2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2008 Red Hat, Inc. All rights reserved.
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.
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/lm_interface.h>
40 static const u32 gfs2_old_fs_formats[] = {
44 static const u32 gfs2_old_multihost_formats[] = {
49 * gfs2_tune_init - Fill a gfs2_tune structure with default values
54 static void gfs2_tune_init(struct gfs2_tune *gt)
56 spin_lock_init(>->gt_spin);
58 gt->gt_demote_secs = 300;
59 gt->gt_incore_log_blocks = 1024;
60 gt->gt_log_flush_secs = 60;
61 gt->gt_recoverd_secs = 60;
63 gt->gt_quota_simul_sync = 64;
64 gt->gt_quota_warn_period = 10;
65 gt->gt_quota_scale_num = 1;
66 gt->gt_quota_scale_den = 1;
67 gt->gt_quota_cache_secs = 300;
68 gt->gt_quota_quantum = 60;
69 gt->gt_new_files_jdata = 0;
70 gt->gt_max_readahead = 1 << 18;
71 gt->gt_stall_secs = 600;
72 gt->gt_complain_secs = 10;
73 gt->gt_statfs_quantum = 30;
74 gt->gt_statfs_slow = 0;
77 static struct gfs2_sbd *init_sbd(struct super_block *sb)
81 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
88 gfs2_tune_init(&sdp->sd_tune);
90 INIT_LIST_HEAD(&sdp->sd_reclaim_list);
91 spin_lock_init(&sdp->sd_reclaim_lock);
92 init_waitqueue_head(&sdp->sd_reclaim_wq);
94 mutex_init(&sdp->sd_inum_mutex);
95 spin_lock_init(&sdp->sd_statfs_spin);
97 spin_lock_init(&sdp->sd_rindex_spin);
98 mutex_init(&sdp->sd_rindex_mutex);
99 INIT_LIST_HEAD(&sdp->sd_rindex_list);
100 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
102 INIT_LIST_HEAD(&sdp->sd_jindex_list);
103 spin_lock_init(&sdp->sd_jindex_spin);
104 mutex_init(&sdp->sd_jindex_mutex);
106 INIT_LIST_HEAD(&sdp->sd_quota_list);
107 spin_lock_init(&sdp->sd_quota_spin);
108 mutex_init(&sdp->sd_quota_mutex);
109 init_waitqueue_head(&sdp->sd_quota_wait);
111 spin_lock_init(&sdp->sd_log_lock);
113 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
114 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
115 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
116 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
117 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
119 mutex_init(&sdp->sd_log_reserve_mutex);
120 INIT_LIST_HEAD(&sdp->sd_ail1_list);
121 INIT_LIST_HEAD(&sdp->sd_ail2_list);
123 init_rwsem(&sdp->sd_log_flush_lock);
124 atomic_set(&sdp->sd_log_in_flight, 0);
125 init_waitqueue_head(&sdp->sd_log_flush_wait);
127 INIT_LIST_HEAD(&sdp->sd_revoke_list);
129 mutex_init(&sdp->sd_freeze_lock);
136 * gfs2_check_sb - Check superblock
137 * @sdp: the filesystem
138 * @sb: The superblock
139 * @silent: Don't print a message if the check fails
141 * Checks the version code of the FS is one that we understand how to
142 * read and that the sizes of the various on-disk structures have not
146 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
150 if (sb->sb_magic != GFS2_MAGIC ||
151 sb->sb_type != GFS2_METATYPE_SB) {
153 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
157 /* If format numbers match exactly, we're done. */
159 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
160 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
163 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
164 for (x = 0; gfs2_old_fs_formats[x]; x++)
165 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
168 if (!gfs2_old_fs_formats[x]) {
170 "GFS2: code version (%u, %u) is incompatible "
171 "with ondisk format (%u, %u)\n",
172 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
173 sb->sb_fs_format, sb->sb_multihost_format);
175 "GFS2: I don't know how to upgrade this FS\n");
180 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
181 for (x = 0; gfs2_old_multihost_formats[x]; x++)
182 if (gfs2_old_multihost_formats[x] ==
183 sb->sb_multihost_format)
186 if (!gfs2_old_multihost_formats[x]) {
188 "GFS2: code version (%u, %u) is incompatible "
189 "with ondisk format (%u, %u)\n",
190 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
191 sb->sb_fs_format, sb->sb_multihost_format);
193 "GFS2: I don't know how to upgrade this FS\n");
198 if (!sdp->sd_args.ar_upgrade) {
200 "GFS2: code version (%u, %u) is incompatible "
201 "with ondisk format (%u, %u)\n",
202 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
203 sb->sb_fs_format, sb->sb_multihost_format);
205 "GFS2: Use the \"upgrade\" mount option to upgrade "
207 printk(KERN_INFO "GFS2: See the manual for more details\n");
214 static void end_bio_io_page(struct bio *bio, int error)
216 struct page *page = bio->bi_private;
219 SetPageUptodate(page);
221 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
225 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
227 const struct gfs2_sb *str = buf;
229 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
230 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
231 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
232 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
233 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
234 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
235 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
236 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
237 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
238 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
239 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
241 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
242 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
246 * gfs2_read_super - Read the gfs2 super block from disk
247 * @sdp: The GFS2 super block
248 * @sector: The location of the super block
249 * @error: The error code to return
251 * This uses the bio functions to read the super block from disk
252 * because we want to be 100% sure that we never read cached data.
253 * A super block is read twice only during each GFS2 mount and is
254 * never written to by the filesystem. The first time its read no
255 * locks are held, and the only details which are looked at are those
256 * relating to the locking protocol. Once locking is up and working,
257 * the sb is read again under the lock to establish the location of
258 * the master directory (contains pointers to journals etc) and the
261 * Returns: 0 on success or error
264 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
266 struct super_block *sb = sdp->sd_vfs;
271 page = alloc_page(GFP_NOFS);
275 ClearPageUptodate(page);
276 ClearPageDirty(page);
279 bio = bio_alloc(GFP_NOFS, 1);
280 if (unlikely(!bio)) {
285 bio->bi_sector = sector * (sb->s_blocksize >> 9);
286 bio->bi_bdev = sb->s_bdev;
287 bio_add_page(bio, page, PAGE_SIZE, 0);
289 bio->bi_end_io = end_bio_io_page;
290 bio->bi_private = page;
291 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
292 wait_on_page_locked(page);
294 if (!PageUptodate(page)) {
299 gfs2_sb_in(&sdp->sd_sb, p);
305 * gfs2_read_sb - Read super block
306 * @sdp: The GFS2 superblock
307 * @gl: the glock for the superblock (assumed to be held)
308 * @silent: Don't print message if mount fails
312 static int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
314 u32 hash_blocks, ind_blocks, leaf_blocks;
319 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
322 fs_err(sdp, "can't read superblock\n");
326 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
330 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
331 GFS2_BASIC_BLOCK_SHIFT;
332 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
333 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
334 sizeof(struct gfs2_dinode)) / sizeof(u64);
335 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
336 sizeof(struct gfs2_meta_header)) / sizeof(u64);
337 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
338 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
339 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
340 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
341 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
342 sizeof(struct gfs2_meta_header)) /
343 sizeof(struct gfs2_quota_change);
345 /* Compute maximum reservation required to add a entry to a directory */
347 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
351 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
352 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
353 ind_blocks += tmp_blocks;
356 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
358 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
360 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
361 sizeof(struct gfs2_dinode);
362 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
367 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
369 m = do_div(d, sdp->sd_inptrs);
371 if (d != sdp->sd_heightsize[x - 1] || m)
373 sdp->sd_heightsize[x] = space;
375 sdp->sd_max_height = x;
376 sdp->sd_heightsize[x] = ~0;
377 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
379 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
380 sizeof(struct gfs2_dinode);
381 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
386 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
388 m = do_div(d, sdp->sd_inptrs);
390 if (d != sdp->sd_jheightsize[x - 1] || m)
392 sdp->sd_jheightsize[x] = space;
394 sdp->sd_max_jheight = x;
395 sdp->sd_jheightsize[x] = ~0;
396 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
401 static int init_names(struct gfs2_sbd *sdp, int silent)
406 proto = sdp->sd_args.ar_lockproto;
407 table = sdp->sd_args.ar_locktable;
409 /* Try to autodetect */
411 if (!proto[0] || !table[0]) {
412 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
416 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
421 proto = sdp->sd_sb.sb_lockproto;
423 table = sdp->sd_sb.sb_locktable;
427 table = sdp->sd_vfs->s_id;
429 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
430 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
432 table = sdp->sd_table_name;
433 while ((table = strchr(table, '/')))
440 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
443 struct task_struct *p;
449 for (sdp->sd_glockd_num = 0;
450 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
451 sdp->sd_glockd_num++) {
452 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
455 fs_err(sdp, "can't start glockd thread: %d\n", error);
458 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
461 error = gfs2_glock_nq_num(sdp,
462 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
463 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
466 fs_err(sdp, "can't acquire mount glock: %d\n", error);
470 error = gfs2_glock_nq_num(sdp,
471 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
473 LM_FLAG_NOEXP | GL_EXACT,
476 fs_err(sdp, "can't acquire live glock: %d\n", error);
480 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
481 CREATE, &sdp->sd_rename_gl);
483 fs_err(sdp, "can't create rename glock: %d\n", error);
487 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
488 CREATE, &sdp->sd_trans_gl);
490 fs_err(sdp, "can't create transaction glock: %d\n", error);
493 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
498 gfs2_glock_put(sdp->sd_trans_gl);
500 gfs2_glock_put(sdp->sd_rename_gl);
502 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
504 gfs2_glock_dq_uninit(mount_gh);
506 while (sdp->sd_glockd_num--)
507 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
512 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
513 u64 no_addr, const char *name)
515 struct gfs2_sbd *sdp = sb->s_fs_info;
516 struct dentry *dentry;
519 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
521 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
522 return PTR_ERR(inode);
524 dentry = d_alloc_root(inode);
526 fs_err(sdp, "can't alloc %s dentry\n", name);
530 dentry->d_op = &gfs2_dops;
535 static int init_sb(struct gfs2_sbd *sdp, int silent)
537 struct super_block *sb = sdp->sd_vfs;
538 struct gfs2_holder sb_gh;
542 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
543 LM_ST_SHARED, 0, &sb_gh);
545 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
549 ret = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
551 fs_err(sdp, "can't read superblock: %d\n", ret);
555 /* Set up the buffer cache and SB for real */
556 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
558 fs_err(sdp, "FS block size (%u) is too small for device "
560 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
563 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
565 fs_err(sdp, "FS block size (%u) is too big for machine "
567 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
570 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
572 /* Get the root inode */
573 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
574 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
578 /* Get the master inode */
579 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
580 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
582 dput(sdp->sd_root_dir);
585 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
587 gfs2_glock_dq_uninit(&sb_gh);
592 * map_journal_extents - create a reusable "extent" mapping from all logical
593 * blocks to all physical blocks for the given journal. This will save
594 * us time when writing journal blocks. Most journals will have only one
595 * extent that maps all their logical blocks. That's because gfs2.mkfs
596 * arranges the journal blocks sequentially to maximize performance.
597 * So the extent would map the first block for the entire file length.
598 * However, gfs2_jadd can happen while file activity is happening, so
599 * those journals may not be sequential. Less likely is the case where
600 * the users created their own journals by mounting the metafs and
601 * laying it out. But it's still possible. These journals might have
604 * TODO: This should be done in bigger chunks rather than one block at a time,
605 * but since it's only done at mount time, I'm not worried about the
608 static int map_journal_extents(struct gfs2_sbd *sdp)
610 struct gfs2_jdesc *jd = sdp->sd_jdesc;
612 u64 db, prev_db; /* logical block, disk block, prev disk block */
613 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
614 struct gfs2_journal_extent *jext = NULL;
615 struct buffer_head bh;
620 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
623 bh.b_size = 1 << ip->i_inode.i_blkbits;
624 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
627 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
628 "%u db=%llu\n", rc, lb, (unsigned long long)db);
631 if (!prev_db || db != prev_db + 1) {
632 jext = kzalloc(sizeof(struct gfs2_journal_extent),
635 printk(KERN_INFO "GFS2 error: out of memory "
636 "mapping journal extents.\n");
643 list_add_tail(&jext->extent_list, &jd->extent_list);
652 static void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp)
654 if (!sdp->sd_lockstruct.ls_ops->lm_others_may_mount)
656 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
657 sdp->sd_lockstruct.ls_ops->lm_others_may_mount(
658 sdp->sd_lockstruct.ls_lockspace);
661 static int init_journal(struct gfs2_sbd *sdp, int undo)
663 struct inode *master = sdp->sd_master_dir->d_inode;
664 struct gfs2_holder ji_gh;
665 struct task_struct *p;
666 struct gfs2_inode *ip;
675 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
676 if (IS_ERR(sdp->sd_jindex)) {
677 fs_err(sdp, "can't lookup journal index: %d\n", error);
678 return PTR_ERR(sdp->sd_jindex);
680 ip = GFS2_I(sdp->sd_jindex);
681 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
683 /* Load in the journal index special file */
685 error = gfs2_jindex_hold(sdp, &ji_gh);
687 fs_err(sdp, "can't read journal index: %d\n", error);
692 if (!gfs2_jindex_size(sdp)) {
693 fs_err(sdp, "no journals!\n");
697 if (sdp->sd_args.ar_spectator) {
698 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
699 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
701 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
702 fs_err(sdp, "can't mount journal #%u\n",
703 sdp->sd_lockstruct.ls_jid);
704 fs_err(sdp, "there are only %u journals (0 - %u)\n",
705 gfs2_jindex_size(sdp),
706 gfs2_jindex_size(sdp) - 1);
709 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
711 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
713 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
714 &sdp->sd_journal_gh);
716 fs_err(sdp, "can't acquire journal glock: %d\n", error);
720 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
721 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
722 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
725 fs_err(sdp, "can't acquire journal inode glock: %d\n",
727 goto fail_journal_gh;
730 error = gfs2_jdesc_check(sdp->sd_jdesc);
732 fs_err(sdp, "my journal (%u) is bad: %d\n",
733 sdp->sd_jdesc->jd_jid, error);
736 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
738 /* Map the extents for this journal's blocks */
739 map_journal_extents(sdp);
742 if (sdp->sd_lockstruct.ls_first) {
744 for (x = 0; x < sdp->sd_journals; x++) {
745 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
747 fs_err(sdp, "error recovering journal %u: %d\n",
753 gfs2_lm_others_may_mount(sdp);
754 } else if (!sdp->sd_args.ar_spectator) {
755 error = gfs2_recover_journal(sdp->sd_jdesc);
757 fs_err(sdp, "error recovering my journal: %d\n", error);
762 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
763 gfs2_glock_dq_uninit(&ji_gh);
766 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
769 fs_err(sdp, "can't start recoverd thread: %d\n", error);
772 sdp->sd_recoverd_process = p;
777 kthread_stop(sdp->sd_recoverd_process);
779 if (!sdp->sd_args.ar_spectator)
780 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
782 if (!sdp->sd_args.ar_spectator)
783 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
785 gfs2_jindex_free(sdp);
787 gfs2_glock_dq_uninit(&ji_gh);
789 iput(sdp->sd_jindex);
794 static int init_inodes(struct gfs2_sbd *sdp, int undo)
797 struct gfs2_inode *ip;
798 struct inode *master = sdp->sd_master_dir->d_inode;
803 error = init_journal(sdp, undo);
807 /* Read in the master inode number inode */
808 sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum");
809 if (IS_ERR(sdp->sd_inum_inode)) {
810 error = PTR_ERR(sdp->sd_inum_inode);
811 fs_err(sdp, "can't read in inum inode: %d\n", error);
816 /* Read in the master statfs inode */
817 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
818 if (IS_ERR(sdp->sd_statfs_inode)) {
819 error = PTR_ERR(sdp->sd_statfs_inode);
820 fs_err(sdp, "can't read in statfs inode: %d\n", error);
824 /* Read in the resource index inode */
825 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
826 if (IS_ERR(sdp->sd_rindex)) {
827 error = PTR_ERR(sdp->sd_rindex);
828 fs_err(sdp, "can't get resource index inode: %d\n", error);
831 ip = GFS2_I(sdp->sd_rindex);
832 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
833 sdp->sd_rindex_uptodate = 0;
835 /* Read in the quota inode */
836 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
837 if (IS_ERR(sdp->sd_quota_inode)) {
838 error = PTR_ERR(sdp->sd_quota_inode);
839 fs_err(sdp, "can't get quota file inode: %d\n", error);
845 iput(sdp->sd_quota_inode);
847 gfs2_clear_rgrpd(sdp);
848 iput(sdp->sd_rindex);
850 iput(sdp->sd_statfs_inode);
852 iput(sdp->sd_inum_inode);
854 init_journal(sdp, UNDO);
859 static int init_per_node(struct gfs2_sbd *sdp, int undo)
861 struct inode *pn = NULL;
864 struct gfs2_inode *ip;
865 struct inode *master = sdp->sd_master_dir->d_inode;
867 if (sdp->sd_args.ar_spectator)
873 pn = gfs2_lookup_simple(master, "per_node");
876 fs_err(sdp, "can't find per_node directory: %d\n", error);
880 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
881 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
882 if (IS_ERR(sdp->sd_ir_inode)) {
883 error = PTR_ERR(sdp->sd_ir_inode);
884 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
888 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
889 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
890 if (IS_ERR(sdp->sd_sc_inode)) {
891 error = PTR_ERR(sdp->sd_sc_inode);
892 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
896 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
897 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
898 if (IS_ERR(sdp->sd_qc_inode)) {
899 error = PTR_ERR(sdp->sd_qc_inode);
900 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
907 ip = GFS2_I(sdp->sd_ir_inode);
908 error = gfs2_glock_nq_init(ip->i_gl,
912 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
916 ip = GFS2_I(sdp->sd_sc_inode);
917 error = gfs2_glock_nq_init(ip->i_gl,
921 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
925 ip = GFS2_I(sdp->sd_qc_inode);
926 error = gfs2_glock_nq_init(ip->i_gl,
930 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
937 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
939 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
941 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
943 iput(sdp->sd_qc_inode);
945 iput(sdp->sd_sc_inode);
947 iput(sdp->sd_ir_inode);
954 static int init_threads(struct gfs2_sbd *sdp, int undo)
956 struct task_struct *p;
962 sdp->sd_log_flush_time = jiffies;
963 sdp->sd_jindex_refresh_time = jiffies;
965 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
968 fs_err(sdp, "can't start logd thread: %d\n", error);
971 sdp->sd_logd_process = p;
973 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
976 fs_err(sdp, "can't start quotad thread: %d\n", error);
979 sdp->sd_quotad_process = p;
985 kthread_stop(sdp->sd_quotad_process);
987 kthread_stop(sdp->sd_logd_process);
992 * gfs2_lm_mount - mount a locking protocol
993 * @sdp: the filesystem
994 * @args: mount arguements
995 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1000 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1002 char *proto = sdp->sd_proto_name;
1003 char *table = sdp->sd_table_name;
1004 int flags = LM_MFLAG_CONV_NODROP;
1007 if (sdp->sd_args.ar_spectator)
1008 flags |= LM_MFLAG_SPECTATOR;
1010 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1012 error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata,
1014 GFS2_MIN_LVB_SIZE, flags,
1015 &sdp->sd_lockstruct, &sdp->sd_kobj);
1017 fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n",
1018 proto, table, sdp->sd_args.ar_hostdata);
1022 if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) ||
1023 gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >=
1024 GFS2_MIN_LVB_SIZE)) {
1025 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1029 if (sdp->sd_args.ar_spectator)
1030 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1032 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1033 sdp->sd_lockstruct.ls_jid);
1035 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1037 if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) &&
1038 !sdp->sd_args.ar_ignore_local_fs) {
1039 sdp->sd_args.ar_localflocks = 1;
1040 sdp->sd_args.ar_localcaching = 1;
1047 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1049 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
1050 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1054 * fill_super - Read in superblock
1055 * @sb: The VFS superblock
1056 * @data: Mount options
1057 * @silent: Don't complain if it's not a GFS2 filesystem
1062 static int fill_super(struct super_block *sb, void *data, int silent)
1064 struct gfs2_sbd *sdp;
1065 struct gfs2_holder mount_gh;
1070 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1074 error = gfs2_mount_args(sdp, (char *)data, 0);
1076 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1080 sb->s_magic = GFS2_MAGIC;
1081 sb->s_op = &gfs2_super_ops;
1082 sb->s_export_op = &gfs2_export_ops;
1083 sb->s_time_gran = 1;
1084 sb->s_maxbytes = MAX_LFS_FILESIZE;
1086 /* Set up the buffer cache and fill in some fake block size values
1087 to allow us to read-in the on-disk superblock. */
1088 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1089 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1090 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1091 GFS2_BASIC_BLOCK_SHIFT;
1092 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1094 error = init_names(sdp, silent);
1098 gfs2_create_debugfs_file(sdp);
1100 error = gfs2_sys_fs_add(sdp);
1104 error = gfs2_lm_mount(sdp, silent);
1108 error = init_locking(sdp, &mount_gh, DO);
1112 error = init_sb(sdp, silent);
1116 error = init_inodes(sdp, DO);
1120 error = init_per_node(sdp, DO);
1124 error = gfs2_statfs_init(sdp);
1126 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1130 error = init_threads(sdp, DO);
1134 if (!(sb->s_flags & MS_RDONLY)) {
1135 error = gfs2_make_fs_rw(sdp);
1137 fs_err(sdp, "can't make FS RW: %d\n", error);
1142 gfs2_glock_dq_uninit(&mount_gh);
1147 init_threads(sdp, UNDO);
1149 init_per_node(sdp, UNDO);
1151 init_inodes(sdp, UNDO);
1153 if (sdp->sd_root_dir)
1154 dput(sdp->sd_root_dir);
1155 if (sdp->sd_master_dir)
1156 dput(sdp->sd_master_dir);
1159 init_locking(sdp, &mount_gh, UNDO);
1161 gfs2_gl_hash_clear(sdp);
1162 gfs2_lm_unmount(sdp);
1163 while (invalidate_inodes(sb))
1166 gfs2_sys_fs_del(sdp);
1168 gfs2_delete_debugfs_file(sdp);
1170 sb->s_fs_info = NULL;
1174 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1175 const char *dev_name, void *data, struct vfsmount *mnt)
1177 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1180 static struct super_block *get_gfs2_sb(const char *dev_name)
1182 struct super_block *sb;
1183 struct nameidata nd;
1186 error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd);
1188 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1192 sb = nd.path.dentry->d_inode->i_sb;
1193 if (sb && (sb->s_type == &gfs2_fs_type))
1194 atomic_inc(&sb->s_active);
1201 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1202 const char *dev_name, void *data, struct vfsmount *mnt)
1204 struct super_block *sb = NULL;
1205 struct gfs2_sbd *sdp;
1207 sb = get_gfs2_sb(dev_name);
1209 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1212 sdp = sb->s_fs_info;
1214 mnt->mnt_root = dget(sdp->sd_master_dir);
1218 static void gfs2_kill_sb(struct super_block *sb)
1220 struct gfs2_sbd *sdp = sb->s_fs_info;
1222 gfs2_meta_syncfs(sdp);
1223 dput(sdp->sd_root_dir);
1224 dput(sdp->sd_master_dir);
1225 sdp->sd_root_dir = NULL;
1226 sdp->sd_master_dir = NULL;
1228 shrink_dcache_sb(sb);
1229 kill_block_super(sb);
1231 gfs2_delete_debugfs_file(sdp);
1234 struct file_system_type gfs2_fs_type = {
1236 .fs_flags = FS_REQUIRES_DEV,
1237 .get_sb = gfs2_get_sb,
1238 .kill_sb = gfs2_kill_sb,
1239 .owner = THIS_MODULE,
1242 struct file_system_type gfs2meta_fs_type = {
1244 .fs_flags = FS_REQUIRES_DEV,
1245 .get_sb = gfs2_get_sb_meta,
1246 .owner = THIS_MODULE,