2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2005 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>
19 #include <asm/semaphore.h>
22 #include "lm_interface.h"
30 #include "ops_export.h"
31 #include "ops_fstype.h"
32 #include "ops_super.h"
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 sdp->sd_gl_hash[x].hb_lock = RW_LOCK_UNLOCKED;
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_unlinked_list);
83 spin_lock_init(&sdp->sd_unlinked_spin);
84 mutex_init(&sdp->sd_unlinked_mutex);
86 INIT_LIST_HEAD(&sdp->sd_quota_list);
87 spin_lock_init(&sdp->sd_quota_spin);
88 mutex_init(&sdp->sd_quota_mutex);
90 spin_lock_init(&sdp->sd_log_lock);
91 init_waitqueue_head(&sdp->sd_log_trans_wq);
92 init_waitqueue_head(&sdp->sd_log_flush_wq);
94 INIT_LIST_HEAD(&sdp->sd_log_le_gl);
95 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
96 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
97 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
98 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
100 mutex_init(&sdp->sd_log_reserve_mutex);
101 INIT_LIST_HEAD(&sdp->sd_ail1_list);
102 INIT_LIST_HEAD(&sdp->sd_ail2_list);
104 mutex_init(&sdp->sd_log_flush_lock);
105 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
107 INIT_LIST_HEAD(&sdp->sd_revoke_list);
109 mutex_init(&sdp->sd_freeze_lock);
114 static void init_vfs(struct super_block *sb, unsigned noatime)
116 struct gfs2_sbd *sdp = sb->s_fs_info;
118 sb->s_magic = GFS2_MAGIC;
119 sb->s_op = &gfs2_super_ops;
120 sb->s_export_op = &gfs2_export_ops;
121 sb->s_maxbytes = MAX_LFS_FILESIZE;
123 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
124 set_bit(noatime, &sdp->sd_flags);
126 /* Don't let the VFS update atimes. GFS2 handles this itself. */
127 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
130 static int init_names(struct gfs2_sbd *sdp, int silent)
132 struct gfs2_sb *sb = NULL;
136 proto = sdp->sd_args.ar_lockproto;
137 table = sdp->sd_args.ar_locktable;
139 /* Try to autodetect */
141 if (!proto[0] || !table[0]) {
142 struct buffer_head *bh;
143 bh = sb_getblk(sdp->sd_vfs,
144 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
146 clear_buffer_uptodate(bh);
147 clear_buffer_dirty(bh);
149 ll_rw_block(READ, 1, &bh);
152 if (!buffer_uptodate(bh)) {
157 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
162 gfs2_sb_in(sb, bh->b_data);
165 error = gfs2_check_sb(sdp, sb, silent);
170 proto = sb->sb_lockproto;
172 table = sb->sb_locktable;
176 table = sdp->sd_vfs->s_id;
178 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
179 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
187 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
190 struct task_struct *p;
196 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
199 fs_err(sdp, "can't start scand thread: %d\n", error);
202 sdp->sd_scand_process = p;
204 for (sdp->sd_glockd_num = 0;
205 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
206 sdp->sd_glockd_num++) {
207 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
210 fs_err(sdp, "can't start glockd thread: %d\n", error);
213 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
216 error = gfs2_glock_nq_num(sdp,
217 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
218 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
221 fs_err(sdp, "can't acquire mount glock: %d\n", error);
225 error = gfs2_glock_nq_num(sdp,
226 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
228 LM_FLAG_NOEXP | GL_EXACT | GL_NEVER_RECURSE,
231 fs_err(sdp, "can't acquire live glock: %d\n", error);
235 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
236 CREATE, &sdp->sd_rename_gl);
238 fs_err(sdp, "can't create rename glock: %d\n", error);
242 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
243 CREATE, &sdp->sd_trans_gl);
245 fs_err(sdp, "can't create transaction glock: %d\n", error);
248 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
253 gfs2_glock_put(sdp->sd_trans_gl);
256 gfs2_glock_put(sdp->sd_rename_gl);
259 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
262 gfs2_glock_dq_uninit(mount_gh);
265 while (sdp->sd_glockd_num--)
266 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
268 kthread_stop(sdp->sd_scand_process);
273 static struct inode *gfs2_lookup_root(struct gfs2_sbd *sdp,
274 const struct gfs2_inum *inum)
277 struct gfs2_glock *gl;
278 struct gfs2_inode *ip;
281 error = gfs2_glock_get(sdp, inum->no_addr,
282 &gfs2_inode_glops, CREATE, &gl);
284 error = gfs2_inode_get(gl, inum, CREATE, &ip);
286 gfs2_inode_min_init(ip, DT_DIR);
287 inode = gfs2_ip2v(ip);
294 return ERR_PTR(error);
297 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
299 struct super_block *sb = sdp->sd_vfs;
300 struct gfs2_holder sb_gh;
301 struct gfs2_inum *inum;
309 error = gfs2_glock_nq_num(sdp,
310 GFS2_SB_LOCK, &gfs2_meta_glops,
311 LM_ST_SHARED, 0, &sb_gh);
313 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
317 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
319 fs_err(sdp, "can't read superblock: %d\n", error);
323 /* Set up the buffer cache and SB for real */
324 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
326 fs_err(sdp, "FS block size (%u) is too small for device "
328 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
331 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
333 fs_err(sdp, "FS block size (%u) is too big for machine "
335 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
339 /* Get rid of buffers from the original block size */
340 sb_gh.gh_gl->gl_ops->go_inval(sb_gh.gh_gl, DIO_METADATA | DIO_DATA);
341 sb_gh.gh_gl->gl_aspace->i_blkbits = sdp->sd_sb.sb_bsize_shift;
343 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
345 /* Get the root inode */
346 inum = &sdp->sd_sb.sb_root_dir;
347 if (sb->s_type == &gfs2meta_fs_type)
348 inum = &sdp->sd_sb.sb_master_dir;
349 inode = gfs2_lookup_root(sdp, inum);
351 error = PTR_ERR(inode);
352 fs_err(sdp, "can't read in root inode: %d\n", error);
356 sb->s_root = d_alloc_root(inode);
358 fs_err(sdp, "can't get root dentry\n");
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),
452 fs_err(sdp, "error recovering journal %u: %d\n",
458 gfs2_lm_others_may_mount(sdp);
459 } else if (!sdp->sd_args.ar_spectator) {
460 error = gfs2_recover_journal(sdp->sd_jdesc, WAIT);
462 fs_err(sdp, "error recovering my journal: %d\n", error);
467 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
468 gfs2_glock_dq_uninit(&ji_gh);
471 /* Disown my Journal glock */
473 sdp->sd_journal_gh.gh_owner = NULL;
474 sdp->sd_jinode_gh.gh_owner = NULL;
476 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
479 fs_err(sdp, "can't start recoverd thread: %d\n", error);
482 sdp->sd_recoverd_process = p;
487 kthread_stop(sdp->sd_recoverd_process);
490 if (!sdp->sd_args.ar_spectator)
491 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
494 if (!sdp->sd_args.ar_spectator)
495 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
498 gfs2_jindex_free(sdp);
500 gfs2_glock_dq_uninit(&ji_gh);
503 iput(sdp->sd_jindex);
509 static int init_inodes(struct gfs2_sbd *sdp, int undo)
512 struct gfs2_inode *ip;
518 inode = gfs2_lookup_root(sdp, &sdp->sd_sb.sb_master_dir);
520 error = PTR_ERR(inode);
521 fs_err(sdp, "can't read in master directory: %d\n", error);
524 sdp->sd_master_dir = inode;
526 error = init_journal(sdp, undo);
530 /* Read in the master inode number inode */
531 sdp->sd_inum_inode = gfs2_lookup_simple(sdp->sd_master_dir, "inum");
532 if (IS_ERR(sdp->sd_inum_inode)) {
533 error = PTR_ERR(sdp->sd_inum_inode);
534 fs_err(sdp, "can't read in inum inode: %d\n", error);
539 /* Read in the master statfs inode */
540 sdp->sd_statfs_inode = gfs2_lookup_simple(sdp->sd_master_dir, "statfs");
541 if (IS_ERR(sdp->sd_statfs_inode)) {
542 error = PTR_ERR(sdp->sd_statfs_inode);
543 fs_err(sdp, "can't read in statfs inode: %d\n", error);
547 /* Read in the resource index inode */
548 sdp->sd_rindex = gfs2_lookup_simple(sdp->sd_master_dir, "rindex");
549 if (IS_ERR(sdp->sd_rindex)) {
550 error = PTR_ERR(sdp->sd_rindex);
551 fs_err(sdp, "can't get resource index inode: %d\n", error);
554 ip = sdp->sd_rindex->u.generic_ip;
555 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
556 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
558 /* Read in the quota inode */
559 sdp->sd_quota_inode = gfs2_lookup_simple(sdp->sd_master_dir, "quota");
560 if (IS_ERR(sdp->sd_quota_inode)) {
561 error = PTR_ERR(sdp->sd_quota_inode);
562 fs_err(sdp, "can't get quota file inode: %d\n", error);
568 iput(sdp->sd_quota_inode);
571 gfs2_clear_rgrpd(sdp);
572 iput(sdp->sd_rindex);
575 iput(sdp->sd_statfs_inode);
578 iput(sdp->sd_inum_inode);
580 init_journal(sdp, UNDO);
582 iput(sdp->sd_master_dir);
587 static int init_per_node(struct gfs2_sbd *sdp, int undo)
589 struct inode *pn = NULL;
592 struct gfs2_inode *ip;
594 if (sdp->sd_args.ar_spectator)
600 pn = gfs2_lookup_simple(sdp->sd_master_dir, "per_node");
603 fs_err(sdp, "can't find per_node directory: %d\n", error);
607 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
608 sdp->sd_ir_inode = gfs2_lookup_simple(pn, buf);
609 if (IS_ERR(sdp->sd_ir_inode)) {
610 error = PTR_ERR(sdp->sd_ir_inode);
611 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
615 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
616 sdp->sd_sc_inode = gfs2_lookup_simple(pn, buf);
617 if (IS_ERR(sdp->sd_sc_inode)) {
618 error = PTR_ERR(sdp->sd_sc_inode);
619 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
623 sprintf(buf, "unlinked_tag%u", sdp->sd_jdesc->jd_jid);
624 sdp->sd_ut_inode = gfs2_lookup_simple(pn, buf);
625 if (IS_ERR(sdp->sd_ut_inode)) {
626 error = PTR_ERR(sdp->sd_ut_inode);
627 fs_err(sdp, "can't find local \"ut\" file: %d\n", error);
631 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
632 sdp->sd_qc_inode = gfs2_lookup_simple(pn, buf);
633 if (IS_ERR(sdp->sd_qc_inode)) {
634 error = PTR_ERR(sdp->sd_qc_inode);
635 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
642 ip = sdp->sd_ir_inode->u.generic_ip;
643 error = gfs2_glock_nq_init(ip->i_gl,
644 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
647 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
651 ip = sdp->sd_sc_inode->u.generic_ip;
652 error = gfs2_glock_nq_init(ip->i_gl,
653 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
656 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
660 ip = sdp->sd_ut_inode->u.generic_ip;
661 error = gfs2_glock_nq_init(ip->i_gl,
662 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
665 fs_err(sdp, "can't lock local \"ut\" file: %d\n", error);
669 ip = sdp->sd_qc_inode->u.generic_ip;
670 error = gfs2_glock_nq_init(ip->i_gl,
671 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
674 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
681 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
684 gfs2_glock_dq_uninit(&sdp->sd_ut_gh);
687 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
690 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
693 iput(sdp->sd_qc_inode);
696 iput(sdp->sd_ut_inode);
699 iput(sdp->sd_sc_inode);
702 iput(sdp->sd_ir_inode);
710 static int init_threads(struct gfs2_sbd *sdp, int undo)
712 struct task_struct *p;
718 sdp->sd_log_flush_time = jiffies;
719 sdp->sd_jindex_refresh_time = jiffies;
721 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
724 fs_err(sdp, "can't start logd thread: %d\n", error);
727 sdp->sd_logd_process = p;
729 sdp->sd_statfs_sync_time = jiffies;
730 sdp->sd_quota_sync_time = jiffies;
732 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
735 fs_err(sdp, "can't start quotad thread: %d\n", error);
738 sdp->sd_quotad_process = p;
740 p = kthread_run(gfs2_inoded, sdp, "gfs2_inoded");
743 fs_err(sdp, "can't start inoded thread: %d\n", error);
746 sdp->sd_inoded_process = p;
751 kthread_stop(sdp->sd_inoded_process);
754 kthread_stop(sdp->sd_quotad_process);
757 kthread_stop(sdp->sd_logd_process);
763 * fill_super - Read in superblock
764 * @sb: The VFS superblock
765 * @data: Mount options
766 * @silent: Don't complain if it's not a GFS2 filesystem
771 static int fill_super(struct super_block *sb, void *data, int silent)
773 struct gfs2_sbd *sdp;
774 struct gfs2_holder mount_gh;
779 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
783 error = gfs2_mount_args(sdp, (char *)data, 0);
785 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
789 init_vfs(sb, SDF_NOATIME);
791 /* Set up the buffer cache and fill in some fake block size values
792 to allow us to read-in the on-disk superblock. */
793 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
794 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
795 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
796 GFS2_BASIC_BLOCK_SHIFT;
797 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
799 error = init_names(sdp, silent);
803 error = gfs2_sys_fs_add(sdp);
807 error = gfs2_lm_mount(sdp, silent);
811 error = init_locking(sdp, &mount_gh, DO);
815 error = init_sb(sdp, silent, DO);
819 error = init_inodes(sdp, DO);
823 error = init_per_node(sdp, DO);
827 error = gfs2_statfs_init(sdp);
829 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
833 error = init_threads(sdp, DO);
837 if (!(sb->s_flags & MS_RDONLY)) {
838 error = gfs2_make_fs_rw(sdp);
840 fs_err(sdp, "can't make FS RW: %d\n", error);
845 gfs2_glock_dq_uninit(&mount_gh);
850 init_threads(sdp, UNDO);
853 init_per_node(sdp, UNDO);
856 init_inodes(sdp, UNDO);
859 init_sb(sdp, 0, UNDO);
862 init_locking(sdp, &mount_gh, UNDO);
865 gfs2_gl_hash_clear(sdp, WAIT);
866 gfs2_lm_unmount(sdp);
867 while (invalidate_inodes(sb))
871 gfs2_sys_fs_del(sdp);
875 sb->s_fs_info = NULL;
880 static struct super_block *gfs2_get_sb(struct file_system_type *fs_type,
881 int flags, const char *dev_name,
884 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super);
887 static void gfs2_kill_sb(struct super_block *sb)
889 kill_block_super(sb);
892 struct file_system_type gfs2_fs_type = {
894 .fs_flags = FS_REQUIRES_DEV,
895 .get_sb = gfs2_get_sb,
896 .kill_sb = gfs2_kill_sb,
897 .owner = THIS_MODULE,
900 struct file_system_type gfs2meta_fs_type = {
902 .fs_flags = FS_REQUIRES_DEV,
903 .get_sb = gfs2_get_sb,
904 .kill_sb = gfs2_kill_sb,
905 .owner = THIS_MODULE,