]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Manage.c
2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2006 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@cse.unsw.edu.au>
24 * School of Computer Science and Engineering
25 * The University of New South Wales
34 #define REGISTER_DEV _IO (MD_MAJOR, 1)
35 #define START_MD _IO (MD_MAJOR, 2)
36 #define STOP_MD _IO (MD_MAJOR, 3)
38 int Manage_ro(char *devname
, int fd
, int readonly
)
40 /* switch to readonly or rw
43 * first check that array is runing
44 * use RESTART_ARRAY_RW or STOP_ARRAY_RO
47 mdu_array_info_t array
;
50 if (md_get_version(fd
) < 9000) {
51 fprintf(stderr
, Name
": need md driver version 0.90.0 or later\n");
54 /* If this is an externally-manage array, we need to modify the
55 * metadata_version so that mdmon doesn't undo our change.
57 mdi
= sysfs_read(fd
, -1, GET_LEVEL
|GET_VERSION
);
59 mdi
->array
.major_version
== -1 &&
60 mdi
->array
.level
> 0 &&
61 is_subarray(mdi
->text_version
)) {
63 strcpy(vers
, "external:");
64 strcat(vers
, mdi
->text_version
);
67 /* We set readonly ourselves. */
69 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
72 rv
= sysfs_set_str(mdi
, NULL
, "array_state", "readonly");
75 fprintf(stderr
, Name
": failed to set readonly for %s: %s\n",
76 devname
, strerror(errno
));
78 vers
[9] = mdi
->text_version
[0];
79 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
84 /* We cannot set read/write - must signal mdmon */
86 sysfs_set_str(mdi
, NULL
, "metadata_version", vers
);
88 cp
= strchr(vers
+10, '/');
91 ping_monitor(vers
+10);
96 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
97 fprintf(stderr
, Name
": %s does not appear to be active.\n",
103 if (ioctl(fd
, STOP_ARRAY_RO
, NULL
)) {
104 fprintf(stderr
, Name
": failed to set readonly for %s: %s\n",
105 devname
, strerror(errno
));
108 } else if (readonly
< 0) {
109 if (ioctl(fd
, RESTART_ARRAY_RW
, NULL
)) {
110 fprintf(stderr
, Name
": failed to set writable for %s: %s\n",
111 devname
, strerror(errno
));
120 int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
)
122 /* Run or stop the array. array must already be configured
124 * Only print failure messages if quiet == 0;
125 * quiet > 0 means really be quiet
126 * quiet < 0 means we will try again if it fails.
128 mdu_param_t param
; /* unused */
130 if (runstop
== -1 && md_get_version(fd
) < 9000) {
131 if (ioctl(fd
, STOP_MD
, 0)) {
132 if (quiet
== 0) fprintf(stderr
,
133 Name
": stopping device %s "
135 devname
, strerror(errno
));
140 if (md_get_version(fd
) < 9000) {
141 fprintf(stderr
, Name
": need md driver version 0.90.0 or later\n");
145 if (ioctl(fd, GET_ARRAY_INFO, &array)) {
146 fprintf(stderr, Name ": %s does not appear to be active.\n",
152 if (ioctl(fd
, RUN_ARRAY
, ¶m
)) {
153 fprintf(stderr
, Name
": failed to run array %s: %s\n",
154 devname
, strerror(errno
));
158 fprintf(stderr
, Name
": started %s\n", devname
);
159 } else if (runstop
< 0){
160 struct map_ent
*map
= NULL
;
163 /* If this is an mdmon managed array, just write 'inactive'
164 * to the array state and let mdmon clear up.
166 mdi
= sysfs_read(fd
, -1, GET_LEVEL
|GET_VERSION
);
168 mdi
->array
.level
> 0 &&
169 is_subarray(mdi
->text_version
)) {
172 /* This is mdmon managed. */
174 if (sysfs_set_str(mdi
, NULL
,
175 "array_state", "inactive") < 0) {
178 ": failed to stop array %s: %s\n",
179 devname
, strerror(errno
));
183 /* Give monitor a chance to act */
184 cp
= strchr(mdi
->text_version
+1, '/');
187 ping_monitor(mdi
->text_version
+1);
189 fd
= open(devname
, O_RDONLY
);
191 mdi
->array
.major_version
== -1 &&
192 mdi
->array
.minor_version
== -2 &&
193 !is_subarray(mdi
->text_version
)) {
194 /* container, possibly mdmon-managed.
195 * Make sure mdmon isn't opening it, which
196 * would interfere with the 'stop'
198 ping_monitor(mdi
->sys_name
);
203 if (fd
>= 0 && ioctl(fd
, STOP_ARRAY
, NULL
)) {
206 ": failed to stop array %s: %s\n",
207 devname
, strerror(errno
));
209 fprintf(stderr
, "Perhaps a running "
210 "process, mounted filesystem "
211 "or active volume group?\n");
217 fprintf(stderr
, Name
": stopped %s\n", devname
);
218 if (fd
>= 0 && fstat(fd
, &stb
) == 0) {
220 if (major(stb
.st_rdev
) == MD_MAJOR
)
221 devnum
= minor(stb
.st_rdev
);
223 devnum
= -1-(minor(stb
.st_rdev
)>>6);
224 map_delete(&map
, devnum
);
232 int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
)
234 mdu_array_info_t info
;
235 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
236 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
237 devname
, strerror(errno
));
243 info
.raid_disks
= raid_disks
;
244 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
245 fprintf(stderr
, Name
": Cannot set device size/shape for %s: %s\n",
246 devname
, strerror(errno
));
252 int Manage_reconfig(char *devname
, int fd
, int layout
)
254 mdu_array_info_t info
;
255 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
256 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
257 devname
, strerror(errno
));
260 info
.layout
= layout
;
261 printf("layout set to %d\n", info
.layout
);
262 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
263 fprintf(stderr
, Name
": Cannot set layout for %s: %s\n",
264 devname
, strerror(errno
));
270 int Manage_subdevs(char *devname
, int fd
,
271 mddev_dev_t devlist
, int verbose
)
273 /* do something to each dev.
275 * 'a' - add the device
277 * If that fails EINVAL, try ADD_NEW_DISK
278 * 'r' - remove the device HOT_REMOVE_DISK
279 * device can be 'faulty' or 'detached' in which case all
280 * matching devices are removed.
281 * 'f' - set the device faulty SET_DISK_FAULTY
282 * device can be 'detached' in which case any device that
283 * is inaccessible will be marked faulty.
285 mdu_array_info_t array
;
286 mdu_disk_info_t disc
;
287 unsigned long long array_size
;
288 mddev_dev_t dv
, next
= NULL
;
292 struct supertype
*st
, *tst
;
297 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
298 fprintf(stderr
, Name
": cannot get array info for %s\n",
303 /* array.size is only 32 bit and may be truncated.
304 * So read from sysfs if possible, and record number of sectors
307 array_size
= get_component_size(fd
);
309 array_size
= array
.size
* 2;
311 tst
= super_by_fd(fd
);
313 fprintf(stderr
, Name
": unsupport array - version %d.%d\n",
314 array
.major_version
, array
.minor_version
);
318 for (dv
= devlist
, j
=0 ; dv
; dv
= next
, j
= jnext
) {
319 unsigned long long ldsize
;
321 char *dnprintable
= dv
->devname
;
327 if (strcmp(dv
->devname
, "failed")==0 ||
328 strcmp(dv
->devname
, "faulty")==0) {
329 if (dv
->disposition
!= 'r') {
330 fprintf(stderr
, Name
": %s only meaningful "
331 "with -r, not -%c\n",
332 dv
->devname
, dv
->disposition
);
335 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
337 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
339 if (disc
.major
== 0 && disc
.minor
== 0)
341 if ((disc
.state
& 1) == 0) /* faulty */
343 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
346 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
347 dnprintable
= dvname
;
352 } else if (strcmp(dv
->devname
, "detached") == 0) {
353 if (dv
->disposition
!= 'r' && dv
->disposition
!= 'f') {
354 fprintf(stderr
, Name
": %s only meaningful "
355 "with -r of -f, not -%c\n",
356 dv
->devname
, dv
->disposition
);
359 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
362 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
364 if (disc
.major
== 0 && disc
.minor
== 0)
366 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
367 sfd
= dev_open(dvname
, O_RDONLY
);
372 if (dv
->disposition
== 'f' &&
373 (disc
.state
& 1) == 1) /* already faulty */
377 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
380 dnprintable
= dvname
;
388 if (stat(dv
->devname
, &stb
)) {
389 fprintf(stderr
, Name
": cannot find %s: %s\n",
390 dv
->devname
, strerror(errno
));
393 if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
394 fprintf(stderr
, Name
": %s is not a "
400 switch(dv
->disposition
){
402 fprintf(stderr
, Name
": internal error - devmode[%s]=%d\n",
403 dv
->devname
, dv
->disposition
);
407 if (tst
->subarray
[0]) {
408 fprintf(stderr
, Name
": Cannot add disks to a"
409 " \'member\' array, perform this"
410 " operation on the parent container\n");
413 /* Make sure it isn't in use (in 2.6 or later) */
414 tfd
= open(dv
->devname
, O_RDONLY
|O_EXCL
|O_DIRECT
);
416 fprintf(stderr
, Name
": Cannot open %s: %s\n",
417 dv
->devname
, strerror(errno
));
420 remove_partitions(tfd
);
424 if (array
.not_persistent
==0)
425 st
->ss
->load_super(st
, tfd
, NULL
);
427 if (!get_dev_size(tfd
, dv
->devname
, &ldsize
)) {
434 if (!tst
->ss
->external
&&
435 array
.major_version
== 0 &&
436 md_get_version(fd
)%100 < 2) {
437 if (ioctl(fd
, HOT_ADD_DISK
,
438 (unsigned long)stb
.st_rdev
)==0) {
440 fprintf(stderr
, Name
": hot added %s\n",
445 fprintf(stderr
, Name
": hot add failed for %s: %s\n",
446 dv
->devname
, strerror(errno
));
450 if (array
.not_persistent
== 0 || tst
->ss
->external
) {
451 /* Make sure device is large enough */
452 if (tst
->ss
->avail_size(tst
, ldsize
/512) <
454 fprintf(stderr
, Name
": %s not large enough to join array\n",
459 /* need to find a sample superblock to copy, and
460 * a spare slot to use.
461 * For 'external' array (well, container based),
462 * We can just load the metadata for the array.
464 if (tst
->ss
->external
) {
465 tst
->ss
->load_super(tst
, fd
, NULL
);
466 } else for (j
= 0; j
< tst
->max_devs
; j
++) {
470 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
472 if (disc
.major
==0 && disc
.minor
==0)
474 if ((disc
.state
& 4)==0) continue; /* sync */
475 /* Looks like a good device to try */
476 dev
= map_dev(disc
.major
, disc
.minor
, 1);
478 dfd
= dev_open(dev
, O_RDONLY
);
479 if (dfd
< 0) continue;
480 if (tst
->ss
->load_super(tst
, dfd
,
488 /* FIXME this is a bad test to be using */
490 fprintf(stderr
, Name
": cannot find valid superblock in this array - HELP\n");
493 /* Possibly this device was recently part of the array
494 * and was temporarily removed, and is now being re-added.
495 * If so, we can simply re-add it.
497 tst
->ss
->uuid_from_super(tst
, duuid
);
499 /* re-add doesn't work for version-1 superblocks
502 if (array
.major_version
== 1 &&
503 get_linux_version() <= 2006018)
506 st
->ss
->uuid_from_super(st
, ouuid
);
507 if (memcmp(duuid
, ouuid
, sizeof(ouuid
))==0) {
508 /* looks close enough for now. Kernel
509 * will worry about whether a bitmap
510 * based reconstruction is possible.
513 st
->ss
->getinfo_super(st
, &mdi
);
514 disc
.major
= major(stb
.st_rdev
);
515 disc
.minor
= minor(stb
.st_rdev
);
516 disc
.number
= mdi
.disk
.number
;
517 disc
.raid_disk
= mdi
.disk
.raid_disk
;
518 disc
.state
= mdi
.disk
.state
;
520 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
521 if (ioctl(fd
, ADD_NEW_DISK
, &disc
) == 0) {
523 fprintf(stderr
, Name
": re-added %s\n", dv
->devname
);
526 /* fall back on normal-add */
530 /* non-persistent. Must ensure that new drive
531 * is at least array.size big.
533 if (ldsize
/512 < array_size
) {
534 fprintf(stderr
, Name
": %s not large enough to join array\n",
539 /* in 2.6.17 and earlier, version-1 superblocks won't
540 * use the number we write, but will choose a free number.
541 * we must choose the same free number, which requires
542 * starting at 'raid_disks' and counting up
544 for (j
= array
.raid_disks
; j
< tst
->max_devs
; j
++) {
546 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
548 if (disc
.major
==0 && disc
.minor
==0)
550 if (disc
.state
& 8) /* removed */
553 disc
.major
= major(stb
.st_rdev
);
554 disc
.minor
= minor(stb
.st_rdev
);
557 if (array
.not_persistent
==0 || tst
->ss
->external
) {
560 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
561 dfd
= open(dv
->devname
, O_RDWR
| O_EXCL
|O_DIRECT
);
562 tst
->ss
->add_to_super(tst
, &disc
, dfd
,
564 /* write_init_super will close 'dfd' */
565 if (tst
->ss
->external
)
566 /* mdmon will write the metadata */
568 else if (tst
->ss
->write_init_super(tst
))
570 } else if (dv
->re_add
) {
571 /* this had better be raid1.
572 * As we are "--re-add"ing we must find a spare slot
575 char *used
= malloc(array
.raid_disks
);
576 memset(used
, 0, array
.raid_disks
);
577 for (j
=0; j
< tst
->max_devs
; j
++) {
578 mdu_disk_info_t disc2
;
580 if (ioctl(fd
, GET_DISK_INFO
, &disc2
))
582 if (disc2
.major
==0 && disc2
.minor
==0)
584 if (disc2
.state
& 8) /* removed */
586 if (disc2
.raid_disk
< 0)
588 if (disc2
.raid_disk
> array
.raid_disks
)
590 used
[disc2
.raid_disk
] = 1;
592 for (j
=0 ; j
<array
.raid_disks
; j
++)
595 disc
.state
|= (1<<MD_DISK_SYNC
);
600 disc
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
601 if (tst
->ss
->external
) {
602 /* add a disk to an external metadata container
603 * only if mdmon is around to see it
605 struct mdinfo new_mdi
;
608 int devnum
= fd2devnum(fd
);
610 container_fd
= open_dev_excl(devnum
);
611 if (container_fd
< 0) {
612 fprintf(stderr
, Name
": add failed for %s:"
613 " could not get exclusive access to container\n",
618 if (!mdmon_running(devnum
)) {
619 fprintf(stderr
, Name
": add failed for %s: mdmon not running\n",
625 sra
= sysfs_read(container_fd
, -1, 0);
627 fprintf(stderr
, Name
": add failed for %s: sysfs_read failed\n",
632 sra
->array
.level
= LEVEL_CONTAINER
;
633 /* Need to set data_offset and component_size */
634 tst
->ss
->getinfo_super(tst
, &new_mdi
);
635 new_mdi
.disk
.major
= disc
.major
;
636 new_mdi
.disk
.minor
= disc
.minor
;
637 if (sysfs_add_disk(sra
, &new_mdi
) != 0) {
638 fprintf(stderr
, Name
": add new device to external metadata"
639 " failed for %s\n", dv
->devname
);
643 ping_monitor(devnum2devname(devnum
));
646 } else if (ioctl(fd
, ADD_NEW_DISK
, &disc
)) {
647 fprintf(stderr
, Name
": add new device failed for %s as %d: %s\n",
648 dv
->devname
, j
, strerror(errno
));
652 fprintf(stderr
, Name
": added %s\n", dv
->devname
);
657 if (tst
->subarray
[0]) {
658 fprintf(stderr
, Name
": Cannot remove disks from a"
659 " \'member\' array, perform this"
660 " operation on the parent container\n");
663 if (tst
->ss
->external
) {
664 /* To remove a device from a container, we must
665 * check that it isn't in use in an array.
666 * This involves looking in the 'holders'
667 * directory - there must be just one entry,
669 * To ensure that it doesn't get used as a
670 * hold spare while we are checking, we
671 * get an O_EXCL open on the container
673 int dnum
= fd2devnum(fd
);
674 lfd
= open_dev_excl(dnum
);
677 ": Cannot get exclusive access "
678 " to container - odd\n");
681 if (!sysfs_unique_holder(dnum
, stb
.st_rdev
)) {
683 ": %s is %s, cannot remove.\n",
685 errno
== EEXIST
? "still in use":
691 /* FIXME check that it is a current member */
692 err
= ioctl(fd
, HOT_REMOVE_DISK
, (unsigned long)stb
.st_rdev
);
693 if (err
&& errno
== ENODEV
) {
694 /* Old kernels rejected this if no personality
696 struct mdinfo
*sra
= sysfs_read(fd
, 0, GET_DEVS
);
697 struct mdinfo
*dv
= NULL
;
700 for ( ; dv
; dv
=dv
->next
)
701 if (dv
->disk
.major
== major(stb
.st_rdev
) &&
702 dv
->disk
.minor
== minor(stb
.st_rdev
))
705 err
= sysfs_set_str(sra
, dv
,
713 fprintf(stderr
, Name
": hot remove failed "
714 "for %s: %s\n", dnprintable
,
722 fprintf(stderr
, Name
": hot removed %s\n",
726 case 'f': /* set faulty */
727 /* FIXME check current member */
728 if (ioctl(fd
, SET_DISK_FAULTY
, (unsigned long) stb
.st_rdev
)) {
729 fprintf(stderr
, Name
": set device faulty failed for %s: %s\n",
730 dnprintable
, strerror(errno
));
734 fprintf(stderr
, Name
": set %s faulty in %s\n",
735 dnprintable
, devname
);
745 /* Open any md device, and issue the RAID_AUTORUN ioctl */
747 int fd
= dev_open("9:0", O_RDONLY
);
749 if (ioctl(fd
, RAID_AUTORUN
, 0) == 0)