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> |
ccd979bd MF |
43 | |
44 | #include <cluster/nodemanager.h> | |
45 | ||
46 | #define MLOG_MASK_PREFIX ML_SUPER | |
47 | #include <cluster/masklog.h> | |
48 | ||
49 | #include "ocfs2.h" | |
50 | ||
51 | /* this should be the only file to include a version 1 header */ | |
52 | #include "ocfs1_fs_compat.h" | |
53 | ||
54 | #include "alloc.h" | |
55 | #include "dlmglue.h" | |
56 | #include "export.h" | |
57 | #include "extent_map.h" | |
58 | #include "heartbeat.h" | |
59 | #include "inode.h" | |
60 | #include "journal.h" | |
61 | #include "localalloc.h" | |
62 | #include "namei.h" | |
63 | #include "slot_map.h" | |
64 | #include "super.h" | |
65 | #include "sysfile.h" | |
66 | #include "uptodate.h" | |
67 | #include "ver.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 | { | |
86 | unsigned long mount_opt; | |
87 | unsigned int atime_quantum; | |
88 | signed short slot; | |
89 | }; | |
90 | ||
ccd979bd | 91 | static int ocfs2_parse_options(struct super_block *sb, char *options, |
c0123ade | 92 | struct mount_options *mopt, |
baf4661a | 93 | int is_remount); |
d550071c | 94 | static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt); |
ccd979bd MF |
95 | static void ocfs2_put_super(struct super_block *sb); |
96 | static int ocfs2_mount_volume(struct super_block *sb); | |
97 | static int ocfs2_remount(struct super_block *sb, int *flags, char *data); | |
98 | static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err); | |
99 | static int ocfs2_initialize_mem_caches(void); | |
100 | static void ocfs2_free_mem_caches(void); | |
101 | static void ocfs2_delete_osb(struct ocfs2_super *osb); | |
102 | ||
726c3342 | 103 | static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf); |
ccd979bd MF |
104 | |
105 | static int ocfs2_sync_fs(struct super_block *sb, int wait); | |
106 | ||
107 | static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb); | |
108 | static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb); | |
bddb8eb3 | 109 | static void ocfs2_release_system_inodes(struct ocfs2_super *osb); |
ccd979bd MF |
110 | static int ocfs2_fill_local_node_info(struct ocfs2_super *osb); |
111 | static int ocfs2_check_volume(struct ocfs2_super *osb); | |
112 | static int ocfs2_verify_volume(struct ocfs2_dinode *di, | |
113 | struct buffer_head *bh, | |
114 | u32 sectsize); | |
115 | static int ocfs2_initialize_super(struct super_block *sb, | |
116 | struct buffer_head *bh, | |
117 | int sector_size); | |
118 | static int ocfs2_get_sector(struct super_block *sb, | |
119 | struct buffer_head **bh, | |
120 | int block, | |
121 | int sect_size); | |
122 | static void ocfs2_write_super(struct super_block *sb); | |
123 | static struct inode *ocfs2_alloc_inode(struct super_block *sb); | |
124 | static void ocfs2_destroy_inode(struct inode *inode); | |
125 | ||
ee9b6d61 | 126 | static const struct super_operations ocfs2_sops = { |
ccd979bd MF |
127 | .statfs = ocfs2_statfs, |
128 | .alloc_inode = ocfs2_alloc_inode, | |
129 | .destroy_inode = ocfs2_destroy_inode, | |
130 | .drop_inode = ocfs2_drop_inode, | |
131 | .clear_inode = ocfs2_clear_inode, | |
132 | .delete_inode = ocfs2_delete_inode, | |
133 | .sync_fs = ocfs2_sync_fs, | |
134 | .write_super = ocfs2_write_super, | |
135 | .put_super = ocfs2_put_super, | |
136 | .remount_fs = ocfs2_remount, | |
d550071c | 137 | .show_options = ocfs2_show_options, |
ccd979bd MF |
138 | }; |
139 | ||
140 | enum { | |
141 | Opt_barrier, | |
142 | Opt_err_panic, | |
143 | Opt_err_ro, | |
144 | Opt_intr, | |
145 | Opt_nointr, | |
146 | Opt_hb_none, | |
147 | Opt_hb_local, | |
148 | Opt_data_ordered, | |
149 | Opt_data_writeback, | |
7f1a37e3 | 150 | Opt_atime_quantum, |
baf4661a | 151 | Opt_slot, |
ccd979bd MF |
152 | Opt_err, |
153 | }; | |
154 | ||
155 | static match_table_t tokens = { | |
156 | {Opt_barrier, "barrier=%u"}, | |
157 | {Opt_err_panic, "errors=panic"}, | |
158 | {Opt_err_ro, "errors=remount-ro"}, | |
159 | {Opt_intr, "intr"}, | |
160 | {Opt_nointr, "nointr"}, | |
161 | {Opt_hb_none, OCFS2_HB_NONE}, | |
162 | {Opt_hb_local, OCFS2_HB_LOCAL}, | |
163 | {Opt_data_ordered, "data=ordered"}, | |
164 | {Opt_data_writeback, "data=writeback"}, | |
7f1a37e3 | 165 | {Opt_atime_quantum, "atime_quantum=%u"}, |
baf4661a | 166 | {Opt_slot, "preferred_slot=%u"}, |
ccd979bd MF |
167 | {Opt_err, NULL} |
168 | }; | |
169 | ||
170 | /* | |
171 | * write_super and sync_fs ripped right out of ext3. | |
172 | */ | |
173 | static void ocfs2_write_super(struct super_block *sb) | |
174 | { | |
7892f2f4 | 175 | if (mutex_trylock(&sb->s_lock) != 0) |
ccd979bd MF |
176 | BUG(); |
177 | sb->s_dirt = 0; | |
178 | } | |
179 | ||
180 | static int ocfs2_sync_fs(struct super_block *sb, int wait) | |
181 | { | |
bddb8eb3 | 182 | int status; |
ccd979bd MF |
183 | tid_t target; |
184 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
185 | ||
186 | sb->s_dirt = 0; | |
187 | ||
188 | if (ocfs2_is_hard_readonly(osb)) | |
189 | return -EROFS; | |
190 | ||
191 | if (wait) { | |
192 | status = ocfs2_flush_truncate_log(osb); | |
193 | if (status < 0) | |
194 | mlog_errno(status); | |
195 | } else { | |
196 | ocfs2_schedule_truncate_log_flush(osb, 0); | |
197 | } | |
198 | ||
199 | if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) { | |
200 | if (wait) | |
201 | log_wait_commit(OCFS2_SB(sb)->journal->j_journal, | |
202 | target); | |
203 | } | |
204 | return 0; | |
205 | } | |
206 | ||
207 | static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb) | |
208 | { | |
209 | struct inode *new = NULL; | |
210 | int status = 0; | |
211 | int i; | |
212 | ||
213 | mlog_entry_void(); | |
214 | ||
24c19ef4 | 215 | new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE); |
ccd979bd MF |
216 | if (IS_ERR(new)) { |
217 | status = PTR_ERR(new); | |
218 | mlog_errno(status); | |
219 | goto bail; | |
220 | } | |
221 | osb->root_inode = new; | |
222 | ||
24c19ef4 | 223 | new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE); |
ccd979bd MF |
224 | if (IS_ERR(new)) { |
225 | status = PTR_ERR(new); | |
226 | mlog_errno(status); | |
227 | goto bail; | |
228 | } | |
229 | osb->sys_root_inode = new; | |
230 | ||
231 | for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE; | |
232 | i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) { | |
233 | new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); | |
234 | if (!new) { | |
235 | ocfs2_release_system_inodes(osb); | |
236 | status = -EINVAL; | |
237 | mlog_errno(status); | |
238 | /* FIXME: Should ERROR_RO_FS */ | |
239 | mlog(ML_ERROR, "Unable to load system inode %d, " | |
240 | "possibly corrupt fs?", i); | |
241 | goto bail; | |
242 | } | |
243 | // the array now has one ref, so drop this one | |
244 | iput(new); | |
245 | } | |
246 | ||
247 | bail: | |
248 | mlog_exit(status); | |
249 | return status; | |
250 | } | |
251 | ||
252 | static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb) | |
253 | { | |
254 | struct inode *new = NULL; | |
255 | int status = 0; | |
256 | int i; | |
257 | ||
258 | mlog_entry_void(); | |
259 | ||
260 | for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1; | |
261 | i < NUM_SYSTEM_INODES; | |
262 | i++) { | |
263 | new = ocfs2_get_system_file_inode(osb, i, osb->slot_num); | |
264 | if (!new) { | |
265 | ocfs2_release_system_inodes(osb); | |
266 | status = -EINVAL; | |
267 | mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n", | |
268 | status, i, osb->slot_num); | |
269 | goto bail; | |
270 | } | |
271 | /* the array now has one ref, so drop this one */ | |
272 | iput(new); | |
273 | } | |
274 | ||
275 | bail: | |
276 | mlog_exit(status); | |
277 | return status; | |
278 | } | |
279 | ||
bddb8eb3 | 280 | static void ocfs2_release_system_inodes(struct ocfs2_super *osb) |
ccd979bd | 281 | { |
bddb8eb3 | 282 | int i; |
ccd979bd MF |
283 | struct inode *inode; |
284 | ||
285 | mlog_entry_void(); | |
286 | ||
287 | for (i = 0; i < NUM_SYSTEM_INODES; i++) { | |
288 | inode = osb->system_inodes[i]; | |
289 | if (inode) { | |
290 | iput(inode); | |
291 | osb->system_inodes[i] = NULL; | |
292 | } | |
293 | } | |
294 | ||
295 | inode = osb->sys_root_inode; | |
296 | if (inode) { | |
297 | iput(inode); | |
298 | osb->sys_root_inode = NULL; | |
299 | } | |
300 | ||
301 | inode = osb->root_inode; | |
302 | if (inode) { | |
303 | iput(inode); | |
304 | osb->root_inode = NULL; | |
305 | } | |
306 | ||
bddb8eb3 | 307 | mlog_exit(0); |
ccd979bd MF |
308 | } |
309 | ||
310 | /* We're allocating fs objects, use GFP_NOFS */ | |
311 | static struct inode *ocfs2_alloc_inode(struct super_block *sb) | |
312 | { | |
313 | struct ocfs2_inode_info *oi; | |
314 | ||
e6b4f8da | 315 | oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS); |
ccd979bd MF |
316 | if (!oi) |
317 | return NULL; | |
318 | ||
319 | return &oi->vfs_inode; | |
320 | } | |
321 | ||
322 | static void ocfs2_destroy_inode(struct inode *inode) | |
323 | { | |
324 | kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode)); | |
325 | } | |
326 | ||
5a254031 MF |
327 | static unsigned long long ocfs2_max_file_offset(unsigned int bbits, |
328 | unsigned int cbits) | |
ccd979bd | 329 | { |
5a254031 MF |
330 | unsigned int bytes = 1 << cbits; |
331 | unsigned int trim = bytes; | |
332 | unsigned int bitshift = 32; | |
333 | ||
334 | /* | |
335 | * i_size and all block offsets in ocfs2 are always 64 bits | |
336 | * wide. i_clusters is 32 bits, in cluster-sized units. So on | |
337 | * 64 bit platforms, cluster size will be the limiting factor. | |
ccd979bd MF |
338 | */ |
339 | ||
340 | #if BITS_PER_LONG == 32 | |
341 | # if defined(CONFIG_LBD) | |
2ecd05ae | 342 | BUILD_BUG_ON(sizeof(sector_t) != 8); |
5a254031 MF |
343 | /* |
344 | * We might be limited by page cache size. | |
345 | */ | |
346 | if (bytes > PAGE_CACHE_SIZE) { | |
347 | bytes = PAGE_CACHE_SIZE; | |
348 | trim = 1; | |
349 | /* | |
350 | * Shift by 31 here so that we don't get larger than | |
351 | * MAX_LFS_FILESIZE | |
352 | */ | |
353 | bitshift = 31; | |
354 | } | |
ccd979bd | 355 | # else |
5a254031 MF |
356 | /* |
357 | * We are limited by the size of sector_t. Use block size, as | |
358 | * that's what we expose to the VFS. | |
359 | */ | |
360 | bytes = 1 << bbits; | |
361 | trim = 1; | |
362 | bitshift = 31; | |
ccd979bd MF |
363 | # endif |
364 | #endif | |
365 | ||
5a254031 MF |
366 | /* |
367 | * Trim by a whole cluster when we can actually approach the | |
368 | * on-disk limits. Otherwise we can overflow i_clusters when | |
369 | * an extent start is at the max offset. | |
370 | */ | |
371 | return (((unsigned long long)bytes) << bitshift) - trim; | |
ccd979bd MF |
372 | } |
373 | ||
374 | static int ocfs2_remount(struct super_block *sb, int *flags, char *data) | |
375 | { | |
376 | int incompat_features; | |
377 | int ret = 0; | |
c0123ade | 378 | struct mount_options parsed_options; |
ccd979bd MF |
379 | struct ocfs2_super *osb = OCFS2_SB(sb); |
380 | ||
c0123ade | 381 | if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) { |
ccd979bd MF |
382 | ret = -EINVAL; |
383 | goto out; | |
384 | } | |
385 | ||
386 | if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) != | |
c0123ade | 387 | (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) { |
ccd979bd MF |
388 | ret = -EINVAL; |
389 | mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n"); | |
390 | goto out; | |
391 | } | |
392 | ||
393 | if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) != | |
c0123ade | 394 | (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) { |
ccd979bd MF |
395 | ret = -EINVAL; |
396 | mlog(ML_ERROR, "Cannot change data mode on remount\n"); | |
397 | goto out; | |
398 | } | |
399 | ||
400 | /* We're going to/from readonly mode. */ | |
401 | if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) { | |
402 | /* Lock here so the check of HARD_RO and the potential | |
403 | * setting of SOFT_RO is atomic. */ | |
404 | spin_lock(&osb->osb_lock); | |
405 | if (osb->osb_flags & OCFS2_OSB_HARD_RO) { | |
406 | mlog(ML_ERROR, "Remount on readonly device is forbidden.\n"); | |
407 | ret = -EROFS; | |
408 | goto unlock_osb; | |
409 | } | |
410 | ||
411 | if (*flags & MS_RDONLY) { | |
412 | mlog(0, "Going to ro mode.\n"); | |
413 | sb->s_flags |= MS_RDONLY; | |
414 | osb->osb_flags |= OCFS2_OSB_SOFT_RO; | |
415 | } else { | |
416 | mlog(0, "Making ro filesystem writeable.\n"); | |
417 | ||
418 | if (osb->osb_flags & OCFS2_OSB_ERROR_FS) { | |
419 | mlog(ML_ERROR, "Cannot remount RDWR " | |
420 | "filesystem due to previous errors.\n"); | |
421 | ret = -EROFS; | |
422 | goto unlock_osb; | |
423 | } | |
424 | incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP); | |
425 | if (incompat_features) { | |
426 | mlog(ML_ERROR, "Cannot remount RDWR because " | |
427 | "of unsupported optional features " | |
428 | "(%x).\n", incompat_features); | |
429 | ret = -EINVAL; | |
430 | goto unlock_osb; | |
431 | } | |
432 | sb->s_flags &= ~MS_RDONLY; | |
433 | osb->osb_flags &= ~OCFS2_OSB_SOFT_RO; | |
434 | } | |
435 | unlock_osb: | |
436 | spin_unlock(&osb->osb_lock); | |
437 | } | |
438 | ||
439 | if (!ret) { | |
ccd979bd MF |
440 | /* Only save off the new mount options in case of a successful |
441 | * remount. */ | |
c0123ade TY |
442 | osb->s_mount_opt = parsed_options.mount_opt; |
443 | osb->s_atime_quantum = parsed_options.atime_quantum; | |
444 | osb->preferred_slot = parsed_options.slot; | |
e001e796 MF |
445 | |
446 | if (!ocfs2_is_hard_readonly(osb)) | |
447 | ocfs2_set_journal_params(osb); | |
ccd979bd MF |
448 | } |
449 | out: | |
450 | return ret; | |
451 | } | |
452 | ||
453 | static int ocfs2_sb_probe(struct super_block *sb, | |
454 | struct buffer_head **bh, | |
455 | int *sector_size) | |
456 | { | |
bddb8eb3 | 457 | int status, tmpstat; |
ccd979bd MF |
458 | struct ocfs1_vol_disk_hdr *hdr; |
459 | struct ocfs2_dinode *di; | |
460 | int blksize; | |
461 | ||
462 | *bh = NULL; | |
463 | ||
464 | /* may be > 512 */ | |
465 | *sector_size = bdev_hardsect_size(sb->s_bdev); | |
466 | if (*sector_size > OCFS2_MAX_BLOCKSIZE) { | |
467 | mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n", | |
468 | *sector_size, OCFS2_MAX_BLOCKSIZE); | |
469 | status = -EINVAL; | |
470 | goto bail; | |
471 | } | |
472 | ||
473 | /* Can this really happen? */ | |
474 | if (*sector_size < OCFS2_MIN_BLOCKSIZE) | |
475 | *sector_size = OCFS2_MIN_BLOCKSIZE; | |
476 | ||
477 | /* check block zero for old format */ | |
478 | status = ocfs2_get_sector(sb, bh, 0, *sector_size); | |
479 | if (status < 0) { | |
480 | mlog_errno(status); | |
481 | goto bail; | |
482 | } | |
483 | hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data; | |
484 | if (hdr->major_version == OCFS1_MAJOR_VERSION) { | |
485 | mlog(ML_ERROR, "incompatible version: %u.%u\n", | |
486 | hdr->major_version, hdr->minor_version); | |
487 | status = -EINVAL; | |
488 | } | |
489 | if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE, | |
490 | strlen(OCFS1_VOLUME_SIGNATURE)) == 0) { | |
491 | mlog(ML_ERROR, "incompatible volume signature: %8s\n", | |
492 | hdr->signature); | |
493 | status = -EINVAL; | |
494 | } | |
495 | brelse(*bh); | |
496 | *bh = NULL; | |
497 | if (status < 0) { | |
498 | mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be " | |
499 | "upgraded before mounting with ocfs v2\n"); | |
500 | goto bail; | |
501 | } | |
502 | ||
503 | /* | |
504 | * Now check at magic offset for 512, 1024, 2048, 4096 | |
505 | * blocksizes. 4096 is the maximum blocksize because it is | |
506 | * the minimum clustersize. | |
507 | */ | |
508 | status = -EINVAL; | |
509 | for (blksize = *sector_size; | |
510 | blksize <= OCFS2_MAX_BLOCKSIZE; | |
511 | blksize <<= 1) { | |
512 | tmpstat = ocfs2_get_sector(sb, bh, | |
513 | OCFS2_SUPER_BLOCK_BLKNO, | |
514 | blksize); | |
515 | if (tmpstat < 0) { | |
516 | status = tmpstat; | |
517 | mlog_errno(status); | |
518 | goto bail; | |
519 | } | |
520 | di = (struct ocfs2_dinode *) (*bh)->b_data; | |
521 | status = ocfs2_verify_volume(di, *bh, blksize); | |
522 | if (status >= 0) | |
523 | goto bail; | |
524 | brelse(*bh); | |
525 | *bh = NULL; | |
526 | if (status != -EAGAIN) | |
527 | break; | |
528 | } | |
529 | ||
530 | bail: | |
531 | return status; | |
532 | } | |
533 | ||
c271c5c2 SM |
534 | static int ocfs2_verify_heartbeat(struct ocfs2_super *osb) |
535 | { | |
536 | if (ocfs2_mount_local(osb)) { | |
537 | if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { | |
538 | mlog(ML_ERROR, "Cannot heartbeat on a locally " | |
539 | "mounted device.\n"); | |
540 | return -EINVAL; | |
541 | } | |
542 | } | |
543 | ||
544 | if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) { | |
545 | if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) { | |
546 | mlog(ML_ERROR, "Heartbeat has to be started to mount " | |
547 | "a read-write clustered device.\n"); | |
548 | return -EINVAL; | |
549 | } | |
550 | } | |
551 | ||
552 | return 0; | |
553 | } | |
554 | ||
ccd979bd MF |
555 | static int ocfs2_fill_super(struct super_block *sb, void *data, int silent) |
556 | { | |
557 | struct dentry *root; | |
558 | int status, sector_size; | |
c0123ade | 559 | struct mount_options parsed_options; |
ccd979bd MF |
560 | struct inode *inode = NULL; |
561 | struct ocfs2_super *osb = NULL; | |
562 | struct buffer_head *bh = NULL; | |
c271c5c2 | 563 | char nodestr[8]; |
ccd979bd MF |
564 | |
565 | mlog_entry("%p, %p, %i", sb, data, silent); | |
566 | ||
c0123ade | 567 | if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) { |
ccd979bd MF |
568 | status = -EINVAL; |
569 | goto read_super_error; | |
570 | } | |
571 | ||
c271c5c2 SM |
572 | /* for now we only have one cluster/node, make sure we see it |
573 | * in the heartbeat universe */ | |
c0123ade | 574 | if (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL) { |
c271c5c2 SM |
575 | if (!o2hb_check_local_node_heartbeating()) { |
576 | status = -EINVAL; | |
577 | goto read_super_error; | |
578 | } | |
579 | } | |
580 | ||
ccd979bd MF |
581 | /* probe for superblock */ |
582 | status = ocfs2_sb_probe(sb, &bh, §or_size); | |
583 | if (status < 0) { | |
584 | mlog(ML_ERROR, "superblock probe failed!\n"); | |
585 | goto read_super_error; | |
586 | } | |
587 | ||
588 | status = ocfs2_initialize_super(sb, bh, sector_size); | |
589 | osb = OCFS2_SB(sb); | |
590 | if (status < 0) { | |
591 | mlog_errno(status); | |
592 | goto read_super_error; | |
593 | } | |
594 | brelse(bh); | |
595 | bh = NULL; | |
c0123ade TY |
596 | osb->s_mount_opt = parsed_options.mount_opt; |
597 | osb->s_atime_quantum = parsed_options.atime_quantum; | |
598 | osb->preferred_slot = parsed_options.slot; | |
ccd979bd MF |
599 | |
600 | sb->s_magic = OCFS2_SUPER_MAGIC; | |
601 | ||
602 | /* Hard readonly mode only if: bdev_read_only, MS_RDONLY, | |
603 | * heartbeat=none */ | |
604 | if (bdev_read_only(sb->s_bdev)) { | |
605 | if (!(sb->s_flags & MS_RDONLY)) { | |
606 | status = -EACCES; | |
607 | mlog(ML_ERROR, "Readonly device detected but readonly " | |
608 | "mount was not specified.\n"); | |
609 | goto read_super_error; | |
610 | } | |
611 | ||
612 | /* You should not be able to start a local heartbeat | |
613 | * on a readonly device. */ | |
614 | if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) { | |
615 | status = -EROFS; | |
616 | mlog(ML_ERROR, "Local heartbeat specified on readonly " | |
617 | "device.\n"); | |
618 | goto read_super_error; | |
619 | } | |
620 | ||
621 | status = ocfs2_check_journals_nolocks(osb); | |
622 | if (status < 0) { | |
623 | if (status == -EROFS) | |
624 | mlog(ML_ERROR, "Recovery required on readonly " | |
625 | "file system, but write access is " | |
626 | "unavailable.\n"); | |
627 | else | |
628 | mlog_errno(status); | |
629 | goto read_super_error; | |
630 | } | |
631 | ||
632 | ocfs2_set_ro_flag(osb, 1); | |
633 | ||
634 | printk(KERN_NOTICE "Readonly device detected. No cluster " | |
635 | "services will be utilized for this mount. Recovery " | |
636 | "will be skipped.\n"); | |
637 | } | |
638 | ||
639 | if (!ocfs2_is_hard_readonly(osb)) { | |
ccd979bd MF |
640 | if (sb->s_flags & MS_RDONLY) |
641 | ocfs2_set_ro_flag(osb, 0); | |
642 | } | |
643 | ||
c271c5c2 SM |
644 | status = ocfs2_verify_heartbeat(osb); |
645 | if (status < 0) { | |
646 | mlog_errno(status); | |
647 | goto read_super_error; | |
648 | } | |
649 | ||
ccd979bd MF |
650 | osb->osb_debug_root = debugfs_create_dir(osb->uuid_str, |
651 | ocfs2_debugfs_root); | |
652 | if (!osb->osb_debug_root) { | |
653 | status = -EINVAL; | |
654 | mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n"); | |
655 | goto read_super_error; | |
656 | } | |
657 | ||
658 | status = ocfs2_mount_volume(sb); | |
659 | if (osb->root_inode) | |
660 | inode = igrab(osb->root_inode); | |
661 | ||
662 | if (status < 0) | |
663 | goto read_super_error; | |
664 | ||
665 | if (!inode) { | |
666 | status = -EIO; | |
667 | mlog_errno(status); | |
668 | goto read_super_error; | |
669 | } | |
670 | ||
671 | root = d_alloc_root(inode); | |
672 | if (!root) { | |
673 | status = -ENOMEM; | |
674 | mlog_errno(status); | |
675 | goto read_super_error; | |
676 | } | |
677 | ||
678 | sb->s_root = root; | |
679 | ||
680 | ocfs2_complete_mount_recovery(osb); | |
681 | ||
c271c5c2 SM |
682 | if (ocfs2_mount_local(osb)) |
683 | snprintf(nodestr, sizeof(nodestr), "local"); | |
684 | else | |
685 | snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num); | |
686 | ||
687 | printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) " | |
781ee3e2 | 688 | "with %s data mode.\n", |
c271c5c2 | 689 | osb->dev_str, nodestr, osb->slot_num, |
ccd979bd MF |
690 | osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" : |
691 | "ordered"); | |
692 | ||
693 | atomic_set(&osb->vol_state, VOLUME_MOUNTED); | |
694 | wake_up(&osb->osb_mount_event); | |
695 | ||
696 | mlog_exit(status); | |
697 | return status; | |
698 | ||
699 | read_super_error: | |
700 | if (bh != NULL) | |
701 | brelse(bh); | |
702 | ||
703 | if (inode) | |
704 | iput(inode); | |
705 | ||
706 | if (osb) { | |
707 | atomic_set(&osb->vol_state, VOLUME_DISABLED); | |
708 | wake_up(&osb->osb_mount_event); | |
709 | ocfs2_dismount_volume(sb, 1); | |
710 | } | |
711 | ||
712 | mlog_exit(status); | |
713 | return status; | |
714 | } | |
715 | ||
454e2398 DH |
716 | static int ocfs2_get_sb(struct file_system_type *fs_type, |
717 | int flags, | |
718 | const char *dev_name, | |
719 | void *data, | |
720 | struct vfsmount *mnt) | |
ccd979bd | 721 | { |
454e2398 DH |
722 | return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super, |
723 | mnt); | |
ccd979bd MF |
724 | } |
725 | ||
726 | static struct file_system_type ocfs2_fs_type = { | |
727 | .owner = THIS_MODULE, | |
728 | .name = "ocfs2", | |
729 | .get_sb = ocfs2_get_sb, /* is this called when we mount | |
730 | * the fs? */ | |
731 | .kill_sb = kill_block_super, /* set to the generic one | |
732 | * right now, but do we | |
733 | * need to change that? */ | |
1ba9da2f | 734 | .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE, |
ccd979bd MF |
735 | .next = NULL |
736 | }; | |
737 | ||
738 | static int ocfs2_parse_options(struct super_block *sb, | |
739 | char *options, | |
c0123ade | 740 | struct mount_options *mopt, |
ccd979bd MF |
741 | int is_remount) |
742 | { | |
743 | int status; | |
744 | char *p; | |
745 | ||
746 | mlog_entry("remount: %d, options: \"%s\"\n", is_remount, | |
747 | options ? options : "(none)"); | |
748 | ||
c0123ade TY |
749 | mopt->mount_opt = 0; |
750 | mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; | |
751 | mopt->slot = OCFS2_INVALID_SLOT; | |
ccd979bd MF |
752 | |
753 | if (!options) { | |
754 | status = 1; | |
755 | goto bail; | |
756 | } | |
757 | ||
758 | while ((p = strsep(&options, ",")) != NULL) { | |
759 | int token, option; | |
760 | substring_t args[MAX_OPT_ARGS]; | |
761 | ||
762 | if (!*p) | |
763 | continue; | |
764 | ||
765 | token = match_token(p, tokens, args); | |
766 | switch (token) { | |
767 | case Opt_hb_local: | |
c0123ade | 768 | mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL; |
ccd979bd MF |
769 | break; |
770 | case Opt_hb_none: | |
c0123ade | 771 | mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL; |
ccd979bd MF |
772 | break; |
773 | case Opt_barrier: | |
774 | if (match_int(&args[0], &option)) { | |
775 | status = 0; | |
776 | goto bail; | |
777 | } | |
778 | if (option) | |
c0123ade | 779 | mopt->mount_opt |= OCFS2_MOUNT_BARRIER; |
ccd979bd | 780 | else |
c0123ade | 781 | mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER; |
ccd979bd MF |
782 | break; |
783 | case Opt_intr: | |
c0123ade | 784 | mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR; |
ccd979bd MF |
785 | break; |
786 | case Opt_nointr: | |
c0123ade | 787 | mopt->mount_opt |= OCFS2_MOUNT_NOINTR; |
ccd979bd MF |
788 | break; |
789 | case Opt_err_panic: | |
c0123ade | 790 | mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; |
ccd979bd MF |
791 | break; |
792 | case Opt_err_ro: | |
c0123ade | 793 | mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC; |
ccd979bd MF |
794 | break; |
795 | case Opt_data_ordered: | |
c0123ade | 796 | mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK; |
ccd979bd MF |
797 | break; |
798 | case Opt_data_writeback: | |
c0123ade | 799 | mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK; |
ccd979bd | 800 | break; |
7f1a37e3 TY |
801 | case Opt_atime_quantum: |
802 | if (match_int(&args[0], &option)) { | |
803 | status = 0; | |
804 | goto bail; | |
805 | } | |
806 | if (option >= 0) | |
c0123ade | 807 | mopt->atime_quantum = option; |
7f1a37e3 | 808 | break; |
baf4661a SM |
809 | case Opt_slot: |
810 | option = 0; | |
811 | if (match_int(&args[0], &option)) { | |
812 | status = 0; | |
813 | goto bail; | |
814 | } | |
815 | if (option) | |
c0123ade | 816 | mopt->slot = (s16)option; |
baf4661a | 817 | break; |
ccd979bd MF |
818 | default: |
819 | mlog(ML_ERROR, | |
820 | "Unrecognized mount option \"%s\" " | |
821 | "or missing value\n", p); | |
822 | status = 0; | |
823 | goto bail; | |
824 | } | |
825 | } | |
826 | ||
827 | status = 1; | |
828 | ||
829 | bail: | |
830 | mlog_exit(status); | |
831 | return status; | |
832 | } | |
833 | ||
d550071c SM |
834 | static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt) |
835 | { | |
836 | struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb); | |
837 | unsigned long opts = osb->s_mount_opt; | |
838 | ||
839 | if (opts & OCFS2_MOUNT_HB_LOCAL) | |
840 | seq_printf(s, ",_netdev,heartbeat=local"); | |
841 | else | |
842 | seq_printf(s, ",heartbeat=none"); | |
843 | ||
844 | if (opts & OCFS2_MOUNT_NOINTR) | |
845 | seq_printf(s, ",nointr"); | |
846 | ||
847 | if (opts & OCFS2_MOUNT_DATA_WRITEBACK) | |
848 | seq_printf(s, ",data=writeback"); | |
849 | else | |
850 | seq_printf(s, ",data=ordered"); | |
851 | ||
852 | if (opts & OCFS2_MOUNT_BARRIER) | |
853 | seq_printf(s, ",barrier=1"); | |
854 | ||
855 | if (opts & OCFS2_MOUNT_ERRORS_PANIC) | |
856 | seq_printf(s, ",errors=panic"); | |
857 | else | |
858 | seq_printf(s, ",errors=remount-ro"); | |
859 | ||
860 | if (osb->preferred_slot != OCFS2_INVALID_SLOT) | |
861 | seq_printf(s, ",preferred_slot=%d", osb->preferred_slot); | |
862 | ||
863 | if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM) | |
864 | seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum); | |
865 | ||
866 | return 0; | |
867 | } | |
868 | ||
ccd979bd MF |
869 | static int __init ocfs2_init(void) |
870 | { | |
871 | int status; | |
872 | ||
873 | mlog_entry_void(); | |
874 | ||
875 | ocfs2_print_version(); | |
876 | ||
ccd979bd MF |
877 | status = init_ocfs2_uptodate_cache(); |
878 | if (status < 0) { | |
879 | mlog_errno(status); | |
880 | goto leave; | |
881 | } | |
882 | ||
883 | status = ocfs2_initialize_mem_caches(); | |
884 | if (status < 0) { | |
885 | mlog_errno(status); | |
886 | goto leave; | |
887 | } | |
888 | ||
889 | ocfs2_wq = create_singlethread_workqueue("ocfs2_wq"); | |
890 | if (!ocfs2_wq) { | |
891 | status = -ENOMEM; | |
892 | goto leave; | |
893 | } | |
894 | ||
ccd979bd MF |
895 | ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL); |
896 | if (!ocfs2_debugfs_root) { | |
897 | status = -EFAULT; | |
898 | mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n"); | |
899 | } | |
900 | ||
901 | leave: | |
902 | if (status < 0) { | |
903 | ocfs2_free_mem_caches(); | |
904 | exit_ocfs2_uptodate_cache(); | |
ccd979bd MF |
905 | } |
906 | ||
907 | mlog_exit(status); | |
908 | ||
909 | if (status >= 0) { | |
910 | return register_filesystem(&ocfs2_fs_type); | |
911 | } else | |
912 | return -1; | |
913 | } | |
914 | ||
915 | static void __exit ocfs2_exit(void) | |
916 | { | |
917 | mlog_entry_void(); | |
918 | ||
919 | if (ocfs2_wq) { | |
920 | flush_workqueue(ocfs2_wq); | |
921 | destroy_workqueue(ocfs2_wq); | |
922 | } | |
923 | ||
924 | debugfs_remove(ocfs2_debugfs_root); | |
925 | ||
926 | ocfs2_free_mem_caches(); | |
927 | ||
928 | unregister_filesystem(&ocfs2_fs_type); | |
929 | ||
ccd979bd MF |
930 | exit_ocfs2_uptodate_cache(); |
931 | ||
932 | mlog_exit_void(); | |
933 | } | |
934 | ||
935 | static void ocfs2_put_super(struct super_block *sb) | |
936 | { | |
937 | mlog_entry("(0x%p)\n", sb); | |
938 | ||
939 | ocfs2_sync_blockdev(sb); | |
940 | ocfs2_dismount_volume(sb, 0); | |
941 | ||
942 | mlog_exit_void(); | |
943 | } | |
944 | ||
726c3342 | 945 | static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf) |
ccd979bd MF |
946 | { |
947 | struct ocfs2_super *osb; | |
948 | u32 numbits, freebits; | |
949 | int status; | |
950 | struct ocfs2_dinode *bm_lock; | |
951 | struct buffer_head *bh = NULL; | |
952 | struct inode *inode = NULL; | |
953 | ||
726c3342 | 954 | mlog_entry("(%p, %p)\n", dentry->d_sb, buf); |
ccd979bd | 955 | |
726c3342 | 956 | osb = OCFS2_SB(dentry->d_sb); |
ccd979bd MF |
957 | |
958 | inode = ocfs2_get_system_file_inode(osb, | |
959 | GLOBAL_BITMAP_SYSTEM_INODE, | |
960 | OCFS2_INVALID_SLOT); | |
961 | if (!inode) { | |
962 | mlog(ML_ERROR, "failed to get bitmap inode\n"); | |
963 | status = -EIO; | |
964 | goto bail; | |
965 | } | |
966 | ||
e63aecb6 | 967 | status = ocfs2_inode_lock(inode, &bh, 0); |
ccd979bd MF |
968 | if (status < 0) { |
969 | mlog_errno(status); | |
970 | goto bail; | |
971 | } | |
972 | ||
973 | bm_lock = (struct ocfs2_dinode *) bh->b_data; | |
974 | ||
975 | numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total); | |
976 | freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used); | |
977 | ||
978 | buf->f_type = OCFS2_SUPER_MAGIC; | |
726c3342 | 979 | buf->f_bsize = dentry->d_sb->s_blocksize; |
ccd979bd MF |
980 | buf->f_namelen = OCFS2_MAX_FILENAME_LEN; |
981 | buf->f_blocks = ((sector_t) numbits) * | |
982 | (osb->s_clustersize >> osb->sb->s_blocksize_bits); | |
983 | buf->f_bfree = ((sector_t) freebits) * | |
984 | (osb->s_clustersize >> osb->sb->s_blocksize_bits); | |
985 | buf->f_bavail = buf->f_bfree; | |
986 | buf->f_files = numbits; | |
987 | buf->f_ffree = freebits; | |
988 | ||
989 | brelse(bh); | |
990 | ||
e63aecb6 | 991 | ocfs2_inode_unlock(inode, 0); |
ccd979bd MF |
992 | status = 0; |
993 | bail: | |
994 | if (inode) | |
995 | iput(inode); | |
996 | ||
997 | mlog_exit(status); | |
998 | ||
999 | return status; | |
1000 | } | |
1001 | ||
4ba9b9d0 | 1002 | static void ocfs2_inode_init_once(struct kmem_cache *cachep, void *data) |
ccd979bd MF |
1003 | { |
1004 | struct ocfs2_inode_info *oi = data; | |
1005 | ||
a35afb83 CL |
1006 | oi->ip_flags = 0; |
1007 | oi->ip_open_count = 0; | |
1008 | spin_lock_init(&oi->ip_lock); | |
1009 | ocfs2_extent_map_init(&oi->vfs_inode); | |
1010 | INIT_LIST_HEAD(&oi->ip_io_markers); | |
1011 | oi->ip_created_trans = 0; | |
1012 | oi->ip_last_trans = 0; | |
1013 | oi->ip_dir_start_lookup = 0; | |
ccd979bd | 1014 | |
a35afb83 CL |
1015 | init_rwsem(&oi->ip_alloc_sem); |
1016 | mutex_init(&oi->ip_io_mutex); | |
ccd979bd | 1017 | |
a35afb83 CL |
1018 | oi->ip_blkno = 0ULL; |
1019 | oi->ip_clusters = 0; | |
ccd979bd | 1020 | |
a35afb83 | 1021 | ocfs2_lock_res_init_once(&oi->ip_rw_lockres); |
e63aecb6 | 1022 | ocfs2_lock_res_init_once(&oi->ip_inode_lockres); |
a35afb83 | 1023 | ocfs2_lock_res_init_once(&oi->ip_open_lockres); |
ccd979bd | 1024 | |
a35afb83 | 1025 | ocfs2_metadata_cache_init(&oi->vfs_inode); |
ccd979bd | 1026 | |
a35afb83 | 1027 | inode_init_once(&oi->vfs_inode); |
ccd979bd MF |
1028 | } |
1029 | ||
1030 | static int ocfs2_initialize_mem_caches(void) | |
1031 | { | |
1032 | ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache", | |
fffb60f9 PJ |
1033 | sizeof(struct ocfs2_inode_info), |
1034 | 0, | |
1035 | (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT| | |
1036 | SLAB_MEM_SPREAD), | |
20c2df83 | 1037 | ocfs2_inode_init_once); |
ccd979bd MF |
1038 | if (!ocfs2_inode_cachep) |
1039 | return -ENOMEM; | |
1040 | ||
ccd979bd MF |
1041 | return 0; |
1042 | } | |
1043 | ||
1044 | static void ocfs2_free_mem_caches(void) | |
1045 | { | |
1046 | if (ocfs2_inode_cachep) | |
1047 | kmem_cache_destroy(ocfs2_inode_cachep); | |
ccd979bd MF |
1048 | |
1049 | ocfs2_inode_cachep = NULL; | |
ccd979bd MF |
1050 | } |
1051 | ||
1052 | static int ocfs2_get_sector(struct super_block *sb, | |
1053 | struct buffer_head **bh, | |
1054 | int block, | |
1055 | int sect_size) | |
1056 | { | |
1057 | if (!sb_set_blocksize(sb, sect_size)) { | |
1058 | mlog(ML_ERROR, "unable to set blocksize\n"); | |
1059 | return -EIO; | |
1060 | } | |
1061 | ||
1062 | *bh = sb_getblk(sb, block); | |
1063 | if (!*bh) { | |
1064 | mlog_errno(-EIO); | |
1065 | return -EIO; | |
1066 | } | |
1067 | lock_buffer(*bh); | |
1068 | if (!buffer_dirty(*bh)) | |
1069 | clear_buffer_uptodate(*bh); | |
1070 | unlock_buffer(*bh); | |
1071 | ll_rw_block(READ, 1, bh); | |
1072 | wait_on_buffer(*bh); | |
1073 | return 0; | |
1074 | } | |
1075 | ||
1076 | /* ocfs2 1.0 only allows one cluster and node identity per kernel image. */ | |
1077 | static int ocfs2_fill_local_node_info(struct ocfs2_super *osb) | |
1078 | { | |
1079 | int status; | |
1080 | ||
1081 | /* XXX hold a ref on the node while mounte? easy enough, if | |
1082 | * desirable. */ | |
c271c5c2 SM |
1083 | if (ocfs2_mount_local(osb)) |
1084 | osb->node_num = 0; | |
1085 | else | |
1086 | osb->node_num = o2nm_this_node(); | |
1087 | ||
ccd979bd MF |
1088 | if (osb->node_num == O2NM_MAX_NODES) { |
1089 | mlog(ML_ERROR, "could not find this host's node number\n"); | |
1090 | status = -ENOENT; | |
1091 | goto bail; | |
1092 | } | |
1093 | ||
781ee3e2 | 1094 | mlog(0, "I am node %d\n", osb->node_num); |
ccd979bd MF |
1095 | |
1096 | status = 0; | |
1097 | bail: | |
1098 | return status; | |
1099 | } | |
1100 | ||
1101 | static int ocfs2_mount_volume(struct super_block *sb) | |
1102 | { | |
1103 | int status = 0; | |
1104 | int unlock_super = 0; | |
1105 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
1106 | ||
1107 | mlog_entry_void(); | |
1108 | ||
1109 | if (ocfs2_is_hard_readonly(osb)) | |
1110 | goto leave; | |
1111 | ||
1112 | status = ocfs2_fill_local_node_info(osb); | |
1113 | if (status < 0) { | |
1114 | mlog_errno(status); | |
1115 | goto leave; | |
1116 | } | |
1117 | ||
ccd979bd MF |
1118 | status = ocfs2_dlm_init(osb); |
1119 | if (status < 0) { | |
1120 | mlog_errno(status); | |
1121 | goto leave; | |
1122 | } | |
1123 | ||
ccd979bd MF |
1124 | status = ocfs2_super_lock(osb, 1); |
1125 | if (status < 0) { | |
1126 | mlog_errno(status); | |
1127 | goto leave; | |
1128 | } | |
1129 | unlock_super = 1; | |
1130 | ||
1131 | /* This will load up the node map and add ourselves to it. */ | |
1132 | status = ocfs2_find_slot(osb); | |
1133 | if (status < 0) { | |
1134 | mlog_errno(status); | |
1135 | goto leave; | |
1136 | } | |
1137 | ||
ccd979bd MF |
1138 | /* load all node-local system inodes */ |
1139 | status = ocfs2_init_local_system_inodes(osb); | |
1140 | if (status < 0) { | |
1141 | mlog_errno(status); | |
1142 | goto leave; | |
1143 | } | |
1144 | ||
1145 | status = ocfs2_check_volume(osb); | |
1146 | if (status < 0) { | |
1147 | mlog_errno(status); | |
1148 | goto leave; | |
1149 | } | |
1150 | ||
1151 | status = ocfs2_truncate_log_init(osb); | |
1152 | if (status < 0) { | |
1153 | mlog_errno(status); | |
1154 | goto leave; | |
1155 | } | |
1156 | ||
c271c5c2 SM |
1157 | if (ocfs2_mount_local(osb)) |
1158 | goto leave; | |
1159 | ||
ccd979bd MF |
1160 | leave: |
1161 | if (unlock_super) | |
1162 | ocfs2_super_unlock(osb, 1); | |
1163 | ||
1164 | mlog_exit(status); | |
1165 | return status; | |
1166 | } | |
1167 | ||
1168 | /* we can't grab the goofy sem lock from inside wait_event, so we use | |
1169 | * memory barriers to make sure that we'll see the null task before | |
1170 | * being woken up */ | |
1171 | static int ocfs2_recovery_thread_running(struct ocfs2_super *osb) | |
1172 | { | |
1173 | mb(); | |
1174 | return osb->recovery_thread_task != NULL; | |
1175 | } | |
1176 | ||
1177 | static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err) | |
1178 | { | |
1179 | int tmp; | |
1180 | struct ocfs2_super *osb = NULL; | |
c271c5c2 | 1181 | char nodestr[8]; |
ccd979bd MF |
1182 | |
1183 | mlog_entry("(0x%p)\n", sb); | |
1184 | ||
1185 | BUG_ON(!sb); | |
1186 | osb = OCFS2_SB(sb); | |
1187 | BUG_ON(!osb); | |
1188 | ||
1189 | ocfs2_shutdown_local_alloc(osb); | |
1190 | ||
1191 | ocfs2_truncate_log_shutdown(osb); | |
1192 | ||
1193 | /* disable any new recovery threads and wait for any currently | |
1194 | * running ones to exit. Do this before setting the vol_state. */ | |
c74ec2f7 | 1195 | mutex_lock(&osb->recovery_lock); |
ccd979bd | 1196 | osb->disable_recovery = 1; |
c74ec2f7 | 1197 | mutex_unlock(&osb->recovery_lock); |
ccd979bd MF |
1198 | wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb)); |
1199 | ||
1200 | /* At this point, we know that no more recovery threads can be | |
1201 | * launched, so wait for any recovery completion work to | |
1202 | * complete. */ | |
1203 | flush_workqueue(ocfs2_wq); | |
1204 | ||
1205 | ocfs2_journal_shutdown(osb); | |
1206 | ||
1207 | ocfs2_sync_blockdev(sb); | |
1208 | ||
1209 | /* No dlm means we've failed during mount, so skip all the | |
1210 | * steps which depended on that to complete. */ | |
1211 | if (osb->dlm) { | |
1212 | tmp = ocfs2_super_lock(osb, 1); | |
1213 | if (tmp < 0) { | |
1214 | mlog_errno(tmp); | |
1215 | return; | |
1216 | } | |
19b613d4 | 1217 | } |
ccd979bd | 1218 | |
19b613d4 MF |
1219 | if (osb->slot_num != OCFS2_INVALID_SLOT) |
1220 | ocfs2_put_slot(osb); | |
ccd979bd | 1221 | |
19b613d4 | 1222 | if (osb->dlm) |
ccd979bd | 1223 | ocfs2_super_unlock(osb, 1); |
ccd979bd MF |
1224 | |
1225 | ocfs2_release_system_inodes(osb); | |
1226 | ||
34d024f8 | 1227 | if (osb->dlm) |
ccd979bd | 1228 | ocfs2_dlm_shutdown(osb); |
ccd979bd | 1229 | |
ccd979bd MF |
1230 | debugfs_remove(osb->osb_debug_root); |
1231 | ||
1232 | if (!mnt_err) | |
1233 | ocfs2_stop_heartbeat(osb); | |
1234 | ||
1235 | atomic_set(&osb->vol_state, VOLUME_DISMOUNTED); | |
1236 | ||
c271c5c2 SM |
1237 | if (ocfs2_mount_local(osb)) |
1238 | snprintf(nodestr, sizeof(nodestr), "local"); | |
1239 | else | |
1240 | snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num); | |
1241 | ||
1242 | printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n", | |
1243 | osb->dev_str, nodestr); | |
ccd979bd MF |
1244 | |
1245 | ocfs2_delete_osb(osb); | |
1246 | kfree(osb); | |
1247 | sb->s_dev = 0; | |
1248 | sb->s_fs_info = NULL; | |
1249 | } | |
1250 | ||
1251 | static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid, | |
1252 | unsigned uuid_bytes) | |
1253 | { | |
1254 | int i, ret; | |
1255 | char *ptr; | |
1256 | ||
1257 | BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN); | |
1258 | ||
cd861280 | 1259 | osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL); |
ccd979bd MF |
1260 | if (osb->uuid_str == NULL) |
1261 | return -ENOMEM; | |
1262 | ||
ccd979bd MF |
1263 | for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) { |
1264 | /* print with null */ | |
1265 | ret = snprintf(ptr, 3, "%02X", uuid[i]); | |
1266 | if (ret != 2) /* drop super cleans up */ | |
1267 | return -EINVAL; | |
1268 | /* then only advance past the last char */ | |
1269 | ptr += 2; | |
1270 | } | |
1271 | ||
1272 | return 0; | |
1273 | } | |
1274 | ||
1275 | static int ocfs2_initialize_super(struct super_block *sb, | |
1276 | struct buffer_head *bh, | |
1277 | int sector_size) | |
1278 | { | |
bddb8eb3 | 1279 | int status; |
5a254031 MF |
1280 | int i, cbits, bbits; |
1281 | struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data; | |
ccd979bd | 1282 | struct inode *inode = NULL; |
ccd979bd MF |
1283 | struct ocfs2_journal *journal; |
1284 | __le32 uuid_net_key; | |
1285 | struct ocfs2_super *osb; | |
1286 | ||
1287 | mlog_entry_void(); | |
1288 | ||
cd861280 | 1289 | osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL); |
ccd979bd MF |
1290 | if (!osb) { |
1291 | status = -ENOMEM; | |
1292 | mlog_errno(status); | |
1293 | goto bail; | |
1294 | } | |
1295 | ||
1296 | sb->s_fs_info = osb; | |
1297 | sb->s_op = &ocfs2_sops; | |
1298 | sb->s_export_op = &ocfs2_export_ops; | |
e0dceaf0 | 1299 | sb->s_time_gran = 1; |
ccd979bd MF |
1300 | sb->s_flags |= MS_NOATIME; |
1301 | /* this is needed to support O_LARGEFILE */ | |
5a254031 MF |
1302 | cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits); |
1303 | bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits); | |
1304 | sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits); | |
ccd979bd MF |
1305 | |
1306 | osb->sb = sb; | |
1307 | /* Save off for ocfs2_rw_direct */ | |
1308 | osb->s_sectsize_bits = blksize_bits(sector_size); | |
ebdec83b | 1309 | BUG_ON(!osb->s_sectsize_bits); |
ccd979bd | 1310 | |
ccd979bd | 1311 | init_waitqueue_head(&osb->recovery_event); |
34d024f8 MF |
1312 | spin_lock_init(&osb->dc_task_lock); |
1313 | init_waitqueue_head(&osb->dc_event); | |
1314 | osb->dc_work_sequence = 0; | |
1315 | osb->dc_wake_sequence = 0; | |
ccd979bd MF |
1316 | INIT_LIST_HEAD(&osb->blocked_lock_list); |
1317 | osb->blocked_lock_count = 0; | |
ccd979bd MF |
1318 | spin_lock_init(&osb->osb_lock); |
1319 | ||
1320 | atomic_set(&osb->alloc_stats.moves, 0); | |
1321 | atomic_set(&osb->alloc_stats.local_data, 0); | |
1322 | atomic_set(&osb->alloc_stats.bitmap_data, 0); | |
1323 | atomic_set(&osb->alloc_stats.bg_allocs, 0); | |
1324 | atomic_set(&osb->alloc_stats.bg_extends, 0); | |
1325 | ||
1326 | ocfs2_init_node_maps(osb); | |
1327 | ||
1328 | snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u", | |
1329 | MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev)); | |
1330 | ||
c74ec2f7 | 1331 | mutex_init(&osb->recovery_lock); |
ccd979bd MF |
1332 | |
1333 | osb->disable_recovery = 0; | |
1334 | osb->recovery_thread_task = NULL; | |
1335 | ||
1336 | init_waitqueue_head(&osb->checkpoint_event); | |
1337 | atomic_set(&osb->needs_checkpoint, 0); | |
1338 | ||
7f1a37e3 TY |
1339 | osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM; |
1340 | ||
ccd979bd MF |
1341 | osb->node_num = O2NM_INVALID_NODE_NUM; |
1342 | osb->slot_num = OCFS2_INVALID_SLOT; | |
1343 | ||
1344 | osb->local_alloc_state = OCFS2_LA_UNUSED; | |
1345 | osb->local_alloc_bh = NULL; | |
1346 | ||
1347 | ocfs2_setup_hb_callbacks(osb); | |
1348 | ||
1349 | init_waitqueue_head(&osb->osb_mount_event); | |
1350 | ||
1351 | osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL); | |
1352 | if (!osb->vol_label) { | |
1353 | mlog(ML_ERROR, "unable to alloc vol label\n"); | |
1354 | status = -ENOMEM; | |
1355 | goto bail; | |
1356 | } | |
1357 | ||
ccd979bd MF |
1358 | osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots); |
1359 | if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) { | |
1360 | mlog(ML_ERROR, "Invalid number of node slots (%u)\n", | |
1361 | osb->max_slots); | |
1362 | status = -EINVAL; | |
1363 | goto bail; | |
1364 | } | |
781ee3e2 | 1365 | mlog(0, "max_slots for this device: %u\n", osb->max_slots); |
ccd979bd | 1366 | |
b4df6ed8 MF |
1367 | init_waitqueue_head(&osb->osb_wipe_event); |
1368 | osb->osb_orphan_wipes = kcalloc(osb->max_slots, | |
1369 | sizeof(*osb->osb_orphan_wipes), | |
1370 | GFP_KERNEL); | |
1371 | if (!osb->osb_orphan_wipes) { | |
1372 | status = -ENOMEM; | |
1373 | mlog_errno(status); | |
1374 | goto bail; | |
1375 | } | |
1376 | ||
ccd979bd MF |
1377 | osb->s_feature_compat = |
1378 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat); | |
1379 | osb->s_feature_ro_compat = | |
1380 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat); | |
1381 | osb->s_feature_incompat = | |
1382 | le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat); | |
1383 | ||
1384 | if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) { | |
1385 | mlog(ML_ERROR, "couldn't mount because of unsupported " | |
1386 | "optional features (%x).\n", i); | |
1387 | status = -EINVAL; | |
1388 | goto bail; | |
1389 | } | |
1390 | if (!(osb->sb->s_flags & MS_RDONLY) && | |
1391 | (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) { | |
1392 | mlog(ML_ERROR, "couldn't mount RDWR because of " | |
1393 | "unsupported optional features (%x).\n", i); | |
1394 | status = -EINVAL; | |
1395 | goto bail; | |
1396 | } | |
1397 | ||
1398 | get_random_bytes(&osb->s_next_generation, sizeof(u32)); | |
1399 | ||
1400 | /* FIXME | |
1401 | * This should be done in ocfs2_journal_init(), but unknown | |
1402 | * ordering issues will cause the filesystem to crash. | |
1403 | * If anyone wants to figure out what part of the code | |
1404 | * refers to osb->journal before ocfs2_journal_init() is run, | |
1405 | * be my guest. | |
1406 | */ | |
1407 | /* initialize our journal structure */ | |
1408 | ||
cd861280 | 1409 | journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL); |
ccd979bd MF |
1410 | if (!journal) { |
1411 | mlog(ML_ERROR, "unable to alloc journal\n"); | |
1412 | status = -ENOMEM; | |
1413 | goto bail; | |
1414 | } | |
1415 | osb->journal = journal; | |
1416 | journal->j_osb = osb; | |
1417 | ||
1418 | atomic_set(&journal->j_num_trans, 0); | |
1419 | init_rwsem(&journal->j_trans_barrier); | |
1420 | init_waitqueue_head(&journal->j_checkpointed); | |
1421 | spin_lock_init(&journal->j_lock); | |
1422 | journal->j_trans_id = (unsigned long) 1; | |
1423 | INIT_LIST_HEAD(&journal->j_la_cleanups); | |
c4028958 | 1424 | INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery); |
ccd979bd MF |
1425 | journal->j_state = OCFS2_JOURNAL_FREE; |
1426 | ||
1427 | /* get some pseudo constants for clustersize bits */ | |
1428 | osb->s_clustersize_bits = | |
1429 | le32_to_cpu(di->id2.i_super.s_clustersize_bits); | |
1430 | osb->s_clustersize = 1 << osb->s_clustersize_bits; | |
1431 | mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits); | |
1432 | ||
1433 | if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE || | |
1434 | osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) { | |
1435 | mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n", | |
1436 | osb->s_clustersize); | |
1437 | status = -EINVAL; | |
1438 | goto bail; | |
1439 | } | |
1440 | ||
1441 | if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1) | |
1442 | > (u32)~0UL) { | |
1443 | mlog(ML_ERROR, "Volume might try to write to blocks beyond " | |
1444 | "what jbd can address in 32 bits.\n"); | |
1445 | status = -EINVAL; | |
1446 | goto bail; | |
1447 | } | |
1448 | ||
1449 | if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid, | |
1450 | sizeof(di->id2.i_super.s_uuid))) { | |
1451 | mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n"); | |
1452 | status = -ENOMEM; | |
1453 | goto bail; | |
1454 | } | |
1455 | ||
78427043 | 1456 | memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key)); |
ccd979bd MF |
1457 | |
1458 | strncpy(osb->vol_label, di->id2.i_super.s_label, 63); | |
1459 | osb->vol_label[63] = '\0'; | |
1460 | osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno); | |
1461 | osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno); | |
1462 | osb->first_cluster_group_blkno = | |
1463 | le64_to_cpu(di->id2.i_super.s_first_cluster_group); | |
1464 | osb->fs_generation = le32_to_cpu(di->i_fs_generation); | |
1465 | mlog(0, "vol_label: %s\n", osb->vol_label); | |
1466 | mlog(0, "uuid: %s\n", osb->uuid_str); | |
b0697053 MF |
1467 | mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n", |
1468 | (unsigned long long)osb->root_blkno, | |
1469 | (unsigned long long)osb->system_dir_blkno); | |
ccd979bd MF |
1470 | |
1471 | osb->osb_dlm_debug = ocfs2_new_dlm_debug(); | |
1472 | if (!osb->osb_dlm_debug) { | |
1473 | status = -ENOMEM; | |
1474 | mlog_errno(status); | |
1475 | goto bail; | |
1476 | } | |
1477 | ||
1478 | atomic_set(&osb->vol_state, VOLUME_INIT); | |
1479 | ||
1480 | /* load root, system_dir, and all global system inodes */ | |
1481 | status = ocfs2_init_global_system_inodes(osb); | |
1482 | if (status < 0) { | |
1483 | mlog_errno(status); | |
1484 | goto bail; | |
1485 | } | |
1486 | ||
1487 | /* | |
1488 | * global bitmap | |
1489 | */ | |
1490 | inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE, | |
1491 | OCFS2_INVALID_SLOT); | |
1492 | if (!inode) { | |
1493 | status = -EINVAL; | |
1494 | mlog_errno(status); | |
1495 | goto bail; | |
1496 | } | |
1497 | ||
1498 | osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno; | |
ccd979bd | 1499 | iput(inode); |
ccd979bd | 1500 | |
e9d578a8 | 1501 | osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8; |
ccd979bd MF |
1502 | |
1503 | status = ocfs2_init_slot_info(osb); | |
1504 | if (status < 0) { | |
1505 | mlog_errno(status); | |
1506 | goto bail; | |
1507 | } | |
1508 | ||
ccd979bd MF |
1509 | bail: |
1510 | mlog_exit(status); | |
1511 | return status; | |
1512 | } | |
1513 | ||
1514 | /* | |
1515 | * will return: -EAGAIN if it is ok to keep searching for superblocks | |
1516 | * -EINVAL if there is a bad superblock | |
1517 | * 0 on success | |
1518 | */ | |
1519 | static int ocfs2_verify_volume(struct ocfs2_dinode *di, | |
1520 | struct buffer_head *bh, | |
1521 | u32 blksz) | |
1522 | { | |
1523 | int status = -EAGAIN; | |
1524 | ||
1525 | mlog_entry_void(); | |
1526 | ||
1527 | if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE, | |
1528 | strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) { | |
1529 | status = -EINVAL; | |
1530 | if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) { | |
1531 | mlog(ML_ERROR, "found superblock with incorrect block " | |
1532 | "size: found %u, should be %u\n", | |
1533 | 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits), | |
1534 | blksz); | |
1535 | } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) != | |
1536 | OCFS2_MAJOR_REV_LEVEL || | |
1537 | le16_to_cpu(di->id2.i_super.s_minor_rev_level) != | |
1538 | OCFS2_MINOR_REV_LEVEL) { | |
1539 | mlog(ML_ERROR, "found superblock with bad version: " | |
1540 | "found %u.%u, should be %u.%u\n", | |
1541 | le16_to_cpu(di->id2.i_super.s_major_rev_level), | |
1542 | le16_to_cpu(di->id2.i_super.s_minor_rev_level), | |
1543 | OCFS2_MAJOR_REV_LEVEL, | |
1544 | OCFS2_MINOR_REV_LEVEL); | |
1545 | } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) { | |
1546 | mlog(ML_ERROR, "bad block number on superblock: " | |
b0697053 | 1547 | "found %llu, should be %llu\n", |
1ca1a111 | 1548 | (unsigned long long)le64_to_cpu(di->i_blkno), |
b0697053 | 1549 | (unsigned long long)bh->b_blocknr); |
ccd979bd MF |
1550 | } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 || |
1551 | le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) { | |
1552 | mlog(ML_ERROR, "bad cluster size found: %u\n", | |
1553 | 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits)); | |
1554 | } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) { | |
1555 | mlog(ML_ERROR, "bad root_blkno: 0\n"); | |
1556 | } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) { | |
1557 | mlog(ML_ERROR, "bad system_dir_blkno: 0\n"); | |
1558 | } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) { | |
1559 | mlog(ML_ERROR, | |
1560 | "Superblock slots found greater than file system " | |
1561 | "maximum: found %u, max %u\n", | |
1562 | le16_to_cpu(di->id2.i_super.s_max_slots), | |
1563 | OCFS2_MAX_SLOTS); | |
1564 | } else { | |
1565 | /* found it! */ | |
1566 | status = 0; | |
1567 | } | |
1568 | } | |
1569 | ||
1570 | mlog_exit(status); | |
1571 | return status; | |
1572 | } | |
1573 | ||
1574 | static int ocfs2_check_volume(struct ocfs2_super *osb) | |
1575 | { | |
bddb8eb3 | 1576 | int status; |
ccd979bd | 1577 | int dirty; |
c271c5c2 | 1578 | int local; |
ccd979bd MF |
1579 | struct ocfs2_dinode *local_alloc = NULL; /* only used if we |
1580 | * recover | |
1581 | * ourselves. */ | |
1582 | ||
1583 | mlog_entry_void(); | |
1584 | ||
1585 | /* Init our journal object. */ | |
1586 | status = ocfs2_journal_init(osb->journal, &dirty); | |
1587 | if (status < 0) { | |
1588 | mlog(ML_ERROR, "Could not initialize journal!\n"); | |
1589 | goto finally; | |
1590 | } | |
1591 | ||
1592 | /* If the journal was unmounted cleanly then we don't want to | |
1593 | * recover anything. Otherwise, journal_load will do that | |
1594 | * dirty work for us :) */ | |
1595 | if (!dirty) { | |
1596 | status = ocfs2_journal_wipe(osb->journal, 0); | |
1597 | if (status < 0) { | |
1598 | mlog_errno(status); | |
1599 | goto finally; | |
1600 | } | |
1601 | } else { | |
1602 | mlog(ML_NOTICE, "File system was not unmounted cleanly, " | |
1603 | "recovering volume.\n"); | |
1604 | } | |
1605 | ||
c271c5c2 SM |
1606 | local = ocfs2_mount_local(osb); |
1607 | ||
ccd979bd | 1608 | /* will play back anything left in the journal. */ |
c271c5c2 | 1609 | ocfs2_journal_load(osb->journal, local); |
ccd979bd MF |
1610 | |
1611 | if (dirty) { | |
1612 | /* recover my local alloc if we didn't unmount cleanly. */ | |
1613 | status = ocfs2_begin_local_alloc_recovery(osb, | |
1614 | osb->slot_num, | |
1615 | &local_alloc); | |
1616 | if (status < 0) { | |
1617 | mlog_errno(status); | |
1618 | goto finally; | |
1619 | } | |
1620 | /* we complete the recovery process after we've marked | |
1621 | * ourselves as mounted. */ | |
1622 | } | |
1623 | ||
1624 | mlog(0, "Journal loaded.\n"); | |
1625 | ||
1626 | status = ocfs2_load_local_alloc(osb); | |
1627 | if (status < 0) { | |
1628 | mlog_errno(status); | |
1629 | goto finally; | |
1630 | } | |
1631 | ||
1632 | if (dirty) { | |
1633 | /* Recovery will be completed after we've mounted the | |
1634 | * rest of the volume. */ | |
1635 | osb->dirty = 1; | |
1636 | osb->local_alloc_copy = local_alloc; | |
1637 | local_alloc = NULL; | |
1638 | } | |
1639 | ||
1640 | /* go through each journal, trylock it and if you get the | |
1641 | * lock, and it's marked as dirty, set the bit in the recover | |
1642 | * map and launch a recovery thread for it. */ | |
1643 | status = ocfs2_mark_dead_nodes(osb); | |
1644 | if (status < 0) | |
1645 | mlog_errno(status); | |
1646 | ||
1647 | finally: | |
1648 | if (local_alloc) | |
1649 | kfree(local_alloc); | |
1650 | ||
1651 | mlog_exit(status); | |
1652 | return status; | |
1653 | } | |
1654 | ||
1655 | /* | |
1656 | * The routine gets called from dismount or close whenever a dismount on | |
1657 | * volume is requested and the osb open count becomes 1. | |
1658 | * It will remove the osb from the global list and also free up all the | |
1659 | * initialized resources and fileobject. | |
1660 | */ | |
1661 | static void ocfs2_delete_osb(struct ocfs2_super *osb) | |
1662 | { | |
1663 | mlog_entry_void(); | |
1664 | ||
1665 | /* This function assumes that the caller has the main osb resource */ | |
1666 | ||
1667 | if (osb->slot_info) | |
1668 | ocfs2_free_slot_info(osb->slot_info); | |
1669 | ||
b4df6ed8 | 1670 | kfree(osb->osb_orphan_wipes); |
ccd979bd MF |
1671 | /* FIXME |
1672 | * This belongs in journal shutdown, but because we have to | |
1673 | * allocate osb->journal at the start of ocfs2_initalize_osb(), | |
1674 | * we free it here. | |
1675 | */ | |
1676 | kfree(osb->journal); | |
1677 | if (osb->local_alloc_copy) | |
1678 | kfree(osb->local_alloc_copy); | |
1679 | kfree(osb->uuid_str); | |
1680 | ocfs2_put_dlm_debug(osb->osb_dlm_debug); | |
1681 | memset(osb, 0, sizeof(struct ocfs2_super)); | |
1682 | ||
1683 | mlog_exit_void(); | |
1684 | } | |
1685 | ||
1686 | /* Put OCFS2 into a readonly state, or (if the user specifies it), | |
1687 | * panic(). We do not support continue-on-error operation. */ | |
1688 | static void ocfs2_handle_error(struct super_block *sb) | |
1689 | { | |
1690 | struct ocfs2_super *osb = OCFS2_SB(sb); | |
1691 | ||
1692 | if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) | |
1693 | panic("OCFS2: (device %s): panic forced after error\n", | |
1694 | sb->s_id); | |
1695 | ||
1696 | ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS); | |
1697 | ||
1698 | if (sb->s_flags & MS_RDONLY && | |
1699 | (ocfs2_is_soft_readonly(osb) || | |
1700 | ocfs2_is_hard_readonly(osb))) | |
1701 | return; | |
1702 | ||
1703 | printk(KERN_CRIT "File system is now read-only due to the potential " | |
1704 | "of on-disk corruption. Please run fsck.ocfs2 once the file " | |
1705 | "system is unmounted.\n"); | |
1706 | sb->s_flags |= MS_RDONLY; | |
1707 | ocfs2_set_ro_flag(osb, 0); | |
1708 | } | |
1709 | ||
1710 | static char error_buf[1024]; | |
1711 | ||
1712 | void __ocfs2_error(struct super_block *sb, | |
1713 | const char *function, | |
1714 | const char *fmt, ...) | |
1715 | { | |
1716 | va_list args; | |
1717 | ||
1718 | va_start(args, fmt); | |
4a6e617a | 1719 | vsnprintf(error_buf, sizeof(error_buf), fmt, args); |
ccd979bd MF |
1720 | va_end(args); |
1721 | ||
1722 | /* Not using mlog here because we want to show the actual | |
1723 | * function the error came from. */ | |
1724 | printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n", | |
1725 | sb->s_id, function, error_buf); | |
1726 | ||
1727 | ocfs2_handle_error(sb); | |
1728 | } | |
1729 | ||
1730 | /* Handle critical errors. This is intentionally more drastic than | |
1731 | * ocfs2_handle_error, so we only use for things like journal errors, | |
1732 | * etc. */ | |
1733 | void __ocfs2_abort(struct super_block* sb, | |
1734 | const char *function, | |
1735 | const char *fmt, ...) | |
1736 | { | |
1737 | va_list args; | |
1738 | ||
1739 | va_start(args, fmt); | |
4a6e617a | 1740 | vsnprintf(error_buf, sizeof(error_buf), fmt, args); |
ccd979bd MF |
1741 | va_end(args); |
1742 | ||
1743 | printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n", | |
1744 | sb->s_id, function, error_buf); | |
1745 | ||
1746 | /* We don't have the cluster support yet to go straight to | |
1747 | * hard readonly in here. Until then, we want to keep | |
1748 | * ocfs2_abort() so that we can at least mark critical | |
1749 | * errors. | |
1750 | * | |
1751 | * TODO: This should abort the journal and alert other nodes | |
1752 | * that our slot needs recovery. */ | |
1753 | ||
1754 | /* Force a panic(). This stinks, but it's better than letting | |
1755 | * things continue without having a proper hard readonly | |
1756 | * here. */ | |
1757 | OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC; | |
1758 | ocfs2_handle_error(sb); | |
1759 | } | |
1760 | ||
1761 | module_init(ocfs2_init); | |
1762 | module_exit(ocfs2_exit); |