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 spin_lock_init(&sdp->sd_statfs_spin);
89 spin_lock_init(&sdp->sd_rindex_spin);
90 mutex_init(&sdp->sd_rindex_mutex);
91 INIT_LIST_HEAD(&sdp->sd_rindex_list);
92 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
94 INIT_LIST_HEAD(&sdp->sd_jindex_list);
95 spin_lock_init(&sdp->sd_jindex_spin);
96 mutex_init(&sdp->sd_jindex_mutex);
98 INIT_LIST_HEAD(&sdp->sd_quota_list);
99 mutex_init(&sdp->sd_quota_mutex);
100 init_waitqueue_head(&sdp->sd_quota_wait);
101 INIT_LIST_HEAD(&sdp->sd_trunc_list);
102 spin_lock_init(&sdp->sd_trunc_lock);
104 spin_lock_init(&sdp->sd_log_lock);
106 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
107 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
108 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
109 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
110 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
112 mutex_init(&sdp->sd_log_reserve_mutex);
113 INIT_LIST_HEAD(&sdp->sd_ail1_list);
114 INIT_LIST_HEAD(&sdp->sd_ail2_list);
116 init_rwsem(&sdp->sd_log_flush_lock);
117 atomic_set(&sdp->sd_log_in_flight, 0);
118 init_waitqueue_head(&sdp->sd_log_flush_wait);
120 INIT_LIST_HEAD(&sdp->sd_revoke_list);
122 mutex_init(&sdp->sd_freeze_lock);
129 * gfs2_check_sb - Check superblock
130 * @sdp: the filesystem
131 * @sb: The superblock
132 * @silent: Don't print a message if the check fails
134 * Checks the version code of the FS is one that we understand how to
135 * read and that the sizes of the various on-disk structures have not
139 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
143 if (sb->sb_magic != GFS2_MAGIC ||
144 sb->sb_type != GFS2_METATYPE_SB) {
146 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
150 /* If format numbers match exactly, we're done. */
152 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
153 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
156 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
157 for (x = 0; gfs2_old_fs_formats[x]; x++)
158 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
161 if (!gfs2_old_fs_formats[x]) {
163 "GFS2: code version (%u, %u) is incompatible "
164 "with ondisk format (%u, %u)\n",
165 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
166 sb->sb_fs_format, sb->sb_multihost_format);
168 "GFS2: I don't know how to upgrade this FS\n");
173 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
174 for (x = 0; gfs2_old_multihost_formats[x]; x++)
175 if (gfs2_old_multihost_formats[x] ==
176 sb->sb_multihost_format)
179 if (!gfs2_old_multihost_formats[x]) {
181 "GFS2: code version (%u, %u) is incompatible "
182 "with ondisk format (%u, %u)\n",
183 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
184 sb->sb_fs_format, sb->sb_multihost_format);
186 "GFS2: I don't know how to upgrade this FS\n");
191 if (!sdp->sd_args.ar_upgrade) {
193 "GFS2: code version (%u, %u) is incompatible "
194 "with ondisk format (%u, %u)\n",
195 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
196 sb->sb_fs_format, sb->sb_multihost_format);
198 "GFS2: Use the \"upgrade\" mount option to upgrade "
200 printk(KERN_INFO "GFS2: See the manual for more details\n");
207 static void end_bio_io_page(struct bio *bio, int error)
209 struct page *page = bio->bi_private;
212 SetPageUptodate(page);
214 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
218 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
220 const struct gfs2_sb *str = buf;
222 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
223 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
224 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
225 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
226 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
227 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
228 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
229 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
230 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
231 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
232 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
234 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
235 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
236 memcpy(sb->sb_uuid, str->sb_uuid, 16);
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 bio->bi_sector = sector * (sb->s_blocksize >> 9);
275 bio->bi_bdev = sb->s_bdev;
276 bio_add_page(bio, page, PAGE_SIZE, 0);
278 bio->bi_end_io = end_bio_io_page;
279 bio->bi_private = page;
280 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
281 wait_on_page_locked(page);
283 if (!PageUptodate(page)) {
288 gfs2_sb_in(&sdp->sd_sb, p);
295 * gfs2_read_sb - Read super block
296 * @sdp: The GFS2 superblock
297 * @silent: Don't print message if mount fails
301 static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
303 u32 hash_blocks, ind_blocks, leaf_blocks;
308 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
311 fs_err(sdp, "can't read superblock\n");
315 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
319 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
320 GFS2_BASIC_BLOCK_SHIFT;
321 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
322 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
323 sizeof(struct gfs2_dinode)) / sizeof(u64);
324 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
325 sizeof(struct gfs2_meta_header)) / sizeof(u64);
326 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
327 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
328 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
329 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
330 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
331 sizeof(struct gfs2_meta_header)) /
332 sizeof(struct gfs2_quota_change);
334 /* Compute maximum reservation required to add a entry to a directory */
336 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
340 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
341 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
342 ind_blocks += tmp_blocks;
345 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
347 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
349 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
350 sizeof(struct gfs2_dinode);
351 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
356 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
358 m = do_div(d, sdp->sd_inptrs);
360 if (d != sdp->sd_heightsize[x - 1] || m)
362 sdp->sd_heightsize[x] = space;
364 sdp->sd_max_height = x;
365 sdp->sd_heightsize[x] = ~0;
366 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
368 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
369 sizeof(struct gfs2_dinode);
370 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
375 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
377 m = do_div(d, sdp->sd_inptrs);
379 if (d != sdp->sd_jheightsize[x - 1] || m)
381 sdp->sd_jheightsize[x] = space;
383 sdp->sd_max_jheight = x;
384 sdp->sd_jheightsize[x] = ~0;
385 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
390 static int init_names(struct gfs2_sbd *sdp, int silent)
395 proto = sdp->sd_args.ar_lockproto;
396 table = sdp->sd_args.ar_locktable;
398 /* Try to autodetect */
400 if (!proto[0] || !table[0]) {
401 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
405 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
410 proto = sdp->sd_sb.sb_lockproto;
412 table = sdp->sd_sb.sb_locktable;
416 table = sdp->sd_vfs->s_id;
418 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
419 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
421 table = sdp->sd_table_name;
422 while ((table = strchr(table, '/')))
429 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
437 error = gfs2_glock_nq_num(sdp,
438 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
439 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
442 fs_err(sdp, "can't acquire mount glock: %d\n", error);
446 error = gfs2_glock_nq_num(sdp,
447 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
449 LM_FLAG_NOEXP | GL_EXACT,
452 fs_err(sdp, "can't acquire live glock: %d\n", error);
456 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
457 CREATE, &sdp->sd_rename_gl);
459 fs_err(sdp, "can't create rename glock: %d\n", error);
463 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
464 CREATE, &sdp->sd_trans_gl);
466 fs_err(sdp, "can't create transaction glock: %d\n", error);
473 gfs2_glock_put(sdp->sd_trans_gl);
475 gfs2_glock_put(sdp->sd_rename_gl);
477 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
479 gfs2_glock_dq_uninit(mount_gh);
484 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
485 u64 no_addr, const char *name)
487 struct gfs2_sbd *sdp = sb->s_fs_info;
488 struct dentry *dentry;
491 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
493 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
494 return PTR_ERR(inode);
496 dentry = d_alloc_root(inode);
498 fs_err(sdp, "can't alloc %s dentry\n", name);
502 dentry->d_op = &gfs2_dops;
507 static int init_sb(struct gfs2_sbd *sdp, int silent)
509 struct super_block *sb = sdp->sd_vfs;
510 struct gfs2_holder sb_gh;
514 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
515 LM_ST_SHARED, 0, &sb_gh);
517 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
521 ret = gfs2_read_sb(sdp, silent);
523 fs_err(sdp, "can't read superblock: %d\n", ret);
527 /* Set up the buffer cache and SB for real */
528 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
530 fs_err(sdp, "FS block size (%u) is too small for device "
532 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
535 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
537 fs_err(sdp, "FS block size (%u) is too big for machine "
539 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
542 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
544 /* Get the root inode */
545 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
546 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
550 /* Get the master inode */
551 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
552 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
554 dput(sdp->sd_root_dir);
557 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
559 gfs2_glock_dq_uninit(&sb_gh);
564 * map_journal_extents - create a reusable "extent" mapping from all logical
565 * blocks to all physical blocks for the given journal. This will save
566 * us time when writing journal blocks. Most journals will have only one
567 * extent that maps all their logical blocks. That's because gfs2.mkfs
568 * arranges the journal blocks sequentially to maximize performance.
569 * So the extent would map the first block for the entire file length.
570 * However, gfs2_jadd can happen while file activity is happening, so
571 * those journals may not be sequential. Less likely is the case where
572 * the users created their own journals by mounting the metafs and
573 * laying it out. But it's still possible. These journals might have
576 * TODO: This should be done in bigger chunks rather than one block at a time,
577 * but since it's only done at mount time, I'm not worried about the
580 static int map_journal_extents(struct gfs2_sbd *sdp)
582 struct gfs2_jdesc *jd = sdp->sd_jdesc;
584 u64 db, prev_db; /* logical block, disk block, prev disk block */
585 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
586 struct gfs2_journal_extent *jext = NULL;
587 struct buffer_head bh;
592 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
595 bh.b_size = 1 << ip->i_inode.i_blkbits;
596 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
599 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
600 "%u db=%llu\n", rc, lb, (unsigned long long)db);
603 if (!prev_db || db != prev_db + 1) {
604 jext = kzalloc(sizeof(struct gfs2_journal_extent),
607 printk(KERN_INFO "GFS2 error: out of memory "
608 "mapping journal extents.\n");
615 list_add_tail(&jext->extent_list, &jd->extent_list);
624 static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
626 char *message = "FIRSTMOUNT=Done";
627 char *envp[] = { message, NULL };
628 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
629 ls->ls_first_done = 1;
630 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
634 * gfs2_jindex_hold - Grab a lock on the jindex
635 * @sdp: The GFS2 superblock
636 * @ji_gh: the holder for the jindex glock
641 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
643 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
646 struct gfs2_jdesc *jd;
651 mutex_lock(&sdp->sd_jindex_mutex);
654 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
658 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
659 name.hash = gfs2_disk_hash(name.name, name.len);
661 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
662 if (error == -ENOENT) {
667 gfs2_glock_dq_uninit(ji_gh);
673 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
677 INIT_LIST_HEAD(&jd->extent_list);
678 slow_work_init(&jd->jd_work, &gfs2_recover_ops);
679 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
680 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
684 error = PTR_ERR(jd->jd_inode);
689 spin_lock(&sdp->sd_jindex_spin);
690 jd->jd_jid = sdp->sd_journals++;
691 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
692 spin_unlock(&sdp->sd_jindex_spin);
695 mutex_unlock(&sdp->sd_jindex_mutex);
700 static int init_journal(struct gfs2_sbd *sdp, int undo)
702 struct inode *master = sdp->sd_master_dir->d_inode;
703 struct gfs2_holder ji_gh;
704 struct gfs2_inode *ip;
713 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
714 if (IS_ERR(sdp->sd_jindex)) {
715 fs_err(sdp, "can't lookup journal index: %d\n", error);
716 return PTR_ERR(sdp->sd_jindex);
718 ip = GFS2_I(sdp->sd_jindex);
720 /* Load in the journal index special file */
722 error = gfs2_jindex_hold(sdp, &ji_gh);
724 fs_err(sdp, "can't read journal index: %d\n", error);
729 if (!gfs2_jindex_size(sdp)) {
730 fs_err(sdp, "no journals!\n");
734 if (sdp->sd_args.ar_spectator) {
735 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
736 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
738 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
739 fs_err(sdp, "can't mount journal #%u\n",
740 sdp->sd_lockstruct.ls_jid);
741 fs_err(sdp, "there are only %u journals (0 - %u)\n",
742 gfs2_jindex_size(sdp),
743 gfs2_jindex_size(sdp) - 1);
746 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
748 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
750 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
751 &sdp->sd_journal_gh);
753 fs_err(sdp, "can't acquire journal glock: %d\n", error);
757 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
758 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
759 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
762 fs_err(sdp, "can't acquire journal inode glock: %d\n",
764 goto fail_journal_gh;
767 error = gfs2_jdesc_check(sdp->sd_jdesc);
769 fs_err(sdp, "my journal (%u) is bad: %d\n",
770 sdp->sd_jdesc->jd_jid, error);
773 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
775 /* Map the extents for this journal's blocks */
776 map_journal_extents(sdp);
778 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
780 if (sdp->sd_lockstruct.ls_first) {
782 for (x = 0; x < sdp->sd_journals; x++) {
783 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
785 fs_err(sdp, "error recovering journal %u: %d\n",
791 gfs2_others_may_mount(sdp);
792 } else if (!sdp->sd_args.ar_spectator) {
793 error = gfs2_recover_journal(sdp->sd_jdesc);
795 fs_err(sdp, "error recovering my journal: %d\n", error);
800 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
801 gfs2_glock_dq_uninit(&ji_gh);
807 if (!sdp->sd_args.ar_spectator)
808 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
810 if (!sdp->sd_args.ar_spectator)
811 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
813 gfs2_jindex_free(sdp);
815 gfs2_glock_dq_uninit(&ji_gh);
817 iput(sdp->sd_jindex);
822 static int init_inodes(struct gfs2_sbd *sdp, int undo)
825 struct gfs2_inode *ip;
826 struct inode *master = sdp->sd_master_dir->d_inode;
831 error = init_journal(sdp, undo);
835 /* Read in the master statfs inode */
836 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
837 if (IS_ERR(sdp->sd_statfs_inode)) {
838 error = PTR_ERR(sdp->sd_statfs_inode);
839 fs_err(sdp, "can't read in statfs inode: %d\n", error);
843 /* Read in the resource index inode */
844 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
845 if (IS_ERR(sdp->sd_rindex)) {
846 error = PTR_ERR(sdp->sd_rindex);
847 fs_err(sdp, "can't get resource index inode: %d\n", error);
850 ip = GFS2_I(sdp->sd_rindex);
851 sdp->sd_rindex_uptodate = 0;
853 /* Read in the quota inode */
854 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
855 if (IS_ERR(sdp->sd_quota_inode)) {
856 error = PTR_ERR(sdp->sd_quota_inode);
857 fs_err(sdp, "can't get quota file inode: %d\n", error);
863 iput(sdp->sd_quota_inode);
865 gfs2_clear_rgrpd(sdp);
866 iput(sdp->sd_rindex);
868 iput(sdp->sd_statfs_inode);
870 init_journal(sdp, UNDO);
875 static int init_per_node(struct gfs2_sbd *sdp, int undo)
877 struct inode *pn = NULL;
880 struct gfs2_inode *ip;
881 struct inode *master = sdp->sd_master_dir->d_inode;
883 if (sdp->sd_args.ar_spectator)
889 pn = gfs2_lookup_simple(master, "per_node");
892 fs_err(sdp, "can't find per_node directory: %d\n", error);
896 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
897 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
898 if (IS_ERR(sdp->sd_sc_inode)) {
899 error = PTR_ERR(sdp->sd_sc_inode);
900 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
904 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
905 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
906 if (IS_ERR(sdp->sd_qc_inode)) {
907 error = PTR_ERR(sdp->sd_qc_inode);
908 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
915 ip = GFS2_I(sdp->sd_sc_inode);
916 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
919 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
923 ip = GFS2_I(sdp->sd_qc_inode);
924 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
927 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
934 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
936 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
938 iput(sdp->sd_qc_inode);
940 iput(sdp->sd_sc_inode);
947 static int init_threads(struct gfs2_sbd *sdp, int undo)
949 struct task_struct *p;
955 sdp->sd_log_flush_time = jiffies;
957 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
960 fs_err(sdp, "can't start logd thread: %d\n", error);
963 sdp->sd_logd_process = p;
965 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
968 fs_err(sdp, "can't start quotad thread: %d\n", error);
971 sdp->sd_quotad_process = p;
977 kthread_stop(sdp->sd_quotad_process);
979 kthread_stop(sdp->sd_logd_process);
983 static const match_table_t nolock_tokens = {
984 { Opt_jid, "jid=%d\n", },
988 static const struct lm_lockops nolock_ops = {
989 .lm_proto_name = "lock_nolock",
990 .lm_put_lock = kmem_cache_free,
991 .lm_tokens = &nolock_tokens,
995 * gfs2_lm_mount - mount a locking protocol
996 * @sdp: the filesystem
997 * @args: mount arguements
998 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1003 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1005 const struct lm_lockops *lm;
1006 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1007 struct gfs2_args *args = &sdp->sd_args;
1008 const char *proto = sdp->sd_proto_name;
1009 const char *table = sdp->sd_table_name;
1014 if (!strcmp("lock_nolock", proto)) {
1016 sdp->sd_args.ar_localflocks = 1;
1017 sdp->sd_args.ar_localcaching = 1;
1018 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1019 } else if (!strcmp("lock_dlm", proto)) {
1023 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1027 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1032 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1033 substring_t tmp[MAX_OPT_ARGS];
1039 token = match_token(o, *lm->lm_tokens, tmp);
1042 ret = match_int(&tmp[0], &option);
1043 if (ret || option < 0)
1044 goto hostdata_error;
1045 ls->ls_jid = option;
1048 /* Obsolete, but left for backward compat purposes */
1051 ret = match_int(&tmp[0], &option);
1052 if (ret || (option != 0 && option != 1))
1053 goto hostdata_error;
1054 ls->ls_first = option;
1057 ret = match_int(&tmp[0], &option);
1058 if (ret || (option != 0 && option != 1))
1059 goto hostdata_error;
1060 ls->ls_nodir = option;
1065 fs_info(sdp, "unknown hostdata (%s)\n", o);
1070 if (sdp->sd_args.ar_spectator)
1071 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1073 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1074 sdp->sd_lockstruct.ls_jid);
1076 fsname = strchr(table, ':');
1079 if (lm->lm_mount == NULL) {
1080 fs_info(sdp, "Now mounting FS...\n");
1083 ret = lm->lm_mount(sdp, fsname);
1085 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1089 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1091 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1092 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1094 lm->lm_unmount(sdp);
1097 void gfs2_online_uevent(struct gfs2_sbd *sdp)
1099 struct super_block *sb = sdp->sd_vfs;
1102 char *envp[] = { ro, spectator, NULL };
1103 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1104 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1105 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1109 * fill_super - Read in superblock
1110 * @sb: The VFS superblock
1111 * @data: Mount options
1112 * @silent: Don't complain if it's not a GFS2 filesystem
1117 static int fill_super(struct super_block *sb, void *data, int silent)
1119 struct gfs2_sbd *sdp;
1120 struct gfs2_holder mount_gh;
1125 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1129 sdp->sd_args.ar_quota = GFS2_QUOTA_DEFAULT;
1130 sdp->sd_args.ar_data = GFS2_DATA_DEFAULT;
1131 sdp->sd_args.ar_commit = 60;
1132 sdp->sd_args.ar_errors = GFS2_ERRORS_DEFAULT;
1134 error = gfs2_mount_args(sdp, &sdp->sd_args, data);
1136 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1140 if (sdp->sd_args.ar_spectator) {
1141 sb->s_flags |= MS_RDONLY;
1142 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1144 if (sdp->sd_args.ar_posix_acl)
1145 sb->s_flags |= MS_POSIXACL;
1147 sb->s_magic = GFS2_MAGIC;
1148 sb->s_op = &gfs2_super_ops;
1149 sb->s_export_op = &gfs2_export_ops;
1150 sb->s_xattr = gfs2_xattr_handlers;
1151 sb->s_time_gran = 1;
1152 sb->s_maxbytes = MAX_LFS_FILESIZE;
1154 /* Set up the buffer cache and fill in some fake block size values
1155 to allow us to read-in the on-disk superblock. */
1156 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1157 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1158 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1159 GFS2_BASIC_BLOCK_SHIFT;
1160 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1162 sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit;
1164 error = init_names(sdp, silent);
1168 gfs2_create_debugfs_file(sdp);
1170 error = gfs2_sys_fs_add(sdp);
1174 error = gfs2_lm_mount(sdp, silent);
1178 error = init_locking(sdp, &mount_gh, DO);
1182 error = init_sb(sdp, silent);
1186 error = init_inodes(sdp, DO);
1190 error = init_per_node(sdp, DO);
1194 error = gfs2_statfs_init(sdp);
1196 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1200 error = init_threads(sdp, DO);
1204 if (!(sb->s_flags & MS_RDONLY)) {
1205 error = gfs2_make_fs_rw(sdp);
1207 fs_err(sdp, "can't make FS RW: %d\n", error);
1212 gfs2_glock_dq_uninit(&mount_gh);
1213 gfs2_online_uevent(sdp);
1217 init_threads(sdp, UNDO);
1219 init_per_node(sdp, UNDO);
1221 init_inodes(sdp, UNDO);
1223 if (sdp->sd_root_dir)
1224 dput(sdp->sd_root_dir);
1225 if (sdp->sd_master_dir)
1226 dput(sdp->sd_master_dir);
1231 init_locking(sdp, &mount_gh, UNDO);
1233 gfs2_gl_hash_clear(sdp);
1234 gfs2_lm_unmount(sdp);
1235 while (invalidate_inodes(sb))
1238 gfs2_sys_fs_del(sdp);
1240 gfs2_delete_debugfs_file(sdp);
1242 sb->s_fs_info = NULL;
1246 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1247 const char *dev_name, void *data, struct vfsmount *mnt)
1249 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1252 static int test_meta_super(struct super_block *s, void *ptr)
1254 struct block_device *bdev = ptr;
1255 return (bdev == s->s_bdev);
1258 static int set_meta_super(struct super_block *s, void *ptr)
1263 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1264 const char *dev_name, void *data, struct vfsmount *mnt)
1266 struct super_block *s;
1267 struct gfs2_sbd *sdp;
1271 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
1273 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1277 s = sget(&gfs2_fs_type, test_meta_super, set_meta_super,
1278 path.dentry->d_inode->i_sb->s_bdev);
1281 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1286 mnt->mnt_root = dget(sdp->sd_master_dir);
1290 static void gfs2_kill_sb(struct super_block *sb)
1292 struct gfs2_sbd *sdp = sb->s_fs_info;
1295 kill_block_super(sb);
1299 gfs2_meta_syncfs(sdp);
1300 dput(sdp->sd_root_dir);
1301 dput(sdp->sd_master_dir);
1302 sdp->sd_root_dir = NULL;
1303 sdp->sd_master_dir = NULL;
1304 shrink_dcache_sb(sb);
1305 kill_block_super(sb);
1306 gfs2_delete_debugfs_file(sdp);
1310 struct file_system_type gfs2_fs_type = {
1312 .fs_flags = FS_REQUIRES_DEV,
1313 .get_sb = gfs2_get_sb,
1314 .kill_sb = gfs2_kill_sb,
1315 .owner = THIS_MODULE,
1318 struct file_system_type gfs2meta_fs_type = {
1320 .fs_flags = FS_REQUIRES_DEV,
1321 .get_sb = gfs2_get_sb_meta,
1322 .owner = THIS_MODULE,