2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2006 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 v.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/vmalloc.h>
16 #include <linux/blkdev.h>
17 #include <linux/kthread.h>
18 #include <linux/gfs2_ondisk.h>
21 #include "lm_interface.h"
29 #include "ops_export.h"
30 #include "ops_fstype.h"
31 #include "ops_super.h"
41 extern struct dentry_operations gfs2_dops;
43 static struct gfs2_sbd *init_sbd(struct super_block *sb)
48 sdp = vmalloc(sizeof(struct gfs2_sbd));
52 memset(sdp, 0, sizeof(struct gfs2_sbd));
57 gfs2_tune_init(&sdp->sd_tune);
59 for (x = 0; x < GFS2_GL_HASH_SIZE; x++) {
60 rwlock_init(&sdp->sd_gl_hash[x].hb_lock);
61 INIT_LIST_HEAD(&sdp->sd_gl_hash[x].hb_list);
63 INIT_LIST_HEAD(&sdp->sd_reclaim_list);
64 spin_lock_init(&sdp->sd_reclaim_lock);
65 init_waitqueue_head(&sdp->sd_reclaim_wq);
66 mutex_init(&sdp->sd_invalidate_inodes_mutex);
68 mutex_init(&sdp->sd_inum_mutex);
69 spin_lock_init(&sdp->sd_statfs_spin);
70 mutex_init(&sdp->sd_statfs_mutex);
72 spin_lock_init(&sdp->sd_rindex_spin);
73 mutex_init(&sdp->sd_rindex_mutex);
74 INIT_LIST_HEAD(&sdp->sd_rindex_list);
75 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
76 INIT_LIST_HEAD(&sdp->sd_rindex_recent_list);
78 INIT_LIST_HEAD(&sdp->sd_jindex_list);
79 spin_lock_init(&sdp->sd_jindex_spin);
80 mutex_init(&sdp->sd_jindex_mutex);
82 INIT_LIST_HEAD(&sdp->sd_quota_list);
83 spin_lock_init(&sdp->sd_quota_spin);
84 mutex_init(&sdp->sd_quota_mutex);
86 spin_lock_init(&sdp->sd_log_lock);
88 INIT_LIST_HEAD(&sdp->sd_log_le_gl);
89 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
90 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
91 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
92 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
94 mutex_init(&sdp->sd_log_reserve_mutex);
95 INIT_LIST_HEAD(&sdp->sd_ail1_list);
96 INIT_LIST_HEAD(&sdp->sd_ail2_list);
98 init_rwsem(&sdp->sd_log_flush_lock);
99 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
101 INIT_LIST_HEAD(&sdp->sd_revoke_list);
103 mutex_init(&sdp->sd_freeze_lock);
108 static void init_vfs(struct super_block *sb, unsigned noatime)
110 struct gfs2_sbd *sdp = sb->s_fs_info;
112 sb->s_magic = GFS2_MAGIC;
113 sb->s_op = &gfs2_super_ops;
114 sb->s_export_op = &gfs2_export_ops;
115 sb->s_maxbytes = MAX_LFS_FILESIZE;
117 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
118 set_bit(noatime, &sdp->sd_flags);
120 /* Don't let the VFS update atimes. GFS2 handles this itself. */
121 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
124 static int init_names(struct gfs2_sbd *sdp, int silent)
126 struct gfs2_sb *sb = NULL;
130 proto = sdp->sd_args.ar_lockproto;
131 table = sdp->sd_args.ar_locktable;
133 /* Try to autodetect */
135 if (!proto[0] || !table[0]) {
136 struct buffer_head *bh;
137 bh = sb_getblk(sdp->sd_vfs,
138 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
140 clear_buffer_uptodate(bh);
141 clear_buffer_dirty(bh);
143 ll_rw_block(READ, 1, &bh);
146 if (!buffer_uptodate(bh)) {
151 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
156 gfs2_sb_in(sb, bh->b_data);
159 error = gfs2_check_sb(sdp, sb, silent);
164 proto = sb->sb_lockproto;
166 table = sb->sb_locktable;
170 table = sdp->sd_vfs->s_id;
172 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
173 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
181 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
184 struct task_struct *p;
190 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
193 fs_err(sdp, "can't start scand thread: %d\n", error);
196 sdp->sd_scand_process = p;
198 for (sdp->sd_glockd_num = 0;
199 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
200 sdp->sd_glockd_num++) {
201 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
204 fs_err(sdp, "can't start glockd thread: %d\n", error);
207 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
210 error = gfs2_glock_nq_num(sdp,
211 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
212 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
215 fs_err(sdp, "can't acquire mount glock: %d\n", error);
219 error = gfs2_glock_nq_num(sdp,
220 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
222 LM_FLAG_NOEXP | GL_EXACT,
225 fs_err(sdp, "can't acquire live glock: %d\n", error);
229 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
230 CREATE, &sdp->sd_rename_gl);
232 fs_err(sdp, "can't create rename glock: %d\n", error);
236 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
237 CREATE, &sdp->sd_trans_gl);
239 fs_err(sdp, "can't create transaction glock: %d\n", error);
242 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
247 gfs2_glock_put(sdp->sd_trans_gl);
250 gfs2_glock_put(sdp->sd_rename_gl);
253 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
256 gfs2_glock_dq_uninit(mount_gh);
259 while (sdp->sd_glockd_num--)
260 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
262 kthread_stop(sdp->sd_scand_process);
267 static struct inode *gfs2_lookup_root(struct super_block *sb,
268 struct gfs2_inum *inum)
270 return gfs2_inode_lookup(sb, inum, DT_DIR);
273 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
275 struct super_block *sb = sdp->sd_vfs;
276 struct gfs2_holder sb_gh;
277 struct gfs2_inum *inum;
289 error = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
290 LM_ST_SHARED, 0, &sb_gh);
292 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
296 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
298 fs_err(sdp, "can't read superblock: %d\n", error);
302 /* Set up the buffer cache and SB for real */
303 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
305 fs_err(sdp, "FS block size (%u) is too small for device "
307 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
310 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
312 fs_err(sdp, "FS block size (%u) is too big for machine "
314 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
317 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
319 /* Get the root inode */
320 inum = &sdp->sd_sb.sb_root_dir;
321 if (sb->s_type == &gfs2meta_fs_type)
322 inum = &sdp->sd_sb.sb_master_dir;
323 inode = gfs2_lookup_root(sb, inum);
325 error = PTR_ERR(inode);
326 fs_err(sdp, "can't read in root inode: %d\n", error);
330 sb->s_root = d_alloc_root(inode);
332 fs_err(sdp, "can't get root dentry\n");
336 sb->s_root->d_op = &gfs2_dops;
338 gfs2_glock_dq_uninit(&sb_gh);
342 static int init_journal(struct gfs2_sbd *sdp, int undo)
344 struct gfs2_holder ji_gh;
345 struct task_struct *p;
346 struct gfs2_inode *ip;
355 sdp->sd_jindex = gfs2_lookup_simple(sdp->sd_master_dir, "jindex");
356 if (IS_ERR(sdp->sd_jindex)) {
357 fs_err(sdp, "can't lookup journal index: %d\n", error);
358 return PTR_ERR(sdp->sd_jindex);
360 ip = GFS2_I(sdp->sd_jindex);
361 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
363 /* Load in the journal index special file */
365 error = gfs2_jindex_hold(sdp, &ji_gh);
367 fs_err(sdp, "can't read journal index: %d\n", error);
372 if (!gfs2_jindex_size(sdp)) {
373 fs_err(sdp, "no journals!\n");
377 if (sdp->sd_args.ar_spectator) {
378 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
379 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
381 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
382 fs_err(sdp, "can't mount journal #%u\n",
383 sdp->sd_lockstruct.ls_jid);
384 fs_err(sdp, "there are only %u journals (0 - %u)\n",
385 gfs2_jindex_size(sdp),
386 gfs2_jindex_size(sdp) - 1);
389 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
391 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
393 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
394 &sdp->sd_journal_gh);
396 fs_err(sdp, "can't acquire journal glock: %d\n", error);
400 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
401 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
402 LM_FLAG_NOEXP | GL_EXACT,
405 fs_err(sdp, "can't acquire journal inode glock: %d\n",
407 goto fail_journal_gh;
410 error = gfs2_jdesc_check(sdp->sd_jdesc);
412 fs_err(sdp, "my journal (%u) is bad: %d\n",
413 sdp->sd_jdesc->jd_jid, error);
416 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
419 if (sdp->sd_lockstruct.ls_first) {
421 for (x = 0; x < sdp->sd_journals; x++) {
422 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
424 fs_err(sdp, "error recovering journal %u: %d\n",
430 gfs2_lm_others_may_mount(sdp);
431 } else if (!sdp->sd_args.ar_spectator) {
432 error = gfs2_recover_journal(sdp->sd_jdesc);
434 fs_err(sdp, "error recovering my journal: %d\n", error);
439 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
440 gfs2_glock_dq_uninit(&ji_gh);
443 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
446 fs_err(sdp, "can't start recoverd thread: %d\n", error);
449 sdp->sd_recoverd_process = p;
454 kthread_stop(sdp->sd_recoverd_process);
457 if (!sdp->sd_args.ar_spectator)
458 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
461 if (!sdp->sd_args.ar_spectator)
462 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
465 gfs2_jindex_free(sdp);
467 gfs2_glock_dq_uninit(&ji_gh);
470 iput(sdp->sd_jindex);
476 static int init_inodes(struct gfs2_sbd *sdp, int undo)
479 struct gfs2_inode *ip;
485 inode = gfs2_lookup_root(sdp->sd_vfs, &sdp->sd_sb.sb_master_dir);
487 error = PTR_ERR(inode);
488 fs_err(sdp, "can't read in master directory: %d\n", error);
491 sdp->sd_master_dir = inode;
493 error = init_journal(sdp, undo);
497 /* Read in the master inode number inode */
498 sdp->sd_inum_inode = gfs2_lookup_simple(sdp->sd_master_dir, "inum");
499 if (IS_ERR(sdp->sd_inum_inode)) {
500 error = PTR_ERR(sdp->sd_inum_inode);
501 fs_err(sdp, "can't read in inum inode: %d\n", error);
506 /* Read in the master statfs inode */
507 sdp->sd_statfs_inode = gfs2_lookup_simple(sdp->sd_master_dir, "statfs");
508 if (IS_ERR(sdp->sd_statfs_inode)) {
509 error = PTR_ERR(sdp->sd_statfs_inode);
510 fs_err(sdp, "can't read in statfs inode: %d\n", error);
514 /* Read in the resource index inode */
515 sdp->sd_rindex = gfs2_lookup_simple(sdp->sd_master_dir, "rindex");
516 if (IS_ERR(sdp->sd_rindex)) {
517 error = PTR_ERR(sdp->sd_rindex);
518 fs_err(sdp, "can't get resource index inode: %d\n", error);
521 ip = GFS2_I(sdp->sd_rindex);
522 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
523 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
525 /* Read in the quota inode */
526 sdp->sd_quota_inode = gfs2_lookup_simple(sdp->sd_master_dir, "quota");
527 if (IS_ERR(sdp->sd_quota_inode)) {
528 error = PTR_ERR(sdp->sd_quota_inode);
529 fs_err(sdp, "can't get quota file inode: %d\n", error);
535 iput(sdp->sd_quota_inode);
538 gfs2_clear_rgrpd(sdp);
539 iput(sdp->sd_rindex);
542 iput(sdp->sd_statfs_inode);
545 iput(sdp->sd_inum_inode);
547 init_journal(sdp, UNDO);
549 iput(sdp->sd_master_dir);
554 static int init_per_node(struct gfs2_sbd *sdp, int undo)
556 struct inode *pn = NULL;
559 struct gfs2_inode *ip;
561 if (sdp->sd_args.ar_spectator)
567 pn = gfs2_lookup_simple(sdp->sd_master_dir, "per_node");
570 fs_err(sdp, "can't find per_node directory: %d\n", error);
574 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
575 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
576 if (IS_ERR(sdp->sd_ir_inode)) {
577 error = PTR_ERR(sdp->sd_ir_inode);
578 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
582 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
583 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
584 if (IS_ERR(sdp->sd_sc_inode)) {
585 error = PTR_ERR(sdp->sd_sc_inode);
586 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
590 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
591 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
592 if (IS_ERR(sdp->sd_qc_inode)) {
593 error = PTR_ERR(sdp->sd_qc_inode);
594 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
601 ip = GFS2_I(sdp->sd_ir_inode);
602 error = gfs2_glock_nq_init(ip->i_gl,
606 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
610 ip = GFS2_I(sdp->sd_sc_inode);
611 error = gfs2_glock_nq_init(ip->i_gl,
615 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
619 ip = GFS2_I(sdp->sd_qc_inode);
620 error = gfs2_glock_nq_init(ip->i_gl,
624 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
631 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
635 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
638 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
641 iput(sdp->sd_qc_inode);
645 iput(sdp->sd_sc_inode);
648 iput(sdp->sd_ir_inode);
656 static int init_threads(struct gfs2_sbd *sdp, int undo)
658 struct task_struct *p;
664 sdp->sd_log_flush_time = jiffies;
665 sdp->sd_jindex_refresh_time = jiffies;
667 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
670 fs_err(sdp, "can't start logd thread: %d\n", error);
673 sdp->sd_logd_process = p;
675 sdp->sd_statfs_sync_time = jiffies;
676 sdp->sd_quota_sync_time = jiffies;
678 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
681 fs_err(sdp, "can't start quotad thread: %d\n", error);
684 sdp->sd_quotad_process = p;
690 kthread_stop(sdp->sd_quotad_process);
692 kthread_stop(sdp->sd_logd_process);
697 * fill_super - Read in superblock
698 * @sb: The VFS superblock
699 * @data: Mount options
700 * @silent: Don't complain if it's not a GFS2 filesystem
705 static int fill_super(struct super_block *sb, void *data, int silent)
707 struct gfs2_sbd *sdp;
708 struct gfs2_holder mount_gh;
713 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
717 error = gfs2_mount_args(sdp, (char *)data, 0);
719 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
723 init_vfs(sb, SDF_NOATIME);
725 /* Set up the buffer cache and fill in some fake block size values
726 to allow us to read-in the on-disk superblock. */
727 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
728 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
729 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
730 GFS2_BASIC_BLOCK_SHIFT;
731 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
733 error = init_names(sdp, silent);
737 error = gfs2_sys_fs_add(sdp);
741 error = gfs2_lm_mount(sdp, silent);
745 error = init_locking(sdp, &mount_gh, DO);
749 error = init_sb(sdp, silent, DO);
753 error = init_inodes(sdp, DO);
757 error = init_per_node(sdp, DO);
761 error = gfs2_statfs_init(sdp);
763 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
767 error = init_threads(sdp, DO);
771 if (!(sb->s_flags & MS_RDONLY)) {
772 error = gfs2_make_fs_rw(sdp);
774 fs_err(sdp, "can't make FS RW: %d\n", error);
779 gfs2_glock_dq_uninit(&mount_gh);
784 init_threads(sdp, UNDO);
787 init_per_node(sdp, UNDO);
790 init_inodes(sdp, UNDO);
793 init_sb(sdp, 0, UNDO);
796 init_locking(sdp, &mount_gh, UNDO);
799 gfs2_gl_hash_clear(sdp, WAIT);
800 gfs2_lm_unmount(sdp);
801 while (invalidate_inodes(sb))
805 gfs2_sys_fs_del(sdp);
809 sb->s_fs_info = NULL;
814 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
815 const char *dev_name, void *data, struct vfsmount *mnt)
817 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
820 static void gfs2_kill_sb(struct super_block *sb)
822 kill_block_super(sb);
825 struct file_system_type gfs2_fs_type = {
827 .fs_flags = FS_REQUIRES_DEV,
828 .get_sb = gfs2_get_sb,
829 .kill_sb = gfs2_kill_sb,
830 .owner = THIS_MODULE,
833 struct file_system_type gfs2meta_fs_type = {
835 .fs_flags = FS_REQUIRES_DEV,
836 .get_sb = gfs2_get_sb,
837 .kill_sb = gfs2_kill_sb,
838 .owner = THIS_MODULE,