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>
33 /* The largest number of disks current arrays can manage is 384
34 * This really should be dynamically, but that will have to wait
35 * At least it isn't MD_SB_DISKS.
40 int devnum
; /* to sync with mdstat info */
44 int active
, working
, failed
, spare
, raid
;
46 int devstate
[MaxDisks
];
47 dev_t devid
[MaxDisks
];
49 int parent_dev
; /* For subarray, devnum of parent.
52 struct supertype
*metadata
;
53 struct state
*subarray
;/* for a container it is a link to first subarray
54 * for a subarray it is a link to next subarray
55 * in the same container */
56 struct state
*parent
; /* for a subarray it is a link to its container
67 static int make_daemon(char *pidfile
);
68 static int check_one_sharer(int scan
);
69 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
);
70 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
71 int test
, struct alert_info
*info
,
73 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
74 int test
, struct alert_info
*info
);
75 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
);
76 static void link_containers_with_subarrays(struct state
*list
);
78 int Monitor(struct mddev_dev
*devlist
,
79 char *mailaddr
, char *alert_cmd
,
80 int period
, int daemonise
, int scan
, int oneshot
,
81 int dosyslog
, int test
, char *pidfile
, int increments
,
85 * Every few seconds, scan every md device looking for changes
86 * When a change is found, log it, possibly run the alert command,
87 * and possibly send Email
89 * For each array, we record:
91 * active/working/failed/spare drives
92 * State of each device.
93 * %rebuilt if rebuilding
95 * If the update time changes, check out all the data again
96 * It is possible that we cannot get the state of each device
97 * due to bugs in the md kernel module.
98 * We also read /proc/mdstat to get rebuild percent,
99 * and to get state on all active devices incase of kernel bug.
103 * An active device had Faulty set or Active/Sync removed
105 * A spare device had Faulty set
107 * An active device had a reverse transition
109 * percent went from -1 to +ve
111 * percent went from below to not-below NN%
113 * Couldn't access a device which was previously visible
115 * if we detect an array with active<raid and spare==0
116 * we look at other arrays that have same spare-group
117 * If we find one with active==raid and spare>0,
118 * and if we can get_disk_info and find a name
119 * Then we hot-remove and hot-add to the other array
121 * If devlist is NULL, then we can monitor everything because --scan
122 * was given. We get an initial list from config file and add anything
123 * that appears in /proc/mdstat
126 struct state
*statelist
= NULL
;
129 struct mdstat_ent
*mdstat
= NULL
;
130 char *mailfrom
= NULL
;
131 struct alert_info info
;
134 mailaddr
= conf_get_mailaddr();
135 if (mailaddr
&& ! scan
)
136 fprintf(stderr
, Name
": Monitor using email address \"%s\" from config file\n",
139 mailfrom
= conf_get_mailfrom();
142 alert_cmd
= conf_get_program();
143 if (alert_cmd
&& ! scan
)
144 fprintf(stderr
, Name
": Monitor using program \"%s\" from config file\n",
147 if (scan
&& !mailaddr
&& !alert_cmd
&& !dosyslog
) {
148 fprintf(stderr
, Name
": No mail address or alert command - not monitoring.\n");
151 info
.alert_cmd
= alert_cmd
;
152 info
.mailaddr
= mailaddr
;
153 info
.mailfrom
= mailfrom
;
154 info
.dosyslog
= dosyslog
;
157 int rv
= make_daemon(pidfile
);
163 if (check_one_sharer(scan
))
166 if (devlist
== NULL
) {
167 struct mddev_ident
*mdlist
= conf_get_ident(NULL
);
168 for (; mdlist
; mdlist
=mdlist
->next
) {
170 if (mdlist
->devname
== NULL
)
172 if (strcasecmp(mdlist
->devname
, "<ignore>") == 0)
174 st
= calloc(1, sizeof *st
);
177 if (mdlist
->devname
[0] == '/')
178 st
->devname
= strdup(mdlist
->devname
);
180 st
->devname
= malloc(8+strlen(mdlist
->devname
)+1);
181 strcpy(strcpy(st
->devname
, "/dev/md/"),
184 st
->next
= statelist
;
185 st
->devnum
= INT_MAX
;
187 st
->expected_spares
= mdlist
->spare_disks
;
188 if (mdlist
->spare_group
)
189 st
->spare_group
= strdup(mdlist
->spare_group
);
193 struct mddev_dev
*dv
;
194 for (dv
=devlist
; dv
; dv
=dv
->next
) {
195 struct mddev_ident
*mdlist
= conf_get_ident(dv
->devname
);
196 struct state
*st
= calloc(1, sizeof *st
);
199 st
->devname
= strdup(dv
->devname
);
200 st
->next
= statelist
;
201 st
->devnum
= INT_MAX
;
203 st
->expected_spares
= -1;
205 st
->expected_spares
= mdlist
->spare_disks
;
206 if (mdlist
->spare_group
)
207 st
->spare_group
= strdup(mdlist
->spare_group
);
221 mdstat
= mdstat_read(oneshot
?0:1, 0);
223 for (st
=statelist
; st
; st
=st
->next
)
224 if (check_array(st
, mdstat
, test
, &info
, increments
))
227 /* now check if there are any new devices found in mdstat */
229 new_found
= add_new_arrays(mdstat
, &statelist
, test
,
232 /* If an array has active < raid && spare == 0 && spare_group != NULL
233 * Look for another array with spare > 0 and active == raid and same spare_group
234 * if found, choose a device and hotremove/hotadd
236 if (share
&& anydegraded
)
237 try_spare_migration(statelist
, &info
);
246 for (st2
= statelist
; st2
; st2
= statelist
) {
247 statelist
= st2
->next
;
256 static int make_daemon(char *pidfile
)
259 * -1 in the forked daemon
262 * so a none-negative becomes the exit code.
270 pid_file
=fopen(pidfile
, "w");
272 perror("cannot create pid file");
274 fprintf(pid_file
,"%d\n", pid
);
285 open("/dev/null", O_RDWR
);
292 static int check_one_sharer(int scan
)
299 sprintf(path
, "%s/autorebuild.pid", MDMON_DIR
);
300 fp
= fopen(path
, "r");
302 if (fscanf(fp
, "%d", &pid
) != 1)
304 sprintf(dir
, "/proc/%d", pid
);
305 rv
= stat(dir
, &buf
);
308 fprintf(stderr
, Name
": Only one "
309 "autorebuild process allowed"
310 " in scan mode, aborting\n");
314 fprintf(stderr
, Name
": Warning: One"
315 " autorebuild process already"
322 if (mkdir(MDMON_DIR
, S_IRWXU
) < 0 &&
324 fprintf(stderr
, Name
": Can't create "
325 "autorebuild.pid file\n");
327 fp
= fopen(path
, "w");
329 fprintf(stderr
, Name
": Cannot create"
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: " Name
" monitoring <root>\n");
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"
386 " mail message from " Name
"\n");
387 fprintf(mp
, "running on %s\n\n", hname
);
390 "A %s event had been detected on"
391 " md device %s.\n\n", event
, dev
);
393 if (disc
&& disc
[0] != ' ')
395 "It could be related to"
396 " component device %s.\n\n", disc
);
397 if (disc
&& disc
[0] == ' ')
398 fprintf(mp
, "Extra information:%s.\n\n", disc
);
400 fprintf(mp
, "Faithfully yours, etc.\n");
402 mdstat
= fopen("/proc/mdstat", "r");
407 "\nP.S. The /proc/mdstat file"
408 " currently contains the following:\n\n");
409 while ( (n
=fread(buf
, 1, sizeof(buf
), mdstat
)) > 0)
410 n
=fwrite(buf
, 1, n
, mp
);
417 /* log the event to syslog maybe */
418 if (info
->dosyslog
) {
419 /* Log at a different severity depending on the event.
421 * These are the critical events: */
422 if (strncmp(event
, "Fail", 4)==0 ||
423 strncmp(event
, "Degrade", 7)==0 ||
424 strncmp(event
, "DeviceDisappeared", 17)==0)
426 /* Good to know about, but are not failures: */
427 else if (strncmp(event
, "Rebuild", 7)==0 ||
428 strncmp(event
, "MoveSpare", 9)==0 ||
429 strncmp(event
, "Spares", 6) != 0)
430 priority
= LOG_WARNING
;
431 /* Everything else: */
437 "%s event detected on md device %s,"
438 " component device %s", event
, dev
, disc
);
441 "%s event detected on md device %s",
446 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
447 int test
, struct alert_info
*ainfo
,
450 /* Update the state 'st' to reflect any changes shown in mdstat,
451 * or found by directly examining the array, and return
452 * '1' if the array is degraded, or '0' if it is optimal (or dead).
454 struct { int state
, major
, minor
; } info
[MaxDisks
];
455 mdu_array_info_t array
;
456 struct mdstat_ent
*mse
= NULL
, *mse2
;
457 char *dev
= st
->devname
;
464 alert("TestMessage", dev
, NULL
, ainfo
);
465 fd
= open(dev
, O_RDONLY
);
468 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
472 fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
473 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)<0) {
475 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
480 /* It's much easier to list what array levels can't
481 * have a device disappear than all of them that can
483 if (array
.level
== 0 || array
.level
== -1) {
485 alert("DeviceDisappeared", dev
, "Wrong-Level", ainfo
);
490 if (st
->devnum
== INT_MAX
) {
492 if (fstat(fd
, &stb
) == 0 &&
493 (S_IFMT
&stb
.st_mode
)==S_IFBLK
) {
494 if (major(stb
.st_rdev
) == MD_MAJOR
)
495 st
->devnum
= minor(stb
.st_rdev
);
497 st
->devnum
= -1- (minor(stb
.st_rdev
)>>6);
501 for (mse2
= mdstat
; mse2
; mse2
=mse2
->next
)
502 if (mse2
->devnum
== st
->devnum
) {
503 mse2
->devnum
= INT_MAX
; /* flag it as "used" */
508 /* duplicated array in statelist
509 * or re-created after reading mdstat*/
514 /* this array is in /proc/mdstat */
515 if (array
.utime
== 0)
516 /* external arrays don't update utime, so
517 * just make sure it is always different. */
518 array
.utime
= st
->utime
+ 1;;
520 if (st
->utime
== array
.utime
&&
521 st
->failed
== array
.failed_disks
&&
522 st
->working
== array
.working_disks
&&
523 st
->spare
== array
.spare_disks
&&
525 mse
->percent
== st
->percent
529 if ((st
->active
< st
->raid
) && st
->spare
== 0)
534 if (st
->utime
== 0 && /* new array */
535 mse
->pattern
&& strchr(mse
->pattern
, '_') /* degraded */
537 alert("DegradedArray", dev
, NULL
, ainfo
);
539 if (st
->utime
== 0 && /* new array */
540 st
->expected_spares
> 0 &&
541 array
.spare_disks
< st
->expected_spares
)
542 alert("SparesMissing", dev
, NULL
, ainfo
);
543 if (st
->percent
== -1 &&
545 alert("RebuildStarted", dev
, NULL
, ainfo
);
546 if (st
->percent
>= 0 &&
548 (mse
->percent
/ increments
) > (st
->percent
/ increments
)) {
549 char percentalert
[15]; // "RebuildNN" (10 chars) or "RebuildStarted" (15 chars)
551 if((mse
->percent
/ increments
) == 0)
552 snprintf(percentalert
, sizeof(percentalert
), "RebuildStarted");
554 snprintf(percentalert
, sizeof(percentalert
), "Rebuild%02d", mse
->percent
);
556 alert(percentalert
, dev
, NULL
, ainfo
);
559 if (mse
->percent
== -1 &&
561 /* Rebuild/sync/whatever just finished.
562 * If there is a number in /mismatch_cnt,
563 * we should report that.
566 sysfs_read(-1, st
->devnum
, GET_MISMATCH
);
567 if (sra
&& sra
->mismatch_cnt
> 0) {
569 snprintf(cnt
, sizeof(cnt
),
570 " mismatches found: %d (on raid level %d)",
571 sra
->mismatch_cnt
, array
.level
);
572 alert("RebuildFinished", dev
, cnt
, ainfo
);
574 alert("RebuildFinished", dev
, NULL
, ainfo
);
578 st
->percent
= mse
->percent
;
580 remaining_disks
= array
.nr_disks
;
581 for (i
=0; i
<MaxDisks
&& remaining_disks
> 0;
583 mdu_disk_info_t disc
;
585 if (ioctl(fd
, GET_DISK_INFO
, &disc
) >= 0) {
586 info
[i
].state
= disc
.state
;
587 info
[i
].major
= disc
.major
;
588 info
[i
].minor
= disc
.minor
;
589 if (disc
.major
|| disc
.minor
)
592 info
[i
].major
= info
[i
].minor
= 0;
596 if (mse
->metadata_version
&&
597 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
598 is_subarray(mse
->metadata_version
+9))
600 devname2devnum(mse
->metadata_version
+10);
602 st
->parent_dev
= NoMdDev
;
603 if (st
->metadata
== NULL
&&
604 st
->parent_dev
== NoMdDev
)
605 st
->metadata
= super_by_fd(fd
, NULL
);
609 for (i
=0; i
<MaxDisks
; i
++) {
610 mdu_disk_info_t disc
= {0,0,0,0,0};
615 if (i
>= last_disk
) {
617 disc
.major
= disc
.minor
= 0;
618 } else if (info
[i
].major
|| info
[i
].minor
) {
619 newstate
= info
[i
].state
;
620 dv
= map_dev(info
[i
].major
, info
[i
].minor
, 1);
621 disc
.state
= newstate
;
622 disc
.major
= info
[i
].major
;
623 disc
.minor
= info
[i
].minor
;
624 } else if (mse
&& mse
->pattern
&& i
< (int)strlen(mse
->pattern
)) {
625 switch(mse
->pattern
[i
]) {
626 case 'U': newstate
= 6 /* ACTIVE/SYNC */; break;
627 case '_': newstate
= 8 /* REMOVED */; break;
629 disc
.major
= disc
.minor
= 0;
631 if (dv
== NULL
&& st
->devid
[i
])
632 dv
= map_dev(major(st
->devid
[i
]),
633 minor(st
->devid
[i
]), 1);
634 change
= newstate
^ st
->devstate
[i
];
635 if (st
->utime
&& change
&& !st
->err
) {
636 if (i
< array
.raid_disks
&&
637 (((newstate
&change
)&(1<<MD_DISK_FAULTY
)) ||
638 ((st
->devstate
[i
]&change
)&(1<<MD_DISK_ACTIVE
)) ||
639 ((st
->devstate
[i
]&change
)&(1<<MD_DISK_SYNC
)))
641 alert("Fail", dev
, dv
, ainfo
);
642 else if (i
>= array
.raid_disks
&&
643 (disc
.major
|| disc
.minor
) &&
644 st
->devid
[i
] == makedev(disc
.major
, disc
.minor
) &&
645 ((newstate
&change
)&(1<<MD_DISK_FAULTY
))
647 alert("FailSpare", dev
, dv
, ainfo
);
648 else if (i
< array
.raid_disks
&&
649 ! (newstate
& (1<<MD_DISK_REMOVED
)) &&
650 (((st
->devstate
[i
]&change
)&(1<<MD_DISK_FAULTY
)) ||
651 ((newstate
&change
)&(1<<MD_DISK_ACTIVE
)) ||
652 ((newstate
&change
)&(1<<MD_DISK_SYNC
)))
654 alert("SpareActive", dev
, dv
, ainfo
);
656 st
->devstate
[i
] = newstate
;
657 st
->devid
[i
] = makedev(disc
.major
, disc
.minor
);
659 st
->active
= array
.active_disks
;
660 st
->working
= array
.working_disks
;
661 st
->spare
= array
.spare_disks
;
662 st
->failed
= array
.failed_disks
;
663 st
->utime
= array
.utime
;
664 st
->raid
= array
.raid_disks
;
666 if ((st
->active
< st
->raid
) && st
->spare
== 0)
671 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
672 int test
, struct alert_info
*info
)
674 struct mdstat_ent
*mse
;
677 for (mse
=mdstat
; mse
; mse
=mse
->next
)
678 if (mse
->devnum
!= INT_MAX
&&
679 (!mse
->level
|| /* retrieve containers */
680 (strcmp(mse
->level
, "raid0") != 0 &&
681 strcmp(mse
->level
, "linear") != 0))
683 struct state
*st
= calloc(1, sizeof *st
);
684 mdu_array_info_t array
;
688 st
->devname
= strdup(get_md_name(mse
->devnum
));
689 if ((fd
= open(st
->devname
, O_RDONLY
)) < 0 ||
690 ioctl(fd
, GET_ARRAY_INFO
, &array
)< 0) {
692 if (fd
>=0) close(fd
);
693 put_md_name(st
->devname
);
696 st
->metadata
->ss
->free_super(st
->metadata
);
703 st
->next
= *statelist
;
705 st
->devnum
= mse
->devnum
;
707 st
->expected_spares
= -1;
708 if (mse
->metadata_version
&&
709 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
710 is_subarray(mse
->metadata_version
+9))
712 devname2devnum(mse
->metadata_version
+10);
714 st
->parent_dev
= NoMdDev
;
717 alert("TestMessage", st
->devname
, NULL
, info
);
718 alert("NewArray", st
->devname
, NULL
, info
);
724 static int get_min_spare_size_required(struct state
*st
, unsigned long long *sizep
)
729 !st
->metadata
->ss
->min_acceptable_spare_size
) {
734 fd
= open(st
->devname
, O_RDONLY
);
737 if (st
->metadata
->ss
->external
)
738 st
->metadata
->ss
->load_container(st
->metadata
, fd
, st
->devname
);
740 st
->metadata
->ss
->load_super(st
->metadata
, fd
, st
->devname
);
742 if (!st
->metadata
->sb
)
744 *sizep
= st
->metadata
->ss
->min_acceptable_spare_size(st
->metadata
);
745 st
->metadata
->ss
->free_super(st
->metadata
);
750 static int check_donor(struct state
*from
, struct state
*to
)
757 /* Cannot move from a member */
761 for (sub
= from
->subarray
; sub
; sub
= sub
->subarray
)
762 /* If source array has degraded subarrays, don't
765 if (sub
->active
< sub
->raid
)
767 if (from
->metadata
->ss
->external
== 0)
768 if (from
->active
< from
->raid
)
770 if (from
->spare
<= 0)
775 static dev_t
choose_spare(struct state
*from
, struct state
*to
,
776 struct domainlist
*domlist
, unsigned long long min_size
)
781 for (d
= from
->raid
; !dev
&& d
< MaxDisks
; d
++) {
782 if (from
->devid
[d
] > 0 &&
783 from
->devstate
[d
] == 0) {
784 struct dev_policy
*pol
;
785 unsigned long long dev_size
;
787 if (to
->metadata
->ss
->external
&&
788 test_partition_from_id(from
->devid
[d
]))
792 dev_size_from_id(from
->devid
[d
], &dev_size
) &&
796 pol
= devnum_policy(from
->devid
[d
]);
797 if (from
->spare_group
)
798 pol_add(&pol
, pol_domain
,
799 from
->spare_group
, NULL
);
800 if (domain_test(domlist
, pol
, to
->metadata
->ss
->name
) == 1)
801 dev
= from
->devid
[d
];
802 dev_policy_free(pol
);
808 static dev_t
container_choose_spare(struct state
*from
, struct state
*to
,
809 struct domainlist
*domlist
,
810 unsigned long long min_size
, int active
)
812 /* This is similar to choose_spare, but we cannot trust devstate,
813 * so we need to read the metadata instead
816 struct supertype
*st
= from
->metadata
;
817 int fd
= open(from
->devname
, O_RDONLY
);
823 if (!st
->ss
->getinfo_super_disks
) {
828 err
= st
->ss
->load_container(st
, fd
, NULL
);
834 /* We must check if number of active disks has not increased
835 * since ioctl in main loop. mdmon may have added spare
836 * to subarray. If so we do not need to look for more spares
837 * so return non zero value */
840 list
= st
->ss
->getinfo_super_disks(st
);
842 st
->ss
->free_super(st
);
847 if (dp
->disk
.state
& (1<<MD_DISK_SYNC
) &&
848 !(dp
->disk
.state
& (1<<MD_DISK_FAULTY
)))
853 if (active
< active_cnt
) {
854 /* Spare just activated.*/
855 st
->ss
->free_super(st
);
860 /* We only need one spare so full list not needed */
861 list
= container_choose_spares(st
, min_size
, domlist
, from
->spare_group
,
862 to
->metadata
->ss
->name
, 1);
864 struct mdinfo
*disks
= list
->devs
;
866 dev
= makedev(disks
->disk
.major
, disks
->disk
.minor
);
869 st
->ss
->free_super(st
);
874 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
)
879 link_containers_with_subarrays(statelist
);
880 for (st
= statelist
; st
; st
= st
->next
)
881 if (st
->active
< st
->raid
&&
882 st
->spare
== 0 && !st
->err
) {
883 struct domainlist
*domlist
= NULL
;
885 struct state
*to
= st
;
886 unsigned long long min_size
;
888 if (to
->parent_dev
!= NoMdDev
&& !to
->parent
)
889 /* subarray monitored without parent container
890 * we can't move spares here */
894 /* member of a container */
897 if (get_min_spare_size_required(to
, &min_size
))
899 if (to
->metadata
->ss
->external
) {
900 /* We must make sure there is
901 * no suitable spare in container already.
902 * If there is we don't add more */
903 dev_t devid
= container_choose_spare(
904 to
, to
, NULL
, min_size
, st
->active
);
908 for (d
= 0; d
< MaxDisks
; d
++)
910 domainlist_add_dev(&domlist
,
912 to
->metadata
->ss
->name
);
914 domain_add(&domlist
, to
->spare_group
);
916 * No spare migration if the destination
917 * has no domain. Skip this array.
921 for (from
=statelist
; from
; from
=from
->next
) {
923 if (!check_donor(from
, to
))
925 if (from
->metadata
->ss
->external
)
926 devid
= container_choose_spare(
927 from
, to
, domlist
, min_size
, 0);
929 devid
= choose_spare(from
, to
, domlist
,
932 && move_spare(from
->devname
, to
->devname
, devid
)) {
933 alert("MoveSpare", to
->devname
, from
->devname
, info
);
937 domain_free(domlist
);
941 /* search the statelist to connect external
942 * metadata subarrays with their containers
943 * We always completely rebuild the tree from scratch as
944 * that is safest considering the possibility of entries
945 * disappearing or changing.
947 static void link_containers_with_subarrays(struct state
*list
)
951 for (st
= list
; st
; st
= st
->next
) {
955 for (st
= list
; st
; st
= st
->next
)
956 if (st
->parent_dev
!= NoMdDev
)
957 for (cont
= list
; cont
; cont
= cont
->next
)
959 cont
->parent_dev
== NoMdDev
&&
960 cont
->devnum
== st
->parent_dev
) {
962 st
->subarray
= cont
->subarray
;
968 /* Not really Monitor but ... */
975 if (stat(dev
, &stb
) != 0) {
976 fprintf(stderr
, Name
": Cannot find %s: %s\n", dev
,
980 devnum
= stat2devnum(&stb
);
983 struct mdstat_ent
*ms
= mdstat_read(1, 0);
984 struct mdstat_ent
*e
;
986 for (e
=ms
; e
; e
=e
->next
)
987 if (e
->devnum
== devnum
)
990 if (!e
|| e
->percent
< 0) {
991 if (e
&& e
->metadata_version
&&
992 strncmp(e
->metadata_version
, "external:", 9) == 0) {
993 if (is_subarray(&e
->metadata_version
[9]))
994 ping_monitor(&e
->metadata_version
[9]);
996 ping_monitor_by_id(devnum
);
1009 static char *clean_states
[] = {
1010 "clear", "inactive", "readonly", "read-auto", "clean", NULL
};
1012 int WaitClean(char *dev
, int sock
, int verbose
)
1019 fd
= open(dev
, O_RDONLY
);
1022 fprintf(stderr
, Name
": Couldn't open %s: %s\n", dev
, strerror(errno
));
1026 devnum
= fd2devnum(fd
);
1027 mdi
= sysfs_read(fd
, devnum
, GET_VERSION
|GET_LEVEL
|GET_SAFEMODE
);
1030 fprintf(stderr
, Name
": Failed to read sysfs attributes for "
1036 switch(mdi
->array
.level
) {
1038 case LEVEL_MULTIPATH
:
1040 /* safemode delay is irrelevant for these levels */
1044 /* for internal metadata the kernel handles the final clean
1045 * transition, containers can never be dirty
1047 if (!is_subarray(mdi
->text_version
))
1050 /* safemode disabled ? */
1051 if (mdi
->safe_mode_delay
== 0)
1055 int state_fd
= sysfs_open(fd2devnum(fd
), NULL
, "array_state");
1060 /* minimize the safe_mode_delay and prepare to wait up to 5s
1061 * for writes to quiesce
1063 sysfs_set_safemode(mdi
, 1);
1069 /* wait for array_state to be clean */
1071 rv
= read(state_fd
, buf
, sizeof(buf
));
1074 if (sysfs_match_word(buf
, clean_states
) <= 4)
1076 FD_SET(state_fd
, &fds
);
1077 rv
= select(state_fd
+ 1, NULL
, NULL
, &fds
, &tm
);
1078 if (rv
< 0 && errno
!= EINTR
)
1080 lseek(state_fd
, 0, SEEK_SET
);
1084 else if (fping_monitor(sock
) == 0 ||
1085 ping_monitor(mdi
->text_version
) == 0) {
1086 /* we need to ping to close the window between array
1087 * state transitioning to clean and the metadata being
1094 fprintf(stderr
, Name
": Error waiting for %s to be clean\n",
1097 /* restore the original safe_mode_delay */
1098 sysfs_set_safemode(mdi
, mdi
->safe_mode_delay
);
1107 #endif /* MDASSEMBLE */