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
;
51 static void super0_swap_endian(struct mdp_superblock_s
*sb
)
53 /* as super0 superblocks are host-endian, it is sometimes
54 * useful to be able to swap the endianness
55 * as (almost) everything is u32's we byte-swap every 4byte
57 * We then also have to swap the events_hi and events_lo
59 char *sbc
= (char *)sb
;
63 for (i
=0; i
< MD_SB_BYTES
; i
+=4) {
72 sb
->events_hi
= sb
->events_lo
;
75 t32
= sb
->cp_events_hi
;
76 sb
->cp_events_hi
= sb
->cp_events_lo
;
77 sb
->cp_events_lo
= t32
;
83 static void examine_super0(struct supertype
*st
, char *homehost
)
85 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) {
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
/ (sb
->layout
&255) / ((sb
->layout
>>8)&255);
127 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks
* sb
->size
,
128 human_size(ddsks
*(long long)sb
->size
<<10));
130 printf(" Raid Devices : %d\n", sb
->raid_disks
);
131 printf(" Total Devices : %d\n", sb
->nr_disks
);
132 printf("Preferred Minor : %d\n", sb
->md_minor
);
134 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
135 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb
->reshape_position
/2, human_size((long long)sb
->reshape_position
<<9));
136 if (sb
->delta_disks
) {
137 printf(" Delta Devices : %d", sb
->delta_disks
);
139 printf(" (%d->%d)\n", sb
->raid_disks
-sb
->delta_disks
, sb
->raid_disks
);
141 printf(" (%d->%d)\n", sb
->raid_disks
, sb
->raid_disks
+sb
->delta_disks
);
143 if (sb
->new_level
!= sb
->level
) {
144 c
= map_num(pers
, sb
->new_level
);
145 printf(" New Level : %s\n", c
?c
:"-unknown-");
147 if (sb
->new_layout
!= sb
->layout
) {
148 if (sb
->level
== 5) {
149 c
= map_num(r5layout
, sb
->new_layout
);
150 printf(" New Layout : %s\n", c
?c
:"-unknown-");
152 if (sb
->level
== 10) {
153 printf(" New Layout : near=%d, %s=%d\n",
155 (sb
->new_layout
&0x10000)?"offset":"far",
156 (sb
->new_layout
>>8)&255);
159 if (sb
->new_chunk
!= sb
->chunk_size
)
160 printf(" New Chunksize : %d\n", sb
->new_chunk
);
164 printf(" Update Time : %.24s\n", ctime(&atime
));
165 printf(" State : %s\n",
166 (sb
->state
&(1<<MD_SB_CLEAN
))?"clean":"active");
167 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
))
168 printf("Internal Bitmap : present\n");
169 printf(" Active Devices : %d\n", sb
->active_disks
);
170 printf("Working Devices : %d\n", sb
->working_disks
);
171 printf(" Failed Devices : %d\n", sb
->failed_disks
);
172 printf(" Spare Devices : %d\n", sb
->spare_disks
);
173 if (calc_sb0_csum(sb
) == sb
->sb_csum
)
174 printf(" Checksum : %x - correct\n", sb
->sb_csum
);
176 printf(" Checksum : %x - expected %lx\n", sb
->sb_csum
, calc_sb0_csum(sb
));
177 printf(" Events : %llu\n",
178 ((unsigned long long)sb
->events_hi
<< 32)
181 if (sb
->level
== 5) {
182 c
= map_num(r5layout
, sb
->layout
);
183 printf(" Layout : %s\n", c
?c
:"-unknown-");
185 if (sb
->level
== 10) {
187 print_r10_layout(sb
->layout
);
196 printf(" Chunk Size : %dK\n", sb
->chunk_size
/1024);
199 printf(" Rounding : %dK\n", sb
->chunk_size
/1024);
204 printf(" Number Major Minor RaidDevice State\n");
205 for (d
= -1; d
<(signed int)(sb
->raid_disks
+sb
->spare_disks
); d
++) {
210 if (d
>=0) dp
= &sb
->disks
[d
];
211 else dp
= &sb
->this_disk
;
212 snprintf(nb
, sizeof(nb
), "%4d", d
);
213 printf("%4s %5d %5d %5d %5d ", d
< 0 ? "this" : nb
,
214 dp
->number
, dp
->major
, dp
->minor
, dp
->raid_disk
);
215 wonly
= dp
->state
& (1<<MD_DISK_WRITEMOSTLY
);
216 dp
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
217 if (dp
->state
& (1<<MD_DISK_FAULTY
)) printf(" faulty");
218 if (dp
->state
& (1<<MD_DISK_ACTIVE
)) printf(" active");
219 if (dp
->state
& (1<<MD_DISK_SYNC
)) printf(" sync");
220 if (dp
->state
& (1<<MD_DISK_REMOVED
)) printf(" removed");
221 if (wonly
) printf(" write-mostly");
222 if (dp
->state
== 0) printf(" spare");
223 if ((dv
=map_dev(dp
->major
, dp
->minor
, 0)))
226 if (d
== -1) printf("\n");
230 static void brief_examine_super0(struct supertype
*st
, int verbose
)
232 mdp_super_t
*sb
= st
->sb
;
233 char *c
=map_num(pers
, sb
->level
);
236 sprintf(devname
, "/dev/md%d", sb
->md_minor
);
239 printf("ARRAY %s level=%s num-devices=%d",
241 c
?c
:"-unknown-", sb
->raid_disks
);
243 printf("ARRAY %s", devname
);
245 if (sb
->minor_version
>= 90)
246 printf(" UUID=%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
247 sb
->set_uuid2
, sb
->set_uuid3
);
249 printf(" UUID=%08x", sb
->set_uuid0
);
253 static void export_examine_super0(struct supertype
*st
)
255 mdp_super_t
*sb
= st
->sb
;
257 printf("MD_LEVEL=%s\n", map_num(pers
, sb
->level
));
258 printf("MD_DEVICES=%d\n", sb
->raid_disks
);
259 if (sb
->minor_version
>= 90)
260 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
261 sb
->set_uuid0
, sb
->set_uuid1
,
262 sb
->set_uuid2
, sb
->set_uuid3
);
264 printf("MD_UUID=%08x\n", sb
->set_uuid0
);
265 printf("MD_UPDATE_TIME=%llu\n",
266 __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
);
267 printf("MD_EVENTS=%llu\n",
268 ((unsigned long long)sb
->events_hi
<< 32)
272 static void detail_super0(struct supertype
*st
, char *homehost
)
274 mdp_super_t
*sb
= st
->sb
;
276 if (sb
->minor_version
>= 90)
277 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
278 sb
->set_uuid2
, sb
->set_uuid3
);
280 printf("%08x", sb
->set_uuid0
);
283 void *hash
= sha1_buffer(homehost
,
286 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
287 printf(" (local to host %s)", homehost
);
289 printf("\n Events : %d.%d\n\n", sb
->events_hi
, sb
->events_lo
);
292 static void brief_detail_super0(struct supertype
*st
)
294 mdp_super_t
*sb
= st
->sb
;
296 if (sb
->minor_version
>= 90)
297 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
298 sb
->set_uuid2
, sb
->set_uuid3
);
300 printf("%08x", sb
->set_uuid0
);
304 static int match_home0(struct supertype
*st
, char *homehost
)
306 mdp_super_t
*sb
= st
->sb
;
312 hash
= sha1_buffer(homehost
,
316 return (memcmp(&sb
->set_uuid2
, hash
, 8)==0);
319 static void uuid_from_super0(struct supertype
*st
, int uuid
[4])
321 mdp_super_t
*super
= st
->sb
;
322 uuid
[0] = super
->set_uuid0
;
323 if (super
->minor_version
>= 90) {
324 uuid
[1] = super
->set_uuid1
;
325 uuid
[2] = super
->set_uuid2
;
326 uuid
[3] = super
->set_uuid3
;
334 static void getinfo_super0(struct supertype
*st
, struct mdinfo
*info
)
336 mdp_super_t
*sb
= st
->sb
;
340 info
->array
.major_version
= sb
->major_version
;
341 info
->array
.minor_version
= sb
->minor_version
;
342 info
->array
.patch_version
= sb
->patch_version
;
343 info
->array
.raid_disks
= sb
->raid_disks
;
344 info
->array
.level
= sb
->level
;
345 info
->array
.layout
= sb
->layout
;
346 info
->array
.md_minor
= sb
->md_minor
;
347 info
->array
.ctime
= sb
->ctime
;
348 info
->array
.utime
= sb
->utime
;
349 info
->array
.chunk_size
= sb
->chunk_size
;
350 info
->array
.state
= sb
->state
;
351 info
->component_size
= sb
->size
*2;
353 info
->disk
.state
= sb
->this_disk
.state
;
354 info
->disk
.major
= sb
->this_disk
.major
;
355 info
->disk
.minor
= sb
->this_disk
.minor
;
356 info
->disk
.raid_disk
= sb
->this_disk
.raid_disk
;
357 info
->disk
.number
= sb
->this_disk
.number
;
359 info
->events
= md_event(sb
);
360 info
->data_offset
= 0;
362 sprintf(info
->text_version
, "0.%d", sb
->minor_version
);
363 info
->safe_mode_delay
= 200;
365 uuid_from_super0(st
, info
->uuid
);
367 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
368 info
->reshape_active
= 1;
369 info
->reshape_progress
= sb
->reshape_position
;
370 info
->new_level
= sb
->new_level
;
371 info
->delta_disks
= sb
->delta_disks
;
372 info
->new_layout
= sb
->new_layout
;
373 info
->new_chunk
= sb
->new_chunk
;
375 info
->reshape_active
= 0;
377 sprintf(info
->name
, "%d", sb
->md_minor
);
378 /* work_disks is calculated rather than read directly */
379 for (i
=0; i
< MD_SB_DISKS
; i
++)
380 if ((sb
->disks
[i
].state
& (1<<MD_DISK_SYNC
)) &&
381 (sb
->disks
[i
].raid_disk
< info
->array
.raid_disks
) &&
382 (sb
->disks
[i
].state
& (1<<MD_DISK_ACTIVE
)) &&
383 !(sb
->disks
[i
].state
& (1<<MD_DISK_FAULTY
)))
385 info
->array
.working_disks
= working
;
389 static int update_super0(struct supertype
*st
, struct mdinfo
*info
,
391 char *devname
, int verbose
,
392 int uuid_set
, char *homehost
)
394 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
395 * For others, the return value is ignored.
398 mdp_super_t
*sb
= st
->sb
;
399 if (strcmp(update
, "sparc2.2")==0 ) {
400 /* 2.2 sparc put the events in the wrong place
401 * So we copy the tail of the superblock
402 * up 4 bytes before continuing
404 __u32
*sb32
= (__u32
*)sb
;
405 memcpy(sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7,
406 sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7+1,
407 (MD_SB_WORDS
- (MD_SB_GENERIC_CONSTANT_WORDS
+7+1))*4);
409 fprintf (stderr
, Name
": adjusting superblock of %s for 2.2/sparc compatability.\n",
412 if (strcmp(update
, "super-minor") ==0) {
413 sb
->md_minor
= info
->array
.md_minor
;
415 fprintf(stderr
, Name
": updating superblock of %s with minor number %d\n",
416 devname
, info
->array
.md_minor
);
418 if (strcmp(update
, "summaries") == 0) {
420 /* set nr_disks, active_disks, working_disks,
421 * failed_disks, spare_disks based on disks[]
422 * array in superblock.
423 * Also make sure extra slots aren't 'failed'
425 sb
->nr_disks
= sb
->active_disks
=
426 sb
->working_disks
= sb
->failed_disks
=
428 for (i
=0; i
< MD_SB_DISKS
; i
++)
429 if (sb
->disks
[i
].major
||
430 sb
->disks
[i
].minor
) {
431 int state
= sb
->disks
[i
].state
;
432 if (state
& (1<<MD_DISK_REMOVED
))
435 if (state
& (1<<MD_DISK_ACTIVE
))
437 if (state
& (1<<MD_DISK_FAULTY
))
443 } else if (i
>= sb
->raid_disks
&& sb
->disks
[i
].number
== 0)
444 sb
->disks
[i
].state
= 0;
446 if (strcmp(update
, "force-one")==0) {
447 /* Not enough devices for a working array, so
448 * bring this one up-to-date.
450 __u32 ehi
= sb
->events_hi
, elo
= sb
->events_lo
;
451 sb
->events_hi
= (info
->events
>>32) & 0xFFFFFFFF;
452 sb
->events_lo
= (info
->events
) & 0xFFFFFFFF;
453 if (sb
->events_hi
!= ehi
||
454 sb
->events_lo
!= elo
)
457 if (strcmp(update
, "force-array")==0) {
458 /* degraded array and 'force' requested, so
459 * maybe need to mark it 'clean'
461 if ((sb
->level
== 5 || sb
->level
== 4 || sb
->level
== 6) &&
462 (sb
->state
& (1 << MD_SB_CLEAN
)) == 0) {
463 /* need to force clean */
464 sb
->state
|= (1 << MD_SB_CLEAN
);
468 if (strcmp(update
, "assemble")==0) {
469 int d
= info
->disk
.number
;
470 int wonly
= sb
->disks
[d
].state
& (1<<MD_DISK_WRITEMOSTLY
);
471 if ((sb
->disks
[d
].state
& ~(1<<MD_DISK_WRITEMOSTLY
))
472 != info
->disk
.state
) {
473 sb
->disks
[d
].state
= info
->disk
.state
| wonly
;
477 if (strcmp(update
, "linear-grow-new") == 0) {
478 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
479 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
480 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
481 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
482 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
483 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
484 sb
->this_disk
= sb
->disks
[info
->disk
.number
];
486 if (strcmp(update
, "linear-grow-update") == 0) {
487 sb
->raid_disks
= info
->array
.raid_disks
;
488 sb
->nr_disks
= info
->array
.nr_disks
;
489 sb
->active_disks
= info
->array
.active_disks
;
490 sb
->working_disks
= info
->array
.working_disks
;
491 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
492 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
493 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
494 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
495 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
496 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
498 if (strcmp(update
, "resync") == 0) {
499 /* make sure resync happens */
500 sb
->state
&= ~(1<<MD_SB_CLEAN
);
503 if (strcmp(update
, "homehost") == 0 &&
507 info
->uuid
[0] = sb
->set_uuid0
;
508 info
->uuid
[1] = sb
->set_uuid1
;
510 if (strcmp(update
, "uuid") == 0) {
511 if (!uuid_set
&& homehost
) {
513 char *hash
= sha1_buffer(homehost
,
516 memcpy(info
->uuid
+2, hash
, 8);
518 sb
->set_uuid0
= info
->uuid
[0];
519 sb
->set_uuid1
= info
->uuid
[1];
520 sb
->set_uuid2
= info
->uuid
[2];
521 sb
->set_uuid3
= info
->uuid
[3];
522 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
523 struct bitmap_super_s
*bm
;
524 bm
= (struct bitmap_super_s
*)(sb
+1);
525 uuid_from_super0(st
, (int*)bm
->uuid
);
528 if (strcmp(update
, "_reshape_progress")==0)
529 sb
->reshape_position
= info
->reshape_progress
;
531 sb
->sb_csum
= calc_sb0_csum(sb
);
536 * For verion-0 superblock, the homehost is 'stored' in the
537 * uuid. 8 bytes for a hash of the host leaving 8 bytes
538 * of random material.
539 * We use the first 8 bytes (64bits) of the sha1 of the
544 static int init_super0(struct supertype
*st
, mdu_array_info_t
*info
,
545 unsigned long long size
, char *ignored_name
, char *homehost
,
551 if (posix_memalign((void**)&sb
, 4096,
552 MD_SB_BYTES
+ ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
553 fprintf(stderr
, Name
": %s could not allocate superblock\n", __func__
);
556 memset(sb
, 0, MD_SB_BYTES
+ sizeof(bitmap_super_t
));
560 /* zeroing the superblock */
564 spares
= info
->working_disks
- info
->active_disks
;
565 if (info
->raid_disks
+ spares
> MD_SB_DISKS
) {
566 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
567 info
->raid_disks
, spares
, MD_SB_DISKS
);
571 sb
->md_magic
= MD_SB_MAGIC
;
572 sb
->major_version
= 0;
573 sb
->minor_version
= 90;
574 sb
->patch_version
= 0;
575 sb
->gvalid_words
= 0; /* ignored */
577 sb
->level
= info
->level
;
578 if (size
!= info
->size
)
580 sb
->size
= info
->size
;
581 sb
->nr_disks
= info
->nr_disks
;
582 sb
->raid_disks
= info
->raid_disks
;
583 sb
->md_minor
= info
->md_minor
;
584 sb
->not_persistent
= 0;
586 sb
->set_uuid0
= uuid
[0];
587 sb
->set_uuid1
= uuid
[1];
588 sb
->set_uuid2
= uuid
[2];
589 sb
->set_uuid3
= uuid
[3];
591 int rfd
= open("/dev/urandom", O_RDONLY
);
592 if (rfd
< 0 || read(rfd
, &sb
->set_uuid0
, 4) != 4)
593 sb
->set_uuid0
= random();
594 if (rfd
< 0 || read(rfd
, &sb
->set_uuid1
, 12) != 12) {
595 sb
->set_uuid1
= random();
596 sb
->set_uuid2
= random();
597 sb
->set_uuid3
= random();
604 char *hash
= sha1_buffer(homehost
,
607 memcpy(&sb
->set_uuid2
, hash
, 8);
610 sb
->utime
= sb
->ctime
;
611 sb
->state
= info
->state
;
612 sb
->active_disks
= info
->active_disks
;
613 sb
->working_disks
= info
->working_disks
;
614 sb
->failed_disks
= info
->failed_disks
;
615 sb
->spare_disks
= info
->spare_disks
;
619 sb
->layout
= info
->layout
;
620 sb
->chunk_size
= info
->chunk_size
;
628 mdu_disk_info_t disk
;
629 struct devinfo
*next
;
633 /* Add a device to the superblock being created */
634 static int add_to_super0(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
635 int fd
, char *devname
)
637 mdp_super_t
*sb
= st
->sb
;
638 mdp_disk_t
*dk
= &sb
->disks
[dinfo
->number
];
639 struct devinfo
*di
, **dip
;
641 dk
->number
= dinfo
->number
;
642 dk
->major
= dinfo
->major
;
643 dk
->minor
= dinfo
->minor
;
644 dk
->raid_disk
= dinfo
->raid_disk
;
645 dk
->state
= dinfo
->state
;
647 sb
->this_disk
= sb
->disks
[dinfo
->number
];
648 sb
->sb_csum
= calc_sb0_csum(sb
);
650 dip
= (struct devinfo
**)&st
->info
;
653 di
= malloc(sizeof(struct devinfo
));
655 di
->devname
= devname
;
664 static int store_super0(struct supertype
*st
, int fd
)
666 unsigned long long dsize
;
667 unsigned long long offset
;
668 mdp_super_t
*super
= st
->sb
;
670 if (!get_dev_size(fd
, NULL
, &dsize
))
673 if (dsize
< MD_RESERVED_SECTORS
*512)
676 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
680 if (lseek64(fd
, offset
, 0)< 0LL)
683 if (write(fd
, super
, sizeof(*super
)) != sizeof(*super
))
686 if (super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
687 struct bitmap_super_s
* bm
= (struct bitmap_super_s
*)(super
+1);
688 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
)
689 if (write(fd
, bm
, ROUND_UP(sizeof(*bm
),4096)) !=
690 ROUND_UP(sizeof(*bm
),4096))
699 static int write_init_super0(struct supertype
*st
)
701 mdp_super_t
*sb
= st
->sb
;
705 for (di
= st
->info
; di
&& ! rv
; di
= di
->next
) {
707 if (di
->disk
.state
== 1)
711 Kill(di
->devname
, 0, 1, 1);
712 Kill(di
->devname
, 0, 1, 1);
714 sb
->disks
[di
->disk
.number
].state
&= ~(1<<MD_DISK_FAULTY
);
716 sb
->this_disk
= sb
->disks
[di
->disk
.number
];
717 sb
->sb_csum
= calc_sb0_csum(sb
);
718 rv
= store_super0(st
, di
->fd
);
720 if (rv
== 0 && (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)))
721 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
725 Name
": failed to write superblock to %s\n",
734 static int compare_super0(struct supertype
*st
, struct supertype
*tst
)
738 * 0 same, or first was empty, and second was copied
739 * 1 second had wrong number
743 mdp_super_t
*first
= st
->sb
;
744 mdp_super_t
*second
= tst
->sb
;
745 int uuid1
[4], uuid2
[4];
747 if (second
->md_magic
!= MD_SB_MAGIC
)
750 if (posix_memalign((void**)&first
, 4096,
752 ROUND_UP(sizeof(struct bitmap_super_s
), 4096)) != 0) {
754 ": %s could not allocate superblock\n", __func__
);
757 memcpy(first
, second
, MD_SB_BYTES
+ sizeof(struct bitmap_super_s
));
762 uuid_from_super0(st
, uuid1
);
763 uuid_from_super0(tst
, uuid2
);
764 if (!same_uuid(uuid1
, uuid2
, 0))
766 if (first
->major_version
!= second
->major_version
||
767 first
->minor_version
!= second
->minor_version
||
768 first
->patch_version
!= second
->patch_version
||
769 first
->gvalid_words
!= second
->gvalid_words
||
770 first
->ctime
!= second
->ctime
||
771 first
->level
!= second
->level
||
772 first
->size
!= second
->size
||
773 first
->raid_disks
!= second
->raid_disks
)
780 static void free_super0(struct supertype
*st
);
782 static int load_super0(struct supertype
*st
, int fd
, char *devname
)
784 /* try to read in the superblock
787 * 1 on cannot get superblock
788 * 2 on superblock meaningless
790 unsigned long long dsize
;
791 unsigned long long offset
;
794 struct bitmap_super_s
*bsb
;
801 if (!get_dev_size(fd
, devname
, &dsize
))
804 if (dsize
< MD_RESERVED_SECTORS
*512) {
807 ": %s is too small for md: size is %llu sectors.\n",
812 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
816 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
818 if (lseek64(fd
, offset
, 0)< 0LL) {
820 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
821 devname
, strerror(errno
));
825 if (posix_memalign((void**)&super
, 4096,
827 ROUND_UP(sizeof(bitmap_super_t
), 4096)) != 0) {
829 ": %s could not allocate superblock\n", __func__
);
833 if (read(fd
, super
, sizeof(*super
)) != MD_SB_BYTES
) {
835 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
841 if (st
->ss
&& st
->minor_version
== 9)
842 super0_swap_endian(super
);
844 if (super
->md_magic
!= MD_SB_MAGIC
) {
846 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
847 devname
, MD_SB_MAGIC
, super
->md_magic
);
852 if (super
->major_version
!= 0) {
854 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
855 devname
, super
->major_version
);
861 if (st
->ss
== NULL
) {
863 st
->minor_version
= super
->minor_version
;
864 st
->max_devs
= MD_SB_DISKS
;
868 /* Now check on the bitmap superblock */
869 if ((super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) == 0)
871 /* Read the bitmap superblock and make sure it looks
872 * valid. If it doesn't clear the bit. An --assemble --force
873 * should get that written out.
875 if (read(fd
, super
+1, ROUND_UP(sizeof(struct bitmap_super_s
),4096))
876 != ROUND_UP(sizeof(struct bitmap_super_s
),4096))
879 uuid_from_super0(st
, uuid
);
880 bsb
= (struct bitmap_super_s
*)(super
+1);
881 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
882 memcmp(bsb
->uuid
, uuid
, 16) != 0)
887 super
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
892 static struct supertype
*match_metadata_desc0(char *arg
)
894 struct supertype
*st
= malloc(sizeof(*st
));
897 memset(st
, 0, sizeof(*st
));
900 st
->minor_version
= 90;
901 st
->max_devs
= MD_SB_DISKS
;
903 /* we sometimes get 00.90 */
904 while (arg
[0] == '0' && arg
[1] == '0')
906 if (strcmp(arg
, "0") == 0 ||
907 strcmp(arg
, "0.90") == 0 ||
908 strcmp(arg
, "default") == 0 ||
909 strcmp(arg
, "") == 0 /* no metadata */
913 st
->minor_version
= 91; /* reshape in progress */
914 if (strcmp(arg
, "0.91") == 0) /* For dup_super support */
917 st
->minor_version
= 9; /* flag for 'byte-swapped' */
918 if (strcmp(arg
, "0.swap")==0 ||
919 strcmp(arg
, "0.9") == 0) /* For dup_super support */
926 static __u64
avail_size0(struct supertype
*st
, __u64 devsize
)
928 if (devsize
< MD_RESERVED_SECTORS
)
930 return MD_NEW_SIZE_SECTORS(devsize
);
933 static int add_internal_bitmap0(struct supertype
*st
, int *chunkp
,
934 int delay
, int write_behind
,
935 unsigned long long size
, int may_change
,
939 * The bitmap comes immediately after the superblock and must be 60K in size
940 * at most. The default size is between 30K and 60K
942 * size is in sectors, chunk is in bytes !!!
944 unsigned long long bits
;
945 unsigned long long max_bits
= 60*1024*8;
946 unsigned long long min_chunk
;
948 mdp_super_t
*sb
= st
->sb
;
949 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MD_SB_BYTES
);
952 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
953 bits
= (size
* 512) / min_chunk
+ 1;
954 while (bits
> max_bits
) {
960 else if (chunk
< min_chunk
)
961 return 0; /* chunk size too small */
963 sb
->state
|= (1<<MD_SB_BITMAP_PRESENT
);
965 memset(bms
, 0, sizeof(*bms
));
966 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
967 bms
->version
= __cpu_to_le32(major
);
968 uuid_from_super0(st
, (int*)bms
->uuid
);
969 bms
->chunksize
= __cpu_to_le32(chunk
);
970 bms
->daemon_sleep
= __cpu_to_le32(delay
);
971 bms
->sync_size
= __cpu_to_le64(size
);
972 bms
->write_behind
= __cpu_to_le32(write_behind
);
978 static void locate_bitmap0(struct supertype
*st
, int fd
)
980 unsigned long long dsize
;
981 unsigned long long offset
;
983 if (!get_dev_size(fd
, NULL
, &dsize
))
986 if (dsize
< MD_RESERVED_SECTORS
*512)
989 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
993 offset
+= MD_SB_BYTES
;
995 lseek64(fd
, offset
, 0);
998 static int write_bitmap0(struct supertype
*st
, int fd
)
1000 unsigned long long dsize
;
1001 unsigned long long offset
;
1002 mdp_super_t
*sb
= st
->sb
;
1007 char abuf
[4096+4096];
1008 char *buf
= (char*)(((long)(abuf
+4096))&~4095L);
1010 if (!get_dev_size(fd
, NULL
, &dsize
))
1014 if (dsize
< MD_RESERVED_SECTORS
*512)
1017 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
1021 if (lseek64(fd
, offset
+ 4096, 0)< 0LL)
1024 memset(buf
, 0xff, 4096);
1025 memcpy(buf
, ((char*)sb
)+MD_SB_BYTES
, sizeof(bitmap_super_t
));
1027 while (towrite
> 0) {
1031 n
= write(fd
, buf
, n
);
1036 memset(buf
, 0xff, 4096);
1045 static void free_super0(struct supertype
*st
)
1053 static int validate_geometry0(struct supertype
*st
, int level
,
1054 int layout
, int raiddisks
,
1055 int chunk
, unsigned long long size
,
1056 char *subdev
, unsigned long long *freesize
,
1059 unsigned long long ldsize
;
1062 if (level
== LEVEL_CONTAINER
)
1064 if (raiddisks
> MD_SB_DISKS
)
1066 if (size
> (0x7fffffffULL
<<9))
1071 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1074 fprintf(stderr
, Name
": super0.90 cannot open %s: %s\n",
1075 subdev
, strerror(errno
));
1079 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1085 if (ldsize
< MD_RESERVED_SECTORS
* 512)
1087 if (size
> (0x7fffffffULL
<<9))
1089 *freesize
= MD_NEW_SIZE_SECTORS(ldsize
>> 9);
1092 #endif /* MDASSEMBLE */
1094 struct superswitch super0
= {
1096 .examine_super
= examine_super0
,
1097 .brief_examine_super
= brief_examine_super0
,
1098 .export_examine_super
= export_examine_super0
,
1099 .detail_super
= detail_super0
,
1100 .brief_detail_super
= brief_detail_super0
,
1101 .write_init_super
= write_init_super0
,
1102 .validate_geometry
= validate_geometry0
,
1103 .add_to_super
= add_to_super0
,
1105 .match_home
= match_home0
,
1106 .uuid_from_super
= uuid_from_super0
,
1107 .getinfo_super
= getinfo_super0
,
1108 .update_super
= update_super0
,
1109 .init_super
= init_super0
,
1110 .store_super
= store_super0
,
1111 .compare_super
= compare_super0
,
1112 .load_super
= load_super0
,
1113 .match_metadata_desc
= match_metadata_desc0
,
1114 .avail_size
= avail_size0
,
1115 .add_internal_bitmap
= add_internal_bitmap0
,
1116 .locate_bitmap
= locate_bitmap0
,
1117 .write_bitmap
= write_bitmap0
,
1118 .free_super
= free_super0
,