projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: clean up localalloc mount option size parsing
[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>
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>
337eb00a 44#include <linux/smp_lock.h>
ccd979bd
MF		 45
46#define MLOG_MASK_PREFIX ML_SUPER
47#include <cluster/masklog.h>
48
49#include "ocfs2.h"
50
51/* this should be the only file to include a version 1 header */
52#include "ocfs1_fs_compat.h"
53
54#include "alloc.h"
d030cc97 55#include "blockcheck.h"
ccd979bd
MF		 56#include "dlmglue.h"
57#include "export.h"
58#include "extent_map.h"
59#include "heartbeat.h"
60#include "inode.h"
61#include "journal.h"
62#include "localalloc.h"
63#include "namei.h"
64#include "slot_map.h"
65#include "super.h"
66#include "sysfile.h"
67#include "uptodate.h"
68#include "ver.h"
cf1d6c76 69#include "xattr.h"
9e33d69f 70#include "quota.h"
374a263e 71#include "refcounttree.h"
b89c5428 72#include "suballoc.h"
ccd979bd
MF		 73
74#include "buffer_head_io.h"
75
e18b890b 76static struct kmem_cache *ocfs2_inode_cachep = NULL;
9e33d69f
JK		 77struct kmem_cache *ocfs2_dquot_cachep;
78struct kmem_cache *ocfs2_qf_chunk_cachep;
ccd979bd 79
ccd979bd
MF		 80/* OCFS2 needs to schedule several differnt types of work which
81 * require cluster locking, disk I/O, recovery waits, etc. Since these
82 * types of work tend to be heavy we avoid using the kernel events
83 * workqueue and schedule on our own. */
84struct workqueue_struct *ocfs2_wq = NULL;
85
86static struct dentry *ocfs2_debugfs_root = NULL;
87
88MODULE_AUTHOR("Oracle");
89MODULE_LICENSE("GPL");
90
c0123ade
TY		 91struct mount_options
92{
d147b3d6 93 unsigned long commit_interval;
c0123ade
TY		 94 unsigned long mount_opt;
95 unsigned int atime_quantum;
96 signed short slot;
73c8a800 97 int localalloc_opt;
d02f00cc 98 unsigned int resv_level;
b61817e1 99 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
c0123ade
TY		 100};
101
ccd979bd 102static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 103 struct mount_options *mopt,
baf4661a 104 int is_remount);
5297aad8
JK		 105static int ocfs2_check_set_options(struct super_block *sb,
106 struct mount_options *options);
d550071c 107static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
ccd979bd
MF		 108static void ocfs2_put_super(struct super_block *sb);
109static int ocfs2_mount_volume(struct super_block *sb);
110static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
111static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
112static int ocfs2_initialize_mem_caches(void);
113static void ocfs2_free_mem_caches(void);
114static void ocfs2_delete_osb(struct ocfs2_super *osb);
115
726c3342 116static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 117
118static int ocfs2_sync_fs(struct super_block *sb, int wait);
119
120static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
121static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 122static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 123static int ocfs2_check_volume(struct ocfs2_super *osb);
124static int ocfs2_verify_volume(struct ocfs2_dinode *di,
125 struct buffer_head *bh,
73be192b
JB		 126 u32 sectsize,
127 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 128static int ocfs2_initialize_super(struct super_block *sb,
129 struct buffer_head *bh,
73be192b
JB		 130 int sector_size,
131 struct ocfs2_blockcheck_stats *stats);
ccd979bd
MF		 132static int ocfs2_get_sector(struct super_block *sb,
133 struct buffer_head **bh,
134 int block,
135 int sect_size);
ccd979bd
MF		 136static struct inode *ocfs2_alloc_inode(struct super_block *sb);
137static void ocfs2_destroy_inode(struct inode *inode);
19ece546
JK		 138static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
139static int ocfs2_enable_quotas(struct ocfs2_super *osb);
140static void ocfs2_disable_quotas(struct ocfs2_super *osb);
ccd979bd 141
ee9b6d61 142static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 143 .statfs = ocfs2_statfs,
144 .alloc_inode = ocfs2_alloc_inode,
145 .destroy_inode = ocfs2_destroy_inode,
146 .drop_inode = ocfs2_drop_inode,
147 .clear_inode = ocfs2_clear_inode,
148 .delete_inode = ocfs2_delete_inode,
149 .sync_fs = ocfs2_sync_fs,
ccd979bd
MF		 150 .put_super = ocfs2_put_super,
151 .remount_fs = ocfs2_remount,
d550071c 152 .show_options = ocfs2_show_options,
9e33d69f
JK		 153 .quota_read = ocfs2_quota_read,
154 .quota_write = ocfs2_quota_write,
ccd979bd
MF		 155};
156
157enum {
158 Opt_barrier,
159 Opt_err_panic,
160 Opt_err_ro,
161 Opt_intr,
162 Opt_nointr,
163 Opt_hb_none,
164 Opt_hb_local,
165 Opt_data_ordered,
166 Opt_data_writeback,
7f1a37e3 167 Opt_atime_quantum,
baf4661a 168 Opt_slot,
d147b3d6 169 Opt_commit,
2fbe8d1e 170 Opt_localalloc,
53fc622b 171 Opt_localflocks,
b61817e1 172 Opt_stack,
cf1d6c76
TY		 173 Opt_user_xattr,
174 Opt_nouser_xattr,
12462f1d 175 Opt_inode64,
a68979b8
TY		 176 Opt_acl,
177 Opt_noacl,
19ece546
JK		 178 Opt_usrquota,
179 Opt_grpquota,
d02f00cc 180 Opt_resv_level,
ccd979bd
MF		 181 Opt_err,
182};
183
a447c093 184static const match_table_t tokens = {
ccd979bd
MF		 185 {Opt_barrier, "barrier=%u"},
186 {Opt_err_panic, "errors=panic"},
187 {Opt_err_ro, "errors=remount-ro"},
188 {Opt_intr, "intr"},
189 {Opt_nointr, "nointr"},
190 {Opt_hb_none, OCFS2_HB_NONE},
191 {Opt_hb_local, OCFS2_HB_LOCAL},
192 {Opt_data_ordered, "data=ordered"},
193 {Opt_data_writeback, "data=writeback"},
7f1a37e3 194 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 195 {Opt_slot, "preferred_slot=%u"},
d147b3d6 196 {Opt_commit, "commit=%u"},
2fbe8d1e 197 {Opt_localalloc, "localalloc=%d"},
53fc622b 198 {Opt_localflocks, "localflocks"},
b61817e1 199 {Opt_stack, "cluster_stack=%s"},
cf1d6c76
TY		 200 {Opt_user_xattr, "user_xattr"},
201 {Opt_nouser_xattr, "nouser_xattr"},
12462f1d 202 {Opt_inode64, "inode64"},
a68979b8
TY		 203 {Opt_acl, "acl"},
204 {Opt_noacl, "noacl"},
19ece546
JK		 205 {Opt_usrquota, "usrquota"},
206 {Opt_grpquota, "grpquota"},
d02f00cc 207 {Opt_resv_level, "resv_level=%u"},
ccd979bd
MF		 208 {Opt_err, NULL}
209};
210
50397507
SM		 211#ifdef CONFIG_DEBUG_FS
212static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
213{
50397507
SM		 214 struct ocfs2_cluster_connection *cconn = osb->cconn;
215 struct ocfs2_recovery_map *rm = osb->recovery_map;
df152c24
SM		 216 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
217 int i, out = 0;
50397507
SM		 218
219 out += snprintf(buf + out, len - out,
220 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
221 "Device", osb->dev_str, osb->uuid_str,
222 osb->fs_generation, osb->vol_label);
223
224 out += snprintf(buf + out, len - out,
225 "%10s => State: %d Flags: 0x%lX\n", "Volume",
226 atomic_read(&osb->vol_state), osb->osb_flags);
227
228 out += snprintf(buf + out, len - out,
229 "%10s => Block: %lu Cluster: %d\n", "Sizes",
230 osb->sb->s_blocksize, osb->s_clustersize);
231
232 out += snprintf(buf + out, len - out,
233 "%10s => Compat: 0x%X Incompat: 0x%X "
234 "ROcompat: 0x%X\n",
235 "Features", osb->s_feature_compat,
236 osb->s_feature_incompat, osb->s_feature_ro_compat);
237
238 out += snprintf(buf + out, len - out,
239 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
240 osb->s_mount_opt, osb->s_atime_quantum);
241
c3d38840
SM		 242 if (cconn) {
243 out += snprintf(buf + out, len - out,
244 "%10s => Stack: %s Name: %*s "
245 "Version: %d.%d\n", "Cluster",
246 (*osb->osb_cluster_stack == '\0' ?
247 "o2cb" : osb->osb_cluster_stack),
248 cconn->cc_namelen, cconn->cc_name,
249 cconn->cc_version.pv_major,
250 cconn->cc_version.pv_minor);
251 }
50397507
SM		 252
253 spin_lock(&osb->dc_task_lock);
254 out += snprintf(buf + out, len - out,
255 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
256 "WorkSeq: %lu\n", "DownCnvt",
c3d38840
SM		 257 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
258 osb->blocked_lock_count, osb->dc_wake_sequence,
259 osb->dc_work_sequence);
50397507
SM		 260 spin_unlock(&osb->dc_task_lock);
261
262 spin_lock(&osb->osb_lock);
263 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
264 "Recovery",
265 (osb->recovery_thread_task ?
266 task_pid_nr(osb->recovery_thread_task) : -1));
267 if (rm->rm_used == 0)
268 out += snprintf(buf + out, len - out, " None\n");
269 else {
270 for (i = 0; i < rm->rm_used; i++)
271 out += snprintf(buf + out, len - out, " %d",
272 rm->rm_entries[i]);
273 out += snprintf(buf + out, len - out, "\n");
274 }
275 spin_unlock(&osb->osb_lock);
276
277 out += snprintf(buf + out, len - out,
278 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
c3d38840
SM		 279 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
280 osb->osb_commit_interval,
50397507
SM		 281 atomic_read(&osb->needs_checkpoint));
282
283 out += snprintf(buf + out, len - out,
c3d38840 284 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
50397507 285 "Journal", osb->journal->j_state,
c3d38840
SM		 286 osb->journal->j_trans_id,
287 atomic_read(&osb->journal->j_num_trans));
50397507
SM		 288
289 out += snprintf(buf + out, len - out,
290 "%10s => GlobalAllocs: %d LocalAllocs: %d "
291 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
292 "Stats",
293 atomic_read(&osb->alloc_stats.bitmap_data),
294 atomic_read(&osb->alloc_stats.local_data),
295 atomic_read(&osb->alloc_stats.bg_allocs),
296 atomic_read(&osb->alloc_stats.moves),
297 atomic_read(&osb->alloc_stats.bg_extends));
298
299 out += snprintf(buf + out, len - out,
300 "%10s => State: %u Descriptor: %llu Size: %u bits "
301 "Default: %u bits\n",
302 "LocalAlloc", osb->local_alloc_state,
303 (unsigned long long)osb->la_last_gd,
304 osb->local_alloc_bits, osb->local_alloc_default_bits);
305
306 spin_lock(&osb->osb_lock);
307 out += snprintf(buf + out, len - out,
b89c5428
TY		 308 "%10s => InodeSlot: %d StolenInodes: %d, "
309 "MetaSlot: %d StolenMeta: %d\n", "Steal",
50397507 310 osb->s_inode_steal_slot,
b89c5428
TY		 311 atomic_read(&osb->s_num_inodes_stolen),
312 osb->s_meta_steal_slot,
313 atomic_read(&osb->s_num_meta_stolen));
50397507
SM		 314 spin_unlock(&osb->osb_lock);
315
df152c24
SM		 316 out += snprintf(buf + out, len - out, "OrphanScan => ");
317 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
318 os->os_count, os->os_seqno);
319 out += snprintf(buf + out, len - out, " Last Scan: ");
320 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
321 out += snprintf(buf + out, len - out, "Disabled\n");
322 else
323 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
324 (get_seconds() - os->os_scantime.tv_sec));
325
50397507
SM		 326 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
327 "Slots", "Num", "RecoGen");
50397507
SM		 328 for (i = 0; i < osb->max_slots; ++i) {
329 out += snprintf(buf + out, len - out,
330 "%10s %c %3d %10d\n",
331 " ",
332 (i == osb->slot_num ? '*' : ' '),
333 i, osb->slot_recovery_generations[i]);
334 }
335
336 return out;
337}
338
339static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
340{
341 struct ocfs2_super *osb = inode->i_private;
342 char *buf = NULL;
343
344 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
345 if (!buf)
346 goto bail;
347
348 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
349
350 file->private_data = buf;
351
352 return 0;
353bail:
354 return -ENOMEM;
355}
356
357static int ocfs2_debug_release(struct inode *inode, struct file *file)
358{
359 kfree(file->private_data);
360 return 0;
361}
362
363static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
364 size_t nbytes, loff_t *ppos)
365{
366 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
367 i_size_read(file->f_mapping->host));
368}
369#else
370static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
371{
372 return 0;
373}
374static int ocfs2_debug_release(struct inode *inode, struct file *file)
375{
376 return 0;
377}
378static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
379 size_t nbytes, loff_t *ppos)
380{
381 return 0;
382}
383#endif /* CONFIG_DEBUG_FS */
384
828c0950 385static const struct file_operations ocfs2_osb_debug_fops = {
50397507
SM		 386 .open = ocfs2_osb_debug_open,
387 .release = ocfs2_debug_release,
388 .read = ocfs2_debug_read,
389 .llseek = generic_file_llseek,
390};
391
ccd979bd
MF		 392static int ocfs2_sync_fs(struct super_block *sb, int wait)
393{
bddb8eb3 394 int status;
ccd979bd
MF		 395 tid_t target;
396 struct ocfs2_super *osb = OCFS2_SB(sb);
397
ccd979bd
MF		 398 if (ocfs2_is_hard_readonly(osb))
399 return -EROFS;
400
401 if (wait) {
402 status = ocfs2_flush_truncate_log(osb);
403 if (status < 0)
404 mlog_errno(status);
405 } else {
406 ocfs2_schedule_truncate_log_flush(osb, 0);
407 }
408
2b4e30fb
JB		 409 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
410 &target)) {
ccd979bd 411 if (wait)
2b4e30fb
JB		 412 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
413 target);
ccd979bd
MF		 414 }
415 return 0;
416}
417
1a224ad1
JK		 418static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
419{
420 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
421 && (ino == USER_QUOTA_SYSTEM_INODE
422 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
423 return 0;
424 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
425 && (ino == GROUP_QUOTA_SYSTEM_INODE
426 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
427 return 0;
428 return 1;
429}
430
ccd979bd
MF		 431static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
432{
433 struct inode *new = NULL;
434 int status = 0;
435 int i;
436
437 mlog_entry_void();
438
5fa0613e 439 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 440 if (IS_ERR(new)) {
441 status = PTR_ERR(new);
442 mlog_errno(status);
443 goto bail;
444 }
445 osb->root_inode = new;
446
5fa0613e 447 new = ocfs2_iget(osb, osb->system_dir_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->sys_root_inode = new;
454
455 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
456 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
1a224ad1
JK		 457 if (!ocfs2_need_system_inode(osb, i))
458 continue;
ccd979bd
MF		 459 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
460 if (!new) {
461 ocfs2_release_system_inodes(osb);
462 status = -EINVAL;
463 mlog_errno(status);
464 /* FIXME: Should ERROR_RO_FS */
465 mlog(ML_ERROR, "Unable to load system inode %d, "
466 "possibly corrupt fs?", i);
467 goto bail;
468 }
469 // the array now has one ref, so drop this one
470 iput(new);
471 }
472
473bail:
474 mlog_exit(status);
475 return status;
476}
477
478static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
479{
480 struct inode *new = NULL;
481 int status = 0;
482 int i;
483
484 mlog_entry_void();
485
486 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
487 i < NUM_SYSTEM_INODES;
488 i++) {
1a224ad1
JK		 489 if (!ocfs2_need_system_inode(osb, i))
490 continue;
ccd979bd
MF		 491 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
492 if (!new) {
493 ocfs2_release_system_inodes(osb);
494 status = -EINVAL;
495 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
496 status, i, osb->slot_num);
497 goto bail;
498 }
499 /* the array now has one ref, so drop this one */
500 iput(new);
501 }
502
503bail:
504 mlog_exit(status);
505 return status;
506}
507
bddb8eb3 508static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 509{
bddb8eb3 510 int i;
ccd979bd
MF		 511 struct inode *inode;
512
513 mlog_entry_void();
514
515 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
516 inode = osb->system_inodes[i];
517 if (inode) {
518 iput(inode);
519 osb->system_inodes[i] = NULL;
520 }
521 }
522
523 inode = osb->sys_root_inode;
524 if (inode) {
525 iput(inode);
526 osb->sys_root_inode = NULL;
527 }
528
529 inode = osb->root_inode;
530 if (inode) {
531 iput(inode);
532 osb->root_inode = NULL;
533 }
534
bddb8eb3 535 mlog_exit(0);
ccd979bd
MF		 536}
537
538/* We're allocating fs objects, use GFP_NOFS */
539static struct inode *ocfs2_alloc_inode(struct super_block *sb)
540{
541 struct ocfs2_inode_info *oi;
542
e6b4f8da 543 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 544 if (!oi)
545 return NULL;
546
2b4e30fb 547 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
ccd979bd
MF		 548 return &oi->vfs_inode;
549}
550
551static void ocfs2_destroy_inode(struct inode *inode)
552{
553 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
554}
555
5a254031
MF		 556static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
557 unsigned int cbits)
ccd979bd 558{
5a254031
MF		 559 unsigned int bytes = 1 << cbits;
560 unsigned int trim = bytes;
561 unsigned int bitshift = 32;
562
563 /*
564 * i_size and all block offsets in ocfs2 are always 64 bits
565 * wide. i_clusters is 32 bits, in cluster-sized units. So on
566 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 567 */
568
569#if BITS_PER_LONG == 32
90c699a9 570# if defined(CONFIG_LBDAF)
2ecd05ae 571 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 572 /*
573 * We might be limited by page cache size.
574 */
575 if (bytes > PAGE_CACHE_SIZE) {
576 bytes = PAGE_CACHE_SIZE;
577 trim = 1;
578 /*
579 * Shift by 31 here so that we don't get larger than
580 * MAX_LFS_FILESIZE
581 */
582 bitshift = 31;
583 }
ccd979bd 584# else
5a254031
MF		 585 /*
586 * We are limited by the size of sector_t. Use block size, as
587 * that's what we expose to the VFS.
588 */
589 bytes = 1 << bbits;
590 trim = 1;
591 bitshift = 31;
ccd979bd
MF		 592# endif
593#endif
594
5a254031
MF		 595 /*
596 * Trim by a whole cluster when we can actually approach the
597 * on-disk limits. Otherwise we can overflow i_clusters when
598 * an extent start is at the max offset.
599 */
600 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 601}
602
603static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
604{
605 int incompat_features;
606 int ret = 0;
c0123ade 607 struct mount_options parsed_options;
ccd979bd
MF		 608 struct ocfs2_super *osb = OCFS2_SB(sb);
609
337eb00a
AIB		 610 lock_kernel();
611
5297aad8
JK		 612 if (!ocfs2_parse_options(sb, data, &parsed_options, 1) ||
613 !ocfs2_check_set_options(sb, &parsed_options)) {
ccd979bd
MF		 614 ret = -EINVAL;
615 goto out;
616 }
617
618 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
c0123ade 619 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
ccd979bd
MF		 620 ret = -EINVAL;
621 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
622 goto out;
623 }
624
625 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 626 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 627 ret = -EINVAL;
628 mlog(ML_ERROR, "Cannot change data mode on remount\n");
629 goto out;
630 }
631
12462f1d
JB		 632 /* Probably don't want this on remount; it might
633 * mess with other nodes */
634 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
635 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
636 ret = -EINVAL;
637 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
638 goto out;
639 }
640
ccd979bd
MF		 641 /* We're going to/from readonly mode. */
642 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
19ece546
JK		 643 /* Disable quota accounting before remounting RO */
644 if (*flags & MS_RDONLY) {
645 ret = ocfs2_susp_quotas(osb, 0);
646 if (ret < 0)
647 goto out;
648 }
ccd979bd
MF		 649 /* Lock here so the check of HARD_RO and the potential
650 * setting of SOFT_RO is atomic. */
651 spin_lock(&osb->osb_lock);
652 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
653 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
654 ret = -EROFS;
655 goto unlock_osb;
656 }
657
658 if (*flags & MS_RDONLY) {
659 mlog(0, "Going to ro mode.\n");
660 sb->s_flags |= MS_RDONLY;
661 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
662 } else {
663 mlog(0, "Making ro filesystem writeable.\n");
664
665 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
666 mlog(ML_ERROR, "Cannot remount RDWR "
667 "filesystem due to previous errors.\n");
668 ret = -EROFS;
669 goto unlock_osb;
670 }
671 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
672 if (incompat_features) {
673 mlog(ML_ERROR, "Cannot remount RDWR because "
674 "of unsupported optional features "
675 "(%x).\n", incompat_features);
676 ret = -EINVAL;
677 goto unlock_osb;
678 }
679 sb->s_flags &= ~MS_RDONLY;
680 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
681 }
682unlock_osb:
683 spin_unlock(&osb->osb_lock);
19ece546
JK		 684 /* Enable quota accounting after remounting RW */
685 if (!ret && !(*flags & MS_RDONLY)) {
686 if (sb_any_quota_suspended(sb))
687 ret = ocfs2_susp_quotas(osb, 1);
688 else
689 ret = ocfs2_enable_quotas(osb);
690 if (ret < 0) {
691 /* Return back changes... */
692 spin_lock(&osb->osb_lock);
693 sb->s_flags |= MS_RDONLY;
694 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
695 spin_unlock(&osb->osb_lock);
696 goto out;
697 }
698 }
ccd979bd
MF		 699 }
700
701 if (!ret) {
ccd979bd
MF		 702 /* Only save off the new mount options in case of a successful
703 * remount. */
c0123ade
TY		 704 osb->s_mount_opt = parsed_options.mount_opt;
705 osb->s_atime_quantum = parsed_options.atime_quantum;
706 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 707 if (parsed_options.commit_interval)
708 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 709
710 if (!ocfs2_is_hard_readonly(osb))
711 ocfs2_set_journal_params(osb);
57b09bb5
JK		 712
713 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
714 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ?
715 MS_POSIXACL : 0);
ccd979bd
MF		 716 }
717out:
337eb00a 718 unlock_kernel();
ccd979bd
MF		 719 return ret;
720}
721
722static int ocfs2_sb_probe(struct super_block *sb,
723 struct buffer_head **bh,
73be192b
JB		 724 int *sector_size,
725 struct ocfs2_blockcheck_stats *stats)
ccd979bd 726{
bddb8eb3 727 int status, tmpstat;
ccd979bd
MF		 728 struct ocfs1_vol_disk_hdr *hdr;
729 struct ocfs2_dinode *di;
730 int blksize;
731
732 *bh = NULL;
733
734 /* may be > 512 */
e1defc4f 735 *sector_size = bdev_logical_block_size(sb->s_bdev);
ccd979bd
MF		 736 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
737 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
738 *sector_size, OCFS2_MAX_BLOCKSIZE);
739 status = -EINVAL;
740 goto bail;
741 }
742
743 /* Can this really happen? */
744 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
745 *sector_size = OCFS2_MIN_BLOCKSIZE;
746
747 /* check block zero for old format */
748 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
749 if (status < 0) {
750 mlog_errno(status);
751 goto bail;
752 }
753 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
754 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
755 mlog(ML_ERROR, "incompatible version: %u.%u\n",
756 hdr->major_version, hdr->minor_version);
757 status = -EINVAL;
758 }
759 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
760 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
761 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
762 hdr->signature);
763 status = -EINVAL;
764 }
765 brelse(*bh);
766 *bh = NULL;
767 if (status < 0) {
768 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
769 "upgraded before mounting with ocfs v2\n");
770 goto bail;
771 }
772
773 /*
774 * Now check at magic offset for 512, 1024, 2048, 4096
775 * blocksizes. 4096 is the maximum blocksize because it is
776 * the minimum clustersize.
777 */
778 status = -EINVAL;
779 for (blksize = *sector_size;
780 blksize <= OCFS2_MAX_BLOCKSIZE;
781 blksize <<= 1) {
782 tmpstat = ocfs2_get_sector(sb, bh,
783 OCFS2_SUPER_BLOCK_BLKNO,
784 blksize);
785 if (tmpstat < 0) {
786 status = tmpstat;
787 mlog_errno(status);
fb5cbe9e 788 break;
ccd979bd
MF		 789 }
790 di = (struct ocfs2_dinode *) (*bh)->b_data;
73be192b 791 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
1c1d9793 792 spin_lock_init(&stats->b_lock);
fb5cbe9e
JB		 793 tmpstat = ocfs2_verify_volume(di, *bh, blksize, stats);
794 if (tmpstat < 0) {
795 brelse(*bh);
796 *bh = NULL;
797 }
798 if (tmpstat != -EAGAIN) {
799 status = tmpstat;
ccd979bd 800 break;
fb5cbe9e 801 }
ccd979bd
MF		 802 }
803
804bail:
805 return status;
806}
807
c271c5c2
SM		 808static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
809{
810 if (ocfs2_mount_local(osb)) {
811 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
812 mlog(ML_ERROR, "Cannot heartbeat on a locally "
813 "mounted device.\n");
814 return -EINVAL;
815 }
816 }
817
b61817e1
JB		 818 if (ocfs2_userspace_stack(osb)) {
819 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
820 mlog(ML_ERROR, "Userspace stack expected, but "
821 "o2cb heartbeat arguments passed to mount\n");
822 return -EINVAL;
823 }
824 }
825
c271c5c2 826 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
b61817e1
JB		 827 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
828 !ocfs2_userspace_stack(osb)) {
c271c5c2
SM		 829 mlog(ML_ERROR, "Heartbeat has to be started to mount "
830 "a read-write clustered device.\n");
831 return -EINVAL;
832 }
833 }
834
835 return 0;
836}
837
b61817e1
JB		 838/*
839 * If we're using a userspace stack, mount should have passed
840 * a name that matches the disk. If not, mount should not
841 * have passed a stack.
842 */
843static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
844 struct mount_options *mopt)
845{
846 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
847 mlog(ML_ERROR,
848 "cluster stack passed to mount, but this filesystem "
849 "does not support it\n");
850 return -EINVAL;
851 }
852
853 if (ocfs2_userspace_stack(osb) &&
854 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
855 OCFS2_STACK_LABEL_LEN)) {
856 mlog(ML_ERROR,
857 "cluster stack passed to mount (\"%s\") does not "
858 "match the filesystem (\"%s\")\n",
859 mopt->cluster_stack,
860 osb->osb_cluster_stack);
861 return -EINVAL;
862 }
863
864 return 0;
865}
866
19ece546
JK		 867static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
868{
869 int type;
870 struct super_block *sb = osb->sb;
871 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
872 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
873 int status = 0;
874
875 for (type = 0; type < MAXQUOTAS; type++) {
876 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
877 continue;
878 if (unsuspend)
879 status = vfs_quota_enable(
880 sb_dqopt(sb)->files[type],
881 type, QFMT_OCFS2,
882 DQUOT_SUSPENDED);
883 else
884 status = vfs_quota_disable(sb, type,
885 DQUOT_SUSPENDED);
886 if (status < 0)
887 break;
888 }
889 if (status < 0)
890 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
891 "remount (error = %d).\n", status);
892 return status;
893}
894
895static int ocfs2_enable_quotas(struct ocfs2_super *osb)
896{
897 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
898 struct super_block *sb = osb->sb;
899 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
900 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
901 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
902 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
903 int status;
904 int type;
905
906 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
907 for (type = 0; type < MAXQUOTAS; type++) {
908 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
909 continue;
910 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
911 osb->slot_num);
912 if (!inode[type]) {
913 status = -ENOENT;
914 goto out_quota_off;
915 }
916 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
917 DQUOT_USAGE_ENABLED);
918 if (status < 0)
919 goto out_quota_off;
920 }
921
922 for (type = 0; type < MAXQUOTAS; type++)
923 iput(inode[type]);
924 return 0;
925out_quota_off:
926 ocfs2_disable_quotas(osb);
927 for (type = 0; type < MAXQUOTAS; type++)
928 iput(inode[type]);
929 mlog_errno(status);
930 return status;
931}
932
933static void ocfs2_disable_quotas(struct ocfs2_super *osb)
934{
935 int type;
936 struct inode *inode;
937 struct super_block *sb = osb->sb;
938
939 /* We mostly ignore errors in this function because there's not much
940 * we can do when we see them */
941 for (type = 0; type < MAXQUOTAS; type++) {
942 if (!sb_has_quota_loaded(sb, type))
943 continue;
944 inode = igrab(sb->s_dquot.files[type]);
945 /* Turn off quotas. This will remove all dquot structures from
946 * memory and so they will be automatically synced to global
947 * quota files */
948 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
949 DQUOT_LIMITS_ENABLED);
950 if (!inode)
951 continue;
952 iput(inode);
953 }
954}
955
956/* Handle quota on quotactl */
957static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount)
959{
960 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
961 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
962
963 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
964 return -EINVAL;
965
966 if (remount)
967 return 0; /* Just ignore it has been handled in
968 * ocfs2_remount() */
969 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
970 format_id, DQUOT_LIMITS_ENABLED);
971}
972
973/* Handle quota off quotactl */
974static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
975{
976 if (remount)
977 return 0; /* Ignore now and handle later in
978 * ocfs2_remount() */
979 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
980}
981
0d54b217 982static const struct quotactl_ops ocfs2_quotactl_ops = {
19ece546
JK		 983 .quota_on = ocfs2_quota_on,
984 .quota_off = ocfs2_quota_off,
985 .quota_sync = vfs_quota_sync,
986 .get_info = vfs_get_dqinfo,
987 .set_info = vfs_set_dqinfo,
988 .get_dqblk = vfs_get_dqblk,
989 .set_dqblk = vfs_set_dqblk,
990};
991
ccd979bd
MF		 992static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
993{
994 struct dentry *root;
995 int status, sector_size;
c0123ade 996 struct mount_options parsed_options;
ccd979bd
MF		 997 struct inode *inode = NULL;
998 struct ocfs2_super *osb = NULL;
999 struct buffer_head *bh = NULL;
c271c5c2 1000 char nodestr[8];
73be192b 1001 struct ocfs2_blockcheck_stats stats;
ccd979bd
MF		 1002
1003 mlog_entry("%p, %p, %i", sb, data, silent);
1004
c0123ade 1005 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 1006 status = -EINVAL;
1007 goto read_super_error;
1008 }
1009
1010 /* probe for superblock */
73be192b 1011 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
ccd979bd
MF		 1012 if (status < 0) {
1013 mlog(ML_ERROR, "superblock probe failed!\n");
1014 goto read_super_error;
1015 }
1016
73be192b 1017 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
ccd979bd
MF		 1018 osb = OCFS2_SB(sb);
1019 if (status < 0) {
1020 mlog_errno(status);
1021 goto read_super_error;
1022 }
1023 brelse(bh);
1024 bh = NULL;
a68979b8 1025
5297aad8
JK		 1026 if (!ocfs2_check_set_options(sb, &parsed_options)) {
1027 status = -EINVAL;
1028 goto read_super_error;
1029 }
c0123ade
TY		 1030 osb->s_mount_opt = parsed_options.mount_opt;
1031 osb->s_atime_quantum = parsed_options.atime_quantum;
1032 osb->preferred_slot = parsed_options.slot;
d147b3d6 1033 osb->osb_commit_interval = parsed_options.commit_interval;
73c8a800
MF		 1034
1035 ocfs2_la_set_sizes(osb, parsed_options.localalloc_opt);
d02f00cc 1036 osb->osb_resv_level = parsed_options.resv_level;
ccd979bd 1037
b61817e1
JB		 1038 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1039 if (status)
1040 goto read_super_error;
1041
ccd979bd
MF		 1042 sb->s_magic = OCFS2_SUPER_MAGIC;
1043
a68979b8
TY		 1044 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1045 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1046
ccd979bd
MF		 1047 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
1048 * heartbeat=none */
1049 if (bdev_read_only(sb->s_bdev)) {
1050 if (!(sb->s_flags & MS_RDONLY)) {
1051 status = -EACCES;
1052 mlog(ML_ERROR, "Readonly device detected but readonly "
1053 "mount was not specified.\n");
1054 goto read_super_error;
1055 }
1056
1057 /* You should not be able to start a local heartbeat
1058 * on a readonly device. */
1059 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
1060 status = -EROFS;
1061 mlog(ML_ERROR, "Local heartbeat specified on readonly "
1062 "device.\n");
1063 goto read_super_error;
1064 }
1065
1066 status = ocfs2_check_journals_nolocks(osb);
1067 if (status < 0) {
1068 if (status == -EROFS)
1069 mlog(ML_ERROR, "Recovery required on readonly "
1070 "file system, but write access is "
1071 "unavailable.\n");
1072 else
2bd63216 1073 mlog_errno(status);
ccd979bd
MF		 1074 goto read_super_error;
1075 }
1076
1077 ocfs2_set_ro_flag(osb, 1);
1078
1079 printk(KERN_NOTICE "Readonly device detected. No cluster "
1080 "services will be utilized for this mount. Recovery "
1081 "will be skipped.\n");
1082 }
1083
1084 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 1085 if (sb->s_flags & MS_RDONLY)
1086 ocfs2_set_ro_flag(osb, 0);
1087 }
1088
c271c5c2
SM		 1089 status = ocfs2_verify_heartbeat(osb);
1090 if (status < 0) {
1091 mlog_errno(status);
1092 goto read_super_error;
1093 }
1094
ccd979bd
MF		 1095 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
1096 ocfs2_debugfs_root);
1097 if (!osb->osb_debug_root) {
1098 status = -EINVAL;
1099 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
1100 goto read_super_error;
1101 }
1102
50397507
SM		 1103 osb->osb_ctxt = debugfs_create_file("fs_state", S_IFREG|S_IRUSR,
1104 osb->osb_debug_root,
1105 osb,
1106 &ocfs2_osb_debug_fops);
1107 if (!osb->osb_ctxt) {
1108 status = -EINVAL;
1109 mlog_errno(status);
1110 goto read_super_error;
1111 }
1112
73be192b
JB		 1113 if (ocfs2_meta_ecc(osb)) {
1114 status = ocfs2_blockcheck_stats_debugfs_install(
1115 &osb->osb_ecc_stats,
1116 osb->osb_debug_root);
1117 if (status) {
1118 mlog(ML_ERROR,
1119 "Unable to create blockcheck statistics "
1120 "files\n");
1121 goto read_super_error;
1122 }
1123 }
1124
ccd979bd
MF		 1125 status = ocfs2_mount_volume(sb);
1126 if (osb->root_inode)
1127 inode = igrab(osb->root_inode);
1128
1129 if (status < 0)
1130 goto read_super_error;
1131
1132 if (!inode) {
1133 status = -EIO;
1134 mlog_errno(status);
1135 goto read_super_error;
1136 }
1137
1138 root = d_alloc_root(inode);
1139 if (!root) {
1140 status = -ENOMEM;
1141 mlog_errno(status);
1142 goto read_super_error;
1143 }
1144
1145 sb->s_root = root;
1146
1147 ocfs2_complete_mount_recovery(osb);
1148
c271c5c2
SM		 1149 if (ocfs2_mount_local(osb))
1150 snprintf(nodestr, sizeof(nodestr), "local");
1151 else
19fdb624 1152 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1153
1154 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 1155 "with %s data mode.\n",
c271c5c2 1156 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 1157 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
1158 "ordered");
1159
1160 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
1161 wake_up(&osb->osb_mount_event);
1162
19ece546
JK		 1163 /* Now we can initialize quotas because we can afford to wait
1164 * for cluster locks recovery now. That also means that truncation
1165 * log recovery can happen but that waits for proper quota setup */
1166 if (!(sb->s_flags & MS_RDONLY)) {
1167 status = ocfs2_enable_quotas(osb);
1168 if (status < 0) {
1169 /* We have to err-out specially here because
1170 * s_root is already set */
1171 mlog_errno(status);
1172 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1173 wake_up(&osb->osb_mount_event);
1174 mlog_exit(status);
1175 return status;
1176 }
1177 }
1178
1179 ocfs2_complete_quota_recovery(osb);
1180
1181 /* Now we wake up again for processes waiting for quotas */
1182 atomic_set(&osb->vol_state, VOLUME_MOUNTED_QUOTAS);
1183 wake_up(&osb->osb_mount_event);
1184
df152c24 1185 /* Start this when the mount is almost sure of being successful */
8b712cd5 1186 ocfs2_orphan_scan_start(osb);
df152c24 1187
ccd979bd
MF		 1188 mlog_exit(status);
1189 return status;
1190
1191read_super_error:
a81cb88b 1192 brelse(bh);
ccd979bd
MF		 1193
1194 if (inode)
1195 iput(inode);
1196
1197 if (osb) {
1198 atomic_set(&osb->vol_state, VOLUME_DISABLED);
1199 wake_up(&osb->osb_mount_event);
1200 ocfs2_dismount_volume(sb, 1);
1201 }
1202
1203 mlog_exit(status);
1204 return status;
1205}
1206
454e2398
DH		 1207static int ocfs2_get_sb(struct file_system_type *fs_type,
1208 int flags,
1209 const char *dev_name,
1210 void *data,
1211 struct vfsmount *mnt)
ccd979bd 1212{
454e2398
DH		 1213 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1214 mnt);
ccd979bd
MF		 1215}
1216
f7b1aa69
JK		 1217static void ocfs2_kill_sb(struct super_block *sb)
1218{
1219 struct ocfs2_super *osb = OCFS2_SB(sb);
1220
5fd13189
JK		 1221 /* Failed mount? */
1222 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1223 goto out;
1224
f7b1aa69
JK		 1225 /* Prevent further queueing of inode drop events */
1226 spin_lock(&dentry_list_lock);
1227 ocfs2_set_osb_flag(osb, OCFS2_OSB_DROP_DENTRY_LOCK_IMMED);
1228 spin_unlock(&dentry_list_lock);
1229 /* Wait for work to finish and/or remove it */
1230 cancel_work_sync(&osb->dentry_lock_work);
5fd13189 1231out:
f7b1aa69
JK		 1232 kill_block_super(sb);
1233}
1234
ccd979bd
MF		 1235static struct file_system_type ocfs2_fs_type = {
1236 .owner = THIS_MODULE,
1237 .name = "ocfs2",
1238 .get_sb = ocfs2_get_sb, /* is this called when we mount
1239 * the fs? */
f7b1aa69
JK		 1240 .kill_sb = ocfs2_kill_sb,
1241
1ba9da2f 1242 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 1243 .next = NULL
1244};
1245
5297aad8
JK		 1246static int ocfs2_check_set_options(struct super_block *sb,
1247 struct mount_options *options)
1248{
1249 if (options->mount_opt & OCFS2_MOUNT_USRQUOTA &&
1250 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1251 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1252 mlog(ML_ERROR, "User quotas were requested, but this "
1253 "filesystem does not have the feature enabled.\n");
1254 return 0;
1255 }
1256 if (options->mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1257 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1258 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1259 mlog(ML_ERROR, "Group quotas were requested, but this "
1260 "filesystem does not have the feature enabled.\n");
1261 return 0;
1262 }
1263 if (options->mount_opt & OCFS2_MOUNT_POSIX_ACL &&
1264 !OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR)) {
1265 mlog(ML_ERROR, "ACL support requested but extended attributes "
1266 "feature is not enabled\n");
1267 return 0;
1268 }
1269 /* No ACL setting specified? Use XATTR feature... */
1270 if (!(options->mount_opt & (OCFS2_MOUNT_POSIX_ACL |
1271 OCFS2_MOUNT_NO_POSIX_ACL))) {
1272 if (OCFS2_HAS_INCOMPAT_FEATURE(sb, OCFS2_FEATURE_INCOMPAT_XATTR))
1273 options->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1274 else
1275 options->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
1276 }
1277 return 1;
1278}
1279
ccd979bd
MF		 1280static int ocfs2_parse_options(struct super_block *sb,
1281 char *options,
c0123ade 1282 struct mount_options *mopt,
ccd979bd
MF		 1283 int is_remount)
1284{
1285 int status;
1286 char *p;
1287
1288 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1289 options ? options : "(none)");
1290
d147b3d6 1291 mopt->commit_interval = 0;
c0123ade
TY		 1292 mopt->mount_opt = 0;
1293 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1294 mopt->slot = OCFS2_INVALID_SLOT;
73c8a800 1295 mopt->localalloc_opt = -1;
b61817e1 1296 mopt->cluster_stack[0] = '\0';
d02f00cc 1297 mopt->resv_level = OCFS2_DEFAULT_RESV_LEVEL;
ccd979bd
MF		 1298
1299 if (!options) {
1300 status = 1;
1301 goto bail;
1302 }
1303
1304 while ((p = strsep(&options, ",")) != NULL) {
1305 int token, option;
1306 substring_t args[MAX_OPT_ARGS];
1307
1308 if (!*p)
1309 continue;
1310
1311 token = match_token(p, tokens, args);
1312 switch (token) {
1313 case Opt_hb_local:
c0123ade 1314 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1315 break;
1316 case Opt_hb_none:
c0123ade 1317 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 1318 break;
1319 case Opt_barrier:
1320 if (match_int(&args[0], &option)) {
1321 status = 0;
1322 goto bail;
1323 }
1324 if (option)
c0123ade 1325 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 1326 else
c0123ade 1327 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 1328 break;
1329 case Opt_intr:
c0123ade 1330 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1331 break;
1332 case Opt_nointr:
c0123ade 1333 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 1334 break;
1335 case Opt_err_panic:
c0123ade 1336 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1337 break;
1338 case Opt_err_ro:
c0123ade 1339 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 1340 break;
1341 case Opt_data_ordered:
c0123ade 1342 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 1343 break;
1344 case Opt_data_writeback:
c0123ade 1345 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 1346 break;
cf1d6c76
TY		 1347 case Opt_user_xattr:
1348 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1349 break;
1350 case Opt_nouser_xattr:
1351 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1352 break;
7f1a37e3
TY		 1353 case Opt_atime_quantum:
1354 if (match_int(&args[0], &option)) {
1355 status = 0;
1356 goto bail;
1357 }
1358 if (option >= 0)
c0123ade 1359 mopt->atime_quantum = option;
7f1a37e3 1360 break;
baf4661a
SM		 1361 case Opt_slot:
1362 option = 0;
1363 if (match_int(&args[0], &option)) {
1364 status = 0;
1365 goto bail;
1366 }
1367 if (option)
c0123ade 1368 mopt->slot = (s16)option;
baf4661a 1369 break;
d147b3d6
MF		 1370 case Opt_commit:
1371 option = 0;
1372 if (match_int(&args[0], &option)) {
1373 status = 0;
1374 goto bail;
1375 }
1376 if (option < 0)
1377 return 0;
1378 if (option == 0)
2b4e30fb 1379 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
d147b3d6
MF		 1380 mopt->commit_interval = HZ * option;
1381 break;
2fbe8d1e
SM		 1382 case Opt_localalloc:
1383 option = 0;
1384 if (match_int(&args[0], &option)) {
1385 status = 0;
1386 goto bail;
1387 }
73c8a800 1388 if (option >= 0)
2fbe8d1e
SM		 1389 mopt->localalloc_opt = option;
1390 break;
53fc622b
MF		 1391 case Opt_localflocks:
1392 /*
1393 * Changing this during remount could race
1394 * flock() requests, or "unbalance" existing
1395 * ones (e.g., a lock is taken in one mode but
1396 * dropped in the other). If users care enough
1397 * to flip locking modes during remount, we
1398 * could add a "local" flag to individual
1399 * flock structures for proper tracking of
1400 * state.
1401 */
1402 if (!is_remount)
1403 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1404 break;
b61817e1
JB		 1405 case Opt_stack:
1406 /* Check both that the option we were passed
1407 * is of the right length and that it is a proper
1408 * string of the right length.
1409 */
1410 if (((args[0].to - args[0].from) !=
1411 OCFS2_STACK_LABEL_LEN) ||
1412 (strnlen(args[0].from,
1413 OCFS2_STACK_LABEL_LEN) !=
1414 OCFS2_STACK_LABEL_LEN)) {
1415 mlog(ML_ERROR,
1416 "Invalid cluster_stack option\n");
1417 status = 0;
1418 goto bail;
1419 }
1420 memcpy(mopt->cluster_stack, args[0].from,
1421 OCFS2_STACK_LABEL_LEN);
1422 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1423 break;
12462f1d
JB		 1424 case Opt_inode64:
1425 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1426 break;
19ece546 1427 case Opt_usrquota:
19ece546
JK		 1428 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1429 break;
1430 case Opt_grpquota:
19ece546
JK		 1431 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1432 break;
a68979b8
TY		 1433 case Opt_acl:
1434 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
5297aad8 1435 mopt->mount_opt &= ~OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1436 break;
1437 case Opt_noacl:
5297aad8 1438 mopt->mount_opt |= OCFS2_MOUNT_NO_POSIX_ACL;
a68979b8
TY		 1439 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1440 break;
d02f00cc
MF		 1441 case Opt_resv_level:
1442 if (is_remount)
1443 break;
1444 if (match_int(&args[0], &option)) {
1445 status = 0;
1446 goto bail;
1447 }
1448 if (option >= OCFS2_MIN_RESV_LEVEL &&
1449 option < OCFS2_MAX_RESV_LEVEL)
1450 mopt->resv_level = option;
1451 break;
ccd979bd
MF		 1452 default:
1453 mlog(ML_ERROR,
1454 "Unrecognized mount option \"%s\" "
1455 "or missing value\n", p);
1456 status = 0;
1457 goto bail;
1458 }
1459 }
1460
1461 status = 1;
1462
1463bail:
1464 mlog_exit(status);
1465 return status;
1466}
1467
d550071c
SM		 1468static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1469{
1470 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1471 unsigned long opts = osb->s_mount_opt;
ebcee4b5 1472 unsigned int local_alloc_megs;
d550071c
SM		 1473
1474 if (opts & OCFS2_MOUNT_HB_LOCAL)
1475 seq_printf(s, ",_netdev,heartbeat=local");
1476 else
1477 seq_printf(s, ",heartbeat=none");
1478
1479 if (opts & OCFS2_MOUNT_NOINTR)
1480 seq_printf(s, ",nointr");
1481
1482 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1483 seq_printf(s, ",data=writeback");
1484 else
1485 seq_printf(s, ",data=ordered");
1486
1487 if (opts & OCFS2_MOUNT_BARRIER)
1488 seq_printf(s, ",barrier=1");
1489
1490 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1491 seq_printf(s, ",errors=panic");
1492 else
1493 seq_printf(s, ",errors=remount-ro");
1494
1495 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1496 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1497
1498 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1499 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1500
d147b3d6
MF		 1501 if (osb->osb_commit_interval)
1502 seq_printf(s, ",commit=%u",
1503 (unsigned) (osb->osb_commit_interval / HZ));
1504
ebcee4b5
MF		 1505 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1506 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1507 seq_printf(s, ",localalloc=%d", local_alloc_megs);
2fbe8d1e 1508
53fc622b
MF		 1509 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1510 seq_printf(s, ",localflocks,");
1511
b61817e1
JB		 1512 if (osb->osb_cluster_stack[0])
1513 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1514 osb->osb_cluster_stack);
19ece546
JK		 1515 if (opts & OCFS2_MOUNT_USRQUOTA)
1516 seq_printf(s, ",usrquota");
1517 if (opts & OCFS2_MOUNT_GRPQUOTA)
1518 seq_printf(s, ",grpquota");
b61817e1 1519
b0f73cfc
SM		 1520 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1521 seq_printf(s, ",nouser_xattr");
1522 else
1523 seq_printf(s, ",user_xattr");
1524
12462f1d
JB		 1525 if (opts & OCFS2_MOUNT_INODE64)
1526 seq_printf(s, ",inode64");
1527
a68979b8
TY		 1528 if (opts & OCFS2_MOUNT_POSIX_ACL)
1529 seq_printf(s, ",acl");
1530 else
1531 seq_printf(s, ",noacl");
a68979b8 1532
d02f00cc
MF		 1533 if (osb->osb_resv_level != OCFS2_DEFAULT_RESV_LEVEL)
1534 seq_printf(s, ",resv_level=%d", osb->osb_resv_level);
1535
d550071c
SM		 1536 return 0;
1537}
1538
ccd979bd
MF		 1539static int __init ocfs2_init(void)
1540{
1541 int status;
1542
1543 mlog_entry_void();
1544
1545 ocfs2_print_version();
1546
ccd979bd
MF		 1547 status = init_ocfs2_uptodate_cache();
1548 if (status < 0) {
1549 mlog_errno(status);
1550 goto leave;
1551 }
1552
1553 status = ocfs2_initialize_mem_caches();
1554 if (status < 0) {
1555 mlog_errno(status);
1556 goto leave;
1557 }
1558
1559 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1560 if (!ocfs2_wq) {
1561 status = -ENOMEM;
1562 goto leave;
1563 }
1564
ccd979bd
MF		 1565 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1566 if (!ocfs2_debugfs_root) {
1567 status = -EFAULT;
1568 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1569 }
1570
171bf93c
MF		 1571 status = ocfs2_quota_setup();
1572 if (status)
1573 goto leave;
1574
63e0c48a
JB		 1575 ocfs2_set_locking_protocol();
1576
9e33d69f 1577 status = register_quota_format(&ocfs2_quota_format);
ccd979bd
MF		 1578leave:
1579 if (status < 0) {
171bf93c 1580 ocfs2_quota_shutdown();
ccd979bd
MF		 1581 ocfs2_free_mem_caches();
1582 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 1583 }
1584
1585 mlog_exit(status);
1586
1587 if (status >= 0) {
1588 return register_filesystem(&ocfs2_fs_type);
1589 } else
1590 return -1;
1591}
1592
1593static void __exit ocfs2_exit(void)
1594{
1595 mlog_entry_void();
1596
171bf93c
MF		 1597 ocfs2_quota_shutdown();
1598
ccd979bd
MF		 1599 if (ocfs2_wq) {
1600 flush_workqueue(ocfs2_wq);
1601 destroy_workqueue(ocfs2_wq);
1602 }
1603
9e33d69f
JK		 1604 unregister_quota_format(&ocfs2_quota_format);
1605
ccd979bd
MF		 1606 debugfs_remove(ocfs2_debugfs_root);
1607
1608 ocfs2_free_mem_caches();
1609
1610 unregister_filesystem(&ocfs2_fs_type);
1611
ccd979bd
MF		 1612 exit_ocfs2_uptodate_cache();
1613
1614 mlog_exit_void();
1615}
1616
1617static void ocfs2_put_super(struct super_block *sb)
1618{
1619 mlog_entry("(0x%p)\n", sb);
1620
6cfd0148
CH		 1621 lock_kernel();
1622
ccd979bd
MF		 1623 ocfs2_sync_blockdev(sb);
1624 ocfs2_dismount_volume(sb, 0);
1625
6cfd0148
CH		 1626 unlock_kernel();
1627
ccd979bd
MF		 1628 mlog_exit_void();
1629}
1630
726c3342 1631static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1632{
1633 struct ocfs2_super *osb;
1634 u32 numbits, freebits;
1635 int status;
1636 struct ocfs2_dinode *bm_lock;
1637 struct buffer_head *bh = NULL;
1638 struct inode *inode = NULL;
1639
726c3342 1640 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
ccd979bd 1641
726c3342 1642 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1643
1644 inode = ocfs2_get_system_file_inode(osb,
1645 GLOBAL_BITMAP_SYSTEM_INODE,
1646 OCFS2_INVALID_SLOT);
1647 if (!inode) {
1648 mlog(ML_ERROR, "failed to get bitmap inode\n");
1649 status = -EIO;
1650 goto bail;
1651 }
1652
e63aecb6 1653 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1654 if (status < 0) {
1655 mlog_errno(status);
1656 goto bail;
1657 }
1658
1659 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1660
1661 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1662 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1663
1664 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1665 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1666 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1667 buf->f_blocks = ((sector_t) numbits) *
1668 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1669 buf->f_bfree = ((sector_t) freebits) *
1670 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1671 buf->f_bavail = buf->f_bfree;
1672 buf->f_files = numbits;
1673 buf->f_ffree = freebits;
837711f8
CL		 1674 buf->f_fsid.val[0] = crc32_le(0, osb->uuid_str, OCFS2_VOL_UUID_LEN)
1675 & 0xFFFFFFFFUL;
1676 buf->f_fsid.val[1] = crc32_le(0, osb->uuid_str + OCFS2_VOL_UUID_LEN,
1677 OCFS2_VOL_UUID_LEN) & 0xFFFFFFFFUL;
ccd979bd
MF		 1678
1679 brelse(bh);
1680
e63aecb6 1681 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1682 status = 0;
1683bail:
1684 if (inode)
1685 iput(inode);
1686
1687 mlog_exit(status);
1688
1689 return status;
1690}
1691
51cc5068 1692static void ocfs2_inode_init_once(void *data)
ccd979bd
MF		 1693{
1694 struct ocfs2_inode_info *oi = data;
1695
a35afb83
CL		 1696 oi->ip_flags = 0;
1697 oi->ip_open_count = 0;
1698 spin_lock_init(&oi->ip_lock);
1699 ocfs2_extent_map_init(&oi->vfs_inode);
1700 INIT_LIST_HEAD(&oi->ip_io_markers);
a35afb83 1701 oi->ip_dir_start_lookup = 0;
ccd979bd 1702
a35afb83 1703 init_rwsem(&oi->ip_alloc_sem);
cf1d6c76 1704 init_rwsem(&oi->ip_xattr_sem);
a35afb83 1705 mutex_init(&oi->ip_io_mutex);
ccd979bd 1706
a35afb83
CL		 1707 oi->ip_blkno = 0ULL;
1708 oi->ip_clusters = 0;
ccd979bd 1709
4fe370af
MF		 1710 ocfs2_resv_init_once(&oi->ip_la_data_resv);
1711
a35afb83 1712 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1713 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1714 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1715
8cb471e8 1716 ocfs2_metadata_cache_init(INODE_CACHE(&oi->vfs_inode),
6e5a3d75 1717 &ocfs2_inode_caching_ops);
ccd979bd 1718
a35afb83 1719 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1720}
1721
1722static int ocfs2_initialize_mem_caches(void)
1723{
1724 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1725 sizeof(struct ocfs2_inode_info),
1726 0,
1727 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1728 SLAB_MEM_SPREAD),
20c2df83 1729 ocfs2_inode_init_once);
9e33d69f
JK		 1730 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1731 sizeof(struct ocfs2_dquot),
1732 0,
1733 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1734 SLAB_MEM_SPREAD),
1735 NULL);
1736 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1737 sizeof(struct ocfs2_quota_chunk),
1738 0,
1739 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1740 NULL);
1741 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1742 !ocfs2_qf_chunk_cachep) {
1743 if (ocfs2_inode_cachep)
1744 kmem_cache_destroy(ocfs2_inode_cachep);
1745 if (ocfs2_dquot_cachep)
1746 kmem_cache_destroy(ocfs2_dquot_cachep);
1747 if (ocfs2_qf_chunk_cachep)
1748 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
ccd979bd 1749 return -ENOMEM;
9e33d69f 1750 }
ccd979bd 1751
ccd979bd
MF		 1752 return 0;
1753}
1754
1755static void ocfs2_free_mem_caches(void)
1756{
1757 if (ocfs2_inode_cachep)
1758 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd 1759 ocfs2_inode_cachep = NULL;
9e33d69f
JK		 1760
1761 if (ocfs2_dquot_cachep)
1762 kmem_cache_destroy(ocfs2_dquot_cachep);
1763 ocfs2_dquot_cachep = NULL;
1764
1765 if (ocfs2_qf_chunk_cachep)
1766 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1767 ocfs2_qf_chunk_cachep = NULL;
ccd979bd
MF		 1768}
1769
1770static int ocfs2_get_sector(struct super_block *sb,
1771 struct buffer_head **bh,
1772 int block,
1773 int sect_size)
1774{
1775 if (!sb_set_blocksize(sb, sect_size)) {
1776 mlog(ML_ERROR, "unable to set blocksize\n");
1777 return -EIO;
1778 }
1779
1780 *bh = sb_getblk(sb, block);
1781 if (!*bh) {
1782 mlog_errno(-EIO);
1783 return -EIO;
1784 }
1785 lock_buffer(*bh);
1786 if (!buffer_dirty(*bh))
1787 clear_buffer_uptodate(*bh);
1788 unlock_buffer(*bh);
1789 ll_rw_block(READ, 1, bh);
1790 wait_on_buffer(*bh);
28d57d43 1791 if (!buffer_uptodate(*bh)) {
1792 mlog_errno(-EIO);
1793 brelse(*bh);
1794 *bh = NULL;
1795 return -EIO;
1796 }
1797
ccd979bd
MF		 1798 return 0;
1799}
1800
ccd979bd
MF		 1801static int ocfs2_mount_volume(struct super_block *sb)
1802{
1803 int status = 0;
1804 int unlock_super = 0;
1805 struct ocfs2_super *osb = OCFS2_SB(sb);
1806
1807 mlog_entry_void();
1808
1809 if (ocfs2_is_hard_readonly(osb))
1810 goto leave;
1811
ccd979bd
MF		 1812 status = ocfs2_dlm_init(osb);
1813 if (status < 0) {
1814 mlog_errno(status);
1815 goto leave;
1816 }
1817
ccd979bd
MF		 1818 status = ocfs2_super_lock(osb, 1);
1819 if (status < 0) {
1820 mlog_errno(status);
1821 goto leave;
1822 }
1823 unlock_super = 1;
1824
1825 /* This will load up the node map and add ourselves to it. */
1826 status = ocfs2_find_slot(osb);
1827 if (status < 0) {
1828 mlog_errno(status);
1829 goto leave;
1830 }
1831
ccd979bd
MF		 1832 /* load all node-local system inodes */
1833 status = ocfs2_init_local_system_inodes(osb);
1834 if (status < 0) {
1835 mlog_errno(status);
1836 goto leave;
1837 }
1838
1839 status = ocfs2_check_volume(osb);
1840 if (status < 0) {
1841 mlog_errno(status);
1842 goto leave;
1843 }
1844
1845 status = ocfs2_truncate_log_init(osb);
06c59bb8 1846 if (status < 0)
ccd979bd 1847 mlog_errno(status);
c271c5c2 1848
ccd979bd
MF		 1849leave:
1850 if (unlock_super)
1851 ocfs2_super_unlock(osb, 1);
1852
1853 mlog_exit(status);
1854 return status;
1855}
1856
ccd979bd
MF		 1857static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1858{
286eaa95 1859 int tmp, hangup_needed = 0;
ccd979bd 1860 struct ocfs2_super *osb = NULL;
c271c5c2 1861 char nodestr[8];
ccd979bd
MF		 1862
1863 mlog_entry("(0x%p)\n", sb);
1864
1865 BUG_ON(!sb);
1866 osb = OCFS2_SB(sb);
1867 BUG_ON(!osb);
1868
50397507
SM		 1869 debugfs_remove(osb->osb_ctxt);
1870
f7b1aa69
JK		 1871 /*
1872 * Flush inode dropping work queue so that deletes are
1873 * performed while the filesystem is still working
1874 */
1875 ocfs2_drop_all_dl_inodes(osb);
1876
692684e1
SM		 1877 /* Orphan scan should be stopped as early as possible */
1878 ocfs2_orphan_scan_stop(osb);
1879
19ece546
JK		 1880 ocfs2_disable_quotas(osb);
1881
ccd979bd
MF		 1882 ocfs2_shutdown_local_alloc(osb);
1883
1884 ocfs2_truncate_log_shutdown(osb);
1885
553abd04
JB		 1886 /* This will disable recovery and flush any recovery work. */
1887 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1888
1889 ocfs2_journal_shutdown(osb);
1890
1891 ocfs2_sync_blockdev(sb);
1892
374a263e
TM		 1893 ocfs2_purge_refcount_trees(osb);
1894
4670c46d
JB		 1895 /* No cluster connection means we've failed during mount, so skip
1896 * all the steps which depended on that to complete. */
1897 if (osb->cconn) {
ccd979bd
MF		 1898 tmp = ocfs2_super_lock(osb, 1);
1899 if (tmp < 0) {
1900 mlog_errno(tmp);
1901 return;
1902 }
19b613d4 1903 }
ccd979bd 1904
19b613d4
MF		 1905 if (osb->slot_num != OCFS2_INVALID_SLOT)
1906 ocfs2_put_slot(osb);
ccd979bd 1907
4670c46d 1908 if (osb->cconn)
ccd979bd 1909 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1910
1911 ocfs2_release_system_inodes(osb);
1912
6953b4c0 1913 /*
6953b4c0
JB		 1914 * If we're dismounting due to mount error, mount.ocfs2 will clean
1915 * up heartbeat. If we're a local mount, there is no heartbeat.
1916 * If we failed before we got a uuid_str yet, we can't stop
1917 * heartbeat. Otherwise, do it.
1918 */
1919 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
286eaa95
JB		 1920 hangup_needed = 1;
1921
1922 if (osb->cconn)
1923 ocfs2_dlm_shutdown(osb, hangup_needed);
1924
73be192b 1925 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
286eaa95
JB		 1926 debugfs_remove(osb->osb_debug_root);
1927
1928 if (hangup_needed)
6953b4c0 1929 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
ccd979bd
MF		 1930
1931 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1932
c271c5c2
SM		 1933 if (ocfs2_mount_local(osb))
1934 snprintf(nodestr, sizeof(nodestr), "local");
1935 else
19fdb624 1936 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1937
1938 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1939 osb->dev_str, nodestr);
ccd979bd
MF		 1940
1941 ocfs2_delete_osb(osb);
1942 kfree(osb);
1943 sb->s_dev = 0;
1944 sb->s_fs_info = NULL;
1945}
1946
1947static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1948 unsigned uuid_bytes)
1949{
1950 int i, ret;
1951 char *ptr;
1952
1953 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1954
cd861280 1955 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 1956 if (osb->uuid_str == NULL)
1957 return -ENOMEM;
1958
ccd979bd
MF		 1959 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1960 /* print with null */
1961 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1962 if (ret != 2) /* drop super cleans up */
1963 return -EINVAL;
1964 /* then only advance past the last char */
1965 ptr += 2;
1966 }
1967
1968 return 0;
1969}
1970
1971static int ocfs2_initialize_super(struct super_block *sb,
1972 struct buffer_head *bh,
73be192b
JB		 1973 int sector_size,
1974 struct ocfs2_blockcheck_stats *stats)
ccd979bd 1975{
bddb8eb3 1976 int status;
5a254031
MF		 1977 int i, cbits, bbits;
1978 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 1979 struct inode *inode = NULL;
ccd979bd
MF		 1980 struct ocfs2_journal *journal;
1981 __le32 uuid_net_key;
1982 struct ocfs2_super *osb;
1983
1984 mlog_entry_void();
1985
cd861280 1986 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 1987 if (!osb) {
1988 status = -ENOMEM;
1989 mlog_errno(status);
1990 goto bail;
1991 }
1992
1993 sb->s_fs_info = osb;
1994 sb->s_op = &ocfs2_sops;
1995 sb->s_export_op = &ocfs2_export_ops;
19ece546
JK		 1996 sb->s_qcop = &ocfs2_quotactl_ops;
1997 sb->dq_op = &ocfs2_quota_operations;
cf1d6c76 1998 sb->s_xattr = ocfs2_xattr_handlers;
e0dceaf0 1999 sb->s_time_gran = 1;
ccd979bd
MF		 2000 sb->s_flags |= MS_NOATIME;
2001 /* this is needed to support O_LARGEFILE */
5a254031
MF		 2002 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2003 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
2004 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd 2005
9b7895ef
MF		 2006 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
2007
2008 for (i = 0; i < 3; i++)
2009 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
2010 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2011
ccd979bd
MF		 2012 osb->sb = sb;
2013 /* Save off for ocfs2_rw_direct */
2014 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 2015 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 2016
34d024f8
MF		 2017 spin_lock_init(&osb->dc_task_lock);
2018 init_waitqueue_head(&osb->dc_event);
2019 osb->dc_work_sequence = 0;
2020 osb->dc_wake_sequence = 0;
ccd979bd
MF		 2021 INIT_LIST_HEAD(&osb->blocked_lock_list);
2022 osb->blocked_lock_count = 0;
ccd979bd 2023 spin_lock_init(&osb->osb_lock);
c8b9cf9a 2024 spin_lock_init(&osb->osb_xattr_lock);
b89c5428 2025 ocfs2_init_steal_slots(osb);
ccd979bd
MF		 2026
2027 atomic_set(&osb->alloc_stats.moves, 0);
2028 atomic_set(&osb->alloc_stats.local_data, 0);
2029 atomic_set(&osb->alloc_stats.bitmap_data, 0);
2030 atomic_set(&osb->alloc_stats.bg_allocs, 0);
2031 atomic_set(&osb->alloc_stats.bg_extends, 0);
2032
73be192b
JB		 2033 /* Copy the blockcheck stats from the superblock probe */
2034 osb->osb_ecc_stats = *stats;
2035
ccd979bd
MF		 2036 ocfs2_init_node_maps(osb);
2037
2038 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2039 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2040
8b712cd5
JM		 2041 ocfs2_orphan_scan_init(osb);
2042
553abd04
JB		 2043 status = ocfs2_recovery_init(osb);
2044 if (status) {
2045 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2046 mlog_errno(status);
2047 goto bail;
2048 }
ccd979bd
MF		 2049
2050 init_waitqueue_head(&osb->checkpoint_event);
2051 atomic_set(&osb->needs_checkpoint, 0);
2052
7f1a37e3
TY		 2053 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2054
ccd979bd
MF		 2055 osb->slot_num = OCFS2_INVALID_SLOT;
2056
8154da3d
TY		 2057 osb->s_xattr_inline_size = le16_to_cpu(
2058 di->id2.i_super.s_xattr_inline_size);
fdd77704 2059
ccd979bd
MF		 2060 osb->local_alloc_state = OCFS2_LA_UNUSED;
2061 osb->local_alloc_bh = NULL;
9c7af40b 2062 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
ccd979bd 2063
ccd979bd
MF		 2064 init_waitqueue_head(&osb->osb_mount_event);
2065
d02f00cc
MF		 2066 status = ocfs2_resmap_init(osb, &osb->osb_la_resmap);
2067 if (status) {
2068 mlog_errno(status);
2069 goto bail;
2070 }
2071
ccd979bd
MF		 2072 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2073 if (!osb->vol_label) {
2074 mlog(ML_ERROR, "unable to alloc vol label\n");
2075 status = -ENOMEM;
2076 goto bail;
2077 }
2078
ccd979bd
MF		 2079 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2080 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2081 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2082 osb->max_slots);
2083 status = -EINVAL;
2084 goto bail;
2085 }
781ee3e2 2086 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
ccd979bd 2087
539d8264
SM		 2088 osb->slot_recovery_generations =
2089 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2090 GFP_KERNEL);
2091 if (!osb->slot_recovery_generations) {
2092 status = -ENOMEM;
2093 mlog_errno(status);
2094 goto bail;
2095 }
2096
b4df6ed8
MF		 2097 init_waitqueue_head(&osb->osb_wipe_event);
2098 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2099 sizeof(*osb->osb_orphan_wipes),
2100 GFP_KERNEL);
2101 if (!osb->osb_orphan_wipes) {
2102 status = -ENOMEM;
2103 mlog_errno(status);
2104 goto bail;
2105 }
2106
374a263e
TM		 2107 osb->osb_rf_lock_tree = RB_ROOT;
2108
ccd979bd
MF		 2109 osb->s_feature_compat =
2110 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2111 osb->s_feature_ro_compat =
2112 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2113 osb->s_feature_incompat =
2114 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2115
2116 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2117 mlog(ML_ERROR, "couldn't mount because of unsupported "
2118 "optional features (%x).\n", i);
2119 status = -EINVAL;
2120 goto bail;
2121 }
2122 if (!(osb->sb->s_flags & MS_RDONLY) &&
2123 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2124 mlog(ML_ERROR, "couldn't mount RDWR because of "
2125 "unsupported optional features (%x).\n", i);
2126 status = -EINVAL;
2127 goto bail;
2128 }
2129
b61817e1
JB		 2130 if (ocfs2_userspace_stack(osb)) {
2131 memcpy(osb->osb_cluster_stack,
2132 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2133 OCFS2_STACK_LABEL_LEN);
2134 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2135 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2136 mlog(ML_ERROR,
2137 "couldn't mount because of an invalid "
2138 "cluster stack label (%s) \n",
2139 osb->osb_cluster_stack);
2140 status = -EINVAL;
2141 goto bail;
2142 }
2143 } else {
2144 /* The empty string is identical with classic tools that
2145 * don't know about s_cluster_info. */
2146 osb->osb_cluster_stack[0] = '\0';
2147 }
2148
ccd979bd
MF		 2149 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2150
2151 /* FIXME
2152 * This should be done in ocfs2_journal_init(), but unknown
2153 * ordering issues will cause the filesystem to crash.
2154 * If anyone wants to figure out what part of the code
2155 * refers to osb->journal before ocfs2_journal_init() is run,
2156 * be my guest.
2157 */
2158 /* initialize our journal structure */
2159
cd861280 2160 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 2161 if (!journal) {
2162 mlog(ML_ERROR, "unable to alloc journal\n");
2163 status = -ENOMEM;
2164 goto bail;
2165 }
2166 osb->journal = journal;
2167 journal->j_osb = osb;
2168
2169 atomic_set(&journal->j_num_trans, 0);
2170 init_rwsem(&journal->j_trans_barrier);
2171 init_waitqueue_head(&journal->j_checkpointed);
2172 spin_lock_init(&journal->j_lock);
2173 journal->j_trans_id = (unsigned long) 1;
2174 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 2175 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 2176 journal->j_state = OCFS2_JOURNAL_FREE;
2177
ea455f8a
JK		 2178 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2179 osb->dentry_lock_list = NULL;
2180
ccd979bd
MF		 2181 /* get some pseudo constants for clustersize bits */
2182 osb->s_clustersize_bits =
2183 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2184 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2185 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2186
2187 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2188 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2189 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2190 osb->s_clustersize);
2191 status = -EINVAL;
2192 goto bail;
2193 }
2194
2195 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2196 > (u32)~0UL) {
2197 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2198 "what jbd can address in 32 bits.\n");
2199 status = -EINVAL;
2200 goto bail;
2201 }
2202
2203 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2204 sizeof(di->id2.i_super.s_uuid))) {
2205 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2206 status = -ENOMEM;
2207 goto bail;
2208 }
2209
78427043 2210 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
ccd979bd
MF		 2211
2212 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2213 osb->vol_label[63] = '\0';
2214 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2215 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2216 osb->first_cluster_group_blkno =
2217 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2218 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
cf1d6c76 2219 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
ccd979bd
MF		 2220 mlog(0, "vol_label: %s\n", osb->vol_label);
2221 mlog(0, "uuid: %s\n", osb->uuid_str);
b0697053
MF		 2222 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2223 (unsigned long long)osb->root_blkno,
2224 (unsigned long long)osb->system_dir_blkno);
ccd979bd
MF		 2225
2226 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2227 if (!osb->osb_dlm_debug) {
2228 status = -ENOMEM;
2229 mlog_errno(status);
2230 goto bail;
2231 }
2232
2233 atomic_set(&osb->vol_state, VOLUME_INIT);
2234
2235 /* load root, system_dir, and all global system inodes */
2236 status = ocfs2_init_global_system_inodes(osb);
2237 if (status < 0) {
2238 mlog_errno(status);
2239 goto bail;
2240 }
2241
2242 /*
2243 * global bitmap
2244 */
2245 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2246 OCFS2_INVALID_SLOT);
2247 if (!inode) {
2248 status = -EINVAL;
2249 mlog_errno(status);
2250 goto bail;
2251 }
2252
2253 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
ccd979bd 2254 iput(inode);
ccd979bd 2255
e9d578a8 2256 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
ccd979bd
MF		 2257
2258 status = ocfs2_init_slot_info(osb);
2259 if (status < 0) {
2260 mlog_errno(status);
2261 goto bail;
2262 }
2263
ccd979bd
MF		 2264bail:
2265 mlog_exit(status);
2266 return status;
2267}
2268
2269/*
2270 * will return: -EAGAIN if it is ok to keep searching for superblocks
2271 * -EINVAL if there is a bad superblock
2272 * 0 on success
2273 */
2274static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2275 struct buffer_head *bh,
73be192b
JB		 2276 u32 blksz,
2277 struct ocfs2_blockcheck_stats *stats)
ccd979bd
MF		 2278{
2279 int status = -EAGAIN;
2280
2281 mlog_entry_void();
2282
2283 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2284 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
d030cc97
JB		 2285 /* We have to do a raw check of the feature here */
2286 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2287 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2288 status = ocfs2_block_check_validate(bh->b_data,
2289 bh->b_size,
73be192b
JB		 2290 &di->i_check,
2291 stats);
d030cc97
JB		 2292 if (status)
2293 goto out;
2294 }
ccd979bd
MF		 2295 status = -EINVAL;
2296 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2297 mlog(ML_ERROR, "found superblock with incorrect block "
2298 "size: found %u, should be %u\n",
2299 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2300 blksz);
2301 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2302 OCFS2_MAJOR_REV_LEVEL ||
2303 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2304 OCFS2_MINOR_REV_LEVEL) {
2305 mlog(ML_ERROR, "found superblock with bad version: "
2306 "found %u.%u, should be %u.%u\n",
2307 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2308 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2309 OCFS2_MAJOR_REV_LEVEL,
2310 OCFS2_MINOR_REV_LEVEL);
2311 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2312 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 2313 "found %llu, should be %llu\n",
1ca1a111 2314 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 2315 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 2316 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2317 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2318 mlog(ML_ERROR, "bad cluster size found: %u\n",
2319 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2320 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2321 mlog(ML_ERROR, "bad root_blkno: 0\n");
2322 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2323 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2324 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2325 mlog(ML_ERROR,
2326 "Superblock slots found greater than file system "
2327 "maximum: found %u, max %u\n",
2328 le16_to_cpu(di->id2.i_super.s_max_slots),
2329 OCFS2_MAX_SLOTS);
2330 } else {
2331 /* found it! */
2332 status = 0;
2333 }
2334 }
2335
d030cc97 2336out:
ccd979bd
MF		 2337 mlog_exit(status);
2338 return status;
2339}
2340
2341static int ocfs2_check_volume(struct ocfs2_super *osb)
2342{
bddb8eb3 2343 int status;
ccd979bd 2344 int dirty;
c271c5c2 2345 int local;
ccd979bd
MF		 2346 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2347 * recover
2348 * ourselves. */
2349
2350 mlog_entry_void();
2351
2352 /* Init our journal object. */
2353 status = ocfs2_journal_init(osb->journal, &dirty);
2354 if (status < 0) {
2355 mlog(ML_ERROR, "Could not initialize journal!\n");
2356 goto finally;
2357 }
2358
2359 /* If the journal was unmounted cleanly then we don't want to
2360 * recover anything. Otherwise, journal_load will do that
2361 * dirty work for us :) */
2362 if (!dirty) {
2363 status = ocfs2_journal_wipe(osb->journal, 0);
2364 if (status < 0) {
2365 mlog_errno(status);
2366 goto finally;
2367 }
2368 } else {
2369 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2370 "recovering volume.\n");
2371 }
2372
c271c5c2
SM		 2373 local = ocfs2_mount_local(osb);
2374
ccd979bd 2375 /* will play back anything left in the journal. */
539d8264 2376 status = ocfs2_journal_load(osb->journal, local, dirty);
01af4820
WW		 2377 if (status < 0) {
2378 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2379 goto finally;
2380 }
ccd979bd
MF		 2381
2382 if (dirty) {
2383 /* recover my local alloc if we didn't unmount cleanly. */
2384 status = ocfs2_begin_local_alloc_recovery(osb,
2385 osb->slot_num,
2386 &local_alloc);
2387 if (status < 0) {
2388 mlog_errno(status);
2389 goto finally;
2390 }
2391 /* we complete the recovery process after we've marked
2392 * ourselves as mounted. */
2393 }
2394
2395 mlog(0, "Journal loaded.\n");
2396
2397 status = ocfs2_load_local_alloc(osb);
2398 if (status < 0) {
2399 mlog_errno(status);
2400 goto finally;
2401 }
2402
2403 if (dirty) {
2404 /* Recovery will be completed after we've mounted the
2405 * rest of the volume. */
2406 osb->dirty = 1;
2407 osb->local_alloc_copy = local_alloc;
2408 local_alloc = NULL;
2409 }
2410
2411 /* go through each journal, trylock it and if you get the
2412 * lock, and it's marked as dirty, set the bit in the recover
2413 * map and launch a recovery thread for it. */
2414 status = ocfs2_mark_dead_nodes(osb);
9140db04
SE		 2415 if (status < 0) {
2416 mlog_errno(status);
2417 goto finally;
2418 }
2419
2420 status = ocfs2_compute_replay_slots(osb);
ccd979bd
MF		 2421 if (status < 0)
2422 mlog_errno(status);
2423
2424finally:
2425 if (local_alloc)
2426 kfree(local_alloc);
2427
2428 mlog_exit(status);
2429 return status;
2430}
2431
2432/*
2433 * The routine gets called from dismount or close whenever a dismount on
2434 * volume is requested and the osb open count becomes 1.
2435 * It will remove the osb from the global list and also free up all the
2436 * initialized resources and fileobject.
2437 */
2438static void ocfs2_delete_osb(struct ocfs2_super *osb)
2439{
2440 mlog_entry_void();
2441
2442 /* This function assumes that the caller has the main osb resource */
2443
8e8a4603 2444 ocfs2_free_slot_info(osb);
ccd979bd 2445
b4df6ed8 2446 kfree(osb->osb_orphan_wipes);
539d8264 2447 kfree(osb->slot_recovery_generations);
ccd979bd
MF		 2448 /* FIXME
2449 * This belongs in journal shutdown, but because we have to
2450 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2451 * we free it here.
2452 */
2453 kfree(osb->journal);
2454 if (osb->local_alloc_copy)
2455 kfree(osb->local_alloc_copy);
2456 kfree(osb->uuid_str);
2457 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2458 memset(osb, 0, sizeof(struct ocfs2_super));
2459
2460 mlog_exit_void();
2461}
2462
2463/* Put OCFS2 into a readonly state, or (if the user specifies it),
2464 * panic(). We do not support continue-on-error operation. */
2465static void ocfs2_handle_error(struct super_block *sb)
2466{
2467 struct ocfs2_super *osb = OCFS2_SB(sb);
2468
2469 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2470 panic("OCFS2: (device %s): panic forced after error\n",
2471 sb->s_id);
2472
2473 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2474
2475 if (sb->s_flags & MS_RDONLY &&
2476 (ocfs2_is_soft_readonly(osb) ||
2477 ocfs2_is_hard_readonly(osb)))
2478 return;
2479
2480 printk(KERN_CRIT "File system is now read-only due to the potential "
2481 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2482 "system is unmounted.\n");
2483 sb->s_flags |= MS_RDONLY;
2484 ocfs2_set_ro_flag(osb, 0);
2485}
2486
2487static char error_buf[1024];
2488
2489void __ocfs2_error(struct super_block *sb,
2490 const char *function,
2491 const char *fmt, ...)
2492{
2493 va_list args;
2494
2495 va_start(args, fmt);
4a6e617a 2496 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2497 va_end(args);
2498
2499 /* Not using mlog here because we want to show the actual
2500 * function the error came from. */
2501 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2502 sb->s_id, function, error_buf);
2503
2504 ocfs2_handle_error(sb);
2505}
2506
2507/* Handle critical errors. This is intentionally more drastic than
2508 * ocfs2_handle_error, so we only use for things like journal errors,
2509 * etc. */
2510void __ocfs2_abort(struct super_block* sb,
2511 const char *function,
2512 const char *fmt, ...)
2513{
2514 va_list args;
2515
2516 va_start(args, fmt);
4a6e617a 2517 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 2518 va_end(args);
2519
2520 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2521 sb->s_id, function, error_buf);
2522
2523 /* We don't have the cluster support yet to go straight to
2524 * hard readonly in here. Until then, we want to keep
2525 * ocfs2_abort() so that we can at least mark critical
2526 * errors.
2527 *
2528 * TODO: This should abort the journal and alert other nodes
2529 * that our slot needs recovery. */
2530
2531 /* Force a panic(). This stinks, but it's better than letting
2532 * things continue without having a proper hard readonly
2533 * here. */
a2f2ddbf
SM		 2534 if (!ocfs2_mount_local(OCFS2_SB(sb)))
2535 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 2536 ocfs2_handle_error(sb);
2537}
2538
2539module_init(ocfs2_init);
2540module_exit(ocfs2_exit);
Linux 6.x block layer and io_uring tree(s)