2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
35 int devnum
; /* to sync with mdstat info */
39 int active
, working
, failed
, spare
, raid
;
42 int devstate
[MAX_DISKS
];
43 dev_t devid
[MAX_DISKS
];
45 int parent_dev
; /* For subarray, devnum of parent.
48 struct supertype
*metadata
;
49 struct state
*subarray
;/* for a container it is a link to first subarray
50 * for a subarray it is a link to next subarray
51 * in the same container */
52 struct state
*parent
; /* for a subarray it is a link to its container
63 static int make_daemon(char *pidfile
);
64 static int check_one_sharer(int scan
);
65 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
);
66 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
67 int test
, struct alert_info
*info
,
68 int increments
, char *prefer
);
69 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
70 int test
, struct alert_info
*info
);
71 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
);
72 static void link_containers_with_subarrays(struct state
*list
);
74 int Monitor(struct mddev_dev
*devlist
,
75 char *mailaddr
, char *alert_cmd
,
77 int daemonise
, int oneshot
,
78 int dosyslog
, char *pidfile
, int increments
,
82 * Every few seconds, scan every md device looking for changes
83 * When a change is found, log it, possibly run the alert command,
84 * and possibly send Email
86 * For each array, we record:
88 * active/working/failed/spare drives
89 * State of each device.
90 * %rebuilt if rebuilding
92 * If the update time changes, check out all the data again
93 * It is possible that we cannot get the state of each device
94 * due to bugs in the md kernel module.
95 * We also read /proc/mdstat to get rebuild percent,
96 * and to get state on all active devices incase of kernel bug.
100 * An active device had Faulty set or Active/Sync removed
102 * A spare device had Faulty set
104 * An active device had a reverse transition
106 * percent went from -1 to +ve
108 * percent went from below to not-below NN%
110 * Couldn't access a device which was previously visible
112 * if we detect an array with active<raid and spare==0
113 * we look at other arrays that have same spare-group
114 * If we find one with active==raid and spare>0,
115 * and if we can get_disk_info and find a name
116 * Then we hot-remove and hot-add to the other array
118 * If devlist is NULL, then we can monitor everything because --scan
119 * was given. We get an initial list from config file and add anything
120 * that appears in /proc/mdstat
123 struct state
*statelist
= NULL
;
126 struct mdstat_ent
*mdstat
= NULL
;
127 char *mailfrom
= NULL
;
128 struct alert_info info
;
131 mailaddr
= conf_get_mailaddr();
132 if (mailaddr
&& ! c
->scan
)
133 pr_err("Monitor using email address \"%s\" from config file\n",
136 mailfrom
= conf_get_mailfrom();
139 alert_cmd
= conf_get_program();
140 if (alert_cmd
&& ! c
->scan
)
141 pr_err("Monitor using program \"%s\" from config file\n",
144 if (c
->scan
&& !mailaddr
&& !alert_cmd
&& !dosyslog
) {
145 pr_err("No mail address or alert command - not monitoring.\n");
148 info
.alert_cmd
= alert_cmd
;
149 info
.mailaddr
= mailaddr
;
150 info
.mailfrom
= mailfrom
;
151 info
.dosyslog
= dosyslog
;
154 int rv
= make_daemon(pidfile
);
160 if (check_one_sharer(c
->scan
))
163 if (devlist
== NULL
) {
164 struct mddev_ident
*mdlist
= conf_get_ident(NULL
);
165 for (; mdlist
; mdlist
=mdlist
->next
) {
167 if (mdlist
->devname
== NULL
)
169 if (strcasecmp(mdlist
->devname
, "<ignore>") == 0)
171 st
= xcalloc(1, sizeof *st
);
172 if (mdlist
->devname
[0] == '/')
173 st
->devname
= xstrdup(mdlist
->devname
);
175 st
->devname
= xmalloc(8+strlen(mdlist
->devname
)+1);
176 strcpy(strcpy(st
->devname
, "/dev/md/"),
179 st
->next
= statelist
;
180 st
->devnum
= INT_MAX
;
181 st
->percent
= RESYNC_UNKNOWN
;
183 st
->expected_spares
= mdlist
->spare_disks
;
184 if (mdlist
->spare_group
)
185 st
->spare_group
= xstrdup(mdlist
->spare_group
);
189 struct mddev_dev
*dv
;
190 for (dv
=devlist
; dv
; dv
=dv
->next
) {
191 struct mddev_ident
*mdlist
= conf_get_ident(dv
->devname
);
192 struct state
*st
= xcalloc(1, sizeof *st
);
193 st
->devname
= xstrdup(dv
->devname
);
194 st
->next
= statelist
;
195 st
->devnum
= INT_MAX
;
196 st
->percent
= RESYNC_UNKNOWN
;
197 st
->expected_spares
= -1;
199 st
->expected_spares
= mdlist
->spare_disks
;
200 if (mdlist
->spare_group
)
201 st
->spare_group
= xstrdup(mdlist
->spare_group
);
215 mdstat
= mdstat_read(oneshot
?0:1, 0);
217 for (st
=statelist
; st
; st
=st
->next
)
218 if (check_array(st
, mdstat
, c
->test
, &info
,
219 increments
, c
->prefer
))
222 /* now check if there are any new devices found in mdstat */
224 new_found
= add_new_arrays(mdstat
, &statelist
, c
->test
,
227 /* If an array has active < raid && spare == 0 && spare_group != NULL
228 * Look for another array with spare > 0 and active == raid and same spare_group
229 * if found, choose a device and hotremove/hotadd
231 if (share
&& anydegraded
)
232 try_spare_migration(statelist
, &info
);
237 mdstat_wait(c
->delay
);
241 for (st2
= statelist
; st2
; st2
= statelist
) {
242 statelist
= st2
->next
;
251 static int make_daemon(char *pidfile
)
254 * -1 in the forked daemon
257 * so a none-negative becomes the exit code.
265 pid_file
=fopen(pidfile
, "w");
267 perror("cannot create pid file");
269 fprintf(pid_file
,"%d\n", pid
);
280 open("/dev/null", O_RDWR
);
287 static int check_one_sharer(int scan
)
294 sprintf(path
, "%s/autorebuild.pid", MDMON_DIR
);
295 fp
= fopen(path
, "r");
297 if (fscanf(fp
, "%d", &pid
) != 1)
299 sprintf(dir
, "/proc/%d", pid
);
300 rv
= stat(dir
, &buf
);
304 "autorebuild process allowed"
305 " in scan mode, aborting\n");
309 pr_err("Warning: One"
310 " autorebuild process already"
317 if (mkdir(MDMON_DIR
, S_IRWXU
) < 0 &&
319 pr_err("Can't create "
320 "autorebuild.pid file\n");
322 fp
= fopen(path
, "w");
324 pr_err("Cannot create"
329 fprintf(fp
, "%d\n", pid
);
337 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
)
341 if (!info
->alert_cmd
&& !info
->mailaddr
&& !info
->dosyslog
) {
342 time_t now
= time(0);
344 printf("%1.15s: %s on %s %s\n", ctime(&now
)+4, event
, dev
, disc
?disc
:"unknown device");
346 if (info
->alert_cmd
) {
350 waitpid(pid
, NULL
, 0);
355 execl(info
->alert_cmd
, info
->alert_cmd
,
356 event
, dev
, disc
, NULL
);
360 if (info
->mailaddr
&&
361 (strncmp(event
, "Fail", 4)==0 ||
362 strncmp(event
, "Test", 4)==0 ||
363 strncmp(event
, "Spares", 6)==0 ||
364 strncmp(event
, "Degrade", 7)==0)) {
365 FILE *mp
= popen(Sendmail
, "w");
369 gethostname(hname
, sizeof(hname
));
370 signal(SIGPIPE
, SIG_IGN
);
372 fprintf(mp
, "From: %s\n", info
->mailfrom
);
374 fprintf(mp
, "From: " Name
" monitoring <root>\n");
375 fprintf(mp
, "To: %s\n", info
->mailaddr
);
376 fprintf(mp
, "Subject: %s event on %s:%s\n\n",
380 "This is an automatically generated"
381 " mail message from " Name
"\n");
382 fprintf(mp
, "running on %s\n\n", hname
);
385 "A %s event had been detected on"
386 " md device %s.\n\n", event
, dev
);
388 if (disc
&& disc
[0] != ' ')
390 "It could be related to"
391 " component device %s.\n\n", disc
);
392 if (disc
&& disc
[0] == ' ')
393 fprintf(mp
, "Extra information:%s.\n\n", disc
);
395 fprintf(mp
, "Faithfully yours, etc.\n");
397 mdstat
= fopen("/proc/mdstat", "r");
402 "\nP.S. The /proc/mdstat file"
403 " currently contains the following:\n\n");
404 while ( (n
=fread(buf
, 1, sizeof(buf
), mdstat
)) > 0)
405 n
=fwrite(buf
, 1, n
, mp
);
412 /* log the event to syslog maybe */
413 if (info
->dosyslog
) {
414 /* Log at a different severity depending on the event.
416 * These are the critical events: */
417 if (strncmp(event
, "Fail", 4)==0 ||
418 strncmp(event
, "Degrade", 7)==0 ||
419 strncmp(event
, "DeviceDisappeared", 17)==0)
421 /* Good to know about, but are not failures: */
422 else if (strncmp(event
, "Rebuild", 7)==0 ||
423 strncmp(event
, "MoveSpare", 9)==0 ||
424 strncmp(event
, "Spares", 6) != 0)
425 priority
= LOG_WARNING
;
426 /* Everything else: */
432 "%s event detected on md device %s,"
433 " component device %s", event
, dev
, disc
);
436 "%s event detected on md device %s",
441 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
442 int test
, struct alert_info
*ainfo
,
443 int increments
, char *prefer
)
445 /* Update the state 'st' to reflect any changes shown in mdstat,
446 * or found by directly examining the array, and return
447 * '1' if the array is degraded, or '0' if it is optimal (or dead).
449 struct { int state
, major
, minor
; } info
[MAX_DISKS
];
450 mdu_array_info_t array
;
451 struct mdstat_ent
*mse
= NULL
, *mse2
;
452 char *dev
= st
->devname
;
460 alert("TestMessage", dev
, NULL
, ainfo
);
461 fd
= open(dev
, O_RDONLY
);
464 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
468 fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
469 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)<0) {
471 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
476 /* It's much easier to list what array levels can't
477 * have a device disappear than all of them that can
479 if (array
.level
== 0 || array
.level
== -1) {
480 if (!st
->err
&& !st
->from_config
)
481 alert("DeviceDisappeared", dev
, "Wrong-Level", ainfo
);
486 if (st
->devnum
== INT_MAX
) {
488 if (fstat(fd
, &stb
) == 0 &&
489 (S_IFMT
&stb
.st_mode
)==S_IFBLK
) {
490 if (major(stb
.st_rdev
) == MD_MAJOR
)
491 st
->devnum
= minor(stb
.st_rdev
);
493 st
->devnum
= -1- (minor(stb
.st_rdev
)>>6);
497 for (mse2
= mdstat
; mse2
; mse2
=mse2
->next
)
498 if (mse2
->devnum
== st
->devnum
) {
499 mse2
->devnum
= INT_MAX
; /* flag it as "used" */
504 /* duplicated array in statelist
505 * or re-created after reading mdstat*/
510 /* this array is in /proc/mdstat */
511 if (array
.utime
== 0)
512 /* external arrays don't update utime, so
513 * just make sure it is always different. */
514 array
.utime
= st
->utime
+ 1;;
517 /* New array appeared where previously had and error */
519 st
->percent
= RESYNC_NONE
;
521 alert("NewArray", st
->devname
, NULL
, ainfo
);
524 if (st
->utime
== array
.utime
&&
525 st
->failed
== array
.failed_disks
&&
526 st
->working
== array
.working_disks
&&
527 st
->spare
== array
.spare_disks
&&
529 mse
->percent
== st
->percent
532 if ((st
->active
< st
->raid
) && st
->spare
== 0)
537 if (st
->utime
== 0 && /* new array */
538 mse
->pattern
&& strchr(mse
->pattern
, '_') /* degraded */
540 alert("DegradedArray", dev
, NULL
, ainfo
);
542 if (st
->utime
== 0 && /* new array */
543 st
->expected_spares
> 0 &&
544 array
.spare_disks
< st
->expected_spares
)
545 alert("SparesMissing", dev
, NULL
, ainfo
);
546 if (st
->percent
< 0 && st
->percent
!= RESYNC_UNKNOWN
&&
548 alert("RebuildStarted", dev
, NULL
, ainfo
);
549 if (st
->percent
>= 0 &&
551 (mse
->percent
/ increments
) > (st
->percent
/ increments
)) {
552 char percentalert
[15]; // "RebuildNN" (10 chars) or "RebuildStarted" (15 chars)
554 if((mse
->percent
/ increments
) == 0)
555 snprintf(percentalert
, sizeof(percentalert
), "RebuildStarted");
557 snprintf(percentalert
, sizeof(percentalert
), "Rebuild%02d", mse
->percent
);
559 alert(percentalert
, dev
, NULL
, ainfo
);
562 if (mse
->percent
== RESYNC_NONE
&&
564 /* Rebuild/sync/whatever just finished.
565 * If there is a number in /mismatch_cnt,
566 * we should report that.
569 sysfs_read(-1, st
->devnum
, GET_MISMATCH
);
570 if (sra
&& sra
->mismatch_cnt
> 0) {
572 snprintf(cnt
, sizeof(cnt
),
573 " mismatches found: %d (on raid level %d)",
574 sra
->mismatch_cnt
, array
.level
);
575 alert("RebuildFinished", dev
, cnt
, ainfo
);
577 alert("RebuildFinished", dev
, NULL
, ainfo
);
581 st
->percent
= mse
->percent
;
583 remaining_disks
= array
.nr_disks
;
584 for (i
=0; i
<MAX_DISKS
&& remaining_disks
> 0;
586 mdu_disk_info_t disc
;
588 if (ioctl(fd
, GET_DISK_INFO
, &disc
) >= 0) {
589 info
[i
].state
= disc
.state
;
590 info
[i
].major
= disc
.major
;
591 info
[i
].minor
= disc
.minor
;
592 if (disc
.major
|| disc
.minor
)
595 info
[i
].major
= info
[i
].minor
= 0;
599 if (mse
->metadata_version
&&
600 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
601 is_subarray(mse
->metadata_version
+9))
603 devname2devnum(mse
->metadata_version
+10);
605 st
->parent_dev
= NoMdDev
;
606 if (st
->metadata
== NULL
&&
607 st
->parent_dev
== NoMdDev
)
608 st
->metadata
= super_by_fd(fd
, NULL
);
612 for (i
=0; i
<MAX_DISKS
; i
++) {
613 mdu_disk_info_t disc
= {0,0,0,0,0};
619 (info
[i
].major
|| info
[i
].minor
)) {
620 newstate
= info
[i
].state
;
621 dv
= map_dev_preferred(
622 info
[i
].major
, info
[i
].minor
, 1,
624 disc
.state
= newstate
;
625 disc
.major
= info
[i
].major
;
626 disc
.minor
= info
[i
].minor
;
628 newstate
= (1 << MD_DISK_REMOVED
);
630 if (dv
== NULL
&& st
->devid
[i
])
631 dv
= map_dev_preferred(
633 minor(st
->devid
[i
]), 1, prefer
);
634 change
= newstate
^ st
->devstate
[i
];
635 if (st
->utime
&& change
&& !st
->err
&& !new_array
) {
636 if ((st
->devstate
[i
]&change
)&(1<<MD_DISK_SYNC
))
637 alert("Fail", dev
, dv
, ainfo
);
638 else if ((newstate
& (1<<MD_DISK_FAULTY
)) &&
639 (disc
.major
|| disc
.minor
) &&
640 st
->devid
[i
] == makedev(disc
.major
, disc
.minor
))
641 alert("FailSpare", dev
, dv
, ainfo
);
642 else if ((newstate
&change
)&(1<<MD_DISK_SYNC
))
643 alert("SpareActive", dev
, dv
, ainfo
);
645 st
->devstate
[i
] = newstate
;
646 st
->devid
[i
] = makedev(disc
.major
, disc
.minor
);
648 st
->active
= array
.active_disks
;
649 st
->working
= array
.working_disks
;
650 st
->spare
= array
.spare_disks
;
651 st
->failed
= array
.failed_disks
;
652 st
->utime
= array
.utime
;
653 st
->raid
= array
.raid_disks
;
655 if ((st
->active
< st
->raid
) && st
->spare
== 0)
660 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
661 int test
, struct alert_info
*info
)
663 struct mdstat_ent
*mse
;
666 for (mse
=mdstat
; mse
; mse
=mse
->next
)
667 if (mse
->devnum
!= INT_MAX
&&
668 (!mse
->level
|| /* retrieve containers */
669 (strcmp(mse
->level
, "raid0") != 0 &&
670 strcmp(mse
->level
, "linear") != 0))
672 struct state
*st
= xcalloc(1, sizeof *st
);
673 mdu_array_info_t array
;
675 st
->devname
= xstrdup(get_md_name(mse
->devnum
));
676 if ((fd
= open(st
->devname
, O_RDONLY
)) < 0 ||
677 ioctl(fd
, GET_ARRAY_INFO
, &array
)< 0) {
679 if (fd
>=0) close(fd
);
680 put_md_name(st
->devname
);
683 st
->metadata
->ss
->free_super(st
->metadata
);
690 st
->next
= *statelist
;
692 st
->devnum
= mse
->devnum
;
693 st
->percent
= RESYNC_UNKNOWN
;
694 st
->expected_spares
= -1;
695 if (mse
->metadata_version
&&
696 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
697 is_subarray(mse
->metadata_version
+9))
699 devname2devnum(mse
->metadata_version
+10);
701 st
->parent_dev
= NoMdDev
;
704 alert("TestMessage", st
->devname
, NULL
, info
);
710 static int get_min_spare_size_required(struct state
*st
, unsigned long long *sizep
)
715 !st
->metadata
->ss
->min_acceptable_spare_size
) {
720 fd
= open(st
->devname
, O_RDONLY
);
723 if (st
->metadata
->ss
->external
)
724 st
->metadata
->ss
->load_container(st
->metadata
, fd
, st
->devname
);
726 st
->metadata
->ss
->load_super(st
->metadata
, fd
, st
->devname
);
728 if (!st
->metadata
->sb
)
730 *sizep
= st
->metadata
->ss
->min_acceptable_spare_size(st
->metadata
);
731 st
->metadata
->ss
->free_super(st
->metadata
);
736 static int check_donor(struct state
*from
, struct state
*to
)
743 /* Cannot move from a member */
747 for (sub
= from
->subarray
; sub
; sub
= sub
->subarray
)
748 /* If source array has degraded subarrays, don't
751 if (sub
->active
< sub
->raid
)
753 if (from
->metadata
->ss
->external
== 0)
754 if (from
->active
< from
->raid
)
756 if (from
->spare
<= 0)
761 static dev_t
choose_spare(struct state
*from
, struct state
*to
,
762 struct domainlist
*domlist
, unsigned long long min_size
)
767 for (d
= from
->raid
; !dev
&& d
< MAX_DISKS
; d
++) {
768 if (from
->devid
[d
] > 0 &&
769 from
->devstate
[d
] == 0) {
770 struct dev_policy
*pol
;
771 unsigned long long dev_size
;
773 if (to
->metadata
->ss
->external
&&
774 test_partition_from_id(from
->devid
[d
]))
778 dev_size_from_id(from
->devid
[d
], &dev_size
) &&
782 pol
= devnum_policy(from
->devid
[d
]);
783 if (from
->spare_group
)
784 pol_add(&pol
, pol_domain
,
785 from
->spare_group
, NULL
);
786 if (domain_test(domlist
, pol
, to
->metadata
->ss
->name
) == 1)
787 dev
= from
->devid
[d
];
788 dev_policy_free(pol
);
794 static dev_t
container_choose_spare(struct state
*from
, struct state
*to
,
795 struct domainlist
*domlist
,
796 unsigned long long min_size
, int active
)
798 /* This is similar to choose_spare, but we cannot trust devstate,
799 * so we need to read the metadata instead
802 struct supertype
*st
= from
->metadata
;
803 int fd
= open(from
->devname
, O_RDONLY
);
809 if (!st
->ss
->getinfo_super_disks
) {
814 err
= st
->ss
->load_container(st
, fd
, NULL
);
820 /* We must check if number of active disks has not increased
821 * since ioctl in main loop. mdmon may have added spare
822 * to subarray. If so we do not need to look for more spares
823 * so return non zero value */
826 list
= st
->ss
->getinfo_super_disks(st
);
828 st
->ss
->free_super(st
);
833 if (dp
->disk
.state
& (1<<MD_DISK_SYNC
) &&
834 !(dp
->disk
.state
& (1<<MD_DISK_FAULTY
)))
839 if (active
< active_cnt
) {
840 /* Spare just activated.*/
841 st
->ss
->free_super(st
);
846 /* We only need one spare so full list not needed */
847 list
= container_choose_spares(st
, min_size
, domlist
, from
->spare_group
,
848 to
->metadata
->ss
->name
, 1);
850 struct mdinfo
*disks
= list
->devs
;
852 dev
= makedev(disks
->disk
.major
, disks
->disk
.minor
);
855 st
->ss
->free_super(st
);
860 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
)
865 link_containers_with_subarrays(statelist
);
866 for (st
= statelist
; st
; st
= st
->next
)
867 if (st
->active
< st
->raid
&&
868 st
->spare
== 0 && !st
->err
) {
869 struct domainlist
*domlist
= NULL
;
871 struct state
*to
= st
;
872 unsigned long long min_size
;
874 if (to
->parent_dev
!= NoMdDev
&& !to
->parent
)
875 /* subarray monitored without parent container
876 * we can't move spares here */
880 /* member of a container */
883 if (get_min_spare_size_required(to
, &min_size
))
885 if (to
->metadata
->ss
->external
) {
886 /* We must make sure there is
887 * no suitable spare in container already.
888 * If there is we don't add more */
889 dev_t devid
= container_choose_spare(
890 to
, to
, NULL
, min_size
, st
->active
);
894 for (d
= 0; d
< MAX_DISKS
; d
++)
896 domainlist_add_dev(&domlist
,
898 to
->metadata
->ss
->name
);
900 domain_add(&domlist
, to
->spare_group
);
902 * No spare migration if the destination
903 * has no domain. Skip this array.
907 for (from
=statelist
; from
; from
=from
->next
) {
909 if (!check_donor(from
, to
))
911 if (from
->metadata
->ss
->external
)
912 devid
= container_choose_spare(
913 from
, to
, domlist
, min_size
, 0);
915 devid
= choose_spare(from
, to
, domlist
,
918 && move_spare(from
->devname
, to
->devname
, devid
)) {
919 alert("MoveSpare", to
->devname
, from
->devname
, info
);
923 domain_free(domlist
);
927 /* search the statelist to connect external
928 * metadata subarrays with their containers
929 * We always completely rebuild the tree from scratch as
930 * that is safest considering the possibility of entries
931 * disappearing or changing.
933 static void link_containers_with_subarrays(struct state
*list
)
937 for (st
= list
; st
; st
= st
->next
) {
941 for (st
= list
; st
; st
= st
->next
)
942 if (st
->parent_dev
!= NoMdDev
)
943 for (cont
= list
; cont
; cont
= cont
->next
)
945 cont
->parent_dev
== NoMdDev
&&
946 cont
->devnum
== st
->parent_dev
) {
948 st
->subarray
= cont
->subarray
;
954 /* Not really Monitor but ... */
961 if (stat(dev
, &stb
) != 0) {
962 pr_err("Cannot find %s: %s\n", dev
,
966 devnum
= stat2devnum(&stb
);
969 struct mdstat_ent
*ms
= mdstat_read(1, 0);
970 struct mdstat_ent
*e
;
972 for (e
=ms
; e
; e
=e
->next
)
973 if (e
->devnum
== devnum
)
976 if (!e
|| e
->percent
< 0) {
977 if (e
&& e
->metadata_version
&&
978 strncmp(e
->metadata_version
, "external:", 9) == 0) {
979 if (is_subarray(&e
->metadata_version
[9]))
980 ping_monitor(&e
->metadata_version
[9]);
982 ping_monitor_by_id(devnum
);
995 static char *clean_states
[] = {
996 "clear", "inactive", "readonly", "read-auto", "clean", NULL
};
998 int WaitClean(char *dev
, int sock
, int verbose
)
1005 fd
= open(dev
, O_RDONLY
);
1008 pr_err("Couldn't open %s: %s\n", dev
, strerror(errno
));
1012 devnum
= fd2devnum(fd
);
1013 mdi
= sysfs_read(fd
, devnum
, GET_VERSION
|GET_LEVEL
|GET_SAFEMODE
);
1016 pr_err("Failed to read sysfs attributes for "
1022 switch(mdi
->array
.level
) {
1024 case LEVEL_MULTIPATH
:
1026 /* safemode delay is irrelevant for these levels */
1030 /* for internal metadata the kernel handles the final clean
1031 * transition, containers can never be dirty
1033 if (!is_subarray(mdi
->text_version
))
1036 /* safemode disabled ? */
1037 if (mdi
->safe_mode_delay
== 0)
1041 int state_fd
= sysfs_open(fd2devnum(fd
), NULL
, "array_state");
1046 /* minimize the safe_mode_delay and prepare to wait up to 5s
1047 * for writes to quiesce
1049 sysfs_set_safemode(mdi
, 1);
1055 /* wait for array_state to be clean */
1057 rv
= read(state_fd
, buf
, sizeof(buf
));
1060 if (sysfs_match_word(buf
, clean_states
) <= 4)
1062 FD_SET(state_fd
, &fds
);
1063 rv
= select(state_fd
+ 1, NULL
, NULL
, &fds
, &tm
);
1064 if (rv
< 0 && errno
!= EINTR
)
1066 lseek(state_fd
, 0, SEEK_SET
);
1070 else if (fping_monitor(sock
) == 0 ||
1071 ping_monitor(mdi
->text_version
) == 0) {
1072 /* we need to ping to close the window between array
1073 * state transitioning to clean and the metadata being
1080 pr_err("Error waiting for %s to be clean\n",
1083 /* restore the original safe_mode_delay */
1084 sysfs_set_safemode(mdi
, mdi
->safe_mode_delay
);
1093 #endif /* MDASSEMBLE */