?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
GUI: Fix Tomato RAF theme for all builds. Compilation typo.
[tomato.git] / release / src-rt-6.x.4708 / linux / linux-2.6.36 / fs / ocfs2 / super.c
blob180c5224c89bf2137d4e34969f58204d638fad50
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/init.h>
32 #include <linux/random.h>
33 #include <linux/statfs.h>
34 #include <linux/moduleparam.h>
35 #include <linux/blkdev.h>
36 #include <linux/socket.h>
37 #include <linux/inet.h>
38 #include <linux/parser.h>
39 #include <linux/crc32.h>
40 #include <linux/debugfs.h>
41 #include <linux/mount.h>
42 #include <linux/seq_file.h>
43 #include <linux/quotaops.h>
44 #include <linux/smp_lock.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 "blockcheck.h"
56 #include "dlmglue.h"
57 #include "export.h"
58 #include "extent_map.h"
59 #include "heartbeat.h"
60 #include "inode.h"
61 #include "journal.h"
62 #include "localalloc.h"
63 #include "namei.h"
64 #include "slot_map.h"
65 #include "super.h"
66 #include "sysfile.h"
67 #include "uptodate.h"
68 #include "ver.h"
69 #include "xattr.h"
70 #include "quota.h"
71 #include "refcounttree.h"
72 #include "suballoc.h"
73
74 #include "buffer_head_io.h"
75
76 static struct kmem_cache *ocfs2_inode_cachep = NULL;
77 struct kmem_cache *ocfs2_dquot_cachep;
78 struct kmem_cache *ocfs2_qf_chunk_cachep;
79
80 /* OCFS2 needs to schedule several differnt types of work which
81 * require cluster locking, disk I/O, recovery waits, etc. Since these
82 * types of work tend to be heavy we avoid using the kernel events
83 * workqueue and schedule on our own. */
84 struct workqueue_struct *ocfs2_wq = NULL;
85
86 static struct dentry *ocfs2_debugfs_root = NULL;
87
88 MODULE_AUTHOR("Oracle");
89 MODULE_LICENSE("GPL");
90
91 struct mount_options
92 {
93 unsigned long commit_interval;
94 unsigned long mount_opt;
95 unsigned int atime_quantum;
96 signed short slot;
97 int localalloc_opt;
98 unsigned int resv_level;
99 int dir_resv_level;
100 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
101 };
102
103 static int ocfs2_parse_options(struct super_block *sb, char *options,
104 struct mount_options *mopt,
105 int is_remount);
106 static int ocfs2_check_set_options(struct super_block *sb,
107 struct mount_options *options);
108 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
109 static void ocfs2_put_super(struct super_block *sb);
110 static int ocfs2_mount_volume(struct super_block *sb);
111 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
112 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
113 static int ocfs2_initialize_mem_caches(void);
114 static void ocfs2_free_mem_caches(void);
115 static void ocfs2_delete_osb(struct ocfs2_super *osb);
116
117 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
118
119 static int ocfs2_sync_fs(struct super_block *sb, int wait);
120
121 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
122 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
123 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
124 static int ocfs2_check_volume(struct ocfs2_super *osb);
125 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
126 struct buffer_head *bh,
127 u32 sectsize,
128 struct ocfs2_blockcheck_stats *stats);
129 static int ocfs2_initialize_super(struct super_block *sb,
130 struct buffer_head *bh,
131 int sector_size,
132 struct ocfs2_blockcheck_stats *stats);
133 static int ocfs2_get_sector(struct super_block *sb,
134 struct buffer_head **bh,
135 int block,
136 int sect_size);
137 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
138 static void ocfs2_destroy_inode(struct inode *inode);
139 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
140 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
141 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
142
143 static const struct super_operations ocfs2_sops = {
144 .statfs = ocfs2_statfs,
145 .alloc_inode = ocfs2_alloc_inode,
146 .destroy_inode = ocfs2_destroy_inode,
147 .drop_inode = ocfs2_drop_inode,
148 .evict_inode = ocfs2_evict_inode,
149 .sync_fs = ocfs2_sync_fs,
150 .put_super = ocfs2_put_super,
151 .remount_fs = ocfs2_remount,
152 .show_options = ocfs2_show_options,
153 .quota_read = ocfs2_quota_read,
154 .quota_write = ocfs2_quota_write,
155 };
156
157 enum {
158 Opt_barrier,
159 Opt_err_panic,
160 Opt_err_ro,
161 Opt_intr,
162 Opt_nointr,
163 Opt_hb_none,
164 Opt_hb_local,
165 Opt_data_ordered,
166 Opt_data_writeback,
167 Opt_atime_quantum,
168 Opt_slot,
169 Opt_commit,
170 Opt_localalloc,
171 Opt_localflocks,
172 Opt_stack,
173 Opt_user_xattr,
174 Opt_nouser_xattr,
175 Opt_inode64,
176 Opt_acl,
177 Opt_noacl,
178 Opt_usrquota,
179 Opt_grpquota,
180 Opt_resv_level,
181 Opt_dir_resv_level,
182 Opt_err,
183 };
184
185 static const match_table_t tokens = {
186 {Opt_barrier, "barrier=%u"},
187 {Opt_err_panic, "errors=panic"},
188 {Opt_err_ro, "errors=remount-ro"},
189 {Opt_intr, "intr"},
190 {Opt_nointr, "nointr"},
191 {Opt_hb_none, OCFS2_HB_NONE},
192 {Opt_hb_local, OCFS2_HB_LOCAL},
193 {Opt_data_ordered, "data=ordered"},
194 {Opt_data_writeback, "data=writeback"},
195 {Opt_atime_quantum, "atime_quantum=%u"},
196 {Opt_slot, "preferred_slot=%u"},
197 {Opt_commit, "commit=%u"},
198 {Opt_localalloc, "localalloc=%d"},
199 {Opt_localflocks, "localflocks"},
200 {Opt_stack, "cluster_stack=%s"},
201 {Opt_user_xattr, "user_xattr"},
202 {Opt_nouser_xattr, "nouser_xattr"},
203 {Opt_inode64, "inode64"},
204 {Opt_acl, "acl"},
205 {Opt_noacl, "noacl"},
206 {Opt_usrquota, "usrquota"},
207 {Opt_grpquota, "grpquota"},
208 {Opt_resv_level, "resv_level=%u"},
209 {Opt_dir_resv_level, "dir_resv_level=%u"},
210 {Opt_err, NULL}
211 };
212
213 #ifdef CONFIG_DEBUG_FS
214 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
215 {
216 struct ocfs2_cluster_connection *cconn = osb->cconn;
217 struct ocfs2_recovery_map *rm = osb->recovery_map;
218 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
219 int i, out = 0;
220
221 out += snprintf(buf + out, len - out,
222 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
223 "Device", osb->dev_str, osb->uuid_str,
224 osb->fs_generation, osb->vol_label);
225
226 out += snprintf(buf + out, len - out,
227 "%10s => State: %d Flags: 0x%lX\n", "Volume",
228 atomic_read(&osb->vol_state), osb->osb_flags);
229
230 out += snprintf(buf + out, len - out,
231 "%10s => Block: %lu Cluster: %d\n", "Sizes",
232 osb->sb->s_blocksize, osb->s_clustersize);
233
234 out += snprintf(buf + out, len - out,
235 "%10s => Compat: 0x%X Incompat: 0x%X "
236 "ROcompat: 0x%X\n",
237 "Features", osb->s_feature_compat,
238 osb->s_feature_incompat, osb->s_feature_ro_compat);
239
240 out += snprintf(buf + out, len - out,
241 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
242 osb->s_mount_opt, osb->s_atime_quantum);
243
244 if (cconn) {
245 out += snprintf(buf + out, len - out,
246 "%10s => Stack: %s Name: %*s "
247 "Version: %d.%d\n", "Cluster",
248 (*osb->osb_cluster_stack == '\0' ?
249 "o2cb" : osb->osb_cluster_stack),
250 cconn->cc_namelen, cconn->cc_name,
251 cconn->cc_version.pv_major,
252 cconn->cc_version.pv_minor);
253 }
254
255 spin_lock(&osb->dc_task_lock);
256 out += snprintf(buf + out, len - out,
257 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
258 "WorkSeq: %lu\n", "DownCnvt",
259 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
260 osb->blocked_lock_count, osb->dc_wake_sequence,
261 osb->dc_work_sequence);
262 spin_unlock(&osb->dc_task_lock);
263
264 spin_lock(&osb->osb_lock);
265 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
266 "Recovery",
267 (osb->recovery_thread_task ?
268 task_pid_nr(osb->recovery_thread_task) : -1));
269 if (rm->rm_used == 0)
270 out += snprintf(buf + out, len - out, " None\n");
271 else {
272 for (i = 0; i < rm->rm_used; i++)
273 out += snprintf(buf + out, len - out, " %d",
274 rm->rm_entries[i]);
275 out += snprintf(buf + out, len - out, "\n");
276 }
277 spin_unlock(&osb->osb_lock);
278
279 out += snprintf(buf + out, len - out,
280 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
281 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
282 osb->osb_commit_interval,
283 atomic_read(&osb->needs_checkpoint));
284
285 out += snprintf(buf + out, len - out,
286 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
287 "Journal", osb->journal->j_state,
288 osb->journal->j_trans_id,
289 atomic_read(&osb->journal->j_num_trans));
290
291 out += snprintf(buf + out, len - out,
292 "%10s => GlobalAllocs: %d LocalAllocs: %d "
293 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
294 "Stats",
295 atomic_read(&osb->alloc_stats.bitmap_data),
296 atomic_read(&osb->alloc_stats.local_data),
297 atomic_read(&osb->alloc_stats.bg_allocs),
298 atomic_read(&osb->alloc_stats.moves),
299 atomic_read(&osb->alloc_stats.bg_extends));
300
301 out += snprintf(buf + out, len - out,
302 "%10s => State: %u Descriptor: %llu Size: %u bits "
303 "Default: %u bits\n",
304 "LocalAlloc", osb->local_alloc_state,
305 (unsigned long long)osb->la_last_gd,
306 osb->local_alloc_bits, osb->local_alloc_default_bits);
307
308 spin_lock(&osb->osb_lock);
309 out += snprintf(buf + out, len - out,
310 "%10s => InodeSlot: %d StolenInodes: %d, "
311 "MetaSlot: %d StolenMeta: %d\n", "Steal",
312 osb->s_inode_steal_slot,
313 atomic_read(&osb->s_num_inodes_stolen),
314 osb->s_meta_steal_slot,
315 atomic_read(&osb->s_num_meta_stolen));
316 spin_unlock(&osb->osb_lock);
317
318 out += snprintf(buf + out, len - out, "OrphanScan => ");
319 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
320 os->os_count, os->os_seqno);
321 out += snprintf(buf + out, len - out, " Last Scan: ");
322 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
323 out += snprintf(buf + out, len - out, "Disabled\n");
324 else
325 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
326 (get_seconds() - os->os_scantime.tv_sec));
327
328 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
329 "Slots", "Num", "RecoGen");
330 for (i = 0; i < osb->max_slots; ++i) {
331 out += snprintf(buf + out, len - out,
332 "%10s %c %3d %10d\n",
333 " ",
334 (i == osb->slot_num ? '*' : ' '),
335 i, osb->slot_recovery_generations[i]);
336 }
337
338 return out;
339 }
340
341 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
342 {
343 struct ocfs2_super *osb = inode->i_private;
344 char *buf = NULL;
345
346 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
347 if (!buf)
348 goto bail;
349
350 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
351
352 file->private_data = buf;
353
354 return 0;
355 bail:
356 return -ENOMEM;
357 }
358
359 static int ocfs2_debug_release(struct inode *inode, struct file *file)
360 {
361 kfree(file->private_data);
362 return 0;
363 }
364
365 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
366 size_t nbytes, loff_t *ppos)
367 {
368 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
369 i_size_read(file->f_mapping->host));
370 }
371 #else
372 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
373 {
374 return 0;
375 }
376 static int ocfs2_debug_release(struct inode *inode, struct file *file)
377 {
378 return 0;
379 }
380 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
381 size_t nbytes, loff_t *ppos)
382 {
383 return 0;
384 }
385 #endif /* CONFIG_DEBUG_FS */
386
387 static const struct file_operations ocfs2_osb_debug_fops = {
388 .open = ocfs2_osb_debug_open,
389 .release = ocfs2_debug_release,
390 .read = ocfs2_debug_read,
391 .llseek = generic_file_llseek,
392 };
393
394 static int ocfs2_sync_fs(struct super_block *sb, int wait)
395 {
396 int status;
397 tid_t target;
398 struct ocfs2_super *osb = OCFS2_SB(sb);
399
400 if (ocfs2_is_hard_readonly(osb))
401 return -EROFS;
402
403 if (wait) {
404 status = ocfs2_flush_truncate_log(osb);
405 if (status < 0)
406 mlog_errno(status);
407 } else {
408 ocfs2_schedule_truncate_log_flush(osb, 0);
409 }
410
411 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
412 &target)) {
413 if (wait)
414 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
415 target);
416 }
417 return 0;
418 }
419
420 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
421 {
422 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
423 && (ino == USER_QUOTA_SYSTEM_INODE
424 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
425 return 0;
426 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
427 && (ino == GROUP_QUOTA_SYSTEM_INODE
428 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
429 return 0;
430 return 1;
431 }
432
433 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
434 {
435 struct inode *new = NULL;
436 int status = 0;
437 int i;
438
439 mlog_entry_void();
440
441 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
442 if (IS_ERR(new)) {
443 status = PTR_ERR(new);
444 mlog_errno(status);
445 goto bail;
446 }
447 osb->root_inode = new;
448
449 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
450 if (IS_ERR(new)) {
451 status = PTR_ERR(new);
452 mlog_errno(status);
453 goto bail;
454 }
455 osb->sys_root_inode = new;
456
457 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
458 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
459 if (!ocfs2_need_system_inode(osb, i))
460 continue;
461 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
462 if (!new) {
463 ocfs2_release_system_inodes(osb);
464 status = -EINVAL;
465 mlog_errno(status);
466 mlog(ML_ERROR, "Unable to load system inode %d, "
467 "possibly corrupt fs?", i);
468 goto bail;
469 }
470 // the array now has one ref, so drop this one
471 iput(new);
472 }
473
474 bail:
475 mlog_exit(status);
476 return status;
477 }
478
479 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
480 {
481 struct inode *new = NULL;
482 int status = 0;
483 int i;
484
485 mlog_entry_void();
486
487 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
488 i < NUM_SYSTEM_INODES;
489 i++) {
490 if (!ocfs2_need_system_inode(osb, i))
491 continue;
492 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
493 if (!new) {
494 ocfs2_release_system_inodes(osb);
495 status = -EINVAL;
496 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
497 status, i, osb->slot_num);
498 goto bail;
499 }
500 /* the array now has one ref, so drop this one */
501 iput(new);
502 }
503
504 bail:
505 mlog_exit(status);
506 return status;
507 }
508
509 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
510 {
511 int i;
512 struct inode *inode;
513
514 mlog_entry_void();
515
516 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
517 inode = osb->system_inodes[i];
518 if (inode) {
519 iput(inode);
520 osb->system_inodes[i] = NULL;
521 }
522 }
523
524 inode = osb->sys_root_inode;
525 if (inode) {
526 iput(inode);
527 osb->sys_root_inode = NULL;
528 }
529
530 inode = osb->root_inode;
531 if (inode) {
532 iput(inode);
533 osb->root_inode = NULL;
534 }
535
536 mlog_exit(0);
537 }
538
539 /* We're allocating fs objects, use GFP_NOFS */
540 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
541 {
542 struct ocfs2_inode_info *oi;
543
544 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
545 if (!oi)
546 return NULL;
547
548 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
549 return &oi->vfs_inode;
550 }
551
552 static void ocfs2_destroy_inode(struct inode *inode)
553 {
554 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
555 }
556
557 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
558 unsigned int cbits)
559 {
560 unsigned int bytes = 1 << cbits;
561 unsigned int trim = bytes;
562 unsigned int bitshift = 32;
563
564 /*
565 * i_size and all block offsets in ocfs2 are always 64 bits
566 * wide. i_clusters is 32 bits, in cluster-sized units. So on
567 * 64 bit platforms, cluster size will be the limiting factor.
568 */
569
570 #if BITS_PER_LONG == 32
571 # if defined(CONFIG_LBDAF)
572 BUILD_BUG_ON(sizeof(sector_t) != 8);
573 /*
574 * We might be limited by page cache size.
575 */
576 if (bytes > PAGE_CACHE_SIZE) {
577 bytes = PAGE_CACHE_SIZE;
578 trim = 1;
579 /*
580 * Shift by 31 here so that we don't get larger than
581 * MAX_LFS_FILESIZE
582 */
583 bitshift = 31;
584 }
585 # else
586 /*
587 * We are limited by the size of sector_t. Use block size, as
588 * that's what we expose to the VFS.
589 */
590 bytes = 1 << bbits;
591 trim = 1;
592 bitshift = 31;
593 # endif
594 #endif
595
596 /*
597 * Trim by a whole cluster when we can actually approach the
598 * on-disk limits. Otherwise we can overflow i_clusters when
599 * an extent start is at the max offset.
600 */
601 return (((unsigned long long)bytes) << bitshift) - trim;
602 }
603
604 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
605 {
606 int incompat_features;
607 int ret = 0;
608 struct mount_options parsed_options;
609 struct ocfs2_super *osb = OCFS2_SB(sb);
610
611 lock_kernel();
612
613 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
614 !ocfs2_check_set_options(sb, &parsed_options)) {
615 ret = -EINVAL;
616 goto out;
617 }
618
619 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
620 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
621 ret = -EINVAL;
622 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
623 goto out;
624 }
625
626 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
627 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
628 ret = -EINVAL;
629 mlog(ML_ERROR, "Cannot change data mode on remount\n");
630 goto out;
631 }
632
633 /* Probably don't want this on remount; it might
634 * mess with other nodes */
635 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
636 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
637 ret = -EINVAL;
638 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
639 goto out;
640 }
641
642 /* We're going to/from readonly mode. */
643 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
644 /* Disable quota accounting before remounting RO */
645 if (*flags & MS_RDONLY) {
646 ret = ocfs2_susp_quotas(osb, 0);
647 if (ret < 0)
648 goto out;
649 }
650 /* Lock here so the check of HARD_RO and the potential
651 * setting of SOFT_RO is atomic. */
652 spin_lock(&osb->osb_lock);
653 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
654 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
655 ret = -EROFS;
656 goto unlock_osb;
657 }
658
659 if (*flags & MS_RDONLY) {
660 mlog(0, "Going to ro mode.\n");
661 sb->s_flags |= MS_RDONLY;
662 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
663 } else {
664 mlog(0, "Making ro filesystem writeable.\n");
665
666 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
667 mlog(ML_ERROR, "Cannot remount RDWR "
668 "filesystem due to previous errors.\n");
669 ret = -EROFS;
670 goto unlock_osb;
671 }
672 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
673 if (incompat_features) {
674 mlog(ML_ERROR, "Cannot remount RDWR because "
675 "of unsupported optional features "
676 "(%x).\n", incompat_features);
677 ret = -EINVAL;
678 goto unlock_osb;
679 }
680 sb->s_flags &= ~MS_RDONLY;
681 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
682 }
683 unlock_osb:
684 spin_unlock(&osb->osb_lock);
685 /* Enable quota accounting after remounting RW */
686 if (!ret && !(*flags & MS_RDONLY)) {
687 if (sb_any_quota_suspended(sb))
688 ret = ocfs2_susp_quotas(osb, 1);
689 else
690 ret = ocfs2_enable_quotas(osb);
691 if (ret < 0) {
692 /* Return back changes... */
693 spin_lock(&osb->osb_lock);
694 sb->s_flags |= MS_RDONLY;
695 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696 spin_unlock(&osb->osb_lock);
697 goto out;
698 }
699 }
700 }
701
702 if (!ret) {
703 /* Only save off the new mount options in case of a successful
704 * remount. */
705 osb->s_mount_opt = parsed_options.mount_opt;
706 osb->s_atime_quantum = parsed_options.atime_quantum;
707 osb->preferred_slot = parsed_options.slot;
708 if (parsed_options.commit_interval)
709 osb->osb_commit_interval = parsed_options.commit_interval;
710
711 if (!ocfs2_is_hard_readonly(osb))
712 ocfs2_set_journal_params(osb);
713
714 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
715 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
716 MS_POSIXACL : 0);
717 }
718 out:
719 unlock_kernel();
720 return ret;
721 }
722
723 static int ocfs2_sb_probe(struct super_block *sb,
724 struct buffer_head **bh,
725 int *sector_size,
726 struct ocfs2_blockcheck_stats *stats)
727 {
728 int status, tmpstat;
729 struct ocfs1_vol_disk_hdr *hdr;
730 struct ocfs2_dinode *di;
731 int blksize;
732
733 *bh = NULL;
734
735 /* may be > 512 */
736 *sector_size = bdev_logical_block_size(sb->s_bdev);
737 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
738 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
739 *sector_size, OCFS2_MAX_BLOCKSIZE);
740 status = -EINVAL;
741 goto bail;
742 }
743
744 /* Can this really happen? */
745 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
746 *sector_size = OCFS2_MIN_BLOCKSIZE;
747
748 /* check block zero for old format */
749 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
750 if (status < 0) {
751 mlog_errno(status);
752 goto bail;
753 }
754 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
755 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
756 mlog(ML_ERROR, "incompatible version: %u.%u\n",
757 hdr->major_version, hdr->minor_version);
758 status = -EINVAL;
759 }
760 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
761 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
762 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
763 hdr->signature);
764 status = -EINVAL;
765 }
766 brelse(*bh);
767 *bh = NULL;
768 if (status < 0) {
769 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
770 "upgraded before mounting with ocfs v2\n");
771 goto bail;
772 }
773
774 /*
775 * Now check at magic offset for 512, 1024, 2048, 4096
776 * blocksizes. 4096 is the maximum blocksize because it is
777 * the minimum clustersize.
778 */
779 status = -EINVAL;
780 for (blksize = *sector_size;
781 blksize <= OCFS2_MAX_BLOCKSIZE;
782 blksize <<= 1) {
783 tmpstat = ocfs2_get_sector(sb, bh,
784 OCFS2_SUPER_BLOCK_BLKNO,
785 blksize);
786 if (tmpstat < 0) {
787 status = tmpstat;
788 mlog_errno(status);
789 break;
790 }
791 di = (struct ocfs2_dinode *) (*bh)->b_data;
792 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
793 spin_lock_init(&stats->b_lock);
794 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
795 if (tmpstat < 0) {
796 brelse(*bh);
797 *bh = NULL;
798 }
799 if (tmpstat != -EAGAIN) {
800 status = tmpstat;
801 break;
802 }
803 }
804
805 bail:
806 return status;
807 }
808
809 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
810 {
811 if (ocfs2_mount_local(osb)) {
812 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
813 mlog(ML_ERROR, "Cannot heartbeat on a locally "
814 "mounted device.\n");
815 return -EINVAL;
816 }
817 }
818
819 if (ocfs2_userspace_stack(osb)) {
820 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
821 mlog(ML_ERROR, "Userspace stack expected, but "
822 "o2cb heartbeat arguments passed to mount\n");
823 return -EINVAL;
824 }
825 }
826
827 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
828 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
829 !ocfs2_userspace_stack(osb)) {
830 mlog(ML_ERROR, "Heartbeat has to be started to mount "
831 "a read-write clustered device.\n");
832 return -EINVAL;
833 }
834 }
835
836 return 0;
837 }
838
839 /*
840 * If we're using a userspace stack, mount should have passed
841 * a name that matches the disk. If not, mount should not
842 * have passed a stack.
843 */
844 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
845 struct mount_options *mopt)
846 {
847 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
848 mlog(ML_ERROR,
849 "cluster stack passed to mount, but this filesystem "
850 "does not support it\n");
851 return -EINVAL;
852 }
853
854 if (ocfs2_userspace_stack(osb) &&
855 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
856 OCFS2_STACK_LABEL_LEN)) {
857 mlog(ML_ERROR,
858 "cluster stack passed to mount (\"%s\") does not "
859 "match the filesystem (\"%s\")\n",
860 mopt->cluster_stack,
861 osb->osb_cluster_stack);
862 return -EINVAL;
863 }
864
865 return 0;
866 }
867
868 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
869 {
870 int type;
871 struct super_block *sb = osb->sb;
872 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
873 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
874 int status = 0;
875
876 for (type = 0; type < MAXQUOTAS; type++) {
877 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
878 continue;
879 if (unsuspend)
880 status = dquot_resume(sb, type);
881 else {
882 struct ocfs2_mem_dqinfo *oinfo;
883
884 /* Cancel periodic syncing before suspending */
885 oinfo = sb_dqinfo(sb, type)->dqi_priv;
886 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
887 status = dquot_suspend(sb, type);
888 }
889 if (status < 0)
890 break;
891 }
892 if (status < 0)
893 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
894 "remount (error = %d).\n", status);
895 return status;
896 }
897
898 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
899 {
900 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
901 struct super_block *sb = osb->sb;
902 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
903 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
904 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
905 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
906 int status;
907 int type;
908
909 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
910 for (type = 0; type < MAXQUOTAS; type++) {
911 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
912 continue;
913 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
914 osb->slot_num);
915 if (!inode[type]) {
916 status = -ENOENT;
917 goto out_quota_off;
918 }
919 status = dquot_enable(inode[type], type, QFMT_OCFS2,
920 DQUOT_USAGE_ENABLED);
921 if (status < 0)
922 goto out_quota_off;
923 }
924
925 for (type = 0; type < MAXQUOTAS; type++)
926 iput(inode[type]);
927 return 0;
928 out_quota_off:
929 ocfs2_disable_quotas(osb);
930 for (type = 0; type < MAXQUOTAS; type++)
931 iput(inode[type]);
932 mlog_errno(status);
933 return status;
934 }
935
936 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
937 {
938 int type;
939 struct inode *inode;
940 struct super_block *sb = osb->sb;
941 struct ocfs2_mem_dqinfo *oinfo;
942
943 /* We mostly ignore errors in this function because there's not much
944 * we can do when we see them */
945 for (type = 0; type < MAXQUOTAS; type++) {
946 if (!sb_has_quota_loaded(sb, type))
947 continue;
948 /* Cancel periodic syncing before we grab dqonoff_mutex */
949 oinfo = sb_dqinfo(sb, type)->dqi_priv;
950 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
951 inode = igrab(sb->s_dquot.files[type]);
952 /* Turn off quotas. This will remove all dquot structures from
953 * memory and so they will be automatically synced to global
954 * quota files */
955 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
956 DQUOT_LIMITS_ENABLED);
957 if (!inode)
958 continue;
959 iput(inode);
960 }
961 }
962
963 /* Handle quota on quotactl */
964 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
965 char *path)
966 {
967 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
968 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
969
970 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
971 return -EINVAL;
972
973 return dquot_enable(sb_dqopt(sb)->files[type], type,
974 format_id, DQUOT_LIMITS_ENABLED);
975 }
976
977 /* Handle quota off quotactl */
978 static int ocfs2_quota_off(struct super_block *sb, int type)
979 {
980 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
981 }
982
983 static const struct quotactl_ops ocfs2_quotactl_ops = {
984 .quota_on = ocfs2_quota_on,
985 .quota_off = ocfs2_quota_off,
986 .quota_sync = dquot_quota_sync,
987 .get_info = dquot_get_dqinfo,
988 .set_info = dquot_set_dqinfo,
989 .get_dqblk = dquot_get_dqblk,
990 .set_dqblk = dquot_set_dqblk,
991 };
992
993 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
994 {
995 struct dentry *root;
996 int status, sector_size;
997 struct mount_options parsed_options;
998 struct inode *inode = NULL;
999 struct ocfs2_super *osb = NULL;
1000 struct buffer_head *bh = NULL;
1001 char nodestr[8];
1002 struct ocfs2_blockcheck_stats stats;
1003
1004 mlog_entry("%p, %p, %i", sb, data, silent);
1005
1006 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1007 status = -EINVAL;
1008 goto read_super_error;
1009 }
1010
1011 /* probe for superblock */
1012 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1013 if (status < 0) {
1014 mlog(ML_ERROR, "superblock probe failed!\n");
1015 goto read_super_error;
1016 }
1017
1018 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1019 osb = OCFS2_SB(sb);
1020 if (status < 0) {
1021 mlog_errno(status);
1022 goto read_super_error;
1023 }
1024 brelse(bh);
1025 bh = NULL;
1026
1027 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1028 status = -EINVAL;
1029 goto read_super_error;
1030 }
1031 osb->s_mount_opt = parsed_options.mount_opt;
1032 osb->s_atime_quantum = parsed_options.atime_quantum;
1033 osb->preferred_slot = parsed_options.slot;
1034 osb->osb_commit_interval = parsed_options.commit_interval;
1035
1036 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1037 osb->osb_resv_level = parsed_options.resv_level;
1038 osb->osb_dir_resv_level = parsed_options.resv_level;
1039 if (parsed_options.dir_resv_level == -1)
1040 osb->osb_dir_resv_level = parsed_options.resv_level;
1041 else
1042 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1043
1044 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1045 if (status)
1046 goto read_super_error;
1047
1048 sb->s_magic = OCFS2_SUPER_MAGIC;
1049
1050 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1051 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1052
1053 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1054 * heartbeat=none */
1055 if (bdev_read_only(sb->s_bdev)) {
1056 if (!(sb->s_flags & MS_RDONLY)) {
1057 status = -EACCES;
1058 mlog(ML_ERROR, "Readonly device detected but readonly "
1059 "mount was not specified.\n");
1060 goto read_super_error;
1061 }
1062
1063 /* You should not be able to start a local heartbeat
1064 * on a readonly device. */
1065 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1066 status = -EROFS;
1067 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1068 "device.\n");
1069 goto read_super_error;
1070 }
1071
1072 status = ocfs2_check_journals_nolocks(osb);
1073 if (status < 0) {
1074 if (status == -EROFS)
1075 mlog(ML_ERROR, "Recovery required on readonly "
1076 "file system, but write access is "
1077 "unavailable.\n");
1078 else
1079 mlog_errno(status);
1080 goto read_super_error;
1081 }
1082
1083 ocfs2_set_ro_flag(osb, 1);
1084
1085 printk(KERN_NOTICE "Readonly device detected. No cluster "
1086 "services will be utilized for this mount. Recovery "
1087 "will be skipped.\n");
1088 }
1089
1090 if (!ocfs2_is_hard_readonly(osb)) {
1091 if (sb->s_flags & MS_RDONLY)
1092 ocfs2_set_ro_flag(osb, 0);
1093 }
1094
1095 status = ocfs2_verify_heartbeat(osb);
1096 if (status < 0) {
1097 mlog_errno(status);
1098 goto read_super_error;
1099 }
1100
1101 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1102 ocfs2_debugfs_root);
1103 if (!osb->osb_debug_root) {
1104 status = -EINVAL;
1105 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1106 goto read_super_error;
1107 }
1108
1109 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1110 osb->osb_debug_root,
1111 osb,
1112 &ocfs2_osb_debug_fops);
1113 if (!osb->osb_ctxt) {
1114 status = -EINVAL;
1115 mlog_errno(status);
1116 goto read_super_error;
1117 }
1118
1119 if (ocfs2_meta_ecc(osb)) {
1120 status = ocfs2_blockcheck_stats_debugfs_install(
1121 &osb->osb_ecc_stats,
1122 osb->osb_debug_root);
1123 if (status) {
1124 mlog(ML_ERROR,
1125 "Unable to create blockcheck statistics "
1126 "files\n");
1127 goto read_super_error;
1128 }
1129 }
1130
1131 status = ocfs2_mount_volume(sb);
1132 if (osb->root_inode)
1133 inode = igrab(osb->root_inode);
1134
1135 if (status < 0)
1136 goto read_super_error;
1137
1138 if (!inode) {
1139 status = -EIO;
1140 mlog_errno(status);
1141 goto read_super_error;
1142 }
1143
1144 root = d_alloc_root(inode);
1145 if (!root) {
1146 status = -ENOMEM;
1147 mlog_errno(status);
1148 goto read_super_error;
1149 }
1150
1151 sb->s_root = root;
1152
1153 ocfs2_complete_mount_recovery(osb);
1154
1155 if (ocfs2_mount_local(osb))
1156 snprintf(nodestr, sizeof(nodestr), "local");
1157 else
1158 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1159
1160 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1161 "with %s data mode.\n",
1162 osb->dev_str, nodestr, osb->slot_num,
1163 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1164 "ordered");
1165
1166 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1167 wake_up(&osb->osb_mount_event);
1168
1169 /* Now we can initialize quotas because we can afford to wait
1170 * for cluster locks recovery now. That also means that truncation
1171 * log recovery can happen but that waits for proper quota setup */
1172 if (!(sb->s_flags & MS_RDONLY)) {
1173 status = ocfs2_enable_quotas(osb);
1174 if (status < 0) {
1175 /* We have to err-out specially here because
1176 * s_root is already set */
1177 mlog_errno(status);
1178 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1179 wake_up(&osb->osb_mount_event);
1180 mlog_exit(status);
1181 return status;
1182 }
1183 }
1184
1185 ocfs2_complete_quota_recovery(osb);
1186
1187 /* Now we wake up again for processes waiting for quotas */
1188 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1189 wake_up(&osb->osb_mount_event);
1190
1191 /* Start this when the mount is almost sure of being successful */
1192 ocfs2_orphan_scan_start(osb);
1193
1194 mlog_exit(status);
1195 return status;
1196
1197 read_super_error:
1198 brelse(bh);
1199
1200 if (inode)
1201 iput(inode);
1202
1203 if (osb) {
1204 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1205 wake_up(&osb->osb_mount_event);
1206 ocfs2_dismount_volume(sb, 1);
1207 }
1208
1209 mlog_exit(status);
1210 return status;
1211 }
1212
1213 static int ocfs2_get_sb(struct file_system_type *fs_type,
1214 int flags,
1215 const char *dev_name,
1216 void *data,
1217 struct vfsmount *mnt)
1218 {
1219 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1220 mnt);
1221 }
1222
1223 static void ocfs2_kill_sb(struct super_block *sb)
1224 {
1225 struct ocfs2_super *osb = OCFS2_SB(sb);
1226
1227 /* Failed mount? */
1228 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1229 goto out;
1230
1231 /* Prevent further queueing of inode drop events */
1232 spin_lock(&dentry_list_lock);
1233 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1234 spin_unlock(&dentry_list_lock);
1235 /* Wait for work to finish and/or remove it */
1236 cancel_work_sync(&osb->dentry_lock_work);
1237 out:
1238 kill_block_super(sb);
1239 }
1240
1241 static struct file_system_type ocfs2_fs_type = {
1242 .owner = THIS_MODULE,
1243 .name = "ocfs2",
1244 .get_sb = ocfs2_get_sb, /* is this called when we mount
1245 * the fs? */
1246 .kill_sb = ocfs2_kill_sb,
1247
1248 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1249 .next = NULL
1250 };
1251
1252 static int ocfs2_check_set_options(struct super_block *sb,
1253 struct mount_options *options)
1254 {
1255 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1256 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1257 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1258 mlog(ML_ERROR, "User quotas were requested, but this "
1259 "filesystem does not have the feature enabled.\n");
1260 return 0;
1261 }
1262 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1263 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1264 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1265 mlog(ML_ERROR, "Group quotas were requested, but this "
1266 "filesystem does not have the feature enabled.\n");
1267 return 0;
1268 }
1269 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1270 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1271 mlog(ML_ERROR, "ACL support requested but extended attributes "
1272 "feature is not enabled\n");
1273 return 0;
1274 }
1275 /* No ACL setting specified? Use XATTR feature... */
1276 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1277 OCFS2_MOUNT_NO_POSIX_ACL))) {
1278 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1279 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1280 else
1281 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1282 }
1283 return 1;
1284 }
1285
1286 static int ocfs2_parse_options(struct super_block *sb,
1287 char *options,
1288 struct mount_options *mopt,
1289 int is_remount)
1290 {
1291 int status;
1292 char *p;
1293
1294 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1295 options ? options : "(none)");
1296
1297 mopt->commit_interval = 0;
1298 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1299 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1300 mopt->slot = OCFS2_INVALID_SLOT;
1301 mopt->localalloc_opt = -1;
1302 mopt->cluster_stack[0] = '\0';
1303 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1304 mopt->dir_resv_level = -1;
1305
1306 if (!options) {
1307 status = 1;
1308 goto bail;
1309 }
1310
1311 while ((p = strsep(&options, ",")) != NULL) {
1312 int token, option;
1313 substring_t args[MAX_OPT_ARGS];
1314
1315 if (!*p)
1316 continue;
1317
1318 token = match_token(p, tokens, args);
1319 switch (token) {
1320 case Opt_hb_local:
1321 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1322 break;
1323 case Opt_hb_none:
1324 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1325 break;
1326 case Opt_barrier:
1327 if (match_int(&args[0], &option)) {
1328 status = 0;
1329 goto bail;
1330 }
1331 if (option)
1332 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1333 else
1334 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1335 break;
1336 case Opt_intr:
1337 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1338 break;
1339 case Opt_nointr:
1340 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1341 break;
1342 case Opt_err_panic:
1343 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1344 break;
1345 case Opt_err_ro:
1346 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1347 break;
1348 case Opt_data_ordered:
1349 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1350 break;
1351 case Opt_data_writeback:
1352 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1353 break;
1354 case Opt_user_xattr:
1355 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1356 break;
1357 case Opt_nouser_xattr:
1358 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1359 break;
1360 case Opt_atime_quantum:
1361 if (match_int(&args[0], &option)) {
1362 status = 0;
1363 goto bail;
1364 }
1365 if (option >= 0)
1366 mopt->atime_quantum = option;
1367 break;
1368 case Opt_slot:
1369 option = 0;
1370 if (match_int(&args[0], &option)) {
1371 status = 0;
1372 goto bail;
1373 }
1374 if (option)
1375 mopt->slot = (s16)option;
1376 break;
1377 case Opt_commit:
1378 option = 0;
1379 if (match_int(&args[0], &option)) {
1380 status = 0;
1381 goto bail;
1382 }
1383 if (option < 0)
1384 return 0;
1385 if (option == 0)
1386 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1387 mopt->commit_interval = HZ * option;
1388 break;
1389 case Opt_localalloc:
1390 option = 0;
1391 if (match_int(&args[0], &option)) {
1392 status = 0;
1393 goto bail;
1394 }
1395 if (option >= 0)
1396 mopt->localalloc_opt = option;
1397 break;
1398 case Opt_localflocks:
1399 /*
1400 * Changing this during remount could race
1401 * flock() requests, or "unbalance" existing
1402 * ones (e.g., a lock is taken in one mode but
1403 * dropped in the other). If users care enough
1404 * to flip locking modes during remount, we
1405 * could add a "local" flag to individual
1406 * flock structures for proper tracking of
1407 * state.
1408 */
1409 if (!is_remount)
1410 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1411 break;
1412 case Opt_stack:
1413 /* Check both that the option we were passed
1414 * is of the right length and that it is a proper
1415 * string of the right length.
1416 */
1417 if (((args[0].to - args[0].from) !=
1418 OCFS2_STACK_LABEL_LEN) ||
1419 (strnlen(args[0].from,
1420 OCFS2_STACK_LABEL_LEN) !=
1421 OCFS2_STACK_LABEL_LEN)) {
1422 mlog(ML_ERROR,
1423 "Invalid cluster_stack option\n");
1424 status = 0;
1425 goto bail;
1426 }
1427 memcpy(mopt->cluster_stack, args[0].from,
1428 OCFS2_STACK_LABEL_LEN);
1429 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1430 break;
1431 case Opt_inode64:
1432 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1433 break;
1434 case Opt_usrquota:
1435 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1436 break;
1437 case Opt_grpquota:
1438 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1439 break;
1440 case Opt_acl:
1441 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1442 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1443 break;
1444 case Opt_noacl:
1445 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1446 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1447 break;
1448 case Opt_resv_level:
1449 if (is_remount)
1450 break;
1451 if (match_int(&args[0], &option)) {
1452 status = 0;
1453 goto bail;
1454 }
1455 if (option >= OCFS2_MIN_RESV_LEVEL &&
1456 option < OCFS2_MAX_RESV_LEVEL)
1457 mopt->resv_level = option;
1458 break;
1459 case Opt_dir_resv_level:
1460 if (is_remount)
1461 break;
1462 if (match_int(&args[0], &option)) {
1463 status = 0;
1464 goto bail;
1465 }
1466 if (option >= OCFS2_MIN_RESV_LEVEL &&
1467 option < OCFS2_MAX_RESV_LEVEL)
1468 mopt->dir_resv_level = option;
1469 break;
1470 default:
1471 mlog(ML_ERROR,
1472 "Unrecognized mount option \"%s\" "
1473 "or missing value\n", p);
1474 status = 0;
1475 goto bail;
1476 }
1477 }
1478
1479 status = 1;
1480
1481 bail:
1482 mlog_exit(status);
1483 return status;
1484 }
1485
1486 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1487 {
1488 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1489 unsigned long opts = osb->s_mount_opt;
1490 unsigned int local_alloc_megs;
1491
1492 if (opts & OCFS2_MOUNT_HB_LOCAL)
1493 seq_printf(s, ",_netdev,heartbeat=local");
1494 else
1495 seq_printf(s, ",heartbeat=none");
1496
1497 if (opts & OCFS2_MOUNT_NOINTR)
1498 seq_printf(s, ",nointr");
1499
1500 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1501 seq_printf(s, ",data=writeback");
1502 else
1503 seq_printf(s, ",data=ordered");
1504
1505 if (opts & OCFS2_MOUNT_BARRIER)
1506 seq_printf(s, ",barrier=1");
1507
1508 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1509 seq_printf(s, ",errors=panic");
1510 else
1511 seq_printf(s, ",errors=remount-ro");
1512
1513 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1514 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1515
1516 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1517 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1518
1519 if (osb->osb_commit_interval)
1520 seq_printf(s, ",commit=%u",
1521 (unsigned) (osb->osb_commit_interval / HZ));
1522
1523 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1524 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1525 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1526
1527 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1528 seq_printf(s, ",localflocks,");
1529
1530 if (osb->osb_cluster_stack[0])
1531 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1532 osb->osb_cluster_stack);
1533 if (opts & OCFS2_MOUNT_USRQUOTA)
1534 seq_printf(s, ",usrquota");
1535 if (opts & OCFS2_MOUNT_GRPQUOTA)
1536 seq_printf(s, ",grpquota");
1537
1538 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1539 seq_printf(s, ",nouser_xattr");
1540 else
1541 seq_printf(s, ",user_xattr");
1542
1543 if (opts & OCFS2_MOUNT_INODE64)
1544 seq_printf(s, ",inode64");
1545
1546 if (opts & OCFS2_MOUNT_POSIX_ACL)
1547 seq_printf(s, ",acl");
1548 else
1549 seq_printf(s, ",noacl");
1550
1551 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1552 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1553
1554 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1555 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1556
1557 return 0;
1558 }
1559
1560 static int __init ocfs2_init(void)
1561 {
1562 int status;
1563
1564 mlog_entry_void();
1565
1566 ocfs2_print_version();
1567
1568 status = init_ocfs2_uptodate_cache();
1569 if (status < 0) {
1570 mlog_errno(status);
1571 goto leave;
1572 }
1573
1574 status = ocfs2_initialize_mem_caches();
1575 if (status < 0) {
1576 mlog_errno(status);
1577 goto leave;
1578 }
1579
1580 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1581 if (!ocfs2_wq) {
1582 status = -ENOMEM;
1583 goto leave;
1584 }
1585
1586 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1587 if (!ocfs2_debugfs_root) {
1588 status = -EFAULT;
1589 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1590 }
1591
1592 status = ocfs2_quota_setup();
1593 if (status)
1594 goto leave;
1595
1596 ocfs2_set_locking_protocol();
1597
1598 status = register_quota_format(&ocfs2_quota_format);
1599 leave:
1600 if (status < 0) {
1601 ocfs2_quota_shutdown();
1602 ocfs2_free_mem_caches();
1603 exit_ocfs2_uptodate_cache();
1604 }
1605
1606 mlog_exit(status);
1607
1608 if (status >= 0) {
1609 return register_filesystem(&ocfs2_fs_type);
1610 } else
1611 return -1;
1612 }
1613
1614 static void __exit ocfs2_exit(void)
1615 {
1616 mlog_entry_void();
1617
1618 ocfs2_quota_shutdown();
1619
1620 if (ocfs2_wq) {
1621 flush_workqueue(ocfs2_wq);
1622 destroy_workqueue(ocfs2_wq);
1623 }
1624
1625 unregister_quota_format(&ocfs2_quota_format);
1626
1627 debugfs_remove(ocfs2_debugfs_root);
1628
1629 ocfs2_free_mem_caches();
1630
1631 unregister_filesystem(&ocfs2_fs_type);
1632
1633 exit_ocfs2_uptodate_cache();
1634
1635 mlog_exit_void();
1636 }
1637
1638 static void ocfs2_put_super(struct super_block *sb)
1639 {
1640 mlog_entry("(0x%p)\n", sb);
1641
1642 lock_kernel();
1643
1644 ocfs2_sync_blockdev(sb);
1645 ocfs2_dismount_volume(sb, 0);
1646
1647 unlock_kernel();
1648
1649 mlog_exit_void();
1650 }
1651
1652 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1653 {
1654 struct ocfs2_super *osb;
1655 u32 numbits, freebits;
1656 int status;
1657 struct ocfs2_dinode *bm_lock;
1658 struct buffer_head *bh = NULL;
1659 struct inode *inode = NULL;
1660
1661 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1662
1663 osb = OCFS2_SB(dentry->d_sb);
1664
1665 inode = ocfs2_get_system_file_inode(osb,
1666 GLOBAL_BITMAP_SYSTEM_INODE,
1667 OCFS2_INVALID_SLOT);
1668 if (!inode) {
1669 mlog(ML_ERROR, "failed to get bitmap inode\n");
1670 status = -EIO;
1671 goto bail;
1672 }
1673
1674 status = ocfs2_inode_lock(inode, &bh, 0);
1675 if (status < 0) {
1676 mlog_errno(status);
1677 goto bail;
1678 }
1679
1680 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1681
1682 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1683 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1684
1685 buf->f_type = OCFS2_SUPER_MAGIC;
1686 buf->f_bsize = dentry->d_sb->s_blocksize;
1687 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1688 buf->f_blocks = ((sector_t) numbits) *
1689 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1690 buf->f_bfree = ((sector_t) freebits) *
1691 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1692 buf->f_bavail = buf->f_bfree;
1693 buf->f_files = numbits;
1694 buf->f_ffree = freebits;
1695 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1696 & 0xFFFFFFFFUL;
1697 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1698 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1699
1700 brelse(bh);
1701
1702 ocfs2_inode_unlock(inode, 0);
1703 status = 0;
1704 bail:
1705 if (inode)
1706 iput(inode);
1707
1708 mlog_exit(status);
1709
1710 return status;
1711 }
1712
1713 static void ocfs2_inode_init_once(void *data)
1714 {
1715 struct ocfs2_inode_info *oi = data;
1716
1717 oi->ip_flags = 0;
1718 oi->ip_open_count = 0;
1719 spin_lock_init(&oi->ip_lock);
1720 ocfs2_extent_map_init(&oi->vfs_inode);
1721 INIT_LIST_HEAD(&oi->ip_io_markers);
1722 oi->ip_dir_start_lookup = 0;
1723
1724 init_rwsem(&oi->ip_alloc_sem);
1725 init_rwsem(&oi->ip_xattr_sem);
1726 mutex_init(&oi->ip_io_mutex);
1727
1728 oi->ip_blkno = 0ULL;
1729 oi->ip_clusters = 0;
1730
1731 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1732
1733 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1734 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1735 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1736
1737 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1738 &ocfs2_inode_caching_ops);
1739
1740 inode_init_once(&oi->vfs_inode);
1741 }
1742
1743 static int ocfs2_initialize_mem_caches(void)
1744 {
1745 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1746 sizeof(struct ocfs2_inode_info),
1747 0,
1748 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1749 SLAB_MEM_SPREAD),
1750 ocfs2_inode_init_once);
1751 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1752 sizeof(struct ocfs2_dquot),
1753 0,
1754 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1755 SLAB_MEM_SPREAD),
1756 NULL);
1757 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1758 sizeof(struct ocfs2_quota_chunk),
1759 0,
1760 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1761 NULL);
1762 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1763 !ocfs2_qf_chunk_cachep) {
1764 if (ocfs2_inode_cachep)
1765 kmem_cache_destroy(ocfs2_inode_cachep);
1766 if (ocfs2_dquot_cachep)
1767 kmem_cache_destroy(ocfs2_dquot_cachep);
1768 if (ocfs2_qf_chunk_cachep)
1769 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1770 return -ENOMEM;
1771 }
1772
1773 return 0;
1774 }
1775
1776 static void ocfs2_free_mem_caches(void)
1777 {
1778 if (ocfs2_inode_cachep)
1779 kmem_cache_destroy(ocfs2_inode_cachep);
1780 ocfs2_inode_cachep = NULL;
1781
1782 if (ocfs2_dquot_cachep)
1783 kmem_cache_destroy(ocfs2_dquot_cachep);
1784 ocfs2_dquot_cachep = NULL;
1785
1786 if (ocfs2_qf_chunk_cachep)
1787 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1788 ocfs2_qf_chunk_cachep = NULL;
1789 }
1790
1791 static int ocfs2_get_sector(struct super_block *sb,
1792 struct buffer_head **bh,
1793 int block,
1794 int sect_size)
1795 {
1796 if (!sb_set_blocksize(sb, sect_size)) {
1797 mlog(ML_ERROR, "unable to set blocksize\n");
1798 return -EIO;
1799 }
1800
1801 *bh = sb_getblk(sb, block);
1802 if (!*bh) {
1803 mlog_errno(-EIO);
1804 return -EIO;
1805 }
1806 lock_buffer(*bh);
1807 if (!buffer_dirty(*bh))
1808 clear_buffer_uptodate(*bh);
1809 unlock_buffer(*bh);
1810 ll_rw_block(READ, 1, bh);
1811 wait_on_buffer(*bh);
1812 if (!buffer_uptodate(*bh)) {
1813 mlog_errno(-EIO);
1814 brelse(*bh);
1815 *bh = NULL;
1816 return -EIO;
1817 }
1818
1819 return 0;
1820 }
1821
1822 static int ocfs2_mount_volume(struct super_block *sb)
1823 {
1824 int status = 0;
1825 int unlock_super = 0;
1826 struct ocfs2_super *osb = OCFS2_SB(sb);
1827
1828 mlog_entry_void();
1829
1830 if (ocfs2_is_hard_readonly(osb))
1831 goto leave;
1832
1833 status = ocfs2_dlm_init(osb);
1834 if (status < 0) {
1835 mlog_errno(status);
1836 goto leave;
1837 }
1838
1839 status = ocfs2_super_lock(osb, 1);
1840 if (status < 0) {
1841 mlog_errno(status);
1842 goto leave;
1843 }
1844 unlock_super = 1;
1845
1846 /* This will load up the node map and add ourselves to it. */
1847 status = ocfs2_find_slot(osb);
1848 if (status < 0) {
1849 mlog_errno(status);
1850 goto leave;
1851 }
1852
1853 /* load all node-local system inodes */
1854 status = ocfs2_init_local_system_inodes(osb);
1855 if (status < 0) {
1856 mlog_errno(status);
1857 goto leave;
1858 }
1859
1860 status = ocfs2_check_volume(osb);
1861 if (status < 0) {
1862 mlog_errno(status);
1863 goto leave;
1864 }
1865
1866 status = ocfs2_truncate_log_init(osb);
1867 if (status < 0)
1868 mlog_errno(status);
1869
1870 leave:
1871 if (unlock_super)
1872 ocfs2_super_unlock(osb, 1);
1873
1874 mlog_exit(status);
1875 return status;
1876 }
1877
1878 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1879 {
1880 int tmp, hangup_needed = 0;
1881 struct ocfs2_super *osb = NULL;
1882 char nodestr[8];
1883
1884 mlog_entry("(0x%p)\n", sb);
1885
1886 BUG_ON(!sb);
1887 osb = OCFS2_SB(sb);
1888 BUG_ON(!osb);
1889
1890 debugfs_remove(osb->osb_ctxt);
1891
1892 /*
1893 * Flush inode dropping work queue so that deletes are
1894 * performed while the filesystem is still working
1895 */
1896 ocfs2_drop_all_dl_inodes(osb);
1897
1898 /* Orphan scan should be stopped as early as possible */
1899 ocfs2_orphan_scan_stop(osb);
1900
1901 ocfs2_disable_quotas(osb);
1902
1903 ocfs2_shutdown_local_alloc(osb);
1904
1905 ocfs2_truncate_log_shutdown(osb);
1906
1907 /* This will disable recovery and flush any recovery work. */
1908 ocfs2_recovery_exit(osb);
1909
1910 ocfs2_journal_shutdown(osb);
1911
1912 ocfs2_sync_blockdev(sb);
1913
1914 ocfs2_purge_refcount_trees(osb);
1915
1916 /* No cluster connection means we've failed during mount, so skip
1917 * all the steps which depended on that to complete. */
1918 if (osb->cconn) {
1919 tmp = ocfs2_super_lock(osb, 1);
1920 if (tmp < 0) {
1921 mlog_errno(tmp);
1922 return;
1923 }
1924 }
1925
1926 if (osb->slot_num != OCFS2_INVALID_SLOT)
1927 ocfs2_put_slot(osb);
1928
1929 if (osb->cconn)
1930 ocfs2_super_unlock(osb, 1);
1931
1932 ocfs2_release_system_inodes(osb);
1933
1934 /*
1935 * If we're dismounting due to mount error, mount.ocfs2 will clean
1936 * up heartbeat. If we're a local mount, there is no heartbeat.
1937 * If we failed before we got a uuid_str yet, we can't stop
1938 * heartbeat. Otherwise, do it.
1939 */
1940 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1941 hangup_needed = 1;
1942
1943 if (osb->cconn)
1944 ocfs2_dlm_shutdown(osb, hangup_needed);
1945
1946 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1947 debugfs_remove(osb->osb_debug_root);
1948
1949 if (hangup_needed)
1950 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1951
1952 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1953
1954 if (ocfs2_mount_local(osb))
1955 snprintf(nodestr, sizeof(nodestr), "local");
1956 else
1957 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1958
1959 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1960 osb->dev_str, nodestr);
1961
1962 ocfs2_delete_osb(osb);
1963 kfree(osb);
1964 sb->s_dev = 0;
1965 sb->s_fs_info = NULL;
1966 }
1967
1968 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1969 unsigned uuid_bytes)
1970 {
1971 int i, ret;
1972 char *ptr;
1973
1974 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1975
1976 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1977 if (osb->uuid_str == NULL)
1978 return -ENOMEM;
1979
1980 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1981 /* print with null */
1982 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1983 if (ret != 2) /* drop super cleans up */
1984 return -EINVAL;
1985 /* then only advance past the last char */
1986 ptr += 2;
1987 }
1988
1989 return 0;
1990 }
1991
1992 static int ocfs2_initialize_super(struct super_block *sb,
1993 struct buffer_head *bh,
1994 int sector_size,
1995 struct ocfs2_blockcheck_stats *stats)
1996 {
1997 int status;
1998 int i, cbits, bbits;
1999 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2000 struct inode *inode = NULL;
2001 struct ocfs2_journal *journal;
2002 __le32 uuid_net_key;
2003 struct ocfs2_super *osb;
2004
2005 mlog_entry_void();
2006
2007 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2008 if (!osb) {
2009 status = -ENOMEM;
2010 mlog_errno(status);
2011 goto bail;
2012 }
2013
2014 sb->s_fs_info = osb;
2015 sb->s_op = &ocfs2_sops;
2016 sb->s_export_op = &ocfs2_export_ops;
2017 sb->s_qcop = &ocfs2_quotactl_ops;
2018 sb->dq_op = &ocfs2_quota_operations;
2019 sb->s_xattr = ocfs2_xattr_handlers;
2020 sb->s_time_gran = 1;
2021 sb->s_flags |= MS_NOATIME;
2022 /* this is needed to support O_LARGEFILE */
2023 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2024 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2025 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2026
2027 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2028
2029 for (i = 0; i < 3; i++)
2030 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2031 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2032
2033 osb->sb = sb;
2034 /* Save off for ocfs2_rw_direct */
2035 osb->s_sectsize_bits = blksize_bits(sector_size);
2036 BUG_ON(!osb->s_sectsize_bits);
2037
2038 spin_lock_init(&osb->dc_task_lock);
2039 init_waitqueue_head(&osb->dc_event);
2040 osb->dc_work_sequence = 0;
2041 osb->dc_wake_sequence = 0;
2042 INIT_LIST_HEAD(&osb->blocked_lock_list);
2043 osb->blocked_lock_count = 0;
2044 spin_lock_init(&osb->osb_lock);
2045 spin_lock_init(&osb->osb_xattr_lock);
2046 ocfs2_init_steal_slots(osb);
2047
2048 atomic_set(&osb->alloc_stats.moves, 0);
2049 atomic_set(&osb->alloc_stats.local_data, 0);
2050 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2051 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2052 atomic_set(&osb->alloc_stats.bg_extends, 0);
2053
2054 /* Copy the blockcheck stats from the superblock probe */
2055 osb->osb_ecc_stats = *stats;
2056
2057 ocfs2_init_node_maps(osb);
2058
2059 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2060 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2061
2062 ocfs2_orphan_scan_init(osb);
2063
2064 status = ocfs2_recovery_init(osb);
2065 if (status) {
2066 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2067 mlog_errno(status);
2068 goto bail;
2069 }
2070
2071 init_waitqueue_head(&osb->checkpoint_event);
2072 atomic_set(&osb->needs_checkpoint, 0);
2073
2074 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2075
2076 osb->slot_num = OCFS2_INVALID_SLOT;
2077
2078 osb->s_xattr_inline_size = le16_to_cpu(
2079 di->id2.i_super.s_xattr_inline_size);
2080
2081 osb->local_alloc_state = OCFS2_LA_UNUSED;
2082 osb->local_alloc_bh = NULL;
2083 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2084
2085 init_waitqueue_head(&osb->osb_mount_event);
2086
2087 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2088 if (status) {
2089 mlog_errno(status);
2090 goto bail;
2091 }
2092
2093 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2094 if (!osb->vol_label) {
2095 mlog(ML_ERROR, "unable to alloc vol label\n");
2096 status = -ENOMEM;
2097 goto bail;
2098 }
2099
2100 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2101 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2102 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2103 osb->max_slots);
2104 status = -EINVAL;
2105 goto bail;
2106 }
2107 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2108
2109 osb->slot_recovery_generations =
2110 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2111 GFP_KERNEL);
2112 if (!osb->slot_recovery_generations) {
2113 status = -ENOMEM;
2114 mlog_errno(status);
2115 goto bail;
2116 }
2117
2118 init_waitqueue_head(&osb->osb_wipe_event);
2119 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2120 sizeof(*osb->osb_orphan_wipes),
2121 GFP_KERNEL);
2122 if (!osb->osb_orphan_wipes) {
2123 status = -ENOMEM;
2124 mlog_errno(status);
2125 goto bail;
2126 }
2127
2128 osb->osb_rf_lock_tree = RB_ROOT;
2129
2130 osb->s_feature_compat =
2131 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2132 osb->s_feature_ro_compat =
2133 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2134 osb->s_feature_incompat =
2135 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2136
2137 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2138 mlog(ML_ERROR, "couldn't mount because of unsupported "
2139 "optional features (%x).\n", i);
2140 status = -EINVAL;
2141 goto bail;
2142 }
2143 if (!(osb->sb->s_flags & MS_RDONLY) &&
2144 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2145 mlog(ML_ERROR, "couldn't mount RDWR because of "
2146 "unsupported optional features (%x).\n", i);
2147 status = -EINVAL;
2148 goto bail;
2149 }
2150
2151 if (ocfs2_userspace_stack(osb)) {
2152 memcpy(osb->osb_cluster_stack,
2153 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2154 OCFS2_STACK_LABEL_LEN);
2155 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2156 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2157 mlog(ML_ERROR,
2158 "couldn't mount because of an invalid "
2159 "cluster stack label (%s) \n",
2160 osb->osb_cluster_stack);
2161 status = -EINVAL;
2162 goto bail;
2163 }
2164 } else {
2165 /* The empty string is identical with classic tools that
2166 * don't know about s_cluster_info. */
2167 osb->osb_cluster_stack[0] = '\0';
2168 }
2169
2170 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2171
2172 /* initialize our journal structure */
2173
2174 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2175 if (!journal) {
2176 mlog(ML_ERROR, "unable to alloc journal\n");
2177 status = -ENOMEM;
2178 goto bail;
2179 }
2180 osb->journal = journal;
2181 journal->j_osb = osb;
2182
2183 atomic_set(&journal->j_num_trans, 0);
2184 init_rwsem(&journal->j_trans_barrier);
2185 init_waitqueue_head(&journal->j_checkpointed);
2186 spin_lock_init(&journal->j_lock);
2187 journal->j_trans_id = (unsigned long) 1;
2188 INIT_LIST_HEAD(&journal->j_la_cleanups);
2189 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2190 journal->j_state = OCFS2_JOURNAL_FREE;
2191
2192 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2193 osb->dentry_lock_list = NULL;
2194
2195 /* get some pseudo constants for clustersize bits */
2196 osb->s_clustersize_bits =
2197 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2198 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2199 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2200
2201 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2202 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2203 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2204 osb->s_clustersize);
2205 status = -EINVAL;
2206 goto bail;
2207 }
2208
2209 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2210 > (u32)~0UL) {
2211 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2212 "what jbd can address in 32 bits.\n");
2213 status = -EINVAL;
2214 goto bail;
2215 }
2216
2217 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2218 sizeof(di->id2.i_super.s_uuid))) {
2219 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2220 status = -ENOMEM;
2221 goto bail;
2222 }
2223
2224 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2225
2226 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2227 osb->vol_label[63] = '\0';
2228 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2229 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2230 osb->first_cluster_group_blkno =
2231 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2232 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2233 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2234 mlog(0, "vol_label: %s\n", osb->vol_label);
2235 mlog(0, "uuid: %s\n", osb->uuid_str);
2236 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2237 (unsigned long long)osb->root_blkno,
2238 (unsigned long long)osb->system_dir_blkno);
2239
2240 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2241 if (!osb->osb_dlm_debug) {
2242 status = -ENOMEM;
2243 mlog_errno(status);
2244 goto bail;
2245 }
2246
2247 atomic_set(&osb->vol_state, VOLUME_INIT);
2248
2249 /* load root, system_dir, and all global system inodes */
2250 status = ocfs2_init_global_system_inodes(osb);
2251 if (status < 0) {
2252 mlog_errno(status);
2253 goto bail;
2254 }
2255
2256 /*
2257 * global bitmap
2258 */
2259 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2260 OCFS2_INVALID_SLOT);
2261 if (!inode) {
2262 status = -EINVAL;
2263 mlog_errno(status);
2264 goto bail;
2265 }
2266
2267 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2268 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2269 iput(inode);
2270
2271 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2272 osb->s_feature_incompat) * 8;
2273
2274 status = ocfs2_init_slot_info(osb);
2275 if (status < 0) {
2276 mlog_errno(status);
2277 goto bail;
2278 }
2279
2280 bail:
2281 mlog_exit(status);
2282 return status;
2283 }
2284
2285 /*
2286 * will return: -EAGAIN if it is ok to keep searching for superblocks
2287 * -EINVAL if there is a bad superblock
2288 * 0 on success
2289 */
2290 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2291 struct buffer_head *bh,
2292 u32 blksz,
2293 struct ocfs2_blockcheck_stats *stats)
2294 {
2295 int status = -EAGAIN;
2296
2297 mlog_entry_void();
2298
2299 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2300 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2301 /* We have to do a raw check of the feature here */
2302 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2303 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2304 status = ocfs2_block_check_validate(bh->b_data,
2305 bh->b_size,
2306 &di->i_check,
2307 stats);
2308 if (status)
2309 goto out;
2310 }
2311 status = -EINVAL;
2312 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2313 mlog(ML_ERROR, "found superblock with incorrect block "
2314 "size: found %u, should be %u\n",
2315 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2316 blksz);
2317 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2318 OCFS2_MAJOR_REV_LEVEL ||
2319 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2320 OCFS2_MINOR_REV_LEVEL) {
2321 mlog(ML_ERROR, "found superblock with bad version: "
2322 "found %u.%u, should be %u.%u\n",
2323 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2324 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2325 OCFS2_MAJOR_REV_LEVEL,
2326 OCFS2_MINOR_REV_LEVEL);
2327 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2328 mlog(ML_ERROR, "bad block number on superblock: "
2329 "found %llu, should be %llu\n",
2330 (unsigned long long)le64_to_cpu(di->i_blkno),
2331 (unsigned long long)bh->b_blocknr);
2332 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2333 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2334 mlog(ML_ERROR, "bad cluster size found: %u\n",
2335 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2336 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2337 mlog(ML_ERROR, "bad root_blkno: 0\n");
2338 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2339 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2340 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2341 mlog(ML_ERROR,
2342 "Superblock slots found greater than file system "
2343 "maximum: found %u, max %u\n",
2344 le16_to_cpu(di->id2.i_super.s_max_slots),
2345 OCFS2_MAX_SLOTS);
2346 } else {
2347 /* found it! */
2348 status = 0;
2349 }
2350 }
2351
2352 out:
2353 mlog_exit(status);
2354 return status;
2355 }
2356
2357 static int ocfs2_check_volume(struct ocfs2_super *osb)
2358 {
2359 int status;
2360 int dirty;
2361 int local;
2362 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2363 * recover
2364 * ourselves. */
2365
2366 mlog_entry_void();
2367
2368 /* Init our journal object. */
2369 status = ocfs2_journal_init(osb->journal, &dirty);
2370 if (status < 0) {
2371 mlog(ML_ERROR, "Could not initialize journal!\n");
2372 goto finally;
2373 }
2374
2375 /* If the journal was unmounted cleanly then we don't want to
2376 * recover anything. Otherwise, journal_load will do that
2377 * dirty work for us :) */
2378 if (!dirty) {
2379 status = ocfs2_journal_wipe(osb->journal, 0);
2380 if (status < 0) {
2381 mlog_errno(status);
2382 goto finally;
2383 }
2384 } else {
2385 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2386 "recovering volume.\n");
2387 }
2388
2389 local = ocfs2_mount_local(osb);
2390
2391 /* will play back anything left in the journal. */
2392 status = ocfs2_journal_load(osb->journal, local, dirty);
2393 if (status < 0) {
2394 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2395 goto finally;
2396 }
2397
2398 if (dirty) {
2399 /* recover my local alloc if we didn't unmount cleanly. */
2400 status = ocfs2_begin_local_alloc_recovery(osb,
2401 osb->slot_num,
2402 &local_alloc);
2403 if (status < 0) {
2404 mlog_errno(status);
2405 goto finally;
2406 }
2407 /* we complete the recovery process after we've marked
2408 * ourselves as mounted. */
2409 }
2410
2411 mlog(0, "Journal loaded.\n");
2412
2413 status = ocfs2_load_local_alloc(osb);
2414 if (status < 0) {
2415 mlog_errno(status);
2416 goto finally;
2417 }
2418
2419 if (dirty) {
2420 /* Recovery will be completed after we've mounted the
2421 * rest of the volume. */
2422 osb->dirty = 1;
2423 osb->local_alloc_copy = local_alloc;
2424 local_alloc = NULL;
2425 }
2426
2427 /* go through each journal, trylock it and if you get the
2428 * lock, and it's marked as dirty, set the bit in the recover
2429 * map and launch a recovery thread for it. */
2430 status = ocfs2_mark_dead_nodes(osb);
2431 if (status < 0) {
2432 mlog_errno(status);
2433 goto finally;
2434 }
2435
2436 status = ocfs2_compute_replay_slots(osb);
2437 if (status < 0)
2438 mlog_errno(status);
2439
2440 finally:
2441 if (local_alloc)
2442 kfree(local_alloc);
2443
2444 mlog_exit(status);
2445 return status;
2446 }
2447
2448 /*
2449 * The routine gets called from dismount or close whenever a dismount on
2450 * volume is requested and the osb open count becomes 1.
2451 * It will remove the osb from the global list and also free up all the
2452 * initialized resources and fileobject.
2453 */
2454 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2455 {
2456 mlog_entry_void();
2457
2458 /* This function assumes that the caller has the main osb resource */
2459
2460 ocfs2_free_slot_info(osb);
2461
2462 kfree(osb->osb_orphan_wipes);
2463 kfree(osb->slot_recovery_generations);
2464 kfree(osb->journal);
2465 if (osb->local_alloc_copy)
2466 kfree(osb->local_alloc_copy);
2467 kfree(osb->uuid_str);
2468 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2469 memset(osb, 0, sizeof(struct ocfs2_super));
2470
2471 mlog_exit_void();
2472 }
2473
2474 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2475 * panic(). We do not support continue-on-error operation. */
2476 static void ocfs2_handle_error(struct super_block *sb)
2477 {
2478 struct ocfs2_super *osb = OCFS2_SB(sb);
2479
2480 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2481 panic("OCFS2: (device %s): panic forced after error\n",
2482 sb->s_id);
2483
2484 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2485
2486 if (sb->s_flags & MS_RDONLY &&
2487 (ocfs2_is_soft_readonly(osb) ||
2488 ocfs2_is_hard_readonly(osb)))
2489 return;
2490
2491 printk(KERN_CRIT "File system is now read-only due to the potential "
2492 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2493 "system is unmounted.\n");
2494 sb->s_flags |= MS_RDONLY;
2495 ocfs2_set_ro_flag(osb, 0);
2496 }
2497
2498 static char error_buf[1024];
2499
2500 void __ocfs2_error(struct super_block *sb,
2501 const char *function,
2502 const char *fmt, ...)
2503 {
2504 va_list args;
2505
2506 va_start(args, fmt);
2507 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2508 va_end(args);
2509
2510 /* Not using mlog here because we want to show the actual
2511 * function the error came from. */
2512 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2513 sb->s_id, function, error_buf);
2514
2515 ocfs2_handle_error(sb);
2516 }
2517
2518 /* Handle critical errors. This is intentionally more drastic than
2519 * ocfs2_handle_error, so we only use for things like journal errors,
2520 * etc. */
2521 void __ocfs2_abort(struct super_block* sb,
2522 const char *function,
2523 const char *fmt, ...)
2524 {
2525 va_list args;
2526
2527 va_start(args, fmt);
2528 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2529 va_end(args);
2530
2531 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2532 sb->s_id, function, error_buf);
2533
2534 /* We don't have the cluster support yet to go straight to
2535 * hard readonly in here. Until then, we want to keep
2536 * ocfs2_abort() so that we can at least mark critical
2537 * errors.
2538 *
2539 * TODO: This should abort the journal and alert other nodes
2540 * that our slot needs recovery. */
2541
2542 /* Force a panic(). This stinks, but it's better than letting
2543 * things continue without having a proper hard readonly
2544 * here. */
2545 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2546 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2547 ocfs2_handle_error(sb);
2548 }
2549
2550 /*
2551 * Void signal blockers, because in-kernel sigprocmask() only fails
2552 * when SIG_* is wrong.
2553 */
2554 void ocfs2_block_signals(sigset_t *oldset)
2555 {
2556 int rc;
2557 sigset_t blocked;
2558
2559 sigfillset(&blocked);
2560 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2561 BUG_ON(rc);
2562 }
2563
2564 void ocfs2_unblock_signals(sigset_t *oldset)
2565 {
2566 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2567 BUG_ON(rc);
2568 }
2569
2570 module_init(ocfs2_init);
2571 module_exit(ocfs2_exit);
Tomato firmware and modifications.