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
40 static unsigned long calc_sb0_csum(mdp_super_t
*super
)
42 unsigned long csum
= super
->sb_csum
;
43 unsigned long newcsum
;
45 newcsum
= calc_csum(super
, MD_SB_BYTES
);
46 super
->sb_csum
= csum
;
50 static void super0_swap_endian(struct mdp_superblock_s
*sb
)
52 /* as super0 superblocks are host-endian, it is sometimes
53 * useful to be able to swap the endianness
54 * as (almost) everything is u32's we byte-swap every 4byte
56 * We then also have to swap the events_hi and events_lo
58 char *sbc
= (char *)sb
;
62 for (i
=0; i
< MD_SB_BYTES
; i
+=4) {
71 sb
->events_hi
= sb
->events_lo
;
74 t32
= sb
->cp_events_hi
;
75 sb
->cp_events_hi
= sb
->cp_events_lo
;
76 sb
->cp_events_lo
= t32
;
82 static void examine_super0(struct supertype
*st
, char *homehost
)
84 mdp_super_t
*sb
= st
->sb
;
90 printf(" Magic : %08x\n", sb
->md_magic
);
91 printf(" Version : %d.%02d.%02d\n", sb
->major_version
, sb
->minor_version
,
93 if (sb
->minor_version
>= 90) {
94 printf(" UUID : %08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
95 sb
->set_uuid2
, sb
->set_uuid3
);
98 void *hash
= sha1_buffer(homehost
,
101 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
102 printf(" (local to host %s)", homehost
);
106 printf(" UUID : %08x\n", sb
->set_uuid0
);
108 if (sb
->not_persistent
)
109 printf(" Eedk : not persistent\n");
112 printf(" Creation Time : %.24s\n", ctime(&atime
));
113 c
=map_num(pers
, sb
->level
);
114 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
115 if ((int)sb
->level
> 0) {
116 int ddsks
= 0, ddsks_denom
= 1;
117 printf(" Used Dev Size : %d%s\n", sb
->size
,
118 human_size((long long)sb
->size
<<10));
120 case 1: ddsks
=1;break;
122 case 5: ddsks
= sb
->raid_disks
-1; break;
123 case 6: ddsks
= sb
->raid_disks
-2; break;
124 case 10: ddsks
= sb
->raid_disks
;
125 ddsks_denom
= (sb
->layout
&255) * ((sb
->layout
>>8)&255);
128 long long asize
= sb
->size
;
129 asize
= (asize
<< 10) * ddsks
/ ddsks_denom
;
130 printf(" Array Size : %llu%s\n",
131 asize
>> 10, human_size(asize
));
134 printf(" Raid Devices : %d\n", sb
->raid_disks
);
135 printf(" Total Devices : %d\n", sb
->nr_disks
);
136 printf("Preferred Minor : %d\n", sb
->md_minor
);
138 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
139 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb
->reshape_position
/2, human_size((long long)sb
->reshape_position
<<9));
140 if (sb
->delta_disks
) {
141 printf(" Delta Devices : %d", sb
->delta_disks
);
142 printf(" (%d->%d)\n", sb
->raid_disks
-sb
->delta_disks
, sb
->raid_disks
);
143 if (((int)sb
->delta_disks
) < 0)
144 delta_extra
= - sb
->delta_disks
;
146 if (sb
->new_level
!= sb
->level
) {
147 c
= map_num(pers
, sb
->new_level
);
148 printf(" New Level : %s\n", c
?c
:"-unknown-");
150 if (sb
->new_layout
!= sb
->layout
) {
151 if (sb
->level
== 5) {
152 c
= map_num(r5layout
, sb
->new_layout
);
153 printf(" New Layout : %s\n", c
?c
:"-unknown-");
155 if (sb
->level
== 6) {
156 c
= map_num(r6layout
, sb
->new_layout
);
157 printf(" New Layout : %s\n", c
?c
:"-unknown-");
159 if (sb
->level
== 10) {
160 printf(" New Layout : near=%d, %s=%d\n",
162 (sb
->new_layout
&0x10000)?"offset":"far",
163 (sb
->new_layout
>>8)&255);
166 if (sb
->new_chunk
!= sb
->chunk_size
)
167 printf(" New Chunksize : %d\n", sb
->new_chunk
);
171 printf(" Update Time : %.24s\n", ctime(&atime
));
172 printf(" State : %s\n",
173 (sb
->state
&(1<<MD_SB_CLEAN
))?"clean":"active");
174 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
))
175 printf("Internal Bitmap : present\n");
176 printf(" Active Devices : %d\n", sb
->active_disks
);
177 printf("Working Devices : %d\n", sb
->working_disks
);
178 printf(" Failed Devices : %d\n", sb
->failed_disks
);
179 printf(" Spare Devices : %d\n", sb
->spare_disks
);
180 if (calc_sb0_csum(sb
) == sb
->sb_csum
)
181 printf(" Checksum : %x - correct\n", sb
->sb_csum
);
183 printf(" Checksum : %x - expected %lx\n", sb
->sb_csum
, calc_sb0_csum(sb
));
184 printf(" Events : %llu\n",
185 ((unsigned long long)sb
->events_hi
<< 32)
188 if (sb
->level
== 5) {
189 c
= map_num(r5layout
, sb
->layout
);
190 printf(" Layout : %s\n", c
?c
:"-unknown-");
192 if (sb
->level
== 6) {
193 c
= map_num(r6layout
, sb
->layout
);
194 printf(" Layout : %s\n", c
?c
:"-unknown-");
196 if (sb
->level
== 10) {
198 print_r10_layout(sb
->layout
);
207 printf(" Chunk Size : %dK\n", sb
->chunk_size
/1024);
210 printf(" Rounding : %dK\n", sb
->chunk_size
/1024);
215 printf(" Number Major Minor RaidDevice State\n");
216 for (d
= -1; d
<(signed int)(sb
->raid_disks
+delta_extra
+ sb
->spare_disks
); d
++) {
221 if (d
>=0) dp
= &sb
->disks
[d
];
222 else dp
= &sb
->this_disk
;
223 snprintf(nb
, sizeof(nb
), "%4d", d
);
224 printf("%4s %5d %5d %5d %5d ", d
< 0 ? "this" : nb
,
225 dp
->number
, dp
->major
, dp
->minor
, dp
->raid_disk
);
226 wonly
= dp
->state
& (1<<MD_DISK_WRITEMOSTLY
);
227 dp
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
228 if (dp
->state
& (1<<MD_DISK_FAULTY
)) printf(" faulty");
229 if (dp
->state
& (1<<MD_DISK_ACTIVE
)) printf(" active");
230 if (dp
->state
& (1<<MD_DISK_SYNC
)) printf(" sync");
231 if (dp
->state
& (1<<MD_DISK_REMOVED
)) printf(" removed");
232 if (wonly
) printf(" write-mostly");
233 if (dp
->state
== 0) printf(" spare");
234 if ((dv
=map_dev(dp
->major
, dp
->minor
, 0)))
237 if (d
== -1) printf("\n");
241 static void brief_examine_super0(struct supertype
*st
, int verbose
)
243 mdp_super_t
*sb
= st
->sb
;
244 char *c
=map_num(pers
, sb
->level
);
247 sprintf(devname
, "/dev/md%d", sb
->md_minor
);
250 printf("ARRAY %s level=%s num-devices=%d",
252 c
?c
:"-unknown-", sb
->raid_disks
);
254 printf("ARRAY %s", devname
);
256 if (sb
->minor_version
>= 90)
257 printf(" UUID=%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
258 sb
->set_uuid2
, sb
->set_uuid3
);
260 printf(" UUID=%08x", sb
->set_uuid0
);
264 static void export_examine_super0(struct supertype
*st
)
266 mdp_super_t
*sb
= st
->sb
;
268 printf("MD_LEVEL=%s\n", map_num(pers
, sb
->level
));
269 printf("MD_DEVICES=%d\n", sb
->raid_disks
);
270 if (sb
->minor_version
>= 90)
271 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
272 sb
->set_uuid0
, sb
->set_uuid1
,
273 sb
->set_uuid2
, sb
->set_uuid3
);
275 printf("MD_UUID=%08x\n", sb
->set_uuid0
);
276 printf("MD_UPDATE_TIME=%llu\n",
277 __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
);
278 printf("MD_EVENTS=%llu\n",
279 ((unsigned long long)sb
->events_hi
<< 32)
283 static int copy_metadata0(struct supertype
*st
, int from
, int to
)
285 /* Read 64K from the appropriate offset of 'from'
286 * and if it looks a little like a 0.90 superblock,
287 * write it to the same offset of 'to'
290 unsigned long long dsize
, offset
;
291 const int bufsize
= 64*1024;
294 if (posix_memalign(&buf
, 4096, bufsize
) != 0)
297 if (!get_dev_size(from
, NULL
, &dsize
))
300 if (dsize
< MD_RESERVED_SECTORS
*512)
303 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
307 if (lseek64(from
, offset
, 0) < 0LL)
309 if (read(from
, buf
, bufsize
) != bufsize
)
312 if (lseek64(to
, offset
, 0) < 0LL)
315 if (super
->md_magic
!= MD_SB_MAGIC
||
316 super
->major_version
!= 0 ||
317 calc_sb0_csum(super
) != super
->sb_csum
)
319 if (write(to
, buf
, bufsize
) != bufsize
)
328 static void detail_super0(struct supertype
*st
, char *homehost
)
330 mdp_super_t
*sb
= st
->sb
;
332 if (sb
->minor_version
>= 90)
333 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
334 sb
->set_uuid2
, sb
->set_uuid3
);
336 printf("%08x", sb
->set_uuid0
);
339 void *hash
= sha1_buffer(homehost
,
342 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
343 printf(" (local to host %s)", homehost
);
345 printf("\n Events : %d.%d\n\n", sb
->events_hi
, sb
->events_lo
);
348 static void brief_detail_super0(struct supertype
*st
)
350 mdp_super_t
*sb
= st
->sb
;
352 if (sb
->minor_version
>= 90)
353 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
354 sb
->set_uuid2
, sb
->set_uuid3
);
356 printf("%08x", sb
->set_uuid0
);
360 static int match_home0(struct supertype
*st
, char *homehost
)
362 mdp_super_t
*sb
= st
->sb
;
368 hash
= sha1_buffer(homehost
,
372 return (memcmp(&sb
->set_uuid2
, hash
, 8)==0);
375 static void uuid_from_super0(struct supertype
*st
, int uuid
[4])
377 mdp_super_t
*super
= st
->sb
;
378 uuid
[0] = super
->set_uuid0
;
379 if (super
->minor_version
>= 90) {
380 uuid
[1] = super
->set_uuid1
;
381 uuid
[2] = super
->set_uuid2
;
382 uuid
[3] = super
->set_uuid3
;
390 static void getinfo_super0(struct supertype
*st
, struct mdinfo
*info
, char *map
)
392 mdp_super_t
*sb
= st
->sb
;
395 int map_disks
= info
->array
.raid_disks
;
397 memset(info
, 0, sizeof(*info
));
398 info
->array
.major_version
= sb
->major_version
;
399 info
->array
.minor_version
= sb
->minor_version
;
400 info
->array
.patch_version
= sb
->patch_version
;
401 info
->array
.raid_disks
= sb
->raid_disks
;
402 info
->array
.level
= sb
->level
;
403 info
->array
.layout
= sb
->layout
;
404 info
->array
.md_minor
= sb
->md_minor
;
405 info
->array
.ctime
= sb
->ctime
;
406 info
->array
.utime
= sb
->utime
;
407 info
->array
.chunk_size
= sb
->chunk_size
;
408 info
->array
.state
= sb
->state
;
409 info
->component_size
= sb
->size
*2;
411 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
))
412 info
->bitmap_offset
= 8;
414 info
->disk
.state
= sb
->this_disk
.state
;
415 info
->disk
.major
= sb
->this_disk
.major
;
416 info
->disk
.minor
= sb
->this_disk
.minor
;
417 info
->disk
.raid_disk
= sb
->this_disk
.raid_disk
;
418 info
->disk
.number
= sb
->this_disk
.number
;
420 info
->events
= md_event(sb
);
421 info
->data_offset
= 0;
423 sprintf(info
->text_version
, "0.%d", sb
->minor_version
);
424 info
->safe_mode_delay
= 200;
426 uuid_from_super0(st
, info
->uuid
);
428 info
->recovery_start
= MaxSector
;
429 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
430 info
->reshape_active
= 1;
431 info
->reshape_progress
= sb
->reshape_position
;
432 info
->new_level
= sb
->new_level
;
433 info
->delta_disks
= sb
->delta_disks
;
434 info
->new_layout
= sb
->new_layout
;
435 info
->new_chunk
= sb
->new_chunk
;
436 if (info
->delta_disks
< 0)
437 info
->array
.raid_disks
-= info
->delta_disks
;
439 info
->reshape_active
= 0;
441 info
->recovery_blocked
= info
->reshape_active
;
443 sprintf(info
->name
, "%d", sb
->md_minor
);
444 /* work_disks is calculated rather than read directly */
445 for (i
=0; i
< MD_SB_DISKS
; i
++)
446 if ((sb
->disks
[i
].state
& (1<<MD_DISK_SYNC
)) &&
447 (sb
->disks
[i
].raid_disk
< (unsigned)info
->array
.raid_disks
) &&
448 (sb
->disks
[i
].state
& (1<<MD_DISK_ACTIVE
)) &&
449 !(sb
->disks
[i
].state
& (1<<MD_DISK_FAULTY
))) {
451 if (map
&& i
< map_disks
)
453 } else if (map
&& i
< map_disks
)
455 info
->array
.working_disks
= working
;
458 static struct mdinfo
*container_content0(struct supertype
*st
, char *subarray
)
465 info
= xmalloc(sizeof(*info
));
466 getinfo_super0(st
, info
, NULL
);
470 static int update_super0(struct supertype
*st
, struct mdinfo
*info
,
472 char *devname
, int verbose
,
473 int uuid_set
, char *homehost
)
475 /* NOTE: for 'assemble' and 'force' we need to return non-zero
476 * if any change was made. For others, the return value is
481 mdp_super_t
*sb
= st
->sb
;
483 if (strcmp(update
, "homehost") == 0 &&
485 /* note that 'homehost' is special as it is really
490 info
->uuid
[0] = sb
->set_uuid0
;
491 info
->uuid
[1] = sb
->set_uuid1
;
494 if (strcmp(update
, "sparc2.2")==0 ) {
495 /* 2.2 sparc put the events in the wrong place
496 * So we copy the tail of the superblock
497 * up 4 bytes before continuing
499 __u32
*sb32
= (__u32
*)sb
;
500 memcpy(sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7,
501 sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7+1,
502 (MD_SB_WORDS
- (MD_SB_GENERIC_CONSTANT_WORDS
+7+1))*4);
504 pr_err("adjusting superblock of %s for 2.2/sparc compatibility.\n",
506 } else if (strcmp(update
, "super-minor") ==0) {
507 sb
->md_minor
= info
->array
.md_minor
;
509 pr_err("updating superblock of %s with minor number %d\n",
510 devname
, info
->array
.md_minor
);
511 } else if (strcmp(update
, "summaries") == 0) {
513 /* set nr_disks, active_disks, working_disks,
514 * failed_disks, spare_disks based on disks[]
515 * array in superblock.
516 * Also make sure extra slots aren't 'failed'
518 sb
->nr_disks
= sb
->active_disks
=
519 sb
->working_disks
= sb
->failed_disks
=
521 for (i
=0; i
< MD_SB_DISKS
; i
++)
522 if (sb
->disks
[i
].major
||
523 sb
->disks
[i
].minor
) {
524 int state
= sb
->disks
[i
].state
;
525 if (state
& (1<<MD_DISK_REMOVED
))
528 if (state
& (1<<MD_DISK_ACTIVE
))
530 if (state
& (1<<MD_DISK_FAULTY
))
536 } else if (i
>= sb
->raid_disks
&& sb
->disks
[i
].number
== 0)
537 sb
->disks
[i
].state
= 0;
538 } else if (strcmp(update
, "force-one")==0) {
539 /* Not enough devices for a working array, so
540 * bring this one up-to-date.
542 __u32 ehi
= sb
->events_hi
, elo
= sb
->events_lo
;
543 sb
->events_hi
= (info
->events
>>32) & 0xFFFFFFFF;
544 sb
->events_lo
= (info
->events
) & 0xFFFFFFFF;
545 if (sb
->events_hi
!= ehi
||
546 sb
->events_lo
!= elo
)
548 } else if (strcmp(update
, "force-array")==0) {
549 /* degraded array and 'force' requested, so
550 * maybe need to mark it 'clean'
552 if ((sb
->level
== 5 || sb
->level
== 4 || sb
->level
== 6) &&
553 (sb
->state
& (1 << MD_SB_CLEAN
)) == 0) {
554 /* need to force clean */
555 sb
->state
|= (1 << MD_SB_CLEAN
);
558 } else if (strcmp(update
, "assemble")==0) {
559 int d
= info
->disk
.number
;
560 int wonly
= sb
->disks
[d
].state
& (1<<MD_DISK_WRITEMOSTLY
);
561 int mask
= (1<<MD_DISK_WRITEMOSTLY
);
563 if (sb
->minor_version
>= 91)
564 /* During reshape we don't insist on everything
565 * being marked 'sync'
567 add
= (1<<MD_DISK_SYNC
);
568 if (((sb
->disks
[d
].state
& ~mask
) | add
)
569 != (unsigned)info
->disk
.state
) {
570 sb
->disks
[d
].state
= info
->disk
.state
| wonly
;
573 if (info
->reshape_active
&&
574 sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0 &&
575 info
->delta_disks
>= 0 &&
576 info
->reshape_progress
< sb
->reshape_position
) {
577 sb
->reshape_position
= info
->reshape_progress
;
580 if (info
->reshape_active
&&
581 sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0 &&
582 info
->delta_disks
< 0 &&
583 info
->reshape_progress
> sb
->reshape_position
) {
584 sb
->reshape_position
= info
->reshape_progress
;
587 } else if (strcmp(update
, "linear-grow-new") == 0) {
588 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
589 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
590 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
591 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
592 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
593 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
594 sb
->this_disk
= sb
->disks
[info
->disk
.number
];
595 } else if (strcmp(update
, "linear-grow-update") == 0) {
596 sb
->raid_disks
= info
->array
.raid_disks
;
597 sb
->nr_disks
= info
->array
.nr_disks
;
598 sb
->active_disks
= info
->array
.active_disks
;
599 sb
->working_disks
= info
->array
.working_disks
;
600 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
601 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
602 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
603 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
604 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
605 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
606 } else if (strcmp(update
, "resync") == 0) {
607 /* make sure resync happens */
608 sb
->state
&= ~(1<<MD_SB_CLEAN
);
610 } else if (strcmp(update
, "uuid") == 0) {
611 if (!uuid_set
&& homehost
) {
613 char *hash
= sha1_buffer(homehost
,
616 memcpy(info
->uuid
+2, hash
, 8);
618 sb
->set_uuid0
= info
->uuid
[0];
619 sb
->set_uuid1
= info
->uuid
[1];
620 sb
->set_uuid2
= info
->uuid
[2];
621 sb
->set_uuid3
= info
->uuid
[3];
622 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
623 struct bitmap_super_s
*bm
;
624 bm
= (struct bitmap_super_s
*)(sb
+1);
625 uuid_from_super0(st
, uuid
);
626 memcpy(bm
->uuid
, uuid
, 16);
628 } else if (strcmp(update
, "metadata") == 0) {
629 /* Create some v1.0 metadata to match ours but make the
630 * ctime bigger. Also update info->array.*_version.
631 * We need to arrange that store_super writes out
633 * Not permitted for unclean array, or array with
636 if (info
->bitmap_offset
) {
637 pr_err("Cannot update metadata when bitmap is present\n");
639 } else if (info
->array
.state
!= 1) {
640 pr_err("Cannot update metadata on unclean array\n");
643 info
->array
.major_version
= 1;
644 info
->array
.minor_version
= 0;
645 uuid_from_super0(st
, info
->uuid
);
646 st
->other
= super1_make_v0(st
, info
, st
->sb
);
648 } else if (strcmp(update
, "revert-reshape") == 0) {
650 if (sb
->minor_version
<= 90)
651 pr_err("No active reshape to revert on %s\n",
653 else if (sb
->delta_disks
== 0)
654 pr_err("%s: Can on revert reshape which changes number of devices\n",
659 sb
->raid_disks
-= sb
->delta_disks
;
660 sb
->delta_disks
= -sb
->delta_disks
;
662 tmp
= sb
->new_layout
;
663 sb
->new_layout
= sb
->layout
;
667 sb
->new_chunk
= sb
->chunk_size
;
668 sb
->chunk_size
= tmp
;
670 } else if (strcmp(update
, "no-bitmap") == 0) {
671 sb
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
672 } else if (strcmp(update
, "_reshape_progress")==0)
673 sb
->reshape_position
= info
->reshape_progress
;
674 else if (strcmp(update
, "writemostly")==0)
675 sb
->state
|= (1<<MD_DISK_WRITEMOSTLY
);
676 else if (strcmp(update
, "readwrite")==0)
677 sb
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
681 sb
->sb_csum
= calc_sb0_csum(sb
);
686 * For verion-0 superblock, the homehost is 'stored' in the
687 * uuid. 8 bytes for a hash of the host leaving 8 bytes
688 * of random material.
689 * We use the first 8 bytes (64bits) of the sha1 of the
694 static int init_super0(struct supertype
*st
, mdu_array_info_t
*info
,
695 unsigned long long size
, char *ignored_name
, char *homehost
,
696 int *uuid
, unsigned long long data_offset
)
701 if (data_offset
!= INVALID_SECTORS
) {
702 pr_err("data-offset not support for 0.90\n");
706 if (posix_memalign((void**)&sb
, 4096,
707 MD_SB_BYTES
+ ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
708 pr_err("%s could not allocate superblock\n", __func__
);
711 memset(sb
, 0, MD_SB_BYTES
+ sizeof(bitmap_super_t
));
715 /* zeroing the superblock */
719 spares
= info
->working_disks
- info
->active_disks
;
720 if (info
->raid_disks
+ spares
> MD_SB_DISKS
) {
721 pr_err("too many devices requested: %d+%d > %d\n",
722 info
->raid_disks
, spares
, MD_SB_DISKS
);
726 sb
->md_magic
= MD_SB_MAGIC
;
727 sb
->major_version
= 0;
728 sb
->minor_version
= 90;
729 sb
->patch_version
= 0;
730 sb
->gvalid_words
= 0; /* ignored */
732 sb
->level
= info
->level
;
734 if (size
!= (unsigned long long)sb
->size
)
736 sb
->nr_disks
= info
->nr_disks
;
737 sb
->raid_disks
= info
->raid_disks
;
738 sb
->md_minor
= info
->md_minor
;
739 sb
->not_persistent
= 0;
741 sb
->set_uuid0
= uuid
[0];
742 sb
->set_uuid1
= uuid
[1];
743 sb
->set_uuid2
= uuid
[2];
744 sb
->set_uuid3
= uuid
[3];
746 int rfd
= open("/dev/urandom", O_RDONLY
);
747 if (rfd
< 0 || read(rfd
, &sb
->set_uuid0
, 4) != 4)
748 sb
->set_uuid0
= random();
749 if (rfd
< 0 || read(rfd
, &sb
->set_uuid1
, 12) != 12) {
750 sb
->set_uuid1
= random();
751 sb
->set_uuid2
= random();
752 sb
->set_uuid3
= random();
757 if (homehost
&& !uuid
) {
759 char *hash
= sha1_buffer(homehost
,
762 memcpy(&sb
->set_uuid2
, hash
, 8);
765 sb
->utime
= sb
->ctime
;
766 sb
->state
= info
->state
;
767 sb
->active_disks
= info
->active_disks
;
768 sb
->working_disks
= info
->working_disks
;
769 sb
->failed_disks
= info
->failed_disks
;
770 sb
->spare_disks
= info
->spare_disks
;
774 sb
->layout
= info
->layout
;
775 sb
->chunk_size
= info
->chunk_size
;
783 mdu_disk_info_t disk
;
784 struct devinfo
*next
;
788 /* Add a device to the superblock being created */
789 static int add_to_super0(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
790 int fd
, char *devname
, unsigned long long data_offset
)
792 mdp_super_t
*sb
= st
->sb
;
793 mdp_disk_t
*dk
= &sb
->disks
[dinfo
->number
];
794 struct devinfo
*di
, **dip
;
796 dk
->number
= dinfo
->number
;
797 dk
->major
= dinfo
->major
;
798 dk
->minor
= dinfo
->minor
;
799 dk
->raid_disk
= dinfo
->raid_disk
;
800 dk
->state
= dinfo
->state
;
801 /* In case our source disk was writemostly, don't copy that bit */
802 dk
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
804 sb
->this_disk
= sb
->disks
[dinfo
->number
];
805 sb
->sb_csum
= calc_sb0_csum(sb
);
807 dip
= (struct devinfo
**)&st
->info
;
810 di
= xmalloc(sizeof(struct devinfo
));
812 di
->devname
= devname
;
821 static int store_super0(struct supertype
*st
, int fd
)
823 unsigned long long dsize
;
824 unsigned long long offset
;
825 mdp_super_t
*super
= st
->sb
;
827 if (!get_dev_size(fd
, NULL
, &dsize
))
830 if (dsize
< MD_RESERVED_SECTORS
*512)
834 /* Writing out v1.0 metadata for --update=metadata */
837 offset
= dsize
/512 - 8*2;
840 if (lseek64(fd
, offset
, 0)< 0LL)
842 else if (write(fd
, st
->other
, 1024) != 1024)
851 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
855 if (lseek64(fd
, offset
, 0)< 0LL)
858 if (write(fd
, super
, sizeof(*super
)) != sizeof(*super
))
861 if (super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
862 struct bitmap_super_s
* bm
= (struct bitmap_super_s
*)(super
+1);
863 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
)
864 if (write(fd
, bm
, ROUND_UP(sizeof(*bm
),4096)) !=
865 ROUND_UP(sizeof(*bm
),4096))
874 static int write_init_super0(struct supertype
*st
)
876 mdp_super_t
*sb
= st
->sb
;
880 for (di
= st
->info
; di
&& ! rv
; di
= di
->next
) {
882 if (di
->disk
.state
== 1)
886 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
889 sb
->disks
[di
->disk
.number
].state
&= ~(1<<MD_DISK_FAULTY
);
891 sb
->this_disk
= sb
->disks
[di
->disk
.number
];
892 sb
->sb_csum
= calc_sb0_csum(sb
);
893 rv
= store_super0(st
, di
->fd
);
895 if (rv
== 0 && (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)))
896 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
899 pr_err("failed to write superblock to %s\n",
906 static int compare_super0(struct supertype
*st
, struct supertype
*tst
)
910 * 0 same, or first was empty, and second was copied
911 * 1 second had wrong number
915 mdp_super_t
*first
= st
->sb
;
916 mdp_super_t
*second
= tst
->sb
;
917 int uuid1
[4], uuid2
[4];
919 if (second
->md_magic
!= MD_SB_MAGIC
)
922 if (posix_memalign((void**)&first
, 4096,
924 ROUND_UP(sizeof(struct bitmap_super_s
), 4096)) != 0) {
925 pr_err("%s could not allocate superblock\n", __func__
);
928 memcpy(first
, second
, MD_SB_BYTES
+ sizeof(struct bitmap_super_s
));
933 uuid_from_super0(st
, uuid1
);
934 uuid_from_super0(tst
, uuid2
);
935 if (!same_uuid(uuid1
, uuid2
, 0))
937 if (first
->major_version
!= second
->major_version
||
938 first
->minor_version
!= second
->minor_version
||
939 first
->patch_version
!= second
->patch_version
||
940 first
->gvalid_words
!= second
->gvalid_words
||
941 first
->ctime
!= second
->ctime
||
942 first
->level
!= second
->level
||
943 first
->size
!= second
->size
||
944 first
->raid_disks
!= second
->raid_disks
)
951 static void free_super0(struct supertype
*st
);
953 static int load_super0(struct supertype
*st
, int fd
, char *devname
)
955 /* try to read in the superblock
958 * 1 on cannot get superblock
959 * 2 on superblock meaningless
961 unsigned long long dsize
;
962 unsigned long long offset
;
965 struct bitmap_super_s
*bsb
;
969 if (!get_dev_size(fd
, devname
, &dsize
))
972 if (dsize
< MD_RESERVED_SECTORS
*512) {
974 pr_err("%s is too small for md: size is %llu sectors.\n",
980 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
984 if (lseek64(fd
, offset
, 0)< 0LL) {
986 pr_err("Cannot seek to superblock on %s: %s\n",
987 devname
, strerror(errno
));
991 if (posix_memalign((void**)&super
, 4096,
993 ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
994 pr_err("%s could not allocate superblock\n", __func__
);
998 if (read(fd
, super
, sizeof(*super
)) != MD_SB_BYTES
) {
1000 pr_err("Cannot read superblock on %s\n",
1006 if (st
->ss
&& st
->minor_version
== 9)
1007 super0_swap_endian(super
);
1009 if (super
->md_magic
!= MD_SB_MAGIC
) {
1011 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1012 devname
, MD_SB_MAGIC
, super
->md_magic
);
1017 if (super
->major_version
!= 0) {
1019 pr_err("Cannot interpret superblock on %s - version is %d\n",
1020 devname
, super
->major_version
);
1026 if (st
->ss
== NULL
) {
1028 st
->minor_version
= super
->minor_version
;
1029 st
->max_devs
= MD_SB_DISKS
;
1033 /* Now check on the bitmap superblock */
1034 if ((super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) == 0)
1036 /* Read the bitmap superblock and make sure it looks
1037 * valid. If it doesn't clear the bit. An --assemble --force
1038 * should get that written out.
1040 if (read(fd
, super
+1, ROUND_UP(sizeof(struct bitmap_super_s
),4096))
1041 != ROUND_UP(sizeof(struct bitmap_super_s
),4096))
1044 uuid_from_super0(st
, uuid
);
1045 bsb
= (struct bitmap_super_s
*)(super
+1);
1046 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1047 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1052 super
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
1057 static struct supertype
*match_metadata_desc0(char *arg
)
1059 struct supertype
*st
= xcalloc(1, sizeof(*st
));
1061 st
->container_devnm
[0] = 0;
1064 st
->minor_version
= 90;
1065 st
->max_devs
= MD_SB_DISKS
;
1067 /* we sometimes get 00.90 */
1068 while (arg
[0] == '0' && arg
[1] == '0')
1070 if (strcmp(arg
, "0") == 0 ||
1071 #ifdef DEFAULT_OLD_METADATA /* ifndef in super1.c */
1072 strcmp(arg
, "default") == 0 ||
1073 #endif /* DEFAULT_OLD_METADATA */
1074 strcmp(arg
, "0.90") == 0 ||
1075 strcmp(arg
, "") == 0 /* no metadata - i.e. non_persistent */
1079 st
->minor_version
= 91; /* reshape in progress */
1080 if (strcmp(arg
, "0.91") == 0) /* For dup_super support */
1083 st
->minor_version
= 9; /* flag for 'byte-swapped' */
1084 if (strcmp(arg
, "0.swap")==0 ||
1085 strcmp(arg
, "0.9") == 0) /* For dup_super support */
1092 static __u64
avail_size0(struct supertype
*st
, __u64 devsize
,
1093 unsigned long long data_offset
)
1095 if (data_offset
!= 0 && data_offset
!= INVALID_SECTORS
)
1097 if (devsize
< MD_RESERVED_SECTORS
)
1099 return MD_NEW_SIZE_SECTORS(devsize
);
1102 static int add_internal_bitmap0(struct supertype
*st
, int *chunkp
,
1103 int delay
, int write_behind
,
1104 unsigned long long size
, int may_change
,
1108 * The bitmap comes immediately after the superblock and must be 60K in size
1109 * at most. The default size is between 30K and 60K
1111 * size is in sectors, chunk is in bytes !!!
1113 unsigned long long bits
;
1114 unsigned long long max_bits
= (60*1024 - sizeof(bitmap_super_t
))*8;
1115 unsigned long long min_chunk
;
1116 int chunk
= *chunkp
;
1117 mdp_super_t
*sb
= st
->sb
;
1118 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MD_SB_BYTES
);
1122 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1123 bits
= (size
* 512) / min_chunk
+ 1;
1124 while (bits
> max_bits
) {
1128 if (chunk
== UnSet
) {
1129 /* A chunk size less than a few Megabytes gives poor
1130 * performance without increasing resync noticeably
1133 if (chunk
< 64*1024*1024)
1134 chunk
= 64*1024*1024;
1135 } else if ((unsigned long long)chunk
< min_chunk
)
1136 return 0; /* chunk size too small */
1138 sb
->state
|= (1<<MD_SB_BITMAP_PRESENT
);
1140 memset(bms
, 0, sizeof(*bms
));
1141 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1142 bms
->version
= __cpu_to_le32(major
);
1143 uuid_from_super0(st
, uuid
);
1144 memcpy(bms
->uuid
, uuid
, 16);
1145 bms
->chunksize
= __cpu_to_le32(chunk
);
1146 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1147 bms
->sync_size
= __cpu_to_le64(size
);
1148 bms
->write_behind
= __cpu_to_le32(write_behind
);
1154 static void locate_bitmap0(struct supertype
*st
, int fd
)
1156 unsigned long long dsize
;
1157 unsigned long long offset
;
1159 if (!get_dev_size(fd
, NULL
, &dsize
))
1162 if (dsize
< MD_RESERVED_SECTORS
*512)
1165 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1169 offset
+= MD_SB_BYTES
;
1171 lseek64(fd
, offset
, 0);
1174 static int write_bitmap0(struct supertype
*st
, int fd
)
1176 unsigned long long dsize
;
1177 unsigned long long offset
;
1178 mdp_super_t
*sb
= st
->sb
;
1185 if (!get_dev_size(fd
, NULL
, &dsize
))
1188 if (dsize
< MD_RESERVED_SECTORS
*512)
1191 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1195 if (lseek64(fd
, offset
+ 4096, 0)< 0LL)
1198 if (posix_memalign(&buf
, 4096, 4096))
1201 memset(buf
, 0xff, 4096);
1202 memcpy(buf
, ((char*)sb
)+MD_SB_BYTES
, sizeof(bitmap_super_t
));
1204 while (towrite
> 0) {
1208 n
= write(fd
, buf
, n
);
1213 memset(buf
, 0xff, 4096);
1223 static void free_super0(struct supertype
*st
)
1228 struct devinfo
*di
= st
->info
;
1229 st
->info
= di
->next
;
1238 static int validate_geometry0(struct supertype
*st
, int level
,
1239 int layout
, int raiddisks
,
1240 int *chunk
, unsigned long long size
,
1241 unsigned long long data_offset
,
1242 char *subdev
, unsigned long long *freesize
,
1245 unsigned long long ldsize
;
1247 unsigned int tbmax
= 4;
1249 /* prior to linux 3.1, a but limits usable device size to 2TB.
1250 * It was introduced in 2.6.29, but we won't worry about that detail
1252 if (get_linux_version() < 3001000)
1255 if (level
== LEVEL_CONTAINER
) {
1257 pr_err("0.90 metadata does not support containers\n");
1260 if (raiddisks
> MD_SB_DISKS
) {
1262 pr_err("0.90 metadata supports at most %d devices per array\n",
1266 if (size
>= tbmax
* 2ULL*1024*1024*1024) {
1268 pr_err("0.90 metadata supports at most "
1269 "%d terabytes per device\n", tbmax
);
1272 if (chunk
&& *chunk
== UnSet
)
1273 *chunk
= DEFAULT_CHUNK
;
1278 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1281 pr_err("super0.90 cannot open %s: %s\n",
1282 subdev
, strerror(errno
));
1286 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1292 if (ldsize
< MD_RESERVED_SECTORS
* 512)
1294 *freesize
= MD_NEW_SIZE_SECTORS(ldsize
>> 9);
1297 #endif /* MDASSEMBLE */
1299 struct superswitch super0
= {
1301 .examine_super
= examine_super0
,
1302 .brief_examine_super
= brief_examine_super0
,
1303 .export_examine_super
= export_examine_super0
,
1304 .detail_super
= detail_super0
,
1305 .brief_detail_super
= brief_detail_super0
,
1306 .write_init_super
= write_init_super0
,
1307 .validate_geometry
= validate_geometry0
,
1308 .add_to_super
= add_to_super0
,
1309 .copy_metadata
= copy_metadata0
,
1311 .match_home
= match_home0
,
1312 .uuid_from_super
= uuid_from_super0
,
1313 .getinfo_super
= getinfo_super0
,
1314 .container_content
= container_content0
,
1315 .update_super
= update_super0
,
1316 .init_super
= init_super0
,
1317 .store_super
= store_super0
,
1318 .compare_super
= compare_super0
,
1319 .load_super
= load_super0
,
1320 .match_metadata_desc
= match_metadata_desc0
,
1321 .avail_size
= avail_size0
,
1322 .add_internal_bitmap
= add_internal_bitmap0
,
1323 .locate_bitmap
= locate_bitmap0
,
1324 .write_bitmap
= write_bitmap0
,
1325 .free_super
= free_super0
,