2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 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>
25 #define HAVE_STDINT_H 1
29 * All handling for the 0.90.0 version superblock is in
32 * - finding, loading, and writing the superblock.
33 * - initialising a new superblock
34 * - printing the superblock for --examine
35 * - printing part of the superblock for --detail
39 static unsigned long calc_sb0_csum(mdp_super_t
*super
)
41 unsigned long csum
= super
->sb_csum
;
42 unsigned long newcsum
;
44 newcsum
= calc_csum(super
, MD_SB_BYTES
);
45 super
->sb_csum
= csum
;
49 static void super0_swap_endian(struct mdp_superblock_s
*sb
)
51 /* as super0 superblocks are host-endian, it is sometimes
52 * useful to be able to swap the endianness
53 * as (almost) everything is u32's we byte-swap every 4byte
55 * We then also have to swap the events_hi and events_lo
57 char *sbc
= (char *)sb
;
61 for (i
=0; i
< MD_SB_BYTES
; i
+=4) {
70 sb
->events_hi
= sb
->events_lo
;
73 t32
= sb
->cp_events_hi
;
74 sb
->cp_events_hi
= sb
->cp_events_lo
;
75 sb
->cp_events_lo
= t32
;
81 static void examine_super0(struct supertype
*st
, char *homehost
)
83 mdp_super_t
*sb
= st
->sb
;
89 printf(" Magic : %08x\n", sb
->md_magic
);
90 printf(" Version : %d.%02d.%02d\n",
91 sb
->major_version
, sb
->minor_version
, sb
->patch_version
);
92 if (sb
->minor_version
>= 90) {
93 printf(" UUID : %08x:%08x:%08x:%08x", sb
->set_uuid0
,
94 sb
->set_uuid1
, sb
->set_uuid2
, sb
->set_uuid3
);
99 hash
= sha1_buffer(homehost
, strlen(homehost
), buf
);
100 if (memcmp(&sb
->set_uuid2
, hash
, 8) == 0)
101 printf(" (local to host %s)", homehost
);
105 printf(" UUID : %08x\n", sb
->set_uuid0
);
107 if (sb
->not_persistent
)
108 printf(" Eedk : not persistent\n");
111 printf(" Creation Time : %.24s\n", ctime(&atime
));
112 c
= map_num(pers
, sb
->level
);
113 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
114 if ((int)sb
->level
> 0) {
115 int ddsks
= 0, ddsks_denom
= 1;
116 printf(" Used Dev Size : %d%s\n", sb
->size
,
117 human_size((long long)sb
->size
<<10));
124 ddsks
= sb
->raid_disks
- 1;
127 ddsks
= sb
->raid_disks
- 2;
130 ddsks
= sb
->raid_disks
;
132 (sb
->layout
& 255) * ((sb
->layout
>> 8) & 255);
135 long long asize
= sb
->size
;
136 asize
= (asize
<< 10) * ddsks
/ ddsks_denom
;
137 printf(" Array Size : %llu%s\n",
138 asize
>> 10, human_size(asize
));
141 printf(" Raid Devices : %d\n", sb
->raid_disks
);
142 printf(" Total Devices : %d\n", sb
->nr_disks
);
143 printf("Preferred Minor : %d\n", sb
->md_minor
);
145 if (sb
->minor_version
> 90 && (sb
->reshape_position
+ 1) != 0) {
146 printf(" Reshape pos'n : %llu%s\n",
147 (unsigned long long)sb
->reshape_position
/ 2,
148 human_size((long long)sb
->reshape_position
<< 9));
149 if (sb
->delta_disks
) {
150 printf(" Delta Devices : %d", sb
->delta_disks
);
151 printf(" (%d->%d)\n", sb
->raid_disks
-sb
->delta_disks
,
153 if (((int)sb
->delta_disks
) < 0)
154 delta_extra
= - sb
->delta_disks
;
156 if (sb
->new_level
!= sb
->level
) {
157 c
= map_num(pers
, sb
->new_level
);
158 printf(" New Level : %s\n", c
?c
:"-unknown-");
160 if (sb
->new_layout
!= sb
->layout
) {
161 if (sb
->level
== 5) {
162 c
= map_num(r5layout
, sb
->new_layout
);
163 printf(" New Layout : %s\n",
166 if (sb
->level
== 6) {
167 c
= map_num(r6layout
, sb
->new_layout
);
168 printf(" New Layout : %s\n",
171 if (sb
->level
== 10) {
172 printf(" New Layout : near=%d, %s=%d\n",
174 (sb
->new_layout
&0x10000)?"offset":"far",
175 (sb
->new_layout
>>8)&255);
178 if (sb
->new_chunk
!= sb
->chunk_size
)
179 printf(" New Chunksize : %d\n", sb
->new_chunk
);
183 printf(" Update Time : %.24s\n", ctime(&atime
));
184 printf(" State : %s\n",
185 (sb
->state
&(1 << MD_SB_CLEAN
)) ? "clean":"active");
186 if (sb
->state
& (1 << MD_SB_BITMAP_PRESENT
))
187 printf("Internal Bitmap : present\n");
188 printf(" Active Devices : %d\n", sb
->active_disks
);
189 printf("Working Devices : %d\n", sb
->working_disks
);
190 printf(" Failed Devices : %d\n", sb
->failed_disks
);
191 printf(" Spare Devices : %d\n", sb
->spare_disks
);
192 if (calc_sb0_csum(sb
) == sb
->sb_csum
)
193 printf(" Checksum : %x - correct\n", sb
->sb_csum
);
195 printf(" Checksum : %x - expected %lx\n",
196 sb
->sb_csum
, calc_sb0_csum(sb
));
197 printf(" Events : %llu\n",
198 ((unsigned long long)sb
->events_hi
<< 32) + sb
->events_lo
);
200 if (sb
->level
== 5) {
201 c
= map_num(r5layout
, sb
->layout
);
202 printf(" Layout : %s\n", c
?c
:"-unknown-");
204 if (sb
->level
== 6) {
205 c
= map_num(r6layout
, sb
->layout
);
206 printf(" Layout : %s\n", c
?c
:"-unknown-");
208 if (sb
->level
== 10) {
210 print_r10_layout(sb
->layout
);
219 printf(" Chunk Size : %dK\n", sb
->chunk_size
/ 1024);
222 printf(" Rounding : %dK\n", sb
->chunk_size
/ 1024);
228 printf(" Number Major Minor RaidDevice State\n");
230 d
< (signed int)(sb
->raid_disks
+ delta_extra
+ sb
->spare_disks
);
236 if (d
>=0) dp
= &sb
->disks
[d
];
237 else dp
= &sb
->this_disk
;
238 snprintf(nb
, sizeof(nb
), "%4d", d
);
239 printf("%4s %5d %5d %5d %5d ", d
< 0 ? "this" : nb
,
240 dp
->number
, dp
->major
, dp
->minor
, dp
->raid_disk
);
241 wonly
= dp
->state
& (1 << MD_DISK_WRITEMOSTLY
);
242 failfast
= dp
->state
& (1<<MD_DISK_FAILFAST
);
243 dp
->state
&= ~(wonly
| failfast
);
244 if (dp
->state
& (1 << MD_DISK_FAULTY
))
246 if (dp
->state
& (1 << MD_DISK_ACTIVE
))
248 if (dp
->state
& (1 << MD_DISK_SYNC
))
250 if (dp
->state
& (1 << MD_DISK_REMOVED
))
253 printf(" write-mostly");
258 if ((dv
= map_dev(dp
->major
, dp
->minor
, 0)))
266 static void brief_examine_super0(struct supertype
*st
, int verbose
)
268 mdp_super_t
*sb
= st
->sb
;
269 char *c
=map_num(pers
, sb
->level
);
272 sprintf(devname
, "/dev/md%d", sb
->md_minor
);
275 printf("ARRAY %s level=%s num-devices=%d",
277 c
?c
:"-unknown-", sb
->raid_disks
);
279 printf("ARRAY %s", devname
);
281 if (sb
->minor_version
>= 90)
282 printf(" UUID=%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
283 sb
->set_uuid2
, sb
->set_uuid3
);
285 printf(" UUID=%08x", sb
->set_uuid0
);
289 static void export_examine_super0(struct supertype
*st
)
291 mdp_super_t
*sb
= st
->sb
;
293 printf("MD_LEVEL=%s\n", map_num(pers
, sb
->level
));
294 printf("MD_DEVICES=%d\n", sb
->raid_disks
);
295 if (sb
->minor_version
>= 90)
296 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
297 sb
->set_uuid0
, sb
->set_uuid1
,
298 sb
->set_uuid2
, sb
->set_uuid3
);
300 printf("MD_UUID=%08x\n", sb
->set_uuid0
);
301 printf("MD_UPDATE_TIME=%llu\n",
302 __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
);
303 printf("MD_EVENTS=%llu\n",
304 ((unsigned long long)sb
->events_hi
<< 32)
308 static int copy_metadata0(struct supertype
*st
, int from
, int to
)
310 /* Read 64K from the appropriate offset of 'from'
311 * and if it looks a little like a 0.90 superblock,
312 * write it to the same offset of 'to'
315 unsigned long long dsize
, offset
;
316 const int bufsize
= 64*1024;
319 if (posix_memalign(&buf
, 4096, bufsize
) != 0)
322 if (!get_dev_size(from
, NULL
, &dsize
))
325 if (dsize
< MD_RESERVED_SECTORS
*512)
328 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
332 if (lseek64(from
, offset
, 0) < 0LL)
334 if (read(from
, buf
, bufsize
) != bufsize
)
337 if (lseek64(to
, offset
, 0) < 0LL)
340 if (super
->md_magic
!= MD_SB_MAGIC
||
341 super
->major_version
!= 0 ||
342 calc_sb0_csum(super
) != super
->sb_csum
)
344 if (write(to
, buf
, bufsize
) != bufsize
)
353 static void detail_super0(struct supertype
*st
, char *homehost
)
355 mdp_super_t
*sb
= st
->sb
;
357 if (sb
->minor_version
>= 90)
358 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
359 sb
->set_uuid2
, sb
->set_uuid3
);
361 printf("%08x", sb
->set_uuid0
);
364 void *hash
= sha1_buffer(homehost
,
367 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
368 printf(" (local to host %s)", homehost
);
370 printf("\n Events : %d.%d\n\n", sb
->events_hi
, sb
->events_lo
);
373 static void brief_detail_super0(struct supertype
*st
)
375 mdp_super_t
*sb
= st
->sb
;
377 if (sb
->minor_version
>= 90)
378 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
379 sb
->set_uuid2
, sb
->set_uuid3
);
381 printf("%08x", sb
->set_uuid0
);
385 static int match_home0(struct supertype
*st
, char *homehost
)
387 mdp_super_t
*sb
= st
->sb
;
393 hash
= sha1_buffer(homehost
,
397 return (memcmp(&sb
->set_uuid2
, hash
, 8)==0);
400 static void uuid_from_super0(struct supertype
*st
, int uuid
[4])
402 mdp_super_t
*super
= st
->sb
;
403 uuid
[0] = super
->set_uuid0
;
404 if (super
->minor_version
>= 90) {
405 uuid
[1] = super
->set_uuid1
;
406 uuid
[2] = super
->set_uuid2
;
407 uuid
[3] = super
->set_uuid3
;
415 static void getinfo_super0(struct supertype
*st
, struct mdinfo
*info
, char *map
)
417 mdp_super_t
*sb
= st
->sb
;
420 int map_disks
= info
->array
.raid_disks
;
422 memset(info
, 0, sizeof(*info
));
423 info
->array
.major_version
= sb
->major_version
;
424 info
->array
.minor_version
= sb
->minor_version
;
425 info
->array
.patch_version
= sb
->patch_version
;
426 info
->array
.raid_disks
= sb
->raid_disks
;
427 info
->array
.level
= sb
->level
;
428 info
->array
.layout
= sb
->layout
;
429 info
->array
.md_minor
= sb
->md_minor
;
430 info
->array
.ctime
= sb
->ctime
;
431 info
->array
.utime
= sb
->utime
;
432 info
->array
.chunk_size
= sb
->chunk_size
;
433 info
->array
.state
= sb
->state
;
434 info
->component_size
= sb
->size
;
435 info
->component_size
*= 2;
437 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
))
438 info
->bitmap_offset
= 8;
440 info
->disk
.state
= sb
->this_disk
.state
;
441 info
->disk
.major
= sb
->this_disk
.major
;
442 info
->disk
.minor
= sb
->this_disk
.minor
;
443 info
->disk
.raid_disk
= sb
->this_disk
.raid_disk
;
444 info
->disk
.number
= sb
->this_disk
.number
;
446 info
->events
= md_event(sb
);
447 info
->data_offset
= 0;
449 sprintf(info
->text_version
, "0.%d", sb
->minor_version
);
450 info
->safe_mode_delay
= 200;
452 uuid_from_super0(st
, info
->uuid
);
454 info
->recovery_start
= MaxSector
;
455 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
456 info
->reshape_active
= 1;
457 info
->reshape_progress
= sb
->reshape_position
;
458 info
->new_level
= sb
->new_level
;
459 info
->delta_disks
= sb
->delta_disks
;
460 info
->new_layout
= sb
->new_layout
;
461 info
->new_chunk
= sb
->new_chunk
;
462 if (info
->delta_disks
< 0)
463 info
->array
.raid_disks
-= info
->delta_disks
;
465 info
->reshape_active
= 0;
467 info
->recovery_blocked
= info
->reshape_active
;
469 sprintf(info
->name
, "%d", sb
->md_minor
);
470 /* work_disks is calculated rather than read directly */
471 for (i
=0; i
< MD_SB_DISKS
; i
++)
472 if ((sb
->disks
[i
].state
& (1<<MD_DISK_SYNC
)) &&
473 (sb
->disks
[i
].raid_disk
< (unsigned)info
->array
.raid_disks
) &&
474 (sb
->disks
[i
].state
& (1<<MD_DISK_ACTIVE
)) &&
475 !(sb
->disks
[i
].state
& (1<<MD_DISK_FAULTY
))) {
477 if (map
&& i
< map_disks
)
479 } else if (map
&& i
< map_disks
)
481 info
->array
.working_disks
= working
;
484 static struct mdinfo
*container_content0(struct supertype
*st
, char *subarray
)
491 info
= xmalloc(sizeof(*info
));
492 getinfo_super0(st
, info
, NULL
);
496 static int update_super0(struct supertype
*st
, struct mdinfo
*info
,
498 char *devname
, int verbose
,
499 int uuid_set
, char *homehost
)
501 /* NOTE: for 'assemble' and 'force' we need to return non-zero
502 * if any change was made. For others, the return value is
507 mdp_super_t
*sb
= st
->sb
;
509 if (strcmp(update
, "homehost") == 0 &&
511 /* note that 'homehost' is special as it is really
516 info
->uuid
[0] = sb
->set_uuid0
;
517 info
->uuid
[1] = sb
->set_uuid1
;
520 if (strcmp(update
, "sparc2.2")==0 ) {
521 /* 2.2 sparc put the events in the wrong place
522 * So we copy the tail of the superblock
523 * up 4 bytes before continuing
525 __u32
*sb32
= (__u32
*)sb
;
526 memcpy(sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7,
527 sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7+1,
528 (MD_SB_WORDS
- (MD_SB_GENERIC_CONSTANT_WORDS
+7+1))*4);
530 pr_err("adjusting superblock of %s for 2.2/sparc compatibility.\n",
532 } else if (strcmp(update
, "super-minor") ==0) {
533 sb
->md_minor
= info
->array
.md_minor
;
535 pr_err("updating superblock of %s with minor number %d\n",
536 devname
, info
->array
.md_minor
);
537 } else if (strcmp(update
, "summaries") == 0) {
539 /* set nr_disks, active_disks, working_disks,
540 * failed_disks, spare_disks based on disks[]
541 * array in superblock.
542 * Also make sure extra slots aren't 'failed'
544 sb
->nr_disks
= sb
->active_disks
=
545 sb
->working_disks
= sb
->failed_disks
=
547 for (i
=0; i
< MD_SB_DISKS
; i
++)
548 if (sb
->disks
[i
].major
||
549 sb
->disks
[i
].minor
) {
550 int state
= sb
->disks
[i
].state
;
551 if (state
& (1<<MD_DISK_REMOVED
))
554 if (state
& (1<<MD_DISK_ACTIVE
))
556 if (state
& (1<<MD_DISK_FAULTY
))
562 } else if (i
>= sb
->raid_disks
&& sb
->disks
[i
].number
== 0)
563 sb
->disks
[i
].state
= 0;
564 } else if (strcmp(update
, "force-one")==0) {
565 /* Not enough devices for a working array, so
566 * bring this one up-to-date.
568 __u32 ehi
= sb
->events_hi
, elo
= sb
->events_lo
;
569 sb
->events_hi
= (info
->events
>>32) & 0xFFFFFFFF;
570 sb
->events_lo
= (info
->events
) & 0xFFFFFFFF;
571 if (sb
->events_hi
!= ehi
||
572 sb
->events_lo
!= elo
)
574 } else if (strcmp(update
, "force-array")==0) {
575 /* degraded array and 'force' requested, so
576 * maybe need to mark it 'clean'
578 if ((sb
->level
== 5 || sb
->level
== 4 || sb
->level
== 6) &&
579 (sb
->state
& (1 << MD_SB_CLEAN
)) == 0) {
580 /* need to force clean */
581 sb
->state
|= (1 << MD_SB_CLEAN
);
584 } else if (strcmp(update
, "assemble")==0) {
585 int d
= info
->disk
.number
;
586 int wonly
= sb
->disks
[d
].state
& (1<<MD_DISK_WRITEMOSTLY
);
587 int failfast
= sb
->disks
[d
].state
& (1<<MD_DISK_FAILFAST
);
588 int mask
= (1<<MD_DISK_WRITEMOSTLY
)|(1<<MD_DISK_FAILFAST
);
590 if (sb
->minor_version
>= 91)
591 /* During reshape we don't insist on everything
592 * being marked 'sync'
594 add
= (1<<MD_DISK_SYNC
);
595 if (((sb
->disks
[d
].state
& ~mask
) | add
)
596 != (unsigned)info
->disk
.state
) {
597 sb
->disks
[d
].state
= info
->disk
.state
| wonly
|failfast
;
600 if (info
->reshape_active
&&
601 sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0 &&
602 info
->delta_disks
>= 0 &&
603 info
->reshape_progress
< sb
->reshape_position
) {
604 sb
->reshape_position
= info
->reshape_progress
;
607 if (info
->reshape_active
&&
608 sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0 &&
609 info
->delta_disks
< 0 &&
610 info
->reshape_progress
> sb
->reshape_position
) {
611 sb
->reshape_position
= info
->reshape_progress
;
614 } else if (strcmp(update
, "linear-grow-new") == 0) {
615 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
616 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
617 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
618 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
619 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
620 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
621 sb
->this_disk
= sb
->disks
[info
->disk
.number
];
622 } else if (strcmp(update
, "linear-grow-update") == 0) {
623 sb
->raid_disks
= info
->array
.raid_disks
;
624 sb
->nr_disks
= info
->array
.nr_disks
;
625 sb
->active_disks
= info
->array
.active_disks
;
626 sb
->working_disks
= info
->array
.working_disks
;
627 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
628 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
629 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
630 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
631 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
632 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
633 } else if (strcmp(update
, "resync") == 0) {
634 /* make sure resync happens */
635 sb
->state
&= ~(1<<MD_SB_CLEAN
);
637 } else if (strcmp(update
, "uuid") == 0) {
638 if (!uuid_set
&& homehost
) {
640 char *hash
= sha1_buffer(homehost
,
643 memcpy(info
->uuid
+2, hash
, 8);
645 sb
->set_uuid0
= info
->uuid
[0];
646 sb
->set_uuid1
= info
->uuid
[1];
647 sb
->set_uuid2
= info
->uuid
[2];
648 sb
->set_uuid3
= info
->uuid
[3];
649 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
650 struct bitmap_super_s
*bm
;
651 bm
= (struct bitmap_super_s
*)(sb
+1);
652 uuid_from_super0(st
, uuid
);
653 memcpy(bm
->uuid
, uuid
, 16);
655 } else if (strcmp(update
, "metadata") == 0) {
656 /* Create some v1.0 metadata to match ours but make the
657 * ctime bigger. Also update info->array.*_version.
658 * We need to arrange that store_super writes out
660 * Not permitted for unclean array, or array with
663 if (info
->bitmap_offset
) {
664 pr_err("Cannot update metadata when bitmap is present\n");
666 } else if (info
->array
.state
!= 1) {
667 pr_err("Cannot update metadata on unclean array\n");
670 info
->array
.major_version
= 1;
671 info
->array
.minor_version
= 0;
672 uuid_from_super0(st
, info
->uuid
);
673 st
->other
= super1_make_v0(st
, info
, st
->sb
);
675 } else if (strcmp(update
, "revert-reshape") == 0) {
677 if (sb
->minor_version
<= 90)
678 pr_err("No active reshape to revert on %s\n",
680 else if (sb
->delta_disks
== 0)
681 pr_err("%s: Can only revert reshape which changes number of devices\n",
685 int parity
= sb
->level
== 6 ? 2 : 1;
688 if (sb
->level
>= 4 && sb
->level
<= 6 &&
689 sb
->reshape_position
% (
691 (sb
->raid_disks
- sb
->delta_disks
- parity
))) {
692 pr_err("Reshape position is not suitably aligned.\n");
693 pr_err("Try normal assembly and stop again\n");
696 sb
->raid_disks
-= sb
->delta_disks
;
697 sb
->delta_disks
= -sb
->delta_disks
;
699 tmp
= sb
->new_layout
;
700 sb
->new_layout
= sb
->layout
;
704 sb
->new_chunk
= sb
->chunk_size
;
705 sb
->chunk_size
= tmp
;
707 } else if (strcmp(update
, "no-bitmap") == 0) {
708 sb
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
709 } else if (strcmp(update
, "_reshape_progress")==0)
710 sb
->reshape_position
= info
->reshape_progress
;
711 else if (strcmp(update
, "writemostly")==0)
712 sb
->state
|= (1<<MD_DISK_WRITEMOSTLY
);
713 else if (strcmp(update
, "readwrite")==0)
714 sb
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
718 sb
->sb_csum
= calc_sb0_csum(sb
);
723 * For version-0 superblock, the homehost is 'stored' in the uuid.
724 * 8 bytes for a hash of the host leaving 8 bytes of random material.
725 * We use the first 8 bytes (64bits) of the sha1 of the host name
727 static int init_super0(struct supertype
*st
, mdu_array_info_t
*info
,
728 struct shape
*s
, char *ignored_name
,
729 char *homehost
, int *uuid
,
730 unsigned long long data_offset
)
735 if (data_offset
!= INVALID_SECTORS
) {
736 pr_err("data-offset not support for 0.90\n");
740 if (posix_memalign((void**)&sb
, 4096,
741 MD_SB_BYTES
+ ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
742 pr_err("could not allocate superblock\n");
745 memset(sb
, 0, MD_SB_BYTES
+ sizeof(bitmap_super_t
));
749 /* zeroing the superblock */
753 spares
= info
->working_disks
- info
->active_disks
;
754 if (info
->raid_disks
+ spares
> MD_SB_DISKS
) {
755 pr_err("too many devices requested: %d+%d > %d\n",
756 info
->raid_disks
, spares
, MD_SB_DISKS
);
760 sb
->md_magic
= MD_SB_MAGIC
;
761 sb
->major_version
= 0;
762 sb
->minor_version
= 90;
763 sb
->patch_version
= 0;
764 sb
->gvalid_words
= 0; /* ignored */
766 sb
->level
= info
->level
;
768 if (s
->size
!= (unsigned long long)sb
->size
)
770 sb
->nr_disks
= info
->nr_disks
;
771 sb
->raid_disks
= info
->raid_disks
;
772 sb
->md_minor
= info
->md_minor
;
773 sb
->not_persistent
= 0;
775 sb
->set_uuid0
= uuid
[0];
776 sb
->set_uuid1
= uuid
[1];
777 sb
->set_uuid2
= uuid
[2];
778 sb
->set_uuid3
= uuid
[3];
781 random_uuid((__u8
*)r
);
782 sb
->set_uuid0
= r
[0];
783 sb
->set_uuid1
= r
[1];
784 sb
->set_uuid2
= r
[2];
785 sb
->set_uuid3
= r
[3];
787 if (homehost
&& !uuid
) {
789 char *hash
= sha1_buffer(homehost
,
792 memcpy(&sb
->set_uuid2
, hash
, 8);
795 sb
->utime
= sb
->ctime
;
796 sb
->state
= info
->state
;
797 sb
->active_disks
= info
->active_disks
;
798 sb
->working_disks
= info
->working_disks
;
799 sb
->failed_disks
= info
->failed_disks
;
800 sb
->spare_disks
= info
->spare_disks
;
804 sb
->layout
= info
->layout
;
805 sb
->chunk_size
= info
->chunk_size
;
813 mdu_disk_info_t disk
;
814 struct devinfo
*next
;
818 /* Add a device to the superblock being created */
819 static int add_to_super0(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
820 int fd
, char *devname
, unsigned long long data_offset
)
822 mdp_super_t
*sb
= st
->sb
;
823 mdp_disk_t
*dk
= &sb
->disks
[dinfo
->number
];
824 struct devinfo
*di
, **dip
;
826 dk
->number
= dinfo
->number
;
827 dk
->major
= dinfo
->major
;
828 dk
->minor
= dinfo
->minor
;
829 dk
->raid_disk
= dinfo
->raid_disk
;
830 dk
->state
= dinfo
->state
& ((1<<MD_DISK_ACTIVE
) |
833 sb
->this_disk
= sb
->disks
[dinfo
->number
];
834 sb
->sb_csum
= calc_sb0_csum(sb
);
836 dip
= (struct devinfo
**)&st
->info
;
839 di
= xmalloc(sizeof(struct devinfo
));
841 di
->devname
= devname
;
850 static int store_super0(struct supertype
*st
, int fd
)
852 unsigned long long dsize
;
853 unsigned long long offset
;
854 mdp_super_t
*super
= st
->sb
;
856 if (!get_dev_size(fd
, NULL
, &dsize
))
859 if (dsize
< MD_RESERVED_SECTORS
*512)
863 /* Writing out v1.0 metadata for --update=metadata */
866 offset
= dsize
/512 - 8*2;
869 if (lseek64(fd
, offset
, 0)< 0LL)
871 else if (write(fd
, st
->other
, 1024) != 1024)
880 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
884 if (lseek64(fd
, offset
, 0)< 0LL)
887 if (write(fd
, super
, sizeof(*super
)) != sizeof(*super
))
890 if (super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
891 struct bitmap_super_s
* bm
= (struct bitmap_super_s
*)(super
+1);
892 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
)
893 if (write(fd
, bm
, ROUND_UP(sizeof(*bm
),4096)) !=
894 ROUND_UP(sizeof(*bm
),4096))
903 static int write_init_super0(struct supertype
*st
)
905 mdp_super_t
*sb
= st
->sb
;
909 for (di
= st
->info
; di
&& ! rv
; di
= di
->next
) {
911 if (di
->disk
.state
& (1 << MD_DISK_FAULTY
))
915 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
918 sb
->disks
[di
->disk
.number
].state
&= ~(1<<MD_DISK_FAULTY
);
920 sb
->this_disk
= sb
->disks
[di
->disk
.number
];
921 sb
->sb_csum
= calc_sb0_csum(sb
);
922 rv
= store_super0(st
, di
->fd
);
924 if (rv
== 0 && (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)))
925 rv
= st
->ss
->write_bitmap(st
, di
->fd
, NoUpdate
);
928 pr_err("failed to write superblock to %s\n",
935 static int compare_super0(struct supertype
*st
, struct supertype
*tst
)
939 * 0 same, or first was empty, and second was copied
940 * 1 second had wrong number
944 mdp_super_t
*first
= st
->sb
;
945 mdp_super_t
*second
= tst
->sb
;
946 int uuid1
[4], uuid2
[4];
948 if (second
->md_magic
!= MD_SB_MAGIC
)
951 if (posix_memalign((void**)&first
, 4096,
953 ROUND_UP(sizeof(struct bitmap_super_s
), 4096)) != 0) {
954 pr_err("could not allocate superblock\n");
957 memcpy(first
, second
, MD_SB_BYTES
+ sizeof(struct bitmap_super_s
));
962 uuid_from_super0(st
, uuid1
);
963 uuid_from_super0(tst
, uuid2
);
964 if (!same_uuid(uuid1
, uuid2
, 0))
966 if (first
->major_version
!= second
->major_version
||
967 first
->minor_version
!= second
->minor_version
||
968 first
->patch_version
!= second
->patch_version
||
969 first
->gvalid_words
!= second
->gvalid_words
||
970 first
->ctime
!= second
->ctime
||
971 first
->level
!= second
->level
||
972 first
->size
!= second
->size
||
973 first
->raid_disks
!= second
->raid_disks
)
979 static void free_super0(struct supertype
*st
);
981 static int load_super0(struct supertype
*st
, int fd
, char *devname
)
983 /* try to read in the superblock
986 * 1 on cannot get superblock
987 * 2 on superblock meaningless
989 unsigned long long dsize
;
990 unsigned long long offset
;
993 struct bitmap_super_s
*bsb
;
997 if (!get_dev_size(fd
, devname
, &dsize
))
1000 if (dsize
< MD_RESERVED_SECTORS
*512) {
1002 pr_err("%s is too small for md: size is %llu sectors.\n",
1006 st
->devsize
= dsize
;
1008 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1012 if (lseek64(fd
, offset
, 0)< 0LL) {
1014 pr_err("Cannot seek to superblock on %s: %s\n",
1015 devname
, strerror(errno
));
1019 if (posix_memalign((void**)&super
, 4096,
1021 ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
1022 pr_err("could not allocate superblock\n");
1026 if (read(fd
, super
, sizeof(*super
)) != MD_SB_BYTES
) {
1028 pr_err("Cannot read superblock on %s\n",
1034 if (st
->ss
&& st
->minor_version
== 9)
1035 super0_swap_endian(super
);
1037 if (super
->md_magic
!= MD_SB_MAGIC
) {
1039 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1040 devname
, MD_SB_MAGIC
, super
->md_magic
);
1045 if (super
->major_version
!= 0) {
1047 pr_err("Cannot interpret superblock on %s - version is %d\n",
1048 devname
, super
->major_version
);
1054 if (st
->ss
== NULL
) {
1056 st
->minor_version
= super
->minor_version
;
1057 st
->max_devs
= MD_SB_DISKS
;
1061 /* Now check on the bitmap superblock */
1062 if ((super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) == 0)
1064 /* Read the bitmap superblock and make sure it looks
1065 * valid. If it doesn't clear the bit. An --assemble --force
1066 * should get that written out.
1068 if (read(fd
, super
+1, ROUND_UP(sizeof(struct bitmap_super_s
),4096))
1069 != ROUND_UP(sizeof(struct bitmap_super_s
),4096))
1072 uuid_from_super0(st
, uuid
);
1073 bsb
= (struct bitmap_super_s
*)(super
+1);
1074 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1075 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1080 super
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
1085 static struct supertype
*match_metadata_desc0(char *arg
)
1087 struct supertype
*st
= xcalloc(1, sizeof(*st
));
1089 st
->container_devnm
[0] = 0;
1092 st
->minor_version
= 90;
1093 st
->max_devs
= MD_SB_DISKS
;
1095 /* we sometimes get 00.90 */
1096 while (arg
[0] == '0' && arg
[1] == '0')
1098 if (strcmp(arg
, "0") == 0 ||
1099 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
1100 strcmp(arg
, "default") == 0 ||
1101 #endif /* DEFAULT_OLD_METADATA */
1102 strcmp(arg
, "0.90") == 0 ||
1103 strcmp(arg
, "") == 0 /* no metadata - i.e. non_persistent */
1107 st
->minor_version
= 91; /* reshape in progress */
1108 if (strcmp(arg
, "0.91") == 0) /* For dup_super support */
1111 st
->minor_version
= 9; /* flag for 'byte-swapped' */
1112 if (strcmp(arg
, "0.swap")==0 ||
1113 strcmp(arg
, "0.9") == 0) /* For dup_super support */
1120 static __u64
avail_size0(struct supertype
*st
, __u64 devsize
,
1121 unsigned long long data_offset
)
1123 if (data_offset
!= 0 && data_offset
!= INVALID_SECTORS
)
1125 if (devsize
< MD_RESERVED_SECTORS
)
1127 return MD_NEW_SIZE_SECTORS(devsize
);
1130 static int add_internal_bitmap0(struct supertype
*st
, int *chunkp
,
1131 int delay
, int write_behind
,
1132 unsigned long long size
, int may_change
,
1136 * The bitmap comes immediately after the superblock and must be 60K in size
1137 * at most. The default size is between 30K and 60K
1139 * size is in sectors, chunk is in bytes !!!
1141 unsigned long long bits
;
1142 unsigned long long max_bits
= (60*1024 - sizeof(bitmap_super_t
))*8;
1143 unsigned long long min_chunk
;
1144 int chunk
= *chunkp
;
1145 mdp_super_t
*sb
= st
->sb
;
1146 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MD_SB_BYTES
);
1149 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1150 bits
= (size
* 512) / min_chunk
+ 1;
1151 while (bits
> max_bits
) {
1155 if (chunk
== UnSet
) {
1156 /* A chunk size less than a few Megabytes gives poor
1157 * performance without increasing resync noticeably
1160 if (chunk
< 64*1024*1024)
1161 chunk
= 64*1024*1024;
1162 } else if ((unsigned long long)chunk
< min_chunk
)
1163 return -EINVAL
; /* chunk size too small */
1165 sb
->state
|= (1<<MD_SB_BITMAP_PRESENT
);
1167 memset(bms
, 0, sizeof(*bms
));
1168 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1169 bms
->version
= __cpu_to_le32(major
);
1170 uuid_from_super0(st
, uuid
);
1171 memcpy(bms
->uuid
, uuid
, 16);
1172 bms
->chunksize
= __cpu_to_le32(chunk
);
1173 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1174 bms
->sync_size
= __cpu_to_le64(size
);
1175 bms
->write_behind
= __cpu_to_le32(write_behind
);
1180 static int locate_bitmap0(struct supertype
*st
, int fd
, int node_num
)
1182 unsigned long long dsize
;
1183 unsigned long long offset
;
1185 if (!get_dev_size(fd
, NULL
, &dsize
))
1188 if (dsize
< MD_RESERVED_SECTORS
*512)
1191 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1195 offset
+= MD_SB_BYTES
;
1197 lseek64(fd
, offset
, 0);
1201 static int write_bitmap0(struct supertype
*st
, int fd
, enum bitmap_update update
)
1203 unsigned long long dsize
;
1204 unsigned long long offset
;
1205 mdp_super_t
*sb
= st
->sb
;
1212 if (!get_dev_size(fd
, NULL
, &dsize
))
1215 if (dsize
< MD_RESERVED_SECTORS
*512)
1218 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1222 if (lseek64(fd
, offset
+ 4096, 0)< 0LL)
1225 if (posix_memalign(&buf
, 4096, 4096))
1228 memset(buf
, 0xff, 4096);
1229 memcpy(buf
, ((char*)sb
)+MD_SB_BYTES
, sizeof(bitmap_super_t
));
1231 while (towrite
> 0) {
1235 n
= write(fd
, buf
, n
);
1240 memset(buf
, 0xff, 4096);
1250 static void free_super0(struct supertype
*st
)
1255 struct devinfo
*di
= st
->info
;
1256 st
->info
= di
->next
;
1265 static int validate_geometry0(struct supertype
*st
, int level
,
1266 int layout
, int raiddisks
,
1267 int *chunk
, unsigned long long size
,
1268 unsigned long long data_offset
,
1269 char *subdev
, unsigned long long *freesize
,
1270 int consistency_policy
, int verbose
)
1272 unsigned long long ldsize
;
1274 unsigned int tbmax
= 4;
1276 /* prior to linux 3.1, a but limits usable device size to 2TB.
1277 * It was introduced in 2.6.29, but we won't worry about that detail
1279 if (get_linux_version() < 3001000)
1282 if (level
== LEVEL_CONTAINER
) {
1284 pr_err("0.90 metadata does not support containers\n");
1287 if (raiddisks
> MD_SB_DISKS
) {
1289 pr_err("0.90 metadata supports at most %d devices per array\n",
1293 if (size
>= tbmax
* 2ULL*1024*1024*1024) {
1295 pr_err("0.90 metadata supports at most %d terabytes per device\n", tbmax
);
1298 if (*chunk
== UnSet
)
1299 *chunk
= DEFAULT_CHUNK
;
1304 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1307 pr_err("super0.90 cannot open %s: %s\n",
1308 subdev
, strerror(errno
));
1312 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1318 if (ldsize
< MD_RESERVED_SECTORS
* 512)
1320 *freesize
= MD_NEW_SIZE_SECTORS(ldsize
>> 9);
1323 #endif /* MDASSEMBLE */
1325 struct superswitch super0
= {
1327 .examine_super
= examine_super0
,
1328 .brief_examine_super
= brief_examine_super0
,
1329 .export_examine_super
= export_examine_super0
,
1330 .detail_super
= detail_super0
,
1331 .brief_detail_super
= brief_detail_super0
,
1332 .write_init_super
= write_init_super0
,
1333 .validate_geometry
= validate_geometry0
,
1334 .add_to_super
= add_to_super0
,
1335 .copy_metadata
= copy_metadata0
,
1337 .match_home
= match_home0
,
1338 .uuid_from_super
= uuid_from_super0
,
1339 .getinfo_super
= getinfo_super0
,
1340 .container_content
= container_content0
,
1341 .update_super
= update_super0
,
1342 .init_super
= init_super0
,
1343 .store_super
= store_super0
,
1344 .compare_super
= compare_super0
,
1345 .load_super
= load_super0
,
1346 .match_metadata_desc
= match_metadata_desc0
,
1347 .avail_size
= avail_size0
,
1348 .add_internal_bitmap
= add_internal_bitmap0
,
1349 .locate_bitmap
= locate_bitmap0
,
1350 .write_bitmap
= write_bitmap0
,
1351 .free_super
= free_super0
,