]>
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 | #include <cluster/masklog.h> |
41 | ||
42 | #include "ocfs2.h" | |
43 | #include "alloc.h" | |
d6b32bbb | 44 | #include "blockcheck.h" |
f56654c4 TM |
45 | #include "dlmglue.h" |
46 | #include "file.h" | |
cf1d6c76 TY |
47 | #include "symlink.h" |
48 | #include "sysfile.h" | |
f56654c4 TM |
49 | #include "inode.h" |
50 | #include "journal.h" | |
51 | #include "ocfs2_fs.h" | |
52 | #include "suballoc.h" | |
53 | #include "uptodate.h" | |
54 | #include "buffer_head_io.h" | |
0c044f0b | 55 | #include "super.h" |
cf1d6c76 | 56 | #include "xattr.h" |
492a8a33 | 57 | #include "refcounttree.h" |
0fe9b66c | 58 | #include "acl.h" |
402b4183 | 59 | #include "ocfs2_trace.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; | |
5f5261ac | 82 | int set_abort; |
78f30c31 TM |
83 | }; |
84 | ||
cf1d6c76 TY |
85 | #define OCFS2_XATTR_ROOT_SIZE (sizeof(struct ocfs2_xattr_def_value_root)) |
86 | #define OCFS2_XATTR_INLINE_SIZE 80 | |
4442f518 | 87 | #define OCFS2_XATTR_HEADER_GAP 4 |
534eaddd TY |
88 | #define OCFS2_XATTR_FREE_IN_IBODY (OCFS2_MIN_XATTR_INLINE_SIZE \ |
89 | - sizeof(struct ocfs2_xattr_header) \ | |
4442f518 | 90 | - OCFS2_XATTR_HEADER_GAP) |
89c38bd0 TY |
91 | #define OCFS2_XATTR_FREE_IN_BLOCK(ptr) ((ptr)->i_sb->s_blocksize \ |
92 | - sizeof(struct ocfs2_xattr_block) \ | |
93 | - sizeof(struct ocfs2_xattr_header) \ | |
4442f518 | 94 | - OCFS2_XATTR_HEADER_GAP) |
cf1d6c76 TY |
95 | |
96 | static struct ocfs2_xattr_def_value_root def_xv = { | |
97 | .xv.xr_list.l_count = cpu_to_le16(1), | |
98 | }; | |
99 | ||
537d81ca | 100 | const struct xattr_handler *ocfs2_xattr_handlers[] = { |
cf1d6c76 | 101 | &ocfs2_xattr_user_handler, |
702e5bc6 CH |
102 | &posix_acl_access_xattr_handler, |
103 | &posix_acl_default_xattr_handler, | |
cf1d6c76 | 104 | &ocfs2_xattr_trusted_handler, |
923f7f31 | 105 | &ocfs2_xattr_security_handler, |
cf1d6c76 TY |
106 | NULL |
107 | }; | |
108 | ||
537d81ca | 109 | static const struct xattr_handler *ocfs2_xattr_handler_map[OCFS2_XATTR_MAX] = { |
cf1d6c76 | 110 | [OCFS2_XATTR_INDEX_USER] = &ocfs2_xattr_user_handler, |
929fb014 | 111 | [OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS] |
702e5bc6 | 112 | = &posix_acl_access_xattr_handler, |
929fb014 | 113 | [OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT] |
702e5bc6 | 114 | = &posix_acl_default_xattr_handler, |
cf1d6c76 | 115 | [OCFS2_XATTR_INDEX_TRUSTED] = &ocfs2_xattr_trusted_handler, |
923f7f31 | 116 | [OCFS2_XATTR_INDEX_SECURITY] = &ocfs2_xattr_security_handler, |
cf1d6c76 TY |
117 | }; |
118 | ||
119 | struct ocfs2_xattr_info { | |
6b240ff6 JB |
120 | int xi_name_index; |
121 | const char *xi_name; | |
18853b95 | 122 | int xi_name_len; |
6b240ff6 JB |
123 | const void *xi_value; |
124 | size_t xi_value_len; | |
cf1d6c76 TY |
125 | }; |
126 | ||
127 | struct ocfs2_xattr_search { | |
128 | struct buffer_head *inode_bh; | |
129 | /* | |
130 | * xattr_bh point to the block buffer head which has extended attribute | |
131 | * when extended attribute in inode, xattr_bh is equal to inode_bh. | |
132 | */ | |
133 | struct buffer_head *xattr_bh; | |
134 | struct ocfs2_xattr_header *header; | |
ba937127 | 135 | struct ocfs2_xattr_bucket *bucket; |
cf1d6c76 TY |
136 | void *base; |
137 | void *end; | |
138 | struct ocfs2_xattr_entry *here; | |
139 | int not_found; | |
140 | }; | |
141 | ||
11179f2c JB |
142 | /* Operations on struct ocfs2_xa_entry */ |
143 | struct ocfs2_xa_loc; | |
144 | struct ocfs2_xa_loc_operations { | |
cf2bc809 JB |
145 | /* |
146 | * Journal functions | |
147 | */ | |
148 | int (*xlo_journal_access)(handle_t *handle, struct ocfs2_xa_loc *loc, | |
149 | int type); | |
150 | void (*xlo_journal_dirty)(handle_t *handle, struct ocfs2_xa_loc *loc); | |
151 | ||
11179f2c JB |
152 | /* |
153 | * Return a pointer to the appropriate buffer in loc->xl_storage | |
154 | * at the given offset from loc->xl_header. | |
155 | */ | |
156 | void *(*xlo_offset_pointer)(struct ocfs2_xa_loc *loc, int offset); | |
157 | ||
69a3e539 JB |
158 | /* Can we reuse the existing entry for the new value? */ |
159 | int (*xlo_can_reuse)(struct ocfs2_xa_loc *loc, | |
160 | struct ocfs2_xattr_info *xi); | |
161 | ||
162 | /* How much space is needed for the new value? */ | |
163 | int (*xlo_check_space)(struct ocfs2_xa_loc *loc, | |
164 | struct ocfs2_xattr_info *xi); | |
165 | ||
166 | /* | |
167 | * Return the offset of the first name+value pair. This is | |
168 | * the start of our downward-filling free space. | |
169 | */ | |
170 | int (*xlo_get_free_start)(struct ocfs2_xa_loc *loc); | |
171 | ||
11179f2c JB |
172 | /* |
173 | * Remove the name+value at this location. Do whatever is | |
174 | * appropriate with the remaining name+value pairs. | |
175 | */ | |
176 | void (*xlo_wipe_namevalue)(struct ocfs2_xa_loc *loc); | |
69a3e539 JB |
177 | |
178 | /* Fill xl_entry with a new entry */ | |
179 | void (*xlo_add_entry)(struct ocfs2_xa_loc *loc, u32 name_hash); | |
180 | ||
181 | /* Add name+value storage to an entry */ | |
182 | void (*xlo_add_namevalue)(struct ocfs2_xa_loc *loc, int size); | |
3fc12afa JB |
183 | |
184 | /* | |
185 | * Initialize the value buf's access and bh fields for this entry. | |
186 | * ocfs2_xa_fill_value_buf() will handle the xv pointer. | |
187 | */ | |
188 | void (*xlo_fill_value_buf)(struct ocfs2_xa_loc *loc, | |
189 | struct ocfs2_xattr_value_buf *vb); | |
11179f2c JB |
190 | }; |
191 | ||
192 | /* | |
193 | * Describes an xattr entry location. This is a memory structure | |
194 | * tracking the on-disk structure. | |
195 | */ | |
196 | struct ocfs2_xa_loc { | |
cf2bc809 JB |
197 | /* This xattr belongs to this inode */ |
198 | struct inode *xl_inode; | |
199 | ||
11179f2c JB |
200 | /* The ocfs2_xattr_header inside the on-disk storage. Not NULL. */ |
201 | struct ocfs2_xattr_header *xl_header; | |
202 | ||
203 | /* Bytes from xl_header to the end of the storage */ | |
204 | int xl_size; | |
205 | ||
206 | /* | |
207 | * The ocfs2_xattr_entry this location describes. If this is | |
208 | * NULL, this location describes the on-disk structure where it | |
209 | * would have been. | |
210 | */ | |
211 | struct ocfs2_xattr_entry *xl_entry; | |
212 | ||
213 | /* | |
214 | * Internal housekeeping | |
215 | */ | |
216 | ||
217 | /* Buffer(s) containing this entry */ | |
218 | void *xl_storage; | |
219 | ||
220 | /* Operations on the storage backing this location */ | |
221 | const struct ocfs2_xa_loc_operations *xl_ops; | |
222 | }; | |
223 | ||
199799a3 JB |
224 | /* |
225 | * Convenience functions to calculate how much space is needed for a | |
226 | * given name+value pair | |
227 | */ | |
228 | static int namevalue_size(int name_len, uint64_t value_len) | |
229 | { | |
230 | if (value_len > OCFS2_XATTR_INLINE_SIZE) | |
231 | return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_ROOT_SIZE; | |
232 | else | |
233 | return OCFS2_XATTR_SIZE(name_len) + OCFS2_XATTR_SIZE(value_len); | |
234 | } | |
235 | ||
236 | static int namevalue_size_xi(struct ocfs2_xattr_info *xi) | |
237 | { | |
238 | return namevalue_size(xi->xi_name_len, xi->xi_value_len); | |
239 | } | |
240 | ||
241 | static int namevalue_size_xe(struct ocfs2_xattr_entry *xe) | |
242 | { | |
243 | u64 value_len = le64_to_cpu(xe->xe_value_size); | |
244 | ||
245 | BUG_ON((value_len > OCFS2_XATTR_INLINE_SIZE) && | |
246 | ocfs2_xattr_is_local(xe)); | |
247 | return namevalue_size(xe->xe_name_len, value_len); | |
248 | } | |
249 | ||
250 | ||
fd68a894 | 251 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
589dc260 TM |
252 | struct ocfs2_xattr_header *xh, |
253 | int index, | |
254 | int *block_off, | |
255 | int *new_offset); | |
256 | ||
54f443f4 JB |
257 | static int ocfs2_xattr_block_find(struct inode *inode, |
258 | int name_index, | |
259 | const char *name, | |
260 | struct ocfs2_xattr_search *xs); | |
589dc260 TM |
261 | static int ocfs2_xattr_index_block_find(struct inode *inode, |
262 | struct buffer_head *root_bh, | |
263 | int name_index, | |
264 | const char *name, | |
265 | struct ocfs2_xattr_search *xs); | |
266 | ||
0c044f0b | 267 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, |
47bca495 | 268 | struct buffer_head *blk_bh, |
0c044f0b TM |
269 | char *buffer, |
270 | size_t buffer_size); | |
271 | ||
01225596 | 272 | static int ocfs2_xattr_create_index_block(struct inode *inode, |
78f30c31 TM |
273 | struct ocfs2_xattr_search *xs, |
274 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 TM |
275 | |
276 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, | |
277 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
278 | struct ocfs2_xattr_search *xs, |
279 | struct ocfs2_xattr_set_ctxt *ctxt); | |
01225596 | 280 | |
47bca495 TM |
281 | typedef int (xattr_tree_rec_func)(struct inode *inode, |
282 | struct buffer_head *root_bh, | |
283 | u64 blkno, u32 cpos, u32 len, void *para); | |
284 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, | |
285 | struct buffer_head *root_bh, | |
286 | xattr_tree_rec_func *rec_func, | |
287 | void *para); | |
288 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
289 | struct ocfs2_xattr_bucket *bucket, | |
290 | void *para); | |
291 | static int ocfs2_rm_xattr_cluster(struct inode *inode, | |
292 | struct buffer_head *root_bh, | |
293 | u64 blkno, | |
294 | u32 cpos, | |
295 | u32 len, | |
296 | void *para); | |
297 | ||
c58b6032 JB |
298 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
299 | u64 src_blk, u64 last_blk, u64 to_blk, | |
300 | unsigned int start_bucket, | |
301 | u32 *first_hash); | |
492a8a33 TM |
302 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, |
303 | struct ocfs2_dinode *di, | |
304 | struct ocfs2_xattr_info *xi, | |
305 | struct ocfs2_xattr_search *xis, | |
306 | struct ocfs2_xattr_search *xbs, | |
307 | struct ocfs2_refcount_tree **ref_tree, | |
308 | int *meta_need, | |
309 | int *credits); | |
ce9c5a54 TM |
310 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, |
311 | struct ocfs2_xattr_bucket *bucket, | |
312 | int offset, | |
313 | struct ocfs2_xattr_value_root **xv, | |
314 | struct buffer_head **bh); | |
a3944256 | 315 | |
0030e001 TY |
316 | static inline u16 ocfs2_xattr_buckets_per_cluster(struct ocfs2_super *osb) |
317 | { | |
318 | return (1 << osb->s_clustersize_bits) / OCFS2_XATTR_BUCKET_SIZE; | |
319 | } | |
320 | ||
321 | static inline u16 ocfs2_blocks_per_xattr_bucket(struct super_block *sb) | |
322 | { | |
323 | return OCFS2_XATTR_BUCKET_SIZE / (1 << sb->s_blocksize_bits); | |
324 | } | |
325 | ||
9c7759aa | 326 | #define bucket_blkno(_b) ((_b)->bu_bhs[0]->b_blocknr) |
51def39f | 327 | #define bucket_block(_b, _n) ((_b)->bu_bhs[(_n)]->b_data) |
3e632946 | 328 | #define bucket_xh(_b) ((struct ocfs2_xattr_header *)bucket_block((_b), 0)) |
9c7759aa | 329 | |
ba937127 | 330 | static struct ocfs2_xattr_bucket *ocfs2_xattr_bucket_new(struct inode *inode) |
6dde41d9 | 331 | { |
ba937127 JB |
332 | struct ocfs2_xattr_bucket *bucket; |
333 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
6dde41d9 | 334 | |
ba937127 JB |
335 | BUG_ON(blks > OCFS2_XATTR_MAX_BLOCKS_PER_BUCKET); |
336 | ||
337 | bucket = kzalloc(sizeof(struct ocfs2_xattr_bucket), GFP_NOFS); | |
338 | if (bucket) { | |
339 | bucket->bu_inode = inode; | |
340 | bucket->bu_blocks = blks; | |
341 | } | |
342 | ||
343 | return bucket; | |
344 | } | |
345 | ||
346 | static void ocfs2_xattr_bucket_relse(struct ocfs2_xattr_bucket *bucket) | |
347 | { | |
348 | int i; | |
349 | ||
350 | for (i = 0; i < bucket->bu_blocks; i++) { | |
6dde41d9 JB |
351 | brelse(bucket->bu_bhs[i]); |
352 | bucket->bu_bhs[i] = NULL; | |
353 | } | |
354 | } | |
355 | ||
ba937127 JB |
356 | static void ocfs2_xattr_bucket_free(struct ocfs2_xattr_bucket *bucket) |
357 | { | |
358 | if (bucket) { | |
359 | ocfs2_xattr_bucket_relse(bucket); | |
360 | bucket->bu_inode = NULL; | |
361 | kfree(bucket); | |
362 | } | |
363 | } | |
364 | ||
784b816a JB |
365 | /* |
366 | * A bucket that has never been written to disk doesn't need to be | |
367 | * read. We just need the buffer_heads. Don't call this for | |
368 | * buckets that are already on disk. ocfs2_read_xattr_bucket() initializes | |
369 | * them fully. | |
370 | */ | |
ba937127 | 371 | static int ocfs2_init_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
9c339255 | 372 | u64 xb_blkno, int new) |
784b816a JB |
373 | { |
374 | int i, rc = 0; | |
784b816a | 375 | |
ba937127 JB |
376 | for (i = 0; i < bucket->bu_blocks; i++) { |
377 | bucket->bu_bhs[i] = sb_getblk(bucket->bu_inode->i_sb, | |
378 | xb_blkno + i); | |
784b816a | 379 | if (!bucket->bu_bhs[i]) { |
7391a294 | 380 | rc = -ENOMEM; |
784b816a JB |
381 | mlog_errno(rc); |
382 | break; | |
383 | } | |
384 | ||
8cb471e8 | 385 | if (!ocfs2_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
9c339255 WW |
386 | bucket->bu_bhs[i])) { |
387 | if (new) | |
388 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode), | |
389 | bucket->bu_bhs[i]); | |
390 | else { | |
391 | set_buffer_uptodate(bucket->bu_bhs[i]); | |
392 | ocfs2_set_buffer_uptodate(INODE_CACHE(bucket->bu_inode), | |
393 | bucket->bu_bhs[i]); | |
394 | } | |
395 | } | |
784b816a JB |
396 | } |
397 | ||
398 | if (rc) | |
ba937127 | 399 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
400 | return rc; |
401 | } | |
402 | ||
403 | /* Read the xattr bucket at xb_blkno */ | |
ba937127 | 404 | static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
405 | u64 xb_blkno) |
406 | { | |
ba937127 | 407 | int rc; |
784b816a | 408 | |
8cb471e8 | 409 | rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno, |
970e4936 JB |
410 | bucket->bu_blocks, bucket->bu_bhs, 0, |
411 | NULL); | |
4d0e214e | 412 | if (!rc) { |
c8b9cf9a | 413 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
414 | rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb, |
415 | bucket->bu_bhs, | |
416 | bucket->bu_blocks, | |
417 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 418 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
419 | if (rc) |
420 | mlog_errno(rc); | |
421 | } | |
422 | ||
784b816a | 423 | if (rc) |
ba937127 | 424 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
425 | return rc; |
426 | } | |
427 | ||
1224be02 | 428 | static int ocfs2_xattr_bucket_journal_access(handle_t *handle, |
1224be02 JB |
429 | struct ocfs2_xattr_bucket *bucket, |
430 | int type) | |
431 | { | |
432 | int i, rc = 0; | |
1224be02 | 433 | |
ba937127 | 434 | for (i = 0; i < bucket->bu_blocks; i++) { |
0cf2f763 JB |
435 | rc = ocfs2_journal_access(handle, |
436 | INODE_CACHE(bucket->bu_inode), | |
1224be02 JB |
437 | bucket->bu_bhs[i], type); |
438 | if (rc) { | |
439 | mlog_errno(rc); | |
440 | break; | |
441 | } | |
442 | } | |
443 | ||
444 | return rc; | |
445 | } | |
446 | ||
447 | static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle, | |
1224be02 JB |
448 | struct ocfs2_xattr_bucket *bucket) |
449 | { | |
ba937127 | 450 | int i; |
1224be02 | 451 | |
c8b9cf9a | 452 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
453 | ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb, |
454 | bucket->bu_bhs, bucket->bu_blocks, | |
455 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 456 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e | 457 | |
ba937127 | 458 | for (i = 0; i < bucket->bu_blocks; i++) |
1224be02 JB |
459 | ocfs2_journal_dirty(handle, bucket->bu_bhs[i]); |
460 | } | |
461 | ||
ba937127 | 462 | static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest, |
4980c6da JB |
463 | struct ocfs2_xattr_bucket *src) |
464 | { | |
465 | int i; | |
ba937127 JB |
466 | int blocksize = src->bu_inode->i_sb->s_blocksize; |
467 | ||
468 | BUG_ON(dest->bu_blocks != src->bu_blocks); | |
469 | BUG_ON(dest->bu_inode != src->bu_inode); | |
4980c6da | 470 | |
ba937127 | 471 | for (i = 0; i < src->bu_blocks; i++) { |
4980c6da JB |
472 | memcpy(bucket_block(dest, i), bucket_block(src, i), |
473 | blocksize); | |
474 | } | |
475 | } | |
1224be02 | 476 | |
4ae1d69b JB |
477 | static int ocfs2_validate_xattr_block(struct super_block *sb, |
478 | struct buffer_head *bh) | |
479 | { | |
d6b32bbb | 480 | int rc; |
4ae1d69b JB |
481 | struct ocfs2_xattr_block *xb = |
482 | (struct ocfs2_xattr_block *)bh->b_data; | |
483 | ||
402b4183 | 484 | trace_ocfs2_validate_xattr_block((unsigned long long)bh->b_blocknr); |
4ae1d69b | 485 | |
d6b32bbb JB |
486 | BUG_ON(!buffer_uptodate(bh)); |
487 | ||
488 | /* | |
489 | * If the ecc fails, we return the error but otherwise | |
490 | * leave the filesystem running. We know any error is | |
491 | * local to this block. | |
492 | */ | |
493 | rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check); | |
494 | if (rc) | |
495 | return rc; | |
496 | ||
497 | /* | |
498 | * Errors after here are fatal | |
499 | */ | |
500 | ||
4ae1d69b | 501 | if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) { |
17a5b9ab | 502 | return ocfs2_error(sb, |
7ecef14a JP |
503 | "Extended attribute block #%llu has bad signature %.*s\n", |
504 | (unsigned long long)bh->b_blocknr, 7, | |
505 | xb->xb_signature); | |
4ae1d69b JB |
506 | } |
507 | ||
508 | if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) { | |
17a5b9ab | 509 | return ocfs2_error(sb, |
7ecef14a JP |
510 | "Extended attribute block #%llu has an invalid xb_blkno of %llu\n", |
511 | (unsigned long long)bh->b_blocknr, | |
512 | (unsigned long long)le64_to_cpu(xb->xb_blkno)); | |
4ae1d69b JB |
513 | } |
514 | ||
515 | if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) { | |
17a5b9ab | 516 | return ocfs2_error(sb, |
7ecef14a JP |
517 | "Extended attribute block #%llu has an invalid xb_fs_generation of #%u\n", |
518 | (unsigned long long)bh->b_blocknr, | |
519 | le32_to_cpu(xb->xb_fs_generation)); | |
4ae1d69b JB |
520 | } |
521 | ||
522 | return 0; | |
523 | } | |
524 | ||
525 | static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, | |
526 | struct buffer_head **bh) | |
527 | { | |
528 | int rc; | |
529 | struct buffer_head *tmp = *bh; | |
530 | ||
8cb471e8 | 531 | rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp, |
970e4936 | 532 | ocfs2_validate_xattr_block); |
4ae1d69b JB |
533 | |
534 | /* If ocfs2_read_block() got us a new bh, pass it up. */ | |
535 | if (!rc && !*bh) | |
536 | *bh = tmp; | |
537 | ||
538 | return rc; | |
539 | } | |
540 | ||
936b8834 | 541 | static inline const char *ocfs2_xattr_prefix(int name_index) |
cf1d6c76 | 542 | { |
537d81ca | 543 | const struct xattr_handler *handler = NULL; |
cf1d6c76 TY |
544 | |
545 | if (name_index > 0 && name_index < OCFS2_XATTR_MAX) | |
546 | handler = ocfs2_xattr_handler_map[name_index]; | |
98e9cb57 | 547 | return handler ? xattr_prefix(handler) : NULL; |
cf1d6c76 TY |
548 | } |
549 | ||
40daa16a | 550 | static u32 ocfs2_xattr_name_hash(struct inode *inode, |
2057e5c6 | 551 | const char *name, |
40daa16a | 552 | int name_len) |
cf1d6c76 TY |
553 | { |
554 | /* Get hash value of uuid from super block */ | |
555 | u32 hash = OCFS2_SB(inode->i_sb)->uuid_hash; | |
556 | int i; | |
557 | ||
cf1d6c76 TY |
558 | /* hash extended attribute name */ |
559 | for (i = 0; i < name_len; i++) { | |
560 | hash = (hash << OCFS2_HASH_SHIFT) ^ | |
561 | (hash >> (8*sizeof(hash) - OCFS2_HASH_SHIFT)) ^ | |
562 | *name++; | |
563 | } | |
564 | ||
565 | return hash; | |
566 | } | |
567 | ||
534eaddd TY |
568 | static int ocfs2_xattr_entry_real_size(int name_len, size_t value_len) |
569 | { | |
199799a3 JB |
570 | return namevalue_size(name_len, value_len) + |
571 | sizeof(struct ocfs2_xattr_entry); | |
572 | } | |
534eaddd | 573 | |
199799a3 JB |
574 | static int ocfs2_xi_entry_usage(struct ocfs2_xattr_info *xi) |
575 | { | |
576 | return namevalue_size_xi(xi) + | |
577 | sizeof(struct ocfs2_xattr_entry); | |
578 | } | |
534eaddd | 579 | |
199799a3 JB |
580 | static int ocfs2_xe_entry_usage(struct ocfs2_xattr_entry *xe) |
581 | { | |
582 | return namevalue_size_xe(xe) + | |
583 | sizeof(struct ocfs2_xattr_entry); | |
534eaddd TY |
584 | } |
585 | ||
586 | int ocfs2_calc_security_init(struct inode *dir, | |
587 | struct ocfs2_security_xattr_info *si, | |
588 | int *want_clusters, | |
589 | int *xattr_credits, | |
590 | struct ocfs2_alloc_context **xattr_ac) | |
591 | { | |
592 | int ret = 0; | |
593 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
594 | int s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
595 | si->value_len); | |
596 | ||
597 | /* | |
598 | * The max space of security xattr taken inline is | |
599 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
600 | * So reserve one metadata block for it is ok. | |
601 | */ | |
602 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
603 | s_size > OCFS2_XATTR_FREE_IN_IBODY) { | |
604 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, xattr_ac); | |
605 | if (ret) { | |
606 | mlog_errno(ret); | |
607 | return ret; | |
608 | } | |
609 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; | |
610 | } | |
611 | ||
612 | /* reserve clusters for xattr value which will be set in B tree*/ | |
0e445b6f TY |
613 | if (si->value_len > OCFS2_XATTR_INLINE_SIZE) { |
614 | int new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
615 | si->value_len); | |
616 | ||
617 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
618 | new_clusters); | |
619 | *want_clusters += new_clusters; | |
620 | } | |
534eaddd TY |
621 | return ret; |
622 | } | |
623 | ||
89c38bd0 TY |
624 | int ocfs2_calc_xattr_init(struct inode *dir, |
625 | struct buffer_head *dir_bh, | |
67697cbd | 626 | umode_t mode, |
89c38bd0 TY |
627 | struct ocfs2_security_xattr_info *si, |
628 | int *want_clusters, | |
629 | int *xattr_credits, | |
9b7895ef | 630 | int *want_meta) |
89c38bd0 TY |
631 | { |
632 | int ret = 0; | |
633 | struct ocfs2_super *osb = OCFS2_SB(dir->i_sb); | |
0e445b6f | 634 | int s_size = 0, a_size = 0, acl_len = 0, new_clusters; |
89c38bd0 TY |
635 | |
636 | if (si->enable) | |
637 | s_size = ocfs2_xattr_entry_real_size(strlen(si->name), | |
638 | si->value_len); | |
639 | ||
640 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) { | |
641 | acl_len = ocfs2_xattr_get_nolock(dir, dir_bh, | |
642 | OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT, | |
643 | "", NULL, 0); | |
644 | if (acl_len > 0) { | |
645 | a_size = ocfs2_xattr_entry_real_size(0, acl_len); | |
646 | if (S_ISDIR(mode)) | |
647 | a_size <<= 1; | |
648 | } else if (acl_len != 0 && acl_len != -ENODATA) { | |
c0a1a6d7 | 649 | ret = acl_len; |
89c38bd0 TY |
650 | mlog_errno(ret); |
651 | return ret; | |
652 | } | |
653 | } | |
654 | ||
655 | if (!(s_size + a_size)) | |
656 | return ret; | |
657 | ||
658 | /* | |
659 | * The max space of security xattr taken inline is | |
660 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
661 | * The max space of acl xattr taken inline is | |
662 | * 80(value) + 16(entry) * 2(if directory) = 192 bytes, | |
663 | * when blocksize = 512, may reserve one more cluser for | |
664 | * xattr bucket, otherwise reserve one metadata block | |
665 | * for them is ok. | |
6c9fd1dc TY |
666 | * If this is a new directory with inline data, |
667 | * we choose to reserve the entire inline area for | |
668 | * directory contents and force an external xattr block. | |
89c38bd0 TY |
669 | */ |
670 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
6c9fd1dc | 671 | (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) || |
89c38bd0 | 672 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) { |
9b7895ef | 673 | *want_meta = *want_meta + 1; |
89c38bd0 TY |
674 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
675 | } | |
676 | ||
677 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE && | |
678 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) { | |
679 | *want_clusters += 1; | |
680 | *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb); | |
681 | } | |
682 | ||
0e445b6f TY |
683 | /* |
684 | * reserve credits and clusters for xattrs which has large value | |
685 | * and have to be set outside | |
686 | */ | |
687 | if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
688 | new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
689 | si->value_len); | |
690 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
691 | new_clusters); | |
692 | *want_clusters += new_clusters; | |
693 | } | |
89c38bd0 TY |
694 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL && |
695 | acl_len > OCFS2_XATTR_INLINE_SIZE) { | |
0e445b6f TY |
696 | /* for directory, it has DEFAULT and ACCESS two types of acls */ |
697 | new_clusters = (S_ISDIR(mode) ? 2 : 1) * | |
698 | ocfs2_clusters_for_bytes(dir->i_sb, acl_len); | |
699 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
700 | new_clusters); | |
701 | *want_clusters += new_clusters; | |
89c38bd0 TY |
702 | } |
703 | ||
704 | return ret; | |
705 | } | |
706 | ||
f56654c4 TM |
707 | static int ocfs2_xattr_extend_allocation(struct inode *inode, |
708 | u32 clusters_to_add, | |
19b801f4 | 709 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 710 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 | 711 | { |
a78f9f46 | 712 | int status = 0, credits; |
85db90e7 | 713 | handle_t *handle = ctxt->handle; |
f56654c4 | 714 | enum ocfs2_alloc_restarted why; |
19b801f4 | 715 | u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters); |
f99b9b7c | 716 | struct ocfs2_extent_tree et; |
f56654c4 | 717 | |
5e404e9e | 718 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f99b9b7c | 719 | |
a78f9f46 | 720 | while (clusters_to_add) { |
402b4183 TM |
721 | trace_ocfs2_xattr_extend_allocation(clusters_to_add); |
722 | ||
a78f9f46 TM |
723 | status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
724 | OCFS2_JOURNAL_ACCESS_WRITE); | |
725 | if (status < 0) { | |
726 | mlog_errno(status); | |
727 | break; | |
728 | } | |
f56654c4 | 729 | |
a78f9f46 TM |
730 | prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
731 | status = ocfs2_add_clusters_in_btree(handle, | |
732 | &et, | |
733 | &logical_start, | |
734 | clusters_to_add, | |
735 | 0, | |
736 | ctxt->data_ac, | |
737 | ctxt->meta_ac, | |
738 | &why); | |
739 | if ((status < 0) && (status != -EAGAIN)) { | |
740 | if (status != -ENOSPC) | |
741 | mlog_errno(status); | |
742 | break; | |
743 | } | |
f56654c4 | 744 | |
a78f9f46 TM |
745 | ocfs2_journal_dirty(handle, vb->vb_bh); |
746 | ||
747 | clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - | |
748 | prev_clusters; | |
749 | ||
750 | if (why != RESTART_NONE && clusters_to_add) { | |
751 | /* | |
752 | * We can only fail in case the alloc file doesn't give | |
753 | * up enough clusters. | |
754 | */ | |
755 | BUG_ON(why == RESTART_META); | |
756 | ||
a78f9f46 | 757 | credits = ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 758 | &vb->vb_xv->xr_list); |
a78f9f46 TM |
759 | status = ocfs2_extend_trans(handle, credits); |
760 | if (status < 0) { | |
761 | status = -ENOMEM; | |
762 | mlog_errno(status); | |
763 | break; | |
764 | } | |
765 | } | |
766 | } | |
f56654c4 TM |
767 | |
768 | return status; | |
769 | } | |
770 | ||
771 | static int __ocfs2_remove_xattr_range(struct inode *inode, | |
d72cc72d | 772 | struct ocfs2_xattr_value_buf *vb, |
f56654c4 | 773 | u32 cpos, u32 phys_cpos, u32 len, |
492a8a33 | 774 | unsigned int ext_flags, |
78f30c31 | 775 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
776 | { |
777 | int ret; | |
778 | u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
85db90e7 | 779 | handle_t *handle = ctxt->handle; |
f99b9b7c JB |
780 | struct ocfs2_extent_tree et; |
781 | ||
5e404e9e | 782 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f56654c4 | 783 | |
0cf2f763 | 784 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
d72cc72d | 785 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
786 | if (ret) { |
787 | mlog_errno(ret); | |
85db90e7 | 788 | goto out; |
f56654c4 TM |
789 | } |
790 | ||
dbdcf6a4 | 791 | ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac, |
78f30c31 | 792 | &ctxt->dealloc); |
f56654c4 TM |
793 | if (ret) { |
794 | mlog_errno(ret); | |
85db90e7 | 795 | goto out; |
f56654c4 TM |
796 | } |
797 | ||
d72cc72d | 798 | le32_add_cpu(&vb->vb_xv->xr_clusters, -len); |
ec20cec7 | 799 | ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 | 800 | |
492a8a33 TM |
801 | if (ext_flags & OCFS2_EXT_REFCOUNTED) |
802 | ret = ocfs2_decrease_refcount(inode, handle, | |
803 | ocfs2_blocks_to_clusters(inode->i_sb, | |
804 | phys_blkno), | |
805 | len, ctxt->meta_ac, &ctxt->dealloc, 1); | |
806 | else | |
807 | ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, | |
808 | phys_blkno, len); | |
f56654c4 TM |
809 | if (ret) |
810 | mlog_errno(ret); | |
811 | ||
f56654c4 | 812 | out: |
f56654c4 TM |
813 | return ret; |
814 | } | |
815 | ||
816 | static int ocfs2_xattr_shrink_size(struct inode *inode, | |
817 | u32 old_clusters, | |
818 | u32 new_clusters, | |
19b801f4 | 819 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 820 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
821 | { |
822 | int ret = 0; | |
492a8a33 | 823 | unsigned int ext_flags; |
f56654c4 TM |
824 | u32 trunc_len, cpos, phys_cpos, alloc_size; |
825 | u64 block; | |
f56654c4 TM |
826 | |
827 | if (old_clusters <= new_clusters) | |
828 | return 0; | |
829 | ||
830 | cpos = new_clusters; | |
831 | trunc_len = old_clusters - new_clusters; | |
832 | while (trunc_len) { | |
833 | ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos, | |
d72cc72d | 834 | &alloc_size, |
492a8a33 | 835 | &vb->vb_xv->xr_list, &ext_flags); |
f56654c4 TM |
836 | if (ret) { |
837 | mlog_errno(ret); | |
838 | goto out; | |
839 | } | |
840 | ||
841 | if (alloc_size > trunc_len) | |
842 | alloc_size = trunc_len; | |
843 | ||
19b801f4 | 844 | ret = __ocfs2_remove_xattr_range(inode, vb, cpos, |
f56654c4 | 845 | phys_cpos, alloc_size, |
492a8a33 | 846 | ext_flags, ctxt); |
f56654c4 TM |
847 | if (ret) { |
848 | mlog_errno(ret); | |
849 | goto out; | |
850 | } | |
851 | ||
852 | block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
8cb471e8 JB |
853 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), |
854 | block, alloc_size); | |
f56654c4 TM |
855 | cpos += alloc_size; |
856 | trunc_len -= alloc_size; | |
857 | } | |
858 | ||
859 | out: | |
f56654c4 TM |
860 | return ret; |
861 | } | |
862 | ||
863 | static int ocfs2_xattr_value_truncate(struct inode *inode, | |
b3e5d379 | 864 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 TM |
865 | int len, |
866 | struct ocfs2_xattr_set_ctxt *ctxt) | |
f56654c4 TM |
867 | { |
868 | int ret; | |
869 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len); | |
b3e5d379 | 870 | u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
f56654c4 TM |
871 | |
872 | if (new_clusters == old_clusters) | |
873 | return 0; | |
874 | ||
875 | if (new_clusters > old_clusters) | |
876 | ret = ocfs2_xattr_extend_allocation(inode, | |
877 | new_clusters - old_clusters, | |
b3e5d379 | 878 | vb, ctxt); |
f56654c4 TM |
879 | else |
880 | ret = ocfs2_xattr_shrink_size(inode, | |
881 | old_clusters, new_clusters, | |
b3e5d379 | 882 | vb, ctxt); |
f56654c4 TM |
883 | |
884 | return ret; | |
885 | } | |
cf1d6c76 | 886 | |
1046cb11 AG |
887 | static int ocfs2_xattr_list_entry(struct super_block *sb, |
888 | char *buffer, size_t size, | |
889 | size_t *result, int type, | |
936b8834 TM |
890 | const char *name, int name_len) |
891 | { | |
892 | char *p = buffer + *result; | |
1046cb11 AG |
893 | const char *prefix; |
894 | int prefix_len; | |
895 | int total_len; | |
936b8834 | 896 | |
1046cb11 AG |
897 | switch(type) { |
898 | case OCFS2_XATTR_INDEX_USER: | |
899 | if (OCFS2_SB(sb)->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
900 | return 0; | |
901 | break; | |
902 | ||
903 | case OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS: | |
904 | case OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT: | |
1751e8a6 | 905 | if (!(sb->s_flags & SB_POSIXACL)) |
1046cb11 AG |
906 | return 0; |
907 | break; | |
908 | ||
909 | case OCFS2_XATTR_INDEX_TRUSTED: | |
910 | if (!capable(CAP_SYS_ADMIN)) | |
911 | return 0; | |
912 | break; | |
913 | } | |
914 | ||
915 | prefix = ocfs2_xattr_prefix(type); | |
916 | if (!prefix) | |
917 | return 0; | |
918 | prefix_len = strlen(prefix); | |
919 | total_len = prefix_len + name_len + 1; | |
936b8834 TM |
920 | *result += total_len; |
921 | ||
922 | /* we are just looking for how big our buffer needs to be */ | |
923 | if (!size) | |
924 | return 0; | |
925 | ||
926 | if (*result > size) | |
927 | return -ERANGE; | |
928 | ||
929 | memcpy(p, prefix, prefix_len); | |
930 | memcpy(p + prefix_len, name, name_len); | |
931 | p[prefix_len + name_len] = '\0'; | |
932 | ||
933 | return 0; | |
934 | } | |
935 | ||
cf1d6c76 TY |
936 | static int ocfs2_xattr_list_entries(struct inode *inode, |
937 | struct ocfs2_xattr_header *header, | |
938 | char *buffer, size_t buffer_size) | |
939 | { | |
936b8834 TM |
940 | size_t result = 0; |
941 | int i, type, ret; | |
1046cb11 | 942 | const char *name; |
cf1d6c76 TY |
943 | |
944 | for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) { | |
945 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
936b8834 | 946 | type = ocfs2_xattr_get_type(entry); |
1046cb11 AG |
947 | name = (const char *)header + |
948 | le16_to_cpu(entry->xe_name_offset); | |
936b8834 | 949 | |
1046cb11 AG |
950 | ret = ocfs2_xattr_list_entry(inode->i_sb, |
951 | buffer, buffer_size, | |
952 | &result, type, name, | |
953 | entry->xe_name_len); | |
954 | if (ret) | |
955 | return ret; | |
cf1d6c76 TY |
956 | } |
957 | ||
936b8834 | 958 | return result; |
cf1d6c76 TY |
959 | } |
960 | ||
8b2c0dba TM |
961 | int ocfs2_has_inline_xattr_value_outside(struct inode *inode, |
962 | struct ocfs2_dinode *di) | |
963 | { | |
964 | struct ocfs2_xattr_header *xh; | |
965 | int i; | |
966 | ||
967 | xh = (struct ocfs2_xattr_header *) | |
968 | ((void *)di + inode->i_sb->s_blocksize - | |
969 | le16_to_cpu(di->i_xattr_inline_size)); | |
970 | ||
971 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) | |
972 | if (!ocfs2_xattr_is_local(&xh->xh_entries[i])) | |
973 | return 1; | |
974 | ||
975 | return 0; | |
976 | } | |
977 | ||
cf1d6c76 TY |
978 | static int ocfs2_xattr_ibody_list(struct inode *inode, |
979 | struct ocfs2_dinode *di, | |
980 | char *buffer, | |
981 | size_t buffer_size) | |
982 | { | |
983 | struct ocfs2_xattr_header *header = NULL; | |
984 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
985 | int ret = 0; | |
986 | ||
987 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
988 | return ret; | |
989 | ||
990 | header = (struct ocfs2_xattr_header *) | |
991 | ((void *)di + inode->i_sb->s_blocksize - | |
992 | le16_to_cpu(di->i_xattr_inline_size)); | |
993 | ||
994 | ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size); | |
995 | ||
996 | return ret; | |
997 | } | |
998 | ||
999 | static int ocfs2_xattr_block_list(struct inode *inode, | |
1000 | struct ocfs2_dinode *di, | |
1001 | char *buffer, | |
1002 | size_t buffer_size) | |
1003 | { | |
1004 | struct buffer_head *blk_bh = NULL; | |
0c044f0b | 1005 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
1006 | int ret = 0; |
1007 | ||
1008 | if (!di->i_xattr_loc) | |
1009 | return ret; | |
1010 | ||
4ae1d69b JB |
1011 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
1012 | &blk_bh); | |
cf1d6c76 TY |
1013 | if (ret < 0) { |
1014 | mlog_errno(ret); | |
1015 | return ret; | |
1016 | } | |
cf1d6c76 | 1017 | |
0c044f0b | 1018 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
0c044f0b TM |
1019 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
1020 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
1021 | ret = ocfs2_xattr_list_entries(inode, header, | |
1022 | buffer, buffer_size); | |
47bca495 TM |
1023 | } else |
1024 | ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh, | |
0c044f0b | 1025 | buffer, buffer_size); |
4ae1d69b | 1026 | |
cf1d6c76 TY |
1027 | brelse(blk_bh); |
1028 | ||
1029 | return ret; | |
1030 | } | |
1031 | ||
1032 | ssize_t ocfs2_listxattr(struct dentry *dentry, | |
1033 | char *buffer, | |
1034 | size_t size) | |
1035 | { | |
1036 | int ret = 0, i_ret = 0, b_ret = 0; | |
1037 | struct buffer_head *di_bh = NULL; | |
1038 | struct ocfs2_dinode *di = NULL; | |
2b0143b5 | 1039 | struct ocfs2_inode_info *oi = OCFS2_I(d_inode(dentry)); |
cf1d6c76 | 1040 | |
8154da3d TY |
1041 | if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb))) |
1042 | return -EOPNOTSUPP; | |
1043 | ||
cf1d6c76 TY |
1044 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1045 | return ret; | |
1046 | ||
2b0143b5 | 1047 | ret = ocfs2_inode_lock(d_inode(dentry), &di_bh, 0); |
cf1d6c76 TY |
1048 | if (ret < 0) { |
1049 | mlog_errno(ret); | |
1050 | return ret; | |
1051 | } | |
1052 | ||
1053 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1054 | ||
1055 | down_read(&oi->ip_xattr_sem); | |
2b0143b5 | 1056 | i_ret = ocfs2_xattr_ibody_list(d_inode(dentry), di, buffer, size); |
cf1d6c76 TY |
1057 | if (i_ret < 0) |
1058 | b_ret = 0; | |
1059 | else { | |
1060 | if (buffer) { | |
1061 | buffer += i_ret; | |
1062 | size -= i_ret; | |
1063 | } | |
2b0143b5 | 1064 | b_ret = ocfs2_xattr_block_list(d_inode(dentry), di, |
cf1d6c76 TY |
1065 | buffer, size); |
1066 | if (b_ret < 0) | |
1067 | i_ret = 0; | |
1068 | } | |
1069 | up_read(&oi->ip_xattr_sem); | |
2b0143b5 | 1070 | ocfs2_inode_unlock(d_inode(dentry), 0); |
cf1d6c76 TY |
1071 | |
1072 | brelse(di_bh); | |
1073 | ||
1074 | return i_ret + b_ret; | |
1075 | } | |
1076 | ||
1077 | static int ocfs2_xattr_find_entry(int name_index, | |
1078 | const char *name, | |
1079 | struct ocfs2_xattr_search *xs) | |
1080 | { | |
1081 | struct ocfs2_xattr_entry *entry; | |
1082 | size_t name_len; | |
1083 | int i, cmp = 1; | |
1084 | ||
1085 | if (name == NULL) | |
1086 | return -EINVAL; | |
1087 | ||
1088 | name_len = strlen(name); | |
1089 | entry = xs->here; | |
1090 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
1091 | cmp = name_index - ocfs2_xattr_get_type(entry); | |
1092 | if (!cmp) | |
1093 | cmp = name_len - entry->xe_name_len; | |
1094 | if (!cmp) | |
1095 | cmp = memcmp(name, (xs->base + | |
1096 | le16_to_cpu(entry->xe_name_offset)), | |
1097 | name_len); | |
1098 | if (cmp == 0) | |
1099 | break; | |
1100 | entry += 1; | |
1101 | } | |
1102 | xs->here = entry; | |
1103 | ||
1104 | return cmp ? -ENODATA : 0; | |
1105 | } | |
1106 | ||
1107 | static int ocfs2_xattr_get_value_outside(struct inode *inode, | |
589dc260 | 1108 | struct ocfs2_xattr_value_root *xv, |
cf1d6c76 TY |
1109 | void *buffer, |
1110 | size_t len) | |
1111 | { | |
1112 | u32 cpos, p_cluster, num_clusters, bpc, clusters; | |
1113 | u64 blkno; | |
1114 | int i, ret = 0; | |
1115 | size_t cplen, blocksize; | |
1116 | struct buffer_head *bh = NULL; | |
cf1d6c76 TY |
1117 | struct ocfs2_extent_list *el; |
1118 | ||
cf1d6c76 TY |
1119 | el = &xv->xr_list; |
1120 | clusters = le32_to_cpu(xv->xr_clusters); | |
1121 | bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1122 | blocksize = inode->i_sb->s_blocksize; | |
1123 | ||
1124 | cpos = 0; | |
1125 | while (cpos < clusters) { | |
1126 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1127 | &num_clusters, el, NULL); |
cf1d6c76 TY |
1128 | if (ret) { |
1129 | mlog_errno(ret); | |
1130 | goto out; | |
1131 | } | |
1132 | ||
1133 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
1134 | /* Copy ocfs2_xattr_value */ | |
1135 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1136 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1137 | &bh, NULL); | |
cf1d6c76 TY |
1138 | if (ret) { |
1139 | mlog_errno(ret); | |
1140 | goto out; | |
1141 | } | |
1142 | ||
1143 | cplen = len >= blocksize ? blocksize : len; | |
1144 | memcpy(buffer, bh->b_data, cplen); | |
1145 | len -= cplen; | |
1146 | buffer += cplen; | |
1147 | ||
1148 | brelse(bh); | |
1149 | bh = NULL; | |
1150 | if (len == 0) | |
1151 | break; | |
1152 | } | |
1153 | cpos += num_clusters; | |
1154 | } | |
1155 | out: | |
1156 | return ret; | |
1157 | } | |
1158 | ||
1159 | static int ocfs2_xattr_ibody_get(struct inode *inode, | |
1160 | int name_index, | |
1161 | const char *name, | |
1162 | void *buffer, | |
1163 | size_t buffer_size, | |
1164 | struct ocfs2_xattr_search *xs) | |
1165 | { | |
1166 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1167 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
589dc260 | 1168 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 TY |
1169 | size_t size; |
1170 | int ret = 0; | |
1171 | ||
1172 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
1173 | return -ENODATA; | |
1174 | ||
1175 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
1176 | xs->header = (struct ocfs2_xattr_header *) | |
1177 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
1178 | xs->base = (void *)xs->header; | |
1179 | xs->here = xs->header->xh_entries; | |
1180 | ||
1181 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
1182 | if (ret) | |
1183 | return ret; | |
1184 | size = le64_to_cpu(xs->here->xe_value_size); | |
1185 | if (buffer) { | |
1186 | if (size > buffer_size) | |
1187 | return -ERANGE; | |
1188 | if (ocfs2_xattr_is_local(xs->here)) { | |
1189 | memcpy(buffer, (void *)xs->base + | |
1190 | le16_to_cpu(xs->here->xe_name_offset) + | |
1191 | OCFS2_XATTR_SIZE(xs->here->xe_name_len), size); | |
1192 | } else { | |
589dc260 TM |
1193 | xv = (struct ocfs2_xattr_value_root *) |
1194 | (xs->base + le16_to_cpu( | |
1195 | xs->here->xe_name_offset) + | |
1196 | OCFS2_XATTR_SIZE(xs->here->xe_name_len)); | |
1197 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1198 | buffer, size); |
1199 | if (ret < 0) { | |
1200 | mlog_errno(ret); | |
1201 | return ret; | |
1202 | } | |
1203 | } | |
1204 | } | |
1205 | ||
1206 | return size; | |
1207 | } | |
1208 | ||
1209 | static int ocfs2_xattr_block_get(struct inode *inode, | |
1210 | int name_index, | |
1211 | const char *name, | |
1212 | void *buffer, | |
1213 | size_t buffer_size, | |
1214 | struct ocfs2_xattr_search *xs) | |
1215 | { | |
cf1d6c76 | 1216 | struct ocfs2_xattr_block *xb; |
589dc260 | 1217 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 | 1218 | size_t size; |
44d8e4e1 SM |
1219 | int ret = -ENODATA, name_offset, name_len, i; |
1220 | int uninitialized_var(block_off); | |
cf1d6c76 | 1221 | |
ba937127 JB |
1222 | xs->bucket = ocfs2_xattr_bucket_new(inode); |
1223 | if (!xs->bucket) { | |
1224 | ret = -ENOMEM; | |
1225 | mlog_errno(ret); | |
1226 | goto cleanup; | |
1227 | } | |
589dc260 | 1228 | |
54f443f4 JB |
1229 | ret = ocfs2_xattr_block_find(inode, name_index, name, xs); |
1230 | if (ret) { | |
cf1d6c76 | 1231 | mlog_errno(ret); |
cf1d6c76 TY |
1232 | goto cleanup; |
1233 | } | |
1234 | ||
6c1e183e TY |
1235 | if (xs->not_found) { |
1236 | ret = -ENODATA; | |
1237 | goto cleanup; | |
1238 | } | |
1239 | ||
54f443f4 | 1240 | xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
1241 | size = le64_to_cpu(xs->here->xe_value_size); |
1242 | if (buffer) { | |
1243 | ret = -ERANGE; | |
1244 | if (size > buffer_size) | |
1245 | goto cleanup; | |
589dc260 TM |
1246 | |
1247 | name_offset = le16_to_cpu(xs->here->xe_name_offset); | |
1248 | name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len); | |
1249 | i = xs->here - xs->header->xh_entries; | |
1250 | ||
1251 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 1252 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
ba937127 | 1253 | bucket_xh(xs->bucket), |
589dc260 TM |
1254 | i, |
1255 | &block_off, | |
1256 | &name_offset); | |
023d4ea3 JQ |
1257 | if (ret) { |
1258 | mlog_errno(ret); | |
1259 | goto cleanup; | |
1260 | } | |
ba937127 | 1261 | xs->base = bucket_block(xs->bucket, block_off); |
589dc260 | 1262 | } |
cf1d6c76 TY |
1263 | if (ocfs2_xattr_is_local(xs->here)) { |
1264 | memcpy(buffer, (void *)xs->base + | |
589dc260 | 1265 | name_offset + name_len, size); |
cf1d6c76 | 1266 | } else { |
589dc260 TM |
1267 | xv = (struct ocfs2_xattr_value_root *) |
1268 | (xs->base + name_offset + name_len); | |
1269 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1270 | buffer, size); |
1271 | if (ret < 0) { | |
1272 | mlog_errno(ret); | |
1273 | goto cleanup; | |
1274 | } | |
1275 | } | |
1276 | } | |
1277 | ret = size; | |
1278 | cleanup: | |
ba937127 | 1279 | ocfs2_xattr_bucket_free(xs->bucket); |
cf1d6c76 | 1280 | |
54f443f4 JB |
1281 | brelse(xs->xattr_bh); |
1282 | xs->xattr_bh = NULL; | |
cf1d6c76 TY |
1283 | return ret; |
1284 | } | |
1285 | ||
4e3e9d02 TY |
1286 | int ocfs2_xattr_get_nolock(struct inode *inode, |
1287 | struct buffer_head *di_bh, | |
0030e001 TY |
1288 | int name_index, |
1289 | const char *name, | |
1290 | void *buffer, | |
1291 | size_t buffer_size) | |
cf1d6c76 TY |
1292 | { |
1293 | int ret; | |
1294 | struct ocfs2_dinode *di = NULL; | |
cf1d6c76 TY |
1295 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
1296 | struct ocfs2_xattr_search xis = { | |
1297 | .not_found = -ENODATA, | |
1298 | }; | |
1299 | struct ocfs2_xattr_search xbs = { | |
1300 | .not_found = -ENODATA, | |
1301 | }; | |
1302 | ||
8154da3d TY |
1303 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
1304 | return -EOPNOTSUPP; | |
1305 | ||
cf1d6c76 | 1306 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
2b693005 | 1307 | return -ENODATA; |
cf1d6c76 | 1308 | |
cf1d6c76 TY |
1309 | xis.inode_bh = xbs.inode_bh = di_bh; |
1310 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1311 | ||
cf1d6c76 TY |
1312 | ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer, |
1313 | buffer_size, &xis); | |
6c1e183e | 1314 | if (ret == -ENODATA && di->i_xattr_loc) |
cf1d6c76 TY |
1315 | ret = ocfs2_xattr_block_get(inode, name_index, name, buffer, |
1316 | buffer_size, &xbs); | |
4e3e9d02 TY |
1317 | |
1318 | return ret; | |
1319 | } | |
1320 | ||
1321 | /* ocfs2_xattr_get() | |
1322 | * | |
1323 | * Copy an extended attribute into the buffer provided. | |
1324 | * Buffer is NULL to compute the size of buffer required. | |
1325 | */ | |
1326 | static int ocfs2_xattr_get(struct inode *inode, | |
1327 | int name_index, | |
1328 | const char *name, | |
1329 | void *buffer, | |
1330 | size_t buffer_size) | |
1331 | { | |
8818efaa | 1332 | int ret, had_lock; |
4e3e9d02 | 1333 | struct buffer_head *di_bh = NULL; |
8818efaa | 1334 | struct ocfs2_lock_holder oh; |
4e3e9d02 | 1335 | |
8818efaa ER |
1336 | had_lock = ocfs2_inode_lock_tracker(inode, &di_bh, 0, &oh); |
1337 | if (had_lock < 0) { | |
1338 | mlog_errno(had_lock); | |
1339 | return had_lock; | |
4e3e9d02 | 1340 | } |
5e64b0d9 | 1341 | down_read(&OCFS2_I(inode)->ip_xattr_sem); |
4e3e9d02 TY |
1342 | ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index, |
1343 | name, buffer, buffer_size); | |
5e64b0d9 | 1344 | up_read(&OCFS2_I(inode)->ip_xattr_sem); |
4e3e9d02 | 1345 | |
8818efaa | 1346 | ocfs2_inode_unlock_tracker(inode, 0, &oh, had_lock); |
cf1d6c76 TY |
1347 | |
1348 | brelse(di_bh); | |
1349 | ||
1350 | return ret; | |
1351 | } | |
1352 | ||
1353 | static int __ocfs2_xattr_set_value_outside(struct inode *inode, | |
85db90e7 | 1354 | handle_t *handle, |
492a8a33 | 1355 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1356 | const void *value, |
1357 | int value_len) | |
1358 | { | |
71d548a6 | 1359 | int ret = 0, i, cp_len; |
cf1d6c76 TY |
1360 | u16 blocksize = inode->i_sb->s_blocksize; |
1361 | u32 p_cluster, num_clusters; | |
1362 | u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1363 | u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len); | |
1364 | u64 blkno; | |
1365 | struct buffer_head *bh = NULL; | |
492a8a33 TM |
1366 | unsigned int ext_flags; |
1367 | struct ocfs2_xattr_value_root *xv = vb->vb_xv; | |
cf1d6c76 TY |
1368 | |
1369 | BUG_ON(clusters > le32_to_cpu(xv->xr_clusters)); | |
1370 | ||
cf1d6c76 TY |
1371 | while (cpos < clusters) { |
1372 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1373 | &num_clusters, &xv->xr_list, |
492a8a33 | 1374 | &ext_flags); |
cf1d6c76 TY |
1375 | if (ret) { |
1376 | mlog_errno(ret); | |
85db90e7 | 1377 | goto out; |
cf1d6c76 TY |
1378 | } |
1379 | ||
492a8a33 TM |
1380 | BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); |
1381 | ||
cf1d6c76 TY |
1382 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); |
1383 | ||
1384 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1385 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1386 | &bh, NULL); | |
cf1d6c76 TY |
1387 | if (ret) { |
1388 | mlog_errno(ret); | |
85db90e7 | 1389 | goto out; |
cf1d6c76 TY |
1390 | } |
1391 | ||
1392 | ret = ocfs2_journal_access(handle, | |
0cf2f763 | 1393 | INODE_CACHE(inode), |
cf1d6c76 TY |
1394 | bh, |
1395 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1396 | if (ret < 0) { | |
1397 | mlog_errno(ret); | |
85db90e7 | 1398 | goto out; |
cf1d6c76 TY |
1399 | } |
1400 | ||
1401 | cp_len = value_len > blocksize ? blocksize : value_len; | |
1402 | memcpy(bh->b_data, value, cp_len); | |
1403 | value_len -= cp_len; | |
1404 | value += cp_len; | |
1405 | if (cp_len < blocksize) | |
1406 | memset(bh->b_data + cp_len, 0, | |
1407 | blocksize - cp_len); | |
1408 | ||
ec20cec7 | 1409 | ocfs2_journal_dirty(handle, bh); |
cf1d6c76 TY |
1410 | brelse(bh); |
1411 | bh = NULL; | |
1412 | ||
1413 | /* | |
1414 | * XXX: do we need to empty all the following | |
1415 | * blocks in this cluster? | |
1416 | */ | |
1417 | if (!value_len) | |
1418 | break; | |
1419 | } | |
1420 | cpos += num_clusters; | |
1421 | } | |
cf1d6c76 TY |
1422 | out: |
1423 | brelse(bh); | |
1424 | ||
1425 | return ret; | |
1426 | } | |
1427 | ||
69a3e539 JB |
1428 | static int ocfs2_xa_check_space_helper(int needed_space, int free_start, |
1429 | int num_entries) | |
1430 | { | |
1431 | int free_space; | |
1432 | ||
1433 | if (!needed_space) | |
1434 | return 0; | |
1435 | ||
1436 | free_space = free_start - | |
1437 | sizeof(struct ocfs2_xattr_header) - | |
1438 | (num_entries * sizeof(struct ocfs2_xattr_entry)) - | |
1439 | OCFS2_XATTR_HEADER_GAP; | |
1440 | if (free_space < 0) | |
1441 | return -EIO; | |
1442 | if (free_space < needed_space) | |
1443 | return -ENOSPC; | |
1444 | ||
1445 | return 0; | |
1446 | } | |
1447 | ||
cf2bc809 JB |
1448 | static int ocfs2_xa_journal_access(handle_t *handle, struct ocfs2_xa_loc *loc, |
1449 | int type) | |
1450 | { | |
1451 | return loc->xl_ops->xlo_journal_access(handle, loc, type); | |
1452 | } | |
1453 | ||
1454 | static void ocfs2_xa_journal_dirty(handle_t *handle, struct ocfs2_xa_loc *loc) | |
1455 | { | |
1456 | loc->xl_ops->xlo_journal_dirty(handle, loc); | |
1457 | } | |
1458 | ||
69a3e539 JB |
1459 | /* Give a pointer into the storage for the given offset */ |
1460 | static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset) | |
1461 | { | |
1462 | BUG_ON(offset >= loc->xl_size); | |
1463 | return loc->xl_ops->xlo_offset_pointer(loc, offset); | |
1464 | } | |
1465 | ||
11179f2c JB |
1466 | /* |
1467 | * Wipe the name+value pair and allow the storage to reclaim it. This | |
1468 | * must be followed by either removal of the entry or a call to | |
1469 | * ocfs2_xa_add_namevalue(). | |
1470 | */ | |
1471 | static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1472 | { | |
1473 | loc->xl_ops->xlo_wipe_namevalue(loc); | |
1474 | } | |
1475 | ||
69a3e539 JB |
1476 | /* |
1477 | * Find lowest offset to a name+value pair. This is the start of our | |
1478 | * downward-growing free space. | |
1479 | */ | |
1480 | static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc) | |
1481 | { | |
1482 | return loc->xl_ops->xlo_get_free_start(loc); | |
1483 | } | |
1484 | ||
1485 | /* Can we reuse loc->xl_entry for xi? */ | |
1486 | static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc, | |
1487 | struct ocfs2_xattr_info *xi) | |
1488 | { | |
1489 | return loc->xl_ops->xlo_can_reuse(loc, xi); | |
1490 | } | |
1491 | ||
1492 | /* How much free space is needed to set the new value */ | |
1493 | static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc, | |
1494 | struct ocfs2_xattr_info *xi) | |
1495 | { | |
1496 | return loc->xl_ops->xlo_check_space(loc, xi); | |
1497 | } | |
1498 | ||
1499 | static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) | |
1500 | { | |
1501 | loc->xl_ops->xlo_add_entry(loc, name_hash); | |
1502 | loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash); | |
1503 | /* | |
1504 | * We can't leave the new entry's xe_name_offset at zero or | |
1505 | * add_namevalue() will go nuts. We set it to the size of our | |
1506 | * storage so that it can never be less than any other entry. | |
1507 | */ | |
1508 | loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size); | |
1509 | } | |
1510 | ||
1511 | static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc, | |
1512 | struct ocfs2_xattr_info *xi) | |
1513 | { | |
1514 | int size = namevalue_size_xi(xi); | |
1515 | int nameval_offset; | |
1516 | char *nameval_buf; | |
1517 | ||
1518 | loc->xl_ops->xlo_add_namevalue(loc, size); | |
1519 | loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); | |
1520 | loc->xl_entry->xe_name_len = xi->xi_name_len; | |
1521 | ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index); | |
1522 | ocfs2_xattr_set_local(loc->xl_entry, | |
1523 | xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE); | |
1524 | ||
1525 | nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1526 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); | |
1527 | memset(nameval_buf, 0, size); | |
1528 | memcpy(nameval_buf, xi->xi_name, xi->xi_name_len); | |
1529 | } | |
1530 | ||
3fc12afa JB |
1531 | static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc, |
1532 | struct ocfs2_xattr_value_buf *vb) | |
1533 | { | |
1534 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1535 | int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); | |
1536 | ||
1537 | /* Value bufs are for value trees */ | |
73857ee0 | 1538 | BUG_ON(ocfs2_xattr_is_local(loc->xl_entry)); |
3fc12afa JB |
1539 | BUG_ON(namevalue_size_xe(loc->xl_entry) != |
1540 | (name_size + OCFS2_XATTR_ROOT_SIZE)); | |
1541 | ||
1542 | loc->xl_ops->xlo_fill_value_buf(loc, vb); | |
1543 | vb->vb_xv = | |
1544 | (struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc, | |
1545 | nameval_offset + | |
1546 | name_size); | |
1547 | } | |
1548 | ||
cf2bc809 JB |
1549 | static int ocfs2_xa_block_journal_access(handle_t *handle, |
1550 | struct ocfs2_xa_loc *loc, int type) | |
1551 | { | |
1552 | struct buffer_head *bh = loc->xl_storage; | |
1553 | ocfs2_journal_access_func access; | |
1554 | ||
1555 | if (loc->xl_size == (bh->b_size - | |
1556 | offsetof(struct ocfs2_xattr_block, | |
1557 | xb_attrs.xb_header))) | |
1558 | access = ocfs2_journal_access_xb; | |
1559 | else | |
1560 | access = ocfs2_journal_access_di; | |
1561 | return access(handle, INODE_CACHE(loc->xl_inode), bh, type); | |
1562 | } | |
1563 | ||
1564 | static void ocfs2_xa_block_journal_dirty(handle_t *handle, | |
1565 | struct ocfs2_xa_loc *loc) | |
1566 | { | |
1567 | struct buffer_head *bh = loc->xl_storage; | |
1568 | ||
1569 | ocfs2_journal_dirty(handle, bh); | |
1570 | } | |
1571 | ||
11179f2c JB |
1572 | static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc, |
1573 | int offset) | |
1574 | { | |
11179f2c JB |
1575 | return (char *)loc->xl_header + offset; |
1576 | } | |
1577 | ||
69a3e539 JB |
1578 | static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc, |
1579 | struct ocfs2_xattr_info *xi) | |
1580 | { | |
1581 | /* | |
1582 | * Block storage is strict. If the sizes aren't exact, we will | |
1583 | * remove the old one and reinsert the new. | |
1584 | */ | |
1585 | return namevalue_size_xe(loc->xl_entry) == | |
1586 | namevalue_size_xi(xi); | |
1587 | } | |
1588 | ||
1589 | static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc) | |
1590 | { | |
1591 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1592 | int i, count = le16_to_cpu(xh->xh_count); | |
1593 | int offset, free_start = loc->xl_size; | |
1594 | ||
1595 | for (i = 0; i < count; i++) { | |
1596 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1597 | if (offset < free_start) | |
1598 | free_start = offset; | |
1599 | } | |
1600 | ||
1601 | return free_start; | |
1602 | } | |
1603 | ||
1604 | static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc, | |
1605 | struct ocfs2_xattr_info *xi) | |
1606 | { | |
1607 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1608 | int free_start = ocfs2_xa_get_free_start(loc); | |
1609 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1610 | ||
1611 | /* | |
1612 | * Block storage will reclaim the original entry before inserting | |
1613 | * the new value, so we only need the difference. If the new | |
1614 | * entry is smaller than the old one, we don't need anything. | |
1615 | */ | |
1616 | if (loc->xl_entry) { | |
1617 | /* Don't need space if we're reusing! */ | |
1618 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1619 | needed_space = 0; | |
1620 | else | |
1621 | needed_space -= ocfs2_xe_entry_usage(loc->xl_entry); | |
1622 | } | |
1623 | if (needed_space < 0) | |
1624 | needed_space = 0; | |
1625 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1626 | } | |
1627 | ||
11179f2c JB |
1628 | /* |
1629 | * Block storage for xattrs keeps the name+value pairs compacted. When | |
1630 | * we remove one, we have to shift any that preceded it towards the end. | |
1631 | */ | |
1632 | static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1633 | { | |
1634 | int i, offset; | |
1635 | int namevalue_offset, first_namevalue_offset, namevalue_size; | |
1636 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1637 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
11179f2c JB |
1638 | int count = le16_to_cpu(xh->xh_count); |
1639 | ||
1640 | namevalue_offset = le16_to_cpu(entry->xe_name_offset); | |
199799a3 | 1641 | namevalue_size = namevalue_size_xe(entry); |
69a3e539 | 1642 | first_namevalue_offset = ocfs2_xa_get_free_start(loc); |
11179f2c JB |
1643 | |
1644 | /* Shift the name+value pairs */ | |
1645 | memmove((char *)xh + first_namevalue_offset + namevalue_size, | |
1646 | (char *)xh + first_namevalue_offset, | |
1647 | namevalue_offset - first_namevalue_offset); | |
1648 | memset((char *)xh + first_namevalue_offset, 0, namevalue_size); | |
1649 | ||
1650 | /* Now tell xh->xh_entries about it */ | |
1651 | for (i = 0; i < count; i++) { | |
1652 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
dfe4d3d6 | 1653 | if (offset <= namevalue_offset) |
11179f2c JB |
1654 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, |
1655 | namevalue_size); | |
1656 | } | |
1657 | ||
1658 | /* | |
1659 | * Note that we don't update xh_free_start or xh_name_value_len | |
1660 | * because they're not used in block-stored xattrs. | |
1661 | */ | |
1662 | } | |
1663 | ||
69a3e539 JB |
1664 | static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1665 | { | |
1666 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1667 | loc->xl_entry = &(loc->xl_header->xh_entries[count]); | |
1668 | le16_add_cpu(&loc->xl_header->xh_count, 1); | |
1669 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1670 | } | |
1671 | ||
1672 | static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1673 | { | |
1674 | int free_start = ocfs2_xa_get_free_start(loc); | |
1675 | ||
1676 | loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size); | |
1677 | } | |
1678 | ||
3fc12afa JB |
1679 | static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc, |
1680 | struct ocfs2_xattr_value_buf *vb) | |
1681 | { | |
1682 | struct buffer_head *bh = loc->xl_storage; | |
1683 | ||
1684 | if (loc->xl_size == (bh->b_size - | |
1685 | offsetof(struct ocfs2_xattr_block, | |
1686 | xb_attrs.xb_header))) | |
1687 | vb->vb_access = ocfs2_journal_access_xb; | |
1688 | else | |
1689 | vb->vb_access = ocfs2_journal_access_di; | |
1690 | vb->vb_bh = bh; | |
1691 | } | |
1692 | ||
11179f2c JB |
1693 | /* |
1694 | * Operations for xattrs stored in blocks. This includes inline inode | |
1695 | * storage and unindexed ocfs2_xattr_blocks. | |
1696 | */ | |
1697 | static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = { | |
cf2bc809 JB |
1698 | .xlo_journal_access = ocfs2_xa_block_journal_access, |
1699 | .xlo_journal_dirty = ocfs2_xa_block_journal_dirty, | |
11179f2c | 1700 | .xlo_offset_pointer = ocfs2_xa_block_offset_pointer, |
69a3e539 JB |
1701 | .xlo_check_space = ocfs2_xa_block_check_space, |
1702 | .xlo_can_reuse = ocfs2_xa_block_can_reuse, | |
1703 | .xlo_get_free_start = ocfs2_xa_block_get_free_start, | |
11179f2c | 1704 | .xlo_wipe_namevalue = ocfs2_xa_block_wipe_namevalue, |
69a3e539 JB |
1705 | .xlo_add_entry = ocfs2_xa_block_add_entry, |
1706 | .xlo_add_namevalue = ocfs2_xa_block_add_namevalue, | |
3fc12afa | 1707 | .xlo_fill_value_buf = ocfs2_xa_block_fill_value_buf, |
11179f2c JB |
1708 | }; |
1709 | ||
cf2bc809 JB |
1710 | static int ocfs2_xa_bucket_journal_access(handle_t *handle, |
1711 | struct ocfs2_xa_loc *loc, int type) | |
1712 | { | |
1713 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1714 | ||
1715 | return ocfs2_xattr_bucket_journal_access(handle, bucket, type); | |
1716 | } | |
1717 | ||
1718 | static void ocfs2_xa_bucket_journal_dirty(handle_t *handle, | |
1719 | struct ocfs2_xa_loc *loc) | |
1720 | { | |
1721 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1722 | ||
1723 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
1724 | } | |
1725 | ||
11179f2c JB |
1726 | static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, |
1727 | int offset) | |
1728 | { | |
1729 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1730 | int block, block_offset; | |
1731 | ||
11179f2c | 1732 | /* The header is at the front of the bucket */ |
cf2bc809 JB |
1733 | block = offset >> loc->xl_inode->i_sb->s_blocksize_bits; |
1734 | block_offset = offset % loc->xl_inode->i_sb->s_blocksize; | |
11179f2c JB |
1735 | |
1736 | return bucket_block(bucket, block) + block_offset; | |
1737 | } | |
1738 | ||
69a3e539 JB |
1739 | static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc, |
1740 | struct ocfs2_xattr_info *xi) | |
1741 | { | |
1742 | return namevalue_size_xe(loc->xl_entry) >= | |
1743 | namevalue_size_xi(xi); | |
1744 | } | |
1745 | ||
1746 | static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc) | |
1747 | { | |
1748 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1749 | return le16_to_cpu(bucket_xh(bucket)->xh_free_start); | |
1750 | } | |
1751 | ||
1752 | static int ocfs2_bucket_align_free_start(struct super_block *sb, | |
1753 | int free_start, int size) | |
1754 | { | |
1755 | /* | |
1756 | * We need to make sure that the name+value pair fits within | |
1757 | * one block. | |
1758 | */ | |
1759 | if (((free_start - size) >> sb->s_blocksize_bits) != | |
1760 | ((free_start - 1) >> sb->s_blocksize_bits)) | |
1761 | free_start -= free_start % sb->s_blocksize; | |
1762 | ||
1763 | return free_start; | |
1764 | } | |
1765 | ||
1766 | static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc, | |
1767 | struct ocfs2_xattr_info *xi) | |
1768 | { | |
1769 | int rc; | |
1770 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1771 | int free_start = ocfs2_xa_get_free_start(loc); | |
1772 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1773 | int size = namevalue_size_xi(xi); | |
cf2bc809 | 1774 | struct super_block *sb = loc->xl_inode->i_sb; |
69a3e539 JB |
1775 | |
1776 | /* | |
1777 | * Bucket storage does not reclaim name+value pairs it cannot | |
1778 | * reuse. They live as holes until the bucket fills, and then | |
1779 | * the bucket is defragmented. However, the bucket can reclaim | |
1780 | * the ocfs2_xattr_entry. | |
1781 | */ | |
1782 | if (loc->xl_entry) { | |
1783 | /* Don't need space if we're reusing! */ | |
1784 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1785 | needed_space = 0; | |
1786 | else | |
1787 | needed_space -= sizeof(struct ocfs2_xattr_entry); | |
1788 | } | |
1789 | BUG_ON(needed_space < 0); | |
1790 | ||
1791 | if (free_start < size) { | |
1792 | if (needed_space) | |
1793 | return -ENOSPC; | |
1794 | } else { | |
1795 | /* | |
1796 | * First we check if it would fit in the first place. | |
1797 | * Below, we align the free start to a block. This may | |
1798 | * slide us below the minimum gap. By checking unaligned | |
1799 | * first, we avoid that error. | |
1800 | */ | |
1801 | rc = ocfs2_xa_check_space_helper(needed_space, free_start, | |
1802 | count); | |
1803 | if (rc) | |
1804 | return rc; | |
1805 | free_start = ocfs2_bucket_align_free_start(sb, free_start, | |
1806 | size); | |
1807 | } | |
1808 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1809 | } | |
1810 | ||
11179f2c JB |
1811 | static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc) |
1812 | { | |
199799a3 JB |
1813 | le16_add_cpu(&loc->xl_header->xh_name_value_len, |
1814 | -namevalue_size_xe(loc->xl_entry)); | |
11179f2c JB |
1815 | } |
1816 | ||
69a3e539 JB |
1817 | static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1818 | { | |
1819 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1820 | int count = le16_to_cpu(xh->xh_count); | |
1821 | int low = 0, high = count - 1, tmp; | |
1822 | struct ocfs2_xattr_entry *tmp_xe; | |
1823 | ||
1824 | /* | |
1825 | * We keep buckets sorted by name_hash, so we need to find | |
1826 | * our insert place. | |
1827 | */ | |
1828 | while (low <= high && count) { | |
1829 | tmp = (low + high) / 2; | |
1830 | tmp_xe = &xh->xh_entries[tmp]; | |
1831 | ||
1832 | if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) | |
1833 | low = tmp + 1; | |
1834 | else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash)) | |
1835 | high = tmp - 1; | |
1836 | else { | |
1837 | low = tmp; | |
1838 | break; | |
1839 | } | |
1840 | } | |
1841 | ||
1842 | if (low != count) | |
1843 | memmove(&xh->xh_entries[low + 1], | |
1844 | &xh->xh_entries[low], | |
1845 | ((count - low) * sizeof(struct ocfs2_xattr_entry))); | |
1846 | ||
1847 | le16_add_cpu(&xh->xh_count, 1); | |
1848 | loc->xl_entry = &xh->xh_entries[low]; | |
1849 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1850 | } | |
1851 | ||
1852 | static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1853 | { | |
1854 | int free_start = ocfs2_xa_get_free_start(loc); | |
1855 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
cf2bc809 | 1856 | struct super_block *sb = loc->xl_inode->i_sb; |
69a3e539 JB |
1857 | int nameval_offset; |
1858 | ||
1859 | free_start = ocfs2_bucket_align_free_start(sb, free_start, size); | |
1860 | nameval_offset = free_start - size; | |
1861 | loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset); | |
1862 | xh->xh_free_start = cpu_to_le16(nameval_offset); | |
1863 | le16_add_cpu(&xh->xh_name_value_len, size); | |
1864 | ||
1865 | } | |
1866 | ||
3fc12afa JB |
1867 | static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc, |
1868 | struct ocfs2_xattr_value_buf *vb) | |
1869 | { | |
1870 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
cf2bc809 | 1871 | struct super_block *sb = loc->xl_inode->i_sb; |
3fc12afa JB |
1872 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); |
1873 | int size = namevalue_size_xe(loc->xl_entry); | |
1874 | int block_offset = nameval_offset >> sb->s_blocksize_bits; | |
1875 | ||
1876 | /* Values are not allowed to straddle block boundaries */ | |
1877 | BUG_ON(block_offset != | |
1878 | ((nameval_offset + size - 1) >> sb->s_blocksize_bits)); | |
1879 | /* We expect the bucket to be filled in */ | |
1880 | BUG_ON(!bucket->bu_bhs[block_offset]); | |
1881 | ||
1882 | vb->vb_access = ocfs2_journal_access; | |
1883 | vb->vb_bh = bucket->bu_bhs[block_offset]; | |
1884 | } | |
1885 | ||
11179f2c JB |
1886 | /* Operations for xattrs stored in buckets. */ |
1887 | static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = { | |
cf2bc809 JB |
1888 | .xlo_journal_access = ocfs2_xa_bucket_journal_access, |
1889 | .xlo_journal_dirty = ocfs2_xa_bucket_journal_dirty, | |
11179f2c | 1890 | .xlo_offset_pointer = ocfs2_xa_bucket_offset_pointer, |
69a3e539 JB |
1891 | .xlo_check_space = ocfs2_xa_bucket_check_space, |
1892 | .xlo_can_reuse = ocfs2_xa_bucket_can_reuse, | |
1893 | .xlo_get_free_start = ocfs2_xa_bucket_get_free_start, | |
11179f2c | 1894 | .xlo_wipe_namevalue = ocfs2_xa_bucket_wipe_namevalue, |
69a3e539 JB |
1895 | .xlo_add_entry = ocfs2_xa_bucket_add_entry, |
1896 | .xlo_add_namevalue = ocfs2_xa_bucket_add_namevalue, | |
3fc12afa | 1897 | .xlo_fill_value_buf = ocfs2_xa_bucket_fill_value_buf, |
11179f2c JB |
1898 | }; |
1899 | ||
399ff3a7 JB |
1900 | static unsigned int ocfs2_xa_value_clusters(struct ocfs2_xa_loc *loc) |
1901 | { | |
1902 | struct ocfs2_xattr_value_buf vb; | |
1903 | ||
1904 | if (ocfs2_xattr_is_local(loc->xl_entry)) | |
1905 | return 0; | |
1906 | ||
1907 | ocfs2_xa_fill_value_buf(loc, &vb); | |
1908 | return le32_to_cpu(vb.vb_xv->xr_clusters); | |
1909 | } | |
1910 | ||
73857ee0 JB |
1911 | static int ocfs2_xa_value_truncate(struct ocfs2_xa_loc *loc, u64 bytes, |
1912 | struct ocfs2_xattr_set_ctxt *ctxt) | |
1913 | { | |
1914 | int trunc_rc, access_rc; | |
1915 | struct ocfs2_xattr_value_buf vb; | |
1916 | ||
1917 | ocfs2_xa_fill_value_buf(loc, &vb); | |
1918 | trunc_rc = ocfs2_xattr_value_truncate(loc->xl_inode, &vb, bytes, | |
1919 | ctxt); | |
1920 | ||
1921 | /* | |
1922 | * The caller of ocfs2_xa_value_truncate() has already called | |
1923 | * ocfs2_xa_journal_access on the loc. However, The truncate code | |
1924 | * calls ocfs2_extend_trans(). This may commit the previous | |
1925 | * transaction and open a new one. If this is a bucket, truncate | |
1926 | * could leave only vb->vb_bh set up for journaling. Meanwhile, | |
1927 | * the caller is expecting to dirty the entire bucket. So we must | |
1928 | * reset the journal work. We do this even if truncate has failed, | |
1929 | * as it could have failed after committing the extend. | |
1930 | */ | |
1931 | access_rc = ocfs2_xa_journal_access(ctxt->handle, loc, | |
1932 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1933 | ||
1934 | /* Errors in truncate take precedence */ | |
1935 | return trunc_rc ? trunc_rc : access_rc; | |
1936 | } | |
1937 | ||
11179f2c JB |
1938 | static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc) |
1939 | { | |
bde1e540 JB |
1940 | int index, count; |
1941 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1942 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1943 | ||
11179f2c | 1944 | ocfs2_xa_wipe_namevalue(loc); |
bde1e540 JB |
1945 | loc->xl_entry = NULL; |
1946 | ||
1947 | le16_add_cpu(&xh->xh_count, -1); | |
1948 | count = le16_to_cpu(xh->xh_count); | |
1949 | ||
1950 | /* | |
1951 | * Only zero out the entry if there are more remaining. This is | |
1952 | * important for an empty bucket, as it keeps track of the | |
1953 | * bucket's hash value. It doesn't hurt empty block storage. | |
1954 | */ | |
1955 | if (count) { | |
1956 | index = ((char *)entry - (char *)&xh->xh_entries) / | |
1957 | sizeof(struct ocfs2_xattr_entry); | |
1958 | memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1], | |
1959 | (count - index) * sizeof(struct ocfs2_xattr_entry)); | |
1960 | memset(&xh->xh_entries[count], 0, | |
1961 | sizeof(struct ocfs2_xattr_entry)); | |
1962 | } | |
11179f2c JB |
1963 | } |
1964 | ||
399ff3a7 JB |
1965 | /* |
1966 | * If we have a problem adjusting the size of an external value during | |
1967 | * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr | |
1968 | * in an intermediate state. For example, the value may be partially | |
1969 | * truncated. | |
1970 | * | |
1971 | * If the value tree hasn't changed, the extend/truncate went nowhere. | |
1972 | * We have nothing to do. The caller can treat it as a straight error. | |
1973 | * | |
1974 | * If the value tree got partially truncated, we now have a corrupted | |
1975 | * extended attribute. We're going to wipe its entry and leak the | |
1976 | * clusters. Better to leak some storage than leave a corrupt entry. | |
1977 | * | |
1978 | * If the value tree grew, it obviously didn't grow enough for the | |
1979 | * new entry. We're not going to try and reclaim those clusters either. | |
1980 | * If there was already an external value there (orig_clusters != 0), | |
1981 | * the new clusters are attached safely and we can just leave the old | |
1982 | * value in place. If there was no external value there, we remove | |
1983 | * the entry. | |
1984 | * | |
1985 | * This way, the xattr block we store in the journal will be consistent. | |
1986 | * If the size change broke because of the journal, no changes will hit | |
1987 | * disk anyway. | |
1988 | */ | |
1989 | static void ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc, | |
1990 | const char *what, | |
1991 | unsigned int orig_clusters) | |
1992 | { | |
1993 | unsigned int new_clusters = ocfs2_xa_value_clusters(loc); | |
1994 | char *nameval_buf = ocfs2_xa_offset_pointer(loc, | |
1995 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
1996 | ||
1997 | if (new_clusters < orig_clusters) { | |
1998 | mlog(ML_ERROR, | |
1999 | "Partial truncate while %s xattr %.*s. Leaking " | |
2000 | "%u clusters and removing the entry\n", | |
2001 | what, loc->xl_entry->xe_name_len, nameval_buf, | |
2002 | orig_clusters - new_clusters); | |
2003 | ocfs2_xa_remove_entry(loc); | |
2004 | } else if (!orig_clusters) { | |
2005 | mlog(ML_ERROR, | |
2006 | "Unable to allocate an external value for xattr " | |
2007 | "%.*s safely. Leaking %u clusters and removing the " | |
2008 | "entry\n", | |
2009 | loc->xl_entry->xe_name_len, nameval_buf, | |
2010 | new_clusters - orig_clusters); | |
2011 | ocfs2_xa_remove_entry(loc); | |
2012 | } else if (new_clusters > orig_clusters) | |
2013 | mlog(ML_ERROR, | |
2014 | "Unable to grow xattr %.*s safely. %u new clusters " | |
2015 | "have been added, but the value will not be " | |
2016 | "modified\n", | |
2017 | loc->xl_entry->xe_name_len, nameval_buf, | |
2018 | new_clusters - orig_clusters); | |
2019 | } | |
2020 | ||
73857ee0 JB |
2021 | static int ocfs2_xa_remove(struct ocfs2_xa_loc *loc, |
2022 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2023 | { | |
2024 | int rc = 0; | |
399ff3a7 | 2025 | unsigned int orig_clusters; |
73857ee0 JB |
2026 | |
2027 | if (!ocfs2_xattr_is_local(loc->xl_entry)) { | |
399ff3a7 | 2028 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2029 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); |
2030 | if (rc) { | |
2031 | mlog_errno(rc); | |
399ff3a7 JB |
2032 | /* |
2033 | * Since this is remove, we can return 0 if | |
2034 | * ocfs2_xa_cleanup_value_truncate() is going to | |
2035 | * wipe the entry anyway. So we check the | |
2036 | * cluster count as well. | |
2037 | */ | |
2038 | if (orig_clusters != ocfs2_xa_value_clusters(loc)) | |
2039 | rc = 0; | |
2040 | ocfs2_xa_cleanup_value_truncate(loc, "removing", | |
2041 | orig_clusters); | |
2042 | if (rc) | |
2043 | goto out; | |
73857ee0 JB |
2044 | } |
2045 | } | |
2046 | ||
2047 | ocfs2_xa_remove_entry(loc); | |
2048 | ||
2049 | out: | |
2050 | return rc; | |
2051 | } | |
2052 | ||
2053 | static void ocfs2_xa_install_value_root(struct ocfs2_xa_loc *loc) | |
2054 | { | |
2055 | int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); | |
2056 | char *nameval_buf; | |
2057 | ||
2058 | nameval_buf = ocfs2_xa_offset_pointer(loc, | |
2059 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
2060 | memcpy(nameval_buf + name_size, &def_xv, OCFS2_XATTR_ROOT_SIZE); | |
2061 | } | |
2062 | ||
2063 | /* | |
2064 | * Take an existing entry and make it ready for the new value. This | |
2065 | * won't allocate space, but it may free space. It should be ready for | |
2066 | * ocfs2_xa_prepare_entry() to finish the work. | |
2067 | */ | |
2068 | static int ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc, | |
2069 | struct ocfs2_xattr_info *xi, | |
2070 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2071 | { | |
2072 | int rc = 0; | |
2073 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
399ff3a7 | 2074 | unsigned int orig_clusters; |
73857ee0 JB |
2075 | char *nameval_buf; |
2076 | int xe_local = ocfs2_xattr_is_local(loc->xl_entry); | |
2077 | int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE; | |
2078 | ||
2079 | BUG_ON(OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len) != | |
2080 | name_size); | |
2081 | ||
2082 | nameval_buf = ocfs2_xa_offset_pointer(loc, | |
2083 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
2084 | if (xe_local) { | |
2085 | memset(nameval_buf + name_size, 0, | |
2086 | namevalue_size_xe(loc->xl_entry) - name_size); | |
2087 | if (!xi_local) | |
2088 | ocfs2_xa_install_value_root(loc); | |
2089 | } else { | |
399ff3a7 | 2090 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2091 | if (xi_local) { |
2092 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); | |
399ff3a7 | 2093 | if (rc < 0) |
73857ee0 | 2094 | mlog_errno(rc); |
399ff3a7 JB |
2095 | else |
2096 | memset(nameval_buf + name_size, 0, | |
2097 | namevalue_size_xe(loc->xl_entry) - | |
2098 | name_size); | |
73857ee0 JB |
2099 | } else if (le64_to_cpu(loc->xl_entry->xe_value_size) > |
2100 | xi->xi_value_len) { | |
2101 | rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, | |
2102 | ctxt); | |
399ff3a7 | 2103 | if (rc < 0) |
73857ee0 | 2104 | mlog_errno(rc); |
399ff3a7 JB |
2105 | } |
2106 | ||
2107 | if (rc) { | |
2108 | ocfs2_xa_cleanup_value_truncate(loc, "reusing", | |
2109 | orig_clusters); | |
2110 | goto out; | |
73857ee0 JB |
2111 | } |
2112 | } | |
2113 | ||
2114 | loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); | |
2115 | ocfs2_xattr_set_local(loc->xl_entry, xi_local); | |
2116 | ||
2117 | out: | |
2118 | return rc; | |
2119 | } | |
2120 | ||
69a3e539 JB |
2121 | /* |
2122 | * Prepares loc->xl_entry to receive the new xattr. This includes | |
2123 | * properly setting up the name+value pair region. If loc->xl_entry | |
2124 | * already exists, it will take care of modifying it appropriately. | |
69a3e539 JB |
2125 | * |
2126 | * Note that this modifies the data. You did journal_access already, | |
2127 | * right? | |
2128 | */ | |
2129 | static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc, | |
2130 | struct ocfs2_xattr_info *xi, | |
73857ee0 JB |
2131 | u32 name_hash, |
2132 | struct ocfs2_xattr_set_ctxt *ctxt) | |
69a3e539 JB |
2133 | { |
2134 | int rc = 0; | |
399ff3a7 JB |
2135 | unsigned int orig_clusters; |
2136 | __le64 orig_value_size = 0; | |
69a3e539 | 2137 | |
69a3e539 JB |
2138 | rc = ocfs2_xa_check_space(loc, xi); |
2139 | if (rc) | |
2140 | goto out; | |
2141 | ||
2142 | if (loc->xl_entry) { | |
2143 | if (ocfs2_xa_can_reuse_entry(loc, xi)) { | |
399ff3a7 | 2144 | orig_value_size = loc->xl_entry->xe_value_size; |
73857ee0 JB |
2145 | rc = ocfs2_xa_reuse_entry(loc, xi, ctxt); |
2146 | if (rc) | |
2147 | goto out; | |
2148 | goto alloc_value; | |
69a3e539 JB |
2149 | } |
2150 | ||
73857ee0 | 2151 | if (!ocfs2_xattr_is_local(loc->xl_entry)) { |
399ff3a7 | 2152 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2153 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); |
2154 | if (rc) { | |
2155 | mlog_errno(rc); | |
399ff3a7 JB |
2156 | ocfs2_xa_cleanup_value_truncate(loc, |
2157 | "overwriting", | |
2158 | orig_clusters); | |
73857ee0 JB |
2159 | goto out; |
2160 | } | |
2161 | } | |
69a3e539 JB |
2162 | ocfs2_xa_wipe_namevalue(loc); |
2163 | } else | |
2164 | ocfs2_xa_add_entry(loc, name_hash); | |
2165 | ||
2166 | /* | |
2167 | * If we get here, we have a blank entry. Fill it. We grow our | |
2168 | * name+value pair back from the end. | |
2169 | */ | |
2170 | ocfs2_xa_add_namevalue(loc, xi); | |
73857ee0 JB |
2171 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
2172 | ocfs2_xa_install_value_root(loc); | |
2173 | ||
2174 | alloc_value: | |
2175 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { | |
399ff3a7 | 2176 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 | 2177 | rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt); |
399ff3a7 | 2178 | if (rc < 0) { |
5f5261ac | 2179 | ctxt->set_abort = 1; |
399ff3a7 JB |
2180 | ocfs2_xa_cleanup_value_truncate(loc, "growing", |
2181 | orig_clusters); | |
d5a7df06 TM |
2182 | /* |
2183 | * If we were growing an existing value, | |
2184 | * ocfs2_xa_cleanup_value_truncate() won't remove | |
2185 | * the entry. We need to restore the original value | |
2186 | * size. | |
2187 | */ | |
2188 | if (loc->xl_entry) { | |
2189 | BUG_ON(!orig_value_size); | |
2190 | loc->xl_entry->xe_value_size = orig_value_size; | |
2191 | } | |
73857ee0 | 2192 | mlog_errno(rc); |
399ff3a7 | 2193 | } |
73857ee0 | 2194 | } |
69a3e539 JB |
2195 | |
2196 | out: | |
2197 | return rc; | |
2198 | } | |
2199 | ||
2200 | /* | |
73857ee0 JB |
2201 | * Store the value portion of the name+value pair. This will skip |
2202 | * values that are stored externally. Their tree roots were set up | |
2203 | * by ocfs2_xa_prepare_entry(). | |
69a3e539 | 2204 | */ |
73857ee0 JB |
2205 | static int ocfs2_xa_store_value(struct ocfs2_xa_loc *loc, |
2206 | struct ocfs2_xattr_info *xi, | |
2207 | struct ocfs2_xattr_set_ctxt *ctxt) | |
69a3e539 | 2208 | { |
73857ee0 | 2209 | int rc = 0; |
69a3e539 JB |
2210 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); |
2211 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
69a3e539 | 2212 | char *nameval_buf; |
73857ee0 | 2213 | struct ocfs2_xattr_value_buf vb; |
69a3e539 | 2214 | |
69a3e539 | 2215 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); |
73857ee0 JB |
2216 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
2217 | ocfs2_xa_fill_value_buf(loc, &vb); | |
2218 | rc = __ocfs2_xattr_set_value_outside(loc->xl_inode, | |
2219 | ctxt->handle, &vb, | |
2220 | xi->xi_value, | |
2221 | xi->xi_value_len); | |
2222 | } else | |
2223 | memcpy(nameval_buf + name_size, xi->xi_value, xi->xi_value_len); | |
2224 | ||
73857ee0 | 2225 | return rc; |
69a3e539 JB |
2226 | } |
2227 | ||
bca5e9bd JB |
2228 | static int ocfs2_xa_set(struct ocfs2_xa_loc *loc, |
2229 | struct ocfs2_xattr_info *xi, | |
2230 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2231 | { | |
2232 | int ret; | |
2233 | u32 name_hash = ocfs2_xattr_name_hash(loc->xl_inode, xi->xi_name, | |
2234 | xi->xi_name_len); | |
2235 | ||
2236 | ret = ocfs2_xa_journal_access(ctxt->handle, loc, | |
2237 | OCFS2_JOURNAL_ACCESS_WRITE); | |
2238 | if (ret) { | |
2239 | mlog_errno(ret); | |
2240 | goto out; | |
2241 | } | |
2242 | ||
399ff3a7 JB |
2243 | /* |
2244 | * From here on out, everything is going to modify the buffer a | |
2245 | * little. Errors are going to leave the xattr header in a | |
2246 | * sane state. Thus, even with errors we dirty the sucker. | |
2247 | */ | |
2248 | ||
bca5e9bd JB |
2249 | /* Don't worry, we are never called with !xi_value and !xl_entry */ |
2250 | if (!xi->xi_value) { | |
2251 | ret = ocfs2_xa_remove(loc, ctxt); | |
399ff3a7 | 2252 | goto out_dirty; |
bca5e9bd JB |
2253 | } |
2254 | ||
2255 | ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt); | |
2256 | if (ret) { | |
2257 | if (ret != -ENOSPC) | |
2258 | mlog_errno(ret); | |
399ff3a7 | 2259 | goto out_dirty; |
bca5e9bd JB |
2260 | } |
2261 | ||
2262 | ret = ocfs2_xa_store_value(loc, xi, ctxt); | |
399ff3a7 | 2263 | if (ret) |
bca5e9bd | 2264 | mlog_errno(ret); |
bca5e9bd | 2265 | |
399ff3a7 | 2266 | out_dirty: |
bca5e9bd JB |
2267 | ocfs2_xa_journal_dirty(ctxt->handle, loc); |
2268 | ||
2269 | out: | |
2270 | return ret; | |
2271 | } | |
2272 | ||
11179f2c JB |
2273 | static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc, |
2274 | struct inode *inode, | |
2275 | struct buffer_head *bh, | |
2276 | struct ocfs2_xattr_entry *entry) | |
2277 | { | |
2278 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
2279 | ||
139fface JB |
2280 | BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL)); |
2281 | ||
cf2bc809 | 2282 | loc->xl_inode = inode; |
11179f2c JB |
2283 | loc->xl_ops = &ocfs2_xa_block_loc_ops; |
2284 | loc->xl_storage = bh; | |
2285 | loc->xl_entry = entry; | |
139fface | 2286 | loc->xl_size = le16_to_cpu(di->i_xattr_inline_size); |
11179f2c JB |
2287 | loc->xl_header = |
2288 | (struct ocfs2_xattr_header *)(bh->b_data + bh->b_size - | |
2289 | loc->xl_size); | |
2290 | } | |
2291 | ||
2292 | static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc, | |
cf2bc809 | 2293 | struct inode *inode, |
11179f2c JB |
2294 | struct buffer_head *bh, |
2295 | struct ocfs2_xattr_entry *entry) | |
2296 | { | |
2297 | struct ocfs2_xattr_block *xb = | |
2298 | (struct ocfs2_xattr_block *)bh->b_data; | |
2299 | ||
2300 | BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED); | |
2301 | ||
cf2bc809 | 2302 | loc->xl_inode = inode; |
11179f2c JB |
2303 | loc->xl_ops = &ocfs2_xa_block_loc_ops; |
2304 | loc->xl_storage = bh; | |
2305 | loc->xl_header = &(xb->xb_attrs.xb_header); | |
2306 | loc->xl_entry = entry; | |
2307 | loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block, | |
2308 | xb_attrs.xb_header); | |
2309 | } | |
2310 | ||
2311 | static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc, | |
2312 | struct ocfs2_xattr_bucket *bucket, | |
2313 | struct ocfs2_xattr_entry *entry) | |
2314 | { | |
cf2bc809 | 2315 | loc->xl_inode = bucket->bu_inode; |
11179f2c JB |
2316 | loc->xl_ops = &ocfs2_xa_bucket_loc_ops; |
2317 | loc->xl_storage = bucket; | |
2318 | loc->xl_header = bucket_xh(bucket); | |
2319 | loc->xl_entry = entry; | |
2320 | loc->xl_size = OCFS2_XATTR_BUCKET_SIZE; | |
2321 | } | |
2322 | ||
ce9c5a54 TM |
2323 | /* |
2324 | * In xattr remove, if it is stored outside and refcounted, we may have | |
2325 | * the chance to split the refcount tree. So need the allocators. | |
2326 | */ | |
2327 | static int ocfs2_lock_xattr_remove_allocators(struct inode *inode, | |
2328 | struct ocfs2_xattr_value_root *xv, | |
2329 | struct ocfs2_caching_info *ref_ci, | |
2330 | struct buffer_head *ref_root_bh, | |
2331 | struct ocfs2_alloc_context **meta_ac, | |
2332 | int *ref_credits) | |
2333 | { | |
2334 | int ret, meta_add = 0; | |
2335 | u32 p_cluster, num_clusters; | |
2336 | unsigned int ext_flags; | |
2337 | ||
2338 | *ref_credits = 0; | |
2339 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
2340 | &num_clusters, | |
2341 | &xv->xr_list, | |
2342 | &ext_flags); | |
2343 | if (ret) { | |
2344 | mlog_errno(ret); | |
2345 | goto out; | |
2346 | } | |
2347 | ||
2348 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
2349 | goto out; | |
2350 | ||
2351 | ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci, | |
2352 | ref_root_bh, xv, | |
2353 | &meta_add, ref_credits); | |
2354 | if (ret) { | |
2355 | mlog_errno(ret); | |
2356 | goto out; | |
2357 | } | |
2358 | ||
2359 | ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), | |
2360 | meta_add, meta_ac); | |
2361 | if (ret) | |
2362 | mlog_errno(ret); | |
2363 | ||
2364 | out: | |
2365 | return ret; | |
2366 | } | |
2367 | ||
cf1d6c76 | 2368 | static int ocfs2_remove_value_outside(struct inode*inode, |
4311901d | 2369 | struct ocfs2_xattr_value_buf *vb, |
ce9c5a54 TM |
2370 | struct ocfs2_xattr_header *header, |
2371 | struct ocfs2_caching_info *ref_ci, | |
2372 | struct buffer_head *ref_root_bh) | |
cf1d6c76 | 2373 | { |
ce9c5a54 | 2374 | int ret = 0, i, ref_credits; |
78f30c31 TM |
2375 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
2376 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; | |
ce9c5a54 | 2377 | void *val; |
78f30c31 TM |
2378 | |
2379 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
cf1d6c76 TY |
2380 | |
2381 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
2382 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
2383 | ||
ce9c5a54 TM |
2384 | if (ocfs2_xattr_is_local(entry)) |
2385 | continue; | |
cf1d6c76 | 2386 | |
ce9c5a54 TM |
2387 | val = (void *)header + |
2388 | le16_to_cpu(entry->xe_name_offset); | |
2389 | vb->vb_xv = (struct ocfs2_xattr_value_root *) | |
2390 | (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); | |
2391 | ||
2392 | ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv, | |
2393 | ref_ci, ref_root_bh, | |
2394 | &ctxt.meta_ac, | |
2395 | &ref_credits); | |
2396 | ||
2397 | ctxt.handle = ocfs2_start_trans(osb, ref_credits + | |
2398 | ocfs2_remove_extent_credits(osb->sb)); | |
2399 | if (IS_ERR(ctxt.handle)) { | |
2400 | ret = PTR_ERR(ctxt.handle); | |
2401 | mlog_errno(ret); | |
2402 | break; | |
2403 | } | |
2404 | ||
2405 | ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); | |
ce9c5a54 TM |
2406 | |
2407 | ocfs2_commit_trans(osb, ctxt.handle); | |
2408 | if (ctxt.meta_ac) { | |
2409 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
2410 | ctxt.meta_ac = NULL; | |
cf1d6c76 | 2411 | } |
b8a0ae57 WW |
2412 | |
2413 | if (ret < 0) { | |
2414 | mlog_errno(ret); | |
2415 | break; | |
2416 | } | |
2417 | ||
cf1d6c76 TY |
2418 | } |
2419 | ||
ce9c5a54 TM |
2420 | if (ctxt.meta_ac) |
2421 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
2422 | ocfs2_schedule_truncate_log_flush(osb, 1); |
2423 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
cf1d6c76 TY |
2424 | return ret; |
2425 | } | |
2426 | ||
2427 | static int ocfs2_xattr_ibody_remove(struct inode *inode, | |
ce9c5a54 TM |
2428 | struct buffer_head *di_bh, |
2429 | struct ocfs2_caching_info *ref_ci, | |
2430 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2431 | { |
2432 | ||
2433 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2434 | struct ocfs2_xattr_header *header; | |
2435 | int ret; | |
4311901d JB |
2436 | struct ocfs2_xattr_value_buf vb = { |
2437 | .vb_bh = di_bh, | |
2438 | .vb_access = ocfs2_journal_access_di, | |
2439 | }; | |
cf1d6c76 TY |
2440 | |
2441 | header = (struct ocfs2_xattr_header *) | |
2442 | ((void *)di + inode->i_sb->s_blocksize - | |
2443 | le16_to_cpu(di->i_xattr_inline_size)); | |
2444 | ||
ce9c5a54 TM |
2445 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2446 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2447 | |
2448 | return ret; | |
2449 | } | |
2450 | ||
ce9c5a54 TM |
2451 | struct ocfs2_rm_xattr_bucket_para { |
2452 | struct ocfs2_caching_info *ref_ci; | |
2453 | struct buffer_head *ref_root_bh; | |
2454 | }; | |
2455 | ||
cf1d6c76 | 2456 | static int ocfs2_xattr_block_remove(struct inode *inode, |
ce9c5a54 TM |
2457 | struct buffer_head *blk_bh, |
2458 | struct ocfs2_caching_info *ref_ci, | |
2459 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2460 | { |
2461 | struct ocfs2_xattr_block *xb; | |
cf1d6c76 | 2462 | int ret = 0; |
4311901d JB |
2463 | struct ocfs2_xattr_value_buf vb = { |
2464 | .vb_bh = blk_bh, | |
2465 | .vb_access = ocfs2_journal_access_xb, | |
2466 | }; | |
ce9c5a54 TM |
2467 | struct ocfs2_rm_xattr_bucket_para args = { |
2468 | .ref_ci = ref_ci, | |
2469 | .ref_root_bh = ref_root_bh, | |
2470 | }; | |
cf1d6c76 TY |
2471 | |
2472 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; | |
a3944256 TM |
2473 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
2474 | struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header); | |
ce9c5a54 TM |
2475 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2476 | ref_ci, ref_root_bh); | |
a3944256 | 2477 | } else |
47bca495 TM |
2478 | ret = ocfs2_iterate_xattr_index_block(inode, |
2479 | blk_bh, | |
2480 | ocfs2_rm_xattr_cluster, | |
ce9c5a54 | 2481 | &args); |
cf1d6c76 TY |
2482 | |
2483 | return ret; | |
2484 | } | |
2485 | ||
08413899 | 2486 | static int ocfs2_xattr_free_block(struct inode *inode, |
ce9c5a54 TM |
2487 | u64 block, |
2488 | struct ocfs2_caching_info *ref_ci, | |
2489 | struct buffer_head *ref_root_bh) | |
08413899 TM |
2490 | { |
2491 | struct inode *xb_alloc_inode; | |
2492 | struct buffer_head *xb_alloc_bh = NULL; | |
2493 | struct buffer_head *blk_bh = NULL; | |
2494 | struct ocfs2_xattr_block *xb; | |
2495 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2496 | handle_t *handle; | |
2497 | int ret = 0; | |
2498 | u64 blk, bg_blkno; | |
2499 | u16 bit; | |
2500 | ||
4ae1d69b | 2501 | ret = ocfs2_read_xattr_block(inode, block, &blk_bh); |
08413899 TM |
2502 | if (ret < 0) { |
2503 | mlog_errno(ret); | |
2504 | goto out; | |
2505 | } | |
2506 | ||
ce9c5a54 | 2507 | ret = ocfs2_xattr_block_remove(inode, blk_bh, ref_ci, ref_root_bh); |
08413899 TM |
2508 | if (ret < 0) { |
2509 | mlog_errno(ret); | |
2510 | goto out; | |
2511 | } | |
2512 | ||
4ae1d69b | 2513 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
08413899 TM |
2514 | blk = le64_to_cpu(xb->xb_blkno); |
2515 | bit = le16_to_cpu(xb->xb_suballoc_bit); | |
74380c47 TM |
2516 | if (xb->xb_suballoc_loc) |
2517 | bg_blkno = le64_to_cpu(xb->xb_suballoc_loc); | |
2518 | else | |
2519 | bg_blkno = ocfs2_which_suballoc_group(blk, bit); | |
08413899 TM |
2520 | |
2521 | xb_alloc_inode = ocfs2_get_system_file_inode(osb, | |
2522 | EXTENT_ALLOC_SYSTEM_INODE, | |
2523 | le16_to_cpu(xb->xb_suballoc_slot)); | |
2524 | if (!xb_alloc_inode) { | |
2525 | ret = -ENOMEM; | |
2526 | mlog_errno(ret); | |
2527 | goto out; | |
2528 | } | |
5955102c | 2529 | inode_lock(xb_alloc_inode); |
08413899 TM |
2530 | |
2531 | ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1); | |
2532 | if (ret < 0) { | |
2533 | mlog_errno(ret); | |
2534 | goto out_mutex; | |
2535 | } | |
2536 | ||
2537 | handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE); | |
2538 | if (IS_ERR(handle)) { | |
2539 | ret = PTR_ERR(handle); | |
2540 | mlog_errno(ret); | |
2541 | goto out_unlock; | |
2542 | } | |
2543 | ||
2544 | ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh, | |
2545 | bit, bg_blkno, 1); | |
2546 | if (ret < 0) | |
2547 | mlog_errno(ret); | |
2548 | ||
2549 | ocfs2_commit_trans(osb, handle); | |
2550 | out_unlock: | |
2551 | ocfs2_inode_unlock(xb_alloc_inode, 1); | |
2552 | brelse(xb_alloc_bh); | |
2553 | out_mutex: | |
5955102c | 2554 | inode_unlock(xb_alloc_inode); |
08413899 TM |
2555 | iput(xb_alloc_inode); |
2556 | out: | |
2557 | brelse(blk_bh); | |
2558 | return ret; | |
2559 | } | |
2560 | ||
cf1d6c76 TY |
2561 | /* |
2562 | * ocfs2_xattr_remove() | |
2563 | * | |
2564 | * Free extended attribute resources associated with this inode. | |
2565 | */ | |
2566 | int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) | |
2567 | { | |
cf1d6c76 TY |
2568 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
2569 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
ce9c5a54 TM |
2570 | struct ocfs2_refcount_tree *ref_tree = NULL; |
2571 | struct buffer_head *ref_root_bh = NULL; | |
2572 | struct ocfs2_caching_info *ref_ci = NULL; | |
cf1d6c76 TY |
2573 | handle_t *handle; |
2574 | int ret; | |
2575 | ||
8154da3d TY |
2576 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
2577 | return 0; | |
2578 | ||
cf1d6c76 TY |
2579 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
2580 | return 0; | |
2581 | ||
84e40080 | 2582 | if (ocfs2_is_refcount_inode(inode)) { |
ce9c5a54 TM |
2583 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(inode->i_sb), |
2584 | le64_to_cpu(di->i_refcount_loc), | |
2585 | 1, &ref_tree, &ref_root_bh); | |
2586 | if (ret) { | |
2587 | mlog_errno(ret); | |
2588 | goto out; | |
2589 | } | |
2590 | ref_ci = &ref_tree->rf_ci; | |
2591 | ||
2592 | } | |
2593 | ||
cf1d6c76 | 2594 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { |
ce9c5a54 TM |
2595 | ret = ocfs2_xattr_ibody_remove(inode, di_bh, |
2596 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2597 | if (ret < 0) { |
2598 | mlog_errno(ret); | |
2599 | goto out; | |
2600 | } | |
2601 | } | |
cf1d6c76 | 2602 | |
08413899 TM |
2603 | if (di->i_xattr_loc) { |
2604 | ret = ocfs2_xattr_free_block(inode, | |
ce9c5a54 TM |
2605 | le64_to_cpu(di->i_xattr_loc), |
2606 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2607 | if (ret < 0) { |
2608 | mlog_errno(ret); | |
2609 | goto out; | |
2610 | } | |
2611 | } | |
2612 | ||
2613 | handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)), | |
2614 | OCFS2_INODE_UPDATE_CREDITS); | |
2615 | if (IS_ERR(handle)) { | |
2616 | ret = PTR_ERR(handle); | |
2617 | mlog_errno(ret); | |
2618 | goto out; | |
2619 | } | |
0cf2f763 | 2620 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, |
84008972 | 2621 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2622 | if (ret) { |
2623 | mlog_errno(ret); | |
2624 | goto out_commit; | |
2625 | } | |
2626 | ||
08413899 | 2627 | di->i_xattr_loc = 0; |
cf1d6c76 TY |
2628 | |
2629 | spin_lock(&oi->ip_lock); | |
2630 | oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL); | |
2631 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2632 | spin_unlock(&oi->ip_lock); | |
6fdb702d | 2633 | ocfs2_update_inode_fsync_trans(handle, inode, 0); |
cf1d6c76 | 2634 | |
ec20cec7 | 2635 | ocfs2_journal_dirty(handle, di_bh); |
cf1d6c76 TY |
2636 | out_commit: |
2637 | ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); | |
2638 | out: | |
ce9c5a54 TM |
2639 | if (ref_tree) |
2640 | ocfs2_unlock_refcount_tree(OCFS2_SB(inode->i_sb), ref_tree, 1); | |
2641 | brelse(ref_root_bh); | |
cf1d6c76 TY |
2642 | return ret; |
2643 | } | |
2644 | ||
2645 | static int ocfs2_xattr_has_space_inline(struct inode *inode, | |
2646 | struct ocfs2_dinode *di) | |
2647 | { | |
2648 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2649 | unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
2650 | int free; | |
2651 | ||
2652 | if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE) | |
2653 | return 0; | |
2654 | ||
2655 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2656 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2657 | free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size); | |
2658 | } else if (ocfs2_inode_is_fast_symlink(inode)) { | |
2659 | free = ocfs2_fast_symlink_chars(inode->i_sb) - | |
2660 | le64_to_cpu(di->i_size); | |
2661 | } else { | |
2662 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2663 | free = (le16_to_cpu(el->l_count) - | |
2664 | le16_to_cpu(el->l_next_free_rec)) * | |
2665 | sizeof(struct ocfs2_extent_rec); | |
2666 | } | |
2667 | if (free >= xattrsize) | |
2668 | return 1; | |
2669 | ||
2670 | return 0; | |
2671 | } | |
2672 | ||
2673 | /* | |
2674 | * ocfs2_xattr_ibody_find() | |
2675 | * | |
2676 | * Find extended attribute in inode block and | |
2677 | * fill search info into struct ocfs2_xattr_search. | |
2678 | */ | |
2679 | static int ocfs2_xattr_ibody_find(struct inode *inode, | |
2680 | int name_index, | |
2681 | const char *name, | |
2682 | struct ocfs2_xattr_search *xs) | |
2683 | { | |
2684 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2685 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2686 | int ret; | |
2687 | int has_space = 0; | |
2688 | ||
2689 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2690 | return 0; | |
2691 | ||
2692 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2693 | down_read(&oi->ip_alloc_sem); | |
2694 | has_space = ocfs2_xattr_has_space_inline(inode, di); | |
2695 | up_read(&oi->ip_alloc_sem); | |
2696 | if (!has_space) | |
2697 | return 0; | |
2698 | } | |
2699 | ||
2700 | xs->xattr_bh = xs->inode_bh; | |
2701 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
2702 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
2703 | xs->header = (struct ocfs2_xattr_header *) | |
2704 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
2705 | else | |
2706 | xs->header = (struct ocfs2_xattr_header *) | |
2707 | (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size); | |
2708 | xs->base = (void *)xs->header; | |
2709 | xs->here = xs->header->xh_entries; | |
2710 | ||
2711 | /* Find the named attribute. */ | |
2712 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
2713 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2714 | if (ret && ret != -ENODATA) | |
2715 | return ret; | |
2716 | xs->not_found = ret; | |
2717 | } | |
2718 | ||
2719 | return 0; | |
2720 | } | |
2721 | ||
139fface JB |
2722 | static int ocfs2_xattr_ibody_init(struct inode *inode, |
2723 | struct buffer_head *di_bh, | |
2724 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2725 | { | |
2726 | int ret; | |
2727 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2728 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2729 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2730 | unsigned int xattrsize = osb->s_xattr_inline_size; | |
2731 | ||
2732 | if (!ocfs2_xattr_has_space_inline(inode, di)) { | |
2733 | ret = -ENOSPC; | |
2734 | goto out; | |
2735 | } | |
2736 | ||
2737 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), di_bh, | |
2738 | OCFS2_JOURNAL_ACCESS_WRITE); | |
2739 | if (ret) { | |
2740 | mlog_errno(ret); | |
2741 | goto out; | |
2742 | } | |
2743 | ||
2744 | /* | |
2745 | * Adjust extent record count or inline data size | |
2746 | * to reserve space for extended attribute. | |
2747 | */ | |
2748 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2749 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2750 | le16_add_cpu(&idata->id_count, -xattrsize); | |
2751 | } else if (!(ocfs2_inode_is_fast_symlink(inode))) { | |
2752 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2753 | le16_add_cpu(&el->l_count, -(xattrsize / | |
2754 | sizeof(struct ocfs2_extent_rec))); | |
2755 | } | |
2756 | di->i_xattr_inline_size = cpu_to_le16(xattrsize); | |
2757 | ||
2758 | spin_lock(&oi->ip_lock); | |
2759 | oi->ip_dyn_features |= OCFS2_INLINE_XATTR_FL|OCFS2_HAS_XATTR_FL; | |
2760 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2761 | spin_unlock(&oi->ip_lock); | |
2762 | ||
ec20cec7 | 2763 | ocfs2_journal_dirty(ctxt->handle, di_bh); |
139fface JB |
2764 | |
2765 | out: | |
2766 | return ret; | |
2767 | } | |
2768 | ||
cf1d6c76 TY |
2769 | /* |
2770 | * ocfs2_xattr_ibody_set() | |
2771 | * | |
2772 | * Set, replace or remove an extended attribute into inode block. | |
2773 | * | |
2774 | */ | |
2775 | static int ocfs2_xattr_ibody_set(struct inode *inode, | |
2776 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2777 | struct ocfs2_xattr_search *xs, |
2778 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 | 2779 | { |
139fface | 2780 | int ret; |
cf1d6c76 | 2781 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
139fface | 2782 | struct ocfs2_xa_loc loc; |
cf1d6c76 TY |
2783 | |
2784 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2785 | return -ENOSPC; | |
2786 | ||
2787 | down_write(&oi->ip_alloc_sem); | |
139fface JB |
2788 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { |
2789 | ret = ocfs2_xattr_ibody_init(inode, xs->inode_bh, ctxt); | |
2790 | if (ret) { | |
2791 | if (ret != -ENOSPC) | |
2792 | mlog_errno(ret); | |
2793 | goto out; | |
2794 | } | |
2795 | } | |
2796 | ||
2797 | ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh, | |
2798 | xs->not_found ? NULL : xs->here); | |
2799 | ret = ocfs2_xa_set(&loc, xi, ctxt); | |
2800 | if (ret) { | |
2801 | if (ret != -ENOSPC) | |
2802 | mlog_errno(ret); | |
2803 | goto out; | |
2804 | } | |
2805 | xs->here = loc.xl_entry; | |
2806 | ||
cf1d6c76 TY |
2807 | out: |
2808 | up_write(&oi->ip_alloc_sem); | |
2809 | ||
2810 | return ret; | |
2811 | } | |
2812 | ||
2813 | /* | |
2814 | * ocfs2_xattr_block_find() | |
2815 | * | |
2816 | * Find extended attribute in external block and | |
2817 | * fill search info into struct ocfs2_xattr_search. | |
2818 | */ | |
2819 | static int ocfs2_xattr_block_find(struct inode *inode, | |
2820 | int name_index, | |
2821 | const char *name, | |
2822 | struct ocfs2_xattr_search *xs) | |
2823 | { | |
2824 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2825 | struct buffer_head *blk_bh = NULL; | |
589dc260 | 2826 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
2827 | int ret = 0; |
2828 | ||
2829 | if (!di->i_xattr_loc) | |
2830 | return ret; | |
2831 | ||
4ae1d69b JB |
2832 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
2833 | &blk_bh); | |
cf1d6c76 TY |
2834 | if (ret < 0) { |
2835 | mlog_errno(ret); | |
2836 | return ret; | |
2837 | } | |
f6087fb7 | 2838 | |
cf1d6c76 | 2839 | xs->xattr_bh = blk_bh; |
4ae1d69b | 2840 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
589dc260 TM |
2841 | |
2842 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2843 | xs->header = &xb->xb_attrs.xb_header; | |
2844 | xs->base = (void *)xs->header; | |
2845 | xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size; | |
2846 | xs->here = xs->header->xh_entries; | |
2847 | ||
2848 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2849 | } else | |
2850 | ret = ocfs2_xattr_index_block_find(inode, blk_bh, | |
2851 | name_index, | |
2852 | name, xs); | |
cf1d6c76 | 2853 | |
cf1d6c76 TY |
2854 | if (ret && ret != -ENODATA) { |
2855 | xs->xattr_bh = NULL; | |
2856 | goto cleanup; | |
2857 | } | |
2858 | xs->not_found = ret; | |
2859 | return 0; | |
cf1d6c76 TY |
2860 | cleanup: |
2861 | brelse(blk_bh); | |
2862 | ||
2863 | return ret; | |
2864 | } | |
2865 | ||
d3981544 | 2866 | static int ocfs2_create_xattr_block(struct inode *inode, |
5aea1f0e | 2867 | struct buffer_head *inode_bh, |
d3981544 JB |
2868 | struct ocfs2_xattr_set_ctxt *ctxt, |
2869 | int indexed, | |
2870 | struct buffer_head **ret_bh) | |
5aea1f0e TM |
2871 | { |
2872 | int ret; | |
2873 | u16 suballoc_bit_start; | |
2874 | u32 num_got; | |
2b6cb576 | 2875 | u64 suballoc_loc, first_blkno; |
5aea1f0e | 2876 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data; |
5aea1f0e TM |
2877 | struct buffer_head *new_bh = NULL; |
2878 | struct ocfs2_xattr_block *xblk; | |
2879 | ||
d3981544 JB |
2880 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
2881 | inode_bh, OCFS2_JOURNAL_ACCESS_CREATE); | |
5aea1f0e TM |
2882 | if (ret < 0) { |
2883 | mlog_errno(ret); | |
2884 | goto end; | |
2885 | } | |
2886 | ||
1ed9b777 | 2887 | ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1, |
2b6cb576 JB |
2888 | &suballoc_loc, &suballoc_bit_start, |
2889 | &num_got, &first_blkno); | |
5aea1f0e TM |
2890 | if (ret < 0) { |
2891 | mlog_errno(ret); | |
2892 | goto end; | |
2893 | } | |
2894 | ||
2895 | new_bh = sb_getblk(inode->i_sb, first_blkno); | |
58796207 RX |
2896 | if (!new_bh) { |
2897 | ret = -ENOMEM; | |
2898 | mlog_errno(ret); | |
2899 | goto end; | |
2900 | } | |
2901 | ||
5aea1f0e TM |
2902 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); |
2903 | ||
d3981544 | 2904 | ret = ocfs2_journal_access_xb(ctxt->handle, INODE_CACHE(inode), |
5aea1f0e TM |
2905 | new_bh, |
2906 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2907 | if (ret < 0) { | |
2908 | mlog_errno(ret); | |
2909 | goto end; | |
2910 | } | |
2911 | ||
2912 | /* Initialize ocfs2_xattr_block */ | |
2913 | xblk = (struct ocfs2_xattr_block *)new_bh->b_data; | |
2914 | memset(xblk, 0, inode->i_sb->s_blocksize); | |
2915 | strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); | |
d3981544 | 2916 | xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot); |
2b6cb576 | 2917 | xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc); |
5aea1f0e | 2918 | xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); |
1ed9b777 JB |
2919 | xblk->xb_fs_generation = |
2920 | cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation); | |
5aea1f0e | 2921 | xblk->xb_blkno = cpu_to_le64(first_blkno); |
a7fe7a3a TM |
2922 | if (indexed) { |
2923 | struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root; | |
2924 | xr->xt_clusters = cpu_to_le32(1); | |
2925 | xr->xt_last_eb_blk = 0; | |
2926 | xr->xt_list.l_tree_depth = 0; | |
2927 | xr->xt_list.l_count = cpu_to_le16( | |
2928 | ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
2929 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
2930 | xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED); | |
2931 | } | |
d3981544 | 2932 | ocfs2_journal_dirty(ctxt->handle, new_bh); |
a7fe7a3a | 2933 | |
d3981544 | 2934 | /* Add it to the inode */ |
5aea1f0e | 2935 | di->i_xattr_loc = cpu_to_le64(first_blkno); |
d3981544 JB |
2936 | |
2937 | spin_lock(&OCFS2_I(inode)->ip_lock); | |
2938 | OCFS2_I(inode)->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
2939 | di->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features); | |
2940 | spin_unlock(&OCFS2_I(inode)->ip_lock); | |
2941 | ||
2942 | ocfs2_journal_dirty(ctxt->handle, inode_bh); | |
5aea1f0e TM |
2943 | |
2944 | *ret_bh = new_bh; | |
2945 | new_bh = NULL; | |
2946 | ||
2947 | end: | |
2948 | brelse(new_bh); | |
2949 | return ret; | |
2950 | } | |
2951 | ||
cf1d6c76 TY |
2952 | /* |
2953 | * ocfs2_xattr_block_set() | |
2954 | * | |
2955 | * Set, replace or remove an extended attribute into external block. | |
2956 | * | |
2957 | */ | |
2958 | static int ocfs2_xattr_block_set(struct inode *inode, | |
2959 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2960 | struct ocfs2_xattr_search *xs, |
2961 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2962 | { |
2963 | struct buffer_head *new_bh = NULL; | |
cf1d6c76 | 2964 | struct ocfs2_xattr_block *xblk = NULL; |
cf1d6c76 | 2965 | int ret; |
d3981544 | 2966 | struct ocfs2_xa_loc loc; |
cf1d6c76 TY |
2967 | |
2968 | if (!xs->xattr_bh) { | |
d3981544 JB |
2969 | ret = ocfs2_create_xattr_block(inode, xs->inode_bh, ctxt, |
2970 | 0, &new_bh); | |
5aea1f0e | 2971 | if (ret) { |
cf1d6c76 | 2972 | mlog_errno(ret); |
85db90e7 | 2973 | goto end; |
cf1d6c76 TY |
2974 | } |
2975 | ||
cf1d6c76 | 2976 | xs->xattr_bh = new_bh; |
5aea1f0e | 2977 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
2978 | xs->header = &xblk->xb_attrs.xb_header; |
2979 | xs->base = (void *)xs->header; | |
2980 | xs->end = (void *)xblk + inode->i_sb->s_blocksize; | |
2981 | xs->here = xs->header->xh_entries; | |
01225596 TM |
2982 | } else |
2983 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; | |
2984 | ||
2985 | if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
d3981544 JB |
2986 | ocfs2_init_xattr_block_xa_loc(&loc, inode, xs->xattr_bh, |
2987 | xs->not_found ? NULL : xs->here); | |
01225596 | 2988 | |
d3981544 JB |
2989 | ret = ocfs2_xa_set(&loc, xi, ctxt); |
2990 | if (!ret) | |
2991 | xs->here = loc.xl_entry; | |
5f5261ac | 2992 | else if ((ret != -ENOSPC) || ctxt->set_abort) |
01225596 | 2993 | goto end; |
d3981544 JB |
2994 | else { |
2995 | ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); | |
2996 | if (ret) | |
2997 | goto end; | |
2998 | } | |
cf1d6c76 TY |
2999 | } |
3000 | ||
d3981544 JB |
3001 | if (le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED) |
3002 | ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); | |
01225596 TM |
3003 | |
3004 | end: | |
cf1d6c76 TY |
3005 | return ret; |
3006 | } | |
3007 | ||
78f30c31 TM |
3008 | /* Check whether the new xattr can be inserted into the inode. */ |
3009 | static int ocfs2_xattr_can_be_in_inode(struct inode *inode, | |
3010 | struct ocfs2_xattr_info *xi, | |
3011 | struct ocfs2_xattr_search *xs) | |
3012 | { | |
78f30c31 TM |
3013 | struct ocfs2_xattr_entry *last; |
3014 | int free, i; | |
3015 | size_t min_offs = xs->end - xs->base; | |
3016 | ||
3017 | if (!xs->header) | |
3018 | return 0; | |
3019 | ||
3020 | last = xs->header->xh_entries; | |
3021 | ||
3022 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
3023 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
3024 | if (offs < min_offs) | |
3025 | min_offs = offs; | |
3026 | last += 1; | |
3027 | } | |
3028 | ||
4442f518 | 3029 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
78f30c31 TM |
3030 | if (free < 0) |
3031 | return 0; | |
3032 | ||
3033 | BUG_ON(!xs->not_found); | |
3034 | ||
199799a3 | 3035 | if (free >= (sizeof(struct ocfs2_xattr_entry) + namevalue_size_xi(xi))) |
78f30c31 TM |
3036 | return 1; |
3037 | ||
3038 | return 0; | |
3039 | } | |
3040 | ||
3041 | static int ocfs2_calc_xattr_set_need(struct inode *inode, | |
3042 | struct ocfs2_dinode *di, | |
3043 | struct ocfs2_xattr_info *xi, | |
3044 | struct ocfs2_xattr_search *xis, | |
3045 | struct ocfs2_xattr_search *xbs, | |
3046 | int *clusters_need, | |
85db90e7 TM |
3047 | int *meta_need, |
3048 | int *credits_need) | |
78f30c31 TM |
3049 | { |
3050 | int ret = 0, old_in_xb = 0; | |
85db90e7 | 3051 | int clusters_add = 0, meta_add = 0, credits = 0; |
78f30c31 TM |
3052 | struct buffer_head *bh = NULL; |
3053 | struct ocfs2_xattr_block *xb = NULL; | |
3054 | struct ocfs2_xattr_entry *xe = NULL; | |
3055 | struct ocfs2_xattr_value_root *xv = NULL; | |
3056 | char *base = NULL; | |
3057 | int name_offset, name_len = 0; | |
3058 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
6b240ff6 | 3059 | xi->xi_value_len); |
78f30c31 TM |
3060 | u64 value_size; |
3061 | ||
71d548a6 TM |
3062 | /* |
3063 | * Calculate the clusters we need to write. | |
3064 | * No matter whether we replace an old one or add a new one, | |
3065 | * we need this for writing. | |
3066 | */ | |
6b240ff6 | 3067 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
71d548a6 TM |
3068 | credits += new_clusters * |
3069 | ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
3070 | ||
78f30c31 | 3071 | if (xis->not_found && xbs->not_found) { |
85db90e7 TM |
3072 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3073 | ||
6b240ff6 | 3074 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 | 3075 | clusters_add += new_clusters; |
85db90e7 | 3076 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3077 | &def_xv.xv.xr_list); |
85db90e7 | 3078 | } |
78f30c31 TM |
3079 | |
3080 | goto meta_guess; | |
3081 | } | |
3082 | ||
3083 | if (!xis->not_found) { | |
3084 | xe = xis->here; | |
3085 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3086 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3087 | base = xis->base; | |
85db90e7 | 3088 | credits += OCFS2_INODE_UPDATE_CREDITS; |
78f30c31 | 3089 | } else { |
970e4936 | 3090 | int i, block_off = 0; |
78f30c31 TM |
3091 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; |
3092 | xe = xbs->here; | |
3093 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3094 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3095 | i = xbs->here - xbs->header->xh_entries; | |
3096 | old_in_xb = 1; | |
3097 | ||
3098 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 3099 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
78f30c31 TM |
3100 | bucket_xh(xbs->bucket), |
3101 | i, &block_off, | |
3102 | &name_offset); | |
3103 | base = bucket_block(xbs->bucket, block_off); | |
85db90e7 TM |
3104 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3105 | } else { | |
78f30c31 | 3106 | base = xbs->base; |
85db90e7 TM |
3107 | credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS; |
3108 | } | |
3109 | } | |
3110 | ||
3111 | /* | |
3112 | * delete a xattr doesn't need metadata and cluster allocation. | |
3113 | * so just calculate the credits and return. | |
3114 | * | |
3115 | * The credits for removing the value tree will be extended | |
3116 | * by ocfs2_remove_extent itself. | |
3117 | */ | |
6b240ff6 | 3118 | if (!xi->xi_value) { |
85db90e7 | 3119 | if (!ocfs2_xattr_is_local(xe)) |
a90714c1 | 3120 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
85db90e7 TM |
3121 | |
3122 | goto out; | |
78f30c31 TM |
3123 | } |
3124 | ||
3125 | /* do cluster allocation guess first. */ | |
3126 | value_size = le64_to_cpu(xe->xe_value_size); | |
3127 | ||
3128 | if (old_in_xb) { | |
3129 | /* | |
3130 | * In xattr set, we always try to set the xe in inode first, | |
3131 | * so if it can be inserted into inode successfully, the old | |
3132 | * one will be removed from the xattr block, and this xattr | |
3133 | * will be inserted into inode as a new xattr in inode. | |
3134 | */ | |
3135 | if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) { | |
3136 | clusters_add += new_clusters; | |
a90714c1 | 3137 | credits += ocfs2_remove_extent_credits(inode->i_sb) + |
85db90e7 TM |
3138 | OCFS2_INODE_UPDATE_CREDITS; |
3139 | if (!ocfs2_xattr_is_local(xe)) | |
3140 | credits += ocfs2_calc_extend_credits( | |
3141 | inode->i_sb, | |
06f9da6e | 3142 | &def_xv.xv.xr_list); |
78f30c31 TM |
3143 | goto out; |
3144 | } | |
3145 | } | |
3146 | ||
6b240ff6 | 3147 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 TM |
3148 | /* the new values will be stored outside. */ |
3149 | u32 old_clusters = 0; | |
3150 | ||
3151 | if (!ocfs2_xattr_is_local(xe)) { | |
3152 | old_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
3153 | value_size); | |
3154 | xv = (struct ocfs2_xattr_value_root *) | |
3155 | (base + name_offset + name_len); | |
97aff52a | 3156 | value_size = OCFS2_XATTR_ROOT_SIZE; |
78f30c31 TM |
3157 | } else |
3158 | xv = &def_xv.xv; | |
3159 | ||
85db90e7 | 3160 | if (old_clusters >= new_clusters) { |
a90714c1 | 3161 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
78f30c31 | 3162 | goto out; |
85db90e7 | 3163 | } else { |
78f30c31 TM |
3164 | meta_add += ocfs2_extend_meta_needed(&xv->xr_list); |
3165 | clusters_add += new_clusters - old_clusters; | |
85db90e7 | 3166 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3167 | &xv->xr_list); |
97aff52a TM |
3168 | if (value_size >= OCFS2_XATTR_ROOT_SIZE) |
3169 | goto out; | |
78f30c31 TM |
3170 | } |
3171 | } else { | |
3172 | /* | |
3173 | * Now the new value will be stored inside. So if the new | |
3174 | * value is smaller than the size of value root or the old | |
3175 | * value, we don't need any allocation, otherwise we have | |
3176 | * to guess metadata allocation. | |
3177 | */ | |
6b240ff6 JB |
3178 | if ((ocfs2_xattr_is_local(xe) && |
3179 | (value_size >= xi->xi_value_len)) || | |
78f30c31 | 3180 | (!ocfs2_xattr_is_local(xe) && |
6b240ff6 | 3181 | OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len)) |
78f30c31 TM |
3182 | goto out; |
3183 | } | |
3184 | ||
3185 | meta_guess: | |
3186 | /* calculate metadata allocation. */ | |
3187 | if (di->i_xattr_loc) { | |
3188 | if (!xbs->xattr_bh) { | |
4ae1d69b JB |
3189 | ret = ocfs2_read_xattr_block(inode, |
3190 | le64_to_cpu(di->i_xattr_loc), | |
3191 | &bh); | |
78f30c31 TM |
3192 | if (ret) { |
3193 | mlog_errno(ret); | |
3194 | goto out; | |
3195 | } | |
3196 | ||
3197 | xb = (struct ocfs2_xattr_block *)bh->b_data; | |
3198 | } else | |
3199 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
3200 | ||
90cb546c TM |
3201 | /* |
3202 | * If there is already an xattr tree, good, we can calculate | |
3203 | * like other b-trees. Otherwise we may have the chance of | |
3204 | * create a tree, the credit calculation is borrowed from | |
3205 | * ocfs2_calc_extend_credits with root_el = NULL. And the | |
3206 | * new tree will be cluster based, so no meta is needed. | |
3207 | */ | |
78f30c31 TM |
3208 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { |
3209 | struct ocfs2_extent_list *el = | |
3210 | &xb->xb_attrs.xb_root.xt_list; | |
3211 | meta_add += ocfs2_extend_meta_needed(el); | |
85db90e7 | 3212 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3213 | el); |
90cb546c TM |
3214 | } else |
3215 | credits += OCFS2_SUBALLOC_ALLOC + 1; | |
78f30c31 TM |
3216 | |
3217 | /* | |
3218 | * This cluster will be used either for new bucket or for | |
3219 | * new xattr block. | |
3220 | * If the cluster size is the same as the bucket size, one | |
3221 | * more is needed since we may need to extend the bucket | |
3222 | * also. | |
3223 | */ | |
3224 | clusters_add += 1; | |
85db90e7 | 3225 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
78f30c31 | 3226 | if (OCFS2_XATTR_BUCKET_SIZE == |
85db90e7 TM |
3227 | OCFS2_SB(inode->i_sb)->s_clustersize) { |
3228 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
78f30c31 | 3229 | clusters_add += 1; |
85db90e7 TM |
3230 | } |
3231 | } else { | |
85db90e7 | 3232 | credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
3ed2be71 TS |
3233 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
3234 | struct ocfs2_extent_list *el = &def_xv.xv.xr_list; | |
3235 | meta_add += ocfs2_extend_meta_needed(el); | |
3236 | credits += ocfs2_calc_extend_credits(inode->i_sb, | |
3237 | el); | |
3238 | } else { | |
3239 | meta_add += 1; | |
3240 | } | |
85db90e7 | 3241 | } |
78f30c31 TM |
3242 | out: |
3243 | if (clusters_need) | |
3244 | *clusters_need = clusters_add; | |
3245 | if (meta_need) | |
3246 | *meta_need = meta_add; | |
85db90e7 TM |
3247 | if (credits_need) |
3248 | *credits_need = credits; | |
78f30c31 TM |
3249 | brelse(bh); |
3250 | return ret; | |
3251 | } | |
3252 | ||
3253 | static int ocfs2_init_xattr_set_ctxt(struct inode *inode, | |
3254 | struct ocfs2_dinode *di, | |
3255 | struct ocfs2_xattr_info *xi, | |
3256 | struct ocfs2_xattr_search *xis, | |
3257 | struct ocfs2_xattr_search *xbs, | |
85db90e7 | 3258 | struct ocfs2_xattr_set_ctxt *ctxt, |
492a8a33 | 3259 | int extra_meta, |
85db90e7 | 3260 | int *credits) |
78f30c31 TM |
3261 | { |
3262 | int clusters_add, meta_add, ret; | |
3263 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
3264 | ||
3265 | memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt)); | |
3266 | ||
3267 | ocfs2_init_dealloc_ctxt(&ctxt->dealloc); | |
3268 | ||
3269 | ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs, | |
85db90e7 | 3270 | &clusters_add, &meta_add, credits); |
78f30c31 TM |
3271 | if (ret) { |
3272 | mlog_errno(ret); | |
3273 | return ret; | |
3274 | } | |
3275 | ||
492a8a33 | 3276 | meta_add += extra_meta; |
402b4183 TM |
3277 | trace_ocfs2_init_xattr_set_ctxt(xi->xi_name, meta_add, |
3278 | clusters_add, *credits); | |
78f30c31 TM |
3279 | |
3280 | if (meta_add) { | |
3281 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, | |
3282 | &ctxt->meta_ac); | |
3283 | if (ret) { | |
3284 | mlog_errno(ret); | |
3285 | goto out; | |
3286 | } | |
3287 | } | |
3288 | ||
3289 | if (clusters_add) { | |
3290 | ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac); | |
3291 | if (ret) | |
3292 | mlog_errno(ret); | |
3293 | } | |
3294 | out: | |
3295 | if (ret) { | |
3296 | if (ctxt->meta_ac) { | |
3297 | ocfs2_free_alloc_context(ctxt->meta_ac); | |
3298 | ctxt->meta_ac = NULL; | |
3299 | } | |
3300 | ||
3301 | /* | |
3302 | * We cannot have an error and a non null ctxt->data_ac. | |
3303 | */ | |
3304 | } | |
3305 | ||
3306 | return ret; | |
3307 | } | |
3308 | ||
85db90e7 TM |
3309 | static int __ocfs2_xattr_set_handle(struct inode *inode, |
3310 | struct ocfs2_dinode *di, | |
3311 | struct ocfs2_xattr_info *xi, | |
3312 | struct ocfs2_xattr_search *xis, | |
3313 | struct ocfs2_xattr_search *xbs, | |
3314 | struct ocfs2_xattr_set_ctxt *ctxt) | |
3315 | { | |
9f868f16 | 3316 | int ret = 0, credits, old_found; |
85db90e7 | 3317 | |
6b240ff6 | 3318 | if (!xi->xi_value) { |
85db90e7 TM |
3319 | /* Remove existing extended attribute */ |
3320 | if (!xis->not_found) | |
3321 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3322 | else if (!xbs->not_found) | |
3323 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3324 | } else { | |
3325 | /* We always try to set extended attribute into inode first*/ | |
3326 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3327 | if (!ret && !xbs->not_found) { | |
3328 | /* | |
3329 | * If succeed and that extended attribute existing in | |
3330 | * external block, then we will remove it. | |
3331 | */ | |
6b240ff6 JB |
3332 | xi->xi_value = NULL; |
3333 | xi->xi_value_len = 0; | |
85db90e7 | 3334 | |
9f868f16 | 3335 | old_found = xis->not_found; |
85db90e7 TM |
3336 | xis->not_found = -ENODATA; |
3337 | ret = ocfs2_calc_xattr_set_need(inode, | |
3338 | di, | |
3339 | xi, | |
3340 | xis, | |
3341 | xbs, | |
3342 | NULL, | |
3343 | NULL, | |
3344 | &credits); | |
9f868f16 | 3345 | xis->not_found = old_found; |
85db90e7 TM |
3346 | if (ret) { |
3347 | mlog_errno(ret); | |
3348 | goto out; | |
3349 | } | |
3350 | ||
c901fb00 | 3351 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3352 | if (ret) { |
3353 | mlog_errno(ret); | |
3354 | goto out; | |
3355 | } | |
3356 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
5f5261ac | 3357 | } else if ((ret == -ENOSPC) && !ctxt->set_abort) { |
85db90e7 TM |
3358 | if (di->i_xattr_loc && !xbs->xattr_bh) { |
3359 | ret = ocfs2_xattr_block_find(inode, | |
6b240ff6 JB |
3360 | xi->xi_name_index, |
3361 | xi->xi_name, xbs); | |
85db90e7 TM |
3362 | if (ret) |
3363 | goto out; | |
3364 | ||
9f868f16 | 3365 | old_found = xis->not_found; |
85db90e7 TM |
3366 | xis->not_found = -ENODATA; |
3367 | ret = ocfs2_calc_xattr_set_need(inode, | |
3368 | di, | |
3369 | xi, | |
3370 | xis, | |
3371 | xbs, | |
3372 | NULL, | |
3373 | NULL, | |
3374 | &credits); | |
9f868f16 | 3375 | xis->not_found = old_found; |
85db90e7 TM |
3376 | if (ret) { |
3377 | mlog_errno(ret); | |
3378 | goto out; | |
3379 | } | |
3380 | ||
c901fb00 | 3381 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3382 | if (ret) { |
3383 | mlog_errno(ret); | |
3384 | goto out; | |
3385 | } | |
3386 | } | |
3387 | /* | |
3388 | * If no space in inode, we will set extended attribute | |
3389 | * into external block. | |
3390 | */ | |
3391 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3392 | if (ret) | |
3393 | goto out; | |
3394 | if (!xis->not_found) { | |
3395 | /* | |
3396 | * If succeed and that extended attribute | |
3397 | * existing in inode, we will remove it. | |
3398 | */ | |
6b240ff6 JB |
3399 | xi->xi_value = NULL; |
3400 | xi->xi_value_len = 0; | |
85db90e7 TM |
3401 | xbs->not_found = -ENODATA; |
3402 | ret = ocfs2_calc_xattr_set_need(inode, | |
3403 | di, | |
3404 | xi, | |
3405 | xis, | |
3406 | xbs, | |
3407 | NULL, | |
3408 | NULL, | |
3409 | &credits); | |
3410 | if (ret) { | |
3411 | mlog_errno(ret); | |
3412 | goto out; | |
3413 | } | |
3414 | ||
c901fb00 | 3415 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3416 | if (ret) { |
3417 | mlog_errno(ret); | |
3418 | goto out; | |
3419 | } | |
3420 | ret = ocfs2_xattr_ibody_set(inode, xi, | |
3421 | xis, ctxt); | |
3422 | } | |
3423 | } | |
3424 | } | |
3425 | ||
4b3f6209 TM |
3426 | if (!ret) { |
3427 | /* Update inode ctime. */ | |
0cf2f763 | 3428 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
89a907af TM |
3429 | xis->inode_bh, |
3430 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4b3f6209 TM |
3431 | if (ret) { |
3432 | mlog_errno(ret); | |
3433 | goto out; | |
3434 | } | |
3435 | ||
078cd827 | 3436 | inode->i_ctime = current_time(inode); |
4b3f6209 TM |
3437 | di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); |
3438 | di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); | |
3439 | ocfs2_journal_dirty(ctxt->handle, xis->inode_bh); | |
3440 | } | |
85db90e7 TM |
3441 | out: |
3442 | return ret; | |
3443 | } | |
3444 | ||
6c3faba4 TY |
3445 | /* |
3446 | * This function only called duing creating inode | |
3447 | * for init security/acl xattrs of the new inode. | |
008aafaf | 3448 | * All transanction credits have been reserved in mknod. |
6c3faba4 TY |
3449 | */ |
3450 | int ocfs2_xattr_set_handle(handle_t *handle, | |
3451 | struct inode *inode, | |
3452 | struct buffer_head *di_bh, | |
3453 | int name_index, | |
3454 | const char *name, | |
3455 | const void *value, | |
3456 | size_t value_len, | |
3457 | int flags, | |
3458 | struct ocfs2_alloc_context *meta_ac, | |
3459 | struct ocfs2_alloc_context *data_ac) | |
3460 | { | |
3461 | struct ocfs2_dinode *di; | |
3462 | int ret; | |
3463 | ||
3464 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3465 | .xi_name_index = name_index, |
3466 | .xi_name = name, | |
18853b95 | 3467 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3468 | .xi_value = value, |
3469 | .xi_value_len = value_len, | |
6c3faba4 TY |
3470 | }; |
3471 | ||
3472 | struct ocfs2_xattr_search xis = { | |
3473 | .not_found = -ENODATA, | |
3474 | }; | |
3475 | ||
3476 | struct ocfs2_xattr_search xbs = { | |
3477 | .not_found = -ENODATA, | |
3478 | }; | |
3479 | ||
3480 | struct ocfs2_xattr_set_ctxt ctxt = { | |
3481 | .handle = handle, | |
3482 | .meta_ac = meta_ac, | |
3483 | .data_ac = data_ac, | |
3484 | }; | |
3485 | ||
3486 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) | |
3487 | return -EOPNOTSUPP; | |
3488 | ||
008aafaf TY |
3489 | /* |
3490 | * In extreme situation, may need xattr bucket when | |
3491 | * block size is too small. And we have already reserved | |
3492 | * the credits for bucket in mknod. | |
3493 | */ | |
3494 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) { | |
3495 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3496 | if (!xbs.bucket) { | |
3497 | mlog_errno(-ENOMEM); | |
3498 | return -ENOMEM; | |
3499 | } | |
3500 | } | |
3501 | ||
6c3faba4 TY |
3502 | xis.inode_bh = xbs.inode_bh = di_bh; |
3503 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3504 | ||
3505 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3506 | ||
3507 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3508 | if (ret) | |
3509 | goto cleanup; | |
3510 | if (xis.not_found) { | |
3511 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3512 | if (ret) | |
3513 | goto cleanup; | |
3514 | } | |
3515 | ||
3516 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3517 | ||
3518 | cleanup: | |
3519 | up_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3520 | brelse(xbs.xattr_bh); | |
008aafaf | 3521 | ocfs2_xattr_bucket_free(xbs.bucket); |
6c3faba4 TY |
3522 | |
3523 | return ret; | |
3524 | } | |
3525 | ||
cf1d6c76 TY |
3526 | /* |
3527 | * ocfs2_xattr_set() | |
3528 | * | |
3529 | * Set, replace or remove an extended attribute for this inode. | |
3530 | * value is NULL to remove an existing extended attribute, else either | |
3531 | * create or replace an extended attribute. | |
3532 | */ | |
3533 | int ocfs2_xattr_set(struct inode *inode, | |
3534 | int name_index, | |
3535 | const char *name, | |
3536 | const void *value, | |
3537 | size_t value_len, | |
3538 | int flags) | |
3539 | { | |
3540 | struct buffer_head *di_bh = NULL; | |
3541 | struct ocfs2_dinode *di; | |
8818efaa | 3542 | int ret, credits, had_lock, ref_meta = 0, ref_credits = 0; |
78f30c31 | 3543 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
85db90e7 | 3544 | struct inode *tl_inode = osb->osb_tl_inode; |
6ea437a3 | 3545 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, NULL, }; |
492a8a33 | 3546 | struct ocfs2_refcount_tree *ref_tree = NULL; |
8818efaa | 3547 | struct ocfs2_lock_holder oh; |
cf1d6c76 TY |
3548 | |
3549 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3550 | .xi_name_index = name_index, |
3551 | .xi_name = name, | |
18853b95 | 3552 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3553 | .xi_value = value, |
3554 | .xi_value_len = value_len, | |
cf1d6c76 TY |
3555 | }; |
3556 | ||
3557 | struct ocfs2_xattr_search xis = { | |
3558 | .not_found = -ENODATA, | |
3559 | }; | |
3560 | ||
3561 | struct ocfs2_xattr_search xbs = { | |
3562 | .not_found = -ENODATA, | |
3563 | }; | |
3564 | ||
8154da3d TY |
3565 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
3566 | return -EOPNOTSUPP; | |
3567 | ||
ba937127 JB |
3568 | /* |
3569 | * Only xbs will be used on indexed trees. xis doesn't need a | |
3570 | * bucket. | |
3571 | */ | |
3572 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3573 | if (!xbs.bucket) { | |
3574 | mlog_errno(-ENOMEM); | |
3575 | return -ENOMEM; | |
3576 | } | |
3577 | ||
8818efaa ER |
3578 | had_lock = ocfs2_inode_lock_tracker(inode, &di_bh, 1, &oh); |
3579 | if (had_lock < 0) { | |
3580 | ret = had_lock; | |
cf1d6c76 | 3581 | mlog_errno(ret); |
ba937127 | 3582 | goto cleanup_nolock; |
cf1d6c76 TY |
3583 | } |
3584 | xis.inode_bh = xbs.inode_bh = di_bh; | |
3585 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3586 | ||
3587 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3588 | /* | |
3589 | * Scan inode and external block to find the same name | |
25985edc | 3590 | * extended attribute and collect search information. |
cf1d6c76 TY |
3591 | */ |
3592 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3593 | if (ret) | |
3594 | goto cleanup; | |
3595 | if (xis.not_found) { | |
3596 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3597 | if (ret) | |
3598 | goto cleanup; | |
3599 | } | |
3600 | ||
3601 | if (xis.not_found && xbs.not_found) { | |
3602 | ret = -ENODATA; | |
3603 | if (flags & XATTR_REPLACE) | |
3604 | goto cleanup; | |
3605 | ret = 0; | |
3606 | if (!value) | |
3607 | goto cleanup; | |
3608 | } else { | |
3609 | ret = -EEXIST; | |
3610 | if (flags & XATTR_CREATE) | |
3611 | goto cleanup; | |
3612 | } | |
3613 | ||
25985edc | 3614 | /* Check whether the value is refcounted and do some preparation. */ |
84e40080 | 3615 | if (ocfs2_is_refcount_inode(inode) && |
492a8a33 TM |
3616 | (!xis.not_found || !xbs.not_found)) { |
3617 | ret = ocfs2_prepare_refcount_xattr(inode, di, &xi, | |
3618 | &xis, &xbs, &ref_tree, | |
3619 | &ref_meta, &ref_credits); | |
3620 | if (ret) { | |
3621 | mlog_errno(ret); | |
3622 | goto cleanup; | |
3623 | } | |
3624 | } | |
85db90e7 | 3625 | |
5955102c | 3626 | inode_lock(tl_inode); |
85db90e7 TM |
3627 | |
3628 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
3629 | ret = __ocfs2_flush_truncate_log(osb); | |
3630 | if (ret < 0) { | |
5955102c | 3631 | inode_unlock(tl_inode); |
85db90e7 TM |
3632 | mlog_errno(ret); |
3633 | goto cleanup; | |
3634 | } | |
3635 | } | |
5955102c | 3636 | inode_unlock(tl_inode); |
85db90e7 TM |
3637 | |
3638 | ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis, | |
492a8a33 | 3639 | &xbs, &ctxt, ref_meta, &credits); |
78f30c31 TM |
3640 | if (ret) { |
3641 | mlog_errno(ret); | |
3642 | goto cleanup; | |
3643 | } | |
3644 | ||
4b3f6209 TM |
3645 | /* we need to update inode's ctime field, so add credit for it. */ |
3646 | credits += OCFS2_INODE_UPDATE_CREDITS; | |
492a8a33 | 3647 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); |
85db90e7 TM |
3648 | if (IS_ERR(ctxt.handle)) { |
3649 | ret = PTR_ERR(ctxt.handle); | |
3650 | mlog_errno(ret); | |
6ea437a3 | 3651 | goto out_free_ac; |
cf1d6c76 | 3652 | } |
85db90e7 TM |
3653 | |
3654 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
6fdb702d | 3655 | ocfs2_update_inode_fsync_trans(ctxt.handle, inode, 0); |
85db90e7 TM |
3656 | |
3657 | ocfs2_commit_trans(osb, ctxt.handle); | |
3658 | ||
6ea437a3 | 3659 | out_free_ac: |
78f30c31 TM |
3660 | if (ctxt.data_ac) |
3661 | ocfs2_free_alloc_context(ctxt.data_ac); | |
3662 | if (ctxt.meta_ac) | |
3663 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
3664 | if (ocfs2_dealloc_has_cluster(&ctxt.dealloc)) | |
3665 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
3666 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
8b2c0dba | 3667 | |
cf1d6c76 | 3668 | cleanup: |
492a8a33 TM |
3669 | if (ref_tree) |
3670 | ocfs2_unlock_refcount_tree(osb, ref_tree, 1); | |
cf1d6c76 | 3671 | up_write(&OCFS2_I(inode)->ip_xattr_sem); |
8b2c0dba TM |
3672 | if (!value && !ret) { |
3673 | ret = ocfs2_try_remove_refcount_tree(inode, di_bh); | |
3674 | if (ret) | |
3675 | mlog_errno(ret); | |
3676 | } | |
8818efaa | 3677 | ocfs2_inode_unlock_tracker(inode, 1, &oh, had_lock); |
ba937127 | 3678 | cleanup_nolock: |
cf1d6c76 TY |
3679 | brelse(di_bh); |
3680 | brelse(xbs.xattr_bh); | |
ba937127 | 3681 | ocfs2_xattr_bucket_free(xbs.bucket); |
cf1d6c76 TY |
3682 | |
3683 | return ret; | |
3684 | } | |
3685 | ||
0c044f0b TM |
3686 | /* |
3687 | * Find the xattr extent rec which may contains name_hash. | |
3688 | * e_cpos will be the first name hash of the xattr rec. | |
3689 | * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list. | |
3690 | */ | |
3691 | static int ocfs2_xattr_get_rec(struct inode *inode, | |
3692 | u32 name_hash, | |
3693 | u64 *p_blkno, | |
3694 | u32 *e_cpos, | |
3695 | u32 *num_clusters, | |
3696 | struct ocfs2_extent_list *el) | |
3697 | { | |
3698 | int ret = 0, i; | |
3699 | struct buffer_head *eb_bh = NULL; | |
3700 | struct ocfs2_extent_block *eb; | |
3701 | struct ocfs2_extent_rec *rec = NULL; | |
3702 | u64 e_blkno = 0; | |
3703 | ||
3704 | if (el->l_tree_depth) { | |
facdb77f JB |
3705 | ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash, |
3706 | &eb_bh); | |
0c044f0b TM |
3707 | if (ret) { |
3708 | mlog_errno(ret); | |
3709 | goto out; | |
3710 | } | |
3711 | ||
3712 | eb = (struct ocfs2_extent_block *) eb_bh->b_data; | |
3713 | el = &eb->h_list; | |
3714 | ||
3715 | if (el->l_tree_depth) { | |
17a5b9ab | 3716 | ret = ocfs2_error(inode->i_sb, |
7ecef14a JP |
3717 | "Inode %lu has non zero tree depth in xattr tree block %llu\n", |
3718 | inode->i_ino, | |
3719 | (unsigned long long)eb_bh->b_blocknr); | |
0c044f0b TM |
3720 | goto out; |
3721 | } | |
3722 | } | |
3723 | ||
3724 | for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { | |
3725 | rec = &el->l_recs[i]; | |
3726 | ||
3727 | if (le32_to_cpu(rec->e_cpos) <= name_hash) { | |
3728 | e_blkno = le64_to_cpu(rec->e_blkno); | |
3729 | break; | |
3730 | } | |
3731 | } | |
3732 | ||
3733 | if (!e_blkno) { | |
7ecef14a JP |
3734 | ret = ocfs2_error(inode->i_sb, "Inode %lu has bad extent record (%u, %u, 0) in xattr\n", |
3735 | inode->i_ino, | |
3736 | le32_to_cpu(rec->e_cpos), | |
3737 | ocfs2_rec_clusters(el, rec)); | |
0c044f0b TM |
3738 | goto out; |
3739 | } | |
3740 | ||
3741 | *p_blkno = le64_to_cpu(rec->e_blkno); | |
3742 | *num_clusters = le16_to_cpu(rec->e_leaf_clusters); | |
3743 | if (e_cpos) | |
3744 | *e_cpos = le32_to_cpu(rec->e_cpos); | |
3745 | out: | |
3746 | brelse(eb_bh); | |
3747 | return ret; | |
3748 | } | |
3749 | ||
3750 | typedef int (xattr_bucket_func)(struct inode *inode, | |
3751 | struct ocfs2_xattr_bucket *bucket, | |
3752 | void *para); | |
3753 | ||
589dc260 | 3754 | static int ocfs2_find_xe_in_bucket(struct inode *inode, |
e2356a3f | 3755 | struct ocfs2_xattr_bucket *bucket, |
589dc260 TM |
3756 | int name_index, |
3757 | const char *name, | |
3758 | u32 name_hash, | |
3759 | u16 *xe_index, | |
3760 | int *found) | |
3761 | { | |
3762 | int i, ret = 0, cmp = 1, block_off, new_offset; | |
e2356a3f | 3763 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
589dc260 TM |
3764 | size_t name_len = strlen(name); |
3765 | struct ocfs2_xattr_entry *xe = NULL; | |
589dc260 TM |
3766 | char *xe_name; |
3767 | ||
3768 | /* | |
3769 | * We don't use binary search in the bucket because there | |
3770 | * may be multiple entries with the same name hash. | |
3771 | */ | |
3772 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
3773 | xe = &xh->xh_entries[i]; | |
3774 | ||
3775 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) | |
3776 | continue; | |
3777 | else if (name_hash < le32_to_cpu(xe->xe_name_hash)) | |
3778 | break; | |
3779 | ||
3780 | cmp = name_index - ocfs2_xattr_get_type(xe); | |
3781 | if (!cmp) | |
3782 | cmp = name_len - xe->xe_name_len; | |
3783 | if (cmp) | |
3784 | continue; | |
3785 | ||
fd68a894 | 3786 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
589dc260 TM |
3787 | xh, |
3788 | i, | |
3789 | &block_off, | |
3790 | &new_offset); | |
3791 | if (ret) { | |
3792 | mlog_errno(ret); | |
3793 | break; | |
3794 | } | |
3795 | ||
970e4936 | 3796 | |
e2356a3f JB |
3797 | xe_name = bucket_block(bucket, block_off) + new_offset; |
3798 | if (!memcmp(name, xe_name, name_len)) { | |
589dc260 TM |
3799 | *xe_index = i; |
3800 | *found = 1; | |
3801 | ret = 0; | |
3802 | break; | |
3803 | } | |
3804 | } | |
3805 | ||
3806 | return ret; | |
3807 | } | |
3808 | ||
3809 | /* | |
3810 | * Find the specified xattr entry in a series of buckets. | |
3811 | * This series start from p_blkno and last for num_clusters. | |
3812 | * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains | |
3813 | * the num of the valid buckets. | |
3814 | * | |
3815 | * Return the buffer_head this xattr should reside in. And if the xattr's | |
3816 | * hash is in the gap of 2 buckets, return the lower bucket. | |
3817 | */ | |
3818 | static int ocfs2_xattr_bucket_find(struct inode *inode, | |
3819 | int name_index, | |
3820 | const char *name, | |
3821 | u32 name_hash, | |
3822 | u64 p_blkno, | |
3823 | u32 first_hash, | |
3824 | u32 num_clusters, | |
3825 | struct ocfs2_xattr_search *xs) | |
3826 | { | |
3827 | int ret, found = 0; | |
589dc260 TM |
3828 | struct ocfs2_xattr_header *xh = NULL; |
3829 | struct ocfs2_xattr_entry *xe = NULL; | |
3830 | u16 index = 0; | |
3831 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
3832 | int low_bucket = 0, bucket, high_bucket; | |
e2356a3f | 3833 | struct ocfs2_xattr_bucket *search; |
589dc260 | 3834 | u32 last_hash; |
e2356a3f | 3835 | u64 blkno, lower_blkno = 0; |
589dc260 | 3836 | |
e2356a3f JB |
3837 | search = ocfs2_xattr_bucket_new(inode); |
3838 | if (!search) { | |
3839 | ret = -ENOMEM; | |
3840 | mlog_errno(ret); | |
3841 | goto out; | |
3842 | } | |
3843 | ||
3844 | ret = ocfs2_read_xattr_bucket(search, p_blkno); | |
589dc260 TM |
3845 | if (ret) { |
3846 | mlog_errno(ret); | |
3847 | goto out; | |
3848 | } | |
3849 | ||
e2356a3f | 3850 | xh = bucket_xh(search); |
589dc260 | 3851 | high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1; |
589dc260 | 3852 | while (low_bucket <= high_bucket) { |
e2356a3f | 3853 | ocfs2_xattr_bucket_relse(search); |
589dc260 | 3854 | |
e2356a3f | 3855 | bucket = (low_bucket + high_bucket) / 2; |
589dc260 | 3856 | blkno = p_blkno + bucket * blk_per_bucket; |
e2356a3f | 3857 | ret = ocfs2_read_xattr_bucket(search, blkno); |
589dc260 TM |
3858 | if (ret) { |
3859 | mlog_errno(ret); | |
3860 | goto out; | |
3861 | } | |
3862 | ||
e2356a3f | 3863 | xh = bucket_xh(search); |
589dc260 TM |
3864 | xe = &xh->xh_entries[0]; |
3865 | if (name_hash < le32_to_cpu(xe->xe_name_hash)) { | |
3866 | high_bucket = bucket - 1; | |
3867 | continue; | |
3868 | } | |
3869 | ||
3870 | /* | |
3871 | * Check whether the hash of the last entry in our | |
5a095611 TM |
3872 | * bucket is larger than the search one. for an empty |
3873 | * bucket, the last one is also the first one. | |
589dc260 | 3874 | */ |
5a095611 TM |
3875 | if (xh->xh_count) |
3876 | xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1]; | |
3877 | ||
589dc260 TM |
3878 | last_hash = le32_to_cpu(xe->xe_name_hash); |
3879 | ||
e2356a3f JB |
3880 | /* record lower_blkno which may be the insert place. */ |
3881 | lower_blkno = blkno; | |
589dc260 TM |
3882 | |
3883 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) { | |
3884 | low_bucket = bucket + 1; | |
3885 | continue; | |
3886 | } | |
3887 | ||
3888 | /* the searched xattr should reside in this bucket if exists. */ | |
e2356a3f | 3889 | ret = ocfs2_find_xe_in_bucket(inode, search, |
589dc260 TM |
3890 | name_index, name, name_hash, |
3891 | &index, &found); | |
3892 | if (ret) { | |
3893 | mlog_errno(ret); | |
3894 | goto out; | |
3895 | } | |
3896 | break; | |
3897 | } | |
3898 | ||
3899 | /* | |
3900 | * Record the bucket we have found. | |
3901 | * When the xattr's hash value is in the gap of 2 buckets, we will | |
3902 | * always set it to the previous bucket. | |
3903 | */ | |
e2356a3f JB |
3904 | if (!lower_blkno) |
3905 | lower_blkno = p_blkno; | |
3906 | ||
3907 | /* This should be in cache - we just read it during the search */ | |
3908 | ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno); | |
3909 | if (ret) { | |
3910 | mlog_errno(ret); | |
3911 | goto out; | |
589dc260 | 3912 | } |
589dc260 | 3913 | |
ba937127 JB |
3914 | xs->header = bucket_xh(xs->bucket); |
3915 | xs->base = bucket_block(xs->bucket, 0); | |
589dc260 TM |
3916 | xs->end = xs->base + inode->i_sb->s_blocksize; |
3917 | ||
3918 | if (found) { | |
589dc260 | 3919 | xs->here = &xs->header->xh_entries[index]; |
402b4183 TM |
3920 | trace_ocfs2_xattr_bucket_find(OCFS2_I(inode)->ip_blkno, |
3921 | name, name_index, name_hash, | |
3922 | (unsigned long long)bucket_blkno(xs->bucket), | |
3923 | index); | |
589dc260 TM |
3924 | } else |
3925 | ret = -ENODATA; | |
3926 | ||
3927 | out: | |
e2356a3f | 3928 | ocfs2_xattr_bucket_free(search); |
589dc260 TM |
3929 | return ret; |
3930 | } | |
3931 | ||
3932 | static int ocfs2_xattr_index_block_find(struct inode *inode, | |
3933 | struct buffer_head *root_bh, | |
3934 | int name_index, | |
3935 | const char *name, | |
3936 | struct ocfs2_xattr_search *xs) | |
3937 | { | |
3938 | int ret; | |
3939 | struct ocfs2_xattr_block *xb = | |
3940 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
3941 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
3942 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
3943 | u64 p_blkno = 0; | |
3944 | u32 first_hash, num_clusters = 0; | |
2057e5c6 | 3945 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
589dc260 TM |
3946 | |
3947 | if (le16_to_cpu(el->l_next_free_rec) == 0) | |
3948 | return -ENODATA; | |
3949 | ||
402b4183 TM |
3950 | trace_ocfs2_xattr_index_block_find(OCFS2_I(inode)->ip_blkno, |
3951 | name, name_index, name_hash, | |
3952 | (unsigned long long)root_bh->b_blocknr, | |
3953 | -1); | |
589dc260 TM |
3954 | |
3955 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash, | |
3956 | &num_clusters, el); | |
3957 | if (ret) { | |
3958 | mlog_errno(ret); | |
3959 | goto out; | |
3960 | } | |
3961 | ||
3962 | BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash); | |
3963 | ||
402b4183 TM |
3964 | trace_ocfs2_xattr_index_block_find_rec(OCFS2_I(inode)->ip_blkno, |
3965 | name, name_index, first_hash, | |
3966 | (unsigned long long)p_blkno, | |
3967 | num_clusters); | |
589dc260 TM |
3968 | |
3969 | ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash, | |
3970 | p_blkno, first_hash, num_clusters, xs); | |
3971 | ||
3972 | out: | |
3973 | return ret; | |
3974 | } | |
3975 | ||
0c044f0b TM |
3976 | static int ocfs2_iterate_xattr_buckets(struct inode *inode, |
3977 | u64 blkno, | |
3978 | u32 clusters, | |
3979 | xattr_bucket_func *func, | |
3980 | void *para) | |
3981 | { | |
6dde41d9 | 3982 | int i, ret = 0; |
0c044f0b TM |
3983 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); |
3984 | u32 num_buckets = clusters * bpc; | |
ba937127 | 3985 | struct ocfs2_xattr_bucket *bucket; |
0c044f0b | 3986 | |
ba937127 JB |
3987 | bucket = ocfs2_xattr_bucket_new(inode); |
3988 | if (!bucket) { | |
3989 | mlog_errno(-ENOMEM); | |
3990 | return -ENOMEM; | |
3991 | } | |
0c044f0b | 3992 | |
402b4183 TM |
3993 | trace_ocfs2_iterate_xattr_buckets( |
3994 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
3995 | (unsigned long long)blkno, clusters); | |
0c044f0b | 3996 | |
ba937127 JB |
3997 | for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) { |
3998 | ret = ocfs2_read_xattr_bucket(bucket, blkno); | |
0c044f0b TM |
3999 | if (ret) { |
4000 | mlog_errno(ret); | |
ba937127 | 4001 | break; |
0c044f0b TM |
4002 | } |
4003 | ||
0c044f0b TM |
4004 | /* |
4005 | * The real bucket num in this series of blocks is stored | |
4006 | * in the 1st bucket. | |
4007 | */ | |
4008 | if (i == 0) | |
ba937127 | 4009 | num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets); |
0c044f0b | 4010 | |
402b4183 | 4011 | trace_ocfs2_iterate_xattr_bucket((unsigned long long)blkno, |
ba937127 | 4012 | le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); |
0c044f0b | 4013 | if (func) { |
ba937127 | 4014 | ret = func(inode, bucket, para); |
a46fa684 | 4015 | if (ret && ret != -ERANGE) |
0c044f0b | 4016 | mlog_errno(ret); |
ba937127 | 4017 | /* Fall through to bucket_relse() */ |
0c044f0b TM |
4018 | } |
4019 | ||
ba937127 JB |
4020 | ocfs2_xattr_bucket_relse(bucket); |
4021 | if (ret) | |
4022 | break; | |
0c044f0b TM |
4023 | } |
4024 | ||
ba937127 | 4025 | ocfs2_xattr_bucket_free(bucket); |
0c044f0b TM |
4026 | return ret; |
4027 | } | |
4028 | ||
4029 | struct ocfs2_xattr_tree_list { | |
4030 | char *buffer; | |
4031 | size_t buffer_size; | |
936b8834 | 4032 | size_t result; |
0c044f0b TM |
4033 | }; |
4034 | ||
fd68a894 | 4035 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
0c044f0b TM |
4036 | struct ocfs2_xattr_header *xh, |
4037 | int index, | |
4038 | int *block_off, | |
4039 | int *new_offset) | |
4040 | { | |
4041 | u16 name_offset; | |
4042 | ||
4043 | if (index < 0 || index >= le16_to_cpu(xh->xh_count)) | |
4044 | return -EINVAL; | |
4045 | ||
4046 | name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset); | |
4047 | ||
fd68a894 TM |
4048 | *block_off = name_offset >> sb->s_blocksize_bits; |
4049 | *new_offset = name_offset % sb->s_blocksize; | |
0c044f0b TM |
4050 | |
4051 | return 0; | |
4052 | } | |
4053 | ||
4054 | static int ocfs2_list_xattr_bucket(struct inode *inode, | |
4055 | struct ocfs2_xattr_bucket *bucket, | |
4056 | void *para) | |
4057 | { | |
936b8834 | 4058 | int ret = 0, type; |
0c044f0b | 4059 | struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para; |
0c044f0b | 4060 | int i, block_off, new_offset; |
1046cb11 | 4061 | const char *name; |
0c044f0b | 4062 | |
3e632946 JB |
4063 | for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) { |
4064 | struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i]; | |
936b8834 | 4065 | type = ocfs2_xattr_get_type(entry); |
0c044f0b | 4066 | |
1046cb11 AG |
4067 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
4068 | bucket_xh(bucket), | |
4069 | i, | |
4070 | &block_off, | |
4071 | &new_offset); | |
4072 | if (ret) | |
4073 | break; | |
936b8834 | 4074 | |
1046cb11 AG |
4075 | name = (const char *)bucket_block(bucket, block_off) + |
4076 | new_offset; | |
4077 | ret = ocfs2_xattr_list_entry(inode->i_sb, | |
4078 | xl->buffer, | |
4079 | xl->buffer_size, | |
4080 | &xl->result, | |
4081 | type, name, | |
4082 | entry->xe_name_len); | |
4083 | if (ret) | |
4084 | break; | |
0c044f0b TM |
4085 | } |
4086 | ||
4087 | return ret; | |
4088 | } | |
4089 | ||
47bca495 TM |
4090 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, |
4091 | struct buffer_head *blk_bh, | |
4092 | xattr_tree_rec_func *rec_func, | |
4093 | void *para) | |
0c044f0b | 4094 | { |
47bca495 TM |
4095 | struct ocfs2_xattr_block *xb = |
4096 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
4097 | struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; | |
0c044f0b TM |
4098 | int ret = 0; |
4099 | u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0; | |
4100 | u64 p_blkno = 0; | |
0c044f0b | 4101 | |
47bca495 | 4102 | if (!el->l_next_free_rec || !rec_func) |
0c044f0b TM |
4103 | return 0; |
4104 | ||
4105 | while (name_hash > 0) { | |
4106 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, | |
4107 | &e_cpos, &num_clusters, el); | |
4108 | if (ret) { | |
4109 | mlog_errno(ret); | |
47bca495 | 4110 | break; |
0c044f0b TM |
4111 | } |
4112 | ||
47bca495 TM |
4113 | ret = rec_func(inode, blk_bh, p_blkno, e_cpos, |
4114 | num_clusters, para); | |
0c044f0b | 4115 | if (ret) { |
a46fa684 TM |
4116 | if (ret != -ERANGE) |
4117 | mlog_errno(ret); | |
47bca495 | 4118 | break; |
0c044f0b TM |
4119 | } |
4120 | ||
4121 | if (e_cpos == 0) | |
4122 | break; | |
4123 | ||
4124 | name_hash = e_cpos - 1; | |
4125 | } | |
4126 | ||
47bca495 TM |
4127 | return ret; |
4128 | ||
4129 | } | |
4130 | ||
4131 | static int ocfs2_list_xattr_tree_rec(struct inode *inode, | |
4132 | struct buffer_head *root_bh, | |
4133 | u64 blkno, u32 cpos, u32 len, void *para) | |
4134 | { | |
4135 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
4136 | ocfs2_list_xattr_bucket, para); | |
4137 | } | |
4138 | ||
4139 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, | |
4140 | struct buffer_head *blk_bh, | |
4141 | char *buffer, | |
4142 | size_t buffer_size) | |
4143 | { | |
4144 | int ret; | |
4145 | struct ocfs2_xattr_tree_list xl = { | |
4146 | .buffer = buffer, | |
4147 | .buffer_size = buffer_size, | |
4148 | .result = 0, | |
4149 | }; | |
4150 | ||
4151 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
4152 | ocfs2_list_xattr_tree_rec, &xl); | |
4153 | if (ret) { | |
4154 | mlog_errno(ret); | |
4155 | goto out; | |
4156 | } | |
4157 | ||
936b8834 | 4158 | ret = xl.result; |
0c044f0b TM |
4159 | out: |
4160 | return ret; | |
4161 | } | |
01225596 TM |
4162 | |
4163 | static int cmp_xe(const void *a, const void *b) | |
4164 | { | |
4165 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4166 | u32 l_hash = le32_to_cpu(l->xe_name_hash); | |
4167 | u32 r_hash = le32_to_cpu(r->xe_name_hash); | |
4168 | ||
4169 | if (l_hash > r_hash) | |
4170 | return 1; | |
4171 | if (l_hash < r_hash) | |
4172 | return -1; | |
4173 | return 0; | |
4174 | } | |
4175 | ||
4176 | static void swap_xe(void *a, void *b, int size) | |
4177 | { | |
4178 | struct ocfs2_xattr_entry *l = a, *r = b, tmp; | |
4179 | ||
4180 | tmp = *l; | |
4181 | memcpy(l, r, sizeof(struct ocfs2_xattr_entry)); | |
4182 | memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry)); | |
4183 | } | |
4184 | ||
4185 | /* | |
4186 | * When the ocfs2_xattr_block is filled up, new bucket will be created | |
4187 | * and all the xattr entries will be moved to the new bucket. | |
178eeac3 JB |
4188 | * The header goes at the start of the bucket, and the names+values are |
4189 | * filled from the end. This is why *target starts as the last buffer. | |
01225596 TM |
4190 | * Note: we need to sort the entries since they are not saved in order |
4191 | * in the ocfs2_xattr_block. | |
4192 | */ | |
4193 | static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, | |
4194 | struct buffer_head *xb_bh, | |
178eeac3 | 4195 | struct ocfs2_xattr_bucket *bucket) |
01225596 TM |
4196 | { |
4197 | int i, blocksize = inode->i_sb->s_blocksize; | |
178eeac3 | 4198 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 TM |
4199 | u16 offset, size, off_change; |
4200 | struct ocfs2_xattr_entry *xe; | |
4201 | struct ocfs2_xattr_block *xb = | |
4202 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4203 | struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header; | |
178eeac3 | 4204 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 4205 | u16 count = le16_to_cpu(xb_xh->xh_count); |
178eeac3 JB |
4206 | char *src = xb_bh->b_data; |
4207 | char *target = bucket_block(bucket, blks - 1); | |
01225596 | 4208 | |
402b4183 TM |
4209 | trace_ocfs2_cp_xattr_block_to_bucket_begin( |
4210 | (unsigned long long)xb_bh->b_blocknr, | |
4211 | (unsigned long long)bucket_blkno(bucket)); | |
178eeac3 JB |
4212 | |
4213 | for (i = 0; i < blks; i++) | |
4214 | memset(bucket_block(bucket, i), 0, blocksize); | |
01225596 | 4215 | |
01225596 TM |
4216 | /* |
4217 | * Since the xe_name_offset is based on ocfs2_xattr_header, | |
4218 | * there is a offset change corresponding to the change of | |
4219 | * ocfs2_xattr_header's position. | |
4220 | */ | |
4221 | off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4222 | xe = &xb_xh->xh_entries[count - 1]; | |
4223 | offset = le16_to_cpu(xe->xe_name_offset) + off_change; | |
4224 | size = blocksize - offset; | |
4225 | ||
4226 | /* copy all the names and values. */ | |
01225596 TM |
4227 | memcpy(target + offset, src + offset, size); |
4228 | ||
4229 | /* Init new header now. */ | |
4230 | xh->xh_count = xb_xh->xh_count; | |
4231 | xh->xh_num_buckets = cpu_to_le16(1); | |
4232 | xh->xh_name_value_len = cpu_to_le16(size); | |
4233 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size); | |
4234 | ||
4235 | /* copy all the entries. */ | |
178eeac3 | 4236 | target = bucket_block(bucket, 0); |
01225596 TM |
4237 | offset = offsetof(struct ocfs2_xattr_header, xh_entries); |
4238 | size = count * sizeof(struct ocfs2_xattr_entry); | |
4239 | memcpy(target + offset, (char *)xb_xh + offset, size); | |
4240 | ||
4241 | /* Change the xe offset for all the xe because of the move. */ | |
4242 | off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize + | |
4243 | offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4244 | for (i = 0; i < count; i++) | |
4245 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change); | |
4246 | ||
402b4183 | 4247 | trace_ocfs2_cp_xattr_block_to_bucket_end(offset, size, off_change); |
01225596 TM |
4248 | |
4249 | sort(target + offset, count, sizeof(struct ocfs2_xattr_entry), | |
4250 | cmp_xe, swap_xe); | |
4251 | } | |
4252 | ||
4253 | /* | |
4254 | * After we move xattr from block to index btree, we have to | |
4255 | * update ocfs2_xattr_search to the new xe and base. | |
4256 | * | |
4257 | * When the entry is in xattr block, xattr_bh indicates the storage place. | |
4258 | * While if the entry is in index b-tree, "bucket" indicates the | |
4259 | * real place of the xattr. | |
4260 | */ | |
178eeac3 JB |
4261 | static void ocfs2_xattr_update_xattr_search(struct inode *inode, |
4262 | struct ocfs2_xattr_search *xs, | |
4263 | struct buffer_head *old_bh) | |
01225596 | 4264 | { |
01225596 TM |
4265 | char *buf = old_bh->b_data; |
4266 | struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf; | |
4267 | struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header; | |
178eeac3 | 4268 | int i; |
01225596 | 4269 | |
ba937127 | 4270 | xs->header = bucket_xh(xs->bucket); |
178eeac3 | 4271 | xs->base = bucket_block(xs->bucket, 0); |
01225596 TM |
4272 | xs->end = xs->base + inode->i_sb->s_blocksize; |
4273 | ||
178eeac3 JB |
4274 | if (xs->not_found) |
4275 | return; | |
01225596 | 4276 | |
178eeac3 JB |
4277 | i = xs->here - old_xh->xh_entries; |
4278 | xs->here = &xs->header->xh_entries[i]; | |
01225596 TM |
4279 | } |
4280 | ||
4281 | static int ocfs2_xattr_create_index_block(struct inode *inode, | |
78f30c31 TM |
4282 | struct ocfs2_xattr_search *xs, |
4283 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 4284 | { |
85db90e7 | 4285 | int ret; |
01225596 TM |
4286 | u32 bit_off, len; |
4287 | u64 blkno; | |
85db90e7 | 4288 | handle_t *handle = ctxt->handle; |
01225596 | 4289 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
01225596 TM |
4290 | struct buffer_head *xb_bh = xs->xattr_bh; |
4291 | struct ocfs2_xattr_block *xb = | |
4292 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4293 | struct ocfs2_xattr_tree_root *xr; | |
4294 | u16 xb_flags = le16_to_cpu(xb->xb_flags); | |
01225596 | 4295 | |
402b4183 TM |
4296 | trace_ocfs2_xattr_create_index_block_begin( |
4297 | (unsigned long long)xb_bh->b_blocknr); | |
01225596 TM |
4298 | |
4299 | BUG_ON(xb_flags & OCFS2_XATTR_INDEXED); | |
178eeac3 | 4300 | BUG_ON(!xs->bucket); |
01225596 | 4301 | |
01225596 TM |
4302 | /* |
4303 | * XXX: | |
4304 | * We can use this lock for now, and maybe move to a dedicated mutex | |
4305 | * if performance becomes a problem later. | |
4306 | */ | |
4307 | down_write(&oi->ip_alloc_sem); | |
4308 | ||
0cf2f763 | 4309 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh, |
84008972 | 4310 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4311 | if (ret) { |
4312 | mlog_errno(ret); | |
85db90e7 | 4313 | goto out; |
01225596 TM |
4314 | } |
4315 | ||
1ed9b777 | 4316 | ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, |
78f30c31 | 4317 | 1, 1, &bit_off, &len); |
01225596 TM |
4318 | if (ret) { |
4319 | mlog_errno(ret); | |
85db90e7 | 4320 | goto out; |
01225596 TM |
4321 | } |
4322 | ||
4323 | /* | |
4324 | * The bucket may spread in many blocks, and | |
4325 | * we will only touch the 1st block and the last block | |
4326 | * in the whole bucket(one for entry and one for data). | |
4327 | */ | |
4328 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); | |
4329 | ||
402b4183 | 4330 | trace_ocfs2_xattr_create_index_block((unsigned long long)blkno); |
01225596 | 4331 | |
9c339255 | 4332 | ret = ocfs2_init_xattr_bucket(xs->bucket, blkno, 1); |
178eeac3 | 4333 | if (ret) { |
01225596 | 4334 | mlog_errno(ret); |
85db90e7 | 4335 | goto out; |
01225596 TM |
4336 | } |
4337 | ||
178eeac3 JB |
4338 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
4339 | OCFS2_JOURNAL_ACCESS_CREATE); | |
01225596 TM |
4340 | if (ret) { |
4341 | mlog_errno(ret); | |
85db90e7 | 4342 | goto out; |
01225596 TM |
4343 | } |
4344 | ||
178eeac3 JB |
4345 | ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket); |
4346 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); | |
01225596 | 4347 | |
178eeac3 | 4348 | ocfs2_xattr_update_xattr_search(inode, xs, xb_bh); |
01225596 TM |
4349 | |
4350 | /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */ | |
4351 | memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize - | |
4352 | offsetof(struct ocfs2_xattr_block, xb_attrs)); | |
4353 | ||
4354 | xr = &xb->xb_attrs.xb_root; | |
4355 | xr->xt_clusters = cpu_to_le32(1); | |
4356 | xr->xt_last_eb_blk = 0; | |
4357 | xr->xt_list.l_tree_depth = 0; | |
4358 | xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
4359 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
4360 | ||
4361 | xr->xt_list.l_recs[0].e_cpos = 0; | |
4362 | xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno); | |
4363 | xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1); | |
4364 | ||
4365 | xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED); | |
4366 | ||
85db90e7 | 4367 | ocfs2_journal_dirty(handle, xb_bh); |
01225596 | 4368 | |
85db90e7 | 4369 | out: |
01225596 TM |
4370 | up_write(&oi->ip_alloc_sem); |
4371 | ||
01225596 TM |
4372 | return ret; |
4373 | } | |
4374 | ||
4375 | static int cmp_xe_offset(const void *a, const void *b) | |
4376 | { | |
4377 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4378 | u32 l_name_offset = le16_to_cpu(l->xe_name_offset); | |
4379 | u32 r_name_offset = le16_to_cpu(r->xe_name_offset); | |
4380 | ||
4381 | if (l_name_offset < r_name_offset) | |
4382 | return 1; | |
4383 | if (l_name_offset > r_name_offset) | |
4384 | return -1; | |
4385 | return 0; | |
4386 | } | |
4387 | ||
4388 | /* | |
4389 | * defrag a xattr bucket if we find that the bucket has some | |
4390 | * holes beteen name/value pairs. | |
4391 | * We will move all the name/value pairs to the end of the bucket | |
4392 | * so that we can spare some space for insertion. | |
4393 | */ | |
4394 | static int ocfs2_defrag_xattr_bucket(struct inode *inode, | |
85db90e7 | 4395 | handle_t *handle, |
01225596 TM |
4396 | struct ocfs2_xattr_bucket *bucket) |
4397 | { | |
4398 | int ret, i; | |
199799a3 | 4399 | size_t end, offset, len; |
01225596 TM |
4400 | struct ocfs2_xattr_header *xh; |
4401 | char *entries, *buf, *bucket_buf = NULL; | |
9c7759aa | 4402 | u64 blkno = bucket_blkno(bucket); |
01225596 | 4403 | u16 xh_free_start; |
01225596 | 4404 | size_t blocksize = inode->i_sb->s_blocksize; |
01225596 | 4405 | struct ocfs2_xattr_entry *xe; |
01225596 TM |
4406 | |
4407 | /* | |
4408 | * In order to make the operation more efficient and generic, | |
4409 | * we copy all the blocks into a contiguous memory and do the | |
4410 | * defragment there, so if anything is error, we will not touch | |
4411 | * the real block. | |
4412 | */ | |
4413 | bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS); | |
4414 | if (!bucket_buf) { | |
4415 | ret = -EIO; | |
4416 | goto out; | |
4417 | } | |
4418 | ||
4419 | buf = bucket_buf; | |
1c32a2fd TM |
4420 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4421 | memcpy(buf, bucket_block(bucket, i), blocksize); | |
01225596 | 4422 | |
1c32a2fd | 4423 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, |
161d6f30 JB |
4424 | OCFS2_JOURNAL_ACCESS_WRITE); |
4425 | if (ret < 0) { | |
4426 | mlog_errno(ret); | |
85db90e7 | 4427 | goto out; |
01225596 TM |
4428 | } |
4429 | ||
4430 | xh = (struct ocfs2_xattr_header *)bucket_buf; | |
4431 | entries = (char *)xh->xh_entries; | |
4432 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4433 | ||
402b4183 | 4434 | trace_ocfs2_defrag_xattr_bucket( |
de29c085 MF |
4435 | (unsigned long long)blkno, le16_to_cpu(xh->xh_count), |
4436 | xh_free_start, le16_to_cpu(xh->xh_name_value_len)); | |
01225596 TM |
4437 | |
4438 | /* | |
4439 | * sort all the entries by their offset. | |
4440 | * the largest will be the first, so that we can | |
4441 | * move them to the end one by one. | |
4442 | */ | |
4443 | sort(entries, le16_to_cpu(xh->xh_count), | |
4444 | sizeof(struct ocfs2_xattr_entry), | |
4445 | cmp_xe_offset, swap_xe); | |
4446 | ||
4447 | /* Move all name/values to the end of the bucket. */ | |
4448 | xe = xh->xh_entries; | |
4449 | end = OCFS2_XATTR_BUCKET_SIZE; | |
4450 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) { | |
4451 | offset = le16_to_cpu(xe->xe_name_offset); | |
199799a3 | 4452 | len = namevalue_size_xe(xe); |
01225596 TM |
4453 | |
4454 | /* | |
4455 | * We must make sure that the name/value pair | |
4456 | * exist in the same block. So adjust end to | |
4457 | * the previous block end if needed. | |
4458 | */ | |
4459 | if (((end - len) / blocksize != | |
4460 | (end - 1) / blocksize)) | |
4461 | end = end - end % blocksize; | |
4462 | ||
4463 | if (end > offset + len) { | |
4464 | memmove(bucket_buf + end - len, | |
4465 | bucket_buf + offset, len); | |
4466 | xe->xe_name_offset = cpu_to_le16(end - len); | |
4467 | } | |
4468 | ||
4469 | mlog_bug_on_msg(end < offset + len, "Defrag check failed for " | |
4470 | "bucket %llu\n", (unsigned long long)blkno); | |
4471 | ||
4472 | end -= len; | |
4473 | } | |
4474 | ||
4475 | mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for " | |
4476 | "bucket %llu\n", (unsigned long long)blkno); | |
4477 | ||
4478 | if (xh_free_start == end) | |
85db90e7 | 4479 | goto out; |
01225596 TM |
4480 | |
4481 | memset(bucket_buf + xh_free_start, 0, end - xh_free_start); | |
4482 | xh->xh_free_start = cpu_to_le16(end); | |
4483 | ||
4484 | /* sort the entries by their name_hash. */ | |
4485 | sort(entries, le16_to_cpu(xh->xh_count), | |
4486 | sizeof(struct ocfs2_xattr_entry), | |
4487 | cmp_xe, swap_xe); | |
4488 | ||
4489 | buf = bucket_buf; | |
1c32a2fd TM |
4490 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4491 | memcpy(bucket_block(bucket, i), buf, blocksize); | |
4492 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
01225596 | 4493 | |
01225596 | 4494 | out: |
01225596 TM |
4495 | kfree(bucket_buf); |
4496 | return ret; | |
4497 | } | |
4498 | ||
4499 | /* | |
b5c03e74 JB |
4500 | * prev_blkno points to the start of an existing extent. new_blkno |
4501 | * points to a newly allocated extent. Because we know each of our | |
4502 | * clusters contains more than bucket, we can easily split one cluster | |
4503 | * at a bucket boundary. So we take the last cluster of the existing | |
4504 | * extent and split it down the middle. We move the last half of the | |
4505 | * buckets in the last cluster of the existing extent over to the new | |
4506 | * extent. | |
01225596 | 4507 | * |
b5c03e74 JB |
4508 | * first_bh is the buffer at prev_blkno so we can update the existing |
4509 | * extent's bucket count. header_bh is the bucket were we were hoping | |
4510 | * to insert our xattr. If the bucket move places the target in the new | |
4511 | * extent, we'll update first_bh and header_bh after modifying the old | |
4512 | * extent. | |
4513 | * | |
4514 | * first_hash will be set as the 1st xe's name_hash in the new extent. | |
01225596 TM |
4515 | */ |
4516 | static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode, | |
4517 | handle_t *handle, | |
41cb8148 JB |
4518 | struct ocfs2_xattr_bucket *first, |
4519 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4520 | u64 new_blkno, |
01225596 TM |
4521 | u32 num_clusters, |
4522 | u32 *first_hash) | |
4523 | { | |
c58b6032 | 4524 | int ret; |
41cb8148 JB |
4525 | struct super_block *sb = inode->i_sb; |
4526 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb); | |
4527 | int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
b5c03e74 | 4528 | int to_move = num_buckets / 2; |
c58b6032 | 4529 | u64 src_blkno; |
41cb8148 JB |
4530 | u64 last_cluster_blkno = bucket_blkno(first) + |
4531 | ((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1)); | |
01225596 | 4532 | |
41cb8148 JB |
4533 | BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets); |
4534 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize); | |
01225596 | 4535 | |
402b4183 TM |
4536 | trace_ocfs2_mv_xattr_bucket_cross_cluster( |
4537 | (unsigned long long)last_cluster_blkno, | |
4538 | (unsigned long long)new_blkno); | |
01225596 | 4539 | |
41cb8148 | 4540 | ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first), |
c58b6032 JB |
4541 | last_cluster_blkno, new_blkno, |
4542 | to_move, first_hash); | |
01225596 TM |
4543 | if (ret) { |
4544 | mlog_errno(ret); | |
4545 | goto out; | |
4546 | } | |
4547 | ||
c58b6032 JB |
4548 | /* This is the first bucket that got moved */ |
4549 | src_blkno = last_cluster_blkno + (to_move * blks_per_bucket); | |
01225596 | 4550 | |
b5c03e74 | 4551 | /* |
c58b6032 | 4552 | * If the target bucket was part of the moved buckets, we need to |
41cb8148 | 4553 | * update first and target. |
b5c03e74 | 4554 | */ |
41cb8148 | 4555 | if (bucket_blkno(target) >= src_blkno) { |
b5c03e74 JB |
4556 | /* Find the block for the new target bucket */ |
4557 | src_blkno = new_blkno + | |
41cb8148 JB |
4558 | (bucket_blkno(target) - src_blkno); |
4559 | ||
4560 | ocfs2_xattr_bucket_relse(first); | |
4561 | ocfs2_xattr_bucket_relse(target); | |
01225596 | 4562 | |
b5c03e74 | 4563 | /* |
c58b6032 | 4564 | * These shouldn't fail - the buffers are in the |
b5c03e74 JB |
4565 | * journal from ocfs2_cp_xattr_bucket(). |
4566 | */ | |
41cb8148 | 4567 | ret = ocfs2_read_xattr_bucket(first, new_blkno); |
c58b6032 JB |
4568 | if (ret) { |
4569 | mlog_errno(ret); | |
4570 | goto out; | |
4571 | } | |
41cb8148 JB |
4572 | ret = ocfs2_read_xattr_bucket(target, src_blkno); |
4573 | if (ret) | |
b5c03e74 | 4574 | mlog_errno(ret); |
01225596 | 4575 | |
01225596 TM |
4576 | } |
4577 | ||
01225596 | 4578 | out: |
01225596 TM |
4579 | return ret; |
4580 | } | |
4581 | ||
01225596 | 4582 | /* |
80bcaf34 TM |
4583 | * Find the suitable pos when we divide a bucket into 2. |
4584 | * We have to make sure the xattrs with the same hash value exist | |
4585 | * in the same bucket. | |
4586 | * | |
4587 | * If this ocfs2_xattr_header covers more than one hash value, find a | |
4588 | * place where the hash value changes. Try to find the most even split. | |
4589 | * The most common case is that all entries have different hash values, | |
4590 | * and the first check we make will find a place to split. | |
4591 | */ | |
4592 | static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh) | |
4593 | { | |
4594 | struct ocfs2_xattr_entry *entries = xh->xh_entries; | |
4595 | int count = le16_to_cpu(xh->xh_count); | |
4596 | int delta, middle = count / 2; | |
4597 | ||
4598 | /* | |
4599 | * We start at the middle. Each step gets farther away in both | |
4600 | * directions. We therefore hit the change in hash value | |
4601 | * nearest to the middle. Note that this loop does not execute for | |
4602 | * count < 2. | |
4603 | */ | |
4604 | for (delta = 0; delta < middle; delta++) { | |
4605 | /* Let's check delta earlier than middle */ | |
4606 | if (cmp_xe(&entries[middle - delta - 1], | |
4607 | &entries[middle - delta])) | |
4608 | return middle - delta; | |
4609 | ||
4610 | /* For even counts, don't walk off the end */ | |
4611 | if ((middle + delta + 1) == count) | |
4612 | continue; | |
4613 | ||
4614 | /* Now try delta past middle */ | |
4615 | if (cmp_xe(&entries[middle + delta], | |
4616 | &entries[middle + delta + 1])) | |
4617 | return middle + delta + 1; | |
4618 | } | |
4619 | ||
4620 | /* Every entry had the same hash */ | |
4621 | return count; | |
4622 | } | |
4623 | ||
4624 | /* | |
4625 | * Move some xattrs in old bucket(blk) to new bucket(new_blk). | |
01225596 | 4626 | * first_hash will record the 1st hash of the new bucket. |
80bcaf34 TM |
4627 | * |
4628 | * Normally half of the xattrs will be moved. But we have to make | |
4629 | * sure that the xattrs with the same hash value are stored in the | |
4630 | * same bucket. If all the xattrs in this bucket have the same hash | |
4631 | * value, the new bucket will be initialized as an empty one and the | |
4632 | * first_hash will be initialized as (hash_value+1). | |
01225596 | 4633 | */ |
80bcaf34 TM |
4634 | static int ocfs2_divide_xattr_bucket(struct inode *inode, |
4635 | handle_t *handle, | |
4636 | u64 blk, | |
4637 | u64 new_blk, | |
4638 | u32 *first_hash, | |
4639 | int new_bucket_head) | |
01225596 TM |
4640 | { |
4641 | int ret, i; | |
199799a3 | 4642 | int count, start, len, name_value_len = 0, name_offset = 0; |
ba937127 | 4643 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4644 | struct ocfs2_xattr_header *xh; |
4645 | struct ocfs2_xattr_entry *xe; | |
4646 | int blocksize = inode->i_sb->s_blocksize; | |
4647 | ||
402b4183 TM |
4648 | trace_ocfs2_divide_xattr_bucket_begin((unsigned long long)blk, |
4649 | (unsigned long long)new_blk); | |
01225596 | 4650 | |
ba937127 JB |
4651 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4652 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4653 | if (!s_bucket || !t_bucket) { | |
4654 | ret = -ENOMEM; | |
4655 | mlog_errno(ret); | |
4656 | goto out; | |
4657 | } | |
01225596 | 4658 | |
ba937127 | 4659 | ret = ocfs2_read_xattr_bucket(s_bucket, blk); |
01225596 TM |
4660 | if (ret) { |
4661 | mlog_errno(ret); | |
4662 | goto out; | |
4663 | } | |
4664 | ||
ba937127 | 4665 | ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket, |
1224be02 | 4666 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4667 | if (ret) { |
4668 | mlog_errno(ret); | |
4669 | goto out; | |
4670 | } | |
4671 | ||
784b816a JB |
4672 | /* |
4673 | * Even if !new_bucket_head, we're overwriting t_bucket. Thus, | |
4674 | * there's no need to read it. | |
4675 | */ | |
9c339255 | 4676 | ret = ocfs2_init_xattr_bucket(t_bucket, new_blk, new_bucket_head); |
01225596 TM |
4677 | if (ret) { |
4678 | mlog_errno(ret); | |
4679 | goto out; | |
4680 | } | |
4681 | ||
2b656c1d JB |
4682 | /* |
4683 | * Hey, if we're overwriting t_bucket, what difference does | |
4684 | * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the | |
4685 | * same part of ocfs2_cp_xattr_bucket(). | |
4686 | */ | |
ba937127 | 4687 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4688 | new_bucket_head ? |
4689 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4690 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4691 | if (ret) { | |
4692 | mlog_errno(ret); | |
4693 | goto out; | |
01225596 TM |
4694 | } |
4695 | ||
ba937127 | 4696 | xh = bucket_xh(s_bucket); |
80bcaf34 TM |
4697 | count = le16_to_cpu(xh->xh_count); |
4698 | start = ocfs2_xattr_find_divide_pos(xh); | |
4699 | ||
4700 | if (start == count) { | |
4701 | xe = &xh->xh_entries[start-1]; | |
4702 | ||
4703 | /* | |
4704 | * initialized a new empty bucket here. | |
4705 | * The hash value is set as one larger than | |
4706 | * that of the last entry in the previous bucket. | |
4707 | */ | |
ba937127 JB |
4708 | for (i = 0; i < t_bucket->bu_blocks; i++) |
4709 | memset(bucket_block(t_bucket, i), 0, blocksize); | |
80bcaf34 | 4710 | |
ba937127 | 4711 | xh = bucket_xh(t_bucket); |
80bcaf34 TM |
4712 | xh->xh_free_start = cpu_to_le16(blocksize); |
4713 | xh->xh_entries[0].xe_name_hash = xe->xe_name_hash; | |
4714 | le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1); | |
4715 | ||
4716 | goto set_num_buckets; | |
4717 | } | |
4718 | ||
01225596 | 4719 | /* copy the whole bucket to the new first. */ |
ba937127 | 4720 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
01225596 TM |
4721 | |
4722 | /* update the new bucket. */ | |
ba937127 | 4723 | xh = bucket_xh(t_bucket); |
01225596 TM |
4724 | |
4725 | /* | |
4726 | * Calculate the total name/value len and xh_free_start for | |
4727 | * the old bucket first. | |
4728 | */ | |
4729 | name_offset = OCFS2_XATTR_BUCKET_SIZE; | |
4730 | name_value_len = 0; | |
4731 | for (i = 0; i < start; i++) { | |
4732 | xe = &xh->xh_entries[i]; | |
199799a3 | 4733 | name_value_len += namevalue_size_xe(xe); |
01225596 TM |
4734 | if (le16_to_cpu(xe->xe_name_offset) < name_offset) |
4735 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
4736 | } | |
4737 | ||
4738 | /* | |
4739 | * Now begin the modification to the new bucket. | |
4740 | * | |
4741 | * In the new bucket, We just move the xattr entry to the beginning | |
4742 | * and don't touch the name/value. So there will be some holes in the | |
4743 | * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is | |
4744 | * called. | |
4745 | */ | |
4746 | xe = &xh->xh_entries[start]; | |
4747 | len = sizeof(struct ocfs2_xattr_entry) * (count - start); | |
402b4183 TM |
4748 | trace_ocfs2_divide_xattr_bucket_move(len, |
4749 | (int)((char *)xe - (char *)xh), | |
4750 | (int)((char *)xh->xh_entries - (char *)xh)); | |
01225596 TM |
4751 | memmove((char *)xh->xh_entries, (char *)xe, len); |
4752 | xe = &xh->xh_entries[count - start]; | |
4753 | len = sizeof(struct ocfs2_xattr_entry) * start; | |
4754 | memset((char *)xe, 0, len); | |
4755 | ||
4756 | le16_add_cpu(&xh->xh_count, -start); | |
4757 | le16_add_cpu(&xh->xh_name_value_len, -name_value_len); | |
4758 | ||
4759 | /* Calculate xh_free_start for the new bucket. */ | |
4760 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); | |
4761 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
4762 | xe = &xh->xh_entries[i]; | |
01225596 TM |
4763 | if (le16_to_cpu(xe->xe_name_offset) < |
4764 | le16_to_cpu(xh->xh_free_start)) | |
4765 | xh->xh_free_start = xe->xe_name_offset; | |
4766 | } | |
4767 | ||
80bcaf34 | 4768 | set_num_buckets: |
01225596 TM |
4769 | /* set xh->xh_num_buckets for the new xh. */ |
4770 | if (new_bucket_head) | |
4771 | xh->xh_num_buckets = cpu_to_le16(1); | |
4772 | else | |
4773 | xh->xh_num_buckets = 0; | |
4774 | ||
ba937127 | 4775 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); |
01225596 TM |
4776 | |
4777 | /* store the first_hash of the new bucket. */ | |
4778 | if (first_hash) | |
4779 | *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash); | |
4780 | ||
4781 | /* | |
80bcaf34 TM |
4782 | * Now only update the 1st block of the old bucket. If we |
4783 | * just added a new empty bucket, there is no need to modify | |
4784 | * it. | |
01225596 | 4785 | */ |
80bcaf34 TM |
4786 | if (start == count) |
4787 | goto out; | |
4788 | ||
ba937127 | 4789 | xh = bucket_xh(s_bucket); |
01225596 TM |
4790 | memset(&xh->xh_entries[start], 0, |
4791 | sizeof(struct ocfs2_xattr_entry) * (count - start)); | |
4792 | xh->xh_count = cpu_to_le16(start); | |
4793 | xh->xh_free_start = cpu_to_le16(name_offset); | |
4794 | xh->xh_name_value_len = cpu_to_le16(name_value_len); | |
4795 | ||
ba937127 | 4796 | ocfs2_xattr_bucket_journal_dirty(handle, s_bucket); |
01225596 TM |
4797 | |
4798 | out: | |
ba937127 JB |
4799 | ocfs2_xattr_bucket_free(s_bucket); |
4800 | ocfs2_xattr_bucket_free(t_bucket); | |
01225596 TM |
4801 | |
4802 | return ret; | |
4803 | } | |
4804 | ||
4805 | /* | |
4806 | * Copy xattr from one bucket to another bucket. | |
4807 | * | |
4808 | * The caller must make sure that the journal transaction | |
4809 | * has enough space for journaling. | |
4810 | */ | |
4811 | static int ocfs2_cp_xattr_bucket(struct inode *inode, | |
4812 | handle_t *handle, | |
4813 | u64 s_blkno, | |
4814 | u64 t_blkno, | |
4815 | int t_is_new) | |
4816 | { | |
4980c6da | 4817 | int ret; |
ba937127 | 4818 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4819 | |
4820 | BUG_ON(s_blkno == t_blkno); | |
4821 | ||
402b4183 TM |
4822 | trace_ocfs2_cp_xattr_bucket((unsigned long long)s_blkno, |
4823 | (unsigned long long)t_blkno, | |
4824 | t_is_new); | |
01225596 | 4825 | |
ba937127 JB |
4826 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4827 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4828 | if (!s_bucket || !t_bucket) { | |
4829 | ret = -ENOMEM; | |
4830 | mlog_errno(ret); | |
4831 | goto out; | |
4832 | } | |
92de109a | 4833 | |
ba937127 | 4834 | ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno); |
01225596 TM |
4835 | if (ret) |
4836 | goto out; | |
4837 | ||
784b816a JB |
4838 | /* |
4839 | * Even if !t_is_new, we're overwriting t_bucket. Thus, | |
4840 | * there's no need to read it. | |
4841 | */ | |
9c339255 | 4842 | ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno, t_is_new); |
01225596 TM |
4843 | if (ret) |
4844 | goto out; | |
4845 | ||
2b656c1d JB |
4846 | /* |
4847 | * Hey, if we're overwriting t_bucket, what difference does | |
4848 | * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new | |
874d65af JB |
4849 | * cluster to fill, we came here from |
4850 | * ocfs2_mv_xattr_buckets(), and it is really new - | |
4851 | * ACCESS_CREATE is required. But we also might have moved data | |
4852 | * out of t_bucket before extending back into it. | |
4853 | * ocfs2_add_new_xattr_bucket() can do this - its call to | |
4854 | * ocfs2_add_new_xattr_cluster() may have created a new extent | |
2b656c1d JB |
4855 | * and copied out the end of the old extent. Then it re-extends |
4856 | * the old extent back to create space for new xattrs. That's | |
4857 | * how we get here, and the bucket isn't really new. | |
4858 | */ | |
ba937127 | 4859 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4860 | t_is_new ? |
4861 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4862 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4863 | if (ret) | |
4864 | goto out; | |
01225596 | 4865 | |
ba937127 JB |
4866 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
4867 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); | |
01225596 TM |
4868 | |
4869 | out: | |
ba937127 JB |
4870 | ocfs2_xattr_bucket_free(t_bucket); |
4871 | ocfs2_xattr_bucket_free(s_bucket); | |
01225596 TM |
4872 | |
4873 | return ret; | |
4874 | } | |
4875 | ||
4876 | /* | |
874d65af JB |
4877 | * src_blk points to the start of an existing extent. last_blk points to |
4878 | * last cluster in that extent. to_blk points to a newly allocated | |
54ecb6b6 JB |
4879 | * extent. We copy the buckets from the cluster at last_blk to the new |
4880 | * extent. If start_bucket is non-zero, we skip that many buckets before | |
4881 | * we start copying. The new extent's xh_num_buckets gets set to the | |
4882 | * number of buckets we copied. The old extent's xh_num_buckets shrinks | |
4883 | * by the same amount. | |
01225596 | 4884 | */ |
54ecb6b6 JB |
4885 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
4886 | u64 src_blk, u64 last_blk, u64 to_blk, | |
4887 | unsigned int start_bucket, | |
4888 | u32 *first_hash) | |
01225596 TM |
4889 | { |
4890 | int i, ret, credits; | |
4891 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
15d60929 | 4892 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 | 4893 | int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); |
15d60929 | 4894 | struct ocfs2_xattr_bucket *old_first, *new_first; |
01225596 | 4895 | |
402b4183 TM |
4896 | trace_ocfs2_mv_xattr_buckets((unsigned long long)last_blk, |
4897 | (unsigned long long)to_blk); | |
01225596 | 4898 | |
54ecb6b6 JB |
4899 | BUG_ON(start_bucket >= num_buckets); |
4900 | if (start_bucket) { | |
4901 | num_buckets -= start_bucket; | |
4902 | last_blk += (start_bucket * blks_per_bucket); | |
4903 | } | |
4904 | ||
15d60929 JB |
4905 | /* The first bucket of the original extent */ |
4906 | old_first = ocfs2_xattr_bucket_new(inode); | |
4907 | /* The first bucket of the new extent */ | |
4908 | new_first = ocfs2_xattr_bucket_new(inode); | |
4909 | if (!old_first || !new_first) { | |
4910 | ret = -ENOMEM; | |
4911 | mlog_errno(ret); | |
4912 | goto out; | |
4913 | } | |
4914 | ||
874d65af | 4915 | ret = ocfs2_read_xattr_bucket(old_first, src_blk); |
15d60929 JB |
4916 | if (ret) { |
4917 | mlog_errno(ret); | |
4918 | goto out; | |
4919 | } | |
4920 | ||
01225596 | 4921 | /* |
54ecb6b6 JB |
4922 | * We need to update the first bucket of the old extent and all |
4923 | * the buckets going to the new extent. | |
01225596 | 4924 | */ |
c901fb00 | 4925 | credits = ((num_buckets + 1) * blks_per_bucket); |
01225596 TM |
4926 | ret = ocfs2_extend_trans(handle, credits); |
4927 | if (ret) { | |
4928 | mlog_errno(ret); | |
4929 | goto out; | |
4930 | } | |
4931 | ||
15d60929 JB |
4932 | ret = ocfs2_xattr_bucket_journal_access(handle, old_first, |
4933 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4934 | if (ret) { |
4935 | mlog_errno(ret); | |
4936 | goto out; | |
4937 | } | |
4938 | ||
4939 | for (i = 0; i < num_buckets; i++) { | |
4940 | ret = ocfs2_cp_xattr_bucket(inode, handle, | |
874d65af | 4941 | last_blk + (i * blks_per_bucket), |
15d60929 JB |
4942 | to_blk + (i * blks_per_bucket), |
4943 | 1); | |
01225596 TM |
4944 | if (ret) { |
4945 | mlog_errno(ret); | |
4946 | goto out; | |
4947 | } | |
01225596 TM |
4948 | } |
4949 | ||
15d60929 JB |
4950 | /* |
4951 | * Get the new bucket ready before we dirty anything | |
4952 | * (This actually shouldn't fail, because we already dirtied | |
4953 | * it once in ocfs2_cp_xattr_bucket()). | |
4954 | */ | |
4955 | ret = ocfs2_read_xattr_bucket(new_first, to_blk); | |
4956 | if (ret) { | |
01225596 TM |
4957 | mlog_errno(ret); |
4958 | goto out; | |
4959 | } | |
15d60929 JB |
4960 | ret = ocfs2_xattr_bucket_journal_access(handle, new_first, |
4961 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4962 | if (ret) { |
4963 | mlog_errno(ret); | |
4964 | goto out; | |
4965 | } | |
4966 | ||
15d60929 JB |
4967 | /* Now update the headers */ |
4968 | le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets); | |
4969 | ocfs2_xattr_bucket_journal_dirty(handle, old_first); | |
01225596 | 4970 | |
15d60929 JB |
4971 | bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(num_buckets); |
4972 | ocfs2_xattr_bucket_journal_dirty(handle, new_first); | |
01225596 TM |
4973 | |
4974 | if (first_hash) | |
15d60929 JB |
4975 | *first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash); |
4976 | ||
01225596 | 4977 | out: |
15d60929 JB |
4978 | ocfs2_xattr_bucket_free(new_first); |
4979 | ocfs2_xattr_bucket_free(old_first); | |
01225596 TM |
4980 | return ret; |
4981 | } | |
4982 | ||
4983 | /* | |
80bcaf34 | 4984 | * Move some xattrs in this cluster to the new cluster. |
01225596 TM |
4985 | * This function should only be called when bucket size == cluster size. |
4986 | * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead. | |
4987 | */ | |
80bcaf34 TM |
4988 | static int ocfs2_divide_xattr_cluster(struct inode *inode, |
4989 | handle_t *handle, | |
4990 | u64 prev_blk, | |
4991 | u64 new_blk, | |
4992 | u32 *first_hash) | |
01225596 TM |
4993 | { |
4994 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
c901fb00 | 4995 | int ret, credits = 2 * blk_per_bucket; |
01225596 TM |
4996 | |
4997 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize); | |
4998 | ||
4999 | ret = ocfs2_extend_trans(handle, credits); | |
5000 | if (ret) { | |
5001 | mlog_errno(ret); | |
5002 | return ret; | |
5003 | } | |
5004 | ||
5005 | /* Move half of the xattr in start_blk to the next bucket. */ | |
80bcaf34 TM |
5006 | return ocfs2_divide_xattr_bucket(inode, handle, prev_blk, |
5007 | new_blk, first_hash, 1); | |
01225596 TM |
5008 | } |
5009 | ||
5010 | /* | |
5011 | * Move some xattrs from the old cluster to the new one since they are not | |
5012 | * contiguous in ocfs2 xattr tree. | |
5013 | * | |
5014 | * new_blk starts a new separate cluster, and we will move some xattrs from | |
5015 | * prev_blk to it. v_start will be set as the first name hash value in this | |
5016 | * new cluster so that it can be used as e_cpos during tree insertion and | |
5017 | * don't collide with our original b-tree operations. first_bh and header_bh | |
5018 | * will also be updated since they will be used in ocfs2_extend_xattr_bucket | |
5019 | * to extend the insert bucket. | |
5020 | * | |
5021 | * The problem is how much xattr should we move to the new one and when should | |
5022 | * we update first_bh and header_bh? | |
5023 | * 1. If cluster size > bucket size, that means the previous cluster has more | |
5024 | * than 1 bucket, so just move half nums of bucket into the new cluster and | |
5025 | * update the first_bh and header_bh if the insert bucket has been moved | |
5026 | * to the new cluster. | |
5027 | * 2. If cluster_size == bucket_size: | |
5028 | * a) If the previous extent rec has more than one cluster and the insert | |
5029 | * place isn't in the last cluster, copy the entire last cluster to the | |
5030 | * new one. This time, we don't need to upate the first_bh and header_bh | |
5031 | * since they will not be moved into the new cluster. | |
5032 | * b) Otherwise, move the bottom half of the xattrs in the last cluster into | |
5033 | * the new one. And we set the extend flag to zero if the insert place is | |
5034 | * moved into the new allocated cluster since no extend is needed. | |
5035 | */ | |
5036 | static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode, | |
5037 | handle_t *handle, | |
012ee910 JB |
5038 | struct ocfs2_xattr_bucket *first, |
5039 | struct ocfs2_xattr_bucket *target, | |
01225596 | 5040 | u64 new_blk, |
01225596 TM |
5041 | u32 prev_clusters, |
5042 | u32 *v_start, | |
5043 | int *extend) | |
5044 | { | |
92cf3adf | 5045 | int ret; |
01225596 | 5046 | |
402b4183 TM |
5047 | trace_ocfs2_adjust_xattr_cross_cluster( |
5048 | (unsigned long long)bucket_blkno(first), | |
5049 | (unsigned long long)new_blk, prev_clusters); | |
01225596 | 5050 | |
41cb8148 | 5051 | if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) { |
01225596 TM |
5052 | ret = ocfs2_mv_xattr_bucket_cross_cluster(inode, |
5053 | handle, | |
41cb8148 | 5054 | first, target, |
01225596 | 5055 | new_blk, |
01225596 TM |
5056 | prev_clusters, |
5057 | v_start); | |
012ee910 | 5058 | if (ret) |
41cb8148 | 5059 | mlog_errno(ret); |
41cb8148 | 5060 | } else { |
92cf3adf JB |
5061 | /* The start of the last cluster in the first extent */ |
5062 | u64 last_blk = bucket_blkno(first) + | |
5063 | ((prev_clusters - 1) * | |
5064 | ocfs2_clusters_to_blocks(inode->i_sb, 1)); | |
01225596 | 5065 | |
012ee910 | 5066 | if (prev_clusters > 1 && bucket_blkno(target) != last_blk) { |
874d65af | 5067 | ret = ocfs2_mv_xattr_buckets(inode, handle, |
92cf3adf | 5068 | bucket_blkno(first), |
54ecb6b6 | 5069 | last_blk, new_blk, 0, |
01225596 | 5070 | v_start); |
012ee910 JB |
5071 | if (ret) |
5072 | mlog_errno(ret); | |
5073 | } else { | |
80bcaf34 TM |
5074 | ret = ocfs2_divide_xattr_cluster(inode, handle, |
5075 | last_blk, new_blk, | |
5076 | v_start); | |
012ee910 JB |
5077 | if (ret) |
5078 | mlog_errno(ret); | |
01225596 | 5079 | |
92cf3adf | 5080 | if ((bucket_blkno(target) == last_blk) && extend) |
01225596 TM |
5081 | *extend = 0; |
5082 | } | |
5083 | } | |
5084 | ||
5085 | return ret; | |
5086 | } | |
5087 | ||
5088 | /* | |
5089 | * Add a new cluster for xattr storage. | |
5090 | * | |
5091 | * If the new cluster is contiguous with the previous one, it will be | |
5092 | * appended to the same extent record, and num_clusters will be updated. | |
5093 | * If not, we will insert a new extent for it and move some xattrs in | |
5094 | * the last cluster into the new allocated one. | |
5095 | * We also need to limit the maximum size of a btree leaf, otherwise we'll | |
5096 | * lose the benefits of hashing because we'll have to search large leaves. | |
5097 | * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize, | |
5098 | * if it's bigger). | |
5099 | * | |
5100 | * first_bh is the first block of the previous extent rec and header_bh | |
5101 | * indicates the bucket we will insert the new xattrs. They will be updated | |
5102 | * when the header_bh is moved into the new cluster. | |
5103 | */ | |
5104 | static int ocfs2_add_new_xattr_cluster(struct inode *inode, | |
5105 | struct buffer_head *root_bh, | |
ed29c0ca JB |
5106 | struct ocfs2_xattr_bucket *first, |
5107 | struct ocfs2_xattr_bucket *target, | |
01225596 TM |
5108 | u32 *num_clusters, |
5109 | u32 prev_cpos, | |
78f30c31 TM |
5110 | int *extend, |
5111 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5112 | { |
85db90e7 | 5113 | int ret; |
01225596 TM |
5114 | u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); |
5115 | u32 prev_clusters = *num_clusters; | |
5116 | u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0; | |
5117 | u64 block; | |
85db90e7 | 5118 | handle_t *handle = ctxt->handle; |
01225596 | 5119 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
f99b9b7c | 5120 | struct ocfs2_extent_tree et; |
01225596 | 5121 | |
402b4183 TM |
5122 | trace_ocfs2_add_new_xattr_cluster_begin( |
5123 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
5124 | (unsigned long long)bucket_blkno(first), | |
5125 | prev_cpos, prev_clusters); | |
01225596 | 5126 | |
5e404e9e | 5127 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
f99b9b7c | 5128 | |
0cf2f763 | 5129 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5130 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5131 | if (ret < 0) { |
5132 | mlog_errno(ret); | |
5133 | goto leave; | |
5134 | } | |
5135 | ||
1ed9b777 | 5136 | ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1, |
01225596 TM |
5137 | clusters_to_add, &bit_off, &num_bits); |
5138 | if (ret < 0) { | |
5139 | if (ret != -ENOSPC) | |
5140 | mlog_errno(ret); | |
5141 | goto leave; | |
5142 | } | |
5143 | ||
5144 | BUG_ON(num_bits > clusters_to_add); | |
5145 | ||
5146 | block = ocfs2_clusters_to_blocks(osb->sb, bit_off); | |
402b4183 | 5147 | trace_ocfs2_add_new_xattr_cluster((unsigned long long)block, num_bits); |
01225596 | 5148 | |
ed29c0ca | 5149 | if (bucket_blkno(first) + (prev_clusters * bpc) == block && |
01225596 TM |
5150 | (prev_clusters + num_bits) << osb->s_clustersize_bits <= |
5151 | OCFS2_MAX_XATTR_TREE_LEAF_SIZE) { | |
5152 | /* | |
5153 | * If this cluster is contiguous with the old one and | |
5154 | * adding this new cluster, we don't surpass the limit of | |
5155 | * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be | |
5156 | * initialized and used like other buckets in the previous | |
5157 | * cluster. | |
5158 | * So add it as a contiguous one. The caller will handle | |
5159 | * its init process. | |
5160 | */ | |
5161 | v_start = prev_cpos + prev_clusters; | |
5162 | *num_clusters = prev_clusters + num_bits; | |
01225596 TM |
5163 | } else { |
5164 | ret = ocfs2_adjust_xattr_cross_cluster(inode, | |
5165 | handle, | |
012ee910 JB |
5166 | first, |
5167 | target, | |
01225596 | 5168 | block, |
01225596 TM |
5169 | prev_clusters, |
5170 | &v_start, | |
5171 | extend); | |
5172 | if (ret) { | |
5173 | mlog_errno(ret); | |
5174 | goto leave; | |
5175 | } | |
5176 | } | |
5177 | ||
402b4183 TM |
5178 | trace_ocfs2_add_new_xattr_cluster_insert((unsigned long long)block, |
5179 | v_start, num_bits); | |
cc79d8c1 | 5180 | ret = ocfs2_insert_extent(handle, &et, v_start, block, |
78f30c31 | 5181 | num_bits, 0, ctxt->meta_ac); |
01225596 TM |
5182 | if (ret < 0) { |
5183 | mlog_errno(ret); | |
5184 | goto leave; | |
5185 | } | |
5186 | ||
ec20cec7 | 5187 | ocfs2_journal_dirty(handle, root_bh); |
01225596 TM |
5188 | |
5189 | leave: | |
01225596 TM |
5190 | return ret; |
5191 | } | |
5192 | ||
5193 | /* | |
92de109a JB |
5194 | * We are given an extent. 'first' is the bucket at the very front of |
5195 | * the extent. The extent has space for an additional bucket past | |
5196 | * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number | |
5197 | * of the target bucket. We wish to shift every bucket past the target | |
5198 | * down one, filling in that additional space. When we get back to the | |
5199 | * target, we split the target between itself and the now-empty bucket | |
5200 | * at target+1 (aka, target_blkno + blks_per_bucket). | |
01225596 TM |
5201 | */ |
5202 | static int ocfs2_extend_xattr_bucket(struct inode *inode, | |
85db90e7 | 5203 | handle_t *handle, |
92de109a JB |
5204 | struct ocfs2_xattr_bucket *first, |
5205 | u64 target_blk, | |
01225596 TM |
5206 | u32 num_clusters) |
5207 | { | |
5208 | int ret, credits; | |
5209 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5210 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
92de109a JB |
5211 | u64 end_blk; |
5212 | u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets); | |
01225596 | 5213 | |
402b4183 TM |
5214 | trace_ocfs2_extend_xattr_bucket((unsigned long long)target_blk, |
5215 | (unsigned long long)bucket_blkno(first), | |
5216 | num_clusters, new_bucket); | |
01225596 | 5217 | |
92de109a JB |
5218 | /* The extent must have room for an additional bucket */ |
5219 | BUG_ON(new_bucket >= | |
5220 | (num_clusters * ocfs2_xattr_buckets_per_cluster(osb))); | |
01225596 | 5221 | |
92de109a JB |
5222 | /* end_blk points to the last existing bucket */ |
5223 | end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket); | |
01225596 TM |
5224 | |
5225 | /* | |
92de109a JB |
5226 | * end_blk is the start of the last existing bucket. |
5227 | * Thus, (end_blk - target_blk) covers the target bucket and | |
5228 | * every bucket after it up to, but not including, the last | |
5229 | * existing bucket. Then we add the last existing bucket, the | |
5230 | * new bucket, and the first bucket (3 * blk_per_bucket). | |
01225596 | 5231 | */ |
c901fb00 | 5232 | credits = (end_blk - target_blk) + (3 * blk_per_bucket); |
85db90e7 TM |
5233 | ret = ocfs2_extend_trans(handle, credits); |
5234 | if (ret) { | |
01225596 TM |
5235 | mlog_errno(ret); |
5236 | goto out; | |
5237 | } | |
5238 | ||
92de109a JB |
5239 | ret = ocfs2_xattr_bucket_journal_access(handle, first, |
5240 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
5241 | if (ret) { |
5242 | mlog_errno(ret); | |
85db90e7 | 5243 | goto out; |
01225596 TM |
5244 | } |
5245 | ||
92de109a | 5246 | while (end_blk != target_blk) { |
01225596 TM |
5247 | ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk, |
5248 | end_blk + blk_per_bucket, 0); | |
5249 | if (ret) | |
85db90e7 | 5250 | goto out; |
01225596 TM |
5251 | end_blk -= blk_per_bucket; |
5252 | } | |
5253 | ||
92de109a JB |
5254 | /* Move half of the xattr in target_blkno to the next bucket. */ |
5255 | ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk, | |
5256 | target_blk + blk_per_bucket, NULL, 0); | |
01225596 | 5257 | |
92de109a JB |
5258 | le16_add_cpu(&bucket_xh(first)->xh_num_buckets, 1); |
5259 | ocfs2_xattr_bucket_journal_dirty(handle, first); | |
01225596 | 5260 | |
01225596 TM |
5261 | out: |
5262 | return ret; | |
5263 | } | |
5264 | ||
5265 | /* | |
91f2033f JB |
5266 | * Add new xattr bucket in an extent record and adjust the buckets |
5267 | * accordingly. xb_bh is the ocfs2_xattr_block, and target is the | |
5268 | * bucket we want to insert into. | |
01225596 | 5269 | * |
91f2033f JB |
5270 | * In the easy case, we will move all the buckets after target down by |
5271 | * one. Half of target's xattrs will be moved to the next bucket. | |
5272 | * | |
5273 | * If current cluster is full, we'll allocate a new one. This may not | |
5274 | * be contiguous. The underlying calls will make sure that there is | |
5275 | * space for the insert, shifting buckets around if necessary. | |
5276 | * 'target' may be moved by those calls. | |
01225596 TM |
5277 | */ |
5278 | static int ocfs2_add_new_xattr_bucket(struct inode *inode, | |
5279 | struct buffer_head *xb_bh, | |
91f2033f | 5280 | struct ocfs2_xattr_bucket *target, |
78f30c31 | 5281 | struct ocfs2_xattr_set_ctxt *ctxt) |
01225596 | 5282 | { |
01225596 TM |
5283 | struct ocfs2_xattr_block *xb = |
5284 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
5285 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
5286 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
91f2033f JB |
5287 | u32 name_hash = |
5288 | le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash); | |
ed29c0ca | 5289 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
01225596 TM |
5290 | int ret, num_buckets, extend = 1; |
5291 | u64 p_blkno; | |
5292 | u32 e_cpos, num_clusters; | |
92de109a | 5293 | /* The bucket at the front of the extent */ |
91f2033f | 5294 | struct ocfs2_xattr_bucket *first; |
01225596 | 5295 | |
402b4183 TM |
5296 | trace_ocfs2_add_new_xattr_bucket( |
5297 | (unsigned long long)bucket_blkno(target)); | |
01225596 | 5298 | |
ed29c0ca | 5299 | /* The first bucket of the original extent */ |
92de109a | 5300 | first = ocfs2_xattr_bucket_new(inode); |
91f2033f | 5301 | if (!first) { |
92de109a JB |
5302 | ret = -ENOMEM; |
5303 | mlog_errno(ret); | |
5304 | goto out; | |
5305 | } | |
5306 | ||
01225596 TM |
5307 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos, |
5308 | &num_clusters, el); | |
5309 | if (ret) { | |
5310 | mlog_errno(ret); | |
5311 | goto out; | |
5312 | } | |
5313 | ||
ed29c0ca | 5314 | ret = ocfs2_read_xattr_bucket(first, p_blkno); |
01225596 TM |
5315 | if (ret) { |
5316 | mlog_errno(ret); | |
5317 | goto out; | |
5318 | } | |
5319 | ||
ed29c0ca JB |
5320 | num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters; |
5321 | if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) { | |
5322 | /* | |
5323 | * This can move first+target if the target bucket moves | |
5324 | * to the new extent. | |
5325 | */ | |
01225596 TM |
5326 | ret = ocfs2_add_new_xattr_cluster(inode, |
5327 | xb_bh, | |
ed29c0ca JB |
5328 | first, |
5329 | target, | |
01225596 TM |
5330 | &num_clusters, |
5331 | e_cpos, | |
78f30c31 TM |
5332 | &extend, |
5333 | ctxt); | |
01225596 TM |
5334 | if (ret) { |
5335 | mlog_errno(ret); | |
5336 | goto out; | |
5337 | } | |
5338 | } | |
5339 | ||
92de109a | 5340 | if (extend) { |
01225596 | 5341 | ret = ocfs2_extend_xattr_bucket(inode, |
85db90e7 | 5342 | ctxt->handle, |
ed29c0ca JB |
5343 | first, |
5344 | bucket_blkno(target), | |
01225596 | 5345 | num_clusters); |
92de109a JB |
5346 | if (ret) |
5347 | mlog_errno(ret); | |
5348 | } | |
5349 | ||
01225596 | 5350 | out: |
92de109a | 5351 | ocfs2_xattr_bucket_free(first); |
ed29c0ca | 5352 | |
01225596 TM |
5353 | return ret; |
5354 | } | |
5355 | ||
01225596 TM |
5356 | /* |
5357 | * Truncate the specified xe_off entry in xattr bucket. | |
5358 | * bucket is indicated by header_bh and len is the new length. | |
5359 | * Both the ocfs2_xattr_value_root and the entry will be updated here. | |
5360 | * | |
5361 | * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed. | |
5362 | */ | |
5363 | static int ocfs2_xattr_bucket_value_truncate(struct inode *inode, | |
548b0f22 | 5364 | struct ocfs2_xattr_bucket *bucket, |
01225596 | 5365 | int xe_off, |
78f30c31 TM |
5366 | int len, |
5367 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5368 | { |
5369 | int ret, offset; | |
5370 | u64 value_blk; | |
01225596 | 5371 | struct ocfs2_xattr_entry *xe; |
548b0f22 | 5372 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 5373 | size_t blocksize = inode->i_sb->s_blocksize; |
b3e5d379 JB |
5374 | struct ocfs2_xattr_value_buf vb = { |
5375 | .vb_access = ocfs2_journal_access, | |
5376 | }; | |
01225596 TM |
5377 | |
5378 | xe = &xh->xh_entries[xe_off]; | |
5379 | ||
5380 | BUG_ON(!xe || ocfs2_xattr_is_local(xe)); | |
5381 | ||
5382 | offset = le16_to_cpu(xe->xe_name_offset) + | |
5383 | OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5384 | ||
5385 | value_blk = offset / blocksize; | |
5386 | ||
5387 | /* We don't allow ocfs2_xattr_value to be stored in different block. */ | |
5388 | BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize); | |
01225596 | 5389 | |
b3e5d379 JB |
5390 | vb.vb_bh = bucket->bu_bhs[value_blk]; |
5391 | BUG_ON(!vb.vb_bh); | |
01225596 | 5392 | |
b3e5d379 JB |
5393 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
5394 | (vb.vb_bh->b_data + offset % blocksize); | |
01225596 | 5395 | |
548b0f22 JB |
5396 | /* |
5397 | * From here on out we have to dirty the bucket. The generic | |
5398 | * value calls only modify one of the bucket's bhs, but we need | |
5399 | * to send the bucket at once. So if they error, they *could* have | |
5400 | * modified something. We have to assume they did, and dirty | |
5401 | * the whole bucket. This leaves us in a consistent state. | |
5402 | */ | |
402b4183 TM |
5403 | trace_ocfs2_xattr_bucket_value_truncate( |
5404 | (unsigned long long)bucket_blkno(bucket), xe_off, len); | |
b3e5d379 | 5405 | ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); |
01225596 TM |
5406 | if (ret) { |
5407 | mlog_errno(ret); | |
554e7f9e TM |
5408 | goto out; |
5409 | } | |
5410 | ||
5411 | ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket, | |
5412 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5413 | if (ret) { | |
5414 | mlog_errno(ret); | |
5415 | goto out; | |
01225596 TM |
5416 | } |
5417 | ||
548b0f22 JB |
5418 | xe->xe_value_size = cpu_to_le64(len); |
5419 | ||
548b0f22 JB |
5420 | ocfs2_xattr_bucket_journal_dirty(ctxt->handle, bucket); |
5421 | ||
01225596 | 5422 | out: |
01225596 TM |
5423 | return ret; |
5424 | } | |
5425 | ||
01225596 TM |
5426 | static int ocfs2_rm_xattr_cluster(struct inode *inode, |
5427 | struct buffer_head *root_bh, | |
5428 | u64 blkno, | |
5429 | u32 cpos, | |
47bca495 TM |
5430 | u32 len, |
5431 | void *para) | |
01225596 TM |
5432 | { |
5433 | int ret; | |
5434 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5435 | struct inode *tl_inode = osb->osb_tl_inode; | |
5436 | handle_t *handle; | |
5437 | struct ocfs2_xattr_block *xb = | |
5438 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
01225596 TM |
5439 | struct ocfs2_alloc_context *meta_ac = NULL; |
5440 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
f99b9b7c JB |
5441 | struct ocfs2_extent_tree et; |
5442 | ||
47bca495 | 5443 | ret = ocfs2_iterate_xattr_buckets(inode, blkno, len, |
ce9c5a54 | 5444 | ocfs2_delete_xattr_in_bucket, para); |
47bca495 TM |
5445 | if (ret) { |
5446 | mlog_errno(ret); | |
5447 | return ret; | |
5448 | } | |
5449 | ||
5e404e9e | 5450 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
01225596 TM |
5451 | |
5452 | ocfs2_init_dealloc_ctxt(&dealloc); | |
5453 | ||
402b4183 TM |
5454 | trace_ocfs2_rm_xattr_cluster( |
5455 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
5456 | (unsigned long long)blkno, cpos, len); | |
01225596 | 5457 | |
8cb471e8 JB |
5458 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), blkno, |
5459 | len); | |
01225596 | 5460 | |
f99b9b7c | 5461 | ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac); |
01225596 TM |
5462 | if (ret) { |
5463 | mlog_errno(ret); | |
5464 | return ret; | |
5465 | } | |
5466 | ||
5955102c | 5467 | inode_lock(tl_inode); |
01225596 TM |
5468 | |
5469 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
5470 | ret = __ocfs2_flush_truncate_log(osb); | |
5471 | if (ret < 0) { | |
5472 | mlog_errno(ret); | |
5473 | goto out; | |
5474 | } | |
5475 | } | |
5476 | ||
a90714c1 | 5477 | handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb)); |
d3264799 | 5478 | if (IS_ERR(handle)) { |
01225596 TM |
5479 | ret = -ENOMEM; |
5480 | mlog_errno(ret); | |
5481 | goto out; | |
5482 | } | |
5483 | ||
0cf2f763 | 5484 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5485 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5486 | if (ret) { |
5487 | mlog_errno(ret); | |
5488 | goto out_commit; | |
5489 | } | |
5490 | ||
dbdcf6a4 | 5491 | ret = ocfs2_remove_extent(handle, &et, cpos, len, meta_ac, |
f99b9b7c | 5492 | &dealloc); |
01225596 TM |
5493 | if (ret) { |
5494 | mlog_errno(ret); | |
5495 | goto out_commit; | |
5496 | } | |
5497 | ||
5498 | le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len); | |
ec20cec7 | 5499 | ocfs2_journal_dirty(handle, root_bh); |
01225596 TM |
5500 | |
5501 | ret = ocfs2_truncate_log_append(osb, handle, blkno, len); | |
5502 | if (ret) | |
5503 | mlog_errno(ret); | |
6fdb702d | 5504 | ocfs2_update_inode_fsync_trans(handle, inode, 0); |
01225596 TM |
5505 | |
5506 | out_commit: | |
5507 | ocfs2_commit_trans(osb, handle); | |
5508 | out: | |
5509 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
5510 | ||
5955102c | 5511 | inode_unlock(tl_inode); |
01225596 TM |
5512 | |
5513 | if (meta_ac) | |
5514 | ocfs2_free_alloc_context(meta_ac); | |
5515 | ||
5516 | ocfs2_run_deallocs(osb, &dealloc); | |
5517 | ||
5518 | return ret; | |
5519 | } | |
5520 | ||
80bcaf34 TM |
5521 | /* |
5522 | * check whether the xattr bucket is filled up with the same hash value. | |
5523 | * If we want to insert the xattr with the same hash, return -ENOSPC. | |
5524 | * If we want to insert a xattr with different hash value, go ahead | |
5525 | * and ocfs2_divide_xattr_bucket will handle this. | |
5526 | */ | |
01225596 | 5527 | static int ocfs2_check_xattr_bucket_collision(struct inode *inode, |
80bcaf34 TM |
5528 | struct ocfs2_xattr_bucket *bucket, |
5529 | const char *name) | |
01225596 | 5530 | { |
3e632946 | 5531 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
80bcaf34 TM |
5532 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
5533 | ||
5534 | if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) | |
5535 | return 0; | |
01225596 TM |
5536 | |
5537 | if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == | |
5538 | xh->xh_entries[0].xe_name_hash) { | |
5539 | mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " | |
5540 | "hash = %u\n", | |
9c7759aa | 5541 | (unsigned long long)bucket_blkno(bucket), |
01225596 TM |
5542 | le32_to_cpu(xh->xh_entries[0].xe_name_hash)); |
5543 | return -ENOSPC; | |
5544 | } | |
5545 | ||
5546 | return 0; | |
5547 | } | |
5548 | ||
c5d95df5 JB |
5549 | /* |
5550 | * Try to set the entry in the current bucket. If we fail, the caller | |
5551 | * will handle getting us another bucket. | |
5552 | */ | |
5553 | static int ocfs2_xattr_set_entry_bucket(struct inode *inode, | |
5554 | struct ocfs2_xattr_info *xi, | |
5555 | struct ocfs2_xattr_search *xs, | |
5556 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5557 | { |
c5d95df5 JB |
5558 | int ret; |
5559 | struct ocfs2_xa_loc loc; | |
01225596 | 5560 | |
402b4183 | 5561 | trace_ocfs2_xattr_set_entry_bucket(xi->xi_name); |
01225596 | 5562 | |
c5d95df5 JB |
5563 | ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket, |
5564 | xs->not_found ? NULL : xs->here); | |
5565 | ret = ocfs2_xa_set(&loc, xi, ctxt); | |
5566 | if (!ret) { | |
5567 | xs->here = loc.xl_entry; | |
5568 | goto out; | |
5569 | } | |
5570 | if (ret != -ENOSPC) { | |
5571 | mlog_errno(ret); | |
5572 | goto out; | |
5573 | } | |
01225596 | 5574 | |
c5d95df5 JB |
5575 | /* Ok, we need space. Let's try defragmenting the bucket. */ |
5576 | ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, | |
5577 | xs->bucket); | |
5578 | if (ret) { | |
5579 | mlog_errno(ret); | |
5580 | goto out; | |
5581 | } | |
01225596 | 5582 | |
c5d95df5 JB |
5583 | ret = ocfs2_xa_set(&loc, xi, ctxt); |
5584 | if (!ret) { | |
5585 | xs->here = loc.xl_entry; | |
5586 | goto out; | |
01225596 | 5587 | } |
c5d95df5 JB |
5588 | if (ret != -ENOSPC) |
5589 | mlog_errno(ret); | |
01225596 | 5590 | |
01225596 | 5591 | |
c5d95df5 | 5592 | out: |
c5d95df5 JB |
5593 | return ret; |
5594 | } | |
01225596 | 5595 | |
c5d95df5 JB |
5596 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, |
5597 | struct ocfs2_xattr_info *xi, | |
5598 | struct ocfs2_xattr_search *xs, | |
5599 | struct ocfs2_xattr_set_ctxt *ctxt) | |
5600 | { | |
5601 | int ret; | |
01225596 | 5602 | |
402b4183 | 5603 | trace_ocfs2_xattr_set_entry_index_block(xi->xi_name); |
01225596 | 5604 | |
c5d95df5 JB |
5605 | ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); |
5606 | if (!ret) | |
5607 | goto out; | |
5608 | if (ret != -ENOSPC) { | |
5609 | mlog_errno(ret); | |
5610 | goto out; | |
5611 | } | |
01225596 | 5612 | |
c5d95df5 | 5613 | /* Ack, need more space. Let's try to get another bucket! */ |
01225596 | 5614 | |
c5d95df5 JB |
5615 | /* |
5616 | * We do not allow for overlapping ranges between buckets. And | |
5617 | * the maximum number of collisions we will allow for then is | |
5618 | * one bucket's worth, so check it here whether we need to | |
5619 | * add a new bucket for the insert. | |
5620 | */ | |
5621 | ret = ocfs2_check_xattr_bucket_collision(inode, | |
91f2033f | 5622 | xs->bucket, |
c5d95df5 JB |
5623 | xi->xi_name); |
5624 | if (ret) { | |
5625 | mlog_errno(ret); | |
5626 | goto out; | |
5627 | } | |
01225596 | 5628 | |
c5d95df5 JB |
5629 | ret = ocfs2_add_new_xattr_bucket(inode, |
5630 | xs->xattr_bh, | |
5631 | xs->bucket, | |
5632 | ctxt); | |
5633 | if (ret) { | |
5634 | mlog_errno(ret); | |
5635 | goto out; | |
01225596 TM |
5636 | } |
5637 | ||
c5d95df5 JB |
5638 | /* |
5639 | * ocfs2_add_new_xattr_bucket() will have updated | |
5640 | * xs->bucket if it moved, but it will not have updated | |
5641 | * any of the other search fields. Thus, we drop it and | |
5642 | * re-search. Everything should be cached, so it'll be | |
5643 | * quick. | |
5644 | */ | |
5645 | ocfs2_xattr_bucket_relse(xs->bucket); | |
5646 | ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, | |
5647 | xi->xi_name_index, | |
5648 | xi->xi_name, xs); | |
5649 | if (ret && ret != -ENODATA) | |
5650 | goto out; | |
5651 | xs->not_found = ret; | |
5652 | ||
5653 | /* Ok, we have a new bucket, let's try again */ | |
5654 | ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); | |
5655 | if (ret && (ret != -ENOSPC)) | |
5656 | mlog_errno(ret); | |
5657 | ||
01225596 | 5658 | out: |
01225596 TM |
5659 | return ret; |
5660 | } | |
a3944256 TM |
5661 | |
5662 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
5663 | struct ocfs2_xattr_bucket *bucket, | |
5664 | void *para) | |
5665 | { | |
ce9c5a54 | 5666 | int ret = 0, ref_credits; |
3e632946 | 5667 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
a3944256 TM |
5668 | u16 i; |
5669 | struct ocfs2_xattr_entry *xe; | |
78f30c31 TM |
5670 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
5671 | struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; | |
548b0f22 JB |
5672 | int credits = ocfs2_remove_extent_credits(osb->sb) + |
5673 | ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
ce9c5a54 TM |
5674 | struct ocfs2_xattr_value_root *xv; |
5675 | struct ocfs2_rm_xattr_bucket_para *args = | |
5676 | (struct ocfs2_rm_xattr_bucket_para *)para; | |
78f30c31 TM |
5677 | |
5678 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
a3944256 TM |
5679 | |
5680 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
5681 | xe = &xh->xh_entries[i]; | |
5682 | if (ocfs2_xattr_is_local(xe)) | |
5683 | continue; | |
5684 | ||
ce9c5a54 TM |
5685 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, |
5686 | i, &xv, NULL); | |
023d4ea3 JQ |
5687 | if (ret) { |
5688 | mlog_errno(ret); | |
5689 | break; | |
5690 | } | |
ce9c5a54 TM |
5691 | |
5692 | ret = ocfs2_lock_xattr_remove_allocators(inode, xv, | |
5693 | args->ref_ci, | |
5694 | args->ref_root_bh, | |
5695 | &ctxt.meta_ac, | |
5696 | &ref_credits); | |
5697 | ||
5698 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); | |
88c3b062 TM |
5699 | if (IS_ERR(ctxt.handle)) { |
5700 | ret = PTR_ERR(ctxt.handle); | |
5701 | mlog_errno(ret); | |
5702 | break; | |
5703 | } | |
5704 | ||
548b0f22 | 5705 | ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, |
78f30c31 | 5706 | i, 0, &ctxt); |
88c3b062 TM |
5707 | |
5708 | ocfs2_commit_trans(osb, ctxt.handle); | |
ce9c5a54 TM |
5709 | if (ctxt.meta_ac) { |
5710 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
5711 | ctxt.meta_ac = NULL; | |
5712 | } | |
a3944256 TM |
5713 | if (ret) { |
5714 | mlog_errno(ret); | |
5715 | break; | |
5716 | } | |
5717 | } | |
5718 | ||
ce9c5a54 TM |
5719 | if (ctxt.meta_ac) |
5720 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
5721 | ocfs2_schedule_truncate_log_flush(osb, 1); |
5722 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
a3944256 TM |
5723 | return ret; |
5724 | } | |
5725 | ||
492a8a33 TM |
5726 | /* |
5727 | * Whenever we modify a xattr value root in the bucket(e.g, CoW | |
5728 | * or change the extent record flag), we need to recalculate | |
5729 | * the metaecc for the whole bucket. So it is done here. | |
5730 | * | |
5731 | * Note: | |
5732 | * We have to give the extra credits for the caller. | |
5733 | */ | |
5734 | static int ocfs2_xattr_bucket_post_refcount(struct inode *inode, | |
5735 | handle_t *handle, | |
5736 | void *para) | |
5737 | { | |
5738 | int ret; | |
5739 | struct ocfs2_xattr_bucket *bucket = | |
5740 | (struct ocfs2_xattr_bucket *)para; | |
5741 | ||
5742 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, | |
5743 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5744 | if (ret) { | |
5745 | mlog_errno(ret); | |
5746 | return ret; | |
5747 | } | |
5748 | ||
5749 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
5750 | ||
5751 | return 0; | |
5752 | } | |
5753 | ||
5754 | /* | |
5755 | * Special action we need if the xattr value is refcounted. | |
5756 | * | |
5757 | * 1. If the xattr is refcounted, lock the tree. | |
5758 | * 2. CoW the xattr if we are setting the new value and the value | |
5759 | * will be stored outside. | |
5760 | * 3. In other case, decrease_refcount will work for us, so just | |
5761 | * lock the refcount tree, calculate the meta and credits is OK. | |
5762 | * | |
5763 | * We have to do CoW before ocfs2_init_xattr_set_ctxt since | |
5764 | * currently CoW is a completed transaction, while this function | |
5765 | * will also lock the allocators and let us deadlock. So we will | |
5766 | * CoW the whole xattr value. | |
5767 | */ | |
5768 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, | |
5769 | struct ocfs2_dinode *di, | |
5770 | struct ocfs2_xattr_info *xi, | |
5771 | struct ocfs2_xattr_search *xis, | |
5772 | struct ocfs2_xattr_search *xbs, | |
5773 | struct ocfs2_refcount_tree **ref_tree, | |
5774 | int *meta_add, | |
5775 | int *credits) | |
5776 | { | |
5777 | int ret = 0; | |
5778 | struct ocfs2_xattr_block *xb; | |
5779 | struct ocfs2_xattr_entry *xe; | |
5780 | char *base; | |
5781 | u32 p_cluster, num_clusters; | |
5782 | unsigned int ext_flags; | |
5783 | int name_offset, name_len; | |
5784 | struct ocfs2_xattr_value_buf vb; | |
5785 | struct ocfs2_xattr_bucket *bucket = NULL; | |
5786 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5787 | struct ocfs2_post_refcount refcount; | |
5788 | struct ocfs2_post_refcount *p = NULL; | |
5789 | struct buffer_head *ref_root_bh = NULL; | |
5790 | ||
5791 | if (!xis->not_found) { | |
5792 | xe = xis->here; | |
5793 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5794 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5795 | base = xis->base; | |
5796 | vb.vb_bh = xis->inode_bh; | |
5797 | vb.vb_access = ocfs2_journal_access_di; | |
5798 | } else { | |
5799 | int i, block_off = 0; | |
5800 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
5801 | xe = xbs->here; | |
5802 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5803 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5804 | i = xbs->here - xbs->header->xh_entries; | |
5805 | ||
5806 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 5807 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
492a8a33 TM |
5808 | bucket_xh(xbs->bucket), |
5809 | i, &block_off, | |
5810 | &name_offset); | |
5811 | if (ret) { | |
5812 | mlog_errno(ret); | |
5813 | goto out; | |
5814 | } | |
5815 | base = bucket_block(xbs->bucket, block_off); | |
5816 | vb.vb_bh = xbs->bucket->bu_bhs[block_off]; | |
5817 | vb.vb_access = ocfs2_journal_access; | |
5818 | ||
5819 | if (ocfs2_meta_ecc(osb)) { | |
5820 | /*create parameters for ocfs2_post_refcount. */ | |
5821 | bucket = xbs->bucket; | |
5822 | refcount.credits = bucket->bu_blocks; | |
5823 | refcount.para = bucket; | |
5824 | refcount.func = | |
5825 | ocfs2_xattr_bucket_post_refcount; | |
5826 | p = &refcount; | |
5827 | } | |
5828 | } else { | |
5829 | base = xbs->base; | |
5830 | vb.vb_bh = xbs->xattr_bh; | |
5831 | vb.vb_access = ocfs2_journal_access_xb; | |
5832 | } | |
5833 | } | |
5834 | ||
5835 | if (ocfs2_xattr_is_local(xe)) | |
5836 | goto out; | |
5837 | ||
5838 | vb.vb_xv = (struct ocfs2_xattr_value_root *) | |
5839 | (base + name_offset + name_len); | |
5840 | ||
5841 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
5842 | &num_clusters, &vb.vb_xv->xr_list, | |
5843 | &ext_flags); | |
5844 | if (ret) { | |
5845 | mlog_errno(ret); | |
5846 | goto out; | |
5847 | } | |
5848 | ||
5849 | /* | |
5850 | * We just need to check the 1st extent record, since we always | |
5851 | * CoW the whole xattr. So there shouldn't be a xattr with | |
5852 | * some REFCOUNT extent recs after the 1st one. | |
5853 | */ | |
5854 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5855 | goto out; | |
5856 | ||
5857 | ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), | |
5858 | 1, ref_tree, &ref_root_bh); | |
5859 | if (ret) { | |
5860 | mlog_errno(ret); | |
5861 | goto out; | |
5862 | } | |
5863 | ||
5864 | /* | |
5865 | * If we are deleting the xattr or the new size will be stored inside, | |
5866 | * cool, leave it there, the xattr truncate process will remove them | |
5867 | * for us(it still needs the refcount tree lock and the meta, credits). | |
5868 | * And the worse case is that every cluster truncate will split the | |
5869 | * refcount tree, and make the original extent become 3. So we will need | |
5870 | * 2 * cluster more extent recs at most. | |
5871 | */ | |
6b240ff6 | 5872 | if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) { |
492a8a33 TM |
5873 | |
5874 | ret = ocfs2_refcounted_xattr_delete_need(inode, | |
5875 | &(*ref_tree)->rf_ci, | |
5876 | ref_root_bh, vb.vb_xv, | |
5877 | meta_add, credits); | |
5878 | if (ret) | |
5879 | mlog_errno(ret); | |
5880 | goto out; | |
5881 | } | |
5882 | ||
5883 | ret = ocfs2_refcount_cow_xattr(inode, di, &vb, | |
5884 | *ref_tree, ref_root_bh, 0, | |
5885 | le32_to_cpu(vb.vb_xv->xr_clusters), p); | |
5886 | if (ret) | |
5887 | mlog_errno(ret); | |
5888 | ||
5889 | out: | |
5890 | brelse(ref_root_bh); | |
5891 | return ret; | |
5892 | } | |
5893 | ||
0129241e TM |
5894 | /* |
5895 | * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. | |
5896 | * The physical clusters will be added to refcount tree. | |
5897 | */ | |
5898 | static int ocfs2_xattr_value_attach_refcount(struct inode *inode, | |
5899 | struct ocfs2_xattr_value_root *xv, | |
5900 | struct ocfs2_extent_tree *value_et, | |
5901 | struct ocfs2_caching_info *ref_ci, | |
5902 | struct buffer_head *ref_root_bh, | |
5903 | struct ocfs2_cached_dealloc_ctxt *dealloc, | |
5904 | struct ocfs2_post_refcount *refcount) | |
5905 | { | |
5906 | int ret = 0; | |
5907 | u32 clusters = le32_to_cpu(xv->xr_clusters); | |
5908 | u32 cpos, p_cluster, num_clusters; | |
5909 | struct ocfs2_extent_list *el = &xv->xr_list; | |
5910 | unsigned int ext_flags; | |
5911 | ||
5912 | cpos = 0; | |
5913 | while (cpos < clusters) { | |
5914 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
5915 | &num_clusters, el, &ext_flags); | |
17caf955 JQ |
5916 | if (ret) { |
5917 | mlog_errno(ret); | |
5918 | break; | |
5919 | } | |
0129241e TM |
5920 | |
5921 | cpos += num_clusters; | |
5922 | if ((ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5923 | continue; | |
5924 | ||
5925 | BUG_ON(!p_cluster); | |
5926 | ||
5927 | ret = ocfs2_add_refcount_flag(inode, value_et, | |
5928 | ref_ci, ref_root_bh, | |
5929 | cpos - num_clusters, | |
5930 | p_cluster, num_clusters, | |
5931 | dealloc, refcount); | |
5932 | if (ret) { | |
5933 | mlog_errno(ret); | |
5934 | break; | |
5935 | } | |
5936 | } | |
5937 | ||
5938 | return ret; | |
5939 | } | |
5940 | ||
5941 | /* | |
5942 | * Given a normal ocfs2_xattr_header, refcount all the entries which | |
5943 | * have value stored outside. | |
5944 | * Used for xattrs stored in inode and ocfs2_xattr_block. | |
5945 | */ | |
5946 | static int ocfs2_xattr_attach_refcount_normal(struct inode *inode, | |
5947 | struct ocfs2_xattr_value_buf *vb, | |
5948 | struct ocfs2_xattr_header *header, | |
5949 | struct ocfs2_caching_info *ref_ci, | |
5950 | struct buffer_head *ref_root_bh, | |
5951 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
5952 | { | |
5953 | ||
5954 | struct ocfs2_xattr_entry *xe; | |
5955 | struct ocfs2_xattr_value_root *xv; | |
5956 | struct ocfs2_extent_tree et; | |
5957 | int i, ret = 0; | |
5958 | ||
5959 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
5960 | xe = &header->xh_entries[i]; | |
5961 | ||
5962 | if (ocfs2_xattr_is_local(xe)) | |
5963 | continue; | |
5964 | ||
5965 | xv = (struct ocfs2_xattr_value_root *)((void *)header + | |
5966 | le16_to_cpu(xe->xe_name_offset) + | |
5967 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
5968 | ||
5969 | vb->vb_xv = xv; | |
5970 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); | |
5971 | ||
5972 | ret = ocfs2_xattr_value_attach_refcount(inode, xv, &et, | |
5973 | ref_ci, ref_root_bh, | |
5974 | dealloc, NULL); | |
5975 | if (ret) { | |
5976 | mlog_errno(ret); | |
5977 | break; | |
5978 | } | |
5979 | } | |
5980 | ||
5981 | return ret; | |
5982 | } | |
5983 | ||
5984 | static int ocfs2_xattr_inline_attach_refcount(struct inode *inode, | |
5985 | struct buffer_head *fe_bh, | |
5986 | struct ocfs2_caching_info *ref_ci, | |
5987 | struct buffer_head *ref_root_bh, | |
5988 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
5989 | { | |
5990 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
5991 | struct ocfs2_xattr_header *header = (struct ocfs2_xattr_header *) | |
5992 | (fe_bh->b_data + inode->i_sb->s_blocksize - | |
5993 | le16_to_cpu(di->i_xattr_inline_size)); | |
5994 | struct ocfs2_xattr_value_buf vb = { | |
5995 | .vb_bh = fe_bh, | |
5996 | .vb_access = ocfs2_journal_access_di, | |
5997 | }; | |
5998 | ||
5999 | return ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6000 | ref_ci, ref_root_bh, dealloc); | |
6001 | } | |
6002 | ||
6003 | struct ocfs2_xattr_tree_value_refcount_para { | |
6004 | struct ocfs2_caching_info *ref_ci; | |
6005 | struct buffer_head *ref_root_bh; | |
6006 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
6007 | }; | |
6008 | ||
6009 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, | |
6010 | struct ocfs2_xattr_bucket *bucket, | |
6011 | int offset, | |
6012 | struct ocfs2_xattr_value_root **xv, | |
6013 | struct buffer_head **bh) | |
6014 | { | |
6015 | int ret, block_off, name_offset; | |
6016 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); | |
6017 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6018 | void *base; | |
6019 | ||
6020 | ret = ocfs2_xattr_bucket_get_name_value(sb, | |
6021 | bucket_xh(bucket), | |
6022 | offset, | |
6023 | &block_off, | |
6024 | &name_offset); | |
6025 | if (ret) { | |
6026 | mlog_errno(ret); | |
6027 | goto out; | |
6028 | } | |
6029 | ||
6030 | base = bucket_block(bucket, block_off); | |
6031 | ||
6032 | *xv = (struct ocfs2_xattr_value_root *)(base + name_offset + | |
6033 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6034 | ||
6035 | if (bh) | |
6036 | *bh = bucket->bu_bhs[block_off]; | |
6037 | out: | |
6038 | return ret; | |
6039 | } | |
6040 | ||
6041 | /* | |
6042 | * For a given xattr bucket, refcount all the entries which | |
6043 | * have value stored outside. | |
6044 | */ | |
6045 | static int ocfs2_xattr_bucket_value_refcount(struct inode *inode, | |
6046 | struct ocfs2_xattr_bucket *bucket, | |
6047 | void *para) | |
6048 | { | |
6049 | int i, ret = 0; | |
6050 | struct ocfs2_extent_tree et; | |
6051 | struct ocfs2_xattr_tree_value_refcount_para *ref = | |
6052 | (struct ocfs2_xattr_tree_value_refcount_para *)para; | |
6053 | struct ocfs2_xattr_header *xh = | |
6054 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6055 | struct ocfs2_xattr_entry *xe; | |
6056 | struct ocfs2_xattr_value_buf vb = { | |
6057 | .vb_access = ocfs2_journal_access, | |
6058 | }; | |
6059 | struct ocfs2_post_refcount refcount = { | |
6060 | .credits = bucket->bu_blocks, | |
6061 | .para = bucket, | |
6062 | .func = ocfs2_xattr_bucket_post_refcount, | |
6063 | }; | |
6064 | struct ocfs2_post_refcount *p = NULL; | |
6065 | ||
6066 | /* We only need post_refcount if we support metaecc. */ | |
6067 | if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb))) | |
6068 | p = &refcount; | |
6069 | ||
402b4183 TM |
6070 | trace_ocfs2_xattr_bucket_value_refcount( |
6071 | (unsigned long long)bucket_blkno(bucket), | |
6072 | le16_to_cpu(xh->xh_count)); | |
0129241e TM |
6073 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { |
6074 | xe = &xh->xh_entries[i]; | |
6075 | ||
6076 | if (ocfs2_xattr_is_local(xe)) | |
6077 | continue; | |
6078 | ||
6079 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i, | |
6080 | &vb.vb_xv, &vb.vb_bh); | |
6081 | if (ret) { | |
6082 | mlog_errno(ret); | |
6083 | break; | |
6084 | } | |
6085 | ||
6086 | ocfs2_init_xattr_value_extent_tree(&et, | |
6087 | INODE_CACHE(inode), &vb); | |
6088 | ||
6089 | ret = ocfs2_xattr_value_attach_refcount(inode, vb.vb_xv, | |
6090 | &et, ref->ref_ci, | |
6091 | ref->ref_root_bh, | |
6092 | ref->dealloc, p); | |
6093 | if (ret) { | |
6094 | mlog_errno(ret); | |
6095 | break; | |
6096 | } | |
6097 | } | |
6098 | ||
6099 | return ret; | |
6100 | ||
6101 | } | |
6102 | ||
6103 | static int ocfs2_refcount_xattr_tree_rec(struct inode *inode, | |
6104 | struct buffer_head *root_bh, | |
6105 | u64 blkno, u32 cpos, u32 len, void *para) | |
6106 | { | |
6107 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
6108 | ocfs2_xattr_bucket_value_refcount, | |
6109 | para); | |
6110 | } | |
6111 | ||
6112 | static int ocfs2_xattr_block_attach_refcount(struct inode *inode, | |
6113 | struct buffer_head *blk_bh, | |
6114 | struct ocfs2_caching_info *ref_ci, | |
6115 | struct buffer_head *ref_root_bh, | |
6116 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6117 | { | |
6118 | int ret = 0; | |
6119 | struct ocfs2_xattr_block *xb = | |
6120 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6121 | ||
6122 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
6123 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
6124 | struct ocfs2_xattr_value_buf vb = { | |
6125 | .vb_bh = blk_bh, | |
6126 | .vb_access = ocfs2_journal_access_xb, | |
6127 | }; | |
6128 | ||
6129 | ret = ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6130 | ref_ci, ref_root_bh, | |
6131 | dealloc); | |
6132 | } else { | |
6133 | struct ocfs2_xattr_tree_value_refcount_para para = { | |
6134 | .ref_ci = ref_ci, | |
6135 | .ref_root_bh = ref_root_bh, | |
6136 | .dealloc = dealloc, | |
6137 | }; | |
6138 | ||
6139 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
6140 | ocfs2_refcount_xattr_tree_rec, | |
6141 | ¶); | |
6142 | } | |
6143 | ||
6144 | return ret; | |
6145 | } | |
6146 | ||
6147 | int ocfs2_xattr_attach_refcount_tree(struct inode *inode, | |
6148 | struct buffer_head *fe_bh, | |
6149 | struct ocfs2_caching_info *ref_ci, | |
6150 | struct buffer_head *ref_root_bh, | |
6151 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6152 | { | |
6153 | int ret = 0; | |
6154 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
6155 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6156 | struct buffer_head *blk_bh = NULL; | |
6157 | ||
6158 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
6159 | ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh, | |
6160 | ref_ci, ref_root_bh, | |
6161 | dealloc); | |
6162 | if (ret) { | |
6163 | mlog_errno(ret); | |
6164 | goto out; | |
6165 | } | |
6166 | } | |
6167 | ||
6168 | if (!di->i_xattr_loc) | |
6169 | goto out; | |
6170 | ||
6171 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), | |
6172 | &blk_bh); | |
6173 | if (ret < 0) { | |
6174 | mlog_errno(ret); | |
6175 | goto out; | |
6176 | } | |
6177 | ||
6178 | ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci, | |
6179 | ref_root_bh, dealloc); | |
6180 | if (ret) | |
6181 | mlog_errno(ret); | |
6182 | ||
6183 | brelse(blk_bh); | |
6184 | out: | |
6185 | ||
6186 | return ret; | |
6187 | } | |
6188 | ||
0fe9b66c | 6189 | typedef int (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); |
2999d12f TM |
6190 | /* |
6191 | * Store the information we need in xattr reflink. | |
6192 | * old_bh and new_bh are inode bh for the old and new inode. | |
6193 | */ | |
6194 | struct ocfs2_xattr_reflink { | |
6195 | struct inode *old_inode; | |
6196 | struct inode *new_inode; | |
6197 | struct buffer_head *old_bh; | |
6198 | struct buffer_head *new_bh; | |
6199 | struct ocfs2_caching_info *ref_ci; | |
6200 | struct buffer_head *ref_root_bh; | |
6201 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
0fe9b66c | 6202 | should_xattr_reflinked *xattr_reflinked; |
2999d12f TM |
6203 | }; |
6204 | ||
6205 | /* | |
6206 | * Given a xattr header and xe offset, | |
6207 | * return the proper xv and the corresponding bh. | |
6208 | * xattr in inode, block and xattr tree have different implementaions. | |
6209 | */ | |
6210 | typedef int (get_xattr_value_root)(struct super_block *sb, | |
6211 | struct buffer_head *bh, | |
6212 | struct ocfs2_xattr_header *xh, | |
6213 | int offset, | |
6214 | struct ocfs2_xattr_value_root **xv, | |
6215 | struct buffer_head **ret_bh, | |
6216 | void *para); | |
6217 | ||
6218 | /* | |
6219 | * Calculate all the xattr value root metadata stored in this xattr header and | |
6220 | * credits we need if we create them from the scratch. | |
6221 | * We use get_xattr_value_root so that all types of xattr container can use it. | |
6222 | */ | |
6223 | static int ocfs2_value_metas_in_xattr_header(struct super_block *sb, | |
6224 | struct buffer_head *bh, | |
6225 | struct ocfs2_xattr_header *xh, | |
6226 | int *metas, int *credits, | |
6227 | int *num_recs, | |
6228 | get_xattr_value_root *func, | |
6229 | void *para) | |
6230 | { | |
6231 | int i, ret = 0; | |
6232 | struct ocfs2_xattr_value_root *xv; | |
6233 | struct ocfs2_xattr_entry *xe; | |
6234 | ||
6235 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6236 | xe = &xh->xh_entries[i]; | |
6237 | if (ocfs2_xattr_is_local(xe)) | |
6238 | continue; | |
6239 | ||
6240 | ret = func(sb, bh, xh, i, &xv, NULL, para); | |
6241 | if (ret) { | |
6242 | mlog_errno(ret); | |
6243 | break; | |
6244 | } | |
6245 | ||
6246 | *metas += le16_to_cpu(xv->xr_list.l_tree_depth) * | |
6247 | le16_to_cpu(xv->xr_list.l_next_free_rec); | |
6248 | ||
6249 | *credits += ocfs2_calc_extend_credits(sb, | |
06f9da6e | 6250 | &def_xv.xv.xr_list); |
2999d12f TM |
6251 | |
6252 | /* | |
6253 | * If the value is a tree with depth > 1, We don't go deep | |
6254 | * to the extent block, so just calculate a maximum record num. | |
6255 | */ | |
6256 | if (!xv->xr_list.l_tree_depth) | |
8ff6af88 | 6257 | *num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); |
2999d12f TM |
6258 | else |
6259 | *num_recs += ocfs2_clusters_for_bytes(sb, | |
6260 | XATTR_SIZE_MAX); | |
6261 | } | |
6262 | ||
6263 | return ret; | |
6264 | } | |
6265 | ||
6266 | /* Used by xattr inode and block to return the right xv and buffer_head. */ | |
6267 | static int ocfs2_get_xattr_value_root(struct super_block *sb, | |
6268 | struct buffer_head *bh, | |
6269 | struct ocfs2_xattr_header *xh, | |
6270 | int offset, | |
6271 | struct ocfs2_xattr_value_root **xv, | |
6272 | struct buffer_head **ret_bh, | |
6273 | void *para) | |
6274 | { | |
6275 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6276 | ||
6277 | *xv = (struct ocfs2_xattr_value_root *)((void *)xh + | |
6278 | le16_to_cpu(xe->xe_name_offset) + | |
6279 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6280 | ||
6281 | if (ret_bh) | |
6282 | *ret_bh = bh; | |
6283 | ||
6284 | return 0; | |
6285 | } | |
6286 | ||
6287 | /* | |
6288 | * Lock the meta_ac and caculate how much credits we need for reflink xattrs. | |
6289 | * It is only used for inline xattr and xattr block. | |
6290 | */ | |
6291 | static int ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, | |
6292 | struct ocfs2_xattr_header *xh, | |
6293 | struct buffer_head *ref_root_bh, | |
6294 | int *credits, | |
6295 | struct ocfs2_alloc_context **meta_ac) | |
6296 | { | |
6297 | int ret, meta_add = 0, num_recs = 0; | |
6298 | struct ocfs2_refcount_block *rb = | |
6299 | (struct ocfs2_refcount_block *)ref_root_bh->b_data; | |
6300 | ||
6301 | *credits = 0; | |
6302 | ||
6303 | ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh, | |
6304 | &meta_add, credits, &num_recs, | |
6305 | ocfs2_get_xattr_value_root, | |
6306 | NULL); | |
6307 | if (ret) { | |
6308 | mlog_errno(ret); | |
6309 | goto out; | |
6310 | } | |
6311 | ||
6312 | /* | |
6313 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6314 | * an approximate number we need for refcount tree change. | |
6315 | * Sometimes we need to split the tree, and after split, half recs | |
6316 | * will be moved to the new block, and a new block can only provide | |
6317 | * half number of recs. So we multiple new blocks by 2. | |
6318 | */ | |
6319 | num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6320 | meta_add += num_recs; | |
6321 | *credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6322 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6323 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6324 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6325 | else | |
6326 | *credits += 1; | |
6327 | ||
6328 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); | |
6329 | if (ret) | |
6330 | mlog_errno(ret); | |
6331 | ||
6332 | out: | |
6333 | return ret; | |
6334 | } | |
6335 | ||
6336 | /* | |
6337 | * Given a xattr header, reflink all the xattrs in this container. | |
6338 | * It can be used for inode, block and bucket. | |
6339 | * | |
6340 | * NOTE: | |
6341 | * Before we call this function, the caller has memcpy the xattr in | |
6342 | * old_xh to the new_xh. | |
0fe9b66c TM |
6343 | * |
6344 | * If args.xattr_reflinked is set, call it to decide whether the xe should | |
6345 | * be reflinked or not. If not, remove it from the new xattr header. | |
2999d12f TM |
6346 | */ |
6347 | static int ocfs2_reflink_xattr_header(handle_t *handle, | |
6348 | struct ocfs2_xattr_reflink *args, | |
6349 | struct buffer_head *old_bh, | |
6350 | struct ocfs2_xattr_header *xh, | |
6351 | struct buffer_head *new_bh, | |
6352 | struct ocfs2_xattr_header *new_xh, | |
6353 | struct ocfs2_xattr_value_buf *vb, | |
6354 | struct ocfs2_alloc_context *meta_ac, | |
6355 | get_xattr_value_root *func, | |
6356 | void *para) | |
6357 | { | |
0fe9b66c | 6358 | int ret = 0, i, j; |
2999d12f TM |
6359 | struct super_block *sb = args->old_inode->i_sb; |
6360 | struct buffer_head *value_bh; | |
0fe9b66c | 6361 | struct ocfs2_xattr_entry *xe, *last; |
2999d12f TM |
6362 | struct ocfs2_xattr_value_root *xv, *new_xv; |
6363 | struct ocfs2_extent_tree data_et; | |
6364 | u32 clusters, cpos, p_cluster, num_clusters; | |
6365 | unsigned int ext_flags = 0; | |
6366 | ||
402b4183 TM |
6367 | trace_ocfs2_reflink_xattr_header((unsigned long long)old_bh->b_blocknr, |
6368 | le16_to_cpu(xh->xh_count)); | |
0fe9b66c TM |
6369 | |
6370 | last = &new_xh->xh_entries[le16_to_cpu(new_xh->xh_count)]; | |
6371 | for (i = 0, j = 0; i < le16_to_cpu(xh->xh_count); i++, j++) { | |
2999d12f TM |
6372 | xe = &xh->xh_entries[i]; |
6373 | ||
0fe9b66c TM |
6374 | if (args->xattr_reflinked && !args->xattr_reflinked(xe)) { |
6375 | xe = &new_xh->xh_entries[j]; | |
6376 | ||
6377 | le16_add_cpu(&new_xh->xh_count, -1); | |
6378 | if (new_xh->xh_count) { | |
6379 | memmove(xe, xe + 1, | |
6380 | (void *)last - (void *)xe); | |
6381 | memset(last, 0, | |
6382 | sizeof(struct ocfs2_xattr_entry)); | |
6383 | } | |
6384 | ||
6385 | /* | |
6386 | * We don't want j to increase in the next round since | |
6387 | * it is already moved ahead. | |
6388 | */ | |
6389 | j--; | |
6390 | continue; | |
6391 | } | |
6392 | ||
2999d12f TM |
6393 | if (ocfs2_xattr_is_local(xe)) |
6394 | continue; | |
6395 | ||
6396 | ret = func(sb, old_bh, xh, i, &xv, NULL, para); | |
6397 | if (ret) { | |
6398 | mlog_errno(ret); | |
6399 | break; | |
6400 | } | |
6401 | ||
0fe9b66c | 6402 | ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); |
2999d12f TM |
6403 | if (ret) { |
6404 | mlog_errno(ret); | |
6405 | break; | |
6406 | } | |
6407 | ||
6408 | /* | |
6409 | * For the xattr which has l_tree_depth = 0, all the extent | |
6410 | * recs have already be copied to the new xh with the | |
6411 | * propriate OCFS2_EXT_REFCOUNTED flag we just need to | |
6412 | * increase the refount count int the refcount tree. | |
6413 | * | |
6414 | * For the xattr which has l_tree_depth > 0, we need | |
6415 | * to initialize it to the empty default value root, | |
6416 | * and then insert the extents one by one. | |
6417 | */ | |
6418 | if (xv->xr_list.l_tree_depth) { | |
32ed0bd7 | 6419 | memcpy(new_xv, &def_xv, OCFS2_XATTR_ROOT_SIZE); |
2999d12f TM |
6420 | vb->vb_xv = new_xv; |
6421 | vb->vb_bh = value_bh; | |
6422 | ocfs2_init_xattr_value_extent_tree(&data_et, | |
6423 | INODE_CACHE(args->new_inode), vb); | |
6424 | } | |
6425 | ||
6426 | clusters = le32_to_cpu(xv->xr_clusters); | |
6427 | cpos = 0; | |
6428 | while (cpos < clusters) { | |
6429 | ret = ocfs2_xattr_get_clusters(args->old_inode, | |
6430 | cpos, | |
6431 | &p_cluster, | |
6432 | &num_clusters, | |
6433 | &xv->xr_list, | |
6434 | &ext_flags); | |
6435 | if (ret) { | |
6436 | mlog_errno(ret); | |
6437 | goto out; | |
6438 | } | |
6439 | ||
6440 | BUG_ON(!p_cluster); | |
6441 | ||
6442 | if (xv->xr_list.l_tree_depth) { | |
6443 | ret = ocfs2_insert_extent(handle, | |
6444 | &data_et, cpos, | |
6445 | ocfs2_clusters_to_blocks( | |
6446 | args->old_inode->i_sb, | |
6447 | p_cluster), | |
6448 | num_clusters, ext_flags, | |
6449 | meta_ac); | |
6450 | if (ret) { | |
6451 | mlog_errno(ret); | |
6452 | goto out; | |
6453 | } | |
6454 | } | |
6455 | ||
6456 | ret = ocfs2_increase_refcount(handle, args->ref_ci, | |
6457 | args->ref_root_bh, | |
6458 | p_cluster, num_clusters, | |
6459 | meta_ac, args->dealloc); | |
6460 | if (ret) { | |
6461 | mlog_errno(ret); | |
6462 | goto out; | |
6463 | } | |
6464 | ||
6465 | cpos += num_clusters; | |
6466 | } | |
6467 | } | |
6468 | ||
6469 | out: | |
6470 | return ret; | |
6471 | } | |
6472 | ||
6473 | static int ocfs2_reflink_xattr_inline(struct ocfs2_xattr_reflink *args) | |
6474 | { | |
6475 | int ret = 0, credits = 0; | |
6476 | handle_t *handle; | |
6477 | struct ocfs2_super *osb = OCFS2_SB(args->old_inode->i_sb); | |
6478 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)args->old_bh->b_data; | |
6479 | int inline_size = le16_to_cpu(di->i_xattr_inline_size); | |
6480 | int header_off = osb->sb->s_blocksize - inline_size; | |
6481 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *) | |
6482 | (args->old_bh->b_data + header_off); | |
6483 | struct ocfs2_xattr_header *new_xh = (struct ocfs2_xattr_header *) | |
6484 | (args->new_bh->b_data + header_off); | |
6485 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6486 | struct ocfs2_inode_info *new_oi; | |
6487 | struct ocfs2_dinode *new_di; | |
6488 | struct ocfs2_xattr_value_buf vb = { | |
6489 | .vb_bh = args->new_bh, | |
6490 | .vb_access = ocfs2_journal_access_di, | |
6491 | }; | |
6492 | ||
6493 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6494 | &credits, &meta_ac); | |
6495 | if (ret) { | |
6496 | mlog_errno(ret); | |
6497 | goto out; | |
6498 | } | |
6499 | ||
6500 | handle = ocfs2_start_trans(osb, credits); | |
6501 | if (IS_ERR(handle)) { | |
6502 | ret = PTR_ERR(handle); | |
6503 | mlog_errno(ret); | |
6504 | goto out; | |
6505 | } | |
6506 | ||
6507 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(args->new_inode), | |
6508 | args->new_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6509 | if (ret) { | |
6510 | mlog_errno(ret); | |
6511 | goto out_commit; | |
6512 | } | |
6513 | ||
6514 | memcpy(args->new_bh->b_data + header_off, | |
6515 | args->old_bh->b_data + header_off, inline_size); | |
6516 | ||
6517 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6518 | new_di->i_xattr_inline_size = cpu_to_le16(inline_size); | |
6519 | ||
6520 | ret = ocfs2_reflink_xattr_header(handle, args, args->old_bh, xh, | |
6521 | args->new_bh, new_xh, &vb, meta_ac, | |
6522 | ocfs2_get_xattr_value_root, NULL); | |
6523 | if (ret) { | |
6524 | mlog_errno(ret); | |
6525 | goto out_commit; | |
6526 | } | |
6527 | ||
6528 | new_oi = OCFS2_I(args->new_inode); | |
ef962df0 JB |
6529 | /* |
6530 | * Adjust extent record count to reserve space for extended attribute. | |
6531 | * Inline data count had been adjusted in ocfs2_duplicate_inline_data(). | |
6532 | */ | |
6533 | if (!(new_oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) && | |
6534 | !(ocfs2_inode_is_fast_symlink(args->new_inode))) { | |
6535 | struct ocfs2_extent_list *el = &new_di->id2.i_list; | |
6536 | le16_add_cpu(&el->l_count, -(inline_size / | |
6537 | sizeof(struct ocfs2_extent_rec))); | |
6538 | } | |
2999d12f TM |
6539 | spin_lock(&new_oi->ip_lock); |
6540 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL | OCFS2_INLINE_XATTR_FL; | |
6541 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6542 | spin_unlock(&new_oi->ip_lock); | |
6543 | ||
6544 | ocfs2_journal_dirty(handle, args->new_bh); | |
6545 | ||
6546 | out_commit: | |
6547 | ocfs2_commit_trans(osb, handle); | |
6548 | ||
6549 | out: | |
6550 | if (meta_ac) | |
6551 | ocfs2_free_alloc_context(meta_ac); | |
6552 | return ret; | |
6553 | } | |
6554 | ||
6555 | static int ocfs2_create_empty_xattr_block(struct inode *inode, | |
6556 | struct buffer_head *fe_bh, | |
6557 | struct buffer_head **ret_bh, | |
6558 | int indexed) | |
6559 | { | |
6560 | int ret; | |
2999d12f | 6561 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
b2317968 | 6562 | struct ocfs2_xattr_set_ctxt ctxt; |
2999d12f | 6563 | |
b2317968 TM |
6564 | memset(&ctxt, 0, sizeof(ctxt)); |
6565 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &ctxt.meta_ac); | |
2999d12f TM |
6566 | if (ret < 0) { |
6567 | mlog_errno(ret); | |
6568 | return ret; | |
6569 | } | |
6570 | ||
d3981544 JB |
6571 | ctxt.handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS); |
6572 | if (IS_ERR(ctxt.handle)) { | |
6573 | ret = PTR_ERR(ctxt.handle); | |
2999d12f TM |
6574 | mlog_errno(ret); |
6575 | goto out; | |
6576 | } | |
6577 | ||
402b4183 TM |
6578 | trace_ocfs2_create_empty_xattr_block( |
6579 | (unsigned long long)fe_bh->b_blocknr, indexed); | |
d3981544 JB |
6580 | ret = ocfs2_create_xattr_block(inode, fe_bh, &ctxt, indexed, |
6581 | ret_bh); | |
2999d12f TM |
6582 | if (ret) |
6583 | mlog_errno(ret); | |
6584 | ||
d3981544 | 6585 | ocfs2_commit_trans(osb, ctxt.handle); |
2999d12f | 6586 | out: |
b2317968 | 6587 | ocfs2_free_alloc_context(ctxt.meta_ac); |
2999d12f TM |
6588 | return ret; |
6589 | } | |
6590 | ||
6591 | static int ocfs2_reflink_xattr_block(struct ocfs2_xattr_reflink *args, | |
6592 | struct buffer_head *blk_bh, | |
6593 | struct buffer_head *new_blk_bh) | |
6594 | { | |
6595 | int ret = 0, credits = 0; | |
6596 | handle_t *handle; | |
6597 | struct ocfs2_inode_info *new_oi = OCFS2_I(args->new_inode); | |
6598 | struct ocfs2_dinode *new_di; | |
6599 | struct ocfs2_super *osb = OCFS2_SB(args->new_inode->i_sb); | |
6600 | int header_off = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
6601 | struct ocfs2_xattr_block *xb = | |
6602 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6603 | struct ocfs2_xattr_header *xh = &xb->xb_attrs.xb_header; | |
6604 | struct ocfs2_xattr_block *new_xb = | |
6605 | (struct ocfs2_xattr_block *)new_blk_bh->b_data; | |
6606 | struct ocfs2_xattr_header *new_xh = &new_xb->xb_attrs.xb_header; | |
6607 | struct ocfs2_alloc_context *meta_ac; | |
6608 | struct ocfs2_xattr_value_buf vb = { | |
6609 | .vb_bh = new_blk_bh, | |
6610 | .vb_access = ocfs2_journal_access_xb, | |
6611 | }; | |
6612 | ||
6613 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6614 | &credits, &meta_ac); | |
6615 | if (ret) { | |
6616 | mlog_errno(ret); | |
6617 | return ret; | |
6618 | } | |
6619 | ||
6620 | /* One more credits in case we need to add xattr flags in new inode. */ | |
6621 | handle = ocfs2_start_trans(osb, credits + 1); | |
6622 | if (IS_ERR(handle)) { | |
6623 | ret = PTR_ERR(handle); | |
6624 | mlog_errno(ret); | |
6625 | goto out; | |
6626 | } | |
6627 | ||
6628 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6629 | ret = ocfs2_journal_access_di(handle, | |
6630 | INODE_CACHE(args->new_inode), | |
6631 | args->new_bh, | |
6632 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6633 | if (ret) { | |
6634 | mlog_errno(ret); | |
6635 | goto out_commit; | |
6636 | } | |
6637 | } | |
6638 | ||
6639 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode), | |
6640 | new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6641 | if (ret) { | |
6642 | mlog_errno(ret); | |
6643 | goto out_commit; | |
6644 | } | |
6645 | ||
6646 | memcpy(new_blk_bh->b_data + header_off, blk_bh->b_data + header_off, | |
6647 | osb->sb->s_blocksize - header_off); | |
6648 | ||
6649 | ret = ocfs2_reflink_xattr_header(handle, args, blk_bh, xh, | |
6650 | new_blk_bh, new_xh, &vb, meta_ac, | |
6651 | ocfs2_get_xattr_value_root, NULL); | |
6652 | if (ret) { | |
6653 | mlog_errno(ret); | |
6654 | goto out_commit; | |
6655 | } | |
6656 | ||
6657 | ocfs2_journal_dirty(handle, new_blk_bh); | |
6658 | ||
6659 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6660 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6661 | spin_lock(&new_oi->ip_lock); | |
6662 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
6663 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6664 | spin_unlock(&new_oi->ip_lock); | |
6665 | ||
6666 | ocfs2_journal_dirty(handle, args->new_bh); | |
6667 | } | |
6668 | ||
6669 | out_commit: | |
6670 | ocfs2_commit_trans(osb, handle); | |
6671 | ||
6672 | out: | |
6673 | ocfs2_free_alloc_context(meta_ac); | |
6674 | return ret; | |
6675 | } | |
6676 | ||
6677 | struct ocfs2_reflink_xattr_tree_args { | |
6678 | struct ocfs2_xattr_reflink *reflink; | |
6679 | struct buffer_head *old_blk_bh; | |
6680 | struct buffer_head *new_blk_bh; | |
6681 | struct ocfs2_xattr_bucket *old_bucket; | |
6682 | struct ocfs2_xattr_bucket *new_bucket; | |
6683 | }; | |
6684 | ||
6685 | /* | |
6686 | * NOTE: | |
6687 | * We have to handle the case that both old bucket and new bucket | |
6688 | * will call this function to get the right ret_bh. | |
6689 | * So The caller must give us the right bh. | |
6690 | */ | |
6691 | static int ocfs2_get_reflink_xattr_value_root(struct super_block *sb, | |
6692 | struct buffer_head *bh, | |
6693 | struct ocfs2_xattr_header *xh, | |
6694 | int offset, | |
6695 | struct ocfs2_xattr_value_root **xv, | |
6696 | struct buffer_head **ret_bh, | |
6697 | void *para) | |
6698 | { | |
6699 | struct ocfs2_reflink_xattr_tree_args *args = | |
6700 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6701 | struct ocfs2_xattr_bucket *bucket; | |
6702 | ||
6703 | if (bh == args->old_bucket->bu_bhs[0]) | |
6704 | bucket = args->old_bucket; | |
6705 | else | |
6706 | bucket = args->new_bucket; | |
6707 | ||
6708 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6709 | xv, ret_bh); | |
6710 | } | |
6711 | ||
6712 | struct ocfs2_value_tree_metas { | |
6713 | int num_metas; | |
6714 | int credits; | |
6715 | int num_recs; | |
6716 | }; | |
6717 | ||
6718 | static int ocfs2_value_tree_metas_in_bucket(struct super_block *sb, | |
6719 | struct buffer_head *bh, | |
6720 | struct ocfs2_xattr_header *xh, | |
6721 | int offset, | |
6722 | struct ocfs2_xattr_value_root **xv, | |
6723 | struct buffer_head **ret_bh, | |
6724 | void *para) | |
6725 | { | |
6726 | struct ocfs2_xattr_bucket *bucket = | |
6727 | (struct ocfs2_xattr_bucket *)para; | |
6728 | ||
6729 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6730 | xv, ret_bh); | |
6731 | } | |
6732 | ||
6733 | static int ocfs2_calc_value_tree_metas(struct inode *inode, | |
6734 | struct ocfs2_xattr_bucket *bucket, | |
6735 | void *para) | |
6736 | { | |
6737 | struct ocfs2_value_tree_metas *metas = | |
6738 | (struct ocfs2_value_tree_metas *)para; | |
6739 | struct ocfs2_xattr_header *xh = | |
6740 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6741 | ||
6742 | /* Add the credits for this bucket first. */ | |
6743 | metas->credits += bucket->bu_blocks; | |
6744 | return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0], | |
6745 | xh, &metas->num_metas, | |
6746 | &metas->credits, &metas->num_recs, | |
6747 | ocfs2_value_tree_metas_in_bucket, | |
6748 | bucket); | |
6749 | } | |
6750 | ||
6751 | /* | |
6752 | * Given a xattr extent rec starting from blkno and having len clusters, | |
6753 | * iterate all the buckets calculate how much metadata we need for reflinking | |
6754 | * all the ocfs2_xattr_value_root and lock the allocators accordingly. | |
6755 | */ | |
6756 | static int ocfs2_lock_reflink_xattr_rec_allocators( | |
6757 | struct ocfs2_reflink_xattr_tree_args *args, | |
6758 | struct ocfs2_extent_tree *xt_et, | |
6759 | u64 blkno, u32 len, int *credits, | |
6760 | struct ocfs2_alloc_context **meta_ac, | |
6761 | struct ocfs2_alloc_context **data_ac) | |
6762 | { | |
6763 | int ret, num_free_extents; | |
6764 | struct ocfs2_value_tree_metas metas; | |
6765 | struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); | |
6766 | struct ocfs2_refcount_block *rb; | |
6767 | ||
6768 | memset(&metas, 0, sizeof(metas)); | |
6769 | ||
6770 | ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len, | |
6771 | ocfs2_calc_value_tree_metas, &metas); | |
6772 | if (ret) { | |
6773 | mlog_errno(ret); | |
6774 | goto out; | |
6775 | } | |
6776 | ||
6777 | *credits = metas.credits; | |
6778 | ||
6779 | /* | |
6780 | * Calculate we need for refcount tree change. | |
6781 | * | |
6782 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6783 | * an approximate number we need for refcount tree change. | |
6784 | * Sometimes we need to split the tree, and after split, half recs | |
6785 | * will be moved to the new block, and a new block can only provide | |
6786 | * half number of recs. So we multiple new blocks by 2. | |
6787 | * In the end, we have to add credits for modifying the already | |
6788 | * existed refcount block. | |
6789 | */ | |
6790 | rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data; | |
6791 | metas.num_recs = | |
6792 | (metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) / | |
6793 | ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6794 | metas.num_metas += metas.num_recs; | |
6795 | *credits += metas.num_recs + | |
6796 | metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6797 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6798 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6799 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6800 | else | |
6801 | *credits += 1; | |
6802 | ||
6803 | /* count in the xattr tree change. */ | |
964f14a0 | 6804 | num_free_extents = ocfs2_num_free_extents(xt_et); |
2999d12f TM |
6805 | if (num_free_extents < 0) { |
6806 | ret = num_free_extents; | |
6807 | mlog_errno(ret); | |
6808 | goto out; | |
6809 | } | |
6810 | ||
6811 | if (num_free_extents < len) | |
6812 | metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el); | |
6813 | ||
6814 | *credits += ocfs2_calc_extend_credits(osb->sb, | |
06f9da6e | 6815 | xt_et->et_root_el); |
2999d12f TM |
6816 | |
6817 | if (metas.num_metas) { | |
6818 | ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas, | |
6819 | meta_ac); | |
6820 | if (ret) { | |
6821 | mlog_errno(ret); | |
6822 | goto out; | |
6823 | } | |
6824 | } | |
6825 | ||
6826 | if (len) { | |
6827 | ret = ocfs2_reserve_clusters(osb, len, data_ac); | |
6828 | if (ret) | |
6829 | mlog_errno(ret); | |
6830 | } | |
6831 | out: | |
6832 | if (ret) { | |
6833 | if (*meta_ac) { | |
6834 | ocfs2_free_alloc_context(*meta_ac); | |
6cae6d31 | 6835 | *meta_ac = NULL; |
2999d12f TM |
6836 | } |
6837 | } | |
6838 | ||
6839 | return ret; | |
6840 | } | |
6841 | ||
121a39bb | 6842 | static int ocfs2_reflink_xattr_bucket(handle_t *handle, |
2999d12f | 6843 | u64 blkno, u64 new_blkno, u32 clusters, |
121a39bb | 6844 | u32 *cpos, int num_buckets, |
2999d12f TM |
6845 | struct ocfs2_alloc_context *meta_ac, |
6846 | struct ocfs2_alloc_context *data_ac, | |
6847 | struct ocfs2_reflink_xattr_tree_args *args) | |
6848 | { | |
6849 | int i, j, ret = 0; | |
6850 | struct super_block *sb = args->reflink->old_inode->i_sb; | |
2999d12f TM |
6851 | int bpb = args->old_bucket->bu_blocks; |
6852 | struct ocfs2_xattr_value_buf vb = { | |
6853 | .vb_access = ocfs2_journal_access, | |
6854 | }; | |
6855 | ||
6856 | for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) { | |
6857 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
6858 | if (ret) { | |
6859 | mlog_errno(ret); | |
6860 | break; | |
6861 | } | |
6862 | ||
9c339255 | 6863 | ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno, 1); |
2999d12f TM |
6864 | if (ret) { |
6865 | mlog_errno(ret); | |
6866 | break; | |
6867 | } | |
6868 | ||
2999d12f TM |
6869 | ret = ocfs2_xattr_bucket_journal_access(handle, |
6870 | args->new_bucket, | |
6871 | OCFS2_JOURNAL_ACCESS_CREATE); | |
6872 | if (ret) { | |
6873 | mlog_errno(ret); | |
6874 | break; | |
6875 | } | |
6876 | ||
6877 | for (j = 0; j < bpb; j++) | |
6878 | memcpy(bucket_block(args->new_bucket, j), | |
6879 | bucket_block(args->old_bucket, j), | |
6880 | sb->s_blocksize); | |
6881 | ||
121a39bb TM |
6882 | /* |
6883 | * Record the start cpos so that we can use it to initialize | |
6884 | * our xattr tree we also set the xh_num_bucket for the new | |
6885 | * bucket. | |
6886 | */ | |
6887 | if (i == 0) { | |
6888 | *cpos = le32_to_cpu(bucket_xh(args->new_bucket)-> | |
6889 | xh_entries[0].xe_name_hash); | |
6890 | bucket_xh(args->new_bucket)->xh_num_buckets = | |
6891 | cpu_to_le16(num_buckets); | |
6892 | } | |
6893 | ||
2999d12f TM |
6894 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); |
6895 | ||
6896 | ret = ocfs2_reflink_xattr_header(handle, args->reflink, | |
6897 | args->old_bucket->bu_bhs[0], | |
6898 | bucket_xh(args->old_bucket), | |
6899 | args->new_bucket->bu_bhs[0], | |
6900 | bucket_xh(args->new_bucket), | |
6901 | &vb, meta_ac, | |
6902 | ocfs2_get_reflink_xattr_value_root, | |
6903 | args); | |
6904 | if (ret) { | |
6905 | mlog_errno(ret); | |
6906 | break; | |
6907 | } | |
6908 | ||
6909 | /* | |
6910 | * Re-access and dirty the bucket to calculate metaecc. | |
6911 | * Because we may extend the transaction in reflink_xattr_header | |
6912 | * which will let the already accessed block gone. | |
6913 | */ | |
6914 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
6915 | args->new_bucket, | |
6916 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6917 | if (ret) { | |
6918 | mlog_errno(ret); | |
6919 | break; | |
6920 | } | |
6921 | ||
6922 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
121a39bb | 6923 | |
2999d12f TM |
6924 | ocfs2_xattr_bucket_relse(args->old_bucket); |
6925 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6926 | } | |
6927 | ||
6928 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6929 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6930 | return ret; | |
6931 | } | |
121a39bb TM |
6932 | |
6933 | static int ocfs2_reflink_xattr_buckets(handle_t *handle, | |
6934 | struct inode *inode, | |
6935 | struct ocfs2_reflink_xattr_tree_args *args, | |
6936 | struct ocfs2_extent_tree *et, | |
6937 | struct ocfs2_alloc_context *meta_ac, | |
6938 | struct ocfs2_alloc_context *data_ac, | |
6939 | u64 blkno, u32 cpos, u32 len) | |
6940 | { | |
6941 | int ret, first_inserted = 0; | |
6942 | u32 p_cluster, num_clusters, reflink_cpos = 0; | |
6943 | u64 new_blkno; | |
6944 | unsigned int num_buckets, reflink_buckets; | |
6945 | unsigned int bpc = | |
6946 | ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); | |
6947 | ||
6948 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
6949 | if (ret) { | |
6950 | mlog_errno(ret); | |
6951 | goto out; | |
6952 | } | |
6953 | num_buckets = le16_to_cpu(bucket_xh(args->old_bucket)->xh_num_buckets); | |
6954 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6955 | ||
6956 | while (len && num_buckets) { | |
6957 | ret = ocfs2_claim_clusters(handle, data_ac, | |
6958 | 1, &p_cluster, &num_clusters); | |
6959 | if (ret) { | |
6960 | mlog_errno(ret); | |
6961 | goto out; | |
6962 | } | |
6963 | ||
6964 | new_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
6965 | reflink_buckets = min(num_buckets, bpc * num_clusters); | |
6966 | ||
6967 | ret = ocfs2_reflink_xattr_bucket(handle, blkno, | |
6968 | new_blkno, num_clusters, | |
6969 | &reflink_cpos, reflink_buckets, | |
6970 | meta_ac, data_ac, args); | |
6971 | if (ret) { | |
6972 | mlog_errno(ret); | |
6973 | goto out; | |
6974 | } | |
6975 | ||
6976 | /* | |
6977 | * For the 1st allocated cluster, we make it use the same cpos | |
6978 | * so that the xattr tree looks the same as the original one | |
6979 | * in the most case. | |
6980 | */ | |
6981 | if (!first_inserted) { | |
6982 | reflink_cpos = cpos; | |
6983 | first_inserted = 1; | |
6984 | } | |
6985 | ret = ocfs2_insert_extent(handle, et, reflink_cpos, new_blkno, | |
6986 | num_clusters, 0, meta_ac); | |
6987 | if (ret) | |
6988 | mlog_errno(ret); | |
6989 | ||
402b4183 TM |
6990 | trace_ocfs2_reflink_xattr_buckets((unsigned long long)new_blkno, |
6991 | num_clusters, reflink_cpos); | |
121a39bb TM |
6992 | |
6993 | len -= num_clusters; | |
6994 | blkno += ocfs2_clusters_to_blocks(inode->i_sb, num_clusters); | |
6995 | num_buckets -= reflink_buckets; | |
6996 | } | |
6997 | out: | |
6998 | return ret; | |
6999 | } | |
7000 | ||
2999d12f TM |
7001 | /* |
7002 | * Create the same xattr extent record in the new inode's xattr tree. | |
7003 | */ | |
7004 | static int ocfs2_reflink_xattr_rec(struct inode *inode, | |
7005 | struct buffer_head *root_bh, | |
7006 | u64 blkno, | |
7007 | u32 cpos, | |
7008 | u32 len, | |
7009 | void *para) | |
7010 | { | |
7011 | int ret, credits = 0; | |
2999d12f TM |
7012 | handle_t *handle; |
7013 | struct ocfs2_reflink_xattr_tree_args *args = | |
7014 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
7015 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
7016 | struct ocfs2_alloc_context *meta_ac = NULL; | |
7017 | struct ocfs2_alloc_context *data_ac = NULL; | |
7018 | struct ocfs2_extent_tree et; | |
7019 | ||
402b4183 | 7020 | trace_ocfs2_reflink_xattr_rec((unsigned long long)blkno, len); |
121a39bb | 7021 | |
2999d12f TM |
7022 | ocfs2_init_xattr_tree_extent_tree(&et, |
7023 | INODE_CACHE(args->reflink->new_inode), | |
7024 | args->new_blk_bh); | |
7025 | ||
7026 | ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno, | |
7027 | len, &credits, | |
7028 | &meta_ac, &data_ac); | |
7029 | if (ret) { | |
7030 | mlog_errno(ret); | |
7031 | goto out; | |
7032 | } | |
7033 | ||
7034 | handle = ocfs2_start_trans(osb, credits); | |
7035 | if (IS_ERR(handle)) { | |
7036 | ret = PTR_ERR(handle); | |
7037 | mlog_errno(ret); | |
7038 | goto out; | |
7039 | } | |
7040 | ||
121a39bb TM |
7041 | ret = ocfs2_reflink_xattr_buckets(handle, inode, args, &et, |
7042 | meta_ac, data_ac, | |
7043 | blkno, cpos, len); | |
2999d12f TM |
7044 | if (ret) |
7045 | mlog_errno(ret); | |
7046 | ||
2999d12f TM |
7047 | ocfs2_commit_trans(osb, handle); |
7048 | ||
7049 | out: | |
7050 | if (meta_ac) | |
7051 | ocfs2_free_alloc_context(meta_ac); | |
7052 | if (data_ac) | |
7053 | ocfs2_free_alloc_context(data_ac); | |
7054 | return ret; | |
7055 | } | |
7056 | ||
7057 | /* | |
7058 | * Create reflinked xattr buckets. | |
7059 | * We will add bucket one by one, and refcount all the xattrs in the bucket | |
7060 | * if they are stored outside. | |
7061 | */ | |
7062 | static int ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, | |
7063 | struct buffer_head *blk_bh, | |
7064 | struct buffer_head *new_blk_bh) | |
7065 | { | |
7066 | int ret; | |
7067 | struct ocfs2_reflink_xattr_tree_args para; | |
7068 | ||
7069 | memset(¶, 0, sizeof(para)); | |
7070 | para.reflink = args; | |
7071 | para.old_blk_bh = blk_bh; | |
7072 | para.new_blk_bh = new_blk_bh; | |
7073 | ||
7074 | para.old_bucket = ocfs2_xattr_bucket_new(args->old_inode); | |
7075 | if (!para.old_bucket) { | |
7076 | mlog_errno(-ENOMEM); | |
7077 | return -ENOMEM; | |
7078 | } | |
7079 | ||
7080 | para.new_bucket = ocfs2_xattr_bucket_new(args->new_inode); | |
7081 | if (!para.new_bucket) { | |
7082 | ret = -ENOMEM; | |
7083 | mlog_errno(ret); | |
7084 | goto out; | |
7085 | } | |
7086 | ||
7087 | ret = ocfs2_iterate_xattr_index_block(args->old_inode, blk_bh, | |
7088 | ocfs2_reflink_xattr_rec, | |
7089 | ¶); | |
7090 | if (ret) | |
7091 | mlog_errno(ret); | |
7092 | ||
7093 | out: | |
7094 | ocfs2_xattr_bucket_free(para.old_bucket); | |
7095 | ocfs2_xattr_bucket_free(para.new_bucket); | |
7096 | return ret; | |
7097 | } | |
7098 | ||
7099 | static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args, | |
7100 | struct buffer_head *blk_bh) | |
7101 | { | |
7102 | int ret, indexed = 0; | |
7103 | struct buffer_head *new_blk_bh = NULL; | |
7104 | struct ocfs2_xattr_block *xb = | |
7105 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
7106 | ||
7107 | ||
7108 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) | |
7109 | indexed = 1; | |
7110 | ||
7111 | ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh, | |
7112 | &new_blk_bh, indexed); | |
7113 | if (ret) { | |
7114 | mlog_errno(ret); | |
7115 | goto out; | |
7116 | } | |
7117 | ||
2decd65a | 7118 | if (!indexed) |
2999d12f TM |
7119 | ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); |
7120 | else | |
7121 | ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); | |
7122 | if (ret) | |
7123 | mlog_errno(ret); | |
7124 | ||
7125 | out: | |
7126 | brelse(new_blk_bh); | |
7127 | return ret; | |
7128 | } | |
7129 | ||
0fe9b66c TM |
7130 | static int ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe) |
7131 | { | |
7132 | int type = ocfs2_xattr_get_type(xe); | |
7133 | ||
7134 | return type != OCFS2_XATTR_INDEX_SECURITY && | |
7135 | type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS && | |
7136 | type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT; | |
7137 | } | |
7138 | ||
2999d12f TM |
7139 | int ocfs2_reflink_xattrs(struct inode *old_inode, |
7140 | struct buffer_head *old_bh, | |
7141 | struct inode *new_inode, | |
0fe9b66c TM |
7142 | struct buffer_head *new_bh, |
7143 | bool preserve_security) | |
2999d12f TM |
7144 | { |
7145 | int ret; | |
7146 | struct ocfs2_xattr_reflink args; | |
7147 | struct ocfs2_inode_info *oi = OCFS2_I(old_inode); | |
7148 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)old_bh->b_data; | |
7149 | struct buffer_head *blk_bh = NULL; | |
7150 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
7151 | struct ocfs2_refcount_tree *ref_tree; | |
7152 | struct buffer_head *ref_root_bh = NULL; | |
7153 | ||
7154 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7155 | le64_to_cpu(di->i_refcount_loc), | |
7156 | 1, &ref_tree, &ref_root_bh); | |
7157 | if (ret) { | |
7158 | mlog_errno(ret); | |
7159 | goto out; | |
7160 | } | |
7161 | ||
7162 | ocfs2_init_dealloc_ctxt(&dealloc); | |
7163 | ||
7164 | args.old_inode = old_inode; | |
7165 | args.new_inode = new_inode; | |
7166 | args.old_bh = old_bh; | |
7167 | args.new_bh = new_bh; | |
7168 | args.ref_ci = &ref_tree->rf_ci; | |
7169 | args.ref_root_bh = ref_root_bh; | |
7170 | args.dealloc = &dealloc; | |
0fe9b66c TM |
7171 | if (preserve_security) |
7172 | args.xattr_reflinked = NULL; | |
7173 | else | |
7174 | args.xattr_reflinked = ocfs2_reflink_xattr_no_security; | |
2999d12f TM |
7175 | |
7176 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
7177 | ret = ocfs2_reflink_xattr_inline(&args); | |
7178 | if (ret) { | |
7179 | mlog_errno(ret); | |
7180 | goto out_unlock; | |
7181 | } | |
7182 | } | |
7183 | ||
7184 | if (!di->i_xattr_loc) | |
7185 | goto out_unlock; | |
7186 | ||
7187 | ret = ocfs2_read_xattr_block(old_inode, le64_to_cpu(di->i_xattr_loc), | |
7188 | &blk_bh); | |
7189 | if (ret < 0) { | |
7190 | mlog_errno(ret); | |
7191 | goto out_unlock; | |
7192 | } | |
7193 | ||
7194 | ret = ocfs2_reflink_xattr_in_block(&args, blk_bh); | |
7195 | if (ret) | |
7196 | mlog_errno(ret); | |
7197 | ||
7198 | brelse(blk_bh); | |
7199 | ||
7200 | out_unlock: | |
7201 | ocfs2_unlock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7202 | ref_tree, 1); | |
7203 | brelse(ref_root_bh); | |
7204 | ||
7205 | if (ocfs2_dealloc_has_cluster(&dealloc)) { | |
7206 | ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1); | |
7207 | ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc); | |
7208 | } | |
7209 | ||
7210 | out: | |
7211 | return ret; | |
7212 | } | |
7213 | ||
0fe9b66c TM |
7214 | /* |
7215 | * Initialize security and acl for a already created inode. | |
7216 | * Used for reflink a non-preserve-security file. | |
7217 | * | |
7218 | * It uses common api like ocfs2_xattr_set, so the caller | |
7219 | * must not hold any lock expect i_mutex. | |
7220 | */ | |
7221 | int ocfs2_init_security_and_acl(struct inode *dir, | |
2a7dba39 | 7222 | struct inode *inode, |
c25a1e06 | 7223 | const struct qstr *qstr) |
0fe9b66c | 7224 | { |
702e5bc6 | 7225 | int ret = 0; |
c25a1e06 | 7226 | struct buffer_head *dir_bh = NULL; |
0fe9b66c | 7227 | |
9d8f13ba | 7228 | ret = ocfs2_init_security_get(inode, dir, qstr, NULL); |
32918dd9 | 7229 | if (ret) { |
0fe9b66c TM |
7230 | mlog_errno(ret); |
7231 | goto leave; | |
7232 | } | |
7233 | ||
7234 | ret = ocfs2_inode_lock(dir, &dir_bh, 0); | |
7235 | if (ret) { | |
7236 | mlog_errno(ret); | |
7237 | goto leave; | |
7238 | } | |
c25a1e06 JB |
7239 | ret = ocfs2_init_acl(NULL, inode, dir, NULL, dir_bh, NULL, NULL); |
7240 | if (ret) | |
7241 | mlog_errno(ret); | |
0fe9b66c TM |
7242 | |
7243 | ocfs2_inode_unlock(dir, 0); | |
7244 | brelse(dir_bh); | |
7245 | leave: | |
7246 | return ret; | |
7247 | } | |
1046cb11 | 7248 | |
923f7f31 TY |
7249 | /* |
7250 | * 'security' attributes support | |
7251 | */ | |
d9a82a04 | 7252 | static int ocfs2_xattr_security_get(const struct xattr_handler *handler, |
b296821a AV |
7253 | struct dentry *unused, struct inode *inode, |
7254 | const char *name, void *buffer, size_t size) | |
923f7f31 | 7255 | { |
b296821a | 7256 | return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_SECURITY, |
431547b3 | 7257 | name, buffer, size); |
923f7f31 TY |
7258 | } |
7259 | ||
d9a82a04 | 7260 | static int ocfs2_xattr_security_set(const struct xattr_handler *handler, |
59301226 AV |
7261 | struct dentry *unused, struct inode *inode, |
7262 | const char *name, const void *value, | |
7263 | size_t size, int flags) | |
923f7f31 | 7264 | { |
59301226 | 7265 | return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, |
431547b3 | 7266 | name, value, size, flags); |
923f7f31 TY |
7267 | } |
7268 | ||
b519ea6d | 7269 | static int ocfs2_initxattrs(struct inode *inode, const struct xattr *xattr_array, |
9d8f13ba MZ |
7270 | void *fs_info) |
7271 | { | |
7272 | const struct xattr *xattr; | |
7273 | int err = 0; | |
7274 | ||
7275 | for (xattr = xattr_array; xattr->name != NULL; xattr++) { | |
7276 | err = ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, | |
7277 | xattr->name, xattr->value, | |
7278 | xattr->value_len, XATTR_CREATE); | |
7279 | if (err) | |
7280 | break; | |
7281 | } | |
7282 | return err; | |
7283 | } | |
7284 | ||
534eaddd TY |
7285 | int ocfs2_init_security_get(struct inode *inode, |
7286 | struct inode *dir, | |
2a7dba39 | 7287 | const struct qstr *qstr, |
534eaddd TY |
7288 | struct ocfs2_security_xattr_info *si) |
7289 | { | |
38d59ef6 TY |
7290 | /* check whether ocfs2 support feature xattr */ |
7291 | if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) | |
7292 | return -EOPNOTSUPP; | |
9d8f13ba MZ |
7293 | if (si) |
7294 | return security_old_inode_init_security(inode, dir, qstr, | |
7295 | &si->name, &si->value, | |
7296 | &si->value_len); | |
7297 | ||
7298 | return security_inode_init_security(inode, dir, qstr, | |
7299 | &ocfs2_initxattrs, NULL); | |
534eaddd TY |
7300 | } |
7301 | ||
7302 | int ocfs2_init_security_set(handle_t *handle, | |
7303 | struct inode *inode, | |
7304 | struct buffer_head *di_bh, | |
7305 | struct ocfs2_security_xattr_info *si, | |
7306 | struct ocfs2_alloc_context *xattr_ac, | |
7307 | struct ocfs2_alloc_context *data_ac) | |
7308 | { | |
7309 | return ocfs2_xattr_set_handle(handle, inode, di_bh, | |
7310 | OCFS2_XATTR_INDEX_SECURITY, | |
7311 | si->name, si->value, si->value_len, 0, | |
7312 | xattr_ac, data_ac); | |
7313 | } | |
7314 | ||
537d81ca | 7315 | const struct xattr_handler ocfs2_xattr_security_handler = { |
923f7f31 | 7316 | .prefix = XATTR_SECURITY_PREFIX, |
923f7f31 TY |
7317 | .get = ocfs2_xattr_security_get, |
7318 | .set = ocfs2_xattr_security_set, | |
7319 | }; | |
7320 | ||
99219aea MF |
7321 | /* |
7322 | * 'trusted' attributes support | |
7323 | */ | |
d9a82a04 | 7324 | static int ocfs2_xattr_trusted_get(const struct xattr_handler *handler, |
b296821a AV |
7325 | struct dentry *unused, struct inode *inode, |
7326 | const char *name, void *buffer, size_t size) | |
99219aea | 7327 | { |
b296821a | 7328 | return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_TRUSTED, |
431547b3 | 7329 | name, buffer, size); |
99219aea MF |
7330 | } |
7331 | ||
d9a82a04 | 7332 | static int ocfs2_xattr_trusted_set(const struct xattr_handler *handler, |
59301226 AV |
7333 | struct dentry *unused, struct inode *inode, |
7334 | const char *name, const void *value, | |
7335 | size_t size, int flags) | |
99219aea | 7336 | { |
59301226 | 7337 | return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_TRUSTED, |
431547b3 | 7338 | name, value, size, flags); |
99219aea MF |
7339 | } |
7340 | ||
537d81ca | 7341 | const struct xattr_handler ocfs2_xattr_trusted_handler = { |
99219aea | 7342 | .prefix = XATTR_TRUSTED_PREFIX, |
99219aea MF |
7343 | .get = ocfs2_xattr_trusted_get, |
7344 | .set = ocfs2_xattr_trusted_set, | |
7345 | }; | |
7346 | ||
99219aea MF |
7347 | /* |
7348 | * 'user' attributes support | |
7349 | */ | |
d9a82a04 | 7350 | static int ocfs2_xattr_user_get(const struct xattr_handler *handler, |
84d86e6d | 7351 | struct dentry *unused, struct inode *inode, |
b296821a | 7352 | const char *name, void *buffer, size_t size) |
99219aea | 7353 | { |
b296821a | 7354 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
99219aea | 7355 | |
99219aea MF |
7356 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) |
7357 | return -EOPNOTSUPP; | |
b296821a | 7358 | return ocfs2_xattr_get(inode, OCFS2_XATTR_INDEX_USER, name, |
99219aea MF |
7359 | buffer, size); |
7360 | } | |
7361 | ||
d9a82a04 | 7362 | static int ocfs2_xattr_user_set(const struct xattr_handler *handler, |
59301226 AV |
7363 | struct dentry *unused, struct inode *inode, |
7364 | const char *name, const void *value, | |
7365 | size_t size, int flags) | |
99219aea | 7366 | { |
59301226 | 7367 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
99219aea | 7368 | |
99219aea MF |
7369 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) |
7370 | return -EOPNOTSUPP; | |
7371 | ||
59301226 | 7372 | return ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_USER, |
431547b3 | 7373 | name, value, size, flags); |
99219aea MF |
7374 | } |
7375 | ||
537d81ca | 7376 | const struct xattr_handler ocfs2_xattr_user_handler = { |
99219aea | 7377 | .prefix = XATTR_USER_PREFIX, |
99219aea MF |
7378 | .get = ocfs2_xattr_user_get, |
7379 | .set = ocfs2_xattr_user_set, | |
7380 | }; |