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>
39 static const u32 gfs2_old_fs_formats[] = {
43 static const u32 gfs2_old_multihost_formats[] = {
48 * gfs2_tune_init - Fill a gfs2_tune structure with default values
53 static void gfs2_tune_init(struct gfs2_tune *gt)
55 spin_lock_init(>->gt_spin);
57 gt->gt_incore_log_blocks = 1024;
58 gt->gt_log_flush_secs = 60;
59 gt->gt_recoverd_secs = 60;
61 gt->gt_quota_simul_sync = 64;
62 gt->gt_quota_warn_period = 10;
63 gt->gt_quota_scale_num = 1;
64 gt->gt_quota_scale_den = 1;
65 gt->gt_quota_quantum = 60;
66 gt->gt_new_files_jdata = 0;
67 gt->gt_max_readahead = 1 << 18;
68 gt->gt_stall_secs = 600;
69 gt->gt_complain_secs = 10;
70 gt->gt_statfs_quantum = 30;
71 gt->gt_statfs_slow = 0;
74 static struct gfs2_sbd *init_sbd(struct super_block *sb)
78 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
85 gfs2_tune_init(&sdp->sd_tune);
87 mutex_init(&sdp->sd_inum_mutex);
88 spin_lock_init(&sdp->sd_statfs_spin);
90 spin_lock_init(&sdp->sd_rindex_spin);
91 mutex_init(&sdp->sd_rindex_mutex);
92 INIT_LIST_HEAD(&sdp->sd_rindex_list);
93 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
95 INIT_LIST_HEAD(&sdp->sd_jindex_list);
96 spin_lock_init(&sdp->sd_jindex_spin);
97 mutex_init(&sdp->sd_jindex_mutex);
99 INIT_LIST_HEAD(&sdp->sd_quota_list);
100 mutex_init(&sdp->sd_quota_mutex);
101 init_waitqueue_head(&sdp->sd_quota_wait);
102 INIT_LIST_HEAD(&sdp->sd_trunc_list);
103 spin_lock_init(&sdp->sd_trunc_lock);
105 spin_lock_init(&sdp->sd_log_lock);
107 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
108 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
109 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
110 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
111 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
113 mutex_init(&sdp->sd_log_reserve_mutex);
114 INIT_LIST_HEAD(&sdp->sd_ail1_list);
115 INIT_LIST_HEAD(&sdp->sd_ail2_list);
117 init_rwsem(&sdp->sd_log_flush_lock);
118 atomic_set(&sdp->sd_log_in_flight, 0);
119 init_waitqueue_head(&sdp->sd_log_flush_wait);
121 INIT_LIST_HEAD(&sdp->sd_revoke_list);
123 mutex_init(&sdp->sd_freeze_lock);
130 * gfs2_check_sb - Check superblock
131 * @sdp: the filesystem
132 * @sb: The superblock
133 * @silent: Don't print a message if the check fails
135 * Checks the version code of the FS is one that we understand how to
136 * read and that the sizes of the various on-disk structures have not
140 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
144 if (sb->sb_magic != GFS2_MAGIC ||
145 sb->sb_type != GFS2_METATYPE_SB) {
147 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
151 /* If format numbers match exactly, we're done. */
153 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
154 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
157 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
158 for (x = 0; gfs2_old_fs_formats[x]; x++)
159 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
162 if (!gfs2_old_fs_formats[x]) {
164 "GFS2: code version (%u, %u) is incompatible "
165 "with ondisk format (%u, %u)\n",
166 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
167 sb->sb_fs_format, sb->sb_multihost_format);
169 "GFS2: I don't know how to upgrade this FS\n");
174 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
175 for (x = 0; gfs2_old_multihost_formats[x]; x++)
176 if (gfs2_old_multihost_formats[x] ==
177 sb->sb_multihost_format)
180 if (!gfs2_old_multihost_formats[x]) {
182 "GFS2: code version (%u, %u) is incompatible "
183 "with ondisk format (%u, %u)\n",
184 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
185 sb->sb_fs_format, sb->sb_multihost_format);
187 "GFS2: I don't know how to upgrade this FS\n");
192 if (!sdp->sd_args.ar_upgrade) {
194 "GFS2: code version (%u, %u) is incompatible "
195 "with ondisk format (%u, %u)\n",
196 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
197 sb->sb_fs_format, sb->sb_multihost_format);
199 "GFS2: Use the \"upgrade\" mount option to upgrade "
201 printk(KERN_INFO "GFS2: See the manual for more details\n");
208 static void end_bio_io_page(struct bio *bio, int error)
210 struct page *page = bio->bi_private;
213 SetPageUptodate(page);
215 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
219 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
221 const struct gfs2_sb *str = buf;
223 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
224 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
225 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
226 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
227 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
228 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
229 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
230 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
231 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
232 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
233 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
235 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
236 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
240 * gfs2_read_super - Read the gfs2 super block from disk
241 * @sdp: The GFS2 super block
242 * @sector: The location of the super block
243 * @error: The error code to return
245 * This uses the bio functions to read the super block from disk
246 * because we want to be 100% sure that we never read cached data.
247 * A super block is read twice only during each GFS2 mount and is
248 * never written to by the filesystem. The first time its read no
249 * locks are held, and the only details which are looked at are those
250 * relating to the locking protocol. Once locking is up and working,
251 * the sb is read again under the lock to establish the location of
252 * the master directory (contains pointers to journals etc) and the
255 * Returns: 0 on success or error
258 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
260 struct super_block *sb = sdp->sd_vfs;
265 page = alloc_page(GFP_NOFS);
269 ClearPageUptodate(page);
270 ClearPageDirty(page);
273 bio = bio_alloc(GFP_NOFS, 1);
274 if (unlikely(!bio)) {
279 bio->bi_sector = sector * (sb->s_blocksize >> 9);
280 bio->bi_bdev = sb->s_bdev;
281 bio_add_page(bio, page, PAGE_SIZE, 0);
283 bio->bi_end_io = end_bio_io_page;
284 bio->bi_private = page;
285 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
286 wait_on_page_locked(page);
288 if (!PageUptodate(page)) {
293 gfs2_sb_in(&sdp->sd_sb, p);
299 * gfs2_read_sb - Read super block
300 * @sdp: The GFS2 superblock
301 * @gl: the glock for the superblock (assumed to be held)
302 * @silent: Don't print message if mount fails
306 static int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
308 u32 hash_blocks, ind_blocks, leaf_blocks;
313 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
316 fs_err(sdp, "can't read superblock\n");
320 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
324 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
325 GFS2_BASIC_BLOCK_SHIFT;
326 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
327 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
328 sizeof(struct gfs2_dinode)) / sizeof(u64);
329 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
330 sizeof(struct gfs2_meta_header)) / sizeof(u64);
331 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
332 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
333 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
334 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
335 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
336 sizeof(struct gfs2_meta_header)) /
337 sizeof(struct gfs2_quota_change);
339 /* Compute maximum reservation required to add a entry to a directory */
341 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
345 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
346 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
347 ind_blocks += tmp_blocks;
350 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
352 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
354 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
355 sizeof(struct gfs2_dinode);
356 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
361 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
363 m = do_div(d, sdp->sd_inptrs);
365 if (d != sdp->sd_heightsize[x - 1] || m)
367 sdp->sd_heightsize[x] = space;
369 sdp->sd_max_height = x;
370 sdp->sd_heightsize[x] = ~0;
371 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
373 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
374 sizeof(struct gfs2_dinode);
375 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
380 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
382 m = do_div(d, sdp->sd_inptrs);
384 if (d != sdp->sd_jheightsize[x - 1] || m)
386 sdp->sd_jheightsize[x] = space;
388 sdp->sd_max_jheight = x;
389 sdp->sd_jheightsize[x] = ~0;
390 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
395 static int init_names(struct gfs2_sbd *sdp, int silent)
400 proto = sdp->sd_args.ar_lockproto;
401 table = sdp->sd_args.ar_locktable;
403 /* Try to autodetect */
405 if (!proto[0] || !table[0]) {
406 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
410 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
415 proto = sdp->sd_sb.sb_lockproto;
417 table = sdp->sd_sb.sb_locktable;
421 table = sdp->sd_vfs->s_id;
423 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
424 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
426 table = sdp->sd_table_name;
427 while ((table = strchr(table, '/')))
434 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
442 error = gfs2_glock_nq_num(sdp,
443 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
444 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
447 fs_err(sdp, "can't acquire mount glock: %d\n", error);
451 error = gfs2_glock_nq_num(sdp,
452 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
454 LM_FLAG_NOEXP | GL_EXACT,
457 fs_err(sdp, "can't acquire live glock: %d\n", error);
461 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
462 CREATE, &sdp->sd_rename_gl);
464 fs_err(sdp, "can't create rename glock: %d\n", error);
468 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
469 CREATE, &sdp->sd_trans_gl);
471 fs_err(sdp, "can't create transaction glock: %d\n", error);
478 gfs2_glock_put(sdp->sd_trans_gl);
480 gfs2_glock_put(sdp->sd_rename_gl);
482 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
484 gfs2_glock_dq_uninit(mount_gh);
489 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
490 u64 no_addr, const char *name)
492 struct gfs2_sbd *sdp = sb->s_fs_info;
493 struct dentry *dentry;
496 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
498 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
499 return PTR_ERR(inode);
501 dentry = d_alloc_root(inode);
503 fs_err(sdp, "can't alloc %s dentry\n", name);
507 dentry->d_op = &gfs2_dops;
512 static int init_sb(struct gfs2_sbd *sdp, int silent)
514 struct super_block *sb = sdp->sd_vfs;
515 struct gfs2_holder sb_gh;
519 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
520 LM_ST_SHARED, 0, &sb_gh);
522 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
526 ret = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
528 fs_err(sdp, "can't read superblock: %d\n", ret);
532 /* Set up the buffer cache and SB for real */
533 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
535 fs_err(sdp, "FS block size (%u) is too small for device "
537 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
540 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
542 fs_err(sdp, "FS block size (%u) is too big for machine "
544 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
547 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
549 /* Get the root inode */
550 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
551 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
555 /* Get the master inode */
556 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
557 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
559 dput(sdp->sd_root_dir);
562 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
564 gfs2_glock_dq_uninit(&sb_gh);
569 * map_journal_extents - create a reusable "extent" mapping from all logical
570 * blocks to all physical blocks for the given journal. This will save
571 * us time when writing journal blocks. Most journals will have only one
572 * extent that maps all their logical blocks. That's because gfs2.mkfs
573 * arranges the journal blocks sequentially to maximize performance.
574 * So the extent would map the first block for the entire file length.
575 * However, gfs2_jadd can happen while file activity is happening, so
576 * those journals may not be sequential. Less likely is the case where
577 * the users created their own journals by mounting the metafs and
578 * laying it out. But it's still possible. These journals might have
581 * TODO: This should be done in bigger chunks rather than one block at a time,
582 * but since it's only done at mount time, I'm not worried about the
585 static int map_journal_extents(struct gfs2_sbd *sdp)
587 struct gfs2_jdesc *jd = sdp->sd_jdesc;
589 u64 db, prev_db; /* logical block, disk block, prev disk block */
590 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
591 struct gfs2_journal_extent *jext = NULL;
592 struct buffer_head bh;
597 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
600 bh.b_size = 1 << ip->i_inode.i_blkbits;
601 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
604 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
605 "%u db=%llu\n", rc, lb, (unsigned long long)db);
608 if (!prev_db || db != prev_db + 1) {
609 jext = kzalloc(sizeof(struct gfs2_journal_extent),
612 printk(KERN_INFO "GFS2 error: out of memory "
613 "mapping journal extents.\n");
620 list_add_tail(&jext->extent_list, &jd->extent_list);
629 static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
631 char *message = "FIRSTMOUNT=Done";
632 char *envp[] = { message, NULL };
633 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
634 ls->ls_first_done = 1;
635 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
639 * gfs2_jindex_hold - Grab a lock on the jindex
640 * @sdp: The GFS2 superblock
641 * @ji_gh: the holder for the jindex glock
646 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
648 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
651 struct gfs2_jdesc *jd;
656 mutex_lock(&sdp->sd_jindex_mutex);
659 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
663 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
664 name.hash = gfs2_disk_hash(name.name, name.len);
666 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
667 if (error == -ENOENT) {
672 gfs2_glock_dq_uninit(ji_gh);
678 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
682 INIT_LIST_HEAD(&jd->extent_list);
683 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
684 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
688 error = PTR_ERR(jd->jd_inode);
693 spin_lock(&sdp->sd_jindex_spin);
694 jd->jd_jid = sdp->sd_journals++;
695 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
696 spin_unlock(&sdp->sd_jindex_spin);
699 mutex_unlock(&sdp->sd_jindex_mutex);
704 static int init_journal(struct gfs2_sbd *sdp, int undo)
706 struct inode *master = sdp->sd_master_dir->d_inode;
707 struct gfs2_holder ji_gh;
708 struct task_struct *p;
709 struct gfs2_inode *ip;
718 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
719 if (IS_ERR(sdp->sd_jindex)) {
720 fs_err(sdp, "can't lookup journal index: %d\n", error);
721 return PTR_ERR(sdp->sd_jindex);
723 ip = GFS2_I(sdp->sd_jindex);
725 /* Load in the journal index special file */
727 error = gfs2_jindex_hold(sdp, &ji_gh);
729 fs_err(sdp, "can't read journal index: %d\n", error);
734 if (!gfs2_jindex_size(sdp)) {
735 fs_err(sdp, "no journals!\n");
739 if (sdp->sd_args.ar_spectator) {
740 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
741 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
743 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
744 fs_err(sdp, "can't mount journal #%u\n",
745 sdp->sd_lockstruct.ls_jid);
746 fs_err(sdp, "there are only %u journals (0 - %u)\n",
747 gfs2_jindex_size(sdp),
748 gfs2_jindex_size(sdp) - 1);
751 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
753 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
755 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
756 &sdp->sd_journal_gh);
758 fs_err(sdp, "can't acquire journal glock: %d\n", error);
762 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
763 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
764 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
767 fs_err(sdp, "can't acquire journal inode glock: %d\n",
769 goto fail_journal_gh;
772 error = gfs2_jdesc_check(sdp->sd_jdesc);
774 fs_err(sdp, "my journal (%u) is bad: %d\n",
775 sdp->sd_jdesc->jd_jid, error);
778 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
780 /* Map the extents for this journal's blocks */
781 map_journal_extents(sdp);
784 if (sdp->sd_lockstruct.ls_first) {
786 for (x = 0; x < sdp->sd_journals; x++) {
787 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
789 fs_err(sdp, "error recovering journal %u: %d\n",
795 gfs2_others_may_mount(sdp);
796 } else if (!sdp->sd_args.ar_spectator) {
797 error = gfs2_recover_journal(sdp->sd_jdesc);
799 fs_err(sdp, "error recovering my journal: %d\n", error);
804 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
805 gfs2_glock_dq_uninit(&ji_gh);
808 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
811 fs_err(sdp, "can't start recoverd thread: %d\n", error);
814 sdp->sd_recoverd_process = p;
819 kthread_stop(sdp->sd_recoverd_process);
821 if (!sdp->sd_args.ar_spectator)
822 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
824 if (!sdp->sd_args.ar_spectator)
825 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
827 gfs2_jindex_free(sdp);
829 gfs2_glock_dq_uninit(&ji_gh);
831 iput(sdp->sd_jindex);
836 static int init_inodes(struct gfs2_sbd *sdp, int undo)
839 struct gfs2_inode *ip;
840 struct inode *master = sdp->sd_master_dir->d_inode;
845 error = init_journal(sdp, undo);
849 /* Read in the master inode number inode */
850 sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum");
851 if (IS_ERR(sdp->sd_inum_inode)) {
852 error = PTR_ERR(sdp->sd_inum_inode);
853 fs_err(sdp, "can't read in inum inode: %d\n", error);
858 /* Read in the master statfs inode */
859 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
860 if (IS_ERR(sdp->sd_statfs_inode)) {
861 error = PTR_ERR(sdp->sd_statfs_inode);
862 fs_err(sdp, "can't read in statfs inode: %d\n", error);
866 /* Read in the resource index inode */
867 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
868 if (IS_ERR(sdp->sd_rindex)) {
869 error = PTR_ERR(sdp->sd_rindex);
870 fs_err(sdp, "can't get resource index inode: %d\n", error);
873 ip = GFS2_I(sdp->sd_rindex);
874 sdp->sd_rindex_uptodate = 0;
876 /* Read in the quota inode */
877 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
878 if (IS_ERR(sdp->sd_quota_inode)) {
879 error = PTR_ERR(sdp->sd_quota_inode);
880 fs_err(sdp, "can't get quota file inode: %d\n", error);
886 iput(sdp->sd_quota_inode);
888 gfs2_clear_rgrpd(sdp);
889 iput(sdp->sd_rindex);
891 iput(sdp->sd_statfs_inode);
893 iput(sdp->sd_inum_inode);
895 init_journal(sdp, UNDO);
900 static int init_per_node(struct gfs2_sbd *sdp, int undo)
902 struct inode *pn = NULL;
905 struct gfs2_inode *ip;
906 struct inode *master = sdp->sd_master_dir->d_inode;
908 if (sdp->sd_args.ar_spectator)
914 pn = gfs2_lookup_simple(master, "per_node");
917 fs_err(sdp, "can't find per_node directory: %d\n", error);
921 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
922 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
923 if (IS_ERR(sdp->sd_ir_inode)) {
924 error = PTR_ERR(sdp->sd_ir_inode);
925 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
929 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
930 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
931 if (IS_ERR(sdp->sd_sc_inode)) {
932 error = PTR_ERR(sdp->sd_sc_inode);
933 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
937 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
938 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
939 if (IS_ERR(sdp->sd_qc_inode)) {
940 error = PTR_ERR(sdp->sd_qc_inode);
941 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
948 ip = GFS2_I(sdp->sd_ir_inode);
949 error = gfs2_glock_nq_init(ip->i_gl,
953 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
957 ip = GFS2_I(sdp->sd_sc_inode);
958 error = gfs2_glock_nq_init(ip->i_gl,
962 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
966 ip = GFS2_I(sdp->sd_qc_inode);
967 error = gfs2_glock_nq_init(ip->i_gl,
971 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
978 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
980 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
982 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
984 iput(sdp->sd_qc_inode);
986 iput(sdp->sd_sc_inode);
988 iput(sdp->sd_ir_inode);
995 static int init_threads(struct gfs2_sbd *sdp, int undo)
997 struct task_struct *p;
1003 sdp->sd_log_flush_time = jiffies;
1005 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
1008 fs_err(sdp, "can't start logd thread: %d\n", error);
1011 sdp->sd_logd_process = p;
1013 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
1016 fs_err(sdp, "can't start quotad thread: %d\n", error);
1019 sdp->sd_quotad_process = p;
1025 kthread_stop(sdp->sd_quotad_process);
1027 kthread_stop(sdp->sd_logd_process);
1031 static const match_table_t nolock_tokens = {
1032 { Opt_jid, "jid=%d\n", },
1036 static const struct lm_lockops nolock_ops = {
1037 .lm_proto_name = "lock_nolock",
1038 .lm_put_lock = kmem_cache_free,
1039 .lm_tokens = &nolock_tokens,
1043 * gfs2_lm_mount - mount a locking protocol
1044 * @sdp: the filesystem
1045 * @args: mount arguements
1046 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1051 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1053 const struct lm_lockops *lm;
1054 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1055 struct gfs2_args *args = &sdp->sd_args;
1056 const char *proto = sdp->sd_proto_name;
1057 const char *table = sdp->sd_table_name;
1062 if (!strcmp("lock_nolock", proto)) {
1064 sdp->sd_args.ar_localflocks = 1;
1065 sdp->sd_args.ar_localcaching = 1;
1066 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1067 } else if (!strcmp("lock_dlm", proto)) {
1071 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1075 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1081 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1082 substring_t tmp[MAX_OPT_ARGS];
1088 token = match_token(o, *lm->lm_tokens, tmp);
1091 ret = match_int(&tmp[0], &option);
1092 if (ret || option < 0)
1093 goto hostdata_error;
1094 ls->ls_jid = option;
1097 ret = match_int(&tmp[0], &option);
1099 goto hostdata_error;
1103 ret = match_int(&tmp[0], &option);
1104 if (ret || (option != 0 && option != 1))
1105 goto hostdata_error;
1106 ls->ls_first = option;
1109 ret = match_int(&tmp[0], &option);
1110 if (ret || (option != 0 && option != 1))
1111 goto hostdata_error;
1112 ls->ls_nodir = option;
1117 fs_info(sdp, "unknown hostdata (%s)\n", o);
1122 if (sdp->sd_args.ar_spectator)
1123 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1125 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1126 sdp->sd_lockstruct.ls_jid);
1128 fsname = strchr(table, ':');
1131 if (lm->lm_mount == NULL) {
1132 fs_info(sdp, "Now mounting FS...\n");
1135 ret = lm->lm_mount(sdp, fsname);
1137 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1141 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1143 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1144 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1146 lm->lm_unmount(sdp);
1150 * fill_super - Read in superblock
1151 * @sb: The VFS superblock
1152 * @data: Mount options
1153 * @silent: Don't complain if it's not a GFS2 filesystem
1158 static int fill_super(struct super_block *sb, void *data, int silent)
1160 struct gfs2_sbd *sdp;
1161 struct gfs2_holder mount_gh;
1166 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1170 sdp->sd_args.ar_quota = GFS2_QUOTA_DEFAULT;
1171 sdp->sd_args.ar_data = GFS2_DATA_DEFAULT;
1173 error = gfs2_mount_args(sdp, &sdp->sd_args, data);
1175 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1179 if (sdp->sd_args.ar_spectator)
1180 sb->s_flags |= MS_RDONLY;
1181 if (sdp->sd_args.ar_posix_acl)
1182 sb->s_flags |= MS_POSIXACL;
1184 sb->s_magic = GFS2_MAGIC;
1185 sb->s_op = &gfs2_super_ops;
1186 sb->s_export_op = &gfs2_export_ops;
1187 sb->s_time_gran = 1;
1188 sb->s_maxbytes = MAX_LFS_FILESIZE;
1190 /* Set up the buffer cache and fill in some fake block size values
1191 to allow us to read-in the on-disk superblock. */
1192 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1193 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1194 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1195 GFS2_BASIC_BLOCK_SHIFT;
1196 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1198 error = init_names(sdp, silent);
1202 gfs2_create_debugfs_file(sdp);
1204 error = gfs2_sys_fs_add(sdp);
1208 error = gfs2_lm_mount(sdp, silent);
1212 error = init_locking(sdp, &mount_gh, DO);
1216 error = init_sb(sdp, silent);
1220 error = init_inodes(sdp, DO);
1224 error = init_per_node(sdp, DO);
1228 error = gfs2_statfs_init(sdp);
1230 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1234 error = init_threads(sdp, DO);
1238 if (!(sb->s_flags & MS_RDONLY)) {
1239 error = gfs2_make_fs_rw(sdp);
1241 fs_err(sdp, "can't make FS RW: %d\n", error);
1246 gfs2_glock_dq_uninit(&mount_gh);
1251 init_threads(sdp, UNDO);
1253 init_per_node(sdp, UNDO);
1255 init_inodes(sdp, UNDO);
1257 if (sdp->sd_root_dir)
1258 dput(sdp->sd_root_dir);
1259 if (sdp->sd_master_dir)
1260 dput(sdp->sd_master_dir);
1265 init_locking(sdp, &mount_gh, UNDO);
1267 gfs2_gl_hash_clear(sdp);
1268 gfs2_lm_unmount(sdp);
1269 while (invalidate_inodes(sb))
1272 gfs2_sys_fs_del(sdp);
1274 gfs2_delete_debugfs_file(sdp);
1276 sb->s_fs_info = NULL;
1280 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1281 const char *dev_name, void *data, struct vfsmount *mnt)
1283 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1286 static struct super_block *get_gfs2_sb(const char *dev_name)
1288 struct super_block *sb;
1289 struct nameidata nd;
1292 error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd);
1294 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1298 sb = nd.path.dentry->d_inode->i_sb;
1299 if (sb && (sb->s_type == &gfs2_fs_type))
1300 atomic_inc(&sb->s_active);
1307 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1308 const char *dev_name, void *data, struct vfsmount *mnt)
1310 struct super_block *sb = NULL;
1311 struct gfs2_sbd *sdp;
1313 sb = get_gfs2_sb(dev_name);
1315 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1318 sdp = sb->s_fs_info;
1320 mnt->mnt_root = dget(sdp->sd_master_dir);
1324 static void gfs2_kill_sb(struct super_block *sb)
1326 struct gfs2_sbd *sdp = sb->s_fs_info;
1329 kill_block_super(sb);
1333 gfs2_meta_syncfs(sdp);
1334 dput(sdp->sd_root_dir);
1335 dput(sdp->sd_master_dir);
1336 sdp->sd_root_dir = NULL;
1337 sdp->sd_master_dir = NULL;
1338 shrink_dcache_sb(sb);
1339 kill_block_super(sb);
1340 gfs2_delete_debugfs_file(sdp);
1344 struct file_system_type gfs2_fs_type = {
1346 .fs_flags = FS_REQUIRES_DEV,
1347 .get_sb = gfs2_get_sb,
1348 .kill_sb = gfs2_kill_sb,
1349 .owner = THIS_MODULE,
1352 struct file_system_type gfs2meta_fs_type = {
1354 .fs_flags = FS_REQUIRES_DEV,
1355 .get_sb = gfs2_get_sb_meta,
1356 .owner = THIS_MODULE,