projects / deliverable / linux.git / blob
? search:
summary | shortlog | log | commit | commitdiff | tree
blame | history | raw | HEAD
ocfs2: add mount option and Kconfig option for acl
[deliverable/linux.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 #include <linux/mount.h>
43 #include <linux/seq_file.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 "xattr.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 struct mount_options
85 {
86 unsigned long commit_interval;
87 unsigned long mount_opt;
88 unsigned int atime_quantum;
89 signed short slot;
90 unsigned int localalloc_opt;
91 char cluster_stack[OCFS2_STACK_LABEL_LEN + 1];
92 };
93
94 static int ocfs2_parse_options(struct super_block *sb, char *options,
95 struct mount_options *mopt,
96 int is_remount);
97 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt);
98 static void ocfs2_put_super(struct super_block *sb);
99 static int ocfs2_mount_volume(struct super_block *sb);
100 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
101 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
102 static int ocfs2_initialize_mem_caches(void);
103 static void ocfs2_free_mem_caches(void);
104 static void ocfs2_delete_osb(struct ocfs2_super *osb);
105
106 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf);
107
108 static int ocfs2_sync_fs(struct super_block *sb, int wait);
109
110 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
111 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
112 static void ocfs2_release_system_inodes(struct ocfs2_super *osb);
113 static int ocfs2_check_volume(struct ocfs2_super *osb);
114 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
115 struct buffer_head *bh,
116 u32 sectsize);
117 static int ocfs2_initialize_super(struct super_block *sb,
118 struct buffer_head *bh,
119 int sector_size);
120 static int ocfs2_get_sector(struct super_block *sb,
121 struct buffer_head **bh,
122 int block,
123 int sect_size);
124 static void ocfs2_write_super(struct super_block *sb);
125 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
126 static void ocfs2_destroy_inode(struct inode *inode);
127
128 static const struct super_operations ocfs2_sops = {
129 .statfs = ocfs2_statfs,
130 .alloc_inode = ocfs2_alloc_inode,
131 .destroy_inode = ocfs2_destroy_inode,
132 .drop_inode = ocfs2_drop_inode,
133 .clear_inode = ocfs2_clear_inode,
134 .delete_inode = ocfs2_delete_inode,
135 .sync_fs = ocfs2_sync_fs,
136 .write_super = ocfs2_write_super,
137 .put_super = ocfs2_put_super,
138 .remount_fs = ocfs2_remount,
139 .show_options = ocfs2_show_options,
140 };
141
142 enum {
143 Opt_barrier,
144 Opt_err_panic,
145 Opt_err_ro,
146 Opt_intr,
147 Opt_nointr,
148 Opt_hb_none,
149 Opt_hb_local,
150 Opt_data_ordered,
151 Opt_data_writeback,
152 Opt_atime_quantum,
153 Opt_slot,
154 Opt_commit,
155 Opt_localalloc,
156 Opt_localflocks,
157 Opt_stack,
158 Opt_user_xattr,
159 Opt_nouser_xattr,
160 Opt_inode64,
161 Opt_acl,
162 Opt_noacl,
163 Opt_err,
164 };
165
166 static const match_table_t tokens = {
167 {Opt_barrier, "barrier=%u"},
168 {Opt_err_panic, "errors=panic"},
169 {Opt_err_ro, "errors=remount-ro"},
170 {Opt_intr, "intr"},
171 {Opt_nointr, "nointr"},
172 {Opt_hb_none, OCFS2_HB_NONE},
173 {Opt_hb_local, OCFS2_HB_LOCAL},
174 {Opt_data_ordered, "data=ordered"},
175 {Opt_data_writeback, "data=writeback"},
176 {Opt_atime_quantum, "atime_quantum=%u"},
177 {Opt_slot, "preferred_slot=%u"},
178 {Opt_commit, "commit=%u"},
179 {Opt_localalloc, "localalloc=%d"},
180 {Opt_localflocks, "localflocks"},
181 {Opt_stack, "cluster_stack=%s"},
182 {Opt_user_xattr, "user_xattr"},
183 {Opt_nouser_xattr, "nouser_xattr"},
184 {Opt_inode64, "inode64"},
185 {Opt_acl, "acl"},
186 {Opt_noacl, "noacl"},
187 {Opt_err, NULL}
188 };
189
190 /*
191 * write_super and sync_fs ripped right out of ext3.
192 */
193 static void ocfs2_write_super(struct super_block *sb)
194 {
195 if (mutex_trylock(&sb->s_lock) != 0)
196 BUG();
197 sb->s_dirt = 0;
198 }
199
200 static int ocfs2_sync_fs(struct super_block *sb, int wait)
201 {
202 int status;
203 tid_t target;
204 struct ocfs2_super *osb = OCFS2_SB(sb);
205
206 sb->s_dirt = 0;
207
208 if (ocfs2_is_hard_readonly(osb))
209 return -EROFS;
210
211 if (wait) {
212 status = ocfs2_flush_truncate_log(osb);
213 if (status < 0)
214 mlog_errno(status);
215 } else {
216 ocfs2_schedule_truncate_log_flush(osb, 0);
217 }
218
219 if (jbd2_journal_start_commit(OCFS2_SB(sb)->journal->j_journal,
220 &target)) {
221 if (wait)
222 jbd2_log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
223 target);
224 }
225 return 0;
226 }
227
228 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
229 {
230 struct inode *new = NULL;
231 int status = 0;
232 int i;
233
234 mlog_entry_void();
235
236 new = ocfs2_iget(osb, osb->root_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
237 if (IS_ERR(new)) {
238 status = PTR_ERR(new);
239 mlog_errno(status);
240 goto bail;
241 }
242 osb->root_inode = new;
243
244 new = ocfs2_iget(osb, osb->system_dir_blkno, OCFS2_FI_FLAG_SYSFILE, 0);
245 if (IS_ERR(new)) {
246 status = PTR_ERR(new);
247 mlog_errno(status);
248 goto bail;
249 }
250 osb->sys_root_inode = new;
251
252 for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
253 i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
254 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
255 if (!new) {
256 ocfs2_release_system_inodes(osb);
257 status = -EINVAL;
258 mlog_errno(status);
259 /* FIXME: Should ERROR_RO_FS */
260 mlog(ML_ERROR, "Unable to load system inode %d, "
261 "possibly corrupt fs?", i);
262 goto bail;
263 }
264 // the array now has one ref, so drop this one
265 iput(new);
266 }
267
268 bail:
269 mlog_exit(status);
270 return status;
271 }
272
273 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
274 {
275 struct inode *new = NULL;
276 int status = 0;
277 int i;
278
279 mlog_entry_void();
280
281 for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
282 i < NUM_SYSTEM_INODES;
283 i++) {
284 new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
285 if (!new) {
286 ocfs2_release_system_inodes(osb);
287 status = -EINVAL;
288 mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
289 status, i, osb->slot_num);
290 goto bail;
291 }
292 /* the array now has one ref, so drop this one */
293 iput(new);
294 }
295
296 bail:
297 mlog_exit(status);
298 return status;
299 }
300
301 static void ocfs2_release_system_inodes(struct ocfs2_super *osb)
302 {
303 int i;
304 struct inode *inode;
305
306 mlog_entry_void();
307
308 for (i = 0; i < NUM_SYSTEM_INODES; i++) {
309 inode = osb->system_inodes[i];
310 if (inode) {
311 iput(inode);
312 osb->system_inodes[i] = NULL;
313 }
314 }
315
316 inode = osb->sys_root_inode;
317 if (inode) {
318 iput(inode);
319 osb->sys_root_inode = NULL;
320 }
321
322 inode = osb->root_inode;
323 if (inode) {
324 iput(inode);
325 osb->root_inode = NULL;
326 }
327
328 mlog_exit(0);
329 }
330
331 /* We're allocating fs objects, use GFP_NOFS */
332 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
333 {
334 struct ocfs2_inode_info *oi;
335
336 oi = kmem_cache_alloc(ocfs2_inode_cachep, GFP_NOFS);
337 if (!oi)
338 return NULL;
339
340 jbd2_journal_init_jbd_inode(&oi->ip_jinode, &oi->vfs_inode);
341 return &oi->vfs_inode;
342 }
343
344 static void ocfs2_destroy_inode(struct inode *inode)
345 {
346 kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
347 }
348
349 static unsigned long long ocfs2_max_file_offset(unsigned int bbits,
350 unsigned int cbits)
351 {
352 unsigned int bytes = 1 << cbits;
353 unsigned int trim = bytes;
354 unsigned int bitshift = 32;
355
356 /*
357 * i_size and all block offsets in ocfs2 are always 64 bits
358 * wide. i_clusters is 32 bits, in cluster-sized units. So on
359 * 64 bit platforms, cluster size will be the limiting factor.
360 */
361
362 #if BITS_PER_LONG == 32
363 # if defined(CONFIG_LBD)
364 BUILD_BUG_ON(sizeof(sector_t) != 8);
365 /*
366 * We might be limited by page cache size.
367 */
368 if (bytes > PAGE_CACHE_SIZE) {
369 bytes = PAGE_CACHE_SIZE;
370 trim = 1;
371 /*
372 * Shift by 31 here so that we don't get larger than
373 * MAX_LFS_FILESIZE
374 */
375 bitshift = 31;
376 }
377 # else
378 /*
379 * We are limited by the size of sector_t. Use block size, as
380 * that's what we expose to the VFS.
381 */
382 bytes = 1 << bbits;
383 trim = 1;
384 bitshift = 31;
385 # endif
386 #endif
387
388 /*
389 * Trim by a whole cluster when we can actually approach the
390 * on-disk limits. Otherwise we can overflow i_clusters when
391 * an extent start is at the max offset.
392 */
393 return (((unsigned long long)bytes) << bitshift) - trim;
394 }
395
396 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
397 {
398 int incompat_features;
399 int ret = 0;
400 struct mount_options parsed_options;
401 struct ocfs2_super *osb = OCFS2_SB(sb);
402
403 if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
404 ret = -EINVAL;
405 goto out;
406 }
407
408 if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
409 (parsed_options.mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
410 ret = -EINVAL;
411 mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
412 goto out;
413 }
414
415 if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
416 (parsed_options.mount_opt & OCFS2_MOUNT_DATA_WRITEBACK)) {
417 ret = -EINVAL;
418 mlog(ML_ERROR, "Cannot change data mode on remount\n");
419 goto out;
420 }
421
422 /* Probably don't want this on remount; it might
423 * mess with other nodes */
424 if (!(osb->s_mount_opt & OCFS2_MOUNT_INODE64) &&
425 (parsed_options.mount_opt & OCFS2_MOUNT_INODE64)) {
426 ret = -EINVAL;
427 mlog(ML_ERROR, "Cannot enable inode64 on remount\n");
428 goto out;
429 }
430
431 /* We're going to/from readonly mode. */
432 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
433 /* Lock here so the check of HARD_RO and the potential
434 * setting of SOFT_RO is atomic. */
435 spin_lock(&osb->osb_lock);
436 if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
437 mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
438 ret = -EROFS;
439 goto unlock_osb;
440 }
441
442 if (*flags & MS_RDONLY) {
443 mlog(0, "Going to ro mode.\n");
444 sb->s_flags |= MS_RDONLY;
445 osb->osb_flags |= OCFS2_OSB_SOFT_RO;
446 } else {
447 mlog(0, "Making ro filesystem writeable.\n");
448
449 if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
450 mlog(ML_ERROR, "Cannot remount RDWR "
451 "filesystem due to previous errors.\n");
452 ret = -EROFS;
453 goto unlock_osb;
454 }
455 incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
456 if (incompat_features) {
457 mlog(ML_ERROR, "Cannot remount RDWR because "
458 "of unsupported optional features "
459 "(%x).\n", incompat_features);
460 ret = -EINVAL;
461 goto unlock_osb;
462 }
463 sb->s_flags &= ~MS_RDONLY;
464 osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
465 }
466 unlock_osb:
467 spin_unlock(&osb->osb_lock);
468 }
469
470 if (!ret) {
471 /* Only save off the new mount options in case of a successful
472 * remount. */
473 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
474 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
475 osb->s_mount_opt = parsed_options.mount_opt;
476 osb->s_atime_quantum = parsed_options.atime_quantum;
477 osb->preferred_slot = parsed_options.slot;
478 if (parsed_options.commit_interval)
479 osb->osb_commit_interval = parsed_options.commit_interval;
480
481 if (!ocfs2_is_hard_readonly(osb))
482 ocfs2_set_journal_params(osb);
483 }
484 out:
485 return ret;
486 }
487
488 static int ocfs2_sb_probe(struct super_block *sb,
489 struct buffer_head **bh,
490 int *sector_size)
491 {
492 int status, tmpstat;
493 struct ocfs1_vol_disk_hdr *hdr;
494 struct ocfs2_dinode *di;
495 int blksize;
496
497 *bh = NULL;
498
499 /* may be > 512 */
500 *sector_size = bdev_hardsect_size(sb->s_bdev);
501 if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
502 mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
503 *sector_size, OCFS2_MAX_BLOCKSIZE);
504 status = -EINVAL;
505 goto bail;
506 }
507
508 /* Can this really happen? */
509 if (*sector_size < OCFS2_MIN_BLOCKSIZE)
510 *sector_size = OCFS2_MIN_BLOCKSIZE;
511
512 /* check block zero for old format */
513 status = ocfs2_get_sector(sb, bh, 0, *sector_size);
514 if (status < 0) {
515 mlog_errno(status);
516 goto bail;
517 }
518 hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
519 if (hdr->major_version == OCFS1_MAJOR_VERSION) {
520 mlog(ML_ERROR, "incompatible version: %u.%u\n",
521 hdr->major_version, hdr->minor_version);
522 status = -EINVAL;
523 }
524 if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
525 strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
526 mlog(ML_ERROR, "incompatible volume signature: %8s\n",
527 hdr->signature);
528 status = -EINVAL;
529 }
530 brelse(*bh);
531 *bh = NULL;
532 if (status < 0) {
533 mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
534 "upgraded before mounting with ocfs v2\n");
535 goto bail;
536 }
537
538 /*
539 * Now check at magic offset for 512, 1024, 2048, 4096
540 * blocksizes. 4096 is the maximum blocksize because it is
541 * the minimum clustersize.
542 */
543 status = -EINVAL;
544 for (blksize = *sector_size;
545 blksize <= OCFS2_MAX_BLOCKSIZE;
546 blksize <<= 1) {
547 tmpstat = ocfs2_get_sector(sb, bh,
548 OCFS2_SUPER_BLOCK_BLKNO,
549 blksize);
550 if (tmpstat < 0) {
551 status = tmpstat;
552 mlog_errno(status);
553 goto bail;
554 }
555 di = (struct ocfs2_dinode *) (*bh)->b_data;
556 status = ocfs2_verify_volume(di, *bh, blksize);
557 if (status >= 0)
558 goto bail;
559 brelse(*bh);
560 *bh = NULL;
561 if (status != -EAGAIN)
562 break;
563 }
564
565 bail:
566 return status;
567 }
568
569 static int ocfs2_verify_heartbeat(struct ocfs2_super *osb)
570 {
571 if (ocfs2_mount_local(osb)) {
572 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
573 mlog(ML_ERROR, "Cannot heartbeat on a locally "
574 "mounted device.\n");
575 return -EINVAL;
576 }
577 }
578
579 if (ocfs2_userspace_stack(osb)) {
580 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
581 mlog(ML_ERROR, "Userspace stack expected, but "
582 "o2cb heartbeat arguments passed to mount\n");
583 return -EINVAL;
584 }
585 }
586
587 if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
588 if (!ocfs2_mount_local(osb) && !ocfs2_is_hard_readonly(osb) &&
589 !ocfs2_userspace_stack(osb)) {
590 mlog(ML_ERROR, "Heartbeat has to be started to mount "
591 "a read-write clustered device.\n");
592 return -EINVAL;
593 }
594 }
595
596 return 0;
597 }
598
599 /*
600 * If we're using a userspace stack, mount should have passed
601 * a name that matches the disk. If not, mount should not
602 * have passed a stack.
603 */
604 static int ocfs2_verify_userspace_stack(struct ocfs2_super *osb,
605 struct mount_options *mopt)
606 {
607 if (!ocfs2_userspace_stack(osb) && mopt->cluster_stack[0]) {
608 mlog(ML_ERROR,
609 "cluster stack passed to mount, but this filesystem "
610 "does not support it\n");
611 return -EINVAL;
612 }
613
614 if (ocfs2_userspace_stack(osb) &&
615 strncmp(osb->osb_cluster_stack, mopt->cluster_stack,
616 OCFS2_STACK_LABEL_LEN)) {
617 mlog(ML_ERROR,
618 "cluster stack passed to mount (\"%s\") does not "
619 "match the filesystem (\"%s\")\n",
620 mopt->cluster_stack,
621 osb->osb_cluster_stack);
622 return -EINVAL;
623 }
624
625 return 0;
626 }
627
628 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
629 {
630 struct dentry *root;
631 int status, sector_size;
632 struct mount_options parsed_options;
633 struct inode *inode = NULL;
634 struct ocfs2_super *osb = NULL;
635 struct buffer_head *bh = NULL;
636 char nodestr[8];
637
638 mlog_entry("%p, %p, %i", sb, data, silent);
639
640 if (!ocfs2_parse_options(sb, data, &parsed_options, 0)) {
641 status = -EINVAL;
642 goto read_super_error;
643 }
644
645 /* probe for superblock */
646 status = ocfs2_sb_probe(sb, &bh, &sector_size);
647 if (status < 0) {
648 mlog(ML_ERROR, "superblock probe failed!\n");
649 goto read_super_error;
650 }
651
652 status = ocfs2_initialize_super(sb, bh, sector_size);
653 osb = OCFS2_SB(sb);
654 if (status < 0) {
655 mlog_errno(status);
656 goto read_super_error;
657 }
658 brelse(bh);
659 bh = NULL;
660
661 if (!(osb->s_feature_incompat & OCFS2_FEATURE_INCOMPAT_XATTR))
662 parsed_options.mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
663
664 osb->s_mount_opt = parsed_options.mount_opt;
665 osb->s_atime_quantum = parsed_options.atime_quantum;
666 osb->preferred_slot = parsed_options.slot;
667 osb->osb_commit_interval = parsed_options.commit_interval;
668 osb->local_alloc_default_bits = ocfs2_megabytes_to_clusters(sb, parsed_options.localalloc_opt);
669 osb->local_alloc_bits = osb->local_alloc_default_bits;
670
671 status = ocfs2_verify_userspace_stack(osb, &parsed_options);
672 if (status)
673 goto read_super_error;
674
675 sb->s_magic = OCFS2_SUPER_MAGIC;
676
677 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
678 ((osb->s_mount_opt & OCFS2_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
679
680 /* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
681 * heartbeat=none */
682 if (bdev_read_only(sb->s_bdev)) {
683 if (!(sb->s_flags & MS_RDONLY)) {
684 status = -EACCES;
685 mlog(ML_ERROR, "Readonly device detected but readonly "
686 "mount was not specified.\n");
687 goto read_super_error;
688 }
689
690 /* You should not be able to start a local heartbeat
691 * on a readonly device. */
692 if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
693 status = -EROFS;
694 mlog(ML_ERROR, "Local heartbeat specified on readonly "
695 "device.\n");
696 goto read_super_error;
697 }
698
699 status = ocfs2_check_journals_nolocks(osb);
700 if (status < 0) {
701 if (status == -EROFS)
702 mlog(ML_ERROR, "Recovery required on readonly "
703 "file system, but write access is "
704 "unavailable.\n");
705 else
706 mlog_errno(status);
707 goto read_super_error;
708 }
709
710 ocfs2_set_ro_flag(osb, 1);
711
712 printk(KERN_NOTICE "Readonly device detected. No cluster "
713 "services will be utilized for this mount. Recovery "
714 "will be skipped.\n");
715 }
716
717 if (!ocfs2_is_hard_readonly(osb)) {
718 if (sb->s_flags & MS_RDONLY)
719 ocfs2_set_ro_flag(osb, 0);
720 }
721
722 status = ocfs2_verify_heartbeat(osb);
723 if (status < 0) {
724 mlog_errno(status);
725 goto read_super_error;
726 }
727
728 osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
729 ocfs2_debugfs_root);
730 if (!osb->osb_debug_root) {
731 status = -EINVAL;
732 mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
733 goto read_super_error;
734 }
735
736 status = ocfs2_mount_volume(sb);
737 if (osb->root_inode)
738 inode = igrab(osb->root_inode);
739
740 if (status < 0)
741 goto read_super_error;
742
743 if (!inode) {
744 status = -EIO;
745 mlog_errno(status);
746 goto read_super_error;
747 }
748
749 root = d_alloc_root(inode);
750 if (!root) {
751 status = -ENOMEM;
752 mlog_errno(status);
753 goto read_super_error;
754 }
755
756 sb->s_root = root;
757
758 ocfs2_complete_mount_recovery(osb);
759
760 if (ocfs2_mount_local(osb))
761 snprintf(nodestr, sizeof(nodestr), "local");
762 else
763 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
764
765 printk(KERN_INFO "ocfs2: Mounting device (%s) on (node %s, slot %d) "
766 "with %s data mode.\n",
767 osb->dev_str, nodestr, osb->slot_num,
768 osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
769 "ordered");
770
771 atomic_set(&osb->vol_state, VOLUME_MOUNTED);
772 wake_up(&osb->osb_mount_event);
773
774 mlog_exit(status);
775 return status;
776
777 read_super_error:
778 brelse(bh);
779
780 if (inode)
781 iput(inode);
782
783 if (osb) {
784 atomic_set(&osb->vol_state, VOLUME_DISABLED);
785 wake_up(&osb->osb_mount_event);
786 ocfs2_dismount_volume(sb, 1);
787 }
788
789 mlog_exit(status);
790 return status;
791 }
792
793 static int ocfs2_get_sb(struct file_system_type *fs_type,
794 int flags,
795 const char *dev_name,
796 void *data,
797 struct vfsmount *mnt)
798 {
799 return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super,
800 mnt);
801 }
802
803 static struct file_system_type ocfs2_fs_type = {
804 .owner = THIS_MODULE,
805 .name = "ocfs2",
806 .get_sb = ocfs2_get_sb, /* is this called when we mount
807 * the fs? */
808 .kill_sb = kill_block_super, /* set to the generic one
809 * right now, but do we
810 * need to change that? */
811 .fs_flags = FS_REQUIRES_DEV|FS_RENAME_DOES_D_MOVE,
812 .next = NULL
813 };
814
815 static int ocfs2_parse_options(struct super_block *sb,
816 char *options,
817 struct mount_options *mopt,
818 int is_remount)
819 {
820 int status;
821 char *p;
822
823 mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
824 options ? options : "(none)");
825
826 mopt->commit_interval = 0;
827 mopt->mount_opt = 0;
828 mopt->atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
829 mopt->slot = OCFS2_INVALID_SLOT;
830 mopt->localalloc_opt = OCFS2_DEFAULT_LOCAL_ALLOC_SIZE;
831 mopt->cluster_stack[0] = '\0';
832
833 if (!options) {
834 status = 1;
835 goto bail;
836 }
837
838 while ((p = strsep(&options, ",")) != NULL) {
839 int token, option;
840 substring_t args[MAX_OPT_ARGS];
841
842 if (!*p)
843 continue;
844
845 token = match_token(p, tokens, args);
846 switch (token) {
847 case Opt_hb_local:
848 mopt->mount_opt |= OCFS2_MOUNT_HB_LOCAL;
849 break;
850 case Opt_hb_none:
851 mopt->mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
852 break;
853 case Opt_barrier:
854 if (match_int(&args[0], &option)) {
855 status = 0;
856 goto bail;
857 }
858 if (option)
859 mopt->mount_opt |= OCFS2_MOUNT_BARRIER;
860 else
861 mopt->mount_opt &= ~OCFS2_MOUNT_BARRIER;
862 break;
863 case Opt_intr:
864 mopt->mount_opt &= ~OCFS2_MOUNT_NOINTR;
865 break;
866 case Opt_nointr:
867 mopt->mount_opt |= OCFS2_MOUNT_NOINTR;
868 break;
869 case Opt_err_panic:
870 mopt->mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
871 break;
872 case Opt_err_ro:
873 mopt->mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
874 break;
875 case Opt_data_ordered:
876 mopt->mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
877 break;
878 case Opt_data_writeback:
879 mopt->mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
880 break;
881 case Opt_user_xattr:
882 mopt->mount_opt &= ~OCFS2_MOUNT_NOUSERXATTR;
883 break;
884 case Opt_nouser_xattr:
885 mopt->mount_opt |= OCFS2_MOUNT_NOUSERXATTR;
886 break;
887 case Opt_atime_quantum:
888 if (match_int(&args[0], &option)) {
889 status = 0;
890 goto bail;
891 }
892 if (option >= 0)
893 mopt->atime_quantum = option;
894 break;
895 case Opt_slot:
896 option = 0;
897 if (match_int(&args[0], &option)) {
898 status = 0;
899 goto bail;
900 }
901 if (option)
902 mopt->slot = (s16)option;
903 break;
904 case Opt_commit:
905 option = 0;
906 if (match_int(&args[0], &option)) {
907 status = 0;
908 goto bail;
909 }
910 if (option < 0)
911 return 0;
912 if (option == 0)
913 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
914 mopt->commit_interval = HZ * option;
915 break;
916 case Opt_localalloc:
917 option = 0;
918 if (match_int(&args[0], &option)) {
919 status = 0;
920 goto bail;
921 }
922 if (option >= 0 && (option <= ocfs2_local_alloc_size(sb) * 8))
923 mopt->localalloc_opt = option;
924 break;
925 case Opt_localflocks:
926 /*
927 * Changing this during remount could race
928 * flock() requests, or "unbalance" existing
929 * ones (e.g., a lock is taken in one mode but
930 * dropped in the other). If users care enough
931 * to flip locking modes during remount, we
932 * could add a "local" flag to individual
933 * flock structures for proper tracking of
934 * state.
935 */
936 if (!is_remount)
937 mopt->mount_opt |= OCFS2_MOUNT_LOCALFLOCKS;
938 break;
939 case Opt_stack:
940 /* Check both that the option we were passed
941 * is of the right length and that it is a proper
942 * string of the right length.
943 */
944 if (((args[0].to - args[0].from) !=
945 OCFS2_STACK_LABEL_LEN) ||
946 (strnlen(args[0].from,
947 OCFS2_STACK_LABEL_LEN) !=
948 OCFS2_STACK_LABEL_LEN)) {
949 mlog(ML_ERROR,
950 "Invalid cluster_stack option\n");
951 status = 0;
952 goto bail;
953 }
954 memcpy(mopt->cluster_stack, args[0].from,
955 OCFS2_STACK_LABEL_LEN);
956 mopt->cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
957 break;
958 case Opt_inode64:
959 mopt->mount_opt |= OCFS2_MOUNT_INODE64;
960 break;
961 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
962 case Opt_acl:
963 mopt->mount_opt |= OCFS2_MOUNT_POSIX_ACL;
964 break;
965 case Opt_noacl:
966 mopt->mount_opt &= ~OCFS2_MOUNT_POSIX_ACL;
967 break;
968 #else
969 case Opt_acl:
970 case Opt_noacl:
971 printk(KERN_INFO "ocfs2 (no)acl options not supported\n");
972 break;
973 #endif
974 default:
975 mlog(ML_ERROR,
976 "Unrecognized mount option \"%s\" "
977 "or missing value\n", p);
978 status = 0;
979 goto bail;
980 }
981 }
982
983 status = 1;
984
985 bail:
986 mlog_exit(status);
987 return status;
988 }
989
990 static int ocfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
991 {
992 struct ocfs2_super *osb = OCFS2_SB(mnt->mnt_sb);
993 unsigned long opts = osb->s_mount_opt;
994 unsigned int local_alloc_megs;
995
996 if (opts & OCFS2_MOUNT_HB_LOCAL)
997 seq_printf(s, ",_netdev,heartbeat=local");
998 else
999 seq_printf(s, ",heartbeat=none");
1000
1001 if (opts & OCFS2_MOUNT_NOINTR)
1002 seq_printf(s, ",nointr");
1003
1004 if (opts & OCFS2_MOUNT_DATA_WRITEBACK)
1005 seq_printf(s, ",data=writeback");
1006 else
1007 seq_printf(s, ",data=ordered");
1008
1009 if (opts & OCFS2_MOUNT_BARRIER)
1010 seq_printf(s, ",barrier=1");
1011
1012 if (opts & OCFS2_MOUNT_ERRORS_PANIC)
1013 seq_printf(s, ",errors=panic");
1014 else
1015 seq_printf(s, ",errors=remount-ro");
1016
1017 if (osb->preferred_slot != OCFS2_INVALID_SLOT)
1018 seq_printf(s, ",preferred_slot=%d", osb->preferred_slot);
1019
1020 if (osb->s_atime_quantum != OCFS2_DEFAULT_ATIME_QUANTUM)
1021 seq_printf(s, ",atime_quantum=%u", osb->s_atime_quantum);
1022
1023 if (osb->osb_commit_interval)
1024 seq_printf(s, ",commit=%u",
1025 (unsigned) (osb->osb_commit_interval / HZ));
1026
1027 local_alloc_megs = osb->local_alloc_bits >> (20 - osb->s_clustersize_bits);
1028 if (local_alloc_megs != OCFS2_DEFAULT_LOCAL_ALLOC_SIZE)
1029 seq_printf(s, ",localalloc=%d", local_alloc_megs);
1030
1031 if (opts & OCFS2_MOUNT_LOCALFLOCKS)
1032 seq_printf(s, ",localflocks,");
1033
1034 if (osb->osb_cluster_stack[0])
1035 seq_printf(s, ",cluster_stack=%.*s", OCFS2_STACK_LABEL_LEN,
1036 osb->osb_cluster_stack);
1037
1038 if (opts & OCFS2_MOUNT_NOUSERXATTR)
1039 seq_printf(s, ",nouser_xattr");
1040 else
1041 seq_printf(s, ",user_xattr");
1042
1043 if (opts & OCFS2_MOUNT_INODE64)
1044 seq_printf(s, ",inode64");
1045
1046 #ifdef CONFIG_OCFS2_FS_POSIX_ACL
1047 if (opts & OCFS2_MOUNT_POSIX_ACL)
1048 seq_printf(s, ",acl");
1049 else
1050 seq_printf(s, ",noacl");
1051 #endif
1052
1053 return 0;
1054 }
1055
1056 static int __init ocfs2_init(void)
1057 {
1058 int status;
1059
1060 mlog_entry_void();
1061
1062 ocfs2_print_version();
1063
1064 status = init_ocfs2_uptodate_cache();
1065 if (status < 0) {
1066 mlog_errno(status);
1067 goto leave;
1068 }
1069
1070 status = ocfs2_initialize_mem_caches();
1071 if (status < 0) {
1072 mlog_errno(status);
1073 goto leave;
1074 }
1075
1076 ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
1077 if (!ocfs2_wq) {
1078 status = -ENOMEM;
1079 goto leave;
1080 }
1081
1082 ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
1083 if (!ocfs2_debugfs_root) {
1084 status = -EFAULT;
1085 mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
1086 }
1087
1088 ocfs2_set_locking_protocol();
1089
1090 leave:
1091 if (status < 0) {
1092 ocfs2_free_mem_caches();
1093 exit_ocfs2_uptodate_cache();
1094 }
1095
1096 mlog_exit(status);
1097
1098 if (status >= 0) {
1099 return register_filesystem(&ocfs2_fs_type);
1100 } else
1101 return -1;
1102 }
1103
1104 static void __exit ocfs2_exit(void)
1105 {
1106 mlog_entry_void();
1107
1108 if (ocfs2_wq) {
1109 flush_workqueue(ocfs2_wq);
1110 destroy_workqueue(ocfs2_wq);
1111 }
1112
1113 debugfs_remove(ocfs2_debugfs_root);
1114
1115 ocfs2_free_mem_caches();
1116
1117 unregister_filesystem(&ocfs2_fs_type);
1118
1119 exit_ocfs2_uptodate_cache();
1120
1121 mlog_exit_void();
1122 }
1123
1124 static void ocfs2_put_super(struct super_block *sb)
1125 {
1126 mlog_entry("(0x%p)\n", sb);
1127
1128 ocfs2_sync_blockdev(sb);
1129 ocfs2_dismount_volume(sb, 0);
1130
1131 mlog_exit_void();
1132 }
1133
1134 static int ocfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1135 {
1136 struct ocfs2_super *osb;
1137 u32 numbits, freebits;
1138 int status;
1139 struct ocfs2_dinode *bm_lock;
1140 struct buffer_head *bh = NULL;
1141 struct inode *inode = NULL;
1142
1143 mlog_entry("(%p, %p)\n", dentry->d_sb, buf);
1144
1145 osb = OCFS2_SB(dentry->d_sb);
1146
1147 inode = ocfs2_get_system_file_inode(osb,
1148 GLOBAL_BITMAP_SYSTEM_INODE,
1149 OCFS2_INVALID_SLOT);
1150 if (!inode) {
1151 mlog(ML_ERROR, "failed to get bitmap inode\n");
1152 status = -EIO;
1153 goto bail;
1154 }
1155
1156 status = ocfs2_inode_lock(inode, &bh, 0);
1157 if (status < 0) {
1158 mlog_errno(status);
1159 goto bail;
1160 }
1161
1162 bm_lock = (struct ocfs2_dinode *) bh->b_data;
1163
1164 numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
1165 freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
1166
1167 buf->f_type = OCFS2_SUPER_MAGIC;
1168 buf->f_bsize = dentry->d_sb->s_blocksize;
1169 buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
1170 buf->f_blocks = ((sector_t) numbits) *
1171 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1172 buf->f_bfree = ((sector_t) freebits) *
1173 (osb->s_clustersize >> osb->sb->s_blocksize_bits);
1174 buf->f_bavail = buf->f_bfree;
1175 buf->f_files = numbits;
1176 buf->f_ffree = freebits;
1177
1178 brelse(bh);
1179
1180 ocfs2_inode_unlock(inode, 0);
1181 status = 0;
1182 bail:
1183 if (inode)
1184 iput(inode);
1185
1186 mlog_exit(status);
1187
1188 return status;
1189 }
1190
1191 static void ocfs2_inode_init_once(void *data)
1192 {
1193 struct ocfs2_inode_info *oi = data;
1194
1195 oi->ip_flags = 0;
1196 oi->ip_open_count = 0;
1197 spin_lock_init(&oi->ip_lock);
1198 ocfs2_extent_map_init(&oi->vfs_inode);
1199 INIT_LIST_HEAD(&oi->ip_io_markers);
1200 oi->ip_created_trans = 0;
1201 oi->ip_last_trans = 0;
1202 oi->ip_dir_start_lookup = 0;
1203
1204 init_rwsem(&oi->ip_alloc_sem);
1205 init_rwsem(&oi->ip_xattr_sem);
1206 mutex_init(&oi->ip_io_mutex);
1207
1208 oi->ip_blkno = 0ULL;
1209 oi->ip_clusters = 0;
1210
1211 ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
1212 ocfs2_lock_res_init_once(&oi->ip_inode_lockres);
1213 ocfs2_lock_res_init_once(&oi->ip_open_lockres);
1214
1215 ocfs2_metadata_cache_init(&oi->vfs_inode);
1216
1217 inode_init_once(&oi->vfs_inode);
1218 }
1219
1220 static int ocfs2_initialize_mem_caches(void)
1221 {
1222 ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
1223 sizeof(struct ocfs2_inode_info),
1224 0,
1225 (SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT|
1226 SLAB_MEM_SPREAD),
1227 ocfs2_inode_init_once);
1228 if (!ocfs2_inode_cachep)
1229 return -ENOMEM;
1230
1231 return 0;
1232 }
1233
1234 static void ocfs2_free_mem_caches(void)
1235 {
1236 if (ocfs2_inode_cachep)
1237 kmem_cache_destroy(ocfs2_inode_cachep);
1238
1239 ocfs2_inode_cachep = NULL;
1240 }
1241
1242 static int ocfs2_get_sector(struct super_block *sb,
1243 struct buffer_head **bh,
1244 int block,
1245 int sect_size)
1246 {
1247 if (!sb_set_blocksize(sb, sect_size)) {
1248 mlog(ML_ERROR, "unable to set blocksize\n");
1249 return -EIO;
1250 }
1251
1252 *bh = sb_getblk(sb, block);
1253 if (!*bh) {
1254 mlog_errno(-EIO);
1255 return -EIO;
1256 }
1257 lock_buffer(*bh);
1258 if (!buffer_dirty(*bh))
1259 clear_buffer_uptodate(*bh);
1260 unlock_buffer(*bh);
1261 ll_rw_block(READ, 1, bh);
1262 wait_on_buffer(*bh);
1263 return 0;
1264 }
1265
1266 static int ocfs2_mount_volume(struct super_block *sb)
1267 {
1268 int status = 0;
1269 int unlock_super = 0;
1270 struct ocfs2_super *osb = OCFS2_SB(sb);
1271
1272 mlog_entry_void();
1273
1274 if (ocfs2_is_hard_readonly(osb))
1275 goto leave;
1276
1277 status = ocfs2_dlm_init(osb);
1278 if (status < 0) {
1279 mlog_errno(status);
1280 goto leave;
1281 }
1282
1283 status = ocfs2_super_lock(osb, 1);
1284 if (status < 0) {
1285 mlog_errno(status);
1286 goto leave;
1287 }
1288 unlock_super = 1;
1289
1290 /* This will load up the node map and add ourselves to it. */
1291 status = ocfs2_find_slot(osb);
1292 if (status < 0) {
1293 mlog_errno(status);
1294 goto leave;
1295 }
1296
1297 /* load all node-local system inodes */
1298 status = ocfs2_init_local_system_inodes(osb);
1299 if (status < 0) {
1300 mlog_errno(status);
1301 goto leave;
1302 }
1303
1304 status = ocfs2_check_volume(osb);
1305 if (status < 0) {
1306 mlog_errno(status);
1307 goto leave;
1308 }
1309
1310 status = ocfs2_truncate_log_init(osb);
1311 if (status < 0) {
1312 mlog_errno(status);
1313 goto leave;
1314 }
1315
1316 if (ocfs2_mount_local(osb))
1317 goto leave;
1318
1319 leave:
1320 if (unlock_super)
1321 ocfs2_super_unlock(osb, 1);
1322
1323 mlog_exit(status);
1324 return status;
1325 }
1326
1327 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1328 {
1329 int tmp, hangup_needed = 0;
1330 struct ocfs2_super *osb = NULL;
1331 char nodestr[8];
1332
1333 mlog_entry("(0x%p)\n", sb);
1334
1335 BUG_ON(!sb);
1336 osb = OCFS2_SB(sb);
1337 BUG_ON(!osb);
1338
1339 ocfs2_shutdown_local_alloc(osb);
1340
1341 ocfs2_truncate_log_shutdown(osb);
1342
1343 /* This will disable recovery and flush any recovery work. */
1344 ocfs2_recovery_exit(osb);
1345
1346 ocfs2_journal_shutdown(osb);
1347
1348 ocfs2_sync_blockdev(sb);
1349
1350 /* No cluster connection means we've failed during mount, so skip
1351 * all the steps which depended on that to complete. */
1352 if (osb->cconn) {
1353 tmp = ocfs2_super_lock(osb, 1);
1354 if (tmp < 0) {
1355 mlog_errno(tmp);
1356 return;
1357 }
1358 }
1359
1360 if (osb->slot_num != OCFS2_INVALID_SLOT)
1361 ocfs2_put_slot(osb);
1362
1363 if (osb->cconn)
1364 ocfs2_super_unlock(osb, 1);
1365
1366 ocfs2_release_system_inodes(osb);
1367
1368 /*
1369 * If we're dismounting due to mount error, mount.ocfs2 will clean
1370 * up heartbeat. If we're a local mount, there is no heartbeat.
1371 * If we failed before we got a uuid_str yet, we can't stop
1372 * heartbeat. Otherwise, do it.
1373 */
1374 if (!mnt_err && !ocfs2_mount_local(osb) && osb->uuid_str)
1375 hangup_needed = 1;
1376
1377 if (osb->cconn)
1378 ocfs2_dlm_shutdown(osb, hangup_needed);
1379
1380 debugfs_remove(osb->osb_debug_root);
1381
1382 if (hangup_needed)
1383 ocfs2_cluster_hangup(osb->uuid_str, strlen(osb->uuid_str));
1384
1385 atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1386
1387 if (ocfs2_mount_local(osb))
1388 snprintf(nodestr, sizeof(nodestr), "local");
1389 else
1390 snprintf(nodestr, sizeof(nodestr), "%u", osb->node_num);
1391
1392 printk(KERN_INFO "ocfs2: Unmounting device (%s) on (node %s)\n",
1393 osb->dev_str, nodestr);
1394
1395 ocfs2_delete_osb(osb);
1396 kfree(osb);
1397 sb->s_dev = 0;
1398 sb->s_fs_info = NULL;
1399 }
1400
1401 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1402 unsigned uuid_bytes)
1403 {
1404 int i, ret;
1405 char *ptr;
1406
1407 BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1408
1409 osb->uuid_str = kzalloc(OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1410 if (osb->uuid_str == NULL)
1411 return -ENOMEM;
1412
1413 for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1414 /* print with null */
1415 ret = snprintf(ptr, 3, "%02X", uuid[i]);
1416 if (ret != 2) /* drop super cleans up */
1417 return -EINVAL;
1418 /* then only advance past the last char */
1419 ptr += 2;
1420 }
1421
1422 return 0;
1423 }
1424
1425 static int ocfs2_initialize_super(struct super_block *sb,
1426 struct buffer_head *bh,
1427 int sector_size)
1428 {
1429 int status;
1430 int i, cbits, bbits;
1431 struct ocfs2_dinode *di = (struct ocfs2_dinode *)bh->b_data;
1432 struct inode *inode = NULL;
1433 struct ocfs2_journal *journal;
1434 __le32 uuid_net_key;
1435 struct ocfs2_super *osb;
1436
1437 mlog_entry_void();
1438
1439 osb = kzalloc(sizeof(struct ocfs2_super), GFP_KERNEL);
1440 if (!osb) {
1441 status = -ENOMEM;
1442 mlog_errno(status);
1443 goto bail;
1444 }
1445
1446 sb->s_fs_info = osb;
1447 sb->s_op = &ocfs2_sops;
1448 sb->s_export_op = &ocfs2_export_ops;
1449 sb->s_xattr = ocfs2_xattr_handlers;
1450 sb->s_time_gran = 1;
1451 sb->s_flags |= MS_NOATIME;
1452 /* this is needed to support O_LARGEFILE */
1453 cbits = le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1454 bbits = le32_to_cpu(di->id2.i_super.s_blocksize_bits);
1455 sb->s_maxbytes = ocfs2_max_file_offset(bbits, cbits);
1456
1457 osb->sb = sb;
1458 /* Save off for ocfs2_rw_direct */
1459 osb->s_sectsize_bits = blksize_bits(sector_size);
1460 BUG_ON(!osb->s_sectsize_bits);
1461
1462 spin_lock_init(&osb->dc_task_lock);
1463 init_waitqueue_head(&osb->dc_event);
1464 osb->dc_work_sequence = 0;
1465 osb->dc_wake_sequence = 0;
1466 INIT_LIST_HEAD(&osb->blocked_lock_list);
1467 osb->blocked_lock_count = 0;
1468 spin_lock_init(&osb->osb_lock);
1469 ocfs2_init_inode_steal_slot(osb);
1470
1471 atomic_set(&osb->alloc_stats.moves, 0);
1472 atomic_set(&osb->alloc_stats.local_data, 0);
1473 atomic_set(&osb->alloc_stats.bitmap_data, 0);
1474 atomic_set(&osb->alloc_stats.bg_allocs, 0);
1475 atomic_set(&osb->alloc_stats.bg_extends, 0);
1476
1477 ocfs2_init_node_maps(osb);
1478
1479 snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1480 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1481
1482 status = ocfs2_recovery_init(osb);
1483 if (status) {
1484 mlog(ML_ERROR, "Unable to initialize recovery state\n");
1485 mlog_errno(status);
1486 goto bail;
1487 }
1488
1489 init_waitqueue_head(&osb->checkpoint_event);
1490 atomic_set(&osb->needs_checkpoint, 0);
1491
1492 osb->s_atime_quantum = OCFS2_DEFAULT_ATIME_QUANTUM;
1493
1494 osb->slot_num = OCFS2_INVALID_SLOT;
1495
1496 osb->s_xattr_inline_size = le16_to_cpu(
1497 di->id2.i_super.s_xattr_inline_size);
1498
1499 osb->local_alloc_state = OCFS2_LA_UNUSED;
1500 osb->local_alloc_bh = NULL;
1501 INIT_DELAYED_WORK(&osb->la_enable_wq, ocfs2_la_enable_worker);
1502
1503 init_waitqueue_head(&osb->osb_mount_event);
1504
1505 osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1506 if (!osb->vol_label) {
1507 mlog(ML_ERROR, "unable to alloc vol label\n");
1508 status = -ENOMEM;
1509 goto bail;
1510 }
1511
1512 osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1513 if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1514 mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1515 osb->max_slots);
1516 status = -EINVAL;
1517 goto bail;
1518 }
1519 mlog(0, "max_slots for this device: %u\n", osb->max_slots);
1520
1521 osb->slot_recovery_generations =
1522 kcalloc(osb->max_slots, sizeof(*osb->slot_recovery_generations),
1523 GFP_KERNEL);
1524 if (!osb->slot_recovery_generations) {
1525 status = -ENOMEM;
1526 mlog_errno(status);
1527 goto bail;
1528 }
1529
1530 init_waitqueue_head(&osb->osb_wipe_event);
1531 osb->osb_orphan_wipes = kcalloc(osb->max_slots,
1532 sizeof(*osb->osb_orphan_wipes),
1533 GFP_KERNEL);
1534 if (!osb->osb_orphan_wipes) {
1535 status = -ENOMEM;
1536 mlog_errno(status);
1537 goto bail;
1538 }
1539
1540 osb->s_feature_compat =
1541 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1542 osb->s_feature_ro_compat =
1543 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1544 osb->s_feature_incompat =
1545 le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1546
1547 if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1548 mlog(ML_ERROR, "couldn't mount because of unsupported "
1549 "optional features (%x).\n", i);
1550 status = -EINVAL;
1551 goto bail;
1552 }
1553 if (!(osb->sb->s_flags & MS_RDONLY) &&
1554 (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1555 mlog(ML_ERROR, "couldn't mount RDWR because of "
1556 "unsupported optional features (%x).\n", i);
1557 status = -EINVAL;
1558 goto bail;
1559 }
1560
1561 if (ocfs2_userspace_stack(osb)) {
1562 memcpy(osb->osb_cluster_stack,
1563 OCFS2_RAW_SB(di)->s_cluster_info.ci_stack,
1564 OCFS2_STACK_LABEL_LEN);
1565 osb->osb_cluster_stack[OCFS2_STACK_LABEL_LEN] = '\0';
1566 if (strlen(osb->osb_cluster_stack) != OCFS2_STACK_LABEL_LEN) {
1567 mlog(ML_ERROR,
1568 "couldn't mount because of an invalid "
1569 "cluster stack label (%s) \n",
1570 osb->osb_cluster_stack);
1571 status = -EINVAL;
1572 goto bail;
1573 }
1574 } else {
1575 /* The empty string is identical with classic tools that
1576 * don't know about s_cluster_info. */
1577 osb->osb_cluster_stack[0] = '\0';
1578 }
1579
1580 get_random_bytes(&osb->s_next_generation, sizeof(u32));
1581
1582 /* FIXME
1583 * This should be done in ocfs2_journal_init(), but unknown
1584 * ordering issues will cause the filesystem to crash.
1585 * If anyone wants to figure out what part of the code
1586 * refers to osb->journal before ocfs2_journal_init() is run,
1587 * be my guest.
1588 */
1589 /* initialize our journal structure */
1590
1591 journal = kzalloc(sizeof(struct ocfs2_journal), GFP_KERNEL);
1592 if (!journal) {
1593 mlog(ML_ERROR, "unable to alloc journal\n");
1594 status = -ENOMEM;
1595 goto bail;
1596 }
1597 osb->journal = journal;
1598 journal->j_osb = osb;
1599
1600 atomic_set(&journal->j_num_trans, 0);
1601 init_rwsem(&journal->j_trans_barrier);
1602 init_waitqueue_head(&journal->j_checkpointed);
1603 spin_lock_init(&journal->j_lock);
1604 journal->j_trans_id = (unsigned long) 1;
1605 INIT_LIST_HEAD(&journal->j_la_cleanups);
1606 INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery);
1607 journal->j_state = OCFS2_JOURNAL_FREE;
1608
1609 /* get some pseudo constants for clustersize bits */
1610 osb->s_clustersize_bits =
1611 le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1612 osb->s_clustersize = 1 << osb->s_clustersize_bits;
1613 mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1614
1615 if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1616 osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1617 mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1618 osb->s_clustersize);
1619 status = -EINVAL;
1620 goto bail;
1621 }
1622
1623 if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1624 > (u32)~0UL) {
1625 mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1626 "what jbd can address in 32 bits.\n");
1627 status = -EINVAL;
1628 goto bail;
1629 }
1630
1631 if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1632 sizeof(di->id2.i_super.s_uuid))) {
1633 mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1634 status = -ENOMEM;
1635 goto bail;
1636 }
1637
1638 memcpy(&uuid_net_key, di->id2.i_super.s_uuid, sizeof(uuid_net_key));
1639
1640 strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1641 osb->vol_label[63] = '\0';
1642 osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1643 osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1644 osb->first_cluster_group_blkno =
1645 le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1646 osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1647 osb->uuid_hash = le32_to_cpu(di->id2.i_super.s_uuid_hash);
1648 mlog(0, "vol_label: %s\n", osb->vol_label);
1649 mlog(0, "uuid: %s\n", osb->uuid_str);
1650 mlog(0, "root_blkno=%llu, system_dir_blkno=%llu\n",
1651 (unsigned long long)osb->root_blkno,
1652 (unsigned long long)osb->system_dir_blkno);
1653
1654 osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1655 if (!osb->osb_dlm_debug) {
1656 status = -ENOMEM;
1657 mlog_errno(status);
1658 goto bail;
1659 }
1660
1661 atomic_set(&osb->vol_state, VOLUME_INIT);
1662
1663 /* load root, system_dir, and all global system inodes */
1664 status = ocfs2_init_global_system_inodes(osb);
1665 if (status < 0) {
1666 mlog_errno(status);
1667 goto bail;
1668 }
1669
1670 /*
1671 * global bitmap
1672 */
1673 inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1674 OCFS2_INVALID_SLOT);
1675 if (!inode) {
1676 status = -EINVAL;
1677 mlog_errno(status);
1678 goto bail;
1679 }
1680
1681 osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1682 iput(inode);
1683
1684 osb->bitmap_cpg = ocfs2_group_bitmap_size(sb) * 8;
1685
1686 status = ocfs2_init_slot_info(osb);
1687 if (status < 0) {
1688 mlog_errno(status);
1689 goto bail;
1690 }
1691
1692 bail:
1693 mlog_exit(status);
1694 return status;
1695 }
1696
1697 /*
1698 * will return: -EAGAIN if it is ok to keep searching for superblocks
1699 * -EINVAL if there is a bad superblock
1700 * 0 on success
1701 */
1702 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1703 struct buffer_head *bh,
1704 u32 blksz)
1705 {
1706 int status = -EAGAIN;
1707
1708 mlog_entry_void();
1709
1710 if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1711 strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1712 status = -EINVAL;
1713 if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1714 mlog(ML_ERROR, "found superblock with incorrect block "
1715 "size: found %u, should be %u\n",
1716 1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1717 blksz);
1718 } else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1719 OCFS2_MAJOR_REV_LEVEL ||
1720 le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1721 OCFS2_MINOR_REV_LEVEL) {
1722 mlog(ML_ERROR, "found superblock with bad version: "
1723 "found %u.%u, should be %u.%u\n",
1724 le16_to_cpu(di->id2.i_super.s_major_rev_level),
1725 le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1726 OCFS2_MAJOR_REV_LEVEL,
1727 OCFS2_MINOR_REV_LEVEL);
1728 } else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1729 mlog(ML_ERROR, "bad block number on superblock: "
1730 "found %llu, should be %llu\n",
1731 (unsigned long long)le64_to_cpu(di->i_blkno),
1732 (unsigned long long)bh->b_blocknr);
1733 } else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1734 le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1735 mlog(ML_ERROR, "bad cluster size found: %u\n",
1736 1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1737 } else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1738 mlog(ML_ERROR, "bad root_blkno: 0\n");
1739 } else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1740 mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1741 } else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1742 mlog(ML_ERROR,
1743 "Superblock slots found greater than file system "
1744 "maximum: found %u, max %u\n",
1745 le16_to_cpu(di->id2.i_super.s_max_slots),
1746 OCFS2_MAX_SLOTS);
1747 } else {
1748 /* found it! */
1749 status = 0;
1750 }
1751 }
1752
1753 mlog_exit(status);
1754 return status;
1755 }
1756
1757 static int ocfs2_check_volume(struct ocfs2_super *osb)
1758 {
1759 int status;
1760 int dirty;
1761 int local;
1762 struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1763 * recover
1764 * ourselves. */
1765
1766 mlog_entry_void();
1767
1768 /* Init our journal object. */
1769 status = ocfs2_journal_init(osb->journal, &dirty);
1770 if (status < 0) {
1771 mlog(ML_ERROR, "Could not initialize journal!\n");
1772 goto finally;
1773 }
1774
1775 /* If the journal was unmounted cleanly then we don't want to
1776 * recover anything. Otherwise, journal_load will do that
1777 * dirty work for us :) */
1778 if (!dirty) {
1779 status = ocfs2_journal_wipe(osb->journal, 0);
1780 if (status < 0) {
1781 mlog_errno(status);
1782 goto finally;
1783 }
1784 } else {
1785 mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1786 "recovering volume.\n");
1787 }
1788
1789 local = ocfs2_mount_local(osb);
1790
1791 /* will play back anything left in the journal. */
1792 status = ocfs2_journal_load(osb->journal, local, dirty);
1793 if (status < 0) {
1794 mlog(ML_ERROR, "ocfs2 journal load failed! %d\n", status);
1795 goto finally;
1796 }
1797
1798 if (dirty) {
1799 /* recover my local alloc if we didn't unmount cleanly. */
1800 status = ocfs2_begin_local_alloc_recovery(osb,
1801 osb->slot_num,
1802 &local_alloc);
1803 if (status < 0) {
1804 mlog_errno(status);
1805 goto finally;
1806 }
1807 /* we complete the recovery process after we've marked
1808 * ourselves as mounted. */
1809 }
1810
1811 mlog(0, "Journal loaded.\n");
1812
1813 status = ocfs2_load_local_alloc(osb);
1814 if (status < 0) {
1815 mlog_errno(status);
1816 goto finally;
1817 }
1818
1819 if (dirty) {
1820 /* Recovery will be completed after we've mounted the
1821 * rest of the volume. */
1822 osb->dirty = 1;
1823 osb->local_alloc_copy = local_alloc;
1824 local_alloc = NULL;
1825 }
1826
1827 /* go through each journal, trylock it and if you get the
1828 * lock, and it's marked as dirty, set the bit in the recover
1829 * map and launch a recovery thread for it. */
1830 status = ocfs2_mark_dead_nodes(osb);
1831 if (status < 0)
1832 mlog_errno(status);
1833
1834 finally:
1835 if (local_alloc)
1836 kfree(local_alloc);
1837
1838 mlog_exit(status);
1839 return status;
1840 }
1841
1842 /*
1843 * The routine gets called from dismount or close whenever a dismount on
1844 * volume is requested and the osb open count becomes 1.
1845 * It will remove the osb from the global list and also free up all the
1846 * initialized resources and fileobject.
1847 */
1848 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1849 {
1850 mlog_entry_void();
1851
1852 /* This function assumes that the caller has the main osb resource */
1853
1854 ocfs2_free_slot_info(osb);
1855
1856 kfree(osb->osb_orphan_wipes);
1857 kfree(osb->slot_recovery_generations);
1858 /* FIXME
1859 * This belongs in journal shutdown, but because we have to
1860 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1861 * we free it here.
1862 */
1863 kfree(osb->journal);
1864 if (osb->local_alloc_copy)
1865 kfree(osb->local_alloc_copy);
1866 kfree(osb->uuid_str);
1867 ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1868 memset(osb, 0, sizeof(struct ocfs2_super));
1869
1870 mlog_exit_void();
1871 }
1872
1873 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1874 * panic(). We do not support continue-on-error operation. */
1875 static void ocfs2_handle_error(struct super_block *sb)
1876 {
1877 struct ocfs2_super *osb = OCFS2_SB(sb);
1878
1879 if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1880 panic("OCFS2: (device %s): panic forced after error\n",
1881 sb->s_id);
1882
1883 ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1884
1885 if (sb->s_flags & MS_RDONLY &&
1886 (ocfs2_is_soft_readonly(osb) ||
1887 ocfs2_is_hard_readonly(osb)))
1888 return;
1889
1890 printk(KERN_CRIT "File system is now read-only due to the potential "
1891 "of on-disk corruption. Please run fsck.ocfs2 once the file "
1892 "system is unmounted.\n");
1893 sb->s_flags |= MS_RDONLY;
1894 ocfs2_set_ro_flag(osb, 0);
1895 }
1896
1897 static char error_buf[1024];
1898
1899 void __ocfs2_error(struct super_block *sb,
1900 const char *function,
1901 const char *fmt, ...)
1902 {
1903 va_list args;
1904
1905 va_start(args, fmt);
1906 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1907 va_end(args);
1908
1909 /* Not using mlog here because we want to show the actual
1910 * function the error came from. */
1911 printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1912 sb->s_id, function, error_buf);
1913
1914 ocfs2_handle_error(sb);
1915 }
1916
1917 /* Handle critical errors. This is intentionally more drastic than
1918 * ocfs2_handle_error, so we only use for things like journal errors,
1919 * etc. */
1920 void __ocfs2_abort(struct super_block* sb,
1921 const char *function,
1922 const char *fmt, ...)
1923 {
1924 va_list args;
1925
1926 va_start(args, fmt);
1927 vsnprintf(error_buf, sizeof(error_buf), fmt, args);
1928 va_end(args);
1929
1930 printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1931 sb->s_id, function, error_buf);
1932
1933 /* We don't have the cluster support yet to go straight to
1934 * hard readonly in here. Until then, we want to keep
1935 * ocfs2_abort() so that we can at least mark critical
1936 * errors.
1937 *
1938 * TODO: This should abort the journal and alert other nodes
1939 * that our slot needs recovery. */
1940
1941 /* Force a panic(). This stinks, but it's better than letting
1942 * things continue without having a proper hard readonly
1943 * here. */
1944 OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1945 ocfs2_handle_error(sb);
1946 }
1947
1948 module_init(ocfs2_init);
1949 module_exit(ocfs2_exit);
Linux kernel - Deliverables
This page took 0.071742 seconds and 5 git commands to generate.