projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-4.6-fixes' of git://git.kernel.org/pub/scm/linux/kernel/git/tj...
[deliverable/linux.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>
ccd979bd 45
80a9a84d
WW		 46#define CREATE_TRACE_POINTS
47#include "ocfs2_trace.h"
48
ccd979bd
MF		 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"
a11f7e63 57#include "aops.h"
d030cc97 58#include "blockcheck.h"
ccd979bd
MF		 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"
cf1d6c76 71#include "xattr.h"
9e33d69f 72#include "quota.h"
374a263e 73#include "refcounttree.h"
b89c5428 74#include "suballoc.h"
ccd979bd
MF		 75
76#include "buffer_head_io.h"
a849d468 77#include "filecheck.h"
ccd979bd 78
1a5c4e2a 79static struct kmem_cache *ocfs2_inode_cachep;
9e33d69f
JK		 80struct kmem_cache *ocfs2_dquot_cachep;
81struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 82
1a5c4e2a 83static struct dentry *ocfs2_debugfs_root;
ccd979bd
MF		 84
85MODULE_AUTHOR("Oracle");
86MODULE_LICENSE("GPL");
ff8fb335 87MODULE_DESCRIPTION("OCFS2 cluster file system");
ccd979bd 88
c0123ade
TY		 89struct mount_options
90{
d147b3d6 91 unsigned long commit_interval;
c0123ade
TY		 92 unsigned long mount_opt;
93 unsigned int atime_quantum;
94 signed short slot;
73c8a800 95 int localalloc_opt;
d02f00cc 96 unsigned int resv_level;
83f92318 97 int dir_resv_level;
b61817e1 98 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 99};
100
ccd979bd 101static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 102 struct mount_options *mopt,
baf4661a 103 int is_remount);
5297aad8
JK		 104static int ocfs2_check_set_options(struct super_block *sb,
105 struct mount_options *options);
34c80b1d 106static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
ccd979bd
MF		 107static void ocfs2_put_super(struct super_block *sb);
108static int ocfs2_mount_volume(struct super_block *sb);
109static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
110static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
111static int ocfs2_initialize_mem_caches(void);
112static void ocfs2_free_mem_caches(void);
113static void ocfs2_delete_osb(struct ocfs2_super *osb);
114
726c3342 115static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 116
117static int ocfs2_sync_fs(struct super_block *sb, int wait);
118
119static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
120static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 121static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 122static int ocfs2_check_volume(struct ocfs2_super *osb);
123static int ocfs2_verify_volume(struct ocfs2_dinode *di,
124 struct buffer_head *bh,
73be192b
JB		 125 u32 sectsize,
126 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 127static int ocfs2_initialize_super(struct super_block *sb,
128 struct buffer_head *bh,
73be192b
JB		 129 int sector_size,
130 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 131static int ocfs2_get_sector(struct super_block *sb,
132 struct buffer_head **bh,
133 int block,
134 int sect_size);
ccd979bd
MF		 135static struct inode *ocfs2_alloc_inode(struct super_block *sb);
136static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 137static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
138static int ocfs2_enable_quotas(struct ocfs2_super *osb);
139static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 140
1c92ec67
JK		 141static struct dquot **ocfs2_get_dquots(struct inode *inode)
142{
143 return OCFS2_I(inode)->i_dquot;
144}
145
ee9b6d61 146static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 147 .statfs = ocfs2_statfs,
148 .alloc_inode = ocfs2_alloc_inode,
149 .destroy_inode = ocfs2_destroy_inode,
150 .drop_inode = ocfs2_drop_inode,
066d92dc 151 .evict_inode = ocfs2_evict_inode,
ccd979bd 152 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 153 .put_super = ocfs2_put_super,
154 .remount_fs = ocfs2_remount,
d550071c 155 .show_options = ocfs2_show_options,
9e33d69f
JK		 156 .quota_read = ocfs2_quota_read,
157 .quota_write = ocfs2_quota_write,
1c92ec67 158 .get_dquots = ocfs2_get_dquots,
ccd979bd
MF		 159};
160
161enum {
162 Opt_barrier,
163 Opt_err_panic,
164 Opt_err_ro,
165 Opt_intr,
166 Opt_nointr,
167 Opt_hb_none,
168 Opt_hb_local,
2c442719 169 Opt_hb_global,
ccd979bd
MF		 170 Opt_data_ordered,
171 Opt_data_writeback,
7f1a37e3 172 Opt_atime_quantum,
baf4661a 173 Opt_slot,
d147b3d6 174 Opt_commit,
2fbe8d1e 175 Opt_localalloc,
53fc622b 176 Opt_localflocks,
b61817e1 177 Opt_stack,
cf1d6c76
TY		 178 Opt_user_xattr,
179 Opt_nouser_xattr,
12462f1d 180 Opt_inode64,
a68979b8
TY		 181 Opt_acl,
182 Opt_noacl,
19ece546
JK		 183 Opt_usrquota,
184 Opt_grpquota,
7bdb0d18
TY		 185 Opt_coherency_buffered,
186 Opt_coherency_full,
d02f00cc 187 Opt_resv_level,
83f92318 188 Opt_dir_resv_level,
1dfeb768 189 Opt_journal_async_commit,
7d0fb914 190 Opt_err_cont,
ccd979bd
MF		 191 Opt_err,
192};
193
a447c093 194static const match_table_t tokens = {
ccd979bd
MF		 195 {Opt_barrier, "barrier=%u"},
196 {Opt_err_panic, "errors=panic"},
197 {Opt_err_ro, "errors=remount-ro"},
198 {Opt_intr, "intr"},
199 {Opt_nointr, "nointr"},
200 {Opt_hb_none, OCFS2_HB_NONE},
201 {Opt_hb_local, OCFS2_HB_LOCAL},
2c442719 202 {Opt_hb_global, OCFS2_HB_GLOBAL},
ccd979bd
MF		 203 {Opt_data_ordered, "data=ordered"},
204 {Opt_data_writeback, "data=writeback"},
7f1a37e3 205 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 206 {Opt_slot, "preferred_slot=%u"},
d147b3d6 207 {Opt_commit, "commit=%u"},
2fbe8d1e 208 {Opt_localalloc, "localalloc=%d"},
53fc622b 209 {Opt_localflocks, "localflocks"},
b61817e1 210 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 211 {Opt_user_xattr, "user_xattr"},
212 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 213 {Opt_inode64, "inode64"},
a68979b8
TY		 214 {Opt_acl, "acl"},
215 {Opt_noacl, "noacl"},
19ece546
JK		 216 {Opt_usrquota, "usrquota"},
217 {Opt_grpquota, "grpquota"},
7bdb0d18
TY		 218 {Opt_coherency_buffered, "coherency=buffered"},
219 {Opt_coherency_full, "coherency=full"},
d02f00cc 220 {Opt_resv_level, "resv_level=%u"},
83f92318 221 {Opt_dir_resv_level, "dir_resv_level=%u"},
1dfeb768 222 {Opt_journal_async_commit, "journal_async_commit"},
7d0fb914 223 {Opt_err_cont, "errors=continue"},
ccd979bd
MF		 224 {Opt_err, NULL}
225};
226
50397507
SM		 227#ifdef CONFIG_DEBUG_FS
228static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
229{
50397507
SM		 230 struct ocfs2_cluster_connection *cconn = osb->cconn;
231 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 232 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
233 int i, out = 0;
bfd97a03 234 unsigned long flags;
50397507
SM		 235
236 out += snprintf(buf + out, len - out,
237 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
238 "Device", osb->dev_str, osb->uuid_str,
239 osb->fs_generation, osb->vol_label);
240
241 out += snprintf(buf + out, len - out,
242 "%10s => State: %d Flags: 0x%lX\n", "Volume",
243 atomic_read(&osb->vol_state), osb->osb_flags);
244
245 out += snprintf(buf + out, len - out,
246 "%10s => Block: %lu Cluster: %d\n", "Sizes",
247 osb->sb->s_blocksize, osb->s_clustersize);
248
249 out += snprintf(buf + out, len - out,
250 "%10s => Compat: 0x%X Incompat: 0x%X "
251 "ROcompat: 0x%X\n",
252 "Features", osb->s_feature_compat,
253 osb->s_feature_incompat, osb->s_feature_ro_compat);
254
255 out += snprintf(buf + out, len - out,
256 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
257 osb->s_mount_opt, osb->s_atime_quantum);
258
c3d38840
SM		 259 if (cconn) {
260 out += snprintf(buf + out, len - out,
261 "%10s => Stack: %s Name: %*s "
262 "Version: %d.%d\n", "Cluster",
263 (*osb->osb_cluster_stack == '\0' ?
264 "o2cb" : osb->osb_cluster_stack),
265 cconn->cc_namelen, cconn->cc_name,
266 cconn->cc_version.pv_major,
267 cconn->cc_version.pv_minor);
268 }
50397507 269
bfd97a03 270 spin_lock_irqsave(&osb->dc_task_lock, flags);
50397507
SM		 271 out += snprintf(buf + out, len - out,
272 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
273 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 274 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
275 osb->blocked_lock_count, osb->dc_wake_sequence,
276 osb->dc_work_sequence);
bfd97a03 277 spin_unlock_irqrestore(&osb->dc_task_lock, flags);
50397507
SM		 278
279 spin_lock(&osb->osb_lock);
280 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
281 "Recovery",
282 (osb->recovery_thread_task ?
283 task_pid_nr(osb->recovery_thread_task) : -1));
284 if (rm->rm_used == 0)
285 out += snprintf(buf + out, len - out, " None\n");
286 else {
287 for (i = 0; i < rm->rm_used; i++)
288 out += snprintf(buf + out, len - out, " %d",
289 rm->rm_entries[i]);
290 out += snprintf(buf + out, len - out, "\n");
291 }
292 spin_unlock(&osb->osb_lock);
293
294 out += snprintf(buf + out, len - out,
8fa9d17f 295 "%10s => Pid: %d Interval: %lu\n", "Commit",
c3d38840 296 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
8fa9d17f 297 osb->osb_commit_interval);
50397507
SM		 298
299 out += snprintf(buf + out, len - out,
c3d38840 300 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 301 "Journal", osb->journal->j_state,
c3d38840
SM		 302 osb->journal->j_trans_id,
303 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 304
305 out += snprintf(buf + out, len - out,
306 "%10s => GlobalAllocs: %d LocalAllocs: %d "
307 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
308 "Stats",
309 atomic_read(&osb->alloc_stats.bitmap_data),
310 atomic_read(&osb->alloc_stats.local_data),
311 atomic_read(&osb->alloc_stats.bg_allocs),
312 atomic_read(&osb->alloc_stats.moves),
313 atomic_read(&osb->alloc_stats.bg_extends));
314
315 out += snprintf(buf + out, len - out,
316 "%10s => State: %u Descriptor: %llu Size: %u bits "
317 "Default: %u bits\n",
318 "LocalAlloc", osb->local_alloc_state,
319 (unsigned long long)osb->la_last_gd,
320 osb->local_alloc_bits, osb->local_alloc_default_bits);
321
322 spin_lock(&osb->osb_lock);
323 out += snprintf(buf + out, len - out,
b89c5428
TY		 324 "%10s => InodeSlot: %d StolenInodes: %d, "
325 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 326 osb->s_inode_steal_slot,
b89c5428
TY		 327 atomic_read(&osb->s_num_inodes_stolen),
328 osb->s_meta_steal_slot,
329 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 330 spin_unlock(&osb->osb_lock);
331
df152c24
SM		 332 out += snprintf(buf + out, len - out, "OrphanScan => ");
333 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
334 os->os_count, os->os_seqno);
335 out += snprintf(buf + out, len - out, " Last Scan: ");
336 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
337 out += snprintf(buf + out, len - out, "Disabled\n");
338 else
339 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
340 (get_seconds() - os->os_scantime.tv_sec));
341
50397507
SM		 342 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
343 "Slots", "Num", "RecoGen");
50397507
SM		 344 for (i = 0; i < osb->max_slots; ++i) {
345 out += snprintf(buf + out, len - out,
346 "%10s %c %3d %10d\n",
347 " ",
348 (i == osb->slot_num ? '*' : ' '),
349 i, osb->slot_recovery_generations[i]);
350 }
351
352 return out;
353}
354
355static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
356{
357 struct ocfs2_super *osb = inode->i_private;
358 char *buf = NULL;
359
360 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
361 if (!buf)
362 goto bail;
363
364 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
365
366 file->private_data = buf;
367
368 return 0;
369bail:
370 return -ENOMEM;
371}
372
373static int ocfs2_debug_release(struct inode *inode, struct file *file)
374{
375 kfree(file->private_data);
376 return 0;
377}
378
379static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
380 size_t nbytes, loff_t *ppos)
381{
382 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
383 i_size_read(file->f_mapping->host));
384}
385#else
386static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
387{
388 return 0;
389}
390static int ocfs2_debug_release(struct inode *inode, struct file *file)
391{
392 return 0;
393}
394static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
395 size_t nbytes, loff_t *ppos)
396{
397 return 0;
398}
399#endif /* CONFIG_DEBUG_FS */
400
828c0950 401static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 402 .open = ocfs2_osb_debug_open,
403 .release = ocfs2_debug_release,
404 .read = ocfs2_debug_read,
405 .llseek = generic_file_llseek,
406};
407
ccd979bd
MF		 408static int ocfs2_sync_fs(struct super_block *sb, int wait)
409{
bddb8eb3 410 int status;
ccd979bd
MF		 411 tid_t target;
412 struct ocfs2_super *osb = OCFS2_SB(sb);
413
ccd979bd
MF		 414 if (ocfs2_is_hard_readonly(osb))
415 return -EROFS;
416
417 if (wait) {
418 status = ocfs2_flush_truncate_log(osb);
419 if (status < 0)
420 mlog_errno(status);
421 } else {
422 ocfs2_schedule_truncate_log_flush(osb, 0);
423 }
424
2b4e30fb
JB		 425 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
426 &target)) {
ccd979bd 427 if (wait)
2b4e30fb
JB		 428 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
429 target);
ccd979bd
MF		 430 }
431 return 0;
432}
433
1a224ad1
JK		 434static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
435{
436 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
437 && (ino == USER_QUOTA_SYSTEM_INODE
438 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
439 return 0;
440 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
441 && (ino == GROUP_QUOTA_SYSTEM_INODE
442 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
443 return 0;
444 return 1;
445}
446
ccd979bd
MF		 447static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
448{
449 struct inode *new = NULL;
450 int status = 0;
451 int i;
452
5fa0613e 453 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 454 if (IS_ERR(new)) {
455 status = PTR_ERR(new);
456 mlog_errno(status);
457 goto bail;
458 }
459 osb->root_inode = new;
460
5fa0613e 461 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 462 if (IS_ERR(new)) {
463 status = PTR_ERR(new);
464 mlog_errno(status);
465 goto bail;
466 }
467 osb->sys_root_inode = new;
468
469 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
470 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 471 if (!ocfs2_need_system_inode(osb, i))
472 continue;
ccd979bd
MF		 473 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
474 if (!new) {
475 ocfs2_release_system_inodes(osb);
476 status = -EINVAL;
477 mlog_errno(status);
478 /* FIXME: Should ERROR_RO_FS */
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);
507 status = -EINVAL;
508 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
509 status, i, osb->slot_num);
510 goto bail;
511 }
512 /* the array now has one ref, so drop this one */
513 iput(new);
514 }
515
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
MF		 676 /* We're going to/from readonly mode. */
677 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 678 /* Disable quota accounting before remounting RO */
679 if (*flags & MS_RDONLY) {
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
693 if (*flags & MS_RDONLY) {
ccd979bd
MF		 694 sb->s_flags |= MS_RDONLY;
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 }
711 sb->s_flags &= ~MS_RDONLY;
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
JK		 717 /* Enable quota accounting after remounting RW */
718 if (!ret && !(*flags & MS_RDONLY)) {
719 if (sb_any_quota_suspended(sb))
720 ret = ocfs2_susp_quotas(osb, 1);
721 else
722 ret = ocfs2_enable_quotas(osb);
723 if (ret < 0) {
724 /* Return back changes... */
725 spin_lock(&osb->osb_lock);
726 sb->s_flags |= MS_RDONLY;
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
JK		 745
746 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
747 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
748 MS_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 /* Cancel periodic syncing before we grab dqonoff_mutex */
989 oinfo = sb_dqinfo(sb, type)->dqi_priv;
990 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK		 991 inode = igrab(sb->s_dquot.files[type]);
992 /* Turn off quotas. This will remove all dquot structures from
993 * memory and so they will be automatically synced to global
994 * quota files */
0f0dd62f
CH		 995 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
996 DQUOT_LIMITS_ENABLED);
19ece546
JK		 997 if (!inode)
998 continue;
999 iput(inode);
1000 }
1001}
1002
ccd979bd
MF		 1003static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1004{
1005 struct dentry *root;
1006 int status, sector_size;
c0123ade 1007 struct mount_options parsed_options;
ccd979bd
MF		 1008 struct inode *inode = NULL;
1009 struct ocfs2_super *osb = NULL;
1010 struct buffer_head *bh = NULL;
49fa8140 1011 char nodestr[12];
73be192b 1012 struct ocfs2_blockcheck_stats stats;
ccd979bd 1013
32a42d39 1014 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1015
c0123ade 1016 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1017 status = -EINVAL;
1018 goto read_super_error;
1019 }
1020
1021 /* probe for superblock */
73be192b 1022 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1023 if (status < 0) {
1024 mlog(ML_ERROR, "superblock probe failed!\n");
1025 goto read_super_error;
1026 }
1027
73be192b 1028 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1029 osb = OCFS2_SB(sb);
1030 if (status < 0) {
1031 mlog_errno(status);
1032 goto read_super_error;
1033 }
1034 brelse(bh);
1035 bh = NULL;
a68979b8 1036
5297aad8
JK		 1037 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1038 status = -EINVAL;
1039 goto read_super_error;
1040 }
c0123ade
TY		 1041 osb->s_mount_opt = parsed_options.mount_opt;
1042 osb->s_atime_quantum = parsed_options.atime_quantum;
1043 osb->preferred_slot = parsed_options.slot;
d147b3d6 1044 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1045
1046 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1047 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF		 1048 osb->osb_dir_resv_level = parsed_options.resv_level;
1049 if (parsed_options.dir_resv_level == -1)
1050 osb->osb_dir_resv_level = parsed_options.resv_level;
1051 else
1052 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1053
b61817e1
JB		 1054 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1055 if (status)
1056 goto read_super_error;
1057
ccd979bd
MF		 1058 sb->s_magic = OCFS2_SUPER_MAGIC;
1059
9e1f1de0 1060 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
a68979b8
TY		 1061 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1062
ccd979bd
MF		 1063 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1064 * heartbeat=none */
1065 if (bdev_read_only(sb->s_bdev)) {
1066 if (!(sb->s_flags & MS_RDONLY)) {
1067 status = -EACCES;
1068 mlog(ML_ERROR, "Readonly device detected but readonly "
1069 "mount was not specified.\n");
1070 goto read_super_error;
1071 }
1072
1073 /* You should not be able to start a local heartbeat
1074 * on a readonly device. */
1075 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1076 status = -EROFS;
1077 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1078 "device.\n");
1079 goto read_super_error;
1080 }
1081
1082 status = ocfs2_check_journals_nolocks(osb);
1083 if (status < 0) {
1084 if (status == -EROFS)
1085 mlog(ML_ERROR, "Recovery required on readonly "
1086 "file system, but write access is "
1087 "unavailable.\n");
1088 else
2bd63216 1089 mlog_errno(status);
ccd979bd
MF		 1090 goto read_super_error;
1091 }
1092
1093 ocfs2_set_ro_flag(osb, 1);
1094
619c200d
SM		 1095 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1096 "Cluster services will not be used for this mount. "
1097 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF		 1098 }
1099
1100 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1101 if (sb->s_flags & MS_RDONLY)
1102 ocfs2_set_ro_flag(osb, 0);
1103 }
1104
c271c5c2
SM		 1105 status = ocfs2_verify_heartbeat(osb);
1106 if (status < 0) {
1107 mlog_errno(status);
1108 goto read_super_error;
1109 }
1110
ccd979bd
MF		 1111 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1112 ocfs2_debugfs_root);
8f443e23 1113 if (!osb->osb_debug_root) {
ccd979bd
MF		 1114 status = -EINVAL;
1115 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1116 goto read_super_error;
1117 }
1118
50397507
SM		 1119 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1120 osb->osb_debug_root,
1121 osb,
1122 &ocfs2_osb_debug_fops);
8f443e23 1123 if (!osb->osb_ctxt) {
50397507
SM		 1124 status = -EINVAL;
1125 mlog_errno(status);
1126 goto read_super_error;
1127 }
1128
73be192b
JB		 1129 if (ocfs2_meta_ecc(osb)) {
1130 status = ocfs2_blockcheck_stats_debugfs_install(
1131 &osb->osb_ecc_stats,
1132 osb->osb_debug_root);
1133 if (status) {
1134 mlog(ML_ERROR,
1135 "Unable to create blockcheck statistics "
1136 "files\n");
1137 goto read_super_error;
1138 }
1139 }
1140
ccd979bd 1141 status = ocfs2_mount_volume(sb);
ccd979bd
MF		 1142 if (status < 0)
1143 goto read_super_error;
1144
be0d93f0
AV		 1145 if (osb->root_inode)
1146 inode = igrab(osb->root_inode);
1147
ccd979bd
MF		 1148 if (!inode) {
1149 status = -EIO;
1150 mlog_errno(status);
1151 goto read_super_error;
1152 }
1153
48fde701 1154 root = d_make_root(inode);
ccd979bd
MF		 1155 if (!root) {
1156 status = -ENOMEM;
1157 mlog_errno(status);
1158 goto read_super_error;
1159 }
1160
1161 sb->s_root = root;
1162
1163 ocfs2_complete_mount_recovery(osb);
1164
c271c5c2
SM		 1165 if (ocfs2_mount_local(osb))
1166 snprintf(nodestr, sizeof(nodestr), "local");
1167 else
19fdb624 1168 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1169
1170 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1171 "with %s data mode.\n",
c271c5c2 1172 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1173 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1174 "ordered");
1175
1176 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1177 wake_up(&osb->osb_mount_event);
1178
19ece546
JK		 1179 /* Now we can initialize quotas because we can afford to wait
1180 * for cluster locks recovery now. That also means that truncation
1181 * log recovery can happen but that waits for proper quota setup */
1182 if (!(sb->s_flags & MS_RDONLY)) {
1183 status = ocfs2_enable_quotas(osb);
1184 if (status < 0) {
1185 /* We have to err-out specially here because
1186 * s_root is already set */
1187 mlog_errno(status);
1188 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1189 wake_up(&osb->osb_mount_event);
19ece546
JK		 1190 return status;
1191 }
1192 }
1193
1194 ocfs2_complete_quota_recovery(osb);
1195
1196 /* Now we wake up again for processes waiting for quotas */
1197 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1198 wake_up(&osb->osb_mount_event);
1199
df152c24 1200 /* Start this when the mount is almost sure of being successful */
8b712cd5 1201 ocfs2_orphan_scan_start(osb);
df152c24 1202
a849d468
GH		 1203 /* Create filecheck sysfile /sys/fs/ocfs2/<devname>/filecheck */
1204 ocfs2_filecheck_create_sysfs(sb);
1205
ccd979bd
MF		 1206 return status;
1207
1208read_super_error:
a81cb88b 1209 brelse(bh);
ccd979bd 1210
ccd979bd
MF		 1211 if (osb) {
1212 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1213 wake_up(&osb->osb_mount_event);
1214 ocfs2_dismount_volume(sb, 1);
1215 }
1216
c1e8d35e
TM		 1217 if (status)
1218 mlog_errno(status);
ccd979bd
MF		 1219 return status;
1220}
1221
152a0836 1222static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH		 1223 int flags,
1224 const char *dev_name,
152a0836 1225 void *data)
ccd979bd 1226{
152a0836 1227 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF		 1228}
1229
1230static struct file_system_type ocfs2_fs_type = {
1231 .owner = THIS_MODULE,
1232 .name = "ocfs2",
152a0836 1233 .mount = ocfs2_mount,
8ed6b237 1234 .kill_sb = kill_block_super,
1ba9da2f 1235 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1236 .next = NULL
1237};
91417705 1238MODULE_ALIAS_FS("ocfs2");
ccd979bd 1239
5297aad8
JK		 1240static int ocfs2_check_set_options(struct super_block *sb,
1241 struct mount_options *options)
1242{
1243 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1244 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1245 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1246 mlog(ML_ERROR, "User quotas were requested, but this "
1247 "filesystem does not have the feature enabled.\n");
1248 return 0;
1249 }
1250 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1251 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1252 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1253 mlog(ML_ERROR, "Group quotas were requested, but this "
1254 "filesystem does not have the feature enabled.\n");
1255 return 0;
1256 }
1257 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1258 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1259 mlog(ML_ERROR, "ACL support requested but extended attributes "
1260 "feature is not enabled\n");
1261 return 0;
1262 }
1263 /* No ACL setting specified? Use XATTR feature... */
1264 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1265 OCFS2_MOUNT_NO_POSIX_ACL))) {
1266 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1267 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1268 else
1269 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1270 }
1271 return 1;
1272}
1273
ccd979bd
MF		 1274static int ocfs2_parse_options(struct super_block *sb,
1275 char *options,
c0123ade 1276 struct mount_options *mopt,
ccd979bd
MF		 1277 int is_remount)
1278{
52c303c5 1279 int status, user_stack = 0;
ccd979bd 1280 char *p;
2c442719 1281 u32 tmp;
3eb5bdf0
NZ		 1282 int token, option;
1283 substring_t args[MAX_OPT_ARGS];
ccd979bd 1284
32a42d39 1285 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1286
d147b3d6 1287 mopt->commit_interval = 0;
4b37fcb7 1288 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY		 1289 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1290 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1291 mopt->localalloc_opt = -1;
b61817e1 1292 mopt->cluster_stack[0] = '\0';
d02f00cc 1293 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1294 mopt->dir_resv_level = -1;
ccd979bd
MF		 1295
1296 if (!options) {
1297 status = 1;
1298 goto bail;
1299 }
1300
1301 while ((p = strsep(&options, ",")) != NULL) {
ccd979bd
MF		 1302 if (!*p)
1303 continue;
1304
1305 token = match_token(p, tokens, args);
1306 switch (token) {
1307 case Opt_hb_local:
c0123ade 1308 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1309 break;
1310 case Opt_hb_none:
2c442719
SM		 1311 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1312 break;
1313 case Opt_hb_global:
1314 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF		 1315 break;
1316 case Opt_barrier:
1317 if (match_int(&args[0], &option)) {
1318 status = 0;
1319 goto bail;
1320 }
1321 if (option)
c0123ade 1322 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1323 else
c0123ade 1324 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1325 break;
1326 case Opt_intr:
c0123ade 1327 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1328 break;
1329 case Opt_nointr:
c0123ade 1330 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1331 break;
1332 case Opt_err_panic:
7d0fb914
GR		 1333 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
1334 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
c0123ade 1335 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1336 break;
1337 case Opt_err_ro:
7d0fb914 1338 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_CONT;
c0123ade 1339 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
7d0fb914
GR		 1340 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_ROFS;
1341 break;
1342 case Opt_err_cont:
1343 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_ROFS;
1344 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1345 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_CONT;
ccd979bd
MF		 1346 break;
1347 case Opt_data_ordered:
c0123ade 1348 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1349 break;
1350 case Opt_data_writeback:
c0123ade 1351 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1352 break;
cf1d6c76
TY		 1353 case Opt_user_xattr:
1354 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1355 break;
1356 case Opt_nouser_xattr:
1357 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1358 break;
7f1a37e3
TY		 1359 case Opt_atime_quantum:
1360 if (match_int(&args[0], &option)) {
1361 status = 0;
1362 goto bail;
1363 }
1364 if (option >= 0)
c0123ade 1365 mopt->atime_quantum = option;
7f1a37e3 1366 break;
baf4661a 1367 case Opt_slot:
baf4661a
SM		 1368 if (match_int(&args[0], &option)) {
1369 status = 0;
1370 goto bail;
1371 }
1372 if (option)
c0123ade 1373 mopt->slot = (s16)option;
baf4661a 1374 break;
d147b3d6 1375 case Opt_commit:
d147b3d6
MF		 1376 if (match_int(&args[0], &option)) {
1377 status = 0;
1378 goto bail;
1379 }
1380 if (option < 0)
1381 return 0;
1382 if (option == 0)
2b4e30fb 1383 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1384 mopt->commit_interval = HZ * option;
1385 break;
2fbe8d1e 1386 case Opt_localalloc:
2fbe8d1e
SM		 1387 if (match_int(&args[0], &option)) {
1388 status = 0;
1389 goto bail;
1390 }
73c8a800 1391 if (option >= 0)
2fbe8d1e
SM		 1392 mopt->localalloc_opt = option;
1393 break;
53fc622b
MF		 1394 case Opt_localflocks:
1395 /*
1396 * Changing this during remount could race
1397 * flock() requests, or "unbalance" existing
1398 * ones (e.g., a lock is taken in one mode but
1399 * dropped in the other). If users care enough
1400 * to flip locking modes during remount, we
1401 * could add a "local" flag to individual
1402 * flock structures for proper tracking of
1403 * state.
1404 */
1405 if (!is_remount)
1406 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1407 break;
b61817e1
JB		 1408 case Opt_stack:
1409 /* Check both that the option we were passed
1410 * is of the right length and that it is a proper
1411 * string of the right length.
1412 */
1413 if (((args[0].to - args[0].from) !=
1414 OCFS2_STACK_LABEL_LEN) ||
1415 (strnlen(args[0].from,
1416 OCFS2_STACK_LABEL_LEN) !=
1417 OCFS2_STACK_LABEL_LEN)) {
1418 mlog(ML_ERROR,
1419 "Invalid cluster_stack option\n");
1420 status = 0;
1421 goto bail;
1422 }
1423 memcpy(mopt->cluster_stack, args[0].from,
1424 OCFS2_STACK_LABEL_LEN);
1425 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF		 1426 /*
1427 * Open code the memcmp here as we don't have
1428 * an osb to pass to
1429 * ocfs2_userspace_stack().
1430 */
1431 if (memcmp(mopt->cluster_stack,
1432 OCFS2_CLASSIC_CLUSTER_STACK,
1433 OCFS2_STACK_LABEL_LEN))
1434 user_stack = 1;
b61817e1 1435 break;
12462f1d
JB		 1436 case Opt_inode64:
1437 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1438 break;
19ece546 1439 case Opt_usrquota:
19ece546
JK		 1440 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1441 break;
1442 case Opt_grpquota:
19ece546
JK		 1443 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1444 break;
7bdb0d18
TY		 1445 case Opt_coherency_buffered:
1446 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1447 break;
1448 case Opt_coherency_full:
1449 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1450 break;
a68979b8
TY		 1451 case Opt_acl:
1452 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1453 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1454 break;
1455 case Opt_noacl:
5297aad8 1456 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1457 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1458 break;
d02f00cc
MF		 1459 case Opt_resv_level:
1460 if (is_remount)
1461 break;
1462 if (match_int(&args[0], &option)) {
1463 status = 0;
1464 goto bail;
1465 }
1466 if (option >= OCFS2_MIN_RESV_LEVEL &&
1467 option < OCFS2_MAX_RESV_LEVEL)
1468 mopt->resv_level = option;
1469 break;
83f92318
MF		 1470 case Opt_dir_resv_level:
1471 if (is_remount)
1472 break;
1473 if (match_int(&args[0], &option)) {
1474 status = 0;
1475 goto bail;
1476 }
1477 if (option >= OCFS2_MIN_RESV_LEVEL &&
1478 option < OCFS2_MAX_RESV_LEVEL)
1479 mopt->dir_resv_level = option;
1480 break;
1dfeb768 1481 case Opt_journal_async_commit:
1482 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1483 break;
ccd979bd
MF		 1484 default:
1485 mlog(ML_ERROR,
1486 "Unrecognized mount option \"%s\" "
1487 "or missing value\n", p);
1488 status = 0;
1489 goto bail;
1490 }
1491 }
1492
52c303c5
MF		 1493 if (user_stack == 0) {
1494 /* Ensure only one heartbeat mode */
1495 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1496 OCFS2_MOUNT_HB_GLOBAL |
1497 OCFS2_MOUNT_HB_NONE);
1498 if (hweight32(tmp) != 1) {
1499 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1500 status = 0;
1501 goto bail;
1502 }
2c442719
SM		 1503 }
1504
ccd979bd
MF		 1505 status = 1;
1506
1507bail:
ccd979bd
MF		 1508 return status;
1509}
1510
34c80b1d 1511static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1512{
34c80b1d 1513 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1514 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1515 unsigned int local_alloc_megs;
d550071c 1516
2c442719
SM		 1517 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1518 seq_printf(s, ",_netdev");
1519 if (opts & OCFS2_MOUNT_HB_LOCAL)
1520 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1521 else
1522 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1523 } else
1524 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM		 1525
1526 if (opts & OCFS2_MOUNT_NOINTR)
1527 seq_printf(s, ",nointr");
1528
1529 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1530 seq_printf(s, ",data=writeback");
1531 else
1532 seq_printf(s, ",data=ordered");
1533
1534 if (opts & OCFS2_MOUNT_BARRIER)
1535 seq_printf(s, ",barrier=1");
1536
1537 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1538 seq_printf(s, ",errors=panic");
7d0fb914
GR		 1539 else if (opts & OCFS2_MOUNT_ERRORS_CONT)
1540 seq_printf(s, ",errors=continue");
d550071c
SM		 1541 else
1542 seq_printf(s, ",errors=remount-ro");
1543
1544 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1545 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1546
34c80b1d 1547 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1548
d147b3d6
MF		 1549 if (osb->osb_commit_interval)
1550 seq_printf(s, ",commit=%u",
1551 (unsigned) (osb->osb_commit_interval / HZ));
1552
ebcee4b5 1553 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1554 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1555 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1556
53fc622b
MF		 1557 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1558 seq_printf(s, ",localflocks,");
1559
b61817e1 1560 if (osb->osb_cluster_stack[0])
a068acf2
KC		 1561 seq_show_option_n(s, "cluster_stack", osb->osb_cluster_stack,
1562 OCFS2_STACK_LABEL_LEN);
19ece546
JK		 1563 if (opts & OCFS2_MOUNT_USRQUOTA)
1564 seq_printf(s, ",usrquota");
1565 if (opts & OCFS2_MOUNT_GRPQUOTA)
1566 seq_printf(s, ",grpquota");
b61817e1 1567
7bdb0d18
TY		 1568 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1569 seq_printf(s, ",coherency=buffered");
1570 else
1571 seq_printf(s, ",coherency=full");
1572
b0f73cfc
SM		 1573 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1574 seq_printf(s, ",nouser_xattr");
1575 else
1576 seq_printf(s, ",user_xattr");
1577
12462f1d
JB		 1578 if (opts & OCFS2_MOUNT_INODE64)
1579 seq_printf(s, ",inode64");
1580
a68979b8
TY		 1581 if (opts & OCFS2_MOUNT_POSIX_ACL)
1582 seq_printf(s, ",acl");
1583 else
1584 seq_printf(s, ",noacl");
a68979b8 1585
d02f00cc
MF		 1586 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1587 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1588
83f92318
MF		 1589 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1590 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1591
1dfeb768 1592 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1593 seq_printf(s, ",journal_async_commit");
1594
d550071c
SM		 1595 return 0;
1596}
1597
ccd979bd
MF		 1598static int __init ocfs2_init(void)
1599{
c18ceab0 1600 int status;
a11f7e63 1601
ccd979bd 1602 status = init_ocfs2_uptodate_cache();
342827d7
AV		 1603 if (status < 0)
1604 goto out1;
ccd979bd
MF		 1605
1606 status = ocfs2_initialize_mem_caches();
342827d7
AV		 1607 if (status < 0)
1608 goto out2;
ccd979bd 1609
ccd979bd 1610 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
8f443e23
LT		 1611 if (!ocfs2_debugfs_root) {
1612 status = -ENOMEM;
ccd979bd 1613 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
35ddf78e 1614 goto out3;
ccd979bd
MF		 1615 }
1616
63e0c48a
JB		 1617 ocfs2_set_locking_protocol();
1618
9e33d69f 1619 status = register_quota_format(&ocfs2_quota_format);
342827d7 1620 if (status < 0)
35ddf78e 1621 goto out3;
342827d7
AV		 1622 status = register_filesystem(&ocfs2_fs_type);
1623 if (!status)
1624 return 0;
ccd979bd 1625
342827d7 1626 unregister_quota_format(&ocfs2_quota_format);
342827d7 1627out3:
35ddf78e 1628 debugfs_remove(ocfs2_debugfs_root);
342827d7
AV		 1629 ocfs2_free_mem_caches();
1630out2:
1631 exit_ocfs2_uptodate_cache();
1632out1:
1633 mlog_errno(status);
1634 return status;
ccd979bd
MF		 1635}
1636
1637static void __exit ocfs2_exit(void)
1638{
9e33d69f
JK		 1639 unregister_quota_format(&ocfs2_quota_format);
1640
ccd979bd
MF		 1641 debugfs_remove(ocfs2_debugfs_root);
1642
1643 ocfs2_free_mem_caches();
1644
1645 unregister_filesystem(&ocfs2_fs_type);
1646
ccd979bd 1647 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1648}
1649
1650static void ocfs2_put_super(struct super_block *sb)
1651{
32a42d39 1652 trace_ocfs2_put_super(sb);
ccd979bd
MF		 1653
1654 ocfs2_sync_blockdev(sb);
1655 ocfs2_dismount_volume(sb, 0);
a849d468 1656 ocfs2_filecheck_remove_sysfs(sb);
ccd979bd
MF		 1657}
1658
726c3342 1659static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1660{
1661 struct ocfs2_super *osb;
1662 u32 numbits, freebits;
1663 int status;
1664 struct ocfs2_dinode *bm_lock;
1665 struct buffer_head *bh = NULL;
1666 struct inode *inode = NULL;
1667
32a42d39 1668 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1669
726c3342 1670 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1671
1672 inode = ocfs2_get_system_file_inode(osb,
1673 GLOBAL_BITMAP_SYSTEM_INODE,
1674 OCFS2_INVALID_SLOT);
1675 if (!inode) {
1676 mlog(ML_ERROR, "failed to get bitmap inode\n");
1677 status = -EIO;
1678 goto bail;
1679 }
1680
e63aecb6 1681 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1682 if (status < 0) {
1683 mlog_errno(status);
1684 goto bail;
1685 }
1686
1687 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1688
1689 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1690 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1691
1692 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1693 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1694 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1695 buf->f_blocks = ((sector_t) numbits) *
1696 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1697 buf->f_bfree = ((sector_t) freebits) *
1698 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1699 buf->f_bavail = buf->f_bfree;
1700 buf->f_files = numbits;
1701 buf->f_ffree = freebits;
837711f8
CL		 1702 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1703 & 0xFFFFFFFFUL;
1704 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1705 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1706
1707 brelse(bh);
1708
e63aecb6 1709 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1710 status = 0;
1711bail:
72865d92 1712 iput(inode);
ccd979bd 1713
c1e8d35e
TM		 1714 if (status)
1715 mlog_errno(status);
ccd979bd
MF		 1716
1717 return status;
1718}
1719
51cc5068 1720static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1721{
1722 struct ocfs2_inode_info *oi = data;
1723
a35afb83
CL		 1724 oi->ip_flags = 0;
1725 oi->ip_open_count = 0;
1726 spin_lock_init(&oi->ip_lock);
1727 ocfs2_extent_map_init(&oi->vfs_inode);
1728 INIT_LIST_HEAD(&oi->ip_io_markers);
4506cfb6 1729 INIT_LIST_HEAD(&oi->ip_unwritten_list);
a35afb83 1730 oi->ip_dir_start_lookup = 0;
a35afb83 1731 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1732 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1733 mutex_init(&oi->ip_io_mutex);
ccd979bd 1734
a35afb83
CL		 1735 oi->ip_blkno = 0ULL;
1736 oi->ip_clusters = 0;
ccd979bd 1737
4fe370af
MF		 1738 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1739
a35afb83 1740 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1741 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1742 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1743
8cb471e8 1744 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1745 &ocfs2_inode_caching_ops);
ccd979bd 1746
a35afb83 1747 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1748}
1749
1750static int ocfs2_initialize_mem_caches(void)
1751{
1752 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1753 sizeof(struct ocfs2_inode_info),
1754 0,
1755 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
5d097056 1756 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
20c2df83 1757 ocfs2_inode_init_once);
9e33d69f
JK		 1758 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1759 sizeof(struct ocfs2_dquot),
1760 0,
1761 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1762 SLAB_MEM_SPREAD),
1763 NULL);
1764 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1765 sizeof(struct ocfs2_quota_chunk),
1766 0,
1767 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1768 NULL);
1769 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1770 !ocfs2_qf_chunk_cachep) {
1771 if (ocfs2_inode_cachep)
1772 kmem_cache_destroy(ocfs2_inode_cachep);
1773 if (ocfs2_dquot_cachep)
1774 kmem_cache_destroy(ocfs2_dquot_cachep);
1775 if (ocfs2_qf_chunk_cachep)
1776 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1777 return -ENOMEM;
9e33d69f 1778 }
ccd979bd 1779
ccd979bd
MF		 1780 return 0;
1781}
1782
1783static void ocfs2_free_mem_caches(void)
1784{
8c0a8537
KS		 1785 /*
1786 * Make sure all delayed rcu free inodes are flushed before we
1787 * destroy cache.
1788 */
1789 rcu_barrier();
ccd979bd
MF		 1790 if (ocfs2_inode_cachep)
1791 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1792 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1793
1794 if (ocfs2_dquot_cachep)
1795 kmem_cache_destroy(ocfs2_dquot_cachep);
1796 ocfs2_dquot_cachep = NULL;
1797
1798 if (ocfs2_qf_chunk_cachep)
1799 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1800 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1801}
1802
1803static int ocfs2_get_sector(struct super_block *sb,
1804 struct buffer_head **bh,
1805 int block,
1806 int sect_size)
1807{
1808 if (!sb_set_blocksize(sb, sect_size)) {
1809 mlog(ML_ERROR, "unable to set blocksize\n");
1810 return -EIO;
1811 }
1812
1813 *bh = sb_getblk(sb, block);
1814 if (!*bh) {
7391a294
RX		 1815 mlog_errno(-ENOMEM);
1816 return -ENOMEM;
ccd979bd
MF		 1817 }
1818 lock_buffer(*bh);
1819 if (!buffer_dirty(*bh))
1820 clear_buffer_uptodate(*bh);
1821 unlock_buffer(*bh);
1822 ll_rw_block(READ, 1, bh);
1823 wait_on_buffer(*bh);
28d57d43 1824 if (!buffer_uptodate(*bh)) {
1825 mlog_errno(-EIO);
1826 brelse(*bh);
1827 *bh = NULL;
1828 return -EIO;
1829 }
1830
ccd979bd
MF		 1831 return 0;
1832}
1833
ccd979bd
MF		 1834static int ocfs2_mount_volume(struct super_block *sb)
1835{
1836 int status = 0;
1837 int unlock_super = 0;
1838 struct ocfs2_super *osb = OCFS2_SB(sb);
1839
ccd979bd
MF		 1840 if (ocfs2_is_hard_readonly(osb))
1841 goto leave;
1842
ccd979bd
MF		 1843 status = ocfs2_dlm_init(osb);
1844 if (status < 0) {
1845 mlog_errno(status);
1846 goto leave;
1847 }
1848
ccd979bd
MF		 1849 status = ocfs2_super_lock(osb, 1);
1850 if (status < 0) {
1851 mlog_errno(status);
1852 goto leave;
1853 }
1854 unlock_super = 1;
1855
1856 /* This will load up the node map and add ourselves to it. */
1857 status = ocfs2_find_slot(osb);
1858 if (status < 0) {
1859 mlog_errno(status);
1860 goto leave;
1861 }
1862
ccd979bd
MF		 1863 /* load all node-local system inodes */
1864 status = ocfs2_init_local_system_inodes(osb);
1865 if (status < 0) {
1866 mlog_errno(status);
1867 goto leave;
1868 }
1869
1870 status = ocfs2_check_volume(osb);
1871 if (status < 0) {
1872 mlog_errno(status);
1873 goto leave;
1874 }
1875
1876 status = ocfs2_truncate_log_init(osb);
06c59bb8 1877 if (status < 0)
ccd979bd 1878 mlog_errno(status);
c271c5c2 1879
ccd979bd
MF		 1880leave:
1881 if (unlock_super)
1882 ocfs2_super_unlock(osb, 1);
1883
ccd979bd
MF		 1884 return status;
1885}
1886
ccd979bd
MF		 1887static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1888{
286eaa95 1889 int tmp, hangup_needed = 0;
ccd979bd 1890 struct ocfs2_super *osb = NULL;
99d7a882 1891 char nodestr[12];
ccd979bd 1892
32a42d39 1893 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF		 1894
1895 BUG_ON(!sb);
1896 osb = OCFS2_SB(sb);
1897 BUG_ON(!osb);
1898
50397507
SM		 1899 debugfs_remove(osb->osb_ctxt);
1900
692684e1
SM		 1901 /* Orphan scan should be stopped as early as possible */
1902 ocfs2_orphan_scan_stop(osb);
1903
19ece546
JK		 1904 ocfs2_disable_quotas(osb);
1905
e3a767b6
JK		 1906 /* All dquots should be freed by now */
1907 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1908 /* Wait for worker to be done with the work structure in osb */
1909 cancel_work_sync(&osb->dquot_drop_work);
1910
ccd979bd
MF		 1911 ocfs2_shutdown_local_alloc(osb);
1912
b253bfd8
X		 1913 ocfs2_truncate_log_shutdown(osb);
1914
553abd04
JB		 1915 /* This will disable recovery and flush any recovery work. */
1916 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1917
1918 ocfs2_journal_shutdown(osb);
1919
1920 ocfs2_sync_blockdev(sb);
1921
374a263e
TM		 1922 ocfs2_purge_refcount_trees(osb);
1923
4670c46d
JB		 1924 /* No cluster connection means we've failed during mount, so skip
1925 * all the steps which depended on that to complete. */
1926 if (osb->cconn) {
ccd979bd
MF		 1927 tmp = ocfs2_super_lock(osb, 1);
1928 if (tmp < 0) {
1929 mlog_errno(tmp);
1930 return;
1931 }
19b613d4 1932 }
ccd979bd 1933
19b613d4
MF		 1934 if (osb->slot_num != OCFS2_INVALID_SLOT)
1935 ocfs2_put_slot(osb);
ccd979bd 1936
4670c46d 1937 if (osb->cconn)
ccd979bd 1938 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1939
1940 ocfs2_release_system_inodes(osb);
1941
6953b4c0 1942 /*
6953b4c0
JB		 1943 * If we're dismounting due to mount error, mount.ocfs2 will clean
1944 * up heartbeat. If we're a local mount, there is no heartbeat.
1945 * If we failed before we got a uuid_str yet, we can't stop
1946 * heartbeat. Otherwise, do it.
1947 */
03efed8a
TY		 1948 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1949 !ocfs2_is_hard_readonly(osb))
286eaa95
JB		 1950 hangup_needed = 1;
1951
1952 if (osb->cconn)
1953 ocfs2_dlm_shutdown(osb, hangup_needed);
1954
73be192b 1955 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1956 debugfs_remove(osb->osb_debug_root);
1957
1958 if (hangup_needed)
6953b4c0 1959 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1960
1961 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1962
c271c5c2
SM		 1963 if (ocfs2_mount_local(osb))
1964 snprintf(nodestr, sizeof(nodestr), "local");
1965 else
19fdb624 1966 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1967
1968 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1969 osb->dev_str, nodestr);
ccd979bd
MF		 1970
1971 ocfs2_delete_osb(osb);
1972 kfree(osb);
1973 sb->s_dev = 0;
1974 sb->s_fs_info = NULL;
1975}
1976
1977static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1978 unsigned uuid_bytes)
1979{
1980 int i, ret;
1981 char *ptr;
1982
1983 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1984
cd861280 1985 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 1986 if (osb->uuid_str == NULL)
1987 return -ENOMEM;
1988
ccd979bd
MF		 1989 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1990 /* print with null */
1991 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1992 if (ret != 2) /* drop super cleans up */
1993 return -EINVAL;
1994 /* then only advance past the last char */
1995 ptr += 2;
1996 }
1997
1998 return 0;
1999}
2000
3bdb8efd
PL		 2001/* Make sure entire volume is addressable by our journal. Requires
2002 osb_clusters_at_boot to be valid and for the journal to have been
2003 initialized by ocfs2_journal_init(). */
2004static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2005{
2006 int status = 0;
2007 u64 max_block =
2008 ocfs2_clusters_to_blocks(osb->sb,
2009 osb->osb_clusters_at_boot) - 1;
2010
2011 /* 32-bit block number is always OK. */
2012 if (max_block <= (u32)~0ULL)
2013 goto out;
2014
2015 /* Volume is "huge", so see if our journal is new enough to
2016 support it. */
2017 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2018 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2019 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2020 JBD2_FEATURE_INCOMPAT_64BIT))) {
2021 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2022 "Enable the 'block64' journal option with tunefs.ocfs2");
2023 status = -EFBIG;
2024 goto out;
2025 }
2026
2027 out:
2028 return status;
2029}
2030
ccd979bd
MF		 2031static int ocfs2_initialize_super(struct super_block *sb,
2032 struct buffer_head *bh,
73be192b
JB		 2033 int sector_size,
2034 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2035{
bddb8eb3 2036 int status;
5a254031
MF		 2037 int i, cbits, bbits;
2038 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2039 struct inode *inode = NULL;
ccd979bd 2040 struct ocfs2_journal *journal;
ccd979bd 2041 struct ocfs2_super *osb;
3bdb8efd 2042 u64 total_blocks;
ccd979bd 2043
cd861280 2044 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 2045 if (!osb) {
2046 status = -ENOMEM;
2047 mlog_errno(status);
2048 goto bail;
2049 }
2050
2051 sb->s_fs_info = osb;
2052 sb->s_op = &ocfs2_sops;
ba87167c 2053 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2054 sb->s_export_op = &ocfs2_export_ops;
664dbd5f 2055 sb->s_qcop = &dquot_quotactl_sysfile_ops;
19ece546 2056 sb->dq_op = &ocfs2_quota_operations;
1c92ec67 2057 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cf1d6c76 2058 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2059 sb->s_time_gran = 1;
ccd979bd
MF		 2060 sb->s_flags |= MS_NOATIME;
2061 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2062 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2063 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2064 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
58be19dc
VD		 2065 memcpy(sb->s_uuid, di->id2.i_super.s_uuid,
2066 sizeof(di->id2.i_super.s_uuid));
ccd979bd 2067
9b7895ef
MF		 2068 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2069
2070 for (i = 0; i < 3; i++)
2071 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2072 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2073
ccd979bd
MF		 2074 osb->sb = sb;
2075 /* Save off for ocfs2_rw_direct */
2076 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2077 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2078
34d024f8
MF		 2079 spin_lock_init(&osb->dc_task_lock);
2080 init_waitqueue_head(&osb->dc_event);
2081 osb->dc_work_sequence = 0;
2082 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2083 INIT_LIST_HEAD(&osb->blocked_lock_list);
2084 osb->blocked_lock_count = 0;
ccd979bd 2085 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2086 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2087 ocfs2_init_steal_slots(osb);
ccd979bd 2088
43b10a20 2089 mutex_init(&osb->system_file_mutex);
2090
ccd979bd
MF		 2091 atomic_set(&osb->alloc_stats.moves, 0);
2092 atomic_set(&osb->alloc_stats.local_data, 0);
2093 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2094 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2095 atomic_set(&osb->alloc_stats.bg_extends, 0);
2096
73be192b
JB		 2097 /* Copy the blockcheck stats from the superblock probe */
2098 osb->osb_ecc_stats = *stats;
2099
ccd979bd
MF		 2100 ocfs2_init_node_maps(osb);
2101
2102 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2103 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2104
75d9bbc7
GR		 2105 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2106 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2107 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2108 osb->max_slots);
2109 status = -EINVAL;
2110 goto bail;
2111 }
75d9bbc7 2112
8b712cd5
JM		 2113 ocfs2_orphan_scan_init(osb);
2114
553abd04
JB		 2115 status = ocfs2_recovery_init(osb);
2116 if (status) {
2117 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2118 mlog_errno(status);
2119 goto bail;
2120 }
ccd979bd
MF		 2121
2122 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2123
7f1a37e3
TY		 2124 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2125
ccd979bd
MF		 2126 osb->slot_num = OCFS2_INVALID_SLOT;
2127
8154da3d
TY		 2128 osb->s_xattr_inline_size = le16_to_cpu(
2129 di->id2.i_super.s_xattr_inline_size);
fdd77704 2130
ccd979bd
MF		 2131 osb->local_alloc_state = OCFS2_LA_UNUSED;
2132 osb->local_alloc_bh = NULL;
9c7af40b 2133 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2134
ccd979bd
MF		 2135 init_waitqueue_head(&osb->osb_mount_event);
2136
d02f00cc
MF		 2137 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2138 if (status) {
2139 mlog_errno(status);
2140 goto bail;
2141 }
2142
ccd979bd
MF		 2143 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2144 if (!osb->vol_label) {
2145 mlog(ML_ERROR, "unable to alloc vol label\n");
2146 status = -ENOMEM;
2147 goto bail;
2148 }
2149
539d8264
SM		 2150 osb->slot_recovery_generations =
2151 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2152 GFP_KERNEL);
2153 if (!osb->slot_recovery_generations) {
2154 status = -ENOMEM;
2155 mlog_errno(status);
2156 goto bail;
2157 }
2158
b4df6ed8
MF		 2159 init_waitqueue_head(&osb->osb_wipe_event);
2160 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2161 sizeof(*osb->osb_orphan_wipes),
2162 GFP_KERNEL);
2163 if (!osb->osb_orphan_wipes) {
2164 status = -ENOMEM;
2165 mlog_errno(status);
2166 goto bail;
2167 }
2168
374a263e
TM		 2169 osb->osb_rf_lock_tree = RB_ROOT;
2170
ccd979bd
MF		 2171 osb->s_feature_compat =
2172 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2173 osb->s_feature_ro_compat =
2174 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2175 osb->s_feature_incompat =
2176 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2177
2178 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2179 mlog(ML_ERROR, "couldn't mount because of unsupported "
2180 "optional features (%x).\n", i);
2181 status = -EINVAL;
2182 goto bail;
2183 }
2184 if (!(osb->sb->s_flags & MS_RDONLY) &&
2185 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2186 mlog(ML_ERROR, "couldn't mount RDWR because of "
2187 "unsupported optional features (%x).\n", i);
2188 status = -EINVAL;
2189 goto bail;
2190 }
2191
98f486f2
SM		 2192 if (ocfs2_clusterinfo_valid(osb)) {
2193 osb->osb_stackflags =
2194 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2195 strlcpy(osb->osb_cluster_stack,
b61817e1 2196 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2197 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB		 2198 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2199 mlog(ML_ERROR,
2200 "couldn't mount because of an invalid "
2201 "cluster stack label (%s) \n",
2202 osb->osb_cluster_stack);
2203 status = -EINVAL;
2204 goto bail;
2205 }
c74a3bdd
GR		 2206 strlcpy(osb->osb_cluster_name,
2207 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2208 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB		 2209 } else {
2210 /* The empty string is identical with classic tools that
2211 * don't know about s_cluster_info. */
2212 osb->osb_cluster_stack[0] = '\0';
2213 }
2214
ccd979bd
MF		 2215 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2216
2217 /* FIXME
2218 * This should be done in ocfs2_journal_init(), but unknown
2219 * ordering issues will cause the filesystem to crash.
2220 * If anyone wants to figure out what part of the code
2221 * refers to osb->journal before ocfs2_journal_init() is run,
2222 * be my guest.
2223 */
2224 /* initialize our journal structure */
2225
cd861280 2226 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2227 if (!journal) {
2228 mlog(ML_ERROR, "unable to alloc journal\n");
2229 status = -ENOMEM;
2230 goto bail;
2231 }
2232 osb->journal = journal;
2233 journal->j_osb = osb;
2234
2235 atomic_set(&journal->j_num_trans, 0);
2236 init_rwsem(&journal->j_trans_barrier);
2237 init_waitqueue_head(&journal->j_checkpointed);
2238 spin_lock_init(&journal->j_lock);
2239 journal->j_trans_id = (unsigned long) 1;
2240 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2241 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2242 journal->j_state = OCFS2_JOURNAL_FREE;
2243
e3a767b6
JK		 2244 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2245 init_llist_head(&osb->dquot_drop_list);
2246
ccd979bd
MF		 2247 /* get some pseudo constants for clustersize bits */
2248 osb->s_clustersize_bits =
2249 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2250 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF		 2251
2252 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2253 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2254 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2255 osb->s_clustersize);
2256 status = -EINVAL;
2257 goto bail;
2258 }
2259
3bdb8efd
PL		 2260 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2261 le32_to_cpu(di->i_clusters));
2262
2263 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2264 total_blocks);
2265 if (status) {
2266 mlog(ML_ERROR, "Volume too large "
2267 "to mount safely on this system");
2268 status = -EFBIG;
ccd979bd
MF		 2269 goto bail;
2270 }
2271
2272 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2273 sizeof(di->id2.i_super.s_uuid))) {
2274 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2275 status = -ENOMEM;
2276 goto bail;
2277 }
2278
69201bb1
FF		 2279 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2280 OCFS2_MAX_VOL_LABEL_LEN);
ccd979bd
MF		 2281 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2282 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2283 osb->first_cluster_group_blkno =
2284 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2285 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2286 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM		 2287 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2288 (unsigned long long)osb->root_blkno,
2289 (unsigned long long)osb->system_dir_blkno,
2290 osb->s_clustersize_bits);
ccd979bd
MF		 2291
2292 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2293 if (!osb->osb_dlm_debug) {
2294 status = -ENOMEM;
2295 mlog_errno(status);
2296 goto bail;
2297 }
2298
2299 atomic_set(&osb->vol_state, VOLUME_INIT);
2300
2301 /* load root, system_dir, and all global system inodes */
2302 status = ocfs2_init_global_system_inodes(osb);
2303 if (status < 0) {
2304 mlog_errno(status);
2305 goto bail;
2306 }
2307
2308 /*
2309 * global bitmap
2310 */
2311 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2312 OCFS2_INVALID_SLOT);
2313 if (!inode) {
2314 status = -EINVAL;
2315 mlog_errno(status);
2316 goto bail;
2317 }
2318
2319 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2320 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2321 iput(inode);
ccd979bd 2322
8571882c
TM		 2323 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2324 osb->s_feature_incompat) * 8;
ccd979bd
MF		 2325
2326 status = ocfs2_init_slot_info(osb);
2327 if (status < 0) {
2328 mlog_errno(status);
2329 goto bail;
2330 }
9de16262 2331 cleancache_init_shared_fs(sb);
ccd979bd 2332
35ddf78e 2333 osb->ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
2334 if (!osb->ocfs2_wq) {
2335 status = -ENOMEM;
2336 mlog_errno(status);
2337 }
2338
ccd979bd 2339bail:
ccd979bd
MF		 2340 return status;
2341}
2342
2343/*
2344 * will return: -EAGAIN if it is ok to keep searching for superblocks
2345 * -EINVAL if there is a bad superblock
2346 * 0 on success
2347 */
2348static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2349 struct buffer_head *bh,
73be192b
JB		 2350 u32 blksz,
2351 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2352{
2353 int status = -EAGAIN;
2354
ccd979bd
MF		 2355 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2356 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2357 /* We have to do a raw check of the feature here */
2358 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2359 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2360 status = ocfs2_block_check_validate(bh->b_data,
2361 bh->b_size,
73be192b
JB		 2362 &di->i_check,
2363 stats);
d030cc97
JB		 2364 if (status)
2365 goto out;
2366 }
ccd979bd
MF		 2367 status = -EINVAL;
2368 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2369 mlog(ML_ERROR, "found superblock with incorrect block "
2370 "size: found %u, should be %u\n",
2371 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2372 blksz);
2373 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2374 OCFS2_MAJOR_REV_LEVEL ||
2375 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2376 OCFS2_MINOR_REV_LEVEL) {
2377 mlog(ML_ERROR, "found superblock with bad version: "
2378 "found %u.%u, should be %u.%u\n",
2379 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2380 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2381 OCFS2_MAJOR_REV_LEVEL,
2382 OCFS2_MINOR_REV_LEVEL);
2383 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2384 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2385 "found %llu, should be %llu\n",
1ca1a111 2386 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2387 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2388 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2389 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2390 mlog(ML_ERROR, "bad cluster size found: %u\n",
2391 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2392 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2393 mlog(ML_ERROR, "bad root_blkno: 0\n");
2394 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2395 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2396 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2397 mlog(ML_ERROR,
2398 "Superblock slots found greater than file system "
2399 "maximum: found %u, max %u\n",
2400 le16_to_cpu(di->id2.i_super.s_max_slots),
2401 OCFS2_MAX_SLOTS);
2402 } else {
2403 /* found it! */
2404 status = 0;
2405 }
2406 }
2407
d030cc97 2408out:
c1e8d35e
TM		 2409 if (status && status != -EAGAIN)
2410 mlog_errno(status);
ccd979bd
MF		 2411 return status;
2412}
2413
2414static int ocfs2_check_volume(struct ocfs2_super *osb)
2415{
bddb8eb3 2416 int status;
ccd979bd 2417 int dirty;
c271c5c2 2418 int local;
ccd979bd
MF		 2419 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2420 * recover
2421 * ourselves. */
2422
ccd979bd
MF		 2423 /* Init our journal object. */
2424 status = ocfs2_journal_init(osb->journal, &dirty);
2425 if (status < 0) {
2426 mlog(ML_ERROR, "Could not initialize journal!\n");
2427 goto finally;
2428 }
2429
3bdb8efd
PL		 2430 /* Now that journal has been initialized, check to make sure
2431 entire volume is addressable. */
2432 status = ocfs2_journal_addressable(osb);
2433 if (status)
2434 goto finally;
2435
ccd979bd
MF		 2436 /* If the journal was unmounted cleanly then we don't want to
2437 * recover anything. Otherwise, journal_load will do that
2438 * dirty work for us :) */
2439 if (!dirty) {
2440 status = ocfs2_journal_wipe(osb->journal, 0);
2441 if (status < 0) {
2442 mlog_errno(status);
2443 goto finally;
2444 }
2445 } else {
619c200d
SM		 2446 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2447 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF		 2448 }
2449
c271c5c2
SM		 2450 local = ocfs2_mount_local(osb);
2451
ccd979bd 2452 /* will play back anything left in the journal. */
539d8264 2453 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2454 if (status < 0) {
2455 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2456 goto finally;
2457 }
ccd979bd 2458
1dfeb768 2459 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2460 jbd2_journal_set_features(osb->journal->j_journal,
2461 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2462 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2463 else
2464 jbd2_journal_clear_features(osb->journal->j_journal,
2465 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2466 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2467
ccd979bd
MF		 2468 if (dirty) {
2469 /* recover my local alloc if we didn't unmount cleanly. */
2470 status = ocfs2_begin_local_alloc_recovery(osb,
2471 osb->slot_num,
2472 &local_alloc);
2473 if (status < 0) {
2474 mlog_errno(status);
2475 goto finally;
2476 }
2477 /* we complete the recovery process after we've marked
2478 * ourselves as mounted. */
2479 }
2480
ccd979bd
MF		 2481 status = ocfs2_load_local_alloc(osb);
2482 if (status < 0) {
2483 mlog_errno(status);
2484 goto finally;
2485 }
2486
2487 if (dirty) {
2488 /* Recovery will be completed after we've mounted the
2489 * rest of the volume. */
2490 osb->dirty = 1;
2491 osb->local_alloc_copy = local_alloc;
2492 local_alloc = NULL;
2493 }
2494
2495 /* go through each journal, trylock it and if you get the
2496 * lock, and it's marked as dirty, set the bit in the recover
2497 * map and launch a recovery thread for it. */
2498 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2499 if (status < 0) {
2500 mlog_errno(status);
2501 goto finally;
2502 }
2503
2504 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2505 if (status < 0)
2506 mlog_errno(status);
2507
2508finally:
d787ab09 2509 kfree(local_alloc);
ccd979bd 2510
c1e8d35e
TM		 2511 if (status)
2512 mlog_errno(status);
ccd979bd
MF		 2513 return status;
2514}
2515
2516/*
2517 * The routine gets called from dismount or close whenever a dismount on
2518 * volume is requested and the osb open count becomes 1.
2519 * It will remove the osb from the global list and also free up all the
2520 * initialized resources and fileobject.
2521 */
2522static void ocfs2_delete_osb(struct ocfs2_super *osb)
2523{
ccd979bd
MF		 2524 /* This function assumes that the caller has the main osb resource */
2525
35ddf78e 2526 /* ocfs2_initializer_super have already created this workqueue */
2527 if (osb->ocfs2_wq) {
2528 flush_workqueue(osb->ocfs2_wq);
2529 destroy_workqueue(osb->ocfs2_wq);
2530 }
2531
8e8a4603 2532 ocfs2_free_slot_info(osb);
ccd979bd 2533
b4df6ed8 2534 kfree(osb->osb_orphan_wipes);
539d8264 2535 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2536 /* FIXME
2537 * This belongs in journal shutdown, but because we have to
421f91d2 2538 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF		 2539 * we free it here.
2540 */
2541 kfree(osb->journal);
d787ab09 2542 kfree(osb->local_alloc_copy);
ccd979bd 2543 kfree(osb->uuid_str);
f13a568e 2544 kfree(osb->vol_label);
ccd979bd
MF		 2545 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2546 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF		 2547}
2548
7d0fb914
GR		 2549/* Depending on the mount option passed, perform one of the following:
2550 * Put OCFS2 into a readonly state (default)
2551 * Return EIO so that only the process errs
2552 * Fix the error as if fsck.ocfs2 -y
2553 * panic
2554 */
2555static int ocfs2_handle_error(struct super_block *sb)
ccd979bd
MF		 2556{
2557 struct ocfs2_super *osb = OCFS2_SB(sb);
7d0fb914 2558 int rv = 0;
ccd979bd
MF		 2559
2560 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
7d0fb914
GR		 2561 pr_crit("On-disk corruption discovered. "
2562 "Please run fsck.ocfs2 once the filesystem is unmounted.\n");
ccd979bd 2563
7d0fb914
GR		 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 } else if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_CONT) {
2568 pr_crit("OCFS2: Returning error to the calling process.\n");
2569 rv = -EIO;
2570 } else { /* default option */
2571 rv = -EROFS;
2572 if (sb->s_flags & MS_RDONLY &&
2573 (ocfs2_is_soft_readonly(osb) ||
2574 ocfs2_is_hard_readonly(osb)))
2575 return rv;
2576
2577 pr_crit("OCFS2: File system is now read-only.\n");
2578 sb->s_flags |= MS_RDONLY;
2579 ocfs2_set_ro_flag(osb, 0);
2580 }
2581
2582 return rv;
ccd979bd
MF		 2583}
2584
7d0fb914 2585int __ocfs2_error(struct super_block *sb, const char *function,
1543306e 2586 const char *fmt, ...)
ccd979bd 2587{
1543306e 2588 struct va_format vaf;
ccd979bd
MF		 2589 va_list args;
2590
2591 va_start(args, fmt);
1543306e
JP		 2592 vaf.fmt = fmt;
2593 vaf.va = &args;
ccd979bd
MF		 2594
2595 /* Not using mlog here because we want to show the actual
2596 * function the error came from. */
7ecef14a 2597 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %pV",
1543306e
JP		 2598 sb->s_id, function, &vaf);
2599
2600 va_end(args);
ccd979bd 2601
7d0fb914 2602 return ocfs2_handle_error(sb);
ccd979bd
MF		 2603}
2604
2605/* Handle critical errors. This is intentionally more drastic than
2606 * ocfs2_handle_error, so we only use for things like journal errors,
2607 * etc. */
1543306e 2608void __ocfs2_abort(struct super_block *sb, const char *function,
ccd979bd
MF		 2609 const char *fmt, ...)
2610{
1543306e 2611 struct va_format vaf;
ccd979bd
MF		 2612 va_list args;
2613
2614 va_start(args, fmt);
ccd979bd 2615
1543306e
JP		 2616 vaf.fmt = fmt;
2617 vaf.va = &args;
2618
7ecef14a 2619 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %pV",
1543306e
JP		 2620 sb->s_id, function, &vaf);
2621
2622 va_end(args);
ccd979bd
MF		 2623
2624 /* We don't have the cluster support yet to go straight to
2625 * hard readonly in here. Until then, we want to keep
2626 * ocfs2_abort() so that we can at least mark critical
2627 * errors.
2628 *
2629 * TODO: This should abort the journal and alert other nodes
2630 * that our slot needs recovery. */
2631
2632 /* Force a panic(). This stinks, but it's better than letting
2633 * things continue without having a proper hard readonly
2634 * here. */
a2f2ddbf
SM		 2635 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2636 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2637 ocfs2_handle_error(sb);
2638}
2639
e4b963f1
JB		 2640/*
2641 * Void signal blockers, because in-kernel sigprocmask() only fails
2642 * when SIG_* is wrong.
2643 */
2644void ocfs2_block_signals(sigset_t *oldset)
2645{
2646 int rc;
2647 sigset_t blocked;
2648
2649 sigfillset(&blocked);
2650 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2651 BUG_ON(rc);
2652}
2653
2654void ocfs2_unblock_signals(sigset_t *oldset)
2655{
2656 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2657 BUG_ON(rc);
2658}
2659
ccd979bd
MF		 2660module_init(ocfs2_init);
2661module_exit(ocfs2_exit);
Linux kernel - Deliverables
This page took 0.890378 seconds and 5 git commands to generate.