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>
44 #define args_neq(a1, a2, x) ((a1)->ar_##x != (a2)->ar_##x)
82 static const match_table_t tokens = {
83 {Opt_lockproto, "lockproto=%s"},
84 {Opt_locktable, "locktable=%s"},
85 {Opt_hostdata, "hostdata=%s"},
86 {Opt_spectator, "spectator"},
87 {Opt_ignore_local_fs, "ignore_local_fs"},
88 {Opt_localflocks, "localflocks"},
89 {Opt_localcaching, "localcaching"},
91 {Opt_nodebug, "nodebug"},
92 {Opt_upgrade, "upgrade"},
95 {Opt_quota_off, "quota=off"},
96 {Opt_quota_account, "quota=account"},
97 {Opt_quota_on, "quota=on"},
99 {Opt_noquota, "noquota"},
100 {Opt_suiddir, "suiddir"},
101 {Opt_nosuiddir, "nosuiddir"},
102 {Opt_data_writeback, "data=writeback"},
103 {Opt_data_ordered, "data=ordered"},
105 {Opt_discard, "discard"},
106 {Opt_nodiscard, "nodiscard"},
107 {Opt_commit, "commit=%d"},
108 {Opt_err_withdraw, "errors=withdraw"},
109 {Opt_err_panic, "errors=panic"},
110 {Opt_statfs_quantum, "statfs_quantum=%d"},
111 {Opt_statfs_percent, "statfs_percent=%d"},
112 {Opt_quota_quantum, "quota_quantum=%d"},
113 {Opt_barrier, "barrier"},
114 {Opt_nobarrier, "nobarrier"},
119 * gfs2_mount_args - Parse mount options
120 * @args: The structure into which the parsed options will be written
121 * @options: The options to parse
126 int gfs2_mount_args(struct gfs2_args *args, char *options)
130 substring_t tmp[MAX_OPT_ARGS];
133 /* Split the options into tokens with the "," character and
137 o = strsep(&options, ",");
143 token = match_token(o, tokens, tmp);
146 match_strlcpy(args->ar_lockproto, &tmp[0],
150 match_strlcpy(args->ar_locktable, &tmp[0],
154 match_strlcpy(args->ar_hostdata, &tmp[0],
158 args->ar_spectator = 1;
160 case Opt_ignore_local_fs:
161 args->ar_ignore_local_fs = 1;
163 case Opt_localflocks:
164 args->ar_localflocks = 1;
166 case Opt_localcaching:
167 args->ar_localcaching = 1;
170 if (args->ar_errors == GFS2_ERRORS_PANIC) {
171 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
172 "are mutually exclusive.\n");
181 args->ar_upgrade = 1;
184 args->ar_posix_acl = 1;
187 args->ar_posix_acl = 0;
191 args->ar_quota = GFS2_QUOTA_OFF;
193 case Opt_quota_account:
194 args->ar_quota = GFS2_QUOTA_ACCOUNT;
198 args->ar_quota = GFS2_QUOTA_ON;
201 args->ar_suiddir = 1;
204 args->ar_suiddir = 0;
206 case Opt_data_writeback:
207 args->ar_data = GFS2_DATA_WRITEBACK;
209 case Opt_data_ordered:
210 args->ar_data = GFS2_DATA_ORDERED;
216 args->ar_discard = 1;
219 args->ar_discard = 0;
222 rv = match_int(&tmp[0], &args->ar_commit);
223 if (rv || args->ar_commit <= 0) {
224 printk(KERN_WARNING "GFS2: commit mount option requires a positive numeric argument\n");
225 return rv ? rv : -EINVAL;
228 case Opt_statfs_quantum:
229 rv = match_int(&tmp[0], &args->ar_statfs_quantum);
230 if (rv || args->ar_statfs_quantum < 0) {
231 printk(KERN_WARNING "GFS2: statfs_quantum mount option requires a non-negative numeric argument\n");
232 return rv ? rv : -EINVAL;
235 case Opt_quota_quantum:
236 rv = match_int(&tmp[0], &args->ar_quota_quantum);
237 if (rv || args->ar_quota_quantum <= 0) {
238 printk(KERN_WARNING "GFS2: quota_quantum mount option requires a positive numeric argument\n");
239 return rv ? rv : -EINVAL;
242 case Opt_statfs_percent:
243 rv = match_int(&tmp[0], &args->ar_statfs_percent);
244 if (rv || args->ar_statfs_percent < 0 ||
245 args->ar_statfs_percent > 100) {
246 printk(KERN_WARNING "statfs_percent mount option requires a numeric argument between 0 and 100\n");
247 return rv ? rv : -EINVAL;
250 case Opt_err_withdraw:
251 args->ar_errors = GFS2_ERRORS_WITHDRAW;
254 if (args->ar_debug) {
255 printk(KERN_WARNING "GFS2: -o debug and -o errors=panic "
256 "are mutually exclusive.\n");
259 args->ar_errors = GFS2_ERRORS_PANIC;
262 args->ar_nobarrier = 0;
265 args->ar_nobarrier = 1;
269 printk(KERN_WARNING "GFS2: invalid mount option: %s\n", o);
278 * gfs2_jindex_free - Clear all the journal index information
279 * @sdp: The GFS2 superblock
283 void gfs2_jindex_free(struct gfs2_sbd *sdp)
285 struct list_head list, *head;
286 struct gfs2_jdesc *jd;
287 struct gfs2_journal_extent *jext;
289 spin_lock(&sdp->sd_jindex_spin);
290 list_add(&list, &sdp->sd_jindex_list);
291 list_del_init(&sdp->sd_jindex_list);
292 sdp->sd_journals = 0;
293 spin_unlock(&sdp->sd_jindex_spin);
295 while (!list_empty(&list)) {
296 jd = list_entry(list.next, struct gfs2_jdesc, jd_list);
297 head = &jd->extent_list;
298 while (!list_empty(head)) {
299 jext = list_entry(head->next,
300 struct gfs2_journal_extent,
302 list_del(&jext->extent_list);
305 list_del(&jd->jd_list);
311 static struct gfs2_jdesc *jdesc_find_i(struct list_head *head, unsigned int jid)
313 struct gfs2_jdesc *jd;
316 list_for_each_entry(jd, head, jd_list) {
317 if (jd->jd_jid == jid) {
329 struct gfs2_jdesc *gfs2_jdesc_find(struct gfs2_sbd *sdp, unsigned int jid)
331 struct gfs2_jdesc *jd;
333 spin_lock(&sdp->sd_jindex_spin);
334 jd = jdesc_find_i(&sdp->sd_jindex_list, jid);
335 spin_unlock(&sdp->sd_jindex_spin);
340 int gfs2_jdesc_check(struct gfs2_jdesc *jd)
342 struct gfs2_inode *ip = GFS2_I(jd->jd_inode);
343 struct gfs2_sbd *sdp = GFS2_SB(jd->jd_inode);
347 if (ip->i_disksize < (8 << 20) || ip->i_disksize > (1 << 30) ||
348 (ip->i_disksize & (sdp->sd_sb.sb_bsize - 1))) {
349 gfs2_consist_inode(ip);
352 jd->jd_blocks = ip->i_disksize >> sdp->sd_sb.sb_bsize_shift;
354 error = gfs2_write_alloc_required(ip, 0, ip->i_disksize, &ar);
356 gfs2_consist_inode(ip);
364 * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one
365 * @sdp: the filesystem
370 int gfs2_make_fs_rw(struct gfs2_sbd *sdp)
372 struct gfs2_inode *ip = GFS2_I(sdp->sd_jdesc->jd_inode);
373 struct gfs2_glock *j_gl = ip->i_gl;
374 struct gfs2_holder t_gh;
375 struct gfs2_log_header_host head;
378 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, 0, &t_gh);
382 j_gl->gl_ops->go_inval(j_gl, DIO_METADATA);
384 error = gfs2_find_jhead(sdp->sd_jdesc, &head);
388 if (!(head.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
394 /* Initialize some head of the log stuff */
395 sdp->sd_log_sequence = head.lh_sequence + 1;
396 gfs2_log_pointers_init(sdp, head.lh_blkno);
398 error = gfs2_quota_init(sdp);
402 set_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
404 gfs2_glock_dq_uninit(&t_gh);
409 t_gh.gh_flags |= GL_NOCACHE;
410 gfs2_glock_dq_uninit(&t_gh);
415 void gfs2_statfs_change_in(struct gfs2_statfs_change_host *sc, const void *buf)
417 const struct gfs2_statfs_change *str = buf;
419 sc->sc_total = be64_to_cpu(str->sc_total);
420 sc->sc_free = be64_to_cpu(str->sc_free);
421 sc->sc_dinodes = be64_to_cpu(str->sc_dinodes);
424 static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host *sc, void *buf)
426 struct gfs2_statfs_change *str = buf;
428 str->sc_total = cpu_to_be64(sc->sc_total);
429 str->sc_free = cpu_to_be64(sc->sc_free);
430 str->sc_dinodes = cpu_to_be64(sc->sc_dinodes);
433 int gfs2_statfs_init(struct gfs2_sbd *sdp)
435 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
436 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
437 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
438 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
439 struct buffer_head *m_bh, *l_bh;
440 struct gfs2_holder gh;
443 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
448 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
452 if (sdp->sd_args.ar_spectator) {
453 spin_lock(&sdp->sd_statfs_spin);
454 gfs2_statfs_change_in(m_sc, m_bh->b_data +
455 sizeof(struct gfs2_dinode));
456 spin_unlock(&sdp->sd_statfs_spin);
458 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
462 spin_lock(&sdp->sd_statfs_spin);
463 gfs2_statfs_change_in(m_sc, m_bh->b_data +
464 sizeof(struct gfs2_dinode));
465 gfs2_statfs_change_in(l_sc, l_bh->b_data +
466 sizeof(struct gfs2_dinode));
467 spin_unlock(&sdp->sd_statfs_spin);
475 gfs2_glock_dq_uninit(&gh);
479 void gfs2_statfs_change(struct gfs2_sbd *sdp, s64 total, s64 free,
482 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
483 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
484 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
485 struct buffer_head *l_bh;
490 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
494 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
496 spin_lock(&sdp->sd_statfs_spin);
497 l_sc->sc_total += total;
498 l_sc->sc_free += free;
499 l_sc->sc_dinodes += dinodes;
500 gfs2_statfs_change_out(l_sc, l_bh->b_data + sizeof(struct gfs2_dinode));
501 if (sdp->sd_args.ar_statfs_percent) {
502 x = 100 * l_sc->sc_free;
503 y = m_sc->sc_free * sdp->sd_args.ar_statfs_percent;
504 if (x >= y || x <= -y)
507 spin_unlock(&sdp->sd_statfs_spin);
511 gfs2_wake_up_statfs(sdp);
514 void update_statfs(struct gfs2_sbd *sdp, struct buffer_head *m_bh,
515 struct buffer_head *l_bh)
517 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
518 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
519 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
520 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
522 gfs2_trans_add_bh(l_ip->i_gl, l_bh, 1);
524 spin_lock(&sdp->sd_statfs_spin);
525 m_sc->sc_total += l_sc->sc_total;
526 m_sc->sc_free += l_sc->sc_free;
527 m_sc->sc_dinodes += l_sc->sc_dinodes;
528 memset(l_sc, 0, sizeof(struct gfs2_statfs_change));
529 memset(l_bh->b_data + sizeof(struct gfs2_dinode),
530 0, sizeof(struct gfs2_statfs_change));
531 spin_unlock(&sdp->sd_statfs_spin);
533 gfs2_trans_add_bh(m_ip->i_gl, m_bh, 1);
534 gfs2_statfs_change_out(m_sc, m_bh->b_data + sizeof(struct gfs2_dinode));
537 int gfs2_statfs_sync(struct super_block *sb, int type)
539 struct gfs2_sbd *sdp = sb->s_fs_info;
540 struct gfs2_inode *m_ip = GFS2_I(sdp->sd_statfs_inode);
541 struct gfs2_inode *l_ip = GFS2_I(sdp->sd_sc_inode);
542 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
543 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
544 struct gfs2_holder gh;
545 struct buffer_head *m_bh, *l_bh;
548 error = gfs2_glock_nq_init(m_ip->i_gl, LM_ST_EXCLUSIVE, GL_NOCACHE,
553 error = gfs2_meta_inode_buffer(m_ip, &m_bh);
557 spin_lock(&sdp->sd_statfs_spin);
558 gfs2_statfs_change_in(m_sc, m_bh->b_data +
559 sizeof(struct gfs2_dinode));
560 if (!l_sc->sc_total && !l_sc->sc_free && !l_sc->sc_dinodes) {
561 spin_unlock(&sdp->sd_statfs_spin);
564 spin_unlock(&sdp->sd_statfs_spin);
566 error = gfs2_meta_inode_buffer(l_ip, &l_bh);
570 error = gfs2_trans_begin(sdp, 2 * RES_DINODE, 0);
574 update_statfs(sdp, m_bh, l_bh);
575 sdp->sd_statfs_force_sync = 0;
584 gfs2_glock_dq_uninit(&gh);
589 struct list_head list;
590 struct gfs2_holder gh;
594 * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all
596 * @sdp: the file system
597 * @state: the state to put the transaction lock into
598 * @t_gh: the hold on the transaction lock
603 static int gfs2_lock_fs_check_clean(struct gfs2_sbd *sdp,
604 struct gfs2_holder *t_gh)
606 struct gfs2_inode *ip;
607 struct gfs2_jdesc *jd;
610 struct gfs2_log_header_host lh;
613 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
614 lfcc = kmalloc(sizeof(struct lfcc), GFP_KERNEL);
619 ip = GFS2_I(jd->jd_inode);
620 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_SHARED, 0, &lfcc->gh);
625 list_add(&lfcc->list, &list);
628 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_DEFERRED,
631 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
632 error = gfs2_jdesc_check(jd);
635 error = gfs2_find_jhead(jd, &lh);
638 if (!(lh.lh_flags & GFS2_LOG_HEAD_UNMOUNT)) {
645 gfs2_glock_dq_uninit(t_gh);
648 while (!list_empty(&list)) {
649 lfcc = list_entry(list.next, struct lfcc, list);
650 list_del(&lfcc->list);
651 gfs2_glock_dq_uninit(&lfcc->gh);
658 * gfs2_freeze_fs - freezes the file system
659 * @sdp: the file system
661 * This function flushes data and meta data for all machines by
662 * aquiring the transaction log exclusively. All journals are
663 * ensured to be in a clean state as well.
668 int gfs2_freeze_fs(struct gfs2_sbd *sdp)
672 mutex_lock(&sdp->sd_freeze_lock);
674 if (!sdp->sd_freeze_count++) {
675 error = gfs2_lock_fs_check_clean(sdp, &sdp->sd_freeze_gh);
677 sdp->sd_freeze_count--;
680 mutex_unlock(&sdp->sd_freeze_lock);
686 * gfs2_unfreeze_fs - unfreezes the file system
687 * @sdp: the file system
689 * This function allows the file system to proceed by unlocking
690 * the exclusively held transaction lock. Other GFS2 nodes are
691 * now free to acquire the lock shared and go on with their lives.
695 void gfs2_unfreeze_fs(struct gfs2_sbd *sdp)
697 mutex_lock(&sdp->sd_freeze_lock);
699 if (sdp->sd_freeze_count && !--sdp->sd_freeze_count)
700 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
702 mutex_unlock(&sdp->sd_freeze_lock);
707 * gfs2_write_inode - Make sure the inode is stable on the disk
709 * @sync: synchronous write flag
714 static int gfs2_write_inode(struct inode *inode, int sync)
716 struct gfs2_inode *ip = GFS2_I(inode);
717 struct gfs2_sbd *sdp = GFS2_SB(inode);
718 struct gfs2_holder gh;
719 struct buffer_head *bh;
720 struct timespec atime;
721 struct gfs2_dinode *di;
724 /* Check this is a "normal" inode, etc */
725 if (!test_bit(GIF_USER, &ip->i_flags) ||
726 (current->flags & PF_MEMALLOC))
728 ret = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
731 ret = gfs2_trans_begin(sdp, RES_DINODE, 0);
734 ret = gfs2_meta_inode_buffer(ip, &bh);
736 di = (struct gfs2_dinode *)bh->b_data;
737 atime.tv_sec = be64_to_cpu(di->di_atime);
738 atime.tv_nsec = be32_to_cpu(di->di_atime_nsec);
739 if (timespec_compare(&inode->i_atime, &atime) > 0) {
740 gfs2_trans_add_bh(ip->i_gl, bh, 1);
741 gfs2_dinode_out(ip, bh->b_data);
747 gfs2_glock_dq_uninit(&gh);
750 gfs2_log_flush(GFS2_SB(inode), ip->i_gl);
755 * gfs2_make_fs_ro - Turn a Read-Write FS into a Read-Only one
756 * @sdp: the filesystem
761 static int gfs2_make_fs_ro(struct gfs2_sbd *sdp)
763 struct gfs2_holder t_gh;
766 flush_workqueue(gfs2_delete_workqueue);
767 gfs2_quota_sync(sdp->sd_vfs, 0);
768 gfs2_statfs_sync(sdp->sd_vfs, 0);
770 error = gfs2_glock_nq_init(sdp->sd_trans_gl, LM_ST_SHARED, GL_NOCACHE,
772 if (error && !test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
775 gfs2_meta_syncfs(sdp);
776 gfs2_log_shutdown(sdp);
778 clear_bit(SDF_JOURNAL_LIVE, &sdp->sd_flags);
781 gfs2_glock_dq_uninit(&t_gh);
783 gfs2_quota_cleanup(sdp);
788 static int gfs2_umount_recovery_wait(void *word)
795 * gfs2_put_super - Unmount the filesystem
796 * @sb: The VFS superblock
800 static void gfs2_put_super(struct super_block *sb)
802 struct gfs2_sbd *sdp = sb->s_fs_info;
804 struct gfs2_jdesc *jd;
806 /* Unfreeze the filesystem, if we need to */
808 mutex_lock(&sdp->sd_freeze_lock);
809 if (sdp->sd_freeze_count)
810 gfs2_glock_dq_uninit(&sdp->sd_freeze_gh);
811 mutex_unlock(&sdp->sd_freeze_lock);
813 /* No more recovery requests */
814 set_bit(SDF_NORECOVERY, &sdp->sd_flags);
817 /* Wait on outstanding recovery */
819 spin_lock(&sdp->sd_jindex_spin);
820 list_for_each_entry(jd, &sdp->sd_jindex_list, jd_list) {
821 if (!test_bit(JDF_RECOVERY, &jd->jd_flags))
823 spin_unlock(&sdp->sd_jindex_spin);
824 wait_on_bit(&jd->jd_flags, JDF_RECOVERY,
825 gfs2_umount_recovery_wait, TASK_UNINTERRUPTIBLE);
828 spin_unlock(&sdp->sd_jindex_spin);
830 kthread_stop(sdp->sd_quotad_process);
831 kthread_stop(sdp->sd_logd_process);
833 if (!(sb->s_flags & MS_RDONLY)) {
834 error = gfs2_make_fs_ro(sdp);
838 /* At this point, we're through modifying the disk */
842 iput(sdp->sd_jindex);
843 iput(sdp->sd_statfs_inode);
844 iput(sdp->sd_rindex);
845 iput(sdp->sd_quota_inode);
847 gfs2_glock_put(sdp->sd_rename_gl);
848 gfs2_glock_put(sdp->sd_trans_gl);
850 if (!sdp->sd_args.ar_spectator) {
851 gfs2_glock_dq_uninit(&sdp->sd_journal_gh);
852 gfs2_glock_dq_uninit(&sdp->sd_jinode_gh);
853 gfs2_glock_dq_uninit(&sdp->sd_sc_gh);
854 gfs2_glock_dq_uninit(&sdp->sd_qc_gh);
855 iput(sdp->sd_sc_inode);
856 iput(sdp->sd_qc_inode);
859 gfs2_glock_dq_uninit(&sdp->sd_live_gh);
860 gfs2_clear_rgrpd(sdp);
861 gfs2_jindex_free(sdp);
862 /* Take apart glock structures and buffer lists */
863 gfs2_gl_hash_clear(sdp);
864 /* Unmount the locking protocol */
865 gfs2_lm_unmount(sdp);
867 /* At this point, we're through participating in the lockspace */
868 gfs2_sys_fs_del(sdp);
872 * gfs2_sync_fs - sync the filesystem
873 * @sb: the superblock
875 * Flushes the log to disk.
878 static int gfs2_sync_fs(struct super_block *sb, int wait)
880 if (wait && sb->s_fs_info)
881 gfs2_log_flush(sb->s_fs_info, NULL);
886 * gfs2_freeze - prevent further writes to the filesystem
887 * @sb: the VFS structure for the filesystem
891 static int gfs2_freeze(struct super_block *sb)
893 struct gfs2_sbd *sdp = sb->s_fs_info;
896 if (test_bit(SDF_SHUTDOWN, &sdp->sd_flags))
900 error = gfs2_freeze_fs(sdp);
906 fs_err(sdp, "waiting for recovery before freeze\n");
910 fs_err(sdp, "error freezing FS: %d\n", error);
914 fs_err(sdp, "retrying...\n");
921 * gfs2_unfreeze - reallow writes to the filesystem
922 * @sb: the VFS structure for the filesystem
926 static int gfs2_unfreeze(struct super_block *sb)
928 gfs2_unfreeze_fs(sb->s_fs_info);
933 * statfs_fill - fill in the sg for a given RG
935 * @sc: the sc structure
937 * Returns: 0 on success, -ESTALE if the LVB is invalid
940 static int statfs_slow_fill(struct gfs2_rgrpd *rgd,
941 struct gfs2_statfs_change_host *sc)
943 gfs2_rgrp_verify(rgd);
944 sc->sc_total += rgd->rd_data;
945 sc->sc_free += rgd->rd_free;
946 sc->sc_dinodes += rgd->rd_dinodes;
951 * gfs2_statfs_slow - Stat a filesystem using asynchronous locking
952 * @sdp: the filesystem
953 * @sc: the sc info that will be returned
955 * Any error (other than a signal) will cause this routine to fall back
956 * to the synchronous version.
958 * FIXME: This really shouldn't busy wait like this.
963 static int gfs2_statfs_slow(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
965 struct gfs2_holder ri_gh;
966 struct gfs2_rgrpd *rgd_next;
967 struct gfs2_holder *gha, *gh;
968 unsigned int slots = 64;
973 memset(sc, 0, sizeof(struct gfs2_statfs_change_host));
974 gha = kcalloc(slots, sizeof(struct gfs2_holder), GFP_KERNEL);
978 error = gfs2_rindex_hold(sdp, &ri_gh);
982 rgd_next = gfs2_rgrpd_get_first(sdp);
987 for (x = 0; x < slots; x++) {
990 if (gh->gh_gl && gfs2_glock_poll(gh)) {
991 err = gfs2_glock_wait(gh);
993 gfs2_holder_uninit(gh);
997 error = statfs_slow_fill(
998 gh->gh_gl->gl_object, sc);
999 gfs2_glock_dq_uninit(gh);
1005 else if (rgd_next && !error) {
1006 error = gfs2_glock_nq_init(rgd_next->rd_gl,
1010 rgd_next = gfs2_rgrpd_get_next(rgd_next);
1014 if (signal_pending(current))
1015 error = -ERESTARTSYS;
1024 gfs2_glock_dq_uninit(&ri_gh);
1032 * gfs2_statfs_i - Do a statfs
1033 * @sdp: the filesystem
1034 * @sg: the sg structure
1039 static int gfs2_statfs_i(struct gfs2_sbd *sdp, struct gfs2_statfs_change_host *sc)
1041 struct gfs2_statfs_change_host *m_sc = &sdp->sd_statfs_master;
1042 struct gfs2_statfs_change_host *l_sc = &sdp->sd_statfs_local;
1044 spin_lock(&sdp->sd_statfs_spin);
1047 sc->sc_total += l_sc->sc_total;
1048 sc->sc_free += l_sc->sc_free;
1049 sc->sc_dinodes += l_sc->sc_dinodes;
1051 spin_unlock(&sdp->sd_statfs_spin);
1053 if (sc->sc_free < 0)
1055 if (sc->sc_free > sc->sc_total)
1056 sc->sc_free = sc->sc_total;
1057 if (sc->sc_dinodes < 0)
1064 * gfs2_statfs - Gather and return stats about the filesystem
1065 * @sb: The superblock
1066 * @statfsbuf: The buffer
1068 * Returns: 0 on success or error code
1071 static int gfs2_statfs(struct dentry *dentry, struct kstatfs *buf)
1073 struct super_block *sb = dentry->d_inode->i_sb;
1074 struct gfs2_sbd *sdp = sb->s_fs_info;
1075 struct gfs2_statfs_change_host sc;
1078 if (gfs2_tune_get(sdp, gt_statfs_slow))
1079 error = gfs2_statfs_slow(sdp, &sc);
1081 error = gfs2_statfs_i(sdp, &sc);
1086 buf->f_type = GFS2_MAGIC;
1087 buf->f_bsize = sdp->sd_sb.sb_bsize;
1088 buf->f_blocks = sc.sc_total;
1089 buf->f_bfree = sc.sc_free;
1090 buf->f_bavail = sc.sc_free;
1091 buf->f_files = sc.sc_dinodes + sc.sc_free;
1092 buf->f_ffree = sc.sc_free;
1093 buf->f_namelen = GFS2_FNAMESIZE;
1099 * gfs2_remount_fs - called when the FS is remounted
1100 * @sb: the filesystem
1101 * @flags: the remount flags
1102 * @data: extra data passed in (not used right now)
1107 static int gfs2_remount_fs(struct super_block *sb, int *flags, char *data)
1109 struct gfs2_sbd *sdp = sb->s_fs_info;
1110 struct gfs2_args args = sdp->sd_args; /* Default to current settings */
1111 struct gfs2_tune *gt = &sdp->sd_tune;
1114 spin_lock(>->gt_spin);
1115 args.ar_commit = gt->gt_log_flush_secs;
1116 args.ar_quota_quantum = gt->gt_quota_quantum;
1117 if (gt->gt_statfs_slow)
1118 args.ar_statfs_quantum = 0;
1120 args.ar_statfs_quantum = gt->gt_statfs_quantum;
1121 spin_unlock(>->gt_spin);
1122 error = gfs2_mount_args(&args, data);
1126 /* Not allowed to change locking details */
1127 if (strcmp(args.ar_lockproto, sdp->sd_args.ar_lockproto) ||
1128 strcmp(args.ar_locktable, sdp->sd_args.ar_locktable) ||
1129 strcmp(args.ar_hostdata, sdp->sd_args.ar_hostdata))
1132 /* Some flags must not be changed */
1133 if (args_neq(&args, &sdp->sd_args, spectator) ||
1134 args_neq(&args, &sdp->sd_args, ignore_local_fs) ||
1135 args_neq(&args, &sdp->sd_args, localflocks) ||
1136 args_neq(&args, &sdp->sd_args, localcaching) ||
1137 args_neq(&args, &sdp->sd_args, meta))
1140 if (sdp->sd_args.ar_spectator)
1141 *flags |= MS_RDONLY;
1143 if ((sb->s_flags ^ *flags) & MS_RDONLY) {
1144 if (*flags & MS_RDONLY)
1145 error = gfs2_make_fs_ro(sdp);
1147 error = gfs2_make_fs_rw(sdp);
1152 sdp->sd_args = args;
1153 if (sdp->sd_args.ar_posix_acl)
1154 sb->s_flags |= MS_POSIXACL;
1156 sb->s_flags &= ~MS_POSIXACL;
1157 if (sdp->sd_args.ar_nobarrier)
1158 set_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1160 clear_bit(SDF_NOBARRIERS, &sdp->sd_flags);
1161 spin_lock(>->gt_spin);
1162 gt->gt_log_flush_secs = args.ar_commit;
1163 gt->gt_quota_quantum = args.ar_quota_quantum;
1164 if (args.ar_statfs_quantum) {
1165 gt->gt_statfs_slow = 0;
1166 gt->gt_statfs_quantum = args.ar_statfs_quantum;
1169 gt->gt_statfs_slow = 1;
1170 gt->gt_statfs_quantum = 30;
1172 spin_unlock(>->gt_spin);
1174 gfs2_online_uevent(sdp);
1179 * gfs2_drop_inode - Drop an inode (test for remote unlink)
1180 * @inode: The inode to drop
1182 * If we've received a callback on an iopen lock then its because a
1183 * remote node tried to deallocate the inode but failed due to this node
1184 * still having the inode open. Here we mark the link count zero
1185 * since we know that it must have reached zero if the GLF_DEMOTE flag
1186 * is set on the iopen glock. If we didn't do a disk read since the
1187 * remote node removed the final link then we might otherwise miss
1188 * this event. This check ensures that this node will deallocate the
1189 * inode's blocks, or alternatively pass the baton on to another
1190 * node for later deallocation.
1193 static void gfs2_drop_inode(struct inode *inode)
1195 struct gfs2_inode *ip = GFS2_I(inode);
1197 if (test_bit(GIF_USER, &ip->i_flags) && inode->i_nlink) {
1198 struct gfs2_glock *gl = ip->i_iopen_gh.gh_gl;
1199 if (gl && test_bit(GLF_DEMOTE, &gl->gl_flags))
1202 generic_drop_inode(inode);
1206 * gfs2_clear_inode - Deallocate an inode when VFS is done with it
1207 * @inode: The VFS inode
1211 static void gfs2_clear_inode(struct inode *inode)
1213 struct gfs2_inode *ip = GFS2_I(inode);
1215 /* This tells us its a "real" inode and not one which only
1216 * serves to contain an address space (see rgrp.c, meta_io.c)
1217 * which therefore doesn't have its own glocks.
1219 if (test_bit(GIF_USER, &ip->i_flags)) {
1220 ip->i_gl->gl_object = NULL;
1221 gfs2_glock_put(ip->i_gl);
1223 if (ip->i_iopen_gh.gh_gl) {
1224 ip->i_iopen_gh.gh_gl->gl_object = NULL;
1225 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1230 static int is_ancestor(const struct dentry *d1, const struct dentry *d2)
1236 } while (!IS_ROOT(d1));
1241 * gfs2_show_options - Show mount options for /proc/mounts
1242 * @s: seq_file structure
1245 * Returns: 0 on success or error code
1248 static int gfs2_show_options(struct seq_file *s, struct vfsmount *mnt)
1250 struct gfs2_sbd *sdp = mnt->mnt_sb->s_fs_info;
1251 struct gfs2_args *args = &sdp->sd_args;
1254 if (is_ancestor(mnt->mnt_root, sdp->sd_master_dir))
1255 seq_printf(s, ",meta");
1256 if (args->ar_lockproto[0])
1257 seq_printf(s, ",lockproto=%s", args->ar_lockproto);
1258 if (args->ar_locktable[0])
1259 seq_printf(s, ",locktable=%s", args->ar_locktable);
1260 if (args->ar_hostdata[0])
1261 seq_printf(s, ",hostdata=%s", args->ar_hostdata);
1262 if (args->ar_spectator)
1263 seq_printf(s, ",spectator");
1264 if (args->ar_ignore_local_fs)
1265 seq_printf(s, ",ignore_local_fs");
1266 if (args->ar_localflocks)
1267 seq_printf(s, ",localflocks");
1268 if (args->ar_localcaching)
1269 seq_printf(s, ",localcaching");
1271 seq_printf(s, ",debug");
1272 if (args->ar_upgrade)
1273 seq_printf(s, ",upgrade");
1274 if (args->ar_posix_acl)
1275 seq_printf(s, ",acl");
1276 if (args->ar_quota != GFS2_QUOTA_DEFAULT) {
1278 switch (args->ar_quota) {
1279 case GFS2_QUOTA_OFF:
1282 case GFS2_QUOTA_ACCOUNT:
1292 seq_printf(s, ",quota=%s", state);
1294 if (args->ar_suiddir)
1295 seq_printf(s, ",suiddir");
1296 if (args->ar_data != GFS2_DATA_DEFAULT) {
1298 switch (args->ar_data) {
1299 case GFS2_DATA_WRITEBACK:
1300 state = "writeback";
1302 case GFS2_DATA_ORDERED:
1309 seq_printf(s, ",data=%s", state);
1311 if (args->ar_discard)
1312 seq_printf(s, ",discard");
1313 val = sdp->sd_tune.gt_log_flush_secs;
1315 seq_printf(s, ",commit=%d", val);
1316 val = sdp->sd_tune.gt_statfs_quantum;
1318 seq_printf(s, ",statfs_quantum=%d", val);
1319 val = sdp->sd_tune.gt_quota_quantum;
1321 seq_printf(s, ",quota_quantum=%d", val);
1322 if (args->ar_statfs_percent)
1323 seq_printf(s, ",statfs_percent=%d", args->ar_statfs_percent);
1324 if (args->ar_errors != GFS2_ERRORS_DEFAULT) {
1327 switch (args->ar_errors) {
1328 case GFS2_ERRORS_WITHDRAW:
1331 case GFS2_ERRORS_PANIC:
1338 seq_printf(s, ",errors=%s", state);
1340 if (test_bit(SDF_NOBARRIERS, &sdp->sd_flags))
1341 seq_printf(s, ",nobarrier");
1347 * We have to (at the moment) hold the inodes main lock to cover
1348 * the gap between unlocking the shared lock on the iopen lock and
1349 * taking the exclusive lock. I'd rather do a shared -> exclusive
1350 * conversion on the iopen lock, but we can change that later. This
1351 * is safe, just less efficient.
1354 static void gfs2_delete_inode(struct inode *inode)
1356 struct gfs2_sbd *sdp = inode->i_sb->s_fs_info;
1357 struct gfs2_inode *ip = GFS2_I(inode);
1358 struct gfs2_holder gh;
1361 if (!test_bit(GIF_USER, &ip->i_flags))
1364 error = gfs2_glock_nq_init(ip->i_gl, LM_ST_EXCLUSIVE, 0, &gh);
1365 if (unlikely(error)) {
1366 gfs2_glock_dq_uninit(&ip->i_iopen_gh);
1370 error = gfs2_check_blk_type(sdp, ip->i_no_addr, GFS2_BLKST_UNLINKED);
1374 gfs2_glock_dq_wait(&ip->i_iopen_gh);
1375 gfs2_holder_reinit(LM_ST_EXCLUSIVE, LM_FLAG_TRY_1CB | GL_NOCACHE, &ip->i_iopen_gh);
1376 error = gfs2_glock_nq(&ip->i_iopen_gh);
1380 if (S_ISDIR(inode->i_mode) &&
1381 (ip->i_diskflags & GFS2_DIF_EXHASH)) {
1382 error = gfs2_dir_exhash_dealloc(ip);
1388 error = gfs2_ea_dealloc(ip);
1393 if (!gfs2_is_stuffed(ip)) {
1394 error = gfs2_file_dealloc(ip);
1399 error = gfs2_dinode_dealloc(ip);
1404 error = gfs2_trans_begin(sdp, 0, sdp->sd_jdesc->jd_blocks);
1407 /* Needs to be done before glock release & also in a transaction */
1408 truncate_inode_pages(&inode->i_data, 0);
1409 gfs2_trans_end(sdp);
1412 if (test_bit(HIF_HOLDER, &ip->i_iopen_gh.gh_iflags))
1413 gfs2_glock_dq(&ip->i_iopen_gh);
1414 gfs2_holder_uninit(&ip->i_iopen_gh);
1415 gfs2_glock_dq_uninit(&gh);
1416 if (error && error != GLR_TRYFAILED && error != -EROFS)
1417 fs_warn(sdp, "gfs2_delete_inode: %d\n", error);
1419 truncate_inode_pages(&inode->i_data, 0);
1423 static struct inode *gfs2_alloc_inode(struct super_block *sb)
1425 struct gfs2_inode *ip;
1427 ip = kmem_cache_alloc(gfs2_inode_cachep, GFP_KERNEL);
1432 return &ip->i_inode;
1435 static void gfs2_destroy_inode(struct inode *inode)
1437 kmem_cache_free(gfs2_inode_cachep, inode);
1440 const struct super_operations gfs2_super_ops = {
1441 .alloc_inode = gfs2_alloc_inode,
1442 .destroy_inode = gfs2_destroy_inode,
1443 .write_inode = gfs2_write_inode,
1444 .delete_inode = gfs2_delete_inode,
1445 .put_super = gfs2_put_super,
1446 .sync_fs = gfs2_sync_fs,
1447 .freeze_fs = gfs2_freeze,
1448 .unfreeze_fs = gfs2_unfreeze,
1449 .statfs = gfs2_statfs,
1450 .remount_fs = gfs2_remount_fs,
1451 .clear_inode = gfs2_clear_inode,
1452 .drop_inode = gfs2_drop_inode,
1453 .show_options = gfs2_show_options,