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/quotaops.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_quota_simul_sync = 64;
60 gt->gt_quota_warn_period = 10;
61 gt->gt_quota_scale_num = 1;
62 gt->gt_quota_scale_den = 1;
63 gt->gt_new_files_jdata = 0;
64 gt->gt_max_readahead = 1 << 18;
65 gt->gt_complain_secs = 10;
68 static struct gfs2_sbd *init_sbd(struct super_block *sb)
72 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
78 set_bit(SDF_NOJOURNALID, &sdp->sd_flags);
79 gfs2_tune_init(&sdp->sd_tune);
81 init_waitqueue_head(&sdp->sd_glock_wait);
82 atomic_set(&sdp->sd_glock_disposal, 0);
83 spin_lock_init(&sdp->sd_statfs_spin);
85 spin_lock_init(&sdp->sd_rindex_spin);
86 mutex_init(&sdp->sd_rindex_mutex);
87 INIT_LIST_HEAD(&sdp->sd_rindex_list);
88 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
90 INIT_LIST_HEAD(&sdp->sd_jindex_list);
91 spin_lock_init(&sdp->sd_jindex_spin);
92 mutex_init(&sdp->sd_jindex_mutex);
94 INIT_LIST_HEAD(&sdp->sd_quota_list);
95 mutex_init(&sdp->sd_quota_mutex);
96 init_waitqueue_head(&sdp->sd_quota_wait);
97 INIT_LIST_HEAD(&sdp->sd_trunc_list);
98 spin_lock_init(&sdp->sd_trunc_lock);
100 spin_lock_init(&sdp->sd_log_lock);
101 atomic_set(&sdp->sd_log_pinned, 0);
102 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
103 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
104 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
105 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
106 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
108 init_waitqueue_head(&sdp->sd_log_waitq);
109 init_waitqueue_head(&sdp->sd_logd_waitq);
110 INIT_LIST_HEAD(&sdp->sd_ail1_list);
111 INIT_LIST_HEAD(&sdp->sd_ail2_list);
113 init_rwsem(&sdp->sd_log_flush_lock);
114 atomic_set(&sdp->sd_log_in_flight, 0);
115 init_waitqueue_head(&sdp->sd_log_flush_wait);
117 INIT_LIST_HEAD(&sdp->sd_revoke_list);
119 mutex_init(&sdp->sd_freeze_lock);
126 * gfs2_check_sb - Check superblock
127 * @sdp: the filesystem
128 * @sb: The superblock
129 * @silent: Don't print a message if the check fails
131 * Checks the version code of the FS is one that we understand how to
132 * read and that the sizes of the various on-disk structures have not
136 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
140 if (sb->sb_magic != GFS2_MAGIC ||
141 sb->sb_type != GFS2_METATYPE_SB) {
143 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
147 /* If format numbers match exactly, we're done. */
149 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
150 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
153 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
154 for (x = 0; gfs2_old_fs_formats[x]; x++)
155 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
158 if (!gfs2_old_fs_formats[x]) {
160 "GFS2: code version (%u, %u) is incompatible "
161 "with ondisk format (%u, %u)\n",
162 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
163 sb->sb_fs_format, sb->sb_multihost_format);
165 "GFS2: I don't know how to upgrade this FS\n");
170 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
171 for (x = 0; gfs2_old_multihost_formats[x]; x++)
172 if (gfs2_old_multihost_formats[x] ==
173 sb->sb_multihost_format)
176 if (!gfs2_old_multihost_formats[x]) {
178 "GFS2: code version (%u, %u) is incompatible "
179 "with ondisk format (%u, %u)\n",
180 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
181 sb->sb_fs_format, sb->sb_multihost_format);
183 "GFS2: I don't know how to upgrade this FS\n");
188 if (!sdp->sd_args.ar_upgrade) {
190 "GFS2: code version (%u, %u) is incompatible "
191 "with ondisk format (%u, %u)\n",
192 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
193 sb->sb_fs_format, sb->sb_multihost_format);
195 "GFS2: Use the \"upgrade\" mount option to upgrade "
197 printk(KERN_INFO "GFS2: See the manual for more details\n");
204 static void end_bio_io_page(struct bio *bio, int error)
206 struct page *page = bio->bi_private;
209 SetPageUptodate(page);
211 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
215 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
217 const struct gfs2_sb *str = buf;
219 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
220 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
221 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
222 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
223 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
224 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
225 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
226 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
227 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
228 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
229 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
231 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
232 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
233 memcpy(sb->sb_uuid, str->sb_uuid, 16);
237 * gfs2_read_super - Read the gfs2 super block from disk
238 * @sdp: The GFS2 super block
239 * @sector: The location of the super block
240 * @error: The error code to return
242 * This uses the bio functions to read the super block from disk
243 * because we want to be 100% sure that we never read cached data.
244 * A super block is read twice only during each GFS2 mount and is
245 * never written to by the filesystem. The first time its read no
246 * locks are held, and the only details which are looked at are those
247 * relating to the locking protocol. Once locking is up and working,
248 * the sb is read again under the lock to establish the location of
249 * the master directory (contains pointers to journals etc) and the
252 * Returns: 0 on success or error
255 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
257 struct super_block *sb = sdp->sd_vfs;
262 page = alloc_page(GFP_NOFS);
266 ClearPageUptodate(page);
267 ClearPageDirty(page);
270 bio = bio_alloc(GFP_NOFS, 1);
271 bio->bi_sector = sector * (sb->s_blocksize >> 9);
272 bio->bi_bdev = sb->s_bdev;
273 bio_add_page(bio, page, PAGE_SIZE, 0);
275 bio->bi_end_io = end_bio_io_page;
276 bio->bi_private = page;
277 submit_bio(READ_SYNC | REQ_META, bio);
278 wait_on_page_locked(page);
280 if (!PageUptodate(page)) {
285 gfs2_sb_in(&sdp->sd_sb, p);
292 * gfs2_read_sb - Read super block
293 * @sdp: The GFS2 superblock
294 * @silent: Don't print message if mount fails
298 static int gfs2_read_sb(struct gfs2_sbd *sdp, int silent)
300 u32 hash_blocks, ind_blocks, leaf_blocks;
305 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
308 fs_err(sdp, "can't read superblock\n");
312 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
316 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
317 GFS2_BASIC_BLOCK_SHIFT;
318 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
319 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
320 sizeof(struct gfs2_dinode)) / sizeof(u64);
321 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
322 sizeof(struct gfs2_meta_header)) / sizeof(u64);
323 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
324 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
325 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
326 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
327 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
328 sizeof(struct gfs2_meta_header)) /
329 sizeof(struct gfs2_quota_change);
331 /* Compute maximum reservation required to add a entry to a directory */
333 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
337 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
338 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
339 ind_blocks += tmp_blocks;
342 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
344 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
346 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
347 sizeof(struct gfs2_dinode);
348 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
353 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
355 m = do_div(d, sdp->sd_inptrs);
357 if (d != sdp->sd_heightsize[x - 1] || m)
359 sdp->sd_heightsize[x] = space;
361 sdp->sd_max_height = x;
362 sdp->sd_heightsize[x] = ~0;
363 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
365 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
366 sizeof(struct gfs2_dinode);
367 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
372 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
374 m = do_div(d, sdp->sd_inptrs);
376 if (d != sdp->sd_jheightsize[x - 1] || m)
378 sdp->sd_jheightsize[x] = space;
380 sdp->sd_max_jheight = x;
381 sdp->sd_jheightsize[x] = ~0;
382 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
387 static int init_names(struct gfs2_sbd *sdp, int silent)
392 proto = sdp->sd_args.ar_lockproto;
393 table = sdp->sd_args.ar_locktable;
395 /* Try to autodetect */
397 if (!proto[0] || !table[0]) {
398 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
402 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
407 proto = sdp->sd_sb.sb_lockproto;
409 table = sdp->sd_sb.sb_locktable;
413 table = sdp->sd_vfs->s_id;
415 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
416 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
418 table = sdp->sd_table_name;
419 while ((table = strchr(table, '/')))
426 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
434 error = gfs2_glock_nq_num(sdp,
435 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
436 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
439 fs_err(sdp, "can't acquire mount glock: %d\n", error);
443 error = gfs2_glock_nq_num(sdp,
444 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
446 LM_FLAG_NOEXP | GL_EXACT,
449 fs_err(sdp, "can't acquire live glock: %d\n", error);
453 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
454 CREATE, &sdp->sd_rename_gl);
456 fs_err(sdp, "can't create rename glock: %d\n", error);
460 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
461 CREATE, &sdp->sd_trans_gl);
463 fs_err(sdp, "can't create transaction glock: %d\n", error);
470 gfs2_glock_put(sdp->sd_trans_gl);
472 gfs2_glock_put(sdp->sd_rename_gl);
474 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
476 gfs2_glock_dq_uninit(mount_gh);
481 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
482 u64 no_addr, const char *name)
484 struct gfs2_sbd *sdp = sb->s_fs_info;
485 struct dentry *dentry;
488 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0);
490 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
491 return PTR_ERR(inode);
493 dentry = d_alloc_root(inode);
495 fs_err(sdp, "can't alloc %s dentry\n", name);
499 dentry->d_op = &gfs2_dops;
504 static int init_sb(struct gfs2_sbd *sdp, int silent)
506 struct super_block *sb = sdp->sd_vfs;
507 struct gfs2_holder sb_gh;
511 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
512 LM_ST_SHARED, 0, &sb_gh);
514 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
518 ret = gfs2_read_sb(sdp, silent);
520 fs_err(sdp, "can't read superblock: %d\n", ret);
524 /* Set up the buffer cache and SB for real */
525 if (sdp->sd_sb.sb_bsize < bdev_logical_block_size(sb->s_bdev)) {
527 fs_err(sdp, "FS block size (%u) is too small for device "
529 sdp->sd_sb.sb_bsize, bdev_logical_block_size(sb->s_bdev));
532 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
534 fs_err(sdp, "FS block size (%u) is too big for machine "
536 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
539 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
541 /* Get the root inode */
542 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
543 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
547 /* Get the master inode */
548 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
549 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
551 dput(sdp->sd_root_dir);
554 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
556 gfs2_glock_dq_uninit(&sb_gh);
561 * map_journal_extents - create a reusable "extent" mapping from all logical
562 * blocks to all physical blocks for the given journal. This will save
563 * us time when writing journal blocks. Most journals will have only one
564 * extent that maps all their logical blocks. That's because gfs2.mkfs
565 * arranges the journal blocks sequentially to maximize performance.
566 * So the extent would map the first block for the entire file length.
567 * However, gfs2_jadd can happen while file activity is happening, so
568 * those journals may not be sequential. Less likely is the case where
569 * the users created their own journals by mounting the metafs and
570 * laying it out. But it's still possible. These journals might have
573 * TODO: This should be done in bigger chunks rather than one block at a time,
574 * but since it's only done at mount time, I'm not worried about the
577 static int map_journal_extents(struct gfs2_sbd *sdp)
579 struct gfs2_jdesc *jd = sdp->sd_jdesc;
581 u64 db, prev_db; /* logical block, disk block, prev disk block */
582 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
583 struct gfs2_journal_extent *jext = NULL;
584 struct buffer_head bh;
589 for (lb = 0; lb < ip->i_disksize >> sdp->sd_sb.sb_bsize_shift; lb++) {
592 bh.b_size = 1 << ip->i_inode.i_blkbits;
593 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
596 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
597 "%u db=%llu\n", rc, lb, (unsigned long long)db);
600 if (!prev_db || db != prev_db + 1) {
601 jext = kzalloc(sizeof(struct gfs2_journal_extent),
604 printk(KERN_INFO "GFS2 error: out of memory "
605 "mapping journal extents.\n");
612 list_add_tail(&jext->extent_list, &jd->extent_list);
621 static void gfs2_others_may_mount(struct gfs2_sbd *sdp)
623 char *message = "FIRSTMOUNT=Done";
624 char *envp[] = { message, NULL };
625 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
626 ls->ls_first_done = 1;
627 kobject_uevent_env(&sdp->sd_kobj, KOBJ_CHANGE, envp);
631 * gfs2_jindex_hold - Grab a lock on the jindex
632 * @sdp: The GFS2 superblock
633 * @ji_gh: the holder for the jindex glock
638 static int gfs2_jindex_hold(struct gfs2_sbd *sdp, struct gfs2_holder *ji_gh)
640 struct gfs2_inode *dip = GFS2_I(sdp->sd_jindex);
643 struct gfs2_jdesc *jd;
648 mutex_lock(&sdp->sd_jindex_mutex);
651 error = gfs2_glock_nq_init(dip->i_gl, LM_ST_SHARED, 0, ji_gh);
655 name.len = sprintf(buf, "journal%u", sdp->sd_journals);
656 name.hash = gfs2_disk_hash(name.name, name.len);
658 error = gfs2_dir_check(sdp->sd_jindex, &name, NULL);
659 if (error == -ENOENT) {
664 gfs2_glock_dq_uninit(ji_gh);
670 jd = kzalloc(sizeof(struct gfs2_jdesc), GFP_KERNEL);
674 INIT_LIST_HEAD(&jd->extent_list);
675 INIT_WORK(&jd->jd_work, gfs2_recover_func);
676 jd->jd_inode = gfs2_lookupi(sdp->sd_jindex, &name, 1);
677 if (!jd->jd_inode || IS_ERR(jd->jd_inode)) {
681 error = PTR_ERR(jd->jd_inode);
686 spin_lock(&sdp->sd_jindex_spin);
687 jd->jd_jid = sdp->sd_journals++;
688 list_add_tail(&jd->jd_list, &sdp->sd_jindex_list);
689 spin_unlock(&sdp->sd_jindex_spin);
692 mutex_unlock(&sdp->sd_jindex_mutex);
697 static int init_journal(struct gfs2_sbd *sdp, int undo)
699 struct inode *master = sdp->sd_master_dir->d_inode;
700 struct gfs2_holder ji_gh;
701 struct gfs2_inode *ip;
710 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
711 if (IS_ERR(sdp->sd_jindex)) {
712 fs_err(sdp, "can't lookup journal index: %d\n", error);
713 return PTR_ERR(sdp->sd_jindex);
715 ip = GFS2_I(sdp->sd_jindex);
717 /* Load in the journal index special file */
719 error = gfs2_jindex_hold(sdp, &ji_gh);
721 fs_err(sdp, "can't read journal index: %d\n", error);
726 if (!gfs2_jindex_size(sdp)) {
727 fs_err(sdp, "no journals!\n");
731 if (sdp->sd_args.ar_spectator) {
732 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
733 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
734 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
735 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
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);
773 atomic_set(&sdp->sd_log_thresh1, 2*sdp->sd_jdesc->jd_blocks/5);
774 atomic_set(&sdp->sd_log_thresh2, 4*sdp->sd_jdesc->jd_blocks/5);
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),
787 fs_err(sdp, "error recovering journal %u: %d\n",
793 gfs2_others_may_mount(sdp);
794 } else if (!sdp->sd_args.ar_spectator) {
795 error = gfs2_recover_journal(sdp->sd_jdesc, true);
797 fs_err(sdp, "error recovering my journal: %d\n", error);
802 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
803 gfs2_glock_dq_uninit(&ji_gh);
809 if (!sdp->sd_args.ar_spectator)
810 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
812 if (!sdp->sd_args.ar_spectator)
813 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
815 gfs2_jindex_free(sdp);
817 gfs2_glock_dq_uninit(&ji_gh);
819 iput(sdp->sd_jindex);
824 static int init_inodes(struct gfs2_sbd *sdp, int undo)
827 struct gfs2_inode *ip;
828 struct inode *master = sdp->sd_master_dir->d_inode;
833 error = init_journal(sdp, undo);
837 /* Read in the master statfs inode */
838 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
839 if (IS_ERR(sdp->sd_statfs_inode)) {
840 error = PTR_ERR(sdp->sd_statfs_inode);
841 fs_err(sdp, "can't read in statfs inode: %d\n", error);
845 /* Read in the resource index inode */
846 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
847 if (IS_ERR(sdp->sd_rindex)) {
848 error = PTR_ERR(sdp->sd_rindex);
849 fs_err(sdp, "can't get resource index inode: %d\n", error);
852 ip = GFS2_I(sdp->sd_rindex);
853 sdp->sd_rindex_uptodate = 0;
855 /* Read in the quota inode */
856 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
857 if (IS_ERR(sdp->sd_quota_inode)) {
858 error = PTR_ERR(sdp->sd_quota_inode);
859 fs_err(sdp, "can't get quota file inode: %d\n", error);
865 iput(sdp->sd_quota_inode);
867 gfs2_clear_rgrpd(sdp);
868 iput(sdp->sd_rindex);
870 iput(sdp->sd_statfs_inode);
872 init_journal(sdp, UNDO);
877 static int init_per_node(struct gfs2_sbd *sdp, int undo)
879 struct inode *pn = NULL;
882 struct gfs2_inode *ip;
883 struct inode *master = sdp->sd_master_dir->d_inode;
885 if (sdp->sd_args.ar_spectator)
891 pn = gfs2_lookup_simple(master, "per_node");
894 fs_err(sdp, "can't find per_node directory: %d\n", error);
898 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
899 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
900 if (IS_ERR(sdp->sd_sc_inode)) {
901 error = PTR_ERR(sdp->sd_sc_inode);
902 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
906 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
907 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
908 if (IS_ERR(sdp->sd_qc_inode)) {
909 error = PTR_ERR(sdp->sd_qc_inode);
910 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
917 ip = GFS2_I(sdp->sd_sc_inode);
918 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
921 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
925 ip = GFS2_I(sdp->sd_qc_inode);
926 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0,
929 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
936 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
938 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
940 iput(sdp->sd_qc_inode);
942 iput(sdp->sd_sc_inode);
949 static int init_threads(struct gfs2_sbd *sdp, int undo)
951 struct task_struct *p;
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 void nolock_put_lock(struct kmem_cache *cachep, struct gfs2_glock *gl)
990 struct gfs2_sbd *sdp = gl->gl_sbd;
991 kmem_cache_free(cachep, gl);
992 if (atomic_dec_and_test(&sdp->sd_glock_disposal))
993 wake_up(&sdp->sd_glock_wait);
996 static const struct lm_lockops nolock_ops = {
997 .lm_proto_name = "lock_nolock",
998 .lm_put_lock = nolock_put_lock,
999 .lm_tokens = &nolock_tokens,
1003 * gfs2_lm_mount - mount a locking protocol
1004 * @sdp: the filesystem
1005 * @args: mount arguments
1006 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1011 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1013 const struct lm_lockops *lm;
1014 struct lm_lockstruct *ls = &sdp->sd_lockstruct;
1015 struct gfs2_args *args = &sdp->sd_args;
1016 const char *proto = sdp->sd_proto_name;
1017 const char *table = sdp->sd_table_name;
1022 if (!strcmp("lock_nolock", proto)) {
1024 sdp->sd_args.ar_localflocks = 1;
1025 sdp->sd_args.ar_localcaching = 1;
1026 #ifdef CONFIG_GFS2_FS_LOCKING_DLM
1027 } else if (!strcmp("lock_dlm", proto)) {
1031 printk(KERN_INFO "GFS2: can't find protocol %s\n", proto);
1035 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1040 for (options = args->ar_hostdata; (o = strsep(&options, ":")); ) {
1041 substring_t tmp[MAX_OPT_ARGS];
1047 token = match_token(o, *lm->lm_tokens, tmp);
1050 ret = match_int(&tmp[0], &option);
1051 if (ret || option < 0)
1052 goto hostdata_error;
1053 if (test_and_clear_bit(SDF_NOJOURNALID, &sdp->sd_flags))
1054 ls->ls_jid = option;
1057 /* Obsolete, but left for backward compat purposes */
1060 ret = match_int(&tmp[0], &option);
1061 if (ret || (option != 0 && option != 1))
1062 goto hostdata_error;
1063 ls->ls_first = option;
1066 ret = match_int(&tmp[0], &option);
1067 if (ret || (option != 0 && option != 1))
1068 goto hostdata_error;
1069 ls->ls_nodir = option;
1074 fs_info(sdp, "unknown hostdata (%s)\n", o);
1079 if (sdp->sd_args.ar_spectator)
1080 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1082 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1083 sdp->sd_lockstruct.ls_jid);
1085 fsname = strchr(table, ':');
1088 if (lm->lm_mount == NULL) {
1089 fs_info(sdp, "Now mounting FS...\n");
1092 ret = lm->lm_mount(sdp, fsname);
1094 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1098 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1100 const struct lm_lockops *lm = sdp->sd_lockstruct.ls_ops;
1101 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)) &&
1103 lm->lm_unmount(sdp);
1106 static int gfs2_journalid_wait(void *word)
1108 if (signal_pending(current))
1114 static int wait_on_journal(struct gfs2_sbd *sdp)
1116 if (sdp->sd_args.ar_spectator)
1118 if (sdp->sd_lockstruct.ls_ops->lm_mount == NULL)
1121 return wait_on_bit(&sdp->sd_flags, SDF_NOJOURNALID, gfs2_journalid_wait, TASK_INTERRUPTIBLE);
1124 void gfs2_online_uevent(struct gfs2_sbd *sdp)
1126 struct super_block *sb = sdp->sd_vfs;
1129 char *envp[] = { ro, spectator, NULL };
1130 sprintf(ro, "RDONLY=%d", (sb->s_flags & MS_RDONLY) ? 1 : 0);
1131 sprintf(spectator, "SPECTATOR=%d", sdp->sd_args.ar_spectator ? 1 : 0);
1132 kobject_uevent_env(&sdp->sd_kobj, KOBJ_ONLINE, envp);
1136 * fill_super - Read in superblock
1137 * @sb: The VFS superblock
1138 * @data: Mount options
1139 * @silent: Don't complain if it's not a GFS2 filesystem
1144 static int fill_super(struct super_block *sb, struct gfs2_args *args, int silent)
1146 struct gfs2_sbd *sdp;
1147 struct gfs2_holder mount_gh;
1152 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1155 sdp->sd_args = *args;
1157 if (sdp->sd_args.ar_spectator) {
1158 sb->s_flags |= MS_RDONLY;
1159 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
1161 if (sdp->sd_args.ar_posix_acl)
1162 sb->s_flags |= MS_POSIXACL;
1163 if (sdp->sd_args.ar_nobarrier)
1164 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1166 sb->s_magic = GFS2_MAGIC;
1167 sb->s_op = &gfs2_super_ops;
1168 sb->s_export_op = &gfs2_export_ops;
1169 sb->s_xattr = gfs2_xattr_handlers;
1170 sb->s_qcop = &gfs2_quotactl_ops;
1171 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
1172 sb->s_time_gran = 1;
1173 sb->s_maxbytes = MAX_LFS_FILESIZE;
1175 /* Set up the buffer cache and fill in some fake block size values
1176 to allow us to read-in the on-disk superblock. */
1177 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1178 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1179 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1180 GFS2_BASIC_BLOCK_SHIFT;
1181 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1183 sdp->sd_tune.gt_logd_secs = sdp->sd_args.ar_commit;
1184 sdp->sd_tune.gt_quota_quantum = sdp->sd_args.ar_quota_quantum;
1185 if (sdp->sd_args.ar_statfs_quantum) {
1186 sdp->sd_tune.gt_statfs_slow = 0;
1187 sdp->sd_tune.gt_statfs_quantum = sdp->sd_args.ar_statfs_quantum;
1190 sdp->sd_tune.gt_statfs_slow = 1;
1191 sdp->sd_tune.gt_statfs_quantum = 30;
1194 error = init_names(sdp, silent);
1198 gfs2_create_debugfs_file(sdp);
1200 error = gfs2_sys_fs_add(sdp);
1204 error = gfs2_lm_mount(sdp, silent);
1208 error = init_locking(sdp, &mount_gh, DO);
1212 error = init_sb(sdp, silent);
1216 error = wait_on_journal(sdp);
1220 error = init_inodes(sdp, DO);
1224 error = init_per_node(sdp, DO);
1228 error = gfs2_statfs_init(sdp);
1230 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1234 error = init_threads(sdp, DO);
1238 if (!(sb->s_flags & MS_RDONLY)) {
1239 error = gfs2_make_fs_rw(sdp);
1241 fs_err(sdp, "can't make FS RW: %d\n", error);
1246 gfs2_glock_dq_uninit(&mount_gh);
1247 gfs2_online_uevent(sdp);
1251 init_threads(sdp, UNDO);
1253 init_per_node(sdp, UNDO);
1255 init_inodes(sdp, UNDO);
1257 if (sdp->sd_root_dir)
1258 dput(sdp->sd_root_dir);
1259 if (sdp->sd_master_dir)
1260 dput(sdp->sd_master_dir);
1265 init_locking(sdp, &mount_gh, UNDO);
1267 invalidate_inodes(sb);
1268 gfs2_gl_hash_clear(sdp);
1269 gfs2_lm_unmount(sdp);
1271 gfs2_sys_fs_del(sdp);
1273 gfs2_delete_debugfs_file(sdp);
1275 sb->s_fs_info = NULL;
1279 static int set_gfs2_super(struct super_block *s, void *data)
1282 s->s_dev = s->s_bdev->bd_dev;
1285 * We set the bdi here to the queue backing, file systems can
1286 * overwrite this in ->fill_super()
1288 s->s_bdi = &bdev_get_queue(s->s_bdev)->backing_dev_info;
1292 static int test_gfs2_super(struct super_block *s, void *ptr)
1294 struct block_device *bdev = ptr;
1295 return (bdev == s->s_bdev);
1299 * gfs2_get_sb - Get the GFS2 superblock
1300 * @fs_type: The GFS2 filesystem type
1301 * @flags: Mount flags
1302 * @dev_name: The name of the device
1303 * @data: The mount arguments
1304 * @mnt: The vfsmnt for this mount
1306 * Q. Why not use get_sb_bdev() ?
1307 * A. We need to select one of two root directories to mount, independent
1308 * of whether this is the initial, or subsequent, mount of this sb
1310 * Returns: 0 or -ve on error
1313 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1314 const char *dev_name, void *data, struct vfsmount *mnt)
1316 struct block_device *bdev;
1317 struct super_block *s;
1318 fmode_t mode = FMODE_READ;
1320 struct gfs2_args args;
1321 struct gfs2_sbd *sdp;
1323 if (!(flags & MS_RDONLY))
1324 mode |= FMODE_WRITE;
1326 bdev = open_bdev_exclusive(dev_name, mode, fs_type);
1328 return PTR_ERR(bdev);
1331 * once the super is inserted into the list by sget, s_umount
1332 * will protect the lockfs code from trying to start a snapshot
1333 * while we are mounting
1335 mutex_lock(&bdev->bd_fsfreeze_mutex);
1336 if (bdev->bd_fsfreeze_count > 0) {
1337 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1341 s = sget(fs_type, test_gfs2_super, set_gfs2_super, bdev);
1342 mutex_unlock(&bdev->bd_fsfreeze_mutex);
1347 memset(&args, 0, sizeof(args));
1348 args.ar_quota = GFS2_QUOTA_DEFAULT;
1349 args.ar_data = GFS2_DATA_DEFAULT;
1350 args.ar_commit = 30;
1351 args.ar_statfs_quantum = 30;
1352 args.ar_quota_quantum = 60;
1353 args.ar_errors = GFS2_ERRORS_DEFAULT;
1355 error = gfs2_mount_args(&args, data);
1357 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1360 deactivate_locked_super(s);
1366 if ((flags ^ s->s_flags) & MS_RDONLY)
1368 close_bdev_exclusive(bdev, mode);
1370 char b[BDEVNAME_SIZE];
1374 strlcpy(s->s_id, bdevname(bdev, b), sizeof(s->s_id));
1375 sb_set_blocksize(s, block_size(bdev));
1376 error = fill_super(s, &args, flags & MS_SILENT ? 1 : 0);
1378 deactivate_locked_super(s);
1381 s->s_flags |= MS_ACTIVE;
1388 mnt->mnt_root = dget(sdp->sd_master_dir);
1390 mnt->mnt_root = dget(sdp->sd_root_dir);
1394 deactivate_locked_super(s);
1396 close_bdev_exclusive(bdev, mode);
1400 static int set_meta_super(struct super_block *s, void *ptr)
1405 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1406 const char *dev_name, void *data, struct vfsmount *mnt)
1408 struct super_block *s;
1409 struct gfs2_sbd *sdp;
1413 error = kern_path(dev_name, LOOKUP_FOLLOW, &path);
1415 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1419 s = sget(&gfs2_fs_type, test_gfs2_super, set_meta_super,
1420 path.dentry->d_inode->i_sb->s_bdev);
1423 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1426 if ((flags ^ s->s_flags) & MS_RDONLY) {
1427 deactivate_locked_super(s);
1432 mnt->mnt_root = dget(sdp->sd_master_dir);
1436 static void gfs2_kill_sb(struct super_block *sb)
1438 struct gfs2_sbd *sdp = sb->s_fs_info;
1441 kill_block_super(sb);
1445 gfs2_meta_syncfs(sdp);
1446 dput(sdp->sd_root_dir);
1447 dput(sdp->sd_master_dir);
1448 sdp->sd_root_dir = NULL;
1449 sdp->sd_master_dir = NULL;
1450 shrink_dcache_sb(sb);
1451 kill_block_super(sb);
1452 gfs2_delete_debugfs_file(sdp);
1456 struct file_system_type gfs2_fs_type = {
1458 .fs_flags = FS_REQUIRES_DEV,
1459 .get_sb = gfs2_get_sb,
1460 .kill_sb = gfs2_kill_sb,
1461 .owner = THIS_MODULE,
1464 struct file_system_type gfs2meta_fs_type = {
1466 .fs_flags = FS_REQUIRES_DEV,
1467 .get_sb = gfs2_get_sb_meta,
1468 .owner = THIS_MODULE,