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"
42 extern struct dentry_operations gfs2_dops;
44 static struct gfs2_sbd *init_sbd(struct super_block *sb)
49 sdp = vmalloc(sizeof(struct gfs2_sbd));
53 memset(sdp, 0, sizeof(struct gfs2_sbd));
58 gfs2_tune_init(&sdp->sd_tune);
60 for (x = 0; x < GFS2_GL_HASH_SIZE; x++) {
61 sdp->sd_gl_hash[x].hb_lock = RW_LOCK_UNLOCKED;
62 INIT_LIST_HEAD(&sdp->sd_gl_hash[x].hb_list);
64 INIT_LIST_HEAD(&sdp->sd_reclaim_list);
65 spin_lock_init(&sdp->sd_reclaim_lock);
66 init_waitqueue_head(&sdp->sd_reclaim_wq);
67 mutex_init(&sdp->sd_invalidate_inodes_mutex);
69 mutex_init(&sdp->sd_inum_mutex);
70 spin_lock_init(&sdp->sd_statfs_spin);
71 mutex_init(&sdp->sd_statfs_mutex);
73 spin_lock_init(&sdp->sd_rindex_spin);
74 mutex_init(&sdp->sd_rindex_mutex);
75 INIT_LIST_HEAD(&sdp->sd_rindex_list);
76 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
77 INIT_LIST_HEAD(&sdp->sd_rindex_recent_list);
79 INIT_LIST_HEAD(&sdp->sd_jindex_list);
80 spin_lock_init(&sdp->sd_jindex_spin);
81 mutex_init(&sdp->sd_jindex_mutex);
83 INIT_LIST_HEAD(&sdp->sd_unlinked_list);
84 spin_lock_init(&sdp->sd_unlinked_spin);
85 mutex_init(&sdp->sd_unlinked_mutex);
87 INIT_LIST_HEAD(&sdp->sd_quota_list);
88 spin_lock_init(&sdp->sd_quota_spin);
89 mutex_init(&sdp->sd_quota_mutex);
91 spin_lock_init(&sdp->sd_log_lock);
93 INIT_LIST_HEAD(&sdp->sd_log_le_gl);
94 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
95 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
96 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
97 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
99 mutex_init(&sdp->sd_log_reserve_mutex);
100 INIT_LIST_HEAD(&sdp->sd_ail1_list);
101 INIT_LIST_HEAD(&sdp->sd_ail2_list);
103 init_rwsem(&sdp->sd_log_flush_lock);
104 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
106 INIT_LIST_HEAD(&sdp->sd_revoke_list);
108 mutex_init(&sdp->sd_freeze_lock);
113 static void init_vfs(struct super_block *sb, unsigned noatime)
115 struct gfs2_sbd *sdp = sb->s_fs_info;
117 sb->s_magic = GFS2_MAGIC;
118 sb->s_op = &gfs2_super_ops;
119 sb->s_export_op = &gfs2_export_ops;
120 sb->s_maxbytes = MAX_LFS_FILESIZE;
122 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
123 set_bit(noatime, &sdp->sd_flags);
125 /* Don't let the VFS update atimes. GFS2 handles this itself. */
126 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
129 static int init_names(struct gfs2_sbd *sdp, int silent)
131 struct gfs2_sb *sb = NULL;
135 proto = sdp->sd_args.ar_lockproto;
136 table = sdp->sd_args.ar_locktable;
138 /* Try to autodetect */
140 if (!proto[0] || !table[0]) {
141 struct buffer_head *bh;
142 bh = sb_getblk(sdp->sd_vfs,
143 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
145 clear_buffer_uptodate(bh);
146 clear_buffer_dirty(bh);
148 ll_rw_block(READ, 1, &bh);
151 if (!buffer_uptodate(bh)) {
156 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
161 gfs2_sb_in(sb, bh->b_data);
164 error = gfs2_check_sb(sdp, sb, silent);
169 proto = sb->sb_lockproto;
171 table = sb->sb_locktable;
175 table = sdp->sd_vfs->s_id;
177 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
178 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
186 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
189 struct task_struct *p;
195 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
198 fs_err(sdp, "can't start scand thread: %d\n", error);
201 sdp->sd_scand_process = p;
203 for (sdp->sd_glockd_num = 0;
204 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
205 sdp->sd_glockd_num++) {
206 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
209 fs_err(sdp, "can't start glockd thread: %d\n", error);
212 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
215 error = gfs2_glock_nq_num(sdp,
216 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
217 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
220 fs_err(sdp, "can't acquire mount glock: %d\n", error);
224 error = gfs2_glock_nq_num(sdp,
225 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
227 LM_FLAG_NOEXP | GL_EXACT,
230 fs_err(sdp, "can't acquire live glock: %d\n", error);
234 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
235 CREATE, &sdp->sd_rename_gl);
237 fs_err(sdp, "can't create rename glock: %d\n", error);
241 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
242 CREATE, &sdp->sd_trans_gl);
244 fs_err(sdp, "can't create transaction glock: %d\n", error);
247 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
252 gfs2_glock_put(sdp->sd_trans_gl);
255 gfs2_glock_put(sdp->sd_rename_gl);
258 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
261 gfs2_glock_dq_uninit(mount_gh);
264 while (sdp->sd_glockd_num--)
265 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
267 kthread_stop(sdp->sd_scand_process);
272 static struct inode *gfs2_lookup_root(struct gfs2_sbd *sdp,
273 const struct gfs2_inum *inum)
276 struct gfs2_glock *gl;
277 struct gfs2_inode *ip;
280 error = gfs2_glock_get(sdp, inum->no_addr, &gfs2_inode_glops,
283 error = gfs2_inode_get(gl, inum, CREATE, &ip);
285 gfs2_inode_min_init(ip, DT_DIR);
286 inode = gfs2_ip2v(ip);
293 return ERR_PTR(error);
296 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
298 struct super_block *sb = sdp->sd_vfs;
299 struct gfs2_holder sb_gh;
300 struct gfs2_inum *inum;
308 error = gfs2_glock_nq_num(sdp,
309 GFS2_SB_LOCK, &gfs2_meta_glops,
310 LM_ST_SHARED, 0, &sb_gh);
312 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
316 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
318 fs_err(sdp, "can't read superblock: %d\n", error);
322 /* Set up the buffer cache and SB for real */
323 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
325 fs_err(sdp, "FS block size (%u) is too small for device "
327 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
330 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
332 fs_err(sdp, "FS block size (%u) is too big for machine "
334 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
338 /* Get rid of buffers from the original block size */
339 sb_gh.gh_gl->gl_ops->go_inval(sb_gh.gh_gl, DIO_METADATA | DIO_DATA);
340 sb_gh.gh_gl->gl_aspace->i_blkbits = sdp->sd_sb.sb_bsize_shift;
342 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
344 /* Get the root inode */
345 inum = &sdp->sd_sb.sb_root_dir;
346 if (sb->s_type == &gfs2meta_fs_type)
347 inum = &sdp->sd_sb.sb_master_dir;
348 inode = gfs2_lookup_root(sdp, inum);
350 error = PTR_ERR(inode);
351 fs_err(sdp, "can't read in root inode: %d\n", error);
355 sb->s_root = d_alloc_root(inode);
357 fs_err(sdp, "can't get root dentry\n");
361 sb->s_root->d_op = &gfs2_dops;
363 gfs2_glock_dq_uninit(&sb_gh);
367 static int init_journal(struct gfs2_sbd *sdp, int undo)
369 struct gfs2_holder ji_gh;
370 struct task_struct *p;
371 struct gfs2_inode *ip;
380 sdp->sd_jindex = gfs2_lookup_simple(sdp->sd_master_dir, "jindex");
381 if (IS_ERR(sdp->sd_jindex)) {
382 fs_err(sdp, "can't lookup journal index: %d\n", error);
383 return PTR_ERR(sdp->sd_jindex);
385 ip = sdp->sd_jindex->u.generic_ip;
386 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
388 /* Load in the journal index special file */
390 error = gfs2_jindex_hold(sdp, &ji_gh);
392 fs_err(sdp, "can't read journal index: %d\n", error);
397 if (!gfs2_jindex_size(sdp)) {
398 fs_err(sdp, "no journals!\n");
402 if (sdp->sd_args.ar_spectator) {
403 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
404 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
406 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
407 fs_err(sdp, "can't mount journal #%u\n",
408 sdp->sd_lockstruct.ls_jid);
409 fs_err(sdp, "there are only %u journals (0 - %u)\n",
410 gfs2_jindex_size(sdp),
411 gfs2_jindex_size(sdp) - 1);
414 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
416 error = gfs2_glock_nq_num(sdp,
417 sdp->sd_lockstruct.ls_jid,
419 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
420 &sdp->sd_journal_gh);
422 fs_err(sdp, "can't acquire journal glock: %d\n", error);
426 ip = sdp->sd_jdesc->jd_inode->u.generic_ip;
427 error = gfs2_glock_nq_init(ip->i_gl,
429 LM_FLAG_NOEXP | GL_EXACT,
432 fs_err(sdp, "can't acquire journal inode glock: %d\n",
434 goto fail_journal_gh;
437 error = gfs2_jdesc_check(sdp->sd_jdesc);
439 fs_err(sdp, "my journal (%u) is bad: %d\n",
440 sdp->sd_jdesc->jd_jid, error);
443 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
446 if (sdp->sd_lockstruct.ls_first) {
448 for (x = 0; x < sdp->sd_journals; x++) {
449 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x));
451 fs_err(sdp, "error recovering journal %u: %d\n",
457 gfs2_lm_others_may_mount(sdp);
458 } else if (!sdp->sd_args.ar_spectator) {
459 error = gfs2_recover_journal(sdp->sd_jdesc);
461 fs_err(sdp, "error recovering my journal: %d\n", error);
466 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
467 gfs2_glock_dq_uninit(&ji_gh);
470 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
473 fs_err(sdp, "can't start recoverd thread: %d\n", error);
476 sdp->sd_recoverd_process = p;
481 kthread_stop(sdp->sd_recoverd_process);
484 if (!sdp->sd_args.ar_spectator)
485 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
488 if (!sdp->sd_args.ar_spectator)
489 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
492 gfs2_jindex_free(sdp);
494 gfs2_glock_dq_uninit(&ji_gh);
497 iput(sdp->sd_jindex);
503 static int init_inodes(struct gfs2_sbd *sdp, int undo)
506 struct gfs2_inode *ip;
512 inode = gfs2_lookup_root(sdp, &sdp->sd_sb.sb_master_dir);
514 error = PTR_ERR(inode);
515 fs_err(sdp, "can't read in master directory: %d\n", error);
518 sdp->sd_master_dir = inode;
520 error = init_journal(sdp, undo);
524 /* Read in the master inode number inode */
525 sdp->sd_inum_inode = gfs2_lookup_simple(sdp->sd_master_dir, "inum");
526 if (IS_ERR(sdp->sd_inum_inode)) {
527 error = PTR_ERR(sdp->sd_inum_inode);
528 fs_err(sdp, "can't read in inum inode: %d\n", error);
533 /* Read in the master statfs inode */
534 sdp->sd_statfs_inode = gfs2_lookup_simple(sdp->sd_master_dir, "statfs");
535 if (IS_ERR(sdp->sd_statfs_inode)) {
536 error = PTR_ERR(sdp->sd_statfs_inode);
537 fs_err(sdp, "can't read in statfs inode: %d\n", error);
541 /* Read in the resource index inode */
542 sdp->sd_rindex = gfs2_lookup_simple(sdp->sd_master_dir, "rindex");
543 if (IS_ERR(sdp->sd_rindex)) {
544 error = PTR_ERR(sdp->sd_rindex);
545 fs_err(sdp, "can't get resource index inode: %d\n", error);
548 ip = sdp->sd_rindex->u.generic_ip;
549 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
550 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
552 /* Read in the quota inode */
553 sdp->sd_quota_inode = gfs2_lookup_simple(sdp->sd_master_dir, "quota");
554 if (IS_ERR(sdp->sd_quota_inode)) {
555 error = PTR_ERR(sdp->sd_quota_inode);
556 fs_err(sdp, "can't get quota file inode: %d\n", error);
562 iput(sdp->sd_quota_inode);
565 gfs2_clear_rgrpd(sdp);
566 iput(sdp->sd_rindex);
569 iput(sdp->sd_statfs_inode);
572 iput(sdp->sd_inum_inode);
574 init_journal(sdp, UNDO);
576 iput(sdp->sd_master_dir);
581 static int init_per_node(struct gfs2_sbd *sdp, int undo)
583 struct inode *pn = NULL;
586 struct gfs2_inode *ip;
588 if (sdp->sd_args.ar_spectator)
594 pn = gfs2_lookup_simple(sdp->sd_master_dir, "per_node");
597 fs_err(sdp, "can't find per_node directory: %d\n", error);
601 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
602 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
603 if (IS_ERR(sdp->sd_ir_inode)) {
604 error = PTR_ERR(sdp->sd_ir_inode);
605 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
609 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
610 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
611 if (IS_ERR(sdp->sd_sc_inode)) {
612 error = PTR_ERR(sdp->sd_sc_inode);
613 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
617 sprintf(buf, "unlinked_tag%u", sdp->sd_jdesc->jd_jid);
618 sdp->sd_ut_inode = gfs2_lookup_simple(pn, buf);
619 if (IS_ERR(sdp->sd_ut_inode)) {
620 error = PTR_ERR(sdp->sd_ut_inode);
621 fs_err(sdp, "can't find local \"ut\" file: %d\n", error);
625 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
626 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
627 if (IS_ERR(sdp->sd_qc_inode)) {
628 error = PTR_ERR(sdp->sd_qc_inode);
629 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
636 ip = sdp->sd_ir_inode->u.generic_ip;
637 error = gfs2_glock_nq_init(ip->i_gl,
641 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
645 ip = sdp->sd_sc_inode->u.generic_ip;
646 error = gfs2_glock_nq_init(ip->i_gl,
650 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
654 ip = sdp->sd_ut_inode->u.generic_ip;
655 error = gfs2_glock_nq_init(ip->i_gl,
659 fs_err(sdp, "can't lock local \"ut\" file: %d\n", error);
663 ip = sdp->sd_qc_inode->u.generic_ip;
664 error = gfs2_glock_nq_init(ip->i_gl,
668 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
675 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
678 gfs2_glock_dq_uninit(&sdp->sd_ut_gh);
681 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
684 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
687 iput(sdp->sd_qc_inode);
690 iput(sdp->sd_ut_inode);
693 iput(sdp->sd_sc_inode);
696 iput(sdp->sd_ir_inode);
704 static int init_threads(struct gfs2_sbd *sdp, int undo)
706 struct task_struct *p;
712 sdp->sd_log_flush_time = jiffies;
713 sdp->sd_jindex_refresh_time = jiffies;
715 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
718 fs_err(sdp, "can't start logd thread: %d\n", error);
721 sdp->sd_logd_process = p;
723 sdp->sd_statfs_sync_time = jiffies;
724 sdp->sd_quota_sync_time = jiffies;
726 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
729 fs_err(sdp, "can't start quotad thread: %d\n", error);
732 sdp->sd_quotad_process = p;
734 p = kthread_run(gfs2_inoded, sdp, "gfs2_inoded");
737 fs_err(sdp, "can't start inoded thread: %d\n", error);
740 sdp->sd_inoded_process = p;
745 kthread_stop(sdp->sd_inoded_process);
748 kthread_stop(sdp->sd_quotad_process);
751 kthread_stop(sdp->sd_logd_process);
757 * fill_super - Read in superblock
758 * @sb: The VFS superblock
759 * @data: Mount options
760 * @silent: Don't complain if it's not a GFS2 filesystem
765 static int fill_super(struct super_block *sb, void *data, int silent)
767 struct gfs2_sbd *sdp;
768 struct gfs2_holder mount_gh;
773 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
777 error = gfs2_mount_args(sdp, (char *)data, 0);
779 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
783 init_vfs(sb, SDF_NOATIME);
785 /* Set up the buffer cache and fill in some fake block size values
786 to allow us to read-in the on-disk superblock. */
787 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
788 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
789 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
790 GFS2_BASIC_BLOCK_SHIFT;
791 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
793 error = init_names(sdp, silent);
797 error = gfs2_sys_fs_add(sdp);
801 error = gfs2_lm_mount(sdp, silent);
805 error = init_locking(sdp, &mount_gh, DO);
809 error = init_sb(sdp, silent, DO);
813 error = init_inodes(sdp, DO);
817 error = init_per_node(sdp, DO);
821 error = gfs2_statfs_init(sdp);
823 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
827 error = init_threads(sdp, DO);
831 if (!(sb->s_flags & MS_RDONLY)) {
832 error = gfs2_make_fs_rw(sdp);
834 fs_err(sdp, "can't make FS RW: %d\n", error);
839 gfs2_glock_dq_uninit(&mount_gh);
844 init_threads(sdp, UNDO);
847 init_per_node(sdp, UNDO);
850 init_inodes(sdp, UNDO);
853 init_sb(sdp, 0, UNDO);
856 init_locking(sdp, &mount_gh, UNDO);
859 gfs2_gl_hash_clear(sdp, WAIT);
860 gfs2_lm_unmount(sdp);
861 while (invalidate_inodes(sb))
865 gfs2_sys_fs_del(sdp);
869 sb->s_fs_info = NULL;
874 static struct super_block *gfs2_get_sb(struct file_system_type *fs_type,
875 int flags, const char *dev_name,
878 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super);
881 static void gfs2_kill_sb(struct super_block *sb)
883 kill_block_super(sb);
886 struct file_system_type gfs2_fs_type = {
888 .fs_flags = FS_REQUIRES_DEV,
889 .get_sb = gfs2_get_sb,
890 .kill_sb = gfs2_kill_sb,
891 .owner = THIS_MODULE,
894 struct file_system_type gfs2meta_fs_type = {
896 .fs_flags = FS_REQUIRES_DEV,
897 .get_sb = gfs2_get_sb,
898 .kill_sb = gfs2_kill_sb,
899 .owner = THIS_MODULE,