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 unsigned 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
ocfs2_get_dquots(struct inode * inode)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
ocfs2_osb_dump(struct ocfs2_super * osb,char * buf,int len)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
ocfs2_osb_debug_open(struct inode * inode,struct file * file)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
ocfs2_debug_release(struct inode * inode,struct file * file)374 static int ocfs2_debug_release(struct inode *inode, struct file *file)
375 {
376 kfree(file->private_data);
377 return 0;
378 }
379
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)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
ocfs2_osb_debug_open(struct inode * inode,struct file * file)387 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
388 {
389 return 0;
390 }
ocfs2_debug_release(struct inode * inode,struct file * file)391 static int ocfs2_debug_release(struct inode *inode, struct file *file)
392 {
393 return 0;
394 }
ocfs2_debug_read(struct file * file,char __user * buf,size_t nbytes,loff_t * ppos)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
ocfs2_sync_fs(struct super_block * sb,int wait)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
ocfs2_need_system_inode(struct ocfs2_super * osb,int ino)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
ocfs2_init_global_system_inodes(struct ocfs2_super * osb)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
ocfs2_init_local_system_inodes(struct ocfs2_super * osb)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
ocfs2_release_system_inodes(struct ocfs2_super * osb)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 */
ocfs2_alloc_inode(struct super_block * sb)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
ocfs2_i_callback(struct rcu_head * head)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
ocfs2_destroy_inode(struct inode * inode)584 static void ocfs2_destroy_inode(struct inode *inode)
585 {
586 call_rcu(&inode->i_rcu, ocfs2_i_callback);
587 }
588
ocfs2_max_file_offset(unsigned int bbits,unsigned int cbits)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 # if defined(CONFIG_LBDAF)
604 BUILD_BUG_ON(sizeof(sector_t) != 8);
605 /*
606 * We might be limited by page cache size.
607 */
608 if (bytes > PAGE_SIZE) {
609 bytes = PAGE_SIZE;
610 trim = 1;
611 /*
612 * Shift by 31 here so that we don't get larger than
613 * MAX_LFS_FILESIZE
614 */
615 bitshift = 31;
616 }
617 # else
618 /*
619 * We are limited by the size of sector_t. Use block size, as
620 * that's what we expose to the VFS.
621 */
622 bytes = 1 << bbits;
623 trim = 1;
624 bitshift = 31;
625 # endif
626 #endif
627
628 /*
629 * Trim by a whole cluster when we can actually approach the
630 * on-disk limits. Otherwise we can overflow i_clusters when
631 * an extent start is at the max offset.
632 */
633 return (((unsigned long long)bytes) << bitshift) - trim;
634 }
635
ocfs2_remount(struct super_block * sb,int * flags,char * data)636 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
637 {
638 int incompat_features;
639 int ret = 0;
640 struct mount_options parsed_options;
641 struct ocfs2_super *osb = OCFS2_SB(sb);
642 u32 tmp;
643
644 sync_filesystem(sb);
645
646 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
647 !ocfs2_check_set_options(sb, &parsed_options)) {
648 ret = -EINVAL;
649 goto out;
650 }
651
652 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
653 OCFS2_MOUNT_HB_NONE;
654 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
655 ret = -EINVAL;
656 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
657 goto out;
658 }
659
660 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
661 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
662 ret = -EINVAL;
663 mlog(ML_ERROR, "Cannot change data mode on remount\n");
664 goto out;
665 }
666
667 /* Probably don't want this on remount; it might
668 * mess with other nodes */
669 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
670 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
671 ret = -EINVAL;
672 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
673 goto out;
674 }
675
676 /* We're going to/from readonly mode. */
677 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
678 /* Disable quota accounting before remounting RO */
679 if (*flags & SB_RDONLY) {
680 ret = ocfs2_susp_quotas(osb, 0);
681 if (ret < 0)
682 goto out;
683 }
684 /* Lock here so the check of HARD_RO and the potential
685 * setting of SOFT_RO is atomic. */
686 spin_lock(&osb->osb_lock);
687 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
688 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
689 ret = -EROFS;
690 goto unlock_osb;
691 }
692
693 if (*flags & SB_RDONLY) {
694 sb->s_flags |= SB_RDONLY;
695 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696 } else {
697 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
698 mlog(ML_ERROR, "Cannot remount RDWR "
699 "filesystem due to previous errors.\n");
700 ret = -EROFS;
701 goto unlock_osb;
702 }
703 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
704 if (incompat_features) {
705 mlog(ML_ERROR, "Cannot remount RDWR because "
706 "of unsupported optional features "
707 "(%x).\n", incompat_features);
708 ret = -EINVAL;
709 goto unlock_osb;
710 }
711 sb->s_flags &= ~SB_RDONLY;
712 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
713 }
714 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
715 unlock_osb:
716 spin_unlock(&osb->osb_lock);
717 /* Enable quota accounting after remounting RW */
718 if (!ret && !(*flags & SB_RDONLY)) {
719 if (sb_any_quota_suspended(sb))
720 ret = ocfs2_susp_quotas(osb, 1);
721 else
722 ret = ocfs2_enable_quotas(osb);
723 if (ret < 0) {
724 /* Return back changes... */
725 spin_lock(&osb->osb_lock);
726 sb->s_flags |= SB_RDONLY;
727 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
728 spin_unlock(&osb->osb_lock);
729 goto out;
730 }
731 }
732 }
733
734 if (!ret) {
735 /* Only save off the new mount options in case of a successful
736 * remount. */
737 osb->s_mount_opt = parsed_options.mount_opt;
738 osb->s_atime_quantum = parsed_options.atime_quantum;
739 osb->preferred_slot = parsed_options.slot;
740 if (parsed_options.commit_interval)
741 osb->osb_commit_interval = parsed_options.commit_interval;
742
743 if (!ocfs2_is_hard_readonly(osb))
744 ocfs2_set_journal_params(osb);
745
746 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
747 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
748 SB_POSIXACL : 0);
749 }
750 out:
751 return ret;
752 }
753
ocfs2_sb_probe(struct super_block * sb,struct buffer_head ** bh,int * sector_size,struct ocfs2_blockcheck_stats * stats)754 static int ocfs2_sb_probe(struct super_block *sb,
755 struct buffer_head **bh,
756 int *sector_size,
757 struct ocfs2_blockcheck_stats *stats)
758 {
759 int status, tmpstat;
760 struct ocfs1_vol_disk_hdr *hdr;
761 struct ocfs2_dinode *di;
762 int blksize;
763
764 *bh = NULL;
765
766 /* may be > 512 */
767 *sector_size = bdev_logical_block_size(sb->s_bdev);
768 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
769 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
770 *sector_size, OCFS2_MAX_BLOCKSIZE);
771 status = -EINVAL;
772 goto bail;
773 }
774
775 /* Can this really happen? */
776 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
777 *sector_size = OCFS2_MIN_BLOCKSIZE;
778
779 /* check block zero for old format */
780 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
781 if (status < 0) {
782 mlog_errno(status);
783 goto bail;
784 }
785 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
786 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
787 mlog(ML_ERROR, "incompatible version: %u.%u\n",
788 hdr->major_version, hdr->minor_version);
789 status = -EINVAL;
790 }
791 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
792 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
793 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
794 hdr->signature);
795 status = -EINVAL;
796 }
797 brelse(*bh);
798 *bh = NULL;
799 if (status < 0) {
800 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
801 "upgraded before mounting with ocfs v2\n");
802 goto bail;
803 }
804
805 /*
806 * Now check at magic offset for 512, 1024, 2048, 4096
807 * blocksizes. 4096 is the maximum blocksize because it is
808 * the minimum clustersize.
809 */
810 status = -EINVAL;
811 for (blksize = *sector_size;
812 blksize <= OCFS2_MAX_BLOCKSIZE;
813 blksize <<= 1) {
814 tmpstat = ocfs2_get_sector(sb, bh,
815 OCFS2_SUPER_BLOCK_BLKNO,
816 blksize);
817 if (tmpstat < 0) {
818 status = tmpstat;
819 mlog_errno(status);
820 break;
821 }
822 di = (struct ocfs2_dinode *) (*bh)->b_data;
823 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
824 spin_lock_init(&stats->b_lock);
825 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
826 if (tmpstat < 0) {
827 brelse(*bh);
828 *bh = NULL;
829 }
830 if (tmpstat != -EAGAIN) {
831 status = tmpstat;
832 break;
833 }
834 }
835
836 bail:
837 return status;
838 }
839
ocfs2_verify_heartbeat(struct ocfs2_super * osb)840 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
841 {
842 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
843
844 if (osb->s_mount_opt & hb_enabled) {
845 if (ocfs2_mount_local(osb)) {
846 mlog(ML_ERROR, "Cannot heartbeat on a locally "
847 "mounted device.\n");
848 return -EINVAL;
849 }
850 if (ocfs2_userspace_stack(osb)) {
851 mlog(ML_ERROR, "Userspace stack expected, but "
852 "o2cb heartbeat arguments passed to mount\n");
853 return -EINVAL;
854 }
855 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
856 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
857 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
858 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
859 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
860 return -EINVAL;
861 }
862 }
863
864 if (!(osb->s_mount_opt & hb_enabled)) {
865 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
866 !ocfs2_userspace_stack(osb)) {
867 mlog(ML_ERROR, "Heartbeat has to be started to mount "
868 "a read-write clustered device.\n");
869 return -EINVAL;
870 }
871 }
872
873 return 0;
874 }
875
876 /*
877 * If we're using a userspace stack, mount should have passed
878 * a name that matches the disk. If not, mount should not
879 * have passed a stack.
880 */
ocfs2_verify_userspace_stack(struct ocfs2_super * osb,struct mount_options * mopt)881 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
882 struct mount_options *mopt)
883 {
884 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
885 mlog(ML_ERROR,
886 "cluster stack passed to mount, but this filesystem "
887 "does not support it\n");
888 return -EINVAL;
889 }
890
891 if (ocfs2_userspace_stack(osb) &&
892 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
893 OCFS2_STACK_LABEL_LEN)) {
894 mlog(ML_ERROR,
895 "cluster stack passed to mount (\"%s\") does not "
896 "match the filesystem (\"%s\")\n",
897 mopt->cluster_stack,
898 osb->osb_cluster_stack);
899 return -EINVAL;
900 }
901
902 return 0;
903 }
904
ocfs2_susp_quotas(struct ocfs2_super * osb,int unsuspend)905 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
906 {
907 int type;
908 struct super_block *sb = osb->sb;
909 unsigned int feature[OCFS2_MAXQUOTAS] = {
910 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
911 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
912 int status = 0;
913
914 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
915 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
916 continue;
917 if (unsuspend)
918 status = dquot_resume(sb, type);
919 else {
920 struct ocfs2_mem_dqinfo *oinfo;
921
922 /* Cancel periodic syncing before suspending */
923 oinfo = sb_dqinfo(sb, type)->dqi_priv;
924 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
925 status = dquot_suspend(sb, type);
926 }
927 if (status < 0)
928 break;
929 }
930 if (status < 0)
931 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
932 "remount (error = %d).\n", status);
933 return status;
934 }
935
ocfs2_enable_quotas(struct ocfs2_super * osb)936 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
937 {
938 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
939 struct super_block *sb = osb->sb;
940 unsigned int feature[OCFS2_MAXQUOTAS] = {
941 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
942 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
943 unsigned int ino[OCFS2_MAXQUOTAS] = {
944 LOCAL_USER_QUOTA_SYSTEM_INODE,
945 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
946 int status;
947 int type;
948
949 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
950 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
951 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
952 continue;
953 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
954 osb->slot_num);
955 if (!inode[type]) {
956 status = -ENOENT;
957 goto out_quota_off;
958 }
959 status = dquot_enable(inode[type], type, QFMT_OCFS2,
960 DQUOT_USAGE_ENABLED);
961 if (status < 0)
962 goto out_quota_off;
963 }
964
965 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
966 iput(inode[type]);
967 return 0;
968 out_quota_off:
969 ocfs2_disable_quotas(osb);
970 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
971 iput(inode[type]);
972 mlog_errno(status);
973 return status;
974 }
975
ocfs2_disable_quotas(struct ocfs2_super * osb)976 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
977 {
978 int type;
979 struct inode *inode;
980 struct super_block *sb = osb->sb;
981 struct ocfs2_mem_dqinfo *oinfo;
982
983 /* We mostly ignore errors in this function because there's not much
984 * we can do when we see them */
985 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
986 if (!sb_has_quota_loaded(sb, type))
987 continue;
988 if (!sb_has_quota_suspended(sb, type)) {
989 oinfo = sb_dqinfo(sb, type)->dqi_priv;
990 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
991 }
992 inode = igrab(sb->s_dquot.files[type]);
993 /* Turn off quotas. This will remove all dquot structures from
994 * memory and so they will be automatically synced to global
995 * quota files */
996 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
997 DQUOT_LIMITS_ENABLED);
998 if (!inode)
999 continue;
1000 iput(inode);
1001 }
1002 }
1003
ocfs2_fill_super(struct super_block * sb,void * data,int silent)1004 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1005 {
1006 struct dentry *root;
1007 int status, sector_size;
1008 struct mount_options parsed_options;
1009 struct inode *inode = NULL;
1010 struct ocfs2_super *osb = NULL;
1011 struct buffer_head *bh = NULL;
1012 char nodestr[12];
1013 struct ocfs2_blockcheck_stats stats;
1014
1015 trace_ocfs2_fill_super(sb, data, silent);
1016
1017 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1018 status = -EINVAL;
1019 goto read_super_error;
1020 }
1021
1022 /* probe for superblock */
1023 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1024 if (status < 0) {
1025 mlog(ML_ERROR, "superblock probe failed!\n");
1026 goto read_super_error;
1027 }
1028
1029 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1030 osb = OCFS2_SB(sb);
1031 if (status < 0) {
1032 mlog_errno(status);
1033 goto read_super_error;
1034 }
1035 brelse(bh);
1036 bh = NULL;
1037
1038 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1039 status = -EINVAL;
1040 goto read_super_error;
1041 }
1042 osb->s_mount_opt = parsed_options.mount_opt;
1043 osb->s_atime_quantum = parsed_options.atime_quantum;
1044 osb->preferred_slot = parsed_options.slot;
1045 osb->osb_commit_interval = parsed_options.commit_interval;
1046
1047 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1048 osb->osb_resv_level = parsed_options.resv_level;
1049 osb->osb_dir_resv_level = parsed_options.resv_level;
1050 if (parsed_options.dir_resv_level == -1)
1051 osb->osb_dir_resv_level = parsed_options.resv_level;
1052 else
1053 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1054
1055 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1056 if (status)
1057 goto read_super_error;
1058
1059 sb->s_magic = OCFS2_SUPER_MAGIC;
1060
1061 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1062 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1063
1064 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1065 * heartbeat=none */
1066 if (bdev_read_only(sb->s_bdev)) {
1067 if (!sb_rdonly(sb)) {
1068 status = -EACCES;
1069 mlog(ML_ERROR, "Readonly device detected but readonly "
1070 "mount was not specified.\n");
1071 goto read_super_error;
1072 }
1073
1074 /* You should not be able to start a local heartbeat
1075 * on a readonly device. */
1076 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1077 status = -EROFS;
1078 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1079 "device.\n");
1080 goto read_super_error;
1081 }
1082
1083 status = ocfs2_check_journals_nolocks(osb);
1084 if (status < 0) {
1085 if (status == -EROFS)
1086 mlog(ML_ERROR, "Recovery required on readonly "
1087 "file system, but write access is "
1088 "unavailable.\n");
1089 else
1090 mlog_errno(status);
1091 goto read_super_error;
1092 }
1093
1094 ocfs2_set_ro_flag(osb, 1);
1095
1096 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1097 "Cluster services will not be used for this mount. "
1098 "Recovery will be skipped.\n", osb->dev_str);
1099 }
1100
1101 if (!ocfs2_is_hard_readonly(osb)) {
1102 if (sb_rdonly(sb))
1103 ocfs2_set_ro_flag(osb, 0);
1104 }
1105
1106 status = ocfs2_verify_heartbeat(osb);
1107 if (status < 0) {
1108 mlog_errno(status);
1109 goto read_super_error;
1110 }
1111
1112 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1113 ocfs2_debugfs_root);
1114 if (!osb->osb_debug_root) {
1115 status = -EINVAL;
1116 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1117 goto read_super_error;
1118 }
1119
1120 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1121 osb->osb_debug_root,
1122 osb,
1123 &ocfs2_osb_debug_fops);
1124 if (!osb->osb_ctxt) {
1125 status = -EINVAL;
1126 mlog_errno(status);
1127 goto read_super_error;
1128 }
1129
1130 if (ocfs2_meta_ecc(osb)) {
1131 status = ocfs2_blockcheck_stats_debugfs_install(
1132 &osb->osb_ecc_stats,
1133 osb->osb_debug_root);
1134 if (status) {
1135 mlog(ML_ERROR,
1136 "Unable to create blockcheck statistics "
1137 "files\n");
1138 goto read_super_error;
1139 }
1140 }
1141
1142 status = ocfs2_mount_volume(sb);
1143 if (status < 0)
1144 goto read_super_error;
1145
1146 if (osb->root_inode)
1147 inode = igrab(osb->root_inode);
1148
1149 if (!inode) {
1150 status = -EIO;
1151 mlog_errno(status);
1152 goto read_super_error;
1153 }
1154
1155 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1156 &ocfs2_kset->kobj);
1157 if (!osb->osb_dev_kset) {
1158 status = -ENOMEM;
1159 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1160 goto read_super_error;
1161 }
1162
1163 /* Create filecheck sysfs related directories/files at
1164 * /sys/fs/ocfs2/<devname>/filecheck */
1165 if (ocfs2_filecheck_create_sysfs(osb)) {
1166 status = -ENOMEM;
1167 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1168 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1169 goto read_super_error;
1170 }
1171
1172 root = d_make_root(inode);
1173 if (!root) {
1174 status = -ENOMEM;
1175 mlog_errno(status);
1176 goto read_super_error;
1177 }
1178
1179 sb->s_root = root;
1180
1181 ocfs2_complete_mount_recovery(osb);
1182
1183 if (ocfs2_mount_local(osb))
1184 snprintf(nodestr, sizeof(nodestr), "local");
1185 else
1186 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1187
1188 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1189 "with %s data mode.\n",
1190 osb->dev_str, nodestr, osb->slot_num,
1191 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1192 "ordered");
1193
1194 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1195 wake_up(&osb->osb_mount_event);
1196
1197 /* Now we can initialize quotas because we can afford to wait
1198 * for cluster locks recovery now. That also means that truncation
1199 * log recovery can happen but that waits for proper quota setup */
1200 if (!sb_rdonly(sb)) {
1201 status = ocfs2_enable_quotas(osb);
1202 if (status < 0) {
1203 /* We have to err-out specially here because
1204 * s_root is already set */
1205 mlog_errno(status);
1206 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1207 wake_up(&osb->osb_mount_event);
1208 return status;
1209 }
1210 }
1211
1212 ocfs2_complete_quota_recovery(osb);
1213
1214 /* Now we wake up again for processes waiting for quotas */
1215 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1216 wake_up(&osb->osb_mount_event);
1217
1218 /* Start this when the mount is almost sure of being successful */
1219 ocfs2_orphan_scan_start(osb);
1220
1221 return status;
1222
1223 read_super_error:
1224 brelse(bh);
1225
1226 if (status)
1227 mlog_errno(status);
1228
1229 if (osb) {
1230 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1231 wake_up(&osb->osb_mount_event);
1232 ocfs2_dismount_volume(sb, 1);
1233 }
1234
1235 return status;
1236 }
1237
ocfs2_mount(struct file_system_type * fs_type,int flags,const char * dev_name,void * data)1238 static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1239 int flags,
1240 const char *dev_name,
1241 void *data)
1242 {
1243 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1244 }
1245
1246 static struct file_system_type ocfs2_fs_type = {
1247 .owner = THIS_MODULE,
1248 .name = "ocfs2",
1249 .mount = ocfs2_mount,
1250 .kill_sb = kill_block_super,
1251 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1252 .next = NULL
1253 };
1254 MODULE_ALIAS_FS("ocfs2");
1255
ocfs2_check_set_options(struct super_block * sb,struct mount_options * options)1256 static int ocfs2_check_set_options(struct super_block *sb,
1257 struct mount_options *options)
1258 {
1259 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1260 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1261 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1262 mlog(ML_ERROR, "User quotas were requested, but this "
1263 "filesystem does not have the feature enabled.\n");
1264 return 0;
1265 }
1266 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1267 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1268 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1269 mlog(ML_ERROR, "Group quotas were requested, but this "
1270 "filesystem does not have the feature enabled.\n");
1271 return 0;
1272 }
1273 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1274 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1275 mlog(ML_ERROR, "ACL support requested but extended attributes "
1276 "feature is not enabled\n");
1277 return 0;
1278 }
1279 /* No ACL setting specified? Use XATTR feature... */
1280 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1281 OCFS2_MOUNT_NO_POSIX_ACL))) {
1282 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1283 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1284 else
1285 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1286 }
1287 return 1;
1288 }
1289
ocfs2_parse_options(struct super_block * sb,char * options,struct mount_options * mopt,int is_remount)1290 static int ocfs2_parse_options(struct super_block *sb,
1291 char *options,
1292 struct mount_options *mopt,
1293 int is_remount)
1294 {
1295 int status, user_stack = 0;
1296 char *p;
1297 u32 tmp;
1298 int token, option;
1299 substring_t args[MAX_OPT_ARGS];
1300
1301 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1302
1303 mopt->commit_interval = 0;
1304 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1305 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1306 mopt->slot = OCFS2_INVALID_SLOT;
1307 mopt->localalloc_opt = -1;
1308 mopt->cluster_stack[0] = '\0';
1309 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1310 mopt->dir_resv_level = -1;
1311
1312 if (!options) {
1313 status = 1;
1314 goto bail;
1315 }
1316
1317 while ((p = strsep(&options, ",")) != NULL) {
1318 if (!*p)
1319 continue;
1320
1321 token = match_token(p, tokens, args);
1322 switch (token) {
1323 case Opt_hb_local:
1324 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1325 break;
1326 case Opt_hb_none:
1327 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1328 break;
1329 case Opt_hb_global:
1330 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1331 break;
1332 case Opt_barrier:
1333 if (match_int(&args[0], &option)) {
1334 status = 0;
1335 goto bail;
1336 }
1337 if (option)
1338 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1339 else
1340 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1341 break;
1342 case Opt_intr:
1343 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1344 break;
1345 case Opt_nointr:
1346 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1347 break;
1348 case Opt_err_panic:
1349 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1350 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1351 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1352 break;
1353 case Opt_err_ro:
1354 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1355 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1356 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1357 break;
1358 case Opt_err_cont:
1359 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1360 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1361 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1362 break;
1363 case Opt_data_ordered:
1364 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1365 break;
1366 case Opt_data_writeback:
1367 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1368 break;
1369 case Opt_user_xattr:
1370 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1371 break;
1372 case Opt_nouser_xattr:
1373 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1374 break;
1375 case Opt_atime_quantum:
1376 if (match_int(&args[0], &option)) {
1377 status = 0;
1378 goto bail;
1379 }
1380 if (option >= 0)
1381 mopt->atime_quantum = option;
1382 break;
1383 case Opt_slot:
1384 if (match_int(&args[0], &option)) {
1385 status = 0;
1386 goto bail;
1387 }
1388 if (option)
1389 mopt->slot = (u16)option;
1390 break;
1391 case Opt_commit:
1392 if (match_int(&args[0], &option)) {
1393 status = 0;
1394 goto bail;
1395 }
1396 if (option < 0)
1397 return 0;
1398 if (option == 0)
1399 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1400 mopt->commit_interval = HZ * option;
1401 break;
1402 case Opt_localalloc:
1403 if (match_int(&args[0], &option)) {
1404 status = 0;
1405 goto bail;
1406 }
1407 if (option >= 0)
1408 mopt->localalloc_opt = option;
1409 break;
1410 case Opt_localflocks:
1411 /*
1412 * Changing this during remount could race
1413 * flock() requests, or "unbalance" existing
1414 * ones (e.g., a lock is taken in one mode but
1415 * dropped in the other). If users care enough
1416 * to flip locking modes during remount, we
1417 * could add a "local" flag to individual
1418 * flock structures for proper tracking of
1419 * state.
1420 */
1421 if (!is_remount)
1422 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1423 break;
1424 case Opt_stack:
1425 /* Check both that the option we were passed
1426 * is of the right length and that it is a proper
1427 * string of the right length.
1428 */
1429 if (((args[0].to - args[0].from) !=
1430 OCFS2_STACK_LABEL_LEN) ||
1431 (strnlen(args[0].from,
1432 OCFS2_STACK_LABEL_LEN) !=
1433 OCFS2_STACK_LABEL_LEN)) {
1434 mlog(ML_ERROR,
1435 "Invalid cluster_stack option\n");
1436 status = 0;
1437 goto bail;
1438 }
1439 memcpy(mopt->cluster_stack, args[0].from,
1440 OCFS2_STACK_LABEL_LEN);
1441 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1442 /*
1443 * Open code the memcmp here as we don't have
1444 * an osb to pass to
1445 * ocfs2_userspace_stack().
1446 */
1447 if (memcmp(mopt->cluster_stack,
1448 OCFS2_CLASSIC_CLUSTER_STACK,
1449 OCFS2_STACK_LABEL_LEN))
1450 user_stack = 1;
1451 break;
1452 case Opt_inode64:
1453 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1454 break;
1455 case Opt_usrquota:
1456 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1457 break;
1458 case Opt_grpquota:
1459 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1460 break;
1461 case Opt_coherency_buffered:
1462 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1463 break;
1464 case Opt_coherency_full:
1465 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1466 break;
1467 case Opt_acl:
1468 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1469 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1470 break;
1471 case Opt_noacl:
1472 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1473 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1474 break;
1475 case Opt_resv_level:
1476 if (is_remount)
1477 break;
1478 if (match_int(&args[0], &option)) {
1479 status = 0;
1480 goto bail;
1481 }
1482 if (option >= OCFS2_MIN_RESV_LEVEL &&
1483 option < OCFS2_MAX_RESV_LEVEL)
1484 mopt->resv_level = option;
1485 break;
1486 case Opt_dir_resv_level:
1487 if (is_remount)
1488 break;
1489 if (match_int(&args[0], &option)) {
1490 status = 0;
1491 goto bail;
1492 }
1493 if (option >= OCFS2_MIN_RESV_LEVEL &&
1494 option < OCFS2_MAX_RESV_LEVEL)
1495 mopt->dir_resv_level = option;
1496 break;
1497 case Opt_journal_async_commit:
1498 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1499 break;
1500 default:
1501 mlog(ML_ERROR,
1502 "Unrecognized mount option \"%s\" "
1503 "or missing value\n", p);
1504 status = 0;
1505 goto bail;
1506 }
1507 }
1508
1509 if (user_stack == 0) {
1510 /* Ensure only one heartbeat mode */
1511 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1512 OCFS2_MOUNT_HB_GLOBAL |
1513 OCFS2_MOUNT_HB_NONE);
1514 if (hweight32(tmp) != 1) {
1515 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1516 status = 0;
1517 goto bail;
1518 }
1519 }
1520
1521 status = 1;
1522
1523 bail:
1524 return status;
1525 }
1526
ocfs2_show_options(struct seq_file * s,struct dentry * root)1527 static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1528 {
1529 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1530 unsigned long opts = osb->s_mount_opt;
1531 unsigned int local_alloc_megs;
1532
1533 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1534 seq_printf(s, ",_netdev");
1535 if (opts & OCFS2_MOUNT_HB_LOCAL)
1536 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1537 else
1538 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1539 } else
1540 seq_printf(s, ",%s", OCFS2_HB_NONE);
1541
1542 if (opts & OCFS2_MOUNT_NOINTR)
1543 seq_printf(s, ",nointr");
1544
1545 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1546 seq_printf(s, ",data=writeback");
1547 else
1548 seq_printf(s, ",data=ordered");
1549
1550 if (opts & OCFS2_MOUNT_BARRIER)
1551 seq_printf(s, ",barrier=1");
1552
1553 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1554 seq_printf(s, ",errors=panic");
1555 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1556 seq_printf(s, ",errors=continue");
1557 else
1558 seq_printf(s, ",errors=remount-ro");
1559
1560 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1561 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1562
1563 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1564
1565 if (osb->osb_commit_interval)
1566 seq_printf(s, ",commit=%u",
1567 (unsigned) (osb->osb_commit_interval / HZ));
1568
1569 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1570 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1571 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1572
1573 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1574 seq_printf(s, ",localflocks,");
1575
1576 if (osb->osb_cluster_stack[0])
1577 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1578 OCFS2_STACK_LABEL_LEN);
1579 if (opts & OCFS2_MOUNT_USRQUOTA)
1580 seq_printf(s, ",usrquota");
1581 if (opts & OCFS2_MOUNT_GRPQUOTA)
1582 seq_printf(s, ",grpquota");
1583
1584 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1585 seq_printf(s, ",coherency=buffered");
1586 else
1587 seq_printf(s, ",coherency=full");
1588
1589 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1590 seq_printf(s, ",nouser_xattr");
1591 else
1592 seq_printf(s, ",user_xattr");
1593
1594 if (opts & OCFS2_MOUNT_INODE64)
1595 seq_printf(s, ",inode64");
1596
1597 if (opts & OCFS2_MOUNT_POSIX_ACL)
1598 seq_printf(s, ",acl");
1599 else
1600 seq_printf(s, ",noacl");
1601
1602 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1603 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1604
1605 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1606 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1607
1608 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1609 seq_printf(s, ",journal_async_commit");
1610
1611 return 0;
1612 }
1613
ocfs2_init(void)1614 static int __init ocfs2_init(void)
1615 {
1616 int status;
1617
1618 status = init_ocfs2_uptodate_cache();
1619 if (status < 0)
1620 goto out1;
1621
1622 status = ocfs2_initialize_mem_caches();
1623 if (status < 0)
1624 goto out2;
1625
1626 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1627 if (!ocfs2_debugfs_root) {
1628 status = -ENOMEM;
1629 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1630 goto out3;
1631 }
1632
1633 ocfs2_set_locking_protocol();
1634
1635 status = register_quota_format(&ocfs2_quota_format);
1636 if (status < 0)
1637 goto out3;
1638 status = register_filesystem(&ocfs2_fs_type);
1639 if (!status)
1640 return 0;
1641
1642 unregister_quota_format(&ocfs2_quota_format);
1643 out3:
1644 debugfs_remove(ocfs2_debugfs_root);
1645 ocfs2_free_mem_caches();
1646 out2:
1647 exit_ocfs2_uptodate_cache();
1648 out1:
1649 mlog_errno(status);
1650 return status;
1651 }
1652
ocfs2_exit(void)1653 static void __exit ocfs2_exit(void)
1654 {
1655 unregister_quota_format(&ocfs2_quota_format);
1656
1657 debugfs_remove(ocfs2_debugfs_root);
1658
1659 ocfs2_free_mem_caches();
1660
1661 unregister_filesystem(&ocfs2_fs_type);
1662
1663 exit_ocfs2_uptodate_cache();
1664 }
1665
ocfs2_put_super(struct super_block * sb)1666 static void ocfs2_put_super(struct super_block *sb)
1667 {
1668 trace_ocfs2_put_super(sb);
1669
1670 ocfs2_sync_blockdev(sb);
1671 ocfs2_dismount_volume(sb, 0);
1672 }
1673
ocfs2_statfs(struct dentry * dentry,struct kstatfs * buf)1674 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1675 {
1676 struct ocfs2_super *osb;
1677 u32 numbits, freebits;
1678 int status;
1679 struct ocfs2_dinode *bm_lock;
1680 struct buffer_head *bh = NULL;
1681 struct inode *inode = NULL;
1682
1683 trace_ocfs2_statfs(dentry->d_sb, buf);
1684
1685 osb = OCFS2_SB(dentry->d_sb);
1686
1687 inode = ocfs2_get_system_file_inode(osb,
1688 GLOBAL_BITMAP_SYSTEM_INODE,
1689 OCFS2_INVALID_SLOT);
1690 if (!inode) {
1691 mlog(ML_ERROR, "failed to get bitmap inode\n");
1692 status = -EIO;
1693 goto bail;
1694 }
1695
1696 status = ocfs2_inode_lock(inode, &bh, 0);
1697 if (status < 0) {
1698 mlog_errno(status);
1699 goto bail;
1700 }
1701
1702 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1703
1704 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1705 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1706
1707 buf->f_type = OCFS2_SUPER_MAGIC;
1708 buf->f_bsize = dentry->d_sb->s_blocksize;
1709 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1710 buf->f_blocks = ((sector_t) numbits) *
1711 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1712 buf->f_bfree = ((sector_t) freebits) *
1713 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1714 buf->f_bavail = buf->f_bfree;
1715 buf->f_files = numbits;
1716 buf->f_ffree = freebits;
1717 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1718 & 0xFFFFFFFFUL;
1719 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1720 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1721
1722 brelse(bh);
1723
1724 ocfs2_inode_unlock(inode, 0);
1725 status = 0;
1726 bail:
1727 iput(inode);
1728
1729 if (status)
1730 mlog_errno(status);
1731
1732 return status;
1733 }
1734
ocfs2_inode_init_once(void * data)1735 static void ocfs2_inode_init_once(void *data)
1736 {
1737 struct ocfs2_inode_info *oi = data;
1738
1739 oi->ip_flags = 0;
1740 oi->ip_open_count = 0;
1741 spin_lock_init(&oi->ip_lock);
1742 ocfs2_extent_map_init(&oi->vfs_inode);
1743 INIT_LIST_HEAD(&oi->ip_io_markers);
1744 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1745 oi->ip_dir_start_lookup = 0;
1746 init_rwsem(&oi->ip_alloc_sem);
1747 init_rwsem(&oi->ip_xattr_sem);
1748 mutex_init(&oi->ip_io_mutex);
1749
1750 oi->ip_blkno = 0ULL;
1751 oi->ip_clusters = 0;
1752 oi->ip_next_orphan = NULL;
1753
1754 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1755
1756 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1757 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1758 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1759
1760 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1761 &ocfs2_inode_caching_ops);
1762
1763 inode_init_once(&oi->vfs_inode);
1764 }
1765
ocfs2_initialize_mem_caches(void)1766 static int ocfs2_initialize_mem_caches(void)
1767 {
1768 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1769 sizeof(struct ocfs2_inode_info),
1770 0,
1771 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1772 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1773 ocfs2_inode_init_once);
1774 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1775 sizeof(struct ocfs2_dquot),
1776 0,
1777 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1778 SLAB_MEM_SPREAD),
1779 NULL);
1780 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1781 sizeof(struct ocfs2_quota_chunk),
1782 0,
1783 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1784 NULL);
1785 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1786 !ocfs2_qf_chunk_cachep) {
1787 kmem_cache_destroy(ocfs2_inode_cachep);
1788 kmem_cache_destroy(ocfs2_dquot_cachep);
1789 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1790 return -ENOMEM;
1791 }
1792
1793 return 0;
1794 }
1795
ocfs2_free_mem_caches(void)1796 static void ocfs2_free_mem_caches(void)
1797 {
1798 /*
1799 * Make sure all delayed rcu free inodes are flushed before we
1800 * destroy cache.
1801 */
1802 rcu_barrier();
1803 kmem_cache_destroy(ocfs2_inode_cachep);
1804 ocfs2_inode_cachep = NULL;
1805
1806 kmem_cache_destroy(ocfs2_dquot_cachep);
1807 ocfs2_dquot_cachep = NULL;
1808
1809 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1810 ocfs2_qf_chunk_cachep = NULL;
1811 }
1812
ocfs2_get_sector(struct super_block * sb,struct buffer_head ** bh,int block,int sect_size)1813 static int ocfs2_get_sector(struct super_block *sb,
1814 struct buffer_head **bh,
1815 int block,
1816 int sect_size)
1817 {
1818 if (!sb_set_blocksize(sb, sect_size)) {
1819 mlog(ML_ERROR, "unable to set blocksize\n");
1820 return -EIO;
1821 }
1822
1823 *bh = sb_getblk(sb, block);
1824 if (!*bh) {
1825 mlog_errno(-ENOMEM);
1826 return -ENOMEM;
1827 }
1828 lock_buffer(*bh);
1829 if (!buffer_dirty(*bh))
1830 clear_buffer_uptodate(*bh);
1831 unlock_buffer(*bh);
1832 ll_rw_block(REQ_OP_READ, 0, 1, bh);
1833 wait_on_buffer(*bh);
1834 if (!buffer_uptodate(*bh)) {
1835 mlog_errno(-EIO);
1836 brelse(*bh);
1837 *bh = NULL;
1838 return -EIO;
1839 }
1840
1841 return 0;
1842 }
1843
ocfs2_mount_volume(struct super_block * sb)1844 static int ocfs2_mount_volume(struct super_block *sb)
1845 {
1846 int status = 0;
1847 int unlock_super = 0;
1848 struct ocfs2_super *osb = OCFS2_SB(sb);
1849
1850 if (ocfs2_is_hard_readonly(osb))
1851 goto leave;
1852
1853 status = ocfs2_dlm_init(osb);
1854 if (status < 0) {
1855 mlog_errno(status);
1856 if (status == -EBADR && ocfs2_userspace_stack(osb))
1857 mlog(ML_ERROR, "couldn't mount because cluster name on"
1858 " disk does not match the running cluster name.\n");
1859 goto leave;
1860 }
1861
1862 status = ocfs2_super_lock(osb, 1);
1863 if (status < 0) {
1864 mlog_errno(status);
1865 goto leave;
1866 }
1867 unlock_super = 1;
1868
1869 /* This will load up the node map and add ourselves to it. */
1870 status = ocfs2_find_slot(osb);
1871 if (status < 0) {
1872 mlog_errno(status);
1873 goto leave;
1874 }
1875
1876 /* load all node-local system inodes */
1877 status = ocfs2_init_local_system_inodes(osb);
1878 if (status < 0) {
1879 mlog_errno(status);
1880 goto leave;
1881 }
1882
1883 status = ocfs2_check_volume(osb);
1884 if (status < 0) {
1885 mlog_errno(status);
1886 goto leave;
1887 }
1888
1889 status = ocfs2_truncate_log_init(osb);
1890 if (status < 0)
1891 mlog_errno(status);
1892
1893 leave:
1894 if (unlock_super)
1895 ocfs2_super_unlock(osb, 1);
1896
1897 return status;
1898 }
1899
ocfs2_dismount_volume(struct super_block * sb,int mnt_err)1900 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1901 {
1902 int tmp, hangup_needed = 0;
1903 struct ocfs2_super *osb = NULL;
1904 char nodestr[12];
1905
1906 trace_ocfs2_dismount_volume(sb);
1907
1908 BUG_ON(!sb);
1909 osb = OCFS2_SB(sb);
1910 BUG_ON(!osb);
1911
1912 /* Remove file check sysfs related directores/files,
1913 * and wait for the pending file check operations */
1914 ocfs2_filecheck_remove_sysfs(osb);
1915
1916 kset_unregister(osb->osb_dev_kset);
1917
1918 debugfs_remove(osb->osb_ctxt);
1919
1920 /* Orphan scan should be stopped as early as possible */
1921 ocfs2_orphan_scan_stop(osb);
1922
1923 ocfs2_disable_quotas(osb);
1924
1925 /* All dquots should be freed by now */
1926 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1927 /* Wait for worker to be done with the work structure in osb */
1928 cancel_work_sync(&osb->dquot_drop_work);
1929
1930 ocfs2_shutdown_local_alloc(osb);
1931
1932 ocfs2_truncate_log_shutdown(osb);
1933
1934 /* This will disable recovery and flush any recovery work. */
1935 ocfs2_recovery_exit(osb);
1936
1937 ocfs2_journal_shutdown(osb);
1938
1939 ocfs2_sync_blockdev(sb);
1940
1941 ocfs2_purge_refcount_trees(osb);
1942
1943 /* No cluster connection means we've failed during mount, so skip
1944 * all the steps which depended on that to complete. */
1945 if (osb->cconn) {
1946 tmp = ocfs2_super_lock(osb, 1);
1947 if (tmp < 0) {
1948 mlog_errno(tmp);
1949 return;
1950 }
1951 }
1952
1953 if (osb->slot_num != OCFS2_INVALID_SLOT)
1954 ocfs2_put_slot(osb);
1955
1956 if (osb->cconn)
1957 ocfs2_super_unlock(osb, 1);
1958
1959 ocfs2_release_system_inodes(osb);
1960
1961 /*
1962 * If we're dismounting due to mount error, mount.ocfs2 will clean
1963 * up heartbeat. If we're a local mount, there is no heartbeat.
1964 * If we failed before we got a uuid_str yet, we can't stop
1965 * heartbeat. Otherwise, do it.
1966 */
1967 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1968 !ocfs2_is_hard_readonly(osb))
1969 hangup_needed = 1;
1970
1971 if (osb->cconn)
1972 ocfs2_dlm_shutdown(osb, hangup_needed);
1973
1974 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1975 debugfs_remove(osb->osb_debug_root);
1976
1977 if (hangup_needed)
1978 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1979
1980 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1981
1982 if (ocfs2_mount_local(osb))
1983 snprintf(nodestr, sizeof(nodestr), "local");
1984 else
1985 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1986
1987 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1988 osb->dev_str, nodestr);
1989
1990 ocfs2_delete_osb(osb);
1991 kfree(osb);
1992 sb->s_dev = 0;
1993 sb->s_fs_info = NULL;
1994 }
1995
ocfs2_setup_osb_uuid(struct ocfs2_super * osb,const unsigned char * uuid,unsigned uuid_bytes)1996 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1997 unsigned uuid_bytes)
1998 {
1999 int i, ret;
2000 char *ptr;
2001
2002 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2003
2004 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2005 if (osb->uuid_str == NULL)
2006 return -ENOMEM;
2007
2008 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2009 /* print with null */
2010 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2011 if (ret != 2) /* drop super cleans up */
2012 return -EINVAL;
2013 /* then only advance past the last char */
2014 ptr += 2;
2015 }
2016
2017 return 0;
2018 }
2019
2020 /* Make sure entire volume is addressable by our journal. Requires
2021 osb_clusters_at_boot to be valid and for the journal to have been
2022 initialized by ocfs2_journal_init(). */
ocfs2_journal_addressable(struct ocfs2_super * osb)2023 static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2024 {
2025 int status = 0;
2026 u64 max_block =
2027 ocfs2_clusters_to_blocks(osb->sb,
2028 osb->osb_clusters_at_boot) - 1;
2029
2030 /* 32-bit block number is always OK. */
2031 if (max_block <= (u32)~0ULL)
2032 goto out;
2033
2034 /* Volume is "huge", so see if our journal is new enough to
2035 support it. */
2036 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2037 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2038 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2039 JBD2_FEATURE_INCOMPAT_64BIT))) {
2040 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2041 "Enable the 'block64' journal option with tunefs.ocfs2");
2042 status = -EFBIG;
2043 goto out;
2044 }
2045
2046 out:
2047 return status;
2048 }
2049
ocfs2_initialize_super(struct super_block * sb,struct buffer_head * bh,int sector_size,struct ocfs2_blockcheck_stats * stats)2050 static int ocfs2_initialize_super(struct super_block *sb,
2051 struct buffer_head *bh,
2052 int sector_size,
2053 struct ocfs2_blockcheck_stats *stats)
2054 {
2055 int status;
2056 int i, cbits, bbits;
2057 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2058 struct inode *inode = NULL;
2059 struct ocfs2_journal *journal;
2060 struct ocfs2_super *osb;
2061 u64 total_blocks;
2062
2063 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2064 if (!osb) {
2065 status = -ENOMEM;
2066 mlog_errno(status);
2067 goto bail;
2068 }
2069
2070 sb->s_fs_info = osb;
2071 sb->s_op = &ocfs2_sops;
2072 sb->s_d_op = &ocfs2_dentry_ops;
2073 sb->s_export_op = &ocfs2_export_ops;
2074 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2075 sb->dq_op = &ocfs2_quota_operations;
2076 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2077 sb->s_xattr = ocfs2_xattr_handlers;
2078 sb->s_time_gran = 1;
2079 sb->s_flags |= SB_NOATIME;
2080 /* this is needed to support O_LARGEFILE */
2081 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2082 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2083 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2084 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2085 sizeof(di->id2.i_super.s_uuid));
2086
2087 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2088
2089 for (i = 0; i < 3; i++)
2090 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2091 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2092
2093 osb->sb = sb;
2094 osb->s_sectsize_bits = blksize_bits(sector_size);
2095 BUG_ON(!osb->s_sectsize_bits);
2096
2097 spin_lock_init(&osb->dc_task_lock);
2098 init_waitqueue_head(&osb->dc_event);
2099 osb->dc_work_sequence = 0;
2100 osb->dc_wake_sequence = 0;
2101 INIT_LIST_HEAD(&osb->blocked_lock_list);
2102 osb->blocked_lock_count = 0;
2103 spin_lock_init(&osb->osb_lock);
2104 spin_lock_init(&osb->osb_xattr_lock);
2105 ocfs2_init_steal_slots(osb);
2106
2107 mutex_init(&osb->system_file_mutex);
2108
2109 atomic_set(&osb->alloc_stats.moves, 0);
2110 atomic_set(&osb->alloc_stats.local_data, 0);
2111 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2112 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2113 atomic_set(&osb->alloc_stats.bg_extends, 0);
2114
2115 /* Copy the blockcheck stats from the superblock probe */
2116 osb->osb_ecc_stats = *stats;
2117
2118 ocfs2_init_node_maps(osb);
2119
2120 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2121 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2122
2123 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2124 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2125 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2126 osb->max_slots);
2127 status = -EINVAL;
2128 goto bail;
2129 }
2130
2131 ocfs2_orphan_scan_init(osb);
2132
2133 status = ocfs2_recovery_init(osb);
2134 if (status) {
2135 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2136 mlog_errno(status);
2137 goto bail;
2138 }
2139
2140 init_waitqueue_head(&osb->checkpoint_event);
2141
2142 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2143
2144 osb->slot_num = OCFS2_INVALID_SLOT;
2145
2146 osb->s_xattr_inline_size = le16_to_cpu(
2147 di->id2.i_super.s_xattr_inline_size);
2148
2149 osb->local_alloc_state = OCFS2_LA_UNUSED;
2150 osb->local_alloc_bh = NULL;
2151 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2152
2153 init_waitqueue_head(&osb->osb_mount_event);
2154
2155 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2156 if (status) {
2157 mlog_errno(status);
2158 goto bail;
2159 }
2160
2161 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2162 if (!osb->vol_label) {
2163 mlog(ML_ERROR, "unable to alloc vol label\n");
2164 status = -ENOMEM;
2165 goto bail;
2166 }
2167
2168 osb->slot_recovery_generations =
2169 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2170 GFP_KERNEL);
2171 if (!osb->slot_recovery_generations) {
2172 status = -ENOMEM;
2173 mlog_errno(status);
2174 goto bail;
2175 }
2176
2177 init_waitqueue_head(&osb->osb_wipe_event);
2178 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2179 sizeof(*osb->osb_orphan_wipes),
2180 GFP_KERNEL);
2181 if (!osb->osb_orphan_wipes) {
2182 status = -ENOMEM;
2183 mlog_errno(status);
2184 goto bail;
2185 }
2186
2187 osb->osb_rf_lock_tree = RB_ROOT;
2188
2189 osb->s_feature_compat =
2190 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2191 osb->s_feature_ro_compat =
2192 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2193 osb->s_feature_incompat =
2194 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2195
2196 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2197 mlog(ML_ERROR, "couldn't mount because of unsupported "
2198 "optional features (%x).\n", i);
2199 status = -EINVAL;
2200 goto bail;
2201 }
2202 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2203 mlog(ML_ERROR, "couldn't mount RDWR because of "
2204 "unsupported optional features (%x).\n", i);
2205 status = -EINVAL;
2206 goto bail;
2207 }
2208
2209 if (ocfs2_clusterinfo_valid(osb)) {
2210 /*
2211 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2212 * terminated, so make sure no overflow happens here by using
2213 * memcpy. Destination strings will always be null terminated
2214 * because osb is allocated using kzalloc.
2215 */
2216 osb->osb_stackflags =
2217 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2218 memcpy(osb->osb_cluster_stack,
2219 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2220 OCFS2_STACK_LABEL_LEN);
2221 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2222 mlog(ML_ERROR,
2223 "couldn't mount because of an invalid "
2224 "cluster stack label (%s) \n",
2225 osb->osb_cluster_stack);
2226 status = -EINVAL;
2227 goto bail;
2228 }
2229 memcpy(osb->osb_cluster_name,
2230 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2231 OCFS2_CLUSTER_NAME_LEN);
2232 } else {
2233 /* The empty string is identical with classic tools that
2234 * don't know about s_cluster_info. */
2235 osb->osb_cluster_stack[0] = '\0';
2236 }
2237
2238 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2239
2240 /* FIXME
2241 * This should be done in ocfs2_journal_init(), but unknown
2242 * ordering issues will cause the filesystem to crash.
2243 * If anyone wants to figure out what part of the code
2244 * refers to osb->journal before ocfs2_journal_init() is run,
2245 * be my guest.
2246 */
2247 /* initialize our journal structure */
2248
2249 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2250 if (!journal) {
2251 mlog(ML_ERROR, "unable to alloc journal\n");
2252 status = -ENOMEM;
2253 goto bail;
2254 }
2255 osb->journal = journal;
2256 journal->j_osb = osb;
2257
2258 atomic_set(&journal->j_num_trans, 0);
2259 init_rwsem(&journal->j_trans_barrier);
2260 init_waitqueue_head(&journal->j_checkpointed);
2261 spin_lock_init(&journal->j_lock);
2262 journal->j_trans_id = (unsigned long) 1;
2263 INIT_LIST_HEAD(&journal->j_la_cleanups);
2264 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2265 journal->j_state = OCFS2_JOURNAL_FREE;
2266
2267 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2268 init_llist_head(&osb->dquot_drop_list);
2269
2270 /* get some pseudo constants for clustersize bits */
2271 osb->s_clustersize_bits =
2272 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2273 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2274
2275 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2276 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2277 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2278 osb->s_clustersize);
2279 status = -EINVAL;
2280 goto bail;
2281 }
2282
2283 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2284 le32_to_cpu(di->i_clusters));
2285
2286 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2287 total_blocks);
2288 if (status) {
2289 mlog(ML_ERROR, "Volume too large "
2290 "to mount safely on this system");
2291 status = -EFBIG;
2292 goto bail;
2293 }
2294
2295 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2296 sizeof(di->id2.i_super.s_uuid))) {
2297 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2298 status = -ENOMEM;
2299 goto bail;
2300 }
2301
2302 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2303 OCFS2_MAX_VOL_LABEL_LEN);
2304 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2305 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2306 osb->first_cluster_group_blkno =
2307 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2308 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2309 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2310 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2311 (unsigned long long)osb->root_blkno,
2312 (unsigned long long)osb->system_dir_blkno,
2313 osb->s_clustersize_bits);
2314
2315 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2316 if (!osb->osb_dlm_debug) {
2317 status = -ENOMEM;
2318 mlog_errno(status);
2319 goto bail;
2320 }
2321
2322 atomic_set(&osb->vol_state, VOLUME_INIT);
2323
2324 /* load root, system_dir, and all global system inodes */
2325 status = ocfs2_init_global_system_inodes(osb);
2326 if (status < 0) {
2327 mlog_errno(status);
2328 goto bail;
2329 }
2330
2331 /*
2332 * global bitmap
2333 */
2334 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2335 OCFS2_INVALID_SLOT);
2336 if (!inode) {
2337 status = -EINVAL;
2338 mlog_errno(status);
2339 goto bail;
2340 }
2341
2342 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2343 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2344 iput(inode);
2345
2346 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2347 osb->s_feature_incompat) * 8;
2348
2349 status = ocfs2_init_slot_info(osb);
2350 if (status < 0) {
2351 mlog_errno(status);
2352 goto bail;
2353 }
2354 cleancache_init_shared_fs(sb);
2355
2356 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2357 if (!osb->ocfs2_wq) {
2358 status = -ENOMEM;
2359 mlog_errno(status);
2360 }
2361
2362 bail:
2363 return status;
2364 }
2365
2366 /*
2367 * will return: -EAGAIN if it is ok to keep searching for superblocks
2368 * -EINVAL if there is a bad superblock
2369 * 0 on success
2370 */
ocfs2_verify_volume(struct ocfs2_dinode * di,struct buffer_head * bh,u32 blksz,struct ocfs2_blockcheck_stats * stats)2371 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2372 struct buffer_head *bh,
2373 u32 blksz,
2374 struct ocfs2_blockcheck_stats *stats)
2375 {
2376 int status = -EAGAIN;
2377
2378 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2379 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2380 /* We have to do a raw check of the feature here */
2381 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2382 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2383 status = ocfs2_block_check_validate(bh->b_data,
2384 bh->b_size,
2385 &di->i_check,
2386 stats);
2387 if (status)
2388 goto out;
2389 }
2390 status = -EINVAL;
2391 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2392 mlog(ML_ERROR, "found superblock with incorrect block "
2393 "size: found %u, should be %u\n",
2394 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2395 blksz);
2396 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2397 OCFS2_MAJOR_REV_LEVEL ||
2398 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2399 OCFS2_MINOR_REV_LEVEL) {
2400 mlog(ML_ERROR, "found superblock with bad version: "
2401 "found %u.%u, should be %u.%u\n",
2402 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2403 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2404 OCFS2_MAJOR_REV_LEVEL,
2405 OCFS2_MINOR_REV_LEVEL);
2406 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2407 mlog(ML_ERROR, "bad block number on superblock: "
2408 "found %llu, should be %llu\n",
2409 (unsigned long long)le64_to_cpu(di->i_blkno),
2410 (unsigned long long)bh->b_blocknr);
2411 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2412 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2413 mlog(ML_ERROR, "bad cluster size found: %u\n",
2414 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2415 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2416 mlog(ML_ERROR, "bad root_blkno: 0\n");
2417 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2418 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2419 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2420 mlog(ML_ERROR,
2421 "Superblock slots found greater than file system "
2422 "maximum: found %u, max %u\n",
2423 le16_to_cpu(di->id2.i_super.s_max_slots),
2424 OCFS2_MAX_SLOTS);
2425 } else {
2426 /* found it! */
2427 status = 0;
2428 }
2429 }
2430
2431 out:
2432 if (status && status != -EAGAIN)
2433 mlog_errno(status);
2434 return status;
2435 }
2436
ocfs2_check_volume(struct ocfs2_super * osb)2437 static int ocfs2_check_volume(struct ocfs2_super *osb)
2438 {
2439 int status;
2440 int dirty;
2441 int local;
2442 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2443 * recover
2444 * ourselves. */
2445
2446 /* Init our journal object. */
2447 status = ocfs2_journal_init(osb->journal, &dirty);
2448 if (status < 0) {
2449 mlog(ML_ERROR, "Could not initialize journal!\n");
2450 goto finally;
2451 }
2452
2453 /* Now that journal has been initialized, check to make sure
2454 entire volume is addressable. */
2455 status = ocfs2_journal_addressable(osb);
2456 if (status)
2457 goto finally;
2458
2459 /* If the journal was unmounted cleanly then we don't want to
2460 * recover anything. Otherwise, journal_load will do that
2461 * dirty work for us :) */
2462 if (!dirty) {
2463 status = ocfs2_journal_wipe(osb->journal, 0);
2464 if (status < 0) {
2465 mlog_errno(status);
2466 goto finally;
2467 }
2468 } else {
2469 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2470 "unmounted cleanly, recovering it.\n", osb->dev_str);
2471 }
2472
2473 local = ocfs2_mount_local(osb);
2474
2475 /* will play back anything left in the journal. */
2476 status = ocfs2_journal_load(osb->journal, local, dirty);
2477 if (status < 0) {
2478 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2479 goto finally;
2480 }
2481
2482 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2483 jbd2_journal_set_features(osb->journal->j_journal,
2484 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2485 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2486 else
2487 jbd2_journal_clear_features(osb->journal->j_journal,
2488 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2489 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2490
2491 if (dirty) {
2492 /* recover my local alloc if we didn't unmount cleanly. */
2493 status = ocfs2_begin_local_alloc_recovery(osb,
2494 osb->slot_num,
2495 &local_alloc);
2496 if (status < 0) {
2497 mlog_errno(status);
2498 goto finally;
2499 }
2500 /* we complete the recovery process after we've marked
2501 * ourselves as mounted. */
2502 }
2503
2504 status = ocfs2_load_local_alloc(osb);
2505 if (status < 0) {
2506 mlog_errno(status);
2507 goto finally;
2508 }
2509
2510 if (dirty) {
2511 /* Recovery will be completed after we've mounted the
2512 * rest of the volume. */
2513 osb->local_alloc_copy = local_alloc;
2514 local_alloc = NULL;
2515 }
2516
2517 /* go through each journal, trylock it and if you get the
2518 * lock, and it's marked as dirty, set the bit in the recover
2519 * map and launch a recovery thread for it. */
2520 status = ocfs2_mark_dead_nodes(osb);
2521 if (status < 0) {
2522 mlog_errno(status);
2523 goto finally;
2524 }
2525
2526 status = ocfs2_compute_replay_slots(osb);
2527 if (status < 0)
2528 mlog_errno(status);
2529
2530 finally:
2531 kfree(local_alloc);
2532
2533 if (status)
2534 mlog_errno(status);
2535 return status;
2536 }
2537
2538 /*
2539 * The routine gets called from dismount or close whenever a dismount on
2540 * volume is requested and the osb open count becomes 1.
2541 * It will remove the osb from the global list and also free up all the
2542 * initialized resources and fileobject.
2543 */
ocfs2_delete_osb(struct ocfs2_super * osb)2544 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2545 {
2546 /* This function assumes that the caller has the main osb resource */
2547
2548 /* ocfs2_initializer_super have already created this workqueue */
2549 if (osb->ocfs2_wq)
2550 destroy_workqueue(osb->ocfs2_wq);
2551
2552 ocfs2_free_slot_info(osb);
2553
2554 kfree(osb->osb_orphan_wipes);
2555 kfree(osb->slot_recovery_generations);
2556 /* FIXME
2557 * This belongs in journal shutdown, but because we have to
2558 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2559 * we free it here.
2560 */
2561 kfree(osb->journal);
2562 kfree(osb->local_alloc_copy);
2563 kfree(osb->uuid_str);
2564 kfree(osb->vol_label);
2565 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2566 memset(osb, 0, sizeof(struct ocfs2_super));
2567 }
2568
2569 /* Depending on the mount option passed, perform one of the following:
2570 * Put OCFS2 into a readonly state (default)
2571 * Return EIO so that only the process errs
2572 * Fix the error as if fsck.ocfs2 -y
2573 * panic
2574 */
ocfs2_handle_error(struct super_block * sb)2575 static int ocfs2_handle_error(struct super_block *sb)
2576 {
2577 struct ocfs2_super *osb = OCFS2_SB(sb);
2578 int rv = 0;
2579
2580 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2581 pr_crit("On-disk corruption discovered. "
2582 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2583
2584 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2585 panic("OCFS2: (device %s): panic forced after error\n",
2586 sb->s_id);
2587 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2588 pr_crit("OCFS2: Returning error to the calling process.\n");
2589 rv = -EIO;
2590 } else { /* default option */
2591 rv = -EROFS;
2592 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2593 return rv;
2594
2595 pr_crit("OCFS2: File system is now read-only.\n");
2596 sb->s_flags |= SB_RDONLY;
2597 ocfs2_set_ro_flag(osb, 0);
2598 }
2599
2600 return rv;
2601 }
2602
__ocfs2_error(struct super_block * sb,const char * function,const char * fmt,...)2603 int __ocfs2_error(struct super_block *sb, const char *function,
2604 const char *fmt, ...)
2605 {
2606 struct va_format vaf;
2607 va_list args;
2608
2609 va_start(args, fmt);
2610 vaf.fmt = fmt;
2611 vaf.va = &args;
2612
2613 /* Not using mlog here because we want to show the actual
2614 * function the error came from. */
2615 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2616 sb->s_id, function, &vaf);
2617
2618 va_end(args);
2619
2620 return ocfs2_handle_error(sb);
2621 }
2622
2623 /* Handle critical errors. This is intentionally more drastic than
2624 * ocfs2_handle_error, so we only use for things like journal errors,
2625 * etc. */
__ocfs2_abort(struct super_block * sb,const char * function,const char * fmt,...)2626 void __ocfs2_abort(struct super_block *sb, const char *function,
2627 const char *fmt, ...)
2628 {
2629 struct va_format vaf;
2630 va_list args;
2631
2632 va_start(args, fmt);
2633
2634 vaf.fmt = fmt;
2635 vaf.va = &args;
2636
2637 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2638 sb->s_id, function, &vaf);
2639
2640 va_end(args);
2641
2642 /* We don't have the cluster support yet to go straight to
2643 * hard readonly in here. Until then, we want to keep
2644 * ocfs2_abort() so that we can at least mark critical
2645 * errors.
2646 *
2647 * TODO: This should abort the journal and alert other nodes
2648 * that our slot needs recovery. */
2649
2650 /* Force a panic(). This stinks, but it's better than letting
2651 * things continue without having a proper hard readonly
2652 * here. */
2653 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2654 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2655 ocfs2_handle_error(sb);
2656 }
2657
2658 /*
2659 * Void signal blockers, because in-kernel sigprocmask() only fails
2660 * when SIG_* is wrong.
2661 */
ocfs2_block_signals(sigset_t * oldset)2662 void ocfs2_block_signals(sigset_t *oldset)
2663 {
2664 int rc;
2665 sigset_t blocked;
2666
2667 sigfillset(&blocked);
2668 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2669 BUG_ON(rc);
2670 }
2671
ocfs2_unblock_signals(sigset_t * oldset)2672 void ocfs2_unblock_signals(sigset_t *oldset)
2673 {
2674 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2675 BUG_ON(rc);
2676 }
2677
2678 module_init(ocfs2_init);
2679 module_exit(ocfs2_exit);
2680