?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge branch 'upstream-linus' of master.kernel.org:/pub/scm/linux/kernel/git/jgarzik...
[linux-2.6/linux-acpi-2.6/ibm-acpi-2.6.git] / fs / ocfs2 / super.c
blob6534f92424dd213a4471a8da553439f9843daa37
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 static struct kmem_cache *ocfs2_inode_cachep = NULL;
72
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. */
77 struct workqueue_struct *ocfs2_wq = NULL;
78
79 static struct dentry *ocfs2_debugfs_root = NULL;
80
81 MODULE_AUTHOR("Oracle");
82 MODULE_LICENSE("GPL");
83
84 static int ocfs2_parse_options(struct super_block *sb, char *options,
85 unsigned long *mount_opt, int is_remount);
86 static void ocfs2_put_super(struct super_block *sb);
87 static int ocfs2_mount_volume(struct super_block *sb);
88 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
89 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
90 static int ocfs2_initialize_mem_caches(void);
91 static void ocfs2_free_mem_caches(void);
92 static void ocfs2_delete_osb(struct ocfs2_super *osb);
93
94 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
95
96 static int ocfs2_sync_fs(struct super_block *sb, int wait);
97
98 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
99 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
100 static int ocfs2_release_system_inodes(struct ocfs2_super *osb);
101 static int ocfs2_fill_local_node_info(struct ocfs2_super *osb);
102 static int ocfs2_check_volume(struct ocfs2_super *osb);
103 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
104 struct buffer_head *bh,
105 u32 sectsize);
106 static int ocfs2_initialize_super(struct super_block *sb,
107 struct buffer_head *bh,
108 int sector_size);
109 static int ocfs2_get_sector(struct super_block *sb,
110 struct buffer_head **bh,
111 int block,
112 int sect_size);
113 static void ocfs2_write_super(struct super_block *sb);
114 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
115 static void ocfs2_destroy_inode(struct inode *inode);
116
117 static unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
118
119 static const struct super_operations ocfs2_sops = {
120 .statfs = ocfs2_statfs,
121 .alloc_inode = ocfs2_alloc_inode,
122 .destroy_inode = ocfs2_destroy_inode,
123 .drop_inode = ocfs2_drop_inode,
124 .clear_inode = ocfs2_clear_inode,
125 .delete_inode = ocfs2_delete_inode,
126 .sync_fs = ocfs2_sync_fs,
127 .write_super = ocfs2_write_super,
128 .put_super = ocfs2_put_super,
129 .remount_fs = ocfs2_remount,
130 };
131
132 enum {
133 Opt_barrier,
134 Opt_err_panic,
135 Opt_err_ro,
136 Opt_intr,
137 Opt_nointr,
138 Opt_hb_none,
139 Opt_hb_local,
140 Opt_data_ordered,
141 Opt_data_writeback,
142 Opt_atime_quantum,
143 Opt_err,
144 };
145
146 static match_table_t tokens = {
147 {Opt_barrier, "barrier=%u"},
148 {Opt_err_panic, "errors=panic"},
149 {Opt_err_ro, "errors=remount-ro"},
150 {Opt_intr, "intr"},
151 {Opt_nointr, "nointr"},
152 {Opt_hb_none, OCFS2_HB_NONE},
153 {Opt_hb_local, OCFS2_HB_LOCAL},
154 {Opt_data_ordered, "data=ordered"},
155 {Opt_data_writeback, "data=writeback"},
156 {Opt_atime_quantum, "atime_quantum=%u"},
157 {Opt_err, NULL}
158 };
159
160 /*
161 * write_super and sync_fs ripped right out of ext3.
162 */
163 static void ocfs2_write_super(struct super_block *sb)
164 {
165 if (mutex_trylock(&sb->s_lock) != 0)
166 BUG();
167 sb->s_dirt = 0;
168 }
169
170 static int ocfs2_sync_fs(struct super_block *sb, int wait)
171 {
172 int status = 0;
173 tid_t target;
174 struct ocfs2_super *osb = OCFS2_SB(sb);
175
176 sb->s_dirt = 0;
177
178 if (ocfs2_is_hard_readonly(osb))
179 return -EROFS;
180
181 if (wait) {
182 status = ocfs2_flush_truncate_log(osb);
183 if (status < 0)
184 mlog_errno(status);
185 } else {
186 ocfs2_schedule_truncate_log_flush(osb, 0);
187 }
188
189 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
190 if (wait)
191 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
192 target);
193 }
194 return 0;
195 }
196
197 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
198 {
199 struct inode *new = NULL;
200 int status = 0;
201 int i;
202
203 mlog_entry_void();
204
205 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE);
206 if (IS_ERR(new)) {
207 status = PTR_ERR(new);
208 mlog_errno(status);
209 goto bail;
210 }
211 osb->root_inode = new;
212
213 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE);
214 if (IS_ERR(new)) {
215 status = PTR_ERR(new);
216 mlog_errno(status);
217 goto bail;
218 }
219 osb->sys_root_inode = new;
220
221 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
222 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
223 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
224 if (!new) {
225 ocfs2_release_system_inodes(osb);
226 status = -EINVAL;
227 mlog_errno(status);
228 /* FIXME: Should ERROR_RO_FS */
229 mlog(ML_ERROR, "Unable to load system inode %d, "
230 "possibly corrupt fs?", i);
231 goto bail;
232 }
233 // the array now has one ref, so drop this one
234 iput(new);
235 }
236
237 bail:
238 mlog_exit(status);
239 return status;
240 }
241
242 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
243 {
244 struct inode *new = NULL;
245 int status = 0;
246 int i;
247
248 mlog_entry_void();
249
250 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
251 i < NUM_SYSTEM_INODES;
252 i++) {
253 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
254 if (!new) {
255 ocfs2_release_system_inodes(osb);
256 status = -EINVAL;
257 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
258 status, i, osb->slot_num);
259 goto bail;
260 }
261 /* the array now has one ref, so drop this one */
262 iput(new);
263 }
264
265 bail:
266 mlog_exit(status);
267 return status;
268 }
269
270 static int ocfs2_release_system_inodes(struct ocfs2_super *osb)
271 {
272 int status = 0, i;
273 struct inode *inode;
274
275 mlog_entry_void();
276
277 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
278 inode = osb->system_inodes[i];
279 if (inode) {
280 iput(inode);
281 osb->system_inodes[i] = NULL;
282 }
283 }
284
285 inode = osb->sys_root_inode;
286 if (inode) {
287 iput(inode);
288 osb->sys_root_inode = NULL;
289 }
290
291 inode = osb->root_inode;
292 if (inode) {
293 iput(inode);
294 osb->root_inode = NULL;
295 }
296
297 mlog_exit(status);
298 return status;
299 }
300
301 /* We're allocating fs objects, use GFP_NOFS */
302 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
303 {
304 struct ocfs2_inode_info *oi;
305
306 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
307 if (!oi)
308 return NULL;
309
310 return &oi->vfs_inode;
311 }
312
313 static void ocfs2_destroy_inode(struct inode *inode)
314 {
315 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
316 }
317
318 /* From xfs_super.c:xfs_max_file_offset
319 * Copyright (c) 2000-2004 Silicon Graphics, Inc.
320 */
321 static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
322 {
323 unsigned int pagefactor = 1;
324 unsigned int bitshift = BITS_PER_LONG - 1;
325
326 /* Figure out maximum filesize, on Linux this can depend on
327 * the filesystem blocksize (on 32 bit platforms).
328 * __block_prepare_write does this in an [unsigned] long...
329 * page->index << (PAGE_CACHE_SHIFT - bbits)
330 * So, for page sized blocks (4K on 32 bit platforms),
331 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
332 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
333 * but for smaller blocksizes it is less (bbits = log2 bsize).
334 * Note1: get_block_t takes a long (implicit cast from above)
335 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
336 * can optionally convert the [unsigned] long from above into
337 * an [unsigned] long long.
338 */
339
340 #if BITS_PER_LONG == 32
341 # if defined(CONFIG_LBD)
342 BUILD_BUG_ON(sizeof(sector_t) != 8);
343 pagefactor = PAGE_CACHE_SIZE;
344 bitshift = BITS_PER_LONG;
345 # else
346 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
347 # endif
348 #endif
349
350 return (((unsigned long long)pagefactor) << bitshift) - 1;
351 }
352
353 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
354 {
355 int incompat_features;
356 int ret = 0;
357 unsigned long parsed_options;
358 struct ocfs2_super *osb = OCFS2_SB(sb);
359
360 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
361 ret = -EINVAL;
362 goto out;
363 }
364
365 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
366 (parsed_options & OCFS2_MOUNT_HB_LOCAL)) {
367 ret = -EINVAL;
368 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
369 goto out;
370 }
371
372 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
373 (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) {
374 ret = -EINVAL;
375 mlog(ML_ERROR, "Cannot change data mode on remount\n");
376 goto out;
377 }
378
379 /* We're going to/from readonly mode. */
380 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
381 /* Lock here so the check of HARD_RO and the potential
382 * setting of SOFT_RO is atomic. */
383 spin_lock(&osb->osb_lock);
384 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
385 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
386 ret = -EROFS;
387 goto unlock_osb;
388 }
389
390 if (*flags & MS_RDONLY) {
391 mlog(0, "Going to ro mode.\n");
392 sb->s_flags |= MS_RDONLY;
393 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
394 } else {
395 mlog(0, "Making ro filesystem writeable.\n");
396
397 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
398 mlog(ML_ERROR, "Cannot remount RDWR "
399 "filesystem due to previous errors.\n");
400 ret = -EROFS;
401 goto unlock_osb;
402 }
403 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
404 if (incompat_features) {
405 mlog(ML_ERROR, "Cannot remount RDWR because "
406 "of unsupported optional features "
407 "(%x).\n", incompat_features);
408 ret = -EINVAL;
409 goto unlock_osb;
410 }
411 sb->s_flags &= ~MS_RDONLY;
412 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
413 }
414 unlock_osb:
415 spin_unlock(&osb->osb_lock);
416 }
417
418 if (!ret) {
419 if (!ocfs2_is_hard_readonly(osb))
420 ocfs2_set_journal_params(osb);
421
422 /* Only save off the new mount options in case of a successful
423 * remount. */
424 osb->s_mount_opt = parsed_options;
425 }
426 out:
427 return ret;
428 }
429
430 static int ocfs2_sb_probe(struct super_block *sb,
431 struct buffer_head **bh,
432 int *sector_size)
433 {
434 int status = 0, tmpstat;
435 struct ocfs1_vol_disk_hdr *hdr;
436 struct ocfs2_dinode *di;
437 int blksize;
438
439 *bh = NULL;
440
441 /* may be > 512 */
442 *sector_size = bdev_hardsect_size(sb->s_bdev);
443 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
444 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
445 *sector_size, OCFS2_MAX_BLOCKSIZE);
446 status = -EINVAL;
447 goto bail;
448 }
449
450 /* Can this really happen? */
451 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
452 *sector_size = OCFS2_MIN_BLOCKSIZE;
453
454 /* check block zero for old format */
455 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
456 if (status < 0) {
457 mlog_errno(status);
458 goto bail;
459 }
460 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
461 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
462 mlog(ML_ERROR, "incompatible version: %u.%u\n",
463 hdr->major_version, hdr->minor_version);
464 status = -EINVAL;
465 }
466 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
467 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
468 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
469 hdr->signature);
470 status = -EINVAL;
471 }
472 brelse(*bh);
473 *bh = NULL;
474 if (status < 0) {
475 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
476 "upgraded before mounting with ocfs v2\n");
477 goto bail;
478 }
479
480 /*
481 * Now check at magic offset for 512, 1024, 2048, 4096
482 * blocksizes. 4096 is the maximum blocksize because it is
483 * the minimum clustersize.
484 */
485 status = -EINVAL;
486 for (blksize = *sector_size;
487 blksize <= OCFS2_MAX_BLOCKSIZE;
488 blksize <<= 1) {
489 tmpstat = ocfs2_get_sector(sb, bh,
490 OCFS2_SUPER_BLOCK_BLKNO,
491 blksize);
492 if (tmpstat < 0) {
493 status = tmpstat;
494 mlog_errno(status);
495 goto bail;
496 }
497 di = (struct ocfs2_dinode *) (*bh)->b_data;
498 status = ocfs2_verify_volume(di, *bh, blksize);
499 if (status >= 0)
500 goto bail;
501 brelse(*bh);
502 *bh = NULL;
503 if (status != -EAGAIN)
504 break;
505 }
506
507 bail:
508 return status;
509 }
510
511 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
512 {
513 if (ocfs2_mount_local(osb)) {
514 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
515 mlog(ML_ERROR, "Cannot heartbeat on a locally "
516 "mounted device.\n");
517 return -EINVAL;
518 }
519 }
520
521 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
522 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb)) {
523 mlog(ML_ERROR, "Heartbeat has to be started to mount "
524 "a read-write clustered device.\n");
525 return -EINVAL;
526 }
527 }
528
529 return 0;
530 }
531
532 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
533 {
534 struct dentry *root;
535 int status, sector_size;
536 unsigned long parsed_opt;
537 struct inode *inode = NULL;
538 struct ocfs2_super *osb = NULL;
539 struct buffer_head *bh = NULL;
540 char nodestr[8];
541
542 mlog_entry("%p, %p, %i", sb, data, silent);
543
544 if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
545 status = -EINVAL;
546 goto read_super_error;
547 }
548
549 /* for now we only have one cluster/node, make sure we see it
550 * in the heartbeat universe */
551 if (parsed_opt & OCFS2_MOUNT_HB_LOCAL) {
552 if (!o2hb_check_local_node_heartbeating()) {
553 status = -EINVAL;
554 goto read_super_error;
555 }
556 }
557
558 /* probe for superblock */
559 status = ocfs2_sb_probe(sb, &bh, &sector_size);
560 if (status < 0) {
561 mlog(ML_ERROR, "superblock probe failed!\n");
562 goto read_super_error;
563 }
564
565 status = ocfs2_initialize_super(sb, bh, sector_size);
566 osb = OCFS2_SB(sb);
567 if (status < 0) {
568 mlog_errno(status);
569 goto read_super_error;
570 }
571 brelse(bh);
572 bh = NULL;
573 osb->s_mount_opt = parsed_opt;
574
575 sb->s_magic = OCFS2_SUPER_MAGIC;
576
577 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
578 * heartbeat=none */
579 if (bdev_read_only(sb->s_bdev)) {
580 if (!(sb->s_flags & MS_RDONLY)) {
581 status = -EACCES;
582 mlog(ML_ERROR, "Readonly device detected but readonly "
583 "mount was not specified.\n");
584 goto read_super_error;
585 }
586
587 /* You should not be able to start a local heartbeat
588 * on a readonly device. */
589 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
590 status = -EROFS;
591 mlog(ML_ERROR, "Local heartbeat specified on readonly "
592 "device.\n");
593 goto read_super_error;
594 }
595
596 status = ocfs2_check_journals_nolocks(osb);
597 if (status < 0) {
598 if (status == -EROFS)
599 mlog(ML_ERROR, "Recovery required on readonly "
600 "file system, but write access is "
601 "unavailable.\n");
602 else
603 mlog_errno(status);
604 goto read_super_error;
605 }
606
607 ocfs2_set_ro_flag(osb, 1);
608
609 printk(KERN_NOTICE "Readonly device detected. No cluster "
610 "services will be utilized for this mount. Recovery "
611 "will be skipped.\n");
612 }
613
614 if (!ocfs2_is_hard_readonly(osb)) {
615 if (sb->s_flags & MS_RDONLY)
616 ocfs2_set_ro_flag(osb, 0);
617 }
618
619 status = ocfs2_verify_heartbeat(osb);
620 if (status < 0) {
621 mlog_errno(status);
622 goto read_super_error;
623 }
624
625 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
626 ocfs2_debugfs_root);
627 if (!osb->osb_debug_root) {
628 status = -EINVAL;
629 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
630 goto read_super_error;
631 }
632
633 status = ocfs2_mount_volume(sb);
634 if (osb->root_inode)
635 inode = igrab(osb->root_inode);
636
637 if (status < 0)
638 goto read_super_error;
639
640 if (!inode) {
641 status = -EIO;
642 mlog_errno(status);
643 goto read_super_error;
644 }
645
646 root = d_alloc_root(inode);
647 if (!root) {
648 status = -ENOMEM;
649 mlog_errno(status);
650 goto read_super_error;
651 }
652
653 sb->s_root = root;
654
655 ocfs2_complete_mount_recovery(osb);
656
657 if (ocfs2_mount_local(osb))
658 snprintf(nodestr, sizeof(nodestr), "local");
659 else
660 snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num);
661
662 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
663 "with %s data mode.\n",
664 osb->dev_str, nodestr, osb->slot_num,
665 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
666 "ordered");
667
668 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
669 wake_up(&osb->osb_mount_event);
670
671 mlog_exit(status);
672 return status;
673
674 read_super_error:
675 if (bh != NULL)
676 brelse(bh);
677
678 if (inode)
679 iput(inode);
680
681 if (osb) {
682 atomic_set(&osb->vol_state, VOLUME_DISABLED);
683 wake_up(&osb->osb_mount_event);
684 ocfs2_dismount_volume(sb, 1);
685 }
686
687 mlog_exit(status);
688 return status;
689 }
690
691 static int ocfs2_get_sb(struct file_system_type *fs_type,
692 int flags,
693 const char *dev_name,
694 void *data,
695 struct vfsmount *mnt)
696 {
697 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
698 mnt);
699 }
700
701 static struct file_system_type ocfs2_fs_type = {
702 .owner = THIS_MODULE,
703 .name = "ocfs2",
704 .get_sb = ocfs2_get_sb, /* is this called when we mount
705 * the fs? */
706 .kill_sb = kill_block_super, /* set to the generic one
707 * right now, but do we
708 * need to change that? */
709 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
710 .next = NULL
711 };
712
713 static int ocfs2_parse_options(struct super_block *sb,
714 char *options,
715 unsigned long *mount_opt,
716 int is_remount)
717 {
718 int status;
719 char *p;
720
721 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
722 options ? options : "(none)");
723
724 *mount_opt = 0;
725
726 if (!options) {
727 status = 1;
728 goto bail;
729 }
730
731 while ((p = strsep(&options, ",")) != NULL) {
732 int token, option;
733 substring_t args[MAX_OPT_ARGS];
734 struct ocfs2_super * osb = OCFS2_SB(sb);
735
736 if (!*p)
737 continue;
738
739 token = match_token(p, tokens, args);
740 switch (token) {
741 case Opt_hb_local:
742 *mount_opt |= OCFS2_MOUNT_HB_LOCAL;
743 break;
744 case Opt_hb_none:
745 *mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
746 break;
747 case Opt_barrier:
748 if (match_int(&args[0], &option)) {
749 status = 0;
750 goto bail;
751 }
752 if (option)
753 *mount_opt |= OCFS2_MOUNT_BARRIER;
754 else
755 *mount_opt &= ~OCFS2_MOUNT_BARRIER;
756 break;
757 case Opt_intr:
758 *mount_opt &= ~OCFS2_MOUNT_NOINTR;
759 break;
760 case Opt_nointr:
761 *mount_opt |= OCFS2_MOUNT_NOINTR;
762 break;
763 case Opt_err_panic:
764 *mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
765 break;
766 case Opt_err_ro:
767 *mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
768 break;
769 case Opt_data_ordered:
770 *mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
771 break;
772 case Opt_data_writeback:
773 *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
774 break;
775 case Opt_atime_quantum:
776 if (match_int(&args[0], &option)) {
777 status = 0;
778 goto bail;
779 }
780 if (option >= 0)
781 osb->s_atime_quantum = option;
782 else
783 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
784 break;
785 default:
786 mlog(ML_ERROR,
787 "Unrecognized mount option \"%s\" "
788 "or missing value\n", p);
789 status = 0;
790 goto bail;
791 }
792 }
793
794 status = 1;
795
796 bail:
797 mlog_exit(status);
798 return status;
799 }
800
801 static int __init ocfs2_init(void)
802 {
803 int status;
804
805 mlog_entry_void();
806
807 ocfs2_print_version();
808
809 if (init_ocfs2_extent_maps())
810 return -ENOMEM;
811
812 status = init_ocfs2_uptodate_cache();
813 if (status < 0) {
814 mlog_errno(status);
815 goto leave;
816 }
817
818 status = ocfs2_initialize_mem_caches();
819 if (status < 0) {
820 mlog_errno(status);
821 goto leave;
822 }
823
824 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
825 if (!ocfs2_wq) {
826 status = -ENOMEM;
827 goto leave;
828 }
829
830 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
831 if (!ocfs2_debugfs_root) {
832 status = -EFAULT;
833 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
834 }
835
836 leave:
837 if (status < 0) {
838 ocfs2_free_mem_caches();
839 exit_ocfs2_uptodate_cache();
840 exit_ocfs2_extent_maps();
841 }
842
843 mlog_exit(status);
844
845 if (status >= 0) {
846 return register_filesystem(&ocfs2_fs_type);
847 } else
848 return -1;
849 }
850
851 static void __exit ocfs2_exit(void)
852 {
853 mlog_entry_void();
854
855 if (ocfs2_wq) {
856 flush_workqueue(ocfs2_wq);
857 destroy_workqueue(ocfs2_wq);
858 }
859
860 debugfs_remove(ocfs2_debugfs_root);
861
862 ocfs2_free_mem_caches();
863
864 unregister_filesystem(&ocfs2_fs_type);
865
866 exit_ocfs2_extent_maps();
867
868 exit_ocfs2_uptodate_cache();
869
870 mlog_exit_void();
871 }
872
873 static void ocfs2_put_super(struct super_block *sb)
874 {
875 mlog_entry("(0x%p)\n", sb);
876
877 ocfs2_sync_blockdev(sb);
878 ocfs2_dismount_volume(sb, 0);
879
880 mlog_exit_void();
881 }
882
883 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
884 {
885 struct ocfs2_super *osb;
886 u32 numbits, freebits;
887 int status;
888 struct ocfs2_dinode *bm_lock;
889 struct buffer_head *bh = NULL;
890 struct inode *inode = NULL;
891
892 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
893
894 osb = OCFS2_SB(dentry->d_sb);
895
896 inode = ocfs2_get_system_file_inode(osb,
897 GLOBAL_BITMAP_SYSTEM_INODE,
898 OCFS2_INVALID_SLOT);
899 if (!inode) {
900 mlog(ML_ERROR, "failed to get bitmap inode\n");
901 status = -EIO;
902 goto bail;
903 }
904
905 status = ocfs2_meta_lock(inode, &bh, 0);
906 if (status < 0) {
907 mlog_errno(status);
908 goto bail;
909 }
910
911 bm_lock = (struct ocfs2_dinode *) bh->b_data;
912
913 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
914 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
915
916 buf->f_type = OCFS2_SUPER_MAGIC;
917 buf->f_bsize = dentry->d_sb->s_blocksize;
918 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
919 buf->f_blocks = ((sector_t) numbits) *
920 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
921 buf->f_bfree = ((sector_t) freebits) *
922 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
923 buf->f_bavail = buf->f_bfree;
924 buf->f_files = numbits;
925 buf->f_ffree = freebits;
926
927 brelse(bh);
928
929 ocfs2_meta_unlock(inode, 0);
930 status = 0;
931 bail:
932 if (inode)
933 iput(inode);
934
935 mlog_exit(status);
936
937 return status;
938 }
939
940 static void ocfs2_inode_init_once(void *data,
941 struct kmem_cache *cachep,
942 unsigned long flags)
943 {
944 struct ocfs2_inode_info *oi = data;
945
946 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
947 SLAB_CTOR_CONSTRUCTOR) {
948 oi->ip_flags = 0;
949 oi->ip_open_count = 0;
950 spin_lock_init(&oi->ip_lock);
951 ocfs2_extent_map_init(&oi->vfs_inode);
952 INIT_LIST_HEAD(&oi->ip_io_markers);
953 oi->ip_created_trans = 0;
954 oi->ip_last_trans = 0;
955 oi->ip_dir_start_lookup = 0;
956
957 init_rwsem(&oi->ip_alloc_sem);
958 mutex_init(&oi->ip_io_mutex);
959
960 oi->ip_blkno = 0ULL;
961 oi->ip_clusters = 0;
962
963 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
964 ocfs2_lock_res_init_once(&oi->ip_meta_lockres);
965 ocfs2_lock_res_init_once(&oi->ip_data_lockres);
966
967 ocfs2_metadata_cache_init(&oi->vfs_inode);
968
969 inode_init_once(&oi->vfs_inode);
970 }
971 }
972
973 static int ocfs2_initialize_mem_caches(void)
974 {
975 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
976 sizeof(struct ocfs2_inode_info),
977 0,
978 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
979 SLAB_MEM_SPREAD),
980 ocfs2_inode_init_once, NULL);
981 if (!ocfs2_inode_cachep)
982 return -ENOMEM;
983
984 return 0;
985 }
986
987 static void ocfs2_free_mem_caches(void)
988 {
989 if (ocfs2_inode_cachep)
990 kmem_cache_destroy(ocfs2_inode_cachep);
991
992 ocfs2_inode_cachep = NULL;
993 }
994
995 static int ocfs2_get_sector(struct super_block *sb,
996 struct buffer_head **bh,
997 int block,
998 int sect_size)
999 {
1000 if (!sb_set_blocksize(sb, sect_size)) {
1001 mlog(ML_ERROR, "unable to set blocksize\n");
1002 return -EIO;
1003 }
1004
1005 *bh = sb_getblk(sb, block);
1006 if (!*bh) {
1007 mlog_errno(-EIO);
1008 return -EIO;
1009 }
1010 lock_buffer(*bh);
1011 if (!buffer_dirty(*bh))
1012 clear_buffer_uptodate(*bh);
1013 unlock_buffer(*bh);
1014 ll_rw_block(READ, 1, bh);
1015 wait_on_buffer(*bh);
1016 return 0;
1017 }
1018
1019 /* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
1020 static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
1021 {
1022 int status;
1023
1024 /* XXX hold a ref on the node while mounte? easy enough, if
1025 * desirable. */
1026 if (ocfs2_mount_local(osb))
1027 osb->node_num = 0;
1028 else
1029 osb->node_num = o2nm_this_node();
1030
1031 if (osb->node_num == O2NM_MAX_NODES) {
1032 mlog(ML_ERROR, "could not find this host's node number\n");
1033 status = -ENOENT;
1034 goto bail;
1035 }
1036
1037 mlog(0, "I am node %d\n", osb->node_num);
1038
1039 status = 0;
1040 bail:
1041 return status;
1042 }
1043
1044 static int ocfs2_mount_volume(struct super_block *sb)
1045 {
1046 int status = 0;
1047 int unlock_super = 0;
1048 struct ocfs2_super *osb = OCFS2_SB(sb);
1049
1050 mlog_entry_void();
1051
1052 if (ocfs2_is_hard_readonly(osb))
1053 goto leave;
1054
1055 status = ocfs2_fill_local_node_info(osb);
1056 if (status < 0) {
1057 mlog_errno(status);
1058 goto leave;
1059 }
1060
1061 status = ocfs2_register_hb_callbacks(osb);
1062 if (status < 0) {
1063 mlog_errno(status);
1064 goto leave;
1065 }
1066
1067 status = ocfs2_dlm_init(osb);
1068 if (status < 0) {
1069 mlog_errno(status);
1070 goto leave;
1071 }
1072
1073 /* requires vote_thread to be running. */
1074 status = ocfs2_register_net_handlers(osb);
1075 if (status < 0) {
1076 mlog_errno(status);
1077 goto leave;
1078 }
1079
1080 status = ocfs2_super_lock(osb, 1);
1081 if (status < 0) {
1082 mlog_errno(status);
1083 goto leave;
1084 }
1085 unlock_super = 1;
1086
1087 /* This will load up the node map and add ourselves to it. */
1088 status = ocfs2_find_slot(osb);
1089 if (status < 0) {
1090 mlog_errno(status);
1091 goto leave;
1092 }
1093
1094 ocfs2_populate_mounted_map(osb);
1095
1096 /* load all node-local system inodes */
1097 status = ocfs2_init_local_system_inodes(osb);
1098 if (status < 0) {
1099 mlog_errno(status);
1100 goto leave;
1101 }
1102
1103 status = ocfs2_check_volume(osb);
1104 if (status < 0) {
1105 mlog_errno(status);
1106 goto leave;
1107 }
1108
1109 status = ocfs2_truncate_log_init(osb);
1110 if (status < 0) {
1111 mlog_errno(status);
1112 goto leave;
1113 }
1114
1115 if (ocfs2_mount_local(osb))
1116 goto leave;
1117
1118 /* This should be sent *after* we recovered our journal as it
1119 * will cause other nodes to unmark us as needing
1120 * recovery. However, we need to send it *before* dropping the
1121 * super block lock as otherwise their recovery threads might
1122 * try to clean us up while we're live! */
1123 status = ocfs2_request_mount_vote(osb);
1124 if (status < 0)
1125 mlog_errno(status);
1126
1127 leave:
1128 if (unlock_super)
1129 ocfs2_super_unlock(osb, 1);
1130
1131 mlog_exit(status);
1132 return status;
1133 }
1134
1135 /* we can't grab the goofy sem lock from inside wait_event, so we use
1136 * memory barriers to make sure that we'll see the null task before
1137 * being woken up */
1138 static int ocfs2_recovery_thread_running(struct ocfs2_super *osb)
1139 {
1140 mb();
1141 return osb->recovery_thread_task != NULL;
1142 }
1143
1144 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1145 {
1146 int tmp;
1147 struct ocfs2_super *osb = NULL;
1148 char nodestr[8];
1149
1150 mlog_entry("(0x%p)\n", sb);
1151
1152 BUG_ON(!sb);
1153 osb = OCFS2_SB(sb);
1154 BUG_ON(!osb);
1155
1156 ocfs2_shutdown_local_alloc(osb);
1157
1158 ocfs2_truncate_log_shutdown(osb);
1159
1160 /* disable any new recovery threads and wait for any currently
1161 * running ones to exit. Do this before setting the vol_state. */
1162 mutex_lock(&osb->recovery_lock);
1163 osb->disable_recovery = 1;
1164 mutex_unlock(&osb->recovery_lock);
1165 wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb));
1166
1167 /* At this point, we know that no more recovery threads can be
1168 * launched, so wait for any recovery completion work to
1169 * complete. */
1170 flush_workqueue(ocfs2_wq);
1171
1172 ocfs2_journal_shutdown(osb);
1173
1174 ocfs2_sync_blockdev(sb);
1175
1176 /* No dlm means we've failed during mount, so skip all the
1177 * steps which depended on that to complete. */
1178 if (osb->dlm) {
1179 tmp = ocfs2_super_lock(osb, 1);
1180 if (tmp < 0) {
1181 mlog_errno(tmp);
1182 return;
1183 }
1184
1185 tmp = ocfs2_request_umount_vote(osb);
1186 if (tmp < 0)
1187 mlog_errno(tmp);
1188
1189 if (osb->slot_num != OCFS2_INVALID_SLOT)
1190 ocfs2_put_slot(osb);
1191
1192 ocfs2_super_unlock(osb, 1);
1193 }
1194
1195 ocfs2_release_system_inodes(osb);
1196
1197 if (osb->dlm) {
1198 ocfs2_unregister_net_handlers(osb);
1199
1200 ocfs2_dlm_shutdown(osb);
1201 }
1202
1203 ocfs2_clear_hb_callbacks(osb);
1204
1205 debugfs_remove(osb->osb_debug_root);
1206
1207 if (!mnt_err)
1208 ocfs2_stop_heartbeat(osb);
1209
1210 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1211
1212 if (ocfs2_mount_local(osb))
1213 snprintf(nodestr, sizeof(nodestr), "local");
1214 else
1215 snprintf(nodestr, sizeof(nodestr), "%d", osb->node_num);
1216
1217 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1218 osb->dev_str, nodestr);
1219
1220 ocfs2_delete_osb(osb);
1221 kfree(osb);
1222 sb->s_dev = 0;
1223 sb->s_fs_info = NULL;
1224 }
1225
1226 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1227 unsigned uuid_bytes)
1228 {
1229 int i, ret;
1230 char *ptr;
1231
1232 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1233
1234 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1235 if (osb->uuid_str == NULL)
1236 return -ENOMEM;
1237
1238 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1239 /* print with null */
1240 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1241 if (ret != 2) /* drop super cleans up */
1242 return -EINVAL;
1243 /* then only advance past the last char */
1244 ptr += 2;
1245 }
1246
1247 return 0;
1248 }
1249
1250 static int ocfs2_initialize_super(struct super_block *sb,
1251 struct buffer_head *bh,
1252 int sector_size)
1253 {
1254 int status = 0;
1255 int i;
1256 struct ocfs2_dinode *di = NULL;
1257 struct inode *inode = NULL;
1258 struct buffer_head *bitmap_bh = NULL;
1259 struct ocfs2_journal *journal;
1260 __le32 uuid_net_key;
1261 struct ocfs2_super *osb;
1262
1263 mlog_entry_void();
1264
1265 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1266 if (!osb) {
1267 status = -ENOMEM;
1268 mlog_errno(status);
1269 goto bail;
1270 }
1271
1272 sb->s_fs_info = osb;
1273 sb->s_op = &ocfs2_sops;
1274 sb->s_export_op = &ocfs2_export_ops;
1275 sb->s_flags |= MS_NOATIME;
1276 /* this is needed to support O_LARGEFILE */
1277 sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits);
1278
1279 osb->sb = sb;
1280 /* Save off for ocfs2_rw_direct */
1281 osb->s_sectsize_bits = blksize_bits(sector_size);
1282 BUG_ON(!osb->s_sectsize_bits);
1283
1284 osb->net_response_ids = 0;
1285 spin_lock_init(&osb->net_response_lock);
1286 INIT_LIST_HEAD(&osb->net_response_list);
1287
1288 INIT_LIST_HEAD(&osb->osb_net_handlers);
1289 init_waitqueue_head(&osb->recovery_event);
1290 spin_lock_init(&osb->vote_task_lock);
1291 init_waitqueue_head(&osb->vote_event);
1292 osb->vote_work_sequence = 0;
1293 osb->vote_wake_sequence = 0;
1294 INIT_LIST_HEAD(&osb->blocked_lock_list);
1295 osb->blocked_lock_count = 0;
1296 INIT_LIST_HEAD(&osb->vote_list);
1297 spin_lock_init(&osb->osb_lock);
1298
1299 atomic_set(&osb->alloc_stats.moves, 0);
1300 atomic_set(&osb->alloc_stats.local_data, 0);
1301 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1302 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1303 atomic_set(&osb->alloc_stats.bg_extends, 0);
1304
1305 ocfs2_init_node_maps(osb);
1306
1307 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1308 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1309
1310 mutex_init(&osb->recovery_lock);
1311
1312 osb->disable_recovery = 0;
1313 osb->recovery_thread_task = NULL;
1314
1315 init_waitqueue_head(&osb->checkpoint_event);
1316 atomic_set(&osb->needs_checkpoint, 0);
1317
1318 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1319
1320 osb->node_num = O2NM_INVALID_NODE_NUM;
1321 osb->slot_num = OCFS2_INVALID_SLOT;
1322
1323 osb->local_alloc_state = OCFS2_LA_UNUSED;
1324 osb->local_alloc_bh = NULL;
1325
1326 ocfs2_setup_hb_callbacks(osb);
1327
1328 init_waitqueue_head(&osb->osb_mount_event);
1329
1330 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1331 if (!osb->vol_label) {
1332 mlog(ML_ERROR, "unable to alloc vol label\n");
1333 status = -ENOMEM;
1334 goto bail;
1335 }
1336
1337 di = (struct ocfs2_dinode *)bh->b_data;
1338
1339 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1340 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1341 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1342 osb->max_slots);
1343 status = -EINVAL;
1344 goto bail;
1345 }
1346 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1347
1348 init_waitqueue_head(&osb->osb_wipe_event);
1349 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1350 sizeof(*osb->osb_orphan_wipes),
1351 GFP_KERNEL);
1352 if (!osb->osb_orphan_wipes) {
1353 status = -ENOMEM;
1354 mlog_errno(status);
1355 goto bail;
1356 }
1357
1358 osb->s_feature_compat =
1359 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1360 osb->s_feature_ro_compat =
1361 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1362 osb->s_feature_incompat =
1363 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1364
1365 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1366 mlog(ML_ERROR, "couldn't mount because of unsupported "
1367 "optional features (%x).\n", i);
1368 status = -EINVAL;
1369 goto bail;
1370 }
1371 if (!(osb->sb->s_flags & MS_RDONLY) &&
1372 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1373 mlog(ML_ERROR, "couldn't mount RDWR because of "
1374 "unsupported optional features (%x).\n", i);
1375 status = -EINVAL;
1376 goto bail;
1377 }
1378
1379 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1380
1381 /* FIXME
1382 * This should be done in ocfs2_journal_init(), but unknown
1383 * ordering issues will cause the filesystem to crash.
1384 * If anyone wants to figure out what part of the code
1385 * refers to osb->journal before ocfs2_journal_init() is run,
1386 * be my guest.
1387 */
1388 /* initialize our journal structure */
1389
1390 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1391 if (!journal) {
1392 mlog(ML_ERROR, "unable to alloc journal\n");
1393 status = -ENOMEM;
1394 goto bail;
1395 }
1396 osb->journal = journal;
1397 journal->j_osb = osb;
1398
1399 atomic_set(&journal->j_num_trans, 0);
1400 init_rwsem(&journal->j_trans_barrier);
1401 init_waitqueue_head(&journal->j_checkpointed);
1402 spin_lock_init(&journal->j_lock);
1403 journal->j_trans_id = (unsigned long) 1;
1404 INIT_LIST_HEAD(&journal->j_la_cleanups);
1405 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1406 journal->j_state = OCFS2_JOURNAL_FREE;
1407
1408 /* get some pseudo constants for clustersize bits */
1409 osb->s_clustersize_bits =
1410 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1411 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1412 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1413
1414 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1415 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1416 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1417 osb->s_clustersize);
1418 status = -EINVAL;
1419 goto bail;
1420 }
1421
1422 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1423 > (u32)~0UL) {
1424 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1425 "what jbd can address in 32 bits.\n");
1426 status = -EINVAL;
1427 goto bail;
1428 }
1429
1430 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1431 sizeof(di->id2.i_super.s_uuid))) {
1432 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1433 status = -ENOMEM;
1434 goto bail;
1435 }
1436
1437 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1438 osb->net_key = le32_to_cpu(uuid_net_key);
1439
1440 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1441 osb->vol_label[63] = '\0';
1442 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1443 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1444 osb->first_cluster_group_blkno =
1445 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1446 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1447 mlog(0, "vol_label: %s\n", osb->vol_label);
1448 mlog(0, "uuid: %s\n", osb->uuid_str);
1449 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1450 (unsigned long long)osb->root_blkno,
1451 (unsigned long long)osb->system_dir_blkno);
1452
1453 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1454 if (!osb->osb_dlm_debug) {
1455 status = -ENOMEM;
1456 mlog_errno(status);
1457 goto bail;
1458 }
1459
1460 atomic_set(&osb->vol_state, VOLUME_INIT);
1461
1462 /* load root, system_dir, and all global system inodes */
1463 status = ocfs2_init_global_system_inodes(osb);
1464 if (status < 0) {
1465 mlog_errno(status);
1466 goto bail;
1467 }
1468
1469 /*
1470 * global bitmap
1471 */
1472 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1473 OCFS2_INVALID_SLOT);
1474 if (!inode) {
1475 status = -EINVAL;
1476 mlog_errno(status);
1477 goto bail;
1478 }
1479
1480 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1481
1482 /* We don't have a cluster lock on the bitmap here because
1483 * we're only interested in static information and the extra
1484 * complexity at mount time isn't worht it. Don't pass the
1485 * inode in to the read function though as we don't want it to
1486 * be put in the cache. */
1487 status = ocfs2_read_block(osb, osb->bitmap_blkno, &bitmap_bh, 0,
1488 NULL);
1489 iput(inode);
1490 if (status < 0) {
1491 mlog_errno(status);
1492 goto bail;
1493 }
1494
1495 di = (struct ocfs2_dinode *) bitmap_bh->b_data;
1496 osb->bitmap_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
1497 brelse(bitmap_bh);
1498 mlog(0, "cluster bitmap inode: %llu, clusters per group: %u\n",
1499 (unsigned long long)osb->bitmap_blkno, osb->bitmap_cpg);
1500
1501 status = ocfs2_init_slot_info(osb);
1502 if (status < 0) {
1503 mlog_errno(status);
1504 goto bail;
1505 }
1506
1507 bail:
1508 mlog_exit(status);
1509 return status;
1510 }
1511
1512 /*
1513 * will return: -EAGAIN if it is ok to keep searching for superblocks
1514 * -EINVAL if there is a bad superblock
1515 * 0 on success
1516 */
1517 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1518 struct buffer_head *bh,
1519 u32 blksz)
1520 {
1521 int status = -EAGAIN;
1522
1523 mlog_entry_void();
1524
1525 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1526 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1527 status = -EINVAL;
1528 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1529 mlog(ML_ERROR, "found superblock with incorrect block "
1530 "size: found %u, should be %u\n",
1531 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1532 blksz);
1533 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1534 OCFS2_MAJOR_REV_LEVEL ||
1535 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1536 OCFS2_MINOR_REV_LEVEL) {
1537 mlog(ML_ERROR, "found superblock with bad version: "
1538 "found %u.%u, should be %u.%u\n",
1539 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1540 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1541 OCFS2_MAJOR_REV_LEVEL,
1542 OCFS2_MINOR_REV_LEVEL);
1543 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1544 mlog(ML_ERROR, "bad block number on superblock: "
1545 "found %llu, should be %llu\n",
1546 (unsigned long long)di->i_blkno,
1547 (unsigned long long)bh->b_blocknr);
1548 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1549 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1550 mlog(ML_ERROR, "bad cluster size found: %u\n",
1551 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1552 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1553 mlog(ML_ERROR, "bad root_blkno: 0\n");
1554 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1555 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1556 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1557 mlog(ML_ERROR,
1558 "Superblock slots found greater than file system "
1559 "maximum: found %u, max %u\n",
1560 le16_to_cpu(di->id2.i_super.s_max_slots),
1561 OCFS2_MAX_SLOTS);
1562 } else {
1563 /* found it! */
1564 status = 0;
1565 }
1566 }
1567
1568 mlog_exit(status);
1569 return status;
1570 }
1571
1572 static int ocfs2_check_volume(struct ocfs2_super *osb)
1573 {
1574 int status = 0;
1575 int dirty;
1576 int local;
1577 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1578 * recover
1579 * ourselves. */
1580
1581 mlog_entry_void();
1582
1583 /* Init our journal object. */
1584 status = ocfs2_journal_init(osb->journal, &dirty);
1585 if (status < 0) {
1586 mlog(ML_ERROR, "Could not initialize journal!\n");
1587 goto finally;
1588 }
1589
1590 /* If the journal was unmounted cleanly then we don't want to
1591 * recover anything. Otherwise, journal_load will do that
1592 * dirty work for us :) */
1593 if (!dirty) {
1594 status = ocfs2_journal_wipe(osb->journal, 0);
1595 if (status < 0) {
1596 mlog_errno(status);
1597 goto finally;
1598 }
1599 } else {
1600 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1601 "recovering volume.\n");
1602 }
1603
1604 local = ocfs2_mount_local(osb);
1605
1606 /* will play back anything left in the journal. */
1607 ocfs2_journal_load(osb->journal, local);
1608
1609 if (dirty) {
1610 /* recover my local alloc if we didn't unmount cleanly. */
1611 status = ocfs2_begin_local_alloc_recovery(osb,
1612 osb->slot_num,
1613 &local_alloc);
1614 if (status < 0) {
1615 mlog_errno(status);
1616 goto finally;
1617 }
1618 /* we complete the recovery process after we've marked
1619 * ourselves as mounted. */
1620 }
1621
1622 mlog(0, "Journal loaded.\n");
1623
1624 status = ocfs2_load_local_alloc(osb);
1625 if (status < 0) {
1626 mlog_errno(status);
1627 goto finally;
1628 }
1629
1630 if (dirty) {
1631 /* Recovery will be completed after we've mounted the
1632 * rest of the volume. */
1633 osb->dirty = 1;
1634 osb->local_alloc_copy = local_alloc;
1635 local_alloc = NULL;
1636 }
1637
1638 /* go through each journal, trylock it and if you get the
1639 * lock, and it's marked as dirty, set the bit in the recover
1640 * map and launch a recovery thread for it. */
1641 status = ocfs2_mark_dead_nodes(osb);
1642 if (status < 0)
1643 mlog_errno(status);
1644
1645 finally:
1646 if (local_alloc)
1647 kfree(local_alloc);
1648
1649 mlog_exit(status);
1650 return status;
1651 }
1652
1653 /*
1654 * The routine gets called from dismount or close whenever a dismount on
1655 * volume is requested and the osb open count becomes 1.
1656 * It will remove the osb from the global list and also free up all the
1657 * initialized resources and fileobject.
1658 */
1659 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1660 {
1661 mlog_entry_void();
1662
1663 /* This function assumes that the caller has the main osb resource */
1664
1665 if (osb->slot_info)
1666 ocfs2_free_slot_info(osb->slot_info);
1667
1668 kfree(osb->osb_orphan_wipes);
1669 /* FIXME
1670 * This belongs in journal shutdown, but because we have to
1671 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1672 * we free it here.
1673 */
1674 kfree(osb->journal);
1675 if (osb->local_alloc_copy)
1676 kfree(osb->local_alloc_copy);
1677 kfree(osb->uuid_str);
1678 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1679 memset(osb, 0, sizeof(struct ocfs2_super));
1680
1681 mlog_exit_void();
1682 }
1683
1684 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1685 * panic(). We do not support continue-on-error operation. */
1686 static void ocfs2_handle_error(struct super_block *sb)
1687 {
1688 struct ocfs2_super *osb = OCFS2_SB(sb);
1689
1690 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1691 panic("OCFS2: (device %s): panic forced after error\n",
1692 sb->s_id);
1693
1694 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1695
1696 if (sb->s_flags & MS_RDONLY &&
1697 (ocfs2_is_soft_readonly(osb) ||
1698 ocfs2_is_hard_readonly(osb)))
1699 return;
1700
1701 printk(KERN_CRIT "File system is now read-only due to the potential "
1702 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1703 "system is unmounted.\n");
1704 sb->s_flags |= MS_RDONLY;
1705 ocfs2_set_ro_flag(osb, 0);
1706 }
1707
1708 static char error_buf[1024];
1709
1710 void __ocfs2_error(struct super_block *sb,
1711 const char *function,
1712 const char *fmt, ...)
1713 {
1714 va_list args;
1715
1716 va_start(args, fmt);
1717 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1718 va_end(args);
1719
1720 /* Not using mlog here because we want to show the actual
1721 * function the error came from. */
1722 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1723 sb->s_id, function, error_buf);
1724
1725 ocfs2_handle_error(sb);
1726 }
1727
1728 /* Handle critical errors. This is intentionally more drastic than
1729 * ocfs2_handle_error, so we only use for things like journal errors,
1730 * etc. */
1731 void __ocfs2_abort(struct super_block* sb,
1732 const char *function,
1733 const char *fmt, ...)
1734 {
1735 va_list args;
1736
1737 va_start(args, fmt);
1738 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1739 va_end(args);
1740
1741 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1742 sb->s_id, function, error_buf);
1743
1744 /* We don't have the cluster support yet to go straight to
1745 * hard readonly in here. Until then, we want to keep
1746 * ocfs2_abort() so that we can at least mark critical
1747 * errors.
1748 *
1749 * TODO: This should abort the journal and alert other nodes
1750 * that our slot needs recovery. */
1751
1752 /* Force a panic(). This stinks, but it's better than letting
1753 * things continue without having a proper hard readonly
1754 * here. */
1755 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1756 ocfs2_handle_error(sb);
1757 }
1758
1759 module_init(ocfs2_init);
1760 module_exit(ocfs2_exit);
Linux 2.6 ibm-acpi/thinkpad-acpi driver git tree