projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
PM / Domains: Factorize dev_pm_genpd_set_performance_state()
[linux-2.6-block.git] / fs / ocfs2 / super.c
CommitLineData
ccd979bd
MF		 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>
ccd979bd
MF		 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>
d550071c 41#include <linux/mount.h>
6953b4c0 42#include <linux/seq_file.h>
19ece546 43#include <linux/quotaops.h>
1cfd8bd0 44#include <linux/cleancache.h>
f361bf4a 45#include <linux/signal.h>
ccd979bd 46
80a9a84d
WW		 47#define CREATE_TRACE_POINTS
48#include "ocfs2_trace.h"
49
ccd979bd
MF		 50#include <cluster/masklog.h>
51
52#include "ocfs2.h"
53
54/* this should be the only file to include a version 1 header */
55#include "ocfs1_fs_compat.h"
56
57#include "alloc.h"
a11f7e63 58#include "aops.h"
d030cc97 59#include "blockcheck.h"
ccd979bd
MF		 60#include "dlmglue.h"
61#include "export.h"
62#include "extent_map.h"
63#include "heartbeat.h"
64#include "inode.h"
65#include "journal.h"
66#include "localalloc.h"
67#include "namei.h"
68#include "slot_map.h"
69#include "super.h"
70#include "sysfile.h"
71#include "uptodate.h"
cf1d6c76 72#include "xattr.h"
9e33d69f 73#include "quota.h"
374a263e 74#include "refcounttree.h"
b89c5428 75#include "suballoc.h"
ccd979bd
MF		 76
77#include "buffer_head_io.h"
a849d468 78#include "filecheck.h"
ccd979bd 79
1a5c4e2a 80static struct kmem_cache *ocfs2_inode_cachep;
9e33d69f
JK		 81struct kmem_cache *ocfs2_dquot_cachep;
82struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 83
1a5c4e2a 84static struct dentry *ocfs2_debugfs_root;
ccd979bd
MF		 85
86MODULE_AUTHOR("Oracle");
87MODULE_LICENSE("GPL");
ff8fb335 88MODULE_DESCRIPTION("OCFS2 cluster file system");
ccd979bd 89
c0123ade
TY		 90struct mount_options
91{
d147b3d6 92 unsigned long commit_interval;
c0123ade
TY		 93 unsigned long mount_opt;
94 unsigned int atime_quantum;
95 signed short slot;
73c8a800 96 int localalloc_opt;
d02f00cc 97 unsigned int resv_level;
83f92318 98 int dir_resv_level;
b61817e1 99 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 100};
101
ccd979bd 102static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 103 struct mount_options *mopt,
baf4661a 104 int is_remount);
5297aad8
JK		 105static int ocfs2_check_set_options(struct super_block *sb,
106 struct mount_options *options);
34c80b1d 107static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
ccd979bd
MF		 108static void ocfs2_put_super(struct super_block *sb);
109static int ocfs2_mount_volume(struct super_block *sb);
110static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112static int ocfs2_initialize_mem_caches(void);
113static void ocfs2_free_mem_caches(void);
114static void ocfs2_delete_osb(struct ocfs2_super *osb);
115
726c3342 116static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 117
118static int ocfs2_sync_fs(struct super_block *sb, int wait);
119
120static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 122static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 123static int ocfs2_check_volume(struct ocfs2_super *osb);
124static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125 struct buffer_head *bh,
73be192b
JB		 126 u32 sectsize,
127 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 128static int ocfs2_initialize_super(struct super_block *sb,
129 struct buffer_head *bh,
73be192b
JB		 130 int sector_size,
131 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 132static int ocfs2_get_sector(struct super_block *sb,
133 struct buffer_head **bh,
134 int block,
135 int sect_size);
ccd979bd
MF		 136static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 138static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 141
1c92ec67
JK		 142static struct dquot **ocfs2_get_dquots(struct inode *inode)
143{
144 return OCFS2_I(inode)->i_dquot;
145}
146
ee9b6d61 147static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 148 .statfs = ocfs2_statfs,
149 .alloc_inode = ocfs2_alloc_inode,
150 .destroy_inode = ocfs2_destroy_inode,
151 .drop_inode = ocfs2_drop_inode,
066d92dc 152 .evict_inode = ocfs2_evict_inode,
ccd979bd 153 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 154 .put_super = ocfs2_put_super,
155 .remount_fs = ocfs2_remount,
d550071c 156 .show_options = ocfs2_show_options,
9e33d69f
JK		 157 .quota_read = ocfs2_quota_read,
158 .quota_write = ocfs2_quota_write,
1c92ec67 159 .get_dquots = ocfs2_get_dquots,
ccd979bd
MF		 160};
161
162enum {
163 Opt_barrier,
164 Opt_err_panic,
165 Opt_err_ro,
166 Opt_intr,
167 Opt_nointr,
168 Opt_hb_none,
169 Opt_hb_local,
2c442719 170 Opt_hb_global,
ccd979bd
MF		 171 Opt_data_ordered,
172 Opt_data_writeback,
7f1a37e3 173 Opt_atime_quantum,
baf4661a 174 Opt_slot,
d147b3d6 175 Opt_commit,
2fbe8d1e 176 Opt_localalloc,
53fc622b 177 Opt_localflocks,
b61817e1 178 Opt_stack,
cf1d6c76
TY		 179 Opt_user_xattr,
180 Opt_nouser_xattr,
12462f1d 181 Opt_inode64,
a68979b8
TY		 182 Opt_acl,
183 Opt_noacl,
19ece546
JK		 184 Opt_usrquota,
185 Opt_grpquota,
7bdb0d18
TY		 186 Opt_coherency_buffered,
187 Opt_coherency_full,
d02f00cc 188 Opt_resv_level,
83f92318 189 Opt_dir_resv_level,
1dfeb768 190 Opt_journal_async_commit,
7d0fb914 191 Opt_err_cont,
ccd979bd
MF		 192 Opt_err,
193};
194
a447c093 195static const match_table_t tokens = {
ccd979bd
MF		 196 {Opt_barrier, "barrier=%u"},
197 {Opt_err_panic, "errors=panic"},
198 {Opt_err_ro, "errors=remount-ro"},
199 {Opt_intr, "intr"},
200 {Opt_nointr, "nointr"},
201 {Opt_hb_none, OCFS2_HB_NONE},
202 {Opt_hb_local, OCFS2_HB_LOCAL},
2c442719 203 {Opt_hb_global, OCFS2_HB_GLOBAL},
ccd979bd
MF		 204 {Opt_data_ordered, "data=ordered"},
205 {Opt_data_writeback, "data=writeback"},
7f1a37e3 206 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 207 {Opt_slot, "preferred_slot=%u"},
d147b3d6 208 {Opt_commit, "commit=%u"},
2fbe8d1e 209 {Opt_localalloc, "localalloc=%d"},
53fc622b 210 {Opt_localflocks, "localflocks"},
b61817e1 211 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 212 {Opt_user_xattr, "user_xattr"},
213 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 214 {Opt_inode64, "inode64"},
a68979b8
TY		 215 {Opt_acl, "acl"},
216 {Opt_noacl, "noacl"},
19ece546
JK		 217 {Opt_usrquota, "usrquota"},
218 {Opt_grpquota, "grpquota"},
7bdb0d18
TY		 219 {Opt_coherency_buffered, "coherency=buffered"},
220 {Opt_coherency_full, "coherency=full"},
d02f00cc 221 {Opt_resv_level, "resv_level=%u"},
83f92318 222 {Opt_dir_resv_level, "dir_resv_level=%u"},
1dfeb768 223 {Opt_journal_async_commit, "journal_async_commit"},
7d0fb914 224 {Opt_err_cont, "errors=continue"},
ccd979bd
MF		 225 {Opt_err, NULL}
226};
227
50397507
SM		 228#ifdef CONFIG_DEBUG_FS
229static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
230{
50397507
SM		 231 struct ocfs2_cluster_connection *cconn = osb->cconn;
232 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 233 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
234 int i, out = 0;
bfd97a03 235 unsigned long flags;
50397507
SM		 236
237 out += snprintf(buf + out, len - out,
238 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
239 "Device", osb->dev_str, osb->uuid_str,
240 osb->fs_generation, osb->vol_label);
241
242 out += snprintf(buf + out, len - out,
243 "%10s => State: %d Flags: 0x%lX\n", "Volume",
244 atomic_read(&osb->vol_state), osb->osb_flags);
245
246 out += snprintf(buf + out, len - out,
247 "%10s => Block: %lu Cluster: %d\n", "Sizes",
248 osb->sb->s_blocksize, osb->s_clustersize);
249
250 out += snprintf(buf + out, len - out,
251 "%10s => Compat: 0x%X Incompat: 0x%X "
252 "ROcompat: 0x%X\n",
253 "Features", osb->s_feature_compat,
254 osb->s_feature_incompat, osb->s_feature_ro_compat);
255
256 out += snprintf(buf + out, len - out,
257 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
258 osb->s_mount_opt, osb->s_atime_quantum);
259
c3d38840
SM		 260 if (cconn) {
261 out += snprintf(buf + out, len - out,
262 "%10s => Stack: %s Name: %*s "
263 "Version: %d.%d\n", "Cluster",
264 (*osb->osb_cluster_stack == '\0' ?
265 "o2cb" : osb->osb_cluster_stack),
266 cconn->cc_namelen, cconn->cc_name,
267 cconn->cc_version.pv_major,
268 cconn->cc_version.pv_minor);
269 }
50397507 270
bfd97a03 271 spin_lock_irqsave(&osb->dc_task_lock, flags);
50397507
SM		 272 out += snprintf(buf + out, len - out,
273 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
274 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 275 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
276 osb->blocked_lock_count, osb->dc_wake_sequence,
277 osb->dc_work_sequence);
bfd97a03 278 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
50397507
SM		 279
280 spin_lock(&osb->osb_lock);
281 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
282 "Recovery",
283 (osb->recovery_thread_task ?
284 task_pid_nr(osb->recovery_thread_task) : -1));
285 if (rm->rm_used == 0)
286 out += snprintf(buf + out, len - out, " None\n");
287 else {
288 for (i = 0; i < rm->rm_used; i++)
289 out += snprintf(buf + out, len - out, " %d",
290 rm->rm_entries[i]);
291 out += snprintf(buf + out, len - out, "\n");
292 }
293 spin_unlock(&osb->osb_lock);
294
295 out += snprintf(buf + out, len - out,
8fa9d17f 296 "%10s => Pid: %d Interval: %lu\n", "Commit",
c3d38840 297 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
8fa9d17f 298 osb->osb_commit_interval);
50397507
SM		 299
300 out += snprintf(buf + out, len - out,
c3d38840 301 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 302 "Journal", osb->journal->j_state,
c3d38840
SM		 303 osb->journal->j_trans_id,
304 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 305
306 out += snprintf(buf + out, len - out,
307 "%10s => GlobalAllocs: %d LocalAllocs: %d "
308 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
309 "Stats",
310 atomic_read(&osb->alloc_stats.bitmap_data),
311 atomic_read(&osb->alloc_stats.local_data),
312 atomic_read(&osb->alloc_stats.bg_allocs),
313 atomic_read(&osb->alloc_stats.moves),
314 atomic_read(&osb->alloc_stats.bg_extends));
315
316 out += snprintf(buf + out, len - out,
317 "%10s => State: %u Descriptor: %llu Size: %u bits "
318 "Default: %u bits\n",
319 "LocalAlloc", osb->local_alloc_state,
320 (unsigned long long)osb->la_last_gd,
321 osb->local_alloc_bits, osb->local_alloc_default_bits);
322
323 spin_lock(&osb->osb_lock);
324 out += snprintf(buf + out, len - out,
b89c5428
TY		 325 "%10s => InodeSlot: %d StolenInodes: %d, "
326 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 327 osb->s_inode_steal_slot,
b89c5428
TY		 328 atomic_read(&osb->s_num_inodes_stolen),
329 osb->s_meta_steal_slot,
330 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 331 spin_unlock(&osb->osb_lock);
332
df152c24
SM		 333 out += snprintf(buf + out, len - out, "OrphanScan => ");
334 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
335 os->os_count, os->os_seqno);
336 out += snprintf(buf + out, len - out, " Last Scan: ");
337 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
338 out += snprintf(buf + out, len - out, "Disabled\n");
339 else
340 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
395627b0 341 (unsigned long)(ktime_get_seconds() - os->os_scantime));
df152c24 342
50397507
SM		 343 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
344 "Slots", "Num", "RecoGen");
50397507
SM		 345 for (i = 0; i < osb->max_slots; ++i) {
346 out += snprintf(buf + out, len - out,
347 "%10s %c %3d %10d\n",
348 " ",
349 (i == osb->slot_num ? '*' : ' '),
350 i, osb->slot_recovery_generations[i]);
351 }
352
353 return out;
354}
355
356static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
357{
358 struct ocfs2_super *osb = inode->i_private;
359 char *buf = NULL;
360
361 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
362 if (!buf)
363 goto bail;
364
365 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
366
367 file->private_data = buf;
368
369 return 0;
370bail:
371 return -ENOMEM;
372}
373
374static int ocfs2_debug_release(struct inode *inode, struct file *file)
375{
376 kfree(file->private_data);
377 return 0;
378}
379
380static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
381 size_t nbytes, loff_t *ppos)
382{
383 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
384 i_size_read(file->f_mapping->host));
385}
386#else
387static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
388{
389 return 0;
390}
391static int ocfs2_debug_release(struct inode *inode, struct file *file)
392{
393 return 0;
394}
395static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
396 size_t nbytes, loff_t *ppos)
397{
398 return 0;
399}
400#endif /* CONFIG_DEBUG_FS */
401
828c0950 402static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 403 .open = ocfs2_osb_debug_open,
404 .release = ocfs2_debug_release,
405 .read = ocfs2_debug_read,
406 .llseek = generic_file_llseek,
407};
408
ccd979bd
MF		 409static int ocfs2_sync_fs(struct super_block *sb, int wait)
410{
bddb8eb3 411 int status;
ccd979bd
MF		 412 tid_t target;
413 struct ocfs2_super *osb = OCFS2_SB(sb);
414
ccd979bd
MF		 415 if (ocfs2_is_hard_readonly(osb))
416 return -EROFS;
417
418 if (wait) {
419 status = ocfs2_flush_truncate_log(osb);
420 if (status < 0)
421 mlog_errno(status);
422 } else {
423 ocfs2_schedule_truncate_log_flush(osb, 0);
424 }
425
1119d3c0 426 if (jbd2_journal_start_commit(osb->journal->j_journal,
2b4e30fb 427 &target)) {
ccd979bd 428 if (wait)
1119d3c0 429 jbd2_log_wait_commit(osb->journal->j_journal,
2b4e30fb 430 target);
ccd979bd
MF		 431 }
432 return 0;
433}
434
1a224ad1
JK		 435static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
436{
437 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
438 && (ino == USER_QUOTA_SYSTEM_INODE
439 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
440 return 0;
441 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
442 && (ino == GROUP_QUOTA_SYSTEM_INODE
443 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
444 return 0;
445 return 1;
446}
447
ccd979bd
MF		 448static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
449{
450 struct inode *new = NULL;
451 int status = 0;
452 int i;
453
5fa0613e 454 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 455 if (IS_ERR(new)) {
456 status = PTR_ERR(new);
457 mlog_errno(status);
458 goto bail;
459 }
460 osb->root_inode = new;
461
5fa0613e 462 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 463 if (IS_ERR(new)) {
464 status = PTR_ERR(new);
465 mlog_errno(status);
466 goto bail;
467 }
468 osb->sys_root_inode = new;
469
470 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
471 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 472 if (!ocfs2_need_system_inode(osb, i))
473 continue;
ccd979bd
MF		 474 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
475 if (!new) {
476 ocfs2_release_system_inodes(osb);
025bcbde 477 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
ccd979bd 478 mlog_errno(status);
ccd979bd
MF		 479 mlog(ML_ERROR, "Unable to load system inode %d, "
480 "possibly corrupt fs?", i);
481 goto bail;
482 }
483 // the array now has one ref, so drop this one
484 iput(new);
485 }
486
487bail:
c1e8d35e
TM		 488 if (status)
489 mlog_errno(status);
ccd979bd
MF		 490 return status;
491}
492
493static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
494{
495 struct inode *new = NULL;
496 int status = 0;
497 int i;
498
ccd979bd
MF		 499 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
500 i < NUM_SYSTEM_INODES;
501 i++) {
1a224ad1
JK		 502 if (!ocfs2_need_system_inode(osb, i))
503 continue;
ccd979bd
MF		 504 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
505 if (!new) {
506 ocfs2_release_system_inodes(osb);
025bcbde 507 status = ocfs2_is_soft_readonly(osb) ? -EROFS : -EINVAL;
ccd979bd
MF		 508 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
509 status, i, osb->slot_num);
510 goto bail;
511 }
512 /* the array now has one ref, so drop this one */
513 iput(new);
514 }
515
516bail:
c1e8d35e
TM		 517 if (status)
518 mlog_errno(status);
ccd979bd
MF		 519 return status;
520}
521
bddb8eb3 522static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 523{
bddb8eb3 524 int i;
ccd979bd
MF		 525 struct inode *inode;
526
b4d693fc
TM		 527 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
528 inode = osb->global_system_inodes[i];
ccd979bd
MF		 529 if (inode) {
530 iput(inode);
b4d693fc 531 osb->global_system_inodes[i] = NULL;
ccd979bd
MF		 532 }
533 }
534
535 inode = osb->sys_root_inode;
536 if (inode) {
537 iput(inode);
538 osb->sys_root_inode = NULL;
539 }
540
541 inode = osb->root_inode;
542 if (inode) {
543 iput(inode);
544 osb->root_inode = NULL;
545 }
546
b4d693fc 547 if (!osb->local_system_inodes)
c1e8d35e 548 return;
b4d693fc
TM		 549
550 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
551 if (osb->local_system_inodes[i]) {
552 iput(osb->local_system_inodes[i]);
553 osb->local_system_inodes[i] = NULL;
554 }
555 }
556
557 kfree(osb->local_system_inodes);
558 osb->local_system_inodes = NULL;
ccd979bd
MF		 559}
560
561/* We're allocating fs objects, use GFP_NOFS */
562static struct inode *ocfs2_alloc_inode(struct super_block *sb)
563{
564 struct ocfs2_inode_info *oi;
565
e6b4f8da 566 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 567 if (!oi)
568 return NULL;
569
2931cdcb
DW		 570 oi->i_sync_tid = 0;
571 oi->i_datasync_tid = 0;
1c92ec67 572 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
2931cdcb 573
2b4e30fb 574 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 575 return &oi->vfs_inode;
576}
577
fa0d7e3d 578static void ocfs2_i_callback(struct rcu_head *head)
ccd979bd 579{
fa0d7e3d 580 struct inode *inode = container_of(head, struct inode, i_rcu);
ccd979bd
MF		 581 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
582}
583
fa0d7e3d
NP		 584static void ocfs2_destroy_inode(struct inode *inode)
585{
586 call_rcu(&inode->i_rcu, ocfs2_i_callback);
587}
588
5a254031
MF		 589static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
590 unsigned int cbits)
ccd979bd 591{
5a254031
MF		 592 unsigned int bytes = 1 << cbits;
593 unsigned int trim = bytes;
594 unsigned int bitshift = 32;
595
596 /*
597 * i_size and all block offsets in ocfs2 are always 64 bits
598 * wide. i_clusters is 32 bits, in cluster-sized units. So on
599 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 600 */
601
602#if BITS_PER_LONG == 32
90c699a9 603# if defined(CONFIG_LBDAF)
2ecd05ae 604 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 605 /*
606 * We might be limited by page cache size.
607 */
09cbfeaf
KS		 608 if (bytes > PAGE_SIZE) {
609 bytes = PAGE_SIZE;
5a254031
MF		 610 trim = 1;
611 /*
612 * Shift by 31 here so that we don't get larger than
613 * MAX_LFS_FILESIZE
614 */
615 bitshift = 31;
616 }
ccd979bd 617# else
5a254031
MF		 618 /*
619 * We are limited by the size of sector_t. Use block size, as
620 * that's what we expose to the VFS.
621 */
622 bytes = 1 << bbits;
623 trim = 1;
624 bitshift = 31;
ccd979bd
MF		 625# endif
626#endif
627
5a254031
MF		 628 /*
629 * Trim by a whole cluster when we can actually approach the
630 * on-disk limits. Otherwise we can overflow i_clusters when
631 * an extent start is at the max offset.
632 */
633 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 634}
635
636static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
637{
638 int incompat_features;
639 int ret = 0;
c0123ade 640 struct mount_options parsed_options;
ccd979bd 641 struct ocfs2_super *osb = OCFS2_SB(sb);
2c442719 642 u32 tmp;
ccd979bd 643
02b9984d
TT		 644 sync_filesystem(sb);
645
5297aad8
JK		 646 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
647 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF		 648 ret = -EINVAL;
649 goto out;
650 }
651
2c442719
SM		 652 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
653 OCFS2_MOUNT_HB_NONE;
654 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
ccd979bd
MF		 655 ret = -EINVAL;
656 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
657 goto out;
658 }
659
660 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 661 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 662 ret = -EINVAL;
663 mlog(ML_ERROR, "Cannot change data mode on remount\n");
664 goto out;
665 }
666
12462f1d
JB		 667 /* Probably don't want this on remount; it might
668 * mess with other nodes */
669 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
670 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
671 ret = -EINVAL;
672 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
673 goto out;
674 }
675
ccd979bd 676 /* We're going to/from readonly mode. */
1751e8a6 677 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
19ece546 678 /* Disable quota accounting before remounting RO */
1751e8a6 679 if (*flags & SB_RDONLY) {
19ece546
JK		 680 ret = ocfs2_susp_quotas(osb, 0);
681 if (ret < 0)
682 goto out;
683 }
ccd979bd
MF		 684 /* Lock here so the check of HARD_RO and the potential
685 * setting of SOFT_RO is atomic. */
686 spin_lock(&osb->osb_lock);
687 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
688 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
689 ret = -EROFS;
690 goto unlock_osb;
691 }
692
1751e8a6
LT		 693 if (*flags & SB_RDONLY) {
694 sb->s_flags |= SB_RDONLY;
ccd979bd
MF		 695 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
696 } else {
ccd979bd
MF		 697 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
698 mlog(ML_ERROR, "Cannot remount RDWR "
699 "filesystem due to previous errors.\n");
700 ret = -EROFS;
701 goto unlock_osb;
702 }
703 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
704 if (incompat_features) {
705 mlog(ML_ERROR, "Cannot remount RDWR because "
706 "of unsupported optional features "
707 "(%x).\n", incompat_features);
708 ret = -EINVAL;
709 goto unlock_osb;
710 }
1751e8a6 711 sb->s_flags &= ~SB_RDONLY;
ccd979bd
MF		 712 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
713 }
32a42d39 714 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
ccd979bd
MF		 715unlock_osb:
716 spin_unlock(&osb->osb_lock);
19ece546 717 /* Enable quota accounting after remounting RW */
1751e8a6 718 if (!ret && !(*flags & SB_RDONLY)) {
19ece546
JK		 719 if (sb_any_quota_suspended(sb))
720 ret = ocfs2_susp_quotas(osb, 1);
721 else
722 ret = ocfs2_enable_quotas(osb);
723 if (ret < 0) {
724 /* Return back changes... */
725 spin_lock(&osb->osb_lock);
1751e8a6 726 sb->s_flags |= SB_RDONLY;
19ece546
JK		 727 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
728 spin_unlock(&osb->osb_lock);
729 goto out;
730 }
731 }
ccd979bd
MF		 732 }
733
734 if (!ret) {
ccd979bd
MF		 735 /* Only save off the new mount options in case of a successful
736 * remount. */
c0123ade
TY		 737 osb->s_mount_opt = parsed_options.mount_opt;
738 osb->s_atime_quantum = parsed_options.atime_quantum;
739 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 740 if (parsed_options.commit_interval)
741 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 742
743 if (!ocfs2_is_hard_readonly(osb))
744 ocfs2_set_journal_params(osb);
57b09bb5 745
1751e8a6 746 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
57b09bb5 747 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
1751e8a6 748 SB_POSIXACL : 0);
ccd979bd
MF		 749 }
750out:
751 return ret;
752}
753
754static int ocfs2_sb_probe(struct super_block *sb,
755 struct buffer_head **bh,
73be192b
JB		 756 int *sector_size,
757 struct ocfs2_blockcheck_stats *stats)
ccd979bd 758{
bddb8eb3 759 int status, tmpstat;
ccd979bd
MF		 760 struct ocfs1_vol_disk_hdr *hdr;
761 struct ocfs2_dinode *di;
762 int blksize;
763
764 *bh = NULL;
765
766 /* may be > 512 */
e1defc4f 767 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 768 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
769 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
770 *sector_size, OCFS2_MAX_BLOCKSIZE);
771 status = -EINVAL;
772 goto bail;
773 }
774
775 /* Can this really happen? */
776 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
777 *sector_size = OCFS2_MIN_BLOCKSIZE;
778
779 /* check block zero for old format */
780 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
781 if (status < 0) {
782 mlog_errno(status);
783 goto bail;
784 }
785 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
786 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
787 mlog(ML_ERROR, "incompatible version: %u.%u\n",
788 hdr->major_version, hdr->minor_version);
789 status = -EINVAL;
790 }
791 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
792 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
793 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
794 hdr->signature);
795 status = -EINVAL;
796 }
797 brelse(*bh);
798 *bh = NULL;
799 if (status < 0) {
800 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
801 "upgraded before mounting with ocfs v2\n");
802 goto bail;
803 }
804
805 /*
806 * Now check at magic offset for 512, 1024, 2048, 4096
807 * blocksizes. 4096 is the maximum blocksize because it is
808 * the minimum clustersize.
809 */
810 status = -EINVAL;
811 for (blksize = *sector_size;
812 blksize <= OCFS2_MAX_BLOCKSIZE;
813 blksize <<= 1) {
814 tmpstat = ocfs2_get_sector(sb, bh,
815 OCFS2_SUPER_BLOCK_BLKNO,
816 blksize);
817 if (tmpstat < 0) {
818 status = tmpstat;
819 mlog_errno(status);
fb5cbe9e 820 break;
ccd979bd
MF		 821 }
822 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 823 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 824 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB		 825 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
826 if (tmpstat < 0) {
827 brelse(*bh);
828 *bh = NULL;
829 }
830 if (tmpstat != -EAGAIN) {
831 status = tmpstat;
ccd979bd 832 break;
fb5cbe9e 833 }
ccd979bd
MF		 834 }
835
836bail:
837 return status;
838}
839
c271c5c2
SM		 840static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
841{
2c442719
SM		 842 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
843
844 if (osb->s_mount_opt & hb_enabled) {
845 if (ocfs2_mount_local(osb)) {
c271c5c2
SM		 846 mlog(ML_ERROR, "Cannot heartbeat on a locally "
847 "mounted device.\n");
848 return -EINVAL;
849 }
2c442719 850 if (ocfs2_userspace_stack(osb)) {
b61817e1
JB		 851 mlog(ML_ERROR, "Userspace stack expected, but "
852 "o2cb heartbeat arguments passed to mount\n");
853 return -EINVAL;
854 }
2c442719
SM		 855 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
856 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
857 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
858 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
859 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
860 return -EINVAL;
861 }
b61817e1
JB		 862 }
863
2c442719 864 if (!(osb->s_mount_opt & hb_enabled)) {
b61817e1
JB		 865 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
866 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 867 mlog(ML_ERROR, "Heartbeat has to be started to mount "
868 "a read-write clustered device.\n");
869 return -EINVAL;
870 }
871 }
872
873 return 0;
874}
875
b61817e1
JB		 876/*
877 * If we're using a userspace stack, mount should have passed
878 * a name that matches the disk. If not, mount should not
879 * have passed a stack.
880 */
881static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
882 struct mount_options *mopt)
883{
884 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
885 mlog(ML_ERROR,
886 "cluster stack passed to mount, but this filesystem "
887 "does not support it\n");
888 return -EINVAL;
889 }
890
891 if (ocfs2_userspace_stack(osb) &&
892 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
893 OCFS2_STACK_LABEL_LEN)) {
894 mlog(ML_ERROR,
895 "cluster stack passed to mount (\"%s\") does not "
896 "match the filesystem (\"%s\")\n",
897 mopt->cluster_stack,
898 osb->osb_cluster_stack);
899 return -EINVAL;
900 }
901
902 return 0;
903}
904
19ece546
JK		 905static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
906{
907 int type;
908 struct super_block *sb = osb->sb;
52362810
JK		 909 unsigned int feature[OCFS2_MAXQUOTAS] = {
910 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
911 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK		 912 int status = 0;
913
52362810 914 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 915 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
916 continue;
917 if (unsuspend)
0f0dd62f 918 status = dquot_resume(sb, type);
eea7feb0
JK		 919 else {
920 struct ocfs2_mem_dqinfo *oinfo;
921
922 /* Cancel periodic syncing before suspending */
923 oinfo = sb_dqinfo(sb, type)->dqi_priv;
924 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
0f0dd62f 925 status = dquot_suspend(sb, type);
eea7feb0 926 }
19ece546
JK		 927 if (status < 0)
928 break;
929 }
930 if (status < 0)
931 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
932 "remount (error = %d).\n", status);
933 return status;
934}
935
936static int ocfs2_enable_quotas(struct ocfs2_super *osb)
937{
52362810 938 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
19ece546 939 struct super_block *sb = osb->sb;
52362810
JK		 940 unsigned int feature[OCFS2_MAXQUOTAS] = {
941 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
942 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
943 unsigned int ino[OCFS2_MAXQUOTAS] = {
944 LOCAL_USER_QUOTA_SYSTEM_INODE,
19ece546
JK		 945 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
946 int status;
947 int type;
948
949 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
52362810 950 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 951 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
952 continue;
953 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
954 osb->slot_num);
955 if (!inode[type]) {
956 status = -ENOENT;
957 goto out_quota_off;
958 }
287a8095
CH		 959 status = dquot_enable(inode[type], type, QFMT_OCFS2,
960 DQUOT_USAGE_ENABLED);
19ece546
JK		 961 if (status < 0)
962 goto out_quota_off;
963 }
964
52362810 965 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 966 iput(inode[type]);
967 return 0;
968out_quota_off:
969 ocfs2_disable_quotas(osb);
52362810 970 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 971 iput(inode[type]);
972 mlog_errno(status);
973 return status;
974}
975
976static void ocfs2_disable_quotas(struct ocfs2_super *osb)
977{
978 int type;
979 struct inode *inode;
980 struct super_block *sb = osb->sb;
c06bcbfa 981 struct ocfs2_mem_dqinfo *oinfo;
19ece546
JK		 982
983 /* We mostly ignore errors in this function because there's not much
984 * we can do when we see them */
52362810 985 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 986 if (!sb_has_quota_loaded(sb, type))
987 continue;
c06bcbfa
JK		 988 oinfo = sb_dqinfo(sb, type)->dqi_priv;
989 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK		 990 inode = igrab(sb->s_dquot.files[type]);
991 /* Turn off quotas. This will remove all dquot structures from
992 * memory and so they will be automatically synced to global
993 * quota files */
0f0dd62f
CH		 994 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
995 DQUOT_LIMITS_ENABLED);
19ece546
JK		 996 if (!inode)
997 continue;
998 iput(inode);
999 }
1000}
1001
ccd979bd
MF		 1002static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1003{
1004 struct dentry *root;
1005 int status, sector_size;
c0123ade 1006 struct mount_options parsed_options;
ccd979bd
MF		 1007 struct inode *inode = NULL;
1008 struct ocfs2_super *osb = NULL;
1009 struct buffer_head *bh = NULL;
49fa8140 1010 char nodestr[12];
73be192b 1011 struct ocfs2_blockcheck_stats stats;
ccd979bd 1012
32a42d39 1013 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1014
c0123ade 1015 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1016 status = -EINVAL;
1017 goto read_super_error;
1018 }
1019
1020 /* probe for superblock */
73be192b 1021 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1022 if (status < 0) {
1023 mlog(ML_ERROR, "superblock probe failed!\n");
1024 goto read_super_error;
1025 }
1026
73be192b 1027 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1028 osb = OCFS2_SB(sb);
1029 if (status < 0) {
1030 mlog_errno(status);
1031 goto read_super_error;
1032 }
1033 brelse(bh);
1034 bh = NULL;
a68979b8 1035
5297aad8
JK		 1036 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1037 status = -EINVAL;
1038 goto read_super_error;
1039 }
c0123ade
TY		 1040 osb->s_mount_opt = parsed_options.mount_opt;
1041 osb->s_atime_quantum = parsed_options.atime_quantum;
1042 osb->preferred_slot = parsed_options.slot;
d147b3d6 1043 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1044
1045 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1046 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF		 1047 osb->osb_dir_resv_level = parsed_options.resv_level;
1048 if (parsed_options.dir_resv_level == -1)
1049 osb->osb_dir_resv_level = parsed_options.resv_level;
1050 else
1051 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1052
b61817e1
JB		 1053 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1054 if (status)
1055 goto read_super_error;
1056
ccd979bd
MF		 1057 sb->s_magic = OCFS2_SUPER_MAGIC;
1058
1751e8a6
LT		 1059 sb->s_flags = (sb->s_flags & ~(SB_POSIXACL | SB_NOSEC)) |
1060 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? SB_POSIXACL : 0);
a68979b8 1061
1751e8a6 1062 /* Hard readonly mode only if: bdev_read_only, SB_RDONLY,
ccd979bd
MF		 1063 * heartbeat=none */
1064 if (bdev_read_only(sb->s_bdev)) {
bc98a42c 1065 if (!sb_rdonly(sb)) {
ccd979bd
MF		 1066 status = -EACCES;
1067 mlog(ML_ERROR, "Readonly device detected but readonly "
1068 "mount was not specified.\n");
1069 goto read_super_error;
1070 }
1071
1072 /* You should not be able to start a local heartbeat
1073 * on a readonly device. */
1074 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1075 status = -EROFS;
1076 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1077 "device.\n");
1078 goto read_super_error;
1079 }
1080
1081 status = ocfs2_check_journals_nolocks(osb);
1082 if (status < 0) {
1083 if (status == -EROFS)
1084 mlog(ML_ERROR, "Recovery required on readonly "
1085 "file system, but write access is "
1086 "unavailable.\n");
1087 else
2bd63216 1088 mlog_errno(status);
ccd979bd
MF		 1089 goto read_super_error;
1090 }
1091
1092 ocfs2_set_ro_flag(osb, 1);
1093
619c200d
SM		 1094 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1095 "Cluster services will not be used for this mount. "
1096 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF		 1097 }
1098
1099 if (!ocfs2_is_hard_readonly(osb)) {
bc98a42c 1100 if (sb_rdonly(sb))
ccd979bd
MF		 1101 ocfs2_set_ro_flag(osb, 0);
1102 }
1103
c271c5c2
SM		 1104 status = ocfs2_verify_heartbeat(osb);
1105 if (status < 0) {
1106 mlog_errno(status);
1107 goto read_super_error;
1108 }
1109
ccd979bd
MF		 1110 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1111 ocfs2_debugfs_root);
8f443e23 1112 if (!osb->osb_debug_root) {
ccd979bd
MF		 1113 status = -EINVAL;
1114 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1115 goto read_super_error;
1116 }
1117
50397507
SM		 1118 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1119 osb->osb_debug_root,
1120 osb,
1121 &ocfs2_osb_debug_fops);
8f443e23 1122 if (!osb->osb_ctxt) {
50397507
SM		 1123 status = -EINVAL;
1124 mlog_errno(status);
1125 goto read_super_error;
1126 }
1127
73be192b
JB		 1128 if (ocfs2_meta_ecc(osb)) {
1129 status = ocfs2_blockcheck_stats_debugfs_install(
1130 &osb->osb_ecc_stats,
1131 osb->osb_debug_root);
1132 if (status) {
1133 mlog(ML_ERROR,
1134 "Unable to create blockcheck statistics "
1135 "files\n");
1136 goto read_super_error;
1137 }
1138 }
1139
ccd979bd 1140 status = ocfs2_mount_volume(sb);
ccd979bd
MF		 1141 if (status < 0)
1142 goto read_super_error;
1143
be0d93f0
AV		 1144 if (osb->root_inode)
1145 inode = igrab(osb->root_inode);
1146
ccd979bd
MF		 1147 if (!inode) {
1148 status = -EIO;
1149 mlog_errno(status);
1150 goto read_super_error;
1151 }
1152
48fde701 1153 root = d_make_root(inode);
ccd979bd
MF		 1154 if (!root) {
1155 status = -ENOMEM;
1156 mlog_errno(status);
1157 goto read_super_error;
1158 }
1159
1160 sb->s_root = root;
1161
1162 ocfs2_complete_mount_recovery(osb);
1163
5f483c4a
GH		 1164 osb->osb_dev_kset = kset_create_and_add(sb->s_id, NULL,
1165 &ocfs2_kset->kobj);
1166 if (!osb->osb_dev_kset) {
1167 status = -ENOMEM;
1168 mlog(ML_ERROR, "Unable to create device kset %s.\n", sb->s_id);
1169 goto read_super_error;
1170 }
1171
1172 /* Create filecheck sysfs related directories/files at
1173 * /sys/fs/ocfs2/<devname>/filecheck */
1174 if (ocfs2_filecheck_create_sysfs(osb)) {
1175 status = -ENOMEM;
1176 mlog(ML_ERROR, "Unable to create filecheck sysfs directory at "
1177 "/sys/fs/ocfs2/%s/filecheck.\n", sb->s_id);
1178 goto read_super_error;
1179 }
1180
c271c5c2
SM		 1181 if (ocfs2_mount_local(osb))
1182 snprintf(nodestr, sizeof(nodestr), "local");
1183 else
19fdb624 1184 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1185
1186 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1187 "with %s data mode.\n",
c271c5c2 1188 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1189 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1190 "ordered");
1191
1192 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1193 wake_up(&osb->osb_mount_event);
1194
19ece546
JK		 1195 /* Now we can initialize quotas because we can afford to wait
1196 * for cluster locks recovery now. That also means that truncation
1197 * log recovery can happen but that waits for proper quota setup */
bc98a42c 1198 if (!sb_rdonly(sb)) {
19ece546
JK		 1199 status = ocfs2_enable_quotas(osb);
1200 if (status < 0) {
1201 /* We have to err-out specially here because
1202 * s_root is already set */
1203 mlog_errno(status);
1204 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1205 wake_up(&osb->osb_mount_event);
19ece546
JK		 1206 return status;
1207 }
1208 }
1209
1210 ocfs2_complete_quota_recovery(osb);
1211
1212 /* Now we wake up again for processes waiting for quotas */
1213 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1214 wake_up(&osb->osb_mount_event);
1215
df152c24 1216 /* Start this when the mount is almost sure of being successful */
8b712cd5 1217 ocfs2_orphan_scan_start(osb);
df152c24 1218
ccd979bd
MF		 1219 return status;
1220
1221read_super_error:
a81cb88b 1222 brelse(bh);
ccd979bd 1223
a52370b3
GH		 1224 if (status)
1225 mlog_errno(status);
1226
ccd979bd
MF		 1227 if (osb) {
1228 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1229 wake_up(&osb->osb_mount_event);
1230 ocfs2_dismount_volume(sb, 1);
1231 }
1232
ccd979bd
MF		 1233 return status;
1234}
1235
152a0836 1236static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH		 1237 int flags,
1238 const char *dev_name,
152a0836 1239 void *data)
ccd979bd 1240{
152a0836 1241 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF		 1242}
1243
1244static struct file_system_type ocfs2_fs_type = {
1245 .owner = THIS_MODULE,
1246 .name = "ocfs2",
152a0836 1247 .mount = ocfs2_mount,
8ed6b237 1248 .kill_sb = kill_block_super,
1ba9da2f 1249 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1250 .next = NULL
1251};
91417705 1252MODULE_ALIAS_FS("ocfs2");
ccd979bd 1253
5297aad8
JK		 1254static int ocfs2_check_set_options(struct super_block *sb,
1255 struct mount_options *options)
1256{
1257 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1258 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1259 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1260 mlog(ML_ERROR, "User quotas were requested, but this "
1261 "filesystem does not have the feature enabled.\n");
1262 return 0;
1263 }
1264 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1265 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1266 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1267 mlog(ML_ERROR, "Group quotas were requested, but this "
1268 "filesystem does not have the feature enabled.\n");
1269 return 0;
1270 }
1271 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1272 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1273 mlog(ML_ERROR, "ACL support requested but extended attributes "
1274 "feature is not enabled\n");
1275 return 0;
1276 }
1277 /* No ACL setting specified? Use XATTR feature... */
1278 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1279 OCFS2_MOUNT_NO_POSIX_ACL))) {
1280 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1281 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1282 else
1283 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1284 }
1285 return 1;
1286}
1287
ccd979bd
MF		 1288static int ocfs2_parse_options(struct super_block *sb,
1289 char *options,
c0123ade 1290 struct mount_options *mopt,
ccd979bd
MF		 1291 int is_remount)
1292{
52c303c5 1293 int status, user_stack = 0;
ccd979bd 1294 char *p;
2c442719 1295 u32 tmp;
3eb5bdf0
NZ		 1296 int token, option;
1297 substring_t args[MAX_OPT_ARGS];
ccd979bd 1298
32a42d39 1299 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1300
d147b3d6 1301 mopt->commit_interval = 0;
4b37fcb7 1302 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY		 1303 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1304 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1305 mopt->localalloc_opt = -1;
b61817e1 1306 mopt->cluster_stack[0] = '\0';
d02f00cc 1307 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1308 mopt->dir_resv_level = -1;
ccd979bd
MF		 1309
1310 if (!options) {
1311 status = 1;
1312 goto bail;
1313 }
1314
1315 while ((p = strsep(&options, ",")) != NULL) {
ccd979bd
MF		 1316 if (!*p)
1317 continue;
1318
1319 token = match_token(p, tokens, args);
1320 switch (token) {
1321 case Opt_hb_local:
c0123ade 1322 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1323 break;
1324 case Opt_hb_none:
2c442719
SM		 1325 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1326 break;
1327 case Opt_hb_global:
1328 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF		 1329 break;
1330 case Opt_barrier:
1331 if (match_int(&args[0], &option)) {
1332 status = 0;
1333 goto bail;
1334 }
1335 if (option)
c0123ade 1336 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1337 else
c0123ade 1338 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1339 break;
1340 case Opt_intr:
c0123ade 1341 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1342 break;
1343 case Opt_nointr:
c0123ade 1344 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1345 break;
1346 case Opt_err_panic:
7d0fb914
GR		 1347 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1348 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
c0123ade 1349 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1350 break;
1351 case Opt_err_ro:
7d0fb914 1352 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
c0123ade 1353 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
7d0fb914
GR		 1354 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1355 break;
1356 case Opt_err_cont:
1357 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1358 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1359 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
ccd979bd
MF		 1360 break;
1361 case Opt_data_ordered:
c0123ade 1362 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1363 break;
1364 case Opt_data_writeback:
c0123ade 1365 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1366 break;
cf1d6c76
TY		 1367 case Opt_user_xattr:
1368 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1369 break;
1370 case Opt_nouser_xattr:
1371 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1372 break;
7f1a37e3
TY		 1373 case Opt_atime_quantum:
1374 if (match_int(&args[0], &option)) {
1375 status = 0;
1376 goto bail;
1377 }
1378 if (option >= 0)
c0123ade 1379 mopt->atime_quantum = option;
7f1a37e3 1380 break;
baf4661a 1381 case Opt_slot:
baf4661a
SM		 1382 if (match_int(&args[0], &option)) {
1383 status = 0;
1384 goto bail;
1385 }
1386 if (option)
c0123ade 1387 mopt->slot = (s16)option;
baf4661a 1388 break;
d147b3d6 1389 case Opt_commit:
d147b3d6
MF		 1390 if (match_int(&args[0], &option)) {
1391 status = 0;
1392 goto bail;
1393 }
1394 if (option < 0)
1395 return 0;
1396 if (option == 0)
2b4e30fb 1397 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1398 mopt->commit_interval = HZ * option;
1399 break;
2fbe8d1e 1400 case Opt_localalloc:
2fbe8d1e
SM		 1401 if (match_int(&args[0], &option)) {
1402 status = 0;
1403 goto bail;
1404 }
73c8a800 1405 if (option >= 0)
2fbe8d1e
SM		 1406 mopt->localalloc_opt = option;
1407 break;
53fc622b
MF		 1408 case Opt_localflocks:
1409 /*
1410 * Changing this during remount could race
1411 * flock() requests, or "unbalance" existing
1412 * ones (e.g., a lock is taken in one mode but
1413 * dropped in the other). If users care enough
1414 * to flip locking modes during remount, we
1415 * could add a "local" flag to individual
1416 * flock structures for proper tracking of
1417 * state.
1418 */
1419 if (!is_remount)
1420 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1421 break;
b61817e1
JB		 1422 case Opt_stack:
1423 /* Check both that the option we were passed
1424 * is of the right length and that it is a proper
1425 * string of the right length.
1426 */
1427 if (((args[0].to - args[0].from) !=
1428 OCFS2_STACK_LABEL_LEN) ||
1429 (strnlen(args[0].from,
1430 OCFS2_STACK_LABEL_LEN) !=
1431 OCFS2_STACK_LABEL_LEN)) {
1432 mlog(ML_ERROR,
1433 "Invalid cluster_stack option\n");
1434 status = 0;
1435 goto bail;
1436 }
1437 memcpy(mopt->cluster_stack, args[0].from,
1438 OCFS2_STACK_LABEL_LEN);
1439 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF		 1440 /*
1441 * Open code the memcmp here as we don't have
1442 * an osb to pass to
1443 * ocfs2_userspace_stack().
1444 */
1445 if (memcmp(mopt->cluster_stack,
1446 OCFS2_CLASSIC_CLUSTER_STACK,
1447 OCFS2_STACK_LABEL_LEN))
1448 user_stack = 1;
b61817e1 1449 break;
12462f1d
JB		 1450 case Opt_inode64:
1451 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1452 break;
19ece546 1453 case Opt_usrquota:
19ece546
JK		 1454 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1455 break;
1456 case Opt_grpquota:
19ece546
JK		 1457 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1458 break;
7bdb0d18
TY		 1459 case Opt_coherency_buffered:
1460 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1461 break;
1462 case Opt_coherency_full:
1463 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1464 break;
a68979b8
TY		 1465 case Opt_acl:
1466 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1467 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1468 break;
1469 case Opt_noacl:
5297aad8 1470 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1471 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1472 break;
d02f00cc
MF		 1473 case Opt_resv_level:
1474 if (is_remount)
1475 break;
1476 if (match_int(&args[0], &option)) {
1477 status = 0;
1478 goto bail;
1479 }
1480 if (option >= OCFS2_MIN_RESV_LEVEL &&
1481 option < OCFS2_MAX_RESV_LEVEL)
1482 mopt->resv_level = option;
1483 break;
83f92318
MF		 1484 case Opt_dir_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->dir_resv_level = option;
1494 break;
1dfeb768 1495 case Opt_journal_async_commit:
1496 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1497 break;
ccd979bd
MF		 1498 default:
1499 mlog(ML_ERROR,
1500 "Unrecognized mount option \"%s\" "
1501 "or missing value\n", p);
1502 status = 0;
1503 goto bail;
1504 }
1505 }
1506
52c303c5
MF		 1507 if (user_stack == 0) {
1508 /* Ensure only one heartbeat mode */
1509 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1510 OCFS2_MOUNT_HB_GLOBAL |
1511 OCFS2_MOUNT_HB_NONE);
1512 if (hweight32(tmp) != 1) {
1513 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1514 status = 0;
1515 goto bail;
1516 }
2c442719
SM		 1517 }
1518
ccd979bd
MF		 1519 status = 1;
1520
1521bail:
ccd979bd
MF		 1522 return status;
1523}
1524
34c80b1d 1525static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1526{
34c80b1d 1527 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1528 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1529 unsigned int local_alloc_megs;
d550071c 1530
2c442719
SM		 1531 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1532 seq_printf(s, ",_netdev");
1533 if (opts & OCFS2_MOUNT_HB_LOCAL)
1534 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1535 else
1536 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1537 } else
1538 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM		 1539
1540 if (opts & OCFS2_MOUNT_NOINTR)
1541 seq_printf(s, ",nointr");
1542
1543 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1544 seq_printf(s, ",data=writeback");
1545 else
1546 seq_printf(s, ",data=ordered");
1547
1548 if (opts & OCFS2_MOUNT_BARRIER)
1549 seq_printf(s, ",barrier=1");
1550
1551 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1552 seq_printf(s, ",errors=panic");
7d0fb914
GR		 1553 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1554 seq_printf(s, ",errors=continue");
d550071c
SM		 1555 else
1556 seq_printf(s, ",errors=remount-ro");
1557
1558 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1559 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1560
34c80b1d 1561 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1562
d147b3d6
MF		 1563 if (osb->osb_commit_interval)
1564 seq_printf(s, ",commit=%u",
1565 (unsigned) (osb->osb_commit_interval / HZ));
1566
ebcee4b5 1567 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1568 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1569 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1570
53fc622b
MF		 1571 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1572 seq_printf(s, ",localflocks,");
1573
b61817e1 1574 if (osb->osb_cluster_stack[0])
a068acf2
KC		 1575 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1576 OCFS2_STACK_LABEL_LEN);
19ece546
JK		 1577 if (opts & OCFS2_MOUNT_USRQUOTA)
1578 seq_printf(s, ",usrquota");
1579 if (opts & OCFS2_MOUNT_GRPQUOTA)
1580 seq_printf(s, ",grpquota");
b61817e1 1581
7bdb0d18
TY		 1582 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1583 seq_printf(s, ",coherency=buffered");
1584 else
1585 seq_printf(s, ",coherency=full");
1586
b0f73cfc
SM		 1587 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1588 seq_printf(s, ",nouser_xattr");
1589 else
1590 seq_printf(s, ",user_xattr");
1591
12462f1d
JB		 1592 if (opts & OCFS2_MOUNT_INODE64)
1593 seq_printf(s, ",inode64");
1594
a68979b8
TY		 1595 if (opts & OCFS2_MOUNT_POSIX_ACL)
1596 seq_printf(s, ",acl");
1597 else
1598 seq_printf(s, ",noacl");
a68979b8 1599
d02f00cc
MF		 1600 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1601 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1602
83f92318
MF		 1603 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1604 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1605
1dfeb768 1606 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1607 seq_printf(s, ",journal_async_commit");
1608
d550071c
SM		 1609 return 0;
1610}
1611
ccd979bd
MF		 1612static int __init ocfs2_init(void)
1613{
c18ceab0 1614 int status;
a11f7e63 1615
ccd979bd 1616 status = init_ocfs2_uptodate_cache();
342827d7
AV		 1617 if (status < 0)
1618 goto out1;
ccd979bd
MF		 1619
1620 status = ocfs2_initialize_mem_caches();
342827d7
AV		 1621 if (status < 0)
1622 goto out2;
ccd979bd 1623
ccd979bd 1624 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
8f443e23
LT		 1625 if (!ocfs2_debugfs_root) {
1626 status = -ENOMEM;
ccd979bd 1627 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
35ddf78e 1628 goto out3;
ccd979bd
MF		 1629 }
1630
63e0c48a
JB		 1631 ocfs2_set_locking_protocol();
1632
9e33d69f 1633 status = register_quota_format(&ocfs2_quota_format);
342827d7 1634 if (status < 0)
35ddf78e 1635 goto out3;
342827d7
AV		 1636 status = register_filesystem(&ocfs2_fs_type);
1637 if (!status)
1638 return 0;
ccd979bd 1639
342827d7 1640 unregister_quota_format(&ocfs2_quota_format);
342827d7 1641out3:
35ddf78e 1642 debugfs_remove(ocfs2_debugfs_root);
342827d7
AV		 1643 ocfs2_free_mem_caches();
1644out2:
1645 exit_ocfs2_uptodate_cache();
1646out1:
1647 mlog_errno(status);
1648 return status;
ccd979bd
MF		 1649}
1650
1651static void __exit ocfs2_exit(void)
1652{
9e33d69f
JK		 1653 unregister_quota_format(&ocfs2_quota_format);
1654
ccd979bd
MF		 1655 debugfs_remove(ocfs2_debugfs_root);
1656
1657 ocfs2_free_mem_caches();
1658
1659 unregister_filesystem(&ocfs2_fs_type);
1660
ccd979bd 1661 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1662}
1663
1664static void ocfs2_put_super(struct super_block *sb)
1665{
32a42d39 1666 trace_ocfs2_put_super(sb);
ccd979bd
MF		 1667
1668 ocfs2_sync_blockdev(sb);
1669 ocfs2_dismount_volume(sb, 0);
ccd979bd
MF		 1670}
1671
726c3342 1672static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1673{
1674 struct ocfs2_super *osb;
1675 u32 numbits, freebits;
1676 int status;
1677 struct ocfs2_dinode *bm_lock;
1678 struct buffer_head *bh = NULL;
1679 struct inode *inode = NULL;
1680
32a42d39 1681 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1682
726c3342 1683 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1684
1685 inode = ocfs2_get_system_file_inode(osb,
1686 GLOBAL_BITMAP_SYSTEM_INODE,
1687 OCFS2_INVALID_SLOT);
1688 if (!inode) {
1689 mlog(ML_ERROR, "failed to get bitmap inode\n");
1690 status = -EIO;
1691 goto bail;
1692 }
1693
e63aecb6 1694 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1695 if (status < 0) {
1696 mlog_errno(status);
1697 goto bail;
1698 }
1699
1700 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1701
1702 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1703 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1704
1705 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1706 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1707 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1708 buf->f_blocks = ((sector_t) numbits) *
1709 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1710 buf->f_bfree = ((sector_t) freebits) *
1711 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1712 buf->f_bavail = buf->f_bfree;
1713 buf->f_files = numbits;
1714 buf->f_ffree = freebits;
837711f8
CL		 1715 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1716 & 0xFFFFFFFFUL;
1717 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1718 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1719
1720 brelse(bh);
1721
e63aecb6 1722 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1723 status = 0;
1724bail:
72865d92 1725 iput(inode);
ccd979bd 1726
c1e8d35e
TM		 1727 if (status)
1728 mlog_errno(status);
ccd979bd
MF		 1729
1730 return status;
1731}
1732
51cc5068 1733static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1734{
1735 struct ocfs2_inode_info *oi = data;
1736
a35afb83
CL		 1737 oi->ip_flags = 0;
1738 oi->ip_open_count = 0;
1739 spin_lock_init(&oi->ip_lock);
1740 ocfs2_extent_map_init(&oi->vfs_inode);
1741 INIT_LIST_HEAD(&oi->ip_io_markers);
4506cfb6 1742 INIT_LIST_HEAD(&oi->ip_unwritten_list);
a35afb83 1743 oi->ip_dir_start_lookup = 0;
a35afb83 1744 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1745 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1746 mutex_init(&oi->ip_io_mutex);
ccd979bd 1747
a35afb83
CL		 1748 oi->ip_blkno = 0ULL;
1749 oi->ip_clusters = 0;
ccd979bd 1750
4fe370af
MF		 1751 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1752
a35afb83 1753 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1754 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1755 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1756
8cb471e8 1757 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1758 &ocfs2_inode_caching_ops);
ccd979bd 1759
a35afb83 1760 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1761}
1762
1763static int ocfs2_initialize_mem_caches(void)
1764{
1765 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1766 sizeof(struct ocfs2_inode_info),
1767 0,
1768 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
5d097056 1769 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1770 ocfs2_inode_init_once);
9e33d69f
JK		 1771 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1772 sizeof(struct ocfs2_dquot),
1773 0,
1774 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1775 SLAB_MEM_SPREAD),
1776 NULL);
1777 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1778 sizeof(struct ocfs2_quota_chunk),
1779 0,
1780 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1781 NULL);
1782 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1783 !ocfs2_qf_chunk_cachep) {
a17b485a 1784 kmem_cache_destroy(ocfs2_inode_cachep);
1785 kmem_cache_destroy(ocfs2_dquot_cachep);
1786 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1787 return -ENOMEM;
9e33d69f 1788 }
ccd979bd 1789
ccd979bd
MF		 1790 return 0;
1791}
1792
1793static void ocfs2_free_mem_caches(void)
1794{
8c0a8537
KS		 1795 /*
1796 * Make sure all delayed rcu free inodes are flushed before we
1797 * destroy cache.
1798 */
1799 rcu_barrier();
a17b485a 1800 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1801 ocfs2_inode_cachep = NULL;
9e33d69f 1802
a17b485a 1803 kmem_cache_destroy(ocfs2_dquot_cachep);
9e33d69f
JK		 1804 ocfs2_dquot_cachep = NULL;
1805
a17b485a 1806 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
9e33d69f 1807 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1808}
1809
1810static int ocfs2_get_sector(struct super_block *sb,
1811 struct buffer_head **bh,
1812 int block,
1813 int sect_size)
1814{
1815 if (!sb_set_blocksize(sb, sect_size)) {
1816 mlog(ML_ERROR, "unable to set blocksize\n");
1817 return -EIO;
1818 }
1819
1820 *bh = sb_getblk(sb, block);
1821 if (!*bh) {
7391a294
RX		 1822 mlog_errno(-ENOMEM);
1823 return -ENOMEM;
ccd979bd
MF		 1824 }
1825 lock_buffer(*bh);
1826 if (!buffer_dirty(*bh))
1827 clear_buffer_uptodate(*bh);
1828 unlock_buffer(*bh);
dfec8a14 1829 ll_rw_block(REQ_OP_READ, 0, 1, bh);
ccd979bd 1830 wait_on_buffer(*bh);
28d57d43 1831 if (!buffer_uptodate(*bh)) {
1832 mlog_errno(-EIO);
1833 brelse(*bh);
1834 *bh = NULL;
1835 return -EIO;
1836 }
1837
ccd979bd
MF		 1838 return 0;
1839}
1840
ccd979bd
MF		 1841static int ocfs2_mount_volume(struct super_block *sb)
1842{
1843 int status = 0;
1844 int unlock_super = 0;
1845 struct ocfs2_super *osb = OCFS2_SB(sb);
1846
ccd979bd
MF		 1847 if (ocfs2_is_hard_readonly(osb))
1848 goto leave;
1849
ccd979bd
MF		 1850 status = ocfs2_dlm_init(osb);
1851 if (status < 0) {
1852 mlog_errno(status);
a52370b3
GH		 1853 if (status == -EBADR && ocfs2_userspace_stack(osb))
1854 mlog(ML_ERROR, "couldn't mount because cluster name on"
1855 " disk does not match the running cluster name.\n");
ccd979bd
MF		 1856 goto leave;
1857 }
1858
ccd979bd
MF		 1859 status = ocfs2_super_lock(osb, 1);
1860 if (status < 0) {
1861 mlog_errno(status);
1862 goto leave;
1863 }
1864 unlock_super = 1;
1865
1866 /* This will load up the node map and add ourselves to it. */
1867 status = ocfs2_find_slot(osb);
1868 if (status < 0) {
1869 mlog_errno(status);
1870 goto leave;
1871 }
1872
ccd979bd
MF		 1873 /* load all node-local system inodes */
1874 status = ocfs2_init_local_system_inodes(osb);
1875 if (status < 0) {
1876 mlog_errno(status);
1877 goto leave;
1878 }
1879
1880 status = ocfs2_check_volume(osb);
1881 if (status < 0) {
1882 mlog_errno(status);
1883 goto leave;
1884 }
1885
1886 status = ocfs2_truncate_log_init(osb);
06c59bb8 1887 if (status < 0)
ccd979bd 1888 mlog_errno(status);
c271c5c2 1889
ccd979bd
MF		 1890leave:
1891 if (unlock_super)
1892 ocfs2_super_unlock(osb, 1);
1893
ccd979bd
MF		 1894 return status;
1895}
1896
ccd979bd
MF		 1897static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1898{
286eaa95 1899 int tmp, hangup_needed = 0;
ccd979bd 1900 struct ocfs2_super *osb = NULL;
99d7a882 1901 char nodestr[12];
ccd979bd 1902
32a42d39 1903 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF		 1904
1905 BUG_ON(!sb);
1906 osb = OCFS2_SB(sb);
1907 BUG_ON(!osb);
1908
5f483c4a
GH		 1909 /* Remove file check sysfs related directores/files,
1910 * and wait for the pending file check operations */
1911 ocfs2_filecheck_remove_sysfs(osb);
1912
1913 kset_unregister(osb->osb_dev_kset);
1914
50397507
SM		 1915 debugfs_remove(osb->osb_ctxt);
1916
692684e1
SM		 1917 /* Orphan scan should be stopped as early as possible */
1918 ocfs2_orphan_scan_stop(osb);
1919
19ece546
JK		 1920 ocfs2_disable_quotas(osb);
1921
e3a767b6
JK		 1922 /* All dquots should be freed by now */
1923 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1924 /* Wait for worker to be done with the work structure in osb */
1925 cancel_work_sync(&osb->dquot_drop_work);
1926
ccd979bd
MF		 1927 ocfs2_shutdown_local_alloc(osb);
1928
b253bfd8
X		 1929 ocfs2_truncate_log_shutdown(osb);
1930
553abd04
JB		 1931 /* This will disable recovery and flush any recovery work. */
1932 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1933
1934 ocfs2_journal_shutdown(osb);
1935
1936 ocfs2_sync_blockdev(sb);
1937
374a263e
TM		 1938 ocfs2_purge_refcount_trees(osb);
1939
4670c46d
JB		 1940 /* No cluster connection means we've failed during mount, so skip
1941 * all the steps which depended on that to complete. */
1942 if (osb->cconn) {
ccd979bd
MF		 1943 tmp = ocfs2_super_lock(osb, 1);
1944 if (tmp < 0) {
1945 mlog_errno(tmp);
1946 return;
1947 }
19b613d4 1948 }
ccd979bd 1949
19b613d4
MF		 1950 if (osb->slot_num != OCFS2_INVALID_SLOT)
1951 ocfs2_put_slot(osb);
ccd979bd 1952
4670c46d 1953 if (osb->cconn)
ccd979bd 1954 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1955
1956 ocfs2_release_system_inodes(osb);
1957
6953b4c0 1958 /*
6953b4c0
JB		 1959 * If we're dismounting due to mount error, mount.ocfs2 will clean
1960 * up heartbeat. If we're a local mount, there is no heartbeat.
1961 * If we failed before we got a uuid_str yet, we can't stop
1962 * heartbeat. Otherwise, do it.
1963 */
03efed8a
TY		 1964 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1965 !ocfs2_is_hard_readonly(osb))
286eaa95
JB		 1966 hangup_needed = 1;
1967
1968 if (osb->cconn)
1969 ocfs2_dlm_shutdown(osb, hangup_needed);
1970
73be192b 1971 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1972 debugfs_remove(osb->osb_debug_root);
1973
1974 if (hangup_needed)
6953b4c0 1975 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1976
1977 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1978
c271c5c2
SM		 1979 if (ocfs2_mount_local(osb))
1980 snprintf(nodestr, sizeof(nodestr), "local");
1981 else
19fdb624 1982 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1983
1984 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1985 osb->dev_str, nodestr);
ccd979bd
MF		 1986
1987 ocfs2_delete_osb(osb);
1988 kfree(osb);
1989 sb->s_dev = 0;
1990 sb->s_fs_info = NULL;
1991}
1992
1993static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1994 unsigned uuid_bytes)
1995{
1996 int i, ret;
1997 char *ptr;
1998
1999 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2000
cd861280 2001 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 2002 if (osb->uuid_str == NULL)
2003 return -ENOMEM;
2004
ccd979bd
MF		 2005 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2006 /* print with null */
2007 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2008 if (ret != 2) /* drop super cleans up */
2009 return -EINVAL;
2010 /* then only advance past the last char */
2011 ptr += 2;
2012 }
2013
2014 return 0;
2015}
2016
3bdb8efd
PL		 2017/* Make sure entire volume is addressable by our journal. Requires
2018 osb_clusters_at_boot to be valid and for the journal to have been
2019 initialized by ocfs2_journal_init(). */
2020static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2021{
2022 int status = 0;
2023 u64 max_block =
2024 ocfs2_clusters_to_blocks(osb->sb,
2025 osb->osb_clusters_at_boot) - 1;
2026
2027 /* 32-bit block number is always OK. */
2028 if (max_block <= (u32)~0ULL)
2029 goto out;
2030
2031 /* Volume is "huge", so see if our journal is new enough to
2032 support it. */
2033 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2034 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2035 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2036 JBD2_FEATURE_INCOMPAT_64BIT))) {
2037 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2038 "Enable the 'block64' journal option with tunefs.ocfs2");
2039 status = -EFBIG;
2040 goto out;
2041 }
2042
2043 out:
2044 return status;
2045}
2046
ccd979bd
MF		 2047static int ocfs2_initialize_super(struct super_block *sb,
2048 struct buffer_head *bh,
73be192b
JB		 2049 int sector_size,
2050 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2051{
bddb8eb3 2052 int status;
5a254031
MF		 2053 int i, cbits, bbits;
2054 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2055 struct inode *inode = NULL;
ccd979bd 2056 struct ocfs2_journal *journal;
ccd979bd 2057 struct ocfs2_super *osb;
3bdb8efd 2058 u64 total_blocks;
ccd979bd 2059
cd861280 2060 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 2061 if (!osb) {
2062 status = -ENOMEM;
2063 mlog_errno(status);
2064 goto bail;
2065 }
2066
2067 sb->s_fs_info = osb;
2068 sb->s_op = &ocfs2_sops;
ba87167c 2069 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2070 sb->s_export_op = &ocfs2_export_ops;
664dbd5f 2071 sb->s_qcop = &dquot_quotactl_sysfile_ops;
19ece546 2072 sb->dq_op = &ocfs2_quota_operations;
1c92ec67 2073 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cf1d6c76 2074 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2075 sb->s_time_gran = 1;
1751e8a6 2076 sb->s_flags |= SB_NOATIME;
ccd979bd 2077 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2078 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2079 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2080 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
85787090 2081 memcpy(&sb->s_uuid, di->id2.i_super.s_uuid,
58be19dc 2082 sizeof(di->id2.i_super.s_uuid));
ccd979bd 2083
9b7895ef
MF		 2084 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2085
2086 for (i = 0; i < 3; i++)
2087 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2088 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2089
ccd979bd 2090 osb->sb = sb;
ccd979bd 2091 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2092 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2093
34d024f8
MF		 2094 spin_lock_init(&osb->dc_task_lock);
2095 init_waitqueue_head(&osb->dc_event);
2096 osb->dc_work_sequence = 0;
2097 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2098 INIT_LIST_HEAD(&osb->blocked_lock_list);
2099 osb->blocked_lock_count = 0;
ccd979bd 2100 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2101 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2102 ocfs2_init_steal_slots(osb);
ccd979bd 2103
43b10a20 2104 mutex_init(&osb->system_file_mutex);
2105
ccd979bd
MF		 2106 atomic_set(&osb->alloc_stats.moves, 0);
2107 atomic_set(&osb->alloc_stats.local_data, 0);
2108 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2109 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2110 atomic_set(&osb->alloc_stats.bg_extends, 0);
2111
73be192b
JB		 2112 /* Copy the blockcheck stats from the superblock probe */
2113 osb->osb_ecc_stats = *stats;
2114
ccd979bd
MF		 2115 ocfs2_init_node_maps(osb);
2116
2117 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2118 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2119
75d9bbc7
GR		 2120 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2121 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2122 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2123 osb->max_slots);
2124 status = -EINVAL;
2125 goto bail;
2126 }
75d9bbc7 2127
8b712cd5
JM		 2128 ocfs2_orphan_scan_init(osb);
2129
553abd04
JB		 2130 status = ocfs2_recovery_init(osb);
2131 if (status) {
2132 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2133 mlog_errno(status);
2134 goto bail;
2135 }
ccd979bd
MF		 2136
2137 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2138
7f1a37e3
TY		 2139 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2140
ccd979bd
MF		 2141 osb->slot_num = OCFS2_INVALID_SLOT;
2142
8154da3d
TY		 2143 osb->s_xattr_inline_size = le16_to_cpu(
2144 di->id2.i_super.s_xattr_inline_size);
fdd77704 2145
ccd979bd
MF		 2146 osb->local_alloc_state = OCFS2_LA_UNUSED;
2147 osb->local_alloc_bh = NULL;
9c7af40b 2148 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2149
ccd979bd
MF		 2150 init_waitqueue_head(&osb->osb_mount_event);
2151
d02f00cc
MF		 2152 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2153 if (status) {
2154 mlog_errno(status);
2155 goto bail;
2156 }
2157
ccd979bd
MF		 2158 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2159 if (!osb->vol_label) {
2160 mlog(ML_ERROR, "unable to alloc vol label\n");
2161 status = -ENOMEM;
2162 goto bail;
2163 }
2164
539d8264
SM		 2165 osb->slot_recovery_generations =
2166 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2167 GFP_KERNEL);
2168 if (!osb->slot_recovery_generations) {
2169 status = -ENOMEM;
2170 mlog_errno(status);
2171 goto bail;
2172 }
2173
b4df6ed8
MF		 2174 init_waitqueue_head(&osb->osb_wipe_event);
2175 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2176 sizeof(*osb->osb_orphan_wipes),
2177 GFP_KERNEL);
2178 if (!osb->osb_orphan_wipes) {
2179 status = -ENOMEM;
2180 mlog_errno(status);
2181 goto bail;
2182 }
2183
374a263e
TM		 2184 osb->osb_rf_lock_tree = RB_ROOT;
2185
ccd979bd
MF		 2186 osb->s_feature_compat =
2187 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2188 osb->s_feature_ro_compat =
2189 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2190 osb->s_feature_incompat =
2191 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2192
2193 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2194 mlog(ML_ERROR, "couldn't mount because of unsupported "
2195 "optional features (%x).\n", i);
2196 status = -EINVAL;
2197 goto bail;
2198 }
bc98a42c 2199 if (!sb_rdonly(osb->sb) && (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
ccd979bd
MF		 2200 mlog(ML_ERROR, "couldn't mount RDWR because of "
2201 "unsupported optional features (%x).\n", i);
2202 status = -EINVAL;
2203 goto bail;
2204 }
2205
98f486f2
SM		 2206 if (ocfs2_clusterinfo_valid(osb)) {
2207 osb->osb_stackflags =
2208 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2209 strlcpy(osb->osb_cluster_stack,
b61817e1 2210 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2211 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB		 2212 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2213 mlog(ML_ERROR,
2214 "couldn't mount because of an invalid "
2215 "cluster stack label (%s) \n",
2216 osb->osb_cluster_stack);
2217 status = -EINVAL;
2218 goto bail;
2219 }
c74a3bdd
GR		 2220 strlcpy(osb->osb_cluster_name,
2221 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2222 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB		 2223 } else {
2224 /* The empty string is identical with classic tools that
2225 * don't know about s_cluster_info. */
2226 osb->osb_cluster_stack[0] = '\0';
2227 }
2228
ccd979bd
MF		 2229 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2230
2231 /* FIXME
2232 * This should be done in ocfs2_journal_init(), but unknown
2233 * ordering issues will cause the filesystem to crash.
2234 * If anyone wants to figure out what part of the code
2235 * refers to osb->journal before ocfs2_journal_init() is run,
2236 * be my guest.
2237 */
2238 /* initialize our journal structure */
2239
cd861280 2240 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2241 if (!journal) {
2242 mlog(ML_ERROR, "unable to alloc journal\n");
2243 status = -ENOMEM;
2244 goto bail;
2245 }
2246 osb->journal = journal;
2247 journal->j_osb = osb;
2248
2249 atomic_set(&journal->j_num_trans, 0);
2250 init_rwsem(&journal->j_trans_barrier);
2251 init_waitqueue_head(&journal->j_checkpointed);
2252 spin_lock_init(&journal->j_lock);
2253 journal->j_trans_id = (unsigned long) 1;
2254 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2255 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2256 journal->j_state = OCFS2_JOURNAL_FREE;
2257
e3a767b6
JK		 2258 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2259 init_llist_head(&osb->dquot_drop_list);
2260
ccd979bd
MF		 2261 /* get some pseudo constants for clustersize bits */
2262 osb->s_clustersize_bits =
2263 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2264 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF		 2265
2266 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2267 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2268 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2269 osb->s_clustersize);
2270 status = -EINVAL;
2271 goto bail;
2272 }
2273
3bdb8efd
PL		 2274 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2275 le32_to_cpu(di->i_clusters));
2276
2277 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2278 total_blocks);
2279 if (status) {
2280 mlog(ML_ERROR, "Volume too large "
2281 "to mount safely on this system");
2282 status = -EFBIG;
ccd979bd
MF		 2283 goto bail;
2284 }
2285
2286 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2287 sizeof(di->id2.i_super.s_uuid))) {
2288 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2289 status = -ENOMEM;
2290 goto bail;
2291 }
2292
69201bb1
FF		 2293 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2294 OCFS2_MAX_VOL_LABEL_LEN);
ccd979bd
MF		 2295 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2296 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2297 osb->first_cluster_group_blkno =
2298 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2299 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2300 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM		 2301 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2302 (unsigned long long)osb->root_blkno,
2303 (unsigned long long)osb->system_dir_blkno,
2304 osb->s_clustersize_bits);
ccd979bd
MF		 2305
2306 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2307 if (!osb->osb_dlm_debug) {
2308 status = -ENOMEM;
2309 mlog_errno(status);
2310 goto bail;
2311 }
2312
2313 atomic_set(&osb->vol_state, VOLUME_INIT);
2314
2315 /* load root, system_dir, and all global system inodes */
2316 status = ocfs2_init_global_system_inodes(osb);
2317 if (status < 0) {
2318 mlog_errno(status);
2319 goto bail;
2320 }
2321
2322 /*
2323 * global bitmap
2324 */
2325 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2326 OCFS2_INVALID_SLOT);
2327 if (!inode) {
2328 status = -EINVAL;
2329 mlog_errno(status);
2330 goto bail;
2331 }
2332
2333 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2334 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2335 iput(inode);
ccd979bd 2336
8571882c
TM		 2337 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2338 osb->s_feature_incompat) * 8;
ccd979bd
MF		 2339
2340 status = ocfs2_init_slot_info(osb);
2341 if (status < 0) {
2342 mlog_errno(status);
2343 goto bail;
2344 }
9de16262 2345 cleancache_init_shared_fs(sb);
ccd979bd 2346
44be9756 2347 osb->ocfs2_wq = alloc_ordered_workqueue("ocfs2_wq", WQ_MEM_RECLAIM);
35ddf78e 2348 if (!osb->ocfs2_wq) {
2349 status = -ENOMEM;
2350 mlog_errno(status);
2351 }
2352
ccd979bd 2353bail:
ccd979bd
MF		 2354 return status;
2355}
2356
2357/*
2358 * will return: -EAGAIN if it is ok to keep searching for superblocks
2359 * -EINVAL if there is a bad superblock
2360 * 0 on success
2361 */
2362static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2363 struct buffer_head *bh,
73be192b
JB		 2364 u32 blksz,
2365 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2366{
2367 int status = -EAGAIN;
2368
ccd979bd
MF		 2369 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2370 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2371 /* We have to do a raw check of the feature here */
2372 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2373 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2374 status = ocfs2_block_check_validate(bh->b_data,
2375 bh->b_size,
73be192b
JB		 2376 &di->i_check,
2377 stats);
d030cc97
JB		 2378 if (status)
2379 goto out;
2380 }
ccd979bd
MF		 2381 status = -EINVAL;
2382 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2383 mlog(ML_ERROR, "found superblock with incorrect block "
2384 "size: found %u, should be %u\n",
2385 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2386 blksz);
2387 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2388 OCFS2_MAJOR_REV_LEVEL ||
2389 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2390 OCFS2_MINOR_REV_LEVEL) {
2391 mlog(ML_ERROR, "found superblock with bad version: "
2392 "found %u.%u, should be %u.%u\n",
2393 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2394 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2395 OCFS2_MAJOR_REV_LEVEL,
2396 OCFS2_MINOR_REV_LEVEL);
2397 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2398 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2399 "found %llu, should be %llu\n",
1ca1a111 2400 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2401 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2402 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2403 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2404 mlog(ML_ERROR, "bad cluster size found: %u\n",
2405 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2406 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2407 mlog(ML_ERROR, "bad root_blkno: 0\n");
2408 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2409 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2410 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2411 mlog(ML_ERROR,
2412 "Superblock slots found greater than file system "
2413 "maximum: found %u, max %u\n",
2414 le16_to_cpu(di->id2.i_super.s_max_slots),
2415 OCFS2_MAX_SLOTS);
2416 } else {
2417 /* found it! */
2418 status = 0;
2419 }
2420 }
2421
d030cc97 2422out:
c1e8d35e
TM		 2423 if (status && status != -EAGAIN)
2424 mlog_errno(status);
ccd979bd
MF		 2425 return status;
2426}
2427
2428static int ocfs2_check_volume(struct ocfs2_super *osb)
2429{
bddb8eb3 2430 int status;
ccd979bd 2431 int dirty;
c271c5c2 2432 int local;
ccd979bd
MF		 2433 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2434 * recover
2435 * ourselves. */
2436
ccd979bd
MF		 2437 /* Init our journal object. */
2438 status = ocfs2_journal_init(osb->journal, &dirty);
2439 if (status < 0) {
2440 mlog(ML_ERROR, "Could not initialize journal!\n");
2441 goto finally;
2442 }
2443
3bdb8efd
PL		 2444 /* Now that journal has been initialized, check to make sure
2445 entire volume is addressable. */
2446 status = ocfs2_journal_addressable(osb);
2447 if (status)
2448 goto finally;
2449
ccd979bd
MF		 2450 /* If the journal was unmounted cleanly then we don't want to
2451 * recover anything. Otherwise, journal_load will do that
2452 * dirty work for us :) */
2453 if (!dirty) {
2454 status = ocfs2_journal_wipe(osb->journal, 0);
2455 if (status < 0) {
2456 mlog_errno(status);
2457 goto finally;
2458 }
2459 } else {
619c200d
SM		 2460 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2461 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF		 2462 }
2463
c271c5c2
SM		 2464 local = ocfs2_mount_local(osb);
2465
ccd979bd 2466 /* will play back anything left in the journal. */
539d8264 2467 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2468 if (status < 0) {
2469 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2470 goto finally;
2471 }
ccd979bd 2472
1dfeb768 2473 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2474 jbd2_journal_set_features(osb->journal->j_journal,
2475 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2476 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2477 else
2478 jbd2_journal_clear_features(osb->journal->j_journal,
2479 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2480 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2481
ccd979bd
MF		 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
ccd979bd
MF		 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. */
ccd979bd
MF		 2504 osb->local_alloc_copy = local_alloc;
2505 local_alloc = NULL;
2506 }
2507
2508 /* go through each journal, trylock it and if you get the
2509 * lock, and it's marked as dirty, set the bit in the recover
2510 * map and launch a recovery thread for it. */
2511 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2512 if (status < 0) {
2513 mlog_errno(status);
2514 goto finally;
2515 }
2516
2517 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2518 if (status < 0)
2519 mlog_errno(status);
2520
2521finally:
d787ab09 2522 kfree(local_alloc);
ccd979bd 2523
c1e8d35e
TM		 2524 if (status)
2525 mlog_errno(status);
ccd979bd
MF		 2526 return status;
2527}
2528
2529/*
2530 * The routine gets called from dismount or close whenever a dismount on
2531 * volume is requested and the osb open count becomes 1.
2532 * It will remove the osb from the global list and also free up all the
2533 * initialized resources and fileobject.
2534 */
2535static void ocfs2_delete_osb(struct ocfs2_super *osb)
2536{
ccd979bd
MF		 2537 /* This function assumes that the caller has the main osb resource */
2538
35ddf78e 2539 /* ocfs2_initializer_super have already created this workqueue */
23e0813a 2540 if (osb->ocfs2_wq)
35ddf78e 2541 destroy_workqueue(osb->ocfs2_wq);
35ddf78e 2542
8e8a4603 2543 ocfs2_free_slot_info(osb);
ccd979bd 2544
b4df6ed8 2545 kfree(osb->osb_orphan_wipes);
539d8264 2546 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2547 /* FIXME
2548 * This belongs in journal shutdown, but because we have to
421f91d2 2549 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF		 2550 * we free it here.
2551 */
2552 kfree(osb->journal);
d787ab09 2553 kfree(osb->local_alloc_copy);
ccd979bd 2554 kfree(osb->uuid_str);
f13a568e 2555 kfree(osb->vol_label);
ccd979bd
MF		 2556 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2557 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF		 2558}
2559
7d0fb914
GR		 2560/* Depending on the mount option passed, perform one of the following:
2561 * Put OCFS2 into a readonly state (default)
2562 * Return EIO so that only the process errs
2563 * Fix the error as if fsck.ocfs2 -y
2564 * panic
2565 */
2566static int ocfs2_handle_error(struct super_block *sb)
ccd979bd
MF		 2567{
2568 struct ocfs2_super *osb = OCFS2_SB(sb);
7d0fb914 2569 int rv = 0;
ccd979bd
MF		 2570
2571 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
7d0fb914
GR		 2572 pr_crit("On-disk corruption discovered. "
2573 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
ccd979bd 2574
7d0fb914
GR		 2575 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC) {
2576 panic("OCFS2: (device %s): panic forced after error\n",
2577 sb->s_id);
2578 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2579 pr_crit("OCFS2: Returning error to the calling process.\n");
2580 rv = -EIO;
2581 } else { /* default option */
2582 rv = -EROFS;
bc98a42c 2583 if (sb_rdonly(sb) && (ocfs2_is_soft_readonly(osb) || ocfs2_is_hard_readonly(osb)))
7d0fb914
GR		 2584 return rv;
2585
2586 pr_crit("OCFS2: File system is now read-only.\n");
1751e8a6 2587 sb->s_flags |= SB_RDONLY;
7d0fb914
GR		 2588 ocfs2_set_ro_flag(osb, 0);
2589 }
2590
2591 return rv;
ccd979bd
MF		 2592}
2593
7d0fb914 2594int __ocfs2_error(struct super_block *sb, const char *function,
1543306e 2595 const char *fmt, ...)
ccd979bd 2596{
1543306e 2597 struct va_format vaf;
ccd979bd
MF		 2598 va_list args;
2599
2600 va_start(args, fmt);
1543306e
JP		 2601 vaf.fmt = fmt;
2602 vaf.va = &args;
ccd979bd
MF		 2603
2604 /* Not using mlog here because we want to show the actual
2605 * function the error came from. */
7ecef14a 2606 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
1543306e
JP		 2607 sb->s_id, function, &vaf);
2608
2609 va_end(args);
ccd979bd 2610
7d0fb914 2611 return ocfs2_handle_error(sb);
ccd979bd
MF		 2612}
2613
2614/* Handle critical errors. This is intentionally more drastic than
2615 * ocfs2_handle_error, so we only use for things like journal errors,
2616 * etc. */
1543306e 2617void __ocfs2_abort(struct super_block *sb, const char *function,
ccd979bd
MF		 2618 const char *fmt, ...)
2619{
1543306e 2620 struct va_format vaf;
ccd979bd
MF		 2621 va_list args;
2622
2623 va_start(args, fmt);
ccd979bd 2624
1543306e
JP		 2625 vaf.fmt = fmt;
2626 vaf.va = &args;
2627
7ecef14a 2628 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
1543306e
JP		 2629 sb->s_id, function, &vaf);
2630
2631 va_end(args);
ccd979bd
MF		 2632
2633 /* We don't have the cluster support yet to go straight to
2634 * hard readonly in here. Until then, we want to keep
2635 * ocfs2_abort() so that we can at least mark critical
2636 * errors.
2637 *
2638 * TODO: This should abort the journal and alert other nodes
2639 * that our slot needs recovery. */
2640
2641 /* Force a panic(). This stinks, but it's better than letting
2642 * things continue without having a proper hard readonly
2643 * here. */
a2f2ddbf
SM		 2644 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2645 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2646 ocfs2_handle_error(sb);
2647}
2648
e4b963f1
JB		 2649/*
2650 * Void signal blockers, because in-kernel sigprocmask() only fails
2651 * when SIG_* is wrong.
2652 */
2653void ocfs2_block_signals(sigset_t *oldset)
2654{
2655 int rc;
2656 sigset_t blocked;
2657
2658 sigfillset(&blocked);
2659 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2660 BUG_ON(rc);
2661}
2662
2663void ocfs2_unblock_signals(sigset_t *oldset)
2664{
2665 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2666 BUG_ON(rc);
2667}
2668
ccd979bd
MF		 2669module_init(ocfs2_init);
2670module_exit(ocfs2_exit);
Linux 6.x block layer and io_uring tree(s)