projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: add a mount option journal_async_commit on ocfs2 filesystem
[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"
77
1a5c4e2a 78static struct kmem_cache *ocfs2_inode_cachep;
9e33d69f
JK		 79struct kmem_cache *ocfs2_dquot_cachep;
80struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 81
25985edc 82/* OCFS2 needs to schedule several different types of work which
ccd979bd
MF		 83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86struct workqueue_struct *ocfs2_wq = NULL;
87
1a5c4e2a 88static struct dentry *ocfs2_debugfs_root;
ccd979bd
MF		 89
90MODULE_AUTHOR("Oracle");
91MODULE_LICENSE("GPL");
ff8fb335 92MODULE_DESCRIPTION("OCFS2 cluster file system");
ccd979bd 93
c0123ade
TY		 94struct mount_options
95{
d147b3d6 96 unsigned long commit_interval;
c0123ade
TY		 97 unsigned long mount_opt;
98 unsigned int atime_quantum;
99 signed short slot;
73c8a800 100 int localalloc_opt;
d02f00cc 101 unsigned int resv_level;
83f92318 102 int dir_resv_level;
b61817e1 103 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 104};
105
ccd979bd 106static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 107 struct mount_options *mopt,
baf4661a 108 int is_remount);
5297aad8
JK		 109static int ocfs2_check_set_options(struct super_block *sb,
110 struct mount_options *options);
34c80b1d 111static int ocfs2_show_options(struct seq_file *s, struct dentry *root);
ccd979bd
MF		 112static void ocfs2_put_super(struct super_block *sb);
113static int ocfs2_mount_volume(struct super_block *sb);
114static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
115static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
116static int ocfs2_initialize_mem_caches(void);
117static void ocfs2_free_mem_caches(void);
118static void ocfs2_delete_osb(struct ocfs2_super *osb);
119
726c3342 120static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 121
122static int ocfs2_sync_fs(struct super_block *sb, int wait);
123
124static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
125static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 126static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 127static int ocfs2_check_volume(struct ocfs2_super *osb);
128static int ocfs2_verify_volume(struct ocfs2_dinode *di,
129 struct buffer_head *bh,
73be192b
JB		 130 u32 sectsize,
131 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 132static int ocfs2_initialize_super(struct super_block *sb,
133 struct buffer_head *bh,
73be192b
JB		 134 int sector_size,
135 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 136static int ocfs2_get_sector(struct super_block *sb,
137 struct buffer_head **bh,
138 int block,
139 int sect_size);
ccd979bd
MF		 140static struct inode *ocfs2_alloc_inode(struct super_block *sb);
141static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 142static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
143static int ocfs2_enable_quotas(struct ocfs2_super *osb);
144static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 145
1c92ec67
JK		 146static struct dquot **ocfs2_get_dquots(struct inode *inode)
147{
148 return OCFS2_I(inode)->i_dquot;
149}
150
ee9b6d61 151static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 152 .statfs = ocfs2_statfs,
153 .alloc_inode = ocfs2_alloc_inode,
154 .destroy_inode = ocfs2_destroy_inode,
155 .drop_inode = ocfs2_drop_inode,
066d92dc 156 .evict_inode = ocfs2_evict_inode,
ccd979bd 157 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 158 .put_super = ocfs2_put_super,
159 .remount_fs = ocfs2_remount,
d550071c 160 .show_options = ocfs2_show_options,
9e33d69f
JK		 161 .quota_read = ocfs2_quota_read,
162 .quota_write = ocfs2_quota_write,
1c92ec67 163 .get_dquots = ocfs2_get_dquots,
ccd979bd
MF		 164};
165
166enum {
167 Opt_barrier,
168 Opt_err_panic,
169 Opt_err_ro,
170 Opt_intr,
171 Opt_nointr,
172 Opt_hb_none,
173 Opt_hb_local,
2c442719 174 Opt_hb_global,
ccd979bd
MF		 175 Opt_data_ordered,
176 Opt_data_writeback,
7f1a37e3 177 Opt_atime_quantum,
baf4661a 178 Opt_slot,
d147b3d6 179 Opt_commit,
2fbe8d1e 180 Opt_localalloc,
53fc622b 181 Opt_localflocks,
b61817e1 182 Opt_stack,
cf1d6c76
TY		 183 Opt_user_xattr,
184 Opt_nouser_xattr,
12462f1d 185 Opt_inode64,
a68979b8
TY		 186 Opt_acl,
187 Opt_noacl,
19ece546
JK		 188 Opt_usrquota,
189 Opt_grpquota,
7bdb0d18
TY		 190 Opt_coherency_buffered,
191 Opt_coherency_full,
d02f00cc 192 Opt_resv_level,
83f92318 193 Opt_dir_resv_level,
1dfeb768 194 Opt_journal_async_commit,
ccd979bd
MF		 195 Opt_err,
196};
197
a447c093 198static const match_table_t tokens = {
ccd979bd
MF		 199 {Opt_barrier, "barrier=%u"},
200 {Opt_err_panic, "errors=panic"},
201 {Opt_err_ro, "errors=remount-ro"},
202 {Opt_intr, "intr"},
203 {Opt_nointr, "nointr"},
204 {Opt_hb_none, OCFS2_HB_NONE},
205 {Opt_hb_local, OCFS2_HB_LOCAL},
2c442719 206 {Opt_hb_global, OCFS2_HB_GLOBAL},
ccd979bd
MF		 207 {Opt_data_ordered, "data=ordered"},
208 {Opt_data_writeback, "data=writeback"},
7f1a37e3 209 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 210 {Opt_slot, "preferred_slot=%u"},
d147b3d6 211 {Opt_commit, "commit=%u"},
2fbe8d1e 212 {Opt_localalloc, "localalloc=%d"},
53fc622b 213 {Opt_localflocks, "localflocks"},
b61817e1 214 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 215 {Opt_user_xattr, "user_xattr"},
216 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 217 {Opt_inode64, "inode64"},
a68979b8
TY		 218 {Opt_acl, "acl"},
219 {Opt_noacl, "noacl"},
19ece546
JK		 220 {Opt_usrquota, "usrquota"},
221 {Opt_grpquota, "grpquota"},
7bdb0d18
TY		 222 {Opt_coherency_buffered, "coherency=buffered"},
223 {Opt_coherency_full, "coherency=full"},
d02f00cc 224 {Opt_resv_level, "resv_level=%u"},
83f92318 225 {Opt_dir_resv_level, "dir_resv_level=%u"},
1dfeb768 226 {Opt_journal_async_commit, "journal_async_commit"},
ccd979bd
MF		 227 {Opt_err, NULL}
228};
229
50397507
SM		 230#ifdef CONFIG_DEBUG_FS
231static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
232{
50397507
SM		 233 struct ocfs2_cluster_connection *cconn = osb->cconn;
234 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 235 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
236 int i, out = 0;
50397507
SM		 237
238 out += snprintf(buf + out, len - out,
239 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
240 "Device", osb->dev_str, osb->uuid_str,
241 osb->fs_generation, osb->vol_label);
242
243 out += snprintf(buf + out, len - out,
244 "%10s => State: %d Flags: 0x%lX\n", "Volume",
245 atomic_read(&osb->vol_state), osb->osb_flags);
246
247 out += snprintf(buf + out, len - out,
248 "%10s => Block: %lu Cluster: %d\n", "Sizes",
249 osb->sb->s_blocksize, osb->s_clustersize);
250
251 out += snprintf(buf + out, len - out,
252 "%10s => Compat: 0x%X Incompat: 0x%X "
253 "ROcompat: 0x%X\n",
254 "Features", osb->s_feature_compat,
255 osb->s_feature_incompat, osb->s_feature_ro_compat);
256
257 out += snprintf(buf + out, len - out,
258 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
259 osb->s_mount_opt, osb->s_atime_quantum);
260
c3d38840
SM		 261 if (cconn) {
262 out += snprintf(buf + out, len - out,
263 "%10s => Stack: %s Name: %*s "
264 "Version: %d.%d\n", "Cluster",
265 (*osb->osb_cluster_stack == '\0' ?
266 "o2cb" : osb->osb_cluster_stack),
267 cconn->cc_namelen, cconn->cc_name,
268 cconn->cc_version.pv_major,
269 cconn->cc_version.pv_minor);
270 }
50397507
SM		 271
272 spin_lock(&osb->dc_task_lock);
273 out += snprintf(buf + out, len - out,
274 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
275 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 276 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
277 osb->blocked_lock_count, osb->dc_wake_sequence,
278 osb->dc_work_sequence);
50397507
SM		 279 spin_unlock(&osb->dc_task_lock);
280
281 spin_lock(&osb->osb_lock);
282 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
283 "Recovery",
284 (osb->recovery_thread_task ?
285 task_pid_nr(osb->recovery_thread_task) : -1));
286 if (rm->rm_used == 0)
287 out += snprintf(buf + out, len - out, " None\n");
288 else {
289 for (i = 0; i < rm->rm_used; i++)
290 out += snprintf(buf + out, len - out, " %d",
291 rm->rm_entries[i]);
292 out += snprintf(buf + out, len - out, "\n");
293 }
294 spin_unlock(&osb->osb_lock);
295
296 out += snprintf(buf + out, len - out,
8fa9d17f 297 "%10s => Pid: %d Interval: %lu\n", "Commit",
c3d38840 298 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
8fa9d17f 299 osb->osb_commit_interval);
50397507
SM		 300
301 out += snprintf(buf + out, len - out,
c3d38840 302 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 303 "Journal", osb->journal->j_state,
c3d38840
SM		 304 osb->journal->j_trans_id,
305 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 306
307 out += snprintf(buf + out, len - out,
308 "%10s => GlobalAllocs: %d LocalAllocs: %d "
309 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
310 "Stats",
311 atomic_read(&osb->alloc_stats.bitmap_data),
312 atomic_read(&osb->alloc_stats.local_data),
313 atomic_read(&osb->alloc_stats.bg_allocs),
314 atomic_read(&osb->alloc_stats.moves),
315 atomic_read(&osb->alloc_stats.bg_extends));
316
317 out += snprintf(buf + out, len - out,
318 "%10s => State: %u Descriptor: %llu Size: %u bits "
319 "Default: %u bits\n",
320 "LocalAlloc", osb->local_alloc_state,
321 (unsigned long long)osb->la_last_gd,
322 osb->local_alloc_bits, osb->local_alloc_default_bits);
323
324 spin_lock(&osb->osb_lock);
325 out += snprintf(buf + out, len - out,
b89c5428
TY		 326 "%10s => InodeSlot: %d StolenInodes: %d, "
327 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 328 osb->s_inode_steal_slot,
b89c5428
TY		 329 atomic_read(&osb->s_num_inodes_stolen),
330 osb->s_meta_steal_slot,
331 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 332 spin_unlock(&osb->osb_lock);
333
df152c24
SM		 334 out += snprintf(buf + out, len - out, "OrphanScan => ");
335 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
336 os->os_count, os->os_seqno);
337 out += snprintf(buf + out, len - out, " Last Scan: ");
338 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
339 out += snprintf(buf + out, len - out, "Disabled\n");
340 else
341 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
342 (get_seconds() - os->os_scantime.tv_sec));
343
50397507
SM		 344 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
345 "Slots", "Num", "RecoGen");
50397507
SM		 346 for (i = 0; i < osb->max_slots; ++i) {
347 out += snprintf(buf + out, len - out,
348 "%10s %c %3d %10d\n",
349 " ",
350 (i == osb->slot_num ? '*' : ' '),
351 i, osb->slot_recovery_generations[i]);
352 }
353
354 return out;
355}
356
357static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
358{
359 struct ocfs2_super *osb = inode->i_private;
360 char *buf = NULL;
361
362 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
363 if (!buf)
364 goto bail;
365
366 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
367
368 file->private_data = buf;
369
370 return 0;
371bail:
372 return -ENOMEM;
373}
374
375static int ocfs2_debug_release(struct inode *inode, struct file *file)
376{
377 kfree(file->private_data);
378 return 0;
379}
380
381static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
382 size_t nbytes, loff_t *ppos)
383{
384 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
385 i_size_read(file->f_mapping->host));
386}
387#else
388static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
389{
390 return 0;
391}
392static int ocfs2_debug_release(struct inode *inode, struct file *file)
393{
394 return 0;
395}
396static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
397 size_t nbytes, loff_t *ppos)
398{
399 return 0;
400}
401#endif /* CONFIG_DEBUG_FS */
402
828c0950 403static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 404 .open = ocfs2_osb_debug_open,
405 .release = ocfs2_debug_release,
406 .read = ocfs2_debug_read,
407 .llseek = generic_file_llseek,
408};
409
ccd979bd
MF		 410static int ocfs2_sync_fs(struct super_block *sb, int wait)
411{
bddb8eb3 412 int status;
ccd979bd
MF		 413 tid_t target;
414 struct ocfs2_super *osb = OCFS2_SB(sb);
415
ccd979bd
MF		 416 if (ocfs2_is_hard_readonly(osb))
417 return -EROFS;
418
419 if (wait) {
420 status = ocfs2_flush_truncate_log(osb);
421 if (status < 0)
422 mlog_errno(status);
423 } else {
424 ocfs2_schedule_truncate_log_flush(osb, 0);
425 }
426
2b4e30fb
JB		 427 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
428 &target)) {
ccd979bd 429 if (wait)
2b4e30fb
JB		 430 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
431 target);
ccd979bd
MF		 432 }
433 return 0;
434}
435
1a224ad1
JK		 436static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
437{
438 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
439 && (ino == USER_QUOTA_SYSTEM_INODE
440 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
441 return 0;
442 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
443 && (ino == GROUP_QUOTA_SYSTEM_INODE
444 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
445 return 0;
446 return 1;
447}
448
ccd979bd
MF		 449static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
450{
451 struct inode *new = NULL;
452 int status = 0;
453 int i;
454
5fa0613e 455 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 456 if (IS_ERR(new)) {
457 status = PTR_ERR(new);
458 mlog_errno(status);
459 goto bail;
460 }
461 osb->root_inode = new;
462
5fa0613e 463 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 464 if (IS_ERR(new)) {
465 status = PTR_ERR(new);
466 mlog_errno(status);
467 goto bail;
468 }
469 osb->sys_root_inode = new;
470
471 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
472 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 473 if (!ocfs2_need_system_inode(osb, i))
474 continue;
ccd979bd
MF		 475 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
476 if (!new) {
477 ocfs2_release_system_inodes(osb);
478 status = -EINVAL;
479 mlog_errno(status);
480 /* FIXME: Should ERROR_RO_FS */
481 mlog(ML_ERROR, "Unable to load system inode %d, "
482 "possibly corrupt fs?", i);
483 goto bail;
484 }
485 // the array now has one ref, so drop this one
486 iput(new);
487 }
488
489bail:
c1e8d35e
TM		 490 if (status)
491 mlog_errno(status);
ccd979bd
MF		 492 return status;
493}
494
495static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
496{
497 struct inode *new = NULL;
498 int status = 0;
499 int i;
500
ccd979bd
MF		 501 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
502 i < NUM_SYSTEM_INODES;
503 i++) {
1a224ad1
JK		 504 if (!ocfs2_need_system_inode(osb, i))
505 continue;
ccd979bd
MF		 506 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
507 if (!new) {
508 ocfs2_release_system_inodes(osb);
509 status = -EINVAL;
510 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
511 status, i, osb->slot_num);
512 goto bail;
513 }
514 /* the array now has one ref, so drop this one */
515 iput(new);
516 }
517
518bail:
c1e8d35e
TM		 519 if (status)
520 mlog_errno(status);
ccd979bd
MF		 521 return status;
522}
523
bddb8eb3 524static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 525{
bddb8eb3 526 int i;
ccd979bd
MF		 527 struct inode *inode;
528
b4d693fc
TM		 529 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
530 inode = osb->global_system_inodes[i];
ccd979bd
MF		 531 if (inode) {
532 iput(inode);
b4d693fc 533 osb->global_system_inodes[i] = NULL;
ccd979bd
MF		 534 }
535 }
536
537 inode = osb->sys_root_inode;
538 if (inode) {
539 iput(inode);
540 osb->sys_root_inode = NULL;
541 }
542
543 inode = osb->root_inode;
544 if (inode) {
545 iput(inode);
546 osb->root_inode = NULL;
547 }
548
b4d693fc 549 if (!osb->local_system_inodes)
c1e8d35e 550 return;
b4d693fc
TM		 551
552 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
553 if (osb->local_system_inodes[i]) {
554 iput(osb->local_system_inodes[i]);
555 osb->local_system_inodes[i] = NULL;
556 }
557 }
558
559 kfree(osb->local_system_inodes);
560 osb->local_system_inodes = NULL;
ccd979bd
MF		 561}
562
563/* We're allocating fs objects, use GFP_NOFS */
564static struct inode *ocfs2_alloc_inode(struct super_block *sb)
565{
566 struct ocfs2_inode_info *oi;
567
e6b4f8da 568 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 569 if (!oi)
570 return NULL;
571
2931cdcb
DW		 572 oi->i_sync_tid = 0;
573 oi->i_datasync_tid = 0;
1c92ec67 574 memset(&oi->i_dquot, 0, sizeof(oi->i_dquot));
2931cdcb 575
2b4e30fb 576 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 577 return &oi->vfs_inode;
578}
579
fa0d7e3d 580static void ocfs2_i_callback(struct rcu_head *head)
ccd979bd 581{
fa0d7e3d 582 struct inode *inode = container_of(head, struct inode, i_rcu);
ccd979bd
MF		 583 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
584}
585
fa0d7e3d
NP		 586static void ocfs2_destroy_inode(struct inode *inode)
587{
588 call_rcu(&inode->i_rcu, ocfs2_i_callback);
589}
590
5a254031
MF		 591static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
592 unsigned int cbits)
ccd979bd 593{
5a254031
MF		 594 unsigned int bytes = 1 << cbits;
595 unsigned int trim = bytes;
596 unsigned int bitshift = 32;
597
598 /*
599 * i_size and all block offsets in ocfs2 are always 64 bits
600 * wide. i_clusters is 32 bits, in cluster-sized units. So on
601 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 602 */
603
604#if BITS_PER_LONG == 32
90c699a9 605# if defined(CONFIG_LBDAF)
2ecd05ae 606 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 607 /*
608 * We might be limited by page cache size.
609 */
610 if (bytes > PAGE_CACHE_SIZE) {
611 bytes = PAGE_CACHE_SIZE;
612 trim = 1;
613 /*
614 * Shift by 31 here so that we don't get larger than
615 * MAX_LFS_FILESIZE
616 */
617 bitshift = 31;
618 }
ccd979bd 619# else
5a254031
MF		 620 /*
621 * We are limited by the size of sector_t. Use block size, as
622 * that's what we expose to the VFS.
623 */
624 bytes = 1 << bbits;
625 trim = 1;
626 bitshift = 31;
ccd979bd
MF		 627# endif
628#endif
629
5a254031
MF		 630 /*
631 * Trim by a whole cluster when we can actually approach the
632 * on-disk limits. Otherwise we can overflow i_clusters when
633 * an extent start is at the max offset.
634 */
635 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 636}
637
638static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
639{
640 int incompat_features;
641 int ret = 0;
c0123ade 642 struct mount_options parsed_options;
ccd979bd 643 struct ocfs2_super *osb = OCFS2_SB(sb);
2c442719 644 u32 tmp;
ccd979bd 645
02b9984d
TT		 646 sync_filesystem(sb);
647
5297aad8
JK		 648 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
649 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF		 650 ret = -EINVAL;
651 goto out;
652 }
653
2c442719
SM		 654 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
655 OCFS2_MOUNT_HB_NONE;
656 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
ccd979bd
MF		 657 ret = -EINVAL;
658 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
659 goto out;
660 }
661
662 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 663 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 664 ret = -EINVAL;
665 mlog(ML_ERROR, "Cannot change data mode on remount\n");
666 goto out;
667 }
668
12462f1d
JB		 669 /* Probably don't want this on remount; it might
670 * mess with other nodes */
671 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
672 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
673 ret = -EINVAL;
674 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
675 goto out;
676 }
677
ccd979bd
MF		 678 /* We're going to/from readonly mode. */
679 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 680 /* Disable quota accounting before remounting RO */
681 if (*flags & MS_RDONLY) {
682 ret = ocfs2_susp_quotas(osb, 0);
683 if (ret < 0)
684 goto out;
685 }
ccd979bd
MF		 686 /* Lock here so the check of HARD_RO and the potential
687 * setting of SOFT_RO is atomic. */
688 spin_lock(&osb->osb_lock);
689 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
690 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
691 ret = -EROFS;
692 goto unlock_osb;
693 }
694
695 if (*flags & MS_RDONLY) {
ccd979bd
MF		 696 sb->s_flags |= MS_RDONLY;
697 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
698 } else {
ccd979bd
MF		 699 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
700 mlog(ML_ERROR, "Cannot remount RDWR "
701 "filesystem due to previous errors.\n");
702 ret = -EROFS;
703 goto unlock_osb;
704 }
705 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
706 if (incompat_features) {
707 mlog(ML_ERROR, "Cannot remount RDWR because "
708 "of unsupported optional features "
709 "(%x).\n", incompat_features);
710 ret = -EINVAL;
711 goto unlock_osb;
712 }
713 sb->s_flags &= ~MS_RDONLY;
714 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
715 }
32a42d39 716 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
ccd979bd
MF		 717unlock_osb:
718 spin_unlock(&osb->osb_lock);
19ece546
JK		 719 /* Enable quota accounting after remounting RW */
720 if (!ret && !(*flags & MS_RDONLY)) {
721 if (sb_any_quota_suspended(sb))
722 ret = ocfs2_susp_quotas(osb, 1);
723 else
724 ret = ocfs2_enable_quotas(osb);
725 if (ret < 0) {
726 /* Return back changes... */
727 spin_lock(&osb->osb_lock);
728 sb->s_flags |= MS_RDONLY;
729 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
730 spin_unlock(&osb->osb_lock);
731 goto out;
732 }
733 }
ccd979bd
MF		 734 }
735
736 if (!ret) {
ccd979bd
MF		 737 /* Only save off the new mount options in case of a successful
738 * remount. */
c0123ade
TY		 739 osb->s_mount_opt = parsed_options.mount_opt;
740 osb->s_atime_quantum = parsed_options.atime_quantum;
741 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 742 if (parsed_options.commit_interval)
743 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 744
745 if (!ocfs2_is_hard_readonly(osb))
746 ocfs2_set_journal_params(osb);
57b09bb5
JK		 747
748 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
749 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
750 MS_POSIXACL : 0);
ccd979bd
MF		 751 }
752out:
753 return ret;
754}
755
756static int ocfs2_sb_probe(struct super_block *sb,
757 struct buffer_head **bh,
73be192b
JB		 758 int *sector_size,
759 struct ocfs2_blockcheck_stats *stats)
ccd979bd 760{
bddb8eb3 761 int status, tmpstat;
ccd979bd
MF		 762 struct ocfs1_vol_disk_hdr *hdr;
763 struct ocfs2_dinode *di;
764 int blksize;
765
766 *bh = NULL;
767
768 /* may be > 512 */
e1defc4f 769 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 770 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
771 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
772 *sector_size, OCFS2_MAX_BLOCKSIZE);
773 status = -EINVAL;
774 goto bail;
775 }
776
777 /* Can this really happen? */
778 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
779 *sector_size = OCFS2_MIN_BLOCKSIZE;
780
781 /* check block zero for old format */
782 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
783 if (status < 0) {
784 mlog_errno(status);
785 goto bail;
786 }
787 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
788 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
789 mlog(ML_ERROR, "incompatible version: %u.%u\n",
790 hdr->major_version, hdr->minor_version);
791 status = -EINVAL;
792 }
793 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
794 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
795 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
796 hdr->signature);
797 status = -EINVAL;
798 }
799 brelse(*bh);
800 *bh = NULL;
801 if (status < 0) {
802 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
803 "upgraded before mounting with ocfs v2\n");
804 goto bail;
805 }
806
807 /*
808 * Now check at magic offset for 512, 1024, 2048, 4096
809 * blocksizes. 4096 is the maximum blocksize because it is
810 * the minimum clustersize.
811 */
812 status = -EINVAL;
813 for (blksize = *sector_size;
814 blksize <= OCFS2_MAX_BLOCKSIZE;
815 blksize <<= 1) {
816 tmpstat = ocfs2_get_sector(sb, bh,
817 OCFS2_SUPER_BLOCK_BLKNO,
818 blksize);
819 if (tmpstat < 0) {
820 status = tmpstat;
821 mlog_errno(status);
fb5cbe9e 822 break;
ccd979bd
MF		 823 }
824 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 825 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 826 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB		 827 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
828 if (tmpstat < 0) {
829 brelse(*bh);
830 *bh = NULL;
831 }
832 if (tmpstat != -EAGAIN) {
833 status = tmpstat;
ccd979bd 834 break;
fb5cbe9e 835 }
ccd979bd
MF		 836 }
837
838bail:
839 return status;
840}
841
c271c5c2
SM		 842static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
843{
2c442719
SM		 844 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
845
846 if (osb->s_mount_opt & hb_enabled) {
847 if (ocfs2_mount_local(osb)) {
c271c5c2
SM		 848 mlog(ML_ERROR, "Cannot heartbeat on a locally "
849 "mounted device.\n");
850 return -EINVAL;
851 }
2c442719 852 if (ocfs2_userspace_stack(osb)) {
b61817e1
JB		 853 mlog(ML_ERROR, "Userspace stack expected, but "
854 "o2cb heartbeat arguments passed to mount\n");
855 return -EINVAL;
856 }
2c442719
SM		 857 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
858 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
859 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
860 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
861 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
862 return -EINVAL;
863 }
b61817e1
JB		 864 }
865
2c442719 866 if (!(osb->s_mount_opt & hb_enabled)) {
b61817e1
JB		 867 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
868 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 869 mlog(ML_ERROR, "Heartbeat has to be started to mount "
870 "a read-write clustered device.\n");
871 return -EINVAL;
872 }
873 }
874
875 return 0;
876}
877
b61817e1
JB		 878/*
879 * If we're using a userspace stack, mount should have passed
880 * a name that matches the disk. If not, mount should not
881 * have passed a stack.
882 */
883static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
884 struct mount_options *mopt)
885{
886 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
887 mlog(ML_ERROR,
888 "cluster stack passed to mount, but this filesystem "
889 "does not support it\n");
890 return -EINVAL;
891 }
892
893 if (ocfs2_userspace_stack(osb) &&
894 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
895 OCFS2_STACK_LABEL_LEN)) {
896 mlog(ML_ERROR,
897 "cluster stack passed to mount (\"%s\") does not "
898 "match the filesystem (\"%s\")\n",
899 mopt->cluster_stack,
900 osb->osb_cluster_stack);
901 return -EINVAL;
902 }
903
904 return 0;
905}
906
19ece546
JK		 907static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
908{
909 int type;
910 struct super_block *sb = osb->sb;
52362810
JK		 911 unsigned int feature[OCFS2_MAXQUOTAS] = {
912 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
913 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK		 914 int status = 0;
915
52362810 916 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 917 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
918 continue;
919 if (unsuspend)
0f0dd62f 920 status = dquot_resume(sb, type);
eea7feb0
JK		 921 else {
922 struct ocfs2_mem_dqinfo *oinfo;
923
924 /* Cancel periodic syncing before suspending */
925 oinfo = sb_dqinfo(sb, type)->dqi_priv;
926 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
0f0dd62f 927 status = dquot_suspend(sb, type);
eea7feb0 928 }
19ece546
JK		 929 if (status < 0)
930 break;
931 }
932 if (status < 0)
933 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
934 "remount (error = %d).\n", status);
935 return status;
936}
937
938static int ocfs2_enable_quotas(struct ocfs2_super *osb)
939{
52362810 940 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
19ece546 941 struct super_block *sb = osb->sb;
52362810
JK		 942 unsigned int feature[OCFS2_MAXQUOTAS] = {
943 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
944 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
945 unsigned int ino[OCFS2_MAXQUOTAS] = {
946 LOCAL_USER_QUOTA_SYSTEM_INODE,
19ece546
JK		 947 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
948 int status;
949 int type;
950
951 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
52362810 952 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 953 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
954 continue;
955 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
956 osb->slot_num);
957 if (!inode[type]) {
958 status = -ENOENT;
959 goto out_quota_off;
960 }
287a8095
CH		 961 status = dquot_enable(inode[type], type, QFMT_OCFS2,
962 DQUOT_USAGE_ENABLED);
19ece546
JK		 963 if (status < 0)
964 goto out_quota_off;
965 }
966
52362810 967 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 968 iput(inode[type]);
969 return 0;
970out_quota_off:
971 ocfs2_disable_quotas(osb);
52362810 972 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK		 973 iput(inode[type]);
974 mlog_errno(status);
975 return status;
976}
977
978static void ocfs2_disable_quotas(struct ocfs2_super *osb)
979{
980 int type;
981 struct inode *inode;
982 struct super_block *sb = osb->sb;
c06bcbfa 983 struct ocfs2_mem_dqinfo *oinfo;
19ece546
JK		 984
985 /* We mostly ignore errors in this function because there's not much
986 * we can do when we see them */
52362810 987 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK		 988 if (!sb_has_quota_loaded(sb, type))
989 continue;
c06bcbfa
JK		 990 /* Cancel periodic syncing before we grab dqonoff_mutex */
991 oinfo = sb_dqinfo(sb, type)->dqi_priv;
992 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK		 993 inode = igrab(sb->s_dquot.files[type]);
994 /* Turn off quotas. This will remove all dquot structures from
995 * memory and so they will be automatically synced to global
996 * quota files */
0f0dd62f
CH		 997 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
998 DQUOT_LIMITS_ENABLED);
19ece546
JK		 999 if (!inode)
1000 continue;
1001 iput(inode);
1002 }
1003}
1004
1005/* Handle quota on quotactl */
f00c9e44 1006static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
19ece546 1007{
52362810
JK		 1008 unsigned int feature[OCFS2_MAXQUOTAS] = {
1009 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
1010 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK		 1011
1012 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1013 return -EINVAL;
1014
287a8095
CH		 1015 return dquot_enable(sb_dqopt(sb)->files[type], type,
1016 format_id, DQUOT_LIMITS_ENABLED);
19ece546
JK		 1017}
1018
1019/* Handle quota off quotactl */
307ae18a 1020static int ocfs2_quota_off(struct super_block *sb, int type)
19ece546 1021{
0f0dd62f 1022 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
19ece546
JK		 1023}
1024
0d54b217 1025static const struct quotactl_ops ocfs2_quotactl_ops = {
f00c9e44 1026 .quota_on_meta = ocfs2_quota_on,
19ece546 1027 .quota_off = ocfs2_quota_off,
287a8095
CH		 1028 .quota_sync = dquot_quota_sync,
1029 .get_info = dquot_get_dqinfo,
1030 .set_info = dquot_set_dqinfo,
1031 .get_dqblk = dquot_get_dqblk,
1032 .set_dqblk = dquot_set_dqblk,
19ece546
JK		 1033};
1034
ccd979bd
MF		 1035static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1036{
1037 struct dentry *root;
1038 int status, sector_size;
c0123ade 1039 struct mount_options parsed_options;
ccd979bd
MF		 1040 struct inode *inode = NULL;
1041 struct ocfs2_super *osb = NULL;
1042 struct buffer_head *bh = NULL;
49fa8140 1043 char nodestr[12];
73be192b 1044 struct ocfs2_blockcheck_stats stats;
ccd979bd 1045
32a42d39 1046 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1047
c0123ade 1048 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1049 status = -EINVAL;
1050 goto read_super_error;
1051 }
1052
1053 /* probe for superblock */
73be192b 1054 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1055 if (status < 0) {
1056 mlog(ML_ERROR, "superblock probe failed!\n");
1057 goto read_super_error;
1058 }
1059
73be192b 1060 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1061 osb = OCFS2_SB(sb);
1062 if (status < 0) {
1063 mlog_errno(status);
1064 goto read_super_error;
1065 }
1066 brelse(bh);
1067 bh = NULL;
a68979b8 1068
5297aad8
JK		 1069 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1070 status = -EINVAL;
1071 goto read_super_error;
1072 }
c0123ade
TY		 1073 osb->s_mount_opt = parsed_options.mount_opt;
1074 osb->s_atime_quantum = parsed_options.atime_quantum;
1075 osb->preferred_slot = parsed_options.slot;
d147b3d6 1076 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1077
1078 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1079 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF		 1080 osb->osb_dir_resv_level = parsed_options.resv_level;
1081 if (parsed_options.dir_resv_level == -1)
1082 osb->osb_dir_resv_level = parsed_options.resv_level;
1083 else
1084 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1085
b61817e1
JB		 1086 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1087 if (status)
1088 goto read_super_error;
1089
ccd979bd
MF		 1090 sb->s_magic = OCFS2_SUPER_MAGIC;
1091
9e1f1de0 1092 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
a68979b8
TY		 1093 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1094
ccd979bd
MF		 1095 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1096 * heartbeat=none */
1097 if (bdev_read_only(sb->s_bdev)) {
1098 if (!(sb->s_flags & MS_RDONLY)) {
1099 status = -EACCES;
1100 mlog(ML_ERROR, "Readonly device detected but readonly "
1101 "mount was not specified.\n");
1102 goto read_super_error;
1103 }
1104
1105 /* You should not be able to start a local heartbeat
1106 * on a readonly device. */
1107 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1108 status = -EROFS;
1109 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1110 "device.\n");
1111 goto read_super_error;
1112 }
1113
1114 status = ocfs2_check_journals_nolocks(osb);
1115 if (status < 0) {
1116 if (status == -EROFS)
1117 mlog(ML_ERROR, "Recovery required on readonly "
1118 "file system, but write access is "
1119 "unavailable.\n");
1120 else
2bd63216 1121 mlog_errno(status);
ccd979bd
MF		 1122 goto read_super_error;
1123 }
1124
1125 ocfs2_set_ro_flag(osb, 1);
1126
619c200d
SM		 1127 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1128 "Cluster services will not be used for this mount. "
1129 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF		 1130 }
1131
1132 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1133 if (sb->s_flags & MS_RDONLY)
1134 ocfs2_set_ro_flag(osb, 0);
1135 }
1136
c271c5c2
SM		 1137 status = ocfs2_verify_heartbeat(osb);
1138 if (status < 0) {
1139 mlog_errno(status);
1140 goto read_super_error;
1141 }
1142
ccd979bd
MF		 1143 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1144 ocfs2_debugfs_root);
1145 if (!osb->osb_debug_root) {
1146 status = -EINVAL;
1147 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1148 goto read_super_error;
1149 }
1150
50397507
SM		 1151 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1152 osb->osb_debug_root,
1153 osb,
1154 &ocfs2_osb_debug_fops);
1155 if (!osb->osb_ctxt) {
1156 status = -EINVAL;
1157 mlog_errno(status);
1158 goto read_super_error;
1159 }
1160
73be192b
JB		 1161 if (ocfs2_meta_ecc(osb)) {
1162 status = ocfs2_blockcheck_stats_debugfs_install(
1163 &osb->osb_ecc_stats,
1164 osb->osb_debug_root);
1165 if (status) {
1166 mlog(ML_ERROR,
1167 "Unable to create blockcheck statistics "
1168 "files\n");
1169 goto read_super_error;
1170 }
1171 }
1172
ccd979bd 1173 status = ocfs2_mount_volume(sb);
ccd979bd
MF		 1174 if (status < 0)
1175 goto read_super_error;
1176
be0d93f0
AV		 1177 if (osb->root_inode)
1178 inode = igrab(osb->root_inode);
1179
ccd979bd
MF		 1180 if (!inode) {
1181 status = -EIO;
1182 mlog_errno(status);
1183 goto read_super_error;
1184 }
1185
48fde701 1186 root = d_make_root(inode);
ccd979bd
MF		 1187 if (!root) {
1188 status = -ENOMEM;
1189 mlog_errno(status);
1190 goto read_super_error;
1191 }
1192
1193 sb->s_root = root;
1194
1195 ocfs2_complete_mount_recovery(osb);
1196
c271c5c2
SM		 1197 if (ocfs2_mount_local(osb))
1198 snprintf(nodestr, sizeof(nodestr), "local");
1199 else
19fdb624 1200 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1201
1202 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1203 "with %s data mode.\n",
c271c5c2 1204 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1205 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1206 "ordered");
1207
1208 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1209 wake_up(&osb->osb_mount_event);
1210
19ece546
JK		 1211 /* Now we can initialize quotas because we can afford to wait
1212 * for cluster locks recovery now. That also means that truncation
1213 * log recovery can happen but that waits for proper quota setup */
1214 if (!(sb->s_flags & MS_RDONLY)) {
1215 status = ocfs2_enable_quotas(osb);
1216 if (status < 0) {
1217 /* We have to err-out specially here because
1218 * s_root is already set */
1219 mlog_errno(status);
1220 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1221 wake_up(&osb->osb_mount_event);
19ece546
JK		 1222 return status;
1223 }
1224 }
1225
1226 ocfs2_complete_quota_recovery(osb);
1227
1228 /* Now we wake up again for processes waiting for quotas */
1229 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1230 wake_up(&osb->osb_mount_event);
1231
df152c24 1232 /* Start this when the mount is almost sure of being successful */
8b712cd5 1233 ocfs2_orphan_scan_start(osb);
df152c24 1234
ccd979bd
MF		 1235 return status;
1236
1237read_super_error:
a81cb88b 1238 brelse(bh);
ccd979bd 1239
ccd979bd
MF		 1240 if (osb) {
1241 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1242 wake_up(&osb->osb_mount_event);
1243 ocfs2_dismount_volume(sb, 1);
1244 }
1245
c1e8d35e
TM		 1246 if (status)
1247 mlog_errno(status);
ccd979bd
MF		 1248 return status;
1249}
1250
152a0836 1251static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH		 1252 int flags,
1253 const char *dev_name,
152a0836 1254 void *data)
ccd979bd 1255{
152a0836 1256 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF		 1257}
1258
1259static struct file_system_type ocfs2_fs_type = {
1260 .owner = THIS_MODULE,
1261 .name = "ocfs2",
152a0836 1262 .mount = ocfs2_mount,
8ed6b237 1263 .kill_sb = kill_block_super,
1ba9da2f 1264 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1265 .next = NULL
1266};
91417705 1267MODULE_ALIAS_FS("ocfs2");
ccd979bd 1268
5297aad8
JK		 1269static int ocfs2_check_set_options(struct super_block *sb,
1270 struct mount_options *options)
1271{
1272 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1273 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1274 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1275 mlog(ML_ERROR, "User quotas were requested, but this "
1276 "filesystem does not have the feature enabled.\n");
1277 return 0;
1278 }
1279 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1280 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1281 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1282 mlog(ML_ERROR, "Group quotas were requested, but this "
1283 "filesystem does not have the feature enabled.\n");
1284 return 0;
1285 }
1286 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1287 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1288 mlog(ML_ERROR, "ACL support requested but extended attributes "
1289 "feature is not enabled\n");
1290 return 0;
1291 }
1292 /* No ACL setting specified? Use XATTR feature... */
1293 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1294 OCFS2_MOUNT_NO_POSIX_ACL))) {
1295 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1296 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1297 else
1298 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1299 }
1300 return 1;
1301}
1302
ccd979bd
MF		 1303static int ocfs2_parse_options(struct super_block *sb,
1304 char *options,
c0123ade 1305 struct mount_options *mopt,
ccd979bd
MF		 1306 int is_remount)
1307{
52c303c5 1308 int status, user_stack = 0;
ccd979bd 1309 char *p;
2c442719 1310 u32 tmp;
ccd979bd 1311
32a42d39 1312 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1313
d147b3d6 1314 mopt->commit_interval = 0;
4b37fcb7 1315 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY		 1316 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1317 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1318 mopt->localalloc_opt = -1;
b61817e1 1319 mopt->cluster_stack[0] = '\0';
d02f00cc 1320 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1321 mopt->dir_resv_level = -1;
ccd979bd
MF		 1322
1323 if (!options) {
1324 status = 1;
1325 goto bail;
1326 }
1327
1328 while ((p = strsep(&options, ",")) != NULL) {
1329 int token, option;
1330 substring_t args[MAX_OPT_ARGS];
1331
1332 if (!*p)
1333 continue;
1334
1335 token = match_token(p, tokens, args);
1336 switch (token) {
1337 case Opt_hb_local:
c0123ade 1338 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1339 break;
1340 case Opt_hb_none:
2c442719
SM		 1341 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1342 break;
1343 case Opt_hb_global:
1344 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF		 1345 break;
1346 case Opt_barrier:
1347 if (match_int(&args[0], &option)) {
1348 status = 0;
1349 goto bail;
1350 }
1351 if (option)
c0123ade 1352 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1353 else
c0123ade 1354 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1355 break;
1356 case Opt_intr:
c0123ade 1357 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1358 break;
1359 case Opt_nointr:
c0123ade 1360 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1361 break;
1362 case Opt_err_panic:
c0123ade 1363 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1364 break;
1365 case Opt_err_ro:
c0123ade 1366 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1367 break;
1368 case Opt_data_ordered:
c0123ade 1369 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1370 break;
1371 case Opt_data_writeback:
c0123ade 1372 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1373 break;
cf1d6c76
TY		 1374 case Opt_user_xattr:
1375 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1376 break;
1377 case Opt_nouser_xattr:
1378 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1379 break;
7f1a37e3
TY		 1380 case Opt_atime_quantum:
1381 if (match_int(&args[0], &option)) {
1382 status = 0;
1383 goto bail;
1384 }
1385 if (option >= 0)
c0123ade 1386 mopt->atime_quantum = option;
7f1a37e3 1387 break;
baf4661a
SM		 1388 case Opt_slot:
1389 option = 0;
1390 if (match_int(&args[0], &option)) {
1391 status = 0;
1392 goto bail;
1393 }
1394 if (option)
c0123ade 1395 mopt->slot = (s16)option;
baf4661a 1396 break;
d147b3d6
MF		 1397 case Opt_commit:
1398 option = 0;
1399 if (match_int(&args[0], &option)) {
1400 status = 0;
1401 goto bail;
1402 }
1403 if (option < 0)
1404 return 0;
1405 if (option == 0)
2b4e30fb 1406 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1407 mopt->commit_interval = HZ * option;
1408 break;
2fbe8d1e
SM		 1409 case Opt_localalloc:
1410 option = 0;
1411 if (match_int(&args[0], &option)) {
1412 status = 0;
1413 goto bail;
1414 }
73c8a800 1415 if (option >= 0)
2fbe8d1e
SM		 1416 mopt->localalloc_opt = option;
1417 break;
53fc622b
MF		 1418 case Opt_localflocks:
1419 /*
1420 * Changing this during remount could race
1421 * flock() requests, or "unbalance" existing
1422 * ones (e.g., a lock is taken in one mode but
1423 * dropped in the other). If users care enough
1424 * to flip locking modes during remount, we
1425 * could add a "local" flag to individual
1426 * flock structures for proper tracking of
1427 * state.
1428 */
1429 if (!is_remount)
1430 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1431 break;
b61817e1
JB		 1432 case Opt_stack:
1433 /* Check both that the option we were passed
1434 * is of the right length and that it is a proper
1435 * string of the right length.
1436 */
1437 if (((args[0].to - args[0].from) !=
1438 OCFS2_STACK_LABEL_LEN) ||
1439 (strnlen(args[0].from,
1440 OCFS2_STACK_LABEL_LEN) !=
1441 OCFS2_STACK_LABEL_LEN)) {
1442 mlog(ML_ERROR,
1443 "Invalid cluster_stack option\n");
1444 status = 0;
1445 goto bail;
1446 }
1447 memcpy(mopt->cluster_stack, args[0].from,
1448 OCFS2_STACK_LABEL_LEN);
1449 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF		 1450 /*
1451 * Open code the memcmp here as we don't have
1452 * an osb to pass to
1453 * ocfs2_userspace_stack().
1454 */
1455 if (memcmp(mopt->cluster_stack,
1456 OCFS2_CLASSIC_CLUSTER_STACK,
1457 OCFS2_STACK_LABEL_LEN))
1458 user_stack = 1;
b61817e1 1459 break;
12462f1d
JB		 1460 case Opt_inode64:
1461 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1462 break;
19ece546 1463 case Opt_usrquota:
19ece546
JK		 1464 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1465 break;
1466 case Opt_grpquota:
19ece546
JK		 1467 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1468 break;
7bdb0d18
TY		 1469 case Opt_coherency_buffered:
1470 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1471 break;
1472 case Opt_coherency_full:
1473 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1474 break;
a68979b8
TY		 1475 case Opt_acl:
1476 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1477 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1478 break;
1479 case Opt_noacl:
5297aad8 1480 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1481 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1482 break;
d02f00cc
MF		 1483 case Opt_resv_level:
1484 if (is_remount)
1485 break;
1486 if (match_int(&args[0], &option)) {
1487 status = 0;
1488 goto bail;
1489 }
1490 if (option >= OCFS2_MIN_RESV_LEVEL &&
1491 option < OCFS2_MAX_RESV_LEVEL)
1492 mopt->resv_level = option;
1493 break;
83f92318
MF		 1494 case Opt_dir_resv_level:
1495 if (is_remount)
1496 break;
1497 if (match_int(&args[0], &option)) {
1498 status = 0;
1499 goto bail;
1500 }
1501 if (option >= OCFS2_MIN_RESV_LEVEL &&
1502 option < OCFS2_MAX_RESV_LEVEL)
1503 mopt->dir_resv_level = option;
1504 break;
1dfeb768 1505 case Opt_journal_async_commit:
1506 mopt->mount_opt |= OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT;
1507 break;
ccd979bd
MF		 1508 default:
1509 mlog(ML_ERROR,
1510 "Unrecognized mount option \"%s\" "
1511 "or missing value\n", p);
1512 status = 0;
1513 goto bail;
1514 }
1515 }
1516
52c303c5
MF		 1517 if (user_stack == 0) {
1518 /* Ensure only one heartbeat mode */
1519 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1520 OCFS2_MOUNT_HB_GLOBAL |
1521 OCFS2_MOUNT_HB_NONE);
1522 if (hweight32(tmp) != 1) {
1523 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1524 status = 0;
1525 goto bail;
1526 }
2c442719
SM		 1527 }
1528
ccd979bd
MF		 1529 status = 1;
1530
1531bail:
ccd979bd
MF		 1532 return status;
1533}
1534
34c80b1d 1535static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1536{
34c80b1d 1537 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1538 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1539 unsigned int local_alloc_megs;
d550071c 1540
2c442719
SM		 1541 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1542 seq_printf(s, ",_netdev");
1543 if (opts & OCFS2_MOUNT_HB_LOCAL)
1544 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1545 else
1546 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1547 } else
1548 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM		 1549
1550 if (opts & OCFS2_MOUNT_NOINTR)
1551 seq_printf(s, ",nointr");
1552
1553 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1554 seq_printf(s, ",data=writeback");
1555 else
1556 seq_printf(s, ",data=ordered");
1557
1558 if (opts & OCFS2_MOUNT_BARRIER)
1559 seq_printf(s, ",barrier=1");
1560
1561 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1562 seq_printf(s, ",errors=panic");
1563 else
1564 seq_printf(s, ",errors=remount-ro");
1565
1566 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1567 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1568
34c80b1d 1569 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1570
d147b3d6
MF		 1571 if (osb->osb_commit_interval)
1572 seq_printf(s, ",commit=%u",
1573 (unsigned) (osb->osb_commit_interval / HZ));
1574
ebcee4b5 1575 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1576 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1577 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1578
53fc622b
MF		 1579 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1580 seq_printf(s, ",localflocks,");
1581
b61817e1
JB		 1582 if (osb->osb_cluster_stack[0])
1583 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1584 osb->osb_cluster_stack);
19ece546
JK		 1585 if (opts & OCFS2_MOUNT_USRQUOTA)
1586 seq_printf(s, ",usrquota");
1587 if (opts & OCFS2_MOUNT_GRPQUOTA)
1588 seq_printf(s, ",grpquota");
b61817e1 1589
7bdb0d18
TY		 1590 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1591 seq_printf(s, ",coherency=buffered");
1592 else
1593 seq_printf(s, ",coherency=full");
1594
b0f73cfc
SM		 1595 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1596 seq_printf(s, ",nouser_xattr");
1597 else
1598 seq_printf(s, ",user_xattr");
1599
12462f1d
JB		 1600 if (opts & OCFS2_MOUNT_INODE64)
1601 seq_printf(s, ",inode64");
1602
a68979b8
TY		 1603 if (opts & OCFS2_MOUNT_POSIX_ACL)
1604 seq_printf(s, ",acl");
1605 else
1606 seq_printf(s, ",noacl");
a68979b8 1607
d02f00cc
MF		 1608 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1609 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1610
83f92318
MF		 1611 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1612 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1613
1dfeb768 1614 if (opts & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
1615 seq_printf(s, ",journal_async_commit");
1616
d550071c
SM		 1617 return 0;
1618}
1619
ccd979bd
MF		 1620static int __init ocfs2_init(void)
1621{
c18ceab0 1622 int status;
a11f7e63 1623
ccd979bd 1624 status = init_ocfs2_uptodate_cache();
342827d7
AV		 1625 if (status < 0)
1626 goto out1;
ccd979bd
MF		 1627
1628 status = ocfs2_initialize_mem_caches();
342827d7
AV		 1629 if (status < 0)
1630 goto out2;
ccd979bd
MF		 1631
1632 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1633 if (!ocfs2_wq) {
1634 status = -ENOMEM;
342827d7 1635 goto out3;
ccd979bd
MF		 1636 }
1637
ccd979bd
MF		 1638 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1639 if (!ocfs2_debugfs_root) {
dc171580 1640 status = -ENOMEM;
ccd979bd 1641 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
dc171580 1642 goto out4;
ccd979bd
MF		 1643 }
1644
63e0c48a
JB		 1645 ocfs2_set_locking_protocol();
1646
9e33d69f 1647 status = register_quota_format(&ocfs2_quota_format);
342827d7
AV		 1648 if (status < 0)
1649 goto out4;
1650 status = register_filesystem(&ocfs2_fs_type);
1651 if (!status)
1652 return 0;
ccd979bd 1653
342827d7
AV		 1654 unregister_quota_format(&ocfs2_quota_format);
1655out4:
1656 destroy_workqueue(ocfs2_wq);
1657 debugfs_remove(ocfs2_debugfs_root);
1658out3:
1659 ocfs2_free_mem_caches();
1660out2:
1661 exit_ocfs2_uptodate_cache();
1662out1:
1663 mlog_errno(status);
1664 return status;
ccd979bd
MF		 1665}
1666
1667static void __exit ocfs2_exit(void)
1668{
ccd979bd
MF		 1669 if (ocfs2_wq) {
1670 flush_workqueue(ocfs2_wq);
1671 destroy_workqueue(ocfs2_wq);
1672 }
1673
9e33d69f
JK		 1674 unregister_quota_format(&ocfs2_quota_format);
1675
ccd979bd
MF		 1676 debugfs_remove(ocfs2_debugfs_root);
1677
1678 ocfs2_free_mem_caches();
1679
1680 unregister_filesystem(&ocfs2_fs_type);
1681
ccd979bd 1682 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1683}
1684
1685static void ocfs2_put_super(struct super_block *sb)
1686{
32a42d39 1687 trace_ocfs2_put_super(sb);
ccd979bd
MF		 1688
1689 ocfs2_sync_blockdev(sb);
1690 ocfs2_dismount_volume(sb, 0);
ccd979bd
MF		 1691}
1692
726c3342 1693static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1694{
1695 struct ocfs2_super *osb;
1696 u32 numbits, freebits;
1697 int status;
1698 struct ocfs2_dinode *bm_lock;
1699 struct buffer_head *bh = NULL;
1700 struct inode *inode = NULL;
1701
32a42d39 1702 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1703
726c3342 1704 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1705
1706 inode = ocfs2_get_system_file_inode(osb,
1707 GLOBAL_BITMAP_SYSTEM_INODE,
1708 OCFS2_INVALID_SLOT);
1709 if (!inode) {
1710 mlog(ML_ERROR, "failed to get bitmap inode\n");
1711 status = -EIO;
1712 goto bail;
1713 }
1714
e63aecb6 1715 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1716 if (status < 0) {
1717 mlog_errno(status);
1718 goto bail;
1719 }
1720
1721 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1722
1723 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1724 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1725
1726 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1727 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1728 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1729 buf->f_blocks = ((sector_t) numbits) *
1730 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1731 buf->f_bfree = ((sector_t) freebits) *
1732 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1733 buf->f_bavail = buf->f_bfree;
1734 buf->f_files = numbits;
1735 buf->f_ffree = freebits;
837711f8
CL		 1736 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1737 & 0xFFFFFFFFUL;
1738 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1739 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1740
1741 brelse(bh);
1742
e63aecb6 1743 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1744 status = 0;
1745bail:
1746 if (inode)
1747 iput(inode);
1748
c1e8d35e
TM		 1749 if (status)
1750 mlog_errno(status);
ccd979bd
MF		 1751
1752 return status;
1753}
1754
51cc5068 1755static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1756{
1757 struct ocfs2_inode_info *oi = data;
1758
a35afb83
CL		 1759 oi->ip_flags = 0;
1760 oi->ip_open_count = 0;
1761 spin_lock_init(&oi->ip_lock);
1762 ocfs2_extent_map_init(&oi->vfs_inode);
1763 INIT_LIST_HEAD(&oi->ip_io_markers);
a35afb83 1764 oi->ip_dir_start_lookup = 0;
c18ceab0 1765 mutex_init(&oi->ip_unaligned_aio);
a35afb83 1766 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1767 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1768 mutex_init(&oi->ip_io_mutex);
ccd979bd 1769
a35afb83
CL		 1770 oi->ip_blkno = 0ULL;
1771 oi->ip_clusters = 0;
ccd979bd 1772
4fe370af
MF		 1773 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1774
a35afb83 1775 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1776 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1777 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1778
8cb471e8 1779 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1780 &ocfs2_inode_caching_ops);
ccd979bd 1781
a35afb83 1782 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1783}
1784
1785static int ocfs2_initialize_mem_caches(void)
1786{
1787 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1788 sizeof(struct ocfs2_inode_info),
1789 0,
1790 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1791 SLAB_MEM_SPREAD),
20c2df83 1792 ocfs2_inode_init_once);
9e33d69f
JK		 1793 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1794 sizeof(struct ocfs2_dquot),
1795 0,
1796 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1797 SLAB_MEM_SPREAD),
1798 NULL);
1799 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1800 sizeof(struct ocfs2_quota_chunk),
1801 0,
1802 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1803 NULL);
1804 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1805 !ocfs2_qf_chunk_cachep) {
1806 if (ocfs2_inode_cachep)
1807 kmem_cache_destroy(ocfs2_inode_cachep);
1808 if (ocfs2_dquot_cachep)
1809 kmem_cache_destroy(ocfs2_dquot_cachep);
1810 if (ocfs2_qf_chunk_cachep)
1811 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1812 return -ENOMEM;
9e33d69f 1813 }
ccd979bd 1814
ccd979bd
MF		 1815 return 0;
1816}
1817
1818static void ocfs2_free_mem_caches(void)
1819{
8c0a8537
KS		 1820 /*
1821 * Make sure all delayed rcu free inodes are flushed before we
1822 * destroy cache.
1823 */
1824 rcu_barrier();
ccd979bd
MF		 1825 if (ocfs2_inode_cachep)
1826 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1827 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1828
1829 if (ocfs2_dquot_cachep)
1830 kmem_cache_destroy(ocfs2_dquot_cachep);
1831 ocfs2_dquot_cachep = NULL;
1832
1833 if (ocfs2_qf_chunk_cachep)
1834 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1835 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1836}
1837
1838static int ocfs2_get_sector(struct super_block *sb,
1839 struct buffer_head **bh,
1840 int block,
1841 int sect_size)
1842{
1843 if (!sb_set_blocksize(sb, sect_size)) {
1844 mlog(ML_ERROR, "unable to set blocksize\n");
1845 return -EIO;
1846 }
1847
1848 *bh = sb_getblk(sb, block);
1849 if (!*bh) {
7391a294
RX		 1850 mlog_errno(-ENOMEM);
1851 return -ENOMEM;
ccd979bd
MF		 1852 }
1853 lock_buffer(*bh);
1854 if (!buffer_dirty(*bh))
1855 clear_buffer_uptodate(*bh);
1856 unlock_buffer(*bh);
1857 ll_rw_block(READ, 1, bh);
1858 wait_on_buffer(*bh);
28d57d43 1859 if (!buffer_uptodate(*bh)) {
1860 mlog_errno(-EIO);
1861 brelse(*bh);
1862 *bh = NULL;
1863 return -EIO;
1864 }
1865
ccd979bd
MF		 1866 return 0;
1867}
1868
ccd979bd
MF		 1869static int ocfs2_mount_volume(struct super_block *sb)
1870{
1871 int status = 0;
1872 int unlock_super = 0;
1873 struct ocfs2_super *osb = OCFS2_SB(sb);
1874
ccd979bd
MF		 1875 if (ocfs2_is_hard_readonly(osb))
1876 goto leave;
1877
ccd979bd
MF		 1878 status = ocfs2_dlm_init(osb);
1879 if (status < 0) {
1880 mlog_errno(status);
1881 goto leave;
1882 }
1883
ccd979bd
MF		 1884 status = ocfs2_super_lock(osb, 1);
1885 if (status < 0) {
1886 mlog_errno(status);
1887 goto leave;
1888 }
1889 unlock_super = 1;
1890
1891 /* This will load up the node map and add ourselves to it. */
1892 status = ocfs2_find_slot(osb);
1893 if (status < 0) {
1894 mlog_errno(status);
1895 goto leave;
1896 }
1897
ccd979bd
MF		 1898 /* load all node-local system inodes */
1899 status = ocfs2_init_local_system_inodes(osb);
1900 if (status < 0) {
1901 mlog_errno(status);
1902 goto leave;
1903 }
1904
1905 status = ocfs2_check_volume(osb);
1906 if (status < 0) {
1907 mlog_errno(status);
1908 goto leave;
1909 }
1910
1911 status = ocfs2_truncate_log_init(osb);
06c59bb8 1912 if (status < 0)
ccd979bd 1913 mlog_errno(status);
c271c5c2 1914
ccd979bd
MF		 1915leave:
1916 if (unlock_super)
1917 ocfs2_super_unlock(osb, 1);
1918
ccd979bd
MF		 1919 return status;
1920}
1921
ccd979bd
MF		 1922static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1923{
286eaa95 1924 int tmp, hangup_needed = 0;
ccd979bd 1925 struct ocfs2_super *osb = NULL;
99d7a882 1926 char nodestr[12];
ccd979bd 1927
32a42d39 1928 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF		 1929
1930 BUG_ON(!sb);
1931 osb = OCFS2_SB(sb);
1932 BUG_ON(!osb);
1933
50397507
SM		 1934 debugfs_remove(osb->osb_ctxt);
1935
692684e1
SM		 1936 /* Orphan scan should be stopped as early as possible */
1937 ocfs2_orphan_scan_stop(osb);
1938
19ece546
JK		 1939 ocfs2_disable_quotas(osb);
1940
e3a767b6
JK		 1941 /* All dquots should be freed by now */
1942 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1943 /* Wait for worker to be done with the work structure in osb */
1944 cancel_work_sync(&osb->dquot_drop_work);
1945
ccd979bd
MF		 1946 ocfs2_shutdown_local_alloc(osb);
1947
b253bfd8
X		 1948 ocfs2_truncate_log_shutdown(osb);
1949
553abd04
JB		 1950 /* This will disable recovery and flush any recovery work. */
1951 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1952
1953 ocfs2_journal_shutdown(osb);
1954
1955 ocfs2_sync_blockdev(sb);
1956
374a263e
TM		 1957 ocfs2_purge_refcount_trees(osb);
1958
4670c46d
JB		 1959 /* No cluster connection means we've failed during mount, so skip
1960 * all the steps which depended on that to complete. */
1961 if (osb->cconn) {
ccd979bd
MF		 1962 tmp = ocfs2_super_lock(osb, 1);
1963 if (tmp < 0) {
1964 mlog_errno(tmp);
1965 return;
1966 }
19b613d4 1967 }
ccd979bd 1968
19b613d4
MF		 1969 if (osb->slot_num != OCFS2_INVALID_SLOT)
1970 ocfs2_put_slot(osb);
ccd979bd 1971
4670c46d 1972 if (osb->cconn)
ccd979bd 1973 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1974
1975 ocfs2_release_system_inodes(osb);
1976
6953b4c0 1977 /*
6953b4c0
JB		 1978 * If we're dismounting due to mount error, mount.ocfs2 will clean
1979 * up heartbeat. If we're a local mount, there is no heartbeat.
1980 * If we failed before we got a uuid_str yet, we can't stop
1981 * heartbeat. Otherwise, do it.
1982 */
03efed8a
TY		 1983 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1984 !ocfs2_is_hard_readonly(osb))
286eaa95
JB		 1985 hangup_needed = 1;
1986
1987 if (osb->cconn)
1988 ocfs2_dlm_shutdown(osb, hangup_needed);
1989
73be192b 1990 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1991 debugfs_remove(osb->osb_debug_root);
1992
1993 if (hangup_needed)
6953b4c0 1994 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1995
1996 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1997
c271c5c2
SM		 1998 if (ocfs2_mount_local(osb))
1999 snprintf(nodestr, sizeof(nodestr), "local");
2000 else
19fdb624 2001 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 2002
2003 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
2004 osb->dev_str, nodestr);
ccd979bd
MF		 2005
2006 ocfs2_delete_osb(osb);
2007 kfree(osb);
2008 sb->s_dev = 0;
2009 sb->s_fs_info = NULL;
2010}
2011
2012static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
2013 unsigned uuid_bytes)
2014{
2015 int i, ret;
2016 char *ptr;
2017
2018 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2019
cd861280 2020 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 2021 if (osb->uuid_str == NULL)
2022 return -ENOMEM;
2023
ccd979bd
MF		 2024 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2025 /* print with null */
2026 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2027 if (ret != 2) /* drop super cleans up */
2028 return -EINVAL;
2029 /* then only advance past the last char */
2030 ptr += 2;
2031 }
2032
2033 return 0;
2034}
2035
3bdb8efd
PL		 2036/* Make sure entire volume is addressable by our journal. Requires
2037 osb_clusters_at_boot to be valid and for the journal to have been
2038 initialized by ocfs2_journal_init(). */
2039static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2040{
2041 int status = 0;
2042 u64 max_block =
2043 ocfs2_clusters_to_blocks(osb->sb,
2044 osb->osb_clusters_at_boot) - 1;
2045
2046 /* 32-bit block number is always OK. */
2047 if (max_block <= (u32)~0ULL)
2048 goto out;
2049
2050 /* Volume is "huge", so see if our journal is new enough to
2051 support it. */
2052 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2053 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2054 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2055 JBD2_FEATURE_INCOMPAT_64BIT))) {
2056 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2057 "Enable the 'block64' journal option with tunefs.ocfs2");
2058 status = -EFBIG;
2059 goto out;
2060 }
2061
2062 out:
2063 return status;
2064}
2065
ccd979bd
MF		 2066static int ocfs2_initialize_super(struct super_block *sb,
2067 struct buffer_head *bh,
73be192b
JB		 2068 int sector_size,
2069 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2070{
bddb8eb3 2071 int status;
5a254031
MF		 2072 int i, cbits, bbits;
2073 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2074 struct inode *inode = NULL;
ccd979bd 2075 struct ocfs2_journal *journal;
ccd979bd 2076 struct ocfs2_super *osb;
3bdb8efd 2077 u64 total_blocks;
ccd979bd 2078
cd861280 2079 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 2080 if (!osb) {
2081 status = -ENOMEM;
2082 mlog_errno(status);
2083 goto bail;
2084 }
2085
2086 sb->s_fs_info = osb;
2087 sb->s_op = &ocfs2_sops;
ba87167c 2088 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2089 sb->s_export_op = &ocfs2_export_ops;
19ece546
JK		 2090 sb->s_qcop = &ocfs2_quotactl_ops;
2091 sb->dq_op = &ocfs2_quota_operations;
1c92ec67 2092 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
cf1d6c76 2093 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2094 sb->s_time_gran = 1;
ccd979bd
MF		 2095 sb->s_flags |= MS_NOATIME;
2096 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2097 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2098 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2099 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd 2100
9b7895ef
MF		 2101 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2102
2103 for (i = 0; i < 3; i++)
2104 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2105 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2106
ccd979bd
MF		 2107 osb->sb = sb;
2108 /* Save off for ocfs2_rw_direct */
2109 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2110 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2111
34d024f8
MF		 2112 spin_lock_init(&osb->dc_task_lock);
2113 init_waitqueue_head(&osb->dc_event);
2114 osb->dc_work_sequence = 0;
2115 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2116 INIT_LIST_HEAD(&osb->blocked_lock_list);
2117 osb->blocked_lock_count = 0;
ccd979bd 2118 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2119 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2120 ocfs2_init_steal_slots(osb);
ccd979bd 2121
43b10a20 2122 mutex_init(&osb->system_file_mutex);
2123
ccd979bd
MF		 2124 atomic_set(&osb->alloc_stats.moves, 0);
2125 atomic_set(&osb->alloc_stats.local_data, 0);
2126 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2127 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2128 atomic_set(&osb->alloc_stats.bg_extends, 0);
2129
73be192b
JB		 2130 /* Copy the blockcheck stats from the superblock probe */
2131 osb->osb_ecc_stats = *stats;
2132
ccd979bd
MF		 2133 ocfs2_init_node_maps(osb);
2134
2135 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2136 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2137
75d9bbc7
GR		 2138 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2139 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2140 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2141 osb->max_slots);
2142 status = -EINVAL;
2143 goto bail;
2144 }
75d9bbc7 2145
8b712cd5
JM		 2146 ocfs2_orphan_scan_init(osb);
2147
553abd04
JB		 2148 status = ocfs2_recovery_init(osb);
2149 if (status) {
2150 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2151 mlog_errno(status);
2152 goto bail;
2153 }
ccd979bd
MF		 2154
2155 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2156
7f1a37e3
TY		 2157 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2158
ccd979bd
MF		 2159 osb->slot_num = OCFS2_INVALID_SLOT;
2160
8154da3d
TY		 2161 osb->s_xattr_inline_size = le16_to_cpu(
2162 di->id2.i_super.s_xattr_inline_size);
fdd77704 2163
ccd979bd
MF		 2164 osb->local_alloc_state = OCFS2_LA_UNUSED;
2165 osb->local_alloc_bh = NULL;
9c7af40b 2166 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2167
ccd979bd
MF		 2168 init_waitqueue_head(&osb->osb_mount_event);
2169
d02f00cc
MF		 2170 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2171 if (status) {
2172 mlog_errno(status);
2173 goto bail;
2174 }
2175
ccd979bd
MF		 2176 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2177 if (!osb->vol_label) {
2178 mlog(ML_ERROR, "unable to alloc vol label\n");
2179 status = -ENOMEM;
2180 goto bail;
2181 }
2182
539d8264
SM		 2183 osb->slot_recovery_generations =
2184 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2185 GFP_KERNEL);
2186 if (!osb->slot_recovery_generations) {
2187 status = -ENOMEM;
2188 mlog_errno(status);
2189 goto bail;
2190 }
2191
b4df6ed8
MF		 2192 init_waitqueue_head(&osb->osb_wipe_event);
2193 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2194 sizeof(*osb->osb_orphan_wipes),
2195 GFP_KERNEL);
2196 if (!osb->osb_orphan_wipes) {
2197 status = -ENOMEM;
2198 mlog_errno(status);
2199 goto bail;
2200 }
2201
374a263e
TM		 2202 osb->osb_rf_lock_tree = RB_ROOT;
2203
ccd979bd
MF		 2204 osb->s_feature_compat =
2205 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2206 osb->s_feature_ro_compat =
2207 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2208 osb->s_feature_incompat =
2209 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2210
2211 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2212 mlog(ML_ERROR, "couldn't mount because of unsupported "
2213 "optional features (%x).\n", i);
2214 status = -EINVAL;
2215 goto bail;
2216 }
2217 if (!(osb->sb->s_flags & MS_RDONLY) &&
2218 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2219 mlog(ML_ERROR, "couldn't mount RDWR because of "
2220 "unsupported optional features (%x).\n", i);
2221 status = -EINVAL;
2222 goto bail;
2223 }
2224
98f486f2
SM		 2225 if (ocfs2_clusterinfo_valid(osb)) {
2226 osb->osb_stackflags =
2227 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2228 strlcpy(osb->osb_cluster_stack,
b61817e1 2229 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2230 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB		 2231 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2232 mlog(ML_ERROR,
2233 "couldn't mount because of an invalid "
2234 "cluster stack label (%s) \n",
2235 osb->osb_cluster_stack);
2236 status = -EINVAL;
2237 goto bail;
2238 }
c74a3bdd
GR		 2239 strlcpy(osb->osb_cluster_name,
2240 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2241 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB		 2242 } else {
2243 /* The empty string is identical with classic tools that
2244 * don't know about s_cluster_info. */
2245 osb->osb_cluster_stack[0] = '\0';
2246 }
2247
ccd979bd
MF		 2248 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2249
2250 /* FIXME
2251 * This should be done in ocfs2_journal_init(), but unknown
2252 * ordering issues will cause the filesystem to crash.
2253 * If anyone wants to figure out what part of the code
2254 * refers to osb->journal before ocfs2_journal_init() is run,
2255 * be my guest.
2256 */
2257 /* initialize our journal structure */
2258
cd861280 2259 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2260 if (!journal) {
2261 mlog(ML_ERROR, "unable to alloc journal\n");
2262 status = -ENOMEM;
2263 goto bail;
2264 }
2265 osb->journal = journal;
2266 journal->j_osb = osb;
2267
2268 atomic_set(&journal->j_num_trans, 0);
2269 init_rwsem(&journal->j_trans_barrier);
2270 init_waitqueue_head(&journal->j_checkpointed);
2271 spin_lock_init(&journal->j_lock);
2272 journal->j_trans_id = (unsigned long) 1;
2273 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2274 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2275 journal->j_state = OCFS2_JOURNAL_FREE;
2276
e3a767b6
JK		 2277 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2278 init_llist_head(&osb->dquot_drop_list);
2279
ccd979bd
MF		 2280 /* get some pseudo constants for clustersize bits */
2281 osb->s_clustersize_bits =
2282 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2283 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF		 2284
2285 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2286 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2287 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2288 osb->s_clustersize);
2289 status = -EINVAL;
2290 goto bail;
2291 }
2292
3bdb8efd
PL		 2293 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2294 le32_to_cpu(di->i_clusters));
2295
2296 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2297 total_blocks);
2298 if (status) {
2299 mlog(ML_ERROR, "Volume too large "
2300 "to mount safely on this system");
2301 status = -EFBIG;
ccd979bd
MF		 2302 goto bail;
2303 }
2304
2305 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2306 sizeof(di->id2.i_super.s_uuid))) {
2307 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2308 status = -ENOMEM;
2309 goto bail;
2310 }
2311
69201bb1
FF		 2312 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2313 OCFS2_MAX_VOL_LABEL_LEN);
ccd979bd
MF		 2314 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2315 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2316 osb->first_cluster_group_blkno =
2317 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2318 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2319 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM		 2320 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2321 (unsigned long long)osb->root_blkno,
2322 (unsigned long long)osb->system_dir_blkno,
2323 osb->s_clustersize_bits);
ccd979bd
MF		 2324
2325 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2326 if (!osb->osb_dlm_debug) {
2327 status = -ENOMEM;
2328 mlog_errno(status);
2329 goto bail;
2330 }
2331
2332 atomic_set(&osb->vol_state, VOLUME_INIT);
2333
2334 /* load root, system_dir, and all global system inodes */
2335 status = ocfs2_init_global_system_inodes(osb);
2336 if (status < 0) {
2337 mlog_errno(status);
2338 goto bail;
2339 }
2340
2341 /*
2342 * global bitmap
2343 */
2344 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2345 OCFS2_INVALID_SLOT);
2346 if (!inode) {
2347 status = -EINVAL;
2348 mlog_errno(status);
2349 goto bail;
2350 }
2351
2352 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2353 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2354 iput(inode);
ccd979bd 2355
8571882c
TM		 2356 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2357 osb->s_feature_incompat) * 8;
ccd979bd
MF		 2358
2359 status = ocfs2_init_slot_info(osb);
2360 if (status < 0) {
2361 mlog_errno(status);
2362 goto bail;
2363 }
e41d33af 2364 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
ccd979bd 2365
ccd979bd 2366bail:
ccd979bd
MF		 2367 return status;
2368}
2369
2370/*
2371 * will return: -EAGAIN if it is ok to keep searching for superblocks
2372 * -EINVAL if there is a bad superblock
2373 * 0 on success
2374 */
2375static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2376 struct buffer_head *bh,
73be192b
JB		 2377 u32 blksz,
2378 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2379{
2380 int status = -EAGAIN;
2381
ccd979bd
MF		 2382 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2383 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2384 /* We have to do a raw check of the feature here */
2385 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2386 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2387 status = ocfs2_block_check_validate(bh->b_data,
2388 bh->b_size,
73be192b
JB		 2389 &di->i_check,
2390 stats);
d030cc97
JB		 2391 if (status)
2392 goto out;
2393 }
ccd979bd
MF		 2394 status = -EINVAL;
2395 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2396 mlog(ML_ERROR, "found superblock with incorrect block "
2397 "size: found %u, should be %u\n",
2398 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2399 blksz);
2400 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2401 OCFS2_MAJOR_REV_LEVEL ||
2402 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2403 OCFS2_MINOR_REV_LEVEL) {
2404 mlog(ML_ERROR, "found superblock with bad version: "
2405 "found %u.%u, should be %u.%u\n",
2406 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2407 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2408 OCFS2_MAJOR_REV_LEVEL,
2409 OCFS2_MINOR_REV_LEVEL);
2410 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2411 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2412 "found %llu, should be %llu\n",
1ca1a111 2413 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2414 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2415 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2416 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2417 mlog(ML_ERROR, "bad cluster size found: %u\n",
2418 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2419 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2420 mlog(ML_ERROR, "bad root_blkno: 0\n");
2421 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2422 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2423 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2424 mlog(ML_ERROR,
2425 "Superblock slots found greater than file system "
2426 "maximum: found %u, max %u\n",
2427 le16_to_cpu(di->id2.i_super.s_max_slots),
2428 OCFS2_MAX_SLOTS);
2429 } else {
2430 /* found it! */
2431 status = 0;
2432 }
2433 }
2434
d030cc97 2435out:
c1e8d35e
TM		 2436 if (status && status != -EAGAIN)
2437 mlog_errno(status);
ccd979bd
MF		 2438 return status;
2439}
2440
2441static int ocfs2_check_volume(struct ocfs2_super *osb)
2442{
bddb8eb3 2443 int status;
ccd979bd 2444 int dirty;
c271c5c2 2445 int local;
ccd979bd
MF		 2446 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2447 * recover
2448 * ourselves. */
2449
ccd979bd
MF		 2450 /* Init our journal object. */
2451 status = ocfs2_journal_init(osb->journal, &dirty);
2452 if (status < 0) {
2453 mlog(ML_ERROR, "Could not initialize journal!\n");
2454 goto finally;
2455 }
2456
3bdb8efd
PL		 2457 /* Now that journal has been initialized, check to make sure
2458 entire volume is addressable. */
2459 status = ocfs2_journal_addressable(osb);
2460 if (status)
2461 goto finally;
2462
ccd979bd
MF		 2463 /* If the journal was unmounted cleanly then we don't want to
2464 * recover anything. Otherwise, journal_load will do that
2465 * dirty work for us :) */
2466 if (!dirty) {
2467 status = ocfs2_journal_wipe(osb->journal, 0);
2468 if (status < 0) {
2469 mlog_errno(status);
2470 goto finally;
2471 }
2472 } else {
619c200d
SM		 2473 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2474 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF		 2475 }
2476
c271c5c2
SM		 2477 local = ocfs2_mount_local(osb);
2478
ccd979bd 2479 /* will play back anything left in the journal. */
539d8264 2480 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2481 if (status < 0) {
2482 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2483 goto finally;
2484 }
ccd979bd 2485
1dfeb768 2486 if (osb->s_mount_opt & OCFS2_MOUNT_JOURNAL_ASYNC_COMMIT)
2487 jbd2_journal_set_features(osb->journal->j_journal,
2488 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2489 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2490 else
2491 jbd2_journal_clear_features(osb->journal->j_journal,
2492 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2493 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2494
ccd979bd
MF		 2495 if (dirty) {
2496 /* recover my local alloc if we didn't unmount cleanly. */
2497 status = ocfs2_begin_local_alloc_recovery(osb,
2498 osb->slot_num,
2499 &local_alloc);
2500 if (status < 0) {
2501 mlog_errno(status);
2502 goto finally;
2503 }
2504 /* we complete the recovery process after we've marked
2505 * ourselves as mounted. */
2506 }
2507
ccd979bd
MF		 2508 status = ocfs2_load_local_alloc(osb);
2509 if (status < 0) {
2510 mlog_errno(status);
2511 goto finally;
2512 }
2513
2514 if (dirty) {
2515 /* Recovery will be completed after we've mounted the
2516 * rest of the volume. */
2517 osb->dirty = 1;
2518 osb->local_alloc_copy = local_alloc;
2519 local_alloc = NULL;
2520 }
2521
2522 /* go through each journal, trylock it and if you get the
2523 * lock, and it's marked as dirty, set the bit in the recover
2524 * map and launch a recovery thread for it. */
2525 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2526 if (status < 0) {
2527 mlog_errno(status);
2528 goto finally;
2529 }
2530
2531 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2532 if (status < 0)
2533 mlog_errno(status);
2534
2535finally:
d787ab09 2536 kfree(local_alloc);
ccd979bd 2537
c1e8d35e
TM		 2538 if (status)
2539 mlog_errno(status);
ccd979bd
MF		 2540 return status;
2541}
2542
2543/*
2544 * The routine gets called from dismount or close whenever a dismount on
2545 * volume is requested and the osb open count becomes 1.
2546 * It will remove the osb from the global list and also free up all the
2547 * initialized resources and fileobject.
2548 */
2549static void ocfs2_delete_osb(struct ocfs2_super *osb)
2550{
ccd979bd
MF		 2551 /* This function assumes that the caller has the main osb resource */
2552
8e8a4603 2553 ocfs2_free_slot_info(osb);
ccd979bd 2554
b4df6ed8 2555 kfree(osb->osb_orphan_wipes);
539d8264 2556 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2557 /* FIXME
2558 * This belongs in journal shutdown, but because we have to
421f91d2 2559 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF		 2560 * we free it here.
2561 */
2562 kfree(osb->journal);
d787ab09 2563 kfree(osb->local_alloc_copy);
ccd979bd 2564 kfree(osb->uuid_str);
f13a568e 2565 kfree(osb->vol_label);
ccd979bd
MF		 2566 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2567 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF		 2568}
2569
2570/* Put OCFS2 into a readonly state, or (if the user specifies it),
2571 * panic(). We do not support continue-on-error operation. */
2572static void ocfs2_handle_error(struct super_block *sb)
2573{
2574 struct ocfs2_super *osb = OCFS2_SB(sb);
2575
2576 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2577 panic("OCFS2: (device %s): panic forced after error\n",
2578 sb->s_id);
2579
2580 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2581
2582 if (sb->s_flags & MS_RDONLY &&
2583 (ocfs2_is_soft_readonly(osb) ||
2584 ocfs2_is_hard_readonly(osb)))
2585 return;
2586
2587 printk(KERN_CRIT "File system is now read-only due to the potential "
2588 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2589 "system is unmounted.\n");
2590 sb->s_flags |= MS_RDONLY;
2591 ocfs2_set_ro_flag(osb, 0);
2592}
2593
2594static char error_buf[1024];
2595
2596void __ocfs2_error(struct super_block *sb,
2597 const char *function,
2598 const char *fmt, ...)
2599{
2600 va_list args;
2601
2602 va_start(args, fmt);
4a6e617a 2603 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2604 va_end(args);
2605
2606 /* Not using mlog here because we want to show the actual
2607 * function the error came from. */
2608 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2609 sb->s_id, function, error_buf);
2610
2611 ocfs2_handle_error(sb);
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. */
2617void __ocfs2_abort(struct super_block* sb,
2618 const char *function,
2619 const char *fmt, ...)
2620{
2621 va_list args;
2622
2623 va_start(args, fmt);
4a6e617a 2624 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2625 va_end(args);
2626
2627 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2628 sb->s_id, function, error_buf);
2629
2630 /* We don't have the cluster support yet to go straight to
2631 * hard readonly in here. Until then, we want to keep
2632 * ocfs2_abort() so that we can at least mark critical
2633 * errors.
2634 *
2635 * TODO: This should abort the journal and alert other nodes
2636 * that our slot needs recovery. */
2637
2638 /* Force a panic(). This stinks, but it's better than letting
2639 * things continue without having a proper hard readonly
2640 * here. */
a2f2ddbf
SM		 2641 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2642 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2643 ocfs2_handle_error(sb);
2644}
2645
e4b963f1
JB		 2646/*
2647 * Void signal blockers, because in-kernel sigprocmask() only fails
2648 * when SIG_* is wrong.
2649 */
2650void ocfs2_block_signals(sigset_t *oldset)
2651{
2652 int rc;
2653 sigset_t blocked;
2654
2655 sigfillset(&blocked);
2656 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2657 BUG_ON(rc);
2658}
2659
2660void ocfs2_unblock_signals(sigset_t *oldset)
2661{
2662 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2663 BUG_ON(rc);
2664}
2665
ccd979bd
MF		 2666module_init(ocfs2_init);
2667module_exit(ocfs2_exit);
Linux kernel - Deliverables
This page took 1.103633 seconds and 5 git commands to generate.