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