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 export_examine_super1(struct supertype
*st
)
380 struct mdp_superblock_1
*sb
= st
->sb
;
384 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
385 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
387 if (sb
->set_name
[i
] == '\n' ||
388 sb
->set_name
[i
] == '\0') {
393 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
395 for (i
=0; i
<16; i
++) {
396 if ((i
&3)==0 && i
!= 0) printf(":");
397 printf("%02x", sb
->set_uuid
[i
]);
400 printf("MD_UPDATE_TIME=%llu\n",
401 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
402 printf("MD_DEV_UUID=");
403 for (i
=0; i
<16; i
++) {
404 if ((i
&3)==0 && i
!= 0) printf(":");
405 printf("%02x", sb
->device_uuid
[i
]);
408 printf("MD_EVENTS=%llu\n",
409 (unsigned long long)__le64_to_cpu(sb
->events
));
412 static void detail_super1(struct supertype
*st
, char *homehost
)
414 struct mdp_superblock_1
*sb
= st
->sb
;
416 int l
= homehost
? strlen(homehost
) : 0;
418 printf(" Name : %.32s", sb
->set_name
);
419 if (l
> 0 && l
< 32 &&
420 sb
->set_name
[l
] == ':' &&
421 strncmp(sb
->set_name
, homehost
, l
) == 0)
422 printf(" (local to host %s)", homehost
);
423 printf("\n UUID : ");
424 for (i
=0; i
<16; i
++) {
425 if ((i
&3)==0 && i
!= 0) printf(":");
426 printf("%02x", sb
->set_uuid
[i
]);
428 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
431 static void brief_detail_super1(struct supertype
*st
)
433 struct mdp_superblock_1
*sb
= st
->sb
;
437 printf(" name=%.32s", sb
->set_name
);
439 for (i
=0; i
<16; i
++) {
440 if ((i
&3)==0 && i
!= 0) printf(":");
441 printf("%02x", sb
->set_uuid
[i
]);
445 static void export_detail_super1(struct supertype
*st
)
447 struct mdp_superblock_1
*sb
= st
->sb
;
452 if (sb
->set_name
[i
] == '\n' ||
453 sb
->set_name
[i
] == '\0') {
458 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
460 for (i
=0; i
<16; i
++) {
461 if ((i
&3)==0 && i
!= 0) printf(":");
462 printf("%02x", sb
->set_uuid
[i
]);
469 static int match_home1(struct supertype
*st
, char *homehost
)
471 struct mdp_superblock_1
*sb
= st
->sb
;
472 int l
= homehost
? strlen(homehost
) : 0;
474 return (l
> 0 && l
< 32 &&
475 sb
->set_name
[l
] == ':' &&
476 strncmp(sb
->set_name
, homehost
, l
) == 0);
479 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
481 struct mdp_superblock_1
*super
= st
->sb
;
482 char *cuuid
= (char*)uuid
;
485 cuuid
[i
] = super
->set_uuid
[i
];
488 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
)
490 struct mdp_superblock_1
*sb
= st
->sb
;
495 info
->array
.major_version
= 1;
496 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
497 info
->array
.patch_version
= 0;
498 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
499 info
->array
.level
= __le32_to_cpu(sb
->level
);
500 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
501 info
->array
.md_minor
= -1;
502 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
503 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
504 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
506 (__le64_to_cpu(sb
->resync_offset
) >= __le64_to_cpu(sb
->size
))
509 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
510 info
->component_size
= __le64_to_cpu(sb
->size
);
512 info
->disk
.major
= 0;
513 info
->disk
.minor
= 0;
514 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
515 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
516 __le32_to_cpu(sb
->max_dev
) > 512)
519 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
521 info
->disk
.raid_disk
= -1;
524 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
527 info
->disk
.state
= 1; /* faulty */
530 info
->disk
.state
= 6; /* active and in sync */
531 info
->disk
.raid_disk
= role
;
533 info
->events
= __le64_to_cpu(sb
->events
);
535 memcpy(info
->uuid
, sb
->set_uuid
, 16);
537 strncpy(info
->name
, sb
->set_name
, 32);
540 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
541 info
->reshape_active
= 1;
542 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
543 info
->new_level
= __le32_to_cpu(sb
->new_level
);
544 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
545 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
546 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
548 info
->reshape_active
= 0;
550 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
551 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
552 if (/*role == 0xFFFF || */role
< info
->array
.raid_disks
)
556 info
->array
.working_disks
= working
;
559 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
561 char *devname
, int verbose
,
562 int uuid_set
, char *homehost
)
564 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
565 * For others, the return value is ignored.
568 struct mdp_superblock_1
*sb
= st
->sb
;
570 if (strcmp(update
, "force-one")==0) {
571 /* Not enough devices for a working array,
572 * so bring this one up-to-date
574 if (sb
->events
!= __cpu_to_le64(info
->events
))
576 sb
->events
= __cpu_to_le64(info
->events
);
578 if (strcmp(update
, "force-array")==0) {
579 /* Degraded array and 'force' requests to
580 * maybe need to mark it 'clean'.
582 switch(__le32_to_cpu(sb
->level
)) {
583 case 5: case 4: case 6:
584 /* need to force clean */
585 if (sb
->resync_offset
!= ~0ULL)
587 sb
->resync_offset
= ~0ULL;
590 if (strcmp(update
, "assemble")==0) {
591 int d
= info
->disk
.number
;
593 if (info
->disk
.state
== 6)
594 want
= __cpu_to_le32(info
->disk
.raid_disk
);
597 if (sb
->dev_roles
[d
] != want
) {
598 sb
->dev_roles
[d
] = want
;
602 if (strcmp(update
, "linear-grow-new") == 0) {
605 int max
= __le32_to_cpu(sb
->max_dev
);
607 for (i
=0 ; i
< max
; i
++)
608 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
610 sb
->dev_number
= __cpu_to_le32(i
);
611 info
->disk
.number
= i
;
612 if (max
>= __le32_to_cpu(sb
->max_dev
))
613 sb
->max_dev
= __cpu_to_le32(max
+1);
615 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
616 read(rfd
, sb
->device_uuid
, 16) != 16) {
617 *(__u32
*)(sb
->device_uuid
) = random();
618 *(__u32
*)(sb
->device_uuid
+4) = random();
619 *(__u32
*)(sb
->device_uuid
+8) = random();
620 *(__u32
*)(sb
->device_uuid
+12) = random();
624 __cpu_to_le16(info
->disk
.raid_disk
);
626 if (strcmp(update
, "linear-grow-update") == 0) {
627 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
628 sb
->dev_roles
[info
->disk
.number
] =
629 __cpu_to_le16(info
->disk
.raid_disk
);
631 if (strcmp(update
, "resync") == 0) {
632 /* make sure resync happens */
633 sb
->resync_offset
= 0ULL;
635 if (strcmp(update
, "uuid") == 0) {
636 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
638 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
639 struct bitmap_super_s
*bm
;
640 bm
= (struct bitmap_super_s
*)(st
->sb
+1024);
641 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
644 if (strcmp(update
, "homehost") == 0 &&
648 c
= strchr(sb
->set_name
, ':');
650 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
652 strncpy(info
->name
, sb
->set_name
, 32);
655 if (strcmp(update
, "name") == 0) {
656 if (info
->name
[0] == 0)
657 sprintf(info
->name
, "%d", info
->array
.md_minor
);
658 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
660 strchr(info
->name
, ':') == NULL
&&
661 strlen(homehost
)+1+strlen(info
->name
) < 32) {
662 strcpy(sb
->set_name
, homehost
);
663 strcat(sb
->set_name
, ":");
664 strcat(sb
->set_name
, info
->name
);
666 strcpy(sb
->set_name
, info
->name
);
668 if (strcmp(update
, "devicesize") == 0 &&
669 __le64_to_cpu(sb
->super_offset
) <
670 __le64_to_cpu(sb
->data_offset
)) {
671 /* set data_size to device size less data_offset */
672 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
673 (st
->sb
+ 1024 + sizeof(struct bitmap_super_s
));
674 printf("Size was %llu\n", (unsigned long long)
675 __le64_to_cpu(sb
->data_size
));
676 sb
->data_size
= __cpu_to_le64(
677 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
678 printf("Size is %llu\n", (unsigned long long)
679 __le64_to_cpu(sb
->data_size
));
681 if (strcmp(update
, "_reshape_progress")==0)
682 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
684 sb
->sb_csum
= calc_sb_1_csum(sb
);
688 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
689 unsigned long long size
, char *name
, char *homehost
, int *uuid
)
691 struct mdp_superblock_1
*sb
= malloc(1024 + sizeof(bitmap_super_t
) +
692 sizeof(struct misc_dev_info
));
699 if (info
->major_version
== -1) {
700 /* zeroing superblock */
704 spares
= info
->working_disks
- info
->active_disks
;
705 if (info
->raid_disks
+ spares
> 384) {
706 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
707 info
->raid_disks
, spares
, 384);
711 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
712 sb
->major_version
= __cpu_to_le32(1);
717 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
719 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
720 read(rfd
, sb
->set_uuid
, 16) != 16) {
721 *(__u32
*)(sb
->set_uuid
) = random();
722 *(__u32
*)(sb
->set_uuid
+4) = random();
723 *(__u32
*)(sb
->set_uuid
+8) = random();
724 *(__u32
*)(sb
->set_uuid
+12) = random();
726 if (rfd
>= 0) close(rfd
);
729 if (name
== NULL
|| *name
== 0) {
730 sprintf(defname
, "%d", info
->md_minor
);
733 memset(sb
->set_name
, 0, 32);
735 strchr(name
, ':')== NULL
&&
736 strlen(homehost
)+1+strlen(name
) < 32) {
737 strcpy(sb
->set_name
, homehost
);
738 strcat(sb
->set_name
, ":");
739 strcat(sb
->set_name
, name
);
741 strcpy(sb
->set_name
, name
);
743 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
744 sb
->level
= __cpu_to_le32(info
->level
);
745 sb
->layout
= __cpu_to_le32(info
->layout
);
746 sb
->size
= __cpu_to_le64(size
*2ULL);
747 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
748 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
750 sb
->data_offset
= __cpu_to_le64(0);
751 sb
->data_size
= __cpu_to_le64(0);
752 sb
->super_offset
= __cpu_to_le64(0);
753 sb
->recovery_offset
= __cpu_to_le64(0);
755 sb
->utime
= sb
->ctime
;
756 sb
->events
= __cpu_to_le64(1);
757 if (info
->state
& (1<<MD_SB_CLEAN
))
758 sb
->resync_offset
= ~0ULL;
760 sb
->resync_offset
= 0;
761 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
762 sizeof(sb
->dev_roles
[0]));
763 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
765 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
770 /* Add a device to the superblock being created */
771 static void add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
)
773 struct mdp_superblock_1
*sb
= st
->sb
;
774 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
775 if ((dk
->state
& 6) == 6) /* active, sync */
776 *rp
= __cpu_to_le16(dk
->raid_disk
);
777 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
783 static void locate_bitmap1(struct supertype
*st
, int fd
);
785 static int store_super1(struct supertype
*st
, int fd
)
787 struct mdp_superblock_1
*sb
= st
->sb
;
788 unsigned long long sb_offset
;
790 unsigned long long dsize
;
792 if (!get_dev_size(fd
, NULL
, &dsize
))
801 * Calculate the position of the superblock.
802 * It is always aligned to a 4K boundary and
803 * depending on minor_version, it can be:
804 * 0: At least 8K, but less than 12K, from end of device
805 * 1: At start of device
806 * 2: 4K from start of device.
808 switch(st
->minor_version
) {
812 sb_offset
&= ~(4*2-1);
826 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
827 0 != __le64_to_cpu(sb
->super_offset
)
829 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
833 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
836 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
838 if (write(fd
, sb
, sbsize
) != sbsize
)
841 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
842 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
844 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
845 locate_bitmap1(st
, fd
);
846 if (write(fd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
854 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
856 static unsigned long choose_bm_space(unsigned long devsize
)
858 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
859 * if bigger than 200Gig, save 128k
861 if (devsize
< 64*2) return 0;
862 if (devsize
- 64*2 >= 200*1024*1024*2)
864 if (devsize
- 4*2 > 8*1024*1024*2)
869 static int write_init_super1(struct supertype
*st
,
870 mdu_disk_info_t
*dinfo
, char *devname
)
872 struct mdp_superblock_1
*sb
= st
->sb
;
873 struct supertype refst
;
874 int fd
= open(devname
, O_RDWR
| O_EXCL
);
879 unsigned long long dsize
, array_size
;
884 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
889 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
890 if (dinfo
->state
& (1<<MD_DISK_WRITEMOSTLY
))
891 sb
->devflags
|= __cpu_to_le32(WriteMostly1
);
893 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
894 read(rfd
, sb
->device_uuid
, 16) != 16) {
895 *(__u32
*)(sb
->device_uuid
) = random();
896 *(__u32
*)(sb
->device_uuid
+4) = random();
897 *(__u32
*)(sb
->device_uuid
+8) = random();
898 *(__u32
*)(sb
->device_uuid
+12) = random();
900 if (rfd
>= 0) close(rfd
);
905 if (load_super1(&refst
, fd
, NULL
)==0) {
906 struct mdp_superblock_1
*refsb
= refst
.sb
;
908 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
909 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
910 /* same array, so preserve events and dev_number */
911 sb
->events
= refsb
->events
;
912 /* bugs in 2.6.17 and earlier mean the dev_number
913 * chosen in Manage must be preserved
915 if (get_linux_version() >= 2006018)
916 sb
->dev_number
= refsb
->dev_number
;
921 if (!get_dev_size(fd
, NULL
, &dsize
))
932 * Calculate the position of the superblock.
933 * It is always aligned to a 4K boundary and
934 * depending on minor_version, it can be:
935 * 0: At least 8K, but less than 12K, from end of device
936 * 1: At start of device
937 * 2: 4K from start of device.
938 * Depending on the array size, we might leave extra space
941 array_size
= __le64_to_cpu(sb
->size
);
942 /* work out how much space we left for a bitmap */
943 bm_space
= choose_bm_space(array_size
);
945 switch(st
->minor_version
) {
949 sb_offset
&= ~(4*2-1);
950 sb
->super_offset
= __cpu_to_le64(sb_offset
);
951 sb
->data_offset
= __cpu_to_le64(0);
952 if (sb_offset
- bm_space
< array_size
)
953 bm_space
= sb_offset
- array_size
;
954 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
957 sb
->super_offset
= __cpu_to_le64(0);
958 if (4*2 + bm_space
+ __le64_to_cpu(sb
->size
) > dsize
)
959 bm_space
= dsize
- __le64_to_cpu(sb
->size
) - 4*2;
960 sb
->data_offset
= __cpu_to_le64(bm_space
+ 4*2);
961 sb
->data_size
= __cpu_to_le64(dsize
- bm_space
- 4*2);
965 sb
->super_offset
= __cpu_to_le64(4*2);
966 if (4*2 + 4*2 + bm_space
+ __le64_to_cpu(sb
->size
) > dsize
)
967 bm_space
= dsize
- __le64_to_cpu(sb
->size
) - 4*2 - 4*2;
968 sb
->data_offset
= __cpu_to_le64(4*2 + 4*2 + bm_space
);
969 sb
->data_size
= __cpu_to_le64(dsize
- 4*2 - 4*2 - bm_space
);
976 sb
->sb_csum
= calc_sb_1_csum(sb
);
977 rv
= store_super1(st
, fd
);
979 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
981 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
982 rv
= st
->ss
->write_bitmap(st
, fd
);
987 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
991 * 0 same, or first was empty, and second was copied
992 * 1 second had wrong number
996 struct mdp_superblock_1
*first
= st
->sb
;
997 struct mdp_superblock_1
*second
= tst
->sb
;
999 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
1001 if (second
->major_version
!= __cpu_to_le32(1))
1005 first
= malloc(1024+sizeof(bitmap_super_t
) +
1006 sizeof(struct misc_dev_info
));
1007 memcpy(first
, second
, 1024+sizeof(bitmap_super_t
) +
1008 sizeof(struct misc_dev_info
));
1012 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
1015 if (first
->ctime
!= second
->ctime
||
1016 first
->level
!= second
->level
||
1017 first
->layout
!= second
->layout
||
1018 first
->size
!= second
->size
||
1019 first
->chunksize
!= second
->chunksize
||
1020 first
->raid_disks
!= second
->raid_disks
)
1025 static void free_super1(struct supertype
*st
);
1027 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
1029 unsigned long long dsize
;
1030 unsigned long long sb_offset
;
1031 struct mdp_superblock_1
*super
;
1033 struct bitmap_super_s
*bsb
;
1034 struct misc_dev_info
*misc
;
1038 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1040 struct supertype tst
;
1041 __u64 bestctime
= 0;
1042 /* guess... choose latest ctime */
1045 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1046 switch(load_super1(&tst
, fd
, devname
)) {
1047 case 0: super
= tst
.sb
;
1048 if (bestvers
== -1 ||
1049 bestctime
< __le64_to_cpu(super
->ctime
)) {
1050 bestvers
= tst
.minor_version
;
1051 bestctime
= __le64_to_cpu(super
->ctime
);
1056 case 1: return 1; /*bad device */
1057 case 2: break; /* bad, try next */
1060 if (bestvers
!= -1) {
1062 tst
.minor_version
= bestvers
;
1065 rv
= load_super1(&tst
, fd
, devname
);
1072 if (!get_dev_size(fd
, devname
, &dsize
))
1078 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
1084 * Calculate the position of the superblock.
1085 * It is always aligned to a 4K boundary and
1086 * depending on minor_version, it can be:
1087 * 0: At least 8K, but less than 12K, from end of device
1088 * 1: At start of device
1089 * 2: 4K from start of device.
1091 switch(st
->minor_version
) {
1095 sb_offset
&= ~(4*2-1);
1107 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
1110 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1112 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
1113 devname
, strerror(errno
));
1117 super
= malloc(1024 + sizeof(bitmap_super_t
) +
1118 sizeof(struct misc_dev_info
));
1120 if (read(fd
, super
, 1024) != 1024) {
1122 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
1128 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1130 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
1131 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1136 if (__le32_to_cpu(super
->major_version
) != 1) {
1138 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
1139 devname
, __le32_to_cpu(super
->major_version
));
1143 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1145 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
1152 bsb
= (struct bitmap_super_s
*)(((char*)super
)+1024);
1154 misc
= (struct misc_dev_info
*) (bsb
+1);
1155 misc
->device_size
= dsize
;
1157 /* Now check on the bitmap superblock */
1158 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1160 /* Read the bitmap superblock and make sure it looks
1161 * valid. If it doesn't clear the bit. An --assemble --force
1162 * should get that written out.
1164 locate_bitmap1(st
, fd
);
1165 if (read(fd
, ((char*)super
)+1024, sizeof(struct bitmap_super_s
))
1166 != sizeof(struct bitmap_super_s
))
1169 uuid_from_super1(st
, uuid
);
1170 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1171 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1176 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
) & ~1);
1181 static struct supertype
*match_metadata_desc1(char *arg
)
1183 struct supertype
*st
= malloc(sizeof(*st
));
1189 if (strcmp(arg
, "1.0") == 0) {
1190 st
->minor_version
= 0;
1193 if (strcmp(arg
, "1.1") == 0) {
1194 st
->minor_version
= 1;
1197 if (strcmp(arg
, "1.2") == 0) {
1198 st
->minor_version
= 2;
1201 if (strcmp(arg
, "1") == 0 ||
1202 strcmp(arg
, "default/large") == 0) {
1203 st
->minor_version
= -1;
1211 /* find available size on device with this devsize, using
1212 * superblock type st, and reserving 'reserve' sectors for
1215 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1220 devsize
-= choose_bm_space(devsize
);
1222 switch(st
->minor_version
) {
1223 case -1: /* no specified. Now time to set default */
1224 st
->minor_version
= 0;
1228 return ((devsize
- 8*2 ) & ~(4*2-1));
1230 /* at start, 4K for superblock and possible bitmap */
1231 return devsize
- 4*2;
1233 /* 4k from start, 4K for superblock and possible bitmap */
1234 return devsize
- (4+4)*2;
1240 add_internal_bitmap1(struct supertype
*st
,
1241 int *chunkp
, int delay
, int write_behind
,
1242 unsigned long long size
,
1243 int may_change
, int major
)
1246 * If not may_change, then this is a 'Grow', and the bitmap
1247 * must fit after the superblock.
1248 * If may_change, then this is create, and we can put the bitmap
1249 * before the superblock if we like, or may move the start.
1250 * If !may_change, the bitmap MUST live at offset of 1K, until
1251 * we get a sysfs interface.
1253 * size is in sectors, chunk is in bytes !!!
1256 unsigned long long bits
;
1257 unsigned long long max_bits
;
1258 unsigned long long min_chunk
;
1260 int chunk
= *chunkp
;
1262 struct mdp_superblock_1
*sb
= st
->sb
;
1263 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + 1024);
1265 switch(st
->minor_version
) {
1267 /* either 3K after the superblock, or some amount of space
1271 /* We are creating array, so we *know* how much room has
1275 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1277 /* make it 3K after the superblock */
1282 room
= __le64_to_cpu(sb
->super_offset
)
1283 - __le64_to_cpu(sb
->data_offset
)
1284 - __le64_to_cpu(sb
->data_size
);
1285 /* remove '1 ||' when we can set offset via sysfs */
1286 if (1 || (room
< 3*2 &&
1287 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1291 offset
= 0; /* means movable offset */
1296 case 2: /* between superblock and data */
1299 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1301 room
= __le64_to_cpu(sb
->data_offset
)
1302 - __le64_to_cpu(sb
->super_offset
);
1303 if (1 || __le32_to_cpu(sb
->max_dev
) <= 384) {
1316 if (chunk
== UnSet
&& room
> 128*2)
1317 /* Limit to 128K of bitmap when chunk size not requested */
1320 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1322 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1323 bits
= (size
*512)/min_chunk
+1;
1324 while (bits
> max_bits
) {
1330 else if (chunk
< min_chunk
)
1331 return 0; /* chunk size too small */
1332 if (chunk
== 0) /* rounding problem */
1336 bits
= (size
*512) / chunk
+ 1;
1337 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +511)/512;
1341 sb
->bitmap_offset
= __cpu_to_le32(offset
);
1343 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) | 1);
1344 memset(bms
, 0, sizeof(*bms
));
1345 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1346 bms
->version
= __cpu_to_le32(major
);
1347 uuid_from_super1(st
, (int*)bms
->uuid
);
1348 bms
->chunksize
= __cpu_to_le32(chunk
);
1349 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1350 bms
->sync_size
= __cpu_to_le64(size
);
1351 bms
->write_behind
= __cpu_to_le32(write_behind
);
1358 static void locate_bitmap1(struct supertype
*st
, int fd
)
1360 unsigned long long offset
;
1361 struct mdp_superblock_1
*sb
;
1365 if (st
->ss
->load_super(st
, fd
, NULL
))
1366 return; /* no error I hope... */
1371 offset
= __le64_to_cpu(sb
->super_offset
);
1372 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
1375 lseek64(fd
, offset
<<9, 0);
1378 static int write_bitmap1(struct supertype
*st
, int fd
)
1380 struct mdp_superblock_1
*sb
= st
->sb
;
1381 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+1024);
1387 locate_bitmap1(st
, fd
);
1389 if (write(fd
, ((char*)sb
)+1024, sizeof(bitmap_super_t
)) !=
1390 sizeof(bitmap_super_t
))
1392 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1393 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1394 memset(buf
, 0xff, sizeof(buf
));
1395 while (towrite
> 0) {
1397 if (n
> sizeof(buf
))
1399 n
= write(fd
, buf
, n
);
1412 static void free_super1(struct supertype
*st
)
1419 struct superswitch super1
= {
1421 .examine_super
= examine_super1
,
1422 .brief_examine_super
= brief_examine_super1
,
1423 .export_examine_super
= export_examine_super1
,
1424 .detail_super
= detail_super1
,
1425 .brief_detail_super
= brief_detail_super1
,
1426 .export_detail_super
= export_detail_super1
,
1428 .match_home
= match_home1
,
1429 .uuid_from_super
= uuid_from_super1
,
1430 .getinfo_super
= getinfo_super1
,
1431 .update_super
= update_super1
,
1432 .init_super
= init_super1
,
1433 .add_to_super
= add_to_super1
,
1434 .store_super
= store_super1
,
1435 .write_init_super
= write_init_super1
,
1436 .compare_super
= compare_super1
,
1437 .load_super
= load_super1
,
1438 .match_metadata_desc
= match_metadata_desc1
,
1439 .avail_size
= avail_size1
,
1440 .add_internal_bitmap
= add_internal_bitmap1
,
1441 .locate_bitmap
= locate_bitmap1
,
1442 .write_bitmap
= write_bitmap1
,
1443 .free_super
= free_super1
,
1445 #if __BYTE_ORDER == BIG_ENDIAN