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 <asm/semaphore.h>
27 #include "ops_export.h"
28 #include "ops_fstype.h"
29 #include "ops_super.h"
39 static struct gfs2_sbd *init_sbd(struct super_block *sb)
44 sdp = vmalloc(sizeof(struct gfs2_sbd));
48 memset(sdp, 0, sizeof(struct gfs2_sbd));
53 gfs2_tune_init(&sdp->sd_tune);
55 for (x = 0; x < GFS2_GL_HASH_SIZE; x++) {
56 sdp->sd_gl_hash[x].hb_lock = RW_LOCK_UNLOCKED;
57 INIT_LIST_HEAD(&sdp->sd_gl_hash[x].hb_list);
59 INIT_LIST_HEAD(&sdp->sd_reclaim_list);
60 spin_lock_init(&sdp->sd_reclaim_lock);
61 init_waitqueue_head(&sdp->sd_reclaim_wq);
62 mutex_init(&sdp->sd_invalidate_inodes_mutex);
64 mutex_init(&sdp->sd_inum_mutex);
65 spin_lock_init(&sdp->sd_statfs_spin);
66 mutex_init(&sdp->sd_statfs_mutex);
68 spin_lock_init(&sdp->sd_rindex_spin);
69 mutex_init(&sdp->sd_rindex_mutex);
70 INIT_LIST_HEAD(&sdp->sd_rindex_list);
71 INIT_LIST_HEAD(&sdp->sd_rindex_mru_list);
72 INIT_LIST_HEAD(&sdp->sd_rindex_recent_list);
74 INIT_LIST_HEAD(&sdp->sd_jindex_list);
75 spin_lock_init(&sdp->sd_jindex_spin);
76 mutex_init(&sdp->sd_jindex_mutex);
78 INIT_LIST_HEAD(&sdp->sd_unlinked_list);
79 spin_lock_init(&sdp->sd_unlinked_spin);
80 mutex_init(&sdp->sd_unlinked_mutex);
82 INIT_LIST_HEAD(&sdp->sd_quota_list);
83 spin_lock_init(&sdp->sd_quota_spin);
84 mutex_init(&sdp->sd_quota_mutex);
86 spin_lock_init(&sdp->sd_log_lock);
87 init_waitqueue_head(&sdp->sd_log_trans_wq);
88 init_waitqueue_head(&sdp->sd_log_flush_wq);
90 INIT_LIST_HEAD(&sdp->sd_log_le_gl);
91 INIT_LIST_HEAD(&sdp->sd_log_le_buf);
92 INIT_LIST_HEAD(&sdp->sd_log_le_revoke);
93 INIT_LIST_HEAD(&sdp->sd_log_le_rg);
94 INIT_LIST_HEAD(&sdp->sd_log_le_databuf);
96 INIT_LIST_HEAD(&sdp->sd_log_blks_list);
97 init_waitqueue_head(&sdp->sd_log_blks_wait);
99 INIT_LIST_HEAD(&sdp->sd_ail1_list);
100 INIT_LIST_HEAD(&sdp->sd_ail2_list);
102 mutex_init(&sdp->sd_log_flush_lock);
103 INIT_LIST_HEAD(&sdp->sd_log_flush_list);
105 INIT_LIST_HEAD(&sdp->sd_revoke_list);
107 mutex_init(&sdp->sd_freeze_lock);
112 static void init_vfs(struct gfs2_sbd *sdp)
114 struct super_block *sb = sdp->sd_vfs;
116 sb->s_magic = GFS2_MAGIC;
117 sb->s_op = &gfs2_super_ops;
118 sb->s_export_op = &gfs2_export_ops;
119 sb->s_maxbytes = MAX_LFS_FILESIZE;
121 if (sb->s_flags & (MS_NOATIME | MS_NODIRATIME))
122 set_bit(SDF_NOATIME, &sdp->sd_flags);
124 /* Don't let the VFS update atimes. GFS2 handles this itself. */
125 sb->s_flags |= MS_NOATIME | MS_NODIRATIME;
127 /* Set up the buffer cache and fill in some fake block size values
128 to allow us to read-in the on-disk superblock. */
129 sdp->sd_sb.sb_bsize = sb_min_blocksize(sb, GFS2_BASIC_BLOCK);
130 sdp->sd_sb.sb_bsize_shift = sb->s_blocksize_bits;
131 sdp->sd_fsb2bb_shift = sdp->sd_sb.sb_bsize_shift -
132 GFS2_BASIC_BLOCK_SHIFT;
133 sdp->sd_fsb2bb = 1 << sdp->sd_fsb2bb_shift;
136 static int init_names(struct gfs2_sbd *sdp, int silent)
138 struct gfs2_sb *sb = NULL;
142 proto = sdp->sd_args.ar_lockproto;
143 table = sdp->sd_args.ar_locktable;
145 /* Try to autodetect */
147 if (!proto[0] || !table[0]) {
148 struct buffer_head *bh;
149 bh = sb_getblk(sdp->sd_vfs,
150 GFS2_SB_ADDR >> sdp->sd_fsb2bb_shift);
152 clear_buffer_uptodate(bh);
153 clear_buffer_dirty(bh);
155 ll_rw_block(READ, 1, &bh);
158 if (!buffer_uptodate(bh)) {
163 sb = kmalloc(sizeof(struct gfs2_sb), GFP_KERNEL);
168 gfs2_sb_in(sb, bh->b_data);
171 error = gfs2_check_sb(sdp, sb, silent);
176 proto = sb->sb_lockproto;
178 table = sb->sb_locktable;
182 table = sdp->sd_vfs->s_id;
184 snprintf(sdp->sd_proto_name, GFS2_FSNAME_LEN, "%s", proto);
185 snprintf(sdp->sd_table_name, GFS2_FSNAME_LEN, "%s", table);
193 static int init_locking(struct gfs2_sbd *sdp, struct gfs2_holder *mount_gh,
196 struct task_struct *p;
202 p = kthread_run(gfs2_scand, sdp, "gfs2_scand");
205 fs_err(sdp, "can't start scand thread: %d\n", error);
208 sdp->sd_scand_process = p;
210 for (sdp->sd_glockd_num = 0;
211 sdp->sd_glockd_num < sdp->sd_args.ar_num_glockd;
212 sdp->sd_glockd_num++) {
213 p = kthread_run(gfs2_glockd, sdp, "gfs2_glockd");
216 fs_err(sdp, "can't start glockd thread: %d\n", error);
219 sdp->sd_glockd_process[sdp->sd_glockd_num] = p;
222 error = gfs2_glock_nq_num(sdp,
223 GFS2_MOUNT_LOCK, &gfs2_nondisk_glops,
224 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP | GL_NOCACHE,
227 fs_err(sdp, "can't acquire mount glock: %d\n", error);
231 error = gfs2_glock_nq_num(sdp,
232 GFS2_LIVE_LOCK, &gfs2_nondisk_glops,
234 LM_FLAG_NOEXP | GL_EXACT | GL_NEVER_RECURSE,
237 fs_err(sdp, "can't acquire live glock: %d\n", error);
241 error = gfs2_glock_get(sdp, GFS2_RENAME_LOCK, &gfs2_nondisk_glops,
242 CREATE, &sdp->sd_rename_gl);
244 fs_err(sdp, "can't create rename glock: %d\n", error);
248 error = gfs2_glock_get(sdp, GFS2_TRANS_LOCK, &gfs2_trans_glops,
249 CREATE, &sdp->sd_trans_gl);
251 fs_err(sdp, "can't create transaction glock: %d\n", error);
254 set_bit(GLF_STICKY, &sdp->sd_trans_gl->gl_flags);
259 gfs2_glock_put(sdp->sd_trans_gl);
262 gfs2_glock_put(sdp->sd_rename_gl);
265 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
268 gfs2_glock_dq_uninit(mount_gh);
271 while (sdp->sd_glockd_num--)
272 kthread_stop(sdp->sd_glockd_process[sdp->sd_glockd_num]);
274 kthread_stop(sdp->sd_scand_process);
279 int gfs2_lookup_root(struct gfs2_sbd *sdp)
282 struct gfs2_glock *gl;
283 struct gfs2_inode *ip;
285 error = gfs2_glock_get(sdp, sdp->sd_sb.sb_root_dir.no_addr,
286 &gfs2_inode_glops, CREATE, &gl);
288 error = gfs2_inode_get(gl, &sdp->sd_sb.sb_root_dir,
292 gfs2_inode_min_init(ip, DT_DIR);
293 sdp->sd_root_dir = gfs2_ip2v(ip);
302 static int init_sb(struct gfs2_sbd *sdp, int silent, int undo)
304 struct super_block *sb = sdp->sd_vfs;
305 struct gfs2_holder sb_gh;
310 iput(sdp->sd_master_dir);
314 error = gfs2_glock_nq_num(sdp,
315 GFS2_SB_LOCK, &gfs2_meta_glops,
316 LM_ST_SHARED, 0, &sb_gh);
318 fs_err(sdp, "can't acquire superblock glock: %d\n", error);
322 error = gfs2_read_sb(sdp, sb_gh.gh_gl, silent);
324 fs_err(sdp, "can't read superblock: %d\n", error);
328 /* Set up the buffer cache and SB for real */
330 if (sdp->sd_sb.sb_bsize < bdev_hardsect_size(sb->s_bdev)) {
331 fs_err(sdp, "FS block size (%u) is too small for device "
333 sdp->sd_sb.sb_bsize, bdev_hardsect_size(sb->s_bdev));
336 if (sdp->sd_sb.sb_bsize > PAGE_SIZE) {
337 fs_err(sdp, "FS block size (%u) is too big for machine "
339 sdp->sd_sb.sb_bsize, (unsigned int)PAGE_SIZE);
343 /* Get rid of buffers from the original block size */
344 sb_gh.gh_gl->gl_ops->go_inval(sb_gh.gh_gl, DIO_METADATA | DIO_DATA);
345 sb_gh.gh_gl->gl_aspace->i_blkbits = sdp->sd_sb.sb_bsize_shift;
347 sb_set_blocksize(sb, sdp->sd_sb.sb_bsize);
349 /* Get the root inode */
350 error = gfs2_lookup_root(sdp);
352 fs_err(sdp, "can't read in root inode: %d\n", error);
356 /* Get the root inode/dentry */
357 inode = sdp->sd_root_dir;
359 fs_err(sdp, "can't get root inode\n");
365 sb->s_root = d_alloc_root(inode);
367 fs_err(sdp, "can't get root dentry\n");
373 gfs2_glock_dq_uninit(&sb_gh);
377 iput(sdp->sd_root_dir);
381 static int init_journal(struct gfs2_sbd *sdp, int undo)
383 struct gfs2_holder ji_gh;
384 struct task_struct *p;
393 error = gfs2_lookup_simple(sdp->sd_master_dir, "jindex",
396 fs_err(sdp, "can't lookup journal index: %d\n", error);
399 set_bit(GLF_STICKY, &get_v2ip(sdp->sd_jindex)->i_gl->gl_flags);
401 /* Load in the journal index special file */
403 error = gfs2_jindex_hold(sdp, &ji_gh);
405 fs_err(sdp, "can't read journal index: %d\n", error);
410 if (!gfs2_jindex_size(sdp)) {
411 fs_err(sdp, "no journals!\n");
415 if (sdp->sd_args.ar_spectator) {
416 sdp->sd_jdesc = gfs2_jdesc_find(sdp, 0);
417 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
419 if (sdp->sd_lockstruct.ls_jid >= gfs2_jindex_size(sdp)) {
420 fs_err(sdp, "can't mount journal #%u\n",
421 sdp->sd_lockstruct.ls_jid);
422 fs_err(sdp, "there are only %u journals (0 - %u)\n",
423 gfs2_jindex_size(sdp),
424 gfs2_jindex_size(sdp) - 1);
427 sdp->sd_jdesc = gfs2_jdesc_find(sdp, sdp->sd_lockstruct.ls_jid);
429 error = gfs2_glock_nq_num(sdp,
430 sdp->sd_lockstruct.ls_jid,
432 LM_ST_EXCLUSIVE, LM_FLAG_NOEXP,
433 &sdp->sd_journal_gh);
435 fs_err(sdp, "can't acquire journal glock: %d\n", error);
439 error = gfs2_glock_nq_init(
440 get_v2ip(sdp->sd_jdesc->jd_inode)->i_gl,
442 LM_FLAG_NOEXP | GL_EXACT,
445 fs_err(sdp, "can't acquire journal inode glock: %d\n",
447 goto fail_journal_gh;
450 error = gfs2_jdesc_check(sdp->sd_jdesc);
452 fs_err(sdp, "my journal (%u) is bad: %d\n",
453 sdp->sd_jdesc->jd_jid, error);
456 sdp->sd_log_blks_free = sdp->sd_jdesc->jd_blocks;
459 if (sdp->sd_lockstruct.ls_first) {
461 for (x = 0; x < sdp->sd_journals; x++) {
462 error = gfs2_recover_journal(gfs2_jdesc_find(sdp, x),
465 fs_err(sdp, "error recovering journal %u: %d\n",
471 gfs2_lm_others_may_mount(sdp);
472 } else if (!sdp->sd_args.ar_spectator) {
473 error = gfs2_recover_journal(sdp->sd_jdesc, WAIT);
475 fs_err(sdp, "error recovering my journal: %d\n", error);
480 set_bit(SDF_JOURNAL_CHECKED, &sdp->sd_flags);
481 gfs2_glock_dq_uninit(&ji_gh);
484 /* Disown my Journal glock */
486 sdp->sd_journal_gh.gh_owner = NULL;
487 sdp->sd_jinode_gh.gh_owner = NULL;
489 p = kthread_run(gfs2_recoverd, sdp, "gfs2_recoverd");
492 fs_err(sdp, "can't start recoverd thread: %d\n", error);
495 sdp->sd_recoverd_process = p;
500 kthread_stop(sdp->sd_recoverd_process);
503 if (!sdp->sd_args.ar_spectator)
504 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
507 if (!sdp->sd_args.ar_spectator)
508 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
511 gfs2_jindex_free(sdp);
513 gfs2_glock_dq_uninit(&ji_gh);
516 iput(sdp->sd_jindex);
522 static int init_inodes(struct gfs2_sbd *sdp, int undo)
529 error = gfs2_lookup_master_dir(sdp);
531 fs_err(sdp, "can't read in master directory: %d\n", error);
535 error = init_journal(sdp, undo);
539 /* Read in the master inode number inode */
540 error = gfs2_lookup_simple(sdp->sd_master_dir, "inum",
541 &sdp->sd_inum_inode);
543 fs_err(sdp, "can't read in inum inode: %d\n", error);
548 /* Read in the master statfs inode */
549 error = gfs2_lookup_simple(sdp->sd_master_dir, "statfs",
550 &sdp->sd_statfs_inode);
552 fs_err(sdp, "can't read in statfs inode: %d\n", error);
556 /* Read in the resource index inode */
557 error = gfs2_lookup_simple(sdp->sd_master_dir, "rindex",
560 fs_err(sdp, "can't get resource index inode: %d\n", error);
563 set_bit(GLF_STICKY, &get_v2ip(sdp->sd_rindex)->i_gl->gl_flags);
564 sdp->sd_rindex_vn = get_v2ip(sdp->sd_rindex)->i_gl->gl_vn - 1;
566 /* Read in the quota inode */
567 error = gfs2_lookup_simple(sdp->sd_master_dir, "quota",
568 &sdp->sd_quota_inode);
570 fs_err(sdp, "can't get quota file inode: %d\n", error);
576 iput(sdp->sd_quota_inode);
579 gfs2_clear_rgrpd(sdp);
580 iput(sdp->sd_rindex);
583 iput(sdp->sd_statfs_inode);
586 iput(sdp->sd_inum_inode);
588 init_journal(sdp, UNDO);
590 iput(sdp->sd_master_dir);
595 static int init_per_node(struct gfs2_sbd *sdp, int undo)
597 struct inode *pn = NULL;
601 if (sdp->sd_args.ar_spectator)
607 error = gfs2_lookup_simple(sdp->sd_master_dir, "per_node", &pn);
609 fs_err(sdp, "can't find per_node directory: %d\n", error);
613 sprintf(buf, "inum_range%u", sdp->sd_jdesc->jd_jid);
614 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ir_inode);
616 fs_err(sdp, "can't find local \"ir\" file: %d\n", error);
620 sprintf(buf, "statfs_change%u", sdp->sd_jdesc->jd_jid);
621 error = gfs2_lookup_simple(pn, buf, &sdp->sd_sc_inode);
623 fs_err(sdp, "can't find local \"sc\" file: %d\n", error);
627 sprintf(buf, "unlinked_tag%u", sdp->sd_jdesc->jd_jid);
628 error = gfs2_lookup_simple(pn, buf, &sdp->sd_ut_inode);
630 fs_err(sdp, "can't find local \"ut\" file: %d\n", error);
634 sprintf(buf, "quota_change%u", sdp->sd_jdesc->jd_jid);
635 error = gfs2_lookup_simple(pn, buf, &sdp->sd_qc_inode);
637 fs_err(sdp, "can't find local \"qc\" file: %d\n", error);
644 error = gfs2_glock_nq_init(get_v2ip(sdp->sd_ir_inode)->i_gl,
645 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
648 fs_err(sdp, "can't lock local \"ir\" file: %d\n", error);
652 error = gfs2_glock_nq_init(get_v2ip(sdp->sd_sc_inode)->i_gl,
653 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
656 fs_err(sdp, "can't lock local \"sc\" file: %d\n", error);
660 error = gfs2_glock_nq_init(get_v2ip(sdp->sd_ut_inode)->i_gl,
661 LM_ST_EXCLUSIVE, GL_NEVER_RECURSE,
664 fs_err(sdp, "can't lock local \"ut\" file: %d\n", error);
668 error = gfs2_glock_nq_init(get_v2ip(sdp->sd_qc_inode)->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");
789 error = init_names(sdp, silent);
793 error = gfs2_sys_fs_add(sdp);
797 error = gfs2_lm_mount(sdp, silent);
801 error = init_locking(sdp, &mount_gh, DO);
805 error = init_sb(sdp, silent, DO);
809 error = init_inodes(sdp, DO);
813 error = init_per_node(sdp, DO);
817 error = gfs2_statfs_init(sdp);
819 fs_err(sdp, "can't initialize statfs subsystem: %d\n", error);
823 error = init_threads(sdp, DO);
827 if (!(sb->s_flags & MS_RDONLY)) {
828 error = gfs2_make_fs_rw(sdp);
830 fs_err(sdp, "can't make FS RW: %d\n", error);
835 gfs2_glock_dq_uninit(&mount_gh);
840 init_threads(sdp, UNDO);
843 init_per_node(sdp, UNDO);
846 init_inodes(sdp, UNDO);
849 init_sb(sdp, 0, UNDO);
852 init_locking(sdp, &mount_gh, UNDO);
855 gfs2_gl_hash_clear(sdp, WAIT);
856 gfs2_lm_unmount(sdp);
857 while (invalidate_inodes(sb))
861 gfs2_sys_fs_del(sdp);
870 static struct super_block *gfs2_get_sb(struct file_system_type *fs_type,
871 int flags, const char *dev_name,
874 return get_sb_bdev(fs_type, flags, dev_name, data, fill_super);
877 struct file_system_type gfs2_fs_type = {
879 .fs_flags = FS_REQUIRES_DEV,
880 .get_sb = gfs2_get_sb,
881 .kill_sb = kill_block_super,
882 .owner = THIS_MODULE,