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 char devnm
[32]; /* to sync with mdstat info */
39 int active
, working
, failed
, spare
, raid
;
43 int devstate
[MAX_DISKS
];
44 dev_t devid
[MAX_DISKS
];
46 char parent_devnm
[32]; /* For subarray, devnm of parent.
49 struct supertype
*metadata
;
50 struct state
*subarray
;/* for a container it is a link to first subarray
51 * for a subarray it is a link to next subarray
52 * in the same container */
53 struct state
*parent
; /* for a subarray it is a link to its container
64 static int make_daemon(char *pidfile
);
65 static int check_one_sharer(int scan
);
66 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
);
67 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
68 int test
, struct alert_info
*info
,
69 int increments
, char *prefer
);
70 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
71 int test
, struct alert_info
*info
);
72 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
);
73 static void link_containers_with_subarrays(struct state
*list
);
75 int Monitor(struct mddev_dev
*devlist
,
76 char *mailaddr
, char *alert_cmd
,
78 int daemonise
, int oneshot
,
79 int dosyslog
, char *pidfile
, int increments
,
83 * Every few seconds, scan every md device looking for changes
84 * When a change is found, log it, possibly run the alert command,
85 * and possibly send Email
87 * For each array, we record:
89 * active/working/failed/spare drives
90 * State of each device.
91 * %rebuilt if rebuilding
93 * If the update time changes, check out all the data again
94 * It is possible that we cannot get the state of each device
95 * due to bugs in the md kernel module.
96 * We also read /proc/mdstat to get rebuild percent,
97 * and to get state on all active devices incase of kernel bug.
101 * An active device had Faulty set or Active/Sync removed
103 * A spare device had Faulty set
105 * An active device had a reverse transition
107 * percent went from -1 to +ve
109 * percent went from below to not-below NN%
111 * Couldn't access a device which was previously visible
113 * if we detect an array with active<raid and spare==0
114 * we look at other arrays that have same spare-group
115 * If we find one with active==raid and spare>0,
116 * and if we can get_disk_info and find a name
117 * Then we hot-remove and hot-add to the other array
119 * If devlist is NULL, then we can monitor everything because --scan
120 * was given. We get an initial list from config file and add anything
121 * that appears in /proc/mdstat
124 struct state
*statelist
= NULL
;
127 struct mdstat_ent
*mdstat
= NULL
;
128 char *mailfrom
= NULL
;
129 struct alert_info info
;
132 mailaddr
= conf_get_mailaddr();
133 if (mailaddr
&& ! c
->scan
)
134 pr_err("Monitor using email address \"%s\" from config file\n",
137 mailfrom
= conf_get_mailfrom();
140 alert_cmd
= conf_get_program();
141 if (alert_cmd
&& ! c
->scan
)
142 pr_err("Monitor using program \"%s\" from config file\n",
145 if (c
->scan
&& !mailaddr
&& !alert_cmd
&& !dosyslog
) {
146 pr_err("No mail address or alert command - not monitoring.\n");
149 info
.alert_cmd
= alert_cmd
;
150 info
.mailaddr
= mailaddr
;
151 info
.mailfrom
= mailfrom
;
152 info
.dosyslog
= dosyslog
;
155 int rv
= make_daemon(pidfile
);
161 if (check_one_sharer(c
->scan
))
164 if (devlist
== NULL
) {
165 struct mddev_ident
*mdlist
= conf_get_ident(NULL
);
166 for (; mdlist
; mdlist
=mdlist
->next
) {
168 if (mdlist
->devname
== NULL
)
170 if (strcasecmp(mdlist
->devname
, "<ignore>") == 0)
172 st
= xcalloc(1, sizeof *st
);
173 if (mdlist
->devname
[0] == '/')
174 st
->devname
= xstrdup(mdlist
->devname
);
176 st
->devname
= xmalloc(8+strlen(mdlist
->devname
)+1);
177 strcpy(strcpy(st
->devname
, "/dev/md/"),
180 st
->next
= statelist
;
182 st
->percent
= RESYNC_UNKNOWN
;
184 st
->expected_spares
= mdlist
->spare_disks
;
185 if (mdlist
->spare_group
)
186 st
->spare_group
= xstrdup(mdlist
->spare_group
);
190 struct mddev_dev
*dv
;
191 for (dv
=devlist
; dv
; dv
=dv
->next
) {
192 struct mddev_ident
*mdlist
= conf_get_ident(dv
->devname
);
193 struct state
*st
= xcalloc(1, sizeof *st
);
194 st
->devname
= xstrdup(dv
->devname
);
195 st
->next
= statelist
;
197 st
->percent
= RESYNC_UNKNOWN
;
198 st
->expected_spares
= -1;
200 st
->expected_spares
= mdlist
->spare_disks
;
201 if (mdlist
->spare_group
)
202 st
->spare_group
= xstrdup(mdlist
->spare_group
);
210 struct state
*st
, **stp
;
215 mdstat
= mdstat_read(oneshot
?0:1, 0);
219 for (st
=statelist
; st
; st
=st
->next
)
220 if (check_array(st
, mdstat
, c
->test
, &info
,
221 increments
, c
->prefer
))
224 /* now check if there are any new devices found in mdstat */
226 new_found
= add_new_arrays(mdstat
, &statelist
, c
->test
,
229 /* If an array has active < raid && spare == 0 && spare_group != NULL
230 * Look for another array with spare > 0 and active == raid and same spare_group
231 * if found, choose a device and hotremove/hotadd
233 if (share
&& anydegraded
)
234 try_spare_migration(statelist
, &info
);
239 mdstat_wait(c
->delay
);
243 for (stp
= &statelist
; (st
= *stp
) != NULL
; ) {
244 if (st
->from_auto
&& st
->err
> 5) {
247 free(st
->spare_group
);
253 for (st2
= statelist
; st2
; st2
= statelist
) {
254 statelist
= st2
->next
;
263 static int make_daemon(char *pidfile
)
266 * -1 in the forked daemon
269 * so a none-negative becomes the exit code.
277 pid_file
=fopen(pidfile
, "w");
279 perror("cannot create pid file");
281 fprintf(pid_file
,"%d\n", pid
);
292 open("/dev/null", O_RDWR
);
299 static int check_one_sharer(int scan
)
306 sprintf(path
, "%s/autorebuild.pid", MDMON_DIR
);
307 fp
= fopen(path
, "r");
309 if (fscanf(fp
, "%d", &pid
) != 1)
311 sprintf(dir
, "/proc/%d", pid
);
312 rv
= stat(dir
, &buf
);
315 pr_err("Only one autorebuild process allowed in scan mode, aborting\n");
319 pr_err("Warning: One autorebuild process already running.\n");
325 if (mkdir(MDMON_DIR
, S_IRWXU
) < 0 &&
327 pr_err("Can't create autorebuild.pid file\n");
329 fp
= fopen(path
, "w");
331 pr_err("Cannot create autorebuild.pidfile\n");
334 fprintf(fp
, "%d\n", pid
);
342 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
)
346 if (!info
->alert_cmd
&& !info
->mailaddr
&& !info
->dosyslog
) {
347 time_t now
= time(0);
349 printf("%1.15s: %s on %s %s\n", ctime(&now
)+4, event
, dev
, disc
?disc
:"unknown device");
351 if (info
->alert_cmd
) {
355 waitpid(pid
, NULL
, 0);
360 execl(info
->alert_cmd
, info
->alert_cmd
,
361 event
, dev
, disc
, NULL
);
365 if (info
->mailaddr
&&
366 (strncmp(event
, "Fail", 4)==0 ||
367 strncmp(event
, "Test", 4)==0 ||
368 strncmp(event
, "Spares", 6)==0 ||
369 strncmp(event
, "Degrade", 7)==0)) {
370 FILE *mp
= popen(Sendmail
, "w");
374 gethostname(hname
, sizeof(hname
));
375 signal(SIGPIPE
, SIG_IGN
);
377 fprintf(mp
, "From: %s\n", info
->mailfrom
);
379 fprintf(mp
, "From: %s monitoring <root>\n", Name
);
380 fprintf(mp
, "To: %s\n", info
->mailaddr
);
381 fprintf(mp
, "Subject: %s event on %s:%s\n\n",
385 "This is an automatically generated mail message from %s\n", Name
);
386 fprintf(mp
, "running on %s\n\n", hname
);
389 "A %s event had been detected on md device %s.\n\n", event
, dev
);
391 if (disc
&& disc
[0] != ' ')
393 "It could be related to component device %s.\n\n", disc
);
394 if (disc
&& disc
[0] == ' ')
395 fprintf(mp
, "Extra information:%s.\n\n", disc
);
397 fprintf(mp
, "Faithfully yours, etc.\n");
399 mdstat
= fopen("/proc/mdstat", "r");
404 "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
405 while ( (n
=fread(buf
, 1, sizeof(buf
), mdstat
)) > 0)
406 n
=fwrite(buf
, 1, n
, mp
);
413 /* log the event to syslog maybe */
414 if (info
->dosyslog
) {
415 /* Log at a different severity depending on the event.
417 * These are the critical events: */
418 if (strncmp(event
, "Fail", 4)==0 ||
419 strncmp(event
, "Degrade", 7)==0 ||
420 strncmp(event
, "DeviceDisappeared", 17)==0)
422 /* Good to know about, but are not failures: */
423 else if (strncmp(event
, "Rebuild", 7)==0 ||
424 strncmp(event
, "MoveSpare", 9)==0 ||
425 strncmp(event
, "Spares", 6) != 0)
426 priority
= LOG_WARNING
;
427 /* Everything else: */
431 if (disc
&& disc
[0] != ' ')
433 "%s event detected on md device %s, component device %s", event
, dev
, disc
);
436 "%s event detected on md device %s: %s",
440 "%s event detected on md device %s",
445 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
446 int test
, struct alert_info
*ainfo
,
447 int increments
, char *prefer
)
449 /* Update the state 'st' to reflect any changes shown in mdstat,
450 * or found by directly examining the array, and return
451 * '1' if the array is degraded, or '0' if it is optimal (or dead).
453 struct { int state
, major
, minor
; } info
[MAX_DISKS
];
454 mdu_array_info_t array
;
455 struct mdstat_ent
*mse
= NULL
, *mse2
;
456 char *dev
= st
->devname
;
464 alert("TestMessage", dev
, NULL
, ainfo
);
466 fd
= open("/sys/block", O_RDONLY
|O_DIRECTORY
);
468 /* Don't open the device unless it is present and
473 fd
= sysfs_open(st
->devnm
, NULL
, "array_state");
475 read(fd
, buf
, 10) < 5 ||
476 strncmp(buf
,"clear",5) == 0 ||
477 strncmp(buf
,"inact",5) == 0) {
480 fd
= sysfs_open(st
->devnm
, NULL
, "level");
481 if (fd
< 0 || read(fd
, buf
, 10) != 0) {
485 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
492 fd
= open(dev
, O_RDONLY
);
495 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
499 fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
500 if (md_get_array_info(fd
, &array
) < 0) {
502 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
507 /* It's much easier to list what array levels can't
508 * have a device disappear than all of them that can
510 if (array
.level
== 0 || array
.level
== -1) {
511 if (!st
->err
&& !st
->from_config
)
512 alert("DeviceDisappeared", dev
, " Wrong-Level", ainfo
);
517 if (st
->devnm
[0] == 0)
518 strcpy(st
->devnm
, fd2devnm(fd
));
520 for (mse2
= mdstat
; mse2
; mse2
=mse2
->next
)
521 if (strcmp(mse2
->devnm
, st
->devnm
) == 0) {
522 mse2
->devnm
[0] = 0; /* flag it as "used" */
527 /* duplicated array in statelist
528 * or re-created after reading mdstat*/
533 /* this array is in /proc/mdstat */
534 if (array
.utime
== 0)
535 /* external arrays don't update utime, so
536 * just make sure it is always different. */
537 array
.utime
= st
->utime
+ 1;;
540 /* New array appeared where previously had an error */
542 st
->percent
= RESYNC_NONE
;
544 alert("NewArray", st
->devname
, NULL
, ainfo
);
547 if (st
->utime
== array
.utime
&&
548 st
->failed
== array
.failed_disks
&&
549 st
->working
== array
.working_disks
&&
550 st
->spare
== array
.spare_disks
&&
552 mse
->percent
== st
->percent
555 if ((st
->active
< st
->raid
) && st
->spare
== 0)
560 if (st
->utime
== 0 && /* new array */
561 mse
->pattern
&& strchr(mse
->pattern
, '_') /* degraded */
563 alert("DegradedArray", dev
, NULL
, ainfo
);
565 if (st
->utime
== 0 && /* new array */
566 st
->expected_spares
> 0 &&
567 array
.spare_disks
< st
->expected_spares
)
568 alert("SparesMissing", dev
, NULL
, ainfo
);
569 if (st
->percent
< 0 && st
->percent
!= RESYNC_UNKNOWN
&&
571 alert("RebuildStarted", dev
, NULL
, ainfo
);
572 if (st
->percent
>= 0 &&
574 (mse
->percent
/ increments
) > (st
->percent
/ increments
)) {
575 char percentalert
[15]; // "RebuildNN" (10 chars) or "RebuildStarted" (15 chars)
577 if((mse
->percent
/ increments
) == 0)
578 snprintf(percentalert
, sizeof(percentalert
), "RebuildStarted");
580 snprintf(percentalert
, sizeof(percentalert
), "Rebuild%02d", mse
->percent
);
582 alert(percentalert
, dev
, NULL
, ainfo
);
585 if (mse
->percent
== RESYNC_NONE
&&
587 /* Rebuild/sync/whatever just finished.
588 * If there is a number in /mismatch_cnt,
589 * we should report that.
592 sysfs_read(-1, st
->devnm
, GET_MISMATCH
);
593 if (sra
&& sra
->mismatch_cnt
> 0) {
595 snprintf(cnt
, sizeof(cnt
),
596 " mismatches found: %d (on raid level %d)",
597 sra
->mismatch_cnt
, array
.level
);
598 alert("RebuildFinished", dev
, cnt
, ainfo
);
600 alert("RebuildFinished", dev
, NULL
, ainfo
);
604 st
->percent
= mse
->percent
;
606 remaining_disks
= array
.nr_disks
;
607 for (i
=0; i
<MAX_DISKS
&& remaining_disks
> 0;
609 mdu_disk_info_t disc
;
611 if (md_get_disk_info(fd
, &disc
) >= 0) {
612 info
[i
].state
= disc
.state
;
613 info
[i
].major
= disc
.major
;
614 info
[i
].minor
= disc
.minor
;
615 if (disc
.major
|| disc
.minor
)
618 info
[i
].major
= info
[i
].minor
= 0;
622 if (mse
->metadata_version
&&
623 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
624 is_subarray(mse
->metadata_version
+9)) {
626 strcpy(st
->parent_devnm
,
627 mse
->metadata_version
+10);
628 sl
= strchr(st
->parent_devnm
, '/');
632 st
->parent_devnm
[0] = 0;
633 if (st
->metadata
== NULL
&&
634 st
->parent_devnm
[0] == 0)
635 st
->metadata
= super_by_fd(fd
, NULL
);
639 for (i
=0; i
<MAX_DISKS
; i
++) {
640 mdu_disk_info_t disc
= {0,0,0,0,0};
646 (info
[i
].major
|| info
[i
].minor
)) {
647 newstate
= info
[i
].state
;
648 dv
= map_dev_preferred(
649 info
[i
].major
, info
[i
].minor
, 1,
651 disc
.state
= newstate
;
652 disc
.major
= info
[i
].major
;
653 disc
.minor
= info
[i
].minor
;
655 newstate
= (1 << MD_DISK_REMOVED
);
657 if (dv
== NULL
&& st
->devid
[i
])
658 dv
= map_dev_preferred(
660 minor(st
->devid
[i
]), 1, prefer
);
661 change
= newstate
^ st
->devstate
[i
];
662 if (st
->utime
&& change
&& !st
->err
&& !new_array
) {
663 if ((st
->devstate
[i
]&change
)&(1<<MD_DISK_SYNC
))
664 alert("Fail", dev
, dv
, ainfo
);
665 else if ((newstate
& (1<<MD_DISK_FAULTY
)) &&
666 (disc
.major
|| disc
.minor
) &&
667 st
->devid
[i
] == makedev(disc
.major
, disc
.minor
))
668 alert("FailSpare", dev
, dv
, ainfo
);
669 else if ((newstate
&change
)&(1<<MD_DISK_SYNC
))
670 alert("SpareActive", dev
, dv
, ainfo
);
672 st
->devstate
[i
] = newstate
;
673 st
->devid
[i
] = makedev(disc
.major
, disc
.minor
);
675 st
->active
= array
.active_disks
;
676 st
->working
= array
.working_disks
;
677 st
->spare
= array
.spare_disks
;
678 st
->failed
= array
.failed_disks
;
679 st
->utime
= array
.utime
;
680 st
->raid
= array
.raid_disks
;
682 if ((st
->active
< st
->raid
) && st
->spare
== 0)
687 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
688 int test
, struct alert_info
*info
)
690 struct mdstat_ent
*mse
;
694 for (mse
=mdstat
; mse
; mse
=mse
->next
)
696 (!mse
->level
|| /* retrieve containers */
697 (strcmp(mse
->level
, "raid0") != 0 &&
698 strcmp(mse
->level
, "linear") != 0))
700 struct state
*st
= xcalloc(1, sizeof *st
);
701 mdu_array_info_t array
;
704 name
= get_md_name(mse
->devnm
);
710 st
->devname
= xstrdup(name
);
711 if ((fd
= open(st
->devname
, O_RDONLY
)) < 0 ||
712 md_get_array_info(fd
, &array
) < 0) {
716 put_md_name(st
->devname
);
719 st
->metadata
->ss
->free_super(st
->metadata
);
726 st
->next
= *statelist
;
729 strcpy(st
->devnm
, mse
->devnm
);
730 st
->percent
= RESYNC_UNKNOWN
;
731 st
->expected_spares
= -1;
732 if (mse
->metadata_version
&&
733 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
734 is_subarray(mse
->metadata_version
+9)) {
736 strcpy(st
->parent_devnm
,
737 mse
->metadata_version
+10);
738 sl
= strchr(st
->parent_devnm
, '/');
741 st
->parent_devnm
[0] = 0;
744 alert("TestMessage", st
->devname
, NULL
, info
);
750 static int get_min_spare_size_required(struct state
*st
, unsigned long long *sizep
)
755 !st
->metadata
->ss
->min_acceptable_spare_size
) {
760 fd
= open(st
->devname
, O_RDONLY
);
763 if (st
->metadata
->ss
->external
)
764 st
->metadata
->ss
->load_container(st
->metadata
, fd
, st
->devname
);
766 st
->metadata
->ss
->load_super(st
->metadata
, fd
, st
->devname
);
768 if (!st
->metadata
->sb
)
770 *sizep
= st
->metadata
->ss
->min_acceptable_spare_size(st
->metadata
);
771 st
->metadata
->ss
->free_super(st
->metadata
);
776 static int check_donor(struct state
*from
, struct state
*to
)
783 /* Cannot move from a member */
787 for (sub
= from
->subarray
; sub
; sub
= sub
->subarray
)
788 /* If source array has degraded subarrays, don't
791 if (sub
->active
< sub
->raid
)
793 if (from
->metadata
->ss
->external
== 0)
794 if (from
->active
< from
->raid
)
796 if (from
->spare
<= 0)
801 static dev_t
choose_spare(struct state
*from
, struct state
*to
,
802 struct domainlist
*domlist
, unsigned long long min_size
)
807 for (d
= from
->raid
; !dev
&& d
< MAX_DISKS
; d
++) {
808 if (from
->devid
[d
] > 0 &&
809 from
->devstate
[d
] == 0) {
810 struct dev_policy
*pol
;
811 unsigned long long dev_size
;
813 if (to
->metadata
->ss
->external
&&
814 test_partition_from_id(from
->devid
[d
]))
818 dev_size_from_id(from
->devid
[d
], &dev_size
) &&
822 pol
= devid_policy(from
->devid
[d
]);
823 if (from
->spare_group
)
824 pol_add(&pol
, pol_domain
,
825 from
->spare_group
, NULL
);
826 if (domain_test(domlist
, pol
, to
->metadata
->ss
->name
) == 1)
827 dev
= from
->devid
[d
];
828 dev_policy_free(pol
);
834 static dev_t
container_choose_spare(struct state
*from
, struct state
*to
,
835 struct domainlist
*domlist
,
836 unsigned long long min_size
, int active
)
838 /* This is similar to choose_spare, but we cannot trust devstate,
839 * so we need to read the metadata instead
842 struct supertype
*st
= from
->metadata
;
843 int fd
= open(from
->devname
, O_RDONLY
);
849 if (!st
->ss
->getinfo_super_disks
) {
854 err
= st
->ss
->load_container(st
, fd
, NULL
);
860 /* We must check if number of active disks has not increased
861 * since ioctl in main loop. mdmon may have added spare
862 * to subarray. If so we do not need to look for more spares
863 * so return non zero value */
866 list
= st
->ss
->getinfo_super_disks(st
);
868 st
->ss
->free_super(st
);
873 if (dp
->disk
.state
& (1<<MD_DISK_SYNC
) &&
874 !(dp
->disk
.state
& (1<<MD_DISK_FAULTY
)))
879 if (active
< active_cnt
) {
880 /* Spare just activated.*/
881 st
->ss
->free_super(st
);
886 /* We only need one spare so full list not needed */
887 list
= container_choose_spares(st
, min_size
, domlist
, from
->spare_group
,
888 to
->metadata
->ss
->name
, 1);
890 struct mdinfo
*disks
= list
->devs
;
892 dev
= makedev(disks
->disk
.major
, disks
->disk
.minor
);
895 st
->ss
->free_super(st
);
899 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
)
904 link_containers_with_subarrays(statelist
);
905 for (st
= statelist
; st
; st
= st
->next
)
906 if (st
->active
< st
->raid
&&
907 st
->spare
== 0 && !st
->err
) {
908 struct domainlist
*domlist
= NULL
;
910 struct state
*to
= st
;
911 unsigned long long min_size
;
913 if (to
->parent_devnm
[0] && !to
->parent
)
914 /* subarray monitored without parent container
915 * we can't move spares here */
919 /* member of a container */
922 if (get_min_spare_size_required(to
, &min_size
))
924 if (to
->metadata
->ss
->external
) {
925 /* We must make sure there is
926 * no suitable spare in container already.
927 * If there is we don't add more */
928 dev_t devid
= container_choose_spare(
929 to
, to
, NULL
, min_size
, st
->active
);
933 for (d
= 0; d
< MAX_DISKS
; d
++)
935 domainlist_add_dev(&domlist
,
937 to
->metadata
->ss
->name
);
939 domain_add(&domlist
, to
->spare_group
);
941 * No spare migration if the destination
942 * has no domain. Skip this array.
946 for (from
=statelist
; from
; from
=from
->next
) {
948 if (!check_donor(from
, to
))
950 if (from
->metadata
->ss
->external
)
951 devid
= container_choose_spare(
952 from
, to
, domlist
, min_size
, 0);
954 devid
= choose_spare(from
, to
, domlist
,
957 && move_spare(from
->devname
, to
->devname
, devid
)) {
958 alert("MoveSpare", to
->devname
, from
->devname
, info
);
962 domain_free(domlist
);
966 /* search the statelist to connect external
967 * metadata subarrays with their containers
968 * We always completely rebuild the tree from scratch as
969 * that is safest considering the possibility of entries
970 * disappearing or changing.
972 static void link_containers_with_subarrays(struct state
*list
)
976 for (st
= list
; st
; st
= st
->next
) {
980 for (st
= list
; st
; st
= st
->next
)
981 if (st
->parent_devnm
[0])
982 for (cont
= list
; cont
; cont
= cont
->next
)
984 cont
->parent_devnm
[0] == 0 &&
985 strcmp(cont
->devnm
, st
->parent_devnm
) == 0) {
987 st
->subarray
= cont
->subarray
;
993 /* Not really Monitor but ... */
1000 int frozen_remaining
= 3;
1002 if (stat(dev
, &stb
) != 0) {
1003 pr_err("Cannot find %s: %s\n", dev
,
1007 tmp
= stat2devnm(&stb
);
1009 pr_err("%s is not a block device.\n", dev
);
1015 struct mdstat_ent
*ms
= mdstat_read(1, 0);
1016 struct mdstat_ent
*e
;
1018 for (e
=ms
; e
; e
=e
->next
)
1019 if (strcmp(e
->devnm
, devnm
) == 0)
1022 if (e
&& e
->percent
== RESYNC_NONE
) {
1023 /* We could be in the brief pause before something
1024 * starts. /proc/mdstat doesn't show that, but
1029 sysfs_init(&mdi
, -1, devnm
);
1030 if (sysfs_get_str(&mdi
, NULL
, "sync_action",
1032 strcmp(buf
,"idle\n") != 0) {
1033 e
->percent
= RESYNC_UNKNOWN
;
1034 if (strcmp(buf
, "frozen\n") == 0) {
1035 if (frozen_remaining
== 0)
1036 e
->percent
= RESYNC_NONE
;
1038 frozen_remaining
-= 1;
1042 if (!e
|| e
->percent
== RESYNC_NONE
) {
1043 if (e
&& e
->metadata_version
&&
1044 strncmp(e
->metadata_version
, "external:", 9) == 0) {
1045 if (is_subarray(&e
->metadata_version
[9]))
1046 ping_monitor(&e
->metadata_version
[9]);
1048 ping_monitor(devnm
);
1061 static char *clean_states
[] = {
1062 "clear", "inactive", "readonly", "read-auto", "clean", NULL
};
1064 int WaitClean(char *dev
, int sock
, int verbose
)
1071 fd
= open(dev
, O_RDONLY
);
1074 pr_err("Couldn't open %s: %s\n", dev
, strerror(errno
));
1078 strcpy(devnm
, fd2devnm(fd
));
1079 mdi
= sysfs_read(fd
, devnm
, GET_VERSION
|GET_LEVEL
|GET_SAFEMODE
);
1082 pr_err("Failed to read sysfs attributes for %s\n", dev
);
1087 switch(mdi
->array
.level
) {
1089 case LEVEL_MULTIPATH
:
1091 /* safemode delay is irrelevant for these levels */
1095 /* for internal metadata the kernel handles the final clean
1096 * transition, containers can never be dirty
1098 if (!is_subarray(mdi
->text_version
))
1101 /* safemode disabled ? */
1102 if (mdi
->safe_mode_delay
== 0)
1106 int state_fd
= sysfs_open(fd2devnm(fd
), NULL
, "array_state");
1110 /* minimize the safe_mode_delay and prepare to wait up to 5s
1111 * for writes to quiesce
1113 sysfs_set_safemode(mdi
, 1);
1115 /* wait for array_state to be clean */
1117 rv
= read(state_fd
, buf
, sizeof(buf
));
1120 if (sysfs_match_word(buf
, clean_states
) <= 4)
1122 rv
= sysfs_wait(state_fd
, &delay
);
1123 if (rv
< 0 && errno
!= EINTR
)
1125 lseek(state_fd
, 0, SEEK_SET
);
1129 else if (fping_monitor(sock
) == 0 ||
1130 ping_monitor(mdi
->text_version
) == 0) {
1131 /* we need to ping to close the window between array
1132 * state transitioning to clean and the metadata being
1139 pr_err("Error waiting for %s to be clean\n",
1142 /* restore the original safe_mode_delay */
1143 sysfs_set_safemode(mdi
, mdi
->safe_mode_delay
);
1152 #endif /* MDASSEMBLE */