projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
printk: add comment about tricky check for text buffer size
[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
e18b890b 78static struct kmem_cache *ocfs2_inode_cachep = NULL;
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
88static struct dentry *ocfs2_debugfs_root = NULL;
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
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,
ccd979bd
MF		 158};
159
160enum {
161 Opt_barrier,
162 Opt_err_panic,
163 Opt_err_ro,
164 Opt_intr,
165 Opt_nointr,
166 Opt_hb_none,
167 Opt_hb_local,
2c442719 168 Opt_hb_global,
ccd979bd
MF		 169 Opt_data_ordered,
170 Opt_data_writeback,
7f1a37e3 171 Opt_atime_quantum,
baf4661a 172 Opt_slot,
d147b3d6 173 Opt_commit,
2fbe8d1e 174 Opt_localalloc,
53fc622b 175 Opt_localflocks,
b61817e1 176 Opt_stack,
cf1d6c76
TY		 177 Opt_user_xattr,
178 Opt_nouser_xattr,
12462f1d 179 Opt_inode64,
a68979b8
TY		 180 Opt_acl,
181 Opt_noacl,
19ece546
JK		 182 Opt_usrquota,
183 Opt_grpquota,
7bdb0d18
TY		 184 Opt_coherency_buffered,
185 Opt_coherency_full,
d02f00cc 186 Opt_resv_level,
83f92318 187 Opt_dir_resv_level,
ccd979bd
MF		 188 Opt_err,
189};
190
a447c093 191static const match_table_t tokens = {
ccd979bd
MF		 192 {Opt_barrier, "barrier=%u"},
193 {Opt_err_panic, "errors=panic"},
194 {Opt_err_ro, "errors=remount-ro"},
195 {Opt_intr, "intr"},
196 {Opt_nointr, "nointr"},
197 {Opt_hb_none, OCFS2_HB_NONE},
198 {Opt_hb_local, OCFS2_HB_LOCAL},
2c442719 199 {Opt_hb_global, OCFS2_HB_GLOBAL},
ccd979bd
MF		 200 {Opt_data_ordered, "data=ordered"},
201 {Opt_data_writeback, "data=writeback"},
7f1a37e3 202 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 203 {Opt_slot, "preferred_slot=%u"},
d147b3d6 204 {Opt_commit, "commit=%u"},
2fbe8d1e 205 {Opt_localalloc, "localalloc=%d"},
53fc622b 206 {Opt_localflocks, "localflocks"},
b61817e1 207 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 208 {Opt_user_xattr, "user_xattr"},
209 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 210 {Opt_inode64, "inode64"},
a68979b8
TY		 211 {Opt_acl, "acl"},
212 {Opt_noacl, "noacl"},
19ece546
JK		 213 {Opt_usrquota, "usrquota"},
214 {Opt_grpquota, "grpquota"},
7bdb0d18
TY		 215 {Opt_coherency_buffered, "coherency=buffered"},
216 {Opt_coherency_full, "coherency=full"},
d02f00cc 217 {Opt_resv_level, "resv_level=%u"},
83f92318 218 {Opt_dir_resv_level, "dir_resv_level=%u"},
ccd979bd
MF		 219 {Opt_err, NULL}
220};
221
50397507
SM		 222#ifdef CONFIG_DEBUG_FS
223static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
224{
50397507
SM		 225 struct ocfs2_cluster_connection *cconn = osb->cconn;
226 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 227 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
228 int i, out = 0;
50397507
SM		 229
230 out += snprintf(buf + out, len - out,
231 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
232 "Device", osb->dev_str, osb->uuid_str,
233 osb->fs_generation, osb->vol_label);
234
235 out += snprintf(buf + out, len - out,
236 "%10s => State: %d Flags: 0x%lX\n", "Volume",
237 atomic_read(&osb->vol_state), osb->osb_flags);
238
239 out += snprintf(buf + out, len - out,
240 "%10s => Block: %lu Cluster: %d\n", "Sizes",
241 osb->sb->s_blocksize, osb->s_clustersize);
242
243 out += snprintf(buf + out, len - out,
244 "%10s => Compat: 0x%X Incompat: 0x%X "
245 "ROcompat: 0x%X\n",
246 "Features", osb->s_feature_compat,
247 osb->s_feature_incompat, osb->s_feature_ro_compat);
248
249 out += snprintf(buf + out, len - out,
250 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
251 osb->s_mount_opt, osb->s_atime_quantum);
252
c3d38840
SM		 253 if (cconn) {
254 out += snprintf(buf + out, len - out,
255 "%10s => Stack: %s Name: %*s "
256 "Version: %d.%d\n", "Cluster",
257 (*osb->osb_cluster_stack == '\0' ?
258 "o2cb" : osb->osb_cluster_stack),
259 cconn->cc_namelen, cconn->cc_name,
260 cconn->cc_version.pv_major,
261 cconn->cc_version.pv_minor);
262 }
50397507
SM		 263
264 spin_lock(&osb->dc_task_lock);
265 out += snprintf(buf + out, len - out,
266 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
267 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 268 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
269 osb->blocked_lock_count, osb->dc_wake_sequence,
270 osb->dc_work_sequence);
50397507
SM		 271 spin_unlock(&osb->dc_task_lock);
272
273 spin_lock(&osb->osb_lock);
274 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
275 "Recovery",
276 (osb->recovery_thread_task ?
277 task_pid_nr(osb->recovery_thread_task) : -1));
278 if (rm->rm_used == 0)
279 out += snprintf(buf + out, len - out, " None\n");
280 else {
281 for (i = 0; i < rm->rm_used; i++)
282 out += snprintf(buf + out, len - out, " %d",
283 rm->rm_entries[i]);
284 out += snprintf(buf + out, len - out, "\n");
285 }
286 spin_unlock(&osb->osb_lock);
287
288 out += snprintf(buf + out, len - out,
8fa9d17f 289 "%10s => Pid: %d Interval: %lu\n", "Commit",
c3d38840 290 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
8fa9d17f 291 osb->osb_commit_interval);
50397507
SM		 292
293 out += snprintf(buf + out, len - out,
c3d38840 294 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 295 "Journal", osb->journal->j_state,
c3d38840
SM		 296 osb->journal->j_trans_id,
297 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 298
299 out += snprintf(buf + out, len - out,
300 "%10s => GlobalAllocs: %d LocalAllocs: %d "
301 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
302 "Stats",
303 atomic_read(&osb->alloc_stats.bitmap_data),
304 atomic_read(&osb->alloc_stats.local_data),
305 atomic_read(&osb->alloc_stats.bg_allocs),
306 atomic_read(&osb->alloc_stats.moves),
307 atomic_read(&osb->alloc_stats.bg_extends));
308
309 out += snprintf(buf + out, len - out,
310 "%10s => State: %u Descriptor: %llu Size: %u bits "
311 "Default: %u bits\n",
312 "LocalAlloc", osb->local_alloc_state,
313 (unsigned long long)osb->la_last_gd,
314 osb->local_alloc_bits, osb->local_alloc_default_bits);
315
316 spin_lock(&osb->osb_lock);
317 out += snprintf(buf + out, len - out,
b89c5428
TY		 318 "%10s => InodeSlot: %d StolenInodes: %d, "
319 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 320 osb->s_inode_steal_slot,
b89c5428
TY		 321 atomic_read(&osb->s_num_inodes_stolen),
322 osb->s_meta_steal_slot,
323 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 324 spin_unlock(&osb->osb_lock);
325
df152c24
SM		 326 out += snprintf(buf + out, len - out, "OrphanScan => ");
327 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
328 os->os_count, os->os_seqno);
329 out += snprintf(buf + out, len - out, " Last Scan: ");
330 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
331 out += snprintf(buf + out, len - out, "Disabled\n");
332 else
333 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
334 (get_seconds() - os->os_scantime.tv_sec));
335
50397507
SM		 336 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
337 "Slots", "Num", "RecoGen");
50397507
SM		 338 for (i = 0; i < osb->max_slots; ++i) {
339 out += snprintf(buf + out, len - out,
340 "%10s %c %3d %10d\n",
341 " ",
342 (i == osb->slot_num ? '*' : ' '),
343 i, osb->slot_recovery_generations[i]);
344 }
345
346 return out;
347}
348
349static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
350{
351 struct ocfs2_super *osb = inode->i_private;
352 char *buf = NULL;
353
354 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
355 if (!buf)
356 goto bail;
357
358 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
359
360 file->private_data = buf;
361
362 return 0;
363bail:
364 return -ENOMEM;
365}
366
367static int ocfs2_debug_release(struct inode *inode, struct file *file)
368{
369 kfree(file->private_data);
370 return 0;
371}
372
373static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
374 size_t nbytes, loff_t *ppos)
375{
376 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
377 i_size_read(file->f_mapping->host));
378}
379#else
380static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
381{
382 return 0;
383}
384static int ocfs2_debug_release(struct inode *inode, struct file *file)
385{
386 return 0;
387}
388static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
389 size_t nbytes, loff_t *ppos)
390{
391 return 0;
392}
393#endif /* CONFIG_DEBUG_FS */
394
828c0950 395static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 396 .open = ocfs2_osb_debug_open,
397 .release = ocfs2_debug_release,
398 .read = ocfs2_debug_read,
399 .llseek = generic_file_llseek,
400};
401
ccd979bd
MF		 402static int ocfs2_sync_fs(struct super_block *sb, int wait)
403{
bddb8eb3 404 int status;
ccd979bd
MF		 405 tid_t target;
406 struct ocfs2_super *osb = OCFS2_SB(sb);
407
ccd979bd
MF		 408 if (ocfs2_is_hard_readonly(osb))
409 return -EROFS;
410
411 if (wait) {
412 status = ocfs2_flush_truncate_log(osb);
413 if (status < 0)
414 mlog_errno(status);
415 } else {
416 ocfs2_schedule_truncate_log_flush(osb, 0);
417 }
418
2b4e30fb
JB		 419 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
420 &target)) {
ccd979bd 421 if (wait)
2b4e30fb
JB		 422 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
423 target);
ccd979bd
MF		 424 }
425 return 0;
426}
427
1a224ad1
JK		 428static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
429{
430 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
431 && (ino == USER_QUOTA_SYSTEM_INODE
432 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
433 return 0;
434 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
435 && (ino == GROUP_QUOTA_SYSTEM_INODE
436 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
437 return 0;
438 return 1;
439}
440
ccd979bd
MF		 441static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
442{
443 struct inode *new = NULL;
444 int status = 0;
445 int i;
446
5fa0613e 447 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 448 if (IS_ERR(new)) {
449 status = PTR_ERR(new);
450 mlog_errno(status);
451 goto bail;
452 }
453 osb->root_inode = new;
454
5fa0613e 455 new = ocfs2_iget(osb, osb->system_dir_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->sys_root_inode = new;
462
463 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
464 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 465 if (!ocfs2_need_system_inode(osb, i))
466 continue;
ccd979bd
MF		 467 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
468 if (!new) {
469 ocfs2_release_system_inodes(osb);
470 status = -EINVAL;
471 mlog_errno(status);
472 /* FIXME: Should ERROR_RO_FS */
473 mlog(ML_ERROR, "Unable to load system inode %d, "
474 "possibly corrupt fs?", i);
475 goto bail;
476 }
477 // the array now has one ref, so drop this one
478 iput(new);
479 }
480
481bail:
c1e8d35e
TM		 482 if (status)
483 mlog_errno(status);
ccd979bd
MF		 484 return status;
485}
486
487static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
488{
489 struct inode *new = NULL;
490 int status = 0;
491 int i;
492
ccd979bd
MF		 493 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
494 i < NUM_SYSTEM_INODES;
495 i++) {
1a224ad1
JK		 496 if (!ocfs2_need_system_inode(osb, i))
497 continue;
ccd979bd
MF		 498 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
499 if (!new) {
500 ocfs2_release_system_inodes(osb);
501 status = -EINVAL;
502 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
503 status, i, osb->slot_num);
504 goto bail;
505 }
506 /* the array now has one ref, so drop this one */
507 iput(new);
508 }
509
510bail:
c1e8d35e
TM		 511 if (status)
512 mlog_errno(status);
ccd979bd
MF		 513 return status;
514}
515
bddb8eb3 516static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 517{
bddb8eb3 518 int i;
ccd979bd
MF		 519 struct inode *inode;
520
b4d693fc
TM		 521 for (i = 0; i < NUM_GLOBAL_SYSTEM_INODES; i++) {
522 inode = osb->global_system_inodes[i];
ccd979bd
MF		 523 if (inode) {
524 iput(inode);
b4d693fc 525 osb->global_system_inodes[i] = NULL;
ccd979bd
MF		 526 }
527 }
528
529 inode = osb->sys_root_inode;
530 if (inode) {
531 iput(inode);
532 osb->sys_root_inode = NULL;
533 }
534
535 inode = osb->root_inode;
536 if (inode) {
537 iput(inode);
538 osb->root_inode = NULL;
539 }
540
b4d693fc 541 if (!osb->local_system_inodes)
c1e8d35e 542 return;
b4d693fc
TM		 543
544 for (i = 0; i < NUM_LOCAL_SYSTEM_INODES * osb->max_slots; i++) {
545 if (osb->local_system_inodes[i]) {
546 iput(osb->local_system_inodes[i]);
547 osb->local_system_inodes[i] = NULL;
548 }
549 }
550
551 kfree(osb->local_system_inodes);
552 osb->local_system_inodes = NULL;
ccd979bd
MF		 553}
554
555/* We're allocating fs objects, use GFP_NOFS */
556static struct inode *ocfs2_alloc_inode(struct super_block *sb)
557{
558 struct ocfs2_inode_info *oi;
559
e6b4f8da 560 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 561 if (!oi)
562 return NULL;
563
2931cdcb
DW		 564 oi->i_sync_tid = 0;
565 oi->i_datasync_tid = 0;
566
2b4e30fb 567 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 568 return &oi->vfs_inode;
569}
570
fa0d7e3d 571static void ocfs2_i_callback(struct rcu_head *head)
ccd979bd 572{
fa0d7e3d 573 struct inode *inode = container_of(head, struct inode, i_rcu);
ccd979bd
MF		 574 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
575}
576
fa0d7e3d
NP		 577static void ocfs2_destroy_inode(struct inode *inode)
578{
579 call_rcu(&inode->i_rcu, ocfs2_i_callback);
580}
581
5a254031
MF		 582static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
583 unsigned int cbits)
ccd979bd 584{
5a254031
MF		 585 unsigned int bytes = 1 << cbits;
586 unsigned int trim = bytes;
587 unsigned int bitshift = 32;
588
589 /*
590 * i_size and all block offsets in ocfs2 are always 64 bits
591 * wide. i_clusters is 32 bits, in cluster-sized units. So on
592 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 593 */
594
595#if BITS_PER_LONG == 32
90c699a9 596# if defined(CONFIG_LBDAF)
2ecd05ae 597 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 598 /*
599 * We might be limited by page cache size.
600 */
601 if (bytes > PAGE_CACHE_SIZE) {
602 bytes = PAGE_CACHE_SIZE;
603 trim = 1;
604 /*
605 * Shift by 31 here so that we don't get larger than
606 * MAX_LFS_FILESIZE
607 */
608 bitshift = 31;
609 }
ccd979bd 610# else
5a254031
MF		 611 /*
612 * We are limited by the size of sector_t. Use block size, as
613 * that's what we expose to the VFS.
614 */
615 bytes = 1 << bbits;
616 trim = 1;
617 bitshift = 31;
ccd979bd
MF		 618# endif
619#endif
620
5a254031
MF		 621 /*
622 * Trim by a whole cluster when we can actually approach the
623 * on-disk limits. Otherwise we can overflow i_clusters when
624 * an extent start is at the max offset.
625 */
626 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 627}
628
629static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
630{
631 int incompat_features;
632 int ret = 0;
c0123ade 633 struct mount_options parsed_options;
ccd979bd 634 struct ocfs2_super *osb = OCFS2_SB(sb);
2c442719 635 u32 tmp;
ccd979bd 636
5297aad8
JK		 637 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
638 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF		 639 ret = -EINVAL;
640 goto out;
641 }
642
2c442719
SM		 643 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
644 OCFS2_MOUNT_HB_NONE;
645 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
ccd979bd
MF		 646 ret = -EINVAL;
647 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
648 goto out;
649 }
650
651 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 652 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 653 ret = -EINVAL;
654 mlog(ML_ERROR, "Cannot change data mode on remount\n");
655 goto out;
656 }
657
12462f1d
JB		 658 /* Probably don't want this on remount; it might
659 * mess with other nodes */
660 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
661 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
662 ret = -EINVAL;
663 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
664 goto out;
665 }
666
ccd979bd
MF		 667 /* We're going to/from readonly mode. */
668 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 669 /* Disable quota accounting before remounting RO */
670 if (*flags & MS_RDONLY) {
671 ret = ocfs2_susp_quotas(osb, 0);
672 if (ret < 0)
673 goto out;
674 }
ccd979bd
MF		 675 /* Lock here so the check of HARD_RO and the potential
676 * setting of SOFT_RO is atomic. */
677 spin_lock(&osb->osb_lock);
678 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
679 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
680 ret = -EROFS;
681 goto unlock_osb;
682 }
683
684 if (*flags & MS_RDONLY) {
ccd979bd
MF		 685 sb->s_flags |= MS_RDONLY;
686 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
687 } else {
ccd979bd
MF		 688 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
689 mlog(ML_ERROR, "Cannot remount RDWR "
690 "filesystem due to previous errors.\n");
691 ret = -EROFS;
692 goto unlock_osb;
693 }
694 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
695 if (incompat_features) {
696 mlog(ML_ERROR, "Cannot remount RDWR because "
697 "of unsupported optional features "
698 "(%x).\n", incompat_features);
699 ret = -EINVAL;
700 goto unlock_osb;
701 }
702 sb->s_flags &= ~MS_RDONLY;
703 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
704 }
32a42d39 705 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
ccd979bd
MF		 706unlock_osb:
707 spin_unlock(&osb->osb_lock);
19ece546
JK		 708 /* Enable quota accounting after remounting RW */
709 if (!ret && !(*flags & MS_RDONLY)) {
710 if (sb_any_quota_suspended(sb))
711 ret = ocfs2_susp_quotas(osb, 1);
712 else
713 ret = ocfs2_enable_quotas(osb);
714 if (ret < 0) {
715 /* Return back changes... */
716 spin_lock(&osb->osb_lock);
717 sb->s_flags |= MS_RDONLY;
718 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
719 spin_unlock(&osb->osb_lock);
720 goto out;
721 }
722 }
ccd979bd
MF		 723 }
724
725 if (!ret) {
ccd979bd
MF		 726 /* Only save off the new mount options in case of a successful
727 * remount. */
c0123ade
TY		 728 osb->s_mount_opt = parsed_options.mount_opt;
729 osb->s_atime_quantum = parsed_options.atime_quantum;
730 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 731 if (parsed_options.commit_interval)
732 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 733
734 if (!ocfs2_is_hard_readonly(osb))
735 ocfs2_set_journal_params(osb);
57b09bb5
JK		 736
737 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
738 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
739 MS_POSIXACL : 0);
ccd979bd
MF		 740 }
741out:
742 return ret;
743}
744
745static int ocfs2_sb_probe(struct super_block *sb,
746 struct buffer_head **bh,
73be192b
JB		 747 int *sector_size,
748 struct ocfs2_blockcheck_stats *stats)
ccd979bd 749{
bddb8eb3 750 int status, tmpstat;
ccd979bd
MF		 751 struct ocfs1_vol_disk_hdr *hdr;
752 struct ocfs2_dinode *di;
753 int blksize;
754
755 *bh = NULL;
756
757 /* may be > 512 */
e1defc4f 758 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 759 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
760 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
761 *sector_size, OCFS2_MAX_BLOCKSIZE);
762 status = -EINVAL;
763 goto bail;
764 }
765
766 /* Can this really happen? */
767 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
768 *sector_size = OCFS2_MIN_BLOCKSIZE;
769
770 /* check block zero for old format */
771 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
772 if (status < 0) {
773 mlog_errno(status);
774 goto bail;
775 }
776 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
777 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
778 mlog(ML_ERROR, "incompatible version: %u.%u\n",
779 hdr->major_version, hdr->minor_version);
780 status = -EINVAL;
781 }
782 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
783 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
784 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
785 hdr->signature);
786 status = -EINVAL;
787 }
788 brelse(*bh);
789 *bh = NULL;
790 if (status < 0) {
791 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
792 "upgraded before mounting with ocfs v2\n");
793 goto bail;
794 }
795
796 /*
797 * Now check at magic offset for 512, 1024, 2048, 4096
798 * blocksizes. 4096 is the maximum blocksize because it is
799 * the minimum clustersize.
800 */
801 status = -EINVAL;
802 for (blksize = *sector_size;
803 blksize <= OCFS2_MAX_BLOCKSIZE;
804 blksize <<= 1) {
805 tmpstat = ocfs2_get_sector(sb, bh,
806 OCFS2_SUPER_BLOCK_BLKNO,
807 blksize);
808 if (tmpstat < 0) {
809 status = tmpstat;
810 mlog_errno(status);
fb5cbe9e 811 break;
ccd979bd
MF		 812 }
813 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 814 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 815 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB		 816 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
817 if (tmpstat < 0) {
818 brelse(*bh);
819 *bh = NULL;
820 }
821 if (tmpstat != -EAGAIN) {
822 status = tmpstat;
ccd979bd 823 break;
fb5cbe9e 824 }
ccd979bd
MF		 825 }
826
827bail:
828 return status;
829}
830
c271c5c2
SM		 831static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
832{
2c442719
SM		 833 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
834
835 if (osb->s_mount_opt & hb_enabled) {
836 if (ocfs2_mount_local(osb)) {
c271c5c2
SM		 837 mlog(ML_ERROR, "Cannot heartbeat on a locally "
838 "mounted device.\n");
839 return -EINVAL;
840 }
2c442719 841 if (ocfs2_userspace_stack(osb)) {
b61817e1
JB		 842 mlog(ML_ERROR, "Userspace stack expected, but "
843 "o2cb heartbeat arguments passed to mount\n");
844 return -EINVAL;
845 }
2c442719
SM		 846 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
847 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
848 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
849 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
850 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
851 return -EINVAL;
852 }
b61817e1
JB		 853 }
854
2c442719 855 if (!(osb->s_mount_opt & hb_enabled)) {
b61817e1
JB		 856 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
857 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 858 mlog(ML_ERROR, "Heartbeat has to be started to mount "
859 "a read-write clustered device.\n");
860 return -EINVAL;
861 }
862 }
863
864 return 0;
865}
866
b61817e1
JB		 867/*
868 * If we're using a userspace stack, mount should have passed
869 * a name that matches the disk. If not, mount should not
870 * have passed a stack.
871 */
872static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
873 struct mount_options *mopt)
874{
875 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
876 mlog(ML_ERROR,
877 "cluster stack passed to mount, but this filesystem "
878 "does not support it\n");
879 return -EINVAL;
880 }
881
882 if (ocfs2_userspace_stack(osb) &&
883 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
884 OCFS2_STACK_LABEL_LEN)) {
885 mlog(ML_ERROR,
886 "cluster stack passed to mount (\"%s\") does not "
887 "match the filesystem (\"%s\")\n",
888 mopt->cluster_stack,
889 osb->osb_cluster_stack);
890 return -EINVAL;
891 }
892
893 return 0;
894}
895
19ece546
JK		 896static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
897{
898 int type;
899 struct super_block *sb = osb->sb;
900 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
901 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
902 int status = 0;
903
904 for (type = 0; type < MAXQUOTAS; type++) {
905 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
906 continue;
907 if (unsuspend)
0f0dd62f 908 status = dquot_resume(sb, type);
eea7feb0
JK		 909 else {
910 struct ocfs2_mem_dqinfo *oinfo;
911
912 /* Cancel periodic syncing before suspending */
913 oinfo = sb_dqinfo(sb, type)->dqi_priv;
914 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
0f0dd62f 915 status = dquot_suspend(sb, type);
eea7feb0 916 }
19ece546
JK		 917 if (status < 0)
918 break;
919 }
920 if (status < 0)
921 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
922 "remount (error = %d).\n", status);
923 return status;
924}
925
926static int ocfs2_enable_quotas(struct ocfs2_super *osb)
927{
928 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
929 struct super_block *sb = osb->sb;
930 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
931 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
932 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
933 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
934 int status;
935 int type;
936
937 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
938 for (type = 0; type < MAXQUOTAS; type++) {
939 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
940 continue;
941 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
942 osb->slot_num);
943 if (!inode[type]) {
944 status = -ENOENT;
945 goto out_quota_off;
946 }
287a8095
CH		 947 status = dquot_enable(inode[type], type, QFMT_OCFS2,
948 DQUOT_USAGE_ENABLED);
19ece546
JK		 949 if (status < 0)
950 goto out_quota_off;
951 }
952
953 for (type = 0; type < MAXQUOTAS; type++)
954 iput(inode[type]);
955 return 0;
956out_quota_off:
957 ocfs2_disable_quotas(osb);
958 for (type = 0; type < MAXQUOTAS; type++)
959 iput(inode[type]);
960 mlog_errno(status);
961 return status;
962}
963
964static void ocfs2_disable_quotas(struct ocfs2_super *osb)
965{
966 int type;
967 struct inode *inode;
968 struct super_block *sb = osb->sb;
c06bcbfa 969 struct ocfs2_mem_dqinfo *oinfo;
19ece546
JK		 970
971 /* We mostly ignore errors in this function because there's not much
972 * we can do when we see them */
973 for (type = 0; type < MAXQUOTAS; type++) {
974 if (!sb_has_quota_loaded(sb, type))
975 continue;
c06bcbfa
JK		 976 /* Cancel periodic syncing before we grab dqonoff_mutex */
977 oinfo = sb_dqinfo(sb, type)->dqi_priv;
978 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK		 979 inode = igrab(sb->s_dquot.files[type]);
980 /* Turn off quotas. This will remove all dquot structures from
981 * memory and so they will be automatically synced to global
982 * quota files */
0f0dd62f
CH		 983 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
984 DQUOT_LIMITS_ENABLED);
19ece546
JK		 985 if (!inode)
986 continue;
987 iput(inode);
988 }
989}
990
991/* Handle quota on quotactl */
f00c9e44 992static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
19ece546
JK		 993{
994 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
995 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
996
997 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
998 return -EINVAL;
999
287a8095
CH		 1000 return dquot_enable(sb_dqopt(sb)->files[type], type,
1001 format_id, DQUOT_LIMITS_ENABLED);
19ece546
JK		 1002}
1003
1004/* Handle quota off quotactl */
307ae18a 1005static int ocfs2_quota_off(struct super_block *sb, int type)
19ece546 1006{
0f0dd62f 1007 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
19ece546
JK		 1008}
1009
0d54b217 1010static const struct quotactl_ops ocfs2_quotactl_ops = {
f00c9e44 1011 .quota_on_meta = ocfs2_quota_on,
19ece546 1012 .quota_off = ocfs2_quota_off,
287a8095
CH		 1013 .quota_sync = dquot_quota_sync,
1014 .get_info = dquot_get_dqinfo,
1015 .set_info = dquot_set_dqinfo,
1016 .get_dqblk = dquot_get_dqblk,
1017 .set_dqblk = dquot_set_dqblk,
19ece546
JK		 1018};
1019
ccd979bd
MF		 1020static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1021{
1022 struct dentry *root;
1023 int status, sector_size;
c0123ade 1024 struct mount_options parsed_options;
ccd979bd
MF		 1025 struct inode *inode = NULL;
1026 struct ocfs2_super *osb = NULL;
1027 struct buffer_head *bh = NULL;
49fa8140 1028 char nodestr[12];
73be192b 1029 struct ocfs2_blockcheck_stats stats;
ccd979bd 1030
32a42d39 1031 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1032
c0123ade 1033 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1034 status = -EINVAL;
1035 goto read_super_error;
1036 }
1037
1038 /* probe for superblock */
73be192b 1039 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1040 if (status < 0) {
1041 mlog(ML_ERROR, "superblock probe failed!\n");
1042 goto read_super_error;
1043 }
1044
73be192b 1045 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1046 osb = OCFS2_SB(sb);
1047 if (status < 0) {
1048 mlog_errno(status);
1049 goto read_super_error;
1050 }
1051 brelse(bh);
1052 bh = NULL;
a68979b8 1053
5297aad8
JK		 1054 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1055 status = -EINVAL;
1056 goto read_super_error;
1057 }
c0123ade
TY		 1058 osb->s_mount_opt = parsed_options.mount_opt;
1059 osb->s_atime_quantum = parsed_options.atime_quantum;
1060 osb->preferred_slot = parsed_options.slot;
d147b3d6 1061 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1062
1063 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1064 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF		 1065 osb->osb_dir_resv_level = parsed_options.resv_level;
1066 if (parsed_options.dir_resv_level == -1)
1067 osb->osb_dir_resv_level = parsed_options.resv_level;
1068 else
1069 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1070
b61817e1
JB		 1071 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1072 if (status)
1073 goto read_super_error;
1074
ccd979bd
MF		 1075 sb->s_magic = OCFS2_SUPER_MAGIC;
1076
9e1f1de0 1077 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
a68979b8
TY		 1078 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1079
ccd979bd
MF		 1080 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1081 * heartbeat=none */
1082 if (bdev_read_only(sb->s_bdev)) {
1083 if (!(sb->s_flags & MS_RDONLY)) {
1084 status = -EACCES;
1085 mlog(ML_ERROR, "Readonly device detected but readonly "
1086 "mount was not specified.\n");
1087 goto read_super_error;
1088 }
1089
1090 /* You should not be able to start a local heartbeat
1091 * on a readonly device. */
1092 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1093 status = -EROFS;
1094 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1095 "device.\n");
1096 goto read_super_error;
1097 }
1098
1099 status = ocfs2_check_journals_nolocks(osb);
1100 if (status < 0) {
1101 if (status == -EROFS)
1102 mlog(ML_ERROR, "Recovery required on readonly "
1103 "file system, but write access is "
1104 "unavailable.\n");
1105 else
2bd63216 1106 mlog_errno(status);
ccd979bd
MF		 1107 goto read_super_error;
1108 }
1109
1110 ocfs2_set_ro_flag(osb, 1);
1111
619c200d
SM		 1112 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1113 "Cluster services will not be used for this mount. "
1114 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF		 1115 }
1116
1117 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1118 if (sb->s_flags & MS_RDONLY)
1119 ocfs2_set_ro_flag(osb, 0);
1120 }
1121
c271c5c2
SM		 1122 status = ocfs2_verify_heartbeat(osb);
1123 if (status < 0) {
1124 mlog_errno(status);
1125 goto read_super_error;
1126 }
1127
ccd979bd
MF		 1128 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1129 ocfs2_debugfs_root);
1130 if (!osb->osb_debug_root) {
1131 status = -EINVAL;
1132 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1133 goto read_super_error;
1134 }
1135
50397507
SM		 1136 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1137 osb->osb_debug_root,
1138 osb,
1139 &ocfs2_osb_debug_fops);
1140 if (!osb->osb_ctxt) {
1141 status = -EINVAL;
1142 mlog_errno(status);
1143 goto read_super_error;
1144 }
1145
73be192b
JB		 1146 if (ocfs2_meta_ecc(osb)) {
1147 status = ocfs2_blockcheck_stats_debugfs_install(
1148 &osb->osb_ecc_stats,
1149 osb->osb_debug_root);
1150 if (status) {
1151 mlog(ML_ERROR,
1152 "Unable to create blockcheck statistics "
1153 "files\n");
1154 goto read_super_error;
1155 }
1156 }
1157
ccd979bd 1158 status = ocfs2_mount_volume(sb);
ccd979bd
MF		 1159 if (status < 0)
1160 goto read_super_error;
1161
be0d93f0
AV		 1162 if (osb->root_inode)
1163 inode = igrab(osb->root_inode);
1164
ccd979bd
MF		 1165 if (!inode) {
1166 status = -EIO;
1167 mlog_errno(status);
1168 goto read_super_error;
1169 }
1170
48fde701 1171 root = d_make_root(inode);
ccd979bd
MF		 1172 if (!root) {
1173 status = -ENOMEM;
1174 mlog_errno(status);
1175 goto read_super_error;
1176 }
1177
1178 sb->s_root = root;
1179
1180 ocfs2_complete_mount_recovery(osb);
1181
c271c5c2
SM		 1182 if (ocfs2_mount_local(osb))
1183 snprintf(nodestr, sizeof(nodestr), "local");
1184 else
19fdb624 1185 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1186
1187 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1188 "with %s data mode.\n",
c271c5c2 1189 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1190 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1191 "ordered");
1192
1193 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1194 wake_up(&osb->osb_mount_event);
1195
19ece546
JK		 1196 /* Now we can initialize quotas because we can afford to wait
1197 * for cluster locks recovery now. That also means that truncation
1198 * log recovery can happen but that waits for proper quota setup */
1199 if (!(sb->s_flags & MS_RDONLY)) {
1200 status = ocfs2_enable_quotas(osb);
1201 if (status < 0) {
1202 /* We have to err-out specially here because
1203 * s_root is already set */
1204 mlog_errno(status);
1205 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1206 wake_up(&osb->osb_mount_event);
19ece546
JK		 1207 return status;
1208 }
1209 }
1210
1211 ocfs2_complete_quota_recovery(osb);
1212
1213 /* Now we wake up again for processes waiting for quotas */
1214 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1215 wake_up(&osb->osb_mount_event);
1216
df152c24 1217 /* Start this when the mount is almost sure of being successful */
8b712cd5 1218 ocfs2_orphan_scan_start(osb);
df152c24 1219
ccd979bd
MF		 1220 return status;
1221
1222read_super_error:
a81cb88b 1223 brelse(bh);
ccd979bd 1224
ccd979bd
MF		 1225 if (osb) {
1226 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1227 wake_up(&osb->osb_mount_event);
1228 ocfs2_dismount_volume(sb, 1);
1229 }
1230
c1e8d35e
TM		 1231 if (status)
1232 mlog_errno(status);
ccd979bd
MF		 1233 return status;
1234}
1235
152a0836 1236static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH		 1237 int flags,
1238 const char *dev_name,
152a0836 1239 void *data)
ccd979bd 1240{
152a0836 1241 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF		 1242}
1243
1244static struct file_system_type ocfs2_fs_type = {
1245 .owner = THIS_MODULE,
1246 .name = "ocfs2",
152a0836 1247 .mount = ocfs2_mount,
8ed6b237 1248 .kill_sb = kill_block_super,
1ba9da2f 1249 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1250 .next = NULL
1251};
91417705 1252MODULE_ALIAS_FS("ocfs2");
ccd979bd 1253
5297aad8
JK		 1254static int ocfs2_check_set_options(struct super_block *sb,
1255 struct mount_options *options)
1256{
1257 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1258 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1259 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1260 mlog(ML_ERROR, "User quotas were requested, but this "
1261 "filesystem does not have the feature enabled.\n");
1262 return 0;
1263 }
1264 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1265 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1266 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1267 mlog(ML_ERROR, "Group quotas were requested, but this "
1268 "filesystem does not have the feature enabled.\n");
1269 return 0;
1270 }
1271 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1272 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1273 mlog(ML_ERROR, "ACL support requested but extended attributes "
1274 "feature is not enabled\n");
1275 return 0;
1276 }
1277 /* No ACL setting specified? Use XATTR feature... */
1278 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1279 OCFS2_MOUNT_NO_POSIX_ACL))) {
1280 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1281 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1282 else
1283 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1284 }
1285 return 1;
1286}
1287
ccd979bd
MF		 1288static int ocfs2_parse_options(struct super_block *sb,
1289 char *options,
c0123ade 1290 struct mount_options *mopt,
ccd979bd
MF		 1291 int is_remount)
1292{
52c303c5 1293 int status, user_stack = 0;
ccd979bd 1294 char *p;
2c442719 1295 u32 tmp;
ccd979bd 1296
32a42d39 1297 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1298
d147b3d6 1299 mopt->commit_interval = 0;
4b37fcb7 1300 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY		 1301 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1302 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1303 mopt->localalloc_opt = -1;
b61817e1 1304 mopt->cluster_stack[0] = '\0';
d02f00cc 1305 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1306 mopt->dir_resv_level = -1;
ccd979bd
MF		 1307
1308 if (!options) {
1309 status = 1;
1310 goto bail;
1311 }
1312
1313 while ((p = strsep(&options, ",")) != NULL) {
1314 int token, option;
1315 substring_t args[MAX_OPT_ARGS];
1316
1317 if (!*p)
1318 continue;
1319
1320 token = match_token(p, tokens, args);
1321 switch (token) {
1322 case Opt_hb_local:
c0123ade 1323 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1324 break;
1325 case Opt_hb_none:
2c442719
SM		 1326 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1327 break;
1328 case Opt_hb_global:
1329 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF		 1330 break;
1331 case Opt_barrier:
1332 if (match_int(&args[0], &option)) {
1333 status = 0;
1334 goto bail;
1335 }
1336 if (option)
c0123ade 1337 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1338 else
c0123ade 1339 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1340 break;
1341 case Opt_intr:
c0123ade 1342 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1343 break;
1344 case Opt_nointr:
c0123ade 1345 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1346 break;
1347 case Opt_err_panic:
c0123ade 1348 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1349 break;
1350 case Opt_err_ro:
c0123ade 1351 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1352 break;
1353 case Opt_data_ordered:
c0123ade 1354 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1355 break;
1356 case Opt_data_writeback:
c0123ade 1357 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1358 break;
cf1d6c76
TY		 1359 case Opt_user_xattr:
1360 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1361 break;
1362 case Opt_nouser_xattr:
1363 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1364 break;
7f1a37e3
TY		 1365 case Opt_atime_quantum:
1366 if (match_int(&args[0], &option)) {
1367 status = 0;
1368 goto bail;
1369 }
1370 if (option >= 0)
c0123ade 1371 mopt->atime_quantum = option;
7f1a37e3 1372 break;
baf4661a
SM		 1373 case Opt_slot:
1374 option = 0;
1375 if (match_int(&args[0], &option)) {
1376 status = 0;
1377 goto bail;
1378 }
1379 if (option)
c0123ade 1380 mopt->slot = (s16)option;
baf4661a 1381 break;
d147b3d6
MF		 1382 case Opt_commit:
1383 option = 0;
1384 if (match_int(&args[0], &option)) {
1385 status = 0;
1386 goto bail;
1387 }
1388 if (option < 0)
1389 return 0;
1390 if (option == 0)
2b4e30fb 1391 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1392 mopt->commit_interval = HZ * option;
1393 break;
2fbe8d1e
SM		 1394 case Opt_localalloc:
1395 option = 0;
1396 if (match_int(&args[0], &option)) {
1397 status = 0;
1398 goto bail;
1399 }
73c8a800 1400 if (option >= 0)
2fbe8d1e
SM		 1401 mopt->localalloc_opt = option;
1402 break;
53fc622b
MF		 1403 case Opt_localflocks:
1404 /*
1405 * Changing this during remount could race
1406 * flock() requests, or "unbalance" existing
1407 * ones (e.g., a lock is taken in one mode but
1408 * dropped in the other). If users care enough
1409 * to flip locking modes during remount, we
1410 * could add a "local" flag to individual
1411 * flock structures for proper tracking of
1412 * state.
1413 */
1414 if (!is_remount)
1415 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1416 break;
b61817e1
JB		 1417 case Opt_stack:
1418 /* Check both that the option we were passed
1419 * is of the right length and that it is a proper
1420 * string of the right length.
1421 */
1422 if (((args[0].to - args[0].from) !=
1423 OCFS2_STACK_LABEL_LEN) ||
1424 (strnlen(args[0].from,
1425 OCFS2_STACK_LABEL_LEN) !=
1426 OCFS2_STACK_LABEL_LEN)) {
1427 mlog(ML_ERROR,
1428 "Invalid cluster_stack option\n");
1429 status = 0;
1430 goto bail;
1431 }
1432 memcpy(mopt->cluster_stack, args[0].from,
1433 OCFS2_STACK_LABEL_LEN);
1434 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF		 1435 /*
1436 * Open code the memcmp here as we don't have
1437 * an osb to pass to
1438 * ocfs2_userspace_stack().
1439 */
1440 if (memcmp(mopt->cluster_stack,
1441 OCFS2_CLASSIC_CLUSTER_STACK,
1442 OCFS2_STACK_LABEL_LEN))
1443 user_stack = 1;
b61817e1 1444 break;
12462f1d
JB		 1445 case Opt_inode64:
1446 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1447 break;
19ece546 1448 case Opt_usrquota:
19ece546
JK		 1449 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1450 break;
1451 case Opt_grpquota:
19ece546
JK		 1452 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1453 break;
7bdb0d18
TY		 1454 case Opt_coherency_buffered:
1455 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1456 break;
1457 case Opt_coherency_full:
1458 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1459 break;
a68979b8
TY		 1460 case Opt_acl:
1461 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1462 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1463 break;
1464 case Opt_noacl:
5297aad8 1465 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1466 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1467 break;
d02f00cc
MF		 1468 case Opt_resv_level:
1469 if (is_remount)
1470 break;
1471 if (match_int(&args[0], &option)) {
1472 status = 0;
1473 goto bail;
1474 }
1475 if (option >= OCFS2_MIN_RESV_LEVEL &&
1476 option < OCFS2_MAX_RESV_LEVEL)
1477 mopt->resv_level = option;
1478 break;
83f92318
MF		 1479 case Opt_dir_resv_level:
1480 if (is_remount)
1481 break;
1482 if (match_int(&args[0], &option)) {
1483 status = 0;
1484 goto bail;
1485 }
1486 if (option >= OCFS2_MIN_RESV_LEVEL &&
1487 option < OCFS2_MAX_RESV_LEVEL)
1488 mopt->dir_resv_level = option;
1489 break;
ccd979bd
MF		 1490 default:
1491 mlog(ML_ERROR,
1492 "Unrecognized mount option \"%s\" "
1493 "or missing value\n", p);
1494 status = 0;
1495 goto bail;
1496 }
1497 }
1498
52c303c5
MF		 1499 if (user_stack == 0) {
1500 /* Ensure only one heartbeat mode */
1501 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1502 OCFS2_MOUNT_HB_GLOBAL |
1503 OCFS2_MOUNT_HB_NONE);
1504 if (hweight32(tmp) != 1) {
1505 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1506 status = 0;
1507 goto bail;
1508 }
2c442719
SM		 1509 }
1510
ccd979bd
MF		 1511 status = 1;
1512
1513bail:
ccd979bd
MF		 1514 return status;
1515}
1516
34c80b1d 1517static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1518{
34c80b1d 1519 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1520 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1521 unsigned int local_alloc_megs;
d550071c 1522
2c442719
SM		 1523 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1524 seq_printf(s, ",_netdev");
1525 if (opts & OCFS2_MOUNT_HB_LOCAL)
1526 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1527 else
1528 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1529 } else
1530 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM		 1531
1532 if (opts & OCFS2_MOUNT_NOINTR)
1533 seq_printf(s, ",nointr");
1534
1535 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1536 seq_printf(s, ",data=writeback");
1537 else
1538 seq_printf(s, ",data=ordered");
1539
1540 if (opts & OCFS2_MOUNT_BARRIER)
1541 seq_printf(s, ",barrier=1");
1542
1543 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1544 seq_printf(s, ",errors=panic");
1545 else
1546 seq_printf(s, ",errors=remount-ro");
1547
1548 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1549 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1550
34c80b1d 1551 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1552
d147b3d6
MF		 1553 if (osb->osb_commit_interval)
1554 seq_printf(s, ",commit=%u",
1555 (unsigned) (osb->osb_commit_interval / HZ));
1556
ebcee4b5 1557 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1558 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1559 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1560
53fc622b
MF		 1561 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1562 seq_printf(s, ",localflocks,");
1563
b61817e1
JB		 1564 if (osb->osb_cluster_stack[0])
1565 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1566 osb->osb_cluster_stack);
19ece546
JK		 1567 if (opts & OCFS2_MOUNT_USRQUOTA)
1568 seq_printf(s, ",usrquota");
1569 if (opts & OCFS2_MOUNT_GRPQUOTA)
1570 seq_printf(s, ",grpquota");
b61817e1 1571
7bdb0d18
TY		 1572 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1573 seq_printf(s, ",coherency=buffered");
1574 else
1575 seq_printf(s, ",coherency=full");
1576
b0f73cfc
SM		 1577 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1578 seq_printf(s, ",nouser_xattr");
1579 else
1580 seq_printf(s, ",user_xattr");
1581
12462f1d
JB		 1582 if (opts & OCFS2_MOUNT_INODE64)
1583 seq_printf(s, ",inode64");
1584
a68979b8
TY		 1585 if (opts & OCFS2_MOUNT_POSIX_ACL)
1586 seq_printf(s, ",acl");
1587 else
1588 seq_printf(s, ",noacl");
a68979b8 1589
d02f00cc
MF		 1590 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1591 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1592
83f92318
MF		 1593 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1594 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1595
d550071c
SM		 1596 return 0;
1597}
1598
ccd979bd
MF		 1599static int __init ocfs2_init(void)
1600{
c18ceab0 1601 int status;
a11f7e63 1602
ccd979bd 1603 status = init_ocfs2_uptodate_cache();
342827d7
AV		 1604 if (status < 0)
1605 goto out1;
ccd979bd
MF		 1606
1607 status = ocfs2_initialize_mem_caches();
342827d7
AV		 1608 if (status < 0)
1609 goto out2;
ccd979bd
MF		 1610
1611 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1612 if (!ocfs2_wq) {
1613 status = -ENOMEM;
342827d7 1614 goto out3;
ccd979bd
MF		 1615 }
1616
ccd979bd
MF		 1617 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1618 if (!ocfs2_debugfs_root) {
1619 status = -EFAULT;
1620 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1621 }
1622
63e0c48a
JB		 1623 ocfs2_set_locking_protocol();
1624
9e33d69f 1625 status = register_quota_format(&ocfs2_quota_format);
342827d7
AV		 1626 if (status < 0)
1627 goto out4;
1628 status = register_filesystem(&ocfs2_fs_type);
1629 if (!status)
1630 return 0;
ccd979bd 1631
342827d7
AV		 1632 unregister_quota_format(&ocfs2_quota_format);
1633out4:
1634 destroy_workqueue(ocfs2_wq);
1635 debugfs_remove(ocfs2_debugfs_root);
1636out3:
1637 ocfs2_free_mem_caches();
1638out2:
1639 exit_ocfs2_uptodate_cache();
1640out1:
1641 mlog_errno(status);
1642 return status;
ccd979bd
MF		 1643}
1644
1645static void __exit ocfs2_exit(void)
1646{
ccd979bd
MF		 1647 if (ocfs2_wq) {
1648 flush_workqueue(ocfs2_wq);
1649 destroy_workqueue(ocfs2_wq);
1650 }
1651
9e33d69f
JK		 1652 unregister_quota_format(&ocfs2_quota_format);
1653
ccd979bd
MF		 1654 debugfs_remove(ocfs2_debugfs_root);
1655
1656 ocfs2_free_mem_caches();
1657
1658 unregister_filesystem(&ocfs2_fs_type);
1659
ccd979bd 1660 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1661}
1662
1663static void ocfs2_put_super(struct super_block *sb)
1664{
32a42d39 1665 trace_ocfs2_put_super(sb);
ccd979bd
MF		 1666
1667 ocfs2_sync_blockdev(sb);
1668 ocfs2_dismount_volume(sb, 0);
ccd979bd
MF		 1669}
1670
726c3342 1671static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1672{
1673 struct ocfs2_super *osb;
1674 u32 numbits, freebits;
1675 int status;
1676 struct ocfs2_dinode *bm_lock;
1677 struct buffer_head *bh = NULL;
1678 struct inode *inode = NULL;
1679
32a42d39 1680 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1681
726c3342 1682 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1683
1684 inode = ocfs2_get_system_file_inode(osb,
1685 GLOBAL_BITMAP_SYSTEM_INODE,
1686 OCFS2_INVALID_SLOT);
1687 if (!inode) {
1688 mlog(ML_ERROR, "failed to get bitmap inode\n");
1689 status = -EIO;
1690 goto bail;
1691 }
1692
e63aecb6 1693 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1694 if (status < 0) {
1695 mlog_errno(status);
1696 goto bail;
1697 }
1698
1699 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1700
1701 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1702 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1703
1704 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1705 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1706 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1707 buf->f_blocks = ((sector_t) numbits) *
1708 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1709 buf->f_bfree = ((sector_t) freebits) *
1710 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1711 buf->f_bavail = buf->f_bfree;
1712 buf->f_files = numbits;
1713 buf->f_ffree = freebits;
837711f8
CL		 1714 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1715 & 0xFFFFFFFFUL;
1716 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1717 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1718
1719 brelse(bh);
1720
e63aecb6 1721 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1722 status = 0;
1723bail:
1724 if (inode)
1725 iput(inode);
1726
c1e8d35e
TM		 1727 if (status)
1728 mlog_errno(status);
ccd979bd
MF		 1729
1730 return status;
1731}
1732
51cc5068 1733static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1734{
1735 struct ocfs2_inode_info *oi = data;
1736
a35afb83
CL		 1737 oi->ip_flags = 0;
1738 oi->ip_open_count = 0;
1739 spin_lock_init(&oi->ip_lock);
1740 ocfs2_extent_map_init(&oi->vfs_inode);
1741 INIT_LIST_HEAD(&oi->ip_io_markers);
a35afb83 1742 oi->ip_dir_start_lookup = 0;
c18ceab0 1743 mutex_init(&oi->ip_unaligned_aio);
a35afb83 1744 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1745 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1746 mutex_init(&oi->ip_io_mutex);
ccd979bd 1747
a35afb83
CL		 1748 oi->ip_blkno = 0ULL;
1749 oi->ip_clusters = 0;
ccd979bd 1750
4fe370af
MF		 1751 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1752
a35afb83 1753 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1754 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1755 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1756
8cb471e8 1757 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1758 &ocfs2_inode_caching_ops);
ccd979bd 1759
a35afb83 1760 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1761}
1762
1763static int ocfs2_initialize_mem_caches(void)
1764{
1765 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1766 sizeof(struct ocfs2_inode_info),
1767 0,
1768 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1769 SLAB_MEM_SPREAD),
20c2df83 1770 ocfs2_inode_init_once);
9e33d69f
JK		 1771 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1772 sizeof(struct ocfs2_dquot),
1773 0,
1774 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1775 SLAB_MEM_SPREAD),
1776 NULL);
1777 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1778 sizeof(struct ocfs2_quota_chunk),
1779 0,
1780 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1781 NULL);
1782 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1783 !ocfs2_qf_chunk_cachep) {
1784 if (ocfs2_inode_cachep)
1785 kmem_cache_destroy(ocfs2_inode_cachep);
1786 if (ocfs2_dquot_cachep)
1787 kmem_cache_destroy(ocfs2_dquot_cachep);
1788 if (ocfs2_qf_chunk_cachep)
1789 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1790 return -ENOMEM;
9e33d69f 1791 }
ccd979bd 1792
ccd979bd
MF		 1793 return 0;
1794}
1795
1796static void ocfs2_free_mem_caches(void)
1797{
8c0a8537
KS		 1798 /*
1799 * Make sure all delayed rcu free inodes are flushed before we
1800 * destroy cache.
1801 */
1802 rcu_barrier();
ccd979bd
MF		 1803 if (ocfs2_inode_cachep)
1804 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1805 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1806
1807 if (ocfs2_dquot_cachep)
1808 kmem_cache_destroy(ocfs2_dquot_cachep);
1809 ocfs2_dquot_cachep = NULL;
1810
1811 if (ocfs2_qf_chunk_cachep)
1812 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1813 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1814}
1815
1816static int ocfs2_get_sector(struct super_block *sb,
1817 struct buffer_head **bh,
1818 int block,
1819 int sect_size)
1820{
1821 if (!sb_set_blocksize(sb, sect_size)) {
1822 mlog(ML_ERROR, "unable to set blocksize\n");
1823 return -EIO;
1824 }
1825
1826 *bh = sb_getblk(sb, block);
1827 if (!*bh) {
7391a294
RX		 1828 mlog_errno(-ENOMEM);
1829 return -ENOMEM;
ccd979bd
MF		 1830 }
1831 lock_buffer(*bh);
1832 if (!buffer_dirty(*bh))
1833 clear_buffer_uptodate(*bh);
1834 unlock_buffer(*bh);
1835 ll_rw_block(READ, 1, bh);
1836 wait_on_buffer(*bh);
28d57d43 1837 if (!buffer_uptodate(*bh)) {
1838 mlog_errno(-EIO);
1839 brelse(*bh);
1840 *bh = NULL;
1841 return -EIO;
1842 }
1843
ccd979bd
MF		 1844 return 0;
1845}
1846
ccd979bd
MF		 1847static int ocfs2_mount_volume(struct super_block *sb)
1848{
1849 int status = 0;
1850 int unlock_super = 0;
1851 struct ocfs2_super *osb = OCFS2_SB(sb);
1852
ccd979bd
MF		 1853 if (ocfs2_is_hard_readonly(osb))
1854 goto leave;
1855
ccd979bd
MF		 1856 status = ocfs2_dlm_init(osb);
1857 if (status < 0) {
1858 mlog_errno(status);
1859 goto leave;
1860 }
1861
ccd979bd
MF		 1862 status = ocfs2_super_lock(osb, 1);
1863 if (status < 0) {
1864 mlog_errno(status);
1865 goto leave;
1866 }
1867 unlock_super = 1;
1868
1869 /* This will load up the node map and add ourselves to it. */
1870 status = ocfs2_find_slot(osb);
1871 if (status < 0) {
1872 mlog_errno(status);
1873 goto leave;
1874 }
1875
ccd979bd
MF		 1876 /* load all node-local system inodes */
1877 status = ocfs2_init_local_system_inodes(osb);
1878 if (status < 0) {
1879 mlog_errno(status);
1880 goto leave;
1881 }
1882
1883 status = ocfs2_check_volume(osb);
1884 if (status < 0) {
1885 mlog_errno(status);
1886 goto leave;
1887 }
1888
1889 status = ocfs2_truncate_log_init(osb);
06c59bb8 1890 if (status < 0)
ccd979bd 1891 mlog_errno(status);
c271c5c2 1892
ccd979bd
MF		 1893leave:
1894 if (unlock_super)
1895 ocfs2_super_unlock(osb, 1);
1896
ccd979bd
MF		 1897 return status;
1898}
1899
ccd979bd
MF		 1900static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1901{
286eaa95 1902 int tmp, hangup_needed = 0;
ccd979bd 1903 struct ocfs2_super *osb = NULL;
99d7a882 1904 char nodestr[12];
ccd979bd 1905
32a42d39 1906 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF		 1907
1908 BUG_ON(!sb);
1909 osb = OCFS2_SB(sb);
1910 BUG_ON(!osb);
1911
50397507
SM		 1912 debugfs_remove(osb->osb_ctxt);
1913
692684e1
SM		 1914 /* Orphan scan should be stopped as early as possible */
1915 ocfs2_orphan_scan_stop(osb);
1916
19ece546
JK		 1917 ocfs2_disable_quotas(osb);
1918
e3a767b6
JK		 1919 /* All dquots should be freed by now */
1920 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1921 /* Wait for worker to be done with the work structure in osb */
1922 cancel_work_sync(&osb->dquot_drop_work);
1923
ccd979bd
MF		 1924 ocfs2_shutdown_local_alloc(osb);
1925
553abd04
JB		 1926 /* This will disable recovery and flush any recovery work. */
1927 ocfs2_recovery_exit(osb);
ccd979bd 1928
75f82eaa
YJ		 1929 /*
1930 * During dismount, when it recovers another node it will call
1931 * ocfs2_recover_orphans and queue delayed work osb_truncate_log_wq.
1932 */
1933 ocfs2_truncate_log_shutdown(osb);
1934
ccd979bd
MF		 1935 ocfs2_journal_shutdown(osb);
1936
1937 ocfs2_sync_blockdev(sb);
1938
374a263e
TM		 1939 ocfs2_purge_refcount_trees(osb);
1940
4670c46d
JB		 1941 /* No cluster connection means we've failed during mount, so skip
1942 * all the steps which depended on that to complete. */
1943 if (osb->cconn) {
ccd979bd
MF		 1944 tmp = ocfs2_super_lock(osb, 1);
1945 if (tmp < 0) {
1946 mlog_errno(tmp);
1947 return;
1948 }
19b613d4 1949 }
ccd979bd 1950
19b613d4
MF		 1951 if (osb->slot_num != OCFS2_INVALID_SLOT)
1952 ocfs2_put_slot(osb);
ccd979bd 1953
4670c46d 1954 if (osb->cconn)
ccd979bd 1955 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1956
1957 ocfs2_release_system_inodes(osb);
1958
6953b4c0 1959 /*
6953b4c0
JB		 1960 * If we're dismounting due to mount error, mount.ocfs2 will clean
1961 * up heartbeat. If we're a local mount, there is no heartbeat.
1962 * If we failed before we got a uuid_str yet, we can't stop
1963 * heartbeat. Otherwise, do it.
1964 */
03efed8a
TY		 1965 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1966 !ocfs2_is_hard_readonly(osb))
286eaa95
JB		 1967 hangup_needed = 1;
1968
1969 if (osb->cconn)
1970 ocfs2_dlm_shutdown(osb, hangup_needed);
1971
73be192b 1972 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1973 debugfs_remove(osb->osb_debug_root);
1974
1975 if (hangup_needed)
6953b4c0 1976 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1977
1978 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1979
c271c5c2
SM		 1980 if (ocfs2_mount_local(osb))
1981 snprintf(nodestr, sizeof(nodestr), "local");
1982 else
19fdb624 1983 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1984
1985 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1986 osb->dev_str, nodestr);
ccd979bd
MF		 1987
1988 ocfs2_delete_osb(osb);
1989 kfree(osb);
1990 sb->s_dev = 0;
1991 sb->s_fs_info = NULL;
1992}
1993
1994static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1995 unsigned uuid_bytes)
1996{
1997 int i, ret;
1998 char *ptr;
1999
2000 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2001
cd861280 2002 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 2003 if (osb->uuid_str == NULL)
2004 return -ENOMEM;
2005
ccd979bd
MF		 2006 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2007 /* print with null */
2008 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2009 if (ret != 2) /* drop super cleans up */
2010 return -EINVAL;
2011 /* then only advance past the last char */
2012 ptr += 2;
2013 }
2014
2015 return 0;
2016}
2017
3bdb8efd
PL		 2018/* Make sure entire volume is addressable by our journal. Requires
2019 osb_clusters_at_boot to be valid and for the journal to have been
2020 initialized by ocfs2_journal_init(). */
2021static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2022{
2023 int status = 0;
2024 u64 max_block =
2025 ocfs2_clusters_to_blocks(osb->sb,
2026 osb->osb_clusters_at_boot) - 1;
2027
2028 /* 32-bit block number is always OK. */
2029 if (max_block <= (u32)~0ULL)
2030 goto out;
2031
2032 /* Volume is "huge", so see if our journal is new enough to
2033 support it. */
2034 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2035 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2036 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2037 JBD2_FEATURE_INCOMPAT_64BIT))) {
2038 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2039 "Enable the 'block64' journal option with tunefs.ocfs2");
2040 status = -EFBIG;
2041 goto out;
2042 }
2043
2044 out:
2045 return status;
2046}
2047
ccd979bd
MF		 2048static int ocfs2_initialize_super(struct super_block *sb,
2049 struct buffer_head *bh,
73be192b
JB		 2050 int sector_size,
2051 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2052{
bddb8eb3 2053 int status;
5a254031
MF		 2054 int i, cbits, bbits;
2055 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2056 struct inode *inode = NULL;
ccd979bd 2057 struct ocfs2_journal *journal;
ccd979bd 2058 struct ocfs2_super *osb;
3bdb8efd 2059 u64 total_blocks;
ccd979bd 2060
cd861280 2061 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 2062 if (!osb) {
2063 status = -ENOMEM;
2064 mlog_errno(status);
2065 goto bail;
2066 }
2067
2068 sb->s_fs_info = osb;
2069 sb->s_op = &ocfs2_sops;
ba87167c 2070 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2071 sb->s_export_op = &ocfs2_export_ops;
19ece546
JK		 2072 sb->s_qcop = &ocfs2_quotactl_ops;
2073 sb->dq_op = &ocfs2_quota_operations;
cf1d6c76 2074 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2075 sb->s_time_gran = 1;
ccd979bd
MF		 2076 sb->s_flags |= MS_NOATIME;
2077 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2078 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2079 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2080 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd 2081
9b7895ef
MF		 2082 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2083
2084 for (i = 0; i < 3; i++)
2085 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2086 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2087
ccd979bd
MF		 2088 osb->sb = sb;
2089 /* Save off for ocfs2_rw_direct */
2090 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2091 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2092
34d024f8
MF		 2093 spin_lock_init(&osb->dc_task_lock);
2094 init_waitqueue_head(&osb->dc_event);
2095 osb->dc_work_sequence = 0;
2096 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2097 INIT_LIST_HEAD(&osb->blocked_lock_list);
2098 osb->blocked_lock_count = 0;
ccd979bd 2099 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2100 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2101 ocfs2_init_steal_slots(osb);
ccd979bd 2102
43b10a20 2103 mutex_init(&osb->system_file_mutex);
2104
ccd979bd
MF		 2105 atomic_set(&osb->alloc_stats.moves, 0);
2106 atomic_set(&osb->alloc_stats.local_data, 0);
2107 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2108 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2109 atomic_set(&osb->alloc_stats.bg_extends, 0);
2110
73be192b
JB		 2111 /* Copy the blockcheck stats from the superblock probe */
2112 osb->osb_ecc_stats = *stats;
2113
ccd979bd
MF		 2114 ocfs2_init_node_maps(osb);
2115
2116 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2117 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2118
75d9bbc7
GR		 2119 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2120 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2121 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2122 osb->max_slots);
2123 status = -EINVAL;
2124 goto bail;
2125 }
75d9bbc7 2126
8b712cd5
JM		 2127 ocfs2_orphan_scan_init(osb);
2128
553abd04
JB		 2129 status = ocfs2_recovery_init(osb);
2130 if (status) {
2131 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2132 mlog_errno(status);
2133 goto bail;
2134 }
ccd979bd
MF		 2135
2136 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2137
7f1a37e3
TY		 2138 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2139
ccd979bd
MF		 2140 osb->slot_num = OCFS2_INVALID_SLOT;
2141
8154da3d
TY		 2142 osb->s_xattr_inline_size = le16_to_cpu(
2143 di->id2.i_super.s_xattr_inline_size);
fdd77704 2144
ccd979bd
MF		 2145 osb->local_alloc_state = OCFS2_LA_UNUSED;
2146 osb->local_alloc_bh = NULL;
9c7af40b 2147 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2148
ccd979bd
MF		 2149 init_waitqueue_head(&osb->osb_mount_event);
2150
d02f00cc
MF		 2151 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2152 if (status) {
2153 mlog_errno(status);
2154 goto bail;
2155 }
2156
ccd979bd
MF		 2157 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2158 if (!osb->vol_label) {
2159 mlog(ML_ERROR, "unable to alloc vol label\n");
2160 status = -ENOMEM;
2161 goto bail;
2162 }
2163
539d8264
SM		 2164 osb->slot_recovery_generations =
2165 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2166 GFP_KERNEL);
2167 if (!osb->slot_recovery_generations) {
2168 status = -ENOMEM;
2169 mlog_errno(status);
2170 goto bail;
2171 }
2172
b4df6ed8
MF		 2173 init_waitqueue_head(&osb->osb_wipe_event);
2174 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2175 sizeof(*osb->osb_orphan_wipes),
2176 GFP_KERNEL);
2177 if (!osb->osb_orphan_wipes) {
2178 status = -ENOMEM;
2179 mlog_errno(status);
2180 goto bail;
2181 }
2182
374a263e
TM		 2183 osb->osb_rf_lock_tree = RB_ROOT;
2184
ccd979bd
MF		 2185 osb->s_feature_compat =
2186 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2187 osb->s_feature_ro_compat =
2188 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2189 osb->s_feature_incompat =
2190 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2191
2192 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2193 mlog(ML_ERROR, "couldn't mount because of unsupported "
2194 "optional features (%x).\n", i);
2195 status = -EINVAL;
2196 goto bail;
2197 }
2198 if (!(osb->sb->s_flags & MS_RDONLY) &&
2199 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2200 mlog(ML_ERROR, "couldn't mount RDWR because of "
2201 "unsupported optional features (%x).\n", i);
2202 status = -EINVAL;
2203 goto bail;
2204 }
2205
98f486f2
SM		 2206 if (ocfs2_clusterinfo_valid(osb)) {
2207 osb->osb_stackflags =
2208 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2209 strlcpy(osb->osb_cluster_stack,
b61817e1 2210 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2211 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB		 2212 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2213 mlog(ML_ERROR,
2214 "couldn't mount because of an invalid "
2215 "cluster stack label (%s) \n",
2216 osb->osb_cluster_stack);
2217 status = -EINVAL;
2218 goto bail;
2219 }
c74a3bdd
GR		 2220 strlcpy(osb->osb_cluster_name,
2221 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2222 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB		 2223 } else {
2224 /* The empty string is identical with classic tools that
2225 * don't know about s_cluster_info. */
2226 osb->osb_cluster_stack[0] = '\0';
2227 }
2228
ccd979bd
MF		 2229 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2230
2231 /* FIXME
2232 * This should be done in ocfs2_journal_init(), but unknown
2233 * ordering issues will cause the filesystem to crash.
2234 * If anyone wants to figure out what part of the code
2235 * refers to osb->journal before ocfs2_journal_init() is run,
2236 * be my guest.
2237 */
2238 /* initialize our journal structure */
2239
cd861280 2240 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2241 if (!journal) {
2242 mlog(ML_ERROR, "unable to alloc journal\n");
2243 status = -ENOMEM;
2244 goto bail;
2245 }
2246 osb->journal = journal;
2247 journal->j_osb = osb;
2248
2249 atomic_set(&journal->j_num_trans, 0);
2250 init_rwsem(&journal->j_trans_barrier);
2251 init_waitqueue_head(&journal->j_checkpointed);
2252 spin_lock_init(&journal->j_lock);
2253 journal->j_trans_id = (unsigned long) 1;
2254 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2255 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2256 journal->j_state = OCFS2_JOURNAL_FREE;
2257
e3a767b6
JK		 2258 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2259 init_llist_head(&osb->dquot_drop_list);
2260
ccd979bd
MF		 2261 /* get some pseudo constants for clustersize bits */
2262 osb->s_clustersize_bits =
2263 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2264 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF		 2265
2266 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2267 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2268 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2269 osb->s_clustersize);
2270 status = -EINVAL;
2271 goto bail;
2272 }
2273
3bdb8efd
PL		 2274 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2275 le32_to_cpu(di->i_clusters));
2276
2277 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2278 total_blocks);
2279 if (status) {
2280 mlog(ML_ERROR, "Volume too large "
2281 "to mount safely on this system");
2282 status = -EFBIG;
ccd979bd
MF		 2283 goto bail;
2284 }
2285
2286 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2287 sizeof(di->id2.i_super.s_uuid))) {
2288 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2289 status = -ENOMEM;
2290 goto bail;
2291 }
2292
ccd979bd
MF		 2293 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2294 osb->vol_label[63] = '\0';
2295 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2296 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2297 osb->first_cluster_group_blkno =
2298 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2299 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2300 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM		 2301 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2302 (unsigned long long)osb->root_blkno,
2303 (unsigned long long)osb->system_dir_blkno,
2304 osb->s_clustersize_bits);
ccd979bd
MF		 2305
2306 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2307 if (!osb->osb_dlm_debug) {
2308 status = -ENOMEM;
2309 mlog_errno(status);
2310 goto bail;
2311 }
2312
2313 atomic_set(&osb->vol_state, VOLUME_INIT);
2314
2315 /* load root, system_dir, and all global system inodes */
2316 status = ocfs2_init_global_system_inodes(osb);
2317 if (status < 0) {
2318 mlog_errno(status);
2319 goto bail;
2320 }
2321
2322 /*
2323 * global bitmap
2324 */
2325 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2326 OCFS2_INVALID_SLOT);
2327 if (!inode) {
2328 status = -EINVAL;
2329 mlog_errno(status);
2330 goto bail;
2331 }
2332
2333 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2334 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2335 iput(inode);
ccd979bd 2336
8571882c
TM		 2337 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2338 osb->s_feature_incompat) * 8;
ccd979bd
MF		 2339
2340 status = ocfs2_init_slot_info(osb);
2341 if (status < 0) {
2342 mlog_errno(status);
2343 goto bail;
2344 }
e41d33af 2345 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
ccd979bd 2346
ccd979bd 2347bail:
ccd979bd
MF		 2348 return status;
2349}
2350
2351/*
2352 * will return: -EAGAIN if it is ok to keep searching for superblocks
2353 * -EINVAL if there is a bad superblock
2354 * 0 on success
2355 */
2356static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2357 struct buffer_head *bh,
73be192b
JB		 2358 u32 blksz,
2359 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2360{
2361 int status = -EAGAIN;
2362
ccd979bd
MF		 2363 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2364 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2365 /* We have to do a raw check of the feature here */
2366 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2367 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2368 status = ocfs2_block_check_validate(bh->b_data,
2369 bh->b_size,
73be192b
JB		 2370 &di->i_check,
2371 stats);
d030cc97
JB		 2372 if (status)
2373 goto out;
2374 }
ccd979bd
MF		 2375 status = -EINVAL;
2376 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2377 mlog(ML_ERROR, "found superblock with incorrect block "
2378 "size: found %u, should be %u\n",
2379 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2380 blksz);
2381 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2382 OCFS2_MAJOR_REV_LEVEL ||
2383 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2384 OCFS2_MINOR_REV_LEVEL) {
2385 mlog(ML_ERROR, "found superblock with bad version: "
2386 "found %u.%u, should be %u.%u\n",
2387 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2388 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2389 OCFS2_MAJOR_REV_LEVEL,
2390 OCFS2_MINOR_REV_LEVEL);
2391 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2392 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2393 "found %llu, should be %llu\n",
1ca1a111 2394 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2395 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2396 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2397 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2398 mlog(ML_ERROR, "bad cluster size found: %u\n",
2399 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2400 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2401 mlog(ML_ERROR, "bad root_blkno: 0\n");
2402 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2403 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2404 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2405 mlog(ML_ERROR,
2406 "Superblock slots found greater than file system "
2407 "maximum: found %u, max %u\n",
2408 le16_to_cpu(di->id2.i_super.s_max_slots),
2409 OCFS2_MAX_SLOTS);
2410 } else {
2411 /* found it! */
2412 status = 0;
2413 }
2414 }
2415
d030cc97 2416out:
c1e8d35e
TM		 2417 if (status && status != -EAGAIN)
2418 mlog_errno(status);
ccd979bd
MF		 2419 return status;
2420}
2421
2422static int ocfs2_check_volume(struct ocfs2_super *osb)
2423{
bddb8eb3 2424 int status;
ccd979bd 2425 int dirty;
c271c5c2 2426 int local;
ccd979bd
MF		 2427 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2428 * recover
2429 * ourselves. */
2430
ccd979bd
MF		 2431 /* Init our journal object. */
2432 status = ocfs2_journal_init(osb->journal, &dirty);
2433 if (status < 0) {
2434 mlog(ML_ERROR, "Could not initialize journal!\n");
2435 goto finally;
2436 }
2437
3bdb8efd
PL		 2438 /* Now that journal has been initialized, check to make sure
2439 entire volume is addressable. */
2440 status = ocfs2_journal_addressable(osb);
2441 if (status)
2442 goto finally;
2443
ccd979bd
MF		 2444 /* If the journal was unmounted cleanly then we don't want to
2445 * recover anything. Otherwise, journal_load will do that
2446 * dirty work for us :) */
2447 if (!dirty) {
2448 status = ocfs2_journal_wipe(osb->journal, 0);
2449 if (status < 0) {
2450 mlog_errno(status);
2451 goto finally;
2452 }
2453 } else {
619c200d
SM		 2454 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2455 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF		 2456 }
2457
c271c5c2
SM		 2458 local = ocfs2_mount_local(osb);
2459
ccd979bd 2460 /* will play back anything left in the journal. */
539d8264 2461 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2462 if (status < 0) {
2463 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2464 goto finally;
2465 }
ccd979bd
MF		 2466
2467 if (dirty) {
2468 /* recover my local alloc if we didn't unmount cleanly. */
2469 status = ocfs2_begin_local_alloc_recovery(osb,
2470 osb->slot_num,
2471 &local_alloc);
2472 if (status < 0) {
2473 mlog_errno(status);
2474 goto finally;
2475 }
2476 /* we complete the recovery process after we've marked
2477 * ourselves as mounted. */
2478 }
2479
ccd979bd
MF		 2480 status = ocfs2_load_local_alloc(osb);
2481 if (status < 0) {
2482 mlog_errno(status);
2483 goto finally;
2484 }
2485
2486 if (dirty) {
2487 /* Recovery will be completed after we've mounted the
2488 * rest of the volume. */
2489 osb->dirty = 1;
2490 osb->local_alloc_copy = local_alloc;
2491 local_alloc = NULL;
2492 }
2493
2494 /* go through each journal, trylock it and if you get the
2495 * lock, and it's marked as dirty, set the bit in the recover
2496 * map and launch a recovery thread for it. */
2497 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2498 if (status < 0) {
2499 mlog_errno(status);
2500 goto finally;
2501 }
2502
2503 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2504 if (status < 0)
2505 mlog_errno(status);
2506
2507finally:
d787ab09 2508 kfree(local_alloc);
ccd979bd 2509
c1e8d35e
TM		 2510 if (status)
2511 mlog_errno(status);
ccd979bd
MF		 2512 return status;
2513}
2514
2515/*
2516 * The routine gets called from dismount or close whenever a dismount on
2517 * volume is requested and the osb open count becomes 1.
2518 * It will remove the osb from the global list and also free up all the
2519 * initialized resources and fileobject.
2520 */
2521static void ocfs2_delete_osb(struct ocfs2_super *osb)
2522{
ccd979bd
MF		 2523 /* This function assumes that the caller has the main osb resource */
2524
8e8a4603 2525 ocfs2_free_slot_info(osb);
ccd979bd 2526
b4df6ed8 2527 kfree(osb->osb_orphan_wipes);
539d8264 2528 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2529 /* FIXME
2530 * This belongs in journal shutdown, but because we have to
421f91d2 2531 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF		 2532 * we free it here.
2533 */
2534 kfree(osb->journal);
d787ab09 2535 kfree(osb->local_alloc_copy);
ccd979bd
MF		 2536 kfree(osb->uuid_str);
2537 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2538 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF		 2539}
2540
2541/* Put OCFS2 into a readonly state, or (if the user specifies it),
2542 * panic(). We do not support continue-on-error operation. */
2543static void ocfs2_handle_error(struct super_block *sb)
2544{
2545 struct ocfs2_super *osb = OCFS2_SB(sb);
2546
2547 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2548 panic("OCFS2: (device %s): panic forced after error\n",
2549 sb->s_id);
2550
2551 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2552
2553 if (sb->s_flags & MS_RDONLY &&
2554 (ocfs2_is_soft_readonly(osb) ||
2555 ocfs2_is_hard_readonly(osb)))
2556 return;
2557
2558 printk(KERN_CRIT "File system is now read-only due to the potential "
2559 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2560 "system is unmounted.\n");
2561 sb->s_flags |= MS_RDONLY;
2562 ocfs2_set_ro_flag(osb, 0);
2563}
2564
2565static char error_buf[1024];
2566
2567void __ocfs2_error(struct super_block *sb,
2568 const char *function,
2569 const char *fmt, ...)
2570{
2571 va_list args;
2572
2573 va_start(args, fmt);
4a6e617a 2574 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2575 va_end(args);
2576
2577 /* Not using mlog here because we want to show the actual
2578 * function the error came from. */
2579 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2580 sb->s_id, function, error_buf);
2581
2582 ocfs2_handle_error(sb);
2583}
2584
2585/* Handle critical errors. This is intentionally more drastic than
2586 * ocfs2_handle_error, so we only use for things like journal errors,
2587 * etc. */
2588void __ocfs2_abort(struct super_block* sb,
2589 const char *function,
2590 const char *fmt, ...)
2591{
2592 va_list args;
2593
2594 va_start(args, fmt);
4a6e617a 2595 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2596 va_end(args);
2597
2598 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2599 sb->s_id, function, error_buf);
2600
2601 /* We don't have the cluster support yet to go straight to
2602 * hard readonly in here. Until then, we want to keep
2603 * ocfs2_abort() so that we can at least mark critical
2604 * errors.
2605 *
2606 * TODO: This should abort the journal and alert other nodes
2607 * that our slot needs recovery. */
2608
2609 /* Force a panic(). This stinks, but it's better than letting
2610 * things continue without having a proper hard readonly
2611 * here. */
a2f2ddbf
SM		 2612 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2613 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2614 ocfs2_handle_error(sb);
2615}
2616
e4b963f1
JB		 2617/*
2618 * Void signal blockers, because in-kernel sigprocmask() only fails
2619 * when SIG_* is wrong.
2620 */
2621void ocfs2_block_signals(sigset_t *oldset)
2622{
2623 int rc;
2624 sigset_t blocked;
2625
2626 sigfillset(&blocked);
2627 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2628 BUG_ON(rc);
2629}
2630
2631void ocfs2_unblock_signals(sigset_t *oldset)
2632{
2633 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2634 BUG_ON(rc);
2635}
2636
ccd979bd
MF		 2637module_init(ocfs2_init);
2638module_exit(ocfs2_exit);
Linux kernel - Deliverables
This page took 0.773698 seconds and 5 git commands to generate.