1/*
2 * Copyright (C) Sistina Software, Inc.  1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc.  All rights reserved.
4 *
5 * This copyrighted material is made available to anyone wishing to use,
6 * modify, copy, or redistribute it subject to the terms and conditions
7 * of the GNU General Public License version 2.
8 */
9
10#include <linux/sched.h>
11#include <linux/slab.h>
12#include <linux/spinlock.h>
13#include <linux/completion.h>
14#include <linux/buffer_head.h>
15#include <linux/blkdev.h>
16#include <linux/kthread.h>
17#include <linux/namei.h>
18#include <linux/mount.h>
19#include <linux/gfs2_ondisk.h>
20#include <linux/quotaops.h>
21
22#include "gfs2.h"
23#include "incore.h"
24#include "bmap.h"
25#include "glock.h"
26#include "glops.h"
27#include "inode.h"
28#include "recovery.h"
29#include "rgrp.h"
30#include "super.h"
31#include "sys.h"
32#include "util.h"
33#include "log.h"
34#include "quota.h"
35#include "dir.h"
36#include "trace_gfs2.h"
37
38#define DO 0
39#define UNDO 1
40
41static const u32 gfs2_old_fs_formats[] = {
42        0
43};
44
45static const u32 gfs2_old_multihost_formats[] = {
46        0
47};
48
49/**
50 * gfs2_tune_init - Fill a gfs2_tune structure with default values
51 * @gt: tune
52 *
53 */
54
55static void gfs2_tune_init(struct gfs2_tune *gt)
56{
57	spin_lock_init(&gt->gt_spin);
58
59	gt->gt_quota_simul_sync = 64;
60	gt->gt_quota_warn_period = 10;
61	gt->gt_quota_scale_num = 1;
62	gt->gt_quota_scale_den = 1;
63	gt->gt_new_files_jdata = 0;
64	gt->gt_max_readahead = 1 << 18;
65	gt->gt_complain_secs = 10;
66}
67
68static struct gfs2_sbd *init_sbd(struct super_block *sb)
69{
70	struct gfs2_sbd *sdp;
71
72	sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
73	if (!sdp)
74		return NULL;
75
76	sb->s_fs_info = sdp;
77	sdp->sd_vfs = sb;
78	set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
79	gfs2_tune_init(&sdp->sd_tune);
80
81	init_waitqueue_head(&sdp->sd_glock_wait);
82	atomic_set(&sdp->sd_glock_disposal, 0);
83	spin_lock_init(&sdp->sd_statfs_spin);
84
85	spin_lock_init(&sdp->sd_rindex_spin);
86	mutex_init(&sdp->sd_rindex_mutex);
87	INIT_LIST_HEAD(&sdp->sd_rindex_list);
88	INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
89
90	INIT_LIST_HEAD(&sdp->sd_jindex_list);
91	spin_lock_init(&sdp->sd_jindex_spin);
92	mutex_init(&sdp->sd_jindex_mutex);
93
94	INIT_LIST_HEAD(&sdp->sd_quota_list);
95	mutex_init(&sdp->sd_quota_mutex);
96	init_waitqueue_head(&sdp->sd_quota_wait);
97	INIT_LIST_HEAD(&sdp->sd_trunc_list);
98	spin_lock_init(&sdp->sd_trunc_lock);
99
100	spin_lock_init(&sdp->sd_log_lock);
101	atomic_set(&sdp->sd_log_pinned, 0);
102	INIT_LIST_HEAD(&sdp->sd_log_le_buf);
103	INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
104	INIT_LIST_HEAD(&sdp->sd_log_le_rg);
105	INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
106	INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
107
108	init_waitqueue_head(&sdp->sd_log_waitq);
109	init_waitqueue_head(&sdp->sd_logd_waitq);
110	INIT_LIST_HEAD(&sdp->sd_ail1_list);
111	INIT_LIST_HEAD(&sdp->sd_ail2_list);
112
113	init_rwsem(&sdp->sd_log_flush_lock);
114	atomic_set(&sdp->sd_log_in_flight, 0);
115	init_waitqueue_head(&sdp->sd_log_flush_wait);
116
117	INIT_LIST_HEAD(&sdp->sd_revoke_list);
118
119	mutex_init(&sdp->sd_freeze_lock);
120
121	return sdp;
122}
123
124
125/**
126 * gfs2_check_sb - Check superblock
127 * @sdp: the filesystem
128 * @sb: The superblock
129 * @silent: Don't print a message if the check fails
130 *
131 * Checks the version code of the FS is one that we understand how to
132 * read and that the sizes of the various on-disk structures have not
133 * changed.
134 */
135
136static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
137{
138	unsigned int x;
139
140	if (sb->sb_magic != GFS2_MAGIC ||
141	    sb->sb_type != GFS2_METATYPE_SB) {
142		if (!silent)
143			printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
144		return -EINVAL;
145	}
146
147	/*  If format numbers match exactly, we're done.  */
148
149	if (sb->sb_fs_format == GFS2_FORMAT_FS &&
150	    sb->sb_multihost_format == GFS2_FORMAT_MULTI)
151		return 0;
152
153	if (sb->sb_fs_format != GFS2_FORMAT_FS) {
154		for (x = 0; gfs2_old_fs_formats[x]; x++)
155			if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
156				break;
157
158		if (!gfs2_old_fs_formats[x]) {
159			printk(KERN_WARNING
160			       "GFS2: code version (%u, %u) is incompatible "
161			       "with ondisk format (%u, %u)\n",
162			       GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
163			       sb->sb_fs_format, sb->sb_multihost_format);
164			printk(KERN_WARNING
165			       "GFS2: I don't know how to upgrade this FS\n");
166			return -EINVAL;
167		}
168	}
169
170	if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
171		for (x = 0; gfs2_old_multihost_formats[x]; x++)
172			if (gfs2_old_multihost_formats[x] ==
173			    sb->sb_multihost_format)
174				break;
175
176		if (!gfs2_old_multihost_formats[x]) {
177			printk(KERN_WARNING
178			       "GFS2: code version (%u, %u) is incompatible "
179			       "with ondisk format (%u, %u)\n",
180			       GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
181			       sb->sb_fs_format, sb->sb_multihost_format);
182			printk(KERN_WARNING
183			       "GFS2: I don't know how to upgrade this FS\n");
184			return -EINVAL;
185		}
186	}
187
188	if (!sdp->sd_args.ar_upgrade) {
189		printk(KERN_WARNING
190		       "GFS2: code version (%u, %u) is incompatible "
191		       "with ondisk format (%u, %u)\n",
192		       GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
193		       sb->sb_fs_format, sb->sb_multihost_format);
194		printk(KERN_INFO
195		       "GFS2: Use the \"upgrade\" mount option to upgrade "
196		       "the FS\n");
197		printk(KERN_INFO "GFS2: See the manual for more details\n");
198		return -EINVAL;
199	}
200
201	return 0;
202}
203
204static void end_bio_io_page(struct bio *bio, int error)
205{
206	struct page *page = bio->bi_private;
207
208	if (!error)
209		SetPageUptodate(page);
210	else
211		printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
212	unlock_page(page);
213}
214
215static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
216{
217	const struct gfs2_sb *str = buf;
218
219	sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
220	sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
221	sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
222	sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
223	sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
224	sb->sb_bsize = be32_to_cpu(str->sb_bsize);
225	sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
226	sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
227	sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
228	sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
229	sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
230
231	memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
232	memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
233	memcpy(sb->sb_uuid, str->sb_uuid, 16);
234}
235
236/**
237 * gfs2_read_super - Read the gfs2 super block from disk
238 * @sdp: The GFS2 super block
239 * @sector: The location of the super block
240 * @error: The error code to return
241 *
242 * This uses the bio functions to read the super block from disk
243 * because we want to be 100% sure that we never read cached data.
244 * A super block is read twice only during each GFS2 mount and is
245 * never written to by the filesystem. The first time its read no
246 * locks are held, and the only details which are looked at are those
247 * relating to the locking protocol. Once locking is up and working,
248 * the sb is read again under the lock to establish the location of
249 * the master directory (contains pointers to journals etc) and the
250 * root directory.
251 *
252 * Returns: 0 on success or error
253 */
254
255static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
256{
257	struct super_block *sb = sdp->sd_vfs;
258	struct gfs2_sb *p;
259	struct page *page;
260	struct bio *bio;
261
262	page = alloc_page(GFP_NOFS);
263	if (unlikely(!page))
264		return -ENOBUFS;
265
266	ClearPageUptodate(page);
267	ClearPageDirty(page);
268	lock_page(page);
269
270	bio = bio_alloc(GFP_NOFS, 1);
271	bio->bi_sector = sector * (sb->s_blocksize >> 9);
272	bio->bi_bdev = sb->s_bdev;
273	bio_add_page(bio, page, PAGE_SIZE, 0);
274
275	bio->bi_end_io = end_bio_io_page;
276	bio->bi_private = page;
277	submit_bio(READ_SYNC | REQ_META, bio);
278	wait_on_page_locked(page);
279	bio_put(bio);
280	if (!PageUptodate(page)) {
281		__free_page(page);
282		return -EIO;
283	}
284	p = kmap(page);
285	gfs2_sb_in(&sdp->sd_sb, p);
286	kunmap(page);
287	__free_page(page);
288	return 0;
289}
290
291/**
292 * gfs2_read_sb - Read super block
293 * @sdp: The GFS2 superblock
294 * @silent: Don't print message if mount fails
295 *
296 */
297
298static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
299{
300	u32 hash_blocks, ind_blocks, leaf_blocks;
301	u32 tmp_blocks;
302	unsigned int x;
303	int error;
304
305	error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
306	if (error) {
307		if (!silent)
308			fs_err(sdp, "can't read superblock\n");
309		return error;
310	}
311
312	error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
313	if (error)
314		return error;
315
316	sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
317			       GFS2_BASIC_BLOCK_SHIFT;
318	sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
319	sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
320			  sizeof(struct gfs2_dinode)) / sizeof(u64);
321	sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
322			  sizeof(struct gfs2_meta_header)) / sizeof(u64);
323	sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
324	sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
325	sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
326	sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
327	sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
328				sizeof(struct gfs2_meta_header)) /
329			        sizeof(struct gfs2_quota_change);
330
331	/* Compute maximum reservation required to add a entry to a directory */
332
333	hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
334			     sdp->sd_jbsize);
335
336	ind_blocks = 0;
337	for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
338		tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
339		ind_blocks += tmp_blocks;
340	}
341
342	leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
343
344	sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
345
346	sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
347				sizeof(struct gfs2_dinode);
348	sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
349	for (x = 2;; x++) {
350		u64 space, d;
351		u32 m;
352
353		space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
354		d = space;
355		m = do_div(d, sdp->sd_inptrs);
356
357		if (d != sdp->sd_heightsize[x - 1] || m)
358			break;
359		sdp->sd_heightsize[x] = space;
360	}
361	sdp->sd_max_height = x;
362	sdp->sd_heightsize[x] = ~0;
363	gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
364
365	sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
366				 sizeof(struct gfs2_dinode);
367	sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
368	for (x = 2;; x++) {
369		u64 space, d;
370		u32 m;
371
372		space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
373		d = space;
374		m = do_div(d, sdp->sd_inptrs);
375
376		if (d != sdp->sd_jheightsize[x - 1] || m)
377			break;
378		sdp->sd_jheightsize[x] = space;
379	}
380	sdp->sd_max_jheight = x;
381	sdp->sd_jheightsize[x] = ~0;
382	gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
383
384	return 0;
385}
386
387static int init_names(struct gfs2_sbd *sdp, int silent)
388{
389	char *proto, *table;
390	int error = 0;
391
392	proto = sdp->sd_args.ar_lockproto;
393	table = sdp->sd_args.ar_locktable;
394
395	/*  Try to autodetect  */
396
397	if (!proto[0] || !table[0]) {
398		error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
399		if (error)
400			return error;
401
402		error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
403		if (error)
404			goto out;
405
406		if (!proto[0])
407			proto = sdp->sd_sb.sb_lockproto;
408		if (!table[0])
409			table = sdp->sd_sb.sb_locktable;
410	}
411
412	if (!table[0])
413		table = sdp->sd_vfs->s_id;
414
415	strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
416	strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
417
418	table = sdp->sd_table_name;
419	while ((table = strchr(table, '/')))
420		*table = '_';
421
422out:
423	return error;
424}
425
426static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
427			int undo)
428{
429	int error = 0;
430
431	if (undo)
432		goto fail_trans;
433
434	error = gfs2_glock_nq_num(sdp,
435				  GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
436				  LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
437				  mount_gh);
438	if (error) {
439		fs_err(sdp, "can't acquire mount glock: %d\n", error);
440		goto fail;
441	}
442
443	error = gfs2_glock_nq_num(sdp,
444				  GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
445				  LM_ST_SHARED,
446				  LM_FLAG_NOEXP | GL_EXACT,
447				  &sdp->sd_live_gh);
448	if (error) {
449		fs_err(sdp, "can't acquire live glock: %d\n", error);
450		goto fail_mount;
451	}
452
453	error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
454			       CREATE, &sdp->sd_rename_gl);
455	if (error) {
456		fs_err(sdp, "can't create rename glock: %d\n", error);
457		goto fail_live;
458	}
459
460	error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
461			       CREATE, &sdp->sd_trans_gl);
462	if (error) {
463		fs_err(sdp, "can't create transaction glock: %d\n", error);
464		goto fail_rename;
465	}
466
467	return 0;
468
469fail_trans:
470	gfs2_glock_put(sdp->sd_trans_gl);
471fail_rename:
472	gfs2_glock_put(sdp->sd_rename_gl);
473fail_live:
474	gfs2_glock_dq_uninit(&sdp->sd_live_gh);
475fail_mount:
476	gfs2_glock_dq_uninit(mount_gh);
477fail:
478	return error;
479}
480
481static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
482			    u64 no_addr, const char *name)
483{
484	struct gfs2_sbd *sdp = sb->s_fs_info;
485	struct dentry *dentry;
486	struct inode *inode;
487
488	inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
489	if (IS_ERR(inode)) {
490		fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
491		return PTR_ERR(inode);
492	}
493	dentry = d_alloc_root(inode);
494	if (!dentry) {
495		fs_err(sdp, "can't alloc %s dentry\n", name);
496		iput(inode);
497		return -ENOMEM;
498	}
499	dentry->d_op = &gfs2_dops;
500	*dptr = dentry;
501	return 0;
502}
503
504static int init_sb(struct gfs2_sbd *sdp, int silent)
505{
506	struct super_block *sb = sdp->sd_vfs;
507	struct gfs2_holder sb_gh;
508	u64 no_addr;
509	int ret;
510
511	ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
512				LM_ST_SHARED, 0, &sb_gh);
513	if (ret) {
514		fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
515		return ret;
516	}
517
518	ret = gfs2_read_sb(sdp, silent);
519	if (ret) {
520		fs_err(sdp, "can't read superblock: %d\n", ret);
521		goto out;
522	}
523
524	/* Set up the buffer cache and SB for real */
525	if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
526		ret = -EINVAL;
527		fs_err(sdp, "FS block size (%u) is too small for device "
528		       "block size (%u)\n",
529		       sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
530		goto out;
531	}
532	if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
533		ret = -EINVAL;
534		fs_err(sdp, "FS block size (%u) is too big for machine "
535		       "page size (%u)\n",
536		       sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
537		goto out;
538	}
539	sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
540
541	/* Get the root inode */
542	no_addr = sdp->sd_sb.sb_root_dir.no_addr;
543	ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
544	if (ret)
545		goto out;
546
547	/* Get the master inode */
548	no_addr = sdp->sd_sb.sb_master_dir.no_addr;
549	ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
550	if (ret) {
551		dput(sdp->sd_root_dir);
552		goto out;
553	}
554	sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
555out:
556	gfs2_glock_dq_uninit(&sb_gh);
557	return ret;
558}
559
560/**
561 * map_journal_extents - create a reusable "extent" mapping from all logical
562 * blocks to all physical blocks for the given journal.  This will save
563 * us time when writing journal blocks.  Most journals will have only one
564 * extent that maps all their logical blocks.  That's because gfs2.mkfs
565 * arranges the journal blocks sequentially to maximize performance.
566 * So the extent would map the first block for the entire file length.
567 * However, gfs2_jadd can happen while file activity is happening, so
568 * those journals may not be sequential.  Less likely is the case where
569 * the users created their own journals by mounting the metafs and
570 * laying it out.  But it's still possible.  These journals might have
571 * several extents.
572 *
573 * TODO: This should be done in bigger chunks rather than one block at a time,
574 *       but since it's only done at mount time, I'm not worried about the
575 *       time it takes.
576 */
577static int map_journal_extents(struct gfs2_sbd *sdp)
578{
579	struct gfs2_jdesc *jd = sdp->sd_jdesc;
580	unsigned int lb;
581	u64 db, prev_db; /* logical block, disk block, prev disk block */
582	struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
583	struct gfs2_journal_extent *jext = NULL;
584	struct buffer_head bh;
585	int rc = 0;
586
587	prev_db = 0;
588
589	for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
590		bh.b_state = 0;
591		bh.b_blocknr = 0;
592		bh.b_size = 1 << ip->i_inode.i_blkbits;
593		rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
594		db = bh.b_blocknr;
595		if (rc || !db) {
596			printk(KERN_INFO "GFS2 journal mapping error %d: lb="
597			       "%u db=%llu\n", rc, lb, (unsigned long long)db);
598			break;
599		}
600		if (!prev_db || db != prev_db + 1) {
601			jext = kzalloc(sizeof(struct gfs2_journal_extent),
602				       GFP_KERNEL);
603			if (!jext) {
604				printk(KERN_INFO "GFS2 error: out of memory "
605				       "mapping journal extents.\n");
606				rc = -ENOMEM;
607				break;
608			}
609			jext->dblock = db;
610			jext->lblock = lb;
611			jext->blocks = 1;
612			list_add_tail(&jext->extent_list, &jd->extent_list);
613		} else {
614			jext->blocks++;
615		}
616		prev_db = db;
617	}
618	return rc;
619}
620
621static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
622{
623	char *message = "FIRSTMOUNT=Done";
624	char *envp[] = { message, NULL };
625	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
626	ls->ls_first_done = 1;
627	kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
628}
629
630/**
631 * gfs2_jindex_hold - Grab a lock on the jindex
632 * @sdp: The GFS2 superblock
633 * @ji_gh: the holder for the jindex glock
634 *
635 * Returns: errno
636 */
637
638static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
639{
640	struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
641	struct qstr name;
642	char buf[20];
643	struct gfs2_jdesc *jd;
644	int error;
645
646	name.name = buf;
647
648	mutex_lock(&sdp->sd_jindex_mutex);
649
650	for (;;) {
651		error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
652		if (error)
653			break;
654
655		name.len = sprintf(buf, "journal%u", sdp->sd_journals);
656		name.hash = gfs2_disk_hash(name.name, name.len);
657
658		error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
659		if (error == -ENOENT) {
660			error = 0;
661			break;
662		}
663
664		gfs2_glock_dq_uninit(ji_gh);
665
666		if (error)
667			break;
668
669		error = -ENOMEM;
670		jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
671		if (!jd)
672			break;
673
674		INIT_LIST_HEAD(&jd->extent_list);
675		INIT_WORK(&jd->jd_work, gfs2_recover_func);
676		jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
677		if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
678			if (!jd->jd_inode)
679				error = -ENOENT;
680			else
681				error = PTR_ERR(jd->jd_inode);
682			kfree(jd);
683			break;
684		}
685
686		spin_lock(&sdp->sd_jindex_spin);
687		jd->jd_jid = sdp->sd_journals++;
688		list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
689		spin_unlock(&sdp->sd_jindex_spin);
690	}
691
692	mutex_unlock(&sdp->sd_jindex_mutex);
693
694	return error;
695}
696
697static int init_journal(struct gfs2_sbd *sdp, int undo)
698{
699	struct inode *master = sdp->sd_master_dir->d_inode;
700	struct gfs2_holder ji_gh;
701	struct gfs2_inode *ip;
702	int jindex = 1;
703	int error = 0;
704
705	if (undo) {
706		jindex = 0;
707		goto fail_jinode_gh;
708	}
709
710	sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
711	if (IS_ERR(sdp->sd_jindex)) {
712		fs_err(sdp, "can't lookup journal index: %d\n", error);
713		return PTR_ERR(sdp->sd_jindex);
714	}
715	ip = GFS2_I(sdp->sd_jindex);
716
717	/* Load in the journal index special file */
718
719	error = gfs2_jindex_hold(sdp, &ji_gh);
720	if (error) {
721		fs_err(sdp, "can't read journal index: %d\n", error);
722		goto fail;
723	}
724
725	error = -EUSERS;
726	if (!gfs2_jindex_size(sdp)) {
727		fs_err(sdp, "no journals!\n");
728		goto fail_jindex;
729	}
730
731	if (sdp->sd_args.ar_spectator) {
732		sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
733		atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
734		atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
735		atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
736	} else {
737		if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
738			fs_err(sdp, "can't mount journal #%u\n",
739			       sdp->sd_lockstruct.ls_jid);
740			fs_err(sdp, "there are only %u journals (0 - %u)\n",
741			       gfs2_jindex_size(sdp),
742			       gfs2_jindex_size(sdp) - 1);
743			goto fail_jindex;
744		}
745		sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
746
747		error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
748					  &gfs2_journal_glops,
749					  LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
750					  &sdp->sd_journal_gh);
751		if (error) {
752			fs_err(sdp, "can't acquire journal glock: %d\n", error);
753			goto fail_jindex;
754		}
755
756		ip = GFS2_I(sdp->sd_jdesc->jd_inode);
757		error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
758					   LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
759					   &sdp->sd_jinode_gh);
760		if (error) {
761			fs_err(sdp, "can't acquire journal inode glock: %d\n",
762			       error);
763			goto fail_journal_gh;
764		}
765
766		error = gfs2_jdesc_check(sdp->sd_jdesc);
767		if (error) {
768			fs_err(sdp, "my journal (%u) is bad: %d\n",
769			       sdp->sd_jdesc->jd_jid, error);
770			goto fail_jinode_gh;
771		}
772		atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
773		atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
774		atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
775
776		/* Map the extents for this journal's blocks */
777		map_journal_extents(sdp);
778	}
779	trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
780
781	if (sdp->sd_lockstruct.ls_first) {
782		unsigned int x;
783		for (x = 0; x < sdp->sd_journals; x++) {
784			error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
785						     true);
786			if (error) {
787				fs_err(sdp, "error recovering journal %u: %d\n",
788				       x, error);
789				goto fail_jinode_gh;
790			}
791		}
792
793		gfs2_others_may_mount(sdp);
794	} else if (!sdp->sd_args.ar_spectator) {
795		error = gfs2_recover_journal(sdp->sd_jdesc, true);
796		if (error) {
797			fs_err(sdp, "error recovering my journal: %d\n", error);
798			goto fail_jinode_gh;
799		}
800	}
801
802	set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
803	gfs2_glock_dq_uninit(&ji_gh);
804	jindex = 0;
805
806	return 0;
807
808fail_jinode_gh:
809	if (!sdp->sd_args.ar_spectator)
810		gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
811fail_journal_gh:
812	if (!sdp->sd_args.ar_spectator)
813		gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
814fail_jindex:
815	gfs2_jindex_free(sdp);
816	if (jindex)
817		gfs2_glock_dq_uninit(&ji_gh);
818fail:
819	iput(sdp->sd_jindex);
820	return error;
821}
822
823
824static int init_inodes(struct gfs2_sbd *sdp, int undo)
825{
826	int error = 0;
827	struct gfs2_inode *ip;
828	struct inode *master = sdp->sd_master_dir->d_inode;
829
830	if (undo)
831		goto fail_qinode;
832
833	error = init_journal(sdp, undo);
834	if (error)
835		goto fail;
836
837	/* Read in the master statfs inode */
838	sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
839	if (IS_ERR(sdp->sd_statfs_inode)) {
840		error = PTR_ERR(sdp->sd_statfs_inode);
841		fs_err(sdp, "can't read in statfs inode: %d\n", error);
842		goto fail_journal;
843	}
844
845	/* Read in the resource index inode */
846	sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
847	if (IS_ERR(sdp->sd_rindex)) {
848		error = PTR_ERR(sdp->sd_rindex);
849		fs_err(sdp, "can't get resource index inode: %d\n", error);
850		goto fail_statfs;
851	}
852	ip = GFS2_I(sdp->sd_rindex);
853	sdp->sd_rindex_uptodate = 0;
854
855	/* Read in the quota inode */
856	sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
857	if (IS_ERR(sdp->sd_quota_inode)) {
858		error = PTR_ERR(sdp->sd_quota_inode);
859		fs_err(sdp, "can't get quota file inode: %d\n", error);
860		goto fail_rindex;
861	}
862	return 0;
863
864fail_qinode:
865	iput(sdp->sd_quota_inode);
866fail_rindex:
867	gfs2_clear_rgrpd(sdp);
868	iput(sdp->sd_rindex);
869fail_statfs:
870	iput(sdp->sd_statfs_inode);
871fail_journal:
872	init_journal(sdp, UNDO);
873fail:
874	return error;
875}
876
877static int init_per_node(struct gfs2_sbd *sdp, int undo)
878{
879	struct inode *pn = NULL;
880	char buf[30];
881	int error = 0;
882	struct gfs2_inode *ip;
883	struct inode *master = sdp->sd_master_dir->d_inode;
884
885	if (sdp->sd_args.ar_spectator)
886		return 0;
887
888	if (undo)
889		goto fail_qc_gh;
890
891	pn = gfs2_lookup_simple(master, "per_node");
892	if (IS_ERR(pn)) {
893		error = PTR_ERR(pn);
894		fs_err(sdp, "can't find per_node directory: %d\n", error);
895		return error;
896	}
897
898	sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
899	sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
900	if (IS_ERR(sdp->sd_sc_inode)) {
901		error = PTR_ERR(sdp->sd_sc_inode);
902		fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
903		goto fail;
904	}
905
906	sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
907	sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
908	if (IS_ERR(sdp->sd_qc_inode)) {
909		error = PTR_ERR(sdp->sd_qc_inode);
910		fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
911		goto fail_ut_i;
912	}
913
914	iput(pn);
915	pn = NULL;
916
917	ip = GFS2_I(sdp->sd_sc_inode);
918	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
919				   &sdp->sd_sc_gh);
920	if (error) {
921		fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
922		goto fail_qc_i;
923	}
924
925	ip = GFS2_I(sdp->sd_qc_inode);
926	error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
927				   &sdp->sd_qc_gh);
928	if (error) {
929		fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
930		goto fail_ut_gh;
931	}
932
933	return 0;
934
935fail_qc_gh:
936	gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
937fail_ut_gh:
938	gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
939fail_qc_i:
940	iput(sdp->sd_qc_inode);
941fail_ut_i:
942	iput(sdp->sd_sc_inode);
943fail:
944	if (pn)
945		iput(pn);
946	return error;
947}
948
949static int init_threads(struct gfs2_sbd *sdp, int undo)
950{
951	struct task_struct *p;
952	int error = 0;
953
954	if (undo)
955		goto fail_quotad;
956
957	p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
958	error = IS_ERR(p);
959	if (error) {
960		fs_err(sdp, "can't start logd thread: %d\n", error);
961		return error;
962	}
963	sdp->sd_logd_process = p;
964
965	p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
966	error = IS_ERR(p);
967	if (error) {
968		fs_err(sdp, "can't start quotad thread: %d\n", error);
969		goto fail;
970	}
971	sdp->sd_quotad_process = p;
972
973	return 0;
974
975
976fail_quotad:
977	kthread_stop(sdp->sd_quotad_process);
978fail:
979	kthread_stop(sdp->sd_logd_process);
980	return error;
981}
982
983static const match_table_t nolock_tokens = {
984	{ Opt_jid, "jid=%d\n", },
985	{ Opt_err, NULL },
986};
987
988static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
989{
990	struct gfs2_sbd *sdp = gl->gl_sbd;
991	kmem_cache_free(cachep, gl);
992	if (atomic_dec_and_test(&sdp->sd_glock_disposal))
993		wake_up(&sdp->sd_glock_wait);
994}
995
996static const struct lm_lockops nolock_ops = {
997	.lm_proto_name = "lock_nolock",
998	.lm_put_lock = nolock_put_lock,
999	.lm_tokens = &nolock_tokens,
1000};
1001
1002/**
1003 * gfs2_lm_mount - mount a locking protocol
1004 * @sdp: the filesystem
1005 * @args: mount arguments
1006 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1007 *
1008 * Returns: errno
1009 */
1010
1011static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1012{
1013	const struct lm_lockops *lm;
1014	struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1015	struct gfs2_args *args = &sdp->sd_args;
1016	const char *proto = sdp->sd_proto_name;
1017	const char *table = sdp->sd_table_name;
1018	const char *fsname;
1019	char *o, *options;
1020	int ret;
1021
1022	if (!strcmp("lock_nolock", proto)) {
1023		lm = &nolock_ops;
1024		sdp->sd_args.ar_localflocks = 1;
1025		sdp->sd_args.ar_localcaching = 1;
1026#ifdef CONFIG_GFS2_FS_LOCKING_DLM
1027	} else if (!strcmp("lock_dlm", proto)) {
1028		lm = &gfs2_dlm_ops;
1029#endif
1030	} else {
1031		printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1032		return -ENOENT;
1033	}
1034
1035	fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1036
1037	ls->ls_ops = lm;
1038	ls->ls_first = 1;
1039
1040	for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1041		substring_t tmp[MAX_OPT_ARGS];
1042		int token, option;
1043
1044		if (!o || !*o)
1045			continue;
1046
1047		token = match_token(o, *lm->lm_tokens, tmp);
1048		switch (token) {
1049		case Opt_jid:
1050			ret = match_int(&tmp[0], &option);
1051			if (ret || option < 0)
1052				goto hostdata_error;
1053			if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
1054				ls->ls_jid = option;
1055			break;
1056		case Opt_id:
1057			/* Obsolete, but left for backward compat purposes */
1058			break;
1059		case Opt_first:
1060			ret = match_int(&tmp[0], &option);
1061			if (ret || (option != 0 && option != 1))
1062				goto hostdata_error;
1063			ls->ls_first = option;
1064			break;
1065		case Opt_nodir:
1066			ret = match_int(&tmp[0], &option);
1067			if (ret || (option != 0 && option != 1))
1068				goto hostdata_error;
1069			ls->ls_nodir = option;
1070			break;
1071		case Opt_err:
1072		default:
1073hostdata_error:
1074			fs_info(sdp, "unknown hostdata (%s)\n", o);
1075			return -EINVAL;
1076		}
1077	}
1078
1079	if (sdp->sd_args.ar_spectator)
1080		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1081	else
1082		snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1083			 sdp->sd_lockstruct.ls_jid);
1084
1085	fsname = strchr(table, ':');
1086	if (fsname)
1087		fsname++;
1088	if (lm->lm_mount == NULL) {
1089		fs_info(sdp, "Now mounting FS...\n");
1090		return 0;
1091	}
1092	ret = lm->lm_mount(sdp, fsname);
1093	if (ret == 0)
1094		fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1095	return ret;
1096}
1097
1098void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1099{
1100	const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1101	if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1102	    lm->lm_unmount)
1103		lm->lm_unmount(sdp);
1104}
1105
1106static int gfs2_journalid_wait(void *word)
1107{
1108	if (signal_pending(current))
1109		return -EINTR;
1110	schedule();
1111	return 0;
1112}
1113
1114static int wait_on_journal(struct gfs2_sbd *sdp)
1115{
1116	if (sdp->sd_args.ar_spectator)
1117		return 0;
1118	if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1119		return 0;
1120
1121	return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1122}
1123
1124void gfs2_online_uevent(struct gfs2_sbd *sdp)
1125{
1126	struct super_block *sb = sdp->sd_vfs;
1127	char ro[20];
1128	char spectator[20];
1129	char *envp[] = { ro, spectator, NULL };
1130	sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1131	sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1132	kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1133}
1134
1135/**
1136 * fill_super - Read in superblock
1137 * @sb: The VFS superblock
1138 * @data: Mount options
1139 * @silent: Don't complain if it's not a GFS2 filesystem
1140 *
1141 * Returns: errno
1142 */
1143
1144static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
1145{
1146	struct gfs2_sbd *sdp;
1147	struct gfs2_holder mount_gh;
1148	int error;
1149
1150	sdp = init_sbd(sb);
1151	if (!sdp) {
1152		printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1153		return -ENOMEM;
1154	}
1155	sdp->sd_args = *args;
1156
1157	if (sdp->sd_args.ar_spectator) {
1158                sb->s_flags |= MS_RDONLY;
1159		set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1160	}
1161	if (sdp->sd_args.ar_posix_acl)
1162		sb->s_flags |= MS_POSIXACL;
1163	if (sdp->sd_args.ar_nobarrier)
1164		set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1165
1166	sb->s_magic = GFS2_MAGIC;
1167	sb->s_op = &gfs2_super_ops;
1168	sb->s_export_op = &gfs2_export_ops;
1169	sb->s_xattr = gfs2_xattr_handlers;
1170	sb->s_qcop = &gfs2_quotactl_ops;
1171	sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1172	sb->s_time_gran = 1;
1173	sb->s_maxbytes = MAX_LFS_FILESIZE;
1174
1175	/* Set up the buffer cache and fill in some fake block size values
1176	   to allow us to read-in the on-disk superblock. */
1177	sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1178	sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1179	sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1180                               GFS2_BASIC_BLOCK_SHIFT;
1181	sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1182
1183	sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
1184	sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1185	if (sdp->sd_args.ar_statfs_quantum) {
1186		sdp->sd_tune.gt_statfs_slow = 0;
1187		sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
1188	}
1189	else {
1190		sdp->sd_tune.gt_statfs_slow = 1;
1191		sdp->sd_tune.gt_statfs_quantum = 30;
1192	}
1193
1194	error = init_names(sdp, silent);
1195	if (error)
1196		goto fail;
1197
1198	gfs2_create_debugfs_file(sdp);
1199
1200	error = gfs2_sys_fs_add(sdp);
1201	if (error)
1202		goto fail;
1203
1204	error = gfs2_lm_mount(sdp, silent);
1205	if (error)
1206		goto fail_sys;
1207
1208	error = init_locking(sdp, &mount_gh, DO);
1209	if (error)
1210		goto fail_lm;
1211
1212	error = init_sb(sdp, silent);
1213	if (error)
1214		goto fail_locking;
1215
1216	error = wait_on_journal(sdp);
1217	if (error)
1218		goto fail_sb;
1219
1220	error = init_inodes(sdp, DO);
1221	if (error)
1222		goto fail_sb;
1223
1224	error = init_per_node(sdp, DO);
1225	if (error)
1226		goto fail_inodes;
1227
1228	error = gfs2_statfs_init(sdp);
1229	if (error) {
1230		fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1231		goto fail_per_node;
1232	}
1233
1234	error = init_threads(sdp, DO);
1235	if (error)
1236		goto fail_per_node;
1237
1238	if (!(sb->s_flags & MS_RDONLY)) {
1239		error = gfs2_make_fs_rw(sdp);
1240		if (error) {
1241			fs_err(sdp, "can't make FS RW: %d\n", error);
1242			goto fail_threads;
1243		}
1244	}
1245
1246	gfs2_glock_dq_uninit(&mount_gh);
1247	gfs2_online_uevent(sdp);
1248	return 0;
1249
1250fail_threads:
1251	init_threads(sdp, UNDO);
1252fail_per_node:
1253	init_per_node(sdp, UNDO);
1254fail_inodes:
1255	init_inodes(sdp, UNDO);
1256fail_sb:
1257	if (sdp->sd_root_dir)
1258		dput(sdp->sd_root_dir);
1259	if (sdp->sd_master_dir)
1260		dput(sdp->sd_master_dir);
1261	if (sb->s_root)
1262		dput(sb->s_root);
1263	sb->s_root = NULL;
1264fail_locking:
1265	init_locking(sdp, &mount_gh, UNDO);
1266fail_lm:
1267	invalidate_inodes(sb);
1268	gfs2_gl_hash_clear(sdp);
1269	gfs2_lm_unmount(sdp);
1270fail_sys:
1271	gfs2_sys_fs_del(sdp);
1272fail:
1273	gfs2_delete_debugfs_file(sdp);
1274	kfree(sdp);
1275	sb->s_fs_info = NULL;
1276	return error;
1277}
1278
1279static int set_gfs2_super(struct super_block *s, void *data)
1280{
1281	s->s_bdev = data;
1282	s->s_dev = s->s_bdev->bd_dev;
1283
1284	/*
1285	 * We set the bdi here to the queue backing, file systems can
1286	 * overwrite this in ->fill_super()
1287	 */
1288	s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1289	return 0;
1290}
1291
1292static int test_gfs2_super(struct super_block *s, void *ptr)
1293{
1294	struct block_device *bdev = ptr;
1295	return (bdev == s->s_bdev);
1296}
1297
1298/**
1299 * gfs2_get_sb - Get the GFS2 superblock
1300 * @fs_type: The GFS2 filesystem type
1301 * @flags: Mount flags
1302 * @dev_name: The name of the device
1303 * @data: The mount arguments
1304 * @mnt: The vfsmnt for this mount
1305 *
1306 * Q. Why not use get_sb_bdev() ?
1307 * A. We need to select one of two root directories to mount, independent
1308 *    of whether this is the initial, or subsequent, mount of this sb
1309 *
1310 * Returns: 0 or -ve on error
1311 */
1312
1313static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1314		       const char *dev_name, void *data, struct vfsmount *mnt)
1315{
1316	struct block_device *bdev;
1317	struct super_block *s;
1318	fmode_t mode = FMODE_READ;
1319	int error;
1320	struct gfs2_args args;
1321	struct gfs2_sbd *sdp;
1322
1323	if (!(flags & MS_RDONLY))
1324		mode |= FMODE_WRITE;
1325
1326	bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1327	if (IS_ERR(bdev))
1328		return PTR_ERR(bdev);
1329
1330	/*
1331	 * once the super is inserted into the list by sget, s_umount
1332	 * will protect the lockfs code from trying to start a snapshot
1333	 * while we are mounting
1334	 */
1335	mutex_lock(&bdev->bd_fsfreeze_mutex);
1336	if (bdev->bd_fsfreeze_count > 0) {
1337		mutex_unlock(&bdev->bd_fsfreeze_mutex);
1338		error = -EBUSY;
1339		goto error_bdev;
1340	}
1341	s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev);
1342	mutex_unlock(&bdev->bd_fsfreeze_mutex);
1343	error = PTR_ERR(s);
1344	if (IS_ERR(s))
1345		goto error_bdev;
1346
1347	memset(&args, 0, sizeof(args));
1348	args.ar_quota = GFS2_QUOTA_DEFAULT;
1349	args.ar_data = GFS2_DATA_DEFAULT;
1350	args.ar_commit = 30;
1351	args.ar_statfs_quantum = 30;
1352	args.ar_quota_quantum = 60;
1353	args.ar_errors = GFS2_ERRORS_DEFAULT;
1354
1355	error = gfs2_mount_args(&args, data);
1356	if (error) {
1357		printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1358		if (s->s_root)
1359			goto error_super;
1360		deactivate_locked_super(s);
1361		return error;
1362	}
1363
1364	if (s->s_root) {
1365		error = -EBUSY;
1366		if ((flags ^ s->s_flags) & MS_RDONLY)
1367			goto error_super;
1368		close_bdev_exclusive(bdev, mode);
1369	} else {
1370		char b[BDEVNAME_SIZE];
1371
1372		s->s_flags = flags;
1373		s->s_mode = mode;
1374		strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1375		sb_set_blocksize(s, block_size(bdev));
1376		error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
1377		if (error) {
1378			deactivate_locked_super(s);
1379			return error;
1380		}
1381		s->s_flags |= MS_ACTIVE;
1382		bdev->bd_super = s;
1383	}
1384
1385	sdp = s->s_fs_info;
1386	mnt->mnt_sb = s;
1387	if (args.ar_meta)
1388		mnt->mnt_root = dget(sdp->sd_master_dir);
1389	else
1390		mnt->mnt_root = dget(sdp->sd_root_dir);
1391	return 0;
1392
1393error_super:
1394	deactivate_locked_super(s);
1395error_bdev:
1396	close_bdev_exclusive(bdev, mode);
1397	return error;
1398}
1399
1400static int set_meta_super(struct super_block *s, void *ptr)
1401{
1402	return -EINVAL;
1403}
1404
1405static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1406			    const char *dev_name, void *data, struct vfsmount *mnt)
1407{
1408	struct super_block *s;
1409	struct gfs2_sbd *sdp;
1410	struct path path;
1411	int error;
1412
1413	error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
1414	if (error) {
1415		printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1416		       dev_name, error);
1417		return error;
1418	}
1419	s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super,
1420		 path.dentry->d_inode->i_sb->s_bdev);
1421	path_put(&path);
1422	if (IS_ERR(s)) {
1423		printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1424		return PTR_ERR(s);
1425	}
1426	if ((flags ^ s->s_flags) & MS_RDONLY) {
1427		deactivate_locked_super(s);
1428		return -EBUSY;
1429	}
1430	sdp = s->s_fs_info;
1431	mnt->mnt_sb = s;
1432	mnt->mnt_root = dget(sdp->sd_master_dir);
1433	return 0;
1434}
1435
1436static void gfs2_kill_sb(struct super_block *sb)
1437{
1438	struct gfs2_sbd *sdp = sb->s_fs_info;
1439
1440	if (sdp == NULL) {
1441		kill_block_super(sb);
1442		return;
1443	}
1444
1445	gfs2_meta_syncfs(sdp);
1446	dput(sdp->sd_root_dir);
1447	dput(sdp->sd_master_dir);
1448	sdp->sd_root_dir = NULL;
1449	sdp->sd_master_dir = NULL;
1450	shrink_dcache_sb(sb);
1451	kill_block_super(sb);
1452	gfs2_delete_debugfs_file(sdp);
1453	kfree(sdp);
1454}
1455
1456struct file_system_type gfs2_fs_type = {
1457	.name = "gfs2",
1458	.fs_flags = FS_REQUIRES_DEV,
1459	.get_sb = gfs2_get_sb,
1460	.kill_sb = gfs2_kill_sb,
1461	.owner = THIS_MODULE,
1462};
1463
1464struct file_system_type gfs2meta_fs_type = {
1465	.name = "gfs2meta",
1466	.fs_flags = FS_REQUIRES_DEV,
1467	.get_sb = gfs2_get_sb_meta,
1468	.owner = THIS_MODULE,
1469};
1470