]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Manage.c
3aa09bcba241caba52b989abd6804de9e65a056e
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)
167 int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
)
169 /* Run or stop the array. array must already be configured
171 * Only print failure messages if quiet == 0;
172 * quiet > 0 means really be quiet
173 * quiet < 0 means we will try again if it fails.
175 mdu_param_t param
; /* unused */
177 if (runstop
== -1 && md_get_version(fd
) < 9000) {
178 if (ioctl(fd
, STOP_MD
, 0)) {
179 if (quiet
== 0) fprintf(stderr
,
180 Name
": stopping device %s "
182 devname
, strerror(errno
));
187 if (md_get_version(fd
) < 9000) {
188 fprintf(stderr
, Name
": need md driver version 0.90.0 or later\n");
192 if (ioctl(fd, GET_ARRAY_INFO, &array)) {
193 fprintf(stderr, Name ": %s does not appear to be active.\n",
199 if (ioctl(fd
, RUN_ARRAY
, ¶m
)) {
200 fprintf(stderr
, Name
": failed to run array %s: %s\n",
201 devname
, strerror(errno
));
205 fprintf(stderr
, Name
": started %s\n", devname
);
206 } else if (runstop
< 0){
207 struct map_ent
*map
= NULL
;
211 /* If this is an mdmon managed array, just write 'inactive'
212 * to the array state and let mdmon clear up.
214 devnum
= fd2devnum(fd
);
215 mdi
= sysfs_read(fd
, -1, GET_LEVEL
|GET_VERSION
);
217 mdi
->array
.level
> 0 &&
218 is_subarray(mdi
->text_version
)) {
219 /* This is mdmon managed. */
221 if (sysfs_set_str(mdi
, NULL
,
222 "array_state", "inactive") < 0) {
225 ": failed to stop array %s: %s\n",
226 devname
, strerror(errno
));
230 /* Give monitor a chance to act */
231 ping_monitor(mdi
->text_version
);
233 fd
= open(devname
, O_RDONLY
);
235 mdi
->array
.major_version
== -1 &&
236 mdi
->array
.minor_version
== -2 &&
237 !is_subarray(mdi
->text_version
)) {
238 /* container, possibly mdmon-managed.
239 * Make sure mdmon isn't opening it, which
240 * would interfere with the 'stop'
242 ping_monitor(mdi
->sys_name
);
245 if (fd
>= 0 && ioctl(fd
, STOP_ARRAY
, NULL
)) {
248 ": failed to stop array %s: %s\n",
249 devname
, strerror(errno
));
251 fprintf(stderr
, "Perhaps a running "
252 "process, mounted filesystem "
253 "or active volume group?\n");
259 /* prior to 2.6.28, KOBJ_CHANGE was not sent when an md array
260 * was stopped, so We'll do it here just to be sure. Drop any
261 * partitions as well...
264 ioctl(fd
, BLKRRPART
, 0);
266 sysfs_uevent(mdi
, "change");
269 if (devnum
!= NoMdDev
&&
270 (stat("/dev/.udev", &stb
) != 0 ||
271 check_env("MDADM_NO_UDEV"))) {
272 struct map_ent
*mp
= map_by_devnum(&map
, devnum
);
273 remove_devices(devnum
, mp
? mp
->path
: NULL
);
278 fprintf(stderr
, Name
": stopped %s\n", devname
);
279 if (devnum
!= NoMdDev
) {
280 map_delete(&map
, devnum
);
288 int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
)
290 mdu_array_info_t info
;
291 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
292 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
293 devname
, strerror(errno
));
299 info
.raid_disks
= raid_disks
;
300 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
301 fprintf(stderr
, Name
": Cannot set device size/shape for %s: %s\n",
302 devname
, strerror(errno
));
308 int Manage_reconfig(char *devname
, int fd
, int layout
)
310 mdu_array_info_t info
;
311 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
312 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
313 devname
, strerror(errno
));
316 info
.layout
= layout
;
317 printf("layout set to %d\n", info
.layout
);
318 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
319 fprintf(stderr
, Name
": Cannot set layout for %s: %s\n",
320 devname
, strerror(errno
));
326 int Manage_subdevs(char *devname
, int fd
,
327 mddev_dev_t devlist
, int verbose
)
329 /* do something to each dev.
331 * 'a' - add the device
333 * If that fails EINVAL, try ADD_NEW_DISK
334 * 'r' - remove the device HOT_REMOVE_DISK
335 * device can be 'faulty' or 'detached' in which case all
336 * matching devices are removed.
337 * 'f' - set the device faulty SET_DISK_FAULTY
338 * device can be 'detached' in which case any device that
339 * is inaccessible will be marked faulty.
341 mdu_array_info_t array
;
342 mdu_disk_info_t disc
;
343 unsigned long long array_size
;
344 mddev_dev_t dv
, next
= NULL
;
348 struct supertype
*st
, *tst
;
353 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
354 fprintf(stderr
, Name
": cannot get array info for %s\n",
359 /* array.size is only 32 bit and may be truncated.
360 * So read from sysfs if possible, and record number of sectors
363 array_size
= get_component_size(fd
);
365 array_size
= array
.size
* 2;
367 tst
= super_by_fd(fd
);
369 fprintf(stderr
, Name
": unsupport array - version %d.%d\n",
370 array
.major_version
, array
.minor_version
);
374 for (dv
= devlist
, j
=0 ; dv
; dv
= next
, j
= jnext
) {
375 unsigned long long ldsize
;
377 char *dnprintable
= dv
->devname
;
383 if (strcmp(dv
->devname
, "failed")==0 ||
384 strcmp(dv
->devname
, "faulty")==0) {
385 if (dv
->disposition
!= 'r') {
386 fprintf(stderr
, Name
": %s only meaningful "
387 "with -r, not -%c\n",
388 dv
->devname
, dv
->disposition
);
391 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
393 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
395 if (disc
.major
== 0 && disc
.minor
== 0)
397 if ((disc
.state
& 1) == 0) /* faulty */
399 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
402 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
403 dnprintable
= dvname
;
408 } else if (strcmp(dv
->devname
, "detached") == 0) {
409 if (dv
->disposition
!= 'r' && dv
->disposition
!= 'f') {
410 fprintf(stderr
, Name
": %s only meaningful "
411 "with -r of -f, not -%c\n",
412 dv
->devname
, dv
->disposition
);
415 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
418 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
420 if (disc
.major
== 0 && disc
.minor
== 0)
422 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
423 sfd
= dev_open(dvname
, O_RDONLY
);
428 if (dv
->disposition
== 'f' &&
429 (disc
.state
& 1) == 1) /* already faulty */
433 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
436 dnprintable
= dvname
;
444 tfd
= dev_open(dv
->devname
, O_RDONLY
);
445 if (tfd
< 0 || fstat(tfd
, &stb
) != 0) {
446 fprintf(stderr
, Name
": cannot find %s: %s\n",
447 dv
->devname
, strerror(errno
));
453 if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
454 fprintf(stderr
, Name
": %s is not a "
460 switch(dv
->disposition
){
462 fprintf(stderr
, Name
": internal error - devmode[%s]=%d\n",
463 dv
->devname
, dv
->disposition
);
467 if (tst
->subarray
[0]) {
468 fprintf(stderr
, Name
": Cannot add disks to a"
469 " \'member\' array, perform this"
470 " operation on the parent container\n");
473 /* Make sure it isn't in use (in 2.6 or later) */
474 tfd
= dev_open(dv
->devname
, O_RDONLY
|O_EXCL
|O_DIRECT
);
476 fprintf(stderr
, Name
": Cannot open %s: %s\n",
477 dv
->devname
, strerror(errno
));
480 remove_partitions(tfd
);
484 if (array
.not_persistent
==0)
485 st
->ss
->load_super(st
, tfd
, NULL
);
487 if (!get_dev_size(tfd
, dv
->devname
, &ldsize
)) {
494 if (!tst
->ss
->external
&&
495 array
.major_version
== 0 &&
496 md_get_version(fd
)%100 < 2) {
497 if (ioctl(fd
, HOT_ADD_DISK
,
498 (unsigned long)stb
.st_rdev
)==0) {
500 fprintf(stderr
, Name
": hot added %s\n",
505 fprintf(stderr
, Name
": hot add failed for %s: %s\n",
506 dv
->devname
, strerror(errno
));
510 if (array
.not_persistent
== 0 || tst
->ss
->external
) {
512 /* need to find a sample superblock to copy, and
513 * a spare slot to use.
514 * For 'external' array (well, container based),
515 * We can just load the metadata for the array.
517 if (tst
->ss
->external
) {
518 tst
->ss
->load_super(tst
, fd
, NULL
);
519 } else for (j
= 0; j
< tst
->max_devs
; j
++) {
523 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
525 if (disc
.major
==0 && disc
.minor
==0)
527 if ((disc
.state
& 4)==0) continue; /* sync */
528 /* Looks like a good device to try */
529 dev
= map_dev(disc
.major
, disc
.minor
, 1);
531 dfd
= dev_open(dev
, O_RDONLY
);
532 if (dfd
< 0) continue;
533 if (tst
->ss
->load_super(tst
, dfd
,
541 /* FIXME this is a bad test to be using */
543 fprintf(stderr
, Name
": cannot find valid superblock in this array - HELP\n");
547 /* Make sure device is large enough */
548 if (tst
->ss
->avail_size(tst
, ldsize
/512) <
550 fprintf(stderr
, Name
": %s not large enough to join array\n",
555 /* Possibly this device was recently part of the array
556 * and was temporarily removed, and is now being re-added.
557 * If so, we can simply re-add it.
559 tst
->ss
->uuid_from_super(tst
, duuid
);
561 /* re-add doesn't work for version-1 superblocks
564 if (array
.major_version
== 1 &&
565 get_linux_version() <= 2006018)
568 st
->ss
->uuid_from_super(st
, ouuid
);
569 if (memcmp(duuid
, ouuid
, sizeof(ouuid
))==0) {
570 /* looks close enough for now. Kernel
571 * will worry about whether a bitmap
572 * based reconstruction is possible.
575 st
->ss
->getinfo_super(st
, &mdi
);
576 disc
.major
= major(stb
.st_rdev
);
577 disc
.minor
= minor(stb
.st_rdev
);
578 disc
.number
= mdi
.disk
.number
;
579 disc
.raid_disk
= mdi
.disk
.raid_disk
;
580 disc
.state
= mdi
.disk
.state
;
581 if (dv
->writemostly
== 1)
582 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
583 if (dv
->writemostly
== 2)
584 disc
.state
&= ~(1 << MD_DISK_WRITEMOSTLY
);
585 if (ioctl(fd
, ADD_NEW_DISK
, &disc
) == 0) {
587 fprintf(stderr
, Name
": re-added %s\n", dv
->devname
);
590 if (errno
== ENOMEM
|| errno
== EROFS
) {
591 fprintf(stderr
, Name
": add new device failed for %s: %s\n",
592 dv
->devname
, strerror(errno
));
595 /* fall back on normal-add */
599 /* non-persistent. Must ensure that new drive
600 * is at least array.size big.
602 if (ldsize
/512 < array_size
) {
603 fprintf(stderr
, Name
": %s not large enough to join array\n",
608 /* in 2.6.17 and earlier, version-1 superblocks won't
609 * use the number we write, but will choose a free number.
610 * we must choose the same free number, which requires
611 * starting at 'raid_disks' and counting up
613 for (j
= array
.raid_disks
; j
< tst
->max_devs
; j
++) {
615 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
617 if (disc
.major
==0 && disc
.minor
==0)
619 if (disc
.state
& 8) /* removed */
622 disc
.major
= major(stb
.st_rdev
);
623 disc
.minor
= minor(stb
.st_rdev
);
626 if (array
.not_persistent
==0 || tst
->ss
->external
) {
628 if (dv
->writemostly
== 1)
629 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
630 dfd
= dev_open(dv
->devname
, O_RDWR
| O_EXCL
|O_DIRECT
);
631 if (tst
->ss
->add_to_super(tst
, &disc
, dfd
,
636 /* write_init_super will close 'dfd' */
637 if (tst
->ss
->external
)
638 /* mdmon will write the metadata */
640 else if (tst
->ss
->write_init_super(tst
))
642 } else if (dv
->re_add
) {
643 /* this had better be raid1.
644 * As we are "--re-add"ing we must find a spare slot
647 char *used
= malloc(array
.raid_disks
);
648 memset(used
, 0, array
.raid_disks
);
649 for (j
=0; j
< tst
->max_devs
; j
++) {
650 mdu_disk_info_t disc2
;
652 if (ioctl(fd
, GET_DISK_INFO
, &disc2
))
654 if (disc2
.major
==0 && disc2
.minor
==0)
656 if (disc2
.state
& 8) /* removed */
658 if (disc2
.raid_disk
< 0)
660 if (disc2
.raid_disk
> array
.raid_disks
)
662 used
[disc2
.raid_disk
] = 1;
664 for (j
=0 ; j
<array
.raid_disks
; j
++)
667 disc
.state
|= (1<<MD_DISK_SYNC
);
671 if (dv
->writemostly
== 1)
672 disc
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
673 if (tst
->ss
->external
) {
674 /* add a disk to an external metadata container
675 * only if mdmon is around to see it
677 struct mdinfo new_mdi
;
680 int devnum
= fd2devnum(fd
);
682 container_fd
= open_dev_excl(devnum
);
683 if (container_fd
< 0) {
684 fprintf(stderr
, Name
": add failed for %s:"
685 " could not get exclusive access to container\n",
690 if (!mdmon_running(devnum
)) {
691 fprintf(stderr
, Name
": add failed for %s: mdmon not running\n",
697 sra
= sysfs_read(container_fd
, -1, 0);
699 fprintf(stderr
, Name
": add failed for %s: sysfs_read failed\n",
704 sra
->array
.level
= LEVEL_CONTAINER
;
705 /* Need to set data_offset and component_size */
706 tst
->ss
->getinfo_super(tst
, &new_mdi
);
707 new_mdi
.disk
.major
= disc
.major
;
708 new_mdi
.disk
.minor
= disc
.minor
;
709 if (sysfs_add_disk(sra
, &new_mdi
, 0) != 0) {
710 fprintf(stderr
, Name
": add new device to external metadata"
711 " failed for %s\n", dv
->devname
);
715 ping_monitor(devnum2devname(devnum
));
718 } else if (ioctl(fd
, ADD_NEW_DISK
, &disc
)) {
719 fprintf(stderr
, Name
": add new device failed for %s as %d: %s\n",
720 dv
->devname
, j
, strerror(errno
));
724 fprintf(stderr
, Name
": added %s\n", dv
->devname
);
729 if (tst
->subarray
[0]) {
730 fprintf(stderr
, Name
": Cannot remove disks from a"
731 " \'member\' array, perform this"
732 " operation on the parent container\n");
735 if (tst
->ss
->external
) {
736 /* To remove a device from a container, we must
737 * check that it isn't in use in an array.
738 * This involves looking in the 'holders'
739 * directory - there must be just one entry,
741 * To ensure that it doesn't get used as a
742 * hold spare while we are checking, we
743 * get an O_EXCL open on the container
745 int dnum
= fd2devnum(fd
);
746 lfd
= open_dev_excl(dnum
);
749 ": Cannot get exclusive access "
750 " to container - odd\n");
753 /* in the detached case it is not possible to
754 * check if we are the unique holder, so just
755 * rely on the 'detached' checks
757 if (strcmp(dv
->devname
, "detached") == 0 ||
758 sysfs_unique_holder(dnum
, stb
.st_rdev
))
762 ": %s is %s, cannot remove.\n",
764 errno
== EEXIST
? "still in use":
770 /* FIXME check that it is a current member */
771 err
= ioctl(fd
, HOT_REMOVE_DISK
, (unsigned long)stb
.st_rdev
);
772 if (err
&& errno
== ENODEV
) {
773 /* Old kernels rejected this if no personality
775 struct mdinfo
*sra
= sysfs_read(fd
, 0, GET_DEVS
);
776 struct mdinfo
*dv
= NULL
;
779 for ( ; dv
; dv
=dv
->next
)
780 if (dv
->disk
.major
== major(stb
.st_rdev
) &&
781 dv
->disk
.minor
== minor(stb
.st_rdev
))
784 err
= sysfs_set_str(sra
, dv
,
792 fprintf(stderr
, Name
": hot remove failed "
793 "for %s: %s\n", dnprintable
,
799 if (tst
->ss
->external
) {
801 * Before dropping our exclusive open we make an
802 * attempt at preventing mdmon from seeing an
803 * 'add' event before reconciling this 'remove'
806 char *name
= devnum2devname(fd2devnum(fd
));
809 fprintf(stderr
, Name
": unable to get container name\n");
818 fprintf(stderr
, Name
": hot removed %s\n",
822 case 'f': /* set faulty */
823 /* FIXME check current member */
824 if (ioctl(fd
, SET_DISK_FAULTY
, (unsigned long) stb
.st_rdev
)) {
825 fprintf(stderr
, Name
": set device faulty failed for %s: %s\n",
826 dnprintable
, strerror(errno
));
830 fprintf(stderr
, Name
": set %s faulty in %s\n",
831 dnprintable
, devname
);
841 /* Open any md device, and issue the RAID_AUTORUN ioctl */
843 int fd
= dev_open("9:0", O_RDONLY
);
845 if (ioctl(fd
, RAID_AUTORUN
, 0) == 0)