projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: fix init of uuid_net_key
[linux-block.git] / fs / ocfs2 / super.c
CommitLineData
ccd979bd
MF		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * super.c
5 *
6 * load/unload driver, mount/dismount volumes
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/module.h>
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
31#include <linux/utsname.h>
32#include <linux/init.h>
33#include <linux/random.h>
34#include <linux/statfs.h>
35#include <linux/moduleparam.h>
36#include <linux/blkdev.h>
37#include <linux/socket.h>
38#include <linux/inet.h>
39#include <linux/parser.h>
40#include <linux/crc32.h>
41#include <linux/debugfs.h>
42
43#include <cluster/nodemanager.h>
44
45#define MLOG_MASK_PREFIX ML_SUPER
46#include <cluster/masklog.h>
47
48#include "ocfs2.h"
49
50/* this should be the only file to include a version 1 header */
51#include "ocfs1_fs_compat.h"
52
53#include "alloc.h"
54#include "dlmglue.h"
55#include "export.h"
56#include "extent_map.h"
57#include "heartbeat.h"
58#include "inode.h"
59#include "journal.h"
60#include "localalloc.h"
61#include "namei.h"
62#include "slot_map.h"
63#include "super.h"
64#include "sysfile.h"
65#include "uptodate.h"
66#include "ver.h"
67#include "vote.h"
68
69#include "buffer_head_io.h"
70
71/*
72 * Globals
73 */
74static spinlock_t ocfs2_globals_lock = SPIN_LOCK_UNLOCKED;
75
76static u32 osb_id; /* Keeps track of next available OSB Id */
77
78static kmem_cache_t *ocfs2_inode_cachep = NULL;
79
80kmem_cache_t *ocfs2_lock_cache = NULL;
81
82/* OCFS2 needs to schedule several differnt types of work which
83 * require cluster locking, disk I/O, recovery waits, etc. Since these
84 * types of work tend to be heavy we avoid using the kernel events
85 * workqueue and schedule on our own. */
86struct workqueue_struct *ocfs2_wq = NULL;
87
88static struct dentry *ocfs2_debugfs_root = NULL;
89
90MODULE_AUTHOR("Oracle");
91MODULE_LICENSE("GPL");
92
93static int ocfs2_parse_options(struct super_block *sb, char *options,
94 unsigned long *mount_opt, int is_remount);
95static void ocfs2_put_super(struct super_block *sb);
96static int ocfs2_mount_volume(struct super_block *sb);
97static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
98static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
99static int ocfs2_initialize_mem_caches(void);
100static void ocfs2_free_mem_caches(void);
101static void ocfs2_delete_osb(struct ocfs2_super *osb);
102
726c3342 103static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 104
105static int ocfs2_sync_fs(struct super_block *sb, int wait);
106
107static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
108static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
109static int ocfs2_release_system_inodes(struct ocfs2_super *osb);
110static int ocfs2_fill_local_node_info(struct ocfs2_super *osb);
111static int ocfs2_check_volume(struct ocfs2_super *osb);
112static int ocfs2_verify_volume(struct ocfs2_dinode *di,
113 struct buffer_head *bh,
114 u32 sectsize);
115static int ocfs2_initialize_super(struct super_block *sb,
116 struct buffer_head *bh,
117 int sector_size);
118static int ocfs2_get_sector(struct super_block *sb,
119 struct buffer_head **bh,
120 int block,
121 int sect_size);
122static void ocfs2_write_super(struct super_block *sb);
123static struct inode *ocfs2_alloc_inode(struct super_block *sb);
124static void ocfs2_destroy_inode(struct inode *inode);
125
126static unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
127
128static struct super_operations ocfs2_sops = {
129 .statfs = ocfs2_statfs,
130 .alloc_inode = ocfs2_alloc_inode,
131 .destroy_inode = ocfs2_destroy_inode,
132 .drop_inode = ocfs2_drop_inode,
133 .clear_inode = ocfs2_clear_inode,
134 .delete_inode = ocfs2_delete_inode,
135 .sync_fs = ocfs2_sync_fs,
136 .write_super = ocfs2_write_super,
137 .put_super = ocfs2_put_super,
138 .remount_fs = ocfs2_remount,
139};
140
141enum {
142 Opt_barrier,
143 Opt_err_panic,
144 Opt_err_ro,
145 Opt_intr,
146 Opt_nointr,
147 Opt_hb_none,
148 Opt_hb_local,
149 Opt_data_ordered,
150 Opt_data_writeback,
151 Opt_err,
152};
153
154static match_table_t tokens = {
155 {Opt_barrier, "barrier=%u"},
156 {Opt_err_panic, "errors=panic"},
157 {Opt_err_ro, "errors=remount-ro"},
158 {Opt_intr, "intr"},
159 {Opt_nointr, "nointr"},
160 {Opt_hb_none, OCFS2_HB_NONE},
161 {Opt_hb_local, OCFS2_HB_LOCAL},
162 {Opt_data_ordered, "data=ordered"},
163 {Opt_data_writeback, "data=writeback"},
164 {Opt_err, NULL}
165};
166
167/*
168 * write_super and sync_fs ripped right out of ext3.
169 */
170static void ocfs2_write_super(struct super_block *sb)
171{
7892f2f4 172 if (mutex_trylock(&sb->s_lock) != 0)
ccd979bd
MF		 173 BUG();
174 sb->s_dirt = 0;
175}
176
177static int ocfs2_sync_fs(struct super_block *sb, int wait)
178{
179 int status = 0;
180 tid_t target;
181 struct ocfs2_super *osb = OCFS2_SB(sb);
182
183 sb->s_dirt = 0;
184
185 if (ocfs2_is_hard_readonly(osb))
186 return -EROFS;
187
188 if (wait) {
189 status = ocfs2_flush_truncate_log(osb);
190 if (status < 0)
191 mlog_errno(status);
192 } else {
193 ocfs2_schedule_truncate_log_flush(osb, 0);
194 }
195
196 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
197 if (wait)
198 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
199 target);
200 }
201 return 0;
202}
203
204static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
205{
206 struct inode *new = NULL;
207 int status = 0;
208 int i;
209
210 mlog_entry_void();
211
212 new = ocfs2_iget(osb, osb->root_blkno);
213 if (IS_ERR(new)) {
214 status = PTR_ERR(new);
215 mlog_errno(status);
216 goto bail;
217 }
218 osb->root_inode = new;
219
220 new = ocfs2_iget(osb, osb->system_dir_blkno);
221 if (IS_ERR(new)) {
222 status = PTR_ERR(new);
223 mlog_errno(status);
224 goto bail;
225 }
226 osb->sys_root_inode = new;
227
228 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
229 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
230 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
231 if (!new) {
232 ocfs2_release_system_inodes(osb);
233 status = -EINVAL;
234 mlog_errno(status);
235 /* FIXME: Should ERROR_RO_FS */
236 mlog(ML_ERROR, "Unable to load system inode %d, "
237 "possibly corrupt fs?", i);
238 goto bail;
239 }
240 // the array now has one ref, so drop this one
241 iput(new);
242 }
243
244bail:
245 mlog_exit(status);
246 return status;
247}
248
249static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
250{
251 struct inode *new = NULL;
252 int status = 0;
253 int i;
254
255 mlog_entry_void();
256
257 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
258 i < NUM_SYSTEM_INODES;
259 i++) {
260 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
261 if (!new) {
262 ocfs2_release_system_inodes(osb);
263 status = -EINVAL;
264 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
265 status, i, osb->slot_num);
266 goto bail;
267 }
268 /* the array now has one ref, so drop this one */
269 iput(new);
270 }
271
272bail:
273 mlog_exit(status);
274 return status;
275}
276
277static int ocfs2_release_system_inodes(struct ocfs2_super *osb)
278{
279 int status = 0, i;
280 struct inode *inode;
281
282 mlog_entry_void();
283
284 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
285 inode = osb->system_inodes[i];
286 if (inode) {
287 iput(inode);
288 osb->system_inodes[i] = NULL;
289 }
290 }
291
292 inode = osb->sys_root_inode;
293 if (inode) {
294 iput(inode);
295 osb->sys_root_inode = NULL;
296 }
297
298 inode = osb->root_inode;
299 if (inode) {
300 iput(inode);
301 osb->root_inode = NULL;
302 }
303
304 mlog_exit(status);
305 return status;
306}
307
308/* We're allocating fs objects, use GFP_NOFS */
309static struct inode *ocfs2_alloc_inode(struct super_block *sb)
310{
311 struct ocfs2_inode_info *oi;
312
313 oi = kmem_cache_alloc(ocfs2_inode_cachep, SLAB_NOFS);
314 if (!oi)
315 return NULL;
316
317 return &oi->vfs_inode;
318}
319
320static void ocfs2_destroy_inode(struct inode *inode)
321{
322 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
323}
324
325/* From xfs_super.c:xfs_max_file_offset
326 * Copyright (c) 2000-2004 Silicon Graphics, Inc.
327 */
328static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
329{
330 unsigned int pagefactor = 1;
331 unsigned int bitshift = BITS_PER_LONG - 1;
332
333 /* Figure out maximum filesize, on Linux this can depend on
334 * the filesystem blocksize (on 32 bit platforms).
335 * __block_prepare_write does this in an [unsigned] long...
336 * page->index << (PAGE_CACHE_SHIFT - bbits)
337 * So, for page sized blocks (4K on 32 bit platforms),
338 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
339 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
340 * but for smaller blocksizes it is less (bbits = log2 bsize).
341 * Note1: get_block_t takes a long (implicit cast from above)
342 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
343 * can optionally convert the [unsigned] long from above into
344 * an [unsigned] long long.
345 */
346
347#if BITS_PER_LONG == 32
348# if defined(CONFIG_LBD)
349 BUG_ON(sizeof(sector_t) != 8);
350 pagefactor = PAGE_CACHE_SIZE;
351 bitshift = BITS_PER_LONG;
352# else
353 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
354# endif
355#endif
356
357 return (((unsigned long long)pagefactor) << bitshift) - 1;
358}
359
360static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
361{
362 int incompat_features;
363 int ret = 0;
364 unsigned long parsed_options;
365 struct ocfs2_super *osb = OCFS2_SB(sb);
366
367 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
368 ret = -EINVAL;
369 goto out;
370 }
371
372 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
373 (parsed_options & OCFS2_MOUNT_HB_LOCAL)) {
374 ret = -EINVAL;
375 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
376 goto out;
377 }
378
379 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
380 (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) {
381 ret = -EINVAL;
382 mlog(ML_ERROR, "Cannot change data mode on remount\n");
383 goto out;
384 }
385
386 /* We're going to/from readonly mode. */
387 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
388 /* Lock here so the check of HARD_RO and the potential
389 * setting of SOFT_RO is atomic. */
390 spin_lock(&osb->osb_lock);
391 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
392 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
393 ret = -EROFS;
394 goto unlock_osb;
395 }
396
397 if (*flags & MS_RDONLY) {
398 mlog(0, "Going to ro mode.\n");
399 sb->s_flags |= MS_RDONLY;
400 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
401 } else {
402 mlog(0, "Making ro filesystem writeable.\n");
403
404 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
405 mlog(ML_ERROR, "Cannot remount RDWR "
406 "filesystem due to previous errors.\n");
407 ret = -EROFS;
408 goto unlock_osb;
409 }
410 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
411 if (incompat_features) {
412 mlog(ML_ERROR, "Cannot remount RDWR because "
413 "of unsupported optional features "
414 "(%x).\n", incompat_features);
415 ret = -EINVAL;
416 goto unlock_osb;
417 }
418 sb->s_flags &= ~MS_RDONLY;
419 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
420 }
421unlock_osb:
422 spin_unlock(&osb->osb_lock);
423 }
424
425 if (!ret) {
426 if (!ocfs2_is_hard_readonly(osb))
427 ocfs2_set_journal_params(osb);
428
429 /* Only save off the new mount options in case of a successful
430 * remount. */
431 osb->s_mount_opt = parsed_options;
432 }
433out:
434 return ret;
435}
436
437static int ocfs2_sb_probe(struct super_block *sb,
438 struct buffer_head **bh,
439 int *sector_size)
440{
441 int status = 0, tmpstat;
442 struct ocfs1_vol_disk_hdr *hdr;
443 struct ocfs2_dinode *di;
444 int blksize;
445
446 *bh = NULL;
447
448 /* may be > 512 */
449 *sector_size = bdev_hardsect_size(sb->s_bdev);
450 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
451 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
452 *sector_size, OCFS2_MAX_BLOCKSIZE);
453 status = -EINVAL;
454 goto bail;
455 }
456
457 /* Can this really happen? */
458 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
459 *sector_size = OCFS2_MIN_BLOCKSIZE;
460
461 /* check block zero for old format */
462 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
463 if (status < 0) {
464 mlog_errno(status);
465 goto bail;
466 }
467 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
468 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
469 mlog(ML_ERROR, "incompatible version: %u.%u\n",
470 hdr->major_version, hdr->minor_version);
471 status = -EINVAL;
472 }
473 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
474 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
475 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
476 hdr->signature);
477 status = -EINVAL;
478 }
479 brelse(*bh);
480 *bh = NULL;
481 if (status < 0) {
482 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
483 "upgraded before mounting with ocfs v2\n");
484 goto bail;
485 }
486
487 /*
488 * Now check at magic offset for 512, 1024, 2048, 4096
489 * blocksizes. 4096 is the maximum blocksize because it is
490 * the minimum clustersize.
491 */
492 status = -EINVAL;
493 for (blksize = *sector_size;
494 blksize <= OCFS2_MAX_BLOCKSIZE;
495 blksize <<= 1) {
496 tmpstat = ocfs2_get_sector(sb, bh,
497 OCFS2_SUPER_BLOCK_BLKNO,
498 blksize);
499 if (tmpstat < 0) {
500 status = tmpstat;
501 mlog_errno(status);
502 goto bail;
503 }
504 di = (struct ocfs2_dinode *) (*bh)->b_data;
505 status = ocfs2_verify_volume(di, *bh, blksize);
506 if (status >= 0)
507 goto bail;
508 brelse(*bh);
509 *bh = NULL;
510 if (status != -EAGAIN)
511 break;
512 }
513
514bail:
515 return status;
516}
517
518static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
519{
520 struct dentry *root;
521 int status, sector_size;
522 unsigned long parsed_opt;
523 struct inode *inode = NULL;
524 struct ocfs2_super *osb = NULL;
525 struct buffer_head *bh = NULL;
526
527 mlog_entry("%p, %p, %i", sb, data, silent);
528
529 /* for now we only have one cluster/node, make sure we see it
530 * in the heartbeat universe */
531 if (!o2hb_check_local_node_heartbeating()) {
532 status = -EINVAL;
533 goto read_super_error;
534 }
535
536 /* probe for superblock */
537 status = ocfs2_sb_probe(sb, &bh, &sector_size);
538 if (status < 0) {
539 mlog(ML_ERROR, "superblock probe failed!\n");
540 goto read_super_error;
541 }
542
543 status = ocfs2_initialize_super(sb, bh, sector_size);
544 osb = OCFS2_SB(sb);
545 if (status < 0) {
546 mlog_errno(status);
547 goto read_super_error;
548 }
549 brelse(bh);
550 bh = NULL;
551
552 if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
553 status = -EINVAL;
554 goto read_super_error;
555 }
556 osb->s_mount_opt = parsed_opt;
557
558 sb->s_magic = OCFS2_SUPER_MAGIC;
559
560 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
561 * heartbeat=none */
562 if (bdev_read_only(sb->s_bdev)) {
563 if (!(sb->s_flags & MS_RDONLY)) {
564 status = -EACCES;
565 mlog(ML_ERROR, "Readonly device detected but readonly "
566 "mount was not specified.\n");
567 goto read_super_error;
568 }
569
570 /* You should not be able to start a local heartbeat
571 * on a readonly device. */
572 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
573 status = -EROFS;
574 mlog(ML_ERROR, "Local heartbeat specified on readonly "
575 "device.\n");
576 goto read_super_error;
577 }
578
579 status = ocfs2_check_journals_nolocks(osb);
580 if (status < 0) {
581 if (status == -EROFS)
582 mlog(ML_ERROR, "Recovery required on readonly "
583 "file system, but write access is "
584 "unavailable.\n");
585 else
586 mlog_errno(status);
587 goto read_super_error;
588 }
589
590 ocfs2_set_ro_flag(osb, 1);
591
592 printk(KERN_NOTICE "Readonly device detected. No cluster "
593 "services will be utilized for this mount. Recovery "
594 "will be skipped.\n");
595 }
596
597 if (!ocfs2_is_hard_readonly(osb)) {
598 /* If this isn't a hard readonly mount, then we need
599 * to make sure that heartbeat is in a valid state,
600 * and that we mark ourselves soft readonly is -oro
601 * was specified. */
602 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
603 mlog(ML_ERROR, "No heartbeat for device (%s)\n",
604 sb->s_id);
605 status = -EINVAL;
606 goto read_super_error;
607 }
608
609 if (sb->s_flags & MS_RDONLY)
610 ocfs2_set_ro_flag(osb, 0);
611 }
612
613 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
614 ocfs2_debugfs_root);
615 if (!osb->osb_debug_root) {
616 status = -EINVAL;
617 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
618 goto read_super_error;
619 }
620
621 status = ocfs2_mount_volume(sb);
622 if (osb->root_inode)
623 inode = igrab(osb->root_inode);
624
625 if (status < 0)
626 goto read_super_error;
627
628 if (!inode) {
629 status = -EIO;
630 mlog_errno(status);
631 goto read_super_error;
632 }
633
634 root = d_alloc_root(inode);
635 if (!root) {
636 status = -ENOMEM;
637 mlog_errno(status);
638 goto read_super_error;
639 }
640
641 sb->s_root = root;
642
643 ocfs2_complete_mount_recovery(osb);
644
781ee3e2
SM		 645 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %d, slot %d) "
646 "with %s data mode.\n",
647 osb->dev_str, osb->node_num, osb->slot_num,
ccd979bd
MF		 648 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
649 "ordered");
650
651 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
652 wake_up(&osb->osb_mount_event);
653
654 mlog_exit(status);
655 return status;
656
657read_super_error:
658 if (bh != NULL)
659 brelse(bh);
660
661 if (inode)
662 iput(inode);
663
664 if (osb) {
665 atomic_set(&osb->vol_state, VOLUME_DISABLED);
666 wake_up(&osb->osb_mount_event);
667 ocfs2_dismount_volume(sb, 1);
668 }
669
670 mlog_exit(status);
671 return status;
672}
673
454e2398
DH		 674static int ocfs2_get_sb(struct file_system_type *fs_type,
675 int flags,
676 const char *dev_name,
677 void *data,
678 struct vfsmount *mnt)
ccd979bd 679{
454e2398
DH		 680 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
681 mnt);
ccd979bd
MF		 682}
683
684static struct file_system_type ocfs2_fs_type = {
685 .owner = THIS_MODULE,
686 .name = "ocfs2",
687 .get_sb = ocfs2_get_sb, /* is this called when we mount
688 * the fs? */
689 .kill_sb = kill_block_super, /* set to the generic one
690 * right now, but do we
691 * need to change that? */
692 .fs_flags = FS_REQUIRES_DEV,
693 .next = NULL
694};
695
696static int ocfs2_parse_options(struct super_block *sb,
697 char *options,
698 unsigned long *mount_opt,
699 int is_remount)
700{
701 int status;
702 char *p;
703
704 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
705 options ? options : "(none)");
706
707 *mount_opt = 0;
708
709 if (!options) {
710 status = 1;
711 goto bail;
712 }
713
714 while ((p = strsep(&options, ",")) != NULL) {
715 int token, option;
716 substring_t args[MAX_OPT_ARGS];
717
718 if (!*p)
719 continue;
720
721 token = match_token(p, tokens, args);
722 switch (token) {
723 case Opt_hb_local:
724 *mount_opt |= OCFS2_MOUNT_HB_LOCAL;
725 break;
726 case Opt_hb_none:
727 *mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
728 break;
729 case Opt_barrier:
730 if (match_int(&args[0], &option)) {
731 status = 0;
732 goto bail;
733 }
734 if (option)
735 *mount_opt |= OCFS2_MOUNT_BARRIER;
736 else
737 *mount_opt &= ~OCFS2_MOUNT_BARRIER;
738 break;
739 case Opt_intr:
740 *mount_opt &= ~OCFS2_MOUNT_NOINTR;
741 break;
742 case Opt_nointr:
743 *mount_opt |= OCFS2_MOUNT_NOINTR;
744 break;
745 case Opt_err_panic:
746 *mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
747 break;
748 case Opt_err_ro:
749 *mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
750 break;
751 case Opt_data_ordered:
752 *mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
753 break;
754 case Opt_data_writeback:
755 *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
756 break;
757 default:
758 mlog(ML_ERROR,
759 "Unrecognized mount option \"%s\" "
760 "or missing value\n", p);
761 status = 0;
762 goto bail;
763 }
764 }
765
766 status = 1;
767
768bail:
769 mlog_exit(status);
770 return status;
771}
772
773static int __init ocfs2_init(void)
774{
775 int status;
776
777 mlog_entry_void();
778
779 ocfs2_print_version();
780
781 if (init_ocfs2_extent_maps())
782 return -ENOMEM;
783
784 status = init_ocfs2_uptodate_cache();
785 if (status < 0) {
786 mlog_errno(status);
787 goto leave;
788 }
789
790 status = ocfs2_initialize_mem_caches();
791 if (status < 0) {
792 mlog_errno(status);
793 goto leave;
794 }
795
796 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
797 if (!ocfs2_wq) {
798 status = -ENOMEM;
799 goto leave;
800 }
801
802 spin_lock(&ocfs2_globals_lock);
803 osb_id = 0;
804 spin_unlock(&ocfs2_globals_lock);
805
806 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
807 if (!ocfs2_debugfs_root) {
808 status = -EFAULT;
809 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
810 }
811
812leave:
813 if (status < 0) {
814 ocfs2_free_mem_caches();
815 exit_ocfs2_uptodate_cache();
816 exit_ocfs2_extent_maps();
817 }
818
819 mlog_exit(status);
820
821 if (status >= 0) {
822 return register_filesystem(&ocfs2_fs_type);
823 } else
824 return -1;
825}
826
827static void __exit ocfs2_exit(void)
828{
829 mlog_entry_void();
830
831 if (ocfs2_wq) {
832 flush_workqueue(ocfs2_wq);
833 destroy_workqueue(ocfs2_wq);
834 }
835
836 debugfs_remove(ocfs2_debugfs_root);
837
838 ocfs2_free_mem_caches();
839
840 unregister_filesystem(&ocfs2_fs_type);
841
842 exit_ocfs2_extent_maps();
843
844 exit_ocfs2_uptodate_cache();
845
846 mlog_exit_void();
847}
848
849static void ocfs2_put_super(struct super_block *sb)
850{
851 mlog_entry("(0x%p)\n", sb);
852
853 ocfs2_sync_blockdev(sb);
854 ocfs2_dismount_volume(sb, 0);
855
856 mlog_exit_void();
857}
858
726c3342 859static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 860{
861 struct ocfs2_super *osb;
862 u32 numbits, freebits;
863 int status;
864 struct ocfs2_dinode *bm_lock;
865 struct buffer_head *bh = NULL;
866 struct inode *inode = NULL;
867
726c3342 868 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
ccd979bd 869
726c3342 870 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 871
872 inode = ocfs2_get_system_file_inode(osb,
873 GLOBAL_BITMAP_SYSTEM_INODE,
874 OCFS2_INVALID_SLOT);
875 if (!inode) {
876 mlog(ML_ERROR, "failed to get bitmap inode\n");
877 status = -EIO;
878 goto bail;
879 }
880
881 status = ocfs2_meta_lock(inode, NULL, &bh, 0);
882 if (status < 0) {
883 mlog_errno(status);
884 goto bail;
885 }
886
887 bm_lock = (struct ocfs2_dinode *) bh->b_data;
888
889 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
890 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
891
892 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 893 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 894 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
895 buf->f_blocks = ((sector_t) numbits) *
896 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
897 buf->f_bfree = ((sector_t) freebits) *
898 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
899 buf->f_bavail = buf->f_bfree;
900 buf->f_files = numbits;
901 buf->f_ffree = freebits;
902
903 brelse(bh);
904
905 ocfs2_meta_unlock(inode, 0);
906 status = 0;
907bail:
908 if (inode)
909 iput(inode);
910
911 mlog_exit(status);
912
913 return status;
914}
915
916static void ocfs2_inode_init_once(void *data,
917 kmem_cache_t *cachep,
918 unsigned long flags)
919{
920 struct ocfs2_inode_info *oi = data;
921
922 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
923 SLAB_CTOR_CONSTRUCTOR) {
924 oi->ip_flags = 0;
925 oi->ip_open_count = 0;
926 spin_lock_init(&oi->ip_lock);
927 ocfs2_extent_map_init(&oi->vfs_inode);
928 INIT_LIST_HEAD(&oi->ip_handle_list);
929 INIT_LIST_HEAD(&oi->ip_io_markers);
930 oi->ip_handle = NULL;
931 oi->ip_created_trans = 0;
932 oi->ip_last_trans = 0;
933 oi->ip_dir_start_lookup = 0;
934
935 init_rwsem(&oi->ip_alloc_sem);
251b6ecc 936 mutex_init(&oi->ip_io_mutex);
ccd979bd
MF		 937
938 oi->ip_blkno = 0ULL;
939 oi->ip_clusters = 0;
940
941 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
942 ocfs2_lock_res_init_once(&oi->ip_meta_lockres);
943 ocfs2_lock_res_init_once(&oi->ip_data_lockres);
944
945 ocfs2_metadata_cache_init(&oi->vfs_inode);
946
947 inode_init_once(&oi->vfs_inode);
948 }
949}
950
951static int ocfs2_initialize_mem_caches(void)
952{
953 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 954 sizeof(struct ocfs2_inode_info),
955 0,
956 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
957 SLAB_MEM_SPREAD),
958 ocfs2_inode_init_once, NULL);
ccd979bd
MF		 959 if (!ocfs2_inode_cachep)
960 return -ENOMEM;
961
962 ocfs2_lock_cache = kmem_cache_create("ocfs2_lock",
963 sizeof(struct ocfs2_journal_lock),
964 0,
ac2b898c 965 SLAB_HWCACHE_ALIGN,
ccd979bd
MF		 966 NULL, NULL);
967 if (!ocfs2_lock_cache)
968 return -ENOMEM;
969
970 return 0;
971}
972
973static void ocfs2_free_mem_caches(void)
974{
975 if (ocfs2_inode_cachep)
976 kmem_cache_destroy(ocfs2_inode_cachep);
977 if (ocfs2_lock_cache)
978 kmem_cache_destroy(ocfs2_lock_cache);
979
980 ocfs2_inode_cachep = NULL;
981 ocfs2_lock_cache = NULL;
982}
983
984static int ocfs2_get_sector(struct super_block *sb,
985 struct buffer_head **bh,
986 int block,
987 int sect_size)
988{
989 if (!sb_set_blocksize(sb, sect_size)) {
990 mlog(ML_ERROR, "unable to set blocksize\n");
991 return -EIO;
992 }
993
994 *bh = sb_getblk(sb, block);
995 if (!*bh) {
996 mlog_errno(-EIO);
997 return -EIO;
998 }
999 lock_buffer(*bh);
1000 if (!buffer_dirty(*bh))
1001 clear_buffer_uptodate(*bh);
1002 unlock_buffer(*bh);
1003 ll_rw_block(READ, 1, bh);
1004 wait_on_buffer(*bh);
1005 return 0;
1006}
1007
1008/* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
1009static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
1010{
1011 int status;
1012
1013 /* XXX hold a ref on the node while mounte? easy enough, if
1014 * desirable. */
1015 osb->node_num = o2nm_this_node();
1016 if (osb->node_num == O2NM_MAX_NODES) {
1017 mlog(ML_ERROR, "could not find this host's node number\n");
1018 status = -ENOENT;
1019 goto bail;
1020 }
1021
781ee3e2 1022 mlog(0, "I am node %d\n", osb->node_num);
ccd979bd
MF		 1023
1024 status = 0;
1025bail:
1026 return status;
1027}
1028
1029static int ocfs2_mount_volume(struct super_block *sb)
1030{
1031 int status = 0;
1032 int unlock_super = 0;
1033 struct ocfs2_super *osb = OCFS2_SB(sb);
1034
1035 mlog_entry_void();
1036
1037 if (ocfs2_is_hard_readonly(osb))
1038 goto leave;
1039
1040 status = ocfs2_fill_local_node_info(osb);
1041 if (status < 0) {
1042 mlog_errno(status);
1043 goto leave;
1044 }
1045
1046 status = ocfs2_register_hb_callbacks(osb);
1047 if (status < 0) {
1048 mlog_errno(status);
1049 goto leave;
1050 }
1051
1052 status = ocfs2_dlm_init(osb);
1053 if (status < 0) {
1054 mlog_errno(status);
1055 goto leave;
1056 }
1057
1058 /* requires vote_thread to be running. */
1059 status = ocfs2_register_net_handlers(osb);
1060 if (status < 0) {
1061 mlog_errno(status);
1062 goto leave;
1063 }
1064
1065 status = ocfs2_super_lock(osb, 1);
1066 if (status < 0) {
1067 mlog_errno(status);
1068 goto leave;
1069 }
1070 unlock_super = 1;
1071
1072 /* This will load up the node map and add ourselves to it. */
1073 status = ocfs2_find_slot(osb);
1074 if (status < 0) {
1075 mlog_errno(status);
1076 goto leave;
1077 }
1078
1079 ocfs2_populate_mounted_map(osb);
1080
1081 /* load all node-local system inodes */
1082 status = ocfs2_init_local_system_inodes(osb);
1083 if (status < 0) {
1084 mlog_errno(status);
1085 goto leave;
1086 }
1087
1088 status = ocfs2_check_volume(osb);
1089 if (status < 0) {
1090 mlog_errno(status);
1091 goto leave;
1092 }
1093
1094 status = ocfs2_truncate_log_init(osb);
1095 if (status < 0) {
1096 mlog_errno(status);
1097 goto leave;
1098 }
1099
1100 /* This should be sent *after* we recovered our journal as it
1101 * will cause other nodes to unmark us as needing
1102 * recovery. However, we need to send it *before* dropping the
1103 * super block lock as otherwise their recovery threads might
1104 * try to clean us up while we're live! */
1105 status = ocfs2_request_mount_vote(osb);
1106 if (status < 0)
1107 mlog_errno(status);
1108
1109leave:
1110 if (unlock_super)
1111 ocfs2_super_unlock(osb, 1);
1112
1113 mlog_exit(status);
1114 return status;
1115}
1116
1117/* we can't grab the goofy sem lock from inside wait_event, so we use
1118 * memory barriers to make sure that we'll see the null task before
1119 * being woken up */
1120static int ocfs2_recovery_thread_running(struct ocfs2_super *osb)
1121{
1122 mb();
1123 return osb->recovery_thread_task != NULL;
1124}
1125
1126static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1127{
1128 int tmp;
1129 struct ocfs2_super *osb = NULL;
1130
1131 mlog_entry("(0x%p)\n", sb);
1132
1133 BUG_ON(!sb);
1134 osb = OCFS2_SB(sb);
1135 BUG_ON(!osb);
1136
1137 ocfs2_shutdown_local_alloc(osb);
1138
1139 ocfs2_truncate_log_shutdown(osb);
1140
1141 /* disable any new recovery threads and wait for any currently
1142 * running ones to exit. Do this before setting the vol_state. */
c74ec2f7 1143 mutex_lock(&osb->recovery_lock);
ccd979bd 1144 osb->disable_recovery = 1;
c74ec2f7 1145 mutex_unlock(&osb->recovery_lock);
ccd979bd
MF		 1146 wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb));
1147
1148 /* At this point, we know that no more recovery threads can be
1149 * launched, so wait for any recovery completion work to
1150 * complete. */
1151 flush_workqueue(ocfs2_wq);
1152
1153 ocfs2_journal_shutdown(osb);
1154
1155 ocfs2_sync_blockdev(sb);
1156
1157 /* No dlm means we've failed during mount, so skip all the
1158 * steps which depended on that to complete. */
1159 if (osb->dlm) {
1160 tmp = ocfs2_super_lock(osb, 1);
1161 if (tmp < 0) {
1162 mlog_errno(tmp);
1163 return;
1164 }
1165
1166 tmp = ocfs2_request_umount_vote(osb);
1167 if (tmp < 0)
1168 mlog_errno(tmp);
1169
1170 if (osb->slot_num != OCFS2_INVALID_SLOT)
1171 ocfs2_put_slot(osb);
1172
1173 ocfs2_super_unlock(osb, 1);
1174 }
1175
1176 ocfs2_release_system_inodes(osb);
1177
1178 if (osb->dlm) {
1179 ocfs2_unregister_net_handlers(osb);
1180
1181 ocfs2_dlm_shutdown(osb);
1182 }
1183
1184 ocfs2_clear_hb_callbacks(osb);
1185
1186 debugfs_remove(osb->osb_debug_root);
1187
1188 if (!mnt_err)
1189 ocfs2_stop_heartbeat(osb);
1190
1191 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1192
781ee3e2
SM		 1193 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %d)\n",
1194 osb->dev_str, osb->node_num);
ccd979bd
MF		 1195
1196 ocfs2_delete_osb(osb);
1197 kfree(osb);
1198 sb->s_dev = 0;
1199 sb->s_fs_info = NULL;
1200}
1201
1202static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1203 unsigned uuid_bytes)
1204{
1205 int i, ret;
1206 char *ptr;
1207
1208 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1209
1210 osb->uuid_str = kcalloc(1, OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1211 if (osb->uuid_str == NULL)
1212 return -ENOMEM;
1213
1214 memcpy(osb->uuid, uuid, OCFS2_VOL_UUID_LEN);
1215
1216 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1217 /* print with null */
1218 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1219 if (ret != 2) /* drop super cleans up */
1220 return -EINVAL;
1221 /* then only advance past the last char */
1222 ptr += 2;
1223 }
1224
1225 return 0;
1226}
1227
1228static int ocfs2_initialize_super(struct super_block *sb,
1229 struct buffer_head *bh,
1230 int sector_size)
1231{
1232 int status = 0;
1233 int i;
1234 struct ocfs2_dinode *di = NULL;
1235 struct inode *inode = NULL;
1236 struct buffer_head *bitmap_bh = NULL;
1237 struct ocfs2_journal *journal;
1238 __le32 uuid_net_key;
1239 struct ocfs2_super *osb;
1240
1241 mlog_entry_void();
1242
1243 osb = kcalloc(1, sizeof(struct ocfs2_super), GFP_KERNEL);
1244 if (!osb) {
1245 status = -ENOMEM;
1246 mlog_errno(status);
1247 goto bail;
1248 }
1249
1250 sb->s_fs_info = osb;
1251 sb->s_op = &ocfs2_sops;
1252 sb->s_export_op = &ocfs2_export_ops;
1253 sb->s_flags |= MS_NOATIME;
1254 /* this is needed to support O_LARGEFILE */
1255 sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits);
1256
1257 osb->sb = sb;
1258 /* Save off for ocfs2_rw_direct */
1259 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 1260 BUG_ON(!osb->s_sectsize_bits);
ccd979bd
MF		 1261
1262 osb->net_response_ids = 0;
1263 spin_lock_init(&osb->net_response_lock);
1264 INIT_LIST_HEAD(&osb->net_response_list);
1265
1266 INIT_LIST_HEAD(&osb->osb_net_handlers);
1267 init_waitqueue_head(&osb->recovery_event);
1268 spin_lock_init(&osb->vote_task_lock);
1269 init_waitqueue_head(&osb->vote_event);
1270 osb->vote_work_sequence = 0;
1271 osb->vote_wake_sequence = 0;
1272 INIT_LIST_HEAD(&osb->blocked_lock_list);
1273 osb->blocked_lock_count = 0;
1274 INIT_LIST_HEAD(&osb->vote_list);
1275 spin_lock_init(&osb->osb_lock);
1276
1277 atomic_set(&osb->alloc_stats.moves, 0);
1278 atomic_set(&osb->alloc_stats.local_data, 0);
1279 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1280 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1281 atomic_set(&osb->alloc_stats.bg_extends, 0);
1282
1283 ocfs2_init_node_maps(osb);
1284
1285 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1286 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1287
c74ec2f7 1288 mutex_init(&osb->recovery_lock);
ccd979bd
MF		 1289
1290 osb->disable_recovery = 0;
1291 osb->recovery_thread_task = NULL;
1292
1293 init_waitqueue_head(&osb->checkpoint_event);
1294 atomic_set(&osb->needs_checkpoint, 0);
1295
1296 osb->node_num = O2NM_INVALID_NODE_NUM;
1297 osb->slot_num = OCFS2_INVALID_SLOT;
1298
1299 osb->local_alloc_state = OCFS2_LA_UNUSED;
1300 osb->local_alloc_bh = NULL;
1301
1302 ocfs2_setup_hb_callbacks(osb);
1303
1304 init_waitqueue_head(&osb->osb_mount_event);
1305
1306 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1307 if (!osb->vol_label) {
1308 mlog(ML_ERROR, "unable to alloc vol label\n");
1309 status = -ENOMEM;
1310 goto bail;
1311 }
1312
1313 osb->uuid = kmalloc(OCFS2_VOL_UUID_LEN, GFP_KERNEL);
1314 if (!osb->uuid) {
1315 mlog(ML_ERROR, "unable to alloc uuid\n");
1316 status = -ENOMEM;
1317 goto bail;
1318 }
1319
1320 di = (struct ocfs2_dinode *)bh->b_data;
1321
1322 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1323 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1324 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1325 osb->max_slots);
1326 status = -EINVAL;
1327 goto bail;
1328 }
781ee3e2 1329 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
ccd979bd 1330
b4df6ed8
MF		 1331 init_waitqueue_head(&osb->osb_wipe_event);
1332 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1333 sizeof(*osb->osb_orphan_wipes),
1334 GFP_KERNEL);
1335 if (!osb->osb_orphan_wipes) {
1336 status = -ENOMEM;
1337 mlog_errno(status);
1338 goto bail;
1339 }
1340
ccd979bd
MF		 1341 osb->s_feature_compat =
1342 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1343 osb->s_feature_ro_compat =
1344 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1345 osb->s_feature_incompat =
1346 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1347
1348 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1349 mlog(ML_ERROR, "couldn't mount because of unsupported "
1350 "optional features (%x).\n", i);
1351 status = -EINVAL;
1352 goto bail;
1353 }
1354 if (!(osb->sb->s_flags & MS_RDONLY) &&
1355 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1356 mlog(ML_ERROR, "couldn't mount RDWR because of "
1357 "unsupported optional features (%x).\n", i);
1358 status = -EINVAL;
1359 goto bail;
1360 }
1361
1362 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1363
1364 /* FIXME
1365 * This should be done in ocfs2_journal_init(), but unknown
1366 * ordering issues will cause the filesystem to crash.
1367 * If anyone wants to figure out what part of the code
1368 * refers to osb->journal before ocfs2_journal_init() is run,
1369 * be my guest.
1370 */
1371 /* initialize our journal structure */
1372
1373 journal = kcalloc(1, sizeof(struct ocfs2_journal), GFP_KERNEL);
1374 if (!journal) {
1375 mlog(ML_ERROR, "unable to alloc journal\n");
1376 status = -ENOMEM;
1377 goto bail;
1378 }
1379 osb->journal = journal;
1380 journal->j_osb = osb;
1381
1382 atomic_set(&journal->j_num_trans, 0);
1383 init_rwsem(&journal->j_trans_barrier);
1384 init_waitqueue_head(&journal->j_checkpointed);
1385 spin_lock_init(&journal->j_lock);
1386 journal->j_trans_id = (unsigned long) 1;
1387 INIT_LIST_HEAD(&journal->j_la_cleanups);
1388 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery, osb);
1389 journal->j_state = OCFS2_JOURNAL_FREE;
1390
1391 /* get some pseudo constants for clustersize bits */
1392 osb->s_clustersize_bits =
1393 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1394 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1395 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1396
1397 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1398 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1399 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1400 osb->s_clustersize);
1401 status = -EINVAL;
1402 goto bail;
1403 }
1404
1405 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1406 > (u32)~0UL) {
1407 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1408 "what jbd can address in 32 bits.\n");
1409 status = -EINVAL;
1410 goto bail;
1411 }
1412
1413 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1414 sizeof(di->id2.i_super.s_uuid))) {
1415 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1416 status = -ENOMEM;
1417 goto bail;
1418 }
1419
a75a6e4c 1420 memcpy(&uuid_net_key, osb->uuid, sizeof(uuid_net_key));
ccd979bd
MF		 1421 osb->net_key = le32_to_cpu(uuid_net_key);
1422
1423 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1424 osb->vol_label[63] = '\0';
1425 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1426 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1427 osb->first_cluster_group_blkno =
1428 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1429 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1430 mlog(0, "vol_label: %s\n", osb->vol_label);
1431 mlog(0, "uuid: %s\n", osb->uuid_str);
b0697053
MF		 1432 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1433 (unsigned long long)osb->root_blkno,
1434 (unsigned long long)osb->system_dir_blkno);
ccd979bd
MF		 1435
1436 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1437 if (!osb->osb_dlm_debug) {
1438 status = -ENOMEM;
1439 mlog_errno(status);
1440 goto bail;
1441 }
1442
1443 atomic_set(&osb->vol_state, VOLUME_INIT);
1444
1445 /* load root, system_dir, and all global system inodes */
1446 status = ocfs2_init_global_system_inodes(osb);
1447 if (status < 0) {
1448 mlog_errno(status);
1449 goto bail;
1450 }
1451
1452 /*
1453 * global bitmap
1454 */
1455 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1456 OCFS2_INVALID_SLOT);
1457 if (!inode) {
1458 status = -EINVAL;
1459 mlog_errno(status);
1460 goto bail;
1461 }
1462
1463 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1464
1465 status = ocfs2_read_block(osb, osb->bitmap_blkno, &bitmap_bh, 0,
1466 inode);
1467 iput(inode);
1468 if (status < 0) {
1469 mlog_errno(status);
1470 goto bail;
1471 }
1472
1473 di = (struct ocfs2_dinode *) bitmap_bh->b_data;
1474 osb->bitmap_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
1475 osb->num_clusters = le32_to_cpu(di->id1.bitmap1.i_total);
1476 brelse(bitmap_bh);
b0697053
MF		 1477 mlog(0, "cluster bitmap inode: %llu, clusters per group: %u\n",
1478 (unsigned long long)osb->bitmap_blkno, osb->bitmap_cpg);
ccd979bd
MF		 1479
1480 status = ocfs2_init_slot_info(osb);
1481 if (status < 0) {
1482 mlog_errno(status);
1483 goto bail;
1484 }
1485
1486 /* Link this osb onto the global linked list of all osb structures. */
1487 /* The Global Link List is mainted for the whole driver . */
1488 spin_lock(&ocfs2_globals_lock);
1489 osb->osb_id = osb_id;
1490 if (osb_id < OCFS2_MAX_OSB_ID)
1491 osb_id++;
1492 else {
1493 mlog(ML_ERROR, "Too many volumes mounted\n");
1494 status = -ENOMEM;
1495 }
1496 spin_unlock(&ocfs2_globals_lock);
1497
1498bail:
1499 mlog_exit(status);
1500 return status;
1501}
1502
1503/*
1504 * will return: -EAGAIN if it is ok to keep searching for superblocks
1505 * -EINVAL if there is a bad superblock
1506 * 0 on success
1507 */
1508static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1509 struct buffer_head *bh,
1510 u32 blksz)
1511{
1512 int status = -EAGAIN;
1513
1514 mlog_entry_void();
1515
1516 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1517 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1518 status = -EINVAL;
1519 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1520 mlog(ML_ERROR, "found superblock with incorrect block "
1521 "size: found %u, should be %u\n",
1522 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1523 blksz);
1524 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1525 OCFS2_MAJOR_REV_LEVEL ||
1526 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1527 OCFS2_MINOR_REV_LEVEL) {
1528 mlog(ML_ERROR, "found superblock with bad version: "
1529 "found %u.%u, should be %u.%u\n",
1530 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1531 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1532 OCFS2_MAJOR_REV_LEVEL,
1533 OCFS2_MINOR_REV_LEVEL);
1534 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1535 mlog(ML_ERROR, "bad block number on superblock: "
b0697053
MF		 1536 "found %llu, should be %llu\n",
1537 (unsigned long long)di->i_blkno,
1538 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 1539 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1540 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1541 mlog(ML_ERROR, "bad cluster size found: %u\n",
1542 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1543 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1544 mlog(ML_ERROR, "bad root_blkno: 0\n");
1545 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1546 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1547 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1548 mlog(ML_ERROR,
1549 "Superblock slots found greater than file system "
1550 "maximum: found %u, max %u\n",
1551 le16_to_cpu(di->id2.i_super.s_max_slots),
1552 OCFS2_MAX_SLOTS);
1553 } else {
1554 /* found it! */
1555 status = 0;
1556 }
1557 }
1558
1559 mlog_exit(status);
1560 return status;
1561}
1562
1563static int ocfs2_check_volume(struct ocfs2_super *osb)
1564{
1565 int status = 0;
1566 int dirty;
1567 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1568 * recover
1569 * ourselves. */
1570
1571 mlog_entry_void();
1572
1573 /* Init our journal object. */
1574 status = ocfs2_journal_init(osb->journal, &dirty);
1575 if (status < 0) {
1576 mlog(ML_ERROR, "Could not initialize journal!\n");
1577 goto finally;
1578 }
1579
1580 /* If the journal was unmounted cleanly then we don't want to
1581 * recover anything. Otherwise, journal_load will do that
1582 * dirty work for us :) */
1583 if (!dirty) {
1584 status = ocfs2_journal_wipe(osb->journal, 0);
1585 if (status < 0) {
1586 mlog_errno(status);
1587 goto finally;
1588 }
1589 } else {
1590 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1591 "recovering volume.\n");
1592 }
1593
1594 /* will play back anything left in the journal. */
1595 ocfs2_journal_load(osb->journal);
1596
1597 if (dirty) {
1598 /* recover my local alloc if we didn't unmount cleanly. */
1599 status = ocfs2_begin_local_alloc_recovery(osb,
1600 osb->slot_num,
1601 &local_alloc);
1602 if (status < 0) {
1603 mlog_errno(status);
1604 goto finally;
1605 }
1606 /* we complete the recovery process after we've marked
1607 * ourselves as mounted. */
1608 }
1609
1610 mlog(0, "Journal loaded.\n");
1611
1612 status = ocfs2_load_local_alloc(osb);
1613 if (status < 0) {
1614 mlog_errno(status);
1615 goto finally;
1616 }
1617
1618 if (dirty) {
1619 /* Recovery will be completed after we've mounted the
1620 * rest of the volume. */
1621 osb->dirty = 1;
1622 osb->local_alloc_copy = local_alloc;
1623 local_alloc = NULL;
1624 }
1625
1626 /* go through each journal, trylock it and if you get the
1627 * lock, and it's marked as dirty, set the bit in the recover
1628 * map and launch a recovery thread for it. */
1629 status = ocfs2_mark_dead_nodes(osb);
1630 if (status < 0)
1631 mlog_errno(status);
1632
1633finally:
1634 if (local_alloc)
1635 kfree(local_alloc);
1636
1637 mlog_exit(status);
1638 return status;
1639}
1640
1641/*
1642 * The routine gets called from dismount or close whenever a dismount on
1643 * volume is requested and the osb open count becomes 1.
1644 * It will remove the osb from the global list and also free up all the
1645 * initialized resources and fileobject.
1646 */
1647static void ocfs2_delete_osb(struct ocfs2_super *osb)
1648{
1649 mlog_entry_void();
1650
1651 /* This function assumes that the caller has the main osb resource */
1652
1653 if (osb->slot_info)
1654 ocfs2_free_slot_info(osb->slot_info);
1655
b4df6ed8 1656 kfree(osb->osb_orphan_wipes);
ccd979bd
MF		 1657 /* FIXME
1658 * This belongs in journal shutdown, but because we have to
1659 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1660 * we free it here.
1661 */
1662 kfree(osb->journal);
1663 if (osb->local_alloc_copy)
1664 kfree(osb->local_alloc_copy);
1665 kfree(osb->uuid_str);
1666 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1667 memset(osb, 0, sizeof(struct ocfs2_super));
1668
1669 mlog_exit_void();
1670}
1671
1672/* Put OCFS2 into a readonly state, or (if the user specifies it),
1673 * panic(). We do not support continue-on-error operation. */
1674static void ocfs2_handle_error(struct super_block *sb)
1675{
1676 struct ocfs2_super *osb = OCFS2_SB(sb);
1677
1678 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1679 panic("OCFS2: (device %s): panic forced after error\n",
1680 sb->s_id);
1681
1682 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1683
1684 if (sb->s_flags & MS_RDONLY &&
1685 (ocfs2_is_soft_readonly(osb) ||
1686 ocfs2_is_hard_readonly(osb)))
1687 return;
1688
1689 printk(KERN_CRIT "File system is now read-only due to the potential "
1690 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1691 "system is unmounted.\n");
1692 sb->s_flags |= MS_RDONLY;
1693 ocfs2_set_ro_flag(osb, 0);
1694}
1695
1696static char error_buf[1024];
1697
1698void __ocfs2_error(struct super_block *sb,
1699 const char *function,
1700 const char *fmt, ...)
1701{
1702 va_list args;
1703
1704 va_start(args, fmt);
1705 vsprintf(error_buf, fmt, args);
1706 va_end(args);
1707
1708 /* Not using mlog here because we want to show the actual
1709 * function the error came from. */
1710 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1711 sb->s_id, function, error_buf);
1712
1713 ocfs2_handle_error(sb);
1714}
1715
1716/* Handle critical errors. This is intentionally more drastic than
1717 * ocfs2_handle_error, so we only use for things like journal errors,
1718 * etc. */
1719void __ocfs2_abort(struct super_block* sb,
1720 const char *function,
1721 const char *fmt, ...)
1722{
1723 va_list args;
1724
1725 va_start(args, fmt);
1726 vsprintf(error_buf, fmt, args);
1727 va_end(args);
1728
1729 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1730 sb->s_id, function, error_buf);
1731
1732 /* We don't have the cluster support yet to go straight to
1733 * hard readonly in here. Until then, we want to keep
1734 * ocfs2_abort() so that we can at least mark critical
1735 * errors.
1736 *
1737 * TODO: This should abort the journal and alert other nodes
1738 * that our slot needs recovery. */
1739
1740 /* Force a panic(). This stinks, but it's better than letting
1741 * things continue without having a proper hard readonly
1742 * here. */
1743 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1744 ocfs2_handle_error(sb);
1745}
1746
1747module_init(ocfs2_init);
1748module_exit(ocfs2_exit);
Linux 6.x block layer and io_uring tree(s)