]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Create.c
2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2013 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 static int default_layout(struct supertype
*st
, int level
, int verbose
)
34 if (st
&& st
->ss
->default_geometry
)
35 st
->ss
->default_geometry(st
, &level
, &layout
, NULL
);
39 default: /* no layout */
43 layout
= 0x102; /* near=2, far=1 */
45 pr_err("layout defaults to n2\n");
49 layout
= map_name(r5layout
, "default");
51 pr_err("layout defaults to %s\n", map_num(r5layout
, layout
));
54 layout
= map_name(faultylayout
, "default");
57 pr_err("layout defaults to %s\n", map_num(faultylayout
, layout
));
64 int Create(struct supertype
*st
, char *mddev
,
65 char *name
, int *uuid
,
66 int subdevs
, struct mddev_dev
*devlist
,
68 struct context
*c
, unsigned long long data_offset
)
71 * Create a new raid array.
73 * First check that necessary details are available
74 * (i.e. level, raid-disks)
76 * Then check each disk to see what might be on it
77 * and report anything interesting.
79 * If anything looks odd, and runstop not set,
82 * SET_ARRAY_INFO and ADD_NEW_DISK, and
83 * if runstop==run, or raiddisks disks were used,
87 unsigned long long minsize
=0, maxsize
=0;
90 int dnum
, raid_disk_num
;
94 int first_missing
= subdevs
* 2;
95 int second_missing
= subdevs
* 2;
96 int missing_disks
= 0;
97 int insert_point
= subdevs
* 2; /* where to insert a missing drive */
103 int have_container
= 0;
104 int container_fd
= -1;
106 unsigned long long bitmapsize
;
107 struct mdinfo info
, *infos
;
109 int do_default_layout
= 0;
110 int do_default_chunk
= 0;
111 unsigned long safe_mode_delay
= 0;
112 char chosen_name
[1024];
113 struct map_ent
*map
= NULL
;
114 unsigned long long newsize
;
116 int major_num
= BITMAP_MAJOR_HI
;
117 if (s
->bitmap_file
&& strcmp(s
->bitmap_file
, "clustered") == 0)
118 major_num
= BITMAP_MAJOR_CLUSTERED
;
120 memset(&info
, 0, sizeof(info
));
121 if (s
->level
== UnSet
&& st
&& st
->ss
->default_geometry
)
122 st
->ss
->default_geometry(st
, &s
->level
, NULL
, NULL
);
123 if (s
->level
== UnSet
) {
124 pr_err("a RAID level is needed to create an array.\n");
127 if (s
->raiddisks
< 4 && s
->level
== 6) {
128 pr_err("at least 4 raid-devices needed for level 6\n");
131 if (s
->raiddisks
> 256 && s
->level
== 6) {
132 pr_err("no more than 256 raid-devices supported for level 6\n");
135 if (s
->raiddisks
< 2 && s
->level
>= 4) {
136 pr_err("at least 2 raid-devices needed for level 4 or 5\n");
139 if (s
->level
<= 0 && s
->sparedisks
) {
140 pr_err("This level does not support spare devices\n");
144 if (subdevs
== 1 && strcmp(devlist
->devname
, "missing") != 0) {
145 /* If given a single device, it might be a container, and we can
146 * extract a device list from there
148 mdu_array_info_t inf
;
151 memset(&inf
, 0, sizeof(inf
));
152 fd
= open(devlist
->devname
, O_RDONLY
);
154 ioctl(fd
, GET_ARRAY_INFO
, &inf
) == 0 &&
155 inf
.raid_disks
== 0) {
156 /* yep, looks like a container */
158 rv
= st
->ss
->load_container(st
, fd
,
163 st
= super_by_fd(fd
, NULL
);
164 if (st
&& !(rv
= st
->ss
->
165 load_container(st
, fd
,
171 if (have_container
) {
172 subdevs
= s
->raiddisks
;
173 first_missing
= subdevs
* 2;
174 second_missing
= subdevs
* 2;
175 insert_point
= subdevs
* 2;
181 if (st
&& st
->ss
->external
&& s
->sparedisks
) {
182 pr_err("This metadata type does not support spare disks at create time\n");
185 if (subdevs
> s
->raiddisks
+s
->sparedisks
+s
->journaldisks
) {
186 pr_err("You have listed more devices (%d) than are in the array(%d)!\n", subdevs
, s
->raiddisks
+s
->sparedisks
);
189 if (!have_container
&& subdevs
< s
->raiddisks
+s
->sparedisks
+s
->journaldisks
) {
190 pr_err("You haven't given enough devices (real or missing) to create this array\n");
193 if (s
->bitmap_file
&& s
->level
<= 0) {
194 pr_err("bitmaps not meaningful with level %s\n",
195 map_num(pers
, s
->level
)?:"given");
199 /* now set some defaults */
201 if (s
->layout
== UnSet
) {
202 do_default_layout
= 1;
203 s
->layout
= default_layout(st
, s
->level
, c
->verbose
);
207 /* check layout fits in array*/
208 if ((s
->layout
&255) * ((s
->layout
>>8)&255) > s
->raiddisks
) {
209 pr_err("that layout requires at least %d devices\n",
210 (s
->layout
&255) * ((s
->layout
>>8)&255));
220 if (s
->chunk
== 0 || s
->chunk
== UnSet
) {
222 do_default_chunk
= 1;
223 /* chunk will be set later */
227 /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */
228 if (get_linux_version() < 2006016 && s
->chunk
== 0) {
231 pr_err("chunk size defaults to 64K\n");
236 case LEVEL_MULTIPATH
:
237 case LEVEL_CONTAINER
:
241 pr_err("chunk size ignored for this level\n");
245 pr_err("unknown level %d\n", s
->level
);
248 if (s
->size
== MAX_SIZE
)
249 /* use '0' to mean 'max' now... */
251 if (s
->size
&& s
->chunk
&& s
->chunk
!= UnSet
)
252 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
253 newsize
= s
->size
* 2;
254 if (st
&& ! st
->ss
->validate_geometry(st
, s
->level
, s
->layout
, s
->raiddisks
,
255 &s
->chunk
, s
->size
*2,
257 &newsize
, c
->verbose
>=0))
260 if (s
->chunk
&& s
->chunk
!= UnSet
) {
261 newsize
&= ~(unsigned long long)(s
->chunk
*2 - 1);
262 if (do_default_chunk
) {
263 /* default chunk was just set */
265 pr_err("chunk size defaults to %dK\n", s
->chunk
);
266 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
267 do_default_chunk
= 0;
272 s
->size
= newsize
/ 2;
274 /* If this is ever reshaped to RAID5, we will
275 * need a chunksize. So round it off a bit
276 * now just to be safe
278 s
->size
&= ~(64ULL-1);
280 if (s
->size
&& c
->verbose
> 0)
281 pr_err("setting size to %lluK\n", s
->size
);
284 /* now look at the subdevs */
285 info
.array
.active_disks
= 0;
286 info
.array
.working_disks
= 0;
288 for (dv
= devlist
; dv
; dv
= dv
->next
)
289 if (data_offset
== VARIABLE_OFFSET
)
290 dv
->data_offset
= INVALID_SECTORS
;
292 dv
->data_offset
= data_offset
;
294 for (dv
=devlist
; dv
&& !have_container
; dv
=dv
->next
, dnum
++) {
295 char *dname
= dv
->devname
;
296 unsigned long long freesize
;
300 if (strcasecmp(dname
, "missing")==0) {
301 if (first_missing
> dnum
)
302 first_missing
= dnum
;
303 if (second_missing
> dnum
&& dnum
> first_missing
)
304 second_missing
= dnum
;
308 if (data_offset
== VARIABLE_OFFSET
) {
309 doff
= strchr(dname
, ':');
312 dv
->data_offset
= parse_size(doff
);
314 dv
->data_offset
= INVALID_SECTORS
;
316 dv
->data_offset
= data_offset
;
318 dfd
= open(dname
, O_RDONLY
);
320 pr_err("cannot open %s: %s\n",
321 dname
, strerror(errno
));
324 if (fstat(dfd
, &stb
) != 0 ||
325 (stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
327 pr_err("%s is not a block device\n",
332 info
.array
.working_disks
++;
333 if (dnum
< s
->raiddisks
&& dv
->disposition
!= 'j')
334 info
.array
.active_disks
++;
336 struct createinfo
*ci
= conf_get_create_info();
341 /* Need to choose a default metadata, which is different
342 * depending on geometry of array.
345 char *name
= "default";
346 for(i
=0; !st
&& superlist
[i
]; i
++) {
347 st
= superlist
[i
]->match_metadata_desc(name
);
350 if (do_default_layout
)
351 s
->layout
= default_layout(st
, s
->level
, c
->verbose
);
352 switch (st
->ss
->validate_geometry(
353 st
, s
->level
, s
->layout
, s
->raiddisks
,
354 &s
->chunk
, s
->size
*2,
355 dv
->data_offset
, dname
,
356 &freesize
, c
->verbose
> 0)) {
357 case -1: /* Not valid, message printed, and not
358 * worth checking any further */
361 case 0: /* Geometry not valid */
364 s
->chunk
= do_default_chunk
? UnSet
: s
->chunk
;
366 case 1: /* All happy */
372 int dfd
= open(dname
, O_RDONLY
|O_EXCL
);
374 pr_err("cannot open %s: %s\n",
375 dname
, strerror(errno
));
378 pr_err("device %s not suitable for any style of array\n",
382 if (st
->ss
!= &super0
||
383 st
->minor_version
!= 90)
386 if (do_default_layout
)
387 s
->layout
= default_layout(st
, s
->level
, 0);
388 if (!st
->ss
->validate_geometry(st
, s
->level
, s
->layout
,
390 &s
->chunk
, s
->size
*2,
395 pr_err("%s is not suitable for this array.\n",
402 if (dv
->disposition
== 'j')
403 goto skip_size_check
; /* skip write journal for size check */
405 freesize
/= 2; /* convert to K */
406 if (s
->chunk
&& s
->chunk
!= UnSet
) {
407 /* round to chunk size */
408 freesize
= freesize
& ~(s
->chunk
-1);
409 if (do_default_chunk
) {
410 /* default chunk was just set */
412 pr_err("chunk size defaults to %dK\n", s
->chunk
);
413 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
414 do_default_chunk
= 0;
418 pr_err("no free space left on %s\n", dname
);
423 if (s
->size
&& freesize
< s
->size
) {
424 pr_err("%s is smaller than given size. %lluK < %lluK + metadata\n",
425 dname
, freesize
, s
->size
);
429 if (maxdisc
== NULL
|| (maxdisc
&& freesize
> maxsize
)) {
433 if (mindisc
==NULL
|| (mindisc
&& freesize
< minsize
)) {
438 if (c
->runstop
!= 1 || c
->verbose
>= 0) {
439 int fd
= open(dname
, O_RDONLY
);
441 pr_err("Cannot open %s: %s\n",
442 dname
, strerror(errno
));
446 warn
|= check_ext2(fd
, dname
);
447 warn
|= check_reiser(fd
, dname
);
448 warn
|= check_raid(fd
, dname
);
449 if (strcmp(st
->ss
->name
, "1.x") == 0 &&
450 st
->minor_version
>= 1)
451 /* metadata at front */
452 warn
|= check_partitions(fd
, dname
, 0, 0);
453 else if (s
->level
== 1 || s
->level
== LEVEL_CONTAINER
454 || (s
->level
== 0 && s
->raiddisks
== 1))
455 /* partitions could be meaningful */
456 warn
|= check_partitions(fd
, dname
, freesize
*2, s
->size
*2);
458 /* partitions cannot be meaningful */
459 warn
|= check_partitions(fd
, dname
, 0, 0);
460 if (strcmp(st
->ss
->name
, "1.x") == 0 &&
461 st
->minor_version
>= 1 &&
464 (warn
& 1024) == 0) {
466 pr_err("Note: this array has metadata at the start and\n"
467 " may not be suitable as a boot device. If you plan to\n"
468 " store '/boot' on this device please ensure that\n"
469 " your boot-loader understands md/v1.x metadata, or use\n"
470 " --metadata=0.90\n");
475 if (s
->raiddisks
+ s
->sparedisks
> st
->max_devs
) {
476 pr_err("Too many devices: %s metadata only supports %d\n",
477 st
->ss
->name
, st
->max_devs
);
481 info
.array
.working_disks
= s
->raiddisks
;
483 pr_err("create aborted\n");
487 if (mindisc
== NULL
&& !have_container
) {
488 pr_err("no size and no drives given - aborting create.\n");
491 if (s
->level
> 0 || s
->level
== LEVEL_MULTIPATH
492 || s
->level
== LEVEL_FAULTY
493 || st
->ss
->external
) {
494 /* size is meaningful */
495 if (!st
->ss
->validate_geometry(st
, s
->level
, s
->layout
,
497 &s
->chunk
, minsize
*2,
500 pr_err("devices too large for RAID level %d\n", s
->level
);
505 /* If this is ever reshaped to RAID5, we will
506 * need a chunksize. So round it off a bit
507 * now just to be safe
509 s
->size
&= ~(64ULL-1);
511 pr_err("size set to %lluK\n", s
->size
);
515 if (!s
->bitmap_file
&&
517 st
->ss
->add_internal_bitmap
&&
518 (s
->write_behind
|| s
->size
> 100*1024*1024ULL)) {
520 pr_err("automatically enabling write-intent bitmap on large array\n");
521 s
->bitmap_file
= "internal";
523 if (s
->bitmap_file
&& strcmp(s
->bitmap_file
, "none") == 0)
524 s
->bitmap_file
= NULL
;
526 if (!have_container
&& s
->level
> 0 && ((maxsize
-s
->size
)*100 > maxsize
)) {
527 if (c
->runstop
!= 1 || c
->verbose
>= 0)
528 pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n",
533 if (st
->ss
->detail_platform
&& st
->ss
->detail_platform(0, 1, NULL
) != 0) {
534 if (c
->runstop
!= 1 || c
->verbose
>= 0)
535 pr_err("%s unable to enumerate platform support\n"
536 " array may not be compatible with hardware/firmware\n",
540 st
->nodes
= c
->nodes
;
541 st
->cluster_name
= c
->homecluster
;
544 if (c
->runstop
!= 1) {
545 if (!ask("Continue creating array? ")) {
546 pr_err("create aborted.\n");
551 pr_err("creation continuing despite oddities due to --run\n");
555 /* If this is raid4/5, we want to configure the last active slot
556 * as missing, so that a reconstruct happens (faster than re-parity)
557 * FIX: Can we do this for raid6 as well?
559 if (st
->ss
->external
== 0 &&
560 s
->assume_clean
==0 && c
->force
== 0 && first_missing
>= s
->raiddisks
) {
561 switch ( s
->level
) {
564 insert_point
= s
->raiddisks
-1;
566 info
.array
.active_disks
--;
573 /* For raid6, if creating with 1 missing drive, make a good drive
574 * into a spare, else the create will fail
576 if (s
->assume_clean
== 0 && c
->force
== 0 && first_missing
< s
->raiddisks
&&
577 st
->ss
->external
== 0 &&
578 second_missing
>= s
->raiddisks
&& s
->level
== 6) {
579 insert_point
= s
->raiddisks
- 1;
580 if (insert_point
== first_missing
)
583 info
.array
.active_disks
--;
587 if (s
->level
<= 0 && first_missing
< subdevs
* 2) {
588 pr_err("This level does not support missing devices\n");
592 /* We need to create the device */
594 mdfd
= create_mddev(mddev
, name
, c
->autof
, LOCAL
, chosen_name
);
599 /* verify if chosen_name is not in use,
600 * it could be in conflict with already existing device
601 * e.g. container, array
603 if (strncmp(chosen_name
, "/dev/md/", 8) == 0
604 && map_by_name(&map
, chosen_name
+8) != NULL
) {
605 pr_err("Array name %s is in use already.\n",
613 vers
= md_get_version(mdfd
);
615 pr_err("Create requires md driver version 0.90.0 or later\n");
618 mdu_array_info_t inf
;
619 memset(&inf
, 0, sizeof(inf
));
620 ioctl(mdfd
, GET_ARRAY_INFO
, &inf
);
621 if (inf
.working_disks
!= 0) {
622 pr_err("another array by this name is already running.\n");
627 /* Ok, lets try some ioctls */
629 info
.array
.level
= s
->level
;
630 info
.array
.size
= s
->size
;
631 info
.array
.raid_disks
= s
->raiddisks
;
632 /* The kernel should *know* what md_minor we are dealing
633 * with, but it chooses to trust me instead. Sigh
635 info
.array
.md_minor
= 0;
636 if (fstat(mdfd
, &stb
)==0)
637 info
.array
.md_minor
= minor(stb
.st_rdev
);
638 info
.array
.not_persistent
= 0;
640 if ( ( (s
->level
== 4 || s
->level
== 5) &&
641 (insert_point
< s
->raiddisks
|| first_missing
< s
->raiddisks
) )
643 ( s
->level
== 6 && (insert_point
< s
->raiddisks
644 || second_missing
< s
->raiddisks
))
650 info
.array
.state
= 1; /* clean, but one+ drive will be missing*/
651 info
.resync_start
= MaxSector
;
653 info
.array
.state
= 0; /* not clean, but no errors */
654 info
.resync_start
= 0;
656 if (s
->level
== 10) {
657 /* for raid10, the bitmap size is the capacity of the array,
658 * which is array.size * raid_disks / ncopies;
659 * .. but convert to sectors.
661 int ncopies
= ((s
->layout
>>8) & 255) * (s
->layout
& 255);
662 bitmapsize
= s
->size
* s
->raiddisks
/ ncopies
* 2;
663 /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/
665 bitmapsize
= s
->size
* 2;
667 /* There is lots of redundancy in these disk counts,
668 * raid_disks is the most meaningful value
669 * it describes the geometry of the array
671 * nr_disks is total number of used slots.
672 * it should be raid_disks+spare_disks
673 * spare_disks is the number of extra disks present
675 * active_disks is the number of working disks in
676 * active slots. (With raid_disks)
677 * working_disks is the total number of working disks,
679 * failed_disks is the number of disks marked failed
681 * Ideally, the kernel would keep these (except raid_disks)
682 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
683 * So for now, we assume that all raid and spare
684 * devices will be given.
686 info
.array
.spare_disks
=s
->sparedisks
;
687 info
.array
.failed_disks
=missing_disks
;
688 info
.array
.nr_disks
= info
.array
.working_disks
689 + info
.array
.failed_disks
;
690 info
.array
.layout
= s
->layout
;
691 info
.array
.chunk_size
= s
->chunk
*1024;
693 if (name
== NULL
|| *name
== 0) {
694 /* base name on mddev */
700 * /dev/md/home -> home
701 * /dev/mdhome -> home
703 /* FIXME compare this with rules in create_mddev */
704 name
= strrchr(mddev
, '/');
707 if (strncmp(name
, "md_", 3)==0 &&
709 (name
-mddev
) == 5 /* /dev/ */)
711 else if (strncmp(name
, "md", 2)==0 &&
714 (name
-mddev
) == 5 /* /dev/ */)
718 if (!st
->ss
->init_super(st
, &info
.array
, s
->size
, name
, c
->homehost
, uuid
,
722 total_slots
= info
.array
.nr_disks
;
723 st
->ss
->getinfo_super(st
, &info
, NULL
);
724 sysfs_init(&info
, mdfd
, NULL
);
726 if (did_default
&& c
->verbose
>= 0) {
727 if (is_subarray(info
.text_version
)) {
732 strncpy(devnm
, info
.text_version
+1, 32);
734 ep
= strchr(devnm
, '/');
738 mdi
= sysfs_read(-1, devnm
, GET_VERSION
);
740 pr_err("Creating array inside %s container %s\n",
741 mdi
?mdi
->text_version
:"managed", devnm
);
744 pr_err("Defaulting to version %s metadata\n", info
.text_version
);
747 map_update(&map
, fd2devnm(mdfd
), info
.text_version
,
748 info
.uuid
, chosen_name
);
749 /* Keep map locked until devices have been added to array
750 * to stop another mdadm from finding and using those devices.
753 if (s
->bitmap_file
&& vers
< 9003) {
754 major_num
= BITMAP_MAJOR_HOSTENDIAN
;
756 pr_err("Warning - bitmaps created on this kernel are not portable\n"
757 " between different architectured. Consider upgrading the Linux kernel.\n");
761 if (s
->bitmap_file
&& (strcmp(s
->bitmap_file
, "internal")==0 ||
762 strcmp(s
->bitmap_file
, "clustered")==0)) {
763 if ((vers
%100) < 2) {
764 pr_err("internal bitmaps not supported by this kernel.\n");
767 if (!st
->ss
->add_internal_bitmap
) {
768 pr_err("internal bitmaps not supported with %s metadata\n",
772 if (!st
->ss
->add_internal_bitmap(st
, &s
->bitmap_chunk
,
773 c
->delay
, s
->write_behind
,
774 bitmapsize
, 1, major_num
)) {
775 pr_err("Given bitmap chunk size not supported.\n");
778 s
->bitmap_file
= NULL
;
781 sysfs_init(&info
, mdfd
, NULL
);
783 if (st
->ss
->external
&& st
->container_devnm
[0]) {
786 /* When creating a member, we need to be careful
787 * to negotiate with mdmon properly.
788 * If it is already running, we cannot write to
789 * the devices and must ask it to do that part.
790 * If it isn't running, we write to the devices,
792 * We hold an exclusive open on the container
793 * device to make sure mdmon doesn't exit after
794 * we checked that it is running.
796 * For now, fail if it is already running.
798 container_fd
= open_dev_excl(st
->container_devnm
);
799 if (container_fd
< 0) {
800 pr_err("Cannot get exclusive open on container - weird.\n");
803 if (mdmon_running(st
->container_devnm
)) {
805 pr_err("reusing mdmon for %s.\n",
806 st
->container_devnm
);
807 st
->update_tail
= &st
->updates
;
811 rv
= set_array_info(mdfd
, st
, &info
);
813 pr_err("failed to set array info for %s: %s\n",
814 mddev
, strerror(errno
));
818 if (s
->bitmap_file
) {
821 st
->ss
->uuid_from_super(st
, uuid
);
822 if (CreateBitmap(s
->bitmap_file
, c
->force
, (char*)uuid
, s
->bitmap_chunk
,
823 c
->delay
, s
->write_behind
,
828 bitmap_fd
= open(s
->bitmap_file
, O_RDWR
);
830 pr_err("weird: %s cannot be openned\n",
834 if (ioctl(mdfd
, SET_BITMAP_FILE
, bitmap_fd
) < 0) {
835 pr_err("Cannot set bitmap file for %s: %s\n",
836 mddev
, strerror(errno
));
841 infos
= xmalloc(sizeof(*infos
) * total_slots
);
842 enable_fds(total_slots
);
843 for (pass
=1; pass
<=2 ; pass
++) {
844 struct mddev_dev
*moved_disk
= NULL
; /* the disk that was moved out of the insert point */
846 for (dnum
=0, raid_disk_num
=0, dv
= devlist
; dv
;
847 dv
=(dv
->next
)?(dv
->next
):moved_disk
, dnum
++) {
850 struct mdinfo
*inf
= &infos
[dnum
];
852 if (dnum
>= total_slots
)
854 if (dnum
== insert_point
) {
859 if (strcasecmp(dv
->devname
, "missing")==0) {
865 if (have_container
&& dnum
< info
.array
.raid_disks
- 1)
866 /* repeatedly use the container */
873 inf
->disk
.number
= dnum
;
874 inf
->disk
.raid_disk
= raid_disk_num
++;
876 if (dv
->disposition
== 'j') {
877 inf
->disk
.raid_disk
= MD_DISK_ROLE_JOURNAL
;
878 inf
->disk
.state
= (1<<MD_DISK_JOURNAL
);
880 } else if (inf
->disk
.raid_disk
< s
->raiddisks
)
881 inf
->disk
.state
= (1<<MD_DISK_ACTIVE
) |
886 if (dv
->writemostly
== 1)
887 inf
->disk
.state
|= (1<<MD_DISK_WRITEMOSTLY
);
892 if (st
->ss
->external
&&
893 st
->container_devnm
[0])
894 fd
= open(dv
->devname
, O_RDWR
);
896 fd
= open(dv
->devname
, O_RDWR
|O_EXCL
);
899 pr_err("failed to open %s after earlier success - aborting\n",
904 inf
->disk
.major
= major(stb
.st_rdev
);
905 inf
->disk
.minor
= minor(stb
.st_rdev
);
908 remove_partitions(fd
);
909 if (st
->ss
->add_to_super(st
, &inf
->disk
,
912 ioctl(mdfd
, STOP_ARRAY
, NULL
);
915 st
->ss
->getinfo_super(st
, inf
, NULL
);
916 safe_mode_delay
= inf
->safe_mode_delay
;
918 if (have_container
&& c
->verbose
> 0)
919 pr_err("Using %s for device %d\n",
920 map_dev(inf
->disk
.major
,
924 if (!have_container
) {
925 /* getinfo_super might have lost these ... */
926 inf
->disk
.major
= major(stb
.st_rdev
);
927 inf
->disk
.minor
= minor(stb
.st_rdev
);
933 rv
= add_disk(mdfd
, st
, &info
, inf
);
936 pr_err("ADD_NEW_DISK for %s failed: %s\n",
937 dv
->devname
, strerror(errno
));
942 if (!have_container
&&
943 dv
== moved_disk
&& dnum
!= insert_point
) break;
946 struct mdinfo info_new
;
947 struct map_ent
*me
= NULL
;
949 /* check to see if the uuid has changed due to these
950 * metadata changes, and if so update the member array
951 * and container uuid. Note ->write_init_super clears
952 * the subarray cursor such that ->getinfo_super once
953 * again returns container info.
955 st
->ss
->getinfo_super(st
, &info_new
, NULL
);
956 if (st
->ss
->external
&& s
->level
!= LEVEL_CONTAINER
&&
957 !same_uuid(info_new
.uuid
, info
.uuid
, 0)) {
958 map_update(&map
, fd2devnm(mdfd
),
959 info_new
.text_version
,
960 info_new
.uuid
, chosen_name
);
961 me
= map_by_devnm(&map
, st
->container_devnm
);
964 if (st
->ss
->write_init_super(st
)) {
965 st
->ss
->free_super(st
);
969 /* update parent container uuid */
971 char *path
= xstrdup(me
->path
);
973 st
->ss
->getinfo_super(st
, &info_new
, NULL
);
974 map_update(&map
, st
->container_devnm
,
975 info_new
.text_version
,
976 info_new
.uuid
, path
);
980 flush_metadata_updates(st
);
981 st
->ss
->free_super(st
);
987 if (s
->level
== LEVEL_CONTAINER
) {
988 /* No need to start. But we should signal udev to
990 sysfs_uevent(&info
, "change");
992 pr_err("container %s prepared.\n", mddev
);
993 wait_for(chosen_name
, mdfd
);
994 } else if (c
->runstop
== 1 || subdevs
>= s
->raiddisks
) {
995 if (st
->ss
->external
) {
999 case LEVEL_MULTIPATH
:
1001 err
= sysfs_set_str(&info
, NULL
, "array_state",
1008 err
= sysfs_set_str(&info
, NULL
, "array_state",
1012 sysfs_set_safemode(&info
, safe_mode_delay
);
1014 pr_err("failed to activate array.\n");
1015 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1018 } else if (c
->readonly
&&
1019 sysfs_attribute_available(
1020 &info
, NULL
, "array_state")) {
1021 if (sysfs_set_str(&info
, NULL
,
1022 "array_state", "readonly") < 0) {
1023 pr_err("Failed to start array: %s\n",
1025 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1029 /* param is not actually used */
1031 if (ioctl(mdfd
, RUN_ARRAY
, ¶m
)) {
1032 pr_err("RUN_ARRAY failed: %s\n",
1034 if (info
.array
.chunk_size
& (info
.array
.chunk_size
-1)) {
1035 cont_err("Problem may be that chunk size is not a power of 2\n");
1037 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1040 /* if start_ro module parameter is set, array is
1041 * auto-read-only, which is bad as the resync won't
1042 * start. So lets make it read-write now.
1044 ioctl(mdfd
, RESTART_ARRAY_RW
, NULL
);
1046 if (c
->verbose
>= 0)
1047 pr_err("array %s started.\n", mddev
);
1048 if (st
->ss
->external
&& st
->container_devnm
[0]) {
1050 start_mdmon(st
->container_devnm
);
1052 ping_monitor(st
->container_devnm
);
1053 close(container_fd
);
1055 wait_for(chosen_name
, mdfd
);
1057 pr_err("not starting array - not enough devices.\n");
1065 map_remove(&map
, fd2devnm(mdfd
));