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
);
348 if (ip
->i_disksize
< (8 << 20) || ip
->i_disksize
> (1 << 30) ||
349 (ip
->i_disksize
& (sdp
->sd_sb
.sb_bsize
- 1))) {
350 gfs2_consist_inode(ip
);
353 jd
->jd_blocks
= ip
->i_disksize
>> sdp
->sd_sb
.sb_bsize_shift
;
355 error
= gfs2_write_alloc_required(ip
, 0, ip
->i_disksize
, &ar
);
357 gfs2_consist_inode(ip
);
365 * gfs2_make_fs_rw - Turn a Read-Only FS into a Read-Write one
366 * @sdp: the filesystem
371 int gfs2_make_fs_rw(struct gfs2_sbd
*sdp
)
373 struct gfs2_inode
*ip
= GFS2_I(sdp
->sd_jdesc
->jd_inode
);
374 struct gfs2_glock
*j_gl
= ip
->i_gl
;
375 struct gfs2_holder t_gh
;
376 struct gfs2_log_header_host head
;
379 error
= gfs2_glock_nq_init(sdp
->sd_trans_gl
, LM_ST_SHARED
, 0, &t_gh
);
383 j_gl
->gl_ops
->go_inval(j_gl
, DIO_METADATA
);
385 error
= gfs2_find_jhead(sdp
->sd_jdesc
, &head
);
389 if (!(head
.lh_flags
& GFS2_LOG_HEAD_UNMOUNT
)) {
395 /* Initialize some head of the log stuff */
396 sdp
->sd_log_sequence
= head
.lh_sequence
+ 1;
397 gfs2_log_pointers_init(sdp
, head
.lh_blkno
);
399 error
= gfs2_quota_init(sdp
);
403 set_bit(SDF_JOURNAL_LIVE
, &sdp
->sd_flags
);
405 gfs2_glock_dq_uninit(&t_gh
);
410 t_gh
.gh_flags
|= GL_NOCACHE
;
411 gfs2_glock_dq_uninit(&t_gh
);
416 void gfs2_statfs_change_in(struct gfs2_statfs_change_host
*sc
, const void *buf
)
418 const struct gfs2_statfs_change
*str
= buf
;
420 sc
->sc_total
= be64_to_cpu(str
->sc_total
);
421 sc
->sc_free
= be64_to_cpu(str
->sc_free
);
422 sc
->sc_dinodes
= be64_to_cpu(str
->sc_dinodes
);
425 static void gfs2_statfs_change_out(const struct gfs2_statfs_change_host
*sc
, void *buf
)
427 struct gfs2_statfs_change
*str
= buf
;
429 str
->sc_total
= cpu_to_be64(sc
->sc_total
);
430 str
->sc_free
= cpu_to_be64(sc
->sc_free
);
431 str
->sc_dinodes
= cpu_to_be64(sc
->sc_dinodes
);
434 int gfs2_statfs_init(struct gfs2_sbd
*sdp
)
436 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
437 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
438 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
439 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
440 struct buffer_head
*m_bh
, *l_bh
;
441 struct gfs2_holder gh
;
444 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, GL_NOCACHE
,
449 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
453 if (sdp
->sd_args
.ar_spectator
) {
454 spin_lock(&sdp
->sd_statfs_spin
);
455 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
456 sizeof(struct gfs2_dinode
));
457 spin_unlock(&sdp
->sd_statfs_spin
);
459 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
463 spin_lock(&sdp
->sd_statfs_spin
);
464 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
465 sizeof(struct gfs2_dinode
));
466 gfs2_statfs_change_in(l_sc
, l_bh
->b_data
+
467 sizeof(struct gfs2_dinode
));
468 spin_unlock(&sdp
->sd_statfs_spin
);
476 gfs2_glock_dq_uninit(&gh
);
480 void gfs2_statfs_change(struct gfs2_sbd
*sdp
, s64 total
, s64 free
,
483 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
484 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
485 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
486 struct buffer_head
*l_bh
;
491 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
495 gfs2_trans_add_bh(l_ip
->i_gl
, l_bh
, 1);
497 spin_lock(&sdp
->sd_statfs_spin
);
498 l_sc
->sc_total
+= total
;
499 l_sc
->sc_free
+= free
;
500 l_sc
->sc_dinodes
+= dinodes
;
501 gfs2_statfs_change_out(l_sc
, l_bh
->b_data
+ sizeof(struct gfs2_dinode
));
502 if (sdp
->sd_args
.ar_statfs_percent
) {
503 x
= 100 * l_sc
->sc_free
;
504 y
= m_sc
->sc_free
* sdp
->sd_args
.ar_statfs_percent
;
505 if (x
>= y
|| x
<= -y
)
508 spin_unlock(&sdp
->sd_statfs_spin
);
512 gfs2_wake_up_statfs(sdp
);
515 void update_statfs(struct gfs2_sbd
*sdp
, struct buffer_head
*m_bh
,
516 struct buffer_head
*l_bh
)
518 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
519 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
520 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
521 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
523 gfs2_trans_add_bh(l_ip
->i_gl
, l_bh
, 1);
525 spin_lock(&sdp
->sd_statfs_spin
);
526 m_sc
->sc_total
+= l_sc
->sc_total
;
527 m_sc
->sc_free
+= l_sc
->sc_free
;
528 m_sc
->sc_dinodes
+= l_sc
->sc_dinodes
;
529 memset(l_sc
, 0, sizeof(struct gfs2_statfs_change
));
530 memset(l_bh
->b_data
+ sizeof(struct gfs2_dinode
),
531 0, sizeof(struct gfs2_statfs_change
));
532 spin_unlock(&sdp
->sd_statfs_spin
);
534 gfs2_trans_add_bh(m_ip
->i_gl
, m_bh
, 1);
535 gfs2_statfs_change_out(m_sc
, m_bh
->b_data
+ sizeof(struct gfs2_dinode
));
538 int gfs2_statfs_sync(struct super_block
*sb
, int type
)
540 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
541 struct gfs2_inode
*m_ip
= GFS2_I(sdp
->sd_statfs_inode
);
542 struct gfs2_inode
*l_ip
= GFS2_I(sdp
->sd_sc_inode
);
543 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
544 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
545 struct gfs2_holder gh
;
546 struct buffer_head
*m_bh
, *l_bh
;
549 error
= gfs2_glock_nq_init(m_ip
->i_gl
, LM_ST_EXCLUSIVE
, GL_NOCACHE
,
554 error
= gfs2_meta_inode_buffer(m_ip
, &m_bh
);
558 spin_lock(&sdp
->sd_statfs_spin
);
559 gfs2_statfs_change_in(m_sc
, m_bh
->b_data
+
560 sizeof(struct gfs2_dinode
));
561 if (!l_sc
->sc_total
&& !l_sc
->sc_free
&& !l_sc
->sc_dinodes
) {
562 spin_unlock(&sdp
->sd_statfs_spin
);
565 spin_unlock(&sdp
->sd_statfs_spin
);
567 error
= gfs2_meta_inode_buffer(l_ip
, &l_bh
);
571 error
= gfs2_trans_begin(sdp
, 2 * RES_DINODE
, 0);
575 update_statfs(sdp
, m_bh
, l_bh
);
576 sdp
->sd_statfs_force_sync
= 0;
585 gfs2_glock_dq_uninit(&gh
);
590 struct list_head list
;
591 struct gfs2_holder gh
;
595 * gfs2_lock_fs_check_clean - Stop all writes to the FS and check that all
597 * @sdp: the file system
598 * @state: the state to put the transaction lock into
599 * @t_gh: the hold on the transaction lock
604 static int gfs2_lock_fs_check_clean(struct gfs2_sbd
*sdp
,
605 struct gfs2_holder
*t_gh
)
607 struct gfs2_inode
*ip
;
608 struct gfs2_jdesc
*jd
;
611 struct gfs2_log_header_host lh
;
614 list_for_each_entry(jd
, &sdp
->sd_jindex_list
, jd_list
) {
615 lfcc
= kmalloc(sizeof(struct lfcc
), GFP_KERNEL
);
620 ip
= GFS2_I(jd
->jd_inode
);
621 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_SHARED
, 0, &lfcc
->gh
);
626 list_add(&lfcc
->list
, &list
);
629 error
= gfs2_glock_nq_init(sdp
->sd_trans_gl
, LM_ST_DEFERRED
,
632 list_for_each_entry(jd
, &sdp
->sd_jindex_list
, jd_list
) {
633 error
= gfs2_jdesc_check(jd
);
636 error
= gfs2_find_jhead(jd
, &lh
);
639 if (!(lh
.lh_flags
& GFS2_LOG_HEAD_UNMOUNT
)) {
646 gfs2_glock_dq_uninit(t_gh
);
649 while (!list_empty(&list
)) {
650 lfcc
= list_entry(list
.next
, struct lfcc
, list
);
651 list_del(&lfcc
->list
);
652 gfs2_glock_dq_uninit(&lfcc
->gh
);
659 * gfs2_freeze_fs - freezes the file system
660 * @sdp: the file system
662 * This function flushes data and meta data for all machines by
663 * aquiring the transaction log exclusively. All journals are
664 * ensured to be in a clean state as well.
669 int gfs2_freeze_fs(struct gfs2_sbd
*sdp
)
673 mutex_lock(&sdp
->sd_freeze_lock
);
675 if (!sdp
->sd_freeze_count
++) {
676 error
= gfs2_lock_fs_check_clean(sdp
, &sdp
->sd_freeze_gh
);
678 sdp
->sd_freeze_count
--;
681 mutex_unlock(&sdp
->sd_freeze_lock
);
687 * gfs2_unfreeze_fs - unfreezes the file system
688 * @sdp: the file system
690 * This function allows the file system to proceed by unlocking
691 * the exclusively held transaction lock. Other GFS2 nodes are
692 * now free to acquire the lock shared and go on with their lives.
696 void gfs2_unfreeze_fs(struct gfs2_sbd
*sdp
)
698 mutex_lock(&sdp
->sd_freeze_lock
);
700 if (sdp
->sd_freeze_count
&& !--sdp
->sd_freeze_count
)
701 gfs2_glock_dq_uninit(&sdp
->sd_freeze_gh
);
703 mutex_unlock(&sdp
->sd_freeze_lock
);
708 * gfs2_write_inode - Make sure the inode is stable on the disk
710 * @sync: synchronous write flag
715 static int gfs2_write_inode(struct inode
*inode
, struct writeback_control
*wbc
)
717 struct gfs2_inode
*ip
= GFS2_I(inode
);
718 struct gfs2_sbd
*sdp
= GFS2_SB(inode
);
719 struct gfs2_holder gh
;
720 struct buffer_head
*bh
;
721 struct timespec atime
;
722 struct gfs2_dinode
*di
;
725 /* Check this is a "normal" inode, etc */
726 if (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
);
749 if (wbc
->sync_mode
== WB_SYNC_ALL
)
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, 1);
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 invalidate_inodes(sdp
->sd_vfs
);
864 gfs2_gl_hash_clear(sdp
);
865 /* Unmount the locking protocol */
866 gfs2_lm_unmount(sdp
);
868 /* At this point, we're through participating in the lockspace */
869 gfs2_sys_fs_del(sdp
);
873 * gfs2_sync_fs - sync the filesystem
874 * @sb: the superblock
876 * Flushes the log to disk.
879 static int gfs2_sync_fs(struct super_block
*sb
, int wait
)
881 if (wait
&& sb
->s_fs_info
)
882 gfs2_log_flush(sb
->s_fs_info
, NULL
);
887 * gfs2_freeze - prevent further writes to the filesystem
888 * @sb: the VFS structure for the filesystem
892 static int gfs2_freeze(struct super_block
*sb
)
894 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
897 if (test_bit(SDF_SHUTDOWN
, &sdp
->sd_flags
))
901 error
= gfs2_freeze_fs(sdp
);
907 fs_err(sdp
, "waiting for recovery before freeze\n");
911 fs_err(sdp
, "error freezing FS: %d\n", error
);
915 fs_err(sdp
, "retrying...\n");
922 * gfs2_unfreeze - reallow writes to the filesystem
923 * @sb: the VFS structure for the filesystem
927 static int gfs2_unfreeze(struct super_block
*sb
)
929 gfs2_unfreeze_fs(sb
->s_fs_info
);
934 * statfs_fill - fill in the sg for a given RG
936 * @sc: the sc structure
938 * Returns: 0 on success, -ESTALE if the LVB is invalid
941 static int statfs_slow_fill(struct gfs2_rgrpd
*rgd
,
942 struct gfs2_statfs_change_host
*sc
)
944 gfs2_rgrp_verify(rgd
);
945 sc
->sc_total
+= rgd
->rd_data
;
946 sc
->sc_free
+= rgd
->rd_free
;
947 sc
->sc_dinodes
+= rgd
->rd_dinodes
;
952 * gfs2_statfs_slow - Stat a filesystem using asynchronous locking
953 * @sdp: the filesystem
954 * @sc: the sc info that will be returned
956 * Any error (other than a signal) will cause this routine to fall back
957 * to the synchronous version.
959 * FIXME: This really shouldn't busy wait like this.
964 static int gfs2_statfs_slow(struct gfs2_sbd
*sdp
, struct gfs2_statfs_change_host
*sc
)
966 struct gfs2_holder ri_gh
;
967 struct gfs2_rgrpd
*rgd_next
;
968 struct gfs2_holder
*gha
, *gh
;
969 unsigned int slots
= 64;
974 memset(sc
, 0, sizeof(struct gfs2_statfs_change_host
));
975 gha
= kcalloc(slots
, sizeof(struct gfs2_holder
), GFP_KERNEL
);
979 error
= gfs2_rindex_hold(sdp
, &ri_gh
);
983 rgd_next
= gfs2_rgrpd_get_first(sdp
);
988 for (x
= 0; x
< slots
; x
++) {
991 if (gh
->gh_gl
&& gfs2_glock_poll(gh
)) {
992 err
= gfs2_glock_wait(gh
);
994 gfs2_holder_uninit(gh
);
998 error
= statfs_slow_fill(
999 gh
->gh_gl
->gl_object
, sc
);
1000 gfs2_glock_dq_uninit(gh
);
1006 else if (rgd_next
&& !error
) {
1007 error
= gfs2_glock_nq_init(rgd_next
->rd_gl
,
1011 rgd_next
= gfs2_rgrpd_get_next(rgd_next
);
1015 if (signal_pending(current
))
1016 error
= -ERESTARTSYS
;
1025 gfs2_glock_dq_uninit(&ri_gh
);
1033 * gfs2_statfs_i - Do a statfs
1034 * @sdp: the filesystem
1035 * @sg: the sg structure
1040 static int gfs2_statfs_i(struct gfs2_sbd
*sdp
, struct gfs2_statfs_change_host
*sc
)
1042 struct gfs2_statfs_change_host
*m_sc
= &sdp
->sd_statfs_master
;
1043 struct gfs2_statfs_change_host
*l_sc
= &sdp
->sd_statfs_local
;
1045 spin_lock(&sdp
->sd_statfs_spin
);
1048 sc
->sc_total
+= l_sc
->sc_total
;
1049 sc
->sc_free
+= l_sc
->sc_free
;
1050 sc
->sc_dinodes
+= l_sc
->sc_dinodes
;
1052 spin_unlock(&sdp
->sd_statfs_spin
);
1054 if (sc
->sc_free
< 0)
1056 if (sc
->sc_free
> sc
->sc_total
)
1057 sc
->sc_free
= sc
->sc_total
;
1058 if (sc
->sc_dinodes
< 0)
1065 * gfs2_statfs - Gather and return stats about the filesystem
1066 * @sb: The superblock
1067 * @statfsbuf: The buffer
1069 * Returns: 0 on success or error code
1072 static int gfs2_statfs(struct dentry
*dentry
, struct kstatfs
*buf
)
1074 struct super_block
*sb
= dentry
->d_inode
->i_sb
;
1075 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1076 struct gfs2_statfs_change_host sc
;
1079 if (gfs2_tune_get(sdp
, gt_statfs_slow
))
1080 error
= gfs2_statfs_slow(sdp
, &sc
);
1082 error
= gfs2_statfs_i(sdp
, &sc
);
1087 buf
->f_type
= GFS2_MAGIC
;
1088 buf
->f_bsize
= sdp
->sd_sb
.sb_bsize
;
1089 buf
->f_blocks
= sc
.sc_total
;
1090 buf
->f_bfree
= sc
.sc_free
;
1091 buf
->f_bavail
= sc
.sc_free
;
1092 buf
->f_files
= sc
.sc_dinodes
+ sc
.sc_free
;
1093 buf
->f_ffree
= sc
.sc_free
;
1094 buf
->f_namelen
= GFS2_FNAMESIZE
;
1100 * gfs2_remount_fs - called when the FS is remounted
1101 * @sb: the filesystem
1102 * @flags: the remount flags
1103 * @data: extra data passed in (not used right now)
1108 static int gfs2_remount_fs(struct super_block
*sb
, int *flags
, char *data
)
1110 struct gfs2_sbd
*sdp
= sb
->s_fs_info
;
1111 struct gfs2_args args
= sdp
->sd_args
; /* Default to current settings */
1112 struct gfs2_tune
*gt
= &sdp
->sd_tune
;
1115 spin_lock(>
->gt_spin
);
1116 args
.ar_commit
= gt
->gt_logd_secs
;
1117 args
.ar_quota_quantum
= gt
->gt_quota_quantum
;
1118 if (gt
->gt_statfs_slow
)
1119 args
.ar_statfs_quantum
= 0;
1121 args
.ar_statfs_quantum
= gt
->gt_statfs_quantum
;
1122 spin_unlock(>
->gt_spin
);
1123 error
= gfs2_mount_args(&args
, data
);
1127 /* Not allowed to change locking details */
1128 if (strcmp(args
.ar_lockproto
, sdp
->sd_args
.ar_lockproto
) ||
1129 strcmp(args
.ar_locktable
, sdp
->sd_args
.ar_locktable
) ||
1130 strcmp(args
.ar_hostdata
, sdp
->sd_args
.ar_hostdata
))
1133 /* Some flags must not be changed */
1134 if (args_neq(&args
, &sdp
->sd_args
, spectator
) ||
1135 args_neq(&args
, &sdp
->sd_args
, ignore_local_fs
) ||
1136 args_neq(&args
, &sdp
->sd_args
, localflocks
) ||
1137 args_neq(&args
, &sdp
->sd_args
, localcaching
) ||
1138 args_neq(&args
, &sdp
->sd_args
, meta
))
1141 if (sdp
->sd_args
.ar_spectator
)
1142 *flags
|= MS_RDONLY
;
1144 if ((sb
->s_flags
^ *flags
) & MS_RDONLY
) {
1145 if (*flags
& MS_RDONLY
)
1146 error
= gfs2_make_fs_ro(sdp
);
1148 error
= gfs2_make_fs_rw(sdp
);
1153 sdp
->sd_args
= args
;
1154 if (sdp
->sd_args
.ar_posix_acl
)
1155 sb
->s_flags
|= MS_POSIXACL
;
1157 sb
->s_flags
&= ~MS_POSIXACL
;
1158 if (sdp
->sd_args
.ar_nobarrier
)
1159 set_bit(SDF_NOBARRIERS
, &sdp
->sd_flags
);
1161 clear_bit(SDF_NOBARRIERS
, &sdp
->sd_flags
);
1162 spin_lock(>
->gt_spin
);
1163 gt
->gt_logd_secs
= args
.ar_commit
;
1164 gt
->gt_quota_quantum
= args
.ar_quota_quantum
;
1165 if (args
.ar_statfs_quantum
) {
1166 gt
->gt_statfs_slow
= 0;
1167 gt
->gt_statfs_quantum
= args
.ar_statfs_quantum
;
1170 gt
->gt_statfs_slow
= 1;
1171 gt
->gt_statfs_quantum
= 30;
1173 spin_unlock(>
->gt_spin
);
1175 gfs2_online_uevent(sdp
);
1180 * gfs2_drop_inode - Drop an inode (test for remote unlink)
1181 * @inode: The inode to drop
1183 * If we've received a callback on an iopen lock then its because a
1184 * remote node tried to deallocate the inode but failed due to this node
1185 * still having the inode open. Here we mark the link count zero
1186 * since we know that it must have reached zero if the GLF_DEMOTE flag
1187 * is set on the iopen glock. If we didn't do a disk read since the
1188 * remote node removed the final link then we might otherwise miss
1189 * this event. This check ensures that this node will deallocate the
1190 * inode's blocks, or alternatively pass the baton on to another
1191 * node for later deallocation.
1194 static void gfs2_drop_inode(struct inode
*inode
)
1196 struct gfs2_inode
*ip
= GFS2_I(inode
);
1198 if (inode
->i_nlink
) {
1199 struct gfs2_glock
*gl
= ip
->i_iopen_gh
.gh_gl
;
1200 if (gl
&& test_bit(GLF_DEMOTE
, &gl
->gl_flags
))
1203 generic_drop_inode(inode
);
1207 * gfs2_clear_inode - Deallocate an inode when VFS is done with it
1208 * @inode: The VFS inode
1212 static void gfs2_clear_inode(struct inode
*inode
)
1214 struct gfs2_inode
*ip
= GFS2_I(inode
);
1216 ip
->i_gl
->gl_object
= NULL
;
1217 gfs2_glock_put(ip
->i_gl
);
1219 if (ip
->i_iopen_gh
.gh_gl
) {
1220 ip
->i_iopen_gh
.gh_gl
->gl_object
= NULL
;
1221 gfs2_glock_dq_uninit(&ip
->i_iopen_gh
);
1225 static int is_ancestor(const struct dentry
*d1
, const struct dentry
*d2
)
1231 } while (!IS_ROOT(d1
));
1236 * gfs2_show_options - Show mount options for /proc/mounts
1237 * @s: seq_file structure
1240 * Returns: 0 on success or error code
1243 static int gfs2_show_options(struct seq_file
*s
, struct vfsmount
*mnt
)
1245 struct gfs2_sbd
*sdp
= mnt
->mnt_sb
->s_fs_info
;
1246 struct gfs2_args
*args
= &sdp
->sd_args
;
1249 if (is_ancestor(mnt
->mnt_root
, sdp
->sd_master_dir
))
1250 seq_printf(s
, ",meta");
1251 if (args
->ar_lockproto
[0])
1252 seq_printf(s
, ",lockproto=%s", args
->ar_lockproto
);
1253 if (args
->ar_locktable
[0])
1254 seq_printf(s
, ",locktable=%s", args
->ar_locktable
);
1255 if (args
->ar_hostdata
[0])
1256 seq_printf(s
, ",hostdata=%s", args
->ar_hostdata
);
1257 if (args
->ar_spectator
)
1258 seq_printf(s
, ",spectator");
1259 if (args
->ar_ignore_local_fs
)
1260 seq_printf(s
, ",ignore_local_fs");
1261 if (args
->ar_localflocks
)
1262 seq_printf(s
, ",localflocks");
1263 if (args
->ar_localcaching
)
1264 seq_printf(s
, ",localcaching");
1266 seq_printf(s
, ",debug");
1267 if (args
->ar_upgrade
)
1268 seq_printf(s
, ",upgrade");
1269 if (args
->ar_posix_acl
)
1270 seq_printf(s
, ",acl");
1271 if (args
->ar_quota
!= GFS2_QUOTA_DEFAULT
) {
1273 switch (args
->ar_quota
) {
1274 case GFS2_QUOTA_OFF
:
1277 case GFS2_QUOTA_ACCOUNT
:
1287 seq_printf(s
, ",quota=%s", state
);
1289 if (args
->ar_suiddir
)
1290 seq_printf(s
, ",suiddir");
1291 if (args
->ar_data
!= GFS2_DATA_DEFAULT
) {
1293 switch (args
->ar_data
) {
1294 case GFS2_DATA_WRITEBACK
:
1295 state
= "writeback";
1297 case GFS2_DATA_ORDERED
:
1304 seq_printf(s
, ",data=%s", state
);
1306 if (args
->ar_discard
)
1307 seq_printf(s
, ",discard");
1308 val
= sdp
->sd_tune
.gt_logd_secs
;
1310 seq_printf(s
, ",commit=%d", val
);
1311 val
= sdp
->sd_tune
.gt_statfs_quantum
;
1313 seq_printf(s
, ",statfs_quantum=%d", val
);
1314 val
= sdp
->sd_tune
.gt_quota_quantum
;
1316 seq_printf(s
, ",quota_quantum=%d", val
);
1317 if (args
->ar_statfs_percent
)
1318 seq_printf(s
, ",statfs_percent=%d", args
->ar_statfs_percent
);
1319 if (args
->ar_errors
!= GFS2_ERRORS_DEFAULT
) {
1322 switch (args
->ar_errors
) {
1323 case GFS2_ERRORS_WITHDRAW
:
1326 case GFS2_ERRORS_PANIC
:
1333 seq_printf(s
, ",errors=%s", state
);
1335 if (test_bit(SDF_NOBARRIERS
, &sdp
->sd_flags
))
1336 seq_printf(s
, ",nobarrier");
1337 if (test_bit(SDF_DEMOTE
, &sdp
->sd_flags
))
1338 seq_printf(s
, ",demote_interface_used");
1343 * We have to (at the moment) hold the inodes main lock to cover
1344 * the gap between unlocking the shared lock on the iopen lock and
1345 * taking the exclusive lock. I'd rather do a shared -> exclusive
1346 * conversion on the iopen lock, but we can change that later. This
1347 * is safe, just less efficient.
1350 static void gfs2_delete_inode(struct inode
*inode
)
1352 struct gfs2_sbd
*sdp
= inode
->i_sb
->s_fs_info
;
1353 struct gfs2_inode
*ip
= GFS2_I(inode
);
1354 struct gfs2_holder gh
;
1357 error
= gfs2_glock_nq_init(ip
->i_gl
, LM_ST_EXCLUSIVE
, 0, &gh
);
1358 if (unlikely(error
)) {
1359 gfs2_glock_dq_uninit(&ip
->i_iopen_gh
);
1363 error
= gfs2_check_blk_type(sdp
, ip
->i_no_addr
, GFS2_BLKST_UNLINKED
);
1367 gfs2_glock_dq_wait(&ip
->i_iopen_gh
);
1368 gfs2_holder_reinit(LM_ST_EXCLUSIVE
, LM_FLAG_TRY_1CB
| GL_NOCACHE
, &ip
->i_iopen_gh
);
1369 error
= gfs2_glock_nq(&ip
->i_iopen_gh
);
1373 if (S_ISDIR(inode
->i_mode
) &&
1374 (ip
->i_diskflags
& GFS2_DIF_EXHASH
)) {
1375 error
= gfs2_dir_exhash_dealloc(ip
);
1381 error
= gfs2_ea_dealloc(ip
);
1386 if (!gfs2_is_stuffed(ip
)) {
1387 error
= gfs2_file_dealloc(ip
);
1392 error
= gfs2_dinode_dealloc(ip
);
1397 error
= gfs2_trans_begin(sdp
, 0, sdp
->sd_jdesc
->jd_blocks
);
1400 /* Needs to be done before glock release & also in a transaction */
1401 truncate_inode_pages(&inode
->i_data
, 0);
1402 gfs2_trans_end(sdp
);
1405 if (test_bit(HIF_HOLDER
, &ip
->i_iopen_gh
.gh_iflags
))
1406 gfs2_glock_dq(&ip
->i_iopen_gh
);
1407 gfs2_holder_uninit(&ip
->i_iopen_gh
);
1408 gfs2_glock_dq_uninit(&gh
);
1409 if (error
&& error
!= GLR_TRYFAILED
&& error
!= -EROFS
)
1410 fs_warn(sdp
, "gfs2_delete_inode: %d\n", error
);
1412 truncate_inode_pages(&inode
->i_data
, 0);
1416 static struct inode
*gfs2_alloc_inode(struct super_block
*sb
)
1418 struct gfs2_inode
*ip
;
1420 ip
= kmem_cache_alloc(gfs2_inode_cachep
, GFP_KERNEL
);
1425 return &ip
->i_inode
;
1428 static void gfs2_destroy_inode(struct inode
*inode
)
1430 kmem_cache_free(gfs2_inode_cachep
, inode
);
1433 const struct super_operations gfs2_super_ops
= {
1434 .alloc_inode
= gfs2_alloc_inode
,
1435 .destroy_inode
= gfs2_destroy_inode
,
1436 .write_inode
= gfs2_write_inode
,
1437 .delete_inode
= gfs2_delete_inode
,
1438 .put_super
= gfs2_put_super
,
1439 .sync_fs
= gfs2_sync_fs
,
1440 .freeze_fs
= gfs2_freeze
,
1441 .unfreeze_fs
= gfs2_unfreeze
,
1442 .statfs
= gfs2_statfs
,
1443 .remount_fs
= gfs2_remount_fs
,
1444 .clear_inode
= gfs2_clear_inode
,
1445 .drop_inode
= gfs2_drop_inode
,
1446 .show_options
= gfs2_show_options
,