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);
67 mutex_init(&sdp->sd_inum_mutex);
68 spin_lock_init(&sdp->sd_statfs_spin);
69 mutex_init(&sdp->sd_statfs_mutex);
71 spin_lock_init(&sdp->sd_rindex_spin);
72 mutex_init(&sdp->sd_rindex_mutex);
73 INIT_LIST_HEAD(&sdp->sd_rindex_list);
74 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
75 INIT_LIST_HEAD(&sdp->sd_rindex_recent_list);
77 INIT_LIST_HEAD(&sdp->sd_jindex_list);
78 spin_lock_init(&sdp->sd_jindex_spin);
79 mutex_init(&sdp->sd_jindex_mutex);
81 INIT_LIST_HEAD(&sdp->sd_quota_list);
82 spin_lock_init(&sdp->sd_quota_spin);
83 mutex_init(&sdp->sd_quota_mutex);
85 spin_lock_init(&sdp->sd_log_lock);
87 INIT_LIST_HEAD(&sdp->sd_log_le_gl);
88 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
89 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
90 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
91 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
93 mutex_init(&sdp->sd_log_reserve_mutex);
94 INIT_LIST_HEAD(&sdp->sd_ail1_list);
95 INIT_LIST_HEAD(&sdp->sd_ail2_list);
97 init_rwsem(&sdp->sd_log_flush_lock);
98 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
100 INIT_LIST_HEAD(&sdp->sd_revoke_list);
102 mutex_init(&sdp->sd_freeze_lock);
107 static void init_vfs(struct super_block *sb, unsigned noatime)
109 struct gfs2_sbd *sdp = sb->s_fs_info;
111 sb->s_magic = GFS2_MAGIC;
112 sb->s_op = &gfs2_super_ops;
113 sb->s_export_op = &gfs2_export_ops;
114 sb->s_maxbytes = MAX_LFS_FILESIZE;
116 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
117 set_bit(noatime, &sdp->sd_flags);
119 /* Don't let the VFS update atimes. GFS2 handles this itself. */
120 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
123 static int init_names(struct gfs2_sbd *sdp, int silent)
125 struct gfs2_sb *sb = NULL;
129 proto = sdp->sd_args.ar_lockproto;
130 table = sdp->sd_args.ar_locktable;
132 /* Try to autodetect */
134 if (!proto[0] || !table[0]) {
135 struct buffer_head *bh;
136 bh = sb_getblk(sdp->sd_vfs,
137 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
139 clear_buffer_uptodate(bh);
140 clear_buffer_dirty(bh);
142 ll_rw_block(READ, 1, &bh);
145 if (!buffer_uptodate(bh)) {
150 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
155 gfs2_sb_in(sb, bh->b_data);
158 error = gfs2_check_sb(sdp, sb, silent);
163 proto = sb->sb_lockproto;
165 table = sb->sb_locktable;
169 table = sdp->sd_vfs->s_id;
171 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
172 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
180 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
183 struct task_struct *p;
189 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
192 fs_err(sdp, "can't start scand thread: %d\n", error);
195 sdp->sd_scand_process = p;
197 for (sdp->sd_glockd_num = 0;
198 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
199 sdp->sd_glockd_num++) {
200 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
203 fs_err(sdp, "can't start glockd thread: %d\n", error);
206 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
209 error = gfs2_glock_nq_num(sdp,
210 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
211 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
214 fs_err(sdp, "can't acquire mount glock: %d\n", error);
218 error = gfs2_glock_nq_num(sdp,
219 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
221 LM_FLAG_NOEXP | GL_EXACT,
224 fs_err(sdp, "can't acquire live glock: %d\n", error);
228 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
229 CREATE, &sdp->sd_rename_gl);
231 fs_err(sdp, "can't create rename glock: %d\n", error);
235 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
236 CREATE, &sdp->sd_trans_gl);
238 fs_err(sdp, "can't create transaction glock: %d\n", error);
241 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
246 gfs2_glock_put(sdp->sd_trans_gl);
249 gfs2_glock_put(sdp->sd_rename_gl);
252 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
255 gfs2_glock_dq_uninit(mount_gh);
258 while (sdp->sd_glockd_num--)
259 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
261 kthread_stop(sdp->sd_scand_process);
266 static struct inode *gfs2_lookup_root(struct super_block *sb,
267 struct gfs2_inum *inum)
269 return gfs2_inode_lookup(sb, inum, DT_DIR);
272 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
274 struct super_block *sb = sdp->sd_vfs;
275 struct gfs2_holder sb_gh;
276 struct gfs2_inum *inum;
288 error = gfs2_glock_nq_num(sdp, GFS2_SB_LOCK, &gfs2_meta_glops,
289 LM_ST_SHARED, 0, &sb_gh);
291 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
295 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
297 fs_err(sdp, "can't read superblock: %d\n", error);
301 /* Set up the buffer cache and SB for real */
302 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
304 fs_err(sdp, "FS block size (%u) is too small for device "
306 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
309 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
311 fs_err(sdp, "FS block size (%u) is too big for machine "
313 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
316 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
318 /* Get the root inode */
319 inum = &sdp->sd_sb.sb_root_dir;
320 if (sb->s_type == &gfs2meta_fs_type)
321 inum = &sdp->sd_sb.sb_master_dir;
322 inode = gfs2_lookup_root(sb, inum);
324 error = PTR_ERR(inode);
325 fs_err(sdp, "can't read in root inode: %d\n", error);
329 sb->s_root = d_alloc_root(inode);
331 fs_err(sdp, "can't get root dentry\n");
335 sb->s_root->d_op = &gfs2_dops;
337 gfs2_glock_dq_uninit(&sb_gh);
341 static int init_journal(struct gfs2_sbd *sdp, int undo)
343 struct gfs2_holder ji_gh;
344 struct task_struct *p;
345 struct gfs2_inode *ip;
354 sdp->sd_jindex = gfs2_lookup_simple(sdp->sd_master_dir, "jindex");
355 if (IS_ERR(sdp->sd_jindex)) {
356 fs_err(sdp, "can't lookup journal index: %d\n", error);
357 return PTR_ERR(sdp->sd_jindex);
359 ip = GFS2_I(sdp->sd_jindex);
360 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
362 /* Load in the journal index special file */
364 error = gfs2_jindex_hold(sdp, &ji_gh);
366 fs_err(sdp, "can't read journal index: %d\n", error);
371 if (!gfs2_jindex_size(sdp)) {
372 fs_err(sdp, "no journals!\n");
376 if (sdp->sd_args.ar_spectator) {
377 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
378 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
380 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
381 fs_err(sdp, "can't mount journal #%u\n",
382 sdp->sd_lockstruct.ls_jid);
383 fs_err(sdp, "there are only %u journals (0 - %u)\n",
384 gfs2_jindex_size(sdp),
385 gfs2_jindex_size(sdp) - 1);
388 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
390 error = gfs2_glock_nq_num(sdp, sdp->sd_lockstruct.ls_jid,
392 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
393 &sdp->sd_journal_gh);
395 fs_err(sdp, "can't acquire journal glock: %d\n", error);
399 ip = GFS2_I(sdp->sd_jdesc->jd_inode);
400 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED,
401 LM_FLAG_NOEXP | GL_EXACT,
404 fs_err(sdp, "can't acquire journal inode glock: %d\n",
406 goto fail_journal_gh;
409 error = gfs2_jdesc_check(sdp->sd_jdesc);
411 fs_err(sdp, "my journal (%u) is bad: %d\n",
412 sdp->sd_jdesc->jd_jid, error);
415 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
418 if (sdp->sd_lockstruct.ls_first) {
420 for (x = 0; x < sdp->sd_journals; x++) {
421 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
423 fs_err(sdp, "error recovering journal %u: %d\n",
429 gfs2_lm_others_may_mount(sdp);
430 } else if (!sdp->sd_args.ar_spectator) {
431 error = gfs2_recover_journal(sdp->sd_jdesc);
433 fs_err(sdp, "error recovering my journal: %d\n", error);
438 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
439 gfs2_glock_dq_uninit(&ji_gh);
442 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
445 fs_err(sdp, "can't start recoverd thread: %d\n", error);
448 sdp->sd_recoverd_process = p;
453 kthread_stop(sdp->sd_recoverd_process);
456 if (!sdp->sd_args.ar_spectator)
457 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
460 if (!sdp->sd_args.ar_spectator)
461 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
464 gfs2_jindex_free(sdp);
466 gfs2_glock_dq_uninit(&ji_gh);
469 iput(sdp->sd_jindex);
475 static int init_inodes(struct gfs2_sbd *sdp, int undo)
478 struct gfs2_inode *ip;
484 inode = gfs2_lookup_root(sdp->sd_vfs, &sdp->sd_sb.sb_master_dir);
486 error = PTR_ERR(inode);
487 fs_err(sdp, "can't read in master directory: %d\n", error);
490 sdp->sd_master_dir = inode;
492 error = init_journal(sdp, undo);
496 /* Read in the master inode number inode */
497 sdp->sd_inum_inode = gfs2_lookup_simple(sdp->sd_master_dir, "inum");
498 if (IS_ERR(sdp->sd_inum_inode)) {
499 error = PTR_ERR(sdp->sd_inum_inode);
500 fs_err(sdp, "can't read in inum inode: %d\n", error);
505 /* Read in the master statfs inode */
506 sdp->sd_statfs_inode = gfs2_lookup_simple(sdp->sd_master_dir, "statfs");
507 if (IS_ERR(sdp->sd_statfs_inode)) {
508 error = PTR_ERR(sdp->sd_statfs_inode);
509 fs_err(sdp, "can't read in statfs inode: %d\n", error);
513 /* Read in the resource index inode */
514 sdp->sd_rindex = gfs2_lookup_simple(sdp->sd_master_dir, "rindex");
515 if (IS_ERR(sdp->sd_rindex)) {
516 error = PTR_ERR(sdp->sd_rindex);
517 fs_err(sdp, "can't get resource index inode: %d\n", error);
520 ip = GFS2_I(sdp->sd_rindex);
521 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
522 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
524 /* Read in the quota inode */
525 sdp->sd_quota_inode = gfs2_lookup_simple(sdp->sd_master_dir, "quota");
526 if (IS_ERR(sdp->sd_quota_inode)) {
527 error = PTR_ERR(sdp->sd_quota_inode);
528 fs_err(sdp, "can't get quota file inode: %d\n", error);
534 iput(sdp->sd_quota_inode);
537 gfs2_clear_rgrpd(sdp);
538 iput(sdp->sd_rindex);
541 iput(sdp->sd_statfs_inode);
544 iput(sdp->sd_inum_inode);
546 init_journal(sdp, UNDO);
548 iput(sdp->sd_master_dir);
553 static int init_per_node(struct gfs2_sbd *sdp, int undo)
555 struct inode *pn = NULL;
558 struct gfs2_inode *ip;
560 if (sdp->sd_args.ar_spectator)
566 pn = gfs2_lookup_simple(sdp->sd_master_dir, "per_node");
569 fs_err(sdp, "can't find per_node directory: %d\n", error);
573 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
574 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
575 if (IS_ERR(sdp->sd_ir_inode)) {
576 error = PTR_ERR(sdp->sd_ir_inode);
577 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
581 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
582 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
583 if (IS_ERR(sdp->sd_sc_inode)) {
584 error = PTR_ERR(sdp->sd_sc_inode);
585 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
589 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
590 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
591 if (IS_ERR(sdp->sd_qc_inode)) {
592 error = PTR_ERR(sdp->sd_qc_inode);
593 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
600 ip = GFS2_I(sdp->sd_ir_inode);
601 error = gfs2_glock_nq_init(ip->i_gl,
605 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
609 ip = GFS2_I(sdp->sd_sc_inode);
610 error = gfs2_glock_nq_init(ip->i_gl,
614 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
618 ip = GFS2_I(sdp->sd_qc_inode);
619 error = gfs2_glock_nq_init(ip->i_gl,
623 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
630 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
634 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
637 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
640 iput(sdp->sd_qc_inode);
644 iput(sdp->sd_sc_inode);
647 iput(sdp->sd_ir_inode);
655 static int init_threads(struct gfs2_sbd *sdp, int undo)
657 struct task_struct *p;
663 sdp->sd_log_flush_time = jiffies;
664 sdp->sd_jindex_refresh_time = jiffies;
666 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
669 fs_err(sdp, "can't start logd thread: %d\n", error);
672 sdp->sd_logd_process = p;
674 sdp->sd_statfs_sync_time = jiffies;
675 sdp->sd_quota_sync_time = jiffies;
677 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
680 fs_err(sdp, "can't start quotad thread: %d\n", error);
683 sdp->sd_quotad_process = p;
689 kthread_stop(sdp->sd_quotad_process);
691 kthread_stop(sdp->sd_logd_process);
696 * fill_super - Read in superblock
697 * @sb: The VFS superblock
698 * @data: Mount options
699 * @silent: Don't complain if it's not a GFS2 filesystem
704 static int fill_super(struct super_block *sb, void *data, int silent)
706 struct gfs2_sbd *sdp;
707 struct gfs2_holder mount_gh;
712 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
716 error = gfs2_mount_args(sdp, (char *)data, 0);
718 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
722 init_vfs(sb, SDF_NOATIME);
724 /* Set up the buffer cache and fill in some fake block size values
725 to allow us to read-in the on-disk superblock. */
726 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
727 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
728 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
729 GFS2_BASIC_BLOCK_SHIFT;
730 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
732 error = init_names(sdp, silent);
736 error = gfs2_sys_fs_add(sdp);
740 error = gfs2_lm_mount(sdp, silent);
744 error = init_locking(sdp, &mount_gh, DO);
748 error = init_sb(sdp, silent, DO);
752 error = init_inodes(sdp, DO);
756 error = init_per_node(sdp, DO);
760 error = gfs2_statfs_init(sdp);
762 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
766 error = init_threads(sdp, DO);
770 if (!(sb->s_flags & MS_RDONLY)) {
771 error = gfs2_make_fs_rw(sdp);
773 fs_err(sdp, "can't make FS RW: %d\n", error);
778 gfs2_glock_dq_uninit(&mount_gh);
783 init_threads(sdp, UNDO);
786 init_per_node(sdp, UNDO);
789 init_inodes(sdp, UNDO);
792 init_sb(sdp, 0, UNDO);
795 init_locking(sdp, &mount_gh, UNDO);
798 gfs2_gl_hash_clear(sdp, WAIT);
799 gfs2_lm_unmount(sdp);
800 while (invalidate_inodes(sb))
804 gfs2_sys_fs_del(sdp);
808 sb->s_fs_info = NULL;
813 static int gfs2_get_sb(struct file_system_type *fs_type, int flags,
814 const char *dev_name, void *data, struct vfsmount *mnt)
816 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super, mnt);
819 static void gfs2_kill_sb(struct super_block *sb)
821 kill_block_super(sb);
824 struct file_system_type gfs2_fs_type = {
826 .fs_flags = FS_REQUIRES_DEV,
827 .get_sb = gfs2_get_sb,
828 .kill_sb = gfs2_kill_sb,
829 .owner = THIS_MODULE,
832 struct file_system_type gfs2meta_fs_type = {
834 .fs_flags = FS_REQUIRES_DEV,
835 .get_sb = gfs2_get_sb,
836 .kill_sb = gfs2_kill_sb,
837 .owner = THIS_MODULE,