]> git.proxmox.com Git - mirror_ubuntu-hirsute-kernel.git/blob - fs/ocfs2/super.c
projects / mirror_ubuntu-hirsute-kernel.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ocfs2: remove unused ocfs2_handle_add_lock()
[mirror_ubuntu-hirsute-kernel.git] / fs / ocfs2 / super.c
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 kmem_cache_t *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 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_err,
143 };
144
145 static match_table_t tokens = {
146 {Opt_barrier, "barrier=%u"},
147 {Opt_err_panic, "errors=panic"},
148 {Opt_err_ro, "errors=remount-ro"},
149 {Opt_intr, "intr"},
150 {Opt_nointr, "nointr"},
151 {Opt_hb_none, OCFS2_HB_NONE},
152 {Opt_hb_local, OCFS2_HB_LOCAL},
153 {Opt_data_ordered, "data=ordered"},
154 {Opt_data_writeback, "data=writeback"},
155 {Opt_err, NULL}
156 };
157
158 /*
159 * write_super and sync_fs ripped right out of ext3.
160 */
161 static void ocfs2_write_super(struct super_block *sb)
162 {
163 if (mutex_trylock(&sb->s_lock) != 0)
164 BUG();
165 sb->s_dirt = 0;
166 }
167
168 static int ocfs2_sync_fs(struct super_block *sb, int wait)
169 {
170 int status = 0;
171 tid_t target;
172 struct ocfs2_super *osb = OCFS2_SB(sb);
173
174 sb->s_dirt = 0;
175
176 if (ocfs2_is_hard_readonly(osb))
177 return -EROFS;
178
179 if (wait) {
180 status = ocfs2_flush_truncate_log(osb);
181 if (status < 0)
182 mlog_errno(status);
183 } else {
184 ocfs2_schedule_truncate_log_flush(osb, 0);
185 }
186
187 if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
188 if (wait)
189 log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
190 target);
191 }
192 return 0;
193 }
194
195 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
196 {
197 struct inode *new = NULL;
198 int status = 0;
199 int i;
200
201 mlog_entry_void();
202
203 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE);
204 if (IS_ERR(new)) {
205 status = PTR_ERR(new);
206 mlog_errno(status);
207 goto bail;
208 }
209 osb->root_inode = new;
210
211 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE);
212 if (IS_ERR(new)) {
213 status = PTR_ERR(new);
214 mlog_errno(status);
215 goto bail;
216 }
217 osb->sys_root_inode = new;
218
219 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
220 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
221 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
222 if (!new) {
223 ocfs2_release_system_inodes(osb);
224 status = -EINVAL;
225 mlog_errno(status);
226 /* FIXME: Should ERROR_RO_FS */
227 mlog(ML_ERROR, "Unable to load system inode %d, "
228 "possibly corrupt fs?", i);
229 goto bail;
230 }
231 // the array now has one ref, so drop this one
232 iput(new);
233 }
234
235 bail:
236 mlog_exit(status);
237 return status;
238 }
239
240 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
241 {
242 struct inode *new = NULL;
243 int status = 0;
244 int i;
245
246 mlog_entry_void();
247
248 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
249 i < NUM_SYSTEM_INODES;
250 i++) {
251 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
252 if (!new) {
253 ocfs2_release_system_inodes(osb);
254 status = -EINVAL;
255 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
256 status, i, osb->slot_num);
257 goto bail;
258 }
259 /* the array now has one ref, so drop this one */
260 iput(new);
261 }
262
263 bail:
264 mlog_exit(status);
265 return status;
266 }
267
268 static int ocfs2_release_system_inodes(struct ocfs2_super *osb)
269 {
270 int status = 0, i;
271 struct inode *inode;
272
273 mlog_entry_void();
274
275 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
276 inode = osb->system_inodes[i];
277 if (inode) {
278 iput(inode);
279 osb->system_inodes[i] = NULL;
280 }
281 }
282
283 inode = osb->sys_root_inode;
284 if (inode) {
285 iput(inode);
286 osb->sys_root_inode = NULL;
287 }
288
289 inode = osb->root_inode;
290 if (inode) {
291 iput(inode);
292 osb->root_inode = NULL;
293 }
294
295 mlog_exit(status);
296 return status;
297 }
298
299 /* We're allocating fs objects, use GFP_NOFS */
300 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
301 {
302 struct ocfs2_inode_info *oi;
303
304 oi = kmem_cache_alloc(ocfs2_inode_cachep, SLAB_NOFS);
305 if (!oi)
306 return NULL;
307
308 return &oi->vfs_inode;
309 }
310
311 static void ocfs2_destroy_inode(struct inode *inode)
312 {
313 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
314 }
315
316 /* From xfs_super.c:xfs_max_file_offset
317 * Copyright (c) 2000-2004 Silicon Graphics, Inc.
318 */
319 static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
320 {
321 unsigned int pagefactor = 1;
322 unsigned int bitshift = BITS_PER_LONG - 1;
323
324 /* Figure out maximum filesize, on Linux this can depend on
325 * the filesystem blocksize (on 32 bit platforms).
326 * __block_prepare_write does this in an [unsigned] long...
327 * page->index << (PAGE_CACHE_SHIFT - bbits)
328 * So, for page sized blocks (4K on 32 bit platforms),
329 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
330 * (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
331 * but for smaller blocksizes it is less (bbits = log2 bsize).
332 * Note1: get_block_t takes a long (implicit cast from above)
333 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
334 * can optionally convert the [unsigned] long from above into
335 * an [unsigned] long long.
336 */
337
338 #if BITS_PER_LONG == 32
339 # if defined(CONFIG_LBD)
340 BUILD_BUG_ON(sizeof(sector_t) != 8);
341 pagefactor = PAGE_CACHE_SIZE;
342 bitshift = BITS_PER_LONG;
343 # else
344 pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
345 # endif
346 #endif
347
348 return (((unsigned long long)pagefactor) << bitshift) - 1;
349 }
350
351 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
352 {
353 int incompat_features;
354 int ret = 0;
355 unsigned long parsed_options;
356 struct ocfs2_super *osb = OCFS2_SB(sb);
357
358 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
359 ret = -EINVAL;
360 goto out;
361 }
362
363 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
364 (parsed_options & OCFS2_MOUNT_HB_LOCAL)) {
365 ret = -EINVAL;
366 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
367 goto out;
368 }
369
370 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
371 (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) {
372 ret = -EINVAL;
373 mlog(ML_ERROR, "Cannot change data mode on remount\n");
374 goto out;
375 }
376
377 /* We're going to/from readonly mode. */
378 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
379 /* Lock here so the check of HARD_RO and the potential
380 * setting of SOFT_RO is atomic. */
381 spin_lock(&osb->osb_lock);
382 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
383 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
384 ret = -EROFS;
385 goto unlock_osb;
386 }
387
388 if (*flags & MS_RDONLY) {
389 mlog(0, "Going to ro mode.\n");
390 sb->s_flags |= MS_RDONLY;
391 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
392 } else {
393 mlog(0, "Making ro filesystem writeable.\n");
394
395 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
396 mlog(ML_ERROR, "Cannot remount RDWR "
397 "filesystem due to previous errors.\n");
398 ret = -EROFS;
399 goto unlock_osb;
400 }
401 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
402 if (incompat_features) {
403 mlog(ML_ERROR, "Cannot remount RDWR because "
404 "of unsupported optional features "
405 "(%x).\n", incompat_features);
406 ret = -EINVAL;
407 goto unlock_osb;
408 }
409 sb->s_flags &= ~MS_RDONLY;
410 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
411 }
412 unlock_osb:
413 spin_unlock(&osb->osb_lock);
414 }
415
416 if (!ret) {
417 if (!ocfs2_is_hard_readonly(osb))
418 ocfs2_set_journal_params(osb);
419
420 /* Only save off the new mount options in case of a successful
421 * remount. */
422 osb->s_mount_opt = parsed_options;
423 }
424 out:
425 return ret;
426 }
427
428 static int ocfs2_sb_probe(struct super_block *sb,
429 struct buffer_head **bh,
430 int *sector_size)
431 {
432 int status = 0, tmpstat;
433 struct ocfs1_vol_disk_hdr *hdr;
434 struct ocfs2_dinode *di;
435 int blksize;
436
437 *bh = NULL;
438
439 /* may be > 512 */
440 *sector_size = bdev_hardsect_size(sb->s_bdev);
441 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
442 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
443 *sector_size, OCFS2_MAX_BLOCKSIZE);
444 status = -EINVAL;
445 goto bail;
446 }
447
448 /* Can this really happen? */
449 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
450 *sector_size = OCFS2_MIN_BLOCKSIZE;
451
452 /* check block zero for old format */
453 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
454 if (status < 0) {
455 mlog_errno(status);
456 goto bail;
457 }
458 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
459 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
460 mlog(ML_ERROR, "incompatible version: %u.%u\n",
461 hdr->major_version, hdr->minor_version);
462 status = -EINVAL;
463 }
464 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
465 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
466 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
467 hdr->signature);
468 status = -EINVAL;
469 }
470 brelse(*bh);
471 *bh = NULL;
472 if (status < 0) {
473 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
474 "upgraded before mounting with ocfs v2\n");
475 goto bail;
476 }
477
478 /*
479 * Now check at magic offset for 512, 1024, 2048, 4096
480 * blocksizes. 4096 is the maximum blocksize because it is
481 * the minimum clustersize.
482 */
483 status = -EINVAL;
484 for (blksize = *sector_size;
485 blksize <= OCFS2_MAX_BLOCKSIZE;
486 blksize <<= 1) {
487 tmpstat = ocfs2_get_sector(sb, bh,
488 OCFS2_SUPER_BLOCK_BLKNO,
489 blksize);
490 if (tmpstat < 0) {
491 status = tmpstat;
492 mlog_errno(status);
493 goto bail;
494 }
495 di = (struct ocfs2_dinode *) (*bh)->b_data;
496 status = ocfs2_verify_volume(di, *bh, blksize);
497 if (status >= 0)
498 goto bail;
499 brelse(*bh);
500 *bh = NULL;
501 if (status != -EAGAIN)
502 break;
503 }
504
505 bail:
506 return status;
507 }
508
509 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
510 {
511 struct dentry *root;
512 int status, sector_size;
513 unsigned long parsed_opt;
514 struct inode *inode = NULL;
515 struct ocfs2_super *osb = NULL;
516 struct buffer_head *bh = NULL;
517
518 mlog_entry("%p, %p, %i", sb, data, silent);
519
520 /* for now we only have one cluster/node, make sure we see it
521 * in the heartbeat universe */
522 if (!o2hb_check_local_node_heartbeating()) {
523 status = -EINVAL;
524 goto read_super_error;
525 }
526
527 /* probe for superblock */
528 status = ocfs2_sb_probe(sb, &bh, &sector_size);
529 if (status < 0) {
530 mlog(ML_ERROR, "superblock probe failed!\n");
531 goto read_super_error;
532 }
533
534 status = ocfs2_initialize_super(sb, bh, sector_size);
535 osb = OCFS2_SB(sb);
536 if (status < 0) {
537 mlog_errno(status);
538 goto read_super_error;
539 }
540 brelse(bh);
541 bh = NULL;
542
543 if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
544 status = -EINVAL;
545 goto read_super_error;
546 }
547 osb->s_mount_opt = parsed_opt;
548
549 sb->s_magic = OCFS2_SUPER_MAGIC;
550
551 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
552 * heartbeat=none */
553 if (bdev_read_only(sb->s_bdev)) {
554 if (!(sb->s_flags & MS_RDONLY)) {
555 status = -EACCES;
556 mlog(ML_ERROR, "Readonly device detected but readonly "
557 "mount was not specified.\n");
558 goto read_super_error;
559 }
560
561 /* You should not be able to start a local heartbeat
562 * on a readonly device. */
563 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
564 status = -EROFS;
565 mlog(ML_ERROR, "Local heartbeat specified on readonly "
566 "device.\n");
567 goto read_super_error;
568 }
569
570 status = ocfs2_check_journals_nolocks(osb);
571 if (status < 0) {
572 if (status == -EROFS)
573 mlog(ML_ERROR, "Recovery required on readonly "
574 "file system, but write access is "
575 "unavailable.\n");
576 else
577 mlog_errno(status);
578 goto read_super_error;
579 }
580
581 ocfs2_set_ro_flag(osb, 1);
582
583 printk(KERN_NOTICE "Readonly device detected. No cluster "
584 "services will be utilized for this mount. Recovery "
585 "will be skipped.\n");
586 }
587
588 if (!ocfs2_is_hard_readonly(osb)) {
589 /* If this isn't a hard readonly mount, then we need
590 * to make sure that heartbeat is in a valid state,
591 * and that we mark ourselves soft readonly is -oro
592 * was specified. */
593 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
594 mlog(ML_ERROR, "No heartbeat for device (%s)\n",
595 sb->s_id);
596 status = -EINVAL;
597 goto read_super_error;
598 }
599
600 if (sb->s_flags & MS_RDONLY)
601 ocfs2_set_ro_flag(osb, 0);
602 }
603
604 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
605 ocfs2_debugfs_root);
606 if (!osb->osb_debug_root) {
607 status = -EINVAL;
608 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
609 goto read_super_error;
610 }
611
612 status = ocfs2_mount_volume(sb);
613 if (osb->root_inode)
614 inode = igrab(osb->root_inode);
615
616 if (status < 0)
617 goto read_super_error;
618
619 if (!inode) {
620 status = -EIO;
621 mlog_errno(status);
622 goto read_super_error;
623 }
624
625 root = d_alloc_root(inode);
626 if (!root) {
627 status = -ENOMEM;
628 mlog_errno(status);
629 goto read_super_error;
630 }
631
632 sb->s_root = root;
633
634 ocfs2_complete_mount_recovery(osb);
635
636 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %d, slot %d) "
637 "with %s data mode.\n",
638 osb->dev_str, osb->node_num, osb->slot_num,
639 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
640 "ordered");
641
642 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
643 wake_up(&osb->osb_mount_event);
644
645 mlog_exit(status);
646 return status;
647
648 read_super_error:
649 if (bh != NULL)
650 brelse(bh);
651
652 if (inode)
653 iput(inode);
654
655 if (osb) {
656 atomic_set(&osb->vol_state, VOLUME_DISABLED);
657 wake_up(&osb->osb_mount_event);
658 ocfs2_dismount_volume(sb, 1);
659 }
660
661 mlog_exit(status);
662 return status;
663 }
664
665 static int ocfs2_get_sb(struct file_system_type *fs_type,
666 int flags,
667 const char *dev_name,
668 void *data,
669 struct vfsmount *mnt)
670 {
671 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
672 mnt);
673 }
674
675 static struct file_system_type ocfs2_fs_type = {
676 .owner = THIS_MODULE,
677 .name = "ocfs2",
678 .get_sb = ocfs2_get_sb, /* is this called when we mount
679 * the fs? */
680 .kill_sb = kill_block_super, /* set to the generic one
681 * right now, but do we
682 * need to change that? */
683 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
684 .next = NULL
685 };
686
687 static int ocfs2_parse_options(struct super_block *sb,
688 char *options,
689 unsigned long *mount_opt,
690 int is_remount)
691 {
692 int status;
693 char *p;
694
695 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
696 options ? options : "(none)");
697
698 *mount_opt = 0;
699
700 if (!options) {
701 status = 1;
702 goto bail;
703 }
704
705 while ((p = strsep(&options, ",")) != NULL) {
706 int token, option;
707 substring_t args[MAX_OPT_ARGS];
708
709 if (!*p)
710 continue;
711
712 token = match_token(p, tokens, args);
713 switch (token) {
714 case Opt_hb_local:
715 *mount_opt |= OCFS2_MOUNT_HB_LOCAL;
716 break;
717 case Opt_hb_none:
718 *mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
719 break;
720 case Opt_barrier:
721 if (match_int(&args[0], &option)) {
722 status = 0;
723 goto bail;
724 }
725 if (option)
726 *mount_opt |= OCFS2_MOUNT_BARRIER;
727 else
728 *mount_opt &= ~OCFS2_MOUNT_BARRIER;
729 break;
730 case Opt_intr:
731 *mount_opt &= ~OCFS2_MOUNT_NOINTR;
732 break;
733 case Opt_nointr:
734 *mount_opt |= OCFS2_MOUNT_NOINTR;
735 break;
736 case Opt_err_panic:
737 *mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
738 break;
739 case Opt_err_ro:
740 *mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
741 break;
742 case Opt_data_ordered:
743 *mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
744 break;
745 case Opt_data_writeback:
746 *mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
747 break;
748 default:
749 mlog(ML_ERROR,
750 "Unrecognized mount option \"%s\" "
751 "or missing value\n", p);
752 status = 0;
753 goto bail;
754 }
755 }
756
757 status = 1;
758
759 bail:
760 mlog_exit(status);
761 return status;
762 }
763
764 static int __init ocfs2_init(void)
765 {
766 int status;
767
768 mlog_entry_void();
769
770 ocfs2_print_version();
771
772 if (init_ocfs2_extent_maps())
773 return -ENOMEM;
774
775 status = init_ocfs2_uptodate_cache();
776 if (status < 0) {
777 mlog_errno(status);
778 goto leave;
779 }
780
781 status = ocfs2_initialize_mem_caches();
782 if (status < 0) {
783 mlog_errno(status);
784 goto leave;
785 }
786
787 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
788 if (!ocfs2_wq) {
789 status = -ENOMEM;
790 goto leave;
791 }
792
793 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
794 if (!ocfs2_debugfs_root) {
795 status = -EFAULT;
796 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
797 }
798
799 leave:
800 if (status < 0) {
801 ocfs2_free_mem_caches();
802 exit_ocfs2_uptodate_cache();
803 exit_ocfs2_extent_maps();
804 }
805
806 mlog_exit(status);
807
808 if (status >= 0) {
809 return register_filesystem(&ocfs2_fs_type);
810 } else
811 return -1;
812 }
813
814 static void __exit ocfs2_exit(void)
815 {
816 mlog_entry_void();
817
818 if (ocfs2_wq) {
819 flush_workqueue(ocfs2_wq);
820 destroy_workqueue(ocfs2_wq);
821 }
822
823 debugfs_remove(ocfs2_debugfs_root);
824
825 ocfs2_free_mem_caches();
826
827 unregister_filesystem(&ocfs2_fs_type);
828
829 exit_ocfs2_extent_maps();
830
831 exit_ocfs2_uptodate_cache();
832
833 mlog_exit_void();
834 }
835
836 static void ocfs2_put_super(struct super_block *sb)
837 {
838 mlog_entry("(0x%p)\n", sb);
839
840 ocfs2_sync_blockdev(sb);
841 ocfs2_dismount_volume(sb, 0);
842
843 mlog_exit_void();
844 }
845
846 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
847 {
848 struct ocfs2_super *osb;
849 u32 numbits, freebits;
850 int status;
851 struct ocfs2_dinode *bm_lock;
852 struct buffer_head *bh = NULL;
853 struct inode *inode = NULL;
854
855 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
856
857 osb = OCFS2_SB(dentry->d_sb);
858
859 inode = ocfs2_get_system_file_inode(osb,
860 GLOBAL_BITMAP_SYSTEM_INODE,
861 OCFS2_INVALID_SLOT);
862 if (!inode) {
863 mlog(ML_ERROR, "failed to get bitmap inode\n");
864 status = -EIO;
865 goto bail;
866 }
867
868 status = ocfs2_meta_lock(inode, NULL, &bh, 0);
869 if (status < 0) {
870 mlog_errno(status);
871 goto bail;
872 }
873
874 bm_lock = (struct ocfs2_dinode *) bh->b_data;
875
876 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
877 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
878
879 buf->f_type = OCFS2_SUPER_MAGIC;
880 buf->f_bsize = dentry->d_sb->s_blocksize;
881 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
882 buf->f_blocks = ((sector_t) numbits) *
883 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
884 buf->f_bfree = ((sector_t) freebits) *
885 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
886 buf->f_bavail = buf->f_bfree;
887 buf->f_files = numbits;
888 buf->f_ffree = freebits;
889
890 brelse(bh);
891
892 ocfs2_meta_unlock(inode, 0);
893 status = 0;
894 bail:
895 if (inode)
896 iput(inode);
897
898 mlog_exit(status);
899
900 return status;
901 }
902
903 static void ocfs2_inode_init_once(void *data,
904 kmem_cache_t *cachep,
905 unsigned long flags)
906 {
907 struct ocfs2_inode_info *oi = data;
908
909 if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
910 SLAB_CTOR_CONSTRUCTOR) {
911 oi->ip_flags = 0;
912 oi->ip_open_count = 0;
913 spin_lock_init(&oi->ip_lock);
914 ocfs2_extent_map_init(&oi->vfs_inode);
915 INIT_LIST_HEAD(&oi->ip_io_markers);
916 oi->ip_created_trans = 0;
917 oi->ip_last_trans = 0;
918 oi->ip_dir_start_lookup = 0;
919
920 init_rwsem(&oi->ip_alloc_sem);
921 mutex_init(&oi->ip_io_mutex);
922
923 oi->ip_blkno = 0ULL;
924 oi->ip_clusters = 0;
925
926 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
927 ocfs2_lock_res_init_once(&oi->ip_meta_lockres);
928 ocfs2_lock_res_init_once(&oi->ip_data_lockres);
929
930 ocfs2_metadata_cache_init(&oi->vfs_inode);
931
932 inode_init_once(&oi->vfs_inode);
933 }
934 }
935
936 static int ocfs2_initialize_mem_caches(void)
937 {
938 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
939 sizeof(struct ocfs2_inode_info),
940 0,
941 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
942 SLAB_MEM_SPREAD),
943 ocfs2_inode_init_once, NULL);
944 if (!ocfs2_inode_cachep)
945 return -ENOMEM;
946
947 return 0;
948 }
949
950 static void ocfs2_free_mem_caches(void)
951 {
952 if (ocfs2_inode_cachep)
953 kmem_cache_destroy(ocfs2_inode_cachep);
954
955 ocfs2_inode_cachep = NULL;
956 }
957
958 static int ocfs2_get_sector(struct super_block *sb,
959 struct buffer_head **bh,
960 int block,
961 int sect_size)
962 {
963 if (!sb_set_blocksize(sb, sect_size)) {
964 mlog(ML_ERROR, "unable to set blocksize\n");
965 return -EIO;
966 }
967
968 *bh = sb_getblk(sb, block);
969 if (!*bh) {
970 mlog_errno(-EIO);
971 return -EIO;
972 }
973 lock_buffer(*bh);
974 if (!buffer_dirty(*bh))
975 clear_buffer_uptodate(*bh);
976 unlock_buffer(*bh);
977 ll_rw_block(READ, 1, bh);
978 wait_on_buffer(*bh);
979 return 0;
980 }
981
982 /* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
983 static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
984 {
985 int status;
986
987 /* XXX hold a ref on the node while mounte? easy enough, if
988 * desirable. */
989 osb->node_num = o2nm_this_node();
990 if (osb->node_num == O2NM_MAX_NODES) {
991 mlog(ML_ERROR, "could not find this host's node number\n");
992 status = -ENOENT;
993 goto bail;
994 }
995
996 mlog(0, "I am node %d\n", osb->node_num);
997
998 status = 0;
999 bail:
1000 return status;
1001 }
1002
1003 static int ocfs2_mount_volume(struct super_block *sb)
1004 {
1005 int status = 0;
1006 int unlock_super = 0;
1007 struct ocfs2_super *osb = OCFS2_SB(sb);
1008
1009 mlog_entry_void();
1010
1011 if (ocfs2_is_hard_readonly(osb))
1012 goto leave;
1013
1014 status = ocfs2_fill_local_node_info(osb);
1015 if (status < 0) {
1016 mlog_errno(status);
1017 goto leave;
1018 }
1019
1020 status = ocfs2_register_hb_callbacks(osb);
1021 if (status < 0) {
1022 mlog_errno(status);
1023 goto leave;
1024 }
1025
1026 status = ocfs2_dlm_init(osb);
1027 if (status < 0) {
1028 mlog_errno(status);
1029 goto leave;
1030 }
1031
1032 /* requires vote_thread to be running. */
1033 status = ocfs2_register_net_handlers(osb);
1034 if (status < 0) {
1035 mlog_errno(status);
1036 goto leave;
1037 }
1038
1039 status = ocfs2_super_lock(osb, 1);
1040 if (status < 0) {
1041 mlog_errno(status);
1042 goto leave;
1043 }
1044 unlock_super = 1;
1045
1046 /* This will load up the node map and add ourselves to it. */
1047 status = ocfs2_find_slot(osb);
1048 if (status < 0) {
1049 mlog_errno(status);
1050 goto leave;
1051 }
1052
1053 ocfs2_populate_mounted_map(osb);
1054
1055 /* load all node-local system inodes */
1056 status = ocfs2_init_local_system_inodes(osb);
1057 if (status < 0) {
1058 mlog_errno(status);
1059 goto leave;
1060 }
1061
1062 status = ocfs2_check_volume(osb);
1063 if (status < 0) {
1064 mlog_errno(status);
1065 goto leave;
1066 }
1067
1068 status = ocfs2_truncate_log_init(osb);
1069 if (status < 0) {
1070 mlog_errno(status);
1071 goto leave;
1072 }
1073
1074 /* This should be sent *after* we recovered our journal as it
1075 * will cause other nodes to unmark us as needing
1076 * recovery. However, we need to send it *before* dropping the
1077 * super block lock as otherwise their recovery threads might
1078 * try to clean us up while we're live! */
1079 status = ocfs2_request_mount_vote(osb);
1080 if (status < 0)
1081 mlog_errno(status);
1082
1083 leave:
1084 if (unlock_super)
1085 ocfs2_super_unlock(osb, 1);
1086
1087 mlog_exit(status);
1088 return status;
1089 }
1090
1091 /* we can't grab the goofy sem lock from inside wait_event, so we use
1092 * memory barriers to make sure that we'll see the null task before
1093 * being woken up */
1094 static int ocfs2_recovery_thread_running(struct ocfs2_super *osb)
1095 {
1096 mb();
1097 return osb->recovery_thread_task != NULL;
1098 }
1099
1100 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1101 {
1102 int tmp;
1103 struct ocfs2_super *osb = NULL;
1104
1105 mlog_entry("(0x%p)\n", sb);
1106
1107 BUG_ON(!sb);
1108 osb = OCFS2_SB(sb);
1109 BUG_ON(!osb);
1110
1111 ocfs2_shutdown_local_alloc(osb);
1112
1113 ocfs2_truncate_log_shutdown(osb);
1114
1115 /* disable any new recovery threads and wait for any currently
1116 * running ones to exit. Do this before setting the vol_state. */
1117 mutex_lock(&osb->recovery_lock);
1118 osb->disable_recovery = 1;
1119 mutex_unlock(&osb->recovery_lock);
1120 wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb));
1121
1122 /* At this point, we know that no more recovery threads can be
1123 * launched, so wait for any recovery completion work to
1124 * complete. */
1125 flush_workqueue(ocfs2_wq);
1126
1127 ocfs2_journal_shutdown(osb);
1128
1129 ocfs2_sync_blockdev(sb);
1130
1131 /* No dlm means we've failed during mount, so skip all the
1132 * steps which depended on that to complete. */
1133 if (osb->dlm) {
1134 tmp = ocfs2_super_lock(osb, 1);
1135 if (tmp < 0) {
1136 mlog_errno(tmp);
1137 return;
1138 }
1139
1140 tmp = ocfs2_request_umount_vote(osb);
1141 if (tmp < 0)
1142 mlog_errno(tmp);
1143
1144 if (osb->slot_num != OCFS2_INVALID_SLOT)
1145 ocfs2_put_slot(osb);
1146
1147 ocfs2_super_unlock(osb, 1);
1148 }
1149
1150 ocfs2_release_system_inodes(osb);
1151
1152 if (osb->dlm) {
1153 ocfs2_unregister_net_handlers(osb);
1154
1155 ocfs2_dlm_shutdown(osb);
1156 }
1157
1158 ocfs2_clear_hb_callbacks(osb);
1159
1160 debugfs_remove(osb->osb_debug_root);
1161
1162 if (!mnt_err)
1163 ocfs2_stop_heartbeat(osb);
1164
1165 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1166
1167 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %d)\n",
1168 osb->dev_str, osb->node_num);
1169
1170 ocfs2_delete_osb(osb);
1171 kfree(osb);
1172 sb->s_dev = 0;
1173 sb->s_fs_info = NULL;
1174 }
1175
1176 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1177 unsigned uuid_bytes)
1178 {
1179 int i, ret;
1180 char *ptr;
1181
1182 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1183
1184 osb->uuid_str = kcalloc(1, OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1185 if (osb->uuid_str == NULL)
1186 return -ENOMEM;
1187
1188 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1189 /* print with null */
1190 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1191 if (ret != 2) /* drop super cleans up */
1192 return -EINVAL;
1193 /* then only advance past the last char */
1194 ptr += 2;
1195 }
1196
1197 return 0;
1198 }
1199
1200 static int ocfs2_initialize_super(struct super_block *sb,
1201 struct buffer_head *bh,
1202 int sector_size)
1203 {
1204 int status = 0;
1205 int i;
1206 struct ocfs2_dinode *di = NULL;
1207 struct inode *inode = NULL;
1208 struct buffer_head *bitmap_bh = NULL;
1209 struct ocfs2_journal *journal;
1210 __le32 uuid_net_key;
1211 struct ocfs2_super *osb;
1212
1213 mlog_entry_void();
1214
1215 osb = kcalloc(1, sizeof(struct ocfs2_super), GFP_KERNEL);
1216 if (!osb) {
1217 status = -ENOMEM;
1218 mlog_errno(status);
1219 goto bail;
1220 }
1221
1222 sb->s_fs_info = osb;
1223 sb->s_op = &ocfs2_sops;
1224 sb->s_export_op = &ocfs2_export_ops;
1225 sb->s_flags |= MS_NOATIME;
1226 /* this is needed to support O_LARGEFILE */
1227 sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits);
1228
1229 osb->sb = sb;
1230 /* Save off for ocfs2_rw_direct */
1231 osb->s_sectsize_bits = blksize_bits(sector_size);
1232 BUG_ON(!osb->s_sectsize_bits);
1233
1234 osb->net_response_ids = 0;
1235 spin_lock_init(&osb->net_response_lock);
1236 INIT_LIST_HEAD(&osb->net_response_list);
1237
1238 INIT_LIST_HEAD(&osb->osb_net_handlers);
1239 init_waitqueue_head(&osb->recovery_event);
1240 spin_lock_init(&osb->vote_task_lock);
1241 init_waitqueue_head(&osb->vote_event);
1242 osb->vote_work_sequence = 0;
1243 osb->vote_wake_sequence = 0;
1244 INIT_LIST_HEAD(&osb->blocked_lock_list);
1245 osb->blocked_lock_count = 0;
1246 INIT_LIST_HEAD(&osb->vote_list);
1247 spin_lock_init(&osb->osb_lock);
1248
1249 atomic_set(&osb->alloc_stats.moves, 0);
1250 atomic_set(&osb->alloc_stats.local_data, 0);
1251 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1252 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1253 atomic_set(&osb->alloc_stats.bg_extends, 0);
1254
1255 ocfs2_init_node_maps(osb);
1256
1257 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1258 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1259
1260 mutex_init(&osb->recovery_lock);
1261
1262 osb->disable_recovery = 0;
1263 osb->recovery_thread_task = NULL;
1264
1265 init_waitqueue_head(&osb->checkpoint_event);
1266 atomic_set(&osb->needs_checkpoint, 0);
1267
1268 osb->node_num = O2NM_INVALID_NODE_NUM;
1269 osb->slot_num = OCFS2_INVALID_SLOT;
1270
1271 osb->local_alloc_state = OCFS2_LA_UNUSED;
1272 osb->local_alloc_bh = NULL;
1273
1274 ocfs2_setup_hb_callbacks(osb);
1275
1276 init_waitqueue_head(&osb->osb_mount_event);
1277
1278 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1279 if (!osb->vol_label) {
1280 mlog(ML_ERROR, "unable to alloc vol label\n");
1281 status = -ENOMEM;
1282 goto bail;
1283 }
1284
1285 di = (struct ocfs2_dinode *)bh->b_data;
1286
1287 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1288 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1289 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1290 osb->max_slots);
1291 status = -EINVAL;
1292 goto bail;
1293 }
1294 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1295
1296 init_waitqueue_head(&osb->osb_wipe_event);
1297 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1298 sizeof(*osb->osb_orphan_wipes),
1299 GFP_KERNEL);
1300 if (!osb->osb_orphan_wipes) {
1301 status = -ENOMEM;
1302 mlog_errno(status);
1303 goto bail;
1304 }
1305
1306 osb->s_feature_compat =
1307 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1308 osb->s_feature_ro_compat =
1309 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1310 osb->s_feature_incompat =
1311 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1312
1313 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1314 mlog(ML_ERROR, "couldn't mount because of unsupported "
1315 "optional features (%x).\n", i);
1316 status = -EINVAL;
1317 goto bail;
1318 }
1319 if (!(osb->sb->s_flags & MS_RDONLY) &&
1320 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1321 mlog(ML_ERROR, "couldn't mount RDWR because of "
1322 "unsupported optional features (%x).\n", i);
1323 status = -EINVAL;
1324 goto bail;
1325 }
1326
1327 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1328
1329 /* FIXME
1330 * This should be done in ocfs2_journal_init(), but unknown
1331 * ordering issues will cause the filesystem to crash.
1332 * If anyone wants to figure out what part of the code
1333 * refers to osb->journal before ocfs2_journal_init() is run,
1334 * be my guest.
1335 */
1336 /* initialize our journal structure */
1337
1338 journal = kcalloc(1, sizeof(struct ocfs2_journal), GFP_KERNEL);
1339 if (!journal) {
1340 mlog(ML_ERROR, "unable to alloc journal\n");
1341 status = -ENOMEM;
1342 goto bail;
1343 }
1344 osb->journal = journal;
1345 journal->j_osb = osb;
1346
1347 atomic_set(&journal->j_num_trans, 0);
1348 init_rwsem(&journal->j_trans_barrier);
1349 init_waitqueue_head(&journal->j_checkpointed);
1350 spin_lock_init(&journal->j_lock);
1351 journal->j_trans_id = (unsigned long) 1;
1352 INIT_LIST_HEAD(&journal->j_la_cleanups);
1353 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery, osb);
1354 journal->j_state = OCFS2_JOURNAL_FREE;
1355
1356 /* get some pseudo constants for clustersize bits */
1357 osb->s_clustersize_bits =
1358 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1359 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1360 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1361
1362 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1363 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1364 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1365 osb->s_clustersize);
1366 status = -EINVAL;
1367 goto bail;
1368 }
1369
1370 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1371 > (u32)~0UL) {
1372 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1373 "what jbd can address in 32 bits.\n");
1374 status = -EINVAL;
1375 goto bail;
1376 }
1377
1378 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1379 sizeof(di->id2.i_super.s_uuid))) {
1380 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1381 status = -ENOMEM;
1382 goto bail;
1383 }
1384
1385 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1386 osb->net_key = le32_to_cpu(uuid_net_key);
1387
1388 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1389 osb->vol_label[63] = '\0';
1390 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1391 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1392 osb->first_cluster_group_blkno =
1393 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1394 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1395 mlog(0, "vol_label: %s\n", osb->vol_label);
1396 mlog(0, "uuid: %s\n", osb->uuid_str);
1397 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1398 (unsigned long long)osb->root_blkno,
1399 (unsigned long long)osb->system_dir_blkno);
1400
1401 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1402 if (!osb->osb_dlm_debug) {
1403 status = -ENOMEM;
1404 mlog_errno(status);
1405 goto bail;
1406 }
1407
1408 atomic_set(&osb->vol_state, VOLUME_INIT);
1409
1410 /* load root, system_dir, and all global system inodes */
1411 status = ocfs2_init_global_system_inodes(osb);
1412 if (status < 0) {
1413 mlog_errno(status);
1414 goto bail;
1415 }
1416
1417 /*
1418 * global bitmap
1419 */
1420 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1421 OCFS2_INVALID_SLOT);
1422 if (!inode) {
1423 status = -EINVAL;
1424 mlog_errno(status);
1425 goto bail;
1426 }
1427
1428 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1429
1430 /* We don't have a cluster lock on the bitmap here because
1431 * we're only interested in static information and the extra
1432 * complexity at mount time isn't worht it. Don't pass the
1433 * inode in to the read function though as we don't want it to
1434 * be put in the cache. */
1435 status = ocfs2_read_block(osb, osb->bitmap_blkno, &bitmap_bh, 0,
1436 NULL);
1437 iput(inode);
1438 if (status < 0) {
1439 mlog_errno(status);
1440 goto bail;
1441 }
1442
1443 di = (struct ocfs2_dinode *) bitmap_bh->b_data;
1444 osb->bitmap_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
1445 brelse(bitmap_bh);
1446 mlog(0, "cluster bitmap inode: %llu, clusters per group: %u\n",
1447 (unsigned long long)osb->bitmap_blkno, osb->bitmap_cpg);
1448
1449 status = ocfs2_init_slot_info(osb);
1450 if (status < 0) {
1451 mlog_errno(status);
1452 goto bail;
1453 }
1454
1455 bail:
1456 mlog_exit(status);
1457 return status;
1458 }
1459
1460 /*
1461 * will return: -EAGAIN if it is ok to keep searching for superblocks
1462 * -EINVAL if there is a bad superblock
1463 * 0 on success
1464 */
1465 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1466 struct buffer_head *bh,
1467 u32 blksz)
1468 {
1469 int status = -EAGAIN;
1470
1471 mlog_entry_void();
1472
1473 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1474 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1475 status = -EINVAL;
1476 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1477 mlog(ML_ERROR, "found superblock with incorrect block "
1478 "size: found %u, should be %u\n",
1479 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1480 blksz);
1481 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1482 OCFS2_MAJOR_REV_LEVEL ||
1483 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1484 OCFS2_MINOR_REV_LEVEL) {
1485 mlog(ML_ERROR, "found superblock with bad version: "
1486 "found %u.%u, should be %u.%u\n",
1487 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1488 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1489 OCFS2_MAJOR_REV_LEVEL,
1490 OCFS2_MINOR_REV_LEVEL);
1491 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1492 mlog(ML_ERROR, "bad block number on superblock: "
1493 "found %llu, should be %llu\n",
1494 (unsigned long long)di->i_blkno,
1495 (unsigned long long)bh->b_blocknr);
1496 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1497 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1498 mlog(ML_ERROR, "bad cluster size found: %u\n",
1499 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1500 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1501 mlog(ML_ERROR, "bad root_blkno: 0\n");
1502 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1503 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1504 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1505 mlog(ML_ERROR,
1506 "Superblock slots found greater than file system "
1507 "maximum: found %u, max %u\n",
1508 le16_to_cpu(di->id2.i_super.s_max_slots),
1509 OCFS2_MAX_SLOTS);
1510 } else {
1511 /* found it! */
1512 status = 0;
1513 }
1514 }
1515
1516 mlog_exit(status);
1517 return status;
1518 }
1519
1520 static int ocfs2_check_volume(struct ocfs2_super *osb)
1521 {
1522 int status = 0;
1523 int dirty;
1524 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1525 * recover
1526 * ourselves. */
1527
1528 mlog_entry_void();
1529
1530 /* Init our journal object. */
1531 status = ocfs2_journal_init(osb->journal, &dirty);
1532 if (status < 0) {
1533 mlog(ML_ERROR, "Could not initialize journal!\n");
1534 goto finally;
1535 }
1536
1537 /* If the journal was unmounted cleanly then we don't want to
1538 * recover anything. Otherwise, journal_load will do that
1539 * dirty work for us :) */
1540 if (!dirty) {
1541 status = ocfs2_journal_wipe(osb->journal, 0);
1542 if (status < 0) {
1543 mlog_errno(status);
1544 goto finally;
1545 }
1546 } else {
1547 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1548 "recovering volume.\n");
1549 }
1550
1551 /* will play back anything left in the journal. */
1552 ocfs2_journal_load(osb->journal);
1553
1554 if (dirty) {
1555 /* recover my local alloc if we didn't unmount cleanly. */
1556 status = ocfs2_begin_local_alloc_recovery(osb,
1557 osb->slot_num,
1558 &local_alloc);
1559 if (status < 0) {
1560 mlog_errno(status);
1561 goto finally;
1562 }
1563 /* we complete the recovery process after we've marked
1564 * ourselves as mounted. */
1565 }
1566
1567 mlog(0, "Journal loaded.\n");
1568
1569 status = ocfs2_load_local_alloc(osb);
1570 if (status < 0) {
1571 mlog_errno(status);
1572 goto finally;
1573 }
1574
1575 if (dirty) {
1576 /* Recovery will be completed after we've mounted the
1577 * rest of the volume. */
1578 osb->dirty = 1;
1579 osb->local_alloc_copy = local_alloc;
1580 local_alloc = NULL;
1581 }
1582
1583 /* go through each journal, trylock it and if you get the
1584 * lock, and it's marked as dirty, set the bit in the recover
1585 * map and launch a recovery thread for it. */
1586 status = ocfs2_mark_dead_nodes(osb);
1587 if (status < 0)
1588 mlog_errno(status);
1589
1590 finally:
1591 if (local_alloc)
1592 kfree(local_alloc);
1593
1594 mlog_exit(status);
1595 return status;
1596 }
1597
1598 /*
1599 * The routine gets called from dismount or close whenever a dismount on
1600 * volume is requested and the osb open count becomes 1.
1601 * It will remove the osb from the global list and also free up all the
1602 * initialized resources and fileobject.
1603 */
1604 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1605 {
1606 mlog_entry_void();
1607
1608 /* This function assumes that the caller has the main osb resource */
1609
1610 if (osb->slot_info)
1611 ocfs2_free_slot_info(osb->slot_info);
1612
1613 kfree(osb->osb_orphan_wipes);
1614 /* FIXME
1615 * This belongs in journal shutdown, but because we have to
1616 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1617 * we free it here.
1618 */
1619 kfree(osb->journal);
1620 if (osb->local_alloc_copy)
1621 kfree(osb->local_alloc_copy);
1622 kfree(osb->uuid_str);
1623 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1624 memset(osb, 0, sizeof(struct ocfs2_super));
1625
1626 mlog_exit_void();
1627 }
1628
1629 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1630 * panic(). We do not support continue-on-error operation. */
1631 static void ocfs2_handle_error(struct super_block *sb)
1632 {
1633 struct ocfs2_super *osb = OCFS2_SB(sb);
1634
1635 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1636 panic("OCFS2: (device %s): panic forced after error\n",
1637 sb->s_id);
1638
1639 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1640
1641 if (sb->s_flags & MS_RDONLY &&
1642 (ocfs2_is_soft_readonly(osb) ||
1643 ocfs2_is_hard_readonly(osb)))
1644 return;
1645
1646 printk(KERN_CRIT "File system is now read-only due to the potential "
1647 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1648 "system is unmounted.\n");
1649 sb->s_flags |= MS_RDONLY;
1650 ocfs2_set_ro_flag(osb, 0);
1651 }
1652
1653 static char error_buf[1024];
1654
1655 void __ocfs2_error(struct super_block *sb,
1656 const char *function,
1657 const char *fmt, ...)
1658 {
1659 va_list args;
1660
1661 va_start(args, fmt);
1662 vsprintf(error_buf, fmt, args);
1663 va_end(args);
1664
1665 /* Not using mlog here because we want to show the actual
1666 * function the error came from. */
1667 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1668 sb->s_id, function, error_buf);
1669
1670 ocfs2_handle_error(sb);
1671 }
1672
1673 /* Handle critical errors. This is intentionally more drastic than
1674 * ocfs2_handle_error, so we only use for things like journal errors,
1675 * etc. */
1676 void __ocfs2_abort(struct super_block* sb,
1677 const char *function,
1678 const char *fmt, ...)
1679 {
1680 va_list args;
1681
1682 va_start(args, fmt);
1683 vsprintf(error_buf, fmt, args);
1684 va_end(args);
1685
1686 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1687 sb->s_id, function, error_buf);
1688
1689 /* We don't have the cluster support yet to go straight to
1690 * hard readonly in here. Until then, we want to keep
1691 * ocfs2_abort() so that we can at least mark critical
1692 * errors.
1693 *
1694 * TODO: This should abort the journal and alert other nodes
1695 * that our slot needs recovery. */
1696
1697 /* Force a panic(). This stinks, but it's better than letting
1698 * things continue without having a proper hard readonly
1699 * here. */
1700 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1701 ocfs2_handle_error(sb);
1702 }
1703
1704 module_init(ocfs2_init);
1705 module_exit(ocfs2_exit);
Ubuntu Hirsute Linux kernel mirror