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
32 * The version-1 superblock :
33 * All numeric fields are little-endian.
35 * total size: 256 bytes plus 2 per device.
36 * 1K allows 384 devices.
38 struct mdp_superblock_1
{
39 /* constant array information - 128 bytes */
40 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
41 __u32 major_version
; /* 1 */
42 __u32 feature_map
; /* 0 for now */
43 __u32 pad0
; /* always set to 0 when writing */
45 __u8 set_uuid
[16]; /* user-space generated. */
46 char set_name
[32]; /* set and interpreted by user-space */
48 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
49 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
50 __u32 layout
; /* only for raid5 currently */
51 __u64 size
; /* used size of component devices, in 512byte sectors */
53 __u32 chunksize
; /* in 512byte sectors */
55 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
56 * NOTE: signed, so bitmap can be before superblock
57 * only meaningful of feature_map[0] is set.
60 /* These are only valid with feature bit '4' */
61 __u32 new_level
; /* new level we are reshaping to */
62 __u64 reshape_position
; /* next address in array-space for reshape */
63 __u32 delta_disks
; /* change in number of raid_disks */
64 __u32 new_layout
; /* new layout */
65 __u32 new_chunk
; /* new chunk size (bytes) */
66 __u8 pad1
[128-124]; /* set to 0 when written */
68 /* constant this-device information - 64 bytes */
69 __u64 data_offset
; /* sector start of data, often 0 */
70 __u64 data_size
; /* sectors in this device that can be used for data */
71 __u64 super_offset
; /* sector start of this superblock */
72 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
73 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
74 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
75 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
76 __u8 devflags
; /* per-device flags. Only one defined...*/
77 #define WriteMostly1 1 /* mask for writemostly flag in above */
78 __u8 pad2
[64-57]; /* set to 0 when writing */
80 /* array state information - 64 bytes */
81 __u64 utime
; /* 40 bits second, 24 btes microseconds */
82 __u64 events
; /* incremented when superblock updated */
83 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
84 __u32 sb_csum
; /* checksum upto devs[max_dev] */
85 __u32 max_dev
; /* size of devs[] array to consider */
86 __u8 pad3
[64-32]; /* set to 0 when writing */
88 /* device state information. Indexed by dev_number.
90 * Note there are no per-device state flags. State information is rolled
91 * into the 'roles' value. If a device is spare or faulty, then it doesn't
92 * have a meaningful role.
94 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
97 struct misc_dev_info
{
101 /* feature_map bits */
102 #define MD_FEATURE_BITMAP_OFFSET 1
103 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
106 #define MD_FEATURE_RESHAPE_ACTIVE 4
108 #define MD_FEATURE_ALL (1|2|4)
111 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
113 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
115 unsigned int disk_csum
, csum
;
116 unsigned long long newcsum
;
117 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
118 unsigned int *isuper
= (unsigned int*)sb
;
121 /* make sure I can count... */
122 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
123 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
124 sizeof(struct mdp_superblock_1
) != 256) {
125 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
128 disk_csum
= sb
->sb_csum
;
131 for (i
=0; size
>=4; size
-= 4 ) {
132 newcsum
+= __le32_to_cpu(*isuper
);
137 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
139 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
140 sb
->sb_csum
= disk_csum
;
141 return __cpu_to_le32(csum
);
145 static void examine_super1(struct supertype
*st
, char *homehost
)
147 struct mdp_superblock_1
*sb
= st
->sb
;
153 int l
= homehost
? strlen(homehost
) : 0;
155 unsigned long long sb_offset
;
157 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
158 printf(" Version : 1");
159 sb_offset
= __le64_to_cpu(sb
->super_offset
);
162 else if (sb_offset
<= 8)
166 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
167 printf(" Array UUID : ");
168 for (i
=0; i
<16; i
++) {
169 if ((i
&3)==0 && i
!= 0) printf(":");
170 printf("%02x", sb
->set_uuid
[i
]);
173 printf(" Name : %.32s", sb
->set_name
);
174 if (l
> 0 && l
< 32 &&
175 sb
->set_name
[l
] == ':' &&
176 strncmp(sb
->set_name
, homehost
, l
) == 0)
177 printf(" (local to host %s)", homehost
);
179 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
180 printf(" Creation Time : %.24s\n", ctime(&atime
));
181 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
182 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
183 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
185 printf(" Avail Dev Size : %llu%s\n",
186 (unsigned long long)__le64_to_cpu(sb
->data_size
),
187 human_size(__le64_to_cpu(sb
->data_size
)<<9));
188 if (__le32_to_cpu(sb
->level
) >= 0) {
190 switch(__le32_to_cpu(sb
->level
)) {
191 case 1: ddsks
=1;break;
193 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
194 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
196 layout
= __le32_to_cpu(sb
->layout
);
197 ddsks
= __le32_to_cpu(sb
->raid_disks
)
198 / (layout
&255) / ((layout
>>8)&255);
201 printf(" Array Size : %llu%s\n",
202 ddsks
*(unsigned long long)__le64_to_cpu(sb
->size
),
203 human_size(ddsks
*__le64_to_cpu(sb
->size
)<<9));
204 if (sb
->size
!= sb
->data_size
)
205 printf(" Used Dev Size : %llu%s\n",
206 (unsigned long long)__le64_to_cpu(sb
->size
),
207 human_size(__le64_to_cpu(sb
->size
)<<9));
210 printf(" Data Offset : %llu sectors\n",
211 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
212 printf(" Super Offset : %llu sectors\n",
213 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
214 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
215 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
216 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
217 printf(" Device UUID : ");
218 for (i
=0; i
<16; i
++) {
219 if ((i
&3)==0 && i
!= 0) printf(":");
220 printf("%02x", sb
->device_uuid
[i
]);
224 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
225 printf("Internal Bitmap : %ld sectors from superblock\n",
226 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
228 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
229 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
230 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
231 if (__le32_to_cpu(sb
->delta_disks
)) {
232 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
233 if (__le32_to_cpu(sb
->delta_disks
))
234 printf(" (%d->%d)\n",
235 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
236 __le32_to_cpu(sb
->raid_disks
));
238 printf(" (%d->%d)\n", __le32_to_cpu(sb
->raid_disks
),
239 __le32_to_cpu(sb
->raid_disks
)+__le32_to_cpu(sb
->delta_disks
));
241 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
242 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
243 printf(" New Level : %s\n", c
?c
:"-unknown-");
245 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
246 if (__le32_to_cpu(sb
->level
) == 5) {
247 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
248 printf(" New Layout : %s\n", c
?c
:"-unknown-");
250 if (__le32_to_cpu(sb
->level
) == 10) {
251 printf(" New Layout : near=%d, %s=%d\n",
252 __le32_to_cpu(sb
->new_layout
)&255,
253 (__le32_to_cpu(sb
->new_layout
)&0x10000)?"offset":"far",
254 (__le32_to_cpu(sb
->new_layout
)>>8)&255);
257 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
258 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
263 if (sb
->devflags
& WriteMostly1
)
264 printf(" write-mostly");
268 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
269 printf(" Update Time : %.24s\n", ctime(&atime
));
271 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
272 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
274 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
275 __le32_to_cpu(calc_sb_1_csum(sb
)));
276 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
278 if (__le32_to_cpu(sb
->level
) == 5) {
279 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
280 printf(" Layout : %s\n", c
?c
:"-unknown-");
282 if (__le32_to_cpu(sb
->level
) == 10) {
283 int lo
= __le32_to_cpu(sb
->layout
);
284 printf(" Layout : near=%d, %s=%d\n",
286 (lo
&0x10000)?"offset":"far",
289 switch(__le32_to_cpu(sb
->level
)) {
295 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
298 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
303 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
304 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
305 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
307 for (d
=0; d
< i
; d
++) {
308 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
310 if (role
== 0xffff) printf("empty");
311 else if(role
== 0xfffe) printf("failed");
312 else printf("%d", role
);
315 printf(" Array State : ");
316 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
); d
++) {
320 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
321 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
323 if (i
== __le32_to_cpu(sb
->dev_number
))
328 if (cnt
> 1) printf("?");
329 else if (cnt
== 1 && me
) printf("U");
330 else if (cnt
== 1) printf("u");
334 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
335 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
339 if (faulty
) printf(" %d failed", faulty
);
344 static void brief_examine_super1(struct supertype
*st
)
346 struct mdp_superblock_1
*sb
= st
->sb
;
348 unsigned long long sb_offset
;
350 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
352 nm
= strchr(sb
->set_name
, ':');
355 else if (sb
->set_name
[0])
360 printf("ARRAY /dev/md/%s level=%s ", nm
, c
?c
:"-unknown-");
361 sb_offset
= __le64_to_cpu(sb
->super_offset
);
363 printf("metadata=1.1 ");
364 else if (sb_offset
<= 8)
365 printf("metadata=1.2 ");
367 printf("metadata=1.0 ");
368 printf("num-devices=%d UUID=", __le32_to_cpu(sb
->raid_disks
));
369 for (i
=0; i
<16; i
++) {
370 if ((i
&3)==0 && i
!= 0) printf(":");
371 printf("%02x", sb
->set_uuid
[i
]);
374 printf(" name=%.32s", sb
->set_name
);
378 static void detail_super1(struct supertype
*st
, char *homehost
)
380 struct mdp_superblock_1
*sb
= st
->sb
;
382 int l
= homehost
? strlen(homehost
) : 0;
384 printf(" Name : %.32s", sb
->set_name
);
385 if (l
> 0 && l
< 32 &&
386 sb
->set_name
[l
] == ':' &&
387 strncmp(sb
->set_name
, homehost
, l
) == 0)
388 printf(" (local to host %s)", homehost
);
389 printf("\n UUID : ");
390 for (i
=0; i
<16; i
++) {
391 if ((i
&3)==0 && i
!= 0) printf(":");
392 printf("%02x", sb
->set_uuid
[i
]);
394 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
397 static void brief_detail_super1(struct supertype
*st
)
399 struct mdp_superblock_1
*sb
= st
->sb
;
403 printf(" name=%.32s", sb
->set_name
);
405 for (i
=0; i
<16; i
++) {
406 if ((i
&3)==0 && i
!= 0) printf(":");
407 printf("%02x", sb
->set_uuid
[i
]);
411 static void export_super1(struct supertype
*st
)
413 struct mdp_superblock_1
*sb
= st
->sb
;
418 if (sb
->set_name
[i
] == '\n' ||
419 sb
->set_name
[i
] == '\0') {
424 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
426 for (i
=0; i
<16; i
++) {
427 if ((i
&3)==0 && i
!= 0) printf(":");
428 printf("%02x", sb
->set_uuid
[i
]);
435 static int match_home1(struct supertype
*st
, char *homehost
)
437 struct mdp_superblock_1
*sb
= st
->sb
;
438 int l
= homehost
? strlen(homehost
) : 0;
440 return (l
> 0 && l
< 32 &&
441 sb
->set_name
[l
] == ':' &&
442 strncmp(sb
->set_name
, homehost
, l
) == 0);
445 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
447 struct mdp_superblock_1
*super
= st
->sb
;
448 char *cuuid
= (char*)uuid
;
451 cuuid
[i
] = super
->set_uuid
[i
];
454 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
)
456 struct mdp_superblock_1
*sb
= st
->sb
;
461 info
->array
.major_version
= 1;
462 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
463 info
->array
.patch_version
= 0;
464 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
465 info
->array
.level
= __le32_to_cpu(sb
->level
);
466 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
467 info
->array
.md_minor
= -1;
468 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
469 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
470 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
472 (__le64_to_cpu(sb
->resync_offset
) >= __le64_to_cpu(sb
->size
))
475 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
476 info
->component_size
= __le64_to_cpu(sb
->size
);
478 info
->disk
.major
= 0;
479 info
->disk
.minor
= 0;
480 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
481 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
482 __le32_to_cpu(sb
->max_dev
) > 512)
485 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
487 info
->disk
.raid_disk
= -1;
490 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
493 info
->disk
.state
= 1; /* faulty */
496 info
->disk
.state
= 6; /* active and in sync */
497 info
->disk
.raid_disk
= role
;
499 info
->events
= __le64_to_cpu(sb
->events
);
501 memcpy(info
->uuid
, sb
->set_uuid
, 16);
503 strncpy(info
->name
, sb
->set_name
, 32);
506 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
507 info
->reshape_active
= 1;
508 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
509 info
->new_level
= __le32_to_cpu(sb
->new_level
);
510 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
511 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
512 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
514 info
->reshape_active
= 0;
516 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
517 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
518 if (/*role == 0xFFFF || */role
< info
->array
.raid_disks
)
522 info
->array
.working_disks
= working
;
525 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
527 char *devname
, int verbose
,
528 int uuid_set
, char *homehost
)
530 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
531 * For others, the return value is ignored.
534 struct mdp_superblock_1
*sb
= st
->sb
;
536 if (strcmp(update
, "force-one")==0) {
537 /* Not enough devices for a working array,
538 * so bring this one up-to-date
540 if (sb
->events
!= __cpu_to_le64(info
->events
))
542 sb
->events
= __cpu_to_le64(info
->events
);
544 if (strcmp(update
, "force-array")==0) {
545 /* Degraded array and 'force' requests to
546 * maybe need to mark it 'clean'.
548 switch(__le32_to_cpu(sb
->level
)) {
549 case 5: case 4: case 6:
550 /* need to force clean */
551 if (sb
->resync_offset
!= ~0ULL)
553 sb
->resync_offset
= ~0ULL;
556 if (strcmp(update
, "assemble")==0) {
557 int d
= info
->disk
.number
;
559 if (info
->disk
.state
== 6)
560 want
= __cpu_to_le32(info
->disk
.raid_disk
);
563 if (sb
->dev_roles
[d
] != want
) {
564 sb
->dev_roles
[d
] = want
;
568 if (strcmp(update
, "linear-grow-new") == 0) {
571 int max
= __le32_to_cpu(sb
->max_dev
);
573 for (i
=0 ; i
< max
; i
++)
574 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
576 sb
->dev_number
= __cpu_to_le32(i
);
577 info
->disk
.number
= i
;
578 if (max
>= __le32_to_cpu(sb
->max_dev
))
579 sb
->max_dev
= __cpu_to_le32(max
+1);
581 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
582 read(rfd
, sb
->device_uuid
, 16) != 16) {
583 *(__u32
*)(sb
->device_uuid
) = random();
584 *(__u32
*)(sb
->device_uuid
+4) = random();
585 *(__u32
*)(sb
->device_uuid
+8) = random();
586 *(__u32
*)(sb
->device_uuid
+12) = random();
590 __cpu_to_le16(info
->disk
.raid_disk
);
592 if (strcmp(update
, "linear-grow-update") == 0) {
593 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
594 sb
->dev_roles
[info
->disk
.number
] =
595 __cpu_to_le16(info
->disk
.raid_disk
);
597 if (strcmp(update
, "resync") == 0) {
598 /* make sure resync happens */
599 sb
->resync_offset
= 0ULL;
601 if (strcmp(update
, "uuid") == 0) {
602 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
604 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
605 struct bitmap_super_s
*bm
;
606 bm
= (struct bitmap_super_s
*)(st
->sb
+1024);
607 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
610 if (strcmp(update
, "homehost") == 0 &&
614 c
= strchr(sb
->set_name
, ':');
616 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
618 strncpy(info
->name
, sb
->set_name
, 32);
621 if (strcmp(update
, "name") == 0) {
622 if (info
->name
[0] == 0)
623 sprintf(info
->name
, "%d", info
->array
.md_minor
);
624 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
626 strchr(info
->name
, ':') == NULL
&&
627 strlen(homehost
)+1+strlen(info
->name
) < 32) {
628 strcpy(sb
->set_name
, homehost
);
629 strcat(sb
->set_name
, ":");
630 strcat(sb
->set_name
, info
->name
);
632 strcpy(sb
->set_name
, info
->name
);
634 if (strcmp(update
, "devicesize") == 0 &&
635 __le64_to_cpu(sb
->super_offset
) <
636 __le64_to_cpu(sb
->data_offset
)) {
637 /* set data_size to device size less data_offset */
638 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
639 (st
->sb
+ 1024 + sizeof(struct bitmap_super_s
));
640 printf("Size was %llu\n", (unsigned long long)
641 __le64_to_cpu(sb
->data_size
));
642 sb
->data_size
= __cpu_to_le64(
643 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
644 printf("Size is %llu\n", (unsigned long long)
645 __le64_to_cpu(sb
->data_size
));
647 if (strcmp(update
, "_reshape_progress")==0)
648 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
650 sb
->sb_csum
= calc_sb_1_csum(sb
);
654 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
655 unsigned long long size
, char *name
, char *homehost
, int *uuid
)
657 struct mdp_superblock_1
*sb
= malloc(1024 + sizeof(bitmap_super_t
) +
658 sizeof(struct misc_dev_info
));
665 if (info
->major_version
== -1) {
666 /* zeroing superblock */
670 spares
= info
->working_disks
- info
->active_disks
;
671 if (info
->raid_disks
+ spares
> 384) {
672 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
673 info
->raid_disks
, spares
, 384);
677 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
678 sb
->major_version
= __cpu_to_le32(1);
683 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
685 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
686 read(rfd
, sb
->set_uuid
, 16) != 16) {
687 *(__u32
*)(sb
->set_uuid
) = random();
688 *(__u32
*)(sb
->set_uuid
+4) = random();
689 *(__u32
*)(sb
->set_uuid
+8) = random();
690 *(__u32
*)(sb
->set_uuid
+12) = random();
692 if (rfd
>= 0) close(rfd
);
695 if (name
== NULL
|| *name
== 0) {
696 sprintf(defname
, "%d", info
->md_minor
);
699 memset(sb
->set_name
, 0, 32);
701 strchr(name
, ':')== NULL
&&
702 strlen(homehost
)+1+strlen(name
) < 32) {
703 strcpy(sb
->set_name
, homehost
);
704 strcat(sb
->set_name
, ":");
705 strcat(sb
->set_name
, name
);
707 strcpy(sb
->set_name
, name
);
709 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
710 sb
->level
= __cpu_to_le32(info
->level
);
711 sb
->layout
= __cpu_to_le32(info
->layout
);
712 sb
->size
= __cpu_to_le64(size
*2ULL);
713 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
714 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
716 sb
->data_offset
= __cpu_to_le64(0);
717 sb
->data_size
= __cpu_to_le64(0);
718 sb
->super_offset
= __cpu_to_le64(0);
719 sb
->recovery_offset
= __cpu_to_le64(0);
721 sb
->utime
= sb
->ctime
;
722 sb
->events
= __cpu_to_le64(1);
723 if (info
->state
& (1<<MD_SB_CLEAN
))
724 sb
->resync_offset
= ~0ULL;
726 sb
->resync_offset
= 0;
727 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
728 sizeof(sb
->dev_roles
[0]));
729 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
731 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
736 /* Add a device to the superblock being created */
737 static void add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
)
739 struct mdp_superblock_1
*sb
= st
->sb
;
740 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
741 if ((dk
->state
& 6) == 6) /* active, sync */
742 *rp
= __cpu_to_le16(dk
->raid_disk
);
743 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
749 static void locate_bitmap1(struct supertype
*st
, int fd
);
751 static int store_super1(struct supertype
*st
, int fd
)
753 struct mdp_superblock_1
*sb
= st
->sb
;
754 unsigned long long sb_offset
;
756 unsigned long long dsize
;
758 if (!get_dev_size(fd
, NULL
, &dsize
))
767 * Calculate the position of the superblock.
768 * It is always aligned to a 4K boundary and
769 * depending on minor_version, it can be:
770 * 0: At least 8K, but less than 12K, from end of device
771 * 1: At start of device
772 * 2: 4K from start of device.
774 switch(st
->minor_version
) {
778 sb_offset
&= ~(4*2-1);
792 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
793 0 != __le64_to_cpu(sb
->super_offset
)
795 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
799 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
802 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
804 if (write(fd
, sb
, sbsize
) != sbsize
)
807 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
808 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
810 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
811 locate_bitmap1(st
, fd
);
812 if (write(fd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
820 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
822 static unsigned long choose_bm_space(unsigned long devsize
)
824 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
825 * if bigger than 200Gig, save 128k
827 if (devsize
< 64*2) return 0;
828 if (devsize
- 64*2 >= 200*1024*1024*2)
830 if (devsize
- 4*2 > 8*1024*1024*2)
835 static int write_init_super1(struct supertype
*st
,
836 mdu_disk_info_t
*dinfo
, char *devname
)
838 struct mdp_superblock_1
*sb
= st
->sb
;
839 struct supertype refst
;
840 int fd
= open(devname
, O_RDWR
| O_EXCL
);
845 unsigned long long dsize
, array_size
;
850 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
855 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
856 if (dinfo
->state
& (1<<MD_DISK_WRITEMOSTLY
))
857 sb
->devflags
|= __cpu_to_le32(WriteMostly1
);
859 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
860 read(rfd
, sb
->device_uuid
, 16) != 16) {
861 *(__u32
*)(sb
->device_uuid
) = random();
862 *(__u32
*)(sb
->device_uuid
+4) = random();
863 *(__u32
*)(sb
->device_uuid
+8) = random();
864 *(__u32
*)(sb
->device_uuid
+12) = random();
866 if (rfd
>= 0) close(rfd
);
871 if (load_super1(&refst
, fd
, NULL
)==0) {
872 struct mdp_superblock_1
*refsb
= refst
.sb
;
874 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
875 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
876 /* same array, so preserve events and dev_number */
877 sb
->events
= refsb
->events
;
878 /* bugs in 2.6.17 and earlier mean the dev_number
879 * chosen in Manage must be preserved
881 if (get_linux_version() >= 2006018)
882 sb
->dev_number
= refsb
->dev_number
;
887 if (!get_dev_size(fd
, NULL
, &dsize
))
898 * Calculate the position of the superblock.
899 * It is always aligned to a 4K boundary and
900 * depending on minor_version, it can be:
901 * 0: At least 8K, but less than 12K, from end of device
902 * 1: At start of device
903 * 2: 4K from start of device.
904 * Depending on the array size, we might leave extra space
907 array_size
= __le64_to_cpu(sb
->size
);
908 /* work out how much space we left of a bitmap */
909 bm_space
= choose_bm_space(array_size
);
911 switch(st
->minor_version
) {
915 sb_offset
&= ~(4*2-1);
916 sb
->super_offset
= __cpu_to_le64(sb_offset
);
917 sb
->data_offset
= __cpu_to_le64(0);
918 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
921 sb
->super_offset
= __cpu_to_le64(0);
922 if (4*2 + bm_space
+ __le64_to_cpu(sb
->size
) > dsize
)
923 bm_space
= dsize
- __le64_to_cpu(sb
->size
) - 4*2;
924 sb
->data_offset
= __cpu_to_le64(bm_space
+ 4*2);
925 sb
->data_size
= __cpu_to_le64(dsize
- bm_space
- 4*2);
929 sb
->super_offset
= __cpu_to_le64(4*2);
930 if (4*2 + 4*2 + bm_space
+ __le64_to_cpu(sb
->size
) > dsize
)
931 bm_space
= dsize
- __le64_to_cpu(sb
->size
) - 4*2 - 4*2;
932 sb
->data_offset
= __cpu_to_le64(4*2 + 4*2 + bm_space
);
933 sb
->data_size
= __cpu_to_le64(dsize
- 4*2 - 4*2 - bm_space
);
940 sb
->sb_csum
= calc_sb_1_csum(sb
);
941 rv
= store_super1(st
, fd
);
943 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
945 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
946 rv
= st
->ss
->write_bitmap(st
, fd
);
951 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
955 * 0 same, or first was empty, and second was copied
956 * 1 second had wrong number
960 struct mdp_superblock_1
*first
= st
->sb
;
961 struct mdp_superblock_1
*second
= tst
->sb
;
963 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
965 if (second
->major_version
!= __cpu_to_le32(1))
969 first
= malloc(1024+sizeof(bitmap_super_t
) +
970 sizeof(struct misc_dev_info
));
971 memcpy(first
, second
, 1024+sizeof(bitmap_super_t
) +
972 sizeof(struct misc_dev_info
));
976 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
979 if (first
->ctime
!= second
->ctime
||
980 first
->level
!= second
->level
||
981 first
->layout
!= second
->layout
||
982 first
->size
!= second
->size
||
983 first
->chunksize
!= second
->chunksize
||
984 first
->raid_disks
!= second
->raid_disks
)
989 static void free_super1(struct supertype
*st
);
991 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
993 unsigned long long dsize
;
994 unsigned long long sb_offset
;
995 struct mdp_superblock_1
*super
;
997 struct bitmap_super_s
*bsb
;
998 struct misc_dev_info
*misc
;
1002 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1004 struct supertype tst
;
1005 __u64 bestctime
= 0;
1006 /* guess... choose latest ctime */
1009 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1010 switch(load_super1(&tst
, fd
, devname
)) {
1011 case 0: super
= tst
.sb
;
1012 if (bestvers
== -1 ||
1013 bestctime
< __le64_to_cpu(super
->ctime
)) {
1014 bestvers
= tst
.minor_version
;
1015 bestctime
= __le64_to_cpu(super
->ctime
);
1020 case 1: return 1; /*bad device */
1021 case 2: break; /* bad, try next */
1024 if (bestvers
!= -1) {
1026 tst
.minor_version
= bestvers
;
1029 rv
= load_super1(&tst
, fd
, devname
);
1036 if (!get_dev_size(fd
, devname
, &dsize
))
1042 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
1048 * Calculate the position of the superblock.
1049 * It is always aligned to a 4K boundary and
1050 * depending on minor_version, it can be:
1051 * 0: At least 8K, but less than 12K, from end of device
1052 * 1: At start of device
1053 * 2: 4K from start of device.
1055 switch(st
->minor_version
) {
1059 sb_offset
&= ~(4*2-1);
1071 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
1074 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1076 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
1077 devname
, strerror(errno
));
1081 super
= malloc(1024 + sizeof(bitmap_super_t
) +
1082 sizeof(struct misc_dev_info
));
1084 if (read(fd
, super
, 1024) != 1024) {
1086 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
1092 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1094 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
1095 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1100 if (__le32_to_cpu(super
->major_version
) != 1) {
1102 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
1103 devname
, __le32_to_cpu(super
->major_version
));
1107 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1109 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
1116 bsb
= (struct bitmap_super_s
*)(((char*)super
)+1024);
1118 misc
= (struct misc_dev_info
*) (bsb
+1);
1119 misc
->device_size
= dsize
;
1121 /* Now check on the bitmap superblock */
1122 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1124 /* Read the bitmap superblock and make sure it looks
1125 * valid. If it doesn't clear the bit. An --assemble --force
1126 * should get that written out.
1128 locate_bitmap1(st
, fd
);
1129 if (read(fd
, ((char*)super
)+1024, sizeof(struct bitmap_super_s
))
1130 != sizeof(struct bitmap_super_s
))
1133 uuid_from_super1(st
, uuid
);
1134 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1135 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1140 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
) & ~1);
1145 static struct supertype
*match_metadata_desc1(char *arg
)
1147 struct supertype
*st
= malloc(sizeof(*st
));
1153 if (strcmp(arg
, "1.0") == 0) {
1154 st
->minor_version
= 0;
1157 if (strcmp(arg
, "1.1") == 0) {
1158 st
->minor_version
= 1;
1161 if (strcmp(arg
, "1.2") == 0) {
1162 st
->minor_version
= 2;
1165 if (strcmp(arg
, "1") == 0 ||
1166 strcmp(arg
, "default/large") == 0) {
1167 st
->minor_version
= -1;
1175 /* find available size on device with this devsize, using
1176 * superblock type st, and reserving 'reserve' sectors for
1179 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1184 devsize
-= choose_bm_space(devsize
);
1186 switch(st
->minor_version
) {
1187 case -1: /* no specified. Now time to set default */
1188 st
->minor_version
= 0;
1192 return ((devsize
- 8*2 ) & ~(4*2-1));
1194 /* at start, 4K for superblock and possible bitmap */
1195 return devsize
- 4*2;
1197 /* 4k from start, 4K for superblock and possible bitmap */
1198 return devsize
- (4+4)*2;
1204 add_internal_bitmap1(struct supertype
*st
,
1205 int *chunkp
, int delay
, int write_behind
,
1206 unsigned long long size
,
1207 int may_change
, int major
)
1210 * If not may_change, then this is a 'Grow', and the bitmap
1211 * must fit after the superblock.
1212 * If may_change, then this is create, and we can put the bitmap
1213 * before the superblock if we like, or may move the start.
1214 * If !may_change, the bitmap MUST live at offset of 1K, until
1215 * we get a sysfs interface.
1217 * size is in sectors, chunk is in bytes !!!
1220 unsigned long long bits
;
1221 unsigned long long max_bits
;
1222 unsigned long long min_chunk
;
1224 int chunk
= *chunkp
;
1226 struct mdp_superblock_1
*sb
= st
->sb
;
1227 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + 1024);
1229 switch(st
->minor_version
) {
1231 /* either 3K after the superblock, or some amount of space
1235 /* We are creating array, so we *know* how much room has
1239 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1241 /* make it 3K after the superblock */
1246 room
= __le64_to_cpu(sb
->super_offset
)
1247 - __le64_to_cpu(sb
->data_offset
)
1248 - __le64_to_cpu(sb
->data_size
);
1249 /* remove '1 ||' when we can set offset via sysfs */
1250 if (1 || (room
< 3*2 &&
1251 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1255 offset
= 0; /* means movable offset */
1260 case 2: /* between superblock and data */
1263 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1265 room
= __le64_to_cpu(sb
->data_offset
)
1266 - __le64_to_cpu(sb
->super_offset
);
1267 if (1 || __le32_to_cpu(sb
->max_dev
) <= 384) {
1280 if (chunk
== UnSet
&& room
> 128*2)
1281 /* Limit to 128K of bitmap when chunk size not requested */
1284 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1286 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1287 bits
= (size
*512)/min_chunk
+1;
1288 while (bits
> max_bits
) {
1294 else if (chunk
< min_chunk
)
1295 return 0; /* chunk size too small */
1296 if (chunk
== 0) /* rounding problem */
1300 bits
= (size
*512) / chunk
+ 1;
1301 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +511)/512;
1305 sb
->bitmap_offset
= __cpu_to_le32(offset
);
1307 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) | 1);
1308 memset(bms
, 0, sizeof(*bms
));
1309 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1310 bms
->version
= __cpu_to_le32(major
);
1311 uuid_from_super1(st
, (int*)bms
->uuid
);
1312 bms
->chunksize
= __cpu_to_le32(chunk
);
1313 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1314 bms
->sync_size
= __cpu_to_le64(size
);
1315 bms
->write_behind
= __cpu_to_le32(write_behind
);
1322 static void locate_bitmap1(struct supertype
*st
, int fd
)
1324 unsigned long long offset
;
1325 struct mdp_superblock_1
*sb
;
1329 if (st
->ss
->load_super(st
, fd
, NULL
))
1330 return; /* no error I hope... */
1335 offset
= __le64_to_cpu(sb
->super_offset
);
1336 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
1339 lseek64(fd
, offset
<<9, 0);
1342 static int write_bitmap1(struct supertype
*st
, int fd
)
1344 struct mdp_superblock_1
*sb
= st
->sb
;
1345 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+1024);
1351 locate_bitmap1(st
, fd
);
1353 if (write(fd
, ((char*)sb
)+1024, sizeof(bitmap_super_t
)) !=
1354 sizeof(bitmap_super_t
))
1356 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1357 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1358 memset(buf
, 0xff, sizeof(buf
));
1359 while (towrite
> 0) {
1361 if (n
> sizeof(buf
))
1363 n
= write(fd
, buf
, n
);
1376 static void free_super1(struct supertype
*st
)
1383 struct superswitch super1
= {
1385 .examine_super
= examine_super1
,
1386 .brief_examine_super
= brief_examine_super1
,
1387 .detail_super
= detail_super1
,
1388 .brief_detail_super
= brief_detail_super1
,
1389 .export_super
= export_super1
,
1391 .match_home
= match_home1
,
1392 .uuid_from_super
= uuid_from_super1
,
1393 .getinfo_super
= getinfo_super1
,
1394 .update_super
= update_super1
,
1395 .init_super
= init_super1
,
1396 .add_to_super
= add_to_super1
,
1397 .store_super
= store_super1
,
1398 .write_init_super
= write_init_super1
,
1399 .compare_super
= compare_super1
,
1400 .load_super
= load_super1
,
1401 .match_metadata_desc
= match_metadata_desc1
,
1402 .avail_size
= avail_size1
,
1403 .add_internal_bitmap
= add_internal_bitmap1
,
1404 .locate_bitmap
= locate_bitmap1
,
1405 .write_bitmap
= write_bitmap1
,
1406 .free_super
= free_super1
,
1408 #if __BYTE_ORDER == BIG_ENDIAN