573 le32_to_cpu(tmp_xe->xe_name_hash))
4574 high = tmp - 1;
06b240d8
TM		 4575 else {
4576 low = tmp;
01225596 4577 break;
06b240d8 4578 }
01225596
TM		 4579 }
4580
4581 xe = &xh->xh_entries[low];
4582 if (low != count)
4583 memmove(xe + 1, xe, (void *)last - (void *)xe);
4584
4585 le16_add_cpu(&xh->xh_count, 1);
4586 memset(xe, 0, sizeof(struct ocfs2_xattr_entry));
4587 xe->xe_name_hash = cpu_to_le32(name_hash);
4588 xe->xe_name_len = name_len;
4589 ocfs2_xattr_set_type(xe, xi->name_index);
4590 }
4591
4592set_new_name_value:
4593 /* Insert the new name+value. */
4594 size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(xi->value_len);
4595
4596 /*
4597 * We must make sure that the name/value pair
4598 * exists in the same block.
4599 */
4600 offs = le16_to_cpu(xh->xh_free_start);
4601 start = offs - size;
4602
4603 if (start >> inode->i_sb->s_blocksize_bits !=
4604 (offs - 1) >> inode->i_sb->s_blocksize_bits) {
4605 offs = offs - offs % blocksize;
4606 xh->xh_free_start = cpu_to_le16(offs);
4607 }
4608
ba937127 4609 val = ocfs2_xattr_bucket_get_val(inode, xs->bucket, offs - size);
01225596
TM		 4610 xe->xe_name_offset = cpu_to_le16(offs - size);
4611
4612 memset(val, 0, size);
4613 memcpy(val, xi->name, name_len);
4614 memcpy(val + OCFS2_XATTR_SIZE(name_len), xi->value, xi->value_len);
4615
4616 xe->xe_value_size = cpu_to_le64(xi->value_len);
4617 ocfs2_xattr_set_local(xe, local);
4618 xs->here = xe;
4619 le16_add_cpu(&xh->xh_free_start, -size);
4620 le16_add_cpu(&xh->xh_name_value_len, size);
4621
4622 return;
4623}
4624
01225596
TM		 4625/*
4626 * Set the xattr entry in the specified bucket.
4627 * The bucket is indicated by xs->bucket and it should have the enough
4628 * space for the xattr insertion.
4629 */
4630static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode,
85db90e7 4631 handle_t *handle,
01225596
TM		 4632 struct ocfs2_xattr_info *xi,
4633 struct ocfs2_xattr_search *xs,
4634 u32 name_hash,
5a095611 4635 int local)
01225596 4636{
1224be02 4637 int ret;
02dbf38d 4638 u64 blkno;
01225596 4639
ff1ec20e
MF		 4640 mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n",
4641 (unsigned long)xi->value_len, xi->name_index,
ba937127 4642 (unsigned long long)bucket_blkno(xs->bucket));
01225596 4643
ba937127 4644 if (!xs->bucket->bu_bhs[1]) {
02dbf38d
JB		 4645 blkno = bucket_blkno(xs->bucket);
4646 ocfs2_xattr_bucket_relse(xs->bucket);
4647 ret = ocfs2_read_xattr_bucket(xs->bucket, blkno);
01225596
TM		 4648 if (ret) {
4649 mlog_errno(ret);
4650 goto out;
4651 }
4652 }
4653
ba937127 4654 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
1224be02
JB		 4655 OCFS2_JOURNAL_ACCESS_WRITE);
4656 if (ret < 0) {
4657 mlog_errno(ret);
4658 goto out;
01225596
TM		 4659 }
4660
5a095611 4661 ocfs2_xattr_set_entry_normal(inode, xi, xs, name_hash, local);
ba937127 4662 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
01225596 4663
01225596 4664out:
01225596
TM		 4665 return ret;
4666}
4667
01225596
TM		 4668/*
4669 * Truncate the specified xe_off entry in xattr bucket.
4670 * bucket is indicated by header_bh and len is the new length.
4671 * Both the ocfs2_xattr_value_root and the entry will be updated here.
4672 *
4673 * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed.
4674 */
4675static int ocfs2_xattr_bucket_value_truncate(struct inode *inode,
548b0f22 4676 struct ocfs2_xattr_bucket *bucket,
01225596 4677 int xe_off,
78f30c31
TM		 4678 int len,
4679 struct ocfs2_xattr_set_ctxt *ctxt)
01225596
TM		 4680{
4681 int ret, offset;
4682 u64 value_blk;
01225596 4683 struct ocfs2_xattr_entry *xe;
548b0f22 4684 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
01225596 4685 size_t blocksize = inode->i_sb->s_blocksize;
b3e5d379
JB		 4686 struct ocfs2_xattr_value_buf vb = {
4687 .vb_access = ocfs2_journal_access,
4688 };
01225596
TM		 4689
4690 xe = &xh->xh_entries[xe_off];
4691
4692 BUG_ON(!xe || ocfs2_xattr_is_local(xe));
4693
4694 offset = le16_to_cpu(xe->xe_name_offset) +
4695 OCFS2_XATTR_SIZE(xe->xe_name_len);
4696
4697 value_blk = offset / blocksize;
4698
4699 /* We don't allow ocfs2_xattr_value to be stored in different block. */
4700 BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize);
01225596 4701
b3e5d379
JB		 4702 vb.vb_bh = bucket->bu_bhs[value_blk];
4703 BUG_ON(!vb.vb_bh);
01225596 4704
b3e5d379
JB		 4705 vb.vb_xv = (struct ocfs2_xattr_value_root *)
4706 (vb.vb_bh->b_data + offset % blocksize);
01225596 4707
548b0f22
JB		 4708 ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket,
4709 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4710 if (ret) {
4711 mlog_errno(ret);
4712 goto out;
4713 }
4714
548b0f22
JB		 4715 /*
4716 * From here on out we have to dirty the bucket. The generic
4717 * value calls only modify one of the bucket's bhs, but we need
4718 * to send the bucket at once. So if they error, they *could* have
4719 * modified something. We have to assume they did, and dirty
4720 * the whole bucket. This leaves us in a consistent state.
4721 */
4722 mlog(0, "truncate %u in xattr bucket %llu to %d bytes.\n",
4723 xe_off, (unsigned long long)bucket_blkno(bucket), len);
b3e5d379 4724 ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt);
01225596
TM		 4725 if (ret) {
4726 mlog_errno(ret);
548b0f22 4727 goto out_dirty;
01225596
TM		 4728 }
4729
548b0f22
JB		 4730 xe->xe_value_size = cpu_to_le64(len);
4731
4732out_dirty:
4733 ocfs2_xattr_bucket_journal_dirty(ctxt->handle, bucket);
4734
01225596 4735out:
01225596
TM		 4736 return ret;
4737}
4738
4739static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode,
78f30c31
TM		 4740 struct ocfs2_xattr_search *xs,
4741 int len,
4742 struct ocfs2_xattr_set_ctxt *ctxt)
01225596
TM		 4743{
4744 int ret, offset;
4745 struct ocfs2_xattr_entry *xe = xs->here;
4746 struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base;
4747
ba937127 4748 BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe));
01225596
TM		 4749
4750 offset = xe - xh->xh_entries;
548b0f22 4751 ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket,
78f30c31 4752 offset, len, ctxt);
01225596
TM		 4753 if (ret)
4754 mlog_errno(ret);
4755
4756 return ret;
4757}
4758
4759static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode,
85db90e7 4760 handle_t *handle,
01225596
TM		 4761 struct ocfs2_xattr_search *xs,
4762 char *val,
4763 int value_len)
4764{
4765 int offset;
4766 struct ocfs2_xattr_value_root *xv;
4767 struct ocfs2_xattr_entry *xe = xs->here;
4768
4769 BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe));
4770
4771 offset = le16_to_cpu(xe->xe_name_offset) +
4772 OCFS2_XATTR_SIZE(xe->xe_name_len);
4773
4774 xv = (struct ocfs2_xattr_value_root *)(xs->base + offset);
4775
85db90e7
TM		 4776 return __ocfs2_xattr_set_value_outside(inode, handle,
4777 xv, val, value_len);
01225596
TM		 4778}
4779
01225596
TM		 4780static int ocfs2_rm_xattr_cluster(struct inode *inode,
4781 struct buffer_head *root_bh,
4782 u64 blkno,
4783 u32 cpos,
4784 u32 len)
4785{
4786 int ret;
4787 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
4788 struct inode *tl_inode = osb->osb_tl_inode;
4789 handle_t *handle;
4790 struct ocfs2_xattr_block *xb =
4791 (struct ocfs2_xattr_block *)root_bh->b_data;
01225596
TM		 4792 struct ocfs2_alloc_context *meta_ac = NULL;
4793 struct ocfs2_cached_dealloc_ctxt dealloc;
f99b9b7c
JB		 4794 struct ocfs2_extent_tree et;
4795
8d6220d6 4796 ocfs2_init_xattr_tree_extent_tree(&et, inode, root_bh);
01225596
TM		 4797
4798 ocfs2_init_dealloc_ctxt(&dealloc);
4799
4800 mlog(0, "rm xattr extent rec at %u len = %u, start from %llu\n",
4801 cpos, len, (unsigned long long)blkno);
4802
4803 ocfs2_remove_xattr_clusters_from_cache(inode, blkno, len);
4804
f99b9b7c 4805 ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac);
01225596
TM		 4806 if (ret) {
4807 mlog_errno(ret);
4808 return ret;
4809 }
4810
4811 mutex_lock(&tl_inode->i_mutex);
4812
4813 if (ocfs2_truncate_log_needs_flush(osb)) {
4814 ret = __ocfs2_flush_truncate_log(osb);
4815 if (ret < 0) {
4816 mlog_errno(ret);
4817 goto out;
4818 }
4819 }
4820
a90714c1 4821 handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb));
d3264799 4822 if (IS_ERR(handle)) {
01225596
TM		 4823 ret = -ENOMEM;
4824 mlog_errno(ret);
4825 goto out;
4826 }
4827
84008972
JB		 4828 ret = ocfs2_journal_access_xb(handle, inode, root_bh,
4829 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4830 if (ret) {
4831 mlog_errno(ret);
4832 goto out_commit;
4833 }
4834
f99b9b7c
JB		 4835 ret = ocfs2_remove_extent(inode, &et, cpos, len, handle, meta_ac,
4836 &dealloc);
01225596
TM		 4837 if (ret) {
4838 mlog_errno(ret);
4839 goto out_commit;
4840 }
4841
4842 le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len);
4843
4844 ret = ocfs2_journal_dirty(handle, root_bh);
4845 if (ret) {
4846 mlog_errno(ret);
4847 goto out_commit;
4848 }
4849
4850 ret = ocfs2_truncate_log_append(osb, handle, blkno, len);
4851 if (ret)
4852 mlog_errno(ret);
4853
4854out_commit:
4855 ocfs2_commit_trans(osb, handle);
4856out:
4857 ocfs2_schedule_truncate_log_flush(osb, 1);
4858
4859 mutex_unlock(&tl_inode->i_mutex);
4860
4861 if (meta_ac)
4862 ocfs2_free_alloc_context(meta_ac);
4863
4864 ocfs2_run_deallocs(osb, &dealloc);
4865
4866 return ret;
4867}
4868
01225596 4869static void ocfs2_xattr_bucket_remove_xs(struct inode *inode,
85db90e7 4870 handle_t *handle,
01225596
TM		 4871 struct ocfs2_xattr_search *xs)
4872{
ba937127 4873 struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket);
01225596
TM		 4874 struct ocfs2_xattr_entry *last = &xh->xh_entries[
4875 le16_to_cpu(xh->xh_count) - 1];
4876 int ret = 0;
4877
ba937127 4878 ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket,
1224be02 4879 OCFS2_JOURNAL_ACCESS_WRITE);
01225596
TM		 4880 if (ret) {
4881 mlog_errno(ret);
85db90e7 4882 return;
01225596
TM		 4883 }
4884
4885 /* Remove the old entry. */
4886 memmove(xs->here, xs->here + 1,
4887 (void *)last - (void *)xs->here);
4888 memset(last, 0, sizeof(struct ocfs2_xattr_entry));
4889 le16_add_cpu(&xh->xh_count, -1);
4890
ba937127 4891 ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket);
01225596
TM		 4892}
4893
4894/*
4895 * Set the xattr name/value in the bucket specified in xs.
4896 *
4897 * As the new value in xi may be stored in the bucket or in an outside cluster,
4898 * we divide the whole process into 3 steps:
4899 * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket)
4900 * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs)
4901 * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside)
4902 * 4. If the clusters for the new outside value can't be allocated, we need
4903 * to free the xattr we allocated in set.
4904 */
4905static int ocfs2_xattr_set_in_bucket(struct inode *inode,
4906 struct ocfs2_xattr_info *xi,
78f30c31
TM		 4907 struct ocfs2_xattr_search *xs,
4908 struct ocfs2_xattr_set_ctxt *ctxt)
01225596 4909{
5a095611 4910 int ret, local = 1;
01225596
TM		 4911 size_t value_len;
4912 char *val = (char *)xi->value;
4913 struct ocfs2_xattr_entry *xe = xs->here;
2057e5c6
TM		 4914 u32 name_hash = ocfs2_xattr_name_hash(inode, xi->name,
4915 strlen(xi->name));
01225596
TM		 4916
4917 if (!xs->not_found && !ocfs2_xattr_is_local(xe)) {
4918 /*
4919 * We need to truncate the xattr storage first.
4920 *
4921 * If both the old and new value are stored to
4922 * outside block, we only need to truncate
4923 * the storage and then set the value outside.
4924 *
4925 * If the new value should be stored within block,
4926 * we should free all the outside block first and
4927 * the modification to the xattr block will be done
4928 * by following steps.
4929 */
4930 if (xi->value_len > OCFS2_XATTR_INLINE_SIZE)
4931 value_len = xi->value_len;
4932 else
4933 value_len = 0;
4934
4935 ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
78f30c31
TM		 4936 value_len,
4937 ctxt);
01225596
TM		 4938 if (ret)
4939 goto out;
4940
4941 if (value_len)
4942 goto set_value_outside;
4943 }
4944
4945 value_len = xi->value_len;
4946 /* So we have to handle the inside block change now. */
4947 if (value_len > OCFS2_XATTR_INLINE_SIZE) {
4948 /*
4949 * If the new value will be stored outside of block,
4950 * initalize a new empty value root and insert it first.
4951 */
4952 local = 0;
4953 xi->value = &def_xv;
4954 xi->value_len = OCFS2_XATTR_ROOT_SIZE;
4955 }
4956
85db90e7
TM		 4957 ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs,
4958 name_hash, local);
01225596
TM		 4959 if (ret) {
4960 mlog_errno(ret);
4961 goto out;
4962 }
4963
5a095611
TM		 4964 if (value_len <= OCFS2_XATTR_INLINE_SIZE)
4965 goto out;
01225596 4966
5a095611
TM		 4967 /* allocate the space now for the outside block storage. */
4968 ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs,
78f30c31 4969 value_len, ctxt);
5a095611
TM		 4970 if (ret) {
4971 mlog_errno(ret);
4972
4973 if (xs->not_found) {
4974 /*
4975 * We can't allocate enough clusters for outside
4976 * storage and we have allocated xattr already,
4977 * so need to remove it.
4978 */
85db90e7 4979 ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs);
01225596 4980 }
01225596
TM		 4981 goto out;
4982 }
4983
4984set_value_outside:
85db90e7
TM		 4985 ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle,
4986 xs, val, value_len);
01225596
TM		 4987out:
4988 return ret;
4989}
4990
80bcaf34
TM		 4991/*
4992 * check whether the xattr bucket is filled up with the same hash value.
4993 * If we want to insert the xattr with the same hash, return -ENOSPC.
4994 * If we want to insert a xattr with different hash value, go ahead
4995 * and ocfs2_divide_xattr_bucket will handle this.
4996 */
01225596 4997static int ocfs2_check_xattr_bucket_collision(struct inode *inode,
80bcaf34
TM		 4998 struct ocfs2_xattr_bucket *bucket,
4999 const char *name)
01225596 5000{
3e632946 5001 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
80bcaf34
TM		 5002 u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name));
5003
5004 if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash))
5005 return 0;
01225596
TM		 5006
5007 if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash ==
5008 xh->xh_entries[0].xe_name_hash) {
5009 mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, "
5010 "hash = %u\n",
9c7759aa 5011 (unsigned long long)bucket_blkno(bucket),
01225596
TM		 5012 le32_to_cpu(xh->xh_entries[0].xe_name_hash));
5013 return -ENOSPC;
5014 }
5015
5016 return 0;
5017}
5018
5019static int ocfs2_xattr_set_entry_index_block(struct inode *inode,
5020 struct ocfs2_xattr_info *xi,
78f30c31
TM		 5021 struct ocfs2_xattr_search *xs,
5022 struct ocfs2_xattr_set_ctxt *ctxt)
01225596
TM		 5023{
5024 struct ocfs2_xattr_header *xh;
5025 struct ocfs2_xattr_entry *xe;
5026 u16 count, header_size, xh_free_start;
6dde41d9 5027 int free, max_free, need, old;
01225596
TM		 5028 size_t value_size = 0, name_len = strlen(xi->name);
5029 size_t blocksize = inode->i_sb->s_blocksize;
5030 int ret, allocation = 0;
01225596
TM		 5031
5032 mlog_entry("Set xattr %s in xattr index block\n", xi->name);
5033
5034try_again:
5035 xh = xs->header;
5036 count = le16_to_cpu(xh->xh_count);
5037 xh_free_start = le16_to_cpu(xh->xh_free_start);
5038 header_size = sizeof(struct ocfs2_xattr_header) +
5039 count * sizeof(struct ocfs2_xattr_entry);
5040 max_free = OCFS2_XATTR_BUCKET_SIZE -
5041 le16_to_cpu(xh->xh_name_value_len) - header_size;
5042
5043 mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size "
5044 "of %u which exceed block size\n",
ba937127 5045 (unsigned long long)bucket_blkno(xs->bucket),
01225596
TM		 5046 header_size);
5047
5048 if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE)
5049 value_size = OCFS2_XATTR_ROOT_SIZE;
5050 else if (xi->value)
5051 value_size = OCFS2_XATTR_SIZE(xi->value_len);
5052
5053 if (xs->not_found)
5054 need = sizeof(struct ocfs2_xattr_entry) +
5055 OCFS2_XATTR_SIZE(name_len) + value_size;
5056 else {
5057 need = value_size + OCFS2_XATTR_SIZE(name_len);
5058
5059 /*
5060 * We only replace the old value if the new length is smaller
5061 * than the old one. Otherwise we will allocate new space in the
5062 * bucket to store it.
5063 */
5064 xe = xs->here;
5065 if (ocfs2_xattr_is_local(xe))
5066 old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size));
5067 else
5068 old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE);
5069
5070 if (old >= value_size)
5071 need = 0;
5072 }
5073
5074 free = xh_free_start - header_size;
5075 /*
5076 * We need to make sure the new name/value pair
5077 * can exist in the same block.
5078 */
5079 if (xh_free_start % blocksize < need)
5080 free -= xh_free_start % blocksize;
5081
5082 mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, "
5083 "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len ="
5084 " %u\n", xs->not_found,
ba937127 5085 (unsigned long long)bucket_blkno(xs->bucket),
01225596
TM		 5086 free, need, max_free, le16_to_cpu(xh->xh_free_start),
5087 le16_to_cpu(xh->xh_name_value_len));
5088
976331d8
TM		 5089 if (free < need ||
5090 (xs->not_found &&
5091 count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) {
01225596
TM		 5092 if (need <= max_free &&
5093 count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) {
5094 /*
5095 * We can create the space by defragment. Since only the
5096 * name/value will be moved, the xe shouldn't be changed
5097 * in xs.
5098 */
85db90e7
TM		 5099 ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle,
5100 xs->bucket);
01225596
TM		 5101 if (ret) {
5102 mlog_errno(ret);
5103 goto out;
5104 }
5105
5106 xh_free_start = le16_to_cpu(xh->xh_free_start);
5107 free = xh_free_start - header_size;
5108 if (xh_free_start % blocksize < need)
5109 free -= xh_free_start % blocksize;
5110
5111 if (free >= need)
5112 goto xattr_set;
5113
5114 mlog(0, "Can't get enough space for xattr insert by "
5115 "defragment. Need %u bytes, but we have %d, so "
5116 "allocate new bucket for it.\n", need, free);
5117 }
5118
5119 /*
5120 * We have to add new buckets or clusters and one
5121 * allocation should leave us enough space for insert.
5122 */
5123 BUG_ON(allocation);
5124
5125 /*
5126 * We do not allow for overlapping ranges between buckets. And
5127 * the maximum number of collisions we will allow for then is
5128 * one bucket's worth, so check it here whether we need to
5129 * add a new bucket for the insert.
5130 */
80bcaf34 5131 ret = ocfs2_check_xattr_bucket_collision(inode,
ba937127 5132 xs->bucket,
80bcaf34 5133 xi->name);
01225596
TM		 5134 if (ret) {
5135 mlog_errno(ret);
5136 goto out;
5137 }
5138
5139 ret = ocfs2_add_new_xattr_bucket(inode,
5140 xs->xattr_bh,
91f2033f 5141 xs->bucket,
78f30c31 5142 ctxt);
01225596
TM		 5143 if (ret) {
5144 mlog_errno(ret);
5145 goto out;
5146 }
5147
91f2033f
JB		 5148 /*
5149 * ocfs2_add_new_xattr_bucket() will have updated
5150 * xs->bucket if it moved, but it will not have updated
5151 * any of the other search fields. Thus, we drop it and
5152 * re-search. Everything should be cached, so it'll be
5153 * quick.
5154 */
ba937127 5155 ocfs2_xattr_bucket_relse(xs->bucket);
01225596
TM		 5156 ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh,
5157 xi->name_index,
5158 xi->name, xs);
5159 if (ret && ret != -ENODATA)
5160 goto out;
5161 xs->not_found = ret;
5162 allocation = 1;
5163 goto try_again;
5164 }
5165
5166xattr_set:
78f30c31 5167 ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt);
01225596
TM		 5168out:
5169 mlog_exit(ret);
5170 return ret;
5171}
a3944256
TM		 5172
5173static int ocfs2_delete_xattr_in_bucket(struct inode *inode,
5174 struct ocfs2_xattr_bucket *bucket,
5175 void *para)
5176{
5177 int ret = 0;
3e632946 5178 struct ocfs2_xattr_header *xh = bucket_xh(bucket);
a3944256
TM		 5179 u16 i;
5180 struct ocfs2_xattr_entry *xe;
78f30c31
TM		 5181 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5182 struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,};
548b0f22
JB		 5183 int credits = ocfs2_remove_extent_credits(osb->sb) +
5184 ocfs2_blocks_per_xattr_bucket(inode->i_sb);
5185
78f30c31
TM		 5186
5187 ocfs2_init_dealloc_ctxt(&ctxt.dealloc);
a3944256
TM		 5188
5189 for (i = 0; i < le16_to_cpu(xh->xh_count); i++) {
5190 xe = &xh->xh_entries[i];
5191 if (ocfs2_xattr_is_local(xe))
5192 continue;
5193
88c3b062
TM		 5194 ctxt.handle = ocfs2_start_trans(osb, credits);
5195 if (IS_ERR(ctxt.handle)) {
5196 ret = PTR_ERR(ctxt.handle);
5197 mlog_errno(ret);
5198 break;
5199 }
5200
548b0f22 5201 ret = ocfs2_xattr_bucket_value_truncate(inode, bucket,
78f30c31 5202 i, 0, &ctxt);
88c3b062
TM		 5203
5204 ocfs2_commit_trans(osb, ctxt.handle);
a3944256
TM		 5205 if (ret) {
5206 mlog_errno(ret);
5207 break;
5208 }
5209 }
5210
78f30c31
TM		 5211 ocfs2_schedule_truncate_log_flush(osb, 1);
5212 ocfs2_run_deallocs(osb, &ctxt.dealloc);
a3944256
TM		 5213 return ret;
5214}
5215
5216static int ocfs2_delete_xattr_index_block(struct inode *inode,
5217 struct buffer_head *xb_bh)
5218{
5219 struct ocfs2_xattr_block *xb =
5220 (struct ocfs2_xattr_block *)xb_bh->b_data;
5221 struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list;
5222 int ret = 0;
5223 u32 name_hash = UINT_MAX, e_cpos, num_clusters;
5224 u64 p_blkno;
5225
5226 if (le16_to_cpu(el->l_next_free_rec) == 0)
5227 return 0;
5228
5229 while (name_hash > 0) {
5230 ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno,
5231 &e_cpos, &num_clusters, el);
5232 if (ret) {
5233 mlog_errno(ret);
5234 goto out;
5235 }
5236
5237 ret = ocfs2_iterate_xattr_buckets(inode, p_blkno, num_clusters,
5238 ocfs2_delete_xattr_in_bucket,
5239 NULL);
5240 if (ret) {
5241 mlog_errno(ret);
5242 goto out;
5243 }
5244
5245 ret = ocfs2_rm_xattr_cluster(inode, xb_bh,
5246 p_blkno, e_cpos, num_clusters);
5247 if (ret) {
5248 mlog_errno(ret);
5249 break;
5250 }
5251
5252 if (e_cpos == 0)
5253 break;
5254
5255 name_hash = e_cpos - 1;
5256 }
5257
5258out:
5259 return ret;
5260}
99219aea 5261
923f7f31
TY		 5262/*
5263 * 'security' attributes support
5264 */
5265static size_t ocfs2_xattr_security_list(struct inode *inode, char *list,
5266 size_t list_size, const char *name,
5267 size_t name_len)
5268{
5269 const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN;
5270 const size_t total_len = prefix_len + name_len + 1;
5271
5272 if (list && total_len <= list_size) {
5273 memcpy(list, XATTR_SECURITY_PREFIX, prefix_len);
5274 memcpy(list + prefix_len, name, name_len);
5275 list[prefix_len + name_len] = '\0';
5276 }
5277 return total_len;
5278}
5279
5280static int ocfs2_xattr_security_get(struct inode *inode, const char *name,
5281 void *buffer, size_t size)
5282{
5283 if (strcmp(name, "") == 0)
5284 return -EINVAL;
5285 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_SECURITY, name,
5286 buffer, size);
5287}
5288
5289static int ocfs2_xattr_security_set(struct inode *inode, const char *name,
5290 const void *value, size_t size, int flags)
5291{
5292 if (strcmp(name, "") == 0)
5293 return -EINVAL;
5294
5295 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, name, value,
5296 size, flags);
5297}
5298
534eaddd
TY		 5299int ocfs2_init_security_get(struct inode *inode,
5300 struct inode *dir,
5301 struct ocfs2_security_xattr_info *si)
5302{
5303 return security_inode_init_security(inode, dir, &si->name, &si->value,
5304 &si->value_len);
5305}
5306
5307int ocfs2_init_security_set(handle_t *handle,
5308 struct inode *inode,
5309 struct buffer_head *di_bh,
5310 struct ocfs2_security_xattr_info *si,
5311 struct ocfs2_alloc_context *xattr_ac,
5312 struct ocfs2_alloc_context *data_ac)
5313{
5314 return ocfs2_xattr_set_handle(handle, inode, di_bh,
5315 OCFS2_XATTR_INDEX_SECURITY,
5316 si->name, si->value, si->value_len, 0,
5317 xattr_ac, data_ac);
5318}
5319
923f7f31
TY		 5320struct xattr_handler ocfs2_xattr_security_handler = {
5321 .prefix = XATTR_SECURITY_PREFIX,
5322 .list = ocfs2_xattr_security_list,
5323 .get = ocfs2_xattr_security_get,
5324 .set = ocfs2_xattr_security_set,
5325};
5326
99219aea
MF		 5327/*
5328 * 'trusted' attributes support
5329 */
99219aea
MF		 5330static size_t ocfs2_xattr_trusted_list(struct inode *inode, char *list,
5331 size_t list_size, const char *name,
5332 size_t name_len)
5333{
ceb1eba3 5334 const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN;
99219aea
MF		 5335 const size_t total_len = prefix_len + name_len + 1;
5336
5337 if (list && total_len <= list_size) {
5338 memcpy(list, XATTR_TRUSTED_PREFIX, prefix_len);
5339 memcpy(list + prefix_len, name, name_len);
5340 list[prefix_len + name_len] = '\0';
5341 }
5342 return total_len;
5343}
5344
5345static int ocfs2_xattr_trusted_get(struct inode *inode, const char *name,
5346 void *buffer, size_t size)
5347{
5348 if (strcmp(name, "") == 0)
5349 return -EINVAL;
5350 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_TRUSTED, name,
5351 buffer, size);
5352}
5353
5354static int ocfs2_xattr_trusted_set(struct inode *inode, const char *name,
5355 const void *value, size_t size, int flags)
5356{
5357 if (strcmp(name, "") == 0)
5358 return -EINVAL;
5359
5360 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_TRUSTED, name, value,
5361 size, flags);
5362}
5363
5364struct xattr_handler ocfs2_xattr_trusted_handler = {
5365 .prefix = XATTR_TRUSTED_PREFIX,
5366 .list = ocfs2_xattr_trusted_list,
5367 .get = ocfs2_xattr_trusted_get,
5368 .set = ocfs2_xattr_trusted_set,
5369};
5370
99219aea
MF		 5371/*
5372 * 'user' attributes support
5373 */
99219aea
MF		 5374static size_t ocfs2_xattr_user_list(struct inode *inode, char *list,
5375 size_t list_size, const char *name,
5376 size_t name_len)
5377{
ceb1eba3 5378 const size_t prefix_len = XATTR_USER_PREFIX_LEN;
99219aea
MF		 5379 const size_t total_len = prefix_len + name_len + 1;
5380 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5381
5382 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5383 return 0;
5384
5385 if (list && total_len <= list_size) {
5386 memcpy(list, XATTR_USER_PREFIX, prefix_len);
5387 memcpy(list + prefix_len, name, name_len);
5388 list[prefix_len + name_len] = '\0';
5389 }
5390 return total_len;
5391}
5392
5393static int ocfs2_xattr_user_get(struct inode *inode, const char *name,
5394 void *buffer, size_t size)
5395{
5396 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5397
5398 if (strcmp(name, "") == 0)
5399 return -EINVAL;
5400 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5401 return -EOPNOTSUPP;
5402 return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_USER, name,
5403 buffer, size);
5404}
5405
5406static int ocfs2_xattr_user_set(struct inode *inode, const char *name,
5407 const void *value, size_t size, int flags)
5408{
5409 struct ocfs2_super *osb = OCFS2_SB(inode->i_sb);
5410
5411 if (strcmp(name, "") == 0)
5412 return -EINVAL;
5413 if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR)
5414 return -EOPNOTSUPP;
5415
5416 return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_USER, name, value,
5417 size, flags);
5418}
5419
5420struct xattr_handler ocfs2_xattr_user_handler = {
5421 .prefix = XATTR_USER_PREFIX,
5422 .list = ocfs2_xattr_user_list,
5423 .get = ocfs2_xattr_user_get,
5424 .set = ocfs2_xattr_user_set,
5425};
Linux 6.x block layer and io_uring tree(s)