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>
21 #include <linux/quotaops.h>
37 #include "trace_gfs2.h"
42 static const u32 gfs2_old_fs_formats[] = {
46 static const u32 gfs2_old_multihost_formats[] = {
51 * gfs2_tune_init - Fill a gfs2_tune structure with default values
56 static void gfs2_tune_init(struct gfs2_tune *gt)
58 spin_lock_init(>->gt_spin);
60 gt->gt_incore_log_blocks = 1024;
62 gt->gt_quota_simul_sync = 64;
63 gt->gt_quota_warn_period = 10;
64 gt->gt_quota_scale_num = 1;
65 gt->gt_quota_scale_den = 1;
66 gt->gt_new_files_jdata = 0;
67 gt->gt_max_readahead = 1 << 18;
68 gt->gt_complain_secs = 10;
71 static struct gfs2_sbd *init_sbd(struct super_block *sb)
75 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
82 gfs2_tune_init(&sdp->sd_tune);
84 init_waitqueue_head(&sdp->sd_glock_wait);
85 atomic_set(&sdp->sd_glock_disposal, 0);
86 spin_lock_init(&sdp->sd_statfs_spin);
88 spin_lock_init(&sdp->sd_rindex_spin);
89 mutex_init(&sdp->sd_rindex_mutex);
90 INIT_LIST_HEAD(&sdp->sd_rindex_list);
91 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
93 INIT_LIST_HEAD(&sdp->sd_jindex_list);
94 spin_lock_init(&sdp->sd_jindex_spin);
95 mutex_init(&sdp->sd_jindex_mutex);
97 INIT_LIST_HEAD(&sdp->sd_quota_list);
98 mutex_init(&sdp->sd_quota_mutex);
99 init_waitqueue_head(&sdp->sd_quota_wait);
100 INIT_LIST_HEAD(&sdp->sd_trunc_list);
101 spin_lock_init(&sdp->sd_trunc_lock);
103 spin_lock_init(&sdp->sd_log_lock);
105 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
106 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
107 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
108 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
109 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
111 mutex_init(&sdp->sd_log_reserve_mutex);
112 INIT_LIST_HEAD(&sdp->sd_ail1_list);
113 INIT_LIST_HEAD(&sdp->sd_ail2_list);
115 init_rwsem(&sdp->sd_log_flush_lock);
116 atomic_set(&sdp->sd_log_in_flight, 0);
117 init_waitqueue_head(&sdp->sd_log_flush_wait);
119 INIT_LIST_HEAD(&sdp->sd_revoke_list);
121 mutex_init(&sdp->sd_freeze_lock);
128 * gfs2_check_sb - Check superblock
129 * @sdp: the filesystem
130 * @sb: The superblock
131 * @silent: Don't print a message if the check fails
133 * Checks the version code of the FS is one that we understand how to
134 * read and that the sizes of the various on-disk structures have not
138 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
142 if (sb->sb_magic != GFS2_MAGIC ||
143 sb->sb_type != GFS2_METATYPE_SB) {
145 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
149 /* If format numbers match exactly, we're done. */
151 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
152 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
155 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
156 for (x = 0; gfs2_old_fs_formats[x]; x++)
157 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
160 if (!gfs2_old_fs_formats[x]) {
162 "GFS2: code version (%u, %u) is incompatible "
163 "with ondisk format (%u, %u)\n",
164 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
165 sb->sb_fs_format, sb->sb_multihost_format);
167 "GFS2: I don't know how to upgrade this FS\n");
172 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
173 for (x = 0; gfs2_old_multihost_formats[x]; x++)
174 if (gfs2_old_multihost_formats[x] ==
175 sb->sb_multihost_format)
178 if (!gfs2_old_multihost_formats[x]) {
180 "GFS2: code version (%u, %u) is incompatible "
181 "with ondisk format (%u, %u)\n",
182 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
183 sb->sb_fs_format, sb->sb_multihost_format);
185 "GFS2: I don't know how to upgrade this FS\n");
190 if (!sdp->sd_args.ar_upgrade) {
192 "GFS2: code version (%u, %u) is incompatible "
193 "with ondisk format (%u, %u)\n",
194 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
195 sb->sb_fs_format, sb->sb_multihost_format);
197 "GFS2: Use the \"upgrade\" mount option to upgrade "
199 printk(KERN_INFO "GFS2: See the manual for more details\n");
206 static void end_bio_io_page(struct bio *bio, int error)
208 struct page *page = bio->bi_private;
211 SetPageUptodate(page);
213 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
217 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
219 const struct gfs2_sb *str = buf;
221 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
222 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
223 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
224 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
225 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
226 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
227 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
228 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
229 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
230 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
231 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
233 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
234 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
235 memcpy(sb->sb_uuid, str->sb_uuid, 16);
239 * gfs2_read_super - Read the gfs2 super block from disk
240 * @sdp: The GFS2 super block
241 * @sector: The location of the super block
242 * @error: The error code to return
244 * This uses the bio functions to read the super block from disk
245 * because we want to be 100% sure that we never read cached data.
246 * A super block is read twice only during each GFS2 mount and is
247 * never written to by the filesystem. The first time its read no
248 * locks are held, and the only details which are looked at are those
249 * relating to the locking protocol. Once locking is up and working,
250 * the sb is read again under the lock to establish the location of
251 * the master directory (contains pointers to journals etc) and the
254 * Returns: 0 on success or error
257 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
259 struct super_block *sb = sdp->sd_vfs;
264 page = alloc_page(GFP_NOFS);
268 ClearPageUptodate(page);
269 ClearPageDirty(page);
272 bio = bio_alloc(GFP_NOFS, 1);
273 bio->bi_sector = sector * (sb->s_blocksize >> 9);
274 bio->bi_bdev = sb->s_bdev;
275 bio_add_page(bio, page, PAGE_SIZE, 0);
277 bio->bi_end_io = end_bio_io_page;
278 bio->bi_private = page;
279 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
280 wait_on_page_locked(page);
282 if (!PageUptodate(page)) {
287 gfs2_sb_in(&sdp->sd_sb, p);
294 * gfs2_read_sb - Read super block
295 * @sdp: The GFS2 superblock
296 * @silent: Don't print message if mount fails
300 static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
302 u32 hash_blocks, ind_blocks, leaf_blocks;
307 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
310 fs_err(sdp, "can't read superblock\n");
314 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
318 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
319 GFS2_BASIC_BLOCK_SHIFT;
320 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
321 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
322 sizeof(struct gfs2_dinode)) / sizeof(u64);
323 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
324 sizeof(struct gfs2_meta_header)) / sizeof(u64);
325 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
326 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
327 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
328 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
329 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
330 sizeof(struct gfs2_meta_header)) /
331 sizeof(struct gfs2_quota_change);
333 /* Compute maximum reservation required to add a entry to a directory */
335 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
339 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
340 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
341 ind_blocks += tmp_blocks;
344 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
346 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
348 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
349 sizeof(struct gfs2_dinode);
350 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
355 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
357 m = do_div(d, sdp->sd_inptrs);
359 if (d != sdp->sd_heightsize[x - 1] || m)
361 sdp->sd_heightsize[x] = space;
363 sdp->sd_max_height = x;
364 sdp->sd_heightsize[x] = ~0;
365 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
367 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
368 sizeof(struct gfs2_dinode);
369 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
374 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
376 m = do_div(d, sdp->sd_inptrs);
378 if (d != sdp->sd_jheightsize[x - 1] || m)
380 sdp->sd_jheightsize[x] = space;
382 sdp->sd_max_jheight = x;
383 sdp->sd_jheightsize[x] = ~0;
384 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
389 static int init_names(struct gfs2_sbd *sdp, int silent)
394 proto = sdp->sd_args.ar_lockproto;
395 table = sdp->sd_args.ar_locktable;
397 /* Try to autodetect */
399 if (!proto[0] || !table[0]) {
400 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
404 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
409 proto = sdp->sd_sb.sb_lockproto;
411 table = sdp->sd_sb.sb_locktable;
415 table = sdp->sd_vfs->s_id;
417 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
418 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
420 table = sdp->sd_table_name;
421 while ((table = strchr(table, '/')))
428 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
436 error = gfs2_glock_nq_num(sdp,
437 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
438 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
441 fs_err(sdp, "can't acquire mount glock: %d\n", error);
445 error = gfs2_glock_nq_num(sdp,
446 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
448 LM_FLAG_NOEXP | GL_EXACT,
451 fs_err(sdp, "can't acquire live glock: %d\n", error);
455 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
456 CREATE, &sdp->sd_rename_gl);
458 fs_err(sdp, "can't create rename glock: %d\n", error);
462 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
463 CREATE, &sdp->sd_trans_gl);
465 fs_err(sdp, "can't create transaction glock: %d\n", error);
472 gfs2_glock_put(sdp->sd_trans_gl);
474 gfs2_glock_put(sdp->sd_rename_gl);
476 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
478 gfs2_glock_dq_uninit(mount_gh);
483 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
484 u64 no_addr, const char *name)
486 struct gfs2_sbd *sdp = sb->s_fs_info;
487 struct dentry *dentry;
490 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
492 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
493 return PTR_ERR(inode);
495 dentry = d_alloc_root(inode);
497 fs_err(sdp, "can't alloc %s dentry\n", name);
501 dentry->d_op = &gfs2_dops;
506 static int init_sb(struct gfs2_sbd *sdp, int silent)
508 struct super_block *sb = sdp->sd_vfs;
509 struct gfs2_holder sb_gh;
513 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
514 LM_ST_SHARED, 0, &sb_gh);
516 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
520 ret = gfs2_read_sb(sdp, silent);
522 fs_err(sdp, "can't read superblock: %d\n", ret);
526 /* Set up the buffer cache and SB for real */
527 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
529 fs_err(sdp, "FS block size (%u) is too small for device "
531 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
534 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
536 fs_err(sdp, "FS block size (%u) is too big for machine "
538 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
541 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
543 /* Get the root inode */
544 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
545 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
549 /* Get the master inode */
550 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
551 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
553 dput(sdp->sd_root_dir);
556 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
558 gfs2_glock_dq_uninit(&sb_gh);
563 * map_journal_extents - create a reusable "extent" mapping from all logical
564 * blocks to all physical blocks for the given journal. This will save
565 * us time when writing journal blocks. Most journals will have only one
566 * extent that maps all their logical blocks. That's because gfs2.mkfs
567 * arranges the journal blocks sequentially to maximize performance.
568 * So the extent would map the first block for the entire file length.
569 * However, gfs2_jadd can happen while file activity is happening, so
570 * those journals may not be sequential. Less likely is the case where
571 * the users created their own journals by mounting the metafs and
572 * laying it out. But it's still possible. These journals might have
575 * TODO: This should be done in bigger chunks rather than one block at a time,
576 * but since it's only done at mount time, I'm not worried about the
579 static int map_journal_extents(struct gfs2_sbd *sdp)
581 struct gfs2_jdesc *jd = sdp->sd_jdesc;
583 u64 db, prev_db; /* logical block, disk block, prev disk block */
584 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
585 struct gfs2_journal_extent *jext = NULL;
586 struct buffer_head bh;
591 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
594 bh.b_size = 1 << ip->i_inode.i_blkbits;
595 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
598 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
599 "%u db=%llu\n", rc, lb, (unsigned long long)db);
602 if (!prev_db || db != prev_db + 1) {
603 jext = kzalloc(sizeof(struct gfs2_journal_extent),
606 printk(KERN_INFO "GFS2 error: out of memory "
607 "mapping journal extents.\n");
614 list_add_tail(&jext->extent_list, &jd->extent_list);
623 static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
625 char *message = "FIRSTMOUNT=Done";
626 char *envp[] = { message, NULL };
627 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
628 ls->ls_first_done = 1;
629 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
633 * gfs2_jindex_hold - Grab a lock on the jindex
634 * @sdp: The GFS2 superblock
635 * @ji_gh: the holder for the jindex glock
640 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
642 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
645 struct gfs2_jdesc *jd;
650 mutex_lock(&sdp->sd_jindex_mutex);
653 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
657 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
658 name.hash = gfs2_disk_hash(name.name, name.len);
660 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
661 if (error == -ENOENT) {
666 gfs2_glock_dq_uninit(ji_gh);
672 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
676 INIT_LIST_HEAD(&jd->extent_list);
677 slow_work_init(&jd->jd_work, &gfs2_recover_ops);
678 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
679 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
683 error = PTR_ERR(jd->jd_inode);
688 spin_lock(&sdp->sd_jindex_spin);
689 jd->jd_jid = sdp->sd_journals++;
690 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
691 spin_unlock(&sdp->sd_jindex_spin);
694 mutex_unlock(&sdp->sd_jindex_mutex);
699 static int init_journal(struct gfs2_sbd *sdp, int undo)
701 struct inode *master = sdp->sd_master_dir->d_inode;
702 struct gfs2_holder ji_gh;
703 struct gfs2_inode *ip;
712 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
713 if (IS_ERR(sdp->sd_jindex)) {
714 fs_err(sdp, "can't lookup journal index: %d\n", error);
715 return PTR_ERR(sdp->sd_jindex);
717 ip = GFS2_I(sdp->sd_jindex);
719 /* Load in the journal index special file */
721 error = gfs2_jindex_hold(sdp, &ji_gh);
723 fs_err(sdp, "can't read journal index: %d\n", error);
728 if (!gfs2_jindex_size(sdp)) {
729 fs_err(sdp, "no journals!\n");
733 if (sdp->sd_args.ar_spectator) {
734 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
735 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
737 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
738 fs_err(sdp, "can't mount journal #%u\n",
739 sdp->sd_lockstruct.ls_jid);
740 fs_err(sdp, "there are only %u journals (0 - %u)\n",
741 gfs2_jindex_size(sdp),
742 gfs2_jindex_size(sdp) - 1);
745 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
747 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
749 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
750 &sdp->sd_journal_gh);
752 fs_err(sdp, "can't acquire journal glock: %d\n", error);
756 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
757 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
758 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
761 fs_err(sdp, "can't acquire journal inode glock: %d\n",
763 goto fail_journal_gh;
766 error = gfs2_jdesc_check(sdp->sd_jdesc);
768 fs_err(sdp, "my journal (%u) is bad: %d\n",
769 sdp->sd_jdesc->jd_jid, error);
772 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
774 /* Map the extents for this journal's blocks */
775 map_journal_extents(sdp);
777 trace_gfs2_log_blocks(sdp, atomic_read(&sdp->sd_log_blks_free));
779 if (sdp->sd_lockstruct.ls_first) {
781 for (x = 0; x < sdp->sd_journals; x++) {
782 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
784 fs_err(sdp, "error recovering journal %u: %d\n",
790 gfs2_others_may_mount(sdp);
791 } else if (!sdp->sd_args.ar_spectator) {
792 error = gfs2_recover_journal(sdp->sd_jdesc);
794 fs_err(sdp, "error recovering my journal: %d\n", error);
799 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
800 gfs2_glock_dq_uninit(&ji_gh);
806 if (!sdp->sd_args.ar_spectator)
807 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
809 if (!sdp->sd_args.ar_spectator)
810 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
812 gfs2_jindex_free(sdp);
814 gfs2_glock_dq_uninit(&ji_gh);
816 iput(sdp->sd_jindex);
821 static int init_inodes(struct gfs2_sbd *sdp, int undo)
824 struct gfs2_inode *ip;
825 struct inode *master = sdp->sd_master_dir->d_inode;
830 error = init_journal(sdp, undo);
834 /* Read in the master statfs inode */
835 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
836 if (IS_ERR(sdp->sd_statfs_inode)) {
837 error = PTR_ERR(sdp->sd_statfs_inode);
838 fs_err(sdp, "can't read in statfs inode: %d\n", error);
842 /* Read in the resource index inode */
843 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
844 if (IS_ERR(sdp->sd_rindex)) {
845 error = PTR_ERR(sdp->sd_rindex);
846 fs_err(sdp, "can't get resource index inode: %d\n", error);
849 ip = GFS2_I(sdp->sd_rindex);
850 sdp->sd_rindex_uptodate = 0;
852 /* Read in the quota inode */
853 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
854 if (IS_ERR(sdp->sd_quota_inode)) {
855 error = PTR_ERR(sdp->sd_quota_inode);
856 fs_err(sdp, "can't get quota file inode: %d\n", error);
862 iput(sdp->sd_quota_inode);
864 gfs2_clear_rgrpd(sdp);
865 iput(sdp->sd_rindex);
867 iput(sdp->sd_statfs_inode);
869 init_journal(sdp, UNDO);
874 static int init_per_node(struct gfs2_sbd *sdp, int undo)
876 struct inode *pn = NULL;
879 struct gfs2_inode *ip;
880 struct inode *master = sdp->sd_master_dir->d_inode;
882 if (sdp->sd_args.ar_spectator)
888 pn = gfs2_lookup_simple(master, "per_node");
891 fs_err(sdp, "can't find per_node directory: %d\n", error);
895 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
896 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
897 if (IS_ERR(sdp->sd_sc_inode)) {
898 error = PTR_ERR(sdp->sd_sc_inode);
899 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
903 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
904 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
905 if (IS_ERR(sdp->sd_qc_inode)) {
906 error = PTR_ERR(sdp->sd_qc_inode);
907 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
914 ip = GFS2_I(sdp->sd_sc_inode);
915 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
918 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
922 ip = GFS2_I(sdp->sd_qc_inode);
923 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
926 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
933 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
935 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
937 iput(sdp->sd_qc_inode);
939 iput(sdp->sd_sc_inode);
946 static int init_threads(struct gfs2_sbd *sdp, int undo)
948 struct task_struct *p;
954 sdp->sd_log_flush_time = jiffies;
956 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
959 fs_err(sdp, "can't start logd thread: %d\n", error);
962 sdp->sd_logd_process = p;
964 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
967 fs_err(sdp, "can't start quotad thread: %d\n", error);
970 sdp->sd_quotad_process = p;
976 kthread_stop(sdp->sd_quotad_process);
978 kthread_stop(sdp->sd_logd_process);
982 static const match_table_t nolock_tokens = {
983 { Opt_jid, "jid=%d\n", },
987 static void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
989 struct gfs2_sbd *sdp = gl->gl_sbd;
990 kmem_cache_free(cachep, gl);
991 if (atomic_dec_and_test(&sdp->sd_glock_disposal))
992 wake_up(&sdp->sd_glock_wait);
995 static const struct lm_lockops nolock_ops = {
996 .lm_proto_name = "lock_nolock",
997 .lm_put_lock = nolock_put_lock,
998 .lm_tokens = &nolock_tokens,
1002 * gfs2_lm_mount - mount a locking protocol
1003 * @sdp: the filesystem
1004 * @args: mount arguments
1005 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1010 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1012 const struct lm_lockops *lm;
1013 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1014 struct gfs2_args *args = &sdp->sd_args;
1015 const char *proto = sdp->sd_proto_name;
1016 const char *table = sdp->sd_table_name;
1021 if (!strcmp("lock_nolock", proto)) {
1023 sdp->sd_args.ar_localflocks = 1;
1024 sdp->sd_args.ar_localcaching = 1;
1025 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1026 } else if (!strcmp("lock_dlm", proto)) {
1030 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1034 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1039 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1040 substring_t tmp[MAX_OPT_ARGS];
1046 token = match_token(o, *lm->lm_tokens, tmp);
1049 ret = match_int(&tmp[0], &option);
1050 if (ret || option < 0)
1051 goto hostdata_error;
1052 ls->ls_jid = option;
1055 /* Obsolete, but left for backward compat purposes */
1058 ret = match_int(&tmp[0], &option);
1059 if (ret || (option != 0 && option != 1))
1060 goto hostdata_error;
1061 ls->ls_first = option;
1064 ret = match_int(&tmp[0], &option);
1065 if (ret || (option != 0 && option != 1))
1066 goto hostdata_error;
1067 ls->ls_nodir = option;
1072 fs_info(sdp, "unknown hostdata (%s)\n", o);
1077 if (sdp->sd_args.ar_spectator)
1078 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1080 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1081 sdp->sd_lockstruct.ls_jid);
1083 fsname = strchr(table, ':');
1086 if (lm->lm_mount == NULL) {
1087 fs_info(sdp, "Now mounting FS...\n");
1090 ret = lm->lm_mount(sdp, fsname);
1092 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1096 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1098 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1099 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1101 lm->lm_unmount(sdp);
1104 void gfs2_online_uevent(struct gfs2_sbd *sdp)
1106 struct super_block *sb = sdp->sd_vfs;
1109 char *envp[] = { ro, spectator, NULL };
1110 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1111 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1112 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1116 * fill_super - Read in superblock
1117 * @sb: The VFS superblock
1118 * @data: Mount options
1119 * @silent: Don't complain if it's not a GFS2 filesystem
1124 static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
1126 struct gfs2_sbd *sdp;
1127 struct gfs2_holder mount_gh;
1132 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1135 sdp->sd_args = *args;
1137 if (sdp->sd_args.ar_spectator) {
1138 sb->s_flags |= MS_RDONLY;
1139 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1141 if (sdp->sd_args.ar_posix_acl)
1142 sb->s_flags |= MS_POSIXACL;
1143 if (sdp->sd_args.ar_nobarrier)
1144 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1146 sb->s_magic = GFS2_MAGIC;
1147 sb->s_op = &gfs2_super_ops;
1148 sb->s_export_op = &gfs2_export_ops;
1149 sb->s_xattr = gfs2_xattr_handlers;
1150 sb->s_qcop = &gfs2_quotactl_ops;
1151 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1152 sb->s_time_gran = 1;
1153 sb->s_maxbytes = MAX_LFS_FILESIZE;
1155 /* Set up the buffer cache and fill in some fake block size values
1156 to allow us to read-in the on-disk superblock. */
1157 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1158 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1159 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1160 GFS2_BASIC_BLOCK_SHIFT;
1161 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1163 sdp->sd_tune.gt_log_flush_secs = sdp->sd_args.ar_commit;
1164 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1165 if (sdp->sd_args.ar_statfs_quantum) {
1166 sdp->sd_tune.gt_statfs_slow = 0;
1167 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
1170 sdp->sd_tune.gt_statfs_slow = 1;
1171 sdp->sd_tune.gt_statfs_quantum = 30;
1174 error = init_names(sdp, silent);
1178 gfs2_create_debugfs_file(sdp);
1180 error = gfs2_sys_fs_add(sdp);
1184 error = gfs2_lm_mount(sdp, silent);
1188 error = init_locking(sdp, &mount_gh, DO);
1192 error = init_sb(sdp, silent);
1196 error = init_inodes(sdp, DO);
1200 error = init_per_node(sdp, DO);
1204 error = gfs2_statfs_init(sdp);
1206 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1210 error = init_threads(sdp, DO);
1214 if (!(sb->s_flags & MS_RDONLY)) {
1215 error = gfs2_make_fs_rw(sdp);
1217 fs_err(sdp, "can't make FS RW: %d\n", error);
1222 gfs2_glock_dq_uninit(&mount_gh);
1223 gfs2_online_uevent(sdp);
1227 init_threads(sdp, UNDO);
1229 init_per_node(sdp, UNDO);
1231 init_inodes(sdp, UNDO);
1233 if (sdp->sd_root_dir)
1234 dput(sdp->sd_root_dir);
1235 if (sdp->sd_master_dir)
1236 dput(sdp->sd_master_dir);
1241 init_locking(sdp, &mount_gh, UNDO);
1243 invalidate_inodes(sb);
1244 gfs2_gl_hash_clear(sdp);
1245 gfs2_lm_unmount(sdp);
1247 gfs2_sys_fs_del(sdp);
1249 gfs2_delete_debugfs_file(sdp);
1251 sb->s_fs_info = NULL;
1255 static int set_gfs2_super(struct super_block *s, void *data)
1258 s->s_dev = s->s_bdev->bd_dev;
1261 * We set the bdi here to the queue backing, file systems can
1262 * overwrite this in ->fill_super()
1264 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1268 static int test_gfs2_super(struct super_block *s, void *ptr)
1270 struct block_device *bdev = ptr;
1271 return (bdev == s->s_bdev);
1275 * gfs2_get_sb - Get the GFS2 superblock
1276 * @fs_type: The GFS2 filesystem type
1277 * @flags: Mount flags
1278 * @dev_name: The name of the device
1279 * @data: The mount arguments
1280 * @mnt: The vfsmnt for this mount
1282 * Q. Why not use get_sb_bdev() ?
1283 * A. We need to select one of two root directories to mount, independent
1284 * of whether this is the initial, or subsequent, mount of this sb
1286 * Returns: 0 or -ve on error
1289 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1290 const char *dev_name, void *data, struct vfsmount *mnt)
1292 struct block_device *bdev;
1293 struct super_block *s;
1294 fmode_t mode = FMODE_READ;
1296 struct gfs2_args args;
1297 struct gfs2_sbd *sdp;
1299 if (!(flags & MS_RDONLY))
1300 mode |= FMODE_WRITE;
1302 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1304 return PTR_ERR(bdev);
1307 * once the super is inserted into the list by sget, s_umount
1308 * will protect the lockfs code from trying to start a snapshot
1309 * while we are mounting
1311 mutex_lock(&bdev->bd_fsfreeze_mutex);
1312 if (bdev->bd_fsfreeze_count > 0) {
1313 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1317 s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev);
1318 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1323 memset(&args, 0, sizeof(args));
1324 args.ar_quota = GFS2_QUOTA_DEFAULT;
1325 args.ar_data = GFS2_DATA_DEFAULT;
1326 args.ar_commit = 60;
1327 args.ar_statfs_quantum = 30;
1328 args.ar_quota_quantum = 60;
1329 args.ar_errors = GFS2_ERRORS_DEFAULT;
1331 error = gfs2_mount_args(&args, data);
1333 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1336 deactivate_locked_super(s);
1342 if ((flags ^ s->s_flags) & MS_RDONLY)
1344 close_bdev_exclusive(bdev, mode);
1346 char b[BDEVNAME_SIZE];
1350 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1351 sb_set_blocksize(s, block_size(bdev));
1352 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
1354 deactivate_locked_super(s);
1357 s->s_flags |= MS_ACTIVE;
1364 mnt->mnt_root = dget(sdp->sd_master_dir);
1366 mnt->mnt_root = dget(sdp->sd_root_dir);
1370 deactivate_locked_super(s);
1372 close_bdev_exclusive(bdev, mode);
1376 static int set_meta_super(struct super_block *s, void *ptr)
1381 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1382 const char *dev_name, void *data, struct vfsmount *mnt)
1384 struct super_block *s;
1385 struct gfs2_sbd *sdp;
1389 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
1391 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1395 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super,
1396 path.dentry->d_inode->i_sb->s_bdev);
1399 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1402 if ((flags ^ s->s_flags) & MS_RDONLY) {
1403 deactivate_locked_super(s);
1408 mnt->mnt_root = dget(sdp->sd_master_dir);
1412 static void gfs2_kill_sb(struct super_block *sb)
1414 struct gfs2_sbd *sdp = sb->s_fs_info;
1417 kill_block_super(sb);
1421 gfs2_meta_syncfs(sdp);
1422 dput(sdp->sd_root_dir);
1423 dput(sdp->sd_master_dir);
1424 sdp->sd_root_dir = NULL;
1425 sdp->sd_master_dir = NULL;
1426 shrink_dcache_sb(sb);
1427 kill_block_super(sb);
1428 gfs2_delete_debugfs_file(sdp);
1432 struct file_system_type gfs2_fs_type = {
1434 .fs_flags = FS_REQUIRES_DEV,
1435 .get_sb = gfs2_get_sb,
1436 .kill_sb = gfs2_kill_sb,
1437 .owner = THIS_MODULE,
1440 struct file_system_type gfs2meta_fs_type = {
1442 .fs_flags = FS_REQUIRES_DEV,
1443 .get_sb = gfs2_get_sb_meta,
1444 .owner = THIS_MODULE,