]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Manage.c
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>
30 #define REGISTER_DEV _IO (MD_MAJOR, 1)
31 #define START_MD _IO (MD_MAJOR, 2)
32 #define STOP_MD _IO (MD_MAJOR, 3)
34 int Manage_ro(char *devname
, int fd
, int readonly
)
36 /* switch to readonly or rw
39 * first check that array is runing
40 * use RESTART_ARRAY_RW or STOP_ARRAY_RO
43 mdu_array_info_t array
;
48 if (md_get_version(fd
) < 9000) {
49 fprintf(stderr
, Name
": need md driver version 0.90.0 or later\n");
53 /* If this is an externally-manage array, we need to modify the
54 * metadata_version so that mdmon doesn't undo our change.
56 mdi
= sysfs_read(fd
, -1, GET_LEVEL
|GET_VERSION
);
58 mdi
->array
.major_version
== -1 &&
59 mdi
->array
.level
> 0 &&
60 is_subarray(mdi
->text_version
)) {
62 strcpy(vers
, "external:");
63 strcat(vers
, mdi
->text_version
);
66 /* We set readonly ourselves. */
68 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
71 rv
= sysfs_set_str(mdi
, NULL
, "array_state", "readonly");
74 fprintf(stderr
, Name
": failed to set readonly for %s: %s\n",
75 devname
, strerror(errno
));
77 vers
[9] = mdi
->text_version
[0];
78 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
83 /* We cannot set read/write - must signal mdmon */
85 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
87 cp
= strchr(vers
+10, '/');
90 ping_monitor(vers
+10);
95 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
96 fprintf(stderr
, Name
": %s does not appear to be active.\n",
102 if (ioctl(fd
, STOP_ARRAY_RO
, NULL
)) {
103 fprintf(stderr
, Name
": failed to set readonly for %s: %s\n",
104 devname
, strerror(errno
));
107 } else if (readonly
< 0) {
108 if (ioctl(fd
, RESTART_ARRAY_RW
, NULL
)) {
109 fprintf(stderr
, Name
": failed to set writable for %s: %s\n",
110 devname
, strerror(errno
));
119 static void remove_devices(int devnum
, char *path
)
121 /* Remove all 'standard' devices for 'devnum', including
122 * partitions. Also remove names at 'path' - possibly with
123 * partition suffixes - which link to those names.
134 sprintf(base
, "/dev/md%d", devnum
);
136 sprintf(base
, "/dev/md_d%d", -1-devnum
);
137 be
= base
+ strlen(base
);
139 path2
= malloc(strlen(path
)+20);
141 pe
= path2
+ strlen(path2
);
145 for (part
= 0; part
< 16; part
++) {
147 sprintf(be
, "p%d", part
);
150 sprintf(pe
, "p%d", part
);
152 sprintf(pe
, "%d", part
);
155 /* FIXME test if really is md device ?? */
158 n
= readlink(path2
, link
, sizeof(link
));
159 if (n
&& strlen(base
) == n
&&
160 strncmp(link
, base
, n
) == 0)
168 int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
)
170 /* Run or stop the array. array must already be configured
172 * Only print failure messages if quiet == 0;
173 * quiet > 0 means really be quiet
174 * quiet < 0 means we will try again if it fails.
176 mdu_param_t param
; /* unused */
178 if (runstop
== -1 && md_get_version(fd
) < 9000) {
179 if (ioctl(fd
, STOP_MD
, 0)) {
180 if (quiet
== 0) fprintf(stderr
,
181 Name
": stopping device %s "
183 devname
, strerror(errno
));
188 if (md_get_version(fd
) < 9000) {
189 fprintf(stderr
, Name
": need md driver version 0.90.0 or later\n");
193 if (ioctl(fd, GET_ARRAY_INFO, &array)) {
194 fprintf(stderr, Name ": %s does not appear to be active.\n",
200 if (ioctl(fd
, RUN_ARRAY
, ¶m
)) {
201 fprintf(stderr
, Name
": failed to run array %s: %s\n",
202 devname
, strerror(errno
));
206 fprintf(stderr
, Name
": started %s\n", devname
);
207 } else if (runstop
< 0){
208 struct map_ent
*map
= NULL
;
212 /* If this is an mdmon managed array, just write 'inactive'
213 * to the array state and let mdmon clear up.
215 devnum
= fd2devnum(fd
);
216 mdi
= sysfs_read(fd
, -1, GET_LEVEL
|GET_VERSION
);
218 mdi
->array
.level
> 0 &&
219 is_subarray(mdi
->text_version
)) {
220 /* This is mdmon managed. */
222 if (sysfs_set_str(mdi
, NULL
,
223 "array_state", "inactive") < 0) {
226 ": failed to stop array %s: %s\n",
227 devname
, strerror(errno
));
231 /* Give monitor a chance to act */
232 ping_monitor(mdi
->text_version
);
234 fd
= open(devname
, O_RDONLY
);
236 mdi
->array
.major_version
== -1 &&
237 mdi
->array
.minor_version
== -2 &&
238 !is_subarray(mdi
->text_version
)) {
239 /* container, possibly mdmon-managed.
240 * Make sure mdmon isn't opening it, which
241 * would interfere with the 'stop'
243 ping_monitor(mdi
->sys_name
);
246 if (fd
>= 0 && ioctl(fd
, STOP_ARRAY
, NULL
)) {
249 ": failed to stop array %s: %s\n",
250 devname
, strerror(errno
));
252 fprintf(stderr
, "Perhaps a running "
253 "process, mounted filesystem "
254 "or active volume group?\n");
260 /* prior to 2.6.28, KOBJ_CHANGE was not sent when an md array
261 * was stopped, so We'll do it here just to be sure. Drop any
262 * partitions as well...
265 ioctl(fd
, BLKRRPART
, 0);
267 sysfs_uevent(mdi
, "change");
270 if (devnum
!= NoMdDev
&&
271 (stat("/dev/.udev", &stb
) != 0 ||
272 check_env("MDADM_NO_UDEV"))) {
273 struct map_ent
*mp
= map_by_devnum(&map
, devnum
);
274 remove_devices(devnum
, mp
? mp
->path
: NULL
);
279 fprintf(stderr
, Name
": stopped %s\n", devname
);
280 if (devnum
!= NoMdDev
) {
281 map_delete(&map
, devnum
);
289 int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
)
291 mdu_array_info_t info
;
292 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
293 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
294 devname
, strerror(errno
));
300 info
.raid_disks
= raid_disks
;
301 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
302 fprintf(stderr
, Name
": Cannot set device size/shape for %s: %s\n",
303 devname
, strerror(errno
));
309 int Manage_subdevs(char *devname
, int fd
,
310 mddev_dev_t devlist
, int verbose
)
312 /* do something to each dev.
314 * 'a' - add the device
316 * If that fails EINVAL, try ADD_NEW_DISK
317 * 'r' - remove the device HOT_REMOVE_DISK
318 * device can be 'faulty' or 'detached' in which case all
319 * matching devices are removed.
320 * 'f' - set the device faulty SET_DISK_FAULTY
321 * device can be 'detached' in which case any device that
322 * is inaccessible will be marked faulty.
324 mdu_array_info_t array
;
325 mdu_disk_info_t disc
;
326 unsigned long long array_size
;
327 mddev_dev_t dv
, next
= NULL
;
331 struct supertype
*st
, *tst
;
336 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
337 fprintf(stderr
, Name
": cannot get array info for %s\n",
342 /* array.size is only 32 bit and may be truncated.
343 * So read from sysfs if possible, and record number of sectors
346 array_size
= get_component_size(fd
);
348 array_size
= array
.size
* 2;
350 tst
= super_by_fd(fd
);
352 fprintf(stderr
, Name
": unsupport array - version %d.%d\n",
353 array
.major_version
, array
.minor_version
);
357 for (dv
= devlist
, j
=0 ; dv
; dv
= next
, j
= jnext
) {
358 unsigned long long ldsize
;
360 char *dnprintable
= dv
->devname
;
366 if (strcmp(dv
->devname
, "failed")==0 ||
367 strcmp(dv
->devname
, "faulty")==0) {
368 if (dv
->disposition
!= 'r') {
369 fprintf(stderr
, Name
": %s only meaningful "
370 "with -r, not -%c\n",
371 dv
->devname
, dv
->disposition
);
374 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
376 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
378 if (disc
.major
== 0 && disc
.minor
== 0)
380 if ((disc
.state
& 1) == 0) /* faulty */
382 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
385 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
386 dnprintable
= dvname
;
391 } else if (strcmp(dv
->devname
, "detached") == 0) {
392 if (dv
->disposition
!= 'r' && dv
->disposition
!= 'f') {
393 fprintf(stderr
, Name
": %s only meaningful "
394 "with -r of -f, not -%c\n",
395 dv
->devname
, dv
->disposition
);
398 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
401 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
403 if (disc
.major
== 0 && disc
.minor
== 0)
405 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
406 sfd
= dev_open(dvname
, O_RDONLY
);
411 if (dv
->disposition
== 'f' &&
412 (disc
.state
& 1) == 1) /* already faulty */
416 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
419 dnprintable
= dvname
;
427 tfd
= dev_open(dv
->devname
, O_RDONLY
);
428 if (tfd
< 0 || fstat(tfd
, &stb
) != 0) {
429 fprintf(stderr
, Name
": cannot find %s: %s\n",
430 dv
->devname
, strerror(errno
));
436 if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
437 fprintf(stderr
, Name
": %s is not a "
443 switch(dv
->disposition
){
445 fprintf(stderr
, Name
": internal error - devmode[%s]=%d\n",
446 dv
->devname
, dv
->disposition
);
450 if (tst
->subarray
[0]) {
451 fprintf(stderr
, Name
": Cannot add disks to a"
452 " \'member\' array, perform this"
453 " operation on the parent container\n");
456 /* Make sure it isn't in use (in 2.6 or later) */
457 tfd
= dev_open(dv
->devname
, O_RDONLY
|O_EXCL
|O_DIRECT
);
459 fprintf(stderr
, Name
": Cannot open %s: %s\n",
460 dv
->devname
, strerror(errno
));
463 remove_partitions(tfd
);
467 if (array
.not_persistent
==0)
468 st
->ss
->load_super(st
, tfd
, NULL
);
470 if (!get_dev_size(tfd
, dv
->devname
, &ldsize
)) {
477 if (!tst
->ss
->external
&&
478 array
.major_version
== 0 &&
479 md_get_version(fd
)%100 < 2) {
480 if (ioctl(fd
, HOT_ADD_DISK
,
481 (unsigned long)stb
.st_rdev
)==0) {
483 fprintf(stderr
, Name
": hot added %s\n",
488 fprintf(stderr
, Name
": hot add failed for %s: %s\n",
489 dv
->devname
, strerror(errno
));
493 if (array
.not_persistent
== 0 || tst
->ss
->external
) {
495 /* need to find a sample superblock to copy, and
496 * a spare slot to use.
497 * For 'external' array (well, container based),
498 * We can just load the metadata for the array.
500 if (tst
->ss
->external
) {
501 tst
->ss
->load_super(tst
, fd
, NULL
);
502 } else for (j
= 0; j
< tst
->max_devs
; j
++) {
506 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
508 if (disc
.major
==0 && disc
.minor
==0)
510 if ((disc
.state
& 4)==0) continue; /* sync */
511 /* Looks like a good device to try */
512 dev
= map_dev(disc
.major
, disc
.minor
, 1);
514 dfd
= dev_open(dev
, O_RDONLY
);
515 if (dfd
< 0) continue;
516 if (tst
->ss
->load_super(tst
, dfd
,
524 /* FIXME this is a bad test to be using */
526 fprintf(stderr
, Name
": cannot find valid superblock in this array - HELP\n");
530 /* Make sure device is large enough */
531 if (tst
->ss
->avail_size(tst
, ldsize
/512) <
533 fprintf(stderr
, Name
": %s not large enough to join array\n",
538 /* Possibly this device was recently part of the array
539 * and was temporarily removed, and is now being re-added.
540 * If so, we can simply re-add it.
542 tst
->ss
->uuid_from_super(tst
, duuid
);
544 /* re-add doesn't work for version-1 superblocks
547 if (array
.major_version
== 1 &&
548 get_linux_version() <= 2006018)
551 st
->ss
->uuid_from_super(st
, ouuid
);
552 if (memcmp(duuid
, ouuid
, sizeof(ouuid
))==0) {
553 /* looks close enough for now. Kernel
554 * will worry about whether a bitmap
555 * based reconstruction is possible.
558 st
->ss
->getinfo_super(st
, &mdi
);
559 disc
.major
= major(stb
.st_rdev
);
560 disc
.minor
= minor(stb
.st_rdev
);
561 disc
.number
= mdi
.disk
.number
;
562 disc
.raid_disk
= mdi
.disk
.raid_disk
;
563 disc
.state
= mdi
.disk
.state
;
564 if (dv
->writemostly
== 1)
565 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
566 if (dv
->writemostly
== 2)
567 disc
.state
&= ~(1 << MD_DISK_WRITEMOSTLY
);
568 /* don't even try if disk is marked as faulty */
570 if ((disc
.state
& 1) == 0 &&
571 ioctl(fd
, ADD_NEW_DISK
, &disc
) == 0) {
573 fprintf(stderr
, Name
": re-added %s\n", dv
->devname
);
576 if (errno
== ENOMEM
|| errno
== EROFS
) {
577 fprintf(stderr
, Name
": add new device failed for %s: %s\n",
578 dv
->devname
, strerror(errno
));
581 /* fall back on normal-add */
586 ": --re-add for %s to %s is not possible\n",
587 dv
->devname
, devname
);
591 /* non-persistent. Must ensure that new drive
592 * is at least array.size big.
594 if (ldsize
/512 < array_size
) {
595 fprintf(stderr
, Name
": %s not large enough to join array\n",
600 /* in 2.6.17 and earlier, version-1 superblocks won't
601 * use the number we write, but will choose a free number.
602 * we must choose the same free number, which requires
603 * starting at 'raid_disks' and counting up
605 for (j
= array
.raid_disks
; j
< tst
->max_devs
; j
++) {
607 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
609 if (disc
.major
==0 && disc
.minor
==0)
611 if (disc
.state
& 8) /* removed */
614 disc
.major
= major(stb
.st_rdev
);
615 disc
.minor
= minor(stb
.st_rdev
);
618 if (array
.not_persistent
==0 || tst
->ss
->external
) {
620 if (dv
->writemostly
== 1)
621 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
622 dfd
= dev_open(dv
->devname
, O_RDWR
| O_EXCL
|O_DIRECT
);
623 if (tst
->ss
->add_to_super(tst
, &disc
, dfd
,
628 /* write_init_super will close 'dfd' */
629 if (tst
->ss
->external
)
630 /* mdmon will write the metadata */
632 else if (tst
->ss
->write_init_super(tst
))
634 } else if (dv
->re_add
) {
635 /* this had better be raid1.
636 * As we are "--re-add"ing we must find a spare slot
639 char *used
= malloc(array
.raid_disks
);
640 memset(used
, 0, array
.raid_disks
);
641 for (j
=0; j
< tst
->max_devs
; j
++) {
642 mdu_disk_info_t disc2
;
644 if (ioctl(fd
, GET_DISK_INFO
, &disc2
))
646 if (disc2
.major
==0 && disc2
.minor
==0)
648 if (disc2
.state
& 8) /* removed */
650 if (disc2
.raid_disk
< 0)
652 if (disc2
.raid_disk
> array
.raid_disks
)
654 used
[disc2
.raid_disk
] = 1;
656 for (j
=0 ; j
<array
.raid_disks
; j
++)
659 disc
.state
|= (1<<MD_DISK_SYNC
);
664 if (dv
->writemostly
== 1)
665 disc
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
666 if (tst
->ss
->external
) {
667 /* add a disk to an external metadata container
668 * only if mdmon is around to see it
670 struct mdinfo new_mdi
;
673 int devnum
= fd2devnum(fd
);
675 container_fd
= open_dev_excl(devnum
);
676 if (container_fd
< 0) {
677 fprintf(stderr
, Name
": add failed for %s:"
678 " could not get exclusive access to container\n",
683 if (!mdmon_running(devnum
)) {
684 fprintf(stderr
, Name
": add failed for %s: mdmon not running\n",
690 sra
= sysfs_read(container_fd
, -1, 0);
692 fprintf(stderr
, Name
": add failed for %s: sysfs_read failed\n",
697 sra
->array
.level
= LEVEL_CONTAINER
;
698 /* Need to set data_offset and component_size */
699 tst
->ss
->getinfo_super(tst
, &new_mdi
);
700 new_mdi
.disk
.major
= disc
.major
;
701 new_mdi
.disk
.minor
= disc
.minor
;
702 new_mdi
.recovery_start
= 0;
703 if (sysfs_add_disk(sra
, &new_mdi
, 0) != 0) {
704 fprintf(stderr
, Name
": add new device to external metadata"
705 " failed for %s\n", dv
->devname
);
709 ping_monitor(devnum2devname(devnum
));
712 } else if (ioctl(fd
, ADD_NEW_DISK
, &disc
)) {
713 fprintf(stderr
, Name
": add new device failed for %s as %d: %s\n",
714 dv
->devname
, j
, strerror(errno
));
718 fprintf(stderr
, Name
": added %s\n", dv
->devname
);
723 if (tst
->subarray
[0]) {
724 fprintf(stderr
, Name
": Cannot remove disks from a"
725 " \'member\' array, perform this"
726 " operation on the parent container\n");
729 if (tst
->ss
->external
) {
730 /* To remove a device from a container, we must
731 * check that it isn't in use in an array.
732 * This involves looking in the 'holders'
733 * directory - there must be just one entry,
735 * To ensure that it doesn't get used as a
736 * hold spare while we are checking, we
737 * get an O_EXCL open on the container
739 int dnum
= fd2devnum(fd
);
740 lfd
= open_dev_excl(dnum
);
743 ": Cannot get exclusive access "
744 " to container - odd\n");
747 /* in the detached case it is not possible to
748 * check if we are the unique holder, so just
749 * rely on the 'detached' checks
751 if (strcmp(dv
->devname
, "detached") == 0 ||
752 sysfs_unique_holder(dnum
, stb
.st_rdev
))
756 ": %s is %s, cannot remove.\n",
758 errno
== EEXIST
? "still in use":
764 /* FIXME check that it is a current member */
765 err
= ioctl(fd
, HOT_REMOVE_DISK
, (unsigned long)stb
.st_rdev
);
766 if (err
&& errno
== ENODEV
) {
767 /* Old kernels rejected this if no personality
769 struct mdinfo
*sra
= sysfs_read(fd
, 0, GET_DEVS
);
770 struct mdinfo
*dv
= NULL
;
773 for ( ; dv
; dv
=dv
->next
)
774 if (dv
->disk
.major
== major(stb
.st_rdev
) &&
775 dv
->disk
.minor
== minor(stb
.st_rdev
))
778 err
= sysfs_set_str(sra
, dv
,
786 fprintf(stderr
, Name
": hot remove failed "
787 "for %s: %s\n", dnprintable
,
793 if (tst
->ss
->external
) {
795 * Before dropping our exclusive open we make an
796 * attempt at preventing mdmon from seeing an
797 * 'add' event before reconciling this 'remove'
800 char *name
= devnum2devname(fd2devnum(fd
));
803 fprintf(stderr
, Name
": unable to get container name\n");
812 fprintf(stderr
, Name
": hot removed %s\n",
816 case 'f': /* set faulty */
817 /* FIXME check current member */
818 if (ioctl(fd
, SET_DISK_FAULTY
, (unsigned long) stb
.st_rdev
)) {
819 fprintf(stderr
, Name
": set device faulty failed for %s: %s\n",
820 dnprintable
, strerror(errno
));
824 fprintf(stderr
, Name
": set %s faulty in %s\n",
825 dnprintable
, devname
);
835 /* Open any md device, and issue the RAID_AUTORUN ioctl */
837 int fd
= dev_open("9:0", O_RDONLY
);
839 if (ioctl(fd
, RAID_AUTORUN
, 0) == 0)