Loading...
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 "ver.h"
72#include "xattr.h"
73#include "quota.h"
74#include "refcounttree.h"
75#include "suballoc.h"
76
77#include "buffer_head_io.h"
78
79static struct kmem_cache *ocfs2_inode_cachep = NULL;
80struct kmem_cache *ocfs2_dquot_cachep;
81struct kmem_cache *ocfs2_qf_chunk_cachep;
82
83/* OCFS2 needs to schedule several different types of work which
84 * require cluster locking, disk I/O, recovery waits, etc. Since these
85 * types of work tend to be heavy we avoid using the kernel events
86 * workqueue and schedule on our own. */
87struct workqueue_struct *ocfs2_wq = NULL;
88
89static struct dentry *ocfs2_debugfs_root = NULL;
90
91MODULE_AUTHOR("Oracle");
92MODULE_LICENSE("GPL");
93
94struct mount_options
95{
96 unsigned long commit_interval;
97 unsigned long mount_opt;
98 unsigned int atime_quantum;
99 signed short slot;
100 int localalloc_opt;
101 unsigned int resv_level;
102 int dir_resv_level;
103 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
104};
105
106static int ocfs2_parse_options(struct super_block *sb, char *options,
107 struct mount_options *mopt,
108 int is_remount);
109static int ocfs2_check_set_options(struct super_block *sb,
110 struct mount_options *options);
111static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
112static void ocfs2_put_super(struct super_block *sb);
113static int ocfs2_mount_volume(struct super_block *sb);
114static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116static int ocfs2_initialize_mem_caches(void);
117static void ocfs2_free_mem_caches(void);
118static void ocfs2_delete_osb(struct ocfs2_super *osb);
119
120static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
121
122static int ocfs2_sync_fs(struct super_block *sb, int wait);
123
124static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
126static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
127static int ocfs2_check_volume(struct ocfs2_super *osb);
128static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 struct buffer_head *bh,
130 u32 sectsize,
131 struct ocfs2_blockcheck_stats *stats);
132static int ocfs2_initialize_super(struct super_block *sb,
133 struct buffer_head *bh,
134 int sector_size,
135 struct ocfs2_blockcheck_stats *stats);
136static int ocfs2_get_sector(struct super_block *sb,
137 struct buffer_head **bh,
138 int block,
139 int sect_size);
140static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141static void ocfs2_destroy_inode(struct inode *inode);
142static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144static void ocfs2_disable_quotas(struct ocfs2_super *osb);
145
146static const struct super_operations ocfs2_sops = {
147 .statfs = ocfs2_statfs,
148 .alloc_inode = ocfs2_alloc_inode,
149 .destroy_inode = ocfs2_destroy_inode,
150 .drop_inode = ocfs2_drop_inode,
151 .evict_inode = ocfs2_evict_inode,
152 .sync_fs = ocfs2_sync_fs,
153 .put_super = ocfs2_put_super,
154 .remount_fs = ocfs2_remount,
155 .show_options = ocfs2_show_options,
156 .quota_read = ocfs2_quota_read,
157 .quota_write = ocfs2_quota_write,
158};
159
160enum {
161 Opt_barrier,
162 Opt_err_panic,
163 Opt_err_ro,
164 Opt_intr,
165 Opt_nointr,
166 Opt_hb_none,
167 Opt_hb_local,
168 Opt_hb_global,
169 Opt_data_ordered,
170 Opt_data_writeback,
171 Opt_atime_quantum,
172 Opt_slot,
173 Opt_commit,
174 Opt_localalloc,
175 Opt_localflocks,
176 Opt_stack,
177 Opt_user_xattr,
178 Opt_nouser_xattr,
179 Opt_inode64,
180 Opt_acl,
181 Opt_noacl,
182 Opt_usrquota,
183 Opt_grpquota,
184 Opt_coherency_buffered,
185 Opt_coherency_full,
186 Opt_resv_level,
187 Opt_dir_resv_level,
188 Opt_err,
189};
190
191static const match_table_t tokens = {
192 {Opt_barrier, "barrier=%u"},
193 {Opt_err_panic, "errors=panic"},
194 {Opt_err_ro, "errors=remount-ro"},
195 {Opt_intr, "intr"},
196 {Opt_nointr, "nointr"},
197 {Opt_hb_none, OCFS2_HB_NONE},
198 {Opt_hb_local, OCFS2_HB_LOCAL},
199 {Opt_hb_global, OCFS2_HB_GLOBAL},
200 {Opt_data_ordered, "data=ordered"},
201 {Opt_data_writeback, "data=writeback"},
202 {Opt_atime_quantum, "atime_quantum=%u"},
203 {Opt_slot, "preferred_slot=%u"},
204 {Opt_commit, "commit=%u"},
205 {Opt_localalloc, "localalloc=%d"},
206 {Opt_localflocks, "localflocks"},
207 {Opt_stack, "cluster_stack=%s"},
208 {Opt_user_xattr, "user_xattr"},
209 {Opt_nouser_xattr, "nouser_xattr"},
210 {Opt_inode64, "inode64"},
211 {Opt_acl, "acl"},
212 {Opt_noacl, "noacl"},
213 {Opt_usrquota, "usrquota"},
214 {Opt_grpquota, "grpquota"},
215 {Opt_coherency_buffered, "coherency=buffered"},
216 {Opt_coherency_full, "coherency=full"},
217 {Opt_resv_level, "resv_level=%u"},
218 {Opt_dir_resv_level, "dir_resv_level=%u"},
219 {Opt_err, NULL}
220};
221
222#ifdef CONFIG_DEBUG_FS
223static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
224{
225 struct ocfs2_cluster_connection *cconn = osb->cconn;
226 struct ocfs2_recovery_map *rm = osb->recovery_map;
227 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
228 int i, out = 0;
229
230 out += snprintf(buf + out, len - out,
231 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
232 "Device", osb->dev_str, osb->uuid_str,
233 osb->fs_generation, osb->vol_label);
234
235 out += snprintf(buf + out, len - out,
236 "%10s => State: %d Flags: 0x%lX\n", "Volume",
237 atomic_read(&osb->vol_state), osb->osb_flags);
238
239 out += snprintf(buf + out, len - out,
240 "%10s => Block: %lu Cluster: %d\n", "Sizes",
241 osb->sb->s_blocksize, osb->s_clustersize);
242
243 out += snprintf(buf + out, len - out,
244 "%10s => Compat: 0x%X Incompat: 0x%X "
245 "ROcompat: 0x%X\n",
246 "Features", osb->s_feature_compat,
247 osb->s_feature_incompat, osb->s_feature_ro_compat);
248
249 out += snprintf(buf + out, len - out,
250 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
251 osb->s_mount_opt, osb->s_atime_quantum);
252
253 if (cconn) {
254 out += snprintf(buf + out, len - out,
255 "%10s => Stack: %s Name: %*s "
256 "Version: %d.%d\n", "Cluster",
257 (*osb->osb_cluster_stack == '\0' ?
258 "o2cb" : osb->osb_cluster_stack),
259 cconn->cc_namelen, cconn->cc_name,
260 cconn->cc_version.pv_major,
261 cconn->cc_version.pv_minor);
262 }
263
264 spin_lock(&osb->dc_task_lock);
265 out += snprintf(buf + out, len - out,
266 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
267 "WorkSeq: %lu\n", "DownCnvt",
268 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
269 osb->blocked_lock_count, osb->dc_wake_sequence,
270 osb->dc_work_sequence);
271 spin_unlock(&osb->dc_task_lock);
272
273 spin_lock(&osb->osb_lock);
274 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
275 "Recovery",
276 (osb->recovery_thread_task ?
277 task_pid_nr(osb->recovery_thread_task) : -1));
278 if (rm->rm_used == 0)
279 out += snprintf(buf + out, len - out, " None\n");
280 else {
281 for (i = 0; i < rm->rm_used; i++)
282 out += snprintf(buf + out, len - out, " %d",
283 rm->rm_entries[i]);
284 out += snprintf(buf + out, len - out, "\n");
285 }
286 spin_unlock(&osb->osb_lock);
287
288 out += snprintf(buf + out, len - out,
289 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
290 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
291 osb->osb_commit_interval,
292 atomic_read(&osb->needs_checkpoint));
293
294 out += snprintf(buf + out, len - out,
295 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
296 "Journal", osb->journal->j_state,
297 osb->journal->j_trans_id,
298 atomic_read(&osb->journal->j_num_trans));
299
300 out += snprintf(buf + out, len - out,
301 "%10s => GlobalAllocs: %d LocalAllocs: %d "
302 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
303 "Stats",
304 atomic_read(&osb->alloc_stats.bitmap_data),
305 atomic_read(&osb->alloc_stats.local_data),
306 atomic_read(&osb->alloc_stats.bg_allocs),
307 atomic_read(&osb->alloc_stats.moves),
308 atomic_read(&osb->alloc_stats.bg_extends));
309
310 out += snprintf(buf + out, len - out,
311 "%10s => State: %u Descriptor: %llu Size: %u bits "
312 "Default: %u bits\n",
313 "LocalAlloc", osb->local_alloc_state,
314 (unsigned long long)osb->la_last_gd,
315 osb->local_alloc_bits, osb->local_alloc_default_bits);
316
317 spin_lock(&osb->osb_lock);
318 out += snprintf(buf + out, len - out,
319 "%10s => InodeSlot: %d StolenInodes: %d, "
320 "MetaSlot: %d StolenMeta: %d\n", "Steal",
321 osb->s_inode_steal_slot,
322 atomic_read(&osb->s_num_inodes_stolen),
323 osb->s_meta_steal_slot,
324 atomic_read(&osb->s_num_meta_stolen));
325 spin_unlock(&osb->osb_lock);
326
327 out += snprintf(buf + out, len - out, "OrphanScan => ");
328 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
329 os->os_count, os->os_seqno);
330 out += snprintf(buf + out, len - out, " Last Scan: ");
331 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
332 out += snprintf(buf + out, len - out, "Disabled\n");
333 else
334 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
335 (get_seconds() - os->os_scantime.tv_sec));
336
337 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
338 "Slots", "Num", "RecoGen");
339 for (i = 0; i < osb->max_slots; ++i) {
340 out += snprintf(buf + out, len - out,
341 "%10s %c %3d %10d\n",
342 " ",
343 (i == osb->slot_num ? '*' : ' '),
344 i, osb->slot_recovery_generations[i]);
345 }
346
347 return out;
348}
349
350static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
351{
352 struct ocfs2_super *osb = inode->i_private;
353 char *buf = NULL;
354
355 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
356 if (!buf)
357 goto bail;
358
359 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
360
361 file->private_data = buf;
362
363 return 0;
364bail:
365 return -ENOMEM;
366}
367
368static int ocfs2_debug_release(struct inode *inode, struct file *file)
369{
370 kfree(file->private_data);
371 return 0;
372}
373
374static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
375 size_t nbytes, loff_t *ppos)
376{
377 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
378 i_size_read(file->f_mapping->host));
379}
380#else
381static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
382{
383 return 0;
384}
385static int ocfs2_debug_release(struct inode *inode, struct file *file)
386{
387 return 0;
388}
389static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
390 size_t nbytes, loff_t *ppos)
391{
392 return 0;
393}
394#endif /* CONFIG_DEBUG_FS */
395
396static const struct file_operations ocfs2_osb_debug_fops = {
397 .open = ocfs2_osb_debug_open,
398 .release = ocfs2_debug_release,
399 .read = ocfs2_debug_read,
400 .llseek = generic_file_llseek,
401};
402
403static int ocfs2_sync_fs(struct super_block *sb, int wait)
404{
405 int status;
406 tid_t target;
407 struct ocfs2_super *osb = OCFS2_SB(sb);
408
409 if (ocfs2_is_hard_readonly(osb))
410 return -EROFS;
411
412 if (wait) {
413 status = ocfs2_flush_truncate_log(osb);
414 if (status < 0)
415 mlog_errno(status);
416 } else {
417 ocfs2_schedule_truncate_log_flush(osb, 0);
418 }
419
420 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
421 &target)) {
422 if (wait)
423 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
424 target);
425 }
426 return 0;
427}
428
429static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
430{
431 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
432 && (ino == USER_QUOTA_SYSTEM_INODE
433 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
434 return 0;
435 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
436 && (ino == GROUP_QUOTA_SYSTEM_INODE
437 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
438 return 0;
439 return 1;
440}
441
442static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
443{
444 struct inode *new = NULL;
445 int status = 0;
446 int i;
447
448 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 if (IS_ERR(new)) {
450 status = PTR_ERR(new);
451 mlog_errno(status);
452 goto bail;
453 }
454 osb->root_inode = new;
455
456 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
457 if (IS_ERR(new)) {
458 status = PTR_ERR(new);
459 mlog_errno(status);
460 goto bail;
461 }
462 osb->sys_root_inode = new;
463
464 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
465 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
466 if (!ocfs2_need_system_inode(osb, i))
467 continue;
468 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
469 if (!new) {
470 ocfs2_release_system_inodes(osb);
471 status = -EINVAL;
472 mlog_errno(status);
473 /* FIXME: Should ERROR_RO_FS */
474 mlog(ML_ERROR, "Unable to load system inode %d, "
475 "possibly corrupt fs?", i);
476 goto bail;
477 }
478 // the array now has one ref, so drop this one
479 iput(new);
480 }
481
482bail:
483 if (status)
484 mlog_errno(status);
485 return status;
486}
487
488static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
489{
490 struct inode *new = NULL;
491 int status = 0;
492 int i;
493
494 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
495 i < NUM_SYSTEM_INODES;
496 i++) {
497 if (!ocfs2_need_system_inode(osb, i))
498 continue;
499 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
500 if (!new) {
501 ocfs2_release_system_inodes(osb);
502 status = -EINVAL;
503 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
504 status, i, osb->slot_num);
505 goto bail;
506 }
507 /* the array now has one ref, so drop this one */
508 iput(new);
509 }
510
511bail:
512 if (status)
513 mlog_errno(status);
514 return status;
515}
516
517static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
518{
519 int i;
520 struct inode *inode;
521
522 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
523 inode = osb->global_system_inodes[i];
524 if (inode) {
525 iput(inode);
526 osb->global_system_inodes[i] = NULL;
527 }
528 }
529
530 inode = osb->sys_root_inode;
531 if (inode) {
532 iput(inode);
533 osb->sys_root_inode = NULL;
534 }
535
536 inode = osb->root_inode;
537 if (inode) {
538 iput(inode);
539 osb->root_inode = NULL;
540 }
541
542 if (!osb->local_system_inodes)
543 return;
544
545 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
546 if (osb->local_system_inodes[i]) {
547 iput(osb->local_system_inodes[i]);
548 osb->local_system_inodes[i] = NULL;
549 }
550 }
551
552 kfree(osb->local_system_inodes);
553 osb->local_system_inodes = NULL;
554}
555
556/* We're allocating fs objects, use GFP_NOFS */
557static struct inode *ocfs2_alloc_inode(struct super_block *sb)
558{
559 struct ocfs2_inode_info *oi;
560
561 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
562 if (!oi)
563 return NULL;
564
565 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
566 return &oi->vfs_inode;
567}
568
569static void ocfs2_i_callback(struct rcu_head *head)
570{
571 struct inode *inode = container_of(head, struct inode, i_rcu);
572 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
573}
574
575static void ocfs2_destroy_inode(struct inode *inode)
576{
577 call_rcu(&inode->i_rcu, ocfs2_i_callback);
578}
579
580static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
581 unsigned int cbits)
582{
583 unsigned int bytes = 1 << cbits;
584 unsigned int trim = bytes;
585 unsigned int bitshift = 32;
586
587 /*
588 * i_size and all block offsets in ocfs2 are always 64 bits
589 * wide. i_clusters is 32 bits, in cluster-sized units. So on
590 * 64 bit platforms, cluster size will be the limiting factor.
591 */
592
593#if BITS_PER_LONG == 32
594# if defined(CONFIG_LBDAF)
595 BUILD_BUG_ON(sizeof(sector_t) != 8);
596 /*
597 * We might be limited by page cache size.
598 */
599 if (bytes > PAGE_CACHE_SIZE) {
600 bytes = PAGE_CACHE_SIZE;
601 trim = 1;
602 /*
603 * Shift by 31 here so that we don't get larger than
604 * MAX_LFS_FILESIZE
605 */
606 bitshift = 31;
607 }
608# else
609 /*
610 * We are limited by the size of sector_t. Use block size, as
611 * that's what we expose to the VFS.
612 */
613 bytes = 1 << bbits;
614 trim = 1;
615 bitshift = 31;
616# endif
617#endif
618
619 /*
620 * Trim by a whole cluster when we can actually approach the
621 * on-disk limits. Otherwise we can overflow i_clusters when
622 * an extent start is at the max offset.
623 */
624 return (((unsigned long long)bytes) << bitshift) - trim;
625}
626
627static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
628{
629 int incompat_features;
630 int ret = 0;
631 struct mount_options parsed_options;
632 struct ocfs2_super *osb = OCFS2_SB(sb);
633 u32 tmp;
634
635 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
636 !ocfs2_check_set_options(sb, &parsed_options)) {
637 ret = -EINVAL;
638 goto out;
639 }
640
641 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
642 OCFS2_MOUNT_HB_NONE;
643 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
644 ret = -EINVAL;
645 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
646 goto out;
647 }
648
649 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
650 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
651 ret = -EINVAL;
652 mlog(ML_ERROR, "Cannot change data mode on remount\n");
653 goto out;
654 }
655
656 /* Probably don't want this on remount; it might
657 * mess with other nodes */
658 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
659 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
660 ret = -EINVAL;
661 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
662 goto out;
663 }
664
665 /* We're going to/from readonly mode. */
666 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
667 /* Disable quota accounting before remounting RO */
668 if (*flags & MS_RDONLY) {
669 ret = ocfs2_susp_quotas(osb, 0);
670 if (ret < 0)
671 goto out;
672 }
673 /* Lock here so the check of HARD_RO and the potential
674 * setting of SOFT_RO is atomic. */
675 spin_lock(&osb->osb_lock);
676 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
677 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
678 ret = -EROFS;
679 goto unlock_osb;
680 }
681
682 if (*flags & MS_RDONLY) {
683 sb->s_flags |= MS_RDONLY;
684 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
685 } else {
686 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
687 mlog(ML_ERROR, "Cannot remount RDWR "
688 "filesystem due to previous errors.\n");
689 ret = -EROFS;
690 goto unlock_osb;
691 }
692 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
693 if (incompat_features) {
694 mlog(ML_ERROR, "Cannot remount RDWR because "
695 "of unsupported optional features "
696 "(%x).\n", incompat_features);
697 ret = -EINVAL;
698 goto unlock_osb;
699 }
700 sb->s_flags &= ~MS_RDONLY;
701 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
702 }
703 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
704unlock_osb:
705 spin_unlock(&osb->osb_lock);
706 /* Enable quota accounting after remounting RW */
707 if (!ret && !(*flags & MS_RDONLY)) {
708 if (sb_any_quota_suspended(sb))
709 ret = ocfs2_susp_quotas(osb, 1);
710 else
711 ret = ocfs2_enable_quotas(osb);
712 if (ret < 0) {
713 /* Return back changes... */
714 spin_lock(&osb->osb_lock);
715 sb->s_flags |= MS_RDONLY;
716 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
717 spin_unlock(&osb->osb_lock);
718 goto out;
719 }
720 }
721 }
722
723 if (!ret) {
724 /* Only save off the new mount options in case of a successful
725 * remount. */
726 osb->s_mount_opt = parsed_options.mount_opt;
727 osb->s_atime_quantum = parsed_options.atime_quantum;
728 osb->preferred_slot = parsed_options.slot;
729 if (parsed_options.commit_interval)
730 osb->osb_commit_interval = parsed_options.commit_interval;
731
732 if (!ocfs2_is_hard_readonly(osb))
733 ocfs2_set_journal_params(osb);
734
735 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
736 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
737 MS_POSIXACL : 0);
738 }
739out:
740 return ret;
741}
742
743static int ocfs2_sb_probe(struct super_block *sb,
744 struct buffer_head **bh,
745 int *sector_size,
746 struct ocfs2_blockcheck_stats *stats)
747{
748 int status, tmpstat;
749 struct ocfs1_vol_disk_hdr *hdr;
750 struct ocfs2_dinode *di;
751 int blksize;
752
753 *bh = NULL;
754
755 /* may be > 512 */
756 *sector_size = bdev_logical_block_size(sb->s_bdev);
757 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
758 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
759 *sector_size, OCFS2_MAX_BLOCKSIZE);
760 status = -EINVAL;
761 goto bail;
762 }
763
764 /* Can this really happen? */
765 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
766 *sector_size = OCFS2_MIN_BLOCKSIZE;
767
768 /* check block zero for old format */
769 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
770 if (status < 0) {
771 mlog_errno(status);
772 goto bail;
773 }
774 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
775 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
776 mlog(ML_ERROR, "incompatible version: %u.%u\n",
777 hdr->major_version, hdr->minor_version);
778 status = -EINVAL;
779 }
780 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
781 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
782 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
783 hdr->signature);
784 status = -EINVAL;
785 }
786 brelse(*bh);
787 *bh = NULL;
788 if (status < 0) {
789 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
790 "upgraded before mounting with ocfs v2\n");
791 goto bail;
792 }
793
794 /*
795 * Now check at magic offset for 512, 1024, 2048, 4096
796 * blocksizes. 4096 is the maximum blocksize because it is
797 * the minimum clustersize.
798 */
799 status = -EINVAL;
800 for (blksize = *sector_size;
801 blksize <= OCFS2_MAX_BLOCKSIZE;
802 blksize <<= 1) {
803 tmpstat = ocfs2_get_sector(sb, bh,
804 OCFS2_SUPER_BLOCK_BLKNO,
805 blksize);
806 if (tmpstat < 0) {
807 status = tmpstat;
808 mlog_errno(status);
809 break;
810 }
811 di = (struct ocfs2_dinode *) (*bh)->b_data;
812 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
813 spin_lock_init(&stats->b_lock);
814 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
815 if (tmpstat < 0) {
816 brelse(*bh);
817 *bh = NULL;
818 }
819 if (tmpstat != -EAGAIN) {
820 status = tmpstat;
821 break;
822 }
823 }
824
825bail:
826 return status;
827}
828
829static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
830{
831 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
832
833 if (osb->s_mount_opt & hb_enabled) {
834 if (ocfs2_mount_local(osb)) {
835 mlog(ML_ERROR, "Cannot heartbeat on a locally "
836 "mounted device.\n");
837 return -EINVAL;
838 }
839 if (ocfs2_userspace_stack(osb)) {
840 mlog(ML_ERROR, "Userspace stack expected, but "
841 "o2cb heartbeat arguments passed to mount\n");
842 return -EINVAL;
843 }
844 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
845 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
846 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
847 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
848 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
849 return -EINVAL;
850 }
851 }
852
853 if (!(osb->s_mount_opt & hb_enabled)) {
854 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
855 !ocfs2_userspace_stack(osb)) {
856 mlog(ML_ERROR, "Heartbeat has to be started to mount "
857 "a read-write clustered device.\n");
858 return -EINVAL;
859 }
860 }
861
862 return 0;
863}
864
865/*
866 * If we're using a userspace stack, mount should have passed
867 * a name that matches the disk. If not, mount should not
868 * have passed a stack.
869 */
870static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
871 struct mount_options *mopt)
872{
873 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
874 mlog(ML_ERROR,
875 "cluster stack passed to mount, but this filesystem "
876 "does not support it\n");
877 return -EINVAL;
878 }
879
880 if (ocfs2_userspace_stack(osb) &&
881 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
882 OCFS2_STACK_LABEL_LEN)) {
883 mlog(ML_ERROR,
884 "cluster stack passed to mount (\"%s\") does not "
885 "match the filesystem (\"%s\")\n",
886 mopt->cluster_stack,
887 osb->osb_cluster_stack);
888 return -EINVAL;
889 }
890
891 return 0;
892}
893
894static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
895{
896 int type;
897 struct super_block *sb = osb->sb;
898 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
899 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
900 int status = 0;
901
902 for (type = 0; type < MAXQUOTAS; type++) {
903 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
904 continue;
905 if (unsuspend)
906 status = dquot_resume(sb, type);
907 else {
908 struct ocfs2_mem_dqinfo *oinfo;
909
910 /* Cancel periodic syncing before suspending */
911 oinfo = sb_dqinfo(sb, type)->dqi_priv;
912 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
913 status = dquot_suspend(sb, type);
914 }
915 if (status < 0)
916 break;
917 }
918 if (status < 0)
919 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
920 "remount (error = %d).\n", status);
921 return status;
922}
923
924static int ocfs2_enable_quotas(struct ocfs2_super *osb)
925{
926 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
927 struct super_block *sb = osb->sb;
928 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
929 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
930 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
931 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
932 int status;
933 int type;
934
935 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
936 for (type = 0; type < MAXQUOTAS; type++) {
937 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
938 continue;
939 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
940 osb->slot_num);
941 if (!inode[type]) {
942 status = -ENOENT;
943 goto out_quota_off;
944 }
945 status = dquot_enable(inode[type], type, QFMT_OCFS2,
946 DQUOT_USAGE_ENABLED);
947 if (status < 0)
948 goto out_quota_off;
949 }
950
951 for (type = 0; type < MAXQUOTAS; type++)
952 iput(inode[type]);
953 return 0;
954out_quota_off:
955 ocfs2_disable_quotas(osb);
956 for (type = 0; type < MAXQUOTAS; type++)
957 iput(inode[type]);
958 mlog_errno(status);
959 return status;
960}
961
962static void ocfs2_disable_quotas(struct ocfs2_super *osb)
963{
964 int type;
965 struct inode *inode;
966 struct super_block *sb = osb->sb;
967 struct ocfs2_mem_dqinfo *oinfo;
968
969 /* We mostly ignore errors in this function because there's not much
970 * we can do when we see them */
971 for (type = 0; type < MAXQUOTAS; type++) {
972 if (!sb_has_quota_loaded(sb, type))
973 continue;
974 /* Cancel periodic syncing before we grab dqonoff_mutex */
975 oinfo = sb_dqinfo(sb, type)->dqi_priv;
976 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
977 inode = igrab(sb->s_dquot.files[type]);
978 /* Turn off quotas. This will remove all dquot structures from
979 * memory and so they will be automatically synced to global
980 * quota files */
981 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
982 DQUOT_LIMITS_ENABLED);
983 if (!inode)
984 continue;
985 iput(inode);
986 }
987}
988
989/* Handle quota on quotactl */
990static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
991{
992 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
993 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
994
995 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
996 return -EINVAL;
997
998 return dquot_enable(sb_dqopt(sb)->files[type], type,
999 format_id, DQUOT_LIMITS_ENABLED);
1000}
1001
1002/* Handle quota off quotactl */
1003static int ocfs2_quota_off(struct super_block *sb, int type)
1004{
1005 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
1006}
1007
1008static const struct quotactl_ops ocfs2_quotactl_ops = {
1009 .quota_on_meta = ocfs2_quota_on,
1010 .quota_off = ocfs2_quota_off,
1011 .quota_sync = dquot_quota_sync,
1012 .get_info = dquot_get_dqinfo,
1013 .set_info = dquot_set_dqinfo,
1014 .get_dqblk = dquot_get_dqblk,
1015 .set_dqblk = dquot_set_dqblk,
1016};
1017
1018static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1019{
1020 struct dentry *root;
1021 int status, sector_size;
1022 struct mount_options parsed_options;
1023 struct inode *inode = NULL;
1024 struct ocfs2_super *osb = NULL;
1025 struct buffer_head *bh = NULL;
1026 char nodestr[8];
1027 struct ocfs2_blockcheck_stats stats;
1028
1029 trace_ocfs2_fill_super(sb, data, silent);
1030
1031 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
1032 status = -EINVAL;
1033 goto read_super_error;
1034 }
1035
1036 /* probe for superblock */
1037 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
1038 if (status < 0) {
1039 mlog(ML_ERROR, "superblock probe failed!\n");
1040 goto read_super_error;
1041 }
1042
1043 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1044 osb = OCFS2_SB(sb);
1045 if (status < 0) {
1046 mlog_errno(status);
1047 goto read_super_error;
1048 }
1049 brelse(bh);
1050 bh = NULL;
1051
1052 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1053 status = -EINVAL;
1054 goto read_super_error;
1055 }
1056 osb->s_mount_opt = parsed_options.mount_opt;
1057 osb->s_atime_quantum = parsed_options.atime_quantum;
1058 osb->preferred_slot = parsed_options.slot;
1059 osb->osb_commit_interval = parsed_options.commit_interval;
1060
1061 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1062 osb->osb_resv_level = parsed_options.resv_level;
1063 osb->osb_dir_resv_level = parsed_options.resv_level;
1064 if (parsed_options.dir_resv_level == -1)
1065 osb->osb_dir_resv_level = parsed_options.resv_level;
1066 else
1067 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1068
1069 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1070 if (status)
1071 goto read_super_error;
1072
1073 sb->s_magic = OCFS2_SUPER_MAGIC;
1074
1075 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
1076 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1077
1078 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1079 * heartbeat=none */
1080 if (bdev_read_only(sb->s_bdev)) {
1081 if (!(sb->s_flags & MS_RDONLY)) {
1082 status = -EACCES;
1083 mlog(ML_ERROR, "Readonly device detected but readonly "
1084 "mount was not specified.\n");
1085 goto read_super_error;
1086 }
1087
1088 /* You should not be able to start a local heartbeat
1089 * on a readonly device. */
1090 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1091 status = -EROFS;
1092 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1093 "device.\n");
1094 goto read_super_error;
1095 }
1096
1097 status = ocfs2_check_journals_nolocks(osb);
1098 if (status < 0) {
1099 if (status == -EROFS)
1100 mlog(ML_ERROR, "Recovery required on readonly "
1101 "file system, but write access is "
1102 "unavailable.\n");
1103 else
1104 mlog_errno(status);
1105 goto read_super_error;
1106 }
1107
1108 ocfs2_set_ro_flag(osb, 1);
1109
1110 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1111 "Cluster services will not be used for this mount. "
1112 "Recovery will be skipped.\n", osb->dev_str);
1113 }
1114
1115 if (!ocfs2_is_hard_readonly(osb)) {
1116 if (sb->s_flags & MS_RDONLY)
1117 ocfs2_set_ro_flag(osb, 0);
1118 }
1119
1120 status = ocfs2_verify_heartbeat(osb);
1121 if (status < 0) {
1122 mlog_errno(status);
1123 goto read_super_error;
1124 }
1125
1126 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1127 ocfs2_debugfs_root);
1128 if (!osb->osb_debug_root) {
1129 status = -EINVAL;
1130 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1131 goto read_super_error;
1132 }
1133
1134 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1135 osb->osb_debug_root,
1136 osb,
1137 &ocfs2_osb_debug_fops);
1138 if (!osb->osb_ctxt) {
1139 status = -EINVAL;
1140 mlog_errno(status);
1141 goto read_super_error;
1142 }
1143
1144 if (ocfs2_meta_ecc(osb)) {
1145 status = ocfs2_blockcheck_stats_debugfs_install(
1146 &osb->osb_ecc_stats,
1147 osb->osb_debug_root);
1148 if (status) {
1149 mlog(ML_ERROR,
1150 "Unable to create blockcheck statistics "
1151 "files\n");
1152 goto read_super_error;
1153 }
1154 }
1155
1156 status = ocfs2_mount_volume(sb);
1157 if (status < 0)
1158 goto read_super_error;
1159
1160 if (osb->root_inode)
1161 inode = igrab(osb->root_inode);
1162
1163 if (!inode) {
1164 status = -EIO;
1165 mlog_errno(status);
1166 goto read_super_error;
1167 }
1168
1169 root = d_make_root(inode);
1170 if (!root) {
1171 status = -ENOMEM;
1172 mlog_errno(status);
1173 goto read_super_error;
1174 }
1175
1176 sb->s_root = root;
1177
1178 ocfs2_complete_mount_recovery(osb);
1179
1180 if (ocfs2_mount_local(osb))
1181 snprintf(nodestr, sizeof(nodestr), "local");
1182 else
1183 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1184
1185 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1186 "with %s data mode.\n",
1187 osb->dev_str, nodestr, osb->slot_num,
1188 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1189 "ordered");
1190
1191 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1192 wake_up(&osb->osb_mount_event);
1193
1194 /* Now we can initialize quotas because we can afford to wait
1195 * for cluster locks recovery now. That also means that truncation
1196 * log recovery can happen but that waits for proper quota setup */
1197 if (!(sb->s_flags & MS_RDONLY)) {
1198 status = ocfs2_enable_quotas(osb);
1199 if (status < 0) {
1200 /* We have to err-out specially here because
1201 * s_root is already set */
1202 mlog_errno(status);
1203 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1204 wake_up(&osb->osb_mount_event);
1205 return status;
1206 }
1207 }
1208
1209 ocfs2_complete_quota_recovery(osb);
1210
1211 /* Now we wake up again for processes waiting for quotas */
1212 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1213 wake_up(&osb->osb_mount_event);
1214
1215 /* Start this when the mount is almost sure of being successful */
1216 ocfs2_orphan_scan_start(osb);
1217
1218 return status;
1219
1220read_super_error:
1221 brelse(bh);
1222
1223 if (osb) {
1224 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1225 wake_up(&osb->osb_mount_event);
1226 ocfs2_dismount_volume(sb, 1);
1227 }
1228
1229 if (status)
1230 mlog_errno(status);
1231 return status;
1232}
1233
1234static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1235 int flags,
1236 const char *dev_name,
1237 void *data)
1238{
1239 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1240}
1241
1242static void ocfs2_kill_sb(struct super_block *sb)
1243{
1244 struct ocfs2_super *osb = OCFS2_SB(sb);
1245
1246 /* Failed mount? */
1247 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1248 goto out;
1249
1250 /* Prevent further queueing of inode drop events */
1251 spin_lock(&dentry_list_lock);
1252 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1253 spin_unlock(&dentry_list_lock);
1254 /* Wait for work to finish and/or remove it */
1255 cancel_work_sync(&osb->dentry_lock_work);
1256out:
1257 kill_block_super(sb);
1258}
1259
1260static struct file_system_type ocfs2_fs_type = {
1261 .owner = THIS_MODULE,
1262 .name = "ocfs2",
1263 .mount = ocfs2_mount,
1264 .kill_sb = ocfs2_kill_sb,
1265
1266 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1267 .next = NULL
1268};
1269
1270static int ocfs2_check_set_options(struct super_block *sb,
1271 struct mount_options *options)
1272{
1273 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1274 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1275 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1276 mlog(ML_ERROR, "User quotas were requested, but this "
1277 "filesystem does not have the feature enabled.\n");
1278 return 0;
1279 }
1280 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1281 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1282 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1283 mlog(ML_ERROR, "Group quotas were requested, but this "
1284 "filesystem does not have the feature enabled.\n");
1285 return 0;
1286 }
1287 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1288 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1289 mlog(ML_ERROR, "ACL support requested but extended attributes "
1290 "feature is not enabled\n");
1291 return 0;
1292 }
1293 /* No ACL setting specified? Use XATTR feature... */
1294 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1295 OCFS2_MOUNT_NO_POSIX_ACL))) {
1296 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1297 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1298 else
1299 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1300 }
1301 return 1;
1302}
1303
1304static int ocfs2_parse_options(struct super_block *sb,
1305 char *options,
1306 struct mount_options *mopt,
1307 int is_remount)
1308{
1309 int status, user_stack = 0;
1310 char *p;
1311 u32 tmp;
1312
1313 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1314
1315 mopt->commit_interval = 0;
1316 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1317 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1318 mopt->slot = OCFS2_INVALID_SLOT;
1319 mopt->localalloc_opt = -1;
1320 mopt->cluster_stack[0] = '\0';
1321 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1322 mopt->dir_resv_level = -1;
1323
1324 if (!options) {
1325 status = 1;
1326 goto bail;
1327 }
1328
1329 while ((p = strsep(&options, ",")) != NULL) {
1330 int token, option;
1331 substring_t args[MAX_OPT_ARGS];
1332
1333 if (!*p)
1334 continue;
1335
1336 token = match_token(p, tokens, args);
1337 switch (token) {
1338 case Opt_hb_local:
1339 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1340 break;
1341 case Opt_hb_none:
1342 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1343 break;
1344 case Opt_hb_global:
1345 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1346 break;
1347 case Opt_barrier:
1348 if (match_int(&args[0], &option)) {
1349 status = 0;
1350 goto bail;
1351 }
1352 if (option)
1353 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1354 else
1355 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1356 break;
1357 case Opt_intr:
1358 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1359 break;
1360 case Opt_nointr:
1361 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1362 break;
1363 case Opt_err_panic:
1364 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1365 break;
1366 case Opt_err_ro:
1367 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1368 break;
1369 case Opt_data_ordered:
1370 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1371 break;
1372 case Opt_data_writeback:
1373 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1374 break;
1375 case Opt_user_xattr:
1376 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1377 break;
1378 case Opt_nouser_xattr:
1379 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1380 break;
1381 case Opt_atime_quantum:
1382 if (match_int(&args[0], &option)) {
1383 status = 0;
1384 goto bail;
1385 }
1386 if (option >= 0)
1387 mopt->atime_quantum = option;
1388 break;
1389 case Opt_slot:
1390 option = 0;
1391 if (match_int(&args[0], &option)) {
1392 status = 0;
1393 goto bail;
1394 }
1395 if (option)
1396 mopt->slot = (s16)option;
1397 break;
1398 case Opt_commit:
1399 option = 0;
1400 if (match_int(&args[0], &option)) {
1401 status = 0;
1402 goto bail;
1403 }
1404 if (option < 0)
1405 return 0;
1406 if (option == 0)
1407 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1408 mopt->commit_interval = HZ * option;
1409 break;
1410 case Opt_localalloc:
1411 option = 0;
1412 if (match_int(&args[0], &option)) {
1413 status = 0;
1414 goto bail;
1415 }
1416 if (option >= 0)
1417 mopt->localalloc_opt = option;
1418 break;
1419 case Opt_localflocks:
1420 /*
1421 * Changing this during remount could race
1422 * flock() requests, or "unbalance" existing
1423 * ones (e.g., a lock is taken in one mode but
1424 * dropped in the other). If users care enough
1425 * to flip locking modes during remount, we
1426 * could add a "local" flag to individual
1427 * flock structures for proper tracking of
1428 * state.
1429 */
1430 if (!is_remount)
1431 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1432 break;
1433 case Opt_stack:
1434 /* Check both that the option we were passed
1435 * is of the right length and that it is a proper
1436 * string of the right length.
1437 */
1438 if (((args[0].to - args[0].from) !=
1439 OCFS2_STACK_LABEL_LEN) ||
1440 (strnlen(args[0].from,
1441 OCFS2_STACK_LABEL_LEN) !=
1442 OCFS2_STACK_LABEL_LEN)) {
1443 mlog(ML_ERROR,
1444 "Invalid cluster_stack option\n");
1445 status = 0;
1446 goto bail;
1447 }
1448 memcpy(mopt->cluster_stack, args[0].from,
1449 OCFS2_STACK_LABEL_LEN);
1450 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1451 /*
1452 * Open code the memcmp here as we don't have
1453 * an osb to pass to
1454 * ocfs2_userspace_stack().
1455 */
1456 if (memcmp(mopt->cluster_stack,
1457 OCFS2_CLASSIC_CLUSTER_STACK,
1458 OCFS2_STACK_LABEL_LEN))
1459 user_stack = 1;
1460 break;
1461 case Opt_inode64:
1462 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1463 break;
1464 case Opt_usrquota:
1465 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1466 break;
1467 case Opt_grpquota:
1468 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1469 break;
1470 case Opt_coherency_buffered:
1471 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1472 break;
1473 case Opt_coherency_full:
1474 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1475 break;
1476 case Opt_acl:
1477 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1478 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1479 break;
1480 case Opt_noacl:
1481 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1482 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1483 break;
1484 case Opt_resv_level:
1485 if (is_remount)
1486 break;
1487 if (match_int(&args[0], &option)) {
1488 status = 0;
1489 goto bail;
1490 }
1491 if (option >= OCFS2_MIN_RESV_LEVEL &&
1492 option < OCFS2_MAX_RESV_LEVEL)
1493 mopt->resv_level = option;
1494 break;
1495 case Opt_dir_resv_level:
1496 if (is_remount)
1497 break;
1498 if (match_int(&args[0], &option)) {
1499 status = 0;
1500 goto bail;
1501 }
1502 if (option >= OCFS2_MIN_RESV_LEVEL &&
1503 option < OCFS2_MAX_RESV_LEVEL)
1504 mopt->dir_resv_level = option;
1505 break;
1506 default:
1507 mlog(ML_ERROR,
1508 "Unrecognized mount option \"%s\" "
1509 "or missing value\n", p);
1510 status = 0;
1511 goto bail;
1512 }
1513 }
1514
1515 if (user_stack == 0) {
1516 /* Ensure only one heartbeat mode */
1517 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1518 OCFS2_MOUNT_HB_GLOBAL |
1519 OCFS2_MOUNT_HB_NONE);
1520 if (hweight32(tmp) != 1) {
1521 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1522 status = 0;
1523 goto bail;
1524 }
1525 }
1526
1527 status = 1;
1528
1529bail:
1530 return status;
1531}
1532
1533static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1534{
1535 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1536 unsigned long opts = osb->s_mount_opt;
1537 unsigned int local_alloc_megs;
1538
1539 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1540 seq_printf(s, ",_netdev");
1541 if (opts & OCFS2_MOUNT_HB_LOCAL)
1542 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1543 else
1544 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1545 } else
1546 seq_printf(s, ",%s", OCFS2_HB_NONE);
1547
1548 if (opts & OCFS2_MOUNT_NOINTR)
1549 seq_printf(s, ",nointr");
1550
1551 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1552 seq_printf(s, ",data=writeback");
1553 else
1554 seq_printf(s, ",data=ordered");
1555
1556 if (opts & OCFS2_MOUNT_BARRIER)
1557 seq_printf(s, ",barrier=1");
1558
1559 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1560 seq_printf(s, ",errors=panic");
1561 else
1562 seq_printf(s, ",errors=remount-ro");
1563
1564 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1565 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1566
1567 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1568
1569 if (osb->osb_commit_interval)
1570 seq_printf(s, ",commit=%u",
1571 (unsigned) (osb->osb_commit_interval / HZ));
1572
1573 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1574 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1575 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1576
1577 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1578 seq_printf(s, ",localflocks,");
1579
1580 if (osb->osb_cluster_stack[0])
1581 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1582 osb->osb_cluster_stack);
1583 if (opts & OCFS2_MOUNT_USRQUOTA)
1584 seq_printf(s, ",usrquota");
1585 if (opts & OCFS2_MOUNT_GRPQUOTA)
1586 seq_printf(s, ",grpquota");
1587
1588 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1589 seq_printf(s, ",coherency=buffered");
1590 else
1591 seq_printf(s, ",coherency=full");
1592
1593 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1594 seq_printf(s, ",nouser_xattr");
1595 else
1596 seq_printf(s, ",user_xattr");
1597
1598 if (opts & OCFS2_MOUNT_INODE64)
1599 seq_printf(s, ",inode64");
1600
1601 if (opts & OCFS2_MOUNT_POSIX_ACL)
1602 seq_printf(s, ",acl");
1603 else
1604 seq_printf(s, ",noacl");
1605
1606 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1607 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1608
1609 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1610 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1611
1612 return 0;
1613}
1614
1615wait_queue_head_t ocfs2__ioend_wq[OCFS2_IOEND_WQ_HASH_SZ];
1616
1617static int __init ocfs2_init(void)
1618{
1619 int status, i;
1620
1621 ocfs2_print_version();
1622
1623 for (i = 0; i < OCFS2_IOEND_WQ_HASH_SZ; i++)
1624 init_waitqueue_head(&ocfs2__ioend_wq[i]);
1625
1626 status = init_ocfs2_uptodate_cache();
1627 if (status < 0)
1628 goto out1;
1629
1630 status = ocfs2_initialize_mem_caches();
1631 if (status < 0)
1632 goto out2;
1633
1634 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1635 if (!ocfs2_wq) {
1636 status = -ENOMEM;
1637 goto out3;
1638 }
1639
1640 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1641 if (!ocfs2_debugfs_root) {
1642 status = -EFAULT;
1643 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1644 }
1645
1646 ocfs2_set_locking_protocol();
1647
1648 status = register_quota_format(&ocfs2_quota_format);
1649 if (status < 0)
1650 goto out4;
1651 status = register_filesystem(&ocfs2_fs_type);
1652 if (!status)
1653 return 0;
1654
1655 unregister_quota_format(&ocfs2_quota_format);
1656out4:
1657 destroy_workqueue(ocfs2_wq);
1658 debugfs_remove(ocfs2_debugfs_root);
1659out3:
1660 ocfs2_free_mem_caches();
1661out2:
1662 exit_ocfs2_uptodate_cache();
1663out1:
1664 mlog_errno(status);
1665 return status;
1666}
1667
1668static void __exit ocfs2_exit(void)
1669{
1670 if (ocfs2_wq) {
1671 flush_workqueue(ocfs2_wq);
1672 destroy_workqueue(ocfs2_wq);
1673 }
1674
1675 unregister_quota_format(&ocfs2_quota_format);
1676
1677 debugfs_remove(ocfs2_debugfs_root);
1678
1679 ocfs2_free_mem_caches();
1680
1681 unregister_filesystem(&ocfs2_fs_type);
1682
1683 exit_ocfs2_uptodate_cache();
1684}
1685
1686static void ocfs2_put_super(struct super_block *sb)
1687{
1688 trace_ocfs2_put_super(sb);
1689
1690 ocfs2_sync_blockdev(sb);
1691 ocfs2_dismount_volume(sb, 0);
1692}
1693
1694static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1695{
1696 struct ocfs2_super *osb;
1697 u32 numbits, freebits;
1698 int status;
1699 struct ocfs2_dinode *bm_lock;
1700 struct buffer_head *bh = NULL;
1701 struct inode *inode = NULL;
1702
1703 trace_ocfs2_statfs(dentry->d_sb, buf);
1704
1705 osb = OCFS2_SB(dentry->d_sb);
1706
1707 inode = ocfs2_get_system_file_inode(osb,
1708 GLOBAL_BITMAP_SYSTEM_INODE,
1709 OCFS2_INVALID_SLOT);
1710 if (!inode) {
1711 mlog(ML_ERROR, "failed to get bitmap inode\n");
1712 status = -EIO;
1713 goto bail;
1714 }
1715
1716 status = ocfs2_inode_lock(inode, &bh, 0);
1717 if (status < 0) {
1718 mlog_errno(status);
1719 goto bail;
1720 }
1721
1722 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1723
1724 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1725 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1726
1727 buf->f_type = OCFS2_SUPER_MAGIC;
1728 buf->f_bsize = dentry->d_sb->s_blocksize;
1729 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1730 buf->f_blocks = ((sector_t) numbits) *
1731 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1732 buf->f_bfree = ((sector_t) freebits) *
1733 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1734 buf->f_bavail = buf->f_bfree;
1735 buf->f_files = numbits;
1736 buf->f_ffree = freebits;
1737 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1738 & 0xFFFFFFFFUL;
1739 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1740 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1741
1742 brelse(bh);
1743
1744 ocfs2_inode_unlock(inode, 0);
1745 status = 0;
1746bail:
1747 if (inode)
1748 iput(inode);
1749
1750 if (status)
1751 mlog_errno(status);
1752
1753 return status;
1754}
1755
1756static void ocfs2_inode_init_once(void *data)
1757{
1758 struct ocfs2_inode_info *oi = data;
1759
1760 oi->ip_flags = 0;
1761 oi->ip_open_count = 0;
1762 spin_lock_init(&oi->ip_lock);
1763 ocfs2_extent_map_init(&oi->vfs_inode);
1764 INIT_LIST_HEAD(&oi->ip_io_markers);
1765 oi->ip_dir_start_lookup = 0;
1766 atomic_set(&oi->ip_unaligned_aio, 0);
1767 init_rwsem(&oi->ip_alloc_sem);
1768 init_rwsem(&oi->ip_xattr_sem);
1769 mutex_init(&oi->ip_io_mutex);
1770
1771 oi->ip_blkno = 0ULL;
1772 oi->ip_clusters = 0;
1773
1774 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1775
1776 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1777 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1778 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1779
1780 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1781 &ocfs2_inode_caching_ops);
1782
1783 inode_init_once(&oi->vfs_inode);
1784}
1785
1786static int ocfs2_initialize_mem_caches(void)
1787{
1788 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1789 sizeof(struct ocfs2_inode_info),
1790 0,
1791 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1792 SLAB_MEM_SPREAD),
1793 ocfs2_inode_init_once);
1794 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1795 sizeof(struct ocfs2_dquot),
1796 0,
1797 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1798 SLAB_MEM_SPREAD),
1799 NULL);
1800 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1801 sizeof(struct ocfs2_quota_chunk),
1802 0,
1803 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1804 NULL);
1805 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1806 !ocfs2_qf_chunk_cachep) {
1807 if (ocfs2_inode_cachep)
1808 kmem_cache_destroy(ocfs2_inode_cachep);
1809 if (ocfs2_dquot_cachep)
1810 kmem_cache_destroy(ocfs2_dquot_cachep);
1811 if (ocfs2_qf_chunk_cachep)
1812 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1813 return -ENOMEM;
1814 }
1815
1816 return 0;
1817}
1818
1819static void ocfs2_free_mem_caches(void)
1820{
1821 if (ocfs2_inode_cachep)
1822 kmem_cache_destroy(ocfs2_inode_cachep);
1823 ocfs2_inode_cachep = NULL;
1824
1825 if (ocfs2_dquot_cachep)
1826 kmem_cache_destroy(ocfs2_dquot_cachep);
1827 ocfs2_dquot_cachep = NULL;
1828
1829 if (ocfs2_qf_chunk_cachep)
1830 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1831 ocfs2_qf_chunk_cachep = NULL;
1832}
1833
1834static int ocfs2_get_sector(struct super_block *sb,
1835 struct buffer_head **bh,
1836 int block,
1837 int sect_size)
1838{
1839 if (!sb_set_blocksize(sb, sect_size)) {
1840 mlog(ML_ERROR, "unable to set blocksize\n");
1841 return -EIO;
1842 }
1843
1844 *bh = sb_getblk(sb, block);
1845 if (!*bh) {
1846 mlog_errno(-EIO);
1847 return -EIO;
1848 }
1849 lock_buffer(*bh);
1850 if (!buffer_dirty(*bh))
1851 clear_buffer_uptodate(*bh);
1852 unlock_buffer(*bh);
1853 ll_rw_block(READ, 1, bh);
1854 wait_on_buffer(*bh);
1855 if (!buffer_uptodate(*bh)) {
1856 mlog_errno(-EIO);
1857 brelse(*bh);
1858 *bh = NULL;
1859 return -EIO;
1860 }
1861
1862 return 0;
1863}
1864
1865static int ocfs2_mount_volume(struct super_block *sb)
1866{
1867 int status = 0;
1868 int unlock_super = 0;
1869 struct ocfs2_super *osb = OCFS2_SB(sb);
1870
1871 if (ocfs2_is_hard_readonly(osb))
1872 goto leave;
1873
1874 status = ocfs2_dlm_init(osb);
1875 if (status < 0) {
1876 mlog_errno(status);
1877 goto leave;
1878 }
1879
1880 status = ocfs2_super_lock(osb, 1);
1881 if (status < 0) {
1882 mlog_errno(status);
1883 goto leave;
1884 }
1885 unlock_super = 1;
1886
1887 /* This will load up the node map and add ourselves to it. */
1888 status = ocfs2_find_slot(osb);
1889 if (status < 0) {
1890 mlog_errno(status);
1891 goto leave;
1892 }
1893
1894 /* load all node-local system inodes */
1895 status = ocfs2_init_local_system_inodes(osb);
1896 if (status < 0) {
1897 mlog_errno(status);
1898 goto leave;
1899 }
1900
1901 status = ocfs2_check_volume(osb);
1902 if (status < 0) {
1903 mlog_errno(status);
1904 goto leave;
1905 }
1906
1907 status = ocfs2_truncate_log_init(osb);
1908 if (status < 0)
1909 mlog_errno(status);
1910
1911leave:
1912 if (unlock_super)
1913 ocfs2_super_unlock(osb, 1);
1914
1915 return status;
1916}
1917
1918static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1919{
1920 int tmp, hangup_needed = 0;
1921 struct ocfs2_super *osb = NULL;
1922 char nodestr[8];
1923
1924 trace_ocfs2_dismount_volume(sb);
1925
1926 BUG_ON(!sb);
1927 osb = OCFS2_SB(sb);
1928 BUG_ON(!osb);
1929
1930 debugfs_remove(osb->osb_ctxt);
1931
1932 /*
1933 * Flush inode dropping work queue so that deletes are
1934 * performed while the filesystem is still working
1935 */
1936 ocfs2_drop_all_dl_inodes(osb);
1937
1938 /* Orphan scan should be stopped as early as possible */
1939 ocfs2_orphan_scan_stop(osb);
1940
1941 ocfs2_disable_quotas(osb);
1942
1943 ocfs2_shutdown_local_alloc(osb);
1944
1945 ocfs2_truncate_log_shutdown(osb);
1946
1947 /* This will disable recovery and flush any recovery work. */
1948 ocfs2_recovery_exit(osb);
1949
1950 ocfs2_journal_shutdown(osb);
1951
1952 ocfs2_sync_blockdev(sb);
1953
1954 ocfs2_purge_refcount_trees(osb);
1955
1956 /* No cluster connection means we've failed during mount, so skip
1957 * all the steps which depended on that to complete. */
1958 if (osb->cconn) {
1959 tmp = ocfs2_super_lock(osb, 1);
1960 if (tmp < 0) {
1961 mlog_errno(tmp);
1962 return;
1963 }
1964 }
1965
1966 if (osb->slot_num != OCFS2_INVALID_SLOT)
1967 ocfs2_put_slot(osb);
1968
1969 if (osb->cconn)
1970 ocfs2_super_unlock(osb, 1);
1971
1972 ocfs2_release_system_inodes(osb);
1973
1974 /*
1975 * If we're dismounting due to mount error, mount.ocfs2 will clean
1976 * up heartbeat. If we're a local mount, there is no heartbeat.
1977 * If we failed before we got a uuid_str yet, we can't stop
1978 * heartbeat. Otherwise, do it.
1979 */
1980 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1981 !ocfs2_is_hard_readonly(osb))
1982 hangup_needed = 1;
1983
1984 if (osb->cconn)
1985 ocfs2_dlm_shutdown(osb, hangup_needed);
1986
1987 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1988 debugfs_remove(osb->osb_debug_root);
1989
1990 if (hangup_needed)
1991 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1992
1993 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1994
1995 if (ocfs2_mount_local(osb))
1996 snprintf(nodestr, sizeof(nodestr), "local");
1997 else
1998 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1999
2000 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2001 osb->dev_str, nodestr);
2002
2003 ocfs2_delete_osb(osb);
2004 kfree(osb);
2005 sb->s_dev = 0;
2006 sb->s_fs_info = NULL;
2007}
2008
2009static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2010 unsigned uuid_bytes)
2011{
2012 int i, ret;
2013 char *ptr;
2014
2015 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2016
2017 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
2018 if (osb->uuid_str == NULL)
2019 return -ENOMEM;
2020
2021 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2022 /* print with null */
2023 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2024 if (ret != 2) /* drop super cleans up */
2025 return -EINVAL;
2026 /* then only advance past the last char */
2027 ptr += 2;
2028 }
2029
2030 return 0;
2031}
2032
2033/* Make sure entire volume is addressable by our journal. Requires
2034 osb_clusters_at_boot to be valid and for the journal to have been
2035 initialized by ocfs2_journal_init(). */
2036static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2037{
2038 int status = 0;
2039 u64 max_block =
2040 ocfs2_clusters_to_blocks(osb->sb,
2041 osb->osb_clusters_at_boot) - 1;
2042
2043 /* 32-bit block number is always OK. */
2044 if (max_block <= (u32)~0ULL)
2045 goto out;
2046
2047 /* Volume is "huge", so see if our journal is new enough to
2048 support it. */
2049 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2050 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2051 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2052 JBD2_FEATURE_INCOMPAT_64BIT))) {
2053 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2054 "Enable the 'block64' journal option with tunefs.ocfs2");
2055 status = -EFBIG;
2056 goto out;
2057 }
2058
2059 out:
2060 return status;
2061}
2062
2063static int ocfs2_initialize_super(struct super_block *sb,
2064 struct buffer_head *bh,
2065 int sector_size,
2066 struct ocfs2_blockcheck_stats *stats)
2067{
2068 int status;
2069 int i, cbits, bbits;
2070 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2071 struct inode *inode = NULL;
2072 struct ocfs2_journal *journal;
2073 __le32 uuid_net_key;
2074 struct ocfs2_super *osb;
2075 u64 total_blocks;
2076
2077 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2078 if (!osb) {
2079 status = -ENOMEM;
2080 mlog_errno(status);
2081 goto bail;
2082 }
2083
2084 sb->s_fs_info = osb;
2085 sb->s_op = &ocfs2_sops;
2086 sb->s_d_op = &ocfs2_dentry_ops;
2087 sb->s_export_op = &ocfs2_export_ops;
2088 sb->s_qcop = &ocfs2_quotactl_ops;
2089 sb->dq_op = &ocfs2_quota_operations;
2090 sb->s_xattr = ocfs2_xattr_handlers;
2091 sb->s_time_gran = 1;
2092 sb->s_flags |= MS_NOATIME;
2093 /* this is needed to support O_LARGEFILE */
2094 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2095 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2096 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2097
2098 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2099
2100 for (i = 0; i < 3; i++)
2101 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2102 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2103
2104 osb->sb = sb;
2105 /* Save off for ocfs2_rw_direct */
2106 osb->s_sectsize_bits = blksize_bits(sector_size);
2107 BUG_ON(!osb->s_sectsize_bits);
2108
2109 spin_lock_init(&osb->dc_task_lock);
2110 init_waitqueue_head(&osb->dc_event);
2111 osb->dc_work_sequence = 0;
2112 osb->dc_wake_sequence = 0;
2113 INIT_LIST_HEAD(&osb->blocked_lock_list);
2114 osb->blocked_lock_count = 0;
2115 spin_lock_init(&osb->osb_lock);
2116 spin_lock_init(&osb->osb_xattr_lock);
2117 ocfs2_init_steal_slots(osb);
2118
2119 atomic_set(&osb->alloc_stats.moves, 0);
2120 atomic_set(&osb->alloc_stats.local_data, 0);
2121 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2122 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2123 atomic_set(&osb->alloc_stats.bg_extends, 0);
2124
2125 /* Copy the blockcheck stats from the superblock probe */
2126 osb->osb_ecc_stats = *stats;
2127
2128 ocfs2_init_node_maps(osb);
2129
2130 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2131 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2132
2133 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2134 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2135 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2136 osb->max_slots);
2137 status = -EINVAL;
2138 goto bail;
2139 }
2140
2141 ocfs2_orphan_scan_init(osb);
2142
2143 status = ocfs2_recovery_init(osb);
2144 if (status) {
2145 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2146 mlog_errno(status);
2147 goto bail;
2148 }
2149
2150 init_waitqueue_head(&osb->checkpoint_event);
2151 atomic_set(&osb->needs_checkpoint, 0);
2152
2153 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2154
2155 osb->slot_num = OCFS2_INVALID_SLOT;
2156
2157 osb->s_xattr_inline_size = le16_to_cpu(
2158 di->id2.i_super.s_xattr_inline_size);
2159
2160 osb->local_alloc_state = OCFS2_LA_UNUSED;
2161 osb->local_alloc_bh = NULL;
2162 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2163
2164 init_waitqueue_head(&osb->osb_mount_event);
2165
2166 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2167 if (status) {
2168 mlog_errno(status);
2169 goto bail;
2170 }
2171
2172 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2173 if (!osb->vol_label) {
2174 mlog(ML_ERROR, "unable to alloc vol label\n");
2175 status = -ENOMEM;
2176 goto bail;
2177 }
2178
2179 osb->slot_recovery_generations =
2180 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2181 GFP_KERNEL);
2182 if (!osb->slot_recovery_generations) {
2183 status = -ENOMEM;
2184 mlog_errno(status);
2185 goto bail;
2186 }
2187
2188 init_waitqueue_head(&osb->osb_wipe_event);
2189 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2190 sizeof(*osb->osb_orphan_wipes),
2191 GFP_KERNEL);
2192 if (!osb->osb_orphan_wipes) {
2193 status = -ENOMEM;
2194 mlog_errno(status);
2195 goto bail;
2196 }
2197
2198 osb->osb_rf_lock_tree = RB_ROOT;
2199
2200 osb->s_feature_compat =
2201 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2202 osb->s_feature_ro_compat =
2203 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2204 osb->s_feature_incompat =
2205 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2206
2207 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2208 mlog(ML_ERROR, "couldn't mount because of unsupported "
2209 "optional features (%x).\n", i);
2210 status = -EINVAL;
2211 goto bail;
2212 }
2213 if (!(osb->sb->s_flags & MS_RDONLY) &&
2214 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2215 mlog(ML_ERROR, "couldn't mount RDWR because of "
2216 "unsupported optional features (%x).\n", i);
2217 status = -EINVAL;
2218 goto bail;
2219 }
2220
2221 if (ocfs2_clusterinfo_valid(osb)) {
2222 osb->osb_stackflags =
2223 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2224 memcpy(osb->osb_cluster_stack,
2225 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2226 OCFS2_STACK_LABEL_LEN);
2227 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2228 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2229 mlog(ML_ERROR,
2230 "couldn't mount because of an invalid "
2231 "cluster stack label (%s) \n",
2232 osb->osb_cluster_stack);
2233 status = -EINVAL;
2234 goto bail;
2235 }
2236 } else {
2237 /* The empty string is identical with classic tools that
2238 * don't know about s_cluster_info. */
2239 osb->osb_cluster_stack[0] = '\0';
2240 }
2241
2242 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2243
2244 /* FIXME
2245 * This should be done in ocfs2_journal_init(), but unknown
2246 * ordering issues will cause the filesystem to crash.
2247 * If anyone wants to figure out what part of the code
2248 * refers to osb->journal before ocfs2_journal_init() is run,
2249 * be my guest.
2250 */
2251 /* initialize our journal structure */
2252
2253 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2254 if (!journal) {
2255 mlog(ML_ERROR, "unable to alloc journal\n");
2256 status = -ENOMEM;
2257 goto bail;
2258 }
2259 osb->journal = journal;
2260 journal->j_osb = osb;
2261
2262 atomic_set(&journal->j_num_trans, 0);
2263 init_rwsem(&journal->j_trans_barrier);
2264 init_waitqueue_head(&journal->j_checkpointed);
2265 spin_lock_init(&journal->j_lock);
2266 journal->j_trans_id = (unsigned long) 1;
2267 INIT_LIST_HEAD(&journal->j_la_cleanups);
2268 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2269 journal->j_state = OCFS2_JOURNAL_FREE;
2270
2271 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2272 osb->dentry_lock_list = NULL;
2273
2274 /* get some pseudo constants for clustersize bits */
2275 osb->s_clustersize_bits =
2276 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2277 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2278
2279 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2280 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2281 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2282 osb->s_clustersize);
2283 status = -EINVAL;
2284 goto bail;
2285 }
2286
2287 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2288 le32_to_cpu(di->i_clusters));
2289
2290 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2291 total_blocks);
2292 if (status) {
2293 mlog(ML_ERROR, "Volume too large "
2294 "to mount safely on this system");
2295 status = -EFBIG;
2296 goto bail;
2297 }
2298
2299 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2300 sizeof(di->id2.i_super.s_uuid))) {
2301 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2302 status = -ENOMEM;
2303 goto bail;
2304 }
2305
2306 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2307
2308 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2309 osb->vol_label[63] = '\0';
2310 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2311 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2312 osb->first_cluster_group_blkno =
2313 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2314 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2315 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2316 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2317 (unsigned long long)osb->root_blkno,
2318 (unsigned long long)osb->system_dir_blkno,
2319 osb->s_clustersize_bits);
2320
2321 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2322 if (!osb->osb_dlm_debug) {
2323 status = -ENOMEM;
2324 mlog_errno(status);
2325 goto bail;
2326 }
2327
2328 atomic_set(&osb->vol_state, VOLUME_INIT);
2329
2330 /* load root, system_dir, and all global system inodes */
2331 status = ocfs2_init_global_system_inodes(osb);
2332 if (status < 0) {
2333 mlog_errno(status);
2334 goto bail;
2335 }
2336
2337 /*
2338 * global bitmap
2339 */
2340 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2341 OCFS2_INVALID_SLOT);
2342 if (!inode) {
2343 status = -EINVAL;
2344 mlog_errno(status);
2345 goto bail;
2346 }
2347
2348 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2349 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2350 iput(inode);
2351
2352 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2353 osb->s_feature_incompat) * 8;
2354
2355 status = ocfs2_init_slot_info(osb);
2356 if (status < 0) {
2357 mlog_errno(status);
2358 goto bail;
2359 }
2360 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
2361
2362bail:
2363 return status;
2364}
2365
2366/*
2367 * will return: -EAGAIN if it is ok to keep searching for superblocks
2368 * -EINVAL if there is a bad superblock
2369 * 0 on success
2370 */
2371static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2372 struct buffer_head *bh,
2373 u32 blksz,
2374 struct ocfs2_blockcheck_stats *stats)
2375{
2376 int status = -EAGAIN;
2377
2378 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2379 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2380 /* We have to do a raw check of the feature here */
2381 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2382 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2383 status = ocfs2_block_check_validate(bh->b_data,
2384 bh->b_size,
2385 &di->i_check,
2386 stats);
2387 if (status)
2388 goto out;
2389 }
2390 status = -EINVAL;
2391 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2392 mlog(ML_ERROR, "found superblock with incorrect block "
2393 "size: found %u, should be %u\n",
2394 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2395 blksz);
2396 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2397 OCFS2_MAJOR_REV_LEVEL ||
2398 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2399 OCFS2_MINOR_REV_LEVEL) {
2400 mlog(ML_ERROR, "found superblock with bad version: "
2401 "found %u.%u, should be %u.%u\n",
2402 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2403 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2404 OCFS2_MAJOR_REV_LEVEL,
2405 OCFS2_MINOR_REV_LEVEL);
2406 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2407 mlog(ML_ERROR, "bad block number on superblock: "
2408 "found %llu, should be %llu\n",
2409 (unsigned long long)le64_to_cpu(di->i_blkno),
2410 (unsigned long long)bh->b_blocknr);
2411 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2412 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2413 mlog(ML_ERROR, "bad cluster size found: %u\n",
2414 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2415 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2416 mlog(ML_ERROR, "bad root_blkno: 0\n");
2417 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2418 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2419 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2420 mlog(ML_ERROR,
2421 "Superblock slots found greater than file system "
2422 "maximum: found %u, max %u\n",
2423 le16_to_cpu(di->id2.i_super.s_max_slots),
2424 OCFS2_MAX_SLOTS);
2425 } else {
2426 /* found it! */
2427 status = 0;
2428 }
2429 }
2430
2431out:
2432 if (status && status != -EAGAIN)
2433 mlog_errno(status);
2434 return status;
2435}
2436
2437static int ocfs2_check_volume(struct ocfs2_super *osb)
2438{
2439 int status;
2440 int dirty;
2441 int local;
2442 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2443 * recover
2444 * ourselves. */
2445
2446 /* Init our journal object. */
2447 status = ocfs2_journal_init(osb->journal, &dirty);
2448 if (status < 0) {
2449 mlog(ML_ERROR, "Could not initialize journal!\n");
2450 goto finally;
2451 }
2452
2453 /* Now that journal has been initialized, check to make sure
2454 entire volume is addressable. */
2455 status = ocfs2_journal_addressable(osb);
2456 if (status)
2457 goto finally;
2458
2459 /* If the journal was unmounted cleanly then we don't want to
2460 * recover anything. Otherwise, journal_load will do that
2461 * dirty work for us :) */
2462 if (!dirty) {
2463 status = ocfs2_journal_wipe(osb->journal, 0);
2464 if (status < 0) {
2465 mlog_errno(status);
2466 goto finally;
2467 }
2468 } else {
2469 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2470 "unmounted cleanly, recovering it.\n", osb->dev_str);
2471 }
2472
2473 local = ocfs2_mount_local(osb);
2474
2475 /* will play back anything left in the journal. */
2476 status = ocfs2_journal_load(osb->journal, local, dirty);
2477 if (status < 0) {
2478 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2479 goto finally;
2480 }
2481
2482 if (dirty) {
2483 /* recover my local alloc if we didn't unmount cleanly. */
2484 status = ocfs2_begin_local_alloc_recovery(osb,
2485 osb->slot_num,
2486 &local_alloc);
2487 if (status < 0) {
2488 mlog_errno(status);
2489 goto finally;
2490 }
2491 /* we complete the recovery process after we've marked
2492 * ourselves as mounted. */
2493 }
2494
2495 status = ocfs2_load_local_alloc(osb);
2496 if (status < 0) {
2497 mlog_errno(status);
2498 goto finally;
2499 }
2500
2501 if (dirty) {
2502 /* Recovery will be completed after we've mounted the
2503 * rest of the volume. */
2504 osb->dirty = 1;
2505 osb->local_alloc_copy = local_alloc;
2506 local_alloc = NULL;
2507 }
2508
2509 /* go through each journal, trylock it and if you get the
2510 * lock, and it's marked as dirty, set the bit in the recover
2511 * map and launch a recovery thread for it. */
2512 status = ocfs2_mark_dead_nodes(osb);
2513 if (status < 0) {
2514 mlog_errno(status);
2515 goto finally;
2516 }
2517
2518 status = ocfs2_compute_replay_slots(osb);
2519 if (status < 0)
2520 mlog_errno(status);
2521
2522finally:
2523 if (local_alloc)
2524 kfree(local_alloc);
2525
2526 if (status)
2527 mlog_errno(status);
2528 return status;
2529}
2530
2531/*
2532 * The routine gets called from dismount or close whenever a dismount on
2533 * volume is requested and the osb open count becomes 1.
2534 * It will remove the osb from the global list and also free up all the
2535 * initialized resources and fileobject.
2536 */
2537static void ocfs2_delete_osb(struct ocfs2_super *osb)
2538{
2539 /* This function assumes that the caller has the main osb resource */
2540
2541 ocfs2_free_slot_info(osb);
2542
2543 kfree(osb->osb_orphan_wipes);
2544 kfree(osb->slot_recovery_generations);
2545 /* FIXME
2546 * This belongs in journal shutdown, but because we have to
2547 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2548 * we free it here.
2549 */
2550 kfree(osb->journal);
2551 if (osb->local_alloc_copy)
2552 kfree(osb->local_alloc_copy);
2553 kfree(osb->uuid_str);
2554 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2555 memset(osb, 0, sizeof(struct ocfs2_super));
2556}
2557
2558/* Put OCFS2 into a readonly state, or (if the user specifies it),
2559 * panic(). We do not support continue-on-error operation. */
2560static void ocfs2_handle_error(struct super_block *sb)
2561{
2562 struct ocfs2_super *osb = OCFS2_SB(sb);
2563
2564 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2565 panic("OCFS2: (device %s): panic forced after error\n",
2566 sb->s_id);
2567
2568 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2569
2570 if (sb->s_flags & MS_RDONLY &&
2571 (ocfs2_is_soft_readonly(osb) ||
2572 ocfs2_is_hard_readonly(osb)))
2573 return;
2574
2575 printk(KERN_CRIT "File system is now read-only due to the potential "
2576 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2577 "system is unmounted.\n");
2578 sb->s_flags |= MS_RDONLY;
2579 ocfs2_set_ro_flag(osb, 0);
2580}
2581
2582static char error_buf[1024];
2583
2584void __ocfs2_error(struct super_block *sb,
2585 const char *function,
2586 const char *fmt, ...)
2587{
2588 va_list args;
2589
2590 va_start(args, fmt);
2591 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2592 va_end(args);
2593
2594 /* Not using mlog here because we want to show the actual
2595 * function the error came from. */
2596 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2597 sb->s_id, function, error_buf);
2598
2599 ocfs2_handle_error(sb);
2600}
2601
2602/* Handle critical errors. This is intentionally more drastic than
2603 * ocfs2_handle_error, so we only use for things like journal errors,
2604 * etc. */
2605void __ocfs2_abort(struct super_block* sb,
2606 const char *function,
2607 const char *fmt, ...)
2608{
2609 va_list args;
2610
2611 va_start(args, fmt);
2612 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2613 va_end(args);
2614
2615 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2616 sb->s_id, function, error_buf);
2617
2618 /* We don't have the cluster support yet to go straight to
2619 * hard readonly in here. Until then, we want to keep
2620 * ocfs2_abort() so that we can at least mark critical
2621 * errors.
2622 *
2623 * TODO: This should abort the journal and alert other nodes
2624 * that our slot needs recovery. */
2625
2626 /* Force a panic(). This stinks, but it's better than letting
2627 * things continue without having a proper hard readonly
2628 * here. */
2629 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2630 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2631 ocfs2_handle_error(sb);
2632}
2633
2634/*
2635 * Void signal blockers, because in-kernel sigprocmask() only fails
2636 * when SIG_* is wrong.
2637 */
2638void ocfs2_block_signals(sigset_t *oldset)
2639{
2640 int rc;
2641 sigset_t blocked;
2642
2643 sigfillset(&blocked);
2644 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2645 BUG_ON(rc);
2646}
2647
2648void ocfs2_unblock_signals(sigset_t *oldset)
2649{
2650 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2651 BUG_ON(rc);
2652}
2653
2654module_init(ocfs2_init);
2655module_exit(ocfs2_exit);
1// SPDX-License-Identifier: GPL-2.0-or-later
2/* -*- mode: c; c-basic-offset: 8; -*-
3 * vim: noexpandtab sw=8 ts=8 sts=0:
4 *
5 * super.c
6 *
7 * load/unload driver, mount/dismount volumes
8 *
9 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
10 */
11
12#include <linux/module.h>
13#include <linux/fs.h>
14#include <linux/types.h>
15#include <linux/slab.h>
16#include <linux/highmem.h>
17#include <linux/init.h>
18#include <linux/random.h>
19#include <linux/statfs.h>
20#include <linux/moduleparam.h>
21#include <linux/blkdev.h>
22#include <linux/socket.h>
23#include <linux/inet.h>
24#include <linux/parser.h>
25#include <linux/crc32.h>
26#include <linux/debugfs.h>
27#include <linux/mount.h>
28#include <linux/seq_file.h>
29#include <linux/quotaops.h>
30#include <linux/cleancache.h>
31#include <linux/signal.h>
32
33#define CREATE_TRACE_POINTS
34#include "ocfs2_trace.h"
35
36#include <cluster/masklog.h>
37
38#include "ocfs2.h"
39
40/* this should be the only file to include a version 1 header */
41#include "ocfs1_fs_compat.h"
42
43#include "alloc.h"
44#include "aops.h"
45#include "blockcheck.h"
46#include "dlmglue.h"
47#include "export.h"
48#include "extent_map.h"
49#include "heartbeat.h"
50#include "inode.h"
51#include "journal.h"
52#include "localalloc.h"
53#include "namei.h"
54#include "slot_map.h"
55#include "super.h"
56#include "sysfile.h"
57#include "uptodate.h"
58#include "xattr.h"
59#include "quota.h"
60#include "refcounttree.h"
61#include "suballoc.h"
62
63#include "buffer_head_io.h"
64#include "filecheck.h"
65
66static struct kmem_cache *ocfs2_inode_cachep;
67struct kmem_cache *ocfs2_dquot_cachep;
68struct kmem_cache *ocfs2_qf_chunk_cachep;
69
70static struct dentry *ocfs2_debugfs_root;
71
72MODULE_AUTHOR("Oracle");
73MODULE_LICENSE("GPL");
74MODULE_DESCRIPTION("OCFS2 cluster file system");
75
76struct mount_options
77{
78 unsigned long commit_interval;
79 unsigned long mount_opt;
80 unsigned int atime_quantum;
81 signed short slot;
82 int localalloc_opt;
83 unsigned int resv_level;
84 int dir_resv_level;
85 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
86};
87
88static int ocfs2_parse_options(struct super_block *sb, char *options,
89 struct mount_options *mopt,
90 int is_remount);
91static int ocfs2_check_set_options(struct super_block *sb,
92 struct mount_options *options);
93static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
94static void ocfs2_put_super(struct super_block *sb);
95static int ocfs2_mount_volume(struct super_block *sb);
96static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
97static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
98static int ocfs2_initialize_mem_caches(void);
99static void ocfs2_free_mem_caches(void);
100static void ocfs2_delete_osb(struct ocfs2_super *osb);
101
102static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
103
104static int ocfs2_sync_fs(struct super_block *sb, int wait);
105
106static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
107static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
108static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
109static int ocfs2_check_volume(struct ocfs2_super *osb);
110static int ocfs2_verify_volume(struct ocfs2_dinode *di,
111 struct buffer_head *bh,
112 u32 sectsize,
113 struct ocfs2_blockcheck_stats *stats);
114static int ocfs2_initialize_super(struct super_block *sb,
115 struct buffer_head *bh,
116 int sector_size,
117 struct ocfs2_blockcheck_stats *stats);
118static int ocfs2_get_sector(struct super_block *sb,
119 struct buffer_head **bh,
120 int block,
121 int sect_size);
122static struct inode *ocfs2_alloc_inode(struct super_block *sb);
123static void ocfs2_free_inode(struct inode *inode);
124static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
125static int ocfs2_enable_quotas(struct ocfs2_super *osb);
126static void ocfs2_disable_quotas(struct ocfs2_super *osb);
127
128static struct dquot **ocfs2_get_dquots(struct inode *inode)
129{
130 return OCFS2_I(inode)->i_dquot;
131}
132
133static const struct super_operations ocfs2_sops = {
134 .statfs = ocfs2_statfs,
135 .alloc_inode = ocfs2_alloc_inode,
136 .free_inode = ocfs2_free_inode,
137 .drop_inode = ocfs2_drop_inode,
138 .evict_inode = ocfs2_evict_inode,
139 .sync_fs = ocfs2_sync_fs,
140 .put_super = ocfs2_put_super,
141 .remount_fs = ocfs2_remount,
142 .show_options = ocfs2_show_options,
143 .quota_read = ocfs2_quota_read,
144 .quota_write = ocfs2_quota_write,
145 .get_dquots = ocfs2_get_dquots,
146};
147
148enum {
149 Opt_barrier,
150 Opt_err_panic,
151 Opt_err_ro,
152 Opt_intr,
153 Opt_nointr,
154 Opt_hb_none,
155 Opt_hb_local,
156 Opt_hb_global,
157 Opt_data_ordered,
158 Opt_data_writeback,
159 Opt_atime_quantum,
160 Opt_slot,
161 Opt_commit,
162 Opt_localalloc,
163 Opt_localflocks,
164 Opt_stack,
165 Opt_user_xattr,
166 Opt_nouser_xattr,
167 Opt_inode64,
168 Opt_acl,
169 Opt_noacl,
170 Opt_usrquota,
171 Opt_grpquota,
172 Opt_coherency_buffered,
173 Opt_coherency_full,
174 Opt_resv_level,
175 Opt_dir_resv_level,
176 Opt_journal_async_commit,
177 Opt_err_cont,
178 Opt_err,
179};
180
181static const match_table_t tokens = {
182 {Opt_barrier, "barrier=%u"},
183 {Opt_err_panic, "errors=panic"},
184 {Opt_err_ro, "errors=remount-ro"},
185 {Opt_intr, "intr"},
186 {Opt_nointr, "nointr"},
187 {Opt_hb_none, OCFS2_HB_NONE},
188 {Opt_hb_local, OCFS2_HB_LOCAL},
189 {Opt_hb_global, OCFS2_HB_GLOBAL},
190 {Opt_data_ordered, "data=ordered"},
191 {Opt_data_writeback, "data=writeback"},
192 {Opt_atime_quantum, "atime_quantum=%u"},
193 {Opt_slot, "preferred_slot=%u"},
194 {Opt_commit, "commit=%u"},
195 {Opt_localalloc, "localalloc=%d"},
196 {Opt_localflocks, "localflocks"},
197 {Opt_stack, "cluster_stack=%s"},
198 {Opt_user_xattr, "user_xattr"},
199 {Opt_nouser_xattr, "nouser_xattr"},
200 {Opt_inode64, "inode64"},
201 {Opt_acl, "acl"},
202 {Opt_noacl, "noacl"},
203 {Opt_usrquota, "usrquota"},
204 {Opt_grpquota, "grpquota"},
205 {Opt_coherency_buffered, "coherency=buffered"},
206 {Opt_coherency_full, "coherency=full"},
207 {Opt_resv_level, "resv_level=%u"},
208 {Opt_dir_resv_level, "dir_resv_level=%u"},
209 {Opt_journal_async_commit, "journal_async_commit"},
210 {Opt_err_cont, "errors=continue"},
211 {Opt_err, NULL}
212};
213
214#ifdef CONFIG_DEBUG_FS
215static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
216{
217 struct ocfs2_cluster_connection *cconn = osb->cconn;
218 struct ocfs2_recovery_map *rm = osb->recovery_map;
219 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
220 int i, out = 0;
221 unsigned long flags;
222
223 out += snprintf(buf + out, len - out,
224 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
225 "Device", osb->dev_str, osb->uuid_str,
226 osb->fs_generation, osb->vol_label);
227
228 out += snprintf(buf + out, len - out,
229 "%10s => State: %d Flags: 0x%lX\n", "Volume",
230 atomic_read(&osb->vol_state), osb->osb_flags);
231
232 out += snprintf(buf + out, len - out,
233 "%10s => Block: %lu Cluster: %d\n", "Sizes",
234 osb->sb->s_blocksize, osb->s_clustersize);
235
236 out += snprintf(buf + out, len - out,
237 "%10s => Compat: 0x%X Incompat: 0x%X "
238 "ROcompat: 0x%X\n",
239 "Features", osb->s_feature_compat,
240 osb->s_feature_incompat, osb->s_feature_ro_compat);
241
242 out += snprintf(buf + out, len - out,
243 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
244 osb->s_mount_opt, osb->s_atime_quantum);
245
246 if (cconn) {
247 out += snprintf(buf + out, len - out,
248 "%10s => Stack: %s Name: %*s "
249 "Version: %d.%d\n", "Cluster",
250 (*osb->osb_cluster_stack == '\0' ?
251 "o2cb" : osb->osb_cluster_stack),
252 cconn->cc_namelen, cconn->cc_name,
253 cconn->cc_version.pv_major,
254 cconn->cc_version.pv_minor);
255 }
256
257 spin_lock_irqsave(&osb->dc_task_lock, flags);
258 out += snprintf(buf + out, len - out,
259 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
260 "WorkSeq: %lu\n", "DownCnvt",
261 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
262 osb->blocked_lock_count, osb->dc_wake_sequence,
263 osb->dc_work_sequence);
264 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
265
266 spin_lock(&osb->osb_lock);
267 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
268 "Recovery",
269 (osb->recovery_thread_task ?
270 task_pid_nr(osb->recovery_thread_task) : -1));
271 if (rm->rm_used == 0)
272 out += snprintf(buf + out, len - out, " None\n");
273 else {
274 for (i = 0; i < rm->rm_used; i++)
275 out += snprintf(buf + out, len - out, " %d",
276 rm->rm_entries[i]);
277 out += snprintf(buf + out, len - out, "\n");
278 }
279 spin_unlock(&osb->osb_lock);
280
281 out += snprintf(buf + out, len - out,
282 "%10s => Pid: %d Interval: %lu\n", "Commit",
283 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
284 osb->osb_commit_interval);
285
286 out += snprintf(buf + out, len - out,
287 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
288 "Journal", osb->journal->j_state,
289 osb->journal->j_trans_id,
290 atomic_read(&osb->journal->j_num_trans));
291
292 out += snprintf(buf + out, len - out,
293 "%10s => GlobalAllocs: %d LocalAllocs: %d "
294 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
295 "Stats",
296 atomic_read(&osb->alloc_stats.bitmap_data),
297 atomic_read(&osb->alloc_stats.local_data),
298 atomic_read(&osb->alloc_stats.bg_allocs),
299 atomic_read(&osb->alloc_stats.moves),
300 atomic_read(&osb->alloc_stats.bg_extends));
301
302 out += snprintf(buf + out, len - out,
303 "%10s => State: %u Descriptor: %llu Size: %u bits "
304 "Default: %u bits\n",
305 "LocalAlloc", osb->local_alloc_state,
306 (unsigned long long)osb->la_last_gd,
307 osb->local_alloc_bits, osb->local_alloc_default_bits);
308
309 spin_lock(&osb->osb_lock);
310 out += snprintf(buf + out, len - out,
311 "%10s => InodeSlot: %d StolenInodes: %d, "
312 "MetaSlot: %d StolenMeta: %d\n", "Steal",
313 osb->s_inode_steal_slot,
314 atomic_read(&osb->s_num_inodes_stolen),
315 osb->s_meta_steal_slot,
316 atomic_read(&osb->s_num_meta_stolen));
317 spin_unlock(&osb->osb_lock);
318
319 out += snprintf(buf + out, len - out, "OrphanScan => ");
320 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
321 os->os_count, os->os_seqno);
322 out += snprintf(buf + out, len - out, " Last Scan: ");
323 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
324 out += snprintf(buf + out, len - out, "Disabled\n");
325 else
326 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
327 (unsigned long)(ktime_get_seconds() - os->os_scantime));
328
329 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
330 "Slots", "Num", "RecoGen");
331 for (i = 0; i < osb->max_slots; ++i) {
332 out += snprintf(buf + out, len - out,
333 "%10s %c %3d %10d\n",
334 " ",
335 (i == osb->slot_num ? '*' : ' '),
336 i, osb->slot_recovery_generations[i]);
337 }
338
339 return out;
340}
341
342static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
343{
344 struct ocfs2_super *osb = inode->i_private;
345 char *buf = NULL;
346
347 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
348 if (!buf)
349 goto bail;
350
351 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
352
353 file->private_data = buf;
354
355 return 0;
356bail:
357 return -ENOMEM;
358}
359
360static int ocfs2_debug_release(struct inode *inode, struct file *file)
361{
362 kfree(file->private_data);
363 return 0;
364}
365
366static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
367 size_t nbytes, loff_t *ppos)
368{
369 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
370 i_size_read(file->f_mapping->host));
371}
372#else
373static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
374{
375 return 0;
376}
377static int ocfs2_debug_release(struct inode *inode, struct file *file)
378{
379 return 0;
380}
381static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
382 size_t nbytes, loff_t *ppos)
383{
384 return 0;
385}
386#endif /* CONFIG_DEBUG_FS */
387
388static const struct file_operations ocfs2_osb_debug_fops = {
389 .open = ocfs2_osb_debug_open,
390 .release = ocfs2_debug_release,
391 .read = ocfs2_debug_read,
392 .llseek = generic_file_llseek,
393};
394
395static int ocfs2_sync_fs(struct super_block *sb, int wait)
396{
397 int status;
398 tid_t target;
399 struct ocfs2_super *osb = OCFS2_SB(sb);
400
401 if (ocfs2_is_hard_readonly(osb))
402 return -EROFS;
403
404 if (wait) {
405 status = ocfs2_flush_truncate_log(osb);
406 if (status < 0)
407 mlog_errno(status);
408 } else {
409 ocfs2_schedule_truncate_log_flush(osb, 0);
410 }
411
412 if (jbd2_journal_start_commit(osb->journal->j_journal,
413 &target)) {
414 if (wait)
415 jbd2_log_wait_commit(osb->journal->j_journal,
416 target);
417 }
418 return 0;
419}
420
421static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
422{
423 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
424 && (ino == USER_QUOTA_SYSTEM_INODE
425 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
426 return 0;
427 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
428 && (ino == GROUP_QUOTA_SYSTEM_INODE
429 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
430 return 0;
431 return 1;
432}
433
434static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
435{
436 struct inode *new = NULL;
437 int status = 0;
438 int i;
439
440 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
441 if (IS_ERR(new)) {
442 status = PTR_ERR(new);
443 mlog_errno(status);
444 goto bail;
445 }
446 osb->root_inode = new;
447
448 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
449 if (IS_ERR(new)) {
450 status = PTR_ERR(new);
451 mlog_errno(status);
452 goto bail;
453 }
454 osb->sys_root_inode = new;
455
456 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
457 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
458 if (!ocfs2_need_system_inode(osb, i))
459 continue;
460 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
461 if (!new) {
462 ocfs2_release_system_inodes(osb);
463 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
464 mlog_errno(status);
465 mlog(ML_ERROR, "Unable to load system inode %d, "
466 "possibly corrupt fs?", i);
467 goto bail;
468 }
469 // the array now has one ref, so drop this one
470 iput(new);
471 }
472
473bail:
474 if (status)
475 mlog_errno(status);
476 return status;
477}
478
479static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
480{
481 struct inode *new = NULL;
482 int status = 0;
483 int i;
484
485 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
486 i < NUM_SYSTEM_INODES;
487 i++) {
488 if (!ocfs2_need_system_inode(osb, i))
489 continue;
490 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
491 if (!new) {
492 ocfs2_release_system_inodes(osb);
493 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
494 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
495 status, i, osb->slot_num);
496 goto bail;
497 }
498 /* the array now has one ref, so drop this one */
499 iput(new);
500 }
501
502bail:
503 if (status)
504 mlog_errno(status);
505 return status;
506}
507
508static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
509{
510 int i;
511 struct inode *inode;
512
513 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
514 inode = osb->global_system_inodes[i];
515 if (inode) {
516 iput(inode);
517 osb->global_system_inodes[i] = NULL;
518 }
519 }
520
521 inode = osb->sys_root_inode;
522 if (inode) {
523 iput(inode);
524 osb->sys_root_inode = NULL;
525 }
526
527 inode = osb->root_inode;
528 if (inode) {
529 iput(inode);
530 osb->root_inode = NULL;
531 }
532
533 if (!osb->local_system_inodes)
534 return;
535
536 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
537 if (osb->local_system_inodes[i]) {
538 iput(osb->local_system_inodes[i]);
539 osb->local_system_inodes[i] = NULL;
540 }
541 }
542
543 kfree(osb->local_system_inodes);
544 osb->local_system_inodes = NULL;
545}
546
547/* We're allocating fs objects, use GFP_NOFS */
548static struct inode *ocfs2_alloc_inode(struct super_block *sb)
549{
550 struct ocfs2_inode_info *oi;
551
552 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
553 if (!oi)
554 return NULL;
555
556 oi->i_sync_tid = 0;
557 oi->i_datasync_tid = 0;
558 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
559
560 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
561 return &oi->vfs_inode;
562}
563
564static void ocfs2_free_inode(struct inode *inode)
565{
566 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
567}
568
569static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
570 unsigned int cbits)
571{
572 unsigned int bytes = 1 << cbits;
573 unsigned int trim = bytes;
574 unsigned int bitshift = 32;
575
576 /*
577 * i_size and all block offsets in ocfs2 are always 64 bits
578 * wide. i_clusters is 32 bits, in cluster-sized units. So on
579 * 64 bit platforms, cluster size will be the limiting factor.
580 */
581
582#if BITS_PER_LONG == 32
583 BUILD_BUG_ON(sizeof(sector_t) != 8);
584 /*
585 * We might be limited by page cache size.
586 */
587 if (bytes > PAGE_SIZE) {
588 bytes = PAGE_SIZE;
589 trim = 1;
590 /*
591 * Shift by 31 here so that we don't get larger than
592 * MAX_LFS_FILESIZE
593 */
594 bitshift = 31;
595 }
596#endif
597
598 /*
599 * Trim by a whole cluster when we can actually approach the
600 * on-disk limits. Otherwise we can overflow i_clusters when
601 * an extent start is at the max offset.
602 */
603 return (((unsigned long long)bytes) << bitshift) - trim;
604}
605
606static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
607{
608 int incompat_features;
609 int ret = 0;
610 struct mount_options parsed_options;
611 struct ocfs2_super *osb = OCFS2_SB(sb);
612 u32 tmp;
613
614 sync_filesystem(sb);
615
616 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
617 !ocfs2_check_set_options(sb, &parsed_options)) {
618 ret = -EINVAL;
619 goto out;
620 }
621
622 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
623 OCFS2_MOUNT_HB_NONE;
624 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
625 ret = -EINVAL;
626 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
627 goto out;
628 }
629
630 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
631 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
632 ret = -EINVAL;
633 mlog(ML_ERROR, "Cannot change data mode on remount\n");
634 goto out;
635 }
636
637 /* Probably don't want this on remount; it might
638 * mess with other nodes */
639 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
640 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
641 ret = -EINVAL;
642 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
643 goto out;
644 }
645
646 /* We're going to/from readonly mode. */
647 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
648 /* Disable quota accounting before remounting RO */
649 if (*flags & SB_RDONLY) {
650 ret = ocfs2_susp_quotas(osb, 0);
651 if (ret < 0)
652 goto out;
653 }
654 /* Lock here so the check of HARD_RO and the potential
655 * setting of SOFT_RO is atomic. */
656 spin_lock(&osb->osb_lock);
657 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
658 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
659 ret = -EROFS;
660 goto unlock_osb;
661 }
662
663 if (*flags & SB_RDONLY) {
664 sb->s_flags |= SB_RDONLY;
665 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
666 } else {
667 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
668 mlog(ML_ERROR, "Cannot remount RDWR "
669 "filesystem due to previous errors.\n");
670 ret = -EROFS;
671 goto unlock_osb;
672 }
673 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
674 if (incompat_features) {
675 mlog(ML_ERROR, "Cannot remount RDWR because "
676 "of unsupported optional features "
677 "(%x).\n", incompat_features);
678 ret = -EINVAL;
679 goto unlock_osb;
680 }
681 sb->s_flags &= ~SB_RDONLY;
682 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
683 }
684 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
685unlock_osb:
686 spin_unlock(&osb->osb_lock);
687 /* Enable quota accounting after remounting RW */
688 if (!ret && !(*flags & SB_RDONLY)) {
689 if (sb_any_quota_suspended(sb))
690 ret = ocfs2_susp_quotas(osb, 1);
691 else
692 ret = ocfs2_enable_quotas(osb);
693 if (ret < 0) {
694 /* Return back changes... */
695 spin_lock(&osb->osb_lock);
696 sb->s_flags |= SB_RDONLY;
697 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
698 spin_unlock(&osb->osb_lock);
699 goto out;
700 }
701 }
702 }
703
704 if (!ret) {
705 /* Only save off the new mount options in case of a successful
706 * remount. */
707 osb->s_mount_opt = parsed_options.mount_opt;
708 osb->s_atime_quantum = parsed_options.atime_quantum;
709 osb->preferred_slot = parsed_options.slot;
710 if (parsed_options.commit_interval)
711 osb->osb_commit_interval = parsed_options.commit_interval;
712
713 if (!ocfs2_is_hard_readonly(osb))
714 ocfs2_set_journal_params(osb);
715
716 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
717 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
718 SB_POSIXACL : 0);
719 }
720out:
721 return ret;
722}
723
724static int ocfs2_sb_probe(struct super_block *sb,
725 struct buffer_head **bh,
726 int *sector_size,
727 struct ocfs2_blockcheck_stats *stats)
728{
729 int status, tmpstat;
730 struct ocfs1_vol_disk_hdr *hdr;
731 struct ocfs2_dinode *di;
732 int blksize;
733
734 *bh = NULL;
735
736 /* may be > 512 */
737 *sector_size = bdev_logical_block_size(sb->s_bdev);
738 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
739 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
740 *sector_size, OCFS2_MAX_BLOCKSIZE);
741 status = -EINVAL;
742 goto bail;
743 }
744
745 /* Can this really happen? */
746 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
747 *sector_size = OCFS2_MIN_BLOCKSIZE;
748
749 /* check block zero for old format */
750 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
751 if (status < 0) {
752 mlog_errno(status);
753 goto bail;
754 }
755 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
756 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
757 mlog(ML_ERROR, "incompatible version: %u.%u\n",
758 hdr->major_version, hdr->minor_version);
759 status = -EINVAL;
760 }
761 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
762 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
763 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
764 hdr->signature);
765 status = -EINVAL;
766 }
767 brelse(*bh);
768 *bh = NULL;
769 if (status < 0) {
770 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
771 "upgraded before mounting with ocfs v2\n");
772 goto bail;
773 }
774
775 /*
776 * Now check at magic offset for 512, 1024, 2048, 4096
777 * blocksizes. 4096 is the maximum blocksize because it is
778 * the minimum clustersize.
779 */
780 status = -EINVAL;
781 for (blksize = *sector_size;
782 blksize <= OCFS2_MAX_BLOCKSIZE;
783 blksize <<= 1) {
784 tmpstat = ocfs2_get_sector(sb, bh,
785 OCFS2_SUPER_BLOCK_BLKNO,
786 blksize);
787 if (tmpstat < 0) {
788 status = tmpstat;
789 mlog_errno(status);
790 break;
791 }
792 di = (struct ocfs2_dinode *) (*bh)->b_data;
793 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
794 spin_lock_init(&stats->b_lock);
795 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
796 if (tmpstat < 0) {
797 brelse(*bh);
798 *bh = NULL;
799 }
800 if (tmpstat != -EAGAIN) {
801 status = tmpstat;
802 break;
803 }
804 }
805
806bail:
807 return status;
808}
809
810static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
811{
812 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
813
814 if (osb->s_mount_opt & hb_enabled) {
815 if (ocfs2_mount_local(osb)) {
816 mlog(ML_ERROR, "Cannot heartbeat on a locally "
817 "mounted device.\n");
818 return -EINVAL;
819 }
820 if (ocfs2_userspace_stack(osb)) {
821 mlog(ML_ERROR, "Userspace stack expected, but "
822 "o2cb heartbeat arguments passed to mount\n");
823 return -EINVAL;
824 }
825 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
826 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
827 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
828 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
829 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
830 return -EINVAL;
831 }
832 }
833
834 if (!(osb->s_mount_opt & hb_enabled)) {
835 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
836 !ocfs2_userspace_stack(osb)) {
837 mlog(ML_ERROR, "Heartbeat has to be started to mount "
838 "a read-write clustered device.\n");
839 return -EINVAL;
840 }
841 }
842
843 return 0;
844}
845
846/*
847 * If we're using a userspace stack, mount should have passed
848 * a name that matches the disk. If not, mount should not
849 * have passed a stack.
850 */
851static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
852 struct mount_options *mopt)
853{
854 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
855 mlog(ML_ERROR,
856 "cluster stack passed to mount, but this filesystem "
857 "does not support it\n");
858 return -EINVAL;
859 }
860
861 if (ocfs2_userspace_stack(osb) &&
862 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
863 OCFS2_STACK_LABEL_LEN)) {
864 mlog(ML_ERROR,
865 "cluster stack passed to mount (\"%s\") does not "
866 "match the filesystem (\"%s\")\n",
867 mopt->cluster_stack,
868 osb->osb_cluster_stack);
869 return -EINVAL;
870 }
871
872 return 0;
873}
874
875static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
876{
877 int type;
878 struct super_block *sb = osb->sb;
879 unsigned int feature[OCFS2_MAXQUOTAS] = {
880 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
881 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
882 int status = 0;
883
884 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
885 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
886 continue;
887 if (unsuspend)
888 status = dquot_resume(sb, type);
889 else {
890 struct ocfs2_mem_dqinfo *oinfo;
891
892 /* Cancel periodic syncing before suspending */
893 oinfo = sb_dqinfo(sb, type)->dqi_priv;
894 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
895 status = dquot_suspend(sb, type);
896 }
897 if (status < 0)
898 break;
899 }
900 if (status < 0)
901 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
902 "remount (error = %d).\n", status);
903 return status;
904}
905
906static int ocfs2_enable_quotas(struct ocfs2_super *osb)
907{
908 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
909 struct super_block *sb = osb->sb;
910 unsigned int feature[OCFS2_MAXQUOTAS] = {
911 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
912 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
913 unsigned int ino[OCFS2_MAXQUOTAS] = {
914 LOCAL_USER_QUOTA_SYSTEM_INODE,
915 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
916 int status;
917 int type;
918
919 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
920 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
921 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
922 continue;
923 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
924 osb->slot_num);
925 if (!inode[type]) {
926 status = -ENOENT;
927 goto out_quota_off;
928 }
929 status = dquot_enable(inode[type], type, QFMT_OCFS2,
930 DQUOT_USAGE_ENABLED);
931 if (status < 0)
932 goto out_quota_off;
933 }
934
935 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
936 iput(inode[type]);
937 return 0;
938out_quota_off:
939 ocfs2_disable_quotas(osb);
940 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
941 iput(inode[type]);
942 mlog_errno(status);
943 return status;
944}
945
946static void ocfs2_disable_quotas(struct ocfs2_super *osb)
947{
948 int type;
949 struct inode *inode;
950 struct super_block *sb = osb->sb;
951 struct ocfs2_mem_dqinfo *oinfo;
952
953 /* We mostly ignore errors in this function because there's not much
954 * we can do when we see them */
955 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
956 if (!sb_has_quota_loaded(sb, type))
957 continue;
958 oinfo = sb_dqinfo(sb, type)->dqi_priv;
959 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
960 inode = igrab(sb->s_dquot.files[type]);
961 /* Turn off quotas. This will remove all dquot structures from
962 * memory and so they will be automatically synced to global
963 * quota files */
964 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
965 DQUOT_LIMITS_ENABLED);
966 if (!inode)
967 continue;
968 iput(inode);
969 }
970}
971
972static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
973{
974 struct dentry *root;
975 int status, sector_size;
976 struct mount_options parsed_options;
977 struct inode *inode = NULL;
978 struct ocfs2_super *osb = NULL;
979 struct buffer_head *bh = NULL;
980 char nodestr[12];
981 struct ocfs2_blockcheck_stats stats;
982
983 trace_ocfs2_fill_super(sb, data, silent);
984
985 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
986 status = -EINVAL;
987 goto read_super_error;
988 }
989
990 /* probe for superblock */
991 status = ocfs2_sb_probe(sb, &bh, §or_size, &stats);
992 if (status < 0) {
993 mlog(ML_ERROR, "superblock probe failed!\n");
994 goto read_super_error;
995 }
996
997 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
998 osb = OCFS2_SB(sb);
999 if (status < 0) {
1000 mlog_errno(status);
1001 goto read_super_error;
1002 }
1003 brelse(bh);
1004 bh = NULL;
1005
1006 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1007 status = -EINVAL;
1008 goto read_super_error;
1009 }
1010 osb->s_mount_opt = parsed_options.mount_opt;
1011 osb->s_atime_quantum = parsed_options.atime_quantum;
1012 osb->preferred_slot = parsed_options.slot;
1013 osb->osb_commit_interval = parsed_options.commit_interval;
1014
1015 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
1016 osb->osb_resv_level = parsed_options.resv_level;
1017 osb->osb_dir_resv_level = parsed_options.resv_level;
1018 if (parsed_options.dir_resv_level == -1)
1019 osb->osb_dir_resv_level = parsed_options.resv_level;
1020 else
1021 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
1022
1023 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1024 if (status)
1025 goto read_super_error;
1026
1027 sb->s_magic = OCFS2_SUPER_MAGIC;
1028
1029 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1030 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
1031
1032 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
1033 * heartbeat=none */
1034 if (bdev_read_only(sb->s_bdev)) {
1035 if (!sb_rdonly(sb)) {
1036 status = -EACCES;
1037 mlog(ML_ERROR, "Readonly device detected but readonly "
1038 "mount was not specified.\n");
1039 goto read_super_error;
1040 }
1041
1042 /* You should not be able to start a local heartbeat
1043 * on a readonly device. */
1044 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1045 status = -EROFS;
1046 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1047 "device.\n");
1048 goto read_super_error;
1049 }
1050
1051 status = ocfs2_check_journals_nolocks(osb);
1052 if (status < 0) {
1053 if (status == -EROFS)
1054 mlog(ML_ERROR, "Recovery required on readonly "
1055 "file system, but write access is "
1056 "unavailable.\n");
1057 else
1058 mlog_errno(status);
1059 goto read_super_error;
1060 }
1061
1062 ocfs2_set_ro_flag(osb, 1);
1063
1064 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1065 "Cluster services will not be used for this mount. "
1066 "Recovery will be skipped.\n", osb->dev_str);
1067 }
1068
1069 if (!ocfs2_is_hard_readonly(osb)) {
1070 if (sb_rdonly(sb))
1071 ocfs2_set_ro_flag(osb, 0);
1072 }
1073
1074 status = ocfs2_verify_heartbeat(osb);
1075 if (status < 0) {
1076 mlog_errno(status);
1077 goto read_super_error;
1078 }
1079
1080 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1081 ocfs2_debugfs_root);
1082
1083 debugfs_create_file("fs_state", S_IFREG|S_IRUSR, osb->osb_debug_root,
1084 osb, &ocfs2_osb_debug_fops);
1085
1086 if (ocfs2_meta_ecc(osb))
1087 ocfs2_blockcheck_stats_debugfs_install( &osb->osb_ecc_stats,
1088 osb->osb_debug_root);
1089
1090 status = ocfs2_mount_volume(sb);
1091 if (status < 0)
1092 goto read_super_error;
1093
1094 if (osb->root_inode)
1095 inode = igrab(osb->root_inode);
1096
1097 if (!inode) {
1098 status = -EIO;
1099 mlog_errno(status);
1100 goto read_super_error;
1101 }
1102
1103 root = d_make_root(inode);
1104 if (!root) {
1105 status = -ENOMEM;
1106 mlog_errno(status);
1107 goto read_super_error;
1108 }
1109
1110 sb->s_root = root;
1111
1112 ocfs2_complete_mount_recovery(osb);
1113
1114 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1115 &ocfs2_kset->kobj);
1116 if (!osb->osb_dev_kset) {
1117 status = -ENOMEM;
1118 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1119 goto read_super_error;
1120 }
1121
1122 /* Create filecheck sysfs related directories/files at
1123 * /sys/fs/ocfs2/<devname>/filecheck */
1124 if (ocfs2_filecheck_create_sysfs(osb)) {
1125 status = -ENOMEM;
1126 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1127 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1128 goto read_super_error;
1129 }
1130
1131 if (ocfs2_mount_local(osb))
1132 snprintf(nodestr, sizeof(nodestr), "local");
1133 else
1134 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1135
1136 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1137 "with %s data mode.\n",
1138 osb->dev_str, nodestr, osb->slot_num,
1139 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1140 "ordered");
1141
1142 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1143 wake_up(&osb->osb_mount_event);
1144
1145 /* Now we can initialize quotas because we can afford to wait
1146 * for cluster locks recovery now. That also means that truncation
1147 * log recovery can happen but that waits for proper quota setup */
1148 if (!sb_rdonly(sb)) {
1149 status = ocfs2_enable_quotas(osb);
1150 if (status < 0) {
1151 /* We have to err-out specially here because
1152 * s_root is already set */
1153 mlog_errno(status);
1154 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1155 wake_up(&osb->osb_mount_event);
1156 return status;
1157 }
1158 }
1159
1160 ocfs2_complete_quota_recovery(osb);
1161
1162 /* Now we wake up again for processes waiting for quotas */
1163 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1164 wake_up(&osb->osb_mount_event);
1165
1166 /* Start this when the mount is almost sure of being successful */
1167 ocfs2_orphan_scan_start(osb);
1168
1169 return status;
1170
1171read_super_error:
1172 brelse(bh);
1173
1174 if (status)
1175 mlog_errno(status);
1176
1177 if (osb) {
1178 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1179 wake_up(&osb->osb_mount_event);
1180 ocfs2_dismount_volume(sb, 1);
1181 }
1182
1183 return status;
1184}
1185
1186static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
1187 int flags,
1188 const char *dev_name,
1189 void *data)
1190{
1191 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
1192}
1193
1194static struct file_system_type ocfs2_fs_type = {
1195 .owner = THIS_MODULE,
1196 .name = "ocfs2",
1197 .mount = ocfs2_mount,
1198 .kill_sb = kill_block_super,
1199 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1200 .next = NULL
1201};
1202MODULE_ALIAS_FS("ocfs2");
1203
1204static int ocfs2_check_set_options(struct super_block *sb,
1205 struct mount_options *options)
1206{
1207 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1208 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1209 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1210 mlog(ML_ERROR, "User quotas were requested, but this "
1211 "filesystem does not have the feature enabled.\n");
1212 return 0;
1213 }
1214 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1215 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1216 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1217 mlog(ML_ERROR, "Group quotas were requested, but this "
1218 "filesystem does not have the feature enabled.\n");
1219 return 0;
1220 }
1221 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1222 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1223 mlog(ML_ERROR, "ACL support requested but extended attributes "
1224 "feature is not enabled\n");
1225 return 0;
1226 }
1227 /* No ACL setting specified? Use XATTR feature... */
1228 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1229 OCFS2_MOUNT_NO_POSIX_ACL))) {
1230 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1231 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1232 else
1233 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1234 }
1235 return 1;
1236}
1237
1238static int ocfs2_parse_options(struct super_block *sb,
1239 char *options,
1240 struct mount_options *mopt,
1241 int is_remount)
1242{
1243 int status, user_stack = 0;
1244 char *p;
1245 u32 tmp;
1246 int token, option;
1247 substring_t args[MAX_OPT_ARGS];
1248
1249 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
1250
1251 mopt->commit_interval = 0;
1252 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
1253 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1254 mopt->slot = OCFS2_INVALID_SLOT;
1255 mopt->localalloc_opt = -1;
1256 mopt->cluster_stack[0] = '\0';
1257 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
1258 mopt->dir_resv_level = -1;
1259
1260 if (!options) {
1261 status = 1;
1262 goto bail;
1263 }
1264
1265 while ((p = strsep(&options, ",")) != NULL) {
1266 if (!*p)
1267 continue;
1268
1269 token = match_token(p, tokens, args);
1270 switch (token) {
1271 case Opt_hb_local:
1272 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1273 break;
1274 case Opt_hb_none:
1275 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1276 break;
1277 case Opt_hb_global:
1278 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
1279 break;
1280 case Opt_barrier:
1281 if (match_int(&args[0], &option)) {
1282 status = 0;
1283 goto bail;
1284 }
1285 if (option)
1286 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1287 else
1288 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1289 break;
1290 case Opt_intr:
1291 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1292 break;
1293 case Opt_nointr:
1294 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1295 break;
1296 case Opt_err_panic:
1297 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1298 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1299 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1300 break;
1301 case Opt_err_ro:
1302 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1303 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1304 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1305 break;
1306 case Opt_err_cont:
1307 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1308 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1309 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
1310 break;
1311 case Opt_data_ordered:
1312 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1313 break;
1314 case Opt_data_writeback:
1315 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1316 break;
1317 case Opt_user_xattr:
1318 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1319 break;
1320 case Opt_nouser_xattr:
1321 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1322 break;
1323 case Opt_atime_quantum:
1324 if (match_int(&args[0], &option)) {
1325 status = 0;
1326 goto bail;
1327 }
1328 if (option >= 0)
1329 mopt->atime_quantum = option;
1330 break;
1331 case Opt_slot:
1332 if (match_int(&args[0], &option)) {
1333 status = 0;
1334 goto bail;
1335 }
1336 if (option)
1337 mopt->slot = (s16)option;
1338 break;
1339 case Opt_commit:
1340 if (match_int(&args[0], &option)) {
1341 status = 0;
1342 goto bail;
1343 }
1344 if (option < 0)
1345 return 0;
1346 if (option == 0)
1347 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1348 mopt->commit_interval = HZ * option;
1349 break;
1350 case Opt_localalloc:
1351 if (match_int(&args[0], &option)) {
1352 status = 0;
1353 goto bail;
1354 }
1355 if (option >= 0)
1356 mopt->localalloc_opt = option;
1357 break;
1358 case Opt_localflocks:
1359 /*
1360 * Changing this during remount could race
1361 * flock() requests, or "unbalance" existing
1362 * ones (e.g., a lock is taken in one mode but
1363 * dropped in the other). If users care enough
1364 * to flip locking modes during remount, we
1365 * could add a "local" flag to individual
1366 * flock structures for proper tracking of
1367 * state.
1368 */
1369 if (!is_remount)
1370 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1371 break;
1372 case Opt_stack:
1373 /* Check both that the option we were passed
1374 * is of the right length and that it is a proper
1375 * string of the right length.
1376 */
1377 if (((args[0].to - args[0].from) !=
1378 OCFS2_STACK_LABEL_LEN) ||
1379 (strnlen(args[0].from,
1380 OCFS2_STACK_LABEL_LEN) !=
1381 OCFS2_STACK_LABEL_LEN)) {
1382 mlog(ML_ERROR,
1383 "Invalid cluster_stack option\n");
1384 status = 0;
1385 goto bail;
1386 }
1387 memcpy(mopt->cluster_stack, args[0].from,
1388 OCFS2_STACK_LABEL_LEN);
1389 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1390 /*
1391 * Open code the memcmp here as we don't have
1392 * an osb to pass to
1393 * ocfs2_userspace_stack().
1394 */
1395 if (memcmp(mopt->cluster_stack,
1396 OCFS2_CLASSIC_CLUSTER_STACK,
1397 OCFS2_STACK_LABEL_LEN))
1398 user_stack = 1;
1399 break;
1400 case Opt_inode64:
1401 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1402 break;
1403 case Opt_usrquota:
1404 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1405 break;
1406 case Opt_grpquota:
1407 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1408 break;
1409 case Opt_coherency_buffered:
1410 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1411 break;
1412 case Opt_coherency_full:
1413 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1414 break;
1415 case Opt_acl:
1416 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1417 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
1418 break;
1419 case Opt_noacl:
1420 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1421 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1422 break;
1423 case Opt_resv_level:
1424 if (is_remount)
1425 break;
1426 if (match_int(&args[0], &option)) {
1427 status = 0;
1428 goto bail;
1429 }
1430 if (option >= OCFS2_MIN_RESV_LEVEL &&
1431 option < OCFS2_MAX_RESV_LEVEL)
1432 mopt->resv_level = option;
1433 break;
1434 case Opt_dir_resv_level:
1435 if (is_remount)
1436 break;
1437 if (match_int(&args[0], &option)) {
1438 status = 0;
1439 goto bail;
1440 }
1441 if (option >= OCFS2_MIN_RESV_LEVEL &&
1442 option < OCFS2_MAX_RESV_LEVEL)
1443 mopt->dir_resv_level = option;
1444 break;
1445 case Opt_journal_async_commit:
1446 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1447 break;
1448 default:
1449 mlog(ML_ERROR,
1450 "Unrecognized mount option \"%s\" "
1451 "or missing value\n", p);
1452 status = 0;
1453 goto bail;
1454 }
1455 }
1456
1457 if (user_stack == 0) {
1458 /* Ensure only one heartbeat mode */
1459 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1460 OCFS2_MOUNT_HB_GLOBAL |
1461 OCFS2_MOUNT_HB_NONE);
1462 if (hweight32(tmp) != 1) {
1463 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1464 status = 0;
1465 goto bail;
1466 }
1467 }
1468
1469 status = 1;
1470
1471bail:
1472 return status;
1473}
1474
1475static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
1476{
1477 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
1478 unsigned long opts = osb->s_mount_opt;
1479 unsigned int local_alloc_megs;
1480
1481 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1482 seq_printf(s, ",_netdev");
1483 if (opts & OCFS2_MOUNT_HB_LOCAL)
1484 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1485 else
1486 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1487 } else
1488 seq_printf(s, ",%s", OCFS2_HB_NONE);
1489
1490 if (opts & OCFS2_MOUNT_NOINTR)
1491 seq_printf(s, ",nointr");
1492
1493 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1494 seq_printf(s, ",data=writeback");
1495 else
1496 seq_printf(s, ",data=ordered");
1497
1498 if (opts & OCFS2_MOUNT_BARRIER)
1499 seq_printf(s, ",barrier=1");
1500
1501 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1502 seq_printf(s, ",errors=panic");
1503 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1504 seq_printf(s, ",errors=continue");
1505 else
1506 seq_printf(s, ",errors=remount-ro");
1507
1508 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1509 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1510
1511 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1512
1513 if (osb->osb_commit_interval)
1514 seq_printf(s, ",commit=%u",
1515 (unsigned) (osb->osb_commit_interval / HZ));
1516
1517 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1518 if (local_alloc_megs != ocfs2_la_default_mb(osb))
1519 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1520
1521 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1522 seq_printf(s, ",localflocks,");
1523
1524 if (osb->osb_cluster_stack[0])
1525 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1526 OCFS2_STACK_LABEL_LEN);
1527 if (opts & OCFS2_MOUNT_USRQUOTA)
1528 seq_printf(s, ",usrquota");
1529 if (opts & OCFS2_MOUNT_GRPQUOTA)
1530 seq_printf(s, ",grpquota");
1531
1532 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1533 seq_printf(s, ",coherency=buffered");
1534 else
1535 seq_printf(s, ",coherency=full");
1536
1537 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1538 seq_printf(s, ",nouser_xattr");
1539 else
1540 seq_printf(s, ",user_xattr");
1541
1542 if (opts & OCFS2_MOUNT_INODE64)
1543 seq_printf(s, ",inode64");
1544
1545 if (opts & OCFS2_MOUNT_POSIX_ACL)
1546 seq_printf(s, ",acl");
1547 else
1548 seq_printf(s, ",noacl");
1549
1550 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1551 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1552
1553 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1554 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1555
1556 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1557 seq_printf(s, ",journal_async_commit");
1558
1559 return 0;
1560}
1561
1562static int __init ocfs2_init(void)
1563{
1564 int status;
1565
1566 status = init_ocfs2_uptodate_cache();
1567 if (status < 0)
1568 goto out1;
1569
1570 status = ocfs2_initialize_mem_caches();
1571 if (status < 0)
1572 goto out2;
1573
1574 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1575
1576 ocfs2_set_locking_protocol();
1577
1578 status = register_quota_format(&ocfs2_quota_format);
1579 if (status < 0)
1580 goto out3;
1581 status = register_filesystem(&ocfs2_fs_type);
1582 if (!status)
1583 return 0;
1584
1585 unregister_quota_format(&ocfs2_quota_format);
1586out3:
1587 debugfs_remove(ocfs2_debugfs_root);
1588 ocfs2_free_mem_caches();
1589out2:
1590 exit_ocfs2_uptodate_cache();
1591out1:
1592 mlog_errno(status);
1593 return status;
1594}
1595
1596static void __exit ocfs2_exit(void)
1597{
1598 unregister_quota_format(&ocfs2_quota_format);
1599
1600 debugfs_remove(ocfs2_debugfs_root);
1601
1602 ocfs2_free_mem_caches();
1603
1604 unregister_filesystem(&ocfs2_fs_type);
1605
1606 exit_ocfs2_uptodate_cache();
1607}
1608
1609static void ocfs2_put_super(struct super_block *sb)
1610{
1611 trace_ocfs2_put_super(sb);
1612
1613 ocfs2_sync_blockdev(sb);
1614 ocfs2_dismount_volume(sb, 0);
1615}
1616
1617static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1618{
1619 struct ocfs2_super *osb;
1620 u32 numbits, freebits;
1621 int status;
1622 struct ocfs2_dinode *bm_lock;
1623 struct buffer_head *bh = NULL;
1624 struct inode *inode = NULL;
1625
1626 trace_ocfs2_statfs(dentry->d_sb, buf);
1627
1628 osb = OCFS2_SB(dentry->d_sb);
1629
1630 inode = ocfs2_get_system_file_inode(osb,
1631 GLOBAL_BITMAP_SYSTEM_INODE,
1632 OCFS2_INVALID_SLOT);
1633 if (!inode) {
1634 mlog(ML_ERROR, "failed to get bitmap inode\n");
1635 status = -EIO;
1636 goto bail;
1637 }
1638
1639 status = ocfs2_inode_lock(inode, &bh, 0);
1640 if (status < 0) {
1641 mlog_errno(status);
1642 goto bail;
1643 }
1644
1645 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1646
1647 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1648 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1649
1650 buf->f_type = OCFS2_SUPER_MAGIC;
1651 buf->f_bsize = dentry->d_sb->s_blocksize;
1652 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1653 buf->f_blocks = ((sector_t) numbits) *
1654 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1655 buf->f_bfree = ((sector_t) freebits) *
1656 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1657 buf->f_bavail = buf->f_bfree;
1658 buf->f_files = numbits;
1659 buf->f_ffree = freebits;
1660 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1661 & 0xFFFFFFFFUL;
1662 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1663 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
1664
1665 brelse(bh);
1666
1667 ocfs2_inode_unlock(inode, 0);
1668 status = 0;
1669bail:
1670 iput(inode);
1671
1672 if (status)
1673 mlog_errno(status);
1674
1675 return status;
1676}
1677
1678static void ocfs2_inode_init_once(void *data)
1679{
1680 struct ocfs2_inode_info *oi = data;
1681
1682 oi->ip_flags = 0;
1683 oi->ip_open_count = 0;
1684 spin_lock_init(&oi->ip_lock);
1685 ocfs2_extent_map_init(&oi->vfs_inode);
1686 INIT_LIST_HEAD(&oi->ip_io_markers);
1687 INIT_LIST_HEAD(&oi->ip_unwritten_list);
1688 oi->ip_dir_start_lookup = 0;
1689 init_rwsem(&oi->ip_alloc_sem);
1690 init_rwsem(&oi->ip_xattr_sem);
1691 mutex_init(&oi->ip_io_mutex);
1692
1693 oi->ip_blkno = 0ULL;
1694 oi->ip_clusters = 0;
1695
1696 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1697
1698 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1699 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1700 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1701
1702 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
1703 &ocfs2_inode_caching_ops);
1704
1705 inode_init_once(&oi->vfs_inode);
1706}
1707
1708static int ocfs2_initialize_mem_caches(void)
1709{
1710 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1711 sizeof(struct ocfs2_inode_info),
1712 0,
1713 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1714 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1715 ocfs2_inode_init_once);
1716 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1717 sizeof(struct ocfs2_dquot),
1718 0,
1719 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1720 SLAB_MEM_SPREAD),
1721 NULL);
1722 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1723 sizeof(struct ocfs2_quota_chunk),
1724 0,
1725 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1726 NULL);
1727 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1728 !ocfs2_qf_chunk_cachep) {
1729 kmem_cache_destroy(ocfs2_inode_cachep);
1730 kmem_cache_destroy(ocfs2_dquot_cachep);
1731 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1732 return -ENOMEM;
1733 }
1734
1735 return 0;
1736}
1737
1738static void ocfs2_free_mem_caches(void)
1739{
1740 /*
1741 * Make sure all delayed rcu free inodes are flushed before we
1742 * destroy cache.
1743 */
1744 rcu_barrier();
1745 kmem_cache_destroy(ocfs2_inode_cachep);
1746 ocfs2_inode_cachep = NULL;
1747
1748 kmem_cache_destroy(ocfs2_dquot_cachep);
1749 ocfs2_dquot_cachep = NULL;
1750
1751 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1752 ocfs2_qf_chunk_cachep = NULL;
1753}
1754
1755static int ocfs2_get_sector(struct super_block *sb,
1756 struct buffer_head **bh,
1757 int block,
1758 int sect_size)
1759{
1760 if (!sb_set_blocksize(sb, sect_size)) {
1761 mlog(ML_ERROR, "unable to set blocksize\n");
1762 return -EIO;
1763 }
1764
1765 *bh = sb_getblk(sb, block);
1766 if (!*bh) {
1767 mlog_errno(-ENOMEM);
1768 return -ENOMEM;
1769 }
1770 lock_buffer(*bh);
1771 if (!buffer_dirty(*bh))
1772 clear_buffer_uptodate(*bh);
1773 unlock_buffer(*bh);
1774 ll_rw_block(REQ_OP_READ, 0, 1, bh);
1775 wait_on_buffer(*bh);
1776 if (!buffer_uptodate(*bh)) {
1777 mlog_errno(-EIO);
1778 brelse(*bh);
1779 *bh = NULL;
1780 return -EIO;
1781 }
1782
1783 return 0;
1784}
1785
1786static int ocfs2_mount_volume(struct super_block *sb)
1787{
1788 int status = 0;
1789 int unlock_super = 0;
1790 struct ocfs2_super *osb = OCFS2_SB(sb);
1791
1792 if (ocfs2_is_hard_readonly(osb))
1793 goto leave;
1794
1795 mutex_init(&osb->obs_trim_fs_mutex);
1796
1797 status = ocfs2_dlm_init(osb);
1798 if (status < 0) {
1799 mlog_errno(status);
1800 if (status == -EBADR && ocfs2_userspace_stack(osb))
1801 mlog(ML_ERROR, "couldn't mount because cluster name on"
1802 " disk does not match the running cluster name.\n");
1803 goto leave;
1804 }
1805
1806 status = ocfs2_super_lock(osb, 1);
1807 if (status < 0) {
1808 mlog_errno(status);
1809 goto leave;
1810 }
1811 unlock_super = 1;
1812
1813 /* This will load up the node map and add ourselves to it. */
1814 status = ocfs2_find_slot(osb);
1815 if (status < 0) {
1816 mlog_errno(status);
1817 goto leave;
1818 }
1819
1820 /* load all node-local system inodes */
1821 status = ocfs2_init_local_system_inodes(osb);
1822 if (status < 0) {
1823 mlog_errno(status);
1824 goto leave;
1825 }
1826
1827 status = ocfs2_check_volume(osb);
1828 if (status < 0) {
1829 mlog_errno(status);
1830 goto leave;
1831 }
1832
1833 status = ocfs2_truncate_log_init(osb);
1834 if (status < 0)
1835 mlog_errno(status);
1836
1837leave:
1838 if (unlock_super)
1839 ocfs2_super_unlock(osb, 1);
1840
1841 return status;
1842}
1843
1844static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1845{
1846 int tmp, hangup_needed = 0;
1847 struct ocfs2_super *osb = NULL;
1848 char nodestr[12];
1849
1850 trace_ocfs2_dismount_volume(sb);
1851
1852 BUG_ON(!sb);
1853 osb = OCFS2_SB(sb);
1854 BUG_ON(!osb);
1855
1856 /* Remove file check sysfs related directores/files,
1857 * and wait for the pending file check operations */
1858 ocfs2_filecheck_remove_sysfs(osb);
1859
1860 kset_unregister(osb->osb_dev_kset);
1861
1862 /* Orphan scan should be stopped as early as possible */
1863 ocfs2_orphan_scan_stop(osb);
1864
1865 ocfs2_disable_quotas(osb);
1866
1867 /* All dquots should be freed by now */
1868 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1869 /* Wait for worker to be done with the work structure in osb */
1870 cancel_work_sync(&osb->dquot_drop_work);
1871
1872 ocfs2_shutdown_local_alloc(osb);
1873
1874 ocfs2_truncate_log_shutdown(osb);
1875
1876 /* This will disable recovery and flush any recovery work. */
1877 ocfs2_recovery_exit(osb);
1878
1879 ocfs2_journal_shutdown(osb);
1880
1881 ocfs2_sync_blockdev(sb);
1882
1883 ocfs2_purge_refcount_trees(osb);
1884
1885 /* No cluster connection means we've failed during mount, so skip
1886 * all the steps which depended on that to complete. */
1887 if (osb->cconn) {
1888 tmp = ocfs2_super_lock(osb, 1);
1889 if (tmp < 0) {
1890 mlog_errno(tmp);
1891 return;
1892 }
1893 }
1894
1895 if (osb->slot_num != OCFS2_INVALID_SLOT)
1896 ocfs2_put_slot(osb);
1897
1898 if (osb->cconn)
1899 ocfs2_super_unlock(osb, 1);
1900
1901 ocfs2_release_system_inodes(osb);
1902
1903 /*
1904 * If we're dismounting due to mount error, mount.ocfs2 will clean
1905 * up heartbeat. If we're a local mount, there is no heartbeat.
1906 * If we failed before we got a uuid_str yet, we can't stop
1907 * heartbeat. Otherwise, do it.
1908 */
1909 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1910 !ocfs2_is_hard_readonly(osb))
1911 hangup_needed = 1;
1912
1913 if (osb->cconn)
1914 ocfs2_dlm_shutdown(osb, hangup_needed);
1915
1916 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1917 debugfs_remove_recursive(osb->osb_debug_root);
1918
1919 if (hangup_needed)
1920 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1921
1922 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1923
1924 if (ocfs2_mount_local(osb))
1925 snprintf(nodestr, sizeof(nodestr), "local");
1926 else
1927 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1928
1929 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1930 osb->dev_str, nodestr);
1931
1932 ocfs2_delete_osb(osb);
1933 kfree(osb);
1934 sb->s_dev = 0;
1935 sb->s_fs_info = NULL;
1936}
1937
1938static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1939 unsigned uuid_bytes)
1940{
1941 int i, ret;
1942 char *ptr;
1943
1944 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1945
1946 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1947 if (osb->uuid_str == NULL)
1948 return -ENOMEM;
1949
1950 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1951 /* print with null */
1952 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1953 if (ret != 2) /* drop super cleans up */
1954 return -EINVAL;
1955 /* then only advance past the last char */
1956 ptr += 2;
1957 }
1958
1959 return 0;
1960}
1961
1962/* Make sure entire volume is addressable by our journal. Requires
1963 osb_clusters_at_boot to be valid and for the journal to have been
1964 initialized by ocfs2_journal_init(). */
1965static int ocfs2_journal_addressable(struct ocfs2_super *osb)
1966{
1967 int status = 0;
1968 u64 max_block =
1969 ocfs2_clusters_to_blocks(osb->sb,
1970 osb->osb_clusters_at_boot) - 1;
1971
1972 /* 32-bit block number is always OK. */
1973 if (max_block <= (u32)~0ULL)
1974 goto out;
1975
1976 /* Volume is "huge", so see if our journal is new enough to
1977 support it. */
1978 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
1979 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
1980 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
1981 JBD2_FEATURE_INCOMPAT_64BIT))) {
1982 mlog(ML_ERROR, "The journal cannot address the entire volume. "
1983 "Enable the 'block64' journal option with tunefs.ocfs2");
1984 status = -EFBIG;
1985 goto out;
1986 }
1987
1988 out:
1989 return status;
1990}
1991
1992static int ocfs2_initialize_super(struct super_block *sb,
1993 struct buffer_head *bh,
1994 int sector_size,
1995 struct ocfs2_blockcheck_stats *stats)
1996{
1997 int status;
1998 int i, cbits, bbits;
1999 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
2000 struct inode *inode = NULL;
2001 struct ocfs2_journal *journal;
2002 struct ocfs2_super *osb;
2003 u64 total_blocks;
2004
2005 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
2006 if (!osb) {
2007 status = -ENOMEM;
2008 mlog_errno(status);
2009 goto bail;
2010 }
2011
2012 sb->s_fs_info = osb;
2013 sb->s_op = &ocfs2_sops;
2014 sb->s_d_op = &ocfs2_dentry_ops;
2015 sb->s_export_op = &ocfs2_export_ops;
2016 sb->s_qcop = &dquot_quotactl_sysfile_ops;
2017 sb->dq_op = &ocfs2_quota_operations;
2018 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
2019 sb->s_xattr = ocfs2_xattr_handlers;
2020 sb->s_time_gran = 1;
2021 sb->s_flags |= SB_NOATIME;
2022 /* this is needed to support O_LARGEFILE */
2023 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2024 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2025 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
2026 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
2027 sizeof(di->id2.i_super.s_uuid));
2028
2029 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2030
2031 for (i = 0; i < 3; i++)
2032 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2033 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2034
2035 osb->sb = sb;
2036 osb->s_sectsize_bits = blksize_bits(sector_size);
2037 BUG_ON(!osb->s_sectsize_bits);
2038
2039 spin_lock_init(&osb->dc_task_lock);
2040 init_waitqueue_head(&osb->dc_event);
2041 osb->dc_work_sequence = 0;
2042 osb->dc_wake_sequence = 0;
2043 INIT_LIST_HEAD(&osb->blocked_lock_list);
2044 osb->blocked_lock_count = 0;
2045 spin_lock_init(&osb->osb_lock);
2046 spin_lock_init(&osb->osb_xattr_lock);
2047 ocfs2_init_steal_slots(osb);
2048
2049 mutex_init(&osb->system_file_mutex);
2050
2051 atomic_set(&osb->alloc_stats.moves, 0);
2052 atomic_set(&osb->alloc_stats.local_data, 0);
2053 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2054 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2055 atomic_set(&osb->alloc_stats.bg_extends, 0);
2056
2057 /* Copy the blockcheck stats from the superblock probe */
2058 osb->osb_ecc_stats = *stats;
2059
2060 ocfs2_init_node_maps(osb);
2061
2062 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2063 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2064
2065 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2066 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2067 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2068 osb->max_slots);
2069 status = -EINVAL;
2070 goto bail;
2071 }
2072
2073 ocfs2_orphan_scan_init(osb);
2074
2075 status = ocfs2_recovery_init(osb);
2076 if (status) {
2077 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2078 mlog_errno(status);
2079 goto bail;
2080 }
2081
2082 init_waitqueue_head(&osb->checkpoint_event);
2083
2084 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2085
2086 osb->slot_num = OCFS2_INVALID_SLOT;
2087
2088 osb->s_xattr_inline_size = le16_to_cpu(
2089 di->id2.i_super.s_xattr_inline_size);
2090
2091 osb->local_alloc_state = OCFS2_LA_UNUSED;
2092 osb->local_alloc_bh = NULL;
2093 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2094
2095 init_waitqueue_head(&osb->osb_mount_event);
2096
2097 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2098 if (status) {
2099 mlog_errno(status);
2100 goto bail;
2101 }
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 bail;
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 bail;
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 bail;
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 bail;
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 bail;
2149 }
2150
2151 if (ocfs2_clusterinfo_valid(osb)) {
2152 osb->osb_stackflags =
2153 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
2154 strlcpy(osb->osb_cluster_stack,
2155 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2156 OCFS2_STACK_LABEL_LEN + 1);
2157 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2158 mlog(ML_ERROR,
2159 "couldn't mount because of an invalid "
2160 "cluster stack label (%s) \n",
2161 osb->osb_cluster_stack);
2162 status = -EINVAL;
2163 goto bail;
2164 }
2165 strlcpy(osb->osb_cluster_name,
2166 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2167 OCFS2_CLUSTER_NAME_LEN + 1);
2168 } else {
2169 /* The empty string is identical with classic tools that
2170 * don't know about s_cluster_info. */
2171 osb->osb_cluster_stack[0] = '\0';
2172 }
2173
2174 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2175
2176 /* FIXME
2177 * This should be done in ocfs2_journal_init(), but unknown
2178 * ordering issues will cause the filesystem to crash.
2179 * If anyone wants to figure out what part of the code
2180 * refers to osb->journal before ocfs2_journal_init() is run,
2181 * be my guest.
2182 */
2183 /* initialize our journal structure */
2184
2185 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2186 if (!journal) {
2187 mlog(ML_ERROR, "unable to alloc journal\n");
2188 status = -ENOMEM;
2189 goto bail;
2190 }
2191 osb->journal = journal;
2192 journal->j_osb = osb;
2193
2194 atomic_set(&journal->j_num_trans, 0);
2195 init_rwsem(&journal->j_trans_barrier);
2196 init_waitqueue_head(&journal->j_checkpointed);
2197 spin_lock_init(&journal->j_lock);
2198 journal->j_trans_id = (unsigned long) 1;
2199 INIT_LIST_HEAD(&journal->j_la_cleanups);
2200 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2201 journal->j_state = OCFS2_JOURNAL_FREE;
2202
2203 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2204 init_llist_head(&osb->dquot_drop_list);
2205
2206 /* get some pseudo constants for clustersize bits */
2207 osb->s_clustersize_bits =
2208 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2209 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2210
2211 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2212 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2213 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2214 osb->s_clustersize);
2215 status = -EINVAL;
2216 goto bail;
2217 }
2218
2219 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2220 le32_to_cpu(di->i_clusters));
2221
2222 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2223 total_blocks);
2224 if (status) {
2225 mlog(ML_ERROR, "Volume too large "
2226 "to mount safely on this system");
2227 status = -EFBIG;
2228 goto bail;
2229 }
2230
2231 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2232 sizeof(di->id2.i_super.s_uuid))) {
2233 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2234 status = -ENOMEM;
2235 goto bail;
2236 }
2237
2238 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2239 OCFS2_MAX_VOL_LABEL_LEN);
2240 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2241 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2242 osb->first_cluster_group_blkno =
2243 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2244 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2245 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2246 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2247 (unsigned long long)osb->root_blkno,
2248 (unsigned long long)osb->system_dir_blkno,
2249 osb->s_clustersize_bits);
2250
2251 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2252 if (!osb->osb_dlm_debug) {
2253 status = -ENOMEM;
2254 mlog_errno(status);
2255 goto bail;
2256 }
2257
2258 atomic_set(&osb->vol_state, VOLUME_INIT);
2259
2260 /* load root, system_dir, and all global system inodes */
2261 status = ocfs2_init_global_system_inodes(osb);
2262 if (status < 0) {
2263 mlog_errno(status);
2264 goto bail;
2265 }
2266
2267 /*
2268 * global bitmap
2269 */
2270 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2271 OCFS2_INVALID_SLOT);
2272 if (!inode) {
2273 status = -EINVAL;
2274 mlog_errno(status);
2275 goto bail;
2276 }
2277
2278 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2279 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
2280 iput(inode);
2281
2282 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2283 osb->s_feature_incompat) * 8;
2284
2285 status = ocfs2_init_slot_info(osb);
2286 if (status < 0) {
2287 mlog_errno(status);
2288 goto bail;
2289 }
2290 cleancache_init_shared_fs(sb);
2291
2292 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
2293 if (!osb->ocfs2_wq) {
2294 status = -ENOMEM;
2295 mlog_errno(status);
2296 }
2297
2298bail:
2299 return status;
2300}
2301
2302/*
2303 * will return: -EAGAIN if it is ok to keep searching for superblocks
2304 * -EINVAL if there is a bad superblock
2305 * 0 on success
2306 */
2307static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2308 struct buffer_head *bh,
2309 u32 blksz,
2310 struct ocfs2_blockcheck_stats *stats)
2311{
2312 int status = -EAGAIN;
2313
2314 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2315 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2316 /* We have to do a raw check of the feature here */
2317 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2318 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2319 status = ocfs2_block_check_validate(bh->b_data,
2320 bh->b_size,
2321 &di->i_check,
2322 stats);
2323 if (status)
2324 goto out;
2325 }
2326 status = -EINVAL;
2327 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2328 mlog(ML_ERROR, "found superblock with incorrect block "
2329 "size: found %u, should be %u\n",
2330 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2331 blksz);
2332 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2333 OCFS2_MAJOR_REV_LEVEL ||
2334 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2335 OCFS2_MINOR_REV_LEVEL) {
2336 mlog(ML_ERROR, "found superblock with bad version: "
2337 "found %u.%u, should be %u.%u\n",
2338 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2339 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2340 OCFS2_MAJOR_REV_LEVEL,
2341 OCFS2_MINOR_REV_LEVEL);
2342 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2343 mlog(ML_ERROR, "bad block number on superblock: "
2344 "found %llu, should be %llu\n",
2345 (unsigned long long)le64_to_cpu(di->i_blkno),
2346 (unsigned long long)bh->b_blocknr);
2347 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2348 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2349 mlog(ML_ERROR, "bad cluster size found: %u\n",
2350 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2351 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2352 mlog(ML_ERROR, "bad root_blkno: 0\n");
2353 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2354 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2355 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2356 mlog(ML_ERROR,
2357 "Superblock slots found greater than file system "
2358 "maximum: found %u, max %u\n",
2359 le16_to_cpu(di->id2.i_super.s_max_slots),
2360 OCFS2_MAX_SLOTS);
2361 } else {
2362 /* found it! */
2363 status = 0;
2364 }
2365 }
2366
2367out:
2368 if (status && status != -EAGAIN)
2369 mlog_errno(status);
2370 return status;
2371}
2372
2373static int ocfs2_check_volume(struct ocfs2_super *osb)
2374{
2375 int status;
2376 int dirty;
2377 int local;
2378 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2379 * recover
2380 * ourselves. */
2381
2382 /* Init our journal object. */
2383 status = ocfs2_journal_init(osb->journal, &dirty);
2384 if (status < 0) {
2385 mlog(ML_ERROR, "Could not initialize journal!\n");
2386 goto finally;
2387 }
2388
2389 /* Now that journal has been initialized, check to make sure
2390 entire volume is addressable. */
2391 status = ocfs2_journal_addressable(osb);
2392 if (status)
2393 goto finally;
2394
2395 /* If the journal was unmounted cleanly then we don't want to
2396 * recover anything. Otherwise, journal_load will do that
2397 * dirty work for us :) */
2398 if (!dirty) {
2399 status = ocfs2_journal_wipe(osb->journal, 0);
2400 if (status < 0) {
2401 mlog_errno(status);
2402 goto finally;
2403 }
2404 } else {
2405 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2406 "unmounted cleanly, recovering it.\n", osb->dev_str);
2407 }
2408
2409 local = ocfs2_mount_local(osb);
2410
2411 /* will play back anything left in the journal. */
2412 status = ocfs2_journal_load(osb->journal, local, dirty);
2413 if (status < 0) {
2414 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2415 goto finally;
2416 }
2417
2418 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2419 jbd2_journal_set_features(osb->journal->j_journal,
2420 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2421 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2422 else
2423 jbd2_journal_clear_features(osb->journal->j_journal,
2424 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2425 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2426
2427 if (dirty) {
2428 /* recover my local alloc if we didn't unmount cleanly. */
2429 status = ocfs2_begin_local_alloc_recovery(osb,
2430 osb->slot_num,
2431 &local_alloc);
2432 if (status < 0) {
2433 mlog_errno(status);
2434 goto finally;
2435 }
2436 /* we complete the recovery process after we've marked
2437 * ourselves as mounted. */
2438 }
2439
2440 status = ocfs2_load_local_alloc(osb);
2441 if (status < 0) {
2442 mlog_errno(status);
2443 goto finally;
2444 }
2445
2446 if (dirty) {
2447 /* Recovery will be completed after we've mounted the
2448 * rest of the volume. */
2449 osb->local_alloc_copy = local_alloc;
2450 local_alloc = NULL;
2451 }
2452
2453 /* go through each journal, trylock it and if you get the
2454 * lock, and it's marked as dirty, set the bit in the recover
2455 * map and launch a recovery thread for it. */
2456 status = ocfs2_mark_dead_nodes(osb);
2457 if (status < 0) {
2458 mlog_errno(status);
2459 goto finally;
2460 }
2461
2462 status = ocfs2_compute_replay_slots(osb);
2463 if (status < 0)
2464 mlog_errno(status);
2465
2466finally:
2467 kfree(local_alloc);
2468
2469 if (status)
2470 mlog_errno(status);
2471 return status;
2472}
2473
2474/*
2475 * The routine gets called from dismount or close whenever a dismount on
2476 * volume is requested and the osb open count becomes 1.
2477 * It will remove the osb from the global list and also free up all the
2478 * initialized resources and fileobject.
2479 */
2480static void ocfs2_delete_osb(struct ocfs2_super *osb)
2481{
2482 /* This function assumes that the caller has the main osb resource */
2483
2484 /* ocfs2_initializer_super have already created this workqueue */
2485 if (osb->ocfs2_wq)
2486 destroy_workqueue(osb->ocfs2_wq);
2487
2488 ocfs2_free_slot_info(osb);
2489
2490 kfree(osb->osb_orphan_wipes);
2491 kfree(osb->slot_recovery_generations);
2492 /* FIXME
2493 * This belongs in journal shutdown, but because we have to
2494 * allocate osb->journal at the start of ocfs2_initialize_osb(),
2495 * we free it here.
2496 */
2497 kfree(osb->journal);
2498 kfree(osb->local_alloc_copy);
2499 kfree(osb->uuid_str);
2500 kfree(osb->vol_label);
2501 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2502 memset(osb, 0, sizeof(struct ocfs2_super));
2503}
2504
2505/* Depending on the mount option passed, perform one of the following:
2506 * Put OCFS2 into a readonly state (default)
2507 * Return EIO so that only the process errs
2508 * Fix the error as if fsck.ocfs2 -y
2509 * panic
2510 */
2511static int ocfs2_handle_error(struct super_block *sb)
2512{
2513 struct ocfs2_super *osb = OCFS2_SB(sb);
2514 int rv = 0;
2515
2516 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2517 pr_crit("On-disk corruption discovered. "
2518 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
2519
2520 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2521 panic("OCFS2: (device %s): panic forced after error\n",
2522 sb->s_id);
2523 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2524 pr_crit("OCFS2: Returning error to the calling process.\n");
2525 rv = -EIO;
2526 } else { /* default option */
2527 rv = -EROFS;
2528 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
2529 return rv;
2530
2531 pr_crit("OCFS2: File system is now read-only.\n");
2532 sb->s_flags |= SB_RDONLY;
2533 ocfs2_set_ro_flag(osb, 0);
2534 }
2535
2536 return rv;
2537}
2538
2539int __ocfs2_error(struct super_block *sb, const char *function,
2540 const char *fmt, ...)
2541{
2542 struct va_format vaf;
2543 va_list args;
2544
2545 va_start(args, fmt);
2546 vaf.fmt = fmt;
2547 vaf.va = &args;
2548
2549 /* Not using mlog here because we want to show the actual
2550 * function the error came from. */
2551 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
2552 sb->s_id, function, &vaf);
2553
2554 va_end(args);
2555
2556 return ocfs2_handle_error(sb);
2557}
2558
2559/* Handle critical errors. This is intentionally more drastic than
2560 * ocfs2_handle_error, so we only use for things like journal errors,
2561 * etc. */
2562void __ocfs2_abort(struct super_block *sb, const char *function,
2563 const char *fmt, ...)
2564{
2565 struct va_format vaf;
2566 va_list args;
2567
2568 va_start(args, fmt);
2569
2570 vaf.fmt = fmt;
2571 vaf.va = &args;
2572
2573 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
2574 sb->s_id, function, &vaf);
2575
2576 va_end(args);
2577
2578 /* We don't have the cluster support yet to go straight to
2579 * hard readonly in here. Until then, we want to keep
2580 * ocfs2_abort() so that we can at least mark critical
2581 * errors.
2582 *
2583 * TODO: This should abort the journal and alert other nodes
2584 * that our slot needs recovery. */
2585
2586 /* Force a panic(). This stinks, but it's better than letting
2587 * things continue without having a proper hard readonly
2588 * here. */
2589 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2590 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2591 ocfs2_handle_error(sb);
2592}
2593
2594/*
2595 * Void signal blockers, because in-kernel sigprocmask() only fails
2596 * when SIG_* is wrong.
2597 */
2598void ocfs2_block_signals(sigset_t *oldset)
2599{
2600 int rc;
2601 sigset_t blocked;
2602
2603 sigfillset(&blocked);
2604 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2605 BUG_ON(rc);
2606}
2607
2608void ocfs2_unblock_signals(sigset_t *oldset)
2609{
2610 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2611 BUG_ON(rc);
2612}
2613
2614module_init(ocfs2_init);
2615module_exit(ocfs2_exit);