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 super_block *sb, unsigned noatime)
118 struct gfs2_sbd *sdp = sb->s_fs_info;
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(noatime, &sdp->sd_flags);
128 /* Don't let the VFS update atimes. GFS2 handles this itself. */
129 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
132 static int init_names(struct gfs2_sbd *sdp, int silent)
134 struct gfs2_sb *sb = NULL;
138 proto = sdp->sd_args.ar_lockproto;
139 table = sdp->sd_args.ar_locktable;
141 /* Try to autodetect */
143 if (!proto[0] || !table[0]) {
144 struct buffer_head *bh;
145 bh = sb_getblk(sdp->sd_vfs,
146 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
148 clear_buffer_uptodate(bh);
149 clear_buffer_dirty(bh);
151 ll_rw_block(READ, 1, &bh);
154 if (!buffer_uptodate(bh)) {
159 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
164 gfs2_sb_in(sb, bh->b_data);
167 error = gfs2_check_sb(sdp, sb, silent);
172 proto = sb->sb_lockproto;
174 table = sb->sb_locktable;
178 table = sdp->sd_vfs->s_id;
180 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
181 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
189 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
192 struct task_struct *p;
198 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
201 fs_err(sdp, "can't start scand thread: %d\n", error);
204 sdp->sd_scand_process = p;
206 for (sdp->sd_glockd_num = 0;
207 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
208 sdp->sd_glockd_num++) {
209 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
212 fs_err(sdp, "can't start glockd thread: %d\n", error);
215 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
218 error = gfs2_glock_nq_num(sdp,
219 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
220 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
223 fs_err(sdp, "can't acquire mount glock: %d\n", error);
227 error = gfs2_glock_nq_num(sdp,
228 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
230 LM_FLAG_NOEXP | GL_EXACT | GL_NEVER_RECURSE,
233 fs_err(sdp, "can't acquire live glock: %d\n", error);
237 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
238 CREATE, &sdp->sd_rename_gl);
240 fs_err(sdp, "can't create rename glock: %d\n", error);
244 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
245 CREATE, &sdp->sd_trans_gl);
247 fs_err(sdp, "can't create transaction glock: %d\n", error);
250 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
255 gfs2_glock_put(sdp->sd_trans_gl);
258 gfs2_glock_put(sdp->sd_rename_gl);
261 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
264 gfs2_glock_dq_uninit(mount_gh);
267 while (sdp->sd_glockd_num--)
268 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
270 kthread_stop(sdp->sd_scand_process);
275 static struct inode *gfs2_lookup_root(struct gfs2_sbd *sdp,
276 const struct gfs2_inum *inum)
279 struct gfs2_glock *gl;
280 struct gfs2_inode *ip;
283 error = gfs2_glock_get(sdp, inum->no_addr,
284 &gfs2_inode_glops, CREATE, &gl);
286 error = gfs2_inode_get(gl, inum, CREATE, &ip);
288 gfs2_inode_min_init(ip, DT_DIR);
289 inode = gfs2_ip2v(ip);
296 return ERR_PTR(error);
299 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
301 struct super_block *sb = sdp->sd_vfs;
302 struct gfs2_holder sb_gh;
303 struct gfs2_inum *inum;
311 error = gfs2_glock_nq_num(sdp,
312 GFS2_SB_LOCK, &gfs2_meta_glops,
313 LM_ST_SHARED, 0, &sb_gh);
315 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
319 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
321 fs_err(sdp, "can't read superblock: %d\n", error);
325 /* Set up the buffer cache and SB for real */
326 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
328 fs_err(sdp, "FS block size (%u) is too small for device "
330 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
333 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
335 fs_err(sdp, "FS block size (%u) is too big for machine "
337 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
341 /* Get rid of buffers from the original block size */
342 sb_gh.gh_gl->gl_ops->go_inval(sb_gh.gh_gl, DIO_METADATA | DIO_DATA);
343 sb_gh.gh_gl->gl_aspace->i_blkbits = sdp->sd_sb.sb_bsize_shift;
345 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
347 /* Get the root inode */
348 inum = &sdp->sd_sb.sb_root_dir;
349 if (sb->s_type == &gfs2meta_fs_type)
350 inum = &sdp->sd_sb.sb_master_dir;
351 inode = gfs2_lookup_root(sdp, inum);
353 error = PTR_ERR(inode);
354 fs_err(sdp, "can't read in root inode: %d\n", error);
358 sb->s_root = d_alloc_root(inode);
360 fs_err(sdp, "can't get root dentry\n");
365 gfs2_glock_dq_uninit(&sb_gh);
369 static int init_journal(struct gfs2_sbd *sdp, int undo)
371 struct gfs2_holder ji_gh;
372 struct task_struct *p;
373 struct gfs2_inode *ip;
382 error = gfs2_lookup_simple(sdp->sd_master_dir, "jindex",
385 fs_err(sdp, "can't lookup journal index: %d\n", error);
388 ip = sdp->sd_jindex->u.generic_ip;
389 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
391 /* Load in the journal index special file */
393 error = gfs2_jindex_hold(sdp, &ji_gh);
395 fs_err(sdp, "can't read journal index: %d\n", error);
400 if (!gfs2_jindex_size(sdp)) {
401 fs_err(sdp, "no journals!\n");
405 if (sdp->sd_args.ar_spectator) {
406 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
407 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
409 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
410 fs_err(sdp, "can't mount journal #%u\n",
411 sdp->sd_lockstruct.ls_jid);
412 fs_err(sdp, "there are only %u journals (0 - %u)\n",
413 gfs2_jindex_size(sdp),
414 gfs2_jindex_size(sdp) - 1);
417 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
419 error = gfs2_glock_nq_num(sdp,
420 sdp->sd_lockstruct.ls_jid,
422 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
423 &sdp->sd_journal_gh);
425 fs_err(sdp, "can't acquire journal glock: %d\n", error);
429 ip = sdp->sd_jdesc->jd_inode->u.generic_ip;
430 error = gfs2_glock_nq_init(ip->i_gl,
432 LM_FLAG_NOEXP | GL_EXACT,
435 fs_err(sdp, "can't acquire journal inode glock: %d\n",
437 goto fail_journal_gh;
440 error = gfs2_jdesc_check(sdp->sd_jdesc);
442 fs_err(sdp, "my journal (%u) is bad: %d\n",
443 sdp->sd_jdesc->jd_jid, error);
446 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
449 if (sdp->sd_lockstruct.ls_first) {
451 for (x = 0; x < sdp->sd_journals; x++) {
452 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
455 fs_err(sdp, "error recovering journal %u: %d\n",
461 gfs2_lm_others_may_mount(sdp);
462 } else if (!sdp->sd_args.ar_spectator) {
463 error = gfs2_recover_journal(sdp->sd_jdesc, WAIT);
465 fs_err(sdp, "error recovering my journal: %d\n", error);
470 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
471 gfs2_glock_dq_uninit(&ji_gh);
474 /* Disown my Journal glock */
476 sdp->sd_journal_gh.gh_owner = NULL;
477 sdp->sd_jinode_gh.gh_owner = NULL;
479 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
482 fs_err(sdp, "can't start recoverd thread: %d\n", error);
485 sdp->sd_recoverd_process = p;
490 kthread_stop(sdp->sd_recoverd_process);
493 if (!sdp->sd_args.ar_spectator)
494 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
497 if (!sdp->sd_args.ar_spectator)
498 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
501 gfs2_jindex_free(sdp);
503 gfs2_glock_dq_uninit(&ji_gh);
506 iput(sdp->sd_jindex);
512 static int init_inodes(struct gfs2_sbd *sdp, int undo)
515 struct gfs2_inode *ip;
521 inode = gfs2_lookup_root(sdp, &sdp->sd_sb.sb_master_dir);
523 error = PTR_ERR(inode);
524 fs_err(sdp, "can't read in master directory: %d\n", error);
527 sdp->sd_master_dir = inode;
529 error = init_journal(sdp, undo);
533 /* Read in the master inode number inode */
534 error = gfs2_lookup_simple(sdp->sd_master_dir, "inum",
535 &sdp->sd_inum_inode);
537 fs_err(sdp, "can't read in inum inode: %d\n", error);
542 /* Read in the master statfs inode */
543 error = gfs2_lookup_simple(sdp->sd_master_dir, "statfs",
544 &sdp->sd_statfs_inode);
546 fs_err(sdp, "can't read in statfs inode: %d\n", error);
550 /* Read in the resource index inode */
551 error = gfs2_lookup_simple(sdp->sd_master_dir, "rindex",
554 fs_err(sdp, "can't get resource index inode: %d\n", error);
557 ip = sdp->sd_rindex->u.generic_ip;
558 set_bit(GLF_STICKY, &ip->i_gl->gl_flags);
559 sdp->sd_rindex_vn = ip->i_gl->gl_vn - 1;
561 /* Read in the quota inode */
562 error = gfs2_lookup_simple(sdp->sd_master_dir, "quota",
563 &sdp->sd_quota_inode);
565 fs_err(sdp, "can't get quota file inode: %d\n", error);
571 iput(sdp->sd_quota_inode);
574 gfs2_clear_rgrpd(sdp);
575 iput(sdp->sd_rindex);
578 iput(sdp->sd_statfs_inode);
581 iput(sdp->sd_inum_inode);
583 init_journal(sdp, UNDO);
585 iput(sdp->sd_master_dir);
590 static int init_per_node(struct gfs2_sbd *sdp, int undo)
592 struct inode *pn = NULL;
595 struct gfs2_inode *ip;
597 if (sdp->sd_args.ar_spectator)
603 error = gfs2_lookup_simple(sdp->sd_master_dir, "per_node", &pn);
605 fs_err(sdp, "can't find per_node directory: %d\n", error);
609 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
610 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ir_inode);
612 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
616 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
617 error = gfs2_lookup_simple(pn, buf, &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 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ut_inode);
626 fs_err(sdp, "can't find local \"ut\" file: %d\n", error);
630 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
631 error = gfs2_lookup_simple(pn, buf, &sdp->sd_qc_inode);
633 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
640 ip = sdp->sd_ir_inode->u.generic_ip;
641 error = gfs2_glock_nq_init(ip->i_gl,
642 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
645 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
649 ip = sdp->sd_sc_inode->u.generic_ip;
650 error = gfs2_glock_nq_init(ip->i_gl,
651 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
654 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
658 ip = sdp->sd_ut_inode->u.generic_ip;
659 error = gfs2_glock_nq_init(ip->i_gl,
660 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
663 fs_err(sdp, "can't lock local \"ut\" file: %d\n", error);
667 ip = sdp->sd_qc_inode->u.generic_ip;
668 error = gfs2_glock_nq_init(ip->i_gl,
669 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
672 fs_err(sdp, "can't lock local \"qc\" file: %d\n", error);
679 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
682 gfs2_glock_dq_uninit(&sdp->sd_ut_gh);
685 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
688 gfs2_glock_dq_uninit(&sdp->sd_ir_gh);
691 iput(sdp->sd_qc_inode);
694 iput(sdp->sd_ut_inode);
697 iput(sdp->sd_sc_inode);
700 iput(sdp->sd_ir_inode);
708 static int init_threads(struct gfs2_sbd *sdp, int undo)
710 struct task_struct *p;
716 sdp->sd_log_flush_time = jiffies;
717 sdp->sd_jindex_refresh_time = jiffies;
719 p = kthread_run(gfs2_logd, sdp, "gfs2_logd");
722 fs_err(sdp, "can't start logd thread: %d\n", error);
725 sdp->sd_logd_process = p;
727 sdp->sd_statfs_sync_time = jiffies;
728 sdp->sd_quota_sync_time = jiffies;
730 p = kthread_run(gfs2_quotad, sdp, "gfs2_quotad");
733 fs_err(sdp, "can't start quotad thread: %d\n", error);
736 sdp->sd_quotad_process = p;
738 p = kthread_run(gfs2_inoded, sdp, "gfs2_inoded");
741 fs_err(sdp, "can't start inoded thread: %d\n", error);
744 sdp->sd_inoded_process = p;
749 kthread_stop(sdp->sd_inoded_process);
752 kthread_stop(sdp->sd_quotad_process);
755 kthread_stop(sdp->sd_logd_process);
761 * fill_super - Read in superblock
762 * @sb: The VFS superblock
763 * @data: Mount options
764 * @silent: Don't complain if it's not a GFS2 filesystem
769 static int fill_super(struct super_block *sb, void *data, int silent)
771 struct gfs2_sbd *sdp;
772 struct gfs2_holder mount_gh;
777 printk(KERN_WARNING "GFS2: can't alloc struct gfs2_sbd\n");
781 error = gfs2_mount_args(sdp, (char *)data, 0);
783 printk(KERN_WARNING "GFS2: can't parse mount arguments\n");
787 init_vfs(sb, SDF_NOATIME);
789 /* Set up the buffer cache and fill in some fake block size values
790 to allow us to read-in the on-disk superblock. */
791 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
792 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
793 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
794 GFS2_BASIC_BLOCK_SHIFT;
795 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
797 error = init_names(sdp, silent);
801 error = gfs2_sys_fs_add(sdp);
805 error = gfs2_lm_mount(sdp, silent);
809 error = init_locking(sdp, &mount_gh, DO);
813 error = init_sb(sdp, silent, DO);
817 error = init_inodes(sdp, DO);
821 error = init_per_node(sdp, DO);
825 error = gfs2_statfs_init(sdp);
827 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
831 error = init_threads(sdp, DO);
835 if (!(sb->s_flags & MS_RDONLY)) {
836 error = gfs2_make_fs_rw(sdp);
838 fs_err(sdp, "can't make FS RW: %d\n", error);
843 gfs2_glock_dq_uninit(&mount_gh);
848 init_threads(sdp, UNDO);
851 init_per_node(sdp, UNDO);
854 init_inodes(sdp, UNDO);
857 init_sb(sdp, 0, UNDO);
860 init_locking(sdp, &mount_gh, UNDO);
863 gfs2_gl_hash_clear(sdp, WAIT);
864 gfs2_lm_unmount(sdp);
865 while (invalidate_inodes(sb))
869 gfs2_sys_fs_del(sdp);
873 sb->s_fs_info = NULL;
878 static struct super_block *gfs2_get_sb(struct file_system_type *fs_type,
879 int flags, const char *dev_name,
882 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super);
885 static void gfs2_kill_sb(struct super_block *sb)
887 kill_block_super(sb);
890 struct file_system_type gfs2_fs_type = {
892 .fs_flags = FS_REQUIRES_DEV,
893 .get_sb = gfs2_get_sb,
894 .kill_sb = gfs2_kill_sb,
895 .owner = THIS_MODULE,
898 struct file_system_type gfs2meta_fs_type = {
900 .fs_flags = FS_REQUIRES_DEV,
901 .get_sb = gfs2_get_sb,
902 .kill_sb = gfs2_kill_sb,
903 .owner = THIS_MODULE,