xref: /openbmc/linux/fs/ocfs2/super.c (revision 87c2ce3b)
1 /* -*- mode: c; c-basic-offset: 8; -*-
2  * vim: noexpandtab sw=8 ts=8 sts=0:
3  *
4  * super.c
5  *
6  * load/unload driver, mount/dismount volumes
7  *
8  * Copyright (C) 2002, 2004 Oracle.  All rights reserved.
9  *
10  * This program is free software; you can redistribute it and/or
11  * modify it under the terms of the GNU General Public
12  * License as published by the Free Software Foundation; either
13  * version 2 of the License, or (at your option) any later version.
14  *
15  * This program is distributed in the hope that it will be useful,
16  * but WITHOUT ANY WARRANTY; without even the implied warranty of
17  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
18  * General Public License for more details.
19  *
20  * You should have received a copy of the GNU General Public
21  * License along with this program; if not, write to the
22  * Free Software Foundation, Inc., 59 Temple Place - Suite 330,
23  * Boston, MA 021110-1307, USA.
24  */
25 
26 #include <linux/module.h>
27 #include <linux/fs.h>
28 #include <linux/types.h>
29 #include <linux/slab.h>
30 #include <linux/highmem.h>
31 #include <linux/utsname.h>
32 #include <linux/init.h>
33 #include <linux/random.h>
34 #include <linux/statfs.h>
35 #include <linux/moduleparam.h>
36 #include <linux/blkdev.h>
37 #include <linux/socket.h>
38 #include <linux/inet.h>
39 #include <linux/parser.h>
40 #include <linux/crc32.h>
41 #include <linux/debugfs.h>
42 
43 #include <cluster/nodemanager.h>
44 
45 #define MLOG_MASK_PREFIX ML_SUPER
46 #include <cluster/masklog.h>
47 
48 #include "ocfs2.h"
49 
50 /* this should be the only file to include a version 1 header */
51 #include "ocfs1_fs_compat.h"
52 
53 #include "alloc.h"
54 #include "dlmglue.h"
55 #include "export.h"
56 #include "extent_map.h"
57 #include "heartbeat.h"
58 #include "inode.h"
59 #include "journal.h"
60 #include "localalloc.h"
61 #include "namei.h"
62 #include "slot_map.h"
63 #include "super.h"
64 #include "sysfile.h"
65 #include "uptodate.h"
66 #include "ver.h"
67 #include "vote.h"
68 
69 #include "buffer_head_io.h"
70 
71 /*
72  * Globals
73  */
74 static spinlock_t ocfs2_globals_lock = SPIN_LOCK_UNLOCKED;
75 
76 static u32 osb_id;             /* Keeps track of next available OSB Id */
77 
78 static kmem_cache_t *ocfs2_inode_cachep = NULL;
79 
80 kmem_cache_t *ocfs2_lock_cache = NULL;
81 
82 /* OCFS2 needs to schedule several differnt types of work which
83  * require cluster locking, disk I/O, recovery waits, etc. Since these
84  * types of work tend to be heavy we avoid using the kernel events
85  * workqueue and schedule on our own. */
86 struct workqueue_struct *ocfs2_wq = NULL;
87 
88 static struct dentry *ocfs2_debugfs_root = NULL;
89 
90 MODULE_AUTHOR("Oracle");
91 MODULE_LICENSE("GPL");
92 
93 static int ocfs2_parse_options(struct super_block *sb, char *options,
94 			       unsigned long *mount_opt, int is_remount);
95 static void ocfs2_put_super(struct super_block *sb);
96 static int ocfs2_mount_volume(struct super_block *sb);
97 static int ocfs2_remount(struct super_block *sb, int *flags, char *data);
98 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err);
99 static int ocfs2_initialize_mem_caches(void);
100 static void ocfs2_free_mem_caches(void);
101 static void ocfs2_delete_osb(struct ocfs2_super *osb);
102 
103 static int ocfs2_statfs(struct super_block *sb, struct kstatfs *buf);
104 
105 static int ocfs2_sync_fs(struct super_block *sb, int wait);
106 
107 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb);
108 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb);
109 static int ocfs2_release_system_inodes(struct ocfs2_super *osb);
110 static int ocfs2_fill_local_node_info(struct ocfs2_super *osb);
111 static int ocfs2_check_volume(struct ocfs2_super *osb);
112 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
113 			       struct buffer_head *bh,
114 			       u32 sectsize);
115 static int ocfs2_initialize_super(struct super_block *sb,
116 				  struct buffer_head *bh,
117 				  int sector_size);
118 static int ocfs2_get_sector(struct super_block *sb,
119 			    struct buffer_head **bh,
120 			    int block,
121 			    int sect_size);
122 static void ocfs2_write_super(struct super_block *sb);
123 static struct inode *ocfs2_alloc_inode(struct super_block *sb);
124 static void ocfs2_destroy_inode(struct inode *inode);
125 
126 static unsigned long long ocfs2_max_file_offset(unsigned int blockshift);
127 
128 static struct super_operations ocfs2_sops = {
129 	.statfs		= ocfs2_statfs,
130 	.alloc_inode	= ocfs2_alloc_inode,
131 	.destroy_inode	= ocfs2_destroy_inode,
132 	.drop_inode	= ocfs2_drop_inode,
133 	.clear_inode	= ocfs2_clear_inode,
134 	.delete_inode	= ocfs2_delete_inode,
135 	.sync_fs	= ocfs2_sync_fs,
136 	.write_super	= ocfs2_write_super,
137 	.put_super	= ocfs2_put_super,
138 	.remount_fs	= ocfs2_remount,
139 };
140 
141 enum {
142 	Opt_barrier,
143 	Opt_err_panic,
144 	Opt_err_ro,
145 	Opt_intr,
146 	Opt_nointr,
147 	Opt_hb_none,
148 	Opt_hb_local,
149 	Opt_data_ordered,
150 	Opt_data_writeback,
151 	Opt_err,
152 };
153 
154 static match_table_t tokens = {
155 	{Opt_barrier, "barrier=%u"},
156 	{Opt_err_panic, "errors=panic"},
157 	{Opt_err_ro, "errors=remount-ro"},
158 	{Opt_intr, "intr"},
159 	{Opt_nointr, "nointr"},
160 	{Opt_hb_none, OCFS2_HB_NONE},
161 	{Opt_hb_local, OCFS2_HB_LOCAL},
162 	{Opt_data_ordered, "data=ordered"},
163 	{Opt_data_writeback, "data=writeback"},
164 	{Opt_err, NULL}
165 };
166 
167 /*
168  * write_super and sync_fs ripped right out of ext3.
169  */
170 static void ocfs2_write_super(struct super_block *sb)
171 {
172 	if (mutex_trylock(&sb->s_lock) != 0)
173 		BUG();
174 	sb->s_dirt = 0;
175 }
176 
177 static int ocfs2_sync_fs(struct super_block *sb, int wait)
178 {
179 	int status = 0;
180 	tid_t target;
181 	struct ocfs2_super *osb = OCFS2_SB(sb);
182 
183 	sb->s_dirt = 0;
184 
185 	if (ocfs2_is_hard_readonly(osb))
186 		return -EROFS;
187 
188 	if (wait) {
189 		status = ocfs2_flush_truncate_log(osb);
190 		if (status < 0)
191 			mlog_errno(status);
192 	} else {
193 		ocfs2_schedule_truncate_log_flush(osb, 0);
194 	}
195 
196 	if (journal_start_commit(OCFS2_SB(sb)->journal->j_journal, &target)) {
197 		if (wait)
198 			log_wait_commit(OCFS2_SB(sb)->journal->j_journal,
199 					target);
200 	}
201 	return 0;
202 }
203 
204 static int ocfs2_init_global_system_inodes(struct ocfs2_super *osb)
205 {
206 	struct inode *new = NULL;
207 	int status = 0;
208 	int i;
209 
210 	mlog_entry_void();
211 
212 	new = ocfs2_iget(osb, osb->root_blkno);
213 	if (IS_ERR(new)) {
214 		status = PTR_ERR(new);
215 		mlog_errno(status);
216 		goto bail;
217 	}
218 	osb->root_inode = new;
219 
220 	new = ocfs2_iget(osb, osb->system_dir_blkno);
221 	if (IS_ERR(new)) {
222 		status = PTR_ERR(new);
223 		mlog_errno(status);
224 		goto bail;
225 	}
226 	osb->sys_root_inode = new;
227 
228 	for (i = OCFS2_FIRST_ONLINE_SYSTEM_INODE;
229 	     i <= OCFS2_LAST_GLOBAL_SYSTEM_INODE; i++) {
230 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
231 		if (!new) {
232 			ocfs2_release_system_inodes(osb);
233 			status = -EINVAL;
234 			mlog_errno(status);
235 			/* FIXME: Should ERROR_RO_FS */
236 			mlog(ML_ERROR, "Unable to load system inode %d, "
237 			     "possibly corrupt fs?", i);
238 			goto bail;
239 		}
240 		// the array now has one ref, so drop this one
241 		iput(new);
242 	}
243 
244 bail:
245 	mlog_exit(status);
246 	return status;
247 }
248 
249 static int ocfs2_init_local_system_inodes(struct ocfs2_super *osb)
250 {
251 	struct inode *new = NULL;
252 	int status = 0;
253 	int i;
254 
255 	mlog_entry_void();
256 
257 	for (i = OCFS2_LAST_GLOBAL_SYSTEM_INODE + 1;
258 	     i < NUM_SYSTEM_INODES;
259 	     i++) {
260 		new = ocfs2_get_system_file_inode(osb, i, osb->slot_num);
261 		if (!new) {
262 			ocfs2_release_system_inodes(osb);
263 			status = -EINVAL;
264 			mlog(ML_ERROR, "status=%d, sysfile=%d, slot=%d\n",
265 			     status, i, osb->slot_num);
266 			goto bail;
267 		}
268 		/* the array now has one ref, so drop this one */
269 		iput(new);
270 	}
271 
272 bail:
273 	mlog_exit(status);
274 	return status;
275 }
276 
277 static int ocfs2_release_system_inodes(struct ocfs2_super *osb)
278 {
279 	int status = 0, i;
280 	struct inode *inode;
281 
282 	mlog_entry_void();
283 
284 	for (i = 0; i < NUM_SYSTEM_INODES; i++) {
285 		inode = osb->system_inodes[i];
286 		if (inode) {
287 			iput(inode);
288 			osb->system_inodes[i] = NULL;
289 		}
290 	}
291 
292 	inode = osb->sys_root_inode;
293 	if (inode) {
294 		iput(inode);
295 		osb->sys_root_inode = NULL;
296 	}
297 
298 	inode = osb->root_inode;
299 	if (inode) {
300 		iput(inode);
301 		osb->root_inode = NULL;
302 	}
303 
304 	mlog_exit(status);
305 	return status;
306 }
307 
308 /* We're allocating fs objects, use GFP_NOFS */
309 static struct inode *ocfs2_alloc_inode(struct super_block *sb)
310 {
311 	struct ocfs2_inode_info *oi;
312 
313 	oi = kmem_cache_alloc(ocfs2_inode_cachep, SLAB_NOFS);
314 	if (!oi)
315 		return NULL;
316 
317 	return &oi->vfs_inode;
318 }
319 
320 static void ocfs2_destroy_inode(struct inode *inode)
321 {
322 	kmem_cache_free(ocfs2_inode_cachep, OCFS2_I(inode));
323 }
324 
325 /* From xfs_super.c:xfs_max_file_offset
326  * Copyright (c) 2000-2004 Silicon Graphics, Inc.
327  */
328 static unsigned long long ocfs2_max_file_offset(unsigned int blockshift)
329 {
330 	unsigned int pagefactor = 1;
331 	unsigned int bitshift = BITS_PER_LONG - 1;
332 
333 	/* Figure out maximum filesize, on Linux this can depend on
334 	 * the filesystem blocksize (on 32 bit platforms).
335 	 * __block_prepare_write does this in an [unsigned] long...
336 	 *      page->index << (PAGE_CACHE_SHIFT - bbits)
337 	 * So, for page sized blocks (4K on 32 bit platforms),
338 	 * this wraps at around 8Tb (hence MAX_LFS_FILESIZE which is
339 	 *      (((u64)PAGE_CACHE_SIZE << (BITS_PER_LONG-1))-1)
340 	 * but for smaller blocksizes it is less (bbits = log2 bsize).
341 	 * Note1: get_block_t takes a long (implicit cast from above)
342 	 * Note2: The Large Block Device (LBD and HAVE_SECTOR_T) patch
343 	 * can optionally convert the [unsigned] long from above into
344 	 * an [unsigned] long long.
345 	 */
346 
347 #if BITS_PER_LONG == 32
348 # if defined(CONFIG_LBD)
349 	BUG_ON(sizeof(sector_t) != 8);
350 	pagefactor = PAGE_CACHE_SIZE;
351 	bitshift = BITS_PER_LONG;
352 # else
353 	pagefactor = PAGE_CACHE_SIZE >> (PAGE_CACHE_SHIFT - blockshift);
354 # endif
355 #endif
356 
357 	return (((unsigned long long)pagefactor) << bitshift) - 1;
358 }
359 
360 static int ocfs2_remount(struct super_block *sb, int *flags, char *data)
361 {
362 	int incompat_features;
363 	int ret = 0;
364 	unsigned long parsed_options;
365 	struct ocfs2_super *osb = OCFS2_SB(sb);
366 
367 	if (!ocfs2_parse_options(sb, data, &parsed_options, 1)) {
368 		ret = -EINVAL;
369 		goto out;
370 	}
371 
372 	if ((osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) !=
373 	    (parsed_options & OCFS2_MOUNT_HB_LOCAL)) {
374 		ret = -EINVAL;
375 		mlog(ML_ERROR, "Cannot change heartbeat mode on remount\n");
376 		goto out;
377 	}
378 
379 	if ((osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK) !=
380 	    (parsed_options & OCFS2_MOUNT_DATA_WRITEBACK)) {
381 		ret = -EINVAL;
382 		mlog(ML_ERROR, "Cannot change data mode on remount\n");
383 		goto out;
384 	}
385 
386 	/* We're going to/from readonly mode. */
387 	if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
388 		/* Lock here so the check of HARD_RO and the potential
389 		 * setting of SOFT_RO is atomic. */
390 		spin_lock(&osb->osb_lock);
391 		if (osb->osb_flags & OCFS2_OSB_HARD_RO) {
392 			mlog(ML_ERROR, "Remount on readonly device is forbidden.\n");
393 			ret = -EROFS;
394 			goto unlock_osb;
395 		}
396 
397 		if (*flags & MS_RDONLY) {
398 			mlog(0, "Going to ro mode.\n");
399 			sb->s_flags |= MS_RDONLY;
400 			osb->osb_flags |= OCFS2_OSB_SOFT_RO;
401 		} else {
402 			mlog(0, "Making ro filesystem writeable.\n");
403 
404 			if (osb->osb_flags & OCFS2_OSB_ERROR_FS) {
405 				mlog(ML_ERROR, "Cannot remount RDWR "
406 				     "filesystem due to previous errors.\n");
407 				ret = -EROFS;
408 				goto unlock_osb;
409 			}
410 			incompat_features = OCFS2_HAS_RO_COMPAT_FEATURE(sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP);
411 			if (incompat_features) {
412 				mlog(ML_ERROR, "Cannot remount RDWR because "
413 				     "of unsupported optional features "
414 				     "(%x).\n", incompat_features);
415 				ret = -EINVAL;
416 				goto unlock_osb;
417 			}
418 			sb->s_flags &= ~MS_RDONLY;
419 			osb->osb_flags &= ~OCFS2_OSB_SOFT_RO;
420 		}
421 unlock_osb:
422 		spin_unlock(&osb->osb_lock);
423 	}
424 
425 	if (!ret) {
426 		if (!ocfs2_is_hard_readonly(osb))
427 			ocfs2_set_journal_params(osb);
428 
429 		/* Only save off the new mount options in case of a successful
430 		 * remount. */
431 		osb->s_mount_opt = parsed_options;
432 	}
433 out:
434 	return ret;
435 }
436 
437 static int ocfs2_sb_probe(struct super_block *sb,
438 			  struct buffer_head **bh,
439 			  int *sector_size)
440 {
441 	int status = 0, tmpstat;
442 	struct ocfs1_vol_disk_hdr *hdr;
443 	struct ocfs2_dinode *di;
444 	int blksize;
445 
446 	*bh = NULL;
447 
448 	/* may be > 512 */
449 	*sector_size = bdev_hardsect_size(sb->s_bdev);
450 	if (*sector_size > OCFS2_MAX_BLOCKSIZE) {
451 		mlog(ML_ERROR, "Hardware sector size too large: %d (max=%d)\n",
452 		     *sector_size, OCFS2_MAX_BLOCKSIZE);
453 		status = -EINVAL;
454 		goto bail;
455 	}
456 
457 	/* Can this really happen? */
458 	if (*sector_size < OCFS2_MIN_BLOCKSIZE)
459 		*sector_size = OCFS2_MIN_BLOCKSIZE;
460 
461 	/* check block zero for old format */
462 	status = ocfs2_get_sector(sb, bh, 0, *sector_size);
463 	if (status < 0) {
464 		mlog_errno(status);
465 		goto bail;
466 	}
467 	hdr = (struct ocfs1_vol_disk_hdr *) (*bh)->b_data;
468 	if (hdr->major_version == OCFS1_MAJOR_VERSION) {
469 		mlog(ML_ERROR, "incompatible version: %u.%u\n",
470 		     hdr->major_version, hdr->minor_version);
471 		status = -EINVAL;
472 	}
473 	if (memcmp(hdr->signature, OCFS1_VOLUME_SIGNATURE,
474 		   strlen(OCFS1_VOLUME_SIGNATURE)) == 0) {
475 		mlog(ML_ERROR, "incompatible volume signature: %8s\n",
476 		     hdr->signature);
477 		status = -EINVAL;
478 	}
479 	brelse(*bh);
480 	*bh = NULL;
481 	if (status < 0) {
482 		mlog(ML_ERROR, "This is an ocfs v1 filesystem which must be "
483 		     "upgraded before mounting with ocfs v2\n");
484 		goto bail;
485 	}
486 
487 	/*
488 	 * Now check at magic offset for 512, 1024, 2048, 4096
489 	 * blocksizes.  4096 is the maximum blocksize because it is
490 	 * the minimum clustersize.
491 	 */
492 	status = -EINVAL;
493 	for (blksize = *sector_size;
494 	     blksize <= OCFS2_MAX_BLOCKSIZE;
495 	     blksize <<= 1) {
496 		tmpstat = ocfs2_get_sector(sb, bh,
497 					   OCFS2_SUPER_BLOCK_BLKNO,
498 					   blksize);
499 		if (tmpstat < 0) {
500 			status = tmpstat;
501 			mlog_errno(status);
502 			goto bail;
503 		}
504 		di = (struct ocfs2_dinode *) (*bh)->b_data;
505 		status = ocfs2_verify_volume(di, *bh, blksize);
506 		if (status >= 0)
507 			goto bail;
508 		brelse(*bh);
509 		*bh = NULL;
510 		if (status != -EAGAIN)
511 			break;
512 	}
513 
514 bail:
515 	return status;
516 }
517 
518 static int ocfs2_fill_super(struct super_block *sb, void *data, int silent)
519 {
520 	struct dentry *root;
521 	int status, sector_size;
522 	unsigned long parsed_opt;
523 	struct inode *inode = NULL;
524 	struct ocfs2_super *osb = NULL;
525 	struct buffer_head *bh = NULL;
526 
527 	mlog_entry("%p, %p, %i", sb, data, silent);
528 
529 	/* for now we only have one cluster/node, make sure we see it
530 	 * in the heartbeat universe */
531 	if (!o2hb_check_local_node_heartbeating()) {
532 		status = -EINVAL;
533 		goto read_super_error;
534 	}
535 
536 	/* probe for superblock */
537 	status = ocfs2_sb_probe(sb, &bh, &sector_size);
538 	if (status < 0) {
539 		mlog(ML_ERROR, "superblock probe failed!\n");
540 		goto read_super_error;
541 	}
542 
543 	status = ocfs2_initialize_super(sb, bh, sector_size);
544 	osb = OCFS2_SB(sb);
545 	if (status < 0) {
546 		mlog_errno(status);
547 		goto read_super_error;
548 	}
549 	brelse(bh);
550 	bh = NULL;
551 
552 	if (!ocfs2_parse_options(sb, data, &parsed_opt, 0)) {
553 		status = -EINVAL;
554 		goto read_super_error;
555 	}
556 	osb->s_mount_opt = parsed_opt;
557 
558 	sb->s_magic = OCFS2_SUPER_MAGIC;
559 
560 	/* Hard readonly mode only if: bdev_read_only, MS_RDONLY,
561 	 * heartbeat=none */
562 	if (bdev_read_only(sb->s_bdev)) {
563 		if (!(sb->s_flags & MS_RDONLY)) {
564 			status = -EACCES;
565 			mlog(ML_ERROR, "Readonly device detected but readonly "
566 			     "mount was not specified.\n");
567 			goto read_super_error;
568 		}
569 
570 		/* You should not be able to start a local heartbeat
571 		 * on a readonly device. */
572 		if (osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL) {
573 			status = -EROFS;
574 			mlog(ML_ERROR, "Local heartbeat specified on readonly "
575 			     "device.\n");
576 			goto read_super_error;
577 		}
578 
579 		status = ocfs2_check_journals_nolocks(osb);
580 		if (status < 0) {
581 			if (status == -EROFS)
582 				mlog(ML_ERROR, "Recovery required on readonly "
583 				     "file system, but write access is "
584 				     "unavailable.\n");
585 			else
586 				mlog_errno(status);
587 			goto read_super_error;
588 		}
589 
590 		ocfs2_set_ro_flag(osb, 1);
591 
592 		printk(KERN_NOTICE "Readonly device detected. No cluster "
593 		       "services will be utilized for this mount. Recovery "
594 		       "will be skipped.\n");
595 	}
596 
597 	if (!ocfs2_is_hard_readonly(osb)) {
598 		/* If this isn't a hard readonly mount, then we need
599 		 * to make sure that heartbeat is in a valid state,
600 		 * and that we mark ourselves soft readonly is -oro
601 		 * was specified. */
602 		if (!(osb->s_mount_opt & OCFS2_MOUNT_HB_LOCAL)) {
603 			mlog(ML_ERROR, "No heartbeat for device (%s)\n",
604 			     sb->s_id);
605 			status = -EINVAL;
606 			goto read_super_error;
607 		}
608 
609 		if (sb->s_flags & MS_RDONLY)
610 			ocfs2_set_ro_flag(osb, 0);
611 	}
612 
613 	osb->osb_debug_root = debugfs_create_dir(osb->uuid_str,
614 						 ocfs2_debugfs_root);
615 	if (!osb->osb_debug_root) {
616 		status = -EINVAL;
617 		mlog(ML_ERROR, "Unable to create per-mount debugfs root.\n");
618 		goto read_super_error;
619 	}
620 
621 	status = ocfs2_mount_volume(sb);
622 	if (osb->root_inode)
623 		inode = igrab(osb->root_inode);
624 
625 	if (status < 0)
626 		goto read_super_error;
627 
628 	if (!inode) {
629 		status = -EIO;
630 		mlog_errno(status);
631 		goto read_super_error;
632 	}
633 
634 	root = d_alloc_root(inode);
635 	if (!root) {
636 		status = -ENOMEM;
637 		mlog_errno(status);
638 		goto read_super_error;
639 	}
640 
641 	sb->s_root = root;
642 
643 	ocfs2_complete_mount_recovery(osb);
644 
645 	printk("ocfs2: Mounting device (%u,%u) on (node %d, slot %d) with %s "
646 	       "data mode.\n",
647 	       MAJOR(sb->s_dev), MINOR(sb->s_dev), osb->node_num,
648 	       osb->slot_num,
649 	       osb->s_mount_opt & OCFS2_MOUNT_DATA_WRITEBACK ? "writeback" :
650 	       "ordered");
651 
652 	atomic_set(&osb->vol_state, VOLUME_MOUNTED);
653 	wake_up(&osb->osb_mount_event);
654 
655 	mlog_exit(status);
656 	return status;
657 
658 read_super_error:
659 	if (bh != NULL)
660 		brelse(bh);
661 
662 	if (inode)
663 		iput(inode);
664 
665 	if (osb) {
666 		atomic_set(&osb->vol_state, VOLUME_DISABLED);
667 		wake_up(&osb->osb_mount_event);
668 		ocfs2_dismount_volume(sb, 1);
669 	}
670 
671 	mlog_exit(status);
672 	return status;
673 }
674 
675 static struct super_block *ocfs2_get_sb(struct file_system_type *fs_type,
676 					int flags,
677 					const char *dev_name,
678 					void *data)
679 {
680 	return get_sb_bdev(fs_type, flags, dev_name, data, ocfs2_fill_super);
681 }
682 
683 static struct file_system_type ocfs2_fs_type = {
684 	.owner          = THIS_MODULE,
685 	.name           = "ocfs2",
686 	.get_sb         = ocfs2_get_sb, /* is this called when we mount
687 					* the fs? */
688 	.kill_sb        = kill_block_super, /* set to the generic one
689 					     * right now, but do we
690 					     * need to change that? */
691 	.fs_flags       = FS_REQUIRES_DEV,
692 	.next           = NULL
693 };
694 
695 static int ocfs2_parse_options(struct super_block *sb,
696 			       char *options,
697 			       unsigned long *mount_opt,
698 			       int is_remount)
699 {
700 	int status;
701 	char *p;
702 
703 	mlog_entry("remount: %d, options: \"%s\"\n", is_remount,
704 		   options ? options : "(none)");
705 
706 	*mount_opt = 0;
707 
708 	if (!options) {
709 		status = 1;
710 		goto bail;
711 	}
712 
713 	while ((p = strsep(&options, ",")) != NULL) {
714 		int token, option;
715 		substring_t args[MAX_OPT_ARGS];
716 
717 		if (!*p)
718 			continue;
719 
720 		token = match_token(p, tokens, args);
721 		switch (token) {
722 		case Opt_hb_local:
723 			*mount_opt |= OCFS2_MOUNT_HB_LOCAL;
724 			break;
725 		case Opt_hb_none:
726 			*mount_opt &= ~OCFS2_MOUNT_HB_LOCAL;
727 			break;
728 		case Opt_barrier:
729 			if (match_int(&args[0], &option)) {
730 				status = 0;
731 				goto bail;
732 			}
733 			if (option)
734 				*mount_opt |= OCFS2_MOUNT_BARRIER;
735 			else
736 				*mount_opt &= ~OCFS2_MOUNT_BARRIER;
737 			break;
738 		case Opt_intr:
739 			*mount_opt &= ~OCFS2_MOUNT_NOINTR;
740 			break;
741 		case Opt_nointr:
742 			*mount_opt |= OCFS2_MOUNT_NOINTR;
743 			break;
744 		case Opt_err_panic:
745 			*mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
746 			break;
747 		case Opt_err_ro:
748 			*mount_opt &= ~OCFS2_MOUNT_ERRORS_PANIC;
749 			break;
750 		case Opt_data_ordered:
751 			*mount_opt &= ~OCFS2_MOUNT_DATA_WRITEBACK;
752 			break;
753 		case Opt_data_writeback:
754 			*mount_opt |= OCFS2_MOUNT_DATA_WRITEBACK;
755 			break;
756 		default:
757 			mlog(ML_ERROR,
758 			     "Unrecognized mount option \"%s\" "
759 			     "or missing value\n", p);
760 			status = 0;
761 			goto bail;
762 		}
763 	}
764 
765 	status = 1;
766 
767 bail:
768 	mlog_exit(status);
769 	return status;
770 }
771 
772 static int __init ocfs2_init(void)
773 {
774 	int status;
775 
776 	mlog_entry_void();
777 
778 	ocfs2_print_version();
779 
780 	if (init_ocfs2_extent_maps())
781 		return -ENOMEM;
782 
783 	status = init_ocfs2_uptodate_cache();
784 	if (status < 0) {
785 		mlog_errno(status);
786 		goto leave;
787 	}
788 
789 	status = ocfs2_initialize_mem_caches();
790 	if (status < 0) {
791 		mlog_errno(status);
792 		goto leave;
793 	}
794 
795 	ocfs2_wq = create_singlethread_workqueue("ocfs2_wq");
796 	if (!ocfs2_wq) {
797 		status = -ENOMEM;
798 		goto leave;
799 	}
800 
801 	spin_lock(&ocfs2_globals_lock);
802 	osb_id = 0;
803 	spin_unlock(&ocfs2_globals_lock);
804 
805 	ocfs2_debugfs_root = debugfs_create_dir("ocfs2", NULL);
806 	if (!ocfs2_debugfs_root) {
807 		status = -EFAULT;
808 		mlog(ML_ERROR, "Unable to create ocfs2 debugfs root.\n");
809 	}
810 
811 leave:
812 	if (status < 0) {
813 		ocfs2_free_mem_caches();
814 		exit_ocfs2_uptodate_cache();
815 		exit_ocfs2_extent_maps();
816 	}
817 
818 	mlog_exit(status);
819 
820 	if (status >= 0) {
821 		return register_filesystem(&ocfs2_fs_type);
822 	} else
823 		return -1;
824 }
825 
826 static void __exit ocfs2_exit(void)
827 {
828 	mlog_entry_void();
829 
830 	if (ocfs2_wq) {
831 		flush_workqueue(ocfs2_wq);
832 		destroy_workqueue(ocfs2_wq);
833 	}
834 
835 	debugfs_remove(ocfs2_debugfs_root);
836 
837 	ocfs2_free_mem_caches();
838 
839 	unregister_filesystem(&ocfs2_fs_type);
840 
841 	exit_ocfs2_extent_maps();
842 
843 	exit_ocfs2_uptodate_cache();
844 
845 	mlog_exit_void();
846 }
847 
848 static void ocfs2_put_super(struct super_block *sb)
849 {
850 	mlog_entry("(0x%p)\n", sb);
851 
852 	ocfs2_sync_blockdev(sb);
853 	ocfs2_dismount_volume(sb, 0);
854 
855 	mlog_exit_void();
856 }
857 
858 static int ocfs2_statfs(struct super_block *sb, struct kstatfs *buf)
859 {
860 	struct ocfs2_super *osb;
861 	u32 numbits, freebits;
862 	int status;
863 	struct ocfs2_dinode *bm_lock;
864 	struct buffer_head *bh = NULL;
865 	struct inode *inode = NULL;
866 
867 	mlog_entry("(%p, %p)\n", sb, buf);
868 
869 	osb = OCFS2_SB(sb);
870 
871 	inode = ocfs2_get_system_file_inode(osb,
872 					    GLOBAL_BITMAP_SYSTEM_INODE,
873 					    OCFS2_INVALID_SLOT);
874 	if (!inode) {
875 		mlog(ML_ERROR, "failed to get bitmap inode\n");
876 		status = -EIO;
877 		goto bail;
878 	}
879 
880 	status = ocfs2_meta_lock(inode, NULL, &bh, 0);
881 	if (status < 0) {
882 		mlog_errno(status);
883 		goto bail;
884 	}
885 
886 	bm_lock = (struct ocfs2_dinode *) bh->b_data;
887 
888 	numbits = le32_to_cpu(bm_lock->id1.bitmap1.i_total);
889 	freebits = numbits - le32_to_cpu(bm_lock->id1.bitmap1.i_used);
890 
891 	buf->f_type = OCFS2_SUPER_MAGIC;
892 	buf->f_bsize = sb->s_blocksize;
893 	buf->f_namelen = OCFS2_MAX_FILENAME_LEN;
894 	buf->f_blocks = ((sector_t) numbits) *
895 			(osb->s_clustersize >> osb->sb->s_blocksize_bits);
896 	buf->f_bfree = ((sector_t) freebits) *
897 		       (osb->s_clustersize >> osb->sb->s_blocksize_bits);
898 	buf->f_bavail = buf->f_bfree;
899 	buf->f_files = numbits;
900 	buf->f_ffree = freebits;
901 
902 	brelse(bh);
903 
904 	ocfs2_meta_unlock(inode, 0);
905 	status = 0;
906 bail:
907 	if (inode)
908 		iput(inode);
909 
910 	mlog_exit(status);
911 
912 	return status;
913 }
914 
915 static void ocfs2_inode_init_once(void *data,
916 				  kmem_cache_t *cachep,
917 				  unsigned long flags)
918 {
919 	struct ocfs2_inode_info *oi = data;
920 
921 	if ((flags & (SLAB_CTOR_VERIFY|SLAB_CTOR_CONSTRUCTOR)) ==
922 	    SLAB_CTOR_CONSTRUCTOR) {
923 		oi->ip_flags = 0;
924 		oi->ip_open_count = 0;
925 		spin_lock_init(&oi->ip_lock);
926 		ocfs2_extent_map_init(&oi->vfs_inode);
927 		INIT_LIST_HEAD(&oi->ip_handle_list);
928 		INIT_LIST_HEAD(&oi->ip_io_markers);
929 		oi->ip_handle = NULL;
930 		oi->ip_created_trans = 0;
931 		oi->ip_last_trans = 0;
932 		oi->ip_dir_start_lookup = 0;
933 
934 		init_rwsem(&oi->ip_alloc_sem);
935 		init_MUTEX(&(oi->ip_io_sem));
936 
937 		oi->ip_blkno = 0ULL;
938 		oi->ip_clusters = 0;
939 
940 		ocfs2_lock_res_init_once(&oi->ip_rw_lockres);
941 		ocfs2_lock_res_init_once(&oi->ip_meta_lockres);
942 		ocfs2_lock_res_init_once(&oi->ip_data_lockres);
943 
944 		ocfs2_metadata_cache_init(&oi->vfs_inode);
945 
946 		inode_init_once(&oi->vfs_inode);
947 	}
948 }
949 
950 static int ocfs2_initialize_mem_caches(void)
951 {
952 	ocfs2_inode_cachep = kmem_cache_create("ocfs2_inode_cache",
953 					       sizeof(struct ocfs2_inode_info),
954 					       0, SLAB_HWCACHE_ALIGN|SLAB_RECLAIM_ACCOUNT,
955 					       ocfs2_inode_init_once, NULL);
956 	if (!ocfs2_inode_cachep)
957 		return -ENOMEM;
958 
959 	ocfs2_lock_cache = kmem_cache_create("ocfs2_lock",
960 					     sizeof(struct ocfs2_journal_lock),
961 					     0,
962 					     SLAB_NO_REAP|SLAB_HWCACHE_ALIGN,
963 					     NULL, NULL);
964 	if (!ocfs2_lock_cache)
965 		return -ENOMEM;
966 
967 	return 0;
968 }
969 
970 static void ocfs2_free_mem_caches(void)
971 {
972 	if (ocfs2_inode_cachep)
973 		kmem_cache_destroy(ocfs2_inode_cachep);
974 	if (ocfs2_lock_cache)
975 		kmem_cache_destroy(ocfs2_lock_cache);
976 
977 	ocfs2_inode_cachep = NULL;
978 	ocfs2_lock_cache = NULL;
979 }
980 
981 static int ocfs2_get_sector(struct super_block *sb,
982 			    struct buffer_head **bh,
983 			    int block,
984 			    int sect_size)
985 {
986 	if (!sb_set_blocksize(sb, sect_size)) {
987 		mlog(ML_ERROR, "unable to set blocksize\n");
988 		return -EIO;
989 	}
990 
991 	*bh = sb_getblk(sb, block);
992 	if (!*bh) {
993 		mlog_errno(-EIO);
994 		return -EIO;
995 	}
996 	lock_buffer(*bh);
997 	if (!buffer_dirty(*bh))
998 		clear_buffer_uptodate(*bh);
999 	unlock_buffer(*bh);
1000 	ll_rw_block(READ, 1, bh);
1001 	wait_on_buffer(*bh);
1002 	return 0;
1003 }
1004 
1005 /* ocfs2 1.0 only allows one cluster and node identity per kernel image. */
1006 static int ocfs2_fill_local_node_info(struct ocfs2_super *osb)
1007 {
1008 	int status;
1009 
1010 	/* XXX hold a ref on the node while mounte?  easy enough, if
1011 	 * desirable. */
1012 	osb->node_num = o2nm_this_node();
1013 	if (osb->node_num == O2NM_MAX_NODES) {
1014 		mlog(ML_ERROR, "could not find this host's node number\n");
1015 		status = -ENOENT;
1016 		goto bail;
1017 	}
1018 
1019 	mlog(ML_NOTICE, "I am node %d\n", osb->node_num);
1020 
1021 	status = 0;
1022 bail:
1023 	return status;
1024 }
1025 
1026 static int ocfs2_mount_volume(struct super_block *sb)
1027 {
1028 	int status = 0;
1029 	int unlock_super = 0;
1030 	struct ocfs2_super *osb = OCFS2_SB(sb);
1031 
1032 	mlog_entry_void();
1033 
1034 	if (ocfs2_is_hard_readonly(osb))
1035 		goto leave;
1036 
1037 	status = ocfs2_fill_local_node_info(osb);
1038 	if (status < 0) {
1039 		mlog_errno(status);
1040 		goto leave;
1041 	}
1042 
1043 	status = ocfs2_register_hb_callbacks(osb);
1044 	if (status < 0) {
1045 		mlog_errno(status);
1046 		goto leave;
1047 	}
1048 
1049 	status = ocfs2_dlm_init(osb);
1050 	if (status < 0) {
1051 		mlog_errno(status);
1052 		goto leave;
1053 	}
1054 
1055 	/* requires vote_thread to be running. */
1056 	status = ocfs2_register_net_handlers(osb);
1057 	if (status < 0) {
1058 		mlog_errno(status);
1059 		goto leave;
1060 	}
1061 
1062 	status = ocfs2_super_lock(osb, 1);
1063 	if (status < 0) {
1064 		mlog_errno(status);
1065 		goto leave;
1066 	}
1067 	unlock_super = 1;
1068 
1069 	/* This will load up the node map and add ourselves to it. */
1070 	status = ocfs2_find_slot(osb);
1071 	if (status < 0) {
1072 		mlog_errno(status);
1073 		goto leave;
1074 	}
1075 
1076 	ocfs2_populate_mounted_map(osb);
1077 
1078 	/* load all node-local system inodes */
1079 	status = ocfs2_init_local_system_inodes(osb);
1080 	if (status < 0) {
1081 		mlog_errno(status);
1082 		goto leave;
1083 	}
1084 
1085 	status = ocfs2_check_volume(osb);
1086 	if (status < 0) {
1087 		mlog_errno(status);
1088 		goto leave;
1089 	}
1090 
1091 	status = ocfs2_truncate_log_init(osb);
1092 	if (status < 0) {
1093 		mlog_errno(status);
1094 		goto leave;
1095 	}
1096 
1097 	/* This should be sent *after* we recovered our journal as it
1098 	 * will cause other nodes to unmark us as needing
1099 	 * recovery. However, we need to send it *before* dropping the
1100 	 * super block lock as otherwise their recovery threads might
1101 	 * try to clean us up while we're live! */
1102 	status = ocfs2_request_mount_vote(osb);
1103 	if (status < 0)
1104 		mlog_errno(status);
1105 
1106 leave:
1107 	if (unlock_super)
1108 		ocfs2_super_unlock(osb, 1);
1109 
1110 	mlog_exit(status);
1111 	return status;
1112 }
1113 
1114 /* we can't grab the goofy sem lock from inside wait_event, so we use
1115  * memory barriers to make sure that we'll see the null task before
1116  * being woken up */
1117 static int ocfs2_recovery_thread_running(struct ocfs2_super *osb)
1118 {
1119 	mb();
1120 	return osb->recovery_thread_task != NULL;
1121 }
1122 
1123 static void ocfs2_dismount_volume(struct super_block *sb, int mnt_err)
1124 {
1125 	int tmp;
1126 	struct ocfs2_super *osb = NULL;
1127 
1128 	mlog_entry("(0x%p)\n", sb);
1129 
1130 	BUG_ON(!sb);
1131 	osb = OCFS2_SB(sb);
1132 	BUG_ON(!osb);
1133 
1134 	ocfs2_shutdown_local_alloc(osb);
1135 
1136 	ocfs2_truncate_log_shutdown(osb);
1137 
1138 	/* disable any new recovery threads and wait for any currently
1139 	 * running ones to exit. Do this before setting the vol_state. */
1140 	down(&osb->recovery_lock);
1141 	osb->disable_recovery = 1;
1142 	up(&osb->recovery_lock);
1143 	wait_event(osb->recovery_event, !ocfs2_recovery_thread_running(osb));
1144 
1145 	/* At this point, we know that no more recovery threads can be
1146 	 * launched, so wait for any recovery completion work to
1147 	 * complete. */
1148 	flush_workqueue(ocfs2_wq);
1149 
1150 	ocfs2_journal_shutdown(osb);
1151 
1152 	ocfs2_sync_blockdev(sb);
1153 
1154 	/* No dlm means we've failed during mount, so skip all the
1155 	 * steps which depended on that to complete. */
1156 	if (osb->dlm) {
1157 		tmp = ocfs2_super_lock(osb, 1);
1158 		if (tmp < 0) {
1159 			mlog_errno(tmp);
1160 			return;
1161 		}
1162 
1163 		tmp = ocfs2_request_umount_vote(osb);
1164 		if (tmp < 0)
1165 			mlog_errno(tmp);
1166 
1167 		if (osb->slot_num != OCFS2_INVALID_SLOT)
1168 			ocfs2_put_slot(osb);
1169 
1170 		ocfs2_super_unlock(osb, 1);
1171 	}
1172 
1173 	ocfs2_release_system_inodes(osb);
1174 
1175 	if (osb->dlm) {
1176 		ocfs2_unregister_net_handlers(osb);
1177 
1178 		ocfs2_dlm_shutdown(osb);
1179 	}
1180 
1181 	ocfs2_clear_hb_callbacks(osb);
1182 
1183 	debugfs_remove(osb->osb_debug_root);
1184 
1185 	if (!mnt_err)
1186 		ocfs2_stop_heartbeat(osb);
1187 
1188 	atomic_set(&osb->vol_state, VOLUME_DISMOUNTED);
1189 
1190 	printk("ocfs2: Unmounting device (%u,%u) on (node %d)\n",
1191 	       MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev), osb->node_num);
1192 
1193 	ocfs2_delete_osb(osb);
1194 	kfree(osb);
1195 	sb->s_dev = 0;
1196 	sb->s_fs_info = NULL;
1197 }
1198 
1199 static int ocfs2_setup_osb_uuid(struct ocfs2_super *osb, const unsigned char *uuid,
1200 				unsigned uuid_bytes)
1201 {
1202 	int i, ret;
1203 	char *ptr;
1204 
1205 	BUG_ON(uuid_bytes != OCFS2_VOL_UUID_LEN);
1206 
1207 	osb->uuid_str = kcalloc(1, OCFS2_VOL_UUID_LEN * 2 + 1, GFP_KERNEL);
1208 	if (osb->uuid_str == NULL)
1209 		return -ENOMEM;
1210 
1211 	memcpy(osb->uuid, uuid, OCFS2_VOL_UUID_LEN);
1212 
1213 	for (i = 0, ptr = osb->uuid_str; i < OCFS2_VOL_UUID_LEN; i++) {
1214 		/* print with null */
1215 		ret = snprintf(ptr, 3, "%02X", uuid[i]);
1216 		if (ret != 2) /* drop super cleans up */
1217 			return -EINVAL;
1218 		/* then only advance past the last char */
1219 		ptr += 2;
1220 	}
1221 
1222 	return 0;
1223 }
1224 
1225 static int ocfs2_initialize_super(struct super_block *sb,
1226 				  struct buffer_head *bh,
1227 				  int sector_size)
1228 {
1229 	int status = 0;
1230 	int i;
1231 	struct ocfs2_dinode *di = NULL;
1232 	struct inode *inode = NULL;
1233 	struct buffer_head *bitmap_bh = NULL;
1234 	struct ocfs2_journal *journal;
1235 	__le32 uuid_net_key;
1236 	struct ocfs2_super *osb;
1237 
1238 	mlog_entry_void();
1239 
1240 	osb = kcalloc(1, sizeof(struct ocfs2_super), GFP_KERNEL);
1241 	if (!osb) {
1242 		status = -ENOMEM;
1243 		mlog_errno(status);
1244 		goto bail;
1245 	}
1246 
1247 	sb->s_fs_info = osb;
1248 	sb->s_op = &ocfs2_sops;
1249 	sb->s_export_op = &ocfs2_export_ops;
1250 	sb->s_flags |= MS_NOATIME;
1251 	/* this is needed to support O_LARGEFILE */
1252 	sb->s_maxbytes = ocfs2_max_file_offset(sb->s_blocksize_bits);
1253 
1254 	osb->sb = sb;
1255 	/* Save off for ocfs2_rw_direct */
1256 	osb->s_sectsize_bits = blksize_bits(sector_size);
1257 	if (!osb->s_sectsize_bits)
1258 		BUG();
1259 
1260 	osb->net_response_ids = 0;
1261 	spin_lock_init(&osb->net_response_lock);
1262 	INIT_LIST_HEAD(&osb->net_response_list);
1263 
1264 	INIT_LIST_HEAD(&osb->osb_net_handlers);
1265 	init_waitqueue_head(&osb->recovery_event);
1266 	spin_lock_init(&osb->vote_task_lock);
1267 	init_waitqueue_head(&osb->vote_event);
1268 	osb->vote_work_sequence = 0;
1269 	osb->vote_wake_sequence = 0;
1270 	INIT_LIST_HEAD(&osb->blocked_lock_list);
1271 	osb->blocked_lock_count = 0;
1272 	INIT_LIST_HEAD(&osb->vote_list);
1273 	spin_lock_init(&osb->osb_lock);
1274 
1275 	atomic_set(&osb->alloc_stats.moves, 0);
1276 	atomic_set(&osb->alloc_stats.local_data, 0);
1277 	atomic_set(&osb->alloc_stats.bitmap_data, 0);
1278 	atomic_set(&osb->alloc_stats.bg_allocs, 0);
1279 	atomic_set(&osb->alloc_stats.bg_extends, 0);
1280 
1281 	ocfs2_init_node_maps(osb);
1282 
1283 	snprintf(osb->dev_str, sizeof(osb->dev_str), "%u,%u",
1284 		 MAJOR(osb->sb->s_dev), MINOR(osb->sb->s_dev));
1285 
1286 	init_MUTEX(&osb->recovery_lock);
1287 
1288 	osb->disable_recovery = 0;
1289 	osb->recovery_thread_task = NULL;
1290 
1291 	init_waitqueue_head(&osb->checkpoint_event);
1292 	atomic_set(&osb->needs_checkpoint, 0);
1293 
1294 	osb->node_num = O2NM_INVALID_NODE_NUM;
1295 	osb->slot_num = OCFS2_INVALID_SLOT;
1296 
1297 	osb->local_alloc_state = OCFS2_LA_UNUSED;
1298 	osb->local_alloc_bh = NULL;
1299 
1300 	ocfs2_setup_hb_callbacks(osb);
1301 
1302 	init_waitqueue_head(&osb->osb_mount_event);
1303 
1304 	osb->vol_label = kmalloc(OCFS2_MAX_VOL_LABEL_LEN, GFP_KERNEL);
1305 	if (!osb->vol_label) {
1306 		mlog(ML_ERROR, "unable to alloc vol label\n");
1307 		status = -ENOMEM;
1308 		goto bail;
1309 	}
1310 
1311 	osb->uuid = kmalloc(OCFS2_VOL_UUID_LEN, GFP_KERNEL);
1312 	if (!osb->uuid) {
1313 		mlog(ML_ERROR, "unable to alloc uuid\n");
1314 		status = -ENOMEM;
1315 		goto bail;
1316 	}
1317 
1318 	di = (struct ocfs2_dinode *)bh->b_data;
1319 
1320 	osb->max_slots = le16_to_cpu(di->id2.i_super.s_max_slots);
1321 	if (osb->max_slots > OCFS2_MAX_SLOTS || osb->max_slots == 0) {
1322 		mlog(ML_ERROR, "Invalid number of node slots (%u)\n",
1323 		     osb->max_slots);
1324 		status = -EINVAL;
1325 		goto bail;
1326 	}
1327 	mlog(ML_NOTICE, "max_slots for this device: %u\n", osb->max_slots);
1328 
1329 	osb->s_feature_compat =
1330 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_compat);
1331 	osb->s_feature_ro_compat =
1332 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_ro_compat);
1333 	osb->s_feature_incompat =
1334 		le32_to_cpu(OCFS2_RAW_SB(di)->s_feature_incompat);
1335 
1336 	if ((i = OCFS2_HAS_INCOMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_INCOMPAT_SUPP))) {
1337 		mlog(ML_ERROR, "couldn't mount because of unsupported "
1338 		     "optional features (%x).\n", i);
1339 		status = -EINVAL;
1340 		goto bail;
1341 	}
1342 	if (!(osb->sb->s_flags & MS_RDONLY) &&
1343 	    (i = OCFS2_HAS_RO_COMPAT_FEATURE(osb->sb, ~OCFS2_FEATURE_RO_COMPAT_SUPP))) {
1344 		mlog(ML_ERROR, "couldn't mount RDWR because of "
1345 		     "unsupported optional features (%x).\n", i);
1346 		status = -EINVAL;
1347 		goto bail;
1348 	}
1349 
1350 	get_random_bytes(&osb->s_next_generation, sizeof(u32));
1351 
1352 	/* FIXME
1353 	 * This should be done in ocfs2_journal_init(), but unknown
1354 	 * ordering issues will cause the filesystem to crash.
1355 	 * If anyone wants to figure out what part of the code
1356 	 * refers to osb->journal before ocfs2_journal_init() is run,
1357 	 * be my guest.
1358 	 */
1359 	/* initialize our journal structure */
1360 
1361 	journal = kcalloc(1, sizeof(struct ocfs2_journal), GFP_KERNEL);
1362 	if (!journal) {
1363 		mlog(ML_ERROR, "unable to alloc journal\n");
1364 		status = -ENOMEM;
1365 		goto bail;
1366 	}
1367 	osb->journal = journal;
1368 	journal->j_osb = osb;
1369 
1370 	atomic_set(&journal->j_num_trans, 0);
1371 	init_rwsem(&journal->j_trans_barrier);
1372 	init_waitqueue_head(&journal->j_checkpointed);
1373 	spin_lock_init(&journal->j_lock);
1374 	journal->j_trans_id = (unsigned long) 1;
1375 	INIT_LIST_HEAD(&journal->j_la_cleanups);
1376 	INIT_WORK(&journal->j_recovery_work, ocfs2_complete_recovery, osb);
1377 	journal->j_state = OCFS2_JOURNAL_FREE;
1378 
1379 	/* get some pseudo constants for clustersize bits */
1380 	osb->s_clustersize_bits =
1381 		le32_to_cpu(di->id2.i_super.s_clustersize_bits);
1382 	osb->s_clustersize = 1 << osb->s_clustersize_bits;
1383 	mlog(0, "clusterbits=%d\n", osb->s_clustersize_bits);
1384 
1385 	if (osb->s_clustersize < OCFS2_MIN_CLUSTERSIZE ||
1386 	    osb->s_clustersize > OCFS2_MAX_CLUSTERSIZE) {
1387 		mlog(ML_ERROR, "Volume has invalid cluster size (%d)\n",
1388 		     osb->s_clustersize);
1389 		status = -EINVAL;
1390 		goto bail;
1391 	}
1392 
1393 	if (ocfs2_clusters_to_blocks(osb->sb, le32_to_cpu(di->i_clusters) - 1)
1394 	    > (u32)~0UL) {
1395 		mlog(ML_ERROR, "Volume might try to write to blocks beyond "
1396 		     "what jbd can address in 32 bits.\n");
1397 		status = -EINVAL;
1398 		goto bail;
1399 	}
1400 
1401 	if (ocfs2_setup_osb_uuid(osb, di->id2.i_super.s_uuid,
1402 				 sizeof(di->id2.i_super.s_uuid))) {
1403 		mlog(ML_ERROR, "Out of memory trying to setup our uuid.\n");
1404 		status = -ENOMEM;
1405 		goto bail;
1406 	}
1407 
1408 	memcpy(&uuid_net_key, &osb->uuid[i], sizeof(osb->net_key));
1409 	osb->net_key = le32_to_cpu(uuid_net_key);
1410 
1411 	strncpy(osb->vol_label, di->id2.i_super.s_label, 63);
1412 	osb->vol_label[63] = '\0';
1413 	osb->root_blkno = le64_to_cpu(di->id2.i_super.s_root_blkno);
1414 	osb->system_dir_blkno = le64_to_cpu(di->id2.i_super.s_system_dir_blkno);
1415 	osb->first_cluster_group_blkno =
1416 		le64_to_cpu(di->id2.i_super.s_first_cluster_group);
1417 	osb->fs_generation = le32_to_cpu(di->i_fs_generation);
1418 	mlog(0, "vol_label: %s\n", osb->vol_label);
1419 	mlog(0, "uuid: %s\n", osb->uuid_str);
1420 	mlog(0, "root_blkno=%"MLFu64", system_dir_blkno=%"MLFu64"\n",
1421 	     osb->root_blkno, osb->system_dir_blkno);
1422 
1423 	osb->osb_dlm_debug = ocfs2_new_dlm_debug();
1424 	if (!osb->osb_dlm_debug) {
1425 		status = -ENOMEM;
1426 		mlog_errno(status);
1427 		goto bail;
1428 	}
1429 
1430 	atomic_set(&osb->vol_state, VOLUME_INIT);
1431 
1432 	/* load root, system_dir, and all global system inodes */
1433 	status = ocfs2_init_global_system_inodes(osb);
1434 	if (status < 0) {
1435 		mlog_errno(status);
1436 		goto bail;
1437 	}
1438 
1439 	/*
1440 	 * global bitmap
1441 	 */
1442 	inode = ocfs2_get_system_file_inode(osb, GLOBAL_BITMAP_SYSTEM_INODE,
1443 					    OCFS2_INVALID_SLOT);
1444 	if (!inode) {
1445 		status = -EINVAL;
1446 		mlog_errno(status);
1447 		goto bail;
1448 	}
1449 
1450 	osb->bitmap_blkno = OCFS2_I(inode)->ip_blkno;
1451 
1452 	status = ocfs2_read_block(osb, osb->bitmap_blkno, &bitmap_bh, 0,
1453 				  inode);
1454 	iput(inode);
1455 	if (status < 0) {
1456 		mlog_errno(status);
1457 		goto bail;
1458 	}
1459 
1460 	di = (struct ocfs2_dinode *) bitmap_bh->b_data;
1461 	osb->bitmap_cpg = le16_to_cpu(di->id2.i_chain.cl_cpg);
1462 	osb->num_clusters = le32_to_cpu(di->id1.bitmap1.i_total);
1463 	brelse(bitmap_bh);
1464 	mlog(0, "cluster bitmap inode: %"MLFu64", clusters per group: %u\n",
1465 	     osb->bitmap_blkno, osb->bitmap_cpg);
1466 
1467 	status = ocfs2_init_slot_info(osb);
1468 	if (status < 0) {
1469 		mlog_errno(status);
1470 		goto bail;
1471 	}
1472 
1473 	/*  Link this osb onto the global linked list of all osb structures. */
1474 	/*  The Global Link List is mainted for the whole driver . */
1475 	spin_lock(&ocfs2_globals_lock);
1476 	osb->osb_id = osb_id;
1477 	if (osb_id < OCFS2_MAX_OSB_ID)
1478 		osb_id++;
1479 	else {
1480 		mlog(ML_ERROR, "Too many volumes mounted\n");
1481 		status = -ENOMEM;
1482 	}
1483 	spin_unlock(&ocfs2_globals_lock);
1484 
1485 bail:
1486 	mlog_exit(status);
1487 	return status;
1488 }
1489 
1490 /*
1491  * will return: -EAGAIN if it is ok to keep searching for superblocks
1492  *              -EINVAL if there is a bad superblock
1493  *              0 on success
1494  */
1495 static int ocfs2_verify_volume(struct ocfs2_dinode *di,
1496 			       struct buffer_head *bh,
1497 			       u32 blksz)
1498 {
1499 	int status = -EAGAIN;
1500 
1501 	mlog_entry_void();
1502 
1503 	if (memcmp(di->i_signature, OCFS2_SUPER_BLOCK_SIGNATURE,
1504 		   strlen(OCFS2_SUPER_BLOCK_SIGNATURE)) == 0) {
1505 		status = -EINVAL;
1506 		if ((1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits)) != blksz) {
1507 			mlog(ML_ERROR, "found superblock with incorrect block "
1508 			     "size: found %u, should be %u\n",
1509 			     1 << le32_to_cpu(di->id2.i_super.s_blocksize_bits),
1510 			       blksz);
1511 		} else if (le16_to_cpu(di->id2.i_super.s_major_rev_level) !=
1512 			   OCFS2_MAJOR_REV_LEVEL ||
1513 			   le16_to_cpu(di->id2.i_super.s_minor_rev_level) !=
1514 			   OCFS2_MINOR_REV_LEVEL) {
1515 			mlog(ML_ERROR, "found superblock with bad version: "
1516 			     "found %u.%u, should be %u.%u\n",
1517 			     le16_to_cpu(di->id2.i_super.s_major_rev_level),
1518 			     le16_to_cpu(di->id2.i_super.s_minor_rev_level),
1519 			     OCFS2_MAJOR_REV_LEVEL,
1520 			     OCFS2_MINOR_REV_LEVEL);
1521 		} else if (bh->b_blocknr != le64_to_cpu(di->i_blkno)) {
1522 			mlog(ML_ERROR, "bad block number on superblock: "
1523 			     "found %"MLFu64", should be %llu\n",
1524 			     di->i_blkno, (unsigned long long)bh->b_blocknr);
1525 		} else if (le32_to_cpu(di->id2.i_super.s_clustersize_bits) < 12 ||
1526 			    le32_to_cpu(di->id2.i_super.s_clustersize_bits) > 20) {
1527 			mlog(ML_ERROR, "bad cluster size found: %u\n",
1528 			     1 << le32_to_cpu(di->id2.i_super.s_clustersize_bits));
1529 		} else if (!le64_to_cpu(di->id2.i_super.s_root_blkno)) {
1530 			mlog(ML_ERROR, "bad root_blkno: 0\n");
1531 		} else if (!le64_to_cpu(di->id2.i_super.s_system_dir_blkno)) {
1532 			mlog(ML_ERROR, "bad system_dir_blkno: 0\n");
1533 		} else if (le16_to_cpu(di->id2.i_super.s_max_slots) > OCFS2_MAX_SLOTS) {
1534 			mlog(ML_ERROR,
1535 			     "Superblock slots found greater than file system "
1536 			     "maximum: found %u, max %u\n",
1537 			     le16_to_cpu(di->id2.i_super.s_max_slots),
1538 			     OCFS2_MAX_SLOTS);
1539 		} else {
1540 			/* found it! */
1541 			status = 0;
1542 		}
1543 	}
1544 
1545 	mlog_exit(status);
1546 	return status;
1547 }
1548 
1549 static int ocfs2_check_volume(struct ocfs2_super *osb)
1550 {
1551 	int status = 0;
1552 	int dirty;
1553 	struct ocfs2_dinode *local_alloc = NULL; /* only used if we
1554 						  * recover
1555 						  * ourselves. */
1556 
1557 	mlog_entry_void();
1558 
1559 	/* Init our journal object. */
1560 	status = ocfs2_journal_init(osb->journal, &dirty);
1561 	if (status < 0) {
1562 		mlog(ML_ERROR, "Could not initialize journal!\n");
1563 		goto finally;
1564 	}
1565 
1566 	/* If the journal was unmounted cleanly then we don't want to
1567 	 * recover anything. Otherwise, journal_load will do that
1568 	 * dirty work for us :) */
1569 	if (!dirty) {
1570 		status = ocfs2_journal_wipe(osb->journal, 0);
1571 		if (status < 0) {
1572 			mlog_errno(status);
1573 			goto finally;
1574 		}
1575 	} else {
1576 		mlog(ML_NOTICE, "File system was not unmounted cleanly, "
1577 		     "recovering volume.\n");
1578 	}
1579 
1580 	/* will play back anything left in the journal. */
1581 	ocfs2_journal_load(osb->journal);
1582 
1583 	if (dirty) {
1584 		/* recover my local alloc if we didn't unmount cleanly. */
1585 		status = ocfs2_begin_local_alloc_recovery(osb,
1586 							  osb->slot_num,
1587 							  &local_alloc);
1588 		if (status < 0) {
1589 			mlog_errno(status);
1590 			goto finally;
1591 		}
1592 		/* we complete the recovery process after we've marked
1593 		 * ourselves as mounted. */
1594 	}
1595 
1596 	mlog(0, "Journal loaded.\n");
1597 
1598 	status = ocfs2_load_local_alloc(osb);
1599 	if (status < 0) {
1600 		mlog_errno(status);
1601 		goto finally;
1602 	}
1603 
1604 	if (dirty) {
1605 		/* Recovery will be completed after we've mounted the
1606 		 * rest of the volume. */
1607 		osb->dirty = 1;
1608 		osb->local_alloc_copy = local_alloc;
1609 		local_alloc = NULL;
1610 	}
1611 
1612 	/* go through each journal, trylock it and if you get the
1613 	 * lock, and it's marked as dirty, set the bit in the recover
1614 	 * map and launch a recovery thread for it. */
1615 	status = ocfs2_mark_dead_nodes(osb);
1616 	if (status < 0)
1617 		mlog_errno(status);
1618 
1619 finally:
1620 	if (local_alloc)
1621 		kfree(local_alloc);
1622 
1623 	mlog_exit(status);
1624 	return status;
1625 }
1626 
1627 /*
1628  * The routine gets called from dismount or close whenever a dismount on
1629  * volume is requested and the osb open count becomes 1.
1630  * It will remove the osb from the global list and also free up all the
1631  * initialized resources and fileobject.
1632  */
1633 static void ocfs2_delete_osb(struct ocfs2_super *osb)
1634 {
1635 	mlog_entry_void();
1636 
1637 	/* This function assumes that the caller has the main osb resource */
1638 
1639 	if (osb->slot_info)
1640 		ocfs2_free_slot_info(osb->slot_info);
1641 
1642 	/* FIXME
1643 	 * This belongs in journal shutdown, but because we have to
1644 	 * allocate osb->journal at the start of ocfs2_initalize_osb(),
1645 	 * we free it here.
1646 	 */
1647 	kfree(osb->journal);
1648 	if (osb->local_alloc_copy)
1649 		kfree(osb->local_alloc_copy);
1650 	kfree(osb->uuid_str);
1651 	ocfs2_put_dlm_debug(osb->osb_dlm_debug);
1652 	memset(osb, 0, sizeof(struct ocfs2_super));
1653 
1654 	mlog_exit_void();
1655 }
1656 
1657 /* Put OCFS2 into a readonly state, or (if the user specifies it),
1658  * panic(). We do not support continue-on-error operation. */
1659 static void ocfs2_handle_error(struct super_block *sb)
1660 {
1661 	struct ocfs2_super *osb = OCFS2_SB(sb);
1662 
1663 	if (osb->s_mount_opt & OCFS2_MOUNT_ERRORS_PANIC)
1664 		panic("OCFS2: (device %s): panic forced after error\n",
1665 		      sb->s_id);
1666 
1667 	ocfs2_set_osb_flag(osb, OCFS2_OSB_ERROR_FS);
1668 
1669 	if (sb->s_flags & MS_RDONLY &&
1670 	    (ocfs2_is_soft_readonly(osb) ||
1671 	     ocfs2_is_hard_readonly(osb)))
1672 		return;
1673 
1674 	printk(KERN_CRIT "File system is now read-only due to the potential "
1675 	       "of on-disk corruption. Please run fsck.ocfs2 once the file "
1676 	       "system is unmounted.\n");
1677 	sb->s_flags |= MS_RDONLY;
1678 	ocfs2_set_ro_flag(osb, 0);
1679 }
1680 
1681 static char error_buf[1024];
1682 
1683 void __ocfs2_error(struct super_block *sb,
1684 		   const char *function,
1685 		   const char *fmt, ...)
1686 {
1687 	va_list args;
1688 
1689 	va_start(args, fmt);
1690 	vsprintf(error_buf, fmt, args);
1691 	va_end(args);
1692 
1693 	/* Not using mlog here because we want to show the actual
1694 	 * function the error came from. */
1695 	printk(KERN_CRIT "OCFS2: ERROR (device %s): %s: %s\n",
1696 	       sb->s_id, function, error_buf);
1697 
1698 	ocfs2_handle_error(sb);
1699 }
1700 
1701 /* Handle critical errors. This is intentionally more drastic than
1702  * ocfs2_handle_error, so we only use for things like journal errors,
1703  * etc. */
1704 void __ocfs2_abort(struct super_block* sb,
1705 		   const char *function,
1706 		   const char *fmt, ...)
1707 {
1708 	va_list args;
1709 
1710 	va_start(args, fmt);
1711 	vsprintf(error_buf, fmt, args);
1712 	va_end(args);
1713 
1714 	printk(KERN_CRIT "OCFS2: abort (device %s): %s: %s\n",
1715 	       sb->s_id, function, error_buf);
1716 
1717 	/* We don't have the cluster support yet to go straight to
1718 	 * hard readonly in here. Until then, we want to keep
1719 	 * ocfs2_abort() so that we can at least mark critical
1720 	 * errors.
1721 	 *
1722 	 * TODO: This should abort the journal and alert other nodes
1723 	 * that our slot needs recovery. */
1724 
1725 	/* Force a panic(). This stinks, but it's better than letting
1726 	 * things continue without having a proper hard readonly
1727 	 * here. */
1728 	OCFS2_SB(sb)->s_mount_opt |= OCFS2_MOUNT_ERRORS_PANIC;
1729 	ocfs2_handle_error(sb);
1730 }
1731 
1732 module_init(ocfs2_init);
1733 module_exit(ocfs2_exit);
1734