]>
Commit | Line | Data |
---|---|---|
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 | 57 | #include "xattr.h" |
492a8a33 | 58 | #include "refcounttree.h" |
0fe9b66c | 59 | #include "acl.h" |
cf1d6c76 TY |
60 | |
61 | struct ocfs2_xattr_def_value_root { | |
62 | struct ocfs2_xattr_value_root xv; | |
63 | struct ocfs2_extent_rec er; | |
64 | }; | |
65 | ||
0c044f0b | 66 | struct ocfs2_xattr_bucket { |
ba937127 JB |
67 | /* The inode these xattrs are associated with */ |
68 | struct inode *bu_inode; | |
69 | ||
70 | /* The actual buffers that make up the bucket */ | |
4ac6032d | 71 | struct buffer_head *bu_bhs[OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET]; |
ba937127 JB |
72 | |
73 | /* How many blocks make up one bucket for this filesystem */ | |
74 | int bu_blocks; | |
0c044f0b TM |
75 | }; |
76 | ||
78f30c31 | 77 | struct ocfs2_xattr_set_ctxt { |
85db90e7 | 78 | handle_t *handle; |
78f30c31 TM |
79 | struct ocfs2_alloc_context *meta_ac; |
80 | struct ocfs2_alloc_context *data_ac; | |
81 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
82 | }; | |
83 | ||
cf1d6c76 TY |
84 | #define OCFS2_XATTR_ROOT_SIZE (sizeof(struct ocfs2_xattr_def_value_root)) |
85 | #define OCFS2_XATTR_INLINE_SIZE 80 | |
4442f518 | 86 | #define OCFS2_XATTR_HEADER_GAP 4 |
534eaddd TY |
87 | #define OCFS2_XATTR_FREE_IN_IBODY (OCFS2_MIN_XATTR_INLINE_SIZE \ |
88 | - sizeof(struct ocfs2_xattr_header) \ | |
4442f518 | 89 | - OCFS2_XATTR_HEADER_GAP) |
89c38bd0 TY |
90 | #define OCFS2_XATTR_FREE_IN_BLOCK(ptr) ((ptr)->i_sb->s_blocksize \ |
91 | - sizeof(struct ocfs2_xattr_block) \ | |
92 | - sizeof(struct ocfs2_xattr_header) \ | |
4442f518 | 93 | - OCFS2_XATTR_HEADER_GAP) |
cf1d6c76 TY |
94 | |
95 | static struct ocfs2_xattr_def_value_root def_xv = { | |
96 | .xv.xr_list.l_count = cpu_to_le16(1), | |
97 | }; | |
98 | ||
99 | struct xattr_handler *ocfs2_xattr_handlers[] = { | |
100 | &ocfs2_xattr_user_handler, | |
929fb014 TY |
101 | &ocfs2_xattr_acl_access_handler, |
102 | &ocfs2_xattr_acl_default_handler, | |
cf1d6c76 | 103 | &ocfs2_xattr_trusted_handler, |
923f7f31 | 104 | &ocfs2_xattr_security_handler, |
cf1d6c76 TY |
105 | NULL |
106 | }; | |
107 | ||
c988fd04 | 108 | static struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = { |
cf1d6c76 | 109 | [OCFS2_XATTR_INDEX_USER] = &ocfs2_xattr_user_handler, |
929fb014 TY |
110 | [OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS] |
111 | = &ocfs2_xattr_acl_access_handler, | |
112 | [OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT] | |
113 | = &ocfs2_xattr_acl_default_handler, | |
cf1d6c76 | 114 | [OCFS2_XATTR_INDEX_TRUSTED] = &ocfs2_xattr_trusted_handler, |
923f7f31 | 115 | [OCFS2_XATTR_INDEX_SECURITY] = &ocfs2_xattr_security_handler, |
cf1d6c76 TY |
116 | }; |
117 | ||
118 | struct ocfs2_xattr_info { | |
119 | int name_index; | |
120 | const char *name; | |
121 | const void *value; | |
122 | size_t value_len; | |
123 | }; | |
124 | ||
125 | struct ocfs2_xattr_search { | |
126 | struct buffer_head *inode_bh; | |
127 | /* | |
128 | * xattr_bh point to the block buffer head which has extended attribute | |
129 | * when extended attribute in inode, xattr_bh is equal to inode_bh. | |
130 | */ | |
131 | struct buffer_head *xattr_bh; | |
132 | struct ocfs2_xattr_header *header; | |
ba937127 | 133 | struct ocfs2_xattr_bucket *bucket; |
cf1d6c76 TY |
134 | void *base; |
135 | void *end; | |
136 | struct ocfs2_xattr_entry *here; | |
137 | int not_found; | |
138 | }; | |
139 | ||
11179f2c JB |
140 | /* Operations on struct ocfs2_xa_entry */ |
141 | struct ocfs2_xa_loc; | |
142 | struct ocfs2_xa_loc_operations { | |
143 | /* | |
144 | * Return a pointer to the appropriate buffer in loc->xl_storage | |
145 | * at the given offset from loc->xl_header. | |
146 | */ | |
147 | void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset); | |
148 | ||
149 | /* | |
150 | * Remove the name+value at this location. Do whatever is | |
151 | * appropriate with the remaining name+value pairs. | |
152 | */ | |
153 | void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc); | |
154 | }; | |
155 | ||
156 | /* | |
157 | * Describes an xattr entry location. This is a memory structure | |
158 | * tracking the on-disk structure. | |
159 | */ | |
160 | struct ocfs2_xa_loc { | |
161 | /* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */ | |
162 | struct ocfs2_xattr_header *xl_header; | |
163 | ||
164 | /* Bytes from xl_header to the end of the storage */ | |
165 | int xl_size; | |
166 | ||
167 | /* | |
168 | * The ocfs2_xattr_entry this location describes. If this is | |
169 | * NULL, this location describes the on-disk structure where it | |
170 | * would have been. | |
171 | */ | |
172 | struct ocfs2_xattr_entry *xl_entry; | |
173 | ||
174 | /* | |
175 | * Internal housekeeping | |
176 | */ | |
177 | ||
178 | /* Buffer(s) containing this entry */ | |
179 | void *xl_storage; | |
180 | ||
181 | /* Operations on the storage backing this location */ | |
182 | const struct ocfs2_xa_loc_operations *xl_ops; | |
183 | }; | |
184 | ||
fd68a894 | 185 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
589dc260 TM |
186 | struct ocfs2_xattr_header *xh, |
187 | int index, | |
188 | int *block_off, | |
189 | int *new_offset); | |
190 | ||
54f443f4 JB |
191 | static int ocfs2_xattr_block_find(struct inode *inode, |
192 | int name_index, | |
193 | const char *name, | |
194 | struct ocfs2_xattr_search *xs); | |
589dc260 TM |
195 | static int ocfs2_xattr_index_block_find(struct inode *inode, |
196 | struct buffer_head *root_bh, | |
197 | int name_index, | |
198 | const char *name, | |
199 | struct ocfs2_xattr_search *xs); | |
200 | ||
0c044f0b | 201 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, |
47bca495 | 202 | struct buffer_head *blk_bh, |
0c044f0b TM |
203 | char *buffer, |
204 | size_t buffer_size); | |
205 | ||
01225596 | 206 | static int ocfs2_xattr_create_index_block(struct inode *inode, |
78f30c31 TM |
207 | struct ocfs2_xattr_search *xs, |
208 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 TM |
209 | |
210 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, | |
211 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
212 | struct ocfs2_xattr_search *xs, |
213 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 | 214 | |
47bca495 TM |
215 | typedef int (xattr_tree_rec_func)(struct inode *inode, |
216 | struct buffer_head *root_bh, | |
217 | u64 blkno, u32 cpos, u32 len, void *para); | |
218 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, | |
219 | struct buffer_head *root_bh, | |
220 | xattr_tree_rec_func *rec_func, | |
221 | void *para); | |
222 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
223 | struct ocfs2_xattr_bucket *bucket, | |
224 | void *para); | |
225 | static int ocfs2_rm_xattr_cluster(struct inode *inode, | |
226 | struct buffer_head *root_bh, | |
227 | u64 blkno, | |
228 | u32 cpos, | |
229 | u32 len, | |
230 | void *para); | |
231 | ||
c58b6032 JB |
232 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
233 | u64 src_blk, u64 last_blk, u64 to_blk, | |
234 | unsigned int start_bucket, | |
235 | u32 *first_hash); | |
492a8a33 TM |
236 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, |
237 | struct ocfs2_dinode *di, | |
238 | struct ocfs2_xattr_info *xi, | |
239 | struct ocfs2_xattr_search *xis, | |
240 | struct ocfs2_xattr_search *xbs, | |
241 | struct ocfs2_refcount_tree **ref_tree, | |
242 | int *meta_need, | |
243 | int *credits); | |
ce9c5a54 TM |
244 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, |
245 | struct ocfs2_xattr_bucket *bucket, | |
246 | int offset, | |
247 | struct ocfs2_xattr_value_root **xv, | |
248 | struct buffer_head **bh); | |
a3944256 | 249 | |
0030e001 TY |
250 | static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb) |
251 | { | |
252 | return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE; | |
253 | } | |
254 | ||
255 | static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb) | |
256 | { | |
257 | return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits); | |
258 | } | |
259 | ||
260 | static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb) | |
261 | { | |
262 | u16 len = sb->s_blocksize - | |
263 | offsetof(struct ocfs2_xattr_header, xh_entries); | |
264 | ||
265 | return len / sizeof(struct ocfs2_xattr_entry); | |
266 | } | |
267 | ||
9c7759aa | 268 | #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr) |
51def39f | 269 | #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data) |
3e632946 | 270 | #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0)) |
9c7759aa | 271 | |
ba937127 | 272 | static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode) |
6dde41d9 | 273 | { |
ba937127 JB |
274 | struct ocfs2_xattr_bucket *bucket; |
275 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
6dde41d9 | 276 | |
ba937127 JB |
277 | BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET); |
278 | ||
279 | bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS); | |
280 | if (bucket) { | |
281 | bucket->bu_inode = inode; | |
282 | bucket->bu_blocks = blks; | |
283 | } | |
284 | ||
285 | return bucket; | |
286 | } | |
287 | ||
288 | static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket) | |
289 | { | |
290 | int i; | |
291 | ||
292 | for (i = 0; i < bucket->bu_blocks; i++) { | |
6dde41d9 JB |
293 | brelse(bucket->bu_bhs[i]); |
294 | bucket->bu_bhs[i] = NULL; | |
295 | } | |
296 | } | |
297 | ||
ba937127 JB |
298 | static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket) |
299 | { | |
300 | if (bucket) { | |
301 | ocfs2_xattr_bucket_relse(bucket); | |
302 | bucket->bu_inode = NULL; | |
303 | kfree(bucket); | |
304 | } | |
305 | } | |
306 | ||
784b816a JB |
307 | /* |
308 | * A bucket that has never been written to disk doesn't need to be | |
309 | * read. We just need the buffer_heads. Don't call this for | |
310 | * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes | |
311 | * them fully. | |
312 | */ | |
ba937127 | 313 | static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
314 | u64 xb_blkno) |
315 | { | |
316 | int i, rc = 0; | |
784b816a | 317 | |
ba937127 JB |
318 | for (i = 0; i < bucket->bu_blocks; i++) { |
319 | bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb, | |
320 | xb_blkno + i); | |
784b816a JB |
321 | if (!bucket->bu_bhs[i]) { |
322 | rc = -EIO; | |
323 | mlog_errno(rc); | |
324 | break; | |
325 | } | |
326 | ||
8cb471e8 | 327 | if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
757055ad | 328 | bucket->bu_bhs[i])) |
8cb471e8 | 329 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
757055ad | 330 | bucket->bu_bhs[i]); |
784b816a JB |
331 | } |
332 | ||
333 | if (rc) | |
ba937127 | 334 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
335 | return rc; |
336 | } | |
337 | ||
338 | /* Read the xattr bucket at xb_blkno */ | |
ba937127 | 339 | static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
340 | u64 xb_blkno) |
341 | { | |
ba937127 | 342 | int rc; |
784b816a | 343 | |
8cb471e8 | 344 | rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno, |
970e4936 JB |
345 | bucket->bu_blocks, bucket->bu_bhs, 0, |
346 | NULL); | |
4d0e214e | 347 | if (!rc) { |
c8b9cf9a | 348 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
349 | rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb, |
350 | bucket->bu_bhs, | |
351 | bucket->bu_blocks, | |
352 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 353 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
354 | if (rc) |
355 | mlog_errno(rc); | |
356 | } | |
357 | ||
784b816a | 358 | if (rc) |
ba937127 | 359 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
360 | return rc; |
361 | } | |
362 | ||
1224be02 | 363 | static int ocfs2_xattr_bucket_journal_access(handle_t *handle, |
1224be02 JB |
364 | struct ocfs2_xattr_bucket *bucket, |
365 | int type) | |
366 | { | |
367 | int i, rc = 0; | |
1224be02 | 368 | |
ba937127 | 369 | for (i = 0; i < bucket->bu_blocks; i++) { |
0cf2f763 JB |
370 | rc = ocfs2_journal_access(handle, |
371 | INODE_CACHE(bucket->bu_inode), | |
1224be02 JB |
372 | bucket->bu_bhs[i], type); |
373 | if (rc) { | |
374 | mlog_errno(rc); | |
375 | break; | |
376 | } | |
377 | } | |
378 | ||
379 | return rc; | |
380 | } | |
381 | ||
382 | static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle, | |
1224be02 JB |
383 | struct ocfs2_xattr_bucket *bucket) |
384 | { | |
ba937127 | 385 | int i; |
1224be02 | 386 | |
c8b9cf9a | 387 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
388 | ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb, |
389 | bucket->bu_bhs, bucket->bu_blocks, | |
390 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 391 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e | 392 | |
ba937127 | 393 | for (i = 0; i < bucket->bu_blocks; i++) |
1224be02 JB |
394 | ocfs2_journal_dirty(handle, bucket->bu_bhs[i]); |
395 | } | |
396 | ||
ba937127 | 397 | static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest, |
4980c6da JB |
398 | struct ocfs2_xattr_bucket *src) |
399 | { | |
400 | int i; | |
ba937127 JB |
401 | int blocksize = src->bu_inode->i_sb->s_blocksize; |
402 | ||
403 | BUG_ON(dest->bu_blocks != src->bu_blocks); | |
404 | BUG_ON(dest->bu_inode != src->bu_inode); | |
4980c6da | 405 | |
ba937127 | 406 | for (i = 0; i < src->bu_blocks; i++) { |
4980c6da JB |
407 | memcpy(bucket_block(dest, i), bucket_block(src, i), |
408 | blocksize); | |
409 | } | |
410 | } | |
1224be02 | 411 | |
4ae1d69b JB |
412 | static int ocfs2_validate_xattr_block(struct super_block *sb, |
413 | struct buffer_head *bh) | |
414 | { | |
d6b32bbb | 415 | int rc; |
4ae1d69b JB |
416 | struct ocfs2_xattr_block *xb = |
417 | (struct ocfs2_xattr_block *)bh->b_data; | |
418 | ||
419 | mlog(0, "Validating xattr block %llu\n", | |
420 | (unsigned long long)bh->b_blocknr); | |
421 | ||
d6b32bbb JB |
422 | BUG_ON(!buffer_uptodate(bh)); |
423 | ||
424 | /* | |
425 | * If the ecc fails, we return the error but otherwise | |
426 | * leave the filesystem running. We know any error is | |
427 | * local to this block. | |
428 | */ | |
429 | rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check); | |
430 | if (rc) | |
431 | return rc; | |
432 | ||
433 | /* | |
434 | * Errors after here are fatal | |
435 | */ | |
436 | ||
4ae1d69b JB |
437 | if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) { |
438 | ocfs2_error(sb, | |
439 | "Extended attribute block #%llu has bad " | |
440 | "signature %.*s", | |
441 | (unsigned long long)bh->b_blocknr, 7, | |
442 | xb->xb_signature); | |
443 | return -EINVAL; | |
444 | } | |
445 | ||
446 | if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) { | |
447 | ocfs2_error(sb, | |
448 | "Extended attribute block #%llu has an " | |
449 | "invalid xb_blkno of %llu", | |
450 | (unsigned long long)bh->b_blocknr, | |
451 | (unsigned long long)le64_to_cpu(xb->xb_blkno)); | |
452 | return -EINVAL; | |
453 | } | |
454 | ||
455 | if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) { | |
456 | ocfs2_error(sb, | |
457 | "Extended attribute block #%llu has an invalid " | |
458 | "xb_fs_generation of #%u", | |
459 | (unsigned long long)bh->b_blocknr, | |
460 | le32_to_cpu(xb->xb_fs_generation)); | |
461 | return -EINVAL; | |
462 | } | |
463 | ||
464 | return 0; | |
465 | } | |
466 | ||
467 | static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, | |
468 | struct buffer_head **bh) | |
469 | { | |
470 | int rc; | |
471 | struct buffer_head *tmp = *bh; | |
472 | ||
8cb471e8 | 473 | rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp, |
970e4936 | 474 | ocfs2_validate_xattr_block); |
4ae1d69b JB |
475 | |
476 | /* If ocfs2_read_block() got us a new bh, pass it up. */ | |
477 | if (!rc && !*bh) | |
478 | *bh = tmp; | |
479 | ||
480 | return rc; | |
481 | } | |
482 | ||
936b8834 | 483 | static inline const char *ocfs2_xattr_prefix(int name_index) |
cf1d6c76 TY |
484 | { |
485 | struct xattr_handler *handler = NULL; | |
486 | ||
487 | if (name_index > 0 && name_index < OCFS2_XATTR_MAX) | |
488 | handler = ocfs2_xattr_handler_map[name_index]; | |
489 | ||
936b8834 | 490 | return handler ? handler->prefix : NULL; |
cf1d6c76 TY |
491 | } |
492 | ||
40daa16a | 493 | static u32 ocfs2_xattr_name_hash(struct inode *inode, |
2057e5c6 | 494 | const char *name, |
40daa16a | 495 | int name_len) |
cf1d6c76 TY |
496 | { |
497 | /* Get hash value of uuid from super block */ | |
498 | u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash; | |
499 | int i; | |
500 | ||
cf1d6c76 TY |
501 | /* hash extended attribute name */ |
502 | for (i = 0; i < name_len; i++) { | |
503 | hash = (hash << OCFS2_HASH_SHIFT) ^ | |
504 | (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^ | |
505 | *name++; | |
506 | } | |
507 | ||
508 | return hash; | |
509 | } | |
510 | ||
511 | /* | |
512 | * ocfs2_xattr_hash_entry() | |
513 | * | |
514 | * Compute the hash of an extended attribute. | |
515 | */ | |
516 | static void ocfs2_xattr_hash_entry(struct inode *inode, | |
517 | struct ocfs2_xattr_header *header, | |
518 | struct ocfs2_xattr_entry *entry) | |
519 | { | |
520 | u32 hash = 0; | |
cf1d6c76 | 521 | char *name = (char *)header + le16_to_cpu(entry->xe_name_offset); |
cf1d6c76 | 522 | |
2057e5c6 | 523 | hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len); |
cf1d6c76 TY |
524 | entry->xe_name_hash = cpu_to_le32(hash); |
525 | ||
526 | return; | |
527 | } | |
f56654c4 | 528 | |
534eaddd TY |
529 | static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len) |
530 | { | |
531 | int size = 0; | |
532 | ||
533 | if (value_len <= OCFS2_XATTR_INLINE_SIZE) | |
534 | size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len); | |
535 | else | |
536 | size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
537 | size += sizeof(struct ocfs2_xattr_entry); | |
538 | ||
539 | return size; | |
540 | } | |
541 | ||
542 | int ocfs2_calc_security_init(struct inode *dir, | |
543 | struct ocfs2_security_xattr_info *si, | |
544 | int *want_clusters, | |
545 | int *xattr_credits, | |
546 | struct ocfs2_alloc_context **xattr_ac) | |
547 | { | |
548 | int ret = 0; | |
549 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
550 | int s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
551 | si->value_len); | |
552 | ||
553 | /* | |
554 | * The max space of security xattr taken inline is | |
555 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
556 | * So reserve one metadata block for it is ok. | |
557 | */ | |
558 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
559 | s_size > OCFS2_XATTR_FREE_IN_IBODY) { | |
560 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac); | |
561 | if (ret) { | |
562 | mlog_errno(ret); | |
563 | return ret; | |
564 | } | |
565 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; | |
566 | } | |
567 | ||
568 | /* reserve clusters for xattr value which will be set in B tree*/ | |
0e445b6f TY |
569 | if (si->value_len > OCFS2_XATTR_INLINE_SIZE) { |
570 | int new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
571 | si->value_len); | |
572 | ||
573 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
574 | new_clusters); | |
575 | *want_clusters += new_clusters; | |
576 | } | |
534eaddd TY |
577 | return ret; |
578 | } | |
579 | ||
89c38bd0 TY |
580 | int ocfs2_calc_xattr_init(struct inode *dir, |
581 | struct buffer_head *dir_bh, | |
582 | int mode, | |
583 | struct ocfs2_security_xattr_info *si, | |
584 | int *want_clusters, | |
585 | int *xattr_credits, | |
9b7895ef | 586 | int *want_meta) |
89c38bd0 TY |
587 | { |
588 | int ret = 0; | |
589 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
0e445b6f | 590 | int s_size = 0, a_size = 0, acl_len = 0, new_clusters; |
89c38bd0 TY |
591 | |
592 | if (si->enable) | |
593 | s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
594 | si->value_len); | |
595 | ||
596 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) { | |
597 | acl_len = ocfs2_xattr_get_nolock(dir, dir_bh, | |
598 | OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT, | |
599 | "", NULL, 0); | |
600 | if (acl_len > 0) { | |
601 | a_size = ocfs2_xattr_entry_real_size(0, acl_len); | |
602 | if (S_ISDIR(mode)) | |
603 | a_size <<= 1; | |
604 | } else if (acl_len != 0 && acl_len != -ENODATA) { | |
605 | mlog_errno(ret); | |
606 | return ret; | |
607 | } | |
608 | } | |
609 | ||
610 | if (!(s_size + a_size)) | |
611 | return ret; | |
612 | ||
613 | /* | |
614 | * The max space of security xattr taken inline is | |
615 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
616 | * The max space of acl xattr taken inline is | |
617 | * 80(value) + 16(entry) * 2(if directory) = 192 bytes, | |
618 | * when blocksize = 512, may reserve one more cluser for | |
619 | * xattr bucket, otherwise reserve one metadata block | |
620 | * for them is ok. | |
6c9fd1dc TY |
621 | * If this is a new directory with inline data, |
622 | * we choose to reserve the entire inline area for | |
623 | * directory contents and force an external xattr block. | |
89c38bd0 TY |
624 | */ |
625 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
6c9fd1dc | 626 | (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) || |
89c38bd0 | 627 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) { |
9b7895ef | 628 | *want_meta = *want_meta + 1; |
89c38bd0 TY |
629 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
630 | } | |
631 | ||
632 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE && | |
633 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) { | |
634 | *want_clusters += 1; | |
635 | *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb); | |
636 | } | |
637 | ||
0e445b6f TY |
638 | /* |
639 | * reserve credits and clusters for xattrs which has large value | |
640 | * and have to be set outside | |
641 | */ | |
642 | if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
643 | new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
644 | si->value_len); | |
645 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
646 | new_clusters); | |
647 | *want_clusters += new_clusters; | |
648 | } | |
89c38bd0 TY |
649 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL && |
650 | acl_len > OCFS2_XATTR_INLINE_SIZE) { | |
0e445b6f TY |
651 | /* for directory, it has DEFAULT and ACCESS two types of acls */ |
652 | new_clusters = (S_ISDIR(mode) ? 2 : 1) * | |
653 | ocfs2_clusters_for_bytes(dir->i_sb, acl_len); | |
654 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
655 | new_clusters); | |
656 | *want_clusters += new_clusters; | |
89c38bd0 TY |
657 | } |
658 | ||
659 | return ret; | |
660 | } | |
661 | ||
f56654c4 TM |
662 | static int ocfs2_xattr_extend_allocation(struct inode *inode, |
663 | u32 clusters_to_add, | |
19b801f4 | 664 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 665 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
666 | { |
667 | int status = 0; | |
85db90e7 | 668 | handle_t *handle = ctxt->handle; |
f56654c4 | 669 | enum ocfs2_alloc_restarted why; |
19b801f4 | 670 | u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters); |
f99b9b7c | 671 | struct ocfs2_extent_tree et; |
f56654c4 TM |
672 | |
673 | mlog(0, "(clusters_to_add for xattr= %u)\n", clusters_to_add); | |
674 | ||
5e404e9e | 675 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f99b9b7c | 676 | |
0cf2f763 | 677 | status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
2a50a743 | 678 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
679 | if (status < 0) { |
680 | mlog_errno(status); | |
681 | goto leave; | |
682 | } | |
683 | ||
19b801f4 | 684 | prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
cbee7e1a JB |
685 | status = ocfs2_add_clusters_in_btree(handle, |
686 | &et, | |
f56654c4 TM |
687 | &logical_start, |
688 | clusters_to_add, | |
689 | 0, | |
78f30c31 TM |
690 | ctxt->data_ac, |
691 | ctxt->meta_ac, | |
f99b9b7c | 692 | &why); |
85db90e7 TM |
693 | if (status < 0) { |
694 | mlog_errno(status); | |
f56654c4 TM |
695 | goto leave; |
696 | } | |
697 | ||
19b801f4 | 698 | status = ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 TM |
699 | if (status < 0) { |
700 | mlog_errno(status); | |
701 | goto leave; | |
702 | } | |
703 | ||
19b801f4 | 704 | clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters; |
f56654c4 | 705 | |
85db90e7 TM |
706 | /* |
707 | * We should have already allocated enough space before the transaction, | |
708 | * so no need to restart. | |
709 | */ | |
710 | BUG_ON(why != RESTART_NONE || clusters_to_add); | |
f56654c4 TM |
711 | |
712 | leave: | |
f56654c4 TM |
713 | |
714 | return status; | |
715 | } | |
716 | ||
717 | static int __ocfs2_remove_xattr_range(struct inode *inode, | |
d72cc72d | 718 | struct ocfs2_xattr_value_buf *vb, |
f56654c4 | 719 | u32 cpos, u32 phys_cpos, u32 len, |
492a8a33 | 720 | unsigned int ext_flags, |
78f30c31 | 721 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
722 | { |
723 | int ret; | |
724 | u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
85db90e7 | 725 | handle_t *handle = ctxt->handle; |
f99b9b7c JB |
726 | struct ocfs2_extent_tree et; |
727 | ||
5e404e9e | 728 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f56654c4 | 729 | |
0cf2f763 | 730 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
d72cc72d | 731 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
732 | if (ret) { |
733 | mlog_errno(ret); | |
85db90e7 | 734 | goto out; |
f56654c4 TM |
735 | } |
736 | ||
dbdcf6a4 | 737 | ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac, |
78f30c31 | 738 | &ctxt->dealloc); |
f56654c4 TM |
739 | if (ret) { |
740 | mlog_errno(ret); | |
85db90e7 | 741 | goto out; |
f56654c4 TM |
742 | } |
743 | ||
d72cc72d | 744 | le32_add_cpu(&vb->vb_xv->xr_clusters, -len); |
f56654c4 | 745 | |
d72cc72d | 746 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 TM |
747 | if (ret) { |
748 | mlog_errno(ret); | |
85db90e7 | 749 | goto out; |
f56654c4 TM |
750 | } |
751 | ||
492a8a33 TM |
752 | if (ext_flags & OCFS2_EXT_REFCOUNTED) |
753 | ret = ocfs2_decrease_refcount(inode, handle, | |
754 | ocfs2_blocks_to_clusters(inode->i_sb, | |
755 | phys_blkno), | |
756 | len, ctxt->meta_ac, &ctxt->dealloc, 1); | |
757 | else | |
758 | ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, | |
759 | phys_blkno, len); | |
f56654c4 TM |
760 | if (ret) |
761 | mlog_errno(ret); | |
762 | ||
f56654c4 | 763 | out: |
f56654c4 TM |
764 | return ret; |
765 | } | |
766 | ||
767 | static int ocfs2_xattr_shrink_size(struct inode *inode, | |
768 | u32 old_clusters, | |
769 | u32 new_clusters, | |
19b801f4 | 770 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 771 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
772 | { |
773 | int ret = 0; | |
492a8a33 | 774 | unsigned int ext_flags; |
f56654c4 TM |
775 | u32 trunc_len, cpos, phys_cpos, alloc_size; |
776 | u64 block; | |
f56654c4 TM |
777 | |
778 | if (old_clusters <= new_clusters) | |
779 | return 0; | |
780 | ||
781 | cpos = new_clusters; | |
782 | trunc_len = old_clusters - new_clusters; | |
783 | while (trunc_len) { | |
784 | ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos, | |
d72cc72d | 785 | &alloc_size, |
492a8a33 | 786 | &vb->vb_xv->xr_list, &ext_flags); |
f56654c4 TM |
787 | if (ret) { |
788 | mlog_errno(ret); | |
789 | goto out; | |
790 | } | |
791 | ||
792 | if (alloc_size > trunc_len) | |
793 | alloc_size = trunc_len; | |
794 | ||
19b801f4 | 795 | ret = __ocfs2_remove_xattr_range(inode, vb, cpos, |
f56654c4 | 796 | phys_cpos, alloc_size, |
492a8a33 | 797 | ext_flags, ctxt); |
f56654c4 TM |
798 | if (ret) { |
799 | mlog_errno(ret); | |
800 | goto out; | |
801 | } | |
802 | ||
803 | block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
8cb471e8 JB |
804 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), |
805 | block, alloc_size); | |
f56654c4 TM |
806 | cpos += alloc_size; |
807 | trunc_len -= alloc_size; | |
808 | } | |
809 | ||
810 | out: | |
f56654c4 TM |
811 | return ret; |
812 | } | |
813 | ||
814 | static int ocfs2_xattr_value_truncate(struct inode *inode, | |
b3e5d379 | 815 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 TM |
816 | int len, |
817 | struct ocfs2_xattr_set_ctxt *ctxt) | |
f56654c4 TM |
818 | { |
819 | int ret; | |
820 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len); | |
b3e5d379 | 821 | u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
f56654c4 TM |
822 | |
823 | if (new_clusters == old_clusters) | |
824 | return 0; | |
825 | ||
826 | if (new_clusters > old_clusters) | |
827 | ret = ocfs2_xattr_extend_allocation(inode, | |
828 | new_clusters - old_clusters, | |
b3e5d379 | 829 | vb, ctxt); |
f56654c4 TM |
830 | else |
831 | ret = ocfs2_xattr_shrink_size(inode, | |
832 | old_clusters, new_clusters, | |
b3e5d379 | 833 | vb, ctxt); |
f56654c4 TM |
834 | |
835 | return ret; | |
836 | } | |
cf1d6c76 | 837 | |
936b8834 TM |
838 | static int ocfs2_xattr_list_entry(char *buffer, size_t size, |
839 | size_t *result, const char *prefix, | |
840 | const char *name, int name_len) | |
841 | { | |
842 | char *p = buffer + *result; | |
843 | int prefix_len = strlen(prefix); | |
844 | int total_len = prefix_len + name_len + 1; | |
845 | ||
846 | *result += total_len; | |
847 | ||
848 | /* we are just looking for how big our buffer needs to be */ | |
849 | if (!size) | |
850 | return 0; | |
851 | ||
852 | if (*result > size) | |
853 | return -ERANGE; | |
854 | ||
855 | memcpy(p, prefix, prefix_len); | |
856 | memcpy(p + prefix_len, name, name_len); | |
857 | p[prefix_len + name_len] = '\0'; | |
858 | ||
859 | return 0; | |
860 | } | |
861 | ||
cf1d6c76 TY |
862 | static int ocfs2_xattr_list_entries(struct inode *inode, |
863 | struct ocfs2_xattr_header *header, | |
864 | char *buffer, size_t buffer_size) | |
865 | { | |
936b8834 TM |
866 | size_t result = 0; |
867 | int i, type, ret; | |
868 | const char *prefix, *name; | |
cf1d6c76 TY |
869 | |
870 | for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) { | |
871 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
936b8834 TM |
872 | type = ocfs2_xattr_get_type(entry); |
873 | prefix = ocfs2_xattr_prefix(type); | |
874 | ||
875 | if (prefix) { | |
876 | name = (const char *)header + | |
877 | le16_to_cpu(entry->xe_name_offset); | |
878 | ||
879 | ret = ocfs2_xattr_list_entry(buffer, buffer_size, | |
880 | &result, prefix, name, | |
881 | entry->xe_name_len); | |
882 | if (ret) | |
883 | return ret; | |
cf1d6c76 TY |
884 | } |
885 | } | |
886 | ||
936b8834 | 887 | return result; |
cf1d6c76 TY |
888 | } |
889 | ||
8b2c0dba TM |
890 | int ocfs2_has_inline_xattr_value_outside(struct inode *inode, |
891 | struct ocfs2_dinode *di) | |
892 | { | |
893 | struct ocfs2_xattr_header *xh; | |
894 | int i; | |
895 | ||
896 | xh = (struct ocfs2_xattr_header *) | |
897 | ((void *)di + inode->i_sb->s_blocksize - | |
898 | le16_to_cpu(di->i_xattr_inline_size)); | |
899 | ||
900 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) | |
901 | if (!ocfs2_xattr_is_local(&xh->xh_entries[i])) | |
902 | return 1; | |
903 | ||
904 | return 0; | |
905 | } | |
906 | ||
cf1d6c76 TY |
907 | static int ocfs2_xattr_ibody_list(struct inode *inode, |
908 | struct ocfs2_dinode *di, | |
909 | char *buffer, | |
910 | size_t buffer_size) | |
911 | { | |
912 | struct ocfs2_xattr_header *header = NULL; | |
913 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
914 | int ret = 0; | |
915 | ||
916 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
917 | return ret; | |
918 | ||
919 | header = (struct ocfs2_xattr_header *) | |
920 | ((void *)di + inode->i_sb->s_blocksize - | |
921 | le16_to_cpu(di->i_xattr_inline_size)); | |
922 | ||
923 | ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size); | |
924 | ||
925 | return ret; | |
926 | } | |
927 | ||
928 | static int ocfs2_xattr_block_list(struct inode *inode, | |
929 | struct ocfs2_dinode *di, | |
930 | char *buffer, | |
931 | size_t buffer_size) | |
932 | { | |
933 | struct buffer_head *blk_bh = NULL; | |
0c044f0b | 934 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
935 | int ret = 0; |
936 | ||
937 | if (!di->i_xattr_loc) | |
938 | return ret; | |
939 | ||
4ae1d69b JB |
940 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
941 | &blk_bh); | |
cf1d6c76 TY |
942 | if (ret < 0) { |
943 | mlog_errno(ret); | |
944 | return ret; | |
945 | } | |
cf1d6c76 | 946 | |
0c044f0b | 947 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
0c044f0b TM |
948 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
949 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
950 | ret = ocfs2_xattr_list_entries(inode, header, | |
951 | buffer, buffer_size); | |
47bca495 TM |
952 | } else |
953 | ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh, | |
0c044f0b | 954 | buffer, buffer_size); |
4ae1d69b | 955 | |
cf1d6c76 TY |
956 | brelse(blk_bh); |
957 | ||
958 | return ret; | |
959 | } | |
960 | ||
961 | ssize_t ocfs2_listxattr(struct dentry *dentry, | |
962 | char *buffer, | |
963 | size_t size) | |
964 | { | |
965 | int ret = 0, i_ret = 0, b_ret = 0; | |
966 | struct buffer_head *di_bh = NULL; | |
967 | struct ocfs2_dinode *di = NULL; | |
968 | struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode); | |
969 | ||
8154da3d TY |
970 | if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb))) |
971 | return -EOPNOTSUPP; | |
972 | ||
cf1d6c76 TY |
973 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
974 | return ret; | |
975 | ||
976 | ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0); | |
977 | if (ret < 0) { | |
978 | mlog_errno(ret); | |
979 | return ret; | |
980 | } | |
981 | ||
982 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
983 | ||
984 | down_read(&oi->ip_xattr_sem); | |
985 | i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size); | |
986 | if (i_ret < 0) | |
987 | b_ret = 0; | |
988 | else { | |
989 | if (buffer) { | |
990 | buffer += i_ret; | |
991 | size -= i_ret; | |
992 | } | |
993 | b_ret = ocfs2_xattr_block_list(dentry->d_inode, di, | |
994 | buffer, size); | |
995 | if (b_ret < 0) | |
996 | i_ret = 0; | |
997 | } | |
998 | up_read(&oi->ip_xattr_sem); | |
999 | ocfs2_inode_unlock(dentry->d_inode, 0); | |
1000 | ||
1001 | brelse(di_bh); | |
1002 | ||
1003 | return i_ret + b_ret; | |
1004 | } | |
1005 | ||
1006 | static int ocfs2_xattr_find_entry(int name_index, | |
1007 | const char *name, | |
1008 | struct ocfs2_xattr_search *xs) | |
1009 | { | |
1010 | struct ocfs2_xattr_entry *entry; | |
1011 | size_t name_len; | |
1012 | int i, cmp = 1; | |
1013 | ||
1014 | if (name == NULL) | |
1015 | return -EINVAL; | |
1016 | ||
1017 | name_len = strlen(name); | |
1018 | entry = xs->here; | |
1019 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
1020 | cmp = name_index - ocfs2_xattr_get_type(entry); | |
1021 | if (!cmp) | |
1022 | cmp = name_len - entry->xe_name_len; | |
1023 | if (!cmp) | |
1024 | cmp = memcmp(name, (xs->base + | |
1025 | le16_to_cpu(entry->xe_name_offset)), | |
1026 | name_len); | |
1027 | if (cmp == 0) | |
1028 | break; | |
1029 | entry += 1; | |
1030 | } | |
1031 | xs->here = entry; | |
1032 | ||
1033 | return cmp ? -ENODATA : 0; | |
1034 | } | |
1035 | ||
1036 | static int ocfs2_xattr_get_value_outside(struct inode *inode, | |
589dc260 | 1037 | struct ocfs2_xattr_value_root *xv, |
cf1d6c76 TY |
1038 | void *buffer, |
1039 | size_t len) | |
1040 | { | |
1041 | u32 cpos, p_cluster, num_clusters, bpc, clusters; | |
1042 | u64 blkno; | |
1043 | int i, ret = 0; | |
1044 | size_t cplen, blocksize; | |
1045 | struct buffer_head *bh = NULL; | |
cf1d6c76 TY |
1046 | struct ocfs2_extent_list *el; |
1047 | ||
cf1d6c76 TY |
1048 | el = &xv->xr_list; |
1049 | clusters = le32_to_cpu(xv->xr_clusters); | |
1050 | bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1051 | blocksize = inode->i_sb->s_blocksize; | |
1052 | ||
1053 | cpos = 0; | |
1054 | while (cpos < clusters) { | |
1055 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1056 | &num_clusters, el, NULL); |
cf1d6c76 TY |
1057 | if (ret) { |
1058 | mlog_errno(ret); | |
1059 | goto out; | |
1060 | } | |
1061 | ||
1062 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
1063 | /* Copy ocfs2_xattr_value */ | |
1064 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1065 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1066 | &bh, NULL); | |
cf1d6c76 TY |
1067 | if (ret) { |
1068 | mlog_errno(ret); | |
1069 | goto out; | |
1070 | } | |
1071 | ||
1072 | cplen = len >= blocksize ? blocksize : len; | |
1073 | memcpy(buffer, bh->b_data, cplen); | |
1074 | len -= cplen; | |
1075 | buffer += cplen; | |
1076 | ||
1077 | brelse(bh); | |
1078 | bh = NULL; | |
1079 | if (len == 0) | |
1080 | break; | |
1081 | } | |
1082 | cpos += num_clusters; | |
1083 | } | |
1084 | out: | |
1085 | return ret; | |
1086 | } | |
1087 | ||
1088 | static int ocfs2_xattr_ibody_get(struct inode *inode, | |
1089 | int name_index, | |
1090 | const char *name, | |
1091 | void *buffer, | |
1092 | size_t buffer_size, | |
1093 | struct ocfs2_xattr_search *xs) | |
1094 | { | |
1095 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1096 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
589dc260 | 1097 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 TY |
1098 | size_t size; |
1099 | int ret = 0; | |
1100 | ||
1101 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
1102 | return -ENODATA; | |
1103 | ||
1104 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
1105 | xs->header = (struct ocfs2_xattr_header *) | |
1106 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
1107 | xs->base = (void *)xs->header; | |
1108 | xs->here = xs->header->xh_entries; | |
1109 | ||
1110 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
1111 | if (ret) | |
1112 | return ret; | |
1113 | size = le64_to_cpu(xs->here->xe_value_size); | |
1114 | if (buffer) { | |
1115 | if (size > buffer_size) | |
1116 | return -ERANGE; | |
1117 | if (ocfs2_xattr_is_local(xs->here)) { | |
1118 | memcpy(buffer, (void *)xs->base + | |
1119 | le16_to_cpu(xs->here->xe_name_offset) + | |
1120 | OCFS2_XATTR_SIZE(xs->here->xe_name_len), size); | |
1121 | } else { | |
589dc260 TM |
1122 | xv = (struct ocfs2_xattr_value_root *) |
1123 | (xs->base + le16_to_cpu( | |
1124 | xs->here->xe_name_offset) + | |
1125 | OCFS2_XATTR_SIZE(xs->here->xe_name_len)); | |
1126 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1127 | buffer, size); |
1128 | if (ret < 0) { | |
1129 | mlog_errno(ret); | |
1130 | return ret; | |
1131 | } | |
1132 | } | |
1133 | } | |
1134 | ||
1135 | return size; | |
1136 | } | |
1137 | ||
1138 | static int ocfs2_xattr_block_get(struct inode *inode, | |
1139 | int name_index, | |
1140 | const char *name, | |
1141 | void *buffer, | |
1142 | size_t buffer_size, | |
1143 | struct ocfs2_xattr_search *xs) | |
1144 | { | |
cf1d6c76 | 1145 | struct ocfs2_xattr_block *xb; |
589dc260 | 1146 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 | 1147 | size_t size; |
44d8e4e1 SM |
1148 | int ret = -ENODATA, name_offset, name_len, i; |
1149 | int uninitialized_var(block_off); | |
cf1d6c76 | 1150 | |
ba937127 JB |
1151 | xs->bucket = ocfs2_xattr_bucket_new(inode); |
1152 | if (!xs->bucket) { | |
1153 | ret = -ENOMEM; | |
1154 | mlog_errno(ret); | |
1155 | goto cleanup; | |
1156 | } | |
589dc260 | 1157 | |
54f443f4 JB |
1158 | ret = ocfs2_xattr_block_find(inode, name_index, name, xs); |
1159 | if (ret) { | |
cf1d6c76 | 1160 | mlog_errno(ret); |
cf1d6c76 TY |
1161 | goto cleanup; |
1162 | } | |
1163 | ||
6c1e183e TY |
1164 | if (xs->not_found) { |
1165 | ret = -ENODATA; | |
1166 | goto cleanup; | |
1167 | } | |
1168 | ||
54f443f4 | 1169 | xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
1170 | size = le64_to_cpu(xs->here->xe_value_size); |
1171 | if (buffer) { | |
1172 | ret = -ERANGE; | |
1173 | if (size > buffer_size) | |
1174 | goto cleanup; | |
589dc260 TM |
1175 | |
1176 | name_offset = le16_to_cpu(xs->here->xe_name_offset); | |
1177 | name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len); | |
1178 | i = xs->here - xs->header->xh_entries; | |
1179 | ||
1180 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 1181 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
ba937127 | 1182 | bucket_xh(xs->bucket), |
589dc260 TM |
1183 | i, |
1184 | &block_off, | |
1185 | &name_offset); | |
ba937127 | 1186 | xs->base = bucket_block(xs->bucket, block_off); |
589dc260 | 1187 | } |
cf1d6c76 TY |
1188 | if (ocfs2_xattr_is_local(xs->here)) { |
1189 | memcpy(buffer, (void *)xs->base + | |
589dc260 | 1190 | name_offset + name_len, size); |
cf1d6c76 | 1191 | } else { |
589dc260 TM |
1192 | xv = (struct ocfs2_xattr_value_root *) |
1193 | (xs->base + name_offset + name_len); | |
1194 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1195 | buffer, size); |
1196 | if (ret < 0) { | |
1197 | mlog_errno(ret); | |
1198 | goto cleanup; | |
1199 | } | |
1200 | } | |
1201 | } | |
1202 | ret = size; | |
1203 | cleanup: | |
ba937127 | 1204 | ocfs2_xattr_bucket_free(xs->bucket); |
cf1d6c76 | 1205 | |
54f443f4 JB |
1206 | brelse(xs->xattr_bh); |
1207 | xs->xattr_bh = NULL; | |
cf1d6c76 TY |
1208 | return ret; |
1209 | } | |
1210 | ||
4e3e9d02 TY |
1211 | int ocfs2_xattr_get_nolock(struct inode *inode, |
1212 | struct buffer_head *di_bh, | |
0030e001 TY |
1213 | int name_index, |
1214 | const char *name, | |
1215 | void *buffer, | |
1216 | size_t buffer_size) | |
cf1d6c76 TY |
1217 | { |
1218 | int ret; | |
1219 | struct ocfs2_dinode *di = NULL; | |
cf1d6c76 TY |
1220 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
1221 | struct ocfs2_xattr_search xis = { | |
1222 | .not_found = -ENODATA, | |
1223 | }; | |
1224 | struct ocfs2_xattr_search xbs = { | |
1225 | .not_found = -ENODATA, | |
1226 | }; | |
1227 | ||
8154da3d TY |
1228 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
1229 | return -EOPNOTSUPP; | |
1230 | ||
cf1d6c76 TY |
1231 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1232 | ret = -ENODATA; | |
1233 | ||
cf1d6c76 TY |
1234 | xis.inode_bh = xbs.inode_bh = di_bh; |
1235 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1236 | ||
1237 | down_read(&oi->ip_xattr_sem); | |
1238 | ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer, | |
1239 | buffer_size, &xis); | |
6c1e183e | 1240 | if (ret == -ENODATA && di->i_xattr_loc) |
cf1d6c76 TY |
1241 | ret = ocfs2_xattr_block_get(inode, name_index, name, buffer, |
1242 | buffer_size, &xbs); | |
1243 | up_read(&oi->ip_xattr_sem); | |
4e3e9d02 TY |
1244 | |
1245 | return ret; | |
1246 | } | |
1247 | ||
1248 | /* ocfs2_xattr_get() | |
1249 | * | |
1250 | * Copy an extended attribute into the buffer provided. | |
1251 | * Buffer is NULL to compute the size of buffer required. | |
1252 | */ | |
1253 | static int ocfs2_xattr_get(struct inode *inode, | |
1254 | int name_index, | |
1255 | const char *name, | |
1256 | void *buffer, | |
1257 | size_t buffer_size) | |
1258 | { | |
1259 | int ret; | |
1260 | struct buffer_head *di_bh = NULL; | |
1261 | ||
1262 | ret = ocfs2_inode_lock(inode, &di_bh, 0); | |
1263 | if (ret < 0) { | |
1264 | mlog_errno(ret); | |
1265 | return ret; | |
1266 | } | |
1267 | ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index, | |
1268 | name, buffer, buffer_size); | |
1269 | ||
cf1d6c76 TY |
1270 | ocfs2_inode_unlock(inode, 0); |
1271 | ||
1272 | brelse(di_bh); | |
1273 | ||
1274 | return ret; | |
1275 | } | |
1276 | ||
1277 | static int __ocfs2_xattr_set_value_outside(struct inode *inode, | |
85db90e7 | 1278 | handle_t *handle, |
492a8a33 | 1279 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1280 | const void *value, |
1281 | int value_len) | |
1282 | { | |
71d548a6 | 1283 | int ret = 0, i, cp_len; |
cf1d6c76 TY |
1284 | u16 blocksize = inode->i_sb->s_blocksize; |
1285 | u32 p_cluster, num_clusters; | |
1286 | u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1287 | u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len); | |
1288 | u64 blkno; | |
1289 | struct buffer_head *bh = NULL; | |
492a8a33 TM |
1290 | unsigned int ext_flags; |
1291 | struct ocfs2_xattr_value_root *xv = vb->vb_xv; | |
cf1d6c76 TY |
1292 | |
1293 | BUG_ON(clusters > le32_to_cpu(xv->xr_clusters)); | |
1294 | ||
cf1d6c76 TY |
1295 | while (cpos < clusters) { |
1296 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1297 | &num_clusters, &xv->xr_list, |
492a8a33 | 1298 | &ext_flags); |
cf1d6c76 TY |
1299 | if (ret) { |
1300 | mlog_errno(ret); | |
85db90e7 | 1301 | goto out; |
cf1d6c76 TY |
1302 | } |
1303 | ||
492a8a33 TM |
1304 | BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); |
1305 | ||
cf1d6c76 TY |
1306 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); |
1307 | ||
1308 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1309 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1310 | &bh, NULL); | |
cf1d6c76 TY |
1311 | if (ret) { |
1312 | mlog_errno(ret); | |
85db90e7 | 1313 | goto out; |
cf1d6c76 TY |
1314 | } |
1315 | ||
1316 | ret = ocfs2_journal_access(handle, | |
0cf2f763 | 1317 | INODE_CACHE(inode), |
cf1d6c76 TY |
1318 | bh, |
1319 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1320 | if (ret < 0) { | |
1321 | mlog_errno(ret); | |
85db90e7 | 1322 | goto out; |
cf1d6c76 TY |
1323 | } |
1324 | ||
1325 | cp_len = value_len > blocksize ? blocksize : value_len; | |
1326 | memcpy(bh->b_data, value, cp_len); | |
1327 | value_len -= cp_len; | |
1328 | value += cp_len; | |
1329 | if (cp_len < blocksize) | |
1330 | memset(bh->b_data + cp_len, 0, | |
1331 | blocksize - cp_len); | |
1332 | ||
1333 | ret = ocfs2_journal_dirty(handle, bh); | |
1334 | if (ret < 0) { | |
1335 | mlog_errno(ret); | |
85db90e7 | 1336 | goto out; |
cf1d6c76 TY |
1337 | } |
1338 | brelse(bh); | |
1339 | bh = NULL; | |
1340 | ||
1341 | /* | |
1342 | * XXX: do we need to empty all the following | |
1343 | * blocks in this cluster? | |
1344 | */ | |
1345 | if (!value_len) | |
1346 | break; | |
1347 | } | |
1348 | cpos += num_clusters; | |
1349 | } | |
cf1d6c76 TY |
1350 | out: |
1351 | brelse(bh); | |
1352 | ||
1353 | return ret; | |
1354 | } | |
1355 | ||
1356 | static int ocfs2_xattr_cleanup(struct inode *inode, | |
85db90e7 | 1357 | handle_t *handle, |
cf1d6c76 TY |
1358 | struct ocfs2_xattr_info *xi, |
1359 | struct ocfs2_xattr_search *xs, | |
512620f4 | 1360 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1361 | size_t offs) |
1362 | { | |
cf1d6c76 TY |
1363 | int ret = 0; |
1364 | size_t name_len = strlen(xi->name); | |
1365 | void *val = xs->base + offs; | |
1366 | size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
1367 | ||
0cf2f763 | 1368 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
512620f4 | 1369 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1370 | if (ret) { |
1371 | mlog_errno(ret); | |
85db90e7 | 1372 | goto out; |
cf1d6c76 TY |
1373 | } |
1374 | /* Decrease xattr count */ | |
1375 | le16_add_cpu(&xs->header->xh_count, -1); | |
1376 | /* Remove the xattr entry and tree root which has already be set*/ | |
1377 | memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry)); | |
1378 | memset(val, 0, size); | |
1379 | ||
512620f4 | 1380 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
cf1d6c76 TY |
1381 | if (ret < 0) |
1382 | mlog_errno(ret); | |
cf1d6c76 TY |
1383 | out: |
1384 | return ret; | |
1385 | } | |
1386 | ||
1387 | static int ocfs2_xattr_update_entry(struct inode *inode, | |
85db90e7 | 1388 | handle_t *handle, |
cf1d6c76 TY |
1389 | struct ocfs2_xattr_info *xi, |
1390 | struct ocfs2_xattr_search *xs, | |
0c748e95 | 1391 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1392 | size_t offs) |
1393 | { | |
85db90e7 | 1394 | int ret; |
cf1d6c76 | 1395 | |
0cf2f763 | 1396 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
0c748e95 | 1397 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1398 | if (ret) { |
1399 | mlog_errno(ret); | |
85db90e7 | 1400 | goto out; |
cf1d6c76 TY |
1401 | } |
1402 | ||
1403 | xs->here->xe_name_offset = cpu_to_le16(offs); | |
1404 | xs->here->xe_value_size = cpu_to_le64(xi->value_len); | |
1405 | if (xi->value_len <= OCFS2_XATTR_INLINE_SIZE) | |
1406 | ocfs2_xattr_set_local(xs->here, 1); | |
1407 | else | |
1408 | ocfs2_xattr_set_local(xs->here, 0); | |
1409 | ocfs2_xattr_hash_entry(inode, xs->header, xs->here); | |
1410 | ||
0c748e95 | 1411 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
cf1d6c76 TY |
1412 | if (ret < 0) |
1413 | mlog_errno(ret); | |
cf1d6c76 TY |
1414 | out: |
1415 | return ret; | |
1416 | } | |
1417 | ||
1418 | /* | |
1419 | * ocfs2_xattr_set_value_outside() | |
1420 | * | |
1421 | * Set large size value in B tree. | |
1422 | */ | |
1423 | static int ocfs2_xattr_set_value_outside(struct inode *inode, | |
1424 | struct ocfs2_xattr_info *xi, | |
1425 | struct ocfs2_xattr_search *xs, | |
78f30c31 | 1426 | struct ocfs2_xattr_set_ctxt *ctxt, |
512620f4 | 1427 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1428 | size_t offs) |
1429 | { | |
1430 | size_t name_len = strlen(xi->name); | |
1431 | void *val = xs->base + offs; | |
1432 | struct ocfs2_xattr_value_root *xv = NULL; | |
1433 | size_t size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
1434 | int ret = 0; | |
1435 | ||
1436 | memset(val, 0, size); | |
1437 | memcpy(val, xi->name, name_len); | |
1438 | xv = (struct ocfs2_xattr_value_root *) | |
1439 | (val + OCFS2_XATTR_SIZE(name_len)); | |
1440 | xv->xr_clusters = 0; | |
1441 | xv->xr_last_eb_blk = 0; | |
1442 | xv->xr_list.l_tree_depth = 0; | |
1443 | xv->xr_list.l_count = cpu_to_le16(1); | |
1444 | xv->xr_list.l_next_free_rec = 0; | |
512620f4 | 1445 | vb->vb_xv = xv; |
cf1d6c76 | 1446 | |
512620f4 | 1447 | ret = ocfs2_xattr_value_truncate(inode, vb, xi->value_len, ctxt); |
cf1d6c76 TY |
1448 | if (ret < 0) { |
1449 | mlog_errno(ret); | |
1450 | return ret; | |
1451 | } | |
512620f4 | 1452 | ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, vb, offs); |
cf1d6c76 TY |
1453 | if (ret < 0) { |
1454 | mlog_errno(ret); | |
1455 | return ret; | |
1456 | } | |
492a8a33 | 1457 | ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, vb, |
85db90e7 | 1458 | xi->value, xi->value_len); |
cf1d6c76 TY |
1459 | if (ret < 0) |
1460 | mlog_errno(ret); | |
1461 | ||
1462 | return ret; | |
1463 | } | |
1464 | ||
11179f2c JB |
1465 | /* |
1466 | * Wipe the name+value pair and allow the storage to reclaim it. This | |
1467 | * must be followed by either removal of the entry or a call to | |
1468 | * ocfs2_xa_add_namevalue(). | |
1469 | */ | |
1470 | static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1471 | { | |
1472 | loc->xl_ops->xlo_wipe_namevalue(loc); | |
1473 | } | |
1474 | ||
1475 | static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc, | |
1476 | int offset) | |
1477 | { | |
1478 | BUG_ON(offset >= loc->xl_size); | |
1479 | return (char *)loc->xl_header + offset; | |
1480 | } | |
1481 | ||
1482 | /* | |
1483 | * Block storage for xattrs keeps the name+value pairs compacted. When | |
1484 | * we remove one, we have to shift any that preceded it towards the end. | |
1485 | */ | |
1486 | static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1487 | { | |
1488 | int i, offset; | |
1489 | int namevalue_offset, first_namevalue_offset, namevalue_size; | |
1490 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1491 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1492 | u64 value_size = le64_to_cpu(entry->xe_value_size); | |
1493 | int count = le16_to_cpu(xh->xh_count); | |
1494 | ||
1495 | namevalue_offset = le16_to_cpu(entry->xe_name_offset); | |
1496 | namevalue_size = OCFS2_XATTR_SIZE(entry->xe_name_len); | |
1497 | if (value_size > OCFS2_XATTR_INLINE_SIZE) | |
1498 | namevalue_size += OCFS2_XATTR_ROOT_SIZE; | |
1499 | else | |
1500 | namevalue_size += OCFS2_XATTR_SIZE(value_size); | |
1501 | ||
1502 | for (i = 0, first_namevalue_offset = loc->xl_size; | |
1503 | i < count; i++) { | |
1504 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1505 | if (offset < first_namevalue_offset) | |
1506 | first_namevalue_offset = offset; | |
1507 | } | |
1508 | ||
1509 | /* Shift the name+value pairs */ | |
1510 | memmove((char *)xh + first_namevalue_offset + namevalue_size, | |
1511 | (char *)xh + first_namevalue_offset, | |
1512 | namevalue_offset - first_namevalue_offset); | |
1513 | memset((char *)xh + first_namevalue_offset, 0, namevalue_size); | |
1514 | ||
1515 | /* Now tell xh->xh_entries about it */ | |
1516 | for (i = 0; i < count; i++) { | |
1517 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1518 | if (offset < namevalue_offset) | |
1519 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, | |
1520 | namevalue_size); | |
1521 | } | |
1522 | ||
1523 | /* | |
1524 | * Note that we don't update xh_free_start or xh_name_value_len | |
1525 | * because they're not used in block-stored xattrs. | |
1526 | */ | |
1527 | } | |
1528 | ||
1529 | /* | |
1530 | * Operations for xattrs stored in blocks. This includes inline inode | |
1531 | * storage and unindexed ocfs2_xattr_blocks. | |
1532 | */ | |
1533 | static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = { | |
1534 | .xlo_offset_pointer = ocfs2_xa_block_offset_pointer, | |
1535 | .xlo_wipe_namevalue = ocfs2_xa_block_wipe_namevalue, | |
1536 | }; | |
1537 | ||
1538 | static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, | |
1539 | int offset) | |
1540 | { | |
1541 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1542 | int block, block_offset; | |
1543 | ||
1544 | BUG_ON(offset >= OCFS2_XATTR_BUCKET_SIZE); | |
1545 | ||
1546 | /* The header is at the front of the bucket */ | |
1547 | block = offset >> bucket->bu_inode->i_sb->s_blocksize_bits; | |
1548 | block_offset = offset % bucket->bu_inode->i_sb->s_blocksize; | |
1549 | ||
1550 | return bucket_block(bucket, block) + block_offset; | |
1551 | } | |
1552 | ||
1553 | static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1554 | { | |
1555 | int namevalue_size; | |
1556 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1557 | u64 value_size = le64_to_cpu(entry->xe_value_size); | |
1558 | ||
1559 | namevalue_size = OCFS2_XATTR_SIZE(entry->xe_name_len); | |
1560 | if (value_size > OCFS2_XATTR_INLINE_SIZE) | |
1561 | namevalue_size += OCFS2_XATTR_ROOT_SIZE; | |
1562 | else | |
1563 | namevalue_size += OCFS2_XATTR_SIZE(value_size); | |
1564 | ||
1565 | le16_add_cpu(&loc->xl_header->xh_name_value_len, -namevalue_size); | |
1566 | } | |
1567 | ||
1568 | /* Operations for xattrs stored in buckets. */ | |
1569 | static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = { | |
1570 | .xlo_offset_pointer = ocfs2_xa_bucket_offset_pointer, | |
1571 | .xlo_wipe_namevalue = ocfs2_xa_bucket_wipe_namevalue, | |
1572 | }; | |
1573 | ||
1574 | static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc) | |
1575 | { | |
bde1e540 JB |
1576 | int index, count; |
1577 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1578 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1579 | ||
11179f2c | 1580 | ocfs2_xa_wipe_namevalue(loc); |
bde1e540 JB |
1581 | loc->xl_entry = NULL; |
1582 | ||
1583 | le16_add_cpu(&xh->xh_count, -1); | |
1584 | count = le16_to_cpu(xh->xh_count); | |
1585 | ||
1586 | /* | |
1587 | * Only zero out the entry if there are more remaining. This is | |
1588 | * important for an empty bucket, as it keeps track of the | |
1589 | * bucket's hash value. It doesn't hurt empty block storage. | |
1590 | */ | |
1591 | if (count) { | |
1592 | index = ((char *)entry - (char *)&xh->xh_entries) / | |
1593 | sizeof(struct ocfs2_xattr_entry); | |
1594 | memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1], | |
1595 | (count - index) * sizeof(struct ocfs2_xattr_entry)); | |
1596 | memset(&xh->xh_entries[count], 0, | |
1597 | sizeof(struct ocfs2_xattr_entry)); | |
1598 | } | |
11179f2c JB |
1599 | } |
1600 | ||
1601 | static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc, | |
1602 | struct inode *inode, | |
1603 | struct buffer_head *bh, | |
1604 | struct ocfs2_xattr_entry *entry) | |
1605 | { | |
1606 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
1607 | ||
1608 | loc->xl_ops = &ocfs2_xa_block_loc_ops; | |
1609 | loc->xl_storage = bh; | |
1610 | loc->xl_entry = entry; | |
1611 | ||
1612 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
1613 | loc->xl_size = le16_to_cpu(di->i_xattr_inline_size); | |
1614 | else { | |
1615 | BUG_ON(entry); | |
1616 | loc->xl_size = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
1617 | } | |
1618 | loc->xl_header = | |
1619 | (struct ocfs2_xattr_header *)(bh->b_data + bh->b_size - | |
1620 | loc->xl_size); | |
1621 | } | |
1622 | ||
1623 | static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc, | |
1624 | struct buffer_head *bh, | |
1625 | struct ocfs2_xattr_entry *entry) | |
1626 | { | |
1627 | struct ocfs2_xattr_block *xb = | |
1628 | (struct ocfs2_xattr_block *)bh->b_data; | |
1629 | ||
1630 | BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED); | |
1631 | ||
1632 | loc->xl_ops = &ocfs2_xa_block_loc_ops; | |
1633 | loc->xl_storage = bh; | |
1634 | loc->xl_header = &(xb->xb_attrs.xb_header); | |
1635 | loc->xl_entry = entry; | |
1636 | loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block, | |
1637 | xb_attrs.xb_header); | |
1638 | } | |
1639 | ||
1640 | static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc, | |
1641 | struct ocfs2_xattr_bucket *bucket, | |
1642 | struct ocfs2_xattr_entry *entry) | |
1643 | { | |
1644 | loc->xl_ops = &ocfs2_xa_bucket_loc_ops; | |
1645 | loc->xl_storage = bucket; | |
1646 | loc->xl_header = bucket_xh(bucket); | |
1647 | loc->xl_entry = entry; | |
1648 | loc->xl_size = OCFS2_XATTR_BUCKET_SIZE; | |
1649 | } | |
1650 | ||
cf1d6c76 TY |
1651 | /* |
1652 | * ocfs2_xattr_set_entry_local() | |
1653 | * | |
1654 | * Set, replace or remove extended attribute in local. | |
1655 | */ | |
1656 | static void ocfs2_xattr_set_entry_local(struct inode *inode, | |
1657 | struct ocfs2_xattr_info *xi, | |
1658 | struct ocfs2_xattr_search *xs, | |
1659 | struct ocfs2_xattr_entry *last, | |
1660 | size_t min_offs) | |
1661 | { | |
1662 | size_t name_len = strlen(xi->name); | |
11179f2c | 1663 | struct ocfs2_xa_loc loc; |
cf1d6c76 | 1664 | |
11179f2c JB |
1665 | if (xs->xattr_bh == xs->inode_bh) |
1666 | ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh, | |
1667 | xs->not_found ? NULL : xs->here); | |
1668 | else | |
1669 | ocfs2_init_xattr_block_xa_loc(&loc, xs->xattr_bh, | |
1670 | xs->not_found ? NULL : xs->here); | |
cf1d6c76 TY |
1671 | if (xi->value && xs->not_found) { |
1672 | /* Insert the new xattr entry. */ | |
1673 | le16_add_cpu(&xs->header->xh_count, 1); | |
1674 | ocfs2_xattr_set_type(last, xi->name_index); | |
1675 | ocfs2_xattr_set_local(last, 1); | |
1676 | last->xe_name_len = name_len; | |
1677 | } else { | |
1678 | void *first_val; | |
1679 | void *val; | |
1680 | size_t offs, size; | |
1681 | ||
1682 | first_val = xs->base + min_offs; | |
1683 | offs = le16_to_cpu(xs->here->xe_name_offset); | |
1684 | val = xs->base + offs; | |
1685 | ||
1686 | if (le64_to_cpu(xs->here->xe_value_size) > | |
1687 | OCFS2_XATTR_INLINE_SIZE) | |
1688 | size = OCFS2_XATTR_SIZE(name_len) + | |
1689 | OCFS2_XATTR_ROOT_SIZE; | |
1690 | else | |
1691 | size = OCFS2_XATTR_SIZE(name_len) + | |
1692 | OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); | |
1693 | ||
1694 | if (xi->value && size == OCFS2_XATTR_SIZE(name_len) + | |
1695 | OCFS2_XATTR_SIZE(xi->value_len)) { | |
1696 | /* The old and the new value have the | |
1697 | same size. Just replace the value. */ | |
1698 | ocfs2_xattr_set_local(xs->here, 1); | |
1699 | xs->here->xe_value_size = cpu_to_le64(xi->value_len); | |
1700 | /* Clear value bytes. */ | |
1701 | memset(val + OCFS2_XATTR_SIZE(name_len), | |
1702 | 0, | |
1703 | OCFS2_XATTR_SIZE(xi->value_len)); | |
1704 | memcpy(val + OCFS2_XATTR_SIZE(name_len), | |
1705 | xi->value, | |
1706 | xi->value_len); | |
1707 | return; | |
1708 | } | |
11179f2c | 1709 | |
bde1e540 JB |
1710 | if (!xi->value) |
1711 | ocfs2_xa_remove_entry(&loc); | |
1712 | else | |
1713 | ocfs2_xa_wipe_namevalue(&loc); | |
cf1d6c76 TY |
1714 | |
1715 | min_offs += size; | |
cf1d6c76 TY |
1716 | } |
1717 | if (xi->value) { | |
1718 | /* Insert the new name+value. */ | |
1719 | size_t size = OCFS2_XATTR_SIZE(name_len) + | |
1720 | OCFS2_XATTR_SIZE(xi->value_len); | |
1721 | void *val = xs->base + min_offs - size; | |
1722 | ||
1723 | xs->here->xe_name_offset = cpu_to_le16(min_offs - size); | |
1724 | memset(val, 0, size); | |
1725 | memcpy(val, xi->name, name_len); | |
1726 | memcpy(val + OCFS2_XATTR_SIZE(name_len), | |
1727 | xi->value, | |
1728 | xi->value_len); | |
1729 | xs->here->xe_value_size = cpu_to_le64(xi->value_len); | |
1730 | ocfs2_xattr_set_local(xs->here, 1); | |
1731 | ocfs2_xattr_hash_entry(inode, xs->header, xs->here); | |
1732 | } | |
1733 | ||
1734 | return; | |
1735 | } | |
1736 | ||
1737 | /* | |
1738 | * ocfs2_xattr_set_entry() | |
1739 | * | |
1740 | * Set extended attribute entry into inode or block. | |
1741 | * | |
1742 | * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE, | |
1743 | * We first insert tree root(ocfs2_xattr_value_root) with set_entry_local(), | |
1744 | * then set value in B tree with set_value_outside(). | |
1745 | */ | |
1746 | static int ocfs2_xattr_set_entry(struct inode *inode, | |
1747 | struct ocfs2_xattr_info *xi, | |
1748 | struct ocfs2_xattr_search *xs, | |
78f30c31 | 1749 | struct ocfs2_xattr_set_ctxt *ctxt, |
cf1d6c76 TY |
1750 | int flag) |
1751 | { | |
1752 | struct ocfs2_xattr_entry *last; | |
1753 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1754 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
1755 | size_t min_offs = xs->end - xs->base, name_len = strlen(xi->name); | |
1756 | size_t size_l = 0; | |
85db90e7 | 1757 | handle_t *handle = ctxt->handle; |
cf1d6c76 TY |
1758 | int free, i, ret; |
1759 | struct ocfs2_xattr_info xi_l = { | |
1760 | .name_index = xi->name_index, | |
1761 | .name = xi->name, | |
1762 | .value = xi->value, | |
1763 | .value_len = xi->value_len, | |
1764 | }; | |
512620f4 JB |
1765 | struct ocfs2_xattr_value_buf vb = { |
1766 | .vb_bh = xs->xattr_bh, | |
1767 | .vb_access = ocfs2_journal_access_di, | |
1768 | }; | |
1769 | ||
1770 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
1771 | BUG_ON(xs->xattr_bh == xs->inode_bh); | |
1772 | vb.vb_access = ocfs2_journal_access_xb; | |
1773 | } else | |
1774 | BUG_ON(xs->xattr_bh != xs->inode_bh); | |
cf1d6c76 TY |
1775 | |
1776 | /* Compute min_offs, last and free space. */ | |
1777 | last = xs->header->xh_entries; | |
1778 | ||
1779 | for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) { | |
1780 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
1781 | if (offs < min_offs) | |
1782 | min_offs = offs; | |
1783 | last += 1; | |
1784 | } | |
1785 | ||
4442f518 | 1786 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
cf1d6c76 | 1787 | if (free < 0) |
b37c4d84 | 1788 | return -EIO; |
cf1d6c76 TY |
1789 | |
1790 | if (!xs->not_found) { | |
1791 | size_t size = 0; | |
1792 | if (ocfs2_xattr_is_local(xs->here)) | |
1793 | size = OCFS2_XATTR_SIZE(name_len) + | |
1794 | OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); | |
1795 | else | |
1796 | size = OCFS2_XATTR_SIZE(name_len) + | |
1797 | OCFS2_XATTR_ROOT_SIZE; | |
1798 | free += (size + sizeof(struct ocfs2_xattr_entry)); | |
1799 | } | |
1800 | /* Check free space in inode or block */ | |
1801 | if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
1802 | if (free < sizeof(struct ocfs2_xattr_entry) + | |
1803 | OCFS2_XATTR_SIZE(name_len) + | |
1804 | OCFS2_XATTR_ROOT_SIZE) { | |
1805 | ret = -ENOSPC; | |
1806 | goto out; | |
1807 | } | |
1808 | size_l = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
1809 | xi_l.value = (void *)&def_xv; | |
1810 | xi_l.value_len = OCFS2_XATTR_ROOT_SIZE; | |
1811 | } else if (xi->value) { | |
1812 | if (free < sizeof(struct ocfs2_xattr_entry) + | |
1813 | OCFS2_XATTR_SIZE(name_len) + | |
1814 | OCFS2_XATTR_SIZE(xi->value_len)) { | |
1815 | ret = -ENOSPC; | |
1816 | goto out; | |
1817 | } | |
1818 | } | |
1819 | ||
1820 | if (!xs->not_found) { | |
1821 | /* For existing extended attribute */ | |
1822 | size_t size = OCFS2_XATTR_SIZE(name_len) + | |
1823 | OCFS2_XATTR_SIZE(le64_to_cpu(xs->here->xe_value_size)); | |
1824 | size_t offs = le16_to_cpu(xs->here->xe_name_offset); | |
1825 | void *val = xs->base + offs; | |
1826 | ||
1827 | if (ocfs2_xattr_is_local(xs->here) && size == size_l) { | |
1828 | /* Replace existing local xattr with tree root */ | |
1829 | ret = ocfs2_xattr_set_value_outside(inode, xi, xs, | |
512620f4 | 1830 | ctxt, &vb, offs); |
cf1d6c76 TY |
1831 | if (ret < 0) |
1832 | mlog_errno(ret); | |
1833 | goto out; | |
1834 | } else if (!ocfs2_xattr_is_local(xs->here)) { | |
1835 | /* For existing xattr which has value outside */ | |
512620f4 JB |
1836 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
1837 | (val + OCFS2_XATTR_SIZE(name_len)); | |
cf1d6c76 TY |
1838 | |
1839 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
1840 | /* | |
1841 | * If new value need set outside also, | |
1842 | * first truncate old value to new value, | |
1843 | * then set new value with set_value_outside(). | |
1844 | */ | |
1845 | ret = ocfs2_xattr_value_truncate(inode, | |
b3e5d379 | 1846 | &vb, |
78f30c31 TM |
1847 | xi->value_len, |
1848 | ctxt); | |
cf1d6c76 TY |
1849 | if (ret < 0) { |
1850 | mlog_errno(ret); | |
1851 | goto out; | |
1852 | } | |
1853 | ||
85db90e7 TM |
1854 | ret = ocfs2_xattr_update_entry(inode, |
1855 | handle, | |
1856 | xi, | |
1857 | xs, | |
0c748e95 | 1858 | &vb, |
85db90e7 | 1859 | offs); |
cf1d6c76 TY |
1860 | if (ret < 0) { |
1861 | mlog_errno(ret); | |
1862 | goto out; | |
1863 | } | |
1864 | ||
85db90e7 TM |
1865 | ret = __ocfs2_xattr_set_value_outside(inode, |
1866 | handle, | |
492a8a33 | 1867 | &vb, |
85db90e7 TM |
1868 | xi->value, |
1869 | xi->value_len); | |
cf1d6c76 TY |
1870 | if (ret < 0) |
1871 | mlog_errno(ret); | |
1872 | goto out; | |
1873 | } else { | |
1874 | /* | |
1875 | * If new value need set in local, | |
1876 | * just trucate old value to zero. | |
1877 | */ | |
1878 | ret = ocfs2_xattr_value_truncate(inode, | |
b3e5d379 | 1879 | &vb, |
85db90e7 TM |
1880 | 0, |
1881 | ctxt); | |
cf1d6c76 TY |
1882 | if (ret < 0) |
1883 | mlog_errno(ret); | |
1884 | } | |
1885 | } | |
1886 | } | |
1887 | ||
0cf2f763 | 1888 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), xs->inode_bh, |
512620f4 | 1889 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1890 | if (ret) { |
1891 | mlog_errno(ret); | |
85db90e7 | 1892 | goto out; |
cf1d6c76 TY |
1893 | } |
1894 | ||
1895 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
0cf2f763 | 1896 | ret = vb.vb_access(handle, INODE_CACHE(inode), vb.vb_bh, |
512620f4 | 1897 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1898 | if (ret) { |
1899 | mlog_errno(ret); | |
85db90e7 | 1900 | goto out; |
cf1d6c76 TY |
1901 | } |
1902 | } | |
1903 | ||
1904 | /* | |
1905 | * Set value in local, include set tree root in local. | |
1906 | * This is the first step for value size >INLINE_SIZE. | |
1907 | */ | |
1908 | ocfs2_xattr_set_entry_local(inode, &xi_l, xs, last, min_offs); | |
1909 | ||
1910 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
1911 | ret = ocfs2_journal_dirty(handle, xs->xattr_bh); | |
1912 | if (ret < 0) { | |
1913 | mlog_errno(ret); | |
85db90e7 | 1914 | goto out; |
cf1d6c76 TY |
1915 | } |
1916 | } | |
1917 | ||
1918 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) && | |
1919 | (flag & OCFS2_INLINE_XATTR_FL)) { | |
1920 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
1921 | unsigned int xattrsize = osb->s_xattr_inline_size; | |
1922 | ||
1923 | /* | |
1924 | * Adjust extent record count or inline data size | |
1925 | * to reserve space for extended attribute. | |
1926 | */ | |
1927 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
1928 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
1929 | le16_add_cpu(&idata->id_count, -xattrsize); | |
1930 | } else if (!(ocfs2_inode_is_fast_symlink(inode))) { | |
1931 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
1932 | le16_add_cpu(&el->l_count, -(xattrsize / | |
1933 | sizeof(struct ocfs2_extent_rec))); | |
1934 | } | |
1935 | di->i_xattr_inline_size = cpu_to_le16(xattrsize); | |
1936 | } | |
1937 | /* Update xattr flag */ | |
1938 | spin_lock(&oi->ip_lock); | |
1939 | oi->ip_dyn_features |= flag; | |
1940 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
1941 | spin_unlock(&oi->ip_lock); | |
cf1d6c76 TY |
1942 | |
1943 | ret = ocfs2_journal_dirty(handle, xs->inode_bh); | |
1944 | if (ret < 0) | |
1945 | mlog_errno(ret); | |
1946 | ||
cf1d6c76 TY |
1947 | if (!ret && xi->value_len > OCFS2_XATTR_INLINE_SIZE) { |
1948 | /* | |
1949 | * Set value outside in B tree. | |
1950 | * This is the second step for value size > INLINE_SIZE. | |
1951 | */ | |
1952 | size_t offs = le16_to_cpu(xs->here->xe_name_offset); | |
512620f4 JB |
1953 | ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt, |
1954 | &vb, offs); | |
cf1d6c76 TY |
1955 | if (ret < 0) { |
1956 | int ret2; | |
1957 | ||
1958 | mlog_errno(ret); | |
1959 | /* | |
1960 | * If set value outside failed, we have to clean | |
1961 | * the junk tree root we have already set in local. | |
1962 | */ | |
85db90e7 | 1963 | ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle, |
512620f4 | 1964 | xi, xs, &vb, offs); |
cf1d6c76 TY |
1965 | if (ret2 < 0) |
1966 | mlog_errno(ret2); | |
1967 | } | |
1968 | } | |
1969 | out: | |
1970 | return ret; | |
cf1d6c76 TY |
1971 | } |
1972 | ||
ce9c5a54 TM |
1973 | /* |
1974 | * In xattr remove, if it is stored outside and refcounted, we may have | |
1975 | * the chance to split the refcount tree. So need the allocators. | |
1976 | */ | |
1977 | static int ocfs2_lock_xattr_remove_allocators(struct inode *inode, | |
1978 | struct ocfs2_xattr_value_root *xv, | |
1979 | struct ocfs2_caching_info *ref_ci, | |
1980 | struct buffer_head *ref_root_bh, | |
1981 | struct ocfs2_alloc_context **meta_ac, | |
1982 | int *ref_credits) | |
1983 | { | |
1984 | int ret, meta_add = 0; | |
1985 | u32 p_cluster, num_clusters; | |
1986 | unsigned int ext_flags; | |
1987 | ||
1988 | *ref_credits = 0; | |
1989 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
1990 | &num_clusters, | |
1991 | &xv->xr_list, | |
1992 | &ext_flags); | |
1993 | if (ret) { | |
1994 | mlog_errno(ret); | |
1995 | goto out; | |
1996 | } | |
1997 | ||
1998 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
1999 | goto out; | |
2000 | ||
2001 | ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci, | |
2002 | ref_root_bh, xv, | |
2003 | &meta_add, ref_credits); | |
2004 | if (ret) { | |
2005 | mlog_errno(ret); | |
2006 | goto out; | |
2007 | } | |
2008 | ||
2009 | ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), | |
2010 | meta_add, meta_ac); | |
2011 | if (ret) | |
2012 | mlog_errno(ret); | |
2013 | ||
2014 | out: | |
2015 | return ret; | |
2016 | } | |
2017 | ||
cf1d6c76 | 2018 | static int ocfs2_remove_value_outside(struct inode*inode, |
4311901d | 2019 | struct ocfs2_xattr_value_buf *vb, |
ce9c5a54 TM |
2020 | struct ocfs2_xattr_header *header, |
2021 | struct ocfs2_caching_info *ref_ci, | |
2022 | struct buffer_head *ref_root_bh) | |
cf1d6c76 | 2023 | { |
ce9c5a54 | 2024 | int ret = 0, i, ref_credits; |
78f30c31 TM |
2025 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
2026 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; | |
ce9c5a54 | 2027 | void *val; |
78f30c31 TM |
2028 | |
2029 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
cf1d6c76 TY |
2030 | |
2031 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
2032 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
2033 | ||
ce9c5a54 TM |
2034 | if (ocfs2_xattr_is_local(entry)) |
2035 | continue; | |
cf1d6c76 | 2036 | |
ce9c5a54 TM |
2037 | val = (void *)header + |
2038 | le16_to_cpu(entry->xe_name_offset); | |
2039 | vb->vb_xv = (struct ocfs2_xattr_value_root *) | |
2040 | (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); | |
2041 | ||
2042 | ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv, | |
2043 | ref_ci, ref_root_bh, | |
2044 | &ctxt.meta_ac, | |
2045 | &ref_credits); | |
2046 | ||
2047 | ctxt.handle = ocfs2_start_trans(osb, ref_credits + | |
2048 | ocfs2_remove_extent_credits(osb->sb)); | |
2049 | if (IS_ERR(ctxt.handle)) { | |
2050 | ret = PTR_ERR(ctxt.handle); | |
2051 | mlog_errno(ret); | |
2052 | break; | |
2053 | } | |
2054 | ||
2055 | ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); | |
2056 | if (ret < 0) { | |
2057 | mlog_errno(ret); | |
2058 | break; | |
2059 | } | |
2060 | ||
2061 | ocfs2_commit_trans(osb, ctxt.handle); | |
2062 | if (ctxt.meta_ac) { | |
2063 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
2064 | ctxt.meta_ac = NULL; | |
cf1d6c76 TY |
2065 | } |
2066 | } | |
2067 | ||
ce9c5a54 TM |
2068 | if (ctxt.meta_ac) |
2069 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
2070 | ocfs2_schedule_truncate_log_flush(osb, 1); |
2071 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
cf1d6c76 TY |
2072 | return ret; |
2073 | } | |
2074 | ||
2075 | static int ocfs2_xattr_ibody_remove(struct inode *inode, | |
ce9c5a54 TM |
2076 | struct buffer_head *di_bh, |
2077 | struct ocfs2_caching_info *ref_ci, | |
2078 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2079 | { |
2080 | ||
2081 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2082 | struct ocfs2_xattr_header *header; | |
2083 | int ret; | |
4311901d JB |
2084 | struct ocfs2_xattr_value_buf vb = { |
2085 | .vb_bh = di_bh, | |
2086 | .vb_access = ocfs2_journal_access_di, | |
2087 | }; | |
cf1d6c76 TY |
2088 | |
2089 | header = (struct ocfs2_xattr_header *) | |
2090 | ((void *)di + inode->i_sb->s_blocksize - | |
2091 | le16_to_cpu(di->i_xattr_inline_size)); | |
2092 | ||
ce9c5a54 TM |
2093 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2094 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2095 | |
2096 | return ret; | |
2097 | } | |
2098 | ||
ce9c5a54 TM |
2099 | struct ocfs2_rm_xattr_bucket_para { |
2100 | struct ocfs2_caching_info *ref_ci; | |
2101 | struct buffer_head *ref_root_bh; | |
2102 | }; | |
2103 | ||
cf1d6c76 | 2104 | static int ocfs2_xattr_block_remove(struct inode *inode, |
ce9c5a54 TM |
2105 | struct buffer_head *blk_bh, |
2106 | struct ocfs2_caching_info *ref_ci, | |
2107 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2108 | { |
2109 | struct ocfs2_xattr_block *xb; | |
cf1d6c76 | 2110 | int ret = 0; |
4311901d JB |
2111 | struct ocfs2_xattr_value_buf vb = { |
2112 | .vb_bh = blk_bh, | |
2113 | .vb_access = ocfs2_journal_access_xb, | |
2114 | }; | |
ce9c5a54 TM |
2115 | struct ocfs2_rm_xattr_bucket_para args = { |
2116 | .ref_ci = ref_ci, | |
2117 | .ref_root_bh = ref_root_bh, | |
2118 | }; | |
cf1d6c76 TY |
2119 | |
2120 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; | |
a3944256 TM |
2121 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
2122 | struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header); | |
ce9c5a54 TM |
2123 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2124 | ref_ci, ref_root_bh); | |
a3944256 | 2125 | } else |
47bca495 TM |
2126 | ret = ocfs2_iterate_xattr_index_block(inode, |
2127 | blk_bh, | |
2128 | ocfs2_rm_xattr_cluster, | |
ce9c5a54 | 2129 | &args); |
cf1d6c76 TY |
2130 | |
2131 | return ret; | |
2132 | } | |
2133 | ||
08413899 | 2134 | static int ocfs2_xattr_free_block(struct inode *inode, |
ce9c5a54 TM |
2135 | u64 block, |
2136 | struct ocfs2_caching_info *ref_ci, | |
2137 | struct buffer_head *ref_root_bh) | |
08413899 TM |
2138 | { |
2139 | struct inode *xb_alloc_inode; | |
2140 | struct buffer_head *xb_alloc_bh = NULL; | |
2141 | struct buffer_head *blk_bh = NULL; | |
2142 | struct ocfs2_xattr_block *xb; | |
2143 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2144 | handle_t *handle; | |
2145 | int ret = 0; | |
2146 | u64 blk, bg_blkno; | |
2147 | u16 bit; | |
2148 | ||
4ae1d69b | 2149 | ret = ocfs2_read_xattr_block(inode, block, &blk_bh); |
08413899 TM |
2150 | if (ret < 0) { |
2151 | mlog_errno(ret); | |
2152 | goto out; | |
2153 | } | |
2154 | ||
ce9c5a54 | 2155 | ret = ocfs2_xattr_block_remove(inode, blk_bh, ref_ci, ref_root_bh); |
08413899 TM |
2156 | if (ret < 0) { |
2157 | mlog_errno(ret); | |
2158 | goto out; | |
2159 | } | |
2160 | ||
4ae1d69b | 2161 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
08413899 TM |
2162 | blk = le64_to_cpu(xb->xb_blkno); |
2163 | bit = le16_to_cpu(xb->xb_suballoc_bit); | |
2164 | bg_blkno = ocfs2_which_suballoc_group(blk, bit); | |
2165 | ||
2166 | xb_alloc_inode = ocfs2_get_system_file_inode(osb, | |
2167 | EXTENT_ALLOC_SYSTEM_INODE, | |
2168 | le16_to_cpu(xb->xb_suballoc_slot)); | |
2169 | if (!xb_alloc_inode) { | |
2170 | ret = -ENOMEM; | |
2171 | mlog_errno(ret); | |
2172 | goto out; | |
2173 | } | |
2174 | mutex_lock(&xb_alloc_inode->i_mutex); | |
2175 | ||
2176 | ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1); | |
2177 | if (ret < 0) { | |
2178 | mlog_errno(ret); | |
2179 | goto out_mutex; | |
2180 | } | |
2181 | ||
2182 | handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE); | |
2183 | if (IS_ERR(handle)) { | |
2184 | ret = PTR_ERR(handle); | |
2185 | mlog_errno(ret); | |
2186 | goto out_unlock; | |
2187 | } | |
2188 | ||
2189 | ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh, | |
2190 | bit, bg_blkno, 1); | |
2191 | if (ret < 0) | |
2192 | mlog_errno(ret); | |
2193 | ||
2194 | ocfs2_commit_trans(osb, handle); | |
2195 | out_unlock: | |
2196 | ocfs2_inode_unlock(xb_alloc_inode, 1); | |
2197 | brelse(xb_alloc_bh); | |
2198 | out_mutex: | |
2199 | mutex_unlock(&xb_alloc_inode->i_mutex); | |
2200 | iput(xb_alloc_inode); | |
2201 | out: | |
2202 | brelse(blk_bh); | |
2203 | return ret; | |
2204 | } | |
2205 | ||
cf1d6c76 TY |
2206 | /* |
2207 | * ocfs2_xattr_remove() | |
2208 | * | |
2209 | * Free extended attribute resources associated with this inode. | |
2210 | */ | |
2211 | int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) | |
2212 | { | |
cf1d6c76 TY |
2213 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
2214 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
ce9c5a54 TM |
2215 | struct ocfs2_refcount_tree *ref_tree = NULL; |
2216 | struct buffer_head *ref_root_bh = NULL; | |
2217 | struct ocfs2_caching_info *ref_ci = NULL; | |
cf1d6c76 TY |
2218 | handle_t *handle; |
2219 | int ret; | |
2220 | ||
8154da3d TY |
2221 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
2222 | return 0; | |
2223 | ||
cf1d6c76 TY |
2224 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
2225 | return 0; | |
2226 | ||
ce9c5a54 TM |
2227 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) { |
2228 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(inode->i_sb), | |
2229 | le64_to_cpu(di->i_refcount_loc), | |
2230 | 1, &ref_tree, &ref_root_bh); | |
2231 | if (ret) { | |
2232 | mlog_errno(ret); | |
2233 | goto out; | |
2234 | } | |
2235 | ref_ci = &ref_tree->rf_ci; | |
2236 | ||
2237 | } | |
2238 | ||
cf1d6c76 | 2239 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { |
ce9c5a54 TM |
2240 | ret = ocfs2_xattr_ibody_remove(inode, di_bh, |
2241 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2242 | if (ret < 0) { |
2243 | mlog_errno(ret); | |
2244 | goto out; | |
2245 | } | |
2246 | } | |
cf1d6c76 | 2247 | |
08413899 TM |
2248 | if (di->i_xattr_loc) { |
2249 | ret = ocfs2_xattr_free_block(inode, | |
ce9c5a54 TM |
2250 | le64_to_cpu(di->i_xattr_loc), |
2251 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2252 | if (ret < 0) { |
2253 | mlog_errno(ret); | |
2254 | goto out; | |
2255 | } | |
2256 | } | |
2257 | ||
2258 | handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)), | |
2259 | OCFS2_INODE_UPDATE_CREDITS); | |
2260 | if (IS_ERR(handle)) { | |
2261 | ret = PTR_ERR(handle); | |
2262 | mlog_errno(ret); | |
2263 | goto out; | |
2264 | } | |
0cf2f763 | 2265 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, |
84008972 | 2266 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2267 | if (ret) { |
2268 | mlog_errno(ret); | |
2269 | goto out_commit; | |
2270 | } | |
2271 | ||
08413899 | 2272 | di->i_xattr_loc = 0; |
cf1d6c76 TY |
2273 | |
2274 | spin_lock(&oi->ip_lock); | |
2275 | oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL); | |
2276 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2277 | spin_unlock(&oi->ip_lock); | |
2278 | ||
2279 | ret = ocfs2_journal_dirty(handle, di_bh); | |
2280 | if (ret < 0) | |
2281 | mlog_errno(ret); | |
2282 | out_commit: | |
2283 | ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); | |
2284 | out: | |
ce9c5a54 TM |
2285 | if (ref_tree) |
2286 | ocfs2_unlock_refcount_tree(OCFS2_SB(inode->i_sb), ref_tree, 1); | |
2287 | brelse(ref_root_bh); | |
cf1d6c76 TY |
2288 | return ret; |
2289 | } | |
2290 | ||
2291 | static int ocfs2_xattr_has_space_inline(struct inode *inode, | |
2292 | struct ocfs2_dinode *di) | |
2293 | { | |
2294 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2295 | unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
2296 | int free; | |
2297 | ||
2298 | if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE) | |
2299 | return 0; | |
2300 | ||
2301 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2302 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2303 | free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size); | |
2304 | } else if (ocfs2_inode_is_fast_symlink(inode)) { | |
2305 | free = ocfs2_fast_symlink_chars(inode->i_sb) - | |
2306 | le64_to_cpu(di->i_size); | |
2307 | } else { | |
2308 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2309 | free = (le16_to_cpu(el->l_count) - | |
2310 | le16_to_cpu(el->l_next_free_rec)) * | |
2311 | sizeof(struct ocfs2_extent_rec); | |
2312 | } | |
2313 | if (free >= xattrsize) | |
2314 | return 1; | |
2315 | ||
2316 | return 0; | |
2317 | } | |
2318 | ||
2319 | /* | |
2320 | * ocfs2_xattr_ibody_find() | |
2321 | * | |
2322 | * Find extended attribute in inode block and | |
2323 | * fill search info into struct ocfs2_xattr_search. | |
2324 | */ | |
2325 | static int ocfs2_xattr_ibody_find(struct inode *inode, | |
2326 | int name_index, | |
2327 | const char *name, | |
2328 | struct ocfs2_xattr_search *xs) | |
2329 | { | |
2330 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2331 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2332 | int ret; | |
2333 | int has_space = 0; | |
2334 | ||
2335 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2336 | return 0; | |
2337 | ||
2338 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2339 | down_read(&oi->ip_alloc_sem); | |
2340 | has_space = ocfs2_xattr_has_space_inline(inode, di); | |
2341 | up_read(&oi->ip_alloc_sem); | |
2342 | if (!has_space) | |
2343 | return 0; | |
2344 | } | |
2345 | ||
2346 | xs->xattr_bh = xs->inode_bh; | |
2347 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
2348 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
2349 | xs->header = (struct ocfs2_xattr_header *) | |
2350 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
2351 | else | |
2352 | xs->header = (struct ocfs2_xattr_header *) | |
2353 | (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size); | |
2354 | xs->base = (void *)xs->header; | |
2355 | xs->here = xs->header->xh_entries; | |
2356 | ||
2357 | /* Find the named attribute. */ | |
2358 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
2359 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2360 | if (ret && ret != -ENODATA) | |
2361 | return ret; | |
2362 | xs->not_found = ret; | |
2363 | } | |
2364 | ||
2365 | return 0; | |
2366 | } | |
2367 | ||
2368 | /* | |
2369 | * ocfs2_xattr_ibody_set() | |
2370 | * | |
2371 | * Set, replace or remove an extended attribute into inode block. | |
2372 | * | |
2373 | */ | |
2374 | static int ocfs2_xattr_ibody_set(struct inode *inode, | |
2375 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2376 | struct ocfs2_xattr_search *xs, |
2377 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2378 | { |
2379 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2380 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2381 | int ret; | |
2382 | ||
2383 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2384 | return -ENOSPC; | |
2385 | ||
2386 | down_write(&oi->ip_alloc_sem); | |
2387 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2388 | if (!ocfs2_xattr_has_space_inline(inode, di)) { | |
2389 | ret = -ENOSPC; | |
2390 | goto out; | |
2391 | } | |
2392 | } | |
2393 | ||
78f30c31 | 2394 | ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, |
cf1d6c76 TY |
2395 | (OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL)); |
2396 | out: | |
2397 | up_write(&oi->ip_alloc_sem); | |
2398 | ||
2399 | return ret; | |
2400 | } | |
2401 | ||
2402 | /* | |
2403 | * ocfs2_xattr_block_find() | |
2404 | * | |
2405 | * Find extended attribute in external block and | |
2406 | * fill search info into struct ocfs2_xattr_search. | |
2407 | */ | |
2408 | static int ocfs2_xattr_block_find(struct inode *inode, | |
2409 | int name_index, | |
2410 | const char *name, | |
2411 | struct ocfs2_xattr_search *xs) | |
2412 | { | |
2413 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2414 | struct buffer_head *blk_bh = NULL; | |
589dc260 | 2415 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
2416 | int ret = 0; |
2417 | ||
2418 | if (!di->i_xattr_loc) | |
2419 | return ret; | |
2420 | ||
4ae1d69b JB |
2421 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
2422 | &blk_bh); | |
cf1d6c76 TY |
2423 | if (ret < 0) { |
2424 | mlog_errno(ret); | |
2425 | return ret; | |
2426 | } | |
f6087fb7 | 2427 | |
cf1d6c76 | 2428 | xs->xattr_bh = blk_bh; |
4ae1d69b | 2429 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
589dc260 TM |
2430 | |
2431 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2432 | xs->header = &xb->xb_attrs.xb_header; | |
2433 | xs->base = (void *)xs->header; | |
2434 | xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size; | |
2435 | xs->here = xs->header->xh_entries; | |
2436 | ||
2437 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2438 | } else | |
2439 | ret = ocfs2_xattr_index_block_find(inode, blk_bh, | |
2440 | name_index, | |
2441 | name, xs); | |
cf1d6c76 | 2442 | |
cf1d6c76 TY |
2443 | if (ret && ret != -ENODATA) { |
2444 | xs->xattr_bh = NULL; | |
2445 | goto cleanup; | |
2446 | } | |
2447 | xs->not_found = ret; | |
2448 | return 0; | |
cf1d6c76 TY |
2449 | cleanup: |
2450 | brelse(blk_bh); | |
2451 | ||
2452 | return ret; | |
2453 | } | |
2454 | ||
5aea1f0e TM |
2455 | static int ocfs2_create_xattr_block(handle_t *handle, |
2456 | struct inode *inode, | |
2457 | struct buffer_head *inode_bh, | |
2458 | struct ocfs2_alloc_context *meta_ac, | |
a7fe7a3a TM |
2459 | struct buffer_head **ret_bh, |
2460 | int indexed) | |
5aea1f0e TM |
2461 | { |
2462 | int ret; | |
2463 | u16 suballoc_bit_start; | |
2464 | u32 num_got; | |
2465 | u64 first_blkno; | |
2466 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data; | |
2467 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2468 | struct buffer_head *new_bh = NULL; | |
2469 | struct ocfs2_xattr_block *xblk; | |
2470 | ||
2471 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), inode_bh, | |
2472 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2473 | if (ret < 0) { | |
2474 | mlog_errno(ret); | |
2475 | goto end; | |
2476 | } | |
2477 | ||
2478 | ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, | |
2479 | &suballoc_bit_start, &num_got, | |
2480 | &first_blkno); | |
2481 | if (ret < 0) { | |
2482 | mlog_errno(ret); | |
2483 | goto end; | |
2484 | } | |
2485 | ||
2486 | new_bh = sb_getblk(inode->i_sb, first_blkno); | |
2487 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); | |
2488 | ||
2489 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), | |
2490 | new_bh, | |
2491 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2492 | if (ret < 0) { | |
2493 | mlog_errno(ret); | |
2494 | goto end; | |
2495 | } | |
2496 | ||
2497 | /* Initialize ocfs2_xattr_block */ | |
2498 | xblk = (struct ocfs2_xattr_block *)new_bh->b_data; | |
2499 | memset(xblk, 0, inode->i_sb->s_blocksize); | |
2500 | strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); | |
b89c5428 | 2501 | xblk->xb_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); |
5aea1f0e TM |
2502 | xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); |
2503 | xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation); | |
2504 | xblk->xb_blkno = cpu_to_le64(first_blkno); | |
2505 | ||
a7fe7a3a TM |
2506 | if (indexed) { |
2507 | struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root; | |
2508 | xr->xt_clusters = cpu_to_le32(1); | |
2509 | xr->xt_last_eb_blk = 0; | |
2510 | xr->xt_list.l_tree_depth = 0; | |
2511 | xr->xt_list.l_count = cpu_to_le16( | |
2512 | ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
2513 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
2514 | xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED); | |
2515 | } | |
2516 | ||
5aea1f0e TM |
2517 | ret = ocfs2_journal_dirty(handle, new_bh); |
2518 | if (ret < 0) { | |
2519 | mlog_errno(ret); | |
2520 | goto end; | |
2521 | } | |
2522 | di->i_xattr_loc = cpu_to_le64(first_blkno); | |
2523 | ocfs2_journal_dirty(handle, inode_bh); | |
2524 | ||
2525 | *ret_bh = new_bh; | |
2526 | new_bh = NULL; | |
2527 | ||
2528 | end: | |
2529 | brelse(new_bh); | |
2530 | return ret; | |
2531 | } | |
2532 | ||
cf1d6c76 TY |
2533 | /* |
2534 | * ocfs2_xattr_block_set() | |
2535 | * | |
2536 | * Set, replace or remove an extended attribute into external block. | |
2537 | * | |
2538 | */ | |
2539 | static int ocfs2_xattr_block_set(struct inode *inode, | |
2540 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2541 | struct ocfs2_xattr_search *xs, |
2542 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2543 | { |
2544 | struct buffer_head *new_bh = NULL; | |
85db90e7 | 2545 | handle_t *handle = ctxt->handle; |
cf1d6c76 | 2546 | struct ocfs2_xattr_block *xblk = NULL; |
cf1d6c76 TY |
2547 | int ret; |
2548 | ||
2549 | if (!xs->xattr_bh) { | |
5aea1f0e | 2550 | ret = ocfs2_create_xattr_block(handle, inode, xs->inode_bh, |
a7fe7a3a | 2551 | ctxt->meta_ac, &new_bh, 0); |
5aea1f0e | 2552 | if (ret) { |
cf1d6c76 | 2553 | mlog_errno(ret); |
85db90e7 | 2554 | goto end; |
cf1d6c76 TY |
2555 | } |
2556 | ||
cf1d6c76 | 2557 | xs->xattr_bh = new_bh; |
5aea1f0e | 2558 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
2559 | xs->header = &xblk->xb_attrs.xb_header; |
2560 | xs->base = (void *)xs->header; | |
2561 | xs->end = (void *)xblk + inode->i_sb->s_blocksize; | |
2562 | xs->here = xs->header->xh_entries; | |
01225596 TM |
2563 | } else |
2564 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; | |
2565 | ||
2566 | if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2567 | /* Set extended attribute into external block */ | |
78f30c31 TM |
2568 | ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, |
2569 | OCFS2_HAS_XATTR_FL); | |
01225596 TM |
2570 | if (!ret || ret != -ENOSPC) |
2571 | goto end; | |
2572 | ||
78f30c31 | 2573 | ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); |
01225596 TM |
2574 | if (ret) |
2575 | goto end; | |
cf1d6c76 TY |
2576 | } |
2577 | ||
78f30c31 | 2578 | ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); |
01225596 TM |
2579 | |
2580 | end: | |
cf1d6c76 TY |
2581 | |
2582 | return ret; | |
2583 | } | |
2584 | ||
78f30c31 TM |
2585 | /* Check whether the new xattr can be inserted into the inode. */ |
2586 | static int ocfs2_xattr_can_be_in_inode(struct inode *inode, | |
2587 | struct ocfs2_xattr_info *xi, | |
2588 | struct ocfs2_xattr_search *xs) | |
2589 | { | |
2590 | u64 value_size; | |
2591 | struct ocfs2_xattr_entry *last; | |
2592 | int free, i; | |
2593 | size_t min_offs = xs->end - xs->base; | |
2594 | ||
2595 | if (!xs->header) | |
2596 | return 0; | |
2597 | ||
2598 | last = xs->header->xh_entries; | |
2599 | ||
2600 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
2601 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
2602 | if (offs < min_offs) | |
2603 | min_offs = offs; | |
2604 | last += 1; | |
2605 | } | |
2606 | ||
4442f518 | 2607 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
78f30c31 TM |
2608 | if (free < 0) |
2609 | return 0; | |
2610 | ||
2611 | BUG_ON(!xs->not_found); | |
2612 | ||
2613 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) | |
2614 | value_size = OCFS2_XATTR_ROOT_SIZE; | |
2615 | else | |
2616 | value_size = OCFS2_XATTR_SIZE(xi->value_len); | |
2617 | ||
2618 | if (free >= sizeof(struct ocfs2_xattr_entry) + | |
2619 | OCFS2_XATTR_SIZE(strlen(xi->name)) + value_size) | |
2620 | return 1; | |
2621 | ||
2622 | return 0; | |
2623 | } | |
2624 | ||
2625 | static int ocfs2_calc_xattr_set_need(struct inode *inode, | |
2626 | struct ocfs2_dinode *di, | |
2627 | struct ocfs2_xattr_info *xi, | |
2628 | struct ocfs2_xattr_search *xis, | |
2629 | struct ocfs2_xattr_search *xbs, | |
2630 | int *clusters_need, | |
85db90e7 TM |
2631 | int *meta_need, |
2632 | int *credits_need) | |
78f30c31 TM |
2633 | { |
2634 | int ret = 0, old_in_xb = 0; | |
85db90e7 | 2635 | int clusters_add = 0, meta_add = 0, credits = 0; |
78f30c31 TM |
2636 | struct buffer_head *bh = NULL; |
2637 | struct ocfs2_xattr_block *xb = NULL; | |
2638 | struct ocfs2_xattr_entry *xe = NULL; | |
2639 | struct ocfs2_xattr_value_root *xv = NULL; | |
2640 | char *base = NULL; | |
2641 | int name_offset, name_len = 0; | |
2642 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
2643 | xi->value_len); | |
2644 | u64 value_size; | |
2645 | ||
71d548a6 TM |
2646 | /* |
2647 | * Calculate the clusters we need to write. | |
2648 | * No matter whether we replace an old one or add a new one, | |
2649 | * we need this for writing. | |
2650 | */ | |
2651 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) | |
2652 | credits += new_clusters * | |
2653 | ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
2654 | ||
78f30c31 | 2655 | if (xis->not_found && xbs->not_found) { |
85db90e7 TM |
2656 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
2657 | ||
2658 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
78f30c31 | 2659 | clusters_add += new_clusters; |
85db90e7 TM |
2660 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
2661 | &def_xv.xv.xr_list, | |
2662 | new_clusters); | |
2663 | } | |
78f30c31 TM |
2664 | |
2665 | goto meta_guess; | |
2666 | } | |
2667 | ||
2668 | if (!xis->not_found) { | |
2669 | xe = xis->here; | |
2670 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
2671 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
2672 | base = xis->base; | |
85db90e7 | 2673 | credits += OCFS2_INODE_UPDATE_CREDITS; |
78f30c31 | 2674 | } else { |
970e4936 | 2675 | int i, block_off = 0; |
78f30c31 TM |
2676 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; |
2677 | xe = xbs->here; | |
2678 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
2679 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
2680 | i = xbs->here - xbs->header->xh_entries; | |
2681 | old_in_xb = 1; | |
2682 | ||
2683 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 2684 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
78f30c31 TM |
2685 | bucket_xh(xbs->bucket), |
2686 | i, &block_off, | |
2687 | &name_offset); | |
2688 | base = bucket_block(xbs->bucket, block_off); | |
85db90e7 TM |
2689 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
2690 | } else { | |
78f30c31 | 2691 | base = xbs->base; |
85db90e7 TM |
2692 | credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS; |
2693 | } | |
2694 | } | |
2695 | ||
2696 | /* | |
2697 | * delete a xattr doesn't need metadata and cluster allocation. | |
2698 | * so just calculate the credits and return. | |
2699 | * | |
2700 | * The credits for removing the value tree will be extended | |
2701 | * by ocfs2_remove_extent itself. | |
2702 | */ | |
2703 | if (!xi->value) { | |
2704 | if (!ocfs2_xattr_is_local(xe)) | |
a90714c1 | 2705 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
85db90e7 TM |
2706 | |
2707 | goto out; | |
78f30c31 TM |
2708 | } |
2709 | ||
2710 | /* do cluster allocation guess first. */ | |
2711 | value_size = le64_to_cpu(xe->xe_value_size); | |
2712 | ||
2713 | if (old_in_xb) { | |
2714 | /* | |
2715 | * In xattr set, we always try to set the xe in inode first, | |
2716 | * so if it can be inserted into inode successfully, the old | |
2717 | * one will be removed from the xattr block, and this xattr | |
2718 | * will be inserted into inode as a new xattr in inode. | |
2719 | */ | |
2720 | if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) { | |
2721 | clusters_add += new_clusters; | |
a90714c1 | 2722 | credits += ocfs2_remove_extent_credits(inode->i_sb) + |
85db90e7 TM |
2723 | OCFS2_INODE_UPDATE_CREDITS; |
2724 | if (!ocfs2_xattr_is_local(xe)) | |
2725 | credits += ocfs2_calc_extend_credits( | |
2726 | inode->i_sb, | |
2727 | &def_xv.xv.xr_list, | |
2728 | new_clusters); | |
78f30c31 TM |
2729 | goto out; |
2730 | } | |
2731 | } | |
2732 | ||
2733 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
2734 | /* the new values will be stored outside. */ | |
2735 | u32 old_clusters = 0; | |
2736 | ||
2737 | if (!ocfs2_xattr_is_local(xe)) { | |
2738 | old_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
2739 | value_size); | |
2740 | xv = (struct ocfs2_xattr_value_root *) | |
2741 | (base + name_offset + name_len); | |
97aff52a | 2742 | value_size = OCFS2_XATTR_ROOT_SIZE; |
78f30c31 TM |
2743 | } else |
2744 | xv = &def_xv.xv; | |
2745 | ||
85db90e7 | 2746 | if (old_clusters >= new_clusters) { |
a90714c1 | 2747 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
78f30c31 | 2748 | goto out; |
85db90e7 | 2749 | } else { |
78f30c31 TM |
2750 | meta_add += ocfs2_extend_meta_needed(&xv->xr_list); |
2751 | clusters_add += new_clusters - old_clusters; | |
85db90e7 TM |
2752 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
2753 | &xv->xr_list, | |
2754 | new_clusters - | |
2755 | old_clusters); | |
97aff52a TM |
2756 | if (value_size >= OCFS2_XATTR_ROOT_SIZE) |
2757 | goto out; | |
78f30c31 TM |
2758 | } |
2759 | } else { | |
2760 | /* | |
2761 | * Now the new value will be stored inside. So if the new | |
2762 | * value is smaller than the size of value root or the old | |
2763 | * value, we don't need any allocation, otherwise we have | |
2764 | * to guess metadata allocation. | |
2765 | */ | |
2766 | if ((ocfs2_xattr_is_local(xe) && value_size >= xi->value_len) || | |
2767 | (!ocfs2_xattr_is_local(xe) && | |
2768 | OCFS2_XATTR_ROOT_SIZE >= xi->value_len)) | |
2769 | goto out; | |
2770 | } | |
2771 | ||
2772 | meta_guess: | |
2773 | /* calculate metadata allocation. */ | |
2774 | if (di->i_xattr_loc) { | |
2775 | if (!xbs->xattr_bh) { | |
4ae1d69b JB |
2776 | ret = ocfs2_read_xattr_block(inode, |
2777 | le64_to_cpu(di->i_xattr_loc), | |
2778 | &bh); | |
78f30c31 TM |
2779 | if (ret) { |
2780 | mlog_errno(ret); | |
2781 | goto out; | |
2782 | } | |
2783 | ||
2784 | xb = (struct ocfs2_xattr_block *)bh->b_data; | |
2785 | } else | |
2786 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
2787 | ||
90cb546c TM |
2788 | /* |
2789 | * If there is already an xattr tree, good, we can calculate | |
2790 | * like other b-trees. Otherwise we may have the chance of | |
2791 | * create a tree, the credit calculation is borrowed from | |
2792 | * ocfs2_calc_extend_credits with root_el = NULL. And the | |
2793 | * new tree will be cluster based, so no meta is needed. | |
2794 | */ | |
78f30c31 TM |
2795 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { |
2796 | struct ocfs2_extent_list *el = | |
2797 | &xb->xb_attrs.xb_root.xt_list; | |
2798 | meta_add += ocfs2_extend_meta_needed(el); | |
85db90e7 TM |
2799 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
2800 | el, 1); | |
90cb546c TM |
2801 | } else |
2802 | credits += OCFS2_SUBALLOC_ALLOC + 1; | |
78f30c31 TM |
2803 | |
2804 | /* | |
2805 | * This cluster will be used either for new bucket or for | |
2806 | * new xattr block. | |
2807 | * If the cluster size is the same as the bucket size, one | |
2808 | * more is needed since we may need to extend the bucket | |
2809 | * also. | |
2810 | */ | |
2811 | clusters_add += 1; | |
85db90e7 | 2812 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
78f30c31 | 2813 | if (OCFS2_XATTR_BUCKET_SIZE == |
85db90e7 TM |
2814 | OCFS2_SB(inode->i_sb)->s_clustersize) { |
2815 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
78f30c31 | 2816 | clusters_add += 1; |
85db90e7 TM |
2817 | } |
2818 | } else { | |
78f30c31 | 2819 | meta_add += 1; |
85db90e7 TM |
2820 | credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
2821 | } | |
78f30c31 TM |
2822 | out: |
2823 | if (clusters_need) | |
2824 | *clusters_need = clusters_add; | |
2825 | if (meta_need) | |
2826 | *meta_need = meta_add; | |
85db90e7 TM |
2827 | if (credits_need) |
2828 | *credits_need = credits; | |
78f30c31 TM |
2829 | brelse(bh); |
2830 | return ret; | |
2831 | } | |
2832 | ||
2833 | static int ocfs2_init_xattr_set_ctxt(struct inode *inode, | |
2834 | struct ocfs2_dinode *di, | |
2835 | struct ocfs2_xattr_info *xi, | |
2836 | struct ocfs2_xattr_search *xis, | |
2837 | struct ocfs2_xattr_search *xbs, | |
85db90e7 | 2838 | struct ocfs2_xattr_set_ctxt *ctxt, |
492a8a33 | 2839 | int extra_meta, |
85db90e7 | 2840 | int *credits) |
78f30c31 TM |
2841 | { |
2842 | int clusters_add, meta_add, ret; | |
2843 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2844 | ||
2845 | memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt)); | |
2846 | ||
2847 | ocfs2_init_dealloc_ctxt(&ctxt->dealloc); | |
2848 | ||
2849 | ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs, | |
85db90e7 | 2850 | &clusters_add, &meta_add, credits); |
78f30c31 TM |
2851 | if (ret) { |
2852 | mlog_errno(ret); | |
2853 | return ret; | |
2854 | } | |
2855 | ||
492a8a33 | 2856 | meta_add += extra_meta; |
85db90e7 TM |
2857 | mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, " |
2858 | "credits = %d\n", xi->name, meta_add, clusters_add, *credits); | |
78f30c31 TM |
2859 | |
2860 | if (meta_add) { | |
2861 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, | |
2862 | &ctxt->meta_ac); | |
2863 | if (ret) { | |
2864 | mlog_errno(ret); | |
2865 | goto out; | |
2866 | } | |
2867 | } | |
2868 | ||
2869 | if (clusters_add) { | |
2870 | ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac); | |
2871 | if (ret) | |
2872 | mlog_errno(ret); | |
2873 | } | |
2874 | out: | |
2875 | if (ret) { | |
2876 | if (ctxt->meta_ac) { | |
2877 | ocfs2_free_alloc_context(ctxt->meta_ac); | |
2878 | ctxt->meta_ac = NULL; | |
2879 | } | |
2880 | ||
2881 | /* | |
2882 | * We cannot have an error and a non null ctxt->data_ac. | |
2883 | */ | |
2884 | } | |
2885 | ||
2886 | return ret; | |
2887 | } | |
2888 | ||
85db90e7 TM |
2889 | static int __ocfs2_xattr_set_handle(struct inode *inode, |
2890 | struct ocfs2_dinode *di, | |
2891 | struct ocfs2_xattr_info *xi, | |
2892 | struct ocfs2_xattr_search *xis, | |
2893 | struct ocfs2_xattr_search *xbs, | |
2894 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2895 | { | |
9f868f16 | 2896 | int ret = 0, credits, old_found; |
85db90e7 TM |
2897 | |
2898 | if (!xi->value) { | |
2899 | /* Remove existing extended attribute */ | |
2900 | if (!xis->not_found) | |
2901 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
2902 | else if (!xbs->not_found) | |
2903 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
2904 | } else { | |
2905 | /* We always try to set extended attribute into inode first*/ | |
2906 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
2907 | if (!ret && !xbs->not_found) { | |
2908 | /* | |
2909 | * If succeed and that extended attribute existing in | |
2910 | * external block, then we will remove it. | |
2911 | */ | |
2912 | xi->value = NULL; | |
2913 | xi->value_len = 0; | |
2914 | ||
9f868f16 | 2915 | old_found = xis->not_found; |
85db90e7 TM |
2916 | xis->not_found = -ENODATA; |
2917 | ret = ocfs2_calc_xattr_set_need(inode, | |
2918 | di, | |
2919 | xi, | |
2920 | xis, | |
2921 | xbs, | |
2922 | NULL, | |
2923 | NULL, | |
2924 | &credits); | |
9f868f16 | 2925 | xis->not_found = old_found; |
85db90e7 TM |
2926 | if (ret) { |
2927 | mlog_errno(ret); | |
2928 | goto out; | |
2929 | } | |
2930 | ||
2931 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
2932 | ctxt->handle->h_buffer_credits); | |
2933 | if (ret) { | |
2934 | mlog_errno(ret); | |
2935 | goto out; | |
2936 | } | |
2937 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
2938 | } else if (ret == -ENOSPC) { | |
2939 | if (di->i_xattr_loc && !xbs->xattr_bh) { | |
2940 | ret = ocfs2_xattr_block_find(inode, | |
2941 | xi->name_index, | |
2942 | xi->name, xbs); | |
2943 | if (ret) | |
2944 | goto out; | |
2945 | ||
9f868f16 | 2946 | old_found = xis->not_found; |
85db90e7 TM |
2947 | xis->not_found = -ENODATA; |
2948 | ret = ocfs2_calc_xattr_set_need(inode, | |
2949 | di, | |
2950 | xi, | |
2951 | xis, | |
2952 | xbs, | |
2953 | NULL, | |
2954 | NULL, | |
2955 | &credits); | |
9f868f16 | 2956 | xis->not_found = old_found; |
85db90e7 TM |
2957 | if (ret) { |
2958 | mlog_errno(ret); | |
2959 | goto out; | |
2960 | } | |
2961 | ||
2962 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
2963 | ctxt->handle->h_buffer_credits); | |
2964 | if (ret) { | |
2965 | mlog_errno(ret); | |
2966 | goto out; | |
2967 | } | |
2968 | } | |
2969 | /* | |
2970 | * If no space in inode, we will set extended attribute | |
2971 | * into external block. | |
2972 | */ | |
2973 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
2974 | if (ret) | |
2975 | goto out; | |
2976 | if (!xis->not_found) { | |
2977 | /* | |
2978 | * If succeed and that extended attribute | |
2979 | * existing in inode, we will remove it. | |
2980 | */ | |
2981 | xi->value = NULL; | |
2982 | xi->value_len = 0; | |
2983 | xbs->not_found = -ENODATA; | |
2984 | ret = ocfs2_calc_xattr_set_need(inode, | |
2985 | di, | |
2986 | xi, | |
2987 | xis, | |
2988 | xbs, | |
2989 | NULL, | |
2990 | NULL, | |
2991 | &credits); | |
2992 | if (ret) { | |
2993 | mlog_errno(ret); | |
2994 | goto out; | |
2995 | } | |
2996 | ||
2997 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
2998 | ctxt->handle->h_buffer_credits); | |
2999 | if (ret) { | |
3000 | mlog_errno(ret); | |
3001 | goto out; | |
3002 | } | |
3003 | ret = ocfs2_xattr_ibody_set(inode, xi, | |
3004 | xis, ctxt); | |
3005 | } | |
3006 | } | |
3007 | } | |
3008 | ||
4b3f6209 TM |
3009 | if (!ret) { |
3010 | /* Update inode ctime. */ | |
0cf2f763 | 3011 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
89a907af TM |
3012 | xis->inode_bh, |
3013 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4b3f6209 TM |
3014 | if (ret) { |
3015 | mlog_errno(ret); | |
3016 | goto out; | |
3017 | } | |
3018 | ||
3019 | inode->i_ctime = CURRENT_TIME; | |
3020 | di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); | |
3021 | di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); | |
3022 | ocfs2_journal_dirty(ctxt->handle, xis->inode_bh); | |
3023 | } | |
85db90e7 TM |
3024 | out: |
3025 | return ret; | |
3026 | } | |
3027 | ||
6c3faba4 TY |
3028 | /* |
3029 | * This function only called duing creating inode | |
3030 | * for init security/acl xattrs of the new inode. | |
008aafaf | 3031 | * All transanction credits have been reserved in mknod. |
6c3faba4 TY |
3032 | */ |
3033 | int ocfs2_xattr_set_handle(handle_t *handle, | |
3034 | struct inode *inode, | |
3035 | struct buffer_head *di_bh, | |
3036 | int name_index, | |
3037 | const char *name, | |
3038 | const void *value, | |
3039 | size_t value_len, | |
3040 | int flags, | |
3041 | struct ocfs2_alloc_context *meta_ac, | |
3042 | struct ocfs2_alloc_context *data_ac) | |
3043 | { | |
3044 | struct ocfs2_dinode *di; | |
3045 | int ret; | |
3046 | ||
3047 | struct ocfs2_xattr_info xi = { | |
3048 | .name_index = name_index, | |
3049 | .name = name, | |
3050 | .value = value, | |
3051 | .value_len = value_len, | |
3052 | }; | |
3053 | ||
3054 | struct ocfs2_xattr_search xis = { | |
3055 | .not_found = -ENODATA, | |
3056 | }; | |
3057 | ||
3058 | struct ocfs2_xattr_search xbs = { | |
3059 | .not_found = -ENODATA, | |
3060 | }; | |
3061 | ||
3062 | struct ocfs2_xattr_set_ctxt ctxt = { | |
3063 | .handle = handle, | |
3064 | .meta_ac = meta_ac, | |
3065 | .data_ac = data_ac, | |
3066 | }; | |
3067 | ||
3068 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) | |
3069 | return -EOPNOTSUPP; | |
3070 | ||
008aafaf TY |
3071 | /* |
3072 | * In extreme situation, may need xattr bucket when | |
3073 | * block size is too small. And we have already reserved | |
3074 | * the credits for bucket in mknod. | |
3075 | */ | |
3076 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) { | |
3077 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3078 | if (!xbs.bucket) { | |
3079 | mlog_errno(-ENOMEM); | |
3080 | return -ENOMEM; | |
3081 | } | |
3082 | } | |
3083 | ||
6c3faba4 TY |
3084 | xis.inode_bh = xbs.inode_bh = di_bh; |
3085 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3086 | ||
3087 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3088 | ||
3089 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3090 | if (ret) | |
3091 | goto cleanup; | |
3092 | if (xis.not_found) { | |
3093 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3094 | if (ret) | |
3095 | goto cleanup; | |
3096 | } | |
3097 | ||
3098 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3099 | ||
3100 | cleanup: | |
3101 | up_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3102 | brelse(xbs.xattr_bh); | |
008aafaf | 3103 | ocfs2_xattr_bucket_free(xbs.bucket); |
6c3faba4 TY |
3104 | |
3105 | return ret; | |
3106 | } | |
3107 | ||
cf1d6c76 TY |
3108 | /* |
3109 | * ocfs2_xattr_set() | |
3110 | * | |
3111 | * Set, replace or remove an extended attribute for this inode. | |
3112 | * value is NULL to remove an existing extended attribute, else either | |
3113 | * create or replace an extended attribute. | |
3114 | */ | |
3115 | int ocfs2_xattr_set(struct inode *inode, | |
3116 | int name_index, | |
3117 | const char *name, | |
3118 | const void *value, | |
3119 | size_t value_len, | |
3120 | int flags) | |
3121 | { | |
3122 | struct buffer_head *di_bh = NULL; | |
3123 | struct ocfs2_dinode *di; | |
492a8a33 | 3124 | int ret, credits, ref_meta = 0, ref_credits = 0; |
78f30c31 | 3125 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
85db90e7 | 3126 | struct inode *tl_inode = osb->osb_tl_inode; |
78f30c31 | 3127 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; |
492a8a33 | 3128 | struct ocfs2_refcount_tree *ref_tree = NULL; |
cf1d6c76 TY |
3129 | |
3130 | struct ocfs2_xattr_info xi = { | |
3131 | .name_index = name_index, | |
3132 | .name = name, | |
3133 | .value = value, | |
3134 | .value_len = value_len, | |
3135 | }; | |
3136 | ||
3137 | struct ocfs2_xattr_search xis = { | |
3138 | .not_found = -ENODATA, | |
3139 | }; | |
3140 | ||
3141 | struct ocfs2_xattr_search xbs = { | |
3142 | .not_found = -ENODATA, | |
3143 | }; | |
3144 | ||
8154da3d TY |
3145 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
3146 | return -EOPNOTSUPP; | |
3147 | ||
ba937127 JB |
3148 | /* |
3149 | * Only xbs will be used on indexed trees. xis doesn't need a | |
3150 | * bucket. | |
3151 | */ | |
3152 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3153 | if (!xbs.bucket) { | |
3154 | mlog_errno(-ENOMEM); | |
3155 | return -ENOMEM; | |
3156 | } | |
3157 | ||
cf1d6c76 TY |
3158 | ret = ocfs2_inode_lock(inode, &di_bh, 1); |
3159 | if (ret < 0) { | |
3160 | mlog_errno(ret); | |
ba937127 | 3161 | goto cleanup_nolock; |
cf1d6c76 TY |
3162 | } |
3163 | xis.inode_bh = xbs.inode_bh = di_bh; | |
3164 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3165 | ||
3166 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3167 | /* | |
3168 | * Scan inode and external block to find the same name | |
3169 | * extended attribute and collect search infomation. | |
3170 | */ | |
3171 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3172 | if (ret) | |
3173 | goto cleanup; | |
3174 | if (xis.not_found) { | |
3175 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3176 | if (ret) | |
3177 | goto cleanup; | |
3178 | } | |
3179 | ||
3180 | if (xis.not_found && xbs.not_found) { | |
3181 | ret = -ENODATA; | |
3182 | if (flags & XATTR_REPLACE) | |
3183 | goto cleanup; | |
3184 | ret = 0; | |
3185 | if (!value) | |
3186 | goto cleanup; | |
3187 | } else { | |
3188 | ret = -EEXIST; | |
3189 | if (flags & XATTR_CREATE) | |
3190 | goto cleanup; | |
3191 | } | |
3192 | ||
492a8a33 TM |
3193 | /* Check whether the value is refcounted and do some prepartion. */ |
3194 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL && | |
3195 | (!xis.not_found || !xbs.not_found)) { | |
3196 | ret = ocfs2_prepare_refcount_xattr(inode, di, &xi, | |
3197 | &xis, &xbs, &ref_tree, | |
3198 | &ref_meta, &ref_credits); | |
3199 | if (ret) { | |
3200 | mlog_errno(ret); | |
3201 | goto cleanup; | |
3202 | } | |
3203 | } | |
85db90e7 TM |
3204 | |
3205 | mutex_lock(&tl_inode->i_mutex); | |
3206 | ||
3207 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
3208 | ret = __ocfs2_flush_truncate_log(osb); | |
3209 | if (ret < 0) { | |
3210 | mutex_unlock(&tl_inode->i_mutex); | |
3211 | mlog_errno(ret); | |
3212 | goto cleanup; | |
3213 | } | |
3214 | } | |
3215 | mutex_unlock(&tl_inode->i_mutex); | |
3216 | ||
3217 | ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis, | |
492a8a33 | 3218 | &xbs, &ctxt, ref_meta, &credits); |
78f30c31 TM |
3219 | if (ret) { |
3220 | mlog_errno(ret); | |
3221 | goto cleanup; | |
3222 | } | |
3223 | ||
4b3f6209 TM |
3224 | /* we need to update inode's ctime field, so add credit for it. */ |
3225 | credits += OCFS2_INODE_UPDATE_CREDITS; | |
492a8a33 | 3226 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); |
85db90e7 TM |
3227 | if (IS_ERR(ctxt.handle)) { |
3228 | ret = PTR_ERR(ctxt.handle); | |
3229 | mlog_errno(ret); | |
3230 | goto cleanup; | |
cf1d6c76 | 3231 | } |
85db90e7 TM |
3232 | |
3233 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3234 | ||
3235 | ocfs2_commit_trans(osb, ctxt.handle); | |
3236 | ||
78f30c31 TM |
3237 | if (ctxt.data_ac) |
3238 | ocfs2_free_alloc_context(ctxt.data_ac); | |
3239 | if (ctxt.meta_ac) | |
3240 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
3241 | if (ocfs2_dealloc_has_cluster(&ctxt.dealloc)) | |
3242 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
3243 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
8b2c0dba | 3244 | |
cf1d6c76 | 3245 | cleanup: |
492a8a33 TM |
3246 | if (ref_tree) |
3247 | ocfs2_unlock_refcount_tree(osb, ref_tree, 1); | |
cf1d6c76 | 3248 | up_write(&OCFS2_I(inode)->ip_xattr_sem); |
8b2c0dba TM |
3249 | if (!value && !ret) { |
3250 | ret = ocfs2_try_remove_refcount_tree(inode, di_bh); | |
3251 | if (ret) | |
3252 | mlog_errno(ret); | |
3253 | } | |
cf1d6c76 | 3254 | ocfs2_inode_unlock(inode, 1); |
ba937127 | 3255 | cleanup_nolock: |
cf1d6c76 TY |
3256 | brelse(di_bh); |
3257 | brelse(xbs.xattr_bh); | |
ba937127 | 3258 | ocfs2_xattr_bucket_free(xbs.bucket); |
cf1d6c76 TY |
3259 | |
3260 | return ret; | |
3261 | } | |
3262 | ||
0c044f0b TM |
3263 | /* |
3264 | * Find the xattr extent rec which may contains name_hash. | |
3265 | * e_cpos will be the first name hash of the xattr rec. | |
3266 | * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list. | |
3267 | */ | |
3268 | static int ocfs2_xattr_get_rec(struct inode *inode, | |
3269 | u32 name_hash, | |
3270 | u64 *p_blkno, | |
3271 | u32 *e_cpos, | |
3272 | u32 *num_clusters, | |
3273 | struct ocfs2_extent_list *el) | |
3274 | { | |
3275 | int ret = 0, i; | |
3276 | struct buffer_head *eb_bh = NULL; | |
3277 | struct ocfs2_extent_block *eb; | |
3278 | struct ocfs2_extent_rec *rec = NULL; | |
3279 | u64 e_blkno = 0; | |
3280 | ||
3281 | if (el->l_tree_depth) { | |
facdb77f JB |
3282 | ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash, |
3283 | &eb_bh); | |
0c044f0b TM |
3284 | if (ret) { |
3285 | mlog_errno(ret); | |
3286 | goto out; | |
3287 | } | |
3288 | ||
3289 | eb = (struct ocfs2_extent_block *) eb_bh->b_data; | |
3290 | el = &eb->h_list; | |
3291 | ||
3292 | if (el->l_tree_depth) { | |
3293 | ocfs2_error(inode->i_sb, | |
3294 | "Inode %lu has non zero tree depth in " | |
3295 | "xattr tree block %llu\n", inode->i_ino, | |
3296 | (unsigned long long)eb_bh->b_blocknr); | |
3297 | ret = -EROFS; | |
3298 | goto out; | |
3299 | } | |
3300 | } | |
3301 | ||
3302 | for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { | |
3303 | rec = &el->l_recs[i]; | |
3304 | ||
3305 | if (le32_to_cpu(rec->e_cpos) <= name_hash) { | |
3306 | e_blkno = le64_to_cpu(rec->e_blkno); | |
3307 | break; | |
3308 | } | |
3309 | } | |
3310 | ||
3311 | if (!e_blkno) { | |
3312 | ocfs2_error(inode->i_sb, "Inode %lu has bad extent " | |
3313 | "record (%u, %u, 0) in xattr", inode->i_ino, | |
3314 | le32_to_cpu(rec->e_cpos), | |
3315 | ocfs2_rec_clusters(el, rec)); | |
3316 | ret = -EROFS; | |
3317 | goto out; | |
3318 | } | |
3319 | ||
3320 | *p_blkno = le64_to_cpu(rec->e_blkno); | |
3321 | *num_clusters = le16_to_cpu(rec->e_leaf_clusters); | |
3322 | if (e_cpos) | |
3323 | *e_cpos = le32_to_cpu(rec->e_cpos); | |
3324 | out: | |
3325 | brelse(eb_bh); | |
3326 | return ret; | |
3327 | } | |
3328 | ||
3329 | typedef int (xattr_bucket_func)(struct inode *inode, | |
3330 | struct ocfs2_xattr_bucket *bucket, | |
3331 | void *para); | |
3332 | ||
589dc260 | 3333 | static int ocfs2_find_xe_in_bucket(struct inode *inode, |
e2356a3f | 3334 | struct ocfs2_xattr_bucket *bucket, |
589dc260 TM |
3335 | int name_index, |
3336 | const char *name, | |
3337 | u32 name_hash, | |
3338 | u16 *xe_index, | |
3339 | int *found) | |
3340 | { | |
3341 | int i, ret = 0, cmp = 1, block_off, new_offset; | |
e2356a3f | 3342 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
589dc260 TM |
3343 | size_t name_len = strlen(name); |
3344 | struct ocfs2_xattr_entry *xe = NULL; | |
589dc260 TM |
3345 | char *xe_name; |
3346 | ||
3347 | /* | |
3348 | * We don't use binary search in the bucket because there | |
3349 | * may be multiple entries with the same name hash. | |
3350 | */ | |
3351 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
3352 | xe = &xh->xh_entries[i]; | |
3353 | ||
3354 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) | |
3355 | continue; | |
3356 | else if (name_hash < le32_to_cpu(xe->xe_name_hash)) | |
3357 | break; | |
3358 | ||
3359 | cmp = name_index - ocfs2_xattr_get_type(xe); | |
3360 | if (!cmp) | |
3361 | cmp = name_len - xe->xe_name_len; | |
3362 | if (cmp) | |
3363 | continue; | |
3364 | ||
fd68a894 | 3365 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
589dc260 TM |
3366 | xh, |
3367 | i, | |
3368 | &block_off, | |
3369 | &new_offset); | |
3370 | if (ret) { | |
3371 | mlog_errno(ret); | |
3372 | break; | |
3373 | } | |
3374 | ||
970e4936 | 3375 | |
e2356a3f JB |
3376 | xe_name = bucket_block(bucket, block_off) + new_offset; |
3377 | if (!memcmp(name, xe_name, name_len)) { | |
589dc260 TM |
3378 | *xe_index = i; |
3379 | *found = 1; | |
3380 | ret = 0; | |
3381 | break; | |
3382 | } | |
3383 | } | |
3384 | ||
3385 | return ret; | |
3386 | } | |
3387 | ||
3388 | /* | |
3389 | * Find the specified xattr entry in a series of buckets. | |
3390 | * This series start from p_blkno and last for num_clusters. | |
3391 | * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains | |
3392 | * the num of the valid buckets. | |
3393 | * | |
3394 | * Return the buffer_head this xattr should reside in. And if the xattr's | |
3395 | * hash is in the gap of 2 buckets, return the lower bucket. | |
3396 | */ | |
3397 | static int ocfs2_xattr_bucket_find(struct inode *inode, | |
3398 | int name_index, | |
3399 | const char *name, | |
3400 | u32 name_hash, | |
3401 | u64 p_blkno, | |
3402 | u32 first_hash, | |
3403 | u32 num_clusters, | |
3404 | struct ocfs2_xattr_search *xs) | |
3405 | { | |
3406 | int ret, found = 0; | |
589dc260 TM |
3407 | struct ocfs2_xattr_header *xh = NULL; |
3408 | struct ocfs2_xattr_entry *xe = NULL; | |
3409 | u16 index = 0; | |
3410 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
3411 | int low_bucket = 0, bucket, high_bucket; | |
e2356a3f | 3412 | struct ocfs2_xattr_bucket *search; |
589dc260 | 3413 | u32 last_hash; |
e2356a3f | 3414 | u64 blkno, lower_blkno = 0; |
589dc260 | 3415 | |
e2356a3f JB |
3416 | search = ocfs2_xattr_bucket_new(inode); |
3417 | if (!search) { | |
3418 | ret = -ENOMEM; | |
3419 | mlog_errno(ret); | |
3420 | goto out; | |
3421 | } | |
3422 | ||
3423 | ret = ocfs2_read_xattr_bucket(search, p_blkno); | |
589dc260 TM |
3424 | if (ret) { |
3425 | mlog_errno(ret); | |
3426 | goto out; | |
3427 | } | |
3428 | ||
e2356a3f | 3429 | xh = bucket_xh(search); |
589dc260 | 3430 | high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1; |
589dc260 | 3431 | while (low_bucket <= high_bucket) { |
e2356a3f | 3432 | ocfs2_xattr_bucket_relse(search); |
589dc260 | 3433 | |
e2356a3f | 3434 | bucket = (low_bucket + high_bucket) / 2; |
589dc260 | 3435 | blkno = p_blkno + bucket * blk_per_bucket; |
e2356a3f | 3436 | ret = ocfs2_read_xattr_bucket(search, blkno); |
589dc260 TM |
3437 | if (ret) { |
3438 | mlog_errno(ret); | |
3439 | goto out; | |
3440 | } | |
3441 | ||
e2356a3f | 3442 | xh = bucket_xh(search); |
589dc260 TM |
3443 | xe = &xh->xh_entries[0]; |
3444 | if (name_hash < le32_to_cpu(xe->xe_name_hash)) { | |
3445 | high_bucket = bucket - 1; | |
3446 | continue; | |
3447 | } | |
3448 | ||
3449 | /* | |
3450 | * Check whether the hash of the last entry in our | |
5a095611 TM |
3451 | * bucket is larger than the search one. for an empty |
3452 | * bucket, the last one is also the first one. | |
589dc260 | 3453 | */ |
5a095611 TM |
3454 | if (xh->xh_count) |
3455 | xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1]; | |
3456 | ||
589dc260 TM |
3457 | last_hash = le32_to_cpu(xe->xe_name_hash); |
3458 | ||
e2356a3f JB |
3459 | /* record lower_blkno which may be the insert place. */ |
3460 | lower_blkno = blkno; | |
589dc260 TM |
3461 | |
3462 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) { | |
3463 | low_bucket = bucket + 1; | |
3464 | continue; | |
3465 | } | |
3466 | ||
3467 | /* the searched xattr should reside in this bucket if exists. */ | |
e2356a3f | 3468 | ret = ocfs2_find_xe_in_bucket(inode, search, |
589dc260 TM |
3469 | name_index, name, name_hash, |
3470 | &index, &found); | |
3471 | if (ret) { | |
3472 | mlog_errno(ret); | |
3473 | goto out; | |
3474 | } | |
3475 | break; | |
3476 | } | |
3477 | ||
3478 | /* | |
3479 | * Record the bucket we have found. | |
3480 | * When the xattr's hash value is in the gap of 2 buckets, we will | |
3481 | * always set it to the previous bucket. | |
3482 | */ | |
e2356a3f JB |
3483 | if (!lower_blkno) |
3484 | lower_blkno = p_blkno; | |
3485 | ||
3486 | /* This should be in cache - we just read it during the search */ | |
3487 | ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno); | |
3488 | if (ret) { | |
3489 | mlog_errno(ret); | |
3490 | goto out; | |
589dc260 | 3491 | } |
589dc260 | 3492 | |
ba937127 JB |
3493 | xs->header = bucket_xh(xs->bucket); |
3494 | xs->base = bucket_block(xs->bucket, 0); | |
589dc260 TM |
3495 | xs->end = xs->base + inode->i_sb->s_blocksize; |
3496 | ||
3497 | if (found) { | |
589dc260 TM |
3498 | xs->here = &xs->header->xh_entries[index]; |
3499 | mlog(0, "find xattr %s in bucket %llu, entry = %u\n", name, | |
ba937127 | 3500 | (unsigned long long)bucket_blkno(xs->bucket), index); |
589dc260 TM |
3501 | } else |
3502 | ret = -ENODATA; | |
3503 | ||
3504 | out: | |
e2356a3f | 3505 | ocfs2_xattr_bucket_free(search); |
589dc260 TM |
3506 | return ret; |
3507 | } | |
3508 | ||
3509 | static int ocfs2_xattr_index_block_find(struct inode *inode, | |
3510 | struct buffer_head *root_bh, | |
3511 | int name_index, | |
3512 | const char *name, | |
3513 | struct ocfs2_xattr_search *xs) | |
3514 | { | |
3515 | int ret; | |
3516 | struct ocfs2_xattr_block *xb = | |
3517 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
3518 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
3519 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
3520 | u64 p_blkno = 0; | |
3521 | u32 first_hash, num_clusters = 0; | |
2057e5c6 | 3522 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
589dc260 TM |
3523 | |
3524 | if (le16_to_cpu(el->l_next_free_rec) == 0) | |
3525 | return -ENODATA; | |
3526 | ||
3527 | mlog(0, "find xattr %s, hash = %u, index = %d in xattr tree\n", | |
3528 | name, name_hash, name_index); | |
3529 | ||
3530 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash, | |
3531 | &num_clusters, el); | |
3532 | if (ret) { | |
3533 | mlog_errno(ret); | |
3534 | goto out; | |
3535 | } | |
3536 | ||
3537 | BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash); | |
3538 | ||
3539 | mlog(0, "find xattr extent rec %u clusters from %llu, the first hash " | |
de29c085 MF |
3540 | "in the rec is %u\n", num_clusters, (unsigned long long)p_blkno, |
3541 | first_hash); | |
589dc260 TM |
3542 | |
3543 | ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash, | |
3544 | p_blkno, first_hash, num_clusters, xs); | |
3545 | ||
3546 | out: | |
3547 | return ret; | |
3548 | } | |
3549 | ||
0c044f0b TM |
3550 | static int ocfs2_iterate_xattr_buckets(struct inode *inode, |
3551 | u64 blkno, | |
3552 | u32 clusters, | |
3553 | xattr_bucket_func *func, | |
3554 | void *para) | |
3555 | { | |
6dde41d9 | 3556 | int i, ret = 0; |
0c044f0b TM |
3557 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); |
3558 | u32 num_buckets = clusters * bpc; | |
ba937127 | 3559 | struct ocfs2_xattr_bucket *bucket; |
0c044f0b | 3560 | |
ba937127 JB |
3561 | bucket = ocfs2_xattr_bucket_new(inode); |
3562 | if (!bucket) { | |
3563 | mlog_errno(-ENOMEM); | |
3564 | return -ENOMEM; | |
3565 | } | |
0c044f0b TM |
3566 | |
3567 | mlog(0, "iterating xattr buckets in %u clusters starting from %llu\n", | |
de29c085 | 3568 | clusters, (unsigned long long)blkno); |
0c044f0b | 3569 | |
ba937127 JB |
3570 | for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) { |
3571 | ret = ocfs2_read_xattr_bucket(bucket, blkno); | |
0c044f0b TM |
3572 | if (ret) { |
3573 | mlog_errno(ret); | |
ba937127 | 3574 | break; |
0c044f0b TM |
3575 | } |
3576 | ||
0c044f0b TM |
3577 | /* |
3578 | * The real bucket num in this series of blocks is stored | |
3579 | * in the 1st bucket. | |
3580 | */ | |
3581 | if (i == 0) | |
ba937127 | 3582 | num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets); |
0c044f0b | 3583 | |
de29c085 MF |
3584 | mlog(0, "iterating xattr bucket %llu, first hash %u\n", |
3585 | (unsigned long long)blkno, | |
ba937127 | 3586 | le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); |
0c044f0b | 3587 | if (func) { |
ba937127 | 3588 | ret = func(inode, bucket, para); |
a46fa684 | 3589 | if (ret && ret != -ERANGE) |
0c044f0b | 3590 | mlog_errno(ret); |
ba937127 | 3591 | /* Fall through to bucket_relse() */ |
0c044f0b TM |
3592 | } |
3593 | ||
ba937127 JB |
3594 | ocfs2_xattr_bucket_relse(bucket); |
3595 | if (ret) | |
3596 | break; | |
0c044f0b TM |
3597 | } |
3598 | ||
ba937127 | 3599 | ocfs2_xattr_bucket_free(bucket); |
0c044f0b TM |
3600 | return ret; |
3601 | } | |
3602 | ||
3603 | struct ocfs2_xattr_tree_list { | |
3604 | char *buffer; | |
3605 | size_t buffer_size; | |
936b8834 | 3606 | size_t result; |
0c044f0b TM |
3607 | }; |
3608 | ||
fd68a894 | 3609 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
0c044f0b TM |
3610 | struct ocfs2_xattr_header *xh, |
3611 | int index, | |
3612 | int *block_off, | |
3613 | int *new_offset) | |
3614 | { | |
3615 | u16 name_offset; | |
3616 | ||
3617 | if (index < 0 || index >= le16_to_cpu(xh->xh_count)) | |
3618 | return -EINVAL; | |
3619 | ||
3620 | name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset); | |
3621 | ||
fd68a894 TM |
3622 | *block_off = name_offset >> sb->s_blocksize_bits; |
3623 | *new_offset = name_offset % sb->s_blocksize; | |
0c044f0b TM |
3624 | |
3625 | return 0; | |
3626 | } | |
3627 | ||
3628 | static int ocfs2_list_xattr_bucket(struct inode *inode, | |
3629 | struct ocfs2_xattr_bucket *bucket, | |
3630 | void *para) | |
3631 | { | |
936b8834 | 3632 | int ret = 0, type; |
0c044f0b | 3633 | struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para; |
0c044f0b | 3634 | int i, block_off, new_offset; |
936b8834 | 3635 | const char *prefix, *name; |
0c044f0b | 3636 | |
3e632946 JB |
3637 | for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) { |
3638 | struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i]; | |
936b8834 TM |
3639 | type = ocfs2_xattr_get_type(entry); |
3640 | prefix = ocfs2_xattr_prefix(type); | |
0c044f0b | 3641 | |
936b8834 | 3642 | if (prefix) { |
fd68a894 | 3643 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
3e632946 | 3644 | bucket_xh(bucket), |
0c044f0b TM |
3645 | i, |
3646 | &block_off, | |
3647 | &new_offset); | |
3648 | if (ret) | |
3649 | break; | |
936b8834 | 3650 | |
51def39f | 3651 | name = (const char *)bucket_block(bucket, block_off) + |
936b8834 TM |
3652 | new_offset; |
3653 | ret = ocfs2_xattr_list_entry(xl->buffer, | |
3654 | xl->buffer_size, | |
3655 | &xl->result, | |
3656 | prefix, name, | |
3657 | entry->xe_name_len); | |
3658 | if (ret) | |
3659 | break; | |
0c044f0b TM |
3660 | } |
3661 | } | |
3662 | ||
3663 | return ret; | |
3664 | } | |
3665 | ||
47bca495 TM |
3666 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, |
3667 | struct buffer_head *blk_bh, | |
3668 | xattr_tree_rec_func *rec_func, | |
3669 | void *para) | |
0c044f0b | 3670 | { |
47bca495 TM |
3671 | struct ocfs2_xattr_block *xb = |
3672 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
3673 | struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; | |
0c044f0b TM |
3674 | int ret = 0; |
3675 | u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0; | |
3676 | u64 p_blkno = 0; | |
0c044f0b | 3677 | |
47bca495 | 3678 | if (!el->l_next_free_rec || !rec_func) |
0c044f0b TM |
3679 | return 0; |
3680 | ||
3681 | while (name_hash > 0) { | |
3682 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, | |
3683 | &e_cpos, &num_clusters, el); | |
3684 | if (ret) { | |
3685 | mlog_errno(ret); | |
47bca495 | 3686 | break; |
0c044f0b TM |
3687 | } |
3688 | ||
47bca495 TM |
3689 | ret = rec_func(inode, blk_bh, p_blkno, e_cpos, |
3690 | num_clusters, para); | |
0c044f0b | 3691 | if (ret) { |
a46fa684 TM |
3692 | if (ret != -ERANGE) |
3693 | mlog_errno(ret); | |
47bca495 | 3694 | break; |
0c044f0b TM |
3695 | } |
3696 | ||
3697 | if (e_cpos == 0) | |
3698 | break; | |
3699 | ||
3700 | name_hash = e_cpos - 1; | |
3701 | } | |
3702 | ||
47bca495 TM |
3703 | return ret; |
3704 | ||
3705 | } | |
3706 | ||
3707 | static int ocfs2_list_xattr_tree_rec(struct inode *inode, | |
3708 | struct buffer_head *root_bh, | |
3709 | u64 blkno, u32 cpos, u32 len, void *para) | |
3710 | { | |
3711 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
3712 | ocfs2_list_xattr_bucket, para); | |
3713 | } | |
3714 | ||
3715 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, | |
3716 | struct buffer_head *blk_bh, | |
3717 | char *buffer, | |
3718 | size_t buffer_size) | |
3719 | { | |
3720 | int ret; | |
3721 | struct ocfs2_xattr_tree_list xl = { | |
3722 | .buffer = buffer, | |
3723 | .buffer_size = buffer_size, | |
3724 | .result = 0, | |
3725 | }; | |
3726 | ||
3727 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
3728 | ocfs2_list_xattr_tree_rec, &xl); | |
3729 | if (ret) { | |
3730 | mlog_errno(ret); | |
3731 | goto out; | |
3732 | } | |
3733 | ||
936b8834 | 3734 | ret = xl.result; |
0c044f0b TM |
3735 | out: |
3736 | return ret; | |
3737 | } | |
01225596 TM |
3738 | |
3739 | static int cmp_xe(const void *a, const void *b) | |
3740 | { | |
3741 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
3742 | u32 l_hash = le32_to_cpu(l->xe_name_hash); | |
3743 | u32 r_hash = le32_to_cpu(r->xe_name_hash); | |
3744 | ||
3745 | if (l_hash > r_hash) | |
3746 | return 1; | |
3747 | if (l_hash < r_hash) | |
3748 | return -1; | |
3749 | return 0; | |
3750 | } | |
3751 | ||
3752 | static void swap_xe(void *a, void *b, int size) | |
3753 | { | |
3754 | struct ocfs2_xattr_entry *l = a, *r = b, tmp; | |
3755 | ||
3756 | tmp = *l; | |
3757 | memcpy(l, r, sizeof(struct ocfs2_xattr_entry)); | |
3758 | memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry)); | |
3759 | } | |
3760 | ||
3761 | /* | |
3762 | * When the ocfs2_xattr_block is filled up, new bucket will be created | |
3763 | * and all the xattr entries will be moved to the new bucket. | |
178eeac3 JB |
3764 | * The header goes at the start of the bucket, and the names+values are |
3765 | * filled from the end. This is why *target starts as the last buffer. | |
01225596 TM |
3766 | * Note: we need to sort the entries since they are not saved in order |
3767 | * in the ocfs2_xattr_block. | |
3768 | */ | |
3769 | static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, | |
3770 | struct buffer_head *xb_bh, | |
178eeac3 | 3771 | struct ocfs2_xattr_bucket *bucket) |
01225596 TM |
3772 | { |
3773 | int i, blocksize = inode->i_sb->s_blocksize; | |
178eeac3 | 3774 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 TM |
3775 | u16 offset, size, off_change; |
3776 | struct ocfs2_xattr_entry *xe; | |
3777 | struct ocfs2_xattr_block *xb = | |
3778 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
3779 | struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header; | |
178eeac3 | 3780 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 3781 | u16 count = le16_to_cpu(xb_xh->xh_count); |
178eeac3 JB |
3782 | char *src = xb_bh->b_data; |
3783 | char *target = bucket_block(bucket, blks - 1); | |
01225596 TM |
3784 | |
3785 | mlog(0, "cp xattr from block %llu to bucket %llu\n", | |
3786 | (unsigned long long)xb_bh->b_blocknr, | |
178eeac3 JB |
3787 | (unsigned long long)bucket_blkno(bucket)); |
3788 | ||
3789 | for (i = 0; i < blks; i++) | |
3790 | memset(bucket_block(bucket, i), 0, blocksize); | |
01225596 | 3791 | |
01225596 TM |
3792 | /* |
3793 | * Since the xe_name_offset is based on ocfs2_xattr_header, | |
3794 | * there is a offset change corresponding to the change of | |
3795 | * ocfs2_xattr_header's position. | |
3796 | */ | |
3797 | off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
3798 | xe = &xb_xh->xh_entries[count - 1]; | |
3799 | offset = le16_to_cpu(xe->xe_name_offset) + off_change; | |
3800 | size = blocksize - offset; | |
3801 | ||
3802 | /* copy all the names and values. */ | |
01225596 TM |
3803 | memcpy(target + offset, src + offset, size); |
3804 | ||
3805 | /* Init new header now. */ | |
3806 | xh->xh_count = xb_xh->xh_count; | |
3807 | xh->xh_num_buckets = cpu_to_le16(1); | |
3808 | xh->xh_name_value_len = cpu_to_le16(size); | |
3809 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size); | |
3810 | ||
3811 | /* copy all the entries. */ | |
178eeac3 | 3812 | target = bucket_block(bucket, 0); |
01225596 TM |
3813 | offset = offsetof(struct ocfs2_xattr_header, xh_entries); |
3814 | size = count * sizeof(struct ocfs2_xattr_entry); | |
3815 | memcpy(target + offset, (char *)xb_xh + offset, size); | |
3816 | ||
3817 | /* Change the xe offset for all the xe because of the move. */ | |
3818 | off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize + | |
3819 | offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
3820 | for (i = 0; i < count; i++) | |
3821 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change); | |
3822 | ||
3823 | mlog(0, "copy entry: start = %u, size = %u, offset_change = %u\n", | |
3824 | offset, size, off_change); | |
3825 | ||
3826 | sort(target + offset, count, sizeof(struct ocfs2_xattr_entry), | |
3827 | cmp_xe, swap_xe); | |
3828 | } | |
3829 | ||
3830 | /* | |
3831 | * After we move xattr from block to index btree, we have to | |
3832 | * update ocfs2_xattr_search to the new xe and base. | |
3833 | * | |
3834 | * When the entry is in xattr block, xattr_bh indicates the storage place. | |
3835 | * While if the entry is in index b-tree, "bucket" indicates the | |
3836 | * real place of the xattr. | |
3837 | */ | |
178eeac3 JB |
3838 | static void ocfs2_xattr_update_xattr_search(struct inode *inode, |
3839 | struct ocfs2_xattr_search *xs, | |
3840 | struct buffer_head *old_bh) | |
01225596 | 3841 | { |
01225596 TM |
3842 | char *buf = old_bh->b_data; |
3843 | struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf; | |
3844 | struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header; | |
178eeac3 | 3845 | int i; |
01225596 | 3846 | |
ba937127 | 3847 | xs->header = bucket_xh(xs->bucket); |
178eeac3 | 3848 | xs->base = bucket_block(xs->bucket, 0); |
01225596 TM |
3849 | xs->end = xs->base + inode->i_sb->s_blocksize; |
3850 | ||
178eeac3 JB |
3851 | if (xs->not_found) |
3852 | return; | |
01225596 | 3853 | |
178eeac3 JB |
3854 | i = xs->here - old_xh->xh_entries; |
3855 | xs->here = &xs->header->xh_entries[i]; | |
01225596 TM |
3856 | } |
3857 | ||
3858 | static int ocfs2_xattr_create_index_block(struct inode *inode, | |
78f30c31 TM |
3859 | struct ocfs2_xattr_search *xs, |
3860 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 3861 | { |
85db90e7 | 3862 | int ret; |
01225596 TM |
3863 | u32 bit_off, len; |
3864 | u64 blkno; | |
85db90e7 | 3865 | handle_t *handle = ctxt->handle; |
01225596 TM |
3866 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
3867 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
01225596 TM |
3868 | struct buffer_head *xb_bh = xs->xattr_bh; |
3869 | struct ocfs2_xattr_block *xb = | |
3870 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
3871 | struct ocfs2_xattr_tree_root *xr; | |
3872 | u16 xb_flags = le16_to_cpu(xb->xb_flags); | |
01225596 TM |
3873 | |
3874 | mlog(0, "create xattr index block for %llu\n", | |
3875 | (unsigned long long)xb_bh->b_blocknr); | |
3876 | ||
3877 | BUG_ON(xb_flags & OCFS2_XATTR_INDEXED); | |
178eeac3 | 3878 | BUG_ON(!xs->bucket); |
01225596 | 3879 | |
01225596 TM |
3880 | /* |
3881 | * XXX: | |
3882 | * We can use this lock for now, and maybe move to a dedicated mutex | |
3883 | * if performance becomes a problem later. | |
3884 | */ | |
3885 | down_write(&oi->ip_alloc_sem); | |
3886 | ||
0cf2f763 | 3887 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh, |
84008972 | 3888 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
3889 | if (ret) { |
3890 | mlog_errno(ret); | |
85db90e7 | 3891 | goto out; |
01225596 TM |
3892 | } |
3893 | ||
78f30c31 TM |
3894 | ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, |
3895 | 1, 1, &bit_off, &len); | |
01225596 TM |
3896 | if (ret) { |
3897 | mlog_errno(ret); | |
85db90e7 | 3898 | goto out; |
01225596 TM |
3899 | } |
3900 | ||
3901 | /* | |
3902 | * The bucket may spread in many blocks, and | |
3903 | * we will only touch the 1st block and the last block | |
3904 | * in the whole bucket(one for entry and one for data). | |
3905 | */ | |
3906 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); | |
3907 | ||
de29c085 MF |
3908 | mlog(0, "allocate 1 cluster from %llu to xattr block\n", |
3909 | (unsigned long long)blkno); | |
01225596 | 3910 | |
178eeac3 JB |
3911 | ret = ocfs2_init_xattr_bucket(xs->bucket, blkno); |
3912 | if (ret) { | |
01225596 | 3913 | mlog_errno(ret); |
85db90e7 | 3914 | goto out; |
01225596 TM |
3915 | } |
3916 | ||
178eeac3 JB |
3917 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
3918 | OCFS2_JOURNAL_ACCESS_CREATE); | |
01225596 TM |
3919 | if (ret) { |
3920 | mlog_errno(ret); | |
85db90e7 | 3921 | goto out; |
01225596 TM |
3922 | } |
3923 | ||
178eeac3 JB |
3924 | ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket); |
3925 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); | |
01225596 | 3926 | |
178eeac3 | 3927 | ocfs2_xattr_update_xattr_search(inode, xs, xb_bh); |
01225596 TM |
3928 | |
3929 | /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */ | |
3930 | memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize - | |
3931 | offsetof(struct ocfs2_xattr_block, xb_attrs)); | |
3932 | ||
3933 | xr = &xb->xb_attrs.xb_root; | |
3934 | xr->xt_clusters = cpu_to_le32(1); | |
3935 | xr->xt_last_eb_blk = 0; | |
3936 | xr->xt_list.l_tree_depth = 0; | |
3937 | xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
3938 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
3939 | ||
3940 | xr->xt_list.l_recs[0].e_cpos = 0; | |
3941 | xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno); | |
3942 | xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1); | |
3943 | ||
3944 | xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED); | |
3945 | ||
85db90e7 | 3946 | ocfs2_journal_dirty(handle, xb_bh); |
01225596 | 3947 | |
85db90e7 | 3948 | out: |
01225596 TM |
3949 | up_write(&oi->ip_alloc_sem); |
3950 | ||
01225596 TM |
3951 | return ret; |
3952 | } | |
3953 | ||
3954 | static int cmp_xe_offset(const void *a, const void *b) | |
3955 | { | |
3956 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
3957 | u32 l_name_offset = le16_to_cpu(l->xe_name_offset); | |
3958 | u32 r_name_offset = le16_to_cpu(r->xe_name_offset); | |
3959 | ||
3960 | if (l_name_offset < r_name_offset) | |
3961 | return 1; | |
3962 | if (l_name_offset > r_name_offset) | |
3963 | return -1; | |
3964 | return 0; | |
3965 | } | |
3966 | ||
3967 | /* | |
3968 | * defrag a xattr bucket if we find that the bucket has some | |
3969 | * holes beteen name/value pairs. | |
3970 | * We will move all the name/value pairs to the end of the bucket | |
3971 | * so that we can spare some space for insertion. | |
3972 | */ | |
3973 | static int ocfs2_defrag_xattr_bucket(struct inode *inode, | |
85db90e7 | 3974 | handle_t *handle, |
01225596 TM |
3975 | struct ocfs2_xattr_bucket *bucket) |
3976 | { | |
3977 | int ret, i; | |
3978 | size_t end, offset, len, value_len; | |
3979 | struct ocfs2_xattr_header *xh; | |
3980 | char *entries, *buf, *bucket_buf = NULL; | |
9c7759aa | 3981 | u64 blkno = bucket_blkno(bucket); |
01225596 | 3982 | u16 xh_free_start; |
01225596 | 3983 | size_t blocksize = inode->i_sb->s_blocksize; |
01225596 | 3984 | struct ocfs2_xattr_entry *xe; |
01225596 TM |
3985 | |
3986 | /* | |
3987 | * In order to make the operation more efficient and generic, | |
3988 | * we copy all the blocks into a contiguous memory and do the | |
3989 | * defragment there, so if anything is error, we will not touch | |
3990 | * the real block. | |
3991 | */ | |
3992 | bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS); | |
3993 | if (!bucket_buf) { | |
3994 | ret = -EIO; | |
3995 | goto out; | |
3996 | } | |
3997 | ||
3998 | buf = bucket_buf; | |
1c32a2fd TM |
3999 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4000 | memcpy(buf, bucket_block(bucket, i), blocksize); | |
01225596 | 4001 | |
1c32a2fd | 4002 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, |
161d6f30 JB |
4003 | OCFS2_JOURNAL_ACCESS_WRITE); |
4004 | if (ret < 0) { | |
4005 | mlog_errno(ret); | |
85db90e7 | 4006 | goto out; |
01225596 TM |
4007 | } |
4008 | ||
4009 | xh = (struct ocfs2_xattr_header *)bucket_buf; | |
4010 | entries = (char *)xh->xh_entries; | |
4011 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4012 | ||
4013 | mlog(0, "adjust xattr bucket in %llu, count = %u, " | |
4014 | "xh_free_start = %u, xh_name_value_len = %u.\n", | |
de29c085 MF |
4015 | (unsigned long long)blkno, le16_to_cpu(xh->xh_count), |
4016 | xh_free_start, le16_to_cpu(xh->xh_name_value_len)); | |
01225596 TM |
4017 | |
4018 | /* | |
4019 | * sort all the entries by their offset. | |
4020 | * the largest will be the first, so that we can | |
4021 | * move them to the end one by one. | |
4022 | */ | |
4023 | sort(entries, le16_to_cpu(xh->xh_count), | |
4024 | sizeof(struct ocfs2_xattr_entry), | |
4025 | cmp_xe_offset, swap_xe); | |
4026 | ||
4027 | /* Move all name/values to the end of the bucket. */ | |
4028 | xe = xh->xh_entries; | |
4029 | end = OCFS2_XATTR_BUCKET_SIZE; | |
4030 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) { | |
4031 | offset = le16_to_cpu(xe->xe_name_offset); | |
4032 | if (ocfs2_xattr_is_local(xe)) | |
4033 | value_len = OCFS2_XATTR_SIZE( | |
4034 | le64_to_cpu(xe->xe_value_size)); | |
4035 | else | |
4036 | value_len = OCFS2_XATTR_ROOT_SIZE; | |
4037 | len = OCFS2_XATTR_SIZE(xe->xe_name_len) + value_len; | |
4038 | ||
4039 | /* | |
4040 | * We must make sure that the name/value pair | |
4041 | * exist in the same block. So adjust end to | |
4042 | * the previous block end if needed. | |
4043 | */ | |
4044 | if (((end - len) / blocksize != | |
4045 | (end - 1) / blocksize)) | |
4046 | end = end - end % blocksize; | |
4047 | ||
4048 | if (end > offset + len) { | |
4049 | memmove(bucket_buf + end - len, | |
4050 | bucket_buf + offset, len); | |
4051 | xe->xe_name_offset = cpu_to_le16(end - len); | |
4052 | } | |
4053 | ||
4054 | mlog_bug_on_msg(end < offset + len, "Defrag check failed for " | |
4055 | "bucket %llu\n", (unsigned long long)blkno); | |
4056 | ||
4057 | end -= len; | |
4058 | } | |
4059 | ||
4060 | mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for " | |
4061 | "bucket %llu\n", (unsigned long long)blkno); | |
4062 | ||
4063 | if (xh_free_start == end) | |
85db90e7 | 4064 | goto out; |
01225596 TM |
4065 | |
4066 | memset(bucket_buf + xh_free_start, 0, end - xh_free_start); | |
4067 | xh->xh_free_start = cpu_to_le16(end); | |
4068 | ||
4069 | /* sort the entries by their name_hash. */ | |
4070 | sort(entries, le16_to_cpu(xh->xh_count), | |
4071 | sizeof(struct ocfs2_xattr_entry), | |
4072 | cmp_xe, swap_xe); | |
4073 | ||
4074 | buf = bucket_buf; | |
1c32a2fd TM |
4075 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4076 | memcpy(bucket_block(bucket, i), buf, blocksize); | |
4077 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
01225596 | 4078 | |
01225596 | 4079 | out: |
01225596 TM |
4080 | kfree(bucket_buf); |
4081 | return ret; | |
4082 | } | |
4083 | ||
4084 | /* | |
b5c03e74 JB |
4085 | * prev_blkno points to the start of an existing extent. new_blkno |
4086 | * points to a newly allocated extent. Because we know each of our | |
4087 | * clusters contains more than bucket, we can easily split one cluster | |
4088 | * at a bucket boundary. So we take the last cluster of the existing | |
4089 | * extent and split it down the middle. We move the last half of the | |
4090 | * buckets in the last cluster of the existing extent over to the new | |
4091 | * extent. | |
01225596 | 4092 | * |
b5c03e74 JB |
4093 | * first_bh is the buffer at prev_blkno so we can update the existing |
4094 | * extent's bucket count. header_bh is the bucket were we were hoping | |
4095 | * to insert our xattr. If the bucket move places the target in the new | |
4096 | * extent, we'll update first_bh and header_bh after modifying the old | |
4097 | * extent. | |
4098 | * | |
4099 | * first_hash will be set as the 1st xe's name_hash in the new extent. | |
01225596 TM |
4100 | */ |
4101 | static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode, | |
4102 | handle_t *handle, | |
41cb8148 JB |
4103 | struct ocfs2_xattr_bucket *first, |
4104 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4105 | u64 new_blkno, |
01225596 TM |
4106 | u32 num_clusters, |
4107 | u32 *first_hash) | |
4108 | { | |
c58b6032 | 4109 | int ret; |
41cb8148 JB |
4110 | struct super_block *sb = inode->i_sb; |
4111 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb); | |
4112 | int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
b5c03e74 | 4113 | int to_move = num_buckets / 2; |
c58b6032 | 4114 | u64 src_blkno; |
41cb8148 JB |
4115 | u64 last_cluster_blkno = bucket_blkno(first) + |
4116 | ((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1)); | |
01225596 | 4117 | |
41cb8148 JB |
4118 | BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets); |
4119 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize); | |
01225596 | 4120 | |
01225596 | 4121 | mlog(0, "move half of xattrs in cluster %llu to %llu\n", |
c58b6032 | 4122 | (unsigned long long)last_cluster_blkno, (unsigned long long)new_blkno); |
01225596 | 4123 | |
41cb8148 | 4124 | ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first), |
c58b6032 JB |
4125 | last_cluster_blkno, new_blkno, |
4126 | to_move, first_hash); | |
01225596 TM |
4127 | if (ret) { |
4128 | mlog_errno(ret); | |
4129 | goto out; | |
4130 | } | |
4131 | ||
c58b6032 JB |
4132 | /* This is the first bucket that got moved */ |
4133 | src_blkno = last_cluster_blkno + (to_move * blks_per_bucket); | |
01225596 | 4134 | |
b5c03e74 | 4135 | /* |
c58b6032 | 4136 | * If the target bucket was part of the moved buckets, we need to |
41cb8148 | 4137 | * update first and target. |
b5c03e74 | 4138 | */ |
41cb8148 | 4139 | if (bucket_blkno(target) >= src_blkno) { |
b5c03e74 JB |
4140 | /* Find the block for the new target bucket */ |
4141 | src_blkno = new_blkno + | |
41cb8148 JB |
4142 | (bucket_blkno(target) - src_blkno); |
4143 | ||
4144 | ocfs2_xattr_bucket_relse(first); | |
4145 | ocfs2_xattr_bucket_relse(target); | |
01225596 | 4146 | |
b5c03e74 | 4147 | /* |
c58b6032 | 4148 | * These shouldn't fail - the buffers are in the |
b5c03e74 JB |
4149 | * journal from ocfs2_cp_xattr_bucket(). |
4150 | */ | |
41cb8148 | 4151 | ret = ocfs2_read_xattr_bucket(first, new_blkno); |
c58b6032 JB |
4152 | if (ret) { |
4153 | mlog_errno(ret); | |
4154 | goto out; | |
4155 | } | |
41cb8148 JB |
4156 | ret = ocfs2_read_xattr_bucket(target, src_blkno); |
4157 | if (ret) | |
b5c03e74 | 4158 | mlog_errno(ret); |
01225596 | 4159 | |
01225596 TM |
4160 | } |
4161 | ||
01225596 | 4162 | out: |
01225596 TM |
4163 | return ret; |
4164 | } | |
4165 | ||
01225596 | 4166 | /* |
80bcaf34 TM |
4167 | * Find the suitable pos when we divide a bucket into 2. |
4168 | * We have to make sure the xattrs with the same hash value exist | |
4169 | * in the same bucket. | |
4170 | * | |
4171 | * If this ocfs2_xattr_header covers more than one hash value, find a | |
4172 | * place where the hash value changes. Try to find the most even split. | |
4173 | * The most common case is that all entries have different hash values, | |
4174 | * and the first check we make will find a place to split. | |
4175 | */ | |
4176 | static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh) | |
4177 | { | |
4178 | struct ocfs2_xattr_entry *entries = xh->xh_entries; | |
4179 | int count = le16_to_cpu(xh->xh_count); | |
4180 | int delta, middle = count / 2; | |
4181 | ||
4182 | /* | |
4183 | * We start at the middle. Each step gets farther away in both | |
4184 | * directions. We therefore hit the change in hash value | |
4185 | * nearest to the middle. Note that this loop does not execute for | |
4186 | * count < 2. | |
4187 | */ | |
4188 | for (delta = 0; delta < middle; delta++) { | |
4189 | /* Let's check delta earlier than middle */ | |
4190 | if (cmp_xe(&entries[middle - delta - 1], | |
4191 | &entries[middle - delta])) | |
4192 | return middle - delta; | |
4193 | ||
4194 | /* For even counts, don't walk off the end */ | |
4195 | if ((middle + delta + 1) == count) | |
4196 | continue; | |
4197 | ||
4198 | /* Now try delta past middle */ | |
4199 | if (cmp_xe(&entries[middle + delta], | |
4200 | &entries[middle + delta + 1])) | |
4201 | return middle + delta + 1; | |
4202 | } | |
4203 | ||
4204 | /* Every entry had the same hash */ | |
4205 | return count; | |
4206 | } | |
4207 | ||
4208 | /* | |
4209 | * Move some xattrs in old bucket(blk) to new bucket(new_blk). | |
01225596 | 4210 | * first_hash will record the 1st hash of the new bucket. |
80bcaf34 TM |
4211 | * |
4212 | * Normally half of the xattrs will be moved. But we have to make | |
4213 | * sure that the xattrs with the same hash value are stored in the | |
4214 | * same bucket. If all the xattrs in this bucket have the same hash | |
4215 | * value, the new bucket will be initialized as an empty one and the | |
4216 | * first_hash will be initialized as (hash_value+1). | |
01225596 | 4217 | */ |
80bcaf34 TM |
4218 | static int ocfs2_divide_xattr_bucket(struct inode *inode, |
4219 | handle_t *handle, | |
4220 | u64 blk, | |
4221 | u64 new_blk, | |
4222 | u32 *first_hash, | |
4223 | int new_bucket_head) | |
01225596 TM |
4224 | { |
4225 | int ret, i; | |
80bcaf34 | 4226 | int count, start, len, name_value_len = 0, xe_len, name_offset = 0; |
ba937127 | 4227 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4228 | struct ocfs2_xattr_header *xh; |
4229 | struct ocfs2_xattr_entry *xe; | |
4230 | int blocksize = inode->i_sb->s_blocksize; | |
4231 | ||
80bcaf34 | 4232 | mlog(0, "move some of xattrs from bucket %llu to %llu\n", |
de29c085 | 4233 | (unsigned long long)blk, (unsigned long long)new_blk); |
01225596 | 4234 | |
ba937127 JB |
4235 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4236 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4237 | if (!s_bucket || !t_bucket) { | |
4238 | ret = -ENOMEM; | |
4239 | mlog_errno(ret); | |
4240 | goto out; | |
4241 | } | |
01225596 | 4242 | |
ba937127 | 4243 | ret = ocfs2_read_xattr_bucket(s_bucket, blk); |
01225596 TM |
4244 | if (ret) { |
4245 | mlog_errno(ret); | |
4246 | goto out; | |
4247 | } | |
4248 | ||
ba937127 | 4249 | ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket, |
1224be02 | 4250 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4251 | if (ret) { |
4252 | mlog_errno(ret); | |
4253 | goto out; | |
4254 | } | |
4255 | ||
784b816a JB |
4256 | /* |
4257 | * Even if !new_bucket_head, we're overwriting t_bucket. Thus, | |
4258 | * there's no need to read it. | |
4259 | */ | |
ba937127 | 4260 | ret = ocfs2_init_xattr_bucket(t_bucket, new_blk); |
01225596 TM |
4261 | if (ret) { |
4262 | mlog_errno(ret); | |
4263 | goto out; | |
4264 | } | |
4265 | ||
2b656c1d JB |
4266 | /* |
4267 | * Hey, if we're overwriting t_bucket, what difference does | |
4268 | * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the | |
4269 | * same part of ocfs2_cp_xattr_bucket(). | |
4270 | */ | |
ba937127 | 4271 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4272 | new_bucket_head ? |
4273 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4274 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4275 | if (ret) { | |
4276 | mlog_errno(ret); | |
4277 | goto out; | |
01225596 TM |
4278 | } |
4279 | ||
ba937127 | 4280 | xh = bucket_xh(s_bucket); |
80bcaf34 TM |
4281 | count = le16_to_cpu(xh->xh_count); |
4282 | start = ocfs2_xattr_find_divide_pos(xh); | |
4283 | ||
4284 | if (start == count) { | |
4285 | xe = &xh->xh_entries[start-1]; | |
4286 | ||
4287 | /* | |
4288 | * initialized a new empty bucket here. | |
4289 | * The hash value is set as one larger than | |
4290 | * that of the last entry in the previous bucket. | |
4291 | */ | |
ba937127 JB |
4292 | for (i = 0; i < t_bucket->bu_blocks; i++) |
4293 | memset(bucket_block(t_bucket, i), 0, blocksize); | |
80bcaf34 | 4294 | |
ba937127 | 4295 | xh = bucket_xh(t_bucket); |
80bcaf34 TM |
4296 | xh->xh_free_start = cpu_to_le16(blocksize); |
4297 | xh->xh_entries[0].xe_name_hash = xe->xe_name_hash; | |
4298 | le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1); | |
4299 | ||
4300 | goto set_num_buckets; | |
4301 | } | |
4302 | ||
01225596 | 4303 | /* copy the whole bucket to the new first. */ |
ba937127 | 4304 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
01225596 TM |
4305 | |
4306 | /* update the new bucket. */ | |
ba937127 | 4307 | xh = bucket_xh(t_bucket); |
01225596 TM |
4308 | |
4309 | /* | |
4310 | * Calculate the total name/value len and xh_free_start for | |
4311 | * the old bucket first. | |
4312 | */ | |
4313 | name_offset = OCFS2_XATTR_BUCKET_SIZE; | |
4314 | name_value_len = 0; | |
4315 | for (i = 0; i < start; i++) { | |
4316 | xe = &xh->xh_entries[i]; | |
4317 | xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
4318 | if (ocfs2_xattr_is_local(xe)) | |
4319 | xe_len += | |
4320 | OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); | |
4321 | else | |
4322 | xe_len += OCFS2_XATTR_ROOT_SIZE; | |
4323 | name_value_len += xe_len; | |
4324 | if (le16_to_cpu(xe->xe_name_offset) < name_offset) | |
4325 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
4326 | } | |
4327 | ||
4328 | /* | |
4329 | * Now begin the modification to the new bucket. | |
4330 | * | |
4331 | * In the new bucket, We just move the xattr entry to the beginning | |
4332 | * and don't touch the name/value. So there will be some holes in the | |
4333 | * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is | |
4334 | * called. | |
4335 | */ | |
4336 | xe = &xh->xh_entries[start]; | |
4337 | len = sizeof(struct ocfs2_xattr_entry) * (count - start); | |
4338 | mlog(0, "mv xattr entry len %d from %d to %d\n", len, | |
ff1ec20e MF |
4339 | (int)((char *)xe - (char *)xh), |
4340 | (int)((char *)xh->xh_entries - (char *)xh)); | |
01225596 TM |
4341 | memmove((char *)xh->xh_entries, (char *)xe, len); |
4342 | xe = &xh->xh_entries[count - start]; | |
4343 | len = sizeof(struct ocfs2_xattr_entry) * start; | |
4344 | memset((char *)xe, 0, len); | |
4345 | ||
4346 | le16_add_cpu(&xh->xh_count, -start); | |
4347 | le16_add_cpu(&xh->xh_name_value_len, -name_value_len); | |
4348 | ||
4349 | /* Calculate xh_free_start for the new bucket. */ | |
4350 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); | |
4351 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
4352 | xe = &xh->xh_entries[i]; | |
4353 | xe_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
4354 | if (ocfs2_xattr_is_local(xe)) | |
4355 | xe_len += | |
4356 | OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); | |
4357 | else | |
4358 | xe_len += OCFS2_XATTR_ROOT_SIZE; | |
4359 | if (le16_to_cpu(xe->xe_name_offset) < | |
4360 | le16_to_cpu(xh->xh_free_start)) | |
4361 | xh->xh_free_start = xe->xe_name_offset; | |
4362 | } | |
4363 | ||
80bcaf34 | 4364 | set_num_buckets: |
01225596 TM |
4365 | /* set xh->xh_num_buckets for the new xh. */ |
4366 | if (new_bucket_head) | |
4367 | xh->xh_num_buckets = cpu_to_le16(1); | |
4368 | else | |
4369 | xh->xh_num_buckets = 0; | |
4370 | ||
ba937127 | 4371 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); |
01225596 TM |
4372 | |
4373 | /* store the first_hash of the new bucket. */ | |
4374 | if (first_hash) | |
4375 | *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash); | |
4376 | ||
4377 | /* | |
80bcaf34 TM |
4378 | * Now only update the 1st block of the old bucket. If we |
4379 | * just added a new empty bucket, there is no need to modify | |
4380 | * it. | |
01225596 | 4381 | */ |
80bcaf34 TM |
4382 | if (start == count) |
4383 | goto out; | |
4384 | ||
ba937127 | 4385 | xh = bucket_xh(s_bucket); |
01225596 TM |
4386 | memset(&xh->xh_entries[start], 0, |
4387 | sizeof(struct ocfs2_xattr_entry) * (count - start)); | |
4388 | xh->xh_count = cpu_to_le16(start); | |
4389 | xh->xh_free_start = cpu_to_le16(name_offset); | |
4390 | xh->xh_name_value_len = cpu_to_le16(name_value_len); | |
4391 | ||
ba937127 | 4392 | ocfs2_xattr_bucket_journal_dirty(handle, s_bucket); |
01225596 TM |
4393 | |
4394 | out: | |
ba937127 JB |
4395 | ocfs2_xattr_bucket_free(s_bucket); |
4396 | ocfs2_xattr_bucket_free(t_bucket); | |
01225596 TM |
4397 | |
4398 | return ret; | |
4399 | } | |
4400 | ||
4401 | /* | |
4402 | * Copy xattr from one bucket to another bucket. | |
4403 | * | |
4404 | * The caller must make sure that the journal transaction | |
4405 | * has enough space for journaling. | |
4406 | */ | |
4407 | static int ocfs2_cp_xattr_bucket(struct inode *inode, | |
4408 | handle_t *handle, | |
4409 | u64 s_blkno, | |
4410 | u64 t_blkno, | |
4411 | int t_is_new) | |
4412 | { | |
4980c6da | 4413 | int ret; |
ba937127 | 4414 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4415 | |
4416 | BUG_ON(s_blkno == t_blkno); | |
4417 | ||
4418 | mlog(0, "cp bucket %llu to %llu, target is %d\n", | |
de29c085 MF |
4419 | (unsigned long long)s_blkno, (unsigned long long)t_blkno, |
4420 | t_is_new); | |
01225596 | 4421 | |
ba937127 JB |
4422 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4423 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4424 | if (!s_bucket || !t_bucket) { | |
4425 | ret = -ENOMEM; | |
4426 | mlog_errno(ret); | |
4427 | goto out; | |
4428 | } | |
92de109a | 4429 | |
ba937127 | 4430 | ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno); |
01225596 TM |
4431 | if (ret) |
4432 | goto out; | |
4433 | ||
784b816a JB |
4434 | /* |
4435 | * Even if !t_is_new, we're overwriting t_bucket. Thus, | |
4436 | * there's no need to read it. | |
4437 | */ | |
ba937127 | 4438 | ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno); |
01225596 TM |
4439 | if (ret) |
4440 | goto out; | |
4441 | ||
2b656c1d JB |
4442 | /* |
4443 | * Hey, if we're overwriting t_bucket, what difference does | |
4444 | * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new | |
874d65af JB |
4445 | * cluster to fill, we came here from |
4446 | * ocfs2_mv_xattr_buckets(), and it is really new - | |
4447 | * ACCESS_CREATE is required. But we also might have moved data | |
4448 | * out of t_bucket before extending back into it. | |
4449 | * ocfs2_add_new_xattr_bucket() can do this - its call to | |
4450 | * ocfs2_add_new_xattr_cluster() may have created a new extent | |
2b656c1d JB |
4451 | * and copied out the end of the old extent. Then it re-extends |
4452 | * the old extent back to create space for new xattrs. That's | |
4453 | * how we get here, and the bucket isn't really new. | |
4454 | */ | |
ba937127 | 4455 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4456 | t_is_new ? |
4457 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4458 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4459 | if (ret) | |
4460 | goto out; | |
01225596 | 4461 | |
ba937127 JB |
4462 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
4463 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); | |
01225596 TM |
4464 | |
4465 | out: | |
ba937127 JB |
4466 | ocfs2_xattr_bucket_free(t_bucket); |
4467 | ocfs2_xattr_bucket_free(s_bucket); | |
01225596 TM |
4468 | |
4469 | return ret; | |
4470 | } | |
4471 | ||
4472 | /* | |
874d65af JB |
4473 | * src_blk points to the start of an existing extent. last_blk points to |
4474 | * last cluster in that extent. to_blk points to a newly allocated | |
54ecb6b6 JB |
4475 | * extent. We copy the buckets from the cluster at last_blk to the new |
4476 | * extent. If start_bucket is non-zero, we skip that many buckets before | |
4477 | * we start copying. The new extent's xh_num_buckets gets set to the | |
4478 | * number of buckets we copied. The old extent's xh_num_buckets shrinks | |
4479 | * by the same amount. | |
01225596 | 4480 | */ |
54ecb6b6 JB |
4481 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
4482 | u64 src_blk, u64 last_blk, u64 to_blk, | |
4483 | unsigned int start_bucket, | |
4484 | u32 *first_hash) | |
01225596 TM |
4485 | { |
4486 | int i, ret, credits; | |
4487 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
15d60929 | 4488 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 | 4489 | int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); |
15d60929 | 4490 | struct ocfs2_xattr_bucket *old_first, *new_first; |
01225596 | 4491 | |
874d65af JB |
4492 | mlog(0, "mv xattrs from cluster %llu to %llu\n", |
4493 | (unsigned long long)last_blk, (unsigned long long)to_blk); | |
01225596 | 4494 | |
54ecb6b6 JB |
4495 | BUG_ON(start_bucket >= num_buckets); |
4496 | if (start_bucket) { | |
4497 | num_buckets -= start_bucket; | |
4498 | last_blk += (start_bucket * blks_per_bucket); | |
4499 | } | |
4500 | ||
15d60929 JB |
4501 | /* The first bucket of the original extent */ |
4502 | old_first = ocfs2_xattr_bucket_new(inode); | |
4503 | /* The first bucket of the new extent */ | |
4504 | new_first = ocfs2_xattr_bucket_new(inode); | |
4505 | if (!old_first || !new_first) { | |
4506 | ret = -ENOMEM; | |
4507 | mlog_errno(ret); | |
4508 | goto out; | |
4509 | } | |
4510 | ||
874d65af | 4511 | ret = ocfs2_read_xattr_bucket(old_first, src_blk); |
15d60929 JB |
4512 | if (ret) { |
4513 | mlog_errno(ret); | |
4514 | goto out; | |
4515 | } | |
4516 | ||
01225596 | 4517 | /* |
54ecb6b6 JB |
4518 | * We need to update the first bucket of the old extent and all |
4519 | * the buckets going to the new extent. | |
01225596 | 4520 | */ |
54ecb6b6 JB |
4521 | credits = ((num_buckets + 1) * blks_per_bucket) + |
4522 | handle->h_buffer_credits; | |
01225596 TM |
4523 | ret = ocfs2_extend_trans(handle, credits); |
4524 | if (ret) { | |
4525 | mlog_errno(ret); | |
4526 | goto out; | |
4527 | } | |
4528 | ||
15d60929 JB |
4529 | ret = ocfs2_xattr_bucket_journal_access(handle, old_first, |
4530 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4531 | if (ret) { |
4532 | mlog_errno(ret); | |
4533 | goto out; | |
4534 | } | |
4535 | ||
4536 | for (i = 0; i < num_buckets; i++) { | |
4537 | ret = ocfs2_cp_xattr_bucket(inode, handle, | |
874d65af | 4538 | last_blk + (i * blks_per_bucket), |
15d60929 JB |
4539 | to_blk + (i * blks_per_bucket), |
4540 | 1); | |
01225596 TM |
4541 | if (ret) { |
4542 | mlog_errno(ret); | |
4543 | goto out; | |
4544 | } | |
01225596 TM |
4545 | } |
4546 | ||
15d60929 JB |
4547 | /* |
4548 | * Get the new bucket ready before we dirty anything | |
4549 | * (This actually shouldn't fail, because we already dirtied | |
4550 | * it once in ocfs2_cp_xattr_bucket()). | |
4551 | */ | |
4552 | ret = ocfs2_read_xattr_bucket(new_first, to_blk); | |
4553 | if (ret) { | |
01225596 TM |
4554 | mlog_errno(ret); |
4555 | goto out; | |
4556 | } | |
15d60929 JB |
4557 | ret = ocfs2_xattr_bucket_journal_access(handle, new_first, |
4558 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4559 | if (ret) { |
4560 | mlog_errno(ret); | |
4561 | goto out; | |
4562 | } | |
4563 | ||
15d60929 JB |
4564 | /* Now update the headers */ |
4565 | le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets); | |
4566 | ocfs2_xattr_bucket_journal_dirty(handle, old_first); | |
01225596 | 4567 | |
15d60929 JB |
4568 | bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(num_buckets); |
4569 | ocfs2_xattr_bucket_journal_dirty(handle, new_first); | |
01225596 TM |
4570 | |
4571 | if (first_hash) | |
15d60929 JB |
4572 | *first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash); |
4573 | ||
01225596 | 4574 | out: |
15d60929 JB |
4575 | ocfs2_xattr_bucket_free(new_first); |
4576 | ocfs2_xattr_bucket_free(old_first); | |
01225596 TM |
4577 | return ret; |
4578 | } | |
4579 | ||
4580 | /* | |
80bcaf34 | 4581 | * Move some xattrs in this cluster to the new cluster. |
01225596 TM |
4582 | * This function should only be called when bucket size == cluster size. |
4583 | * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead. | |
4584 | */ | |
80bcaf34 TM |
4585 | static int ocfs2_divide_xattr_cluster(struct inode *inode, |
4586 | handle_t *handle, | |
4587 | u64 prev_blk, | |
4588 | u64 new_blk, | |
4589 | u32 *first_hash) | |
01225596 TM |
4590 | { |
4591 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
85db90e7 | 4592 | int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits; |
01225596 TM |
4593 | |
4594 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize); | |
4595 | ||
4596 | ret = ocfs2_extend_trans(handle, credits); | |
4597 | if (ret) { | |
4598 | mlog_errno(ret); | |
4599 | return ret; | |
4600 | } | |
4601 | ||
4602 | /* Move half of the xattr in start_blk to the next bucket. */ | |
80bcaf34 TM |
4603 | return ocfs2_divide_xattr_bucket(inode, handle, prev_blk, |
4604 | new_blk, first_hash, 1); | |
01225596 TM |
4605 | } |
4606 | ||
4607 | /* | |
4608 | * Move some xattrs from the old cluster to the new one since they are not | |
4609 | * contiguous in ocfs2 xattr tree. | |
4610 | * | |
4611 | * new_blk starts a new separate cluster, and we will move some xattrs from | |
4612 | * prev_blk to it. v_start will be set as the first name hash value in this | |
4613 | * new cluster so that it can be used as e_cpos during tree insertion and | |
4614 | * don't collide with our original b-tree operations. first_bh and header_bh | |
4615 | * will also be updated since they will be used in ocfs2_extend_xattr_bucket | |
4616 | * to extend the insert bucket. | |
4617 | * | |
4618 | * The problem is how much xattr should we move to the new one and when should | |
4619 | * we update first_bh and header_bh? | |
4620 | * 1. If cluster size > bucket size, that means the previous cluster has more | |
4621 | * than 1 bucket, so just move half nums of bucket into the new cluster and | |
4622 | * update the first_bh and header_bh if the insert bucket has been moved | |
4623 | * to the new cluster. | |
4624 | * 2. If cluster_size == bucket_size: | |
4625 | * a) If the previous extent rec has more than one cluster and the insert | |
4626 | * place isn't in the last cluster, copy the entire last cluster to the | |
4627 | * new one. This time, we don't need to upate the first_bh and header_bh | |
4628 | * since they will not be moved into the new cluster. | |
4629 | * b) Otherwise, move the bottom half of the xattrs in the last cluster into | |
4630 | * the new one. And we set the extend flag to zero if the insert place is | |
4631 | * moved into the new allocated cluster since no extend is needed. | |
4632 | */ | |
4633 | static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode, | |
4634 | handle_t *handle, | |
012ee910 JB |
4635 | struct ocfs2_xattr_bucket *first, |
4636 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4637 | u64 new_blk, |
01225596 TM |
4638 | u32 prev_clusters, |
4639 | u32 *v_start, | |
4640 | int *extend) | |
4641 | { | |
92cf3adf | 4642 | int ret; |
01225596 TM |
4643 | |
4644 | mlog(0, "adjust xattrs from cluster %llu len %u to %llu\n", | |
012ee910 | 4645 | (unsigned long long)bucket_blkno(first), prev_clusters, |
de29c085 | 4646 | (unsigned long long)new_blk); |
01225596 | 4647 | |
41cb8148 | 4648 | if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) { |
01225596 TM |
4649 | ret = ocfs2_mv_xattr_bucket_cross_cluster(inode, |
4650 | handle, | |
41cb8148 | 4651 | first, target, |
01225596 | 4652 | new_blk, |
01225596 TM |
4653 | prev_clusters, |
4654 | v_start); | |
012ee910 | 4655 | if (ret) |
41cb8148 | 4656 | mlog_errno(ret); |
41cb8148 | 4657 | } else { |
92cf3adf JB |
4658 | /* The start of the last cluster in the first extent */ |
4659 | u64 last_blk = bucket_blkno(first) + | |
4660 | ((prev_clusters - 1) * | |
4661 | ocfs2_clusters_to_blocks(inode->i_sb, 1)); | |
01225596 | 4662 | |
012ee910 | 4663 | if (prev_clusters > 1 && bucket_blkno(target) != last_blk) { |
874d65af | 4664 | ret = ocfs2_mv_xattr_buckets(inode, handle, |
92cf3adf | 4665 | bucket_blkno(first), |
54ecb6b6 | 4666 | last_blk, new_blk, 0, |
01225596 | 4667 | v_start); |
012ee910 JB |
4668 | if (ret) |
4669 | mlog_errno(ret); | |
4670 | } else { | |
80bcaf34 TM |
4671 | ret = ocfs2_divide_xattr_cluster(inode, handle, |
4672 | last_blk, new_blk, | |
4673 | v_start); | |
012ee910 JB |
4674 | if (ret) |
4675 | mlog_errno(ret); | |
01225596 | 4676 | |
92cf3adf | 4677 | if ((bucket_blkno(target) == last_blk) && extend) |
01225596 TM |
4678 | *extend = 0; |
4679 | } | |
4680 | } | |
4681 | ||
4682 | return ret; | |
4683 | } | |
4684 | ||
4685 | /* | |
4686 | * Add a new cluster for xattr storage. | |
4687 | * | |
4688 | * If the new cluster is contiguous with the previous one, it will be | |
4689 | * appended to the same extent record, and num_clusters will be updated. | |
4690 | * If not, we will insert a new extent for it and move some xattrs in | |
4691 | * the last cluster into the new allocated one. | |
4692 | * We also need to limit the maximum size of a btree leaf, otherwise we'll | |
4693 | * lose the benefits of hashing because we'll have to search large leaves. | |
4694 | * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize, | |
4695 | * if it's bigger). | |
4696 | * | |
4697 | * first_bh is the first block of the previous extent rec and header_bh | |
4698 | * indicates the bucket we will insert the new xattrs. They will be updated | |
4699 | * when the header_bh is moved into the new cluster. | |
4700 | */ | |
4701 | static int ocfs2_add_new_xattr_cluster(struct inode *inode, | |
4702 | struct buffer_head *root_bh, | |
ed29c0ca JB |
4703 | struct ocfs2_xattr_bucket *first, |
4704 | struct ocfs2_xattr_bucket *target, | |
01225596 TM |
4705 | u32 *num_clusters, |
4706 | u32 prev_cpos, | |
78f30c31 TM |
4707 | int *extend, |
4708 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 4709 | { |
85db90e7 | 4710 | int ret; |
01225596 TM |
4711 | u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); |
4712 | u32 prev_clusters = *num_clusters; | |
4713 | u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0; | |
4714 | u64 block; | |
85db90e7 | 4715 | handle_t *handle = ctxt->handle; |
01225596 | 4716 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
f99b9b7c | 4717 | struct ocfs2_extent_tree et; |
01225596 TM |
4718 | |
4719 | mlog(0, "Add new xattr cluster for %llu, previous xattr hash = %u, " | |
4720 | "previous xattr blkno = %llu\n", | |
4721 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ed29c0ca | 4722 | prev_cpos, (unsigned long long)bucket_blkno(first)); |
01225596 | 4723 | |
5e404e9e | 4724 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
f99b9b7c | 4725 | |
0cf2f763 | 4726 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 4727 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4728 | if (ret < 0) { |
4729 | mlog_errno(ret); | |
4730 | goto leave; | |
4731 | } | |
4732 | ||
78f30c31 | 4733 | ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1, |
01225596 TM |
4734 | clusters_to_add, &bit_off, &num_bits); |
4735 | if (ret < 0) { | |
4736 | if (ret != -ENOSPC) | |
4737 | mlog_errno(ret); | |
4738 | goto leave; | |
4739 | } | |
4740 | ||
4741 | BUG_ON(num_bits > clusters_to_add); | |
4742 | ||
4743 | block = ocfs2_clusters_to_blocks(osb->sb, bit_off); | |
4744 | mlog(0, "Allocating %u clusters at block %u for xattr in inode %llu\n", | |
4745 | num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
4746 | ||
ed29c0ca | 4747 | if (bucket_blkno(first) + (prev_clusters * bpc) == block && |
01225596 TM |
4748 | (prev_clusters + num_bits) << osb->s_clustersize_bits <= |
4749 | OCFS2_MAX_XATTR_TREE_LEAF_SIZE) { | |
4750 | /* | |
4751 | * If this cluster is contiguous with the old one and | |
4752 | * adding this new cluster, we don't surpass the limit of | |
4753 | * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be | |
4754 | * initialized and used like other buckets in the previous | |
4755 | * cluster. | |
4756 | * So add it as a contiguous one. The caller will handle | |
4757 | * its init process. | |
4758 | */ | |
4759 | v_start = prev_cpos + prev_clusters; | |
4760 | *num_clusters = prev_clusters + num_bits; | |
4761 | mlog(0, "Add contiguous %u clusters to previous extent rec.\n", | |
4762 | num_bits); | |
4763 | } else { | |
4764 | ret = ocfs2_adjust_xattr_cross_cluster(inode, | |
4765 | handle, | |
012ee910 JB |
4766 | first, |
4767 | target, | |
01225596 | 4768 | block, |
01225596 TM |
4769 | prev_clusters, |
4770 | &v_start, | |
4771 | extend); | |
4772 | if (ret) { | |
4773 | mlog_errno(ret); | |
4774 | goto leave; | |
4775 | } | |
4776 | } | |
4777 | ||
4778 | mlog(0, "Insert %u clusters at block %llu for xattr at %u\n", | |
de29c085 | 4779 | num_bits, (unsigned long long)block, v_start); |
cc79d8c1 | 4780 | ret = ocfs2_insert_extent(handle, &et, v_start, block, |
78f30c31 | 4781 | num_bits, 0, ctxt->meta_ac); |
01225596 TM |
4782 | if (ret < 0) { |
4783 | mlog_errno(ret); | |
4784 | goto leave; | |
4785 | } | |
4786 | ||
4787 | ret = ocfs2_journal_dirty(handle, root_bh); | |
85db90e7 | 4788 | if (ret < 0) |
01225596 | 4789 | mlog_errno(ret); |
01225596 TM |
4790 | |
4791 | leave: | |
01225596 TM |
4792 | return ret; |
4793 | } | |
4794 | ||
4795 | /* | |
92de109a JB |
4796 | * We are given an extent. 'first' is the bucket at the very front of |
4797 | * the extent. The extent has space for an additional bucket past | |
4798 | * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number | |
4799 | * of the target bucket. We wish to shift every bucket past the target | |
4800 | * down one, filling in that additional space. When we get back to the | |
4801 | * target, we split the target between itself and the now-empty bucket | |
4802 | * at target+1 (aka, target_blkno + blks_per_bucket). | |
01225596 TM |
4803 | */ |
4804 | static int ocfs2_extend_xattr_bucket(struct inode *inode, | |
85db90e7 | 4805 | handle_t *handle, |
92de109a JB |
4806 | struct ocfs2_xattr_bucket *first, |
4807 | u64 target_blk, | |
01225596 TM |
4808 | u32 num_clusters) |
4809 | { | |
4810 | int ret, credits; | |
4811 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
4812 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
92de109a JB |
4813 | u64 end_blk; |
4814 | u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets); | |
01225596 TM |
4815 | |
4816 | mlog(0, "extend xattr bucket in %llu, xattr extend rec starting " | |
92de109a JB |
4817 | "from %llu, len = %u\n", (unsigned long long)target_blk, |
4818 | (unsigned long long)bucket_blkno(first), num_clusters); | |
01225596 | 4819 | |
92de109a JB |
4820 | /* The extent must have room for an additional bucket */ |
4821 | BUG_ON(new_bucket >= | |
4822 | (num_clusters * ocfs2_xattr_buckets_per_cluster(osb))); | |
01225596 | 4823 | |
92de109a JB |
4824 | /* end_blk points to the last existing bucket */ |
4825 | end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket); | |
01225596 TM |
4826 | |
4827 | /* | |
92de109a JB |
4828 | * end_blk is the start of the last existing bucket. |
4829 | * Thus, (end_blk - target_blk) covers the target bucket and | |
4830 | * every bucket after it up to, but not including, the last | |
4831 | * existing bucket. Then we add the last existing bucket, the | |
4832 | * new bucket, and the first bucket (3 * blk_per_bucket). | |
01225596 | 4833 | */ |
92de109a | 4834 | credits = (end_blk - target_blk) + (3 * blk_per_bucket) + |
85db90e7 TM |
4835 | handle->h_buffer_credits; |
4836 | ret = ocfs2_extend_trans(handle, credits); | |
4837 | if (ret) { | |
01225596 TM |
4838 | mlog_errno(ret); |
4839 | goto out; | |
4840 | } | |
4841 | ||
92de109a JB |
4842 | ret = ocfs2_xattr_bucket_journal_access(handle, first, |
4843 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4844 | if (ret) { |
4845 | mlog_errno(ret); | |
85db90e7 | 4846 | goto out; |
01225596 TM |
4847 | } |
4848 | ||
92de109a | 4849 | while (end_blk != target_blk) { |
01225596 TM |
4850 | ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk, |
4851 | end_blk + blk_per_bucket, 0); | |
4852 | if (ret) | |
85db90e7 | 4853 | goto out; |
01225596 TM |
4854 | end_blk -= blk_per_bucket; |
4855 | } | |
4856 | ||
92de109a JB |
4857 | /* Move half of the xattr in target_blkno to the next bucket. */ |
4858 | ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk, | |
4859 | target_blk + blk_per_bucket, NULL, 0); | |
01225596 | 4860 | |
92de109a JB |
4861 | le16_add_cpu(&bucket_xh(first)->xh_num_buckets, 1); |
4862 | ocfs2_xattr_bucket_journal_dirty(handle, first); | |
01225596 | 4863 | |
01225596 TM |
4864 | out: |
4865 | return ret; | |
4866 | } | |
4867 | ||
4868 | /* | |
91f2033f JB |
4869 | * Add new xattr bucket in an extent record and adjust the buckets |
4870 | * accordingly. xb_bh is the ocfs2_xattr_block, and target is the | |
4871 | * bucket we want to insert into. | |
01225596 | 4872 | * |
91f2033f JB |
4873 | * In the easy case, we will move all the buckets after target down by |
4874 | * one. Half of target's xattrs will be moved to the next bucket. | |
4875 | * | |
4876 | * If current cluster is full, we'll allocate a new one. This may not | |
4877 | * be contiguous. The underlying calls will make sure that there is | |
4878 | * space for the insert, shifting buckets around if necessary. | |
4879 | * 'target' may be moved by those calls. | |
01225596 TM |
4880 | */ |
4881 | static int ocfs2_add_new_xattr_bucket(struct inode *inode, | |
4882 | struct buffer_head *xb_bh, | |
91f2033f | 4883 | struct ocfs2_xattr_bucket *target, |
78f30c31 | 4884 | struct ocfs2_xattr_set_ctxt *ctxt) |
01225596 | 4885 | { |
01225596 TM |
4886 | struct ocfs2_xattr_block *xb = |
4887 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4888 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
4889 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
91f2033f JB |
4890 | u32 name_hash = |
4891 | le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash); | |
ed29c0ca | 4892 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
01225596 TM |
4893 | int ret, num_buckets, extend = 1; |
4894 | u64 p_blkno; | |
4895 | u32 e_cpos, num_clusters; | |
92de109a | 4896 | /* The bucket at the front of the extent */ |
91f2033f | 4897 | struct ocfs2_xattr_bucket *first; |
01225596 | 4898 | |
91f2033f JB |
4899 | mlog(0, "Add new xattr bucket starting from %llu\n", |
4900 | (unsigned long long)bucket_blkno(target)); | |
01225596 | 4901 | |
ed29c0ca | 4902 | /* The first bucket of the original extent */ |
92de109a | 4903 | first = ocfs2_xattr_bucket_new(inode); |
91f2033f | 4904 | if (!first) { |
92de109a JB |
4905 | ret = -ENOMEM; |
4906 | mlog_errno(ret); | |
4907 | goto out; | |
4908 | } | |
4909 | ||
01225596 TM |
4910 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos, |
4911 | &num_clusters, el); | |
4912 | if (ret) { | |
4913 | mlog_errno(ret); | |
4914 | goto out; | |
4915 | } | |
4916 | ||
ed29c0ca | 4917 | ret = ocfs2_read_xattr_bucket(first, p_blkno); |
01225596 TM |
4918 | if (ret) { |
4919 | mlog_errno(ret); | |
4920 | goto out; | |
4921 | } | |
4922 | ||
ed29c0ca JB |
4923 | num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters; |
4924 | if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) { | |
4925 | /* | |
4926 | * This can move first+target if the target bucket moves | |
4927 | * to the new extent. | |
4928 | */ | |
01225596 TM |
4929 | ret = ocfs2_add_new_xattr_cluster(inode, |
4930 | xb_bh, | |
ed29c0ca JB |
4931 | first, |
4932 | target, | |
01225596 TM |
4933 | &num_clusters, |
4934 | e_cpos, | |
78f30c31 TM |
4935 | &extend, |
4936 | ctxt); | |
01225596 TM |
4937 | if (ret) { |
4938 | mlog_errno(ret); | |
4939 | goto out; | |
4940 | } | |
4941 | } | |
4942 | ||
92de109a | 4943 | if (extend) { |
01225596 | 4944 | ret = ocfs2_extend_xattr_bucket(inode, |
85db90e7 | 4945 | ctxt->handle, |
ed29c0ca JB |
4946 | first, |
4947 | bucket_blkno(target), | |
01225596 | 4948 | num_clusters); |
92de109a JB |
4949 | if (ret) |
4950 | mlog_errno(ret); | |
4951 | } | |
4952 | ||
01225596 | 4953 | out: |
92de109a | 4954 | ocfs2_xattr_bucket_free(first); |
ed29c0ca | 4955 | |
01225596 TM |
4956 | return ret; |
4957 | } | |
4958 | ||
4959 | static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode, | |
4960 | struct ocfs2_xattr_bucket *bucket, | |
4961 | int offs) | |
4962 | { | |
4963 | int block_off = offs >> inode->i_sb->s_blocksize_bits; | |
4964 | ||
4965 | offs = offs % inode->i_sb->s_blocksize; | |
51def39f | 4966 | return bucket_block(bucket, block_off) + offs; |
01225596 TM |
4967 | } |
4968 | ||
4969 | /* | |
4970 | * Handle the normal xattr set, including replace, delete and new. | |
01225596 TM |
4971 | * |
4972 | * Note: "local" indicates the real data's locality. So we can't | |
4973 | * just its bucket locality by its length. | |
4974 | */ | |
4975 | static void ocfs2_xattr_set_entry_normal(struct inode *inode, | |
4976 | struct ocfs2_xattr_info *xi, | |
4977 | struct ocfs2_xattr_search *xs, | |
4978 | u32 name_hash, | |
5a095611 | 4979 | int local) |
01225596 TM |
4980 | { |
4981 | struct ocfs2_xattr_entry *last, *xe; | |
4982 | int name_len = strlen(xi->name); | |
4983 | struct ocfs2_xattr_header *xh = xs->header; | |
4984 | u16 count = le16_to_cpu(xh->xh_count), start; | |
4985 | size_t blocksize = inode->i_sb->s_blocksize; | |
4986 | char *val; | |
4987 | size_t offs, size, new_size; | |
11179f2c | 4988 | struct ocfs2_xa_loc loc; |
01225596 | 4989 | |
11179f2c JB |
4990 | ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket, |
4991 | xs->not_found ? NULL : xs->here); | |
01225596 TM |
4992 | last = &xh->xh_entries[count]; |
4993 | if (!xs->not_found) { | |
4994 | xe = xs->here; | |
4995 | offs = le16_to_cpu(xe->xe_name_offset); | |
4996 | if (ocfs2_xattr_is_local(xe)) | |
4997 | size = OCFS2_XATTR_SIZE(name_len) + | |
4998 | OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); | |
4999 | else | |
5000 | size = OCFS2_XATTR_SIZE(name_len) + | |
5001 | OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE); | |
5002 | ||
5003 | /* | |
5004 | * If the new value will be stored outside, xi->value has been | |
5005 | * initalized as an empty ocfs2_xattr_value_root, and the same | |
5006 | * goes with xi->value_len, so we can set new_size safely here. | |
5007 | * See ocfs2_xattr_set_in_bucket. | |
5008 | */ | |
5009 | new_size = OCFS2_XATTR_SIZE(name_len) + | |
5010 | OCFS2_XATTR_SIZE(xi->value_len); | |
5011 | ||
01225596 | 5012 | if (xi->value) { |
bde1e540 | 5013 | ocfs2_xa_wipe_namevalue(&loc); |
01225596 TM |
5014 | if (new_size > size) |
5015 | goto set_new_name_value; | |
5016 | ||
5017 | /* Now replace the old value with new one. */ | |
5018 | if (local) | |
5019 | xe->xe_value_size = cpu_to_le64(xi->value_len); | |
5020 | else | |
5021 | xe->xe_value_size = 0; | |
5022 | ||
5023 | val = ocfs2_xattr_bucket_get_val(inode, | |
ba937127 | 5024 | xs->bucket, offs); |
01225596 TM |
5025 | memset(val + OCFS2_XATTR_SIZE(name_len), 0, |
5026 | size - OCFS2_XATTR_SIZE(name_len)); | |
5027 | if (OCFS2_XATTR_SIZE(xi->value_len) > 0) | |
5028 | memcpy(val + OCFS2_XATTR_SIZE(name_len), | |
5029 | xi->value, xi->value_len); | |
5030 | ||
5031 | le16_add_cpu(&xh->xh_name_value_len, new_size); | |
5032 | ocfs2_xattr_set_local(xe, local); | |
5033 | return; | |
5034 | } else { | |
bde1e540 JB |
5035 | ocfs2_xa_remove_entry(&loc); |
5036 | if (!xh->xh_count) | |
5a095611 TM |
5037 | xh->xh_free_start = |
5038 | cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); | |
5039 | ||
01225596 TM |
5040 | return; |
5041 | } | |
5042 | } else { | |
5043 | /* find a new entry for insert. */ | |
5044 | int low = 0, high = count - 1, tmp; | |
5045 | struct ocfs2_xattr_entry *tmp_xe; | |
5046 | ||
5a095611 | 5047 | while (low <= high && count) { |
01225596 TM |
5048 | tmp = (low + high) / 2; |
5049 | tmp_xe = &xh->xh_entries[tmp]; | |
5050 | ||
5051 | if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) | |
5052 | low = tmp + 1; | |
5053 | else if (name_hash < | |
5054 | le32_to_cpu(tmp_xe->xe_name_hash)) | |
5055 | high = tmp - 1; | |
06b240d8 TM |
5056 | else { |
5057 | low = tmp; | |
01225596 | 5058 | break; |
06b240d8 | 5059 | } |
01225596 TM |
5060 | } |
5061 | ||
5062 | xe = &xh->xh_entries[low]; | |
5063 | if (low != count) | |
5064 | memmove(xe + 1, xe, (void *)last - (void *)xe); | |
5065 | ||
5066 | le16_add_cpu(&xh->xh_count, 1); | |
5067 | memset(xe, 0, sizeof(struct ocfs2_xattr_entry)); | |
5068 | xe->xe_name_hash = cpu_to_le32(name_hash); | |
5069 | xe->xe_name_len = name_len; | |
5070 | ocfs2_xattr_set_type(xe, xi->name_index); | |
5071 | } | |
5072 | ||
5073 | set_new_name_value: | |
5074 | /* Insert the new name+value. */ | |
5075 | size = OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(xi->value_len); | |
5076 | ||
5077 | /* | |
5078 | * We must make sure that the name/value pair | |
5079 | * exists in the same block. | |
5080 | */ | |
5081 | offs = le16_to_cpu(xh->xh_free_start); | |
5082 | start = offs - size; | |
5083 | ||
5084 | if (start >> inode->i_sb->s_blocksize_bits != | |
5085 | (offs - 1) >> inode->i_sb->s_blocksize_bits) { | |
5086 | offs = offs - offs % blocksize; | |
5087 | xh->xh_free_start = cpu_to_le16(offs); | |
5088 | } | |
5089 | ||
ba937127 | 5090 | val = ocfs2_xattr_bucket_get_val(inode, xs->bucket, offs - size); |
01225596 TM |
5091 | xe->xe_name_offset = cpu_to_le16(offs - size); |
5092 | ||
5093 | memset(val, 0, size); | |
5094 | memcpy(val, xi->name, name_len); | |
5095 | memcpy(val + OCFS2_XATTR_SIZE(name_len), xi->value, xi->value_len); | |
5096 | ||
5097 | xe->xe_value_size = cpu_to_le64(xi->value_len); | |
5098 | ocfs2_xattr_set_local(xe, local); | |
5099 | xs->here = xe; | |
5100 | le16_add_cpu(&xh->xh_free_start, -size); | |
5101 | le16_add_cpu(&xh->xh_name_value_len, size); | |
5102 | ||
5103 | return; | |
5104 | } | |
5105 | ||
01225596 TM |
5106 | /* |
5107 | * Set the xattr entry in the specified bucket. | |
5108 | * The bucket is indicated by xs->bucket and it should have the enough | |
5109 | * space for the xattr insertion. | |
5110 | */ | |
5111 | static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode, | |
85db90e7 | 5112 | handle_t *handle, |
01225596 TM |
5113 | struct ocfs2_xattr_info *xi, |
5114 | struct ocfs2_xattr_search *xs, | |
5115 | u32 name_hash, | |
5a095611 | 5116 | int local) |
01225596 | 5117 | { |
1224be02 | 5118 | int ret; |
02dbf38d | 5119 | u64 blkno; |
01225596 | 5120 | |
ff1ec20e MF |
5121 | mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n", |
5122 | (unsigned long)xi->value_len, xi->name_index, | |
ba937127 | 5123 | (unsigned long long)bucket_blkno(xs->bucket)); |
01225596 | 5124 | |
ba937127 | 5125 | if (!xs->bucket->bu_bhs[1]) { |
02dbf38d JB |
5126 | blkno = bucket_blkno(xs->bucket); |
5127 | ocfs2_xattr_bucket_relse(xs->bucket); | |
5128 | ret = ocfs2_read_xattr_bucket(xs->bucket, blkno); | |
01225596 TM |
5129 | if (ret) { |
5130 | mlog_errno(ret); | |
5131 | goto out; | |
5132 | } | |
5133 | } | |
5134 | ||
ba937127 | 5135 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
1224be02 JB |
5136 | OCFS2_JOURNAL_ACCESS_WRITE); |
5137 | if (ret < 0) { | |
5138 | mlog_errno(ret); | |
5139 | goto out; | |
01225596 TM |
5140 | } |
5141 | ||
5a095611 | 5142 | ocfs2_xattr_set_entry_normal(inode, xi, xs, name_hash, local); |
ba937127 | 5143 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); |
01225596 | 5144 | |
01225596 | 5145 | out: |
01225596 TM |
5146 | return ret; |
5147 | } | |
5148 | ||
01225596 TM |
5149 | /* |
5150 | * Truncate the specified xe_off entry in xattr bucket. | |
5151 | * bucket is indicated by header_bh and len is the new length. | |
5152 | * Both the ocfs2_xattr_value_root and the entry will be updated here. | |
5153 | * | |
5154 | * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed. | |
5155 | */ | |
5156 | static int ocfs2_xattr_bucket_value_truncate(struct inode *inode, | |
548b0f22 | 5157 | struct ocfs2_xattr_bucket *bucket, |
01225596 | 5158 | int xe_off, |
78f30c31 TM |
5159 | int len, |
5160 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5161 | { |
5162 | int ret, offset; | |
5163 | u64 value_blk; | |
01225596 | 5164 | struct ocfs2_xattr_entry *xe; |
548b0f22 | 5165 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 5166 | size_t blocksize = inode->i_sb->s_blocksize; |
b3e5d379 JB |
5167 | struct ocfs2_xattr_value_buf vb = { |
5168 | .vb_access = ocfs2_journal_access, | |
5169 | }; | |
01225596 TM |
5170 | |
5171 | xe = &xh->xh_entries[xe_off]; | |
5172 | ||
5173 | BUG_ON(!xe || ocfs2_xattr_is_local(xe)); | |
5174 | ||
5175 | offset = le16_to_cpu(xe->xe_name_offset) + | |
5176 | OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5177 | ||
5178 | value_blk = offset / blocksize; | |
5179 | ||
5180 | /* We don't allow ocfs2_xattr_value to be stored in different block. */ | |
5181 | BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize); | |
01225596 | 5182 | |
b3e5d379 JB |
5183 | vb.vb_bh = bucket->bu_bhs[value_blk]; |
5184 | BUG_ON(!vb.vb_bh); | |
01225596 | 5185 | |
b3e5d379 JB |
5186 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
5187 | (vb.vb_bh->b_data + offset % blocksize); | |
01225596 | 5188 | |
548b0f22 JB |
5189 | /* |
5190 | * From here on out we have to dirty the bucket. The generic | |
5191 | * value calls only modify one of the bucket's bhs, but we need | |
5192 | * to send the bucket at once. So if they error, they *could* have | |
5193 | * modified something. We have to assume they did, and dirty | |
5194 | * the whole bucket. This leaves us in a consistent state. | |
5195 | */ | |
5196 | mlog(0, "truncate %u in xattr bucket %llu to %d bytes.\n", | |
5197 | xe_off, (unsigned long long)bucket_blkno(bucket), len); | |
b3e5d379 | 5198 | ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); |
01225596 TM |
5199 | if (ret) { |
5200 | mlog_errno(ret); | |
554e7f9e TM |
5201 | goto out; |
5202 | } | |
5203 | ||
5204 | ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket, | |
5205 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5206 | if (ret) { | |
5207 | mlog_errno(ret); | |
5208 | goto out; | |
01225596 TM |
5209 | } |
5210 | ||
548b0f22 JB |
5211 | xe->xe_value_size = cpu_to_le64(len); |
5212 | ||
548b0f22 JB |
5213 | ocfs2_xattr_bucket_journal_dirty(ctxt->handle, bucket); |
5214 | ||
01225596 | 5215 | out: |
01225596 TM |
5216 | return ret; |
5217 | } | |
5218 | ||
5219 | static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode, | |
78f30c31 TM |
5220 | struct ocfs2_xattr_search *xs, |
5221 | int len, | |
5222 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5223 | { |
5224 | int ret, offset; | |
5225 | struct ocfs2_xattr_entry *xe = xs->here; | |
5226 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base; | |
5227 | ||
ba937127 | 5228 | BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe)); |
01225596 TM |
5229 | |
5230 | offset = xe - xh->xh_entries; | |
548b0f22 | 5231 | ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket, |
78f30c31 | 5232 | offset, len, ctxt); |
01225596 TM |
5233 | if (ret) |
5234 | mlog_errno(ret); | |
5235 | ||
5236 | return ret; | |
5237 | } | |
5238 | ||
5239 | static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode, | |
85db90e7 | 5240 | handle_t *handle, |
01225596 TM |
5241 | struct ocfs2_xattr_search *xs, |
5242 | char *val, | |
5243 | int value_len) | |
5244 | { | |
712e53e4 | 5245 | int ret, offset, block_off; |
01225596 TM |
5246 | struct ocfs2_xattr_value_root *xv; |
5247 | struct ocfs2_xattr_entry *xe = xs->here; | |
712e53e4 TM |
5248 | struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); |
5249 | void *base; | |
492a8a33 TM |
5250 | struct ocfs2_xattr_value_buf vb = { |
5251 | .vb_access = ocfs2_journal_access, | |
5252 | }; | |
01225596 TM |
5253 | |
5254 | BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe)); | |
5255 | ||
fd68a894 | 5256 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, xh, |
712e53e4 TM |
5257 | xe - xh->xh_entries, |
5258 | &block_off, | |
5259 | &offset); | |
5260 | if (ret) { | |
5261 | mlog_errno(ret); | |
5262 | goto out; | |
5263 | } | |
01225596 | 5264 | |
712e53e4 TM |
5265 | base = bucket_block(xs->bucket, block_off); |
5266 | xv = (struct ocfs2_xattr_value_root *)(base + offset + | |
5267 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
01225596 | 5268 | |
492a8a33 TM |
5269 | vb.vb_xv = xv; |
5270 | vb.vb_bh = xs->bucket->bu_bhs[block_off]; | |
712e53e4 | 5271 | ret = __ocfs2_xattr_set_value_outside(inode, handle, |
492a8a33 | 5272 | &vb, val, value_len); |
712e53e4 TM |
5273 | if (ret) |
5274 | mlog_errno(ret); | |
5275 | out: | |
5276 | return ret; | |
01225596 TM |
5277 | } |
5278 | ||
01225596 TM |
5279 | static int ocfs2_rm_xattr_cluster(struct inode *inode, |
5280 | struct buffer_head *root_bh, | |
5281 | u64 blkno, | |
5282 | u32 cpos, | |
47bca495 TM |
5283 | u32 len, |
5284 | void *para) | |
01225596 TM |
5285 | { |
5286 | int ret; | |
5287 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5288 | struct inode *tl_inode = osb->osb_tl_inode; | |
5289 | handle_t *handle; | |
5290 | struct ocfs2_xattr_block *xb = | |
5291 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
01225596 TM |
5292 | struct ocfs2_alloc_context *meta_ac = NULL; |
5293 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
f99b9b7c JB |
5294 | struct ocfs2_extent_tree et; |
5295 | ||
47bca495 | 5296 | ret = ocfs2_iterate_xattr_buckets(inode, blkno, len, |
ce9c5a54 | 5297 | ocfs2_delete_xattr_in_bucket, para); |
47bca495 TM |
5298 | if (ret) { |
5299 | mlog_errno(ret); | |
5300 | return ret; | |
5301 | } | |
5302 | ||
5e404e9e | 5303 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
01225596 TM |
5304 | |
5305 | ocfs2_init_dealloc_ctxt(&dealloc); | |
5306 | ||
5307 | mlog(0, "rm xattr extent rec at %u len = %u, start from %llu\n", | |
5308 | cpos, len, (unsigned long long)blkno); | |
5309 | ||
8cb471e8 JB |
5310 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), blkno, |
5311 | len); | |
01225596 | 5312 | |
f99b9b7c | 5313 | ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac); |
01225596 TM |
5314 | if (ret) { |
5315 | mlog_errno(ret); | |
5316 | return ret; | |
5317 | } | |
5318 | ||
5319 | mutex_lock(&tl_inode->i_mutex); | |
5320 | ||
5321 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
5322 | ret = __ocfs2_flush_truncate_log(osb); | |
5323 | if (ret < 0) { | |
5324 | mlog_errno(ret); | |
5325 | goto out; | |
5326 | } | |
5327 | } | |
5328 | ||
a90714c1 | 5329 | handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb)); |
d3264799 | 5330 | if (IS_ERR(handle)) { |
01225596 TM |
5331 | ret = -ENOMEM; |
5332 | mlog_errno(ret); | |
5333 | goto out; | |
5334 | } | |
5335 | ||
0cf2f763 | 5336 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5337 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5338 | if (ret) { |
5339 | mlog_errno(ret); | |
5340 | goto out_commit; | |
5341 | } | |
5342 | ||
dbdcf6a4 | 5343 | ret = ocfs2_remove_extent(handle, &et, cpos, len, meta_ac, |
f99b9b7c | 5344 | &dealloc); |
01225596 TM |
5345 | if (ret) { |
5346 | mlog_errno(ret); | |
5347 | goto out_commit; | |
5348 | } | |
5349 | ||
5350 | le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len); | |
5351 | ||
5352 | ret = ocfs2_journal_dirty(handle, root_bh); | |
5353 | if (ret) { | |
5354 | mlog_errno(ret); | |
5355 | goto out_commit; | |
5356 | } | |
5357 | ||
5358 | ret = ocfs2_truncate_log_append(osb, handle, blkno, len); | |
5359 | if (ret) | |
5360 | mlog_errno(ret); | |
5361 | ||
5362 | out_commit: | |
5363 | ocfs2_commit_trans(osb, handle); | |
5364 | out: | |
5365 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
5366 | ||
5367 | mutex_unlock(&tl_inode->i_mutex); | |
5368 | ||
5369 | if (meta_ac) | |
5370 | ocfs2_free_alloc_context(meta_ac); | |
5371 | ||
5372 | ocfs2_run_deallocs(osb, &dealloc); | |
5373 | ||
5374 | return ret; | |
5375 | } | |
5376 | ||
01225596 | 5377 | static void ocfs2_xattr_bucket_remove_xs(struct inode *inode, |
85db90e7 | 5378 | handle_t *handle, |
01225596 TM |
5379 | struct ocfs2_xattr_search *xs) |
5380 | { | |
ba937127 | 5381 | struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); |
01225596 TM |
5382 | struct ocfs2_xattr_entry *last = &xh->xh_entries[ |
5383 | le16_to_cpu(xh->xh_count) - 1]; | |
5384 | int ret = 0; | |
5385 | ||
ba937127 | 5386 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
1224be02 | 5387 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5388 | if (ret) { |
5389 | mlog_errno(ret); | |
85db90e7 | 5390 | return; |
01225596 TM |
5391 | } |
5392 | ||
5393 | /* Remove the old entry. */ | |
5394 | memmove(xs->here, xs->here + 1, | |
5395 | (void *)last - (void *)xs->here); | |
5396 | memset(last, 0, sizeof(struct ocfs2_xattr_entry)); | |
5397 | le16_add_cpu(&xh->xh_count, -1); | |
5398 | ||
ba937127 | 5399 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); |
01225596 TM |
5400 | } |
5401 | ||
5402 | /* | |
5403 | * Set the xattr name/value in the bucket specified in xs. | |
5404 | * | |
5405 | * As the new value in xi may be stored in the bucket or in an outside cluster, | |
5406 | * we divide the whole process into 3 steps: | |
5407 | * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket) | |
5408 | * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs) | |
5409 | * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside) | |
5410 | * 4. If the clusters for the new outside value can't be allocated, we need | |
5411 | * to free the xattr we allocated in set. | |
5412 | */ | |
5413 | static int ocfs2_xattr_set_in_bucket(struct inode *inode, | |
5414 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
5415 | struct ocfs2_xattr_search *xs, |
5416 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5417 | { |
5a095611 | 5418 | int ret, local = 1; |
01225596 TM |
5419 | size_t value_len; |
5420 | char *val = (char *)xi->value; | |
5421 | struct ocfs2_xattr_entry *xe = xs->here; | |
2057e5c6 TM |
5422 | u32 name_hash = ocfs2_xattr_name_hash(inode, xi->name, |
5423 | strlen(xi->name)); | |
01225596 TM |
5424 | |
5425 | if (!xs->not_found && !ocfs2_xattr_is_local(xe)) { | |
5426 | /* | |
5427 | * We need to truncate the xattr storage first. | |
5428 | * | |
5429 | * If both the old and new value are stored to | |
5430 | * outside block, we only need to truncate | |
5431 | * the storage and then set the value outside. | |
5432 | * | |
5433 | * If the new value should be stored within block, | |
5434 | * we should free all the outside block first and | |
5435 | * the modification to the xattr block will be done | |
5436 | * by following steps. | |
5437 | */ | |
5438 | if (xi->value_len > OCFS2_XATTR_INLINE_SIZE) | |
5439 | value_len = xi->value_len; | |
5440 | else | |
5441 | value_len = 0; | |
5442 | ||
5443 | ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, | |
78f30c31 TM |
5444 | value_len, |
5445 | ctxt); | |
01225596 TM |
5446 | if (ret) |
5447 | goto out; | |
5448 | ||
5449 | if (value_len) | |
5450 | goto set_value_outside; | |
5451 | } | |
5452 | ||
5453 | value_len = xi->value_len; | |
5454 | /* So we have to handle the inside block change now. */ | |
5455 | if (value_len > OCFS2_XATTR_INLINE_SIZE) { | |
5456 | /* | |
5457 | * If the new value will be stored outside of block, | |
5458 | * initalize a new empty value root and insert it first. | |
5459 | */ | |
5460 | local = 0; | |
5461 | xi->value = &def_xv; | |
5462 | xi->value_len = OCFS2_XATTR_ROOT_SIZE; | |
5463 | } | |
5464 | ||
85db90e7 TM |
5465 | ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs, |
5466 | name_hash, local); | |
01225596 TM |
5467 | if (ret) { |
5468 | mlog_errno(ret); | |
5469 | goto out; | |
5470 | } | |
5471 | ||
5a095611 TM |
5472 | if (value_len <= OCFS2_XATTR_INLINE_SIZE) |
5473 | goto out; | |
01225596 | 5474 | |
5a095611 TM |
5475 | /* allocate the space now for the outside block storage. */ |
5476 | ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, | |
78f30c31 | 5477 | value_len, ctxt); |
5a095611 TM |
5478 | if (ret) { |
5479 | mlog_errno(ret); | |
5480 | ||
5481 | if (xs->not_found) { | |
5482 | /* | |
5483 | * We can't allocate enough clusters for outside | |
5484 | * storage and we have allocated xattr already, | |
5485 | * so need to remove it. | |
5486 | */ | |
85db90e7 | 5487 | ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs); |
01225596 | 5488 | } |
01225596 TM |
5489 | goto out; |
5490 | } | |
5491 | ||
5492 | set_value_outside: | |
85db90e7 TM |
5493 | ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle, |
5494 | xs, val, value_len); | |
01225596 TM |
5495 | out: |
5496 | return ret; | |
5497 | } | |
5498 | ||
80bcaf34 TM |
5499 | /* |
5500 | * check whether the xattr bucket is filled up with the same hash value. | |
5501 | * If we want to insert the xattr with the same hash, return -ENOSPC. | |
5502 | * If we want to insert a xattr with different hash value, go ahead | |
5503 | * and ocfs2_divide_xattr_bucket will handle this. | |
5504 | */ | |
01225596 | 5505 | static int ocfs2_check_xattr_bucket_collision(struct inode *inode, |
80bcaf34 TM |
5506 | struct ocfs2_xattr_bucket *bucket, |
5507 | const char *name) | |
01225596 | 5508 | { |
3e632946 | 5509 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
80bcaf34 TM |
5510 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
5511 | ||
5512 | if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) | |
5513 | return 0; | |
01225596 TM |
5514 | |
5515 | if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == | |
5516 | xh->xh_entries[0].xe_name_hash) { | |
5517 | mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " | |
5518 | "hash = %u\n", | |
9c7759aa | 5519 | (unsigned long long)bucket_blkno(bucket), |
01225596 TM |
5520 | le32_to_cpu(xh->xh_entries[0].xe_name_hash)); |
5521 | return -ENOSPC; | |
5522 | } | |
5523 | ||
5524 | return 0; | |
5525 | } | |
5526 | ||
5527 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, | |
5528 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
5529 | struct ocfs2_xattr_search *xs, |
5530 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5531 | { |
5532 | struct ocfs2_xattr_header *xh; | |
5533 | struct ocfs2_xattr_entry *xe; | |
5534 | u16 count, header_size, xh_free_start; | |
6dde41d9 | 5535 | int free, max_free, need, old; |
01225596 TM |
5536 | size_t value_size = 0, name_len = strlen(xi->name); |
5537 | size_t blocksize = inode->i_sb->s_blocksize; | |
5538 | int ret, allocation = 0; | |
01225596 TM |
5539 | |
5540 | mlog_entry("Set xattr %s in xattr index block\n", xi->name); | |
5541 | ||
5542 | try_again: | |
5543 | xh = xs->header; | |
5544 | count = le16_to_cpu(xh->xh_count); | |
5545 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
5546 | header_size = sizeof(struct ocfs2_xattr_header) + | |
5547 | count * sizeof(struct ocfs2_xattr_entry); | |
4442f518 TY |
5548 | max_free = OCFS2_XATTR_BUCKET_SIZE - header_size - |
5549 | le16_to_cpu(xh->xh_name_value_len) - OCFS2_XATTR_HEADER_GAP; | |
01225596 TM |
5550 | |
5551 | mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size " | |
5552 | "of %u which exceed block size\n", | |
ba937127 | 5553 | (unsigned long long)bucket_blkno(xs->bucket), |
01225596 TM |
5554 | header_size); |
5555 | ||
5556 | if (xi->value && xi->value_len > OCFS2_XATTR_INLINE_SIZE) | |
5557 | value_size = OCFS2_XATTR_ROOT_SIZE; | |
5558 | else if (xi->value) | |
5559 | value_size = OCFS2_XATTR_SIZE(xi->value_len); | |
5560 | ||
5561 | if (xs->not_found) | |
5562 | need = sizeof(struct ocfs2_xattr_entry) + | |
5563 | OCFS2_XATTR_SIZE(name_len) + value_size; | |
5564 | else { | |
5565 | need = value_size + OCFS2_XATTR_SIZE(name_len); | |
5566 | ||
5567 | /* | |
5568 | * We only replace the old value if the new length is smaller | |
5569 | * than the old one. Otherwise we will allocate new space in the | |
5570 | * bucket to store it. | |
5571 | */ | |
5572 | xe = xs->here; | |
5573 | if (ocfs2_xattr_is_local(xe)) | |
5574 | old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); | |
5575 | else | |
5576 | old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE); | |
5577 | ||
5578 | if (old >= value_size) | |
5579 | need = 0; | |
5580 | } | |
5581 | ||
4442f518 | 5582 | free = xh_free_start - header_size - OCFS2_XATTR_HEADER_GAP; |
01225596 TM |
5583 | /* |
5584 | * We need to make sure the new name/value pair | |
5585 | * can exist in the same block. | |
5586 | */ | |
5587 | if (xh_free_start % blocksize < need) | |
5588 | free -= xh_free_start % blocksize; | |
5589 | ||
5590 | mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, " | |
5591 | "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len =" | |
5592 | " %u\n", xs->not_found, | |
ba937127 | 5593 | (unsigned long long)bucket_blkno(xs->bucket), |
01225596 TM |
5594 | free, need, max_free, le16_to_cpu(xh->xh_free_start), |
5595 | le16_to_cpu(xh->xh_name_value_len)); | |
5596 | ||
976331d8 TM |
5597 | if (free < need || |
5598 | (xs->not_found && | |
5599 | count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) { | |
01225596 TM |
5600 | if (need <= max_free && |
5601 | count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) { | |
5602 | /* | |
5603 | * We can create the space by defragment. Since only the | |
5604 | * name/value will be moved, the xe shouldn't be changed | |
5605 | * in xs. | |
5606 | */ | |
85db90e7 TM |
5607 | ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, |
5608 | xs->bucket); | |
01225596 TM |
5609 | if (ret) { |
5610 | mlog_errno(ret); | |
5611 | goto out; | |
5612 | } | |
5613 | ||
5614 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4442f518 TY |
5615 | free = xh_free_start - header_size |
5616 | - OCFS2_XATTR_HEADER_GAP; | |
01225596 TM |
5617 | if (xh_free_start % blocksize < need) |
5618 | free -= xh_free_start % blocksize; | |
5619 | ||
5620 | if (free >= need) | |
5621 | goto xattr_set; | |
5622 | ||
5623 | mlog(0, "Can't get enough space for xattr insert by " | |
5624 | "defragment. Need %u bytes, but we have %d, so " | |
5625 | "allocate new bucket for it.\n", need, free); | |
5626 | } | |
5627 | ||
5628 | /* | |
5629 | * We have to add new buckets or clusters and one | |
5630 | * allocation should leave us enough space for insert. | |
5631 | */ | |
5632 | BUG_ON(allocation); | |
5633 | ||
5634 | /* | |
5635 | * We do not allow for overlapping ranges between buckets. And | |
5636 | * the maximum number of collisions we will allow for then is | |
5637 | * one bucket's worth, so check it here whether we need to | |
5638 | * add a new bucket for the insert. | |
5639 | */ | |
80bcaf34 | 5640 | ret = ocfs2_check_xattr_bucket_collision(inode, |
ba937127 | 5641 | xs->bucket, |
80bcaf34 | 5642 | xi->name); |
01225596 TM |
5643 | if (ret) { |
5644 | mlog_errno(ret); | |
5645 | goto out; | |
5646 | } | |
5647 | ||
5648 | ret = ocfs2_add_new_xattr_bucket(inode, | |
5649 | xs->xattr_bh, | |
91f2033f | 5650 | xs->bucket, |
78f30c31 | 5651 | ctxt); |
01225596 TM |
5652 | if (ret) { |
5653 | mlog_errno(ret); | |
5654 | goto out; | |
5655 | } | |
5656 | ||
91f2033f JB |
5657 | /* |
5658 | * ocfs2_add_new_xattr_bucket() will have updated | |
5659 | * xs->bucket if it moved, but it will not have updated | |
5660 | * any of the other search fields. Thus, we drop it and | |
5661 | * re-search. Everything should be cached, so it'll be | |
5662 | * quick. | |
5663 | */ | |
ba937127 | 5664 | ocfs2_xattr_bucket_relse(xs->bucket); |
01225596 TM |
5665 | ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, |
5666 | xi->name_index, | |
5667 | xi->name, xs); | |
5668 | if (ret && ret != -ENODATA) | |
5669 | goto out; | |
5670 | xs->not_found = ret; | |
5671 | allocation = 1; | |
5672 | goto try_again; | |
5673 | } | |
5674 | ||
5675 | xattr_set: | |
78f30c31 | 5676 | ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt); |
01225596 TM |
5677 | out: |
5678 | mlog_exit(ret); | |
5679 | return ret; | |
5680 | } | |
a3944256 TM |
5681 | |
5682 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
5683 | struct ocfs2_xattr_bucket *bucket, | |
5684 | void *para) | |
5685 | { | |
ce9c5a54 | 5686 | int ret = 0, ref_credits; |
3e632946 | 5687 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
a3944256 TM |
5688 | u16 i; |
5689 | struct ocfs2_xattr_entry *xe; | |
78f30c31 TM |
5690 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
5691 | struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; | |
548b0f22 JB |
5692 | int credits = ocfs2_remove_extent_credits(osb->sb) + |
5693 | ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
ce9c5a54 TM |
5694 | struct ocfs2_xattr_value_root *xv; |
5695 | struct ocfs2_rm_xattr_bucket_para *args = | |
5696 | (struct ocfs2_rm_xattr_bucket_para *)para; | |
78f30c31 TM |
5697 | |
5698 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
a3944256 TM |
5699 | |
5700 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
5701 | xe = &xh->xh_entries[i]; | |
5702 | if (ocfs2_xattr_is_local(xe)) | |
5703 | continue; | |
5704 | ||
ce9c5a54 TM |
5705 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, |
5706 | i, &xv, NULL); | |
5707 | ||
5708 | ret = ocfs2_lock_xattr_remove_allocators(inode, xv, | |
5709 | args->ref_ci, | |
5710 | args->ref_root_bh, | |
5711 | &ctxt.meta_ac, | |
5712 | &ref_credits); | |
5713 | ||
5714 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); | |
88c3b062 TM |
5715 | if (IS_ERR(ctxt.handle)) { |
5716 | ret = PTR_ERR(ctxt.handle); | |
5717 | mlog_errno(ret); | |
5718 | break; | |
5719 | } | |
5720 | ||
548b0f22 | 5721 | ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, |
78f30c31 | 5722 | i, 0, &ctxt); |
88c3b062 TM |
5723 | |
5724 | ocfs2_commit_trans(osb, ctxt.handle); | |
ce9c5a54 TM |
5725 | if (ctxt.meta_ac) { |
5726 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
5727 | ctxt.meta_ac = NULL; | |
5728 | } | |
a3944256 TM |
5729 | if (ret) { |
5730 | mlog_errno(ret); | |
5731 | break; | |
5732 | } | |
5733 | } | |
5734 | ||
ce9c5a54 TM |
5735 | if (ctxt.meta_ac) |
5736 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
5737 | ocfs2_schedule_truncate_log_flush(osb, 1); |
5738 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
a3944256 TM |
5739 | return ret; |
5740 | } | |
5741 | ||
492a8a33 TM |
5742 | /* |
5743 | * Whenever we modify a xattr value root in the bucket(e.g, CoW | |
5744 | * or change the extent record flag), we need to recalculate | |
5745 | * the metaecc for the whole bucket. So it is done here. | |
5746 | * | |
5747 | * Note: | |
5748 | * We have to give the extra credits for the caller. | |
5749 | */ | |
5750 | static int ocfs2_xattr_bucket_post_refcount(struct inode *inode, | |
5751 | handle_t *handle, | |
5752 | void *para) | |
5753 | { | |
5754 | int ret; | |
5755 | struct ocfs2_xattr_bucket *bucket = | |
5756 | (struct ocfs2_xattr_bucket *)para; | |
5757 | ||
5758 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, | |
5759 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5760 | if (ret) { | |
5761 | mlog_errno(ret); | |
5762 | return ret; | |
5763 | } | |
5764 | ||
5765 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
5766 | ||
5767 | return 0; | |
5768 | } | |
5769 | ||
5770 | /* | |
5771 | * Special action we need if the xattr value is refcounted. | |
5772 | * | |
5773 | * 1. If the xattr is refcounted, lock the tree. | |
5774 | * 2. CoW the xattr if we are setting the new value and the value | |
5775 | * will be stored outside. | |
5776 | * 3. In other case, decrease_refcount will work for us, so just | |
5777 | * lock the refcount tree, calculate the meta and credits is OK. | |
5778 | * | |
5779 | * We have to do CoW before ocfs2_init_xattr_set_ctxt since | |
5780 | * currently CoW is a completed transaction, while this function | |
5781 | * will also lock the allocators and let us deadlock. So we will | |
5782 | * CoW the whole xattr value. | |
5783 | */ | |
5784 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, | |
5785 | struct ocfs2_dinode *di, | |
5786 | struct ocfs2_xattr_info *xi, | |
5787 | struct ocfs2_xattr_search *xis, | |
5788 | struct ocfs2_xattr_search *xbs, | |
5789 | struct ocfs2_refcount_tree **ref_tree, | |
5790 | int *meta_add, | |
5791 | int *credits) | |
5792 | { | |
5793 | int ret = 0; | |
5794 | struct ocfs2_xattr_block *xb; | |
5795 | struct ocfs2_xattr_entry *xe; | |
5796 | char *base; | |
5797 | u32 p_cluster, num_clusters; | |
5798 | unsigned int ext_flags; | |
5799 | int name_offset, name_len; | |
5800 | struct ocfs2_xattr_value_buf vb; | |
5801 | struct ocfs2_xattr_bucket *bucket = NULL; | |
5802 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5803 | struct ocfs2_post_refcount refcount; | |
5804 | struct ocfs2_post_refcount *p = NULL; | |
5805 | struct buffer_head *ref_root_bh = NULL; | |
5806 | ||
5807 | if (!xis->not_found) { | |
5808 | xe = xis->here; | |
5809 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5810 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5811 | base = xis->base; | |
5812 | vb.vb_bh = xis->inode_bh; | |
5813 | vb.vb_access = ocfs2_journal_access_di; | |
5814 | } else { | |
5815 | int i, block_off = 0; | |
5816 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
5817 | xe = xbs->here; | |
5818 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5819 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5820 | i = xbs->here - xbs->header->xh_entries; | |
5821 | ||
5822 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 5823 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
492a8a33 TM |
5824 | bucket_xh(xbs->bucket), |
5825 | i, &block_off, | |
5826 | &name_offset); | |
5827 | if (ret) { | |
5828 | mlog_errno(ret); | |
5829 | goto out; | |
5830 | } | |
5831 | base = bucket_block(xbs->bucket, block_off); | |
5832 | vb.vb_bh = xbs->bucket->bu_bhs[block_off]; | |
5833 | vb.vb_access = ocfs2_journal_access; | |
5834 | ||
5835 | if (ocfs2_meta_ecc(osb)) { | |
5836 | /*create parameters for ocfs2_post_refcount. */ | |
5837 | bucket = xbs->bucket; | |
5838 | refcount.credits = bucket->bu_blocks; | |
5839 | refcount.para = bucket; | |
5840 | refcount.func = | |
5841 | ocfs2_xattr_bucket_post_refcount; | |
5842 | p = &refcount; | |
5843 | } | |
5844 | } else { | |
5845 | base = xbs->base; | |
5846 | vb.vb_bh = xbs->xattr_bh; | |
5847 | vb.vb_access = ocfs2_journal_access_xb; | |
5848 | } | |
5849 | } | |
5850 | ||
5851 | if (ocfs2_xattr_is_local(xe)) | |
5852 | goto out; | |
5853 | ||
5854 | vb.vb_xv = (struct ocfs2_xattr_value_root *) | |
5855 | (base + name_offset + name_len); | |
5856 | ||
5857 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
5858 | &num_clusters, &vb.vb_xv->xr_list, | |
5859 | &ext_flags); | |
5860 | if (ret) { | |
5861 | mlog_errno(ret); | |
5862 | goto out; | |
5863 | } | |
5864 | ||
5865 | /* | |
5866 | * We just need to check the 1st extent record, since we always | |
5867 | * CoW the whole xattr. So there shouldn't be a xattr with | |
5868 | * some REFCOUNT extent recs after the 1st one. | |
5869 | */ | |
5870 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5871 | goto out; | |
5872 | ||
5873 | ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), | |
5874 | 1, ref_tree, &ref_root_bh); | |
5875 | if (ret) { | |
5876 | mlog_errno(ret); | |
5877 | goto out; | |
5878 | } | |
5879 | ||
5880 | /* | |
5881 | * If we are deleting the xattr or the new size will be stored inside, | |
5882 | * cool, leave it there, the xattr truncate process will remove them | |
5883 | * for us(it still needs the refcount tree lock and the meta, credits). | |
5884 | * And the worse case is that every cluster truncate will split the | |
5885 | * refcount tree, and make the original extent become 3. So we will need | |
5886 | * 2 * cluster more extent recs at most. | |
5887 | */ | |
5888 | if (!xi->value || xi->value_len <= OCFS2_XATTR_INLINE_SIZE) { | |
5889 | ||
5890 | ret = ocfs2_refcounted_xattr_delete_need(inode, | |
5891 | &(*ref_tree)->rf_ci, | |
5892 | ref_root_bh, vb.vb_xv, | |
5893 | meta_add, credits); | |
5894 | if (ret) | |
5895 | mlog_errno(ret); | |
5896 | goto out; | |
5897 | } | |
5898 | ||
5899 | ret = ocfs2_refcount_cow_xattr(inode, di, &vb, | |
5900 | *ref_tree, ref_root_bh, 0, | |
5901 | le32_to_cpu(vb.vb_xv->xr_clusters), p); | |
5902 | if (ret) | |
5903 | mlog_errno(ret); | |
5904 | ||
5905 | out: | |
5906 | brelse(ref_root_bh); | |
5907 | return ret; | |
5908 | } | |
5909 | ||
0129241e TM |
5910 | /* |
5911 | * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. | |
5912 | * The physical clusters will be added to refcount tree. | |
5913 | */ | |
5914 | static int ocfs2_xattr_value_attach_refcount(struct inode *inode, | |
5915 | struct ocfs2_xattr_value_root *xv, | |
5916 | struct ocfs2_extent_tree *value_et, | |
5917 | struct ocfs2_caching_info *ref_ci, | |
5918 | struct buffer_head *ref_root_bh, | |
5919 | struct ocfs2_cached_dealloc_ctxt *dealloc, | |
5920 | struct ocfs2_post_refcount *refcount) | |
5921 | { | |
5922 | int ret = 0; | |
5923 | u32 clusters = le32_to_cpu(xv->xr_clusters); | |
5924 | u32 cpos, p_cluster, num_clusters; | |
5925 | struct ocfs2_extent_list *el = &xv->xr_list; | |
5926 | unsigned int ext_flags; | |
5927 | ||
5928 | cpos = 0; | |
5929 | while (cpos < clusters) { | |
5930 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
5931 | &num_clusters, el, &ext_flags); | |
5932 | ||
5933 | cpos += num_clusters; | |
5934 | if ((ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5935 | continue; | |
5936 | ||
5937 | BUG_ON(!p_cluster); | |
5938 | ||
5939 | ret = ocfs2_add_refcount_flag(inode, value_et, | |
5940 | ref_ci, ref_root_bh, | |
5941 | cpos - num_clusters, | |
5942 | p_cluster, num_clusters, | |
5943 | dealloc, refcount); | |
5944 | if (ret) { | |
5945 | mlog_errno(ret); | |
5946 | break; | |
5947 | } | |
5948 | } | |
5949 | ||
5950 | return ret; | |
5951 | } | |
5952 | ||
5953 | /* | |
5954 | * Given a normal ocfs2_xattr_header, refcount all the entries which | |
5955 | * have value stored outside. | |
5956 | * Used for xattrs stored in inode and ocfs2_xattr_block. | |
5957 | */ | |
5958 | static int ocfs2_xattr_attach_refcount_normal(struct inode *inode, | |
5959 | struct ocfs2_xattr_value_buf *vb, | |
5960 | struct ocfs2_xattr_header *header, | |
5961 | struct ocfs2_caching_info *ref_ci, | |
5962 | struct buffer_head *ref_root_bh, | |
5963 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
5964 | { | |
5965 | ||
5966 | struct ocfs2_xattr_entry *xe; | |
5967 | struct ocfs2_xattr_value_root *xv; | |
5968 | struct ocfs2_extent_tree et; | |
5969 | int i, ret = 0; | |
5970 | ||
5971 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
5972 | xe = &header->xh_entries[i]; | |
5973 | ||
5974 | if (ocfs2_xattr_is_local(xe)) | |
5975 | continue; | |
5976 | ||
5977 | xv = (struct ocfs2_xattr_value_root *)((void *)header + | |
5978 | le16_to_cpu(xe->xe_name_offset) + | |
5979 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
5980 | ||
5981 | vb->vb_xv = xv; | |
5982 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); | |
5983 | ||
5984 | ret = ocfs2_xattr_value_attach_refcount(inode, xv, &et, | |
5985 | ref_ci, ref_root_bh, | |
5986 | dealloc, NULL); | |
5987 | if (ret) { | |
5988 | mlog_errno(ret); | |
5989 | break; | |
5990 | } | |
5991 | } | |
5992 | ||
5993 | return ret; | |
5994 | } | |
5995 | ||
5996 | static int ocfs2_xattr_inline_attach_refcount(struct inode *inode, | |
5997 | struct buffer_head *fe_bh, | |
5998 | struct ocfs2_caching_info *ref_ci, | |
5999 | struct buffer_head *ref_root_bh, | |
6000 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6001 | { | |
6002 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6003 | struct ocfs2_xattr_header *header = (struct ocfs2_xattr_header *) | |
6004 | (fe_bh->b_data + inode->i_sb->s_blocksize - | |
6005 | le16_to_cpu(di->i_xattr_inline_size)); | |
6006 | struct ocfs2_xattr_value_buf vb = { | |
6007 | .vb_bh = fe_bh, | |
6008 | .vb_access = ocfs2_journal_access_di, | |
6009 | }; | |
6010 | ||
6011 | return ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6012 | ref_ci, ref_root_bh, dealloc); | |
6013 | } | |
6014 | ||
6015 | struct ocfs2_xattr_tree_value_refcount_para { | |
6016 | struct ocfs2_caching_info *ref_ci; | |
6017 | struct buffer_head *ref_root_bh; | |
6018 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
6019 | }; | |
6020 | ||
6021 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, | |
6022 | struct ocfs2_xattr_bucket *bucket, | |
6023 | int offset, | |
6024 | struct ocfs2_xattr_value_root **xv, | |
6025 | struct buffer_head **bh) | |
6026 | { | |
6027 | int ret, block_off, name_offset; | |
6028 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); | |
6029 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6030 | void *base; | |
6031 | ||
6032 | ret = ocfs2_xattr_bucket_get_name_value(sb, | |
6033 | bucket_xh(bucket), | |
6034 | offset, | |
6035 | &block_off, | |
6036 | &name_offset); | |
6037 | if (ret) { | |
6038 | mlog_errno(ret); | |
6039 | goto out; | |
6040 | } | |
6041 | ||
6042 | base = bucket_block(bucket, block_off); | |
6043 | ||
6044 | *xv = (struct ocfs2_xattr_value_root *)(base + name_offset + | |
6045 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6046 | ||
6047 | if (bh) | |
6048 | *bh = bucket->bu_bhs[block_off]; | |
6049 | out: | |
6050 | return ret; | |
6051 | } | |
6052 | ||
6053 | /* | |
6054 | * For a given xattr bucket, refcount all the entries which | |
6055 | * have value stored outside. | |
6056 | */ | |
6057 | static int ocfs2_xattr_bucket_value_refcount(struct inode *inode, | |
6058 | struct ocfs2_xattr_bucket *bucket, | |
6059 | void *para) | |
6060 | { | |
6061 | int i, ret = 0; | |
6062 | struct ocfs2_extent_tree et; | |
6063 | struct ocfs2_xattr_tree_value_refcount_para *ref = | |
6064 | (struct ocfs2_xattr_tree_value_refcount_para *)para; | |
6065 | struct ocfs2_xattr_header *xh = | |
6066 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6067 | struct ocfs2_xattr_entry *xe; | |
6068 | struct ocfs2_xattr_value_buf vb = { | |
6069 | .vb_access = ocfs2_journal_access, | |
6070 | }; | |
6071 | struct ocfs2_post_refcount refcount = { | |
6072 | .credits = bucket->bu_blocks, | |
6073 | .para = bucket, | |
6074 | .func = ocfs2_xattr_bucket_post_refcount, | |
6075 | }; | |
6076 | struct ocfs2_post_refcount *p = NULL; | |
6077 | ||
6078 | /* We only need post_refcount if we support metaecc. */ | |
6079 | if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb))) | |
6080 | p = &refcount; | |
6081 | ||
6082 | mlog(0, "refcount bucket %llu, count = %u\n", | |
6083 | (unsigned long long)bucket_blkno(bucket), | |
6084 | le16_to_cpu(xh->xh_count)); | |
6085 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6086 | xe = &xh->xh_entries[i]; | |
6087 | ||
6088 | if (ocfs2_xattr_is_local(xe)) | |
6089 | continue; | |
6090 | ||
6091 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i, | |
6092 | &vb.vb_xv, &vb.vb_bh); | |
6093 | if (ret) { | |
6094 | mlog_errno(ret); | |
6095 | break; | |
6096 | } | |
6097 | ||
6098 | ocfs2_init_xattr_value_extent_tree(&et, | |
6099 | INODE_CACHE(inode), &vb); | |
6100 | ||
6101 | ret = ocfs2_xattr_value_attach_refcount(inode, vb.vb_xv, | |
6102 | &et, ref->ref_ci, | |
6103 | ref->ref_root_bh, | |
6104 | ref->dealloc, p); | |
6105 | if (ret) { | |
6106 | mlog_errno(ret); | |
6107 | break; | |
6108 | } | |
6109 | } | |
6110 | ||
6111 | return ret; | |
6112 | ||
6113 | } | |
6114 | ||
6115 | static int ocfs2_refcount_xattr_tree_rec(struct inode *inode, | |
6116 | struct buffer_head *root_bh, | |
6117 | u64 blkno, u32 cpos, u32 len, void *para) | |
6118 | { | |
6119 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
6120 | ocfs2_xattr_bucket_value_refcount, | |
6121 | para); | |
6122 | } | |
6123 | ||
6124 | static int ocfs2_xattr_block_attach_refcount(struct inode *inode, | |
6125 | struct buffer_head *blk_bh, | |
6126 | struct ocfs2_caching_info *ref_ci, | |
6127 | struct buffer_head *ref_root_bh, | |
6128 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6129 | { | |
6130 | int ret = 0; | |
6131 | struct ocfs2_xattr_block *xb = | |
6132 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6133 | ||
6134 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
6135 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
6136 | struct ocfs2_xattr_value_buf vb = { | |
6137 | .vb_bh = blk_bh, | |
6138 | .vb_access = ocfs2_journal_access_xb, | |
6139 | }; | |
6140 | ||
6141 | ret = ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6142 | ref_ci, ref_root_bh, | |
6143 | dealloc); | |
6144 | } else { | |
6145 | struct ocfs2_xattr_tree_value_refcount_para para = { | |
6146 | .ref_ci = ref_ci, | |
6147 | .ref_root_bh = ref_root_bh, | |
6148 | .dealloc = dealloc, | |
6149 | }; | |
6150 | ||
6151 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
6152 | ocfs2_refcount_xattr_tree_rec, | |
6153 | ¶); | |
6154 | } | |
6155 | ||
6156 | return ret; | |
6157 | } | |
6158 | ||
6159 | int ocfs2_xattr_attach_refcount_tree(struct inode *inode, | |
6160 | struct buffer_head *fe_bh, | |
6161 | struct ocfs2_caching_info *ref_ci, | |
6162 | struct buffer_head *ref_root_bh, | |
6163 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6164 | { | |
6165 | int ret = 0; | |
6166 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
6167 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6168 | struct buffer_head *blk_bh = NULL; | |
6169 | ||
6170 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
6171 | ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh, | |
6172 | ref_ci, ref_root_bh, | |
6173 | dealloc); | |
6174 | if (ret) { | |
6175 | mlog_errno(ret); | |
6176 | goto out; | |
6177 | } | |
6178 | } | |
6179 | ||
6180 | if (!di->i_xattr_loc) | |
6181 | goto out; | |
6182 | ||
6183 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), | |
6184 | &blk_bh); | |
6185 | if (ret < 0) { | |
6186 | mlog_errno(ret); | |
6187 | goto out; | |
6188 | } | |
6189 | ||
6190 | ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci, | |
6191 | ref_root_bh, dealloc); | |
6192 | if (ret) | |
6193 | mlog_errno(ret); | |
6194 | ||
6195 | brelse(blk_bh); | |
6196 | out: | |
6197 | ||
6198 | return ret; | |
6199 | } | |
6200 | ||
0fe9b66c | 6201 | typedef int (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); |
2999d12f TM |
6202 | /* |
6203 | * Store the information we need in xattr reflink. | |
6204 | * old_bh and new_bh are inode bh for the old and new inode. | |
6205 | */ | |
6206 | struct ocfs2_xattr_reflink { | |
6207 | struct inode *old_inode; | |
6208 | struct inode *new_inode; | |
6209 | struct buffer_head *old_bh; | |
6210 | struct buffer_head *new_bh; | |
6211 | struct ocfs2_caching_info *ref_ci; | |
6212 | struct buffer_head *ref_root_bh; | |
6213 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
0fe9b66c | 6214 | should_xattr_reflinked *xattr_reflinked; |
2999d12f TM |
6215 | }; |
6216 | ||
6217 | /* | |
6218 | * Given a xattr header and xe offset, | |
6219 | * return the proper xv and the corresponding bh. | |
6220 | * xattr in inode, block and xattr tree have different implementaions. | |
6221 | */ | |
6222 | typedef int (get_xattr_value_root)(struct super_block *sb, | |
6223 | struct buffer_head *bh, | |
6224 | struct ocfs2_xattr_header *xh, | |
6225 | int offset, | |
6226 | struct ocfs2_xattr_value_root **xv, | |
6227 | struct buffer_head **ret_bh, | |
6228 | void *para); | |
6229 | ||
6230 | /* | |
6231 | * Calculate all the xattr value root metadata stored in this xattr header and | |
6232 | * credits we need if we create them from the scratch. | |
6233 | * We use get_xattr_value_root so that all types of xattr container can use it. | |
6234 | */ | |
6235 | static int ocfs2_value_metas_in_xattr_header(struct super_block *sb, | |
6236 | struct buffer_head *bh, | |
6237 | struct ocfs2_xattr_header *xh, | |
6238 | int *metas, int *credits, | |
6239 | int *num_recs, | |
6240 | get_xattr_value_root *func, | |
6241 | void *para) | |
6242 | { | |
6243 | int i, ret = 0; | |
6244 | struct ocfs2_xattr_value_root *xv; | |
6245 | struct ocfs2_xattr_entry *xe; | |
6246 | ||
6247 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6248 | xe = &xh->xh_entries[i]; | |
6249 | if (ocfs2_xattr_is_local(xe)) | |
6250 | continue; | |
6251 | ||
6252 | ret = func(sb, bh, xh, i, &xv, NULL, para); | |
6253 | if (ret) { | |
6254 | mlog_errno(ret); | |
6255 | break; | |
6256 | } | |
6257 | ||
6258 | *metas += le16_to_cpu(xv->xr_list.l_tree_depth) * | |
6259 | le16_to_cpu(xv->xr_list.l_next_free_rec); | |
6260 | ||
6261 | *credits += ocfs2_calc_extend_credits(sb, | |
6262 | &def_xv.xv.xr_list, | |
6263 | le32_to_cpu(xv->xr_clusters)); | |
6264 | ||
6265 | /* | |
6266 | * If the value is a tree with depth > 1, We don't go deep | |
6267 | * to the extent block, so just calculate a maximum record num. | |
6268 | */ | |
6269 | if (!xv->xr_list.l_tree_depth) | |
8ff6af88 | 6270 | *num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); |
2999d12f TM |
6271 | else |
6272 | *num_recs += ocfs2_clusters_for_bytes(sb, | |
6273 | XATTR_SIZE_MAX); | |
6274 | } | |
6275 | ||
6276 | return ret; | |
6277 | } | |
6278 | ||
6279 | /* Used by xattr inode and block to return the right xv and buffer_head. */ | |
6280 | static int ocfs2_get_xattr_value_root(struct super_block *sb, | |
6281 | struct buffer_head *bh, | |
6282 | struct ocfs2_xattr_header *xh, | |
6283 | int offset, | |
6284 | struct ocfs2_xattr_value_root **xv, | |
6285 | struct buffer_head **ret_bh, | |
6286 | void *para) | |
6287 | { | |
6288 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6289 | ||
6290 | *xv = (struct ocfs2_xattr_value_root *)((void *)xh + | |
6291 | le16_to_cpu(xe->xe_name_offset) + | |
6292 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6293 | ||
6294 | if (ret_bh) | |
6295 | *ret_bh = bh; | |
6296 | ||
6297 | return 0; | |
6298 | } | |
6299 | ||
6300 | /* | |
6301 | * Lock the meta_ac and caculate how much credits we need for reflink xattrs. | |
6302 | * It is only used for inline xattr and xattr block. | |
6303 | */ | |
6304 | static int ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, | |
6305 | struct ocfs2_xattr_header *xh, | |
6306 | struct buffer_head *ref_root_bh, | |
6307 | int *credits, | |
6308 | struct ocfs2_alloc_context **meta_ac) | |
6309 | { | |
6310 | int ret, meta_add = 0, num_recs = 0; | |
6311 | struct ocfs2_refcount_block *rb = | |
6312 | (struct ocfs2_refcount_block *)ref_root_bh->b_data; | |
6313 | ||
6314 | *credits = 0; | |
6315 | ||
6316 | ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh, | |
6317 | &meta_add, credits, &num_recs, | |
6318 | ocfs2_get_xattr_value_root, | |
6319 | NULL); | |
6320 | if (ret) { | |
6321 | mlog_errno(ret); | |
6322 | goto out; | |
6323 | } | |
6324 | ||
6325 | /* | |
6326 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6327 | * an approximate number we need for refcount tree change. | |
6328 | * Sometimes we need to split the tree, and after split, half recs | |
6329 | * will be moved to the new block, and a new block can only provide | |
6330 | * half number of recs. So we multiple new blocks by 2. | |
6331 | */ | |
6332 | num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6333 | meta_add += num_recs; | |
6334 | *credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6335 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6336 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6337 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6338 | else | |
6339 | *credits += 1; | |
6340 | ||
6341 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); | |
6342 | if (ret) | |
6343 | mlog_errno(ret); | |
6344 | ||
6345 | out: | |
6346 | return ret; | |
6347 | } | |
6348 | ||
6349 | /* | |
6350 | * Given a xattr header, reflink all the xattrs in this container. | |
6351 | * It can be used for inode, block and bucket. | |
6352 | * | |
6353 | * NOTE: | |
6354 | * Before we call this function, the caller has memcpy the xattr in | |
6355 | * old_xh to the new_xh. | |
0fe9b66c TM |
6356 | * |
6357 | * If args.xattr_reflinked is set, call it to decide whether the xe should | |
6358 | * be reflinked or not. If not, remove it from the new xattr header. | |
2999d12f TM |
6359 | */ |
6360 | static int ocfs2_reflink_xattr_header(handle_t *handle, | |
6361 | struct ocfs2_xattr_reflink *args, | |
6362 | struct buffer_head *old_bh, | |
6363 | struct ocfs2_xattr_header *xh, | |
6364 | struct buffer_head *new_bh, | |
6365 | struct ocfs2_xattr_header *new_xh, | |
6366 | struct ocfs2_xattr_value_buf *vb, | |
6367 | struct ocfs2_alloc_context *meta_ac, | |
6368 | get_xattr_value_root *func, | |
6369 | void *para) | |
6370 | { | |
0fe9b66c | 6371 | int ret = 0, i, j; |
2999d12f TM |
6372 | struct super_block *sb = args->old_inode->i_sb; |
6373 | struct buffer_head *value_bh; | |
0fe9b66c | 6374 | struct ocfs2_xattr_entry *xe, *last; |
2999d12f TM |
6375 | struct ocfs2_xattr_value_root *xv, *new_xv; |
6376 | struct ocfs2_extent_tree data_et; | |
6377 | u32 clusters, cpos, p_cluster, num_clusters; | |
6378 | unsigned int ext_flags = 0; | |
6379 | ||
6380 | mlog(0, "reflink xattr in container %llu, count = %u\n", | |
6381 | (unsigned long long)old_bh->b_blocknr, le16_to_cpu(xh->xh_count)); | |
0fe9b66c TM |
6382 | |
6383 | last = &new_xh->xh_entries[le16_to_cpu(new_xh->xh_count)]; | |
6384 | for (i = 0, j = 0; i < le16_to_cpu(xh->xh_count); i++, j++) { | |
2999d12f TM |
6385 | xe = &xh->xh_entries[i]; |
6386 | ||
0fe9b66c TM |
6387 | if (args->xattr_reflinked && !args->xattr_reflinked(xe)) { |
6388 | xe = &new_xh->xh_entries[j]; | |
6389 | ||
6390 | le16_add_cpu(&new_xh->xh_count, -1); | |
6391 | if (new_xh->xh_count) { | |
6392 | memmove(xe, xe + 1, | |
6393 | (void *)last - (void *)xe); | |
6394 | memset(last, 0, | |
6395 | sizeof(struct ocfs2_xattr_entry)); | |
6396 | } | |
6397 | ||
6398 | /* | |
6399 | * We don't want j to increase in the next round since | |
6400 | * it is already moved ahead. | |
6401 | */ | |
6402 | j--; | |
6403 | continue; | |
6404 | } | |
6405 | ||
2999d12f TM |
6406 | if (ocfs2_xattr_is_local(xe)) |
6407 | continue; | |
6408 | ||
6409 | ret = func(sb, old_bh, xh, i, &xv, NULL, para); | |
6410 | if (ret) { | |
6411 | mlog_errno(ret); | |
6412 | break; | |
6413 | } | |
6414 | ||
0fe9b66c | 6415 | ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); |
2999d12f TM |
6416 | if (ret) { |
6417 | mlog_errno(ret); | |
6418 | break; | |
6419 | } | |
6420 | ||
6421 | /* | |
6422 | * For the xattr which has l_tree_depth = 0, all the extent | |
6423 | * recs have already be copied to the new xh with the | |
6424 | * propriate OCFS2_EXT_REFCOUNTED flag we just need to | |
6425 | * increase the refount count int the refcount tree. | |
6426 | * | |
6427 | * For the xattr which has l_tree_depth > 0, we need | |
6428 | * to initialize it to the empty default value root, | |
6429 | * and then insert the extents one by one. | |
6430 | */ | |
6431 | if (xv->xr_list.l_tree_depth) { | |
6432 | memcpy(new_xv, &def_xv, sizeof(def_xv)); | |
6433 | vb->vb_xv = new_xv; | |
6434 | vb->vb_bh = value_bh; | |
6435 | ocfs2_init_xattr_value_extent_tree(&data_et, | |
6436 | INODE_CACHE(args->new_inode), vb); | |
6437 | } | |
6438 | ||
6439 | clusters = le32_to_cpu(xv->xr_clusters); | |
6440 | cpos = 0; | |
6441 | while (cpos < clusters) { | |
6442 | ret = ocfs2_xattr_get_clusters(args->old_inode, | |
6443 | cpos, | |
6444 | &p_cluster, | |
6445 | &num_clusters, | |
6446 | &xv->xr_list, | |
6447 | &ext_flags); | |
6448 | if (ret) { | |
6449 | mlog_errno(ret); | |
6450 | goto out; | |
6451 | } | |
6452 | ||
6453 | BUG_ON(!p_cluster); | |
6454 | ||
6455 | if (xv->xr_list.l_tree_depth) { | |
6456 | ret = ocfs2_insert_extent(handle, | |
6457 | &data_et, cpos, | |
6458 | ocfs2_clusters_to_blocks( | |
6459 | args->old_inode->i_sb, | |
6460 | p_cluster), | |
6461 | num_clusters, ext_flags, | |
6462 | meta_ac); | |
6463 | if (ret) { | |
6464 | mlog_errno(ret); | |
6465 | goto out; | |
6466 | } | |
6467 | } | |
6468 | ||
6469 | ret = ocfs2_increase_refcount(handle, args->ref_ci, | |
6470 | args->ref_root_bh, | |
6471 | p_cluster, num_clusters, | |
6472 | meta_ac, args->dealloc); | |
6473 | if (ret) { | |
6474 | mlog_errno(ret); | |
6475 | goto out; | |
6476 | } | |
6477 | ||
6478 | cpos += num_clusters; | |
6479 | } | |
6480 | } | |
6481 | ||
6482 | out: | |
6483 | return ret; | |
6484 | } | |
6485 | ||
6486 | static int ocfs2_reflink_xattr_inline(struct ocfs2_xattr_reflink *args) | |
6487 | { | |
6488 | int ret = 0, credits = 0; | |
6489 | handle_t *handle; | |
6490 | struct ocfs2_super *osb = OCFS2_SB(args->old_inode->i_sb); | |
6491 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)args->old_bh->b_data; | |
6492 | int inline_size = le16_to_cpu(di->i_xattr_inline_size); | |
6493 | int header_off = osb->sb->s_blocksize - inline_size; | |
6494 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *) | |
6495 | (args->old_bh->b_data + header_off); | |
6496 | struct ocfs2_xattr_header *new_xh = (struct ocfs2_xattr_header *) | |
6497 | (args->new_bh->b_data + header_off); | |
6498 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6499 | struct ocfs2_inode_info *new_oi; | |
6500 | struct ocfs2_dinode *new_di; | |
6501 | struct ocfs2_xattr_value_buf vb = { | |
6502 | .vb_bh = args->new_bh, | |
6503 | .vb_access = ocfs2_journal_access_di, | |
6504 | }; | |
6505 | ||
6506 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6507 | &credits, &meta_ac); | |
6508 | if (ret) { | |
6509 | mlog_errno(ret); | |
6510 | goto out; | |
6511 | } | |
6512 | ||
6513 | handle = ocfs2_start_trans(osb, credits); | |
6514 | if (IS_ERR(handle)) { | |
6515 | ret = PTR_ERR(handle); | |
6516 | mlog_errno(ret); | |
6517 | goto out; | |
6518 | } | |
6519 | ||
6520 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(args->new_inode), | |
6521 | args->new_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6522 | if (ret) { | |
6523 | mlog_errno(ret); | |
6524 | goto out_commit; | |
6525 | } | |
6526 | ||
6527 | memcpy(args->new_bh->b_data + header_off, | |
6528 | args->old_bh->b_data + header_off, inline_size); | |
6529 | ||
6530 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6531 | new_di->i_xattr_inline_size = cpu_to_le16(inline_size); | |
6532 | ||
6533 | ret = ocfs2_reflink_xattr_header(handle, args, args->old_bh, xh, | |
6534 | args->new_bh, new_xh, &vb, meta_ac, | |
6535 | ocfs2_get_xattr_value_root, NULL); | |
6536 | if (ret) { | |
6537 | mlog_errno(ret); | |
6538 | goto out_commit; | |
6539 | } | |
6540 | ||
6541 | new_oi = OCFS2_I(args->new_inode); | |
6542 | spin_lock(&new_oi->ip_lock); | |
6543 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL | OCFS2_INLINE_XATTR_FL; | |
6544 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6545 | spin_unlock(&new_oi->ip_lock); | |
6546 | ||
6547 | ocfs2_journal_dirty(handle, args->new_bh); | |
6548 | ||
6549 | out_commit: | |
6550 | ocfs2_commit_trans(osb, handle); | |
6551 | ||
6552 | out: | |
6553 | if (meta_ac) | |
6554 | ocfs2_free_alloc_context(meta_ac); | |
6555 | return ret; | |
6556 | } | |
6557 | ||
6558 | static int ocfs2_create_empty_xattr_block(struct inode *inode, | |
6559 | struct buffer_head *fe_bh, | |
6560 | struct buffer_head **ret_bh, | |
6561 | int indexed) | |
6562 | { | |
6563 | int ret; | |
6564 | handle_t *handle; | |
6565 | struct ocfs2_alloc_context *meta_ac; | |
6566 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
6567 | ||
6568 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac); | |
6569 | if (ret < 0) { | |
6570 | mlog_errno(ret); | |
6571 | return ret; | |
6572 | } | |
6573 | ||
6574 | handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS); | |
6575 | if (IS_ERR(handle)) { | |
6576 | ret = PTR_ERR(handle); | |
6577 | mlog_errno(ret); | |
6578 | goto out; | |
6579 | } | |
6580 | ||
6581 | mlog(0, "create new xattr block for inode %llu, index = %d\n", | |
6582 | (unsigned long long)fe_bh->b_blocknr, indexed); | |
6583 | ret = ocfs2_create_xattr_block(handle, inode, fe_bh, | |
6584 | meta_ac, ret_bh, indexed); | |
6585 | if (ret) | |
6586 | mlog_errno(ret); | |
6587 | ||
6588 | ocfs2_commit_trans(osb, handle); | |
6589 | out: | |
6590 | ocfs2_free_alloc_context(meta_ac); | |
6591 | return ret; | |
6592 | } | |
6593 | ||
6594 | static int ocfs2_reflink_xattr_block(struct ocfs2_xattr_reflink *args, | |
6595 | struct buffer_head *blk_bh, | |
6596 | struct buffer_head *new_blk_bh) | |
6597 | { | |
6598 | int ret = 0, credits = 0; | |
6599 | handle_t *handle; | |
6600 | struct ocfs2_inode_info *new_oi = OCFS2_I(args->new_inode); | |
6601 | struct ocfs2_dinode *new_di; | |
6602 | struct ocfs2_super *osb = OCFS2_SB(args->new_inode->i_sb); | |
6603 | int header_off = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
6604 | struct ocfs2_xattr_block *xb = | |
6605 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6606 | struct ocfs2_xattr_header *xh = &xb->xb_attrs.xb_header; | |
6607 | struct ocfs2_xattr_block *new_xb = | |
6608 | (struct ocfs2_xattr_block *)new_blk_bh->b_data; | |
6609 | struct ocfs2_xattr_header *new_xh = &new_xb->xb_attrs.xb_header; | |
6610 | struct ocfs2_alloc_context *meta_ac; | |
6611 | struct ocfs2_xattr_value_buf vb = { | |
6612 | .vb_bh = new_blk_bh, | |
6613 | .vb_access = ocfs2_journal_access_xb, | |
6614 | }; | |
6615 | ||
6616 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6617 | &credits, &meta_ac); | |
6618 | if (ret) { | |
6619 | mlog_errno(ret); | |
6620 | return ret; | |
6621 | } | |
6622 | ||
6623 | /* One more credits in case we need to add xattr flags in new inode. */ | |
6624 | handle = ocfs2_start_trans(osb, credits + 1); | |
6625 | if (IS_ERR(handle)) { | |
6626 | ret = PTR_ERR(handle); | |
6627 | mlog_errno(ret); | |
6628 | goto out; | |
6629 | } | |
6630 | ||
6631 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6632 | ret = ocfs2_journal_access_di(handle, | |
6633 | INODE_CACHE(args->new_inode), | |
6634 | args->new_bh, | |
6635 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6636 | if (ret) { | |
6637 | mlog_errno(ret); | |
6638 | goto out_commit; | |
6639 | } | |
6640 | } | |
6641 | ||
6642 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode), | |
6643 | new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6644 | if (ret) { | |
6645 | mlog_errno(ret); | |
6646 | goto out_commit; | |
6647 | } | |
6648 | ||
6649 | memcpy(new_blk_bh->b_data + header_off, blk_bh->b_data + header_off, | |
6650 | osb->sb->s_blocksize - header_off); | |
6651 | ||
6652 | ret = ocfs2_reflink_xattr_header(handle, args, blk_bh, xh, | |
6653 | new_blk_bh, new_xh, &vb, meta_ac, | |
6654 | ocfs2_get_xattr_value_root, NULL); | |
6655 | if (ret) { | |
6656 | mlog_errno(ret); | |
6657 | goto out_commit; | |
6658 | } | |
6659 | ||
6660 | ocfs2_journal_dirty(handle, new_blk_bh); | |
6661 | ||
6662 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6663 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6664 | spin_lock(&new_oi->ip_lock); | |
6665 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
6666 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6667 | spin_unlock(&new_oi->ip_lock); | |
6668 | ||
6669 | ocfs2_journal_dirty(handle, args->new_bh); | |
6670 | } | |
6671 | ||
6672 | out_commit: | |
6673 | ocfs2_commit_trans(osb, handle); | |
6674 | ||
6675 | out: | |
6676 | ocfs2_free_alloc_context(meta_ac); | |
6677 | return ret; | |
6678 | } | |
6679 | ||
6680 | struct ocfs2_reflink_xattr_tree_args { | |
6681 | struct ocfs2_xattr_reflink *reflink; | |
6682 | struct buffer_head *old_blk_bh; | |
6683 | struct buffer_head *new_blk_bh; | |
6684 | struct ocfs2_xattr_bucket *old_bucket; | |
6685 | struct ocfs2_xattr_bucket *new_bucket; | |
6686 | }; | |
6687 | ||
6688 | /* | |
6689 | * NOTE: | |
6690 | * We have to handle the case that both old bucket and new bucket | |
6691 | * will call this function to get the right ret_bh. | |
6692 | * So The caller must give us the right bh. | |
6693 | */ | |
6694 | static int ocfs2_get_reflink_xattr_value_root(struct super_block *sb, | |
6695 | struct buffer_head *bh, | |
6696 | struct ocfs2_xattr_header *xh, | |
6697 | int offset, | |
6698 | struct ocfs2_xattr_value_root **xv, | |
6699 | struct buffer_head **ret_bh, | |
6700 | void *para) | |
6701 | { | |
6702 | struct ocfs2_reflink_xattr_tree_args *args = | |
6703 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6704 | struct ocfs2_xattr_bucket *bucket; | |
6705 | ||
6706 | if (bh == args->old_bucket->bu_bhs[0]) | |
6707 | bucket = args->old_bucket; | |
6708 | else | |
6709 | bucket = args->new_bucket; | |
6710 | ||
6711 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6712 | xv, ret_bh); | |
6713 | } | |
6714 | ||
6715 | struct ocfs2_value_tree_metas { | |
6716 | int num_metas; | |
6717 | int credits; | |
6718 | int num_recs; | |
6719 | }; | |
6720 | ||
6721 | static int ocfs2_value_tree_metas_in_bucket(struct super_block *sb, | |
6722 | struct buffer_head *bh, | |
6723 | struct ocfs2_xattr_header *xh, | |
6724 | int offset, | |
6725 | struct ocfs2_xattr_value_root **xv, | |
6726 | struct buffer_head **ret_bh, | |
6727 | void *para) | |
6728 | { | |
6729 | struct ocfs2_xattr_bucket *bucket = | |
6730 | (struct ocfs2_xattr_bucket *)para; | |
6731 | ||
6732 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6733 | xv, ret_bh); | |
6734 | } | |
6735 | ||
6736 | static int ocfs2_calc_value_tree_metas(struct inode *inode, | |
6737 | struct ocfs2_xattr_bucket *bucket, | |
6738 | void *para) | |
6739 | { | |
6740 | struct ocfs2_value_tree_metas *metas = | |
6741 | (struct ocfs2_value_tree_metas *)para; | |
6742 | struct ocfs2_xattr_header *xh = | |
6743 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6744 | ||
6745 | /* Add the credits for this bucket first. */ | |
6746 | metas->credits += bucket->bu_blocks; | |
6747 | return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0], | |
6748 | xh, &metas->num_metas, | |
6749 | &metas->credits, &metas->num_recs, | |
6750 | ocfs2_value_tree_metas_in_bucket, | |
6751 | bucket); | |
6752 | } | |
6753 | ||
6754 | /* | |
6755 | * Given a xattr extent rec starting from blkno and having len clusters, | |
6756 | * iterate all the buckets calculate how much metadata we need for reflinking | |
6757 | * all the ocfs2_xattr_value_root and lock the allocators accordingly. | |
6758 | */ | |
6759 | static int ocfs2_lock_reflink_xattr_rec_allocators( | |
6760 | struct ocfs2_reflink_xattr_tree_args *args, | |
6761 | struct ocfs2_extent_tree *xt_et, | |
6762 | u64 blkno, u32 len, int *credits, | |
6763 | struct ocfs2_alloc_context **meta_ac, | |
6764 | struct ocfs2_alloc_context **data_ac) | |
6765 | { | |
6766 | int ret, num_free_extents; | |
6767 | struct ocfs2_value_tree_metas metas; | |
6768 | struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); | |
6769 | struct ocfs2_refcount_block *rb; | |
6770 | ||
6771 | memset(&metas, 0, sizeof(metas)); | |
6772 | ||
6773 | ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len, | |
6774 | ocfs2_calc_value_tree_metas, &metas); | |
6775 | if (ret) { | |
6776 | mlog_errno(ret); | |
6777 | goto out; | |
6778 | } | |
6779 | ||
6780 | *credits = metas.credits; | |
6781 | ||
6782 | /* | |
6783 | * Calculate we need for refcount tree change. | |
6784 | * | |
6785 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6786 | * an approximate number we need for refcount tree change. | |
6787 | * Sometimes we need to split the tree, and after split, half recs | |
6788 | * will be moved to the new block, and a new block can only provide | |
6789 | * half number of recs. So we multiple new blocks by 2. | |
6790 | * In the end, we have to add credits for modifying the already | |
6791 | * existed refcount block. | |
6792 | */ | |
6793 | rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data; | |
6794 | metas.num_recs = | |
6795 | (metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) / | |
6796 | ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6797 | metas.num_metas += metas.num_recs; | |
6798 | *credits += metas.num_recs + | |
6799 | metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6800 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6801 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6802 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6803 | else | |
6804 | *credits += 1; | |
6805 | ||
6806 | /* count in the xattr tree change. */ | |
6807 | num_free_extents = ocfs2_num_free_extents(osb, xt_et); | |
6808 | if (num_free_extents < 0) { | |
6809 | ret = num_free_extents; | |
6810 | mlog_errno(ret); | |
6811 | goto out; | |
6812 | } | |
6813 | ||
6814 | if (num_free_extents < len) | |
6815 | metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el); | |
6816 | ||
6817 | *credits += ocfs2_calc_extend_credits(osb->sb, | |
6818 | xt_et->et_root_el, len); | |
6819 | ||
6820 | if (metas.num_metas) { | |
6821 | ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas, | |
6822 | meta_ac); | |
6823 | if (ret) { | |
6824 | mlog_errno(ret); | |
6825 | goto out; | |
6826 | } | |
6827 | } | |
6828 | ||
6829 | if (len) { | |
6830 | ret = ocfs2_reserve_clusters(osb, len, data_ac); | |
6831 | if (ret) | |
6832 | mlog_errno(ret); | |
6833 | } | |
6834 | out: | |
6835 | if (ret) { | |
6836 | if (*meta_ac) { | |
6837 | ocfs2_free_alloc_context(*meta_ac); | |
6838 | meta_ac = NULL; | |
6839 | } | |
6840 | } | |
6841 | ||
6842 | return ret; | |
6843 | } | |
6844 | ||
6845 | static int ocfs2_reflink_xattr_buckets(handle_t *handle, | |
6846 | u64 blkno, u64 new_blkno, u32 clusters, | |
6847 | struct ocfs2_alloc_context *meta_ac, | |
6848 | struct ocfs2_alloc_context *data_ac, | |
6849 | struct ocfs2_reflink_xattr_tree_args *args) | |
6850 | { | |
6851 | int i, j, ret = 0; | |
6852 | struct super_block *sb = args->reflink->old_inode->i_sb; | |
6853 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
6854 | u32 num_buckets = clusters * bpc; | |
6855 | int bpb = args->old_bucket->bu_blocks; | |
6856 | struct ocfs2_xattr_value_buf vb = { | |
6857 | .vb_access = ocfs2_journal_access, | |
6858 | }; | |
6859 | ||
6860 | for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) { | |
6861 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
6862 | if (ret) { | |
6863 | mlog_errno(ret); | |
6864 | break; | |
6865 | } | |
6866 | ||
6867 | ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno); | |
6868 | if (ret) { | |
6869 | mlog_errno(ret); | |
6870 | break; | |
6871 | } | |
6872 | ||
6873 | /* | |
6874 | * The real bucket num in this series of blocks is stored | |
6875 | * in the 1st bucket. | |
6876 | */ | |
6877 | if (i == 0) | |
6878 | num_buckets = le16_to_cpu( | |
6879 | bucket_xh(args->old_bucket)->xh_num_buckets); | |
6880 | ||
6881 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
6882 | args->new_bucket, | |
6883 | OCFS2_JOURNAL_ACCESS_CREATE); | |
6884 | if (ret) { | |
6885 | mlog_errno(ret); | |
6886 | break; | |
6887 | } | |
6888 | ||
6889 | for (j = 0; j < bpb; j++) | |
6890 | memcpy(bucket_block(args->new_bucket, j), | |
6891 | bucket_block(args->old_bucket, j), | |
6892 | sb->s_blocksize); | |
6893 | ||
6894 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
6895 | ||
6896 | ret = ocfs2_reflink_xattr_header(handle, args->reflink, | |
6897 | args->old_bucket->bu_bhs[0], | |
6898 | bucket_xh(args->old_bucket), | |
6899 | args->new_bucket->bu_bhs[0], | |
6900 | bucket_xh(args->new_bucket), | |
6901 | &vb, meta_ac, | |
6902 | ocfs2_get_reflink_xattr_value_root, | |
6903 | args); | |
6904 | if (ret) { | |
6905 | mlog_errno(ret); | |
6906 | break; | |
6907 | } | |
6908 | ||
6909 | /* | |
6910 | * Re-access and dirty the bucket to calculate metaecc. | |
6911 | * Because we may extend the transaction in reflink_xattr_header | |
6912 | * which will let the already accessed block gone. | |
6913 | */ | |
6914 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
6915 | args->new_bucket, | |
6916 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6917 | if (ret) { | |
6918 | mlog_errno(ret); | |
6919 | break; | |
6920 | } | |
6921 | ||
6922 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
6923 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6924 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6925 | } | |
6926 | ||
6927 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6928 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6929 | return ret; | |
6930 | } | |
6931 | /* | |
6932 | * Create the same xattr extent record in the new inode's xattr tree. | |
6933 | */ | |
6934 | static int ocfs2_reflink_xattr_rec(struct inode *inode, | |
6935 | struct buffer_head *root_bh, | |
6936 | u64 blkno, | |
6937 | u32 cpos, | |
6938 | u32 len, | |
6939 | void *para) | |
6940 | { | |
6941 | int ret, credits = 0; | |
6942 | u32 p_cluster, num_clusters; | |
6943 | u64 new_blkno; | |
6944 | handle_t *handle; | |
6945 | struct ocfs2_reflink_xattr_tree_args *args = | |
6946 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6947 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
6948 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6949 | struct ocfs2_alloc_context *data_ac = NULL; | |
6950 | struct ocfs2_extent_tree et; | |
6951 | ||
6952 | ocfs2_init_xattr_tree_extent_tree(&et, | |
6953 | INODE_CACHE(args->reflink->new_inode), | |
6954 | args->new_blk_bh); | |
6955 | ||
6956 | ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno, | |
6957 | len, &credits, | |
6958 | &meta_ac, &data_ac); | |
6959 | if (ret) { | |
6960 | mlog_errno(ret); | |
6961 | goto out; | |
6962 | } | |
6963 | ||
6964 | handle = ocfs2_start_trans(osb, credits); | |
6965 | if (IS_ERR(handle)) { | |
6966 | ret = PTR_ERR(handle); | |
6967 | mlog_errno(ret); | |
6968 | goto out; | |
6969 | } | |
6970 | ||
6971 | ret = ocfs2_claim_clusters(osb, handle, data_ac, | |
6972 | len, &p_cluster, &num_clusters); | |
6973 | if (ret) { | |
6974 | mlog_errno(ret); | |
6975 | goto out_commit; | |
6976 | } | |
6977 | ||
6978 | new_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cluster); | |
6979 | ||
6980 | mlog(0, "reflink xattr buckets %llu to %llu, len %u\n", | |
6981 | (unsigned long long)blkno, (unsigned long long)new_blkno, len); | |
6982 | ret = ocfs2_reflink_xattr_buckets(handle, blkno, new_blkno, len, | |
6983 | meta_ac, data_ac, args); | |
6984 | if (ret) { | |
6985 | mlog_errno(ret); | |
6986 | goto out_commit; | |
6987 | } | |
6988 | ||
6989 | mlog(0, "insert new xattr extent rec start %llu len %u to %u\n", | |
6990 | (unsigned long long)new_blkno, len, cpos); | |
6991 | ret = ocfs2_insert_extent(handle, &et, cpos, new_blkno, | |
6992 | len, 0, meta_ac); | |
6993 | if (ret) | |
6994 | mlog_errno(ret); | |
6995 | ||
6996 | out_commit: | |
6997 | ocfs2_commit_trans(osb, handle); | |
6998 | ||
6999 | out: | |
7000 | if (meta_ac) | |
7001 | ocfs2_free_alloc_context(meta_ac); | |
7002 | if (data_ac) | |
7003 | ocfs2_free_alloc_context(data_ac); | |
7004 | return ret; | |
7005 | } | |
7006 | ||
7007 | /* | |
7008 | * Create reflinked xattr buckets. | |
7009 | * We will add bucket one by one, and refcount all the xattrs in the bucket | |
7010 | * if they are stored outside. | |
7011 | */ | |
7012 | static int ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, | |
7013 | struct buffer_head *blk_bh, | |
7014 | struct buffer_head *new_blk_bh) | |
7015 | { | |
7016 | int ret; | |
7017 | struct ocfs2_reflink_xattr_tree_args para; | |
7018 | ||
7019 | memset(¶, 0, sizeof(para)); | |
7020 | para.reflink = args; | |
7021 | para.old_blk_bh = blk_bh; | |
7022 | para.new_blk_bh = new_blk_bh; | |
7023 | ||
7024 | para.old_bucket = ocfs2_xattr_bucket_new(args->old_inode); | |
7025 | if (!para.old_bucket) { | |
7026 | mlog_errno(-ENOMEM); | |
7027 | return -ENOMEM; | |
7028 | } | |
7029 | ||
7030 | para.new_bucket = ocfs2_xattr_bucket_new(args->new_inode); | |
7031 | if (!para.new_bucket) { | |
7032 | ret = -ENOMEM; | |
7033 | mlog_errno(ret); | |
7034 | goto out; | |
7035 | } | |
7036 | ||
7037 | ret = ocfs2_iterate_xattr_index_block(args->old_inode, blk_bh, | |
7038 | ocfs2_reflink_xattr_rec, | |
7039 | ¶); | |
7040 | if (ret) | |
7041 | mlog_errno(ret); | |
7042 | ||
7043 | out: | |
7044 | ocfs2_xattr_bucket_free(para.old_bucket); | |
7045 | ocfs2_xattr_bucket_free(para.new_bucket); | |
7046 | return ret; | |
7047 | } | |
7048 | ||
7049 | static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args, | |
7050 | struct buffer_head *blk_bh) | |
7051 | { | |
7052 | int ret, indexed = 0; | |
7053 | struct buffer_head *new_blk_bh = NULL; | |
7054 | struct ocfs2_xattr_block *xb = | |
7055 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
7056 | ||
7057 | ||
7058 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) | |
7059 | indexed = 1; | |
7060 | ||
7061 | ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh, | |
7062 | &new_blk_bh, indexed); | |
7063 | if (ret) { | |
7064 | mlog_errno(ret); | |
7065 | goto out; | |
7066 | } | |
7067 | ||
7068 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) | |
7069 | ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); | |
7070 | else | |
7071 | ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); | |
7072 | if (ret) | |
7073 | mlog_errno(ret); | |
7074 | ||
7075 | out: | |
7076 | brelse(new_blk_bh); | |
7077 | return ret; | |
7078 | } | |
7079 | ||
0fe9b66c TM |
7080 | static int ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe) |
7081 | { | |
7082 | int type = ocfs2_xattr_get_type(xe); | |
7083 | ||
7084 | return type != OCFS2_XATTR_INDEX_SECURITY && | |
7085 | type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS && | |
7086 | type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT; | |
7087 | } | |
7088 | ||
2999d12f TM |
7089 | int ocfs2_reflink_xattrs(struct inode *old_inode, |
7090 | struct buffer_head *old_bh, | |
7091 | struct inode *new_inode, | |
0fe9b66c TM |
7092 | struct buffer_head *new_bh, |
7093 | bool preserve_security) | |
2999d12f TM |
7094 | { |
7095 | int ret; | |
7096 | struct ocfs2_xattr_reflink args; | |
7097 | struct ocfs2_inode_info *oi = OCFS2_I(old_inode); | |
7098 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)old_bh->b_data; | |
7099 | struct buffer_head *blk_bh = NULL; | |
7100 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
7101 | struct ocfs2_refcount_tree *ref_tree; | |
7102 | struct buffer_head *ref_root_bh = NULL; | |
7103 | ||
7104 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7105 | le64_to_cpu(di->i_refcount_loc), | |
7106 | 1, &ref_tree, &ref_root_bh); | |
7107 | if (ret) { | |
7108 | mlog_errno(ret); | |
7109 | goto out; | |
7110 | } | |
7111 | ||
7112 | ocfs2_init_dealloc_ctxt(&dealloc); | |
7113 | ||
7114 | args.old_inode = old_inode; | |
7115 | args.new_inode = new_inode; | |
7116 | args.old_bh = old_bh; | |
7117 | args.new_bh = new_bh; | |
7118 | args.ref_ci = &ref_tree->rf_ci; | |
7119 | args.ref_root_bh = ref_root_bh; | |
7120 | args.dealloc = &dealloc; | |
0fe9b66c TM |
7121 | if (preserve_security) |
7122 | args.xattr_reflinked = NULL; | |
7123 | else | |
7124 | args.xattr_reflinked = ocfs2_reflink_xattr_no_security; | |
2999d12f TM |
7125 | |
7126 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
7127 | ret = ocfs2_reflink_xattr_inline(&args); | |
7128 | if (ret) { | |
7129 | mlog_errno(ret); | |
7130 | goto out_unlock; | |
7131 | } | |
7132 | } | |
7133 | ||
7134 | if (!di->i_xattr_loc) | |
7135 | goto out_unlock; | |
7136 | ||
7137 | ret = ocfs2_read_xattr_block(old_inode, le64_to_cpu(di->i_xattr_loc), | |
7138 | &blk_bh); | |
7139 | if (ret < 0) { | |
7140 | mlog_errno(ret); | |
7141 | goto out_unlock; | |
7142 | } | |
7143 | ||
7144 | ret = ocfs2_reflink_xattr_in_block(&args, blk_bh); | |
7145 | if (ret) | |
7146 | mlog_errno(ret); | |
7147 | ||
7148 | brelse(blk_bh); | |
7149 | ||
7150 | out_unlock: | |
7151 | ocfs2_unlock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7152 | ref_tree, 1); | |
7153 | brelse(ref_root_bh); | |
7154 | ||
7155 | if (ocfs2_dealloc_has_cluster(&dealloc)) { | |
7156 | ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1); | |
7157 | ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc); | |
7158 | } | |
7159 | ||
7160 | out: | |
7161 | return ret; | |
7162 | } | |
7163 | ||
0fe9b66c TM |
7164 | /* |
7165 | * Initialize security and acl for a already created inode. | |
7166 | * Used for reflink a non-preserve-security file. | |
7167 | * | |
7168 | * It uses common api like ocfs2_xattr_set, so the caller | |
7169 | * must not hold any lock expect i_mutex. | |
7170 | */ | |
7171 | int ocfs2_init_security_and_acl(struct inode *dir, | |
7172 | struct inode *inode) | |
7173 | { | |
7174 | int ret = 0; | |
7175 | struct buffer_head *dir_bh = NULL; | |
7176 | struct ocfs2_security_xattr_info si = { | |
7177 | .enable = 1, | |
7178 | }; | |
7179 | ||
7180 | ret = ocfs2_init_security_get(inode, dir, &si); | |
7181 | if (!ret) { | |
431547b3 CH |
7182 | ret = ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, |
7183 | si.name, si.value, si.value_len, | |
7184 | XATTR_CREATE); | |
0fe9b66c TM |
7185 | if (ret) { |
7186 | mlog_errno(ret); | |
7187 | goto leave; | |
7188 | } | |
7189 | } else if (ret != -EOPNOTSUPP) { | |
7190 | mlog_errno(ret); | |
7191 | goto leave; | |
7192 | } | |
7193 | ||
7194 | ret = ocfs2_inode_lock(dir, &dir_bh, 0); | |
7195 | if (ret) { | |
7196 | mlog_errno(ret); | |
7197 | goto leave; | |
7198 | } | |
7199 | ||
7200 | ret = ocfs2_init_acl(NULL, inode, dir, NULL, dir_bh, NULL, NULL); | |
7201 | if (ret) | |
7202 | mlog_errno(ret); | |
7203 | ||
7204 | ocfs2_inode_unlock(dir, 0); | |
7205 | brelse(dir_bh); | |
7206 | leave: | |
7207 | return ret; | |
7208 | } | |
923f7f31 TY |
7209 | /* |
7210 | * 'security' attributes support | |
7211 | */ | |
431547b3 | 7212 | static size_t ocfs2_xattr_security_list(struct dentry *dentry, char *list, |
923f7f31 | 7213 | size_t list_size, const char *name, |
431547b3 | 7214 | size_t name_len, int type) |
923f7f31 TY |
7215 | { |
7216 | const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN; | |
7217 | const size_t total_len = prefix_len + name_len + 1; | |
7218 | ||
7219 | if (list && total_len <= list_size) { | |
7220 | memcpy(list, XATTR_SECURITY_PREFIX, prefix_len); | |
7221 | memcpy(list + prefix_len, name, name_len); | |
7222 | list[prefix_len + name_len] = '\0'; | |
7223 | } | |
7224 | return total_len; | |
7225 | } | |
7226 | ||
431547b3 CH |
7227 | static int ocfs2_xattr_security_get(struct dentry *dentry, const char *name, |
7228 | void *buffer, size_t size, int type) | |
923f7f31 TY |
7229 | { |
7230 | if (strcmp(name, "") == 0) | |
7231 | return -EINVAL; | |
431547b3 CH |
7232 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7233 | name, buffer, size); | |
923f7f31 TY |
7234 | } |
7235 | ||
431547b3 CH |
7236 | static int ocfs2_xattr_security_set(struct dentry *dentry, const char *name, |
7237 | const void *value, size_t size, int flags, int type) | |
923f7f31 TY |
7238 | { |
7239 | if (strcmp(name, "") == 0) | |
7240 | return -EINVAL; | |
7241 | ||
431547b3 CH |
7242 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7243 | name, value, size, flags); | |
923f7f31 TY |
7244 | } |
7245 | ||
534eaddd TY |
7246 | int ocfs2_init_security_get(struct inode *inode, |
7247 | struct inode *dir, | |
7248 | struct ocfs2_security_xattr_info *si) | |
7249 | { | |
38d59ef6 TY |
7250 | /* check whether ocfs2 support feature xattr */ |
7251 | if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) | |
7252 | return -EOPNOTSUPP; | |
534eaddd TY |
7253 | return security_inode_init_security(inode, dir, &si->name, &si->value, |
7254 | &si->value_len); | |
7255 | } | |
7256 | ||
7257 | int ocfs2_init_security_set(handle_t *handle, | |
7258 | struct inode *inode, | |
7259 | struct buffer_head *di_bh, | |
7260 | struct ocfs2_security_xattr_info *si, | |
7261 | struct ocfs2_alloc_context *xattr_ac, | |
7262 | struct ocfs2_alloc_context *data_ac) | |
7263 | { | |
7264 | return ocfs2_xattr_set_handle(handle, inode, di_bh, | |
7265 | OCFS2_XATTR_INDEX_SECURITY, | |
7266 | si->name, si->value, si->value_len, 0, | |
7267 | xattr_ac, data_ac); | |
7268 | } | |
7269 | ||
923f7f31 TY |
7270 | struct xattr_handler ocfs2_xattr_security_handler = { |
7271 | .prefix = XATTR_SECURITY_PREFIX, | |
7272 | .list = ocfs2_xattr_security_list, | |
7273 | .get = ocfs2_xattr_security_get, | |
7274 | .set = ocfs2_xattr_security_set, | |
7275 | }; | |
7276 | ||
99219aea MF |
7277 | /* |
7278 | * 'trusted' attributes support | |
7279 | */ | |
431547b3 | 7280 | static size_t ocfs2_xattr_trusted_list(struct dentry *dentry, char *list, |
99219aea | 7281 | size_t list_size, const char *name, |
431547b3 | 7282 | size_t name_len, int type) |
99219aea | 7283 | { |
ceb1eba3 | 7284 | const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN; |
99219aea MF |
7285 | const size_t total_len = prefix_len + name_len + 1; |
7286 | ||
7287 | if (list && total_len <= list_size) { | |
7288 | memcpy(list, XATTR_TRUSTED_PREFIX, prefix_len); | |
7289 | memcpy(list + prefix_len, name, name_len); | |
7290 | list[prefix_len + name_len] = '\0'; | |
7291 | } | |
7292 | return total_len; | |
7293 | } | |
7294 | ||
431547b3 CH |
7295 | static int ocfs2_xattr_trusted_get(struct dentry *dentry, const char *name, |
7296 | void *buffer, size_t size, int type) | |
99219aea MF |
7297 | { |
7298 | if (strcmp(name, "") == 0) | |
7299 | return -EINVAL; | |
431547b3 CH |
7300 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7301 | name, buffer, size); | |
99219aea MF |
7302 | } |
7303 | ||
431547b3 CH |
7304 | static int ocfs2_xattr_trusted_set(struct dentry *dentry, const char *name, |
7305 | const void *value, size_t size, int flags, int type) | |
99219aea MF |
7306 | { |
7307 | if (strcmp(name, "") == 0) | |
7308 | return -EINVAL; | |
7309 | ||
431547b3 CH |
7310 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7311 | name, value, size, flags); | |
99219aea MF |
7312 | } |
7313 | ||
7314 | struct xattr_handler ocfs2_xattr_trusted_handler = { | |
7315 | .prefix = XATTR_TRUSTED_PREFIX, | |
7316 | .list = ocfs2_xattr_trusted_list, | |
7317 | .get = ocfs2_xattr_trusted_get, | |
7318 | .set = ocfs2_xattr_trusted_set, | |
7319 | }; | |
7320 | ||
99219aea MF |
7321 | /* |
7322 | * 'user' attributes support | |
7323 | */ | |
431547b3 | 7324 | static size_t ocfs2_xattr_user_list(struct dentry *dentry, char *list, |
99219aea | 7325 | size_t list_size, const char *name, |
431547b3 | 7326 | size_t name_len, int type) |
99219aea | 7327 | { |
ceb1eba3 | 7328 | const size_t prefix_len = XATTR_USER_PREFIX_LEN; |
99219aea | 7329 | const size_t total_len = prefix_len + name_len + 1; |
431547b3 | 7330 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7331 | |
7332 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7333 | return 0; | |
7334 | ||
7335 | if (list && total_len <= list_size) { | |
7336 | memcpy(list, XATTR_USER_PREFIX, prefix_len); | |
7337 | memcpy(list + prefix_len, name, name_len); | |
7338 | list[prefix_len + name_len] = '\0'; | |
7339 | } | |
7340 | return total_len; | |
7341 | } | |
7342 | ||
431547b3 CH |
7343 | static int ocfs2_xattr_user_get(struct dentry *dentry, const char *name, |
7344 | void *buffer, size_t size, int type) | |
99219aea | 7345 | { |
431547b3 | 7346 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7347 | |
7348 | if (strcmp(name, "") == 0) | |
7349 | return -EINVAL; | |
7350 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7351 | return -EOPNOTSUPP; | |
431547b3 | 7352 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_USER, name, |
99219aea MF |
7353 | buffer, size); |
7354 | } | |
7355 | ||
431547b3 CH |
7356 | static int ocfs2_xattr_user_set(struct dentry *dentry, const char *name, |
7357 | const void *value, size_t size, int flags, int type) | |
99219aea | 7358 | { |
431547b3 | 7359 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7360 | |
7361 | if (strcmp(name, "") == 0) | |
7362 | return -EINVAL; | |
7363 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7364 | return -EOPNOTSUPP; | |
7365 | ||
431547b3 CH |
7366 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_USER, |
7367 | name, value, size, flags); | |
99219aea MF |
7368 | } |
7369 | ||
7370 | struct xattr_handler ocfs2_xattr_user_handler = { | |
7371 | .prefix = XATTR_USER_PREFIX, | |
7372 | .list = ocfs2_xattr_user_list, | |
7373 | .get = ocfs2_xattr_user_get, | |
7374 | .set = ocfs2_xattr_user_set, | |
7375 | }; |