]> git.ipfire.org Git - thirdparty/kernel/linux.git/blob - fs/ocfs2/super.c
projects / thirdparty / kernel / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
block: remove CONFIG_LBDAF
[thirdparty/kernel/linux.git] / fs / ocfs2 / super.c
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/cleancache.h>
45 #include <linux/signal.h>
46
47 #define CREATE_TRACE_POINTS
48 #include "ocfs2_trace.h"
49
50 #include <cluster/masklog.h>
51
52 #include "ocfs2.h"
53
54 /* this should be the only file to include a version 1 header */
55 #include "ocfs1_fs_compat.h"
56
57 #include "alloc.h"
58 #include "aops.h"
59 #include "blockcheck.h"
60 #include "dlmglue.h"
61 #include "export.h"
62 #include "extent_map.h"
63 #include "heartbeat.h"
64 #include "inode.h"
65 #include "journal.h"
66 #include "localalloc.h"
67 #include "namei.h"
68 #include "slot_map.h"
69 #include "super.h"
70 #include "sysfile.h"
71 #include "uptodate.h"
72 #include "xattr.h"
73 #include "quota.h"
74 #include "refcounttree.h"
75 #include "suballoc.h"
76
77 #include "buffer_head_io.h"
78 #include "filecheck.h"
79
80 static struct kmem_cache *ocfs2_inode_cachep;
81 struct kmem_cache *ocfs2_dquot_cachep;
82 struct kmem_cache *ocfs2_qf_chunk_cachep;
83
84 static struct dentry *ocfs2_debugfs_root;
85
86 MODULE_AUTHOR("Oracle");
87 MODULE_LICENSE("GPL");
88 MODULE_DESCRIPTION("OCFS2 cluster file system");
89
90 struct mount_options
91 {
92 unsigned long commit_interval;
93 unsigned long mount_opt;
94 unsigned int atime_quantum;
95 signed short slot;
96 int localalloc_opt;
97 unsigned int resv_level;
98 int dir_resv_level;
99 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
100 };
101
102 static int ocfs2_parse_options(struct super_block *sb, char *options,
103 struct mount_options *mopt,
104 int is_remount);
105 static int ocfs2_check_set_options(struct super_block *sb,
106 struct mount_options *options);
107 static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
108 static void ocfs2_put_super(struct super_block *sb);
109 static int ocfs2_mount_volume(struct super_block *sb);
110 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112 static int ocfs2_initialize_mem_caches(void);
113 static void ocfs2_free_mem_caches(void);
114 static void ocfs2_delete_osb(struct ocfs2_super *osb);
115
116 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
117
118 static int ocfs2_sync_fs(struct super_block *sb, int wait);
119
120 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
122 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
123 static int ocfs2_check_volume(struct ocfs2_super *osb);
124 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125 struct buffer_head *bh,
126 u32 sectsize,
127 struct ocfs2_blockcheck_stats *stats);
128 static int ocfs2_initialize_super(struct super_block *sb,
129 struct buffer_head *bh,
130 int sector_size,
131 struct ocfs2_blockcheck_stats *stats);
132 static int ocfs2_get_sector(struct super_block *sb,
133 struct buffer_head **bh,
134 int block,
135 int sect_size);
136 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137 static void ocfs2_destroy_inode(struct inode *inode);
138 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
141
142 static struct dquot **ocfs2_get_dquots(struct inode *inode)
143 {
144 return OCFS2_I(inode)->i_dquot;
145 }
146
147 static const struct super_operations ocfs2_sops = {
148 .statfs = ocfs2_statfs,
149 .alloc_inode = ocfs2_alloc_inode,
150 .destroy_inode = ocfs2_destroy_inode,
151 .drop_inode = ocfs2_drop_inode,
152 .evict_inode = ocfs2_evict_inode,
153 .sync_fs = ocfs2_sync_fs,
154 .put_super = ocfs2_put_super,
155 .remount_fs = ocfs2_remount,
156 .show_options = ocfs2_show_options,
157 .quota_read = ocfs2_quota_read,
158 .quota_write = ocfs2_quota_write,
159 .get_dquots = ocfs2_get_dquots,
160 };
161
162 enum {
163 Opt_barrier,
164 Opt_err_panic,
165 Opt_err_ro,
166 Opt_intr,
167 Opt_nointr,
168 Opt_hb_none,
169 Opt_hb_local,
170 Opt_hb_global,
171 Opt_data_ordered,
172 Opt_data_writeback,
173 Opt_atime_quantum,
174 Opt_slot,
175 Opt_commit,
176 Opt_localalloc,
177 Opt_localflocks,
178 Opt_stack,
179 Opt_user_xattr,
180 Opt_nouser_xattr,
181 Opt_inode64,
182 Opt_acl,
183 Opt_noacl,
184 Opt_usrquota,
185 Opt_grpquota,
186 Opt_coherency_buffered,
187 Opt_coherency_full,
188 Opt_resv_level,
189 Opt_dir_resv_level,
190 Opt_journal_async_commit,
191 Opt_err_cont,
192 Opt_err,
193 };
194
195 static const match_table_t tokens = {
196 {Opt_barrier, "barrier=%u"},
197 {Opt_err_panic, "errors=panic"},
198 {Opt_err_ro, "errors=remount-ro"},
199 {Opt_intr, "intr"},
200 {Opt_nointr, "nointr"},
201 {Opt_hb_none, OCFS2_HB_NONE},
202 {Opt_hb_local, OCFS2_HB_LOCAL},
203 {Opt_hb_global, OCFS2_HB_GLOBAL},
204 {Opt_data_ordered, "data=ordered"},
205 {Opt_data_writeback, "data=writeback"},
206 {Opt_atime_quantum, "atime_quantum=%u"},
207 {Opt_slot, "preferred_slot=%u"},
208 {Opt_commit, "commit=%u"},
209 {Opt_localalloc, "localalloc=%d"},
210 {Opt_localflocks, "localflocks"},
211 {Opt_stack, "cluster_stack=%s"},
212 {Opt_user_xattr, "user_xattr"},
213 {Opt_nouser_xattr, "nouser_xattr"},
214 {Opt_inode64, "inode64"},
215 {Opt_acl, "acl"},
216 {Opt_noacl, "noacl"},
217 {Opt_usrquota, "usrquota"},
218 {Opt_grpquota, "grpquota"},
219 {Opt_coherency_buffered, "coherency=buffered"},
220 {Opt_coherency_full, "coherency=full"},
221 {Opt_resv_level, "resv_level=%u"},
222 {Opt_dir_resv_level, "dir_resv_level=%u"},
223 {Opt_journal_async_commit, "journal_async_commit"},
224 {Opt_err_cont, "errors=continue"},
225 {Opt_err, NULL}
226 };
227
228 #ifdef CONFIG_DEBUG_FS
229 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
230 {
231 struct ocfs2_cluster_connection *cconn = osb->cconn;
232 struct ocfs2_recovery_map *rm = osb->recovery_map;
233 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
234 int i, out = 0;
235 unsigned long flags;
236
237 out += snprintf(buf + out, len - out,
238 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
239 "Device", osb->dev_str, osb->uuid_str,
240 osb->fs_generation, osb->vol_label);
241
242 out += snprintf(buf + out, len - out,
243 "%10s => State: %d Flags: 0x%lX\n", "Volume",
244 atomic_read(&osb->vol_state), osb->osb_flags);
245
246 out += snprintf(buf + out, len - out,
247 "%10s => Block: %lu Cluster: %d\n", "Sizes",
248 osb->sb->s_blocksize, osb->s_clustersize);
249
250 out += snprintf(buf + out, len - out,
251 "%10s => Compat: 0x%X Incompat: 0x%X "
252 "ROcompat: 0x%X\n",
253 "Features", osb->s_feature_compat,
254 osb->s_feature_incompat, osb->s_feature_ro_compat);
255
256 out += snprintf(buf + out, len - out,
257 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
258 osb->s_mount_opt, osb->s_atime_quantum);
259
260 if (cconn) {
261 out += snprintf(buf + out, len - out,
262 "%10s => Stack: %s Name: %*s "
263 "Version: %d.%d\n", "Cluster",
264 (*osb->osb_cluster_stack == '\0' ?
265 "o2cb" : osb->osb_cluster_stack),
266 cconn->cc_namelen, cconn->cc_name,
267 cconn->cc_version.pv_major,
268 cconn->cc_version.pv_minor);
269 }
270
271 spin_lock_irqsave(&osb->dc_task_lock, flags);
272 out += snprintf(buf + out, len - out,
273 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
274 "WorkSeq: %lu\n", "DownCnvt",
275 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
276 osb->blocked_lock_count, osb->dc_wake_sequence,
277 osb->dc_work_sequence);
278 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
279
280 spin_lock(&osb->osb_lock);
281 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
282 "Recovery",
283 (osb->recovery_thread_task ?
284 task_pid_nr(osb->recovery_thread_task) : -1));
285 if (rm->rm_used == 0)
286 out += snprintf(buf + out, len - out, " None\n");
287 else {
288 for (i = 0; i < rm->rm_used; i++)
289 out += snprintf(buf + out, len - out, " %d",
290 rm->rm_entries[i]);
291 out += snprintf(buf + out, len - out, "\n");
292 }
293 spin_unlock(&osb->osb_lock);
294
295 out += snprintf(buf + out, len - out,
296 "%10s => Pid: %d Interval: %lu\n", "Commit",
297 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
298 osb->osb_commit_interval);
299
300 out += snprintf(buf + out, len - out,
301 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
302 "Journal", osb->journal->j_state,
303 osb->journal->j_trans_id,
304 atomic_read(&osb->journal->j_num_trans));
305
306 out += snprintf(buf + out, len - out,
307 "%10s => GlobalAllocs: %d LocalAllocs: %d "
308 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
309 "Stats",
310 atomic_read(&osb->alloc_stats.bitmap_data),
311 atomic_read(&osb->alloc_stats.local_data),
312 atomic_read(&osb->alloc_stats.bg_allocs),
313 atomic_read(&osb->alloc_stats.moves),
314 atomic_read(&osb->alloc_stats.bg_extends));
315
316 out += snprintf(buf + out, len - out,
317 "%10s => State: %u Descriptor: %llu Size: %u bits "
318 "Default: %u bits\n",
319 "LocalAlloc", osb->local_alloc_state,
320 (unsigned long long)osb->la_last_gd,
321 osb->local_alloc_bits, osb->local_alloc_default_bits);
322
323 spin_lock(&osb->osb_lock);
324 out += snprintf(buf + out, len - out,
325 "%10s => InodeSlot: %d StolenInodes: %d, "
326 "MetaSlot: %d StolenMeta: %d\n", "Steal",
327 osb->s_inode_steal_slot,
328 atomic_read(&osb->s_num_inodes_stolen),
329 osb->s_meta_steal_slot,
330 atomic_read(&osb->s_num_meta_stolen));
331 spin_unlock(&osb->osb_lock);
332
333 out += snprintf(buf + out, len - out, "OrphanScan => ");
334 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
335 os->os_count, os->os_seqno);
336 out += snprintf(buf + out, len - out, " Last Scan: ");
337 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
338 out += snprintf(buf + out, len - out, "Disabled\n");
339 else
340 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
341 (unsigned long)(ktime_get_seconds() - os->os_scantime));
342
343 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
344 "Slots", "Num", "RecoGen");
345 for (i = 0; i < osb->max_slots; ++i) {
346 out += snprintf(buf + out, len - out,
347 "%10s %c %3d %10d\n",
348 " ",
349 (i == osb->slot_num ? '*' : ' '),
350 i, osb->slot_recovery_generations[i]);
351 }
352
353 return out;
354 }
355
356 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
357 {
358 struct ocfs2_super *osb = inode->i_private;
359 char *buf = NULL;
360
361 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
362 if (!buf)
363 goto bail;
364
365 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
366
367 file->private_data = buf;
368
369 return 0;
370 bail:
371 return -ENOMEM;
372 }
373
374 static int ocfs2_debug_release(struct inode *inode, struct file *file)
375 {
376 kfree(file->private_data);
377 return 0;
378 }
379
380 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
381 size_t nbytes, loff_t *ppos)
382 {
383 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
384 i_size_read(file->f_mapping->host));
385 }
386 #else
387 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
388 {
389 return 0;
390 }
391 static int ocfs2_debug_release(struct inode *inode, struct file *file)
392 {
393 return 0;
394 }
395 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
396 size_t nbytes, loff_t *ppos)
397 {
398 return 0;
399 }
400 #endif /* CONFIG_DEBUG_FS */
401
402 static const struct file_operations ocfs2_osb_debug_fops = {
403 .open = ocfs2_osb_debug_open,
404 .release = ocfs2_debug_release,
405 .read = ocfs2_debug_read,
406 .llseek = generic_file_llseek,
407 };
408
409 static int ocfs2_sync_fs(struct super_block *sb, int wait)
410 {
411 int status;
412 tid_t target;
413 struct ocfs2_super *osb = OCFS2_SB(sb);
414
415 if (ocfs2_is_hard_readonly(osb))
416 return -EROFS;
417
418 if (wait) {
419 status = ocfs2_flush_truncate_log(osb);
420 if (status < 0)
421 mlog_errno(status);
422 } else {
423 ocfs2_schedule_truncate_log_flush(osb, 0);
424 }
425
426 if (jbd2_journal_start_commit(osb->journal->j_journal,
427 &target)) {
428 if (wait)
429 jbd2_log_wait_commit(osb->journal->j_journal,
430 target);
431 }
432 return 0;
433 }
434
435 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
436 {
437 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
438 && (ino == USER_QUOTA_SYSTEM_INODE
439 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
440 return 0;
441 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
442 && (ino == GROUP_QUOTA_SYSTEM_INODE
443 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
444 return 0;
445 return 1;
446 }
447
448 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
449 {
450 struct inode *new = NULL;
451 int status = 0;
452 int i;
453
454 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
455 if (IS_ERR(new)) {
456 status = PTR_ERR(new);
457 mlog_errno(status);
458 goto bail;
459 }
460 osb->root_inode = new;
461
462 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
463 if (IS_ERR(new)) {
464 status = PTR_ERR(new);
465 mlog_errno(status);
466 goto bail;
467 }
468 osb->sys_root_inode = new;
469
470 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
471 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
472 if (!ocfs2_need_system_inode(osb, i))
473 continue;
474 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
475 if (!new) {
476 ocfs2_release_system_inodes(osb);
477 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
478 mlog_errno(status);
479 mlog(ML_ERROR, "Unable to load system inode %d, "
480 "possibly corrupt fs?", i);
481 goto bail;
482 }
483 // the array now has one ref, so drop this one
484 iput(new);
485 }
486
487 bail:
488 if (status)
489 mlog_errno(status);
490 return status;
491 }
492
493 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
494 {
495 struct inode *new = NULL;
496 int status = 0;
497 int i;
498
499 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
500 i < NUM_SYSTEM_INODES;
501 i++) {
502 if (!ocfs2_need_system_inode(osb, i))
503 continue;
504 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
505 if (!new) {
506 ocfs2_release_system_inodes(osb);
507 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
508 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
509 status, i, osb->slot_num);
510 goto bail;
511 }
512 /* the array now has one ref, so drop this one */
513 iput(new);
514 }
515
516 bail:
517 if (status)
518 mlog_errno(status);
519 return status;
520 }
521
522 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
523 {
524 int i;
525 struct inode *inode;
526
527 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
528 inode = osb->global_system_inodes[i];
529 if (inode) {
530 iput(inode);
531 osb->global_system_inodes[i] = NULL;
532 }
533 }
534
535 inode = osb->sys_root_inode;
536 if (inode) {
537 iput(inode);
538 osb->sys_root_inode = NULL;
539 }
540
541 inode = osb->root_inode;
542 if (inode) {
543 iput(inode);
544 osb->root_inode = NULL;
545 }
546
547 if (!osb->local_system_inodes)
548 return;
549
550 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
551 if (osb->local_system_inodes[i]) {
552 iput(osb->local_system_inodes[i]);
553 osb->local_system_inodes[i] = NULL;
554 }
555 }
556
557 kfree(osb->local_system_inodes);
558 osb->local_system_inodes = NULL;
559 }
560
561 /* We're allocating fs objects, use GFP_NOFS */
562 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
563 {
564 struct ocfs2_inode_info *oi;
565
566 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
567 if (!oi)
568 return NULL;
569
570 oi->i_sync_tid = 0;
571 oi->i_datasync_tid = 0;
572 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
573
574 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
575 return &oi->vfs_inode;
576 }
577
578 static void ocfs2_i_callback(struct rcu_head *head)
579 {
580 struct inode *inode = container_of(head, struct inode, i_rcu);
581 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
582 }
583
584 static void ocfs2_destroy_inode(struct inode *inode)
585 {
586 call_rcu(&inode->i_rcu, ocfs2_i_callback);
587 }
588
589 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
590 unsigned int cbits)
591 {
592 unsigned int bytes = 1 << cbits;
593 unsigned int trim = bytes;
594 unsigned int bitshift = 32;
595
596 /*
597 * i_size and all block offsets in ocfs2 are always 64 bits
598 * wide. i_clusters is 32 bits, in cluster-sized units. So on
599 * 64 bit platforms, cluster size will be the limiting factor.
600 */
601
602 #if BITS_PER_LONG == 32
603 BUILD_BUG_ON(sizeof(sector_t) != 8);
604 /*
605 * We might be limited by page cache size.
606 */
607 if (bytes > PAGE_SIZE) {
608 bytes = PAGE_SIZE;
609 trim = 1;
610 /*
611 * Shift by 31 here so that we don't get larger than
612 * MAX_LFS_FILESIZE
613 */
614 bitshift = 31;
615 }
616 #endif
617
618 /*
619 * Trim by a whole cluster when we can actually approach the
620 * on-disk limits. Otherwise we can overflow i_clusters when
621 * an extent start is at the max offset.
622 */
623 return (((unsigned long long)bytes) << bitshift) - trim;
624 }
625
626 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
627 {
628 int incompat_features;
629 int ret = 0;
630 struct mount_options parsed_options;
631 struct ocfs2_super *osb = OCFS2_SB(sb);
632 u32 tmp;
633
634 sync_filesystem(sb);
635
636 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
637 !ocfs2_check_set_options(sb, &parsed_options)) {
638 ret = -EINVAL;
639 goto out;
640 }
641
642 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
643 OCFS2_MOUNT_HB_NONE;
644 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
645 ret = -EINVAL;
646 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
647 goto out;
648 }
649
650 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
651 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
652 ret = -EINVAL;
653 mlog(ML_ERROR, "Cannot change data mode on remount\n");
654 goto out;
655 }
656
657 /* Probably don't want this on remount; it might
658 * mess with other nodes */
659 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
660 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
661 ret = -EINVAL;
662 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
663 goto out;
664 }
665
666 /* We're going to/from readonly mode. */
667 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
668 /* Disable quota accounting before remounting RO */
669 if (*flags & SB_RDONLY) {
670 ret = ocfs2_susp_quotas(osb, 0);
671 if (ret < 0)
672 goto out;
673 }
674 /* Lock here so the check of HARD_RO and the potential
675 * setting of SOFT_RO is atomic. */
676 spin_lock(&osb->osb_lock);
677 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
678 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
679 ret = -EROFS;
680 goto unlock_osb;
681 }
682
683 if (*flags & SB_RDONLY) {
684 sb->s_flags |= SB_RDONLY;
685 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
686 } else {
687 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
688 mlog(ML_ERROR, "Cannot remount RDWR "
689 "filesystem due to previous errors.\n");
690 ret = -EROFS;
691 goto unlock_osb;
692 }
693 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
694 if (incompat_features) {
695 mlog(ML_ERROR, "Cannot remount RDWR because "
696 "of unsupported optional features "
697 "(%x).\n", incompat_features);
698 ret = -EINVAL;
699 goto unlock_osb;
700 }
701 sb->s_flags &= ~SB_RDONLY;
702 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
703 }
704 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
705 unlock_osb:
706 spin_unlock(&osb->osb_lock);
707 /* Enable quota accounting after remounting RW */
708 if (!ret && !(*flags & SB_RDONLY)) {
709 if (sb_any_quota_suspended(sb))
710 ret = ocfs2_susp_quotas(osb, 1);
711 else
712 ret = ocfs2_enable_quotas(osb);
713 if (ret < 0) {
714 /* Return back changes... */
715 spin_lock(&osb->osb_lock);
716 sb->s_flags |= SB_RDONLY;
717 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
718 spin_unlock(&osb->osb_lock);
719 goto out;
720 }
721 }
722 }
723
724 if (!ret) {
725 /* Only save off the new mount options in case of a successful
726 * remount. */
727 osb->s_mount_opt = parsed_options.mount_opt;
728 osb->s_atime_quantum = parsed_options.atime_quantum;
729 osb->preferred_slot = parsed_options.slot;
730 if (parsed_options.commit_interval)
731 osb->osb_commit_interval = parsed_options.commit_interval;
732
733 if (!ocfs2_is_hard_readonly(osb))
734 ocfs2_set_journal_params(osb);
735
736 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
737 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
738 SB_POSIXACL : 0);
739 }
740 out:
741 return ret;
742 }
743
744 static int ocfs2_sb_probe(struct super_block *sb,
745 struct buffer_head **bh,
746 int *sector_size,
747 struct ocfs2_blockcheck_stats *stats)
748 {
749 int status, tmpstat;
750 struct ocfs1_vol_disk_hdr *hdr;
751 struct ocfs2_dinode *di;
752 int blksize;
753
754 *bh = NULL;
755
756 /* may be > 512 */
757 *sector_size = bdev_logical_block_size(sb->s_bdev);
758 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
759 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
760 *sector_size, OCFS2_MAX_BLOCKSIZE);
761 status = -EINVAL;
762 goto bail;
763 }
764
765 /* Can this really happen? */
766 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
767 *sector_size = OCFS2_MIN_BLOCKSIZE;
768
769 /* check block zero for old format */
770 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
771 if (status < 0) {
772 mlog_errno(status);
773 goto bail;
774 }
775 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
776 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
777 mlog(ML_ERROR, "incompatible version: %u.%u\n",
778 hdr->major_version, hdr->minor_version);
779 status = -EINVAL;
780 }
781 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
782 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
783 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
784 hdr->signature);
785 status = -EINVAL;
786 }
787 brelse(*bh);
788 *bh = NULL;
789 if (status < 0) {
790 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
791 "upgraded before mounting with ocfs v2\n");
792 goto bail;
793 }
794
795 /*
796 * Now check at magic offset for 512, 1024, 2048, 4096
797 * blocksizes. 4096 is the maximum blocksize because it is
798 * the minimum clustersize.
799 */
800 status = -EINVAL;
801 for (blksize = *sector_size;
802 blksize <= OCFS2_MAX_BLOCKSIZE;
803 blksize <<= 1) {
804 tmpstat = ocfs2_get_sector(sb, bh,
805 OCFS2_SUPER_BLOCK_BLKNO,
806 blksize);
807 if (tmpstat < 0) {
808 status = tmpstat;
809 mlog_errno(status);
810 break;
811 }
812 di = (struct ocfs2_dinode *) (*bh)->b_data;
813 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
814 spin_lock_init(&stats->b_lock);
815 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
816 if (tmpstat < 0) {
817 brelse(*bh);
818 *bh = NULL;
819 }
820 if (tmpstat != -EAGAIN) {
821 status = tmpstat;
822 break;
823 }
824 }
825
826 bail:
827 return status;
828 }
829
830 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
831 {
832 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
833
834 if (osb->s_mount_opt & hb_enabled) {
835 if (ocfs2_mount_local(osb)) {
836 mlog(ML_ERROR, "Cannot heartbeat on a locally "
837 "mounted device.\n");
838 return -EINVAL;
839 }
840 if (ocfs2_userspace_stack(osb)) {
841 mlog(ML_ERROR, "Userspace stack expected, but "
842 "o2cb heartbeat arguments passed to mount\n");
843 return -EINVAL;
844 }
845 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
846 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
847 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
848 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
849 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
850 return -EINVAL;
851 }
852 }
853
854 if (!(osb->s_mount_opt & hb_enabled)) {
855 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
856 !ocfs2_userspace_stack(osb)) {
857 mlog(ML_ERROR, "Heartbeat has to be started to mount "
858 "a read-write clustered device.\n");
859 return -EINVAL;
860 }
861 }
862
863 return 0;
864 }
865
866 /*
867 * If we're using a userspace stack, mount should have passed
868 * a name that matches the disk. If not, mount should not
869 * have passed a stack.
870 */
871 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
872 struct mount_options *mopt)
873 {
874 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
875 mlog(ML_ERROR,
876 "cluster stack passed to mount, but this filesystem "
877 "does not support it\n");
878 return -EINVAL;
879 }
880
881 if (ocfs2_userspace_stack(osb) &&
882 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
883 OCFS2_STACK_LABEL_LEN)) {
884 mlog(ML_ERROR,
885 "cluster stack passed to mount (\"%s\") does not "
886 "match the filesystem (\"%s\")\n",
887 mopt->cluster_stack,
888 osb->osb_cluster_stack);
889 return -EINVAL;
890 }
891
892 return 0;
893 }
894
895 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
896 {
897 int type;
898 struct super_block *sb = osb->sb;
899 unsigned int feature[OCFS2_MAXQUOTAS] = {
900 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
901 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
902 int status = 0;
903
904 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
905 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
906 continue;
907 if (unsuspend)
908 status = dquot_resume(sb, type);
909 else {
910 struct ocfs2_mem_dqinfo *oinfo;
911
912 /* Cancel periodic syncing before suspending */
913 oinfo = sb_dqinfo(sb, type)->dqi_priv;
914 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
915 status = dquot_suspend(sb, type);
916 }
917 if (status < 0)
918 break;
919 }
920 if (status < 0)
921 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
922 "remount (error = %d).\n", status);
923 return status;
924 }
925
926 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
927 {
928 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
929 struct super_block *sb = osb->sb;
930 unsigned int feature[OCFS2_MAXQUOTAS] = {
931 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
932 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
933 unsigned int ino[OCFS2_MAXQUOTAS] = {
934 LOCAL_USER_QUOTA_SYSTEM_INODE,
935 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
936 int status;
937 int type;
938
939 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
940 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
941 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
942 continue;
943 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
944 osb->slot_num);
945 if (!inode[type]) {
946 status = -ENOENT;
947 goto out_quota_off;
948 }
949 status = dquot_enable(inode[type], type, QFMT_OCFS2,
950 DQUOT_USAGE_ENABLED);
951 if (status < 0)
952 goto out_quota_off;
953 }
954
955 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
956 iput(inode[type]);
957 return 0;
958 out_quota_off:
959 ocfs2_disable_quotas(osb);
960 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
961 iput(inode[type]);
962 mlog_errno(status);
963 return status;
964 }
965
966 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
967 {
968 int type;
969 struct inode *inode;
970 struct super_block *sb = osb->sb;
971 struct ocfs2_mem_dqinfo *oinfo;
972
973 /* We mostly ignore errors in this function because there's not much
974 * we can do when we see them */
975 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
976 if (!sb_has_quota_loaded(sb, type))
977 continue;
978 oinfo = sb_dqinfo(sb, type)->dqi_priv;
979 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
980 inode = igrab(sb->s_dquot.files[type]);
981 /* Turn off quotas. This will remove all dquot structures from
982 * memory and so they will be automatically synced to global
983 * quota files */
984 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
985 DQUOT_LIMITS_ENABLED);
986 if (!inode)
987 continue;
988 iput(inode);
989 }
990 }
991
992 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
993 {
994 struct dentry *root;
995 int status, sector_size;
996 struct mount_options parsed_options;
997 struct inode *inode = NULL;
998 struct ocfs2_super *osb = NULL;
999 struct buffer_head *bh = NULL;
1000 char nodestr[12];
1001 struct ocfs2_blockcheck_stats stats;
1002
1003 trace_ocfs2_fill_super(sb, data, silent);
1004
1005 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1006 status = -EINVAL;
1007 goto read_super_error;
1008 }
1009
1010 /* probe for superblock */
1011 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
1012 if (status < 0) {
1013 mlog(ML_ERROR, "superblock probe failed!\n");
1014 goto read_super_error;
1015 }
1016
1017 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1018 osb = OCFS2_SB(sb);
1019 if (status < 0) {
1020 mlog_errno(status);
1021 goto read_super_error;
1022 }
1023 brelse(bh);
1024 bh = NULL;
1025
1026 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1027 status = -EINVAL;
1028 goto read_super_error;
1029 }
1030 osb->s_mount_opt = parsed_options.mount_opt;
1031 osb->s_atime_quantum = parsed_options.atime_quantum;
1032 osb->preferred_slot = parsed_options.slot;
1033 osb->osb_commit_interval = parsed_options.commit_interval;
1034
1035 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1036 osb->osb_resv_level = parsed_options.resv_level;
1037 osb->osb_dir_resv_level = parsed_options.resv_level;
1038 if (parsed_options.dir_resv_level == -1)
1039 osb->osb_dir_resv_level = parsed_options.resv_level;
1040 else
1041 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1042
1043 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1044 if (status)
1045 goto read_super_error;
1046
1047 sb->s_magic = OCFS2_SUPER_MAGIC;
1048
1049 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1050 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1051
1052 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1053 * heartbeat=none */
1054 if (bdev_read_only(sb->s_bdev)) {
1055 if (!sb_rdonly(sb)) {
1056 status = -EACCES;
1057 mlog(ML_ERROR, "Readonly device detected but readonly "
1058 "mount was not specified.\n");
1059 goto read_super_error;
1060 }
1061
1062 /* You should not be able to start a local heartbeat
1063 * on a readonly device. */
1064 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1065 status = -EROFS;
1066 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1067 "device.\n");
1068 goto read_super_error;
1069 }
1070
1071 status = ocfs2_check_journals_nolocks(osb);
1072 if (status < 0) {
1073 if (status == -EROFS)
1074 mlog(ML_ERROR, "Recovery required on readonly "
1075 "file system, but write access is "
1076 "unavailable.\n");
1077 else
1078 mlog_errno(status);
1079 goto read_super_error;
1080 }
1081
1082 ocfs2_set_ro_flag(osb, 1);
1083
1084 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1085 "Cluster services will not be used for this mount. "
1086 "Recovery will be skipped.\n", osb->dev_str);
1087 }
1088
1089 if (!ocfs2_is_hard_readonly(osb)) {
1090 if (sb_rdonly(sb))
1091 ocfs2_set_ro_flag(osb, 0);
1092 }
1093
1094 status = ocfs2_verify_heartbeat(osb);
1095 if (status < 0) {
1096 mlog_errno(status);
1097 goto read_super_error;
1098 }
1099
1100 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1101 ocfs2_debugfs_root);
1102 if (!osb->osb_debug_root) {
1103 status = -EINVAL;
1104 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1105 goto read_super_error;
1106 }
1107
1108 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1109 osb->osb_debug_root,
1110 osb,
1111 &ocfs2_osb_debug_fops);
1112 if (!osb->osb_ctxt) {
1113 status = -EINVAL;
1114 mlog_errno(status);
1115 goto read_super_error;
1116 }
1117
1118 if (ocfs2_meta_ecc(osb)) {
1119 status = ocfs2_blockcheck_stats_debugfs_install(
1120 &osb->osb_ecc_stats,
1121 osb->osb_debug_root);
1122 if (status) {
1123 mlog(ML_ERROR,
1124 "Unable to create blockcheck statistics "
1125 "files\n");
1126 goto read_super_error;
1127 }
1128 }
1129
1130 status = ocfs2_mount_volume(sb);
1131 if (status < 0)
1132 goto read_super_error;
1133
1134 if (osb->root_inode)
1135 inode = igrab(osb->root_inode);
1136
1137 if (!inode) {
1138 status = -EIO;
1139 mlog_errno(status);
1140 goto read_super_error;
1141 }
1142
1143 root = d_make_root(inode);
1144 if (!root) {
1145 status = -ENOMEM;
1146 mlog_errno(status);
1147 goto read_super_error;
1148 }
1149
1150 sb->s_root = root;
1151
1152 ocfs2_complete_mount_recovery(osb);
1153
1154 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1155 &ocfs2_kset->kobj);
1156 if (!osb->osb_dev_kset) {
1157 status = -ENOMEM;
1158 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1159 goto read_super_error;
1160 }
1161
1162 /* Create filecheck sysfs related directories/files at
1163 * /sys/fs/ocfs2/<devname>/filecheck */
1164 if (ocfs2_filecheck_create_sysfs(osb)) {
1165 status = -ENOMEM;
1166 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1167 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1168 goto read_super_error;
1169 }
1170
1171 if (ocfs2_mount_local(osb))
1172 snprintf(nodestr, sizeof(nodestr), "local");
1173 else
1174 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1175
1176 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1177 "with %s data mode.\n",
1178 osb->dev_str, nodestr, osb->slot_num,
1179 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1180 "ordered");
1181
1182 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1183 wake_up(&osb->osb_mount_event);
1184
1185 /* Now we can initialize quotas because we can afford to wait
1186 * for cluster locks recovery now. That also means that truncation
1187 * log recovery can happen but that waits for proper quota setup */
1188 if (!sb_rdonly(sb)) {
1189 status = ocfs2_enable_quotas(osb);
1190 if (status < 0) {
1191 /* We have to err-out specially here because
1192 * s_root is already set */
1193 mlog_errno(status);
1194 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1195 wake_up(&osb->osb_mount_event);
1196 return status;
1197 }
1198 }
1199
1200 ocfs2_complete_quota_recovery(osb);
1201
1202 /* Now we wake up again for processes waiting for quotas */
1203 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1204 wake_up(&osb->osb_mount_event);
1205
1206 /* Start this when the mount is almost sure of being successful */
1207 ocfs2_orphan_scan_start(osb);
1208
1209 return status;
1210
1211 read_super_error:
1212 brelse(bh);
1213
1214 if (status)
1215 mlog_errno(status);
1216
1217 if (osb) {
1218 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1219 wake_up(&osb->osb_mount_event);
1220 ocfs2_dismount_volume(sb, 1);
1221 }
1222
1223 return status;
1224 }
1225
1226 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1227 int flags,
1228 const char *dev_name,
1229 void *data)
1230 {
1231 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1232 }
1233
1234 static struct file_system_type ocfs2_fs_type = {
1235 .owner = THIS_MODULE,
1236 .name = "ocfs2",
1237 .mount = ocfs2_mount,
1238 .kill_sb = kill_block_super,
1239 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1240 .next = NULL
1241 };
1242 MODULE_ALIAS_FS("ocfs2");
1243
1244 static int ocfs2_check_set_options(struct super_block *sb,
1245 struct mount_options *options)
1246 {
1247 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1248 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1249 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1250 mlog(ML_ERROR, "User quotas were requested, but this "
1251 "filesystem does not have the feature enabled.\n");
1252 return 0;
1253 }
1254 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1255 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1256 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1257 mlog(ML_ERROR, "Group quotas were requested, but this "
1258 "filesystem does not have the feature enabled.\n");
1259 return 0;
1260 }
1261 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1262 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1263 mlog(ML_ERROR, "ACL support requested but extended attributes "
1264 "feature is not enabled\n");
1265 return 0;
1266 }
1267 /* No ACL setting specified? Use XATTR feature... */
1268 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1269 OCFS2_MOUNT_NO_POSIX_ACL))) {
1270 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1271 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1272 else
1273 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1274 }
1275 return 1;
1276 }
1277
1278 static int ocfs2_parse_options(struct super_block *sb,
1279 char *options,
1280 struct mount_options *mopt,
1281 int is_remount)
1282 {
1283 int status, user_stack = 0;
1284 char *p;
1285 u32 tmp;
1286 int token, option;
1287 substring_t args[MAX_OPT_ARGS];
1288
1289 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1290
1291 mopt->commit_interval = 0;
1292 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1293 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1294 mopt->slot = OCFS2_INVALID_SLOT;
1295 mopt->localalloc_opt = -1;
1296 mopt->cluster_stack[0] = '\0';
1297 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1298 mopt->dir_resv_level = -1;
1299
1300 if (!options) {
1301 status = 1;
1302 goto bail;
1303 }
1304
1305 while ((p = strsep(&options, ",")) != NULL) {
1306 if (!*p)
1307 continue;
1308
1309 token = match_token(p, tokens, args);
1310 switch (token) {
1311 case Opt_hb_local:
1312 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1313 break;
1314 case Opt_hb_none:
1315 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1316 break;
1317 case Opt_hb_global:
1318 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1319 break;
1320 case Opt_barrier:
1321 if (match_int(&args[0], &option)) {
1322 status = 0;
1323 goto bail;
1324 }
1325 if (option)
1326 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1327 else
1328 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1329 break;
1330 case Opt_intr:
1331 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1332 break;
1333 case Opt_nointr:
1334 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1335 break;
1336 case Opt_err_panic:
1337 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1338 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1339 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1340 break;
1341 case Opt_err_ro:
1342 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1343 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1344 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1345 break;
1346 case Opt_err_cont:
1347 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1348 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1349 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1350 break;
1351 case Opt_data_ordered:
1352 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1353 break;
1354 case Opt_data_writeback:
1355 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1356 break;
1357 case Opt_user_xattr:
1358 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1359 break;
1360 case Opt_nouser_xattr:
1361 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1362 break;
1363 case Opt_atime_quantum:
1364 if (match_int(&args[0], &option)) {
1365 status = 0;
1366 goto bail;
1367 }
1368 if (option >= 0)
1369 mopt->atime_quantum = option;
1370 break;
1371 case Opt_slot:
1372 if (match_int(&args[0], &option)) {
1373 status = 0;
1374 goto bail;
1375 }
1376 if (option)
1377 mopt->slot = (s16)option;
1378 break;
1379 case Opt_commit:
1380 if (match_int(&args[0], &option)) {
1381 status = 0;
1382 goto bail;
1383 }
1384 if (option < 0)
1385 return 0;
1386 if (option == 0)
1387 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1388 mopt->commit_interval = HZ * option;
1389 break;
1390 case Opt_localalloc:
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 /*
1431 * Open code the memcmp here as we don't have
1432 * an osb to pass to
1433 * ocfs2_userspace_stack().
1434 */
1435 if (memcmp(mopt->cluster_stack,
1436 OCFS2_CLASSIC_CLUSTER_STACK,
1437 OCFS2_STACK_LABEL_LEN))
1438 user_stack = 1;
1439 break;
1440 case Opt_inode64:
1441 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1442 break;
1443 case Opt_usrquota:
1444 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1445 break;
1446 case Opt_grpquota:
1447 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1448 break;
1449 case Opt_coherency_buffered:
1450 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1451 break;
1452 case Opt_coherency_full:
1453 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1454 break;
1455 case Opt_acl:
1456 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1457 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1458 break;
1459 case Opt_noacl:
1460 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1461 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1462 break;
1463 case Opt_resv_level:
1464 if (is_remount)
1465 break;
1466 if (match_int(&args[0], &option)) {
1467 status = 0;
1468 goto bail;
1469 }
1470 if (option >= OCFS2_MIN_RESV_LEVEL &&
1471 option < OCFS2_MAX_RESV_LEVEL)
1472 mopt->resv_level = option;
1473 break;
1474 case Opt_dir_resv_level:
1475 if (is_remount)
1476 break;
1477 if (match_int(&args[0], &option)) {
1478 status = 0;
1479 goto bail;
1480 }
1481 if (option >= OCFS2_MIN_RESV_LEVEL &&
1482 option < OCFS2_MAX_RESV_LEVEL)
1483 mopt->dir_resv_level = option;
1484 break;
1485 case Opt_journal_async_commit:
1486 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1487 break;
1488 default:
1489 mlog(ML_ERROR,
1490 "Unrecognized mount option \"%s\" "
1491 "or missing value\n", p);
1492 status = 0;
1493 goto bail;
1494 }
1495 }
1496
1497 if (user_stack == 0) {
1498 /* Ensure only one heartbeat mode */
1499 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1500 OCFS2_MOUNT_HB_GLOBAL |
1501 OCFS2_MOUNT_HB_NONE);
1502 if (hweight32(tmp) != 1) {
1503 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1504 status = 0;
1505 goto bail;
1506 }
1507 }
1508
1509 status = 1;
1510
1511 bail:
1512 return status;
1513 }
1514
1515 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1516 {
1517 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1518 unsigned long opts = osb->s_mount_opt;
1519 unsigned int local_alloc_megs;
1520
1521 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1522 seq_printf(s, ",_netdev");
1523 if (opts & OCFS2_MOUNT_HB_LOCAL)
1524 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1525 else
1526 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1527 } else
1528 seq_printf(s, ",%s", OCFS2_HB_NONE);
1529
1530 if (opts & OCFS2_MOUNT_NOINTR)
1531 seq_printf(s, ",nointr");
1532
1533 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1534 seq_printf(s, ",data=writeback");
1535 else
1536 seq_printf(s, ",data=ordered");
1537
1538 if (opts & OCFS2_MOUNT_BARRIER)
1539 seq_printf(s, ",barrier=1");
1540
1541 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1542 seq_printf(s, ",errors=panic");
1543 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1544 seq_printf(s, ",errors=continue");
1545 else
1546 seq_printf(s, ",errors=remount-ro");
1547
1548 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1549 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1550
1551 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1552
1553 if (osb->osb_commit_interval)
1554 seq_printf(s, ",commit=%u",
1555 (unsigned) (osb->osb_commit_interval / HZ));
1556
1557 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1558 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1559 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1560
1561 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1562 seq_printf(s, ",localflocks,");
1563
1564 if (osb->osb_cluster_stack[0])
1565 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1566 OCFS2_STACK_LABEL_LEN);
1567 if (opts & OCFS2_MOUNT_USRQUOTA)
1568 seq_printf(s, ",usrquota");
1569 if (opts & OCFS2_MOUNT_GRPQUOTA)
1570 seq_printf(s, ",grpquota");
1571
1572 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1573 seq_printf(s, ",coherency=buffered");
1574 else
1575 seq_printf(s, ",coherency=full");
1576
1577 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1578 seq_printf(s, ",nouser_xattr");
1579 else
1580 seq_printf(s, ",user_xattr");
1581
1582 if (opts & OCFS2_MOUNT_INODE64)
1583 seq_printf(s, ",inode64");
1584
1585 if (opts & OCFS2_MOUNT_POSIX_ACL)
1586 seq_printf(s, ",acl");
1587 else
1588 seq_printf(s, ",noacl");
1589
1590 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1591 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1592
1593 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1594 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1595
1596 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1597 seq_printf(s, ",journal_async_commit");
1598
1599 return 0;
1600 }
1601
1602 static int __init ocfs2_init(void)
1603 {
1604 int status;
1605
1606 status = init_ocfs2_uptodate_cache();
1607 if (status < 0)
1608 goto out1;
1609
1610 status = ocfs2_initialize_mem_caches();
1611 if (status < 0)
1612 goto out2;
1613
1614 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1615 if (!ocfs2_debugfs_root) {
1616 status = -ENOMEM;
1617 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1618 goto out3;
1619 }
1620
1621 ocfs2_set_locking_protocol();
1622
1623 status = register_quota_format(&ocfs2_quota_format);
1624 if (status < 0)
1625 goto out3;
1626 status = register_filesystem(&ocfs2_fs_type);
1627 if (!status)
1628 return 0;
1629
1630 unregister_quota_format(&ocfs2_quota_format);
1631 out3:
1632 debugfs_remove(ocfs2_debugfs_root);
1633 ocfs2_free_mem_caches();
1634 out2:
1635 exit_ocfs2_uptodate_cache();
1636 out1:
1637 mlog_errno(status);
1638 return status;
1639 }
1640
1641 static void __exit ocfs2_exit(void)
1642 {
1643 unregister_quota_format(&ocfs2_quota_format);
1644
1645 debugfs_remove(ocfs2_debugfs_root);
1646
1647 ocfs2_free_mem_caches();
1648
1649 unregister_filesystem(&ocfs2_fs_type);
1650
1651 exit_ocfs2_uptodate_cache();
1652 }
1653
1654 static void ocfs2_put_super(struct super_block *sb)
1655 {
1656 trace_ocfs2_put_super(sb);
1657
1658 ocfs2_sync_blockdev(sb);
1659 ocfs2_dismount_volume(sb, 0);
1660 }
1661
1662 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1663 {
1664 struct ocfs2_super *osb;
1665 u32 numbits, freebits;
1666 int status;
1667 struct ocfs2_dinode *bm_lock;
1668 struct buffer_head *bh = NULL;
1669 struct inode *inode = NULL;
1670
1671 trace_ocfs2_statfs(dentry->d_sb, buf);
1672
1673 osb = OCFS2_SB(dentry->d_sb);
1674
1675 inode = ocfs2_get_system_file_inode(osb,
1676 GLOBAL_BITMAP_SYSTEM_INODE,
1677 OCFS2_INVALID_SLOT);
1678 if (!inode) {
1679 mlog(ML_ERROR, "failed to get bitmap inode\n");
1680 status = -EIO;
1681 goto bail;
1682 }
1683
1684 status = ocfs2_inode_lock(inode, &bh, 0);
1685 if (status < 0) {
1686 mlog_errno(status);
1687 goto bail;
1688 }
1689
1690 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1691
1692 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1693 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1694
1695 buf->f_type = OCFS2_SUPER_MAGIC;
1696 buf->f_bsize = dentry->d_sb->s_blocksize;
1697 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1698 buf->f_blocks = ((sector_t) numbits) *
1699 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1700 buf->f_bfree = ((sector_t) freebits) *
1701 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1702 buf->f_bavail = buf->f_bfree;
1703 buf->f_files = numbits;
1704 buf->f_ffree = freebits;
1705 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1706 & 0xFFFFFFFFUL;
1707 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1708 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1709
1710 brelse(bh);
1711
1712 ocfs2_inode_unlock(inode, 0);
1713 status = 0;
1714 bail:
1715 iput(inode);
1716
1717 if (status)
1718 mlog_errno(status);
1719
1720 return status;
1721 }
1722
1723 static void ocfs2_inode_init_once(void *data)
1724 {
1725 struct ocfs2_inode_info *oi = data;
1726
1727 oi->ip_flags = 0;
1728 oi->ip_open_count = 0;
1729 spin_lock_init(&oi->ip_lock);
1730 ocfs2_extent_map_init(&oi->vfs_inode);
1731 INIT_LIST_HEAD(&oi->ip_io_markers);
1732 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1733 oi->ip_dir_start_lookup = 0;
1734 init_rwsem(&oi->ip_alloc_sem);
1735 init_rwsem(&oi->ip_xattr_sem);
1736 mutex_init(&oi->ip_io_mutex);
1737
1738 oi->ip_blkno = 0ULL;
1739 oi->ip_clusters = 0;
1740
1741 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1742
1743 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1744 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1745 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1746
1747 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1748 &ocfs2_inode_caching_ops);
1749
1750 inode_init_once(&oi->vfs_inode);
1751 }
1752
1753 static int ocfs2_initialize_mem_caches(void)
1754 {
1755 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1756 sizeof(struct ocfs2_inode_info),
1757 0,
1758 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1759 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1760 ocfs2_inode_init_once);
1761 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1762 sizeof(struct ocfs2_dquot),
1763 0,
1764 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1765 SLAB_MEM_SPREAD),
1766 NULL);
1767 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1768 sizeof(struct ocfs2_quota_chunk),
1769 0,
1770 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1771 NULL);
1772 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1773 !ocfs2_qf_chunk_cachep) {
1774 kmem_cache_destroy(ocfs2_inode_cachep);
1775 kmem_cache_destroy(ocfs2_dquot_cachep);
1776 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1777 return -ENOMEM;
1778 }
1779
1780 return 0;
1781 }
1782
1783 static void ocfs2_free_mem_caches(void)
1784 {
1785 /*
1786 * Make sure all delayed rcu free inodes are flushed before we
1787 * destroy cache.
1788 */
1789 rcu_barrier();
1790 kmem_cache_destroy(ocfs2_inode_cachep);
1791 ocfs2_inode_cachep = NULL;
1792
1793 kmem_cache_destroy(ocfs2_dquot_cachep);
1794 ocfs2_dquot_cachep = NULL;
1795
1796 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1797 ocfs2_qf_chunk_cachep = NULL;
1798 }
1799
1800 static int ocfs2_get_sector(struct super_block *sb,
1801 struct buffer_head **bh,
1802 int block,
1803 int sect_size)
1804 {
1805 if (!sb_set_blocksize(sb, sect_size)) {
1806 mlog(ML_ERROR, "unable to set blocksize\n");
1807 return -EIO;
1808 }
1809
1810 *bh = sb_getblk(sb, block);
1811 if (!*bh) {
1812 mlog_errno(-ENOMEM);
1813 return -ENOMEM;
1814 }
1815 lock_buffer(*bh);
1816 if (!buffer_dirty(*bh))
1817 clear_buffer_uptodate(*bh);
1818 unlock_buffer(*bh);
1819 ll_rw_block(REQ_OP_READ, 0, 1, bh);
1820 wait_on_buffer(*bh);
1821 if (!buffer_uptodate(*bh)) {
1822 mlog_errno(-EIO);
1823 brelse(*bh);
1824 *bh = NULL;
1825 return -EIO;
1826 }
1827
1828 return 0;
1829 }
1830
1831 static int ocfs2_mount_volume(struct super_block *sb)
1832 {
1833 int status = 0;
1834 int unlock_super = 0;
1835 struct ocfs2_super *osb = OCFS2_SB(sb);
1836
1837 if (ocfs2_is_hard_readonly(osb))
1838 goto leave;
1839
1840 mutex_init(&osb->obs_trim_fs_mutex);
1841
1842 status = ocfs2_dlm_init(osb);
1843 if (status < 0) {
1844 mlog_errno(status);
1845 if (status == -EBADR && ocfs2_userspace_stack(osb))
1846 mlog(ML_ERROR, "couldn't mount because cluster name on"
1847 " disk does not match the running cluster name.\n");
1848 goto leave;
1849 }
1850
1851 status = ocfs2_super_lock(osb, 1);
1852 if (status < 0) {
1853 mlog_errno(status);
1854 goto leave;
1855 }
1856 unlock_super = 1;
1857
1858 /* This will load up the node map and add ourselves to it. */
1859 status = ocfs2_find_slot(osb);
1860 if (status < 0) {
1861 mlog_errno(status);
1862 goto leave;
1863 }
1864
1865 /* load all node-local system inodes */
1866 status = ocfs2_init_local_system_inodes(osb);
1867 if (status < 0) {
1868 mlog_errno(status);
1869 goto leave;
1870 }
1871
1872 status = ocfs2_check_volume(osb);
1873 if (status < 0) {
1874 mlog_errno(status);
1875 goto leave;
1876 }
1877
1878 status = ocfs2_truncate_log_init(osb);
1879 if (status < 0)
1880 mlog_errno(status);
1881
1882 leave:
1883 if (unlock_super)
1884 ocfs2_super_unlock(osb, 1);
1885
1886 return status;
1887 }
1888
1889 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1890 {
1891 int tmp, hangup_needed = 0;
1892 struct ocfs2_super *osb = NULL;
1893 char nodestr[12];
1894
1895 trace_ocfs2_dismount_volume(sb);
1896
1897 BUG_ON(!sb);
1898 osb = OCFS2_SB(sb);
1899 BUG_ON(!osb);
1900
1901 /* Remove file check sysfs related directores/files,
1902 * and wait for the pending file check operations */
1903 ocfs2_filecheck_remove_sysfs(osb);
1904
1905 kset_unregister(osb->osb_dev_kset);
1906
1907 debugfs_remove(osb->osb_ctxt);
1908
1909 /* Orphan scan should be stopped as early as possible */
1910 ocfs2_orphan_scan_stop(osb);
1911
1912 ocfs2_disable_quotas(osb);
1913
1914 /* All dquots should be freed by now */
1915 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1916 /* Wait for worker to be done with the work structure in osb */
1917 cancel_work_sync(&osb->dquot_drop_work);
1918
1919 ocfs2_shutdown_local_alloc(osb);
1920
1921 ocfs2_truncate_log_shutdown(osb);
1922
1923 /* This will disable recovery and flush any recovery work. */
1924 ocfs2_recovery_exit(osb);
1925
1926 ocfs2_journal_shutdown(osb);
1927
1928 ocfs2_sync_blockdev(sb);
1929
1930 ocfs2_purge_refcount_trees(osb);
1931
1932 /* No cluster connection means we've failed during mount, so skip
1933 * all the steps which depended on that to complete. */
1934 if (osb->cconn) {
1935 tmp = ocfs2_super_lock(osb, 1);
1936 if (tmp < 0) {
1937 mlog_errno(tmp);
1938 return;
1939 }
1940 }
1941
1942 if (osb->slot_num != OCFS2_INVALID_SLOT)
1943 ocfs2_put_slot(osb);
1944
1945 if (osb->cconn)
1946 ocfs2_super_unlock(osb, 1);
1947
1948 ocfs2_release_system_inodes(osb);
1949
1950 /*
1951 * If we're dismounting due to mount error, mount.ocfs2 will clean
1952 * up heartbeat. If we're a local mount, there is no heartbeat.
1953 * If we failed before we got a uuid_str yet, we can't stop
1954 * heartbeat. Otherwise, do it.
1955 */
1956 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1957 !ocfs2_is_hard_readonly(osb))
1958 hangup_needed = 1;
1959
1960 if (osb->cconn)
1961 ocfs2_dlm_shutdown(osb, hangup_needed);
1962
1963 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1964 debugfs_remove(osb->osb_debug_root);
1965
1966 if (hangup_needed)
1967 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1968
1969 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1970
1971 if (ocfs2_mount_local(osb))
1972 snprintf(nodestr, sizeof(nodestr), "local");
1973 else
1974 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1975
1976 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1977 osb->dev_str, nodestr);
1978
1979 ocfs2_delete_osb(osb);
1980 kfree(osb);
1981 sb->s_dev = 0;
1982 sb->s_fs_info = NULL;
1983 }
1984
1985 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1986 unsigned uuid_bytes)
1987 {
1988 int i, ret;
1989 char *ptr;
1990
1991 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1992
1993 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1994 if (osb->uuid_str == NULL)
1995 return -ENOMEM;
1996
1997 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1998 /* print with null */
1999 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2000 if (ret != 2) /* drop super cleans up */
2001 return -EINVAL;
2002 /* then only advance past the last char */
2003 ptr += 2;
2004 }
2005
2006 return 0;
2007 }
2008
2009 /* Make sure entire volume is addressable by our journal. Requires
2010 osb_clusters_at_boot to be valid and for the journal to have been
2011 initialized by ocfs2_journal_init(). */
2012 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2013 {
2014 int status = 0;
2015 u64 max_block =
2016 ocfs2_clusters_to_blocks(osb->sb,
2017 osb->osb_clusters_at_boot) - 1;
2018
2019 /* 32-bit block number is always OK. */
2020 if (max_block <= (u32)~0ULL)
2021 goto out;
2022
2023 /* Volume is "huge", so see if our journal is new enough to
2024 support it. */
2025 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2026 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2027 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2028 JBD2_FEATURE_INCOMPAT_64BIT))) {
2029 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2030 "Enable the 'block64' journal option with tunefs.ocfs2");
2031 status = -EFBIG;
2032 goto out;
2033 }
2034
2035 out:
2036 return status;
2037 }
2038
2039 static int ocfs2_initialize_super(struct super_block *sb,
2040 struct buffer_head *bh,
2041 int sector_size,
2042 struct ocfs2_blockcheck_stats *stats)
2043 {
2044 int status;
2045 int i, cbits, bbits;
2046 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2047 struct inode *inode = NULL;
2048 struct ocfs2_journal *journal;
2049 struct ocfs2_super *osb;
2050 u64 total_blocks;
2051
2052 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2053 if (!osb) {
2054 status = -ENOMEM;
2055 mlog_errno(status);
2056 goto bail;
2057 }
2058
2059 sb->s_fs_info = osb;
2060 sb->s_op = &ocfs2_sops;
2061 sb->s_d_op = &ocfs2_dentry_ops;
2062 sb->s_export_op = &ocfs2_export_ops;
2063 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2064 sb->dq_op = &ocfs2_quota_operations;
2065 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2066 sb->s_xattr = ocfs2_xattr_handlers;
2067 sb->s_time_gran = 1;
2068 sb->s_flags |= SB_NOATIME;
2069 /* this is needed to support O_LARGEFILE */
2070 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2071 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2072 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2073 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2074 sizeof(di->id2.i_super.s_uuid));
2075
2076 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2077
2078 for (i = 0; i < 3; i++)
2079 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2080 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2081
2082 osb->sb = sb;
2083 osb->s_sectsize_bits = blksize_bits(sector_size);
2084 BUG_ON(!osb->s_sectsize_bits);
2085
2086 spin_lock_init(&osb->dc_task_lock);
2087 init_waitqueue_head(&osb->dc_event);
2088 osb->dc_work_sequence = 0;
2089 osb->dc_wake_sequence = 0;
2090 INIT_LIST_HEAD(&osb->blocked_lock_list);
2091 osb->blocked_lock_count = 0;
2092 spin_lock_init(&osb->osb_lock);
2093 spin_lock_init(&osb->osb_xattr_lock);
2094 ocfs2_init_steal_slots(osb);
2095
2096 mutex_init(&osb->system_file_mutex);
2097
2098 atomic_set(&osb->alloc_stats.moves, 0);
2099 atomic_set(&osb->alloc_stats.local_data, 0);
2100 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2101 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2102 atomic_set(&osb->alloc_stats.bg_extends, 0);
2103
2104 /* Copy the blockcheck stats from the superblock probe */
2105 osb->osb_ecc_stats = *stats;
2106
2107 ocfs2_init_node_maps(osb);
2108
2109 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2110 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2111
2112 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2113 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2114 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2115 osb->max_slots);
2116 status = -EINVAL;
2117 goto bail;
2118 }
2119
2120 ocfs2_orphan_scan_init(osb);
2121
2122 status = ocfs2_recovery_init(osb);
2123 if (status) {
2124 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2125 mlog_errno(status);
2126 goto bail;
2127 }
2128
2129 init_waitqueue_head(&osb->checkpoint_event);
2130
2131 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2132
2133 osb->slot_num = OCFS2_INVALID_SLOT;
2134
2135 osb->s_xattr_inline_size = le16_to_cpu(
2136 di->id2.i_super.s_xattr_inline_size);
2137
2138 osb->local_alloc_state = OCFS2_LA_UNUSED;
2139 osb->local_alloc_bh = NULL;
2140 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2141
2142 init_waitqueue_head(&osb->osb_mount_event);
2143
2144 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2145 if (status) {
2146 mlog_errno(status);
2147 goto bail;
2148 }
2149
2150 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2151 if (!osb->vol_label) {
2152 mlog(ML_ERROR, "unable to alloc vol label\n");
2153 status = -ENOMEM;
2154 goto bail;
2155 }
2156
2157 osb->slot_recovery_generations =
2158 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2159 GFP_KERNEL);
2160 if (!osb->slot_recovery_generations) {
2161 status = -ENOMEM;
2162 mlog_errno(status);
2163 goto bail;
2164 }
2165
2166 init_waitqueue_head(&osb->osb_wipe_event);
2167 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2168 sizeof(*osb->osb_orphan_wipes),
2169 GFP_KERNEL);
2170 if (!osb->osb_orphan_wipes) {
2171 status = -ENOMEM;
2172 mlog_errno(status);
2173 goto bail;
2174 }
2175
2176 osb->osb_rf_lock_tree = RB_ROOT;
2177
2178 osb->s_feature_compat =
2179 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2180 osb->s_feature_ro_compat =
2181 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2182 osb->s_feature_incompat =
2183 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2184
2185 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2186 mlog(ML_ERROR, "couldn't mount because of unsupported "
2187 "optional features (%x).\n", i);
2188 status = -EINVAL;
2189 goto bail;
2190 }
2191 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2192 mlog(ML_ERROR, "couldn't mount RDWR because of "
2193 "unsupported optional features (%x).\n", i);
2194 status = -EINVAL;
2195 goto bail;
2196 }
2197
2198 if (ocfs2_clusterinfo_valid(osb)) {
2199 osb->osb_stackflags =
2200 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2201 strlcpy(osb->osb_cluster_stack,
2202 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2203 OCFS2_STACK_LABEL_LEN + 1);
2204 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2205 mlog(ML_ERROR,
2206 "couldn't mount because of an invalid "
2207 "cluster stack label (%s) \n",
2208 osb->osb_cluster_stack);
2209 status = -EINVAL;
2210 goto bail;
2211 }
2212 strlcpy(osb->osb_cluster_name,
2213 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2214 OCFS2_CLUSTER_NAME_LEN + 1);
2215 } else {
2216 /* The empty string is identical with classic tools that
2217 * don't know about s_cluster_info. */
2218 osb->osb_cluster_stack[0] = '\0';
2219 }
2220
2221 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2222
2223 /* FIXME
2224 * This should be done in ocfs2_journal_init(), but unknown
2225 * ordering issues will cause the filesystem to crash.
2226 * If anyone wants to figure out what part of the code
2227 * refers to osb->journal before ocfs2_journal_init() is run,
2228 * be my guest.
2229 */
2230 /* initialize our journal structure */
2231
2232 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2233 if (!journal) {
2234 mlog(ML_ERROR, "unable to alloc journal\n");
2235 status = -ENOMEM;
2236 goto bail;
2237 }
2238 osb->journal = journal;
2239 journal->j_osb = osb;
2240
2241 atomic_set(&journal->j_num_trans, 0);
2242 init_rwsem(&journal->j_trans_barrier);
2243 init_waitqueue_head(&journal->j_checkpointed);
2244 spin_lock_init(&journal->j_lock);
2245 journal->j_trans_id = (unsigned long) 1;
2246 INIT_LIST_HEAD(&journal->j_la_cleanups);
2247 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2248 journal->j_state = OCFS2_JOURNAL_FREE;
2249
2250 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2251 init_llist_head(&osb->dquot_drop_list);
2252
2253 /* get some pseudo constants for clustersize bits */
2254 osb->s_clustersize_bits =
2255 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2256 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2257
2258 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2259 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2260 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2261 osb->s_clustersize);
2262 status = -EINVAL;
2263 goto bail;
2264 }
2265
2266 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2267 le32_to_cpu(di->i_clusters));
2268
2269 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2270 total_blocks);
2271 if (status) {
2272 mlog(ML_ERROR, "Volume too large "
2273 "to mount safely on this system");
2274 status = -EFBIG;
2275 goto bail;
2276 }
2277
2278 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2279 sizeof(di->id2.i_super.s_uuid))) {
2280 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2281 status = -ENOMEM;
2282 goto bail;
2283 }
2284
2285 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2286 OCFS2_MAX_VOL_LABEL_LEN);
2287 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2288 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2289 osb->first_cluster_group_blkno =
2290 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2291 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2292 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2293 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2294 (unsigned long long)osb->root_blkno,
2295 (unsigned long long)osb->system_dir_blkno,
2296 osb->s_clustersize_bits);
2297
2298 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2299 if (!osb->osb_dlm_debug) {
2300 status = -ENOMEM;
2301 mlog_errno(status);
2302 goto bail;
2303 }
2304
2305 atomic_set(&osb->vol_state, VOLUME_INIT);
2306
2307 /* load root, system_dir, and all global system inodes */
2308 status = ocfs2_init_global_system_inodes(osb);
2309 if (status < 0) {
2310 mlog_errno(status);
2311 goto bail;
2312 }
2313
2314 /*
2315 * global bitmap
2316 */
2317 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2318 OCFS2_INVALID_SLOT);
2319 if (!inode) {
2320 status = -EINVAL;
2321 mlog_errno(status);
2322 goto bail;
2323 }
2324
2325 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2326 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2327 iput(inode);
2328
2329 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2330 osb->s_feature_incompat) * 8;
2331
2332 status = ocfs2_init_slot_info(osb);
2333 if (status < 0) {
2334 mlog_errno(status);
2335 goto bail;
2336 }
2337 cleancache_init_shared_fs(sb);
2338
2339 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2340 if (!osb->ocfs2_wq) {
2341 status = -ENOMEM;
2342 mlog_errno(status);
2343 }
2344
2345 bail:
2346 return status;
2347 }
2348
2349 /*
2350 * will return: -EAGAIN if it is ok to keep searching for superblocks
2351 * -EINVAL if there is a bad superblock
2352 * 0 on success
2353 */
2354 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2355 struct buffer_head *bh,
2356 u32 blksz,
2357 struct ocfs2_blockcheck_stats *stats)
2358 {
2359 int status = -EAGAIN;
2360
2361 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2362 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2363 /* We have to do a raw check of the feature here */
2364 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2365 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2366 status = ocfs2_block_check_validate(bh->b_data,
2367 bh->b_size,
2368 &di->i_check,
2369 stats);
2370 if (status)
2371 goto out;
2372 }
2373 status = -EINVAL;
2374 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2375 mlog(ML_ERROR, "found superblock with incorrect block "
2376 "size: found %u, should be %u\n",
2377 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2378 blksz);
2379 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2380 OCFS2_MAJOR_REV_LEVEL ||
2381 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2382 OCFS2_MINOR_REV_LEVEL) {
2383 mlog(ML_ERROR, "found superblock with bad version: "
2384 "found %u.%u, should be %u.%u\n",
2385 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2386 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2387 OCFS2_MAJOR_REV_LEVEL,
2388 OCFS2_MINOR_REV_LEVEL);
2389 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2390 mlog(ML_ERROR, "bad block number on superblock: "
2391 "found %llu, should be %llu\n",
2392 (unsigned long long)le64_to_cpu(di->i_blkno),
2393 (unsigned long long)bh->b_blocknr);
2394 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2395 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2396 mlog(ML_ERROR, "bad cluster size found: %u\n",
2397 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2398 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2399 mlog(ML_ERROR, "bad root_blkno: 0\n");
2400 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2401 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2402 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2403 mlog(ML_ERROR,
2404 "Superblock slots found greater than file system "
2405 "maximum: found %u, max %u\n",
2406 le16_to_cpu(di->id2.i_super.s_max_slots),
2407 OCFS2_MAX_SLOTS);
2408 } else {
2409 /* found it! */
2410 status = 0;
2411 }
2412 }
2413
2414 out:
2415 if (status && status != -EAGAIN)
2416 mlog_errno(status);
2417 return status;
2418 }
2419
2420 static int ocfs2_check_volume(struct ocfs2_super *osb)
2421 {
2422 int status;
2423 int dirty;
2424 int local;
2425 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2426 * recover
2427 * ourselves. */
2428
2429 /* Init our journal object. */
2430 status = ocfs2_journal_init(osb->journal, &dirty);
2431 if (status < 0) {
2432 mlog(ML_ERROR, "Could not initialize journal!\n");
2433 goto finally;
2434 }
2435
2436 /* Now that journal has been initialized, check to make sure
2437 entire volume is addressable. */
2438 status = ocfs2_journal_addressable(osb);
2439 if (status)
2440 goto finally;
2441
2442 /* If the journal was unmounted cleanly then we don't want to
2443 * recover anything. Otherwise, journal_load will do that
2444 * dirty work for us :) */
2445 if (!dirty) {
2446 status = ocfs2_journal_wipe(osb->journal, 0);
2447 if (status < 0) {
2448 mlog_errno(status);
2449 goto finally;
2450 }
2451 } else {
2452 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2453 "unmounted cleanly, recovering it.\n", osb->dev_str);
2454 }
2455
2456 local = ocfs2_mount_local(osb);
2457
2458 /* will play back anything left in the journal. */
2459 status = ocfs2_journal_load(osb->journal, local, dirty);
2460 if (status < 0) {
2461 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2462 goto finally;
2463 }
2464
2465 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2466 jbd2_journal_set_features(osb->journal->j_journal,
2467 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2468 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2469 else
2470 jbd2_journal_clear_features(osb->journal->j_journal,
2471 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2472 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2473
2474 if (dirty) {
2475 /* recover my local alloc if we didn't unmount cleanly. */
2476 status = ocfs2_begin_local_alloc_recovery(osb,
2477 osb->slot_num,
2478 &local_alloc);
2479 if (status < 0) {
2480 mlog_errno(status);
2481 goto finally;
2482 }
2483 /* we complete the recovery process after we've marked
2484 * ourselves as mounted. */
2485 }
2486
2487 status = ocfs2_load_local_alloc(osb);
2488 if (status < 0) {
2489 mlog_errno(status);
2490 goto finally;
2491 }
2492
2493 if (dirty) {
2494 /* Recovery will be completed after we've mounted the
2495 * rest of the volume. */
2496 osb->local_alloc_copy = local_alloc;
2497 local_alloc = NULL;
2498 }
2499
2500 /* go through each journal, trylock it and if you get the
2501 * lock, and it's marked as dirty, set the bit in the recover
2502 * map and launch a recovery thread for it. */
2503 status = ocfs2_mark_dead_nodes(osb);
2504 if (status < 0) {
2505 mlog_errno(status);
2506 goto finally;
2507 }
2508
2509 status = ocfs2_compute_replay_slots(osb);
2510 if (status < 0)
2511 mlog_errno(status);
2512
2513 finally:
2514 kfree(local_alloc);
2515
2516 if (status)
2517 mlog_errno(status);
2518 return status;
2519 }
2520
2521 /*
2522 * The routine gets called from dismount or close whenever a dismount on
2523 * volume is requested and the osb open count becomes 1.
2524 * It will remove the osb from the global list and also free up all the
2525 * initialized resources and fileobject.
2526 */
2527 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2528 {
2529 /* This function assumes that the caller has the main osb resource */
2530
2531 /* ocfs2_initializer_super have already created this workqueue */
2532 if (osb->ocfs2_wq)
2533 destroy_workqueue(osb->ocfs2_wq);
2534
2535 ocfs2_free_slot_info(osb);
2536
2537 kfree(osb->osb_orphan_wipes);
2538 kfree(osb->slot_recovery_generations);
2539 /* FIXME
2540 * This belongs in journal shutdown, but because we have to
2541 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2542 * we free it here.
2543 */
2544 kfree(osb->journal);
2545 kfree(osb->local_alloc_copy);
2546 kfree(osb->uuid_str);
2547 kfree(osb->vol_label);
2548 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2549 memset(osb, 0, sizeof(struct ocfs2_super));
2550 }
2551
2552 /* Depending on the mount option passed, perform one of the following:
2553 * Put OCFS2 into a readonly state (default)
2554 * Return EIO so that only the process errs
2555 * Fix the error as if fsck.ocfs2 -y
2556 * panic
2557 */
2558 static int ocfs2_handle_error(struct super_block *sb)
2559 {
2560 struct ocfs2_super *osb = OCFS2_SB(sb);
2561 int rv = 0;
2562
2563 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2564 pr_crit("On-disk corruption discovered. "
2565 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2566
2567 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2568 panic("OCFS2: (device %s): panic forced after error\n",
2569 sb->s_id);
2570 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2571 pr_crit("OCFS2: Returning error to the calling process.\n");
2572 rv = -EIO;
2573 } else { /* default option */
2574 rv = -EROFS;
2575 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2576 return rv;
2577
2578 pr_crit("OCFS2: File system is now read-only.\n");
2579 sb->s_flags |= SB_RDONLY;
2580 ocfs2_set_ro_flag(osb, 0);
2581 }
2582
2583 return rv;
2584 }
2585
2586 int __ocfs2_error(struct super_block *sb, const char *function,
2587 const char *fmt, ...)
2588 {
2589 struct va_format vaf;
2590 va_list args;
2591
2592 va_start(args, fmt);
2593 vaf.fmt = fmt;
2594 vaf.va = &args;
2595
2596 /* Not using mlog here because we want to show the actual
2597 * function the error came from. */
2598 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2599 sb->s_id, function, &vaf);
2600
2601 va_end(args);
2602
2603 return ocfs2_handle_error(sb);
2604 }
2605
2606 /* Handle critical errors. This is intentionally more drastic than
2607 * ocfs2_handle_error, so we only use for things like journal errors,
2608 * etc. */
2609 void __ocfs2_abort(struct super_block *sb, const char *function,
2610 const char *fmt, ...)
2611 {
2612 struct va_format vaf;
2613 va_list args;
2614
2615 va_start(args, fmt);
2616
2617 vaf.fmt = fmt;
2618 vaf.va = &args;
2619
2620 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2621 sb->s_id, function, &vaf);
2622
2623 va_end(args);
2624
2625 /* We don't have the cluster support yet to go straight to
2626 * hard readonly in here. Until then, we want to keep
2627 * ocfs2_abort() so that we can at least mark critical
2628 * errors.
2629 *
2630 * TODO: This should abort the journal and alert other nodes
2631 * that our slot needs recovery. */
2632
2633 /* Force a panic(). This stinks, but it's better than letting
2634 * things continue without having a proper hard readonly
2635 * here. */
2636 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2637 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2638 ocfs2_handle_error(sb);
2639 }
2640
2641 /*
2642 * Void signal blockers, because in-kernel sigprocmask() only fails
2643 * when SIG_* is wrong.
2644 */
2645 void ocfs2_block_signals(sigset_t *oldset)
2646 {
2647 int rc;
2648 sigset_t blocked;
2649
2650 sigfillset(&blocked);
2651 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2652 BUG_ON(rc);
2653 }
2654
2655 void ocfs2_unblock_signals(sigset_t *oldset)
2656 {
2657 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2658 BUG_ON(rc);
2659 }
2660
2661 module_init(ocfs2_init);
2662 module_exit(ocfs2_exit);
A mirror of Linus' kernel repository