]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ocfs2/super.c
ocfs2: fix error handling when creating debugfs root in ocfs2_init()
[mirror_ubuntu-bionic-kernel.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
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
02b9984d
TT
637 sync_filesystem(sb);
638
5297aad8
JK
639 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
640 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF
641 ret = -EINVAL;
642 goto out;
643 }
644
2c442719
SM
645 tmp = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL |
646 OCFS2_MOUNT_HB_NONE;
647 if ((osb->s_mount_opt & tmp) != (parsed_options.mount_opt & tmp)) {
ccd979bd
MF
648 ret = -EINVAL;
649 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
650 goto out;
651 }
652
653 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 654 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF
655 ret = -EINVAL;
656 mlog(ML_ERROR, "Cannot change data mode on remount\n");
657 goto out;
658 }
659
12462f1d
JB
660 /* Probably don't want this on remount; it might
661 * mess with other nodes */
662 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
663 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
664 ret = -EINVAL;
665 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
666 goto out;
667 }
668
ccd979bd
MF
669 /* We're going to/from readonly mode. */
670 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK
671 /* Disable quota accounting before remounting RO */
672 if (*flags & MS_RDONLY) {
673 ret = ocfs2_susp_quotas(osb, 0);
674 if (ret < 0)
675 goto out;
676 }
ccd979bd
MF
677 /* Lock here so the check of HARD_RO and the potential
678 * setting of SOFT_RO is atomic. */
679 spin_lock(&osb->osb_lock);
680 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
681 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
682 ret = -EROFS;
683 goto unlock_osb;
684 }
685
686 if (*flags & MS_RDONLY) {
ccd979bd
MF
687 sb->s_flags |= MS_RDONLY;
688 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
689 } else {
ccd979bd
MF
690 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
691 mlog(ML_ERROR, "Cannot remount RDWR "
692 "filesystem due to previous errors.\n");
693 ret = -EROFS;
694 goto unlock_osb;
695 }
696 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
697 if (incompat_features) {
698 mlog(ML_ERROR, "Cannot remount RDWR because "
699 "of unsupported optional features "
700 "(%x).\n", incompat_features);
701 ret = -EINVAL;
702 goto unlock_osb;
703 }
704 sb->s_flags &= ~MS_RDONLY;
705 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
706 }
32a42d39 707 trace_ocfs2_remount(sb->s_flags, osb->osb_flags, *flags);
ccd979bd
MF
708unlock_osb:
709 spin_unlock(&osb->osb_lock);
19ece546
JK
710 /* Enable quota accounting after remounting RW */
711 if (!ret && !(*flags & MS_RDONLY)) {
712 if (sb_any_quota_suspended(sb))
713 ret = ocfs2_susp_quotas(osb, 1);
714 else
715 ret = ocfs2_enable_quotas(osb);
716 if (ret < 0) {
717 /* Return back changes... */
718 spin_lock(&osb->osb_lock);
719 sb->s_flags |= MS_RDONLY;
720 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
721 spin_unlock(&osb->osb_lock);
722 goto out;
723 }
724 }
ccd979bd
MF
725 }
726
727 if (!ret) {
ccd979bd
MF
728 /* Only save off the new mount options in case of a successful
729 * remount. */
c0123ade
TY
730 osb->s_mount_opt = parsed_options.mount_opt;
731 osb->s_atime_quantum = parsed_options.atime_quantum;
732 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF
733 if (parsed_options.commit_interval)
734 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF
735
736 if (!ocfs2_is_hard_readonly(osb))
737 ocfs2_set_journal_params(osb);
57b09bb5
JK
738
739 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
740 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
741 MS_POSIXACL : 0);
ccd979bd
MF
742 }
743out:
744 return ret;
745}
746
747static int ocfs2_sb_probe(struct super_block *sb,
748 struct buffer_head **bh,
73be192b
JB
749 int *sector_size,
750 struct ocfs2_blockcheck_stats *stats)
ccd979bd 751{
bddb8eb3 752 int status, tmpstat;
ccd979bd
MF
753 struct ocfs1_vol_disk_hdr *hdr;
754 struct ocfs2_dinode *di;
755 int blksize;
756
757 *bh = NULL;
758
759 /* may be > 512 */
e1defc4f 760 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF
761 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
762 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
763 *sector_size, OCFS2_MAX_BLOCKSIZE);
764 status = -EINVAL;
765 goto bail;
766 }
767
768 /* Can this really happen? */
769 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
770 *sector_size = OCFS2_MIN_BLOCKSIZE;
771
772 /* check block zero for old format */
773 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
774 if (status < 0) {
775 mlog_errno(status);
776 goto bail;
777 }
778 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
779 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
780 mlog(ML_ERROR, "incompatible version: %u.%u\n",
781 hdr->major_version, hdr->minor_version);
782 status = -EINVAL;
783 }
784 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
785 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
786 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
787 hdr->signature);
788 status = -EINVAL;
789 }
790 brelse(*bh);
791 *bh = NULL;
792 if (status < 0) {
793 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
794 "upgraded before mounting with ocfs v2\n");
795 goto bail;
796 }
797
798 /*
799 * Now check at magic offset for 512, 1024, 2048, 4096
800 * blocksizes. 4096 is the maximum blocksize because it is
801 * the minimum clustersize.
802 */
803 status = -EINVAL;
804 for (blksize = *sector_size;
805 blksize <= OCFS2_MAX_BLOCKSIZE;
806 blksize <<= 1) {
807 tmpstat = ocfs2_get_sector(sb, bh,
808 OCFS2_SUPER_BLOCK_BLKNO,
809 blksize);
810 if (tmpstat < 0) {
811 status = tmpstat;
812 mlog_errno(status);
fb5cbe9e 813 break;
ccd979bd
MF
814 }
815 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 816 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 817 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB
818 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
819 if (tmpstat < 0) {
820 brelse(*bh);
821 *bh = NULL;
822 }
823 if (tmpstat != -EAGAIN) {
824 status = tmpstat;
ccd979bd 825 break;
fb5cbe9e 826 }
ccd979bd
MF
827 }
828
829bail:
830 return status;
831}
832
c271c5c2
SM
833static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
834{
2c442719
SM
835 u32 hb_enabled = OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL;
836
837 if (osb->s_mount_opt & hb_enabled) {
838 if (ocfs2_mount_local(osb)) {
c271c5c2
SM
839 mlog(ML_ERROR, "Cannot heartbeat on a locally "
840 "mounted device.\n");
841 return -EINVAL;
842 }
2c442719 843 if (ocfs2_userspace_stack(osb)) {
b61817e1
JB
844 mlog(ML_ERROR, "Userspace stack expected, but "
845 "o2cb heartbeat arguments passed to mount\n");
846 return -EINVAL;
847 }
2c442719
SM
848 if (((osb->s_mount_opt & OCFS2_MOUNT_HB_GLOBAL) &&
849 !ocfs2_cluster_o2cb_global_heartbeat(osb)) ||
850 ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) &&
851 ocfs2_cluster_o2cb_global_heartbeat(osb))) {
852 mlog(ML_ERROR, "Mismatching o2cb heartbeat modes\n");
853 return -EINVAL;
854 }
b61817e1
JB
855 }
856
2c442719 857 if (!(osb->s_mount_opt & hb_enabled)) {
b61817e1
JB
858 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
859 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM
860 mlog(ML_ERROR, "Heartbeat has to be started to mount "
861 "a read-write clustered device.\n");
862 return -EINVAL;
863 }
864 }
865
866 return 0;
867}
868
b61817e1
JB
869/*
870 * If we're using a userspace stack, mount should have passed
871 * a name that matches the disk. If not, mount should not
872 * have passed a stack.
873 */
874static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
875 struct mount_options *mopt)
876{
877 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
878 mlog(ML_ERROR,
879 "cluster stack passed to mount, but this filesystem "
880 "does not support it\n");
881 return -EINVAL;
882 }
883
884 if (ocfs2_userspace_stack(osb) &&
885 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
886 OCFS2_STACK_LABEL_LEN)) {
887 mlog(ML_ERROR,
888 "cluster stack passed to mount (\"%s\") does not "
889 "match the filesystem (\"%s\")\n",
890 mopt->cluster_stack,
891 osb->osb_cluster_stack);
892 return -EINVAL;
893 }
894
895 return 0;
896}
897
19ece546
JK
898static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
899{
900 int type;
901 struct super_block *sb = osb->sb;
52362810
JK
902 unsigned int feature[OCFS2_MAXQUOTAS] = {
903 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
904 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK
905 int status = 0;
906
52362810 907 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK
908 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
909 continue;
910 if (unsuspend)
0f0dd62f 911 status = dquot_resume(sb, type);
eea7feb0
JK
912 else {
913 struct ocfs2_mem_dqinfo *oinfo;
914
915 /* Cancel periodic syncing before suspending */
916 oinfo = sb_dqinfo(sb, type)->dqi_priv;
917 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
0f0dd62f 918 status = dquot_suspend(sb, type);
eea7feb0 919 }
19ece546
JK
920 if (status < 0)
921 break;
922 }
923 if (status < 0)
924 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
925 "remount (error = %d).\n", status);
926 return status;
927}
928
929static int ocfs2_enable_quotas(struct ocfs2_super *osb)
930{
52362810 931 struct inode *inode[OCFS2_MAXQUOTAS] = { NULL, NULL };
19ece546 932 struct super_block *sb = osb->sb;
52362810
JK
933 unsigned int feature[OCFS2_MAXQUOTAS] = {
934 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
935 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
936 unsigned int ino[OCFS2_MAXQUOTAS] = {
937 LOCAL_USER_QUOTA_SYSTEM_INODE,
19ece546
JK
938 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
939 int status;
940 int type;
941
942 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
52362810 943 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK
944 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
945 continue;
946 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
947 osb->slot_num);
948 if (!inode[type]) {
949 status = -ENOENT;
950 goto out_quota_off;
951 }
287a8095
CH
952 status = dquot_enable(inode[type], type, QFMT_OCFS2,
953 DQUOT_USAGE_ENABLED);
19ece546
JK
954 if (status < 0)
955 goto out_quota_off;
956 }
957
52362810 958 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK
959 iput(inode[type]);
960 return 0;
961out_quota_off:
962 ocfs2_disable_quotas(osb);
52362810 963 for (type = 0; type < OCFS2_MAXQUOTAS; type++)
19ece546
JK
964 iput(inode[type]);
965 mlog_errno(status);
966 return status;
967}
968
969static void ocfs2_disable_quotas(struct ocfs2_super *osb)
970{
971 int type;
972 struct inode *inode;
973 struct super_block *sb = osb->sb;
c06bcbfa 974 struct ocfs2_mem_dqinfo *oinfo;
19ece546
JK
975
976 /* We mostly ignore errors in this function because there's not much
977 * we can do when we see them */
52362810 978 for (type = 0; type < OCFS2_MAXQUOTAS; type++) {
19ece546
JK
979 if (!sb_has_quota_loaded(sb, type))
980 continue;
c06bcbfa
JK
981 /* Cancel periodic syncing before we grab dqonoff_mutex */
982 oinfo = sb_dqinfo(sb, type)->dqi_priv;
983 cancel_delayed_work_sync(&oinfo->dqi_sync_work);
19ece546
JK
984 inode = igrab(sb->s_dquot.files[type]);
985 /* Turn off quotas. This will remove all dquot structures from
986 * memory and so they will be automatically synced to global
987 * quota files */
0f0dd62f
CH
988 dquot_disable(sb, type, DQUOT_USAGE_ENABLED |
989 DQUOT_LIMITS_ENABLED);
19ece546
JK
990 if (!inode)
991 continue;
992 iput(inode);
993 }
994}
995
996/* Handle quota on quotactl */
f00c9e44 997static int ocfs2_quota_on(struct super_block *sb, int type, int format_id)
19ece546 998{
52362810
JK
999 unsigned int feature[OCFS2_MAXQUOTAS] = {
1000 OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
1001 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
19ece546
JK
1002
1003 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
1004 return -EINVAL;
1005
287a8095
CH
1006 return dquot_enable(sb_dqopt(sb)->files[type], type,
1007 format_id, DQUOT_LIMITS_ENABLED);
19ece546
JK
1008}
1009
1010/* Handle quota off quotactl */
307ae18a 1011static int ocfs2_quota_off(struct super_block *sb, int type)
19ece546 1012{
0f0dd62f 1013 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
19ece546
JK
1014}
1015
0d54b217 1016static const struct quotactl_ops ocfs2_quotactl_ops = {
f00c9e44 1017 .quota_on_meta = ocfs2_quota_on,
19ece546 1018 .quota_off = ocfs2_quota_off,
287a8095
CH
1019 .quota_sync = dquot_quota_sync,
1020 .get_info = dquot_get_dqinfo,
1021 .set_info = dquot_set_dqinfo,
1022 .get_dqblk = dquot_get_dqblk,
1023 .set_dqblk = dquot_set_dqblk,
19ece546
JK
1024};
1025
ccd979bd
MF
1026static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
1027{
1028 struct dentry *root;
1029 int status, sector_size;
c0123ade 1030 struct mount_options parsed_options;
ccd979bd
MF
1031 struct inode *inode = NULL;
1032 struct ocfs2_super *osb = NULL;
1033 struct buffer_head *bh = NULL;
49fa8140 1034 char nodestr[12];
73be192b 1035 struct ocfs2_blockcheck_stats stats;
ccd979bd 1036
32a42d39 1037 trace_ocfs2_fill_super(sb, data, silent);
ccd979bd 1038
c0123ade 1039 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF
1040 status = -EINVAL;
1041 goto read_super_error;
1042 }
1043
1044 /* probe for superblock */
73be192b 1045 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF
1046 if (status < 0) {
1047 mlog(ML_ERROR, "superblock probe failed!\n");
1048 goto read_super_error;
1049 }
1050
73be192b 1051 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF
1052 osb = OCFS2_SB(sb);
1053 if (status < 0) {
1054 mlog_errno(status);
1055 goto read_super_error;
1056 }
1057 brelse(bh);
1058 bh = NULL;
a68979b8 1059
5297aad8
JK
1060 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1061 status = -EINVAL;
1062 goto read_super_error;
1063 }
c0123ade
TY
1064 osb->s_mount_opt = parsed_options.mount_opt;
1065 osb->s_atime_quantum = parsed_options.atime_quantum;
1066 osb->preferred_slot = parsed_options.slot;
d147b3d6 1067 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF
1068
1069 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1070 osb->osb_resv_level = parsed_options.resv_level;
83f92318
MF
1071 osb->osb_dir_resv_level = parsed_options.resv_level;
1072 if (parsed_options.dir_resv_level == -1)
1073 osb->osb_dir_resv_level = parsed_options.resv_level;
1074 else
1075 osb->osb_dir_resv_level = parsed_options.dir_resv_level;
ccd979bd 1076
b61817e1
JB
1077 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1078 if (status)
1079 goto read_super_error;
1080
ccd979bd
MF
1081 sb->s_magic = OCFS2_SUPER_MAGIC;
1082
9e1f1de0 1083 sb->s_flags = (sb->s_flags & ~(MS_POSIXACL | MS_NOSEC)) |
a68979b8
TY
1084 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1085
ccd979bd
MF
1086 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1087 * heartbeat=none */
1088 if (bdev_read_only(sb->s_bdev)) {
1089 if (!(sb->s_flags & MS_RDONLY)) {
1090 status = -EACCES;
1091 mlog(ML_ERROR, "Readonly device detected but readonly "
1092 "mount was not specified.\n");
1093 goto read_super_error;
1094 }
1095
1096 /* You should not be able to start a local heartbeat
1097 * on a readonly device. */
1098 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1099 status = -EROFS;
1100 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1101 "device.\n");
1102 goto read_super_error;
1103 }
1104
1105 status = ocfs2_check_journals_nolocks(osb);
1106 if (status < 0) {
1107 if (status == -EROFS)
1108 mlog(ML_ERROR, "Recovery required on readonly "
1109 "file system, but write access is "
1110 "unavailable.\n");
1111 else
2bd63216 1112 mlog_errno(status);
ccd979bd
MF
1113 goto read_super_error;
1114 }
1115
1116 ocfs2_set_ro_flag(osb, 1);
1117
619c200d
SM
1118 printk(KERN_NOTICE "ocfs2: Readonly device (%s) detected. "
1119 "Cluster services will not be used for this mount. "
1120 "Recovery will be skipped.\n", osb->dev_str);
ccd979bd
MF
1121 }
1122
1123 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF
1124 if (sb->s_flags & MS_RDONLY)
1125 ocfs2_set_ro_flag(osb, 0);
1126 }
1127
c271c5c2
SM
1128 status = ocfs2_verify_heartbeat(osb);
1129 if (status < 0) {
1130 mlog_errno(status);
1131 goto read_super_error;
1132 }
1133
ccd979bd
MF
1134 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1135 ocfs2_debugfs_root);
1136 if (!osb->osb_debug_root) {
1137 status = -EINVAL;
1138 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1139 goto read_super_error;
1140 }
1141
50397507
SM
1142 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1143 osb->osb_debug_root,
1144 osb,
1145 &ocfs2_osb_debug_fops);
1146 if (!osb->osb_ctxt) {
1147 status = -EINVAL;
1148 mlog_errno(status);
1149 goto read_super_error;
1150 }
1151
73be192b
JB
1152 if (ocfs2_meta_ecc(osb)) {
1153 status = ocfs2_blockcheck_stats_debugfs_install(
1154 &osb->osb_ecc_stats,
1155 osb->osb_debug_root);
1156 if (status) {
1157 mlog(ML_ERROR,
1158 "Unable to create blockcheck statistics "
1159 "files\n");
1160 goto read_super_error;
1161 }
1162 }
1163
ccd979bd 1164 status = ocfs2_mount_volume(sb);
ccd979bd
MF
1165 if (status < 0)
1166 goto read_super_error;
1167
be0d93f0
AV
1168 if (osb->root_inode)
1169 inode = igrab(osb->root_inode);
1170
ccd979bd
MF
1171 if (!inode) {
1172 status = -EIO;
1173 mlog_errno(status);
1174 goto read_super_error;
1175 }
1176
48fde701 1177 root = d_make_root(inode);
ccd979bd
MF
1178 if (!root) {
1179 status = -ENOMEM;
1180 mlog_errno(status);
1181 goto read_super_error;
1182 }
1183
1184 sb->s_root = root;
1185
1186 ocfs2_complete_mount_recovery(osb);
1187
c271c5c2
SM
1188 if (ocfs2_mount_local(osb))
1189 snprintf(nodestr, sizeof(nodestr), "local");
1190 else
19fdb624 1191 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM
1192
1193 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1194 "with %s data mode.\n",
c271c5c2 1195 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF
1196 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1197 "ordered");
1198
1199 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1200 wake_up(&osb->osb_mount_event);
1201
19ece546
JK
1202 /* Now we can initialize quotas because we can afford to wait
1203 * for cluster locks recovery now. That also means that truncation
1204 * log recovery can happen but that waits for proper quota setup */
1205 if (!(sb->s_flags & MS_RDONLY)) {
1206 status = ocfs2_enable_quotas(osb);
1207 if (status < 0) {
1208 /* We have to err-out specially here because
1209 * s_root is already set */
1210 mlog_errno(status);
1211 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1212 wake_up(&osb->osb_mount_event);
19ece546
JK
1213 return status;
1214 }
1215 }
1216
1217 ocfs2_complete_quota_recovery(osb);
1218
1219 /* Now we wake up again for processes waiting for quotas */
1220 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1221 wake_up(&osb->osb_mount_event);
1222
df152c24 1223 /* Start this when the mount is almost sure of being successful */
8b712cd5 1224 ocfs2_orphan_scan_start(osb);
df152c24 1225
ccd979bd
MF
1226 return status;
1227
1228read_super_error:
a81cb88b 1229 brelse(bh);
ccd979bd 1230
ccd979bd
MF
1231 if (osb) {
1232 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1233 wake_up(&osb->osb_mount_event);
1234 ocfs2_dismount_volume(sb, 1);
1235 }
1236
c1e8d35e
TM
1237 if (status)
1238 mlog_errno(status);
ccd979bd
MF
1239 return status;
1240}
1241
152a0836 1242static struct dentry *ocfs2_mount(struct file_system_type *fs_type,
454e2398
DH
1243 int flags,
1244 const char *dev_name,
152a0836 1245 void *data)
ccd979bd 1246{
152a0836 1247 return mount_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
ccd979bd
MF
1248}
1249
1250static struct file_system_type ocfs2_fs_type = {
1251 .owner = THIS_MODULE,
1252 .name = "ocfs2",
152a0836 1253 .mount = ocfs2_mount,
8ed6b237 1254 .kill_sb = kill_block_super,
1ba9da2f 1255 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF
1256 .next = NULL
1257};
91417705 1258MODULE_ALIAS_FS("ocfs2");
ccd979bd 1259
5297aad8
JK
1260static int ocfs2_check_set_options(struct super_block *sb,
1261 struct mount_options *options)
1262{
1263 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1264 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1265 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1266 mlog(ML_ERROR, "User quotas were requested, but this "
1267 "filesystem does not have the feature enabled.\n");
1268 return 0;
1269 }
1270 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1271 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1272 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1273 mlog(ML_ERROR, "Group quotas were requested, but this "
1274 "filesystem does not have the feature enabled.\n");
1275 return 0;
1276 }
1277 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1278 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1279 mlog(ML_ERROR, "ACL support requested but extended attributes "
1280 "feature is not enabled\n");
1281 return 0;
1282 }
1283 /* No ACL setting specified? Use XATTR feature... */
1284 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1285 OCFS2_MOUNT_NO_POSIX_ACL))) {
1286 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1287 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1288 else
1289 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1290 }
1291 return 1;
1292}
1293
ccd979bd
MF
1294static int ocfs2_parse_options(struct super_block *sb,
1295 char *options,
c0123ade 1296 struct mount_options *mopt,
ccd979bd
MF
1297 int is_remount)
1298{
52c303c5 1299 int status, user_stack = 0;
ccd979bd 1300 char *p;
2c442719 1301 u32 tmp;
ccd979bd 1302
32a42d39 1303 trace_ocfs2_parse_options(is_remount, options ? options : "(none)");
ccd979bd 1304
d147b3d6 1305 mopt->commit_interval = 0;
4b37fcb7 1306 mopt->mount_opt = OCFS2_MOUNT_NOINTR;
c0123ade
TY
1307 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1308 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1309 mopt->localalloc_opt = -1;
b61817e1 1310 mopt->cluster_stack[0] = '\0';
d02f00cc 1311 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
83f92318 1312 mopt->dir_resv_level = -1;
ccd979bd
MF
1313
1314 if (!options) {
1315 status = 1;
1316 goto bail;
1317 }
1318
1319 while ((p = strsep(&options, ",")) != NULL) {
1320 int token, option;
1321 substring_t args[MAX_OPT_ARGS];
1322
1323 if (!*p)
1324 continue;
1325
1326 token = match_token(p, tokens, args);
1327 switch (token) {
1328 case Opt_hb_local:
c0123ade 1329 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF
1330 break;
1331 case Opt_hb_none:
2c442719
SM
1332 mopt->mount_opt |= OCFS2_MOUNT_HB_NONE;
1333 break;
1334 case Opt_hb_global:
1335 mopt->mount_opt |= OCFS2_MOUNT_HB_GLOBAL;
ccd979bd
MF
1336 break;
1337 case Opt_barrier:
1338 if (match_int(&args[0], &option)) {
1339 status = 0;
1340 goto bail;
1341 }
1342 if (option)
c0123ade 1343 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1344 else
c0123ade 1345 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF
1346 break;
1347 case Opt_intr:
c0123ade 1348 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF
1349 break;
1350 case Opt_nointr:
c0123ade 1351 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF
1352 break;
1353 case Opt_err_panic:
c0123ade 1354 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF
1355 break;
1356 case Opt_err_ro:
c0123ade 1357 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF
1358 break;
1359 case Opt_data_ordered:
c0123ade 1360 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF
1361 break;
1362 case Opt_data_writeback:
c0123ade 1363 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1364 break;
cf1d6c76
TY
1365 case Opt_user_xattr:
1366 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1367 break;
1368 case Opt_nouser_xattr:
1369 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1370 break;
7f1a37e3
TY
1371 case Opt_atime_quantum:
1372 if (match_int(&args[0], &option)) {
1373 status = 0;
1374 goto bail;
1375 }
1376 if (option >= 0)
c0123ade 1377 mopt->atime_quantum = option;
7f1a37e3 1378 break;
baf4661a
SM
1379 case Opt_slot:
1380 option = 0;
1381 if (match_int(&args[0], &option)) {
1382 status = 0;
1383 goto bail;
1384 }
1385 if (option)
c0123ade 1386 mopt->slot = (s16)option;
baf4661a 1387 break;
d147b3d6
MF
1388 case Opt_commit:
1389 option = 0;
1390 if (match_int(&args[0], &option)) {
1391 status = 0;
1392 goto bail;
1393 }
1394 if (option < 0)
1395 return 0;
1396 if (option == 0)
2b4e30fb 1397 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF
1398 mopt->commit_interval = HZ * option;
1399 break;
2fbe8d1e
SM
1400 case Opt_localalloc:
1401 option = 0;
1402 if (match_int(&args[0], &option)) {
1403 status = 0;
1404 goto bail;
1405 }
73c8a800 1406 if (option >= 0)
2fbe8d1e
SM
1407 mopt->localalloc_opt = option;
1408 break;
53fc622b
MF
1409 case Opt_localflocks:
1410 /*
1411 * Changing this during remount could race
1412 * flock() requests, or "unbalance" existing
1413 * ones (e.g., a lock is taken in one mode but
1414 * dropped in the other). If users care enough
1415 * to flip locking modes during remount, we
1416 * could add a "local" flag to individual
1417 * flock structures for proper tracking of
1418 * state.
1419 */
1420 if (!is_remount)
1421 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1422 break;
b61817e1
JB
1423 case Opt_stack:
1424 /* Check both that the option we were passed
1425 * is of the right length and that it is a proper
1426 * string of the right length.
1427 */
1428 if (((args[0].to - args[0].from) !=
1429 OCFS2_STACK_LABEL_LEN) ||
1430 (strnlen(args[0].from,
1431 OCFS2_STACK_LABEL_LEN) !=
1432 OCFS2_STACK_LABEL_LEN)) {
1433 mlog(ML_ERROR,
1434 "Invalid cluster_stack option\n");
1435 status = 0;
1436 goto bail;
1437 }
1438 memcpy(mopt->cluster_stack, args[0].from,
1439 OCFS2_STACK_LABEL_LEN);
1440 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
52c303c5
MF
1441 /*
1442 * Open code the memcmp here as we don't have
1443 * an osb to pass to
1444 * ocfs2_userspace_stack().
1445 */
1446 if (memcmp(mopt->cluster_stack,
1447 OCFS2_CLASSIC_CLUSTER_STACK,
1448 OCFS2_STACK_LABEL_LEN))
1449 user_stack = 1;
b61817e1 1450 break;
12462f1d
JB
1451 case Opt_inode64:
1452 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1453 break;
19ece546 1454 case Opt_usrquota:
19ece546
JK
1455 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1456 break;
1457 case Opt_grpquota:
19ece546
JK
1458 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1459 break;
7bdb0d18
TY
1460 case Opt_coherency_buffered:
1461 mopt->mount_opt |= OCFS2_MOUNT_COHERENCY_BUFFERED;
1462 break;
1463 case Opt_coherency_full:
1464 mopt->mount_opt &= ~OCFS2_MOUNT_COHERENCY_BUFFERED;
1465 break;
a68979b8
TY
1466 case Opt_acl:
1467 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1468 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY
1469 break;
1470 case Opt_noacl:
5297aad8 1471 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY
1472 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1473 break;
d02f00cc
MF
1474 case Opt_resv_level:
1475 if (is_remount)
1476 break;
1477 if (match_int(&args[0], &option)) {
1478 status = 0;
1479 goto bail;
1480 }
1481 if (option >= OCFS2_MIN_RESV_LEVEL &&
1482 option < OCFS2_MAX_RESV_LEVEL)
1483 mopt->resv_level = option;
1484 break;
83f92318
MF
1485 case Opt_dir_resv_level:
1486 if (is_remount)
1487 break;
1488 if (match_int(&args[0], &option)) {
1489 status = 0;
1490 goto bail;
1491 }
1492 if (option >= OCFS2_MIN_RESV_LEVEL &&
1493 option < OCFS2_MAX_RESV_LEVEL)
1494 mopt->dir_resv_level = option;
1495 break;
ccd979bd
MF
1496 default:
1497 mlog(ML_ERROR,
1498 "Unrecognized mount option \"%s\" "
1499 "or missing value\n", p);
1500 status = 0;
1501 goto bail;
1502 }
1503 }
1504
52c303c5
MF
1505 if (user_stack == 0) {
1506 /* Ensure only one heartbeat mode */
1507 tmp = mopt->mount_opt & (OCFS2_MOUNT_HB_LOCAL |
1508 OCFS2_MOUNT_HB_GLOBAL |
1509 OCFS2_MOUNT_HB_NONE);
1510 if (hweight32(tmp) != 1) {
1511 mlog(ML_ERROR, "Invalid heartbeat mount options\n");
1512 status = 0;
1513 goto bail;
1514 }
2c442719
SM
1515 }
1516
ccd979bd
MF
1517 status = 1;
1518
1519bail:
ccd979bd
MF
1520 return status;
1521}
1522
34c80b1d 1523static int ocfs2_show_options(struct seq_file *s, struct dentry *root)
d550071c 1524{
34c80b1d 1525 struct ocfs2_super *osb = OCFS2_SB(root->d_sb);
d550071c 1526 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1527 unsigned int local_alloc_megs;
d550071c 1528
2c442719
SM
1529 if (opts & (OCFS2_MOUNT_HB_LOCAL | OCFS2_MOUNT_HB_GLOBAL)) {
1530 seq_printf(s, ",_netdev");
1531 if (opts & OCFS2_MOUNT_HB_LOCAL)
1532 seq_printf(s, ",%s", OCFS2_HB_LOCAL);
1533 else
1534 seq_printf(s, ",%s", OCFS2_HB_GLOBAL);
1535 } else
1536 seq_printf(s, ",%s", OCFS2_HB_NONE);
d550071c
SM
1537
1538 if (opts & OCFS2_MOUNT_NOINTR)
1539 seq_printf(s, ",nointr");
1540
1541 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1542 seq_printf(s, ",data=writeback");
1543 else
1544 seq_printf(s, ",data=ordered");
1545
1546 if (opts & OCFS2_MOUNT_BARRIER)
1547 seq_printf(s, ",barrier=1");
1548
1549 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1550 seq_printf(s, ",errors=panic");
1551 else
1552 seq_printf(s, ",errors=remount-ro");
1553
1554 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1555 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1556
34c80b1d 1557 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
d550071c 1558
d147b3d6
MF
1559 if (osb->osb_commit_interval)
1560 seq_printf(s, ",commit=%u",
1561 (unsigned) (osb->osb_commit_interval / HZ));
1562
ebcee4b5 1563 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
6b82021b 1564 if (local_alloc_megs != ocfs2_la_default_mb(osb))
ebcee4b5 1565 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1566
53fc622b
MF
1567 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1568 seq_printf(s, ",localflocks,");
1569
b61817e1
JB
1570 if (osb->osb_cluster_stack[0])
1571 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1572 osb->osb_cluster_stack);
19ece546
JK
1573 if (opts & OCFS2_MOUNT_USRQUOTA)
1574 seq_printf(s, ",usrquota");
1575 if (opts & OCFS2_MOUNT_GRPQUOTA)
1576 seq_printf(s, ",grpquota");
b61817e1 1577
7bdb0d18
TY
1578 if (opts & OCFS2_MOUNT_COHERENCY_BUFFERED)
1579 seq_printf(s, ",coherency=buffered");
1580 else
1581 seq_printf(s, ",coherency=full");
1582
b0f73cfc
SM
1583 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1584 seq_printf(s, ",nouser_xattr");
1585 else
1586 seq_printf(s, ",user_xattr");
1587
12462f1d
JB
1588 if (opts & OCFS2_MOUNT_INODE64)
1589 seq_printf(s, ",inode64");
1590
a68979b8
TY
1591 if (opts & OCFS2_MOUNT_POSIX_ACL)
1592 seq_printf(s, ",acl");
1593 else
1594 seq_printf(s, ",noacl");
a68979b8 1595
d02f00cc
MF
1596 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1597 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1598
83f92318
MF
1599 if (osb->osb_dir_resv_level != osb->osb_resv_level)
1600 seq_printf(s, ",dir_resv_level=%d", osb->osb_resv_level);
1601
d550071c
SM
1602 return 0;
1603}
1604
ccd979bd
MF
1605static int __init ocfs2_init(void)
1606{
c18ceab0 1607 int status;
a11f7e63 1608
ccd979bd 1609 status = init_ocfs2_uptodate_cache();
342827d7
AV
1610 if (status < 0)
1611 goto out1;
ccd979bd
MF
1612
1613 status = ocfs2_initialize_mem_caches();
342827d7
AV
1614 if (status < 0)
1615 goto out2;
ccd979bd
MF
1616
1617 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1618 if (!ocfs2_wq) {
1619 status = -ENOMEM;
342827d7 1620 goto out3;
ccd979bd
MF
1621 }
1622
ccd979bd
MF
1623 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1624 if (!ocfs2_debugfs_root) {
dc171580 1625 status = -ENOMEM;
ccd979bd 1626 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
dc171580 1627 goto out4;
ccd979bd
MF
1628 }
1629
63e0c48a
JB
1630 ocfs2_set_locking_protocol();
1631
9e33d69f 1632 status = register_quota_format(&ocfs2_quota_format);
342827d7
AV
1633 if (status < 0)
1634 goto out4;
1635 status = register_filesystem(&ocfs2_fs_type);
1636 if (!status)
1637 return 0;
ccd979bd 1638
342827d7
AV
1639 unregister_quota_format(&ocfs2_quota_format);
1640out4:
1641 destroy_workqueue(ocfs2_wq);
1642 debugfs_remove(ocfs2_debugfs_root);
1643out3:
1644 ocfs2_free_mem_caches();
1645out2:
1646 exit_ocfs2_uptodate_cache();
1647out1:
1648 mlog_errno(status);
1649 return status;
ccd979bd
MF
1650}
1651
1652static void __exit ocfs2_exit(void)
1653{
ccd979bd
MF
1654 if (ocfs2_wq) {
1655 flush_workqueue(ocfs2_wq);
1656 destroy_workqueue(ocfs2_wq);
1657 }
1658
9e33d69f
JK
1659 unregister_quota_format(&ocfs2_quota_format);
1660
ccd979bd
MF
1661 debugfs_remove(ocfs2_debugfs_root);
1662
1663 ocfs2_free_mem_caches();
1664
1665 unregister_filesystem(&ocfs2_fs_type);
1666
ccd979bd 1667 exit_ocfs2_uptodate_cache();
ccd979bd
MF
1668}
1669
1670static void ocfs2_put_super(struct super_block *sb)
1671{
32a42d39 1672 trace_ocfs2_put_super(sb);
ccd979bd
MF
1673
1674 ocfs2_sync_blockdev(sb);
1675 ocfs2_dismount_volume(sb, 0);
ccd979bd
MF
1676}
1677
726c3342 1678static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF
1679{
1680 struct ocfs2_super *osb;
1681 u32 numbits, freebits;
1682 int status;
1683 struct ocfs2_dinode *bm_lock;
1684 struct buffer_head *bh = NULL;
1685 struct inode *inode = NULL;
1686
32a42d39 1687 trace_ocfs2_statfs(dentry->d_sb, buf);
ccd979bd 1688
726c3342 1689 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF
1690
1691 inode = ocfs2_get_system_file_inode(osb,
1692 GLOBAL_BITMAP_SYSTEM_INODE,
1693 OCFS2_INVALID_SLOT);
1694 if (!inode) {
1695 mlog(ML_ERROR, "failed to get bitmap inode\n");
1696 status = -EIO;
1697 goto bail;
1698 }
1699
e63aecb6 1700 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF
1701 if (status < 0) {
1702 mlog_errno(status);
1703 goto bail;
1704 }
1705
1706 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1707
1708 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1709 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1710
1711 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1712 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF
1713 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1714 buf->f_blocks = ((sector_t) numbits) *
1715 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1716 buf->f_bfree = ((sector_t) freebits) *
1717 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1718 buf->f_bavail = buf->f_bfree;
1719 buf->f_files = numbits;
1720 buf->f_ffree = freebits;
837711f8
CL
1721 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1722 & 0xFFFFFFFFUL;
1723 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1724 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF
1725
1726 brelse(bh);
1727
e63aecb6 1728 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF
1729 status = 0;
1730bail:
1731 if (inode)
1732 iput(inode);
1733
c1e8d35e
TM
1734 if (status)
1735 mlog_errno(status);
ccd979bd
MF
1736
1737 return status;
1738}
1739
51cc5068 1740static void ocfs2_inode_init_once(void *data)
ccd979bd
MF
1741{
1742 struct ocfs2_inode_info *oi = data;
1743
a35afb83
CL
1744 oi->ip_flags = 0;
1745 oi->ip_open_count = 0;
1746 spin_lock_init(&oi->ip_lock);
1747 ocfs2_extent_map_init(&oi->vfs_inode);
1748 INIT_LIST_HEAD(&oi->ip_io_markers);
a35afb83 1749 oi->ip_dir_start_lookup = 0;
c18ceab0 1750 mutex_init(&oi->ip_unaligned_aio);
a35afb83 1751 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1752 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1753 mutex_init(&oi->ip_io_mutex);
ccd979bd 1754
a35afb83
CL
1755 oi->ip_blkno = 0ULL;
1756 oi->ip_clusters = 0;
ccd979bd 1757
4fe370af
MF
1758 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1759
a35afb83 1760 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1761 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1762 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1763
8cb471e8 1764 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1765 &ocfs2_inode_caching_ops);
ccd979bd 1766
a35afb83 1767 inode_init_once(&oi->vfs_inode);
ccd979bd
MF
1768}
1769
1770static int ocfs2_initialize_mem_caches(void)
1771{
1772 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ
1773 sizeof(struct ocfs2_inode_info),
1774 0,
1775 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1776 SLAB_MEM_SPREAD),
20c2df83 1777 ocfs2_inode_init_once);
9e33d69f
JK
1778 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1779 sizeof(struct ocfs2_dquot),
1780 0,
1781 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1782 SLAB_MEM_SPREAD),
1783 NULL);
1784 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1785 sizeof(struct ocfs2_quota_chunk),
1786 0,
1787 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1788 NULL);
1789 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1790 !ocfs2_qf_chunk_cachep) {
1791 if (ocfs2_inode_cachep)
1792 kmem_cache_destroy(ocfs2_inode_cachep);
1793 if (ocfs2_dquot_cachep)
1794 kmem_cache_destroy(ocfs2_dquot_cachep);
1795 if (ocfs2_qf_chunk_cachep)
1796 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1797 return -ENOMEM;
9e33d69f 1798 }
ccd979bd 1799
ccd979bd
MF
1800 return 0;
1801}
1802
1803static void ocfs2_free_mem_caches(void)
1804{
8c0a8537
KS
1805 /*
1806 * Make sure all delayed rcu free inodes are flushed before we
1807 * destroy cache.
1808 */
1809 rcu_barrier();
ccd979bd
MF
1810 if (ocfs2_inode_cachep)
1811 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1812 ocfs2_inode_cachep = NULL;
9e33d69f
JK
1813
1814 if (ocfs2_dquot_cachep)
1815 kmem_cache_destroy(ocfs2_dquot_cachep);
1816 ocfs2_dquot_cachep = NULL;
1817
1818 if (ocfs2_qf_chunk_cachep)
1819 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1820 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF
1821}
1822
1823static int ocfs2_get_sector(struct super_block *sb,
1824 struct buffer_head **bh,
1825 int block,
1826 int sect_size)
1827{
1828 if (!sb_set_blocksize(sb, sect_size)) {
1829 mlog(ML_ERROR, "unable to set blocksize\n");
1830 return -EIO;
1831 }
1832
1833 *bh = sb_getblk(sb, block);
1834 if (!*bh) {
7391a294
RX
1835 mlog_errno(-ENOMEM);
1836 return -ENOMEM;
ccd979bd
MF
1837 }
1838 lock_buffer(*bh);
1839 if (!buffer_dirty(*bh))
1840 clear_buffer_uptodate(*bh);
1841 unlock_buffer(*bh);
1842 ll_rw_block(READ, 1, bh);
1843 wait_on_buffer(*bh);
28d57d43 1844 if (!buffer_uptodate(*bh)) {
1845 mlog_errno(-EIO);
1846 brelse(*bh);
1847 *bh = NULL;
1848 return -EIO;
1849 }
1850
ccd979bd
MF
1851 return 0;
1852}
1853
ccd979bd
MF
1854static int ocfs2_mount_volume(struct super_block *sb)
1855{
1856 int status = 0;
1857 int unlock_super = 0;
1858 struct ocfs2_super *osb = OCFS2_SB(sb);
1859
ccd979bd
MF
1860 if (ocfs2_is_hard_readonly(osb))
1861 goto leave;
1862
ccd979bd
MF
1863 status = ocfs2_dlm_init(osb);
1864 if (status < 0) {
1865 mlog_errno(status);
1866 goto leave;
1867 }
1868
ccd979bd
MF
1869 status = ocfs2_super_lock(osb, 1);
1870 if (status < 0) {
1871 mlog_errno(status);
1872 goto leave;
1873 }
1874 unlock_super = 1;
1875
1876 /* This will load up the node map and add ourselves to it. */
1877 status = ocfs2_find_slot(osb);
1878 if (status < 0) {
1879 mlog_errno(status);
1880 goto leave;
1881 }
1882
ccd979bd
MF
1883 /* load all node-local system inodes */
1884 status = ocfs2_init_local_system_inodes(osb);
1885 if (status < 0) {
1886 mlog_errno(status);
1887 goto leave;
1888 }
1889
1890 status = ocfs2_check_volume(osb);
1891 if (status < 0) {
1892 mlog_errno(status);
1893 goto leave;
1894 }
1895
1896 status = ocfs2_truncate_log_init(osb);
06c59bb8 1897 if (status < 0)
ccd979bd 1898 mlog_errno(status);
c271c5c2 1899
ccd979bd
MF
1900leave:
1901 if (unlock_super)
1902 ocfs2_super_unlock(osb, 1);
1903
ccd979bd
MF
1904 return status;
1905}
1906
ccd979bd
MF
1907static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1908{
286eaa95 1909 int tmp, hangup_needed = 0;
ccd979bd 1910 struct ocfs2_super *osb = NULL;
99d7a882 1911 char nodestr[12];
ccd979bd 1912
32a42d39 1913 trace_ocfs2_dismount_volume(sb);
ccd979bd
MF
1914
1915 BUG_ON(!sb);
1916 osb = OCFS2_SB(sb);
1917 BUG_ON(!osb);
1918
50397507
SM
1919 debugfs_remove(osb->osb_ctxt);
1920
692684e1
SM
1921 /* Orphan scan should be stopped as early as possible */
1922 ocfs2_orphan_scan_stop(osb);
1923
19ece546
JK
1924 ocfs2_disable_quotas(osb);
1925
e3a767b6
JK
1926 /* All dquots should be freed by now */
1927 WARN_ON(!llist_empty(&osb->dquot_drop_list));
1928 /* Wait for worker to be done with the work structure in osb */
1929 cancel_work_sync(&osb->dquot_drop_work);
1930
ccd979bd
MF
1931 ocfs2_shutdown_local_alloc(osb);
1932
b253bfd8
X
1933 ocfs2_truncate_log_shutdown(osb);
1934
553abd04
JB
1935 /* This will disable recovery and flush any recovery work. */
1936 ocfs2_recovery_exit(osb);
ccd979bd
MF
1937
1938 ocfs2_journal_shutdown(osb);
1939
1940 ocfs2_sync_blockdev(sb);
1941
374a263e
TM
1942 ocfs2_purge_refcount_trees(osb);
1943
4670c46d
JB
1944 /* No cluster connection means we've failed during mount, so skip
1945 * all the steps which depended on that to complete. */
1946 if (osb->cconn) {
ccd979bd
MF
1947 tmp = ocfs2_super_lock(osb, 1);
1948 if (tmp < 0) {
1949 mlog_errno(tmp);
1950 return;
1951 }
19b613d4 1952 }
ccd979bd 1953
19b613d4
MF
1954 if (osb->slot_num != OCFS2_INVALID_SLOT)
1955 ocfs2_put_slot(osb);
ccd979bd 1956
4670c46d 1957 if (osb->cconn)
ccd979bd 1958 ocfs2_super_unlock(osb, 1);
ccd979bd
MF
1959
1960 ocfs2_release_system_inodes(osb);
1961
6953b4c0 1962 /*
6953b4c0
JB
1963 * If we're dismounting due to mount error, mount.ocfs2 will clean
1964 * up heartbeat. If we're a local mount, there is no heartbeat.
1965 * If we failed before we got a uuid_str yet, we can't stop
1966 * heartbeat. Otherwise, do it.
1967 */
03efed8a
TY
1968 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str &&
1969 !ocfs2_is_hard_readonly(osb))
286eaa95
JB
1970 hangup_needed = 1;
1971
1972 if (osb->cconn)
1973 ocfs2_dlm_shutdown(osb, hangup_needed);
1974
73be192b 1975 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB
1976 debugfs_remove(osb->osb_debug_root);
1977
1978 if (hangup_needed)
6953b4c0 1979 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF
1980
1981 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1982
c271c5c2
SM
1983 if (ocfs2_mount_local(osb))
1984 snprintf(nodestr, sizeof(nodestr), "local");
1985 else
19fdb624 1986 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM
1987
1988 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1989 osb->dev_str, nodestr);
ccd979bd
MF
1990
1991 ocfs2_delete_osb(osb);
1992 kfree(osb);
1993 sb->s_dev = 0;
1994 sb->s_fs_info = NULL;
1995}
1996
1997static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1998 unsigned uuid_bytes)
1999{
2000 int i, ret;
2001 char *ptr;
2002
2003 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
2004
cd861280 2005 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF
2006 if (osb->uuid_str == NULL)
2007 return -ENOMEM;
2008
ccd979bd
MF
2009 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
2010 /* print with null */
2011 ret = snprintf(ptr, 3, "%02X", uuid[i]);
2012 if (ret != 2) /* drop super cleans up */
2013 return -EINVAL;
2014 /* then only advance past the last char */
2015 ptr += 2;
2016 }
2017
2018 return 0;
2019}
2020
3bdb8efd
PL
2021/* Make sure entire volume is addressable by our journal. Requires
2022 osb_clusters_at_boot to be valid and for the journal to have been
2023 initialized by ocfs2_journal_init(). */
2024static int ocfs2_journal_addressable(struct ocfs2_super *osb)
2025{
2026 int status = 0;
2027 u64 max_block =
2028 ocfs2_clusters_to_blocks(osb->sb,
2029 osb->osb_clusters_at_boot) - 1;
2030
2031 /* 32-bit block number is always OK. */
2032 if (max_block <= (u32)~0ULL)
2033 goto out;
2034
2035 /* Volume is "huge", so see if our journal is new enough to
2036 support it. */
2037 if (!(OCFS2_HAS_COMPAT_FEATURE(osb->sb,
2038 OCFS2_FEATURE_COMPAT_JBD2_SB) &&
2039 jbd2_journal_check_used_features(osb->journal->j_journal, 0, 0,
2040 JBD2_FEATURE_INCOMPAT_64BIT))) {
2041 mlog(ML_ERROR, "The journal cannot address the entire volume. "
2042 "Enable the 'block64' journal option with tunefs.ocfs2");
2043 status = -EFBIG;
2044 goto out;
2045 }
2046
2047 out:
2048 return status;
2049}
2050
ccd979bd
MF
2051static int ocfs2_initialize_super(struct super_block *sb,
2052 struct buffer_head *bh,
73be192b
JB
2053 int sector_size,
2054 struct ocfs2_blockcheck_stats *stats)
ccd979bd 2055{
bddb8eb3 2056 int status;
5a254031
MF
2057 int i, cbits, bbits;
2058 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 2059 struct inode *inode = NULL;
ccd979bd 2060 struct ocfs2_journal *journal;
ccd979bd 2061 struct ocfs2_super *osb;
3bdb8efd 2062 u64 total_blocks;
ccd979bd 2063
cd861280 2064 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF
2065 if (!osb) {
2066 status = -ENOMEM;
2067 mlog_errno(status);
2068 goto bail;
2069 }
2070
2071 sb->s_fs_info = osb;
2072 sb->s_op = &ocfs2_sops;
ba87167c 2073 sb->s_d_op = &ocfs2_dentry_ops;
ccd979bd 2074 sb->s_export_op = &ocfs2_export_ops;
19ece546
JK
2075 sb->s_qcop = &ocfs2_quotactl_ops;
2076 sb->dq_op = &ocfs2_quota_operations;
cf1d6c76 2077 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 2078 sb->s_time_gran = 1;
ccd979bd
MF
2079 sb->s_flags |= MS_NOATIME;
2080 /* this is needed to support O_LARGEFILE */
5a254031
MF
2081 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2082 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2083 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd 2084
9b7895ef
MF
2085 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2086
2087 for (i = 0; i < 3; i++)
2088 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2089 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2090
ccd979bd
MF
2091 osb->sb = sb;
2092 /* Save off for ocfs2_rw_direct */
2093 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2094 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2095
34d024f8
MF
2096 spin_lock_init(&osb->dc_task_lock);
2097 init_waitqueue_head(&osb->dc_event);
2098 osb->dc_work_sequence = 0;
2099 osb->dc_wake_sequence = 0;
ccd979bd
MF
2100 INIT_LIST_HEAD(&osb->blocked_lock_list);
2101 osb->blocked_lock_count = 0;
ccd979bd 2102 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2103 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2104 ocfs2_init_steal_slots(osb);
ccd979bd 2105
43b10a20 2106 mutex_init(&osb->system_file_mutex);
2107
ccd979bd
MF
2108 atomic_set(&osb->alloc_stats.moves, 0);
2109 atomic_set(&osb->alloc_stats.local_data, 0);
2110 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2111 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2112 atomic_set(&osb->alloc_stats.bg_extends, 0);
2113
73be192b
JB
2114 /* Copy the blockcheck stats from the superblock probe */
2115 osb->osb_ecc_stats = *stats;
2116
ccd979bd
MF
2117 ocfs2_init_node_maps(osb);
2118
2119 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2120 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2121
75d9bbc7
GR
2122 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2123 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2124 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2125 osb->max_slots);
2126 status = -EINVAL;
2127 goto bail;
2128 }
75d9bbc7 2129
8b712cd5
JM
2130 ocfs2_orphan_scan_init(osb);
2131
553abd04
JB
2132 status = ocfs2_recovery_init(osb);
2133 if (status) {
2134 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2135 mlog_errno(status);
2136 goto bail;
2137 }
ccd979bd
MF
2138
2139 init_waitqueue_head(&osb->checkpoint_event);
ccd979bd 2140
7f1a37e3
TY
2141 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2142
ccd979bd
MF
2143 osb->slot_num = OCFS2_INVALID_SLOT;
2144
8154da3d
TY
2145 osb->s_xattr_inline_size = le16_to_cpu(
2146 di->id2.i_super.s_xattr_inline_size);
fdd77704 2147
ccd979bd
MF
2148 osb->local_alloc_state = OCFS2_LA_UNUSED;
2149 osb->local_alloc_bh = NULL;
9c7af40b 2150 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2151
ccd979bd
MF
2152 init_waitqueue_head(&osb->osb_mount_event);
2153
d02f00cc
MF
2154 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2155 if (status) {
2156 mlog_errno(status);
2157 goto bail;
2158 }
2159
ccd979bd
MF
2160 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2161 if (!osb->vol_label) {
2162 mlog(ML_ERROR, "unable to alloc vol label\n");
2163 status = -ENOMEM;
2164 goto bail;
2165 }
2166
539d8264
SM
2167 osb->slot_recovery_generations =
2168 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2169 GFP_KERNEL);
2170 if (!osb->slot_recovery_generations) {
2171 status = -ENOMEM;
2172 mlog_errno(status);
2173 goto bail;
2174 }
2175
b4df6ed8
MF
2176 init_waitqueue_head(&osb->osb_wipe_event);
2177 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2178 sizeof(*osb->osb_orphan_wipes),
2179 GFP_KERNEL);
2180 if (!osb->osb_orphan_wipes) {
2181 status = -ENOMEM;
2182 mlog_errno(status);
2183 goto bail;
2184 }
2185
374a263e
TM
2186 osb->osb_rf_lock_tree = RB_ROOT;
2187
ccd979bd
MF
2188 osb->s_feature_compat =
2189 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2190 osb->s_feature_ro_compat =
2191 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2192 osb->s_feature_incompat =
2193 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2194
2195 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2196 mlog(ML_ERROR, "couldn't mount because of unsupported "
2197 "optional features (%x).\n", i);
2198 status = -EINVAL;
2199 goto bail;
2200 }
2201 if (!(osb->sb->s_flags & MS_RDONLY) &&
2202 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2203 mlog(ML_ERROR, "couldn't mount RDWR because of "
2204 "unsupported optional features (%x).\n", i);
2205 status = -EINVAL;
2206 goto bail;
2207 }
2208
98f486f2
SM
2209 if (ocfs2_clusterinfo_valid(osb)) {
2210 osb->osb_stackflags =
2211 OCFS2_RAW_SB(di)->s_cluster_info.ci_stackflags;
c74a3bdd 2212 strlcpy(osb->osb_cluster_stack,
b61817e1 2213 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
c74a3bdd 2214 OCFS2_STACK_LABEL_LEN + 1);
b61817e1
JB
2215 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2216 mlog(ML_ERROR,
2217 "couldn't mount because of an invalid "
2218 "cluster stack label (%s) \n",
2219 osb->osb_cluster_stack);
2220 status = -EINVAL;
2221 goto bail;
2222 }
c74a3bdd
GR
2223 strlcpy(osb->osb_cluster_name,
2224 OCFS2_RAW_SB(di)->s_cluster_info.ci_cluster,
2225 OCFS2_CLUSTER_NAME_LEN + 1);
b61817e1
JB
2226 } else {
2227 /* The empty string is identical with classic tools that
2228 * don't know about s_cluster_info. */
2229 osb->osb_cluster_stack[0] = '\0';
2230 }
2231
ccd979bd
MF
2232 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2233
2234 /* FIXME
2235 * This should be done in ocfs2_journal_init(), but unknown
2236 * ordering issues will cause the filesystem to crash.
2237 * If anyone wants to figure out what part of the code
2238 * refers to osb->journal before ocfs2_journal_init() is run,
2239 * be my guest.
2240 */
2241 /* initialize our journal structure */
2242
cd861280 2243 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF
2244 if (!journal) {
2245 mlog(ML_ERROR, "unable to alloc journal\n");
2246 status = -ENOMEM;
2247 goto bail;
2248 }
2249 osb->journal = journal;
2250 journal->j_osb = osb;
2251
2252 atomic_set(&journal->j_num_trans, 0);
2253 init_rwsem(&journal->j_trans_barrier);
2254 init_waitqueue_head(&journal->j_checkpointed);
2255 spin_lock_init(&journal->j_lock);
2256 journal->j_trans_id = (unsigned long) 1;
2257 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2258 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF
2259 journal->j_state = OCFS2_JOURNAL_FREE;
2260
e3a767b6
JK
2261 INIT_WORK(&osb->dquot_drop_work, ocfs2_drop_dquot_refs);
2262 init_llist_head(&osb->dquot_drop_list);
2263
ccd979bd
MF
2264 /* get some pseudo constants for clustersize bits */
2265 osb->s_clustersize_bits =
2266 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2267 osb->s_clustersize = 1 << osb->s_clustersize_bits;
ccd979bd
MF
2268
2269 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2270 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2271 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2272 osb->s_clustersize);
2273 status = -EINVAL;
2274 goto bail;
2275 }
2276
3bdb8efd
PL
2277 total_blocks = ocfs2_clusters_to_blocks(osb->sb,
2278 le32_to_cpu(di->i_clusters));
2279
2280 status = generic_check_addressable(osb->sb->s_blocksize_bits,
2281 total_blocks);
2282 if (status) {
2283 mlog(ML_ERROR, "Volume too large "
2284 "to mount safely on this system");
2285 status = -EFBIG;
ccd979bd
MF
2286 goto bail;
2287 }
2288
2289 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2290 sizeof(di->id2.i_super.s_uuid))) {
2291 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2292 status = -ENOMEM;
2293 goto bail;
2294 }
2295
69201bb1
FF
2296 strlcpy(osb->vol_label, di->id2.i_super.s_label,
2297 OCFS2_MAX_VOL_LABEL_LEN);
ccd979bd
MF
2298 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2299 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2300 osb->first_cluster_group_blkno =
2301 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2302 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2303 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
32a42d39
TM
2304 trace_ocfs2_initialize_super(osb->vol_label, osb->uuid_str,
2305 (unsigned long long)osb->root_blkno,
2306 (unsigned long long)osb->system_dir_blkno,
2307 osb->s_clustersize_bits);
ccd979bd
MF
2308
2309 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2310 if (!osb->osb_dlm_debug) {
2311 status = -ENOMEM;
2312 mlog_errno(status);
2313 goto bail;
2314 }
2315
2316 atomic_set(&osb->vol_state, VOLUME_INIT);
2317
2318 /* load root, system_dir, and all global system inodes */
2319 status = ocfs2_init_global_system_inodes(osb);
2320 if (status < 0) {
2321 mlog_errno(status);
2322 goto bail;
2323 }
2324
2325 /*
2326 * global bitmap
2327 */
2328 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2329 OCFS2_INVALID_SLOT);
2330 if (!inode) {
2331 status = -EINVAL;
2332 mlog_errno(status);
2333 goto bail;
2334 }
2335
2336 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
6b82021b 2337 osb->osb_clusters_at_boot = OCFS2_I(inode)->ip_clusters;
ccd979bd 2338 iput(inode);
ccd979bd 2339
8571882c
TM
2340 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb, 0,
2341 osb->s_feature_incompat) * 8;
ccd979bd
MF
2342
2343 status = ocfs2_init_slot_info(osb);
2344 if (status < 0) {
2345 mlog_errno(status);
2346 goto bail;
2347 }
e41d33af 2348 cleancache_init_shared_fs((char *)&di->id2.i_super.s_uuid, sb);
ccd979bd 2349
ccd979bd 2350bail:
ccd979bd
MF
2351 return status;
2352}
2353
2354/*
2355 * will return: -EAGAIN if it is ok to keep searching for superblocks
2356 * -EINVAL if there is a bad superblock
2357 * 0 on success
2358 */
2359static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2360 struct buffer_head *bh,
73be192b
JB
2361 u32 blksz,
2362 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF
2363{
2364 int status = -EAGAIN;
2365
ccd979bd
MF
2366 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2367 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB
2368 /* We have to do a raw check of the feature here */
2369 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2370 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2371 status = ocfs2_block_check_validate(bh->b_data,
2372 bh->b_size,
73be192b
JB
2373 &di->i_check,
2374 stats);
d030cc97
JB
2375 if (status)
2376 goto out;
2377 }
ccd979bd
MF
2378 status = -EINVAL;
2379 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2380 mlog(ML_ERROR, "found superblock with incorrect block "
2381 "size: found %u, should be %u\n",
2382 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2383 blksz);
2384 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2385 OCFS2_MAJOR_REV_LEVEL ||
2386 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2387 OCFS2_MINOR_REV_LEVEL) {
2388 mlog(ML_ERROR, "found superblock with bad version: "
2389 "found %u.%u, should be %u.%u\n",
2390 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2391 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2392 OCFS2_MAJOR_REV_LEVEL,
2393 OCFS2_MINOR_REV_LEVEL);
2394 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2395 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2396 "found %llu, should be %llu\n",
1ca1a111 2397 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2398 (unsigned long long)bh->b_blocknr);
ccd979bd
MF
2399 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2400 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2401 mlog(ML_ERROR, "bad cluster size found: %u\n",
2402 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2403 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2404 mlog(ML_ERROR, "bad root_blkno: 0\n");
2405 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2406 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2407 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2408 mlog(ML_ERROR,
2409 "Superblock slots found greater than file system "
2410 "maximum: found %u, max %u\n",
2411 le16_to_cpu(di->id2.i_super.s_max_slots),
2412 OCFS2_MAX_SLOTS);
2413 } else {
2414 /* found it! */
2415 status = 0;
2416 }
2417 }
2418
d030cc97 2419out:
c1e8d35e
TM
2420 if (status && status != -EAGAIN)
2421 mlog_errno(status);
ccd979bd
MF
2422 return status;
2423}
2424
2425static int ocfs2_check_volume(struct ocfs2_super *osb)
2426{
bddb8eb3 2427 int status;
ccd979bd 2428 int dirty;
c271c5c2 2429 int local;
ccd979bd
MF
2430 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2431 * recover
2432 * ourselves. */
2433
ccd979bd
MF
2434 /* Init our journal object. */
2435 status = ocfs2_journal_init(osb->journal, &dirty);
2436 if (status < 0) {
2437 mlog(ML_ERROR, "Could not initialize journal!\n");
2438 goto finally;
2439 }
2440
3bdb8efd
PL
2441 /* Now that journal has been initialized, check to make sure
2442 entire volume is addressable. */
2443 status = ocfs2_journal_addressable(osb);
2444 if (status)
2445 goto finally;
2446
ccd979bd
MF
2447 /* If the journal was unmounted cleanly then we don't want to
2448 * recover anything. Otherwise, journal_load will do that
2449 * dirty work for us :) */
2450 if (!dirty) {
2451 status = ocfs2_journal_wipe(osb->journal, 0);
2452 if (status < 0) {
2453 mlog_errno(status);
2454 goto finally;
2455 }
2456 } else {
619c200d
SM
2457 printk(KERN_NOTICE "ocfs2: File system on device (%s) was not "
2458 "unmounted cleanly, recovering it.\n", osb->dev_str);
ccd979bd
MF
2459 }
2460
c271c5c2
SM
2461 local = ocfs2_mount_local(osb);
2462
ccd979bd 2463 /* will play back anything left in the journal. */
539d8264 2464 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW
2465 if (status < 0) {
2466 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2467 goto finally;
2468 }
ccd979bd
MF
2469
2470 if (dirty) {
2471 /* recover my local alloc if we didn't unmount cleanly. */
2472 status = ocfs2_begin_local_alloc_recovery(osb,
2473 osb->slot_num,
2474 &local_alloc);
2475 if (status < 0) {
2476 mlog_errno(status);
2477 goto finally;
2478 }
2479 /* we complete the recovery process after we've marked
2480 * ourselves as mounted. */
2481 }
2482
ccd979bd
MF
2483 status = ocfs2_load_local_alloc(osb);
2484 if (status < 0) {
2485 mlog_errno(status);
2486 goto finally;
2487 }
2488
2489 if (dirty) {
2490 /* Recovery will be completed after we've mounted the
2491 * rest of the volume. */
2492 osb->dirty = 1;
2493 osb->local_alloc_copy = local_alloc;
2494 local_alloc = NULL;
2495 }
2496
2497 /* go through each journal, trylock it and if you get the
2498 * lock, and it's marked as dirty, set the bit in the recover
2499 * map and launch a recovery thread for it. */
2500 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE
2501 if (status < 0) {
2502 mlog_errno(status);
2503 goto finally;
2504 }
2505
2506 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF
2507 if (status < 0)
2508 mlog_errno(status);
2509
2510finally:
d787ab09 2511 kfree(local_alloc);
ccd979bd 2512
c1e8d35e
TM
2513 if (status)
2514 mlog_errno(status);
ccd979bd
MF
2515 return status;
2516}
2517
2518/*
2519 * The routine gets called from dismount or close whenever a dismount on
2520 * volume is requested and the osb open count becomes 1.
2521 * It will remove the osb from the global list and also free up all the
2522 * initialized resources and fileobject.
2523 */
2524static void ocfs2_delete_osb(struct ocfs2_super *osb)
2525{
ccd979bd
MF
2526 /* This function assumes that the caller has the main osb resource */
2527
8e8a4603 2528 ocfs2_free_slot_info(osb);
ccd979bd 2529
b4df6ed8 2530 kfree(osb->osb_orphan_wipes);
539d8264 2531 kfree(osb->slot_recovery_generations);
ccd979bd
MF
2532 /* FIXME
2533 * This belongs in journal shutdown, but because we have to
421f91d2 2534 * allocate osb->journal at the start of ocfs2_initialize_osb(),
ccd979bd
MF
2535 * we free it here.
2536 */
2537 kfree(osb->journal);
d787ab09 2538 kfree(osb->local_alloc_copy);
ccd979bd 2539 kfree(osb->uuid_str);
f13a568e 2540 kfree(osb->vol_label);
ccd979bd
MF
2541 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2542 memset(osb, 0, sizeof(struct ocfs2_super));
ccd979bd
MF
2543}
2544
2545/* Put OCFS2 into a readonly state, or (if the user specifies it),
2546 * panic(). We do not support continue-on-error operation. */
2547static void ocfs2_handle_error(struct super_block *sb)
2548{
2549 struct ocfs2_super *osb = OCFS2_SB(sb);
2550
2551 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2552 panic("OCFS2: (device %s): panic forced after error\n",
2553 sb->s_id);
2554
2555 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2556
2557 if (sb->s_flags & MS_RDONLY &&
2558 (ocfs2_is_soft_readonly(osb) ||
2559 ocfs2_is_hard_readonly(osb)))
2560 return;
2561
2562 printk(KERN_CRIT "File system is now read-only due to the potential "
2563 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2564 "system is unmounted.\n");
2565 sb->s_flags |= MS_RDONLY;
2566 ocfs2_set_ro_flag(osb, 0);
2567}
2568
2569static char error_buf[1024];
2570
2571void __ocfs2_error(struct super_block *sb,
2572 const char *function,
2573 const char *fmt, ...)
2574{
2575 va_list args;
2576
2577 va_start(args, fmt);
4a6e617a 2578 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF
2579 va_end(args);
2580
2581 /* Not using mlog here because we want to show the actual
2582 * function the error came from. */
2583 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2584 sb->s_id, function, error_buf);
2585
2586 ocfs2_handle_error(sb);
2587}
2588
2589/* Handle critical errors. This is intentionally more drastic than
2590 * ocfs2_handle_error, so we only use for things like journal errors,
2591 * etc. */
2592void __ocfs2_abort(struct super_block* sb,
2593 const char *function,
2594 const char *fmt, ...)
2595{
2596 va_list args;
2597
2598 va_start(args, fmt);
4a6e617a 2599 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF
2600 va_end(args);
2601
2602 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2603 sb->s_id, function, error_buf);
2604
2605 /* We don't have the cluster support yet to go straight to
2606 * hard readonly in here. Until then, we want to keep
2607 * ocfs2_abort() so that we can at least mark critical
2608 * errors.
2609 *
2610 * TODO: This should abort the journal and alert other nodes
2611 * that our slot needs recovery. */
2612
2613 /* Force a panic(). This stinks, but it's better than letting
2614 * things continue without having a proper hard readonly
2615 * here. */
a2f2ddbf
SM
2616 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2617 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF
2618 ocfs2_handle_error(sb);
2619}
2620
e4b963f1
JB
2621/*
2622 * Void signal blockers, because in-kernel sigprocmask() only fails
2623 * when SIG_* is wrong.
2624 */
2625void ocfs2_block_signals(sigset_t *oldset)
2626{
2627 int rc;
2628 sigset_t blocked;
2629
2630 sigfillset(&blocked);
2631 rc = sigprocmask(SIG_BLOCK, &blocked, oldset);
2632 BUG_ON(rc);
2633}
2634
2635void ocfs2_unblock_signals(sigset_t *oldset)
2636{
2637 int rc = sigprocmask(SIG_SETMASK, oldset, NULL);
2638 BUG_ON(rc);
2639}
2640
ccd979bd
MF
2641module_init(ocfs2_init);
2642module_exit(ocfs2_exit);