2 * Copyright (C) Sistina Software, Inc. 1997-2003 All rights reserved.
3 * Copyright (C) 2004-2007 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 version 2.
10 #include <linux/bio.h>
11 #include <linux/sched.h>
12 #include <linux/slab.h>
13 #include <linux/spinlock.h>
14 #include <linux/completion.h>
15 #include <linux/buffer_head.h>
16 #include <linux/statfs.h>
17 #include <linux/seq_file.h>
18 #include <linux/mount.h>
19 #include <linux/kthread.h>
20 #include <linux/delay.h>
21 #include <linux/gfs2_ondisk.h>
22 #include <linux/crc32.h>
23 #include <linux/time.h>
43 #define args_neq(a1, a2, x) ((a1)->ar_##x != (a2)->ar_##x)
79 static const match_table_t tokens = {
80 {Opt_lockproto, "lockproto=%s"},
81 {Opt_locktable, "locktable=%s"},
82 {Opt_hostdata, "hostdata=%s"},
83 {Opt_spectator, "spectator"},
84 {Opt_ignore_local_fs, "ignore_local_fs"},
85 {Opt_localflocks, "localflocks"},
86 {Opt_localcaching, "localcaching"},
88 {Opt_nodebug, "nodebug"},
89 {Opt_upgrade, "upgrade"},
92 {Opt_quota_off, "quota=off"},
93 {Opt_quota_account, "quota=account"},
94 {Opt_quota_on, "quota=on"},
96 {Opt_noquota, "noquota"},
97 {Opt_suiddir, "suiddir"},
98 {Opt_nosuiddir, "nosuiddir"},
99 {Opt_data_writeback, "data=writeback"},
100 {Opt_data_ordered, "data=ordered"},
102 {Opt_discard, "discard"},
103 {Opt_nodiscard, "nodiscard"},
104 {Opt_commit, "commit=%d"},
105 {Opt_err_withdraw, "errors=withdraw"},
106 {Opt_err_panic, "errors=panic"},
107 {Opt_statfs_quantum, "statfs_quantum=%d"},
108 {Opt_statfs_percent, "statfs_percent=%d"},
109 {Opt_quota_quantum, "quota_quantum=%d"},
114 * gfs2_mount_args - Parse mount options
115 * @args: The structure into which the parsed options will be written
116 * @options: The options to parse
121 int gfs2_mount_args(struct gfs2_args *args, char *options)
125 substring_t tmp[MAX_OPT_ARGS];
128 /* Split the options into tokens with the "," character and
132 o = strsep(&options, ",");
138 token = match_token(o, tokens, tmp);
141 match_strlcpy(args->ar_lockproto, &tmp[0],
145 match_strlcpy(args->ar_locktable, &tmp[0],
149 match_strlcpy(args->ar_hostdata, &tmp[0],
153 args->ar_spectator = 1;
155 case Opt_ignore_local_fs:
156 args->ar_ignore_local_fs = 1;
158 case Opt_localflocks:
159 args->ar_localflocks = 1;
161 case Opt_localcaching:
162 args->ar_localcaching = 1;
165 if (args->ar_errors == GFS2_ERRORS_PANIC) {
166 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
167 "are mutually exclusive.\n");
176 args->ar_upgrade = 1;
179 args->ar_posix_acl = 1;
182 args->ar_posix_acl = 0;
186 args->ar_quota = GFS2_QUOTA_OFF;
188 case Opt_quota_account:
189 args->ar_quota = GFS2_QUOTA_ACCOUNT;
193 args->ar_quota = GFS2_QUOTA_ON;
196 args->ar_suiddir = 1;
199 args->ar_suiddir = 0;
201 case Opt_data_writeback:
202 args->ar_data = GFS2_DATA_WRITEBACK;
204 case Opt_data_ordered:
205 args->ar_data = GFS2_DATA_ORDERED;
211 args->ar_discard = 1;
214 args->ar_discard = 0;
217 rv = match_int(&tmp[0], &args->ar_commit);
218 if (rv || args->ar_commit <= 0) {
219 printk(KERN_WARNING "GFS2: commit mount option requires a positive numeric argument\n");
220 return rv ? rv : -EINVAL;
223 case Opt_statfs_quantum:
224 rv = match_int(&tmp[0], &args->ar_statfs_quantum);
225 if (rv || args->ar_statfs_quantum < 0) {
226 printk(KERN_WARNING "GFS2: statfs_quantum mount option requires a non-negative numeric argument\n");
227 return rv ? rv : -EINVAL;
230 case Opt_quota_quantum:
231 rv = match_int(&tmp[0], &args->ar_quota_quantum);
232 if (rv || args->ar_quota_quantum <= 0) {
233 printk(KERN_WARNING "GFS2: quota_quantum mount option requires a positive numeric argument\n");
234 return rv ? rv : -EINVAL;
237 case Opt_statfs_percent:
238 rv = match_int(&tmp[0], &args->ar_statfs_percent);
239 if (rv || args->ar_statfs_percent < 0 ||
240 args->ar_statfs_percent > 100) {
241 printk(KERN_WARNING "statfs_percent mount option requires a numeric argument between 0 and 100\n");
242 return rv ? rv : -EINVAL;
245 case Opt_err_withdraw:
246 args->ar_errors = GFS2_ERRORS_WITHDRAW;
249 if (args->ar_debug) {
250 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
251 "are mutually exclusive.\n");
254 args->ar_errors = GFS2_ERRORS_PANIC;
258 printk(KERN_WARNING "GFS2: invalid mount option: %s\n", o);
267 * gfs2_jindex_free - Clear all the journal index information
268 * @sdp: The GFS2 superblock
272 void gfs2_jindex_free(struct gfs2_sbd *sdp)
274 struct list_head list, *head;
275 struct gfs2_jdesc *jd;
276 struct gfs2_journal_extent *jext;
278 spin_lock(&sdp->sd_jindex_spin);
279 list_add(&list, &sdp->sd_jindex_list);
280 list_del_init(&sdp->sd_jindex_list);
281 sdp->sd_journals = 0;
282 spin_unlock(&sdp->sd_jindex_spin);
284 while (!list_empty(&list)) {
285 jd = list_entry(list.next, struct gfs2_jdesc, jd_list);
286 head = &jd->extent_list;
287 while (!list_empty(head)) {
288 jext = list_entry(head->next,
289 struct gfs2_journal_extent,
291 list_del(&jext->extent_list);
294 list_del(&jd->jd_list);
300 static struct gfs2_jdesc *jdesc_find_i(struct list_head *head, unsigned int jid)
302 struct gfs2_jdesc *jd;
305 list_for_each_entry(jd, head, jd_list) {
306 if (jd->jd_jid == jid) {
318 struct gfs2_jdesc *gfs2_jdesc_find(struct gfs2_sbd *sdp, unsigned int jid)
320 struct gfs2_jdesc *jd;
322 spin_lock(&sdp->sd_jindex_spin);
323 jd = jdesc_find_i(&sdp->sd_jindex_list, jid);
324 spin_unlock(&sdp->sd_jindex_spin);
329 int gfs2_jdesc_check(struct gfs2_jdesc *jd)
331 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
332 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
336 if (ip->i_disksize < (8 << 20) || ip->i_disksize > (1 << 30) ||
337 (ip->i_disksize & (sdp->sd_sb.sb_bsize - 1))) {
338 gfs2_consist_inode(ip);
341 jd->jd_blocks = ip->i_disksize >> sdp->sd_sb.sb_bsize_shift;
343 error = gfs2_write_alloc_required(ip, 0, ip->i_disksize, &ar);
345 gfs2_consist_inode(ip);
353 * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one
354 * @sdp: the filesystem
359 int gfs2_make_fs_rw(struct gfs2_sbd *sdp)
361 struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode);
362 struct gfs2_glock *j_gl = ip->i_gl;
363 struct gfs2_holder t_gh;
364 struct gfs2_log_header_host head;
367 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, 0, &t_gh);
371 j_gl->gl_ops->go_inval(j_gl, DIO_METADATA);
373 error = gfs2_find_jhead(sdp->sd_jdesc, &head);
377 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
383 /* Initialize some head of the log stuff */
384 sdp->sd_log_sequence = head.lh_sequence + 1;
385 gfs2_log_pointers_init(sdp, head.lh_blkno);
387 error = gfs2_quota_init(sdp);
391 set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
393 gfs2_glock_dq_uninit(&t_gh);
398 t_gh.gh_flags |= GL_NOCACHE;
399 gfs2_glock_dq_uninit(&t_gh);
404 void gfs2_statfs_change_in(struct gfs2_statfs_change_host *sc, const void *buf)
406 const struct gfs2_statfs_change *str = buf;
408 sc->sc_total = be64_to_cpu(str->sc_total);
409 sc->sc_free = be64_to_cpu(str->sc_free);
410 sc->sc_dinodes = be64_to_cpu(str->sc_dinodes);
413 static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host *sc, void *buf)
415 struct gfs2_statfs_change *str = buf;
417 str->sc_total = cpu_to_be64(sc->sc_total);
418 str->sc_free = cpu_to_be64(sc->sc_free);
419 str->sc_dinodes = cpu_to_be64(sc->sc_dinodes);
422 int gfs2_statfs_init(struct gfs2_sbd *sdp)
424 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
425 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
426 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
427 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
428 struct buffer_head *m_bh, *l_bh;
429 struct gfs2_holder gh;
432 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
437 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
441 if (sdp->sd_args.ar_spectator) {
442 spin_lock(&sdp->sd_statfs_spin);
443 gfs2_statfs_change_in(m_sc, m_bh->b_data +
444 sizeof(struct gfs2_dinode));
445 spin_unlock(&sdp->sd_statfs_spin);
447 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
451 spin_lock(&sdp->sd_statfs_spin);
452 gfs2_statfs_change_in(m_sc, m_bh->b_data +
453 sizeof(struct gfs2_dinode));
454 gfs2_statfs_change_in(l_sc, l_bh->b_data +
455 sizeof(struct gfs2_dinode));
456 spin_unlock(&sdp->sd_statfs_spin);
464 gfs2_glock_dq_uninit(&gh);
468 void gfs2_statfs_change(struct gfs2_sbd *sdp, s64 total, s64 free,
471 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
472 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
473 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
474 struct buffer_head *l_bh;
479 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
483 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
485 spin_lock(&sdp->sd_statfs_spin);
486 l_sc->sc_total += total;
487 l_sc->sc_free += free;
488 l_sc->sc_dinodes += dinodes;
489 gfs2_statfs_change_out(l_sc, l_bh->b_data + sizeof(struct gfs2_dinode));
490 if (sdp->sd_args.ar_statfs_percent) {
491 x = 100 * l_sc->sc_free;
492 y = m_sc->sc_free * sdp->sd_args.ar_statfs_percent;
493 if (x >= y || x <= -y)
496 spin_unlock(&sdp->sd_statfs_spin);
500 gfs2_wake_up_statfs(sdp);
503 void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
504 struct buffer_head *l_bh)
506 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
507 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
508 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
509 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
511 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
513 spin_lock(&sdp->sd_statfs_spin);
514 m_sc->sc_total += l_sc->sc_total;
515 m_sc->sc_free += l_sc->sc_free;
516 m_sc->sc_dinodes += l_sc->sc_dinodes;
517 memset(l_sc, 0, sizeof(struct gfs2_statfs_change));
518 memset(l_bh->b_data + sizeof(struct gfs2_dinode),
519 0, sizeof(struct gfs2_statfs_change));
520 spin_unlock(&sdp->sd_statfs_spin);
522 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
523 gfs2_statfs_change_out(m_sc, m_bh->b_data + sizeof(struct gfs2_dinode));
526 int gfs2_statfs_sync(struct super_block *sb, int type)
528 struct gfs2_sbd *sdp = sb->s_fs_info;
529 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
530 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
531 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
532 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
533 struct gfs2_holder gh;
534 struct buffer_head *m_bh, *l_bh;
537 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
542 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
546 spin_lock(&sdp->sd_statfs_spin);
547 gfs2_statfs_change_in(m_sc, m_bh->b_data +
548 sizeof(struct gfs2_dinode));
549 if (!l_sc->sc_total && !l_sc->sc_free && !l_sc->sc_dinodes) {
550 spin_unlock(&sdp->sd_statfs_spin);
553 spin_unlock(&sdp->sd_statfs_spin);
555 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
559 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
563 update_statfs(sdp, m_bh, l_bh);
564 sdp->sd_statfs_force_sync = 0;
573 gfs2_glock_dq_uninit(&gh);
578 struct list_head list;
579 struct gfs2_holder gh;
583 * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all
585 * @sdp: the file system
586 * @state: the state to put the transaction lock into
587 * @t_gh: the hold on the transaction lock
592 static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp,
593 struct gfs2_holder *t_gh)
595 struct gfs2_inode *ip;
596 struct gfs2_jdesc *jd;
599 struct gfs2_log_header_host lh;
602 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
603 lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL);
608 ip = GFS2_I(jd->jd_inode);
609 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &lfcc->gh);
614 list_add(&lfcc->list, &list);
617 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_DEFERRED,
620 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
621 error = gfs2_jdesc_check(jd);
624 error = gfs2_find_jhead(jd, &lh);
627 if (!(lh.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
634 gfs2_glock_dq_uninit(t_gh);
637 while (!list_empty(&list)) {
638 lfcc = list_entry(list.next, struct lfcc, list);
639 list_del(&lfcc->list);
640 gfs2_glock_dq_uninit(&lfcc->gh);
647 * gfs2_freeze_fs - freezes the file system
648 * @sdp: the file system
650 * This function flushes data and meta data for all machines by
651 * aquiring the transaction log exclusively. All journals are
652 * ensured to be in a clean state as well.
657 int gfs2_freeze_fs(struct gfs2_sbd *sdp)
661 mutex_lock(&sdp->sd_freeze_lock);
663 if (!sdp->sd_freeze_count++) {
664 error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh);
666 sdp->sd_freeze_count--;
669 mutex_unlock(&sdp->sd_freeze_lock);
675 * gfs2_unfreeze_fs - unfreezes the file system
676 * @sdp: the file system
678 * This function allows the file system to proceed by unlocking
679 * the exclusively held transaction lock. Other GFS2 nodes are
680 * now free to acquire the lock shared and go on with their lives.
684 void gfs2_unfreeze_fs(struct gfs2_sbd *sdp)
686 mutex_lock(&sdp->sd_freeze_lock);
688 if (sdp->sd_freeze_count && !--sdp->sd_freeze_count)
689 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
691 mutex_unlock(&sdp->sd_freeze_lock);
696 * gfs2_write_inode - Make sure the inode is stable on the disk
698 * @sync: synchronous write flag
703 static int gfs2_write_inode(struct inode *inode, int sync)
705 struct gfs2_inode *ip = GFS2_I(inode);
706 struct gfs2_sbd *sdp = GFS2_SB(inode);
707 struct gfs2_holder gh;
708 struct buffer_head *bh;
709 struct timespec atime;
710 struct gfs2_dinode *di;
713 /* Check this is a "normal" inode, etc */
714 if (!test_bit(GIF_USER, &ip->i_flags) ||
715 (current->flags & PF_MEMALLOC))
717 ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
720 ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
723 ret = gfs2_meta_inode_buffer(ip, &bh);
725 di = (struct gfs2_dinode *)bh->b_data;
726 atime.tv_sec = be64_to_cpu(di->di_atime);
727 atime.tv_nsec = be32_to_cpu(di->di_atime_nsec);
728 if (timespec_compare(&inode->i_atime, &atime) > 0) {
729 gfs2_trans_add_bh(ip->i_gl, bh, 1);
730 gfs2_dinode_out(ip, bh->b_data);
736 gfs2_glock_dq_uninit(&gh);
739 gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
744 * gfs2_make_fs_ro - Turn a Read-Write FS into a Read-Only one
745 * @sdp: the filesystem
750 static int gfs2_make_fs_ro(struct gfs2_sbd *sdp)
752 struct gfs2_holder t_gh;
755 flush_workqueue(gfs2_delete_workqueue);
756 gfs2_quota_sync(sdp->sd_vfs, 0);
757 gfs2_statfs_sync(sdp->sd_vfs, 0);
759 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, GL_NOCACHE,
761 if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
764 gfs2_meta_syncfs(sdp);
765 gfs2_log_shutdown(sdp);
767 clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
770 gfs2_glock_dq_uninit(&t_gh);
772 gfs2_quota_cleanup(sdp);
777 static int gfs2_umount_recovery_wait(void *word)
784 * gfs2_put_super - Unmount the filesystem
785 * @sb: The VFS superblock
789 static void gfs2_put_super(struct super_block *sb)
791 struct gfs2_sbd *sdp = sb->s_fs_info;
793 struct gfs2_jdesc *jd;
795 /* Unfreeze the filesystem, if we need to */
797 mutex_lock(&sdp->sd_freeze_lock);
798 if (sdp->sd_freeze_count)
799 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
800 mutex_unlock(&sdp->sd_freeze_lock);
802 /* No more recovery requests */
803 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
806 /* Wait on outstanding recovery */
808 spin_lock(&sdp->sd_jindex_spin);
809 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
810 if (!test_bit(JDF_RECOVERY, &jd->jd_flags))
812 spin_unlock(&sdp->sd_jindex_spin);
813 wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
814 gfs2_umount_recovery_wait, TASK_UNINTERRUPTIBLE);
817 spin_unlock(&sdp->sd_jindex_spin);
819 kthread_stop(sdp->sd_quotad_process);
820 kthread_stop(sdp->sd_logd_process);
822 if (!(sb->s_flags & MS_RDONLY)) {
823 error = gfs2_make_fs_ro(sdp);
827 /* At this point, we're through modifying the disk */
831 iput(sdp->sd_jindex);
832 iput(sdp->sd_statfs_inode);
833 iput(sdp->sd_rindex);
834 iput(sdp->sd_quota_inode);
836 gfs2_glock_put(sdp->sd_rename_gl);
837 gfs2_glock_put(sdp->sd_trans_gl);
839 if (!sdp->sd_args.ar_spectator) {
840 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
841 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
842 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
843 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
844 iput(sdp->sd_sc_inode);
845 iput(sdp->sd_qc_inode);
848 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
849 gfs2_clear_rgrpd(sdp);
850 gfs2_jindex_free(sdp);
851 /* Take apart glock structures and buffer lists */
852 gfs2_gl_hash_clear(sdp);
853 /* Unmount the locking protocol */
854 gfs2_lm_unmount(sdp);
856 /* At this point, we're through participating in the lockspace */
857 gfs2_sys_fs_del(sdp);
861 * gfs2_sync_fs - sync the filesystem
862 * @sb: the superblock
864 * Flushes the log to disk.
867 static int gfs2_sync_fs(struct super_block *sb, int wait)
869 if (wait && sb->s_fs_info)
870 gfs2_log_flush(sb->s_fs_info, NULL);
875 * gfs2_freeze - prevent further writes to the filesystem
876 * @sb: the VFS structure for the filesystem
880 static int gfs2_freeze(struct super_block *sb)
882 struct gfs2_sbd *sdp = sb->s_fs_info;
885 if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
889 error = gfs2_freeze_fs(sdp);
895 fs_err(sdp, "waiting for recovery before freeze\n");
899 fs_err(sdp, "error freezing FS: %d\n", error);
903 fs_err(sdp, "retrying...\n");
910 * gfs2_unfreeze - reallow writes to the filesystem
911 * @sb: the VFS structure for the filesystem
915 static int gfs2_unfreeze(struct super_block *sb)
917 gfs2_unfreeze_fs(sb->s_fs_info);
922 * statfs_fill - fill in the sg for a given RG
924 * @sc: the sc structure
926 * Returns: 0 on success, -ESTALE if the LVB is invalid
929 static int statfs_slow_fill(struct gfs2_rgrpd *rgd,
930 struct gfs2_statfs_change_host *sc)
932 gfs2_rgrp_verify(rgd);
933 sc->sc_total += rgd->rd_data;
934 sc->sc_free += rgd->rd_free;
935 sc->sc_dinodes += rgd->rd_dinodes;
940 * gfs2_statfs_slow - Stat a filesystem using asynchronous locking
941 * @sdp: the filesystem
942 * @sc: the sc info that will be returned
944 * Any error (other than a signal) will cause this routine to fall back
945 * to the synchronous version.
947 * FIXME: This really shouldn't busy wait like this.
952 static int gfs2_statfs_slow(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
954 struct gfs2_holder ri_gh;
955 struct gfs2_rgrpd *rgd_next;
956 struct gfs2_holder *gha, *gh;
957 unsigned int slots = 64;
962 memset(sc, 0, sizeof(struct gfs2_statfs_change_host));
963 gha = kcalloc(slots, sizeof(struct gfs2_holder), GFP_KERNEL);
967 error = gfs2_rindex_hold(sdp, &ri_gh);
971 rgd_next = gfs2_rgrpd_get_first(sdp);
976 for (x = 0; x < slots; x++) {
979 if (gh->gh_gl && gfs2_glock_poll(gh)) {
980 err = gfs2_glock_wait(gh);
982 gfs2_holder_uninit(gh);
986 error = statfs_slow_fill(
987 gh->gh_gl->gl_object, sc);
988 gfs2_glock_dq_uninit(gh);
994 else if (rgd_next && !error) {
995 error = gfs2_glock_nq_init(rgd_next->rd_gl,
999 rgd_next = gfs2_rgrpd_get_next(rgd_next);
1003 if (signal_pending(current))
1004 error = -ERESTARTSYS;
1013 gfs2_glock_dq_uninit(&ri_gh);
1021 * gfs2_statfs_i - Do a statfs
1022 * @sdp: the filesystem
1023 * @sg: the sg structure
1028 static int gfs2_statfs_i(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
1030 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
1031 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
1033 spin_lock(&sdp->sd_statfs_spin);
1036 sc->sc_total += l_sc->sc_total;
1037 sc->sc_free += l_sc->sc_free;
1038 sc->sc_dinodes += l_sc->sc_dinodes;
1040 spin_unlock(&sdp->sd_statfs_spin);
1042 if (sc->sc_free < 0)
1044 if (sc->sc_free > sc->sc_total)
1045 sc->sc_free = sc->sc_total;
1046 if (sc->sc_dinodes < 0)
1053 * gfs2_statfs - Gather and return stats about the filesystem
1054 * @sb: The superblock
1055 * @statfsbuf: The buffer
1057 * Returns: 0 on success or error code
1060 static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1062 struct super_block *sb = dentry->d_inode->i_sb;
1063 struct gfs2_sbd *sdp = sb->s_fs_info;
1064 struct gfs2_statfs_change_host sc;
1067 if (gfs2_tune_get(sdp, gt_statfs_slow))
1068 error = gfs2_statfs_slow(sdp, &sc);
1070 error = gfs2_statfs_i(sdp, &sc);
1075 buf->f_type = GFS2_MAGIC;
1076 buf->f_bsize = sdp->sd_sb.sb_bsize;
1077 buf->f_blocks = sc.sc_total;
1078 buf->f_bfree = sc.sc_free;
1079 buf->f_bavail = sc.sc_free;
1080 buf->f_files = sc.sc_dinodes + sc.sc_free;
1081 buf->f_ffree = sc.sc_free;
1082 buf->f_namelen = GFS2_FNAMESIZE;
1088 * gfs2_remount_fs - called when the FS is remounted
1089 * @sb: the filesystem
1090 * @flags: the remount flags
1091 * @data: extra data passed in (not used right now)
1096 static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
1098 struct gfs2_sbd *sdp = sb->s_fs_info;
1099 struct gfs2_args args = sdp->sd_args; /* Default to current settings */
1100 struct gfs2_tune *gt = &sdp->sd_tune;
1103 spin_lock(>->gt_spin);
1104 args.ar_commit = gt->gt_log_flush_secs;
1105 args.ar_quota_quantum = gt->gt_quota_quantum;
1106 if (gt->gt_statfs_slow)
1107 args.ar_statfs_quantum = 0;
1109 args.ar_statfs_quantum = gt->gt_statfs_quantum;
1110 spin_unlock(>->gt_spin);
1111 error = gfs2_mount_args(&args, data);
1115 /* Not allowed to change locking details */
1116 if (strcmp(args.ar_lockproto, sdp->sd_args.ar_lockproto) ||
1117 strcmp(args.ar_locktable, sdp->sd_args.ar_locktable) ||
1118 strcmp(args.ar_hostdata, sdp->sd_args.ar_hostdata))
1121 /* Some flags must not be changed */
1122 if (args_neq(&args, &sdp->sd_args, spectator) ||
1123 args_neq(&args, &sdp->sd_args, ignore_local_fs) ||
1124 args_neq(&args, &sdp->sd_args, localflocks) ||
1125 args_neq(&args, &sdp->sd_args, localcaching) ||
1126 args_neq(&args, &sdp->sd_args, meta))
1129 if (sdp->sd_args.ar_spectator)
1130 *flags |= MS_RDONLY;
1132 if ((sb->s_flags ^ *flags) & MS_RDONLY) {
1133 if (*flags & MS_RDONLY)
1134 error = gfs2_make_fs_ro(sdp);
1136 error = gfs2_make_fs_rw(sdp);
1141 sdp->sd_args = args;
1142 if (sdp->sd_args.ar_posix_acl)
1143 sb->s_flags |= MS_POSIXACL;
1145 sb->s_flags &= ~MS_POSIXACL;
1146 spin_lock(>->gt_spin);
1147 gt->gt_log_flush_secs = args.ar_commit;
1148 gt->gt_quota_quantum = args.ar_quota_quantum;
1149 if (args.ar_statfs_quantum) {
1150 gt->gt_statfs_slow = 0;
1151 gt->gt_statfs_quantum = args.ar_statfs_quantum;
1154 gt->gt_statfs_slow = 1;
1155 gt->gt_statfs_quantum = 30;
1157 spin_unlock(>->gt_spin);
1159 gfs2_online_uevent(sdp);
1164 * gfs2_drop_inode - Drop an inode (test for remote unlink)
1165 * @inode: The inode to drop
1167 * If we've received a callback on an iopen lock then its because a
1168 * remote node tried to deallocate the inode but failed due to this node
1169 * still having the inode open. Here we mark the link count zero
1170 * since we know that it must have reached zero if the GLF_DEMOTE flag
1171 * is set on the iopen glock. If we didn't do a disk read since the
1172 * remote node removed the final link then we might otherwise miss
1173 * this event. This check ensures that this node will deallocate the
1174 * inode's blocks, or alternatively pass the baton on to another
1175 * node for later deallocation.
1178 static void gfs2_drop_inode(struct inode *inode)
1180 struct gfs2_inode *ip = GFS2_I(inode);
1182 if (test_bit(GIF_USER, &ip->i_flags) && inode->i_nlink) {
1183 struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
1184 if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags))
1187 generic_drop_inode(inode);
1191 * gfs2_clear_inode - Deallocate an inode when VFS is done with it
1192 * @inode: The VFS inode
1196 static void gfs2_clear_inode(struct inode *inode)
1198 struct gfs2_inode *ip = GFS2_I(inode);
1200 /* This tells us its a "real" inode and not one which only
1201 * serves to contain an address space (see rgrp.c, meta_io.c)
1202 * which therefore doesn't have its own glocks.
1204 if (test_bit(GIF_USER, &ip->i_flags)) {
1205 ip->i_gl->gl_object = NULL;
1206 gfs2_glock_put(ip->i_gl);
1208 if (ip->i_iopen_gh.gh_gl) {
1209 ip->i_iopen_gh.gh_gl->gl_object = NULL;
1210 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1215 static int is_ancestor(const struct dentry *d1, const struct dentry *d2)
1221 } while (!IS_ROOT(d1));
1226 * gfs2_show_options - Show mount options for /proc/mounts
1227 * @s: seq_file structure
1230 * Returns: 0 on success or error code
1233 static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1235 struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info;
1236 struct gfs2_args *args = &sdp->sd_args;
1239 if (is_ancestor(mnt->mnt_root, sdp->sd_master_dir))
1240 seq_printf(s, ",meta");
1241 if (args->ar_lockproto[0])
1242 seq_printf(s, ",lockproto=%s", args->ar_lockproto);
1243 if (args->ar_locktable[0])
1244 seq_printf(s, ",locktable=%s", args->ar_locktable);
1245 if (args->ar_hostdata[0])
1246 seq_printf(s, ",hostdata=%s", args->ar_hostdata);
1247 if (args->ar_spectator)
1248 seq_printf(s, ",spectator");
1249 if (args->ar_ignore_local_fs)
1250 seq_printf(s, ",ignore_local_fs");
1251 if (args->ar_localflocks)
1252 seq_printf(s, ",localflocks");
1253 if (args->ar_localcaching)
1254 seq_printf(s, ",localcaching");
1256 seq_printf(s, ",debug");
1257 if (args->ar_upgrade)
1258 seq_printf(s, ",upgrade");
1259 if (args->ar_posix_acl)
1260 seq_printf(s, ",acl");
1261 if (args->ar_quota != GFS2_QUOTA_DEFAULT) {
1263 switch (args->ar_quota) {
1264 case GFS2_QUOTA_OFF:
1267 case GFS2_QUOTA_ACCOUNT:
1277 seq_printf(s, ",quota=%s", state);
1279 if (args->ar_suiddir)
1280 seq_printf(s, ",suiddir");
1281 if (args->ar_data != GFS2_DATA_DEFAULT) {
1283 switch (args->ar_data) {
1284 case GFS2_DATA_WRITEBACK:
1285 state = "writeback";
1287 case GFS2_DATA_ORDERED:
1294 seq_printf(s, ",data=%s", state);
1296 if (args->ar_discard)
1297 seq_printf(s, ",discard");
1298 val = sdp->sd_tune.gt_log_flush_secs;
1300 seq_printf(s, ",commit=%d", val);
1301 val = sdp->sd_tune.gt_statfs_quantum;
1303 seq_printf(s, ",statfs_quantum=%d", val);
1304 val = sdp->sd_tune.gt_quota_quantum;
1306 seq_printf(s, ",quota_quantum=%d", val);
1307 if (args->ar_statfs_percent)
1308 seq_printf(s, ",statfs_percent=%d", args->ar_statfs_percent);
1309 if (args->ar_errors != GFS2_ERRORS_DEFAULT) {
1312 switch (args->ar_errors) {
1313 case GFS2_ERRORS_WITHDRAW:
1316 case GFS2_ERRORS_PANIC:
1323 seq_printf(s, ",errors=%s", state);
1329 * We have to (at the moment) hold the inodes main lock to cover
1330 * the gap between unlocking the shared lock on the iopen lock and
1331 * taking the exclusive lock. I'd rather do a shared -> exclusive
1332 * conversion on the iopen lock, but we can change that later. This
1333 * is safe, just less efficient.
1336 static void gfs2_delete_inode(struct inode *inode)
1338 struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
1339 struct gfs2_inode *ip = GFS2_I(inode);
1340 struct gfs2_holder gh;
1343 if (!test_bit(GIF_USER, &ip->i_flags))
1346 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
1347 if (unlikely(error)) {
1348 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1352 error = gfs2_check_blk_type(sdp, ip->i_no_addr, GFS2_BLKST_UNLINKED);
1356 gfs2_glock_dq_wait(&ip->i_iopen_gh);
1357 gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh);
1358 error = gfs2_glock_nq(&ip->i_iopen_gh);
1362 if (S_ISDIR(inode->i_mode) &&
1363 (ip->i_diskflags & GFS2_DIF_EXHASH)) {
1364 error = gfs2_dir_exhash_dealloc(ip);
1370 error = gfs2_ea_dealloc(ip);
1375 if (!gfs2_is_stuffed(ip)) {
1376 error = gfs2_file_dealloc(ip);
1381 error = gfs2_dinode_dealloc(ip);
1386 error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks);
1389 /* Needs to be done before glock release & also in a transaction */
1390 truncate_inode_pages(&inode->i_data, 0);
1391 gfs2_trans_end(sdp);
1394 if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags))
1395 gfs2_glock_dq(&ip->i_iopen_gh);
1396 gfs2_holder_uninit(&ip->i_iopen_gh);
1397 gfs2_glock_dq_uninit(&gh);
1398 if (error && error != GLR_TRYFAILED && error != -EROFS)
1399 fs_warn(sdp, "gfs2_delete_inode: %d\n", error);
1401 truncate_inode_pages(&inode->i_data, 0);
1405 static struct inode *gfs2_alloc_inode(struct super_block *sb)
1407 struct gfs2_inode *ip;
1409 ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL);
1414 return &ip->i_inode;
1417 static void gfs2_destroy_inode(struct inode *inode)
1419 kmem_cache_free(gfs2_inode_cachep, inode);
1422 const struct super_operations gfs2_super_ops = {
1423 .alloc_inode = gfs2_alloc_inode,
1424 .destroy_inode = gfs2_destroy_inode,
1425 .write_inode = gfs2_write_inode,
1426 .delete_inode = gfs2_delete_inode,
1427 .put_super = gfs2_put_super,
1428 .sync_fs = gfs2_sync_fs,
1429 .freeze_fs = gfs2_freeze,
1430 .unfreeze_fs = gfs2_unfreeze,
1431 .statfs = gfs2_statfs,
1432 .remount_fs = gfs2_remount_fs,
1433 .clear_inode = gfs2_clear_inode,
1434 .drop_inode = gfs2_drop_inode,
1435 .show_options = gfs2_show_options,