]>
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, |
784b816a JB |
372 | u64 xb_blkno) |
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), |
757055ad | 386 | bucket->bu_bhs[i])) |
8cb471e8 | 387 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(bucket->bu_inode), |
757055ad | 388 | bucket->bu_bhs[i]); |
784b816a JB |
389 | } |
390 | ||
391 | if (rc) | |
ba937127 | 392 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
393 | return rc; |
394 | } | |
395 | ||
396 | /* Read the xattr bucket at xb_blkno */ | |
ba937127 | 397 | static int ocfs2_read_xattr_bucket(struct ocfs2_xattr_bucket *bucket, |
784b816a JB |
398 | u64 xb_blkno) |
399 | { | |
ba937127 | 400 | int rc; |
784b816a | 401 | |
8cb471e8 | 402 | rc = ocfs2_read_blocks(INODE_CACHE(bucket->bu_inode), xb_blkno, |
970e4936 JB |
403 | bucket->bu_blocks, bucket->bu_bhs, 0, |
404 | NULL); | |
4d0e214e | 405 | if (!rc) { |
c8b9cf9a | 406 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
407 | rc = ocfs2_validate_meta_ecc_bhs(bucket->bu_inode->i_sb, |
408 | bucket->bu_bhs, | |
409 | bucket->bu_blocks, | |
410 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 411 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
412 | if (rc) |
413 | mlog_errno(rc); | |
414 | } | |
415 | ||
784b816a | 416 | if (rc) |
ba937127 | 417 | ocfs2_xattr_bucket_relse(bucket); |
784b816a JB |
418 | return rc; |
419 | } | |
420 | ||
1224be02 | 421 | static int ocfs2_xattr_bucket_journal_access(handle_t *handle, |
1224be02 JB |
422 | struct ocfs2_xattr_bucket *bucket, |
423 | int type) | |
424 | { | |
425 | int i, rc = 0; | |
1224be02 | 426 | |
ba937127 | 427 | for (i = 0; i < bucket->bu_blocks; i++) { |
0cf2f763 JB |
428 | rc = ocfs2_journal_access(handle, |
429 | INODE_CACHE(bucket->bu_inode), | |
1224be02 JB |
430 | bucket->bu_bhs[i], type); |
431 | if (rc) { | |
432 | mlog_errno(rc); | |
433 | break; | |
434 | } | |
435 | } | |
436 | ||
437 | return rc; | |
438 | } | |
439 | ||
440 | static void ocfs2_xattr_bucket_journal_dirty(handle_t *handle, | |
1224be02 JB |
441 | struct ocfs2_xattr_bucket *bucket) |
442 | { | |
ba937127 | 443 | int i; |
1224be02 | 444 | |
c8b9cf9a | 445 | spin_lock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e JB |
446 | ocfs2_compute_meta_ecc_bhs(bucket->bu_inode->i_sb, |
447 | bucket->bu_bhs, bucket->bu_blocks, | |
448 | &bucket_xh(bucket)->xh_check); | |
c8b9cf9a | 449 | spin_unlock(&OCFS2_SB(bucket->bu_inode->i_sb)->osb_xattr_lock); |
4d0e214e | 450 | |
ba937127 | 451 | for (i = 0; i < bucket->bu_blocks; i++) |
1224be02 JB |
452 | ocfs2_journal_dirty(handle, bucket->bu_bhs[i]); |
453 | } | |
454 | ||
ba937127 | 455 | static void ocfs2_xattr_bucket_copy_data(struct ocfs2_xattr_bucket *dest, |
4980c6da JB |
456 | struct ocfs2_xattr_bucket *src) |
457 | { | |
458 | int i; | |
ba937127 JB |
459 | int blocksize = src->bu_inode->i_sb->s_blocksize; |
460 | ||
461 | BUG_ON(dest->bu_blocks != src->bu_blocks); | |
462 | BUG_ON(dest->bu_inode != src->bu_inode); | |
4980c6da | 463 | |
ba937127 | 464 | for (i = 0; i < src->bu_blocks; i++) { |
4980c6da JB |
465 | memcpy(bucket_block(dest, i), bucket_block(src, i), |
466 | blocksize); | |
467 | } | |
468 | } | |
1224be02 | 469 | |
4ae1d69b JB |
470 | static int ocfs2_validate_xattr_block(struct super_block *sb, |
471 | struct buffer_head *bh) | |
472 | { | |
d6b32bbb | 473 | int rc; |
4ae1d69b JB |
474 | struct ocfs2_xattr_block *xb = |
475 | (struct ocfs2_xattr_block *)bh->b_data; | |
476 | ||
402b4183 | 477 | trace_ocfs2_validate_xattr_block((unsigned long long)bh->b_blocknr); |
4ae1d69b | 478 | |
d6b32bbb JB |
479 | BUG_ON(!buffer_uptodate(bh)); |
480 | ||
481 | /* | |
482 | * If the ecc fails, we return the error but otherwise | |
483 | * leave the filesystem running. We know any error is | |
484 | * local to this block. | |
485 | */ | |
486 | rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &xb->xb_check); | |
487 | if (rc) | |
488 | return rc; | |
489 | ||
490 | /* | |
491 | * Errors after here are fatal | |
492 | */ | |
493 | ||
4ae1d69b JB |
494 | if (!OCFS2_IS_VALID_XATTR_BLOCK(xb)) { |
495 | ocfs2_error(sb, | |
496 | "Extended attribute block #%llu has bad " | |
497 | "signature %.*s", | |
498 | (unsigned long long)bh->b_blocknr, 7, | |
499 | xb->xb_signature); | |
500 | return -EINVAL; | |
501 | } | |
502 | ||
503 | if (le64_to_cpu(xb->xb_blkno) != bh->b_blocknr) { | |
504 | ocfs2_error(sb, | |
505 | "Extended attribute block #%llu has an " | |
506 | "invalid xb_blkno of %llu", | |
507 | (unsigned long long)bh->b_blocknr, | |
508 | (unsigned long long)le64_to_cpu(xb->xb_blkno)); | |
509 | return -EINVAL; | |
510 | } | |
511 | ||
512 | if (le32_to_cpu(xb->xb_fs_generation) != OCFS2_SB(sb)->fs_generation) { | |
513 | ocfs2_error(sb, | |
514 | "Extended attribute block #%llu has an invalid " | |
515 | "xb_fs_generation of #%u", | |
516 | (unsigned long long)bh->b_blocknr, | |
517 | le32_to_cpu(xb->xb_fs_generation)); | |
518 | return -EINVAL; | |
519 | } | |
520 | ||
521 | return 0; | |
522 | } | |
523 | ||
524 | static int ocfs2_read_xattr_block(struct inode *inode, u64 xb_blkno, | |
525 | struct buffer_head **bh) | |
526 | { | |
527 | int rc; | |
528 | struct buffer_head *tmp = *bh; | |
529 | ||
8cb471e8 | 530 | rc = ocfs2_read_block(INODE_CACHE(inode), xb_blkno, &tmp, |
970e4936 | 531 | ocfs2_validate_xattr_block); |
4ae1d69b JB |
532 | |
533 | /* If ocfs2_read_block() got us a new bh, pass it up. */ | |
534 | if (!rc && !*bh) | |
535 | *bh = tmp; | |
536 | ||
537 | return rc; | |
538 | } | |
539 | ||
936b8834 | 540 | static inline const char *ocfs2_xattr_prefix(int name_index) |
cf1d6c76 | 541 | { |
537d81ca | 542 | const struct xattr_handler *handler = NULL; |
cf1d6c76 TY |
543 | |
544 | if (name_index > 0 && name_index < OCFS2_XATTR_MAX) | |
545 | handler = ocfs2_xattr_handler_map[name_index]; | |
546 | ||
936b8834 | 547 | return handler ? handler->prefix : 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) { | |
649 | mlog_errno(ret); | |
650 | return ret; | |
651 | } | |
652 | } | |
653 | ||
654 | if (!(s_size + a_size)) | |
655 | return ret; | |
656 | ||
657 | /* | |
658 | * The max space of security xattr taken inline is | |
659 | * 256(name) + 80(value) + 16(entry) = 352 bytes, | |
660 | * The max space of acl xattr taken inline is | |
661 | * 80(value) + 16(entry) * 2(if directory) = 192 bytes, | |
662 | * when blocksize = 512, may reserve one more cluser for | |
663 | * xattr bucket, otherwise reserve one metadata block | |
664 | * for them is ok. | |
6c9fd1dc TY |
665 | * If this is a new directory with inline data, |
666 | * we choose to reserve the entire inline area for | |
667 | * directory contents and force an external xattr block. | |
89c38bd0 TY |
668 | */ |
669 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE || | |
6c9fd1dc | 670 | (S_ISDIR(mode) && ocfs2_supports_inline_data(osb)) || |
89c38bd0 | 671 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_IBODY) { |
9b7895ef | 672 | *want_meta = *want_meta + 1; |
89c38bd0 TY |
673 | *xattr_credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
674 | } | |
675 | ||
676 | if (dir->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE && | |
677 | (s_size + a_size) > OCFS2_XATTR_FREE_IN_BLOCK(dir)) { | |
678 | *want_clusters += 1; | |
679 | *xattr_credits += ocfs2_blocks_per_xattr_bucket(dir->i_sb); | |
680 | } | |
681 | ||
0e445b6f TY |
682 | /* |
683 | * reserve credits and clusters for xattrs which has large value | |
684 | * and have to be set outside | |
685 | */ | |
686 | if (si->enable && si->value_len > OCFS2_XATTR_INLINE_SIZE) { | |
687 | new_clusters = ocfs2_clusters_for_bytes(dir->i_sb, | |
688 | si->value_len); | |
689 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
690 | new_clusters); | |
691 | *want_clusters += new_clusters; | |
692 | } | |
89c38bd0 TY |
693 | if (osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL && |
694 | acl_len > OCFS2_XATTR_INLINE_SIZE) { | |
0e445b6f TY |
695 | /* for directory, it has DEFAULT and ACCESS two types of acls */ |
696 | new_clusters = (S_ISDIR(mode) ? 2 : 1) * | |
697 | ocfs2_clusters_for_bytes(dir->i_sb, acl_len); | |
698 | *xattr_credits += ocfs2_clusters_to_blocks(dir->i_sb, | |
699 | new_clusters); | |
700 | *want_clusters += new_clusters; | |
89c38bd0 TY |
701 | } |
702 | ||
703 | return ret; | |
704 | } | |
705 | ||
f56654c4 TM |
706 | static int ocfs2_xattr_extend_allocation(struct inode *inode, |
707 | u32 clusters_to_add, | |
19b801f4 | 708 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 709 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 | 710 | { |
a78f9f46 | 711 | int status = 0, credits; |
85db90e7 | 712 | handle_t *handle = ctxt->handle; |
f56654c4 | 713 | enum ocfs2_alloc_restarted why; |
19b801f4 | 714 | u32 prev_clusters, logical_start = le32_to_cpu(vb->vb_xv->xr_clusters); |
f99b9b7c | 715 | struct ocfs2_extent_tree et; |
f56654c4 | 716 | |
5e404e9e | 717 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f99b9b7c | 718 | |
a78f9f46 | 719 | while (clusters_to_add) { |
402b4183 TM |
720 | trace_ocfs2_xattr_extend_allocation(clusters_to_add); |
721 | ||
a78f9f46 TM |
722 | status = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
723 | OCFS2_JOURNAL_ACCESS_WRITE); | |
724 | if (status < 0) { | |
725 | mlog_errno(status); | |
726 | break; | |
727 | } | |
f56654c4 | 728 | |
a78f9f46 TM |
729 | prev_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
730 | status = ocfs2_add_clusters_in_btree(handle, | |
731 | &et, | |
732 | &logical_start, | |
733 | clusters_to_add, | |
734 | 0, | |
735 | ctxt->data_ac, | |
736 | ctxt->meta_ac, | |
737 | &why); | |
738 | if ((status < 0) && (status != -EAGAIN)) { | |
739 | if (status != -ENOSPC) | |
740 | mlog_errno(status); | |
741 | break; | |
742 | } | |
f56654c4 | 743 | |
a78f9f46 TM |
744 | ocfs2_journal_dirty(handle, vb->vb_bh); |
745 | ||
746 | clusters_to_add -= le32_to_cpu(vb->vb_xv->xr_clusters) - | |
747 | prev_clusters; | |
748 | ||
749 | if (why != RESTART_NONE && clusters_to_add) { | |
750 | /* | |
751 | * We can only fail in case the alloc file doesn't give | |
752 | * up enough clusters. | |
753 | */ | |
754 | BUG_ON(why == RESTART_META); | |
755 | ||
a78f9f46 | 756 | credits = ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 757 | &vb->vb_xv->xr_list); |
a78f9f46 TM |
758 | status = ocfs2_extend_trans(handle, credits); |
759 | if (status < 0) { | |
760 | status = -ENOMEM; | |
761 | mlog_errno(status); | |
762 | break; | |
763 | } | |
764 | } | |
765 | } | |
f56654c4 TM |
766 | |
767 | return status; | |
768 | } | |
769 | ||
770 | static int __ocfs2_remove_xattr_range(struct inode *inode, | |
d72cc72d | 771 | struct ocfs2_xattr_value_buf *vb, |
f56654c4 | 772 | u32 cpos, u32 phys_cpos, u32 len, |
492a8a33 | 773 | unsigned int ext_flags, |
78f30c31 | 774 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
775 | { |
776 | int ret; | |
777 | u64 phys_blkno = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
85db90e7 | 778 | handle_t *handle = ctxt->handle; |
f99b9b7c JB |
779 | struct ocfs2_extent_tree et; |
780 | ||
5e404e9e | 781 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); |
f56654c4 | 782 | |
0cf2f763 | 783 | ret = vb->vb_access(handle, INODE_CACHE(inode), vb->vb_bh, |
d72cc72d | 784 | OCFS2_JOURNAL_ACCESS_WRITE); |
f56654c4 TM |
785 | if (ret) { |
786 | mlog_errno(ret); | |
85db90e7 | 787 | goto out; |
f56654c4 TM |
788 | } |
789 | ||
dbdcf6a4 | 790 | ret = ocfs2_remove_extent(handle, &et, cpos, len, ctxt->meta_ac, |
78f30c31 | 791 | &ctxt->dealloc); |
f56654c4 TM |
792 | if (ret) { |
793 | mlog_errno(ret); | |
85db90e7 | 794 | goto out; |
f56654c4 TM |
795 | } |
796 | ||
d72cc72d | 797 | le32_add_cpu(&vb->vb_xv->xr_clusters, -len); |
ec20cec7 | 798 | ocfs2_journal_dirty(handle, vb->vb_bh); |
f56654c4 | 799 | |
492a8a33 TM |
800 | if (ext_flags & OCFS2_EXT_REFCOUNTED) |
801 | ret = ocfs2_decrease_refcount(inode, handle, | |
802 | ocfs2_blocks_to_clusters(inode->i_sb, | |
803 | phys_blkno), | |
804 | len, ctxt->meta_ac, &ctxt->dealloc, 1); | |
805 | else | |
806 | ret = ocfs2_cache_cluster_dealloc(&ctxt->dealloc, | |
807 | phys_blkno, len); | |
f56654c4 TM |
808 | if (ret) |
809 | mlog_errno(ret); | |
810 | ||
f56654c4 | 811 | out: |
f56654c4 TM |
812 | return ret; |
813 | } | |
814 | ||
815 | static int ocfs2_xattr_shrink_size(struct inode *inode, | |
816 | u32 old_clusters, | |
817 | u32 new_clusters, | |
19b801f4 | 818 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 | 819 | struct ocfs2_xattr_set_ctxt *ctxt) |
f56654c4 TM |
820 | { |
821 | int ret = 0; | |
492a8a33 | 822 | unsigned int ext_flags; |
f56654c4 TM |
823 | u32 trunc_len, cpos, phys_cpos, alloc_size; |
824 | u64 block; | |
f56654c4 TM |
825 | |
826 | if (old_clusters <= new_clusters) | |
827 | return 0; | |
828 | ||
829 | cpos = new_clusters; | |
830 | trunc_len = old_clusters - new_clusters; | |
831 | while (trunc_len) { | |
832 | ret = ocfs2_xattr_get_clusters(inode, cpos, &phys_cpos, | |
d72cc72d | 833 | &alloc_size, |
492a8a33 | 834 | &vb->vb_xv->xr_list, &ext_flags); |
f56654c4 TM |
835 | if (ret) { |
836 | mlog_errno(ret); | |
837 | goto out; | |
838 | } | |
839 | ||
840 | if (alloc_size > trunc_len) | |
841 | alloc_size = trunc_len; | |
842 | ||
19b801f4 | 843 | ret = __ocfs2_remove_xattr_range(inode, vb, cpos, |
f56654c4 | 844 | phys_cpos, alloc_size, |
492a8a33 | 845 | ext_flags, ctxt); |
f56654c4 TM |
846 | if (ret) { |
847 | mlog_errno(ret); | |
848 | goto out; | |
849 | } | |
850 | ||
851 | block = ocfs2_clusters_to_blocks(inode->i_sb, phys_cpos); | |
8cb471e8 JB |
852 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), |
853 | block, alloc_size); | |
f56654c4 TM |
854 | cpos += alloc_size; |
855 | trunc_len -= alloc_size; | |
856 | } | |
857 | ||
858 | out: | |
f56654c4 TM |
859 | return ret; |
860 | } | |
861 | ||
862 | static int ocfs2_xattr_value_truncate(struct inode *inode, | |
b3e5d379 | 863 | struct ocfs2_xattr_value_buf *vb, |
78f30c31 TM |
864 | int len, |
865 | struct ocfs2_xattr_set_ctxt *ctxt) | |
f56654c4 TM |
866 | { |
867 | int ret; | |
868 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, len); | |
b3e5d379 | 869 | u32 old_clusters = le32_to_cpu(vb->vb_xv->xr_clusters); |
f56654c4 TM |
870 | |
871 | if (new_clusters == old_clusters) | |
872 | return 0; | |
873 | ||
874 | if (new_clusters > old_clusters) | |
875 | ret = ocfs2_xattr_extend_allocation(inode, | |
876 | new_clusters - old_clusters, | |
b3e5d379 | 877 | vb, ctxt); |
f56654c4 TM |
878 | else |
879 | ret = ocfs2_xattr_shrink_size(inode, | |
880 | old_clusters, new_clusters, | |
b3e5d379 | 881 | vb, ctxt); |
f56654c4 TM |
882 | |
883 | return ret; | |
884 | } | |
cf1d6c76 | 885 | |
936b8834 TM |
886 | static int ocfs2_xattr_list_entry(char *buffer, size_t size, |
887 | size_t *result, const char *prefix, | |
888 | const char *name, int name_len) | |
889 | { | |
890 | char *p = buffer + *result; | |
891 | int prefix_len = strlen(prefix); | |
892 | int total_len = prefix_len + name_len + 1; | |
893 | ||
894 | *result += total_len; | |
895 | ||
896 | /* we are just looking for how big our buffer needs to be */ | |
897 | if (!size) | |
898 | return 0; | |
899 | ||
900 | if (*result > size) | |
901 | return -ERANGE; | |
902 | ||
903 | memcpy(p, prefix, prefix_len); | |
904 | memcpy(p + prefix_len, name, name_len); | |
905 | p[prefix_len + name_len] = '\0'; | |
906 | ||
907 | return 0; | |
908 | } | |
909 | ||
cf1d6c76 TY |
910 | static int ocfs2_xattr_list_entries(struct inode *inode, |
911 | struct ocfs2_xattr_header *header, | |
912 | char *buffer, size_t buffer_size) | |
913 | { | |
936b8834 TM |
914 | size_t result = 0; |
915 | int i, type, ret; | |
916 | const char *prefix, *name; | |
cf1d6c76 TY |
917 | |
918 | for (i = 0 ; i < le16_to_cpu(header->xh_count); i++) { | |
919 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
936b8834 TM |
920 | type = ocfs2_xattr_get_type(entry); |
921 | prefix = ocfs2_xattr_prefix(type); | |
922 | ||
923 | if (prefix) { | |
924 | name = (const char *)header + | |
925 | le16_to_cpu(entry->xe_name_offset); | |
926 | ||
927 | ret = ocfs2_xattr_list_entry(buffer, buffer_size, | |
928 | &result, prefix, name, | |
929 | entry->xe_name_len); | |
930 | if (ret) | |
931 | return ret; | |
cf1d6c76 TY |
932 | } |
933 | } | |
934 | ||
936b8834 | 935 | return result; |
cf1d6c76 TY |
936 | } |
937 | ||
8b2c0dba TM |
938 | int ocfs2_has_inline_xattr_value_outside(struct inode *inode, |
939 | struct ocfs2_dinode *di) | |
940 | { | |
941 | struct ocfs2_xattr_header *xh; | |
942 | int i; | |
943 | ||
944 | xh = (struct ocfs2_xattr_header *) | |
945 | ((void *)di + inode->i_sb->s_blocksize - | |
946 | le16_to_cpu(di->i_xattr_inline_size)); | |
947 | ||
948 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) | |
949 | if (!ocfs2_xattr_is_local(&xh->xh_entries[i])) | |
950 | return 1; | |
951 | ||
952 | return 0; | |
953 | } | |
954 | ||
cf1d6c76 TY |
955 | static int ocfs2_xattr_ibody_list(struct inode *inode, |
956 | struct ocfs2_dinode *di, | |
957 | char *buffer, | |
958 | size_t buffer_size) | |
959 | { | |
960 | struct ocfs2_xattr_header *header = NULL; | |
961 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
962 | int ret = 0; | |
963 | ||
964 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
965 | return ret; | |
966 | ||
967 | header = (struct ocfs2_xattr_header *) | |
968 | ((void *)di + inode->i_sb->s_blocksize - | |
969 | le16_to_cpu(di->i_xattr_inline_size)); | |
970 | ||
971 | ret = ocfs2_xattr_list_entries(inode, header, buffer, buffer_size); | |
972 | ||
973 | return ret; | |
974 | } | |
975 | ||
976 | static int ocfs2_xattr_block_list(struct inode *inode, | |
977 | struct ocfs2_dinode *di, | |
978 | char *buffer, | |
979 | size_t buffer_size) | |
980 | { | |
981 | struct buffer_head *blk_bh = NULL; | |
0c044f0b | 982 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
983 | int ret = 0; |
984 | ||
985 | if (!di->i_xattr_loc) | |
986 | return ret; | |
987 | ||
4ae1d69b JB |
988 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
989 | &blk_bh); | |
cf1d6c76 TY |
990 | if (ret < 0) { |
991 | mlog_errno(ret); | |
992 | return ret; | |
993 | } | |
cf1d6c76 | 994 | |
0c044f0b | 995 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
0c044f0b TM |
996 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
997 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
998 | ret = ocfs2_xattr_list_entries(inode, header, | |
999 | buffer, buffer_size); | |
47bca495 TM |
1000 | } else |
1001 | ret = ocfs2_xattr_tree_list_index_block(inode, blk_bh, | |
0c044f0b | 1002 | buffer, buffer_size); |
4ae1d69b | 1003 | |
cf1d6c76 TY |
1004 | brelse(blk_bh); |
1005 | ||
1006 | return ret; | |
1007 | } | |
1008 | ||
1009 | ssize_t ocfs2_listxattr(struct dentry *dentry, | |
1010 | char *buffer, | |
1011 | size_t size) | |
1012 | { | |
1013 | int ret = 0, i_ret = 0, b_ret = 0; | |
1014 | struct buffer_head *di_bh = NULL; | |
1015 | struct ocfs2_dinode *di = NULL; | |
1016 | struct ocfs2_inode_info *oi = OCFS2_I(dentry->d_inode); | |
1017 | ||
8154da3d TY |
1018 | if (!ocfs2_supports_xattr(OCFS2_SB(dentry->d_sb))) |
1019 | return -EOPNOTSUPP; | |
1020 | ||
cf1d6c76 TY |
1021 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1022 | return ret; | |
1023 | ||
1024 | ret = ocfs2_inode_lock(dentry->d_inode, &di_bh, 0); | |
1025 | if (ret < 0) { | |
1026 | mlog_errno(ret); | |
1027 | return ret; | |
1028 | } | |
1029 | ||
1030 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1031 | ||
1032 | down_read(&oi->ip_xattr_sem); | |
1033 | i_ret = ocfs2_xattr_ibody_list(dentry->d_inode, di, buffer, size); | |
1034 | if (i_ret < 0) | |
1035 | b_ret = 0; | |
1036 | else { | |
1037 | if (buffer) { | |
1038 | buffer += i_ret; | |
1039 | size -= i_ret; | |
1040 | } | |
1041 | b_ret = ocfs2_xattr_block_list(dentry->d_inode, di, | |
1042 | buffer, size); | |
1043 | if (b_ret < 0) | |
1044 | i_ret = 0; | |
1045 | } | |
1046 | up_read(&oi->ip_xattr_sem); | |
1047 | ocfs2_inode_unlock(dentry->d_inode, 0); | |
1048 | ||
1049 | brelse(di_bh); | |
1050 | ||
1051 | return i_ret + b_ret; | |
1052 | } | |
1053 | ||
1054 | static int ocfs2_xattr_find_entry(int name_index, | |
1055 | const char *name, | |
1056 | struct ocfs2_xattr_search *xs) | |
1057 | { | |
1058 | struct ocfs2_xattr_entry *entry; | |
1059 | size_t name_len; | |
1060 | int i, cmp = 1; | |
1061 | ||
1062 | if (name == NULL) | |
1063 | return -EINVAL; | |
1064 | ||
1065 | name_len = strlen(name); | |
1066 | entry = xs->here; | |
1067 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
1068 | cmp = name_index - ocfs2_xattr_get_type(entry); | |
1069 | if (!cmp) | |
1070 | cmp = name_len - entry->xe_name_len; | |
1071 | if (!cmp) | |
1072 | cmp = memcmp(name, (xs->base + | |
1073 | le16_to_cpu(entry->xe_name_offset)), | |
1074 | name_len); | |
1075 | if (cmp == 0) | |
1076 | break; | |
1077 | entry += 1; | |
1078 | } | |
1079 | xs->here = entry; | |
1080 | ||
1081 | return cmp ? -ENODATA : 0; | |
1082 | } | |
1083 | ||
1084 | static int ocfs2_xattr_get_value_outside(struct inode *inode, | |
589dc260 | 1085 | struct ocfs2_xattr_value_root *xv, |
cf1d6c76 TY |
1086 | void *buffer, |
1087 | size_t len) | |
1088 | { | |
1089 | u32 cpos, p_cluster, num_clusters, bpc, clusters; | |
1090 | u64 blkno; | |
1091 | int i, ret = 0; | |
1092 | size_t cplen, blocksize; | |
1093 | struct buffer_head *bh = NULL; | |
cf1d6c76 TY |
1094 | struct ocfs2_extent_list *el; |
1095 | ||
cf1d6c76 TY |
1096 | el = &xv->xr_list; |
1097 | clusters = le32_to_cpu(xv->xr_clusters); | |
1098 | bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1099 | blocksize = inode->i_sb->s_blocksize; | |
1100 | ||
1101 | cpos = 0; | |
1102 | while (cpos < clusters) { | |
1103 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1104 | &num_clusters, el, NULL); |
cf1d6c76 TY |
1105 | if (ret) { |
1106 | mlog_errno(ret); | |
1107 | goto out; | |
1108 | } | |
1109 | ||
1110 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
1111 | /* Copy ocfs2_xattr_value */ | |
1112 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1113 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1114 | &bh, NULL); | |
cf1d6c76 TY |
1115 | if (ret) { |
1116 | mlog_errno(ret); | |
1117 | goto out; | |
1118 | } | |
1119 | ||
1120 | cplen = len >= blocksize ? blocksize : len; | |
1121 | memcpy(buffer, bh->b_data, cplen); | |
1122 | len -= cplen; | |
1123 | buffer += cplen; | |
1124 | ||
1125 | brelse(bh); | |
1126 | bh = NULL; | |
1127 | if (len == 0) | |
1128 | break; | |
1129 | } | |
1130 | cpos += num_clusters; | |
1131 | } | |
1132 | out: | |
1133 | return ret; | |
1134 | } | |
1135 | ||
1136 | static int ocfs2_xattr_ibody_get(struct inode *inode, | |
1137 | int name_index, | |
1138 | const char *name, | |
1139 | void *buffer, | |
1140 | size_t buffer_size, | |
1141 | struct ocfs2_xattr_search *xs) | |
1142 | { | |
1143 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1144 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
589dc260 | 1145 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 TY |
1146 | size_t size; |
1147 | int ret = 0; | |
1148 | ||
1149 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) | |
1150 | return -ENODATA; | |
1151 | ||
1152 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
1153 | xs->header = (struct ocfs2_xattr_header *) | |
1154 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
1155 | xs->base = (void *)xs->header; | |
1156 | xs->here = xs->header->xh_entries; | |
1157 | ||
1158 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
1159 | if (ret) | |
1160 | return ret; | |
1161 | size = le64_to_cpu(xs->here->xe_value_size); | |
1162 | if (buffer) { | |
1163 | if (size > buffer_size) | |
1164 | return -ERANGE; | |
1165 | if (ocfs2_xattr_is_local(xs->here)) { | |
1166 | memcpy(buffer, (void *)xs->base + | |
1167 | le16_to_cpu(xs->here->xe_name_offset) + | |
1168 | OCFS2_XATTR_SIZE(xs->here->xe_name_len), size); | |
1169 | } else { | |
589dc260 TM |
1170 | xv = (struct ocfs2_xattr_value_root *) |
1171 | (xs->base + le16_to_cpu( | |
1172 | xs->here->xe_name_offset) + | |
1173 | OCFS2_XATTR_SIZE(xs->here->xe_name_len)); | |
1174 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1175 | buffer, size); |
1176 | if (ret < 0) { | |
1177 | mlog_errno(ret); | |
1178 | return ret; | |
1179 | } | |
1180 | } | |
1181 | } | |
1182 | ||
1183 | return size; | |
1184 | } | |
1185 | ||
1186 | static int ocfs2_xattr_block_get(struct inode *inode, | |
1187 | int name_index, | |
1188 | const char *name, | |
1189 | void *buffer, | |
1190 | size_t buffer_size, | |
1191 | struct ocfs2_xattr_search *xs) | |
1192 | { | |
cf1d6c76 | 1193 | struct ocfs2_xattr_block *xb; |
589dc260 | 1194 | struct ocfs2_xattr_value_root *xv; |
cf1d6c76 | 1195 | size_t size; |
44d8e4e1 SM |
1196 | int ret = -ENODATA, name_offset, name_len, i; |
1197 | int uninitialized_var(block_off); | |
cf1d6c76 | 1198 | |
ba937127 JB |
1199 | xs->bucket = ocfs2_xattr_bucket_new(inode); |
1200 | if (!xs->bucket) { | |
1201 | ret = -ENOMEM; | |
1202 | mlog_errno(ret); | |
1203 | goto cleanup; | |
1204 | } | |
589dc260 | 1205 | |
54f443f4 JB |
1206 | ret = ocfs2_xattr_block_find(inode, name_index, name, xs); |
1207 | if (ret) { | |
cf1d6c76 | 1208 | mlog_errno(ret); |
cf1d6c76 TY |
1209 | goto cleanup; |
1210 | } | |
1211 | ||
6c1e183e TY |
1212 | if (xs->not_found) { |
1213 | ret = -ENODATA; | |
1214 | goto cleanup; | |
1215 | } | |
1216 | ||
54f443f4 | 1217 | xb = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
1218 | size = le64_to_cpu(xs->here->xe_value_size); |
1219 | if (buffer) { | |
1220 | ret = -ERANGE; | |
1221 | if (size > buffer_size) | |
1222 | goto cleanup; | |
589dc260 TM |
1223 | |
1224 | name_offset = le16_to_cpu(xs->here->xe_name_offset); | |
1225 | name_len = OCFS2_XATTR_SIZE(xs->here->xe_name_len); | |
1226 | i = xs->here - xs->header->xh_entries; | |
1227 | ||
1228 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 1229 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
ba937127 | 1230 | bucket_xh(xs->bucket), |
589dc260 TM |
1231 | i, |
1232 | &block_off, | |
1233 | &name_offset); | |
ba937127 | 1234 | xs->base = bucket_block(xs->bucket, block_off); |
589dc260 | 1235 | } |
cf1d6c76 TY |
1236 | if (ocfs2_xattr_is_local(xs->here)) { |
1237 | memcpy(buffer, (void *)xs->base + | |
589dc260 | 1238 | name_offset + name_len, size); |
cf1d6c76 | 1239 | } else { |
589dc260 TM |
1240 | xv = (struct ocfs2_xattr_value_root *) |
1241 | (xs->base + name_offset + name_len); | |
1242 | ret = ocfs2_xattr_get_value_outside(inode, xv, | |
cf1d6c76 TY |
1243 | buffer, size); |
1244 | if (ret < 0) { | |
1245 | mlog_errno(ret); | |
1246 | goto cleanup; | |
1247 | } | |
1248 | } | |
1249 | } | |
1250 | ret = size; | |
1251 | cleanup: | |
ba937127 | 1252 | ocfs2_xattr_bucket_free(xs->bucket); |
cf1d6c76 | 1253 | |
54f443f4 JB |
1254 | brelse(xs->xattr_bh); |
1255 | xs->xattr_bh = NULL; | |
cf1d6c76 TY |
1256 | return ret; |
1257 | } | |
1258 | ||
4e3e9d02 TY |
1259 | int ocfs2_xattr_get_nolock(struct inode *inode, |
1260 | struct buffer_head *di_bh, | |
0030e001 TY |
1261 | int name_index, |
1262 | const char *name, | |
1263 | void *buffer, | |
1264 | size_t buffer_size) | |
cf1d6c76 TY |
1265 | { |
1266 | int ret; | |
1267 | struct ocfs2_dinode *di = NULL; | |
cf1d6c76 TY |
1268 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
1269 | struct ocfs2_xattr_search xis = { | |
1270 | .not_found = -ENODATA, | |
1271 | }; | |
1272 | struct ocfs2_xattr_search xbs = { | |
1273 | .not_found = -ENODATA, | |
1274 | }; | |
1275 | ||
8154da3d TY |
1276 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
1277 | return -EOPNOTSUPP; | |
1278 | ||
cf1d6c76 TY |
1279 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
1280 | ret = -ENODATA; | |
1281 | ||
cf1d6c76 TY |
1282 | xis.inode_bh = xbs.inode_bh = di_bh; |
1283 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
1284 | ||
cf1d6c76 TY |
1285 | ret = ocfs2_xattr_ibody_get(inode, name_index, name, buffer, |
1286 | buffer_size, &xis); | |
6c1e183e | 1287 | if (ret == -ENODATA && di->i_xattr_loc) |
cf1d6c76 TY |
1288 | ret = ocfs2_xattr_block_get(inode, name_index, name, buffer, |
1289 | buffer_size, &xbs); | |
4e3e9d02 TY |
1290 | |
1291 | return ret; | |
1292 | } | |
1293 | ||
1294 | /* ocfs2_xattr_get() | |
1295 | * | |
1296 | * Copy an extended attribute into the buffer provided. | |
1297 | * Buffer is NULL to compute the size of buffer required. | |
1298 | */ | |
1299 | static int ocfs2_xattr_get(struct inode *inode, | |
1300 | int name_index, | |
1301 | const char *name, | |
1302 | void *buffer, | |
1303 | size_t buffer_size) | |
1304 | { | |
1305 | int ret; | |
1306 | struct buffer_head *di_bh = NULL; | |
1307 | ||
1308 | ret = ocfs2_inode_lock(inode, &di_bh, 0); | |
1309 | if (ret < 0) { | |
1310 | mlog_errno(ret); | |
1311 | return ret; | |
1312 | } | |
5e64b0d9 | 1313 | down_read(&OCFS2_I(inode)->ip_xattr_sem); |
4e3e9d02 TY |
1314 | ret = ocfs2_xattr_get_nolock(inode, di_bh, name_index, |
1315 | name, buffer, buffer_size); | |
5e64b0d9 | 1316 | up_read(&OCFS2_I(inode)->ip_xattr_sem); |
4e3e9d02 | 1317 | |
cf1d6c76 TY |
1318 | ocfs2_inode_unlock(inode, 0); |
1319 | ||
1320 | brelse(di_bh); | |
1321 | ||
1322 | return ret; | |
1323 | } | |
1324 | ||
1325 | static int __ocfs2_xattr_set_value_outside(struct inode *inode, | |
85db90e7 | 1326 | handle_t *handle, |
492a8a33 | 1327 | struct ocfs2_xattr_value_buf *vb, |
cf1d6c76 TY |
1328 | const void *value, |
1329 | int value_len) | |
1330 | { | |
71d548a6 | 1331 | int ret = 0, i, cp_len; |
cf1d6c76 TY |
1332 | u16 blocksize = inode->i_sb->s_blocksize; |
1333 | u32 p_cluster, num_clusters; | |
1334 | u32 cpos = 0, bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
1335 | u32 clusters = ocfs2_clusters_for_bytes(inode->i_sb, value_len); | |
1336 | u64 blkno; | |
1337 | struct buffer_head *bh = NULL; | |
492a8a33 TM |
1338 | unsigned int ext_flags; |
1339 | struct ocfs2_xattr_value_root *xv = vb->vb_xv; | |
cf1d6c76 TY |
1340 | |
1341 | BUG_ON(clusters > le32_to_cpu(xv->xr_clusters)); | |
1342 | ||
cf1d6c76 TY |
1343 | while (cpos < clusters) { |
1344 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
1061f9c1 | 1345 | &num_clusters, &xv->xr_list, |
492a8a33 | 1346 | &ext_flags); |
cf1d6c76 TY |
1347 | if (ret) { |
1348 | mlog_errno(ret); | |
85db90e7 | 1349 | goto out; |
cf1d6c76 TY |
1350 | } |
1351 | ||
492a8a33 TM |
1352 | BUG_ON(ext_flags & OCFS2_EXT_REFCOUNTED); |
1353 | ||
cf1d6c76 TY |
1354 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); |
1355 | ||
1356 | for (i = 0; i < num_clusters * bpc; i++, blkno++) { | |
8cb471e8 JB |
1357 | ret = ocfs2_read_block(INODE_CACHE(inode), blkno, |
1358 | &bh, NULL); | |
cf1d6c76 TY |
1359 | if (ret) { |
1360 | mlog_errno(ret); | |
85db90e7 | 1361 | goto out; |
cf1d6c76 TY |
1362 | } |
1363 | ||
1364 | ret = ocfs2_journal_access(handle, | |
0cf2f763 | 1365 | INODE_CACHE(inode), |
cf1d6c76 TY |
1366 | bh, |
1367 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1368 | if (ret < 0) { | |
1369 | mlog_errno(ret); | |
85db90e7 | 1370 | goto out; |
cf1d6c76 TY |
1371 | } |
1372 | ||
1373 | cp_len = value_len > blocksize ? blocksize : value_len; | |
1374 | memcpy(bh->b_data, value, cp_len); | |
1375 | value_len -= cp_len; | |
1376 | value += cp_len; | |
1377 | if (cp_len < blocksize) | |
1378 | memset(bh->b_data + cp_len, 0, | |
1379 | blocksize - cp_len); | |
1380 | ||
ec20cec7 | 1381 | ocfs2_journal_dirty(handle, bh); |
cf1d6c76 TY |
1382 | brelse(bh); |
1383 | bh = NULL; | |
1384 | ||
1385 | /* | |
1386 | * XXX: do we need to empty all the following | |
1387 | * blocks in this cluster? | |
1388 | */ | |
1389 | if (!value_len) | |
1390 | break; | |
1391 | } | |
1392 | cpos += num_clusters; | |
1393 | } | |
cf1d6c76 TY |
1394 | out: |
1395 | brelse(bh); | |
1396 | ||
1397 | return ret; | |
1398 | } | |
1399 | ||
69a3e539 JB |
1400 | static int ocfs2_xa_check_space_helper(int needed_space, int free_start, |
1401 | int num_entries) | |
1402 | { | |
1403 | int free_space; | |
1404 | ||
1405 | if (!needed_space) | |
1406 | return 0; | |
1407 | ||
1408 | free_space = free_start - | |
1409 | sizeof(struct ocfs2_xattr_header) - | |
1410 | (num_entries * sizeof(struct ocfs2_xattr_entry)) - | |
1411 | OCFS2_XATTR_HEADER_GAP; | |
1412 | if (free_space < 0) | |
1413 | return -EIO; | |
1414 | if (free_space < needed_space) | |
1415 | return -ENOSPC; | |
1416 | ||
1417 | return 0; | |
1418 | } | |
1419 | ||
cf2bc809 JB |
1420 | static int ocfs2_xa_journal_access(handle_t *handle, struct ocfs2_xa_loc *loc, |
1421 | int type) | |
1422 | { | |
1423 | return loc->xl_ops->xlo_journal_access(handle, loc, type); | |
1424 | } | |
1425 | ||
1426 | static void ocfs2_xa_journal_dirty(handle_t *handle, struct ocfs2_xa_loc *loc) | |
1427 | { | |
1428 | loc->xl_ops->xlo_journal_dirty(handle, loc); | |
1429 | } | |
1430 | ||
69a3e539 JB |
1431 | /* Give a pointer into the storage for the given offset */ |
1432 | static void *ocfs2_xa_offset_pointer(struct ocfs2_xa_loc *loc, int offset) | |
1433 | { | |
1434 | BUG_ON(offset >= loc->xl_size); | |
1435 | return loc->xl_ops->xlo_offset_pointer(loc, offset); | |
1436 | } | |
1437 | ||
11179f2c JB |
1438 | /* |
1439 | * Wipe the name+value pair and allow the storage to reclaim it. This | |
1440 | * must be followed by either removal of the entry or a call to | |
1441 | * ocfs2_xa_add_namevalue(). | |
1442 | */ | |
1443 | static void ocfs2_xa_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1444 | { | |
1445 | loc->xl_ops->xlo_wipe_namevalue(loc); | |
1446 | } | |
1447 | ||
69a3e539 JB |
1448 | /* |
1449 | * Find lowest offset to a name+value pair. This is the start of our | |
1450 | * downward-growing free space. | |
1451 | */ | |
1452 | static int ocfs2_xa_get_free_start(struct ocfs2_xa_loc *loc) | |
1453 | { | |
1454 | return loc->xl_ops->xlo_get_free_start(loc); | |
1455 | } | |
1456 | ||
1457 | /* Can we reuse loc->xl_entry for xi? */ | |
1458 | static int ocfs2_xa_can_reuse_entry(struct ocfs2_xa_loc *loc, | |
1459 | struct ocfs2_xattr_info *xi) | |
1460 | { | |
1461 | return loc->xl_ops->xlo_can_reuse(loc, xi); | |
1462 | } | |
1463 | ||
1464 | /* How much free space is needed to set the new value */ | |
1465 | static int ocfs2_xa_check_space(struct ocfs2_xa_loc *loc, | |
1466 | struct ocfs2_xattr_info *xi) | |
1467 | { | |
1468 | return loc->xl_ops->xlo_check_space(loc, xi); | |
1469 | } | |
1470 | ||
1471 | static void ocfs2_xa_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) | |
1472 | { | |
1473 | loc->xl_ops->xlo_add_entry(loc, name_hash); | |
1474 | loc->xl_entry->xe_name_hash = cpu_to_le32(name_hash); | |
1475 | /* | |
1476 | * We can't leave the new entry's xe_name_offset at zero or | |
1477 | * add_namevalue() will go nuts. We set it to the size of our | |
1478 | * storage so that it can never be less than any other entry. | |
1479 | */ | |
1480 | loc->xl_entry->xe_name_offset = cpu_to_le16(loc->xl_size); | |
1481 | } | |
1482 | ||
1483 | static void ocfs2_xa_add_namevalue(struct ocfs2_xa_loc *loc, | |
1484 | struct ocfs2_xattr_info *xi) | |
1485 | { | |
1486 | int size = namevalue_size_xi(xi); | |
1487 | int nameval_offset; | |
1488 | char *nameval_buf; | |
1489 | ||
1490 | loc->xl_ops->xlo_add_namevalue(loc, size); | |
1491 | loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); | |
1492 | loc->xl_entry->xe_name_len = xi->xi_name_len; | |
1493 | ocfs2_xattr_set_type(loc->xl_entry, xi->xi_name_index); | |
1494 | ocfs2_xattr_set_local(loc->xl_entry, | |
1495 | xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE); | |
1496 | ||
1497 | nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1498 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); | |
1499 | memset(nameval_buf, 0, size); | |
1500 | memcpy(nameval_buf, xi->xi_name, xi->xi_name_len); | |
1501 | } | |
1502 | ||
3fc12afa JB |
1503 | static void ocfs2_xa_fill_value_buf(struct ocfs2_xa_loc *loc, |
1504 | struct ocfs2_xattr_value_buf *vb) | |
1505 | { | |
1506 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); | |
1507 | int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); | |
1508 | ||
1509 | /* Value bufs are for value trees */ | |
73857ee0 | 1510 | BUG_ON(ocfs2_xattr_is_local(loc->xl_entry)); |
3fc12afa JB |
1511 | BUG_ON(namevalue_size_xe(loc->xl_entry) != |
1512 | (name_size + OCFS2_XATTR_ROOT_SIZE)); | |
1513 | ||
1514 | loc->xl_ops->xlo_fill_value_buf(loc, vb); | |
1515 | vb->vb_xv = | |
1516 | (struct ocfs2_xattr_value_root *)ocfs2_xa_offset_pointer(loc, | |
1517 | nameval_offset + | |
1518 | name_size); | |
1519 | } | |
1520 | ||
cf2bc809 JB |
1521 | static int ocfs2_xa_block_journal_access(handle_t *handle, |
1522 | struct ocfs2_xa_loc *loc, int type) | |
1523 | { | |
1524 | struct buffer_head *bh = loc->xl_storage; | |
1525 | ocfs2_journal_access_func access; | |
1526 | ||
1527 | if (loc->xl_size == (bh->b_size - | |
1528 | offsetof(struct ocfs2_xattr_block, | |
1529 | xb_attrs.xb_header))) | |
1530 | access = ocfs2_journal_access_xb; | |
1531 | else | |
1532 | access = ocfs2_journal_access_di; | |
1533 | return access(handle, INODE_CACHE(loc->xl_inode), bh, type); | |
1534 | } | |
1535 | ||
1536 | static void ocfs2_xa_block_journal_dirty(handle_t *handle, | |
1537 | struct ocfs2_xa_loc *loc) | |
1538 | { | |
1539 | struct buffer_head *bh = loc->xl_storage; | |
1540 | ||
1541 | ocfs2_journal_dirty(handle, bh); | |
1542 | } | |
1543 | ||
11179f2c JB |
1544 | static void *ocfs2_xa_block_offset_pointer(struct ocfs2_xa_loc *loc, |
1545 | int offset) | |
1546 | { | |
11179f2c JB |
1547 | return (char *)loc->xl_header + offset; |
1548 | } | |
1549 | ||
69a3e539 JB |
1550 | static int ocfs2_xa_block_can_reuse(struct ocfs2_xa_loc *loc, |
1551 | struct ocfs2_xattr_info *xi) | |
1552 | { | |
1553 | /* | |
1554 | * Block storage is strict. If the sizes aren't exact, we will | |
1555 | * remove the old one and reinsert the new. | |
1556 | */ | |
1557 | return namevalue_size_xe(loc->xl_entry) == | |
1558 | namevalue_size_xi(xi); | |
1559 | } | |
1560 | ||
1561 | static int ocfs2_xa_block_get_free_start(struct ocfs2_xa_loc *loc) | |
1562 | { | |
1563 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1564 | int i, count = le16_to_cpu(xh->xh_count); | |
1565 | int offset, free_start = loc->xl_size; | |
1566 | ||
1567 | for (i = 0; i < count; i++) { | |
1568 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
1569 | if (offset < free_start) | |
1570 | free_start = offset; | |
1571 | } | |
1572 | ||
1573 | return free_start; | |
1574 | } | |
1575 | ||
1576 | static int ocfs2_xa_block_check_space(struct ocfs2_xa_loc *loc, | |
1577 | struct ocfs2_xattr_info *xi) | |
1578 | { | |
1579 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1580 | int free_start = ocfs2_xa_get_free_start(loc); | |
1581 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1582 | ||
1583 | /* | |
1584 | * Block storage will reclaim the original entry before inserting | |
1585 | * the new value, so we only need the difference. If the new | |
1586 | * entry is smaller than the old one, we don't need anything. | |
1587 | */ | |
1588 | if (loc->xl_entry) { | |
1589 | /* Don't need space if we're reusing! */ | |
1590 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1591 | needed_space = 0; | |
1592 | else | |
1593 | needed_space -= ocfs2_xe_entry_usage(loc->xl_entry); | |
1594 | } | |
1595 | if (needed_space < 0) | |
1596 | needed_space = 0; | |
1597 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1598 | } | |
1599 | ||
11179f2c JB |
1600 | /* |
1601 | * Block storage for xattrs keeps the name+value pairs compacted. When | |
1602 | * we remove one, we have to shift any that preceded it towards the end. | |
1603 | */ | |
1604 | static void ocfs2_xa_block_wipe_namevalue(struct ocfs2_xa_loc *loc) | |
1605 | { | |
1606 | int i, offset; | |
1607 | int namevalue_offset, first_namevalue_offset, namevalue_size; | |
1608 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1609 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
11179f2c JB |
1610 | int count = le16_to_cpu(xh->xh_count); |
1611 | ||
1612 | namevalue_offset = le16_to_cpu(entry->xe_name_offset); | |
199799a3 | 1613 | namevalue_size = namevalue_size_xe(entry); |
69a3e539 | 1614 | first_namevalue_offset = ocfs2_xa_get_free_start(loc); |
11179f2c JB |
1615 | |
1616 | /* Shift the name+value pairs */ | |
1617 | memmove((char *)xh + first_namevalue_offset + namevalue_size, | |
1618 | (char *)xh + first_namevalue_offset, | |
1619 | namevalue_offset - first_namevalue_offset); | |
1620 | memset((char *)xh + first_namevalue_offset, 0, namevalue_size); | |
1621 | ||
1622 | /* Now tell xh->xh_entries about it */ | |
1623 | for (i = 0; i < count; i++) { | |
1624 | offset = le16_to_cpu(xh->xh_entries[i].xe_name_offset); | |
dfe4d3d6 | 1625 | if (offset <= namevalue_offset) |
11179f2c JB |
1626 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, |
1627 | namevalue_size); | |
1628 | } | |
1629 | ||
1630 | /* | |
1631 | * Note that we don't update xh_free_start or xh_name_value_len | |
1632 | * because they're not used in block-stored xattrs. | |
1633 | */ | |
1634 | } | |
1635 | ||
69a3e539 JB |
1636 | static void ocfs2_xa_block_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1637 | { | |
1638 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1639 | loc->xl_entry = &(loc->xl_header->xh_entries[count]); | |
1640 | le16_add_cpu(&loc->xl_header->xh_count, 1); | |
1641 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1642 | } | |
1643 | ||
1644 | static void ocfs2_xa_block_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1645 | { | |
1646 | int free_start = ocfs2_xa_get_free_start(loc); | |
1647 | ||
1648 | loc->xl_entry->xe_name_offset = cpu_to_le16(free_start - size); | |
1649 | } | |
1650 | ||
3fc12afa JB |
1651 | static void ocfs2_xa_block_fill_value_buf(struct ocfs2_xa_loc *loc, |
1652 | struct ocfs2_xattr_value_buf *vb) | |
1653 | { | |
1654 | struct buffer_head *bh = loc->xl_storage; | |
1655 | ||
1656 | if (loc->xl_size == (bh->b_size - | |
1657 | offsetof(struct ocfs2_xattr_block, | |
1658 | xb_attrs.xb_header))) | |
1659 | vb->vb_access = ocfs2_journal_access_xb; | |
1660 | else | |
1661 | vb->vb_access = ocfs2_journal_access_di; | |
1662 | vb->vb_bh = bh; | |
1663 | } | |
1664 | ||
11179f2c JB |
1665 | /* |
1666 | * Operations for xattrs stored in blocks. This includes inline inode | |
1667 | * storage and unindexed ocfs2_xattr_blocks. | |
1668 | */ | |
1669 | static const struct ocfs2_xa_loc_operations ocfs2_xa_block_loc_ops = { | |
cf2bc809 JB |
1670 | .xlo_journal_access = ocfs2_xa_block_journal_access, |
1671 | .xlo_journal_dirty = ocfs2_xa_block_journal_dirty, | |
11179f2c | 1672 | .xlo_offset_pointer = ocfs2_xa_block_offset_pointer, |
69a3e539 JB |
1673 | .xlo_check_space = ocfs2_xa_block_check_space, |
1674 | .xlo_can_reuse = ocfs2_xa_block_can_reuse, | |
1675 | .xlo_get_free_start = ocfs2_xa_block_get_free_start, | |
11179f2c | 1676 | .xlo_wipe_namevalue = ocfs2_xa_block_wipe_namevalue, |
69a3e539 JB |
1677 | .xlo_add_entry = ocfs2_xa_block_add_entry, |
1678 | .xlo_add_namevalue = ocfs2_xa_block_add_namevalue, | |
3fc12afa | 1679 | .xlo_fill_value_buf = ocfs2_xa_block_fill_value_buf, |
11179f2c JB |
1680 | }; |
1681 | ||
cf2bc809 JB |
1682 | static int ocfs2_xa_bucket_journal_access(handle_t *handle, |
1683 | struct ocfs2_xa_loc *loc, int type) | |
1684 | { | |
1685 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1686 | ||
1687 | return ocfs2_xattr_bucket_journal_access(handle, bucket, type); | |
1688 | } | |
1689 | ||
1690 | static void ocfs2_xa_bucket_journal_dirty(handle_t *handle, | |
1691 | struct ocfs2_xa_loc *loc) | |
1692 | { | |
1693 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1694 | ||
1695 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
1696 | } | |
1697 | ||
11179f2c JB |
1698 | static void *ocfs2_xa_bucket_offset_pointer(struct ocfs2_xa_loc *loc, |
1699 | int offset) | |
1700 | { | |
1701 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1702 | int block, block_offset; | |
1703 | ||
11179f2c | 1704 | /* The header is at the front of the bucket */ |
cf2bc809 JB |
1705 | block = offset >> loc->xl_inode->i_sb->s_blocksize_bits; |
1706 | block_offset = offset % loc->xl_inode->i_sb->s_blocksize; | |
11179f2c JB |
1707 | |
1708 | return bucket_block(bucket, block) + block_offset; | |
1709 | } | |
1710 | ||
69a3e539 JB |
1711 | static int ocfs2_xa_bucket_can_reuse(struct ocfs2_xa_loc *loc, |
1712 | struct ocfs2_xattr_info *xi) | |
1713 | { | |
1714 | return namevalue_size_xe(loc->xl_entry) >= | |
1715 | namevalue_size_xi(xi); | |
1716 | } | |
1717 | ||
1718 | static int ocfs2_xa_bucket_get_free_start(struct ocfs2_xa_loc *loc) | |
1719 | { | |
1720 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
1721 | return le16_to_cpu(bucket_xh(bucket)->xh_free_start); | |
1722 | } | |
1723 | ||
1724 | static int ocfs2_bucket_align_free_start(struct super_block *sb, | |
1725 | int free_start, int size) | |
1726 | { | |
1727 | /* | |
1728 | * We need to make sure that the name+value pair fits within | |
1729 | * one block. | |
1730 | */ | |
1731 | if (((free_start - size) >> sb->s_blocksize_bits) != | |
1732 | ((free_start - 1) >> sb->s_blocksize_bits)) | |
1733 | free_start -= free_start % sb->s_blocksize; | |
1734 | ||
1735 | return free_start; | |
1736 | } | |
1737 | ||
1738 | static int ocfs2_xa_bucket_check_space(struct ocfs2_xa_loc *loc, | |
1739 | struct ocfs2_xattr_info *xi) | |
1740 | { | |
1741 | int rc; | |
1742 | int count = le16_to_cpu(loc->xl_header->xh_count); | |
1743 | int free_start = ocfs2_xa_get_free_start(loc); | |
1744 | int needed_space = ocfs2_xi_entry_usage(xi); | |
1745 | int size = namevalue_size_xi(xi); | |
cf2bc809 | 1746 | struct super_block *sb = loc->xl_inode->i_sb; |
69a3e539 JB |
1747 | |
1748 | /* | |
1749 | * Bucket storage does not reclaim name+value pairs it cannot | |
1750 | * reuse. They live as holes until the bucket fills, and then | |
1751 | * the bucket is defragmented. However, the bucket can reclaim | |
1752 | * the ocfs2_xattr_entry. | |
1753 | */ | |
1754 | if (loc->xl_entry) { | |
1755 | /* Don't need space if we're reusing! */ | |
1756 | if (ocfs2_xa_can_reuse_entry(loc, xi)) | |
1757 | needed_space = 0; | |
1758 | else | |
1759 | needed_space -= sizeof(struct ocfs2_xattr_entry); | |
1760 | } | |
1761 | BUG_ON(needed_space < 0); | |
1762 | ||
1763 | if (free_start < size) { | |
1764 | if (needed_space) | |
1765 | return -ENOSPC; | |
1766 | } else { | |
1767 | /* | |
1768 | * First we check if it would fit in the first place. | |
1769 | * Below, we align the free start to a block. This may | |
1770 | * slide us below the minimum gap. By checking unaligned | |
1771 | * first, we avoid that error. | |
1772 | */ | |
1773 | rc = ocfs2_xa_check_space_helper(needed_space, free_start, | |
1774 | count); | |
1775 | if (rc) | |
1776 | return rc; | |
1777 | free_start = ocfs2_bucket_align_free_start(sb, free_start, | |
1778 | size); | |
1779 | } | |
1780 | return ocfs2_xa_check_space_helper(needed_space, free_start, count); | |
1781 | } | |
1782 | ||
11179f2c JB |
1783 | static void ocfs2_xa_bucket_wipe_namevalue(struct ocfs2_xa_loc *loc) |
1784 | { | |
199799a3 JB |
1785 | le16_add_cpu(&loc->xl_header->xh_name_value_len, |
1786 | -namevalue_size_xe(loc->xl_entry)); | |
11179f2c JB |
1787 | } |
1788 | ||
69a3e539 JB |
1789 | static void ocfs2_xa_bucket_add_entry(struct ocfs2_xa_loc *loc, u32 name_hash) |
1790 | { | |
1791 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1792 | int count = le16_to_cpu(xh->xh_count); | |
1793 | int low = 0, high = count - 1, tmp; | |
1794 | struct ocfs2_xattr_entry *tmp_xe; | |
1795 | ||
1796 | /* | |
1797 | * We keep buckets sorted by name_hash, so we need to find | |
1798 | * our insert place. | |
1799 | */ | |
1800 | while (low <= high && count) { | |
1801 | tmp = (low + high) / 2; | |
1802 | tmp_xe = &xh->xh_entries[tmp]; | |
1803 | ||
1804 | if (name_hash > le32_to_cpu(tmp_xe->xe_name_hash)) | |
1805 | low = tmp + 1; | |
1806 | else if (name_hash < le32_to_cpu(tmp_xe->xe_name_hash)) | |
1807 | high = tmp - 1; | |
1808 | else { | |
1809 | low = tmp; | |
1810 | break; | |
1811 | } | |
1812 | } | |
1813 | ||
1814 | if (low != count) | |
1815 | memmove(&xh->xh_entries[low + 1], | |
1816 | &xh->xh_entries[low], | |
1817 | ((count - low) * sizeof(struct ocfs2_xattr_entry))); | |
1818 | ||
1819 | le16_add_cpu(&xh->xh_count, 1); | |
1820 | loc->xl_entry = &xh->xh_entries[low]; | |
1821 | memset(loc->xl_entry, 0, sizeof(struct ocfs2_xattr_entry)); | |
1822 | } | |
1823 | ||
1824 | static void ocfs2_xa_bucket_add_namevalue(struct ocfs2_xa_loc *loc, int size) | |
1825 | { | |
1826 | int free_start = ocfs2_xa_get_free_start(loc); | |
1827 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
cf2bc809 | 1828 | struct super_block *sb = loc->xl_inode->i_sb; |
69a3e539 JB |
1829 | int nameval_offset; |
1830 | ||
1831 | free_start = ocfs2_bucket_align_free_start(sb, free_start, size); | |
1832 | nameval_offset = free_start - size; | |
1833 | loc->xl_entry->xe_name_offset = cpu_to_le16(nameval_offset); | |
1834 | xh->xh_free_start = cpu_to_le16(nameval_offset); | |
1835 | le16_add_cpu(&xh->xh_name_value_len, size); | |
1836 | ||
1837 | } | |
1838 | ||
3fc12afa JB |
1839 | static void ocfs2_xa_bucket_fill_value_buf(struct ocfs2_xa_loc *loc, |
1840 | struct ocfs2_xattr_value_buf *vb) | |
1841 | { | |
1842 | struct ocfs2_xattr_bucket *bucket = loc->xl_storage; | |
cf2bc809 | 1843 | struct super_block *sb = loc->xl_inode->i_sb; |
3fc12afa JB |
1844 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); |
1845 | int size = namevalue_size_xe(loc->xl_entry); | |
1846 | int block_offset = nameval_offset >> sb->s_blocksize_bits; | |
1847 | ||
1848 | /* Values are not allowed to straddle block boundaries */ | |
1849 | BUG_ON(block_offset != | |
1850 | ((nameval_offset + size - 1) >> sb->s_blocksize_bits)); | |
1851 | /* We expect the bucket to be filled in */ | |
1852 | BUG_ON(!bucket->bu_bhs[block_offset]); | |
1853 | ||
1854 | vb->vb_access = ocfs2_journal_access; | |
1855 | vb->vb_bh = bucket->bu_bhs[block_offset]; | |
1856 | } | |
1857 | ||
11179f2c JB |
1858 | /* Operations for xattrs stored in buckets. */ |
1859 | static const struct ocfs2_xa_loc_operations ocfs2_xa_bucket_loc_ops = { | |
cf2bc809 JB |
1860 | .xlo_journal_access = ocfs2_xa_bucket_journal_access, |
1861 | .xlo_journal_dirty = ocfs2_xa_bucket_journal_dirty, | |
11179f2c | 1862 | .xlo_offset_pointer = ocfs2_xa_bucket_offset_pointer, |
69a3e539 JB |
1863 | .xlo_check_space = ocfs2_xa_bucket_check_space, |
1864 | .xlo_can_reuse = ocfs2_xa_bucket_can_reuse, | |
1865 | .xlo_get_free_start = ocfs2_xa_bucket_get_free_start, | |
11179f2c | 1866 | .xlo_wipe_namevalue = ocfs2_xa_bucket_wipe_namevalue, |
69a3e539 JB |
1867 | .xlo_add_entry = ocfs2_xa_bucket_add_entry, |
1868 | .xlo_add_namevalue = ocfs2_xa_bucket_add_namevalue, | |
3fc12afa | 1869 | .xlo_fill_value_buf = ocfs2_xa_bucket_fill_value_buf, |
11179f2c JB |
1870 | }; |
1871 | ||
399ff3a7 JB |
1872 | static unsigned int ocfs2_xa_value_clusters(struct ocfs2_xa_loc *loc) |
1873 | { | |
1874 | struct ocfs2_xattr_value_buf vb; | |
1875 | ||
1876 | if (ocfs2_xattr_is_local(loc->xl_entry)) | |
1877 | return 0; | |
1878 | ||
1879 | ocfs2_xa_fill_value_buf(loc, &vb); | |
1880 | return le32_to_cpu(vb.vb_xv->xr_clusters); | |
1881 | } | |
1882 | ||
73857ee0 JB |
1883 | static int ocfs2_xa_value_truncate(struct ocfs2_xa_loc *loc, u64 bytes, |
1884 | struct ocfs2_xattr_set_ctxt *ctxt) | |
1885 | { | |
1886 | int trunc_rc, access_rc; | |
1887 | struct ocfs2_xattr_value_buf vb; | |
1888 | ||
1889 | ocfs2_xa_fill_value_buf(loc, &vb); | |
1890 | trunc_rc = ocfs2_xattr_value_truncate(loc->xl_inode, &vb, bytes, | |
1891 | ctxt); | |
1892 | ||
1893 | /* | |
1894 | * The caller of ocfs2_xa_value_truncate() has already called | |
1895 | * ocfs2_xa_journal_access on the loc. However, The truncate code | |
1896 | * calls ocfs2_extend_trans(). This may commit the previous | |
1897 | * transaction and open a new one. If this is a bucket, truncate | |
1898 | * could leave only vb->vb_bh set up for journaling. Meanwhile, | |
1899 | * the caller is expecting to dirty the entire bucket. So we must | |
1900 | * reset the journal work. We do this even if truncate has failed, | |
1901 | * as it could have failed after committing the extend. | |
1902 | */ | |
1903 | access_rc = ocfs2_xa_journal_access(ctxt->handle, loc, | |
1904 | OCFS2_JOURNAL_ACCESS_WRITE); | |
1905 | ||
1906 | /* Errors in truncate take precedence */ | |
1907 | return trunc_rc ? trunc_rc : access_rc; | |
1908 | } | |
1909 | ||
11179f2c JB |
1910 | static void ocfs2_xa_remove_entry(struct ocfs2_xa_loc *loc) |
1911 | { | |
bde1e540 JB |
1912 | int index, count; |
1913 | struct ocfs2_xattr_header *xh = loc->xl_header; | |
1914 | struct ocfs2_xattr_entry *entry = loc->xl_entry; | |
1915 | ||
11179f2c | 1916 | ocfs2_xa_wipe_namevalue(loc); |
bde1e540 JB |
1917 | loc->xl_entry = NULL; |
1918 | ||
1919 | le16_add_cpu(&xh->xh_count, -1); | |
1920 | count = le16_to_cpu(xh->xh_count); | |
1921 | ||
1922 | /* | |
1923 | * Only zero out the entry if there are more remaining. This is | |
1924 | * important for an empty bucket, as it keeps track of the | |
1925 | * bucket's hash value. It doesn't hurt empty block storage. | |
1926 | */ | |
1927 | if (count) { | |
1928 | index = ((char *)entry - (char *)&xh->xh_entries) / | |
1929 | sizeof(struct ocfs2_xattr_entry); | |
1930 | memmove(&xh->xh_entries[index], &xh->xh_entries[index + 1], | |
1931 | (count - index) * sizeof(struct ocfs2_xattr_entry)); | |
1932 | memset(&xh->xh_entries[count], 0, | |
1933 | sizeof(struct ocfs2_xattr_entry)); | |
1934 | } | |
11179f2c JB |
1935 | } |
1936 | ||
399ff3a7 JB |
1937 | /* |
1938 | * If we have a problem adjusting the size of an external value during | |
1939 | * ocfs2_xa_prepare_entry() or ocfs2_xa_remove(), we may have an xattr | |
1940 | * in an intermediate state. For example, the value may be partially | |
1941 | * truncated. | |
1942 | * | |
1943 | * If the value tree hasn't changed, the extend/truncate went nowhere. | |
1944 | * We have nothing to do. The caller can treat it as a straight error. | |
1945 | * | |
1946 | * If the value tree got partially truncated, we now have a corrupted | |
1947 | * extended attribute. We're going to wipe its entry and leak the | |
1948 | * clusters. Better to leak some storage than leave a corrupt entry. | |
1949 | * | |
1950 | * If the value tree grew, it obviously didn't grow enough for the | |
1951 | * new entry. We're not going to try and reclaim those clusters either. | |
1952 | * If there was already an external value there (orig_clusters != 0), | |
1953 | * the new clusters are attached safely and we can just leave the old | |
1954 | * value in place. If there was no external value there, we remove | |
1955 | * the entry. | |
1956 | * | |
1957 | * This way, the xattr block we store in the journal will be consistent. | |
1958 | * If the size change broke because of the journal, no changes will hit | |
1959 | * disk anyway. | |
1960 | */ | |
1961 | static void ocfs2_xa_cleanup_value_truncate(struct ocfs2_xa_loc *loc, | |
1962 | const char *what, | |
1963 | unsigned int orig_clusters) | |
1964 | { | |
1965 | unsigned int new_clusters = ocfs2_xa_value_clusters(loc); | |
1966 | char *nameval_buf = ocfs2_xa_offset_pointer(loc, | |
1967 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
1968 | ||
1969 | if (new_clusters < orig_clusters) { | |
1970 | mlog(ML_ERROR, | |
1971 | "Partial truncate while %s xattr %.*s. Leaking " | |
1972 | "%u clusters and removing the entry\n", | |
1973 | what, loc->xl_entry->xe_name_len, nameval_buf, | |
1974 | orig_clusters - new_clusters); | |
1975 | ocfs2_xa_remove_entry(loc); | |
1976 | } else if (!orig_clusters) { | |
1977 | mlog(ML_ERROR, | |
1978 | "Unable to allocate an external value for xattr " | |
1979 | "%.*s safely. Leaking %u clusters and removing the " | |
1980 | "entry\n", | |
1981 | loc->xl_entry->xe_name_len, nameval_buf, | |
1982 | new_clusters - orig_clusters); | |
1983 | ocfs2_xa_remove_entry(loc); | |
1984 | } else if (new_clusters > orig_clusters) | |
1985 | mlog(ML_ERROR, | |
1986 | "Unable to grow xattr %.*s safely. %u new clusters " | |
1987 | "have been added, but the value will not be " | |
1988 | "modified\n", | |
1989 | loc->xl_entry->xe_name_len, nameval_buf, | |
1990 | new_clusters - orig_clusters); | |
1991 | } | |
1992 | ||
73857ee0 JB |
1993 | static int ocfs2_xa_remove(struct ocfs2_xa_loc *loc, |
1994 | struct ocfs2_xattr_set_ctxt *ctxt) | |
1995 | { | |
1996 | int rc = 0; | |
399ff3a7 | 1997 | unsigned int orig_clusters; |
73857ee0 JB |
1998 | |
1999 | if (!ocfs2_xattr_is_local(loc->xl_entry)) { | |
399ff3a7 | 2000 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2001 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); |
2002 | if (rc) { | |
2003 | mlog_errno(rc); | |
399ff3a7 JB |
2004 | /* |
2005 | * Since this is remove, we can return 0 if | |
2006 | * ocfs2_xa_cleanup_value_truncate() is going to | |
2007 | * wipe the entry anyway. So we check the | |
2008 | * cluster count as well. | |
2009 | */ | |
2010 | if (orig_clusters != ocfs2_xa_value_clusters(loc)) | |
2011 | rc = 0; | |
2012 | ocfs2_xa_cleanup_value_truncate(loc, "removing", | |
2013 | orig_clusters); | |
2014 | if (rc) | |
2015 | goto out; | |
73857ee0 JB |
2016 | } |
2017 | } | |
2018 | ||
2019 | ocfs2_xa_remove_entry(loc); | |
2020 | ||
2021 | out: | |
2022 | return rc; | |
2023 | } | |
2024 | ||
2025 | static void ocfs2_xa_install_value_root(struct ocfs2_xa_loc *loc) | |
2026 | { | |
2027 | int name_size = OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len); | |
2028 | char *nameval_buf; | |
2029 | ||
2030 | nameval_buf = ocfs2_xa_offset_pointer(loc, | |
2031 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
2032 | memcpy(nameval_buf + name_size, &def_xv, OCFS2_XATTR_ROOT_SIZE); | |
2033 | } | |
2034 | ||
2035 | /* | |
2036 | * Take an existing entry and make it ready for the new value. This | |
2037 | * won't allocate space, but it may free space. It should be ready for | |
2038 | * ocfs2_xa_prepare_entry() to finish the work. | |
2039 | */ | |
2040 | static int ocfs2_xa_reuse_entry(struct ocfs2_xa_loc *loc, | |
2041 | struct ocfs2_xattr_info *xi, | |
2042 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2043 | { | |
2044 | int rc = 0; | |
2045 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
399ff3a7 | 2046 | unsigned int orig_clusters; |
73857ee0 JB |
2047 | char *nameval_buf; |
2048 | int xe_local = ocfs2_xattr_is_local(loc->xl_entry); | |
2049 | int xi_local = xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE; | |
2050 | ||
2051 | BUG_ON(OCFS2_XATTR_SIZE(loc->xl_entry->xe_name_len) != | |
2052 | name_size); | |
2053 | ||
2054 | nameval_buf = ocfs2_xa_offset_pointer(loc, | |
2055 | le16_to_cpu(loc->xl_entry->xe_name_offset)); | |
2056 | if (xe_local) { | |
2057 | memset(nameval_buf + name_size, 0, | |
2058 | namevalue_size_xe(loc->xl_entry) - name_size); | |
2059 | if (!xi_local) | |
2060 | ocfs2_xa_install_value_root(loc); | |
2061 | } else { | |
399ff3a7 | 2062 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2063 | if (xi_local) { |
2064 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); | |
399ff3a7 | 2065 | if (rc < 0) |
73857ee0 | 2066 | mlog_errno(rc); |
399ff3a7 JB |
2067 | else |
2068 | memset(nameval_buf + name_size, 0, | |
2069 | namevalue_size_xe(loc->xl_entry) - | |
2070 | name_size); | |
73857ee0 JB |
2071 | } else if (le64_to_cpu(loc->xl_entry->xe_value_size) > |
2072 | xi->xi_value_len) { | |
2073 | rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, | |
2074 | ctxt); | |
399ff3a7 | 2075 | if (rc < 0) |
73857ee0 | 2076 | mlog_errno(rc); |
399ff3a7 JB |
2077 | } |
2078 | ||
2079 | if (rc) { | |
2080 | ocfs2_xa_cleanup_value_truncate(loc, "reusing", | |
2081 | orig_clusters); | |
2082 | goto out; | |
73857ee0 JB |
2083 | } |
2084 | } | |
2085 | ||
2086 | loc->xl_entry->xe_value_size = cpu_to_le64(xi->xi_value_len); | |
2087 | ocfs2_xattr_set_local(loc->xl_entry, xi_local); | |
2088 | ||
2089 | out: | |
2090 | return rc; | |
2091 | } | |
2092 | ||
69a3e539 JB |
2093 | /* |
2094 | * Prepares loc->xl_entry to receive the new xattr. This includes | |
2095 | * properly setting up the name+value pair region. If loc->xl_entry | |
2096 | * already exists, it will take care of modifying it appropriately. | |
69a3e539 JB |
2097 | * |
2098 | * Note that this modifies the data. You did journal_access already, | |
2099 | * right? | |
2100 | */ | |
2101 | static int ocfs2_xa_prepare_entry(struct ocfs2_xa_loc *loc, | |
2102 | struct ocfs2_xattr_info *xi, | |
73857ee0 JB |
2103 | u32 name_hash, |
2104 | struct ocfs2_xattr_set_ctxt *ctxt) | |
69a3e539 JB |
2105 | { |
2106 | int rc = 0; | |
399ff3a7 JB |
2107 | unsigned int orig_clusters; |
2108 | __le64 orig_value_size = 0; | |
69a3e539 | 2109 | |
69a3e539 JB |
2110 | rc = ocfs2_xa_check_space(loc, xi); |
2111 | if (rc) | |
2112 | goto out; | |
2113 | ||
2114 | if (loc->xl_entry) { | |
2115 | if (ocfs2_xa_can_reuse_entry(loc, xi)) { | |
399ff3a7 | 2116 | orig_value_size = loc->xl_entry->xe_value_size; |
73857ee0 JB |
2117 | rc = ocfs2_xa_reuse_entry(loc, xi, ctxt); |
2118 | if (rc) | |
2119 | goto out; | |
2120 | goto alloc_value; | |
69a3e539 JB |
2121 | } |
2122 | ||
73857ee0 | 2123 | if (!ocfs2_xattr_is_local(loc->xl_entry)) { |
399ff3a7 | 2124 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 JB |
2125 | rc = ocfs2_xa_value_truncate(loc, 0, ctxt); |
2126 | if (rc) { | |
2127 | mlog_errno(rc); | |
399ff3a7 JB |
2128 | ocfs2_xa_cleanup_value_truncate(loc, |
2129 | "overwriting", | |
2130 | orig_clusters); | |
73857ee0 JB |
2131 | goto out; |
2132 | } | |
2133 | } | |
69a3e539 JB |
2134 | ocfs2_xa_wipe_namevalue(loc); |
2135 | } else | |
2136 | ocfs2_xa_add_entry(loc, name_hash); | |
2137 | ||
2138 | /* | |
2139 | * If we get here, we have a blank entry. Fill it. We grow our | |
2140 | * name+value pair back from the end. | |
2141 | */ | |
2142 | ocfs2_xa_add_namevalue(loc, xi); | |
73857ee0 JB |
2143 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
2144 | ocfs2_xa_install_value_root(loc); | |
2145 | ||
2146 | alloc_value: | |
2147 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { | |
399ff3a7 | 2148 | orig_clusters = ocfs2_xa_value_clusters(loc); |
73857ee0 | 2149 | rc = ocfs2_xa_value_truncate(loc, xi->xi_value_len, ctxt); |
399ff3a7 | 2150 | if (rc < 0) { |
5f5261ac | 2151 | ctxt->set_abort = 1; |
399ff3a7 JB |
2152 | ocfs2_xa_cleanup_value_truncate(loc, "growing", |
2153 | orig_clusters); | |
d5a7df06 TM |
2154 | /* |
2155 | * If we were growing an existing value, | |
2156 | * ocfs2_xa_cleanup_value_truncate() won't remove | |
2157 | * the entry. We need to restore the original value | |
2158 | * size. | |
2159 | */ | |
2160 | if (loc->xl_entry) { | |
2161 | BUG_ON(!orig_value_size); | |
2162 | loc->xl_entry->xe_value_size = orig_value_size; | |
2163 | } | |
73857ee0 | 2164 | mlog_errno(rc); |
399ff3a7 | 2165 | } |
73857ee0 | 2166 | } |
69a3e539 JB |
2167 | |
2168 | out: | |
2169 | return rc; | |
2170 | } | |
2171 | ||
2172 | /* | |
73857ee0 JB |
2173 | * Store the value portion of the name+value pair. This will skip |
2174 | * values that are stored externally. Their tree roots were set up | |
2175 | * by ocfs2_xa_prepare_entry(). | |
69a3e539 | 2176 | */ |
73857ee0 JB |
2177 | static int ocfs2_xa_store_value(struct ocfs2_xa_loc *loc, |
2178 | struct ocfs2_xattr_info *xi, | |
2179 | struct ocfs2_xattr_set_ctxt *ctxt) | |
69a3e539 | 2180 | { |
73857ee0 | 2181 | int rc = 0; |
69a3e539 JB |
2182 | int nameval_offset = le16_to_cpu(loc->xl_entry->xe_name_offset); |
2183 | int name_size = OCFS2_XATTR_SIZE(xi->xi_name_len); | |
69a3e539 | 2184 | char *nameval_buf; |
73857ee0 | 2185 | struct ocfs2_xattr_value_buf vb; |
69a3e539 | 2186 | |
69a3e539 | 2187 | nameval_buf = ocfs2_xa_offset_pointer(loc, nameval_offset); |
73857ee0 JB |
2188 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
2189 | ocfs2_xa_fill_value_buf(loc, &vb); | |
2190 | rc = __ocfs2_xattr_set_value_outside(loc->xl_inode, | |
2191 | ctxt->handle, &vb, | |
2192 | xi->xi_value, | |
2193 | xi->xi_value_len); | |
2194 | } else | |
2195 | memcpy(nameval_buf + name_size, xi->xi_value, xi->xi_value_len); | |
2196 | ||
73857ee0 | 2197 | return rc; |
69a3e539 JB |
2198 | } |
2199 | ||
bca5e9bd JB |
2200 | static int ocfs2_xa_set(struct ocfs2_xa_loc *loc, |
2201 | struct ocfs2_xattr_info *xi, | |
2202 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2203 | { | |
2204 | int ret; | |
2205 | u32 name_hash = ocfs2_xattr_name_hash(loc->xl_inode, xi->xi_name, | |
2206 | xi->xi_name_len); | |
2207 | ||
2208 | ret = ocfs2_xa_journal_access(ctxt->handle, loc, | |
2209 | OCFS2_JOURNAL_ACCESS_WRITE); | |
2210 | if (ret) { | |
2211 | mlog_errno(ret); | |
2212 | goto out; | |
2213 | } | |
2214 | ||
399ff3a7 JB |
2215 | /* |
2216 | * From here on out, everything is going to modify the buffer a | |
2217 | * little. Errors are going to leave the xattr header in a | |
2218 | * sane state. Thus, even with errors we dirty the sucker. | |
2219 | */ | |
2220 | ||
bca5e9bd JB |
2221 | /* Don't worry, we are never called with !xi_value and !xl_entry */ |
2222 | if (!xi->xi_value) { | |
2223 | ret = ocfs2_xa_remove(loc, ctxt); | |
399ff3a7 | 2224 | goto out_dirty; |
bca5e9bd JB |
2225 | } |
2226 | ||
2227 | ret = ocfs2_xa_prepare_entry(loc, xi, name_hash, ctxt); | |
2228 | if (ret) { | |
2229 | if (ret != -ENOSPC) | |
2230 | mlog_errno(ret); | |
399ff3a7 | 2231 | goto out_dirty; |
bca5e9bd JB |
2232 | } |
2233 | ||
2234 | ret = ocfs2_xa_store_value(loc, xi, ctxt); | |
399ff3a7 | 2235 | if (ret) |
bca5e9bd | 2236 | mlog_errno(ret); |
bca5e9bd | 2237 | |
399ff3a7 | 2238 | out_dirty: |
bca5e9bd JB |
2239 | ocfs2_xa_journal_dirty(ctxt->handle, loc); |
2240 | ||
2241 | out: | |
2242 | return ret; | |
2243 | } | |
2244 | ||
11179f2c JB |
2245 | static void ocfs2_init_dinode_xa_loc(struct ocfs2_xa_loc *loc, |
2246 | struct inode *inode, | |
2247 | struct buffer_head *bh, | |
2248 | struct ocfs2_xattr_entry *entry) | |
2249 | { | |
2250 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
2251 | ||
139fface JB |
2252 | BUG_ON(!(OCFS2_I(inode)->ip_dyn_features & OCFS2_INLINE_XATTR_FL)); |
2253 | ||
cf2bc809 | 2254 | loc->xl_inode = inode; |
11179f2c JB |
2255 | loc->xl_ops = &ocfs2_xa_block_loc_ops; |
2256 | loc->xl_storage = bh; | |
2257 | loc->xl_entry = entry; | |
139fface | 2258 | loc->xl_size = le16_to_cpu(di->i_xattr_inline_size); |
11179f2c JB |
2259 | loc->xl_header = |
2260 | (struct ocfs2_xattr_header *)(bh->b_data + bh->b_size - | |
2261 | loc->xl_size); | |
2262 | } | |
2263 | ||
2264 | static void ocfs2_init_xattr_block_xa_loc(struct ocfs2_xa_loc *loc, | |
cf2bc809 | 2265 | struct inode *inode, |
11179f2c JB |
2266 | struct buffer_head *bh, |
2267 | struct ocfs2_xattr_entry *entry) | |
2268 | { | |
2269 | struct ocfs2_xattr_block *xb = | |
2270 | (struct ocfs2_xattr_block *)bh->b_data; | |
2271 | ||
2272 | BUG_ON(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED); | |
2273 | ||
cf2bc809 | 2274 | loc->xl_inode = inode; |
11179f2c JB |
2275 | loc->xl_ops = &ocfs2_xa_block_loc_ops; |
2276 | loc->xl_storage = bh; | |
2277 | loc->xl_header = &(xb->xb_attrs.xb_header); | |
2278 | loc->xl_entry = entry; | |
2279 | loc->xl_size = bh->b_size - offsetof(struct ocfs2_xattr_block, | |
2280 | xb_attrs.xb_header); | |
2281 | } | |
2282 | ||
2283 | static void ocfs2_init_xattr_bucket_xa_loc(struct ocfs2_xa_loc *loc, | |
2284 | struct ocfs2_xattr_bucket *bucket, | |
2285 | struct ocfs2_xattr_entry *entry) | |
2286 | { | |
cf2bc809 | 2287 | loc->xl_inode = bucket->bu_inode; |
11179f2c JB |
2288 | loc->xl_ops = &ocfs2_xa_bucket_loc_ops; |
2289 | loc->xl_storage = bucket; | |
2290 | loc->xl_header = bucket_xh(bucket); | |
2291 | loc->xl_entry = entry; | |
2292 | loc->xl_size = OCFS2_XATTR_BUCKET_SIZE; | |
2293 | } | |
2294 | ||
ce9c5a54 TM |
2295 | /* |
2296 | * In xattr remove, if it is stored outside and refcounted, we may have | |
2297 | * the chance to split the refcount tree. So need the allocators. | |
2298 | */ | |
2299 | static int ocfs2_lock_xattr_remove_allocators(struct inode *inode, | |
2300 | struct ocfs2_xattr_value_root *xv, | |
2301 | struct ocfs2_caching_info *ref_ci, | |
2302 | struct buffer_head *ref_root_bh, | |
2303 | struct ocfs2_alloc_context **meta_ac, | |
2304 | int *ref_credits) | |
2305 | { | |
2306 | int ret, meta_add = 0; | |
2307 | u32 p_cluster, num_clusters; | |
2308 | unsigned int ext_flags; | |
2309 | ||
2310 | *ref_credits = 0; | |
2311 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
2312 | &num_clusters, | |
2313 | &xv->xr_list, | |
2314 | &ext_flags); | |
2315 | if (ret) { | |
2316 | mlog_errno(ret); | |
2317 | goto out; | |
2318 | } | |
2319 | ||
2320 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
2321 | goto out; | |
2322 | ||
2323 | ret = ocfs2_refcounted_xattr_delete_need(inode, ref_ci, | |
2324 | ref_root_bh, xv, | |
2325 | &meta_add, ref_credits); | |
2326 | if (ret) { | |
2327 | mlog_errno(ret); | |
2328 | goto out; | |
2329 | } | |
2330 | ||
2331 | ret = ocfs2_reserve_new_metadata_blocks(OCFS2_SB(inode->i_sb), | |
2332 | meta_add, meta_ac); | |
2333 | if (ret) | |
2334 | mlog_errno(ret); | |
2335 | ||
2336 | out: | |
2337 | return ret; | |
2338 | } | |
2339 | ||
cf1d6c76 | 2340 | static int ocfs2_remove_value_outside(struct inode*inode, |
4311901d | 2341 | struct ocfs2_xattr_value_buf *vb, |
ce9c5a54 TM |
2342 | struct ocfs2_xattr_header *header, |
2343 | struct ocfs2_caching_info *ref_ci, | |
2344 | struct buffer_head *ref_root_bh) | |
cf1d6c76 | 2345 | { |
ce9c5a54 | 2346 | int ret = 0, i, ref_credits; |
78f30c31 TM |
2347 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
2348 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, }; | |
ce9c5a54 | 2349 | void *val; |
78f30c31 TM |
2350 | |
2351 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
cf1d6c76 TY |
2352 | |
2353 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
2354 | struct ocfs2_xattr_entry *entry = &header->xh_entries[i]; | |
2355 | ||
ce9c5a54 TM |
2356 | if (ocfs2_xattr_is_local(entry)) |
2357 | continue; | |
cf1d6c76 | 2358 | |
ce9c5a54 TM |
2359 | val = (void *)header + |
2360 | le16_to_cpu(entry->xe_name_offset); | |
2361 | vb->vb_xv = (struct ocfs2_xattr_value_root *) | |
2362 | (val + OCFS2_XATTR_SIZE(entry->xe_name_len)); | |
2363 | ||
2364 | ret = ocfs2_lock_xattr_remove_allocators(inode, vb->vb_xv, | |
2365 | ref_ci, ref_root_bh, | |
2366 | &ctxt.meta_ac, | |
2367 | &ref_credits); | |
2368 | ||
2369 | ctxt.handle = ocfs2_start_trans(osb, ref_credits + | |
2370 | ocfs2_remove_extent_credits(osb->sb)); | |
2371 | if (IS_ERR(ctxt.handle)) { | |
2372 | ret = PTR_ERR(ctxt.handle); | |
2373 | mlog_errno(ret); | |
2374 | break; | |
2375 | } | |
2376 | ||
2377 | ret = ocfs2_xattr_value_truncate(inode, vb, 0, &ctxt); | |
ce9c5a54 TM |
2378 | |
2379 | ocfs2_commit_trans(osb, ctxt.handle); | |
2380 | if (ctxt.meta_ac) { | |
2381 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
2382 | ctxt.meta_ac = NULL; | |
cf1d6c76 | 2383 | } |
b8a0ae57 WW |
2384 | |
2385 | if (ret < 0) { | |
2386 | mlog_errno(ret); | |
2387 | break; | |
2388 | } | |
2389 | ||
cf1d6c76 TY |
2390 | } |
2391 | ||
ce9c5a54 TM |
2392 | if (ctxt.meta_ac) |
2393 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
2394 | ocfs2_schedule_truncate_log_flush(osb, 1); |
2395 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
cf1d6c76 TY |
2396 | return ret; |
2397 | } | |
2398 | ||
2399 | static int ocfs2_xattr_ibody_remove(struct inode *inode, | |
ce9c5a54 TM |
2400 | struct buffer_head *di_bh, |
2401 | struct ocfs2_caching_info *ref_ci, | |
2402 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2403 | { |
2404 | ||
2405 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2406 | struct ocfs2_xattr_header *header; | |
2407 | int ret; | |
4311901d JB |
2408 | struct ocfs2_xattr_value_buf vb = { |
2409 | .vb_bh = di_bh, | |
2410 | .vb_access = ocfs2_journal_access_di, | |
2411 | }; | |
cf1d6c76 TY |
2412 | |
2413 | header = (struct ocfs2_xattr_header *) | |
2414 | ((void *)di + inode->i_sb->s_blocksize - | |
2415 | le16_to_cpu(di->i_xattr_inline_size)); | |
2416 | ||
ce9c5a54 TM |
2417 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2418 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2419 | |
2420 | return ret; | |
2421 | } | |
2422 | ||
ce9c5a54 TM |
2423 | struct ocfs2_rm_xattr_bucket_para { |
2424 | struct ocfs2_caching_info *ref_ci; | |
2425 | struct buffer_head *ref_root_bh; | |
2426 | }; | |
2427 | ||
cf1d6c76 | 2428 | static int ocfs2_xattr_block_remove(struct inode *inode, |
ce9c5a54 TM |
2429 | struct buffer_head *blk_bh, |
2430 | struct ocfs2_caching_info *ref_ci, | |
2431 | struct buffer_head *ref_root_bh) | |
cf1d6c76 TY |
2432 | { |
2433 | struct ocfs2_xattr_block *xb; | |
cf1d6c76 | 2434 | int ret = 0; |
4311901d JB |
2435 | struct ocfs2_xattr_value_buf vb = { |
2436 | .vb_bh = blk_bh, | |
2437 | .vb_access = ocfs2_journal_access_xb, | |
2438 | }; | |
ce9c5a54 TM |
2439 | struct ocfs2_rm_xattr_bucket_para args = { |
2440 | .ref_ci = ref_ci, | |
2441 | .ref_root_bh = ref_root_bh, | |
2442 | }; | |
cf1d6c76 TY |
2443 | |
2444 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; | |
a3944256 TM |
2445 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { |
2446 | struct ocfs2_xattr_header *header = &(xb->xb_attrs.xb_header); | |
ce9c5a54 TM |
2447 | ret = ocfs2_remove_value_outside(inode, &vb, header, |
2448 | ref_ci, ref_root_bh); | |
a3944256 | 2449 | } else |
47bca495 TM |
2450 | ret = ocfs2_iterate_xattr_index_block(inode, |
2451 | blk_bh, | |
2452 | ocfs2_rm_xattr_cluster, | |
ce9c5a54 | 2453 | &args); |
cf1d6c76 TY |
2454 | |
2455 | return ret; | |
2456 | } | |
2457 | ||
08413899 | 2458 | static int ocfs2_xattr_free_block(struct inode *inode, |
ce9c5a54 TM |
2459 | u64 block, |
2460 | struct ocfs2_caching_info *ref_ci, | |
2461 | struct buffer_head *ref_root_bh) | |
08413899 TM |
2462 | { |
2463 | struct inode *xb_alloc_inode; | |
2464 | struct buffer_head *xb_alloc_bh = NULL; | |
2465 | struct buffer_head *blk_bh = NULL; | |
2466 | struct ocfs2_xattr_block *xb; | |
2467 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2468 | handle_t *handle; | |
2469 | int ret = 0; | |
2470 | u64 blk, bg_blkno; | |
2471 | u16 bit; | |
2472 | ||
4ae1d69b | 2473 | ret = ocfs2_read_xattr_block(inode, block, &blk_bh); |
08413899 TM |
2474 | if (ret < 0) { |
2475 | mlog_errno(ret); | |
2476 | goto out; | |
2477 | } | |
2478 | ||
ce9c5a54 | 2479 | ret = ocfs2_xattr_block_remove(inode, blk_bh, ref_ci, ref_root_bh); |
08413899 TM |
2480 | if (ret < 0) { |
2481 | mlog_errno(ret); | |
2482 | goto out; | |
2483 | } | |
2484 | ||
4ae1d69b | 2485 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
08413899 TM |
2486 | blk = le64_to_cpu(xb->xb_blkno); |
2487 | bit = le16_to_cpu(xb->xb_suballoc_bit); | |
74380c47 TM |
2488 | if (xb->xb_suballoc_loc) |
2489 | bg_blkno = le64_to_cpu(xb->xb_suballoc_loc); | |
2490 | else | |
2491 | bg_blkno = ocfs2_which_suballoc_group(blk, bit); | |
08413899 TM |
2492 | |
2493 | xb_alloc_inode = ocfs2_get_system_file_inode(osb, | |
2494 | EXTENT_ALLOC_SYSTEM_INODE, | |
2495 | le16_to_cpu(xb->xb_suballoc_slot)); | |
2496 | if (!xb_alloc_inode) { | |
2497 | ret = -ENOMEM; | |
2498 | mlog_errno(ret); | |
2499 | goto out; | |
2500 | } | |
2501 | mutex_lock(&xb_alloc_inode->i_mutex); | |
2502 | ||
2503 | ret = ocfs2_inode_lock(xb_alloc_inode, &xb_alloc_bh, 1); | |
2504 | if (ret < 0) { | |
2505 | mlog_errno(ret); | |
2506 | goto out_mutex; | |
2507 | } | |
2508 | ||
2509 | handle = ocfs2_start_trans(osb, OCFS2_SUBALLOC_FREE); | |
2510 | if (IS_ERR(handle)) { | |
2511 | ret = PTR_ERR(handle); | |
2512 | mlog_errno(ret); | |
2513 | goto out_unlock; | |
2514 | } | |
2515 | ||
2516 | ret = ocfs2_free_suballoc_bits(handle, xb_alloc_inode, xb_alloc_bh, | |
2517 | bit, bg_blkno, 1); | |
2518 | if (ret < 0) | |
2519 | mlog_errno(ret); | |
2520 | ||
2521 | ocfs2_commit_trans(osb, handle); | |
2522 | out_unlock: | |
2523 | ocfs2_inode_unlock(xb_alloc_inode, 1); | |
2524 | brelse(xb_alloc_bh); | |
2525 | out_mutex: | |
2526 | mutex_unlock(&xb_alloc_inode->i_mutex); | |
2527 | iput(xb_alloc_inode); | |
2528 | out: | |
2529 | brelse(blk_bh); | |
2530 | return ret; | |
2531 | } | |
2532 | ||
cf1d6c76 TY |
2533 | /* |
2534 | * ocfs2_xattr_remove() | |
2535 | * | |
2536 | * Free extended attribute resources associated with this inode. | |
2537 | */ | |
2538 | int ocfs2_xattr_remove(struct inode *inode, struct buffer_head *di_bh) | |
2539 | { | |
cf1d6c76 TY |
2540 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
2541 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
ce9c5a54 TM |
2542 | struct ocfs2_refcount_tree *ref_tree = NULL; |
2543 | struct buffer_head *ref_root_bh = NULL; | |
2544 | struct ocfs2_caching_info *ref_ci = NULL; | |
cf1d6c76 TY |
2545 | handle_t *handle; |
2546 | int ret; | |
2547 | ||
8154da3d TY |
2548 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
2549 | return 0; | |
2550 | ||
cf1d6c76 TY |
2551 | if (!(oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) |
2552 | return 0; | |
2553 | ||
ce9c5a54 TM |
2554 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL) { |
2555 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(inode->i_sb), | |
2556 | le64_to_cpu(di->i_refcount_loc), | |
2557 | 1, &ref_tree, &ref_root_bh); | |
2558 | if (ret) { | |
2559 | mlog_errno(ret); | |
2560 | goto out; | |
2561 | } | |
2562 | ref_ci = &ref_tree->rf_ci; | |
2563 | ||
2564 | } | |
2565 | ||
cf1d6c76 | 2566 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { |
ce9c5a54 TM |
2567 | ret = ocfs2_xattr_ibody_remove(inode, di_bh, |
2568 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2569 | if (ret < 0) { |
2570 | mlog_errno(ret); | |
2571 | goto out; | |
2572 | } | |
2573 | } | |
cf1d6c76 | 2574 | |
08413899 TM |
2575 | if (di->i_xattr_loc) { |
2576 | ret = ocfs2_xattr_free_block(inode, | |
ce9c5a54 TM |
2577 | le64_to_cpu(di->i_xattr_loc), |
2578 | ref_ci, ref_root_bh); | |
cf1d6c76 TY |
2579 | if (ret < 0) { |
2580 | mlog_errno(ret); | |
2581 | goto out; | |
2582 | } | |
2583 | } | |
2584 | ||
2585 | handle = ocfs2_start_trans((OCFS2_SB(inode->i_sb)), | |
2586 | OCFS2_INODE_UPDATE_CREDITS); | |
2587 | if (IS_ERR(handle)) { | |
2588 | ret = PTR_ERR(handle); | |
2589 | mlog_errno(ret); | |
2590 | goto out; | |
2591 | } | |
0cf2f763 | 2592 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, |
84008972 | 2593 | OCFS2_JOURNAL_ACCESS_WRITE); |
cf1d6c76 TY |
2594 | if (ret) { |
2595 | mlog_errno(ret); | |
2596 | goto out_commit; | |
2597 | } | |
2598 | ||
08413899 | 2599 | di->i_xattr_loc = 0; |
cf1d6c76 TY |
2600 | |
2601 | spin_lock(&oi->ip_lock); | |
2602 | oi->ip_dyn_features &= ~(OCFS2_INLINE_XATTR_FL | OCFS2_HAS_XATTR_FL); | |
2603 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2604 | spin_unlock(&oi->ip_lock); | |
2605 | ||
ec20cec7 | 2606 | ocfs2_journal_dirty(handle, di_bh); |
cf1d6c76 TY |
2607 | out_commit: |
2608 | ocfs2_commit_trans(OCFS2_SB(inode->i_sb), handle); | |
2609 | out: | |
ce9c5a54 TM |
2610 | if (ref_tree) |
2611 | ocfs2_unlock_refcount_tree(OCFS2_SB(inode->i_sb), ref_tree, 1); | |
2612 | brelse(ref_root_bh); | |
cf1d6c76 TY |
2613 | return ret; |
2614 | } | |
2615 | ||
2616 | static int ocfs2_xattr_has_space_inline(struct inode *inode, | |
2617 | struct ocfs2_dinode *di) | |
2618 | { | |
2619 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2620 | unsigned int xattrsize = OCFS2_SB(inode->i_sb)->s_xattr_inline_size; | |
2621 | int free; | |
2622 | ||
2623 | if (xattrsize < OCFS2_MIN_XATTR_INLINE_SIZE) | |
2624 | return 0; | |
2625 | ||
2626 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2627 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2628 | free = le16_to_cpu(idata->id_count) - le64_to_cpu(di->i_size); | |
2629 | } else if (ocfs2_inode_is_fast_symlink(inode)) { | |
2630 | free = ocfs2_fast_symlink_chars(inode->i_sb) - | |
2631 | le64_to_cpu(di->i_size); | |
2632 | } else { | |
2633 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2634 | free = (le16_to_cpu(el->l_count) - | |
2635 | le16_to_cpu(el->l_next_free_rec)) * | |
2636 | sizeof(struct ocfs2_extent_rec); | |
2637 | } | |
2638 | if (free >= xattrsize) | |
2639 | return 1; | |
2640 | ||
2641 | return 0; | |
2642 | } | |
2643 | ||
2644 | /* | |
2645 | * ocfs2_xattr_ibody_find() | |
2646 | * | |
2647 | * Find extended attribute in inode block and | |
2648 | * fill search info into struct ocfs2_xattr_search. | |
2649 | */ | |
2650 | static int ocfs2_xattr_ibody_find(struct inode *inode, | |
2651 | int name_index, | |
2652 | const char *name, | |
2653 | struct ocfs2_xattr_search *xs) | |
2654 | { | |
2655 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2656 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2657 | int ret; | |
2658 | int has_space = 0; | |
2659 | ||
2660 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2661 | return 0; | |
2662 | ||
2663 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { | |
2664 | down_read(&oi->ip_alloc_sem); | |
2665 | has_space = ocfs2_xattr_has_space_inline(inode, di); | |
2666 | up_read(&oi->ip_alloc_sem); | |
2667 | if (!has_space) | |
2668 | return 0; | |
2669 | } | |
2670 | ||
2671 | xs->xattr_bh = xs->inode_bh; | |
2672 | xs->end = (void *)di + inode->i_sb->s_blocksize; | |
2673 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) | |
2674 | xs->header = (struct ocfs2_xattr_header *) | |
2675 | (xs->end - le16_to_cpu(di->i_xattr_inline_size)); | |
2676 | else | |
2677 | xs->header = (struct ocfs2_xattr_header *) | |
2678 | (xs->end - OCFS2_SB(inode->i_sb)->s_xattr_inline_size); | |
2679 | xs->base = (void *)xs->header; | |
2680 | xs->here = xs->header->xh_entries; | |
2681 | ||
2682 | /* Find the named attribute. */ | |
2683 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
2684 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2685 | if (ret && ret != -ENODATA) | |
2686 | return ret; | |
2687 | xs->not_found = ret; | |
2688 | } | |
2689 | ||
2690 | return 0; | |
2691 | } | |
2692 | ||
139fface JB |
2693 | static int ocfs2_xattr_ibody_init(struct inode *inode, |
2694 | struct buffer_head *di_bh, | |
2695 | struct ocfs2_xattr_set_ctxt *ctxt) | |
2696 | { | |
2697 | int ret; | |
2698 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
2699 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)di_bh->b_data; | |
2700 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
2701 | unsigned int xattrsize = osb->s_xattr_inline_size; | |
2702 | ||
2703 | if (!ocfs2_xattr_has_space_inline(inode, di)) { | |
2704 | ret = -ENOSPC; | |
2705 | goto out; | |
2706 | } | |
2707 | ||
2708 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), di_bh, | |
2709 | OCFS2_JOURNAL_ACCESS_WRITE); | |
2710 | if (ret) { | |
2711 | mlog_errno(ret); | |
2712 | goto out; | |
2713 | } | |
2714 | ||
2715 | /* | |
2716 | * Adjust extent record count or inline data size | |
2717 | * to reserve space for extended attribute. | |
2718 | */ | |
2719 | if (oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) { | |
2720 | struct ocfs2_inline_data *idata = &di->id2.i_data; | |
2721 | le16_add_cpu(&idata->id_count, -xattrsize); | |
2722 | } else if (!(ocfs2_inode_is_fast_symlink(inode))) { | |
2723 | struct ocfs2_extent_list *el = &di->id2.i_list; | |
2724 | le16_add_cpu(&el->l_count, -(xattrsize / | |
2725 | sizeof(struct ocfs2_extent_rec))); | |
2726 | } | |
2727 | di->i_xattr_inline_size = cpu_to_le16(xattrsize); | |
2728 | ||
2729 | spin_lock(&oi->ip_lock); | |
2730 | oi->ip_dyn_features |= OCFS2_INLINE_XATTR_FL|OCFS2_HAS_XATTR_FL; | |
2731 | di->i_dyn_features = cpu_to_le16(oi->ip_dyn_features); | |
2732 | spin_unlock(&oi->ip_lock); | |
2733 | ||
ec20cec7 | 2734 | ocfs2_journal_dirty(ctxt->handle, di_bh); |
139fface JB |
2735 | |
2736 | out: | |
2737 | return ret; | |
2738 | } | |
2739 | ||
cf1d6c76 TY |
2740 | /* |
2741 | * ocfs2_xattr_ibody_set() | |
2742 | * | |
2743 | * Set, replace or remove an extended attribute into inode block. | |
2744 | * | |
2745 | */ | |
2746 | static int ocfs2_xattr_ibody_set(struct inode *inode, | |
2747 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2748 | struct ocfs2_xattr_search *xs, |
2749 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 | 2750 | { |
139fface | 2751 | int ret; |
cf1d6c76 | 2752 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
139fface | 2753 | struct ocfs2_xa_loc loc; |
cf1d6c76 TY |
2754 | |
2755 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) | |
2756 | return -ENOSPC; | |
2757 | ||
2758 | down_write(&oi->ip_alloc_sem); | |
139fface JB |
2759 | if (!(oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL)) { |
2760 | ret = ocfs2_xattr_ibody_init(inode, xs->inode_bh, ctxt); | |
2761 | if (ret) { | |
2762 | if (ret != -ENOSPC) | |
2763 | mlog_errno(ret); | |
2764 | goto out; | |
2765 | } | |
2766 | } | |
2767 | ||
2768 | ocfs2_init_dinode_xa_loc(&loc, inode, xs->inode_bh, | |
2769 | xs->not_found ? NULL : xs->here); | |
2770 | ret = ocfs2_xa_set(&loc, xi, ctxt); | |
2771 | if (ret) { | |
2772 | if (ret != -ENOSPC) | |
2773 | mlog_errno(ret); | |
2774 | goto out; | |
2775 | } | |
2776 | xs->here = loc.xl_entry; | |
2777 | ||
cf1d6c76 TY |
2778 | out: |
2779 | up_write(&oi->ip_alloc_sem); | |
2780 | ||
2781 | return ret; | |
2782 | } | |
2783 | ||
2784 | /* | |
2785 | * ocfs2_xattr_block_find() | |
2786 | * | |
2787 | * Find extended attribute in external block and | |
2788 | * fill search info into struct ocfs2_xattr_search. | |
2789 | */ | |
2790 | static int ocfs2_xattr_block_find(struct inode *inode, | |
2791 | int name_index, | |
2792 | const char *name, | |
2793 | struct ocfs2_xattr_search *xs) | |
2794 | { | |
2795 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)xs->inode_bh->b_data; | |
2796 | struct buffer_head *blk_bh = NULL; | |
589dc260 | 2797 | struct ocfs2_xattr_block *xb; |
cf1d6c76 TY |
2798 | int ret = 0; |
2799 | ||
2800 | if (!di->i_xattr_loc) | |
2801 | return ret; | |
2802 | ||
4ae1d69b JB |
2803 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), |
2804 | &blk_bh); | |
cf1d6c76 TY |
2805 | if (ret < 0) { |
2806 | mlog_errno(ret); | |
2807 | return ret; | |
2808 | } | |
f6087fb7 | 2809 | |
cf1d6c76 | 2810 | xs->xattr_bh = blk_bh; |
4ae1d69b | 2811 | xb = (struct ocfs2_xattr_block *)blk_bh->b_data; |
589dc260 TM |
2812 | |
2813 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
2814 | xs->header = &xb->xb_attrs.xb_header; | |
2815 | xs->base = (void *)xs->header; | |
2816 | xs->end = (void *)(blk_bh->b_data) + blk_bh->b_size; | |
2817 | xs->here = xs->header->xh_entries; | |
2818 | ||
2819 | ret = ocfs2_xattr_find_entry(name_index, name, xs); | |
2820 | } else | |
2821 | ret = ocfs2_xattr_index_block_find(inode, blk_bh, | |
2822 | name_index, | |
2823 | name, xs); | |
cf1d6c76 | 2824 | |
cf1d6c76 TY |
2825 | if (ret && ret != -ENODATA) { |
2826 | xs->xattr_bh = NULL; | |
2827 | goto cleanup; | |
2828 | } | |
2829 | xs->not_found = ret; | |
2830 | return 0; | |
cf1d6c76 TY |
2831 | cleanup: |
2832 | brelse(blk_bh); | |
2833 | ||
2834 | return ret; | |
2835 | } | |
2836 | ||
d3981544 | 2837 | static int ocfs2_create_xattr_block(struct inode *inode, |
5aea1f0e | 2838 | struct buffer_head *inode_bh, |
d3981544 JB |
2839 | struct ocfs2_xattr_set_ctxt *ctxt, |
2840 | int indexed, | |
2841 | struct buffer_head **ret_bh) | |
5aea1f0e TM |
2842 | { |
2843 | int ret; | |
2844 | u16 suballoc_bit_start; | |
2845 | u32 num_got; | |
2b6cb576 | 2846 | u64 suballoc_loc, first_blkno; |
5aea1f0e | 2847 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)inode_bh->b_data; |
5aea1f0e TM |
2848 | struct buffer_head *new_bh = NULL; |
2849 | struct ocfs2_xattr_block *xblk; | |
2850 | ||
d3981544 JB |
2851 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
2852 | inode_bh, OCFS2_JOURNAL_ACCESS_CREATE); | |
5aea1f0e TM |
2853 | if (ret < 0) { |
2854 | mlog_errno(ret); | |
2855 | goto end; | |
2856 | } | |
2857 | ||
1ed9b777 | 2858 | ret = ocfs2_claim_metadata(ctxt->handle, ctxt->meta_ac, 1, |
2b6cb576 JB |
2859 | &suballoc_loc, &suballoc_bit_start, |
2860 | &num_got, &first_blkno); | |
5aea1f0e TM |
2861 | if (ret < 0) { |
2862 | mlog_errno(ret); | |
2863 | goto end; | |
2864 | } | |
2865 | ||
2866 | new_bh = sb_getblk(inode->i_sb, first_blkno); | |
58796207 RX |
2867 | if (!new_bh) { |
2868 | ret = -ENOMEM; | |
2869 | mlog_errno(ret); | |
2870 | goto end; | |
2871 | } | |
2872 | ||
5aea1f0e TM |
2873 | ocfs2_set_new_buffer_uptodate(INODE_CACHE(inode), new_bh); |
2874 | ||
d3981544 | 2875 | ret = ocfs2_journal_access_xb(ctxt->handle, INODE_CACHE(inode), |
5aea1f0e TM |
2876 | new_bh, |
2877 | OCFS2_JOURNAL_ACCESS_CREATE); | |
2878 | if (ret < 0) { | |
2879 | mlog_errno(ret); | |
2880 | goto end; | |
2881 | } | |
2882 | ||
2883 | /* Initialize ocfs2_xattr_block */ | |
2884 | xblk = (struct ocfs2_xattr_block *)new_bh->b_data; | |
2885 | memset(xblk, 0, inode->i_sb->s_blocksize); | |
2886 | strcpy((void *)xblk, OCFS2_XATTR_BLOCK_SIGNATURE); | |
d3981544 | 2887 | xblk->xb_suballoc_slot = cpu_to_le16(ctxt->meta_ac->ac_alloc_slot); |
2b6cb576 | 2888 | xblk->xb_suballoc_loc = cpu_to_le64(suballoc_loc); |
5aea1f0e | 2889 | xblk->xb_suballoc_bit = cpu_to_le16(suballoc_bit_start); |
1ed9b777 JB |
2890 | xblk->xb_fs_generation = |
2891 | cpu_to_le32(OCFS2_SB(inode->i_sb)->fs_generation); | |
5aea1f0e | 2892 | xblk->xb_blkno = cpu_to_le64(first_blkno); |
a7fe7a3a TM |
2893 | if (indexed) { |
2894 | struct ocfs2_xattr_tree_root *xr = &xblk->xb_attrs.xb_root; | |
2895 | xr->xt_clusters = cpu_to_le32(1); | |
2896 | xr->xt_last_eb_blk = 0; | |
2897 | xr->xt_list.l_tree_depth = 0; | |
2898 | xr->xt_list.l_count = cpu_to_le16( | |
2899 | ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
2900 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
2901 | xblk->xb_flags = cpu_to_le16(OCFS2_XATTR_INDEXED); | |
2902 | } | |
d3981544 | 2903 | ocfs2_journal_dirty(ctxt->handle, new_bh); |
a7fe7a3a | 2904 | |
d3981544 | 2905 | /* Add it to the inode */ |
5aea1f0e | 2906 | di->i_xattr_loc = cpu_to_le64(first_blkno); |
d3981544 JB |
2907 | |
2908 | spin_lock(&OCFS2_I(inode)->ip_lock); | |
2909 | OCFS2_I(inode)->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
2910 | di->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features); | |
2911 | spin_unlock(&OCFS2_I(inode)->ip_lock); | |
2912 | ||
2913 | ocfs2_journal_dirty(ctxt->handle, inode_bh); | |
5aea1f0e TM |
2914 | |
2915 | *ret_bh = new_bh; | |
2916 | new_bh = NULL; | |
2917 | ||
2918 | end: | |
2919 | brelse(new_bh); | |
2920 | return ret; | |
2921 | } | |
2922 | ||
cf1d6c76 TY |
2923 | /* |
2924 | * ocfs2_xattr_block_set() | |
2925 | * | |
2926 | * Set, replace or remove an extended attribute into external block. | |
2927 | * | |
2928 | */ | |
2929 | static int ocfs2_xattr_block_set(struct inode *inode, | |
2930 | struct ocfs2_xattr_info *xi, | |
78f30c31 TM |
2931 | struct ocfs2_xattr_search *xs, |
2932 | struct ocfs2_xattr_set_ctxt *ctxt) | |
cf1d6c76 TY |
2933 | { |
2934 | struct buffer_head *new_bh = NULL; | |
cf1d6c76 | 2935 | struct ocfs2_xattr_block *xblk = NULL; |
cf1d6c76 | 2936 | int ret; |
d3981544 | 2937 | struct ocfs2_xa_loc loc; |
cf1d6c76 TY |
2938 | |
2939 | if (!xs->xattr_bh) { | |
d3981544 JB |
2940 | ret = ocfs2_create_xattr_block(inode, xs->inode_bh, ctxt, |
2941 | 0, &new_bh); | |
5aea1f0e | 2942 | if (ret) { |
cf1d6c76 | 2943 | mlog_errno(ret); |
85db90e7 | 2944 | goto end; |
cf1d6c76 TY |
2945 | } |
2946 | ||
cf1d6c76 | 2947 | xs->xattr_bh = new_bh; |
5aea1f0e | 2948 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; |
cf1d6c76 TY |
2949 | xs->header = &xblk->xb_attrs.xb_header; |
2950 | xs->base = (void *)xs->header; | |
2951 | xs->end = (void *)xblk + inode->i_sb->s_blocksize; | |
2952 | xs->here = xs->header->xh_entries; | |
01225596 TM |
2953 | } else |
2954 | xblk = (struct ocfs2_xattr_block *)xs->xattr_bh->b_data; | |
2955 | ||
2956 | if (!(le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
d3981544 JB |
2957 | ocfs2_init_xattr_block_xa_loc(&loc, inode, xs->xattr_bh, |
2958 | xs->not_found ? NULL : xs->here); | |
01225596 | 2959 | |
d3981544 JB |
2960 | ret = ocfs2_xa_set(&loc, xi, ctxt); |
2961 | if (!ret) | |
2962 | xs->here = loc.xl_entry; | |
5f5261ac | 2963 | else if ((ret != -ENOSPC) || ctxt->set_abort) |
01225596 | 2964 | goto end; |
d3981544 JB |
2965 | else { |
2966 | ret = ocfs2_xattr_create_index_block(inode, xs, ctxt); | |
2967 | if (ret) | |
2968 | goto end; | |
2969 | } | |
cf1d6c76 TY |
2970 | } |
2971 | ||
d3981544 JB |
2972 | if (le16_to_cpu(xblk->xb_flags) & OCFS2_XATTR_INDEXED) |
2973 | ret = ocfs2_xattr_set_entry_index_block(inode, xi, xs, ctxt); | |
01225596 TM |
2974 | |
2975 | end: | |
cf1d6c76 TY |
2976 | return ret; |
2977 | } | |
2978 | ||
78f30c31 TM |
2979 | /* Check whether the new xattr can be inserted into the inode. */ |
2980 | static int ocfs2_xattr_can_be_in_inode(struct inode *inode, | |
2981 | struct ocfs2_xattr_info *xi, | |
2982 | struct ocfs2_xattr_search *xs) | |
2983 | { | |
78f30c31 TM |
2984 | struct ocfs2_xattr_entry *last; |
2985 | int free, i; | |
2986 | size_t min_offs = xs->end - xs->base; | |
2987 | ||
2988 | if (!xs->header) | |
2989 | return 0; | |
2990 | ||
2991 | last = xs->header->xh_entries; | |
2992 | ||
2993 | for (i = 0; i < le16_to_cpu(xs->header->xh_count); i++) { | |
2994 | size_t offs = le16_to_cpu(last->xe_name_offset); | |
2995 | if (offs < min_offs) | |
2996 | min_offs = offs; | |
2997 | last += 1; | |
2998 | } | |
2999 | ||
4442f518 | 3000 | free = min_offs - ((void *)last - xs->base) - OCFS2_XATTR_HEADER_GAP; |
78f30c31 TM |
3001 | if (free < 0) |
3002 | return 0; | |
3003 | ||
3004 | BUG_ON(!xs->not_found); | |
3005 | ||
199799a3 | 3006 | if (free >= (sizeof(struct ocfs2_xattr_entry) + namevalue_size_xi(xi))) |
78f30c31 TM |
3007 | return 1; |
3008 | ||
3009 | return 0; | |
3010 | } | |
3011 | ||
3012 | static int ocfs2_calc_xattr_set_need(struct inode *inode, | |
3013 | struct ocfs2_dinode *di, | |
3014 | struct ocfs2_xattr_info *xi, | |
3015 | struct ocfs2_xattr_search *xis, | |
3016 | struct ocfs2_xattr_search *xbs, | |
3017 | int *clusters_need, | |
85db90e7 TM |
3018 | int *meta_need, |
3019 | int *credits_need) | |
78f30c31 TM |
3020 | { |
3021 | int ret = 0, old_in_xb = 0; | |
85db90e7 | 3022 | int clusters_add = 0, meta_add = 0, credits = 0; |
78f30c31 TM |
3023 | struct buffer_head *bh = NULL; |
3024 | struct ocfs2_xattr_block *xb = NULL; | |
3025 | struct ocfs2_xattr_entry *xe = NULL; | |
3026 | struct ocfs2_xattr_value_root *xv = NULL; | |
3027 | char *base = NULL; | |
3028 | int name_offset, name_len = 0; | |
3029 | u32 new_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
6b240ff6 | 3030 | xi->xi_value_len); |
78f30c31 TM |
3031 | u64 value_size; |
3032 | ||
71d548a6 TM |
3033 | /* |
3034 | * Calculate the clusters we need to write. | |
3035 | * No matter whether we replace an old one or add a new one, | |
3036 | * we need this for writing. | |
3037 | */ | |
6b240ff6 | 3038 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) |
71d548a6 TM |
3039 | credits += new_clusters * |
3040 | ocfs2_clusters_to_blocks(inode->i_sb, 1); | |
3041 | ||
78f30c31 | 3042 | if (xis->not_found && xbs->not_found) { |
85db90e7 TM |
3043 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3044 | ||
6b240ff6 | 3045 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 | 3046 | clusters_add += new_clusters; |
85db90e7 | 3047 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3048 | &def_xv.xv.xr_list); |
85db90e7 | 3049 | } |
78f30c31 TM |
3050 | |
3051 | goto meta_guess; | |
3052 | } | |
3053 | ||
3054 | if (!xis->not_found) { | |
3055 | xe = xis->here; | |
3056 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3057 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3058 | base = xis->base; | |
85db90e7 | 3059 | credits += OCFS2_INODE_UPDATE_CREDITS; |
78f30c31 | 3060 | } else { |
970e4936 | 3061 | int i, block_off = 0; |
78f30c31 TM |
3062 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; |
3063 | xe = xbs->here; | |
3064 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
3065 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
3066 | i = xbs->here - xbs->header->xh_entries; | |
3067 | old_in_xb = 1; | |
3068 | ||
3069 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 3070 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
78f30c31 TM |
3071 | bucket_xh(xbs->bucket), |
3072 | i, &block_off, | |
3073 | &name_offset); | |
3074 | base = bucket_block(xbs->bucket, block_off); | |
85db90e7 TM |
3075 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
3076 | } else { | |
78f30c31 | 3077 | base = xbs->base; |
85db90e7 TM |
3078 | credits += OCFS2_XATTR_BLOCK_UPDATE_CREDITS; |
3079 | } | |
3080 | } | |
3081 | ||
3082 | /* | |
3083 | * delete a xattr doesn't need metadata and cluster allocation. | |
3084 | * so just calculate the credits and return. | |
3085 | * | |
3086 | * The credits for removing the value tree will be extended | |
3087 | * by ocfs2_remove_extent itself. | |
3088 | */ | |
6b240ff6 | 3089 | if (!xi->xi_value) { |
85db90e7 | 3090 | if (!ocfs2_xattr_is_local(xe)) |
a90714c1 | 3091 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
85db90e7 TM |
3092 | |
3093 | goto out; | |
78f30c31 TM |
3094 | } |
3095 | ||
3096 | /* do cluster allocation guess first. */ | |
3097 | value_size = le64_to_cpu(xe->xe_value_size); | |
3098 | ||
3099 | if (old_in_xb) { | |
3100 | /* | |
3101 | * In xattr set, we always try to set the xe in inode first, | |
3102 | * so if it can be inserted into inode successfully, the old | |
3103 | * one will be removed from the xattr block, and this xattr | |
3104 | * will be inserted into inode as a new xattr in inode. | |
3105 | */ | |
3106 | if (ocfs2_xattr_can_be_in_inode(inode, xi, xis)) { | |
3107 | clusters_add += new_clusters; | |
a90714c1 | 3108 | credits += ocfs2_remove_extent_credits(inode->i_sb) + |
85db90e7 TM |
3109 | OCFS2_INODE_UPDATE_CREDITS; |
3110 | if (!ocfs2_xattr_is_local(xe)) | |
3111 | credits += ocfs2_calc_extend_credits( | |
3112 | inode->i_sb, | |
06f9da6e | 3113 | &def_xv.xv.xr_list); |
78f30c31 TM |
3114 | goto out; |
3115 | } | |
3116 | } | |
3117 | ||
6b240ff6 | 3118 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
78f30c31 TM |
3119 | /* the new values will be stored outside. */ |
3120 | u32 old_clusters = 0; | |
3121 | ||
3122 | if (!ocfs2_xattr_is_local(xe)) { | |
3123 | old_clusters = ocfs2_clusters_for_bytes(inode->i_sb, | |
3124 | value_size); | |
3125 | xv = (struct ocfs2_xattr_value_root *) | |
3126 | (base + name_offset + name_len); | |
97aff52a | 3127 | value_size = OCFS2_XATTR_ROOT_SIZE; |
78f30c31 TM |
3128 | } else |
3129 | xv = &def_xv.xv; | |
3130 | ||
85db90e7 | 3131 | if (old_clusters >= new_clusters) { |
a90714c1 | 3132 | credits += ocfs2_remove_extent_credits(inode->i_sb); |
78f30c31 | 3133 | goto out; |
85db90e7 | 3134 | } else { |
78f30c31 TM |
3135 | meta_add += ocfs2_extend_meta_needed(&xv->xr_list); |
3136 | clusters_add += new_clusters - old_clusters; | |
85db90e7 | 3137 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3138 | &xv->xr_list); |
97aff52a TM |
3139 | if (value_size >= OCFS2_XATTR_ROOT_SIZE) |
3140 | goto out; | |
78f30c31 TM |
3141 | } |
3142 | } else { | |
3143 | /* | |
3144 | * Now the new value will be stored inside. So if the new | |
3145 | * value is smaller than the size of value root or the old | |
3146 | * value, we don't need any allocation, otherwise we have | |
3147 | * to guess metadata allocation. | |
3148 | */ | |
6b240ff6 JB |
3149 | if ((ocfs2_xattr_is_local(xe) && |
3150 | (value_size >= xi->xi_value_len)) || | |
78f30c31 | 3151 | (!ocfs2_xattr_is_local(xe) && |
6b240ff6 | 3152 | OCFS2_XATTR_ROOT_SIZE >= xi->xi_value_len)) |
78f30c31 TM |
3153 | goto out; |
3154 | } | |
3155 | ||
3156 | meta_guess: | |
3157 | /* calculate metadata allocation. */ | |
3158 | if (di->i_xattr_loc) { | |
3159 | if (!xbs->xattr_bh) { | |
4ae1d69b JB |
3160 | ret = ocfs2_read_xattr_block(inode, |
3161 | le64_to_cpu(di->i_xattr_loc), | |
3162 | &bh); | |
78f30c31 TM |
3163 | if (ret) { |
3164 | mlog_errno(ret); | |
3165 | goto out; | |
3166 | } | |
3167 | ||
3168 | xb = (struct ocfs2_xattr_block *)bh->b_data; | |
3169 | } else | |
3170 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
3171 | ||
90cb546c TM |
3172 | /* |
3173 | * If there is already an xattr tree, good, we can calculate | |
3174 | * like other b-trees. Otherwise we may have the chance of | |
3175 | * create a tree, the credit calculation is borrowed from | |
3176 | * ocfs2_calc_extend_credits with root_el = NULL. And the | |
3177 | * new tree will be cluster based, so no meta is needed. | |
3178 | */ | |
78f30c31 TM |
3179 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { |
3180 | struct ocfs2_extent_list *el = | |
3181 | &xb->xb_attrs.xb_root.xt_list; | |
3182 | meta_add += ocfs2_extend_meta_needed(el); | |
85db90e7 | 3183 | credits += ocfs2_calc_extend_credits(inode->i_sb, |
06f9da6e | 3184 | el); |
90cb546c TM |
3185 | } else |
3186 | credits += OCFS2_SUBALLOC_ALLOC + 1; | |
78f30c31 TM |
3187 | |
3188 | /* | |
3189 | * This cluster will be used either for new bucket or for | |
3190 | * new xattr block. | |
3191 | * If the cluster size is the same as the bucket size, one | |
3192 | * more is needed since we may need to extend the bucket | |
3193 | * also. | |
3194 | */ | |
3195 | clusters_add += 1; | |
85db90e7 | 3196 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
78f30c31 | 3197 | if (OCFS2_XATTR_BUCKET_SIZE == |
85db90e7 TM |
3198 | OCFS2_SB(inode->i_sb)->s_clustersize) { |
3199 | credits += ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
78f30c31 | 3200 | clusters_add += 1; |
85db90e7 TM |
3201 | } |
3202 | } else { | |
85db90e7 | 3203 | credits += OCFS2_XATTR_BLOCK_CREATE_CREDITS; |
3ed2be71 TS |
3204 | if (xi->xi_value_len > OCFS2_XATTR_INLINE_SIZE) { |
3205 | struct ocfs2_extent_list *el = &def_xv.xv.xr_list; | |
3206 | meta_add += ocfs2_extend_meta_needed(el); | |
3207 | credits += ocfs2_calc_extend_credits(inode->i_sb, | |
3208 | el); | |
3209 | } else { | |
3210 | meta_add += 1; | |
3211 | } | |
85db90e7 | 3212 | } |
78f30c31 TM |
3213 | out: |
3214 | if (clusters_need) | |
3215 | *clusters_need = clusters_add; | |
3216 | if (meta_need) | |
3217 | *meta_need = meta_add; | |
85db90e7 TM |
3218 | if (credits_need) |
3219 | *credits_need = credits; | |
78f30c31 TM |
3220 | brelse(bh); |
3221 | return ret; | |
3222 | } | |
3223 | ||
3224 | static int ocfs2_init_xattr_set_ctxt(struct inode *inode, | |
3225 | struct ocfs2_dinode *di, | |
3226 | struct ocfs2_xattr_info *xi, | |
3227 | struct ocfs2_xattr_search *xis, | |
3228 | struct ocfs2_xattr_search *xbs, | |
85db90e7 | 3229 | struct ocfs2_xattr_set_ctxt *ctxt, |
492a8a33 | 3230 | int extra_meta, |
85db90e7 | 3231 | int *credits) |
78f30c31 TM |
3232 | { |
3233 | int clusters_add, meta_add, ret; | |
3234 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
3235 | ||
3236 | memset(ctxt, 0, sizeof(struct ocfs2_xattr_set_ctxt)); | |
3237 | ||
3238 | ocfs2_init_dealloc_ctxt(&ctxt->dealloc); | |
3239 | ||
3240 | ret = ocfs2_calc_xattr_set_need(inode, di, xi, xis, xbs, | |
85db90e7 | 3241 | &clusters_add, &meta_add, credits); |
78f30c31 TM |
3242 | if (ret) { |
3243 | mlog_errno(ret); | |
3244 | return ret; | |
3245 | } | |
3246 | ||
492a8a33 | 3247 | meta_add += extra_meta; |
402b4183 TM |
3248 | trace_ocfs2_init_xattr_set_ctxt(xi->xi_name, meta_add, |
3249 | clusters_add, *credits); | |
78f30c31 TM |
3250 | |
3251 | if (meta_add) { | |
3252 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, | |
3253 | &ctxt->meta_ac); | |
3254 | if (ret) { | |
3255 | mlog_errno(ret); | |
3256 | goto out; | |
3257 | } | |
3258 | } | |
3259 | ||
3260 | if (clusters_add) { | |
3261 | ret = ocfs2_reserve_clusters(osb, clusters_add, &ctxt->data_ac); | |
3262 | if (ret) | |
3263 | mlog_errno(ret); | |
3264 | } | |
3265 | out: | |
3266 | if (ret) { | |
3267 | if (ctxt->meta_ac) { | |
3268 | ocfs2_free_alloc_context(ctxt->meta_ac); | |
3269 | ctxt->meta_ac = NULL; | |
3270 | } | |
3271 | ||
3272 | /* | |
3273 | * We cannot have an error and a non null ctxt->data_ac. | |
3274 | */ | |
3275 | } | |
3276 | ||
3277 | return ret; | |
3278 | } | |
3279 | ||
85db90e7 TM |
3280 | static int __ocfs2_xattr_set_handle(struct inode *inode, |
3281 | struct ocfs2_dinode *di, | |
3282 | struct ocfs2_xattr_info *xi, | |
3283 | struct ocfs2_xattr_search *xis, | |
3284 | struct ocfs2_xattr_search *xbs, | |
3285 | struct ocfs2_xattr_set_ctxt *ctxt) | |
3286 | { | |
9f868f16 | 3287 | int ret = 0, credits, old_found; |
85db90e7 | 3288 | |
6b240ff6 | 3289 | if (!xi->xi_value) { |
85db90e7 TM |
3290 | /* Remove existing extended attribute */ |
3291 | if (!xis->not_found) | |
3292 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3293 | else if (!xbs->not_found) | |
3294 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3295 | } else { | |
3296 | /* We always try to set extended attribute into inode first*/ | |
3297 | ret = ocfs2_xattr_ibody_set(inode, xi, xis, ctxt); | |
3298 | if (!ret && !xbs->not_found) { | |
3299 | /* | |
3300 | * If succeed and that extended attribute existing in | |
3301 | * external block, then we will remove it. | |
3302 | */ | |
6b240ff6 JB |
3303 | xi->xi_value = NULL; |
3304 | xi->xi_value_len = 0; | |
85db90e7 | 3305 | |
9f868f16 | 3306 | old_found = xis->not_found; |
85db90e7 TM |
3307 | xis->not_found = -ENODATA; |
3308 | ret = ocfs2_calc_xattr_set_need(inode, | |
3309 | di, | |
3310 | xi, | |
3311 | xis, | |
3312 | xbs, | |
3313 | NULL, | |
3314 | NULL, | |
3315 | &credits); | |
9f868f16 | 3316 | xis->not_found = old_found; |
85db90e7 TM |
3317 | if (ret) { |
3318 | mlog_errno(ret); | |
3319 | goto out; | |
3320 | } | |
3321 | ||
c901fb00 | 3322 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3323 | if (ret) { |
3324 | mlog_errno(ret); | |
3325 | goto out; | |
3326 | } | |
3327 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
5f5261ac | 3328 | } else if ((ret == -ENOSPC) && !ctxt->set_abort) { |
85db90e7 TM |
3329 | if (di->i_xattr_loc && !xbs->xattr_bh) { |
3330 | ret = ocfs2_xattr_block_find(inode, | |
6b240ff6 JB |
3331 | xi->xi_name_index, |
3332 | xi->xi_name, xbs); | |
85db90e7 TM |
3333 | if (ret) |
3334 | goto out; | |
3335 | ||
9f868f16 | 3336 | old_found = xis->not_found; |
85db90e7 TM |
3337 | xis->not_found = -ENODATA; |
3338 | ret = ocfs2_calc_xattr_set_need(inode, | |
3339 | di, | |
3340 | xi, | |
3341 | xis, | |
3342 | xbs, | |
3343 | NULL, | |
3344 | NULL, | |
3345 | &credits); | |
9f868f16 | 3346 | xis->not_found = old_found; |
85db90e7 TM |
3347 | if (ret) { |
3348 | mlog_errno(ret); | |
3349 | goto out; | |
3350 | } | |
3351 | ||
c901fb00 | 3352 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3353 | if (ret) { |
3354 | mlog_errno(ret); | |
3355 | goto out; | |
3356 | } | |
3357 | } | |
3358 | /* | |
3359 | * If no space in inode, we will set extended attribute | |
3360 | * into external block. | |
3361 | */ | |
3362 | ret = ocfs2_xattr_block_set(inode, xi, xbs, ctxt); | |
3363 | if (ret) | |
3364 | goto out; | |
3365 | if (!xis->not_found) { | |
3366 | /* | |
3367 | * If succeed and that extended attribute | |
3368 | * existing in inode, we will remove it. | |
3369 | */ | |
6b240ff6 JB |
3370 | xi->xi_value = NULL; |
3371 | xi->xi_value_len = 0; | |
85db90e7 TM |
3372 | xbs->not_found = -ENODATA; |
3373 | ret = ocfs2_calc_xattr_set_need(inode, | |
3374 | di, | |
3375 | xi, | |
3376 | xis, | |
3377 | xbs, | |
3378 | NULL, | |
3379 | NULL, | |
3380 | &credits); | |
3381 | if (ret) { | |
3382 | mlog_errno(ret); | |
3383 | goto out; | |
3384 | } | |
3385 | ||
c901fb00 | 3386 | ret = ocfs2_extend_trans(ctxt->handle, credits); |
85db90e7 TM |
3387 | if (ret) { |
3388 | mlog_errno(ret); | |
3389 | goto out; | |
3390 | } | |
3391 | ret = ocfs2_xattr_ibody_set(inode, xi, | |
3392 | xis, ctxt); | |
3393 | } | |
3394 | } | |
3395 | } | |
3396 | ||
4b3f6209 TM |
3397 | if (!ret) { |
3398 | /* Update inode ctime. */ | |
0cf2f763 | 3399 | ret = ocfs2_journal_access_di(ctxt->handle, INODE_CACHE(inode), |
89a907af TM |
3400 | xis->inode_bh, |
3401 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4b3f6209 TM |
3402 | if (ret) { |
3403 | mlog_errno(ret); | |
3404 | goto out; | |
3405 | } | |
3406 | ||
3407 | inode->i_ctime = CURRENT_TIME; | |
3408 | di->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); | |
3409 | di->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); | |
3410 | ocfs2_journal_dirty(ctxt->handle, xis->inode_bh); | |
3411 | } | |
85db90e7 TM |
3412 | out: |
3413 | return ret; | |
3414 | } | |
3415 | ||
6c3faba4 TY |
3416 | /* |
3417 | * This function only called duing creating inode | |
3418 | * for init security/acl xattrs of the new inode. | |
008aafaf | 3419 | * All transanction credits have been reserved in mknod. |
6c3faba4 TY |
3420 | */ |
3421 | int ocfs2_xattr_set_handle(handle_t *handle, | |
3422 | struct inode *inode, | |
3423 | struct buffer_head *di_bh, | |
3424 | int name_index, | |
3425 | const char *name, | |
3426 | const void *value, | |
3427 | size_t value_len, | |
3428 | int flags, | |
3429 | struct ocfs2_alloc_context *meta_ac, | |
3430 | struct ocfs2_alloc_context *data_ac) | |
3431 | { | |
3432 | struct ocfs2_dinode *di; | |
3433 | int ret; | |
3434 | ||
3435 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3436 | .xi_name_index = name_index, |
3437 | .xi_name = name, | |
18853b95 | 3438 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3439 | .xi_value = value, |
3440 | .xi_value_len = value_len, | |
6c3faba4 TY |
3441 | }; |
3442 | ||
3443 | struct ocfs2_xattr_search xis = { | |
3444 | .not_found = -ENODATA, | |
3445 | }; | |
3446 | ||
3447 | struct ocfs2_xattr_search xbs = { | |
3448 | .not_found = -ENODATA, | |
3449 | }; | |
3450 | ||
3451 | struct ocfs2_xattr_set_ctxt ctxt = { | |
3452 | .handle = handle, | |
3453 | .meta_ac = meta_ac, | |
3454 | .data_ac = data_ac, | |
3455 | }; | |
3456 | ||
3457 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) | |
3458 | return -EOPNOTSUPP; | |
3459 | ||
008aafaf TY |
3460 | /* |
3461 | * In extreme situation, may need xattr bucket when | |
3462 | * block size is too small. And we have already reserved | |
3463 | * the credits for bucket in mknod. | |
3464 | */ | |
3465 | if (inode->i_sb->s_blocksize == OCFS2_MIN_BLOCKSIZE) { | |
3466 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3467 | if (!xbs.bucket) { | |
3468 | mlog_errno(-ENOMEM); | |
3469 | return -ENOMEM; | |
3470 | } | |
3471 | } | |
3472 | ||
6c3faba4 TY |
3473 | xis.inode_bh = xbs.inode_bh = di_bh; |
3474 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3475 | ||
3476 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3477 | ||
3478 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3479 | if (ret) | |
3480 | goto cleanup; | |
3481 | if (xis.not_found) { | |
3482 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3483 | if (ret) | |
3484 | goto cleanup; | |
3485 | } | |
3486 | ||
3487 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3488 | ||
3489 | cleanup: | |
3490 | up_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3491 | brelse(xbs.xattr_bh); | |
008aafaf | 3492 | ocfs2_xattr_bucket_free(xbs.bucket); |
6c3faba4 TY |
3493 | |
3494 | return ret; | |
3495 | } | |
3496 | ||
cf1d6c76 TY |
3497 | /* |
3498 | * ocfs2_xattr_set() | |
3499 | * | |
3500 | * Set, replace or remove an extended attribute for this inode. | |
3501 | * value is NULL to remove an existing extended attribute, else either | |
3502 | * create or replace an extended attribute. | |
3503 | */ | |
3504 | int ocfs2_xattr_set(struct inode *inode, | |
3505 | int name_index, | |
3506 | const char *name, | |
3507 | const void *value, | |
3508 | size_t value_len, | |
3509 | int flags) | |
3510 | { | |
3511 | struct buffer_head *di_bh = NULL; | |
3512 | struct ocfs2_dinode *di; | |
492a8a33 | 3513 | int ret, credits, ref_meta = 0, ref_credits = 0; |
78f30c31 | 3514 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
85db90e7 | 3515 | struct inode *tl_inode = osb->osb_tl_inode; |
6ea437a3 | 3516 | struct ocfs2_xattr_set_ctxt ctxt = { NULL, NULL, NULL, }; |
492a8a33 | 3517 | struct ocfs2_refcount_tree *ref_tree = NULL; |
cf1d6c76 TY |
3518 | |
3519 | struct ocfs2_xattr_info xi = { | |
6b240ff6 JB |
3520 | .xi_name_index = name_index, |
3521 | .xi_name = name, | |
18853b95 | 3522 | .xi_name_len = strlen(name), |
6b240ff6 JB |
3523 | .xi_value = value, |
3524 | .xi_value_len = value_len, | |
cf1d6c76 TY |
3525 | }; |
3526 | ||
3527 | struct ocfs2_xattr_search xis = { | |
3528 | .not_found = -ENODATA, | |
3529 | }; | |
3530 | ||
3531 | struct ocfs2_xattr_search xbs = { | |
3532 | .not_found = -ENODATA, | |
3533 | }; | |
3534 | ||
8154da3d TY |
3535 | if (!ocfs2_supports_xattr(OCFS2_SB(inode->i_sb))) |
3536 | return -EOPNOTSUPP; | |
3537 | ||
ba937127 JB |
3538 | /* |
3539 | * Only xbs will be used on indexed trees. xis doesn't need a | |
3540 | * bucket. | |
3541 | */ | |
3542 | xbs.bucket = ocfs2_xattr_bucket_new(inode); | |
3543 | if (!xbs.bucket) { | |
3544 | mlog_errno(-ENOMEM); | |
3545 | return -ENOMEM; | |
3546 | } | |
3547 | ||
cf1d6c76 TY |
3548 | ret = ocfs2_inode_lock(inode, &di_bh, 1); |
3549 | if (ret < 0) { | |
3550 | mlog_errno(ret); | |
ba937127 | 3551 | goto cleanup_nolock; |
cf1d6c76 TY |
3552 | } |
3553 | xis.inode_bh = xbs.inode_bh = di_bh; | |
3554 | di = (struct ocfs2_dinode *)di_bh->b_data; | |
3555 | ||
3556 | down_write(&OCFS2_I(inode)->ip_xattr_sem); | |
3557 | /* | |
3558 | * Scan inode and external block to find the same name | |
25985edc | 3559 | * extended attribute and collect search information. |
cf1d6c76 TY |
3560 | */ |
3561 | ret = ocfs2_xattr_ibody_find(inode, name_index, name, &xis); | |
3562 | if (ret) | |
3563 | goto cleanup; | |
3564 | if (xis.not_found) { | |
3565 | ret = ocfs2_xattr_block_find(inode, name_index, name, &xbs); | |
3566 | if (ret) | |
3567 | goto cleanup; | |
3568 | } | |
3569 | ||
3570 | if (xis.not_found && xbs.not_found) { | |
3571 | ret = -ENODATA; | |
3572 | if (flags & XATTR_REPLACE) | |
3573 | goto cleanup; | |
3574 | ret = 0; | |
3575 | if (!value) | |
3576 | goto cleanup; | |
3577 | } else { | |
3578 | ret = -EEXIST; | |
3579 | if (flags & XATTR_CREATE) | |
3580 | goto cleanup; | |
3581 | } | |
3582 | ||
25985edc | 3583 | /* Check whether the value is refcounted and do some preparation. */ |
492a8a33 TM |
3584 | if (OCFS2_I(inode)->ip_dyn_features & OCFS2_HAS_REFCOUNT_FL && |
3585 | (!xis.not_found || !xbs.not_found)) { | |
3586 | ret = ocfs2_prepare_refcount_xattr(inode, di, &xi, | |
3587 | &xis, &xbs, &ref_tree, | |
3588 | &ref_meta, &ref_credits); | |
3589 | if (ret) { | |
3590 | mlog_errno(ret); | |
3591 | goto cleanup; | |
3592 | } | |
3593 | } | |
85db90e7 TM |
3594 | |
3595 | mutex_lock(&tl_inode->i_mutex); | |
3596 | ||
3597 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
3598 | ret = __ocfs2_flush_truncate_log(osb); | |
3599 | if (ret < 0) { | |
3600 | mutex_unlock(&tl_inode->i_mutex); | |
3601 | mlog_errno(ret); | |
3602 | goto cleanup; | |
3603 | } | |
3604 | } | |
3605 | mutex_unlock(&tl_inode->i_mutex); | |
3606 | ||
3607 | ret = ocfs2_init_xattr_set_ctxt(inode, di, &xi, &xis, | |
492a8a33 | 3608 | &xbs, &ctxt, ref_meta, &credits); |
78f30c31 TM |
3609 | if (ret) { |
3610 | mlog_errno(ret); | |
3611 | goto cleanup; | |
3612 | } | |
3613 | ||
4b3f6209 TM |
3614 | /* we need to update inode's ctime field, so add credit for it. */ |
3615 | credits += OCFS2_INODE_UPDATE_CREDITS; | |
492a8a33 | 3616 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); |
85db90e7 TM |
3617 | if (IS_ERR(ctxt.handle)) { |
3618 | ret = PTR_ERR(ctxt.handle); | |
3619 | mlog_errno(ret); | |
6ea437a3 | 3620 | goto out_free_ac; |
cf1d6c76 | 3621 | } |
85db90e7 TM |
3622 | |
3623 | ret = __ocfs2_xattr_set_handle(inode, di, &xi, &xis, &xbs, &ctxt); | |
3624 | ||
3625 | ocfs2_commit_trans(osb, ctxt.handle); | |
3626 | ||
6ea437a3 | 3627 | out_free_ac: |
78f30c31 TM |
3628 | if (ctxt.data_ac) |
3629 | ocfs2_free_alloc_context(ctxt.data_ac); | |
3630 | if (ctxt.meta_ac) | |
3631 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
3632 | if (ocfs2_dealloc_has_cluster(&ctxt.dealloc)) | |
3633 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
3634 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
8b2c0dba | 3635 | |
cf1d6c76 | 3636 | cleanup: |
492a8a33 TM |
3637 | if (ref_tree) |
3638 | ocfs2_unlock_refcount_tree(osb, ref_tree, 1); | |
cf1d6c76 | 3639 | up_write(&OCFS2_I(inode)->ip_xattr_sem); |
8b2c0dba TM |
3640 | if (!value && !ret) { |
3641 | ret = ocfs2_try_remove_refcount_tree(inode, di_bh); | |
3642 | if (ret) | |
3643 | mlog_errno(ret); | |
3644 | } | |
cf1d6c76 | 3645 | ocfs2_inode_unlock(inode, 1); |
ba937127 | 3646 | cleanup_nolock: |
cf1d6c76 TY |
3647 | brelse(di_bh); |
3648 | brelse(xbs.xattr_bh); | |
ba937127 | 3649 | ocfs2_xattr_bucket_free(xbs.bucket); |
cf1d6c76 TY |
3650 | |
3651 | return ret; | |
3652 | } | |
3653 | ||
0c044f0b TM |
3654 | /* |
3655 | * Find the xattr extent rec which may contains name_hash. | |
3656 | * e_cpos will be the first name hash of the xattr rec. | |
3657 | * el must be the ocfs2_xattr_header.xb_attrs.xb_root.xt_list. | |
3658 | */ | |
3659 | static int ocfs2_xattr_get_rec(struct inode *inode, | |
3660 | u32 name_hash, | |
3661 | u64 *p_blkno, | |
3662 | u32 *e_cpos, | |
3663 | u32 *num_clusters, | |
3664 | struct ocfs2_extent_list *el) | |
3665 | { | |
3666 | int ret = 0, i; | |
3667 | struct buffer_head *eb_bh = NULL; | |
3668 | struct ocfs2_extent_block *eb; | |
3669 | struct ocfs2_extent_rec *rec = NULL; | |
3670 | u64 e_blkno = 0; | |
3671 | ||
3672 | if (el->l_tree_depth) { | |
facdb77f JB |
3673 | ret = ocfs2_find_leaf(INODE_CACHE(inode), el, name_hash, |
3674 | &eb_bh); | |
0c044f0b TM |
3675 | if (ret) { |
3676 | mlog_errno(ret); | |
3677 | goto out; | |
3678 | } | |
3679 | ||
3680 | eb = (struct ocfs2_extent_block *) eb_bh->b_data; | |
3681 | el = &eb->h_list; | |
3682 | ||
3683 | if (el->l_tree_depth) { | |
3684 | ocfs2_error(inode->i_sb, | |
3685 | "Inode %lu has non zero tree depth in " | |
3686 | "xattr tree block %llu\n", inode->i_ino, | |
3687 | (unsigned long long)eb_bh->b_blocknr); | |
3688 | ret = -EROFS; | |
3689 | goto out; | |
3690 | } | |
3691 | } | |
3692 | ||
3693 | for (i = le16_to_cpu(el->l_next_free_rec) - 1; i >= 0; i--) { | |
3694 | rec = &el->l_recs[i]; | |
3695 | ||
3696 | if (le32_to_cpu(rec->e_cpos) <= name_hash) { | |
3697 | e_blkno = le64_to_cpu(rec->e_blkno); | |
3698 | break; | |
3699 | } | |
3700 | } | |
3701 | ||
3702 | if (!e_blkno) { | |
3703 | ocfs2_error(inode->i_sb, "Inode %lu has bad extent " | |
3704 | "record (%u, %u, 0) in xattr", inode->i_ino, | |
3705 | le32_to_cpu(rec->e_cpos), | |
3706 | ocfs2_rec_clusters(el, rec)); | |
3707 | ret = -EROFS; | |
3708 | goto out; | |
3709 | } | |
3710 | ||
3711 | *p_blkno = le64_to_cpu(rec->e_blkno); | |
3712 | *num_clusters = le16_to_cpu(rec->e_leaf_clusters); | |
3713 | if (e_cpos) | |
3714 | *e_cpos = le32_to_cpu(rec->e_cpos); | |
3715 | out: | |
3716 | brelse(eb_bh); | |
3717 | return ret; | |
3718 | } | |
3719 | ||
3720 | typedef int (xattr_bucket_func)(struct inode *inode, | |
3721 | struct ocfs2_xattr_bucket *bucket, | |
3722 | void *para); | |
3723 | ||
589dc260 | 3724 | static int ocfs2_find_xe_in_bucket(struct inode *inode, |
e2356a3f | 3725 | struct ocfs2_xattr_bucket *bucket, |
589dc260 TM |
3726 | int name_index, |
3727 | const char *name, | |
3728 | u32 name_hash, | |
3729 | u16 *xe_index, | |
3730 | int *found) | |
3731 | { | |
3732 | int i, ret = 0, cmp = 1, block_off, new_offset; | |
e2356a3f | 3733 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
589dc260 TM |
3734 | size_t name_len = strlen(name); |
3735 | struct ocfs2_xattr_entry *xe = NULL; | |
589dc260 TM |
3736 | char *xe_name; |
3737 | ||
3738 | /* | |
3739 | * We don't use binary search in the bucket because there | |
3740 | * may be multiple entries with the same name hash. | |
3741 | */ | |
3742 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
3743 | xe = &xh->xh_entries[i]; | |
3744 | ||
3745 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) | |
3746 | continue; | |
3747 | else if (name_hash < le32_to_cpu(xe->xe_name_hash)) | |
3748 | break; | |
3749 | ||
3750 | cmp = name_index - ocfs2_xattr_get_type(xe); | |
3751 | if (!cmp) | |
3752 | cmp = name_len - xe->xe_name_len; | |
3753 | if (cmp) | |
3754 | continue; | |
3755 | ||
fd68a894 | 3756 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
589dc260 TM |
3757 | xh, |
3758 | i, | |
3759 | &block_off, | |
3760 | &new_offset); | |
3761 | if (ret) { | |
3762 | mlog_errno(ret); | |
3763 | break; | |
3764 | } | |
3765 | ||
970e4936 | 3766 | |
e2356a3f JB |
3767 | xe_name = bucket_block(bucket, block_off) + new_offset; |
3768 | if (!memcmp(name, xe_name, name_len)) { | |
589dc260 TM |
3769 | *xe_index = i; |
3770 | *found = 1; | |
3771 | ret = 0; | |
3772 | break; | |
3773 | } | |
3774 | } | |
3775 | ||
3776 | return ret; | |
3777 | } | |
3778 | ||
3779 | /* | |
3780 | * Find the specified xattr entry in a series of buckets. | |
3781 | * This series start from p_blkno and last for num_clusters. | |
3782 | * The ocfs2_xattr_header.xh_num_buckets of the first bucket contains | |
3783 | * the num of the valid buckets. | |
3784 | * | |
3785 | * Return the buffer_head this xattr should reside in. And if the xattr's | |
3786 | * hash is in the gap of 2 buckets, return the lower bucket. | |
3787 | */ | |
3788 | static int ocfs2_xattr_bucket_find(struct inode *inode, | |
3789 | int name_index, | |
3790 | const char *name, | |
3791 | u32 name_hash, | |
3792 | u64 p_blkno, | |
3793 | u32 first_hash, | |
3794 | u32 num_clusters, | |
3795 | struct ocfs2_xattr_search *xs) | |
3796 | { | |
3797 | int ret, found = 0; | |
589dc260 TM |
3798 | struct ocfs2_xattr_header *xh = NULL; |
3799 | struct ocfs2_xattr_entry *xe = NULL; | |
3800 | u16 index = 0; | |
3801 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
3802 | int low_bucket = 0, bucket, high_bucket; | |
e2356a3f | 3803 | struct ocfs2_xattr_bucket *search; |
589dc260 | 3804 | u32 last_hash; |
e2356a3f | 3805 | u64 blkno, lower_blkno = 0; |
589dc260 | 3806 | |
e2356a3f JB |
3807 | search = ocfs2_xattr_bucket_new(inode); |
3808 | if (!search) { | |
3809 | ret = -ENOMEM; | |
3810 | mlog_errno(ret); | |
3811 | goto out; | |
3812 | } | |
3813 | ||
3814 | ret = ocfs2_read_xattr_bucket(search, p_blkno); | |
589dc260 TM |
3815 | if (ret) { |
3816 | mlog_errno(ret); | |
3817 | goto out; | |
3818 | } | |
3819 | ||
e2356a3f | 3820 | xh = bucket_xh(search); |
589dc260 | 3821 | high_bucket = le16_to_cpu(xh->xh_num_buckets) - 1; |
589dc260 | 3822 | while (low_bucket <= high_bucket) { |
e2356a3f | 3823 | ocfs2_xattr_bucket_relse(search); |
589dc260 | 3824 | |
e2356a3f | 3825 | bucket = (low_bucket + high_bucket) / 2; |
589dc260 | 3826 | blkno = p_blkno + bucket * blk_per_bucket; |
e2356a3f | 3827 | ret = ocfs2_read_xattr_bucket(search, blkno); |
589dc260 TM |
3828 | if (ret) { |
3829 | mlog_errno(ret); | |
3830 | goto out; | |
3831 | } | |
3832 | ||
e2356a3f | 3833 | xh = bucket_xh(search); |
589dc260 TM |
3834 | xe = &xh->xh_entries[0]; |
3835 | if (name_hash < le32_to_cpu(xe->xe_name_hash)) { | |
3836 | high_bucket = bucket - 1; | |
3837 | continue; | |
3838 | } | |
3839 | ||
3840 | /* | |
3841 | * Check whether the hash of the last entry in our | |
5a095611 TM |
3842 | * bucket is larger than the search one. for an empty |
3843 | * bucket, the last one is also the first one. | |
589dc260 | 3844 | */ |
5a095611 TM |
3845 | if (xh->xh_count) |
3846 | xe = &xh->xh_entries[le16_to_cpu(xh->xh_count) - 1]; | |
3847 | ||
589dc260 TM |
3848 | last_hash = le32_to_cpu(xe->xe_name_hash); |
3849 | ||
e2356a3f JB |
3850 | /* record lower_blkno which may be the insert place. */ |
3851 | lower_blkno = blkno; | |
589dc260 TM |
3852 | |
3853 | if (name_hash > le32_to_cpu(xe->xe_name_hash)) { | |
3854 | low_bucket = bucket + 1; | |
3855 | continue; | |
3856 | } | |
3857 | ||
3858 | /* the searched xattr should reside in this bucket if exists. */ | |
e2356a3f | 3859 | ret = ocfs2_find_xe_in_bucket(inode, search, |
589dc260 TM |
3860 | name_index, name, name_hash, |
3861 | &index, &found); | |
3862 | if (ret) { | |
3863 | mlog_errno(ret); | |
3864 | goto out; | |
3865 | } | |
3866 | break; | |
3867 | } | |
3868 | ||
3869 | /* | |
3870 | * Record the bucket we have found. | |
3871 | * When the xattr's hash value is in the gap of 2 buckets, we will | |
3872 | * always set it to the previous bucket. | |
3873 | */ | |
e2356a3f JB |
3874 | if (!lower_blkno) |
3875 | lower_blkno = p_blkno; | |
3876 | ||
3877 | /* This should be in cache - we just read it during the search */ | |
3878 | ret = ocfs2_read_xattr_bucket(xs->bucket, lower_blkno); | |
3879 | if (ret) { | |
3880 | mlog_errno(ret); | |
3881 | goto out; | |
589dc260 | 3882 | } |
589dc260 | 3883 | |
ba937127 JB |
3884 | xs->header = bucket_xh(xs->bucket); |
3885 | xs->base = bucket_block(xs->bucket, 0); | |
589dc260 TM |
3886 | xs->end = xs->base + inode->i_sb->s_blocksize; |
3887 | ||
3888 | if (found) { | |
589dc260 | 3889 | xs->here = &xs->header->xh_entries[index]; |
402b4183 TM |
3890 | trace_ocfs2_xattr_bucket_find(OCFS2_I(inode)->ip_blkno, |
3891 | name, name_index, name_hash, | |
3892 | (unsigned long long)bucket_blkno(xs->bucket), | |
3893 | index); | |
589dc260 TM |
3894 | } else |
3895 | ret = -ENODATA; | |
3896 | ||
3897 | out: | |
e2356a3f | 3898 | ocfs2_xattr_bucket_free(search); |
589dc260 TM |
3899 | return ret; |
3900 | } | |
3901 | ||
3902 | static int ocfs2_xattr_index_block_find(struct inode *inode, | |
3903 | struct buffer_head *root_bh, | |
3904 | int name_index, | |
3905 | const char *name, | |
3906 | struct ocfs2_xattr_search *xs) | |
3907 | { | |
3908 | int ret; | |
3909 | struct ocfs2_xattr_block *xb = | |
3910 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
3911 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
3912 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
3913 | u64 p_blkno = 0; | |
3914 | u32 first_hash, num_clusters = 0; | |
2057e5c6 | 3915 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
589dc260 TM |
3916 | |
3917 | if (le16_to_cpu(el->l_next_free_rec) == 0) | |
3918 | return -ENODATA; | |
3919 | ||
402b4183 TM |
3920 | trace_ocfs2_xattr_index_block_find(OCFS2_I(inode)->ip_blkno, |
3921 | name, name_index, name_hash, | |
3922 | (unsigned long long)root_bh->b_blocknr, | |
3923 | -1); | |
589dc260 TM |
3924 | |
3925 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &first_hash, | |
3926 | &num_clusters, el); | |
3927 | if (ret) { | |
3928 | mlog_errno(ret); | |
3929 | goto out; | |
3930 | } | |
3931 | ||
3932 | BUG_ON(p_blkno == 0 || num_clusters == 0 || first_hash > name_hash); | |
3933 | ||
402b4183 TM |
3934 | trace_ocfs2_xattr_index_block_find_rec(OCFS2_I(inode)->ip_blkno, |
3935 | name, name_index, first_hash, | |
3936 | (unsigned long long)p_blkno, | |
3937 | num_clusters); | |
589dc260 TM |
3938 | |
3939 | ret = ocfs2_xattr_bucket_find(inode, name_index, name, name_hash, | |
3940 | p_blkno, first_hash, num_clusters, xs); | |
3941 | ||
3942 | out: | |
3943 | return ret; | |
3944 | } | |
3945 | ||
0c044f0b TM |
3946 | static int ocfs2_iterate_xattr_buckets(struct inode *inode, |
3947 | u64 blkno, | |
3948 | u32 clusters, | |
3949 | xattr_bucket_func *func, | |
3950 | void *para) | |
3951 | { | |
6dde41d9 | 3952 | int i, ret = 0; |
0c044f0b TM |
3953 | u32 bpc = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); |
3954 | u32 num_buckets = clusters * bpc; | |
ba937127 | 3955 | struct ocfs2_xattr_bucket *bucket; |
0c044f0b | 3956 | |
ba937127 JB |
3957 | bucket = ocfs2_xattr_bucket_new(inode); |
3958 | if (!bucket) { | |
3959 | mlog_errno(-ENOMEM); | |
3960 | return -ENOMEM; | |
3961 | } | |
0c044f0b | 3962 | |
402b4183 TM |
3963 | trace_ocfs2_iterate_xattr_buckets( |
3964 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
3965 | (unsigned long long)blkno, clusters); | |
0c044f0b | 3966 | |
ba937127 JB |
3967 | for (i = 0; i < num_buckets; i++, blkno += bucket->bu_blocks) { |
3968 | ret = ocfs2_read_xattr_bucket(bucket, blkno); | |
0c044f0b TM |
3969 | if (ret) { |
3970 | mlog_errno(ret); | |
ba937127 | 3971 | break; |
0c044f0b TM |
3972 | } |
3973 | ||
0c044f0b TM |
3974 | /* |
3975 | * The real bucket num in this series of blocks is stored | |
3976 | * in the 1st bucket. | |
3977 | */ | |
3978 | if (i == 0) | |
ba937127 | 3979 | num_buckets = le16_to_cpu(bucket_xh(bucket)->xh_num_buckets); |
0c044f0b | 3980 | |
402b4183 | 3981 | trace_ocfs2_iterate_xattr_bucket((unsigned long long)blkno, |
ba937127 | 3982 | le32_to_cpu(bucket_xh(bucket)->xh_entries[0].xe_name_hash)); |
0c044f0b | 3983 | if (func) { |
ba937127 | 3984 | ret = func(inode, bucket, para); |
a46fa684 | 3985 | if (ret && ret != -ERANGE) |
0c044f0b | 3986 | mlog_errno(ret); |
ba937127 | 3987 | /* Fall through to bucket_relse() */ |
0c044f0b TM |
3988 | } |
3989 | ||
ba937127 JB |
3990 | ocfs2_xattr_bucket_relse(bucket); |
3991 | if (ret) | |
3992 | break; | |
0c044f0b TM |
3993 | } |
3994 | ||
ba937127 | 3995 | ocfs2_xattr_bucket_free(bucket); |
0c044f0b TM |
3996 | return ret; |
3997 | } | |
3998 | ||
3999 | struct ocfs2_xattr_tree_list { | |
4000 | char *buffer; | |
4001 | size_t buffer_size; | |
936b8834 | 4002 | size_t result; |
0c044f0b TM |
4003 | }; |
4004 | ||
fd68a894 | 4005 | static int ocfs2_xattr_bucket_get_name_value(struct super_block *sb, |
0c044f0b TM |
4006 | struct ocfs2_xattr_header *xh, |
4007 | int index, | |
4008 | int *block_off, | |
4009 | int *new_offset) | |
4010 | { | |
4011 | u16 name_offset; | |
4012 | ||
4013 | if (index < 0 || index >= le16_to_cpu(xh->xh_count)) | |
4014 | return -EINVAL; | |
4015 | ||
4016 | name_offset = le16_to_cpu(xh->xh_entries[index].xe_name_offset); | |
4017 | ||
fd68a894 TM |
4018 | *block_off = name_offset >> sb->s_blocksize_bits; |
4019 | *new_offset = name_offset % sb->s_blocksize; | |
0c044f0b TM |
4020 | |
4021 | return 0; | |
4022 | } | |
4023 | ||
4024 | static int ocfs2_list_xattr_bucket(struct inode *inode, | |
4025 | struct ocfs2_xattr_bucket *bucket, | |
4026 | void *para) | |
4027 | { | |
936b8834 | 4028 | int ret = 0, type; |
0c044f0b | 4029 | struct ocfs2_xattr_tree_list *xl = (struct ocfs2_xattr_tree_list *)para; |
0c044f0b | 4030 | int i, block_off, new_offset; |
936b8834 | 4031 | const char *prefix, *name; |
0c044f0b | 4032 | |
3e632946 JB |
4033 | for (i = 0 ; i < le16_to_cpu(bucket_xh(bucket)->xh_count); i++) { |
4034 | struct ocfs2_xattr_entry *entry = &bucket_xh(bucket)->xh_entries[i]; | |
936b8834 TM |
4035 | type = ocfs2_xattr_get_type(entry); |
4036 | prefix = ocfs2_xattr_prefix(type); | |
0c044f0b | 4037 | |
936b8834 | 4038 | if (prefix) { |
fd68a894 | 4039 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
3e632946 | 4040 | bucket_xh(bucket), |
0c044f0b TM |
4041 | i, |
4042 | &block_off, | |
4043 | &new_offset); | |
4044 | if (ret) | |
4045 | break; | |
936b8834 | 4046 | |
51def39f | 4047 | name = (const char *)bucket_block(bucket, block_off) + |
936b8834 TM |
4048 | new_offset; |
4049 | ret = ocfs2_xattr_list_entry(xl->buffer, | |
4050 | xl->buffer_size, | |
4051 | &xl->result, | |
4052 | prefix, name, | |
4053 | entry->xe_name_len); | |
4054 | if (ret) | |
4055 | break; | |
0c044f0b TM |
4056 | } |
4057 | } | |
4058 | ||
4059 | return ret; | |
4060 | } | |
4061 | ||
47bca495 TM |
4062 | static int ocfs2_iterate_xattr_index_block(struct inode *inode, |
4063 | struct buffer_head *blk_bh, | |
4064 | xattr_tree_rec_func *rec_func, | |
4065 | void *para) | |
0c044f0b | 4066 | { |
47bca495 TM |
4067 | struct ocfs2_xattr_block *xb = |
4068 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
4069 | struct ocfs2_extent_list *el = &xb->xb_attrs.xb_root.xt_list; | |
0c044f0b TM |
4070 | int ret = 0; |
4071 | u32 name_hash = UINT_MAX, e_cpos = 0, num_clusters = 0; | |
4072 | u64 p_blkno = 0; | |
0c044f0b | 4073 | |
47bca495 | 4074 | if (!el->l_next_free_rec || !rec_func) |
0c044f0b TM |
4075 | return 0; |
4076 | ||
4077 | while (name_hash > 0) { | |
4078 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, | |
4079 | &e_cpos, &num_clusters, el); | |
4080 | if (ret) { | |
4081 | mlog_errno(ret); | |
47bca495 | 4082 | break; |
0c044f0b TM |
4083 | } |
4084 | ||
47bca495 TM |
4085 | ret = rec_func(inode, blk_bh, p_blkno, e_cpos, |
4086 | num_clusters, para); | |
0c044f0b | 4087 | if (ret) { |
a46fa684 TM |
4088 | if (ret != -ERANGE) |
4089 | mlog_errno(ret); | |
47bca495 | 4090 | break; |
0c044f0b TM |
4091 | } |
4092 | ||
4093 | if (e_cpos == 0) | |
4094 | break; | |
4095 | ||
4096 | name_hash = e_cpos - 1; | |
4097 | } | |
4098 | ||
47bca495 TM |
4099 | return ret; |
4100 | ||
4101 | } | |
4102 | ||
4103 | static int ocfs2_list_xattr_tree_rec(struct inode *inode, | |
4104 | struct buffer_head *root_bh, | |
4105 | u64 blkno, u32 cpos, u32 len, void *para) | |
4106 | { | |
4107 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
4108 | ocfs2_list_xattr_bucket, para); | |
4109 | } | |
4110 | ||
4111 | static int ocfs2_xattr_tree_list_index_block(struct inode *inode, | |
4112 | struct buffer_head *blk_bh, | |
4113 | char *buffer, | |
4114 | size_t buffer_size) | |
4115 | { | |
4116 | int ret; | |
4117 | struct ocfs2_xattr_tree_list xl = { | |
4118 | .buffer = buffer, | |
4119 | .buffer_size = buffer_size, | |
4120 | .result = 0, | |
4121 | }; | |
4122 | ||
4123 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
4124 | ocfs2_list_xattr_tree_rec, &xl); | |
4125 | if (ret) { | |
4126 | mlog_errno(ret); | |
4127 | goto out; | |
4128 | } | |
4129 | ||
936b8834 | 4130 | ret = xl.result; |
0c044f0b TM |
4131 | out: |
4132 | return ret; | |
4133 | } | |
01225596 TM |
4134 | |
4135 | static int cmp_xe(const void *a, const void *b) | |
4136 | { | |
4137 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4138 | u32 l_hash = le32_to_cpu(l->xe_name_hash); | |
4139 | u32 r_hash = le32_to_cpu(r->xe_name_hash); | |
4140 | ||
4141 | if (l_hash > r_hash) | |
4142 | return 1; | |
4143 | if (l_hash < r_hash) | |
4144 | return -1; | |
4145 | return 0; | |
4146 | } | |
4147 | ||
4148 | static void swap_xe(void *a, void *b, int size) | |
4149 | { | |
4150 | struct ocfs2_xattr_entry *l = a, *r = b, tmp; | |
4151 | ||
4152 | tmp = *l; | |
4153 | memcpy(l, r, sizeof(struct ocfs2_xattr_entry)); | |
4154 | memcpy(r, &tmp, sizeof(struct ocfs2_xattr_entry)); | |
4155 | } | |
4156 | ||
4157 | /* | |
4158 | * When the ocfs2_xattr_block is filled up, new bucket will be created | |
4159 | * and all the xattr entries will be moved to the new bucket. | |
178eeac3 JB |
4160 | * The header goes at the start of the bucket, and the names+values are |
4161 | * filled from the end. This is why *target starts as the last buffer. | |
01225596 TM |
4162 | * Note: we need to sort the entries since they are not saved in order |
4163 | * in the ocfs2_xattr_block. | |
4164 | */ | |
4165 | static void ocfs2_cp_xattr_block_to_bucket(struct inode *inode, | |
4166 | struct buffer_head *xb_bh, | |
178eeac3 | 4167 | struct ocfs2_xattr_bucket *bucket) |
01225596 TM |
4168 | { |
4169 | int i, blocksize = inode->i_sb->s_blocksize; | |
178eeac3 | 4170 | int blks = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 TM |
4171 | u16 offset, size, off_change; |
4172 | struct ocfs2_xattr_entry *xe; | |
4173 | struct ocfs2_xattr_block *xb = | |
4174 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4175 | struct ocfs2_xattr_header *xb_xh = &xb->xb_attrs.xb_header; | |
178eeac3 | 4176 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 4177 | u16 count = le16_to_cpu(xb_xh->xh_count); |
178eeac3 JB |
4178 | char *src = xb_bh->b_data; |
4179 | char *target = bucket_block(bucket, blks - 1); | |
01225596 | 4180 | |
402b4183 TM |
4181 | trace_ocfs2_cp_xattr_block_to_bucket_begin( |
4182 | (unsigned long long)xb_bh->b_blocknr, | |
4183 | (unsigned long long)bucket_blkno(bucket)); | |
178eeac3 JB |
4184 | |
4185 | for (i = 0; i < blks; i++) | |
4186 | memset(bucket_block(bucket, i), 0, blocksize); | |
01225596 | 4187 | |
01225596 TM |
4188 | /* |
4189 | * Since the xe_name_offset is based on ocfs2_xattr_header, | |
4190 | * there is a offset change corresponding to the change of | |
4191 | * ocfs2_xattr_header's position. | |
4192 | */ | |
4193 | off_change = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4194 | xe = &xb_xh->xh_entries[count - 1]; | |
4195 | offset = le16_to_cpu(xe->xe_name_offset) + off_change; | |
4196 | size = blocksize - offset; | |
4197 | ||
4198 | /* copy all the names and values. */ | |
01225596 TM |
4199 | memcpy(target + offset, src + offset, size); |
4200 | ||
4201 | /* Init new header now. */ | |
4202 | xh->xh_count = xb_xh->xh_count; | |
4203 | xh->xh_num_buckets = cpu_to_le16(1); | |
4204 | xh->xh_name_value_len = cpu_to_le16(size); | |
4205 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE - size); | |
4206 | ||
4207 | /* copy all the entries. */ | |
178eeac3 | 4208 | target = bucket_block(bucket, 0); |
01225596 TM |
4209 | offset = offsetof(struct ocfs2_xattr_header, xh_entries); |
4210 | size = count * sizeof(struct ocfs2_xattr_entry); | |
4211 | memcpy(target + offset, (char *)xb_xh + offset, size); | |
4212 | ||
4213 | /* Change the xe offset for all the xe because of the move. */ | |
4214 | off_change = OCFS2_XATTR_BUCKET_SIZE - blocksize + | |
4215 | offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
4216 | for (i = 0; i < count; i++) | |
4217 | le16_add_cpu(&xh->xh_entries[i].xe_name_offset, off_change); | |
4218 | ||
402b4183 | 4219 | trace_ocfs2_cp_xattr_block_to_bucket_end(offset, size, off_change); |
01225596 TM |
4220 | |
4221 | sort(target + offset, count, sizeof(struct ocfs2_xattr_entry), | |
4222 | cmp_xe, swap_xe); | |
4223 | } | |
4224 | ||
4225 | /* | |
4226 | * After we move xattr from block to index btree, we have to | |
4227 | * update ocfs2_xattr_search to the new xe and base. | |
4228 | * | |
4229 | * When the entry is in xattr block, xattr_bh indicates the storage place. | |
4230 | * While if the entry is in index b-tree, "bucket" indicates the | |
4231 | * real place of the xattr. | |
4232 | */ | |
178eeac3 JB |
4233 | static void ocfs2_xattr_update_xattr_search(struct inode *inode, |
4234 | struct ocfs2_xattr_search *xs, | |
4235 | struct buffer_head *old_bh) | |
01225596 | 4236 | { |
01225596 TM |
4237 | char *buf = old_bh->b_data; |
4238 | struct ocfs2_xattr_block *old_xb = (struct ocfs2_xattr_block *)buf; | |
4239 | struct ocfs2_xattr_header *old_xh = &old_xb->xb_attrs.xb_header; | |
178eeac3 | 4240 | int i; |
01225596 | 4241 | |
ba937127 | 4242 | xs->header = bucket_xh(xs->bucket); |
178eeac3 | 4243 | xs->base = bucket_block(xs->bucket, 0); |
01225596 TM |
4244 | xs->end = xs->base + inode->i_sb->s_blocksize; |
4245 | ||
178eeac3 JB |
4246 | if (xs->not_found) |
4247 | return; | |
01225596 | 4248 | |
178eeac3 JB |
4249 | i = xs->here - old_xh->xh_entries; |
4250 | xs->here = &xs->header->xh_entries[i]; | |
01225596 TM |
4251 | } |
4252 | ||
4253 | static int ocfs2_xattr_create_index_block(struct inode *inode, | |
78f30c31 TM |
4254 | struct ocfs2_xattr_search *xs, |
4255 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 4256 | { |
85db90e7 | 4257 | int ret; |
01225596 TM |
4258 | u32 bit_off, len; |
4259 | u64 blkno; | |
85db90e7 | 4260 | handle_t *handle = ctxt->handle; |
01225596 | 4261 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
01225596 TM |
4262 | struct buffer_head *xb_bh = xs->xattr_bh; |
4263 | struct ocfs2_xattr_block *xb = | |
4264 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
4265 | struct ocfs2_xattr_tree_root *xr; | |
4266 | u16 xb_flags = le16_to_cpu(xb->xb_flags); | |
01225596 | 4267 | |
402b4183 TM |
4268 | trace_ocfs2_xattr_create_index_block_begin( |
4269 | (unsigned long long)xb_bh->b_blocknr); | |
01225596 TM |
4270 | |
4271 | BUG_ON(xb_flags & OCFS2_XATTR_INDEXED); | |
178eeac3 | 4272 | BUG_ON(!xs->bucket); |
01225596 | 4273 | |
01225596 TM |
4274 | /* |
4275 | * XXX: | |
4276 | * We can use this lock for now, and maybe move to a dedicated mutex | |
4277 | * if performance becomes a problem later. | |
4278 | */ | |
4279 | down_write(&oi->ip_alloc_sem); | |
4280 | ||
0cf2f763 | 4281 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), xb_bh, |
84008972 | 4282 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4283 | if (ret) { |
4284 | mlog_errno(ret); | |
85db90e7 | 4285 | goto out; |
01225596 TM |
4286 | } |
4287 | ||
1ed9b777 | 4288 | ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, |
78f30c31 | 4289 | 1, 1, &bit_off, &len); |
01225596 TM |
4290 | if (ret) { |
4291 | mlog_errno(ret); | |
85db90e7 | 4292 | goto out; |
01225596 TM |
4293 | } |
4294 | ||
4295 | /* | |
4296 | * The bucket may spread in many blocks, and | |
4297 | * we will only touch the 1st block and the last block | |
4298 | * in the whole bucket(one for entry and one for data). | |
4299 | */ | |
4300 | blkno = ocfs2_clusters_to_blocks(inode->i_sb, bit_off); | |
4301 | ||
402b4183 | 4302 | trace_ocfs2_xattr_create_index_block((unsigned long long)blkno); |
01225596 | 4303 | |
178eeac3 JB |
4304 | ret = ocfs2_init_xattr_bucket(xs->bucket, blkno); |
4305 | if (ret) { | |
01225596 | 4306 | mlog_errno(ret); |
85db90e7 | 4307 | goto out; |
01225596 TM |
4308 | } |
4309 | ||
178eeac3 JB |
4310 | ret = ocfs2_xattr_bucket_journal_access(handle, xs->bucket, |
4311 | OCFS2_JOURNAL_ACCESS_CREATE); | |
01225596 TM |
4312 | if (ret) { |
4313 | mlog_errno(ret); | |
85db90e7 | 4314 | goto out; |
01225596 TM |
4315 | } |
4316 | ||
178eeac3 JB |
4317 | ocfs2_cp_xattr_block_to_bucket(inode, xb_bh, xs->bucket); |
4318 | ocfs2_xattr_bucket_journal_dirty(handle, xs->bucket); | |
01225596 | 4319 | |
178eeac3 | 4320 | ocfs2_xattr_update_xattr_search(inode, xs, xb_bh); |
01225596 TM |
4321 | |
4322 | /* Change from ocfs2_xattr_header to ocfs2_xattr_tree_root */ | |
4323 | memset(&xb->xb_attrs, 0, inode->i_sb->s_blocksize - | |
4324 | offsetof(struct ocfs2_xattr_block, xb_attrs)); | |
4325 | ||
4326 | xr = &xb->xb_attrs.xb_root; | |
4327 | xr->xt_clusters = cpu_to_le32(1); | |
4328 | xr->xt_last_eb_blk = 0; | |
4329 | xr->xt_list.l_tree_depth = 0; | |
4330 | xr->xt_list.l_count = cpu_to_le16(ocfs2_xattr_recs_per_xb(inode->i_sb)); | |
4331 | xr->xt_list.l_next_free_rec = cpu_to_le16(1); | |
4332 | ||
4333 | xr->xt_list.l_recs[0].e_cpos = 0; | |
4334 | xr->xt_list.l_recs[0].e_blkno = cpu_to_le64(blkno); | |
4335 | xr->xt_list.l_recs[0].e_leaf_clusters = cpu_to_le16(1); | |
4336 | ||
4337 | xb->xb_flags = cpu_to_le16(xb_flags | OCFS2_XATTR_INDEXED); | |
4338 | ||
85db90e7 | 4339 | ocfs2_journal_dirty(handle, xb_bh); |
01225596 | 4340 | |
85db90e7 | 4341 | out: |
01225596 TM |
4342 | up_write(&oi->ip_alloc_sem); |
4343 | ||
01225596 TM |
4344 | return ret; |
4345 | } | |
4346 | ||
4347 | static int cmp_xe_offset(const void *a, const void *b) | |
4348 | { | |
4349 | const struct ocfs2_xattr_entry *l = a, *r = b; | |
4350 | u32 l_name_offset = le16_to_cpu(l->xe_name_offset); | |
4351 | u32 r_name_offset = le16_to_cpu(r->xe_name_offset); | |
4352 | ||
4353 | if (l_name_offset < r_name_offset) | |
4354 | return 1; | |
4355 | if (l_name_offset > r_name_offset) | |
4356 | return -1; | |
4357 | return 0; | |
4358 | } | |
4359 | ||
4360 | /* | |
4361 | * defrag a xattr bucket if we find that the bucket has some | |
4362 | * holes beteen name/value pairs. | |
4363 | * We will move all the name/value pairs to the end of the bucket | |
4364 | * so that we can spare some space for insertion. | |
4365 | */ | |
4366 | static int ocfs2_defrag_xattr_bucket(struct inode *inode, | |
85db90e7 | 4367 | handle_t *handle, |
01225596 TM |
4368 | struct ocfs2_xattr_bucket *bucket) |
4369 | { | |
4370 | int ret, i; | |
199799a3 | 4371 | size_t end, offset, len; |
01225596 TM |
4372 | struct ocfs2_xattr_header *xh; |
4373 | char *entries, *buf, *bucket_buf = NULL; | |
9c7759aa | 4374 | u64 blkno = bucket_blkno(bucket); |
01225596 | 4375 | u16 xh_free_start; |
01225596 | 4376 | size_t blocksize = inode->i_sb->s_blocksize; |
01225596 | 4377 | struct ocfs2_xattr_entry *xe; |
01225596 TM |
4378 | |
4379 | /* | |
4380 | * In order to make the operation more efficient and generic, | |
4381 | * we copy all the blocks into a contiguous memory and do the | |
4382 | * defragment there, so if anything is error, we will not touch | |
4383 | * the real block. | |
4384 | */ | |
4385 | bucket_buf = kmalloc(OCFS2_XATTR_BUCKET_SIZE, GFP_NOFS); | |
4386 | if (!bucket_buf) { | |
4387 | ret = -EIO; | |
4388 | goto out; | |
4389 | } | |
4390 | ||
4391 | buf = bucket_buf; | |
1c32a2fd TM |
4392 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4393 | memcpy(buf, bucket_block(bucket, i), blocksize); | |
01225596 | 4394 | |
1c32a2fd | 4395 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, |
161d6f30 JB |
4396 | OCFS2_JOURNAL_ACCESS_WRITE); |
4397 | if (ret < 0) { | |
4398 | mlog_errno(ret); | |
85db90e7 | 4399 | goto out; |
01225596 TM |
4400 | } |
4401 | ||
4402 | xh = (struct ocfs2_xattr_header *)bucket_buf; | |
4403 | entries = (char *)xh->xh_entries; | |
4404 | xh_free_start = le16_to_cpu(xh->xh_free_start); | |
4405 | ||
402b4183 | 4406 | trace_ocfs2_defrag_xattr_bucket( |
de29c085 MF |
4407 | (unsigned long long)blkno, le16_to_cpu(xh->xh_count), |
4408 | xh_free_start, le16_to_cpu(xh->xh_name_value_len)); | |
01225596 TM |
4409 | |
4410 | /* | |
4411 | * sort all the entries by their offset. | |
4412 | * the largest will be the first, so that we can | |
4413 | * move them to the end one by one. | |
4414 | */ | |
4415 | sort(entries, le16_to_cpu(xh->xh_count), | |
4416 | sizeof(struct ocfs2_xattr_entry), | |
4417 | cmp_xe_offset, swap_xe); | |
4418 | ||
4419 | /* Move all name/values to the end of the bucket. */ | |
4420 | xe = xh->xh_entries; | |
4421 | end = OCFS2_XATTR_BUCKET_SIZE; | |
4422 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++, xe++) { | |
4423 | offset = le16_to_cpu(xe->xe_name_offset); | |
199799a3 | 4424 | len = namevalue_size_xe(xe); |
01225596 TM |
4425 | |
4426 | /* | |
4427 | * We must make sure that the name/value pair | |
4428 | * exist in the same block. So adjust end to | |
4429 | * the previous block end if needed. | |
4430 | */ | |
4431 | if (((end - len) / blocksize != | |
4432 | (end - 1) / blocksize)) | |
4433 | end = end - end % blocksize; | |
4434 | ||
4435 | if (end > offset + len) { | |
4436 | memmove(bucket_buf + end - len, | |
4437 | bucket_buf + offset, len); | |
4438 | xe->xe_name_offset = cpu_to_le16(end - len); | |
4439 | } | |
4440 | ||
4441 | mlog_bug_on_msg(end < offset + len, "Defrag check failed for " | |
4442 | "bucket %llu\n", (unsigned long long)blkno); | |
4443 | ||
4444 | end -= len; | |
4445 | } | |
4446 | ||
4447 | mlog_bug_on_msg(xh_free_start > end, "Defrag check failed for " | |
4448 | "bucket %llu\n", (unsigned long long)blkno); | |
4449 | ||
4450 | if (xh_free_start == end) | |
85db90e7 | 4451 | goto out; |
01225596 TM |
4452 | |
4453 | memset(bucket_buf + xh_free_start, 0, end - xh_free_start); | |
4454 | xh->xh_free_start = cpu_to_le16(end); | |
4455 | ||
4456 | /* sort the entries by their name_hash. */ | |
4457 | sort(entries, le16_to_cpu(xh->xh_count), | |
4458 | sizeof(struct ocfs2_xattr_entry), | |
4459 | cmp_xe, swap_xe); | |
4460 | ||
4461 | buf = bucket_buf; | |
1c32a2fd TM |
4462 | for (i = 0; i < bucket->bu_blocks; i++, buf += blocksize) |
4463 | memcpy(bucket_block(bucket, i), buf, blocksize); | |
4464 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
01225596 | 4465 | |
01225596 | 4466 | out: |
01225596 TM |
4467 | kfree(bucket_buf); |
4468 | return ret; | |
4469 | } | |
4470 | ||
4471 | /* | |
b5c03e74 JB |
4472 | * prev_blkno points to the start of an existing extent. new_blkno |
4473 | * points to a newly allocated extent. Because we know each of our | |
4474 | * clusters contains more than bucket, we can easily split one cluster | |
4475 | * at a bucket boundary. So we take the last cluster of the existing | |
4476 | * extent and split it down the middle. We move the last half of the | |
4477 | * buckets in the last cluster of the existing extent over to the new | |
4478 | * extent. | |
01225596 | 4479 | * |
b5c03e74 JB |
4480 | * first_bh is the buffer at prev_blkno so we can update the existing |
4481 | * extent's bucket count. header_bh is the bucket were we were hoping | |
4482 | * to insert our xattr. If the bucket move places the target in the new | |
4483 | * extent, we'll update first_bh and header_bh after modifying the old | |
4484 | * extent. | |
4485 | * | |
4486 | * first_hash will be set as the 1st xe's name_hash in the new extent. | |
01225596 TM |
4487 | */ |
4488 | static int ocfs2_mv_xattr_bucket_cross_cluster(struct inode *inode, | |
4489 | handle_t *handle, | |
41cb8148 JB |
4490 | struct ocfs2_xattr_bucket *first, |
4491 | struct ocfs2_xattr_bucket *target, | |
01225596 | 4492 | u64 new_blkno, |
01225596 TM |
4493 | u32 num_clusters, |
4494 | u32 *first_hash) | |
4495 | { | |
c58b6032 | 4496 | int ret; |
41cb8148 JB |
4497 | struct super_block *sb = inode->i_sb; |
4498 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(sb); | |
4499 | int num_buckets = ocfs2_xattr_buckets_per_cluster(OCFS2_SB(sb)); | |
b5c03e74 | 4500 | int to_move = num_buckets / 2; |
c58b6032 | 4501 | u64 src_blkno; |
41cb8148 JB |
4502 | u64 last_cluster_blkno = bucket_blkno(first) + |
4503 | ((num_clusters - 1) * ocfs2_clusters_to_blocks(sb, 1)); | |
01225596 | 4504 | |
41cb8148 JB |
4505 | BUG_ON(le16_to_cpu(bucket_xh(first)->xh_num_buckets) < num_buckets); |
4506 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE == OCFS2_SB(sb)->s_clustersize); | |
01225596 | 4507 | |
402b4183 TM |
4508 | trace_ocfs2_mv_xattr_bucket_cross_cluster( |
4509 | (unsigned long long)last_cluster_blkno, | |
4510 | (unsigned long long)new_blkno); | |
01225596 | 4511 | |
41cb8148 | 4512 | ret = ocfs2_mv_xattr_buckets(inode, handle, bucket_blkno(first), |
c58b6032 JB |
4513 | last_cluster_blkno, new_blkno, |
4514 | to_move, first_hash); | |
01225596 TM |
4515 | if (ret) { |
4516 | mlog_errno(ret); | |
4517 | goto out; | |
4518 | } | |
4519 | ||
c58b6032 JB |
4520 | /* This is the first bucket that got moved */ |
4521 | src_blkno = last_cluster_blkno + (to_move * blks_per_bucket); | |
01225596 | 4522 | |
b5c03e74 | 4523 | /* |
c58b6032 | 4524 | * If the target bucket was part of the moved buckets, we need to |
41cb8148 | 4525 | * update first and target. |
b5c03e74 | 4526 | */ |
41cb8148 | 4527 | if (bucket_blkno(target) >= src_blkno) { |
b5c03e74 JB |
4528 | /* Find the block for the new target bucket */ |
4529 | src_blkno = new_blkno + | |
41cb8148 JB |
4530 | (bucket_blkno(target) - src_blkno); |
4531 | ||
4532 | ocfs2_xattr_bucket_relse(first); | |
4533 | ocfs2_xattr_bucket_relse(target); | |
01225596 | 4534 | |
b5c03e74 | 4535 | /* |
c58b6032 | 4536 | * These shouldn't fail - the buffers are in the |
b5c03e74 JB |
4537 | * journal from ocfs2_cp_xattr_bucket(). |
4538 | */ | |
41cb8148 | 4539 | ret = ocfs2_read_xattr_bucket(first, new_blkno); |
c58b6032 JB |
4540 | if (ret) { |
4541 | mlog_errno(ret); | |
4542 | goto out; | |
4543 | } | |
41cb8148 JB |
4544 | ret = ocfs2_read_xattr_bucket(target, src_blkno); |
4545 | if (ret) | |
b5c03e74 | 4546 | mlog_errno(ret); |
01225596 | 4547 | |
01225596 TM |
4548 | } |
4549 | ||
01225596 | 4550 | out: |
01225596 TM |
4551 | return ret; |
4552 | } | |
4553 | ||
01225596 | 4554 | /* |
80bcaf34 TM |
4555 | * Find the suitable pos when we divide a bucket into 2. |
4556 | * We have to make sure the xattrs with the same hash value exist | |
4557 | * in the same bucket. | |
4558 | * | |
4559 | * If this ocfs2_xattr_header covers more than one hash value, find a | |
4560 | * place where the hash value changes. Try to find the most even split. | |
4561 | * The most common case is that all entries have different hash values, | |
4562 | * and the first check we make will find a place to split. | |
4563 | */ | |
4564 | static int ocfs2_xattr_find_divide_pos(struct ocfs2_xattr_header *xh) | |
4565 | { | |
4566 | struct ocfs2_xattr_entry *entries = xh->xh_entries; | |
4567 | int count = le16_to_cpu(xh->xh_count); | |
4568 | int delta, middle = count / 2; | |
4569 | ||
4570 | /* | |
4571 | * We start at the middle. Each step gets farther away in both | |
4572 | * directions. We therefore hit the change in hash value | |
4573 | * nearest to the middle. Note that this loop does not execute for | |
4574 | * count < 2. | |
4575 | */ | |
4576 | for (delta = 0; delta < middle; delta++) { | |
4577 | /* Let's check delta earlier than middle */ | |
4578 | if (cmp_xe(&entries[middle - delta - 1], | |
4579 | &entries[middle - delta])) | |
4580 | return middle - delta; | |
4581 | ||
4582 | /* For even counts, don't walk off the end */ | |
4583 | if ((middle + delta + 1) == count) | |
4584 | continue; | |
4585 | ||
4586 | /* Now try delta past middle */ | |
4587 | if (cmp_xe(&entries[middle + delta], | |
4588 | &entries[middle + delta + 1])) | |
4589 | return middle + delta + 1; | |
4590 | } | |
4591 | ||
4592 | /* Every entry had the same hash */ | |
4593 | return count; | |
4594 | } | |
4595 | ||
4596 | /* | |
4597 | * Move some xattrs in old bucket(blk) to new bucket(new_blk). | |
01225596 | 4598 | * first_hash will record the 1st hash of the new bucket. |
80bcaf34 TM |
4599 | * |
4600 | * Normally half of the xattrs will be moved. But we have to make | |
4601 | * sure that the xattrs with the same hash value are stored in the | |
4602 | * same bucket. If all the xattrs in this bucket have the same hash | |
4603 | * value, the new bucket will be initialized as an empty one and the | |
4604 | * first_hash will be initialized as (hash_value+1). | |
01225596 | 4605 | */ |
80bcaf34 TM |
4606 | static int ocfs2_divide_xattr_bucket(struct inode *inode, |
4607 | handle_t *handle, | |
4608 | u64 blk, | |
4609 | u64 new_blk, | |
4610 | u32 *first_hash, | |
4611 | int new_bucket_head) | |
01225596 TM |
4612 | { |
4613 | int ret, i; | |
199799a3 | 4614 | int count, start, len, name_value_len = 0, name_offset = 0; |
ba937127 | 4615 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4616 | struct ocfs2_xattr_header *xh; |
4617 | struct ocfs2_xattr_entry *xe; | |
4618 | int blocksize = inode->i_sb->s_blocksize; | |
4619 | ||
402b4183 TM |
4620 | trace_ocfs2_divide_xattr_bucket_begin((unsigned long long)blk, |
4621 | (unsigned long long)new_blk); | |
01225596 | 4622 | |
ba937127 JB |
4623 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4624 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4625 | if (!s_bucket || !t_bucket) { | |
4626 | ret = -ENOMEM; | |
4627 | mlog_errno(ret); | |
4628 | goto out; | |
4629 | } | |
01225596 | 4630 | |
ba937127 | 4631 | ret = ocfs2_read_xattr_bucket(s_bucket, blk); |
01225596 TM |
4632 | if (ret) { |
4633 | mlog_errno(ret); | |
4634 | goto out; | |
4635 | } | |
4636 | ||
ba937127 | 4637 | ret = ocfs2_xattr_bucket_journal_access(handle, s_bucket, |
1224be02 | 4638 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
4639 | if (ret) { |
4640 | mlog_errno(ret); | |
4641 | goto out; | |
4642 | } | |
4643 | ||
784b816a JB |
4644 | /* |
4645 | * Even if !new_bucket_head, we're overwriting t_bucket. Thus, | |
4646 | * there's no need to read it. | |
4647 | */ | |
ba937127 | 4648 | ret = ocfs2_init_xattr_bucket(t_bucket, new_blk); |
01225596 TM |
4649 | if (ret) { |
4650 | mlog_errno(ret); | |
4651 | goto out; | |
4652 | } | |
4653 | ||
2b656c1d JB |
4654 | /* |
4655 | * Hey, if we're overwriting t_bucket, what difference does | |
4656 | * ACCESS_CREATE vs ACCESS_WRITE make? See the comment in the | |
4657 | * same part of ocfs2_cp_xattr_bucket(). | |
4658 | */ | |
ba937127 | 4659 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4660 | new_bucket_head ? |
4661 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4662 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4663 | if (ret) { | |
4664 | mlog_errno(ret); | |
4665 | goto out; | |
01225596 TM |
4666 | } |
4667 | ||
ba937127 | 4668 | xh = bucket_xh(s_bucket); |
80bcaf34 TM |
4669 | count = le16_to_cpu(xh->xh_count); |
4670 | start = ocfs2_xattr_find_divide_pos(xh); | |
4671 | ||
4672 | if (start == count) { | |
4673 | xe = &xh->xh_entries[start-1]; | |
4674 | ||
4675 | /* | |
4676 | * initialized a new empty bucket here. | |
4677 | * The hash value is set as one larger than | |
4678 | * that of the last entry in the previous bucket. | |
4679 | */ | |
ba937127 JB |
4680 | for (i = 0; i < t_bucket->bu_blocks; i++) |
4681 | memset(bucket_block(t_bucket, i), 0, blocksize); | |
80bcaf34 | 4682 | |
ba937127 | 4683 | xh = bucket_xh(t_bucket); |
80bcaf34 TM |
4684 | xh->xh_free_start = cpu_to_le16(blocksize); |
4685 | xh->xh_entries[0].xe_name_hash = xe->xe_name_hash; | |
4686 | le32_add_cpu(&xh->xh_entries[0].xe_name_hash, 1); | |
4687 | ||
4688 | goto set_num_buckets; | |
4689 | } | |
4690 | ||
01225596 | 4691 | /* copy the whole bucket to the new first. */ |
ba937127 | 4692 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
01225596 TM |
4693 | |
4694 | /* update the new bucket. */ | |
ba937127 | 4695 | xh = bucket_xh(t_bucket); |
01225596 TM |
4696 | |
4697 | /* | |
4698 | * Calculate the total name/value len and xh_free_start for | |
4699 | * the old bucket first. | |
4700 | */ | |
4701 | name_offset = OCFS2_XATTR_BUCKET_SIZE; | |
4702 | name_value_len = 0; | |
4703 | for (i = 0; i < start; i++) { | |
4704 | xe = &xh->xh_entries[i]; | |
199799a3 | 4705 | name_value_len += namevalue_size_xe(xe); |
01225596 TM |
4706 | if (le16_to_cpu(xe->xe_name_offset) < name_offset) |
4707 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
4708 | } | |
4709 | ||
4710 | /* | |
4711 | * Now begin the modification to the new bucket. | |
4712 | * | |
4713 | * In the new bucket, We just move the xattr entry to the beginning | |
4714 | * and don't touch the name/value. So there will be some holes in the | |
4715 | * bucket, and they will be removed when ocfs2_defrag_xattr_bucket is | |
4716 | * called. | |
4717 | */ | |
4718 | xe = &xh->xh_entries[start]; | |
4719 | len = sizeof(struct ocfs2_xattr_entry) * (count - start); | |
402b4183 TM |
4720 | trace_ocfs2_divide_xattr_bucket_move(len, |
4721 | (int)((char *)xe - (char *)xh), | |
4722 | (int)((char *)xh->xh_entries - (char *)xh)); | |
01225596 TM |
4723 | memmove((char *)xh->xh_entries, (char *)xe, len); |
4724 | xe = &xh->xh_entries[count - start]; | |
4725 | len = sizeof(struct ocfs2_xattr_entry) * start; | |
4726 | memset((char *)xe, 0, len); | |
4727 | ||
4728 | le16_add_cpu(&xh->xh_count, -start); | |
4729 | le16_add_cpu(&xh->xh_name_value_len, -name_value_len); | |
4730 | ||
4731 | /* Calculate xh_free_start for the new bucket. */ | |
4732 | xh->xh_free_start = cpu_to_le16(OCFS2_XATTR_BUCKET_SIZE); | |
4733 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
4734 | xe = &xh->xh_entries[i]; | |
01225596 TM |
4735 | if (le16_to_cpu(xe->xe_name_offset) < |
4736 | le16_to_cpu(xh->xh_free_start)) | |
4737 | xh->xh_free_start = xe->xe_name_offset; | |
4738 | } | |
4739 | ||
80bcaf34 | 4740 | set_num_buckets: |
01225596 TM |
4741 | /* set xh->xh_num_buckets for the new xh. */ |
4742 | if (new_bucket_head) | |
4743 | xh->xh_num_buckets = cpu_to_le16(1); | |
4744 | else | |
4745 | xh->xh_num_buckets = 0; | |
4746 | ||
ba937127 | 4747 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); |
01225596 TM |
4748 | |
4749 | /* store the first_hash of the new bucket. */ | |
4750 | if (first_hash) | |
4751 | *first_hash = le32_to_cpu(xh->xh_entries[0].xe_name_hash); | |
4752 | ||
4753 | /* | |
80bcaf34 TM |
4754 | * Now only update the 1st block of the old bucket. If we |
4755 | * just added a new empty bucket, there is no need to modify | |
4756 | * it. | |
01225596 | 4757 | */ |
80bcaf34 TM |
4758 | if (start == count) |
4759 | goto out; | |
4760 | ||
ba937127 | 4761 | xh = bucket_xh(s_bucket); |
01225596 TM |
4762 | memset(&xh->xh_entries[start], 0, |
4763 | sizeof(struct ocfs2_xattr_entry) * (count - start)); | |
4764 | xh->xh_count = cpu_to_le16(start); | |
4765 | xh->xh_free_start = cpu_to_le16(name_offset); | |
4766 | xh->xh_name_value_len = cpu_to_le16(name_value_len); | |
4767 | ||
ba937127 | 4768 | ocfs2_xattr_bucket_journal_dirty(handle, s_bucket); |
01225596 TM |
4769 | |
4770 | out: | |
ba937127 JB |
4771 | ocfs2_xattr_bucket_free(s_bucket); |
4772 | ocfs2_xattr_bucket_free(t_bucket); | |
01225596 TM |
4773 | |
4774 | return ret; | |
4775 | } | |
4776 | ||
4777 | /* | |
4778 | * Copy xattr from one bucket to another bucket. | |
4779 | * | |
4780 | * The caller must make sure that the journal transaction | |
4781 | * has enough space for journaling. | |
4782 | */ | |
4783 | static int ocfs2_cp_xattr_bucket(struct inode *inode, | |
4784 | handle_t *handle, | |
4785 | u64 s_blkno, | |
4786 | u64 t_blkno, | |
4787 | int t_is_new) | |
4788 | { | |
4980c6da | 4789 | int ret; |
ba937127 | 4790 | struct ocfs2_xattr_bucket *s_bucket = NULL, *t_bucket = NULL; |
01225596 TM |
4791 | |
4792 | BUG_ON(s_blkno == t_blkno); | |
4793 | ||
402b4183 TM |
4794 | trace_ocfs2_cp_xattr_bucket((unsigned long long)s_blkno, |
4795 | (unsigned long long)t_blkno, | |
4796 | t_is_new); | |
01225596 | 4797 | |
ba937127 JB |
4798 | s_bucket = ocfs2_xattr_bucket_new(inode); |
4799 | t_bucket = ocfs2_xattr_bucket_new(inode); | |
4800 | if (!s_bucket || !t_bucket) { | |
4801 | ret = -ENOMEM; | |
4802 | mlog_errno(ret); | |
4803 | goto out; | |
4804 | } | |
92de109a | 4805 | |
ba937127 | 4806 | ret = ocfs2_read_xattr_bucket(s_bucket, s_blkno); |
01225596 TM |
4807 | if (ret) |
4808 | goto out; | |
4809 | ||
784b816a JB |
4810 | /* |
4811 | * Even if !t_is_new, we're overwriting t_bucket. Thus, | |
4812 | * there's no need to read it. | |
4813 | */ | |
ba937127 | 4814 | ret = ocfs2_init_xattr_bucket(t_bucket, t_blkno); |
01225596 TM |
4815 | if (ret) |
4816 | goto out; | |
4817 | ||
2b656c1d JB |
4818 | /* |
4819 | * Hey, if we're overwriting t_bucket, what difference does | |
4820 | * ACCESS_CREATE vs ACCESS_WRITE make? Well, if we allocated a new | |
874d65af JB |
4821 | * cluster to fill, we came here from |
4822 | * ocfs2_mv_xattr_buckets(), and it is really new - | |
4823 | * ACCESS_CREATE is required. But we also might have moved data | |
4824 | * out of t_bucket before extending back into it. | |
4825 | * ocfs2_add_new_xattr_bucket() can do this - its call to | |
4826 | * ocfs2_add_new_xattr_cluster() may have created a new extent | |
2b656c1d JB |
4827 | * and copied out the end of the old extent. Then it re-extends |
4828 | * the old extent back to create space for new xattrs. That's | |
4829 | * how we get here, and the bucket isn't really new. | |
4830 | */ | |
ba937127 | 4831 | ret = ocfs2_xattr_bucket_journal_access(handle, t_bucket, |
1224be02 JB |
4832 | t_is_new ? |
4833 | OCFS2_JOURNAL_ACCESS_CREATE : | |
4834 | OCFS2_JOURNAL_ACCESS_WRITE); | |
4835 | if (ret) | |
4836 | goto out; | |
01225596 | 4837 | |
ba937127 JB |
4838 | ocfs2_xattr_bucket_copy_data(t_bucket, s_bucket); |
4839 | ocfs2_xattr_bucket_journal_dirty(handle, t_bucket); | |
01225596 TM |
4840 | |
4841 | out: | |
ba937127 JB |
4842 | ocfs2_xattr_bucket_free(t_bucket); |
4843 | ocfs2_xattr_bucket_free(s_bucket); | |
01225596 TM |
4844 | |
4845 | return ret; | |
4846 | } | |
4847 | ||
4848 | /* | |
874d65af JB |
4849 | * src_blk points to the start of an existing extent. last_blk points to |
4850 | * last cluster in that extent. to_blk points to a newly allocated | |
54ecb6b6 JB |
4851 | * extent. We copy the buckets from the cluster at last_blk to the new |
4852 | * extent. If start_bucket is non-zero, we skip that many buckets before | |
4853 | * we start copying. The new extent's xh_num_buckets gets set to the | |
4854 | * number of buckets we copied. The old extent's xh_num_buckets shrinks | |
4855 | * by the same amount. | |
01225596 | 4856 | */ |
54ecb6b6 JB |
4857 | static int ocfs2_mv_xattr_buckets(struct inode *inode, handle_t *handle, |
4858 | u64 src_blk, u64 last_blk, u64 to_blk, | |
4859 | unsigned int start_bucket, | |
4860 | u32 *first_hash) | |
01225596 TM |
4861 | { |
4862 | int i, ret, credits; | |
4863 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
15d60929 | 4864 | int blks_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); |
01225596 | 4865 | int num_buckets = ocfs2_xattr_buckets_per_cluster(osb); |
15d60929 | 4866 | struct ocfs2_xattr_bucket *old_first, *new_first; |
01225596 | 4867 | |
402b4183 TM |
4868 | trace_ocfs2_mv_xattr_buckets((unsigned long long)last_blk, |
4869 | (unsigned long long)to_blk); | |
01225596 | 4870 | |
54ecb6b6 JB |
4871 | BUG_ON(start_bucket >= num_buckets); |
4872 | if (start_bucket) { | |
4873 | num_buckets -= start_bucket; | |
4874 | last_blk += (start_bucket * blks_per_bucket); | |
4875 | } | |
4876 | ||
15d60929 JB |
4877 | /* The first bucket of the original extent */ |
4878 | old_first = ocfs2_xattr_bucket_new(inode); | |
4879 | /* The first bucket of the new extent */ | |
4880 | new_first = ocfs2_xattr_bucket_new(inode); | |
4881 | if (!old_first || !new_first) { | |
4882 | ret = -ENOMEM; | |
4883 | mlog_errno(ret); | |
4884 | goto out; | |
4885 | } | |
4886 | ||
874d65af | 4887 | ret = ocfs2_read_xattr_bucket(old_first, src_blk); |
15d60929 JB |
4888 | if (ret) { |
4889 | mlog_errno(ret); | |
4890 | goto out; | |
4891 | } | |
4892 | ||
01225596 | 4893 | /* |
54ecb6b6 JB |
4894 | * We need to update the first bucket of the old extent and all |
4895 | * the buckets going to the new extent. | |
01225596 | 4896 | */ |
c901fb00 | 4897 | credits = ((num_buckets + 1) * blks_per_bucket); |
01225596 TM |
4898 | ret = ocfs2_extend_trans(handle, credits); |
4899 | if (ret) { | |
4900 | mlog_errno(ret); | |
4901 | goto out; | |
4902 | } | |
4903 | ||
15d60929 JB |
4904 | ret = ocfs2_xattr_bucket_journal_access(handle, old_first, |
4905 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4906 | if (ret) { |
4907 | mlog_errno(ret); | |
4908 | goto out; | |
4909 | } | |
4910 | ||
4911 | for (i = 0; i < num_buckets; i++) { | |
4912 | ret = ocfs2_cp_xattr_bucket(inode, handle, | |
874d65af | 4913 | last_blk + (i * blks_per_bucket), |
15d60929 JB |
4914 | to_blk + (i * blks_per_bucket), |
4915 | 1); | |
01225596 TM |
4916 | if (ret) { |
4917 | mlog_errno(ret); | |
4918 | goto out; | |
4919 | } | |
01225596 TM |
4920 | } |
4921 | ||
15d60929 JB |
4922 | /* |
4923 | * Get the new bucket ready before we dirty anything | |
4924 | * (This actually shouldn't fail, because we already dirtied | |
4925 | * it once in ocfs2_cp_xattr_bucket()). | |
4926 | */ | |
4927 | ret = ocfs2_read_xattr_bucket(new_first, to_blk); | |
4928 | if (ret) { | |
01225596 TM |
4929 | mlog_errno(ret); |
4930 | goto out; | |
4931 | } | |
15d60929 JB |
4932 | ret = ocfs2_xattr_bucket_journal_access(handle, new_first, |
4933 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
4934 | if (ret) { |
4935 | mlog_errno(ret); | |
4936 | goto out; | |
4937 | } | |
4938 | ||
15d60929 JB |
4939 | /* Now update the headers */ |
4940 | le16_add_cpu(&bucket_xh(old_first)->xh_num_buckets, -num_buckets); | |
4941 | ocfs2_xattr_bucket_journal_dirty(handle, old_first); | |
01225596 | 4942 | |
15d60929 JB |
4943 | bucket_xh(new_first)->xh_num_buckets = cpu_to_le16(num_buckets); |
4944 | ocfs2_xattr_bucket_journal_dirty(handle, new_first); | |
01225596 TM |
4945 | |
4946 | if (first_hash) | |
15d60929 JB |
4947 | *first_hash = le32_to_cpu(bucket_xh(new_first)->xh_entries[0].xe_name_hash); |
4948 | ||
01225596 | 4949 | out: |
15d60929 JB |
4950 | ocfs2_xattr_bucket_free(new_first); |
4951 | ocfs2_xattr_bucket_free(old_first); | |
01225596 TM |
4952 | return ret; |
4953 | } | |
4954 | ||
4955 | /* | |
80bcaf34 | 4956 | * Move some xattrs in this cluster to the new cluster. |
01225596 TM |
4957 | * This function should only be called when bucket size == cluster size. |
4958 | * Otherwise ocfs2_mv_xattr_bucket_cross_cluster should be used instead. | |
4959 | */ | |
80bcaf34 TM |
4960 | static int ocfs2_divide_xattr_cluster(struct inode *inode, |
4961 | handle_t *handle, | |
4962 | u64 prev_blk, | |
4963 | u64 new_blk, | |
4964 | u32 *first_hash) | |
01225596 TM |
4965 | { |
4966 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
c901fb00 | 4967 | int ret, credits = 2 * blk_per_bucket; |
01225596 TM |
4968 | |
4969 | BUG_ON(OCFS2_XATTR_BUCKET_SIZE < OCFS2_SB(inode->i_sb)->s_clustersize); | |
4970 | ||
4971 | ret = ocfs2_extend_trans(handle, credits); | |
4972 | if (ret) { | |
4973 | mlog_errno(ret); | |
4974 | return ret; | |
4975 | } | |
4976 | ||
4977 | /* Move half of the xattr in start_blk to the next bucket. */ | |
80bcaf34 TM |
4978 | return ocfs2_divide_xattr_bucket(inode, handle, prev_blk, |
4979 | new_blk, first_hash, 1); | |
01225596 TM |
4980 | } |
4981 | ||
4982 | /* | |
4983 | * Move some xattrs from the old cluster to the new one since they are not | |
4984 | * contiguous in ocfs2 xattr tree. | |
4985 | * | |
4986 | * new_blk starts a new separate cluster, and we will move some xattrs from | |
4987 | * prev_blk to it. v_start will be set as the first name hash value in this | |
4988 | * new cluster so that it can be used as e_cpos during tree insertion and | |
4989 | * don't collide with our original b-tree operations. first_bh and header_bh | |
4990 | * will also be updated since they will be used in ocfs2_extend_xattr_bucket | |
4991 | * to extend the insert bucket. | |
4992 | * | |
4993 | * The problem is how much xattr should we move to the new one and when should | |
4994 | * we update first_bh and header_bh? | |
4995 | * 1. If cluster size > bucket size, that means the previous cluster has more | |
4996 | * than 1 bucket, so just move half nums of bucket into the new cluster and | |
4997 | * update the first_bh and header_bh if the insert bucket has been moved | |
4998 | * to the new cluster. | |
4999 | * 2. If cluster_size == bucket_size: | |
5000 | * a) If the previous extent rec has more than one cluster and the insert | |
5001 | * place isn't in the last cluster, copy the entire last cluster to the | |
5002 | * new one. This time, we don't need to upate the first_bh and header_bh | |
5003 | * since they will not be moved into the new cluster. | |
5004 | * b) Otherwise, move the bottom half of the xattrs in the last cluster into | |
5005 | * the new one. And we set the extend flag to zero if the insert place is | |
5006 | * moved into the new allocated cluster since no extend is needed. | |
5007 | */ | |
5008 | static int ocfs2_adjust_xattr_cross_cluster(struct inode *inode, | |
5009 | handle_t *handle, | |
012ee910 JB |
5010 | struct ocfs2_xattr_bucket *first, |
5011 | struct ocfs2_xattr_bucket *target, | |
01225596 | 5012 | u64 new_blk, |
01225596 TM |
5013 | u32 prev_clusters, |
5014 | u32 *v_start, | |
5015 | int *extend) | |
5016 | { | |
92cf3adf | 5017 | int ret; |
01225596 | 5018 | |
402b4183 TM |
5019 | trace_ocfs2_adjust_xattr_cross_cluster( |
5020 | (unsigned long long)bucket_blkno(first), | |
5021 | (unsigned long long)new_blk, prev_clusters); | |
01225596 | 5022 | |
41cb8148 | 5023 | if (ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)) > 1) { |
01225596 TM |
5024 | ret = ocfs2_mv_xattr_bucket_cross_cluster(inode, |
5025 | handle, | |
41cb8148 | 5026 | first, target, |
01225596 | 5027 | new_blk, |
01225596 TM |
5028 | prev_clusters, |
5029 | v_start); | |
012ee910 | 5030 | if (ret) |
41cb8148 | 5031 | mlog_errno(ret); |
41cb8148 | 5032 | } else { |
92cf3adf JB |
5033 | /* The start of the last cluster in the first extent */ |
5034 | u64 last_blk = bucket_blkno(first) + | |
5035 | ((prev_clusters - 1) * | |
5036 | ocfs2_clusters_to_blocks(inode->i_sb, 1)); | |
01225596 | 5037 | |
012ee910 | 5038 | if (prev_clusters > 1 && bucket_blkno(target) != last_blk) { |
874d65af | 5039 | ret = ocfs2_mv_xattr_buckets(inode, handle, |
92cf3adf | 5040 | bucket_blkno(first), |
54ecb6b6 | 5041 | last_blk, new_blk, 0, |
01225596 | 5042 | v_start); |
012ee910 JB |
5043 | if (ret) |
5044 | mlog_errno(ret); | |
5045 | } else { | |
80bcaf34 TM |
5046 | ret = ocfs2_divide_xattr_cluster(inode, handle, |
5047 | last_blk, new_blk, | |
5048 | v_start); | |
012ee910 JB |
5049 | if (ret) |
5050 | mlog_errno(ret); | |
01225596 | 5051 | |
92cf3adf | 5052 | if ((bucket_blkno(target) == last_blk) && extend) |
01225596 TM |
5053 | *extend = 0; |
5054 | } | |
5055 | } | |
5056 | ||
5057 | return ret; | |
5058 | } | |
5059 | ||
5060 | /* | |
5061 | * Add a new cluster for xattr storage. | |
5062 | * | |
5063 | * If the new cluster is contiguous with the previous one, it will be | |
5064 | * appended to the same extent record, and num_clusters will be updated. | |
5065 | * If not, we will insert a new extent for it and move some xattrs in | |
5066 | * the last cluster into the new allocated one. | |
5067 | * We also need to limit the maximum size of a btree leaf, otherwise we'll | |
5068 | * lose the benefits of hashing because we'll have to search large leaves. | |
5069 | * So now the maximum size is OCFS2_MAX_XATTR_TREE_LEAF_SIZE(or clustersize, | |
5070 | * if it's bigger). | |
5071 | * | |
5072 | * first_bh is the first block of the previous extent rec and header_bh | |
5073 | * indicates the bucket we will insert the new xattrs. They will be updated | |
5074 | * when the header_bh is moved into the new cluster. | |
5075 | */ | |
5076 | static int ocfs2_add_new_xattr_cluster(struct inode *inode, | |
5077 | struct buffer_head *root_bh, | |
ed29c0ca JB |
5078 | struct ocfs2_xattr_bucket *first, |
5079 | struct ocfs2_xattr_bucket *target, | |
01225596 TM |
5080 | u32 *num_clusters, |
5081 | u32 prev_cpos, | |
78f30c31 TM |
5082 | int *extend, |
5083 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5084 | { |
85db90e7 | 5085 | int ret; |
01225596 TM |
5086 | u16 bpc = ocfs2_clusters_to_blocks(inode->i_sb, 1); |
5087 | u32 prev_clusters = *num_clusters; | |
5088 | u32 clusters_to_add = 1, bit_off, num_bits, v_start = 0; | |
5089 | u64 block; | |
85db90e7 | 5090 | handle_t *handle = ctxt->handle; |
01225596 | 5091 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
f99b9b7c | 5092 | struct ocfs2_extent_tree et; |
01225596 | 5093 | |
402b4183 TM |
5094 | trace_ocfs2_add_new_xattr_cluster_begin( |
5095 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
5096 | (unsigned long long)bucket_blkno(first), | |
5097 | prev_cpos, prev_clusters); | |
01225596 | 5098 | |
5e404e9e | 5099 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
f99b9b7c | 5100 | |
0cf2f763 | 5101 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5102 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5103 | if (ret < 0) { |
5104 | mlog_errno(ret); | |
5105 | goto leave; | |
5106 | } | |
5107 | ||
1ed9b777 | 5108 | ret = __ocfs2_claim_clusters(handle, ctxt->data_ac, 1, |
01225596 TM |
5109 | clusters_to_add, &bit_off, &num_bits); |
5110 | if (ret < 0) { | |
5111 | if (ret != -ENOSPC) | |
5112 | mlog_errno(ret); | |
5113 | goto leave; | |
5114 | } | |
5115 | ||
5116 | BUG_ON(num_bits > clusters_to_add); | |
5117 | ||
5118 | block = ocfs2_clusters_to_blocks(osb->sb, bit_off); | |
402b4183 | 5119 | trace_ocfs2_add_new_xattr_cluster((unsigned long long)block, num_bits); |
01225596 | 5120 | |
ed29c0ca | 5121 | if (bucket_blkno(first) + (prev_clusters * bpc) == block && |
01225596 TM |
5122 | (prev_clusters + num_bits) << osb->s_clustersize_bits <= |
5123 | OCFS2_MAX_XATTR_TREE_LEAF_SIZE) { | |
5124 | /* | |
5125 | * If this cluster is contiguous with the old one and | |
5126 | * adding this new cluster, we don't surpass the limit of | |
5127 | * OCFS2_MAX_XATTR_TREE_LEAF_SIZE, cool. We will let it be | |
5128 | * initialized and used like other buckets in the previous | |
5129 | * cluster. | |
5130 | * So add it as a contiguous one. The caller will handle | |
5131 | * its init process. | |
5132 | */ | |
5133 | v_start = prev_cpos + prev_clusters; | |
5134 | *num_clusters = prev_clusters + num_bits; | |
01225596 TM |
5135 | } else { |
5136 | ret = ocfs2_adjust_xattr_cross_cluster(inode, | |
5137 | handle, | |
012ee910 JB |
5138 | first, |
5139 | target, | |
01225596 | 5140 | block, |
01225596 TM |
5141 | prev_clusters, |
5142 | &v_start, | |
5143 | extend); | |
5144 | if (ret) { | |
5145 | mlog_errno(ret); | |
5146 | goto leave; | |
5147 | } | |
5148 | } | |
5149 | ||
402b4183 TM |
5150 | trace_ocfs2_add_new_xattr_cluster_insert((unsigned long long)block, |
5151 | v_start, num_bits); | |
cc79d8c1 | 5152 | ret = ocfs2_insert_extent(handle, &et, v_start, block, |
78f30c31 | 5153 | num_bits, 0, ctxt->meta_ac); |
01225596 TM |
5154 | if (ret < 0) { |
5155 | mlog_errno(ret); | |
5156 | goto leave; | |
5157 | } | |
5158 | ||
ec20cec7 | 5159 | ocfs2_journal_dirty(handle, root_bh); |
01225596 TM |
5160 | |
5161 | leave: | |
01225596 TM |
5162 | return ret; |
5163 | } | |
5164 | ||
5165 | /* | |
92de109a JB |
5166 | * We are given an extent. 'first' is the bucket at the very front of |
5167 | * the extent. The extent has space for an additional bucket past | |
5168 | * bucket_xh(first)->xh_num_buckets. 'target_blkno' is the block number | |
5169 | * of the target bucket. We wish to shift every bucket past the target | |
5170 | * down one, filling in that additional space. When we get back to the | |
5171 | * target, we split the target between itself and the now-empty bucket | |
5172 | * at target+1 (aka, target_blkno + blks_per_bucket). | |
01225596 TM |
5173 | */ |
5174 | static int ocfs2_extend_xattr_bucket(struct inode *inode, | |
85db90e7 | 5175 | handle_t *handle, |
92de109a JB |
5176 | struct ocfs2_xattr_bucket *first, |
5177 | u64 target_blk, | |
01225596 TM |
5178 | u32 num_clusters) |
5179 | { | |
5180 | int ret, credits; | |
5181 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5182 | u16 blk_per_bucket = ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
92de109a JB |
5183 | u64 end_blk; |
5184 | u16 new_bucket = le16_to_cpu(bucket_xh(first)->xh_num_buckets); | |
01225596 | 5185 | |
402b4183 TM |
5186 | trace_ocfs2_extend_xattr_bucket((unsigned long long)target_blk, |
5187 | (unsigned long long)bucket_blkno(first), | |
5188 | num_clusters, new_bucket); | |
01225596 | 5189 | |
92de109a JB |
5190 | /* The extent must have room for an additional bucket */ |
5191 | BUG_ON(new_bucket >= | |
5192 | (num_clusters * ocfs2_xattr_buckets_per_cluster(osb))); | |
01225596 | 5193 | |
92de109a JB |
5194 | /* end_blk points to the last existing bucket */ |
5195 | end_blk = bucket_blkno(first) + ((new_bucket - 1) * blk_per_bucket); | |
01225596 TM |
5196 | |
5197 | /* | |
92de109a JB |
5198 | * end_blk is the start of the last existing bucket. |
5199 | * Thus, (end_blk - target_blk) covers the target bucket and | |
5200 | * every bucket after it up to, but not including, the last | |
5201 | * existing bucket. Then we add the last existing bucket, the | |
5202 | * new bucket, and the first bucket (3 * blk_per_bucket). | |
01225596 | 5203 | */ |
c901fb00 | 5204 | credits = (end_blk - target_blk) + (3 * blk_per_bucket); |
85db90e7 TM |
5205 | ret = ocfs2_extend_trans(handle, credits); |
5206 | if (ret) { | |
01225596 TM |
5207 | mlog_errno(ret); |
5208 | goto out; | |
5209 | } | |
5210 | ||
92de109a JB |
5211 | ret = ocfs2_xattr_bucket_journal_access(handle, first, |
5212 | OCFS2_JOURNAL_ACCESS_WRITE); | |
01225596 TM |
5213 | if (ret) { |
5214 | mlog_errno(ret); | |
85db90e7 | 5215 | goto out; |
01225596 TM |
5216 | } |
5217 | ||
92de109a | 5218 | while (end_blk != target_blk) { |
01225596 TM |
5219 | ret = ocfs2_cp_xattr_bucket(inode, handle, end_blk, |
5220 | end_blk + blk_per_bucket, 0); | |
5221 | if (ret) | |
85db90e7 | 5222 | goto out; |
01225596 TM |
5223 | end_blk -= blk_per_bucket; |
5224 | } | |
5225 | ||
92de109a JB |
5226 | /* Move half of the xattr in target_blkno to the next bucket. */ |
5227 | ret = ocfs2_divide_xattr_bucket(inode, handle, target_blk, | |
5228 | target_blk + blk_per_bucket, NULL, 0); | |
01225596 | 5229 | |
92de109a JB |
5230 | le16_add_cpu(&bucket_xh(first)->xh_num_buckets, 1); |
5231 | ocfs2_xattr_bucket_journal_dirty(handle, first); | |
01225596 | 5232 | |
01225596 TM |
5233 | out: |
5234 | return ret; | |
5235 | } | |
5236 | ||
5237 | /* | |
91f2033f JB |
5238 | * Add new xattr bucket in an extent record and adjust the buckets |
5239 | * accordingly. xb_bh is the ocfs2_xattr_block, and target is the | |
5240 | * bucket we want to insert into. | |
01225596 | 5241 | * |
91f2033f JB |
5242 | * In the easy case, we will move all the buckets after target down by |
5243 | * one. Half of target's xattrs will be moved to the next bucket. | |
5244 | * | |
5245 | * If current cluster is full, we'll allocate a new one. This may not | |
5246 | * be contiguous. The underlying calls will make sure that there is | |
5247 | * space for the insert, shifting buckets around if necessary. | |
5248 | * 'target' may be moved by those calls. | |
01225596 TM |
5249 | */ |
5250 | static int ocfs2_add_new_xattr_bucket(struct inode *inode, | |
5251 | struct buffer_head *xb_bh, | |
91f2033f | 5252 | struct ocfs2_xattr_bucket *target, |
78f30c31 | 5253 | struct ocfs2_xattr_set_ctxt *ctxt) |
01225596 | 5254 | { |
01225596 TM |
5255 | struct ocfs2_xattr_block *xb = |
5256 | (struct ocfs2_xattr_block *)xb_bh->b_data; | |
5257 | struct ocfs2_xattr_tree_root *xb_root = &xb->xb_attrs.xb_root; | |
5258 | struct ocfs2_extent_list *el = &xb_root->xt_list; | |
91f2033f JB |
5259 | u32 name_hash = |
5260 | le32_to_cpu(bucket_xh(target)->xh_entries[0].xe_name_hash); | |
ed29c0ca | 5261 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
01225596 TM |
5262 | int ret, num_buckets, extend = 1; |
5263 | u64 p_blkno; | |
5264 | u32 e_cpos, num_clusters; | |
92de109a | 5265 | /* The bucket at the front of the extent */ |
91f2033f | 5266 | struct ocfs2_xattr_bucket *first; |
01225596 | 5267 | |
402b4183 TM |
5268 | trace_ocfs2_add_new_xattr_bucket( |
5269 | (unsigned long long)bucket_blkno(target)); | |
01225596 | 5270 | |
ed29c0ca | 5271 | /* The first bucket of the original extent */ |
92de109a | 5272 | first = ocfs2_xattr_bucket_new(inode); |
91f2033f | 5273 | if (!first) { |
92de109a JB |
5274 | ret = -ENOMEM; |
5275 | mlog_errno(ret); | |
5276 | goto out; | |
5277 | } | |
5278 | ||
01225596 TM |
5279 | ret = ocfs2_xattr_get_rec(inode, name_hash, &p_blkno, &e_cpos, |
5280 | &num_clusters, el); | |
5281 | if (ret) { | |
5282 | mlog_errno(ret); | |
5283 | goto out; | |
5284 | } | |
5285 | ||
ed29c0ca | 5286 | ret = ocfs2_read_xattr_bucket(first, p_blkno); |
01225596 TM |
5287 | if (ret) { |
5288 | mlog_errno(ret); | |
5289 | goto out; | |
5290 | } | |
5291 | ||
ed29c0ca JB |
5292 | num_buckets = ocfs2_xattr_buckets_per_cluster(osb) * num_clusters; |
5293 | if (num_buckets == le16_to_cpu(bucket_xh(first)->xh_num_buckets)) { | |
5294 | /* | |
5295 | * This can move first+target if the target bucket moves | |
5296 | * to the new extent. | |
5297 | */ | |
01225596 TM |
5298 | ret = ocfs2_add_new_xattr_cluster(inode, |
5299 | xb_bh, | |
ed29c0ca JB |
5300 | first, |
5301 | target, | |
01225596 TM |
5302 | &num_clusters, |
5303 | e_cpos, | |
78f30c31 TM |
5304 | &extend, |
5305 | ctxt); | |
01225596 TM |
5306 | if (ret) { |
5307 | mlog_errno(ret); | |
5308 | goto out; | |
5309 | } | |
5310 | } | |
5311 | ||
92de109a | 5312 | if (extend) { |
01225596 | 5313 | ret = ocfs2_extend_xattr_bucket(inode, |
85db90e7 | 5314 | ctxt->handle, |
ed29c0ca JB |
5315 | first, |
5316 | bucket_blkno(target), | |
01225596 | 5317 | num_clusters); |
92de109a JB |
5318 | if (ret) |
5319 | mlog_errno(ret); | |
5320 | } | |
5321 | ||
01225596 | 5322 | out: |
92de109a | 5323 | ocfs2_xattr_bucket_free(first); |
ed29c0ca | 5324 | |
01225596 TM |
5325 | return ret; |
5326 | } | |
5327 | ||
5328 | static inline char *ocfs2_xattr_bucket_get_val(struct inode *inode, | |
5329 | struct ocfs2_xattr_bucket *bucket, | |
5330 | int offs) | |
5331 | { | |
5332 | int block_off = offs >> inode->i_sb->s_blocksize_bits; | |
5333 | ||
5334 | offs = offs % inode->i_sb->s_blocksize; | |
51def39f | 5335 | return bucket_block(bucket, block_off) + offs; |
01225596 TM |
5336 | } |
5337 | ||
01225596 TM |
5338 | /* |
5339 | * Truncate the specified xe_off entry in xattr bucket. | |
5340 | * bucket is indicated by header_bh and len is the new length. | |
5341 | * Both the ocfs2_xattr_value_root and the entry will be updated here. | |
5342 | * | |
5343 | * Copy the new updated xe and xe_value_root to new_xe and new_xv if needed. | |
5344 | */ | |
5345 | static int ocfs2_xattr_bucket_value_truncate(struct inode *inode, | |
548b0f22 | 5346 | struct ocfs2_xattr_bucket *bucket, |
01225596 | 5347 | int xe_off, |
78f30c31 TM |
5348 | int len, |
5349 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 TM |
5350 | { |
5351 | int ret, offset; | |
5352 | u64 value_blk; | |
01225596 | 5353 | struct ocfs2_xattr_entry *xe; |
548b0f22 | 5354 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
01225596 | 5355 | size_t blocksize = inode->i_sb->s_blocksize; |
b3e5d379 JB |
5356 | struct ocfs2_xattr_value_buf vb = { |
5357 | .vb_access = ocfs2_journal_access, | |
5358 | }; | |
01225596 TM |
5359 | |
5360 | xe = &xh->xh_entries[xe_off]; | |
5361 | ||
5362 | BUG_ON(!xe || ocfs2_xattr_is_local(xe)); | |
5363 | ||
5364 | offset = le16_to_cpu(xe->xe_name_offset) + | |
5365 | OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5366 | ||
5367 | value_blk = offset / blocksize; | |
5368 | ||
5369 | /* We don't allow ocfs2_xattr_value to be stored in different block. */ | |
5370 | BUG_ON(value_blk != (offset + OCFS2_XATTR_ROOT_SIZE - 1) / blocksize); | |
01225596 | 5371 | |
b3e5d379 JB |
5372 | vb.vb_bh = bucket->bu_bhs[value_blk]; |
5373 | BUG_ON(!vb.vb_bh); | |
01225596 | 5374 | |
b3e5d379 JB |
5375 | vb.vb_xv = (struct ocfs2_xattr_value_root *) |
5376 | (vb.vb_bh->b_data + offset % blocksize); | |
01225596 | 5377 | |
548b0f22 JB |
5378 | /* |
5379 | * From here on out we have to dirty the bucket. The generic | |
5380 | * value calls only modify one of the bucket's bhs, but we need | |
5381 | * to send the bucket at once. So if they error, they *could* have | |
5382 | * modified something. We have to assume they did, and dirty | |
5383 | * the whole bucket. This leaves us in a consistent state. | |
5384 | */ | |
402b4183 TM |
5385 | trace_ocfs2_xattr_bucket_value_truncate( |
5386 | (unsigned long long)bucket_blkno(bucket), xe_off, len); | |
b3e5d379 | 5387 | ret = ocfs2_xattr_value_truncate(inode, &vb, len, ctxt); |
01225596 TM |
5388 | if (ret) { |
5389 | mlog_errno(ret); | |
554e7f9e TM |
5390 | goto out; |
5391 | } | |
5392 | ||
5393 | ret = ocfs2_xattr_bucket_journal_access(ctxt->handle, bucket, | |
5394 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5395 | if (ret) { | |
5396 | mlog_errno(ret); | |
5397 | goto out; | |
01225596 TM |
5398 | } |
5399 | ||
548b0f22 JB |
5400 | xe->xe_value_size = cpu_to_le64(len); |
5401 | ||
548b0f22 JB |
5402 | ocfs2_xattr_bucket_journal_dirty(ctxt->handle, bucket); |
5403 | ||
01225596 | 5404 | out: |
01225596 TM |
5405 | return ret; |
5406 | } | |
5407 | ||
01225596 TM |
5408 | static int ocfs2_rm_xattr_cluster(struct inode *inode, |
5409 | struct buffer_head *root_bh, | |
5410 | u64 blkno, | |
5411 | u32 cpos, | |
47bca495 TM |
5412 | u32 len, |
5413 | void *para) | |
01225596 TM |
5414 | { |
5415 | int ret; | |
5416 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5417 | struct inode *tl_inode = osb->osb_tl_inode; | |
5418 | handle_t *handle; | |
5419 | struct ocfs2_xattr_block *xb = | |
5420 | (struct ocfs2_xattr_block *)root_bh->b_data; | |
01225596 TM |
5421 | struct ocfs2_alloc_context *meta_ac = NULL; |
5422 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
f99b9b7c JB |
5423 | struct ocfs2_extent_tree et; |
5424 | ||
47bca495 | 5425 | ret = ocfs2_iterate_xattr_buckets(inode, blkno, len, |
ce9c5a54 | 5426 | ocfs2_delete_xattr_in_bucket, para); |
47bca495 TM |
5427 | if (ret) { |
5428 | mlog_errno(ret); | |
5429 | return ret; | |
5430 | } | |
5431 | ||
5e404e9e | 5432 | ocfs2_init_xattr_tree_extent_tree(&et, INODE_CACHE(inode), root_bh); |
01225596 TM |
5433 | |
5434 | ocfs2_init_dealloc_ctxt(&dealloc); | |
5435 | ||
402b4183 TM |
5436 | trace_ocfs2_rm_xattr_cluster( |
5437 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
5438 | (unsigned long long)blkno, cpos, len); | |
01225596 | 5439 | |
8cb471e8 JB |
5440 | ocfs2_remove_xattr_clusters_from_cache(INODE_CACHE(inode), blkno, |
5441 | len); | |
01225596 | 5442 | |
f99b9b7c | 5443 | ret = ocfs2_lock_allocators(inode, &et, 0, 1, NULL, &meta_ac); |
01225596 TM |
5444 | if (ret) { |
5445 | mlog_errno(ret); | |
5446 | return ret; | |
5447 | } | |
5448 | ||
5449 | mutex_lock(&tl_inode->i_mutex); | |
5450 | ||
5451 | if (ocfs2_truncate_log_needs_flush(osb)) { | |
5452 | ret = __ocfs2_flush_truncate_log(osb); | |
5453 | if (ret < 0) { | |
5454 | mlog_errno(ret); | |
5455 | goto out; | |
5456 | } | |
5457 | } | |
5458 | ||
a90714c1 | 5459 | handle = ocfs2_start_trans(osb, ocfs2_remove_extent_credits(osb->sb)); |
d3264799 | 5460 | if (IS_ERR(handle)) { |
01225596 TM |
5461 | ret = -ENOMEM; |
5462 | mlog_errno(ret); | |
5463 | goto out; | |
5464 | } | |
5465 | ||
0cf2f763 | 5466 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(inode), root_bh, |
84008972 | 5467 | OCFS2_JOURNAL_ACCESS_WRITE); |
01225596 TM |
5468 | if (ret) { |
5469 | mlog_errno(ret); | |
5470 | goto out_commit; | |
5471 | } | |
5472 | ||
dbdcf6a4 | 5473 | ret = ocfs2_remove_extent(handle, &et, cpos, len, meta_ac, |
f99b9b7c | 5474 | &dealloc); |
01225596 TM |
5475 | if (ret) { |
5476 | mlog_errno(ret); | |
5477 | goto out_commit; | |
5478 | } | |
5479 | ||
5480 | le32_add_cpu(&xb->xb_attrs.xb_root.xt_clusters, -len); | |
ec20cec7 | 5481 | ocfs2_journal_dirty(handle, root_bh); |
01225596 TM |
5482 | |
5483 | ret = ocfs2_truncate_log_append(osb, handle, blkno, len); | |
5484 | if (ret) | |
5485 | mlog_errno(ret); | |
5486 | ||
5487 | out_commit: | |
5488 | ocfs2_commit_trans(osb, handle); | |
5489 | out: | |
5490 | ocfs2_schedule_truncate_log_flush(osb, 1); | |
5491 | ||
5492 | mutex_unlock(&tl_inode->i_mutex); | |
5493 | ||
5494 | if (meta_ac) | |
5495 | ocfs2_free_alloc_context(meta_ac); | |
5496 | ||
5497 | ocfs2_run_deallocs(osb, &dealloc); | |
5498 | ||
5499 | return ret; | |
5500 | } | |
5501 | ||
80bcaf34 TM |
5502 | /* |
5503 | * check whether the xattr bucket is filled up with the same hash value. | |
5504 | * If we want to insert the xattr with the same hash, return -ENOSPC. | |
5505 | * If we want to insert a xattr with different hash value, go ahead | |
5506 | * and ocfs2_divide_xattr_bucket will handle this. | |
5507 | */ | |
01225596 | 5508 | static int ocfs2_check_xattr_bucket_collision(struct inode *inode, |
80bcaf34 TM |
5509 | struct ocfs2_xattr_bucket *bucket, |
5510 | const char *name) | |
01225596 | 5511 | { |
3e632946 | 5512 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
80bcaf34 TM |
5513 | u32 name_hash = ocfs2_xattr_name_hash(inode, name, strlen(name)); |
5514 | ||
5515 | if (name_hash != le32_to_cpu(xh->xh_entries[0].xe_name_hash)) | |
5516 | return 0; | |
01225596 TM |
5517 | |
5518 | if (xh->xh_entries[le16_to_cpu(xh->xh_count) - 1].xe_name_hash == | |
5519 | xh->xh_entries[0].xe_name_hash) { | |
5520 | mlog(ML_ERROR, "Too much hash collision in xattr bucket %llu, " | |
5521 | "hash = %u\n", | |
9c7759aa | 5522 | (unsigned long long)bucket_blkno(bucket), |
01225596 TM |
5523 | le32_to_cpu(xh->xh_entries[0].xe_name_hash)); |
5524 | return -ENOSPC; | |
5525 | } | |
5526 | ||
5527 | return 0; | |
5528 | } | |
5529 | ||
c5d95df5 JB |
5530 | /* |
5531 | * Try to set the entry in the current bucket. If we fail, the caller | |
5532 | * will handle getting us another bucket. | |
5533 | */ | |
5534 | static int ocfs2_xattr_set_entry_bucket(struct inode *inode, | |
5535 | struct ocfs2_xattr_info *xi, | |
5536 | struct ocfs2_xattr_search *xs, | |
5537 | struct ocfs2_xattr_set_ctxt *ctxt) | |
01225596 | 5538 | { |
c5d95df5 JB |
5539 | int ret; |
5540 | struct ocfs2_xa_loc loc; | |
01225596 | 5541 | |
402b4183 | 5542 | trace_ocfs2_xattr_set_entry_bucket(xi->xi_name); |
01225596 | 5543 | |
c5d95df5 JB |
5544 | ocfs2_init_xattr_bucket_xa_loc(&loc, xs->bucket, |
5545 | xs->not_found ? NULL : xs->here); | |
5546 | ret = ocfs2_xa_set(&loc, xi, ctxt); | |
5547 | if (!ret) { | |
5548 | xs->here = loc.xl_entry; | |
5549 | goto out; | |
5550 | } | |
5551 | if (ret != -ENOSPC) { | |
5552 | mlog_errno(ret); | |
5553 | goto out; | |
5554 | } | |
01225596 | 5555 | |
c5d95df5 JB |
5556 | /* Ok, we need space. Let's try defragmenting the bucket. */ |
5557 | ret = ocfs2_defrag_xattr_bucket(inode, ctxt->handle, | |
5558 | xs->bucket); | |
5559 | if (ret) { | |
5560 | mlog_errno(ret); | |
5561 | goto out; | |
5562 | } | |
01225596 | 5563 | |
c5d95df5 JB |
5564 | ret = ocfs2_xa_set(&loc, xi, ctxt); |
5565 | if (!ret) { | |
5566 | xs->here = loc.xl_entry; | |
5567 | goto out; | |
01225596 | 5568 | } |
c5d95df5 JB |
5569 | if (ret != -ENOSPC) |
5570 | mlog_errno(ret); | |
01225596 | 5571 | |
01225596 | 5572 | |
c5d95df5 | 5573 | out: |
c5d95df5 JB |
5574 | return ret; |
5575 | } | |
01225596 | 5576 | |
c5d95df5 JB |
5577 | static int ocfs2_xattr_set_entry_index_block(struct inode *inode, |
5578 | struct ocfs2_xattr_info *xi, | |
5579 | struct ocfs2_xattr_search *xs, | |
5580 | struct ocfs2_xattr_set_ctxt *ctxt) | |
5581 | { | |
5582 | int ret; | |
01225596 | 5583 | |
402b4183 | 5584 | trace_ocfs2_xattr_set_entry_index_block(xi->xi_name); |
01225596 | 5585 | |
c5d95df5 JB |
5586 | ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); |
5587 | if (!ret) | |
5588 | goto out; | |
5589 | if (ret != -ENOSPC) { | |
5590 | mlog_errno(ret); | |
5591 | goto out; | |
5592 | } | |
01225596 | 5593 | |
c5d95df5 | 5594 | /* Ack, need more space. Let's try to get another bucket! */ |
01225596 | 5595 | |
c5d95df5 JB |
5596 | /* |
5597 | * We do not allow for overlapping ranges between buckets. And | |
5598 | * the maximum number of collisions we will allow for then is | |
5599 | * one bucket's worth, so check it here whether we need to | |
5600 | * add a new bucket for the insert. | |
5601 | */ | |
5602 | ret = ocfs2_check_xattr_bucket_collision(inode, | |
91f2033f | 5603 | xs->bucket, |
c5d95df5 JB |
5604 | xi->xi_name); |
5605 | if (ret) { | |
5606 | mlog_errno(ret); | |
5607 | goto out; | |
5608 | } | |
01225596 | 5609 | |
c5d95df5 JB |
5610 | ret = ocfs2_add_new_xattr_bucket(inode, |
5611 | xs->xattr_bh, | |
5612 | xs->bucket, | |
5613 | ctxt); | |
5614 | if (ret) { | |
5615 | mlog_errno(ret); | |
5616 | goto out; | |
01225596 TM |
5617 | } |
5618 | ||
c5d95df5 JB |
5619 | /* |
5620 | * ocfs2_add_new_xattr_bucket() will have updated | |
5621 | * xs->bucket if it moved, but it will not have updated | |
5622 | * any of the other search fields. Thus, we drop it and | |
5623 | * re-search. Everything should be cached, so it'll be | |
5624 | * quick. | |
5625 | */ | |
5626 | ocfs2_xattr_bucket_relse(xs->bucket); | |
5627 | ret = ocfs2_xattr_index_block_find(inode, xs->xattr_bh, | |
5628 | xi->xi_name_index, | |
5629 | xi->xi_name, xs); | |
5630 | if (ret && ret != -ENODATA) | |
5631 | goto out; | |
5632 | xs->not_found = ret; | |
5633 | ||
5634 | /* Ok, we have a new bucket, let's try again */ | |
5635 | ret = ocfs2_xattr_set_entry_bucket(inode, xi, xs, ctxt); | |
5636 | if (ret && (ret != -ENOSPC)) | |
5637 | mlog_errno(ret); | |
5638 | ||
01225596 | 5639 | out: |
01225596 TM |
5640 | return ret; |
5641 | } | |
a3944256 TM |
5642 | |
5643 | static int ocfs2_delete_xattr_in_bucket(struct inode *inode, | |
5644 | struct ocfs2_xattr_bucket *bucket, | |
5645 | void *para) | |
5646 | { | |
ce9c5a54 | 5647 | int ret = 0, ref_credits; |
3e632946 | 5648 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); |
a3944256 TM |
5649 | u16 i; |
5650 | struct ocfs2_xattr_entry *xe; | |
78f30c31 TM |
5651 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
5652 | struct ocfs2_xattr_set_ctxt ctxt = {NULL, NULL,}; | |
548b0f22 JB |
5653 | int credits = ocfs2_remove_extent_credits(osb->sb) + |
5654 | ocfs2_blocks_per_xattr_bucket(inode->i_sb); | |
ce9c5a54 TM |
5655 | struct ocfs2_xattr_value_root *xv; |
5656 | struct ocfs2_rm_xattr_bucket_para *args = | |
5657 | (struct ocfs2_rm_xattr_bucket_para *)para; | |
78f30c31 TM |
5658 | |
5659 | ocfs2_init_dealloc_ctxt(&ctxt.dealloc); | |
a3944256 TM |
5660 | |
5661 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
5662 | xe = &xh->xh_entries[i]; | |
5663 | if (ocfs2_xattr_is_local(xe)) | |
5664 | continue; | |
5665 | ||
ce9c5a54 TM |
5666 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, |
5667 | i, &xv, NULL); | |
5668 | ||
5669 | ret = ocfs2_lock_xattr_remove_allocators(inode, xv, | |
5670 | args->ref_ci, | |
5671 | args->ref_root_bh, | |
5672 | &ctxt.meta_ac, | |
5673 | &ref_credits); | |
5674 | ||
5675 | ctxt.handle = ocfs2_start_trans(osb, credits + ref_credits); | |
88c3b062 TM |
5676 | if (IS_ERR(ctxt.handle)) { |
5677 | ret = PTR_ERR(ctxt.handle); | |
5678 | mlog_errno(ret); | |
5679 | break; | |
5680 | } | |
5681 | ||
548b0f22 | 5682 | ret = ocfs2_xattr_bucket_value_truncate(inode, bucket, |
78f30c31 | 5683 | i, 0, &ctxt); |
88c3b062 TM |
5684 | |
5685 | ocfs2_commit_trans(osb, ctxt.handle); | |
ce9c5a54 TM |
5686 | if (ctxt.meta_ac) { |
5687 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
5688 | ctxt.meta_ac = NULL; | |
5689 | } | |
a3944256 TM |
5690 | if (ret) { |
5691 | mlog_errno(ret); | |
5692 | break; | |
5693 | } | |
5694 | } | |
5695 | ||
ce9c5a54 TM |
5696 | if (ctxt.meta_ac) |
5697 | ocfs2_free_alloc_context(ctxt.meta_ac); | |
78f30c31 TM |
5698 | ocfs2_schedule_truncate_log_flush(osb, 1); |
5699 | ocfs2_run_deallocs(osb, &ctxt.dealloc); | |
a3944256 TM |
5700 | return ret; |
5701 | } | |
5702 | ||
492a8a33 TM |
5703 | /* |
5704 | * Whenever we modify a xattr value root in the bucket(e.g, CoW | |
5705 | * or change the extent record flag), we need to recalculate | |
5706 | * the metaecc for the whole bucket. So it is done here. | |
5707 | * | |
5708 | * Note: | |
5709 | * We have to give the extra credits for the caller. | |
5710 | */ | |
5711 | static int ocfs2_xattr_bucket_post_refcount(struct inode *inode, | |
5712 | handle_t *handle, | |
5713 | void *para) | |
5714 | { | |
5715 | int ret; | |
5716 | struct ocfs2_xattr_bucket *bucket = | |
5717 | (struct ocfs2_xattr_bucket *)para; | |
5718 | ||
5719 | ret = ocfs2_xattr_bucket_journal_access(handle, bucket, | |
5720 | OCFS2_JOURNAL_ACCESS_WRITE); | |
5721 | if (ret) { | |
5722 | mlog_errno(ret); | |
5723 | return ret; | |
5724 | } | |
5725 | ||
5726 | ocfs2_xattr_bucket_journal_dirty(handle, bucket); | |
5727 | ||
5728 | return 0; | |
5729 | } | |
5730 | ||
5731 | /* | |
5732 | * Special action we need if the xattr value is refcounted. | |
5733 | * | |
5734 | * 1. If the xattr is refcounted, lock the tree. | |
5735 | * 2. CoW the xattr if we are setting the new value and the value | |
5736 | * will be stored outside. | |
5737 | * 3. In other case, decrease_refcount will work for us, so just | |
5738 | * lock the refcount tree, calculate the meta and credits is OK. | |
5739 | * | |
5740 | * We have to do CoW before ocfs2_init_xattr_set_ctxt since | |
5741 | * currently CoW is a completed transaction, while this function | |
5742 | * will also lock the allocators and let us deadlock. So we will | |
5743 | * CoW the whole xattr value. | |
5744 | */ | |
5745 | static int ocfs2_prepare_refcount_xattr(struct inode *inode, | |
5746 | struct ocfs2_dinode *di, | |
5747 | struct ocfs2_xattr_info *xi, | |
5748 | struct ocfs2_xattr_search *xis, | |
5749 | struct ocfs2_xattr_search *xbs, | |
5750 | struct ocfs2_refcount_tree **ref_tree, | |
5751 | int *meta_add, | |
5752 | int *credits) | |
5753 | { | |
5754 | int ret = 0; | |
5755 | struct ocfs2_xattr_block *xb; | |
5756 | struct ocfs2_xattr_entry *xe; | |
5757 | char *base; | |
5758 | u32 p_cluster, num_clusters; | |
5759 | unsigned int ext_flags; | |
5760 | int name_offset, name_len; | |
5761 | struct ocfs2_xattr_value_buf vb; | |
5762 | struct ocfs2_xattr_bucket *bucket = NULL; | |
5763 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
5764 | struct ocfs2_post_refcount refcount; | |
5765 | struct ocfs2_post_refcount *p = NULL; | |
5766 | struct buffer_head *ref_root_bh = NULL; | |
5767 | ||
5768 | if (!xis->not_found) { | |
5769 | xe = xis->here; | |
5770 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5771 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5772 | base = xis->base; | |
5773 | vb.vb_bh = xis->inode_bh; | |
5774 | vb.vb_access = ocfs2_journal_access_di; | |
5775 | } else { | |
5776 | int i, block_off = 0; | |
5777 | xb = (struct ocfs2_xattr_block *)xbs->xattr_bh->b_data; | |
5778 | xe = xbs->here; | |
5779 | name_offset = le16_to_cpu(xe->xe_name_offset); | |
5780 | name_len = OCFS2_XATTR_SIZE(xe->xe_name_len); | |
5781 | i = xbs->here - xbs->header->xh_entries; | |
5782 | ||
5783 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) { | |
fd68a894 | 5784 | ret = ocfs2_xattr_bucket_get_name_value(inode->i_sb, |
492a8a33 TM |
5785 | bucket_xh(xbs->bucket), |
5786 | i, &block_off, | |
5787 | &name_offset); | |
5788 | if (ret) { | |
5789 | mlog_errno(ret); | |
5790 | goto out; | |
5791 | } | |
5792 | base = bucket_block(xbs->bucket, block_off); | |
5793 | vb.vb_bh = xbs->bucket->bu_bhs[block_off]; | |
5794 | vb.vb_access = ocfs2_journal_access; | |
5795 | ||
5796 | if (ocfs2_meta_ecc(osb)) { | |
5797 | /*create parameters for ocfs2_post_refcount. */ | |
5798 | bucket = xbs->bucket; | |
5799 | refcount.credits = bucket->bu_blocks; | |
5800 | refcount.para = bucket; | |
5801 | refcount.func = | |
5802 | ocfs2_xattr_bucket_post_refcount; | |
5803 | p = &refcount; | |
5804 | } | |
5805 | } else { | |
5806 | base = xbs->base; | |
5807 | vb.vb_bh = xbs->xattr_bh; | |
5808 | vb.vb_access = ocfs2_journal_access_xb; | |
5809 | } | |
5810 | } | |
5811 | ||
5812 | if (ocfs2_xattr_is_local(xe)) | |
5813 | goto out; | |
5814 | ||
5815 | vb.vb_xv = (struct ocfs2_xattr_value_root *) | |
5816 | (base + name_offset + name_len); | |
5817 | ||
5818 | ret = ocfs2_xattr_get_clusters(inode, 0, &p_cluster, | |
5819 | &num_clusters, &vb.vb_xv->xr_list, | |
5820 | &ext_flags); | |
5821 | if (ret) { | |
5822 | mlog_errno(ret); | |
5823 | goto out; | |
5824 | } | |
5825 | ||
5826 | /* | |
5827 | * We just need to check the 1st extent record, since we always | |
5828 | * CoW the whole xattr. So there shouldn't be a xattr with | |
5829 | * some REFCOUNT extent recs after the 1st one. | |
5830 | */ | |
5831 | if (!(ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5832 | goto out; | |
5833 | ||
5834 | ret = ocfs2_lock_refcount_tree(osb, le64_to_cpu(di->i_refcount_loc), | |
5835 | 1, ref_tree, &ref_root_bh); | |
5836 | if (ret) { | |
5837 | mlog_errno(ret); | |
5838 | goto out; | |
5839 | } | |
5840 | ||
5841 | /* | |
5842 | * If we are deleting the xattr or the new size will be stored inside, | |
5843 | * cool, leave it there, the xattr truncate process will remove them | |
5844 | * for us(it still needs the refcount tree lock and the meta, credits). | |
5845 | * And the worse case is that every cluster truncate will split the | |
5846 | * refcount tree, and make the original extent become 3. So we will need | |
5847 | * 2 * cluster more extent recs at most. | |
5848 | */ | |
6b240ff6 | 5849 | if (!xi->xi_value || xi->xi_value_len <= OCFS2_XATTR_INLINE_SIZE) { |
492a8a33 TM |
5850 | |
5851 | ret = ocfs2_refcounted_xattr_delete_need(inode, | |
5852 | &(*ref_tree)->rf_ci, | |
5853 | ref_root_bh, vb.vb_xv, | |
5854 | meta_add, credits); | |
5855 | if (ret) | |
5856 | mlog_errno(ret); | |
5857 | goto out; | |
5858 | } | |
5859 | ||
5860 | ret = ocfs2_refcount_cow_xattr(inode, di, &vb, | |
5861 | *ref_tree, ref_root_bh, 0, | |
5862 | le32_to_cpu(vb.vb_xv->xr_clusters), p); | |
5863 | if (ret) | |
5864 | mlog_errno(ret); | |
5865 | ||
5866 | out: | |
5867 | brelse(ref_root_bh); | |
5868 | return ret; | |
5869 | } | |
5870 | ||
0129241e TM |
5871 | /* |
5872 | * Add the REFCOUNTED flags for all the extent rec in ocfs2_xattr_value_root. | |
5873 | * The physical clusters will be added to refcount tree. | |
5874 | */ | |
5875 | static int ocfs2_xattr_value_attach_refcount(struct inode *inode, | |
5876 | struct ocfs2_xattr_value_root *xv, | |
5877 | struct ocfs2_extent_tree *value_et, | |
5878 | struct ocfs2_caching_info *ref_ci, | |
5879 | struct buffer_head *ref_root_bh, | |
5880 | struct ocfs2_cached_dealloc_ctxt *dealloc, | |
5881 | struct ocfs2_post_refcount *refcount) | |
5882 | { | |
5883 | int ret = 0; | |
5884 | u32 clusters = le32_to_cpu(xv->xr_clusters); | |
5885 | u32 cpos, p_cluster, num_clusters; | |
5886 | struct ocfs2_extent_list *el = &xv->xr_list; | |
5887 | unsigned int ext_flags; | |
5888 | ||
5889 | cpos = 0; | |
5890 | while (cpos < clusters) { | |
5891 | ret = ocfs2_xattr_get_clusters(inode, cpos, &p_cluster, | |
5892 | &num_clusters, el, &ext_flags); | |
17caf955 JQ |
5893 | if (ret) { |
5894 | mlog_errno(ret); | |
5895 | break; | |
5896 | } | |
0129241e TM |
5897 | |
5898 | cpos += num_clusters; | |
5899 | if ((ext_flags & OCFS2_EXT_REFCOUNTED)) | |
5900 | continue; | |
5901 | ||
5902 | BUG_ON(!p_cluster); | |
5903 | ||
5904 | ret = ocfs2_add_refcount_flag(inode, value_et, | |
5905 | ref_ci, ref_root_bh, | |
5906 | cpos - num_clusters, | |
5907 | p_cluster, num_clusters, | |
5908 | dealloc, refcount); | |
5909 | if (ret) { | |
5910 | mlog_errno(ret); | |
5911 | break; | |
5912 | } | |
5913 | } | |
5914 | ||
5915 | return ret; | |
5916 | } | |
5917 | ||
5918 | /* | |
5919 | * Given a normal ocfs2_xattr_header, refcount all the entries which | |
5920 | * have value stored outside. | |
5921 | * Used for xattrs stored in inode and ocfs2_xattr_block. | |
5922 | */ | |
5923 | static int ocfs2_xattr_attach_refcount_normal(struct inode *inode, | |
5924 | struct ocfs2_xattr_value_buf *vb, | |
5925 | struct ocfs2_xattr_header *header, | |
5926 | struct ocfs2_caching_info *ref_ci, | |
5927 | struct buffer_head *ref_root_bh, | |
5928 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
5929 | { | |
5930 | ||
5931 | struct ocfs2_xattr_entry *xe; | |
5932 | struct ocfs2_xattr_value_root *xv; | |
5933 | struct ocfs2_extent_tree et; | |
5934 | int i, ret = 0; | |
5935 | ||
5936 | for (i = 0; i < le16_to_cpu(header->xh_count); i++) { | |
5937 | xe = &header->xh_entries[i]; | |
5938 | ||
5939 | if (ocfs2_xattr_is_local(xe)) | |
5940 | continue; | |
5941 | ||
5942 | xv = (struct ocfs2_xattr_value_root *)((void *)header + | |
5943 | le16_to_cpu(xe->xe_name_offset) + | |
5944 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
5945 | ||
5946 | vb->vb_xv = xv; | |
5947 | ocfs2_init_xattr_value_extent_tree(&et, INODE_CACHE(inode), vb); | |
5948 | ||
5949 | ret = ocfs2_xattr_value_attach_refcount(inode, xv, &et, | |
5950 | ref_ci, ref_root_bh, | |
5951 | dealloc, NULL); | |
5952 | if (ret) { | |
5953 | mlog_errno(ret); | |
5954 | break; | |
5955 | } | |
5956 | } | |
5957 | ||
5958 | return ret; | |
5959 | } | |
5960 | ||
5961 | static int ocfs2_xattr_inline_attach_refcount(struct inode *inode, | |
5962 | struct buffer_head *fe_bh, | |
5963 | struct ocfs2_caching_info *ref_ci, | |
5964 | struct buffer_head *ref_root_bh, | |
5965 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
5966 | { | |
5967 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
5968 | struct ocfs2_xattr_header *header = (struct ocfs2_xattr_header *) | |
5969 | (fe_bh->b_data + inode->i_sb->s_blocksize - | |
5970 | le16_to_cpu(di->i_xattr_inline_size)); | |
5971 | struct ocfs2_xattr_value_buf vb = { | |
5972 | .vb_bh = fe_bh, | |
5973 | .vb_access = ocfs2_journal_access_di, | |
5974 | }; | |
5975 | ||
5976 | return ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
5977 | ref_ci, ref_root_bh, dealloc); | |
5978 | } | |
5979 | ||
5980 | struct ocfs2_xattr_tree_value_refcount_para { | |
5981 | struct ocfs2_caching_info *ref_ci; | |
5982 | struct buffer_head *ref_root_bh; | |
5983 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
5984 | }; | |
5985 | ||
5986 | static int ocfs2_get_xattr_tree_value_root(struct super_block *sb, | |
5987 | struct ocfs2_xattr_bucket *bucket, | |
5988 | int offset, | |
5989 | struct ocfs2_xattr_value_root **xv, | |
5990 | struct buffer_head **bh) | |
5991 | { | |
5992 | int ret, block_off, name_offset; | |
5993 | struct ocfs2_xattr_header *xh = bucket_xh(bucket); | |
5994 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
5995 | void *base; | |
5996 | ||
5997 | ret = ocfs2_xattr_bucket_get_name_value(sb, | |
5998 | bucket_xh(bucket), | |
5999 | offset, | |
6000 | &block_off, | |
6001 | &name_offset); | |
6002 | if (ret) { | |
6003 | mlog_errno(ret); | |
6004 | goto out; | |
6005 | } | |
6006 | ||
6007 | base = bucket_block(bucket, block_off); | |
6008 | ||
6009 | *xv = (struct ocfs2_xattr_value_root *)(base + name_offset + | |
6010 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6011 | ||
6012 | if (bh) | |
6013 | *bh = bucket->bu_bhs[block_off]; | |
6014 | out: | |
6015 | return ret; | |
6016 | } | |
6017 | ||
6018 | /* | |
6019 | * For a given xattr bucket, refcount all the entries which | |
6020 | * have value stored outside. | |
6021 | */ | |
6022 | static int ocfs2_xattr_bucket_value_refcount(struct inode *inode, | |
6023 | struct ocfs2_xattr_bucket *bucket, | |
6024 | void *para) | |
6025 | { | |
6026 | int i, ret = 0; | |
6027 | struct ocfs2_extent_tree et; | |
6028 | struct ocfs2_xattr_tree_value_refcount_para *ref = | |
6029 | (struct ocfs2_xattr_tree_value_refcount_para *)para; | |
6030 | struct ocfs2_xattr_header *xh = | |
6031 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6032 | struct ocfs2_xattr_entry *xe; | |
6033 | struct ocfs2_xattr_value_buf vb = { | |
6034 | .vb_access = ocfs2_journal_access, | |
6035 | }; | |
6036 | struct ocfs2_post_refcount refcount = { | |
6037 | .credits = bucket->bu_blocks, | |
6038 | .para = bucket, | |
6039 | .func = ocfs2_xattr_bucket_post_refcount, | |
6040 | }; | |
6041 | struct ocfs2_post_refcount *p = NULL; | |
6042 | ||
6043 | /* We only need post_refcount if we support metaecc. */ | |
6044 | if (ocfs2_meta_ecc(OCFS2_SB(inode->i_sb))) | |
6045 | p = &refcount; | |
6046 | ||
402b4183 TM |
6047 | trace_ocfs2_xattr_bucket_value_refcount( |
6048 | (unsigned long long)bucket_blkno(bucket), | |
6049 | le16_to_cpu(xh->xh_count)); | |
0129241e TM |
6050 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { |
6051 | xe = &xh->xh_entries[i]; | |
6052 | ||
6053 | if (ocfs2_xattr_is_local(xe)) | |
6054 | continue; | |
6055 | ||
6056 | ret = ocfs2_get_xattr_tree_value_root(inode->i_sb, bucket, i, | |
6057 | &vb.vb_xv, &vb.vb_bh); | |
6058 | if (ret) { | |
6059 | mlog_errno(ret); | |
6060 | break; | |
6061 | } | |
6062 | ||
6063 | ocfs2_init_xattr_value_extent_tree(&et, | |
6064 | INODE_CACHE(inode), &vb); | |
6065 | ||
6066 | ret = ocfs2_xattr_value_attach_refcount(inode, vb.vb_xv, | |
6067 | &et, ref->ref_ci, | |
6068 | ref->ref_root_bh, | |
6069 | ref->dealloc, p); | |
6070 | if (ret) { | |
6071 | mlog_errno(ret); | |
6072 | break; | |
6073 | } | |
6074 | } | |
6075 | ||
6076 | return ret; | |
6077 | ||
6078 | } | |
6079 | ||
6080 | static int ocfs2_refcount_xattr_tree_rec(struct inode *inode, | |
6081 | struct buffer_head *root_bh, | |
6082 | u64 blkno, u32 cpos, u32 len, void *para) | |
6083 | { | |
6084 | return ocfs2_iterate_xattr_buckets(inode, blkno, len, | |
6085 | ocfs2_xattr_bucket_value_refcount, | |
6086 | para); | |
6087 | } | |
6088 | ||
6089 | static int ocfs2_xattr_block_attach_refcount(struct inode *inode, | |
6090 | struct buffer_head *blk_bh, | |
6091 | struct ocfs2_caching_info *ref_ci, | |
6092 | struct buffer_head *ref_root_bh, | |
6093 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6094 | { | |
6095 | int ret = 0; | |
6096 | struct ocfs2_xattr_block *xb = | |
6097 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6098 | ||
6099 | if (!(le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED)) { | |
6100 | struct ocfs2_xattr_header *header = &xb->xb_attrs.xb_header; | |
6101 | struct ocfs2_xattr_value_buf vb = { | |
6102 | .vb_bh = blk_bh, | |
6103 | .vb_access = ocfs2_journal_access_xb, | |
6104 | }; | |
6105 | ||
6106 | ret = ocfs2_xattr_attach_refcount_normal(inode, &vb, header, | |
6107 | ref_ci, ref_root_bh, | |
6108 | dealloc); | |
6109 | } else { | |
6110 | struct ocfs2_xattr_tree_value_refcount_para para = { | |
6111 | .ref_ci = ref_ci, | |
6112 | .ref_root_bh = ref_root_bh, | |
6113 | .dealloc = dealloc, | |
6114 | }; | |
6115 | ||
6116 | ret = ocfs2_iterate_xattr_index_block(inode, blk_bh, | |
6117 | ocfs2_refcount_xattr_tree_rec, | |
6118 | ¶); | |
6119 | } | |
6120 | ||
6121 | return ret; | |
6122 | } | |
6123 | ||
6124 | int ocfs2_xattr_attach_refcount_tree(struct inode *inode, | |
6125 | struct buffer_head *fe_bh, | |
6126 | struct ocfs2_caching_info *ref_ci, | |
6127 | struct buffer_head *ref_root_bh, | |
6128 | struct ocfs2_cached_dealloc_ctxt *dealloc) | |
6129 | { | |
6130 | int ret = 0; | |
6131 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
6132 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)fe_bh->b_data; | |
6133 | struct buffer_head *blk_bh = NULL; | |
6134 | ||
6135 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
6136 | ret = ocfs2_xattr_inline_attach_refcount(inode, fe_bh, | |
6137 | ref_ci, ref_root_bh, | |
6138 | dealloc); | |
6139 | if (ret) { | |
6140 | mlog_errno(ret); | |
6141 | goto out; | |
6142 | } | |
6143 | } | |
6144 | ||
6145 | if (!di->i_xattr_loc) | |
6146 | goto out; | |
6147 | ||
6148 | ret = ocfs2_read_xattr_block(inode, le64_to_cpu(di->i_xattr_loc), | |
6149 | &blk_bh); | |
6150 | if (ret < 0) { | |
6151 | mlog_errno(ret); | |
6152 | goto out; | |
6153 | } | |
6154 | ||
6155 | ret = ocfs2_xattr_block_attach_refcount(inode, blk_bh, ref_ci, | |
6156 | ref_root_bh, dealloc); | |
6157 | if (ret) | |
6158 | mlog_errno(ret); | |
6159 | ||
6160 | brelse(blk_bh); | |
6161 | out: | |
6162 | ||
6163 | return ret; | |
6164 | } | |
6165 | ||
0fe9b66c | 6166 | typedef int (should_xattr_reflinked)(struct ocfs2_xattr_entry *xe); |
2999d12f TM |
6167 | /* |
6168 | * Store the information we need in xattr reflink. | |
6169 | * old_bh and new_bh are inode bh for the old and new inode. | |
6170 | */ | |
6171 | struct ocfs2_xattr_reflink { | |
6172 | struct inode *old_inode; | |
6173 | struct inode *new_inode; | |
6174 | struct buffer_head *old_bh; | |
6175 | struct buffer_head *new_bh; | |
6176 | struct ocfs2_caching_info *ref_ci; | |
6177 | struct buffer_head *ref_root_bh; | |
6178 | struct ocfs2_cached_dealloc_ctxt *dealloc; | |
0fe9b66c | 6179 | should_xattr_reflinked *xattr_reflinked; |
2999d12f TM |
6180 | }; |
6181 | ||
6182 | /* | |
6183 | * Given a xattr header and xe offset, | |
6184 | * return the proper xv and the corresponding bh. | |
6185 | * xattr in inode, block and xattr tree have different implementaions. | |
6186 | */ | |
6187 | typedef int (get_xattr_value_root)(struct super_block *sb, | |
6188 | struct buffer_head *bh, | |
6189 | struct ocfs2_xattr_header *xh, | |
6190 | int offset, | |
6191 | struct ocfs2_xattr_value_root **xv, | |
6192 | struct buffer_head **ret_bh, | |
6193 | void *para); | |
6194 | ||
6195 | /* | |
6196 | * Calculate all the xattr value root metadata stored in this xattr header and | |
6197 | * credits we need if we create them from the scratch. | |
6198 | * We use get_xattr_value_root so that all types of xattr container can use it. | |
6199 | */ | |
6200 | static int ocfs2_value_metas_in_xattr_header(struct super_block *sb, | |
6201 | struct buffer_head *bh, | |
6202 | struct ocfs2_xattr_header *xh, | |
6203 | int *metas, int *credits, | |
6204 | int *num_recs, | |
6205 | get_xattr_value_root *func, | |
6206 | void *para) | |
6207 | { | |
6208 | int i, ret = 0; | |
6209 | struct ocfs2_xattr_value_root *xv; | |
6210 | struct ocfs2_xattr_entry *xe; | |
6211 | ||
6212 | for (i = 0; i < le16_to_cpu(xh->xh_count); i++) { | |
6213 | xe = &xh->xh_entries[i]; | |
6214 | if (ocfs2_xattr_is_local(xe)) | |
6215 | continue; | |
6216 | ||
6217 | ret = func(sb, bh, xh, i, &xv, NULL, para); | |
6218 | if (ret) { | |
6219 | mlog_errno(ret); | |
6220 | break; | |
6221 | } | |
6222 | ||
6223 | *metas += le16_to_cpu(xv->xr_list.l_tree_depth) * | |
6224 | le16_to_cpu(xv->xr_list.l_next_free_rec); | |
6225 | ||
6226 | *credits += ocfs2_calc_extend_credits(sb, | |
06f9da6e | 6227 | &def_xv.xv.xr_list); |
2999d12f TM |
6228 | |
6229 | /* | |
6230 | * If the value is a tree with depth > 1, We don't go deep | |
6231 | * to the extent block, so just calculate a maximum record num. | |
6232 | */ | |
6233 | if (!xv->xr_list.l_tree_depth) | |
8ff6af88 | 6234 | *num_recs += le16_to_cpu(xv->xr_list.l_next_free_rec); |
2999d12f TM |
6235 | else |
6236 | *num_recs += ocfs2_clusters_for_bytes(sb, | |
6237 | XATTR_SIZE_MAX); | |
6238 | } | |
6239 | ||
6240 | return ret; | |
6241 | } | |
6242 | ||
6243 | /* Used by xattr inode and block to return the right xv and buffer_head. */ | |
6244 | static int ocfs2_get_xattr_value_root(struct super_block *sb, | |
6245 | struct buffer_head *bh, | |
6246 | struct ocfs2_xattr_header *xh, | |
6247 | int offset, | |
6248 | struct ocfs2_xattr_value_root **xv, | |
6249 | struct buffer_head **ret_bh, | |
6250 | void *para) | |
6251 | { | |
6252 | struct ocfs2_xattr_entry *xe = &xh->xh_entries[offset]; | |
6253 | ||
6254 | *xv = (struct ocfs2_xattr_value_root *)((void *)xh + | |
6255 | le16_to_cpu(xe->xe_name_offset) + | |
6256 | OCFS2_XATTR_SIZE(xe->xe_name_len)); | |
6257 | ||
6258 | if (ret_bh) | |
6259 | *ret_bh = bh; | |
6260 | ||
6261 | return 0; | |
6262 | } | |
6263 | ||
6264 | /* | |
6265 | * Lock the meta_ac and caculate how much credits we need for reflink xattrs. | |
6266 | * It is only used for inline xattr and xattr block. | |
6267 | */ | |
6268 | static int ocfs2_reflink_lock_xattr_allocators(struct ocfs2_super *osb, | |
6269 | struct ocfs2_xattr_header *xh, | |
6270 | struct buffer_head *ref_root_bh, | |
6271 | int *credits, | |
6272 | struct ocfs2_alloc_context **meta_ac) | |
6273 | { | |
6274 | int ret, meta_add = 0, num_recs = 0; | |
6275 | struct ocfs2_refcount_block *rb = | |
6276 | (struct ocfs2_refcount_block *)ref_root_bh->b_data; | |
6277 | ||
6278 | *credits = 0; | |
6279 | ||
6280 | ret = ocfs2_value_metas_in_xattr_header(osb->sb, NULL, xh, | |
6281 | &meta_add, credits, &num_recs, | |
6282 | ocfs2_get_xattr_value_root, | |
6283 | NULL); | |
6284 | if (ret) { | |
6285 | mlog_errno(ret); | |
6286 | goto out; | |
6287 | } | |
6288 | ||
6289 | /* | |
6290 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6291 | * an approximate number we need for refcount tree change. | |
6292 | * Sometimes we need to split the tree, and after split, half recs | |
6293 | * will be moved to the new block, and a new block can only provide | |
6294 | * half number of recs. So we multiple new blocks by 2. | |
6295 | */ | |
6296 | num_recs = num_recs / ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6297 | meta_add += num_recs; | |
6298 | *credits += num_recs + num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6299 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6300 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6301 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6302 | else | |
6303 | *credits += 1; | |
6304 | ||
6305 | ret = ocfs2_reserve_new_metadata_blocks(osb, meta_add, meta_ac); | |
6306 | if (ret) | |
6307 | mlog_errno(ret); | |
6308 | ||
6309 | out: | |
6310 | return ret; | |
6311 | } | |
6312 | ||
6313 | /* | |
6314 | * Given a xattr header, reflink all the xattrs in this container. | |
6315 | * It can be used for inode, block and bucket. | |
6316 | * | |
6317 | * NOTE: | |
6318 | * Before we call this function, the caller has memcpy the xattr in | |
6319 | * old_xh to the new_xh. | |
0fe9b66c TM |
6320 | * |
6321 | * If args.xattr_reflinked is set, call it to decide whether the xe should | |
6322 | * be reflinked or not. If not, remove it from the new xattr header. | |
2999d12f TM |
6323 | */ |
6324 | static int ocfs2_reflink_xattr_header(handle_t *handle, | |
6325 | struct ocfs2_xattr_reflink *args, | |
6326 | struct buffer_head *old_bh, | |
6327 | struct ocfs2_xattr_header *xh, | |
6328 | struct buffer_head *new_bh, | |
6329 | struct ocfs2_xattr_header *new_xh, | |
6330 | struct ocfs2_xattr_value_buf *vb, | |
6331 | struct ocfs2_alloc_context *meta_ac, | |
6332 | get_xattr_value_root *func, | |
6333 | void *para) | |
6334 | { | |
0fe9b66c | 6335 | int ret = 0, i, j; |
2999d12f TM |
6336 | struct super_block *sb = args->old_inode->i_sb; |
6337 | struct buffer_head *value_bh; | |
0fe9b66c | 6338 | struct ocfs2_xattr_entry *xe, *last; |
2999d12f TM |
6339 | struct ocfs2_xattr_value_root *xv, *new_xv; |
6340 | struct ocfs2_extent_tree data_et; | |
6341 | u32 clusters, cpos, p_cluster, num_clusters; | |
6342 | unsigned int ext_flags = 0; | |
6343 | ||
402b4183 TM |
6344 | trace_ocfs2_reflink_xattr_header((unsigned long long)old_bh->b_blocknr, |
6345 | le16_to_cpu(xh->xh_count)); | |
0fe9b66c TM |
6346 | |
6347 | last = &new_xh->xh_entries[le16_to_cpu(new_xh->xh_count)]; | |
6348 | for (i = 0, j = 0; i < le16_to_cpu(xh->xh_count); i++, j++) { | |
2999d12f TM |
6349 | xe = &xh->xh_entries[i]; |
6350 | ||
0fe9b66c TM |
6351 | if (args->xattr_reflinked && !args->xattr_reflinked(xe)) { |
6352 | xe = &new_xh->xh_entries[j]; | |
6353 | ||
6354 | le16_add_cpu(&new_xh->xh_count, -1); | |
6355 | if (new_xh->xh_count) { | |
6356 | memmove(xe, xe + 1, | |
6357 | (void *)last - (void *)xe); | |
6358 | memset(last, 0, | |
6359 | sizeof(struct ocfs2_xattr_entry)); | |
6360 | } | |
6361 | ||
6362 | /* | |
6363 | * We don't want j to increase in the next round since | |
6364 | * it is already moved ahead. | |
6365 | */ | |
6366 | j--; | |
6367 | continue; | |
6368 | } | |
6369 | ||
2999d12f TM |
6370 | if (ocfs2_xattr_is_local(xe)) |
6371 | continue; | |
6372 | ||
6373 | ret = func(sb, old_bh, xh, i, &xv, NULL, para); | |
6374 | if (ret) { | |
6375 | mlog_errno(ret); | |
6376 | break; | |
6377 | } | |
6378 | ||
0fe9b66c | 6379 | ret = func(sb, new_bh, new_xh, j, &new_xv, &value_bh, para); |
2999d12f TM |
6380 | if (ret) { |
6381 | mlog_errno(ret); | |
6382 | break; | |
6383 | } | |
6384 | ||
6385 | /* | |
6386 | * For the xattr which has l_tree_depth = 0, all the extent | |
6387 | * recs have already be copied to the new xh with the | |
6388 | * propriate OCFS2_EXT_REFCOUNTED flag we just need to | |
6389 | * increase the refount count int the refcount tree. | |
6390 | * | |
6391 | * For the xattr which has l_tree_depth > 0, we need | |
6392 | * to initialize it to the empty default value root, | |
6393 | * and then insert the extents one by one. | |
6394 | */ | |
6395 | if (xv->xr_list.l_tree_depth) { | |
6396 | memcpy(new_xv, &def_xv, sizeof(def_xv)); | |
6397 | vb->vb_xv = new_xv; | |
6398 | vb->vb_bh = value_bh; | |
6399 | ocfs2_init_xattr_value_extent_tree(&data_et, | |
6400 | INODE_CACHE(args->new_inode), vb); | |
6401 | } | |
6402 | ||
6403 | clusters = le32_to_cpu(xv->xr_clusters); | |
6404 | cpos = 0; | |
6405 | while (cpos < clusters) { | |
6406 | ret = ocfs2_xattr_get_clusters(args->old_inode, | |
6407 | cpos, | |
6408 | &p_cluster, | |
6409 | &num_clusters, | |
6410 | &xv->xr_list, | |
6411 | &ext_flags); | |
6412 | if (ret) { | |
6413 | mlog_errno(ret); | |
6414 | goto out; | |
6415 | } | |
6416 | ||
6417 | BUG_ON(!p_cluster); | |
6418 | ||
6419 | if (xv->xr_list.l_tree_depth) { | |
6420 | ret = ocfs2_insert_extent(handle, | |
6421 | &data_et, cpos, | |
6422 | ocfs2_clusters_to_blocks( | |
6423 | args->old_inode->i_sb, | |
6424 | p_cluster), | |
6425 | num_clusters, ext_flags, | |
6426 | meta_ac); | |
6427 | if (ret) { | |
6428 | mlog_errno(ret); | |
6429 | goto out; | |
6430 | } | |
6431 | } | |
6432 | ||
6433 | ret = ocfs2_increase_refcount(handle, args->ref_ci, | |
6434 | args->ref_root_bh, | |
6435 | p_cluster, num_clusters, | |
6436 | meta_ac, args->dealloc); | |
6437 | if (ret) { | |
6438 | mlog_errno(ret); | |
6439 | goto out; | |
6440 | } | |
6441 | ||
6442 | cpos += num_clusters; | |
6443 | } | |
6444 | } | |
6445 | ||
6446 | out: | |
6447 | return ret; | |
6448 | } | |
6449 | ||
6450 | static int ocfs2_reflink_xattr_inline(struct ocfs2_xattr_reflink *args) | |
6451 | { | |
6452 | int ret = 0, credits = 0; | |
6453 | handle_t *handle; | |
6454 | struct ocfs2_super *osb = OCFS2_SB(args->old_inode->i_sb); | |
6455 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)args->old_bh->b_data; | |
6456 | int inline_size = le16_to_cpu(di->i_xattr_inline_size); | |
6457 | int header_off = osb->sb->s_blocksize - inline_size; | |
6458 | struct ocfs2_xattr_header *xh = (struct ocfs2_xattr_header *) | |
6459 | (args->old_bh->b_data + header_off); | |
6460 | struct ocfs2_xattr_header *new_xh = (struct ocfs2_xattr_header *) | |
6461 | (args->new_bh->b_data + header_off); | |
6462 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6463 | struct ocfs2_inode_info *new_oi; | |
6464 | struct ocfs2_dinode *new_di; | |
6465 | struct ocfs2_xattr_value_buf vb = { | |
6466 | .vb_bh = args->new_bh, | |
6467 | .vb_access = ocfs2_journal_access_di, | |
6468 | }; | |
6469 | ||
6470 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6471 | &credits, &meta_ac); | |
6472 | if (ret) { | |
6473 | mlog_errno(ret); | |
6474 | goto out; | |
6475 | } | |
6476 | ||
6477 | handle = ocfs2_start_trans(osb, credits); | |
6478 | if (IS_ERR(handle)) { | |
6479 | ret = PTR_ERR(handle); | |
6480 | mlog_errno(ret); | |
6481 | goto out; | |
6482 | } | |
6483 | ||
6484 | ret = ocfs2_journal_access_di(handle, INODE_CACHE(args->new_inode), | |
6485 | args->new_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6486 | if (ret) { | |
6487 | mlog_errno(ret); | |
6488 | goto out_commit; | |
6489 | } | |
6490 | ||
6491 | memcpy(args->new_bh->b_data + header_off, | |
6492 | args->old_bh->b_data + header_off, inline_size); | |
6493 | ||
6494 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6495 | new_di->i_xattr_inline_size = cpu_to_le16(inline_size); | |
6496 | ||
6497 | ret = ocfs2_reflink_xattr_header(handle, args, args->old_bh, xh, | |
6498 | args->new_bh, new_xh, &vb, meta_ac, | |
6499 | ocfs2_get_xattr_value_root, NULL); | |
6500 | if (ret) { | |
6501 | mlog_errno(ret); | |
6502 | goto out_commit; | |
6503 | } | |
6504 | ||
6505 | new_oi = OCFS2_I(args->new_inode); | |
ef962df0 JB |
6506 | /* |
6507 | * Adjust extent record count to reserve space for extended attribute. | |
6508 | * Inline data count had been adjusted in ocfs2_duplicate_inline_data(). | |
6509 | */ | |
6510 | if (!(new_oi->ip_dyn_features & OCFS2_INLINE_DATA_FL) && | |
6511 | !(ocfs2_inode_is_fast_symlink(args->new_inode))) { | |
6512 | struct ocfs2_extent_list *el = &new_di->id2.i_list; | |
6513 | le16_add_cpu(&el->l_count, -(inline_size / | |
6514 | sizeof(struct ocfs2_extent_rec))); | |
6515 | } | |
2999d12f TM |
6516 | spin_lock(&new_oi->ip_lock); |
6517 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL | OCFS2_INLINE_XATTR_FL; | |
6518 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6519 | spin_unlock(&new_oi->ip_lock); | |
6520 | ||
6521 | ocfs2_journal_dirty(handle, args->new_bh); | |
6522 | ||
6523 | out_commit: | |
6524 | ocfs2_commit_trans(osb, handle); | |
6525 | ||
6526 | out: | |
6527 | if (meta_ac) | |
6528 | ocfs2_free_alloc_context(meta_ac); | |
6529 | return ret; | |
6530 | } | |
6531 | ||
6532 | static int ocfs2_create_empty_xattr_block(struct inode *inode, | |
6533 | struct buffer_head *fe_bh, | |
6534 | struct buffer_head **ret_bh, | |
6535 | int indexed) | |
6536 | { | |
6537 | int ret; | |
2999d12f | 6538 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
b2317968 | 6539 | struct ocfs2_xattr_set_ctxt ctxt; |
2999d12f | 6540 | |
b2317968 TM |
6541 | memset(&ctxt, 0, sizeof(ctxt)); |
6542 | ret = ocfs2_reserve_new_metadata_blocks(osb, 1, &ctxt.meta_ac); | |
2999d12f TM |
6543 | if (ret < 0) { |
6544 | mlog_errno(ret); | |
6545 | return ret; | |
6546 | } | |
6547 | ||
d3981544 JB |
6548 | ctxt.handle = ocfs2_start_trans(osb, OCFS2_XATTR_BLOCK_CREATE_CREDITS); |
6549 | if (IS_ERR(ctxt.handle)) { | |
6550 | ret = PTR_ERR(ctxt.handle); | |
2999d12f TM |
6551 | mlog_errno(ret); |
6552 | goto out; | |
6553 | } | |
6554 | ||
402b4183 TM |
6555 | trace_ocfs2_create_empty_xattr_block( |
6556 | (unsigned long long)fe_bh->b_blocknr, indexed); | |
d3981544 JB |
6557 | ret = ocfs2_create_xattr_block(inode, fe_bh, &ctxt, indexed, |
6558 | ret_bh); | |
2999d12f TM |
6559 | if (ret) |
6560 | mlog_errno(ret); | |
6561 | ||
d3981544 | 6562 | ocfs2_commit_trans(osb, ctxt.handle); |
2999d12f | 6563 | out: |
b2317968 | 6564 | ocfs2_free_alloc_context(ctxt.meta_ac); |
2999d12f TM |
6565 | return ret; |
6566 | } | |
6567 | ||
6568 | static int ocfs2_reflink_xattr_block(struct ocfs2_xattr_reflink *args, | |
6569 | struct buffer_head *blk_bh, | |
6570 | struct buffer_head *new_blk_bh) | |
6571 | { | |
6572 | int ret = 0, credits = 0; | |
6573 | handle_t *handle; | |
6574 | struct ocfs2_inode_info *new_oi = OCFS2_I(args->new_inode); | |
6575 | struct ocfs2_dinode *new_di; | |
6576 | struct ocfs2_super *osb = OCFS2_SB(args->new_inode->i_sb); | |
6577 | int header_off = offsetof(struct ocfs2_xattr_block, xb_attrs.xb_header); | |
6578 | struct ocfs2_xattr_block *xb = | |
6579 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
6580 | struct ocfs2_xattr_header *xh = &xb->xb_attrs.xb_header; | |
6581 | struct ocfs2_xattr_block *new_xb = | |
6582 | (struct ocfs2_xattr_block *)new_blk_bh->b_data; | |
6583 | struct ocfs2_xattr_header *new_xh = &new_xb->xb_attrs.xb_header; | |
6584 | struct ocfs2_alloc_context *meta_ac; | |
6585 | struct ocfs2_xattr_value_buf vb = { | |
6586 | .vb_bh = new_blk_bh, | |
6587 | .vb_access = ocfs2_journal_access_xb, | |
6588 | }; | |
6589 | ||
6590 | ret = ocfs2_reflink_lock_xattr_allocators(osb, xh, args->ref_root_bh, | |
6591 | &credits, &meta_ac); | |
6592 | if (ret) { | |
6593 | mlog_errno(ret); | |
6594 | return ret; | |
6595 | } | |
6596 | ||
6597 | /* One more credits in case we need to add xattr flags in new inode. */ | |
6598 | handle = ocfs2_start_trans(osb, credits + 1); | |
6599 | if (IS_ERR(handle)) { | |
6600 | ret = PTR_ERR(handle); | |
6601 | mlog_errno(ret); | |
6602 | goto out; | |
6603 | } | |
6604 | ||
6605 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6606 | ret = ocfs2_journal_access_di(handle, | |
6607 | INODE_CACHE(args->new_inode), | |
6608 | args->new_bh, | |
6609 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6610 | if (ret) { | |
6611 | mlog_errno(ret); | |
6612 | goto out_commit; | |
6613 | } | |
6614 | } | |
6615 | ||
6616 | ret = ocfs2_journal_access_xb(handle, INODE_CACHE(args->new_inode), | |
6617 | new_blk_bh, OCFS2_JOURNAL_ACCESS_WRITE); | |
6618 | if (ret) { | |
6619 | mlog_errno(ret); | |
6620 | goto out_commit; | |
6621 | } | |
6622 | ||
6623 | memcpy(new_blk_bh->b_data + header_off, blk_bh->b_data + header_off, | |
6624 | osb->sb->s_blocksize - header_off); | |
6625 | ||
6626 | ret = ocfs2_reflink_xattr_header(handle, args, blk_bh, xh, | |
6627 | new_blk_bh, new_xh, &vb, meta_ac, | |
6628 | ocfs2_get_xattr_value_root, NULL); | |
6629 | if (ret) { | |
6630 | mlog_errno(ret); | |
6631 | goto out_commit; | |
6632 | } | |
6633 | ||
6634 | ocfs2_journal_dirty(handle, new_blk_bh); | |
6635 | ||
6636 | if (!(new_oi->ip_dyn_features & OCFS2_HAS_XATTR_FL)) { | |
6637 | new_di = (struct ocfs2_dinode *)args->new_bh->b_data; | |
6638 | spin_lock(&new_oi->ip_lock); | |
6639 | new_oi->ip_dyn_features |= OCFS2_HAS_XATTR_FL; | |
6640 | new_di->i_dyn_features = cpu_to_le16(new_oi->ip_dyn_features); | |
6641 | spin_unlock(&new_oi->ip_lock); | |
6642 | ||
6643 | ocfs2_journal_dirty(handle, args->new_bh); | |
6644 | } | |
6645 | ||
6646 | out_commit: | |
6647 | ocfs2_commit_trans(osb, handle); | |
6648 | ||
6649 | out: | |
6650 | ocfs2_free_alloc_context(meta_ac); | |
6651 | return ret; | |
6652 | } | |
6653 | ||
6654 | struct ocfs2_reflink_xattr_tree_args { | |
6655 | struct ocfs2_xattr_reflink *reflink; | |
6656 | struct buffer_head *old_blk_bh; | |
6657 | struct buffer_head *new_blk_bh; | |
6658 | struct ocfs2_xattr_bucket *old_bucket; | |
6659 | struct ocfs2_xattr_bucket *new_bucket; | |
6660 | }; | |
6661 | ||
6662 | /* | |
6663 | * NOTE: | |
6664 | * We have to handle the case that both old bucket and new bucket | |
6665 | * will call this function to get the right ret_bh. | |
6666 | * So The caller must give us the right bh. | |
6667 | */ | |
6668 | static int ocfs2_get_reflink_xattr_value_root(struct super_block *sb, | |
6669 | struct buffer_head *bh, | |
6670 | struct ocfs2_xattr_header *xh, | |
6671 | int offset, | |
6672 | struct ocfs2_xattr_value_root **xv, | |
6673 | struct buffer_head **ret_bh, | |
6674 | void *para) | |
6675 | { | |
6676 | struct ocfs2_reflink_xattr_tree_args *args = | |
6677 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6678 | struct ocfs2_xattr_bucket *bucket; | |
6679 | ||
6680 | if (bh == args->old_bucket->bu_bhs[0]) | |
6681 | bucket = args->old_bucket; | |
6682 | else | |
6683 | bucket = args->new_bucket; | |
6684 | ||
6685 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6686 | xv, ret_bh); | |
6687 | } | |
6688 | ||
6689 | struct ocfs2_value_tree_metas { | |
6690 | int num_metas; | |
6691 | int credits; | |
6692 | int num_recs; | |
6693 | }; | |
6694 | ||
6695 | static int ocfs2_value_tree_metas_in_bucket(struct super_block *sb, | |
6696 | struct buffer_head *bh, | |
6697 | struct ocfs2_xattr_header *xh, | |
6698 | int offset, | |
6699 | struct ocfs2_xattr_value_root **xv, | |
6700 | struct buffer_head **ret_bh, | |
6701 | void *para) | |
6702 | { | |
6703 | struct ocfs2_xattr_bucket *bucket = | |
6704 | (struct ocfs2_xattr_bucket *)para; | |
6705 | ||
6706 | return ocfs2_get_xattr_tree_value_root(sb, bucket, offset, | |
6707 | xv, ret_bh); | |
6708 | } | |
6709 | ||
6710 | static int ocfs2_calc_value_tree_metas(struct inode *inode, | |
6711 | struct ocfs2_xattr_bucket *bucket, | |
6712 | void *para) | |
6713 | { | |
6714 | struct ocfs2_value_tree_metas *metas = | |
6715 | (struct ocfs2_value_tree_metas *)para; | |
6716 | struct ocfs2_xattr_header *xh = | |
6717 | (struct ocfs2_xattr_header *)bucket->bu_bhs[0]->b_data; | |
6718 | ||
6719 | /* Add the credits for this bucket first. */ | |
6720 | metas->credits += bucket->bu_blocks; | |
6721 | return ocfs2_value_metas_in_xattr_header(inode->i_sb, bucket->bu_bhs[0], | |
6722 | xh, &metas->num_metas, | |
6723 | &metas->credits, &metas->num_recs, | |
6724 | ocfs2_value_tree_metas_in_bucket, | |
6725 | bucket); | |
6726 | } | |
6727 | ||
6728 | /* | |
6729 | * Given a xattr extent rec starting from blkno and having len clusters, | |
6730 | * iterate all the buckets calculate how much metadata we need for reflinking | |
6731 | * all the ocfs2_xattr_value_root and lock the allocators accordingly. | |
6732 | */ | |
6733 | static int ocfs2_lock_reflink_xattr_rec_allocators( | |
6734 | struct ocfs2_reflink_xattr_tree_args *args, | |
6735 | struct ocfs2_extent_tree *xt_et, | |
6736 | u64 blkno, u32 len, int *credits, | |
6737 | struct ocfs2_alloc_context **meta_ac, | |
6738 | struct ocfs2_alloc_context **data_ac) | |
6739 | { | |
6740 | int ret, num_free_extents; | |
6741 | struct ocfs2_value_tree_metas metas; | |
6742 | struct ocfs2_super *osb = OCFS2_SB(args->reflink->old_inode->i_sb); | |
6743 | struct ocfs2_refcount_block *rb; | |
6744 | ||
6745 | memset(&metas, 0, sizeof(metas)); | |
6746 | ||
6747 | ret = ocfs2_iterate_xattr_buckets(args->reflink->old_inode, blkno, len, | |
6748 | ocfs2_calc_value_tree_metas, &metas); | |
6749 | if (ret) { | |
6750 | mlog_errno(ret); | |
6751 | goto out; | |
6752 | } | |
6753 | ||
6754 | *credits = metas.credits; | |
6755 | ||
6756 | /* | |
6757 | * Calculate we need for refcount tree change. | |
6758 | * | |
6759 | * We need to add/modify num_recs in refcount tree, so just calculate | |
6760 | * an approximate number we need for refcount tree change. | |
6761 | * Sometimes we need to split the tree, and after split, half recs | |
6762 | * will be moved to the new block, and a new block can only provide | |
6763 | * half number of recs. So we multiple new blocks by 2. | |
6764 | * In the end, we have to add credits for modifying the already | |
6765 | * existed refcount block. | |
6766 | */ | |
6767 | rb = (struct ocfs2_refcount_block *)args->reflink->ref_root_bh->b_data; | |
6768 | metas.num_recs = | |
6769 | (metas.num_recs + ocfs2_refcount_recs_per_rb(osb->sb) - 1) / | |
6770 | ocfs2_refcount_recs_per_rb(osb->sb) * 2; | |
6771 | metas.num_metas += metas.num_recs; | |
6772 | *credits += metas.num_recs + | |
6773 | metas.num_recs * OCFS2_EXPAND_REFCOUNT_TREE_CREDITS; | |
6774 | if (le32_to_cpu(rb->rf_flags) & OCFS2_REFCOUNT_TREE_FL) | |
6775 | *credits += le16_to_cpu(rb->rf_list.l_tree_depth) * | |
6776 | le16_to_cpu(rb->rf_list.l_next_free_rec) + 1; | |
6777 | else | |
6778 | *credits += 1; | |
6779 | ||
6780 | /* count in the xattr tree change. */ | |
6781 | num_free_extents = ocfs2_num_free_extents(osb, xt_et); | |
6782 | if (num_free_extents < 0) { | |
6783 | ret = num_free_extents; | |
6784 | mlog_errno(ret); | |
6785 | goto out; | |
6786 | } | |
6787 | ||
6788 | if (num_free_extents < len) | |
6789 | metas.num_metas += ocfs2_extend_meta_needed(xt_et->et_root_el); | |
6790 | ||
6791 | *credits += ocfs2_calc_extend_credits(osb->sb, | |
06f9da6e | 6792 | xt_et->et_root_el); |
2999d12f TM |
6793 | |
6794 | if (metas.num_metas) { | |
6795 | ret = ocfs2_reserve_new_metadata_blocks(osb, metas.num_metas, | |
6796 | meta_ac); | |
6797 | if (ret) { | |
6798 | mlog_errno(ret); | |
6799 | goto out; | |
6800 | } | |
6801 | } | |
6802 | ||
6803 | if (len) { | |
6804 | ret = ocfs2_reserve_clusters(osb, len, data_ac); | |
6805 | if (ret) | |
6806 | mlog_errno(ret); | |
6807 | } | |
6808 | out: | |
6809 | if (ret) { | |
6810 | if (*meta_ac) { | |
6811 | ocfs2_free_alloc_context(*meta_ac); | |
6cae6d31 | 6812 | *meta_ac = NULL; |
2999d12f TM |
6813 | } |
6814 | } | |
6815 | ||
6816 | return ret; | |
6817 | } | |
6818 | ||
121a39bb | 6819 | static int ocfs2_reflink_xattr_bucket(handle_t *handle, |
2999d12f | 6820 | u64 blkno, u64 new_blkno, u32 clusters, |
121a39bb | 6821 | u32 *cpos, int num_buckets, |
2999d12f TM |
6822 | struct ocfs2_alloc_context *meta_ac, |
6823 | struct ocfs2_alloc_context *data_ac, | |
6824 | struct ocfs2_reflink_xattr_tree_args *args) | |
6825 | { | |
6826 | int i, j, ret = 0; | |
6827 | struct super_block *sb = args->reflink->old_inode->i_sb; | |
2999d12f TM |
6828 | int bpb = args->old_bucket->bu_blocks; |
6829 | struct ocfs2_xattr_value_buf vb = { | |
6830 | .vb_access = ocfs2_journal_access, | |
6831 | }; | |
6832 | ||
6833 | for (i = 0; i < num_buckets; i++, blkno += bpb, new_blkno += bpb) { | |
6834 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
6835 | if (ret) { | |
6836 | mlog_errno(ret); | |
6837 | break; | |
6838 | } | |
6839 | ||
6840 | ret = ocfs2_init_xattr_bucket(args->new_bucket, new_blkno); | |
6841 | if (ret) { | |
6842 | mlog_errno(ret); | |
6843 | break; | |
6844 | } | |
6845 | ||
2999d12f TM |
6846 | ret = ocfs2_xattr_bucket_journal_access(handle, |
6847 | args->new_bucket, | |
6848 | OCFS2_JOURNAL_ACCESS_CREATE); | |
6849 | if (ret) { | |
6850 | mlog_errno(ret); | |
6851 | break; | |
6852 | } | |
6853 | ||
6854 | for (j = 0; j < bpb; j++) | |
6855 | memcpy(bucket_block(args->new_bucket, j), | |
6856 | bucket_block(args->old_bucket, j), | |
6857 | sb->s_blocksize); | |
6858 | ||
121a39bb TM |
6859 | /* |
6860 | * Record the start cpos so that we can use it to initialize | |
6861 | * our xattr tree we also set the xh_num_bucket for the new | |
6862 | * bucket. | |
6863 | */ | |
6864 | if (i == 0) { | |
6865 | *cpos = le32_to_cpu(bucket_xh(args->new_bucket)-> | |
6866 | xh_entries[0].xe_name_hash); | |
6867 | bucket_xh(args->new_bucket)->xh_num_buckets = | |
6868 | cpu_to_le16(num_buckets); | |
6869 | } | |
6870 | ||
2999d12f TM |
6871 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); |
6872 | ||
6873 | ret = ocfs2_reflink_xattr_header(handle, args->reflink, | |
6874 | args->old_bucket->bu_bhs[0], | |
6875 | bucket_xh(args->old_bucket), | |
6876 | args->new_bucket->bu_bhs[0], | |
6877 | bucket_xh(args->new_bucket), | |
6878 | &vb, meta_ac, | |
6879 | ocfs2_get_reflink_xattr_value_root, | |
6880 | args); | |
6881 | if (ret) { | |
6882 | mlog_errno(ret); | |
6883 | break; | |
6884 | } | |
6885 | ||
6886 | /* | |
6887 | * Re-access and dirty the bucket to calculate metaecc. | |
6888 | * Because we may extend the transaction in reflink_xattr_header | |
6889 | * which will let the already accessed block gone. | |
6890 | */ | |
6891 | ret = ocfs2_xattr_bucket_journal_access(handle, | |
6892 | args->new_bucket, | |
6893 | OCFS2_JOURNAL_ACCESS_WRITE); | |
6894 | if (ret) { | |
6895 | mlog_errno(ret); | |
6896 | break; | |
6897 | } | |
6898 | ||
6899 | ocfs2_xattr_bucket_journal_dirty(handle, args->new_bucket); | |
121a39bb | 6900 | |
2999d12f TM |
6901 | ocfs2_xattr_bucket_relse(args->old_bucket); |
6902 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6903 | } | |
6904 | ||
6905 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6906 | ocfs2_xattr_bucket_relse(args->new_bucket); | |
6907 | return ret; | |
6908 | } | |
121a39bb TM |
6909 | |
6910 | static int ocfs2_reflink_xattr_buckets(handle_t *handle, | |
6911 | struct inode *inode, | |
6912 | struct ocfs2_reflink_xattr_tree_args *args, | |
6913 | struct ocfs2_extent_tree *et, | |
6914 | struct ocfs2_alloc_context *meta_ac, | |
6915 | struct ocfs2_alloc_context *data_ac, | |
6916 | u64 blkno, u32 cpos, u32 len) | |
6917 | { | |
6918 | int ret, first_inserted = 0; | |
6919 | u32 p_cluster, num_clusters, reflink_cpos = 0; | |
6920 | u64 new_blkno; | |
6921 | unsigned int num_buckets, reflink_buckets; | |
6922 | unsigned int bpc = | |
6923 | ocfs2_xattr_buckets_per_cluster(OCFS2_SB(inode->i_sb)); | |
6924 | ||
6925 | ret = ocfs2_read_xattr_bucket(args->old_bucket, blkno); | |
6926 | if (ret) { | |
6927 | mlog_errno(ret); | |
6928 | goto out; | |
6929 | } | |
6930 | num_buckets = le16_to_cpu(bucket_xh(args->old_bucket)->xh_num_buckets); | |
6931 | ocfs2_xattr_bucket_relse(args->old_bucket); | |
6932 | ||
6933 | while (len && num_buckets) { | |
6934 | ret = ocfs2_claim_clusters(handle, data_ac, | |
6935 | 1, &p_cluster, &num_clusters); | |
6936 | if (ret) { | |
6937 | mlog_errno(ret); | |
6938 | goto out; | |
6939 | } | |
6940 | ||
6941 | new_blkno = ocfs2_clusters_to_blocks(inode->i_sb, p_cluster); | |
6942 | reflink_buckets = min(num_buckets, bpc * num_clusters); | |
6943 | ||
6944 | ret = ocfs2_reflink_xattr_bucket(handle, blkno, | |
6945 | new_blkno, num_clusters, | |
6946 | &reflink_cpos, reflink_buckets, | |
6947 | meta_ac, data_ac, args); | |
6948 | if (ret) { | |
6949 | mlog_errno(ret); | |
6950 | goto out; | |
6951 | } | |
6952 | ||
6953 | /* | |
6954 | * For the 1st allocated cluster, we make it use the same cpos | |
6955 | * so that the xattr tree looks the same as the original one | |
6956 | * in the most case. | |
6957 | */ | |
6958 | if (!first_inserted) { | |
6959 | reflink_cpos = cpos; | |
6960 | first_inserted = 1; | |
6961 | } | |
6962 | ret = ocfs2_insert_extent(handle, et, reflink_cpos, new_blkno, | |
6963 | num_clusters, 0, meta_ac); | |
6964 | if (ret) | |
6965 | mlog_errno(ret); | |
6966 | ||
402b4183 TM |
6967 | trace_ocfs2_reflink_xattr_buckets((unsigned long long)new_blkno, |
6968 | num_clusters, reflink_cpos); | |
121a39bb TM |
6969 | |
6970 | len -= num_clusters; | |
6971 | blkno += ocfs2_clusters_to_blocks(inode->i_sb, num_clusters); | |
6972 | num_buckets -= reflink_buckets; | |
6973 | } | |
6974 | out: | |
6975 | return ret; | |
6976 | } | |
6977 | ||
2999d12f TM |
6978 | /* |
6979 | * Create the same xattr extent record in the new inode's xattr tree. | |
6980 | */ | |
6981 | static int ocfs2_reflink_xattr_rec(struct inode *inode, | |
6982 | struct buffer_head *root_bh, | |
6983 | u64 blkno, | |
6984 | u32 cpos, | |
6985 | u32 len, | |
6986 | void *para) | |
6987 | { | |
6988 | int ret, credits = 0; | |
2999d12f TM |
6989 | handle_t *handle; |
6990 | struct ocfs2_reflink_xattr_tree_args *args = | |
6991 | (struct ocfs2_reflink_xattr_tree_args *)para; | |
6992 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
6993 | struct ocfs2_alloc_context *meta_ac = NULL; | |
6994 | struct ocfs2_alloc_context *data_ac = NULL; | |
6995 | struct ocfs2_extent_tree et; | |
6996 | ||
402b4183 | 6997 | trace_ocfs2_reflink_xattr_rec((unsigned long long)blkno, len); |
121a39bb | 6998 | |
2999d12f TM |
6999 | ocfs2_init_xattr_tree_extent_tree(&et, |
7000 | INODE_CACHE(args->reflink->new_inode), | |
7001 | args->new_blk_bh); | |
7002 | ||
7003 | ret = ocfs2_lock_reflink_xattr_rec_allocators(args, &et, blkno, | |
7004 | len, &credits, | |
7005 | &meta_ac, &data_ac); | |
7006 | if (ret) { | |
7007 | mlog_errno(ret); | |
7008 | goto out; | |
7009 | } | |
7010 | ||
7011 | handle = ocfs2_start_trans(osb, credits); | |
7012 | if (IS_ERR(handle)) { | |
7013 | ret = PTR_ERR(handle); | |
7014 | mlog_errno(ret); | |
7015 | goto out; | |
7016 | } | |
7017 | ||
121a39bb TM |
7018 | ret = ocfs2_reflink_xattr_buckets(handle, inode, args, &et, |
7019 | meta_ac, data_ac, | |
7020 | blkno, cpos, len); | |
2999d12f TM |
7021 | if (ret) |
7022 | mlog_errno(ret); | |
7023 | ||
2999d12f TM |
7024 | ocfs2_commit_trans(osb, handle); |
7025 | ||
7026 | out: | |
7027 | if (meta_ac) | |
7028 | ocfs2_free_alloc_context(meta_ac); | |
7029 | if (data_ac) | |
7030 | ocfs2_free_alloc_context(data_ac); | |
7031 | return ret; | |
7032 | } | |
7033 | ||
7034 | /* | |
7035 | * Create reflinked xattr buckets. | |
7036 | * We will add bucket one by one, and refcount all the xattrs in the bucket | |
7037 | * if they are stored outside. | |
7038 | */ | |
7039 | static int ocfs2_reflink_xattr_tree(struct ocfs2_xattr_reflink *args, | |
7040 | struct buffer_head *blk_bh, | |
7041 | struct buffer_head *new_blk_bh) | |
7042 | { | |
7043 | int ret; | |
7044 | struct ocfs2_reflink_xattr_tree_args para; | |
7045 | ||
7046 | memset(¶, 0, sizeof(para)); | |
7047 | para.reflink = args; | |
7048 | para.old_blk_bh = blk_bh; | |
7049 | para.new_blk_bh = new_blk_bh; | |
7050 | ||
7051 | para.old_bucket = ocfs2_xattr_bucket_new(args->old_inode); | |
7052 | if (!para.old_bucket) { | |
7053 | mlog_errno(-ENOMEM); | |
7054 | return -ENOMEM; | |
7055 | } | |
7056 | ||
7057 | para.new_bucket = ocfs2_xattr_bucket_new(args->new_inode); | |
7058 | if (!para.new_bucket) { | |
7059 | ret = -ENOMEM; | |
7060 | mlog_errno(ret); | |
7061 | goto out; | |
7062 | } | |
7063 | ||
7064 | ret = ocfs2_iterate_xattr_index_block(args->old_inode, blk_bh, | |
7065 | ocfs2_reflink_xattr_rec, | |
7066 | ¶); | |
7067 | if (ret) | |
7068 | mlog_errno(ret); | |
7069 | ||
7070 | out: | |
7071 | ocfs2_xattr_bucket_free(para.old_bucket); | |
7072 | ocfs2_xattr_bucket_free(para.new_bucket); | |
7073 | return ret; | |
7074 | } | |
7075 | ||
7076 | static int ocfs2_reflink_xattr_in_block(struct ocfs2_xattr_reflink *args, | |
7077 | struct buffer_head *blk_bh) | |
7078 | { | |
7079 | int ret, indexed = 0; | |
7080 | struct buffer_head *new_blk_bh = NULL; | |
7081 | struct ocfs2_xattr_block *xb = | |
7082 | (struct ocfs2_xattr_block *)blk_bh->b_data; | |
7083 | ||
7084 | ||
7085 | if (le16_to_cpu(xb->xb_flags) & OCFS2_XATTR_INDEXED) | |
7086 | indexed = 1; | |
7087 | ||
7088 | ret = ocfs2_create_empty_xattr_block(args->new_inode, args->new_bh, | |
7089 | &new_blk_bh, indexed); | |
7090 | if (ret) { | |
7091 | mlog_errno(ret); | |
7092 | goto out; | |
7093 | } | |
7094 | ||
2decd65a | 7095 | if (!indexed) |
2999d12f TM |
7096 | ret = ocfs2_reflink_xattr_block(args, blk_bh, new_blk_bh); |
7097 | else | |
7098 | ret = ocfs2_reflink_xattr_tree(args, blk_bh, new_blk_bh); | |
7099 | if (ret) | |
7100 | mlog_errno(ret); | |
7101 | ||
7102 | out: | |
7103 | brelse(new_blk_bh); | |
7104 | return ret; | |
7105 | } | |
7106 | ||
0fe9b66c TM |
7107 | static int ocfs2_reflink_xattr_no_security(struct ocfs2_xattr_entry *xe) |
7108 | { | |
7109 | int type = ocfs2_xattr_get_type(xe); | |
7110 | ||
7111 | return type != OCFS2_XATTR_INDEX_SECURITY && | |
7112 | type != OCFS2_XATTR_INDEX_POSIX_ACL_ACCESS && | |
7113 | type != OCFS2_XATTR_INDEX_POSIX_ACL_DEFAULT; | |
7114 | } | |
7115 | ||
2999d12f TM |
7116 | int ocfs2_reflink_xattrs(struct inode *old_inode, |
7117 | struct buffer_head *old_bh, | |
7118 | struct inode *new_inode, | |
0fe9b66c TM |
7119 | struct buffer_head *new_bh, |
7120 | bool preserve_security) | |
2999d12f TM |
7121 | { |
7122 | int ret; | |
7123 | struct ocfs2_xattr_reflink args; | |
7124 | struct ocfs2_inode_info *oi = OCFS2_I(old_inode); | |
7125 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)old_bh->b_data; | |
7126 | struct buffer_head *blk_bh = NULL; | |
7127 | struct ocfs2_cached_dealloc_ctxt dealloc; | |
7128 | struct ocfs2_refcount_tree *ref_tree; | |
7129 | struct buffer_head *ref_root_bh = NULL; | |
7130 | ||
7131 | ret = ocfs2_lock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7132 | le64_to_cpu(di->i_refcount_loc), | |
7133 | 1, &ref_tree, &ref_root_bh); | |
7134 | if (ret) { | |
7135 | mlog_errno(ret); | |
7136 | goto out; | |
7137 | } | |
7138 | ||
7139 | ocfs2_init_dealloc_ctxt(&dealloc); | |
7140 | ||
7141 | args.old_inode = old_inode; | |
7142 | args.new_inode = new_inode; | |
7143 | args.old_bh = old_bh; | |
7144 | args.new_bh = new_bh; | |
7145 | args.ref_ci = &ref_tree->rf_ci; | |
7146 | args.ref_root_bh = ref_root_bh; | |
7147 | args.dealloc = &dealloc; | |
0fe9b66c TM |
7148 | if (preserve_security) |
7149 | args.xattr_reflinked = NULL; | |
7150 | else | |
7151 | args.xattr_reflinked = ocfs2_reflink_xattr_no_security; | |
2999d12f TM |
7152 | |
7153 | if (oi->ip_dyn_features & OCFS2_INLINE_XATTR_FL) { | |
7154 | ret = ocfs2_reflink_xattr_inline(&args); | |
7155 | if (ret) { | |
7156 | mlog_errno(ret); | |
7157 | goto out_unlock; | |
7158 | } | |
7159 | } | |
7160 | ||
7161 | if (!di->i_xattr_loc) | |
7162 | goto out_unlock; | |
7163 | ||
7164 | ret = ocfs2_read_xattr_block(old_inode, le64_to_cpu(di->i_xattr_loc), | |
7165 | &blk_bh); | |
7166 | if (ret < 0) { | |
7167 | mlog_errno(ret); | |
7168 | goto out_unlock; | |
7169 | } | |
7170 | ||
7171 | ret = ocfs2_reflink_xattr_in_block(&args, blk_bh); | |
7172 | if (ret) | |
7173 | mlog_errno(ret); | |
7174 | ||
7175 | brelse(blk_bh); | |
7176 | ||
7177 | out_unlock: | |
7178 | ocfs2_unlock_refcount_tree(OCFS2_SB(old_inode->i_sb), | |
7179 | ref_tree, 1); | |
7180 | brelse(ref_root_bh); | |
7181 | ||
7182 | if (ocfs2_dealloc_has_cluster(&dealloc)) { | |
7183 | ocfs2_schedule_truncate_log_flush(OCFS2_SB(old_inode->i_sb), 1); | |
7184 | ocfs2_run_deallocs(OCFS2_SB(old_inode->i_sb), &dealloc); | |
7185 | } | |
7186 | ||
7187 | out: | |
7188 | return ret; | |
7189 | } | |
7190 | ||
0fe9b66c TM |
7191 | /* |
7192 | * Initialize security and acl for a already created inode. | |
7193 | * Used for reflink a non-preserve-security file. | |
7194 | * | |
7195 | * It uses common api like ocfs2_xattr_set, so the caller | |
7196 | * must not hold any lock expect i_mutex. | |
7197 | */ | |
7198 | int ocfs2_init_security_and_acl(struct inode *dir, | |
2a7dba39 | 7199 | struct inode *inode, |
702e5bc6 CH |
7200 | const struct qstr *qstr, |
7201 | struct posix_acl *default_acl, | |
7202 | struct posix_acl *acl) | |
0fe9b66c | 7203 | { |
0fe9b66c | 7204 | struct buffer_head *dir_bh = NULL; |
702e5bc6 | 7205 | int ret = 0; |
0fe9b66c | 7206 | |
9d8f13ba | 7207 | ret = ocfs2_init_security_get(inode, dir, qstr, NULL); |
32918dd9 | 7208 | if (ret) { |
0fe9b66c TM |
7209 | mlog_errno(ret); |
7210 | goto leave; | |
7211 | } | |
7212 | ||
7213 | ret = ocfs2_inode_lock(dir, &dir_bh, 0); | |
7214 | if (ret) { | |
7215 | mlog_errno(ret); | |
7216 | goto leave; | |
7217 | } | |
7218 | ||
702e5bc6 CH |
7219 | if (!ret && default_acl) |
7220 | ret = ocfs2_iop_set_acl(inode, default_acl, ACL_TYPE_DEFAULT); | |
7221 | if (!ret && acl) | |
7222 | ret = ocfs2_iop_set_acl(inode, acl, ACL_TYPE_ACCESS); | |
0fe9b66c TM |
7223 | |
7224 | ocfs2_inode_unlock(dir, 0); | |
7225 | brelse(dir_bh); | |
7226 | leave: | |
7227 | return ret; | |
7228 | } | |
923f7f31 TY |
7229 | /* |
7230 | * 'security' attributes support | |
7231 | */ | |
431547b3 | 7232 | static size_t ocfs2_xattr_security_list(struct dentry *dentry, char *list, |
923f7f31 | 7233 | size_t list_size, const char *name, |
431547b3 | 7234 | size_t name_len, int type) |
923f7f31 TY |
7235 | { |
7236 | const size_t prefix_len = XATTR_SECURITY_PREFIX_LEN; | |
7237 | const size_t total_len = prefix_len + name_len + 1; | |
7238 | ||
7239 | if (list && total_len <= list_size) { | |
7240 | memcpy(list, XATTR_SECURITY_PREFIX, prefix_len); | |
7241 | memcpy(list + prefix_len, name, name_len); | |
7242 | list[prefix_len + name_len] = '\0'; | |
7243 | } | |
7244 | return total_len; | |
7245 | } | |
7246 | ||
431547b3 CH |
7247 | static int ocfs2_xattr_security_get(struct dentry *dentry, const char *name, |
7248 | void *buffer, size_t size, int type) | |
923f7f31 TY |
7249 | { |
7250 | if (strcmp(name, "") == 0) | |
7251 | return -EINVAL; | |
431547b3 CH |
7252 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7253 | name, buffer, size); | |
923f7f31 TY |
7254 | } |
7255 | ||
431547b3 CH |
7256 | static int ocfs2_xattr_security_set(struct dentry *dentry, const char *name, |
7257 | const void *value, size_t size, int flags, int type) | |
923f7f31 TY |
7258 | { |
7259 | if (strcmp(name, "") == 0) | |
7260 | return -EINVAL; | |
7261 | ||
431547b3 CH |
7262 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_SECURITY, |
7263 | name, value, size, flags); | |
923f7f31 TY |
7264 | } |
7265 | ||
9d8f13ba MZ |
7266 | int ocfs2_initxattrs(struct inode *inode, const struct xattr *xattr_array, |
7267 | void *fs_info) | |
7268 | { | |
7269 | const struct xattr *xattr; | |
7270 | int err = 0; | |
7271 | ||
7272 | for (xattr = xattr_array; xattr->name != NULL; xattr++) { | |
7273 | err = ocfs2_xattr_set(inode, OCFS2_XATTR_INDEX_SECURITY, | |
7274 | xattr->name, xattr->value, | |
7275 | xattr->value_len, XATTR_CREATE); | |
7276 | if (err) | |
7277 | break; | |
7278 | } | |
7279 | return err; | |
7280 | } | |
7281 | ||
534eaddd TY |
7282 | int ocfs2_init_security_get(struct inode *inode, |
7283 | struct inode *dir, | |
2a7dba39 | 7284 | const struct qstr *qstr, |
534eaddd TY |
7285 | struct ocfs2_security_xattr_info *si) |
7286 | { | |
38d59ef6 TY |
7287 | /* check whether ocfs2 support feature xattr */ |
7288 | if (!ocfs2_supports_xattr(OCFS2_SB(dir->i_sb))) | |
7289 | return -EOPNOTSUPP; | |
9d8f13ba MZ |
7290 | if (si) |
7291 | return security_old_inode_init_security(inode, dir, qstr, | |
7292 | &si->name, &si->value, | |
7293 | &si->value_len); | |
7294 | ||
7295 | return security_inode_init_security(inode, dir, qstr, | |
7296 | &ocfs2_initxattrs, NULL); | |
534eaddd TY |
7297 | } |
7298 | ||
7299 | int ocfs2_init_security_set(handle_t *handle, | |
7300 | struct inode *inode, | |
7301 | struct buffer_head *di_bh, | |
7302 | struct ocfs2_security_xattr_info *si, | |
7303 | struct ocfs2_alloc_context *xattr_ac, | |
7304 | struct ocfs2_alloc_context *data_ac) | |
7305 | { | |
7306 | return ocfs2_xattr_set_handle(handle, inode, di_bh, | |
7307 | OCFS2_XATTR_INDEX_SECURITY, | |
7308 | si->name, si->value, si->value_len, 0, | |
7309 | xattr_ac, data_ac); | |
7310 | } | |
7311 | ||
537d81ca | 7312 | const struct xattr_handler ocfs2_xattr_security_handler = { |
923f7f31 TY |
7313 | .prefix = XATTR_SECURITY_PREFIX, |
7314 | .list = ocfs2_xattr_security_list, | |
7315 | .get = ocfs2_xattr_security_get, | |
7316 | .set = ocfs2_xattr_security_set, | |
7317 | }; | |
7318 | ||
99219aea MF |
7319 | /* |
7320 | * 'trusted' attributes support | |
7321 | */ | |
431547b3 | 7322 | static size_t ocfs2_xattr_trusted_list(struct dentry *dentry, char *list, |
99219aea | 7323 | size_t list_size, const char *name, |
431547b3 | 7324 | size_t name_len, int type) |
99219aea | 7325 | { |
ceb1eba3 | 7326 | const size_t prefix_len = XATTR_TRUSTED_PREFIX_LEN; |
99219aea MF |
7327 | const size_t total_len = prefix_len + name_len + 1; |
7328 | ||
7329 | if (list && total_len <= list_size) { | |
7330 | memcpy(list, XATTR_TRUSTED_PREFIX, prefix_len); | |
7331 | memcpy(list + prefix_len, name, name_len); | |
7332 | list[prefix_len + name_len] = '\0'; | |
7333 | } | |
7334 | return total_len; | |
7335 | } | |
7336 | ||
431547b3 CH |
7337 | static int ocfs2_xattr_trusted_get(struct dentry *dentry, const char *name, |
7338 | void *buffer, size_t size, int type) | |
99219aea MF |
7339 | { |
7340 | if (strcmp(name, "") == 0) | |
7341 | return -EINVAL; | |
431547b3 CH |
7342 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7343 | name, buffer, size); | |
99219aea MF |
7344 | } |
7345 | ||
431547b3 CH |
7346 | static int ocfs2_xattr_trusted_set(struct dentry *dentry, const char *name, |
7347 | const void *value, size_t size, int flags, int type) | |
99219aea MF |
7348 | { |
7349 | if (strcmp(name, "") == 0) | |
7350 | return -EINVAL; | |
7351 | ||
431547b3 CH |
7352 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_TRUSTED, |
7353 | name, value, size, flags); | |
99219aea MF |
7354 | } |
7355 | ||
537d81ca | 7356 | const struct xattr_handler ocfs2_xattr_trusted_handler = { |
99219aea MF |
7357 | .prefix = XATTR_TRUSTED_PREFIX, |
7358 | .list = ocfs2_xattr_trusted_list, | |
7359 | .get = ocfs2_xattr_trusted_get, | |
7360 | .set = ocfs2_xattr_trusted_set, | |
7361 | }; | |
7362 | ||
99219aea MF |
7363 | /* |
7364 | * 'user' attributes support | |
7365 | */ | |
431547b3 | 7366 | static size_t ocfs2_xattr_user_list(struct dentry *dentry, char *list, |
99219aea | 7367 | size_t list_size, const char *name, |
431547b3 | 7368 | size_t name_len, int type) |
99219aea | 7369 | { |
ceb1eba3 | 7370 | const size_t prefix_len = XATTR_USER_PREFIX_LEN; |
99219aea | 7371 | const size_t total_len = prefix_len + name_len + 1; |
431547b3 | 7372 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7373 | |
7374 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7375 | return 0; | |
7376 | ||
7377 | if (list && total_len <= list_size) { | |
7378 | memcpy(list, XATTR_USER_PREFIX, prefix_len); | |
7379 | memcpy(list + prefix_len, name, name_len); | |
7380 | list[prefix_len + name_len] = '\0'; | |
7381 | } | |
7382 | return total_len; | |
7383 | } | |
7384 | ||
431547b3 CH |
7385 | static int ocfs2_xattr_user_get(struct dentry *dentry, const char *name, |
7386 | void *buffer, size_t size, int type) | |
99219aea | 7387 | { |
431547b3 | 7388 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7389 | |
7390 | if (strcmp(name, "") == 0) | |
7391 | return -EINVAL; | |
7392 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7393 | return -EOPNOTSUPP; | |
431547b3 | 7394 | return ocfs2_xattr_get(dentry->d_inode, OCFS2_XATTR_INDEX_USER, name, |
99219aea MF |
7395 | buffer, size); |
7396 | } | |
7397 | ||
431547b3 CH |
7398 | static int ocfs2_xattr_user_set(struct dentry *dentry, const char *name, |
7399 | const void *value, size_t size, int flags, int type) | |
99219aea | 7400 | { |
431547b3 | 7401 | struct ocfs2_super *osb = OCFS2_SB(dentry->d_sb); |
99219aea MF |
7402 | |
7403 | if (strcmp(name, "") == 0) | |
7404 | return -EINVAL; | |
7405 | if (osb->s_mount_opt & OCFS2_MOUNT_NOUSERXATTR) | |
7406 | return -EOPNOTSUPP; | |
7407 | ||
431547b3 CH |
7408 | return ocfs2_xattr_set(dentry->d_inode, OCFS2_XATTR_INDEX_USER, |
7409 | name, value, size, flags); | |
99219aea MF |
7410 | } |
7411 | ||
537d81ca | 7412 | const struct xattr_handler ocfs2_xattr_user_handler = { |
99219aea MF |
7413 | .prefix = XATTR_USER_PREFIX, |
7414 | .list = ocfs2_xattr_user_list, | |
7415 | .get = ocfs2_xattr_user_get, | |
7416 | .set = ocfs2_xattr_user_set, | |
7417 | }; |