projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ocfs2: Abstract out node number queries.
[linux-2.6-block.git] / fs / ocfs2 / super.c
CommitLineData
ccd979bd
MF		 1/* -*- mode: c; c-basic-offset: 8; -*-
2 * vim: noexpandtab sw=8 ts=8 sts=0:
3 *
4 * super.c
5 *
6 * load/unload driver, mount/dismount volumes
7 *
8 * Copyright (C) 2002, 2004 Oracle. All rights reserved.
9 *
10 * This program is free software; you can redistribute it and/or
11 * modify it under the terms of the GNU General Public
12 * License as published by the Free Software Foundation; either
13 * version 2 of the License, or (at your option) any later version.
14 *
15 * This program is distributed in the hope that it will be useful,
16 * but WITHOUT ANY WARRANTY; without even the implied warranty of
17 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
18 * General Public License for more details.
19 *
20 * You should have received a copy of the GNU General Public
21 * License along with this program; if not, write to the
22 * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23 * Boston, MA 021110-1307, USA.
24 */
25
26#include <linux/module.h>
27#include <linux/fs.h>
28#include <linux/types.h>
29#include <linux/slab.h>
30#include <linux/highmem.h>
31#include <linux/utsname.h>
32#include <linux/init.h>
33#include <linux/random.h>
34#include <linux/statfs.h>
35#include <linux/moduleparam.h>
36#include <linux/blkdev.h>
37#include <linux/socket.h>
38#include <linux/inet.h>
39#include <linux/parser.h>
40#include <linux/crc32.h>
41#include <linux/debugfs.h>
d550071c 42#include <linux/mount.h>
ccd979bd
MF		 43
44#include <cluster/nodemanager.h>
45
46#define MLOG_MASK_PREFIX ML_SUPER
47#include <cluster/masklog.h>
48
49#include "ocfs2.h"
50
51/* this should be the only file to include a version 1 header */
52#include "ocfs1_fs_compat.h"
53
54#include "alloc.h"
55#include "dlmglue.h"
56#include "export.h"
57#include "extent_map.h"
58#include "heartbeat.h"
59#include "inode.h"
60#include "journal.h"
61#include "localalloc.h"
62#include "namei.h"
63#include "slot_map.h"
64#include "super.h"
65#include "sysfile.h"
66#include "uptodate.h"
67#include "ver.h"
ccd979bd
MF		 68
69#include "buffer_head_io.h"
70
e18b890b 71static struct kmem_cache *ocfs2_inode_cachep = NULL;
ccd979bd 72
ccd979bd
MF		 73/* OCFS2 needs to schedule several differnt types of work which
74 * require cluster locking, disk I/O, recovery waits, etc. Since these
75 * types of work tend to be heavy we avoid using the kernel events
76 * workqueue and schedule on our own. */
77struct workqueue_struct *ocfs2_wq = NULL;
78
79static struct dentry *ocfs2_debugfs_root = NULL;
80
81MODULE_AUTHOR("Oracle");
82MODULE_LICENSE("GPL");
83
c0123ade
TY		 84struct mount_options
85{
d147b3d6 86 unsigned long commit_interval;
c0123ade
TY		 87 unsigned long mount_opt;
88 unsigned int atime_quantum;
89 signed short slot;
2fbe8d1e 90 unsigned int localalloc_opt;
c0123ade
TY		 91};
92
ccd979bd 93static int ocfs2_parse_options(struct super_block *sb, char *options,
c0123ade 94 struct mount_options *mopt,
baf4661a 95 int is_remount);
d550071c 96static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
ccd979bd
MF		 97static void ocfs2_put_super(struct super_block *sb);
98static int ocfs2_mount_volume(struct super_block *sb);
99static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
100static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
101static int ocfs2_initialize_mem_caches(void);
102static void ocfs2_free_mem_caches(void);
103static void ocfs2_delete_osb(struct ocfs2_super *osb);
104
726c3342 105static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
ccd979bd
MF		 106
107static int ocfs2_sync_fs(struct super_block *sb, int wait);
108
109static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
110static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
bddb8eb3 111static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
ccd979bd
MF		 112static int ocfs2_fill_local_node_info(struct ocfs2_super *osb);
113static int ocfs2_check_volume(struct ocfs2_super *osb);
114static int ocfs2_verify_volume(struct ocfs2_dinode *di,
115 struct buffer_head *bh,
116 u32 sectsize);
117static int ocfs2_initialize_super(struct super_block *sb,
118 struct buffer_head *bh,
119 int sector_size);
120static int ocfs2_get_sector(struct super_block *sb,
121 struct buffer_head **bh,
122 int block,
123 int sect_size);
124static void ocfs2_write_super(struct super_block *sb);
125static struct inode *ocfs2_alloc_inode(struct super_block *sb);
126static void ocfs2_destroy_inode(struct inode *inode);
127
ee9b6d61 128static const struct super_operations ocfs2_sops = {
ccd979bd
MF		 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,
d550071c 139 .show_options = ocfs2_show_options,
ccd979bd
MF		 140};
141
142enum {
143 Opt_barrier,
144 Opt_err_panic,
145 Opt_err_ro,
146 Opt_intr,
147 Opt_nointr,
148 Opt_hb_none,
149 Opt_hb_local,
150 Opt_data_ordered,
151 Opt_data_writeback,
7f1a37e3 152 Opt_atime_quantum,
baf4661a 153 Opt_slot,
d147b3d6 154 Opt_commit,
2fbe8d1e 155 Opt_localalloc,
53fc622b 156 Opt_localflocks,
ccd979bd
MF		 157 Opt_err,
158};
159
160static match_table_t tokens = {
161 {Opt_barrier, "barrier=%u"},
162 {Opt_err_panic, "errors=panic"},
163 {Opt_err_ro, "errors=remount-ro"},
164 {Opt_intr, "intr"},
165 {Opt_nointr, "nointr"},
166 {Opt_hb_none, OCFS2_HB_NONE},
167 {Opt_hb_local, OCFS2_HB_LOCAL},
168 {Opt_data_ordered, "data=ordered"},
169 {Opt_data_writeback, "data=writeback"},
7f1a37e3 170 {Opt_atime_quantum, "atime_quantum=%u"},
baf4661a 171 {Opt_slot, "preferred_slot=%u"},
d147b3d6 172 {Opt_commit, "commit=%u"},
2fbe8d1e 173 {Opt_localalloc, "localalloc=%d"},
53fc622b 174 {Opt_localflocks, "localflocks"},
ccd979bd
MF		 175 {Opt_err, NULL}
176};
177
178/*
179 * write_super and sync_fs ripped right out of ext3.
180 */
181static void ocfs2_write_super(struct super_block *sb)
182{
7892f2f4 183 if (mutex_trylock(&sb->s_lock) != 0)
ccd979bd
MF		 184 BUG();
185 sb->s_dirt = 0;
186}
187
188static int ocfs2_sync_fs(struct super_block *sb, int wait)
189{
bddb8eb3 190 int status;
ccd979bd
MF		 191 tid_t target;
192 struct ocfs2_super *osb = OCFS2_SB(sb);
193
194 sb->s_dirt = 0;
195
196 if (ocfs2_is_hard_readonly(osb))
197 return -EROFS;
198
199 if (wait) {
200 status = ocfs2_flush_truncate_log(osb);
201 if (status < 0)
202 mlog_errno(status);
203 } else {
204 ocfs2_schedule_truncate_log_flush(osb, 0);
205 }
206
207 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
208 if (wait)
209 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
210 target);
211 }
212 return 0;
213}
214
215static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
216{
217 struct inode *new = NULL;
218 int status = 0;
219 int i;
220
221 mlog_entry_void();
222
5fa0613e 223 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 224 if (IS_ERR(new)) {
225 status = PTR_ERR(new);
226 mlog_errno(status);
227 goto bail;
228 }
229 osb->root_inode = new;
230
5fa0613e 231 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
ccd979bd
MF		 232 if (IS_ERR(new)) {
233 status = PTR_ERR(new);
234 mlog_errno(status);
235 goto bail;
236 }
237 osb->sys_root_inode = new;
238
239 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
240 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
241 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
242 if (!new) {
243 ocfs2_release_system_inodes(osb);
244 status = -EINVAL;
245 mlog_errno(status);
246 /* FIXME: Should ERROR_RO_FS */
247 mlog(ML_ERROR, "Unable to load system inode %d, "
248 "possibly corrupt fs?", i);
249 goto bail;
250 }
251 // the array now has one ref, so drop this one
252 iput(new);
253 }
254
255bail:
256 mlog_exit(status);
257 return status;
258}
259
260static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
261{
262 struct inode *new = NULL;
263 int status = 0;
264 int i;
265
266 mlog_entry_void();
267
268 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
269 i < NUM_SYSTEM_INODES;
270 i++) {
271 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
272 if (!new) {
273 ocfs2_release_system_inodes(osb);
274 status = -EINVAL;
275 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
276 status, i, osb->slot_num);
277 goto bail;
278 }
279 /* the array now has one ref, so drop this one */
280 iput(new);
281 }
282
283bail:
284 mlog_exit(status);
285 return status;
286}
287
bddb8eb3 288static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
ccd979bd 289{
bddb8eb3 290 int i;
ccd979bd
MF		 291 struct inode *inode;
292
293 mlog_entry_void();
294
295 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
296 inode = osb->system_inodes[i];
297 if (inode) {
298 iput(inode);
299 osb->system_inodes[i] = NULL;
300 }
301 }
302
303 inode = osb->sys_root_inode;
304 if (inode) {
305 iput(inode);
306 osb->sys_root_inode = NULL;
307 }
308
309 inode = osb->root_inode;
310 if (inode) {
311 iput(inode);
312 osb->root_inode = NULL;
313 }
314
bddb8eb3 315 mlog_exit(0);
ccd979bd
MF		 316}
317
318/* We're allocating fs objects, use GFP_NOFS */
319static struct inode *ocfs2_alloc_inode(struct super_block *sb)
320{
321 struct ocfs2_inode_info *oi;
322
e6b4f8da 323 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
ccd979bd
MF		 324 if (!oi)
325 return NULL;
326
327 return &oi->vfs_inode;
328}
329
330static void ocfs2_destroy_inode(struct inode *inode)
331{
332 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
333}
334
5a254031
MF		 335static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
336 unsigned int cbits)
ccd979bd 337{
5a254031
MF		 338 unsigned int bytes = 1 << cbits;
339 unsigned int trim = bytes;
340 unsigned int bitshift = 32;
341
342 /*
343 * i_size and all block offsets in ocfs2 are always 64 bits
344 * wide. i_clusters is 32 bits, in cluster-sized units. So on
345 * 64 bit platforms, cluster size will be the limiting factor.
ccd979bd
MF		 346 */
347
348#if BITS_PER_LONG == 32
349# if defined(CONFIG_LBD)
2ecd05ae 350 BUILD_BUG_ON(sizeof(sector_t) != 8);
5a254031
MF		 351 /*
352 * We might be limited by page cache size.
353 */
354 if (bytes > PAGE_CACHE_SIZE) {
355 bytes = PAGE_CACHE_SIZE;
356 trim = 1;
357 /*
358 * Shift by 31 here so that we don't get larger than
359 * MAX_LFS_FILESIZE
360 */
361 bitshift = 31;
362 }
ccd979bd 363# else
5a254031
MF		 364 /*
365 * We are limited by the size of sector_t. Use block size, as
366 * that's what we expose to the VFS.
367 */
368 bytes = 1 << bbits;
369 trim = 1;
370 bitshift = 31;
ccd979bd
MF		 371# endif
372#endif
373
5a254031
MF		 374 /*
375 * Trim by a whole cluster when we can actually approach the
376 * on-disk limits. Otherwise we can overflow i_clusters when
377 * an extent start is at the max offset.
378 */
379 return (((unsigned long long)bytes) << bitshift) - trim;
ccd979bd
MF		 380}
381
382static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
383{
384 int incompat_features;
385 int ret = 0;
c0123ade 386 struct mount_options parsed_options;
ccd979bd
MF		 387 struct ocfs2_super *osb = OCFS2_SB(sb);
388
c0123ade 389 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
ccd979bd
MF		 390 ret = -EINVAL;
391 goto out;
392 }
393
394 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
c0123ade 395 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
ccd979bd
MF		 396 ret = -EINVAL;
397 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
398 goto out;
399 }
400
401 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
c0123ade 402 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
ccd979bd
MF		 403 ret = -EINVAL;
404 mlog(ML_ERROR, "Cannot change data mode on remount\n");
405 goto out;
406 }
407
408 /* We're going to/from readonly mode. */
409 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
410 /* Lock here so the check of HARD_RO and the potential
411 * setting of SOFT_RO is atomic. */
412 spin_lock(&osb->osb_lock);
413 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
414 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
415 ret = -EROFS;
416 goto unlock_osb;
417 }
418
419 if (*flags & MS_RDONLY) {
420 mlog(0, "Going to ro mode.\n");
421 sb->s_flags |= MS_RDONLY;
422 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
423 } else {
424 mlog(0, "Making ro filesystem writeable.\n");
425
426 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
427 mlog(ML_ERROR, "Cannot remount RDWR "
428 "filesystem due to previous errors.\n");
429 ret = -EROFS;
430 goto unlock_osb;
431 }
432 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
433 if (incompat_features) {
434 mlog(ML_ERROR, "Cannot remount RDWR because "
435 "of unsupported optional features "
436 "(%x).\n", incompat_features);
437 ret = -EINVAL;
438 goto unlock_osb;
439 }
440 sb->s_flags &= ~MS_RDONLY;
441 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
442 }
443unlock_osb:
444 spin_unlock(&osb->osb_lock);
445 }
446
447 if (!ret) {
ccd979bd
MF		 448 /* Only save off the new mount options in case of a successful
449 * remount. */
c0123ade
TY		 450 osb->s_mount_opt = parsed_options.mount_opt;
451 osb->s_atime_quantum = parsed_options.atime_quantum;
452 osb->preferred_slot = parsed_options.slot;
d147b3d6
MF		 453 if (parsed_options.commit_interval)
454 osb->osb_commit_interval = parsed_options.commit_interval;
e001e796
MF		 455
456 if (!ocfs2_is_hard_readonly(osb))
457 ocfs2_set_journal_params(osb);
ccd979bd
MF		 458 }
459out:
460 return ret;
461}
462
463static int ocfs2_sb_probe(struct super_block *sb,
464 struct buffer_head **bh,
465 int *sector_size)
466{
bddb8eb3 467 int status, tmpstat;
ccd979bd
MF		 468 struct ocfs1_vol_disk_hdr *hdr;
469 struct ocfs2_dinode *di;
470 int blksize;
471
472 *bh = NULL;
473
474 /* may be > 512 */
475 *sector_size = bdev_hardsect_size(sb->s_bdev);
476 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
477 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
478 *sector_size, OCFS2_MAX_BLOCKSIZE);
479 status = -EINVAL;
480 goto bail;
481 }
482
483 /* Can this really happen? */
484 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
485 *sector_size = OCFS2_MIN_BLOCKSIZE;
486
487 /* check block zero for old format */
488 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
489 if (status < 0) {
490 mlog_errno(status);
491 goto bail;
492 }
493 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
494 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
495 mlog(ML_ERROR, "incompatible version: %u.%u\n",
496 hdr->major_version, hdr->minor_version);
497 status = -EINVAL;
498 }
499 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
500 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
501 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
502 hdr->signature);
503 status = -EINVAL;
504 }
505 brelse(*bh);
506 *bh = NULL;
507 if (status < 0) {
508 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
509 "upgraded before mounting with ocfs v2\n");
510 goto bail;
511 }
512
513 /*
514 * Now check at magic offset for 512, 1024, 2048, 4096
515 * blocksizes. 4096 is the maximum blocksize because it is
516 * the minimum clustersize.
517 */
518 status = -EINVAL;
519 for (blksize = *sector_size;
520 blksize <= OCFS2_MAX_BLOCKSIZE;
521 blksize <<= 1) {
522 tmpstat = ocfs2_get_sector(sb, bh,
523 OCFS2_SUPER_BLOCK_BLKNO,
524 blksize);
525 if (tmpstat < 0) {
526 status = tmpstat;
527 mlog_errno(status);
528 goto bail;
529 }
530 di = (struct ocfs2_dinode *) (*bh)->b_data;
531 status = ocfs2_verify_volume(di, *bh, blksize);
532 if (status >= 0)
533 goto bail;
534 brelse(*bh);
535 *bh = NULL;
536 if (status != -EAGAIN)
537 break;
538 }
539
540bail:
541 return status;
542}
543
c271c5c2
SM		 544static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
545{
546 if (ocfs2_mount_local(osb)) {
547 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
548 mlog(ML_ERROR, "Cannot heartbeat on a locally "
549 "mounted device.\n");
550 return -EINVAL;
551 }
552 }
553
554 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
555 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
556 mlog(ML_ERROR, "Heartbeat has to be started to mount "
557 "a read-write clustered device.\n");
558 return -EINVAL;
559 }
560 }
561
562 return 0;
563}
564
ccd979bd
MF		 565static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
566{
567 struct dentry *root;
568 int status, sector_size;
c0123ade 569 struct mount_options parsed_options;
ccd979bd
MF		 570 struct inode *inode = NULL;
571 struct ocfs2_super *osb = NULL;
572 struct buffer_head *bh = NULL;
c271c5c2 573 char nodestr[8];
ccd979bd
MF		 574
575 mlog_entry("%p, %p, %i", sb, data, silent);
576
c0123ade 577 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
ccd979bd
MF		 578 status = -EINVAL;
579 goto read_super_error;
580 }
581
c271c5c2
SM		 582 /* for now we only have one cluster/node, make sure we see it
583 * in the heartbeat universe */
c0123ade 584 if (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL) {
c271c5c2
SM		 585 if (!o2hb_check_local_node_heartbeating()) {
586 status = -EINVAL;
587 goto read_super_error;
588 }
589 }
590
ccd979bd
MF		 591 /* probe for superblock */
592 status = ocfs2_sb_probe(sb, &bh, &sector_size);
593 if (status < 0) {
594 mlog(ML_ERROR, "superblock probe failed!\n");
595 goto read_super_error;
596 }
597
598 status = ocfs2_initialize_super(sb, bh, sector_size);
599 osb = OCFS2_SB(sb);
600 if (status < 0) {
601 mlog_errno(status);
602 goto read_super_error;
603 }
604 brelse(bh);
605 bh = NULL;
c0123ade
TY		 606 osb->s_mount_opt = parsed_options.mount_opt;
607 osb->s_atime_quantum = parsed_options.atime_quantum;
608 osb->preferred_slot = parsed_options.slot;
d147b3d6 609 osb->osb_commit_interval = parsed_options.commit_interval;
2fbe8d1e 610 osb->local_alloc_size = parsed_options.localalloc_opt;
ccd979bd
MF		 611
612 sb->s_magic = OCFS2_SUPER_MAGIC;
613
614 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
615 * heartbeat=none */
616 if (bdev_read_only(sb->s_bdev)) {
617 if (!(sb->s_flags & MS_RDONLY)) {
618 status = -EACCES;
619 mlog(ML_ERROR, "Readonly device detected but readonly "
620 "mount was not specified.\n");
621 goto read_super_error;
622 }
623
624 /* You should not be able to start a local heartbeat
625 * on a readonly device. */
626 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
627 status = -EROFS;
628 mlog(ML_ERROR, "Local heartbeat specified on readonly "
629 "device.\n");
630 goto read_super_error;
631 }
632
633 status = ocfs2_check_journals_nolocks(osb);
634 if (status < 0) {
635 if (status == -EROFS)
636 mlog(ML_ERROR, "Recovery required on readonly "
637 "file system, but write access is "
638 "unavailable.\n");
639 else
640 mlog_errno(status);
641 goto read_super_error;
642 }
643
644 ocfs2_set_ro_flag(osb, 1);
645
646 printk(KERN_NOTICE "Readonly device detected. No cluster "
647 "services will be utilized for this mount. Recovery "
648 "will be skipped.\n");
649 }
650
651 if (!ocfs2_is_hard_readonly(osb)) {
ccd979bd
MF		 652 if (sb->s_flags & MS_RDONLY)
653 ocfs2_set_ro_flag(osb, 0);
654 }
655
c271c5c2
SM		 656 status = ocfs2_verify_heartbeat(osb);
657 if (status < 0) {
658 mlog_errno(status);
659 goto read_super_error;
660 }
661
ccd979bd
MF		 662 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
663 ocfs2_debugfs_root);
664 if (!osb->osb_debug_root) {
665 status = -EINVAL;
666 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
667 goto read_super_error;
668 }
669
670 status = ocfs2_mount_volume(sb);
671 if (osb->root_inode)
672 inode = igrab(osb->root_inode);
673
674 if (status < 0)
675 goto read_super_error;
676
677 if (!inode) {
678 status = -EIO;
679 mlog_errno(status);
680 goto read_super_error;
681 }
682
683 root = d_alloc_root(inode);
684 if (!root) {
685 status = -ENOMEM;
686 mlog_errno(status);
687 goto read_super_error;
688 }
689
690 sb->s_root = root;
691
692 ocfs2_complete_mount_recovery(osb);
693
c271c5c2
SM		 694 if (ocfs2_mount_local(osb))
695 snprintf(nodestr, sizeof(nodestr), "local");
696 else
19fdb624 697 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 698
699 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
781ee3e2 700 "with %s data mode.\n",
c271c5c2 701 osb->dev_str, nodestr, osb->slot_num,
ccd979bd
MF		 702 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
703 "ordered");
704
705 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
706 wake_up(&osb->osb_mount_event);
707
708 mlog_exit(status);
709 return status;
710
711read_super_error:
712 if (bh != NULL)
713 brelse(bh);
714
715 if (inode)
716 iput(inode);
717
718 if (osb) {
719 atomic_set(&osb->vol_state, VOLUME_DISABLED);
720 wake_up(&osb->osb_mount_event);
721 ocfs2_dismount_volume(sb, 1);
722 }
723
724 mlog_exit(status);
725 return status;
726}
727
454e2398
DH		 728static int ocfs2_get_sb(struct file_system_type *fs_type,
729 int flags,
730 const char *dev_name,
731 void *data,
732 struct vfsmount *mnt)
ccd979bd 733{
454e2398
DH		 734 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
735 mnt);
ccd979bd
MF		 736}
737
738static struct file_system_type ocfs2_fs_type = {
739 .owner = THIS_MODULE,
740 .name = "ocfs2",
741 .get_sb = ocfs2_get_sb, /* is this called when we mount
742 * the fs? */
743 .kill_sb = kill_block_super, /* set to the generic one
744 * right now, but do we
745 * need to change that? */
1ba9da2f 746 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
ccd979bd
MF		 747 .next = NULL
748};
749
750static int ocfs2_parse_options(struct super_block *sb,
751 char *options,
c0123ade 752 struct mount_options *mopt,
ccd979bd
MF		 753 int is_remount)
754{
755 int status;
756 char *p;
757
758 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
759 options ? options : "(none)");
760
d147b3d6 761 mopt->commit_interval = 0;
c0123ade
TY		 762 mopt->mount_opt = 0;
763 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
764 mopt->slot = OCFS2_INVALID_SLOT;
2fbe8d1e 765 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
ccd979bd
MF		 766
767 if (!options) {
768 status = 1;
769 goto bail;
770 }
771
772 while ((p = strsep(&options, ",")) != NULL) {
773 int token, option;
774 substring_t args[MAX_OPT_ARGS];
775
776 if (!*p)
777 continue;
778
779 token = match_token(p, tokens, args);
780 switch (token) {
781 case Opt_hb_local:
c0123ade 782 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 783 break;
784 case Opt_hb_none:
c0123ade 785 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
ccd979bd
MF		 786 break;
787 case Opt_barrier:
788 if (match_int(&args[0], &option)) {
789 status = 0;
790 goto bail;
791 }
792 if (option)
c0123ade 793 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
ccd979bd 794 else
c0123ade 795 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
ccd979bd
MF		 796 break;
797 case Opt_intr:
c0123ade 798 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 799 break;
800 case Opt_nointr:
c0123ade 801 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
ccd979bd
MF		 802 break;
803 case Opt_err_panic:
c0123ade 804 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 805 break;
806 case Opt_err_ro:
c0123ade 807 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
ccd979bd
MF		 808 break;
809 case Opt_data_ordered:
c0123ade 810 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd
MF		 811 break;
812 case Opt_data_writeback:
c0123ade 813 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
ccd979bd 814 break;
7f1a37e3
TY		 815 case Opt_atime_quantum:
816 if (match_int(&args[0], &option)) {
817 status = 0;
818 goto bail;
819 }
820 if (option >= 0)
c0123ade 821 mopt->atime_quantum = option;
7f1a37e3 822 break;
baf4661a
SM		 823 case Opt_slot:
824 option = 0;
825 if (match_int(&args[0], &option)) {
826 status = 0;
827 goto bail;
828 }
829 if (option)
c0123ade 830 mopt->slot = (s16)option;
baf4661a 831 break;
d147b3d6
MF		 832 case Opt_commit:
833 option = 0;
834 if (match_int(&args[0], &option)) {
835 status = 0;
836 goto bail;
837 }
838 if (option < 0)
839 return 0;
840 if (option == 0)
841 option = JBD_DEFAULT_MAX_COMMIT_AGE;
842 mopt->commit_interval = HZ * option;
843 break;
2fbe8d1e
SM		 844 case Opt_localalloc:
845 option = 0;
846 if (match_int(&args[0], &option)) {
847 status = 0;
848 goto bail;
849 }
850 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
851 mopt->localalloc_opt = option;
852 break;
53fc622b
MF		 853 case Opt_localflocks:
854 /*
855 * Changing this during remount could race
856 * flock() requests, or "unbalance" existing
857 * ones (e.g., a lock is taken in one mode but
858 * dropped in the other). If users care enough
859 * to flip locking modes during remount, we
860 * could add a "local" flag to individual
861 * flock structures for proper tracking of
862 * state.
863 */
864 if (!is_remount)
865 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
866 break;
ccd979bd
MF		 867 default:
868 mlog(ML_ERROR,
869 "Unrecognized mount option \"%s\" "
870 "or missing value\n", p);
871 status = 0;
872 goto bail;
873 }
874 }
875
876 status = 1;
877
878bail:
879 mlog_exit(status);
880 return status;
881}
882
d550071c
SM		 883static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
884{
885 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
886 unsigned long opts = osb->s_mount_opt;
887
888 if (opts & OCFS2_MOUNT_HB_LOCAL)
889 seq_printf(s, ",_netdev,heartbeat=local");
890 else
891 seq_printf(s, ",heartbeat=none");
892
893 if (opts & OCFS2_MOUNT_NOINTR)
894 seq_printf(s, ",nointr");
895
896 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
897 seq_printf(s, ",data=writeback");
898 else
899 seq_printf(s, ",data=ordered");
900
901 if (opts & OCFS2_MOUNT_BARRIER)
902 seq_printf(s, ",barrier=1");
903
904 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
905 seq_printf(s, ",errors=panic");
906 else
907 seq_printf(s, ",errors=remount-ro");
908
909 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
910 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
911
912 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
913 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
914
d147b3d6
MF		 915 if (osb->osb_commit_interval)
916 seq_printf(s, ",commit=%u",
917 (unsigned) (osb->osb_commit_interval / HZ));
918
2fbe8d1e
SM		 919 if (osb->local_alloc_size != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
920 seq_printf(s, ",localalloc=%d", osb->local_alloc_size);
921
53fc622b
MF		 922 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
923 seq_printf(s, ",localflocks,");
924
d550071c
SM		 925 return 0;
926}
927
ccd979bd
MF		 928static int __init ocfs2_init(void)
929{
930 int status;
931
932 mlog_entry_void();
933
934 ocfs2_print_version();
935
24ef1815
JB		 936 dlmglue_init_stack();
937
ccd979bd
MF		 938 status = init_ocfs2_uptodate_cache();
939 if (status < 0) {
940 mlog_errno(status);
941 goto leave;
942 }
943
944 status = ocfs2_initialize_mem_caches();
945 if (status < 0) {
946 mlog_errno(status);
947 goto leave;
948 }
949
950 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
951 if (!ocfs2_wq) {
952 status = -ENOMEM;
953 goto leave;
954 }
955
ccd979bd
MF		 956 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
957 if (!ocfs2_debugfs_root) {
958 status = -EFAULT;
959 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
960 }
961
962leave:
963 if (status < 0) {
964 ocfs2_free_mem_caches();
965 exit_ocfs2_uptodate_cache();
ccd979bd
MF		 966 }
967
968 mlog_exit(status);
969
970 if (status >= 0) {
971 return register_filesystem(&ocfs2_fs_type);
972 } else
973 return -1;
974}
975
976static void __exit ocfs2_exit(void)
977{
978 mlog_entry_void();
979
980 if (ocfs2_wq) {
981 flush_workqueue(ocfs2_wq);
982 destroy_workqueue(ocfs2_wq);
983 }
984
985 debugfs_remove(ocfs2_debugfs_root);
986
987 ocfs2_free_mem_caches();
988
989 unregister_filesystem(&ocfs2_fs_type);
990
ccd979bd
MF		 991 exit_ocfs2_uptodate_cache();
992
24ef1815
JB		 993 dlmglue_exit_stack();
994
ccd979bd
MF		 995 mlog_exit_void();
996}
997
998static void ocfs2_put_super(struct super_block *sb)
999{
1000 mlog_entry("(0x%p)\n", sb);
1001
1002 ocfs2_sync_blockdev(sb);
1003 ocfs2_dismount_volume(sb, 0);
1004
1005 mlog_exit_void();
1006}
1007
726c3342 1008static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
ccd979bd
MF		 1009{
1010 struct ocfs2_super *osb;
1011 u32 numbits, freebits;
1012 int status;
1013 struct ocfs2_dinode *bm_lock;
1014 struct buffer_head *bh = NULL;
1015 struct inode *inode = NULL;
1016
726c3342 1017 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
ccd979bd 1018
726c3342 1019 osb = OCFS2_SB(dentry->d_sb);
ccd979bd
MF		 1020
1021 inode = ocfs2_get_system_file_inode(osb,
1022 GLOBAL_BITMAP_SYSTEM_INODE,
1023 OCFS2_INVALID_SLOT);
1024 if (!inode) {
1025 mlog(ML_ERROR, "failed to get bitmap inode\n");
1026 status = -EIO;
1027 goto bail;
1028 }
1029
e63aecb6 1030 status = ocfs2_inode_lock(inode, &bh, 0);
ccd979bd
MF		 1031 if (status < 0) {
1032 mlog_errno(status);
1033 goto bail;
1034 }
1035
1036 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1037
1038 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1039 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1040
1041 buf->f_type = OCFS2_SUPER_MAGIC;
726c3342 1042 buf->f_bsize = dentry->d_sb->s_blocksize;
ccd979bd
MF		 1043 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1044 buf->f_blocks = ((sector_t) numbits) *
1045 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1046 buf->f_bfree = ((sector_t) freebits) *
1047 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1048 buf->f_bavail = buf->f_bfree;
1049 buf->f_files = numbits;
1050 buf->f_ffree = freebits;
1051
1052 brelse(bh);
1053
e63aecb6 1054 ocfs2_inode_unlock(inode, 0);
ccd979bd
MF		 1055 status = 0;
1056bail:
1057 if (inode)
1058 iput(inode);
1059
1060 mlog_exit(status);
1061
1062 return status;
1063}
1064
4ba9b9d0 1065static void ocfs2_inode_init_once(struct kmem_cache *cachep, void *data)
ccd979bd
MF		 1066{
1067 struct ocfs2_inode_info *oi = data;
1068
a35afb83
CL		 1069 oi->ip_flags = 0;
1070 oi->ip_open_count = 0;
1071 spin_lock_init(&oi->ip_lock);
1072 ocfs2_extent_map_init(&oi->vfs_inode);
1073 INIT_LIST_HEAD(&oi->ip_io_markers);
1074 oi->ip_created_trans = 0;
1075 oi->ip_last_trans = 0;
1076 oi->ip_dir_start_lookup = 0;
ccd979bd 1077
a35afb83
CL		 1078 init_rwsem(&oi->ip_alloc_sem);
1079 mutex_init(&oi->ip_io_mutex);
ccd979bd 1080
a35afb83
CL		 1081 oi->ip_blkno = 0ULL;
1082 oi->ip_clusters = 0;
ccd979bd 1083
a35afb83 1084 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
e63aecb6 1085 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
a35afb83 1086 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
ccd979bd 1087
a35afb83 1088 ocfs2_metadata_cache_init(&oi->vfs_inode);
ccd979bd 1089
a35afb83 1090 inode_init_once(&oi->vfs_inode);
ccd979bd
MF		 1091}
1092
1093static int ocfs2_initialize_mem_caches(void)
1094{
1095 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
fffb60f9
PJ		 1096 sizeof(struct ocfs2_inode_info),
1097 0,
1098 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1099 SLAB_MEM_SPREAD),
20c2df83 1100 ocfs2_inode_init_once);
ccd979bd
MF		 1101 if (!ocfs2_inode_cachep)
1102 return -ENOMEM;
1103
ccd979bd
MF		 1104 return 0;
1105}
1106
1107static void ocfs2_free_mem_caches(void)
1108{
1109 if (ocfs2_inode_cachep)
1110 kmem_cache_destroy(ocfs2_inode_cachep);
ccd979bd
MF		 1111
1112 ocfs2_inode_cachep = NULL;
ccd979bd
MF		 1113}
1114
1115static int ocfs2_get_sector(struct super_block *sb,
1116 struct buffer_head **bh,
1117 int block,
1118 int sect_size)
1119{
1120 if (!sb_set_blocksize(sb, sect_size)) {
1121 mlog(ML_ERROR, "unable to set blocksize\n");
1122 return -EIO;
1123 }
1124
1125 *bh = sb_getblk(sb, block);
1126 if (!*bh) {
1127 mlog_errno(-EIO);
1128 return -EIO;
1129 }
1130 lock_buffer(*bh);
1131 if (!buffer_dirty(*bh))
1132 clear_buffer_uptodate(*bh);
1133 unlock_buffer(*bh);
1134 ll_rw_block(READ, 1, bh);
1135 wait_on_buffer(*bh);
1136 return 0;
1137}
1138
1139/* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
1140static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
1141{
1142 int status;
1143
1144 /* XXX hold a ref on the node while mounte? easy enough, if
1145 * desirable. */
c271c5c2
SM		 1146 if (ocfs2_mount_local(osb))
1147 osb->node_num = 0;
19fdb624
JB		 1148 else {
1149 status = ocfs2_cluster_this_node(&osb->node_num);
1150 if (status < 0) {
1151 mlog_errno(status);
1152 mlog(ML_ERROR,
1153 "could not find this host's node number\n");
1154 goto bail;
1155 }
ccd979bd
MF		 1156 }
1157
19fdb624 1158 mlog(0, "I am node %u\n", osb->node_num);
ccd979bd
MF		 1159
1160 status = 0;
1161bail:
1162 return status;
1163}
1164
1165static int ocfs2_mount_volume(struct super_block *sb)
1166{
1167 int status = 0;
1168 int unlock_super = 0;
1169 struct ocfs2_super *osb = OCFS2_SB(sb);
1170
1171 mlog_entry_void();
1172
1173 if (ocfs2_is_hard_readonly(osb))
1174 goto leave;
1175
1176 status = ocfs2_fill_local_node_info(osb);
1177 if (status < 0) {
1178 mlog_errno(status);
1179 goto leave;
1180 }
1181
ccd979bd
MF		 1182 status = ocfs2_dlm_init(osb);
1183 if (status < 0) {
1184 mlog_errno(status);
1185 goto leave;
1186 }
1187
ccd979bd
MF		 1188 status = ocfs2_super_lock(osb, 1);
1189 if (status < 0) {
1190 mlog_errno(status);
1191 goto leave;
1192 }
1193 unlock_super = 1;
1194
1195 /* This will load up the node map and add ourselves to it. */
1196 status = ocfs2_find_slot(osb);
1197 if (status < 0) {
1198 mlog_errno(status);
1199 goto leave;
1200 }
1201
ccd979bd
MF		 1202 /* load all node-local system inodes */
1203 status = ocfs2_init_local_system_inodes(osb);
1204 if (status < 0) {
1205 mlog_errno(status);
1206 goto leave;
1207 }
1208
1209 status = ocfs2_check_volume(osb);
1210 if (status < 0) {
1211 mlog_errno(status);
1212 goto leave;
1213 }
1214
1215 status = ocfs2_truncate_log_init(osb);
1216 if (status < 0) {
1217 mlog_errno(status);
1218 goto leave;
1219 }
1220
c271c5c2
SM		 1221 if (ocfs2_mount_local(osb))
1222 goto leave;
1223
ccd979bd
MF		 1224leave:
1225 if (unlock_super)
1226 ocfs2_super_unlock(osb, 1);
1227
1228 mlog_exit(status);
1229 return status;
1230}
1231
ccd979bd
MF		 1232static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1233{
1234 int tmp;
1235 struct ocfs2_super *osb = NULL;
c271c5c2 1236 char nodestr[8];
ccd979bd
MF		 1237
1238 mlog_entry("(0x%p)\n", sb);
1239
1240 BUG_ON(!sb);
1241 osb = OCFS2_SB(sb);
1242 BUG_ON(!osb);
1243
1244 ocfs2_shutdown_local_alloc(osb);
1245
1246 ocfs2_truncate_log_shutdown(osb);
1247
553abd04
JB		 1248 /* This will disable recovery and flush any recovery work. */
1249 ocfs2_recovery_exit(osb);
ccd979bd
MF		 1250
1251 ocfs2_journal_shutdown(osb);
1252
1253 ocfs2_sync_blockdev(sb);
1254
4670c46d
JB		 1255 /* No cluster connection means we've failed during mount, so skip
1256 * all the steps which depended on that to complete. */
1257 if (osb->cconn) {
ccd979bd
MF		 1258 tmp = ocfs2_super_lock(osb, 1);
1259 if (tmp < 0) {
1260 mlog_errno(tmp);
1261 return;
1262 }
19b613d4 1263 }
ccd979bd 1264
19b613d4
MF		 1265 if (osb->slot_num != OCFS2_INVALID_SLOT)
1266 ocfs2_put_slot(osb);
ccd979bd 1267
4670c46d 1268 if (osb->cconn)
ccd979bd 1269 ocfs2_super_unlock(osb, 1);
ccd979bd
MF		 1270
1271 ocfs2_release_system_inodes(osb);
1272
4670c46d 1273 if (osb->cconn)
ccd979bd 1274 ocfs2_dlm_shutdown(osb);
ccd979bd 1275
ccd979bd
MF		 1276 debugfs_remove(osb->osb_debug_root);
1277
1278 if (!mnt_err)
1279 ocfs2_stop_heartbeat(osb);
1280
1281 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1282
c271c5c2
SM		 1283 if (ocfs2_mount_local(osb))
1284 snprintf(nodestr, sizeof(nodestr), "local");
1285 else
19fdb624 1286 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
c271c5c2
SM		 1287
1288 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1289 osb->dev_str, nodestr);
ccd979bd
MF		 1290
1291 ocfs2_delete_osb(osb);
1292 kfree(osb);
1293 sb->s_dev = 0;
1294 sb->s_fs_info = NULL;
1295}
1296
1297static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1298 unsigned uuid_bytes)
1299{
1300 int i, ret;
1301 char *ptr;
1302
1303 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1304
cd861280 1305 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
ccd979bd
MF		 1306 if (osb->uuid_str == NULL)
1307 return -ENOMEM;
1308
ccd979bd
MF		 1309 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1310 /* print with null */
1311 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1312 if (ret != 2) /* drop super cleans up */
1313 return -EINVAL;
1314 /* then only advance past the last char */
1315 ptr += 2;
1316 }
1317
1318 return 0;
1319}
1320
1321static int ocfs2_initialize_super(struct super_block *sb,
1322 struct buffer_head *bh,
1323 int sector_size)
1324{
bddb8eb3 1325 int status;
5a254031
MF		 1326 int i, cbits, bbits;
1327 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
ccd979bd 1328 struct inode *inode = NULL;
ccd979bd
MF		 1329 struct ocfs2_journal *journal;
1330 __le32 uuid_net_key;
1331 struct ocfs2_super *osb;
1332
1333 mlog_entry_void();
1334
cd861280 1335 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
ccd979bd
MF		 1336 if (!osb) {
1337 status = -ENOMEM;
1338 mlog_errno(status);
1339 goto bail;
1340 }
1341
1342 sb->s_fs_info = osb;
1343 sb->s_op = &ocfs2_sops;
1344 sb->s_export_op = &ocfs2_export_ops;
e0dceaf0 1345 sb->s_time_gran = 1;
ccd979bd
MF		 1346 sb->s_flags |= MS_NOATIME;
1347 /* this is needed to support O_LARGEFILE */
5a254031
MF		 1348 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1349 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1350 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
ccd979bd
MF		 1351
1352 osb->sb = sb;
1353 /* Save off for ocfs2_rw_direct */
1354 osb->s_sectsize_bits = blksize_bits(sector_size);
ebdec83b 1355 BUG_ON(!osb->s_sectsize_bits);
ccd979bd 1356
34d024f8
MF		 1357 spin_lock_init(&osb->dc_task_lock);
1358 init_waitqueue_head(&osb->dc_event);
1359 osb->dc_work_sequence = 0;
1360 osb->dc_wake_sequence = 0;
ccd979bd
MF		 1361 INIT_LIST_HEAD(&osb->blocked_lock_list);
1362 osb->blocked_lock_count = 0;
ccd979bd
MF		 1363 spin_lock_init(&osb->osb_lock);
1364
1365 atomic_set(&osb->alloc_stats.moves, 0);
1366 atomic_set(&osb->alloc_stats.local_data, 0);
1367 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1368 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1369 atomic_set(&osb->alloc_stats.bg_extends, 0);
1370
1371 ocfs2_init_node_maps(osb);
1372
1373 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1374 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1375
553abd04
JB		 1376 status = ocfs2_recovery_init(osb);
1377 if (status) {
1378 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1379 mlog_errno(status);
1380 goto bail;
1381 }
ccd979bd
MF		 1382
1383 init_waitqueue_head(&osb->checkpoint_event);
1384 atomic_set(&osb->needs_checkpoint, 0);
1385
7f1a37e3
TY		 1386 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1387
ccd979bd
MF		 1388 osb->slot_num = OCFS2_INVALID_SLOT;
1389
1390 osb->local_alloc_state = OCFS2_LA_UNUSED;
1391 osb->local_alloc_bh = NULL;
1392
ccd979bd
MF		 1393 init_waitqueue_head(&osb->osb_mount_event);
1394
1395 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1396 if (!osb->vol_label) {
1397 mlog(ML_ERROR, "unable to alloc vol label\n");
1398 status = -ENOMEM;
1399 goto bail;
1400 }
1401
ccd979bd
MF		 1402 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1403 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1404 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1405 osb->max_slots);
1406 status = -EINVAL;
1407 goto bail;
1408 }
781ee3e2 1409 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
ccd979bd 1410
b4df6ed8
MF		 1411 init_waitqueue_head(&osb->osb_wipe_event);
1412 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1413 sizeof(*osb->osb_orphan_wipes),
1414 GFP_KERNEL);
1415 if (!osb->osb_orphan_wipes) {
1416 status = -ENOMEM;
1417 mlog_errno(status);
1418 goto bail;
1419 }
1420
ccd979bd
MF		 1421 osb->s_feature_compat =
1422 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1423 osb->s_feature_ro_compat =
1424 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1425 osb->s_feature_incompat =
1426 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1427
1428 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1429 mlog(ML_ERROR, "couldn't mount because of unsupported "
1430 "optional features (%x).\n", i);
1431 status = -EINVAL;
1432 goto bail;
1433 }
1434 if (!(osb->sb->s_flags & MS_RDONLY) &&
1435 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1436 mlog(ML_ERROR, "couldn't mount RDWR because of "
1437 "unsupported optional features (%x).\n", i);
1438 status = -EINVAL;
1439 goto bail;
1440 }
1441
1442 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1443
1444 /* FIXME
1445 * This should be done in ocfs2_journal_init(), but unknown
1446 * ordering issues will cause the filesystem to crash.
1447 * If anyone wants to figure out what part of the code
1448 * refers to osb->journal before ocfs2_journal_init() is run,
1449 * be my guest.
1450 */
1451 /* initialize our journal structure */
1452
cd861280 1453 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
ccd979bd
MF		 1454 if (!journal) {
1455 mlog(ML_ERROR, "unable to alloc journal\n");
1456 status = -ENOMEM;
1457 goto bail;
1458 }
1459 osb->journal = journal;
1460 journal->j_osb = osb;
1461
1462 atomic_set(&journal->j_num_trans, 0);
1463 init_rwsem(&journal->j_trans_barrier);
1464 init_waitqueue_head(&journal->j_checkpointed);
1465 spin_lock_init(&journal->j_lock);
1466 journal->j_trans_id = (unsigned long) 1;
1467 INIT_LIST_HEAD(&journal->j_la_cleanups);
c4028958 1468 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
ccd979bd
MF		 1469 journal->j_state = OCFS2_JOURNAL_FREE;
1470
1471 /* get some pseudo constants for clustersize bits */
1472 osb->s_clustersize_bits =
1473 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1474 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1475 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1476
1477 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1478 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1479 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1480 osb->s_clustersize);
1481 status = -EINVAL;
1482 goto bail;
1483 }
1484
1485 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1486 > (u32)~0UL) {
1487 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1488 "what jbd can address in 32 bits.\n");
1489 status = -EINVAL;
1490 goto bail;
1491 }
1492
1493 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1494 sizeof(di->id2.i_super.s_uuid))) {
1495 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1496 status = -ENOMEM;
1497 goto bail;
1498 }
1499
78427043 1500 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
ccd979bd
MF		 1501
1502 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1503 osb->vol_label[63] = '\0';
1504 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1505 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1506 osb->first_cluster_group_blkno =
1507 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1508 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1509 mlog(0, "vol_label: %s\n", osb->vol_label);
1510 mlog(0, "uuid: %s\n", osb->uuid_str);
b0697053
MF		 1511 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1512 (unsigned long long)osb->root_blkno,
1513 (unsigned long long)osb->system_dir_blkno);
ccd979bd
MF		 1514
1515 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1516 if (!osb->osb_dlm_debug) {
1517 status = -ENOMEM;
1518 mlog_errno(status);
1519 goto bail;
1520 }
1521
1522 atomic_set(&osb->vol_state, VOLUME_INIT);
1523
1524 /* load root, system_dir, and all global system inodes */
1525 status = ocfs2_init_global_system_inodes(osb);
1526 if (status < 0) {
1527 mlog_errno(status);
1528 goto bail;
1529 }
1530
1531 /*
1532 * global bitmap
1533 */
1534 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1535 OCFS2_INVALID_SLOT);
1536 if (!inode) {
1537 status = -EINVAL;
1538 mlog_errno(status);
1539 goto bail;
1540 }
1541
1542 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
ccd979bd 1543 iput(inode);
ccd979bd 1544
e9d578a8 1545 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
ccd979bd
MF		 1546
1547 status = ocfs2_init_slot_info(osb);
1548 if (status < 0) {
1549 mlog_errno(status);
1550 goto bail;
1551 }
1552
ccd979bd
MF		 1553bail:
1554 mlog_exit(status);
1555 return status;
1556}
1557
1558/*
1559 * will return: -EAGAIN if it is ok to keep searching for superblocks
1560 * -EINVAL if there is a bad superblock
1561 * 0 on success
1562 */
1563static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1564 struct buffer_head *bh,
1565 u32 blksz)
1566{
1567 int status = -EAGAIN;
1568
1569 mlog_entry_void();
1570
1571 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1572 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1573 status = -EINVAL;
1574 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1575 mlog(ML_ERROR, "found superblock with incorrect block "
1576 "size: found %u, should be %u\n",
1577 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1578 blksz);
1579 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1580 OCFS2_MAJOR_REV_LEVEL ||
1581 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1582 OCFS2_MINOR_REV_LEVEL) {
1583 mlog(ML_ERROR, "found superblock with bad version: "
1584 "found %u.%u, should be %u.%u\n",
1585 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1586 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1587 OCFS2_MAJOR_REV_LEVEL,
1588 OCFS2_MINOR_REV_LEVEL);
1589 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1590 mlog(ML_ERROR, "bad block number on superblock: "
b0697053 1591 "found %llu, should be %llu\n",
1ca1a111 1592 (unsigned long long)le64_to_cpu(di->i_blkno),
b0697053 1593 (unsigned long long)bh->b_blocknr);
ccd979bd
MF		 1594 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1595 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1596 mlog(ML_ERROR, "bad cluster size found: %u\n",
1597 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1598 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1599 mlog(ML_ERROR, "bad root_blkno: 0\n");
1600 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1601 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1602 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1603 mlog(ML_ERROR,
1604 "Superblock slots found greater than file system "
1605 "maximum: found %u, max %u\n",
1606 le16_to_cpu(di->id2.i_super.s_max_slots),
1607 OCFS2_MAX_SLOTS);
1608 } else {
1609 /* found it! */
1610 status = 0;
1611 }
1612 }
1613
1614 mlog_exit(status);
1615 return status;
1616}
1617
1618static int ocfs2_check_volume(struct ocfs2_super *osb)
1619{
bddb8eb3 1620 int status;
ccd979bd 1621 int dirty;
c271c5c2 1622 int local;
ccd979bd
MF		 1623 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1624 * recover
1625 * ourselves. */
1626
1627 mlog_entry_void();
1628
1629 /* Init our journal object. */
1630 status = ocfs2_journal_init(osb->journal, &dirty);
1631 if (status < 0) {
1632 mlog(ML_ERROR, "Could not initialize journal!\n");
1633 goto finally;
1634 }
1635
1636 /* If the journal was unmounted cleanly then we don't want to
1637 * recover anything. Otherwise, journal_load will do that
1638 * dirty work for us :) */
1639 if (!dirty) {
1640 status = ocfs2_journal_wipe(osb->journal, 0);
1641 if (status < 0) {
1642 mlog_errno(status);
1643 goto finally;
1644 }
1645 } else {
1646 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1647 "recovering volume.\n");
1648 }
1649
c271c5c2
SM		 1650 local = ocfs2_mount_local(osb);
1651
ccd979bd 1652 /* will play back anything left in the journal. */
c271c5c2 1653 ocfs2_journal_load(osb->journal, local);
ccd979bd
MF		 1654
1655 if (dirty) {
1656 /* recover my local alloc if we didn't unmount cleanly. */
1657 status = ocfs2_begin_local_alloc_recovery(osb,
1658 osb->slot_num,
1659 &local_alloc);
1660 if (status < 0) {
1661 mlog_errno(status);
1662 goto finally;
1663 }
1664 /* we complete the recovery process after we've marked
1665 * ourselves as mounted. */
1666 }
1667
1668 mlog(0, "Journal loaded.\n");
1669
1670 status = ocfs2_load_local_alloc(osb);
1671 if (status < 0) {
1672 mlog_errno(status);
1673 goto finally;
1674 }
1675
1676 if (dirty) {
1677 /* Recovery will be completed after we've mounted the
1678 * rest of the volume. */
1679 osb->dirty = 1;
1680 osb->local_alloc_copy = local_alloc;
1681 local_alloc = NULL;
1682 }
1683
1684 /* go through each journal, trylock it and if you get the
1685 * lock, and it's marked as dirty, set the bit in the recover
1686 * map and launch a recovery thread for it. */
1687 status = ocfs2_mark_dead_nodes(osb);
1688 if (status < 0)
1689 mlog_errno(status);
1690
1691finally:
1692 if (local_alloc)
1693 kfree(local_alloc);
1694
1695 mlog_exit(status);
1696 return status;
1697}
1698
1699/*
1700 * The routine gets called from dismount or close whenever a dismount on
1701 * volume is requested and the osb open count becomes 1.
1702 * It will remove the osb from the global list and also free up all the
1703 * initialized resources and fileobject.
1704 */
1705static void ocfs2_delete_osb(struct ocfs2_super *osb)
1706{
1707 mlog_entry_void();
1708
1709 /* This function assumes that the caller has the main osb resource */
1710
8e8a4603 1711 ocfs2_free_slot_info(osb);
ccd979bd 1712
b4df6ed8 1713 kfree(osb->osb_orphan_wipes);
ccd979bd
MF		 1714 /* FIXME
1715 * This belongs in journal shutdown, but because we have to
1716 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1717 * we free it here.
1718 */
1719 kfree(osb->journal);
1720 if (osb->local_alloc_copy)
1721 kfree(osb->local_alloc_copy);
1722 kfree(osb->uuid_str);
1723 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1724 memset(osb, 0, sizeof(struct ocfs2_super));
1725
1726 mlog_exit_void();
1727}
1728
1729/* Put OCFS2 into a readonly state, or (if the user specifies it),
1730 * panic(). We do not support continue-on-error operation. */
1731static void ocfs2_handle_error(struct super_block *sb)
1732{
1733 struct ocfs2_super *osb = OCFS2_SB(sb);
1734
1735 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1736 panic("OCFS2: (device %s): panic forced after error\n",
1737 sb->s_id);
1738
1739 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1740
1741 if (sb->s_flags & MS_RDONLY &&
1742 (ocfs2_is_soft_readonly(osb) ||
1743 ocfs2_is_hard_readonly(osb)))
1744 return;
1745
1746 printk(KERN_CRIT "File system is now read-only due to the potential "
1747 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1748 "system is unmounted.\n");
1749 sb->s_flags |= MS_RDONLY;
1750 ocfs2_set_ro_flag(osb, 0);
1751}
1752
1753static char error_buf[1024];
1754
1755void __ocfs2_error(struct super_block *sb,
1756 const char *function,
1757 const char *fmt, ...)
1758{
1759 va_list args;
1760
1761 va_start(args, fmt);
4a6e617a 1762 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 1763 va_end(args);
1764
1765 /* Not using mlog here because we want to show the actual
1766 * function the error came from. */
1767 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1768 sb->s_id, function, error_buf);
1769
1770 ocfs2_handle_error(sb);
1771}
1772
1773/* Handle critical errors. This is intentionally more drastic than
1774 * ocfs2_handle_error, so we only use for things like journal errors,
1775 * etc. */
1776void __ocfs2_abort(struct super_block* sb,
1777 const char *function,
1778 const char *fmt, ...)
1779{
1780 va_list args;
1781
1782 va_start(args, fmt);
4a6e617a 1783 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
ccd979bd
MF		 1784 va_end(args);
1785
1786 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1787 sb->s_id, function, error_buf);
1788
1789 /* We don't have the cluster support yet to go straight to
1790 * hard readonly in here. Until then, we want to keep
1791 * ocfs2_abort() so that we can at least mark critical
1792 * errors.
1793 *
1794 * TODO: This should abort the journal and alert other nodes
1795 * that our slot needs recovery. */
1796
1797 /* Force a panic(). This stinks, but it's better than letting
1798 * things continue without having a proper hard readonly
1799 * here. */
1800 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1801 ocfs2_handle_error(sb);
1802}
1803
1804module_init(ocfs2_init);
1805module_exit(ocfs2_exit);
Linux 6.x block layer and io_uring tree(s)