]>
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
)) {
170 /* This is mdmon managed. */
172 if (sysfs_set_str(mdi
, NULL
,
173 "array_state", "inactive") < 0) {
176 ": failed to stop array %s: %s\n",
177 devname
, strerror(errno
));
181 /* Give monitor a chance to act */
182 ping_monitor(mdi
->text_version
);
184 fd
= open(devname
, O_RDONLY
);
186 mdi
->array
.major_version
== -1 &&
187 mdi
->array
.minor_version
== -2 &&
188 !is_subarray(mdi
->text_version
)) {
189 /* container, possibly mdmon-managed.
190 * Make sure mdmon isn't opening it, which
191 * would interfere with the 'stop'
193 ping_monitor(mdi
->sys_name
);
198 if (fd
>= 0 && ioctl(fd
, STOP_ARRAY
, NULL
)) {
201 ": failed to stop array %s: %s\n",
202 devname
, strerror(errno
));
204 fprintf(stderr
, "Perhaps a running "
205 "process, mounted filesystem "
206 "or active volume group?\n");
212 fprintf(stderr
, Name
": stopped %s\n", devname
);
213 if (fd
>= 0 && fstat(fd
, &stb
) == 0) {
215 if (major(stb
.st_rdev
) == MD_MAJOR
)
216 devnum
= minor(stb
.st_rdev
);
218 devnum
= -1-(minor(stb
.st_rdev
)>>6);
219 map_delete(&map
, devnum
);
227 int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
)
229 mdu_array_info_t info
;
230 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
231 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
232 devname
, strerror(errno
));
238 info
.raid_disks
= raid_disks
;
239 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
240 fprintf(stderr
, Name
": Cannot set device size/shape for %s: %s\n",
241 devname
, strerror(errno
));
247 int Manage_reconfig(char *devname
, int fd
, int layout
)
249 mdu_array_info_t info
;
250 if (ioctl(fd
, GET_ARRAY_INFO
, &info
) != 0) {
251 fprintf(stderr
, Name
": Cannot get array information for %s: %s\n",
252 devname
, strerror(errno
));
255 info
.layout
= layout
;
256 printf("layout set to %d\n", info
.layout
);
257 if (ioctl(fd
, SET_ARRAY_INFO
, &info
) != 0) {
258 fprintf(stderr
, Name
": Cannot set layout for %s: %s\n",
259 devname
, strerror(errno
));
265 int Manage_subdevs(char *devname
, int fd
,
266 mddev_dev_t devlist
, int verbose
)
268 /* do something to each dev.
270 * 'a' - add the device
272 * If that fails EINVAL, try ADD_NEW_DISK
273 * 'r' - remove the device HOT_REMOVE_DISK
274 * device can be 'faulty' or 'detached' in which case all
275 * matching devices are removed.
276 * 'f' - set the device faulty SET_DISK_FAULTY
277 * device can be 'detached' in which case any device that
278 * is inaccessible will be marked faulty.
280 mdu_array_info_t array
;
281 mdu_disk_info_t disc
;
282 unsigned long long array_size
;
283 mddev_dev_t dv
, next
= NULL
;
287 struct supertype
*st
, *tst
;
292 if (ioctl(fd
, GET_ARRAY_INFO
, &array
)) {
293 fprintf(stderr
, Name
": cannot get array info for %s\n",
298 /* array.size is only 32 bit and may be truncated.
299 * So read from sysfs if possible, and record number of sectors
302 array_size
= get_component_size(fd
);
304 array_size
= array
.size
* 2;
306 tst
= super_by_fd(fd
);
308 fprintf(stderr
, Name
": unsupport array - version %d.%d\n",
309 array
.major_version
, array
.minor_version
);
313 for (dv
= devlist
, j
=0 ; dv
; dv
= next
, j
= jnext
) {
314 unsigned long long ldsize
;
316 char *dnprintable
= dv
->devname
;
322 if (strcmp(dv
->devname
, "failed")==0 ||
323 strcmp(dv
->devname
, "faulty")==0) {
324 if (dv
->disposition
!= 'r') {
325 fprintf(stderr
, Name
": %s only meaningful "
326 "with -r, not -%c\n",
327 dv
->devname
, dv
->disposition
);
330 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
332 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
334 if (disc
.major
== 0 && disc
.minor
== 0)
336 if ((disc
.state
& 1) == 0) /* faulty */
338 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
341 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
342 dnprintable
= dvname
;
347 } else if (strcmp(dv
->devname
, "detached") == 0) {
348 if (dv
->disposition
!= 'r' && dv
->disposition
!= 'f') {
349 fprintf(stderr
, Name
": %s only meaningful "
350 "with -r of -f, not -%c\n",
351 dv
->devname
, dv
->disposition
);
354 for (; j
< array
.raid_disks
+ array
.nr_disks
; j
++) {
357 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
359 if (disc
.major
== 0 && disc
.minor
== 0)
361 sprintf(dvname
,"%d:%d", disc
.major
, disc
.minor
);
362 sfd
= dev_open(dvname
, O_RDONLY
);
367 if (dv
->disposition
== 'f' &&
368 (disc
.state
& 1) == 1) /* already faulty */
372 stb
.st_rdev
= makedev(disc
.major
, disc
.minor
);
375 dnprintable
= dvname
;
383 if (stat(dv
->devname
, &stb
)) {
384 fprintf(stderr
, Name
": cannot find %s: %s\n",
385 dv
->devname
, strerror(errno
));
388 if ((stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
389 fprintf(stderr
, Name
": %s is not a "
395 switch(dv
->disposition
){
397 fprintf(stderr
, Name
": internal error - devmode[%s]=%d\n",
398 dv
->devname
, dv
->disposition
);
402 if (tst
->subarray
[0]) {
403 fprintf(stderr
, Name
": Cannot add disks to a"
404 " \'member\' array, perform this"
405 " operation on the parent container\n");
408 /* Make sure it isn't in use (in 2.6 or later) */
409 tfd
= open(dv
->devname
, O_RDONLY
|O_EXCL
|O_DIRECT
);
411 fprintf(stderr
, Name
": Cannot open %s: %s\n",
412 dv
->devname
, strerror(errno
));
415 remove_partitions(tfd
);
419 if (array
.not_persistent
==0)
420 st
->ss
->load_super(st
, tfd
, NULL
);
422 if (!get_dev_size(tfd
, dv
->devname
, &ldsize
)) {
429 if (!tst
->ss
->external
&&
430 array
.major_version
== 0 &&
431 md_get_version(fd
)%100 < 2) {
432 if (ioctl(fd
, HOT_ADD_DISK
,
433 (unsigned long)stb
.st_rdev
)==0) {
435 fprintf(stderr
, Name
": hot added %s\n",
440 fprintf(stderr
, Name
": hot add failed for %s: %s\n",
441 dv
->devname
, strerror(errno
));
445 if (array
.not_persistent
== 0 || tst
->ss
->external
) {
446 /* Make sure device is large enough */
447 if (tst
->ss
->avail_size(tst
, ldsize
/512) <
449 fprintf(stderr
, Name
": %s not large enough to join array\n",
454 /* need to find a sample superblock to copy, and
455 * a spare slot to use.
456 * For 'external' array (well, container based),
457 * We can just load the metadata for the array.
459 if (tst
->ss
->external
) {
460 tst
->ss
->load_super(tst
, fd
, NULL
);
461 } else for (j
= 0; j
< tst
->max_devs
; j
++) {
465 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
467 if (disc
.major
==0 && disc
.minor
==0)
469 if ((disc
.state
& 4)==0) continue; /* sync */
470 /* Looks like a good device to try */
471 dev
= map_dev(disc
.major
, disc
.minor
, 1);
473 dfd
= dev_open(dev
, O_RDONLY
);
474 if (dfd
< 0) continue;
475 if (tst
->ss
->load_super(tst
, dfd
,
483 /* FIXME this is a bad test to be using */
485 fprintf(stderr
, Name
": cannot find valid superblock in this array - HELP\n");
488 /* Possibly this device was recently part of the array
489 * and was temporarily removed, and is now being re-added.
490 * If so, we can simply re-add it.
492 tst
->ss
->uuid_from_super(tst
, duuid
);
494 /* re-add doesn't work for version-1 superblocks
497 if (array
.major_version
== 1 &&
498 get_linux_version() <= 2006018)
501 st
->ss
->uuid_from_super(st
, ouuid
);
502 if (memcmp(duuid
, ouuid
, sizeof(ouuid
))==0) {
503 /* looks close enough for now. Kernel
504 * will worry about whether a bitmap
505 * based reconstruction is possible.
508 st
->ss
->getinfo_super(st
, &mdi
);
509 disc
.major
= major(stb
.st_rdev
);
510 disc
.minor
= minor(stb
.st_rdev
);
511 disc
.number
= mdi
.disk
.number
;
512 disc
.raid_disk
= mdi
.disk
.raid_disk
;
513 disc
.state
= mdi
.disk
.state
;
515 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
516 if (ioctl(fd
, ADD_NEW_DISK
, &disc
) == 0) {
518 fprintf(stderr
, Name
": re-added %s\n", dv
->devname
);
521 /* fall back on normal-add */
525 /* non-persistent. Must ensure that new drive
526 * is at least array.size big.
528 if (ldsize
/512 < array_size
) {
529 fprintf(stderr
, Name
": %s not large enough to join array\n",
534 /* in 2.6.17 and earlier, version-1 superblocks won't
535 * use the number we write, but will choose a free number.
536 * we must choose the same free number, which requires
537 * starting at 'raid_disks' and counting up
539 for (j
= array
.raid_disks
; j
< tst
->max_devs
; j
++) {
541 if (ioctl(fd
, GET_DISK_INFO
, &disc
))
543 if (disc
.major
==0 && disc
.minor
==0)
545 if (disc
.state
& 8) /* removed */
548 disc
.major
= major(stb
.st_rdev
);
549 disc
.minor
= minor(stb
.st_rdev
);
552 if (array
.not_persistent
==0 || tst
->ss
->external
) {
555 disc
.state
|= 1 << MD_DISK_WRITEMOSTLY
;
556 dfd
= open(dv
->devname
, O_RDWR
| O_EXCL
|O_DIRECT
);
557 tst
->ss
->add_to_super(tst
, &disc
, dfd
,
559 /* write_init_super will close 'dfd' */
560 if (tst
->ss
->external
)
561 /* mdmon will write the metadata */
563 else if (tst
->ss
->write_init_super(tst
))
565 } else if (dv
->re_add
) {
566 /* this had better be raid1.
567 * As we are "--re-add"ing we must find a spare slot
570 char *used
= malloc(array
.raid_disks
);
571 memset(used
, 0, array
.raid_disks
);
572 for (j
=0; j
< tst
->max_devs
; j
++) {
573 mdu_disk_info_t disc2
;
575 if (ioctl(fd
, GET_DISK_INFO
, &disc2
))
577 if (disc2
.major
==0 && disc2
.minor
==0)
579 if (disc2
.state
& 8) /* removed */
581 if (disc2
.raid_disk
< 0)
583 if (disc2
.raid_disk
> array
.raid_disks
)
585 used
[disc2
.raid_disk
] = 1;
587 for (j
=0 ; j
<array
.raid_disks
; j
++)
590 disc
.state
|= (1<<MD_DISK_SYNC
);
595 disc
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
596 if (tst
->ss
->external
) {
597 /* add a disk to an external metadata container
598 * only if mdmon is around to see it
600 struct mdinfo new_mdi
;
603 int devnum
= fd2devnum(fd
);
605 container_fd
= open_dev_excl(devnum
);
606 if (container_fd
< 0) {
607 fprintf(stderr
, Name
": add failed for %s:"
608 " could not get exclusive access to container\n",
613 if (!mdmon_running(devnum
)) {
614 fprintf(stderr
, Name
": add failed for %s: mdmon not running\n",
620 sra
= sysfs_read(container_fd
, -1, 0);
622 fprintf(stderr
, Name
": add failed for %s: sysfs_read failed\n",
627 sra
->array
.level
= LEVEL_CONTAINER
;
628 /* Need to set data_offset and component_size */
629 tst
->ss
->getinfo_super(tst
, &new_mdi
);
630 new_mdi
.disk
.major
= disc
.major
;
631 new_mdi
.disk
.minor
= disc
.minor
;
632 if (sysfs_add_disk(sra
, &new_mdi
) != 0) {
633 fprintf(stderr
, Name
": add new device to external metadata"
634 " failed for %s\n", dv
->devname
);
638 ping_monitor(devnum2devname(devnum
));
641 } else if (ioctl(fd
, ADD_NEW_DISK
, &disc
)) {
642 fprintf(stderr
, Name
": add new device failed for %s as %d: %s\n",
643 dv
->devname
, j
, strerror(errno
));
647 fprintf(stderr
, Name
": added %s\n", dv
->devname
);
652 if (tst
->subarray
[0]) {
653 fprintf(stderr
, Name
": Cannot remove disks from a"
654 " \'member\' array, perform this"
655 " operation on the parent container\n");
658 if (tst
->ss
->external
) {
659 /* To remove a device from a container, we must
660 * check that it isn't in use in an array.
661 * This involves looking in the 'holders'
662 * directory - there must be just one entry,
664 * To ensure that it doesn't get used as a
665 * hold spare while we are checking, we
666 * get an O_EXCL open on the container
668 int dnum
= fd2devnum(fd
);
669 lfd
= open_dev_excl(dnum
);
672 ": Cannot get exclusive access "
673 " to container - odd\n");
676 if (!sysfs_unique_holder(dnum
, stb
.st_rdev
)) {
678 ": %s is %s, cannot remove.\n",
680 errno
== EEXIST
? "still in use":
686 /* FIXME check that it is a current member */
687 err
= ioctl(fd
, HOT_REMOVE_DISK
, (unsigned long)stb
.st_rdev
);
688 if (err
&& errno
== ENODEV
) {
689 /* Old kernels rejected this if no personality
691 struct mdinfo
*sra
= sysfs_read(fd
, 0, GET_DEVS
);
692 struct mdinfo
*dv
= NULL
;
695 for ( ; dv
; dv
=dv
->next
)
696 if (dv
->disk
.major
== major(stb
.st_rdev
) &&
697 dv
->disk
.minor
== minor(stb
.st_rdev
))
700 err
= sysfs_set_str(sra
, dv
,
708 fprintf(stderr
, Name
": hot remove failed "
709 "for %s: %s\n", dnprintable
,
717 fprintf(stderr
, Name
": hot removed %s\n",
721 case 'f': /* set faulty */
722 /* FIXME check current member */
723 if (ioctl(fd
, SET_DISK_FAULTY
, (unsigned long) stb
.st_rdev
)) {
724 fprintf(stderr
, Name
": set device faulty failed for %s: %s\n",
725 dnprintable
, strerror(errno
));
729 fprintf(stderr
, Name
": set %s faulty in %s\n",
730 dnprintable
, devname
);
740 /* Open any md device, and issue the RAID_AUTORUN ioctl */
742 int fd
= dev_open("9:0", O_RDONLY
);
744 if (ioctl(fd
, RAID_AUTORUN
, 0) == 0)