Commit | Line | Data |
---|---|---|
ccd979bd MF |
1 | /* -*- mode: c; c-basic-offset: 8; -*- |
2 | * vim: noexpandtab sw=8 ts=8 sts=0: | |
3 | * | |
4 | * super.c | |
5 | * | |
6 | * load/unload driver, mount/dismount volumes | |
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/module.h> | |
27 | #include <linux/fs.h> | |
28 | #include <linux/types.h> | |
29 | #include <linux/slab.h> | |
30 | #include <linux/highmem.h> | |
31 | #include <linux/utsname.h> | |
32 | #include <linux/init.h> | |
33 | #include <linux/random.h> | |
34 | #include <linux/statfs.h> | |
35 | #include <linux/moduleparam.h> | |
36 | #include <linux/blkdev.h> | |
37 | #include <linux/socket.h> | |
38 | #include <linux/inet.h> | |
39 | #include <linux/parser.h> | |
40 | #include <linux/crc32.h> | |
41 | #include <linux/debugfs.h> | |
d550071c | 42 | #include <linux/mount.h> |
6953b4c0 | 43 | #include <linux/seq_file.h> |
ccd979bd MF |
44 | |
45 | #define MLOG_MASK_PREFIX ML_SUPER | |
46 | #include <cluster/masklog.h> | |
47 | ||
48 | #include "ocfs2.h" | |
49 | ||
50 | /* this should be the only file to include a version 1 header */ | |
51 | #include "ocfs1_fs_compat.h" | |
52 | ||
53 | #include "alloc.h" | |
54 | #include "dlmglue.h" | |
55 | #include "export.h" | |
56 | #include "extent_map.h" | |
57 | #include "heartbeat.h" | |
58 | #include "inode.h" | |
59 | #include "journal.h" | |
60 | #include "localalloc.h" | |
61 | #include "namei.h" | |
62 | #include "slot_map.h" | |
63 | #include "super.h" | |
64 | #include "sysfile.h" | |
65 | #include "uptodate.h" | |
66 | #include "ver.h" | |
cf1d6c76 | 67 | #include "xattr.h" |
9e33d69f | 68 | #include "quota.h" |
ccd979bd MF |
69 | |
70 | #include "buffer_head_io.h" | |
71 | ||
e18b890b | 72 | static struct kmem_cache *ocfs2_inode_cachep = NULL; |
9e33d69f JK |
73 | struct kmem_cache *ocfs2_dquot_cachep; |
74 | struct kmem_cache *ocfs2_qf_chunk_cachep; | |
ccd979bd | 75 | |
ccd979bd MF |
76 | /* OCFS2 needs to schedule several differnt types of work which |
77 | * require cluster locking, disk I/O, recovery waits, etc. Since these | |
78 | * types of work tend to be heavy we avoid using the kernel events | |
79 | * workqueue and schedule on our own. */ | |
80 | struct workqueue_struct *ocfs2_wq = NULL; | |
81 | ||
82 | static struct dentry *ocfs2_debugfs_root = NULL; | |
83 | ||
84 | MODULE_AUTHOR("Oracle"); | |
85 | MODULE_LICENSE("GPL"); | |
86 | ||
c0123ade TY |
87 | struct mount_options |
88 | { | |
d147b3d6 | 89 | unsigned long commit_interval; |
c0123ade TY |
90 | unsigned long mount_opt; |
91 | unsigned int atime_quantum; | |
92 | signed short slot; | |
2fbe8d1e | 93 | unsigned int localalloc_opt; |
b61817e1 | 94 | char cluster_stack[OCFS2_STACK_LABEL_LEN + 1]; |
c0123ade TY |
95 | }; |
96 | ||
ccd979bd | 97 | static int ocfs2_parse_options(struct super_block *sb, char *options, |
c0123ade | 98 | struct mount_options *mopt, |
baf4661a | 99 | int is_remount); |
d550071c | 100 | static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); |
ccd979bd MF |
101 | static void ocfs2_put_super(struct super_block *sb); |
102 | static int ocfs2_mount_volume(struct super_block *sb); | |
103 | static int ocfs2_remount(struct super_block *sb, int *flags, char *data); | |
104 | static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); | |
105 | static int ocfs2_initialize_mem_caches(void); | |
106 | static void ocfs2_free_mem_caches(void); | |
107 | static void ocfs2_delete_osb(struct ocfs2_super *osb); | |
108 | ||
726c3342 | 109 | static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); |
ccd979bd MF |
110 | |
111 | static int ocfs2_sync_fs(struct super_block *sb, int wait); | |
112 | ||
113 | static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); | |
114 | static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); | |
bddb8eb3 | 115 | static void ocfs2_release_system_inodes(struct ocfs2_super *osb); |
ccd979bd MF |
116 | static int ocfs2_check_volume(struct ocfs2_super *osb); |
117 | static int ocfs2_verify_volume(struct ocfs2_dinode *di, | |
118 | struct buffer_head *bh, | |
119 | u32 sectsize); | |
120 | static int ocfs2_initialize_super(struct super_block *sb, | |
121 | struct buffer_head *bh, | |
122 | int sector_size); | |
123 | static int ocfs2_get_sector(struct super_block *sb, | |
124 | struct buffer_head **bh, | |
125 | int block, | |
126 | int sect_size); | |
127 | static void ocfs2_write_super(struct super_block *sb); | |
128 | static struct inode *ocfs2_alloc_inode(struct super_block *sb); | |
129 | static void ocfs2_destroy_inode(struct inode *inode); | |
130 | ||
ee9b6d61 | 131 | static const struct super_operations ocfs2_sops = { |
ccd979bd MF |
132 | .statfs = ocfs2_statfs, |
133 | .alloc_inode = ocfs2_alloc_inode, | |
134 | .destroy_inode = ocfs2_destroy_inode, | |
135 | .drop_inode = ocfs2_drop_inode, | |
136 | .clear_inode = ocfs2_clear_inode, | |
137 | .delete_inode = ocfs2_delete_inode, | |
138 | .sync_fs = ocfs2_sync_fs, | |
139 | .write_super = ocfs2_write_super, | |
140 | .put_super = ocfs2_put_super, | |
141 | .remount_fs = ocfs2_remount, | |
d550071c | 142 | .show_options = ocfs2_show_options, |
9e33d69f JK |
143 | .quota_read = ocfs2_quota_read, |
144 | .quota_write = ocfs2_quota_write, | |
ccd979bd MF |
145 | }; |
146 | ||
147 | enum { | |
148 | Opt_barrier, | |
149 | Opt_err_panic, | |
150 | Opt_err_ro, | |
151 | Opt_intr, | |
152 | Opt_nointr, | |
153 | Opt_hb_none, | |
154 | Opt_hb_local, | |
155 | Opt_data_ordered, | |
156 | Opt_data_writeback, | |
7f1a37e3 | 157 | Opt_atime_quantum, |
baf4661a | 158 | Opt_slot, |
d147b3d6 | 159 | Opt_commit, |
2fbe8d1e | 160 | Opt_localalloc, |
53fc622b | 161 | Opt_localflocks, |
b61817e1 | 162 | Opt_stack, |
cf1d6c76 TY |
163 | Opt_user_xattr, |
164 | Opt_nouser_xattr, | |
12462f1d | 165 | Opt_inode64, |
a68979b8 TY |
166 | Opt_acl, |
167 | Opt_noacl, | |
ccd979bd MF |
168 | Opt_err, |
169 | }; | |
170 | ||
a447c093 | 171 | static const match_table_t tokens = { |
ccd979bd MF |
172 | {Opt_barrier, "barrier=%u"}, |
173 | {Opt_err_panic, "errors=panic"}, | |
174 | {Opt_err_ro, "errors=remount-ro"}, | |
175 | {Opt_intr, "intr"}, | |
176 | {Opt_nointr, "nointr"}, | |
177 | {Opt_hb_none, OCFS2_HB_NONE}, | |
178 | {Opt_hb_local, OCFS2_HB_LOCAL}, | |
179 | {Opt_data_ordered, "data=ordered"}, | |
180 | {Opt_data_writeback, "data=writeback"}, | |
7f1a37e3 | 181 | {Opt_atime_quantum, "atime_quantum=%u"}, |
baf4661a | 182 | {Opt_slot, "preferred_slot=%u"}, |
d147b3d6 | 183 | {Opt_commit, "commit=%u"}, |
2fbe8d1e | 184 | {Opt_localalloc, "localalloc=%d"}, |
53fc622b | 185 | {Opt_localflocks, "localflocks"}, |
b61817e1 | 186 | {Opt_stack, "cluster_stack=%s"}, |
cf1d6c76 TY |
187 | {Opt_user_xattr, "user_xattr"}, |
188 | {Opt_nouser_xattr, "nouser_xattr"}, | |
12462f1d | 189 | {Opt_inode64, "inode64"}, |
a68979b8 TY |
190 | {Opt_acl, "acl"}, |
191 | {Opt_noacl, "noacl"}, | |
ccd979bd MF |
192 | {Opt_err, NULL} |
193 | }; | |
194 | ||
195 | /* | |
196 | * write_super and sync_fs ripped right out of ext3. | |
197 | */ | |
198 | static void ocfs2_write_super(struct super_block *sb) | |
199 | { | |
7892f2f4 | 200 | if (mutex_trylock(&sb->s_lock) != 0) |
ccd979bd MF |
201 | BUG(); |
202 | sb->s_dirt = 0; | |
203 | } | |
204 | ||
205 | static int ocfs2_sync_fs(struct super_block *sb, int wait) | |
206 | { | |
bddb8eb3 | 207 | int status; |
ccd979bd MF |
208 | tid_t target; |
209 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
210 | ||
211 | sb->s_dirt = 0; | |
212 | ||
213 | if (ocfs2_is_hard_readonly(osb)) | |
214 | return -EROFS; | |
215 | ||
216 | if (wait) { | |
217 | status = ocfs2_flush_truncate_log(osb); | |
218 | if (status < 0) | |
219 | mlog_errno(status); | |
220 | } else { | |
221 | ocfs2_schedule_truncate_log_flush(osb, 0); | |
222 | } | |
223 | ||
2b4e30fb JB |
224 | if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal, |
225 | &target)) { | |
ccd979bd | 226 | if (wait) |
2b4e30fb JB |
227 | jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal, |
228 | target); | |
ccd979bd MF |
229 | } |
230 | return 0; | |
231 | } | |
232 | ||
1a224ad1 JK |
233 | static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino) |
234 | { | |
235 | if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA) | |
236 | && (ino == USER_QUOTA_SYSTEM_INODE | |
237 | || ino == LOCAL_USER_QUOTA_SYSTEM_INODE)) | |
238 | return 0; | |
239 | if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA) | |
240 | && (ino == GROUP_QUOTA_SYSTEM_INODE | |
241 | || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE)) | |
242 | return 0; | |
243 | return 1; | |
244 | } | |
245 | ||
ccd979bd MF |
246 | static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) |
247 | { | |
248 | struct inode *new = NULL; | |
249 | int status = 0; | |
250 | int i; | |
251 | ||
252 | mlog_entry_void(); | |
253 | ||
5fa0613e | 254 | new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0); |
ccd979bd MF |
255 | if (IS_ERR(new)) { |
256 | status = PTR_ERR(new); | |
257 | mlog_errno(status); | |
258 | goto bail; | |
259 | } | |
260 | osb->root_inode = new; | |
261 | ||
5fa0613e | 262 | new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0); |
ccd979bd MF |
263 | if (IS_ERR(new)) { |
264 | status = PTR_ERR(new); | |
265 | mlog_errno(status); | |
266 | goto bail; | |
267 | } | |
268 | osb->sys_root_inode = new; | |
269 | ||
270 | for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; | |
271 | i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { | |
1a224ad1 JK |
272 | if (!ocfs2_need_system_inode(osb, i)) |
273 | continue; | |
ccd979bd MF |
274 | new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); |
275 | if (!new) { | |
276 | ocfs2_release_system_inodes(osb); | |
277 | status = -EINVAL; | |
278 | mlog_errno(status); | |
279 | /* FIXME: Should ERROR_RO_FS */ | |
280 | mlog(ML_ERROR, "Unable to load system inode %d, " | |
281 | "possibly corrupt fs?", i); | |
282 | goto bail; | |
283 | } | |
284 | // the array now has one ref, so drop this one | |
285 | iput(new); | |
286 | } | |
287 | ||
288 | bail: | |
289 | mlog_exit(status); | |
290 | return status; | |
291 | } | |
292 | ||
293 | static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) | |
294 | { | |
295 | struct inode *new = NULL; | |
296 | int status = 0; | |
297 | int i; | |
298 | ||
299 | mlog_entry_void(); | |
300 | ||
301 | for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; | |
302 | i < NUM_SYSTEM_INODES; | |
303 | i++) { | |
1a224ad1 JK |
304 | if (!ocfs2_need_system_inode(osb, i)) |
305 | continue; | |
ccd979bd MF |
306 | new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); |
307 | if (!new) { | |
308 | ocfs2_release_system_inodes(osb); | |
309 | status = -EINVAL; | |
310 | mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", | |
311 | status, i, osb->slot_num); | |
312 | goto bail; | |
313 | } | |
314 | /* the array now has one ref, so drop this one */ | |
315 | iput(new); | |
316 | } | |
317 | ||
318 | bail: | |
319 | mlog_exit(status); | |
320 | return status; | |
321 | } | |
322 | ||
bddb8eb3 | 323 | static void ocfs2_release_system_inodes(struct ocfs2_super *osb) |
ccd979bd | 324 | { |
bddb8eb3 | 325 | int i; |
ccd979bd MF |
326 | struct inode *inode; |
327 | ||
328 | mlog_entry_void(); | |
329 | ||
330 | for (i = 0; i < NUM_SYSTEM_INODES; i++) { | |
331 | inode = osb->system_inodes[i]; | |
332 | if (inode) { | |
333 | iput(inode); | |
334 | osb->system_inodes[i] = NULL; | |
335 | } | |
336 | } | |
337 | ||
338 | inode = osb->sys_root_inode; | |
339 | if (inode) { | |
340 | iput(inode); | |
341 | osb->sys_root_inode = NULL; | |
342 | } | |
343 | ||
344 | inode = osb->root_inode; | |
345 | if (inode) { | |
346 | iput(inode); | |
347 | osb->root_inode = NULL; | |
348 | } | |
349 | ||
bddb8eb3 | 350 | mlog_exit(0); |
ccd979bd MF |
351 | } |
352 | ||
353 | /* We're allocating fs objects, use GFP_NOFS */ | |
354 | static struct inode *ocfs2_alloc_inode(struct super_block *sb) | |
355 | { | |
356 | struct ocfs2_inode_info *oi; | |
357 | ||
e6b4f8da | 358 | oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS); |
ccd979bd MF |
359 | if (!oi) |
360 | return NULL; | |
361 | ||
2b4e30fb | 362 | jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode); |
ccd979bd MF |
363 | return &oi->vfs_inode; |
364 | } | |
365 | ||
366 | static void ocfs2_destroy_inode(struct inode *inode) | |
367 | { | |
368 | kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); | |
369 | } | |
370 | ||
5a254031 MF |
371 | static unsigned long long ocfs2_max_file_offset(unsigned int bbits, |
372 | unsigned int cbits) | |
ccd979bd | 373 | { |
5a254031 MF |
374 | unsigned int bytes = 1 << cbits; |
375 | unsigned int trim = bytes; | |
376 | unsigned int bitshift = 32; | |
377 | ||
378 | /* | |
379 | * i_size and all block offsets in ocfs2 are always 64 bits | |
380 | * wide. i_clusters is 32 bits, in cluster-sized units. So on | |
381 | * 64 bit platforms, cluster size will be the limiting factor. | |
ccd979bd MF |
382 | */ |
383 | ||
384 | #if BITS_PER_LONG == 32 | |
385 | # if defined(CONFIG_LBD) | |
2ecd05ae | 386 | BUILD_BUG_ON(sizeof(sector_t) != 8); |
5a254031 MF |
387 | /* |
388 | * We might be limited by page cache size. | |
389 | */ | |
390 | if (bytes > PAGE_CACHE_SIZE) { | |
391 | bytes = PAGE_CACHE_SIZE; | |
392 | trim = 1; | |
393 | /* | |
394 | * Shift by 31 here so that we don't get larger than | |
395 | * MAX_LFS_FILESIZE | |
396 | */ | |
397 | bitshift = 31; | |
398 | } | |
ccd979bd | 399 | # else |
5a254031 MF |
400 | /* |
401 | * We are limited by the size of sector_t. Use block size, as | |
402 | * that's what we expose to the VFS. | |
403 | */ | |
404 | bytes = 1 << bbits; | |
405 | trim = 1; | |
406 | bitshift = 31; | |
ccd979bd MF |
407 | # endif |
408 | #endif | |
409 | ||
5a254031 MF |
410 | /* |
411 | * Trim by a whole cluster when we can actually approach the | |
412 | * on-disk limits. Otherwise we can overflow i_clusters when | |
413 | * an extent start is at the max offset. | |
414 | */ | |
415 | return (((unsigned long long)bytes) << bitshift) - trim; | |
ccd979bd MF |
416 | } |
417 | ||
418 | static int ocfs2_remount(struct super_block *sb, int *flags, char *data) | |
419 | { | |
420 | int incompat_features; | |
421 | int ret = 0; | |
c0123ade | 422 | struct mount_options parsed_options; |
ccd979bd MF |
423 | struct ocfs2_super *osb = OCFS2_SB(sb); |
424 | ||
c0123ade | 425 | if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { |
ccd979bd MF |
426 | ret = -EINVAL; |
427 | goto out; | |
428 | } | |
429 | ||
430 | if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != | |
c0123ade | 431 | (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { |
ccd979bd MF |
432 | ret = -EINVAL; |
433 | mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); | |
434 | goto out; | |
435 | } | |
436 | ||
437 | if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != | |
c0123ade | 438 | (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { |
ccd979bd MF |
439 | ret = -EINVAL; |
440 | mlog(ML_ERROR, "Cannot change data mode on remount\n"); | |
441 | goto out; | |
442 | } | |
443 | ||
12462f1d JB |
444 | /* Probably don't want this on remount; it might |
445 | * mess with other nodes */ | |
446 | if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) && | |
447 | (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) { | |
448 | ret = -EINVAL; | |
449 | mlog(ML_ERROR, "Cannot enable inode64 on remount\n"); | |
450 | goto out; | |
451 | } | |
452 | ||
ccd979bd MF |
453 | /* We're going to/from readonly mode. */ |
454 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { | |
455 | /* Lock here so the check of HARD_RO and the potential | |
456 | * setting of SOFT_RO is atomic. */ | |
457 | spin_lock(&osb->osb_lock); | |
458 | if (osb->osb_flags & OCFS2_OSB_HARD_RO) { | |
459 | mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); | |
460 | ret = -EROFS; | |
461 | goto unlock_osb; | |
462 | } | |
463 | ||
464 | if (*flags & MS_RDONLY) { | |
465 | mlog(0, "Going to ro mode.\n"); | |
466 | sb->s_flags |= MS_RDONLY; | |
467 | osb->osb_flags |= OCFS2_OSB_SOFT_RO; | |
468 | } else { | |
469 | mlog(0, "Making ro filesystem writeable.\n"); | |
470 | ||
471 | if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { | |
472 | mlog(ML_ERROR, "Cannot remount RDWR " | |
473 | "filesystem due to previous errors.\n"); | |
474 | ret = -EROFS; | |
475 | goto unlock_osb; | |
476 | } | |
477 | incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); | |
478 | if (incompat_features) { | |
479 | mlog(ML_ERROR, "Cannot remount RDWR because " | |
480 | "of unsupported optional features " | |
481 | "(%x).\n", incompat_features); | |
482 | ret = -EINVAL; | |
483 | goto unlock_osb; | |
484 | } | |
485 | sb->s_flags &= ~MS_RDONLY; | |
486 | osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; | |
487 | } | |
488 | unlock_osb: | |
489 | spin_unlock(&osb->osb_lock); | |
490 | } | |
491 | ||
492 | if (!ret) { | |
ccd979bd MF |
493 | /* Only save off the new mount options in case of a successful |
494 | * remount. */ | |
a68979b8 TY |
495 | if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) |
496 | parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; | |
c0123ade TY |
497 | osb->s_mount_opt = parsed_options.mount_opt; |
498 | osb->s_atime_quantum = parsed_options.atime_quantum; | |
499 | osb->preferred_slot = parsed_options.slot; | |
d147b3d6 MF |
500 | if (parsed_options.commit_interval) |
501 | osb->osb_commit_interval = parsed_options.commit_interval; | |
e001e796 MF |
502 | |
503 | if (!ocfs2_is_hard_readonly(osb)) | |
504 | ocfs2_set_journal_params(osb); | |
ccd979bd MF |
505 | } |
506 | out: | |
507 | return ret; | |
508 | } | |
509 | ||
510 | static int ocfs2_sb_probe(struct super_block *sb, | |
511 | struct buffer_head **bh, | |
512 | int *sector_size) | |
513 | { | |
bddb8eb3 | 514 | int status, tmpstat; |
ccd979bd MF |
515 | struct ocfs1_vol_disk_hdr *hdr; |
516 | struct ocfs2_dinode *di; | |
517 | int blksize; | |
518 | ||
519 | *bh = NULL; | |
520 | ||
521 | /* may be > 512 */ | |
522 | *sector_size = bdev_hardsect_size(sb->s_bdev); | |
523 | if (*sector_size > OCFS2_MAX_BLOCKSIZE) { | |
524 | mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", | |
525 | *sector_size, OCFS2_MAX_BLOCKSIZE); | |
526 | status = -EINVAL; | |
527 | goto bail; | |
528 | } | |
529 | ||
530 | /* Can this really happen? */ | |
531 | if (*sector_size < OCFS2_MIN_BLOCKSIZE) | |
532 | *sector_size = OCFS2_MIN_BLOCKSIZE; | |
533 | ||
534 | /* check block zero for old format */ | |
535 | status = ocfs2_get_sector(sb, bh, 0, *sector_size); | |
536 | if (status < 0) { | |
537 | mlog_errno(status); | |
538 | goto bail; | |
539 | } | |
540 | hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; | |
541 | if (hdr->major_version == OCFS1_MAJOR_VERSION) { | |
542 | mlog(ML_ERROR, "incompatible version: %u.%u\n", | |
543 | hdr->major_version, hdr->minor_version); | |
544 | status = -EINVAL; | |
545 | } | |
546 | if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, | |
547 | strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { | |
548 | mlog(ML_ERROR, "incompatible volume signature: %8s\n", | |
549 | hdr->signature); | |
550 | status = -EINVAL; | |
551 | } | |
552 | brelse(*bh); | |
553 | *bh = NULL; | |
554 | if (status < 0) { | |
555 | mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " | |
556 | "upgraded before mounting with ocfs v2\n"); | |
557 | goto bail; | |
558 | } | |
559 | ||
560 | /* | |
561 | * Now check at magic offset for 512, 1024, 2048, 4096 | |
562 | * blocksizes. 4096 is the maximum blocksize because it is | |
563 | * the minimum clustersize. | |
564 | */ | |
565 | status = -EINVAL; | |
566 | for (blksize = *sector_size; | |
567 | blksize <= OCFS2_MAX_BLOCKSIZE; | |
568 | blksize <<= 1) { | |
569 | tmpstat = ocfs2_get_sector(sb, bh, | |
570 | OCFS2_SUPER_BLOCK_BLKNO, | |
571 | blksize); | |
572 | if (tmpstat < 0) { | |
573 | status = tmpstat; | |
574 | mlog_errno(status); | |
575 | goto bail; | |
576 | } | |
577 | di = (struct ocfs2_dinode *) (*bh)->b_data; | |
578 | status = ocfs2_verify_volume(di, *bh, blksize); | |
579 | if (status >= 0) | |
580 | goto bail; | |
581 | brelse(*bh); | |
582 | *bh = NULL; | |
583 | if (status != -EAGAIN) | |
584 | break; | |
585 | } | |
586 | ||
587 | bail: | |
588 | return status; | |
589 | } | |
590 | ||
c271c5c2 SM |
591 | static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) |
592 | { | |
593 | if (ocfs2_mount_local(osb)) { | |
594 | if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { | |
595 | mlog(ML_ERROR, "Cannot heartbeat on a locally " | |
596 | "mounted device.\n"); | |
597 | return -EINVAL; | |
598 | } | |
599 | } | |
600 | ||
b61817e1 JB |
601 | if (ocfs2_userspace_stack(osb)) { |
602 | if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { | |
603 | mlog(ML_ERROR, "Userspace stack expected, but " | |
604 | "o2cb heartbeat arguments passed to mount\n"); | |
605 | return -EINVAL; | |
606 | } | |
607 | } | |
608 | ||
c271c5c2 | 609 | if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { |
b61817e1 JB |
610 | if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) && |
611 | !ocfs2_userspace_stack(osb)) { | |
c271c5c2 SM |
612 | mlog(ML_ERROR, "Heartbeat has to be started to mount " |
613 | "a read-write clustered device.\n"); | |
614 | return -EINVAL; | |
615 | } | |
616 | } | |
617 | ||
618 | return 0; | |
619 | } | |
620 | ||
b61817e1 JB |
621 | /* |
622 | * If we're using a userspace stack, mount should have passed | |
623 | * a name that matches the disk. If not, mount should not | |
624 | * have passed a stack. | |
625 | */ | |
626 | static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb, | |
627 | struct mount_options *mopt) | |
628 | { | |
629 | if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) { | |
630 | mlog(ML_ERROR, | |
631 | "cluster stack passed to mount, but this filesystem " | |
632 | "does not support it\n"); | |
633 | return -EINVAL; | |
634 | } | |
635 | ||
636 | if (ocfs2_userspace_stack(osb) && | |
637 | strncmp(osb->osb_cluster_stack, mopt->cluster_stack, | |
638 | OCFS2_STACK_LABEL_LEN)) { | |
639 | mlog(ML_ERROR, | |
640 | "cluster stack passed to mount (\"%s\") does not " | |
641 | "match the filesystem (\"%s\")\n", | |
642 | mopt->cluster_stack, | |
643 | osb->osb_cluster_stack); | |
644 | return -EINVAL; | |
645 | } | |
646 | ||
647 | return 0; | |
648 | } | |
649 | ||
ccd979bd MF |
650 | static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) |
651 | { | |
652 | struct dentry *root; | |
653 | int status, sector_size; | |
c0123ade | 654 | struct mount_options parsed_options; |
ccd979bd MF |
655 | struct inode *inode = NULL; |
656 | struct ocfs2_super *osb = NULL; | |
657 | struct buffer_head *bh = NULL; | |
c271c5c2 | 658 | char nodestr[8]; |
ccd979bd MF |
659 | |
660 | mlog_entry("%p, %p, %i", sb, data, silent); | |
661 | ||
c0123ade | 662 | if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { |
ccd979bd MF |
663 | status = -EINVAL; |
664 | goto read_super_error; | |
665 | } | |
666 | ||
667 | /* probe for superblock */ | |
668 | status = ocfs2_sb_probe(sb, &bh, §or_size); | |
669 | if (status < 0) { | |
670 | mlog(ML_ERROR, "superblock probe failed!\n"); | |
671 | goto read_super_error; | |
672 | } | |
673 | ||
674 | status = ocfs2_initialize_super(sb, bh, sector_size); | |
675 | osb = OCFS2_SB(sb); | |
676 | if (status < 0) { | |
677 | mlog_errno(status); | |
678 | goto read_super_error; | |
679 | } | |
680 | brelse(bh); | |
681 | bh = NULL; | |
a68979b8 TY |
682 | |
683 | if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR)) | |
684 | parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; | |
685 | ||
c0123ade TY |
686 | osb->s_mount_opt = parsed_options.mount_opt; |
687 | osb->s_atime_quantum = parsed_options.atime_quantum; | |
688 | osb->preferred_slot = parsed_options.slot; | |
d147b3d6 | 689 | osb->osb_commit_interval = parsed_options.commit_interval; |
9c7af40b MF |
690 | osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt); |
691 | osb->local_alloc_bits = osb->local_alloc_default_bits; | |
ccd979bd | 692 | |
b61817e1 JB |
693 | status = ocfs2_verify_userspace_stack(osb, &parsed_options); |
694 | if (status) | |
695 | goto read_super_error; | |
696 | ||
ccd979bd MF |
697 | sb->s_magic = OCFS2_SUPER_MAGIC; |
698 | ||
a68979b8 TY |
699 | sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | |
700 | ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); | |
701 | ||
ccd979bd MF |
702 | /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, |
703 | * heartbeat=none */ | |
704 | if (bdev_read_only(sb->s_bdev)) { | |
705 | if (!(sb->s_flags & MS_RDONLY)) { | |
706 | status = -EACCES; | |
707 | mlog(ML_ERROR, "Readonly device detected but readonly " | |
708 | "mount was not specified.\n"); | |
709 | goto read_super_error; | |
710 | } | |
711 | ||
712 | /* You should not be able to start a local heartbeat | |
713 | * on a readonly device. */ | |
714 | if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { | |
715 | status = -EROFS; | |
716 | mlog(ML_ERROR, "Local heartbeat specified on readonly " | |
717 | "device.\n"); | |
718 | goto read_super_error; | |
719 | } | |
720 | ||
721 | status = ocfs2_check_journals_nolocks(osb); | |
722 | if (status < 0) { | |
723 | if (status == -EROFS) | |
724 | mlog(ML_ERROR, "Recovery required on readonly " | |
725 | "file system, but write access is " | |
726 | "unavailable.\n"); | |
727 | else | |
728 | mlog_errno(status); | |
729 | goto read_super_error; | |
730 | } | |
731 | ||
732 | ocfs2_set_ro_flag(osb, 1); | |
733 | ||
734 | printk(KERN_NOTICE "Readonly device detected. No cluster " | |
735 | "services will be utilized for this mount. Recovery " | |
736 | "will be skipped.\n"); | |
737 | } | |
738 | ||
739 | if (!ocfs2_is_hard_readonly(osb)) { | |
ccd979bd MF |
740 | if (sb->s_flags & MS_RDONLY) |
741 | ocfs2_set_ro_flag(osb, 0); | |
742 | } | |
743 | ||
c271c5c2 SM |
744 | status = ocfs2_verify_heartbeat(osb); |
745 | if (status < 0) { | |
746 | mlog_errno(status); | |
747 | goto read_super_error; | |
748 | } | |
749 | ||
ccd979bd MF |
750 | osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, |
751 | ocfs2_debugfs_root); | |
752 | if (!osb->osb_debug_root) { | |
753 | status = -EINVAL; | |
754 | mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); | |
755 | goto read_super_error; | |
756 | } | |
757 | ||
758 | status = ocfs2_mount_volume(sb); | |
759 | if (osb->root_inode) | |
760 | inode = igrab(osb->root_inode); | |
761 | ||
762 | if (status < 0) | |
763 | goto read_super_error; | |
764 | ||
765 | if (!inode) { | |
766 | status = -EIO; | |
767 | mlog_errno(status); | |
768 | goto read_super_error; | |
769 | } | |
770 | ||
771 | root = d_alloc_root(inode); | |
772 | if (!root) { | |
773 | status = -ENOMEM; | |
774 | mlog_errno(status); | |
775 | goto read_super_error; | |
776 | } | |
777 | ||
778 | sb->s_root = root; | |
779 | ||
780 | ocfs2_complete_mount_recovery(osb); | |
781 | ||
c271c5c2 SM |
782 | if (ocfs2_mount_local(osb)) |
783 | snprintf(nodestr, sizeof(nodestr), "local"); | |
784 | else | |
19fdb624 | 785 | snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); |
c271c5c2 SM |
786 | |
787 | printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " | |
781ee3e2 | 788 | "with %s data mode.\n", |
c271c5c2 | 789 | osb->dev_str, nodestr, osb->slot_num, |
ccd979bd MF |
790 | osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : |
791 | "ordered"); | |
792 | ||
793 | atomic_set(&osb->vol_state, VOLUME_MOUNTED); | |
794 | wake_up(&osb->osb_mount_event); | |
795 | ||
796 | mlog_exit(status); | |
797 | return status; | |
798 | ||
799 | read_super_error: | |
a81cb88b | 800 | brelse(bh); |
ccd979bd MF |
801 | |
802 | if (inode) | |
803 | iput(inode); | |
804 | ||
805 | if (osb) { | |
806 | atomic_set(&osb->vol_state, VOLUME_DISABLED); | |
807 | wake_up(&osb->osb_mount_event); | |
808 | ocfs2_dismount_volume(sb, 1); | |
809 | } | |
810 | ||
811 | mlog_exit(status); | |
812 | return status; | |
813 | } | |
814 | ||
454e2398 DH |
815 | static int ocfs2_get_sb(struct file_system_type *fs_type, |
816 | int flags, | |
817 | const char *dev_name, | |
818 | void *data, | |
819 | struct vfsmount *mnt) | |
ccd979bd | 820 | { |
454e2398 DH |
821 | return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, |
822 | mnt); | |
ccd979bd MF |
823 | } |
824 | ||
825 | static struct file_system_type ocfs2_fs_type = { | |
826 | .owner = THIS_MODULE, | |
827 | .name = "ocfs2", | |
828 | .get_sb = ocfs2_get_sb, /* is this called when we mount | |
829 | * the fs? */ | |
830 | .kill_sb = kill_block_super, /* set to the generic one | |
831 | * right now, but do we | |
832 | * need to change that? */ | |
1ba9da2f | 833 | .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, |
ccd979bd MF |
834 | .next = NULL |
835 | }; | |
836 | ||
837 | static int ocfs2_parse_options(struct super_block *sb, | |
838 | char *options, | |
c0123ade | 839 | struct mount_options *mopt, |
ccd979bd MF |
840 | int is_remount) |
841 | { | |
842 | int status; | |
843 | char *p; | |
844 | ||
845 | mlog_entry("remount: %d, options: \"%s\"\n", is_remount, | |
846 | options ? options : "(none)"); | |
847 | ||
d147b3d6 | 848 | mopt->commit_interval = 0; |
c0123ade TY |
849 | mopt->mount_opt = 0; |
850 | mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; | |
851 | mopt->slot = OCFS2_INVALID_SLOT; | |
2fbe8d1e | 852 | mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE; |
b61817e1 | 853 | mopt->cluster_stack[0] = '\0'; |
ccd979bd MF |
854 | |
855 | if (!options) { | |
856 | status = 1; | |
857 | goto bail; | |
858 | } | |
859 | ||
860 | while ((p = strsep(&options, ",")) != NULL) { | |
861 | int token, option; | |
862 | substring_t args[MAX_OPT_ARGS]; | |
863 | ||
864 | if (!*p) | |
865 | continue; | |
866 | ||
867 | token = match_token(p, tokens, args); | |
868 | switch (token) { | |
869 | case Opt_hb_local: | |
c0123ade | 870 | mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; |
ccd979bd MF |
871 | break; |
872 | case Opt_hb_none: | |
c0123ade | 873 | mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; |
ccd979bd MF |
874 | break; |
875 | case Opt_barrier: | |
876 | if (match_int(&args[0], &option)) { | |
877 | status = 0; | |
878 | goto bail; | |
879 | } | |
880 | if (option) | |
c0123ade | 881 | mopt->mount_opt |= OCFS2_MOUNT_BARRIER; |
ccd979bd | 882 | else |
c0123ade | 883 | mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; |
ccd979bd MF |
884 | break; |
885 | case Opt_intr: | |
c0123ade | 886 | mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; |
ccd979bd MF |
887 | break; |
888 | case Opt_nointr: | |
c0123ade | 889 | mopt->mount_opt |= OCFS2_MOUNT_NOINTR; |
ccd979bd MF |
890 | break; |
891 | case Opt_err_panic: | |
c0123ade | 892 | mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; |
ccd979bd MF |
893 | break; |
894 | case Opt_err_ro: | |
c0123ade | 895 | mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; |
ccd979bd MF |
896 | break; |
897 | case Opt_data_ordered: | |
c0123ade | 898 | mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; |
ccd979bd MF |
899 | break; |
900 | case Opt_data_writeback: | |
c0123ade | 901 | mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; |
ccd979bd | 902 | break; |
cf1d6c76 TY |
903 | case Opt_user_xattr: |
904 | mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR; | |
905 | break; | |
906 | case Opt_nouser_xattr: | |
907 | mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR; | |
908 | break; | |
7f1a37e3 TY |
909 | case Opt_atime_quantum: |
910 | if (match_int(&args[0], &option)) { | |
911 | status = 0; | |
912 | goto bail; | |
913 | } | |
914 | if (option >= 0) | |
c0123ade | 915 | mopt->atime_quantum = option; |
7f1a37e3 | 916 | break; |
baf4661a SM |
917 | case Opt_slot: |
918 | option = 0; | |
919 | if (match_int(&args[0], &option)) { | |
920 | status = 0; | |
921 | goto bail; | |
922 | } | |
923 | if (option) | |
c0123ade | 924 | mopt->slot = (s16)option; |
baf4661a | 925 | break; |
d147b3d6 MF |
926 | case Opt_commit: |
927 | option = 0; | |
928 | if (match_int(&args[0], &option)) { | |
929 | status = 0; | |
930 | goto bail; | |
931 | } | |
932 | if (option < 0) | |
933 | return 0; | |
934 | if (option == 0) | |
2b4e30fb | 935 | option = JBD2_DEFAULT_MAX_COMMIT_AGE; |
d147b3d6 MF |
936 | mopt->commit_interval = HZ * option; |
937 | break; | |
2fbe8d1e SM |
938 | case Opt_localalloc: |
939 | option = 0; | |
940 | if (match_int(&args[0], &option)) { | |
941 | status = 0; | |
942 | goto bail; | |
943 | } | |
944 | if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8)) | |
945 | mopt->localalloc_opt = option; | |
946 | break; | |
53fc622b MF |
947 | case Opt_localflocks: |
948 | /* | |
949 | * Changing this during remount could race | |
950 | * flock() requests, or "unbalance" existing | |
951 | * ones (e.g., a lock is taken in one mode but | |
952 | * dropped in the other). If users care enough | |
953 | * to flip locking modes during remount, we | |
954 | * could add a "local" flag to individual | |
955 | * flock structures for proper tracking of | |
956 | * state. | |
957 | */ | |
958 | if (!is_remount) | |
959 | mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS; | |
960 | break; | |
b61817e1 JB |
961 | case Opt_stack: |
962 | /* Check both that the option we were passed | |
963 | * is of the right length and that it is a proper | |
964 | * string of the right length. | |
965 | */ | |
966 | if (((args[0].to - args[0].from) != | |
967 | OCFS2_STACK_LABEL_LEN) || | |
968 | (strnlen(args[0].from, | |
969 | OCFS2_STACK_LABEL_LEN) != | |
970 | OCFS2_STACK_LABEL_LEN)) { | |
971 | mlog(ML_ERROR, | |
972 | "Invalid cluster_stack option\n"); | |
973 | status = 0; | |
974 | goto bail; | |
975 | } | |
976 | memcpy(mopt->cluster_stack, args[0].from, | |
977 | OCFS2_STACK_LABEL_LEN); | |
978 | mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; | |
979 | break; | |
12462f1d JB |
980 | case Opt_inode64: |
981 | mopt->mount_opt |= OCFS2_MOUNT_INODE64; | |
982 | break; | |
a68979b8 TY |
983 | #ifdef CONFIG_OCFS2_FS_POSIX_ACL |
984 | case Opt_acl: | |
985 | mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL; | |
986 | break; | |
987 | case Opt_noacl: | |
988 | mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL; | |
989 | break; | |
990 | #else | |
991 | case Opt_acl: | |
992 | case Opt_noacl: | |
993 | printk(KERN_INFO "ocfs2 (no)acl options not supported\n"); | |
994 | break; | |
995 | #endif | |
ccd979bd MF |
996 | default: |
997 | mlog(ML_ERROR, | |
998 | "Unrecognized mount option \"%s\" " | |
999 | "or missing value\n", p); | |
1000 | status = 0; | |
1001 | goto bail; | |
1002 | } | |
1003 | } | |
1004 | ||
1005 | status = 1; | |
1006 | ||
1007 | bail: | |
1008 | mlog_exit(status); | |
1009 | return status; | |
1010 | } | |
1011 | ||
d550071c SM |
1012 | static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) |
1013 | { | |
1014 | struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); | |
1015 | unsigned long opts = osb->s_mount_opt; | |
ebcee4b5 | 1016 | unsigned int local_alloc_megs; |
d550071c SM |
1017 | |
1018 | if (opts & OCFS2_MOUNT_HB_LOCAL) | |
1019 | seq_printf(s, ",_netdev,heartbeat=local"); | |
1020 | else | |
1021 | seq_printf(s, ",heartbeat=none"); | |
1022 | ||
1023 | if (opts & OCFS2_MOUNT_NOINTR) | |
1024 | seq_printf(s, ",nointr"); | |
1025 | ||
1026 | if (opts & OCFS2_MOUNT_DATA_WRITEBACK) | |
1027 | seq_printf(s, ",data=writeback"); | |
1028 | else | |
1029 | seq_printf(s, ",data=ordered"); | |
1030 | ||
1031 | if (opts & OCFS2_MOUNT_BARRIER) | |
1032 | seq_printf(s, ",barrier=1"); | |
1033 | ||
1034 | if (opts & OCFS2_MOUNT_ERRORS_PANIC) | |
1035 | seq_printf(s, ",errors=panic"); | |
1036 | else | |
1037 | seq_printf(s, ",errors=remount-ro"); | |
1038 | ||
1039 | if (osb->preferred_slot != OCFS2_INVALID_SLOT) | |
1040 | seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); | |
1041 | ||
1042 | if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) | |
1043 | seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); | |
1044 | ||
d147b3d6 MF |
1045 | if (osb->osb_commit_interval) |
1046 | seq_printf(s, ",commit=%u", | |
1047 | (unsigned) (osb->osb_commit_interval / HZ)); | |
1048 | ||
ebcee4b5 MF |
1049 | local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits); |
1050 | if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE) | |
1051 | seq_printf(s, ",localalloc=%d", local_alloc_megs); | |
2fbe8d1e | 1052 | |
53fc622b MF |
1053 | if (opts & OCFS2_MOUNT_LOCALFLOCKS) |
1054 | seq_printf(s, ",localflocks,"); | |
1055 | ||
b61817e1 JB |
1056 | if (osb->osb_cluster_stack[0]) |
1057 | seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN, | |
1058 | osb->osb_cluster_stack); | |
1059 | ||
b0f73cfc SM |
1060 | if (opts & OCFS2_MOUNT_NOUSERXATTR) |
1061 | seq_printf(s, ",nouser_xattr"); | |
1062 | else | |
1063 | seq_printf(s, ",user_xattr"); | |
1064 | ||
12462f1d JB |
1065 | if (opts & OCFS2_MOUNT_INODE64) |
1066 | seq_printf(s, ",inode64"); | |
1067 | ||
a68979b8 TY |
1068 | #ifdef CONFIG_OCFS2_FS_POSIX_ACL |
1069 | if (opts & OCFS2_MOUNT_POSIX_ACL) | |
1070 | seq_printf(s, ",acl"); | |
1071 | else | |
1072 | seq_printf(s, ",noacl"); | |
1073 | #endif | |
1074 | ||
d550071c SM |
1075 | return 0; |
1076 | } | |
1077 | ||
ccd979bd MF |
1078 | static int __init ocfs2_init(void) |
1079 | { | |
1080 | int status; | |
1081 | ||
1082 | mlog_entry_void(); | |
1083 | ||
1084 | ocfs2_print_version(); | |
1085 | ||
ccd979bd MF |
1086 | status = init_ocfs2_uptodate_cache(); |
1087 | if (status < 0) { | |
1088 | mlog_errno(status); | |
1089 | goto leave; | |
1090 | } | |
1091 | ||
1092 | status = ocfs2_initialize_mem_caches(); | |
1093 | if (status < 0) { | |
1094 | mlog_errno(status); | |
1095 | goto leave; | |
1096 | } | |
1097 | ||
1098 | ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); | |
1099 | if (!ocfs2_wq) { | |
1100 | status = -ENOMEM; | |
1101 | goto leave; | |
1102 | } | |
1103 | ||
ccd979bd MF |
1104 | ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); |
1105 | if (!ocfs2_debugfs_root) { | |
1106 | status = -EFAULT; | |
1107 | mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); | |
1108 | } | |
1109 | ||
171bf93c MF |
1110 | status = ocfs2_quota_setup(); |
1111 | if (status) | |
1112 | goto leave; | |
1113 | ||
63e0c48a JB |
1114 | ocfs2_set_locking_protocol(); |
1115 | ||
9e33d69f | 1116 | status = register_quota_format(&ocfs2_quota_format); |
ccd979bd MF |
1117 | leave: |
1118 | if (status < 0) { | |
171bf93c | 1119 | ocfs2_quota_shutdown(); |
ccd979bd MF |
1120 | ocfs2_free_mem_caches(); |
1121 | exit_ocfs2_uptodate_cache(); | |
ccd979bd MF |
1122 | } |
1123 | ||
1124 | mlog_exit(status); | |
1125 | ||
1126 | if (status >= 0) { | |
1127 | return register_filesystem(&ocfs2_fs_type); | |
1128 | } else | |
1129 | return -1; | |
1130 | } | |
1131 | ||
1132 | static void __exit ocfs2_exit(void) | |
1133 | { | |
1134 | mlog_entry_void(); | |
1135 | ||
171bf93c MF |
1136 | ocfs2_quota_shutdown(); |
1137 | ||
ccd979bd MF |
1138 | if (ocfs2_wq) { |
1139 | flush_workqueue(ocfs2_wq); | |
1140 | destroy_workqueue(ocfs2_wq); | |
1141 | } | |
1142 | ||
9e33d69f JK |
1143 | unregister_quota_format(&ocfs2_quota_format); |
1144 | ||
ccd979bd MF |
1145 | debugfs_remove(ocfs2_debugfs_root); |
1146 | ||
1147 | ocfs2_free_mem_caches(); | |
1148 | ||
1149 | unregister_filesystem(&ocfs2_fs_type); | |
1150 | ||
ccd979bd MF |
1151 | exit_ocfs2_uptodate_cache(); |
1152 | ||
1153 | mlog_exit_void(); | |
1154 | } | |
1155 | ||
1156 | static void ocfs2_put_super(struct super_block *sb) | |
1157 | { | |
1158 | mlog_entry("(0x%p)\n", sb); | |
1159 | ||
1160 | ocfs2_sync_blockdev(sb); | |
1161 | ocfs2_dismount_volume(sb, 0); | |
1162 | ||
1163 | mlog_exit_void(); | |
1164 | } | |
1165 | ||
726c3342 | 1166 | static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) |
ccd979bd MF |
1167 | { |
1168 | struct ocfs2_super *osb; | |
1169 | u32 numbits, freebits; | |
1170 | int status; | |
1171 | struct ocfs2_dinode *bm_lock; | |
1172 | struct buffer_head *bh = NULL; | |
1173 | struct inode *inode = NULL; | |
1174 | ||
726c3342 | 1175 | mlog_entry("(%p, %p)\n", dentry->d_sb, buf); |
ccd979bd | 1176 | |
726c3342 | 1177 | osb = OCFS2_SB(dentry->d_sb); |
ccd979bd MF |
1178 | |
1179 | inode = ocfs2_get_system_file_inode(osb, | |
1180 | GLOBAL_BITMAP_SYSTEM_INODE, | |
1181 | OCFS2_INVALID_SLOT); | |
1182 | if (!inode) { | |
1183 | mlog(ML_ERROR, "failed to get bitmap inode\n"); | |
1184 | status = -EIO; | |
1185 | goto bail; | |
1186 | } | |
1187 | ||
e63aecb6 | 1188 | status = ocfs2_inode_lock(inode, &bh, 0); |
ccd979bd MF |
1189 | if (status < 0) { |
1190 | mlog_errno(status); | |
1191 | goto bail; | |
1192 | } | |
1193 | ||
1194 | bm_lock = (struct ocfs2_dinode *) bh->b_data; | |
1195 | ||
1196 | numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); | |
1197 | freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); | |
1198 | ||
1199 | buf->f_type = OCFS2_SUPER_MAGIC; | |
726c3342 | 1200 | buf->f_bsize = dentry->d_sb->s_blocksize; |
ccd979bd MF |
1201 | buf->f_namelen = OCFS2_MAX_FILENAME_LEN; |
1202 | buf->f_blocks = ((sector_t) numbits) * | |
1203 | (osb->s_clustersize >> osb->sb->s_blocksize_bits); | |
1204 | buf->f_bfree = ((sector_t) freebits) * | |
1205 | (osb->s_clustersize >> osb->sb->s_blocksize_bits); | |
1206 | buf->f_bavail = buf->f_bfree; | |
1207 | buf->f_files = numbits; | |
1208 | buf->f_ffree = freebits; | |
1209 | ||
1210 | brelse(bh); | |
1211 | ||
e63aecb6 | 1212 | ocfs2_inode_unlock(inode, 0); |
ccd979bd MF |
1213 | status = 0; |
1214 | bail: | |
1215 | if (inode) | |
1216 | iput(inode); | |
1217 | ||
1218 | mlog_exit(status); | |
1219 | ||
1220 | return status; | |
1221 | } | |
1222 | ||
51cc5068 | 1223 | static void ocfs2_inode_init_once(void *data) |
ccd979bd MF |
1224 | { |
1225 | struct ocfs2_inode_info *oi = data; | |
1226 | ||
a35afb83 CL |
1227 | oi->ip_flags = 0; |
1228 | oi->ip_open_count = 0; | |
1229 | spin_lock_init(&oi->ip_lock); | |
1230 | ocfs2_extent_map_init(&oi->vfs_inode); | |
1231 | INIT_LIST_HEAD(&oi->ip_io_markers); | |
1232 | oi->ip_created_trans = 0; | |
1233 | oi->ip_last_trans = 0; | |
1234 | oi->ip_dir_start_lookup = 0; | |
ccd979bd | 1235 | |
a35afb83 | 1236 | init_rwsem(&oi->ip_alloc_sem); |
cf1d6c76 | 1237 | init_rwsem(&oi->ip_xattr_sem); |
a35afb83 | 1238 | mutex_init(&oi->ip_io_mutex); |
ccd979bd | 1239 | |
a35afb83 CL |
1240 | oi->ip_blkno = 0ULL; |
1241 | oi->ip_clusters = 0; | |
ccd979bd | 1242 | |
a35afb83 | 1243 | ocfs2_lock_res_init_once(&oi->ip_rw_lockres); |
e63aecb6 | 1244 | ocfs2_lock_res_init_once(&oi->ip_inode_lockres); |
a35afb83 | 1245 | ocfs2_lock_res_init_once(&oi->ip_open_lockres); |
ccd979bd | 1246 | |
a35afb83 | 1247 | ocfs2_metadata_cache_init(&oi->vfs_inode); |
ccd979bd | 1248 | |
a35afb83 | 1249 | inode_init_once(&oi->vfs_inode); |
ccd979bd MF |
1250 | } |
1251 | ||
1252 | static int ocfs2_initialize_mem_caches(void) | |
1253 | { | |
1254 | ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", | |
fffb60f9 PJ |
1255 | sizeof(struct ocfs2_inode_info), |
1256 | 0, | |
1257 | (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| | |
1258 | SLAB_MEM_SPREAD), | |
20c2df83 | 1259 | ocfs2_inode_init_once); |
9e33d69f JK |
1260 | ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache", |
1261 | sizeof(struct ocfs2_dquot), | |
1262 | 0, | |
1263 | (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| | |
1264 | SLAB_MEM_SPREAD), | |
1265 | NULL); | |
1266 | ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache", | |
1267 | sizeof(struct ocfs2_quota_chunk), | |
1268 | 0, | |
1269 | (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD), | |
1270 | NULL); | |
1271 | if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep || | |
1272 | !ocfs2_qf_chunk_cachep) { | |
1273 | if (ocfs2_inode_cachep) | |
1274 | kmem_cache_destroy(ocfs2_inode_cachep); | |
1275 | if (ocfs2_dquot_cachep) | |
1276 | kmem_cache_destroy(ocfs2_dquot_cachep); | |
1277 | if (ocfs2_qf_chunk_cachep) | |
1278 | kmem_cache_destroy(ocfs2_qf_chunk_cachep); | |
ccd979bd | 1279 | return -ENOMEM; |
9e33d69f | 1280 | } |
ccd979bd | 1281 | |
ccd979bd MF |
1282 | return 0; |
1283 | } | |
1284 | ||
1285 | static void ocfs2_free_mem_caches(void) | |
1286 | { | |
1287 | if (ocfs2_inode_cachep) | |
1288 | kmem_cache_destroy(ocfs2_inode_cachep); | |
ccd979bd | 1289 | ocfs2_inode_cachep = NULL; |
9e33d69f JK |
1290 | |
1291 | if (ocfs2_dquot_cachep) | |
1292 | kmem_cache_destroy(ocfs2_dquot_cachep); | |
1293 | ocfs2_dquot_cachep = NULL; | |
1294 | ||
1295 | if (ocfs2_qf_chunk_cachep) | |
1296 | kmem_cache_destroy(ocfs2_qf_chunk_cachep); | |
1297 | ocfs2_qf_chunk_cachep = NULL; | |
ccd979bd MF |
1298 | } |
1299 | ||
1300 | static int ocfs2_get_sector(struct super_block *sb, | |
1301 | struct buffer_head **bh, | |
1302 | int block, | |
1303 | int sect_size) | |
1304 | { | |
1305 | if (!sb_set_blocksize(sb, sect_size)) { | |
1306 | mlog(ML_ERROR, "unable to set blocksize\n"); | |
1307 | return -EIO; | |
1308 | } | |
1309 | ||
1310 | *bh = sb_getblk(sb, block); | |
1311 | if (!*bh) { | |
1312 | mlog_errno(-EIO); | |
1313 | return -EIO; | |
1314 | } | |
1315 | lock_buffer(*bh); | |
1316 | if (!buffer_dirty(*bh)) | |
1317 | clear_buffer_uptodate(*bh); | |
1318 | unlock_buffer(*bh); | |
1319 | ll_rw_block(READ, 1, bh); | |
1320 | wait_on_buffer(*bh); | |
1321 | return 0; | |
1322 | } | |
1323 | ||
ccd979bd MF |
1324 | static int ocfs2_mount_volume(struct super_block *sb) |
1325 | { | |
1326 | int status = 0; | |
1327 | int unlock_super = 0; | |
1328 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
1329 | ||
1330 | mlog_entry_void(); | |
1331 | ||
1332 | if (ocfs2_is_hard_readonly(osb)) | |
1333 | goto leave; | |
1334 | ||
ccd979bd MF |
1335 | status = ocfs2_dlm_init(osb); |
1336 | if (status < 0) { | |
1337 | mlog_errno(status); | |
1338 | goto leave; | |
1339 | } | |
1340 | ||
ccd979bd MF |
1341 | status = ocfs2_super_lock(osb, 1); |
1342 | if (status < 0) { | |
1343 | mlog_errno(status); | |
1344 | goto leave; | |
1345 | } | |
1346 | unlock_super = 1; | |
1347 | ||
1348 | /* This will load up the node map and add ourselves to it. */ | |
1349 | status = ocfs2_find_slot(osb); | |
1350 | if (status < 0) { | |
1351 | mlog_errno(status); | |
1352 | goto leave; | |
1353 | } | |
1354 | ||
ccd979bd MF |
1355 | /* load all node-local system inodes */ |
1356 | status = ocfs2_init_local_system_inodes(osb); | |
1357 | if (status < 0) { | |
1358 | mlog_errno(status); | |
1359 | goto leave; | |
1360 | } | |
1361 | ||
1362 | status = ocfs2_check_volume(osb); | |
1363 | if (status < 0) { | |
1364 | mlog_errno(status); | |
1365 | goto leave; | |
1366 | } | |
1367 | ||
1368 | status = ocfs2_truncate_log_init(osb); | |
1369 | if (status < 0) { | |
1370 | mlog_errno(status); | |
1371 | goto leave; | |
1372 | } | |
1373 | ||
c271c5c2 SM |
1374 | if (ocfs2_mount_local(osb)) |
1375 | goto leave; | |
1376 | ||
ccd979bd MF |
1377 | leave: |
1378 | if (unlock_super) | |
1379 | ocfs2_super_unlock(osb, 1); | |
1380 | ||
1381 | mlog_exit(status); | |
1382 | return status; | |
1383 | } | |
1384 | ||
ccd979bd MF |
1385 | static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) |
1386 | { | |
286eaa95 | 1387 | int tmp, hangup_needed = 0; |
ccd979bd | 1388 | struct ocfs2_super *osb = NULL; |
c271c5c2 | 1389 | char nodestr[8]; |
ccd979bd MF |
1390 | |
1391 | mlog_entry("(0x%p)\n", sb); | |
1392 | ||
1393 | BUG_ON(!sb); | |
1394 | osb = OCFS2_SB(sb); | |
1395 | BUG_ON(!osb); | |
1396 | ||
1397 | ocfs2_shutdown_local_alloc(osb); | |
1398 | ||
1399 | ocfs2_truncate_log_shutdown(osb); | |
1400 | ||
553abd04 JB |
1401 | /* This will disable recovery and flush any recovery work. */ |
1402 | ocfs2_recovery_exit(osb); | |
ccd979bd MF |
1403 | |
1404 | ocfs2_journal_shutdown(osb); | |
1405 | ||
1406 | ocfs2_sync_blockdev(sb); | |
1407 | ||
4670c46d JB |
1408 | /* No cluster connection means we've failed during mount, so skip |
1409 | * all the steps which depended on that to complete. */ | |
1410 | if (osb->cconn) { | |
ccd979bd MF |
1411 | tmp = ocfs2_super_lock(osb, 1); |
1412 | if (tmp < 0) { | |
1413 | mlog_errno(tmp); | |
1414 | return; | |
1415 | } | |
19b613d4 | 1416 | } |
ccd979bd | 1417 | |
19b613d4 MF |
1418 | if (osb->slot_num != OCFS2_INVALID_SLOT) |
1419 | ocfs2_put_slot(osb); | |
ccd979bd | 1420 | |
4670c46d | 1421 | if (osb->cconn) |
ccd979bd | 1422 | ocfs2_super_unlock(osb, 1); |
ccd979bd MF |
1423 | |
1424 | ocfs2_release_system_inodes(osb); | |
1425 | ||
6953b4c0 | 1426 | /* |
6953b4c0 JB |
1427 | * If we're dismounting due to mount error, mount.ocfs2 will clean |
1428 | * up heartbeat. If we're a local mount, there is no heartbeat. | |
1429 | * If we failed before we got a uuid_str yet, we can't stop | |
1430 | * heartbeat. Otherwise, do it. | |
1431 | */ | |
1432 | if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str) | |
286eaa95 JB |
1433 | hangup_needed = 1; |
1434 | ||
1435 | if (osb->cconn) | |
1436 | ocfs2_dlm_shutdown(osb, hangup_needed); | |
1437 | ||
1438 | debugfs_remove(osb->osb_debug_root); | |
1439 | ||
1440 | if (hangup_needed) | |
6953b4c0 | 1441 | ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str)); |
ccd979bd MF |
1442 | |
1443 | atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); | |
1444 | ||
c271c5c2 SM |
1445 | if (ocfs2_mount_local(osb)) |
1446 | snprintf(nodestr, sizeof(nodestr), "local"); | |
1447 | else | |
19fdb624 | 1448 | snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num); |
c271c5c2 SM |
1449 | |
1450 | printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", | |
1451 | osb->dev_str, nodestr); | |
ccd979bd MF |
1452 | |
1453 | ocfs2_delete_osb(osb); | |
1454 | kfree(osb); | |
1455 | sb->s_dev = 0; | |
1456 | sb->s_fs_info = NULL; | |
1457 | } | |
1458 | ||
1459 | static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, | |
1460 | unsigned uuid_bytes) | |
1461 | { | |
1462 | int i, ret; | |
1463 | char *ptr; | |
1464 | ||
1465 | BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); | |
1466 | ||
cd861280 | 1467 | osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); |
ccd979bd MF |
1468 | if (osb->uuid_str == NULL) |
1469 | return -ENOMEM; | |
1470 | ||
ccd979bd MF |
1471 | for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { |
1472 | /* print with null */ | |
1473 | ret = snprintf(ptr, 3, "%02X", uuid[i]); | |
1474 | if (ret != 2) /* drop super cleans up */ | |
1475 | return -EINVAL; | |
1476 | /* then only advance past the last char */ | |
1477 | ptr += 2; | |
1478 | } | |
1479 | ||
1480 | return 0; | |
1481 | } | |
1482 | ||
1483 | static int ocfs2_initialize_super(struct super_block *sb, | |
1484 | struct buffer_head *bh, | |
1485 | int sector_size) | |
1486 | { | |
bddb8eb3 | 1487 | int status; |
5a254031 MF |
1488 | int i, cbits, bbits; |
1489 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
ccd979bd | 1490 | struct inode *inode = NULL; |
ccd979bd MF |
1491 | struct ocfs2_journal *journal; |
1492 | __le32 uuid_net_key; | |
1493 | struct ocfs2_super *osb; | |
1494 | ||
1495 | mlog_entry_void(); | |
1496 | ||
cd861280 | 1497 | osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); |
ccd979bd MF |
1498 | if (!osb) { |
1499 | status = -ENOMEM; | |
1500 | mlog_errno(status); | |
1501 | goto bail; | |
1502 | } | |
1503 | ||
1504 | sb->s_fs_info = osb; | |
1505 | sb->s_op = &ocfs2_sops; | |
1506 | sb->s_export_op = &ocfs2_export_ops; | |
cf1d6c76 | 1507 | sb->s_xattr = ocfs2_xattr_handlers; |
e0dceaf0 | 1508 | sb->s_time_gran = 1; |
ccd979bd MF |
1509 | sb->s_flags |= MS_NOATIME; |
1510 | /* this is needed to support O_LARGEFILE */ | |
5a254031 MF |
1511 | cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); |
1512 | bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); | |
1513 | sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); | |
ccd979bd MF |
1514 | |
1515 | osb->sb = sb; | |
1516 | /* Save off for ocfs2_rw_direct */ | |
1517 | osb->s_sectsize_bits = blksize_bits(sector_size); | |
ebdec83b | 1518 | BUG_ON(!osb->s_sectsize_bits); |
ccd979bd | 1519 | |
34d024f8 MF |
1520 | spin_lock_init(&osb->dc_task_lock); |
1521 | init_waitqueue_head(&osb->dc_event); | |
1522 | osb->dc_work_sequence = 0; | |
1523 | osb->dc_wake_sequence = 0; | |
ccd979bd MF |
1524 | INIT_LIST_HEAD(&osb->blocked_lock_list); |
1525 | osb->blocked_lock_count = 0; | |
ccd979bd | 1526 | spin_lock_init(&osb->osb_lock); |
4d0ddb2c | 1527 | ocfs2_init_inode_steal_slot(osb); |
ccd979bd MF |
1528 | |
1529 | atomic_set(&osb->alloc_stats.moves, 0); | |
1530 | atomic_set(&osb->alloc_stats.local_data, 0); | |
1531 | atomic_set(&osb->alloc_stats.bitmap_data, 0); | |
1532 | atomic_set(&osb->alloc_stats.bg_allocs, 0); | |
1533 | atomic_set(&osb->alloc_stats.bg_extends, 0); | |
1534 | ||
1535 | ocfs2_init_node_maps(osb); | |
1536 | ||
1537 | snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", | |
1538 | MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); | |
1539 | ||
553abd04 JB |
1540 | status = ocfs2_recovery_init(osb); |
1541 | if (status) { | |
1542 | mlog(ML_ERROR, "Unable to initialize recovery state\n"); | |
1543 | mlog_errno(status); | |
1544 | goto bail; | |
1545 | } | |
ccd979bd MF |
1546 | |
1547 | init_waitqueue_head(&osb->checkpoint_event); | |
1548 | atomic_set(&osb->needs_checkpoint, 0); | |
1549 | ||
7f1a37e3 TY |
1550 | osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; |
1551 | ||
ccd979bd MF |
1552 | osb->slot_num = OCFS2_INVALID_SLOT; |
1553 | ||
8154da3d TY |
1554 | osb->s_xattr_inline_size = le16_to_cpu( |
1555 | di->id2.i_super.s_xattr_inline_size); | |
fdd77704 | 1556 | |
ccd979bd MF |
1557 | osb->local_alloc_state = OCFS2_LA_UNUSED; |
1558 | osb->local_alloc_bh = NULL; | |
9c7af40b | 1559 | INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker); |
ccd979bd | 1560 | |
ccd979bd MF |
1561 | init_waitqueue_head(&osb->osb_mount_event); |
1562 | ||
1563 | osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); | |
1564 | if (!osb->vol_label) { | |
1565 | mlog(ML_ERROR, "unable to alloc vol label\n"); | |
1566 | status = -ENOMEM; | |
1567 | goto bail; | |
1568 | } | |
1569 | ||
ccd979bd MF |
1570 | osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); |
1571 | if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { | |
1572 | mlog(ML_ERROR, "Invalid number of node slots (%u)\n", | |
1573 | osb->max_slots); | |
1574 | status = -EINVAL; | |
1575 | goto bail; | |
1576 | } | |
781ee3e2 | 1577 | mlog(0, "max_slots for this device: %u\n", osb->max_slots); |
ccd979bd | 1578 | |
539d8264 SM |
1579 | osb->slot_recovery_generations = |
1580 | kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations), | |
1581 | GFP_KERNEL); | |
1582 | if (!osb->slot_recovery_generations) { | |
1583 | status = -ENOMEM; | |
1584 | mlog_errno(status); | |
1585 | goto bail; | |
1586 | } | |
1587 | ||
b4df6ed8 MF |
1588 | init_waitqueue_head(&osb->osb_wipe_event); |
1589 | osb->osb_orphan_wipes = kcalloc(osb->max_slots, | |
1590 | sizeof(*osb->osb_orphan_wipes), | |
1591 | GFP_KERNEL); | |
1592 | if (!osb->osb_orphan_wipes) { | |
1593 | status = -ENOMEM; | |
1594 | mlog_errno(status); | |
1595 | goto bail; | |
1596 | } | |
1597 | ||
ccd979bd MF |
1598 | osb->s_feature_compat = |
1599 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); | |
1600 | osb->s_feature_ro_compat = | |
1601 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); | |
1602 | osb->s_feature_incompat = | |
1603 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); | |
1604 | ||
1605 | if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { | |
1606 | mlog(ML_ERROR, "couldn't mount because of unsupported " | |
1607 | "optional features (%x).\n", i); | |
1608 | status = -EINVAL; | |
1609 | goto bail; | |
1610 | } | |
1611 | if (!(osb->sb->s_flags & MS_RDONLY) && | |
1612 | (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { | |
1613 | mlog(ML_ERROR, "couldn't mount RDWR because of " | |
1614 | "unsupported optional features (%x).\n", i); | |
1615 | status = -EINVAL; | |
1616 | goto bail; | |
1617 | } | |
1618 | ||
b61817e1 JB |
1619 | if (ocfs2_userspace_stack(osb)) { |
1620 | memcpy(osb->osb_cluster_stack, | |
1621 | OCFS2_RAW_SB(di)->s_cluster_info.ci_stack, | |
1622 | OCFS2_STACK_LABEL_LEN); | |
1623 | osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0'; | |
1624 | if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) { | |
1625 | mlog(ML_ERROR, | |
1626 | "couldn't mount because of an invalid " | |
1627 | "cluster stack label (%s) \n", | |
1628 | osb->osb_cluster_stack); | |
1629 | status = -EINVAL; | |
1630 | goto bail; | |
1631 | } | |
1632 | } else { | |
1633 | /* The empty string is identical with classic tools that | |
1634 | * don't know about s_cluster_info. */ | |
1635 | osb->osb_cluster_stack[0] = '\0'; | |
1636 | } | |
1637 | ||
ccd979bd MF |
1638 | get_random_bytes(&osb->s_next_generation, sizeof(u32)); |
1639 | ||
1640 | /* FIXME | |
1641 | * This should be done in ocfs2_journal_init(), but unknown | |
1642 | * ordering issues will cause the filesystem to crash. | |
1643 | * If anyone wants to figure out what part of the code | |
1644 | * refers to osb->journal before ocfs2_journal_init() is run, | |
1645 | * be my guest. | |
1646 | */ | |
1647 | /* initialize our journal structure */ | |
1648 | ||
cd861280 | 1649 | journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); |
ccd979bd MF |
1650 | if (!journal) { |
1651 | mlog(ML_ERROR, "unable to alloc journal\n"); | |
1652 | status = -ENOMEM; | |
1653 | goto bail; | |
1654 | } | |
1655 | osb->journal = journal; | |
1656 | journal->j_osb = osb; | |
1657 | ||
1658 | atomic_set(&journal->j_num_trans, 0); | |
1659 | init_rwsem(&journal->j_trans_barrier); | |
1660 | init_waitqueue_head(&journal->j_checkpointed); | |
1661 | spin_lock_init(&journal->j_lock); | |
1662 | journal->j_trans_id = (unsigned long) 1; | |
1663 | INIT_LIST_HEAD(&journal->j_la_cleanups); | |
c4028958 | 1664 | INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); |
ccd979bd MF |
1665 | journal->j_state = OCFS2_JOURNAL_FREE; |
1666 | ||
1667 | /* get some pseudo constants for clustersize bits */ | |
1668 | osb->s_clustersize_bits = | |
1669 | le32_to_cpu(di->id2.i_super.s_clustersize_bits); | |
1670 | osb->s_clustersize = 1 << osb->s_clustersize_bits; | |
1671 | mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); | |
1672 | ||
1673 | if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || | |
1674 | osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { | |
1675 | mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", | |
1676 | osb->s_clustersize); | |
1677 | status = -EINVAL; | |
1678 | goto bail; | |
1679 | } | |
1680 | ||
1681 | if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) | |
1682 | > (u32)~0UL) { | |
1683 | mlog(ML_ERROR, "Volume might try to write to blocks beyond " | |
1684 | "what jbd can address in 32 bits.\n"); | |
1685 | status = -EINVAL; | |
1686 | goto bail; | |
1687 | } | |
1688 | ||
1689 | if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, | |
1690 | sizeof(di->id2.i_super.s_uuid))) { | |
1691 | mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); | |
1692 | status = -ENOMEM; | |
1693 | goto bail; | |
1694 | } | |
1695 | ||
78427043 | 1696 | memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); |
ccd979bd MF |
1697 | |
1698 | strncpy(osb->vol_label, di->id2.i_super.s_label, 63); | |
1699 | osb->vol_label[63] = '\0'; | |
1700 | osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); | |
1701 | osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); | |
1702 | osb->first_cluster_group_blkno = | |
1703 | le64_to_cpu(di->id2.i_super.s_first_cluster_group); | |
1704 | osb->fs_generation = le32_to_cpu(di->i_fs_generation); | |
cf1d6c76 | 1705 | osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash); |
ccd979bd MF |
1706 | mlog(0, "vol_label: %s\n", osb->vol_label); |
1707 | mlog(0, "uuid: %s\n", osb->uuid_str); | |
b0697053 MF |
1708 | mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", |
1709 | (unsigned long long)osb->root_blkno, | |
1710 | (unsigned long long)osb->system_dir_blkno); | |
ccd979bd MF |
1711 | |
1712 | osb->osb_dlm_debug = ocfs2_new_dlm_debug(); | |
1713 | if (!osb->osb_dlm_debug) { | |
1714 | status = -ENOMEM; | |
1715 | mlog_errno(status); | |
1716 | goto bail; | |
1717 | } | |
1718 | ||
1719 | atomic_set(&osb->vol_state, VOLUME_INIT); | |
1720 | ||
1721 | /* load root, system_dir, and all global system inodes */ | |
1722 | status = ocfs2_init_global_system_inodes(osb); | |
1723 | if (status < 0) { | |
1724 | mlog_errno(status); | |
1725 | goto bail; | |
1726 | } | |
1727 | ||
1728 | /* | |
1729 | * global bitmap | |
1730 | */ | |
1731 | inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, | |
1732 | OCFS2_INVALID_SLOT); | |
1733 | if (!inode) { | |
1734 | status = -EINVAL; | |
1735 | mlog_errno(status); | |
1736 | goto bail; | |
1737 | } | |
1738 | ||
1739 | osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; | |
ccd979bd | 1740 | iput(inode); |
ccd979bd | 1741 | |
e9d578a8 | 1742 | osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; |
ccd979bd MF |
1743 | |
1744 | status = ocfs2_init_slot_info(osb); | |
1745 | if (status < 0) { | |
1746 | mlog_errno(status); | |
1747 | goto bail; | |
1748 | } | |
1749 | ||
ccd979bd MF |
1750 | bail: |
1751 | mlog_exit(status); | |
1752 | return status; | |
1753 | } | |
1754 | ||
1755 | /* | |
1756 | * will return: -EAGAIN if it is ok to keep searching for superblocks | |
1757 | * -EINVAL if there is a bad superblock | |
1758 | * 0 on success | |
1759 | */ | |
1760 | static int ocfs2_verify_volume(struct ocfs2_dinode *di, | |
1761 | struct buffer_head *bh, | |
1762 | u32 blksz) | |
1763 | { | |
1764 | int status = -EAGAIN; | |
1765 | ||
1766 | mlog_entry_void(); | |
1767 | ||
1768 | if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, | |
1769 | strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { | |
1770 | status = -EINVAL; | |
1771 | if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { | |
1772 | mlog(ML_ERROR, "found superblock with incorrect block " | |
1773 | "size: found %u, should be %u\n", | |
1774 | 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), | |
1775 | blksz); | |
1776 | } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != | |
1777 | OCFS2_MAJOR_REV_LEVEL || | |
1778 | le16_to_cpu(di->id2.i_super.s_minor_rev_level) != | |
1779 | OCFS2_MINOR_REV_LEVEL) { | |
1780 | mlog(ML_ERROR, "found superblock with bad version: " | |
1781 | "found %u.%u, should be %u.%u\n", | |
1782 | le16_to_cpu(di->id2.i_super.s_major_rev_level), | |
1783 | le16_to_cpu(di->id2.i_super.s_minor_rev_level), | |
1784 | OCFS2_MAJOR_REV_LEVEL, | |
1785 | OCFS2_MINOR_REV_LEVEL); | |
1786 | } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { | |
1787 | mlog(ML_ERROR, "bad block number on superblock: " | |
b0697053 | 1788 | "found %llu, should be %llu\n", |
1ca1a111 | 1789 | (unsigned long long)le64_to_cpu(di->i_blkno), |
b0697053 | 1790 | (unsigned long long)bh->b_blocknr); |
ccd979bd MF |
1791 | } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || |
1792 | le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { | |
1793 | mlog(ML_ERROR, "bad cluster size found: %u\n", | |
1794 | 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); | |
1795 | } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { | |
1796 | mlog(ML_ERROR, "bad root_blkno: 0\n"); | |
1797 | } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { | |
1798 | mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); | |
1799 | } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { | |
1800 | mlog(ML_ERROR, | |
1801 | "Superblock slots found greater than file system " | |
1802 | "maximum: found %u, max %u\n", | |
1803 | le16_to_cpu(di->id2.i_super.s_max_slots), | |
1804 | OCFS2_MAX_SLOTS); | |
1805 | } else { | |
1806 | /* found it! */ | |
1807 | status = 0; | |
1808 | } | |
1809 | } | |
1810 | ||
1811 | mlog_exit(status); | |
1812 | return status; | |
1813 | } | |
1814 | ||
1815 | static int ocfs2_check_volume(struct ocfs2_super *osb) | |
1816 | { | |
bddb8eb3 | 1817 | int status; |
ccd979bd | 1818 | int dirty; |
c271c5c2 | 1819 | int local; |
ccd979bd MF |
1820 | struct ocfs2_dinode *local_alloc = NULL; /* only used if we |
1821 | * recover | |
1822 | * ourselves. */ | |
1823 | ||
1824 | mlog_entry_void(); | |
1825 | ||
1826 | /* Init our journal object. */ | |
1827 | status = ocfs2_journal_init(osb->journal, &dirty); | |
1828 | if (status < 0) { | |
1829 | mlog(ML_ERROR, "Could not initialize journal!\n"); | |
1830 | goto finally; | |
1831 | } | |
1832 | ||
1833 | /* If the journal was unmounted cleanly then we don't want to | |
1834 | * recover anything. Otherwise, journal_load will do that | |
1835 | * dirty work for us :) */ | |
1836 | if (!dirty) { | |
1837 | status = ocfs2_journal_wipe(osb->journal, 0); | |
1838 | if (status < 0) { | |
1839 | mlog_errno(status); | |
1840 | goto finally; | |
1841 | } | |
1842 | } else { | |
1843 | mlog(ML_NOTICE, "File system was not unmounted cleanly, " | |
1844 | "recovering volume.\n"); | |
1845 | } | |
1846 | ||
c271c5c2 SM |
1847 | local = ocfs2_mount_local(osb); |
1848 | ||
ccd979bd | 1849 | /* will play back anything left in the journal. */ |
539d8264 | 1850 | status = ocfs2_journal_load(osb->journal, local, dirty); |
01af4820 WW |
1851 | if (status < 0) { |
1852 | mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status); | |
1853 | goto finally; | |
1854 | } | |
ccd979bd MF |
1855 | |
1856 | if (dirty) { | |
1857 | /* recover my local alloc if we didn't unmount cleanly. */ | |
1858 | status = ocfs2_begin_local_alloc_recovery(osb, | |
1859 | osb->slot_num, | |
1860 | &local_alloc); | |
1861 | if (status < 0) { | |
1862 | mlog_errno(status); | |
1863 | goto finally; | |
1864 | } | |
1865 | /* we complete the recovery process after we've marked | |
1866 | * ourselves as mounted. */ | |
1867 | } | |
1868 | ||
1869 | mlog(0, "Journal loaded.\n"); | |
1870 | ||
1871 | status = ocfs2_load_local_alloc(osb); | |
1872 | if (status < 0) { | |
1873 | mlog_errno(status); | |
1874 | goto finally; | |
1875 | } | |
1876 | ||
1877 | if (dirty) { | |
1878 | /* Recovery will be completed after we've mounted the | |
1879 | * rest of the volume. */ | |
1880 | osb->dirty = 1; | |
1881 | osb->local_alloc_copy = local_alloc; | |
1882 | local_alloc = NULL; | |
1883 | } | |
1884 | ||
1885 | /* go through each journal, trylock it and if you get the | |
1886 | * lock, and it's marked as dirty, set the bit in the recover | |
1887 | * map and launch a recovery thread for it. */ | |
1888 | status = ocfs2_mark_dead_nodes(osb); | |
1889 | if (status < 0) | |
1890 | mlog_errno(status); | |
1891 | ||
1892 | finally: | |
1893 | if (local_alloc) | |
1894 | kfree(local_alloc); | |
1895 | ||
1896 | mlog_exit(status); | |
1897 | return status; | |
1898 | } | |
1899 | ||
1900 | /* | |
1901 | * The routine gets called from dismount or close whenever a dismount on | |
1902 | * volume is requested and the osb open count becomes 1. | |
1903 | * It will remove the osb from the global list and also free up all the | |
1904 | * initialized resources and fileobject. | |
1905 | */ | |
1906 | static void ocfs2_delete_osb(struct ocfs2_super *osb) | |
1907 | { | |
1908 | mlog_entry_void(); | |
1909 | ||
1910 | /* This function assumes that the caller has the main osb resource */ | |
1911 | ||
8e8a4603 | 1912 | ocfs2_free_slot_info(osb); |
ccd979bd | 1913 | |
b4df6ed8 | 1914 | kfree(osb->osb_orphan_wipes); |
539d8264 | 1915 | kfree(osb->slot_recovery_generations); |
ccd979bd MF |
1916 | /* FIXME |
1917 | * This belongs in journal shutdown, but because we have to | |
1918 | * allocate osb->journal at the start of ocfs2_initalize_osb(), | |
1919 | * we free it here. | |
1920 | */ | |
1921 | kfree(osb->journal); | |
1922 | if (osb->local_alloc_copy) | |
1923 | kfree(osb->local_alloc_copy); | |
1924 | kfree(osb->uuid_str); | |
1925 | ocfs2_put_dlm_debug(osb->osb_dlm_debug); | |
1926 | memset(osb, 0, sizeof(struct ocfs2_super)); | |
1927 | ||
1928 | mlog_exit_void(); | |
1929 | } | |
1930 | ||
1931 | /* Put OCFS2 into a readonly state, or (if the user specifies it), | |
1932 | * panic(). We do not support continue-on-error operation. */ | |
1933 | static void ocfs2_handle_error(struct super_block *sb) | |
1934 | { | |
1935 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
1936 | ||
1937 | if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) | |
1938 | panic("OCFS2: (device %s): panic forced after error\n", | |
1939 | sb->s_id); | |
1940 | ||
1941 | ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); | |
1942 | ||
1943 | if (sb->s_flags & MS_RDONLY && | |
1944 | (ocfs2_is_soft_readonly(osb) || | |
1945 | ocfs2_is_hard_readonly(osb))) | |
1946 | return; | |
1947 | ||
1948 | printk(KERN_CRIT "File system is now read-only due to the potential " | |
1949 | "of on-disk corruption. Please run fsck.ocfs2 once the file " | |
1950 | "system is unmounted.\n"); | |
1951 | sb->s_flags |= MS_RDONLY; | |
1952 | ocfs2_set_ro_flag(osb, 0); | |
1953 | } | |
1954 | ||
1955 | static char error_buf[1024]; | |
1956 | ||
1957 | void __ocfs2_error(struct super_block *sb, | |
1958 | const char *function, | |
1959 | const char *fmt, ...) | |
1960 | { | |
1961 | va_list args; | |
1962 | ||
1963 | va_start(args, fmt); | |
4a6e617a | 1964 | vsnprintf(error_buf, sizeof(error_buf), fmt, args); |
ccd979bd MF |
1965 | va_end(args); |
1966 | ||
1967 | /* Not using mlog here because we want to show the actual | |
1968 | * function the error came from. */ | |
1969 | printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n", | |
1970 | sb->s_id, function, error_buf); | |
1971 | ||
1972 | ocfs2_handle_error(sb); | |
1973 | } | |
1974 | ||
1975 | /* Handle critical errors. This is intentionally more drastic than | |
1976 | * ocfs2_handle_error, so we only use for things like journal errors, | |
1977 | * etc. */ | |
1978 | void __ocfs2_abort(struct super_block* sb, | |
1979 | const char *function, | |
1980 | const char *fmt, ...) | |
1981 | { | |
1982 | va_list args; | |
1983 | ||
1984 | va_start(args, fmt); | |
4a6e617a | 1985 | vsnprintf(error_buf, sizeof(error_buf), fmt, args); |
ccd979bd MF |
1986 | va_end(args); |
1987 | ||
1988 | printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", | |
1989 | sb->s_id, function, error_buf); | |
1990 | ||
1991 | /* We don't have the cluster support yet to go straight to | |
1992 | * hard readonly in here. Until then, we want to keep | |
1993 | * ocfs2_abort() so that we can at least mark critical | |
1994 | * errors. | |
1995 | * | |
1996 | * TODO: This should abort the journal and alert other nodes | |
1997 | * that our slot needs recovery. */ | |
1998 | ||
1999 | /* Force a panic(). This stinks, but it's better than letting | |
2000 | * things continue without having a proper hard readonly | |
2001 | * here. */ | |
2002 | OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; | |
2003 | ocfs2_handle_error(sb); | |
2004 | } | |
2005 | ||
2006 | module_init(ocfs2_init); | |
2007 | module_exit(ocfs2_exit); |