projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: Fix ocfs2_osb_dump()
[linux-2.6-block.git] / fs / ocfs2 / super.c
CommitLineData
ccd979bd
MF		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * super.c
5 *
6 * load/unload driver, mount/dismount volumes
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/module.h>
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
31#include <linux/utsname.h>
32#include <linux/init.h>
33#include <linux/random.h>
34#include <linux/statfs.h>
35#include <linux/moduleparam.h>
36#include <linux/blkdev.h>
37#include <linux/socket.h>
38#include <linux/inet.h>
39#include <linux/parser.h>
40#include <linux/crc32.h>
41#include <linux/debugfs.h>
d550071c 42#include <linux/mount.h>
6953b4c0 43#include <linux/seq_file.h>
19ece546 44#include <linux/quotaops.h>
337eb00a 45#include <linux/smp_lock.h>
ccd979bd
MF		 46
47#define MLOG_MASK_PREFIX ML_SUPER
48#include <cluster/masklog.h>
49
50#include "ocfs2.h"
51
52/* this should be the only file to include a version 1 header */
53#include "ocfs1_fs_compat.h"
54
55#include "alloc.h"
d030cc97 56#include "blockcheck.h"
ccd979bd
MF		 57#include "dlmglue.h"
58#include "export.h"
59#include "extent_map.h"
60#include "heartbeat.h"
61#include "inode.h"
62#include "journal.h"
63#include "localalloc.h"
64#include "namei.h"
65#include "slot_map.h"
66#include "super.h"
67#include "sysfile.h"
68#include "uptodate.h"
69#include "ver.h"
cf1d6c76 70#include "xattr.h"
9e33d69f 71#include "quota.h"
ccd979bd
MF		 72
73#include "buffer_head_io.h"
74
e18b890b 75static struct kmem_cache *ocfs2_inode_cachep = NULL;
9e33d69f
JK		 76struct kmem_cache *ocfs2_dquot_cachep;
77struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 78
ccd979bd
MF		 79/* OCFS2 needs to schedule several differnt types of work which
80 * require cluster locking, disk I/O, recovery waits, etc. Since these
81 * types of work tend to be heavy we avoid using the kernel events
82 * workqueue and schedule on our own. */
83struct workqueue_struct *ocfs2_wq = NULL;
84
85static struct dentry *ocfs2_debugfs_root = NULL;
86
87MODULE_AUTHOR("Oracle");
88MODULE_LICENSE("GPL");
89
c0123ade
TY		 90struct mount_options
91{
d147b3d6 92 unsigned long commit_interval;
c0123ade
TY		 93 unsigned long mount_opt;
94 unsigned int atime_quantum;
95 signed short slot;
2fbe8d1e 96 unsigned int localalloc_opt;
b61817e1 97 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 98};
99
ccd979bd 100static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 101 struct mount_options *mopt,
baf4661a 102 int is_remount);
d550071c 103static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
ccd979bd
MF		 104static void ocfs2_put_super(struct super_block *sb);
105static int ocfs2_mount_volume(struct super_block *sb);
106static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
107static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
108static int ocfs2_initialize_mem_caches(void);
109static void ocfs2_free_mem_caches(void);
110static void ocfs2_delete_osb(struct ocfs2_super *osb);
111
726c3342 112static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 113
114static int ocfs2_sync_fs(struct super_block *sb, int wait);
115
116static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
117static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 118static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 119static int ocfs2_check_volume(struct ocfs2_super *osb);
120static int ocfs2_verify_volume(struct ocfs2_dinode *di,
121 struct buffer_head *bh,
73be192b
JB		 122 u32 sectsize,
123 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 124static int ocfs2_initialize_super(struct super_block *sb,
125 struct buffer_head *bh,
73be192b
JB		 126 int sector_size,
127 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 128static int ocfs2_get_sector(struct super_block *sb,
129 struct buffer_head **bh,
130 int block,
131 int sect_size);
ccd979bd
MF		 132static struct inode *ocfs2_alloc_inode(struct super_block *sb);
133static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 134static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
135static int ocfs2_enable_quotas(struct ocfs2_super *osb);
136static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 137
ee9b6d61 138static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 139 .statfs = ocfs2_statfs,
140 .alloc_inode = ocfs2_alloc_inode,
141 .destroy_inode = ocfs2_destroy_inode,
142 .drop_inode = ocfs2_drop_inode,
143 .clear_inode = ocfs2_clear_inode,
144 .delete_inode = ocfs2_delete_inode,
145 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 146 .put_super = ocfs2_put_super,
147 .remount_fs = ocfs2_remount,
d550071c 148 .show_options = ocfs2_show_options,
9e33d69f
JK		 149 .quota_read = ocfs2_quota_read,
150 .quota_write = ocfs2_quota_write,
ccd979bd
MF		 151};
152
153enum {
154 Opt_barrier,
155 Opt_err_panic,
156 Opt_err_ro,
157 Opt_intr,
158 Opt_nointr,
159 Opt_hb_none,
160 Opt_hb_local,
161 Opt_data_ordered,
162 Opt_data_writeback,
7f1a37e3 163 Opt_atime_quantum,
baf4661a 164 Opt_slot,
d147b3d6 165 Opt_commit,
2fbe8d1e 166 Opt_localalloc,
53fc622b 167 Opt_localflocks,
b61817e1 168 Opt_stack,
cf1d6c76
TY		 169 Opt_user_xattr,
170 Opt_nouser_xattr,
12462f1d 171 Opt_inode64,
a68979b8
TY		 172 Opt_acl,
173 Opt_noacl,
19ece546
JK		 174 Opt_usrquota,
175 Opt_grpquota,
ccd979bd
MF		 176 Opt_err,
177};
178
a447c093 179static const match_table_t tokens = {
ccd979bd
MF		 180 {Opt_barrier, "barrier=%u"},
181 {Opt_err_panic, "errors=panic"},
182 {Opt_err_ro, "errors=remount-ro"},
183 {Opt_intr, "intr"},
184 {Opt_nointr, "nointr"},
185 {Opt_hb_none, OCFS2_HB_NONE},
186 {Opt_hb_local, OCFS2_HB_LOCAL},
187 {Opt_data_ordered, "data=ordered"},
188 {Opt_data_writeback, "data=writeback"},
7f1a37e3 189 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 190 {Opt_slot, "preferred_slot=%u"},
d147b3d6 191 {Opt_commit, "commit=%u"},
2fbe8d1e 192 {Opt_localalloc, "localalloc=%d"},
53fc622b 193 {Opt_localflocks, "localflocks"},
b61817e1 194 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 195 {Opt_user_xattr, "user_xattr"},
196 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 197 {Opt_inode64, "inode64"},
a68979b8
TY		 198 {Opt_acl, "acl"},
199 {Opt_noacl, "noacl"},
19ece546
JK		 200 {Opt_usrquota, "usrquota"},
201 {Opt_grpquota, "grpquota"},
ccd979bd
MF		 202 {Opt_err, NULL}
203};
204
50397507
SM		 205#ifdef CONFIG_DEBUG_FS
206static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
207{
208 int out = 0;
209 int i;
210 struct ocfs2_cluster_connection *cconn = osb->cconn;
211 struct ocfs2_recovery_map *rm = osb->recovery_map;
15633a22 212 struct ocfs2_orphan_scan *os;
50397507
SM		 213
214 out += snprintf(buf + out, len - out,
215 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
216 "Device", osb->dev_str, osb->uuid_str,
217 osb->fs_generation, osb->vol_label);
218
219 out += snprintf(buf + out, len - out,
220 "%10s => State: %d Flags: 0x%lX\n", "Volume",
221 atomic_read(&osb->vol_state), osb->osb_flags);
222
223 out += snprintf(buf + out, len - out,
224 "%10s => Block: %lu Cluster: %d\n", "Sizes",
225 osb->sb->s_blocksize, osb->s_clustersize);
226
227 out += snprintf(buf + out, len - out,
228 "%10s => Compat: 0x%X Incompat: 0x%X "
229 "ROcompat: 0x%X\n",
230 "Features", osb->s_feature_compat,
231 osb->s_feature_incompat, osb->s_feature_ro_compat);
232
233 out += snprintf(buf + out, len - out,
234 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
235 osb->s_mount_opt, osb->s_atime_quantum);
236
c3d38840
SM		 237 if (cconn) {
238 out += snprintf(buf + out, len - out,
239 "%10s => Stack: %s Name: %*s "
240 "Version: %d.%d\n", "Cluster",
241 (*osb->osb_cluster_stack == '\0' ?
242 "o2cb" : osb->osb_cluster_stack),
243 cconn->cc_namelen, cconn->cc_name,
244 cconn->cc_version.pv_major,
245 cconn->cc_version.pv_minor);
246 }
50397507
SM		 247
248 spin_lock(&osb->dc_task_lock);
249 out += snprintf(buf + out, len - out,
250 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
251 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 252 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
253 osb->blocked_lock_count, osb->dc_wake_sequence,
254 osb->dc_work_sequence);
50397507
SM		 255 spin_unlock(&osb->dc_task_lock);
256
257 spin_lock(&osb->osb_lock);
258 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
259 "Recovery",
260 (osb->recovery_thread_task ?
261 task_pid_nr(osb->recovery_thread_task) : -1));
262 if (rm->rm_used == 0)
263 out += snprintf(buf + out, len - out, " None\n");
264 else {
265 for (i = 0; i < rm->rm_used; i++)
266 out += snprintf(buf + out, len - out, " %d",
267 rm->rm_entries[i]);
268 out += snprintf(buf + out, len - out, "\n");
269 }
270 spin_unlock(&osb->osb_lock);
271
272 out += snprintf(buf + out, len - out,
273 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
c3d38840
SM		 274 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
275 osb->osb_commit_interval,
50397507
SM		 276 atomic_read(&osb->needs_checkpoint));
277
278 out += snprintf(buf + out, len - out,
c3d38840 279 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 280 "Journal", osb->journal->j_state,
c3d38840
SM		 281 osb->journal->j_trans_id,
282 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 283
284 out += snprintf(buf + out, len - out,
285 "%10s => GlobalAllocs: %d LocalAllocs: %d "
286 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
287 "Stats",
288 atomic_read(&osb->alloc_stats.bitmap_data),
289 atomic_read(&osb->alloc_stats.local_data),
290 atomic_read(&osb->alloc_stats.bg_allocs),
291 atomic_read(&osb->alloc_stats.moves),
292 atomic_read(&osb->alloc_stats.bg_extends));
293
294 out += snprintf(buf + out, len - out,
295 "%10s => State: %u Descriptor: %llu Size: %u bits "
296 "Default: %u bits\n",
297 "LocalAlloc", osb->local_alloc_state,
298 (unsigned long long)osb->la_last_gd,
299 osb->local_alloc_bits, osb->local_alloc_default_bits);
300
301 spin_lock(&osb->osb_lock);
302 out += snprintf(buf + out, len - out,
303 "%10s => Slot: %d NumStolen: %d\n", "Steal",
304 osb->s_inode_steal_slot,
305 atomic_read(&osb->s_num_inodes_stolen));
306 spin_unlock(&osb->osb_lock);
307
308 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
309 "Slots", "Num", "RecoGen");
50397507
SM		 310 for (i = 0; i < osb->max_slots; ++i) {
311 out += snprintf(buf + out, len - out,
312 "%10s %c %3d %10d\n",
313 " ",
314 (i == osb->slot_num ? '*' : ' '),
315 i, osb->slot_recovery_generations[i]);
316 }
317
15633a22
SE		 318 os = &osb->osb_orphan_scan;
319 out += snprintf(buf + out, len - out, "Orphan Scan=> ");
320 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
321 os->os_count, os->os_seqno);
322 out += snprintf(buf + out, len - out, " Last Scan: %lu seconds ago\n",
323 (get_seconds() - os->os_scantime.tv_sec));
324
50397507
SM		 325 return out;
326}
327
328static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
329{
330 struct ocfs2_super *osb = inode->i_private;
331 char *buf = NULL;
332
333 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
334 if (!buf)
335 goto bail;
336
337 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
338
339 file->private_data = buf;
340
341 return 0;
342bail:
343 return -ENOMEM;
344}
345
346static int ocfs2_debug_release(struct inode *inode, struct file *file)
347{
348 kfree(file->private_data);
349 return 0;
350}
351
352static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
353 size_t nbytes, loff_t *ppos)
354{
355 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
356 i_size_read(file->f_mapping->host));
357}
358#else
359static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
360{
361 return 0;
362}
363static int ocfs2_debug_release(struct inode *inode, struct file *file)
364{
365 return 0;
366}
367static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
368 size_t nbytes, loff_t *ppos)
369{
370 return 0;
371}
372#endif /* CONFIG_DEBUG_FS */
373
374static struct file_operations ocfs2_osb_debug_fops = {
375 .open = ocfs2_osb_debug_open,
376 .release = ocfs2_debug_release,
377 .read = ocfs2_debug_read,
378 .llseek = generic_file_llseek,
379};
380
ccd979bd
MF		 381static int ocfs2_sync_fs(struct super_block *sb, int wait)
382{
bddb8eb3 383 int status;
ccd979bd
MF		 384 tid_t target;
385 struct ocfs2_super *osb = OCFS2_SB(sb);
386
ccd979bd
MF		 387 if (ocfs2_is_hard_readonly(osb))
388 return -EROFS;
389
390 if (wait) {
391 status = ocfs2_flush_truncate_log(osb);
392 if (status < 0)
393 mlog_errno(status);
394 } else {
395 ocfs2_schedule_truncate_log_flush(osb, 0);
396 }
397
2b4e30fb
JB		 398 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
399 &target)) {
ccd979bd 400 if (wait)
2b4e30fb
JB		 401 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
402 target);
ccd979bd
MF		 403 }
404 return 0;
405}
406
1a224ad1
JK		 407static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
408{
409 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
410 && (ino == USER_QUOTA_SYSTEM_INODE
411 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
412 return 0;
413 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
414 && (ino == GROUP_QUOTA_SYSTEM_INODE
415 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
416 return 0;
417 return 1;
418}
419
ccd979bd
MF		 420static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
421{
422 struct inode *new = NULL;
423 int status = 0;
424 int i;
425
426 mlog_entry_void();
427
5fa0613e 428 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 429 if (IS_ERR(new)) {
430 status = PTR_ERR(new);
431 mlog_errno(status);
432 goto bail;
433 }
434 osb->root_inode = new;
435
5fa0613e 436 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 437 if (IS_ERR(new)) {
438 status = PTR_ERR(new);
439 mlog_errno(status);
440 goto bail;
441 }
442 osb->sys_root_inode = new;
443
444 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
445 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 446 if (!ocfs2_need_system_inode(osb, i))
447 continue;
ccd979bd
MF		 448 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
449 if (!new) {
450 ocfs2_release_system_inodes(osb);
451 status = -EINVAL;
452 mlog_errno(status);
453 /* FIXME: Should ERROR_RO_FS */
454 mlog(ML_ERROR, "Unable to load system inode %d, "
455 "possibly corrupt fs?", i);
456 goto bail;
457 }
458 // the array now has one ref, so drop this one
459 iput(new);
460 }
461
462bail:
463 mlog_exit(status);
464 return status;
465}
466
467static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
468{
469 struct inode *new = NULL;
470 int status = 0;
471 int i;
472
473 mlog_entry_void();
474
475 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
476 i < NUM_SYSTEM_INODES;
477 i++) {
1a224ad1
JK		 478 if (!ocfs2_need_system_inode(osb, i))
479 continue;
ccd979bd
MF		 480 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
481 if (!new) {
482 ocfs2_release_system_inodes(osb);
483 status = -EINVAL;
484 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
485 status, i, osb->slot_num);
486 goto bail;
487 }
488 /* the array now has one ref, so drop this one */
489 iput(new);
490 }
491
492bail:
493 mlog_exit(status);
494 return status;
495}
496
bddb8eb3 497static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 498{
bddb8eb3 499 int i;
ccd979bd
MF		 500 struct inode *inode;
501
502 mlog_entry_void();
503
504 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
505 inode = osb->system_inodes[i];
506 if (inode) {
507 iput(inode);
508 osb->system_inodes[i] = NULL;
509 }
510 }
511
512 inode = osb->sys_root_inode;
513 if (inode) {
514 iput(inode);
515 osb->sys_root_inode = NULL;
516 }
517
518 inode = osb->root_inode;
519 if (inode) {
520 iput(inode);
521 osb->root_inode = NULL;
522 }
523
bddb8eb3 524 mlog_exit(0);
ccd979bd
MF		 525}
526
527/* We're allocating fs objects, use GFP_NOFS */
528static struct inode *ocfs2_alloc_inode(struct super_block *sb)
529{
530 struct ocfs2_inode_info *oi;
531
e6b4f8da 532 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 533 if (!oi)
534 return NULL;
535
2b4e30fb 536 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 537 return &oi->vfs_inode;
538}
539
540static void ocfs2_destroy_inode(struct inode *inode)
541{
542 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
543}
544
5a254031
MF		 545static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
546 unsigned int cbits)
ccd979bd 547{
5a254031
MF		 548 unsigned int bytes = 1 << cbits;
549 unsigned int trim = bytes;
550 unsigned int bitshift = 32;
551
552 /*
553 * i_size and all block offsets in ocfs2 are always 64 bits
554 * wide. i_clusters is 32 bits, in cluster-sized units. So on
555 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 556 */
557
558#if BITS_PER_LONG == 32
559# if defined(CONFIG_LBD)
2ecd05ae 560 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 561 /*
562 * We might be limited by page cache size.
563 */
564 if (bytes > PAGE_CACHE_SIZE) {
565 bytes = PAGE_CACHE_SIZE;
566 trim = 1;
567 /*
568 * Shift by 31 here so that we don't get larger than
569 * MAX_LFS_FILESIZE
570 */
571 bitshift = 31;
572 }
ccd979bd 573# else
5a254031
MF		 574 /*
575 * We are limited by the size of sector_t. Use block size, as
576 * that's what we expose to the VFS.
577 */
578 bytes = 1 << bbits;
579 trim = 1;
580 bitshift = 31;
ccd979bd
MF		 581# endif
582#endif
583
5a254031
MF		 584 /*
585 * Trim by a whole cluster when we can actually approach the
586 * on-disk limits. Otherwise we can overflow i_clusters when
587 * an extent start is at the max offset.
588 */
589 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 590}
591
592static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
593{
594 int incompat_features;
595 int ret = 0;
c0123ade 596 struct mount_options parsed_options;
ccd979bd
MF		 597 struct ocfs2_super *osb = OCFS2_SB(sb);
598
337eb00a
AIB		 599 lock_kernel();
600
c0123ade 601 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
ccd979bd
MF		 602 ret = -EINVAL;
603 goto out;
604 }
605
606 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
c0123ade 607 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
ccd979bd
MF		 608 ret = -EINVAL;
609 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
610 goto out;
611 }
612
613 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 614 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 615 ret = -EINVAL;
616 mlog(ML_ERROR, "Cannot change data mode on remount\n");
617 goto out;
618 }
619
12462f1d
JB		 620 /* Probably don't want this on remount; it might
621 * mess with other nodes */
622 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
623 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
624 ret = -EINVAL;
625 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
626 goto out;
627 }
628
ccd979bd
MF		 629 /* We're going to/from readonly mode. */
630 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 631 /* Disable quota accounting before remounting RO */
632 if (*flags & MS_RDONLY) {
633 ret = ocfs2_susp_quotas(osb, 0);
634 if (ret < 0)
635 goto out;
636 }
ccd979bd
MF		 637 /* Lock here so the check of HARD_RO and the potential
638 * setting of SOFT_RO is atomic. */
639 spin_lock(&osb->osb_lock);
640 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
641 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
642 ret = -EROFS;
643 goto unlock_osb;
644 }
645
646 if (*flags & MS_RDONLY) {
647 mlog(0, "Going to ro mode.\n");
648 sb->s_flags |= MS_RDONLY;
649 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
650 } else {
651 mlog(0, "Making ro filesystem writeable.\n");
652
653 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
654 mlog(ML_ERROR, "Cannot remount RDWR "
655 "filesystem due to previous errors.\n");
656 ret = -EROFS;
657 goto unlock_osb;
658 }
659 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
660 if (incompat_features) {
661 mlog(ML_ERROR, "Cannot remount RDWR because "
662 "of unsupported optional features "
663 "(%x).\n", incompat_features);
664 ret = -EINVAL;
665 goto unlock_osb;
666 }
667 sb->s_flags &= ~MS_RDONLY;
668 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
669 }
670unlock_osb:
671 spin_unlock(&osb->osb_lock);
19ece546
JK		 672 /* Enable quota accounting after remounting RW */
673 if (!ret && !(*flags & MS_RDONLY)) {
674 if (sb_any_quota_suspended(sb))
675 ret = ocfs2_susp_quotas(osb, 1);
676 else
677 ret = ocfs2_enable_quotas(osb);
678 if (ret < 0) {
679 /* Return back changes... */
680 spin_lock(&osb->osb_lock);
681 sb->s_flags |= MS_RDONLY;
682 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
683 spin_unlock(&osb->osb_lock);
684 goto out;
685 }
686 }
ccd979bd
MF		 687 }
688
689 if (!ret) {
ccd979bd
MF		 690 /* Only save off the new mount options in case of a successful
691 * remount. */
a68979b8
TY		 692 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
693 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
c0123ade
TY		 694 osb->s_mount_opt = parsed_options.mount_opt;
695 osb->s_atime_quantum = parsed_options.atime_quantum;
696 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 697 if (parsed_options.commit_interval)
698 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 699
700 if (!ocfs2_is_hard_readonly(osb))
701 ocfs2_set_journal_params(osb);
ccd979bd
MF		 702 }
703out:
337eb00a 704 unlock_kernel();
ccd979bd
MF		 705 return ret;
706}
707
708static int ocfs2_sb_probe(struct super_block *sb,
709 struct buffer_head **bh,
73be192b
JB		 710 int *sector_size,
711 struct ocfs2_blockcheck_stats *stats)
ccd979bd 712{
bddb8eb3 713 int status, tmpstat;
ccd979bd
MF		 714 struct ocfs1_vol_disk_hdr *hdr;
715 struct ocfs2_dinode *di;
716 int blksize;
717
718 *bh = NULL;
719
720 /* may be > 512 */
e1defc4f 721 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 722 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
723 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
724 *sector_size, OCFS2_MAX_BLOCKSIZE);
725 status = -EINVAL;
726 goto bail;
727 }
728
729 /* Can this really happen? */
730 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
731 *sector_size = OCFS2_MIN_BLOCKSIZE;
732
733 /* check block zero for old format */
734 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
735 if (status < 0) {
736 mlog_errno(status);
737 goto bail;
738 }
739 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
740 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
741 mlog(ML_ERROR, "incompatible version: %u.%u\n",
742 hdr->major_version, hdr->minor_version);
743 status = -EINVAL;
744 }
745 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
746 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
747 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
748 hdr->signature);
749 status = -EINVAL;
750 }
751 brelse(*bh);
752 *bh = NULL;
753 if (status < 0) {
754 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
755 "upgraded before mounting with ocfs v2\n");
756 goto bail;
757 }
758
759 /*
760 * Now check at magic offset for 512, 1024, 2048, 4096
761 * blocksizes. 4096 is the maximum blocksize because it is
762 * the minimum clustersize.
763 */
764 status = -EINVAL;
765 for (blksize = *sector_size;
766 blksize <= OCFS2_MAX_BLOCKSIZE;
767 blksize <<= 1) {
768 tmpstat = ocfs2_get_sector(sb, bh,
769 OCFS2_SUPER_BLOCK_BLKNO,
770 blksize);
771 if (tmpstat < 0) {
772 status = tmpstat;
773 mlog_errno(status);
774 goto bail;
775 }
776 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b
JB		 777 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
778 status = ocfs2_verify_volume(di, *bh, blksize, stats);
ccd979bd
MF		 779 if (status >= 0)
780 goto bail;
781 brelse(*bh);
782 *bh = NULL;
783 if (status != -EAGAIN)
784 break;
785 }
786
787bail:
788 return status;
789}
790
c271c5c2
SM		 791static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
792{
793 if (ocfs2_mount_local(osb)) {
794 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
795 mlog(ML_ERROR, "Cannot heartbeat on a locally "
796 "mounted device.\n");
797 return -EINVAL;
798 }
799 }
800
b61817e1
JB		 801 if (ocfs2_userspace_stack(osb)) {
802 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
803 mlog(ML_ERROR, "Userspace stack expected, but "
804 "o2cb heartbeat arguments passed to mount\n");
805 return -EINVAL;
806 }
807 }
808
c271c5c2 809 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
b61817e1
JB		 810 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
811 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 812 mlog(ML_ERROR, "Heartbeat has to be started to mount "
813 "a read-write clustered device.\n");
814 return -EINVAL;
815 }
816 }
817
818 return 0;
819}
820
b61817e1
JB		 821/*
822 * If we're using a userspace stack, mount should have passed
823 * a name that matches the disk. If not, mount should not
824 * have passed a stack.
825 */
826static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
827 struct mount_options *mopt)
828{
829 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
830 mlog(ML_ERROR,
831 "cluster stack passed to mount, but this filesystem "
832 "does not support it\n");
833 return -EINVAL;
834 }
835
836 if (ocfs2_userspace_stack(osb) &&
837 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
838 OCFS2_STACK_LABEL_LEN)) {
839 mlog(ML_ERROR,
840 "cluster stack passed to mount (\"%s\") does not "
841 "match the filesystem (\"%s\")\n",
842 mopt->cluster_stack,
843 osb->osb_cluster_stack);
844 return -EINVAL;
845 }
846
847 return 0;
848}
849
19ece546
JK		 850static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
851{
852 int type;
853 struct super_block *sb = osb->sb;
854 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
855 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
856 int status = 0;
857
858 for (type = 0; type < MAXQUOTAS; type++) {
859 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
860 continue;
861 if (unsuspend)
862 status = vfs_quota_enable(
863 sb_dqopt(sb)->files[type],
864 type, QFMT_OCFS2,
865 DQUOT_SUSPENDED);
866 else
867 status = vfs_quota_disable(sb, type,
868 DQUOT_SUSPENDED);
869 if (status < 0)
870 break;
871 }
872 if (status < 0)
873 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
874 "remount (error = %d).\n", status);
875 return status;
876}
877
878static int ocfs2_enable_quotas(struct ocfs2_super *osb)
879{
880 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
881 struct super_block *sb = osb->sb;
882 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
883 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
884 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
885 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
886 int status;
887 int type;
888
889 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
890 for (type = 0; type < MAXQUOTAS; type++) {
891 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
892 continue;
893 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
894 osb->slot_num);
895 if (!inode[type]) {
896 status = -ENOENT;
897 goto out_quota_off;
898 }
899 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
900 DQUOT_USAGE_ENABLED);
901 if (status < 0)
902 goto out_quota_off;
903 }
904
905 for (type = 0; type < MAXQUOTAS; type++)
906 iput(inode[type]);
907 return 0;
908out_quota_off:
909 ocfs2_disable_quotas(osb);
910 for (type = 0; type < MAXQUOTAS; type++)
911 iput(inode[type]);
912 mlog_errno(status);
913 return status;
914}
915
916static void ocfs2_disable_quotas(struct ocfs2_super *osb)
917{
918 int type;
919 struct inode *inode;
920 struct super_block *sb = osb->sb;
921
922 /* We mostly ignore errors in this function because there's not much
923 * we can do when we see them */
924 for (type = 0; type < MAXQUOTAS; type++) {
925 if (!sb_has_quota_loaded(sb, type))
926 continue;
927 inode = igrab(sb->s_dquot.files[type]);
928 /* Turn off quotas. This will remove all dquot structures from
929 * memory and so they will be automatically synced to global
930 * quota files */
931 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
932 DQUOT_LIMITS_ENABLED);
933 if (!inode)
934 continue;
935 iput(inode);
936 }
937}
938
939/* Handle quota on quotactl */
940static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
941 char *path, int remount)
942{
943 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
944 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
945
946 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
947 return -EINVAL;
948
949 if (remount)
950 return 0; /* Just ignore it has been handled in
951 * ocfs2_remount() */
952 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
953 format_id, DQUOT_LIMITS_ENABLED);
954}
955
956/* Handle quota off quotactl */
957static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
958{
959 if (remount)
960 return 0; /* Ignore now and handle later in
961 * ocfs2_remount() */
962 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
963}
964
965static struct quotactl_ops ocfs2_quotactl_ops = {
966 .quota_on = ocfs2_quota_on,
967 .quota_off = ocfs2_quota_off,
968 .quota_sync = vfs_quota_sync,
969 .get_info = vfs_get_dqinfo,
970 .set_info = vfs_set_dqinfo,
971 .get_dqblk = vfs_get_dqblk,
972 .set_dqblk = vfs_set_dqblk,
973};
974
ccd979bd
MF		 975static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
976{
977 struct dentry *root;
978 int status, sector_size;
c0123ade 979 struct mount_options parsed_options;
ccd979bd
MF		 980 struct inode *inode = NULL;
981 struct ocfs2_super *osb = NULL;
982 struct buffer_head *bh = NULL;
c271c5c2 983 char nodestr[8];
73be192b 984 struct ocfs2_blockcheck_stats stats;
ccd979bd
MF		 985
986 mlog_entry("%p, %p, %i", sb, data, silent);
987
c0123ade 988 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 989 status = -EINVAL;
990 goto read_super_error;
991 }
992
993 /* probe for superblock */
73be192b 994 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 995 if (status < 0) {
996 mlog(ML_ERROR, "superblock probe failed!\n");
997 goto read_super_error;
998 }
999
73be192b 1000 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1001 osb = OCFS2_SB(sb);
1002 if (status < 0) {
1003 mlog_errno(status);
1004 goto read_super_error;
1005 }
1006 brelse(bh);
1007 bh = NULL;
a68979b8
TY		 1008
1009 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
1010 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1011
c0123ade
TY		 1012 osb->s_mount_opt = parsed_options.mount_opt;
1013 osb->s_atime_quantum = parsed_options.atime_quantum;
1014 osb->preferred_slot = parsed_options.slot;
d147b3d6 1015 osb->osb_commit_interval = parsed_options.commit_interval;
9c7af40b
MF		 1016 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
1017 osb->local_alloc_bits = osb->local_alloc_default_bits;
19ece546
JK		 1018 if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
1019 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1020 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1021 status = -EINVAL;
1022 mlog(ML_ERROR, "User quotas were requested, but this "
1023 "filesystem does not have the feature enabled.\n");
1024 goto read_super_error;
1025 }
1026 if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1027 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1028 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1029 status = -EINVAL;
1030 mlog(ML_ERROR, "Group quotas were requested, but this "
1031 "filesystem does not have the feature enabled.\n");
1032 goto read_super_error;
1033 }
ccd979bd 1034
b61817e1
JB		 1035 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1036 if (status)
1037 goto read_super_error;
1038
ccd979bd
MF		 1039 sb->s_magic = OCFS2_SUPER_MAGIC;
1040
a68979b8
TY		 1041 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1042 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1043
ccd979bd
MF		 1044 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1045 * heartbeat=none */
1046 if (bdev_read_only(sb->s_bdev)) {
1047 if (!(sb->s_flags & MS_RDONLY)) {
1048 status = -EACCES;
1049 mlog(ML_ERROR, "Readonly device detected but readonly "
1050 "mount was not specified.\n");
1051 goto read_super_error;
1052 }
1053
1054 /* You should not be able to start a local heartbeat
1055 * on a readonly device. */
1056 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1057 status = -EROFS;
1058 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1059 "device.\n");
1060 goto read_super_error;
1061 }
1062
1063 status = ocfs2_check_journals_nolocks(osb);
1064 if (status < 0) {
1065 if (status == -EROFS)
1066 mlog(ML_ERROR, "Recovery required on readonly "
1067 "file system, but write access is "
1068 "unavailable.\n");
1069 else
1070 mlog_errno(status);
1071 goto read_super_error;
1072 }
1073
1074 ocfs2_set_ro_flag(osb, 1);
1075
1076 printk(KERN_NOTICE "Readonly device detected. No cluster "
1077 "services will be utilized for this mount. Recovery "
1078 "will be skipped.\n");
1079 }
1080
1081 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1082 if (sb->s_flags & MS_RDONLY)
1083 ocfs2_set_ro_flag(osb, 0);
1084 }
1085
c271c5c2
SM		 1086 status = ocfs2_verify_heartbeat(osb);
1087 if (status < 0) {
1088 mlog_errno(status);
1089 goto read_super_error;
1090 }
1091
ccd979bd
MF		 1092 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1093 ocfs2_debugfs_root);
1094 if (!osb->osb_debug_root) {
1095 status = -EINVAL;
1096 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1097 goto read_super_error;
1098 }
1099
50397507
SM		 1100 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1101 osb->osb_debug_root,
1102 osb,
1103 &ocfs2_osb_debug_fops);
1104 if (!osb->osb_ctxt) {
1105 status = -EINVAL;
1106 mlog_errno(status);
1107 goto read_super_error;
1108 }
1109
73be192b
JB		 1110 if (ocfs2_meta_ecc(osb)) {
1111 status = ocfs2_blockcheck_stats_debugfs_install(
1112 &osb->osb_ecc_stats,
1113 osb->osb_debug_root);
1114 if (status) {
1115 mlog(ML_ERROR,
1116 "Unable to create blockcheck statistics "
1117 "files\n");
1118 goto read_super_error;
1119 }
1120 }
1121
ccd979bd
MF		 1122 status = ocfs2_mount_volume(sb);
1123 if (osb->root_inode)
1124 inode = igrab(osb->root_inode);
1125
1126 if (status < 0)
1127 goto read_super_error;
1128
1129 if (!inode) {
1130 status = -EIO;
1131 mlog_errno(status);
1132 goto read_super_error;
1133 }
1134
1135 root = d_alloc_root(inode);
1136 if (!root) {
1137 status = -ENOMEM;
1138 mlog_errno(status);
1139 goto read_super_error;
1140 }
1141
1142 sb->s_root = root;
1143
1144 ocfs2_complete_mount_recovery(osb);
1145
c271c5c2
SM		 1146 if (ocfs2_mount_local(osb))
1147 snprintf(nodestr, sizeof(nodestr), "local");
1148 else
19fdb624 1149 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1150
1151 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1152 "with %s data mode.\n",
c271c5c2 1153 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1154 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1155 "ordered");
1156
1157 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1158 wake_up(&osb->osb_mount_event);
1159
19ece546
JK		 1160 /* Now we can initialize quotas because we can afford to wait
1161 * for cluster locks recovery now. That also means that truncation
1162 * log recovery can happen but that waits for proper quota setup */
1163 if (!(sb->s_flags & MS_RDONLY)) {
1164 status = ocfs2_enable_quotas(osb);
1165 if (status < 0) {
1166 /* We have to err-out specially here because
1167 * s_root is already set */
1168 mlog_errno(status);
1169 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1170 wake_up(&osb->osb_mount_event);
1171 mlog_exit(status);
1172 return status;
1173 }
1174 }
1175
1176 ocfs2_complete_quota_recovery(osb);
1177
1178 /* Now we wake up again for processes waiting for quotas */
1179 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1180 wake_up(&osb->osb_mount_event);
1181
ccd979bd
MF		 1182 mlog_exit(status);
1183 return status;
1184
1185read_super_error:
a81cb88b 1186 brelse(bh);
ccd979bd
MF		 1187
1188 if (inode)
1189 iput(inode);
1190
1191 if (osb) {
1192 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1193 wake_up(&osb->osb_mount_event);
1194 ocfs2_dismount_volume(sb, 1);
1195 }
1196
1197 mlog_exit(status);
1198 return status;
1199}
1200
454e2398
DH		 1201static int ocfs2_get_sb(struct file_system_type *fs_type,
1202 int flags,
1203 const char *dev_name,
1204 void *data,
1205 struct vfsmount *mnt)
ccd979bd 1206{
454e2398
DH		 1207 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1208 mnt);
ccd979bd
MF		 1209}
1210
1211static struct file_system_type ocfs2_fs_type = {
1212 .owner = THIS_MODULE,
1213 .name = "ocfs2",
1214 .get_sb = ocfs2_get_sb, /* is this called when we mount
1215 * the fs? */
1216 .kill_sb = kill_block_super, /* set to the generic one
1217 * right now, but do we
1218 * need to change that? */
1ba9da2f 1219 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1220 .next = NULL
1221};
1222
1223static int ocfs2_parse_options(struct super_block *sb,
1224 char *options,
c0123ade 1225 struct mount_options *mopt,
ccd979bd
MF		 1226 int is_remount)
1227{
1228 int status;
1229 char *p;
1230
1231 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1232 options ? options : "(none)");
1233
d147b3d6 1234 mopt->commit_interval = 0;
c0123ade
TY		 1235 mopt->mount_opt = 0;
1236 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1237 mopt->slot = OCFS2_INVALID_SLOT;
2fbe8d1e 1238 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
b61817e1 1239 mopt->cluster_stack[0] = '\0';
ccd979bd
MF		 1240
1241 if (!options) {
1242 status = 1;
1243 goto bail;
1244 }
1245
1246 while ((p = strsep(&options, ",")) != NULL) {
1247 int token, option;
1248 substring_t args[MAX_OPT_ARGS];
1249
1250 if (!*p)
1251 continue;
1252
1253 token = match_token(p, tokens, args);
1254 switch (token) {
1255 case Opt_hb_local:
c0123ade 1256 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1257 break;
1258 case Opt_hb_none:
c0123ade 1259 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1260 break;
1261 case Opt_barrier:
1262 if (match_int(&args[0], &option)) {
1263 status = 0;
1264 goto bail;
1265 }
1266 if (option)
c0123ade 1267 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1268 else
c0123ade 1269 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1270 break;
1271 case Opt_intr:
c0123ade 1272 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1273 break;
1274 case Opt_nointr:
c0123ade 1275 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1276 break;
1277 case Opt_err_panic:
c0123ade 1278 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1279 break;
1280 case Opt_err_ro:
c0123ade 1281 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1282 break;
1283 case Opt_data_ordered:
c0123ade 1284 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1285 break;
1286 case Opt_data_writeback:
c0123ade 1287 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1288 break;
cf1d6c76
TY		 1289 case Opt_user_xattr:
1290 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1291 break;
1292 case Opt_nouser_xattr:
1293 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1294 break;
7f1a37e3
TY		 1295 case Opt_atime_quantum:
1296 if (match_int(&args[0], &option)) {
1297 status = 0;
1298 goto bail;
1299 }
1300 if (option >= 0)
c0123ade 1301 mopt->atime_quantum = option;
7f1a37e3 1302 break;
baf4661a
SM		 1303 case Opt_slot:
1304 option = 0;
1305 if (match_int(&args[0], &option)) {
1306 status = 0;
1307 goto bail;
1308 }
1309 if (option)
c0123ade 1310 mopt->slot = (s16)option;
baf4661a 1311 break;
d147b3d6
MF		 1312 case Opt_commit:
1313 option = 0;
1314 if (match_int(&args[0], &option)) {
1315 status = 0;
1316 goto bail;
1317 }
1318 if (option < 0)
1319 return 0;
1320 if (option == 0)
2b4e30fb 1321 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1322 mopt->commit_interval = HZ * option;
1323 break;
2fbe8d1e
SM		 1324 case Opt_localalloc:
1325 option = 0;
1326 if (match_int(&args[0], &option)) {
1327 status = 0;
1328 goto bail;
1329 }
1330 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1331 mopt->localalloc_opt = option;
1332 break;
53fc622b
MF		 1333 case Opt_localflocks:
1334 /*
1335 * Changing this during remount could race
1336 * flock() requests, or "unbalance" existing
1337 * ones (e.g., a lock is taken in one mode but
1338 * dropped in the other). If users care enough
1339 * to flip locking modes during remount, we
1340 * could add a "local" flag to individual
1341 * flock structures for proper tracking of
1342 * state.
1343 */
1344 if (!is_remount)
1345 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1346 break;
b61817e1
JB		 1347 case Opt_stack:
1348 /* Check both that the option we were passed
1349 * is of the right length and that it is a proper
1350 * string of the right length.
1351 */
1352 if (((args[0].to - args[0].from) !=
1353 OCFS2_STACK_LABEL_LEN) ||
1354 (strnlen(args[0].from,
1355 OCFS2_STACK_LABEL_LEN) !=
1356 OCFS2_STACK_LABEL_LEN)) {
1357 mlog(ML_ERROR,
1358 "Invalid cluster_stack option\n");
1359 status = 0;
1360 goto bail;
1361 }
1362 memcpy(mopt->cluster_stack, args[0].from,
1363 OCFS2_STACK_LABEL_LEN);
1364 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1365 break;
12462f1d
JB		 1366 case Opt_inode64:
1367 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1368 break;
19ece546
JK		 1369 case Opt_usrquota:
1370 /* We check only on remount, otherwise features
1371 * aren't yet initialized. */
1372 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1373 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1374 mlog(ML_ERROR, "User quota requested but "
1375 "filesystem feature is not set\n");
1376 status = 0;
1377 goto bail;
1378 }
1379 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1380 break;
1381 case Opt_grpquota:
1382 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1383 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1384 mlog(ML_ERROR, "Group quota requested but "
1385 "filesystem feature is not set\n");
1386 status = 0;
1387 goto bail;
1388 }
1389 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1390 break;
a68979b8
TY		 1391#ifdef CONFIG_OCFS2_FS_POSIX_ACL
1392 case Opt_acl:
1393 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1394 break;
1395 case Opt_noacl:
1396 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1397 break;
1398#else
1399 case Opt_acl:
1400 case Opt_noacl:
1401 printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
1402 break;
1403#endif
ccd979bd
MF		 1404 default:
1405 mlog(ML_ERROR,
1406 "Unrecognized mount option \"%s\" "
1407 "or missing value\n", p);
1408 status = 0;
1409 goto bail;
1410 }
1411 }
1412
1413 status = 1;
1414
1415bail:
1416 mlog_exit(status);
1417 return status;
1418}
1419
d550071c
SM		 1420static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1421{
1422 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1423 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1424 unsigned int local_alloc_megs;
d550071c
SM		 1425
1426 if (opts & OCFS2_MOUNT_HB_LOCAL)
1427 seq_printf(s, ",_netdev,heartbeat=local");
1428 else
1429 seq_printf(s, ",heartbeat=none");
1430
1431 if (opts & OCFS2_MOUNT_NOINTR)
1432 seq_printf(s, ",nointr");
1433
1434 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1435 seq_printf(s, ",data=writeback");
1436 else
1437 seq_printf(s, ",data=ordered");
1438
1439 if (opts & OCFS2_MOUNT_BARRIER)
1440 seq_printf(s, ",barrier=1");
1441
1442 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1443 seq_printf(s, ",errors=panic");
1444 else
1445 seq_printf(s, ",errors=remount-ro");
1446
1447 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1448 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1449
1450 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1451 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1452
d147b3d6
MF		 1453 if (osb->osb_commit_interval)
1454 seq_printf(s, ",commit=%u",
1455 (unsigned) (osb->osb_commit_interval / HZ));
1456
ebcee4b5
MF		 1457 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1458 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1459 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1460
53fc622b
MF		 1461 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1462 seq_printf(s, ",localflocks,");
1463
b61817e1
JB		 1464 if (osb->osb_cluster_stack[0])
1465 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1466 osb->osb_cluster_stack);
19ece546
JK		 1467 if (opts & OCFS2_MOUNT_USRQUOTA)
1468 seq_printf(s, ",usrquota");
1469 if (opts & OCFS2_MOUNT_GRPQUOTA)
1470 seq_printf(s, ",grpquota");
b61817e1 1471
b0f73cfc
SM		 1472 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1473 seq_printf(s, ",nouser_xattr");
1474 else
1475 seq_printf(s, ",user_xattr");
1476
12462f1d
JB		 1477 if (opts & OCFS2_MOUNT_INODE64)
1478 seq_printf(s, ",inode64");
1479
a68979b8
TY		 1480#ifdef CONFIG_OCFS2_FS_POSIX_ACL
1481 if (opts & OCFS2_MOUNT_POSIX_ACL)
1482 seq_printf(s, ",acl");
1483 else
1484 seq_printf(s, ",noacl");
1485#endif
1486
d550071c
SM		 1487 return 0;
1488}
1489
ccd979bd
MF		 1490static int __init ocfs2_init(void)
1491{
1492 int status;
1493
1494 mlog_entry_void();
1495
1496 ocfs2_print_version();
1497
ccd979bd
MF		 1498 status = init_ocfs2_uptodate_cache();
1499 if (status < 0) {
1500 mlog_errno(status);
1501 goto leave;
1502 }
1503
1504 status = ocfs2_initialize_mem_caches();
1505 if (status < 0) {
1506 mlog_errno(status);
1507 goto leave;
1508 }
1509
1510 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1511 if (!ocfs2_wq) {
1512 status = -ENOMEM;
1513 goto leave;
1514 }
1515
ccd979bd
MF		 1516 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1517 if (!ocfs2_debugfs_root) {
1518 status = -EFAULT;
1519 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1520 }
1521
171bf93c
MF		 1522 status = ocfs2_quota_setup();
1523 if (status)
1524 goto leave;
1525
63e0c48a
JB		 1526 ocfs2_set_locking_protocol();
1527
9e33d69f 1528 status = register_quota_format(&ocfs2_quota_format);
ccd979bd
MF		 1529leave:
1530 if (status < 0) {
171bf93c 1531 ocfs2_quota_shutdown();
ccd979bd
MF		 1532 ocfs2_free_mem_caches();
1533 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1534 }
1535
1536 mlog_exit(status);
1537
1538 if (status >= 0) {
1539 return register_filesystem(&ocfs2_fs_type);
1540 } else
1541 return -1;
1542}
1543
1544static void __exit ocfs2_exit(void)
1545{
1546 mlog_entry_void();
1547
171bf93c
MF		 1548 ocfs2_quota_shutdown();
1549
ccd979bd
MF		 1550 if (ocfs2_wq) {
1551 flush_workqueue(ocfs2_wq);
1552 destroy_workqueue(ocfs2_wq);
1553 }
1554
9e33d69f
JK		 1555 unregister_quota_format(&ocfs2_quota_format);
1556
ccd979bd
MF		 1557 debugfs_remove(ocfs2_debugfs_root);
1558
1559 ocfs2_free_mem_caches();
1560
1561 unregister_filesystem(&ocfs2_fs_type);
1562
ccd979bd
MF		 1563 exit_ocfs2_uptodate_cache();
1564
1565 mlog_exit_void();
1566}
1567
1568static void ocfs2_put_super(struct super_block *sb)
1569{
1570 mlog_entry("(0x%p)\n", sb);
1571
6cfd0148
CH		 1572 lock_kernel();
1573
ccd979bd
MF		 1574 ocfs2_sync_blockdev(sb);
1575 ocfs2_dismount_volume(sb, 0);
1576
6cfd0148
CH		 1577 unlock_kernel();
1578
ccd979bd
MF		 1579 mlog_exit_void();
1580}
1581
726c3342 1582static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1583{
1584 struct ocfs2_super *osb;
1585 u32 numbits, freebits;
1586 int status;
1587 struct ocfs2_dinode *bm_lock;
1588 struct buffer_head *bh = NULL;
1589 struct inode *inode = NULL;
1590
726c3342 1591 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
ccd979bd 1592
726c3342 1593 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1594
1595 inode = ocfs2_get_system_file_inode(osb,
1596 GLOBAL_BITMAP_SYSTEM_INODE,
1597 OCFS2_INVALID_SLOT);
1598 if (!inode) {
1599 mlog(ML_ERROR, "failed to get bitmap inode\n");
1600 status = -EIO;
1601 goto bail;
1602 }
1603
e63aecb6 1604 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1605 if (status < 0) {
1606 mlog_errno(status);
1607 goto bail;
1608 }
1609
1610 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1611
1612 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1613 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1614
1615 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1616 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1617 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1618 buf->f_blocks = ((sector_t) numbits) *
1619 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1620 buf->f_bfree = ((sector_t) freebits) *
1621 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1622 buf->f_bavail = buf->f_bfree;
1623 buf->f_files = numbits;
1624 buf->f_ffree = freebits;
1625
1626 brelse(bh);
1627
e63aecb6 1628 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1629 status = 0;
1630bail:
1631 if (inode)
1632 iput(inode);
1633
1634 mlog_exit(status);
1635
1636 return status;
1637}
1638
51cc5068 1639static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1640{
1641 struct ocfs2_inode_info *oi = data;
1642
a35afb83
CL		 1643 oi->ip_flags = 0;
1644 oi->ip_open_count = 0;
1645 spin_lock_init(&oi->ip_lock);
1646 ocfs2_extent_map_init(&oi->vfs_inode);
1647 INIT_LIST_HEAD(&oi->ip_io_markers);
1648 oi->ip_created_trans = 0;
1649 oi->ip_last_trans = 0;
1650 oi->ip_dir_start_lookup = 0;
ccd979bd 1651
a35afb83 1652 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1653 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1654 mutex_init(&oi->ip_io_mutex);
ccd979bd 1655
a35afb83
CL		 1656 oi->ip_blkno = 0ULL;
1657 oi->ip_clusters = 0;
ccd979bd 1658
a35afb83 1659 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1660 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1661 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1662
a35afb83 1663 ocfs2_metadata_cache_init(&oi->vfs_inode);
ccd979bd 1664
a35afb83 1665 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1666}
1667
1668static int ocfs2_initialize_mem_caches(void)
1669{
1670 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1671 sizeof(struct ocfs2_inode_info),
1672 0,
1673 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1674 SLAB_MEM_SPREAD),
20c2df83 1675 ocfs2_inode_init_once);
9e33d69f
JK		 1676 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1677 sizeof(struct ocfs2_dquot),
1678 0,
1679 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1680 SLAB_MEM_SPREAD),
1681 NULL);
1682 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1683 sizeof(struct ocfs2_quota_chunk),
1684 0,
1685 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1686 NULL);
1687 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1688 !ocfs2_qf_chunk_cachep) {
1689 if (ocfs2_inode_cachep)
1690 kmem_cache_destroy(ocfs2_inode_cachep);
1691 if (ocfs2_dquot_cachep)
1692 kmem_cache_destroy(ocfs2_dquot_cachep);
1693 if (ocfs2_qf_chunk_cachep)
1694 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1695 return -ENOMEM;
9e33d69f 1696 }
ccd979bd 1697
ccd979bd
MF		 1698 return 0;
1699}
1700
1701static void ocfs2_free_mem_caches(void)
1702{
1703 if (ocfs2_inode_cachep)
1704 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1705 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1706
1707 if (ocfs2_dquot_cachep)
1708 kmem_cache_destroy(ocfs2_dquot_cachep);
1709 ocfs2_dquot_cachep = NULL;
1710
1711 if (ocfs2_qf_chunk_cachep)
1712 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1713 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1714}
1715
1716static int ocfs2_get_sector(struct super_block *sb,
1717 struct buffer_head **bh,
1718 int block,
1719 int sect_size)
1720{
1721 if (!sb_set_blocksize(sb, sect_size)) {
1722 mlog(ML_ERROR, "unable to set blocksize\n");
1723 return -EIO;
1724 }
1725
1726 *bh = sb_getblk(sb, block);
1727 if (!*bh) {
1728 mlog_errno(-EIO);
1729 return -EIO;
1730 }
1731 lock_buffer(*bh);
1732 if (!buffer_dirty(*bh))
1733 clear_buffer_uptodate(*bh);
1734 unlock_buffer(*bh);
1735 ll_rw_block(READ, 1, bh);
1736 wait_on_buffer(*bh);
28d57d43 1737 if (!buffer_uptodate(*bh)) {
1738 mlog_errno(-EIO);
1739 brelse(*bh);
1740 *bh = NULL;
1741 return -EIO;
1742 }
1743
ccd979bd
MF		 1744 return 0;
1745}
1746
ccd979bd
MF		 1747static int ocfs2_mount_volume(struct super_block *sb)
1748{
1749 int status = 0;
1750 int unlock_super = 0;
1751 struct ocfs2_super *osb = OCFS2_SB(sb);
1752
1753 mlog_entry_void();
1754
1755 if (ocfs2_is_hard_readonly(osb))
1756 goto leave;
1757
ccd979bd
MF		 1758 status = ocfs2_dlm_init(osb);
1759 if (status < 0) {
1760 mlog_errno(status);
1761 goto leave;
1762 }
1763
ccd979bd
MF		 1764 status = ocfs2_super_lock(osb, 1);
1765 if (status < 0) {
1766 mlog_errno(status);
1767 goto leave;
1768 }
1769 unlock_super = 1;
1770
1771 /* This will load up the node map and add ourselves to it. */
1772 status = ocfs2_find_slot(osb);
1773 if (status < 0) {
1774 mlog_errno(status);
1775 goto leave;
1776 }
1777
ccd979bd
MF		 1778 /* load all node-local system inodes */
1779 status = ocfs2_init_local_system_inodes(osb);
1780 if (status < 0) {
1781 mlog_errno(status);
1782 goto leave;
1783 }
1784
1785 status = ocfs2_check_volume(osb);
1786 if (status < 0) {
1787 mlog_errno(status);
1788 goto leave;
1789 }
1790
1791 status = ocfs2_truncate_log_init(osb);
06c59bb8 1792 if (status < 0)
ccd979bd 1793 mlog_errno(status);
c271c5c2 1794
ccd979bd
MF		 1795leave:
1796 if (unlock_super)
1797 ocfs2_super_unlock(osb, 1);
1798
1799 mlog_exit(status);
1800 return status;
1801}
1802
ccd979bd
MF		 1803static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1804{
286eaa95 1805 int tmp, hangup_needed = 0;
ccd979bd 1806 struct ocfs2_super *osb = NULL;
c271c5c2 1807 char nodestr[8];
ccd979bd
MF		 1808
1809 mlog_entry("(0x%p)\n", sb);
1810
1811 BUG_ON(!sb);
1812 osb = OCFS2_SB(sb);
1813 BUG_ON(!osb);
1814
50397507
SM		 1815 debugfs_remove(osb->osb_ctxt);
1816
19ece546
JK		 1817 ocfs2_disable_quotas(osb);
1818
ccd979bd
MF		 1819 ocfs2_shutdown_local_alloc(osb);
1820
1821 ocfs2_truncate_log_shutdown(osb);
1822
83273932
SE		 1823 ocfs2_orphan_scan_stop(osb);
1824
553abd04
JB		 1825 /* This will disable recovery and flush any recovery work. */
1826 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1827
1828 ocfs2_journal_shutdown(osb);
1829
1830 ocfs2_sync_blockdev(sb);
1831
4670c46d
JB		 1832 /* No cluster connection means we've failed during mount, so skip
1833 * all the steps which depended on that to complete. */
1834 if (osb->cconn) {
ccd979bd
MF		 1835 tmp = ocfs2_super_lock(osb, 1);
1836 if (tmp < 0) {
1837 mlog_errno(tmp);
1838 return;
1839 }
19b613d4 1840 }
ccd979bd 1841
19b613d4
MF		 1842 if (osb->slot_num != OCFS2_INVALID_SLOT)
1843 ocfs2_put_slot(osb);
ccd979bd 1844
4670c46d 1845 if (osb->cconn)
ccd979bd 1846 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1847
1848 ocfs2_release_system_inodes(osb);
1849
6953b4c0 1850 /*
6953b4c0
JB		 1851 * If we're dismounting due to mount error, mount.ocfs2 will clean
1852 * up heartbeat. If we're a local mount, there is no heartbeat.
1853 * If we failed before we got a uuid_str yet, we can't stop
1854 * heartbeat. Otherwise, do it.
1855 */
1856 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
286eaa95
JB		 1857 hangup_needed = 1;
1858
1859 if (osb->cconn)
1860 ocfs2_dlm_shutdown(osb, hangup_needed);
1861
73be192b 1862 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1863 debugfs_remove(osb->osb_debug_root);
1864
1865 if (hangup_needed)
6953b4c0 1866 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1867
1868 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1869
c271c5c2
SM		 1870 if (ocfs2_mount_local(osb))
1871 snprintf(nodestr, sizeof(nodestr), "local");
1872 else
19fdb624 1873 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1874
1875 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1876 osb->dev_str, nodestr);
ccd979bd
MF		 1877
1878 ocfs2_delete_osb(osb);
1879 kfree(osb);
1880 sb->s_dev = 0;
1881 sb->s_fs_info = NULL;
1882}
1883
1884static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1885 unsigned uuid_bytes)
1886{
1887 int i, ret;
1888 char *ptr;
1889
1890 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1891
cd861280 1892 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 1893 if (osb->uuid_str == NULL)
1894 return -ENOMEM;
1895
ccd979bd
MF		 1896 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1897 /* print with null */
1898 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1899 if (ret != 2) /* drop super cleans up */
1900 return -EINVAL;
1901 /* then only advance past the last char */
1902 ptr += 2;
1903 }
1904
1905 return 0;
1906}
1907
1908static int ocfs2_initialize_super(struct super_block *sb,
1909 struct buffer_head *bh,
73be192b
JB		 1910 int sector_size,
1911 struct ocfs2_blockcheck_stats *stats)
ccd979bd 1912{
bddb8eb3 1913 int status;
5a254031
MF		 1914 int i, cbits, bbits;
1915 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 1916 struct inode *inode = NULL;
ccd979bd
MF		 1917 struct ocfs2_journal *journal;
1918 __le32 uuid_net_key;
1919 struct ocfs2_super *osb;
1920
1921 mlog_entry_void();
1922
cd861280 1923 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 1924 if (!osb) {
1925 status = -ENOMEM;
1926 mlog_errno(status);
1927 goto bail;
1928 }
1929
1930 sb->s_fs_info = osb;
1931 sb->s_op = &ocfs2_sops;
1932 sb->s_export_op = &ocfs2_export_ops;
19ece546
JK		 1933 sb->s_qcop = &ocfs2_quotactl_ops;
1934 sb->dq_op = &ocfs2_quota_operations;
cf1d6c76 1935 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 1936 sb->s_time_gran = 1;
ccd979bd
MF		 1937 sb->s_flags |= MS_NOATIME;
1938 /* this is needed to support O_LARGEFILE */
5a254031
MF		 1939 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1940 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1941 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd 1942
9b7895ef
MF		 1943 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1944
1945 for (i = 0; i < 3; i++)
1946 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1947 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1948
ccd979bd
MF		 1949 osb->sb = sb;
1950 /* Save off for ocfs2_rw_direct */
1951 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 1952 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 1953
34d024f8
MF		 1954 spin_lock_init(&osb->dc_task_lock);
1955 init_waitqueue_head(&osb->dc_event);
1956 osb->dc_work_sequence = 0;
1957 osb->dc_wake_sequence = 0;
ccd979bd
MF		 1958 INIT_LIST_HEAD(&osb->blocked_lock_list);
1959 osb->blocked_lock_count = 0;
ccd979bd 1960 spin_lock_init(&osb->osb_lock);
c8b9cf9a 1961 spin_lock_init(&osb->osb_xattr_lock);
4d0ddb2c 1962 ocfs2_init_inode_steal_slot(osb);
ccd979bd
MF		 1963
1964 atomic_set(&osb->alloc_stats.moves, 0);
1965 atomic_set(&osb->alloc_stats.local_data, 0);
1966 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1967 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1968 atomic_set(&osb->alloc_stats.bg_extends, 0);
1969
73be192b
JB		 1970 /* Copy the blockcheck stats from the superblock probe */
1971 osb->osb_ecc_stats = *stats;
1972
ccd979bd
MF		 1973 ocfs2_init_node_maps(osb);
1974
1975 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1976 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1977
553abd04
JB		 1978 status = ocfs2_recovery_init(osb);
1979 if (status) {
1980 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1981 mlog_errno(status);
1982 goto bail;
1983 }
ccd979bd 1984
83273932
SE		 1985 status = ocfs2_orphan_scan_init(osb);
1986 if (status) {
1987 mlog(ML_ERROR, "Unable to initialize delayed orphan scan\n");
1988 mlog_errno(status);
1989 goto bail;
1990 }
1991
ccd979bd
MF		 1992 init_waitqueue_head(&osb->checkpoint_event);
1993 atomic_set(&osb->needs_checkpoint, 0);
1994
7f1a37e3
TY		 1995 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1996
ccd979bd
MF		 1997 osb->slot_num = OCFS2_INVALID_SLOT;
1998
8154da3d
TY		 1999 osb->s_xattr_inline_size = le16_to_cpu(
2000 di->id2.i_super.s_xattr_inline_size);
fdd77704 2001
ccd979bd
MF		 2002 osb->local_alloc_state = OCFS2_LA_UNUSED;
2003 osb->local_alloc_bh = NULL;
9c7af40b 2004 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2005
ccd979bd
MF		 2006 init_waitqueue_head(&osb->osb_mount_event);
2007
2008 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2009 if (!osb->vol_label) {
2010 mlog(ML_ERROR, "unable to alloc vol label\n");
2011 status = -ENOMEM;
2012 goto bail;
2013 }
2014
ccd979bd
MF		 2015 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2016 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2017 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2018 osb->max_slots);
2019 status = -EINVAL;
2020 goto bail;
2021 }
781ee3e2 2022 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
ccd979bd 2023
539d8264
SM		 2024 osb->slot_recovery_generations =
2025 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2026 GFP_KERNEL);
2027 if (!osb->slot_recovery_generations) {
2028 status = -ENOMEM;
2029 mlog_errno(status);
2030 goto bail;
2031 }
2032
b4df6ed8
MF		 2033 init_waitqueue_head(&osb->osb_wipe_event);
2034 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2035 sizeof(*osb->osb_orphan_wipes),
2036 GFP_KERNEL);
2037 if (!osb->osb_orphan_wipes) {
2038 status = -ENOMEM;
2039 mlog_errno(status);
2040 goto bail;
2041 }
2042
ccd979bd
MF		 2043 osb->s_feature_compat =
2044 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2045 osb->s_feature_ro_compat =
2046 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2047 osb->s_feature_incompat =
2048 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2049
2050 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2051 mlog(ML_ERROR, "couldn't mount because of unsupported "
2052 "optional features (%x).\n", i);
2053 status = -EINVAL;
2054 goto bail;
2055 }
2056 if (!(osb->sb->s_flags & MS_RDONLY) &&
2057 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2058 mlog(ML_ERROR, "couldn't mount RDWR because of "
2059 "unsupported optional features (%x).\n", i);
2060 status = -EINVAL;
2061 goto bail;
2062 }
2063
b61817e1
JB		 2064 if (ocfs2_userspace_stack(osb)) {
2065 memcpy(osb->osb_cluster_stack,
2066 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2067 OCFS2_STACK_LABEL_LEN);
2068 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2069 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2070 mlog(ML_ERROR,
2071 "couldn't mount because of an invalid "
2072 "cluster stack label (%s) \n",
2073 osb->osb_cluster_stack);
2074 status = -EINVAL;
2075 goto bail;
2076 }
2077 } else {
2078 /* The empty string is identical with classic tools that
2079 * don't know about s_cluster_info. */
2080 osb->osb_cluster_stack[0] = '\0';
2081 }
2082
ccd979bd
MF		 2083 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2084
2085 /* FIXME
2086 * This should be done in ocfs2_journal_init(), but unknown
2087 * ordering issues will cause the filesystem to crash.
2088 * If anyone wants to figure out what part of the code
2089 * refers to osb->journal before ocfs2_journal_init() is run,
2090 * be my guest.
2091 */
2092 /* initialize our journal structure */
2093
cd861280 2094 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2095 if (!journal) {
2096 mlog(ML_ERROR, "unable to alloc journal\n");
2097 status = -ENOMEM;
2098 goto bail;
2099 }
2100 osb->journal = journal;
2101 journal->j_osb = osb;
2102
2103 atomic_set(&journal->j_num_trans, 0);
2104 init_rwsem(&journal->j_trans_barrier);
2105 init_waitqueue_head(&journal->j_checkpointed);
2106 spin_lock_init(&journal->j_lock);
2107 journal->j_trans_id = (unsigned long) 1;
2108 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2109 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2110 journal->j_state = OCFS2_JOURNAL_FREE;
2111
ea455f8a
JK		 2112 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2113 osb->dentry_lock_list = NULL;
2114
ccd979bd
MF		 2115 /* get some pseudo constants for clustersize bits */
2116 osb->s_clustersize_bits =
2117 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2118 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2119 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2120
2121 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2122 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2123 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2124 osb->s_clustersize);
2125 status = -EINVAL;
2126 goto bail;
2127 }
2128
2129 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2130 > (u32)~0UL) {
2131 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2132 "what jbd can address in 32 bits.\n");
2133 status = -EINVAL;
2134 goto bail;
2135 }
2136
2137 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2138 sizeof(di->id2.i_super.s_uuid))) {
2139 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2140 status = -ENOMEM;
2141 goto bail;
2142 }
2143
78427043 2144 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
ccd979bd
MF		 2145
2146 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2147 osb->vol_label[63] = '\0';
2148 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2149 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2150 osb->first_cluster_group_blkno =
2151 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2152 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2153 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
ccd979bd
MF		 2154 mlog(0, "vol_label: %s\n", osb->vol_label);
2155 mlog(0, "uuid: %s\n", osb->uuid_str);
b0697053
MF		 2156 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2157 (unsigned long long)osb->root_blkno,
2158 (unsigned long long)osb->system_dir_blkno);
ccd979bd
MF		 2159
2160 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2161 if (!osb->osb_dlm_debug) {
2162 status = -ENOMEM;
2163 mlog_errno(status);
2164 goto bail;
2165 }
2166
2167 atomic_set(&osb->vol_state, VOLUME_INIT);
2168
2169 /* load root, system_dir, and all global system inodes */
2170 status = ocfs2_init_global_system_inodes(osb);
2171 if (status < 0) {
2172 mlog_errno(status);
2173 goto bail;
2174 }
2175
2176 /*
2177 * global bitmap
2178 */
2179 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2180 OCFS2_INVALID_SLOT);
2181 if (!inode) {
2182 status = -EINVAL;
2183 mlog_errno(status);
2184 goto bail;
2185 }
2186
2187 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
ccd979bd 2188 iput(inode);
ccd979bd 2189
e9d578a8 2190 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
ccd979bd
MF		 2191
2192 status = ocfs2_init_slot_info(osb);
2193 if (status < 0) {
2194 mlog_errno(status);
2195 goto bail;
2196 }
2197
ccd979bd
MF		 2198bail:
2199 mlog_exit(status);
2200 return status;
2201}
2202
2203/*
2204 * will return: -EAGAIN if it is ok to keep searching for superblocks
2205 * -EINVAL if there is a bad superblock
2206 * 0 on success
2207 */
2208static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2209 struct buffer_head *bh,
73be192b
JB		 2210 u32 blksz,
2211 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2212{
2213 int status = -EAGAIN;
2214
2215 mlog_entry_void();
2216
2217 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2218 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2219 /* We have to do a raw check of the feature here */
2220 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2221 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2222 status = ocfs2_block_check_validate(bh->b_data,
2223 bh->b_size,
73be192b
JB		 2224 &di->i_check,
2225 stats);
d030cc97
JB		 2226 if (status)
2227 goto out;
2228 }
ccd979bd
MF		 2229 status = -EINVAL;
2230 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2231 mlog(ML_ERROR, "found superblock with incorrect block "
2232 "size: found %u, should be %u\n",
2233 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2234 blksz);
2235 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2236 OCFS2_MAJOR_REV_LEVEL ||
2237 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2238 OCFS2_MINOR_REV_LEVEL) {
2239 mlog(ML_ERROR, "found superblock with bad version: "
2240 "found %u.%u, should be %u.%u\n",
2241 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2242 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2243 OCFS2_MAJOR_REV_LEVEL,
2244 OCFS2_MINOR_REV_LEVEL);
2245 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2246 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2247 "found %llu, should be %llu\n",
1ca1a111 2248 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2249 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2250 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2251 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2252 mlog(ML_ERROR, "bad cluster size found: %u\n",
2253 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2254 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2255 mlog(ML_ERROR, "bad root_blkno: 0\n");
2256 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2257 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2258 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2259 mlog(ML_ERROR,
2260 "Superblock slots found greater than file system "
2261 "maximum: found %u, max %u\n",
2262 le16_to_cpu(di->id2.i_super.s_max_slots),
2263 OCFS2_MAX_SLOTS);
2264 } else {
2265 /* found it! */
2266 status = 0;
2267 }
2268 }
2269
d030cc97 2270out:
ccd979bd
MF		 2271 mlog_exit(status);
2272 return status;
2273}
2274
2275static int ocfs2_check_volume(struct ocfs2_super *osb)
2276{
bddb8eb3 2277 int status;
ccd979bd 2278 int dirty;
c271c5c2 2279 int local;
ccd979bd
MF		 2280 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2281 * recover
2282 * ourselves. */
2283
2284 mlog_entry_void();
2285
2286 /* Init our journal object. */
2287 status = ocfs2_journal_init(osb->journal, &dirty);
2288 if (status < 0) {
2289 mlog(ML_ERROR, "Could not initialize journal!\n");
2290 goto finally;
2291 }
2292
2293 /* If the journal was unmounted cleanly then we don't want to
2294 * recover anything. Otherwise, journal_load will do that
2295 * dirty work for us :) */
2296 if (!dirty) {
2297 status = ocfs2_journal_wipe(osb->journal, 0);
2298 if (status < 0) {
2299 mlog_errno(status);
2300 goto finally;
2301 }
2302 } else {
2303 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2304 "recovering volume.\n");
2305 }
2306
c271c5c2
SM		 2307 local = ocfs2_mount_local(osb);
2308
ccd979bd 2309 /* will play back anything left in the journal. */
539d8264 2310 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2311 if (status < 0) {
2312 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2313 goto finally;
2314 }
ccd979bd
MF		 2315
2316 if (dirty) {
2317 /* recover my local alloc if we didn't unmount cleanly. */
2318 status = ocfs2_begin_local_alloc_recovery(osb,
2319 osb->slot_num,
2320 &local_alloc);
2321 if (status < 0) {
2322 mlog_errno(status);
2323 goto finally;
2324 }
2325 /* we complete the recovery process after we've marked
2326 * ourselves as mounted. */
2327 }
2328
2329 mlog(0, "Journal loaded.\n");
2330
2331 status = ocfs2_load_local_alloc(osb);
2332 if (status < 0) {
2333 mlog_errno(status);
2334 goto finally;
2335 }
2336
2337 if (dirty) {
2338 /* Recovery will be completed after we've mounted the
2339 * rest of the volume. */
2340 osb->dirty = 1;
2341 osb->local_alloc_copy = local_alloc;
2342 local_alloc = NULL;
2343 }
2344
2345 /* go through each journal, trylock it and if you get the
2346 * lock, and it's marked as dirty, set the bit in the recover
2347 * map and launch a recovery thread for it. */
2348 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2349 if (status < 0) {
2350 mlog_errno(status);
2351 goto finally;
2352 }
2353
2354 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2355 if (status < 0)
2356 mlog_errno(status);
2357
2358finally:
2359 if (local_alloc)
2360 kfree(local_alloc);
2361
2362 mlog_exit(status);
2363 return status;
2364}
2365
2366/*
2367 * The routine gets called from dismount or close whenever a dismount on
2368 * volume is requested and the osb open count becomes 1.
2369 * It will remove the osb from the global list and also free up all the
2370 * initialized resources and fileobject.
2371 */
2372static void ocfs2_delete_osb(struct ocfs2_super *osb)
2373{
2374 mlog_entry_void();
2375
2376 /* This function assumes that the caller has the main osb resource */
2377
8e8a4603 2378 ocfs2_free_slot_info(osb);
ccd979bd 2379
b4df6ed8 2380 kfree(osb->osb_orphan_wipes);
539d8264 2381 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2382 /* FIXME
2383 * This belongs in journal shutdown, but because we have to
2384 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2385 * we free it here.
2386 */
2387 kfree(osb->journal);
2388 if (osb->local_alloc_copy)
2389 kfree(osb->local_alloc_copy);
2390 kfree(osb->uuid_str);
2391 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2392 memset(osb, 0, sizeof(struct ocfs2_super));
2393
2394 mlog_exit_void();
2395}
2396
2397/* Put OCFS2 into a readonly state, or (if the user specifies it),
2398 * panic(). We do not support continue-on-error operation. */
2399static void ocfs2_handle_error(struct super_block *sb)
2400{
2401 struct ocfs2_super *osb = OCFS2_SB(sb);
2402
2403 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2404 panic("OCFS2: (device %s): panic forced after error\n",
2405 sb->s_id);
2406
2407 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2408
2409 if (sb->s_flags & MS_RDONLY &&
2410 (ocfs2_is_soft_readonly(osb) ||
2411 ocfs2_is_hard_readonly(osb)))
2412 return;
2413
2414 printk(KERN_CRIT "File system is now read-only due to the potential "
2415 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2416 "system is unmounted.\n");
2417 sb->s_flags |= MS_RDONLY;
2418 ocfs2_set_ro_flag(osb, 0);
2419}
2420
2421static char error_buf[1024];
2422
2423void __ocfs2_error(struct super_block *sb,
2424 const char *function,
2425 const char *fmt, ...)
2426{
2427 va_list args;
2428
2429 va_start(args, fmt);
4a6e617a 2430 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2431 va_end(args);
2432
2433 /* Not using mlog here because we want to show the actual
2434 * function the error came from. */
2435 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2436 sb->s_id, function, error_buf);
2437
2438 ocfs2_handle_error(sb);
2439}
2440
2441/* Handle critical errors. This is intentionally more drastic than
2442 * ocfs2_handle_error, so we only use for things like journal errors,
2443 * etc. */
2444void __ocfs2_abort(struct super_block* sb,
2445 const char *function,
2446 const char *fmt, ...)
2447{
2448 va_list args;
2449
2450 va_start(args, fmt);
4a6e617a 2451 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2452 va_end(args);
2453
2454 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2455 sb->s_id, function, error_buf);
2456
2457 /* We don't have the cluster support yet to go straight to
2458 * hard readonly in here. Until then, we want to keep
2459 * ocfs2_abort() so that we can at least mark critical
2460 * errors.
2461 *
2462 * TODO: This should abort the journal and alert other nodes
2463 * that our slot needs recovery. */
2464
2465 /* Force a panic(). This stinks, but it's better than letting
2466 * things continue without having a proper hard readonly
2467 * here. */
2468 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2469 ocfs2_handle_error(sb);
2470}
2471
2472module_init(ocfs2_init);
2473module_exit(ocfs2_exit);
Linux 6.x block layer and io_uring tree(s)