]>
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];
84 int major_num
= BITMAP_MAJOR_HI
;
86 memset(&info
, 0, sizeof(info
));
89 /* "ddf" and "imsm" metadata only supports one level - should possibly
90 * push this into metadata handler??
92 if (st
&& (st
->ss
== &super_ddf
|| st
->ss
== &super_imsm
))
93 level
= LEVEL_CONTAINER
;
98 Name
": a RAID level is needed to create an array.\n");
101 if (raiddisks
< 4 && level
== 6) {
103 Name
": at least 4 raid-devices needed for level 6\n");
106 if (raiddisks
> 256 && level
== 6) {
108 Name
": no more than 256 raid-devices supported for level 6\n");
111 if (raiddisks
< 2 && level
>= 4) {
113 Name
": at least 2 raid-devices needed for level 4 or 5\n");
116 if (level
<= 0 && sparedisks
) {
118 Name
": This level does not support spare devices\n");
122 if (subdevs
== 1 && strcmp(devlist
->devname
, "missing") != 0) {
123 /* If given a single device, it might be a container, and we can
124 * extract a device list from there
126 mdu_array_info_t inf
;
129 memset(&inf
, 0, sizeof(inf
));
130 fd
= open(devlist
->devname
, O_RDONLY
);
132 ioctl(fd
, GET_ARRAY_INFO
, &inf
) == 0 &&
133 inf
.raid_disks
== 0) {
134 /* yep, looks like a container */
136 rv
= st
->ss
->load_super(st
, fd
,
141 st
= guess_super(fd
);
142 if (st
&& !(rv
= st
->ss
->
152 if (have_container
) {
157 if (st
&& st
->ss
->external
&& sparedisks
) {
159 Name
": This metadata type does not support "
160 "spare disks are create time\n");
163 if (subdevs
> raiddisks
+sparedisks
) {
164 fprintf(stderr
, Name
": You have listed more devices (%d) than are in the array(%d)!\n", subdevs
, raiddisks
+sparedisks
);
167 if (!have_container
&& subdevs
< raiddisks
+sparedisks
) {
168 fprintf(stderr
, Name
": You haven't given enough devices (real or missing) to create this array\n");
171 if (bitmap_file
&& level
<= 0) {
172 fprintf(stderr
, Name
": bitmaps not meaningful with level %s\n",
173 map_num(pers
, level
)?:"given");
177 /* now set some defaults */
180 default: /* no layout */
184 layout
= 0x102; /* near=2, far=1 */
187 Name
": layout defaults to n1\n");
191 layout
= map_name(r5layout
, "default");
194 Name
": layout defaults to %s\n", map_num(r5layout
, layout
));
197 layout
= map_name(faultylayout
, "default");
201 Name
": layout defaults to %s\n", map_num(faultylayout
, layout
));
206 /* check layout fits in array*/
207 if ((layout
&255) * ((layout
>>8)&255) > raiddisks
) {
208 fprintf(stderr
, Name
": that layout requires at least %d devices\n",
209 (layout
&255) * ((layout
>>8)&255));
219 case LEVEL_LINEAR
: /* linear */
223 fprintf(stderr
, Name
": chunk size defaults to 64K\n");
228 case LEVEL_MULTIPATH
:
229 case LEVEL_CONTAINER
:
233 fprintf(stderr
, Name
": chunk size ignored for this level\n");
237 fprintf(stderr
, Name
": unknown level %d\n", level
);
241 if (st
&& ! st
->ss
->validate_geometry(st
, level
, layout
, raiddisks
,
242 chunk
, size
, NULL
, NULL
, verbose
>=0))
245 /* now look at the subdevs */
246 info
.array
.active_disks
= 0;
247 info
.array
.working_disks
= 0;
249 for (dv
=devlist
; dv
; dv
=dv
->next
, dnum
++) {
250 char *dname
= dv
->devname
;
251 unsigned long long freesize
;
252 if (strcasecmp(dname
, "missing")==0) {
253 if (first_missing
> dnum
)
254 first_missing
= dnum
;
255 if (second_missing
> dnum
&& dnum
> first_missing
)
256 second_missing
= dnum
;
260 info
.array
.working_disks
++;
261 if (dnum
< raiddisks
)
262 info
.array
.active_disks
++;
264 struct createinfo
*ci
= conf_get_create_info();
269 /* Need to choose a default metadata, which is different
270 * depending on geometry of array.
273 char *name
= "default";
274 for(i
=0; !st
&& superlist
[i
]; i
++) {
275 st
= superlist
[i
]->match_metadata_desc(name
);
276 if (st
&& !st
->ss
->validate_geometry
277 (st
, level
, layout
, raiddisks
,
278 chunk
, size
, dname
, &freesize
,
284 fprintf(stderr
, Name
": device %s not suitable "
285 "for any style of array\n",
289 if (st
->ss
!= &super0
||
290 st
->minor_version
!= 90)
293 if (!st
->ss
->validate_geometry(st
, level
, layout
,
300 Name
": %s is not suitable for "
308 freesize
/= 2; /* convert to K */
310 /* round to chunk size */
311 freesize
= freesize
& ~(chunk
-1);
314 if (size
&& freesize
< size
) {
315 fprintf(stderr
, Name
": %s is smaller that given size."
316 " %lluK < %lluK + metadata\n",
317 dname
, freesize
, size
);
321 if (maxdisc
== NULL
|| (maxdisc
&& freesize
> maxsize
)) {
325 if (mindisc
==NULL
|| (mindisc
&& freesize
< minsize
)) {
329 if (runstop
!= 1 || verbose
>= 0) {
330 int fd
= open(dname
, O_RDONLY
);
332 fprintf(stderr
, Name
": Cannot open %s: %s\n",
333 dname
, strerror(errno
));
337 warn
|= check_ext2(fd
, dname
);
338 warn
|= check_reiser(fd
, dname
);
339 warn
|= check_raid(fd
, dname
);
344 fprintf(stderr
, Name
": create aborted\n");
348 if (mindisc
== NULL
&& !have_container
) {
349 fprintf(stderr
, Name
": no size and no drives given - aborting create.\n");
352 if (level
> 0 || level
== LEVEL_MULTIPATH
353 || level
== LEVEL_FAULTY
354 || st
->ss
->external
) {
355 /* size is meaningful */
356 if (!st
->ss
->validate_geometry(st
, level
, layout
,
360 fprintf(stderr
, Name
": devices too large for RAID level %d\n", level
);
365 fprintf(stderr
, Name
": size set to %lluK\n", size
);
368 if (level
> 0 && ((maxsize
-size
)*100 > maxsize
)) {
369 if (runstop
!= 1 || verbose
>= 0)
370 fprintf(stderr
, Name
": largest drive (%s) exceed size (%lluK) by more than 1%%\n",
377 if (!ask("Continue creating array? ")) {
378 fprintf(stderr
, Name
": create aborted.\n");
383 fprintf(stderr
, Name
": creation continuing despite oddities due to --run\n");
387 /* If this is raid4/5, we want to configure the last active slot
388 * as missing, so that a reconstruct happens (faster than re-parity)
389 * FIX: Can we do this for raid6 as well?
391 if (assume_clean
==0 && force
== 0 && first_missing
>= raiddisks
) {
395 insert_point
= raiddisks
-1;
397 info
.array
.active_disks
--;
404 /* For raid6, if creating with 1 missing drive, make a good drive
405 * into a spare, else the create will fail
407 if (assume_clean
== 0 && force
== 0 && first_missing
< raiddisks
&&
408 st
->ss
->external
== 0 &&
409 second_missing
>= raiddisks
&& level
== 6) {
410 insert_point
= raiddisks
- 1;
411 if (insert_point
== first_missing
)
414 info
.array
.active_disks
--;
418 if (level
<= 0 && first_missing
< subdevs
* 2) {
420 Name
": This level does not support missing devices\n");
424 /* We need to create the device */
425 mdfd
= create_mddev(mddev
, name
, autof
, LOCAL
, chosen_name
);
429 vers
= md_get_version(mdfd
);
431 fprintf(stderr
, Name
": Create requires md driver version 0.90.0 or later\n");
434 mdu_array_info_t inf
;
435 memset(&inf
, 0, sizeof(inf
));
436 ioctl(mdfd
, GET_ARRAY_INFO
, &inf
);
437 if (inf
.working_disks
!= 0) {
438 fprintf(stderr
, Name
": another array by this name"
439 " is already running.\n");
444 /* Ok, lets try some ioctls */
446 info
.array
.level
= level
;
447 info
.array
.size
= size
;
448 info
.array
.raid_disks
= raiddisks
;
449 /* The kernel should *know* what md_minor we are dealing
450 * with, but it chooses to trust me instead. Sigh
452 info
.array
.md_minor
= 0;
453 if (fstat(mdfd
, &stb
)==0)
454 info
.array
.md_minor
= minor(stb
.st_rdev
);
455 info
.array
.not_persistent
= 0;
457 if ( ( (level
== 4 || level
== 5) &&
458 (insert_point
< raiddisks
|| first_missing
< raiddisks
) )
460 ( level
== 6 && (insert_point
< raiddisks
461 || second_missing
< raiddisks
))
467 info
.array
.state
= 1; /* clean, but one+ drive will be missing*/
468 info
.resync_start
= ~0ULL;
470 info
.array
.state
= 0; /* not clean, but no errors */
471 info
.resync_start
= 0;
474 /* for raid10, the bitmap size is the capacity of the array,
475 * which is array.size * raid_disks / ncopies;
476 * .. but convert to sectors.
478 int ncopies
= ((layout
>>8) & 255) * (layout
& 255);
479 bitmapsize
= (unsigned long long)size
* raiddisks
/ ncopies
* 2;
480 /* printf("bms=%llu as=%d rd=%d nc=%d\n", bitmapsize, size, raiddisks, ncopies);*/
482 bitmapsize
= (unsigned long long)size
* 2;
484 /* There is lots of redundancy in these disk counts,
485 * raid_disks is the most meaningful value
486 * it describes the geometry of the array
488 * nr_disks is total number of used slots.
489 * it should be raid_disks+spare_disks
490 * spare_disks is the number of extra disks present
492 * active_disks is the number of working disks in
493 * active slots. (With raid_disks)
494 * working_disks is the total number of working disks,
496 * failed_disks is the number of disks marked failed
498 * Ideally, the kernel would keep these (except raid_disks)
499 * up-to-date as we ADD_NEW_DISK, but it doesn't (yet).
500 * So for now, we assume that all raid and spare
501 * devices will be given.
503 info
.array
.spare_disks
=sparedisks
;
504 info
.array
.failed_disks
=missing_disks
;
505 info
.array
.nr_disks
= info
.array
.working_disks
506 + info
.array
.failed_disks
;
507 info
.array
.layout
= layout
;
508 info
.array
.chunk_size
= chunk
*1024;
510 if (name
== NULL
|| *name
== 0) {
511 /* base name on mddev */
516 * /dev/md/home -> home
517 * /dev/mdhome -> home
519 /* FIXME compare this with rules in create_mddev */
520 name
= strrchr(mddev
, '/');
523 if (strncmp(name
, "md_d", 4)==0 &&
526 (name
-mddev
) == 5 /* /dev/ */)
528 else if (strncmp(name
, "md", 2)==0 &&
531 (name
-mddev
) == 5 /* /dev/ */)
535 if (!st
->ss
->init_super(st
, &info
.array
, size
, name
, homehost
, uuid
))
538 total_slots
= info
.array
.nr_disks
;
539 sysfs_init(&info
, mdfd
, 0);
540 st
->ss
->getinfo_super(st
, &info
);
542 if (did_default
&& verbose
>= 0) {
543 if (is_subarray(info
.text_version
)) {
544 int dnum
= devname2devnum(info
.text_version
+1);
546 int mdp
= get_mdp_major();
549 path
= map_dev(MD_MAJOR
, dnum
, 1);
551 path
= map_dev(mdp
, (-1-dnum
)<< 6, 1);
553 mdi
= sysfs_read(-1, dnum
, GET_VERSION
);
555 fprintf(stderr
, Name
": Creating array inside "
557 mdi
?mdi
->text_version
:"managed", path
);
560 fprintf(stderr
, Name
": Defaulting to version"
561 " %s metadata\n", info
.text_version
);
564 map_update(NULL
, fd2devnum(mdfd
), info
.text_version
,
565 info
.uuid
, chosen_name
);
567 if (bitmap_file
&& vers
< 9003) {
568 major_num
= BITMAP_MAJOR_HOSTENDIAN
;
570 fprintf(stderr
, Name
": Warning - bitmaps created on this kernel are not portable\n"
571 " between different architectured. Consider upgrading the Linux kernel.\n");
575 if (bitmap_file
&& strcmp(bitmap_file
, "internal")==0) {
576 if ((vers
%100) < 2) {
577 fprintf(stderr
, Name
": internal bitmaps not supported by this kernel.\n");
580 if (!st
->ss
->add_internal_bitmap(st
, &bitmap_chunk
,
582 bitmapsize
, 1, major_num
)) {
583 fprintf(stderr
, Name
": Given bitmap chunk size not supported.\n");
590 sysfs_init(&info
, mdfd
, 0);
592 if (st
->ss
->external
&& st
->subarray
[0]) {
595 /* When creating a member, we need to be careful
596 * to negotiate with mdmon properly.
597 * If it is already running, we cannot write to
598 * the devices and must ask it to do that part.
599 * If it isn't running, we write to the devices,
601 * We hold an exclusive open on the container
602 * device to make sure mdmon doesn't exit after
603 * we checked that it is running.
605 * For now, fail if it is already running.
607 container_fd
= open_dev_excl(st
->container_dev
);
608 if (container_fd
< 0) {
609 fprintf(stderr
, Name
": Cannot get exclusive "
610 "open on container - weird.\n");
613 if (mdmon_running(st
->container_dev
)) {
615 fprintf(stderr
, Name
": reusing mdmon "
617 devnum2devname(st
->container_dev
));
618 st
->update_tail
= &st
->updates
;
622 rv
= set_array_info(mdfd
, st
, &info
);
624 fprintf(stderr
, Name
": failed to set array info for %s: %s\n",
625 mddev
, strerror(errno
));
632 st
->ss
->uuid_from_super(st
, uuid
);
633 if (CreateBitmap(bitmap_file
, force
, (char*)uuid
, bitmap_chunk
,
639 bitmap_fd
= open(bitmap_file
, O_RDWR
);
641 fprintf(stderr
, Name
": weird: %s cannot be openned\n",
645 if (ioctl(mdfd
, SET_BITMAP_FILE
, bitmap_fd
) < 0) {
646 fprintf(stderr
, Name
": Cannot set bitmap file for %s: %s\n",
647 mddev
, strerror(errno
));
652 infos
= malloc(sizeof(*infos
) * total_slots
);
654 for (pass
=1; pass
<=2 ; pass
++) {
655 mddev_dev_t moved_disk
= NULL
; /* the disk that was moved out of the insert point */
657 for (dnum
=0, dv
= devlist
; dv
;
658 dv
=(dv
->next
)?(dv
->next
):moved_disk
, dnum
++) {
661 struct mdinfo
*inf
= &infos
[dnum
];
663 if (dnum
>= total_slots
)
665 if (dnum
== insert_point
) {
668 if (dnum
== insert_point
||
669 strcasecmp(dv
->devname
, "missing")==0)
676 inf
->disk
.number
= dnum
;
677 inf
->disk
.raid_disk
= dnum
;
678 if (inf
->disk
.raid_disk
< raiddisks
)
679 inf
->disk
.state
= (1<<MD_DISK_ACTIVE
) |
684 if (dv
->writemostly
== 1)
685 inf
->disk
.state
|= (1<<MD_DISK_WRITEMOSTLY
);
687 if (st
->ss
->external
&& st
->subarray
[0])
688 fd
= open(dv
->devname
, O_RDWR
);
690 fd
= open(dv
->devname
, O_RDWR
|O_EXCL
);
693 fprintf(stderr
, Name
": failed to open %s "
694 "after earlier success - aborting\n",
699 inf
->disk
.major
= major(stb
.st_rdev
);
700 inf
->disk
.minor
= minor(stb
.st_rdev
);
702 remove_partitions(fd
);
703 st
->ss
->add_to_super(st
, &inf
->disk
,
705 st
->ss
->getinfo_super(st
, inf
);
706 safe_mode_delay
= inf
->safe_mode_delay
;
708 /* getinfo_super might have lost these ... */
709 inf
->disk
.major
= major(stb
.st_rdev
);
710 inf
->disk
.minor
= minor(stb
.st_rdev
);
716 rv
= add_disk(mdfd
, st
, &info
, inf
);
720 Name
": ADD_NEW_DISK for %s "
722 dv
->devname
, strerror(errno
));
723 st
->ss
->free_super(st
);
728 if (dv
== moved_disk
&& dnum
!= insert_point
) break;
731 st
->ss
->write_init_super(st
);
732 flush_metadata_updates(st
);
736 st
->ss
->free_super(st
);
738 if (level
== LEVEL_CONTAINER
) {
739 /* No need to start. But we should signal udev to
741 sysfs_uevent(&info
, "change");
743 } else if (runstop
== 1 || subdevs
>= raiddisks
) {
744 if (st
->ss
->external
) {
747 case LEVEL_MULTIPATH
:
749 sysfs_set_str(&info
, NULL
, "array_state",
754 sysfs_set_str(&info
, NULL
, "array_state",
758 sysfs_set_safemode(&info
, safe_mode_delay
);
760 /* param is not actually used */
762 if (ioctl(mdfd
, RUN_ARRAY
, ¶m
)) {
763 fprintf(stderr
, Name
": RUN_ARRAY failed: %s\n",
765 Manage_runstop(mddev
, mdfd
, -1, 0);
770 fprintf(stderr
, Name
": array %s started.\n", mddev
);
771 if (st
->ss
->external
&& st
->subarray
[0]) {
773 start_mdmon(st
->container_dev
);
775 ping_monitor(devnum2devname(st
->container_dev
));
779 fprintf(stderr
, Name
": not starting array - not enough devices.\n");