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 INIT_LIST_HEAD(&sdp->sd_log_blks_list);
101 init_waitqueue_head(&sdp->sd_log_blks_wait);
103 INIT_LIST_HEAD(&sdp->sd_ail1_list);
104 INIT_LIST_HEAD(&sdp->sd_ail2_list);
106 mutex_init(&sdp->sd_log_flush_lock);
107 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
109 INIT_LIST_HEAD(&sdp->sd_revoke_list);
111 mutex_init(&sdp->sd_freeze_lock);
116 static void init_vfs(struct gfs2_sbd *sdp)
118 struct super_block *sb = sdp->sd_vfs;
120 sb->s_magic = GFS2_MAGIC;
121 sb->s_op = &gfs2_super_ops;
122 sb->s_export_op = &gfs2_export_ops;
123 sb->s_maxbytes = MAX_LFS_FILESIZE;
125 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
126 set_bit(SDF_NOATIME, &sdp->sd_flags);
128 /* Don't let the VFS update atimes. GFS2 handles this itself. */
129 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
131 /* Set up the buffer cache and fill in some fake block size values
132 to allow us to read-in the on-disk superblock. */
133 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
134 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
135 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
136 GFS2_BASIC_BLOCK_SHIFT;
137 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
140 static int init_names(struct gfs2_sbd *sdp, int silent)
142 struct gfs2_sb *sb = NULL;
146 proto = sdp->sd_args.ar_lockproto;
147 table = sdp->sd_args.ar_locktable;
149 /* Try to autodetect */
151 if (!proto[0] || !table[0]) {
152 struct buffer_head *bh;
153 bh = sb_getblk(sdp->sd_vfs,
154 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
156 clear_buffer_uptodate(bh);
157 clear_buffer_dirty(bh);
159 ll_rw_block(READ, 1, &bh);
162 if (!buffer_uptodate(bh)) {
167 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
172 gfs2_sb_in(sb, bh->b_data);
175 error = gfs2_check_sb(sdp, sb, silent);
180 proto = sb->sb_lockproto;
182 table = sb->sb_locktable;
186 table = sdp->sd_vfs->s_id;
188 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
189 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
197 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
200 struct task_struct *p;
206 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
209 fs_err(sdp, "can't start scand thread: %d\n", error);
212 sdp->sd_scand_process = p;
214 for (sdp->sd_glockd_num = 0;
215 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
216 sdp->sd_glockd_num++) {
217 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
220 fs_err(sdp, "can't start glockd thread: %d\n", error);
223 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
226 error = gfs2_glock_nq_num(sdp,
227 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
228 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
231 fs_err(sdp, "can't acquire mount glock: %d\n", error);
235 error = gfs2_glock_nq_num(sdp,
236 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
238 LM_FLAG_NOEXP | GL_EXACT | GL_NEVER_RECURSE,
241 fs_err(sdp, "can't acquire live glock: %d\n", error);
245 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
246 CREATE, &sdp->sd_rename_gl);
248 fs_err(sdp, "can't create rename glock: %d\n", error);
252 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
253 CREATE, &sdp->sd_trans_gl);
255 fs_err(sdp, "can't create transaction glock: %d\n", error);
258 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
263 gfs2_glock_put(sdp->sd_trans_gl);
266 gfs2_glock_put(sdp->sd_rename_gl);
269 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
272 gfs2_glock_dq_uninit(mount_gh);
275 while (sdp->sd_glockd_num--)
276 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
278 kthread_stop(sdp->sd_scand_process);
283 int gfs2_lookup_root(struct gfs2_sbd *sdp)
286 struct gfs2_glock *gl;
287 struct gfs2_inode *ip;
289 error = gfs2_glock_get(sdp, sdp->sd_sb.sb_root_dir.no_addr,
290 &gfs2_inode_glops, CREATE, &gl);
292 error = gfs2_inode_get(gl, &sdp->sd_sb.sb_root_dir,
296 gfs2_inode_min_init(ip, DT_DIR);
297 sdp->sd_root_dir = gfs2_ip2v(ip);
306 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
308 struct super_block *sb = sdp->sd_vfs;
309 struct gfs2_holder sb_gh;
314 iput(sdp->sd_master_dir);
318 error = gfs2_glock_nq_num(sdp,
319 GFS2_SB_LOCK, &gfs2_meta_glops,
320 LM_ST_SHARED, 0, &sb_gh);
322 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
326 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
328 fs_err(sdp, "can't read superblock: %d\n", error);
332 /* Set up the buffer cache and SB for real */
334 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
335 fs_err(sdp, "FS block size (%u) is too small for device "
337 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
340 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
341 fs_err(sdp, "FS block size (%u) is too big for machine "
343 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
347 /* Get rid of buffers from the original block size */
348 sb_gh.gh_gl->gl_ops->go_inval(sb_gh.gh_gl, DIO_METADATA | DIO_DATA);
349 sb_gh.gh_gl->gl_aspace->i_blkbits = sdp->sd_sb.sb_bsize_shift;
351 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
353 /* Get the root inode */
354 error = gfs2_lookup_root(sdp);
356 fs_err(sdp, "can't read in root inode: %d\n", error);
360 /* Get the root inode/dentry */
361 inode = sdp->sd_root_dir;
363 fs_err(sdp, "can't get root inode\n");
369 sb->s_root = d_alloc_root(inode);
371 fs_err(sdp, "can't get root dentry\n");
377 gfs2_glock_dq_uninit(&sb_gh);
381 iput(sdp->sd_root_dir);
385 static int init_journal(struct gfs2_sbd *sdp, int undo)
387 struct gfs2_holder ji_gh;
388 struct task_struct *p;
389 struct gfs2_inode *ip;
398 error = gfs2_lookup_simple(sdp->sd_master_dir, "jindex",
401 fs_err(sdp, "can't lookup journal index: %d\n", error);
404 ip = sdp->sd_jindex->u.generic_ip;
405 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
407 /* Load in the journal index special file */
409 error = gfs2_jindex_hold(sdp, &ji_gh);
411 fs_err(sdp, "can't read journal index: %d\n", error);
416 if (!gfs2_jindex_size(sdp)) {
417 fs_err(sdp, "no journals!\n");
421 if (sdp->sd_args.ar_spectator) {
422 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
423 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
425 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
426 fs_err(sdp, "can't mount journal #%u\n",
427 sdp->sd_lockstruct.ls_jid);
428 fs_err(sdp, "there are only %u journals (0 - %u)\n",
429 gfs2_jindex_size(sdp),
430 gfs2_jindex_size(sdp) - 1);
433 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
435 error = gfs2_glock_nq_num(sdp,
436 sdp->sd_lockstruct.ls_jid,
438 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
439 &sdp->sd_journal_gh);
441 fs_err(sdp, "can't acquire journal glock: %d\n", error);
445 ip = sdp->sd_jdesc->jd_inode->u.generic_ip;
446 error = gfs2_glock_nq_init(ip->i_gl,
448 LM_FLAG_NOEXP | GL_EXACT,
451 fs_err(sdp, "can't acquire journal inode glock: %d\n",
453 goto fail_journal_gh;
456 error = gfs2_jdesc_check(sdp->sd_jdesc);
458 fs_err(sdp, "my journal (%u) is bad: %d\n",
459 sdp->sd_jdesc->jd_jid, error);
462 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
465 if (sdp->sd_lockstruct.ls_first) {
467 for (x = 0; x < sdp->sd_journals; x++) {
468 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
471 fs_err(sdp, "error recovering journal %u: %d\n",
477 gfs2_lm_others_may_mount(sdp);
478 } else if (!sdp->sd_args.ar_spectator) {
479 error = gfs2_recover_journal(sdp->sd_jdesc, WAIT);
481 fs_err(sdp, "error recovering my journal: %d\n", error);
486 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
487 gfs2_glock_dq_uninit(&ji_gh);
490 /* Disown my Journal glock */
492 sdp->sd_journal_gh.gh_owner = NULL;
493 sdp->sd_jinode_gh.gh_owner = NULL;
495 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
498 fs_err(sdp, "can't start recoverd thread: %d\n", error);
501 sdp->sd_recoverd_process = p;
506 kthread_stop(sdp->sd_recoverd_process);
509 if (!sdp->sd_args.ar_spectator)
510 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
513 if (!sdp->sd_args.ar_spectator)
514 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
517 gfs2_jindex_free(sdp);
519 gfs2_glock_dq_uninit(&ji_gh);
522 iput(sdp->sd_jindex);
528 static int init_inodes(struct gfs2_sbd *sdp, int undo)
531 struct gfs2_inode *ip;
536 error = gfs2_lookup_master_dir(sdp);
538 fs_err(sdp, "can't read in master directory: %d\n", error);
542 error = init_journal(sdp, undo);
546 /* Read in the master inode number inode */
547 error = gfs2_lookup_simple(sdp->sd_master_dir, "inum",
548 &sdp->sd_inum_inode);
550 fs_err(sdp, "can't read in inum inode: %d\n", error);
555 /* Read in the master statfs inode */
556 error = gfs2_lookup_simple(sdp->sd_master_dir, "statfs",
557 &sdp->sd_statfs_inode);
559 fs_err(sdp, "can't read in statfs inode: %d\n", error);
563 /* Read in the resource index inode */
564 error = gfs2_lookup_simple(sdp->sd_master_dir, "rindex",
567 fs_err(sdp, "can't get resource index inode: %d\n", error);
570 ip = sdp->sd_rindex->u.generic_ip;
571 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
572 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
574 /* Read in the quota inode */
575 error = gfs2_lookup_simple(sdp->sd_master_dir, "quota",
576 &sdp->sd_quota_inode);
578 fs_err(sdp, "can't get quota file inode: %d\n", error);
584 iput(sdp->sd_quota_inode);
587 gfs2_clear_rgrpd(sdp);
588 iput(sdp->sd_rindex);
591 iput(sdp->sd_statfs_inode);
594 iput(sdp->sd_inum_inode);
596 init_journal(sdp, UNDO);
598 iput(sdp->sd_master_dir);
603 static int init_per_node(struct gfs2_sbd *sdp, int undo)
605 struct inode *pn = NULL;
608 struct gfs2_inode *ip;
610 if (sdp->sd_args.ar_spectator)
616 error = gfs2_lookup_simple(sdp->sd_master_dir, "per_node", &pn);
618 fs_err(sdp, "can't find per_node directory: %d\n", error);
622 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
623 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ir_inode);
625 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
629 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
630 error = gfs2_lookup_simple(pn, buf, &sdp->sd_sc_inode);
632 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
636 sprintf(buf, "unlinked_tag%u", sdp->sd_jdesc->jd_jid);
637 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ut_inode);
639 fs_err(sdp, "can't find local \"ut\" file: %d\n", error);
643 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
644 error = gfs2_lookup_simple(pn, buf, &sdp->sd_qc_inode);
646 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
653 ip = sdp->sd_ir_inode->u.generic_ip;
654 error = gfs2_glock_nq_init(ip->i_gl,
655 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
658 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
662 ip = sdp->sd_sc_inode->u.generic_ip;
663 error = gfs2_glock_nq_init(ip->i_gl,
664 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
667 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
671 ip = sdp->sd_ut_inode->u.generic_ip;
672 error = gfs2_glock_nq_init(ip->i_gl,
673 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
676 fs_err(sdp, "can't lock local \"ut\" file: %d\n", error);
680 ip = sdp->sd_qc_inode->u.generic_ip;
681 error = gfs2_glock_nq_init(ip->i_gl,
682 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
685 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
692 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
695 gfs2_glock_dq_uninit(&sdp->sd_ut_gh);
698 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
701 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
704 iput(sdp->sd_qc_inode);
707 iput(sdp->sd_ut_inode);
710 iput(sdp->sd_sc_inode);
713 iput(sdp->sd_ir_inode);
721 static int init_threads(struct gfs2_sbd *sdp, int undo)
723 struct task_struct *p;
729 sdp->sd_log_flush_time = jiffies;
730 sdp->sd_jindex_refresh_time = jiffies;
732 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
735 fs_err(sdp, "can't start logd thread: %d\n", error);
738 sdp->sd_logd_process = p;
740 sdp->sd_statfs_sync_time = jiffies;
741 sdp->sd_quota_sync_time = jiffies;
743 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
746 fs_err(sdp, "can't start quotad thread: %d\n", error);
749 sdp->sd_quotad_process = p;
751 p = kthread_run(gfs2_inoded, sdp, "gfs2_inoded");
754 fs_err(sdp, "can't start inoded thread: %d\n", error);
757 sdp->sd_inoded_process = p;
762 kthread_stop(sdp->sd_inoded_process);
765 kthread_stop(sdp->sd_quotad_process);
768 kthread_stop(sdp->sd_logd_process);
774 * fill_super - Read in superblock
775 * @sb: The VFS superblock
776 * @data: Mount options
777 * @silent: Don't complain if it's not a GFS2 filesystem
782 static int fill_super(struct super_block *sb, void *data, int silent)
784 struct gfs2_sbd *sdp;
785 struct gfs2_holder mount_gh;
790 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
794 error = gfs2_mount_args(sdp, (char *)data, 0);
796 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
802 error = init_names(sdp, silent);
806 error = gfs2_sys_fs_add(sdp);
810 error = gfs2_lm_mount(sdp, silent);
814 error = init_locking(sdp, &mount_gh, DO);
818 error = init_sb(sdp, silent, DO);
822 error = init_inodes(sdp, DO);
826 error = init_per_node(sdp, DO);
830 error = gfs2_statfs_init(sdp);
832 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
836 error = init_threads(sdp, DO);
840 if (!(sb->s_flags & MS_RDONLY)) {
841 error = gfs2_make_fs_rw(sdp);
843 fs_err(sdp, "can't make FS RW: %d\n", error);
848 gfs2_glock_dq_uninit(&mount_gh);
853 init_threads(sdp, UNDO);
856 init_per_node(sdp, UNDO);
859 init_inodes(sdp, UNDO);
862 init_sb(sdp, 0, UNDO);
865 init_locking(sdp, &mount_gh, UNDO);
868 gfs2_gl_hash_clear(sdp, WAIT);
869 gfs2_lm_unmount(sdp);
870 while (invalidate_inodes(sb))
874 gfs2_sys_fs_del(sdp);
878 sb->s_fs_info = NULL;
883 static struct super_block *gfs2_get_sb(struct file_system_type *fs_type,
884 int flags, const char *dev_name,
887 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super);
890 struct file_system_type gfs2_fs_type = {
892 .fs_flags = FS_REQUIRES_DEV,
893 .get_sb = gfs2_get_sb,
894 .kill_sb = kill_block_super,
895 .owner = THIS_MODULE,