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>
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;
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 mutex_init(&sdp->sd_inum_mutex);
91 spin_lock_init(&sdp->sd_statfs_spin);
93 spin_lock_init(&sdp->sd_rindex_spin);
94 mutex_init(&sdp->sd_rindex_mutex);
95 INIT_LIST_HEAD(&sdp->sd_rindex_list);
96 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
98 INIT_LIST_HEAD(&sdp->sd_jindex_list);
99 spin_lock_init(&sdp->sd_jindex_spin);
100 mutex_init(&sdp->sd_jindex_mutex);
102 INIT_LIST_HEAD(&sdp->sd_quota_list);
103 spin_lock_init(&sdp->sd_quota_spin);
104 mutex_init(&sdp->sd_quota_mutex);
105 init_waitqueue_head(&sdp->sd_quota_wait);
106 INIT_LIST_HEAD(&sdp->sd_trunc_list);
107 spin_lock_init(&sdp->sd_trunc_lock);
109 spin_lock_init(&sdp->sd_log_lock);
111 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
112 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
113 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
114 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
115 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
117 mutex_init(&sdp->sd_log_reserve_mutex);
118 INIT_LIST_HEAD(&sdp->sd_ail1_list);
119 INIT_LIST_HEAD(&sdp->sd_ail2_list);
121 init_rwsem(&sdp->sd_log_flush_lock);
122 atomic_set(&sdp->sd_log_in_flight, 0);
123 init_waitqueue_head(&sdp->sd_log_flush_wait);
125 INIT_LIST_HEAD(&sdp->sd_revoke_list);
127 mutex_init(&sdp->sd_freeze_lock);
134 * gfs2_check_sb - Check superblock
135 * @sdp: the filesystem
136 * @sb: The superblock
137 * @silent: Don't print a message if the check fails
139 * Checks the version code of the FS is one that we understand how to
140 * read and that the sizes of the various on-disk structures have not
144 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
148 if (sb->sb_magic != GFS2_MAGIC ||
149 sb->sb_type != GFS2_METATYPE_SB) {
151 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
155 /* If format numbers match exactly, we're done. */
157 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
158 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
161 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
162 for (x = 0; gfs2_old_fs_formats[x]; x++)
163 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
166 if (!gfs2_old_fs_formats[x]) {
168 "GFS2: code version (%u, %u) is incompatible "
169 "with ondisk format (%u, %u)\n",
170 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
171 sb->sb_fs_format, sb->sb_multihost_format);
173 "GFS2: I don't know how to upgrade this FS\n");
178 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
179 for (x = 0; gfs2_old_multihost_formats[x]; x++)
180 if (gfs2_old_multihost_formats[x] ==
181 sb->sb_multihost_format)
184 if (!gfs2_old_multihost_formats[x]) {
186 "GFS2: code version (%u, %u) is incompatible "
187 "with ondisk format (%u, %u)\n",
188 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
189 sb->sb_fs_format, sb->sb_multihost_format);
191 "GFS2: I don't know how to upgrade this FS\n");
196 if (!sdp->sd_args.ar_upgrade) {
198 "GFS2: code version (%u, %u) is incompatible "
199 "with ondisk format (%u, %u)\n",
200 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
201 sb->sb_fs_format, sb->sb_multihost_format);
203 "GFS2: Use the \"upgrade\" mount option to upgrade "
205 printk(KERN_INFO "GFS2: See the manual for more details\n");
212 static void end_bio_io_page(struct bio *bio, int error)
214 struct page *page = bio->bi_private;
217 SetPageUptodate(page);
219 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
223 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
225 const struct gfs2_sb *str = buf;
227 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
228 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
229 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
230 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
231 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
232 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
233 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
234 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
235 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
236 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
237 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
239 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
240 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
244 * gfs2_read_super - Read the gfs2 super block from disk
245 * @sdp: The GFS2 super block
246 * @sector: The location of the super block
247 * @error: The error code to return
249 * This uses the bio functions to read the super block from disk
250 * because we want to be 100% sure that we never read cached data.
251 * A super block is read twice only during each GFS2 mount and is
252 * never written to by the filesystem. The first time its read no
253 * locks are held, and the only details which are looked at are those
254 * relating to the locking protocol. Once locking is up and working,
255 * the sb is read again under the lock to establish the location of
256 * the master directory (contains pointers to journals etc) and the
259 * Returns: 0 on success or error
262 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
264 struct super_block *sb = sdp->sd_vfs;
269 page = alloc_page(GFP_NOFS);
273 ClearPageUptodate(page);
274 ClearPageDirty(page);
277 bio = bio_alloc(GFP_NOFS, 1);
278 if (unlikely(!bio)) {
283 bio->bi_sector = sector * (sb->s_blocksize >> 9);
284 bio->bi_bdev = sb->s_bdev;
285 bio_add_page(bio, page, PAGE_SIZE, 0);
287 bio->bi_end_io = end_bio_io_page;
288 bio->bi_private = page;
289 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
290 wait_on_page_locked(page);
292 if (!PageUptodate(page)) {
297 gfs2_sb_in(&sdp->sd_sb, p);
303 * gfs2_read_sb - Read super block
304 * @sdp: The GFS2 superblock
305 * @gl: the glock for the superblock (assumed to be held)
306 * @silent: Don't print message if mount fails
310 static int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
312 u32 hash_blocks, ind_blocks, leaf_blocks;
317 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
320 fs_err(sdp, "can't read superblock\n");
324 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
328 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
329 GFS2_BASIC_BLOCK_SHIFT;
330 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
331 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
332 sizeof(struct gfs2_dinode)) / sizeof(u64);
333 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
334 sizeof(struct gfs2_meta_header)) / sizeof(u64);
335 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
336 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
337 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
338 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
339 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
340 sizeof(struct gfs2_meta_header)) /
341 sizeof(struct gfs2_quota_change);
343 /* Compute maximum reservation required to add a entry to a directory */
345 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
349 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
350 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
351 ind_blocks += tmp_blocks;
354 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
356 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
358 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
359 sizeof(struct gfs2_dinode);
360 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
365 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
367 m = do_div(d, sdp->sd_inptrs);
369 if (d != sdp->sd_heightsize[x - 1] || m)
371 sdp->sd_heightsize[x] = space;
373 sdp->sd_max_height = x;
374 sdp->sd_heightsize[x] = ~0;
375 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
377 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
378 sizeof(struct gfs2_dinode);
379 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
384 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
386 m = do_div(d, sdp->sd_inptrs);
388 if (d != sdp->sd_jheightsize[x - 1] || m)
390 sdp->sd_jheightsize[x] = space;
392 sdp->sd_max_jheight = x;
393 sdp->sd_jheightsize[x] = ~0;
394 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
399 static int init_names(struct gfs2_sbd *sdp, int silent)
404 proto = sdp->sd_args.ar_lockproto;
405 table = sdp->sd_args.ar_locktable;
407 /* Try to autodetect */
409 if (!proto[0] || !table[0]) {
410 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
414 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
419 proto = sdp->sd_sb.sb_lockproto;
421 table = sdp->sd_sb.sb_locktable;
425 table = sdp->sd_vfs->s_id;
427 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
428 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
430 table = sdp->sd_table_name;
431 while ((table = strchr(table, '/')))
438 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
446 error = gfs2_glock_nq_num(sdp,
447 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
448 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
451 fs_err(sdp, "can't acquire mount glock: %d\n", error);
455 error = gfs2_glock_nq_num(sdp,
456 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
458 LM_FLAG_NOEXP | GL_EXACT,
461 fs_err(sdp, "can't acquire live glock: %d\n", error);
465 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
466 CREATE, &sdp->sd_rename_gl);
468 fs_err(sdp, "can't create rename glock: %d\n", error);
472 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
473 CREATE, &sdp->sd_trans_gl);
475 fs_err(sdp, "can't create transaction glock: %d\n", error);
482 gfs2_glock_put(sdp->sd_trans_gl);
484 gfs2_glock_put(sdp->sd_rename_gl);
486 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
488 gfs2_glock_dq_uninit(mount_gh);
493 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
494 u64 no_addr, const char *name)
496 struct gfs2_sbd *sdp = sb->s_fs_info;
497 struct dentry *dentry;
500 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
502 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
503 return PTR_ERR(inode);
505 dentry = d_alloc_root(inode);
507 fs_err(sdp, "can't alloc %s dentry\n", name);
511 dentry->d_op = &gfs2_dops;
516 static int init_sb(struct gfs2_sbd *sdp, int silent)
518 struct super_block *sb = sdp->sd_vfs;
519 struct gfs2_holder sb_gh;
523 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
524 LM_ST_SHARED, 0, &sb_gh);
526 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
530 ret = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
532 fs_err(sdp, "can't read superblock: %d\n", ret);
536 /* Set up the buffer cache and SB for real */
537 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
539 fs_err(sdp, "FS block size (%u) is too small for device "
541 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
544 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
546 fs_err(sdp, "FS block size (%u) is too big for machine "
548 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
551 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
553 /* Get the root inode */
554 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
555 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
559 /* Get the master inode */
560 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
561 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
563 dput(sdp->sd_root_dir);
566 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
568 gfs2_glock_dq_uninit(&sb_gh);
573 * map_journal_extents - create a reusable "extent" mapping from all logical
574 * blocks to all physical blocks for the given journal. This will save
575 * us time when writing journal blocks. Most journals will have only one
576 * extent that maps all their logical blocks. That's because gfs2.mkfs
577 * arranges the journal blocks sequentially to maximize performance.
578 * So the extent would map the first block for the entire file length.
579 * However, gfs2_jadd can happen while file activity is happening, so
580 * those journals may not be sequential. Less likely is the case where
581 * the users created their own journals by mounting the metafs and
582 * laying it out. But it's still possible. These journals might have
585 * TODO: This should be done in bigger chunks rather than one block at a time,
586 * but since it's only done at mount time, I'm not worried about the
589 static int map_journal_extents(struct gfs2_sbd *sdp)
591 struct gfs2_jdesc *jd = sdp->sd_jdesc;
593 u64 db, prev_db; /* logical block, disk block, prev disk block */
594 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
595 struct gfs2_journal_extent *jext = NULL;
596 struct buffer_head bh;
601 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
604 bh.b_size = 1 << ip->i_inode.i_blkbits;
605 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
608 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
609 "%u db=%llu\n", rc, lb, (unsigned long long)db);
612 if (!prev_db || db != prev_db + 1) {
613 jext = kzalloc(sizeof(struct gfs2_journal_extent),
616 printk(KERN_INFO "GFS2 error: out of memory "
617 "mapping journal extents.\n");
624 list_add_tail(&jext->extent_list, &jd->extent_list);
633 static void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp)
635 if (!sdp->sd_lockstruct.ls_ops->lm_others_may_mount)
637 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
638 sdp->sd_lockstruct.ls_ops->lm_others_may_mount(
639 sdp->sd_lockstruct.ls_lockspace);
643 * gfs2_jindex_hold - Grab a lock on the jindex
644 * @sdp: The GFS2 superblock
645 * @ji_gh: the holder for the jindex glock
650 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
652 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
655 struct gfs2_jdesc *jd;
660 mutex_lock(&sdp->sd_jindex_mutex);
663 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
667 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
668 name.hash = gfs2_disk_hash(name.name, name.len);
670 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
671 if (error == -ENOENT) {
676 gfs2_glock_dq_uninit(ji_gh);
682 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
686 INIT_LIST_HEAD(&jd->extent_list);
687 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
688 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
692 error = PTR_ERR(jd->jd_inode);
697 spin_lock(&sdp->sd_jindex_spin);
698 jd->jd_jid = sdp->sd_journals++;
699 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
700 spin_unlock(&sdp->sd_jindex_spin);
703 mutex_unlock(&sdp->sd_jindex_mutex);
709 * gfs2_jindex_free - Clear all the journal index information
710 * @sdp: The GFS2 superblock
714 static void gfs2_jindex_free(struct gfs2_sbd *sdp)
716 struct list_head list, *head;
717 struct gfs2_jdesc *jd;
718 struct gfs2_journal_extent *jext;
720 spin_lock(&sdp->sd_jindex_spin);
721 list_add(&list, &sdp->sd_jindex_list);
722 list_del_init(&sdp->sd_jindex_list);
723 sdp->sd_journals = 0;
724 spin_unlock(&sdp->sd_jindex_spin);
726 while (!list_empty(&list)) {
727 jd = list_entry(list.next, struct gfs2_jdesc, jd_list);
728 head = &jd->extent_list;
729 while (!list_empty(head)) {
730 jext = list_entry(head->next,
731 struct gfs2_journal_extent,
733 list_del(&jext->extent_list);
736 list_del(&jd->jd_list);
742 static int init_journal(struct gfs2_sbd *sdp, int undo)
744 struct inode *master = sdp->sd_master_dir->d_inode;
745 struct gfs2_holder ji_gh;
746 struct task_struct *p;
747 struct gfs2_inode *ip;
756 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
757 if (IS_ERR(sdp->sd_jindex)) {
758 fs_err(sdp, "can't lookup journal index: %d\n", error);
759 return PTR_ERR(sdp->sd_jindex);
761 ip = GFS2_I(sdp->sd_jindex);
763 /* Load in the journal index special file */
765 error = gfs2_jindex_hold(sdp, &ji_gh);
767 fs_err(sdp, "can't read journal index: %d\n", error);
772 if (!gfs2_jindex_size(sdp)) {
773 fs_err(sdp, "no journals!\n");
777 if (sdp->sd_args.ar_spectator) {
778 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
779 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
781 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
782 fs_err(sdp, "can't mount journal #%u\n",
783 sdp->sd_lockstruct.ls_jid);
784 fs_err(sdp, "there are only %u journals (0 - %u)\n",
785 gfs2_jindex_size(sdp),
786 gfs2_jindex_size(sdp) - 1);
789 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
791 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
793 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
794 &sdp->sd_journal_gh);
796 fs_err(sdp, "can't acquire journal glock: %d\n", error);
800 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
801 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
802 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
805 fs_err(sdp, "can't acquire journal inode glock: %d\n",
807 goto fail_journal_gh;
810 error = gfs2_jdesc_check(sdp->sd_jdesc);
812 fs_err(sdp, "my journal (%u) is bad: %d\n",
813 sdp->sd_jdesc->jd_jid, error);
816 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
818 /* Map the extents for this journal's blocks */
819 map_journal_extents(sdp);
822 if (sdp->sd_lockstruct.ls_first) {
824 for (x = 0; x < sdp->sd_journals; x++) {
825 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
827 fs_err(sdp, "error recovering journal %u: %d\n",
833 gfs2_lm_others_may_mount(sdp);
834 } else if (!sdp->sd_args.ar_spectator) {
835 error = gfs2_recover_journal(sdp->sd_jdesc);
837 fs_err(sdp, "error recovering my journal: %d\n", error);
842 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
843 gfs2_glock_dq_uninit(&ji_gh);
846 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
849 fs_err(sdp, "can't start recoverd thread: %d\n", error);
852 sdp->sd_recoverd_process = p;
857 kthread_stop(sdp->sd_recoverd_process);
859 if (!sdp->sd_args.ar_spectator)
860 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
862 if (!sdp->sd_args.ar_spectator)
863 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
865 gfs2_jindex_free(sdp);
867 gfs2_glock_dq_uninit(&ji_gh);
869 iput(sdp->sd_jindex);
874 static int init_inodes(struct gfs2_sbd *sdp, int undo)
877 struct gfs2_inode *ip;
878 struct inode *master = sdp->sd_master_dir->d_inode;
883 error = init_journal(sdp, undo);
887 /* Read in the master inode number inode */
888 sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum");
889 if (IS_ERR(sdp->sd_inum_inode)) {
890 error = PTR_ERR(sdp->sd_inum_inode);
891 fs_err(sdp, "can't read in inum inode: %d\n", error);
896 /* Read in the master statfs inode */
897 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
898 if (IS_ERR(sdp->sd_statfs_inode)) {
899 error = PTR_ERR(sdp->sd_statfs_inode);
900 fs_err(sdp, "can't read in statfs inode: %d\n", error);
904 /* Read in the resource index inode */
905 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
906 if (IS_ERR(sdp->sd_rindex)) {
907 error = PTR_ERR(sdp->sd_rindex);
908 fs_err(sdp, "can't get resource index inode: %d\n", error);
911 ip = GFS2_I(sdp->sd_rindex);
912 sdp->sd_rindex_uptodate = 0;
914 /* Read in the quota inode */
915 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
916 if (IS_ERR(sdp->sd_quota_inode)) {
917 error = PTR_ERR(sdp->sd_quota_inode);
918 fs_err(sdp, "can't get quota file inode: %d\n", error);
924 iput(sdp->sd_quota_inode);
926 gfs2_clear_rgrpd(sdp);
927 iput(sdp->sd_rindex);
929 iput(sdp->sd_statfs_inode);
931 iput(sdp->sd_inum_inode);
933 init_journal(sdp, UNDO);
938 static int init_per_node(struct gfs2_sbd *sdp, int undo)
940 struct inode *pn = NULL;
943 struct gfs2_inode *ip;
944 struct inode *master = sdp->sd_master_dir->d_inode;
946 if (sdp->sd_args.ar_spectator)
952 pn = gfs2_lookup_simple(master, "per_node");
955 fs_err(sdp, "can't find per_node directory: %d\n", error);
959 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
960 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
961 if (IS_ERR(sdp->sd_ir_inode)) {
962 error = PTR_ERR(sdp->sd_ir_inode);
963 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
967 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
968 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
969 if (IS_ERR(sdp->sd_sc_inode)) {
970 error = PTR_ERR(sdp->sd_sc_inode);
971 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
975 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
976 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
977 if (IS_ERR(sdp->sd_qc_inode)) {
978 error = PTR_ERR(sdp->sd_qc_inode);
979 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
986 ip = GFS2_I(sdp->sd_ir_inode);
987 error = gfs2_glock_nq_init(ip->i_gl,
991 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
995 ip = GFS2_I(sdp->sd_sc_inode);
996 error = gfs2_glock_nq_init(ip->i_gl,
1000 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
1004 ip = GFS2_I(sdp->sd_qc_inode);
1005 error = gfs2_glock_nq_init(ip->i_gl,
1009 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
1016 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
1018 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
1020 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
1022 iput(sdp->sd_qc_inode);
1024 iput(sdp->sd_sc_inode);
1026 iput(sdp->sd_ir_inode);
1033 static int init_threads(struct gfs2_sbd *sdp, int undo)
1035 struct task_struct *p;
1041 sdp->sd_log_flush_time = jiffies;
1042 sdp->sd_jindex_refresh_time = jiffies;
1044 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
1047 fs_err(sdp, "can't start logd thread: %d\n", error);
1050 sdp->sd_logd_process = p;
1052 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
1055 fs_err(sdp, "can't start quotad thread: %d\n", error);
1058 sdp->sd_quotad_process = p;
1064 kthread_stop(sdp->sd_quotad_process);
1066 kthread_stop(sdp->sd_logd_process);
1071 * gfs2_lm_mount - mount a locking protocol
1072 * @sdp: the filesystem
1073 * @args: mount arguements
1074 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1079 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1081 char *proto = sdp->sd_proto_name;
1082 char *table = sdp->sd_table_name;
1083 int flags = LM_MFLAG_CONV_NODROP;
1086 if (sdp->sd_args.ar_spectator)
1087 flags |= LM_MFLAG_SPECTATOR;
1089 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1091 error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata,
1093 GFS2_MIN_LVB_SIZE, flags,
1094 &sdp->sd_lockstruct, &sdp->sd_kobj);
1096 fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n",
1097 proto, table, sdp->sd_args.ar_hostdata);
1101 if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) ||
1102 gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >=
1103 GFS2_MIN_LVB_SIZE)) {
1104 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1108 if (sdp->sd_args.ar_spectator)
1109 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1111 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1112 sdp->sd_lockstruct.ls_jid);
1114 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1116 if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) &&
1117 !sdp->sd_args.ar_ignore_local_fs) {
1118 sdp->sd_args.ar_localflocks = 1;
1119 sdp->sd_args.ar_localcaching = 1;
1126 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1128 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
1129 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1133 * fill_super - Read in superblock
1134 * @sb: The VFS superblock
1135 * @data: Mount options
1136 * @silent: Don't complain if it's not a GFS2 filesystem
1141 static int fill_super(struct super_block *sb, void *data, int silent)
1143 struct gfs2_sbd *sdp;
1144 struct gfs2_holder mount_gh;
1149 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1153 error = gfs2_mount_args(sdp, (char *)data, 0);
1155 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1159 sb->s_magic = GFS2_MAGIC;
1160 sb->s_op = &gfs2_super_ops;
1161 sb->s_export_op = &gfs2_export_ops;
1162 sb->s_time_gran = 1;
1163 sb->s_maxbytes = MAX_LFS_FILESIZE;
1165 /* Set up the buffer cache and fill in some fake block size values
1166 to allow us to read-in the on-disk superblock. */
1167 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1168 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1169 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1170 GFS2_BASIC_BLOCK_SHIFT;
1171 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1173 error = init_names(sdp, silent);
1177 gfs2_create_debugfs_file(sdp);
1179 error = gfs2_sys_fs_add(sdp);
1183 error = gfs2_lm_mount(sdp, silent);
1187 error = init_locking(sdp, &mount_gh, DO);
1191 error = init_sb(sdp, silent);
1195 error = init_inodes(sdp, DO);
1199 error = init_per_node(sdp, DO);
1203 error = gfs2_statfs_init(sdp);
1205 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1209 error = init_threads(sdp, DO);
1213 if (!(sb->s_flags & MS_RDONLY)) {
1214 error = gfs2_make_fs_rw(sdp);
1216 fs_err(sdp, "can't make FS RW: %d\n", error);
1221 gfs2_glock_dq_uninit(&mount_gh);
1226 init_threads(sdp, UNDO);
1228 init_per_node(sdp, UNDO);
1230 init_inodes(sdp, UNDO);
1232 if (sdp->sd_root_dir)
1233 dput(sdp->sd_root_dir);
1234 if (sdp->sd_master_dir)
1235 dput(sdp->sd_master_dir);
1238 init_locking(sdp, &mount_gh, UNDO);
1240 gfs2_gl_hash_clear(sb);
1241 gfs2_lm_unmount(sdp);
1242 while (invalidate_inodes(sb))
1245 gfs2_sys_fs_del(sdp);
1247 gfs2_delete_debugfs_file(sdp);
1249 sb->s_fs_info = NULL;
1253 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1254 const char *dev_name, void *data, struct vfsmount *mnt)
1256 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1259 static struct super_block *get_gfs2_sb(const char *dev_name)
1261 struct super_block *sb;
1262 struct nameidata nd;
1265 error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd);
1267 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1271 sb = nd.path.dentry->d_inode->i_sb;
1272 if (sb && (sb->s_type == &gfs2_fs_type))
1273 atomic_inc(&sb->s_active);
1280 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1281 const char *dev_name, void *data, struct vfsmount *mnt)
1283 struct super_block *sb = NULL;
1284 struct gfs2_sbd *sdp;
1286 sb = get_gfs2_sb(dev_name);
1288 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1291 sdp = sb->s_fs_info;
1293 mnt->mnt_root = dget(sdp->sd_master_dir);
1297 static void gfs2_kill_sb(struct super_block *sb)
1299 struct gfs2_sbd *sdp = sb->s_fs_info;
1302 kill_block_super(sb);
1305 gfs2_meta_syncfs(sdp);
1306 dput(sdp->sd_root_dir);
1307 dput(sdp->sd_master_dir);
1308 sdp->sd_root_dir = NULL;
1309 sdp->sd_master_dir = NULL;
1311 /* Unfreeze the filesystem, if we need to */
1312 mutex_lock(&sdp->sd_freeze_lock);
1313 if (sdp->sd_freeze_count)
1314 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
1315 mutex_unlock(&sdp->sd_freeze_lock);
1317 kthread_stop(sdp->sd_quotad_process);
1318 kthread_stop(sdp->sd_logd_process);
1319 kthread_stop(sdp->sd_recoverd_process);
1321 if (!(sb->s_flags & MS_RDONLY)) {
1322 int error = gfs2_make_fs_ro(sdp);
1327 /* At this point, we're through modifying the disk */
1328 gfs2_jindex_free(sdp);
1329 gfs2_clear_rgrpd(sdp);
1330 iput(sdp->sd_jindex);
1331 iput(sdp->sd_inum_inode);
1332 iput(sdp->sd_statfs_inode);
1333 iput(sdp->sd_rindex);
1334 iput(sdp->sd_quota_inode);
1336 gfs2_glock_put(sdp->sd_rename_gl);
1337 gfs2_glock_put(sdp->sd_trans_gl);
1339 if (!sdp->sd_args.ar_spectator) {
1340 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
1341 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
1342 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
1343 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
1344 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
1345 iput(sdp->sd_ir_inode);
1346 iput(sdp->sd_sc_inode);
1347 iput(sdp->sd_qc_inode);
1349 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
1350 kill_block_super(sb);
1351 gfs2_lm_unmount(sdp);
1352 gfs2_sys_fs_del(sdp);
1353 gfs2_delete_debugfs_file(sdp);
1357 struct file_system_type gfs2_fs_type = {
1359 .fs_flags = FS_REQUIRES_DEV,
1360 .get_sb = gfs2_get_sb,
1361 .kill_sb = gfs2_kill_sb,
1362 .owner = THIS_MODULE,
1365 struct file_system_type gfs2meta_fs_type = {
1367 .fs_flags = FS_REQUIRES_DEV,
1368 .get_sb = gfs2_get_sb_meta,
1369 .owner = THIS_MODULE,