]>
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 { | |
6b240ff6 JB |
119 | int xi_name_index; |
120 | const char *xi_name; | |
18853b95 | 121 | int xi_name_len; |
6b240ff6 JB |
122 | const void *xi_value; |
123 | size_t xi_value_len; | |
cf1d6c76 TY |
124 | }; |
125 | ||
126 | struct ocfs2_xattr_search { | |
127 | struct buffer_head *inode_bh; | |
128 | /* | |
129 | * xattr_bh point to the block buffer head which has extended attribute | |
130 | * when extended attribute in inode, xattr_bh is equal to inode_bh. | |
131 | */ | |
132 | struct buffer_head *xattr_bh; | |
133 | struct ocfs2_xattr_header *header; | |
ba937127 | 134 | struct ocfs2_xattr_bucket *bucket; |
cf1d6c76 TY |
135 | void *base; |
136 | void *end; | |
137 | struct ocfs2_xattr_entry *here; | |
138 | int not_found; | |
139 | }; | |
140 | ||
11179f2c JB |
141 | /* Operations on struct ocfs2_xa_entry */ |
142 | struct ocfs2_xa_loc; | |
143 | struct ocfs2_xa_loc_operations { | |
144 | /* | |
145 | * Return a pointer to the appropriate buffer in loc->xl_storage | |
146 | * at the given offset from loc->xl_header. | |
147 | */ | |
148 | void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset); | |
149 | ||
69a3e539 JB |
150 | /* Can we reuse the existing entry for the new value? */ |
151 | int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc, | |
152 | struct ocfs2_xattr_info *xi); | |
153 | ||
154 | /* How much space is needed for the new value? */ | |
155 | int (*xlo_check_space)(struct ocfs2_xa_loc *loc, | |
156 | struct ocfs2_xattr_info *xi); | |
157 | ||
158 | /* | |
159 | * Return the offset of the first name+value pair. This is | |
160 | * the start of our downward-filling free space. | |
161 | */ | |
162 | int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc); | |
163 | ||
11179f2c JB |
164 | /* |
165 | * Remove the name+value at this location. Do whatever is | |
166 | * appropriate with the remaining name+value pairs. | |
167 | */ | |
168 | void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc); | |
69a3e539 JB |
169 | |
170 | /* Fill xl_entry with a new entry */ | |
171 | void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash); | |
172 | ||
173 | /* Add name+value storage to an entry */ | |
174 | void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size); | |
3fc12afa JB |
175 | |
176 | /* | |
177 | * Initialize the value buf's access and bh fields for this entry. | |
178 | * ocfs2_xa_fill_value_buf() will handle the xv pointer. | |
179 | */ | |
180 | void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc, | |
181 | struct ocfs2_xattr_value_buf *vb); | |
11179f2c JB |
182 | }; |
183 | ||
184 | /* | |
185 | * Describes an xattr entry location. This is a memory structure | |
186 | * tracking the on-disk structure. | |
187 | */ | |
188 | struct ocfs2_xa_loc { | |
189 | /* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */ | |
190 | struct ocfs2_xattr_header *xl_header; | |
191 | ||
192 | /* Bytes from xl_header to the end of the storage */ | |
193 | int xl_size; | |
194 | ||
195 | /* | |
196 | * The ocfs2_xattr_entry this location describes. If this is | |
197 | * NULL, this location describes the on-disk structure where it | |
198 | * would have been. | |
199 | */ | |
200 | struct ocfs2_xattr_entry *xl_entry; | |
201 | ||
202 | /* | |
203 | * Internal housekeeping | |
204 | */ | |
205 | ||
206 | /* Buffer(s) containing this entry */ | |
207 | void *xl_storage; | |
208 | ||
209 | /* Operations on the storage backing this location */ | |
210 | const struct ocfs2_xa_loc_operations *xl_ops; | |
211 | }; | |
212 | ||
199799a3 JB |
213 | /* |
214 | * Convenience functions to calculate how much space is needed for a | |
215 | * given name+value pair | |
216 | */ | |
217 | static int namevalue_size(int name_len, uint64_t value_len) | |
218 | { | |
219 | if (value_len > OCFS2_XATTR_INLINE_SIZE) | |
220 | return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
221 | else | |
222 | return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len); | |
223 | } | |
224 | ||
225 | static int namevalue_size_xi(struct ocfs2_xattr_info *xi) | |
226 | { | |
227 | return namevalue_size(xi->xi_name_len, xi->xi_value_len); | |
228 | } | |
229 | ||
230 | static int namevalue_size_xe(struct ocfs2_xattr_entry *xe) | |
231 | { | |
232 | u64 value_len = le64_to_cpu(xe->xe_value_size); | |
233 | ||
234 | BUG_ON((value_len > OCFS2_XATTR_INLINE_SIZE) && | |
235 | ocfs2_xattr_is_local(xe)); | |
236 | return namevalue_size(xe->xe_name_len, value_len); | |
237 | } | |
238 | ||
239 | ||
fd68a894 | 240 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
589dc260 TM |
241 | struct ocfs2_xattr_header *xh, |
242 | int index, | |
243 | int *block_off, | |
244 | int *new_offset); | |
245 | ||
54f443f4 JB |
246 | static int ocfs2_xattr_block_find(struct inode *inode, |
247 | int name_index, | |
248 | const char *name, | |
249 | struct ocfs2_xattr_search *xs); | |
589dc260 TM |
250 | static int ocfs2_xattr_index_block_find(struct inode *inode, |
251 | struct buffer_head *root_bh, | |
252 | int name_index, | |
253 | const char *name, | |
254 | struct ocfs2_xattr_search *xs); | |
255 | ||
0c044f0b | 256 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, |
47bca495 | 257 | struct buffer_head *blk_bh, |
0c044f0b TM |
258 | char *buffer, |
259 | size_t buffer_size); | |
260 | ||
01225596 | 261 | static int ocfs2_xattr_create_index_block(struct inode *inode, |
78f30c31 TM |
262 | struct ocfs2_xattr_search *xs, |
263 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 TM |
264 | |
265 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, | |
266 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
267 | struct ocfs2_xattr_search *xs, |
268 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 | 269 | |
47bca495 TM |
270 | typedef int (xattr_tree_rec_func)(struct inode *inode, |
271 | struct buffer_head *root_bh, | |
272 | u64 blkno, u32 cpos, u32 len, void *para); | |
273 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, | |
274 | struct buffer_head *root_bh, | |
275 | xattr_tree_rec_func *rec_func, | |
276 | void *para); | |
277 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
278 | struct ocfs2_xattr_bucket *bucket, | |
279 | void *para); | |
280 | static int ocfs2_rm_xattr_cluster(struct inode *inode, | |
281 | struct buffer_head *root_bh, | |
282 | u64 blkno, | |
283 | u32 cpos, | |
284 | u32 len, | |
285 | void *para); | |
286 | ||
c58b6032 JB |
287 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
288 | u64 src_blk, u64 last_blk, u64 to_blk, | |
289 | unsigned int start_bucket, | |
290 | u32 *first_hash); | |
492a8a33 TM |
291 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, |
292 | struct ocfs2_dinode *di, | |
293 | struct ocfs2_xattr_info *xi, | |
294 | struct ocfs2_xattr_search *xis, | |
295 | struct ocfs2_xattr_search *xbs, | |
296 | struct ocfs2_refcount_tree **ref_tree, | |
297 | int *meta_need, | |
298 | int *credits); | |
ce9c5a54 TM |
299 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, |
300 | struct ocfs2_xattr_bucket *bucket, | |
301 | int offset, | |
302 | struct ocfs2_xattr_value_root **xv, | |
303 | struct buffer_head **bh); | |
a3944256 | 304 | |
0030e001 TY |
305 | static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb) |
306 | { | |
307 | return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE; | |
308 | } | |
309 | ||
310 | static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb) | |
311 | { | |
312 | return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits); | |
313 | } | |
314 | ||
315 | static inline u16 ocfs2_xattr_max_xe_in_bucket(struct super_block *sb) | |
316 | { | |
317 | u16 len = sb->s_blocksize - | |
318 | offsetof(struct ocfs2_xattr_header, xh_entries); | |
319 | ||
320 | return len / sizeof(struct ocfs2_xattr_entry); | |
321 | } | |
322 | ||
9c7759aa | 323 | #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr) |
51def39f | 324 | #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data) |
3e632946 | 325 | #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0)) |
9c7759aa | 326 | |
ba937127 | 327 | static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode) |
6dde41d9 | 328 | { |
ba937127 JB |
329 | struct ocfs2_xattr_bucket *bucket; |
330 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
6dde41d9 | 331 | |
ba937127 JB |
332 | BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET); |
333 | ||
334 | bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS); | |
335 | if (bucket) { | |
336 | bucket->bu_inode = inode; | |
337 | bucket->bu_blocks = blks; | |
338 | } | |
339 | ||
340 | return bucket; | |
341 | } | |
342 | ||
343 | static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket) | |
344 | { | |
345 | int i; | |
346 | ||
347 | for (i = 0; i < bucket->bu_blocks; i++) { | |
6dde41d9 JB |
348 | brelse(bucket->bu_bhs[i]); |
349 | bucket->bu_bhs[i] = NULL; | |
350 | } | |
351 | } | |
352 | ||
ba937127 JB |
353 | static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket) |
354 | { | |
355 | if (bucket) { | |
356 | ocfs2_xattr_bucket_relse(bucket); | |
357 | bucket->bu_inode = NULL; | |
358 | kfree(bucket); | |
359 | } | |
360 | } | |
361 | ||
784b816a JB |
362 | /* |
363 | * A bucket that has never been written to disk doesn't need to be | |
364 | * read. We just need the buffer_heads. Don't call this for | |
365 | * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes | |
366 | * them fully. | |
367 | */ | |
ba937127 | 368 | static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
369 | u64 xb_blkno) |
370 | { | |
371 | int i, rc = 0; | |
784b816a | 372 | |
ba937127 JB |
373 | for (i = 0; i < bucket->bu_blocks; i++) { |
374 | bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb, | |
375 | xb_blkno + i); | |
784b816a JB |
376 | if (!bucket->bu_bhs[i]) { |
377 | rc = -EIO; | |
378 | mlog_errno(rc); | |
379 | break; | |
380 | } | |
381 | ||
8cb471e8 | 382 | if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
757055ad | 383 | bucket->bu_bhs[i])) |
8cb471e8 | 384 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
757055ad | 385 | bucket->bu_bhs[i]); |
784b816a JB |
386 | } |
387 | ||
388 | if (rc) | |
ba937127 | 389 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
390 | return rc; |
391 | } | |
392 | ||
393 | /* Read the xattr bucket at xb_blkno */ | |
ba937127 | 394 | static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
395 | u64 xb_blkno) |
396 | { | |
ba937127 | 397 | int rc; |
784b816a | 398 | |
8cb471e8 | 399 | rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno, |
970e4936 JB |
400 | bucket->bu_blocks, bucket->bu_bhs, 0, |
401 | NULL); | |
4d0e214e | 402 | if (!rc) { |
c8b9cf9a | 403 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
404 | rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb, |
405 | bucket->bu_bhs, | |
406 | bucket->bu_blocks, | |
407 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 408 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
409 | if (rc) |
410 | mlog_errno(rc); | |
411 | } | |
412 | ||
784b816a | 413 | if (rc) |
ba937127 | 414 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
415 | return rc; |
416 | } | |
417 | ||
1224be02 | 418 | static int ocfs2_xattr_bucket_journal_access(handle_t *handle, |
1224be02 JB |
419 | struct ocfs2_xattr_bucket *bucket, |
420 | int type) | |
421 | { | |
422 | int i, rc = 0; | |
1224be02 | 423 | |
ba937127 | 424 | for (i = 0; i < bucket->bu_blocks; i++) { |
0cf2f763 JB |
425 | rc = ocfs2_journal_access(handle, |
426 | INODE_CACHE(bucket->bu_inode), | |
1224be02 JB |
427 | bucket->bu_bhs[i], type); |
428 | if (rc) { | |
429 | mlog_errno(rc); | |
430 | break; | |
431 | } | |
432 | } | |
433 | ||
434 | return rc; | |
435 | } | |
436 | ||
437 | static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle, | |
1224be02 JB |
438 | struct ocfs2_xattr_bucket *bucket) |
439 | { | |
ba937127 | 440 | int i; |
1224be02 | 441 | |
c8b9cf9a | 442 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
443 | ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb, |
444 | bucket->bu_bhs, bucket->bu_blocks, | |
445 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 446 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e | 447 | |
ba937127 | 448 | for (i = 0; i < bucket->bu_blocks; i++) |
1224be02 JB |
449 | ocfs2_journal_dirty(handle, bucket->bu_bhs[i]); |
450 | } | |
451 | ||
ba937127 | 452 | static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest, |
4980c6da JB |
453 | struct ocfs2_xattr_bucket *src) |
454 | { | |
455 | int i; | |
ba937127 JB |
456 | int blocksize = src->bu_inode->i_sb->s_blocksize; |
457 | ||
458 | BUG_ON(dest->bu_blocks != src->bu_blocks); | |
459 | BUG_ON(dest->bu_inode != src->bu_inode); | |
4980c6da | 460 | |
ba937127 | 461 | for (i = 0; i < src->bu_blocks; i++) { |
4980c6da JB |
462 | memcpy(bucket_block(dest, i), bucket_block(src, i), |
463 | blocksize); | |
464 | } | |
465 | } | |
1224be02 | 466 | |
4ae1d69b JB |
467 | static int ocfs2_validate_xattr_block(struct super_block *sb, |
468 | struct buffer_head *bh) | |
469 | { | |
d6b32bbb | 470 | int rc; |
4ae1d69b JB |
471 | struct ocfs2_xattr_block *xb = |
472 | (struct ocfs2_xattr_block *)bh->b_data; | |
473 | ||
474 | mlog(0, "Validating xattr block %llu\n", | |
475 | (unsigned long long)bh->b_blocknr); | |
476 | ||
d6b32bbb JB |
477 | BUG_ON(!buffer_uptodate(bh)); |
478 | ||
479 | /* | |
480 | * If the ecc fails, we return the error but otherwise | |
481 | * leave the filesystem running. We know any error is | |
482 | * local to this block. | |
483 | */ | |
484 | rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check); | |
485 | if (rc) | |
486 | return rc; | |
487 | ||
488 | /* | |
489 | * Errors after here are fatal | |
490 | */ | |
491 | ||
4ae1d69b JB |
492 | if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) { |
493 | ocfs2_error(sb, | |
494 | "Extended attribute block #%llu has bad " | |
495 | "signature %.*s", | |
496 | (unsigned long long)bh->b_blocknr, 7, | |
497 | xb->xb_signature); | |
498 | return -EINVAL; | |
499 | } | |
500 | ||
501 | if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) { | |
502 | ocfs2_error(sb, | |
503 | "Extended attribute block #%llu has an " | |
504 | "invalid xb_blkno of %llu", | |
505 | (unsigned long long)bh->b_blocknr, | |
506 | (unsigned long long)le64_to_cpu(xb->xb_blkno)); | |
507 | return -EINVAL; | |
508 | } | |
509 | ||
510 | if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) { | |
511 | ocfs2_error(sb, | |
512 | "Extended attribute block #%llu has an invalid " | |
513 | "xb_fs_generation of #%u", | |
514 | (unsigned long long)bh->b_blocknr, | |
515 | le32_to_cpu(xb->xb_fs_generation)); | |
516 | return -EINVAL; | |
517 | } | |
518 | ||
519 | return 0; | |
520 | } | |
521 | ||
522 | static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, | |
523 | struct buffer_head **bh) | |
524 | { | |
525 | int rc; | |
526 | struct buffer_head *tmp = *bh; | |
527 | ||
8cb471e8 | 528 | rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp, |
970e4936 | 529 | ocfs2_validate_xattr_block); |
4ae1d69b JB |
530 | |
531 | /* If ocfs2_read_block() got us a new bh, pass it up. */ | |
532 | if (!rc && !*bh) | |
533 | *bh = tmp; | |
534 | ||
535 | return rc; | |
536 | } | |
537 | ||
936b8834 | 538 | static inline const char *ocfs2_xattr_prefix(int name_index) |
cf1d6c76 TY |
539 | { |
540 | struct xattr_handler *handler = NULL; | |
541 | ||
542 | if (name_index > 0 && name_index < OCFS2_XATTR_MAX) | |
543 | handler = ocfs2_xattr_handler_map[name_index]; | |
544 | ||
936b8834 | 545 | return handler ? handler->prefix : NULL; |
cf1d6c76 TY |
546 | } |
547 | ||
40daa16a | 548 | static u32 ocfs2_xattr_name_hash(struct inode *inode, |
2057e5c6 | 549 | const char *name, |
40daa16a | 550 | int name_len) |
cf1d6c76 TY |
551 | { |
552 | /* Get hash value of uuid from super block */ | |
553 | u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash; | |
554 | int i; | |
555 | ||
cf1d6c76 TY |
556 | /* hash extended attribute name */ |
557 | for (i = 0; i < name_len; i++) { | |
558 | hash = (hash << OCFS2_HASH_SHIFT) ^ | |
559 | (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^ | |
560 | *name++; | |
561 | } | |
562 | ||
563 | return hash; | |
564 | } | |
565 | ||
566 | /* | |
567 | * ocfs2_xattr_hash_entry() | |
568 | * | |
569 | * Compute the hash of an extended attribute. | |
570 | */ | |
571 | static void ocfs2_xattr_hash_entry(struct inode *inode, | |
572 | struct ocfs2_xattr_header *header, | |
573 | struct ocfs2_xattr_entry *entry) | |
574 | { | |
575 | u32 hash = 0; | |
cf1d6c76 | 576 | char *name = (char *)header + le16_to_cpu(entry->xe_name_offset); |
cf1d6c76 | 577 | |
2057e5c6 | 578 | hash = ocfs2_xattr_name_hash(inode, name, entry->xe_name_len); |
cf1d6c76 TY |
579 | entry->xe_name_hash = cpu_to_le32(hash); |
580 | ||
581 | return; | |
582 | } | |
f56654c4 | 583 | |
534eaddd TY |
584 | static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len) |
585 | { | |
199799a3 JB |
586 | return namevalue_size(name_len, value_len) + |
587 | sizeof(struct ocfs2_xattr_entry); | |
588 | } | |
534eaddd | 589 | |
199799a3 JB |
590 | static int ocfs2_xi_entry_usage(struct ocfs2_xattr_info *xi) |
591 | { | |
592 | return namevalue_size_xi(xi) + | |
593 | sizeof(struct ocfs2_xattr_entry); | |
594 | } | |
534eaddd | 595 | |
199799a3 JB |
596 | static int ocfs2_xe_entry_usage(struct ocfs2_xattr_entry *xe) |
597 | { | |
598 | return namevalue_size_xe(xe) + | |
599 | sizeof(struct ocfs2_xattr_entry); | |
534eaddd TY |
600 | } |
601 | ||
602 | int ocfs2_calc_security_init(struct inode *dir, | |
603 | struct ocfs2_security_xattr_info *si, | |
604 | int *want_clusters, | |
605 | int *xattr_credits, | |
606 | struct ocfs2_alloc_context **xattr_ac) | |
607 | { | |
608 | int ret = 0; | |
609 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
610 | int s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
611 | si->value_len); | |
612 | ||
613 | /* | |
614 | * The max space of security xattr taken inline is | |
615 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
616 | * So reserve one metadata block for it is ok. | |
617 | */ | |
618 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
619 | s_size > OCFS2_XATTR_FREE_IN_IBODY) { | |
620 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac); | |
621 | if (ret) { | |
622 | mlog_errno(ret); | |
623 | return ret; | |
624 | } | |
625 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; | |
626 | } | |
627 | ||
628 | /* reserve clusters for xattr value which will be set in B tree*/ | |
0e445b6f TY |
629 | if (si->value_len > OCFS2_XATTR_INLINE_SIZE) { |
630 | int new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
631 | si->value_len); | |
632 | ||
633 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
634 | new_clusters); | |
635 | *want_clusters += new_clusters; | |
636 | } | |
534eaddd TY |
637 | return ret; |
638 | } | |
639 | ||
89c38bd0 TY |
640 | int ocfs2_calc_xattr_init(struct inode *dir, |
641 | struct buffer_head *dir_bh, | |
642 | int mode, | |
643 | struct ocfs2_security_xattr_info *si, | |
644 | int *want_clusters, | |
645 | int *xattr_credits, | |
9b7895ef | 646 | int *want_meta) |
89c38bd0 TY |
647 | { |
648 | int ret = 0; | |
649 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
0e445b6f | 650 | int s_size = 0, a_size = 0, acl_len = 0, new_clusters; |
89c38bd0 TY |
651 | |
652 | if (si->enable) | |
653 | s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
654 | si->value_len); | |
655 | ||
656 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) { | |
657 | acl_len = ocfs2_xattr_get_nolock(dir, dir_bh, | |
658 | OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT, | |
659 | "", NULL, 0); | |
660 | if (acl_len > 0) { | |
661 | a_size = ocfs2_xattr_entry_real_size(0, acl_len); | |
662 | if (S_ISDIR(mode)) | |
663 | a_size <<= 1; | |
664 | } else if (acl_len != 0 && acl_len != -ENODATA) { | |
665 | mlog_errno(ret); | |
666 | return ret; | |
667 | } | |
668 | } | |
669 | ||
670 | if (!(s_size + a_size)) | |
671 | return ret; | |
672 | ||
673 | /* | |
674 | * The max space of security xattr taken inline is | |
675 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
676 | * The max space of acl xattr taken inline is | |
677 | * 80(value) + 16(entry) * 2(if directory) = 192 bytes, | |
678 | * when blocksize = 512, may reserve one more cluser for | |
679 | * xattr bucket, otherwise reserve one metadata block | |
680 | * for them is ok. | |
6c9fd1dc TY |
681 | * If this is a new directory with inline data, |
682 | * we choose to reserve the entire inline area for | |
683 | * directory contents and force an external xattr block. | |
89c38bd0 TY |
684 | */ |
685 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
6c9fd1dc | 686 | (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) || |
89c38bd0 | 687 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) { |
9b7895ef | 688 | *want_meta = *want_meta + 1; |
89c38bd0 TY |
689 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
690 | } | |
691 | ||
692 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE && | |
693 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) { | |
694 | *want_clusters += 1; | |
695 | *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb); | |
696 | } | |
697 | ||
0e445b6f TY |
698 | /* |
699 | * reserve credits and clusters for xattrs which has large value | |
700 | * and have to be set outside | |
701 | */ | |
702 | if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
703 | new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
704 | si->value_len); | |
705 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
706 | new_clusters); | |
707 | *want_clusters += new_clusters; | |
708 | } | |
89c38bd0 TY |
709 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL && |
710 | acl_len > OCFS2_XATTR_INLINE_SIZE) { | |
0e445b6f TY |
711 | /* for directory, it has DEFAULT and ACCESS two types of acls */ |
712 | new_clusters = (S_ISDIR(mode) ? 2 : 1) * | |
713 | ocfs2_clusters_for_bytes(dir->i_sb, acl_len); | |
714 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
715 | new_clusters); | |
716 | *want_clusters += new_clusters; | |
89c38bd0 TY |
717 | } |
718 | ||
719 | return ret; | |
720 | } | |
721 | ||
f56654c4 TM |
722 | static int ocfs2_xattr_extend_allocation(struct inode *inode, |
723 | u32 clusters_to_add, | |
19b801f4 | 724 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 725 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
726 | { |
727 | int status = 0; | |
85db90e7 | 728 | handle_t *handle = ctxt->handle; |
f56654c4 | 729 | enum ocfs2_alloc_restarted why; |
19b801f4 | 730 | u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters); |
f99b9b7c | 731 | struct ocfs2_extent_tree et; |
f56654c4 TM |
732 | |
733 | mlog(0, "(clusters_to_add for xattr= %u)\n", clusters_to_add); | |
734 | ||
5e404e9e | 735 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f99b9b7c | 736 | |
0cf2f763 | 737 | status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
2a50a743 | 738 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
739 | if (status < 0) { |
740 | mlog_errno(status); | |
741 | goto leave; | |
742 | } | |
743 | ||
19b801f4 | 744 | prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
cbee7e1a JB |
745 | status = ocfs2_add_clusters_in_btree(handle, |
746 | &et, | |
f56654c4 TM |
747 | &logical_start, |
748 | clusters_to_add, | |
749 | 0, | |
78f30c31 TM |
750 | ctxt->data_ac, |
751 | ctxt->meta_ac, | |
f99b9b7c | 752 | &why); |
85db90e7 TM |
753 | if (status < 0) { |
754 | mlog_errno(status); | |
f56654c4 TM |
755 | goto leave; |
756 | } | |
757 | ||
19b801f4 | 758 | status = ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 TM |
759 | if (status < 0) { |
760 | mlog_errno(status); | |
761 | goto leave; | |
762 | } | |
763 | ||
19b801f4 | 764 | clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - prev_clusters; |
f56654c4 | 765 | |
85db90e7 TM |
766 | /* |
767 | * We should have already allocated enough space before the transaction, | |
768 | * so no need to restart. | |
769 | */ | |
770 | BUG_ON(why != RESTART_NONE || clusters_to_add); | |
f56654c4 TM |
771 | |
772 | leave: | |
f56654c4 TM |
773 | |
774 | return status; | |
775 | } | |
776 | ||
777 | static int __ocfs2_remove_xattr_range(struct inode *inode, | |
d72cc72d | 778 | struct ocfs2_xattr_value_buf *vb, |
f56654c4 | 779 | u32 cpos, u32 phys_cpos, u32 len, |
492a8a33 | 780 | unsigned int ext_flags, |
78f30c31 | 781 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
782 | { |
783 | int ret; | |
784 | u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
85db90e7 | 785 | handle_t *handle = ctxt->handle; |
f99b9b7c JB |
786 | struct ocfs2_extent_tree et; |
787 | ||
5e404e9e | 788 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f56654c4 | 789 | |
0cf2f763 | 790 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
d72cc72d | 791 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
792 | if (ret) { |
793 | mlog_errno(ret); | |
85db90e7 | 794 | goto out; |
f56654c4 TM |
795 | } |
796 | ||
dbdcf6a4 | 797 | ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac, |
78f30c31 | 798 | &ctxt->dealloc); |
f56654c4 TM |
799 | if (ret) { |
800 | mlog_errno(ret); | |
85db90e7 | 801 | goto out; |
f56654c4 TM |
802 | } |
803 | ||
d72cc72d | 804 | le32_add_cpu(&vb->vb_xv->xr_clusters, -len); |
f56654c4 | 805 | |
d72cc72d | 806 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 TM |
807 | if (ret) { |
808 | mlog_errno(ret); | |
85db90e7 | 809 | goto out; |
f56654c4 TM |
810 | } |
811 | ||
492a8a33 TM |
812 | if (ext_flags & OCFS2_EXT_REFCOUNTED) |
813 | ret = ocfs2_decrease_refcount(inode, handle, | |
814 | ocfs2_blocks_to_clusters(inode->i_sb, | |
815 | phys_blkno), | |
816 | len, ctxt->meta_ac, &ctxt->dealloc, 1); | |
817 | else | |
818 | ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, | |
819 | phys_blkno, len); | |
f56654c4 TM |
820 | if (ret) |
821 | mlog_errno(ret); | |
822 | ||
f56654c4 | 823 | out: |
f56654c4 TM |
824 | return ret; |
825 | } | |
826 | ||
827 | static int ocfs2_xattr_shrink_size(struct inode *inode, | |
828 | u32 old_clusters, | |
829 | u32 new_clusters, | |
19b801f4 | 830 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 831 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
832 | { |
833 | int ret = 0; | |
492a8a33 | 834 | unsigned int ext_flags; |
f56654c4 TM |
835 | u32 trunc_len, cpos, phys_cpos, alloc_size; |
836 | u64 block; | |
f56654c4 TM |
837 | |
838 | if (old_clusters <= new_clusters) | |
839 | return 0; | |
840 | ||
841 | cpos = new_clusters; | |
842 | trunc_len = old_clusters - new_clusters; | |
843 | while (trunc_len) { | |
844 | ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos, | |
d72cc72d | 845 | &alloc_size, |
492a8a33 | 846 | &vb->vb_xv->xr_list, &ext_flags); |
f56654c4 TM |
847 | if (ret) { |
848 | mlog_errno(ret); | |
849 | goto out; | |
850 | } | |
851 | ||
852 | if (alloc_size > trunc_len) | |
853 | alloc_size = trunc_len; | |
854 | ||
19b801f4 | 855 | ret = __ocfs2_remove_xattr_range(inode, vb, cpos, |
f56654c4 | 856 | phys_cpos, alloc_size, |
492a8a33 | 857 | ext_flags, ctxt); |
f56654c4 TM |
858 | if (ret) { |
859 | mlog_errno(ret); | |
860 | goto out; | |
861 | } | |
862 | ||
863 | block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
8cb471e8 JB |
864 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), |
865 | block, alloc_size); | |
f56654c4 TM |
866 | cpos += alloc_size; |
867 | trunc_len -= alloc_size; | |
868 | } | |
869 | ||
870 | out: | |
f56654c4 TM |
871 | return ret; |
872 | } | |
873 | ||
874 | static int ocfs2_xattr_value_truncate(struct inode *inode, | |
b3e5d379 | 875 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 TM |
876 | int len, |
877 | struct ocfs2_xattr_set_ctxt *ctxt) | |
f56654c4 TM |
878 | { |
879 | int ret; | |
880 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len); | |
b3e5d379 | 881 | u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
f56654c4 TM |
882 | |
883 | if (new_clusters == old_clusters) | |
884 | return 0; | |
885 | ||
886 | if (new_clusters > old_clusters) | |
887 | ret = ocfs2_xattr_extend_allocation(inode, | |
888 | new_clusters - old_clusters, | |
b3e5d379 | 889 | vb, ctxt); |
f56654c4 TM |
890 | else |
891 | ret = ocfs2_xattr_shrink_size(inode, | |
892 | old_clusters, new_clusters, | |
b3e5d379 | 893 | vb, ctxt); |
f56654c4 TM |
894 | |
895 | return ret; | |
896 | } | |
cf1d6c76 | 897 | |
936b8834 TM |
898 | static int ocfs2_xattr_list_entry(char *buffer, size_t size, |
899 | size_t *result, const char *prefix, | |
900 | const char *name, int name_len) | |
901 | { | |
902 | char *p = buffer + *result; | |
903 | int prefix_len = strlen(prefix); | |
904 | int total_len = prefix_len + name_len + 1; | |
905 | ||
906 | *result += total_len; | |
907 | ||
908 | /* we are just looking for how big our buffer needs to be */ | |
909 | if (!size) | |
910 | return 0; | |
911 | ||
912 | if (*result > size) | |
913 | return -ERANGE; | |
914 | ||
915 | memcpy(p, prefix, prefix_len); | |
916 | memcpy(p + prefix_len, name, name_len); | |
917 | p[prefix_len + name_len] = '\0'; | |
918 | ||
919 | return 0; | |
920 | } | |
921 | ||
cf1d6c76 TY |
922 | static int ocfs2_xattr_list_entries(struct inode *inode, |
923 | struct ocfs2_xattr_header *header, | |
924 | char *buffer, size_t buffer_size) | |
925 | { | |
936b8834 TM |
926 | size_t result = 0; |
927 | int i, type, ret; | |
928 | const char *prefix, *name; | |
cf1d6c76 TY |
929 | |
930 | for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) { | |
931 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
936b8834 TM |
932 | type = ocfs2_xattr_get_type(entry); |
933 | prefix = ocfs2_xattr_prefix(type); | |
934 | ||
935 | if (prefix) { | |
936 | name = (const char *)header + | |
937 | le16_to_cpu(entry->xe_name_offset); | |
938 | ||
939 | ret = ocfs2_xattr_list_entry(buffer, buffer_size, | |
940 | &result, prefix, name, | |
941 | entry->xe_name_len); | |
942 | if (ret) | |
943 | return ret; | |
cf1d6c76 TY |
944 | } |
945 | } | |
946 | ||
936b8834 | 947 | return result; |
cf1d6c76 TY |
948 | } |
949 | ||
8b2c0dba TM |
950 | int ocfs2_has_inline_xattr_value_outside(struct inode *inode, |
951 | struct ocfs2_dinode *di) | |
952 | { | |
953 | struct ocfs2_xattr_header *xh; | |
954 | int i; | |
955 | ||
956 | xh = (struct ocfs2_xattr_header *) | |
957 | ((void *)di + inode->i_sb->s_blocksize - | |
958 | le16_to_cpu(di->i_xattr_inline_size)); | |
959 | ||
960 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) | |
961 | if (!ocfs2_xattr_is_local(&xh->xh_entries[i])) | |
962 | return 1; | |
963 | ||
964 | return 0; | |
965 | } | |
966 | ||
cf1d6c76 TY |
967 | static int ocfs2_xattr_ibody_list(struct inode *inode, |
968 | struct ocfs2_dinode *di, | |
969 | char *buffer, | |
970 | size_t buffer_size) | |
971 | { | |
972 | struct ocfs2_xattr_header *header = NULL; | |
973 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
974 | int ret = 0; | |
975 | ||
976 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
977 | return ret; | |
978 | ||
979 | header = (struct ocfs2_xattr_header *) | |
980 | ((void *)di + inode->i_sb->s_blocksize - | |
981 | le16_to_cpu(di->i_xattr_inline_size)); | |
982 | ||
983 | ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size); | |
984 | ||
985 | return ret; | |
986 | } | |
987 | ||
988 | static int ocfs2_xattr_block_list(struct inode *inode, | |
989 | struct ocfs2_dinode *di, | |
990 | char *buffer, | |
991 | size_t buffer_size) | |
992 | { | |
993 | struct buffer_head *blk_bh = NULL; | |
0c044f0b | 994 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
995 | int ret = 0; |
996 | ||
997 | if (!di->i_xattr_loc) | |
998 | return ret; | |
999 | ||
4ae1d69b JB |
1000 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
1001 | &blk_bh); | |
cf1d6c76 TY |
1002 | if (ret < 0) { |
1003 | mlog_errno(ret); | |
1004 | return ret; | |
1005 | } | |
cf1d6c76 | 1006 | |
0c044f0b | 1007 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
0c044f0b TM |
1008 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
1009 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
1010 | ret = ocfs2_xattr_list_entries(inode, header, | |
1011 | buffer, buffer_size); | |
47bca495 TM |
1012 | } else |
1013 | ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh, | |
0c044f0b | 1014 | buffer, buffer_size); |
4ae1d69b | 1015 | |
cf1d6c76 TY |
1016 | brelse(blk_bh); |
1017 | ||
1018 | return ret; | |
1019 | } | |
1020 | ||
1021 | ssize_t ocfs2_listxattr(struct dentry *dentry, | |
1022 | char *buffer, | |
1023 | size_t size) | |
1024 | { | |
1025 | int ret = 0, i_ret = 0, b_ret = 0; | |
1026 | struct buffer_head *di_bh = NULL; | |
1027 | struct ocfs2_dinode *di = NULL; | |
1028 | struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode); | |
1029 | ||
8154da3d TY |
1030 | if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb))) |
1031 | return -EOPNOTSUPP; | |
1032 | ||
cf1d6c76 TY |
1033 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1034 | return ret; | |
1035 | ||
1036 | ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0); | |
1037 | if (ret < 0) { | |
1038 | mlog_errno(ret); | |
1039 | return ret; | |
1040 | } | |
1041 | ||
1042 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1043 | ||
1044 | down_read(&oi->ip_xattr_sem); | |
1045 | i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size); | |
1046 | if (i_ret < 0) | |
1047 | b_ret = 0; | |
1048 | else { | |
1049 | if (buffer) { | |
1050 | buffer += i_ret; | |
1051 | size -= i_ret; | |
1052 | } | |
1053 | b_ret = ocfs2_xattr_block_list(dentry->d_inode, di, | |
1054 | buffer, size); | |
1055 | if (b_ret < 0) | |
1056 | i_ret = 0; | |
1057 | } | |
1058 | up_read(&oi->ip_xattr_sem); | |
1059 | ocfs2_inode_unlock(dentry->d_inode, 0); | |
1060 | ||
1061 | brelse(di_bh); | |
1062 | ||
1063 | return i_ret + b_ret; | |
1064 | } | |
1065 | ||
1066 | static int ocfs2_xattr_find_entry(int name_index, | |
1067 | const char *name, | |
1068 | struct ocfs2_xattr_search *xs) | |
1069 | { | |
1070 | struct ocfs2_xattr_entry *entry; | |
1071 | size_t name_len; | |
1072 | int i, cmp = 1; | |
1073 | ||
1074 | if (name == NULL) | |
1075 | return -EINVAL; | |
1076 | ||
1077 | name_len = strlen(name); | |
1078 | entry = xs->here; | |
1079 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
1080 | cmp = name_index - ocfs2_xattr_get_type(entry); | |
1081 | if (!cmp) | |
1082 | cmp = name_len - entry->xe_name_len; | |
1083 | if (!cmp) | |
1084 | cmp = memcmp(name, (xs->base + | |
1085 | le16_to_cpu(entry->xe_name_offset)), | |
1086 | name_len); | |
1087 | if (cmp == 0) | |
1088 | break; | |
1089 | entry += 1; | |
1090 | } | |
1091 | xs->here = entry; | |
1092 | ||
1093 | return cmp ? -ENODATA : 0; | |
1094 | } | |
1095 | ||
1096 | static int ocfs2_xattr_get_value_outside(struct inode *inode, | |
589dc260 | 1097 | struct ocfs2_xattr_value_root *xv, |
cf1d6c76 TY |
1098 | void *buffer, |
1099 | size_t len) | |
1100 | { | |
1101 | u32 cpos, p_cluster, num_clusters, bpc, clusters; | |
1102 | u64 blkno; | |
1103 | int i, ret = 0; | |
1104 | size_t cplen, blocksize; | |
1105 | struct buffer_head *bh = NULL; | |
cf1d6c76 TY |
1106 | struct ocfs2_extent_list *el; |
1107 | ||
cf1d6c76 TY |
1108 | el = &xv->xr_list; |
1109 | clusters = le32_to_cpu(xv->xr_clusters); | |
1110 | bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1111 | blocksize = inode->i_sb->s_blocksize; | |
1112 | ||
1113 | cpos = 0; | |
1114 | while (cpos < clusters) { | |
1115 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1116 | &num_clusters, el, NULL); |
cf1d6c76 TY |
1117 | if (ret) { |
1118 | mlog_errno(ret); | |
1119 | goto out; | |
1120 | } | |
1121 | ||
1122 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
1123 | /* Copy ocfs2_xattr_value */ | |
1124 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1125 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1126 | &bh, NULL); | |
cf1d6c76 TY |
1127 | if (ret) { |
1128 | mlog_errno(ret); | |
1129 | goto out; | |
1130 | } | |
1131 | ||
1132 | cplen = len >= blocksize ? blocksize : len; | |
1133 | memcpy(buffer, bh->b_data, cplen); | |
1134 | len -= cplen; | |
1135 | buffer += cplen; | |
1136 | ||
1137 | brelse(bh); | |
1138 | bh = NULL; | |
1139 | if (len == 0) | |
1140 | break; | |
1141 | } | |
1142 | cpos += num_clusters; | |
1143 | } | |
1144 | out: | |
1145 | return ret; | |
1146 | } | |
1147 | ||
1148 | static int ocfs2_xattr_ibody_get(struct inode *inode, | |
1149 | int name_index, | |
1150 | const char *name, | |
1151 | void *buffer, | |
1152 | size_t buffer_size, | |
1153 | struct ocfs2_xattr_search *xs) | |
1154 | { | |
1155 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1156 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
589dc260 | 1157 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 TY |
1158 | size_t size; |
1159 | int ret = 0; | |
1160 | ||
1161 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
1162 | return -ENODATA; | |
1163 | ||
1164 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
1165 | xs->header = (struct ocfs2_xattr_header *) | |
1166 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
1167 | xs->base = (void *)xs->header; | |
1168 | xs->here = xs->header->xh_entries; | |
1169 | ||
1170 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
1171 | if (ret) | |
1172 | return ret; | |
1173 | size = le64_to_cpu(xs->here->xe_value_size); | |
1174 | if (buffer) { | |
1175 | if (size > buffer_size) | |
1176 | return -ERANGE; | |
1177 | if (ocfs2_xattr_is_local(xs->here)) { | |
1178 | memcpy(buffer, (void *)xs->base + | |
1179 | le16_to_cpu(xs->here->xe_name_offset) + | |
1180 | OCFS2_XATTR_SIZE(xs->here->xe_name_len), size); | |
1181 | } else { | |
589dc260 TM |
1182 | xv = (struct ocfs2_xattr_value_root *) |
1183 | (xs->base + le16_to_cpu( | |
1184 | xs->here->xe_name_offset) + | |
1185 | OCFS2_XATTR_SIZE(xs->here->xe_name_len)); | |
1186 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1187 | buffer, size); |
1188 | if (ret < 0) { | |
1189 | mlog_errno(ret); | |
1190 | return ret; | |
1191 | } | |
1192 | } | |
1193 | } | |
1194 | ||
1195 | return size; | |
1196 | } | |
1197 | ||
1198 | static int ocfs2_xattr_block_get(struct inode *inode, | |
1199 | int name_index, | |
1200 | const char *name, | |
1201 | void *buffer, | |
1202 | size_t buffer_size, | |
1203 | struct ocfs2_xattr_search *xs) | |
1204 | { | |
cf1d6c76 | 1205 | struct ocfs2_xattr_block *xb; |
589dc260 | 1206 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 | 1207 | size_t size; |
44d8e4e1 SM |
1208 | int ret = -ENODATA, name_offset, name_len, i; |
1209 | int uninitialized_var(block_off); | |
cf1d6c76 | 1210 | |
ba937127 JB |
1211 | xs->bucket = ocfs2_xattr_bucket_new(inode); |
1212 | if (!xs->bucket) { | |
1213 | ret = -ENOMEM; | |
1214 | mlog_errno(ret); | |
1215 | goto cleanup; | |
1216 | } | |
589dc260 | 1217 | |
54f443f4 JB |
1218 | ret = ocfs2_xattr_block_find(inode, name_index, name, xs); |
1219 | if (ret) { | |
cf1d6c76 | 1220 | mlog_errno(ret); |
cf1d6c76 TY |
1221 | goto cleanup; |
1222 | } | |
1223 | ||
6c1e183e TY |
1224 | if (xs->not_found) { |
1225 | ret = -ENODATA; | |
1226 | goto cleanup; | |
1227 | } | |
1228 | ||
54f443f4 | 1229 | xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
1230 | size = le64_to_cpu(xs->here->xe_value_size); |
1231 | if (buffer) { | |
1232 | ret = -ERANGE; | |
1233 | if (size > buffer_size) | |
1234 | goto cleanup; | |
589dc260 TM |
1235 | |
1236 | name_offset = le16_to_cpu(xs->here->xe_name_offset); | |
1237 | name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len); | |
1238 | i = xs->here - xs->header->xh_entries; | |
1239 | ||
1240 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 1241 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
ba937127 | 1242 | bucket_xh(xs->bucket), |
589dc260 TM |
1243 | i, |
1244 | &block_off, | |
1245 | &name_offset); | |
ba937127 | 1246 | xs->base = bucket_block(xs->bucket, block_off); |
589dc260 | 1247 | } |
cf1d6c76 TY |
1248 | if (ocfs2_xattr_is_local(xs->here)) { |
1249 | memcpy(buffer, (void *)xs->base + | |
589dc260 | 1250 | name_offset + name_len, size); |
cf1d6c76 | 1251 | } else { |
589dc260 TM |
1252 | xv = (struct ocfs2_xattr_value_root *) |
1253 | (xs->base + name_offset + name_len); | |
1254 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1255 | buffer, size); |
1256 | if (ret < 0) { | |
1257 | mlog_errno(ret); | |
1258 | goto cleanup; | |
1259 | } | |
1260 | } | |
1261 | } | |
1262 | ret = size; | |
1263 | cleanup: | |
ba937127 | 1264 | ocfs2_xattr_bucket_free(xs->bucket); |
cf1d6c76 | 1265 | |
54f443f4 JB |
1266 | brelse(xs->xattr_bh); |
1267 | xs->xattr_bh = NULL; | |
cf1d6c76 TY |
1268 | return ret; |
1269 | } | |
1270 | ||
4e3e9d02 TY |
1271 | int ocfs2_xattr_get_nolock(struct inode *inode, |
1272 | struct buffer_head *di_bh, | |
0030e001 TY |
1273 | int name_index, |
1274 | const char *name, | |
1275 | void *buffer, | |
1276 | size_t buffer_size) | |
cf1d6c76 TY |
1277 | { |
1278 | int ret; | |
1279 | struct ocfs2_dinode *di = NULL; | |
cf1d6c76 TY |
1280 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
1281 | struct ocfs2_xattr_search xis = { | |
1282 | .not_found = -ENODATA, | |
1283 | }; | |
1284 | struct ocfs2_xattr_search xbs = { | |
1285 | .not_found = -ENODATA, | |
1286 | }; | |
1287 | ||
8154da3d TY |
1288 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
1289 | return -EOPNOTSUPP; | |
1290 | ||
cf1d6c76 TY |
1291 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1292 | ret = -ENODATA; | |
1293 | ||
cf1d6c76 TY |
1294 | xis.inode_bh = xbs.inode_bh = di_bh; |
1295 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1296 | ||
1297 | down_read(&oi->ip_xattr_sem); | |
1298 | ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer, | |
1299 | buffer_size, &xis); | |
6c1e183e | 1300 | if (ret == -ENODATA && di->i_xattr_loc) |
cf1d6c76 TY |
1301 | ret = ocfs2_xattr_block_get(inode, name_index, name, buffer, |
1302 | buffer_size, &xbs); | |
1303 | up_read(&oi->ip_xattr_sem); | |
4e3e9d02 TY |
1304 | |
1305 | return ret; | |
1306 | } | |
1307 | ||
1308 | /* ocfs2_xattr_get() | |
1309 | * | |
1310 | * Copy an extended attribute into the buffer provided. | |
1311 | * Buffer is NULL to compute the size of buffer required. | |
1312 | */ | |
1313 | static int ocfs2_xattr_get(struct inode *inode, | |
1314 | int name_index, | |
1315 | const char *name, | |
1316 | void *buffer, | |
1317 | size_t buffer_size) | |
1318 | { | |
1319 | int ret; | |
1320 | struct buffer_head *di_bh = NULL; | |
1321 | ||
1322 | ret = ocfs2_inode_lock(inode, &di_bh, 0); | |
1323 | if (ret < 0) { | |
1324 | mlog_errno(ret); | |
1325 | return ret; | |
1326 | } | |
1327 | ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index, | |
1328 | name, buffer, buffer_size); | |
1329 | ||
cf1d6c76 TY |
1330 | ocfs2_inode_unlock(inode, 0); |
1331 | ||
1332 | brelse(di_bh); | |
1333 | ||
1334 | return ret; | |
1335 | } | |
1336 | ||
1337 | static int __ocfs2_xattr_set_value_outside(struct inode *inode, | |
85db90e7 | 1338 | handle_t *handle, |
492a8a33 | 1339 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1340 | const void *value, |
1341 | int value_len) | |
1342 | { | |
71d548a6 | 1343 | int ret = 0, i, cp_len; |
cf1d6c76 TY |
1344 | u16 blocksize = inode->i_sb->s_blocksize; |
1345 | u32 p_cluster, num_clusters; | |
1346 | u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1347 | u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len); | |
1348 | u64 blkno; | |
1349 | struct buffer_head *bh = NULL; | |
492a8a33 TM |
1350 | unsigned int ext_flags; |
1351 | struct ocfs2_xattr_value_root *xv = vb->vb_xv; | |
cf1d6c76 TY |
1352 | |
1353 | BUG_ON(clusters > le32_to_cpu(xv->xr_clusters)); | |
1354 | ||
cf1d6c76 TY |
1355 | while (cpos < clusters) { |
1356 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1357 | &num_clusters, &xv->xr_list, |
492a8a33 | 1358 | &ext_flags); |
cf1d6c76 TY |
1359 | if (ret) { |
1360 | mlog_errno(ret); | |
85db90e7 | 1361 | goto out; |
cf1d6c76 TY |
1362 | } |
1363 | ||
492a8a33 TM |
1364 | BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); |
1365 | ||
cf1d6c76 TY |
1366 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); |
1367 | ||
1368 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1369 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1370 | &bh, NULL); | |
cf1d6c76 TY |
1371 | if (ret) { |
1372 | mlog_errno(ret); | |
85db90e7 | 1373 | goto out; |
cf1d6c76 TY |
1374 | } |
1375 | ||
1376 | ret = ocfs2_journal_access(handle, | |
0cf2f763 | 1377 | INODE_CACHE(inode), |
cf1d6c76 TY |
1378 | bh, |
1379 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1380 | if (ret < 0) { | |
1381 | mlog_errno(ret); | |
85db90e7 | 1382 | goto out; |
cf1d6c76 TY |
1383 | } |
1384 | ||
1385 | cp_len = value_len > blocksize ? blocksize : value_len; | |
1386 | memcpy(bh->b_data, value, cp_len); | |
1387 | value_len -= cp_len; | |
1388 | value += cp_len; | |
1389 | if (cp_len < blocksize) | |
1390 | memset(bh->b_data + cp_len, 0, | |
1391 | blocksize - cp_len); | |
1392 | ||
1393 | ret = ocfs2_journal_dirty(handle, bh); | |
1394 | if (ret < 0) { | |
1395 | mlog_errno(ret); | |
85db90e7 | 1396 | goto out; |
cf1d6c76 TY |
1397 | } |
1398 | brelse(bh); | |
1399 | bh = NULL; | |
1400 | ||
1401 | /* | |
1402 | * XXX: do we need to empty all the following | |
1403 | * blocks in this cluster? | |
1404 | */ | |
1405 | if (!value_len) | |
1406 | break; | |
1407 | } | |
1408 | cpos += num_clusters; | |
1409 | } | |
cf1d6c76 TY |
1410 | out: |
1411 | brelse(bh); | |
1412 | ||
1413 | return ret; | |
1414 | } | |
1415 | ||
1416 | static int ocfs2_xattr_cleanup(struct inode *inode, | |
85db90e7 | 1417 | handle_t *handle, |
cf1d6c76 TY |
1418 | struct ocfs2_xattr_info *xi, |
1419 | struct ocfs2_xattr_search *xs, | |
512620f4 | 1420 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1421 | size_t offs) |
1422 | { | |
cf1d6c76 | 1423 | int ret = 0; |
cf1d6c76 | 1424 | void *val = xs->base + offs; |
199799a3 | 1425 | size_t size = namevalue_size_xi(xi); |
cf1d6c76 | 1426 | |
0cf2f763 | 1427 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
512620f4 | 1428 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1429 | if (ret) { |
1430 | mlog_errno(ret); | |
85db90e7 | 1431 | goto out; |
cf1d6c76 TY |
1432 | } |
1433 | /* Decrease xattr count */ | |
1434 | le16_add_cpu(&xs->header->xh_count, -1); | |
1435 | /* Remove the xattr entry and tree root which has already be set*/ | |
1436 | memset((void *)xs->here, 0, sizeof(struct ocfs2_xattr_entry)); | |
1437 | memset(val, 0, size); | |
1438 | ||
512620f4 | 1439 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
cf1d6c76 TY |
1440 | if (ret < 0) |
1441 | mlog_errno(ret); | |
cf1d6c76 TY |
1442 | out: |
1443 | return ret; | |
1444 | } | |
1445 | ||
1446 | static int ocfs2_xattr_update_entry(struct inode *inode, | |
85db90e7 | 1447 | handle_t *handle, |
cf1d6c76 TY |
1448 | struct ocfs2_xattr_info *xi, |
1449 | struct ocfs2_xattr_search *xs, | |
0c748e95 | 1450 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1451 | size_t offs) |
1452 | { | |
85db90e7 | 1453 | int ret; |
cf1d6c76 | 1454 | |
0cf2f763 | 1455 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
0c748e95 | 1456 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
1457 | if (ret) { |
1458 | mlog_errno(ret); | |
85db90e7 | 1459 | goto out; |
cf1d6c76 TY |
1460 | } |
1461 | ||
1462 | xs->here->xe_name_offset = cpu_to_le16(offs); | |
6b240ff6 JB |
1463 | xs->here->xe_value_size = cpu_to_le64(xi->xi_value_len); |
1464 | if (xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) | |
cf1d6c76 TY |
1465 | ocfs2_xattr_set_local(xs->here, 1); |
1466 | else | |
1467 | ocfs2_xattr_set_local(xs->here, 0); | |
1468 | ocfs2_xattr_hash_entry(inode, xs->header, xs->here); | |
1469 | ||
0c748e95 | 1470 | ret = ocfs2_journal_dirty(handle, vb->vb_bh); |
cf1d6c76 TY |
1471 | if (ret < 0) |
1472 | mlog_errno(ret); | |
cf1d6c76 TY |
1473 | out: |
1474 | return ret; | |
1475 | } | |
1476 | ||
1477 | /* | |
1478 | * ocfs2_xattr_set_value_outside() | |
1479 | * | |
1480 | * Set large size value in B tree. | |
1481 | */ | |
1482 | static int ocfs2_xattr_set_value_outside(struct inode *inode, | |
1483 | struct ocfs2_xattr_info *xi, | |
1484 | struct ocfs2_xattr_search *xs, | |
78f30c31 | 1485 | struct ocfs2_xattr_set_ctxt *ctxt, |
512620f4 | 1486 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1487 | size_t offs) |
1488 | { | |
cf1d6c76 TY |
1489 | void *val = xs->base + offs; |
1490 | struct ocfs2_xattr_value_root *xv = NULL; | |
199799a3 | 1491 | size_t size = namevalue_size_xi(xi); |
cf1d6c76 TY |
1492 | int ret = 0; |
1493 | ||
1494 | memset(val, 0, size); | |
18853b95 | 1495 | memcpy(val, xi->xi_name, xi->xi_name_len); |
cf1d6c76 | 1496 | xv = (struct ocfs2_xattr_value_root *) |
18853b95 | 1497 | (val + OCFS2_XATTR_SIZE(xi->xi_name_len)); |
cf1d6c76 TY |
1498 | xv->xr_clusters = 0; |
1499 | xv->xr_last_eb_blk = 0; | |
1500 | xv->xr_list.l_tree_depth = 0; | |
1501 | xv->xr_list.l_count = cpu_to_le16(1); | |
1502 | xv->xr_list.l_next_free_rec = 0; | |
512620f4 | 1503 | vb->vb_xv = xv; |
cf1d6c76 | 1504 | |
6b240ff6 | 1505 | ret = ocfs2_xattr_value_truncate(inode, vb, xi->xi_value_len, ctxt); |
cf1d6c76 TY |
1506 | if (ret < 0) { |
1507 | mlog_errno(ret); | |
1508 | return ret; | |
1509 | } | |
512620f4 | 1510 | ret = ocfs2_xattr_update_entry(inode, ctxt->handle, xi, xs, vb, offs); |
cf1d6c76 TY |
1511 | if (ret < 0) { |
1512 | mlog_errno(ret); | |
1513 | return ret; | |
1514 | } | |
492a8a33 | 1515 | ret = __ocfs2_xattr_set_value_outside(inode, ctxt->handle, vb, |
6b240ff6 | 1516 | xi->xi_value, xi->xi_value_len); |
cf1d6c76 TY |
1517 | if (ret < 0) |
1518 | mlog_errno(ret); | |
1519 | ||
1520 | return ret; | |
1521 | } | |
1522 | ||
69a3e539 JB |
1523 | static int ocfs2_xa_check_space_helper(int needed_space, int free_start, |
1524 | int num_entries) | |
1525 | { | |
1526 | int free_space; | |
1527 | ||
1528 | if (!needed_space) | |
1529 | return 0; | |
1530 | ||
1531 | free_space = free_start - | |
1532 | sizeof(struct ocfs2_xattr_header) - | |
1533 | (num_entries * sizeof(struct ocfs2_xattr_entry)) - | |
1534 | OCFS2_XATTR_HEADER_GAP; | |
1535 | if (free_space < 0) | |
1536 | return -EIO; | |
1537 | if (free_space < needed_space) | |
1538 | return -ENOSPC; | |
1539 | ||
1540 | return 0; | |
1541 | } | |
1542 | ||
1543 | /* Give a pointer into the storage for the given offset */ | |
1544 | static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset) | |
1545 | { | |
1546 | BUG_ON(offset >= loc->xl_size); | |
1547 | return loc->xl_ops->xlo_offset_pointer(loc, offset); | |
1548 | } | |
1549 | ||
11179f2c JB |
1550 | /* |
1551 | * Wipe the name+value pair and allow the storage to reclaim it. This | |
1552 | * must be followed by either removal of the entry or a call to | |
1553 | * ocfs2_xa_add_namevalue(). | |
1554 | */ | |
1555 | static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1556 | { | |
1557 | loc->xl_ops->xlo_wipe_namevalue(loc); | |
1558 | } | |
1559 | ||
69a3e539 JB |
1560 | /* |
1561 | * Find lowest offset to a name+value pair. This is the start of our | |
1562 | * downward-growing free space. | |
1563 | */ | |
1564 | static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc) | |
1565 | { | |
1566 | return loc->xl_ops->xlo_get_free_start(loc); | |
1567 | } | |
1568 | ||
1569 | /* Can we reuse loc->xl_entry for xi? */ | |
1570 | static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc, | |
1571 | struct ocfs2_xattr_info *xi) | |
1572 | { | |
1573 | return loc->xl_ops->xlo_can_reuse(loc, xi); | |
1574 | } | |
1575 | ||
1576 | /* How much free space is needed to set the new value */ | |
1577 | static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc, | |
1578 | struct ocfs2_xattr_info *xi) | |
1579 | { | |
1580 | return loc->xl_ops->xlo_check_space(loc, xi); | |
1581 | } | |
1582 | ||
1583 | static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) | |
1584 | { | |
1585 | loc->xl_ops->xlo_add_entry(loc, name_hash); | |
1586 | loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash); | |
1587 | /* | |
1588 | * We can't leave the new entry's xe_name_offset at zero or | |
1589 | * add_namevalue() will go nuts. We set it to the size of our | |
1590 | * storage so that it can never be less than any other entry. | |
1591 | */ | |
1592 | loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size); | |
1593 | } | |
1594 | ||
1595 | static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc, | |
1596 | struct ocfs2_xattr_info *xi) | |
1597 | { | |
1598 | int size = namevalue_size_xi(xi); | |
1599 | int nameval_offset; | |
1600 | char *nameval_buf; | |
1601 | ||
1602 | loc->xl_ops->xlo_add_namevalue(loc, size); | |
1603 | loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); | |
1604 | loc->xl_entry->xe_name_len = xi->xi_name_len; | |
1605 | ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index); | |
1606 | ocfs2_xattr_set_local(loc->xl_entry, | |
1607 | xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE); | |
1608 | ||
1609 | nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1610 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); | |
1611 | memset(nameval_buf, 0, size); | |
1612 | memcpy(nameval_buf, xi->xi_name, xi->xi_name_len); | |
1613 | } | |
1614 | ||
3fc12afa JB |
1615 | static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc, |
1616 | struct ocfs2_xattr_value_buf *vb) | |
1617 | { | |
1618 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1619 | int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); | |
1620 | ||
1621 | /* Value bufs are for value trees */ | |
1622 | BUG_ON(namevalue_size_xe(loc->xl_entry) != | |
1623 | (name_size + OCFS2_XATTR_ROOT_SIZE)); | |
1624 | ||
1625 | loc->xl_ops->xlo_fill_value_buf(loc, vb); | |
1626 | vb->vb_xv = | |
1627 | (struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc, | |
1628 | nameval_offset + | |
1629 | name_size); | |
1630 | } | |
1631 | ||
11179f2c JB |
1632 | static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc, |
1633 | int offset) | |
1634 | { | |
11179f2c JB |
1635 | return (char *)loc->xl_header + offset; |
1636 | } | |
1637 | ||
69a3e539 JB |
1638 | static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc, |
1639 | struct ocfs2_xattr_info *xi) | |
1640 | { | |
1641 | /* | |
1642 | * Block storage is strict. If the sizes aren't exact, we will | |
1643 | * remove the old one and reinsert the new. | |
1644 | */ | |
1645 | return namevalue_size_xe(loc->xl_entry) == | |
1646 | namevalue_size_xi(xi); | |
1647 | } | |
1648 | ||
1649 | static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc) | |
1650 | { | |
1651 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1652 | int i, count = le16_to_cpu(xh->xh_count); | |
1653 | int offset, free_start = loc->xl_size; | |
1654 | ||
1655 | for (i = 0; i < count; i++) { | |
1656 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1657 | if (offset < free_start) | |
1658 | free_start = offset; | |
1659 | } | |
1660 | ||
1661 | return free_start; | |
1662 | } | |
1663 | ||
1664 | static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc, | |
1665 | struct ocfs2_xattr_info *xi) | |
1666 | { | |
1667 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1668 | int free_start = ocfs2_xa_get_free_start(loc); | |
1669 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1670 | ||
1671 | /* | |
1672 | * Block storage will reclaim the original entry before inserting | |
1673 | * the new value, so we only need the difference. If the new | |
1674 | * entry is smaller than the old one, we don't need anything. | |
1675 | */ | |
1676 | if (loc->xl_entry) { | |
1677 | /* Don't need space if we're reusing! */ | |
1678 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1679 | needed_space = 0; | |
1680 | else | |
1681 | needed_space -= ocfs2_xe_entry_usage(loc->xl_entry); | |
1682 | } | |
1683 | if (needed_space < 0) | |
1684 | needed_space = 0; | |
1685 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1686 | } | |
1687 | ||
11179f2c JB |
1688 | /* |
1689 | * Block storage for xattrs keeps the name+value pairs compacted. When | |
1690 | * we remove one, we have to shift any that preceded it towards the end. | |
1691 | */ | |
1692 | static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1693 | { | |
1694 | int i, offset; | |
1695 | int namevalue_offset, first_namevalue_offset, namevalue_size; | |
1696 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1697 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
11179f2c JB |
1698 | int count = le16_to_cpu(xh->xh_count); |
1699 | ||
1700 | namevalue_offset = le16_to_cpu(entry->xe_name_offset); | |
199799a3 | 1701 | namevalue_size = namevalue_size_xe(entry); |
69a3e539 | 1702 | first_namevalue_offset = ocfs2_xa_get_free_start(loc); |
11179f2c JB |
1703 | |
1704 | /* Shift the name+value pairs */ | |
1705 | memmove((char *)xh + first_namevalue_offset + namevalue_size, | |
1706 | (char *)xh + first_namevalue_offset, | |
1707 | namevalue_offset - first_namevalue_offset); | |
1708 | memset((char *)xh + first_namevalue_offset, 0, namevalue_size); | |
1709 | ||
1710 | /* Now tell xh->xh_entries about it */ | |
1711 | for (i = 0; i < count; i++) { | |
1712 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1713 | if (offset < namevalue_offset) | |
1714 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, | |
1715 | namevalue_size); | |
1716 | } | |
1717 | ||
1718 | /* | |
1719 | * Note that we don't update xh_free_start or xh_name_value_len | |
1720 | * because they're not used in block-stored xattrs. | |
1721 | */ | |
1722 | } | |
1723 | ||
69a3e539 JB |
1724 | static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1725 | { | |
1726 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1727 | loc->xl_entry = &(loc->xl_header->xh_entries[count]); | |
1728 | le16_add_cpu(&loc->xl_header->xh_count, 1); | |
1729 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1730 | } | |
1731 | ||
1732 | static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1733 | { | |
1734 | int free_start = ocfs2_xa_get_free_start(loc); | |
1735 | ||
1736 | loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size); | |
1737 | } | |
1738 | ||
3fc12afa JB |
1739 | static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc, |
1740 | struct ocfs2_xattr_value_buf *vb) | |
1741 | { | |
1742 | struct buffer_head *bh = loc->xl_storage; | |
1743 | ||
1744 | if (loc->xl_size == (bh->b_size - | |
1745 | offsetof(struct ocfs2_xattr_block, | |
1746 | xb_attrs.xb_header))) | |
1747 | vb->vb_access = ocfs2_journal_access_xb; | |
1748 | else | |
1749 | vb->vb_access = ocfs2_journal_access_di; | |
1750 | vb->vb_bh = bh; | |
1751 | } | |
1752 | ||
11179f2c JB |
1753 | /* |
1754 | * Operations for xattrs stored in blocks. This includes inline inode | |
1755 | * storage and unindexed ocfs2_xattr_blocks. | |
1756 | */ | |
1757 | static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = { | |
1758 | .xlo_offset_pointer = ocfs2_xa_block_offset_pointer, | |
69a3e539 JB |
1759 | .xlo_check_space = ocfs2_xa_block_check_space, |
1760 | .xlo_can_reuse = ocfs2_xa_block_can_reuse, | |
1761 | .xlo_get_free_start = ocfs2_xa_block_get_free_start, | |
11179f2c | 1762 | .xlo_wipe_namevalue = ocfs2_xa_block_wipe_namevalue, |
69a3e539 JB |
1763 | .xlo_add_entry = ocfs2_xa_block_add_entry, |
1764 | .xlo_add_namevalue = ocfs2_xa_block_add_namevalue, | |
3fc12afa | 1765 | .xlo_fill_value_buf = ocfs2_xa_block_fill_value_buf, |
11179f2c JB |
1766 | }; |
1767 | ||
1768 | static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, | |
1769 | int offset) | |
1770 | { | |
1771 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1772 | int block, block_offset; | |
1773 | ||
11179f2c JB |
1774 | /* The header is at the front of the bucket */ |
1775 | block = offset >> bucket->bu_inode->i_sb->s_blocksize_bits; | |
1776 | block_offset = offset % bucket->bu_inode->i_sb->s_blocksize; | |
1777 | ||
1778 | return bucket_block(bucket, block) + block_offset; | |
1779 | } | |
1780 | ||
69a3e539 JB |
1781 | static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc, |
1782 | struct ocfs2_xattr_info *xi) | |
1783 | { | |
1784 | return namevalue_size_xe(loc->xl_entry) >= | |
1785 | namevalue_size_xi(xi); | |
1786 | } | |
1787 | ||
1788 | static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc) | |
1789 | { | |
1790 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1791 | return le16_to_cpu(bucket_xh(bucket)->xh_free_start); | |
1792 | } | |
1793 | ||
1794 | static int ocfs2_bucket_align_free_start(struct super_block *sb, | |
1795 | int free_start, int size) | |
1796 | { | |
1797 | /* | |
1798 | * We need to make sure that the name+value pair fits within | |
1799 | * one block. | |
1800 | */ | |
1801 | if (((free_start - size) >> sb->s_blocksize_bits) != | |
1802 | ((free_start - 1) >> sb->s_blocksize_bits)) | |
1803 | free_start -= free_start % sb->s_blocksize; | |
1804 | ||
1805 | return free_start; | |
1806 | } | |
1807 | ||
1808 | static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc, | |
1809 | struct ocfs2_xattr_info *xi) | |
1810 | { | |
1811 | int rc; | |
1812 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1813 | int free_start = ocfs2_xa_get_free_start(loc); | |
1814 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1815 | int size = namevalue_size_xi(xi); | |
1816 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1817 | struct super_block *sb = bucket->bu_inode->i_sb; | |
1818 | ||
1819 | /* | |
1820 | * Bucket storage does not reclaim name+value pairs it cannot | |
1821 | * reuse. They live as holes until the bucket fills, and then | |
1822 | * the bucket is defragmented. However, the bucket can reclaim | |
1823 | * the ocfs2_xattr_entry. | |
1824 | */ | |
1825 | if (loc->xl_entry) { | |
1826 | /* Don't need space if we're reusing! */ | |
1827 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1828 | needed_space = 0; | |
1829 | else | |
1830 | needed_space -= sizeof(struct ocfs2_xattr_entry); | |
1831 | } | |
1832 | BUG_ON(needed_space < 0); | |
1833 | ||
1834 | if (free_start < size) { | |
1835 | if (needed_space) | |
1836 | return -ENOSPC; | |
1837 | } else { | |
1838 | /* | |
1839 | * First we check if it would fit in the first place. | |
1840 | * Below, we align the free start to a block. This may | |
1841 | * slide us below the minimum gap. By checking unaligned | |
1842 | * first, we avoid that error. | |
1843 | */ | |
1844 | rc = ocfs2_xa_check_space_helper(needed_space, free_start, | |
1845 | count); | |
1846 | if (rc) | |
1847 | return rc; | |
1848 | free_start = ocfs2_bucket_align_free_start(sb, free_start, | |
1849 | size); | |
1850 | } | |
1851 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1852 | } | |
1853 | ||
11179f2c JB |
1854 | static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc) |
1855 | { | |
199799a3 JB |
1856 | le16_add_cpu(&loc->xl_header->xh_name_value_len, |
1857 | -namevalue_size_xe(loc->xl_entry)); | |
11179f2c JB |
1858 | } |
1859 | ||
69a3e539 JB |
1860 | static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1861 | { | |
1862 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1863 | int count = le16_to_cpu(xh->xh_count); | |
1864 | int low = 0, high = count - 1, tmp; | |
1865 | struct ocfs2_xattr_entry *tmp_xe; | |
1866 | ||
1867 | /* | |
1868 | * We keep buckets sorted by name_hash, so we need to find | |
1869 | * our insert place. | |
1870 | */ | |
1871 | while (low <= high && count) { | |
1872 | tmp = (low + high) / 2; | |
1873 | tmp_xe = &xh->xh_entries[tmp]; | |
1874 | ||
1875 | if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) | |
1876 | low = tmp + 1; | |
1877 | else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash)) | |
1878 | high = tmp - 1; | |
1879 | else { | |
1880 | low = tmp; | |
1881 | break; | |
1882 | } | |
1883 | } | |
1884 | ||
1885 | if (low != count) | |
1886 | memmove(&xh->xh_entries[low + 1], | |
1887 | &xh->xh_entries[low], | |
1888 | ((count - low) * sizeof(struct ocfs2_xattr_entry))); | |
1889 | ||
1890 | le16_add_cpu(&xh->xh_count, 1); | |
1891 | loc->xl_entry = &xh->xh_entries[low]; | |
1892 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1893 | } | |
1894 | ||
1895 | static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1896 | { | |
1897 | int free_start = ocfs2_xa_get_free_start(loc); | |
1898 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1899 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1900 | struct super_block *sb = bucket->bu_inode->i_sb; | |
1901 | int nameval_offset; | |
1902 | ||
1903 | free_start = ocfs2_bucket_align_free_start(sb, free_start, size); | |
1904 | nameval_offset = free_start - size; | |
1905 | loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset); | |
1906 | xh->xh_free_start = cpu_to_le16(nameval_offset); | |
1907 | le16_add_cpu(&xh->xh_name_value_len, size); | |
1908 | ||
1909 | } | |
1910 | ||
3fc12afa JB |
1911 | static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc, |
1912 | struct ocfs2_xattr_value_buf *vb) | |
1913 | { | |
1914 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1915 | struct super_block *sb = bucket->bu_inode->i_sb; | |
1916 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1917 | int size = namevalue_size_xe(loc->xl_entry); | |
1918 | int block_offset = nameval_offset >> sb->s_blocksize_bits; | |
1919 | ||
1920 | /* Values are not allowed to straddle block boundaries */ | |
1921 | BUG_ON(block_offset != | |
1922 | ((nameval_offset + size - 1) >> sb->s_blocksize_bits)); | |
1923 | /* We expect the bucket to be filled in */ | |
1924 | BUG_ON(!bucket->bu_bhs[block_offset]); | |
1925 | ||
1926 | vb->vb_access = ocfs2_journal_access; | |
1927 | vb->vb_bh = bucket->bu_bhs[block_offset]; | |
1928 | } | |
1929 | ||
11179f2c JB |
1930 | /* Operations for xattrs stored in buckets. */ |
1931 | static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = { | |
1932 | .xlo_offset_pointer = ocfs2_xa_bucket_offset_pointer, | |
69a3e539 JB |
1933 | .xlo_check_space = ocfs2_xa_bucket_check_space, |
1934 | .xlo_can_reuse = ocfs2_xa_bucket_can_reuse, | |
1935 | .xlo_get_free_start = ocfs2_xa_bucket_get_free_start, | |
11179f2c | 1936 | .xlo_wipe_namevalue = ocfs2_xa_bucket_wipe_namevalue, |
69a3e539 JB |
1937 | .xlo_add_entry = ocfs2_xa_bucket_add_entry, |
1938 | .xlo_add_namevalue = ocfs2_xa_bucket_add_namevalue, | |
3fc12afa | 1939 | .xlo_fill_value_buf = ocfs2_xa_bucket_fill_value_buf, |
11179f2c JB |
1940 | }; |
1941 | ||
1942 | static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc) | |
1943 | { | |
bde1e540 JB |
1944 | int index, count; |
1945 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1946 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1947 | ||
11179f2c | 1948 | ocfs2_xa_wipe_namevalue(loc); |
bde1e540 JB |
1949 | loc->xl_entry = NULL; |
1950 | ||
1951 | le16_add_cpu(&xh->xh_count, -1); | |
1952 | count = le16_to_cpu(xh->xh_count); | |
1953 | ||
1954 | /* | |
1955 | * Only zero out the entry if there are more remaining. This is | |
1956 | * important for an empty bucket, as it keeps track of the | |
1957 | * bucket's hash value. It doesn't hurt empty block storage. | |
1958 | */ | |
1959 | if (count) { | |
1960 | index = ((char *)entry - (char *)&xh->xh_entries) / | |
1961 | sizeof(struct ocfs2_xattr_entry); | |
1962 | memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1], | |
1963 | (count - index) * sizeof(struct ocfs2_xattr_entry)); | |
1964 | memset(&xh->xh_entries[count], 0, | |
1965 | sizeof(struct ocfs2_xattr_entry)); | |
1966 | } | |
11179f2c JB |
1967 | } |
1968 | ||
69a3e539 JB |
1969 | /* |
1970 | * Prepares loc->xl_entry to receive the new xattr. This includes | |
1971 | * properly setting up the name+value pair region. If loc->xl_entry | |
1972 | * already exists, it will take care of modifying it appropriately. | |
1973 | * This also includes deleting entries, but don't call this to remove | |
1974 | * a non-existant entry. That's just a bug. | |
1975 | * | |
1976 | * Note that this modifies the data. You did journal_access already, | |
1977 | * right? | |
1978 | */ | |
1979 | static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc, | |
1980 | struct ocfs2_xattr_info *xi, | |
1981 | u32 name_hash) | |
1982 | { | |
1983 | int rc = 0; | |
1984 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
1985 | char *nameval_buf; | |
1986 | ||
1987 | if (!xi->xi_value) { | |
1988 | ocfs2_xa_remove_entry(loc); | |
1989 | goto out; | |
1990 | } | |
1991 | ||
1992 | rc = ocfs2_xa_check_space(loc, xi); | |
1993 | if (rc) | |
1994 | goto out; | |
1995 | ||
1996 | if (loc->xl_entry) { | |
1997 | if (ocfs2_xa_can_reuse_entry(loc, xi)) { | |
1998 | nameval_buf = ocfs2_xa_offset_pointer(loc, | |
1999 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
2000 | memset(nameval_buf + name_size, 0, | |
2001 | namevalue_size_xe(loc->xl_entry) - name_size); | |
2002 | loc->xl_entry->xe_value_size = | |
2003 | cpu_to_le64(xi->xi_value_len); | |
2004 | goto out; | |
2005 | } | |
2006 | ||
2007 | ocfs2_xa_wipe_namevalue(loc); | |
2008 | } else | |
2009 | ocfs2_xa_add_entry(loc, name_hash); | |
2010 | ||
2011 | /* | |
2012 | * If we get here, we have a blank entry. Fill it. We grow our | |
2013 | * name+value pair back from the end. | |
2014 | */ | |
2015 | ocfs2_xa_add_namevalue(loc, xi); | |
2016 | ||
2017 | out: | |
2018 | return rc; | |
2019 | } | |
2020 | ||
2021 | /* | |
2022 | * Store the value portion of the name+value pair. This is either an | |
2023 | * inline value or the tree root of an external value. | |
2024 | */ | |
2025 | static void ocfs2_xa_store_inline_value(struct ocfs2_xa_loc *loc, | |
2026 | struct ocfs2_xattr_info *xi) | |
2027 | { | |
2028 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
2029 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
9dc47400 JB |
2030 | int inline_value_size = namevalue_size_xi(xi) - name_size; |
2031 | const void *value = xi->xi_value; | |
69a3e539 JB |
2032 | char *nameval_buf; |
2033 | ||
2034 | if (!xi->xi_value) | |
2035 | return; | |
2036 | ||
9dc47400 JB |
2037 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
2038 | value = &def_xv; | |
2039 | inline_value_size = OCFS2_XATTR_ROOT_SIZE; | |
2040 | } | |
69a3e539 | 2041 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); |
9dc47400 | 2042 | memcpy(nameval_buf + name_size, value, inline_value_size); |
69a3e539 JB |
2043 | } |
2044 | ||
11179f2c JB |
2045 | static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc, |
2046 | struct inode *inode, | |
2047 | struct buffer_head *bh, | |
2048 | struct ocfs2_xattr_entry *entry) | |
2049 | { | |
2050 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
2051 | ||
2052 | loc->xl_ops = &ocfs2_xa_block_loc_ops; | |
2053 | loc->xl_storage = bh; | |
2054 | loc->xl_entry = entry; | |
2055 | ||
2056 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
2057 | loc->xl_size = le16_to_cpu(di->i_xattr_inline_size); | |
2058 | else { | |
2059 | BUG_ON(entry); | |
2060 | loc->xl_size = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
2061 | } | |
2062 | loc->xl_header = | |
2063 | (struct ocfs2_xattr_header *)(bh->b_data + bh->b_size - | |
2064 | loc->xl_size); | |
2065 | } | |
2066 | ||
2067 | static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc, | |
2068 | struct buffer_head *bh, | |
2069 | struct ocfs2_xattr_entry *entry) | |
2070 | { | |
2071 | struct ocfs2_xattr_block *xb = | |
2072 | (struct ocfs2_xattr_block *)bh->b_data; | |
2073 | ||
2074 | BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED); | |
2075 | ||
2076 | loc->xl_ops = &ocfs2_xa_block_loc_ops; | |
2077 | loc->xl_storage = bh; | |
2078 | loc->xl_header = &(xb->xb_attrs.xb_header); | |
2079 | loc->xl_entry = entry; | |
2080 | loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block, | |
2081 | xb_attrs.xb_header); | |
2082 | } | |
2083 | ||
2084 | static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc, | |
2085 | struct ocfs2_xattr_bucket *bucket, | |
2086 | struct ocfs2_xattr_entry *entry) | |
2087 | { | |
2088 | loc->xl_ops = &ocfs2_xa_bucket_loc_ops; | |
2089 | loc->xl_storage = bucket; | |
2090 | loc->xl_header = bucket_xh(bucket); | |
2091 | loc->xl_entry = entry; | |
2092 | loc->xl_size = OCFS2_XATTR_BUCKET_SIZE; | |
2093 | } | |
2094 | ||
cf1d6c76 TY |
2095 | |
2096 | /* | |
2097 | * ocfs2_xattr_set_entry() | |
2098 | * | |
2099 | * Set extended attribute entry into inode or block. | |
2100 | * | |
2101 | * If extended attribute value size > OCFS2_XATTR_INLINE_SIZE, | |
69a3e539 | 2102 | * We first insert tree root(ocfs2_xattr_value_root) like a normal value, |
cf1d6c76 TY |
2103 | * then set value in B tree with set_value_outside(). |
2104 | */ | |
2105 | static int ocfs2_xattr_set_entry(struct inode *inode, | |
2106 | struct ocfs2_xattr_info *xi, | |
2107 | struct ocfs2_xattr_search *xs, | |
78f30c31 | 2108 | struct ocfs2_xattr_set_ctxt *ctxt, |
cf1d6c76 TY |
2109 | int flag) |
2110 | { | |
2111 | struct ocfs2_xattr_entry *last; | |
2112 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2113 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
18853b95 | 2114 | size_t min_offs = xs->end - xs->base; |
cf1d6c76 | 2115 | size_t size_l = 0; |
85db90e7 | 2116 | handle_t *handle = ctxt->handle; |
cf1d6c76 | 2117 | int free, i, ret; |
69a3e539 JB |
2118 | u32 name_hash = ocfs2_xattr_name_hash(inode, xi->xi_name, |
2119 | xi->xi_name_len); | |
2120 | struct ocfs2_xa_loc loc; | |
512620f4 JB |
2121 | struct ocfs2_xattr_value_buf vb = { |
2122 | .vb_bh = xs->xattr_bh, | |
2123 | .vb_access = ocfs2_journal_access_di, | |
2124 | }; | |
2125 | ||
2126 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
2127 | BUG_ON(xs->xattr_bh == xs->inode_bh); | |
2128 | vb.vb_access = ocfs2_journal_access_xb; | |
2129 | } else | |
2130 | BUG_ON(xs->xattr_bh != xs->inode_bh); | |
cf1d6c76 TY |
2131 | |
2132 | /* Compute min_offs, last and free space. */ | |
2133 | last = xs->header->xh_entries; | |
2134 | ||
2135 | for (i = 0 ; i < le16_to_cpu(xs->header->xh_count); i++) { | |
2136 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
2137 | if (offs < min_offs) | |
2138 | min_offs = offs; | |
2139 | last += 1; | |
2140 | } | |
2141 | ||
4442f518 | 2142 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
cf1d6c76 | 2143 | if (free < 0) |
b37c4d84 | 2144 | return -EIO; |
cf1d6c76 | 2145 | |
199799a3 JB |
2146 | if (!xs->not_found) |
2147 | free += ocfs2_xe_entry_usage(xs->here); | |
2148 | ||
cf1d6c76 | 2149 | /* Check free space in inode or block */ |
9dc47400 JB |
2150 | if (xi->xi_value && (free < ocfs2_xi_entry_usage(xi))) { |
2151 | ret = -ENOSPC; | |
2152 | goto out; | |
cf1d6c76 TY |
2153 | } |
2154 | ||
2155 | if (!xs->not_found) { | |
2156 | /* For existing extended attribute */ | |
199799a3 | 2157 | size_t size = namevalue_size_xe(xs->here); |
cf1d6c76 TY |
2158 | size_t offs = le16_to_cpu(xs->here->xe_name_offset); |
2159 | void *val = xs->base + offs; | |
2160 | ||
2161 | if (ocfs2_xattr_is_local(xs->here) && size == size_l) { | |
2162 | /* Replace existing local xattr with tree root */ | |
2163 | ret = ocfs2_xattr_set_value_outside(inode, xi, xs, | |
512620f4 | 2164 | ctxt, &vb, offs); |
cf1d6c76 TY |
2165 | if (ret < 0) |
2166 | mlog_errno(ret); | |
2167 | goto out; | |
2168 | } else if (!ocfs2_xattr_is_local(xs->here)) { | |
2169 | /* For existing xattr which has value outside */ | |
512620f4 | 2170 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
18853b95 | 2171 | (val + OCFS2_XATTR_SIZE(xi->xi_name_len)); |
cf1d6c76 | 2172 | |
6b240ff6 | 2173 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
cf1d6c76 TY |
2174 | /* |
2175 | * If new value need set outside also, | |
2176 | * first truncate old value to new value, | |
2177 | * then set new value with set_value_outside(). | |
2178 | */ | |
2179 | ret = ocfs2_xattr_value_truncate(inode, | |
6b240ff6 JB |
2180 | &vb, |
2181 | xi->xi_value_len, | |
2182 | ctxt); | |
cf1d6c76 TY |
2183 | if (ret < 0) { |
2184 | mlog_errno(ret); | |
2185 | goto out; | |
2186 | } | |
2187 | ||
85db90e7 TM |
2188 | ret = ocfs2_xattr_update_entry(inode, |
2189 | handle, | |
2190 | xi, | |
2191 | xs, | |
0c748e95 | 2192 | &vb, |
85db90e7 | 2193 | offs); |
cf1d6c76 TY |
2194 | if (ret < 0) { |
2195 | mlog_errno(ret); | |
2196 | goto out; | |
2197 | } | |
2198 | ||
85db90e7 | 2199 | ret = __ocfs2_xattr_set_value_outside(inode, |
6b240ff6 JB |
2200 | handle, |
2201 | &vb, | |
2202 | xi->xi_value, | |
2203 | xi->xi_value_len); | |
cf1d6c76 TY |
2204 | if (ret < 0) |
2205 | mlog_errno(ret); | |
2206 | goto out; | |
2207 | } else { | |
2208 | /* | |
2209 | * If new value need set in local, | |
2210 | * just trucate old value to zero. | |
2211 | */ | |
2212 | ret = ocfs2_xattr_value_truncate(inode, | |
b3e5d379 | 2213 | &vb, |
85db90e7 TM |
2214 | 0, |
2215 | ctxt); | |
cf1d6c76 TY |
2216 | if (ret < 0) |
2217 | mlog_errno(ret); | |
2218 | } | |
2219 | } | |
2220 | } | |
2221 | ||
0cf2f763 | 2222 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), xs->inode_bh, |
512620f4 | 2223 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2224 | if (ret) { |
2225 | mlog_errno(ret); | |
85db90e7 | 2226 | goto out; |
cf1d6c76 TY |
2227 | } |
2228 | ||
2229 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
0cf2f763 | 2230 | ret = vb.vb_access(handle, INODE_CACHE(inode), vb.vb_bh, |
512620f4 | 2231 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2232 | if (ret) { |
2233 | mlog_errno(ret); | |
85db90e7 | 2234 | goto out; |
cf1d6c76 TY |
2235 | } |
2236 | } | |
2237 | ||
69a3e539 JB |
2238 | if (xs->xattr_bh == xs->inode_bh) |
2239 | ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh, | |
2240 | xs->not_found ? NULL : xs->here); | |
2241 | else | |
2242 | ocfs2_init_xattr_block_xa_loc(&loc, xs->xattr_bh, | |
2243 | xs->not_found ? NULL : xs->here); | |
2244 | ||
cf1d6c76 | 2245 | /* |
69a3e539 JB |
2246 | * Prepare our entry and insert the inline value. This will |
2247 | * be a value tree root for values that are larger than | |
2248 | * OCFS2_XATTR_INLINE_SIZE. | |
cf1d6c76 | 2249 | */ |
69a3e539 JB |
2250 | ret = ocfs2_xa_prepare_entry(&loc, xi, name_hash); |
2251 | if (ret) { | |
2252 | if (ret != -ENOSPC) | |
2253 | mlog_errno(ret); | |
2254 | goto out; | |
2255 | } | |
2256 | /* XXX For now, until we make ocfs2_xa_prepare_entry() primary */ | |
2257 | BUG_ON(ret == -ENOSPC); | |
2258 | ocfs2_xa_store_inline_value(&loc, xi); | |
2259 | xs->here = loc.xl_entry; | |
cf1d6c76 TY |
2260 | |
2261 | if (!(flag & OCFS2_INLINE_XATTR_FL)) { | |
2262 | ret = ocfs2_journal_dirty(handle, xs->xattr_bh); | |
2263 | if (ret < 0) { | |
2264 | mlog_errno(ret); | |
85db90e7 | 2265 | goto out; |
cf1d6c76 TY |
2266 | } |
2267 | } | |
2268 | ||
2269 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) && | |
2270 | (flag & OCFS2_INLINE_XATTR_FL)) { | |
2271 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2272 | unsigned int xattrsize = osb->s_xattr_inline_size; | |
2273 | ||
2274 | /* | |
2275 | * Adjust extent record count or inline data size | |
2276 | * to reserve space for extended attribute. | |
2277 | */ | |
2278 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2279 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2280 | le16_add_cpu(&idata->id_count, -xattrsize); | |
2281 | } else if (!(ocfs2_inode_is_fast_symlink(inode))) { | |
2282 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2283 | le16_add_cpu(&el->l_count, -(xattrsize / | |
2284 | sizeof(struct ocfs2_extent_rec))); | |
2285 | } | |
2286 | di->i_xattr_inline_size = cpu_to_le16(xattrsize); | |
2287 | } | |
2288 | /* Update xattr flag */ | |
2289 | spin_lock(&oi->ip_lock); | |
2290 | oi->ip_dyn_features |= flag; | |
2291 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2292 | spin_unlock(&oi->ip_lock); | |
cf1d6c76 TY |
2293 | |
2294 | ret = ocfs2_journal_dirty(handle, xs->inode_bh); | |
2295 | if (ret < 0) | |
2296 | mlog_errno(ret); | |
2297 | ||
6b240ff6 | 2298 | if (!ret && xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
cf1d6c76 TY |
2299 | /* |
2300 | * Set value outside in B tree. | |
2301 | * This is the second step for value size > INLINE_SIZE. | |
2302 | */ | |
2303 | size_t offs = le16_to_cpu(xs->here->xe_name_offset); | |
512620f4 JB |
2304 | ret = ocfs2_xattr_set_value_outside(inode, xi, xs, ctxt, |
2305 | &vb, offs); | |
cf1d6c76 TY |
2306 | if (ret < 0) { |
2307 | int ret2; | |
2308 | ||
2309 | mlog_errno(ret); | |
2310 | /* | |
2311 | * If set value outside failed, we have to clean | |
2312 | * the junk tree root we have already set in local. | |
2313 | */ | |
85db90e7 | 2314 | ret2 = ocfs2_xattr_cleanup(inode, ctxt->handle, |
512620f4 | 2315 | xi, xs, &vb, offs); |
cf1d6c76 TY |
2316 | if (ret2 < 0) |
2317 | mlog_errno(ret2); | |
2318 | } | |
2319 | } | |
2320 | out: | |
2321 | return ret; | |
cf1d6c76 TY |
2322 | } |
2323 | ||
ce9c5a54 TM |
2324 | /* |
2325 | * In xattr remove, if it is stored outside and refcounted, we may have | |
2326 | * the chance to split the refcount tree. So need the allocators. | |
2327 | */ | |
2328 | static int ocfs2_lock_xattr_remove_allocators(struct inode *inode, | |
2329 | struct ocfs2_xattr_value_root *xv, | |
2330 | struct ocfs2_caching_info *ref_ci, | |
2331 | struct buffer_head *ref_root_bh, | |
2332 | struct ocfs2_alloc_context **meta_ac, | |
2333 | int *ref_credits) | |
2334 | { | |
2335 | int ret, meta_add = 0; | |
2336 | u32 p_cluster, num_clusters; | |
2337 | unsigned int ext_flags; | |
2338 | ||
2339 | *ref_credits = 0; | |
2340 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
2341 | &num_clusters, | |
2342 | &xv->xr_list, | |
2343 | &ext_flags); | |
2344 | if (ret) { | |
2345 | mlog_errno(ret); | |
2346 | goto out; | |
2347 | } | |
2348 | ||
2349 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
2350 | goto out; | |
2351 | ||
2352 | ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci, | |
2353 | ref_root_bh, xv, | |
2354 | &meta_add, ref_credits); | |
2355 | if (ret) { | |
2356 | mlog_errno(ret); | |
2357 | goto out; | |
2358 | } | |
2359 | ||
2360 | ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), | |
2361 | meta_add, meta_ac); | |
2362 | if (ret) | |
2363 | mlog_errno(ret); | |
2364 | ||
2365 | out: | |
2366 | return ret; | |
2367 | } | |
2368 | ||
cf1d6c76 | 2369 | static int ocfs2_remove_value_outside(struct inode*inode, |
4311901d | 2370 | struct ocfs2_xattr_value_buf *vb, |
ce9c5a54 TM |
2371 | struct ocfs2_xattr_header *header, |
2372 | struct ocfs2_caching_info *ref_ci, | |
2373 | struct buffer_head *ref_root_bh) | |
cf1d6c76 | 2374 | { |
ce9c5a54 | 2375 | int ret = 0, i, ref_credits; |
78f30c31 TM |
2376 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
2377 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; | |
ce9c5a54 | 2378 | void *val; |
78f30c31 TM |
2379 | |
2380 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
cf1d6c76 TY |
2381 | |
2382 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
2383 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
2384 | ||
ce9c5a54 TM |
2385 | if (ocfs2_xattr_is_local(entry)) |
2386 | continue; | |
cf1d6c76 | 2387 | |
ce9c5a54 TM |
2388 | val = (void *)header + |
2389 | le16_to_cpu(entry->xe_name_offset); | |
2390 | vb->vb_xv = (struct ocfs2_xattr_value_root *) | |
2391 | (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); | |
2392 | ||
2393 | ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv, | |
2394 | ref_ci, ref_root_bh, | |
2395 | &ctxt.meta_ac, | |
2396 | &ref_credits); | |
2397 | ||
2398 | ctxt.handle = ocfs2_start_trans(osb, ref_credits + | |
2399 | ocfs2_remove_extent_credits(osb->sb)); | |
2400 | if (IS_ERR(ctxt.handle)) { | |
2401 | ret = PTR_ERR(ctxt.handle); | |
2402 | mlog_errno(ret); | |
2403 | break; | |
2404 | } | |
2405 | ||
2406 | ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); | |
2407 | if (ret < 0) { | |
2408 | mlog_errno(ret); | |
2409 | break; | |
2410 | } | |
2411 | ||
2412 | ocfs2_commit_trans(osb, ctxt.handle); | |
2413 | if (ctxt.meta_ac) { | |
2414 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
2415 | ctxt.meta_ac = NULL; | |
cf1d6c76 TY |
2416 | } |
2417 | } | |
2418 | ||
ce9c5a54 TM |
2419 | if (ctxt.meta_ac) |
2420 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
2421 | ocfs2_schedule_truncate_log_flush(osb, 1); |
2422 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
cf1d6c76 TY |
2423 | return ret; |
2424 | } | |
2425 | ||
2426 | static int ocfs2_xattr_ibody_remove(struct inode *inode, | |
ce9c5a54 TM |
2427 | struct buffer_head *di_bh, |
2428 | struct ocfs2_caching_info *ref_ci, | |
2429 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2430 | { |
2431 | ||
2432 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2433 | struct ocfs2_xattr_header *header; | |
2434 | int ret; | |
4311901d JB |
2435 | struct ocfs2_xattr_value_buf vb = { |
2436 | .vb_bh = di_bh, | |
2437 | .vb_access = ocfs2_journal_access_di, | |
2438 | }; | |
cf1d6c76 TY |
2439 | |
2440 | header = (struct ocfs2_xattr_header *) | |
2441 | ((void *)di + inode->i_sb->s_blocksize - | |
2442 | le16_to_cpu(di->i_xattr_inline_size)); | |
2443 | ||
ce9c5a54 TM |
2444 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2445 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2446 | |
2447 | return ret; | |
2448 | } | |
2449 | ||
ce9c5a54 TM |
2450 | struct ocfs2_rm_xattr_bucket_para { |
2451 | struct ocfs2_caching_info *ref_ci; | |
2452 | struct buffer_head *ref_root_bh; | |
2453 | }; | |
2454 | ||
cf1d6c76 | 2455 | static int ocfs2_xattr_block_remove(struct inode *inode, |
ce9c5a54 TM |
2456 | struct buffer_head *blk_bh, |
2457 | struct ocfs2_caching_info *ref_ci, | |
2458 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2459 | { |
2460 | struct ocfs2_xattr_block *xb; | |
cf1d6c76 | 2461 | int ret = 0; |
4311901d JB |
2462 | struct ocfs2_xattr_value_buf vb = { |
2463 | .vb_bh = blk_bh, | |
2464 | .vb_access = ocfs2_journal_access_xb, | |
2465 | }; | |
ce9c5a54 TM |
2466 | struct ocfs2_rm_xattr_bucket_para args = { |
2467 | .ref_ci = ref_ci, | |
2468 | .ref_root_bh = ref_root_bh, | |
2469 | }; | |
cf1d6c76 TY |
2470 | |
2471 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; | |
a3944256 TM |
2472 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
2473 | struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header); | |
ce9c5a54 TM |
2474 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2475 | ref_ci, ref_root_bh); | |
a3944256 | 2476 | } else |
47bca495 TM |
2477 | ret = ocfs2_iterate_xattr_index_block(inode, |
2478 | blk_bh, | |
2479 | ocfs2_rm_xattr_cluster, | |
ce9c5a54 | 2480 | &args); |
cf1d6c76 TY |
2481 | |
2482 | return ret; | |
2483 | } | |
2484 | ||
08413899 | 2485 | static int ocfs2_xattr_free_block(struct inode *inode, |
ce9c5a54 TM |
2486 | u64 block, |
2487 | struct ocfs2_caching_info *ref_ci, | |
2488 | struct buffer_head *ref_root_bh) | |
08413899 TM |
2489 | { |
2490 | struct inode *xb_alloc_inode; | |
2491 | struct buffer_head *xb_alloc_bh = NULL; | |
2492 | struct buffer_head *blk_bh = NULL; | |
2493 | struct ocfs2_xattr_block *xb; | |
2494 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2495 | handle_t *handle; | |
2496 | int ret = 0; | |
2497 | u64 blk, bg_blkno; | |
2498 | u16 bit; | |
2499 | ||
4ae1d69b | 2500 | ret = ocfs2_read_xattr_block(inode, block, &blk_bh); |
08413899 TM |
2501 | if (ret < 0) { |
2502 | mlog_errno(ret); | |
2503 | goto out; | |
2504 | } | |
2505 | ||
ce9c5a54 | 2506 | ret = ocfs2_xattr_block_remove(inode, blk_bh, ref_ci, ref_root_bh); |
08413899 TM |
2507 | if (ret < 0) { |
2508 | mlog_errno(ret); | |
2509 | goto out; | |
2510 | } | |
2511 | ||
4ae1d69b | 2512 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
08413899 TM |
2513 | blk = le64_to_cpu(xb->xb_blkno); |
2514 | bit = le16_to_cpu(xb->xb_suballoc_bit); | |
2515 | bg_blkno = ocfs2_which_suballoc_group(blk, bit); | |
2516 | ||
2517 | xb_alloc_inode = ocfs2_get_system_file_inode(osb, | |
2518 | EXTENT_ALLOC_SYSTEM_INODE, | |
2519 | le16_to_cpu(xb->xb_suballoc_slot)); | |
2520 | if (!xb_alloc_inode) { | |
2521 | ret = -ENOMEM; | |
2522 | mlog_errno(ret); | |
2523 | goto out; | |
2524 | } | |
2525 | mutex_lock(&xb_alloc_inode->i_mutex); | |
2526 | ||
2527 | ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1); | |
2528 | if (ret < 0) { | |
2529 | mlog_errno(ret); | |
2530 | goto out_mutex; | |
2531 | } | |
2532 | ||
2533 | handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE); | |
2534 | if (IS_ERR(handle)) { | |
2535 | ret = PTR_ERR(handle); | |
2536 | mlog_errno(ret); | |
2537 | goto out_unlock; | |
2538 | } | |
2539 | ||
2540 | ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh, | |
2541 | bit, bg_blkno, 1); | |
2542 | if (ret < 0) | |
2543 | mlog_errno(ret); | |
2544 | ||
2545 | ocfs2_commit_trans(osb, handle); | |
2546 | out_unlock: | |
2547 | ocfs2_inode_unlock(xb_alloc_inode, 1); | |
2548 | brelse(xb_alloc_bh); | |
2549 | out_mutex: | |
2550 | mutex_unlock(&xb_alloc_inode->i_mutex); | |
2551 | iput(xb_alloc_inode); | |
2552 | out: | |
2553 | brelse(blk_bh); | |
2554 | return ret; | |
2555 | } | |
2556 | ||
cf1d6c76 TY |
2557 | /* |
2558 | * ocfs2_xattr_remove() | |
2559 | * | |
2560 | * Free extended attribute resources associated with this inode. | |
2561 | */ | |
2562 | int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) | |
2563 | { | |
cf1d6c76 TY |
2564 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
2565 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
ce9c5a54 TM |
2566 | struct ocfs2_refcount_tree *ref_tree = NULL; |
2567 | struct buffer_head *ref_root_bh = NULL; | |
2568 | struct ocfs2_caching_info *ref_ci = NULL; | |
cf1d6c76 TY |
2569 | handle_t *handle; |
2570 | int ret; | |
2571 | ||
8154da3d TY |
2572 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
2573 | return 0; | |
2574 | ||
cf1d6c76 TY |
2575 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
2576 | return 0; | |
2577 | ||
ce9c5a54 TM |
2578 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) { |
2579 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(inode->i_sb), | |
2580 | le64_to_cpu(di->i_refcount_loc), | |
2581 | 1, &ref_tree, &ref_root_bh); | |
2582 | if (ret) { | |
2583 | mlog_errno(ret); | |
2584 | goto out; | |
2585 | } | |
2586 | ref_ci = &ref_tree->rf_ci; | |
2587 | ||
2588 | } | |
2589 | ||
cf1d6c76 | 2590 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { |
ce9c5a54 TM |
2591 | ret = ocfs2_xattr_ibody_remove(inode, di_bh, |
2592 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2593 | if (ret < 0) { |
2594 | mlog_errno(ret); | |
2595 | goto out; | |
2596 | } | |
2597 | } | |
cf1d6c76 | 2598 | |
08413899 TM |
2599 | if (di->i_xattr_loc) { |
2600 | ret = ocfs2_xattr_free_block(inode, | |
ce9c5a54 TM |
2601 | le64_to_cpu(di->i_xattr_loc), |
2602 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2603 | if (ret < 0) { |
2604 | mlog_errno(ret); | |
2605 | goto out; | |
2606 | } | |
2607 | } | |
2608 | ||
2609 | handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)), | |
2610 | OCFS2_INODE_UPDATE_CREDITS); | |
2611 | if (IS_ERR(handle)) { | |
2612 | ret = PTR_ERR(handle); | |
2613 | mlog_errno(ret); | |
2614 | goto out; | |
2615 | } | |
0cf2f763 | 2616 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, |
84008972 | 2617 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2618 | if (ret) { |
2619 | mlog_errno(ret); | |
2620 | goto out_commit; | |
2621 | } | |
2622 | ||
08413899 | 2623 | di->i_xattr_loc = 0; |
cf1d6c76 TY |
2624 | |
2625 | spin_lock(&oi->ip_lock); | |
2626 | oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL); | |
2627 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2628 | spin_unlock(&oi->ip_lock); | |
2629 | ||
2630 | ret = ocfs2_journal_dirty(handle, di_bh); | |
2631 | if (ret < 0) | |
2632 | mlog_errno(ret); | |
2633 | out_commit: | |
2634 | ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); | |
2635 | out: | |
ce9c5a54 TM |
2636 | if (ref_tree) |
2637 | ocfs2_unlock_refcount_tree(OCFS2_SB(inode->i_sb), ref_tree, 1); | |
2638 | brelse(ref_root_bh); | |
cf1d6c76 TY |
2639 | return ret; |
2640 | } | |
2641 | ||
2642 | static int ocfs2_xattr_has_space_inline(struct inode *inode, | |
2643 | struct ocfs2_dinode *di) | |
2644 | { | |
2645 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2646 | unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
2647 | int free; | |
2648 | ||
2649 | if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE) | |
2650 | return 0; | |
2651 | ||
2652 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2653 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2654 | free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size); | |
2655 | } else if (ocfs2_inode_is_fast_symlink(inode)) { | |
2656 | free = ocfs2_fast_symlink_chars(inode->i_sb) - | |
2657 | le64_to_cpu(di->i_size); | |
2658 | } else { | |
2659 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2660 | free = (le16_to_cpu(el->l_count) - | |
2661 | le16_to_cpu(el->l_next_free_rec)) * | |
2662 | sizeof(struct ocfs2_extent_rec); | |
2663 | } | |
2664 | if (free >= xattrsize) | |
2665 | return 1; | |
2666 | ||
2667 | return 0; | |
2668 | } | |
2669 | ||
2670 | /* | |
2671 | * ocfs2_xattr_ibody_find() | |
2672 | * | |
2673 | * Find extended attribute in inode block and | |
2674 | * fill search info into struct ocfs2_xattr_search. | |
2675 | */ | |
2676 | static int ocfs2_xattr_ibody_find(struct inode *inode, | |
2677 | int name_index, | |
2678 | const char *name, | |
2679 | struct ocfs2_xattr_search *xs) | |
2680 | { | |
2681 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2682 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2683 | int ret; | |
2684 | int has_space = 0; | |
2685 | ||
2686 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2687 | return 0; | |
2688 | ||
2689 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2690 | down_read(&oi->ip_alloc_sem); | |
2691 | has_space = ocfs2_xattr_has_space_inline(inode, di); | |
2692 | up_read(&oi->ip_alloc_sem); | |
2693 | if (!has_space) | |
2694 | return 0; | |
2695 | } | |
2696 | ||
2697 | xs->xattr_bh = xs->inode_bh; | |
2698 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
2699 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
2700 | xs->header = (struct ocfs2_xattr_header *) | |
2701 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
2702 | else | |
2703 | xs->header = (struct ocfs2_xattr_header *) | |
2704 | (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size); | |
2705 | xs->base = (void *)xs->header; | |
2706 | xs->here = xs->header->xh_entries; | |
2707 | ||
2708 | /* Find the named attribute. */ | |
2709 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
2710 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2711 | if (ret && ret != -ENODATA) | |
2712 | return ret; | |
2713 | xs->not_found = ret; | |
2714 | } | |
2715 | ||
2716 | return 0; | |
2717 | } | |
2718 | ||
2719 | /* | |
2720 | * ocfs2_xattr_ibody_set() | |
2721 | * | |
2722 | * Set, replace or remove an extended attribute into inode block. | |
2723 | * | |
2724 | */ | |
2725 | static int ocfs2_xattr_ibody_set(struct inode *inode, | |
2726 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2727 | struct ocfs2_xattr_search *xs, |
2728 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2729 | { |
2730 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2731 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2732 | int ret; | |
2733 | ||
2734 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2735 | return -ENOSPC; | |
2736 | ||
2737 | down_write(&oi->ip_alloc_sem); | |
2738 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2739 | if (!ocfs2_xattr_has_space_inline(inode, di)) { | |
2740 | ret = -ENOSPC; | |
2741 | goto out; | |
2742 | } | |
2743 | } | |
2744 | ||
78f30c31 | 2745 | ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, |
cf1d6c76 TY |
2746 | (OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL)); |
2747 | out: | |
2748 | up_write(&oi->ip_alloc_sem); | |
2749 | ||
2750 | return ret; | |
2751 | } | |
2752 | ||
2753 | /* | |
2754 | * ocfs2_xattr_block_find() | |
2755 | * | |
2756 | * Find extended attribute in external block and | |
2757 | * fill search info into struct ocfs2_xattr_search. | |
2758 | */ | |
2759 | static int ocfs2_xattr_block_find(struct inode *inode, | |
2760 | int name_index, | |
2761 | const char *name, | |
2762 | struct ocfs2_xattr_search *xs) | |
2763 | { | |
2764 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2765 | struct buffer_head *blk_bh = NULL; | |
589dc260 | 2766 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
2767 | int ret = 0; |
2768 | ||
2769 | if (!di->i_xattr_loc) | |
2770 | return ret; | |
2771 | ||
4ae1d69b JB |
2772 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
2773 | &blk_bh); | |
cf1d6c76 TY |
2774 | if (ret < 0) { |
2775 | mlog_errno(ret); | |
2776 | return ret; | |
2777 | } | |
f6087fb7 | 2778 | |
cf1d6c76 | 2779 | xs->xattr_bh = blk_bh; |
4ae1d69b | 2780 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
589dc260 TM |
2781 | |
2782 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2783 | xs->header = &xb->xb_attrs.xb_header; | |
2784 | xs->base = (void *)xs->header; | |
2785 | xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size; | |
2786 | xs->here = xs->header->xh_entries; | |
2787 | ||
2788 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2789 | } else | |
2790 | ret = ocfs2_xattr_index_block_find(inode, blk_bh, | |
2791 | name_index, | |
2792 | name, xs); | |
cf1d6c76 | 2793 | |
cf1d6c76 TY |
2794 | if (ret && ret != -ENODATA) { |
2795 | xs->xattr_bh = NULL; | |
2796 | goto cleanup; | |
2797 | } | |
2798 | xs->not_found = ret; | |
2799 | return 0; | |
cf1d6c76 TY |
2800 | cleanup: |
2801 | brelse(blk_bh); | |
2802 | ||
2803 | return ret; | |
2804 | } | |
2805 | ||
5aea1f0e TM |
2806 | static int ocfs2_create_xattr_block(handle_t *handle, |
2807 | struct inode *inode, | |
2808 | struct buffer_head *inode_bh, | |
2809 | struct ocfs2_alloc_context *meta_ac, | |
a7fe7a3a TM |
2810 | struct buffer_head **ret_bh, |
2811 | int indexed) | |
5aea1f0e TM |
2812 | { |
2813 | int ret; | |
2814 | u16 suballoc_bit_start; | |
2815 | u32 num_got; | |
2816 | u64 first_blkno; | |
2817 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data; | |
2818 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2819 | struct buffer_head *new_bh = NULL; | |
2820 | struct ocfs2_xattr_block *xblk; | |
2821 | ||
2822 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), inode_bh, | |
2823 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2824 | if (ret < 0) { | |
2825 | mlog_errno(ret); | |
2826 | goto end; | |
2827 | } | |
2828 | ||
2829 | ret = ocfs2_claim_metadata(osb, handle, meta_ac, 1, | |
2830 | &suballoc_bit_start, &num_got, | |
2831 | &first_blkno); | |
2832 | if (ret < 0) { | |
2833 | mlog_errno(ret); | |
2834 | goto end; | |
2835 | } | |
2836 | ||
2837 | new_bh = sb_getblk(inode->i_sb, first_blkno); | |
2838 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); | |
2839 | ||
2840 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), | |
2841 | new_bh, | |
2842 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2843 | if (ret < 0) { | |
2844 | mlog_errno(ret); | |
2845 | goto end; | |
2846 | } | |
2847 | ||
2848 | /* Initialize ocfs2_xattr_block */ | |
2849 | xblk = (struct ocfs2_xattr_block *)new_bh->b_data; | |
2850 | memset(xblk, 0, inode->i_sb->s_blocksize); | |
2851 | strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); | |
b89c5428 | 2852 | xblk->xb_suballoc_slot = cpu_to_le16(meta_ac->ac_alloc_slot); |
5aea1f0e TM |
2853 | xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); |
2854 | xblk->xb_fs_generation = cpu_to_le32(osb->fs_generation); | |
2855 | xblk->xb_blkno = cpu_to_le64(first_blkno); | |
2856 | ||
a7fe7a3a TM |
2857 | if (indexed) { |
2858 | struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root; | |
2859 | xr->xt_clusters = cpu_to_le32(1); | |
2860 | xr->xt_last_eb_blk = 0; | |
2861 | xr->xt_list.l_tree_depth = 0; | |
2862 | xr->xt_list.l_count = cpu_to_le16( | |
2863 | ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
2864 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
2865 | xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED); | |
2866 | } | |
2867 | ||
5aea1f0e TM |
2868 | ret = ocfs2_journal_dirty(handle, new_bh); |
2869 | if (ret < 0) { | |
2870 | mlog_errno(ret); | |
2871 | goto end; | |
2872 | } | |
2873 | di->i_xattr_loc = cpu_to_le64(first_blkno); | |
2874 | ocfs2_journal_dirty(handle, inode_bh); | |
2875 | ||
2876 | *ret_bh = new_bh; | |
2877 | new_bh = NULL; | |
2878 | ||
2879 | end: | |
2880 | brelse(new_bh); | |
2881 | return ret; | |
2882 | } | |
2883 | ||
cf1d6c76 TY |
2884 | /* |
2885 | * ocfs2_xattr_block_set() | |
2886 | * | |
2887 | * Set, replace or remove an extended attribute into external block. | |
2888 | * | |
2889 | */ | |
2890 | static int ocfs2_xattr_block_set(struct inode *inode, | |
2891 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2892 | struct ocfs2_xattr_search *xs, |
2893 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2894 | { |
2895 | struct buffer_head *new_bh = NULL; | |
85db90e7 | 2896 | handle_t *handle = ctxt->handle; |
cf1d6c76 | 2897 | struct ocfs2_xattr_block *xblk = NULL; |
cf1d6c76 TY |
2898 | int ret; |
2899 | ||
2900 | if (!xs->xattr_bh) { | |
5aea1f0e | 2901 | ret = ocfs2_create_xattr_block(handle, inode, xs->inode_bh, |
a7fe7a3a | 2902 | ctxt->meta_ac, &new_bh, 0); |
5aea1f0e | 2903 | if (ret) { |
cf1d6c76 | 2904 | mlog_errno(ret); |
85db90e7 | 2905 | goto end; |
cf1d6c76 TY |
2906 | } |
2907 | ||
cf1d6c76 | 2908 | xs->xattr_bh = new_bh; |
5aea1f0e | 2909 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
2910 | xs->header = &xblk->xb_attrs.xb_header; |
2911 | xs->base = (void *)xs->header; | |
2912 | xs->end = (void *)xblk + inode->i_sb->s_blocksize; | |
2913 | xs->here = xs->header->xh_entries; | |
01225596 TM |
2914 | } else |
2915 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; | |
2916 | ||
2917 | if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2918 | /* Set extended attribute into external block */ | |
78f30c31 TM |
2919 | ret = ocfs2_xattr_set_entry(inode, xi, xs, ctxt, |
2920 | OCFS2_HAS_XATTR_FL); | |
01225596 TM |
2921 | if (!ret || ret != -ENOSPC) |
2922 | goto end; | |
2923 | ||
78f30c31 | 2924 | ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); |
01225596 TM |
2925 | if (ret) |
2926 | goto end; | |
cf1d6c76 TY |
2927 | } |
2928 | ||
78f30c31 | 2929 | ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); |
01225596 TM |
2930 | |
2931 | end: | |
cf1d6c76 TY |
2932 | |
2933 | return ret; | |
2934 | } | |
2935 | ||
78f30c31 TM |
2936 | /* Check whether the new xattr can be inserted into the inode. */ |
2937 | static int ocfs2_xattr_can_be_in_inode(struct inode *inode, | |
2938 | struct ocfs2_xattr_info *xi, | |
2939 | struct ocfs2_xattr_search *xs) | |
2940 | { | |
78f30c31 TM |
2941 | struct ocfs2_xattr_entry *last; |
2942 | int free, i; | |
2943 | size_t min_offs = xs->end - xs->base; | |
2944 | ||
2945 | if (!xs->header) | |
2946 | return 0; | |
2947 | ||
2948 | last = xs->header->xh_entries; | |
2949 | ||
2950 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
2951 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
2952 | if (offs < min_offs) | |
2953 | min_offs = offs; | |
2954 | last += 1; | |
2955 | } | |
2956 | ||
4442f518 | 2957 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
78f30c31 TM |
2958 | if (free < 0) |
2959 | return 0; | |
2960 | ||
2961 | BUG_ON(!xs->not_found); | |
2962 | ||
199799a3 | 2963 | if (free >= (sizeof(struct ocfs2_xattr_entry) + namevalue_size_xi(xi))) |
78f30c31 TM |
2964 | return 1; |
2965 | ||
2966 | return 0; | |
2967 | } | |
2968 | ||
2969 | static int ocfs2_calc_xattr_set_need(struct inode *inode, | |
2970 | struct ocfs2_dinode *di, | |
2971 | struct ocfs2_xattr_info *xi, | |
2972 | struct ocfs2_xattr_search *xis, | |
2973 | struct ocfs2_xattr_search *xbs, | |
2974 | int *clusters_need, | |
85db90e7 TM |
2975 | int *meta_need, |
2976 | int *credits_need) | |
78f30c31 TM |
2977 | { |
2978 | int ret = 0, old_in_xb = 0; | |
85db90e7 | 2979 | int clusters_add = 0, meta_add = 0, credits = 0; |
78f30c31 TM |
2980 | struct buffer_head *bh = NULL; |
2981 | struct ocfs2_xattr_block *xb = NULL; | |
2982 | struct ocfs2_xattr_entry *xe = NULL; | |
2983 | struct ocfs2_xattr_value_root *xv = NULL; | |
2984 | char *base = NULL; | |
2985 | int name_offset, name_len = 0; | |
2986 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
6b240ff6 | 2987 | xi->xi_value_len); |
78f30c31 TM |
2988 | u64 value_size; |
2989 | ||
71d548a6 TM |
2990 | /* |
2991 | * Calculate the clusters we need to write. | |
2992 | * No matter whether we replace an old one or add a new one, | |
2993 | * we need this for writing. | |
2994 | */ | |
6b240ff6 | 2995 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
71d548a6 TM |
2996 | credits += new_clusters * |
2997 | ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
2998 | ||
78f30c31 | 2999 | if (xis->not_found && xbs->not_found) { |
85db90e7 TM |
3000 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3001 | ||
6b240ff6 | 3002 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 | 3003 | clusters_add += new_clusters; |
85db90e7 TM |
3004 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
3005 | &def_xv.xv.xr_list, | |
3006 | new_clusters); | |
3007 | } | |
78f30c31 TM |
3008 | |
3009 | goto meta_guess; | |
3010 | } | |
3011 | ||
3012 | if (!xis->not_found) { | |
3013 | xe = xis->here; | |
3014 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3015 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3016 | base = xis->base; | |
85db90e7 | 3017 | credits += OCFS2_INODE_UPDATE_CREDITS; |
78f30c31 | 3018 | } else { |
970e4936 | 3019 | int i, block_off = 0; |
78f30c31 TM |
3020 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; |
3021 | xe = xbs->here; | |
3022 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3023 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3024 | i = xbs->here - xbs->header->xh_entries; | |
3025 | old_in_xb = 1; | |
3026 | ||
3027 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 3028 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
78f30c31 TM |
3029 | bucket_xh(xbs->bucket), |
3030 | i, &block_off, | |
3031 | &name_offset); | |
3032 | base = bucket_block(xbs->bucket, block_off); | |
85db90e7 TM |
3033 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3034 | } else { | |
78f30c31 | 3035 | base = xbs->base; |
85db90e7 TM |
3036 | credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS; |
3037 | } | |
3038 | } | |
3039 | ||
3040 | /* | |
3041 | * delete a xattr doesn't need metadata and cluster allocation. | |
3042 | * so just calculate the credits and return. | |
3043 | * | |
3044 | * The credits for removing the value tree will be extended | |
3045 | * by ocfs2_remove_extent itself. | |
3046 | */ | |
6b240ff6 | 3047 | if (!xi->xi_value) { |
85db90e7 | 3048 | if (!ocfs2_xattr_is_local(xe)) |
a90714c1 | 3049 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
85db90e7 TM |
3050 | |
3051 | goto out; | |
78f30c31 TM |
3052 | } |
3053 | ||
3054 | /* do cluster allocation guess first. */ | |
3055 | value_size = le64_to_cpu(xe->xe_value_size); | |
3056 | ||
3057 | if (old_in_xb) { | |
3058 | /* | |
3059 | * In xattr set, we always try to set the xe in inode first, | |
3060 | * so if it can be inserted into inode successfully, the old | |
3061 | * one will be removed from the xattr block, and this xattr | |
3062 | * will be inserted into inode as a new xattr in inode. | |
3063 | */ | |
3064 | if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) { | |
3065 | clusters_add += new_clusters; | |
a90714c1 | 3066 | credits += ocfs2_remove_extent_credits(inode->i_sb) + |
85db90e7 TM |
3067 | OCFS2_INODE_UPDATE_CREDITS; |
3068 | if (!ocfs2_xattr_is_local(xe)) | |
3069 | credits += ocfs2_calc_extend_credits( | |
3070 | inode->i_sb, | |
3071 | &def_xv.xv.xr_list, | |
3072 | new_clusters); | |
78f30c31 TM |
3073 | goto out; |
3074 | } | |
3075 | } | |
3076 | ||
6b240ff6 | 3077 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 TM |
3078 | /* the new values will be stored outside. */ |
3079 | u32 old_clusters = 0; | |
3080 | ||
3081 | if (!ocfs2_xattr_is_local(xe)) { | |
3082 | old_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
3083 | value_size); | |
3084 | xv = (struct ocfs2_xattr_value_root *) | |
3085 | (base + name_offset + name_len); | |
97aff52a | 3086 | value_size = OCFS2_XATTR_ROOT_SIZE; |
78f30c31 TM |
3087 | } else |
3088 | xv = &def_xv.xv; | |
3089 | ||
85db90e7 | 3090 | if (old_clusters >= new_clusters) { |
a90714c1 | 3091 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
78f30c31 | 3092 | goto out; |
85db90e7 | 3093 | } else { |
78f30c31 TM |
3094 | meta_add += ocfs2_extend_meta_needed(&xv->xr_list); |
3095 | clusters_add += new_clusters - old_clusters; | |
85db90e7 TM |
3096 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
3097 | &xv->xr_list, | |
3098 | new_clusters - | |
3099 | old_clusters); | |
97aff52a TM |
3100 | if (value_size >= OCFS2_XATTR_ROOT_SIZE) |
3101 | goto out; | |
78f30c31 TM |
3102 | } |
3103 | } else { | |
3104 | /* | |
3105 | * Now the new value will be stored inside. So if the new | |
3106 | * value is smaller than the size of value root or the old | |
3107 | * value, we don't need any allocation, otherwise we have | |
3108 | * to guess metadata allocation. | |
3109 | */ | |
6b240ff6 JB |
3110 | if ((ocfs2_xattr_is_local(xe) && |
3111 | (value_size >= xi->xi_value_len)) || | |
78f30c31 | 3112 | (!ocfs2_xattr_is_local(xe) && |
6b240ff6 | 3113 | OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len)) |
78f30c31 TM |
3114 | goto out; |
3115 | } | |
3116 | ||
3117 | meta_guess: | |
3118 | /* calculate metadata allocation. */ | |
3119 | if (di->i_xattr_loc) { | |
3120 | if (!xbs->xattr_bh) { | |
4ae1d69b JB |
3121 | ret = ocfs2_read_xattr_block(inode, |
3122 | le64_to_cpu(di->i_xattr_loc), | |
3123 | &bh); | |
78f30c31 TM |
3124 | if (ret) { |
3125 | mlog_errno(ret); | |
3126 | goto out; | |
3127 | } | |
3128 | ||
3129 | xb = (struct ocfs2_xattr_block *)bh->b_data; | |
3130 | } else | |
3131 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
3132 | ||
90cb546c TM |
3133 | /* |
3134 | * If there is already an xattr tree, good, we can calculate | |
3135 | * like other b-trees. Otherwise we may have the chance of | |
3136 | * create a tree, the credit calculation is borrowed from | |
3137 | * ocfs2_calc_extend_credits with root_el = NULL. And the | |
3138 | * new tree will be cluster based, so no meta is needed. | |
3139 | */ | |
78f30c31 TM |
3140 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { |
3141 | struct ocfs2_extent_list *el = | |
3142 | &xb->xb_attrs.xb_root.xt_list; | |
3143 | meta_add += ocfs2_extend_meta_needed(el); | |
85db90e7 TM |
3144 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
3145 | el, 1); | |
90cb546c TM |
3146 | } else |
3147 | credits += OCFS2_SUBALLOC_ALLOC + 1; | |
78f30c31 TM |
3148 | |
3149 | /* | |
3150 | * This cluster will be used either for new bucket or for | |
3151 | * new xattr block. | |
3152 | * If the cluster size is the same as the bucket size, one | |
3153 | * more is needed since we may need to extend the bucket | |
3154 | * also. | |
3155 | */ | |
3156 | clusters_add += 1; | |
85db90e7 | 3157 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
78f30c31 | 3158 | if (OCFS2_XATTR_BUCKET_SIZE == |
85db90e7 TM |
3159 | OCFS2_SB(inode->i_sb)->s_clustersize) { |
3160 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
78f30c31 | 3161 | clusters_add += 1; |
85db90e7 TM |
3162 | } |
3163 | } else { | |
78f30c31 | 3164 | meta_add += 1; |
85db90e7 TM |
3165 | credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
3166 | } | |
78f30c31 TM |
3167 | out: |
3168 | if (clusters_need) | |
3169 | *clusters_need = clusters_add; | |
3170 | if (meta_need) | |
3171 | *meta_need = meta_add; | |
85db90e7 TM |
3172 | if (credits_need) |
3173 | *credits_need = credits; | |
78f30c31 TM |
3174 | brelse(bh); |
3175 | return ret; | |
3176 | } | |
3177 | ||
3178 | static int ocfs2_init_xattr_set_ctxt(struct inode *inode, | |
3179 | struct ocfs2_dinode *di, | |
3180 | struct ocfs2_xattr_info *xi, | |
3181 | struct ocfs2_xattr_search *xis, | |
3182 | struct ocfs2_xattr_search *xbs, | |
85db90e7 | 3183 | struct ocfs2_xattr_set_ctxt *ctxt, |
492a8a33 | 3184 | int extra_meta, |
85db90e7 | 3185 | int *credits) |
78f30c31 TM |
3186 | { |
3187 | int clusters_add, meta_add, ret; | |
3188 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
3189 | ||
3190 | memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt)); | |
3191 | ||
3192 | ocfs2_init_dealloc_ctxt(&ctxt->dealloc); | |
3193 | ||
3194 | ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs, | |
85db90e7 | 3195 | &clusters_add, &meta_add, credits); |
78f30c31 TM |
3196 | if (ret) { |
3197 | mlog_errno(ret); | |
3198 | return ret; | |
3199 | } | |
3200 | ||
492a8a33 | 3201 | meta_add += extra_meta; |
85db90e7 | 3202 | mlog(0, "Set xattr %s, reserve meta blocks = %d, clusters = %d, " |
6b240ff6 | 3203 | "credits = %d\n", xi->xi_name, meta_add, clusters_add, *credits); |
78f30c31 TM |
3204 | |
3205 | if (meta_add) { | |
3206 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, | |
3207 | &ctxt->meta_ac); | |
3208 | if (ret) { | |
3209 | mlog_errno(ret); | |
3210 | goto out; | |
3211 | } | |
3212 | } | |
3213 | ||
3214 | if (clusters_add) { | |
3215 | ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac); | |
3216 | if (ret) | |
3217 | mlog_errno(ret); | |
3218 | } | |
3219 | out: | |
3220 | if (ret) { | |
3221 | if (ctxt->meta_ac) { | |
3222 | ocfs2_free_alloc_context(ctxt->meta_ac); | |
3223 | ctxt->meta_ac = NULL; | |
3224 | } | |
3225 | ||
3226 | /* | |
3227 | * We cannot have an error and a non null ctxt->data_ac. | |
3228 | */ | |
3229 | } | |
3230 | ||
3231 | return ret; | |
3232 | } | |
3233 | ||
85db90e7 TM |
3234 | static int __ocfs2_xattr_set_handle(struct inode *inode, |
3235 | struct ocfs2_dinode *di, | |
3236 | struct ocfs2_xattr_info *xi, | |
3237 | struct ocfs2_xattr_search *xis, | |
3238 | struct ocfs2_xattr_search *xbs, | |
3239 | struct ocfs2_xattr_set_ctxt *ctxt) | |
3240 | { | |
9f868f16 | 3241 | int ret = 0, credits, old_found; |
85db90e7 | 3242 | |
6b240ff6 | 3243 | if (!xi->xi_value) { |
85db90e7 TM |
3244 | /* Remove existing extended attribute */ |
3245 | if (!xis->not_found) | |
3246 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3247 | else if (!xbs->not_found) | |
3248 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3249 | } else { | |
3250 | /* We always try to set extended attribute into inode first*/ | |
3251 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3252 | if (!ret && !xbs->not_found) { | |
3253 | /* | |
3254 | * If succeed and that extended attribute existing in | |
3255 | * external block, then we will remove it. | |
3256 | */ | |
6b240ff6 JB |
3257 | xi->xi_value = NULL; |
3258 | xi->xi_value_len = 0; | |
85db90e7 | 3259 | |
9f868f16 | 3260 | old_found = xis->not_found; |
85db90e7 TM |
3261 | xis->not_found = -ENODATA; |
3262 | ret = ocfs2_calc_xattr_set_need(inode, | |
3263 | di, | |
3264 | xi, | |
3265 | xis, | |
3266 | xbs, | |
3267 | NULL, | |
3268 | NULL, | |
3269 | &credits); | |
9f868f16 | 3270 | xis->not_found = old_found; |
85db90e7 TM |
3271 | if (ret) { |
3272 | mlog_errno(ret); | |
3273 | goto out; | |
3274 | } | |
3275 | ||
3276 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
3277 | ctxt->handle->h_buffer_credits); | |
3278 | if (ret) { | |
3279 | mlog_errno(ret); | |
3280 | goto out; | |
3281 | } | |
3282 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3283 | } else if (ret == -ENOSPC) { | |
3284 | if (di->i_xattr_loc && !xbs->xattr_bh) { | |
3285 | ret = ocfs2_xattr_block_find(inode, | |
6b240ff6 JB |
3286 | xi->xi_name_index, |
3287 | xi->xi_name, xbs); | |
85db90e7 TM |
3288 | if (ret) |
3289 | goto out; | |
3290 | ||
9f868f16 | 3291 | old_found = xis->not_found; |
85db90e7 TM |
3292 | xis->not_found = -ENODATA; |
3293 | ret = ocfs2_calc_xattr_set_need(inode, | |
3294 | di, | |
3295 | xi, | |
3296 | xis, | |
3297 | xbs, | |
3298 | NULL, | |
3299 | NULL, | |
3300 | &credits); | |
9f868f16 | 3301 | xis->not_found = old_found; |
85db90e7 TM |
3302 | if (ret) { |
3303 | mlog_errno(ret); | |
3304 | goto out; | |
3305 | } | |
3306 | ||
3307 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
3308 | ctxt->handle->h_buffer_credits); | |
3309 | if (ret) { | |
3310 | mlog_errno(ret); | |
3311 | goto out; | |
3312 | } | |
3313 | } | |
3314 | /* | |
3315 | * If no space in inode, we will set extended attribute | |
3316 | * into external block. | |
3317 | */ | |
3318 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3319 | if (ret) | |
3320 | goto out; | |
3321 | if (!xis->not_found) { | |
3322 | /* | |
3323 | * If succeed and that extended attribute | |
3324 | * existing in inode, we will remove it. | |
3325 | */ | |
6b240ff6 JB |
3326 | xi->xi_value = NULL; |
3327 | xi->xi_value_len = 0; | |
85db90e7 TM |
3328 | xbs->not_found = -ENODATA; |
3329 | ret = ocfs2_calc_xattr_set_need(inode, | |
3330 | di, | |
3331 | xi, | |
3332 | xis, | |
3333 | xbs, | |
3334 | NULL, | |
3335 | NULL, | |
3336 | &credits); | |
3337 | if (ret) { | |
3338 | mlog_errno(ret); | |
3339 | goto out; | |
3340 | } | |
3341 | ||
3342 | ret = ocfs2_extend_trans(ctxt->handle, credits + | |
3343 | ctxt->handle->h_buffer_credits); | |
3344 | if (ret) { | |
3345 | mlog_errno(ret); | |
3346 | goto out; | |
3347 | } | |
3348 | ret = ocfs2_xattr_ibody_set(inode, xi, | |
3349 | xis, ctxt); | |
3350 | } | |
3351 | } | |
3352 | } | |
3353 | ||
4b3f6209 TM |
3354 | if (!ret) { |
3355 | /* Update inode ctime. */ | |
0cf2f763 | 3356 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
89a907af TM |
3357 | xis->inode_bh, |
3358 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4b3f6209 TM |
3359 | if (ret) { |
3360 | mlog_errno(ret); | |
3361 | goto out; | |
3362 | } | |
3363 | ||
3364 | inode->i_ctime = CURRENT_TIME; | |
3365 | di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); | |
3366 | di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); | |
3367 | ocfs2_journal_dirty(ctxt->handle, xis->inode_bh); | |
3368 | } | |
85db90e7 TM |
3369 | out: |
3370 | return ret; | |
3371 | } | |
3372 | ||
6c3faba4 TY |
3373 | /* |
3374 | * This function only called duing creating inode | |
3375 | * for init security/acl xattrs of the new inode. | |
008aafaf | 3376 | * All transanction credits have been reserved in mknod. |
6c3faba4 TY |
3377 | */ |
3378 | int ocfs2_xattr_set_handle(handle_t *handle, | |
3379 | struct inode *inode, | |
3380 | struct buffer_head *di_bh, | |
3381 | int name_index, | |
3382 | const char *name, | |
3383 | const void *value, | |
3384 | size_t value_len, | |
3385 | int flags, | |
3386 | struct ocfs2_alloc_context *meta_ac, | |
3387 | struct ocfs2_alloc_context *data_ac) | |
3388 | { | |
3389 | struct ocfs2_dinode *di; | |
3390 | int ret; | |
3391 | ||
3392 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3393 | .xi_name_index = name_index, |
3394 | .xi_name = name, | |
18853b95 | 3395 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3396 | .xi_value = value, |
3397 | .xi_value_len = value_len, | |
6c3faba4 TY |
3398 | }; |
3399 | ||
3400 | struct ocfs2_xattr_search xis = { | |
3401 | .not_found = -ENODATA, | |
3402 | }; | |
3403 | ||
3404 | struct ocfs2_xattr_search xbs = { | |
3405 | .not_found = -ENODATA, | |
3406 | }; | |
3407 | ||
3408 | struct ocfs2_xattr_set_ctxt ctxt = { | |
3409 | .handle = handle, | |
3410 | .meta_ac = meta_ac, | |
3411 | .data_ac = data_ac, | |
3412 | }; | |
3413 | ||
3414 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) | |
3415 | return -EOPNOTSUPP; | |
3416 | ||
008aafaf TY |
3417 | /* |
3418 | * In extreme situation, may need xattr bucket when | |
3419 | * block size is too small. And we have already reserved | |
3420 | * the credits for bucket in mknod. | |
3421 | */ | |
3422 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) { | |
3423 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3424 | if (!xbs.bucket) { | |
3425 | mlog_errno(-ENOMEM); | |
3426 | return -ENOMEM; | |
3427 | } | |
3428 | } | |
3429 | ||
6c3faba4 TY |
3430 | xis.inode_bh = xbs.inode_bh = di_bh; |
3431 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3432 | ||
3433 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3434 | ||
3435 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3436 | if (ret) | |
3437 | goto cleanup; | |
3438 | if (xis.not_found) { | |
3439 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3440 | if (ret) | |
3441 | goto cleanup; | |
3442 | } | |
3443 | ||
3444 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3445 | ||
3446 | cleanup: | |
3447 | up_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3448 | brelse(xbs.xattr_bh); | |
008aafaf | 3449 | ocfs2_xattr_bucket_free(xbs.bucket); |
6c3faba4 TY |
3450 | |
3451 | return ret; | |
3452 | } | |
3453 | ||
cf1d6c76 TY |
3454 | /* |
3455 | * ocfs2_xattr_set() | |
3456 | * | |
3457 | * Set, replace or remove an extended attribute for this inode. | |
3458 | * value is NULL to remove an existing extended attribute, else either | |
3459 | * create or replace an extended attribute. | |
3460 | */ | |
3461 | int ocfs2_xattr_set(struct inode *inode, | |
3462 | int name_index, | |
3463 | const char *name, | |
3464 | const void *value, | |
3465 | size_t value_len, | |
3466 | int flags) | |
3467 | { | |
3468 | struct buffer_head *di_bh = NULL; | |
3469 | struct ocfs2_dinode *di; | |
492a8a33 | 3470 | int ret, credits, ref_meta = 0, ref_credits = 0; |
78f30c31 | 3471 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
85db90e7 | 3472 | struct inode *tl_inode = osb->osb_tl_inode; |
78f30c31 | 3473 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; |
492a8a33 | 3474 | struct ocfs2_refcount_tree *ref_tree = NULL; |
cf1d6c76 TY |
3475 | |
3476 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3477 | .xi_name_index = name_index, |
3478 | .xi_name = name, | |
18853b95 | 3479 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3480 | .xi_value = value, |
3481 | .xi_value_len = value_len, | |
cf1d6c76 TY |
3482 | }; |
3483 | ||
3484 | struct ocfs2_xattr_search xis = { | |
3485 | .not_found = -ENODATA, | |
3486 | }; | |
3487 | ||
3488 | struct ocfs2_xattr_search xbs = { | |
3489 | .not_found = -ENODATA, | |
3490 | }; | |
3491 | ||
8154da3d TY |
3492 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
3493 | return -EOPNOTSUPP; | |
3494 | ||
ba937127 JB |
3495 | /* |
3496 | * Only xbs will be used on indexed trees. xis doesn't need a | |
3497 | * bucket. | |
3498 | */ | |
3499 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3500 | if (!xbs.bucket) { | |
3501 | mlog_errno(-ENOMEM); | |
3502 | return -ENOMEM; | |
3503 | } | |
3504 | ||
cf1d6c76 TY |
3505 | ret = ocfs2_inode_lock(inode, &di_bh, 1); |
3506 | if (ret < 0) { | |
3507 | mlog_errno(ret); | |
ba937127 | 3508 | goto cleanup_nolock; |
cf1d6c76 TY |
3509 | } |
3510 | xis.inode_bh = xbs.inode_bh = di_bh; | |
3511 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3512 | ||
3513 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3514 | /* | |
3515 | * Scan inode and external block to find the same name | |
3516 | * extended attribute and collect search infomation. | |
3517 | */ | |
3518 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3519 | if (ret) | |
3520 | goto cleanup; | |
3521 | if (xis.not_found) { | |
3522 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3523 | if (ret) | |
3524 | goto cleanup; | |
3525 | } | |
3526 | ||
3527 | if (xis.not_found && xbs.not_found) { | |
3528 | ret = -ENODATA; | |
3529 | if (flags & XATTR_REPLACE) | |
3530 | goto cleanup; | |
3531 | ret = 0; | |
3532 | if (!value) | |
3533 | goto cleanup; | |
3534 | } else { | |
3535 | ret = -EEXIST; | |
3536 | if (flags & XATTR_CREATE) | |
3537 | goto cleanup; | |
3538 | } | |
3539 | ||
492a8a33 TM |
3540 | /* Check whether the value is refcounted and do some prepartion. */ |
3541 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL && | |
3542 | (!xis.not_found || !xbs.not_found)) { | |
3543 | ret = ocfs2_prepare_refcount_xattr(inode, di, &xi, | |
3544 | &xis, &xbs, &ref_tree, | |
3545 | &ref_meta, &ref_credits); | |
3546 | if (ret) { | |
3547 | mlog_errno(ret); | |
3548 | goto cleanup; | |
3549 | } | |
3550 | } | |
85db90e7 TM |
3551 | |
3552 | mutex_lock(&tl_inode->i_mutex); | |
3553 | ||
3554 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
3555 | ret = __ocfs2_flush_truncate_log(osb); | |
3556 | if (ret < 0) { | |
3557 | mutex_unlock(&tl_inode->i_mutex); | |
3558 | mlog_errno(ret); | |
3559 | goto cleanup; | |
3560 | } | |
3561 | } | |
3562 | mutex_unlock(&tl_inode->i_mutex); | |
3563 | ||
3564 | ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis, | |
492a8a33 | 3565 | &xbs, &ctxt, ref_meta, &credits); |
78f30c31 TM |
3566 | if (ret) { |
3567 | mlog_errno(ret); | |
3568 | goto cleanup; | |
3569 | } | |
3570 | ||
4b3f6209 TM |
3571 | /* we need to update inode's ctime field, so add credit for it. */ |
3572 | credits += OCFS2_INODE_UPDATE_CREDITS; | |
492a8a33 | 3573 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); |
85db90e7 TM |
3574 | if (IS_ERR(ctxt.handle)) { |
3575 | ret = PTR_ERR(ctxt.handle); | |
3576 | mlog_errno(ret); | |
3577 | goto cleanup; | |
cf1d6c76 | 3578 | } |
85db90e7 TM |
3579 | |
3580 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3581 | ||
3582 | ocfs2_commit_trans(osb, ctxt.handle); | |
3583 | ||
78f30c31 TM |
3584 | if (ctxt.data_ac) |
3585 | ocfs2_free_alloc_context(ctxt.data_ac); | |
3586 | if (ctxt.meta_ac) | |
3587 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
3588 | if (ocfs2_dealloc_has_cluster(&ctxt.dealloc)) | |
3589 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
3590 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
8b2c0dba | 3591 | |
cf1d6c76 | 3592 | cleanup: |
492a8a33 TM |
3593 | if (ref_tree) |
3594 | ocfs2_unlock_refcount_tree(osb, ref_tree, 1); | |
cf1d6c76 | 3595 | up_write(&OCFS2_I(inode)->ip_xattr_sem); |
8b2c0dba TM |
3596 | if (!value && !ret) { |
3597 | ret = ocfs2_try_remove_refcount_tree(inode, di_bh); | |
3598 | if (ret) | |
3599 | mlog_errno(ret); | |
3600 | } | |
cf1d6c76 | 3601 | ocfs2_inode_unlock(inode, 1); |
ba937127 | 3602 | cleanup_nolock: |
cf1d6c76 TY |
3603 | brelse(di_bh); |
3604 | brelse(xbs.xattr_bh); | |
ba937127 | 3605 | ocfs2_xattr_bucket_free(xbs.bucket); |
cf1d6c76 TY |
3606 | |
3607 | return ret; | |
3608 | } | |
3609 | ||
0c044f0b TM |
3610 | /* |
3611 | * Find the xattr extent rec which may contains name_hash. | |
3612 | * e_cpos will be the first name hash of the xattr rec. | |
3613 | * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list. | |
3614 | */ | |
3615 | static int ocfs2_xattr_get_rec(struct inode *inode, | |
3616 | u32 name_hash, | |
3617 | u64 *p_blkno, | |
3618 | u32 *e_cpos, | |
3619 | u32 *num_clusters, | |
3620 | struct ocfs2_extent_list *el) | |
3621 | { | |
3622 | int ret = 0, i; | |
3623 | struct buffer_head *eb_bh = NULL; | |
3624 | struct ocfs2_extent_block *eb; | |
3625 | struct ocfs2_extent_rec *rec = NULL; | |
3626 | u64 e_blkno = 0; | |
3627 | ||
3628 | if (el->l_tree_depth) { | |
facdb77f JB |
3629 | ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash, |
3630 | &eb_bh); | |
0c044f0b TM |
3631 | if (ret) { |
3632 | mlog_errno(ret); | |
3633 | goto out; | |
3634 | } | |
3635 | ||
3636 | eb = (struct ocfs2_extent_block *) eb_bh->b_data; | |
3637 | el = &eb->h_list; | |
3638 | ||
3639 | if (el->l_tree_depth) { | |
3640 | ocfs2_error(inode->i_sb, | |
3641 | "Inode %lu has non zero tree depth in " | |
3642 | "xattr tree block %llu\n", inode->i_ino, | |
3643 | (unsigned long long)eb_bh->b_blocknr); | |
3644 | ret = -EROFS; | |
3645 | goto out; | |
3646 | } | |
3647 | } | |
3648 | ||
3649 | for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { | |
3650 | rec = &el->l_recs[i]; | |
3651 | ||
3652 | if (le32_to_cpu(rec->e_cpos) <= name_hash) { | |
3653 | e_blkno = le64_to_cpu(rec->e_blkno); | |
3654 | break; | |
3655 | } | |
3656 | } | |
3657 | ||
3658 | if (!e_blkno) { | |
3659 | ocfs2_error(inode->i_sb, "Inode %lu has bad extent " | |
3660 | "record (%u, %u, 0) in xattr", inode->i_ino, | |
3661 | le32_to_cpu(rec->e_cpos), | |
3662 | ocfs2_rec_clusters(el, rec)); | |
3663 | ret = -EROFS; | |
3664 | goto out; | |
3665 | } | |
3666 | ||
3667 | *p_blkno = le64_to_cpu(rec->e_blkno); | |
3668 | *num_clusters = le16_to_cpu(rec->e_leaf_clusters); | |
3669 | if (e_cpos) | |
3670 | *e_cpos = le32_to_cpu(rec->e_cpos); | |
3671 | out: | |
3672 | brelse(eb_bh); | |
3673 | return ret; | |
3674 | } | |
3675 | ||
3676 | typedef int (xattr_bucket_func)(struct inode *inode, | |
3677 | struct ocfs2_xattr_bucket *bucket, | |
3678 | void *para); | |
3679 | ||
589dc260 | 3680 | static int ocfs2_find_xe_in_bucket(struct inode *inode, |
e2356a3f | 3681 | struct ocfs2_xattr_bucket *bucket, |
589dc260 TM |
3682 | int name_index, |
3683 | const char *name, | |
3684 | u32 name_hash, | |
3685 | u16 *xe_index, | |
3686 | int *found) | |
3687 | { | |
3688 | int i, ret = 0, cmp = 1, block_off, new_offset; | |
e2356a3f | 3689 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
589dc260 TM |
3690 | size_t name_len = strlen(name); |
3691 | struct ocfs2_xattr_entry *xe = NULL; | |
589dc260 TM |
3692 | char *xe_name; |
3693 | ||
3694 | /* | |
3695 | * We don't use binary search in the bucket because there | |
3696 | * may be multiple entries with the same name hash. | |
3697 | */ | |
3698 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
3699 | xe = &xh->xh_entries[i]; | |
3700 | ||
3701 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) | |
3702 | continue; | |
3703 | else if (name_hash < le32_to_cpu(xe->xe_name_hash)) | |
3704 | break; | |
3705 | ||
3706 | cmp = name_index - ocfs2_xattr_get_type(xe); | |
3707 | if (!cmp) | |
3708 | cmp = name_len - xe->xe_name_len; | |
3709 | if (cmp) | |
3710 | continue; | |
3711 | ||
fd68a894 | 3712 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
589dc260 TM |
3713 | xh, |
3714 | i, | |
3715 | &block_off, | |
3716 | &new_offset); | |
3717 | if (ret) { | |
3718 | mlog_errno(ret); | |
3719 | break; | |
3720 | } | |
3721 | ||
970e4936 | 3722 | |
e2356a3f JB |
3723 | xe_name = bucket_block(bucket, block_off) + new_offset; |
3724 | if (!memcmp(name, xe_name, name_len)) { | |
589dc260 TM |
3725 | *xe_index = i; |
3726 | *found = 1; | |
3727 | ret = 0; | |
3728 | break; | |
3729 | } | |
3730 | } | |
3731 | ||
3732 | return ret; | |
3733 | } | |
3734 | ||
3735 | /* | |
3736 | * Find the specified xattr entry in a series of buckets. | |
3737 | * This series start from p_blkno and last for num_clusters. | |
3738 | * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains | |
3739 | * the num of the valid buckets. | |
3740 | * | |
3741 | * Return the buffer_head this xattr should reside in. And if the xattr's | |
3742 | * hash is in the gap of 2 buckets, return the lower bucket. | |
3743 | */ | |
3744 | static int ocfs2_xattr_bucket_find(struct inode *inode, | |
3745 | int name_index, | |
3746 | const char *name, | |
3747 | u32 name_hash, | |
3748 | u64 p_blkno, | |
3749 | u32 first_hash, | |
3750 | u32 num_clusters, | |
3751 | struct ocfs2_xattr_search *xs) | |
3752 | { | |
3753 | int ret, found = 0; | |
589dc260 TM |
3754 | struct ocfs2_xattr_header *xh = NULL; |
3755 | struct ocfs2_xattr_entry *xe = NULL; | |
3756 | u16 index = 0; | |
3757 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
3758 | int low_bucket = 0, bucket, high_bucket; | |
e2356a3f | 3759 | struct ocfs2_xattr_bucket *search; |
589dc260 | 3760 | u32 last_hash; |
e2356a3f | 3761 | u64 blkno, lower_blkno = 0; |
589dc260 | 3762 | |
e2356a3f JB |
3763 | search = ocfs2_xattr_bucket_new(inode); |
3764 | if (!search) { | |
3765 | ret = -ENOMEM; | |
3766 | mlog_errno(ret); | |
3767 | goto out; | |
3768 | } | |
3769 | ||
3770 | ret = ocfs2_read_xattr_bucket(search, p_blkno); | |
589dc260 TM |
3771 | if (ret) { |
3772 | mlog_errno(ret); | |
3773 | goto out; | |
3774 | } | |
3775 | ||
e2356a3f | 3776 | xh = bucket_xh(search); |
589dc260 | 3777 | high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1; |
589dc260 | 3778 | while (low_bucket <= high_bucket) { |
e2356a3f | 3779 | ocfs2_xattr_bucket_relse(search); |
589dc260 | 3780 | |
e2356a3f | 3781 | bucket = (low_bucket + high_bucket) / 2; |
589dc260 | 3782 | blkno = p_blkno + bucket * blk_per_bucket; |
e2356a3f | 3783 | ret = ocfs2_read_xattr_bucket(search, blkno); |
589dc260 TM |
3784 | if (ret) { |
3785 | mlog_errno(ret); | |
3786 | goto out; | |
3787 | } | |
3788 | ||
e2356a3f | 3789 | xh = bucket_xh(search); |
589dc260 TM |
3790 | xe = &xh->xh_entries[0]; |
3791 | if (name_hash < le32_to_cpu(xe->xe_name_hash)) { | |
3792 | high_bucket = bucket - 1; | |
3793 | continue; | |
3794 | } | |
3795 | ||
3796 | /* | |
3797 | * Check whether the hash of the last entry in our | |
5a095611 TM |
3798 | * bucket is larger than the search one. for an empty |
3799 | * bucket, the last one is also the first one. | |
589dc260 | 3800 | */ |
5a095611 TM |
3801 | if (xh->xh_count) |
3802 | xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1]; | |
3803 | ||
589dc260 TM |
3804 | last_hash = le32_to_cpu(xe->xe_name_hash); |
3805 | ||
e2356a3f JB |
3806 | /* record lower_blkno which may be the insert place. */ |
3807 | lower_blkno = blkno; | |
589dc260 TM |
3808 | |
3809 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) { | |
3810 | low_bucket = bucket + 1; | |
3811 | continue; | |
3812 | } | |
3813 | ||
3814 | /* the searched xattr should reside in this bucket if exists. */ | |
e2356a3f | 3815 | ret = ocfs2_find_xe_in_bucket(inode, search, |
589dc260 TM |
3816 | name_index, name, name_hash, |
3817 | &index, &found); | |
3818 | if (ret) { | |
3819 | mlog_errno(ret); | |
3820 | goto out; | |
3821 | } | |
3822 | break; | |
3823 | } | |
3824 | ||
3825 | /* | |
3826 | * Record the bucket we have found. | |
3827 | * When the xattr's hash value is in the gap of 2 buckets, we will | |
3828 | * always set it to the previous bucket. | |
3829 | */ | |
e2356a3f JB |
3830 | if (!lower_blkno) |
3831 | lower_blkno = p_blkno; | |
3832 | ||
3833 | /* This should be in cache - we just read it during the search */ | |
3834 | ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno); | |
3835 | if (ret) { | |
3836 | mlog_errno(ret); | |
3837 | goto out; | |
589dc260 | 3838 | } |
589dc260 | 3839 | |
ba937127 JB |
3840 | xs->header = bucket_xh(xs->bucket); |
3841 | xs->base = bucket_block(xs->bucket, 0); | |
589dc260 TM |
3842 | xs->end = xs->base + inode->i_sb->s_blocksize; |
3843 | ||
3844 | if (found) { | |
589dc260 TM |
3845 | xs->here = &xs->header->xh_entries[index]; |
3846 | mlog(0, "find xattr %s in bucket %llu, entry = %u\n", name, | |
ba937127 | 3847 | (unsigned long long)bucket_blkno(xs->bucket), index); |
589dc260 TM |
3848 | } else |
3849 | ret = -ENODATA; | |
3850 | ||
3851 | out: | |
e2356a3f | 3852 | ocfs2_xattr_bucket_free(search); |
589dc260 TM |
3853 | return ret; |
3854 | } | |
3855 | ||
3856 | static int ocfs2_xattr_index_block_find(struct inode *inode, | |
3857 | struct buffer_head *root_bh, | |
3858 | int name_index, | |
3859 | const char *name, | |
3860 | struct ocfs2_xattr_search *xs) | |
3861 | { | |
3862 | int ret; | |
3863 | struct ocfs2_xattr_block *xb = | |
3864 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
3865 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
3866 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
3867 | u64 p_blkno = 0; | |
3868 | u32 first_hash, num_clusters = 0; | |
2057e5c6 | 3869 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
589dc260 TM |
3870 | |
3871 | if (le16_to_cpu(el->l_next_free_rec) == 0) | |
3872 | return -ENODATA; | |
3873 | ||
3874 | mlog(0, "find xattr %s, hash = %u, index = %d in xattr tree\n", | |
3875 | name, name_hash, name_index); | |
3876 | ||
3877 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash, | |
3878 | &num_clusters, el); | |
3879 | if (ret) { | |
3880 | mlog_errno(ret); | |
3881 | goto out; | |
3882 | } | |
3883 | ||
3884 | BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash); | |
3885 | ||
3886 | mlog(0, "find xattr extent rec %u clusters from %llu, the first hash " | |
de29c085 MF |
3887 | "in the rec is %u\n", num_clusters, (unsigned long long)p_blkno, |
3888 | first_hash); | |
589dc260 TM |
3889 | |
3890 | ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash, | |
3891 | p_blkno, first_hash, num_clusters, xs); | |
3892 | ||
3893 | out: | |
3894 | return ret; | |
3895 | } | |
3896 | ||
0c044f0b TM |
3897 | static int ocfs2_iterate_xattr_buckets(struct inode *inode, |
3898 | u64 blkno, | |
3899 | u32 clusters, | |
3900 | xattr_bucket_func *func, | |
3901 | void *para) | |
3902 | { | |
6dde41d9 | 3903 | int i, ret = 0; |
0c044f0b TM |
3904 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); |
3905 | u32 num_buckets = clusters * bpc; | |
ba937127 | 3906 | struct ocfs2_xattr_bucket *bucket; |
0c044f0b | 3907 | |
ba937127 JB |
3908 | bucket = ocfs2_xattr_bucket_new(inode); |
3909 | if (!bucket) { | |
3910 | mlog_errno(-ENOMEM); | |
3911 | return -ENOMEM; | |
3912 | } | |
0c044f0b TM |
3913 | |
3914 | mlog(0, "iterating xattr buckets in %u clusters starting from %llu\n", | |
de29c085 | 3915 | clusters, (unsigned long long)blkno); |
0c044f0b | 3916 | |
ba937127 JB |
3917 | for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) { |
3918 | ret = ocfs2_read_xattr_bucket(bucket, blkno); | |
0c044f0b TM |
3919 | if (ret) { |
3920 | mlog_errno(ret); | |
ba937127 | 3921 | break; |
0c044f0b TM |
3922 | } |
3923 | ||
0c044f0b TM |
3924 | /* |
3925 | * The real bucket num in this series of blocks is stored | |
3926 | * in the 1st bucket. | |
3927 | */ | |
3928 | if (i == 0) | |
ba937127 | 3929 | num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets); |
0c044f0b | 3930 | |
de29c085 MF |
3931 | mlog(0, "iterating xattr bucket %llu, first hash %u\n", |
3932 | (unsigned long long)blkno, | |
ba937127 | 3933 | le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); |
0c044f0b | 3934 | if (func) { |
ba937127 | 3935 | ret = func(inode, bucket, para); |
a46fa684 | 3936 | if (ret && ret != -ERANGE) |
0c044f0b | 3937 | mlog_errno(ret); |
ba937127 | 3938 | /* Fall through to bucket_relse() */ |
0c044f0b TM |
3939 | } |
3940 | ||
ba937127 JB |
3941 | ocfs2_xattr_bucket_relse(bucket); |
3942 | if (ret) | |
3943 | break; | |
0c044f0b TM |
3944 | } |
3945 | ||
ba937127 | 3946 | ocfs2_xattr_bucket_free(bucket); |
0c044f0b TM |
3947 | return ret; |
3948 | } | |
3949 | ||
3950 | struct ocfs2_xattr_tree_list { | |
3951 | char *buffer; | |
3952 | size_t buffer_size; | |
936b8834 | 3953 | size_t result; |
0c044f0b TM |
3954 | }; |
3955 | ||
fd68a894 | 3956 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
0c044f0b TM |
3957 | struct ocfs2_xattr_header *xh, |
3958 | int index, | |
3959 | int *block_off, | |
3960 | int *new_offset) | |
3961 | { | |
3962 | u16 name_offset; | |
3963 | ||
3964 | if (index < 0 || index >= le16_to_cpu(xh->xh_count)) | |
3965 | return -EINVAL; | |
3966 | ||
3967 | name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset); | |
3968 | ||
fd68a894 TM |
3969 | *block_off = name_offset >> sb->s_blocksize_bits; |
3970 | *new_offset = name_offset % sb->s_blocksize; | |
0c044f0b TM |
3971 | |
3972 | return 0; | |
3973 | } | |
3974 | ||
3975 | static int ocfs2_list_xattr_bucket(struct inode *inode, | |
3976 | struct ocfs2_xattr_bucket *bucket, | |
3977 | void *para) | |
3978 | { | |
936b8834 | 3979 | int ret = 0, type; |
0c044f0b | 3980 | struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para; |
0c044f0b | 3981 | int i, block_off, new_offset; |
936b8834 | 3982 | const char *prefix, *name; |
0c044f0b | 3983 | |
3e632946 JB |
3984 | for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) { |
3985 | struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i]; | |
936b8834 TM |
3986 | type = ocfs2_xattr_get_type(entry); |
3987 | prefix = ocfs2_xattr_prefix(type); | |
0c044f0b | 3988 | |
936b8834 | 3989 | if (prefix) { |
fd68a894 | 3990 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
3e632946 | 3991 | bucket_xh(bucket), |
0c044f0b TM |
3992 | i, |
3993 | &block_off, | |
3994 | &new_offset); | |
3995 | if (ret) | |
3996 | break; | |
936b8834 | 3997 | |
51def39f | 3998 | name = (const char *)bucket_block(bucket, block_off) + |
936b8834 TM |
3999 | new_offset; |
4000 | ret = ocfs2_xattr_list_entry(xl->buffer, | |
4001 | xl->buffer_size, | |
4002 | &xl->result, | |
4003 | prefix, name, | |
4004 | entry->xe_name_len); | |
4005 | if (ret) | |
4006 | break; | |
0c044f0b TM |
4007 | } |
4008 | } | |
4009 | ||
4010 | return ret; | |
4011 | } | |
4012 | ||
47bca495 TM |
4013 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, |
4014 | struct buffer_head *blk_bh, | |
4015 | xattr_tree_rec_func *rec_func, | |
4016 | void *para) | |
0c044f0b | 4017 | { |
47bca495 TM |
4018 | struct ocfs2_xattr_block *xb = |
4019 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
4020 | struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; | |
0c044f0b TM |
4021 | int ret = 0; |
4022 | u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0; | |
4023 | u64 p_blkno = 0; | |
0c044f0b | 4024 | |
47bca495 | 4025 | if (!el->l_next_free_rec || !rec_func) |
0c044f0b TM |
4026 | return 0; |
4027 | ||
4028 | while (name_hash > 0) { | |
4029 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, | |
4030 | &e_cpos, &num_clusters, el); | |
4031 | if (ret) { | |
4032 | mlog_errno(ret); | |
47bca495 | 4033 | break; |
0c044f0b TM |
4034 | } |
4035 | ||
47bca495 TM |
4036 | ret = rec_func(inode, blk_bh, p_blkno, e_cpos, |
4037 | num_clusters, para); | |
0c044f0b | 4038 | if (ret) { |
a46fa684 TM |
4039 | if (ret != -ERANGE) |
4040 | mlog_errno(ret); | |
47bca495 | 4041 | break; |
0c044f0b TM |
4042 | } |
4043 | ||
4044 | if (e_cpos == 0) | |
4045 | break; | |
4046 | ||
4047 | name_hash = e_cpos - 1; | |
4048 | } | |
4049 | ||
47bca495 TM |
4050 | return ret; |
4051 | ||
4052 | } | |
4053 | ||
4054 | static int ocfs2_list_xattr_tree_rec(struct inode *inode, | |
4055 | struct buffer_head *root_bh, | |
4056 | u64 blkno, u32 cpos, u32 len, void *para) | |
4057 | { | |
4058 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
4059 | ocfs2_list_xattr_bucket, para); | |
4060 | } | |
4061 | ||
4062 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, | |
4063 | struct buffer_head *blk_bh, | |
4064 | char *buffer, | |
4065 | size_t buffer_size) | |
4066 | { | |
4067 | int ret; | |
4068 | struct ocfs2_xattr_tree_list xl = { | |
4069 | .buffer = buffer, | |
4070 | .buffer_size = buffer_size, | |
4071 | .result = 0, | |
4072 | }; | |
4073 | ||
4074 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
4075 | ocfs2_list_xattr_tree_rec, &xl); | |
4076 | if (ret) { | |
4077 | mlog_errno(ret); | |
4078 | goto out; | |
4079 | } | |
4080 | ||
936b8834 | 4081 | ret = xl.result; |
0c044f0b TM |
4082 | out: |
4083 | return ret; | |
4084 | } | |
01225596 TM |
4085 | |
4086 | static int cmp_xe(const void *a, const void *b) | |
4087 | { | |
4088 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4089 | u32 l_hash = le32_to_cpu(l->xe_name_hash); | |
4090 | u32 r_hash = le32_to_cpu(r->xe_name_hash); | |
4091 | ||
4092 | if (l_hash > r_hash) | |
4093 | return 1; | |
4094 | if (l_hash < r_hash) | |
4095 | return -1; | |
4096 | return 0; | |
4097 | } | |
4098 | ||
4099 | static void swap_xe(void *a, void *b, int size) | |
4100 | { | |
4101 | struct ocfs2_xattr_entry *l = a, *r = b, tmp; | |
4102 | ||
4103 | tmp = *l; | |
4104 | memcpy(l, r, sizeof(struct ocfs2_xattr_entry)); | |
4105 | memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry)); | |
4106 | } | |
4107 | ||
4108 | /* | |
4109 | * When the ocfs2_xattr_block is filled up, new bucket will be created | |
4110 | * and all the xattr entries will be moved to the new bucket. | |
178eeac3 JB |
4111 | * The header goes at the start of the bucket, and the names+values are |
4112 | * filled from the end. This is why *target starts as the last buffer. | |
01225596 TM |
4113 | * Note: we need to sort the entries since they are not saved in order |
4114 | * in the ocfs2_xattr_block. | |
4115 | */ | |
4116 | static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, | |
4117 | struct buffer_head *xb_bh, | |
178eeac3 | 4118 | struct ocfs2_xattr_bucket *bucket) |
01225596 TM |
4119 | { |
4120 | int i, blocksize = inode->i_sb->s_blocksize; | |
178eeac3 | 4121 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 TM |
4122 | u16 offset, size, off_change; |
4123 | struct ocfs2_xattr_entry *xe; | |
4124 | struct ocfs2_xattr_block *xb = | |
4125 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4126 | struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header; | |
178eeac3 | 4127 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 4128 | u16 count = le16_to_cpu(xb_xh->xh_count); |
178eeac3 JB |
4129 | char *src = xb_bh->b_data; |
4130 | char *target = bucket_block(bucket, blks - 1); | |
01225596 TM |
4131 | |
4132 | mlog(0, "cp xattr from block %llu to bucket %llu\n", | |
4133 | (unsigned long long)xb_bh->b_blocknr, | |
178eeac3 JB |
4134 | (unsigned long long)bucket_blkno(bucket)); |
4135 | ||
4136 | for (i = 0; i < blks; i++) | |
4137 | memset(bucket_block(bucket, i), 0, blocksize); | |
01225596 | 4138 | |
01225596 TM |
4139 | /* |
4140 | * Since the xe_name_offset is based on ocfs2_xattr_header, | |
4141 | * there is a offset change corresponding to the change of | |
4142 | * ocfs2_xattr_header's position. | |
4143 | */ | |
4144 | off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4145 | xe = &xb_xh->xh_entries[count - 1]; | |
4146 | offset = le16_to_cpu(xe->xe_name_offset) + off_change; | |
4147 | size = blocksize - offset; | |
4148 | ||
4149 | /* copy all the names and values. */ | |
01225596 TM |
4150 | memcpy(target + offset, src + offset, size); |
4151 | ||
4152 | /* Init new header now. */ | |
4153 | xh->xh_count = xb_xh->xh_count; | |
4154 | xh->xh_num_buckets = cpu_to_le16(1); | |
4155 | xh->xh_name_value_len = cpu_to_le16(size); | |
4156 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size); | |
4157 | ||
4158 | /* copy all the entries. */ | |
178eeac3 | 4159 | target = bucket_block(bucket, 0); |
01225596 TM |
4160 | offset = offsetof(struct ocfs2_xattr_header, xh_entries); |
4161 | size = count * sizeof(struct ocfs2_xattr_entry); | |
4162 | memcpy(target + offset, (char *)xb_xh + offset, size); | |
4163 | ||
4164 | /* Change the xe offset for all the xe because of the move. */ | |
4165 | off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize + | |
4166 | offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4167 | for (i = 0; i < count; i++) | |
4168 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change); | |
4169 | ||
4170 | mlog(0, "copy entry: start = %u, size = %u, offset_change = %u\n", | |
4171 | offset, size, off_change); | |
4172 | ||
4173 | sort(target + offset, count, sizeof(struct ocfs2_xattr_entry), | |
4174 | cmp_xe, swap_xe); | |
4175 | } | |
4176 | ||
4177 | /* | |
4178 | * After we move xattr from block to index btree, we have to | |
4179 | * update ocfs2_xattr_search to the new xe and base. | |
4180 | * | |
4181 | * When the entry is in xattr block, xattr_bh indicates the storage place. | |
4182 | * While if the entry is in index b-tree, "bucket" indicates the | |
4183 | * real place of the xattr. | |
4184 | */ | |
178eeac3 JB |
4185 | static void ocfs2_xattr_update_xattr_search(struct inode *inode, |
4186 | struct ocfs2_xattr_search *xs, | |
4187 | struct buffer_head *old_bh) | |
01225596 | 4188 | { |
01225596 TM |
4189 | char *buf = old_bh->b_data; |
4190 | struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf; | |
4191 | struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header; | |
178eeac3 | 4192 | int i; |
01225596 | 4193 | |
ba937127 | 4194 | xs->header = bucket_xh(xs->bucket); |
178eeac3 | 4195 | xs->base = bucket_block(xs->bucket, 0); |
01225596 TM |
4196 | xs->end = xs->base + inode->i_sb->s_blocksize; |
4197 | ||
178eeac3 JB |
4198 | if (xs->not_found) |
4199 | return; | |
01225596 | 4200 | |
178eeac3 JB |
4201 | i = xs->here - old_xh->xh_entries; |
4202 | xs->here = &xs->header->xh_entries[i]; | |
01225596 TM |
4203 | } |
4204 | ||
4205 | static int ocfs2_xattr_create_index_block(struct inode *inode, | |
78f30c31 TM |
4206 | struct ocfs2_xattr_search *xs, |
4207 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 4208 | { |
85db90e7 | 4209 | int ret; |
01225596 TM |
4210 | u32 bit_off, len; |
4211 | u64 blkno; | |
85db90e7 | 4212 | handle_t *handle = ctxt->handle; |
01225596 TM |
4213 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
4214 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
01225596 TM |
4215 | struct buffer_head *xb_bh = xs->xattr_bh; |
4216 | struct ocfs2_xattr_block *xb = | |
4217 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4218 | struct ocfs2_xattr_tree_root *xr; | |
4219 | u16 xb_flags = le16_to_cpu(xb->xb_flags); | |
01225596 TM |
4220 | |
4221 | mlog(0, "create xattr index block for %llu\n", | |
4222 | (unsigned long long)xb_bh->b_blocknr); | |
4223 | ||
4224 | BUG_ON(xb_flags & OCFS2_XATTR_INDEXED); | |
178eeac3 | 4225 | BUG_ON(!xs->bucket); |
01225596 | 4226 | |
01225596 TM |
4227 | /* |
4228 | * XXX: | |
4229 | * We can use this lock for now, and maybe move to a dedicated mutex | |
4230 | * if performance becomes a problem later. | |
4231 | */ | |
4232 | down_write(&oi->ip_alloc_sem); | |
4233 | ||
0cf2f763 | 4234 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh, |
84008972 | 4235 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4236 | if (ret) { |
4237 | mlog_errno(ret); | |
85db90e7 | 4238 | goto out; |
01225596 TM |
4239 | } |
4240 | ||
78f30c31 TM |
4241 | ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, |
4242 | 1, 1, &bit_off, &len); | |
01225596 TM |
4243 | if (ret) { |
4244 | mlog_errno(ret); | |
85db90e7 | 4245 | goto out; |
01225596 TM |
4246 | } |
4247 | ||
4248 | /* | |
4249 | * The bucket may spread in many blocks, and | |
4250 | * we will only touch the 1st block and the last block | |
4251 | * in the whole bucket(one for entry and one for data). | |
4252 | */ | |
4253 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); | |
4254 | ||
de29c085 MF |
4255 | mlog(0, "allocate 1 cluster from %llu to xattr block\n", |
4256 | (unsigned long long)blkno); | |
01225596 | 4257 | |
178eeac3 JB |
4258 | ret = ocfs2_init_xattr_bucket(xs->bucket, blkno); |
4259 | if (ret) { | |
01225596 | 4260 | mlog_errno(ret); |
85db90e7 | 4261 | goto out; |
01225596 TM |
4262 | } |
4263 | ||
178eeac3 JB |
4264 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
4265 | OCFS2_JOURNAL_ACCESS_CREATE); | |
01225596 TM |
4266 | if (ret) { |
4267 | mlog_errno(ret); | |
85db90e7 | 4268 | goto out; |
01225596 TM |
4269 | } |
4270 | ||
178eeac3 JB |
4271 | ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket); |
4272 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); | |
01225596 | 4273 | |
178eeac3 | 4274 | ocfs2_xattr_update_xattr_search(inode, xs, xb_bh); |
01225596 TM |
4275 | |
4276 | /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */ | |
4277 | memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize - | |
4278 | offsetof(struct ocfs2_xattr_block, xb_attrs)); | |
4279 | ||
4280 | xr = &xb->xb_attrs.xb_root; | |
4281 | xr->xt_clusters = cpu_to_le32(1); | |
4282 | xr->xt_last_eb_blk = 0; | |
4283 | xr->xt_list.l_tree_depth = 0; | |
4284 | xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
4285 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
4286 | ||
4287 | xr->xt_list.l_recs[0].e_cpos = 0; | |
4288 | xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno); | |
4289 | xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1); | |
4290 | ||
4291 | xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED); | |
4292 | ||
85db90e7 | 4293 | ocfs2_journal_dirty(handle, xb_bh); |
01225596 | 4294 | |
85db90e7 | 4295 | out: |
01225596 TM |
4296 | up_write(&oi->ip_alloc_sem); |
4297 | ||
01225596 TM |
4298 | return ret; |
4299 | } | |
4300 | ||
4301 | static int cmp_xe_offset(const void *a, const void *b) | |
4302 | { | |
4303 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4304 | u32 l_name_offset = le16_to_cpu(l->xe_name_offset); | |
4305 | u32 r_name_offset = le16_to_cpu(r->xe_name_offset); | |
4306 | ||
4307 | if (l_name_offset < r_name_offset) | |
4308 | return 1; | |
4309 | if (l_name_offset > r_name_offset) | |
4310 | return -1; | |
4311 | return 0; | |
4312 | } | |
4313 | ||
4314 | /* | |
4315 | * defrag a xattr bucket if we find that the bucket has some | |
4316 | * holes beteen name/value pairs. | |
4317 | * We will move all the name/value pairs to the end of the bucket | |
4318 | * so that we can spare some space for insertion. | |
4319 | */ | |
4320 | static int ocfs2_defrag_xattr_bucket(struct inode *inode, | |
85db90e7 | 4321 | handle_t *handle, |
01225596 TM |
4322 | struct ocfs2_xattr_bucket *bucket) |
4323 | { | |
4324 | int ret, i; | |
199799a3 | 4325 | size_t end, offset, len; |
01225596 TM |
4326 | struct ocfs2_xattr_header *xh; |
4327 | char *entries, *buf, *bucket_buf = NULL; | |
9c7759aa | 4328 | u64 blkno = bucket_blkno(bucket); |
01225596 | 4329 | u16 xh_free_start; |
01225596 | 4330 | size_t blocksize = inode->i_sb->s_blocksize; |
01225596 | 4331 | struct ocfs2_xattr_entry *xe; |
01225596 TM |
4332 | |
4333 | /* | |
4334 | * In order to make the operation more efficient and generic, | |
4335 | * we copy all the blocks into a contiguous memory and do the | |
4336 | * defragment there, so if anything is error, we will not touch | |
4337 | * the real block. | |
4338 | */ | |
4339 | bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS); | |
4340 | if (!bucket_buf) { | |
4341 | ret = -EIO; | |
4342 | goto out; | |
4343 | } | |
4344 | ||
4345 | buf = bucket_buf; | |
1c32a2fd TM |
4346 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4347 | memcpy(buf, bucket_block(bucket, i), blocksize); | |
01225596 | 4348 | |
1c32a2fd | 4349 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, |
161d6f30 JB |
4350 | OCFS2_JOURNAL_ACCESS_WRITE); |
4351 | if (ret < 0) { | |
4352 | mlog_errno(ret); | |
85db90e7 | 4353 | goto out; |
01225596 TM |
4354 | } |
4355 | ||
4356 | xh = (struct ocfs2_xattr_header *)bucket_buf; | |
4357 | entries = (char *)xh->xh_entries; | |
4358 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4359 | ||
4360 | mlog(0, "adjust xattr bucket in %llu, count = %u, " | |
4361 | "xh_free_start = %u, xh_name_value_len = %u.\n", | |
de29c085 MF |
4362 | (unsigned long long)blkno, le16_to_cpu(xh->xh_count), |
4363 | xh_free_start, le16_to_cpu(xh->xh_name_value_len)); | |
01225596 TM |
4364 | |
4365 | /* | |
4366 | * sort all the entries by their offset. | |
4367 | * the largest will be the first, so that we can | |
4368 | * move them to the end one by one. | |
4369 | */ | |
4370 | sort(entries, le16_to_cpu(xh->xh_count), | |
4371 | sizeof(struct ocfs2_xattr_entry), | |
4372 | cmp_xe_offset, swap_xe); | |
4373 | ||
4374 | /* Move all name/values to the end of the bucket. */ | |
4375 | xe = xh->xh_entries; | |
4376 | end = OCFS2_XATTR_BUCKET_SIZE; | |
4377 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) { | |
4378 | offset = le16_to_cpu(xe->xe_name_offset); | |
199799a3 | 4379 | len = namevalue_size_xe(xe); |
01225596 TM |
4380 | |
4381 | /* | |
4382 | * We must make sure that the name/value pair | |
4383 | * exist in the same block. So adjust end to | |
4384 | * the previous block end if needed. | |
4385 | */ | |
4386 | if (((end - len) / blocksize != | |
4387 | (end - 1) / blocksize)) | |
4388 | end = end - end % blocksize; | |
4389 | ||
4390 | if (end > offset + len) { | |
4391 | memmove(bucket_buf + end - len, | |
4392 | bucket_buf + offset, len); | |
4393 | xe->xe_name_offset = cpu_to_le16(end - len); | |
4394 | } | |
4395 | ||
4396 | mlog_bug_on_msg(end < offset + len, "Defrag check failed for " | |
4397 | "bucket %llu\n", (unsigned long long)blkno); | |
4398 | ||
4399 | end -= len; | |
4400 | } | |
4401 | ||
4402 | mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for " | |
4403 | "bucket %llu\n", (unsigned long long)blkno); | |
4404 | ||
4405 | if (xh_free_start == end) | |
85db90e7 | 4406 | goto out; |
01225596 TM |
4407 | |
4408 | memset(bucket_buf + xh_free_start, 0, end - xh_free_start); | |
4409 | xh->xh_free_start = cpu_to_le16(end); | |
4410 | ||
4411 | /* sort the entries by their name_hash. */ | |
4412 | sort(entries, le16_to_cpu(xh->xh_count), | |
4413 | sizeof(struct ocfs2_xattr_entry), | |
4414 | cmp_xe, swap_xe); | |
4415 | ||
4416 | buf = bucket_buf; | |
1c32a2fd TM |
4417 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4418 | memcpy(bucket_block(bucket, i), buf, blocksize); | |
4419 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
01225596 | 4420 | |
01225596 | 4421 | out: |
01225596 TM |
4422 | kfree(bucket_buf); |
4423 | return ret; | |
4424 | } | |
4425 | ||
4426 | /* | |
b5c03e74 JB |
4427 | * prev_blkno points to the start of an existing extent. new_blkno |
4428 | * points to a newly allocated extent. Because we know each of our | |
4429 | * clusters contains more than bucket, we can easily split one cluster | |
4430 | * at a bucket boundary. So we take the last cluster of the existing | |
4431 | * extent and split it down the middle. We move the last half of the | |
4432 | * buckets in the last cluster of the existing extent over to the new | |
4433 | * extent. | |
01225596 | 4434 | * |
b5c03e74 JB |
4435 | * first_bh is the buffer at prev_blkno so we can update the existing |
4436 | * extent's bucket count. header_bh is the bucket were we were hoping | |
4437 | * to insert our xattr. If the bucket move places the target in the new | |
4438 | * extent, we'll update first_bh and header_bh after modifying the old | |
4439 | * extent. | |
4440 | * | |
4441 | * first_hash will be set as the 1st xe's name_hash in the new extent. | |
01225596 TM |
4442 | */ |
4443 | static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode, | |
4444 | handle_t *handle, | |
41cb8148 JB |
4445 | struct ocfs2_xattr_bucket *first, |
4446 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4447 | u64 new_blkno, |
01225596 TM |
4448 | u32 num_clusters, |
4449 | u32 *first_hash) | |
4450 | { | |
c58b6032 | 4451 | int ret; |
41cb8148 JB |
4452 | struct super_block *sb = inode->i_sb; |
4453 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb); | |
4454 | int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
b5c03e74 | 4455 | int to_move = num_buckets / 2; |
c58b6032 | 4456 | u64 src_blkno; |
41cb8148 JB |
4457 | u64 last_cluster_blkno = bucket_blkno(first) + |
4458 | ((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1)); | |
01225596 | 4459 | |
41cb8148 JB |
4460 | BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets); |
4461 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize); | |
01225596 | 4462 | |
01225596 | 4463 | mlog(0, "move half of xattrs in cluster %llu to %llu\n", |
c58b6032 | 4464 | (unsigned long long)last_cluster_blkno, (unsigned long long)new_blkno); |
01225596 | 4465 | |
41cb8148 | 4466 | ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first), |
c58b6032 JB |
4467 | last_cluster_blkno, new_blkno, |
4468 | to_move, first_hash); | |
01225596 TM |
4469 | if (ret) { |
4470 | mlog_errno(ret); | |
4471 | goto out; | |
4472 | } | |
4473 | ||
c58b6032 JB |
4474 | /* This is the first bucket that got moved */ |
4475 | src_blkno = last_cluster_blkno + (to_move * blks_per_bucket); | |
01225596 | 4476 | |
b5c03e74 | 4477 | /* |
c58b6032 | 4478 | * If the target bucket was part of the moved buckets, we need to |
41cb8148 | 4479 | * update first and target. |
b5c03e74 | 4480 | */ |
41cb8148 | 4481 | if (bucket_blkno(target) >= src_blkno) { |
b5c03e74 JB |
4482 | /* Find the block for the new target bucket */ |
4483 | src_blkno = new_blkno + | |
41cb8148 JB |
4484 | (bucket_blkno(target) - src_blkno); |
4485 | ||
4486 | ocfs2_xattr_bucket_relse(first); | |
4487 | ocfs2_xattr_bucket_relse(target); | |
01225596 | 4488 | |
b5c03e74 | 4489 | /* |
c58b6032 | 4490 | * These shouldn't fail - the buffers are in the |
b5c03e74 JB |
4491 | * journal from ocfs2_cp_xattr_bucket(). |
4492 | */ | |
41cb8148 | 4493 | ret = ocfs2_read_xattr_bucket(first, new_blkno); |
c58b6032 JB |
4494 | if (ret) { |
4495 | mlog_errno(ret); | |
4496 | goto out; | |
4497 | } | |
41cb8148 JB |
4498 | ret = ocfs2_read_xattr_bucket(target, src_blkno); |
4499 | if (ret) | |
b5c03e74 | 4500 | mlog_errno(ret); |
01225596 | 4501 | |
01225596 TM |
4502 | } |
4503 | ||
01225596 | 4504 | out: |
01225596 TM |
4505 | return ret; |
4506 | } | |
4507 | ||
01225596 | 4508 | /* |
80bcaf34 TM |
4509 | * Find the suitable pos when we divide a bucket into 2. |
4510 | * We have to make sure the xattrs with the same hash value exist | |
4511 | * in the same bucket. | |
4512 | * | |
4513 | * If this ocfs2_xattr_header covers more than one hash value, find a | |
4514 | * place where the hash value changes. Try to find the most even split. | |
4515 | * The most common case is that all entries have different hash values, | |
4516 | * and the first check we make will find a place to split. | |
4517 | */ | |
4518 | static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh) | |
4519 | { | |
4520 | struct ocfs2_xattr_entry *entries = xh->xh_entries; | |
4521 | int count = le16_to_cpu(xh->xh_count); | |
4522 | int delta, middle = count / 2; | |
4523 | ||
4524 | /* | |
4525 | * We start at the middle. Each step gets farther away in both | |
4526 | * directions. We therefore hit the change in hash value | |
4527 | * nearest to the middle. Note that this loop does not execute for | |
4528 | * count < 2. | |
4529 | */ | |
4530 | for (delta = 0; delta < middle; delta++) { | |
4531 | /* Let's check delta earlier than middle */ | |
4532 | if (cmp_xe(&entries[middle - delta - 1], | |
4533 | &entries[middle - delta])) | |
4534 | return middle - delta; | |
4535 | ||
4536 | /* For even counts, don't walk off the end */ | |
4537 | if ((middle + delta + 1) == count) | |
4538 | continue; | |
4539 | ||
4540 | /* Now try delta past middle */ | |
4541 | if (cmp_xe(&entries[middle + delta], | |
4542 | &entries[middle + delta + 1])) | |
4543 | return middle + delta + 1; | |
4544 | } | |
4545 | ||
4546 | /* Every entry had the same hash */ | |
4547 | return count; | |
4548 | } | |
4549 | ||
4550 | /* | |
4551 | * Move some xattrs in old bucket(blk) to new bucket(new_blk). | |
01225596 | 4552 | * first_hash will record the 1st hash of the new bucket. |
80bcaf34 TM |
4553 | * |
4554 | * Normally half of the xattrs will be moved. But we have to make | |
4555 | * sure that the xattrs with the same hash value are stored in the | |
4556 | * same bucket. If all the xattrs in this bucket have the same hash | |
4557 | * value, the new bucket will be initialized as an empty one and the | |
4558 | * first_hash will be initialized as (hash_value+1). | |
01225596 | 4559 | */ |
80bcaf34 TM |
4560 | static int ocfs2_divide_xattr_bucket(struct inode *inode, |
4561 | handle_t *handle, | |
4562 | u64 blk, | |
4563 | u64 new_blk, | |
4564 | u32 *first_hash, | |
4565 | int new_bucket_head) | |
01225596 TM |
4566 | { |
4567 | int ret, i; | |
199799a3 | 4568 | int count, start, len, name_value_len = 0, name_offset = 0; |
ba937127 | 4569 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4570 | struct ocfs2_xattr_header *xh; |
4571 | struct ocfs2_xattr_entry *xe; | |
4572 | int blocksize = inode->i_sb->s_blocksize; | |
4573 | ||
80bcaf34 | 4574 | mlog(0, "move some of xattrs from bucket %llu to %llu\n", |
de29c085 | 4575 | (unsigned long long)blk, (unsigned long long)new_blk); |
01225596 | 4576 | |
ba937127 JB |
4577 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4578 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4579 | if (!s_bucket || !t_bucket) { | |
4580 | ret = -ENOMEM; | |
4581 | mlog_errno(ret); | |
4582 | goto out; | |
4583 | } | |
01225596 | 4584 | |
ba937127 | 4585 | ret = ocfs2_read_xattr_bucket(s_bucket, blk); |
01225596 TM |
4586 | if (ret) { |
4587 | mlog_errno(ret); | |
4588 | goto out; | |
4589 | } | |
4590 | ||
ba937127 | 4591 | ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket, |
1224be02 | 4592 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4593 | if (ret) { |
4594 | mlog_errno(ret); | |
4595 | goto out; | |
4596 | } | |
4597 | ||
784b816a JB |
4598 | /* |
4599 | * Even if !new_bucket_head, we're overwriting t_bucket. Thus, | |
4600 | * there's no need to read it. | |
4601 | */ | |
ba937127 | 4602 | ret = ocfs2_init_xattr_bucket(t_bucket, new_blk); |
01225596 TM |
4603 | if (ret) { |
4604 | mlog_errno(ret); | |
4605 | goto out; | |
4606 | } | |
4607 | ||
2b656c1d JB |
4608 | /* |
4609 | * Hey, if we're overwriting t_bucket, what difference does | |
4610 | * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the | |
4611 | * same part of ocfs2_cp_xattr_bucket(). | |
4612 | */ | |
ba937127 | 4613 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4614 | new_bucket_head ? |
4615 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4616 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4617 | if (ret) { | |
4618 | mlog_errno(ret); | |
4619 | goto out; | |
01225596 TM |
4620 | } |
4621 | ||
ba937127 | 4622 | xh = bucket_xh(s_bucket); |
80bcaf34 TM |
4623 | count = le16_to_cpu(xh->xh_count); |
4624 | start = ocfs2_xattr_find_divide_pos(xh); | |
4625 | ||
4626 | if (start == count) { | |
4627 | xe = &xh->xh_entries[start-1]; | |
4628 | ||
4629 | /* | |
4630 | * initialized a new empty bucket here. | |
4631 | * The hash value is set as one larger than | |
4632 | * that of the last entry in the previous bucket. | |
4633 | */ | |
ba937127 JB |
4634 | for (i = 0; i < t_bucket->bu_blocks; i++) |
4635 | memset(bucket_block(t_bucket, i), 0, blocksize); | |
80bcaf34 | 4636 | |
ba937127 | 4637 | xh = bucket_xh(t_bucket); |
80bcaf34 TM |
4638 | xh->xh_free_start = cpu_to_le16(blocksize); |
4639 | xh->xh_entries[0].xe_name_hash = xe->xe_name_hash; | |
4640 | le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1); | |
4641 | ||
4642 | goto set_num_buckets; | |
4643 | } | |
4644 | ||
01225596 | 4645 | /* copy the whole bucket to the new first. */ |
ba937127 | 4646 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
01225596 TM |
4647 | |
4648 | /* update the new bucket. */ | |
ba937127 | 4649 | xh = bucket_xh(t_bucket); |
01225596 TM |
4650 | |
4651 | /* | |
4652 | * Calculate the total name/value len and xh_free_start for | |
4653 | * the old bucket first. | |
4654 | */ | |
4655 | name_offset = OCFS2_XATTR_BUCKET_SIZE; | |
4656 | name_value_len = 0; | |
4657 | for (i = 0; i < start; i++) { | |
4658 | xe = &xh->xh_entries[i]; | |
199799a3 | 4659 | name_value_len += namevalue_size_xe(xe); |
01225596 TM |
4660 | if (le16_to_cpu(xe->xe_name_offset) < name_offset) |
4661 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
4662 | } | |
4663 | ||
4664 | /* | |
4665 | * Now begin the modification to the new bucket. | |
4666 | * | |
4667 | * In the new bucket, We just move the xattr entry to the beginning | |
4668 | * and don't touch the name/value. So there will be some holes in the | |
4669 | * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is | |
4670 | * called. | |
4671 | */ | |
4672 | xe = &xh->xh_entries[start]; | |
4673 | len = sizeof(struct ocfs2_xattr_entry) * (count - start); | |
4674 | mlog(0, "mv xattr entry len %d from %d to %d\n", len, | |
ff1ec20e MF |
4675 | (int)((char *)xe - (char *)xh), |
4676 | (int)((char *)xh->xh_entries - (char *)xh)); | |
01225596 TM |
4677 | memmove((char *)xh->xh_entries, (char *)xe, len); |
4678 | xe = &xh->xh_entries[count - start]; | |
4679 | len = sizeof(struct ocfs2_xattr_entry) * start; | |
4680 | memset((char *)xe, 0, len); | |
4681 | ||
4682 | le16_add_cpu(&xh->xh_count, -start); | |
4683 | le16_add_cpu(&xh->xh_name_value_len, -name_value_len); | |
4684 | ||
4685 | /* Calculate xh_free_start for the new bucket. */ | |
4686 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); | |
4687 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
4688 | xe = &xh->xh_entries[i]; | |
01225596 TM |
4689 | if (le16_to_cpu(xe->xe_name_offset) < |
4690 | le16_to_cpu(xh->xh_free_start)) | |
4691 | xh->xh_free_start = xe->xe_name_offset; | |
4692 | } | |
4693 | ||
80bcaf34 | 4694 | set_num_buckets: |
01225596 TM |
4695 | /* set xh->xh_num_buckets for the new xh. */ |
4696 | if (new_bucket_head) | |
4697 | xh->xh_num_buckets = cpu_to_le16(1); | |
4698 | else | |
4699 | xh->xh_num_buckets = 0; | |
4700 | ||
ba937127 | 4701 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); |
01225596 TM |
4702 | |
4703 | /* store the first_hash of the new bucket. */ | |
4704 | if (first_hash) | |
4705 | *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash); | |
4706 | ||
4707 | /* | |
80bcaf34 TM |
4708 | * Now only update the 1st block of the old bucket. If we |
4709 | * just added a new empty bucket, there is no need to modify | |
4710 | * it. | |
01225596 | 4711 | */ |
80bcaf34 TM |
4712 | if (start == count) |
4713 | goto out; | |
4714 | ||
ba937127 | 4715 | xh = bucket_xh(s_bucket); |
01225596 TM |
4716 | memset(&xh->xh_entries[start], 0, |
4717 | sizeof(struct ocfs2_xattr_entry) * (count - start)); | |
4718 | xh->xh_count = cpu_to_le16(start); | |
4719 | xh->xh_free_start = cpu_to_le16(name_offset); | |
4720 | xh->xh_name_value_len = cpu_to_le16(name_value_len); | |
4721 | ||
ba937127 | 4722 | ocfs2_xattr_bucket_journal_dirty(handle, s_bucket); |
01225596 TM |
4723 | |
4724 | out: | |
ba937127 JB |
4725 | ocfs2_xattr_bucket_free(s_bucket); |
4726 | ocfs2_xattr_bucket_free(t_bucket); | |
01225596 TM |
4727 | |
4728 | return ret; | |
4729 | } | |
4730 | ||
4731 | /* | |
4732 | * Copy xattr from one bucket to another bucket. | |
4733 | * | |
4734 | * The caller must make sure that the journal transaction | |
4735 | * has enough space for journaling. | |
4736 | */ | |
4737 | static int ocfs2_cp_xattr_bucket(struct inode *inode, | |
4738 | handle_t *handle, | |
4739 | u64 s_blkno, | |
4740 | u64 t_blkno, | |
4741 | int t_is_new) | |
4742 | { | |
4980c6da | 4743 | int ret; |
ba937127 | 4744 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4745 | |
4746 | BUG_ON(s_blkno == t_blkno); | |
4747 | ||
4748 | mlog(0, "cp bucket %llu to %llu, target is %d\n", | |
de29c085 MF |
4749 | (unsigned long long)s_blkno, (unsigned long long)t_blkno, |
4750 | t_is_new); | |
01225596 | 4751 | |
ba937127 JB |
4752 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4753 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4754 | if (!s_bucket || !t_bucket) { | |
4755 | ret = -ENOMEM; | |
4756 | mlog_errno(ret); | |
4757 | goto out; | |
4758 | } | |
92de109a | 4759 | |
ba937127 | 4760 | ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno); |
01225596 TM |
4761 | if (ret) |
4762 | goto out; | |
4763 | ||
784b816a JB |
4764 | /* |
4765 | * Even if !t_is_new, we're overwriting t_bucket. Thus, | |
4766 | * there's no need to read it. | |
4767 | */ | |
ba937127 | 4768 | ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno); |
01225596 TM |
4769 | if (ret) |
4770 | goto out; | |
4771 | ||
2b656c1d JB |
4772 | /* |
4773 | * Hey, if we're overwriting t_bucket, what difference does | |
4774 | * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new | |
874d65af JB |
4775 | * cluster to fill, we came here from |
4776 | * ocfs2_mv_xattr_buckets(), and it is really new - | |
4777 | * ACCESS_CREATE is required. But we also might have moved data | |
4778 | * out of t_bucket before extending back into it. | |
4779 | * ocfs2_add_new_xattr_bucket() can do this - its call to | |
4780 | * ocfs2_add_new_xattr_cluster() may have created a new extent | |
2b656c1d JB |
4781 | * and copied out the end of the old extent. Then it re-extends |
4782 | * the old extent back to create space for new xattrs. That's | |
4783 | * how we get here, and the bucket isn't really new. | |
4784 | */ | |
ba937127 | 4785 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4786 | t_is_new ? |
4787 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4788 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4789 | if (ret) | |
4790 | goto out; | |
01225596 | 4791 | |
ba937127 JB |
4792 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
4793 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); | |
01225596 TM |
4794 | |
4795 | out: | |
ba937127 JB |
4796 | ocfs2_xattr_bucket_free(t_bucket); |
4797 | ocfs2_xattr_bucket_free(s_bucket); | |
01225596 TM |
4798 | |
4799 | return ret; | |
4800 | } | |
4801 | ||
4802 | /* | |
874d65af JB |
4803 | * src_blk points to the start of an existing extent. last_blk points to |
4804 | * last cluster in that extent. to_blk points to a newly allocated | |
54ecb6b6 JB |
4805 | * extent. We copy the buckets from the cluster at last_blk to the new |
4806 | * extent. If start_bucket is non-zero, we skip that many buckets before | |
4807 | * we start copying. The new extent's xh_num_buckets gets set to the | |
4808 | * number of buckets we copied. The old extent's xh_num_buckets shrinks | |
4809 | * by the same amount. | |
01225596 | 4810 | */ |
54ecb6b6 JB |
4811 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
4812 | u64 src_blk, u64 last_blk, u64 to_blk, | |
4813 | unsigned int start_bucket, | |
4814 | u32 *first_hash) | |
01225596 TM |
4815 | { |
4816 | int i, ret, credits; | |
4817 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
15d60929 | 4818 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 | 4819 | int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); |
15d60929 | 4820 | struct ocfs2_xattr_bucket *old_first, *new_first; |
01225596 | 4821 | |
874d65af JB |
4822 | mlog(0, "mv xattrs from cluster %llu to %llu\n", |
4823 | (unsigned long long)last_blk, (unsigned long long)to_blk); | |
01225596 | 4824 | |
54ecb6b6 JB |
4825 | BUG_ON(start_bucket >= num_buckets); |
4826 | if (start_bucket) { | |
4827 | num_buckets -= start_bucket; | |
4828 | last_blk += (start_bucket * blks_per_bucket); | |
4829 | } | |
4830 | ||
15d60929 JB |
4831 | /* The first bucket of the original extent */ |
4832 | old_first = ocfs2_xattr_bucket_new(inode); | |
4833 | /* The first bucket of the new extent */ | |
4834 | new_first = ocfs2_xattr_bucket_new(inode); | |
4835 | if (!old_first || !new_first) { | |
4836 | ret = -ENOMEM; | |
4837 | mlog_errno(ret); | |
4838 | goto out; | |
4839 | } | |
4840 | ||
874d65af | 4841 | ret = ocfs2_read_xattr_bucket(old_first, src_blk); |
15d60929 JB |
4842 | if (ret) { |
4843 | mlog_errno(ret); | |
4844 | goto out; | |
4845 | } | |
4846 | ||
01225596 | 4847 | /* |
54ecb6b6 JB |
4848 | * We need to update the first bucket of the old extent and all |
4849 | * the buckets going to the new extent. | |
01225596 | 4850 | */ |
54ecb6b6 JB |
4851 | credits = ((num_buckets + 1) * blks_per_bucket) + |
4852 | handle->h_buffer_credits; | |
01225596 TM |
4853 | ret = ocfs2_extend_trans(handle, credits); |
4854 | if (ret) { | |
4855 | mlog_errno(ret); | |
4856 | goto out; | |
4857 | } | |
4858 | ||
15d60929 JB |
4859 | ret = ocfs2_xattr_bucket_journal_access(handle, old_first, |
4860 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4861 | if (ret) { |
4862 | mlog_errno(ret); | |
4863 | goto out; | |
4864 | } | |
4865 | ||
4866 | for (i = 0; i < num_buckets; i++) { | |
4867 | ret = ocfs2_cp_xattr_bucket(inode, handle, | |
874d65af | 4868 | last_blk + (i * blks_per_bucket), |
15d60929 JB |
4869 | to_blk + (i * blks_per_bucket), |
4870 | 1); | |
01225596 TM |
4871 | if (ret) { |
4872 | mlog_errno(ret); | |
4873 | goto out; | |
4874 | } | |
01225596 TM |
4875 | } |
4876 | ||
15d60929 JB |
4877 | /* |
4878 | * Get the new bucket ready before we dirty anything | |
4879 | * (This actually shouldn't fail, because we already dirtied | |
4880 | * it once in ocfs2_cp_xattr_bucket()). | |
4881 | */ | |
4882 | ret = ocfs2_read_xattr_bucket(new_first, to_blk); | |
4883 | if (ret) { | |
01225596 TM |
4884 | mlog_errno(ret); |
4885 | goto out; | |
4886 | } | |
15d60929 JB |
4887 | ret = ocfs2_xattr_bucket_journal_access(handle, new_first, |
4888 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4889 | if (ret) { |
4890 | mlog_errno(ret); | |
4891 | goto out; | |
4892 | } | |
4893 | ||
15d60929 JB |
4894 | /* Now update the headers */ |
4895 | le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets); | |
4896 | ocfs2_xattr_bucket_journal_dirty(handle, old_first); | |
01225596 | 4897 | |
15d60929 JB |
4898 | bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(num_buckets); |
4899 | ocfs2_xattr_bucket_journal_dirty(handle, new_first); | |
01225596 TM |
4900 | |
4901 | if (first_hash) | |
15d60929 JB |
4902 | *first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash); |
4903 | ||
01225596 | 4904 | out: |
15d60929 JB |
4905 | ocfs2_xattr_bucket_free(new_first); |
4906 | ocfs2_xattr_bucket_free(old_first); | |
01225596 TM |
4907 | return ret; |
4908 | } | |
4909 | ||
4910 | /* | |
80bcaf34 | 4911 | * Move some xattrs in this cluster to the new cluster. |
01225596 TM |
4912 | * This function should only be called when bucket size == cluster size. |
4913 | * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead. | |
4914 | */ | |
80bcaf34 TM |
4915 | static int ocfs2_divide_xattr_cluster(struct inode *inode, |
4916 | handle_t *handle, | |
4917 | u64 prev_blk, | |
4918 | u64 new_blk, | |
4919 | u32 *first_hash) | |
01225596 TM |
4920 | { |
4921 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
85db90e7 | 4922 | int ret, credits = 2 * blk_per_bucket + handle->h_buffer_credits; |
01225596 TM |
4923 | |
4924 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize); | |
4925 | ||
4926 | ret = ocfs2_extend_trans(handle, credits); | |
4927 | if (ret) { | |
4928 | mlog_errno(ret); | |
4929 | return ret; | |
4930 | } | |
4931 | ||
4932 | /* Move half of the xattr in start_blk to the next bucket. */ | |
80bcaf34 TM |
4933 | return ocfs2_divide_xattr_bucket(inode, handle, prev_blk, |
4934 | new_blk, first_hash, 1); | |
01225596 TM |
4935 | } |
4936 | ||
4937 | /* | |
4938 | * Move some xattrs from the old cluster to the new one since they are not | |
4939 | * contiguous in ocfs2 xattr tree. | |
4940 | * | |
4941 | * new_blk starts a new separate cluster, and we will move some xattrs from | |
4942 | * prev_blk to it. v_start will be set as the first name hash value in this | |
4943 | * new cluster so that it can be used as e_cpos during tree insertion and | |
4944 | * don't collide with our original b-tree operations. first_bh and header_bh | |
4945 | * will also be updated since they will be used in ocfs2_extend_xattr_bucket | |
4946 | * to extend the insert bucket. | |
4947 | * | |
4948 | * The problem is how much xattr should we move to the new one and when should | |
4949 | * we update first_bh and header_bh? | |
4950 | * 1. If cluster size > bucket size, that means the previous cluster has more | |
4951 | * than 1 bucket, so just move half nums of bucket into the new cluster and | |
4952 | * update the first_bh and header_bh if the insert bucket has been moved | |
4953 | * to the new cluster. | |
4954 | * 2. If cluster_size == bucket_size: | |
4955 | * a) If the previous extent rec has more than one cluster and the insert | |
4956 | * place isn't in the last cluster, copy the entire last cluster to the | |
4957 | * new one. This time, we don't need to upate the first_bh and header_bh | |
4958 | * since they will not be moved into the new cluster. | |
4959 | * b) Otherwise, move the bottom half of the xattrs in the last cluster into | |
4960 | * the new one. And we set the extend flag to zero if the insert place is | |
4961 | * moved into the new allocated cluster since no extend is needed. | |
4962 | */ | |
4963 | static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode, | |
4964 | handle_t *handle, | |
012ee910 JB |
4965 | struct ocfs2_xattr_bucket *first, |
4966 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4967 | u64 new_blk, |
01225596 TM |
4968 | u32 prev_clusters, |
4969 | u32 *v_start, | |
4970 | int *extend) | |
4971 | { | |
92cf3adf | 4972 | int ret; |
01225596 TM |
4973 | |
4974 | mlog(0, "adjust xattrs from cluster %llu len %u to %llu\n", | |
012ee910 | 4975 | (unsigned long long)bucket_blkno(first), prev_clusters, |
de29c085 | 4976 | (unsigned long long)new_blk); |
01225596 | 4977 | |
41cb8148 | 4978 | if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) { |
01225596 TM |
4979 | ret = ocfs2_mv_xattr_bucket_cross_cluster(inode, |
4980 | handle, | |
41cb8148 | 4981 | first, target, |
01225596 | 4982 | new_blk, |
01225596 TM |
4983 | prev_clusters, |
4984 | v_start); | |
012ee910 | 4985 | if (ret) |
41cb8148 | 4986 | mlog_errno(ret); |
41cb8148 | 4987 | } else { |
92cf3adf JB |
4988 | /* The start of the last cluster in the first extent */ |
4989 | u64 last_blk = bucket_blkno(first) + | |
4990 | ((prev_clusters - 1) * | |
4991 | ocfs2_clusters_to_blocks(inode->i_sb, 1)); | |
01225596 | 4992 | |
012ee910 | 4993 | if (prev_clusters > 1 && bucket_blkno(target) != last_blk) { |
874d65af | 4994 | ret = ocfs2_mv_xattr_buckets(inode, handle, |
92cf3adf | 4995 | bucket_blkno(first), |
54ecb6b6 | 4996 | last_blk, new_blk, 0, |
01225596 | 4997 | v_start); |
012ee910 JB |
4998 | if (ret) |
4999 | mlog_errno(ret); | |
5000 | } else { | |
80bcaf34 TM |
5001 | ret = ocfs2_divide_xattr_cluster(inode, handle, |
5002 | last_blk, new_blk, | |
5003 | v_start); | |
012ee910 JB |
5004 | if (ret) |
5005 | mlog_errno(ret); | |
01225596 | 5006 | |
92cf3adf | 5007 | if ((bucket_blkno(target) == last_blk) && extend) |
01225596 TM |
5008 | *extend = 0; |
5009 | } | |
5010 | } | |
5011 | ||
5012 | return ret; | |
5013 | } | |
5014 | ||
5015 | /* | |
5016 | * Add a new cluster for xattr storage. | |
5017 | * | |
5018 | * If the new cluster is contiguous with the previous one, it will be | |
5019 | * appended to the same extent record, and num_clusters will be updated. | |
5020 | * If not, we will insert a new extent for it and move some xattrs in | |
5021 | * the last cluster into the new allocated one. | |
5022 | * We also need to limit the maximum size of a btree leaf, otherwise we'll | |
5023 | * lose the benefits of hashing because we'll have to search large leaves. | |
5024 | * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize, | |
5025 | * if it's bigger). | |
5026 | * | |
5027 | * first_bh is the first block of the previous extent rec and header_bh | |
5028 | * indicates the bucket we will insert the new xattrs. They will be updated | |
5029 | * when the header_bh is moved into the new cluster. | |
5030 | */ | |
5031 | static int ocfs2_add_new_xattr_cluster(struct inode *inode, | |
5032 | struct buffer_head *root_bh, | |
ed29c0ca JB |
5033 | struct ocfs2_xattr_bucket *first, |
5034 | struct ocfs2_xattr_bucket *target, | |
01225596 TM |
5035 | u32 *num_clusters, |
5036 | u32 prev_cpos, | |
78f30c31 TM |
5037 | int *extend, |
5038 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5039 | { |
85db90e7 | 5040 | int ret; |
01225596 TM |
5041 | u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); |
5042 | u32 prev_clusters = *num_clusters; | |
5043 | u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0; | |
5044 | u64 block; | |
85db90e7 | 5045 | handle_t *handle = ctxt->handle; |
01225596 | 5046 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
f99b9b7c | 5047 | struct ocfs2_extent_tree et; |
01225596 TM |
5048 | |
5049 | mlog(0, "Add new xattr cluster for %llu, previous xattr hash = %u, " | |
5050 | "previous xattr blkno = %llu\n", | |
5051 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
ed29c0ca | 5052 | prev_cpos, (unsigned long long)bucket_blkno(first)); |
01225596 | 5053 | |
5e404e9e | 5054 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
f99b9b7c | 5055 | |
0cf2f763 | 5056 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5057 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5058 | if (ret < 0) { |
5059 | mlog_errno(ret); | |
5060 | goto leave; | |
5061 | } | |
5062 | ||
78f30c31 | 5063 | ret = __ocfs2_claim_clusters(osb, handle, ctxt->data_ac, 1, |
01225596 TM |
5064 | clusters_to_add, &bit_off, &num_bits); |
5065 | if (ret < 0) { | |
5066 | if (ret != -ENOSPC) | |
5067 | mlog_errno(ret); | |
5068 | goto leave; | |
5069 | } | |
5070 | ||
5071 | BUG_ON(num_bits > clusters_to_add); | |
5072 | ||
5073 | block = ocfs2_clusters_to_blocks(osb->sb, bit_off); | |
5074 | mlog(0, "Allocating %u clusters at block %u for xattr in inode %llu\n", | |
5075 | num_bits, bit_off, (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
5076 | ||
ed29c0ca | 5077 | if (bucket_blkno(first) + (prev_clusters * bpc) == block && |
01225596 TM |
5078 | (prev_clusters + num_bits) << osb->s_clustersize_bits <= |
5079 | OCFS2_MAX_XATTR_TREE_LEAF_SIZE) { | |
5080 | /* | |
5081 | * If this cluster is contiguous with the old one and | |
5082 | * adding this new cluster, we don't surpass the limit of | |
5083 | * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be | |
5084 | * initialized and used like other buckets in the previous | |
5085 | * cluster. | |
5086 | * So add it as a contiguous one. The caller will handle | |
5087 | * its init process. | |
5088 | */ | |
5089 | v_start = prev_cpos + prev_clusters; | |
5090 | *num_clusters = prev_clusters + num_bits; | |
5091 | mlog(0, "Add contiguous %u clusters to previous extent rec.\n", | |
5092 | num_bits); | |
5093 | } else { | |
5094 | ret = ocfs2_adjust_xattr_cross_cluster(inode, | |
5095 | handle, | |
012ee910 JB |
5096 | first, |
5097 | target, | |
01225596 | 5098 | block, |
01225596 TM |
5099 | prev_clusters, |
5100 | &v_start, | |
5101 | extend); | |
5102 | if (ret) { | |
5103 | mlog_errno(ret); | |
5104 | goto leave; | |
5105 | } | |
5106 | } | |
5107 | ||
5108 | mlog(0, "Insert %u clusters at block %llu for xattr at %u\n", | |
de29c085 | 5109 | num_bits, (unsigned long long)block, v_start); |
cc79d8c1 | 5110 | ret = ocfs2_insert_extent(handle, &et, v_start, block, |
78f30c31 | 5111 | num_bits, 0, ctxt->meta_ac); |
01225596 TM |
5112 | if (ret < 0) { |
5113 | mlog_errno(ret); | |
5114 | goto leave; | |
5115 | } | |
5116 | ||
5117 | ret = ocfs2_journal_dirty(handle, root_bh); | |
85db90e7 | 5118 | if (ret < 0) |
01225596 | 5119 | mlog_errno(ret); |
01225596 TM |
5120 | |
5121 | leave: | |
01225596 TM |
5122 | return ret; |
5123 | } | |
5124 | ||
5125 | /* | |
92de109a JB |
5126 | * We are given an extent. 'first' is the bucket at the very front of |
5127 | * the extent. The extent has space for an additional bucket past | |
5128 | * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number | |
5129 | * of the target bucket. We wish to shift every bucket past the target | |
5130 | * down one, filling in that additional space. When we get back to the | |
5131 | * target, we split the target between itself and the now-empty bucket | |
5132 | * at target+1 (aka, target_blkno + blks_per_bucket). | |
01225596 TM |
5133 | */ |
5134 | static int ocfs2_extend_xattr_bucket(struct inode *inode, | |
85db90e7 | 5135 | handle_t *handle, |
92de109a JB |
5136 | struct ocfs2_xattr_bucket *first, |
5137 | u64 target_blk, | |
01225596 TM |
5138 | u32 num_clusters) |
5139 | { | |
5140 | int ret, credits; | |
5141 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5142 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
92de109a JB |
5143 | u64 end_blk; |
5144 | u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets); | |
01225596 TM |
5145 | |
5146 | mlog(0, "extend xattr bucket in %llu, xattr extend rec starting " | |
92de109a JB |
5147 | "from %llu, len = %u\n", (unsigned long long)target_blk, |
5148 | (unsigned long long)bucket_blkno(first), num_clusters); | |
01225596 | 5149 | |
92de109a JB |
5150 | /* The extent must have room for an additional bucket */ |
5151 | BUG_ON(new_bucket >= | |
5152 | (num_clusters * ocfs2_xattr_buckets_per_cluster(osb))); | |
01225596 | 5153 | |
92de109a JB |
5154 | /* end_blk points to the last existing bucket */ |
5155 | end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket); | |
01225596 TM |
5156 | |
5157 | /* | |
92de109a JB |
5158 | * end_blk is the start of the last existing bucket. |
5159 | * Thus, (end_blk - target_blk) covers the target bucket and | |
5160 | * every bucket after it up to, but not including, the last | |
5161 | * existing bucket. Then we add the last existing bucket, the | |
5162 | * new bucket, and the first bucket (3 * blk_per_bucket). | |
01225596 | 5163 | */ |
92de109a | 5164 | credits = (end_blk - target_blk) + (3 * blk_per_bucket) + |
85db90e7 TM |
5165 | handle->h_buffer_credits; |
5166 | ret = ocfs2_extend_trans(handle, credits); | |
5167 | if (ret) { | |
01225596 TM |
5168 | mlog_errno(ret); |
5169 | goto out; | |
5170 | } | |
5171 | ||
92de109a JB |
5172 | ret = ocfs2_xattr_bucket_journal_access(handle, first, |
5173 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
5174 | if (ret) { |
5175 | mlog_errno(ret); | |
85db90e7 | 5176 | goto out; |
01225596 TM |
5177 | } |
5178 | ||
92de109a | 5179 | while (end_blk != target_blk) { |
01225596 TM |
5180 | ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk, |
5181 | end_blk + blk_per_bucket, 0); | |
5182 | if (ret) | |
85db90e7 | 5183 | goto out; |
01225596 TM |
5184 | end_blk -= blk_per_bucket; |
5185 | } | |
5186 | ||
92de109a JB |
5187 | /* Move half of the xattr in target_blkno to the next bucket. */ |
5188 | ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk, | |
5189 | target_blk + blk_per_bucket, NULL, 0); | |
01225596 | 5190 | |
92de109a JB |
5191 | le16_add_cpu(&bucket_xh(first)->xh_num_buckets, 1); |
5192 | ocfs2_xattr_bucket_journal_dirty(handle, first); | |
01225596 | 5193 | |
01225596 TM |
5194 | out: |
5195 | return ret; | |
5196 | } | |
5197 | ||
5198 | /* | |
91f2033f JB |
5199 | * Add new xattr bucket in an extent record and adjust the buckets |
5200 | * accordingly. xb_bh is the ocfs2_xattr_block, and target is the | |
5201 | * bucket we want to insert into. | |
01225596 | 5202 | * |
91f2033f JB |
5203 | * In the easy case, we will move all the buckets after target down by |
5204 | * one. Half of target's xattrs will be moved to the next bucket. | |
5205 | * | |
5206 | * If current cluster is full, we'll allocate a new one. This may not | |
5207 | * be contiguous. The underlying calls will make sure that there is | |
5208 | * space for the insert, shifting buckets around if necessary. | |
5209 | * 'target' may be moved by those calls. | |
01225596 TM |
5210 | */ |
5211 | static int ocfs2_add_new_xattr_bucket(struct inode *inode, | |
5212 | struct buffer_head *xb_bh, | |
91f2033f | 5213 | struct ocfs2_xattr_bucket *target, |
78f30c31 | 5214 | struct ocfs2_xattr_set_ctxt *ctxt) |
01225596 | 5215 | { |
01225596 TM |
5216 | struct ocfs2_xattr_block *xb = |
5217 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
5218 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
5219 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
91f2033f JB |
5220 | u32 name_hash = |
5221 | le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash); | |
ed29c0ca | 5222 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
01225596 TM |
5223 | int ret, num_buckets, extend = 1; |
5224 | u64 p_blkno; | |
5225 | u32 e_cpos, num_clusters; | |
92de109a | 5226 | /* The bucket at the front of the extent */ |
91f2033f | 5227 | struct ocfs2_xattr_bucket *first; |
01225596 | 5228 | |
91f2033f JB |
5229 | mlog(0, "Add new xattr bucket starting from %llu\n", |
5230 | (unsigned long long)bucket_blkno(target)); | |
01225596 | 5231 | |
ed29c0ca | 5232 | /* The first bucket of the original extent */ |
92de109a | 5233 | first = ocfs2_xattr_bucket_new(inode); |
91f2033f | 5234 | if (!first) { |
92de109a JB |
5235 | ret = -ENOMEM; |
5236 | mlog_errno(ret); | |
5237 | goto out; | |
5238 | } | |
5239 | ||
01225596 TM |
5240 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos, |
5241 | &num_clusters, el); | |
5242 | if (ret) { | |
5243 | mlog_errno(ret); | |
5244 | goto out; | |
5245 | } | |
5246 | ||
ed29c0ca | 5247 | ret = ocfs2_read_xattr_bucket(first, p_blkno); |
01225596 TM |
5248 | if (ret) { |
5249 | mlog_errno(ret); | |
5250 | goto out; | |
5251 | } | |
5252 | ||
ed29c0ca JB |
5253 | num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters; |
5254 | if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) { | |
5255 | /* | |
5256 | * This can move first+target if the target bucket moves | |
5257 | * to the new extent. | |
5258 | */ | |
01225596 TM |
5259 | ret = ocfs2_add_new_xattr_cluster(inode, |
5260 | xb_bh, | |
ed29c0ca JB |
5261 | first, |
5262 | target, | |
01225596 TM |
5263 | &num_clusters, |
5264 | e_cpos, | |
78f30c31 TM |
5265 | &extend, |
5266 | ctxt); | |
01225596 TM |
5267 | if (ret) { |
5268 | mlog_errno(ret); | |
5269 | goto out; | |
5270 | } | |
5271 | } | |
5272 | ||
92de109a | 5273 | if (extend) { |
01225596 | 5274 | ret = ocfs2_extend_xattr_bucket(inode, |
85db90e7 | 5275 | ctxt->handle, |
ed29c0ca JB |
5276 | first, |
5277 | bucket_blkno(target), | |
01225596 | 5278 | num_clusters); |
92de109a JB |
5279 | if (ret) |
5280 | mlog_errno(ret); | |
5281 | } | |
5282 | ||
01225596 | 5283 | out: |
92de109a | 5284 | ocfs2_xattr_bucket_free(first); |
ed29c0ca | 5285 | |
01225596 TM |
5286 | return ret; |
5287 | } | |
5288 | ||
5289 | static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode, | |
5290 | struct ocfs2_xattr_bucket *bucket, | |
5291 | int offs) | |
5292 | { | |
5293 | int block_off = offs >> inode->i_sb->s_blocksize_bits; | |
5294 | ||
5295 | offs = offs % inode->i_sb->s_blocksize; | |
51def39f | 5296 | return bucket_block(bucket, block_off) + offs; |
01225596 TM |
5297 | } |
5298 | ||
01225596 TM |
5299 | /* |
5300 | * Set the xattr entry in the specified bucket. | |
5301 | * The bucket is indicated by xs->bucket and it should have the enough | |
5302 | * space for the xattr insertion. | |
5303 | */ | |
5304 | static int ocfs2_xattr_set_entry_in_bucket(struct inode *inode, | |
85db90e7 | 5305 | handle_t *handle, |
01225596 TM |
5306 | struct ocfs2_xattr_info *xi, |
5307 | struct ocfs2_xattr_search *xs, | |
9dc47400 | 5308 | u32 name_hash) |
01225596 | 5309 | { |
1224be02 | 5310 | int ret; |
02dbf38d | 5311 | u64 blkno; |
69a3e539 | 5312 | struct ocfs2_xa_loc loc; |
01225596 | 5313 | |
ff1ec20e | 5314 | mlog(0, "Set xattr entry len = %lu index = %d in bucket %llu\n", |
6b240ff6 | 5315 | (unsigned long)xi->xi_value_len, xi->xi_name_index, |
ba937127 | 5316 | (unsigned long long)bucket_blkno(xs->bucket)); |
01225596 | 5317 | |
ba937127 | 5318 | if (!xs->bucket->bu_bhs[1]) { |
02dbf38d JB |
5319 | blkno = bucket_blkno(xs->bucket); |
5320 | ocfs2_xattr_bucket_relse(xs->bucket); | |
5321 | ret = ocfs2_read_xattr_bucket(xs->bucket, blkno); | |
01225596 TM |
5322 | if (ret) { |
5323 | mlog_errno(ret); | |
5324 | goto out; | |
5325 | } | |
5326 | } | |
5327 | ||
ba937127 | 5328 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
1224be02 JB |
5329 | OCFS2_JOURNAL_ACCESS_WRITE); |
5330 | if (ret < 0) { | |
5331 | mlog_errno(ret); | |
5332 | goto out; | |
01225596 TM |
5333 | } |
5334 | ||
69a3e539 JB |
5335 | ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket, |
5336 | xs->not_found ? NULL : xs->here); | |
5337 | ret = ocfs2_xa_prepare_entry(&loc, xi, name_hash); | |
5338 | if (ret) { | |
5339 | if (ret != -ENOSPC) | |
5340 | mlog_errno(ret); | |
5341 | goto out; | |
5342 | } | |
5343 | /* XXX For now, until we make ocfs2_xa_prepare_entry() primary */ | |
5344 | BUG_ON(ret == -ENOSPC); | |
5345 | ocfs2_xa_store_inline_value(&loc, xi); | |
5346 | xs->here = loc.xl_entry; | |
5347 | ||
ba937127 | 5348 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); |
01225596 | 5349 | |
01225596 | 5350 | out: |
01225596 TM |
5351 | return ret; |
5352 | } | |
5353 | ||
01225596 TM |
5354 | /* |
5355 | * Truncate the specified xe_off entry in xattr bucket. | |
5356 | * bucket is indicated by header_bh and len is the new length. | |
5357 | * Both the ocfs2_xattr_value_root and the entry will be updated here. | |
5358 | * | |
5359 | * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed. | |
5360 | */ | |
5361 | static int ocfs2_xattr_bucket_value_truncate(struct inode *inode, | |
548b0f22 | 5362 | struct ocfs2_xattr_bucket *bucket, |
01225596 | 5363 | int xe_off, |
78f30c31 TM |
5364 | int len, |
5365 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5366 | { |
5367 | int ret, offset; | |
5368 | u64 value_blk; | |
01225596 | 5369 | struct ocfs2_xattr_entry *xe; |
548b0f22 | 5370 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 5371 | size_t blocksize = inode->i_sb->s_blocksize; |
b3e5d379 JB |
5372 | struct ocfs2_xattr_value_buf vb = { |
5373 | .vb_access = ocfs2_journal_access, | |
5374 | }; | |
01225596 TM |
5375 | |
5376 | xe = &xh->xh_entries[xe_off]; | |
5377 | ||
5378 | BUG_ON(!xe || ocfs2_xattr_is_local(xe)); | |
5379 | ||
5380 | offset = le16_to_cpu(xe->xe_name_offset) + | |
5381 | OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5382 | ||
5383 | value_blk = offset / blocksize; | |
5384 | ||
5385 | /* We don't allow ocfs2_xattr_value to be stored in different block. */ | |
5386 | BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize); | |
01225596 | 5387 | |
b3e5d379 JB |
5388 | vb.vb_bh = bucket->bu_bhs[value_blk]; |
5389 | BUG_ON(!vb.vb_bh); | |
01225596 | 5390 | |
b3e5d379 JB |
5391 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
5392 | (vb.vb_bh->b_data + offset % blocksize); | |
01225596 | 5393 | |
548b0f22 JB |
5394 | /* |
5395 | * From here on out we have to dirty the bucket. The generic | |
5396 | * value calls only modify one of the bucket's bhs, but we need | |
5397 | * to send the bucket at once. So if they error, they *could* have | |
5398 | * modified something. We have to assume they did, and dirty | |
5399 | * the whole bucket. This leaves us in a consistent state. | |
5400 | */ | |
5401 | mlog(0, "truncate %u in xattr bucket %llu to %d bytes.\n", | |
5402 | xe_off, (unsigned long long)bucket_blkno(bucket), len); | |
b3e5d379 | 5403 | ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); |
01225596 TM |
5404 | if (ret) { |
5405 | mlog_errno(ret); | |
554e7f9e TM |
5406 | goto out; |
5407 | } | |
5408 | ||
5409 | ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket, | |
5410 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5411 | if (ret) { | |
5412 | mlog_errno(ret); | |
5413 | goto out; | |
01225596 TM |
5414 | } |
5415 | ||
548b0f22 JB |
5416 | xe->xe_value_size = cpu_to_le64(len); |
5417 | ||
548b0f22 JB |
5418 | ocfs2_xattr_bucket_journal_dirty(ctxt->handle, bucket); |
5419 | ||
01225596 | 5420 | out: |
01225596 TM |
5421 | return ret; |
5422 | } | |
5423 | ||
5424 | static int ocfs2_xattr_bucket_value_truncate_xs(struct inode *inode, | |
78f30c31 TM |
5425 | struct ocfs2_xattr_search *xs, |
5426 | int len, | |
5427 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5428 | { |
5429 | int ret, offset; | |
5430 | struct ocfs2_xattr_entry *xe = xs->here; | |
5431 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *)xs->base; | |
5432 | ||
ba937127 | 5433 | BUG_ON(!xs->bucket->bu_bhs[0] || !xe || ocfs2_xattr_is_local(xe)); |
01225596 TM |
5434 | |
5435 | offset = xe - xh->xh_entries; | |
548b0f22 | 5436 | ret = ocfs2_xattr_bucket_value_truncate(inode, xs->bucket, |
78f30c31 | 5437 | offset, len, ctxt); |
01225596 TM |
5438 | if (ret) |
5439 | mlog_errno(ret); | |
5440 | ||
5441 | return ret; | |
5442 | } | |
5443 | ||
5444 | static int ocfs2_xattr_bucket_set_value_outside(struct inode *inode, | |
85db90e7 | 5445 | handle_t *handle, |
01225596 TM |
5446 | struct ocfs2_xattr_search *xs, |
5447 | char *val, | |
5448 | int value_len) | |
5449 | { | |
712e53e4 | 5450 | int ret, offset, block_off; |
01225596 TM |
5451 | struct ocfs2_xattr_value_root *xv; |
5452 | struct ocfs2_xattr_entry *xe = xs->here; | |
712e53e4 TM |
5453 | struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); |
5454 | void *base; | |
492a8a33 TM |
5455 | struct ocfs2_xattr_value_buf vb = { |
5456 | .vb_access = ocfs2_journal_access, | |
5457 | }; | |
01225596 TM |
5458 | |
5459 | BUG_ON(!xs->base || !xe || ocfs2_xattr_is_local(xe)); | |
5460 | ||
fd68a894 | 5461 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, xh, |
712e53e4 TM |
5462 | xe - xh->xh_entries, |
5463 | &block_off, | |
5464 | &offset); | |
5465 | if (ret) { | |
5466 | mlog_errno(ret); | |
5467 | goto out; | |
5468 | } | |
01225596 | 5469 | |
712e53e4 TM |
5470 | base = bucket_block(xs->bucket, block_off); |
5471 | xv = (struct ocfs2_xattr_value_root *)(base + offset + | |
5472 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
01225596 | 5473 | |
492a8a33 TM |
5474 | vb.vb_xv = xv; |
5475 | vb.vb_bh = xs->bucket->bu_bhs[block_off]; | |
712e53e4 | 5476 | ret = __ocfs2_xattr_set_value_outside(inode, handle, |
492a8a33 | 5477 | &vb, val, value_len); |
712e53e4 TM |
5478 | if (ret) |
5479 | mlog_errno(ret); | |
5480 | out: | |
5481 | return ret; | |
01225596 TM |
5482 | } |
5483 | ||
01225596 TM |
5484 | static int ocfs2_rm_xattr_cluster(struct inode *inode, |
5485 | struct buffer_head *root_bh, | |
5486 | u64 blkno, | |
5487 | u32 cpos, | |
47bca495 TM |
5488 | u32 len, |
5489 | void *para) | |
01225596 TM |
5490 | { |
5491 | int ret; | |
5492 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5493 | struct inode *tl_inode = osb->osb_tl_inode; | |
5494 | handle_t *handle; | |
5495 | struct ocfs2_xattr_block *xb = | |
5496 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
01225596 TM |
5497 | struct ocfs2_alloc_context *meta_ac = NULL; |
5498 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
f99b9b7c JB |
5499 | struct ocfs2_extent_tree et; |
5500 | ||
47bca495 | 5501 | ret = ocfs2_iterate_xattr_buckets(inode, blkno, len, |
ce9c5a54 | 5502 | ocfs2_delete_xattr_in_bucket, para); |
47bca495 TM |
5503 | if (ret) { |
5504 | mlog_errno(ret); | |
5505 | return ret; | |
5506 | } | |
5507 | ||
5e404e9e | 5508 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
01225596 TM |
5509 | |
5510 | ocfs2_init_dealloc_ctxt(&dealloc); | |
5511 | ||
5512 | mlog(0, "rm xattr extent rec at %u len = %u, start from %llu\n", | |
5513 | cpos, len, (unsigned long long)blkno); | |
5514 | ||
8cb471e8 JB |
5515 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), blkno, |
5516 | len); | |
01225596 | 5517 | |
f99b9b7c | 5518 | ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac); |
01225596 TM |
5519 | if (ret) { |
5520 | mlog_errno(ret); | |
5521 | return ret; | |
5522 | } | |
5523 | ||
5524 | mutex_lock(&tl_inode->i_mutex); | |
5525 | ||
5526 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
5527 | ret = __ocfs2_flush_truncate_log(osb); | |
5528 | if (ret < 0) { | |
5529 | mlog_errno(ret); | |
5530 | goto out; | |
5531 | } | |
5532 | } | |
5533 | ||
a90714c1 | 5534 | handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb)); |
d3264799 | 5535 | if (IS_ERR(handle)) { |
01225596 TM |
5536 | ret = -ENOMEM; |
5537 | mlog_errno(ret); | |
5538 | goto out; | |
5539 | } | |
5540 | ||
0cf2f763 | 5541 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5542 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5543 | if (ret) { |
5544 | mlog_errno(ret); | |
5545 | goto out_commit; | |
5546 | } | |
5547 | ||
dbdcf6a4 | 5548 | ret = ocfs2_remove_extent(handle, &et, cpos, len, meta_ac, |
f99b9b7c | 5549 | &dealloc); |
01225596 TM |
5550 | if (ret) { |
5551 | mlog_errno(ret); | |
5552 | goto out_commit; | |
5553 | } | |
5554 | ||
5555 | le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len); | |
5556 | ||
5557 | ret = ocfs2_journal_dirty(handle, root_bh); | |
5558 | if (ret) { | |
5559 | mlog_errno(ret); | |
5560 | goto out_commit; | |
5561 | } | |
5562 | ||
5563 | ret = ocfs2_truncate_log_append(osb, handle, blkno, len); | |
5564 | if (ret) | |
5565 | mlog_errno(ret); | |
5566 | ||
5567 | out_commit: | |
5568 | ocfs2_commit_trans(osb, handle); | |
5569 | out: | |
5570 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
5571 | ||
5572 | mutex_unlock(&tl_inode->i_mutex); | |
5573 | ||
5574 | if (meta_ac) | |
5575 | ocfs2_free_alloc_context(meta_ac); | |
5576 | ||
5577 | ocfs2_run_deallocs(osb, &dealloc); | |
5578 | ||
5579 | return ret; | |
5580 | } | |
5581 | ||
01225596 | 5582 | static void ocfs2_xattr_bucket_remove_xs(struct inode *inode, |
85db90e7 | 5583 | handle_t *handle, |
01225596 TM |
5584 | struct ocfs2_xattr_search *xs) |
5585 | { | |
ba937127 | 5586 | struct ocfs2_xattr_header *xh = bucket_xh(xs->bucket); |
01225596 TM |
5587 | struct ocfs2_xattr_entry *last = &xh->xh_entries[ |
5588 | le16_to_cpu(xh->xh_count) - 1]; | |
5589 | int ret = 0; | |
5590 | ||
ba937127 | 5591 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
1224be02 | 5592 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5593 | if (ret) { |
5594 | mlog_errno(ret); | |
85db90e7 | 5595 | return; |
01225596 TM |
5596 | } |
5597 | ||
5598 | /* Remove the old entry. */ | |
5599 | memmove(xs->here, xs->here + 1, | |
5600 | (void *)last - (void *)xs->here); | |
5601 | memset(last, 0, sizeof(struct ocfs2_xattr_entry)); | |
5602 | le16_add_cpu(&xh->xh_count, -1); | |
5603 | ||
ba937127 | 5604 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); |
01225596 TM |
5605 | } |
5606 | ||
5607 | /* | |
5608 | * Set the xattr name/value in the bucket specified in xs. | |
5609 | * | |
5610 | * As the new value in xi may be stored in the bucket or in an outside cluster, | |
5611 | * we divide the whole process into 3 steps: | |
5612 | * 1. insert name/value in the bucket(ocfs2_xattr_set_entry_in_bucket) | |
5613 | * 2. truncate of the outside cluster(ocfs2_xattr_bucket_value_truncate_xs) | |
5614 | * 3. Set the value to the outside cluster(ocfs2_xattr_bucket_set_value_outside) | |
5615 | * 4. If the clusters for the new outside value can't be allocated, we need | |
5616 | * to free the xattr we allocated in set. | |
5617 | */ | |
5618 | static int ocfs2_xattr_set_in_bucket(struct inode *inode, | |
5619 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
5620 | struct ocfs2_xattr_search *xs, |
5621 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5622 | { |
9dc47400 | 5623 | int ret; |
01225596 | 5624 | size_t value_len; |
6b240ff6 | 5625 | char *val = (char *)xi->xi_value; |
01225596 | 5626 | struct ocfs2_xattr_entry *xe = xs->here; |
6b240ff6 | 5627 | u32 name_hash = ocfs2_xattr_name_hash(inode, xi->xi_name, |
18853b95 | 5628 | xi->xi_name_len); |
01225596 | 5629 | |
9dc47400 | 5630 | value_len = xi->xi_value_len; |
01225596 TM |
5631 | if (!xs->not_found && !ocfs2_xattr_is_local(xe)) { |
5632 | /* | |
5633 | * We need to truncate the xattr storage first. | |
5634 | * | |
5635 | * If both the old and new value are stored to | |
5636 | * outside block, we only need to truncate | |
5637 | * the storage and then set the value outside. | |
5638 | * | |
5639 | * If the new value should be stored within block, | |
5640 | * we should free all the outside block first and | |
5641 | * the modification to the xattr block will be done | |
5642 | * by following steps. | |
5643 | */ | |
9dc47400 | 5644 | if (xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) |
01225596 TM |
5645 | value_len = 0; |
5646 | ||
5647 | ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, | |
78f30c31 TM |
5648 | value_len, |
5649 | ctxt); | |
01225596 TM |
5650 | if (ret) |
5651 | goto out; | |
5652 | ||
5653 | if (value_len) | |
5654 | goto set_value_outside; | |
5655 | } | |
5656 | ||
01225596 | 5657 | /* So we have to handle the inside block change now. */ |
85db90e7 | 5658 | ret = ocfs2_xattr_set_entry_in_bucket(inode, ctxt->handle, xi, xs, |
9dc47400 | 5659 | name_hash); |
01225596 TM |
5660 | if (ret) { |
5661 | mlog_errno(ret); | |
5662 | goto out; | |
5663 | } | |
5664 | ||
9dc47400 | 5665 | if (xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) |
5a095611 | 5666 | goto out; |
01225596 | 5667 | |
5a095611 TM |
5668 | /* allocate the space now for the outside block storage. */ |
5669 | ret = ocfs2_xattr_bucket_value_truncate_xs(inode, xs, | |
78f30c31 | 5670 | value_len, ctxt); |
5a095611 TM |
5671 | if (ret) { |
5672 | mlog_errno(ret); | |
5673 | ||
5674 | if (xs->not_found) { | |
5675 | /* | |
5676 | * We can't allocate enough clusters for outside | |
5677 | * storage and we have allocated xattr already, | |
5678 | * so need to remove it. | |
5679 | */ | |
85db90e7 | 5680 | ocfs2_xattr_bucket_remove_xs(inode, ctxt->handle, xs); |
01225596 | 5681 | } |
01225596 TM |
5682 | goto out; |
5683 | } | |
5684 | ||
5685 | set_value_outside: | |
85db90e7 TM |
5686 | ret = ocfs2_xattr_bucket_set_value_outside(inode, ctxt->handle, |
5687 | xs, val, value_len); | |
01225596 TM |
5688 | out: |
5689 | return ret; | |
5690 | } | |
5691 | ||
80bcaf34 TM |
5692 | /* |
5693 | * check whether the xattr bucket is filled up with the same hash value. | |
5694 | * If we want to insert the xattr with the same hash, return -ENOSPC. | |
5695 | * If we want to insert a xattr with different hash value, go ahead | |
5696 | * and ocfs2_divide_xattr_bucket will handle this. | |
5697 | */ | |
01225596 | 5698 | static int ocfs2_check_xattr_bucket_collision(struct inode *inode, |
80bcaf34 TM |
5699 | struct ocfs2_xattr_bucket *bucket, |
5700 | const char *name) | |
01225596 | 5701 | { |
3e632946 | 5702 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
80bcaf34 TM |
5703 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
5704 | ||
5705 | if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) | |
5706 | return 0; | |
01225596 TM |
5707 | |
5708 | if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == | |
5709 | xh->xh_entries[0].xe_name_hash) { | |
5710 | mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " | |
5711 | "hash = %u\n", | |
9c7759aa | 5712 | (unsigned long long)bucket_blkno(bucket), |
01225596 TM |
5713 | le32_to_cpu(xh->xh_entries[0].xe_name_hash)); |
5714 | return -ENOSPC; | |
5715 | } | |
5716 | ||
5717 | return 0; | |
5718 | } | |
5719 | ||
5720 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, | |
5721 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
5722 | struct ocfs2_xattr_search *xs, |
5723 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5724 | { |
5725 | struct ocfs2_xattr_header *xh; | |
5726 | struct ocfs2_xattr_entry *xe; | |
5727 | u16 count, header_size, xh_free_start; | |
6dde41d9 | 5728 | int free, max_free, need, old; |
18853b95 | 5729 | size_t value_size = 0; |
01225596 TM |
5730 | size_t blocksize = inode->i_sb->s_blocksize; |
5731 | int ret, allocation = 0; | |
01225596 | 5732 | |
6b240ff6 | 5733 | mlog_entry("Set xattr %s in xattr index block\n", xi->xi_name); |
01225596 TM |
5734 | |
5735 | try_again: | |
5736 | xh = xs->header; | |
5737 | count = le16_to_cpu(xh->xh_count); | |
5738 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
5739 | header_size = sizeof(struct ocfs2_xattr_header) + | |
5740 | count * sizeof(struct ocfs2_xattr_entry); | |
4442f518 TY |
5741 | max_free = OCFS2_XATTR_BUCKET_SIZE - header_size - |
5742 | le16_to_cpu(xh->xh_name_value_len) - OCFS2_XATTR_HEADER_GAP; | |
01225596 TM |
5743 | |
5744 | mlog_bug_on_msg(header_size > blocksize, "bucket %llu has header size " | |
5745 | "of %u which exceed block size\n", | |
ba937127 | 5746 | (unsigned long long)bucket_blkno(xs->bucket), |
01225596 TM |
5747 | header_size); |
5748 | ||
6b240ff6 | 5749 | if (xi->xi_value && xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
01225596 | 5750 | value_size = OCFS2_XATTR_ROOT_SIZE; |
6b240ff6 JB |
5751 | else if (xi->xi_value) |
5752 | value_size = OCFS2_XATTR_SIZE(xi->xi_value_len); | |
01225596 TM |
5753 | |
5754 | if (xs->not_found) | |
5755 | need = sizeof(struct ocfs2_xattr_entry) + | |
18853b95 | 5756 | OCFS2_XATTR_SIZE(xi->xi_name_len) + value_size; |
01225596 | 5757 | else { |
18853b95 | 5758 | need = value_size + OCFS2_XATTR_SIZE(xi->xi_name_len); |
01225596 TM |
5759 | |
5760 | /* | |
5761 | * We only replace the old value if the new length is smaller | |
5762 | * than the old one. Otherwise we will allocate new space in the | |
5763 | * bucket to store it. | |
5764 | */ | |
5765 | xe = xs->here; | |
5766 | if (ocfs2_xattr_is_local(xe)) | |
5767 | old = OCFS2_XATTR_SIZE(le64_to_cpu(xe->xe_value_size)); | |
5768 | else | |
5769 | old = OCFS2_XATTR_SIZE(OCFS2_XATTR_ROOT_SIZE); | |
5770 | ||
5771 | if (old >= value_size) | |
5772 | need = 0; | |
5773 | } | |
5774 | ||
4442f518 | 5775 | free = xh_free_start - header_size - OCFS2_XATTR_HEADER_GAP; |
01225596 TM |
5776 | /* |
5777 | * We need to make sure the new name/value pair | |
5778 | * can exist in the same block. | |
5779 | */ | |
5780 | if (xh_free_start % blocksize < need) | |
5781 | free -= xh_free_start % blocksize; | |
5782 | ||
5783 | mlog(0, "xs->not_found = %d, in xattr bucket %llu: free = %d, " | |
5784 | "need = %d, max_free = %d, xh_free_start = %u, xh_name_value_len =" | |
5785 | " %u\n", xs->not_found, | |
ba937127 | 5786 | (unsigned long long)bucket_blkno(xs->bucket), |
01225596 TM |
5787 | free, need, max_free, le16_to_cpu(xh->xh_free_start), |
5788 | le16_to_cpu(xh->xh_name_value_len)); | |
5789 | ||
976331d8 TM |
5790 | if (free < need || |
5791 | (xs->not_found && | |
5792 | count == ocfs2_xattr_max_xe_in_bucket(inode->i_sb))) { | |
01225596 TM |
5793 | if (need <= max_free && |
5794 | count < ocfs2_xattr_max_xe_in_bucket(inode->i_sb)) { | |
5795 | /* | |
5796 | * We can create the space by defragment. Since only the | |
5797 | * name/value will be moved, the xe shouldn't be changed | |
5798 | * in xs. | |
5799 | */ | |
85db90e7 TM |
5800 | ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, |
5801 | xs->bucket); | |
01225596 TM |
5802 | if (ret) { |
5803 | mlog_errno(ret); | |
5804 | goto out; | |
5805 | } | |
5806 | ||
5807 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4442f518 TY |
5808 | free = xh_free_start - header_size |
5809 | - OCFS2_XATTR_HEADER_GAP; | |
01225596 TM |
5810 | if (xh_free_start % blocksize < need) |
5811 | free -= xh_free_start % blocksize; | |
5812 | ||
5813 | if (free >= need) | |
5814 | goto xattr_set; | |
5815 | ||
5816 | mlog(0, "Can't get enough space for xattr insert by " | |
5817 | "defragment. Need %u bytes, but we have %d, so " | |
5818 | "allocate new bucket for it.\n", need, free); | |
5819 | } | |
5820 | ||
5821 | /* | |
5822 | * We have to add new buckets or clusters and one | |
5823 | * allocation should leave us enough space for insert. | |
5824 | */ | |
5825 | BUG_ON(allocation); | |
5826 | ||
5827 | /* | |
5828 | * We do not allow for overlapping ranges between buckets. And | |
5829 | * the maximum number of collisions we will allow for then is | |
5830 | * one bucket's worth, so check it here whether we need to | |
5831 | * add a new bucket for the insert. | |
5832 | */ | |
80bcaf34 | 5833 | ret = ocfs2_check_xattr_bucket_collision(inode, |
ba937127 | 5834 | xs->bucket, |
6b240ff6 | 5835 | xi->xi_name); |
01225596 TM |
5836 | if (ret) { |
5837 | mlog_errno(ret); | |
5838 | goto out; | |
5839 | } | |
5840 | ||
5841 | ret = ocfs2_add_new_xattr_bucket(inode, | |
5842 | xs->xattr_bh, | |
91f2033f | 5843 | xs->bucket, |
78f30c31 | 5844 | ctxt); |
01225596 TM |
5845 | if (ret) { |
5846 | mlog_errno(ret); | |
5847 | goto out; | |
5848 | } | |
5849 | ||
91f2033f JB |
5850 | /* |
5851 | * ocfs2_add_new_xattr_bucket() will have updated | |
5852 | * xs->bucket if it moved, but it will not have updated | |
5853 | * any of the other search fields. Thus, we drop it and | |
5854 | * re-search. Everything should be cached, so it'll be | |
5855 | * quick. | |
5856 | */ | |
ba937127 | 5857 | ocfs2_xattr_bucket_relse(xs->bucket); |
01225596 | 5858 | ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, |
6b240ff6 JB |
5859 | xi->xi_name_index, |
5860 | xi->xi_name, xs); | |
01225596 TM |
5861 | if (ret && ret != -ENODATA) |
5862 | goto out; | |
5863 | xs->not_found = ret; | |
5864 | allocation = 1; | |
5865 | goto try_again; | |
5866 | } | |
5867 | ||
5868 | xattr_set: | |
78f30c31 | 5869 | ret = ocfs2_xattr_set_in_bucket(inode, xi, xs, ctxt); |
01225596 TM |
5870 | out: |
5871 | mlog_exit(ret); | |
5872 | return ret; | |
5873 | } | |
a3944256 TM |
5874 | |
5875 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
5876 | struct ocfs2_xattr_bucket *bucket, | |
5877 | void *para) | |
5878 | { | |
ce9c5a54 | 5879 | int ret = 0, ref_credits; |
3e632946 | 5880 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
a3944256 TM |
5881 | u16 i; |
5882 | struct ocfs2_xattr_entry *xe; | |
78f30c31 TM |
5883 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
5884 | struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; | |
548b0f22 JB |
5885 | int credits = ocfs2_remove_extent_credits(osb->sb) + |
5886 | ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
ce9c5a54 TM |
5887 | struct ocfs2_xattr_value_root *xv; |
5888 | struct ocfs2_rm_xattr_bucket_para *args = | |
5889 | (struct ocfs2_rm_xattr_bucket_para *)para; | |
78f30c31 TM |
5890 | |
5891 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
a3944256 TM |
5892 | |
5893 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
5894 | xe = &xh->xh_entries[i]; | |
5895 | if (ocfs2_xattr_is_local(xe)) | |
5896 | continue; | |
5897 | ||
ce9c5a54 TM |
5898 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, |
5899 | i, &xv, NULL); | |
5900 | ||
5901 | ret = ocfs2_lock_xattr_remove_allocators(inode, xv, | |
5902 | args->ref_ci, | |
5903 | args->ref_root_bh, | |
5904 | &ctxt.meta_ac, | |
5905 | &ref_credits); | |
5906 | ||
5907 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); | |
88c3b062 TM |
5908 | if (IS_ERR(ctxt.handle)) { |
5909 | ret = PTR_ERR(ctxt.handle); | |
5910 | mlog_errno(ret); | |
5911 | break; | |
5912 | } | |
5913 | ||
548b0f22 | 5914 | ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, |
78f30c31 | 5915 | i, 0, &ctxt); |
88c3b062 TM |
5916 | |
5917 | ocfs2_commit_trans(osb, ctxt.handle); | |
ce9c5a54 TM |
5918 | if (ctxt.meta_ac) { |
5919 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
5920 | ctxt.meta_ac = NULL; | |
5921 | } | |
a3944256 TM |
5922 | if (ret) { |
5923 | mlog_errno(ret); | |
5924 | break; | |
5925 | } | |
5926 | } | |
5927 | ||
ce9c5a54 TM |
5928 | if (ctxt.meta_ac) |
5929 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
5930 | ocfs2_schedule_truncate_log_flush(osb, 1); |
5931 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
a3944256 TM |
5932 | return ret; |
5933 | } | |
5934 | ||
492a8a33 TM |
5935 | /* |
5936 | * Whenever we modify a xattr value root in the bucket(e.g, CoW | |
5937 | * or change the extent record flag), we need to recalculate | |
5938 | * the metaecc for the whole bucket. So it is done here. | |
5939 | * | |
5940 | * Note: | |
5941 | * We have to give the extra credits for the caller. | |
5942 | */ | |
5943 | static int ocfs2_xattr_bucket_post_refcount(struct inode *inode, | |
5944 | handle_t *handle, | |
5945 | void *para) | |
5946 | { | |
5947 | int ret; | |
5948 | struct ocfs2_xattr_bucket *bucket = | |
5949 | (struct ocfs2_xattr_bucket *)para; | |
5950 | ||
5951 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, | |
5952 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5953 | if (ret) { | |
5954 | mlog_errno(ret); | |
5955 | return ret; | |
5956 | } | |
5957 | ||
5958 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
5959 | ||
5960 | return 0; | |
5961 | } | |
5962 | ||
5963 | /* | |
5964 | * Special action we need if the xattr value is refcounted. | |
5965 | * | |
5966 | * 1. If the xattr is refcounted, lock the tree. | |
5967 | * 2. CoW the xattr if we are setting the new value and the value | |
5968 | * will be stored outside. | |
5969 | * 3. In other case, decrease_refcount will work for us, so just | |
5970 | * lock the refcount tree, calculate the meta and credits is OK. | |
5971 | * | |
5972 | * We have to do CoW before ocfs2_init_xattr_set_ctxt since | |
5973 | * currently CoW is a completed transaction, while this function | |
5974 | * will also lock the allocators and let us deadlock. So we will | |
5975 | * CoW the whole xattr value. | |
5976 | */ | |
5977 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, | |
5978 | struct ocfs2_dinode *di, | |
5979 | struct ocfs2_xattr_info *xi, | |
5980 | struct ocfs2_xattr_search *xis, | |
5981 | struct ocfs2_xattr_search *xbs, | |
5982 | struct ocfs2_refcount_tree **ref_tree, | |
5983 | int *meta_add, | |
5984 | int *credits) | |
5985 | { | |
5986 | int ret = 0; | |
5987 | struct ocfs2_xattr_block *xb; | |
5988 | struct ocfs2_xattr_entry *xe; | |
5989 | char *base; | |
5990 | u32 p_cluster, num_clusters; | |
5991 | unsigned int ext_flags; | |
5992 | int name_offset, name_len; | |
5993 | struct ocfs2_xattr_value_buf vb; | |
5994 | struct ocfs2_xattr_bucket *bucket = NULL; | |
5995 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5996 | struct ocfs2_post_refcount refcount; | |
5997 | struct ocfs2_post_refcount *p = NULL; | |
5998 | struct buffer_head *ref_root_bh = NULL; | |
5999 | ||
6000 | if (!xis->not_found) { | |
6001 | xe = xis->here; | |
6002 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
6003 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
6004 | base = xis->base; | |
6005 | vb.vb_bh = xis->inode_bh; | |
6006 | vb.vb_access = ocfs2_journal_access_di; | |
6007 | } else { | |
6008 | int i, block_off = 0; | |
6009 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
6010 | xe = xbs->here; | |
6011 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
6012 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
6013 | i = xbs->here - xbs->header->xh_entries; | |
6014 | ||
6015 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 6016 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
492a8a33 TM |
6017 | bucket_xh(xbs->bucket), |
6018 | i, &block_off, | |
6019 | &name_offset); | |
6020 | if (ret) { | |
6021 | mlog_errno(ret); | |
6022 | goto out; | |
6023 | } | |
6024 | base = bucket_block(xbs->bucket, block_off); | |
6025 | vb.vb_bh = xbs->bucket->bu_bhs[block_off]; | |
6026 | vb.vb_access = ocfs2_journal_access; | |
6027 | ||
6028 | if (ocfs2_meta_ecc(osb)) { | |
6029 | /*create parameters for ocfs2_post_refcount. */ | |
6030 | bucket = xbs->bucket; | |
6031 | refcount.credits = bucket->bu_blocks; | |
6032 | refcount.para = bucket; | |
6033 | refcount.func = | |
6034 | ocfs2_xattr_bucket_post_refcount; | |
6035 | p = &refcount; | |
6036 | } | |
6037 | } else { | |
6038 | base = xbs->base; | |
6039 | vb.vb_bh = xbs->xattr_bh; | |
6040 | vb.vb_access = ocfs2_journal_access_xb; | |
6041 | } | |
6042 | } | |
6043 | ||
6044 | if (ocfs2_xattr_is_local(xe)) | |
6045 | goto out; | |
6046 | ||
6047 | vb.vb_xv = (struct ocfs2_xattr_value_root *) | |
6048 | (base + name_offset + name_len); | |
6049 | ||
6050 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
6051 | &num_clusters, &vb.vb_xv->xr_list, | |
6052 | &ext_flags); | |
6053 | if (ret) { | |
6054 | mlog_errno(ret); | |
6055 | goto out; | |
6056 | } | |
6057 | ||
6058 | /* | |
6059 | * We just need to check the 1st extent record, since we always | |
6060 | * CoW the whole xattr. So there shouldn't be a xattr with | |
6061 | * some REFCOUNT extent recs after the 1st one. | |
6062 | */ | |
6063 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
6064 | goto out; | |
6065 | ||
6066 | ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), | |
6067 | 1, ref_tree, &ref_root_bh); | |
6068 | if (ret) { | |
6069 | mlog_errno(ret); | |
6070 | goto out; | |
6071 | } | |
6072 | ||
6073 | /* | |
6074 | * If we are deleting the xattr or the new size will be stored inside, | |
6075 | * cool, leave it there, the xattr truncate process will remove them | |
6076 | * for us(it still needs the refcount tree lock and the meta, credits). | |
6077 | * And the worse case is that every cluster truncate will split the | |
6078 | * refcount tree, and make the original extent become 3. So we will need | |
6079 | * 2 * cluster more extent recs at most. | |
6080 | */ | |
6b240ff6 | 6081 | if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) { |
492a8a33 TM |
6082 | |
6083 | ret = ocfs2_refcounted_xattr_delete_need(inode, | |
6084 | &(*ref_tree)->rf_ci, | |
6085 | ref_root_bh, vb.vb_xv, | |
6086 | meta_add, credits); | |
6087 | if (ret) | |
6088 | mlog_errno(ret); | |
6089 | goto out; | |
6090 | } | |
6091 | ||
6092 | ret = ocfs2_refcount_cow_xattr(inode, di, &vb, | |
6093 | *ref_tree, ref_root_bh, 0, | |
6094 | le32_to_cpu(vb.vb_xv->xr_clusters), p); | |
6095 | if (ret) | |
6096 | mlog_errno(ret); | |
6097 | ||
6098 | out: | |
6099 | brelse(ref_root_bh); | |
6100 | return ret; | |
6101 | } | |
6102 | ||
0129241e TM |
6103 | /* |
6104 | * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. | |
6105 | * The physical clusters will be added to refcount tree. | |
6106 | */ | |
6107 | static int ocfs2_xattr_value_attach_refcount(struct inode *inode, | |
6108 | struct ocfs2_xattr_value_root *xv, | |
6109 | struct ocfs2_extent_tree *value_et, | |
6110 | struct ocfs2_caching_info *ref_ci, | |
6111 | struct buffer_head *ref_root_bh, | |
6112 | struct ocfs2_cached_dealloc_ctxt *dealloc, | |
6113 | struct ocfs2_post_refcount *refcount) | |
6114 | { | |
6115 | int ret = 0; | |
6116 | u32 clusters = le32_to_cpu(xv->xr_clusters); | |
6117 | u32 cpos, p_cluster, num_clusters; | |
6118 | struct ocfs2_extent_list *el = &xv->xr_list; | |
6119 | unsigned int ext_flags; | |
6120 | ||
6121 | cpos = 0; | |
6122 | while (cpos < clusters) { | |
6123 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
6124 | &num_clusters, el, &ext_flags); | |
6125 | ||
6126 | cpos += num_clusters; | |
6127 | if ((ext_flags & OCFS2_EXT_REFCOUNTED)) | |
6128 | continue; | |
6129 | ||
6130 | BUG_ON(!p_cluster); | |
6131 | ||
6132 | ret = ocfs2_add_refcount_flag(inode, value_et, | |
6133 | ref_ci, ref_root_bh, | |
6134 | cpos - num_clusters, | |
6135 | p_cluster, num_clusters, | |
6136 | dealloc, refcount); | |
6137 | if (ret) { | |
6138 | mlog_errno(ret); | |
6139 | break; | |
6140 | } | |
6141 | } | |
6142 | ||
6143 | return ret; | |
6144 | } | |
6145 | ||
6146 | /* | |
6147 | * Given a normal ocfs2_xattr_header, refcount all the entries which | |
6148 | * have value stored outside. | |
6149 | * Used for xattrs stored in inode and ocfs2_xattr_block. | |
6150 | */ | |
6151 | static int ocfs2_xattr_attach_refcount_normal(struct inode *inode, | |
6152 | struct ocfs2_xattr_value_buf *vb, | |
6153 | struct ocfs2_xattr_header *header, | |
6154 | struct ocfs2_caching_info *ref_ci, | |
6155 | struct buffer_head *ref_root_bh, | |
6156 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6157 | { | |
6158 | ||
6159 | struct ocfs2_xattr_entry *xe; | |
6160 | struct ocfs2_xattr_value_root *xv; | |
6161 | struct ocfs2_extent_tree et; | |
6162 | int i, ret = 0; | |
6163 | ||
6164 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
6165 | xe = &header->xh_entries[i]; | |
6166 | ||
6167 | if (ocfs2_xattr_is_local(xe)) | |
6168 | continue; | |
6169 | ||
6170 | xv = (struct ocfs2_xattr_value_root *)((void *)header + | |
6171 | le16_to_cpu(xe->xe_name_offset) + | |
6172 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6173 | ||
6174 | vb->vb_xv = xv; | |
6175 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); | |
6176 | ||
6177 | ret = ocfs2_xattr_value_attach_refcount(inode, xv, &et, | |
6178 | ref_ci, ref_root_bh, | |
6179 | dealloc, NULL); | |
6180 | if (ret) { | |
6181 | mlog_errno(ret); | |
6182 | break; | |
6183 | } | |
6184 | } | |
6185 | ||
6186 | return ret; | |
6187 | } | |
6188 | ||
6189 | static int ocfs2_xattr_inline_attach_refcount(struct inode *inode, | |
6190 | struct buffer_head *fe_bh, | |
6191 | struct ocfs2_caching_info *ref_ci, | |
6192 | struct buffer_head *ref_root_bh, | |
6193 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6194 | { | |
6195 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6196 | struct ocfs2_xattr_header *header = (struct ocfs2_xattr_header *) | |
6197 | (fe_bh->b_data + inode->i_sb->s_blocksize - | |
6198 | le16_to_cpu(di->i_xattr_inline_size)); | |
6199 | struct ocfs2_xattr_value_buf vb = { | |
6200 | .vb_bh = fe_bh, | |
6201 | .vb_access = ocfs2_journal_access_di, | |
6202 | }; | |
6203 | ||
6204 | return ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6205 | ref_ci, ref_root_bh, dealloc); | |
6206 | } | |
6207 | ||
6208 | struct ocfs2_xattr_tree_value_refcount_para { | |
6209 | struct ocfs2_caching_info *ref_ci; | |
6210 | struct buffer_head *ref_root_bh; | |
6211 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
6212 | }; | |
6213 | ||
6214 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, | |
6215 | struct ocfs2_xattr_bucket *bucket, | |
6216 | int offset, | |
6217 | struct ocfs2_xattr_value_root **xv, | |
6218 | struct buffer_head **bh) | |
6219 | { | |
6220 | int ret, block_off, name_offset; | |
6221 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); | |
6222 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6223 | void *base; | |
6224 | ||
6225 | ret = ocfs2_xattr_bucket_get_name_value(sb, | |
6226 | bucket_xh(bucket), | |
6227 | offset, | |
6228 | &block_off, | |
6229 | &name_offset); | |
6230 | if (ret) { | |
6231 | mlog_errno(ret); | |
6232 | goto out; | |
6233 | } | |
6234 | ||
6235 | base = bucket_block(bucket, block_off); | |
6236 | ||
6237 | *xv = (struct ocfs2_xattr_value_root *)(base + name_offset + | |
6238 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6239 | ||
6240 | if (bh) | |
6241 | *bh = bucket->bu_bhs[block_off]; | |
6242 | out: | |
6243 | return ret; | |
6244 | } | |
6245 | ||
6246 | /* | |
6247 | * For a given xattr bucket, refcount all the entries which | |
6248 | * have value stored outside. | |
6249 | */ | |
6250 | static int ocfs2_xattr_bucket_value_refcount(struct inode *inode, | |
6251 | struct ocfs2_xattr_bucket *bucket, | |
6252 | void *para) | |
6253 | { | |
6254 | int i, ret = 0; | |
6255 | struct ocfs2_extent_tree et; | |
6256 | struct ocfs2_xattr_tree_value_refcount_para *ref = | |
6257 | (struct ocfs2_xattr_tree_value_refcount_para *)para; | |
6258 | struct ocfs2_xattr_header *xh = | |
6259 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6260 | struct ocfs2_xattr_entry *xe; | |
6261 | struct ocfs2_xattr_value_buf vb = { | |
6262 | .vb_access = ocfs2_journal_access, | |
6263 | }; | |
6264 | struct ocfs2_post_refcount refcount = { | |
6265 | .credits = bucket->bu_blocks, | |
6266 | .para = bucket, | |
6267 | .func = ocfs2_xattr_bucket_post_refcount, | |
6268 | }; | |
6269 | struct ocfs2_post_refcount *p = NULL; | |
6270 | ||
6271 | /* We only need post_refcount if we support metaecc. */ | |
6272 | if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb))) | |
6273 | p = &refcount; | |
6274 | ||
6275 | mlog(0, "refcount bucket %llu, count = %u\n", | |
6276 | (unsigned long long)bucket_blkno(bucket), | |
6277 | le16_to_cpu(xh->xh_count)); | |
6278 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6279 | xe = &xh->xh_entries[i]; | |
6280 | ||
6281 | if (ocfs2_xattr_is_local(xe)) | |
6282 | continue; | |
6283 | ||
6284 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i, | |
6285 | &vb.vb_xv, &vb.vb_bh); | |
6286 | if (ret) { | |
6287 | mlog_errno(ret); | |
6288 | break; | |
6289 | } | |
6290 | ||
6291 | ocfs2_init_xattr_value_extent_tree(&et, | |
6292 | INODE_CACHE(inode), &vb); | |
6293 | ||
6294 | ret = ocfs2_xattr_value_attach_refcount(inode, vb.vb_xv, | |
6295 | &et, ref->ref_ci, | |
6296 | ref->ref_root_bh, | |
6297 | ref->dealloc, p); | |
6298 | if (ret) { | |
6299 | mlog_errno(ret); | |
6300 | break; | |
6301 | } | |
6302 | } | |
6303 | ||
6304 | return ret; | |
6305 | ||
6306 | } | |
6307 | ||
6308 | static int ocfs2_refcount_xattr_tree_rec(struct inode *inode, | |
6309 | struct buffer_head *root_bh, | |
6310 | u64 blkno, u32 cpos, u32 len, void *para) | |
6311 | { | |
6312 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
6313 | ocfs2_xattr_bucket_value_refcount, | |
6314 | para); | |
6315 | } | |
6316 | ||
6317 | static int ocfs2_xattr_block_attach_refcount(struct inode *inode, | |
6318 | struct buffer_head *blk_bh, | |
6319 | struct ocfs2_caching_info *ref_ci, | |
6320 | struct buffer_head *ref_root_bh, | |
6321 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6322 | { | |
6323 | int ret = 0; | |
6324 | struct ocfs2_xattr_block *xb = | |
6325 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6326 | ||
6327 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
6328 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
6329 | struct ocfs2_xattr_value_buf vb = { | |
6330 | .vb_bh = blk_bh, | |
6331 | .vb_access = ocfs2_journal_access_xb, | |
6332 | }; | |
6333 | ||
6334 | ret = ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6335 | ref_ci, ref_root_bh, | |
6336 | dealloc); | |
6337 | } else { | |
6338 | struct ocfs2_xattr_tree_value_refcount_para para = { | |
6339 | .ref_ci = ref_ci, | |
6340 | .ref_root_bh = ref_root_bh, | |
6341 | .dealloc = dealloc, | |
6342 | }; | |
6343 | ||
6344 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
6345 | ocfs2_refcount_xattr_tree_rec, | |
6346 | ¶); | |
6347 | } | |
6348 | ||
6349 | return ret; | |
6350 | } | |
6351 | ||
6352 | int ocfs2_xattr_attach_refcount_tree(struct inode *inode, | |
6353 | struct buffer_head *fe_bh, | |
6354 | struct ocfs2_caching_info *ref_ci, | |
6355 | struct buffer_head *ref_root_bh, | |
6356 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6357 | { | |
6358 | int ret = 0; | |
6359 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
6360 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6361 | struct buffer_head *blk_bh = NULL; | |
6362 | ||
6363 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
6364 | ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh, | |
6365 | ref_ci, ref_root_bh, | |
6366 | dealloc); | |
6367 | if (ret) { | |
6368 | mlog_errno(ret); | |
6369 | goto out; | |
6370 | } | |
6371 | } | |
6372 | ||
6373 | if (!di->i_xattr_loc) | |
6374 | goto out; | |
6375 | ||
6376 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), | |
6377 | &blk_bh); | |
6378 | if (ret < 0) { | |
6379 | mlog_errno(ret); | |
6380 | goto out; | |
6381 | } | |
6382 | ||
6383 | ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci, | |
6384 | ref_root_bh, dealloc); | |
6385 | if (ret) | |
6386 | mlog_errno(ret); | |
6387 | ||
6388 | brelse(blk_bh); | |
6389 | out: | |
6390 | ||
6391 | return ret; | |
6392 | } | |
6393 | ||
0fe9b66c | 6394 | typedef int (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); |
2999d12f TM |
6395 | /* |
6396 | * Store the information we need in xattr reflink. | |
6397 | * old_bh and new_bh are inode bh for the old and new inode. | |
6398 | */ | |
6399 | struct ocfs2_xattr_reflink { | |
6400 | struct inode *old_inode; | |
6401 | struct inode *new_inode; | |
6402 | struct buffer_head *old_bh; | |
6403 | struct buffer_head *new_bh; | |
6404 | struct ocfs2_caching_info *ref_ci; | |
6405 | struct buffer_head *ref_root_bh; | |
6406 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
0fe9b66c | 6407 | should_xattr_reflinked *xattr_reflinked; |
2999d12f TM |
6408 | }; |
6409 | ||
6410 | /* | |
6411 | * Given a xattr header and xe offset, | |
6412 | * return the proper xv and the corresponding bh. | |
6413 | * xattr in inode, block and xattr tree have different implementaions. | |
6414 | */ | |
6415 | typedef int (get_xattr_value_root)(struct super_block *sb, | |
6416 | struct buffer_head *bh, | |
6417 | struct ocfs2_xattr_header *xh, | |
6418 | int offset, | |
6419 | struct ocfs2_xattr_value_root **xv, | |
6420 | struct buffer_head **ret_bh, | |
6421 | void *para); | |
6422 | ||
6423 | /* | |
6424 | * Calculate all the xattr value root metadata stored in this xattr header and | |
6425 | * credits we need if we create them from the scratch. | |
6426 | * We use get_xattr_value_root so that all types of xattr container can use it. | |
6427 | */ | |
6428 | static int ocfs2_value_metas_in_xattr_header(struct super_block *sb, | |
6429 | struct buffer_head *bh, | |
6430 | struct ocfs2_xattr_header *xh, | |
6431 | int *metas, int *credits, | |
6432 | int *num_recs, | |
6433 | get_xattr_value_root *func, | |
6434 | void *para) | |
6435 | { | |
6436 | int i, ret = 0; | |
6437 | struct ocfs2_xattr_value_root *xv; | |
6438 | struct ocfs2_xattr_entry *xe; | |
6439 | ||
6440 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6441 | xe = &xh->xh_entries[i]; | |
6442 | if (ocfs2_xattr_is_local(xe)) | |
6443 | continue; | |
6444 | ||
6445 | ret = func(sb, bh, xh, i, &xv, NULL, para); | |
6446 | if (ret) { | |
6447 | mlog_errno(ret); | |
6448 | break; | |
6449 | } | |
6450 | ||
6451 | *metas += le16_to_cpu(xv->xr_list.l_tree_depth) * | |
6452 | le16_to_cpu(xv->xr_list.l_next_free_rec); | |
6453 | ||
6454 | *credits += ocfs2_calc_extend_credits(sb, | |
6455 | &def_xv.xv.xr_list, | |
6456 | le32_to_cpu(xv->xr_clusters)); | |
6457 | ||
6458 | /* | |
6459 | * If the value is a tree with depth > 1, We don't go deep | |
6460 | * to the extent block, so just calculate a maximum record num. | |
6461 | */ | |
6462 | if (!xv->xr_list.l_tree_depth) | |
8ff6af88 | 6463 | *num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); |
2999d12f TM |
6464 | else |
6465 | *num_recs += ocfs2_clusters_for_bytes(sb, | |
6466 | XATTR_SIZE_MAX); | |
6467 | } | |
6468 | ||
6469 | return ret; | |
6470 | } | |
6471 | ||
6472 | /* Used by xattr inode and block to return the right xv and buffer_head. */ | |
6473 | static int ocfs2_get_xattr_value_root(struct super_block *sb, | |
6474 | struct buffer_head *bh, | |
6475 | struct ocfs2_xattr_header *xh, | |
6476 | int offset, | |
6477 | struct ocfs2_xattr_value_root **xv, | |
6478 | struct buffer_head **ret_bh, | |
6479 | void *para) | |
6480 | { | |
6481 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6482 | ||
6483 | *xv = (struct ocfs2_xattr_value_root *)((void *)xh + | |
6484 | le16_to_cpu(xe->xe_name_offset) + | |
6485 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6486 | ||
6487 | if (ret_bh) | |
6488 | *ret_bh = bh; | |
6489 | ||
6490 | return 0; | |
6491 | } | |
6492 | ||
6493 | /* | |
6494 | * Lock the meta_ac and caculate how much credits we need for reflink xattrs. | |
6495 | * It is only used for inline xattr and xattr block. | |
6496 | */ | |
6497 | static int ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, | |
6498 | struct ocfs2_xattr_header *xh, | |
6499 | struct buffer_head *ref_root_bh, | |
6500 | int *credits, | |
6501 | struct ocfs2_alloc_context **meta_ac) | |
6502 | { | |
6503 | int ret, meta_add = 0, num_recs = 0; | |
6504 | struct ocfs2_refcount_block *rb = | |
6505 | (struct ocfs2_refcount_block *)ref_root_bh->b_data; | |
6506 | ||
6507 | *credits = 0; | |
6508 | ||
6509 | ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh, | |
6510 | &meta_add, credits, &num_recs, | |
6511 | ocfs2_get_xattr_value_root, | |
6512 | NULL); | |
6513 | if (ret) { | |
6514 | mlog_errno(ret); | |
6515 | goto out; | |
6516 | } | |
6517 | ||
6518 | /* | |
6519 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6520 | * an approximate number we need for refcount tree change. | |
6521 | * Sometimes we need to split the tree, and after split, half recs | |
6522 | * will be moved to the new block, and a new block can only provide | |
6523 | * half number of recs. So we multiple new blocks by 2. | |
6524 | */ | |
6525 | num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6526 | meta_add += num_recs; | |
6527 | *credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6528 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6529 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6530 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6531 | else | |
6532 | *credits += 1; | |
6533 | ||
6534 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); | |
6535 | if (ret) | |
6536 | mlog_errno(ret); | |
6537 | ||
6538 | out: | |
6539 | return ret; | |
6540 | } | |
6541 | ||
6542 | /* | |
6543 | * Given a xattr header, reflink all the xattrs in this container. | |
6544 | * It can be used for inode, block and bucket. | |
6545 | * | |
6546 | * NOTE: | |
6547 | * Before we call this function, the caller has memcpy the xattr in | |
6548 | * old_xh to the new_xh. | |
0fe9b66c TM |
6549 | * |
6550 | * If args.xattr_reflinked is set, call it to decide whether the xe should | |
6551 | * be reflinked or not. If not, remove it from the new xattr header. | |
2999d12f TM |
6552 | */ |
6553 | static int ocfs2_reflink_xattr_header(handle_t *handle, | |
6554 | struct ocfs2_xattr_reflink *args, | |
6555 | struct buffer_head *old_bh, | |
6556 | struct ocfs2_xattr_header *xh, | |
6557 | struct buffer_head *new_bh, | |
6558 | struct ocfs2_xattr_header *new_xh, | |
6559 | struct ocfs2_xattr_value_buf *vb, | |
6560 | struct ocfs2_alloc_context *meta_ac, | |
6561 | get_xattr_value_root *func, | |
6562 | void *para) | |
6563 | { | |
0fe9b66c | 6564 | int ret = 0, i, j; |
2999d12f TM |
6565 | struct super_block *sb = args->old_inode->i_sb; |
6566 | struct buffer_head *value_bh; | |
0fe9b66c | 6567 | struct ocfs2_xattr_entry *xe, *last; |
2999d12f TM |
6568 | struct ocfs2_xattr_value_root *xv, *new_xv; |
6569 | struct ocfs2_extent_tree data_et; | |
6570 | u32 clusters, cpos, p_cluster, num_clusters; | |
6571 | unsigned int ext_flags = 0; | |
6572 | ||
6573 | mlog(0, "reflink xattr in container %llu, count = %u\n", | |
6574 | (unsigned long long)old_bh->b_blocknr, le16_to_cpu(xh->xh_count)); | |
0fe9b66c TM |
6575 | |
6576 | last = &new_xh->xh_entries[le16_to_cpu(new_xh->xh_count)]; | |
6577 | for (i = 0, j = 0; i < le16_to_cpu(xh->xh_count); i++, j++) { | |
2999d12f TM |
6578 | xe = &xh->xh_entries[i]; |
6579 | ||
0fe9b66c TM |
6580 | if (args->xattr_reflinked && !args->xattr_reflinked(xe)) { |
6581 | xe = &new_xh->xh_entries[j]; | |
6582 | ||
6583 | le16_add_cpu(&new_xh->xh_count, -1); | |
6584 | if (new_xh->xh_count) { | |
6585 | memmove(xe, xe + 1, | |
6586 | (void *)last - (void *)xe); | |
6587 | memset(last, 0, | |
6588 | sizeof(struct ocfs2_xattr_entry)); | |
6589 | } | |
6590 | ||
6591 | /* | |
6592 | * We don't want j to increase in the next round since | |
6593 | * it is already moved ahead. | |
6594 | */ | |
6595 | j--; | |
6596 | continue; | |
6597 | } | |
6598 | ||
2999d12f TM |
6599 | if (ocfs2_xattr_is_local(xe)) |
6600 | continue; | |
6601 | ||
6602 | ret = func(sb, old_bh, xh, i, &xv, NULL, para); | |
6603 | if (ret) { | |
6604 | mlog_errno(ret); | |
6605 | break; | |
6606 | } | |
6607 | ||
0fe9b66c | 6608 | ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); |
2999d12f TM |
6609 | if (ret) { |
6610 | mlog_errno(ret); | |
6611 | break; | |
6612 | } | |
6613 | ||
6614 | /* | |
6615 | * For the xattr which has l_tree_depth = 0, all the extent | |
6616 | * recs have already be copied to the new xh with the | |
6617 | * propriate OCFS2_EXT_REFCOUNTED flag we just need to | |
6618 | * increase the refount count int the refcount tree. | |
6619 | * | |
6620 | * For the xattr which has l_tree_depth > 0, we need | |
6621 | * to initialize it to the empty default value root, | |
6622 | * and then insert the extents one by one. | |
6623 | */ | |
6624 | if (xv->xr_list.l_tree_depth) { | |
6625 | memcpy(new_xv, &def_xv, sizeof(def_xv)); | |
6626 | vb->vb_xv = new_xv; | |
6627 | vb->vb_bh = value_bh; | |
6628 | ocfs2_init_xattr_value_extent_tree(&data_et, | |
6629 | INODE_CACHE(args->new_inode), vb); | |
6630 | } | |
6631 | ||
6632 | clusters = le32_to_cpu(xv->xr_clusters); | |
6633 | cpos = 0; | |
6634 | while (cpos < clusters) { | |
6635 | ret = ocfs2_xattr_get_clusters(args->old_inode, | |
6636 | cpos, | |
6637 | &p_cluster, | |
6638 | &num_clusters, | |
6639 | &xv->xr_list, | |
6640 | &ext_flags); | |
6641 | if (ret) { | |
6642 | mlog_errno(ret); | |
6643 | goto out; | |
6644 | } | |
6645 | ||
6646 | BUG_ON(!p_cluster); | |
6647 | ||
6648 | if (xv->xr_list.l_tree_depth) { | |
6649 | ret = ocfs2_insert_extent(handle, | |
6650 | &data_et, cpos, | |
6651 | ocfs2_clusters_to_blocks( | |
6652 | args->old_inode->i_sb, | |
6653 | p_cluster), | |
6654 | num_clusters, ext_flags, | |
6655 | meta_ac); | |
6656 | if (ret) { | |
6657 | mlog_errno(ret); | |
6658 | goto out; | |
6659 | } | |
6660 | } | |
6661 | ||
6662 | ret = ocfs2_increase_refcount(handle, args->ref_ci, | |
6663 | args->ref_root_bh, | |
6664 | p_cluster, num_clusters, | |
6665 | meta_ac, args->dealloc); | |
6666 | if (ret) { | |
6667 | mlog_errno(ret); | |
6668 | goto out; | |
6669 | } | |
6670 | ||
6671 | cpos += num_clusters; | |
6672 | } | |
6673 | } | |
6674 | ||
6675 | out: | |
6676 | return ret; | |
6677 | } | |
6678 | ||
6679 | static int ocfs2_reflink_xattr_inline(struct ocfs2_xattr_reflink *args) | |
6680 | { | |
6681 | int ret = 0, credits = 0; | |
6682 | handle_t *handle; | |
6683 | struct ocfs2_super *osb = OCFS2_SB(args->old_inode->i_sb); | |
6684 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)args->old_bh->b_data; | |
6685 | int inline_size = le16_to_cpu(di->i_xattr_inline_size); | |
6686 | int header_off = osb->sb->s_blocksize - inline_size; | |
6687 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *) | |
6688 | (args->old_bh->b_data + header_off); | |
6689 | struct ocfs2_xattr_header *new_xh = (struct ocfs2_xattr_header *) | |
6690 | (args->new_bh->b_data + header_off); | |
6691 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6692 | struct ocfs2_inode_info *new_oi; | |
6693 | struct ocfs2_dinode *new_di; | |
6694 | struct ocfs2_xattr_value_buf vb = { | |
6695 | .vb_bh = args->new_bh, | |
6696 | .vb_access = ocfs2_journal_access_di, | |
6697 | }; | |
6698 | ||
6699 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6700 | &credits, &meta_ac); | |
6701 | if (ret) { | |
6702 | mlog_errno(ret); | |
6703 | goto out; | |
6704 | } | |
6705 | ||
6706 | handle = ocfs2_start_trans(osb, credits); | |
6707 | if (IS_ERR(handle)) { | |
6708 | ret = PTR_ERR(handle); | |
6709 | mlog_errno(ret); | |
6710 | goto out; | |
6711 | } | |
6712 | ||
6713 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(args->new_inode), | |
6714 | args->new_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6715 | if (ret) { | |
6716 | mlog_errno(ret); | |
6717 | goto out_commit; | |
6718 | } | |
6719 | ||
6720 | memcpy(args->new_bh->b_data + header_off, | |
6721 | args->old_bh->b_data + header_off, inline_size); | |
6722 | ||
6723 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6724 | new_di->i_xattr_inline_size = cpu_to_le16(inline_size); | |
6725 | ||
6726 | ret = ocfs2_reflink_xattr_header(handle, args, args->old_bh, xh, | |
6727 | args->new_bh, new_xh, &vb, meta_ac, | |
6728 | ocfs2_get_xattr_value_root, NULL); | |
6729 | if (ret) { | |
6730 | mlog_errno(ret); | |
6731 | goto out_commit; | |
6732 | } | |
6733 | ||
6734 | new_oi = OCFS2_I(args->new_inode); | |
6735 | spin_lock(&new_oi->ip_lock); | |
6736 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL | OCFS2_INLINE_XATTR_FL; | |
6737 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6738 | spin_unlock(&new_oi->ip_lock); | |
6739 | ||
6740 | ocfs2_journal_dirty(handle, args->new_bh); | |
6741 | ||
6742 | out_commit: | |
6743 | ocfs2_commit_trans(osb, handle); | |
6744 | ||
6745 | out: | |
6746 | if (meta_ac) | |
6747 | ocfs2_free_alloc_context(meta_ac); | |
6748 | return ret; | |
6749 | } | |
6750 | ||
6751 | static int ocfs2_create_empty_xattr_block(struct inode *inode, | |
6752 | struct buffer_head *fe_bh, | |
6753 | struct buffer_head **ret_bh, | |
6754 | int indexed) | |
6755 | { | |
6756 | int ret; | |
6757 | handle_t *handle; | |
6758 | struct ocfs2_alloc_context *meta_ac; | |
6759 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
6760 | ||
6761 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &meta_ac); | |
6762 | if (ret < 0) { | |
6763 | mlog_errno(ret); | |
6764 | return ret; | |
6765 | } | |
6766 | ||
6767 | handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS); | |
6768 | if (IS_ERR(handle)) { | |
6769 | ret = PTR_ERR(handle); | |
6770 | mlog_errno(ret); | |
6771 | goto out; | |
6772 | } | |
6773 | ||
6774 | mlog(0, "create new xattr block for inode %llu, index = %d\n", | |
6775 | (unsigned long long)fe_bh->b_blocknr, indexed); | |
6776 | ret = ocfs2_create_xattr_block(handle, inode, fe_bh, | |
6777 | meta_ac, ret_bh, indexed); | |
6778 | if (ret) | |
6779 | mlog_errno(ret); | |
6780 | ||
6781 | ocfs2_commit_trans(osb, handle); | |
6782 | out: | |
6783 | ocfs2_free_alloc_context(meta_ac); | |
6784 | return ret; | |
6785 | } | |
6786 | ||
6787 | static int ocfs2_reflink_xattr_block(struct ocfs2_xattr_reflink *args, | |
6788 | struct buffer_head *blk_bh, | |
6789 | struct buffer_head *new_blk_bh) | |
6790 | { | |
6791 | int ret = 0, credits = 0; | |
6792 | handle_t *handle; | |
6793 | struct ocfs2_inode_info *new_oi = OCFS2_I(args->new_inode); | |
6794 | struct ocfs2_dinode *new_di; | |
6795 | struct ocfs2_super *osb = OCFS2_SB(args->new_inode->i_sb); | |
6796 | int header_off = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
6797 | struct ocfs2_xattr_block *xb = | |
6798 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6799 | struct ocfs2_xattr_header *xh = &xb->xb_attrs.xb_header; | |
6800 | struct ocfs2_xattr_block *new_xb = | |
6801 | (struct ocfs2_xattr_block *)new_blk_bh->b_data; | |
6802 | struct ocfs2_xattr_header *new_xh = &new_xb->xb_attrs.xb_header; | |
6803 | struct ocfs2_alloc_context *meta_ac; | |
6804 | struct ocfs2_xattr_value_buf vb = { | |
6805 | .vb_bh = new_blk_bh, | |
6806 | .vb_access = ocfs2_journal_access_xb, | |
6807 | }; | |
6808 | ||
6809 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6810 | &credits, &meta_ac); | |
6811 | if (ret) { | |
6812 | mlog_errno(ret); | |
6813 | return ret; | |
6814 | } | |
6815 | ||
6816 | /* One more credits in case we need to add xattr flags in new inode. */ | |
6817 | handle = ocfs2_start_trans(osb, credits + 1); | |
6818 | if (IS_ERR(handle)) { | |
6819 | ret = PTR_ERR(handle); | |
6820 | mlog_errno(ret); | |
6821 | goto out; | |
6822 | } | |
6823 | ||
6824 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6825 | ret = ocfs2_journal_access_di(handle, | |
6826 | INODE_CACHE(args->new_inode), | |
6827 | args->new_bh, | |
6828 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6829 | if (ret) { | |
6830 | mlog_errno(ret); | |
6831 | goto out_commit; | |
6832 | } | |
6833 | } | |
6834 | ||
6835 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode), | |
6836 | new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6837 | if (ret) { | |
6838 | mlog_errno(ret); | |
6839 | goto out_commit; | |
6840 | } | |
6841 | ||
6842 | memcpy(new_blk_bh->b_data + header_off, blk_bh->b_data + header_off, | |
6843 | osb->sb->s_blocksize - header_off); | |
6844 | ||
6845 | ret = ocfs2_reflink_xattr_header(handle, args, blk_bh, xh, | |
6846 | new_blk_bh, new_xh, &vb, meta_ac, | |
6847 | ocfs2_get_xattr_value_root, NULL); | |
6848 | if (ret) { | |
6849 | mlog_errno(ret); | |
6850 | goto out_commit; | |
6851 | } | |
6852 | ||
6853 | ocfs2_journal_dirty(handle, new_blk_bh); | |
6854 | ||
6855 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6856 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6857 | spin_lock(&new_oi->ip_lock); | |
6858 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
6859 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6860 | spin_unlock(&new_oi->ip_lock); | |
6861 | ||
6862 | ocfs2_journal_dirty(handle, args->new_bh); | |
6863 | } | |
6864 | ||
6865 | out_commit: | |
6866 | ocfs2_commit_trans(osb, handle); | |
6867 | ||
6868 | out: | |
6869 | ocfs2_free_alloc_context(meta_ac); | |
6870 | return ret; | |
6871 | } | |
6872 | ||
6873 | struct ocfs2_reflink_xattr_tree_args { | |
6874 | struct ocfs2_xattr_reflink *reflink; | |
6875 | struct buffer_head *old_blk_bh; | |
6876 | struct buffer_head *new_blk_bh; | |
6877 | struct ocfs2_xattr_bucket *old_bucket; | |
6878 | struct ocfs2_xattr_bucket *new_bucket; | |
6879 | }; | |
6880 | ||
6881 | /* | |
6882 | * NOTE: | |
6883 | * We have to handle the case that both old bucket and new bucket | |
6884 | * will call this function to get the right ret_bh. | |
6885 | * So The caller must give us the right bh. | |
6886 | */ | |
6887 | static int ocfs2_get_reflink_xattr_value_root(struct super_block *sb, | |
6888 | struct buffer_head *bh, | |
6889 | struct ocfs2_xattr_header *xh, | |
6890 | int offset, | |
6891 | struct ocfs2_xattr_value_root **xv, | |
6892 | struct buffer_head **ret_bh, | |
6893 | void *para) | |
6894 | { | |
6895 | struct ocfs2_reflink_xattr_tree_args *args = | |
6896 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6897 | struct ocfs2_xattr_bucket *bucket; | |
6898 | ||
6899 | if (bh == args->old_bucket->bu_bhs[0]) | |
6900 | bucket = args->old_bucket; | |
6901 | else | |
6902 | bucket = args->new_bucket; | |
6903 | ||
6904 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6905 | xv, ret_bh); | |
6906 | } | |
6907 | ||
6908 | struct ocfs2_value_tree_metas { | |
6909 | int num_metas; | |
6910 | int credits; | |
6911 | int num_recs; | |
6912 | }; | |
6913 | ||
6914 | static int ocfs2_value_tree_metas_in_bucket(struct super_block *sb, | |
6915 | struct buffer_head *bh, | |
6916 | struct ocfs2_xattr_header *xh, | |
6917 | int offset, | |
6918 | struct ocfs2_xattr_value_root **xv, | |
6919 | struct buffer_head **ret_bh, | |
6920 | void *para) | |
6921 | { | |
6922 | struct ocfs2_xattr_bucket *bucket = | |
6923 | (struct ocfs2_xattr_bucket *)para; | |
6924 | ||
6925 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6926 | xv, ret_bh); | |
6927 | } | |
6928 | ||
6929 | static int ocfs2_calc_value_tree_metas(struct inode *inode, | |
6930 | struct ocfs2_xattr_bucket *bucket, | |
6931 | void *para) | |
6932 | { | |
6933 | struct ocfs2_value_tree_metas *metas = | |
6934 | (struct ocfs2_value_tree_metas *)para; | |
6935 | struct ocfs2_xattr_header *xh = | |
6936 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6937 | ||
6938 | /* Add the credits for this bucket first. */ | |
6939 | metas->credits += bucket->bu_blocks; | |
6940 | return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0], | |
6941 | xh, &metas->num_metas, | |
6942 | &metas->credits, &metas->num_recs, | |
6943 | ocfs2_value_tree_metas_in_bucket, | |
6944 | bucket); | |
6945 | } | |
6946 | ||
6947 | /* | |
6948 | * Given a xattr extent rec starting from blkno and having len clusters, | |
6949 | * iterate all the buckets calculate how much metadata we need for reflinking | |
6950 | * all the ocfs2_xattr_value_root and lock the allocators accordingly. | |
6951 | */ | |
6952 | static int ocfs2_lock_reflink_xattr_rec_allocators( | |
6953 | struct ocfs2_reflink_xattr_tree_args *args, | |
6954 | struct ocfs2_extent_tree *xt_et, | |
6955 | u64 blkno, u32 len, int *credits, | |
6956 | struct ocfs2_alloc_context **meta_ac, | |
6957 | struct ocfs2_alloc_context **data_ac) | |
6958 | { | |
6959 | int ret, num_free_extents; | |
6960 | struct ocfs2_value_tree_metas metas; | |
6961 | struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); | |
6962 | struct ocfs2_refcount_block *rb; | |
6963 | ||
6964 | memset(&metas, 0, sizeof(metas)); | |
6965 | ||
6966 | ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len, | |
6967 | ocfs2_calc_value_tree_metas, &metas); | |
6968 | if (ret) { | |
6969 | mlog_errno(ret); | |
6970 | goto out; | |
6971 | } | |
6972 | ||
6973 | *credits = metas.credits; | |
6974 | ||
6975 | /* | |
6976 | * Calculate we need for refcount tree change. | |
6977 | * | |
6978 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6979 | * an approximate number we need for refcount tree change. | |
6980 | * Sometimes we need to split the tree, and after split, half recs | |
6981 | * will be moved to the new block, and a new block can only provide | |
6982 | * half number of recs. So we multiple new blocks by 2. | |
6983 | * In the end, we have to add credits for modifying the already | |
6984 | * existed refcount block. | |
6985 | */ | |
6986 | rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data; | |
6987 | metas.num_recs = | |
6988 | (metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) / | |
6989 | ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6990 | metas.num_metas += metas.num_recs; | |
6991 | *credits += metas.num_recs + | |
6992 | metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6993 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6994 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6995 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6996 | else | |
6997 | *credits += 1; | |
6998 | ||
6999 | /* count in the xattr tree change. */ | |
7000 | num_free_extents = ocfs2_num_free_extents(osb, xt_et); | |
7001 | if (num_free_extents < 0) { | |
7002 | ret = num_free_extents; | |
7003 | mlog_errno(ret); | |
7004 | goto out; | |
7005 | } | |
7006 | ||
7007 | if (num_free_extents < len) | |
7008 | metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el); | |
7009 | ||
7010 | *credits += ocfs2_calc_extend_credits(osb->sb, | |
7011 | xt_et->et_root_el, len); | |
7012 | ||
7013 | if (metas.num_metas) { | |
7014 | ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas, | |
7015 | meta_ac); | |
7016 | if (ret) { | |
7017 | mlog_errno(ret); | |
7018 | goto out; | |
7019 | } | |
7020 | } | |
7021 | ||
7022 | if (len) { | |
7023 | ret = ocfs2_reserve_clusters(osb, len, data_ac); | |
7024 | if (ret) | |
7025 | mlog_errno(ret); | |
7026 | } | |
7027 | out: | |
7028 | if (ret) { | |
7029 | if (*meta_ac) { | |
7030 | ocfs2_free_alloc_context(*meta_ac); | |
7031 | meta_ac = NULL; | |
7032 | } | |
7033 | } | |
7034 | ||
7035 | return ret; | |
7036 | } | |
7037 | ||
7038 | static int ocfs2_reflink_xattr_buckets(handle_t *handle, | |
7039 | u64 blkno, u64 new_blkno, u32 clusters, | |
7040 | struct ocfs2_alloc_context *meta_ac, | |
7041 | struct ocfs2_alloc_context *data_ac, | |
7042 | struct ocfs2_reflink_xattr_tree_args *args) | |
7043 | { | |
7044 | int i, j, ret = 0; | |
7045 | struct super_block *sb = args->reflink->old_inode->i_sb; | |
7046 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
7047 | u32 num_buckets = clusters * bpc; | |
7048 | int bpb = args->old_bucket->bu_blocks; | |
7049 | struct ocfs2_xattr_value_buf vb = { | |
7050 | .vb_access = ocfs2_journal_access, | |
7051 | }; | |
7052 | ||
7053 | for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) { | |
7054 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
7055 | if (ret) { | |
7056 | mlog_errno(ret); | |
7057 | break; | |
7058 | } | |
7059 | ||
7060 | ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno); | |
7061 | if (ret) { | |
7062 | mlog_errno(ret); | |
7063 | break; | |
7064 | } | |
7065 | ||
7066 | /* | |
7067 | * The real bucket num in this series of blocks is stored | |
7068 | * in the 1st bucket. | |
7069 | */ | |
7070 | if (i == 0) | |
7071 | num_buckets = le16_to_cpu( | |
7072 | bucket_xh(args->old_bucket)->xh_num_buckets); | |
7073 | ||
7074 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
7075 | args->new_bucket, | |
7076 | OCFS2_JOURNAL_ACCESS_CREATE); | |
7077 | if (ret) { | |
7078 | mlog_errno(ret); | |
7079 | break; | |
7080 | } | |
7081 | ||
7082 | for (j = 0; j < bpb; j++) | |
7083 | memcpy(bucket_block(args->new_bucket, j), | |
7084 | bucket_block(args->old_bucket, j), | |
7085 | sb->s_blocksize); | |
7086 | ||
7087 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
7088 | ||
7089 | ret = ocfs2_reflink_xattr_header(handle, args->reflink, | |
7090 | args->old_bucket->bu_bhs[0], | |
7091 | bucket_xh(args->old_bucket), | |
7092 | args->new_bucket->bu_bhs[0], | |
7093 | bucket_xh(args->new_bucket), | |
7094 | &vb, meta_ac, | |
7095 | ocfs2_get_reflink_xattr_value_root, | |
7096 | args); | |
7097 | if (ret) { | |
7098 | mlog_errno(ret); | |
7099 | break; | |
7100 | } | |
7101 | ||
7102 | /* | |
7103 | * Re-access and dirty the bucket to calculate metaecc. | |
7104 | * Because we may extend the transaction in reflink_xattr_header | |
7105 | * which will let the already accessed block gone. | |
7106 | */ | |
7107 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
7108 | args->new_bucket, | |
7109 | OCFS2_JOURNAL_ACCESS_WRITE); | |
7110 | if (ret) { | |
7111 | mlog_errno(ret); | |
7112 | break; | |
7113 | } | |
7114 | ||
7115 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
7116 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
7117 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
7118 | } | |
7119 | ||
7120 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
7121 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
7122 | return ret; | |
7123 | } | |
7124 | /* | |
7125 | * Create the same xattr extent record in the new inode's xattr tree. | |
7126 | */ | |
7127 | static int ocfs2_reflink_xattr_rec(struct inode *inode, | |
7128 | struct buffer_head *root_bh, | |
7129 | u64 blkno, | |
7130 | u32 cpos, | |
7131 | u32 len, | |
7132 | void *para) | |
7133 | { | |
7134 | int ret, credits = 0; | |
7135 | u32 p_cluster, num_clusters; | |
7136 | u64 new_blkno; | |
7137 | handle_t *handle; | |
7138 | struct ocfs2_reflink_xattr_tree_args *args = | |
7139 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
7140 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
7141 | struct ocfs2_alloc_context *meta_ac = NULL; | |
7142 | struct ocfs2_alloc_context *data_ac = NULL; | |
7143 | struct ocfs2_extent_tree et; | |
7144 | ||
7145 | ocfs2_init_xattr_tree_extent_tree(&et, | |
7146 | INODE_CACHE(args->reflink->new_inode), | |
7147 | args->new_blk_bh); | |
7148 | ||
7149 | ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno, | |
7150 | len, &credits, | |
7151 | &meta_ac, &data_ac); | |
7152 | if (ret) { | |
7153 | mlog_errno(ret); | |
7154 | goto out; | |
7155 | } | |
7156 | ||
7157 | handle = ocfs2_start_trans(osb, credits); | |
7158 | if (IS_ERR(handle)) { | |
7159 | ret = PTR_ERR(handle); | |
7160 | mlog_errno(ret); | |
7161 | goto out; | |
7162 | } | |
7163 | ||
7164 | ret = ocfs2_claim_clusters(osb, handle, data_ac, | |
7165 | len, &p_cluster, &num_clusters); | |
7166 | if (ret) { | |
7167 | mlog_errno(ret); | |
7168 | goto out_commit; | |
7169 | } | |
7170 | ||
7171 | new_blkno = ocfs2_clusters_to_blocks(osb->sb, p_cluster); | |
7172 | ||
7173 | mlog(0, "reflink xattr buckets %llu to %llu, len %u\n", | |
7174 | (unsigned long long)blkno, (unsigned long long)new_blkno, len); | |
7175 | ret = ocfs2_reflink_xattr_buckets(handle, blkno, new_blkno, len, | |
7176 | meta_ac, data_ac, args); | |
7177 | if (ret) { | |
7178 | mlog_errno(ret); | |
7179 | goto out_commit; | |
7180 | } | |
7181 | ||
7182 | mlog(0, "insert new xattr extent rec start %llu len %u to %u\n", | |
7183 | (unsigned long long)new_blkno, len, cpos); | |
7184 | ret = ocfs2_insert_extent(handle, &et, cpos, new_blkno, | |
7185 | len, 0, meta_ac); | |
7186 | if (ret) | |
7187 | mlog_errno(ret); | |
7188 | ||
7189 | out_commit: | |
7190 | ocfs2_commit_trans(osb, handle); | |
7191 | ||
7192 | out: | |
7193 | if (meta_ac) | |
7194 | ocfs2_free_alloc_context(meta_ac); | |
7195 | if (data_ac) | |
7196 | ocfs2_free_alloc_context(data_ac); | |
7197 | return ret; | |
7198 | } | |
7199 | ||
7200 | /* | |
7201 | * Create reflinked xattr buckets. | |
7202 | * We will add bucket one by one, and refcount all the xattrs in the bucket | |
7203 | * if they are stored outside. | |
7204 | */ | |
7205 | static int ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, | |
7206 | struct buffer_head *blk_bh, | |
7207 | struct buffer_head *new_blk_bh) | |
7208 | { | |
7209 | int ret; | |
7210 | struct ocfs2_reflink_xattr_tree_args para; | |
7211 | ||
7212 | memset(¶, 0, sizeof(para)); | |
7213 | para.reflink = args; | |
7214 | para.old_blk_bh = blk_bh; | |
7215 | para.new_blk_bh = new_blk_bh; | |
7216 | ||
7217 | para.old_bucket = ocfs2_xattr_bucket_new(args->old_inode); | |
7218 | if (!para.old_bucket) { | |
7219 | mlog_errno(-ENOMEM); | |
7220 | return -ENOMEM; | |
7221 | } | |
7222 | ||
7223 | para.new_bucket = ocfs2_xattr_bucket_new(args->new_inode); | |
7224 | if (!para.new_bucket) { | |
7225 | ret = -ENOMEM; | |
7226 | mlog_errno(ret); | |
7227 | goto out; | |
7228 | } | |
7229 | ||
7230 | ret = ocfs2_iterate_xattr_index_block(args->old_inode, blk_bh, | |
7231 | ocfs2_reflink_xattr_rec, | |
7232 | ¶); | |
7233 | if (ret) | |
7234 | mlog_errno(ret); | |
7235 | ||
7236 | out: | |
7237 | ocfs2_xattr_bucket_free(para.old_bucket); | |
7238 | ocfs2_xattr_bucket_free(para.new_bucket); | |
7239 | return ret; | |
7240 | } | |
7241 | ||
7242 | static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args, | |
7243 | struct buffer_head *blk_bh) | |
7244 | { | |
7245 | int ret, indexed = 0; | |
7246 | struct buffer_head *new_blk_bh = NULL; | |
7247 | struct ocfs2_xattr_block *xb = | |
7248 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
7249 | ||
7250 | ||
7251 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) | |
7252 | indexed = 1; | |
7253 | ||
7254 | ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh, | |
7255 | &new_blk_bh, indexed); | |
7256 | if (ret) { | |
7257 | mlog_errno(ret); | |
7258 | goto out; | |
7259 | } | |
7260 | ||
7261 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) | |
7262 | ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); | |
7263 | else | |
7264 | ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); | |
7265 | if (ret) | |
7266 | mlog_errno(ret); | |
7267 | ||
7268 | out: | |
7269 | brelse(new_blk_bh); | |
7270 | return ret; | |
7271 | } | |
7272 | ||
0fe9b66c TM |
7273 | static int ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe) |
7274 | { | |
7275 | int type = ocfs2_xattr_get_type(xe); | |
7276 | ||
7277 | return type != OCFS2_XATTR_INDEX_SECURITY && | |
7278 | type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS && | |
7279 | type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT; | |
7280 | } | |
7281 | ||
2999d12f TM |
7282 | int ocfs2_reflink_xattrs(struct inode *old_inode, |
7283 | struct buffer_head *old_bh, | |
7284 | struct inode *new_inode, | |
0fe9b66c TM |
7285 | struct buffer_head *new_bh, |
7286 | bool preserve_security) | |
2999d12f TM |
7287 | { |
7288 | int ret; | |
7289 | struct ocfs2_xattr_reflink args; | |
7290 | struct ocfs2_inode_info *oi = OCFS2_I(old_inode); | |
7291 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)old_bh->b_data; | |
7292 | struct buffer_head *blk_bh = NULL; | |
7293 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
7294 | struct ocfs2_refcount_tree *ref_tree; | |
7295 | struct buffer_head *ref_root_bh = NULL; | |
7296 | ||
7297 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7298 | le64_to_cpu(di->i_refcount_loc), | |
7299 | 1, &ref_tree, &ref_root_bh); | |
7300 | if (ret) { | |
7301 | mlog_errno(ret); | |
7302 | goto out; | |
7303 | } | |
7304 | ||
7305 | ocfs2_init_dealloc_ctxt(&dealloc); | |
7306 | ||
7307 | args.old_inode = old_inode; | |
7308 | args.new_inode = new_inode; | |
7309 | args.old_bh = old_bh; | |
7310 | args.new_bh = new_bh; | |
7311 | args.ref_ci = &ref_tree->rf_ci; | |
7312 | args.ref_root_bh = ref_root_bh; | |
7313 | args.dealloc = &dealloc; | |
0fe9b66c TM |
7314 | if (preserve_security) |
7315 | args.xattr_reflinked = NULL; | |
7316 | else | |
7317 | args.xattr_reflinked = ocfs2_reflink_xattr_no_security; | |
2999d12f TM |
7318 | |
7319 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
7320 | ret = ocfs2_reflink_xattr_inline(&args); | |
7321 | if (ret) { | |
7322 | mlog_errno(ret); | |
7323 | goto out_unlock; | |
7324 | } | |
7325 | } | |
7326 | ||
7327 | if (!di->i_xattr_loc) | |
7328 | goto out_unlock; | |
7329 | ||
7330 | ret = ocfs2_read_xattr_block(old_inode, le64_to_cpu(di->i_xattr_loc), | |
7331 | &blk_bh); | |
7332 | if (ret < 0) { | |
7333 | mlog_errno(ret); | |
7334 | goto out_unlock; | |
7335 | } | |
7336 | ||
7337 | ret = ocfs2_reflink_xattr_in_block(&args, blk_bh); | |
7338 | if (ret) | |
7339 | mlog_errno(ret); | |
7340 | ||
7341 | brelse(blk_bh); | |
7342 | ||
7343 | out_unlock: | |
7344 | ocfs2_unlock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7345 | ref_tree, 1); | |
7346 | brelse(ref_root_bh); | |
7347 | ||
7348 | if (ocfs2_dealloc_has_cluster(&dealloc)) { | |
7349 | ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1); | |
7350 | ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc); | |
7351 | } | |
7352 | ||
7353 | out: | |
7354 | return ret; | |
7355 | } | |
7356 | ||
0fe9b66c TM |
7357 | /* |
7358 | * Initialize security and acl for a already created inode. | |
7359 | * Used for reflink a non-preserve-security file. | |
7360 | * | |
7361 | * It uses common api like ocfs2_xattr_set, so the caller | |
7362 | * must not hold any lock expect i_mutex. | |
7363 | */ | |
7364 | int ocfs2_init_security_and_acl(struct inode *dir, | |
7365 | struct inode *inode) | |
7366 | { | |
7367 | int ret = 0; | |
7368 | struct buffer_head *dir_bh = NULL; | |
7369 | struct ocfs2_security_xattr_info si = { | |
7370 | .enable = 1, | |
7371 | }; | |
7372 | ||
7373 | ret = ocfs2_init_security_get(inode, dir, &si); | |
7374 | if (!ret) { | |
431547b3 CH |
7375 | ret = ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, |
7376 | si.name, si.value, si.value_len, | |
7377 | XATTR_CREATE); | |
0fe9b66c TM |
7378 | if (ret) { |
7379 | mlog_errno(ret); | |
7380 | goto leave; | |
7381 | } | |
7382 | } else if (ret != -EOPNOTSUPP) { | |
7383 | mlog_errno(ret); | |
7384 | goto leave; | |
7385 | } | |
7386 | ||
7387 | ret = ocfs2_inode_lock(dir, &dir_bh, 0); | |
7388 | if (ret) { | |
7389 | mlog_errno(ret); | |
7390 | goto leave; | |
7391 | } | |
7392 | ||
7393 | ret = ocfs2_init_acl(NULL, inode, dir, NULL, dir_bh, NULL, NULL); | |
7394 | if (ret) | |
7395 | mlog_errno(ret); | |
7396 | ||
7397 | ocfs2_inode_unlock(dir, 0); | |
7398 | brelse(dir_bh); | |
7399 | leave: | |
7400 | return ret; | |
7401 | } | |
923f7f31 TY |
7402 | /* |
7403 | * 'security' attributes support | |
7404 | */ | |
431547b3 | 7405 | static size_t ocfs2_xattr_security_list(struct dentry *dentry, char *list, |
923f7f31 | 7406 | size_t list_size, const char *name, |
431547b3 | 7407 | size_t name_len, int type) |
923f7f31 TY |
7408 | { |
7409 | const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN; | |
7410 | const size_t total_len = prefix_len + name_len + 1; | |
7411 | ||
7412 | if (list && total_len <= list_size) { | |
7413 | memcpy(list, XATTR_SECURITY_PREFIX, prefix_len); | |
7414 | memcpy(list + prefix_len, name, name_len); | |
7415 | list[prefix_len + name_len] = '\0'; | |
7416 | } | |
7417 | return total_len; | |
7418 | } | |
7419 | ||
431547b3 CH |
7420 | static int ocfs2_xattr_security_get(struct dentry *dentry, const char *name, |
7421 | void *buffer, size_t size, int type) | |
923f7f31 TY |
7422 | { |
7423 | if (strcmp(name, "") == 0) | |
7424 | return -EINVAL; | |
431547b3 CH |
7425 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7426 | name, buffer, size); | |
923f7f31 TY |
7427 | } |
7428 | ||
431547b3 CH |
7429 | static int ocfs2_xattr_security_set(struct dentry *dentry, const char *name, |
7430 | const void *value, size_t size, int flags, int type) | |
923f7f31 TY |
7431 | { |
7432 | if (strcmp(name, "") == 0) | |
7433 | return -EINVAL; | |
7434 | ||
431547b3 CH |
7435 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7436 | name, value, size, flags); | |
923f7f31 TY |
7437 | } |
7438 | ||
534eaddd TY |
7439 | int ocfs2_init_security_get(struct inode *inode, |
7440 | struct inode *dir, | |
7441 | struct ocfs2_security_xattr_info *si) | |
7442 | { | |
38d59ef6 TY |
7443 | /* check whether ocfs2 support feature xattr */ |
7444 | if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) | |
7445 | return -EOPNOTSUPP; | |
534eaddd TY |
7446 | return security_inode_init_security(inode, dir, &si->name, &si->value, |
7447 | &si->value_len); | |
7448 | } | |
7449 | ||
7450 | int ocfs2_init_security_set(handle_t *handle, | |
7451 | struct inode *inode, | |
7452 | struct buffer_head *di_bh, | |
7453 | struct ocfs2_security_xattr_info *si, | |
7454 | struct ocfs2_alloc_context *xattr_ac, | |
7455 | struct ocfs2_alloc_context *data_ac) | |
7456 | { | |
7457 | return ocfs2_xattr_set_handle(handle, inode, di_bh, | |
7458 | OCFS2_XATTR_INDEX_SECURITY, | |
7459 | si->name, si->value, si->value_len, 0, | |
7460 | xattr_ac, data_ac); | |
7461 | } | |
7462 | ||
923f7f31 TY |
7463 | struct xattr_handler ocfs2_xattr_security_handler = { |
7464 | .prefix = XATTR_SECURITY_PREFIX, | |
7465 | .list = ocfs2_xattr_security_list, | |
7466 | .get = ocfs2_xattr_security_get, | |
7467 | .set = ocfs2_xattr_security_set, | |
7468 | }; | |
7469 | ||
99219aea MF |
7470 | /* |
7471 | * 'trusted' attributes support | |
7472 | */ | |
431547b3 | 7473 | static size_t ocfs2_xattr_trusted_list(struct dentry *dentry, char *list, |
99219aea | 7474 | size_t list_size, const char *name, |
431547b3 | 7475 | size_t name_len, int type) |
99219aea | 7476 | { |
ceb1eba3 | 7477 | const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN; |
99219aea MF |
7478 | const size_t total_len = prefix_len + name_len + 1; |
7479 | ||
7480 | if (list && total_len <= list_size) { | |
7481 | memcpy(list, XATTR_TRUSTED_PREFIX, prefix_len); | |
7482 | memcpy(list + prefix_len, name, name_len); | |
7483 | list[prefix_len + name_len] = '\0'; | |
7484 | } | |
7485 | return total_len; | |
7486 | } | |
7487 | ||
431547b3 CH |
7488 | static int ocfs2_xattr_trusted_get(struct dentry *dentry, const char *name, |
7489 | void *buffer, size_t size, int type) | |
99219aea MF |
7490 | { |
7491 | if (strcmp(name, "") == 0) | |
7492 | return -EINVAL; | |
431547b3 CH |
7493 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7494 | name, buffer, size); | |
99219aea MF |
7495 | } |
7496 | ||
431547b3 CH |
7497 | static int ocfs2_xattr_trusted_set(struct dentry *dentry, const char *name, |
7498 | const void *value, size_t size, int flags, int type) | |
99219aea MF |
7499 | { |
7500 | if (strcmp(name, "") == 0) | |
7501 | return -EINVAL; | |
7502 | ||
431547b3 CH |
7503 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7504 | name, value, size, flags); | |
99219aea MF |
7505 | } |
7506 | ||
7507 | struct xattr_handler ocfs2_xattr_trusted_handler = { | |
7508 | .prefix = XATTR_TRUSTED_PREFIX, | |
7509 | .list = ocfs2_xattr_trusted_list, | |
7510 | .get = ocfs2_xattr_trusted_get, | |
7511 | .set = ocfs2_xattr_trusted_set, | |
7512 | }; | |
7513 | ||
99219aea MF |
7514 | /* |
7515 | * 'user' attributes support | |
7516 | */ | |
431547b3 | 7517 | static size_t ocfs2_xattr_user_list(struct dentry *dentry, char *list, |
99219aea | 7518 | size_t list_size, const char *name, |
431547b3 | 7519 | size_t name_len, int type) |
99219aea | 7520 | { |
ceb1eba3 | 7521 | const size_t prefix_len = XATTR_USER_PREFIX_LEN; |
99219aea | 7522 | const size_t total_len = prefix_len + name_len + 1; |
431547b3 | 7523 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7524 | |
7525 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7526 | return 0; | |
7527 | ||
7528 | if (list && total_len <= list_size) { | |
7529 | memcpy(list, XATTR_USER_PREFIX, prefix_len); | |
7530 | memcpy(list + prefix_len, name, name_len); | |
7531 | list[prefix_len + name_len] = '\0'; | |
7532 | } | |
7533 | return total_len; | |
7534 | } | |
7535 | ||
431547b3 CH |
7536 | static int ocfs2_xattr_user_get(struct dentry *dentry, const char *name, |
7537 | void *buffer, size_t size, int type) | |
99219aea | 7538 | { |
431547b3 | 7539 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7540 | |
7541 | if (strcmp(name, "") == 0) | |
7542 | return -EINVAL; | |
7543 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7544 | return -EOPNOTSUPP; | |
431547b3 | 7545 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_USER, name, |
99219aea MF |
7546 | buffer, size); |
7547 | } | |
7548 | ||
431547b3 CH |
7549 | static int ocfs2_xattr_user_set(struct dentry *dentry, const char *name, |
7550 | const void *value, size_t size, int flags, int type) | |
99219aea | 7551 | { |
431547b3 | 7552 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7553 | |
7554 | if (strcmp(name, "") == 0) | |
7555 | return -EINVAL; | |
7556 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7557 | return -EOPNOTSUPP; | |
7558 | ||
431547b3 CH |
7559 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_USER, |
7560 | name, value, size, flags); | |
99219aea MF |
7561 | } |
7562 | ||
7563 | struct xattr_handler ocfs2_xattr_user_handler = { | |
7564 | .prefix = XATTR_USER_PREFIX, | |
7565 | .list = ocfs2_xattr_user_list, | |
7566 | .get = ocfs2_xattr_user_get, | |
7567 | .set = ocfs2_xattr_user_set, | |
7568 | }; |