]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - Create.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
35 int Create(struct supertype
*st
, char *mddev
,
36 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
37 char *name
, char *homehost
, int *uuid
,
38 int subdevs
, mddev_dev_t devlist
,
39 int runstop
, int verbose
, int force
, int assume_clean
,
40 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
)
43 * Create a new raid array.
45 * First check that necessary details are available
46 * (i.e. level, raid-disks)
48 * Then check each disk to see what might be on it
49 * and report anything interesting.
51 * If anything looks odd, and runstop not set,
54 * SET_ARRAY_INFO and ADD_NEW_DISK, and
55 * if runstop==run, or raiddisks disks were used,
59 unsigned long long minsize
=0, maxsize
=0;
66 int first_missing
= subdevs
* 2;
67 int second_missing
= subdevs
* 2;
68 int missing_disks
= 0;
69 int insert_point
= subdevs
* 2; /* where to insert a missing drive */
75 int have_container
= 0;
76 int container_fd
= -1;
78 unsigned long long bitmapsize
;
79 struct mdinfo info
, *infos
;
81 unsigned long safe_mode_delay
= 0;
82 char chosen_name
[1024];
83 struct map_ent
*map
= NULL
;
85 int major_num
= BITMAP_MAJOR_HI
;
87 memset(&info
, 0, sizeof(info
));
90 /* "ddf" and "imsm" metadata only supports one level - should possibly
91 * push this into metadata handler??
93 if (st
&& (st
->ss
== &super_ddf
|| st
->ss
== &super_imsm
))
94 level
= LEVEL_CONTAINER
;
99 Name
": a RAID level is needed to create an array.\n");
102 if (raiddisks
< 4 && level
== 6) {
104 Name
": at least 4 raid-devices needed for level 6\n");
107 if (raiddisks
> 256 && level
== 6) {
109 Name
": no more than 256 raid-devices supported for level 6\n");
112 if (raiddisks
< 2 && level
>= 4) {
114 Name
": at least 2 raid-devices needed for level 4 or 5\n");
117 if (level
<= 0 && sparedisks
) {
119 Name
": This level does not support spare devices\n");
123 if (subdevs
== 1 && strcmp(devlist
->devname
, "missing") != 0) {
124 /* If given a single device, it might be a container, and we can
125 * extract a device list from there
127 mdu_array_info_t inf
;
130 memset(&inf
, 0, sizeof(inf
));
131 fd
= open(devlist
->devname
, O_RDONLY
);
133 ioctl(fd
, GET_ARRAY_INFO
, &inf
) == 0 &&
134 inf
.raid_disks
== 0) {
135 /* yep, looks like a container */
137 rv
= st
->ss
->load_super(st
, fd
,
142 st
= guess_super(fd
);
143 if (st
&& !(rv
= st
->ss
->
153 if (have_container
) {
158 if (st
&& st
->ss
->external
&& sparedisks
) {
160 Name
": This metadata type does not support "
161 "spare disks are create time\n");
164 if (subdevs
> raiddisks
+sparedisks
) {
165 fprintf(stderr
, Name
": You have listed more devices (%d) than are in the array(%d)!\n", subdevs
, raiddisks
+sparedisks
);
168 if (!have_container
&& subdevs
< raiddisks
+sparedisks
) {
169 fprintf(stderr
, Name
": You haven't given enough devices (real or missing) to create this array\n");
172 if (bitmap_file
&& level
<= 0) {
173 fprintf(stderr
, Name
": bitmaps not meaningful with level %s\n",
174 map_num(pers
, level
)?:"given");
178 /* now set some defaults */
181 default: /* no layout */
185 layout
= 0x102; /* near=2, far=1 */
188 Name
": layout defaults to n1\n");
192 layout
= map_name(r5layout
, "default");
195 Name
": layout defaults to %s\n", map_num(r5layout
, layout
));
198 layout
= map_name(faultylayout
, "default");
202 Name
": layout defaults to %s\n", map_num(faultylayout
, layout
));
207 /* check layout fits in array*/
208 if ((layout
&255) * ((layout
>>8)&255) > raiddisks
) {
209 fprintf(stderr
, Name
": that layout requires at least %d devices\n",
210 (layout
&255) * ((layout
>>8)&255));
220 case LEVEL_LINEAR
: /* linear */
224 fprintf(stderr
, Name
": chunk size defaults to 64K\n");
229 case LEVEL_MULTIPATH
:
230 case LEVEL_CONTAINER
:
234 fprintf(stderr
, Name
": chunk size ignored for this level\n");
238 fprintf(stderr
, Name
": unknown level %d\n", level
);
242 if (st
&& ! st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
243 chunk
, size
, NULL
, NULL
, verbose
>=0))
246 /* now look at the subdevs */
247 info
.array
.active_disks
= 0;
248 info
.array
.working_disks
= 0;
250 for (dv
=devlist
; dv
; dv
=dv
->next
, dnum
++) {
251 char *dname
= dv
->devname
;
252 unsigned long long freesize
;
253 if (strcasecmp(dname
, "missing")==0) {
254 if (first_missing
> dnum
)
255 first_missing
= dnum
;
256 if (second_missing
> dnum
&& dnum
> first_missing
)
257 second_missing
= dnum
;
261 info
.array
.working_disks
++;
262 if (dnum
< raiddisks
)
263 info
.array
.active_disks
++;
265 struct createinfo
*ci
= conf_get_create_info();
270 /* Need to choose a default metadata, which is different
271 * depending on geometry of array.
274 char *name
= "default";
275 for(i
=0; !st
&& superlist
[i
]; i
++) {
276 st
= superlist
[i
]->match_metadata_desc(name
);
277 if (st
&& !st
->ss
->validate_geometry
278 (st
, level
, layout
, raiddisks
,
279 chunk
, size
, dname
, &freesize
,
285 fprintf(stderr
, Name
": device %s not suitable "
286 "for any style of array\n",
290 if (st
->ss
!= &super0
||
291 st
->minor_version
!= 90)
294 if (!st
->ss
->validate_geometry(st
, level
, layout
,
301 Name
": %s is not suitable for "
309 freesize
/= 2; /* convert to K */
311 /* round to chunk size */
312 freesize
= freesize
& ~(chunk
-1);
315 if (size
&& freesize
< size
) {
316 fprintf(stderr
, Name
": %s is smaller than given size."
317 " %lluK < %lluK + metadata\n",
318 dname
, freesize
, size
);
322 if (maxdisc
== NULL
|| (maxdisc
&& freesize
> maxsize
)) {
326 if (mindisc
==NULL
|| (mindisc
&& freesize
< minsize
)) {
330 if (runstop
!= 1 || verbose
>= 0) {
331 int fd
= open(dname
, O_RDONLY
);
333 fprintf(stderr
, Name
": Cannot open %s: %s\n",
334 dname
, strerror(errno
));
338 warn
|= check_ext2(fd
, dname
);
339 warn
|= check_reiser(fd
, dname
);
340 warn
|= check_raid(fd
, dname
);
345 fprintf(stderr
, Name
": create aborted\n");
349 if (mindisc
== NULL
&& !have_container
) {
350 fprintf(stderr
, Name
": no size and no drives given - aborting create.\n");
353 if (level
> 0 || level
== LEVEL_MULTIPATH
354 || level
== LEVEL_FAULTY
355 || st
->ss
->external
) {
356 /* size is meaningful */
357 if (!st
->ss
->validate_geometry(st
, level
, layout
,
361 fprintf(stderr
, Name
": devices too large for RAID level %d\n", level
);
366 fprintf(stderr
, Name
": size set to %lluK\n", size
);
369 if (level
> 0 && ((maxsize
-size
)*100 > maxsize
)) {
370 if (runstop
!= 1 || verbose
>= 0)
371 fprintf(stderr
, Name
": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
378 if (!ask("Continue creating array? ")) {
379 fprintf(stderr
, Name
": create aborted.\n");
384 fprintf(stderr
, Name
": creation continuing despite oddities due to --run\n");
388 /* If this is raid4/5, we want to configure the last active slot
389 * as missing, so that a reconstruct happens (faster than re-parity)
390 * FIX: Can we do this for raid6 as well?
392 if (st
->ss
->external
== 0 &&
393 assume_clean
==0 && force
== 0 && first_missing
>= raiddisks
) {
397 insert_point
= raiddisks
-1;
399 info
.array
.active_disks
--;
406 /* For raid6, if creating with 1 missing drive, make a good drive
407 * into a spare, else the create will fail
409 if (assume_clean
== 0 && force
== 0 && first_missing
< raiddisks
&&
410 st
->ss
->external
== 0 &&
411 second_missing
>= raiddisks
&& level
== 6) {
412 insert_point
= raiddisks
- 1;
413 if (insert_point
== first_missing
)
416 info
.array
.active_disks
--;
420 if (level
<= 0 && first_missing
< subdevs
* 2) {
422 Name
": This level does not support missing devices\n");
426 /* We need to create the device */
428 mdfd
= create_mddev(mddev
, name
, autof
, LOCAL
, chosen_name
);
433 vers
= md_get_version(mdfd
);
435 fprintf(stderr
, Name
": Create requires md driver version 0.90.0 or later\n");
438 mdu_array_info_t inf
;
439 memset(&inf
, 0, sizeof(inf
));
440 ioctl(mdfd
, GET_ARRAY_INFO
, &inf
);
441 if (inf
.working_disks
!= 0) {
442 fprintf(stderr
, Name
": another array by this name"
443 " is already running.\n");
448 /* Ok, lets try some ioctls */
450 info
.array
.level
= level
;
451 info
.array
.size
= size
;
452 info
.array
.raid_disks
= raiddisks
;
453 /* The kernel should *know* what md_minor we are dealing
454 * with, but it chooses to trust me instead. Sigh
456 info
.array
.md_minor
= 0;
457 if (fstat(mdfd
, &stb
)==0)
458 info
.array
.md_minor
= minor(stb
.st_rdev
);
459 info
.array
.not_persistent
= 0;
461 if ( ( (level
== 4 || level
== 5) &&
462 (insert_point
< raiddisks
|| first_missing
< raiddisks
) )
464 ( level
== 6 && (insert_point
< raiddisks
465 || second_missing
< raiddisks
))
471 info
.array
.state
= 1; /* clean, but one+ drive will be missing*/
472 info
.resync_start
= ~0ULL;
474 info
.array
.state
= 0; /* not clean, but no errors */
475 info
.resync_start
= 0;
478 /* for raid10, the bitmap size is the capacity of the array,
479 * which is array.size * raid_disks / ncopies;
480 * .. but convert to sectors.
482 int ncopies
= ((layout
>>8) & 255) * (layout
& 255);
483 bitmapsize
= (unsigned long long)size
* raiddisks
/ ncopies
* 2;
484 /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
486 bitmapsize
= (unsigned long long)size
* 2;
488 /* There is lots of redundancy in these disk counts,
489 * raid_disks is the most meaningful value
490 * it describes the geometry of the array
492 * nr_disks is total number of used slots.
493 * it should be raid_disks+spare_disks
494 * spare_disks is the number of extra disks present
496 * active_disks is the number of working disks in
497 * active slots. (With raid_disks)
498 * working_disks is the total number of working disks,
500 * failed_disks is the number of disks marked failed
502 * Ideally, the kernel would keep these (except raid_disks)
503 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
504 * So for now, we assume that all raid and spare
505 * devices will be given.
507 info
.array
.spare_disks
=sparedisks
;
508 info
.array
.failed_disks
=missing_disks
;
509 info
.array
.nr_disks
= info
.array
.working_disks
510 + info
.array
.failed_disks
;
511 info
.array
.layout
= layout
;
512 info
.array
.chunk_size
= chunk
*1024;
514 if (name
== NULL
|| *name
== 0) {
515 /* base name on mddev */
520 * /dev/md/home -> home
521 * /dev/mdhome -> home
523 /* FIXME compare this with rules in create_mddev */
524 name
= strrchr(mddev
, '/');
527 if (strncmp(name
, "md_d", 4)==0 &&
530 (name
-mddev
) == 5 /* /dev/ */)
532 else if (strncmp(name
, "md", 2)==0 &&
535 (name
-mddev
) == 5 /* /dev/ */)
539 if (!st
->ss
->init_super(st
, &info
.array
, size
, name
, homehost
, uuid
))
542 total_slots
= info
.array
.nr_disks
;
543 sysfs_init(&info
, mdfd
, 0);
544 st
->ss
->getinfo_super(st
, &info
);
546 if (did_default
&& verbose
>= 0) {
547 if (is_subarray(info
.text_version
)) {
548 int dnum
= devname2devnum(info
.text_version
+1);
550 int mdp
= get_mdp_major();
553 path
= map_dev(MD_MAJOR
, dnum
, 1);
555 path
= map_dev(mdp
, (-1-dnum
)<< 6, 1);
557 mdi
= sysfs_read(-1, dnum
, GET_VERSION
);
559 fprintf(stderr
, Name
": Creating array inside "
561 mdi
?mdi
->text_version
:"managed", path
);
564 fprintf(stderr
, Name
": Defaulting to version"
565 " %s metadata\n", info
.text_version
);
568 map_update(&map
, fd2devnum(mdfd
), info
.text_version
,
569 info
.uuid
, chosen_name
);
572 if (bitmap_file
&& vers
< 9003) {
573 major_num
= BITMAP_MAJOR_HOSTENDIAN
;
575 fprintf(stderr
, Name
": Warning - bitmaps created on this kernel are not portable\n"
576 " between different architectured. Consider upgrading the Linux kernel.\n");
580 if (bitmap_file
&& strcmp(bitmap_file
, "internal")==0) {
581 if ((vers
%100) < 2) {
582 fprintf(stderr
, Name
": internal bitmaps not supported by this kernel.\n");
585 if (!st
->ss
->add_internal_bitmap(st
, &bitmap_chunk
,
587 bitmapsize
, 1, major_num
)) {
588 fprintf(stderr
, Name
": Given bitmap chunk size not supported.\n");
595 sysfs_init(&info
, mdfd
, 0);
597 if (st
->ss
->external
&& st
->subarray
[0]) {
600 /* When creating a member, we need to be careful
601 * to negotiate with mdmon properly.
602 * If it is already running, we cannot write to
603 * the devices and must ask it to do that part.
604 * If it isn't running, we write to the devices,
606 * We hold an exclusive open on the container
607 * device to make sure mdmon doesn't exit after
608 * we checked that it is running.
610 * For now, fail if it is already running.
612 container_fd
= open_dev_excl(st
->container_dev
);
613 if (container_fd
< 0) {
614 fprintf(stderr
, Name
": Cannot get exclusive "
615 "open on container - weird.\n");
618 if (mdmon_running(st
->container_dev
)) {
620 fprintf(stderr
, Name
": reusing mdmon "
622 devnum2devname(st
->container_dev
));
623 st
->update_tail
= &st
->updates
;
627 rv
= set_array_info(mdfd
, st
, &info
);
629 fprintf(stderr
, Name
": failed to set array info for %s: %s\n",
630 mddev
, strerror(errno
));
637 st
->ss
->uuid_from_super(st
, uuid
);
638 if (CreateBitmap(bitmap_file
, force
, (char*)uuid
, bitmap_chunk
,
644 bitmap_fd
= open(bitmap_file
, O_RDWR
);
646 fprintf(stderr
, Name
": weird: %s cannot be openned\n",
650 if (ioctl(mdfd
, SET_BITMAP_FILE
, bitmap_fd
) < 0) {
651 fprintf(stderr
, Name
": Cannot set bitmap file for %s: %s\n",
652 mddev
, strerror(errno
));
657 infos
= malloc(sizeof(*infos
) * total_slots
);
659 for (pass
=1; pass
<=2 ; pass
++) {
660 mddev_dev_t moved_disk
= NULL
; /* the disk that was moved out of the insert point */
662 for (dnum
=0, dv
= devlist
; dv
;
663 dv
=(dv
->next
)?(dv
->next
):moved_disk
, dnum
++) {
666 struct mdinfo
*inf
= &infos
[dnum
];
668 if (dnum
>= total_slots
)
670 if (dnum
== insert_point
) {
673 if (dnum
== insert_point
||
674 strcasecmp(dv
->devname
, "missing")==0)
681 inf
->disk
.number
= dnum
;
682 inf
->disk
.raid_disk
= dnum
;
683 if (inf
->disk
.raid_disk
< raiddisks
)
684 inf
->disk
.state
= (1<<MD_DISK_ACTIVE
) |
689 if (dv
->writemostly
== 1)
690 inf
->disk
.state
|= (1<<MD_DISK_WRITEMOSTLY
);
692 if (st
->ss
->external
&& st
->subarray
[0])
693 fd
= open(dv
->devname
, O_RDWR
);
695 fd
= open(dv
->devname
, O_RDWR
|O_EXCL
);
698 fprintf(stderr
, Name
": failed to open %s "
699 "after earlier success - aborting\n",
704 inf
->disk
.major
= major(stb
.st_rdev
);
705 inf
->disk
.minor
= minor(stb
.st_rdev
);
707 remove_partitions(fd
);
708 if (st
->ss
->add_to_super(st
, &inf
->disk
,
711 st
->ss
->getinfo_super(st
, inf
);
712 safe_mode_delay
= inf
->safe_mode_delay
;
714 /* getinfo_super might have lost these ... */
715 inf
->disk
.major
= major(stb
.st_rdev
);
716 inf
->disk
.minor
= minor(stb
.st_rdev
);
722 rv
= add_disk(mdfd
, st
, &info
, inf
);
726 Name
": ADD_NEW_DISK for %s "
728 dv
->devname
, strerror(errno
));
729 st
->ss
->free_super(st
);
734 if (dv
== moved_disk
&& dnum
!= insert_point
) break;
737 st
->ss
->write_init_super(st
);
738 flush_metadata_updates(st
);
742 st
->ss
->free_super(st
);
744 if (level
== LEVEL_CONTAINER
) {
745 /* No need to start. But we should signal udev to
747 sysfs_uevent(&info
, "change");
749 } else if (runstop
== 1 || subdevs
>= raiddisks
) {
750 if (st
->ss
->external
) {
753 case LEVEL_MULTIPATH
:
755 sysfs_set_str(&info
, NULL
, "array_state",
760 sysfs_set_str(&info
, NULL
, "array_state",
764 sysfs_set_safemode(&info
, safe_mode_delay
);
766 /* param is not actually used */
768 if (ioctl(mdfd
, RUN_ARRAY
, ¶m
)) {
769 fprintf(stderr
, Name
": RUN_ARRAY failed: %s\n",
771 Manage_runstop(mddev
, mdfd
, -1, 0);
776 fprintf(stderr
, Name
": array %s started.\n", mddev
);
777 if (st
->ss
->external
&& st
->subarray
[0]) {
779 start_mdmon(st
->container_dev
);
781 ping_monitor(devnum2devname(st
->container_dev
));
784 wait_for(chosen_name
);
786 fprintf(stderr
, Name
": not starting array - not enough devices.\n");