]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Monitor.c
a82e99d693abc287553fae52536571a2825cb6d2
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 write_autorebuild_pid(void);
67 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
);
68 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
69 int test
, struct alert_info
*info
,
70 int increments
, char *prefer
);
71 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
72 int test
, struct alert_info
*info
);
73 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
);
74 static void link_containers_with_subarrays(struct state
*list
);
76 int Monitor(struct mddev_dev
*devlist
,
77 char *mailaddr
, char *alert_cmd
,
79 int daemonise
, int oneshot
,
80 int dosyslog
, char *pidfile
, int increments
,
84 * Every few seconds, scan every md device looking for changes
85 * When a change is found, log it, possibly run the alert command,
86 * and possibly send Email
88 * For each array, we record:
90 * active/working/failed/spare drives
91 * State of each device.
92 * %rebuilt if rebuilding
94 * If the update time changes, check out all the data again
95 * It is possible that we cannot get the state of each device
96 * due to bugs in the md kernel module.
97 * We also read /proc/mdstat to get rebuild percent,
98 * and to get state on all active devices incase of kernel bug.
102 * An active device had Faulty set or Active/Sync removed
104 * A spare device had Faulty set
106 * An active device had a reverse transition
108 * percent went from -1 to +ve
110 * percent went from below to not-below NN%
112 * Couldn't access a device which was previously visible
114 * if we detect an array with active<raid and spare==0
115 * we look at other arrays that have same spare-group
116 * If we find one with active==raid and spare>0,
117 * and if we can get_disk_info and find a name
118 * Then we hot-remove and hot-add to the other array
120 * If devlist is NULL, then we can monitor everything because --scan
121 * was given. We get an initial list from config file and add anything
122 * that appears in /proc/mdstat
125 struct state
*statelist
= NULL
;
128 struct mdstat_ent
*mdstat
= NULL
;
130 struct alert_info info
;
131 struct mddev_ident
*mdlist
;
132 int delay_for_event
= c
->delay
;
135 mailaddr
= conf_get_mailaddr();
136 if (mailaddr
&& ! c
->scan
)
137 pr_err("Monitor using email address \"%s\" from config file\n",
140 mailfrom
= conf_get_mailfrom();
143 alert_cmd
= conf_get_program();
144 if (alert_cmd
&& !c
->scan
)
145 pr_err("Monitor using program \"%s\" from config file\n",
148 if (c
->scan
&& !mailaddr
&& !alert_cmd
&& !dosyslog
) {
149 pr_err("No mail address or alert command - not monitoring.\n");
152 info
.alert_cmd
= alert_cmd
;
153 info
.mailaddr
= mailaddr
;
154 info
.mailfrom
= mailfrom
;
155 info
.dosyslog
= dosyslog
;
158 if (check_one_sharer(c
->scan
))
163 int rv
= make_daemon(pidfile
);
169 write_autorebuild_pid();
171 if (devlist
== NULL
) {
172 mdlist
= conf_get_ident(NULL
);
173 for (; mdlist
; mdlist
= mdlist
->next
) {
176 if (mdlist
->devname
== NULL
)
178 if (strcasecmp(mdlist
->devname
, "<ignore>") == 0)
180 st
= xcalloc(1, sizeof *st
);
181 if (mdlist
->devname
[0] == '/')
182 st
->devname
= xstrdup(mdlist
->devname
);
184 st
->devname
= xmalloc(8+strlen(mdlist
->devname
)+1);
185 strcpy(strcpy(st
->devname
, "/dev/md/"),
188 st
->next
= statelist
;
190 st
->percent
= RESYNC_UNKNOWN
;
192 st
->expected_spares
= mdlist
->spare_disks
;
193 if (mdlist
->spare_group
)
194 st
->spare_group
= xstrdup(mdlist
->spare_group
);
198 struct mddev_dev
*dv
;
200 for (dv
= devlist
; dv
; dv
= dv
->next
) {
201 struct state
*st
= xcalloc(1, sizeof *st
);
202 mdlist
= conf_get_ident(dv
->devname
);
203 st
->devname
= xstrdup(dv
->devname
);
204 st
->next
= statelist
;
206 st
->percent
= RESYNC_UNKNOWN
;
207 st
->expected_spares
= -1;
209 st
->expected_spares
= mdlist
->spare_disks
;
210 if (mdlist
->spare_group
)
211 st
->spare_group
= xstrdup(mdlist
->spare_group
);
219 struct state
*st
, **stp
;
221 int anyredundant
= 0;
225 mdstat
= mdstat_read(oneshot
? 0 : 1, 0);
227 for (st
= statelist
; st
; st
= st
->next
) {
228 if (check_array(st
, mdstat
, c
->test
, &info
,
229 increments
, c
->prefer
))
231 /* for external arrays, metadata is filled for
234 if (st
->metadata
&& st
->metadata
->ss
->external
)
236 if (st
->err
== 0 && !anyredundant
)
240 /* now check if there are any new devices found in mdstat */
242 new_found
= add_new_arrays(mdstat
, &statelist
, c
->test
,
245 /* If an array has active < raid && spare == 0 && spare_group != NULL
246 * Look for another array with spare > 0 and active == raid and same spare_group
247 * if found, choose a device and hotremove/hotadd
249 if (share
&& anydegraded
)
250 try_spare_migration(statelist
, &info
);
254 else if (!anyredundant
) {
258 int wait_result
= mdstat_wait(delay_for_event
);
261 * If mdmonitor is awaken by event, set small delay once
262 * to deal with udev and mdadm.
264 if (wait_result
!= 0) {
268 delay_for_event
= c
->delay
;
275 for (stp
= &statelist
; (st
= *stp
) != NULL
; ) {
276 if (st
->from_auto
&& st
->err
> 5) {
279 free(st
->spare_group
);
285 for (st2
= statelist
; st2
; st2
= statelist
) {
286 statelist
= st2
->next
;
295 static int make_daemon(char *pidfile
)
298 * -1 in the forked daemon
301 * so a none-negative becomes the exit code.
308 FILE *pid_file
= NULL
;
309 int fd
= open(pidfile
, O_WRONLY
| O_CREAT
| O_TRUNC
,
312 pid_file
= fdopen(fd
, "w");
314 perror("cannot create pid file");
316 fprintf(pid_file
,"%d\n", pid
);
327 open("/dev/null", O_RDWR
);
334 static int check_one_sharer(int scan
)
339 char comm_path
[PATH_MAX
];
343 sprintf(path
, "%s/autorebuild.pid", MDMON_DIR
);
344 fp
= fopen(path
, "r");
346 if (fscanf(fp
, "%d", &pid
) != 1)
348 snprintf(comm_path
, sizeof(comm_path
),
349 "/proc/%d/comm", pid
);
350 comm_fp
= fopen(comm_path
, "r");
352 if (fscanf(comm_fp
, "%s", comm
) &&
353 strncmp(basename(comm
), Name
, strlen(Name
)) == 0) {
355 pr_err("Only one autorebuild process allowed in scan mode, aborting\n");
360 pr_err("Warning: One autorebuild process already running.\n");
370 static void write_autorebuild_pid()
375 sprintf(path
, "%s/autorebuild.pid", MDMON_DIR
);
377 if (mkdir(MDMON_DIR
, 0700) < 0 && errno
!= EEXIST
) {
378 pr_err("Can't create autorebuild.pid file\n");
380 int fd
= open(path
, O_WRONLY
| O_CREAT
| O_TRUNC
, 0700);
383 fp
= fdopen(fd
, "w");
386 pr_err("Can't create autorebuild.pid file\n");
389 fprintf(fp
, "%d\n", pid
);
395 static void alert(char *event
, char *dev
, char *disc
, struct alert_info
*info
)
399 if (!info
->alert_cmd
&& !info
->mailaddr
&& !info
->dosyslog
) {
400 time_t now
= time(0);
402 printf("%1.15s: %s on %s %s\n", ctime(&now
) + 4,
403 event
, dev
, disc
?disc
:"unknown device");
405 if (info
->alert_cmd
) {
409 waitpid(pid
, NULL
, 0);
414 execl(info
->alert_cmd
, info
->alert_cmd
,
415 event
, dev
, disc
, NULL
);
419 if (info
->mailaddr
&& (strncmp(event
, "Fail", 4) == 0 ||
420 strncmp(event
, "Test", 4) == 0 ||
421 strncmp(event
, "Spares", 6) == 0 ||
422 strncmp(event
, "Degrade", 7) == 0)) {
423 FILE *mp
= popen(Sendmail
, "w");
427 gethostname(hname
, sizeof(hname
));
428 signal(SIGPIPE
, SIG_IGN
);
430 fprintf(mp
, "From: %s\n", info
->mailfrom
);
432 fprintf(mp
, "From: %s monitoring <root>\n",
434 fprintf(mp
, "To: %s\n", info
->mailaddr
);
435 fprintf(mp
, "Subject: %s event on %s:%s\n\n",
439 "This is an automatically generated mail message from %s\n", Name
);
440 fprintf(mp
, "running on %s\n\n", hname
);
443 "A %s event had been detected on md device %s.\n\n", event
, dev
);
445 if (disc
&& disc
[0] != ' ')
447 "It could be related to component device %s.\n\n", disc
);
448 if (disc
&& disc
[0] == ' ')
449 fprintf(mp
, "Extra information:%s.\n\n", disc
);
451 fprintf(mp
, "Faithfully yours, etc.\n");
453 mdstat
= fopen("/proc/mdstat", "r");
458 "\nP.S. The /proc/mdstat file currently contains the following:\n\n");
459 while ((n
= fread(buf
, 1, sizeof(buf
),
461 n
= fwrite(buf
, 1, n
, mp
);
468 /* log the event to syslog maybe */
469 if (info
->dosyslog
) {
470 /* Log at a different severity depending on the event.
472 * These are the critical events: */
473 if (strncmp(event
, "Fail", 4) == 0 ||
474 strncmp(event
, "Degrade", 7) == 0 ||
475 strncmp(event
, "DeviceDisappeared", 17) == 0)
477 /* Good to know about, but are not failures: */
478 else if (strncmp(event
, "Rebuild", 7) == 0 ||
479 strncmp(event
, "MoveSpare", 9) == 0 ||
480 strncmp(event
, "Spares", 6) != 0)
481 priority
= LOG_WARNING
;
482 /* Everything else: */
486 if (disc
&& disc
[0] != ' ')
488 "%s event detected on md device %s, component device %s", event
, dev
, disc
);
491 "%s event detected on md device %s: %s",
495 "%s event detected on md device %s",
500 static int check_array(struct state
*st
, struct mdstat_ent
*mdstat
,
501 int test
, struct alert_info
*ainfo
,
502 int increments
, char *prefer
)
504 /* Update the state 'st' to reflect any changes shown in mdstat,
505 * or found by directly examining the array, and return
506 * '1' if the array is degraded, or '0' if it is optimal (or dead).
508 struct { int state
, major
, minor
; } info
[MAX_DISKS
];
509 struct mdinfo
*sra
= NULL
;
510 mdu_array_info_t array
;
511 struct mdstat_ent
*mse
= NULL
, *mse2
;
512 char *dev
= st
->devname
;
519 int is_container
= 0;
520 unsigned long redundancy_only_flags
= 0;
523 alert("TestMessage", dev
, NULL
, ainfo
);
527 fd
= open(dev
, O_RDONLY
);
531 if (st
->devnm
[0] == 0)
532 strcpy(st
->devnm
, fd2devnm(fd
));
534 for (mse2
= mdstat
; mse2
; mse2
= mse2
->next
)
535 if (strcmp(mse2
->devnm
, st
->devnm
) == 0) {
536 mse2
->devnm
[0] = 0; /* flag it as "used" */
541 /* duplicated array in statelist
542 * or re-created after reading mdstat
548 if (mse
->level
== NULL
)
551 if (!is_container
&& !md_array_active(fd
))
554 fcntl(fd
, F_SETFD
, FD_CLOEXEC
);
555 if (md_get_array_info(fd
, &array
) < 0)
558 if (!is_container
&& map_name(pers
, mse
->level
) > 0)
559 redundancy_only_flags
|= GET_MISMATCH
;
561 sra
= sysfs_read(-1, st
->devnm
, GET_LEVEL
| GET_DISKS
| GET_DEVS
|
562 GET_STATE
| redundancy_only_flags
);
567 /* It's much easier to list what array levels can't
568 * have a device disappear than all of them that can
570 if (sra
->array
.level
== 0 || sra
->array
.level
== -1) {
571 if (!st
->err
&& !st
->from_config
)
572 alert("DeviceDisappeared", dev
, " Wrong-Level", ainfo
);
577 /* this array is in /proc/mdstat */
578 if (array
.utime
== 0)
579 /* external arrays don't update utime, so
580 * just make sure it is always different. */
581 array
.utime
= st
->utime
+ 1;;
584 /* New array appeared where previously had an error */
586 st
->percent
= RESYNC_NONE
;
589 alert("NewArray", st
->devname
, NULL
, ainfo
);
592 if (st
->utime
== array
.utime
&& st
->failed
== sra
->array
.failed_disks
&&
593 st
->working
== sra
->array
.working_disks
&&
594 st
->spare
== sra
->array
.spare_disks
&&
595 (mse
== NULL
|| (mse
->percent
== st
->percent
))) {
596 if ((st
->active
< st
->raid
) && st
->spare
== 0)
600 if (st
->utime
== 0 && /* new array */
601 mse
->pattern
&& strchr(mse
->pattern
, '_') /* degraded */)
602 alert("DegradedArray", dev
, NULL
, ainfo
);
604 if (st
->utime
== 0 && /* new array */ st
->expected_spares
> 0 &&
605 sra
->array
.spare_disks
< st
->expected_spares
)
606 alert("SparesMissing", dev
, NULL
, ainfo
);
607 if (st
->percent
< 0 && st
->percent
!= RESYNC_UNKNOWN
&&
609 alert("RebuildStarted", dev
, NULL
, ainfo
);
610 if (st
->percent
>= 0 && mse
->percent
>= 0 &&
611 (mse
->percent
/ increments
) > (st
->percent
/ increments
)) {
612 char percentalert
[18];
614 * "RebuildNN" (10 chars) or "RebuildStarted" (15 chars)
617 if((mse
->percent
/ increments
) == 0)
618 snprintf(percentalert
, sizeof(percentalert
),
621 snprintf(percentalert
, sizeof(percentalert
),
622 "Rebuild%02d", mse
->percent
);
624 alert(percentalert
, dev
, NULL
, ainfo
);
627 if (mse
->percent
== RESYNC_NONE
&& st
->percent
>= 0) {
628 /* Rebuild/sync/whatever just finished.
629 * If there is a number in /mismatch_cnt,
630 * we should report that.
632 if (sra
&& sra
->mismatch_cnt
> 0) {
634 snprintf(cnt
, sizeof(cnt
),
635 " mismatches found: %d (on raid level %d)",
636 sra
->mismatch_cnt
, sra
->array
.level
);
637 alert("RebuildFinished", dev
, cnt
, ainfo
);
639 alert("RebuildFinished", dev
, NULL
, ainfo
);
641 st
->percent
= mse
->percent
;
643 remaining_disks
= sra
->array
.nr_disks
;
644 for (i
= 0; i
< MAX_DISKS
&& remaining_disks
> 0; i
++) {
645 mdu_disk_info_t disc
;
647 if (md_get_disk_info(fd
, &disc
) >= 0) {
648 info
[i
].state
= disc
.state
;
649 info
[i
].major
= disc
.major
;
650 info
[i
].minor
= disc
.minor
;
651 if (disc
.major
|| disc
.minor
)
654 info
[i
].major
= info
[i
].minor
= 0;
658 if (mse
->metadata_version
&&
659 strncmp(mse
->metadata_version
, "external:", 9) == 0 &&
660 is_subarray(mse
->metadata_version
+9)) {
662 strcpy(st
->parent_devnm
, mse
->metadata_version
+ 10);
663 sl
= strchr(st
->parent_devnm
, '/');
667 st
->parent_devnm
[0] = 0;
668 if (st
->metadata
== NULL
&& st
->parent_devnm
[0] == 0)
669 st
->metadata
= super_by_fd(fd
, NULL
);
671 for (i
= 0; i
< MAX_DISKS
; i
++) {
672 mdu_disk_info_t disc
= {0, 0, 0, 0, 0};
677 if (i
< last_disk
&& (info
[i
].major
|| info
[i
].minor
)) {
678 newstate
= info
[i
].state
;
679 dv
= map_dev_preferred(info
[i
].major
, info
[i
].minor
, 1,
681 disc
.state
= newstate
;
682 disc
.major
= info
[i
].major
;
683 disc
.minor
= info
[i
].minor
;
685 newstate
= (1 << MD_DISK_REMOVED
);
687 if (dv
== NULL
&& st
->devid
[i
])
688 dv
= map_dev_preferred(major(st
->devid
[i
]),
689 minor(st
->devid
[i
]), 1, prefer
);
690 change
= newstate
^ st
->devstate
[i
];
691 if (st
->utime
&& change
&& !st
->err
&& !new_array
) {
692 if ((st
->devstate
[i
]&change
) & (1 << MD_DISK_SYNC
))
693 alert("Fail", dev
, dv
, ainfo
);
694 else if ((newstate
& (1 << MD_DISK_FAULTY
)) &&
695 (disc
.major
|| disc
.minor
) &&
696 st
->devid
[i
] == makedev(disc
.major
,
698 alert("FailSpare", dev
, dv
, ainfo
);
699 else if ((newstate
&change
) & (1 << MD_DISK_SYNC
))
700 alert("SpareActive", dev
, dv
, ainfo
);
702 st
->devstate
[i
] = newstate
;
703 st
->devid
[i
] = makedev(disc
.major
, disc
.minor
);
705 st
->active
= sra
->array
.active_disks
;
706 st
->working
= sra
->array
.working_disks
;
707 st
->spare
= sra
->array
.spare_disks
;
708 st
->failed
= sra
->array
.failed_disks
;
709 st
->utime
= array
.utime
;
710 st
->raid
= sra
->array
.raid_disks
;
712 if ((st
->active
< st
->raid
) && st
->spare
== 0)
723 if (!st
->err
&& !is_container
)
724 alert("DeviceDisappeared", dev
, NULL
, ainfo
);
729 static int add_new_arrays(struct mdstat_ent
*mdstat
, struct state
**statelist
,
730 int test
, struct alert_info
*info
)
732 struct mdstat_ent
*mse
;
736 for (mse
= mdstat
; mse
; mse
= mse
->next
)
737 if (mse
->devnm
[0] && (!mse
->level
|| /* retrieve containers */
738 (strcmp(mse
->level
, "raid0") != 0 &&
739 strcmp(mse
->level
, "linear") != 0))) {
740 struct state
*st
= xcalloc(1, sizeof *st
);
741 mdu_array_info_t array
;
744 name
= get_md_name(mse
->devnm
);
750 st
->devname
= xstrdup(name
);
751 if ((fd
= open(st
->devname
, O_RDONLY
)) < 0 ||
752 md_get_array_info(fd
, &array
) < 0) {
756 put_md_name(st
->devname
);
759 st
->metadata
->ss
->free_super(st
->metadata
);
766 st
->next
= *statelist
;
769 strcpy(st
->devnm
, mse
->devnm
);
770 st
->percent
= RESYNC_UNKNOWN
;
771 st
->expected_spares
= -1;
772 if (mse
->metadata_version
&&
773 strncmp(mse
->metadata_version
,
774 "external:", 9) == 0 &&
775 is_subarray(mse
->metadata_version
+9)) {
777 strcpy(st
->parent_devnm
,
778 mse
->metadata_version
+10);
779 sl
= strchr(st
->parent_devnm
, '/');
782 st
->parent_devnm
[0] = 0;
785 alert("TestMessage", st
->devname
, NULL
, info
);
791 static int get_required_spare_criteria(struct state
*st
,
792 struct spare_criteria
*sc
)
796 if (!st
->metadata
|| !st
->metadata
->ss
->get_spare_criteria
) {
802 fd
= open(st
->devname
, O_RDONLY
);
805 if (st
->metadata
->ss
->external
)
806 st
->metadata
->ss
->load_container(st
->metadata
, fd
, st
->devname
);
808 st
->metadata
->ss
->load_super(st
->metadata
, fd
, st
->devname
);
810 if (!st
->metadata
->sb
)
813 st
->metadata
->ss
->get_spare_criteria(st
->metadata
, sc
);
814 st
->metadata
->ss
->free_super(st
->metadata
);
819 static int check_donor(struct state
*from
, struct state
*to
)
826 /* Cannot move from a member */
830 for (sub
= from
->subarray
; sub
; sub
= sub
->subarray
)
831 /* If source array has degraded subarrays, don't
834 if (sub
->active
< sub
->raid
)
836 if (from
->metadata
->ss
->external
== 0)
837 if (from
->active
< from
->raid
)
839 if (from
->spare
<= 0)
844 static dev_t
choose_spare(struct state
*from
, struct state
*to
,
845 struct domainlist
*domlist
, struct spare_criteria
*sc
)
850 for (d
= from
->raid
; !dev
&& d
< MAX_DISKS
; d
++) {
851 if (from
->devid
[d
] > 0 && from
->devstate
[d
] == 0) {
852 struct dev_policy
*pol
;
853 unsigned long long dev_size
;
854 unsigned int dev_sector_size
;
856 if (to
->metadata
->ss
->external
&&
857 test_partition_from_id(from
->devid
[d
]))
861 dev_size_from_id(from
->devid
[d
], &dev_size
) &&
862 dev_size
< sc
->min_size
)
865 if (sc
->sector_size
&&
866 dev_sector_size_from_id(from
->devid
[d
],
868 sc
->sector_size
!= dev_sector_size
)
871 pol
= devid_policy(from
->devid
[d
]);
872 if (from
->spare_group
)
873 pol_add(&pol
, pol_domain
,
874 from
->spare_group
, NULL
);
875 if (domain_test(domlist
, pol
,
876 to
->metadata
->ss
->name
) == 1)
877 dev
= from
->devid
[d
];
878 dev_policy_free(pol
);
884 static dev_t
container_choose_spare(struct state
*from
, struct state
*to
,
885 struct domainlist
*domlist
,
886 struct spare_criteria
*sc
, int active
)
888 /* This is similar to choose_spare, but we cannot trust devstate,
889 * so we need to read the metadata instead
892 struct supertype
*st
= from
->metadata
;
893 int fd
= open(from
->devname
, O_RDONLY
);
899 if (!st
->ss
->getinfo_super_disks
) {
904 err
= st
->ss
->load_container(st
, fd
, NULL
);
910 /* We must check if number of active disks has not increased
911 * since ioctl in main loop. mdmon may have added spare
912 * to subarray. If so we do not need to look for more spares
913 * so return non zero value */
916 list
= st
->ss
->getinfo_super_disks(st
);
918 st
->ss
->free_super(st
);
923 if (dp
->disk
.state
& (1 << MD_DISK_SYNC
) &&
924 !(dp
->disk
.state
& (1 << MD_DISK_FAULTY
)))
929 if (active
< active_cnt
) {
930 /* Spare just activated.*/
931 st
->ss
->free_super(st
);
936 /* We only need one spare so full list not needed */
937 list
= container_choose_spares(st
, sc
, domlist
, from
->spare_group
,
938 to
->metadata
->ss
->name
, 1);
940 struct mdinfo
*disks
= list
->devs
;
942 dev
= makedev(disks
->disk
.major
, disks
->disk
.minor
);
945 st
->ss
->free_super(st
);
949 static void try_spare_migration(struct state
*statelist
, struct alert_info
*info
)
953 struct spare_criteria sc
;
955 link_containers_with_subarrays(statelist
);
956 for (st
= statelist
; st
; st
= st
->next
)
957 if (st
->active
< st
->raid
&& st
->spare
== 0 && !st
->err
) {
958 struct domainlist
*domlist
= NULL
;
960 struct state
*to
= st
;
962 if (to
->parent_devnm
[0] && !to
->parent
)
963 /* subarray monitored without parent container
964 * we can't move spares here */
968 /* member of a container */
971 if (get_required_spare_criteria(to
, &sc
))
973 if (to
->metadata
->ss
->external
) {
974 /* We must make sure there is
975 * no suitable spare in container already.
976 * If there is we don't add more */
977 dev_t devid
= container_choose_spare(
978 to
, to
, NULL
, &sc
, st
->active
);
982 for (d
= 0; d
< MAX_DISKS
; d
++)
984 domainlist_add_dev(&domlist
,
986 to
->metadata
->ss
->name
);
988 domain_add(&domlist
, to
->spare_group
);
990 * No spare migration if the destination
991 * has no domain. Skip this array.
995 for (from
=statelist
; from
; from
=from
->next
) {
997 if (!check_donor(from
, to
))
999 if (from
->metadata
->ss
->external
)
1000 devid
= container_choose_spare(
1001 from
, to
, domlist
, &sc
, 0);
1003 devid
= choose_spare(from
, to
, domlist
,
1006 move_spare(from
->devname
, to
->devname
,
1008 alert("MoveSpare", to
->devname
,
1009 from
->devname
, info
);
1013 domain_free(domlist
);
1017 /* search the statelist to connect external
1018 * metadata subarrays with their containers
1019 * We always completely rebuild the tree from scratch as
1020 * that is safest considering the possibility of entries
1021 * disappearing or changing.
1023 static void link_containers_with_subarrays(struct state
*list
)
1027 for (st
= list
; st
; st
= st
->next
) {
1029 st
->subarray
= NULL
;
1031 for (st
= list
; st
; st
= st
->next
)
1032 if (st
->parent_devnm
[0])
1033 for (cont
= list
; cont
; cont
= cont
->next
)
1034 if (!cont
->err
&& cont
->parent_devnm
[0] == 0 &&
1035 strcmp(cont
->devnm
, st
->parent_devnm
) == 0) {
1037 st
->subarray
= cont
->subarray
;
1038 cont
->subarray
= st
;
1043 /* Not really Monitor but ... */
1050 int frozen_remaining
= 3;
1052 if (!stat_is_blkdev(dev
, &rdev
))
1055 tmp
= devid2devnm(rdev
);
1057 pr_err("Cannot get md device name.\n");
1064 struct mdstat_ent
*ms
= mdstat_read(1, 0);
1065 struct mdstat_ent
*e
;
1067 for (e
= ms
; e
; e
= e
->next
)
1068 if (strcmp(e
->devnm
, devnm
) == 0)
1071 if (e
&& e
->percent
== RESYNC_NONE
) {
1072 /* We could be in the brief pause before something
1073 * starts. /proc/mdstat doesn't show that, but
1079 if (sysfs_init(&mdi
, -1, devnm
))
1081 if (sysfs_get_str(&mdi
, NULL
, "sync_action",
1083 strcmp(buf
,"idle\n") != 0) {
1084 e
->percent
= RESYNC_UNKNOWN
;
1085 if (strcmp(buf
, "frozen\n") == 0) {
1086 if (frozen_remaining
== 0)
1087 e
->percent
= RESYNC_NONE
;
1089 frozen_remaining
-= 1;
1093 if (!e
|| e
->percent
== RESYNC_NONE
) {
1094 if (e
&& e
->metadata_version
&&
1095 strncmp(e
->metadata_version
, "external:", 9) == 0) {
1096 if (is_subarray(&e
->metadata_version
[9]))
1097 ping_monitor(&e
->metadata_version
[9]);
1099 ping_monitor(devnm
);
1110 /* The state "broken" is used only for RAID0/LINEAR - it's the same as
1111 * "clean", but used in case the array has one or more members missing.
1113 static char *clean_states
[] = {
1114 "clear", "inactive", "readonly", "read-auto", "clean", "broken", NULL
};
1116 int WaitClean(char *dev
, int verbose
)
1123 if (!stat_is_blkdev(dev
, NULL
))
1125 fd
= open(dev
, O_RDONLY
);
1128 pr_err("Couldn't open %s: %s\n", dev
, strerror(errno
));
1132 strcpy(devnm
, fd2devnm(fd
));
1133 mdi
= sysfs_read(fd
, devnm
, GET_VERSION
|GET_LEVEL
|GET_SAFEMODE
);
1136 pr_err("Failed to read sysfs attributes for %s\n", dev
);
1141 switch(mdi
->array
.level
) {
1143 case LEVEL_MULTIPATH
:
1145 /* safemode delay is irrelevant for these levels */
1149 /* for internal metadata the kernel handles the final clean
1150 * transition, containers can never be dirty
1152 if (!is_subarray(mdi
->text_version
))
1155 /* safemode disabled ? */
1156 if (mdi
->safe_mode_delay
== 0)
1160 int state_fd
= sysfs_open(fd2devnm(fd
), NULL
, "array_state");
1164 /* minimize the safe_mode_delay and prepare to wait up to 5s
1165 * for writes to quiesce
1167 sysfs_set_safemode(mdi
, 1);
1169 /* wait for array_state to be clean */
1171 rv
= read(state_fd
, buf
, sizeof(buf
));
1174 if (sysfs_match_word(buf
, clean_states
) <
1175 (int)ARRAY_SIZE(clean_states
) - 1)
1177 rv
= sysfs_wait(state_fd
, &delay
);
1178 if (rv
< 0 && errno
!= EINTR
)
1180 lseek(state_fd
, 0, SEEK_SET
);
1184 else if (ping_monitor(mdi
->text_version
) == 0) {
1185 /* we need to ping to close the window between array
1186 * state transitioning to clean and the metadata being
1192 pr_err("Error connecting monitor with %s\n", dev
);
1195 pr_err("Error waiting for %s to be clean\n", dev
);
1197 /* restore the original safe_mode_delay */
1198 sysfs_set_safemode(mdi
, mdi
->safe_mode_delay
);