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>
24 #include <linux/wait.h>
25 #include <linux/writeback.h>
45 #define args_neq(a1, a2, x) ((a1)->ar_##x != (a2)->ar_##x)
83 static const match_table_t tokens = {
84 {Opt_lockproto, "lockproto=%s"},
85 {Opt_locktable, "locktable=%s"},
86 {Opt_hostdata, "hostdata=%s"},
87 {Opt_spectator, "spectator"},
88 {Opt_ignore_local_fs, "ignore_local_fs"},
89 {Opt_localflocks, "localflocks"},
90 {Opt_localcaching, "localcaching"},
92 {Opt_nodebug, "nodebug"},
93 {Opt_upgrade, "upgrade"},
96 {Opt_quota_off, "quota=off"},
97 {Opt_quota_account, "quota=account"},
98 {Opt_quota_on, "quota=on"},
100 {Opt_noquota, "noquota"},
101 {Opt_suiddir, "suiddir"},
102 {Opt_nosuiddir, "nosuiddir"},
103 {Opt_data_writeback, "data=writeback"},
104 {Opt_data_ordered, "data=ordered"},
106 {Opt_discard, "discard"},
107 {Opt_nodiscard, "nodiscard"},
108 {Opt_commit, "commit=%d"},
109 {Opt_err_withdraw, "errors=withdraw"},
110 {Opt_err_panic, "errors=panic"},
111 {Opt_statfs_quantum, "statfs_quantum=%d"},
112 {Opt_statfs_percent, "statfs_percent=%d"},
113 {Opt_quota_quantum, "quota_quantum=%d"},
114 {Opt_barrier, "barrier"},
115 {Opt_nobarrier, "nobarrier"},
120 * gfs2_mount_args - Parse mount options
121 * @args: The structure into which the parsed options will be written
122 * @options: The options to parse
127 int gfs2_mount_args(struct gfs2_args *args, char *options)
131 substring_t tmp[MAX_OPT_ARGS];
134 /* Split the options into tokens with the "," character and
138 o = strsep(&options, ",");
144 token = match_token(o, tokens, tmp);
147 match_strlcpy(args->ar_lockproto, &tmp[0],
151 match_strlcpy(args->ar_locktable, &tmp[0],
155 match_strlcpy(args->ar_hostdata, &tmp[0],
159 args->ar_spectator = 1;
161 case Opt_ignore_local_fs:
162 args->ar_ignore_local_fs = 1;
164 case Opt_localflocks:
165 args->ar_localflocks = 1;
167 case Opt_localcaching:
168 args->ar_localcaching = 1;
171 if (args->ar_errors == GFS2_ERRORS_PANIC) {
172 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
173 "are mutually exclusive.\n");
182 args->ar_upgrade = 1;
185 args->ar_posix_acl = 1;
188 args->ar_posix_acl = 0;
192 args->ar_quota = GFS2_QUOTA_OFF;
194 case Opt_quota_account:
195 args->ar_quota = GFS2_QUOTA_ACCOUNT;
199 args->ar_quota = GFS2_QUOTA_ON;
202 args->ar_suiddir = 1;
205 args->ar_suiddir = 0;
207 case Opt_data_writeback:
208 args->ar_data = GFS2_DATA_WRITEBACK;
210 case Opt_data_ordered:
211 args->ar_data = GFS2_DATA_ORDERED;
217 args->ar_discard = 1;
220 args->ar_discard = 0;
223 rv = match_int(&tmp[0], &args->ar_commit);
224 if (rv || args->ar_commit <= 0) {
225 printk(KERN_WARNING "GFS2: commit mount option requires a positive numeric argument\n");
226 return rv ? rv : -EINVAL;
229 case Opt_statfs_quantum:
230 rv = match_int(&tmp[0], &args->ar_statfs_quantum);
231 if (rv || args->ar_statfs_quantum < 0) {
232 printk(KERN_WARNING "GFS2: statfs_quantum mount option requires a non-negative numeric argument\n");
233 return rv ? rv : -EINVAL;
236 case Opt_quota_quantum:
237 rv = match_int(&tmp[0], &args->ar_quota_quantum);
238 if (rv || args->ar_quota_quantum <= 0) {
239 printk(KERN_WARNING "GFS2: quota_quantum mount option requires a positive numeric argument\n");
240 return rv ? rv : -EINVAL;
243 case Opt_statfs_percent:
244 rv = match_int(&tmp[0], &args->ar_statfs_percent);
245 if (rv || args->ar_statfs_percent < 0 ||
246 args->ar_statfs_percent > 100) {
247 printk(KERN_WARNING "statfs_percent mount option requires a numeric argument between 0 and 100\n");
248 return rv ? rv : -EINVAL;
251 case Opt_err_withdraw:
252 args->ar_errors = GFS2_ERRORS_WITHDRAW;
255 if (args->ar_debug) {
256 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
257 "are mutually exclusive.\n");
260 args->ar_errors = GFS2_ERRORS_PANIC;
263 args->ar_nobarrier = 0;
266 args->ar_nobarrier = 1;
270 printk(KERN_WARNING "GFS2: invalid mount option: %s\n", o);
279 * gfs2_jindex_free - Clear all the journal index information
280 * @sdp: The GFS2 superblock
284 void gfs2_jindex_free(struct gfs2_sbd *sdp)
286 struct list_head list, *head;
287 struct gfs2_jdesc *jd;
288 struct gfs2_journal_extent *jext;
290 spin_lock(&sdp->sd_jindex_spin);
291 list_add(&list, &sdp->sd_jindex_list);
292 list_del_init(&sdp->sd_jindex_list);
293 sdp->sd_journals = 0;
294 spin_unlock(&sdp->sd_jindex_spin);
296 while (!list_empty(&list)) {
297 jd = list_entry(list.next, struct gfs2_jdesc, jd_list);
298 head = &jd->extent_list;
299 while (!list_empty(head)) {
300 jext = list_entry(head->next,
301 struct gfs2_journal_extent,
303 list_del(&jext->extent_list);
306 list_del(&jd->jd_list);
312 static struct gfs2_jdesc *jdesc_find_i(struct list_head *head, unsigned int jid)
314 struct gfs2_jdesc *jd;
317 list_for_each_entry(jd, head, jd_list) {
318 if (jd->jd_jid == jid) {
330 struct gfs2_jdesc *gfs2_jdesc_find(struct gfs2_sbd *sdp, unsigned int jid)
332 struct gfs2_jdesc *jd;
334 spin_lock(&sdp->sd_jindex_spin);
335 jd = jdesc_find_i(&sdp->sd_jindex_list, jid);
336 spin_unlock(&sdp->sd_jindex_spin);
341 int gfs2_jdesc_check(struct gfs2_jdesc *jd)
343 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
344 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
346 if (ip->i_disksize < (8 << 20) || ip->i_disksize > (1 << 30) ||
347 (ip->i_disksize & (sdp->sd_sb.sb_bsize - 1))) {
348 gfs2_consist_inode(ip);
351 jd->jd_blocks = ip->i_disksize >> sdp->sd_sb.sb_bsize_shift;
353 if (gfs2_write_alloc_required(ip, 0, ip->i_disksize)) {
354 gfs2_consist_inode(ip);
362 * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one
363 * @sdp: the filesystem
368 int gfs2_make_fs_rw(struct gfs2_sbd *sdp)
370 struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode);
371 struct gfs2_glock *j_gl = ip->i_gl;
372 struct gfs2_holder t_gh;
373 struct gfs2_log_header_host head;
376 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, 0, &t_gh);
380 j_gl->gl_ops->go_inval(j_gl, DIO_METADATA);
382 error = gfs2_find_jhead(sdp->sd_jdesc, &head);
386 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
392 /* Initialize some head of the log stuff */
393 sdp->sd_log_sequence = head.lh_sequence + 1;
394 gfs2_log_pointers_init(sdp, head.lh_blkno);
396 error = gfs2_quota_init(sdp);
400 set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
402 gfs2_glock_dq_uninit(&t_gh);
407 t_gh.gh_flags |= GL_NOCACHE;
408 gfs2_glock_dq_uninit(&t_gh);
413 void gfs2_statfs_change_in(struct gfs2_statfs_change_host *sc, const void *buf)
415 const struct gfs2_statfs_change *str = buf;
417 sc->sc_total = be64_to_cpu(str->sc_total);
418 sc->sc_free = be64_to_cpu(str->sc_free);
419 sc->sc_dinodes = be64_to_cpu(str->sc_dinodes);
422 static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host *sc, void *buf)
424 struct gfs2_statfs_change *str = buf;
426 str->sc_total = cpu_to_be64(sc->sc_total);
427 str->sc_free = cpu_to_be64(sc->sc_free);
428 str->sc_dinodes = cpu_to_be64(sc->sc_dinodes);
431 int gfs2_statfs_init(struct gfs2_sbd *sdp)
433 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
434 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
435 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
436 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
437 struct buffer_head *m_bh, *l_bh;
438 struct gfs2_holder gh;
441 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
446 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
450 if (sdp->sd_args.ar_spectator) {
451 spin_lock(&sdp->sd_statfs_spin);
452 gfs2_statfs_change_in(m_sc, m_bh->b_data +
453 sizeof(struct gfs2_dinode));
454 spin_unlock(&sdp->sd_statfs_spin);
456 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
460 spin_lock(&sdp->sd_statfs_spin);
461 gfs2_statfs_change_in(m_sc, m_bh->b_data +
462 sizeof(struct gfs2_dinode));
463 gfs2_statfs_change_in(l_sc, l_bh->b_data +
464 sizeof(struct gfs2_dinode));
465 spin_unlock(&sdp->sd_statfs_spin);
473 gfs2_glock_dq_uninit(&gh);
477 void gfs2_statfs_change(struct gfs2_sbd *sdp, s64 total, s64 free,
480 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
481 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
482 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
483 struct buffer_head *l_bh;
488 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
492 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
494 spin_lock(&sdp->sd_statfs_spin);
495 l_sc->sc_total += total;
496 l_sc->sc_free += free;
497 l_sc->sc_dinodes += dinodes;
498 gfs2_statfs_change_out(l_sc, l_bh->b_data + sizeof(struct gfs2_dinode));
499 if (sdp->sd_args.ar_statfs_percent) {
500 x = 100 * l_sc->sc_free;
501 y = m_sc->sc_free * sdp->sd_args.ar_statfs_percent;
502 if (x >= y || x <= -y)
505 spin_unlock(&sdp->sd_statfs_spin);
509 gfs2_wake_up_statfs(sdp);
512 void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
513 struct buffer_head *l_bh)
515 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
516 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
517 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
518 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
520 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
522 spin_lock(&sdp->sd_statfs_spin);
523 m_sc->sc_total += l_sc->sc_total;
524 m_sc->sc_free += l_sc->sc_free;
525 m_sc->sc_dinodes += l_sc->sc_dinodes;
526 memset(l_sc, 0, sizeof(struct gfs2_statfs_change));
527 memset(l_bh->b_data + sizeof(struct gfs2_dinode),
528 0, sizeof(struct gfs2_statfs_change));
529 spin_unlock(&sdp->sd_statfs_spin);
531 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
532 gfs2_statfs_change_out(m_sc, m_bh->b_data + sizeof(struct gfs2_dinode));
535 int gfs2_statfs_sync(struct super_block *sb, int type)
537 struct gfs2_sbd *sdp = sb->s_fs_info;
538 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
539 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
540 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
541 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
542 struct gfs2_holder gh;
543 struct buffer_head *m_bh, *l_bh;
546 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
551 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
555 spin_lock(&sdp->sd_statfs_spin);
556 gfs2_statfs_change_in(m_sc, m_bh->b_data +
557 sizeof(struct gfs2_dinode));
558 if (!l_sc->sc_total && !l_sc->sc_free && !l_sc->sc_dinodes) {
559 spin_unlock(&sdp->sd_statfs_spin);
562 spin_unlock(&sdp->sd_statfs_spin);
564 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
568 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
572 update_statfs(sdp, m_bh, l_bh);
573 sdp->sd_statfs_force_sync = 0;
582 gfs2_glock_dq_uninit(&gh);
587 struct list_head list;
588 struct gfs2_holder gh;
592 * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all
594 * @sdp: the file system
595 * @state: the state to put the transaction lock into
596 * @t_gh: the hold on the transaction lock
601 static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp,
602 struct gfs2_holder *t_gh)
604 struct gfs2_inode *ip;
605 struct gfs2_jdesc *jd;
608 struct gfs2_log_header_host lh;
611 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
612 lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL);
617 ip = GFS2_I(jd->jd_inode);
618 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &lfcc->gh);
623 list_add(&lfcc->list, &list);
626 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_DEFERRED,
629 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
630 error = gfs2_jdesc_check(jd);
633 error = gfs2_find_jhead(jd, &lh);
636 if (!(lh.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
643 gfs2_glock_dq_uninit(t_gh);
646 while (!list_empty(&list)) {
647 lfcc = list_entry(list.next, struct lfcc, list);
648 list_del(&lfcc->list);
649 gfs2_glock_dq_uninit(&lfcc->gh);
656 * gfs2_freeze_fs - freezes the file system
657 * @sdp: the file system
659 * This function flushes data and meta data for all machines by
660 * aquiring the transaction log exclusively. All journals are
661 * ensured to be in a clean state as well.
666 int gfs2_freeze_fs(struct gfs2_sbd *sdp)
670 mutex_lock(&sdp->sd_freeze_lock);
672 if (!sdp->sd_freeze_count++) {
673 error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh);
675 sdp->sd_freeze_count--;
678 mutex_unlock(&sdp->sd_freeze_lock);
684 * gfs2_unfreeze_fs - unfreezes the file system
685 * @sdp: the file system
687 * This function allows the file system to proceed by unlocking
688 * the exclusively held transaction lock. Other GFS2 nodes are
689 * now free to acquire the lock shared and go on with their lives.
693 void gfs2_unfreeze_fs(struct gfs2_sbd *sdp)
695 mutex_lock(&sdp->sd_freeze_lock);
697 if (sdp->sd_freeze_count && !--sdp->sd_freeze_count)
698 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
700 mutex_unlock(&sdp->sd_freeze_lock);
705 * gfs2_write_inode - Make sure the inode is stable on the disk
707 * @sync: synchronous write flag
712 static int gfs2_write_inode(struct inode *inode, struct writeback_control *wbc)
714 struct gfs2_inode *ip = GFS2_I(inode);
715 struct gfs2_sbd *sdp = GFS2_SB(inode);
716 struct gfs2_holder gh;
717 struct buffer_head *bh;
718 struct timespec atime;
719 struct gfs2_dinode *di;
722 /* Check this is a "normal" inode, etc */
723 if (current->flags & PF_MEMALLOC)
725 ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
728 ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
731 ret = gfs2_meta_inode_buffer(ip, &bh);
733 di = (struct gfs2_dinode *)bh->b_data;
734 atime.tv_sec = be64_to_cpu(di->di_atime);
735 atime.tv_nsec = be32_to_cpu(di->di_atime_nsec);
736 if (timespec_compare(&inode->i_atime, &atime) > 0) {
737 gfs2_trans_add_bh(ip->i_gl, bh, 1);
738 gfs2_dinode_out(ip, bh->b_data);
744 gfs2_glock_dq_uninit(&gh);
746 if (wbc->sync_mode == WB_SYNC_ALL)
747 gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
752 * gfs2_make_fs_ro - Turn a Read-Write FS into a Read-Only one
753 * @sdp: the filesystem
758 static int gfs2_make_fs_ro(struct gfs2_sbd *sdp)
760 struct gfs2_holder t_gh;
763 flush_workqueue(gfs2_delete_workqueue);
764 gfs2_quota_sync(sdp->sd_vfs, 0, 1);
765 gfs2_statfs_sync(sdp->sd_vfs, 0);
767 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, GL_NOCACHE,
769 if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
772 gfs2_meta_syncfs(sdp);
773 gfs2_log_shutdown(sdp);
775 clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
778 gfs2_glock_dq_uninit(&t_gh);
780 gfs2_quota_cleanup(sdp);
785 static int gfs2_umount_recovery_wait(void *word)
792 * gfs2_put_super - Unmount the filesystem
793 * @sb: The VFS superblock
797 static void gfs2_put_super(struct super_block *sb)
799 struct gfs2_sbd *sdp = sb->s_fs_info;
801 struct gfs2_jdesc *jd;
803 /* Unfreeze the filesystem, if we need to */
805 mutex_lock(&sdp->sd_freeze_lock);
806 if (sdp->sd_freeze_count)
807 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
808 mutex_unlock(&sdp->sd_freeze_lock);
810 /* No more recovery requests */
811 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
814 /* Wait on outstanding recovery */
816 spin_lock(&sdp->sd_jindex_spin);
817 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
818 if (!test_bit(JDF_RECOVERY, &jd->jd_flags))
820 spin_unlock(&sdp->sd_jindex_spin);
821 wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
822 gfs2_umount_recovery_wait, TASK_UNINTERRUPTIBLE);
825 spin_unlock(&sdp->sd_jindex_spin);
827 kthread_stop(sdp->sd_quotad_process);
828 kthread_stop(sdp->sd_logd_process);
830 if (!(sb->s_flags & MS_RDONLY)) {
831 error = gfs2_make_fs_ro(sdp);
835 /* At this point, we're through modifying the disk */
839 iput(sdp->sd_jindex);
840 iput(sdp->sd_statfs_inode);
841 iput(sdp->sd_rindex);
842 iput(sdp->sd_quota_inode);
844 gfs2_glock_put(sdp->sd_rename_gl);
845 gfs2_glock_put(sdp->sd_trans_gl);
847 if (!sdp->sd_args.ar_spectator) {
848 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
849 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
850 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
851 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
852 iput(sdp->sd_sc_inode);
853 iput(sdp->sd_qc_inode);
856 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
857 gfs2_clear_rgrpd(sdp);
858 gfs2_jindex_free(sdp);
859 /* Take apart glock structures and buffer lists */
860 invalidate_inodes(sdp->sd_vfs);
861 gfs2_gl_hash_clear(sdp);
862 /* Unmount the locking protocol */
863 gfs2_lm_unmount(sdp);
865 /* At this point, we're through participating in the lockspace */
866 gfs2_sys_fs_del(sdp);
870 * gfs2_sync_fs - sync the filesystem
871 * @sb: the superblock
873 * Flushes the log to disk.
876 static int gfs2_sync_fs(struct super_block *sb, int wait)
878 if (wait && sb->s_fs_info)
879 gfs2_log_flush(sb->s_fs_info, NULL);
884 * gfs2_freeze - prevent further writes to the filesystem
885 * @sb: the VFS structure for the filesystem
889 static int gfs2_freeze(struct super_block *sb)
891 struct gfs2_sbd *sdp = sb->s_fs_info;
894 if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
898 error = gfs2_freeze_fs(sdp);
904 fs_err(sdp, "waiting for recovery before freeze\n");
908 fs_err(sdp, "error freezing FS: %d\n", error);
912 fs_err(sdp, "retrying...\n");
919 * gfs2_unfreeze - reallow writes to the filesystem
920 * @sb: the VFS structure for the filesystem
924 static int gfs2_unfreeze(struct super_block *sb)
926 gfs2_unfreeze_fs(sb->s_fs_info);
931 * statfs_fill - fill in the sg for a given RG
933 * @sc: the sc structure
935 * Returns: 0 on success, -ESTALE if the LVB is invalid
938 static int statfs_slow_fill(struct gfs2_rgrpd *rgd,
939 struct gfs2_statfs_change_host *sc)
941 gfs2_rgrp_verify(rgd);
942 sc->sc_total += rgd->rd_data;
943 sc->sc_free += rgd->rd_free;
944 sc->sc_dinodes += rgd->rd_dinodes;
949 * gfs2_statfs_slow - Stat a filesystem using asynchronous locking
950 * @sdp: the filesystem
951 * @sc: the sc info that will be returned
953 * Any error (other than a signal) will cause this routine to fall back
954 * to the synchronous version.
956 * FIXME: This really shouldn't busy wait like this.
961 static int gfs2_statfs_slow(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
963 struct gfs2_holder ri_gh;
964 struct gfs2_rgrpd *rgd_next;
965 struct gfs2_holder *gha, *gh;
966 unsigned int slots = 64;
971 memset(sc, 0, sizeof(struct gfs2_statfs_change_host));
972 gha = kcalloc(slots, sizeof(struct gfs2_holder), GFP_KERNEL);
976 error = gfs2_rindex_hold(sdp, &ri_gh);
980 rgd_next = gfs2_rgrpd_get_first(sdp);
985 for (x = 0; x < slots; x++) {
988 if (gh->gh_gl && gfs2_glock_poll(gh)) {
989 err = gfs2_glock_wait(gh);
991 gfs2_holder_uninit(gh);
995 error = statfs_slow_fill(
996 gh->gh_gl->gl_object, sc);
997 gfs2_glock_dq_uninit(gh);
1003 else if (rgd_next && !error) {
1004 error = gfs2_glock_nq_init(rgd_next->rd_gl,
1008 rgd_next = gfs2_rgrpd_get_next(rgd_next);
1012 if (signal_pending(current))
1013 error = -ERESTARTSYS;
1022 gfs2_glock_dq_uninit(&ri_gh);
1030 * gfs2_statfs_i - Do a statfs
1031 * @sdp: the filesystem
1032 * @sg: the sg structure
1037 static int gfs2_statfs_i(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
1039 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
1040 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
1042 spin_lock(&sdp->sd_statfs_spin);
1045 sc->sc_total += l_sc->sc_total;
1046 sc->sc_free += l_sc->sc_free;
1047 sc->sc_dinodes += l_sc->sc_dinodes;
1049 spin_unlock(&sdp->sd_statfs_spin);
1051 if (sc->sc_free < 0)
1053 if (sc->sc_free > sc->sc_total)
1054 sc->sc_free = sc->sc_total;
1055 if (sc->sc_dinodes < 0)
1062 * gfs2_statfs - Gather and return stats about the filesystem
1063 * @sb: The superblock
1064 * @statfsbuf: The buffer
1066 * Returns: 0 on success or error code
1069 static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1071 struct super_block *sb = dentry->d_inode->i_sb;
1072 struct gfs2_sbd *sdp = sb->s_fs_info;
1073 struct gfs2_statfs_change_host sc;
1076 if (gfs2_tune_get(sdp, gt_statfs_slow))
1077 error = gfs2_statfs_slow(sdp, &sc);
1079 error = gfs2_statfs_i(sdp, &sc);
1084 buf->f_type = GFS2_MAGIC;
1085 buf->f_bsize = sdp->sd_sb.sb_bsize;
1086 buf->f_blocks = sc.sc_total;
1087 buf->f_bfree = sc.sc_free;
1088 buf->f_bavail = sc.sc_free;
1089 buf->f_files = sc.sc_dinodes + sc.sc_free;
1090 buf->f_ffree = sc.sc_free;
1091 buf->f_namelen = GFS2_FNAMESIZE;
1097 * gfs2_remount_fs - called when the FS is remounted
1098 * @sb: the filesystem
1099 * @flags: the remount flags
1100 * @data: extra data passed in (not used right now)
1105 static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
1107 struct gfs2_sbd *sdp = sb->s_fs_info;
1108 struct gfs2_args args = sdp->sd_args; /* Default to current settings */
1109 struct gfs2_tune *gt = &sdp->sd_tune;
1112 spin_lock(>->gt_spin);
1113 args.ar_commit = gt->gt_logd_secs;
1114 args.ar_quota_quantum = gt->gt_quota_quantum;
1115 if (gt->gt_statfs_slow)
1116 args.ar_statfs_quantum = 0;
1118 args.ar_statfs_quantum = gt->gt_statfs_quantum;
1119 spin_unlock(>->gt_spin);
1120 error = gfs2_mount_args(&args, data);
1124 /* Not allowed to change locking details */
1125 if (strcmp(args.ar_lockproto, sdp->sd_args.ar_lockproto) ||
1126 strcmp(args.ar_locktable, sdp->sd_args.ar_locktable) ||
1127 strcmp(args.ar_hostdata, sdp->sd_args.ar_hostdata))
1130 /* Some flags must not be changed */
1131 if (args_neq(&args, &sdp->sd_args, spectator) ||
1132 args_neq(&args, &sdp->sd_args, ignore_local_fs) ||
1133 args_neq(&args, &sdp->sd_args, localflocks) ||
1134 args_neq(&args, &sdp->sd_args, localcaching) ||
1135 args_neq(&args, &sdp->sd_args, meta))
1138 if (sdp->sd_args.ar_spectator)
1139 *flags |= MS_RDONLY;
1141 if ((sb->s_flags ^ *flags) & MS_RDONLY) {
1142 if (*flags & MS_RDONLY)
1143 error = gfs2_make_fs_ro(sdp);
1145 error = gfs2_make_fs_rw(sdp);
1150 sdp->sd_args = args;
1151 if (sdp->sd_args.ar_posix_acl)
1152 sb->s_flags |= MS_POSIXACL;
1154 sb->s_flags &= ~MS_POSIXACL;
1155 if (sdp->sd_args.ar_nobarrier)
1156 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1158 clear_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1159 spin_lock(>->gt_spin);
1160 gt->gt_logd_secs = args.ar_commit;
1161 gt->gt_quota_quantum = args.ar_quota_quantum;
1162 if (args.ar_statfs_quantum) {
1163 gt->gt_statfs_slow = 0;
1164 gt->gt_statfs_quantum = args.ar_statfs_quantum;
1167 gt->gt_statfs_slow = 1;
1168 gt->gt_statfs_quantum = 30;
1170 spin_unlock(>->gt_spin);
1172 gfs2_online_uevent(sdp);
1177 * gfs2_drop_inode - Drop an inode (test for remote unlink)
1178 * @inode: The inode to drop
1180 * If we've received a callback on an iopen lock then its because a
1181 * remote node tried to deallocate the inode but failed due to this node
1182 * still having the inode open. Here we mark the link count zero
1183 * since we know that it must have reached zero if the GLF_DEMOTE flag
1184 * is set on the iopen glock. If we didn't do a disk read since the
1185 * remote node removed the final link then we might otherwise miss
1186 * this event. This check ensures that this node will deallocate the
1187 * inode's blocks, or alternatively pass the baton on to another
1188 * node for later deallocation.
1191 static void gfs2_drop_inode(struct inode *inode)
1193 struct gfs2_inode *ip = GFS2_I(inode);
1195 if (inode->i_nlink) {
1196 struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
1197 if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags))
1200 generic_drop_inode(inode);
1204 * gfs2_clear_inode - Deallocate an inode when VFS is done with it
1205 * @inode: The VFS inode
1209 static void gfs2_clear_inode(struct inode *inode)
1211 struct gfs2_inode *ip = GFS2_I(inode);
1213 ip->i_gl->gl_object = NULL;
1214 gfs2_glock_put(ip->i_gl);
1216 if (ip->i_iopen_gh.gh_gl) {
1217 ip->i_iopen_gh.gh_gl->gl_object = NULL;
1218 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1222 static int is_ancestor(const struct dentry *d1, const struct dentry *d2)
1228 } while (!IS_ROOT(d1));
1233 * gfs2_show_options - Show mount options for /proc/mounts
1234 * @s: seq_file structure
1237 * Returns: 0 on success or error code
1240 static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1242 struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info;
1243 struct gfs2_args *args = &sdp->sd_args;
1246 if (is_ancestor(mnt->mnt_root, sdp->sd_master_dir))
1247 seq_printf(s, ",meta");
1248 if (args->ar_lockproto[0])
1249 seq_printf(s, ",lockproto=%s", args->ar_lockproto);
1250 if (args->ar_locktable[0])
1251 seq_printf(s, ",locktable=%s", args->ar_locktable);
1252 if (args->ar_hostdata[0])
1253 seq_printf(s, ",hostdata=%s", args->ar_hostdata);
1254 if (args->ar_spectator)
1255 seq_printf(s, ",spectator");
1256 if (args->ar_ignore_local_fs)
1257 seq_printf(s, ",ignore_local_fs");
1258 if (args->ar_localflocks)
1259 seq_printf(s, ",localflocks");
1260 if (args->ar_localcaching)
1261 seq_printf(s, ",localcaching");
1263 seq_printf(s, ",debug");
1264 if (args->ar_upgrade)
1265 seq_printf(s, ",upgrade");
1266 if (args->ar_posix_acl)
1267 seq_printf(s, ",acl");
1268 if (args->ar_quota != GFS2_QUOTA_DEFAULT) {
1270 switch (args->ar_quota) {
1271 case GFS2_QUOTA_OFF:
1274 case GFS2_QUOTA_ACCOUNT:
1284 seq_printf(s, ",quota=%s", state);
1286 if (args->ar_suiddir)
1287 seq_printf(s, ",suiddir");
1288 if (args->ar_data != GFS2_DATA_DEFAULT) {
1290 switch (args->ar_data) {
1291 case GFS2_DATA_WRITEBACK:
1292 state = "writeback";
1294 case GFS2_DATA_ORDERED:
1301 seq_printf(s, ",data=%s", state);
1303 if (args->ar_discard)
1304 seq_printf(s, ",discard");
1305 val = sdp->sd_tune.gt_logd_secs;
1307 seq_printf(s, ",commit=%d", val);
1308 val = sdp->sd_tune.gt_statfs_quantum;
1310 seq_printf(s, ",statfs_quantum=%d", val);
1311 val = sdp->sd_tune.gt_quota_quantum;
1313 seq_printf(s, ",quota_quantum=%d", val);
1314 if (args->ar_statfs_percent)
1315 seq_printf(s, ",statfs_percent=%d", args->ar_statfs_percent);
1316 if (args->ar_errors != GFS2_ERRORS_DEFAULT) {
1319 switch (args->ar_errors) {
1320 case GFS2_ERRORS_WITHDRAW:
1323 case GFS2_ERRORS_PANIC:
1330 seq_printf(s, ",errors=%s", state);
1332 if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags))
1333 seq_printf(s, ",nobarrier");
1334 if (test_bit(SDF_DEMOTE, &sdp->sd_flags))
1335 seq_printf(s, ",demote_interface_used");
1340 * We have to (at the moment) hold the inodes main lock to cover
1341 * the gap between unlocking the shared lock on the iopen lock and
1342 * taking the exclusive lock. I'd rather do a shared -> exclusive
1343 * conversion on the iopen lock, but we can change that later. This
1344 * is safe, just less efficient.
1347 static void gfs2_delete_inode(struct inode *inode)
1349 struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
1350 struct gfs2_inode *ip = GFS2_I(inode);
1351 struct gfs2_holder gh;
1354 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
1355 if (unlikely(error)) {
1356 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1360 error = gfs2_check_blk_type(sdp, ip->i_no_addr, GFS2_BLKST_UNLINKED);
1364 gfs2_glock_dq_wait(&ip->i_iopen_gh);
1365 gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh);
1366 error = gfs2_glock_nq(&ip->i_iopen_gh);
1370 if (S_ISDIR(inode->i_mode) &&
1371 (ip->i_diskflags & GFS2_DIF_EXHASH)) {
1372 error = gfs2_dir_exhash_dealloc(ip);
1378 error = gfs2_ea_dealloc(ip);
1383 if (!gfs2_is_stuffed(ip)) {
1384 error = gfs2_file_dealloc(ip);
1389 error = gfs2_dinode_dealloc(ip);
1394 error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks);
1397 /* Needs to be done before glock release & also in a transaction */
1398 truncate_inode_pages(&inode->i_data, 0);
1399 gfs2_trans_end(sdp);
1402 if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags))
1403 gfs2_glock_dq(&ip->i_iopen_gh);
1404 gfs2_holder_uninit(&ip->i_iopen_gh);
1405 gfs2_glock_dq_uninit(&gh);
1406 if (error && error != GLR_TRYFAILED && error != -EROFS)
1407 fs_warn(sdp, "gfs2_delete_inode: %d\n", error);
1409 truncate_inode_pages(&inode->i_data, 0);
1413 static struct inode *gfs2_alloc_inode(struct super_block *sb)
1415 struct gfs2_inode *ip;
1417 ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL);
1422 return &ip->i_inode;
1425 static void gfs2_destroy_inode(struct inode *inode)
1427 kmem_cache_free(gfs2_inode_cachep, inode);
1430 const struct super_operations gfs2_super_ops = {
1431 .alloc_inode = gfs2_alloc_inode,
1432 .destroy_inode = gfs2_destroy_inode,
1433 .write_inode = gfs2_write_inode,
1434 .delete_inode = gfs2_delete_inode,
1435 .put_super = gfs2_put_super,
1436 .sync_fs = gfs2_sync_fs,
1437 .freeze_fs = gfs2_freeze,
1438 .unfreeze_fs = gfs2_unfreeze,
1439 .statfs = gfs2_statfs,
1440 .remount_fs = gfs2_remount_fs,
1441 .clear_inode = gfs2_clear_inode,
1442 .drop_inode = gfs2_drop_inode,
1443 .show_options = gfs2_show_options,