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
30 #define HAVE_STDINT_H 1
34 * All handling for the 0.90.0 version superblock is in
37 * - finding, loading, and writing the superblock.
38 * - initialising a new superblock
39 * - printing the superblock for --examine
40 * - printing part of the superblock for --detail
45 static unsigned long calc_sb0_csum(mdp_super_t
*super
)
47 unsigned long csum
= super
->sb_csum
;
48 unsigned long newcsum
;
50 newcsum
= calc_csum(super
, MD_SB_BYTES
);
51 super
->sb_csum
= csum
;
56 void super0_swap_endian(struct mdp_superblock_s
*sb
)
58 /* as super0 superblocks are host-endian, it is sometimes
59 * useful to be able to swap the endianness
60 * as (almost) everything is u32's we byte-swap every 4byte
62 * We then also have to swap the events_hi and events_lo
64 char *sbc
= (char *)sb
;
68 for (i
=0; i
< MD_SB_BYTES
; i
+=4) {
77 sb
->events_hi
= sb
->events_lo
;
80 t32
= sb
->cp_events_hi
;
81 sb
->cp_events_hi
= sb
->cp_events_lo
;
82 sb
->cp_events_lo
= t32
;
88 static void examine_super0(struct supertype
*st
, char *homehost
)
90 mdp_super_t
*sb
= st
->sb
;
95 printf(" Magic : %08x\n", sb
->md_magic
);
96 printf(" Version : %02d.%02d.%02d\n", sb
->major_version
, sb
->minor_version
,
98 if (sb
->minor_version
>= 90) {
99 printf(" UUID : %08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
100 sb
->set_uuid2
, sb
->set_uuid3
);
103 void *hash
= sha1_buffer(homehost
,
106 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
107 printf(" (local to host %s)", homehost
);
111 printf(" UUID : %08x\n", sb
->set_uuid0
);
113 if (sb
->not_persistent
)
114 printf(" Eedk : not persistent\n");
117 printf(" Creation Time : %.24s\n", ctime(&atime
));
118 c
=map_num(pers
, sb
->level
);
119 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
120 if ((int)sb
->level
>= 0) {
122 printf(" Used Dev Size : %d%s\n", sb
->size
,
123 human_size((long long)sb
->size
<<10));
125 case 1: ddsks
=1;break;
127 case 5: ddsks
= sb
->raid_disks
-1; break;
128 case 6: ddsks
= sb
->raid_disks
-2; break;
129 case 10: ddsks
= sb
->raid_disks
/ (sb
->layout
&255) / ((sb
->layout
>>8)&255);
132 printf(" Array Size : %llu%s\n", (unsigned long long)ddsks
* sb
->size
,
133 human_size(ddsks
*(long long)sb
->size
<<10));
135 printf(" Raid Devices : %d\n", sb
->raid_disks
);
136 printf(" Total Devices : %d\n", sb
->nr_disks
);
137 printf("Preferred Minor : %d\n", sb
->md_minor
);
139 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
140 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)sb
->reshape_position
/2, human_size((long long)sb
->reshape_position
<<9));
141 if (sb
->delta_disks
) {
142 printf(" Delta Devices : %d", sb
->delta_disks
);
144 printf(" (%d->%d)\n", sb
->raid_disks
-sb
->delta_disks
, sb
->raid_disks
);
146 printf(" (%d->%d)\n", sb
->raid_disks
, sb
->raid_disks
+sb
->delta_disks
);
148 if (sb
->new_level
!= sb
->level
) {
149 c
= map_num(pers
, sb
->new_level
);
150 printf(" New Level : %s\n", c
?c
:"-unknown-");
152 if (sb
->new_layout
!= sb
->layout
) {
153 if (sb
->level
== 5) {
154 c
= map_num(r5layout
, sb
->new_layout
);
155 printf(" New Layout : %s\n", c
?c
:"-unknown-");
157 if (sb
->level
== 10) {
158 printf(" New Layout : near=%d, %s=%d\n",
160 (sb
->new_layout
&0x10000)?"offset":"far",
161 (sb
->new_layout
>>8)&255);
164 if (sb
->new_chunk
!= sb
->chunk_size
)
165 printf(" New Chunksize : %d\n", sb
->new_chunk
);
169 printf(" Update Time : %.24s\n", ctime(&atime
));
170 printf(" State : %s\n",
171 (sb
->state
&(1<<MD_SB_CLEAN
))?"clean":"active");
172 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
))
173 printf("Internal Bitmap : present\n");
174 printf(" Active Devices : %d\n", sb
->active_disks
);
175 printf("Working Devices : %d\n", sb
->working_disks
);
176 printf(" Failed Devices : %d\n", sb
->failed_disks
);
177 printf(" Spare Devices : %d\n", sb
->spare_disks
);
178 if (calc_sb0_csum(sb
) == sb
->sb_csum
)
179 printf(" Checksum : %x - correct\n", sb
->sb_csum
);
181 printf(" Checksum : %x - expected %lx\n", sb
->sb_csum
, calc_sb0_csum(sb
));
182 printf(" Events : %llu\n",
183 ((unsigned long long)sb
->events_hi
<< 32)
186 if (sb
->level
== 5) {
187 c
= map_num(r5layout
, sb
->layout
);
188 printf(" Layout : %s\n", c
?c
:"-unknown-");
190 if (sb
->level
== 10) {
191 printf(" Layout : near=%d, %s=%d\n",
193 (sb
->layout
&0x10000)?"offset":"far",
194 (sb
->layout
>>8)&255);
202 printf(" Chunk Size : %dK\n", sb
->chunk_size
/1024);
205 printf(" Rounding : %dK\n", sb
->chunk_size
/1024);
210 printf(" Number Major Minor RaidDevice State\n");
211 for (d
= -1; d
<(signed int)(sb
->raid_disks
+sb
->spare_disks
); d
++) {
216 if (d
>=0) dp
= &sb
->disks
[d
];
217 else dp
= &sb
->this_disk
;
218 snprintf(nb
, sizeof(nb
), "%4d", d
);
219 printf("%4s %5d %5d %5d %5d ", d
< 0 ? "this" : nb
,
220 dp
->number
, dp
->major
, dp
->minor
, dp
->raid_disk
);
221 wonly
= dp
->state
& (1<<MD_DISK_WRITEMOSTLY
);
222 dp
->state
&= ~(1<<MD_DISK_WRITEMOSTLY
);
223 if (dp
->state
& (1<<MD_DISK_FAULTY
)) printf(" faulty");
224 if (dp
->state
& (1<<MD_DISK_ACTIVE
)) printf(" active");
225 if (dp
->state
& (1<<MD_DISK_SYNC
)) printf(" sync");
226 if (dp
->state
& (1<<MD_DISK_REMOVED
)) printf(" removed");
227 if (wonly
) printf(" write-mostly");
228 if (dp
->state
== 0) printf(" spare");
229 if ((dv
=map_dev(dp
->major
, dp
->minor
, 0)))
232 if (d
== -1) printf("\n");
236 static void brief_examine_super0(struct supertype
*st
)
238 mdp_super_t
*sb
= st
->sb
;
239 char *c
=map_num(pers
, sb
->level
);
242 sprintf(devname
, "/dev/md%d", sb
->md_minor
);
244 printf("ARRAY %s level=%s num-devices=%d UUID=",
246 c
?c
:"-unknown-", sb
->raid_disks
);
247 if (sb
->minor_version
>= 90)
248 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
249 sb
->set_uuid2
, sb
->set_uuid3
);
251 printf("%08x", sb
->set_uuid0
);
255 static void export_examine_super0(struct supertype
*st
)
257 mdp_super_t
*sb
= st
->sb
;
259 printf("MD_LEVEL=%s\n", map_num(pers
, sb
->level
));
260 printf("MD_DEVICES=%d\n", sb
->raid_disks
);
261 if (sb
->minor_version
>= 90)
262 printf("MD_UUID=%08x:%08x:%08x:%08x\n",
263 sb
->set_uuid0
, sb
->set_uuid1
,
264 sb
->set_uuid2
, sb
->set_uuid3
);
266 printf("MD_UUID=%08x\n", sb
->set_uuid0
);
267 printf("MD_UPDATE_TIME=%llu\n",
268 __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
);
269 printf("MD_EVENTS=%llu\n",
270 ((unsigned long long)sb
->events_hi
<< 32)
274 static void detail_super0(struct supertype
*st
, char *homehost
)
276 mdp_super_t
*sb
= st
->sb
;
278 if (sb
->minor_version
>= 90)
279 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
280 sb
->set_uuid2
, sb
->set_uuid3
);
282 printf("%08x", sb
->set_uuid0
);
285 void *hash
= sha1_buffer(homehost
,
288 if (memcmp(&sb
->set_uuid2
, hash
, 8)==0)
289 printf(" (local to host %s)", homehost
);
291 printf("\n Events : %d.%d\n\n", sb
->events_hi
, sb
->events_lo
);
294 static void brief_detail_super0(struct supertype
*st
)
296 mdp_super_t
*sb
= st
->sb
;
298 if (sb
->minor_version
>= 90)
299 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
300 sb
->set_uuid2
, sb
->set_uuid3
);
302 printf("%08x", sb
->set_uuid0
);
305 static void export_detail_super0(struct supertype
*st
)
307 mdp_super_t
*sb
= st
->sb
;
309 if (sb
->minor_version
>= 90)
310 printf("%08x:%08x:%08x:%08x", sb
->set_uuid0
, sb
->set_uuid1
,
311 sb
->set_uuid2
, sb
->set_uuid3
);
313 printf("%08x", sb
->set_uuid0
);
318 static int match_home0(struct supertype
*st
, char *homehost
)
320 mdp_super_t
*sb
= st
->sb
;
322 char *hash
= sha1_buffer(homehost
,
326 return (memcmp(&sb
->set_uuid2
, hash
, 8)==0);
329 static void uuid_from_super0(struct supertype
*st
, int uuid
[4])
331 mdp_super_t
*super
= st
->sb
;
332 uuid
[0] = super
->set_uuid0
;
333 if (super
->minor_version
>= 90) {
334 uuid
[1] = super
->set_uuid1
;
335 uuid
[2] = super
->set_uuid2
;
336 uuid
[3] = super
->set_uuid3
;
344 static void getinfo_super0(struct supertype
*st
, struct mdinfo
*info
)
346 mdp_super_t
*sb
= st
->sb
;
350 info
->array
.major_version
= sb
->major_version
;
351 info
->array
.minor_version
= sb
->minor_version
;
352 info
->array
.patch_version
= sb
->patch_version
;
353 info
->array
.raid_disks
= sb
->raid_disks
;
354 info
->array
.level
= sb
->level
;
355 info
->array
.layout
= sb
->layout
;
356 info
->array
.md_minor
= sb
->md_minor
;
357 info
->array
.ctime
= sb
->ctime
;
358 info
->array
.utime
= sb
->utime
;
359 info
->array
.chunk_size
= sb
->chunk_size
;
360 info
->array
.state
= sb
->state
;
361 info
->component_size
= sb
->size
*2;
363 info
->disk
.state
= sb
->this_disk
.state
;
364 info
->disk
.major
= sb
->this_disk
.major
;
365 info
->disk
.minor
= sb
->this_disk
.minor
;
366 info
->disk
.raid_disk
= sb
->this_disk
.raid_disk
;
367 info
->disk
.number
= sb
->this_disk
.number
;
369 info
->events
= md_event(sb
);
370 info
->data_offset
= 0;
372 uuid_from_super0(st
, info
->uuid
);
374 if (sb
->minor_version
> 90 && (sb
->reshape_position
+1) != 0) {
375 info
->reshape_active
= 1;
376 info
->reshape_progress
= sb
->reshape_position
;
377 info
->new_level
= sb
->new_level
;
378 info
->delta_disks
= sb
->delta_disks
;
379 info
->new_layout
= sb
->new_layout
;
380 info
->new_chunk
= sb
->new_chunk
;
382 info
->reshape_active
= 0;
384 sprintf(info
->name
, "%d", sb
->md_minor
);
385 /* work_disks is calculated rather than read directly */
386 for (i
=0; i
< MD_SB_DISKS
; i
++)
387 if ((sb
->disks
[i
].state
& (1<<MD_DISK_SYNC
)) &&
388 (sb
->disks
[i
].raid_disk
< info
->array
.raid_disks
) &&
389 (sb
->disks
[i
].state
& (1<<MD_DISK_ACTIVE
)) &&
390 !(sb
->disks
[i
].state
& (1<<MD_DISK_FAULTY
)))
392 info
->array
.working_disks
= working
;
396 static int update_super0(struct supertype
*st
, struct mdinfo
*info
,
398 char *devname
, int verbose
,
399 int uuid_set
, char *homehost
)
401 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
402 * For others, the return value is ignored.
405 mdp_super_t
*sb
= st
->sb
;
406 if (strcmp(update
, "sparc2.2")==0 ) {
407 /* 2.2 sparc put the events in the wrong place
408 * So we copy the tail of the superblock
409 * up 4 bytes before continuing
411 __u32
*sb32
= (__u32
*)sb
;
412 memcpy(sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7,
413 sb32
+MD_SB_GENERIC_CONSTANT_WORDS
+7+1,
414 (MD_SB_WORDS
- (MD_SB_GENERIC_CONSTANT_WORDS
+7+1))*4);
416 fprintf (stderr
, Name
": adjusting superblock of %s for 2.2/sparc compatability.\n",
419 if (strcmp(update
, "super-minor") ==0) {
420 sb
->md_minor
= info
->array
.md_minor
;
422 fprintf(stderr
, Name
": updating superblock of %s with minor number %d\n",
423 devname
, info
->array
.md_minor
);
425 if (strcmp(update
, "summaries") == 0) {
427 /* set nr_disks, active_disks, working_disks,
428 * failed_disks, spare_disks based on disks[]
429 * array in superblock.
430 * Also make sure extra slots aren't 'failed'
432 sb
->nr_disks
= sb
->active_disks
=
433 sb
->working_disks
= sb
->failed_disks
=
435 for (i
=0; i
< MD_SB_DISKS
; i
++)
436 if (sb
->disks
[i
].major
||
437 sb
->disks
[i
].minor
) {
438 int state
= sb
->disks
[i
].state
;
439 if (state
& (1<<MD_DISK_REMOVED
))
442 if (state
& (1<<MD_DISK_ACTIVE
))
444 if (state
& (1<<MD_DISK_FAULTY
))
450 } else if (i
>= sb
->raid_disks
&& sb
->disks
[i
].number
== 0)
451 sb
->disks
[i
].state
= 0;
453 if (strcmp(update
, "force-one")==0) {
454 /* Not enough devices for a working array, so
455 * bring this one up-to-date.
457 __u32 ehi
= sb
->events_hi
, elo
= sb
->events_lo
;
458 sb
->events_hi
= (info
->events
>>32) & 0xFFFFFFFF;
459 sb
->events_lo
= (info
->events
) & 0xFFFFFFFF;
460 if (sb
->events_hi
!= ehi
||
461 sb
->events_lo
!= elo
)
464 if (strcmp(update
, "force-array")==0) {
465 /* degraded array and 'force' requested, so
466 * maybe need to mark it 'clean'
468 if ((sb
->level
== 5 || sb
->level
== 4 || sb
->level
== 6) &&
469 (sb
->state
& (1 << MD_SB_CLEAN
)) == 0) {
470 /* need to force clean */
471 sb
->state
|= (1 << MD_SB_CLEAN
);
475 if (strcmp(update
, "assemble")==0) {
476 int d
= info
->disk
.number
;
477 int wonly
= sb
->disks
[d
].state
& (1<<MD_DISK_WRITEMOSTLY
);
478 if ((sb
->disks
[d
].state
& ~(1<<MD_DISK_WRITEMOSTLY
))
479 != info
->disk
.state
) {
480 sb
->disks
[d
].state
= info
->disk
.state
| wonly
;
484 if (strcmp(update
, "linear-grow-new") == 0) {
485 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
486 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
487 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
488 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
489 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
490 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
491 sb
->this_disk
= sb
->disks
[info
->disk
.number
];
493 if (strcmp(update
, "linear-grow-update") == 0) {
494 sb
->raid_disks
= info
->array
.raid_disks
;
495 sb
->nr_disks
= info
->array
.nr_disks
;
496 sb
->active_disks
= info
->array
.active_disks
;
497 sb
->working_disks
= info
->array
.working_disks
;
498 memset(&sb
->disks
[info
->disk
.number
], 0, sizeof(sb
->disks
[0]));
499 sb
->disks
[info
->disk
.number
].number
= info
->disk
.number
;
500 sb
->disks
[info
->disk
.number
].major
= info
->disk
.major
;
501 sb
->disks
[info
->disk
.number
].minor
= info
->disk
.minor
;
502 sb
->disks
[info
->disk
.number
].raid_disk
= info
->disk
.raid_disk
;
503 sb
->disks
[info
->disk
.number
].state
= info
->disk
.state
;
505 if (strcmp(update
, "resync") == 0) {
506 /* make sure resync happens */
507 sb
->state
&= ~(1<<MD_SB_CLEAN
);
510 if (strcmp(update
, "homehost") == 0 &&
514 info
->uuid
[0] = sb
->set_uuid0
;
515 info
->uuid
[1] = sb
->set_uuid1
;
517 if (strcmp(update
, "uuid") == 0) {
518 if (!uuid_set
&& homehost
) {
520 char *hash
= sha1_buffer(homehost
,
523 memcpy(info
->uuid
+2, hash
, 8);
525 sb
->set_uuid0
= info
->uuid
[0];
526 sb
->set_uuid1
= info
->uuid
[1];
527 sb
->set_uuid2
= info
->uuid
[2];
528 sb
->set_uuid3
= info
->uuid
[3];
529 if (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
530 struct bitmap_super_s
*bm
;
531 bm
= (struct bitmap_super_s
*)(sb
+1);
532 uuid_from_super0(st
, (int*)bm
->uuid
);
535 if (strcmp(update
, "_reshape_progress")==0)
536 sb
->reshape_position
= info
->reshape_progress
;
538 sb
->sb_csum
= calc_sb0_csum(sb
);
543 * For verion-0 superblock, the homehost is 'stored' in the
544 * uuid. 8 bytes for a hash of the host leaving 8 bytes
545 * of random material.
546 * We use the first 8 bytes (64bits) of the sha1 of the
551 static int init_super0(struct supertype
*st
, mdu_array_info_t
*info
,
552 unsigned long long size
, char *ignored_name
, char *homehost
,
555 mdp_super_t
*sb
= malloc(MD_SB_BYTES
+ sizeof(bitmap_super_t
));
557 memset(sb
, 0, MD_SB_BYTES
+ sizeof(bitmap_super_t
));
560 if (info
->major_version
== -1) {
561 /* zeroing the superblock */
565 spares
= info
->working_disks
- info
->active_disks
;
566 if (info
->raid_disks
+ spares
> MD_SB_DISKS
) {
567 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
568 info
->raid_disks
, spares
, MD_SB_DISKS
);
572 sb
->md_magic
= MD_SB_MAGIC
;
573 sb
->major_version
= 0;
574 sb
->minor_version
= 90;
575 sb
->patch_version
= 0;
576 sb
->gvalid_words
= 0; /* ignored */
578 sb
->level
= info
->level
;
579 if (size
!= info
->size
)
581 sb
->size
= info
->size
;
582 sb
->nr_disks
= info
->nr_disks
;
583 sb
->raid_disks
= info
->raid_disks
;
584 sb
->md_minor
= info
->md_minor
;
585 sb
->not_persistent
= 0;
587 sb
->set_uuid0
= uuid
[0];
588 sb
->set_uuid1
= uuid
[1];
589 sb
->set_uuid2
= uuid
[2];
590 sb
->set_uuid3
= uuid
[3];
592 int rfd
= open("/dev/urandom", O_RDONLY
);
593 if (rfd
< 0 || read(rfd
, &sb
->set_uuid0
, 4) != 4)
594 sb
->set_uuid0
= random();
595 if (rfd
< 0 || read(rfd
, &sb
->set_uuid1
, 12) != 12) {
596 sb
->set_uuid1
= random();
597 sb
->set_uuid2
= random();
598 sb
->set_uuid3
= random();
605 char *hash
= sha1_buffer(homehost
,
608 memcpy(&sb
->set_uuid2
, hash
, 8);
611 sb
->utime
= sb
->ctime
;
612 sb
->state
= info
->state
;
613 sb
->active_disks
= info
->active_disks
;
614 sb
->working_disks
= info
->working_disks
;
615 sb
->failed_disks
= info
->failed_disks
;
616 sb
->spare_disks
= info
->spare_disks
;
620 sb
->layout
= info
->layout
;
621 sb
->chunk_size
= info
->chunk_size
;
629 mdu_disk_info_t disk
;
630 struct devinfo
*next
;
632 /* Add a device to the superblock being created */
633 static void add_to_super0(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
634 int fd
, char *devname
)
636 mdp_super_t
*sb
= st
->sb
;
637 mdp_disk_t
*dk
= &sb
->disks
[dinfo
->number
];
638 struct devinfo
*di
, **dip
;
640 dk
->number
= dinfo
->number
;
641 dk
->major
= dinfo
->major
;
642 dk
->minor
= dinfo
->minor
;
643 dk
->raid_disk
= dinfo
->raid_disk
;
644 dk
->state
= dinfo
->state
;
646 dip
= (struct devinfo
**)&st
->info
;
649 di
= malloc(sizeof(struct devinfo
));
651 di
->devname
= devname
;
657 static int store_super0(struct supertype
*st
, int fd
)
659 unsigned long long dsize
;
660 unsigned long long offset
;
661 mdp_super_t
*super
= st
->sb
;
663 if (!get_dev_size(fd
, NULL
, &dsize
))
666 if (dsize
< MD_RESERVED_SECTORS
*512)
669 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
673 if (lseek64(fd
, offset
, 0)< 0LL)
676 if (write(fd
, super
, sizeof(*super
)) != sizeof(*super
))
679 if (super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) {
680 struct bitmap_super_s
* bm
= (struct bitmap_super_s
*)(super
+1);
681 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
)
682 if (write(fd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
691 static int write_init_super0(struct supertype
*st
)
693 mdp_super_t
*sb
= st
->sb
;
697 for (di
= st
->info
; di
&& ! rv
; di
= di
->next
) {
699 if (di
->disk
.state
== 1)
701 Kill(di
->devname
, 0, 1, 1);
702 Kill(di
->devname
, 0, 1, 1);
704 sb
->disks
[di
->disk
.number
].state
&= ~(1<<MD_DISK_FAULTY
);
706 sb
->this_disk
= sb
->disks
[di
->disk
.number
];
707 sb
->sb_csum
= calc_sb0_csum(sb
);
708 rv
= store_super0(st
, di
->fd
);
710 if (rv
== 0 && (sb
->state
& (1<<MD_SB_BITMAP_PRESENT
)))
711 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
715 Name
": failed to write superblock to %s\n",
724 static int compare_super0(struct supertype
*st
, struct supertype
*tst
)
728 * 0 same, or first was empty, and second was copied
729 * 1 second had wrong number
733 mdp_super_t
*first
= st
->sb
;
734 mdp_super_t
*second
= tst
->sb
;
735 int uuid1
[4], uuid2
[4];
737 if (second
->md_magic
!= MD_SB_MAGIC
)
740 first
= malloc(MD_SB_BYTES
+ sizeof(struct bitmap_super_s
));
741 memcpy(first
, second
, MD_SB_BYTES
+ sizeof(struct bitmap_super_s
));
746 uuid_from_super0(st
, uuid1
);
747 uuid_from_super0(tst
, uuid2
);
748 if (!same_uuid(uuid1
, uuid2
, 0))
750 if (first
->major_version
!= second
->major_version
||
751 first
->minor_version
!= second
->minor_version
||
752 first
->patch_version
!= second
->patch_version
||
753 first
->gvalid_words
!= second
->gvalid_words
||
754 first
->ctime
!= second
->ctime
||
755 first
->level
!= second
->level
||
756 first
->size
!= second
->size
||
757 first
->raid_disks
!= second
->raid_disks
)
764 static void free_super0(struct supertype
*st
);
766 static int load_super0(struct supertype
*st
, int fd
, char *devname
)
768 /* try to read in the superblock
771 * 1 on cannot get superblock
772 * 2 on superblock meaningless
774 unsigned long long dsize
;
775 unsigned long long offset
;
778 struct bitmap_super_s
*bsb
;
782 if (!get_dev_size(fd
, devname
, &dsize
))
785 if (dsize
< MD_RESERVED_SECTORS
*512) {
788 ": %s is too small for md: size is %llu sectors.\n",
793 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
797 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
799 if (lseek64(fd
, offset
, 0)< 0LL) {
801 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
802 devname
, strerror(errno
));
806 super
= malloc(MD_SB_BYTES
+ sizeof(bitmap_super_t
));
808 if (read(fd
, super
, sizeof(*super
)) != MD_SB_BYTES
) {
810 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
816 if (st
->ss
&& st
->minor_version
== 9)
817 super0_swap_endian(super
);
819 if (super
->md_magic
!= MD_SB_MAGIC
) {
821 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
822 devname
, MD_SB_MAGIC
, super
->md_magic
);
827 if (super
->major_version
!= 0) {
829 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
830 devname
, super
->major_version
);
836 if (st
->ss
== NULL
) {
838 st
->minor_version
= super
->minor_version
;
839 st
->max_devs
= MD_SB_DISKS
;
843 /* Now check on the bitmap superblock */
844 if ((super
->state
& (1<<MD_SB_BITMAP_PRESENT
)) == 0)
846 /* Read the bitmap superblock and make sure it looks
847 * valid. If it doesn't clear the bit. An --assemble --force
848 * should get that written out.
850 if (read(fd
, super
+1, sizeof(struct bitmap_super_s
))
851 != sizeof(struct bitmap_super_s
))
854 uuid_from_super0(st
, uuid
);
855 bsb
= (struct bitmap_super_s
*)(super
+1);
856 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
857 memcmp(bsb
->uuid
, uuid
, 16) != 0)
862 super
->state
&= ~(1<<MD_SB_BITMAP_PRESENT
);
867 static struct supertype
*match_metadata_desc0(char *arg
)
869 struct supertype
*st
= malloc(sizeof(*st
));
874 st
->minor_version
= 90;
875 st
->max_devs
= MD_SB_DISKS
;
877 if (strcmp(arg
, "0") == 0 ||
878 strcmp(arg
, "0.90") == 0 ||
879 strcmp(arg
, "0.91") == 0 ||
880 strcmp(arg
, "default") == 0 ||
881 strcmp(arg
, "") == 0 /* no metadata */
885 st
->minor_version
= 9; /* flag for 'byte-swapped' */
886 if (strcmp(arg
, "0.swap")==0 ||
887 strcmp(arg
, "0.9") == 0) /* For dup_super support */
894 static __u64
avail_size0(struct supertype
*st
, __u64 devsize
)
896 if (devsize
< MD_RESERVED_SECTORS
)
898 return MD_NEW_SIZE_SECTORS(devsize
);
901 static int add_internal_bitmap0(struct supertype
*st
, int *chunkp
,
902 int delay
, int write_behind
,
903 unsigned long long size
, int may_change
,
907 * The bitmap comes immediately after the superblock and must be 60K in size
908 * at most. The default size is between 30K and 60K
910 * size is in sectors, chunk is in bytes !!!
912 unsigned long long bits
;
913 unsigned long long max_bits
= 60*1024*8;
914 unsigned long long min_chunk
;
916 mdp_super_t
*sb
= st
->sb
;
917 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MD_SB_BYTES
);
920 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
921 bits
= (size
* 512) / min_chunk
+ 1;
922 while (bits
> max_bits
) {
928 else if (chunk
< min_chunk
)
929 return 0; /* chunk size too small */
931 sb
->state
|= (1<<MD_SB_BITMAP_PRESENT
);
933 memset(bms
, 0, sizeof(*bms
));
934 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
935 bms
->version
= __cpu_to_le32(major
);
936 uuid_from_super0(st
, (int*)bms
->uuid
);
937 bms
->chunksize
= __cpu_to_le32(chunk
);
938 bms
->daemon_sleep
= __cpu_to_le32(delay
);
939 bms
->sync_size
= __cpu_to_le64(size
);
940 bms
->write_behind
= __cpu_to_le32(write_behind
);
946 void locate_bitmap0(struct supertype
*st
, int fd
)
948 unsigned long long dsize
;
949 unsigned long long offset
;
951 if (!get_dev_size(fd
, NULL
, &dsize
))
954 if (dsize
< MD_RESERVED_SECTORS
*512)
957 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
961 offset
+= MD_SB_BYTES
;
963 lseek64(fd
, offset
, 0);
966 int write_bitmap0(struct supertype
*st
, int fd
)
968 unsigned long long dsize
;
969 unsigned long long offset
;
970 mdp_super_t
*sb
= st
->sb
;
977 if (!get_dev_size(fd
, NULL
, &dsize
))
981 if (dsize
< MD_RESERVED_SECTORS
*512)
984 offset
= MD_NEW_SIZE_SECTORS(dsize
>>9);
988 if (lseek64(fd
, offset
+ 4096, 0)< 0LL)
992 if (write(fd
, ((char*)sb
)+MD_SB_BYTES
, sizeof(bitmap_super_t
)) !=
993 sizeof(bitmap_super_t
))
995 towrite
= 64*1024 - MD_SB_BYTES
- sizeof(bitmap_super_t
);
996 memset(buf
, 0xff, sizeof(buf
));
997 while (towrite
> 0) {
1001 n
= write(fd
, buf
, n
);
1014 static void free_super0(struct supertype
*st
)
1021 static int validate_geometry0(struct supertype
*st
, int level
,
1022 int layout
, int raiddisks
,
1023 int chunk
, unsigned long long size
,
1024 char *subdev
, unsigned long long *freesize
)
1026 unsigned long long ldsize
;
1029 if (level
== LEVEL_CONTAINER
)
1031 if (raiddisks
> MD_SB_DISKS
)
1033 if (size
> (0x7fffffffULL
<<10))
1038 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1040 fprintf(stderr
, Name
": Cannot open %s: %s\n",
1041 subdev
, strerror(errno
));
1044 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1050 if (ldsize
< MD_RESERVED_SECTORS
* 512)
1052 if (size
> (0x7fffffffULL
<<10))
1054 *freesize
= MD_NEW_SIZE_SECTORS(ldsize
>> 9);
1058 struct superswitch super0
= {
1060 .examine_super
= examine_super0
,
1061 .brief_examine_super
= brief_examine_super0
,
1062 .export_examine_super
= export_examine_super0
,
1063 .detail_super
= detail_super0
,
1064 .brief_detail_super
= brief_detail_super0
,
1065 .export_detail_super
= export_detail_super0
,
1066 .write_init_super
= write_init_super0
,
1068 .match_home
= match_home0
,
1069 .uuid_from_super
= uuid_from_super0
,
1070 .getinfo_super
= getinfo_super0
,
1071 .update_super
= update_super0
,
1072 .init_super
= init_super0
,
1073 .add_to_super
= add_to_super0
,
1074 .store_super
= store_super0
,
1075 .compare_super
= compare_super0
,
1076 .load_super
= load_super0
,
1077 .match_metadata_desc
= match_metadata_desc0
,
1078 .avail_size
= avail_size0
,
1079 .add_internal_bitmap
= add_internal_bitmap0
,
1080 .locate_bitmap
= locate_bitmap0
,
1081 .write_bitmap
= write_bitmap0
,
1082 .free_super
= free_super0
,
1083 .validate_geometry
= validate_geometry0
,