?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
PM / yenta: Split resume into early and late parts (rev. 4)
[linux/fpc-iii.git] / fs / ocfs2 / super.c
bloba3f8871d21fd516fb29aefbc040e30e6be016069
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>
42 #include <linux/mount.h>
43 #include <linux/seq_file.h>
44 #include <linux/quotaops.h>
45 #include <linux/smp_lock.h>
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"
56 #include "blockcheck.h"
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"
70 #include "xattr.h"
71 #include "quota.h"
72
73 #include "buffer_head_io.h"
74
75 static struct kmem_cache *ocfs2_inode_cachep = NULL;
76 struct kmem_cache *ocfs2_dquot_cachep;
77 struct kmem_cache *ocfs2_qf_chunk_cachep;
78
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. */
83 struct workqueue_struct *ocfs2_wq = NULL;
84
85 static struct dentry *ocfs2_debugfs_root = NULL;
86
87 MODULE_AUTHOR("Oracle");
88 MODULE_LICENSE("GPL");
89
90 struct mount_options
91 {
92 unsigned long commit_interval;
93 unsigned long mount_opt;
94 unsigned int atime_quantum;
95 signed short slot;
96 unsigned int localalloc_opt;
97 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
98 };
99
100 static int ocfs2_parse_options(struct super_block *sb, char *options,
101 struct mount_options *mopt,
102 int is_remount);
103 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
104 static void ocfs2_put_super(struct super_block *sb);
105 static int ocfs2_mount_volume(struct super_block *sb);
106 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
107 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
108 static int ocfs2_initialize_mem_caches(void);
109 static void ocfs2_free_mem_caches(void);
110 static void ocfs2_delete_osb(struct ocfs2_super *osb);
111
112 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
113
114 static int ocfs2_sync_fs(struct super_block *sb, int wait);
115
116 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
117 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
118 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
119 static int ocfs2_check_volume(struct ocfs2_super *osb);
120 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
121 struct buffer_head *bh,
122 u32 sectsize,
123 struct ocfs2_blockcheck_stats *stats);
124 static int ocfs2_initialize_super(struct super_block *sb,
125 struct buffer_head *bh,
126 int sector_size,
127 struct ocfs2_blockcheck_stats *stats);
128 static int ocfs2_get_sector(struct super_block *sb,
129 struct buffer_head **bh,
130 int block,
131 int sect_size);
132 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
133 static void ocfs2_destroy_inode(struct inode *inode);
134 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend);
135 static int ocfs2_enable_quotas(struct ocfs2_super *osb);
136 static void ocfs2_disable_quotas(struct ocfs2_super *osb);
137
138 static const struct super_operations ocfs2_sops = {
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,
146 .put_super = ocfs2_put_super,
147 .remount_fs = ocfs2_remount,
148 .show_options = ocfs2_show_options,
149 .quota_read = ocfs2_quota_read,
150 .quota_write = ocfs2_quota_write,
151 };
152
153 enum {
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,
163 Opt_atime_quantum,
164 Opt_slot,
165 Opt_commit,
166 Opt_localalloc,
167 Opt_localflocks,
168 Opt_stack,
169 Opt_user_xattr,
170 Opt_nouser_xattr,
171 Opt_inode64,
172 Opt_acl,
173 Opt_noacl,
174 Opt_usrquota,
175 Opt_grpquota,
176 Opt_err,
177 };
178
179 static const match_table_t tokens = {
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"},
189 {Opt_atime_quantum, "atime_quantum=%u"},
190 {Opt_slot, "preferred_slot=%u"},
191 {Opt_commit, "commit=%u"},
192 {Opt_localalloc, "localalloc=%d"},
193 {Opt_localflocks, "localflocks"},
194 {Opt_stack, "cluster_stack=%s"},
195 {Opt_user_xattr, "user_xattr"},
196 {Opt_nouser_xattr, "nouser_xattr"},
197 {Opt_inode64, "inode64"},
198 {Opt_acl, "acl"},
199 {Opt_noacl, "noacl"},
200 {Opt_usrquota, "usrquota"},
201 {Opt_grpquota, "grpquota"},
202 {Opt_err, NULL}
203 };
204
205 #ifdef CONFIG_DEBUG_FS
206 static int ocfs2_osb_dump(struct ocfs2_super *osb, char *buf, int len)
207 {
208 struct ocfs2_cluster_connection *cconn = osb->cconn;
209 struct ocfs2_recovery_map *rm = osb->recovery_map;
210 struct ocfs2_orphan_scan *os = &osb->osb_orphan_scan;
211 int i, out = 0;
212
213 out += snprintf(buf + out, len - out,
214 "%10s => Id: %-s Uuid: %-s Gen: 0x%X Label: %-s\n",
215 "Device", osb->dev_str, osb->uuid_str,
216 osb->fs_generation, osb->vol_label);
217
218 out += snprintf(buf + out, len - out,
219 "%10s => State: %d Flags: 0x%lX\n", "Volume",
220 atomic_read(&osb->vol_state), osb->osb_flags);
221
222 out += snprintf(buf + out, len - out,
223 "%10s => Block: %lu Cluster: %d\n", "Sizes",
224 osb->sb->s_blocksize, osb->s_clustersize);
225
226 out += snprintf(buf + out, len - out,
227 "%10s => Compat: 0x%X Incompat: 0x%X "
228 "ROcompat: 0x%X\n",
229 "Features", osb->s_feature_compat,
230 osb->s_feature_incompat, osb->s_feature_ro_compat);
231
232 out += snprintf(buf + out, len - out,
233 "%10s => Opts: 0x%lX AtimeQuanta: %u\n", "Mount",
234 osb->s_mount_opt, osb->s_atime_quantum);
235
236 if (cconn) {
237 out += snprintf(buf + out, len - out,
238 "%10s => Stack: %s Name: %*s "
239 "Version: %d.%d\n", "Cluster",
240 (*osb->osb_cluster_stack == '\0' ?
241 "o2cb" : osb->osb_cluster_stack),
242 cconn->cc_namelen, cconn->cc_name,
243 cconn->cc_version.pv_major,
244 cconn->cc_version.pv_minor);
245 }
246
247 spin_lock(&osb->dc_task_lock);
248 out += snprintf(buf + out, len - out,
249 "%10s => Pid: %d Count: %lu WakeSeq: %lu "
250 "WorkSeq: %lu\n", "DownCnvt",
251 (osb->dc_task ? task_pid_nr(osb->dc_task) : -1),
252 osb->blocked_lock_count, osb->dc_wake_sequence,
253 osb->dc_work_sequence);
254 spin_unlock(&osb->dc_task_lock);
255
256 spin_lock(&osb->osb_lock);
257 out += snprintf(buf + out, len - out, "%10s => Pid: %d Nodes:",
258 "Recovery",
259 (osb->recovery_thread_task ?
260 task_pid_nr(osb->recovery_thread_task) : -1));
261 if (rm->rm_used == 0)
262 out += snprintf(buf + out, len - out, " None\n");
263 else {
264 for (i = 0; i < rm->rm_used; i++)
265 out += snprintf(buf + out, len - out, " %d",
266 rm->rm_entries[i]);
267 out += snprintf(buf + out, len - out, "\n");
268 }
269 spin_unlock(&osb->osb_lock);
270
271 out += snprintf(buf + out, len - out,
272 "%10s => Pid: %d Interval: %lu Needs: %d\n", "Commit",
273 (osb->commit_task ? task_pid_nr(osb->commit_task) : -1),
274 osb->osb_commit_interval,
275 atomic_read(&osb->needs_checkpoint));
276
277 out += snprintf(buf + out, len - out,
278 "%10s => State: %d TxnId: %lu NumTxns: %d\n",
279 "Journal", osb->journal->j_state,
280 osb->journal->j_trans_id,
281 atomic_read(&osb->journal->j_num_trans));
282
283 out += snprintf(buf + out, len - out,
284 "%10s => GlobalAllocs: %d LocalAllocs: %d "
285 "SubAllocs: %d LAWinMoves: %d SAExtends: %d\n",
286 "Stats",
287 atomic_read(&osb->alloc_stats.bitmap_data),
288 atomic_read(&osb->alloc_stats.local_data),
289 atomic_read(&osb->alloc_stats.bg_allocs),
290 atomic_read(&osb->alloc_stats.moves),
291 atomic_read(&osb->alloc_stats.bg_extends));
292
293 out += snprintf(buf + out, len - out,
294 "%10s => State: %u Descriptor: %llu Size: %u bits "
295 "Default: %u bits\n",
296 "LocalAlloc", osb->local_alloc_state,
297 (unsigned long long)osb->la_last_gd,
298 osb->local_alloc_bits, osb->local_alloc_default_bits);
299
300 spin_lock(&osb->osb_lock);
301 out += snprintf(buf + out, len - out,
302 "%10s => Slot: %d NumStolen: %d\n", "Steal",
303 osb->s_inode_steal_slot,
304 atomic_read(&osb->s_num_inodes_stolen));
305 spin_unlock(&osb->osb_lock);
306
307 out += snprintf(buf + out, len - out, "OrphanScan => ");
308 out += snprintf(buf + out, len - out, "Local: %u Global: %u ",
309 os->os_count, os->os_seqno);
310 out += snprintf(buf + out, len - out, " Last Scan: ");
311 if (atomic_read(&os->os_state) == ORPHAN_SCAN_INACTIVE)
312 out += snprintf(buf + out, len - out, "Disabled\n");
313 else
314 out += snprintf(buf + out, len - out, "%lu seconds ago\n",
315 (get_seconds() - os->os_scantime.tv_sec));
316
317 out += snprintf(buf + out, len - out, "%10s => %3s %10s\n",
318 "Slots", "Num", "RecoGen");
319 for (i = 0; i < osb->max_slots; ++i) {
320 out += snprintf(buf + out, len - out,
321 "%10s %c %3d %10d\n",
322 " ",
323 (i == osb->slot_num ? '*' : ' '),
324 i, osb->slot_recovery_generations[i]);
325 }
326
327 return out;
328 }
329
330 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
331 {
332 struct ocfs2_super *osb = inode->i_private;
333 char *buf = NULL;
334
335 buf = kmalloc(PAGE_SIZE, GFP_KERNEL);
336 if (!buf)
337 goto bail;
338
339 i_size_write(inode, ocfs2_osb_dump(osb, buf, PAGE_SIZE));
340
341 file->private_data = buf;
342
343 return 0;
344 bail:
345 return -ENOMEM;
346 }
347
348 static int ocfs2_debug_release(struct inode *inode, struct file *file)
349 {
350 kfree(file->private_data);
351 return 0;
352 }
353
354 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
355 size_t nbytes, loff_t *ppos)
356 {
357 return simple_read_from_buffer(buf, nbytes, ppos, file->private_data,
358 i_size_read(file->f_mapping->host));
359 }
360 #else
361 static int ocfs2_osb_debug_open(struct inode *inode, struct file *file)
362 {
363 return 0;
364 }
365 static int ocfs2_debug_release(struct inode *inode, struct file *file)
366 {
367 return 0;
368 }
369 static ssize_t ocfs2_debug_read(struct file *file, char __user *buf,
370 size_t nbytes, loff_t *ppos)
371 {
372 return 0;
373 }
374 #endif /* CONFIG_DEBUG_FS */
375
376 static struct file_operations ocfs2_osb_debug_fops = {
377 .open = ocfs2_osb_debug_open,
378 .release = ocfs2_debug_release,
379 .read = ocfs2_debug_read,
380 .llseek = generic_file_llseek,
381 };
382
383 static int ocfs2_sync_fs(struct super_block *sb, int wait)
384 {
385 int status;
386 tid_t target;
387 struct ocfs2_super *osb = OCFS2_SB(sb);
388
389 if (ocfs2_is_hard_readonly(osb))
390 return -EROFS;
391
392 if (wait) {
393 status = ocfs2_flush_truncate_log(osb);
394 if (status < 0)
395 mlog_errno(status);
396 } else {
397 ocfs2_schedule_truncate_log_flush(osb, 0);
398 }
399
400 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
401 &target)) {
402 if (wait)
403 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
404 target);
405 }
406 return 0;
407 }
408
409 static int ocfs2_need_system_inode(struct ocfs2_super *osb, int ino)
410 {
411 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_USRQUOTA)
412 && (ino == USER_QUOTA_SYSTEM_INODE
413 || ino == LOCAL_USER_QUOTA_SYSTEM_INODE))
414 return 0;
415 if (!OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)
416 && (ino == GROUP_QUOTA_SYSTEM_INODE
417 || ino == LOCAL_GROUP_QUOTA_SYSTEM_INODE))
418 return 0;
419 return 1;
420 }
421
422 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
423 {
424 struct inode *new = NULL;
425 int status = 0;
426 int i;
427
428 mlog_entry_void();
429
430 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
431 if (IS_ERR(new)) {
432 status = PTR_ERR(new);
433 mlog_errno(status);
434 goto bail;
435 }
436 osb->root_inode = new;
437
438 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
439 if (IS_ERR(new)) {
440 status = PTR_ERR(new);
441 mlog_errno(status);
442 goto bail;
443 }
444 osb->sys_root_inode = new;
445
446 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
447 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
448 if (!ocfs2_need_system_inode(osb, i))
449 continue;
450 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
451 if (!new) {
452 ocfs2_release_system_inodes(osb);
453 status = -EINVAL;
454 mlog_errno(status);
455 /* FIXME: Should ERROR_RO_FS */
456 mlog(ML_ERROR, "Unable to load system inode %d, "
457 "possibly corrupt fs?", i);
458 goto bail;
459 }
460 // the array now has one ref, so drop this one
461 iput(new);
462 }
463
464 bail:
465 mlog_exit(status);
466 return status;
467 }
468
469 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
470 {
471 struct inode *new = NULL;
472 int status = 0;
473 int i;
474
475 mlog_entry_void();
476
477 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
478 i < NUM_SYSTEM_INODES;
479 i++) {
480 if (!ocfs2_need_system_inode(osb, i))
481 continue;
482 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
483 if (!new) {
484 ocfs2_release_system_inodes(osb);
485 status = -EINVAL;
486 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
487 status, i, osb->slot_num);
488 goto bail;
489 }
490 /* the array now has one ref, so drop this one */
491 iput(new);
492 }
493
494 bail:
495 mlog_exit(status);
496 return status;
497 }
498
499 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
500 {
501 int i;
502 struct inode *inode;
503
504 mlog_entry_void();
505
506 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
507 inode = osb->system_inodes[i];
508 if (inode) {
509 iput(inode);
510 osb->system_inodes[i] = NULL;
511 }
512 }
513
514 inode = osb->sys_root_inode;
515 if (inode) {
516 iput(inode);
517 osb->sys_root_inode = NULL;
518 }
519
520 inode = osb->root_inode;
521 if (inode) {
522 iput(inode);
523 osb->root_inode = NULL;
524 }
525
526 mlog_exit(0);
527 }
528
529 /* We're allocating fs objects, use GFP_NOFS */
530 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
531 {
532 struct ocfs2_inode_info *oi;
533
534 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
535 if (!oi)
536 return NULL;
537
538 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
539 return &oi->vfs_inode;
540 }
541
542 static void ocfs2_destroy_inode(struct inode *inode)
543 {
544 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
545 }
546
547 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
548 unsigned int cbits)
549 {
550 unsigned int bytes = 1 << cbits;
551 unsigned int trim = bytes;
552 unsigned int bitshift = 32;
553
554 /*
555 * i_size and all block offsets in ocfs2 are always 64 bits
556 * wide. i_clusters is 32 bits, in cluster-sized units. So on
557 * 64 bit platforms, cluster size will be the limiting factor.
558 */
559
560 #if BITS_PER_LONG == 32
561 # if defined(CONFIG_LBDAF)
562 BUILD_BUG_ON(sizeof(sector_t) != 8);
563 /*
564 * We might be limited by page cache size.
565 */
566 if (bytes > PAGE_CACHE_SIZE) {
567 bytes = PAGE_CACHE_SIZE;
568 trim = 1;
569 /*
570 * Shift by 31 here so that we don't get larger than
571 * MAX_LFS_FILESIZE
572 */
573 bitshift = 31;
574 }
575 # else
576 /*
577 * We are limited by the size of sector_t. Use block size, as
578 * that's what we expose to the VFS.
579 */
580 bytes = 1 << bbits;
581 trim = 1;
582 bitshift = 31;
583 # endif
584 #endif
585
586 /*
587 * Trim by a whole cluster when we can actually approach the
588 * on-disk limits. Otherwise we can overflow i_clusters when
589 * an extent start is at the max offset.
590 */
591 return (((unsigned long long)bytes) << bitshift) - trim;
592 }
593
594 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
595 {
596 int incompat_features;
597 int ret = 0;
598 struct mount_options parsed_options;
599 struct ocfs2_super *osb = OCFS2_SB(sb);
600
601 lock_kernel();
602
603 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
604 ret = -EINVAL;
605 goto out;
606 }
607
608 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
609 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
610 ret = -EINVAL;
611 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
612 goto out;
613 }
614
615 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
616 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
617 ret = -EINVAL;
618 mlog(ML_ERROR, "Cannot change data mode on remount\n");
619 goto out;
620 }
621
622 /* Probably don't want this on remount; it might
623 * mess with other nodes */
624 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
625 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
626 ret = -EINVAL;
627 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
628 goto out;
629 }
630
631 /* We're going to/from readonly mode. */
632 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
633 /* Disable quota accounting before remounting RO */
634 if (*flags & MS_RDONLY) {
635 ret = ocfs2_susp_quotas(osb, 0);
636 if (ret < 0)
637 goto out;
638 }
639 /* Lock here so the check of HARD_RO and the potential
640 * setting of SOFT_RO is atomic. */
641 spin_lock(&osb->osb_lock);
642 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
643 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
644 ret = -EROFS;
645 goto unlock_osb;
646 }
647
648 if (*flags & MS_RDONLY) {
649 mlog(0, "Going to ro mode.\n");
650 sb->s_flags |= MS_RDONLY;
651 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
652 } else {
653 mlog(0, "Making ro filesystem writeable.\n");
654
655 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
656 mlog(ML_ERROR, "Cannot remount RDWR "
657 "filesystem due to previous errors.\n");
658 ret = -EROFS;
659 goto unlock_osb;
660 }
661 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
662 if (incompat_features) {
663 mlog(ML_ERROR, "Cannot remount RDWR because "
664 "of unsupported optional features "
665 "(%x).\n", incompat_features);
666 ret = -EINVAL;
667 goto unlock_osb;
668 }
669 sb->s_flags &= ~MS_RDONLY;
670 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
671 }
672 unlock_osb:
673 spin_unlock(&osb->osb_lock);
674 /* Enable quota accounting after remounting RW */
675 if (!ret && !(*flags & MS_RDONLY)) {
676 if (sb_any_quota_suspended(sb))
677 ret = ocfs2_susp_quotas(osb, 1);
678 else
679 ret = ocfs2_enable_quotas(osb);
680 if (ret < 0) {
681 /* Return back changes... */
682 spin_lock(&osb->osb_lock);
683 sb->s_flags |= MS_RDONLY;
684 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
685 spin_unlock(&osb->osb_lock);
686 goto out;
687 }
688 }
689 }
690
691 if (!ret) {
692 /* Only save off the new mount options in case of a successful
693 * remount. */
694 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
695 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
696 osb->s_mount_opt = parsed_options.mount_opt;
697 osb->s_atime_quantum = parsed_options.atime_quantum;
698 osb->preferred_slot = parsed_options.slot;
699 if (parsed_options.commit_interval)
700 osb->osb_commit_interval = parsed_options.commit_interval;
701
702 if (!ocfs2_is_hard_readonly(osb))
703 ocfs2_set_journal_params(osb);
704 }
705 out:
706 unlock_kernel();
707 return ret;
708 }
709
710 static int ocfs2_sb_probe(struct super_block *sb,
711 struct buffer_head **bh,
712 int *sector_size,
713 struct ocfs2_blockcheck_stats *stats)
714 {
715 int status, tmpstat;
716 struct ocfs1_vol_disk_hdr *hdr;
717 struct ocfs2_dinode *di;
718 int blksize;
719
720 *bh = NULL;
721
722 /* may be > 512 */
723 *sector_size = bdev_logical_block_size(sb->s_bdev);
724 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
725 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
726 *sector_size, OCFS2_MAX_BLOCKSIZE);
727 status = -EINVAL;
728 goto bail;
729 }
730
731 /* Can this really happen? */
732 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
733 *sector_size = OCFS2_MIN_BLOCKSIZE;
734
735 /* check block zero for old format */
736 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
737 if (status < 0) {
738 mlog_errno(status);
739 goto bail;
740 }
741 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
742 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
743 mlog(ML_ERROR, "incompatible version: %u.%u\n",
744 hdr->major_version, hdr->minor_version);
745 status = -EINVAL;
746 }
747 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
748 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
749 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
750 hdr->signature);
751 status = -EINVAL;
752 }
753 brelse(*bh);
754 *bh = NULL;
755 if (status < 0) {
756 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
757 "upgraded before mounting with ocfs v2\n");
758 goto bail;
759 }
760
761 /*
762 * Now check at magic offset for 512, 1024, 2048, 4096
763 * blocksizes. 4096 is the maximum blocksize because it is
764 * the minimum clustersize.
765 */
766 status = -EINVAL;
767 for (blksize = *sector_size;
768 blksize <= OCFS2_MAX_BLOCKSIZE;
769 blksize <<= 1) {
770 tmpstat = ocfs2_get_sector(sb, bh,
771 OCFS2_SUPER_BLOCK_BLKNO,
772 blksize);
773 if (tmpstat < 0) {
774 status = tmpstat;
775 mlog_errno(status);
776 goto bail;
777 }
778 di = (struct ocfs2_dinode *) (*bh)->b_data;
779 memset(stats, 0, sizeof(struct ocfs2_blockcheck_stats));
780 spin_lock_init(&stats->b_lock);
781 status = ocfs2_verify_volume(di, *bh, blksize, stats);
782 if (status >= 0)
783 goto bail;
784 brelse(*bh);
785 *bh = NULL;
786 if (status != -EAGAIN)
787 break;
788 }
789
790 bail:
791 return status;
792 }
793
794 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
795 {
796 if (ocfs2_mount_local(osb)) {
797 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
798 mlog(ML_ERROR, "Cannot heartbeat on a locally "
799 "mounted device.\n");
800 return -EINVAL;
801 }
802 }
803
804 if (ocfs2_userspace_stack(osb)) {
805 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
806 mlog(ML_ERROR, "Userspace stack expected, but "
807 "o2cb heartbeat arguments passed to mount\n");
808 return -EINVAL;
809 }
810 }
811
812 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
813 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
814 !ocfs2_userspace_stack(osb)) {
815 mlog(ML_ERROR, "Heartbeat has to be started to mount "
816 "a read-write clustered device.\n");
817 return -EINVAL;
818 }
819 }
820
821 return 0;
822 }
823
824 /*
825 * If we're using a userspace stack, mount should have passed
826 * a name that matches the disk. If not, mount should not
827 * have passed a stack.
828 */
829 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
830 struct mount_options *mopt)
831 {
832 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
833 mlog(ML_ERROR,
834 "cluster stack passed to mount, but this filesystem "
835 "does not support it\n");
836 return -EINVAL;
837 }
838
839 if (ocfs2_userspace_stack(osb) &&
840 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
841 OCFS2_STACK_LABEL_LEN)) {
842 mlog(ML_ERROR,
843 "cluster stack passed to mount (\"%s\") does not "
844 "match the filesystem (\"%s\")\n",
845 mopt->cluster_stack,
846 osb->osb_cluster_stack);
847 return -EINVAL;
848 }
849
850 return 0;
851 }
852
853 static int ocfs2_susp_quotas(struct ocfs2_super *osb, int unsuspend)
854 {
855 int type;
856 struct super_block *sb = osb->sb;
857 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
858 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
859 int status = 0;
860
861 for (type = 0; type < MAXQUOTAS; type++) {
862 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
863 continue;
864 if (unsuspend)
865 status = vfs_quota_enable(
866 sb_dqopt(sb)->files[type],
867 type, QFMT_OCFS2,
868 DQUOT_SUSPENDED);
869 else
870 status = vfs_quota_disable(sb, type,
871 DQUOT_SUSPENDED);
872 if (status < 0)
873 break;
874 }
875 if (status < 0)
876 mlog(ML_ERROR, "Failed to suspend/unsuspend quotas on "
877 "remount (error = %d).\n", status);
878 return status;
879 }
880
881 static int ocfs2_enable_quotas(struct ocfs2_super *osb)
882 {
883 struct inode *inode[MAXQUOTAS] = { NULL, NULL };
884 struct super_block *sb = osb->sb;
885 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
886 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
887 unsigned int ino[MAXQUOTAS] = { LOCAL_USER_QUOTA_SYSTEM_INODE,
888 LOCAL_GROUP_QUOTA_SYSTEM_INODE };
889 int status;
890 int type;
891
892 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NEGATIVE_USAGE;
893 for (type = 0; type < MAXQUOTAS; type++) {
894 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
895 continue;
896 inode[type] = ocfs2_get_system_file_inode(osb, ino[type],
897 osb->slot_num);
898 if (!inode[type]) {
899 status = -ENOENT;
900 goto out_quota_off;
901 }
902 status = vfs_quota_enable(inode[type], type, QFMT_OCFS2,
903 DQUOT_USAGE_ENABLED);
904 if (status < 0)
905 goto out_quota_off;
906 }
907
908 for (type = 0; type < MAXQUOTAS; type++)
909 iput(inode[type]);
910 return 0;
911 out_quota_off:
912 ocfs2_disable_quotas(osb);
913 for (type = 0; type < MAXQUOTAS; type++)
914 iput(inode[type]);
915 mlog_errno(status);
916 return status;
917 }
918
919 static void ocfs2_disable_quotas(struct ocfs2_super *osb)
920 {
921 int type;
922 struct inode *inode;
923 struct super_block *sb = osb->sb;
924
925 /* We mostly ignore errors in this function because there's not much
926 * we can do when we see them */
927 for (type = 0; type < MAXQUOTAS; type++) {
928 if (!sb_has_quota_loaded(sb, type))
929 continue;
930 inode = igrab(sb->s_dquot.files[type]);
931 /* Turn off quotas. This will remove all dquot structures from
932 * memory and so they will be automatically synced to global
933 * quota files */
934 vfs_quota_disable(sb, type, DQUOT_USAGE_ENABLED |
935 DQUOT_LIMITS_ENABLED);
936 if (!inode)
937 continue;
938 iput(inode);
939 }
940 }
941
942 /* Handle quota on quotactl */
943 static int ocfs2_quota_on(struct super_block *sb, int type, int format_id,
944 char *path, int remount)
945 {
946 unsigned int feature[MAXQUOTAS] = { OCFS2_FEATURE_RO_COMPAT_USRQUOTA,
947 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA};
948
949 if (!OCFS2_HAS_RO_COMPAT_FEATURE(sb, feature[type]))
950 return -EINVAL;
951
952 if (remount)
953 return 0; /* Just ignore it has been handled in
954 * ocfs2_remount() */
955 return vfs_quota_enable(sb_dqopt(sb)->files[type], type,
956 format_id, DQUOT_LIMITS_ENABLED);
957 }
958
959 /* Handle quota off quotactl */
960 static int ocfs2_quota_off(struct super_block *sb, int type, int remount)
961 {
962 if (remount)
963 return 0; /* Ignore now and handle later in
964 * ocfs2_remount() */
965 return vfs_quota_disable(sb, type, DQUOT_LIMITS_ENABLED);
966 }
967
968 static struct quotactl_ops ocfs2_quotactl_ops = {
969 .quota_on = ocfs2_quota_on,
970 .quota_off = ocfs2_quota_off,
971 .quota_sync = vfs_quota_sync,
972 .get_info = vfs_get_dqinfo,
973 .set_info = vfs_set_dqinfo,
974 .get_dqblk = vfs_get_dqblk,
975 .set_dqblk = vfs_set_dqblk,
976 };
977
978 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
979 {
980 struct dentry *root;
981 int status, sector_size;
982 struct mount_options parsed_options;
983 struct inode *inode = NULL;
984 struct ocfs2_super *osb = NULL;
985 struct buffer_head *bh = NULL;
986 char nodestr[8];
987 struct ocfs2_blockcheck_stats stats;
988
989 mlog_entry("%p, %p, %i", sb, data, silent);
990
991 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
992 status = -EINVAL;
993 goto read_super_error;
994 }
995
996 /* probe for superblock */
997 status = ocfs2_sb_probe(sb, &bh, &sector_size, &stats);
998 if (status < 0) {
999 mlog(ML_ERROR, "superblock probe failed!\n");
1000 goto read_super_error;
1001 }
1002
1003 status = ocfs2_initialize_super(sb, bh, sector_size, &stats);
1004 osb = OCFS2_SB(sb);
1005 if (status < 0) {
1006 mlog_errno(status);
1007 goto read_super_error;
1008 }
1009 brelse(bh);
1010 bh = NULL;
1011
1012 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
1013 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1014
1015 osb->s_mount_opt = parsed_options.mount_opt;
1016 osb->s_atime_quantum = parsed_options.atime_quantum;
1017 osb->preferred_slot = parsed_options.slot;
1018 osb->osb_commit_interval = parsed_options.commit_interval;
1019 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
1020 osb->local_alloc_bits = osb->local_alloc_default_bits;
1021 if (osb->s_mount_opt & OCFS2_MOUNT_USRQUOTA &&
1022 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1023 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1024 status = -EINVAL;
1025 mlog(ML_ERROR, "User quotas were requested, but this "
1026 "filesystem does not have the feature enabled.\n");
1027 goto read_super_error;
1028 }
1029 if (osb->s_mount_opt & OCFS2_MOUNT_GRPQUOTA &&
1030 !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1031 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1032 status = -EINVAL;
1033 mlog(ML_ERROR, "Group quotas were requested, but this "
1034 "filesystem does not have the feature enabled.\n");
1035 goto read_super_error;
1036 }
1037
1038 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
1039 if (status)
1040 goto read_super_error;
1041
1042 sb->s_magic = OCFS2_SUPER_MAGIC;
1043
1044 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
1045 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
1046
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
1073 mlog_errno(status);
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)) {
1085 if (sb->s_flags & MS_RDONLY)
1086 ocfs2_set_ro_flag(osb, 0);
1087 }
1088
1089 status = ocfs2_verify_heartbeat(osb);
1090 if (status < 0) {
1091 mlog_errno(status);
1092 goto read_super_error;
1093 }
1094
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
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
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
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
1149 if (ocfs2_mount_local(osb))
1150 snprintf(nodestr, sizeof(nodestr), "local");
1151 else
1152 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1153
1154 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
1155 "with %s data mode.\n",
1156 osb->dev_str, nodestr, osb->slot_num,
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
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
1185 /* Start this when the mount is almost sure of being successful */
1186 ocfs2_orphan_scan_start(osb);
1187
1188 mlog_exit(status);
1189 return status;
1190
1191 read_super_error:
1192 brelse(bh);
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
1207 static 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)
1212 {
1213 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
1214 mnt);
1215 }
1216
1217 static void ocfs2_kill_sb(struct super_block *sb)
1218 {
1219 struct ocfs2_super *osb = OCFS2_SB(sb);
1220
1221 /* Failed mount? */
1222 if (!osb || atomic_read(&osb->vol_state) == VOLUME_DISABLED)
1223 goto out;
1224
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);
1231 out:
1232 kill_block_super(sb);
1233 }
1234
1235 static 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? */
1240 .kill_sb = ocfs2_kill_sb,
1241
1242 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
1243 .next = NULL
1244 };
1245
1246 static int ocfs2_parse_options(struct super_block *sb,
1247 char *options,
1248 struct mount_options *mopt,
1249 int is_remount)
1250 {
1251 int status;
1252 char *p;
1253
1254 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
1255 options ? options : "(none)");
1256
1257 mopt->commit_interval = 0;
1258 mopt->mount_opt = 0;
1259 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1260 mopt->slot = OCFS2_INVALID_SLOT;
1261 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
1262 mopt->cluster_stack[0] = '\0';
1263
1264 if (!options) {
1265 status = 1;
1266 goto bail;
1267 }
1268
1269 while ((p = strsep(&options, ",")) != NULL) {
1270 int token, option;
1271 substring_t args[MAX_OPT_ARGS];
1272
1273 if (!*p)
1274 continue;
1275
1276 token = match_token(p, tokens, args);
1277 switch (token) {
1278 case Opt_hb_local:
1279 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
1280 break;
1281 case Opt_hb_none:
1282 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
1283 break;
1284 case Opt_barrier:
1285 if (match_int(&args[0], &option)) {
1286 status = 0;
1287 goto bail;
1288 }
1289 if (option)
1290 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
1291 else
1292 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
1293 break;
1294 case Opt_intr:
1295 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
1296 break;
1297 case Opt_nointr:
1298 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
1299 break;
1300 case Opt_err_panic:
1301 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1302 break;
1303 case Opt_err_ro:
1304 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
1305 break;
1306 case Opt_data_ordered:
1307 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
1308 break;
1309 case Opt_data_writeback:
1310 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
1311 break;
1312 case Opt_user_xattr:
1313 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
1314 break;
1315 case Opt_nouser_xattr:
1316 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
1317 break;
1318 case Opt_atime_quantum:
1319 if (match_int(&args[0], &option)) {
1320 status = 0;
1321 goto bail;
1322 }
1323 if (option >= 0)
1324 mopt->atime_quantum = option;
1325 break;
1326 case Opt_slot:
1327 option = 0;
1328 if (match_int(&args[0], &option)) {
1329 status = 0;
1330 goto bail;
1331 }
1332 if (option)
1333 mopt->slot = (s16)option;
1334 break;
1335 case Opt_commit:
1336 option = 0;
1337 if (match_int(&args[0], &option)) {
1338 status = 0;
1339 goto bail;
1340 }
1341 if (option < 0)
1342 return 0;
1343 if (option == 0)
1344 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
1345 mopt->commit_interval = HZ * option;
1346 break;
1347 case Opt_localalloc:
1348 option = 0;
1349 if (match_int(&args[0], &option)) {
1350 status = 0;
1351 goto bail;
1352 }
1353 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
1354 mopt->localalloc_opt = option;
1355 break;
1356 case Opt_localflocks:
1357 /*
1358 * Changing this during remount could race
1359 * flock() requests, or "unbalance" existing
1360 * ones (e.g., a lock is taken in one mode but
1361 * dropped in the other). If users care enough
1362 * to flip locking modes during remount, we
1363 * could add a "local" flag to individual
1364 * flock structures for proper tracking of
1365 * state.
1366 */
1367 if (!is_remount)
1368 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
1369 break;
1370 case Opt_stack:
1371 /* Check both that the option we were passed
1372 * is of the right length and that it is a proper
1373 * string of the right length.
1374 */
1375 if (((args[0].to - args[0].from) !=
1376 OCFS2_STACK_LABEL_LEN) ||
1377 (strnlen(args[0].from,
1378 OCFS2_STACK_LABEL_LEN) !=
1379 OCFS2_STACK_LABEL_LEN)) {
1380 mlog(ML_ERROR,
1381 "Invalid cluster_stack option\n");
1382 status = 0;
1383 goto bail;
1384 }
1385 memcpy(mopt->cluster_stack, args[0].from,
1386 OCFS2_STACK_LABEL_LEN);
1387 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1388 break;
1389 case Opt_inode64:
1390 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
1391 break;
1392 case Opt_usrquota:
1393 /* We check only on remount, otherwise features
1394 * aren't yet initialized. */
1395 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1396 OCFS2_FEATURE_RO_COMPAT_USRQUOTA)) {
1397 mlog(ML_ERROR, "User quota requested but "
1398 "filesystem feature is not set\n");
1399 status = 0;
1400 goto bail;
1401 }
1402 mopt->mount_opt |= OCFS2_MOUNT_USRQUOTA;
1403 break;
1404 case Opt_grpquota:
1405 if (is_remount && !OCFS2_HAS_RO_COMPAT_FEATURE(sb,
1406 OCFS2_FEATURE_RO_COMPAT_GRPQUOTA)) {
1407 mlog(ML_ERROR, "Group quota requested but "
1408 "filesystem feature is not set\n");
1409 status = 0;
1410 goto bail;
1411 }
1412 mopt->mount_opt |= OCFS2_MOUNT_GRPQUOTA;
1413 break;
1414 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1415 case Opt_acl:
1416 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
1417 break;
1418 case Opt_noacl:
1419 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
1420 break;
1421 #else
1422 case Opt_acl:
1423 case Opt_noacl:
1424 printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
1425 break;
1426 #endif
1427 default:
1428 mlog(ML_ERROR,
1429 "Unrecognized mount option \"%s\" "
1430 "or missing value\n", p);
1431 status = 0;
1432 goto bail;
1433 }
1434 }
1435
1436 status = 1;
1437
1438 bail:
1439 mlog_exit(status);
1440 return status;
1441 }
1442
1443 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1444 {
1445 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
1446 unsigned long opts = osb->s_mount_opt;
1447 unsigned int local_alloc_megs;
1448
1449 if (opts & OCFS2_MOUNT_HB_LOCAL)
1450 seq_printf(s, ",_netdev,heartbeat=local");
1451 else
1452 seq_printf(s, ",heartbeat=none");
1453
1454 if (opts & OCFS2_MOUNT_NOINTR)
1455 seq_printf(s, ",nointr");
1456
1457 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1458 seq_printf(s, ",data=writeback");
1459 else
1460 seq_printf(s, ",data=ordered");
1461
1462 if (opts & OCFS2_MOUNT_BARRIER)
1463 seq_printf(s, ",barrier=1");
1464
1465 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1466 seq_printf(s, ",errors=panic");
1467 else
1468 seq_printf(s, ",errors=remount-ro");
1469
1470 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1471 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1472
1473 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1474 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1475
1476 if (osb->osb_commit_interval)
1477 seq_printf(s, ",commit=%u",
1478 (unsigned) (osb->osb_commit_interval / HZ));
1479
1480 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1481 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1482 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1483
1484 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1485 seq_printf(s, ",localflocks,");
1486
1487 if (osb->osb_cluster_stack[0])
1488 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1489 osb->osb_cluster_stack);
1490 if (opts & OCFS2_MOUNT_USRQUOTA)
1491 seq_printf(s, ",usrquota");
1492 if (opts & OCFS2_MOUNT_GRPQUOTA)
1493 seq_printf(s, ",grpquota");
1494
1495 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1496 seq_printf(s, ",nouser_xattr");
1497 else
1498 seq_printf(s, ",user_xattr");
1499
1500 if (opts & OCFS2_MOUNT_INODE64)
1501 seq_printf(s, ",inode64");
1502
1503 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1504 if (opts & OCFS2_MOUNT_POSIX_ACL)
1505 seq_printf(s, ",acl");
1506 else
1507 seq_printf(s, ",noacl");
1508 #endif
1509
1510 return 0;
1511 }
1512
1513 static int __init ocfs2_init(void)
1514 {
1515 int status;
1516
1517 mlog_entry_void();
1518
1519 ocfs2_print_version();
1520
1521 status = init_ocfs2_uptodate_cache();
1522 if (status < 0) {
1523 mlog_errno(status);
1524 goto leave;
1525 }
1526
1527 status = ocfs2_initialize_mem_caches();
1528 if (status < 0) {
1529 mlog_errno(status);
1530 goto leave;
1531 }
1532
1533 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1534 if (!ocfs2_wq) {
1535 status = -ENOMEM;
1536 goto leave;
1537 }
1538
1539 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1540 if (!ocfs2_debugfs_root) {
1541 status = -EFAULT;
1542 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1543 }
1544
1545 status = ocfs2_quota_setup();
1546 if (status)
1547 goto leave;
1548
1549 ocfs2_set_locking_protocol();
1550
1551 status = register_quota_format(&ocfs2_quota_format);
1552 leave:
1553 if (status < 0) {
1554 ocfs2_quota_shutdown();
1555 ocfs2_free_mem_caches();
1556 exit_ocfs2_uptodate_cache();
1557 }
1558
1559 mlog_exit(status);
1560
1561 if (status >= 0) {
1562 return register_filesystem(&ocfs2_fs_type);
1563 } else
1564 return -1;
1565 }
1566
1567 static void __exit ocfs2_exit(void)
1568 {
1569 mlog_entry_void();
1570
1571 ocfs2_quota_shutdown();
1572
1573 if (ocfs2_wq) {
1574 flush_workqueue(ocfs2_wq);
1575 destroy_workqueue(ocfs2_wq);
1576 }
1577
1578 unregister_quota_format(&ocfs2_quota_format);
1579
1580 debugfs_remove(ocfs2_debugfs_root);
1581
1582 ocfs2_free_mem_caches();
1583
1584 unregister_filesystem(&ocfs2_fs_type);
1585
1586 exit_ocfs2_uptodate_cache();
1587
1588 mlog_exit_void();
1589 }
1590
1591 static void ocfs2_put_super(struct super_block *sb)
1592 {
1593 mlog_entry("(0x%p)\n", sb);
1594
1595 lock_kernel();
1596
1597 ocfs2_sync_blockdev(sb);
1598 ocfs2_dismount_volume(sb, 0);
1599
1600 unlock_kernel();
1601
1602 mlog_exit_void();
1603 }
1604
1605 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1606 {
1607 struct ocfs2_super *osb;
1608 u32 numbits, freebits;
1609 int status;
1610 struct ocfs2_dinode *bm_lock;
1611 struct buffer_head *bh = NULL;
1612 struct inode *inode = NULL;
1613
1614 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1615
1616 osb = OCFS2_SB(dentry->d_sb);
1617
1618 inode = ocfs2_get_system_file_inode(osb,
1619 GLOBAL_BITMAP_SYSTEM_INODE,
1620 OCFS2_INVALID_SLOT);
1621 if (!inode) {
1622 mlog(ML_ERROR, "failed to get bitmap inode\n");
1623 status = -EIO;
1624 goto bail;
1625 }
1626
1627 status = ocfs2_inode_lock(inode, &bh, 0);
1628 if (status < 0) {
1629 mlog_errno(status);
1630 goto bail;
1631 }
1632
1633 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1634
1635 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1636 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1637
1638 buf->f_type = OCFS2_SUPER_MAGIC;
1639 buf->f_bsize = dentry->d_sb->s_blocksize;
1640 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1641 buf->f_blocks = ((sector_t) numbits) *
1642 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1643 buf->f_bfree = ((sector_t) freebits) *
1644 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1645 buf->f_bavail = buf->f_bfree;
1646 buf->f_files = numbits;
1647 buf->f_ffree = freebits;
1648
1649 brelse(bh);
1650
1651 ocfs2_inode_unlock(inode, 0);
1652 status = 0;
1653 bail:
1654 if (inode)
1655 iput(inode);
1656
1657 mlog_exit(status);
1658
1659 return status;
1660 }
1661
1662 static void ocfs2_inode_init_once(void *data)
1663 {
1664 struct ocfs2_inode_info *oi = data;
1665
1666 oi->ip_flags = 0;
1667 oi->ip_open_count = 0;
1668 spin_lock_init(&oi->ip_lock);
1669 ocfs2_extent_map_init(&oi->vfs_inode);
1670 INIT_LIST_HEAD(&oi->ip_io_markers);
1671 oi->ip_created_trans = 0;
1672 oi->ip_last_trans = 0;
1673 oi->ip_dir_start_lookup = 0;
1674
1675 init_rwsem(&oi->ip_alloc_sem);
1676 init_rwsem(&oi->ip_xattr_sem);
1677 mutex_init(&oi->ip_io_mutex);
1678
1679 oi->ip_blkno = 0ULL;
1680 oi->ip_clusters = 0;
1681
1682 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1683 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1684 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1685
1686 ocfs2_metadata_cache_init(&oi->vfs_inode);
1687
1688 inode_init_once(&oi->vfs_inode);
1689 }
1690
1691 static int ocfs2_initialize_mem_caches(void)
1692 {
1693 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1694 sizeof(struct ocfs2_inode_info),
1695 0,
1696 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1697 SLAB_MEM_SPREAD),
1698 ocfs2_inode_init_once);
1699 ocfs2_dquot_cachep = kmem_cache_create("ocfs2_dquot_cache",
1700 sizeof(struct ocfs2_dquot),
1701 0,
1702 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1703 SLAB_MEM_SPREAD),
1704 NULL);
1705 ocfs2_qf_chunk_cachep = kmem_cache_create("ocfs2_qf_chunk_cache",
1706 sizeof(struct ocfs2_quota_chunk),
1707 0,
1708 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD),
1709 NULL);
1710 if (!ocfs2_inode_cachep || !ocfs2_dquot_cachep ||
1711 !ocfs2_qf_chunk_cachep) {
1712 if (ocfs2_inode_cachep)
1713 kmem_cache_destroy(ocfs2_inode_cachep);
1714 if (ocfs2_dquot_cachep)
1715 kmem_cache_destroy(ocfs2_dquot_cachep);
1716 if (ocfs2_qf_chunk_cachep)
1717 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1718 return -ENOMEM;
1719 }
1720
1721 return 0;
1722 }
1723
1724 static void ocfs2_free_mem_caches(void)
1725 {
1726 if (ocfs2_inode_cachep)
1727 kmem_cache_destroy(ocfs2_inode_cachep);
1728 ocfs2_inode_cachep = NULL;
1729
1730 if (ocfs2_dquot_cachep)
1731 kmem_cache_destroy(ocfs2_dquot_cachep);
1732 ocfs2_dquot_cachep = NULL;
1733
1734 if (ocfs2_qf_chunk_cachep)
1735 kmem_cache_destroy(ocfs2_qf_chunk_cachep);
1736 ocfs2_qf_chunk_cachep = NULL;
1737 }
1738
1739 static int ocfs2_get_sector(struct super_block *sb,
1740 struct buffer_head **bh,
1741 int block,
1742 int sect_size)
1743 {
1744 if (!sb_set_blocksize(sb, sect_size)) {
1745 mlog(ML_ERROR, "unable to set blocksize\n");
1746 return -EIO;
1747 }
1748
1749 *bh = sb_getblk(sb, block);
1750 if (!*bh) {
1751 mlog_errno(-EIO);
1752 return -EIO;
1753 }
1754 lock_buffer(*bh);
1755 if (!buffer_dirty(*bh))
1756 clear_buffer_uptodate(*bh);
1757 unlock_buffer(*bh);
1758 ll_rw_block(READ, 1, bh);
1759 wait_on_buffer(*bh);
1760 if (!buffer_uptodate(*bh)) {
1761 mlog_errno(-EIO);
1762 brelse(*bh);
1763 *bh = NULL;
1764 return -EIO;
1765 }
1766
1767 return 0;
1768 }
1769
1770 static int ocfs2_mount_volume(struct super_block *sb)
1771 {
1772 int status = 0;
1773 int unlock_super = 0;
1774 struct ocfs2_super *osb = OCFS2_SB(sb);
1775
1776 mlog_entry_void();
1777
1778 if (ocfs2_is_hard_readonly(osb))
1779 goto leave;
1780
1781 status = ocfs2_dlm_init(osb);
1782 if (status < 0) {
1783 mlog_errno(status);
1784 goto leave;
1785 }
1786
1787 status = ocfs2_super_lock(osb, 1);
1788 if (status < 0) {
1789 mlog_errno(status);
1790 goto leave;
1791 }
1792 unlock_super = 1;
1793
1794 /* This will load up the node map and add ourselves to it. */
1795 status = ocfs2_find_slot(osb);
1796 if (status < 0) {
1797 mlog_errno(status);
1798 goto leave;
1799 }
1800
1801 /* load all node-local system inodes */
1802 status = ocfs2_init_local_system_inodes(osb);
1803 if (status < 0) {
1804 mlog_errno(status);
1805 goto leave;
1806 }
1807
1808 status = ocfs2_check_volume(osb);
1809 if (status < 0) {
1810 mlog_errno(status);
1811 goto leave;
1812 }
1813
1814 status = ocfs2_truncate_log_init(osb);
1815 if (status < 0)
1816 mlog_errno(status);
1817
1818 leave:
1819 if (unlock_super)
1820 ocfs2_super_unlock(osb, 1);
1821
1822 mlog_exit(status);
1823 return status;
1824 }
1825
1826 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1827 {
1828 int tmp, hangup_needed = 0;
1829 struct ocfs2_super *osb = NULL;
1830 char nodestr[8];
1831
1832 mlog_entry("(0x%p)\n", sb);
1833
1834 BUG_ON(!sb);
1835 osb = OCFS2_SB(sb);
1836 BUG_ON(!osb);
1837
1838 debugfs_remove(osb->osb_ctxt);
1839
1840 /*
1841 * Flush inode dropping work queue so that deletes are
1842 * performed while the filesystem is still working
1843 */
1844 ocfs2_drop_all_dl_inodes(osb);
1845
1846 /* Orphan scan should be stopped as early as possible */
1847 ocfs2_orphan_scan_stop(osb);
1848
1849 ocfs2_disable_quotas(osb);
1850
1851 ocfs2_shutdown_local_alloc(osb);
1852
1853 ocfs2_truncate_log_shutdown(osb);
1854
1855 /* This will disable recovery and flush any recovery work. */
1856 ocfs2_recovery_exit(osb);
1857
1858 ocfs2_journal_shutdown(osb);
1859
1860 ocfs2_sync_blockdev(sb);
1861
1862 /* No cluster connection means we've failed during mount, so skip
1863 * all the steps which depended on that to complete. */
1864 if (osb->cconn) {
1865 tmp = ocfs2_super_lock(osb, 1);
1866 if (tmp < 0) {
1867 mlog_errno(tmp);
1868 return;
1869 }
1870 }
1871
1872 if (osb->slot_num != OCFS2_INVALID_SLOT)
1873 ocfs2_put_slot(osb);
1874
1875 if (osb->cconn)
1876 ocfs2_super_unlock(osb, 1);
1877
1878 ocfs2_release_system_inodes(osb);
1879
1880 /*
1881 * If we're dismounting due to mount error, mount.ocfs2 will clean
1882 * up heartbeat. If we're a local mount, there is no heartbeat.
1883 * If we failed before we got a uuid_str yet, we can't stop
1884 * heartbeat. Otherwise, do it.
1885 */
1886 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1887 hangup_needed = 1;
1888
1889 if (osb->cconn)
1890 ocfs2_dlm_shutdown(osb, hangup_needed);
1891
1892 ocfs2_blockcheck_stats_debugfs_remove(&osb->osb_ecc_stats);
1893 debugfs_remove(osb->osb_debug_root);
1894
1895 if (hangup_needed)
1896 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1897
1898 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1899
1900 if (ocfs2_mount_local(osb))
1901 snprintf(nodestr, sizeof(nodestr), "local");
1902 else
1903 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1904
1905 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1906 osb->dev_str, nodestr);
1907
1908 ocfs2_delete_osb(osb);
1909 kfree(osb);
1910 sb->s_dev = 0;
1911 sb->s_fs_info = NULL;
1912 }
1913
1914 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1915 unsigned uuid_bytes)
1916 {
1917 int i, ret;
1918 char *ptr;
1919
1920 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1921
1922 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1923 if (osb->uuid_str == NULL)
1924 return -ENOMEM;
1925
1926 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1927 /* print with null */
1928 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1929 if (ret != 2) /* drop super cleans up */
1930 return -EINVAL;
1931 /* then only advance past the last char */
1932 ptr += 2;
1933 }
1934
1935 return 0;
1936 }
1937
1938 static int ocfs2_initialize_super(struct super_block *sb,
1939 struct buffer_head *bh,
1940 int sector_size,
1941 struct ocfs2_blockcheck_stats *stats)
1942 {
1943 int status;
1944 int i, cbits, bbits;
1945 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1946 struct inode *inode = NULL;
1947 struct ocfs2_journal *journal;
1948 __le32 uuid_net_key;
1949 struct ocfs2_super *osb;
1950
1951 mlog_entry_void();
1952
1953 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1954 if (!osb) {
1955 status = -ENOMEM;
1956 mlog_errno(status);
1957 goto bail;
1958 }
1959
1960 sb->s_fs_info = osb;
1961 sb->s_op = &ocfs2_sops;
1962 sb->s_export_op = &ocfs2_export_ops;
1963 sb->s_qcop = &ocfs2_quotactl_ops;
1964 sb->dq_op = &ocfs2_quota_operations;
1965 sb->s_xattr = ocfs2_xattr_handlers;
1966 sb->s_time_gran = 1;
1967 sb->s_flags |= MS_NOATIME;
1968 /* this is needed to support O_LARGEFILE */
1969 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1970 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1971 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1972
1973 osb->osb_dx_mask = (1 << (cbits - bbits)) - 1;
1974
1975 for (i = 0; i < 3; i++)
1976 osb->osb_dx_seed[i] = le32_to_cpu(di->id2.i_super.s_dx_seed[i]);
1977 osb->osb_dx_seed[3] = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1978
1979 osb->sb = sb;
1980 /* Save off for ocfs2_rw_direct */
1981 osb->s_sectsize_bits = blksize_bits(sector_size);
1982 BUG_ON(!osb->s_sectsize_bits);
1983
1984 spin_lock_init(&osb->dc_task_lock);
1985 init_waitqueue_head(&osb->dc_event);
1986 osb->dc_work_sequence = 0;
1987 osb->dc_wake_sequence = 0;
1988 INIT_LIST_HEAD(&osb->blocked_lock_list);
1989 osb->blocked_lock_count = 0;
1990 spin_lock_init(&osb->osb_lock);
1991 spin_lock_init(&osb->osb_xattr_lock);
1992 ocfs2_init_inode_steal_slot(osb);
1993
1994 atomic_set(&osb->alloc_stats.moves, 0);
1995 atomic_set(&osb->alloc_stats.local_data, 0);
1996 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1997 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1998 atomic_set(&osb->alloc_stats.bg_extends, 0);
1999
2000 /* Copy the blockcheck stats from the superblock probe */
2001 osb->osb_ecc_stats = *stats;
2002
2003 ocfs2_init_node_maps(osb);
2004
2005 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
2006 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
2007
2008 ocfs2_orphan_scan_init(osb);
2009
2010 status = ocfs2_recovery_init(osb);
2011 if (status) {
2012 mlog(ML_ERROR, "Unable to initialize recovery state\n");
2013 mlog_errno(status);
2014 goto bail;
2015 }
2016
2017 init_waitqueue_head(&osb->checkpoint_event);
2018 atomic_set(&osb->needs_checkpoint, 0);
2019
2020 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
2021
2022 osb->slot_num = OCFS2_INVALID_SLOT;
2023
2024 osb->s_xattr_inline_size = le16_to_cpu(
2025 di->id2.i_super.s_xattr_inline_size);
2026
2027 osb->local_alloc_state = OCFS2_LA_UNUSED;
2028 osb->local_alloc_bh = NULL;
2029 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
2030
2031 init_waitqueue_head(&osb->osb_mount_event);
2032
2033 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
2034 if (!osb->vol_label) {
2035 mlog(ML_ERROR, "unable to alloc vol label\n");
2036 status = -ENOMEM;
2037 goto bail;
2038 }
2039
2040 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
2041 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
2042 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
2043 osb->max_slots);
2044 status = -EINVAL;
2045 goto bail;
2046 }
2047 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
2048
2049 osb->slot_recovery_generations =
2050 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
2051 GFP_KERNEL);
2052 if (!osb->slot_recovery_generations) {
2053 status = -ENOMEM;
2054 mlog_errno(status);
2055 goto bail;
2056 }
2057
2058 init_waitqueue_head(&osb->osb_wipe_event);
2059 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
2060 sizeof(*osb->osb_orphan_wipes),
2061 GFP_KERNEL);
2062 if (!osb->osb_orphan_wipes) {
2063 status = -ENOMEM;
2064 mlog_errno(status);
2065 goto bail;
2066 }
2067
2068 osb->s_feature_compat =
2069 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
2070 osb->s_feature_ro_compat =
2071 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
2072 osb->s_feature_incompat =
2073 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
2074
2075 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
2076 mlog(ML_ERROR, "couldn't mount because of unsupported "
2077 "optional features (%x).\n", i);
2078 status = -EINVAL;
2079 goto bail;
2080 }
2081 if (!(osb->sb->s_flags & MS_RDONLY) &&
2082 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
2083 mlog(ML_ERROR, "couldn't mount RDWR because of "
2084 "unsupported optional features (%x).\n", i);
2085 status = -EINVAL;
2086 goto bail;
2087 }
2088
2089 if (ocfs2_userspace_stack(osb)) {
2090 memcpy(osb->osb_cluster_stack,
2091 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
2092 OCFS2_STACK_LABEL_LEN);
2093 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
2094 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
2095 mlog(ML_ERROR,
2096 "couldn't mount because of an invalid "
2097 "cluster stack label (%s) \n",
2098 osb->osb_cluster_stack);
2099 status = -EINVAL;
2100 goto bail;
2101 }
2102 } else {
2103 /* The empty string is identical with classic tools that
2104 * don't know about s_cluster_info. */
2105 osb->osb_cluster_stack[0] = '\0';
2106 }
2107
2108 get_random_bytes(&osb->s_next_generation, sizeof(u32));
2109
2110 /* FIXME
2111 * This should be done in ocfs2_journal_init(), but unknown
2112 * ordering issues will cause the filesystem to crash.
2113 * If anyone wants to figure out what part of the code
2114 * refers to osb->journal before ocfs2_journal_init() is run,
2115 * be my guest.
2116 */
2117 /* initialize our journal structure */
2118
2119 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
2120 if (!journal) {
2121 mlog(ML_ERROR, "unable to alloc journal\n");
2122 status = -ENOMEM;
2123 goto bail;
2124 }
2125 osb->journal = journal;
2126 journal->j_osb = osb;
2127
2128 atomic_set(&journal->j_num_trans, 0);
2129 init_rwsem(&journal->j_trans_barrier);
2130 init_waitqueue_head(&journal->j_checkpointed);
2131 spin_lock_init(&journal->j_lock);
2132 journal->j_trans_id = (unsigned long) 1;
2133 INIT_LIST_HEAD(&journal->j_la_cleanups);
2134 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
2135 journal->j_state = OCFS2_JOURNAL_FREE;
2136
2137 INIT_WORK(&osb->dentry_lock_work, ocfs2_drop_dl_inodes);
2138 osb->dentry_lock_list = NULL;
2139
2140 /* get some pseudo constants for clustersize bits */
2141 osb->s_clustersize_bits =
2142 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
2143 osb->s_clustersize = 1 << osb->s_clustersize_bits;
2144 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
2145
2146 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
2147 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
2148 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
2149 osb->s_clustersize);
2150 status = -EINVAL;
2151 goto bail;
2152 }
2153
2154 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
2155 > (u32)~0UL) {
2156 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
2157 "what jbd can address in 32 bits.\n");
2158 status = -EINVAL;
2159 goto bail;
2160 }
2161
2162 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
2163 sizeof(di->id2.i_super.s_uuid))) {
2164 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
2165 status = -ENOMEM;
2166 goto bail;
2167 }
2168
2169 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
2170
2171 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
2172 osb->vol_label[63] = '\0';
2173 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
2174 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
2175 osb->first_cluster_group_blkno =
2176 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
2177 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
2178 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
2179 mlog(0, "vol_label: %s\n", osb->vol_label);
2180 mlog(0, "uuid: %s\n", osb->uuid_str);
2181 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
2182 (unsigned long long)osb->root_blkno,
2183 (unsigned long long)osb->system_dir_blkno);
2184
2185 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
2186 if (!osb->osb_dlm_debug) {
2187 status = -ENOMEM;
2188 mlog_errno(status);
2189 goto bail;
2190 }
2191
2192 atomic_set(&osb->vol_state, VOLUME_INIT);
2193
2194 /* load root, system_dir, and all global system inodes */
2195 status = ocfs2_init_global_system_inodes(osb);
2196 if (status < 0) {
2197 mlog_errno(status);
2198 goto bail;
2199 }
2200
2201 /*
2202 * global bitmap
2203 */
2204 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
2205 OCFS2_INVALID_SLOT);
2206 if (!inode) {
2207 status = -EINVAL;
2208 mlog_errno(status);
2209 goto bail;
2210 }
2211
2212 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
2213 iput(inode);
2214
2215 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
2216
2217 status = ocfs2_init_slot_info(osb);
2218 if (status < 0) {
2219 mlog_errno(status);
2220 goto bail;
2221 }
2222
2223 bail:
2224 mlog_exit(status);
2225 return status;
2226 }
2227
2228 /*
2229 * will return: -EAGAIN if it is ok to keep searching for superblocks
2230 * -EINVAL if there is a bad superblock
2231 * 0 on success
2232 */
2233 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
2234 struct buffer_head *bh,
2235 u32 blksz,
2236 struct ocfs2_blockcheck_stats *stats)
2237 {
2238 int status = -EAGAIN;
2239
2240 mlog_entry_void();
2241
2242 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
2243 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
2244 /* We have to do a raw check of the feature here */
2245 if (le32_to_cpu(di->id2.i_super.s_feature_incompat) &
2246 OCFS2_FEATURE_INCOMPAT_META_ECC) {
2247 status = ocfs2_block_check_validate(bh->b_data,
2248 bh->b_size,
2249 &di->i_check,
2250 stats);
2251 if (status)
2252 goto out;
2253 }
2254 status = -EINVAL;
2255 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
2256 mlog(ML_ERROR, "found superblock with incorrect block "
2257 "size: found %u, should be %u\n",
2258 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
2259 blksz);
2260 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
2261 OCFS2_MAJOR_REV_LEVEL ||
2262 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
2263 OCFS2_MINOR_REV_LEVEL) {
2264 mlog(ML_ERROR, "found superblock with bad version: "
2265 "found %u.%u, should be %u.%u\n",
2266 le16_to_cpu(di->id2.i_super.s_major_rev_level),
2267 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
2268 OCFS2_MAJOR_REV_LEVEL,
2269 OCFS2_MINOR_REV_LEVEL);
2270 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
2271 mlog(ML_ERROR, "bad block number on superblock: "
2272 "found %llu, should be %llu\n",
2273 (unsigned long long)le64_to_cpu(di->i_blkno),
2274 (unsigned long long)bh->b_blocknr);
2275 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
2276 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
2277 mlog(ML_ERROR, "bad cluster size found: %u\n",
2278 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
2279 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
2280 mlog(ML_ERROR, "bad root_blkno: 0\n");
2281 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
2282 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
2283 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
2284 mlog(ML_ERROR,
2285 "Superblock slots found greater than file system "
2286 "maximum: found %u, max %u\n",
2287 le16_to_cpu(di->id2.i_super.s_max_slots),
2288 OCFS2_MAX_SLOTS);
2289 } else {
2290 /* found it! */
2291 status = 0;
2292 }
2293 }
2294
2295 out:
2296 mlog_exit(status);
2297 return status;
2298 }
2299
2300 static int ocfs2_check_volume(struct ocfs2_super *osb)
2301 {
2302 int status;
2303 int dirty;
2304 int local;
2305 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
2306 * recover
2307 * ourselves. */
2308
2309 mlog_entry_void();
2310
2311 /* Init our journal object. */
2312 status = ocfs2_journal_init(osb->journal, &dirty);
2313 if (status < 0) {
2314 mlog(ML_ERROR, "Could not initialize journal!\n");
2315 goto finally;
2316 }
2317
2318 /* If the journal was unmounted cleanly then we don't want to
2319 * recover anything. Otherwise, journal_load will do that
2320 * dirty work for us :) */
2321 if (!dirty) {
2322 status = ocfs2_journal_wipe(osb->journal, 0);
2323 if (status < 0) {
2324 mlog_errno(status);
2325 goto finally;
2326 }
2327 } else {
2328 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
2329 "recovering volume.\n");
2330 }
2331
2332 local = ocfs2_mount_local(osb);
2333
2334 /* will play back anything left in the journal. */
2335 status = ocfs2_journal_load(osb->journal, local, dirty);
2336 if (status < 0) {
2337 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
2338 goto finally;
2339 }
2340
2341 if (dirty) {
2342 /* recover my local alloc if we didn't unmount cleanly. */
2343 status = ocfs2_begin_local_alloc_recovery(osb,
2344 osb->slot_num,
2345 &local_alloc);
2346 if (status < 0) {
2347 mlog_errno(status);
2348 goto finally;
2349 }
2350 /* we complete the recovery process after we've marked
2351 * ourselves as mounted. */
2352 }
2353
2354 mlog(0, "Journal loaded.\n");
2355
2356 status = ocfs2_load_local_alloc(osb);
2357 if (status < 0) {
2358 mlog_errno(status);
2359 goto finally;
2360 }
2361
2362 if (dirty) {
2363 /* Recovery will be completed after we've mounted the
2364 * rest of the volume. */
2365 osb->dirty = 1;
2366 osb->local_alloc_copy = local_alloc;
2367 local_alloc = NULL;
2368 }
2369
2370 /* go through each journal, trylock it and if you get the
2371 * lock, and it's marked as dirty, set the bit in the recover
2372 * map and launch a recovery thread for it. */
2373 status = ocfs2_mark_dead_nodes(osb);
2374 if (status < 0) {
2375 mlog_errno(status);
2376 goto finally;
2377 }
2378
2379 status = ocfs2_compute_replay_slots(osb);
2380 if (status < 0)
2381 mlog_errno(status);
2382
2383 finally:
2384 if (local_alloc)
2385 kfree(local_alloc);
2386
2387 mlog_exit(status);
2388 return status;
2389 }
2390
2391 /*
2392 * The routine gets called from dismount or close whenever a dismount on
2393 * volume is requested and the osb open count becomes 1.
2394 * It will remove the osb from the global list and also free up all the
2395 * initialized resources and fileobject.
2396 */
2397 static void ocfs2_delete_osb(struct ocfs2_super *osb)
2398 {
2399 mlog_entry_void();
2400
2401 /* This function assumes that the caller has the main osb resource */
2402
2403 ocfs2_free_slot_info(osb);
2404
2405 kfree(osb->osb_orphan_wipes);
2406 kfree(osb->slot_recovery_generations);
2407 /* FIXME
2408 * This belongs in journal shutdown, but because we have to
2409 * allocate osb->journal at the start of ocfs2_initalize_osb(),
2410 * we free it here.
2411 */
2412 kfree(osb->journal);
2413 if (osb->local_alloc_copy)
2414 kfree(osb->local_alloc_copy);
2415 kfree(osb->uuid_str);
2416 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
2417 memset(osb, 0, sizeof(struct ocfs2_super));
2418
2419 mlog_exit_void();
2420 }
2421
2422 /* Put OCFS2 into a readonly state, or (if the user specifies it),
2423 * panic(). We do not support continue-on-error operation. */
2424 static void ocfs2_handle_error(struct super_block *sb)
2425 {
2426 struct ocfs2_super *osb = OCFS2_SB(sb);
2427
2428 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
2429 panic("OCFS2: (device %s): panic forced after error\n",
2430 sb->s_id);
2431
2432 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
2433
2434 if (sb->s_flags & MS_RDONLY &&
2435 (ocfs2_is_soft_readonly(osb) ||
2436 ocfs2_is_hard_readonly(osb)))
2437 return;
2438
2439 printk(KERN_CRIT "File system is now read-only due to the potential "
2440 "of on-disk corruption. Please run fsck.ocfs2 once the file "
2441 "system is unmounted.\n");
2442 sb->s_flags |= MS_RDONLY;
2443 ocfs2_set_ro_flag(osb, 0);
2444 }
2445
2446 static char error_buf[1024];
2447
2448 void __ocfs2_error(struct super_block *sb,
2449 const char *function,
2450 const char *fmt, ...)
2451 {
2452 va_list args;
2453
2454 va_start(args, fmt);
2455 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2456 va_end(args);
2457
2458 /* Not using mlog here because we want to show the actual
2459 * function the error came from. */
2460 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
2461 sb->s_id, function, error_buf);
2462
2463 ocfs2_handle_error(sb);
2464 }
2465
2466 /* Handle critical errors. This is intentionally more drastic than
2467 * ocfs2_handle_error, so we only use for things like journal errors,
2468 * etc. */
2469 void __ocfs2_abort(struct super_block* sb,
2470 const char *function,
2471 const char *fmt, ...)
2472 {
2473 va_list args;
2474
2475 va_start(args, fmt);
2476 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
2477 va_end(args);
2478
2479 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
2480 sb->s_id, function, error_buf);
2481
2482 /* We don't have the cluster support yet to go straight to
2483 * hard readonly in here. Until then, we want to keep
2484 * ocfs2_abort() so that we can at least mark critical
2485 * errors.
2486 *
2487 * TODO: This should abort the journal and alert other nodes
2488 * that our slot needs recovery. */
2489
2490 /* Force a panic(). This stinks, but it's better than letting
2491 * things continue without having a proper hard readonly
2492 * here. */
2493 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
2494 ocfs2_handle_error(sb);
2495 }
2496
2497 module_init(ocfs2_init);
2498 module_exit(ocfs2_exit);
Linux with added FPC-III support