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/slow-work.h>
36 #include "trace_gfs2.h"
41 static const u32 gfs2_old_fs_formats[] = {
45 static const u32 gfs2_old_multihost_formats[] = {
50 * gfs2_tune_init - Fill a gfs2_tune structure with default values
55 static void gfs2_tune_init(struct gfs2_tune *gt)
57 spin_lock_init(>->gt_spin);
59 gt->gt_incore_log_blocks = 1024;
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);
237 memcpy(sb->sb_uuid, str->sb_uuid, 16);
241 * gfs2_read_super - Read the gfs2 super block from disk
242 * @sdp: The GFS2 super block
243 * @sector: The location of the super block
244 * @error: The error code to return
246 * This uses the bio functions to read the super block from disk
247 * because we want to be 100% sure that we never read cached data.
248 * A super block is read twice only during each GFS2 mount and is
249 * never written to by the filesystem. The first time its read no
250 * locks are held, and the only details which are looked at are those
251 * relating to the locking protocol. Once locking is up and working,
252 * the sb is read again under the lock to establish the location of
253 * the master directory (contains pointers to journals etc) and the
256 * Returns: 0 on success or error
259 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
261 struct super_block *sb = sdp->sd_vfs;
266 page = alloc_page(GFP_NOFS);
270 ClearPageUptodate(page);
271 ClearPageDirty(page);
274 bio = bio_alloc(GFP_NOFS, 1);
275 bio->bi_sector = sector * (sb->s_blocksize >> 9);
276 bio->bi_bdev = sb->s_bdev;
277 bio_add_page(bio, page, PAGE_SIZE, 0);
279 bio->bi_end_io = end_bio_io_page;
280 bio->bi_private = page;
281 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
282 wait_on_page_locked(page);
284 if (!PageUptodate(page)) {
289 gfs2_sb_in(&sdp->sd_sb, p);
296 * gfs2_read_sb - Read super block
297 * @sdp: The GFS2 superblock
298 * @silent: Don't print message if mount fails
302 static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
304 u32 hash_blocks, ind_blocks, leaf_blocks;
309 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
312 fs_err(sdp, "can't read superblock\n");
316 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
320 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
321 GFS2_BASIC_BLOCK_SHIFT;
322 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
323 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
324 sizeof(struct gfs2_dinode)) / sizeof(u64);
325 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
326 sizeof(struct gfs2_meta_header)) / sizeof(u64);
327 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
328 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
329 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
330 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
331 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
332 sizeof(struct gfs2_meta_header)) /
333 sizeof(struct gfs2_quota_change);
335 /* Compute maximum reservation required to add a entry to a directory */
337 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
341 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
342 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
343 ind_blocks += tmp_blocks;
346 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
348 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
350 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
351 sizeof(struct gfs2_dinode);
352 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
357 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
359 m = do_div(d, sdp->sd_inptrs);
361 if (d != sdp->sd_heightsize[x - 1] || m)
363 sdp->sd_heightsize[x] = space;
365 sdp->sd_max_height = x;
366 sdp->sd_heightsize[x] = ~0;
367 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
369 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
370 sizeof(struct gfs2_dinode);
371 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
376 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
378 m = do_div(d, sdp->sd_inptrs);
380 if (d != sdp->sd_jheightsize[x - 1] || m)
382 sdp->sd_jheightsize[x] = space;
384 sdp->sd_max_jheight = x;
385 sdp->sd_jheightsize[x] = ~0;
386 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
391 static int init_names(struct gfs2_sbd *sdp, int silent)
396 proto = sdp->sd_args.ar_lockproto;
397 table = sdp->sd_args.ar_locktable;
399 /* Try to autodetect */
401 if (!proto[0] || !table[0]) {
402 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
406 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
411 proto = sdp->sd_sb.sb_lockproto;
413 table = sdp->sd_sb.sb_locktable;
417 table = sdp->sd_vfs->s_id;
419 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
420 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
422 table = sdp->sd_table_name;
423 while ((table = strchr(table, '/')))
430 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
438 error = gfs2_glock_nq_num(sdp,
439 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
440 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
443 fs_err(sdp, "can't acquire mount glock: %d\n", error);
447 error = gfs2_glock_nq_num(sdp,
448 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
450 LM_FLAG_NOEXP | GL_EXACT,
453 fs_err(sdp, "can't acquire live glock: %d\n", error);
457 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
458 CREATE, &sdp->sd_rename_gl);
460 fs_err(sdp, "can't create rename glock: %d\n", error);
464 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
465 CREATE, &sdp->sd_trans_gl);
467 fs_err(sdp, "can't create transaction glock: %d\n", error);
474 gfs2_glock_put(sdp->sd_trans_gl);
476 gfs2_glock_put(sdp->sd_rename_gl);
478 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
480 gfs2_glock_dq_uninit(mount_gh);
485 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
486 u64 no_addr, const char *name)
488 struct gfs2_sbd *sdp = sb->s_fs_info;
489 struct dentry *dentry;
492 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
494 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
495 return PTR_ERR(inode);
497 dentry = d_alloc_root(inode);
499 fs_err(sdp, "can't alloc %s dentry\n", name);
503 dentry->d_op = &gfs2_dops;
508 static int init_sb(struct gfs2_sbd *sdp, int silent)
510 struct super_block *sb = sdp->sd_vfs;
511 struct gfs2_holder sb_gh;
515 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
516 LM_ST_SHARED, 0, &sb_gh);
518 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
522 ret = gfs2_read_sb(sdp, silent);
524 fs_err(sdp, "can't read superblock: %d\n", ret);
528 /* Set up the buffer cache and SB for real */
529 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
531 fs_err(sdp, "FS block size (%u) is too small for device "
533 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
536 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
538 fs_err(sdp, "FS block size (%u) is too big for machine "
540 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
543 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
545 /* Get the root inode */
546 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
547 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
551 /* Get the master inode */
552 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
553 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
555 dput(sdp->sd_root_dir);
558 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
560 gfs2_glock_dq_uninit(&sb_gh);
565 * map_journal_extents - create a reusable "extent" mapping from all logical
566 * blocks to all physical blocks for the given journal. This will save
567 * us time when writing journal blocks. Most journals will have only one
568 * extent that maps all their logical blocks. That's because gfs2.mkfs
569 * arranges the journal blocks sequentially to maximize performance.
570 * So the extent would map the first block for the entire file length.
571 * However, gfs2_jadd can happen while file activity is happening, so
572 * those journals may not be sequential. Less likely is the case where
573 * the users created their own journals by mounting the metafs and
574 * laying it out. But it's still possible. These journals might have
577 * TODO: This should be done in bigger chunks rather than one block at a time,
578 * but since it's only done at mount time, I'm not worried about the
581 static int map_journal_extents(struct gfs2_sbd *sdp)
583 struct gfs2_jdesc *jd = sdp->sd_jdesc;
585 u64 db, prev_db; /* logical block, disk block, prev disk block */
586 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
587 struct gfs2_journal_extent *jext = NULL;
588 struct buffer_head bh;
593 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
596 bh.b_size = 1 << ip->i_inode.i_blkbits;
597 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
600 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
601 "%u db=%llu\n", rc, lb, (unsigned long long)db);
604 if (!prev_db || db != prev_db + 1) {
605 jext = kzalloc(sizeof(struct gfs2_journal_extent),
608 printk(KERN_INFO "GFS2 error: out of memory "
609 "mapping journal extents.\n");
616 list_add_tail(&jext->extent_list, &jd->extent_list);
625 static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
627 char *message = "FIRSTMOUNT=Done";
628 char *envp[] = { message, NULL };
629 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
630 ls->ls_first_done = 1;
631 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
635 * gfs2_jindex_hold - Grab a lock on the jindex
636 * @sdp: The GFS2 superblock
637 * @ji_gh: the holder for the jindex glock
642 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
644 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
647 struct gfs2_jdesc *jd;
652 mutex_lock(&sdp->sd_jindex_mutex);
655 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
659 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
660 name.hash = gfs2_disk_hash(name.name, name.len);
662 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
663 if (error == -ENOENT) {
668 gfs2_glock_dq_uninit(ji_gh);
674 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
678 INIT_LIST_HEAD(&jd->extent_list);
679 slow_work_init(&jd->jd_work, &gfs2_recover_ops);
680 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
681 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
685 error = PTR_ERR(jd->jd_inode);
690 spin_lock(&sdp->sd_jindex_spin);
691 jd->jd_jid = sdp->sd_journals++;
692 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
693 spin_unlock(&sdp->sd_jindex_spin);
696 mutex_unlock(&sdp->sd_jindex_mutex);
701 static int init_journal(struct gfs2_sbd *sdp, int undo)
703 struct inode *master = sdp->sd_master_dir->d_inode;
704 struct gfs2_holder ji_gh;
705 struct gfs2_inode *ip;
714 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
715 if (IS_ERR(sdp->sd_jindex)) {
716 fs_err(sdp, "can't lookup journal index: %d\n", error);
717 return PTR_ERR(sdp->sd_jindex);
719 ip = GFS2_I(sdp->sd_jindex);
721 /* Load in the journal index special file */
723 error = gfs2_jindex_hold(sdp, &ji_gh);
725 fs_err(sdp, "can't read journal index: %d\n", error);
730 if (!gfs2_jindex_size(sdp)) {
731 fs_err(sdp, "no journals!\n");
735 if (sdp->sd_args.ar_spectator) {
736 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
737 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
739 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
740 fs_err(sdp, "can't mount journal #%u\n",
741 sdp->sd_lockstruct.ls_jid);
742 fs_err(sdp, "there are only %u journals (0 - %u)\n",
743 gfs2_jindex_size(sdp),
744 gfs2_jindex_size(sdp) - 1);
747 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
749 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
751 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
752 &sdp->sd_journal_gh);
754 fs_err(sdp, "can't acquire journal glock: %d\n", error);
758 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
759 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
760 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
763 fs_err(sdp, "can't acquire journal inode glock: %d\n",
765 goto fail_journal_gh;
768 error = gfs2_jdesc_check(sdp->sd_jdesc);
770 fs_err(sdp, "my journal (%u) is bad: %d\n",
771 sdp->sd_jdesc->jd_jid, error);
774 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
776 /* Map the extents for this journal's blocks */
777 map_journal_extents(sdp);
779 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
781 if (sdp->sd_lockstruct.ls_first) {
783 for (x = 0; x < sdp->sd_journals; x++) {
784 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
786 fs_err(sdp, "error recovering journal %u: %d\n",
792 gfs2_others_may_mount(sdp);
793 } else if (!sdp->sd_args.ar_spectator) {
794 error = gfs2_recover_journal(sdp->sd_jdesc);
796 fs_err(sdp, "error recovering my journal: %d\n", error);
801 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
802 gfs2_glock_dq_uninit(&ji_gh);
808 if (!sdp->sd_args.ar_spectator)
809 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
811 if (!sdp->sd_args.ar_spectator)
812 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
814 gfs2_jindex_free(sdp);
816 gfs2_glock_dq_uninit(&ji_gh);
818 iput(sdp->sd_jindex);
823 static int init_inodes(struct gfs2_sbd *sdp, int undo)
826 struct gfs2_inode *ip;
827 struct inode *master = sdp->sd_master_dir->d_inode;
832 error = init_journal(sdp, undo);
836 /* Read in the master inode number inode */
837 sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum");
838 if (IS_ERR(sdp->sd_inum_inode)) {
839 error = PTR_ERR(sdp->sd_inum_inode);
840 fs_err(sdp, "can't read in inum inode: %d\n", error);
845 /* Read in the master statfs inode */
846 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
847 if (IS_ERR(sdp->sd_statfs_inode)) {
848 error = PTR_ERR(sdp->sd_statfs_inode);
849 fs_err(sdp, "can't read in statfs inode: %d\n", error);
853 /* Read in the resource index inode */
854 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
855 if (IS_ERR(sdp->sd_rindex)) {
856 error = PTR_ERR(sdp->sd_rindex);
857 fs_err(sdp, "can't get resource index inode: %d\n", error);
860 ip = GFS2_I(sdp->sd_rindex);
861 sdp->sd_rindex_uptodate = 0;
863 /* Read in the quota inode */
864 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
865 if (IS_ERR(sdp->sd_quota_inode)) {
866 error = PTR_ERR(sdp->sd_quota_inode);
867 fs_err(sdp, "can't get quota file inode: %d\n", error);
873 iput(sdp->sd_quota_inode);
875 gfs2_clear_rgrpd(sdp);
876 iput(sdp->sd_rindex);
878 iput(sdp->sd_statfs_inode);
880 iput(sdp->sd_inum_inode);
882 init_journal(sdp, UNDO);
887 static int init_per_node(struct gfs2_sbd *sdp, int undo)
889 struct inode *pn = NULL;
892 struct gfs2_inode *ip;
893 struct inode *master = sdp->sd_master_dir->d_inode;
895 if (sdp->sd_args.ar_spectator)
901 pn = gfs2_lookup_simple(master, "per_node");
904 fs_err(sdp, "can't find per_node directory: %d\n", error);
908 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
909 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
910 if (IS_ERR(sdp->sd_ir_inode)) {
911 error = PTR_ERR(sdp->sd_ir_inode);
912 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
916 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
917 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
918 if (IS_ERR(sdp->sd_sc_inode)) {
919 error = PTR_ERR(sdp->sd_sc_inode);
920 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
924 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
925 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
926 if (IS_ERR(sdp->sd_qc_inode)) {
927 error = PTR_ERR(sdp->sd_qc_inode);
928 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
935 ip = GFS2_I(sdp->sd_ir_inode);
936 error = gfs2_glock_nq_init(ip->i_gl,
940 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
944 ip = GFS2_I(sdp->sd_sc_inode);
945 error = gfs2_glock_nq_init(ip->i_gl,
949 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
953 ip = GFS2_I(sdp->sd_qc_inode);
954 error = gfs2_glock_nq_init(ip->i_gl,
958 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
965 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
967 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
969 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
971 iput(sdp->sd_qc_inode);
973 iput(sdp->sd_sc_inode);
975 iput(sdp->sd_ir_inode);
982 static int init_threads(struct gfs2_sbd *sdp, int undo)
984 struct task_struct *p;
990 sdp->sd_log_flush_time = jiffies;
992 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
995 fs_err(sdp, "can't start logd thread: %d\n", error);
998 sdp->sd_logd_process = p;
1000 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
1003 fs_err(sdp, "can't start quotad thread: %d\n", error);
1006 sdp->sd_quotad_process = p;
1012 kthread_stop(sdp->sd_quotad_process);
1014 kthread_stop(sdp->sd_logd_process);
1018 static const match_table_t nolock_tokens = {
1019 { Opt_jid, "jid=%d\n", },
1023 static const struct lm_lockops nolock_ops = {
1024 .lm_proto_name = "lock_nolock",
1025 .lm_put_lock = kmem_cache_free,
1026 .lm_tokens = &nolock_tokens,
1030 * gfs2_lm_mount - mount a locking protocol
1031 * @sdp: the filesystem
1032 * @args: mount arguements
1033 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1038 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1040 const struct lm_lockops *lm;
1041 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1042 struct gfs2_args *args = &sdp->sd_args;
1043 const char *proto = sdp->sd_proto_name;
1044 const char *table = sdp->sd_table_name;
1049 if (!strcmp("lock_nolock", proto)) {
1051 sdp->sd_args.ar_localflocks = 1;
1052 sdp->sd_args.ar_localcaching = 1;
1053 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1054 } else if (!strcmp("lock_dlm", proto)) {
1058 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1062 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1068 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1069 substring_t tmp[MAX_OPT_ARGS];
1075 token = match_token(o, *lm->lm_tokens, tmp);
1078 ret = match_int(&tmp[0], &option);
1079 if (ret || option < 0)
1080 goto hostdata_error;
1081 ls->ls_jid = option;
1084 ret = match_int(&tmp[0], &option);
1086 goto hostdata_error;
1090 ret = match_int(&tmp[0], &option);
1091 if (ret || (option != 0 && option != 1))
1092 goto hostdata_error;
1093 ls->ls_first = option;
1096 ret = match_int(&tmp[0], &option);
1097 if (ret || (option != 0 && option != 1))
1098 goto hostdata_error;
1099 ls->ls_nodir = option;
1104 fs_info(sdp, "unknown hostdata (%s)\n", o);
1109 if (sdp->sd_args.ar_spectator)
1110 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1112 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1113 sdp->sd_lockstruct.ls_jid);
1115 fsname = strchr(table, ':');
1118 if (lm->lm_mount == NULL) {
1119 fs_info(sdp, "Now mounting FS...\n");
1122 ret = lm->lm_mount(sdp, fsname);
1124 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1128 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1130 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1131 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1133 lm->lm_unmount(sdp);
1137 * fill_super - Read in superblock
1138 * @sb: The VFS superblock
1139 * @data: Mount options
1140 * @silent: Don't complain if it's not a GFS2 filesystem
1145 static int fill_super(struct super_block *sb, void *data, int silent)
1147 struct gfs2_sbd *sdp;
1148 struct gfs2_holder mount_gh;
1153 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1157 sdp->sd_args.ar_quota = GFS2_QUOTA_DEFAULT;
1158 sdp->sd_args.ar_data = GFS2_DATA_DEFAULT;
1159 sdp->sd_args.ar_commit = 60;
1161 error = gfs2_mount_args(sdp, &sdp->sd_args, data);
1163 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1167 if (sdp->sd_args.ar_spectator) {
1168 sb->s_flags |= MS_RDONLY;
1169 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1171 if (sdp->sd_args.ar_posix_acl)
1172 sb->s_flags |= MS_POSIXACL;
1174 sb->s_magic = GFS2_MAGIC;
1175 sb->s_op = &gfs2_super_ops;
1176 sb->s_export_op = &gfs2_export_ops;
1177 sb->s_time_gran = 1;
1178 sb->s_maxbytes = MAX_LFS_FILESIZE;
1180 /* Set up the buffer cache and fill in some fake block size values
1181 to allow us to read-in the on-disk superblock. */
1182 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1183 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1184 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1185 GFS2_BASIC_BLOCK_SHIFT;
1186 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1188 sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit;
1190 error = init_names(sdp, silent);
1194 gfs2_create_debugfs_file(sdp);
1196 error = gfs2_sys_fs_add(sdp);
1200 error = gfs2_lm_mount(sdp, silent);
1204 error = init_locking(sdp, &mount_gh, DO);
1208 error = init_sb(sdp, silent);
1212 error = init_inodes(sdp, DO);
1216 error = init_per_node(sdp, DO);
1220 error = gfs2_statfs_init(sdp);
1222 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1226 error = init_threads(sdp, DO);
1230 if (!(sb->s_flags & MS_RDONLY)) {
1231 error = gfs2_make_fs_rw(sdp);
1233 fs_err(sdp, "can't make FS RW: %d\n", error);
1238 gfs2_glock_dq_uninit(&mount_gh);
1243 init_threads(sdp, UNDO);
1245 init_per_node(sdp, UNDO);
1247 init_inodes(sdp, UNDO);
1249 if (sdp->sd_root_dir)
1250 dput(sdp->sd_root_dir);
1251 if (sdp->sd_master_dir)
1252 dput(sdp->sd_master_dir);
1257 init_locking(sdp, &mount_gh, UNDO);
1259 gfs2_gl_hash_clear(sdp);
1260 gfs2_lm_unmount(sdp);
1261 while (invalidate_inodes(sb))
1264 gfs2_sys_fs_del(sdp);
1266 gfs2_delete_debugfs_file(sdp);
1268 sb->s_fs_info = NULL;
1272 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1273 const char *dev_name, void *data, struct vfsmount *mnt)
1275 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1278 static int test_meta_super(struct super_block *s, void *ptr)
1280 struct block_device *bdev = ptr;
1281 return (bdev == s->s_bdev);
1284 static int set_meta_super(struct super_block *s, void *ptr)
1289 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1290 const char *dev_name, void *data, struct vfsmount *mnt)
1292 struct super_block *s;
1293 struct gfs2_sbd *sdp;
1297 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
1299 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1303 s = sget(&gfs2_fs_type, test_meta_super, set_meta_super,
1304 path.dentry->d_inode->i_sb->s_bdev);
1307 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1312 mnt->mnt_root = dget(sdp->sd_master_dir);
1316 static void gfs2_kill_sb(struct super_block *sb)
1318 struct gfs2_sbd *sdp = sb->s_fs_info;
1321 kill_block_super(sb);
1325 gfs2_meta_syncfs(sdp);
1326 dput(sdp->sd_root_dir);
1327 dput(sdp->sd_master_dir);
1328 sdp->sd_root_dir = NULL;
1329 sdp->sd_master_dir = NULL;
1330 shrink_dcache_sb(sb);
1331 kill_block_super(sb);
1332 gfs2_delete_debugfs_file(sdp);
1336 struct file_system_type gfs2_fs_type = {
1338 .fs_flags = FS_REQUIRES_DEV,
1339 .get_sb = gfs2_get_sb,
1340 .kill_sb = gfs2_kill_sb,
1341 .owner = THIS_MODULE,
1344 struct file_system_type gfs2meta_fs_type = {
1346 .fs_flags = FS_REQUIRES_DEV,
1347 .get_sb = gfs2_get_sb_meta,
1348 .owner = THIS_MODULE,