]>
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 pr_err("At least 2 nodes are needed for cluster-md\n");
125 memset(&info
, 0, sizeof(info
));
126 if (s
->level
== UnSet
&& st
&& st
->ss
->default_geometry
)
127 st
->ss
->default_geometry(st
, &s
->level
, NULL
, NULL
);
128 if (s
->level
== UnSet
) {
129 pr_err("a RAID level is needed to create an array.\n");
132 if (s
->raiddisks
< 4 && s
->level
== 6) {
133 pr_err("at least 4 raid-devices needed for level 6\n");
136 if (s
->raiddisks
> 256 && s
->level
== 6) {
137 pr_err("no more than 256 raid-devices supported for level 6\n");
140 if (s
->raiddisks
< 2 && s
->level
>= 4) {
141 pr_err("at least 2 raid-devices needed for level 4 or 5\n");
144 if (s
->level
<= 0 && s
->sparedisks
) {
145 pr_err("This level does not support spare devices\n");
149 if (subdevs
== 1 && strcmp(devlist
->devname
, "missing") != 0) {
150 /* If given a single device, it might be a container, and we can
151 * extract a device list from there
153 mdu_array_info_t inf
;
156 memset(&inf
, 0, sizeof(inf
));
157 fd
= open(devlist
->devname
, O_RDONLY
);
159 ioctl(fd
, GET_ARRAY_INFO
, &inf
) == 0 &&
160 inf
.raid_disks
== 0) {
161 /* yep, looks like a container */
163 rv
= st
->ss
->load_container(st
, fd
,
168 st
= super_by_fd(fd
, NULL
);
169 if (st
&& !(rv
= st
->ss
->
170 load_container(st
, fd
,
176 if (have_container
) {
177 subdevs
= s
->raiddisks
;
178 first_missing
= subdevs
* 2;
179 second_missing
= subdevs
* 2;
180 insert_point
= subdevs
* 2;
186 if (st
&& st
->ss
->external
&& s
->sparedisks
) {
187 pr_err("This metadata type does not support spare disks at create time\n");
190 if (subdevs
> s
->raiddisks
+s
->sparedisks
+s
->journaldisks
) {
191 pr_err("You have listed more devices (%d) than are in the array(%d)!\n", subdevs
, s
->raiddisks
+s
->sparedisks
);
194 if (!have_container
&& subdevs
< s
->raiddisks
+s
->sparedisks
+s
->journaldisks
) {
195 pr_err("You haven't given enough devices (real or missing) to create this array\n");
198 if (s
->bitmap_file
&& s
->level
<= 0) {
199 pr_err("bitmaps not meaningful with level %s\n",
200 map_num(pers
, s
->level
)?:"given");
204 /* now set some defaults */
206 if (s
->layout
== UnSet
) {
207 do_default_layout
= 1;
208 s
->layout
= default_layout(st
, s
->level
, c
->verbose
);
212 /* check layout fits in array*/
213 if ((s
->layout
&255) * ((s
->layout
>>8)&255) > s
->raiddisks
) {
214 pr_err("that layout requires at least %d devices\n",
215 (s
->layout
&255) * ((s
->layout
>>8)&255));
225 if (s
->chunk
== 0 || s
->chunk
== UnSet
) {
227 do_default_chunk
= 1;
228 /* chunk will be set later */
232 /* a chunksize of zero 0s perfectly valid (and preferred) since 2.6.16 */
233 if (get_linux_version() < 2006016 && s
->chunk
== 0) {
236 pr_err("chunk size defaults to 64K\n");
241 case LEVEL_MULTIPATH
:
242 case LEVEL_CONTAINER
:
246 pr_err("chunk size ignored for this level\n");
250 pr_err("unknown level %d\n", s
->level
);
253 if (s
->size
== MAX_SIZE
)
254 /* use '0' to mean 'max' now... */
256 if (s
->size
&& s
->chunk
&& s
->chunk
!= UnSet
)
257 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
258 newsize
= s
->size
* 2;
259 if (st
&& ! st
->ss
->validate_geometry(st
, s
->level
, s
->layout
, s
->raiddisks
,
260 &s
->chunk
, s
->size
*2,
262 &newsize
, c
->verbose
>=0))
265 if (s
->chunk
&& s
->chunk
!= UnSet
) {
266 newsize
&= ~(unsigned long long)(s
->chunk
*2 - 1);
267 if (do_default_chunk
) {
268 /* default chunk was just set */
270 pr_err("chunk size defaults to %dK\n", s
->chunk
);
271 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
272 do_default_chunk
= 0;
277 s
->size
= newsize
/ 2;
279 /* If this is ever reshaped to RAID5, we will
280 * need a chunksize. So round it off a bit
281 * now just to be safe
283 s
->size
&= ~(64ULL-1);
285 if (s
->size
&& c
->verbose
> 0)
286 pr_err("setting size to %lluK\n", s
->size
);
289 /* now look at the subdevs */
290 info
.array
.active_disks
= 0;
291 info
.array
.working_disks
= 0;
293 for (dv
= devlist
; dv
; dv
= dv
->next
)
294 if (data_offset
== VARIABLE_OFFSET
)
295 dv
->data_offset
= INVALID_SECTORS
;
297 dv
->data_offset
= data_offset
;
299 for (dv
=devlist
; dv
&& !have_container
; dv
=dv
->next
, dnum
++) {
300 char *dname
= dv
->devname
;
301 unsigned long long freesize
;
305 if (strcasecmp(dname
, "missing")==0) {
306 if (first_missing
> dnum
)
307 first_missing
= dnum
;
308 if (second_missing
> dnum
&& dnum
> first_missing
)
309 second_missing
= dnum
;
313 if (data_offset
== VARIABLE_OFFSET
) {
314 doff
= strchr(dname
, ':');
317 dv
->data_offset
= parse_size(doff
);
319 dv
->data_offset
= INVALID_SECTORS
;
321 dv
->data_offset
= data_offset
;
323 dfd
= open(dname
, O_RDONLY
);
325 pr_err("cannot open %s: %s\n",
326 dname
, strerror(errno
));
329 if (fstat(dfd
, &stb
) != 0 ||
330 (stb
.st_mode
& S_IFMT
) != S_IFBLK
) {
332 pr_err("%s is not a block device\n",
337 info
.array
.working_disks
++;
338 if (dnum
< s
->raiddisks
&& dv
->disposition
!= 'j')
339 info
.array
.active_disks
++;
341 struct createinfo
*ci
= conf_get_create_info();
346 /* Need to choose a default metadata, which is different
347 * depending on geometry of array.
350 char *name
= "default";
351 for(i
=0; !st
&& superlist
[i
]; i
++) {
352 st
= superlist
[i
]->match_metadata_desc(name
);
355 if (do_default_layout
)
356 s
->layout
= default_layout(st
, s
->level
, c
->verbose
);
357 switch (st
->ss
->validate_geometry(
358 st
, s
->level
, s
->layout
, s
->raiddisks
,
359 &s
->chunk
, s
->size
*2,
360 dv
->data_offset
, dname
,
361 &freesize
, c
->verbose
> 0)) {
362 case -1: /* Not valid, message printed, and not
363 * worth checking any further */
366 case 0: /* Geometry not valid */
369 s
->chunk
= do_default_chunk
? UnSet
: s
->chunk
;
371 case 1: /* All happy */
377 int dfd
= open(dname
, O_RDONLY
|O_EXCL
);
379 pr_err("cannot open %s: %s\n",
380 dname
, strerror(errno
));
383 pr_err("device %s not suitable for any style of array\n",
387 if (st
->ss
!= &super0
||
388 st
->minor_version
!= 90)
391 if (do_default_layout
)
392 s
->layout
= default_layout(st
, s
->level
, 0);
393 if (!st
->ss
->validate_geometry(st
, s
->level
, s
->layout
,
395 &s
->chunk
, s
->size
*2,
400 pr_err("%s is not suitable for this array.\n",
407 if (dv
->disposition
== 'j')
408 goto skip_size_check
; /* skip write journal for size check */
410 freesize
/= 2; /* convert to K */
411 if (s
->chunk
&& s
->chunk
!= UnSet
) {
412 /* round to chunk size */
413 freesize
= freesize
& ~(s
->chunk
-1);
414 if (do_default_chunk
) {
415 /* default chunk was just set */
417 pr_err("chunk size defaults to %dK\n", s
->chunk
);
418 s
->size
&= ~(unsigned long long)(s
->chunk
- 1);
419 do_default_chunk
= 0;
423 pr_err("no free space left on %s\n", dname
);
428 if (s
->size
&& freesize
< s
->size
) {
429 pr_err("%s is smaller than given size. %lluK < %lluK + metadata\n",
430 dname
, freesize
, s
->size
);
434 if (maxdisc
== NULL
|| (maxdisc
&& freesize
> maxsize
)) {
438 if (mindisc
==NULL
|| (mindisc
&& freesize
< minsize
)) {
443 if (c
->runstop
!= 1 || c
->verbose
>= 0) {
444 int fd
= open(dname
, O_RDONLY
);
446 pr_err("Cannot open %s: %s\n",
447 dname
, strerror(errno
));
451 warn
|= check_ext2(fd
, dname
);
452 warn
|= check_reiser(fd
, dname
);
453 warn
|= check_raid(fd
, dname
);
454 if (strcmp(st
->ss
->name
, "1.x") == 0 &&
455 st
->minor_version
>= 1)
456 /* metadata at front */
457 warn
|= check_partitions(fd
, dname
, 0, 0);
458 else if (s
->level
== 1 || s
->level
== LEVEL_CONTAINER
459 || (s
->level
== 0 && s
->raiddisks
== 1))
460 /* partitions could be meaningful */
461 warn
|= check_partitions(fd
, dname
, freesize
*2, s
->size
*2);
463 /* partitions cannot be meaningful */
464 warn
|= check_partitions(fd
, dname
, 0, 0);
465 if (strcmp(st
->ss
->name
, "1.x") == 0 &&
466 st
->minor_version
>= 1 &&
469 (warn
& 1024) == 0) {
471 pr_err("Note: this array has metadata at the start and\n"
472 " may not be suitable as a boot device. If you plan to\n"
473 " store '/boot' on this device please ensure that\n"
474 " your boot-loader understands md/v1.x metadata, or use\n"
475 " --metadata=0.90\n");
480 if (s
->raiddisks
+ s
->sparedisks
> st
->max_devs
) {
481 pr_err("Too many devices: %s metadata only supports %d\n",
482 st
->ss
->name
, st
->max_devs
);
486 info
.array
.working_disks
= s
->raiddisks
;
488 pr_err("create aborted\n");
492 if (mindisc
== NULL
&& !have_container
) {
493 pr_err("no size and no drives given - aborting create.\n");
496 if (s
->level
> 0 || s
->level
== LEVEL_MULTIPATH
497 || s
->level
== LEVEL_FAULTY
498 || st
->ss
->external
) {
499 /* size is meaningful */
500 if (!st
->ss
->validate_geometry(st
, s
->level
, s
->layout
,
502 &s
->chunk
, minsize
*2,
505 pr_err("devices too large for RAID level %d\n", s
->level
);
510 /* If this is ever reshaped to RAID5, we will
511 * need a chunksize. So round it off a bit
512 * now just to be safe
514 s
->size
&= ~(64ULL-1);
516 pr_err("size set to %lluK\n", s
->size
);
520 if (!s
->bitmap_file
&&
522 st
->ss
->add_internal_bitmap
&&
523 (s
->write_behind
|| s
->size
> 100*1024*1024ULL)) {
525 pr_err("automatically enabling write-intent bitmap on large array\n");
526 s
->bitmap_file
= "internal";
528 if (s
->bitmap_file
&& strcmp(s
->bitmap_file
, "none") == 0)
529 s
->bitmap_file
= NULL
;
531 if (!have_container
&& s
->level
> 0 && ((maxsize
-s
->size
)*100 > maxsize
)) {
532 if (c
->runstop
!= 1 || c
->verbose
>= 0)
533 pr_err("largest drive (%s) exceeds size (%lluK) by more than 1%%\n",
538 if (st
->ss
->detail_platform
&& st
->ss
->detail_platform(0, 1, NULL
) != 0) {
539 if (c
->runstop
!= 1 || c
->verbose
>= 0)
540 pr_err("%s unable to enumerate platform support\n"
541 " array may not be compatible with hardware/firmware\n",
545 st
->nodes
= c
->nodes
;
546 st
->cluster_name
= c
->homecluster
;
549 if (c
->runstop
!= 1) {
550 if (!ask("Continue creating array? ")) {
551 pr_err("create aborted.\n");
556 pr_err("creation continuing despite oddities due to --run\n");
560 /* If this is raid4/5, we want to configure the last active slot
561 * as missing, so that a reconstruct happens (faster than re-parity)
562 * FIX: Can we do this for raid6 as well?
564 if (st
->ss
->external
== 0 &&
565 s
->assume_clean
==0 && c
->force
== 0 && first_missing
>= s
->raiddisks
) {
566 switch ( s
->level
) {
569 insert_point
= s
->raiddisks
-1;
571 info
.array
.active_disks
--;
578 /* For raid6, if creating with 1 missing drive, make a good drive
579 * into a spare, else the create will fail
581 if (s
->assume_clean
== 0 && c
->force
== 0 && first_missing
< s
->raiddisks
&&
582 st
->ss
->external
== 0 &&
583 second_missing
>= s
->raiddisks
&& s
->level
== 6) {
584 insert_point
= s
->raiddisks
- 1;
585 if (insert_point
== first_missing
)
588 info
.array
.active_disks
--;
592 if (s
->level
<= 0 && first_missing
< subdevs
* 2) {
593 pr_err("This level does not support missing devices\n");
597 /* We need to create the device */
599 mdfd
= create_mddev(mddev
, name
, c
->autof
, LOCAL
, chosen_name
);
604 /* verify if chosen_name is not in use,
605 * it could be in conflict with already existing device
606 * e.g. container, array
608 if (strncmp(chosen_name
, "/dev/md/", 8) == 0
609 && map_by_name(&map
, chosen_name
+8) != NULL
) {
610 pr_err("Array name %s is in use already.\n",
618 vers
= md_get_version(mdfd
);
620 pr_err("Create requires md driver version 0.90.0 or later\n");
623 mdu_array_info_t inf
;
624 memset(&inf
, 0, sizeof(inf
));
625 ioctl(mdfd
, GET_ARRAY_INFO
, &inf
);
626 if (inf
.working_disks
!= 0) {
627 pr_err("another array by this name is already running.\n");
632 /* Ok, lets try some ioctls */
634 info
.array
.level
= s
->level
;
635 info
.array
.size
= s
->size
;
636 info
.array
.raid_disks
= s
->raiddisks
;
637 /* The kernel should *know* what md_minor we are dealing
638 * with, but it chooses to trust me instead. Sigh
640 info
.array
.md_minor
= 0;
641 if (fstat(mdfd
, &stb
)==0)
642 info
.array
.md_minor
= minor(stb
.st_rdev
);
643 info
.array
.not_persistent
= 0;
645 if ( ( (s
->level
== 4 || s
->level
== 5) &&
646 (insert_point
< s
->raiddisks
|| first_missing
< s
->raiddisks
) )
648 ( s
->level
== 6 && (insert_point
< s
->raiddisks
649 || second_missing
< s
->raiddisks
))
655 info
.array
.state
= 1; /* clean, but one+ drive will be missing*/
656 info
.resync_start
= MaxSector
;
658 info
.array
.state
= 0; /* not clean, but no errors */
659 info
.resync_start
= 0;
661 if (s
->level
== 10) {
662 /* for raid10, the bitmap size is the capacity of the array,
663 * which is array.size * raid_disks / ncopies;
664 * .. but convert to sectors.
666 int ncopies
= ((s
->layout
>>8) & 255) * (s
->layout
& 255);
667 bitmapsize
= s
->size
* s
->raiddisks
/ ncopies
* 2;
668 /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, s->size, s->raiddisks, ncopies);*/
670 bitmapsize
= s
->size
* 2;
672 /* There is lots of redundancy in these disk counts,
673 * raid_disks is the most meaningful value
674 * it describes the geometry of the array
676 * nr_disks is total number of used slots.
677 * it should be raid_disks+spare_disks
678 * spare_disks is the number of extra disks present
680 * active_disks is the number of working disks in
681 * active slots. (With raid_disks)
682 * working_disks is the total number of working disks,
684 * failed_disks is the number of disks marked failed
686 * Ideally, the kernel would keep these (except raid_disks)
687 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
688 * So for now, we assume that all raid and spare
689 * devices will be given.
691 info
.array
.spare_disks
=s
->sparedisks
;
692 info
.array
.failed_disks
=missing_disks
;
693 info
.array
.nr_disks
= info
.array
.working_disks
694 + info
.array
.failed_disks
;
695 info
.array
.layout
= s
->layout
;
696 info
.array
.chunk_size
= s
->chunk
*1024;
698 if (name
== NULL
|| *name
== 0) {
699 /* base name on mddev */
705 * /dev/md/home -> home
706 * /dev/mdhome -> home
708 /* FIXME compare this with rules in create_mddev */
709 name
= strrchr(mddev
, '/');
712 if (strncmp(name
, "md_", 3)==0 &&
714 (name
-mddev
) == 5 /* /dev/ */)
716 else if (strncmp(name
, "md", 2)==0 &&
719 (name
-mddev
) == 5 /* /dev/ */)
723 if (!st
->ss
->init_super(st
, &info
.array
, s
->size
, name
, c
->homehost
, uuid
,
727 total_slots
= info
.array
.nr_disks
;
728 st
->ss
->getinfo_super(st
, &info
, NULL
);
729 sysfs_init(&info
, mdfd
, NULL
);
731 if (did_default
&& c
->verbose
>= 0) {
732 if (is_subarray(info
.text_version
)) {
737 strncpy(devnm
, info
.text_version
+1, 32);
739 ep
= strchr(devnm
, '/');
743 mdi
= sysfs_read(-1, devnm
, GET_VERSION
);
745 pr_err("Creating array inside %s container %s\n",
746 mdi
?mdi
->text_version
:"managed", devnm
);
749 pr_err("Defaulting to version %s metadata\n", info
.text_version
);
752 map_update(&map
, fd2devnm(mdfd
), info
.text_version
,
753 info
.uuid
, chosen_name
);
754 /* Keep map locked until devices have been added to array
755 * to stop another mdadm from finding and using those devices.
758 if (s
->bitmap_file
&& vers
< 9003) {
759 major_num
= BITMAP_MAJOR_HOSTENDIAN
;
761 pr_err("Warning - bitmaps created on this kernel are not portable\n"
762 " between different architectured. Consider upgrading the Linux kernel.\n");
766 if (s
->bitmap_file
&& (strcmp(s
->bitmap_file
, "internal")==0 ||
767 strcmp(s
->bitmap_file
, "clustered")==0)) {
768 if ((vers
%100) < 2) {
769 pr_err("internal bitmaps not supported by this kernel.\n");
772 if (!st
->ss
->add_internal_bitmap
) {
773 pr_err("internal bitmaps not supported with %s metadata\n",
777 if (st
->ss
->add_internal_bitmap(st
, &s
->bitmap_chunk
,
778 c
->delay
, s
->write_behind
,
779 bitmapsize
, 1, major_num
)) {
780 pr_err("Given bitmap chunk size not supported.\n");
783 s
->bitmap_file
= NULL
;
786 sysfs_init(&info
, mdfd
, NULL
);
788 if (st
->ss
->external
&& st
->container_devnm
[0]) {
791 /* When creating a member, we need to be careful
792 * to negotiate with mdmon properly.
793 * If it is already running, we cannot write to
794 * the devices and must ask it to do that part.
795 * If it isn't running, we write to the devices,
797 * We hold an exclusive open on the container
798 * device to make sure mdmon doesn't exit after
799 * we checked that it is running.
801 * For now, fail if it is already running.
803 container_fd
= open_dev_excl(st
->container_devnm
);
804 if (container_fd
< 0) {
805 pr_err("Cannot get exclusive open on container - weird.\n");
808 if (mdmon_running(st
->container_devnm
)) {
810 pr_err("reusing mdmon for %s.\n",
811 st
->container_devnm
);
812 st
->update_tail
= &st
->updates
;
816 rv
= set_array_info(mdfd
, st
, &info
);
818 pr_err("failed to set array info for %s: %s\n",
819 mddev
, strerror(errno
));
823 if (s
->bitmap_file
) {
826 st
->ss
->uuid_from_super(st
, uuid
);
827 if (CreateBitmap(s
->bitmap_file
, c
->force
, (char*)uuid
, s
->bitmap_chunk
,
828 c
->delay
, s
->write_behind
,
833 bitmap_fd
= open(s
->bitmap_file
, O_RDWR
);
835 pr_err("weird: %s cannot be openned\n",
839 if (ioctl(mdfd
, SET_BITMAP_FILE
, bitmap_fd
) < 0) {
840 pr_err("Cannot set bitmap file for %s: %s\n",
841 mddev
, strerror(errno
));
846 infos
= xmalloc(sizeof(*infos
) * total_slots
);
847 enable_fds(total_slots
);
848 for (pass
=1; pass
<=2 ; pass
++) {
849 struct mddev_dev
*moved_disk
= NULL
; /* the disk that was moved out of the insert point */
851 for (dnum
=0, raid_disk_num
=0, dv
= devlist
; dv
;
852 dv
=(dv
->next
)?(dv
->next
):moved_disk
, dnum
++) {
855 struct mdinfo
*inf
= &infos
[dnum
];
857 if (dnum
>= total_slots
)
859 if (dnum
== insert_point
) {
864 if (strcasecmp(dv
->devname
, "missing")==0) {
870 if (have_container
&& dnum
< info
.array
.raid_disks
- 1)
871 /* repeatedly use the container */
878 inf
->disk
.number
= dnum
;
879 inf
->disk
.raid_disk
= raid_disk_num
++;
881 if (dv
->disposition
== 'j') {
882 inf
->disk
.raid_disk
= MD_DISK_ROLE_JOURNAL
;
883 inf
->disk
.state
= (1<<MD_DISK_JOURNAL
);
885 } else if (inf
->disk
.raid_disk
< s
->raiddisks
)
886 inf
->disk
.state
= (1<<MD_DISK_ACTIVE
) |
891 if (dv
->writemostly
== 1)
892 inf
->disk
.state
|= (1<<MD_DISK_WRITEMOSTLY
);
897 if (st
->ss
->external
&&
898 st
->container_devnm
[0])
899 fd
= open(dv
->devname
, O_RDWR
);
901 fd
= open(dv
->devname
, O_RDWR
|O_EXCL
);
904 pr_err("failed to open %s after earlier success - aborting\n",
909 inf
->disk
.major
= major(stb
.st_rdev
);
910 inf
->disk
.minor
= minor(stb
.st_rdev
);
913 remove_partitions(fd
);
914 if (st
->ss
->add_to_super(st
, &inf
->disk
,
917 ioctl(mdfd
, STOP_ARRAY
, NULL
);
920 st
->ss
->getinfo_super(st
, inf
, NULL
);
921 safe_mode_delay
= inf
->safe_mode_delay
;
923 if (have_container
&& c
->verbose
> 0)
924 pr_err("Using %s for device %d\n",
925 map_dev(inf
->disk
.major
,
929 if (!have_container
) {
930 /* getinfo_super might have lost these ... */
931 inf
->disk
.major
= major(stb
.st_rdev
);
932 inf
->disk
.minor
= minor(stb
.st_rdev
);
938 rv
= add_disk(mdfd
, st
, &info
, inf
);
941 pr_err("ADD_NEW_DISK for %s failed: %s\n",
942 dv
->devname
, strerror(errno
));
947 if (!have_container
&&
948 dv
== moved_disk
&& dnum
!= insert_point
) break;
951 struct mdinfo info_new
;
952 struct map_ent
*me
= NULL
;
954 /* check to see if the uuid has changed due to these
955 * metadata changes, and if so update the member array
956 * and container uuid. Note ->write_init_super clears
957 * the subarray cursor such that ->getinfo_super once
958 * again returns container info.
960 st
->ss
->getinfo_super(st
, &info_new
, NULL
);
961 if (st
->ss
->external
&& s
->level
!= LEVEL_CONTAINER
&&
962 !same_uuid(info_new
.uuid
, info
.uuid
, 0)) {
963 map_update(&map
, fd2devnm(mdfd
),
964 info_new
.text_version
,
965 info_new
.uuid
, chosen_name
);
966 me
= map_by_devnm(&map
, st
->container_devnm
);
969 if (st
->ss
->write_init_super(st
)) {
970 st
->ss
->free_super(st
);
974 /* update parent container uuid */
976 char *path
= xstrdup(me
->path
);
978 st
->ss
->getinfo_super(st
, &info_new
, NULL
);
979 map_update(&map
, st
->container_devnm
,
980 info_new
.text_version
,
981 info_new
.uuid
, path
);
985 flush_metadata_updates(st
);
986 st
->ss
->free_super(st
);
992 if (s
->level
== LEVEL_CONTAINER
) {
993 /* No need to start. But we should signal udev to
995 sysfs_uevent(&info
, "change");
997 pr_err("container %s prepared.\n", mddev
);
998 wait_for(chosen_name
, mdfd
);
999 } else if (c
->runstop
== 1 || subdevs
>= s
->raiddisks
) {
1000 if (st
->ss
->external
) {
1004 case LEVEL_MULTIPATH
:
1006 err
= sysfs_set_str(&info
, NULL
, "array_state",
1013 err
= sysfs_set_str(&info
, NULL
, "array_state",
1017 sysfs_set_safemode(&info
, safe_mode_delay
);
1019 pr_err("failed to activate array.\n");
1020 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1023 } else if (c
->readonly
&&
1024 sysfs_attribute_available(
1025 &info
, NULL
, "array_state")) {
1026 if (sysfs_set_str(&info
, NULL
,
1027 "array_state", "readonly") < 0) {
1028 pr_err("Failed to start array: %s\n",
1030 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1034 /* param is not actually used */
1036 if (ioctl(mdfd
, RUN_ARRAY
, ¶m
)) {
1037 pr_err("RUN_ARRAY failed: %s\n",
1039 if (info
.array
.chunk_size
& (info
.array
.chunk_size
-1)) {
1040 cont_err("Problem may be that chunk size is not a power of 2\n");
1042 ioctl(mdfd
, STOP_ARRAY
, NULL
);
1045 /* if start_ro module parameter is set, array is
1046 * auto-read-only, which is bad as the resync won't
1047 * start. So lets make it read-write now.
1049 ioctl(mdfd
, RESTART_ARRAY_RW
, NULL
);
1051 if (c
->verbose
>= 0)
1052 pr_err("array %s started.\n", mddev
);
1053 if (st
->ss
->external
&& st
->container_devnm
[0]) {
1055 start_mdmon(st
->container_devnm
);
1057 ping_monitor(st
->container_devnm
);
1058 close(container_fd
);
1060 wait_for(chosen_name
, mdfd
);
1062 pr_err("not starting array - not enough devices.\n");
1070 map_remove(&map
, fd2devnm(mdfd
));