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)
76 static const match_table_t tokens
= {
77 {Opt_lockproto
, "lockproto=%s"},
78 {Opt_locktable
, "locktable=%s"},
79 {Opt_hostdata
, "hostdata=%s"},
80 {Opt_spectator
, "spectator"},
81 {Opt_ignore_local_fs
, "ignore_local_fs"},
82 {Opt_localflocks
, "localflocks"},
83 {Opt_localcaching
, "localcaching"},
85 {Opt_nodebug
, "nodebug"},
86 {Opt_upgrade
, "upgrade"},
89 {Opt_quota_off
, "quota=off"},
90 {Opt_quota_account
, "quota=account"},
91 {Opt_quota_on
, "quota=on"},
93 {Opt_noquota
, "noquota"},
94 {Opt_suiddir
, "suiddir"},
95 {Opt_nosuiddir
, "nosuiddir"},
96 {Opt_data_writeback
, "data=writeback"},
97 {Opt_data_ordered
, "data=ordered"},
99 {Opt_discard
, "discard"},
100 {Opt_nodiscard
, "nodiscard"},
101 {Opt_commit
, "commit=%d"},
102 {Opt_err_withdraw
, "errors=withdraw"},
103 {Opt_err_panic
, "errors=panic"},
108 * gfs2_mount_args - Parse mount options
115 int gfs2_mount_args(struct gfs2_sbd
*sdp
, struct gfs2_args
*args
, char *options
)
119 substring_t tmp
[MAX_OPT_ARGS
];
122 /* Split the options into tokens with the "," character and
126 o
= strsep(&options
, ",");
132 token
= match_token(o
, tokens
, tmp
);
135 match_strlcpy(args
->ar_lockproto
, &tmp
[0],
139 match_strlcpy(args
->ar_locktable
, &tmp
[0],
143 match_strlcpy(args
->ar_hostdata
, &tmp
[0],
147 args
->ar_spectator
= 1;
149 case Opt_ignore_local_fs
:
150 args
->ar_ignore_local_fs
= 1;
152 case Opt_localflocks
:
153 args
->ar_localflocks
= 1;
155 case Opt_localcaching
:
156 args
->ar_localcaching
= 1;
159 if (args
->ar_errors
== GFS2_ERRORS_PANIC
) {
160 fs_info(sdp
, "-o debug and -o errors=panic "
161 "are mutually exclusive.\n");
170 args
->ar_upgrade
= 1;
173 args
->ar_posix_acl
= 1;
176 args
->ar_posix_acl
= 0;
180 args
->ar_quota
= GFS2_QUOTA_OFF
;
182 case Opt_quota_account
:
183 args
->ar_quota
= GFS2_QUOTA_ACCOUNT
;
187 args
->ar_quota
= GFS2_QUOTA_ON
;
190 args
->ar_suiddir
= 1;
193 args
->ar_suiddir
= 0;
195 case Opt_data_writeback
:
196 args
->ar_data
= GFS2_DATA_WRITEBACK
;
198 case Opt_data_ordered
:
199 args
->ar_data
= GFS2_DATA_ORDERED
;
205 args
->ar_discard
= 1;
208 args
->ar_discard
= 0;
211 rv
= match_int(&tmp
[0], &args
->ar_commit
);
212 if (rv
|| args
->ar_commit
<= 0) {
213 fs_info(sdp
, "commit mount option requires a positive numeric argument\n");
214 return rv
? rv
: -EINVAL
;
217 case Opt_err_withdraw
:
218 args
->ar_errors
= GFS2_ERRORS_WITHDRAW
;
221 if (args
->ar_debug
) {
222 fs_info(sdp
, "-o debug and -o errors=panic "
223 "are mutually exclusive.\n");
226 args
->ar_errors
= GFS2_ERRORS_PANIC
;
230 fs_info(sdp
, "invalid mount option: %s\n", o
);
239 * gfs2_jindex_free - Clear all the journal index information
240 * @sdp: The GFS2 superblock
244 void gfs2_jindex_free(struct gfs2_sbd
*sdp
)
246 struct list_head list
, *head
;
247 struct gfs2_jdesc
*jd
;
248 struct gfs2_journal_extent
*jext
;
250 spin_lock(&sdp
->sd_jindex_spin
);
251 list_add(&list
, &sdp
->sd_jindex_list
);
252 list_del_init(&sdp
->sd_jindex_list
);
253 sdp
->sd_journals
= 0;
254 spin_unlock(&sdp
->sd_jindex_spin
);
256 while (!list_empty(&list
)) {
257 jd
= list_entry(list
.next
, struct gfs2_jdesc
, jd_list
);
258 head
= &jd
->extent_list
;
259 while (!list_empty(head
)) {
260 jext
= list_entry(head
->next
,
261 struct gfs2_journal_extent
,
263 list_del(&jext
->extent_list
);
266 list_del(&jd
->jd_list
);
272 static struct gfs2_jdesc
*jdesc_find_i(struct list_head
*head
, unsigned int jid
)
274 struct gfs2_jdesc
*jd
;
277 list_for_each_entry(jd
, head
, jd_list
) {
278 if (jd
->jd_jid
== jid
) {
290 struct gfs2_jdesc
*gfs2_jdesc_find(struct gfs2_sbd
*sdp
, unsigned int jid
)
292 struct gfs2_jdesc
*jd
;
294 spin_lock(&sdp
->sd_jindex_spin
);
295 jd
= jdesc_find_i(&sdp
->sd_jindex_list
, jid
);
296 spin_unlock(&sdp
->sd_jindex_spin
);
301 int gfs2_jdesc_check(struct gfs2_jdesc
*jd
)
303 struct gfs2_inode
*ip
= GFS2_I(jd
->jd_inode
);
304 struct gfs2_sbd
*sdp
= GFS2_SB(jd
->jd_inode
);
308 if (ip
->i_disksize
< (8 << 20) || ip
->i_disksize
> (1 << 30) ||
309 (ip
->i_disksize
& (sdp
->sd_sb
.sb_bsize
- 1))) {
310 gfs2_consist_inode(ip
);
313 jd
->jd_blocks
= ip
->i_disksize
>> sdp
->sd_sb
.sb_bsize_shift
;
315 error
= gfs2_write_alloc_required(ip
, 0, ip
->i_disksize
, &ar
);
317 gfs2_consist_inode(ip
);
325 * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one
326 * @sdp: the filesystem
331 int gfs2_make_fs_rw(struct gfs2_sbd
*sdp
)
333 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_jdesc
->jd_inode
);
334 struct gfs2_glock
*j_gl
= ip
->i_gl
;
335 struct gfs2_holder t_gh
;
336 struct gfs2_log_header_host head
;
339 error
= gfs2_glock_nq_init(sdp
->sd_trans_gl
, LM_ST_SHARED
, 0, &t_gh
);
343 j_gl
->gl_ops
->go_inval(j_gl
, DIO_METADATA
);
345 error
= gfs2_find_jhead(sdp
->sd_jdesc
, &head
);
349 if (!(head
.lh_flags
& GFS2_LOG_HEAD_UNMOUNT
)) {
355 /* Initialize some head of the log stuff */
356 sdp
->sd_log_sequence
= head
.lh_sequence
+ 1;
357 gfs2_log_pointers_init(sdp
, head
.lh_blkno
);
359 error
= gfs2_quota_init(sdp
);
363 set_bit(SDF_JOURNAL_LIVE
, &sdp
->sd_flags
);
365 gfs2_glock_dq_uninit(&t_gh
);
370 t_gh
.gh_flags
|= GL_NOCACHE
;
371 gfs2_glock_dq_uninit(&t_gh
);
376 void gfs2_statfs_change_in(struct gfs2_statfs_change_host
*sc
, const void *buf
)
378 const struct gfs2_statfs_change
*str
= buf
;
380 sc
->sc_total
= be64_to_cpu(str
->sc_total
);
381 sc
->sc_free
= be64_to_cpu(str
->sc_free
);
382 sc
->sc_dinodes
= be64_to_cpu(str
->sc_dinodes
);
385 static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host
*sc
, void *buf
)
387 struct gfs2_statfs_change
*str
= buf
;
389 str
->sc_total
= cpu_to_be64(sc
->sc_total
);
390 str
->sc_free
= cpu_to_be64(sc
->sc_free
);
391 str
->sc_dinodes
= cpu_to_be64(sc
->sc_dinodes
);
394 int gfs2_statfs_init(struct gfs2_sbd
*sdp
)
396 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
397 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
398 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
399 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
400 struct buffer_head
*m_bh
, *l_bh
;
401 struct gfs2_holder gh
;
404 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, GL_NOCACHE
,
409 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
413 if (sdp
->sd_args
.ar_spectator
) {
414 spin_lock(&sdp
->sd_statfs_spin
);
415 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
416 sizeof(struct gfs2_dinode
));
417 spin_unlock(&sdp
->sd_statfs_spin
);
419 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
423 spin_lock(&sdp
->sd_statfs_spin
);
424 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
425 sizeof(struct gfs2_dinode
));
426 gfs2_statfs_change_in(l_sc
, l_bh
->b_data
+
427 sizeof(struct gfs2_dinode
));
428 spin_unlock(&sdp
->sd_statfs_spin
);
436 gfs2_glock_dq_uninit(&gh
);
440 void gfs2_statfs_change(struct gfs2_sbd
*sdp
, s64 total
, s64 free
,
443 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
444 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
445 struct buffer_head
*l_bh
;
448 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
452 gfs2_trans_add_bh(l_ip
->i_gl
, l_bh
, 1);
454 spin_lock(&sdp
->sd_statfs_spin
);
455 l_sc
->sc_total
+= total
;
456 l_sc
->sc_free
+= free
;
457 l_sc
->sc_dinodes
+= dinodes
;
458 gfs2_statfs_change_out(l_sc
, l_bh
->b_data
+ sizeof(struct gfs2_dinode
));
459 spin_unlock(&sdp
->sd_statfs_spin
);
464 void update_statfs(struct gfs2_sbd
*sdp
, struct buffer_head
*m_bh
,
465 struct buffer_head
*l_bh
)
467 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
468 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
469 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
470 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
472 gfs2_trans_add_bh(l_ip
->i_gl
, l_bh
, 1);
474 spin_lock(&sdp
->sd_statfs_spin
);
475 m_sc
->sc_total
+= l_sc
->sc_total
;
476 m_sc
->sc_free
+= l_sc
->sc_free
;
477 m_sc
->sc_dinodes
+= l_sc
->sc_dinodes
;
478 memset(l_sc
, 0, sizeof(struct gfs2_statfs_change
));
479 memset(l_bh
->b_data
+ sizeof(struct gfs2_dinode
),
480 0, sizeof(struct gfs2_statfs_change
));
481 spin_unlock(&sdp
->sd_statfs_spin
);
483 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
484 gfs2_statfs_change_out(m_sc
, m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
487 int gfs2_statfs_sync(struct gfs2_sbd
*sdp
)
489 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
490 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
491 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
492 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
493 struct gfs2_holder gh
;
494 struct buffer_head
*m_bh
, *l_bh
;
497 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, GL_NOCACHE
,
502 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
506 spin_lock(&sdp
->sd_statfs_spin
);
507 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
508 sizeof(struct gfs2_dinode
));
509 if (!l_sc
->sc_total
&& !l_sc
->sc_free
&& !l_sc
->sc_dinodes
) {
510 spin_unlock(&sdp
->sd_statfs_spin
);
513 spin_unlock(&sdp
->sd_statfs_spin
);
515 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
519 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
523 update_statfs(sdp
, m_bh
, l_bh
);
532 gfs2_glock_dq_uninit(&gh
);
537 struct list_head list
;
538 struct gfs2_holder gh
;
542 * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all
544 * @sdp: the file system
545 * @state: the state to put the transaction lock into
546 * @t_gh: the hold on the transaction lock
551 static int gfs2_lock_fs_check_clean(struct gfs2_sbd
*sdp
,
552 struct gfs2_holder
*t_gh
)
554 struct gfs2_inode
*ip
;
555 struct gfs2_jdesc
*jd
;
558 struct gfs2_log_header_host lh
;
561 list_for_each_entry(jd
, &sdp
->sd_jindex_list
, jd_list
) {
562 lfcc
= kmalloc(sizeof(struct lfcc
), GFP_KERNEL
);
567 ip
= GFS2_I(jd
->jd_inode
);
568 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, 0, &lfcc
->gh
);
573 list_add(&lfcc
->list
, &list
);
576 error
= gfs2_glock_nq_init(sdp
->sd_trans_gl
, LM_ST_DEFERRED
,
579 list_for_each_entry(jd
, &sdp
->sd_jindex_list
, jd_list
) {
580 error
= gfs2_jdesc_check(jd
);
583 error
= gfs2_find_jhead(jd
, &lh
);
586 if (!(lh
.lh_flags
& GFS2_LOG_HEAD_UNMOUNT
)) {
593 gfs2_glock_dq_uninit(t_gh
);
596 while (!list_empty(&list
)) {
597 lfcc
= list_entry(list
.next
, struct lfcc
, list
);
598 list_del(&lfcc
->list
);
599 gfs2_glock_dq_uninit(&lfcc
->gh
);
606 * gfs2_freeze_fs - freezes the file system
607 * @sdp: the file system
609 * This function flushes data and meta data for all machines by
610 * aquiring the transaction log exclusively. All journals are
611 * ensured to be in a clean state as well.
616 int gfs2_freeze_fs(struct gfs2_sbd
*sdp
)
620 mutex_lock(&sdp
->sd_freeze_lock
);
622 if (!sdp
->sd_freeze_count
++) {
623 error
= gfs2_lock_fs_check_clean(sdp
, &sdp
->sd_freeze_gh
);
625 sdp
->sd_freeze_count
--;
628 mutex_unlock(&sdp
->sd_freeze_lock
);
634 * gfs2_unfreeze_fs - unfreezes the file system
635 * @sdp: the file system
637 * This function allows the file system to proceed by unlocking
638 * the exclusively held transaction lock. Other GFS2 nodes are
639 * now free to acquire the lock shared and go on with their lives.
643 void gfs2_unfreeze_fs(struct gfs2_sbd
*sdp
)
645 mutex_lock(&sdp
->sd_freeze_lock
);
647 if (sdp
->sd_freeze_count
&& !--sdp
->sd_freeze_count
)
648 gfs2_glock_dq_uninit(&sdp
->sd_freeze_gh
);
650 mutex_unlock(&sdp
->sd_freeze_lock
);
655 * gfs2_write_inode - Make sure the inode is stable on the disk
657 * @sync: synchronous write flag
662 static int gfs2_write_inode(struct inode
*inode
, int sync
)
664 struct gfs2_inode
*ip
= GFS2_I(inode
);
665 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
666 struct gfs2_holder gh
;
667 struct buffer_head
*bh
;
668 struct timespec atime
;
669 struct gfs2_dinode
*di
;
672 /* Check this is a "normal" inode, etc */
673 if (!test_bit(GIF_USER
, &ip
->i_flags
) ||
674 (current
->flags
& PF_MEMALLOC
))
676 ret
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
679 ret
= gfs2_trans_begin(sdp
, RES_DINODE
, 0);
682 ret
= gfs2_meta_inode_buffer(ip
, &bh
);
684 di
= (struct gfs2_dinode
*)bh
->b_data
;
685 atime
.tv_sec
= be64_to_cpu(di
->di_atime
);
686 atime
.tv_nsec
= be32_to_cpu(di
->di_atime_nsec
);
687 if (timespec_compare(&inode
->i_atime
, &atime
) > 0) {
688 gfs2_trans_add_bh(ip
->i_gl
, bh
, 1);
689 gfs2_dinode_out(ip
, bh
->b_data
);
695 gfs2_glock_dq_uninit(&gh
);
698 gfs2_log_flush(GFS2_SB(inode
), ip
->i_gl
);
703 * gfs2_make_fs_ro - Turn a Read-Write FS into a Read-Only one
704 * @sdp: the filesystem
709 static int gfs2_make_fs_ro(struct gfs2_sbd
*sdp
)
711 struct gfs2_holder t_gh
;
714 flush_workqueue(gfs2_delete_workqueue
);
715 gfs2_quota_sync(sdp
);
716 gfs2_statfs_sync(sdp
);
718 error
= gfs2_glock_nq_init(sdp
->sd_trans_gl
, LM_ST_SHARED
, GL_NOCACHE
,
720 if (error
&& !test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
))
723 gfs2_meta_syncfs(sdp
);
724 gfs2_log_shutdown(sdp
);
726 clear_bit(SDF_JOURNAL_LIVE
, &sdp
->sd_flags
);
729 gfs2_glock_dq_uninit(&t_gh
);
731 gfs2_quota_cleanup(sdp
);
736 static int gfs2_umount_recovery_wait(void *word
)
743 * gfs2_put_super - Unmount the filesystem
744 * @sb: The VFS superblock
748 static void gfs2_put_super(struct super_block
*sb
)
750 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
752 struct gfs2_jdesc
*jd
;
754 /* Unfreeze the filesystem, if we need to */
756 mutex_lock(&sdp
->sd_freeze_lock
);
757 if (sdp
->sd_freeze_count
)
758 gfs2_glock_dq_uninit(&sdp
->sd_freeze_gh
);
759 mutex_unlock(&sdp
->sd_freeze_lock
);
761 /* No more recovery requests */
762 set_bit(SDF_NORECOVERY
, &sdp
->sd_flags
);
765 /* Wait on outstanding recovery */
767 spin_lock(&sdp
->sd_jindex_spin
);
768 list_for_each_entry(jd
, &sdp
->sd_jindex_list
, jd_list
) {
769 if (!test_bit(JDF_RECOVERY
, &jd
->jd_flags
))
771 spin_unlock(&sdp
->sd_jindex_spin
);
772 wait_on_bit(&jd
->jd_flags
, JDF_RECOVERY
,
773 gfs2_umount_recovery_wait
, TASK_UNINTERRUPTIBLE
);
776 spin_unlock(&sdp
->sd_jindex_spin
);
778 kthread_stop(sdp
->sd_quotad_process
);
779 kthread_stop(sdp
->sd_logd_process
);
781 if (!(sb
->s_flags
& MS_RDONLY
)) {
782 error
= gfs2_make_fs_ro(sdp
);
786 /* At this point, we're through modifying the disk */
790 iput(sdp
->sd_jindex
);
791 iput(sdp
->sd_statfs_inode
);
792 iput(sdp
->sd_rindex
);
793 iput(sdp
->sd_quota_inode
);
795 gfs2_glock_put(sdp
->sd_rename_gl
);
796 gfs2_glock_put(sdp
->sd_trans_gl
);
798 if (!sdp
->sd_args
.ar_spectator
) {
799 gfs2_glock_dq_uninit(&sdp
->sd_journal_gh
);
800 gfs2_glock_dq_uninit(&sdp
->sd_jinode_gh
);
801 gfs2_glock_dq_uninit(&sdp
->sd_sc_gh
);
802 gfs2_glock_dq_uninit(&sdp
->sd_qc_gh
);
803 iput(sdp
->sd_sc_inode
);
804 iput(sdp
->sd_qc_inode
);
807 gfs2_glock_dq_uninit(&sdp
->sd_live_gh
);
808 gfs2_clear_rgrpd(sdp
);
809 gfs2_jindex_free(sdp
);
810 /* Take apart glock structures and buffer lists */
811 gfs2_gl_hash_clear(sdp
);
812 /* Unmount the locking protocol */
813 gfs2_lm_unmount(sdp
);
815 /* At this point, we're through participating in the lockspace */
816 gfs2_sys_fs_del(sdp
);
820 * gfs2_sync_fs - sync the filesystem
821 * @sb: the superblock
823 * Flushes the log to disk.
826 static int gfs2_sync_fs(struct super_block
*sb
, int wait
)
828 if (wait
&& sb
->s_fs_info
)
829 gfs2_log_flush(sb
->s_fs_info
, NULL
);
834 * gfs2_freeze - prevent further writes to the filesystem
835 * @sb: the VFS structure for the filesystem
839 static int gfs2_freeze(struct super_block
*sb
)
841 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
844 if (test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
))
848 error
= gfs2_freeze_fs(sdp
);
854 fs_err(sdp
, "waiting for recovery before freeze\n");
858 fs_err(sdp
, "error freezing FS: %d\n", error
);
862 fs_err(sdp
, "retrying...\n");
869 * gfs2_unfreeze - reallow writes to the filesystem
870 * @sb: the VFS structure for the filesystem
874 static int gfs2_unfreeze(struct super_block
*sb
)
876 gfs2_unfreeze_fs(sb
->s_fs_info
);
881 * statfs_fill - fill in the sg for a given RG
883 * @sc: the sc structure
885 * Returns: 0 on success, -ESTALE if the LVB is invalid
888 static int statfs_slow_fill(struct gfs2_rgrpd
*rgd
,
889 struct gfs2_statfs_change_host
*sc
)
891 gfs2_rgrp_verify(rgd
);
892 sc
->sc_total
+= rgd
->rd_data
;
893 sc
->sc_free
+= rgd
->rd_free
;
894 sc
->sc_dinodes
+= rgd
->rd_dinodes
;
899 * gfs2_statfs_slow - Stat a filesystem using asynchronous locking
900 * @sdp: the filesystem
901 * @sc: the sc info that will be returned
903 * Any error (other than a signal) will cause this routine to fall back
904 * to the synchronous version.
906 * FIXME: This really shouldn't busy wait like this.
911 static int gfs2_statfs_slow(struct gfs2_sbd
*sdp
, struct gfs2_statfs_change_host
*sc
)
913 struct gfs2_holder ri_gh
;
914 struct gfs2_rgrpd
*rgd_next
;
915 struct gfs2_holder
*gha
, *gh
;
916 unsigned int slots
= 64;
921 memset(sc
, 0, sizeof(struct gfs2_statfs_change_host
));
922 gha
= kcalloc(slots
, sizeof(struct gfs2_holder
), GFP_KERNEL
);
926 error
= gfs2_rindex_hold(sdp
, &ri_gh
);
930 rgd_next
= gfs2_rgrpd_get_first(sdp
);
935 for (x
= 0; x
< slots
; x
++) {
938 if (gh
->gh_gl
&& gfs2_glock_poll(gh
)) {
939 err
= gfs2_glock_wait(gh
);
941 gfs2_holder_uninit(gh
);
945 error
= statfs_slow_fill(
946 gh
->gh_gl
->gl_object
, sc
);
947 gfs2_glock_dq_uninit(gh
);
953 else if (rgd_next
&& !error
) {
954 error
= gfs2_glock_nq_init(rgd_next
->rd_gl
,
958 rgd_next
= gfs2_rgrpd_get_next(rgd_next
);
962 if (signal_pending(current
))
963 error
= -ERESTARTSYS
;
972 gfs2_glock_dq_uninit(&ri_gh
);
980 * gfs2_statfs_i - Do a statfs
981 * @sdp: the filesystem
982 * @sg: the sg structure
987 static int gfs2_statfs_i(struct gfs2_sbd
*sdp
, struct gfs2_statfs_change_host
*sc
)
989 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
990 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
992 spin_lock(&sdp
->sd_statfs_spin
);
995 sc
->sc_total
+= l_sc
->sc_total
;
996 sc
->sc_free
+= l_sc
->sc_free
;
997 sc
->sc_dinodes
+= l_sc
->sc_dinodes
;
999 spin_unlock(&sdp
->sd_statfs_spin
);
1001 if (sc
->sc_free
< 0)
1003 if (sc
->sc_free
> sc
->sc_total
)
1004 sc
->sc_free
= sc
->sc_total
;
1005 if (sc
->sc_dinodes
< 0)
1012 * gfs2_statfs - Gather and return stats about the filesystem
1013 * @sb: The superblock
1014 * @statfsbuf: The buffer
1016 * Returns: 0 on success or error code
1019 static int gfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1021 struct super_block
*sb
= dentry
->d_inode
->i_sb
;
1022 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1023 struct gfs2_statfs_change_host sc
;
1026 if (gfs2_tune_get(sdp
, gt_statfs_slow
))
1027 error
= gfs2_statfs_slow(sdp
, &sc
);
1029 error
= gfs2_statfs_i(sdp
, &sc
);
1034 buf
->f_type
= GFS2_MAGIC
;
1035 buf
->f_bsize
= sdp
->sd_sb
.sb_bsize
;
1036 buf
->f_blocks
= sc
.sc_total
;
1037 buf
->f_bfree
= sc
.sc_free
;
1038 buf
->f_bavail
= sc
.sc_free
;
1039 buf
->f_files
= sc
.sc_dinodes
+ sc
.sc_free
;
1040 buf
->f_ffree
= sc
.sc_free
;
1041 buf
->f_namelen
= GFS2_FNAMESIZE
;
1047 * gfs2_remount_fs - called when the FS is remounted
1048 * @sb: the filesystem
1049 * @flags: the remount flags
1050 * @data: extra data passed in (not used right now)
1055 static int gfs2_remount_fs(struct super_block
*sb
, int *flags
, char *data
)
1057 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1058 struct gfs2_args args
= sdp
->sd_args
; /* Default to current settings */
1059 struct gfs2_tune
*gt
= &sdp
->sd_tune
;
1062 spin_lock(>
->gt_spin
);
1063 args
.ar_commit
= gt
->gt_log_flush_secs
;
1064 spin_unlock(>
->gt_spin
);
1065 error
= gfs2_mount_args(sdp
, &args
, data
);
1069 /* Not allowed to change locking details */
1070 if (strcmp(args
.ar_lockproto
, sdp
->sd_args
.ar_lockproto
) ||
1071 strcmp(args
.ar_locktable
, sdp
->sd_args
.ar_locktable
) ||
1072 strcmp(args
.ar_hostdata
, sdp
->sd_args
.ar_hostdata
))
1075 /* Some flags must not be changed */
1076 if (args_neq(&args
, &sdp
->sd_args
, spectator
) ||
1077 args_neq(&args
, &sdp
->sd_args
, ignore_local_fs
) ||
1078 args_neq(&args
, &sdp
->sd_args
, localflocks
) ||
1079 args_neq(&args
, &sdp
->sd_args
, localcaching
) ||
1080 args_neq(&args
, &sdp
->sd_args
, meta
))
1083 if (sdp
->sd_args
.ar_spectator
)
1084 *flags
|= MS_RDONLY
;
1086 if ((sb
->s_flags
^ *flags
) & MS_RDONLY
) {
1087 if (*flags
& MS_RDONLY
)
1088 error
= gfs2_make_fs_ro(sdp
);
1090 error
= gfs2_make_fs_rw(sdp
);
1095 sdp
->sd_args
= args
;
1096 if (sdp
->sd_args
.ar_posix_acl
)
1097 sb
->s_flags
|= MS_POSIXACL
;
1099 sb
->s_flags
&= ~MS_POSIXACL
;
1100 spin_lock(>
->gt_spin
);
1101 gt
->gt_log_flush_secs
= args
.ar_commit
;
1102 spin_unlock(>
->gt_spin
);
1104 gfs2_online_uevent(sdp
);
1109 * gfs2_drop_inode - Drop an inode (test for remote unlink)
1110 * @inode: The inode to drop
1112 * If we've received a callback on an iopen lock then its because a
1113 * remote node tried to deallocate the inode but failed due to this node
1114 * still having the inode open. Here we mark the link count zero
1115 * since we know that it must have reached zero if the GLF_DEMOTE flag
1116 * is set on the iopen glock. If we didn't do a disk read since the
1117 * remote node removed the final link then we might otherwise miss
1118 * this event. This check ensures that this node will deallocate the
1119 * inode's blocks, or alternatively pass the baton on to another
1120 * node for later deallocation.
1123 static void gfs2_drop_inode(struct inode
*inode
)
1125 struct gfs2_inode
*ip
= GFS2_I(inode
);
1127 if (test_bit(GIF_USER
, &ip
->i_flags
) && inode
->i_nlink
) {
1128 struct gfs2_glock
*gl
= ip
->i_iopen_gh
.gh_gl
;
1129 if (gl
&& test_bit(GLF_DEMOTE
, &gl
->gl_flags
))
1132 generic_drop_inode(inode
);
1136 * gfs2_clear_inode - Deallocate an inode when VFS is done with it
1137 * @inode: The VFS inode
1141 static void gfs2_clear_inode(struct inode
*inode
)
1143 struct gfs2_inode
*ip
= GFS2_I(inode
);
1145 /* This tells us its a "real" inode and not one which only
1146 * serves to contain an address space (see rgrp.c, meta_io.c)
1147 * which therefore doesn't have its own glocks.
1149 if (test_bit(GIF_USER
, &ip
->i_flags
)) {
1150 ip
->i_gl
->gl_object
= NULL
;
1151 gfs2_glock_put(ip
->i_gl
);
1153 if (ip
->i_iopen_gh
.gh_gl
) {
1154 ip
->i_iopen_gh
.gh_gl
->gl_object
= NULL
;
1155 gfs2_glock_dq_uninit(&ip
->i_iopen_gh
);
1160 static int is_ancestor(const struct dentry
*d1
, const struct dentry
*d2
)
1166 } while (!IS_ROOT(d1
));
1171 * gfs2_show_options - Show mount options for /proc/mounts
1172 * @s: seq_file structure
1175 * Returns: 0 on success or error code
1178 static int gfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1180 struct gfs2_sbd
*sdp
= mnt
->mnt_sb
->s_fs_info
;
1181 struct gfs2_args
*args
= &sdp
->sd_args
;
1184 if (is_ancestor(mnt
->mnt_root
, sdp
->sd_master_dir
))
1185 seq_printf(s
, ",meta");
1186 if (args
->ar_lockproto
[0])
1187 seq_printf(s
, ",lockproto=%s", args
->ar_lockproto
);
1188 if (args
->ar_locktable
[0])
1189 seq_printf(s
, ",locktable=%s", args
->ar_locktable
);
1190 if (args
->ar_hostdata
[0])
1191 seq_printf(s
, ",hostdata=%s", args
->ar_hostdata
);
1192 if (args
->ar_spectator
)
1193 seq_printf(s
, ",spectator");
1194 if (args
->ar_ignore_local_fs
)
1195 seq_printf(s
, ",ignore_local_fs");
1196 if (args
->ar_localflocks
)
1197 seq_printf(s
, ",localflocks");
1198 if (args
->ar_localcaching
)
1199 seq_printf(s
, ",localcaching");
1201 seq_printf(s
, ",debug");
1202 if (args
->ar_upgrade
)
1203 seq_printf(s
, ",upgrade");
1204 if (args
->ar_posix_acl
)
1205 seq_printf(s
, ",acl");
1206 if (args
->ar_quota
!= GFS2_QUOTA_DEFAULT
) {
1208 switch (args
->ar_quota
) {
1209 case GFS2_QUOTA_OFF
:
1212 case GFS2_QUOTA_ACCOUNT
:
1222 seq_printf(s
, ",quota=%s", state
);
1224 if (args
->ar_suiddir
)
1225 seq_printf(s
, ",suiddir");
1226 if (args
->ar_data
!= GFS2_DATA_DEFAULT
) {
1228 switch (args
->ar_data
) {
1229 case GFS2_DATA_WRITEBACK
:
1230 state
= "writeback";
1232 case GFS2_DATA_ORDERED
:
1239 seq_printf(s
, ",data=%s", state
);
1241 if (args
->ar_discard
)
1242 seq_printf(s
, ",discard");
1243 lfsecs
= sdp
->sd_tune
.gt_log_flush_secs
;
1245 seq_printf(s
, ",commit=%d", lfsecs
);
1246 if (args
->ar_errors
!= GFS2_ERRORS_DEFAULT
) {
1249 switch (args
->ar_errors
) {
1250 case GFS2_ERRORS_WITHDRAW
:
1253 case GFS2_ERRORS_PANIC
:
1260 seq_printf(s
, ",errors=%s", state
);
1266 * We have to (at the moment) hold the inodes main lock to cover
1267 * the gap between unlocking the shared lock on the iopen lock and
1268 * taking the exclusive lock. I'd rather do a shared -> exclusive
1269 * conversion on the iopen lock, but we can change that later. This
1270 * is safe, just less efficient.
1273 static void gfs2_delete_inode(struct inode
*inode
)
1275 struct gfs2_sbd
*sdp
= inode
->i_sb
->s_fs_info
;
1276 struct gfs2_inode
*ip
= GFS2_I(inode
);
1277 struct gfs2_holder gh
;
1280 if (!test_bit(GIF_USER
, &ip
->i_flags
))
1283 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
1284 if (unlikely(error
)) {
1285 gfs2_glock_dq_uninit(&ip
->i_iopen_gh
);
1289 error
= gfs2_check_blk_type(sdp
, ip
->i_no_addr
, GFS2_BLKST_UNLINKED
);
1293 gfs2_glock_dq_wait(&ip
->i_iopen_gh
);
1294 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, LM_FLAG_TRY_1CB
| GL_NOCACHE
, &ip
->i_iopen_gh
);
1295 error
= gfs2_glock_nq(&ip
->i_iopen_gh
);
1299 if (S_ISDIR(inode
->i_mode
) &&
1300 (ip
->i_diskflags
& GFS2_DIF_EXHASH
)) {
1301 error
= gfs2_dir_exhash_dealloc(ip
);
1307 error
= gfs2_ea_dealloc(ip
);
1312 if (!gfs2_is_stuffed(ip
)) {
1313 error
= gfs2_file_dealloc(ip
);
1318 error
= gfs2_dinode_dealloc(ip
);
1323 error
= gfs2_trans_begin(sdp
, 0, sdp
->sd_jdesc
->jd_blocks
);
1326 /* Needs to be done before glock release & also in a transaction */
1327 truncate_inode_pages(&inode
->i_data
, 0);
1328 gfs2_trans_end(sdp
);
1331 if (test_bit(HIF_HOLDER
, &ip
->i_iopen_gh
.gh_iflags
))
1332 gfs2_glock_dq(&ip
->i_iopen_gh
);
1333 gfs2_holder_uninit(&ip
->i_iopen_gh
);
1334 gfs2_glock_dq_uninit(&gh
);
1335 if (error
&& error
!= GLR_TRYFAILED
&& error
!= -EROFS
)
1336 fs_warn(sdp
, "gfs2_delete_inode: %d\n", error
);
1338 truncate_inode_pages(&inode
->i_data
, 0);
1342 static struct inode
*gfs2_alloc_inode(struct super_block
*sb
)
1344 struct gfs2_inode
*ip
;
1346 ip
= kmem_cache_alloc(gfs2_inode_cachep
, GFP_KERNEL
);
1351 return &ip
->i_inode
;
1354 static void gfs2_destroy_inode(struct inode
*inode
)
1356 kmem_cache_free(gfs2_inode_cachep
, inode
);
1359 const struct super_operations gfs2_super_ops
= {
1360 .alloc_inode
= gfs2_alloc_inode
,
1361 .destroy_inode
= gfs2_destroy_inode
,
1362 .write_inode
= gfs2_write_inode
,
1363 .delete_inode
= gfs2_delete_inode
,
1364 .put_super
= gfs2_put_super
,
1365 .sync_fs
= gfs2_sync_fs
,
1366 .freeze_fs
= gfs2_freeze
,
1367 .unfreeze_fs
= gfs2_unfreeze
,
1368 .statfs
= gfs2_statfs
,
1369 .remount_fs
= gfs2_remount_fs
,
1370 .clear_inode
= gfs2_clear_inode
,
1371 .drop_inode
= gfs2_drop_inode
,
1372 .show_options
= gfs2_show_options
,