Commit | Line | Data |
---|---|---|
ccd979bd MF |
1 | /* -*- mode: c; c-basic-offset: 8; -*- |
2 | * vim: noexpandtab sw=8 ts=8 sts=0: | |
3 | * | |
4 | * inode.c | |
5 | * | |
6 | * vfs' aops, fops, dops and iops | |
7 | * | |
8 | * Copyright (C) 2002, 2004 Oracle. All rights reserved. | |
9 | * | |
10 | * This program is free software; you can redistribute it and/or | |
11 | * modify it under the terms of the GNU General Public | |
12 | * License as published by the Free Software Foundation; either | |
13 | * version 2 of the License, or (at your option) any later version. | |
14 | * | |
15 | * This program is distributed in the hope that it will be useful, | |
16 | * but WITHOUT ANY WARRANTY; without even the implied warranty of | |
17 | * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU | |
18 | * General Public License for more details. | |
19 | * | |
20 | * You should have received a copy of the GNU General Public | |
21 | * License along with this program; if not, write to the | |
22 | * Free Software Foundation, Inc., 59 Temple Place - Suite 330, | |
23 | * Boston, MA 021110-1307, USA. | |
24 | */ | |
25 | ||
26 | #include <linux/fs.h> | |
27 | #include <linux/types.h> | |
ccd979bd MF |
28 | #include <linux/highmem.h> |
29 | #include <linux/pagemap.h> | |
a90714c1 | 30 | #include <linux/quotaops.h> |
ccd979bd MF |
31 | |
32 | #include <asm/byteorder.h> | |
33 | ||
ccd979bd MF |
34 | #include <cluster/masklog.h> |
35 | ||
36 | #include "ocfs2.h" | |
37 | ||
38 | #include "alloc.h" | |
9b7895ef | 39 | #include "dir.h" |
d6b32bbb | 40 | #include "blockcheck.h" |
ccd979bd MF |
41 | #include "dlmglue.h" |
42 | #include "extent_map.h" | |
43 | #include "file.h" | |
b4df6ed8 | 44 | #include "heartbeat.h" |
ccd979bd MF |
45 | #include "inode.h" |
46 | #include "journal.h" | |
47 | #include "namei.h" | |
48 | #include "suballoc.h" | |
49 | #include "super.h" | |
50 | #include "symlink.h" | |
51 | #include "sysfile.h" | |
52 | #include "uptodate.h" | |
cf1d6c76 | 53 | #include "xattr.h" |
8b2c0dba | 54 | #include "refcounttree.h" |
64f3b269 | 55 | #include "ocfs2_trace.h" |
ccd979bd MF |
56 | |
57 | #include "buffer_head_io.h" | |
58 | ||
ccd979bd MF |
59 | struct ocfs2_find_inode_args |
60 | { | |
61 | u64 fi_blkno; | |
62 | unsigned long fi_ino; | |
63 | unsigned int fi_flags; | |
5fa0613e | 64 | unsigned int fi_sysfile_type; |
ccd979bd MF |
65 | }; |
66 | ||
5fa0613e JK |
67 | static struct lock_class_key ocfs2_sysfile_lock_key[NUM_SYSTEM_INODES]; |
68 | ||
ccd979bd MF |
69 | static int ocfs2_read_locked_inode(struct inode *inode, |
70 | struct ocfs2_find_inode_args *args); | |
71 | static int ocfs2_init_locked_inode(struct inode *inode, void *opaque); | |
72 | static int ocfs2_find_actor(struct inode *inode, void *opaque); | |
73 | static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, | |
74 | struct inode *inode, | |
75 | struct buffer_head *fe_bh); | |
76 | ||
ca4d147e HP |
77 | void ocfs2_set_inode_flags(struct inode *inode) |
78 | { | |
79 | unsigned int flags = OCFS2_I(inode)->ip_attr; | |
80 | ||
81 | inode->i_flags &= ~(S_IMMUTABLE | | |
82 | S_SYNC | S_APPEND | S_NOATIME | S_DIRSYNC); | |
83 | ||
84 | if (flags & OCFS2_IMMUTABLE_FL) | |
85 | inode->i_flags |= S_IMMUTABLE; | |
86 | ||
87 | if (flags & OCFS2_SYNC_FL) | |
88 | inode->i_flags |= S_SYNC; | |
89 | if (flags & OCFS2_APPEND_FL) | |
90 | inode->i_flags |= S_APPEND; | |
91 | if (flags & OCFS2_NOATIME_FL) | |
92 | inode->i_flags |= S_NOATIME; | |
93 | if (flags & OCFS2_DIRSYNC_FL) | |
94 | inode->i_flags |= S_DIRSYNC; | |
95 | } | |
96 | ||
6e4b0d56 JK |
97 | /* Propagate flags from i_flags to OCFS2_I(inode)->ip_attr */ |
98 | void ocfs2_get_inode_flags(struct ocfs2_inode_info *oi) | |
99 | { | |
100 | unsigned int flags = oi->vfs_inode.i_flags; | |
101 | ||
102 | oi->ip_attr &= ~(OCFS2_SYNC_FL|OCFS2_APPEND_FL| | |
103 | OCFS2_IMMUTABLE_FL|OCFS2_NOATIME_FL|OCFS2_DIRSYNC_FL); | |
104 | if (flags & S_SYNC) | |
105 | oi->ip_attr |= OCFS2_SYNC_FL; | |
106 | if (flags & S_APPEND) | |
107 | oi->ip_attr |= OCFS2_APPEND_FL; | |
108 | if (flags & S_IMMUTABLE) | |
109 | oi->ip_attr |= OCFS2_IMMUTABLE_FL; | |
110 | if (flags & S_NOATIME) | |
111 | oi->ip_attr |= OCFS2_NOATIME_FL; | |
112 | if (flags & S_DIRSYNC) | |
113 | oi->ip_attr |= OCFS2_DIRSYNC_FL; | |
114 | } | |
115 | ||
6ca497a8 | 116 | struct inode *ocfs2_ilookup(struct super_block *sb, u64 blkno) |
117 | { | |
118 | struct ocfs2_find_inode_args args; | |
119 | ||
120 | args.fi_blkno = blkno; | |
121 | args.fi_flags = 0; | |
122 | args.fi_ino = ino_from_blkno(sb, blkno); | |
123 | args.fi_sysfile_type = 0; | |
124 | ||
125 | return ilookup5(sb, blkno, ocfs2_find_actor, &args); | |
126 | } | |
5fa0613e JK |
127 | struct inode *ocfs2_iget(struct ocfs2_super *osb, u64 blkno, unsigned flags, |
128 | int sysfile_type) | |
ccd979bd MF |
129 | { |
130 | struct inode *inode = NULL; | |
131 | struct super_block *sb = osb->sb; | |
132 | struct ocfs2_find_inode_args args; | |
133 | ||
64f3b269 TM |
134 | trace_ocfs2_iget_begin((unsigned long long)blkno, flags, |
135 | sysfile_type); | |
ccd979bd MF |
136 | |
137 | /* Ok. By now we've either got the offsets passed to us by the | |
138 | * caller, or we just pulled them off the bh. Lets do some | |
139 | * sanity checks to make sure they're OK. */ | |
140 | if (blkno == 0) { | |
141 | inode = ERR_PTR(-EINVAL); | |
142 | mlog_errno(PTR_ERR(inode)); | |
143 | goto bail; | |
144 | } | |
145 | ||
146 | args.fi_blkno = blkno; | |
24c19ef4 | 147 | args.fi_flags = flags; |
ccd979bd | 148 | args.fi_ino = ino_from_blkno(sb, blkno); |
5fa0613e | 149 | args.fi_sysfile_type = sysfile_type; |
ccd979bd MF |
150 | |
151 | inode = iget5_locked(sb, args.fi_ino, ocfs2_find_actor, | |
152 | ocfs2_init_locked_inode, &args); | |
153 | /* inode was *not* in the inode cache. 2.6.x requires | |
154 | * us to do our own read_inode call and unlock it | |
155 | * afterwards. */ | |
ccd979bd MF |
156 | if (inode == NULL) { |
157 | inode = ERR_PTR(-ENOMEM); | |
158 | mlog_errno(PTR_ERR(inode)); | |
159 | goto bail; | |
160 | } | |
64f3b269 | 161 | trace_ocfs2_iget5_locked(inode->i_state); |
6218b90e | 162 | if (inode->i_state & I_NEW) { |
6218b90e TM |
163 | ocfs2_read_locked_inode(inode, &args); |
164 | unlock_new_inode(inode); | |
165 | } | |
ccd979bd MF |
166 | if (is_bad_inode(inode)) { |
167 | iput(inode); | |
168 | inode = ERR_PTR(-ESTALE); | |
ccd979bd MF |
169 | goto bail; |
170 | } | |
171 | ||
172 | bail: | |
173 | if (!IS_ERR(inode)) { | |
64f3b269 TM |
174 | trace_ocfs2_iget_end(inode, |
175 | (unsigned long long)OCFS2_I(inode)->ip_blkno); | |
6a1bd4a5 | 176 | } |
ccd979bd MF |
177 | |
178 | return inode; | |
179 | } | |
180 | ||
181 | ||
182 | /* | |
183 | * here's how inodes get read from disk: | |
184 | * iget5_locked -> find_actor -> OCFS2_FIND_ACTOR | |
185 | * found? : return the in-memory inode | |
186 | * not found? : get_new_inode -> OCFS2_INIT_LOCKED_INODE | |
187 | */ | |
188 | ||
189 | static int ocfs2_find_actor(struct inode *inode, void *opaque) | |
190 | { | |
191 | struct ocfs2_find_inode_args *args = NULL; | |
192 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
193 | int ret = 0; | |
194 | ||
ccd979bd MF |
195 | args = opaque; |
196 | ||
197 | mlog_bug_on_msg(!inode, "No inode in find actor!\n"); | |
198 | ||
64f3b269 TM |
199 | trace_ocfs2_find_actor(inode, inode->i_ino, opaque, args->fi_blkno); |
200 | ||
ccd979bd MF |
201 | if (oi->ip_blkno != args->fi_blkno) |
202 | goto bail; | |
203 | ||
ccd979bd MF |
204 | ret = 1; |
205 | bail: | |
ccd979bd MF |
206 | return ret; |
207 | } | |
208 | ||
209 | /* | |
210 | * initialize the new inode, but don't do anything that would cause | |
211 | * us to sleep. | |
212 | * return 0 on success, 1 on failure | |
213 | */ | |
214 | static int ocfs2_init_locked_inode(struct inode *inode, void *opaque) | |
215 | { | |
216 | struct ocfs2_find_inode_args *args = opaque; | |
cb25797d JK |
217 | static struct lock_class_key ocfs2_quota_ip_alloc_sem_key, |
218 | ocfs2_file_ip_alloc_sem_key; | |
ccd979bd | 219 | |
ccd979bd MF |
220 | inode->i_ino = args->fi_ino; |
221 | OCFS2_I(inode)->ip_blkno = args->fi_blkno; | |
5fa0613e JK |
222 | if (args->fi_sysfile_type != 0) |
223 | lockdep_set_class(&inode->i_mutex, | |
224 | &ocfs2_sysfile_lock_key[args->fi_sysfile_type]); | |
cb25797d JK |
225 | if (args->fi_sysfile_type == USER_QUOTA_SYSTEM_INODE || |
226 | args->fi_sysfile_type == GROUP_QUOTA_SYSTEM_INODE || | |
227 | args->fi_sysfile_type == LOCAL_USER_QUOTA_SYSTEM_INODE || | |
228 | args->fi_sysfile_type == LOCAL_GROUP_QUOTA_SYSTEM_INODE) | |
229 | lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem, | |
230 | &ocfs2_quota_ip_alloc_sem_key); | |
231 | else | |
232 | lockdep_set_class(&OCFS2_I(inode)->ip_alloc_sem, | |
233 | &ocfs2_file_ip_alloc_sem_key); | |
ccd979bd | 234 | |
ccd979bd MF |
235 | return 0; |
236 | } | |
237 | ||
b657c95c JB |
238 | void ocfs2_populate_inode(struct inode *inode, struct ocfs2_dinode *fe, |
239 | int create_ino) | |
ccd979bd MF |
240 | { |
241 | struct super_block *sb; | |
242 | struct ocfs2_super *osb; | |
53da4939 | 243 | int use_plocks = 1; |
ccd979bd | 244 | |
ccd979bd MF |
245 | sb = inode->i_sb; |
246 | osb = OCFS2_SB(sb); | |
247 | ||
53da4939 MF |
248 | if ((osb->s_mount_opt & OCFS2_MOUNT_LOCALFLOCKS) || |
249 | ocfs2_mount_local(osb) || !ocfs2_stack_supports_plocks()) | |
250 | use_plocks = 0; | |
251 | ||
b657c95c JB |
252 | /* |
253 | * These have all been checked by ocfs2_read_inode_block() or set | |
254 | * by ocfs2_mknod_locked(), so a failure is a code bug. | |
255 | */ | |
256 | BUG_ON(!OCFS2_IS_VALID_DINODE(fe)); /* This means that read_inode | |
257 | cannot create a superblock | |
258 | inode today. change if | |
259 | that is needed. */ | |
260 | BUG_ON(!(fe->i_flags & cpu_to_le32(OCFS2_VALID_FL))); | |
261 | BUG_ON(le32_to_cpu(fe->i_fs_generation) != osb->fs_generation); | |
ccd979bd | 262 | |
ccd979bd | 263 | |
8110b073 MF |
264 | OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); |
265 | OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr); | |
15b1e36b | 266 | OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features); |
8110b073 | 267 | |
ccd979bd MF |
268 | inode->i_version = 1; |
269 | inode->i_generation = le32_to_cpu(fe->i_generation); | |
270 | inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev)); | |
271 | inode->i_mode = le16_to_cpu(fe->i_mode); | |
2c034176 EB |
272 | i_uid_write(inode, le32_to_cpu(fe->i_uid)); |
273 | i_gid_write(inode, le32_to_cpu(fe->i_gid)); | |
ccd979bd MF |
274 | |
275 | /* Fast symlinks will have i_size but no allocated clusters. */ | |
ea022dfb | 276 | if (S_ISLNK(inode->i_mode) && !fe->i_clusters) { |
ccd979bd | 277 | inode->i_blocks = 0; |
ea022dfb AV |
278 | inode->i_mapping->a_ops = &ocfs2_fast_symlink_aops; |
279 | } else { | |
8110b073 | 280 | inode->i_blocks = ocfs2_inode_sector_count(inode); |
ea022dfb AV |
281 | inode->i_mapping->a_ops = &ocfs2_aops; |
282 | } | |
ccd979bd MF |
283 | inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime); |
284 | inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec); | |
285 | inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime); | |
286 | inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec); | |
287 | inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime); | |
288 | inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec); | |
289 | ||
290 | if (OCFS2_I(inode)->ip_blkno != le64_to_cpu(fe->i_blkno)) | |
291 | mlog(ML_ERROR, | |
b0697053 MF |
292 | "ip_blkno %llu != i_blkno %llu!\n", |
293 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
1ca1a111 | 294 | (unsigned long long)le64_to_cpu(fe->i_blkno)); |
ccd979bd | 295 | |
bfe86848 | 296 | set_nlink(inode, ocfs2_read_links_count(fe)); |
ccd979bd | 297 | |
64f3b269 TM |
298 | trace_ocfs2_populate_inode(OCFS2_I(inode)->ip_blkno, |
299 | le32_to_cpu(fe->i_flags)); | |
bbbd0eb3 | 300 | if (fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) { |
24c19ef4 | 301 | OCFS2_I(inode)->ip_flags |= OCFS2_INODE_SYSTEM_FILE; |
bbbd0eb3 JK |
302 | inode->i_flags |= S_NOQUOTA; |
303 | } | |
64f3b269 | 304 | |
ccd979bd MF |
305 | if (fe->i_flags & cpu_to_le32(OCFS2_LOCAL_ALLOC_FL)) { |
306 | OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP; | |
ccd979bd MF |
307 | } else if (fe->i_flags & cpu_to_le32(OCFS2_BITMAP_FL)) { |
308 | OCFS2_I(inode)->ip_flags |= OCFS2_INODE_BITMAP; | |
1a224ad1 JK |
309 | } else if (fe->i_flags & cpu_to_le32(OCFS2_QUOTA_FL)) { |
310 | inode->i_flags |= S_NOQUOTA; | |
ccd979bd | 311 | } else if (fe->i_flags & cpu_to_le32(OCFS2_SUPER_BLOCK_FL)) { |
ccd979bd MF |
312 | /* we can't actually hit this as read_inode can't |
313 | * handle superblocks today ;-) */ | |
314 | BUG(); | |
315 | } | |
316 | ||
317 | switch (inode->i_mode & S_IFMT) { | |
318 | case S_IFREG: | |
53da4939 MF |
319 | if (use_plocks) |
320 | inode->i_fop = &ocfs2_fops; | |
321 | else | |
322 | inode->i_fop = &ocfs2_fops_no_plocks; | |
ccd979bd MF |
323 | inode->i_op = &ocfs2_file_iops; |
324 | i_size_write(inode, le64_to_cpu(fe->i_size)); | |
325 | break; | |
326 | case S_IFDIR: | |
327 | inode->i_op = &ocfs2_dir_iops; | |
53da4939 MF |
328 | if (use_plocks) |
329 | inode->i_fop = &ocfs2_dops; | |
330 | else | |
331 | inode->i_fop = &ocfs2_dops_no_plocks; | |
ccd979bd | 332 | i_size_write(inode, le64_to_cpu(fe->i_size)); |
5e98d492 | 333 | OCFS2_I(inode)->ip_dir_lock_gen = 1; |
ccd979bd MF |
334 | break; |
335 | case S_IFLNK: | |
ea022dfb | 336 | inode->i_op = &ocfs2_symlink_inode_operations; |
ccd979bd MF |
337 | i_size_write(inode, le64_to_cpu(fe->i_size)); |
338 | break; | |
339 | default: | |
340 | inode->i_op = &ocfs2_special_file_iops; | |
341 | init_special_inode(inode, inode->i_mode, | |
342 | inode->i_rdev); | |
343 | break; | |
344 | } | |
345 | ||
24c19ef4 MF |
346 | if (create_ino) { |
347 | inode->i_ino = ino_from_blkno(inode->i_sb, | |
348 | le64_to_cpu(fe->i_blkno)); | |
349 | ||
350 | /* | |
351 | * If we ever want to create system files from kernel, | |
352 | * the generation argument to | |
353 | * ocfs2_inode_lock_res_init() will have to change. | |
354 | */ | |
1ca1a111 | 355 | BUG_ON(le32_to_cpu(fe->i_flags) & OCFS2_SYSTEM_FL); |
24c19ef4 | 356 | |
e63aecb6 | 357 | ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres, |
24c19ef4 | 358 | OCFS2_LOCK_TYPE_META, 0, inode); |
50008630 TY |
359 | |
360 | ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres, | |
361 | OCFS2_LOCK_TYPE_OPEN, 0, inode); | |
24c19ef4 MF |
362 | } |
363 | ||
ccd979bd | 364 | ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_rw_lockres, |
24c19ef4 MF |
365 | OCFS2_LOCK_TYPE_RW, inode->i_generation, |
366 | inode); | |
367 | ||
ca4d147e | 368 | ocfs2_set_inode_flags(inode); |
ca4d147e | 369 | |
13821151 TM |
370 | OCFS2_I(inode)->ip_last_used_slot = 0; |
371 | OCFS2_I(inode)->ip_last_used_group = 0; | |
e3b4a97d MF |
372 | |
373 | if (S_ISDIR(inode->i_mode)) | |
374 | ocfs2_resv_set_type(&OCFS2_I(inode)->ip_la_data_resv, | |
375 | OCFS2_RESV_FLAG_DIR); | |
ccd979bd MF |
376 | } |
377 | ||
378 | static int ocfs2_read_locked_inode(struct inode *inode, | |
379 | struct ocfs2_find_inode_args *args) | |
380 | { | |
381 | struct super_block *sb; | |
382 | struct ocfs2_super *osb; | |
383 | struct ocfs2_dinode *fe; | |
384 | struct buffer_head *bh = NULL; | |
24c19ef4 MF |
385 | int status, can_lock; |
386 | u32 generation = 0; | |
ccd979bd | 387 | |
ccd979bd | 388 | status = -EINVAL; |
ccd979bd MF |
389 | sb = inode->i_sb; |
390 | osb = OCFS2_SB(sb); | |
391 | ||
24c19ef4 MF |
392 | /* |
393 | * To improve performance of cold-cache inode stats, we take | |
394 | * the cluster lock here if possible. | |
395 | * | |
396 | * Generally, OCFS2 never trusts the contents of an inode | |
397 | * unless it's holding a cluster lock, so taking it here isn't | |
398 | * a correctness issue as much as it is a performance | |
399 | * improvement. | |
400 | * | |
401 | * There are three times when taking the lock is not a good idea: | |
402 | * | |
403 | * 1) During startup, before we have initialized the DLM. | |
404 | * | |
405 | * 2) If we are reading certain system files which never get | |
406 | * cluster locks (local alloc, truncate log). | |
407 | * | |
408 | * 3) If the process doing the iget() is responsible for | |
409 | * orphan dir recovery. We're holding the orphan dir lock and | |
410 | * can get into a deadlock with another process on another | |
411 | * node in ->delete_inode(). | |
412 | * | |
413 | * #1 and #2 can be simply solved by never taking the lock | |
414 | * here for system files (which are the only type we read | |
415 | * during mount). It's a heavier approach, but our main | |
b595076a | 416 | * concern is user-accessible files anyway. |
24c19ef4 MF |
417 | * |
418 | * #3 works itself out because we'll eventually take the | |
419 | * cluster lock before trusting anything anyway. | |
420 | */ | |
421 | can_lock = !(args->fi_flags & OCFS2_FI_FLAG_SYSFILE) | |
50008630 | 422 | && !(args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) |
c271c5c2 | 423 | && !ocfs2_mount_local(osb); |
24c19ef4 | 424 | |
64f3b269 TM |
425 | trace_ocfs2_read_locked_inode( |
426 | (unsigned long long)OCFS2_I(inode)->ip_blkno, can_lock); | |
427 | ||
24c19ef4 MF |
428 | /* |
429 | * To maintain backwards compatibility with older versions of | |
430 | * ocfs2-tools, we still store the generation value for system | |
431 | * files. The only ones that actually matter to userspace are | |
432 | * the journals, but it's easier and inexpensive to just flag | |
433 | * all system files similarly. | |
434 | */ | |
435 | if (args->fi_flags & OCFS2_FI_FLAG_SYSFILE) | |
436 | generation = osb->fs_generation; | |
437 | ||
e63aecb6 | 438 | ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_inode_lockres, |
24c19ef4 MF |
439 | OCFS2_LOCK_TYPE_META, |
440 | generation, inode); | |
441 | ||
50008630 TY |
442 | ocfs2_inode_lock_res_init(&OCFS2_I(inode)->ip_open_lockres, |
443 | OCFS2_LOCK_TYPE_OPEN, | |
444 | 0, inode); | |
445 | ||
24c19ef4 | 446 | if (can_lock) { |
50008630 TY |
447 | status = ocfs2_open_lock(inode); |
448 | if (status) { | |
449 | make_bad_inode(inode); | |
450 | mlog_errno(status); | |
451 | return status; | |
452 | } | |
e63aecb6 | 453 | status = ocfs2_inode_lock(inode, NULL, 0); |
24c19ef4 MF |
454 | if (status) { |
455 | make_bad_inode(inode); | |
456 | mlog_errno(status); | |
457 | return status; | |
458 | } | |
ccd979bd MF |
459 | } |
460 | ||
50008630 TY |
461 | if (args->fi_flags & OCFS2_FI_FLAG_ORPHAN_RECOVERY) { |
462 | status = ocfs2_try_open_lock(inode, 0); | |
463 | if (status) { | |
2bd63216 | 464 | make_bad_inode(inode); |
50008630 TY |
465 | return status; |
466 | } | |
467 | } | |
468 | ||
b657c95c JB |
469 | if (can_lock) { |
470 | status = ocfs2_read_inode_block_full(inode, &bh, | |
471 | OCFS2_BH_IGNORE_CACHE); | |
472 | } else { | |
da1e9098 | 473 | status = ocfs2_read_blocks_sync(osb, args->fi_blkno, 1, &bh); |
f5ce5a08 SM |
474 | /* |
475 | * If buffer is in jbd, then its checksum may not have been | |
476 | * computed as yet. | |
477 | */ | |
478 | if (!status && !buffer_jbd(bh)) | |
b657c95c JB |
479 | status = ocfs2_validate_inode_block(osb->sb, bh); |
480 | } | |
ccd979bd MF |
481 | if (status < 0) { |
482 | mlog_errno(status); | |
ccd979bd MF |
483 | goto bail; |
484 | } | |
485 | ||
24c19ef4 | 486 | status = -EINVAL; |
ccd979bd | 487 | fe = (struct ocfs2_dinode *) bh->b_data; |
ccd979bd | 488 | |
24c19ef4 MF |
489 | /* |
490 | * This is a code bug. Right now the caller needs to | |
491 | * understand whether it is asking for a system file inode or | |
492 | * not so the proper lock names can be built. | |
493 | */ | |
494 | mlog_bug_on_msg(!!(fe->i_flags & cpu_to_le32(OCFS2_SYSTEM_FL)) != | |
495 | !!(args->fi_flags & OCFS2_FI_FLAG_SYSFILE), | |
496 | "Inode %llu: system file state is ambigous\n", | |
497 | (unsigned long long)args->fi_blkno); | |
ccd979bd MF |
498 | |
499 | if (S_ISCHR(le16_to_cpu(fe->i_mode)) || | |
500 | S_ISBLK(le16_to_cpu(fe->i_mode))) | |
b657c95c | 501 | inode->i_rdev = huge_decode_dev(le64_to_cpu(fe->id1.dev1.i_rdev)); |
ccd979bd | 502 | |
b657c95c | 503 | ocfs2_populate_inode(inode, fe, 0); |
ccd979bd MF |
504 | |
505 | BUG_ON(args->fi_blkno != le64_to_cpu(fe->i_blkno)); | |
506 | ||
ccd979bd MF |
507 | status = 0; |
508 | ||
509 | bail: | |
24c19ef4 | 510 | if (can_lock) |
e63aecb6 | 511 | ocfs2_inode_unlock(inode, 0); |
24c19ef4 MF |
512 | |
513 | if (status < 0) | |
514 | make_bad_inode(inode); | |
515 | ||
ccd979bd MF |
516 | if (args && bh) |
517 | brelse(bh); | |
518 | ||
ccd979bd MF |
519 | return status; |
520 | } | |
521 | ||
522 | void ocfs2_sync_blockdev(struct super_block *sb) | |
523 | { | |
524 | sync_blockdev(sb->s_bdev); | |
525 | } | |
526 | ||
527 | static int ocfs2_truncate_for_delete(struct ocfs2_super *osb, | |
528 | struct inode *inode, | |
529 | struct buffer_head *fe_bh) | |
530 | { | |
531 | int status = 0; | |
ccd979bd | 532 | struct ocfs2_dinode *fe; |
60b11392 | 533 | handle_t *handle = NULL; |
ccd979bd | 534 | |
ccd979bd MF |
535 | fe = (struct ocfs2_dinode *) fe_bh->b_data; |
536 | ||
1afc32b9 MF |
537 | /* |
538 | * This check will also skip truncate of inodes with inline | |
539 | * data and fast symlinks. | |
540 | */ | |
3a0782d0 | 541 | if (fe->i_clusters) { |
2b4e30fb JB |
542 | if (ocfs2_should_order_data(inode)) |
543 | ocfs2_begin_ordered_truncate(inode, 0); | |
544 | ||
60b11392 MF |
545 | handle = ocfs2_start_trans(osb, OCFS2_INODE_UPDATE_CREDITS); |
546 | if (IS_ERR(handle)) { | |
547 | status = PTR_ERR(handle); | |
0350cb07 | 548 | handle = NULL; |
60b11392 MF |
549 | mlog_errno(status); |
550 | goto out; | |
551 | } | |
552 | ||
0cf2f763 JB |
553 | status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), |
554 | fe_bh, | |
13723d00 | 555 | OCFS2_JOURNAL_ACCESS_WRITE); |
60b11392 MF |
556 | if (status < 0) { |
557 | mlog_errno(status); | |
558 | goto out; | |
559 | } | |
560 | ||
561 | i_size_write(inode, 0); | |
562 | ||
563 | status = ocfs2_mark_inode_dirty(handle, inode, fe_bh); | |
564 | if (status < 0) { | |
565 | mlog_errno(status); | |
566 | goto out; | |
567 | } | |
568 | ||
569 | ocfs2_commit_trans(osb, handle); | |
570 | handle = NULL; | |
571 | ||
78f94673 | 572 | status = ocfs2_commit_truncate(osb, inode, fe_bh); |
3a0782d0 MF |
573 | if (status < 0) { |
574 | mlog_errno(status); | |
575 | goto out; | |
576 | } | |
ccd979bd | 577 | } |
ccd979bd | 578 | |
60b11392 MF |
579 | out: |
580 | if (handle) | |
581 | ocfs2_commit_trans(osb, handle); | |
ccd979bd MF |
582 | return status; |
583 | } | |
584 | ||
585 | static int ocfs2_remove_inode(struct inode *inode, | |
586 | struct buffer_head *di_bh, | |
587 | struct inode *orphan_dir_inode, | |
588 | struct buffer_head *orphan_dir_bh) | |
589 | { | |
590 | int status; | |
591 | struct inode *inode_alloc_inode = NULL; | |
592 | struct buffer_head *inode_alloc_bh = NULL; | |
1fabe148 | 593 | handle_t *handle; |
ccd979bd MF |
594 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); |
595 | struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; | |
596 | ||
597 | inode_alloc_inode = | |
598 | ocfs2_get_system_file_inode(osb, INODE_ALLOC_SYSTEM_INODE, | |
599 | le16_to_cpu(di->i_suballoc_slot)); | |
600 | if (!inode_alloc_inode) { | |
601 | status = -EEXIST; | |
602 | mlog_errno(status); | |
603 | goto bail; | |
604 | } | |
605 | ||
1b1dcc1b | 606 | mutex_lock(&inode_alloc_inode->i_mutex); |
e63aecb6 | 607 | status = ocfs2_inode_lock(inode_alloc_inode, &inode_alloc_bh, 1); |
ccd979bd | 608 | if (status < 0) { |
1b1dcc1b | 609 | mutex_unlock(&inode_alloc_inode->i_mutex); |
ccd979bd MF |
610 | |
611 | mlog_errno(status); | |
612 | goto bail; | |
613 | } | |
614 | ||
a90714c1 | 615 | handle = ocfs2_start_trans(osb, OCFS2_DELETE_INODE_CREDITS + |
9b7895ef | 616 | ocfs2_quota_trans_credits(inode->i_sb)); |
ccd979bd MF |
617 | if (IS_ERR(handle)) { |
618 | status = PTR_ERR(handle); | |
619 | mlog_errno(status); | |
620 | goto bail_unlock; | |
621 | } | |
622 | ||
d4cd1871 LD |
623 | if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { |
624 | status = ocfs2_orphan_del(osb, handle, orphan_dir_inode, inode, | |
625 | orphan_dir_bh); | |
626 | if (status < 0) { | |
627 | mlog_errno(status); | |
628 | goto bail_commit; | |
629 | } | |
ccd979bd MF |
630 | } |
631 | ||
632 | /* set the inodes dtime */ | |
0cf2f763 | 633 | status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), di_bh, |
13723d00 | 634 | OCFS2_JOURNAL_ACCESS_WRITE); |
ccd979bd MF |
635 | if (status < 0) { |
636 | mlog_errno(status); | |
637 | goto bail_commit; | |
638 | } | |
639 | ||
640 | di->i_dtime = cpu_to_le64(CURRENT_TIME.tv_sec); | |
4092d49f | 641 | di->i_flags &= cpu_to_le32(~(OCFS2_VALID_FL | OCFS2_ORPHANED_FL)); |
ec20cec7 | 642 | ocfs2_journal_dirty(handle, di_bh); |
ccd979bd | 643 | |
8cb471e8 | 644 | ocfs2_remove_from_cache(INODE_CACHE(inode), di_bh); |
63936dda | 645 | dquot_free_inode(inode); |
ccd979bd MF |
646 | |
647 | status = ocfs2_free_dinode(handle, inode_alloc_inode, | |
648 | inode_alloc_bh, di); | |
649 | if (status < 0) | |
650 | mlog_errno(status); | |
651 | ||
652 | bail_commit: | |
02dc1af4 | 653 | ocfs2_commit_trans(osb, handle); |
ccd979bd | 654 | bail_unlock: |
e63aecb6 | 655 | ocfs2_inode_unlock(inode_alloc_inode, 1); |
1b1dcc1b | 656 | mutex_unlock(&inode_alloc_inode->i_mutex); |
ccd979bd MF |
657 | brelse(inode_alloc_bh); |
658 | bail: | |
659 | iput(inode_alloc_inode); | |
660 | ||
661 | return status; | |
662 | } | |
663 | ||
2bd63216 | 664 | /* |
b4df6ed8 MF |
665 | * Serialize with orphan dir recovery. If the process doing |
666 | * recovery on this orphan dir does an iget() with the dir | |
667 | * i_mutex held, we'll deadlock here. Instead we detect this | |
668 | * and exit early - recovery will wipe this inode for us. | |
669 | */ | |
670 | static int ocfs2_check_orphan_recovery_state(struct ocfs2_super *osb, | |
671 | int slot) | |
672 | { | |
673 | int ret = 0; | |
674 | ||
675 | spin_lock(&osb->osb_lock); | |
676 | if (ocfs2_node_map_test_bit(osb, &osb->osb_recovering_orphan_dirs, slot)) { | |
b4df6ed8 MF |
677 | ret = -EDEADLK; |
678 | goto out; | |
679 | } | |
680 | /* This signals to the orphan recovery process that it should | |
681 | * wait for us to handle the wipe. */ | |
682 | osb->osb_orphan_wipes[slot]++; | |
683 | out: | |
684 | spin_unlock(&osb->osb_lock); | |
64f3b269 | 685 | trace_ocfs2_check_orphan_recovery_state(slot, ret); |
b4df6ed8 MF |
686 | return ret; |
687 | } | |
688 | ||
689 | static void ocfs2_signal_wipe_completion(struct ocfs2_super *osb, | |
690 | int slot) | |
691 | { | |
692 | spin_lock(&osb->osb_lock); | |
693 | osb->osb_orphan_wipes[slot]--; | |
694 | spin_unlock(&osb->osb_lock); | |
695 | ||
696 | wake_up(&osb->osb_wipe_event); | |
697 | } | |
698 | ||
ccd979bd MF |
699 | static int ocfs2_wipe_inode(struct inode *inode, |
700 | struct buffer_head *di_bh) | |
701 | { | |
d577632e | 702 | int status, orphaned_slot = -1; |
ccd979bd MF |
703 | struct inode *orphan_dir_inode = NULL; |
704 | struct buffer_head *orphan_dir_bh = NULL; | |
705 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
d4cd1871 | 706 | struct ocfs2_dinode *di = (struct ocfs2_dinode *) di_bh->b_data; |
ccd979bd | 707 | |
d4cd1871 LD |
708 | if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { |
709 | orphaned_slot = le16_to_cpu(di->i_orphaned_slot); | |
b4df6ed8 | 710 | |
d4cd1871 LD |
711 | status = ocfs2_check_orphan_recovery_state(osb, orphaned_slot); |
712 | if (status) | |
713 | return status; | |
b4df6ed8 | 714 | |
d4cd1871 LD |
715 | orphan_dir_inode = ocfs2_get_system_file_inode(osb, |
716 | ORPHAN_DIR_SYSTEM_INODE, | |
717 | orphaned_slot); | |
718 | if (!orphan_dir_inode) { | |
719 | status = -EEXIST; | |
720 | mlog_errno(status); | |
721 | goto bail; | |
722 | } | |
ccd979bd | 723 | |
d4cd1871 LD |
724 | /* Lock the orphan dir. The lock will be held for the entire |
725 | * delete_inode operation. We do this now to avoid races with | |
726 | * recovery completion on other nodes. */ | |
727 | mutex_lock(&orphan_dir_inode->i_mutex); | |
728 | status = ocfs2_inode_lock(orphan_dir_inode, &orphan_dir_bh, 1); | |
729 | if (status < 0) { | |
730 | mutex_unlock(&orphan_dir_inode->i_mutex); | |
ccd979bd | 731 | |
d4cd1871 LD |
732 | mlog_errno(status); |
733 | goto bail; | |
734 | } | |
ccd979bd MF |
735 | } |
736 | ||
737 | /* we do this while holding the orphan dir lock because we | |
34d024f8 MF |
738 | * don't want recovery being run from another node to try an |
739 | * inode delete underneath us -- this will result in two nodes | |
ccd979bd MF |
740 | * truncating the same file! */ |
741 | status = ocfs2_truncate_for_delete(osb, inode, di_bh); | |
742 | if (status < 0) { | |
743 | mlog_errno(status); | |
744 | goto bail_unlock_dir; | |
745 | } | |
746 | ||
9b7895ef MF |
747 | /* Remove any dir index tree */ |
748 | if (S_ISDIR(inode->i_mode)) { | |
749 | status = ocfs2_dx_dir_truncate(inode, di_bh); | |
750 | if (status) { | |
751 | mlog_errno(status); | |
752 | goto bail_unlock_dir; | |
753 | } | |
754 | } | |
755 | ||
cf1d6c76 TY |
756 | /*Free extended attribute resources associated with this inode.*/ |
757 | status = ocfs2_xattr_remove(inode, di_bh); | |
758 | if (status < 0) { | |
759 | mlog_errno(status); | |
760 | goto bail_unlock_dir; | |
761 | } | |
762 | ||
8b2c0dba TM |
763 | status = ocfs2_remove_refcount_tree(inode, di_bh); |
764 | if (status < 0) { | |
765 | mlog_errno(status); | |
766 | goto bail_unlock_dir; | |
767 | } | |
768 | ||
ccd979bd MF |
769 | status = ocfs2_remove_inode(inode, di_bh, orphan_dir_inode, |
770 | orphan_dir_bh); | |
771 | if (status < 0) | |
772 | mlog_errno(status); | |
773 | ||
774 | bail_unlock_dir: | |
d4cd1871 LD |
775 | if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR) |
776 | return status; | |
777 | ||
e63aecb6 | 778 | ocfs2_inode_unlock(orphan_dir_inode, 1); |
1b1dcc1b | 779 | mutex_unlock(&orphan_dir_inode->i_mutex); |
ccd979bd MF |
780 | brelse(orphan_dir_bh); |
781 | bail: | |
782 | iput(orphan_dir_inode); | |
b4df6ed8 | 783 | ocfs2_signal_wipe_completion(osb, orphaned_slot); |
ccd979bd MF |
784 | |
785 | return status; | |
786 | } | |
787 | ||
788 | /* There is a series of simple checks that should be done before a | |
34d024f8 | 789 | * trylock is even considered. Encapsulate those in this function. */ |
ccd979bd MF |
790 | static int ocfs2_inode_is_valid_to_delete(struct inode *inode) |
791 | { | |
792 | int ret = 0; | |
793 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
794 | struct ocfs2_super *osb = OCFS2_SB(inode->i_sb); | |
795 | ||
64f3b269 TM |
796 | trace_ocfs2_inode_is_valid_to_delete(current, osb->dc_task, |
797 | (unsigned long long)oi->ip_blkno, | |
798 | oi->ip_flags); | |
799 | ||
ccd979bd MF |
800 | /* We shouldn't be getting here for the root directory |
801 | * inode.. */ | |
802 | if (inode == osb->root_inode) { | |
803 | mlog(ML_ERROR, "Skipping delete of root inode.\n"); | |
804 | goto bail; | |
805 | } | |
806 | ||
34d024f8 | 807 | /* If we're coming from downconvert_thread we can't go into our own |
ccd979bd MF |
808 | * voting [hello, deadlock city!], so unforuntately we just |
809 | * have to skip deleting this guy. That's OK though because | |
810 | * the node who's doing the actual deleting should handle it | |
811 | * anyway. */ | |
64f3b269 | 812 | if (current == osb->dc_task) |
ccd979bd | 813 | goto bail; |
ccd979bd MF |
814 | |
815 | spin_lock(&oi->ip_lock); | |
816 | /* OCFS2 *never* deletes system files. This should technically | |
817 | * never get here as system file inodes should always have a | |
818 | * positive link count. */ | |
819 | if (oi->ip_flags & OCFS2_INODE_SYSTEM_FILE) { | |
b0697053 MF |
820 | mlog(ML_ERROR, "Skipping delete of system file %llu\n", |
821 | (unsigned long long)oi->ip_blkno); | |
ccd979bd MF |
822 | goto bail_unlock; |
823 | } | |
824 | ||
34d024f8 MF |
825 | /* If we have allowd wipe of this inode for another node, it |
826 | * will be marked here so we can safely skip it. Recovery will | |
25985edc | 827 | * cleanup any inodes we might inadvertently skip here. */ |
64f3b269 | 828 | if (oi->ip_flags & OCFS2_INODE_SKIP_DELETE) |
ccd979bd | 829 | goto bail_unlock; |
ccd979bd MF |
830 | |
831 | ret = 1; | |
832 | bail_unlock: | |
833 | spin_unlock(&oi->ip_lock); | |
834 | bail: | |
835 | return ret; | |
836 | } | |
837 | ||
838 | /* Query the cluster to determine whether we should wipe an inode from | |
839 | * disk or not. | |
840 | * | |
841 | * Requires the inode to have the cluster lock. */ | |
842 | static int ocfs2_query_inode_wipe(struct inode *inode, | |
843 | struct buffer_head *di_bh, | |
844 | int *wipe) | |
845 | { | |
64f3b269 | 846 | int status = 0, reason = 0; |
ccd979bd MF |
847 | struct ocfs2_inode_info *oi = OCFS2_I(inode); |
848 | struct ocfs2_dinode *di; | |
849 | ||
850 | *wipe = 0; | |
851 | ||
64f3b269 TM |
852 | trace_ocfs2_query_inode_wipe_begin((unsigned long long)oi->ip_blkno, |
853 | inode->i_nlink); | |
854 | ||
ccd979bd MF |
855 | /* While we were waiting for the cluster lock in |
856 | * ocfs2_delete_inode, another node might have asked to delete | |
857 | * the inode. Recheck our flags to catch this. */ | |
858 | if (!ocfs2_inode_is_valid_to_delete(inode)) { | |
64f3b269 | 859 | reason = 1; |
ccd979bd MF |
860 | goto bail; |
861 | } | |
862 | ||
863 | /* Now that we have an up to date inode, we can double check | |
864 | * the link count. */ | |
64f3b269 | 865 | if (inode->i_nlink) |
ccd979bd | 866 | goto bail; |
ccd979bd MF |
867 | |
868 | /* Do some basic inode verification... */ | |
869 | di = (struct ocfs2_dinode *) di_bh->b_data; | |
d4cd1871 LD |
870 | if (!(di->i_flags & cpu_to_le32(OCFS2_ORPHANED_FL)) && |
871 | !(oi->ip_flags & OCFS2_INODE_SKIP_ORPHAN_DIR)) { | |
b54c2ca4 TY |
872 | /* |
873 | * Inodes in the orphan dir must have ORPHANED_FL. The only | |
874 | * inodes that come back out of the orphan dir are reflink | |
875 | * targets. A reflink target may be moved out of the orphan | |
876 | * dir between the time we scan the directory and the time we | |
877 | * process it. This would lead to HAS_REFCOUNT_FL being set but | |
878 | * ORPHANED_FL not. | |
879 | */ | |
880 | if (di->i_dyn_features & cpu_to_le16(OCFS2_HAS_REFCOUNT_FL)) { | |
64f3b269 | 881 | reason = 2; |
b54c2ca4 TY |
882 | goto bail; |
883 | } | |
884 | ||
ccd979bd MF |
885 | /* for lack of a better error? */ |
886 | status = -EEXIST; | |
887 | mlog(ML_ERROR, | |
b0697053 | 888 | "Inode %llu (on-disk %llu) not orphaned! " |
ccd979bd | 889 | "Disk flags 0x%x, inode flags 0x%x\n", |
b0697053 | 890 | (unsigned long long)oi->ip_blkno, |
1ca1a111 MF |
891 | (unsigned long long)le64_to_cpu(di->i_blkno), |
892 | le32_to_cpu(di->i_flags), oi->ip_flags); | |
ccd979bd MF |
893 | goto bail; |
894 | } | |
895 | ||
896 | /* has someone already deleted us?! baaad... */ | |
897 | if (di->i_dtime) { | |
898 | status = -EEXIST; | |
899 | mlog_errno(status); | |
900 | goto bail; | |
901 | } | |
902 | ||
6f16bf65 MF |
903 | /* |
904 | * This is how ocfs2 determines whether an inode is still live | |
905 | * within the cluster. Every node takes a shared read lock on | |
906 | * the inode open lock in ocfs2_read_locked_inode(). When we | |
907 | * get to ->delete_inode(), each node tries to convert it's | |
908 | * lock to an exclusive. Trylocks are serialized by the inode | |
25985edc | 909 | * meta data lock. If the upconvert succeeds, we know the inode |
6f16bf65 MF |
910 | * is no longer live and can be deleted. |
911 | * | |
912 | * Though we call this with the meta data lock held, the | |
913 | * trylock keeps us from ABBA deadlock. | |
914 | */ | |
915 | status = ocfs2_try_open_lock(inode, 1); | |
50008630 | 916 | if (status == -EAGAIN) { |
ccd979bd | 917 | status = 0; |
64f3b269 | 918 | reason = 3; |
ccd979bd MF |
919 | goto bail; |
920 | } | |
921 | if (status < 0) { | |
922 | mlog_errno(status); | |
923 | goto bail; | |
924 | } | |
925 | ||
50008630 | 926 | *wipe = 1; |
64f3b269 | 927 | trace_ocfs2_query_inode_wipe_succ(le16_to_cpu(di->i_orphaned_slot)); |
ccd979bd MF |
928 | |
929 | bail: | |
64f3b269 | 930 | trace_ocfs2_query_inode_wipe_end(status, reason); |
ccd979bd MF |
931 | return status; |
932 | } | |
933 | ||
934 | /* Support function for ocfs2_delete_inode. Will help us keep the | |
935 | * inode data in a consistent state for clear_inode. Always truncates | |
936 | * pages, optionally sync's them first. */ | |
937 | static void ocfs2_cleanup_delete_inode(struct inode *inode, | |
938 | int sync_data) | |
939 | { | |
64f3b269 TM |
940 | trace_ocfs2_cleanup_delete_inode( |
941 | (unsigned long long)OCFS2_I(inode)->ip_blkno, sync_data); | |
ccd979bd | 942 | if (sync_data) |
249ec93c | 943 | filemap_write_and_wait(inode->i_mapping); |
ccd979bd MF |
944 | truncate_inode_pages(&inode->i_data, 0); |
945 | } | |
946 | ||
066d92dc | 947 | static void ocfs2_delete_inode(struct inode *inode) |
ccd979bd MF |
948 | { |
949 | int wipe, status; | |
e4b963f1 | 950 | sigset_t oldset; |
ccd979bd MF |
951 | struct buffer_head *di_bh = NULL; |
952 | ||
64f3b269 TM |
953 | trace_ocfs2_delete_inode(inode->i_ino, |
954 | (unsigned long long)OCFS2_I(inode)->ip_blkno, | |
955 | is_bad_inode(inode)); | |
ccd979bd | 956 | |
a90714c1 JK |
957 | /* When we fail in read_inode() we mark inode as bad. The second test |
958 | * catches the case when inode allocation fails before allocating | |
959 | * a block for inode. */ | |
64f3b269 | 960 | if (is_bad_inode(inode) || !OCFS2_I(inode)->ip_blkno) |
ccd979bd | 961 | goto bail; |
ccd979bd | 962 | |
871a2931 | 963 | dquot_initialize(inode); |
907f4554 | 964 | |
ccd979bd MF |
965 | if (!ocfs2_inode_is_valid_to_delete(inode)) { |
966 | /* It's probably not necessary to truncate_inode_pages | |
967 | * here but we do it for safety anyway (it will most | |
968 | * likely be a no-op anyway) */ | |
969 | ocfs2_cleanup_delete_inode(inode, 0); | |
970 | goto bail; | |
971 | } | |
972 | ||
973 | /* We want to block signals in delete_inode as the lock and | |
974 | * messaging paths may return us -ERESTARTSYS. Which would | |
975 | * cause us to exit early, resulting in inodes being orphaned | |
976 | * forever. */ | |
e4b963f1 | 977 | ocfs2_block_signals(&oldset); |
ccd979bd | 978 | |
6ca497a8 | 979 | /* |
980 | * Synchronize us against ocfs2_get_dentry. We take this in | |
981 | * shared mode so that all nodes can still concurrently | |
982 | * process deletes. | |
983 | */ | |
984 | status = ocfs2_nfs_sync_lock(OCFS2_SB(inode->i_sb), 0); | |
985 | if (status < 0) { | |
986 | mlog(ML_ERROR, "getting nfs sync lock(PR) failed %d\n", status); | |
987 | ocfs2_cleanup_delete_inode(inode, 0); | |
988 | goto bail_unblock; | |
989 | } | |
ccd979bd MF |
990 | /* Lock down the inode. This gives us an up to date view of |
991 | * it's metadata (for verification), and allows us to | |
34d024f8 | 992 | * serialize delete_inode on multiple nodes. |
ccd979bd MF |
993 | * |
994 | * Even though we might be doing a truncate, we don't take the | |
995 | * allocation lock here as it won't be needed - nobody will | |
996 | * have the file open. | |
997 | */ | |
e63aecb6 | 998 | status = ocfs2_inode_lock(inode, &di_bh, 1); |
ccd979bd MF |
999 | if (status < 0) { |
1000 | if (status != -ENOENT) | |
1001 | mlog_errno(status); | |
1002 | ocfs2_cleanup_delete_inode(inode, 0); | |
6ca497a8 | 1003 | goto bail_unlock_nfs_sync; |
ccd979bd MF |
1004 | } |
1005 | ||
1006 | /* Query the cluster. This will be the final decision made | |
1007 | * before we go ahead and wipe the inode. */ | |
1008 | status = ocfs2_query_inode_wipe(inode, di_bh, &wipe); | |
1009 | if (!wipe || status < 0) { | |
34d024f8 | 1010 | /* Error and remote inode busy both mean we won't be |
ccd979bd MF |
1011 | * removing the inode, so they take almost the same |
1012 | * path. */ | |
1013 | if (status < 0) | |
1014 | mlog_errno(status); | |
1015 | ||
34d024f8 MF |
1016 | /* Someone in the cluster has disallowed a wipe of |
1017 | * this inode, or it was never completely | |
1018 | * orphaned. Write out the pages and exit now. */ | |
ccd979bd MF |
1019 | ocfs2_cleanup_delete_inode(inode, 1); |
1020 | goto bail_unlock_inode; | |
1021 | } | |
1022 | ||
1023 | ocfs2_cleanup_delete_inode(inode, 0); | |
1024 | ||
1025 | status = ocfs2_wipe_inode(inode, di_bh); | |
1026 | if (status < 0) { | |
b4df6ed8 MF |
1027 | if (status != -EDEADLK) |
1028 | mlog_errno(status); | |
ccd979bd MF |
1029 | goto bail_unlock_inode; |
1030 | } | |
1031 | ||
24c19ef4 MF |
1032 | /* |
1033 | * Mark the inode as successfully deleted. | |
1034 | * | |
1035 | * This is important for ocfs2_clear_inode() as it will check | |
1036 | * this flag and skip any checkpointing work | |
1037 | * | |
1038 | * ocfs2_stuff_meta_lvb() also uses this flag to invalidate | |
1039 | * the LVB for other nodes. | |
1040 | */ | |
ccd979bd MF |
1041 | OCFS2_I(inode)->ip_flags |= OCFS2_INODE_DELETED; |
1042 | ||
1043 | bail_unlock_inode: | |
e63aecb6 | 1044 | ocfs2_inode_unlock(inode, 1); |
ccd979bd | 1045 | brelse(di_bh); |
6ca497a8 | 1046 | |
1047 | bail_unlock_nfs_sync: | |
1048 | ocfs2_nfs_sync_unlock(OCFS2_SB(inode->i_sb), 0); | |
1049 | ||
ccd979bd | 1050 | bail_unblock: |
e4b963f1 | 1051 | ocfs2_unblock_signals(&oldset); |
ccd979bd | 1052 | bail: |
c1e8d35e | 1053 | return; |
ccd979bd MF |
1054 | } |
1055 | ||
066d92dc | 1056 | static void ocfs2_clear_inode(struct inode *inode) |
ccd979bd MF |
1057 | { |
1058 | int status; | |
1059 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
1060 | ||
dbd5768f | 1061 | clear_inode(inode); |
64f3b269 TM |
1062 | trace_ocfs2_clear_inode((unsigned long long)oi->ip_blkno, |
1063 | inode->i_nlink); | |
ccd979bd MF |
1064 | |
1065 | mlog_bug_on_msg(OCFS2_SB(inode->i_sb) == NULL, | |
1066 | "Inode=%lu\n", inode->i_ino); | |
1067 | ||
9f754758 | 1068 | dquot_drop(inode); |
257ba15c | 1069 | |
34d024f8 MF |
1070 | /* To preven remote deletes we hold open lock before, now it |
1071 | * is time to unlock PR and EX open locks. */ | |
50008630 TY |
1072 | ocfs2_open_unlock(inode); |
1073 | ||
ccd979bd | 1074 | /* Do these before all the other work so that we don't bounce |
34d024f8 | 1075 | * the downconvert thread while waiting to destroy the locks. */ |
ccd979bd | 1076 | ocfs2_mark_lockres_freeing(&oi->ip_rw_lockres); |
e63aecb6 | 1077 | ocfs2_mark_lockres_freeing(&oi->ip_inode_lockres); |
50008630 | 1078 | ocfs2_mark_lockres_freeing(&oi->ip_open_lockres); |
ccd979bd | 1079 | |
4fe370af MF |
1080 | ocfs2_resv_discard(&OCFS2_SB(inode->i_sb)->osb_la_resmap, |
1081 | &oi->ip_la_data_resv); | |
1082 | ocfs2_resv_init_once(&oi->ip_la_data_resv); | |
1083 | ||
ccd979bd MF |
1084 | /* We very well may get a clear_inode before all an inodes |
1085 | * metadata has hit disk. Of course, we can't drop any cluster | |
1086 | * locks until the journal has finished with it. The only | |
1087 | * exception here are successfully wiped inodes - their | |
1088 | * metadata can now be considered to be part of the system | |
1089 | * inodes from which it came. */ | |
1090 | if (!(OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED)) | |
1091 | ocfs2_checkpoint_inode(inode); | |
1092 | ||
1093 | mlog_bug_on_msg(!list_empty(&oi->ip_io_markers), | |
b0697053 MF |
1094 | "Clear inode of %llu, inode has io markers\n", |
1095 | (unsigned long long)oi->ip_blkno); | |
ccd979bd | 1096 | |
83418978 MF |
1097 | ocfs2_extent_map_trunc(inode, 0); |
1098 | ||
ccd979bd MF |
1099 | status = ocfs2_drop_inode_locks(inode); |
1100 | if (status < 0) | |
1101 | mlog_errno(status); | |
1102 | ||
1103 | ocfs2_lock_res_free(&oi->ip_rw_lockres); | |
e63aecb6 | 1104 | ocfs2_lock_res_free(&oi->ip_inode_lockres); |
50008630 | 1105 | ocfs2_lock_res_free(&oi->ip_open_lockres); |
ccd979bd | 1106 | |
66fb345d | 1107 | ocfs2_metadata_cache_exit(INODE_CACHE(inode)); |
ccd979bd | 1108 | |
8cb471e8 | 1109 | mlog_bug_on_msg(INODE_CACHE(inode)->ci_num_cached, |
b0697053 | 1110 | "Clear inode of %llu, inode has %u cache items\n", |
8cb471e8 JB |
1111 | (unsigned long long)oi->ip_blkno, |
1112 | INODE_CACHE(inode)->ci_num_cached); | |
ccd979bd | 1113 | |
8cb471e8 | 1114 | mlog_bug_on_msg(!(INODE_CACHE(inode)->ci_flags & OCFS2_CACHE_FL_INLINE), |
b0697053 MF |
1115 | "Clear inode of %llu, inode has a bad flag\n", |
1116 | (unsigned long long)oi->ip_blkno); | |
ccd979bd MF |
1117 | |
1118 | mlog_bug_on_msg(spin_is_locked(&oi->ip_lock), | |
b0697053 MF |
1119 | "Clear inode of %llu, inode is locked\n", |
1120 | (unsigned long long)oi->ip_blkno); | |
ccd979bd | 1121 | |
251b6ecc | 1122 | mlog_bug_on_msg(!mutex_trylock(&oi->ip_io_mutex), |
b0697053 MF |
1123 | "Clear inode of %llu, io_mutex is locked\n", |
1124 | (unsigned long long)oi->ip_blkno); | |
251b6ecc | 1125 | mutex_unlock(&oi->ip_io_mutex); |
ccd979bd MF |
1126 | |
1127 | /* | |
1128 | * down_trylock() returns 0, down_write_trylock() returns 1 | |
1129 | * kernel 1, world 0 | |
1130 | */ | |
1131 | mlog_bug_on_msg(!down_write_trylock(&oi->ip_alloc_sem), | |
b0697053 MF |
1132 | "Clear inode of %llu, alloc_sem is locked\n", |
1133 | (unsigned long long)oi->ip_blkno); | |
ccd979bd MF |
1134 | up_write(&oi->ip_alloc_sem); |
1135 | ||
1136 | mlog_bug_on_msg(oi->ip_open_count, | |
b0697053 MF |
1137 | "Clear inode of %llu has open count %d\n", |
1138 | (unsigned long long)oi->ip_blkno, oi->ip_open_count); | |
ccd979bd MF |
1139 | |
1140 | /* Clear all other flags. */ | |
47460d65 | 1141 | oi->ip_flags = 0; |
ccd979bd MF |
1142 | oi->ip_dir_start_lookup = 0; |
1143 | oi->ip_blkno = 0ULL; | |
ae0dff68 SM |
1144 | |
1145 | /* | |
1146 | * ip_jinode is used to track txns against this inode. We ensure that | |
1147 | * the journal is flushed before journal shutdown. Thus it is safe to | |
1148 | * have inodes get cleaned up after journal shutdown. | |
1149 | */ | |
2b4e30fb JB |
1150 | jbd2_journal_release_jbd_inode(OCFS2_SB(inode->i_sb)->journal->j_journal, |
1151 | &oi->ip_jinode); | |
ccd979bd MF |
1152 | } |
1153 | ||
066d92dc AV |
1154 | void ocfs2_evict_inode(struct inode *inode) |
1155 | { | |
1156 | if (!inode->i_nlink || | |
1157 | (OCFS2_I(inode)->ip_flags & OCFS2_INODE_MAYBE_ORPHANED)) { | |
1158 | ocfs2_delete_inode(inode); | |
1159 | } else { | |
1160 | truncate_inode_pages(&inode->i_data, 0); | |
1161 | } | |
1162 | ocfs2_clear_inode(inode); | |
1163 | } | |
1164 | ||
ccd979bd MF |
1165 | /* Called under inode_lock, with no more references on the |
1166 | * struct inode, so it's safe here to check the flags field | |
1167 | * and to manipulate i_nlink without any other locks. */ | |
45321ac5 | 1168 | int ocfs2_drop_inode(struct inode *inode) |
ccd979bd MF |
1169 | { |
1170 | struct ocfs2_inode_info *oi = OCFS2_I(inode); | |
45321ac5 | 1171 | int res; |
ccd979bd | 1172 | |
64f3b269 TM |
1173 | trace_ocfs2_drop_inode((unsigned long long)oi->ip_blkno, |
1174 | inode->i_nlink, oi->ip_flags); | |
ccd979bd | 1175 | |
379dfe9d | 1176 | if (oi->ip_flags & OCFS2_INODE_MAYBE_ORPHANED) |
45321ac5 | 1177 | res = 1; |
379dfe9d | 1178 | else |
45321ac5 | 1179 | res = generic_drop_inode(inode); |
ccd979bd | 1180 | |
45321ac5 | 1181 | return res; |
ccd979bd MF |
1182 | } |
1183 | ||
ccd979bd MF |
1184 | /* |
1185 | * This is called from our getattr. | |
1186 | */ | |
1187 | int ocfs2_inode_revalidate(struct dentry *dentry) | |
1188 | { | |
1189 | struct inode *inode = dentry->d_inode; | |
1190 | int status = 0; | |
1191 | ||
64f3b269 TM |
1192 | trace_ocfs2_inode_revalidate(inode, |
1193 | inode ? (unsigned long long)OCFS2_I(inode)->ip_blkno : 0ULL, | |
1194 | inode ? (unsigned long long)OCFS2_I(inode)->ip_flags : 0); | |
ccd979bd MF |
1195 | |
1196 | if (!inode) { | |
ccd979bd MF |
1197 | status = -ENOENT; |
1198 | goto bail; | |
1199 | } | |
1200 | ||
1201 | spin_lock(&OCFS2_I(inode)->ip_lock); | |
1202 | if (OCFS2_I(inode)->ip_flags & OCFS2_INODE_DELETED) { | |
1203 | spin_unlock(&OCFS2_I(inode)->ip_lock); | |
ccd979bd MF |
1204 | status = -ENOENT; |
1205 | goto bail; | |
1206 | } | |
1207 | spin_unlock(&OCFS2_I(inode)->ip_lock); | |
1208 | ||
e63aecb6 | 1209 | /* Let ocfs2_inode_lock do the work of updating our struct |
ccd979bd | 1210 | * inode for us. */ |
e63aecb6 | 1211 | status = ocfs2_inode_lock(inode, NULL, 0); |
ccd979bd MF |
1212 | if (status < 0) { |
1213 | if (status != -ENOENT) | |
1214 | mlog_errno(status); | |
1215 | goto bail; | |
1216 | } | |
e63aecb6 | 1217 | ocfs2_inode_unlock(inode, 0); |
ccd979bd | 1218 | bail: |
ccd979bd MF |
1219 | return status; |
1220 | } | |
1221 | ||
1222 | /* | |
1223 | * Updates a disk inode from a | |
1224 | * struct inode. | |
1225 | * Only takes ip_lock. | |
1226 | */ | |
1fabe148 | 1227 | int ocfs2_mark_inode_dirty(handle_t *handle, |
ccd979bd MF |
1228 | struct inode *inode, |
1229 | struct buffer_head *bh) | |
1230 | { | |
1231 | int status; | |
1232 | struct ocfs2_dinode *fe = (struct ocfs2_dinode *) bh->b_data; | |
1233 | ||
64f3b269 | 1234 | trace_ocfs2_mark_inode_dirty((unsigned long long)OCFS2_I(inode)->ip_blkno); |
ccd979bd | 1235 | |
0cf2f763 | 1236 | status = ocfs2_journal_access_di(handle, INODE_CACHE(inode), bh, |
13723d00 | 1237 | OCFS2_JOURNAL_ACCESS_WRITE); |
ccd979bd MF |
1238 | if (status < 0) { |
1239 | mlog_errno(status); | |
1240 | goto leave; | |
1241 | } | |
1242 | ||
1243 | spin_lock(&OCFS2_I(inode)->ip_lock); | |
1244 | fe->i_clusters = cpu_to_le32(OCFS2_I(inode)->ip_clusters); | |
6e4b0d56 | 1245 | ocfs2_get_inode_flags(OCFS2_I(inode)); |
ca4d147e | 1246 | fe->i_attr = cpu_to_le32(OCFS2_I(inode)->ip_attr); |
15b1e36b | 1247 | fe->i_dyn_features = cpu_to_le16(OCFS2_I(inode)->ip_dyn_features); |
ccd979bd MF |
1248 | spin_unlock(&OCFS2_I(inode)->ip_lock); |
1249 | ||
1250 | fe->i_size = cpu_to_le64(i_size_read(inode)); | |
198a1ca3 | 1251 | ocfs2_set_links_count(fe, inode->i_nlink); |
2c034176 EB |
1252 | fe->i_uid = cpu_to_le32(i_uid_read(inode)); |
1253 | fe->i_gid = cpu_to_le32(i_gid_read(inode)); | |
ccd979bd MF |
1254 | fe->i_mode = cpu_to_le16(inode->i_mode); |
1255 | fe->i_atime = cpu_to_le64(inode->i_atime.tv_sec); | |
1256 | fe->i_atime_nsec = cpu_to_le32(inode->i_atime.tv_nsec); | |
1257 | fe->i_ctime = cpu_to_le64(inode->i_ctime.tv_sec); | |
1258 | fe->i_ctime_nsec = cpu_to_le32(inode->i_ctime.tv_nsec); | |
1259 | fe->i_mtime = cpu_to_le64(inode->i_mtime.tv_sec); | |
1260 | fe->i_mtime_nsec = cpu_to_le32(inode->i_mtime.tv_nsec); | |
1261 | ||
ec20cec7 | 1262 | ocfs2_journal_dirty(handle, bh); |
ccd979bd | 1263 | leave: |
ccd979bd MF |
1264 | return status; |
1265 | } | |
1266 | ||
1267 | /* | |
1268 | * | |
1269 | * Updates a struct inode from a disk inode. | |
1270 | * does no i/o, only takes ip_lock. | |
1271 | */ | |
1272 | void ocfs2_refresh_inode(struct inode *inode, | |
1273 | struct ocfs2_dinode *fe) | |
1274 | { | |
ccd979bd MF |
1275 | spin_lock(&OCFS2_I(inode)->ip_lock); |
1276 | ||
1277 | OCFS2_I(inode)->ip_clusters = le32_to_cpu(fe->i_clusters); | |
ca4d147e | 1278 | OCFS2_I(inode)->ip_attr = le32_to_cpu(fe->i_attr); |
15b1e36b | 1279 | OCFS2_I(inode)->ip_dyn_features = le16_to_cpu(fe->i_dyn_features); |
ca4d147e | 1280 | ocfs2_set_inode_flags(inode); |
ccd979bd | 1281 | i_size_write(inode, le64_to_cpu(fe->i_size)); |
bfe86848 | 1282 | set_nlink(inode, ocfs2_read_links_count(fe)); |
2c034176 EB |
1283 | i_uid_write(inode, le32_to_cpu(fe->i_uid)); |
1284 | i_gid_write(inode, le32_to_cpu(fe->i_gid)); | |
ccd979bd | 1285 | inode->i_mode = le16_to_cpu(fe->i_mode); |
ccd979bd MF |
1286 | if (S_ISLNK(inode->i_mode) && le32_to_cpu(fe->i_clusters) == 0) |
1287 | inode->i_blocks = 0; | |
1288 | else | |
8110b073 | 1289 | inode->i_blocks = ocfs2_inode_sector_count(inode); |
ccd979bd MF |
1290 | inode->i_atime.tv_sec = le64_to_cpu(fe->i_atime); |
1291 | inode->i_atime.tv_nsec = le32_to_cpu(fe->i_atime_nsec); | |
1292 | inode->i_mtime.tv_sec = le64_to_cpu(fe->i_mtime); | |
1293 | inode->i_mtime.tv_nsec = le32_to_cpu(fe->i_mtime_nsec); | |
1294 | inode->i_ctime.tv_sec = le64_to_cpu(fe->i_ctime); | |
1295 | inode->i_ctime.tv_nsec = le32_to_cpu(fe->i_ctime_nsec); | |
1296 | ||
1297 | spin_unlock(&OCFS2_I(inode)->ip_lock); | |
1298 | } | |
b657c95c JB |
1299 | |
1300 | int ocfs2_validate_inode_block(struct super_block *sb, | |
1301 | struct buffer_head *bh) | |
1302 | { | |
d6b32bbb | 1303 | int rc; |
b657c95c JB |
1304 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; |
1305 | ||
64f3b269 | 1306 | trace_ocfs2_validate_inode_block((unsigned long long)bh->b_blocknr); |
970e4936 | 1307 | |
b657c95c JB |
1308 | BUG_ON(!buffer_uptodate(bh)); |
1309 | ||
d6b32bbb JB |
1310 | /* |
1311 | * If the ecc fails, we return the error but otherwise | |
1312 | * leave the filesystem running. We know any error is | |
1313 | * local to this block. | |
1314 | */ | |
1315 | rc = ocfs2_validate_meta_ecc(sb, bh->b_data, &di->i_check); | |
13723d00 JB |
1316 | if (rc) { |
1317 | mlog(ML_ERROR, "Checksum failed for dinode %llu\n", | |
1318 | (unsigned long long)bh->b_blocknr); | |
d6b32bbb | 1319 | goto bail; |
13723d00 | 1320 | } |
d6b32bbb JB |
1321 | |
1322 | /* | |
1323 | * Errors after here are fatal. | |
1324 | */ | |
1325 | ||
1326 | rc = -EINVAL; | |
1327 | ||
b657c95c JB |
1328 | if (!OCFS2_IS_VALID_DINODE(di)) { |
1329 | ocfs2_error(sb, "Invalid dinode #%llu: signature = %.*s\n", | |
1330 | (unsigned long long)bh->b_blocknr, 7, | |
1331 | di->i_signature); | |
1332 | goto bail; | |
1333 | } | |
1334 | ||
1335 | if (le64_to_cpu(di->i_blkno) != bh->b_blocknr) { | |
1336 | ocfs2_error(sb, "Invalid dinode #%llu: i_blkno is %llu\n", | |
1337 | (unsigned long long)bh->b_blocknr, | |
1338 | (unsigned long long)le64_to_cpu(di->i_blkno)); | |
1339 | goto bail; | |
1340 | } | |
1341 | ||
1342 | if (!(di->i_flags & cpu_to_le32(OCFS2_VALID_FL))) { | |
1343 | ocfs2_error(sb, | |
1344 | "Invalid dinode #%llu: OCFS2_VALID_FL not set\n", | |
1345 | (unsigned long long)bh->b_blocknr); | |
1346 | goto bail; | |
1347 | } | |
1348 | ||
1349 | if (le32_to_cpu(di->i_fs_generation) != | |
1350 | OCFS2_SB(sb)->fs_generation) { | |
1351 | ocfs2_error(sb, | |
1352 | "Invalid dinode #%llu: fs_generation is %u\n", | |
1353 | (unsigned long long)bh->b_blocknr, | |
1354 | le32_to_cpu(di->i_fs_generation)); | |
1355 | goto bail; | |
1356 | } | |
1357 | ||
1358 | rc = 0; | |
1359 | ||
1360 | bail: | |
1361 | return rc; | |
1362 | } | |
1363 | ||
1364 | int ocfs2_read_inode_block_full(struct inode *inode, struct buffer_head **bh, | |
1365 | int flags) | |
1366 | { | |
1367 | int rc; | |
1368 | struct buffer_head *tmp = *bh; | |
1369 | ||
8cb471e8 JB |
1370 | rc = ocfs2_read_blocks(INODE_CACHE(inode), OCFS2_I(inode)->ip_blkno, |
1371 | 1, &tmp, flags, ocfs2_validate_inode_block); | |
b657c95c JB |
1372 | |
1373 | /* If ocfs2_read_blocks() got us a new bh, pass it up. */ | |
970e4936 | 1374 | if (!rc && !*bh) |
b657c95c JB |
1375 | *bh = tmp; |
1376 | ||
b657c95c JB |
1377 | return rc; |
1378 | } | |
1379 | ||
1380 | int ocfs2_read_inode_block(struct inode *inode, struct buffer_head **bh) | |
1381 | { | |
1382 | return ocfs2_read_inode_block_full(inode, bh, 0); | |
1383 | } | |
6e5a3d75 | 1384 | |
6e5a3d75 JB |
1385 | |
1386 | static u64 ocfs2_inode_cache_owner(struct ocfs2_caching_info *ci) | |
1387 | { | |
1388 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1389 | ||
1390 | return oi->ip_blkno; | |
1391 | } | |
1392 | ||
8cb471e8 JB |
1393 | static struct super_block *ocfs2_inode_cache_get_super(struct ocfs2_caching_info *ci) |
1394 | { | |
1395 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1396 | ||
1397 | return oi->vfs_inode.i_sb; | |
1398 | } | |
1399 | ||
6e5a3d75 JB |
1400 | static void ocfs2_inode_cache_lock(struct ocfs2_caching_info *ci) |
1401 | { | |
1402 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1403 | ||
1404 | spin_lock(&oi->ip_lock); | |
1405 | } | |
1406 | ||
1407 | static void ocfs2_inode_cache_unlock(struct ocfs2_caching_info *ci) | |
1408 | { | |
1409 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1410 | ||
1411 | spin_unlock(&oi->ip_lock); | |
1412 | } | |
1413 | ||
1414 | static void ocfs2_inode_cache_io_lock(struct ocfs2_caching_info *ci) | |
1415 | { | |
1416 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1417 | ||
1418 | mutex_lock(&oi->ip_io_mutex); | |
1419 | } | |
1420 | ||
1421 | static void ocfs2_inode_cache_io_unlock(struct ocfs2_caching_info *ci) | |
1422 | { | |
1423 | struct ocfs2_inode_info *oi = cache_info_to_inode(ci); | |
1424 | ||
1425 | mutex_unlock(&oi->ip_io_mutex); | |
1426 | } | |
1427 | ||
1428 | const struct ocfs2_caching_operations ocfs2_inode_caching_ops = { | |
1429 | .co_owner = ocfs2_inode_cache_owner, | |
8cb471e8 | 1430 | .co_get_super = ocfs2_inode_cache_get_super, |
6e5a3d75 JB |
1431 | .co_cache_lock = ocfs2_inode_cache_lock, |
1432 | .co_cache_unlock = ocfs2_inode_cache_unlock, | |
1433 | .co_io_lock = ocfs2_inode_cache_io_lock, | |
1434 | .co_io_unlock = ocfs2_inode_cache_io_unlock, | |
1435 | }; | |
1436 |