Loading...
1// SPDX-License-Identifier: GPL-2.0-or-later
2/*
3 * super.c
4 *
5 * load/unload driver, mount/dismount volumes
6 *
7 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
8 */
9
10#include <linux/module.h>
11#include <linux/fs.h>
12#include <linux/types.h>
13#include <linux/slab.h>
14#include <linux/highmem.h>
15#include <linux/init.h>
16#include <linux/random.h>
17#include <linux/statfs.h>
18#include <linux/moduleparam.h>
19#include <linux/blkdev.h>
20#include <linux/socket.h>
21#include <linux/inet.h>
22#include <linux/parser.h>
23#include <linux/crc32.h>
24#include <linux/debugfs.h>
25#include <linux/mount.h>
26#include <linux/seq_file.h>
27#include <linux/quotaops.h>
28#include <linux/signal.h>
29
30#define CREATE_TRACE_POINTS
31#include "ocfs2_trace.h"
32
33#include <cluster/masklog.h>
34
35#include "ocfs2.h"
36
37/* this should be the only file to include a version 1 header */
38#include "ocfs1_fs_compat.h"
39
40#include "alloc.h"
41#include "aops.h"
42#include "blockcheck.h"
43#include "dlmglue.h"
44#include "export.h"
45#include "extent_map.h"
46#include "heartbeat.h"
47#include "inode.h"
48#include "journal.h"
49#include "localalloc.h"
50#include "namei.h"
51#include "slot_map.h"
52#include "super.h"
53#include "sysfile.h"
54#include "uptodate.h"
55#include "xattr.h"
56#include "quota.h"
57#include "refcounttree.h"
58#include "suballoc.h"
59
60#include "buffer_head_io.h"
61#include "filecheck.h"
62
63static struct kmem_cache *ocfs2_inode_cachep;
64struct kmem_cache *ocfs2_dquot_cachep;
65struct kmem_cache *ocfs2_qf_chunk_cachep;
66
67static struct dentry *ocfs2_debugfs_root;
68
69MODULE_AUTHOR("Oracle");
70MODULE_LICENSE("GPL");
71MODULE_DESCRIPTION("OCFS2 cluster file system");
72
73struct mount_options
74{
75 unsigned long commit_interval;
76 unsigned long mount_opt;
77 unsigned int atime_quantum;
78 unsigned short slot;
79 int localalloc_opt;
80 unsigned int resv_level;
81 int dir_resv_level;
82 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
83};
84
85static int ocfs2_parse_options(struct super_block *sb, char *options,
86 struct mount_options *mopt,
87 int is_remount);
88static int ocfs2_check_set_options(struct super_block *sb,
89 struct mount_options *options);
90static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
91static void ocfs2_put_super(struct super_block *sb);
92static int ocfs2_mount_volume(struct super_block *sb);
93static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
94static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
95static int ocfs2_initialize_mem_caches(void);
96static void ocfs2_free_mem_caches(void);
97static void ocfs2_delete_osb(struct ocfs2_super *osb);
98
99static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
100
101static int ocfs2_sync_fs(struct super_block *sb, int wait);
102
103static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
104static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
105static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
106static int ocfs2_check_volume(struct ocfs2_super *osb);
107static int ocfs2_verify_volume(struct ocfs2_dinode *di,
108 struct buffer_head *bh,
109 u32 sectsize,
110 struct ocfs2_blockcheck_stats *stats);
111static int ocfs2_initialize_super(struct super_block *sb,
112 struct buffer_head *bh,
113 int sector_size,
114 struct ocfs2_blockcheck_stats *stats);
115static int ocfs2_get_sector(struct super_block *sb,
116 struct buffer_head **bh,
117 int block,
118 int sect_size);
119static struct inode *ocfs2_alloc_inode(struct super_block *sb);
120static void ocfs2_free_inode(struct inode *inode);
121static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
122static int ocfs2_enable_quotas(struct ocfs2_super *osb);
123static void ocfs2_disable_quotas(struct ocfs2_super *osb);
124
125static struct dquot **ocfs2_get_dquots(struct inode *inode)
126{
127 return OCFS2_I(inode)->i_dquot;
128}
129
130static const struct super_operations ocfs2_sops = {
131 .statfs = ocfs2_statfs,
132 .alloc_inode = ocfs2_alloc_inode,
133 .free_inode = ocfs2_free_inode,
134 .drop_inode = ocfs2_drop_inode,
135 .evict_inode = ocfs2_evict_inode,
136 .sync_fs = ocfs2_sync_fs,
137 .put_super = ocfs2_put_super,
138 .remount_fs = ocfs2_remount,
139 .show_options = ocfs2_show_options,
140 .quota_read = ocfs2_quota_read,
141 .quota_write = ocfs2_quota_write,
142 .get_dquots = ocfs2_get_dquots,
143};
144
145enum {
146 Opt_barrier,
147 Opt_err_panic,
148 Opt_err_ro,
149 Opt_intr,
150 Opt_nointr,
151 Opt_hb_none,
152 Opt_hb_local,
153 Opt_hb_global,
154 Opt_data_ordered,
155 Opt_data_writeback,
156 Opt_atime_quantum,
157 Opt_slot,
158 Opt_commit,
159 Opt_localalloc,
160 Opt_localflocks,
161 Opt_stack,
162 Opt_user_xattr,
163 Opt_nouser_xattr,
164 Opt_inode64,
165 Opt_acl,
166 Opt_noacl,
167 Opt_usrquota,
168 Opt_grpquota,
169 Opt_coherency_buffered,
170 Opt_coherency_full,
171 Opt_resv_level,
172 Opt_dir_resv_level,
173 Opt_journal_async_commit,
174 Opt_err_cont,
175 Opt_err,
176};
177
178static const match_table_t tokens = {
179 {Opt_barrier, "barrier=%u"},
180 {Opt_err_panic, "errors=panic"},
181 {Opt_err_ro, "errors=remount-ro"},
182 {Opt_intr, "intr"},
183 {Opt_nointr, "nointr"},
184 {Opt_hb_none, OCFS2_HB_NONE},
185 {Opt_hb_local, OCFS2_HB_LOCAL},
186 {Opt_hb_global, OCFS2_HB_GLOBAL},
187 {Opt_data_ordered, "data=ordered"},
188 {Opt_data_writeback, "data=writeback"},
189 {Opt_atime_quantum, "atime_quantum=%u"},
190 {Opt_slot, "preferred_slot=%u"},
191 {Opt_commit, "commit=%u"},
192 {Opt_localalloc, "localalloc=%d"},
193 {Opt_localflocks, "localflocks"},
194 {Opt_stack, "cluster_stack=%s"},
195 {Opt_user_xattr, "user_xattr"},
196 {Opt_nouser_xattr, "nouser_xattr"},
197 {Opt_inode64, "inode64"},
198 {Opt_acl, "acl"},
199 {Opt_noacl, "noacl"},
200 {Opt_usrquota, "usrquota"},
201 {Opt_grpquota, "grpquota"},
202 {Opt_coherency_buffered, "coherency=buffered"},
203 {Opt_coherency_full, "coherency=full"},
204 {Opt_resv_level, "resv_level=%u"},
205 {Opt_dir_resv_level, "dir_resv_level=%u"},
206 {Opt_journal_async_commit, "journal_async_commit"},
207 {Opt_err_cont, "errors=continue"},
208 {Opt_err, NULL}
209};
210
211#ifdef CONFIG_DEBUG_FS
212static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
213{
214 struct ocfs2_cluster_connection *cconn = osb->cconn;
215 struct ocfs2_recovery_map *rm = osb->recovery_map;
216 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
217 int i, out = 0;
218 unsigned long flags;
219
220 out += scnprintf(buf + out, len - out,
221 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
222 "Device", osb->dev_str, osb->uuid_str,
223 osb->fs_generation, osb->vol_label);
224
225 out += scnprintf(buf + out, len - out,
226 "%10s => State: %d Flags: 0x%lX\n", "Volume",
227 atomic_read(&osb->vol_state), osb->osb_flags);
228
229 out += scnprintf(buf + out, len - out,
230 "%10s => Block: %lu Cluster: %d\n", "Sizes",
231 osb->sb->s_blocksize, osb->s_clustersize);
232
233 out += scnprintf(buf + out, len - out,
234 "%10s => Compat: 0x%X Incompat: 0x%X "
235 "ROcompat: 0x%X\n",
236 "Features", osb->s_feature_compat,
237 osb->s_feature_incompat, osb->s_feature_ro_compat);
238
239 out += scnprintf(buf + out, len - out,
240 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
241 osb->s_mount_opt, osb->s_atime_quantum);
242
243 if (cconn) {
244 out += scnprintf(buf + out, len - out,
245 "%10s => Stack: %s Name: %*s "
246 "Version: %d.%d\n", "Cluster",
247 (*osb->osb_cluster_stack == '\0' ?
248 "o2cb" : osb->osb_cluster_stack),
249 cconn->cc_namelen, cconn->cc_name,
250 cconn->cc_version.pv_major,
251 cconn->cc_version.pv_minor);
252 }
253
254 spin_lock_irqsave(&osb->dc_task_lock, flags);
255 out += scnprintf(buf + out, len - out,
256 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
257 "WorkSeq: %lu\n", "DownCnvt",
258 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
259 osb->blocked_lock_count, osb->dc_wake_sequence,
260 osb->dc_work_sequence);
261 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
262
263 spin_lock(&osb->osb_lock);
264 out += scnprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
265 "Recovery",
266 (osb->recovery_thread_task ?
267 task_pid_nr(osb->recovery_thread_task) : -1));
268 if (rm->rm_used == 0)
269 out += scnprintf(buf + out, len - out, " None\n");
270 else {
271 for (i = 0; i < rm->rm_used; i++)
272 out += scnprintf(buf + out, len - out, " %d",
273 rm->rm_entries[i]);
274 out += scnprintf(buf + out, len - out, "\n");
275 }
276 spin_unlock(&osb->osb_lock);
277
278 out += scnprintf(buf + out, len - out,
279 "%10s => Pid: %d Interval: %lu\n", "Commit",
280 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
281 osb->osb_commit_interval);
282
283 out += scnprintf(buf + out, len - out,
284 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
285 "Journal", osb->journal->j_state,
286 osb->journal->j_trans_id,
287 atomic_read(&osb->journal->j_num_trans));
288
289 out += scnprintf(buf + out, len - out,
290 "%10s => GlobalAllocs: %d LocalAllocs: %d "
291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
292 "Stats",
293 atomic_read(&osb->alloc_stats.bitmap_data),
294 atomic_read(&osb->alloc_stats.local_data),
295 atomic_read(&osb->alloc_stats.bg_allocs),
296 atomic_read(&osb->alloc_stats.moves),
297 atomic_read(&osb->alloc_stats.bg_extends));
298
299 out += scnprintf(buf + out, len - out,
300 "%10s => State: %u Descriptor: %llu Size: %u bits "
301 "Default: %u bits\n",
302 "LocalAlloc", osb->local_alloc_state,
303 (unsigned long long)osb->la_last_gd,
304 osb->local_alloc_bits, osb->local_alloc_default_bits);
305
306 spin_lock(&osb->osb_lock);
307 out += scnprintf(buf + out, len - out,
308 "%10s => InodeSlot: %d StolenInodes: %d, "
309 "MetaSlot: %d StolenMeta: %d\n", "Steal",
310 osb->s_inode_steal_slot,
311 atomic_read(&osb->s_num_inodes_stolen),
312 osb->s_meta_steal_slot,
313 atomic_read(&osb->s_num_meta_stolen));
314 spin_unlock(&osb->osb_lock);
315
316 out += scnprintf(buf + out, len - out, "OrphanScan => ");
317 out += scnprintf(buf + out, len - out, "Local: %u Global: %u ",
318 os->os_count, os->os_seqno);
319 out += scnprintf(buf + out, len - out, " Last Scan: ");
320 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
321 out += scnprintf(buf + out, len - out, "Disabled\n");
322 else
323 out += scnprintf(buf + out, len - out, "%lu seconds ago\n",
324 (unsigned long)(ktime_get_seconds() - os->os_scantime));
325
326 out += scnprintf(buf + out, len - out, "%10s => %3s %10s\n",
327 "Slots", "Num", "RecoGen");
328 for (i = 0; i < osb->max_slots; ++i) {
329 out += scnprintf(buf + out, len - out,
330 "%10s %c %3d %10d\n",
331 " ",
332 (i == osb->slot_num ? '*' : ' '),
333 i, osb->slot_recovery_generations[i]);
334 }
335
336 return out;
337}
338
339static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
340{
341 struct ocfs2_super *osb = inode->i_private;
342 char *buf = NULL;
343
344 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
345 if (!buf)
346 goto bail;
347
348 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
349
350 file->private_data = buf;
351
352 return 0;
353bail:
354 return -ENOMEM;
355}
356
357static int ocfs2_debug_release(struct inode *inode, struct file *file)
358{
359 kfree(file->private_data);
360 return 0;
361}
362
363static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
364 size_t nbytes, loff_t *ppos)
365{
366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
367 i_size_read(file->f_mapping->host));
368}
369#else
370static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
371{
372 return 0;
373}
374static int ocfs2_debug_release(struct inode *inode, struct file *file)
375{
376 return 0;
377}
378static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
379 size_t nbytes, loff_t *ppos)
380{
381 return 0;
382}
383#endif /* CONFIG_DEBUG_FS */
384
385static const struct file_operations ocfs2_osb_debug_fops = {
386 .open = ocfs2_osb_debug_open,
387 .release = ocfs2_debug_release,
388 .read = ocfs2_debug_read,
389 .llseek = generic_file_llseek,
390};
391
392static int ocfs2_sync_fs(struct super_block *sb, int wait)
393{
394 int status;
395 tid_t target;
396 struct ocfs2_super *osb = OCFS2_SB(sb);
397
398 if (ocfs2_is_hard_readonly(osb))
399 return -EROFS;
400
401 if (wait) {
402 status = ocfs2_flush_truncate_log(osb);
403 if (status < 0)
404 mlog_errno(status);
405 } else {
406 ocfs2_schedule_truncate_log_flush(osb, 0);
407 }
408
409 if (jbd2_journal_start_commit(osb->journal->j_journal,
410 &target)) {
411 if (wait)
412 jbd2_log_wait_commit(osb->journal->j_journal,
413 target);
414 }
415 return 0;
416}
417
418static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
419{
420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
421 && (ino == USER_QUOTA_SYSTEM_INODE
422 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
423 return 0;
424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
425 && (ino == GROUP_QUOTA_SYSTEM_INODE
426 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
427 return 0;
428 return 1;
429}
430
431static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
432{
433 struct inode *new = NULL;
434 int status = 0;
435 int i;
436
437 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
438 if (IS_ERR(new)) {
439 status = PTR_ERR(new);
440 mlog_errno(status);
441 goto bail;
442 }
443 osb->root_inode = new;
444
445 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
446 if (IS_ERR(new)) {
447 status = PTR_ERR(new);
448 mlog_errno(status);
449 goto bail;
450 }
451 osb->sys_root_inode = new;
452
453 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
454 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
455 if (!ocfs2_need_system_inode(osb, i))
456 continue;
457 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
458 if (!new) {
459 ocfs2_release_system_inodes(osb);
460 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
461 mlog_errno(status);
462 mlog(ML_ERROR, "Unable to load system inode %d, "
463 "possibly corrupt fs?", i);
464 goto bail;
465 }
466 // the array now has one ref, so drop this one
467 iput(new);
468 }
469
470bail:
471 if (status)
472 mlog_errno(status);
473 return status;
474}
475
476static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
477{
478 struct inode *new = NULL;
479 int status = 0;
480 int i;
481
482 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
483 i < NUM_SYSTEM_INODES;
484 i++) {
485 if (!ocfs2_need_system_inode(osb, i))
486 continue;
487 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
488 if (!new) {
489 ocfs2_release_system_inodes(osb);
490 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
491 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
492 status, i, osb->slot_num);
493 goto bail;
494 }
495 /* the array now has one ref, so drop this one */
496 iput(new);
497 }
498
499bail:
500 if (status)
501 mlog_errno(status);
502 return status;
503}
504
505static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
506{
507 int i;
508 struct inode *inode;
509
510 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
511 inode = osb->global_system_inodes[i];
512 if (inode) {
513 iput(inode);
514 osb->global_system_inodes[i] = NULL;
515 }
516 }
517
518 inode = osb->sys_root_inode;
519 if (inode) {
520 iput(inode);
521 osb->sys_root_inode = NULL;
522 }
523
524 inode = osb->root_inode;
525 if (inode) {
526 iput(inode);
527 osb->root_inode = NULL;
528 }
529
530 if (!osb->local_system_inodes)
531 return;
532
533 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
534 if (osb->local_system_inodes[i]) {
535 iput(osb->local_system_inodes[i]);
536 osb->local_system_inodes[i] = NULL;
537 }
538 }
539
540 kfree(osb->local_system_inodes);
541 osb->local_system_inodes = NULL;
542}
543
544/* We're allocating fs objects, use GFP_NOFS */
545static struct inode *ocfs2_alloc_inode(struct super_block *sb)
546{
547 struct ocfs2_inode_info *oi;
548
549 oi = alloc_inode_sb(sb, ocfs2_inode_cachep, GFP_NOFS);
550 if (!oi)
551 return NULL;
552
553 oi->i_sync_tid = 0;
554 oi->i_datasync_tid = 0;
555 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
556
557 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
558 return &oi->vfs_inode;
559}
560
561static void ocfs2_free_inode(struct inode *inode)
562{
563 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
564}
565
566static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
567 unsigned int cbits)
568{
569 unsigned int bytes = 1 << cbits;
570 unsigned int trim = bytes;
571 unsigned int bitshift = 32;
572
573 /*
574 * i_size and all block offsets in ocfs2 are always 64 bits
575 * wide. i_clusters is 32 bits, in cluster-sized units. So on
576 * 64 bit platforms, cluster size will be the limiting factor.
577 */
578
579#if BITS_PER_LONG == 32
580 BUILD_BUG_ON(sizeof(sector_t) != 8);
581 /*
582 * We might be limited by page cache size.
583 */
584 if (bytes > PAGE_SIZE) {
585 bytes = PAGE_SIZE;
586 trim = 1;
587 /*
588 * Shift by 31 here so that we don't get larger than
589 * MAX_LFS_FILESIZE
590 */
591 bitshift = 31;
592 }
593#endif
594
595 /*
596 * Trim by a whole cluster when we can actually approach the
597 * on-disk limits. Otherwise we can overflow i_clusters when
598 * an extent start is at the max offset.
599 */
600 return (((unsigned long long)bytes) << bitshift) - trim;
601}
602
603static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
604{
605 int incompat_features;
606 int ret = 0;
607 struct mount_options parsed_options;
608 struct ocfs2_super *osb = OCFS2_SB(sb);
609 u32 tmp;
610
611 sync_filesystem(sb);
612
613 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
614 !ocfs2_check_set_options(sb, &parsed_options)) {
615 ret = -EINVAL;
616 goto out;
617 }
618
619 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
620 OCFS2_MOUNT_HB_NONE;
621 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
622 ret = -EINVAL;
623 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
624 goto out;
625 }
626
627 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
628 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
629 ret = -EINVAL;
630 mlog(ML_ERROR, "Cannot change data mode on remount\n");
631 goto out;
632 }
633
634 /* Probably don't want this on remount; it might
635 * mess with other nodes */
636 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
637 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
638 ret = -EINVAL;
639 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
640 goto out;
641 }
642
643 /* We're going to/from readonly mode. */
644 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
645 /* Disable quota accounting before remounting RO */
646 if (*flags & SB_RDONLY) {
647 ret = ocfs2_susp_quotas(osb, 0);
648 if (ret < 0)
649 goto out;
650 }
651 /* Lock here so the check of HARD_RO and the potential
652 * setting of SOFT_RO is atomic. */
653 spin_lock(&osb->osb_lock);
654 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
655 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
656 ret = -EROFS;
657 goto unlock_osb;
658 }
659
660 if (*flags & SB_RDONLY) {
661 sb->s_flags |= SB_RDONLY;
662 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
663 } else {
664 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
665 mlog(ML_ERROR, "Cannot remount RDWR "
666 "filesystem due to previous errors.\n");
667 ret = -EROFS;
668 goto unlock_osb;
669 }
670 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
671 if (incompat_features) {
672 mlog(ML_ERROR, "Cannot remount RDWR because "
673 "of unsupported optional features "
674 "(%x).\n", incompat_features);
675 ret = -EINVAL;
676 goto unlock_osb;
677 }
678 sb->s_flags &= ~SB_RDONLY;
679 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
680 }
681 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
682unlock_osb:
683 spin_unlock(&osb->osb_lock);
684 /* Enable quota accounting after remounting RW */
685 if (!ret && !(*flags & SB_RDONLY)) {
686 if (sb_any_quota_suspended(sb))
687 ret = ocfs2_susp_quotas(osb, 1);
688 else
689 ret = ocfs2_enable_quotas(osb);
690 if (ret < 0) {
691 /* Return back changes... */
692 spin_lock(&osb->osb_lock);
693 sb->s_flags |= SB_RDONLY;
694 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
695 spin_unlock(&osb->osb_lock);
696 goto out;
697 }
698 }
699 }
700
701 if (!ret) {
702 /* Only save off the new mount options in case of a successful
703 * remount. */
704 osb->s_mount_opt = parsed_options.mount_opt;
705 osb->s_atime_quantum = parsed_options.atime_quantum;
706 osb->preferred_slot = parsed_options.slot;
707 if (parsed_options.commit_interval)
708 osb->osb_commit_interval = parsed_options.commit_interval;
709
710 if (!ocfs2_is_hard_readonly(osb))
711 ocfs2_set_journal_params(osb);
712
713 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
714 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
715 SB_POSIXACL : 0);
716 }
717out:
718 return ret;
719}
720
721static int ocfs2_sb_probe(struct super_block *sb,
722 struct buffer_head **bh,
723 int *sector_size,
724 struct ocfs2_blockcheck_stats *stats)
725{
726 int status, tmpstat;
727 struct ocfs1_vol_disk_hdr *hdr;
728 struct ocfs2_dinode *di;
729 int blksize;
730
731 *bh = NULL;
732
733 /* may be > 512 */
734 *sector_size = bdev_logical_block_size(sb->s_bdev);
735 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
736 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
737 *sector_size, OCFS2_MAX_BLOCKSIZE);
738 status = -EINVAL;
739 goto bail;
740 }
741
742 /* Can this really happen? */
743 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
744 *sector_size = OCFS2_MIN_BLOCKSIZE;
745
746 /* check block zero for old format */
747 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
748 if (status < 0) {
749 mlog_errno(status);
750 goto bail;
751 }
752 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
753 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
754 mlog(ML_ERROR, "incompatible version: %u.%u\n",
755 hdr->major_version, hdr->minor_version);
756 status = -EINVAL;
757 }
758 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
759 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
760 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
761 hdr->signature);
762 status = -EINVAL;
763 }
764 brelse(*bh);
765 *bh = NULL;
766 if (status < 0) {
767 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
768 "upgraded before mounting with ocfs v2\n");
769 goto bail;
770 }
771
772 /*
773 * Now check at magic offset for 512, 1024, 2048, 4096
774 * blocksizes. 4096 is the maximum blocksize because it is
775 * the minimum clustersize.
776 */
777 status = -EINVAL;
778 for (blksize = *sector_size;
779 blksize <= OCFS2_MAX_BLOCKSIZE;
780 blksize <<= 1) {
781 tmpstat = ocfs2_get_sector(sb, bh,
782 OCFS2_SUPER_BLOCK_BLKNO,
783 blksize);
784 if (tmpstat < 0) {
785 status = tmpstat;
786 mlog_errno(status);
787 break;
788 }
789 di = (struct ocfs2_dinode *) (*bh)->b_data;
790 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
791 spin_lock_init(&stats->b_lock);
792 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
793 if (tmpstat < 0) {
794 brelse(*bh);
795 *bh = NULL;
796 }
797 if (tmpstat != -EAGAIN) {
798 status = tmpstat;
799 break;
800 }
801 }
802
803bail:
804 return status;
805}
806
807static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
808{
809 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
810
811 if (osb->s_mount_opt & hb_enabled) {
812 if (ocfs2_mount_local(osb)) {
813 mlog(ML_ERROR, "Cannot heartbeat on a locally "
814 "mounted device.\n");
815 return -EINVAL;
816 }
817 if (ocfs2_userspace_stack(osb)) {
818 mlog(ML_ERROR, "Userspace stack expected, but "
819 "o2cb heartbeat arguments passed to mount\n");
820 return -EINVAL;
821 }
822 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
823 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
824 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
825 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
826 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
827 return -EINVAL;
828 }
829 }
830
831 if (!(osb->s_mount_opt & hb_enabled)) {
832 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
833 !ocfs2_userspace_stack(osb)) {
834 mlog(ML_ERROR, "Heartbeat has to be started to mount "
835 "a read-write clustered device.\n");
836 return -EINVAL;
837 }
838 }
839
840 return 0;
841}
842
843/*
844 * If we're using a userspace stack, mount should have passed
845 * a name that matches the disk. If not, mount should not
846 * have passed a stack.
847 */
848static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
849 struct mount_options *mopt)
850{
851 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
852 mlog(ML_ERROR,
853 "cluster stack passed to mount, but this filesystem "
854 "does not support it\n");
855 return -EINVAL;
856 }
857
858 if (ocfs2_userspace_stack(osb) &&
859 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
860 OCFS2_STACK_LABEL_LEN)) {
861 mlog(ML_ERROR,
862 "cluster stack passed to mount (\"%s\") does not "
863 "match the filesystem (\"%s\")\n",
864 mopt->cluster_stack,
865 osb->osb_cluster_stack);
866 return -EINVAL;
867 }
868
869 return 0;
870}
871
872static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
873{
874 int type;
875 struct super_block *sb = osb->sb;
876 unsigned int feature[OCFS2_MAXQUOTAS] = {
877 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
878 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
879 int status = 0;
880
881 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
882 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
883 continue;
884 if (unsuspend)
885 status = dquot_resume(sb, type);
886 else {
887 struct ocfs2_mem_dqinfo *oinfo;
888
889 /* Cancel periodic syncing before suspending */
890 oinfo = sb_dqinfo(sb, type)->dqi_priv;
891 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
892 status = dquot_suspend(sb, type);
893 }
894 if (status < 0)
895 break;
896 }
897 if (status < 0)
898 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
899 "remount (error = %d).\n", status);
900 return status;
901}
902
903static int ocfs2_enable_quotas(struct ocfs2_super *osb)
904{
905 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
906 struct super_block *sb = osb->sb;
907 unsigned int feature[OCFS2_MAXQUOTAS] = {
908 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
909 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
910 unsigned int ino[OCFS2_MAXQUOTAS] = {
911 LOCAL_USER_QUOTA_SYSTEM_INODE,
912 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
913 int status;
914 int type;
915
916 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
917 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
918 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
919 continue;
920 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
921 osb->slot_num);
922 if (!inode[type]) {
923 status = -ENOENT;
924 goto out_quota_off;
925 }
926 status = dquot_load_quota_inode(inode[type], type, QFMT_OCFS2,
927 DQUOT_USAGE_ENABLED);
928 if (status < 0)
929 goto out_quota_off;
930 }
931
932 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
933 iput(inode[type]);
934 return 0;
935out_quota_off:
936 ocfs2_disable_quotas(osb);
937 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
938 iput(inode[type]);
939 mlog_errno(status);
940 return status;
941}
942
943static void ocfs2_disable_quotas(struct ocfs2_super *osb)
944{
945 int type;
946 struct inode *inode;
947 struct super_block *sb = osb->sb;
948 struct ocfs2_mem_dqinfo *oinfo;
949
950 /* We mostly ignore errors in this function because there's not much
951 * we can do when we see them */
952 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
953 if (!sb_has_quota_loaded(sb, type))
954 continue;
955 if (!sb_has_quota_suspended(sb, type)) {
956 oinfo = sb_dqinfo(sb, type)->dqi_priv;
957 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
958 }
959 inode = igrab(sb->s_dquot.files[type]);
960 /* Turn off quotas. This will remove all dquot structures from
961 * memory and so they will be automatically synced to global
962 * quota files */
963 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
964 DQUOT_LIMITS_ENABLED);
965 iput(inode);
966 }
967}
968
969static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
970{
971 struct dentry *root;
972 int status, sector_size;
973 struct mount_options parsed_options;
974 struct inode *inode = NULL;
975 struct ocfs2_super *osb = NULL;
976 struct buffer_head *bh = NULL;
977 char nodestr[12];
978 struct ocfs2_blockcheck_stats stats;
979
980 trace_ocfs2_fill_super(sb, data, silent);
981
982 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
983 status = -EINVAL;
984 goto out;
985 }
986
987 /* probe for superblock */
988 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
989 if (status < 0) {
990 mlog(ML_ERROR, "superblock probe failed!\n");
991 goto out;
992 }
993
994 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
995 brelse(bh);
996 bh = NULL;
997 if (status < 0)
998 goto out;
999
1000 osb = OCFS2_SB(sb);
1001
1002 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1003 status = -EINVAL;
1004 goto out_super;
1005 }
1006 osb->s_mount_opt = parsed_options.mount_opt;
1007 osb->s_atime_quantum = parsed_options.atime_quantum;
1008 osb->preferred_slot = parsed_options.slot;
1009 osb->osb_commit_interval = parsed_options.commit_interval;
1010
1011 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1012 osb->osb_resv_level = parsed_options.resv_level;
1013 osb->osb_dir_resv_level = parsed_options.resv_level;
1014 if (parsed_options.dir_resv_level == -1)
1015 osb->osb_dir_resv_level = parsed_options.resv_level;
1016 else
1017 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1018
1019 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1020 if (status)
1021 goto out_super;
1022
1023 sb->s_magic = OCFS2_SUPER_MAGIC;
1024
1025 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1026 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1027
1028 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1029 * heartbeat=none */
1030 if (bdev_read_only(sb->s_bdev)) {
1031 if (!sb_rdonly(sb)) {
1032 status = -EACCES;
1033 mlog(ML_ERROR, "Readonly device detected but readonly "
1034 "mount was not specified.\n");
1035 goto out_super;
1036 }
1037
1038 /* You should not be able to start a local heartbeat
1039 * on a readonly device. */
1040 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1041 status = -EROFS;
1042 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1043 "device.\n");
1044 goto out_super;
1045 }
1046
1047 status = ocfs2_check_journals_nolocks(osb);
1048 if (status < 0) {
1049 if (status == -EROFS)
1050 mlog(ML_ERROR, "Recovery required on readonly "
1051 "file system, but write access is "
1052 "unavailable.\n");
1053 goto out_super;
1054 }
1055
1056 ocfs2_set_ro_flag(osb, 1);
1057
1058 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1059 "Cluster services will not be used for this mount. "
1060 "Recovery will be skipped.\n", osb->dev_str);
1061 }
1062
1063 if (!ocfs2_is_hard_readonly(osb)) {
1064 if (sb_rdonly(sb))
1065 ocfs2_set_ro_flag(osb, 0);
1066 }
1067
1068 status = ocfs2_verify_heartbeat(osb);
1069 if (status < 0)
1070 goto out_super;
1071
1072 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1073 ocfs2_debugfs_root);
1074
1075 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1076 osb, &ocfs2_osb_debug_fops);
1077
1078 if (ocfs2_meta_ecc(osb))
1079 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1080 osb->osb_debug_root);
1081
1082 status = ocfs2_mount_volume(sb);
1083 if (status < 0)
1084 goto out_debugfs;
1085
1086 if (osb->root_inode)
1087 inode = igrab(osb->root_inode);
1088
1089 if (!inode) {
1090 status = -EIO;
1091 goto out_dismount;
1092 }
1093
1094 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1095 &ocfs2_kset->kobj);
1096 if (!osb->osb_dev_kset) {
1097 status = -ENOMEM;
1098 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1099 goto out_dismount;
1100 }
1101
1102 /* Create filecheck sysfs related directories/files at
1103 * /sys/fs/ocfs2/<devname>/filecheck */
1104 if (ocfs2_filecheck_create_sysfs(osb)) {
1105 status = -ENOMEM;
1106 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1107 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1108 goto out_dismount;
1109 }
1110
1111 root = d_make_root(inode);
1112 if (!root) {
1113 status = -ENOMEM;
1114 goto out_dismount;
1115 }
1116
1117 sb->s_root = root;
1118
1119 ocfs2_complete_mount_recovery(osb);
1120
1121 if (ocfs2_mount_local(osb))
1122 snprintf(nodestr, sizeof(nodestr), "local");
1123 else
1124 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1125
1126 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1127 "with %s data mode.\n",
1128 osb->dev_str, nodestr, osb->slot_num,
1129 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1130 "ordered");
1131
1132 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1133 wake_up(&osb->osb_mount_event);
1134
1135 /* Now we can initialize quotas because we can afford to wait
1136 * for cluster locks recovery now. That also means that truncation
1137 * log recovery can happen but that waits for proper quota setup */
1138 if (!sb_rdonly(sb)) {
1139 status = ocfs2_enable_quotas(osb);
1140 if (status < 0) {
1141 /* We have to err-out specially here because
1142 * s_root is already set */
1143 mlog_errno(status);
1144 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1145 wake_up(&osb->osb_mount_event);
1146 return status;
1147 }
1148 }
1149
1150 ocfs2_complete_quota_recovery(osb);
1151
1152 /* Now we wake up again for processes waiting for quotas */
1153 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1154 wake_up(&osb->osb_mount_event);
1155
1156 /* Start this when the mount is almost sure of being successful */
1157 ocfs2_orphan_scan_start(osb);
1158
1159 return status;
1160
1161out_dismount:
1162 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1163 wake_up(&osb->osb_mount_event);
1164 ocfs2_free_replay_slots(osb);
1165 ocfs2_dismount_volume(sb, 1);
1166 goto out;
1167
1168out_debugfs:
1169 debugfs_remove_recursive(osb->osb_debug_root);
1170out_super:
1171 ocfs2_release_system_inodes(osb);
1172 kfree(osb->recovery_map);
1173 ocfs2_delete_osb(osb);
1174 kfree(osb);
1175out:
1176 mlog_errno(status);
1177
1178 return status;
1179}
1180
1181static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1182 int flags,
1183 const char *dev_name,
1184 void *data)
1185{
1186 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1187}
1188
1189static struct file_system_type ocfs2_fs_type = {
1190 .owner = THIS_MODULE,
1191 .name = "ocfs2",
1192 .mount = ocfs2_mount,
1193 .kill_sb = kill_block_super,
1194 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1195 .next = NULL
1196};
1197MODULE_ALIAS_FS("ocfs2");
1198
1199static int ocfs2_check_set_options(struct super_block *sb,
1200 struct mount_options *options)
1201{
1202 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1203 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1204 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1205 mlog(ML_ERROR, "User quotas were requested, but this "
1206 "filesystem does not have the feature enabled.\n");
1207 return 0;
1208 }
1209 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1210 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1211 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1212 mlog(ML_ERROR, "Group quotas were requested, but this "
1213 "filesystem does not have the feature enabled.\n");
1214 return 0;
1215 }
1216 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1217 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1218 mlog(ML_ERROR, "ACL support requested but extended attributes "
1219 "feature is not enabled\n");
1220 return 0;
1221 }
1222 /* No ACL setting specified? Use XATTR feature... */
1223 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1224 OCFS2_MOUNT_NO_POSIX_ACL))) {
1225 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1226 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1227 else
1228 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1229 }
1230 return 1;
1231}
1232
1233static int ocfs2_parse_options(struct super_block *sb,
1234 char *options,
1235 struct mount_options *mopt,
1236 int is_remount)
1237{
1238 int status, user_stack = 0;
1239 char *p;
1240 u32 tmp;
1241 int token, option;
1242 substring_t args[MAX_OPT_ARGS];
1243
1244 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1245
1246 mopt->commit_interval = 0;
1247 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1248 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1249 mopt->slot = OCFS2_INVALID_SLOT;
1250 mopt->localalloc_opt = -1;
1251 mopt->cluster_stack[0] = '\0';
1252 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1253 mopt->dir_resv_level = -1;
1254
1255 if (!options) {
1256 status = 1;
1257 goto bail;
1258 }
1259
1260 while ((p = strsep(&options, ",")) != NULL) {
1261 if (!*p)
1262 continue;
1263
1264 token = match_token(p, tokens, args);
1265 switch (token) {
1266 case Opt_hb_local:
1267 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1268 break;
1269 case Opt_hb_none:
1270 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1271 break;
1272 case Opt_hb_global:
1273 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1274 break;
1275 case Opt_barrier:
1276 if (match_int(&args[0], &option)) {
1277 status = 0;
1278 goto bail;
1279 }
1280 if (option)
1281 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1282 else
1283 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1284 break;
1285 case Opt_intr:
1286 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1287 break;
1288 case Opt_nointr:
1289 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1290 break;
1291 case Opt_err_panic:
1292 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1293 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1294 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1295 break;
1296 case Opt_err_ro:
1297 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1298 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1299 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1300 break;
1301 case Opt_err_cont:
1302 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1303 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1304 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1305 break;
1306 case Opt_data_ordered:
1307 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1308 break;
1309 case Opt_data_writeback:
1310 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1311 break;
1312 case Opt_user_xattr:
1313 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1314 break;
1315 case Opt_nouser_xattr:
1316 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1317 break;
1318 case Opt_atime_quantum:
1319 if (match_int(&args[0], &option)) {
1320 status = 0;
1321 goto bail;
1322 }
1323 if (option >= 0)
1324 mopt->atime_quantum = option;
1325 break;
1326 case Opt_slot:
1327 if (match_int(&args[0], &option)) {
1328 status = 0;
1329 goto bail;
1330 }
1331 if (option)
1332 mopt->slot = (u16)option;
1333 break;
1334 case Opt_commit:
1335 if (match_int(&args[0], &option)) {
1336 status = 0;
1337 goto bail;
1338 }
1339 if (option < 0)
1340 return 0;
1341 if (option == 0)
1342 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1343 mopt->commit_interval = HZ * option;
1344 break;
1345 case Opt_localalloc:
1346 if (match_int(&args[0], &option)) {
1347 status = 0;
1348 goto bail;
1349 }
1350 if (option >= 0)
1351 mopt->localalloc_opt = option;
1352 break;
1353 case Opt_localflocks:
1354 /*
1355 * Changing this during remount could race
1356 * flock() requests, or "unbalance" existing
1357 * ones (e.g., a lock is taken in one mode but
1358 * dropped in the other). If users care enough
1359 * to flip locking modes during remount, we
1360 * could add a "local" flag to individual
1361 * flock structures for proper tracking of
1362 * state.
1363 */
1364 if (!is_remount)
1365 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1366 break;
1367 case Opt_stack:
1368 /* Check both that the option we were passed
1369 * is of the right length and that it is a proper
1370 * string of the right length.
1371 */
1372 if (((args[0].to - args[0].from) !=
1373 OCFS2_STACK_LABEL_LEN) ||
1374 (strnlen(args[0].from,
1375 OCFS2_STACK_LABEL_LEN) !=
1376 OCFS2_STACK_LABEL_LEN)) {
1377 mlog(ML_ERROR,
1378 "Invalid cluster_stack option\n");
1379 status = 0;
1380 goto bail;
1381 }
1382 memcpy(mopt->cluster_stack, args[0].from,
1383 OCFS2_STACK_LABEL_LEN);
1384 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1385 /*
1386 * Open code the memcmp here as we don't have
1387 * an osb to pass to
1388 * ocfs2_userspace_stack().
1389 */
1390 if (memcmp(mopt->cluster_stack,
1391 OCFS2_CLASSIC_CLUSTER_STACK,
1392 OCFS2_STACK_LABEL_LEN))
1393 user_stack = 1;
1394 break;
1395 case Opt_inode64:
1396 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1397 break;
1398 case Opt_usrquota:
1399 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1400 break;
1401 case Opt_grpquota:
1402 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1403 break;
1404 case Opt_coherency_buffered:
1405 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1406 break;
1407 case Opt_coherency_full:
1408 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1409 break;
1410 case Opt_acl:
1411 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1412 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1413 break;
1414 case Opt_noacl:
1415 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1416 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1417 break;
1418 case Opt_resv_level:
1419 if (is_remount)
1420 break;
1421 if (match_int(&args[0], &option)) {
1422 status = 0;
1423 goto bail;
1424 }
1425 if (option >= OCFS2_MIN_RESV_LEVEL &&
1426 option < OCFS2_MAX_RESV_LEVEL)
1427 mopt->resv_level = option;
1428 break;
1429 case Opt_dir_resv_level:
1430 if (is_remount)
1431 break;
1432 if (match_int(&args[0], &option)) {
1433 status = 0;
1434 goto bail;
1435 }
1436 if (option >= OCFS2_MIN_RESV_LEVEL &&
1437 option < OCFS2_MAX_RESV_LEVEL)
1438 mopt->dir_resv_level = option;
1439 break;
1440 case Opt_journal_async_commit:
1441 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1442 break;
1443 default:
1444 mlog(ML_ERROR,
1445 "Unrecognized mount option \"%s\" "
1446 "or missing value\n", p);
1447 status = 0;
1448 goto bail;
1449 }
1450 }
1451
1452 if (user_stack == 0) {
1453 /* Ensure only one heartbeat mode */
1454 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1455 OCFS2_MOUNT_HB_GLOBAL |
1456 OCFS2_MOUNT_HB_NONE);
1457 if (hweight32(tmp) != 1) {
1458 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1459 status = 0;
1460 goto bail;
1461 }
1462 }
1463
1464 status = 1;
1465
1466bail:
1467 return status;
1468}
1469
1470static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1471{
1472 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1473 unsigned long opts = osb->s_mount_opt;
1474 unsigned int local_alloc_megs;
1475
1476 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1477 seq_printf(s, ",_netdev");
1478 if (opts & OCFS2_MOUNT_HB_LOCAL)
1479 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1480 else
1481 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1482 } else
1483 seq_printf(s, ",%s", OCFS2_HB_NONE);
1484
1485 if (opts & OCFS2_MOUNT_NOINTR)
1486 seq_printf(s, ",nointr");
1487
1488 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1489 seq_printf(s, ",data=writeback");
1490 else
1491 seq_printf(s, ",data=ordered");
1492
1493 if (opts & OCFS2_MOUNT_BARRIER)
1494 seq_printf(s, ",barrier=1");
1495
1496 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1497 seq_printf(s, ",errors=panic");
1498 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1499 seq_printf(s, ",errors=continue");
1500 else
1501 seq_printf(s, ",errors=remount-ro");
1502
1503 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1504 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1505
1506 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1507
1508 if (osb->osb_commit_interval)
1509 seq_printf(s, ",commit=%u",
1510 (unsigned) (osb->osb_commit_interval / HZ));
1511
1512 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1513 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1514 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1515
1516 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1517 seq_printf(s, ",localflocks,");
1518
1519 if (osb->osb_cluster_stack[0])
1520 seq_show_option(s, "cluster_stack", osb->osb_cluster_stack);
1521 if (opts & OCFS2_MOUNT_USRQUOTA)
1522 seq_printf(s, ",usrquota");
1523 if (opts & OCFS2_MOUNT_GRPQUOTA)
1524 seq_printf(s, ",grpquota");
1525
1526 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1527 seq_printf(s, ",coherency=buffered");
1528 else
1529 seq_printf(s, ",coherency=full");
1530
1531 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1532 seq_printf(s, ",nouser_xattr");
1533 else
1534 seq_printf(s, ",user_xattr");
1535
1536 if (opts & OCFS2_MOUNT_INODE64)
1537 seq_printf(s, ",inode64");
1538
1539 if (opts & OCFS2_MOUNT_POSIX_ACL)
1540 seq_printf(s, ",acl");
1541 else
1542 seq_printf(s, ",noacl");
1543
1544 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1545 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1546
1547 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1548 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1549
1550 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1551 seq_printf(s, ",journal_async_commit");
1552
1553 return 0;
1554}
1555
1556static int __init ocfs2_init(void)
1557{
1558 int status;
1559
1560 status = init_ocfs2_uptodate_cache();
1561 if (status < 0)
1562 goto out1;
1563
1564 status = ocfs2_initialize_mem_caches();
1565 if (status < 0)
1566 goto out2;
1567
1568 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1569
1570 ocfs2_set_locking_protocol();
1571
1572 status = register_quota_format(&ocfs2_quota_format);
1573 if (status < 0)
1574 goto out3;
1575 status = register_filesystem(&ocfs2_fs_type);
1576 if (!status)
1577 return 0;
1578
1579 unregister_quota_format(&ocfs2_quota_format);
1580out3:
1581 debugfs_remove(ocfs2_debugfs_root);
1582 ocfs2_free_mem_caches();
1583out2:
1584 exit_ocfs2_uptodate_cache();
1585out1:
1586 mlog_errno(status);
1587 return status;
1588}
1589
1590static void __exit ocfs2_exit(void)
1591{
1592 unregister_quota_format(&ocfs2_quota_format);
1593
1594 debugfs_remove(ocfs2_debugfs_root);
1595
1596 ocfs2_free_mem_caches();
1597
1598 unregister_filesystem(&ocfs2_fs_type);
1599
1600 exit_ocfs2_uptodate_cache();
1601}
1602
1603static void ocfs2_put_super(struct super_block *sb)
1604{
1605 trace_ocfs2_put_super(sb);
1606
1607 ocfs2_sync_blockdev(sb);
1608 ocfs2_dismount_volume(sb, 0);
1609}
1610
1611static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1612{
1613 struct ocfs2_super *osb;
1614 u32 numbits, freebits;
1615 int status;
1616 struct ocfs2_dinode *bm_lock;
1617 struct buffer_head *bh = NULL;
1618 struct inode *inode = NULL;
1619
1620 trace_ocfs2_statfs(dentry->d_sb, buf);
1621
1622 osb = OCFS2_SB(dentry->d_sb);
1623
1624 inode = ocfs2_get_system_file_inode(osb,
1625 GLOBAL_BITMAP_SYSTEM_INODE,
1626 OCFS2_INVALID_SLOT);
1627 if (!inode) {
1628 mlog(ML_ERROR, "failed to get bitmap inode\n");
1629 status = -EIO;
1630 goto bail;
1631 }
1632
1633 status = ocfs2_inode_lock(inode, &bh, 0);
1634 if (status < 0) {
1635 mlog_errno(status);
1636 goto bail;
1637 }
1638
1639 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1640
1641 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1642 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1643
1644 buf->f_type = OCFS2_SUPER_MAGIC;
1645 buf->f_bsize = dentry->d_sb->s_blocksize;
1646 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1647 buf->f_blocks = ((sector_t) numbits) *
1648 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1649 buf->f_bfree = ((sector_t) freebits) *
1650 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1651 buf->f_bavail = buf->f_bfree;
1652 buf->f_files = numbits;
1653 buf->f_ffree = freebits;
1654 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1655 & 0xFFFFFFFFUL;
1656 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1657 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1658
1659 brelse(bh);
1660
1661 ocfs2_inode_unlock(inode, 0);
1662 status = 0;
1663bail:
1664 iput(inode);
1665
1666 if (status)
1667 mlog_errno(status);
1668
1669 return status;
1670}
1671
1672static void ocfs2_inode_init_once(void *data)
1673{
1674 struct ocfs2_inode_info *oi = data;
1675
1676 oi->ip_flags = 0;
1677 oi->ip_open_count = 0;
1678 spin_lock_init(&oi->ip_lock);
1679 ocfs2_extent_map_init(&oi->vfs_inode);
1680 INIT_LIST_HEAD(&oi->ip_io_markers);
1681 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1682 oi->ip_dir_start_lookup = 0;
1683 init_rwsem(&oi->ip_alloc_sem);
1684 init_rwsem(&oi->ip_xattr_sem);
1685 mutex_init(&oi->ip_io_mutex);
1686
1687 oi->ip_blkno = 0ULL;
1688 oi->ip_clusters = 0;
1689 oi->ip_next_orphan = NULL;
1690
1691 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1692
1693 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1694 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1695 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1696
1697 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1698 &ocfs2_inode_caching_ops);
1699
1700 inode_init_once(&oi->vfs_inode);
1701}
1702
1703static int ocfs2_initialize_mem_caches(void)
1704{
1705 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1706 sizeof(struct ocfs2_inode_info),
1707 0,
1708 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1709 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1710 ocfs2_inode_init_once);
1711 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1712 sizeof(struct ocfs2_dquot),
1713 0,
1714 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1715 SLAB_MEM_SPREAD),
1716 NULL);
1717 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1718 sizeof(struct ocfs2_quota_chunk),
1719 0,
1720 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1721 NULL);
1722 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1723 !ocfs2_qf_chunk_cachep) {
1724 kmem_cache_destroy(ocfs2_inode_cachep);
1725 kmem_cache_destroy(ocfs2_dquot_cachep);
1726 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1727 return -ENOMEM;
1728 }
1729
1730 return 0;
1731}
1732
1733static void ocfs2_free_mem_caches(void)
1734{
1735 /*
1736 * Make sure all delayed rcu free inodes are flushed before we
1737 * destroy cache.
1738 */
1739 rcu_barrier();
1740 kmem_cache_destroy(ocfs2_inode_cachep);
1741 ocfs2_inode_cachep = NULL;
1742
1743 kmem_cache_destroy(ocfs2_dquot_cachep);
1744 ocfs2_dquot_cachep = NULL;
1745
1746 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1747 ocfs2_qf_chunk_cachep = NULL;
1748}
1749
1750static int ocfs2_get_sector(struct super_block *sb,
1751 struct buffer_head **bh,
1752 int block,
1753 int sect_size)
1754{
1755 if (!sb_set_blocksize(sb, sect_size)) {
1756 mlog(ML_ERROR, "unable to set blocksize\n");
1757 return -EIO;
1758 }
1759
1760 *bh = sb_getblk(sb, block);
1761 if (!*bh) {
1762 mlog_errno(-ENOMEM);
1763 return -ENOMEM;
1764 }
1765 lock_buffer(*bh);
1766 if (!buffer_dirty(*bh))
1767 clear_buffer_uptodate(*bh);
1768 unlock_buffer(*bh);
1769 if (bh_read(*bh, 0) < 0) {
1770 mlog_errno(-EIO);
1771 brelse(*bh);
1772 *bh = NULL;
1773 return -EIO;
1774 }
1775
1776 return 0;
1777}
1778
1779static int ocfs2_mount_volume(struct super_block *sb)
1780{
1781 int status = 0;
1782 struct ocfs2_super *osb = OCFS2_SB(sb);
1783
1784 if (ocfs2_is_hard_readonly(osb))
1785 goto out;
1786
1787 mutex_init(&osb->obs_trim_fs_mutex);
1788
1789 status = ocfs2_dlm_init(osb);
1790 if (status < 0) {
1791 mlog_errno(status);
1792 if (status == -EBADR && ocfs2_userspace_stack(osb))
1793 mlog(ML_ERROR, "couldn't mount because cluster name on"
1794 " disk does not match the running cluster name.\n");
1795 goto out;
1796 }
1797
1798 status = ocfs2_super_lock(osb, 1);
1799 if (status < 0) {
1800 mlog_errno(status);
1801 goto out_dlm;
1802 }
1803
1804 /* This will load up the node map and add ourselves to it. */
1805 status = ocfs2_find_slot(osb);
1806 if (status < 0) {
1807 mlog_errno(status);
1808 goto out_super_lock;
1809 }
1810
1811 /* load all node-local system inodes */
1812 status = ocfs2_init_local_system_inodes(osb);
1813 if (status < 0) {
1814 mlog_errno(status);
1815 goto out_super_lock;
1816 }
1817
1818 status = ocfs2_check_volume(osb);
1819 if (status < 0) {
1820 mlog_errno(status);
1821 goto out_system_inodes;
1822 }
1823
1824 status = ocfs2_truncate_log_init(osb);
1825 if (status < 0) {
1826 mlog_errno(status);
1827 goto out_check_volume;
1828 }
1829
1830 ocfs2_super_unlock(osb, 1);
1831 return 0;
1832
1833out_check_volume:
1834 ocfs2_free_replay_slots(osb);
1835out_system_inodes:
1836 if (osb->local_alloc_state == OCFS2_LA_ENABLED)
1837 ocfs2_shutdown_local_alloc(osb);
1838 ocfs2_release_system_inodes(osb);
1839 /* before journal shutdown, we should release slot_info */
1840 ocfs2_free_slot_info(osb);
1841 ocfs2_journal_shutdown(osb);
1842out_super_lock:
1843 ocfs2_super_unlock(osb, 1);
1844out_dlm:
1845 ocfs2_dlm_shutdown(osb, 0);
1846out:
1847 return status;
1848}
1849
1850static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1851{
1852 int tmp, hangup_needed = 0;
1853 struct ocfs2_super *osb = NULL;
1854 char nodestr[12];
1855
1856 trace_ocfs2_dismount_volume(sb);
1857
1858 BUG_ON(!sb);
1859 osb = OCFS2_SB(sb);
1860 BUG_ON(!osb);
1861
1862 /* Remove file check sysfs related directores/files,
1863 * and wait for the pending file check operations */
1864 ocfs2_filecheck_remove_sysfs(osb);
1865
1866 kset_unregister(osb->osb_dev_kset);
1867
1868 /* Orphan scan should be stopped as early as possible */
1869 ocfs2_orphan_scan_stop(osb);
1870
1871 ocfs2_disable_quotas(osb);
1872
1873 /* All dquots should be freed by now */
1874 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1875 /* Wait for worker to be done with the work structure in osb */
1876 cancel_work_sync(&osb->dquot_drop_work);
1877
1878 ocfs2_shutdown_local_alloc(osb);
1879
1880 ocfs2_truncate_log_shutdown(osb);
1881
1882 /* This will disable recovery and flush any recovery work. */
1883 ocfs2_recovery_exit(osb);
1884
1885 ocfs2_sync_blockdev(sb);
1886
1887 ocfs2_purge_refcount_trees(osb);
1888
1889 /* No cluster connection means we've failed during mount, so skip
1890 * all the steps which depended on that to complete. */
1891 if (osb->cconn) {
1892 tmp = ocfs2_super_lock(osb, 1);
1893 if (tmp < 0) {
1894 mlog_errno(tmp);
1895 return;
1896 }
1897 }
1898
1899 if (osb->slot_num != OCFS2_INVALID_SLOT)
1900 ocfs2_put_slot(osb);
1901
1902 if (osb->cconn)
1903 ocfs2_super_unlock(osb, 1);
1904
1905 ocfs2_release_system_inodes(osb);
1906
1907 ocfs2_journal_shutdown(osb);
1908
1909 /*
1910 * If we're dismounting due to mount error, mount.ocfs2 will clean
1911 * up heartbeat. If we're a local mount, there is no heartbeat.
1912 * If we failed before we got a uuid_str yet, we can't stop
1913 * heartbeat. Otherwise, do it.
1914 */
1915 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1916 !ocfs2_is_hard_readonly(osb))
1917 hangup_needed = 1;
1918
1919 ocfs2_dlm_shutdown(osb, hangup_needed);
1920
1921 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1922 debugfs_remove_recursive(osb->osb_debug_root);
1923
1924 if (hangup_needed)
1925 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1926
1927 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1928
1929 if (ocfs2_mount_local(osb))
1930 snprintf(nodestr, sizeof(nodestr), "local");
1931 else
1932 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1933
1934 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1935 osb->dev_str, nodestr);
1936
1937 ocfs2_delete_osb(osb);
1938 kfree(osb);
1939 sb->s_dev = 0;
1940 sb->s_fs_info = NULL;
1941}
1942
1943static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1944 unsigned uuid_bytes)
1945{
1946 int i, ret;
1947 char *ptr;
1948
1949 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1950
1951 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1952 if (osb->uuid_str == NULL)
1953 return -ENOMEM;
1954
1955 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1956 /* print with null */
1957 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1958 if (ret != 2) /* drop super cleans up */
1959 return -EINVAL;
1960 /* then only advance past the last char */
1961 ptr += 2;
1962 }
1963
1964 return 0;
1965}
1966
1967/* Make sure entire volume is addressable by our journal. Requires
1968 osb_clusters_at_boot to be valid and for the journal to have been
1969 initialized by ocfs2_journal_init(). */
1970static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1971{
1972 int status = 0;
1973 u64 max_block =
1974 ocfs2_clusters_to_blocks(osb->sb,
1975 osb->osb_clusters_at_boot) - 1;
1976
1977 /* 32-bit block number is always OK. */
1978 if (max_block <= (u32)~0ULL)
1979 goto out;
1980
1981 /* Volume is "huge", so see if our journal is new enough to
1982 support it. */
1983 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1984 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1985 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1986 JBD2_FEATURE_INCOMPAT_64BIT))) {
1987 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1988 "Enable the 'block64' journal option with tunefs.ocfs2");
1989 status = -EFBIG;
1990 goto out;
1991 }
1992
1993 out:
1994 return status;
1995}
1996
1997static int ocfs2_initialize_super(struct super_block *sb,
1998 struct buffer_head *bh,
1999 int sector_size,
2000 struct ocfs2_blockcheck_stats *stats)
2001{
2002 int status;
2003 int i, cbits, bbits;
2004 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2005 struct inode *inode = NULL;
2006 struct ocfs2_super *osb;
2007 u64 total_blocks;
2008
2009 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2010 if (!osb) {
2011 status = -ENOMEM;
2012 mlog_errno(status);
2013 goto out;
2014 }
2015
2016 sb->s_fs_info = osb;
2017 sb->s_op = &ocfs2_sops;
2018 sb->s_d_op = &ocfs2_dentry_ops;
2019 sb->s_export_op = &ocfs2_export_ops;
2020 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2021 sb->dq_op = &ocfs2_quota_operations;
2022 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2023 sb->s_xattr = ocfs2_xattr_handlers;
2024 sb->s_time_gran = 1;
2025 sb->s_flags |= SB_NOATIME;
2026 /* this is needed to support O_LARGEFILE */
2027 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2028 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2029 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2030 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2031 sizeof(di->id2.i_super.s_uuid));
2032
2033 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2034
2035 for (i = 0; i < 3; i++)
2036 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2037 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2038
2039 osb->sb = sb;
2040 osb->s_sectsize_bits = blksize_bits(sector_size);
2041 BUG_ON(!osb->s_sectsize_bits);
2042
2043 spin_lock_init(&osb->dc_task_lock);
2044 init_waitqueue_head(&osb->dc_event);
2045 osb->dc_work_sequence = 0;
2046 osb->dc_wake_sequence = 0;
2047 INIT_LIST_HEAD(&osb->blocked_lock_list);
2048 osb->blocked_lock_count = 0;
2049 spin_lock_init(&osb->osb_lock);
2050 spin_lock_init(&osb->osb_xattr_lock);
2051 ocfs2_init_steal_slots(osb);
2052
2053 mutex_init(&osb->system_file_mutex);
2054
2055 atomic_set(&osb->alloc_stats.moves, 0);
2056 atomic_set(&osb->alloc_stats.local_data, 0);
2057 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2058 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2059 atomic_set(&osb->alloc_stats.bg_extends, 0);
2060
2061 /* Copy the blockcheck stats from the superblock probe */
2062 osb->osb_ecc_stats = *stats;
2063
2064 ocfs2_init_node_maps(osb);
2065
2066 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2067 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2068
2069 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2070 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2071 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2072 osb->max_slots);
2073 status = -EINVAL;
2074 goto out;
2075 }
2076
2077 ocfs2_orphan_scan_init(osb);
2078
2079 status = ocfs2_recovery_init(osb);
2080 if (status) {
2081 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2082 mlog_errno(status);
2083 goto out;
2084 }
2085
2086 init_waitqueue_head(&osb->checkpoint_event);
2087
2088 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2089
2090 osb->slot_num = OCFS2_INVALID_SLOT;
2091
2092 osb->s_xattr_inline_size = le16_to_cpu(
2093 di->id2.i_super.s_xattr_inline_size);
2094
2095 osb->local_alloc_state = OCFS2_LA_UNUSED;
2096 osb->local_alloc_bh = NULL;
2097 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2098
2099 init_waitqueue_head(&osb->osb_mount_event);
2100
2101 ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2102
2103 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2104 if (!osb->vol_label) {
2105 mlog(ML_ERROR, "unable to alloc vol label\n");
2106 status = -ENOMEM;
2107 goto out_recovery_map;
2108 }
2109
2110 osb->slot_recovery_generations =
2111 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2112 GFP_KERNEL);
2113 if (!osb->slot_recovery_generations) {
2114 status = -ENOMEM;
2115 mlog_errno(status);
2116 goto out_vol_label;
2117 }
2118
2119 init_waitqueue_head(&osb->osb_wipe_event);
2120 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2121 sizeof(*osb->osb_orphan_wipes),
2122 GFP_KERNEL);
2123 if (!osb->osb_orphan_wipes) {
2124 status = -ENOMEM;
2125 mlog_errno(status);
2126 goto out_slot_recovery_gen;
2127 }
2128
2129 osb->osb_rf_lock_tree = RB_ROOT;
2130
2131 osb->s_feature_compat =
2132 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2133 osb->s_feature_ro_compat =
2134 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2135 osb->s_feature_incompat =
2136 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2137
2138 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2139 mlog(ML_ERROR, "couldn't mount because of unsupported "
2140 "optional features (%x).\n", i);
2141 status = -EINVAL;
2142 goto out_orphan_wipes;
2143 }
2144 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2145 mlog(ML_ERROR, "couldn't mount RDWR because of "
2146 "unsupported optional features (%x).\n", i);
2147 status = -EINVAL;
2148 goto out_orphan_wipes;
2149 }
2150
2151 if (ocfs2_clusterinfo_valid(osb)) {
2152 /*
2153 * ci_stack and ci_cluster in ocfs2_cluster_info may not be null
2154 * terminated, so make sure no overflow happens here by using
2155 * memcpy. Destination strings will always be null terminated
2156 * because osb is allocated using kzalloc.
2157 */
2158 osb->osb_stackflags =
2159 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2160 memcpy(osb->osb_cluster_stack,
2161 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2162 OCFS2_STACK_LABEL_LEN);
2163 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2164 mlog(ML_ERROR,
2165 "couldn't mount because of an invalid "
2166 "cluster stack label (%s) \n",
2167 osb->osb_cluster_stack);
2168 status = -EINVAL;
2169 goto out_orphan_wipes;
2170 }
2171 memcpy(osb->osb_cluster_name,
2172 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2173 OCFS2_CLUSTER_NAME_LEN);
2174 } else {
2175 /* The empty string is identical with classic tools that
2176 * don't know about s_cluster_info. */
2177 osb->osb_cluster_stack[0] = '\0';
2178 }
2179
2180 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2181
2182 /*
2183 * FIXME
2184 * This should be done in ocfs2_journal_init(), but any inode
2185 * writes back operation will cause the filesystem to crash.
2186 */
2187 status = ocfs2_journal_alloc(osb);
2188 if (status < 0)
2189 goto out_orphan_wipes;
2190
2191 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2192 init_llist_head(&osb->dquot_drop_list);
2193
2194 /* get some pseudo constants for clustersize bits */
2195 osb->s_clustersize_bits =
2196 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2197 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2198
2199 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2200 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2201 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2202 osb->s_clustersize);
2203 status = -EINVAL;
2204 goto out_journal;
2205 }
2206
2207 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2208 le32_to_cpu(di->i_clusters));
2209
2210 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2211 total_blocks);
2212 if (status) {
2213 mlog(ML_ERROR, "Volume too large "
2214 "to mount safely on this system");
2215 status = -EFBIG;
2216 goto out_journal;
2217 }
2218
2219 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2220 sizeof(di->id2.i_super.s_uuid))) {
2221 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2222 status = -ENOMEM;
2223 goto out_journal;
2224 }
2225
2226 strscpy(osb->vol_label, di->id2.i_super.s_label,
2227 OCFS2_MAX_VOL_LABEL_LEN);
2228 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2229 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2230 osb->first_cluster_group_blkno =
2231 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2232 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2233 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2234 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2235 (unsigned long long)osb->root_blkno,
2236 (unsigned long long)osb->system_dir_blkno,
2237 osb->s_clustersize_bits);
2238
2239 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2240 if (!osb->osb_dlm_debug) {
2241 status = -ENOMEM;
2242 mlog_errno(status);
2243 goto out_uuid_str;
2244 }
2245
2246 atomic_set(&osb->vol_state, VOLUME_INIT);
2247
2248 /* load root, system_dir, and all global system inodes */
2249 status = ocfs2_init_global_system_inodes(osb);
2250 if (status < 0) {
2251 mlog_errno(status);
2252 goto out_dlm_out;
2253 }
2254
2255 /*
2256 * global bitmap
2257 */
2258 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2259 OCFS2_INVALID_SLOT);
2260 if (!inode) {
2261 status = -EINVAL;
2262 mlog_errno(status);
2263 goto out_system_inodes;
2264 }
2265
2266 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2267 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2268 iput(inode);
2269
2270 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2271 osb->s_feature_incompat) * 8;
2272
2273 status = ocfs2_init_slot_info(osb);
2274 if (status < 0) {
2275 mlog_errno(status);
2276 goto out_system_inodes;
2277 }
2278
2279 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2280 if (!osb->ocfs2_wq) {
2281 status = -ENOMEM;
2282 mlog_errno(status);
2283 goto out_slot_info;
2284 }
2285
2286 return status;
2287
2288out_slot_info:
2289 ocfs2_free_slot_info(osb);
2290out_system_inodes:
2291 ocfs2_release_system_inodes(osb);
2292out_dlm_out:
2293 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2294out_uuid_str:
2295 kfree(osb->uuid_str);
2296out_journal:
2297 kfree(osb->journal);
2298out_orphan_wipes:
2299 kfree(osb->osb_orphan_wipes);
2300out_slot_recovery_gen:
2301 kfree(osb->slot_recovery_generations);
2302out_vol_label:
2303 kfree(osb->vol_label);
2304out_recovery_map:
2305 kfree(osb->recovery_map);
2306out:
2307 kfree(osb);
2308 sb->s_fs_info = NULL;
2309 return status;
2310}
2311
2312/*
2313 * will return: -EAGAIN if it is ok to keep searching for superblocks
2314 * -EINVAL if there is a bad superblock
2315 * 0 on success
2316 */
2317static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2318 struct buffer_head *bh,
2319 u32 blksz,
2320 struct ocfs2_blockcheck_stats *stats)
2321{
2322 int status = -EAGAIN;
2323
2324 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2325 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2326 /* We have to do a raw check of the feature here */
2327 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2328 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2329 status = ocfs2_block_check_validate(bh->b_data,
2330 bh->b_size,
2331 &di->i_check,
2332 stats);
2333 if (status)
2334 goto out;
2335 }
2336 status = -EINVAL;
2337 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2338 mlog(ML_ERROR, "found superblock with incorrect block "
2339 "size: found %u, should be %u\n",
2340 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2341 blksz);
2342 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2343 OCFS2_MAJOR_REV_LEVEL ||
2344 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2345 OCFS2_MINOR_REV_LEVEL) {
2346 mlog(ML_ERROR, "found superblock with bad version: "
2347 "found %u.%u, should be %u.%u\n",
2348 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2349 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2350 OCFS2_MAJOR_REV_LEVEL,
2351 OCFS2_MINOR_REV_LEVEL);
2352 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2353 mlog(ML_ERROR, "bad block number on superblock: "
2354 "found %llu, should be %llu\n",
2355 (unsigned long long)le64_to_cpu(di->i_blkno),
2356 (unsigned long long)bh->b_blocknr);
2357 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2358 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2359 mlog(ML_ERROR, "bad cluster size found: %u\n",
2360 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2361 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2362 mlog(ML_ERROR, "bad root_blkno: 0\n");
2363 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2364 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2365 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2366 mlog(ML_ERROR,
2367 "Superblock slots found greater than file system "
2368 "maximum: found %u, max %u\n",
2369 le16_to_cpu(di->id2.i_super.s_max_slots),
2370 OCFS2_MAX_SLOTS);
2371 } else {
2372 /* found it! */
2373 status = 0;
2374 }
2375 }
2376
2377out:
2378 if (status && status != -EAGAIN)
2379 mlog_errno(status);
2380 return status;
2381}
2382
2383static int ocfs2_check_volume(struct ocfs2_super *osb)
2384{
2385 int status;
2386 int dirty;
2387 int local;
2388 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2389 * recover
2390 * ourselves. */
2391
2392 /* Init our journal object. */
2393 status = ocfs2_journal_init(osb, &dirty);
2394 if (status < 0) {
2395 mlog(ML_ERROR, "Could not initialize journal!\n");
2396 goto finally;
2397 }
2398
2399 /* Now that journal has been initialized, check to make sure
2400 entire volume is addressable. */
2401 status = ocfs2_journal_addressable(osb);
2402 if (status)
2403 goto finally;
2404
2405 /* If the journal was unmounted cleanly then we don't want to
2406 * recover anything. Otherwise, journal_load will do that
2407 * dirty work for us :) */
2408 if (!dirty) {
2409 status = ocfs2_journal_wipe(osb->journal, 0);
2410 if (status < 0) {
2411 mlog_errno(status);
2412 goto finally;
2413 }
2414 } else {
2415 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2416 "unmounted cleanly, recovering it.\n", osb->dev_str);
2417 }
2418
2419 local = ocfs2_mount_local(osb);
2420
2421 /* will play back anything left in the journal. */
2422 status = ocfs2_journal_load(osb->journal, local, dirty);
2423 if (status < 0) {
2424 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2425 goto finally;
2426 }
2427
2428 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2429 jbd2_journal_set_features(osb->journal->j_journal,
2430 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2431 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2432 else
2433 jbd2_journal_clear_features(osb->journal->j_journal,
2434 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2435 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2436
2437 if (dirty) {
2438 /* recover my local alloc if we didn't unmount cleanly. */
2439 status = ocfs2_begin_local_alloc_recovery(osb,
2440 osb->slot_num,
2441 &local_alloc);
2442 if (status < 0) {
2443 mlog_errno(status);
2444 goto finally;
2445 }
2446 /* we complete the recovery process after we've marked
2447 * ourselves as mounted. */
2448 }
2449
2450 status = ocfs2_load_local_alloc(osb);
2451 if (status < 0) {
2452 mlog_errno(status);
2453 goto finally;
2454 }
2455
2456 if (dirty) {
2457 /* Recovery will be completed after we've mounted the
2458 * rest of the volume. */
2459 osb->local_alloc_copy = local_alloc;
2460 local_alloc = NULL;
2461 }
2462
2463 /* go through each journal, trylock it and if you get the
2464 * lock, and it's marked as dirty, set the bit in the recover
2465 * map and launch a recovery thread for it. */
2466 status = ocfs2_mark_dead_nodes(osb);
2467 if (status < 0) {
2468 mlog_errno(status);
2469 goto finally;
2470 }
2471
2472 status = ocfs2_compute_replay_slots(osb);
2473 if (status < 0)
2474 mlog_errno(status);
2475
2476finally:
2477 kfree(local_alloc);
2478
2479 if (status)
2480 mlog_errno(status);
2481 return status;
2482}
2483
2484/*
2485 * The routine gets called from dismount or close whenever a dismount on
2486 * volume is requested and the osb open count becomes 1.
2487 * It will remove the osb from the global list and also free up all the
2488 * initialized resources and fileobject.
2489 */
2490static void ocfs2_delete_osb(struct ocfs2_super *osb)
2491{
2492 /* This function assumes that the caller has the main osb resource */
2493
2494 /* ocfs2_initializer_super have already created this workqueue */
2495 if (osb->ocfs2_wq)
2496 destroy_workqueue(osb->ocfs2_wq);
2497
2498 ocfs2_free_slot_info(osb);
2499
2500 kfree(osb->osb_orphan_wipes);
2501 kfree(osb->slot_recovery_generations);
2502 /* FIXME
2503 * This belongs in journal shutdown, but because we have to
2504 * allocate osb->journal at the middle of ocfs2_initialize_super(),
2505 * we free it here.
2506 */
2507 kfree(osb->journal);
2508 kfree(osb->local_alloc_copy);
2509 kfree(osb->uuid_str);
2510 kfree(osb->vol_label);
2511 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2512 memset(osb, 0, sizeof(struct ocfs2_super));
2513}
2514
2515/* Depending on the mount option passed, perform one of the following:
2516 * Put OCFS2 into a readonly state (default)
2517 * Return EIO so that only the process errs
2518 * Fix the error as if fsck.ocfs2 -y
2519 * panic
2520 */
2521static int ocfs2_handle_error(struct super_block *sb)
2522{
2523 struct ocfs2_super *osb = OCFS2_SB(sb);
2524 int rv = 0;
2525
2526 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2527 pr_crit("On-disk corruption discovered. "
2528 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2529
2530 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2531 panic("OCFS2: (device %s): panic forced after error\n",
2532 sb->s_id);
2533 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2534 pr_crit("OCFS2: Returning error to the calling process.\n");
2535 rv = -EIO;
2536 } else { /* default option */
2537 rv = -EROFS;
2538 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2539 return rv;
2540
2541 pr_crit("OCFS2: File system is now read-only.\n");
2542 sb->s_flags |= SB_RDONLY;
2543 ocfs2_set_ro_flag(osb, 0);
2544 }
2545
2546 return rv;
2547}
2548
2549int __ocfs2_error(struct super_block *sb, const char *function,
2550 const char *fmt, ...)
2551{
2552 struct va_format vaf;
2553 va_list args;
2554
2555 va_start(args, fmt);
2556 vaf.fmt = fmt;
2557 vaf.va = &args;
2558
2559 /* Not using mlog here because we want to show the actual
2560 * function the error came from. */
2561 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2562 sb->s_id, function, &vaf);
2563
2564 va_end(args);
2565
2566 return ocfs2_handle_error(sb);
2567}
2568
2569/* Handle critical errors. This is intentionally more drastic than
2570 * ocfs2_handle_error, so we only use for things like journal errors,
2571 * etc. */
2572void __ocfs2_abort(struct super_block *sb, const char *function,
2573 const char *fmt, ...)
2574{
2575 struct va_format vaf;
2576 va_list args;
2577
2578 va_start(args, fmt);
2579
2580 vaf.fmt = fmt;
2581 vaf.va = &args;
2582
2583 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2584 sb->s_id, function, &vaf);
2585
2586 va_end(args);
2587
2588 /* We don't have the cluster support yet to go straight to
2589 * hard readonly in here. Until then, we want to keep
2590 * ocfs2_abort() so that we can at least mark critical
2591 * errors.
2592 *
2593 * TODO: This should abort the journal and alert other nodes
2594 * that our slot needs recovery. */
2595
2596 /* Force a panic(). This stinks, but it's better than letting
2597 * things continue without having a proper hard readonly
2598 * here. */
2599 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2600 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2601 ocfs2_handle_error(sb);
2602}
2603
2604/*
2605 * Void signal blockers, because in-kernel sigprocmask() only fails
2606 * when SIG_* is wrong.
2607 */
2608void ocfs2_block_signals(sigset_t *oldset)
2609{
2610 int rc;
2611 sigset_t blocked;
2612
2613 sigfillset(&blocked);
2614 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2615 BUG_ON(rc);
2616}
2617
2618void ocfs2_unblock_signals(sigset_t *oldset)
2619{
2620 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2621 BUG_ON(rc);
2622}
2623
2624module_init(ocfs2_init);
2625module_exit(ocfs2_exit);
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
46#define CREATE_TRACE_POINTS
47#include "ocfs2_trace.h"
48
49#include <cluster/masklog.h>
50
51#include "ocfs2.h"
52
53/* this should be the only file to include a version 1 header */
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 "xattr.h"
72#include "quota.h"
73#include "refcounttree.h"
74#include "suballoc.h"
75
76#include "buffer_head_io.h"
77
78static struct kmem_cache *ocfs2_inode_cachep = NULL;
79struct kmem_cache *ocfs2_dquot_cachep;
80struct kmem_cache *ocfs2_qf_chunk_cachep;
81
82/* OCFS2 needs to schedule several different types of work which
83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86struct workqueue_struct *ocfs2_wq = NULL;
87
88static struct dentry *ocfs2_debugfs_root = NULL;
89
90MODULE_AUTHOR("Oracle");
91MODULE_LICENSE("GPL");
92MODULE_DESCRIPTION("OCFS2 cluster file system");
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\n", "Commit",
290 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 osb->osb_commit_interval);
292
293 out += snprintf(buf + out, len - out,
294 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
295 "Journal", osb->journal->j_state,
296 osb->journal->j_trans_id,
297 atomic_read(&osb->journal->j_num_trans));
298
299 out += snprintf(buf + out, len - out,
300 "%10s => GlobalAllocs: %d LocalAllocs: %d "
301 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
302 "Stats",
303 atomic_read(&osb->alloc_stats.bitmap_data),
304 atomic_read(&osb->alloc_stats.local_data),
305 atomic_read(&osb->alloc_stats.bg_allocs),
306 atomic_read(&osb->alloc_stats.moves),
307 atomic_read(&osb->alloc_stats.bg_extends));
308
309 out += snprintf(buf + out, len - out,
310 "%10s => State: %u Descriptor: %llu Size: %u bits "
311 "Default: %u bits\n",
312 "LocalAlloc", osb->local_alloc_state,
313 (unsigned long long)osb->la_last_gd,
314 osb->local_alloc_bits, osb->local_alloc_default_bits);
315
316 spin_lock(&osb->osb_lock);
317 out += snprintf(buf + out, len - out,
318 "%10s => InodeSlot: %d StolenInodes: %d, "
319 "MetaSlot: %d StolenMeta: %d\n", "Steal",
320 osb->s_inode_steal_slot,
321 atomic_read(&osb->s_num_inodes_stolen),
322 osb->s_meta_steal_slot,
323 atomic_read(&osb->s_num_meta_stolen));
324 spin_unlock(&osb->osb_lock);
325
326 out += snprintf(buf + out, len - out, "OrphanScan => ");
327 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
328 os->os_count, os->os_seqno);
329 out += snprintf(buf + out, len - out, " Last Scan: ");
330 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
331 out += snprintf(buf + out, len - out, "Disabled\n");
332 else
333 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
334 (get_seconds() - os->os_scantime.tv_sec));
335
336 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
337 "Slots", "Num", "RecoGen");
338 for (i = 0; i < osb->max_slots; ++i) {
339 out += snprintf(buf + out, len - out,
340 "%10s %c %3d %10d\n",
341 " ",
342 (i == osb->slot_num ? '*' : ' '),
343 i, osb->slot_recovery_generations[i]);
344 }
345
346 return out;
347}
348
349static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
350{
351 struct ocfs2_super *osb = inode->i_private;
352 char *buf = NULL;
353
354 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
355 if (!buf)
356 goto bail;
357
358 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
359
360 file->private_data = buf;
361
362 return 0;
363bail:
364 return -ENOMEM;
365}
366
367static int ocfs2_debug_release(struct inode *inode, struct file *file)
368{
369 kfree(file->private_data);
370 return 0;
371}
372
373static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
374 size_t nbytes, loff_t *ppos)
375{
376 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
377 i_size_read(file->f_mapping->host));
378}
379#else
380static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
381{
382 return 0;
383}
384static int ocfs2_debug_release(struct inode *inode, struct file *file)
385{
386 return 0;
387}
388static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
389 size_t nbytes, loff_t *ppos)
390{
391 return 0;
392}
393#endif /* CONFIG_DEBUG_FS */
394
395static const struct file_operations ocfs2_osb_debug_fops = {
396 .open = ocfs2_osb_debug_open,
397 .release = ocfs2_debug_release,
398 .read = ocfs2_debug_read,
399 .llseek = generic_file_llseek,
400};
401
402static int ocfs2_sync_fs(struct super_block *sb, int wait)
403{
404 int status;
405 tid_t target;
406 struct ocfs2_super *osb = OCFS2_SB(sb);
407
408 if (ocfs2_is_hard_readonly(osb))
409 return -EROFS;
410
411 if (wait) {
412 status = ocfs2_flush_truncate_log(osb);
413 if (status < 0)
414 mlog_errno(status);
415 } else {
416 ocfs2_schedule_truncate_log_flush(osb, 0);
417 }
418
419 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
420 &target)) {
421 if (wait)
422 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
423 target);
424 }
425 return 0;
426}
427
428static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
429{
430 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
431 && (ino == USER_QUOTA_SYSTEM_INODE
432 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
433 return 0;
434 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
435 && (ino == GROUP_QUOTA_SYSTEM_INODE
436 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
437 return 0;
438 return 1;
439}
440
441static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
442{
443 struct inode *new = NULL;
444 int status = 0;
445 int i;
446
447 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
448 if (IS_ERR(new)) {
449 status = PTR_ERR(new);
450 mlog_errno(status);
451 goto bail;
452 }
453 osb->root_inode = new;
454
455 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
456 if (IS_ERR(new)) {
457 status = PTR_ERR(new);
458 mlog_errno(status);
459 goto bail;
460 }
461 osb->sys_root_inode = new;
462
463 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
464 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
465 if (!ocfs2_need_system_inode(osb, i))
466 continue;
467 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
468 if (!new) {
469 ocfs2_release_system_inodes(osb);
470 status = -EINVAL;
471 mlog_errno(status);
472 /* FIXME: Should ERROR_RO_FS */
473 mlog(ML_ERROR, "Unable to load system inode %d, "
474 "possibly corrupt fs?", i);
475 goto bail;
476 }
477 // the array now has one ref, so drop this one
478 iput(new);
479 }
480
481bail:
482 if (status)
483 mlog_errno(status);
484 return status;
485}
486
487static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
488{
489 struct inode *new = NULL;
490 int status = 0;
491 int i;
492
493 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
494 i < NUM_SYSTEM_INODES;
495 i++) {
496 if (!ocfs2_need_system_inode(osb, i))
497 continue;
498 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
499 if (!new) {
500 ocfs2_release_system_inodes(osb);
501 status = -EINVAL;
502 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
503 status, i, osb->slot_num);
504 goto bail;
505 }
506 /* the array now has one ref, so drop this one */
507 iput(new);
508 }
509
510bail:
511 if (status)
512 mlog_errno(status);
513 return status;
514}
515
516static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
517{
518 int i;
519 struct inode *inode;
520
521 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
522 inode = osb->global_system_inodes[i];
523 if (inode) {
524 iput(inode);
525 osb->global_system_inodes[i] = NULL;
526 }
527 }
528
529 inode = osb->sys_root_inode;
530 if (inode) {
531 iput(inode);
532 osb->sys_root_inode = NULL;
533 }
534
535 inode = osb->root_inode;
536 if (inode) {
537 iput(inode);
538 osb->root_inode = NULL;
539 }
540
541 if (!osb->local_system_inodes)
542 return;
543
544 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
545 if (osb->local_system_inodes[i]) {
546 iput(osb->local_system_inodes[i]);
547 osb->local_system_inodes[i] = NULL;
548 }
549 }
550
551 kfree(osb->local_system_inodes);
552 osb->local_system_inodes = NULL;
553}
554
555/* We're allocating fs objects, use GFP_NOFS */
556static struct inode *ocfs2_alloc_inode(struct super_block *sb)
557{
558 struct ocfs2_inode_info *oi;
559
560 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
561 if (!oi)
562 return NULL;
563
564 oi->i_sync_tid = 0;
565 oi->i_datasync_tid = 0;
566
567 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
568 return &oi->vfs_inode;
569}
570
571static void ocfs2_i_callback(struct rcu_head *head)
572{
573 struct inode *inode = container_of(head, struct inode, i_rcu);
574 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
575}
576
577static void ocfs2_destroy_inode(struct inode *inode)
578{
579 call_rcu(&inode->i_rcu, ocfs2_i_callback);
580}
581
582static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
583 unsigned int cbits)
584{
585 unsigned int bytes = 1 << cbits;
586 unsigned int trim = bytes;
587 unsigned int bitshift = 32;
588
589 /*
590 * i_size and all block offsets in ocfs2 are always 64 bits
591 * wide. i_clusters is 32 bits, in cluster-sized units. So on
592 * 64 bit platforms, cluster size will be the limiting factor.
593 */
594
595#if BITS_PER_LONG == 32
596# if defined(CONFIG_LBDAF)
597 BUILD_BUG_ON(sizeof(sector_t) != 8);
598 /*
599 * We might be limited by page cache size.
600 */
601 if (bytes > PAGE_CACHE_SIZE) {
602 bytes = PAGE_CACHE_SIZE;
603 trim = 1;
604 /*
605 * Shift by 31 here so that we don't get larger than
606 * MAX_LFS_FILESIZE
607 */
608 bitshift = 31;
609 }
610# else
611 /*
612 * We are limited by the size of sector_t. Use block size, as
613 * that's what we expose to the VFS.
614 */
615 bytes = 1 << bbits;
616 trim = 1;
617 bitshift = 31;
618# endif
619#endif
620
621 /*
622 * Trim by a whole cluster when we can actually approach the
623 * on-disk limits. Otherwise we can overflow i_clusters when
624 * an extent start is at the max offset.
625 */
626 return (((unsigned long long)bytes) << bitshift) - trim;
627}
628
629static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
630{
631 int incompat_features;
632 int ret = 0;
633 struct mount_options parsed_options;
634 struct ocfs2_super *osb = OCFS2_SB(sb);
635 u32 tmp;
636
637 sync_filesystem(sb);
638
639 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
640 !ocfs2_check_set_options(sb, &parsed_options)) {
641 ret = -EINVAL;
642 goto out;
643 }
644
645 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
646 OCFS2_MOUNT_HB_NONE;
647 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
648 ret = -EINVAL;
649 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
650 goto out;
651 }
652
653 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
654 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
655 ret = -EINVAL;
656 mlog(ML_ERROR, "Cannot change data mode on remount\n");
657 goto out;
658 }
659
660 /* Probably don't want this on remount; it might
661 * mess with other nodes */
662 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
663 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
664 ret = -EINVAL;
665 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
666 goto out;
667 }
668
669 /* We're going to/from readonly mode. */
670 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
671 /* Disable quota accounting before remounting RO */
672 if (*flags & MS_RDONLY) {
673 ret = ocfs2_susp_quotas(osb, 0);
674 if (ret < 0)
675 goto out;
676 }
677 /* Lock here so the check of HARD_RO and the potential
678 * setting of SOFT_RO is atomic. */
679 spin_lock(&osb->osb_lock);
680 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
681 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
682 ret = -EROFS;
683 goto unlock_osb;
684 }
685
686 if (*flags & MS_RDONLY) {
687 sb->s_flags |= MS_RDONLY;
688 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
689 } else {
690 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
691 mlog(ML_ERROR, "Cannot remount RDWR "
692 "filesystem due to previous errors.\n");
693 ret = -EROFS;
694 goto unlock_osb;
695 }
696 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
697 if (incompat_features) {
698 mlog(ML_ERROR, "Cannot remount RDWR because "
699 "of unsupported optional features "
700 "(%x).\n", incompat_features);
701 ret = -EINVAL;
702 goto unlock_osb;
703 }
704 sb->s_flags &= ~MS_RDONLY;
705 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
706 }
707 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
708unlock_osb:
709 spin_unlock(&osb->osb_lock);
710 /* Enable quota accounting after remounting RW */
711 if (!ret && !(*flags & MS_RDONLY)) {
712 if (sb_any_quota_suspended(sb))
713 ret = ocfs2_susp_quotas(osb, 1);
714 else
715 ret = ocfs2_enable_quotas(osb);
716 if (ret < 0) {
717 /* Return back changes... */
718 spin_lock(&osb->osb_lock);
719 sb->s_flags |= MS_RDONLY;
720 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
721 spin_unlock(&osb->osb_lock);
722 goto out;
723 }
724 }
725 }
726
727 if (!ret) {
728 /* Only save off the new mount options in case of a successful
729 * remount. */
730 osb->s_mount_opt = parsed_options.mount_opt;
731 osb->s_atime_quantum = parsed_options.atime_quantum;
732 osb->preferred_slot = parsed_options.slot;
733 if (parsed_options.commit_interval)
734 osb->osb_commit_interval = parsed_options.commit_interval;
735
736 if (!ocfs2_is_hard_readonly(osb))
737 ocfs2_set_journal_params(osb);
738
739 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
740 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
741 MS_POSIXACL : 0);
742 }
743out:
744 return ret;
745}
746
747static int ocfs2_sb_probe(struct super_block *sb,
748 struct buffer_head **bh,
749 int *sector_size,
750 struct ocfs2_blockcheck_stats *stats)
751{
752 int status, tmpstat;
753 struct ocfs1_vol_disk_hdr *hdr;
754 struct ocfs2_dinode *di;
755 int blksize;
756
757 *bh = NULL;
758
759 /* may be > 512 */
760 *sector_size = bdev_logical_block_size(sb->s_bdev);
761 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
762 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
763 *sector_size, OCFS2_MAX_BLOCKSIZE);
764 status = -EINVAL;
765 goto bail;
766 }
767
768 /* Can this really happen? */
769 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
770 *sector_size = OCFS2_MIN_BLOCKSIZE;
771
772 /* check block zero for old format */
773 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
774 if (status < 0) {
775 mlog_errno(status);
776 goto bail;
777 }
778 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
779 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
780 mlog(ML_ERROR, "incompatible version: %u.%u\n",
781 hdr->major_version, hdr->minor_version);
782 status = -EINVAL;
783 }
784 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
785 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
786 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
787 hdr->signature);
788 status = -EINVAL;
789 }
790 brelse(*bh);
791 *bh = NULL;
792 if (status < 0) {
793 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
794 "upgraded before mounting with ocfs v2\n");
795 goto bail;
796 }
797
798 /*
799 * Now check at magic offset for 512, 1024, 2048, 4096
800 * blocksizes. 4096 is the maximum blocksize because it is
801 * the minimum clustersize.
802 */
803 status = -EINVAL;
804 for (blksize = *sector_size;
805 blksize <= OCFS2_MAX_BLOCKSIZE;
806 blksize <<= 1) {
807 tmpstat = ocfs2_get_sector(sb, bh,
808 OCFS2_SUPER_BLOCK_BLKNO,
809 blksize);
810 if (tmpstat < 0) {
811 status = tmpstat;
812 mlog_errno(status);
813 break;
814 }
815 di = (struct ocfs2_dinode *) (*bh)->b_data;
816 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
817 spin_lock_init(&stats->b_lock);
818 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
819 if (tmpstat < 0) {
820 brelse(*bh);
821 *bh = NULL;
822 }
823 if (tmpstat != -EAGAIN) {
824 status = tmpstat;
825 break;
826 }
827 }
828
829bail:
830 return status;
831}
832
833static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
834{
835 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
836
837 if (osb->s_mount_opt & hb_enabled) {
838 if (ocfs2_mount_local(osb)) {
839 mlog(ML_ERROR, "Cannot heartbeat on a locally "
840 "mounted device.\n");
841 return -EINVAL;
842 }
843 if (ocfs2_userspace_stack(osb)) {
844 mlog(ML_ERROR, "Userspace stack expected, but "
845 "o2cb heartbeat arguments passed to mount\n");
846 return -EINVAL;
847 }
848 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
849 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
850 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
851 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
852 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
853 return -EINVAL;
854 }
855 }
856
857 if (!(osb->s_mount_opt & hb_enabled)) {
858 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
859 !ocfs2_userspace_stack(osb)) {
860 mlog(ML_ERROR, "Heartbeat has to be started to mount "
861 "a read-write clustered device.\n");
862 return -EINVAL;
863 }
864 }
865
866 return 0;
867}
868
869/*
870 * If we're using a userspace stack, mount should have passed
871 * a name that matches the disk. If not, mount should not
872 * have passed a stack.
873 */
874static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
875 struct mount_options *mopt)
876{
877 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
878 mlog(ML_ERROR,
879 "cluster stack passed to mount, but this filesystem "
880 "does not support it\n");
881 return -EINVAL;
882 }
883
884 if (ocfs2_userspace_stack(osb) &&
885 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
886 OCFS2_STACK_LABEL_LEN)) {
887 mlog(ML_ERROR,
888 "cluster stack passed to mount (\"%s\") does not "
889 "match the filesystem (\"%s\")\n",
890 mopt->cluster_stack,
891 osb->osb_cluster_stack);
892 return -EINVAL;
893 }
894
895 return 0;
896}
897
898static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
899{
900 int type;
901 struct super_block *sb = osb->sb;
902 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
903 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
904 int status = 0;
905
906 for (type = 0; type < MAXQUOTAS; type++) {
907 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
908 continue;
909 if (unsuspend)
910 status = dquot_resume(sb, type);
911 else {
912 struct ocfs2_mem_dqinfo *oinfo;
913
914 /* Cancel periodic syncing before suspending */
915 oinfo = sb_dqinfo(sb, type)->dqi_priv;
916 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
917 status = dquot_suspend(sb, type);
918 }
919 if (status < 0)
920 break;
921 }
922 if (status < 0)
923 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
924 "remount (error = %d).\n", status);
925 return status;
926}
927
928static int ocfs2_enable_quotas(struct ocfs2_super *osb)
929{
930 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
931 struct super_block *sb = osb->sb;
932 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
933 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
934 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
935 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
936 int status;
937 int type;
938
939 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
940 for (type = 0; type < MAXQUOTAS; type++) {
941 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
942 continue;
943 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
944 osb->slot_num);
945 if (!inode[type]) {
946 status = -ENOENT;
947 goto out_quota_off;
948 }
949 status = dquot_enable(inode[type], type, QFMT_OCFS2,
950 DQUOT_USAGE_ENABLED);
951 if (status < 0)
952 goto out_quota_off;
953 }
954
955 for (type = 0; type < MAXQUOTAS; type++)
956 iput(inode[type]);
957 return 0;
958out_quota_off:
959 ocfs2_disable_quotas(osb);
960 for (type = 0; type < MAXQUOTAS; type++)
961 iput(inode[type]);
962 mlog_errno(status);
963 return status;
964}
965
966static void ocfs2_disable_quotas(struct ocfs2_super *osb)
967{
968 int type;
969 struct inode *inode;
970 struct super_block *sb = osb->sb;
971 struct ocfs2_mem_dqinfo *oinfo;
972
973 /* We mostly ignore errors in this function because there's not much
974 * we can do when we see them */
975 for (type = 0; type < MAXQUOTAS; type++) {
976 if (!sb_has_quota_loaded(sb, type))
977 continue;
978 /* Cancel periodic syncing before we grab dqonoff_mutex */
979 oinfo = sb_dqinfo(sb, type)->dqi_priv;
980 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
981 inode = igrab(sb->s_dquot.files[type]);
982 /* Turn off quotas. This will remove all dquot structures from
983 * memory and so they will be automatically synced to global
984 * quota files */
985 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
986 DQUOT_LIMITS_ENABLED);
987 if (!inode)
988 continue;
989 iput(inode);
990 }
991}
992
993/* Handle quota on quotactl */
994static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
995{
996 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
997 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
998
999 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1000 return -EINVAL;
1001
1002 return dquot_enable(sb_dqopt(sb)->files[type], type,
1003 format_id, DQUOT_LIMITS_ENABLED);
1004}
1005
1006/* Handle quota off quotactl */
1007static int ocfs2_quota_off(struct super_block *sb, int type)
1008{
1009 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1010}
1011
1012static const struct quotactl_ops ocfs2_quotactl_ops = {
1013 .quota_on_meta = ocfs2_quota_on,
1014 .quota_off = ocfs2_quota_off,
1015 .quota_sync = dquot_quota_sync,
1016 .get_info = dquot_get_dqinfo,
1017 .set_info = dquot_set_dqinfo,
1018 .get_dqblk = dquot_get_dqblk,
1019 .set_dqblk = dquot_set_dqblk,
1020};
1021
1022static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1023{
1024 struct dentry *root;
1025 int status, sector_size;
1026 struct mount_options parsed_options;
1027 struct inode *inode = NULL;
1028 struct ocfs2_super *osb = NULL;
1029 struct buffer_head *bh = NULL;
1030 char nodestr[12];
1031 struct ocfs2_blockcheck_stats stats;
1032
1033 trace_ocfs2_fill_super(sb, data, silent);
1034
1035 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1036 status = -EINVAL;
1037 goto read_super_error;
1038 }
1039
1040 /* probe for superblock */
1041 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1042 if (status < 0) {
1043 mlog(ML_ERROR, "superblock probe failed!\n");
1044 goto read_super_error;
1045 }
1046
1047 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1048 osb = OCFS2_SB(sb);
1049 if (status < 0) {
1050 mlog_errno(status);
1051 goto read_super_error;
1052 }
1053 brelse(bh);
1054 bh = NULL;
1055
1056 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1057 status = -EINVAL;
1058 goto read_super_error;
1059 }
1060 osb->s_mount_opt = parsed_options.mount_opt;
1061 osb->s_atime_quantum = parsed_options.atime_quantum;
1062 osb->preferred_slot = parsed_options.slot;
1063 osb->osb_commit_interval = parsed_options.commit_interval;
1064
1065 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1066 osb->osb_resv_level = parsed_options.resv_level;
1067 osb->osb_dir_resv_level = parsed_options.resv_level;
1068 if (parsed_options.dir_resv_level == -1)
1069 osb->osb_dir_resv_level = parsed_options.resv_level;
1070 else
1071 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1072
1073 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1074 if (status)
1075 goto read_super_error;
1076
1077 sb->s_magic = OCFS2_SUPER_MAGIC;
1078
1079 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1080 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1081
1082 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1083 * heartbeat=none */
1084 if (bdev_read_only(sb->s_bdev)) {
1085 if (!(sb->s_flags & MS_RDONLY)) {
1086 status = -EACCES;
1087 mlog(ML_ERROR, "Readonly device detected but readonly "
1088 "mount was not specified.\n");
1089 goto read_super_error;
1090 }
1091
1092 /* You should not be able to start a local heartbeat
1093 * on a readonly device. */
1094 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1095 status = -EROFS;
1096 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1097 "device.\n");
1098 goto read_super_error;
1099 }
1100
1101 status = ocfs2_check_journals_nolocks(osb);
1102 if (status < 0) {
1103 if (status == -EROFS)
1104 mlog(ML_ERROR, "Recovery required on readonly "
1105 "file system, but write access is "
1106 "unavailable.\n");
1107 else
1108 mlog_errno(status);
1109 goto read_super_error;
1110 }
1111
1112 ocfs2_set_ro_flag(osb, 1);
1113
1114 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1115 "Cluster services will not be used for this mount. "
1116 "Recovery will be skipped.\n", osb->dev_str);
1117 }
1118
1119 if (!ocfs2_is_hard_readonly(osb)) {
1120 if (sb->s_flags & MS_RDONLY)
1121 ocfs2_set_ro_flag(osb, 0);
1122 }
1123
1124 status = ocfs2_verify_heartbeat(osb);
1125 if (status < 0) {
1126 mlog_errno(status);
1127 goto read_super_error;
1128 }
1129
1130 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1131 ocfs2_debugfs_root);
1132 if (!osb->osb_debug_root) {
1133 status = -EINVAL;
1134 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1135 goto read_super_error;
1136 }
1137
1138 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1139 osb->osb_debug_root,
1140 osb,
1141 &ocfs2_osb_debug_fops);
1142 if (!osb->osb_ctxt) {
1143 status = -EINVAL;
1144 mlog_errno(status);
1145 goto read_super_error;
1146 }
1147
1148 if (ocfs2_meta_ecc(osb)) {
1149 status = ocfs2_blockcheck_stats_debugfs_install(
1150 &osb->osb_ecc_stats,
1151 osb->osb_debug_root);
1152 if (status) {
1153 mlog(ML_ERROR,
1154 "Unable to create blockcheck statistics "
1155 "files\n");
1156 goto read_super_error;
1157 }
1158 }
1159
1160 status = ocfs2_mount_volume(sb);
1161 if (status < 0)
1162 goto read_super_error;
1163
1164 if (osb->root_inode)
1165 inode = igrab(osb->root_inode);
1166
1167 if (!inode) {
1168 status = -EIO;
1169 mlog_errno(status);
1170 goto read_super_error;
1171 }
1172
1173 root = d_make_root(inode);
1174 if (!root) {
1175 status = -ENOMEM;
1176 mlog_errno(status);
1177 goto read_super_error;
1178 }
1179
1180 sb->s_root = root;
1181
1182 ocfs2_complete_mount_recovery(osb);
1183
1184 if (ocfs2_mount_local(osb))
1185 snprintf(nodestr, sizeof(nodestr), "local");
1186 else
1187 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1188
1189 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1190 "with %s data mode.\n",
1191 osb->dev_str, nodestr, osb->slot_num,
1192 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1193 "ordered");
1194
1195 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1196 wake_up(&osb->osb_mount_event);
1197
1198 /* Now we can initialize quotas because we can afford to wait
1199 * for cluster locks recovery now. That also means that truncation
1200 * log recovery can happen but that waits for proper quota setup */
1201 if (!(sb->s_flags & MS_RDONLY)) {
1202 status = ocfs2_enable_quotas(osb);
1203 if (status < 0) {
1204 /* We have to err-out specially here because
1205 * s_root is already set */
1206 mlog_errno(status);
1207 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1208 wake_up(&osb->osb_mount_event);
1209 return status;
1210 }
1211 }
1212
1213 ocfs2_complete_quota_recovery(osb);
1214
1215 /* Now we wake up again for processes waiting for quotas */
1216 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1217 wake_up(&osb->osb_mount_event);
1218
1219 /* Start this when the mount is almost sure of being successful */
1220 ocfs2_orphan_scan_start(osb);
1221
1222 return status;
1223
1224read_super_error:
1225 brelse(bh);
1226
1227 if (osb) {
1228 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1229 wake_up(&osb->osb_mount_event);
1230 ocfs2_dismount_volume(sb, 1);
1231 }
1232
1233 if (status)
1234 mlog_errno(status);
1235 return status;
1236}
1237
1238static 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
1246static 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};
1254MODULE_ALIAS_FS("ocfs2");
1255
1256static 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
1290static 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
1299 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1300
1301 mopt->commit_interval = 0;
1302 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1303 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1304 mopt->slot = OCFS2_INVALID_SLOT;
1305 mopt->localalloc_opt = -1;
1306 mopt->cluster_stack[0] = '\0';
1307 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1308 mopt->dir_resv_level = -1;
1309
1310 if (!options) {
1311 status = 1;
1312 goto bail;
1313 }
1314
1315 while ((p = strsep(&options, ",")) != NULL) {
1316 int token, option;
1317 substring_t args[MAX_OPT_ARGS];
1318
1319 if (!*p)
1320 continue;
1321
1322 token = match_token(p, tokens, args);
1323 switch (token) {
1324 case Opt_hb_local:
1325 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1326 break;
1327 case Opt_hb_none:
1328 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1329 break;
1330 case Opt_hb_global:
1331 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1332 break;
1333 case Opt_barrier:
1334 if (match_int(&args[0], &option)) {
1335 status = 0;
1336 goto bail;
1337 }
1338 if (option)
1339 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1340 else
1341 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1342 break;
1343 case Opt_intr:
1344 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1345 break;
1346 case Opt_nointr:
1347 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1348 break;
1349 case Opt_err_panic:
1350 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1351 break;
1352 case Opt_err_ro:
1353 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1354 break;
1355 case Opt_data_ordered:
1356 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1357 break;
1358 case Opt_data_writeback:
1359 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1360 break;
1361 case Opt_user_xattr:
1362 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1363 break;
1364 case Opt_nouser_xattr:
1365 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1366 break;
1367 case Opt_atime_quantum:
1368 if (match_int(&args[0], &option)) {
1369 status = 0;
1370 goto bail;
1371 }
1372 if (option >= 0)
1373 mopt->atime_quantum = option;
1374 break;
1375 case Opt_slot:
1376 option = 0;
1377 if (match_int(&args[0], &option)) {
1378 status = 0;
1379 goto bail;
1380 }
1381 if (option)
1382 mopt->slot = (s16)option;
1383 break;
1384 case Opt_commit:
1385 option = 0;
1386 if (match_int(&args[0], &option)) {
1387 status = 0;
1388 goto bail;
1389 }
1390 if (option < 0)
1391 return 0;
1392 if (option == 0)
1393 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1394 mopt->commit_interval = HZ * option;
1395 break;
1396 case Opt_localalloc:
1397 option = 0;
1398 if (match_int(&args[0], &option)) {
1399 status = 0;
1400 goto bail;
1401 }
1402 if (option >= 0)
1403 mopt->localalloc_opt = option;
1404 break;
1405 case Opt_localflocks:
1406 /*
1407 * Changing this during remount could race
1408 * flock() requests, or "unbalance" existing
1409 * ones (e.g., a lock is taken in one mode but
1410 * dropped in the other). If users care enough
1411 * to flip locking modes during remount, we
1412 * could add a "local" flag to individual
1413 * flock structures for proper tracking of
1414 * state.
1415 */
1416 if (!is_remount)
1417 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1418 break;
1419 case Opt_stack:
1420 /* Check both that the option we were passed
1421 * is of the right length and that it is a proper
1422 * string of the right length.
1423 */
1424 if (((args[0].to - args[0].from) !=
1425 OCFS2_STACK_LABEL_LEN) ||
1426 (strnlen(args[0].from,
1427 OCFS2_STACK_LABEL_LEN) !=
1428 OCFS2_STACK_LABEL_LEN)) {
1429 mlog(ML_ERROR,
1430 "Invalid cluster_stack option\n");
1431 status = 0;
1432 goto bail;
1433 }
1434 memcpy(mopt->cluster_stack, args[0].from,
1435 OCFS2_STACK_LABEL_LEN);
1436 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1437 /*
1438 * Open code the memcmp here as we don't have
1439 * an osb to pass to
1440 * ocfs2_userspace_stack().
1441 */
1442 if (memcmp(mopt->cluster_stack,
1443 OCFS2_CLASSIC_CLUSTER_STACK,
1444 OCFS2_STACK_LABEL_LEN))
1445 user_stack = 1;
1446 break;
1447 case Opt_inode64:
1448 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1449 break;
1450 case Opt_usrquota:
1451 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1452 break;
1453 case Opt_grpquota:
1454 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1455 break;
1456 case Opt_coherency_buffered:
1457 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1458 break;
1459 case Opt_coherency_full:
1460 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1461 break;
1462 case Opt_acl:
1463 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1464 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1465 break;
1466 case Opt_noacl:
1467 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1468 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1469 break;
1470 case Opt_resv_level:
1471 if (is_remount)
1472 break;
1473 if (match_int(&args[0], &option)) {
1474 status = 0;
1475 goto bail;
1476 }
1477 if (option >= OCFS2_MIN_RESV_LEVEL &&
1478 option < OCFS2_MAX_RESV_LEVEL)
1479 mopt->resv_level = option;
1480 break;
1481 case Opt_dir_resv_level:
1482 if (is_remount)
1483 break;
1484 if (match_int(&args[0], &option)) {
1485 status = 0;
1486 goto bail;
1487 }
1488 if (option >= OCFS2_MIN_RESV_LEVEL &&
1489 option < OCFS2_MAX_RESV_LEVEL)
1490 mopt->dir_resv_level = option;
1491 break;
1492 default:
1493 mlog(ML_ERROR,
1494 "Unrecognized mount option \"%s\" "
1495 "or missing value\n", p);
1496 status = 0;
1497 goto bail;
1498 }
1499 }
1500
1501 if (user_stack == 0) {
1502 /* Ensure only one heartbeat mode */
1503 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1504 OCFS2_MOUNT_HB_GLOBAL |
1505 OCFS2_MOUNT_HB_NONE);
1506 if (hweight32(tmp) != 1) {
1507 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1508 status = 0;
1509 goto bail;
1510 }
1511 }
1512
1513 status = 1;
1514
1515bail:
1516 return status;
1517}
1518
1519static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1520{
1521 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1522 unsigned long opts = osb->s_mount_opt;
1523 unsigned int local_alloc_megs;
1524
1525 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1526 seq_printf(s, ",_netdev");
1527 if (opts & OCFS2_MOUNT_HB_LOCAL)
1528 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1529 else
1530 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1531 } else
1532 seq_printf(s, ",%s", OCFS2_HB_NONE);
1533
1534 if (opts & OCFS2_MOUNT_NOINTR)
1535 seq_printf(s, ",nointr");
1536
1537 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1538 seq_printf(s, ",data=writeback");
1539 else
1540 seq_printf(s, ",data=ordered");
1541
1542 if (opts & OCFS2_MOUNT_BARRIER)
1543 seq_printf(s, ",barrier=1");
1544
1545 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1546 seq_printf(s, ",errors=panic");
1547 else
1548 seq_printf(s, ",errors=remount-ro");
1549
1550 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1551 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1552
1553 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1554
1555 if (osb->osb_commit_interval)
1556 seq_printf(s, ",commit=%u",
1557 (unsigned) (osb->osb_commit_interval / HZ));
1558
1559 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1560 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1561 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1562
1563 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1564 seq_printf(s, ",localflocks,");
1565
1566 if (osb->osb_cluster_stack[0])
1567 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1568 osb->osb_cluster_stack);
1569 if (opts & OCFS2_MOUNT_USRQUOTA)
1570 seq_printf(s, ",usrquota");
1571 if (opts & OCFS2_MOUNT_GRPQUOTA)
1572 seq_printf(s, ",grpquota");
1573
1574 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1575 seq_printf(s, ",coherency=buffered");
1576 else
1577 seq_printf(s, ",coherency=full");
1578
1579 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1580 seq_printf(s, ",nouser_xattr");
1581 else
1582 seq_printf(s, ",user_xattr");
1583
1584 if (opts & OCFS2_MOUNT_INODE64)
1585 seq_printf(s, ",inode64");
1586
1587 if (opts & OCFS2_MOUNT_POSIX_ACL)
1588 seq_printf(s, ",acl");
1589 else
1590 seq_printf(s, ",noacl");
1591
1592 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1593 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1594
1595 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1596 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1597
1598 return 0;
1599}
1600
1601static int __init ocfs2_init(void)
1602{
1603 int status;
1604
1605 status = init_ocfs2_uptodate_cache();
1606 if (status < 0)
1607 goto out1;
1608
1609 status = ocfs2_initialize_mem_caches();
1610 if (status < 0)
1611 goto out2;
1612
1613 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1614 if (!ocfs2_wq) {
1615 status = -ENOMEM;
1616 goto out3;
1617 }
1618
1619 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1620 if (!ocfs2_debugfs_root) {
1621 status = -EFAULT;
1622 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1623 }
1624
1625 ocfs2_set_locking_protocol();
1626
1627 status = register_quota_format(&ocfs2_quota_format);
1628 if (status < 0)
1629 goto out4;
1630 status = register_filesystem(&ocfs2_fs_type);
1631 if (!status)
1632 return 0;
1633
1634 unregister_quota_format(&ocfs2_quota_format);
1635out4:
1636 destroy_workqueue(ocfs2_wq);
1637 debugfs_remove(ocfs2_debugfs_root);
1638out3:
1639 ocfs2_free_mem_caches();
1640out2:
1641 exit_ocfs2_uptodate_cache();
1642out1:
1643 mlog_errno(status);
1644 return status;
1645}
1646
1647static void __exit ocfs2_exit(void)
1648{
1649 if (ocfs2_wq) {
1650 flush_workqueue(ocfs2_wq);
1651 destroy_workqueue(ocfs2_wq);
1652 }
1653
1654 unregister_quota_format(&ocfs2_quota_format);
1655
1656 debugfs_remove(ocfs2_debugfs_root);
1657
1658 ocfs2_free_mem_caches();
1659
1660 unregister_filesystem(&ocfs2_fs_type);
1661
1662 exit_ocfs2_uptodate_cache();
1663}
1664
1665static void ocfs2_put_super(struct super_block *sb)
1666{
1667 trace_ocfs2_put_super(sb);
1668
1669 ocfs2_sync_blockdev(sb);
1670 ocfs2_dismount_volume(sb, 0);
1671}
1672
1673static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1674{
1675 struct ocfs2_super *osb;
1676 u32 numbits, freebits;
1677 int status;
1678 struct ocfs2_dinode *bm_lock;
1679 struct buffer_head *bh = NULL;
1680 struct inode *inode = NULL;
1681
1682 trace_ocfs2_statfs(dentry->d_sb, buf);
1683
1684 osb = OCFS2_SB(dentry->d_sb);
1685
1686 inode = ocfs2_get_system_file_inode(osb,
1687 GLOBAL_BITMAP_SYSTEM_INODE,
1688 OCFS2_INVALID_SLOT);
1689 if (!inode) {
1690 mlog(ML_ERROR, "failed to get bitmap inode\n");
1691 status = -EIO;
1692 goto bail;
1693 }
1694
1695 status = ocfs2_inode_lock(inode, &bh, 0);
1696 if (status < 0) {
1697 mlog_errno(status);
1698 goto bail;
1699 }
1700
1701 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1702
1703 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1704 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1705
1706 buf->f_type = OCFS2_SUPER_MAGIC;
1707 buf->f_bsize = dentry->d_sb->s_blocksize;
1708 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1709 buf->f_blocks = ((sector_t) numbits) *
1710 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1711 buf->f_bfree = ((sector_t) freebits) *
1712 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1713 buf->f_bavail = buf->f_bfree;
1714 buf->f_files = numbits;
1715 buf->f_ffree = freebits;
1716 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1717 & 0xFFFFFFFFUL;
1718 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1719 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1720
1721 brelse(bh);
1722
1723 ocfs2_inode_unlock(inode, 0);
1724 status = 0;
1725bail:
1726 if (inode)
1727 iput(inode);
1728
1729 if (status)
1730 mlog_errno(status);
1731
1732 return status;
1733}
1734
1735static 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 oi->ip_dir_start_lookup = 0;
1745 mutex_init(&oi->ip_unaligned_aio);
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
1753 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1754
1755 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1756 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1757 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1758
1759 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1760 &ocfs2_inode_caching_ops);
1761
1762 inode_init_once(&oi->vfs_inode);
1763}
1764
1765static int ocfs2_initialize_mem_caches(void)
1766{
1767 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1768 sizeof(struct ocfs2_inode_info),
1769 0,
1770 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1771 SLAB_MEM_SPREAD),
1772 ocfs2_inode_init_once);
1773 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1774 sizeof(struct ocfs2_dquot),
1775 0,
1776 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1777 SLAB_MEM_SPREAD),
1778 NULL);
1779 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1780 sizeof(struct ocfs2_quota_chunk),
1781 0,
1782 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1783 NULL);
1784 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1785 !ocfs2_qf_chunk_cachep) {
1786 if (ocfs2_inode_cachep)
1787 kmem_cache_destroy(ocfs2_inode_cachep);
1788 if (ocfs2_dquot_cachep)
1789 kmem_cache_destroy(ocfs2_dquot_cachep);
1790 if (ocfs2_qf_chunk_cachep)
1791 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1792 return -ENOMEM;
1793 }
1794
1795 return 0;
1796}
1797
1798static void ocfs2_free_mem_caches(void)
1799{
1800 /*
1801 * Make sure all delayed rcu free inodes are flushed before we
1802 * destroy cache.
1803 */
1804 rcu_barrier();
1805 if (ocfs2_inode_cachep)
1806 kmem_cache_destroy(ocfs2_inode_cachep);
1807 ocfs2_inode_cachep = NULL;
1808
1809 if (ocfs2_dquot_cachep)
1810 kmem_cache_destroy(ocfs2_dquot_cachep);
1811 ocfs2_dquot_cachep = NULL;
1812
1813 if (ocfs2_qf_chunk_cachep)
1814 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1815 ocfs2_qf_chunk_cachep = NULL;
1816}
1817
1818static int ocfs2_get_sector(struct super_block *sb,
1819 struct buffer_head **bh,
1820 int block,
1821 int sect_size)
1822{
1823 if (!sb_set_blocksize(sb, sect_size)) {
1824 mlog(ML_ERROR, "unable to set blocksize\n");
1825 return -EIO;
1826 }
1827
1828 *bh = sb_getblk(sb, block);
1829 if (!*bh) {
1830 mlog_errno(-ENOMEM);
1831 return -ENOMEM;
1832 }
1833 lock_buffer(*bh);
1834 if (!buffer_dirty(*bh))
1835 clear_buffer_uptodate(*bh);
1836 unlock_buffer(*bh);
1837 ll_rw_block(READ, 1, bh);
1838 wait_on_buffer(*bh);
1839 if (!buffer_uptodate(*bh)) {
1840 mlog_errno(-EIO);
1841 brelse(*bh);
1842 *bh = NULL;
1843 return -EIO;
1844 }
1845
1846 return 0;
1847}
1848
1849static int ocfs2_mount_volume(struct super_block *sb)
1850{
1851 int status = 0;
1852 int unlock_super = 0;
1853 struct ocfs2_super *osb = OCFS2_SB(sb);
1854
1855 if (ocfs2_is_hard_readonly(osb))
1856 goto leave;
1857
1858 status = ocfs2_dlm_init(osb);
1859 if (status < 0) {
1860 mlog_errno(status);
1861 goto leave;
1862 }
1863
1864 status = ocfs2_super_lock(osb, 1);
1865 if (status < 0) {
1866 mlog_errno(status);
1867 goto leave;
1868 }
1869 unlock_super = 1;
1870
1871 /* This will load up the node map and add ourselves to it. */
1872 status = ocfs2_find_slot(osb);
1873 if (status < 0) {
1874 mlog_errno(status);
1875 goto leave;
1876 }
1877
1878 /* load all node-local system inodes */
1879 status = ocfs2_init_local_system_inodes(osb);
1880 if (status < 0) {
1881 mlog_errno(status);
1882 goto leave;
1883 }
1884
1885 status = ocfs2_check_volume(osb);
1886 if (status < 0) {
1887 mlog_errno(status);
1888 goto leave;
1889 }
1890
1891 status = ocfs2_truncate_log_init(osb);
1892 if (status < 0)
1893 mlog_errno(status);
1894
1895leave:
1896 if (unlock_super)
1897 ocfs2_super_unlock(osb, 1);
1898
1899 return status;
1900}
1901
1902static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1903{
1904 int tmp, hangup_needed = 0;
1905 struct ocfs2_super *osb = NULL;
1906 char nodestr[12];
1907
1908 trace_ocfs2_dismount_volume(sb);
1909
1910 BUG_ON(!sb);
1911 osb = OCFS2_SB(sb);
1912 BUG_ON(!osb);
1913
1914 debugfs_remove(osb->osb_ctxt);
1915
1916 /* Orphan scan should be stopped as early as possible */
1917 ocfs2_orphan_scan_stop(osb);
1918
1919 ocfs2_disable_quotas(osb);
1920
1921 /* All dquots should be freed by now */
1922 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1923 /* Wait for worker to be done with the work structure in osb */
1924 cancel_work_sync(&osb->dquot_drop_work);
1925
1926 ocfs2_shutdown_local_alloc(osb);
1927
1928 /* This will disable recovery and flush any recovery work. */
1929 ocfs2_recovery_exit(osb);
1930
1931 /*
1932 * During dismount, when it recovers another node it will call
1933 * ocfs2_recover_orphans and queue delayed work osb_truncate_log_wq.
1934 */
1935 ocfs2_truncate_log_shutdown(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
1996static 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(). */
2023static 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
2050static 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 = &ocfs2_quotactl_ops;
2075 sb->dq_op = &ocfs2_quota_operations;
2076 sb->s_xattr = ocfs2_xattr_handlers;
2077 sb->s_time_gran = 1;
2078 sb->s_flags |= MS_NOATIME;
2079 /* this is needed to support O_LARGEFILE */
2080 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2081 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2082 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2083
2084 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2085
2086 for (i = 0; i < 3; i++)
2087 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2088 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2089
2090 osb->sb = sb;
2091 /* Save off for ocfs2_rw_direct */
2092 osb->s_sectsize_bits = blksize_bits(sector_size);
2093 BUG_ON(!osb->s_sectsize_bits);
2094
2095 spin_lock_init(&osb->dc_task_lock);
2096 init_waitqueue_head(&osb->dc_event);
2097 osb->dc_work_sequence = 0;
2098 osb->dc_wake_sequence = 0;
2099 INIT_LIST_HEAD(&osb->blocked_lock_list);
2100 osb->blocked_lock_count = 0;
2101 spin_lock_init(&osb->osb_lock);
2102 spin_lock_init(&osb->osb_xattr_lock);
2103 ocfs2_init_steal_slots(osb);
2104
2105 mutex_init(&osb->system_file_mutex);
2106
2107 atomic_set(&osb->alloc_stats.moves, 0);
2108 atomic_set(&osb->alloc_stats.local_data, 0);
2109 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2110 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2111 atomic_set(&osb->alloc_stats.bg_extends, 0);
2112
2113 /* Copy the blockcheck stats from the superblock probe */
2114 osb->osb_ecc_stats = *stats;
2115
2116 ocfs2_init_node_maps(osb);
2117
2118 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2119 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2120
2121 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2122 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2123 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2124 osb->max_slots);
2125 status = -EINVAL;
2126 goto bail;
2127 }
2128
2129 ocfs2_orphan_scan_init(osb);
2130
2131 status = ocfs2_recovery_init(osb);
2132 if (status) {
2133 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2134 mlog_errno(status);
2135 goto bail;
2136 }
2137
2138 init_waitqueue_head(&osb->checkpoint_event);
2139
2140 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2141
2142 osb->slot_num = OCFS2_INVALID_SLOT;
2143
2144 osb->s_xattr_inline_size = le16_to_cpu(
2145 di->id2.i_super.s_xattr_inline_size);
2146
2147 osb->local_alloc_state = OCFS2_LA_UNUSED;
2148 osb->local_alloc_bh = NULL;
2149 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2150
2151 init_waitqueue_head(&osb->osb_mount_event);
2152
2153 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2154 if (status) {
2155 mlog_errno(status);
2156 goto bail;
2157 }
2158
2159 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2160 if (!osb->vol_label) {
2161 mlog(ML_ERROR, "unable to alloc vol label\n");
2162 status = -ENOMEM;
2163 goto bail;
2164 }
2165
2166 osb->slot_recovery_generations =
2167 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2168 GFP_KERNEL);
2169 if (!osb->slot_recovery_generations) {
2170 status = -ENOMEM;
2171 mlog_errno(status);
2172 goto bail;
2173 }
2174
2175 init_waitqueue_head(&osb->osb_wipe_event);
2176 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2177 sizeof(*osb->osb_orphan_wipes),
2178 GFP_KERNEL);
2179 if (!osb->osb_orphan_wipes) {
2180 status = -ENOMEM;
2181 mlog_errno(status);
2182 goto bail;
2183 }
2184
2185 osb->osb_rf_lock_tree = RB_ROOT;
2186
2187 osb->s_feature_compat =
2188 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2189 osb->s_feature_ro_compat =
2190 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2191 osb->s_feature_incompat =
2192 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2193
2194 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2195 mlog(ML_ERROR, "couldn't mount because of unsupported "
2196 "optional features (%x).\n", i);
2197 status = -EINVAL;
2198 goto bail;
2199 }
2200 if (!(osb->sb->s_flags & MS_RDONLY) &&
2201 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2202 mlog(ML_ERROR, "couldn't mount RDWR because of "
2203 "unsupported optional features (%x).\n", i);
2204 status = -EINVAL;
2205 goto bail;
2206 }
2207
2208 if (ocfs2_clusterinfo_valid(osb)) {
2209 osb->osb_stackflags =
2210 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2211 strlcpy(osb->osb_cluster_stack,
2212 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2213 OCFS2_STACK_LABEL_LEN + 1);
2214 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2215 mlog(ML_ERROR,
2216 "couldn't mount because of an invalid "
2217 "cluster stack label (%s) \n",
2218 osb->osb_cluster_stack);
2219 status = -EINVAL;
2220 goto bail;
2221 }
2222 strlcpy(osb->osb_cluster_name,
2223 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2224 OCFS2_CLUSTER_NAME_LEN + 1);
2225 } else {
2226 /* The empty string is identical with classic tools that
2227 * don't know about s_cluster_info. */
2228 osb->osb_cluster_stack[0] = '\0';
2229 }
2230
2231 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2232
2233 /* FIXME
2234 * This should be done in ocfs2_journal_init(), but unknown
2235 * ordering issues will cause the filesystem to crash.
2236 * If anyone wants to figure out what part of the code
2237 * refers to osb->journal before ocfs2_journal_init() is run,
2238 * be my guest.
2239 */
2240 /* initialize our journal structure */
2241
2242 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2243 if (!journal) {
2244 mlog(ML_ERROR, "unable to alloc journal\n");
2245 status = -ENOMEM;
2246 goto bail;
2247 }
2248 osb->journal = journal;
2249 journal->j_osb = osb;
2250
2251 atomic_set(&journal->j_num_trans, 0);
2252 init_rwsem(&journal->j_trans_barrier);
2253 init_waitqueue_head(&journal->j_checkpointed);
2254 spin_lock_init(&journal->j_lock);
2255 journal->j_trans_id = (unsigned long) 1;
2256 INIT_LIST_HEAD(&journal->j_la_cleanups);
2257 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2258 journal->j_state = OCFS2_JOURNAL_FREE;
2259
2260 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2261 init_llist_head(&osb->dquot_drop_list);
2262
2263 /* get some pseudo constants for clustersize bits */
2264 osb->s_clustersize_bits =
2265 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2266 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2267
2268 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2269 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2270 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2271 osb->s_clustersize);
2272 status = -EINVAL;
2273 goto bail;
2274 }
2275
2276 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2277 le32_to_cpu(di->i_clusters));
2278
2279 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2280 total_blocks);
2281 if (status) {
2282 mlog(ML_ERROR, "Volume too large "
2283 "to mount safely on this system");
2284 status = -EFBIG;
2285 goto bail;
2286 }
2287
2288 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2289 sizeof(di->id2.i_super.s_uuid))) {
2290 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2291 status = -ENOMEM;
2292 goto bail;
2293 }
2294
2295 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2296 osb->vol_label[63] = '\0';
2297 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2298 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2299 osb->first_cluster_group_blkno =
2300 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2301 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2302 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2303 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2304 (unsigned long long)osb->root_blkno,
2305 (unsigned long long)osb->system_dir_blkno,
2306 osb->s_clustersize_bits);
2307
2308 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2309 if (!osb->osb_dlm_debug) {
2310 status = -ENOMEM;
2311 mlog_errno(status);
2312 goto bail;
2313 }
2314
2315 atomic_set(&osb->vol_state, VOLUME_INIT);
2316
2317 /* load root, system_dir, and all global system inodes */
2318 status = ocfs2_init_global_system_inodes(osb);
2319 if (status < 0) {
2320 mlog_errno(status);
2321 goto bail;
2322 }
2323
2324 /*
2325 * global bitmap
2326 */
2327 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2328 OCFS2_INVALID_SLOT);
2329 if (!inode) {
2330 status = -EINVAL;
2331 mlog_errno(status);
2332 goto bail;
2333 }
2334
2335 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2336 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2337 iput(inode);
2338
2339 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2340 osb->s_feature_incompat) * 8;
2341
2342 status = ocfs2_init_slot_info(osb);
2343 if (status < 0) {
2344 mlog_errno(status);
2345 goto bail;
2346 }
2347 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2348
2349bail:
2350 return status;
2351}
2352
2353/*
2354 * will return: -EAGAIN if it is ok to keep searching for superblocks
2355 * -EINVAL if there is a bad superblock
2356 * 0 on success
2357 */
2358static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2359 struct buffer_head *bh,
2360 u32 blksz,
2361 struct ocfs2_blockcheck_stats *stats)
2362{
2363 int status = -EAGAIN;
2364
2365 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2366 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2367 /* We have to do a raw check of the feature here */
2368 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2369 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2370 status = ocfs2_block_check_validate(bh->b_data,
2371 bh->b_size,
2372 &di->i_check,
2373 stats);
2374 if (status)
2375 goto out;
2376 }
2377 status = -EINVAL;
2378 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2379 mlog(ML_ERROR, "found superblock with incorrect block "
2380 "size: found %u, should be %u\n",
2381 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2382 blksz);
2383 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2384 OCFS2_MAJOR_REV_LEVEL ||
2385 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2386 OCFS2_MINOR_REV_LEVEL) {
2387 mlog(ML_ERROR, "found superblock with bad version: "
2388 "found %u.%u, should be %u.%u\n",
2389 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2390 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2391 OCFS2_MAJOR_REV_LEVEL,
2392 OCFS2_MINOR_REV_LEVEL);
2393 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2394 mlog(ML_ERROR, "bad block number on superblock: "
2395 "found %llu, should be %llu\n",
2396 (unsigned long long)le64_to_cpu(di->i_blkno),
2397 (unsigned long long)bh->b_blocknr);
2398 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2399 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2400 mlog(ML_ERROR, "bad cluster size found: %u\n",
2401 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2402 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2403 mlog(ML_ERROR, "bad root_blkno: 0\n");
2404 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2405 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2406 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2407 mlog(ML_ERROR,
2408 "Superblock slots found greater than file system "
2409 "maximum: found %u, max %u\n",
2410 le16_to_cpu(di->id2.i_super.s_max_slots),
2411 OCFS2_MAX_SLOTS);
2412 } else {
2413 /* found it! */
2414 status = 0;
2415 }
2416 }
2417
2418out:
2419 if (status && status != -EAGAIN)
2420 mlog_errno(status);
2421 return status;
2422}
2423
2424static int ocfs2_check_volume(struct ocfs2_super *osb)
2425{
2426 int status;
2427 int dirty;
2428 int local;
2429 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2430 * recover
2431 * ourselves. */
2432
2433 /* Init our journal object. */
2434 status = ocfs2_journal_init(osb->journal, &dirty);
2435 if (status < 0) {
2436 mlog(ML_ERROR, "Could not initialize journal!\n");
2437 goto finally;
2438 }
2439
2440 /* Now that journal has been initialized, check to make sure
2441 entire volume is addressable. */
2442 status = ocfs2_journal_addressable(osb);
2443 if (status)
2444 goto finally;
2445
2446 /* If the journal was unmounted cleanly then we don't want to
2447 * recover anything. Otherwise, journal_load will do that
2448 * dirty work for us :) */
2449 if (!dirty) {
2450 status = ocfs2_journal_wipe(osb->journal, 0);
2451 if (status < 0) {
2452 mlog_errno(status);
2453 goto finally;
2454 }
2455 } else {
2456 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2457 "unmounted cleanly, recovering it.\n", osb->dev_str);
2458 }
2459
2460 local = ocfs2_mount_local(osb);
2461
2462 /* will play back anything left in the journal. */
2463 status = ocfs2_journal_load(osb->journal, local, dirty);
2464 if (status < 0) {
2465 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2466 goto finally;
2467 }
2468
2469 if (dirty) {
2470 /* recover my local alloc if we didn't unmount cleanly. */
2471 status = ocfs2_begin_local_alloc_recovery(osb,
2472 osb->slot_num,
2473 &local_alloc);
2474 if (status < 0) {
2475 mlog_errno(status);
2476 goto finally;
2477 }
2478 /* we complete the recovery process after we've marked
2479 * ourselves as mounted. */
2480 }
2481
2482 status = ocfs2_load_local_alloc(osb);
2483 if (status < 0) {
2484 mlog_errno(status);
2485 goto finally;
2486 }
2487
2488 if (dirty) {
2489 /* Recovery will be completed after we've mounted the
2490 * rest of the volume. */
2491 osb->dirty = 1;
2492 osb->local_alloc_copy = local_alloc;
2493 local_alloc = NULL;
2494 }
2495
2496 /* go through each journal, trylock it and if you get the
2497 * lock, and it's marked as dirty, set the bit in the recover
2498 * map and launch a recovery thread for it. */
2499 status = ocfs2_mark_dead_nodes(osb);
2500 if (status < 0) {
2501 mlog_errno(status);
2502 goto finally;
2503 }
2504
2505 status = ocfs2_compute_replay_slots(osb);
2506 if (status < 0)
2507 mlog_errno(status);
2508
2509finally:
2510 kfree(local_alloc);
2511
2512 if (status)
2513 mlog_errno(status);
2514 return status;
2515}
2516
2517/*
2518 * The routine gets called from dismount or close whenever a dismount on
2519 * volume is requested and the osb open count becomes 1.
2520 * It will remove the osb from the global list and also free up all the
2521 * initialized resources and fileobject.
2522 */
2523static void ocfs2_delete_osb(struct ocfs2_super *osb)
2524{
2525 /* This function assumes that the caller has the main osb resource */
2526
2527 ocfs2_free_slot_info(osb);
2528
2529 kfree(osb->osb_orphan_wipes);
2530 kfree(osb->slot_recovery_generations);
2531 /* FIXME
2532 * This belongs in journal shutdown, but because we have to
2533 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2534 * we free it here.
2535 */
2536 kfree(osb->journal);
2537 kfree(osb->local_alloc_copy);
2538 kfree(osb->uuid_str);
2539 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2540 memset(osb, 0, sizeof(struct ocfs2_super));
2541}
2542
2543/* Put OCFS2 into a readonly state, or (if the user specifies it),
2544 * panic(). We do not support continue-on-error operation. */
2545static void ocfs2_handle_error(struct super_block *sb)
2546{
2547 struct ocfs2_super *osb = OCFS2_SB(sb);
2548
2549 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2550 panic("OCFS2: (device %s): panic forced after error\n",
2551 sb->s_id);
2552
2553 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2554
2555 if (sb->s_flags & MS_RDONLY &&
2556 (ocfs2_is_soft_readonly(osb) ||
2557 ocfs2_is_hard_readonly(osb)))
2558 return;
2559
2560 printk(KERN_CRIT "File system is now read-only due to the potential "
2561 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2562 "system is unmounted.\n");
2563 sb->s_flags |= MS_RDONLY;
2564 ocfs2_set_ro_flag(osb, 0);
2565}
2566
2567static char error_buf[1024];
2568
2569void __ocfs2_error(struct super_block *sb,
2570 const char *function,
2571 const char *fmt, ...)
2572{
2573 va_list args;
2574
2575 va_start(args, fmt);
2576 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2577 va_end(args);
2578
2579 /* Not using mlog here because we want to show the actual
2580 * function the error came from. */
2581 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2582 sb->s_id, function, error_buf);
2583
2584 ocfs2_handle_error(sb);
2585}
2586
2587/* Handle critical errors. This is intentionally more drastic than
2588 * ocfs2_handle_error, so we only use for things like journal errors,
2589 * etc. */
2590void __ocfs2_abort(struct super_block* sb,
2591 const char *function,
2592 const char *fmt, ...)
2593{
2594 va_list args;
2595
2596 va_start(args, fmt);
2597 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2598 va_end(args);
2599
2600 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2601 sb->s_id, function, error_buf);
2602
2603 /* We don't have the cluster support yet to go straight to
2604 * hard readonly in here. Until then, we want to keep
2605 * ocfs2_abort() so that we can at least mark critical
2606 * errors.
2607 *
2608 * TODO: This should abort the journal and alert other nodes
2609 * that our slot needs recovery. */
2610
2611 /* Force a panic(). This stinks, but it's better than letting
2612 * things continue without having a proper hard readonly
2613 * here. */
2614 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2615 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2616 ocfs2_handle_error(sb);
2617}
2618
2619/*
2620 * Void signal blockers, because in-kernel sigprocmask() only fails
2621 * when SIG_* is wrong.
2622 */
2623void ocfs2_block_signals(sigset_t *oldset)
2624{
2625 int rc;
2626 sigset_t blocked;
2627
2628 sigfillset(&blocked);
2629 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2630 BUG_ON(rc);
2631}
2632
2633void ocfs2_unblock_signals(sigset_t *oldset)
2634{
2635 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2636 BUG_ON(rc);
2637}
2638
2639module_init(ocfs2_init);
2640module_exit(ocfs2_exit);