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>
30 #include "ops_fstype.h"
31 #include "ops_dentry.h"
32 #include "ops_super.h"
43 static const u32 gfs2_old_fs_formats[] = {
47 static const u32 gfs2_old_multihost_formats[] = {
52 * gfs2_tune_init - Fill a gfs2_tune structure with default values
57 static void gfs2_tune_init(struct gfs2_tune *gt)
59 spin_lock_init(>->gt_spin);
61 gt->gt_demote_secs = 300;
62 gt->gt_incore_log_blocks = 1024;
63 gt->gt_log_flush_secs = 60;
64 gt->gt_recoverd_secs = 60;
66 gt->gt_quotad_secs = 5;
67 gt->gt_quota_simul_sync = 64;
68 gt->gt_quota_warn_period = 10;
69 gt->gt_quota_scale_num = 1;
70 gt->gt_quota_scale_den = 1;
71 gt->gt_quota_cache_secs = 300;
72 gt->gt_quota_quantum = 60;
73 gt->gt_new_files_jdata = 0;
74 gt->gt_max_readahead = 1 << 18;
75 gt->gt_stall_secs = 600;
76 gt->gt_complain_secs = 10;
77 gt->gt_statfs_quantum = 30;
78 gt->gt_statfs_slow = 0;
81 static struct gfs2_sbd *init_sbd(struct super_block *sb)
85 sdp = kzalloc(sizeof(struct gfs2_sbd), GFP_KERNEL);
92 gfs2_tune_init(&sdp->sd_tune);
94 INIT_LIST_HEAD(&sdp->sd_reclaim_list);
95 spin_lock_init(&sdp->sd_reclaim_lock);
96 init_waitqueue_head(&sdp->sd_reclaim_wq);
98 mutex_init(&sdp->sd_inum_mutex);
99 spin_lock_init(&sdp->sd_statfs_spin);
101 spin_lock_init(&sdp->sd_rindex_spin);
102 mutex_init(&sdp->sd_rindex_mutex);
103 INIT_LIST_HEAD(&sdp->sd_rindex_list);
104 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
106 INIT_LIST_HEAD(&sdp->sd_jindex_list);
107 spin_lock_init(&sdp->sd_jindex_spin);
108 mutex_init(&sdp->sd_jindex_mutex);
110 INIT_LIST_HEAD(&sdp->sd_quota_list);
111 spin_lock_init(&sdp->sd_quota_spin);
112 mutex_init(&sdp->sd_quota_mutex);
114 spin_lock_init(&sdp->sd_log_lock);
116 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
117 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
118 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
119 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
120 INIT_LIST_HEAD(&sdp->sd_log_le_ordered);
122 mutex_init(&sdp->sd_log_reserve_mutex);
123 INIT_LIST_HEAD(&sdp->sd_ail1_list);
124 INIT_LIST_HEAD(&sdp->sd_ail2_list);
126 init_rwsem(&sdp->sd_log_flush_lock);
127 atomic_set(&sdp->sd_log_in_flight, 0);
128 init_waitqueue_head(&sdp->sd_log_flush_wait);
130 INIT_LIST_HEAD(&sdp->sd_revoke_list);
132 mutex_init(&sdp->sd_freeze_lock);
139 * gfs2_check_sb - Check superblock
140 * @sdp: the filesystem
141 * @sb: The superblock
142 * @silent: Don't print a message if the check fails
144 * Checks the version code of the FS is one that we understand how to
145 * read and that the sizes of the various on-disk structures have not
149 static int gfs2_check_sb(struct gfs2_sbd *sdp, struct gfs2_sb_host *sb, int silent)
153 if (sb->sb_magic != GFS2_MAGIC ||
154 sb->sb_type != GFS2_METATYPE_SB) {
156 printk(KERN_WARNING "GFS2: not a GFS2 filesystem\n");
160 /* If format numbers match exactly, we're done. */
162 if (sb->sb_fs_format == GFS2_FORMAT_FS &&
163 sb->sb_multihost_format == GFS2_FORMAT_MULTI)
166 if (sb->sb_fs_format != GFS2_FORMAT_FS) {
167 for (x = 0; gfs2_old_fs_formats[x]; x++)
168 if (gfs2_old_fs_formats[x] == sb->sb_fs_format)
171 if (!gfs2_old_fs_formats[x]) {
173 "GFS2: code version (%u, %u) is incompatible "
174 "with ondisk format (%u, %u)\n",
175 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
176 sb->sb_fs_format, sb->sb_multihost_format);
178 "GFS2: I don't know how to upgrade this FS\n");
183 if (sb->sb_multihost_format != GFS2_FORMAT_MULTI) {
184 for (x = 0; gfs2_old_multihost_formats[x]; x++)
185 if (gfs2_old_multihost_formats[x] ==
186 sb->sb_multihost_format)
189 if (!gfs2_old_multihost_formats[x]) {
191 "GFS2: code version (%u, %u) is incompatible "
192 "with ondisk format (%u, %u)\n",
193 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
194 sb->sb_fs_format, sb->sb_multihost_format);
196 "GFS2: I don't know how to upgrade this FS\n");
201 if (!sdp->sd_args.ar_upgrade) {
203 "GFS2: code version (%u, %u) is incompatible "
204 "with ondisk format (%u, %u)\n",
205 GFS2_FORMAT_FS, GFS2_FORMAT_MULTI,
206 sb->sb_fs_format, sb->sb_multihost_format);
208 "GFS2: Use the \"upgrade\" mount option to upgrade "
210 printk(KERN_INFO "GFS2: See the manual for more details\n");
217 static void end_bio_io_page(struct bio *bio, int error)
219 struct page *page = bio->bi_private;
222 SetPageUptodate(page);
224 printk(KERN_WARNING "gfs2: error %d reading superblock\n", error);
228 static void gfs2_sb_in(struct gfs2_sb_host *sb, const void *buf)
230 const struct gfs2_sb *str = buf;
232 sb->sb_magic = be32_to_cpu(str->sb_header.mh_magic);
233 sb->sb_type = be32_to_cpu(str->sb_header.mh_type);
234 sb->sb_format = be32_to_cpu(str->sb_header.mh_format);
235 sb->sb_fs_format = be32_to_cpu(str->sb_fs_format);
236 sb->sb_multihost_format = be32_to_cpu(str->sb_multihost_format);
237 sb->sb_bsize = be32_to_cpu(str->sb_bsize);
238 sb->sb_bsize_shift = be32_to_cpu(str->sb_bsize_shift);
239 sb->sb_master_dir.no_addr = be64_to_cpu(str->sb_master_dir.no_addr);
240 sb->sb_master_dir.no_formal_ino = be64_to_cpu(str->sb_master_dir.no_formal_ino);
241 sb->sb_root_dir.no_addr = be64_to_cpu(str->sb_root_dir.no_addr);
242 sb->sb_root_dir.no_formal_ino = be64_to_cpu(str->sb_root_dir.no_formal_ino);
244 memcpy(sb->sb_lockproto, str->sb_lockproto, GFS2_LOCKNAME_LEN);
245 memcpy(sb->sb_locktable, str->sb_locktable, GFS2_LOCKNAME_LEN);
249 * gfs2_read_super - Read the gfs2 super block from disk
250 * @sdp: The GFS2 super block
251 * @sector: The location of the super block
252 * @error: The error code to return
254 * This uses the bio functions to read the super block from disk
255 * because we want to be 100% sure that we never read cached data.
256 * A super block is read twice only during each GFS2 mount and is
257 * never written to by the filesystem. The first time its read no
258 * locks are held, and the only details which are looked at are those
259 * relating to the locking protocol. Once locking is up and working,
260 * the sb is read again under the lock to establish the location of
261 * the master directory (contains pointers to journals etc) and the
264 * Returns: 0 on success or error
267 static int gfs2_read_super(struct gfs2_sbd *sdp, sector_t sector)
269 struct super_block *sb = sdp->sd_vfs;
274 page = alloc_page(GFP_NOFS);
278 ClearPageUptodate(page);
279 ClearPageDirty(page);
282 bio = bio_alloc(GFP_NOFS, 1);
283 if (unlikely(!bio)) {
288 bio->bi_sector = sector * (sb->s_blocksize >> 9);
289 bio->bi_bdev = sb->s_bdev;
290 bio_add_page(bio, page, PAGE_SIZE, 0);
292 bio->bi_end_io = end_bio_io_page;
293 bio->bi_private = page;
294 submit_bio(READ_SYNC | (1 << BIO_RW_META), bio);
295 wait_on_page_locked(page);
297 if (!PageUptodate(page)) {
302 gfs2_sb_in(&sdp->sd_sb, p);
308 * gfs2_read_sb - Read super block
309 * @sdp: The GFS2 superblock
310 * @gl: the glock for the superblock (assumed to be held)
311 * @silent: Don't print message if mount fails
315 static int gfs2_read_sb(struct gfs2_sbd *sdp, struct gfs2_glock *gl, int silent)
317 u32 hash_blocks, ind_blocks, leaf_blocks;
322 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
325 fs_err(sdp, "can't read superblock\n");
329 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
333 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
334 GFS2_BASIC_BLOCK_SHIFT;
335 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
336 sdp->sd_diptrs = (sdp->sd_sb.sb_bsize -
337 sizeof(struct gfs2_dinode)) / sizeof(u64);
338 sdp->sd_inptrs = (sdp->sd_sb.sb_bsize -
339 sizeof(struct gfs2_meta_header)) / sizeof(u64);
340 sdp->sd_jbsize = sdp->sd_sb.sb_bsize - sizeof(struct gfs2_meta_header);
341 sdp->sd_hash_bsize = sdp->sd_sb.sb_bsize / 2;
342 sdp->sd_hash_bsize_shift = sdp->sd_sb.sb_bsize_shift - 1;
343 sdp->sd_hash_ptrs = sdp->sd_hash_bsize / sizeof(u64);
344 sdp->sd_qc_per_block = (sdp->sd_sb.sb_bsize -
345 sizeof(struct gfs2_meta_header)) /
346 sizeof(struct gfs2_quota_change);
348 /* Compute maximum reservation required to add a entry to a directory */
350 hash_blocks = DIV_ROUND_UP(sizeof(u64) * (1 << GFS2_DIR_MAX_DEPTH),
354 for (tmp_blocks = hash_blocks; tmp_blocks > sdp->sd_diptrs;) {
355 tmp_blocks = DIV_ROUND_UP(tmp_blocks, sdp->sd_inptrs);
356 ind_blocks += tmp_blocks;
359 leaf_blocks = 2 + GFS2_DIR_MAX_DEPTH;
361 sdp->sd_max_dirres = hash_blocks + ind_blocks + leaf_blocks;
363 sdp->sd_heightsize[0] = sdp->sd_sb.sb_bsize -
364 sizeof(struct gfs2_dinode);
365 sdp->sd_heightsize[1] = sdp->sd_sb.sb_bsize * sdp->sd_diptrs;
370 space = sdp->sd_heightsize[x - 1] * sdp->sd_inptrs;
372 m = do_div(d, sdp->sd_inptrs);
374 if (d != sdp->sd_heightsize[x - 1] || m)
376 sdp->sd_heightsize[x] = space;
378 sdp->sd_max_height = x;
379 sdp->sd_heightsize[x] = ~0;
380 gfs2_assert(sdp, sdp->sd_max_height <= GFS2_MAX_META_HEIGHT);
382 sdp->sd_jheightsize[0] = sdp->sd_sb.sb_bsize -
383 sizeof(struct gfs2_dinode);
384 sdp->sd_jheightsize[1] = sdp->sd_jbsize * sdp->sd_diptrs;
389 space = sdp->sd_jheightsize[x - 1] * sdp->sd_inptrs;
391 m = do_div(d, sdp->sd_inptrs);
393 if (d != sdp->sd_jheightsize[x - 1] || m)
395 sdp->sd_jheightsize[x] = space;
397 sdp->sd_max_jheight = x;
398 sdp->sd_jheightsize[x] = ~0;
399 gfs2_assert(sdp, sdp->sd_max_jheight <= GFS2_MAX_META_HEIGHT);
404 static int init_names(struct gfs2_sbd *sdp, int silent)
409 proto = sdp->sd_args.ar_lockproto;
410 table = sdp->sd_args.ar_locktable;
412 /* Try to autodetect */
414 if (!proto[0] || !table[0]) {
415 error = gfs2_read_super(sdp, GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
419 error = gfs2_check_sb(sdp, &sdp->sd_sb, silent);
424 proto = sdp->sd_sb.sb_lockproto;
426 table = sdp->sd_sb.sb_locktable;
430 table = sdp->sd_vfs->s_id;
432 strlcpy(sdp->sd_proto_name, proto, GFS2_FSNAME_LEN);
433 strlcpy(sdp->sd_table_name, table, GFS2_FSNAME_LEN);
435 table = sdp->sd_table_name;
436 while ((table = strchr(table, '/')))
443 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
446 struct task_struct *p;
452 for (sdp->sd_glockd_num = 0;
453 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
454 sdp->sd_glockd_num++) {
455 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
458 fs_err(sdp, "can't start glockd thread: %d\n", error);
461 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
464 error = gfs2_glock_nq_num(sdp,
465 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
466 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
469 fs_err(sdp, "can't acquire mount glock: %d\n", error);
473 error = gfs2_glock_nq_num(sdp,
474 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
476 LM_FLAG_NOEXP | GL_EXACT,
479 fs_err(sdp, "can't acquire live glock: %d\n", error);
483 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
484 CREATE, &sdp->sd_rename_gl);
486 fs_err(sdp, "can't create rename glock: %d\n", error);
490 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
491 CREATE, &sdp->sd_trans_gl);
493 fs_err(sdp, "can't create transaction glock: %d\n", error);
496 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
501 gfs2_glock_put(sdp->sd_trans_gl);
503 gfs2_glock_put(sdp->sd_rename_gl);
505 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
507 gfs2_glock_dq_uninit(mount_gh);
509 while (sdp->sd_glockd_num--)
510 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
515 static int gfs2_lookup_root(struct super_block *sb, struct dentry **dptr,
516 u64 no_addr, const char *name)
518 struct gfs2_sbd *sdp = sb->s_fs_info;
519 struct dentry *dentry;
522 inode = gfs2_inode_lookup(sb, DT_DIR, no_addr, 0, 0);
524 fs_err(sdp, "can't read in %s inode: %ld\n", name, PTR_ERR(inode));
525 return PTR_ERR(inode);
527 dentry = d_alloc_root(inode);
529 fs_err(sdp, "can't alloc %s dentry\n", name);
533 dentry->d_op = &gfs2_dops;
538 static int init_sb(struct gfs2_sbd *sdp, int silent)
540 struct super_block *sb = sdp->sd_vfs;
541 struct gfs2_holder sb_gh;
545 ret = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
546 LM_ST_SHARED, 0, &sb_gh);
548 fs_err(sdp, "can't acquire superblock glock: %d\n", ret);
552 ret = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
554 fs_err(sdp, "can't read superblock: %d\n", ret);
558 /* Set up the buffer cache and SB for real */
559 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
561 fs_err(sdp, "FS block size (%u) is too small for device "
563 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
566 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
568 fs_err(sdp, "FS block size (%u) is too big for machine "
570 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
573 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
575 /* Get the root inode */
576 no_addr = sdp->sd_sb.sb_root_dir.no_addr;
577 ret = gfs2_lookup_root(sb, &sdp->sd_root_dir, no_addr, "root");
581 /* Get the master inode */
582 no_addr = sdp->sd_sb.sb_master_dir.no_addr;
583 ret = gfs2_lookup_root(sb, &sdp->sd_master_dir, no_addr, "master");
585 dput(sdp->sd_root_dir);
588 sb->s_root = dget(sdp->sd_args.ar_meta ? sdp->sd_master_dir : sdp->sd_root_dir);
590 gfs2_glock_dq_uninit(&sb_gh);
595 * map_journal_extents - create a reusable "extent" mapping from all logical
596 * blocks to all physical blocks for the given journal. This will save
597 * us time when writing journal blocks. Most journals will have only one
598 * extent that maps all their logical blocks. That's because gfs2.mkfs
599 * arranges the journal blocks sequentially to maximize performance.
600 * So the extent would map the first block for the entire file length.
601 * However, gfs2_jadd can happen while file activity is happening, so
602 * those journals may not be sequential. Less likely is the case where
603 * the users created their own journals by mounting the metafs and
604 * laying it out. But it's still possible. These journals might have
607 * TODO: This should be done in bigger chunks rather than one block at a time,
608 * but since it's only done at mount time, I'm not worried about the
611 static int map_journal_extents(struct gfs2_sbd *sdp)
613 struct gfs2_jdesc *jd = sdp->sd_jdesc;
615 u64 db, prev_db; /* logical block, disk block, prev disk block */
616 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
617 struct gfs2_journal_extent *jext = NULL;
618 struct buffer_head bh;
623 for (lb = 0; lb < ip->i_di.di_size >> sdp->sd_sb.sb_bsize_shift; lb++) {
626 bh.b_size = 1 << ip->i_inode.i_blkbits;
627 rc = gfs2_block_map(jd->jd_inode, lb, &bh, 0);
630 printk(KERN_INFO "GFS2 journal mapping error %d: lb="
631 "%u db=%llu\n", rc, lb, (unsigned long long)db);
634 if (!prev_db || db != prev_db + 1) {
635 jext = kzalloc(sizeof(struct gfs2_journal_extent),
638 printk(KERN_INFO "GFS2 error: out of memory "
639 "mapping journal extents.\n");
646 list_add_tail(&jext->extent_list, &jd->extent_list);
655 static void gfs2_lm_others_may_mount(struct gfs2_sbd *sdp)
657 if (!sdp->sd_lockstruct.ls_ops->lm_others_may_mount)
659 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
660 sdp->sd_lockstruct.ls_ops->lm_others_may_mount(
661 sdp->sd_lockstruct.ls_lockspace);
664 static int init_journal(struct gfs2_sbd *sdp, int undo)
666 struct inode *master = sdp->sd_master_dir->d_inode;
667 struct gfs2_holder ji_gh;
668 struct task_struct *p;
669 struct gfs2_inode *ip;
678 sdp->sd_jindex = gfs2_lookup_simple(master, "jindex");
679 if (IS_ERR(sdp->sd_jindex)) {
680 fs_err(sdp, "can't lookup journal index: %d\n", error);
681 return PTR_ERR(sdp->sd_jindex);
683 ip = GFS2_I(sdp->sd_jindex);
684 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
686 /* Load in the journal index special file */
688 error = gfs2_jindex_hold(sdp, &ji_gh);
690 fs_err(sdp, "can't read journal index: %d\n", error);
695 if (!gfs2_jindex_size(sdp)) {
696 fs_err(sdp, "no journals!\n");
700 if (sdp->sd_args.ar_spectator) {
701 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
702 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
704 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
705 fs_err(sdp, "can't mount journal #%u\n",
706 sdp->sd_lockstruct.ls_jid);
707 fs_err(sdp, "there are only %u journals (0 - %u)\n",
708 gfs2_jindex_size(sdp),
709 gfs2_jindex_size(sdp) - 1);
712 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
714 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
716 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
717 &sdp->sd_journal_gh);
719 fs_err(sdp, "can't acquire journal glock: %d\n", error);
723 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
724 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
725 LM_FLAG_NOEXP | GL_EXACT | GL_NOCACHE,
728 fs_err(sdp, "can't acquire journal inode glock: %d\n",
730 goto fail_journal_gh;
733 error = gfs2_jdesc_check(sdp->sd_jdesc);
735 fs_err(sdp, "my journal (%u) is bad: %d\n",
736 sdp->sd_jdesc->jd_jid, error);
739 atomic_set(&sdp->sd_log_blks_free, sdp->sd_jdesc->jd_blocks);
741 /* Map the extents for this journal's blocks */
742 map_journal_extents(sdp);
745 if (sdp->sd_lockstruct.ls_first) {
747 for (x = 0; x < sdp->sd_journals; x++) {
748 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
750 fs_err(sdp, "error recovering journal %u: %d\n",
756 gfs2_lm_others_may_mount(sdp);
757 } else if (!sdp->sd_args.ar_spectator) {
758 error = gfs2_recover_journal(sdp->sd_jdesc);
760 fs_err(sdp, "error recovering my journal: %d\n", error);
765 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
766 gfs2_glock_dq_uninit(&ji_gh);
769 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
772 fs_err(sdp, "can't start recoverd thread: %d\n", error);
775 sdp->sd_recoverd_process = p;
780 kthread_stop(sdp->sd_recoverd_process);
782 if (!sdp->sd_args.ar_spectator)
783 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
785 if (!sdp->sd_args.ar_spectator)
786 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
788 gfs2_jindex_free(sdp);
790 gfs2_glock_dq_uninit(&ji_gh);
792 iput(sdp->sd_jindex);
797 static int init_inodes(struct gfs2_sbd *sdp, int undo)
800 struct gfs2_inode *ip;
801 struct inode *master = sdp->sd_master_dir->d_inode;
806 error = init_journal(sdp, undo);
810 /* Read in the master inode number inode */
811 sdp->sd_inum_inode = gfs2_lookup_simple(master, "inum");
812 if (IS_ERR(sdp->sd_inum_inode)) {
813 error = PTR_ERR(sdp->sd_inum_inode);
814 fs_err(sdp, "can't read in inum inode: %d\n", error);
819 /* Read in the master statfs inode */
820 sdp->sd_statfs_inode = gfs2_lookup_simple(master, "statfs");
821 if (IS_ERR(sdp->sd_statfs_inode)) {
822 error = PTR_ERR(sdp->sd_statfs_inode);
823 fs_err(sdp, "can't read in statfs inode: %d\n", error);
827 /* Read in the resource index inode */
828 sdp->sd_rindex = gfs2_lookup_simple(master, "rindex");
829 if (IS_ERR(sdp->sd_rindex)) {
830 error = PTR_ERR(sdp->sd_rindex);
831 fs_err(sdp, "can't get resource index inode: %d\n", error);
834 ip = GFS2_I(sdp->sd_rindex);
835 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
836 sdp->sd_rindex_uptodate = 0;
838 /* Read in the quota inode */
839 sdp->sd_quota_inode = gfs2_lookup_simple(master, "quota");
840 if (IS_ERR(sdp->sd_quota_inode)) {
841 error = PTR_ERR(sdp->sd_quota_inode);
842 fs_err(sdp, "can't get quota file inode: %d\n", error);
848 iput(sdp->sd_quota_inode);
850 gfs2_clear_rgrpd(sdp);
851 iput(sdp->sd_rindex);
853 iput(sdp->sd_statfs_inode);
855 iput(sdp->sd_inum_inode);
857 init_journal(sdp, UNDO);
862 static int init_per_node(struct gfs2_sbd *sdp, int undo)
864 struct inode *pn = NULL;
867 struct gfs2_inode *ip;
868 struct inode *master = sdp->sd_master_dir->d_inode;
870 if (sdp->sd_args.ar_spectator)
876 pn = gfs2_lookup_simple(master, "per_node");
879 fs_err(sdp, "can't find per_node directory: %d\n", error);
883 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
884 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
885 if (IS_ERR(sdp->sd_ir_inode)) {
886 error = PTR_ERR(sdp->sd_ir_inode);
887 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
891 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
892 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
893 if (IS_ERR(sdp->sd_sc_inode)) {
894 error = PTR_ERR(sdp->sd_sc_inode);
895 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
899 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
900 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
901 if (IS_ERR(sdp->sd_qc_inode)) {
902 error = PTR_ERR(sdp->sd_qc_inode);
903 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
910 ip = GFS2_I(sdp->sd_ir_inode);
911 error = gfs2_glock_nq_init(ip->i_gl,
915 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
919 ip = GFS2_I(sdp->sd_sc_inode);
920 error = gfs2_glock_nq_init(ip->i_gl,
924 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
928 ip = GFS2_I(sdp->sd_qc_inode);
929 error = gfs2_glock_nq_init(ip->i_gl,
933 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
940 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
942 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
944 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
946 iput(sdp->sd_qc_inode);
948 iput(sdp->sd_sc_inode);
950 iput(sdp->sd_ir_inode);
957 static int init_threads(struct gfs2_sbd *sdp, int undo)
959 struct task_struct *p;
965 sdp->sd_log_flush_time = jiffies;
966 sdp->sd_jindex_refresh_time = jiffies;
968 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
971 fs_err(sdp, "can't start logd thread: %d\n", error);
974 sdp->sd_logd_process = p;
976 sdp->sd_statfs_sync_time = jiffies;
977 sdp->sd_quota_sync_time = jiffies;
979 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
982 fs_err(sdp, "can't start quotad thread: %d\n", error);
985 sdp->sd_quotad_process = p;
991 kthread_stop(sdp->sd_quotad_process);
993 kthread_stop(sdp->sd_logd_process);
998 * gfs2_lm_mount - mount a locking protocol
999 * @sdp: the filesystem
1000 * @args: mount arguements
1001 * @silent: if 1, don't complain if the FS isn't a GFS2 fs
1006 static int gfs2_lm_mount(struct gfs2_sbd *sdp, int silent)
1008 char *proto = sdp->sd_proto_name;
1009 char *table = sdp->sd_table_name;
1010 int flags = LM_MFLAG_CONV_NODROP;
1013 if (sdp->sd_args.ar_spectator)
1014 flags |= LM_MFLAG_SPECTATOR;
1016 fs_info(sdp, "Trying to join cluster \"%s\", \"%s\"\n", proto, table);
1018 error = gfs2_mount_lockproto(proto, table, sdp->sd_args.ar_hostdata,
1020 GFS2_MIN_LVB_SIZE, flags,
1021 &sdp->sd_lockstruct, &sdp->sd_kobj);
1023 fs_info(sdp, "can't mount proto=%s, table=%s, hostdata=%s\n",
1024 proto, table, sdp->sd_args.ar_hostdata);
1028 if (gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_ops) ||
1029 gfs2_assert_warn(sdp, sdp->sd_lockstruct.ls_lvb_size >=
1030 GFS2_MIN_LVB_SIZE)) {
1031 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1035 if (sdp->sd_args.ar_spectator)
1036 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.s", table);
1038 snprintf(sdp->sd_fsname, GFS2_FSNAME_LEN, "%s.%u", table,
1039 sdp->sd_lockstruct.ls_jid);
1041 fs_info(sdp, "Joined cluster. Now mounting FS...\n");
1043 if ((sdp->sd_lockstruct.ls_flags & LM_LSFLAG_LOCAL) &&
1044 !sdp->sd_args.ar_ignore_local_fs) {
1045 sdp->sd_args.ar_localflocks = 1;
1046 sdp->sd_args.ar_localcaching = 1;
1053 void gfs2_lm_unmount(struct gfs2_sbd *sdp)
1055 if (likely(!test_bit(SDF_SHUTDOWN, &sdp->sd_flags)))
1056 gfs2_unmount_lockproto(&sdp->sd_lockstruct);
1060 * fill_super - Read in superblock
1061 * @sb: The VFS superblock
1062 * @data: Mount options
1063 * @silent: Don't complain if it's not a GFS2 filesystem
1068 static int fill_super(struct super_block *sb, void *data, int silent)
1070 struct gfs2_sbd *sdp;
1071 struct gfs2_holder mount_gh;
1076 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
1080 error = gfs2_mount_args(sdp, (char *)data, 0);
1082 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
1086 sb->s_magic = GFS2_MAGIC;
1087 sb->s_op = &gfs2_super_ops;
1088 sb->s_export_op = &gfs2_export_ops;
1089 sb->s_time_gran = 1;
1090 sb->s_maxbytes = MAX_LFS_FILESIZE;
1092 /* Set up the buffer cache and fill in some fake block size values
1093 to allow us to read-in the on-disk superblock. */
1094 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
1095 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
1096 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
1097 GFS2_BASIC_BLOCK_SHIFT;
1098 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
1100 error = init_names(sdp, silent);
1104 gfs2_create_debugfs_file(sdp);
1106 error = gfs2_sys_fs_add(sdp);
1110 error = gfs2_lm_mount(sdp, silent);
1114 error = init_locking(sdp, &mount_gh, DO);
1118 error = init_sb(sdp, silent);
1122 error = init_inodes(sdp, DO);
1126 error = init_per_node(sdp, DO);
1130 error = gfs2_statfs_init(sdp);
1132 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
1136 error = init_threads(sdp, DO);
1140 if (!(sb->s_flags & MS_RDONLY)) {
1141 error = gfs2_make_fs_rw(sdp);
1143 fs_err(sdp, "can't make FS RW: %d\n", error);
1148 gfs2_glock_dq_uninit(&mount_gh);
1153 init_threads(sdp, UNDO);
1155 init_per_node(sdp, UNDO);
1157 init_inodes(sdp, UNDO);
1159 if (sdp->sd_root_dir)
1160 dput(sdp->sd_root_dir);
1161 if (sdp->sd_master_dir)
1162 dput(sdp->sd_master_dir);
1165 init_locking(sdp, &mount_gh, UNDO);
1167 gfs2_gl_hash_clear(sdp);
1168 gfs2_lm_unmount(sdp);
1169 while (invalidate_inodes(sb))
1172 gfs2_sys_fs_del(sdp);
1174 gfs2_delete_debugfs_file(sdp);
1176 sb->s_fs_info = NULL;
1180 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
1181 const char *dev_name, void *data, struct vfsmount *mnt)
1183 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
1186 static struct super_block *get_gfs2_sb(const char *dev_name)
1188 struct super_block *sb;
1189 struct nameidata nd;
1192 error = path_lookup(dev_name, LOOKUP_FOLLOW, &nd);
1194 printk(KERN_WARNING "GFS2: path_lookup on %s returned error %d\n",
1198 sb = nd.path.dentry->d_inode->i_sb;
1199 if (sb && (sb->s_type == &gfs2_fs_type))
1200 atomic_inc(&sb->s_active);
1207 static int gfs2_get_sb_meta(struct file_system_type *fs_type, int flags,
1208 const char *dev_name, void *data, struct vfsmount *mnt)
1210 struct super_block *sb = NULL;
1211 struct gfs2_sbd *sdp;
1213 sb = get_gfs2_sb(dev_name);
1215 printk(KERN_WARNING "GFS2: gfs2 mount does not exist\n");
1218 sdp = sb->s_fs_info;
1220 mnt->mnt_root = dget(sdp->sd_master_dir);
1224 static void gfs2_kill_sb(struct super_block *sb)
1226 struct gfs2_sbd *sdp = sb->s_fs_info;
1228 gfs2_meta_syncfs(sdp);
1229 dput(sdp->sd_root_dir);
1230 dput(sdp->sd_master_dir);
1231 sdp->sd_root_dir = NULL;
1232 sdp->sd_master_dir = NULL;
1234 shrink_dcache_sb(sb);
1235 kill_block_super(sb);
1237 gfs2_delete_debugfs_file(sdp);
1240 struct file_system_type gfs2_fs_type = {
1242 .fs_flags = FS_REQUIRES_DEV,
1243 .get_sb = gfs2_get_sb,
1244 .kill_sb = gfs2_kill_sb,
1245 .owner = THIS_MODULE,
1248 struct file_system_type gfs2meta_fs_type = {
1250 .fs_flags = FS_REQUIRES_DEV,
1251 .get_sb = gfs2_get_sb_meta,
1252 .owner = THIS_MODULE,