1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
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
46#define CREATE_TRACE_POINTS
47#include "ocfs2_trace.h"
48
49#include <cluster/masklog.h>
50
51#include "ocfs2.h"
52
53
54#include "ocfs1_fs_compat.h"
55
56#include "alloc.h"
57#include "aops.h"
58#include "blockcheck.h"
59#include "dlmglue.h"
60#include "export.h"
61#include "extent_map.h"
62#include "heartbeat.h"
63#include "inode.h"
64#include "journal.h"
65#include "localalloc.h"
66#include "namei.h"
67#include "slot_map.h"
68#include "super.h"
69#include "sysfile.h"
70#include "uptodate.h"
71#include "ver.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
79static struct kmem_cache *ocfs2_inode_cachep = NULL;
80struct kmem_cache *ocfs2_dquot_cachep;
81struct kmem_cache *ocfs2_qf_chunk_cachep;
82
83
84
85
86
87struct workqueue_struct *ocfs2_wq = NULL;
88
89static struct dentry *ocfs2_debugfs_root = NULL;
90
91MODULE_AUTHOR("Oracle");
92MODULE_LICENSE("GPL");
93
94struct mount_options
95{
96 unsigned long commit_interval;
97 unsigned long mount_opt;
98 unsigned int atime_quantum;
99 signed short slot;
100 int localalloc_opt;
101 unsigned int resv_level;
102 int dir_resv_level;
103 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
104};
105
106static int ocfs2_parse_options(struct super_block *sb, char *options,
107 struct mount_options *mopt,
108 int is_remount);
109static int ocfs2_check_set_options(struct super_block *sb,
110 struct mount_options *options);
111static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
112static void ocfs2_put_super(struct super_block *sb);
113static int ocfs2_mount_volume(struct super_block *sb);
114static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116static int ocfs2_initialize_mem_caches(void);
117static void ocfs2_free_mem_caches(void);
118static void ocfs2_delete_osb(struct ocfs2_super *osb);
119
120static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
121
122static int ocfs2_sync_fs(struct super_block *sb, int wait);
123
124static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
126static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
127static int ocfs2_check_volume(struct ocfs2_super *osb);
128static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 struct buffer_head *bh,
130 u32 sectsize,
131 struct ocfs2_blockcheck_stats *stats);
132static int ocfs2_initialize_super(struct super_block *sb,
133 struct buffer_head *bh,
134 int sector_size,
135 struct ocfs2_blockcheck_stats *stats);
136static int ocfs2_get_sector(struct super_block *sb,
137 struct buffer_head **bh,
138 int block,
139 int sect_size);
140static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141static void ocfs2_destroy_inode(struct inode *inode);
142static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144static void ocfs2_disable_quotas(struct ocfs2_super *osb);
145
146static const struct super_operations ocfs2_sops = {
147 .statfs = ocfs2_statfs,
148 .alloc_inode = ocfs2_alloc_inode,
149 .destroy_inode = ocfs2_destroy_inode,
150 .drop_inode = ocfs2_drop_inode,
151 .evict_inode = ocfs2_evict_inode,
152 .sync_fs = ocfs2_sync_fs,
153 .put_super = ocfs2_put_super,
154 .remount_fs = ocfs2_remount,
155 .show_options = ocfs2_show_options,
156 .quota_read = ocfs2_quota_read,
157 .quota_write = ocfs2_quota_write,
158};
159
160enum {
161 Opt_barrier,
162 Opt_err_panic,
163 Opt_err_ro,
164 Opt_intr,
165 Opt_nointr,
166 Opt_hb_none,
167 Opt_hb_local,
168 Opt_hb_global,
169 Opt_data_ordered,
170 Opt_data_writeback,
171 Opt_atime_quantum,
172 Opt_slot,
173 Opt_commit,
174 Opt_localalloc,
175 Opt_localflocks,
176 Opt_stack,
177 Opt_user_xattr,
178 Opt_nouser_xattr,
179 Opt_inode64,
180 Opt_acl,
181 Opt_noacl,
182 Opt_usrquota,
183 Opt_grpquota,
184 Opt_coherency_buffered,
185 Opt_coherency_full,
186 Opt_resv_level,
187 Opt_dir_resv_level,
188 Opt_err,
189};
190
191static const match_table_t tokens = {
192 {Opt_barrier, "barrier=%u"},
193 {Opt_err_panic, "errors=panic"},
194 {Opt_err_ro, "errors=remount-ro"},
195 {Opt_intr, "intr"},
196 {Opt_nointr, "nointr"},
197 {Opt_hb_none, OCFS2_HB_NONE},
198 {Opt_hb_local, OCFS2_HB_LOCAL},
199 {Opt_hb_global, OCFS2_HB_GLOBAL},
200 {Opt_data_ordered, "data=ordered"},
201 {Opt_data_writeback, "data=writeback"},
202 {Opt_atime_quantum, "atime_quantum=%u"},
203 {Opt_slot, "preferred_slot=%u"},
204 {Opt_commit, "commit=%u"},
205 {Opt_localalloc, "localalloc=%d"},
206 {Opt_localflocks, "localflocks"},
207 {Opt_stack, "cluster_stack=%s"},
208 {Opt_user_xattr, "user_xattr"},
209 {Opt_nouser_xattr, "nouser_xattr"},
210 {Opt_inode64, "inode64"},
211 {Opt_acl, "acl"},
212 {Opt_noacl, "noacl"},
213 {Opt_usrquota, "usrquota"},
214 {Opt_grpquota, "grpquota"},
215 {Opt_coherency_buffered, "coherency=buffered"},
216 {Opt_coherency_full, "coherency=full"},
217 {Opt_resv_level, "resv_level=%u"},
218 {Opt_dir_resv_level, "dir_resv_level=%u"},
219 {Opt_err, NULL}
220};
221
222#ifdef CONFIG_DEBUG_FS
223static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
224{
225 struct ocfs2_cluster_connection *cconn = osb->cconn;
226 struct ocfs2_recovery_map *rm = osb->recovery_map;
227 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
228 int i, out = 0;
229
230 out += snprintf(buf + out, len - out,
231 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
232 "Device", osb->dev_str, osb->uuid_str,
233 osb->fs_generation, osb->vol_label);
234
235 out += snprintf(buf + out, len - out,
236 "%10s => State: %d Flags: 0x%lX\n", "Volume",
237 atomic_read(&osb->vol_state), osb->osb_flags);
238
239 out += snprintf(buf + out, len - out,
240 "%10s => Block: %lu Cluster: %d\n", "Sizes",
241 osb->sb->s_blocksize, osb->s_clustersize);
242
243 out += snprintf(buf + out, len - out,
244 "%10s => Compat: 0x%X Incompat: 0x%X "
245 "ROcompat: 0x%X\n",
246 "Features", osb->s_feature_compat,
247 osb->s_feature_incompat, osb->s_feature_ro_compat);
248
249 out += snprintf(buf + out, len - out,
250 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
251 osb->s_mount_opt, osb->s_atime_quantum);
252
253 if (cconn) {
254 out += snprintf(buf + out, len - out,
255 "%10s => Stack: %s Name: %*s "
256 "Version: %d.%d\n", "Cluster",
257 (*osb->osb_cluster_stack == '\0' ?
258 "o2cb" : osb->osb_cluster_stack),
259 cconn->cc_namelen, cconn->cc_name,
260 cconn->cc_version.pv_major,
261 cconn->cc_version.pv_minor);
262 }
263
264 spin_lock(&osb->dc_task_lock);
265 out += snprintf(buf + out, len - out,
266 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
267 "WorkSeq: %lu\n", "DownCnvt",
268 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
269 osb->blocked_lock_count, osb->dc_wake_sequence,
270 osb->dc_work_sequence);
271 spin_unlock(&osb->dc_task_lock);
272
273 spin_lock(&osb->osb_lock);
274 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
275 "Recovery",
276 (osb->recovery_thread_task ?
277 task_pid_nr(osb->recovery_thread_task) : -1));
278 if (rm->rm_used == 0)
279 out += snprintf(buf + out, len - out, " None\n");
280 else {
281 for (i = 0; i < rm->rm_used; i++)
282 out += snprintf(buf + out, len - out, " %d",
283 rm->rm_entries[i]);
284 out += snprintf(buf + out, len - out, "\n");
285 }
286 spin_unlock(&osb->osb_lock);
287
288 out += snprintf(buf + out, len - out,
289 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
290 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 osb->osb_commit_interval,
292 atomic_read(&osb->needs_checkpoint));
293
294 out += snprintf(buf + out, len - out,
295 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
296 "Journal", osb->journal->j_state,
297 osb->journal->j_trans_id,
298 atomic_read(&osb->journal->j_num_trans));
299
300 out += snprintf(buf + out, len - out,
301 "%10s => GlobalAllocs: %d LocalAllocs: %d "
302 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
303 "Stats",
304 atomic_read(&osb->alloc_stats.bitmap_data),
305 atomic_read(&osb->alloc_stats.local_data),
306 atomic_read(&osb->alloc_stats.bg_allocs),
307 atomic_read(&osb->alloc_stats.moves),
308 atomic_read(&osb->alloc_stats.bg_extends));
309
310 out += snprintf(buf + out, len - out,
311 "%10s => State: %u Descriptor: %llu Size: %u bits "
312 "Default: %u bits\n",
313 "LocalAlloc", osb->local_alloc_state,
314 (unsigned long long)osb->la_last_gd,
315 osb->local_alloc_bits, osb->local_alloc_default_bits);
316
317 spin_lock(&osb->osb_lock);
318 out += snprintf(buf + out, len - out,
319 "%10s => InodeSlot: %d StolenInodes: %d, "
320 "MetaSlot: %d StolenMeta: %d\n", "Steal",
321 osb->s_inode_steal_slot,
322 atomic_read(&osb->s_num_inodes_stolen),
323 osb->s_meta_steal_slot,
324 atomic_read(&osb->s_num_meta_stolen));
325 spin_unlock(&osb->osb_lock);
326
327 out += snprintf(buf + out, len - out, "OrphanScan => ");
328 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
329 os->os_count, os->os_seqno);
330 out += snprintf(buf + out, len - out, " Last Scan: ");
331 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
332 out += snprintf(buf + out, len - out, "Disabled\n");
333 else
334 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
335 (get_seconds() - os->os_scantime.tv_sec));
336
337 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
338 "Slots", "Num", "RecoGen");
339 for (i = 0; i < osb->max_slots; ++i) {
340 out += snprintf(buf + out, len - out,
341 "%10s %c %3d %10d\n",
342 " ",
343 (i == osb->slot_num ? '*' : ' '),
344 i, osb->slot_recovery_generations[i]);
345 }
346
347 return out;
348}
349
350static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
351{
352 struct ocfs2_super *osb = inode->i_private;
353 char *buf = NULL;
354
355 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
356 if (!buf)
357 goto bail;
358
359 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
360
361 file->private_data = buf;
362
363 return 0;
364bail:
365 return -ENOMEM;
366}
367
368static int ocfs2_debug_release(struct inode *inode, struct file *file)
369{
370 kfree(file->private_data);
371 return 0;
372}
373
374static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
375 size_t nbytes, loff_t *ppos)
376{
377 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
378 i_size_read(file->f_mapping->host));
379}
380#else
381static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
382{
383 return 0;
384}
385static int ocfs2_debug_release(struct inode *inode, struct file *file)
386{
387 return 0;
388}
389static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
390 size_t nbytes, loff_t *ppos)
391{
392 return 0;
393}
394#endif
395
396static const struct file_operations ocfs2_osb_debug_fops = {
397 .open = ocfs2_osb_debug_open,
398 .release = ocfs2_debug_release,
399 .read = ocfs2_debug_read,
400 .llseek = generic_file_llseek,
401};
402
403static int ocfs2_sync_fs(struct super_block *sb, int wait)
404{
405 int status;
406 tid_t target;
407 struct ocfs2_super *osb = OCFS2_SB(sb);
408
409 if (ocfs2_is_hard_readonly(osb))
410 return -EROFS;
411
412 if (wait) {
413 status = ocfs2_flush_truncate_log(osb);
414 if (status < 0)
415 mlog_errno(status);
416 } else {
417 ocfs2_schedule_truncate_log_flush(osb, 0);
418 }
419
420 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
421 &target)) {
422 if (wait)
423 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
424 target);
425 }
426 return 0;
427}
428
429static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
430{
431 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
432 && (ino == USER_QUOTA_SYSTEM_INODE
433 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
434 return 0;
435 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
436 && (ino == GROUP_QUOTA_SYSTEM_INODE
437 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
438 return 0;
439 return 1;
440}
441
442static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
443{
444 struct inode *new = NULL;
445 int status = 0;
446 int i;
447
448 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 if (IS_ERR(new)) {
450 status = PTR_ERR(new);
451 mlog_errno(status);
452 goto bail;
453 }
454 osb->root_inode = new;
455
456 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
457 if (IS_ERR(new)) {
458 status = PTR_ERR(new);
459 mlog_errno(status);
460 goto bail;
461 }
462 osb->sys_root_inode = new;
463
464 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
465 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
466 if (!ocfs2_need_system_inode(osb, i))
467 continue;
468 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
469 if (!new) {
470 ocfs2_release_system_inodes(osb);
471 status = -EINVAL;
472 mlog_errno(status);
473
474 mlog(ML_ERROR, "Unable to load system inode %d, "
475 "possibly corrupt fs?", i);
476 goto bail;
477 }
478
479 iput(new);
480 }
481
482bail:
483 if (status)
484 mlog_errno(status);
485 return status;
486}
487
488static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
489{
490 struct inode *new = NULL;
491 int status = 0;
492 int i;
493
494 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
495 i < NUM_SYSTEM_INODES;
496 i++) {
497 if (!ocfs2_need_system_inode(osb, i))
498 continue;
499 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
500 if (!new) {
501 ocfs2_release_system_inodes(osb);
502 status = -EINVAL;
503 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
504 status, i, osb->slot_num);
505 goto bail;
506 }
507
508 iput(new);
509 }
510
511bail:
512 if (status)
513 mlog_errno(status);
514 return status;
515}
516
517static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
518{
519 int i;
520 struct inode *inode;
521
522 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
523 inode = osb->global_system_inodes[i];
524 if (inode) {
525 iput(inode);
526 osb->global_system_inodes[i] = NULL;
527 }
528 }
529
530 inode = osb->sys_root_inode;
531 if (inode) {
532 iput(inode);
533 osb->sys_root_inode = NULL;
534 }
535
536 inode = osb->root_inode;
537 if (inode) {
538 iput(inode);
539 osb->root_inode = NULL;
540 }
541
542 if (!osb->local_system_inodes)
543 return;
544
545 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
546 if (osb->local_system_inodes[i]) {
547 iput(osb->local_system_inodes[i]);
548 osb->local_system_inodes[i] = NULL;
549 }
550 }
551
552 kfree(osb->local_system_inodes);
553 osb->local_system_inodes = NULL;
554}
555
556
557static struct inode *ocfs2_alloc_inode(struct super_block *sb)
558{
559 struct ocfs2_inode_info *oi;
560
561 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
562 if (!oi)
563 return NULL;
564
565 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
566 return &oi->vfs_inode;
567}
568
569static void ocfs2_i_callback(struct rcu_head *head)
570{
571 struct inode *inode = container_of(head, struct inode, i_rcu);
572 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
573}
574
575static void ocfs2_destroy_inode(struct inode *inode)
576{
577 call_rcu(&inode->i_rcu, ocfs2_i_callback);
578}
579
580static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
581 unsigned int cbits)
582{
583 unsigned int bytes = 1 << cbits;
584 unsigned int trim = bytes;
585 unsigned int bitshift = 32;
586
587
588
589
590
591
592
593#if BITS_PER_LONG == 32
594# if defined(CONFIG_LBDAF)
595 BUILD_BUG_ON(sizeof(sector_t) != 8);
596
597
598
599 if (bytes > PAGE_CACHE_SIZE) {
600 bytes = PAGE_CACHE_SIZE;
601 trim = 1;
602
603
604
605
606 bitshift = 31;
607 }
608# else
609
610
611
612
613 bytes = 1 << bbits;
614 trim = 1;
615 bitshift = 31;
616# endif
617#endif
618
619
620
621
622
623
624 return (((unsigned long long)bytes) << bitshift) - trim;
625}
626
627static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
628{
629 int incompat_features;
630 int ret = 0;
631 struct mount_options parsed_options;
632 struct ocfs2_super *osb = OCFS2_SB(sb);
633 u32 tmp;
634
635 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
636 !ocfs2_check_set_options(sb, &parsed_options)) {
637 ret = -EINVAL;
638 goto out;
639 }
640
641 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
642 OCFS2_MOUNT_HB_NONE;
643 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
644 ret = -EINVAL;
645 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
646 goto out;
647 }
648
649 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
650 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
651 ret = -EINVAL;
652 mlog(ML_ERROR, "Cannot change data mode on remount\n");
653 goto out;
654 }
655
656
657
658 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
659 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
660 ret = -EINVAL;
661 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
662 goto out;
663 }
664
665
666 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
667
668 if (*flags & MS_RDONLY) {
669 ret = ocfs2_susp_quotas(osb, 0);
670 if (ret < 0)
671 goto out;
672 }
673
674
675 spin_lock(&osb->osb_lock);
676 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
677 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
678 ret = -EROFS;
679 goto unlock_osb;
680 }
681
682 if (*flags & MS_RDONLY) {
683 sb->s_flags |= MS_RDONLY;
684 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
685 } else {
686 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
687 mlog(ML_ERROR, "Cannot remount RDWR "
688 "filesystem due to previous errors.\n");
689 ret = -EROFS;
690 goto unlock_osb;
691 }
692 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
693 if (incompat_features) {
694 mlog(ML_ERROR, "Cannot remount RDWR because "
695 "of unsupported optional features "
696 "(%x).\n", incompat_features);
697 ret = -EINVAL;
698 goto unlock_osb;
699 }
700 sb->s_flags &= ~MS_RDONLY;
701 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
702 }
703 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
704unlock_osb:
705 spin_unlock(&osb->osb_lock);
706
707 if (!ret && !(*flags & MS_RDONLY)) {
708 if (sb_any_quota_suspended(sb))
709 ret = ocfs2_susp_quotas(osb, 1);
710 else
711 ret = ocfs2_enable_quotas(osb);
712 if (ret < 0) {
713
714 spin_lock(&osb->osb_lock);
715 sb->s_flags |= MS_RDONLY;
716 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
717 spin_unlock(&osb->osb_lock);
718 goto out;
719 }
720 }
721 }
722
723 if (!ret) {
724
725
726 osb->s_mount_opt = parsed_options.mount_opt;
727 osb->s_atime_quantum = parsed_options.atime_quantum;
728 osb->preferred_slot = parsed_options.slot;
729 if (parsed_options.commit_interval)
730 osb->osb_commit_interval = parsed_options.commit_interval;
731
732 if (!ocfs2_is_hard_readonly(osb))
733 ocfs2_set_journal_params(osb);
734
735 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
736 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
737 MS_POSIXACL : 0);
738 }
739out:
740 return ret;
741}
742
743static int ocfs2_sb_probe(struct super_block *sb,
744 struct buffer_head **bh,
745 int *sector_size,
746 struct ocfs2_blockcheck_stats *stats)
747{
748 int status, tmpstat;
749 struct ocfs1_vol_disk_hdr *hdr;
750 struct ocfs2_dinode *di;
751 int blksize;
752
753 *bh = NULL;
754
755
756 *sector_size = bdev_logical_block_size(sb->s_bdev);
757 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
758 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
759 *sector_size, OCFS2_MAX_BLOCKSIZE);
760 status = -EINVAL;
761 goto bail;
762 }
763
764
765 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
766 *sector_size = OCFS2_MIN_BLOCKSIZE;
767
768
769 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
770 if (status < 0) {
771 mlog_errno(status);
772 goto bail;
773 }
774 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
775 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
776 mlog(ML_ERROR, "incompatible version: %u.%u\n",
777 hdr->major_version, hdr->minor_version);
778 status = -EINVAL;
779 }
780 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
781 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
782 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
783 hdr->signature);
784 status = -EINVAL;
785 }
786 brelse(*bh);
787 *bh = NULL;
788 if (status < 0) {
789 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
790 "upgraded before mounting with ocfs v2\n");
791 goto bail;
792 }
793
794
795
796
797
798
799 status = -EINVAL;
800 for (blksize = *sector_size;
801 blksize <= OCFS2_MAX_BLOCKSIZE;
802 blksize <<= 1) {
803 tmpstat = ocfs2_get_sector(sb, bh,
804 OCFS2_SUPER_BLOCK_BLKNO,
805 blksize);
806 if (tmpstat < 0) {
807 status = tmpstat;
808 mlog_errno(status);
809 break;
810 }
811 di = (struct ocfs2_dinode *) (*bh)->b_data;
812 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
813 spin_lock_init(&stats->b_lock);
814 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
815 if (tmpstat < 0) {
816 brelse(*bh);
817 *bh = NULL;
818 }
819 if (tmpstat != -EAGAIN) {
820 status = tmpstat;
821 break;
822 }
823 }
824
825bail:
826 return status;
827}
828
829static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
830{
831 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
832
833 if (osb->s_mount_opt & hb_enabled) {
834 if (ocfs2_mount_local(osb)) {
835 mlog(ML_ERROR, "Cannot heartbeat on a locally "
836 "mounted device.\n");
837 return -EINVAL;
838 }
839 if (ocfs2_userspace_stack(osb)) {
840 mlog(ML_ERROR, "Userspace stack expected, but "
841 "o2cb heartbeat arguments passed to mount\n");
842 return -EINVAL;
843 }
844 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
845 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
846 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
847 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
848 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
849 return -EINVAL;
850 }
851 }
852
853 if (!(osb->s_mount_opt & hb_enabled)) {
854 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
855 !ocfs2_userspace_stack(osb)) {
856 mlog(ML_ERROR, "Heartbeat has to be started to mount "
857 "a read-write clustered device.\n");
858 return -EINVAL;
859 }
860 }
861
862 return 0;
863}
864
865
866
867
868
869
870static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
871 struct mount_options *mopt)
872{
873 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
874 mlog(ML_ERROR,
875 "cluster stack passed to mount, but this filesystem "
876 "does not support it\n");
877 return -EINVAL;
878 }
879
880 if (ocfs2_userspace_stack(osb) &&
881 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
882 OCFS2_STACK_LABEL_LEN)) {
883 mlog(ML_ERROR,
884 "cluster stack passed to mount (\"%s\") does not "
885 "match the filesystem (\"%s\")\n",
886 mopt->cluster_stack,
887 osb->osb_cluster_stack);
888 return -EINVAL;
889 }
890
891 return 0;
892}
893
894static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
895{
896 int type;
897 struct super_block *sb = osb->sb;
898 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
899 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
900 int status = 0;
901
902 for (type = 0; type < MAXQUOTAS; type++) {
903 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
904 continue;
905 if (unsuspend)
906 status = dquot_resume(sb, type);
907 else {
908 struct ocfs2_mem_dqinfo *oinfo;
909
910
911 oinfo = sb_dqinfo(sb, type)->dqi_priv;
912 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
913 status = dquot_suspend(sb, type);
914 }
915 if (status < 0)
916 break;
917 }
918 if (status < 0)
919 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
920 "remount (error = %d).\n", status);
921 return status;
922}
923
924static int ocfs2_enable_quotas(struct ocfs2_super *osb)
925{
926 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
927 struct super_block *sb = osb->sb;
928 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
929 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
930 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
931 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
932 int status;
933 int type;
934
935 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
936 for (type = 0; type < MAXQUOTAS; type++) {
937 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
938 continue;
939 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
940 osb->slot_num);
941 if (!inode[type]) {
942 status = -ENOENT;
943 goto out_quota_off;
944 }
945 status = dquot_enable(inode[type], type, QFMT_OCFS2,
946 DQUOT_USAGE_ENABLED);
947 if (status < 0)
948 goto out_quota_off;
949 }
950
951 for (type = 0; type < MAXQUOTAS; type++)
952 iput(inode[type]);
953 return 0;
954out_quota_off:
955 ocfs2_disable_quotas(osb);
956 for (type = 0; type < MAXQUOTAS; type++)
957 iput(inode[type]);
958 mlog_errno(status);
959 return status;
960}
961
962static void ocfs2_disable_quotas(struct ocfs2_super *osb)
963{
964 int type;
965 struct inode *inode;
966 struct super_block *sb = osb->sb;
967 struct ocfs2_mem_dqinfo *oinfo;
968
969
970
971 for (type = 0; type < MAXQUOTAS; type++) {
972 if (!sb_has_quota_loaded(sb, type))
973 continue;
974
975 oinfo = sb_dqinfo(sb, type)->dqi_priv;
976 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
977 inode = igrab(sb->s_dquot.files[type]);
978
979
980
981 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
982 DQUOT_LIMITS_ENABLED);
983 if (!inode)
984 continue;
985 iput(inode);
986 }
987}
988
989
990static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
991{
992 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
993 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
994
995 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
996 return -EINVAL;
997
998 return dquot_enable(sb_dqopt(sb)->files[type], type,
999 format_id, DQUOT_LIMITS_ENABLED);
1000}
1001
1002
1003static int ocfs2_quota_off(struct super_block *sb, int type)
1004{
1005 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1006}
1007
1008static const struct quotactl_ops ocfs2_quotactl_ops = {
1009 .quota_on_meta = ocfs2_quota_on,
1010 .quota_off = ocfs2_quota_off,
1011 .quota_sync = dquot_quota_sync,
1012 .get_info = dquot_get_dqinfo,
1013 .set_info = dquot_set_dqinfo,
1014 .get_dqblk = dquot_get_dqblk,
1015 .set_dqblk = dquot_set_dqblk,
1016};
1017
1018static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1019{
1020 struct dentry *root;
1021 int status, sector_size;
1022 struct mount_options parsed_options;
1023 struct inode *inode = NULL;
1024 struct ocfs2_super *osb = NULL;
1025 struct buffer_head *bh = NULL;
1026 char nodestr[8];
1027 struct ocfs2_blockcheck_stats stats;
1028
1029 trace_ocfs2_fill_super(sb, data, silent);
1030
1031 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1032 status = -EINVAL;
1033 goto read_super_error;
1034 }
1035
1036
1037 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1038 if (status < 0) {
1039 mlog(ML_ERROR, "superblock probe failed!\n");
1040 goto read_super_error;
1041 }
1042
1043 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1044 osb = OCFS2_SB(sb);
1045 if (status < 0) {
1046 mlog_errno(status);
1047 goto read_super_error;
1048 }
1049 brelse(bh);
1050 bh = NULL;
1051
1052 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1053 status = -EINVAL;
1054 goto read_super_error;
1055 }
1056 osb->s_mount_opt = parsed_options.mount_opt;
1057 osb->s_atime_quantum = parsed_options.atime_quantum;
1058 osb->preferred_slot = parsed_options.slot;
1059 osb->osb_commit_interval = parsed_options.commit_interval;
1060
1061 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1062 osb->osb_resv_level = parsed_options.resv_level;
1063 osb->osb_dir_resv_level = parsed_options.resv_level;
1064 if (parsed_options.dir_resv_level == -1)
1065 osb->osb_dir_resv_level = parsed_options.resv_level;
1066 else
1067 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1068
1069 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1070 if (status)
1071 goto read_super_error;
1072
1073 sb->s_magic = OCFS2_SUPER_MAGIC;
1074
1075 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1076 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1077
1078
1079
1080 if (bdev_read_only(sb->s_bdev)) {
1081 if (!(sb->s_flags & MS_RDONLY)) {
1082 status = -EACCES;
1083 mlog(ML_ERROR, "Readonly device detected but readonly "
1084 "mount was not specified.\n");
1085 goto read_super_error;
1086 }
1087
1088
1089
1090 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1091 status = -EROFS;
1092 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1093 "device.\n");
1094 goto read_super_error;
1095 }
1096
1097 status = ocfs2_check_journals_nolocks(osb);
1098 if (status < 0) {
1099 if (status == -EROFS)
1100 mlog(ML_ERROR, "Recovery required on readonly "
1101 "file system, but write access is "
1102 "unavailable.\n");
1103 else
1104 mlog_errno(status);
1105 goto read_super_error;
1106 }
1107
1108 ocfs2_set_ro_flag(osb, 1);
1109
1110 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1111 "Cluster services will not be used for this mount. "
1112 "Recovery will be skipped.\n", osb->dev_str);
1113 }
1114
1115 if (!ocfs2_is_hard_readonly(osb)) {
1116 if (sb->s_flags & MS_RDONLY)
1117 ocfs2_set_ro_flag(osb, 0);
1118 }
1119
1120 status = ocfs2_verify_heartbeat(osb);
1121 if (status < 0) {
1122 mlog_errno(status);
1123 goto read_super_error;
1124 }
1125
1126 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1127 ocfs2_debugfs_root);
1128 if (!osb->osb_debug_root) {
1129 status = -EINVAL;
1130 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1131 goto read_super_error;
1132 }
1133
1134 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1135 osb->osb_debug_root,
1136 osb,
1137 &ocfs2_osb_debug_fops);
1138 if (!osb->osb_ctxt) {
1139 status = -EINVAL;
1140 mlog_errno(status);
1141 goto read_super_error;
1142 }
1143
1144 if (ocfs2_meta_ecc(osb)) {
1145 status = ocfs2_blockcheck_stats_debugfs_install(
1146 &osb->osb_ecc_stats,
1147 osb->osb_debug_root);
1148 if (status) {
1149 mlog(ML_ERROR,
1150 "Unable to create blockcheck statistics "
1151 "files\n");
1152 goto read_super_error;
1153 }
1154 }
1155
1156 status = ocfs2_mount_volume(sb);
1157 if (status < 0)
1158 goto read_super_error;
1159
1160 if (osb->root_inode)
1161 inode = igrab(osb->root_inode);
1162
1163 if (!inode) {
1164 status = -EIO;
1165 mlog_errno(status);
1166 goto read_super_error;
1167 }
1168
1169 root = d_make_root(inode);
1170 if (!root) {
1171 status = -ENOMEM;
1172 mlog_errno(status);
1173 goto read_super_error;
1174 }
1175
1176 sb->s_root = root;
1177
1178 ocfs2_complete_mount_recovery(osb);
1179
1180 if (ocfs2_mount_local(osb))
1181 snprintf(nodestr, sizeof(nodestr), "local");
1182 else
1183 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1184
1185 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1186 "with %s data mode.\n",
1187 osb->dev_str, nodestr, osb->slot_num,
1188 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1189 "ordered");
1190
1191 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1192 wake_up(&osb->osb_mount_event);
1193
1194
1195
1196
1197 if (!(sb->s_flags & MS_RDONLY)) {
1198 status = ocfs2_enable_quotas(osb);
1199 if (status < 0) {
1200
1201
1202 mlog_errno(status);
1203 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1204 wake_up(&osb->osb_mount_event);
1205 return status;
1206 }
1207 }
1208
1209 ocfs2_complete_quota_recovery(osb);
1210
1211
1212 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1213 wake_up(&osb->osb_mount_event);
1214
1215
1216 ocfs2_orphan_scan_start(osb);
1217
1218 return status;
1219
1220read_super_error:
1221 brelse(bh);
1222
1223 if (osb) {
1224 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1225 wake_up(&osb->osb_mount_event);
1226 ocfs2_dismount_volume(sb, 1);
1227 }
1228
1229 if (status)
1230 mlog_errno(status);
1231 return status;
1232}
1233
1234static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1235 int flags,
1236 const char *dev_name,
1237 void *data)
1238{
1239 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1240}
1241
1242static void ocfs2_kill_sb(struct super_block *sb)
1243{
1244 struct ocfs2_super *osb = OCFS2_SB(sb);
1245
1246
1247 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1248 goto out;
1249
1250
1251 spin_lock(&dentry_list_lock);
1252 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1253 spin_unlock(&dentry_list_lock);
1254
1255 cancel_work_sync(&osb->dentry_lock_work);
1256out:
1257 kill_block_super(sb);
1258}
1259
1260static struct file_system_type ocfs2_fs_type = {
1261 .owner = THIS_MODULE,
1262 .name = "ocfs2",
1263 .mount = ocfs2_mount,
1264 .kill_sb = ocfs2_kill_sb,
1265
1266 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1267 .next = NULL
1268};
1269MODULE_ALIAS_FS("ocfs2");
1270
1271static int ocfs2_check_set_options(struct super_block *sb,
1272 struct mount_options *options)
1273{
1274 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1275 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1276 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1277 mlog(ML_ERROR, "User quotas were requested, but this "
1278 "filesystem does not have the feature enabled.\n");
1279 return 0;
1280 }
1281 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1282 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1283 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1284 mlog(ML_ERROR, "Group quotas were requested, but this "
1285 "filesystem does not have the feature enabled.\n");
1286 return 0;
1287 }
1288 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1289 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1290 mlog(ML_ERROR, "ACL support requested but extended attributes "
1291 "feature is not enabled\n");
1292 return 0;
1293 }
1294
1295 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1296 OCFS2_MOUNT_NO_POSIX_ACL))) {
1297 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1298 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1299 else
1300 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1301 }
1302 return 1;
1303}
1304
1305static int ocfs2_parse_options(struct super_block *sb,
1306 char *options,
1307 struct mount_options *mopt,
1308 int is_remount)
1309{
1310 int status, user_stack = 0;
1311 char *p;
1312 u32 tmp;
1313
1314 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1315
1316 mopt->commit_interval = 0;
1317 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1318 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1319 mopt->slot = OCFS2_INVALID_SLOT;
1320 mopt->localalloc_opt = -1;
1321 mopt->cluster_stack[0] = '\0';
1322 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1323 mopt->dir_resv_level = -1;
1324
1325 if (!options) {
1326 status = 1;
1327 goto bail;
1328 }
1329
1330 while ((p = strsep(&options, ",")) != NULL) {
1331 int token, option;
1332 substring_t args[MAX_OPT_ARGS];
1333
1334 if (!*p)
1335 continue;
1336
1337 token = match_token(p, tokens, args);
1338 switch (token) {
1339 case Opt_hb_local:
1340 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1341 break;
1342 case Opt_hb_none:
1343 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1344 break;
1345 case Opt_hb_global:
1346 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1347 break;
1348 case Opt_barrier:
1349 if (match_int(&args[0], &option)) {
1350 status = 0;
1351 goto bail;
1352 }
1353 if (option)
1354 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1355 else
1356 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1357 break;
1358 case Opt_intr:
1359 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1360 break;
1361 case Opt_nointr:
1362 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1363 break;
1364 case Opt_err_panic:
1365 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1366 break;
1367 case Opt_err_ro:
1368 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1369 break;
1370 case Opt_data_ordered:
1371 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1372 break;
1373 case Opt_data_writeback:
1374 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1375 break;
1376 case Opt_user_xattr:
1377 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1378 break;
1379 case Opt_nouser_xattr:
1380 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1381 break;
1382 case Opt_atime_quantum:
1383 if (match_int(&args[0], &option)) {
1384 status = 0;
1385 goto bail;
1386 }
1387 if (option >= 0)
1388 mopt->atime_quantum = option;
1389 break;
1390 case Opt_slot:
1391 option = 0;
1392 if (match_int(&args[0], &option)) {
1393 status = 0;
1394 goto bail;
1395 }
1396 if (option)
1397 mopt->slot = (s16)option;
1398 break;
1399 case Opt_commit:
1400 option = 0;
1401 if (match_int(&args[0], &option)) {
1402 status = 0;
1403 goto bail;
1404 }
1405 if (option < 0)
1406 return 0;
1407 if (option == 0)
1408 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1409 mopt->commit_interval = HZ * option;
1410 break;
1411 case Opt_localalloc:
1412 option = 0;
1413 if (match_int(&args[0], &option)) {
1414 status = 0;
1415 goto bail;
1416 }
1417 if (option >= 0)
1418 mopt->localalloc_opt = option;
1419 break;
1420 case Opt_localflocks:
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431 if (!is_remount)
1432 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1433 break;
1434 case Opt_stack:
1435
1436
1437
1438
1439 if (((args[0].to - args[0].from) !=
1440 OCFS2_STACK_LABEL_LEN) ||
1441 (strnlen(args[0].from,
1442 OCFS2_STACK_LABEL_LEN) !=
1443 OCFS2_STACK_LABEL_LEN)) {
1444 mlog(ML_ERROR,
1445 "Invalid cluster_stack option\n");
1446 status = 0;
1447 goto bail;
1448 }
1449 memcpy(mopt->cluster_stack, args[0].from,
1450 OCFS2_STACK_LABEL_LEN);
1451 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1452
1453
1454
1455
1456
1457 if (memcmp(mopt->cluster_stack,
1458 OCFS2_CLASSIC_CLUSTER_STACK,
1459 OCFS2_STACK_LABEL_LEN))
1460 user_stack = 1;
1461 break;
1462 case Opt_inode64:
1463 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1464 break;
1465 case Opt_usrquota:
1466 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1467 break;
1468 case Opt_grpquota:
1469 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1470 break;
1471 case Opt_coherency_buffered:
1472 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1473 break;
1474 case Opt_coherency_full:
1475 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1476 break;
1477 case Opt_acl:
1478 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1479 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1480 break;
1481 case Opt_noacl:
1482 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1483 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1484 break;
1485 case Opt_resv_level:
1486 if (is_remount)
1487 break;
1488 if (match_int(&args[0], &option)) {
1489 status = 0;
1490 goto bail;
1491 }
1492 if (option >= OCFS2_MIN_RESV_LEVEL &&
1493 option < OCFS2_MAX_RESV_LEVEL)
1494 mopt->resv_level = option;
1495 break;
1496 case Opt_dir_resv_level:
1497 if (is_remount)
1498 break;
1499 if (match_int(&args[0], &option)) {
1500 status = 0;
1501 goto bail;
1502 }
1503 if (option >= OCFS2_MIN_RESV_LEVEL &&
1504 option < OCFS2_MAX_RESV_LEVEL)
1505 mopt->dir_resv_level = option;
1506 break;
1507 default:
1508 mlog(ML_ERROR,
1509 "Unrecognized mount option \"%s\" "
1510 "or missing value\n", p);
1511 status = 0;
1512 goto bail;
1513 }
1514 }
1515
1516 if (user_stack == 0) {
1517
1518 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1519 OCFS2_MOUNT_HB_GLOBAL |
1520 OCFS2_MOUNT_HB_NONE);
1521 if (hweight32(tmp) != 1) {
1522 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1523 status = 0;
1524 goto bail;
1525 }
1526 }
1527
1528 status = 1;
1529
1530bail:
1531 return status;
1532}
1533
1534static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1535{
1536 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1537 unsigned long opts = osb->s_mount_opt;
1538 unsigned int local_alloc_megs;
1539
1540 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1541 seq_printf(s, ",_netdev");
1542 if (opts & OCFS2_MOUNT_HB_LOCAL)
1543 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1544 else
1545 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1546 } else
1547 seq_printf(s, ",%s", OCFS2_HB_NONE);
1548
1549 if (opts & OCFS2_MOUNT_NOINTR)
1550 seq_printf(s, ",nointr");
1551
1552 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1553 seq_printf(s, ",data=writeback");
1554 else
1555 seq_printf(s, ",data=ordered");
1556
1557 if (opts & OCFS2_MOUNT_BARRIER)
1558 seq_printf(s, ",barrier=1");
1559
1560 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1561 seq_printf(s, ",errors=panic");
1562 else
1563 seq_printf(s, ",errors=remount-ro");
1564
1565 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1566 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1567
1568 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1569
1570 if (osb->osb_commit_interval)
1571 seq_printf(s, ",commit=%u",
1572 (unsigned) (osb->osb_commit_interval / HZ));
1573
1574 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1575 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1576 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1577
1578 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1579 seq_printf(s, ",localflocks,");
1580
1581 if (osb->osb_cluster_stack[0])
1582 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1583 osb->osb_cluster_stack);
1584 if (opts & OCFS2_MOUNT_USRQUOTA)
1585 seq_printf(s, ",usrquota");
1586 if (opts & OCFS2_MOUNT_GRPQUOTA)
1587 seq_printf(s, ",grpquota");
1588
1589 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1590 seq_printf(s, ",coherency=buffered");
1591 else
1592 seq_printf(s, ",coherency=full");
1593
1594 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1595 seq_printf(s, ",nouser_xattr");
1596 else
1597 seq_printf(s, ",user_xattr");
1598
1599 if (opts & OCFS2_MOUNT_INODE64)
1600 seq_printf(s, ",inode64");
1601
1602 if (opts & OCFS2_MOUNT_POSIX_ACL)
1603 seq_printf(s, ",acl");
1604 else
1605 seq_printf(s, ",noacl");
1606
1607 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1608 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1609
1610 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1611 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1612
1613 return 0;
1614}
1615
1616wait_queue_head_t ocfs2__ioend_wq[OCFS2_IOEND_WQ_HASH_SZ];
1617
1618static int __init ocfs2_init(void)
1619{
1620 int status, i;
1621
1622 ocfs2_print_version();
1623
1624 for (i = 0; i < OCFS2_IOEND_WQ_HASH_SZ; i++)
1625 init_waitqueue_head(&ocfs2__ioend_wq[i]);
1626
1627 status = init_ocfs2_uptodate_cache();
1628 if (status < 0)
1629 goto out1;
1630
1631 status = ocfs2_initialize_mem_caches();
1632 if (status < 0)
1633 goto out2;
1634
1635 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1636 if (!ocfs2_wq) {
1637 status = -ENOMEM;
1638 goto out3;
1639 }
1640
1641 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1642 if (!ocfs2_debugfs_root) {
1643 status = -EFAULT;
1644 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1645 }
1646
1647 ocfs2_set_locking_protocol();
1648
1649 status = register_quota_format(&ocfs2_quota_format);
1650 if (status < 0)
1651 goto out4;
1652 status = register_filesystem(&ocfs2_fs_type);
1653 if (!status)
1654 return 0;
1655
1656 unregister_quota_format(&ocfs2_quota_format);
1657out4:
1658 destroy_workqueue(ocfs2_wq);
1659 debugfs_remove(ocfs2_debugfs_root);
1660out3:
1661 ocfs2_free_mem_caches();
1662out2:
1663 exit_ocfs2_uptodate_cache();
1664out1:
1665 mlog_errno(status);
1666 return status;
1667}
1668
1669static void __exit ocfs2_exit(void)
1670{
1671 if (ocfs2_wq) {
1672 flush_workqueue(ocfs2_wq);
1673 destroy_workqueue(ocfs2_wq);
1674 }
1675
1676 unregister_quota_format(&ocfs2_quota_format);
1677
1678 debugfs_remove(ocfs2_debugfs_root);
1679
1680 ocfs2_free_mem_caches();
1681
1682 unregister_filesystem(&ocfs2_fs_type);
1683
1684 exit_ocfs2_uptodate_cache();
1685}
1686
1687static void ocfs2_put_super(struct super_block *sb)
1688{
1689 trace_ocfs2_put_super(sb);
1690
1691 ocfs2_sync_blockdev(sb);
1692 ocfs2_dismount_volume(sb, 0);
1693}
1694
1695static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1696{
1697 struct ocfs2_super *osb;
1698 u32 numbits, freebits;
1699 int status;
1700 struct ocfs2_dinode *bm_lock;
1701 struct buffer_head *bh = NULL;
1702 struct inode *inode = NULL;
1703
1704 trace_ocfs2_statfs(dentry->d_sb, buf);
1705
1706 osb = OCFS2_SB(dentry->d_sb);
1707
1708 inode = ocfs2_get_system_file_inode(osb,
1709 GLOBAL_BITMAP_SYSTEM_INODE,
1710 OCFS2_INVALID_SLOT);
1711 if (!inode) {
1712 mlog(ML_ERROR, "failed to get bitmap inode\n");
1713 status = -EIO;
1714 goto bail;
1715 }
1716
1717 status = ocfs2_inode_lock(inode, &bh, 0);
1718 if (status < 0) {
1719 mlog_errno(status);
1720 goto bail;
1721 }
1722
1723 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1724
1725 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1726 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1727
1728 buf->f_type = OCFS2_SUPER_MAGIC;
1729 buf->f_bsize = dentry->d_sb->s_blocksize;
1730 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1731 buf->f_blocks = ((sector_t) numbits) *
1732 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1733 buf->f_bfree = ((sector_t) freebits) *
1734 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1735 buf->f_bavail = buf->f_bfree;
1736 buf->f_files = numbits;
1737 buf->f_ffree = freebits;
1738 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1739 & 0xFFFFFFFFUL;
1740 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1741 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1742
1743 brelse(bh);
1744
1745 ocfs2_inode_unlock(inode, 0);
1746 status = 0;
1747bail:
1748 if (inode)
1749 iput(inode);
1750
1751 if (status)
1752 mlog_errno(status);
1753
1754 return status;
1755}
1756
1757static void ocfs2_inode_init_once(void *data)
1758{
1759 struct ocfs2_inode_info *oi = data;
1760
1761 oi->ip_flags = 0;
1762 oi->ip_open_count = 0;
1763 spin_lock_init(&oi->ip_lock);
1764 ocfs2_extent_map_init(&oi->vfs_inode);
1765 INIT_LIST_HEAD(&oi->ip_io_markers);
1766 oi->ip_dir_start_lookup = 0;
1767 atomic_set(&oi->ip_unaligned_aio, 0);
1768 init_rwsem(&oi->ip_alloc_sem);
1769 init_rwsem(&oi->ip_xattr_sem);
1770 mutex_init(&oi->ip_io_mutex);
1771
1772 oi->ip_blkno = 0ULL;
1773 oi->ip_clusters = 0;
1774
1775 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1776
1777 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1778 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1779 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1780
1781 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1782 &ocfs2_inode_caching_ops);
1783
1784 inode_init_once(&oi->vfs_inode);
1785}
1786
1787static int ocfs2_initialize_mem_caches(void)
1788{
1789 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1790 sizeof(struct ocfs2_inode_info),
1791 0,
1792 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1793 SLAB_MEM_SPREAD),
1794 ocfs2_inode_init_once);
1795 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1796 sizeof(struct ocfs2_dquot),
1797 0,
1798 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1799 SLAB_MEM_SPREAD),
1800 NULL);
1801 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1802 sizeof(struct ocfs2_quota_chunk),
1803 0,
1804 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1805 NULL);
1806 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1807 !ocfs2_qf_chunk_cachep) {
1808 if (ocfs2_inode_cachep)
1809 kmem_cache_destroy(ocfs2_inode_cachep);
1810 if (ocfs2_dquot_cachep)
1811 kmem_cache_destroy(ocfs2_dquot_cachep);
1812 if (ocfs2_qf_chunk_cachep)
1813 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1814 return -ENOMEM;
1815 }
1816
1817 return 0;
1818}
1819
1820static void ocfs2_free_mem_caches(void)
1821{
1822
1823
1824
1825
1826 rcu_barrier();
1827 if (ocfs2_inode_cachep)
1828 kmem_cache_destroy(ocfs2_inode_cachep);
1829 ocfs2_inode_cachep = NULL;
1830
1831 if (ocfs2_dquot_cachep)
1832 kmem_cache_destroy(ocfs2_dquot_cachep);
1833 ocfs2_dquot_cachep = NULL;
1834
1835 if (ocfs2_qf_chunk_cachep)
1836 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1837 ocfs2_qf_chunk_cachep = NULL;
1838}
1839
1840static int ocfs2_get_sector(struct super_block *sb,
1841 struct buffer_head **bh,
1842 int block,
1843 int sect_size)
1844{
1845 if (!sb_set_blocksize(sb, sect_size)) {
1846 mlog(ML_ERROR, "unable to set blocksize\n");
1847 return -EIO;
1848 }
1849
1850 *bh = sb_getblk(sb, block);
1851 if (!*bh) {
1852 mlog_errno(-EIO);
1853 return -EIO;
1854 }
1855 lock_buffer(*bh);
1856 if (!buffer_dirty(*bh))
1857 clear_buffer_uptodate(*bh);
1858 unlock_buffer(*bh);
1859 ll_rw_block(READ, 1, bh);
1860 wait_on_buffer(*bh);
1861 if (!buffer_uptodate(*bh)) {
1862 mlog_errno(-EIO);
1863 brelse(*bh);
1864 *bh = NULL;
1865 return -EIO;
1866 }
1867
1868 return 0;
1869}
1870
1871static int ocfs2_mount_volume(struct super_block *sb)
1872{
1873 int status = 0;
1874 int unlock_super = 0;
1875 struct ocfs2_super *osb = OCFS2_SB(sb);
1876
1877 if (ocfs2_is_hard_readonly(osb))
1878 goto leave;
1879
1880 status = ocfs2_dlm_init(osb);
1881 if (status < 0) {
1882 mlog_errno(status);
1883 goto leave;
1884 }
1885
1886 status = ocfs2_super_lock(osb, 1);
1887 if (status < 0) {
1888 mlog_errno(status);
1889 goto leave;
1890 }
1891 unlock_super = 1;
1892
1893
1894 status = ocfs2_find_slot(osb);
1895 if (status < 0) {
1896 mlog_errno(status);
1897 goto leave;
1898 }
1899
1900
1901 status = ocfs2_init_local_system_inodes(osb);
1902 if (status < 0) {
1903 mlog_errno(status);
1904 goto leave;
1905 }
1906
1907 status = ocfs2_check_volume(osb);
1908 if (status < 0) {
1909 mlog_errno(status);
1910 goto leave;
1911 }
1912
1913 status = ocfs2_truncate_log_init(osb);
1914 if (status < 0)
1915 mlog_errno(status);
1916
1917leave:
1918 if (unlock_super)
1919 ocfs2_super_unlock(osb, 1);
1920
1921 return status;
1922}
1923
1924static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1925{
1926 int tmp, hangup_needed = 0;
1927 struct ocfs2_super *osb = NULL;
1928 char nodestr[8];
1929
1930 trace_ocfs2_dismount_volume(sb);
1931
1932 BUG_ON(!sb);
1933 osb = OCFS2_SB(sb);
1934 BUG_ON(!osb);
1935
1936 debugfs_remove(osb->osb_ctxt);
1937
1938
1939
1940
1941
1942 ocfs2_drop_all_dl_inodes(osb);
1943
1944
1945 ocfs2_orphan_scan_stop(osb);
1946
1947 ocfs2_disable_quotas(osb);
1948
1949 ocfs2_shutdown_local_alloc(osb);
1950
1951 ocfs2_truncate_log_shutdown(osb);
1952
1953
1954 ocfs2_recovery_exit(osb);
1955
1956 ocfs2_journal_shutdown(osb);
1957
1958 ocfs2_sync_blockdev(sb);
1959
1960 ocfs2_purge_refcount_trees(osb);
1961
1962
1963
1964 if (osb->cconn) {
1965 tmp = ocfs2_super_lock(osb, 1);
1966 if (tmp < 0) {
1967 mlog_errno(tmp);
1968 return;
1969 }
1970 }
1971
1972 if (osb->slot_num != OCFS2_INVALID_SLOT)
1973 ocfs2_put_slot(osb);
1974
1975 if (osb->cconn)
1976 ocfs2_super_unlock(osb, 1);
1977
1978 ocfs2_release_system_inodes(osb);
1979
1980
1981
1982
1983
1984
1985
1986 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1987 !ocfs2_is_hard_readonly(osb))
1988 hangup_needed = 1;
1989
1990 if (osb->cconn)
1991 ocfs2_dlm_shutdown(osb, hangup_needed);
1992
1993 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1994 debugfs_remove(osb->osb_debug_root);
1995
1996 if (hangup_needed)
1997 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1998
1999 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
2000
2001 if (ocfs2_mount_local(osb))
2002 snprintf(nodestr, sizeof(nodestr), "local");
2003 else
2004 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
2005
2006 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2007 osb->dev_str, nodestr);
2008
2009 ocfs2_delete_osb(osb);
2010 kfree(osb);
2011 sb->s_dev = 0;
2012 sb->s_fs_info = NULL;
2013}
2014
2015static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2016 unsigned uuid_bytes)
2017{
2018 int i, ret;
2019 char *ptr;
2020
2021 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2022
2023 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2024 if (osb->uuid_str == NULL)
2025 return -ENOMEM;
2026
2027 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2028
2029 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2030 if (ret != 2)
2031 return -EINVAL;
2032
2033 ptr += 2;
2034 }
2035
2036 return 0;
2037}
2038
2039
2040
2041
2042static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2043{
2044 int status = 0;
2045 u64 max_block =
2046 ocfs2_clusters_to_blocks(osb->sb,
2047 osb->osb_clusters_at_boot) - 1;
2048
2049
2050 if (max_block <= (u32)~0ULL)
2051 goto out;
2052
2053
2054
2055 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2056 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2057 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2058 JBD2_FEATURE_INCOMPAT_64BIT))) {
2059 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2060 "Enable the 'block64' journal option with tunefs.ocfs2");
2061 status = -EFBIG;
2062 goto out;
2063 }
2064
2065 out:
2066 return status;
2067}
2068
2069static int ocfs2_initialize_super(struct super_block *sb,
2070 struct buffer_head *bh,
2071 int sector_size,
2072 struct ocfs2_blockcheck_stats *stats)
2073{
2074 int status;
2075 int i, cbits, bbits;
2076 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2077 struct inode *inode = NULL;
2078 struct ocfs2_journal *journal;
2079 __le32 uuid_net_key;
2080 struct ocfs2_super *osb;
2081 u64 total_blocks;
2082
2083 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2084 if (!osb) {
2085 status = -ENOMEM;
2086 mlog_errno(status);
2087 goto bail;
2088 }
2089
2090 sb->s_fs_info = osb;
2091 sb->s_op = &ocfs2_sops;
2092 sb->s_d_op = &ocfs2_dentry_ops;
2093 sb->s_export_op = &ocfs2_export_ops;
2094 sb->s_qcop = &ocfs2_quotactl_ops;
2095 sb->dq_op = &ocfs2_quota_operations;
2096 sb->s_xattr = ocfs2_xattr_handlers;
2097 sb->s_time_gran = 1;
2098 sb->s_flags |= MS_NOATIME;
2099
2100 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2101 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2102 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2103
2104 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2105
2106 for (i = 0; i < 3; i++)
2107 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2108 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2109
2110 osb->sb = sb;
2111
2112 osb->s_sectsize_bits = blksize_bits(sector_size);
2113 BUG_ON(!osb->s_sectsize_bits);
2114
2115 spin_lock_init(&osb->dc_task_lock);
2116 init_waitqueue_head(&osb->dc_event);
2117 osb->dc_work_sequence = 0;
2118 osb->dc_wake_sequence = 0;
2119 INIT_LIST_HEAD(&osb->blocked_lock_list);
2120 osb->blocked_lock_count = 0;
2121 spin_lock_init(&osb->osb_lock);
2122 spin_lock_init(&osb->osb_xattr_lock);
2123 ocfs2_init_steal_slots(osb);
2124
2125 atomic_set(&osb->alloc_stats.moves, 0);
2126 atomic_set(&osb->alloc_stats.local_data, 0);
2127 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2128 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2129 atomic_set(&osb->alloc_stats.bg_extends, 0);
2130
2131
2132 osb->osb_ecc_stats = *stats;
2133
2134 ocfs2_init_node_maps(osb);
2135
2136 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2137 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2138
2139 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2140 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2141 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2142 osb->max_slots);
2143 status = -EINVAL;
2144 goto bail;
2145 }
2146
2147 ocfs2_orphan_scan_init(osb);
2148
2149 status = ocfs2_recovery_init(osb);
2150 if (status) {
2151 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2152 mlog_errno(status);
2153 goto bail;
2154 }
2155
2156 init_waitqueue_head(&osb->checkpoint_event);
2157 atomic_set(&osb->needs_checkpoint, 0);
2158
2159 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2160
2161 osb->slot_num = OCFS2_INVALID_SLOT;
2162
2163 osb->s_xattr_inline_size = le16_to_cpu(
2164 di->id2.i_super.s_xattr_inline_size);
2165
2166 osb->local_alloc_state = OCFS2_LA_UNUSED;
2167 osb->local_alloc_bh = NULL;
2168 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2169
2170 init_waitqueue_head(&osb->osb_mount_event);
2171
2172 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2173 if (status) {
2174 mlog_errno(status);
2175 goto bail;
2176 }
2177
2178 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2179 if (!osb->vol_label) {
2180 mlog(ML_ERROR, "unable to alloc vol label\n");
2181 status = -ENOMEM;
2182 goto bail;
2183 }
2184
2185 osb->slot_recovery_generations =
2186 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2187 GFP_KERNEL);
2188 if (!osb->slot_recovery_generations) {
2189 status = -ENOMEM;
2190 mlog_errno(status);
2191 goto bail;
2192 }
2193
2194 init_waitqueue_head(&osb->osb_wipe_event);
2195 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2196 sizeof(*osb->osb_orphan_wipes),
2197 GFP_KERNEL);
2198 if (!osb->osb_orphan_wipes) {
2199 status = -ENOMEM;
2200 mlog_errno(status);
2201 goto bail;
2202 }
2203
2204 osb->osb_rf_lock_tree = RB_ROOT;
2205
2206 osb->s_feature_compat =
2207 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2208 osb->s_feature_ro_compat =
2209 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2210 osb->s_feature_incompat =
2211 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2212
2213 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2214 mlog(ML_ERROR, "couldn't mount because of unsupported "
2215 "optional features (%x).\n", i);
2216 status = -EINVAL;
2217 goto bail;
2218 }
2219 if (!(osb->sb->s_flags & MS_RDONLY) &&
2220 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2221 mlog(ML_ERROR, "couldn't mount RDWR because of "
2222 "unsupported optional features (%x).\n", i);
2223 status = -EINVAL;
2224 goto bail;
2225 }
2226
2227 if (ocfs2_clusterinfo_valid(osb)) {
2228 osb->osb_stackflags =
2229 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2230 memcpy(osb->osb_cluster_stack,
2231 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2232 OCFS2_STACK_LABEL_LEN);
2233 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2234 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2235 mlog(ML_ERROR,
2236 "couldn't mount because of an invalid "
2237 "cluster stack label (%s) \n",
2238 osb->osb_cluster_stack);
2239 status = -EINVAL;
2240 goto bail;
2241 }
2242 } else {
2243
2244
2245 osb->osb_cluster_stack[0] = '\0';
2246 }
2247
2248 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2249
2250
2251
2252
2253
2254
2255
2256
2257
2258
2259 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2260 if (!journal) {
2261 mlog(ML_ERROR, "unable to alloc journal\n");
2262 status = -ENOMEM;
2263 goto bail;
2264 }
2265 osb->journal = journal;
2266 journal->j_osb = osb;
2267
2268 atomic_set(&journal->j_num_trans, 0);
2269 init_rwsem(&journal->j_trans_barrier);
2270 init_waitqueue_head(&journal->j_checkpointed);
2271 spin_lock_init(&journal->j_lock);
2272 journal->j_trans_id = (unsigned long) 1;
2273 INIT_LIST_HEAD(&journal->j_la_cleanups);
2274 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2275 journal->j_state = OCFS2_JOURNAL_FREE;
2276
2277 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2278 osb->dentry_lock_list = NULL;
2279
2280
2281 osb->s_clustersize_bits =
2282 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2283 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2284
2285 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2286 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2287 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2288 osb->s_clustersize);
2289 status = -EINVAL;
2290 goto bail;
2291 }
2292
2293 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2294 le32_to_cpu(di->i_clusters));
2295
2296 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2297 total_blocks);
2298 if (status) {
2299 mlog(ML_ERROR, "Volume too large "
2300 "to mount safely on this system");
2301 status = -EFBIG;
2302 goto bail;
2303 }
2304
2305 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2306 sizeof(di->id2.i_super.s_uuid))) {
2307 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2308 status = -ENOMEM;
2309 goto bail;
2310 }
2311
2312 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2313
2314 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2315 osb->vol_label[63] = '\0';
2316 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2317 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2318 osb->first_cluster_group_blkno =
2319 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2320 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2321 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2322 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2323 (unsigned long long)osb->root_blkno,
2324 (unsigned long long)osb->system_dir_blkno,
2325 osb->s_clustersize_bits);
2326
2327 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2328 if (!osb->osb_dlm_debug) {
2329 status = -ENOMEM;
2330 mlog_errno(status);
2331 goto bail;
2332 }
2333
2334 atomic_set(&osb->vol_state, VOLUME_INIT);
2335
2336
2337 status = ocfs2_init_global_system_inodes(osb);
2338 if (status < 0) {
2339 mlog_errno(status);
2340 goto bail;
2341 }
2342
2343
2344
2345
2346 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2347 OCFS2_INVALID_SLOT);
2348 if (!inode) {
2349 status = -EINVAL;
2350 mlog_errno(status);
2351 goto bail;
2352 }
2353
2354 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2355 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2356 iput(inode);
2357
2358 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2359 osb->s_feature_incompat) * 8;
2360
2361 status = ocfs2_init_slot_info(osb);
2362 if (status < 0) {
2363 mlog_errno(status);
2364 goto bail;
2365 }
2366 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2367
2368bail:
2369 return status;
2370}
2371
2372
2373
2374
2375
2376
2377static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2378 struct buffer_head *bh,
2379 u32 blksz,
2380 struct ocfs2_blockcheck_stats *stats)
2381{
2382 int status = -EAGAIN;
2383
2384 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2385 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2386
2387 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2388 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2389 status = ocfs2_block_check_validate(bh->b_data,
2390 bh->b_size,
2391 &di->i_check,
2392 stats);
2393 if (status)
2394 goto out;
2395 }
2396 status = -EINVAL;
2397 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2398 mlog(ML_ERROR, "found superblock with incorrect block "
2399 "size: found %u, should be %u\n",
2400 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2401 blksz);
2402 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2403 OCFS2_MAJOR_REV_LEVEL ||
2404 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2405 OCFS2_MINOR_REV_LEVEL) {
2406 mlog(ML_ERROR, "found superblock with bad version: "
2407 "found %u.%u, should be %u.%u\n",
2408 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2409 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2410 OCFS2_MAJOR_REV_LEVEL,
2411 OCFS2_MINOR_REV_LEVEL);
2412 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2413 mlog(ML_ERROR, "bad block number on superblock: "
2414 "found %llu, should be %llu\n",
2415 (unsigned long long)le64_to_cpu(di->i_blkno),
2416 (unsigned long long)bh->b_blocknr);
2417 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2418 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2419 mlog(ML_ERROR, "bad cluster size found: %u\n",
2420 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2421 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2422 mlog(ML_ERROR, "bad root_blkno: 0\n");
2423 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2424 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2425 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2426 mlog(ML_ERROR,
2427 "Superblock slots found greater than file system "
2428 "maximum: found %u, max %u\n",
2429 le16_to_cpu(di->id2.i_super.s_max_slots),
2430 OCFS2_MAX_SLOTS);
2431 } else {
2432
2433 status = 0;
2434 }
2435 }
2436
2437out:
2438 if (status && status != -EAGAIN)
2439 mlog_errno(status);
2440 return status;
2441}
2442
2443static int ocfs2_check_volume(struct ocfs2_super *osb)
2444{
2445 int status;
2446 int dirty;
2447 int local;
2448 struct ocfs2_dinode *local_alloc = NULL;
2449
2450
2451
2452
2453 status = ocfs2_journal_init(osb->journal, &dirty);
2454 if (status < 0) {
2455 mlog(ML_ERROR, "Could not initialize journal!\n");
2456 goto finally;
2457 }
2458
2459
2460
2461 status = ocfs2_journal_addressable(osb);
2462 if (status)
2463 goto finally;
2464
2465
2466
2467
2468 if (!dirty) {
2469 status = ocfs2_journal_wipe(osb->journal, 0);
2470 if (status < 0) {
2471 mlog_errno(status);
2472 goto finally;
2473 }
2474 } else {
2475 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2476 "unmounted cleanly, recovering it.\n", osb->dev_str);
2477 }
2478
2479 local = ocfs2_mount_local(osb);
2480
2481
2482 status = ocfs2_journal_load(osb->journal, local, dirty);
2483 if (status < 0) {
2484 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2485 goto finally;
2486 }
2487
2488 if (dirty) {
2489
2490 status = ocfs2_begin_local_alloc_recovery(osb,
2491 osb->slot_num,
2492 &local_alloc);
2493 if (status < 0) {
2494 mlog_errno(status);
2495 goto finally;
2496 }
2497
2498
2499 }
2500
2501 status = ocfs2_load_local_alloc(osb);
2502 if (status < 0) {
2503 mlog_errno(status);
2504 goto finally;
2505 }
2506
2507 if (dirty) {
2508
2509
2510 osb->dirty = 1;
2511 osb->local_alloc_copy = local_alloc;
2512 local_alloc = NULL;
2513 }
2514
2515
2516
2517
2518 status = ocfs2_mark_dead_nodes(osb);
2519 if (status < 0) {
2520 mlog_errno(status);
2521 goto finally;
2522 }
2523
2524 status = ocfs2_compute_replay_slots(osb);
2525 if (status < 0)
2526 mlog_errno(status);
2527
2528finally:
2529 kfree(local_alloc);
2530
2531 if (status)
2532 mlog_errno(status);
2533 return status;
2534}
2535
2536
2537
2538
2539
2540
2541
2542static void ocfs2_delete_osb(struct ocfs2_super *osb)
2543{
2544
2545
2546 ocfs2_free_slot_info(osb);
2547
2548 kfree(osb->osb_orphan_wipes);
2549 kfree(osb->slot_recovery_generations);
2550
2551
2552
2553
2554
2555 kfree(osb->journal);
2556 kfree(osb->local_alloc_copy);
2557 kfree(osb->uuid_str);
2558 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2559 memset(osb, 0, sizeof(struct ocfs2_super));
2560}
2561
2562
2563
2564static void ocfs2_handle_error(struct super_block *sb)
2565{
2566 struct ocfs2_super *osb = OCFS2_SB(sb);
2567
2568 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2569 panic("OCFS2: (device %s): panic forced after error\n",
2570 sb->s_id);
2571
2572 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2573
2574 if (sb->s_flags & MS_RDONLY &&
2575 (ocfs2_is_soft_readonly(osb) ||
2576 ocfs2_is_hard_readonly(osb)))
2577 return;
2578
2579 printk(KERN_CRIT "File system is now read-only due to the potential "
2580 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2581 "system is unmounted.\n");
2582 sb->s_flags |= MS_RDONLY;
2583 ocfs2_set_ro_flag(osb, 0);
2584}
2585
2586static char error_buf[1024];
2587
2588void __ocfs2_error(struct super_block *sb,
2589 const char *function,
2590 const char *fmt, ...)
2591{
2592 va_list args;
2593
2594 va_start(args, fmt);
2595 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2596 va_end(args);
2597
2598
2599
2600 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2601 sb->s_id, function, error_buf);
2602
2603 ocfs2_handle_error(sb);
2604}
2605
2606
2607
2608
2609void __ocfs2_abort(struct super_block* sb,
2610 const char *function,
2611 const char *fmt, ...)
2612{
2613 va_list args;
2614
2615 va_start(args, fmt);
2616 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2617 va_end(args);
2618
2619 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2620 sb->s_id, function, error_buf);
2621
2622
2623
2624
2625
2626
2627
2628
2629
2630
2631
2632
2633 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2634 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2635 ocfs2_handle_error(sb);
2636}
2637
2638
2639
2640
2641
2642void ocfs2_block_signals(sigset_t *oldset)
2643{
2644 int rc;
2645 sigset_t blocked;
2646
2647 sigfillset(&blocked);
2648 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2649 BUG_ON(rc);
2650}
2651
2652void ocfs2_unblock_signals(sigset_t *oldset)
2653{
2654 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2655 BUG_ON(rc);
2656}
2657
2658module_init(ocfs2_init);
2659module_exit(ocfs2_exit);
2660
