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(void *sbv
, char *homehost
)
147 struct mdp_superblock_1
*sb
= sbv
;
153 int l
= homehost
? strlen(homehost
) : 0;
156 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
157 printf(" Version : %02d\n", 1);
158 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
159 printf(" Array UUID : ");
160 for (i
=0; i
<16; i
++) {
161 if ((i
&3)==0 && i
!= 0) printf(":");
162 printf("%02x", sb
->set_uuid
[i
]);
165 printf(" Name : %.32s", sb
->set_name
);
166 if (l
> 0 && l
< 32 &&
167 sb
->set_name
[l
] == ':' &&
168 strncmp(sb
->set_name
, homehost
, l
) == 0)
169 printf(" (local to host %s)", homehost
);
171 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
172 printf(" Creation Time : %.24s\n", ctime(&atime
));
173 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
174 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
175 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
177 printf(" Used Dev Size : %llu%s\n",
178 (unsigned long long)sb
->data_size
,
179 human_size(sb
->data_size
<<9));
180 if (__le32_to_cpu(sb
->level
) >= 0) {
182 switch(__le32_to_cpu(sb
->level
)) {
183 case 1: ddsks
=1;break;
185 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
186 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
188 layout
= __le32_to_cpu(sb
->layout
);
189 ddsks
= __le32_to_cpu(sb
->raid_disks
)
190 / (layout
&255) / ((layout
>>8)&255);
193 printf(" Array Size : %llu%s\n",
194 ddsks
*(unsigned long long)__le64_to_cpu(sb
->size
),
195 human_size(ddsks
*__le64_to_cpu(sb
->size
)<<9));
196 if (sb
->size
!= sb
->data_size
)
197 printf(" Used Size : %llu%s\n",
198 (unsigned long long)__le64_to_cpu(sb
->size
),
199 human_size(__le64_to_cpu(sb
->size
)<<9));
202 printf(" Data Offset : %llu sectors\n",
203 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
204 printf(" Super Offset : %llu sectors\n",
205 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
206 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
207 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
208 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
209 printf(" Device UUID : ");
210 for (i
=0; i
<16; i
++) {
211 if ((i
&3)==0 && i
!= 0) printf(":");
212 printf("%02x", sb
->device_uuid
[i
]);
216 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
217 printf("Internal Bitmap : %ld sectors from superblock\n",
218 (long)__le32_to_cpu(sb
->bitmap_offset
));
220 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
221 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
222 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
223 if (__le32_to_cpu(sb
->delta_disks
)) {
224 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
225 if (__le32_to_cpu(sb
->delta_disks
))
226 printf(" (%d->%d)\n",
227 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
228 __le32_to_cpu(sb
->raid_disks
));
230 printf(" (%d->%d)\n", __le32_to_cpu(sb
->raid_disks
),
231 __le32_to_cpu(sb
->raid_disks
)+__le32_to_cpu(sb
->delta_disks
));
233 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
234 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
235 printf(" New Level : %s\n", c
?c
:"-unknown-");
237 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
238 if (__le32_to_cpu(sb
->level
) == 5) {
239 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
240 printf(" New Layout : %s\n", c
?c
:"-unknown-");
242 if (__le32_to_cpu(sb
->level
) == 10) {
243 printf(" New Layout : near=%d, %s=%d\n",
244 __le32_to_cpu(sb
->new_layout
)&255,
245 (__le32_to_cpu(sb
->new_layout
)&0x10000)?"offset":"far",
246 (__le32_to_cpu(sb
->new_layout
)>>8)&255);
249 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
250 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
255 if (sb
->devflags
& WriteMostly1
)
256 printf(" write-mostly");
260 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
261 printf(" Update Time : %.24s\n", ctime(&atime
));
263 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
264 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
266 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
267 __le32_to_cpu(calc_sb_1_csum(sb
)));
268 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
270 if (__le32_to_cpu(sb
->level
) == 5) {
271 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
272 printf(" Layout : %s\n", c
?c
:"-unknown-");
274 if (__le32_to_cpu(sb
->level
) == 10) {
275 int lo
= __le32_to_cpu(sb
->layout
);
276 printf(" Layout : near=%d, %s=%d\n",
278 (lo
&0x10000)?"offset":"far",
281 switch(__le32_to_cpu(sb
->level
)) {
287 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
290 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
295 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
296 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
297 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
299 for (d
=0; d
< i
; d
++) {
300 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
302 if (role
== 0xffff) printf("empty");
303 else if(role
== 0xfffe) printf("failed");
304 else printf("%d", role
);
307 printf(" Array State : ");
308 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
); d
++) {
312 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
313 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
315 if (i
== __le32_to_cpu(sb
->dev_number
))
320 if (cnt
> 1) printf("?");
321 else if (cnt
== 1 && me
) printf("U");
322 else if (cnt
== 1) printf("u");
326 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
327 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
331 if (faulty
) printf(" %d failed", faulty
);
336 static void brief_examine_super1(void *sbv
)
338 struct mdp_superblock_1
*sb
= sbv
;
341 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
343 nm
= strchr(sb
->set_name
, ':');
346 else if (sb
->set_name
[0])
351 printf("ARRAY /dev/md/%s level=%s metadata=1 num-devices=%d UUID=",
353 c
?c
:"-unknown-", __le32_to_cpu(sb
->raid_disks
));
354 for (i
=0; i
<16; i
++) {
355 if ((i
&3)==0 && i
!= 0) printf(":");
356 printf("%02x", sb
->set_uuid
[i
]);
359 printf(" name=%.32s", sb
->set_name
);
363 static void detail_super1(void *sbv
, char *homehost
)
365 struct mdp_superblock_1
*sb
= sbv
;
367 int l
= homehost
? strlen(homehost
) : 0;
369 printf(" Name : %.32s", sb
->set_name
);
370 if (l
> 0 && l
< 32 &&
371 sb
->set_name
[l
] == ':' &&
372 strncmp(sb
->set_name
, homehost
, l
) == 0)
373 printf(" (local to host %s)", homehost
);
374 printf("\n UUID : ");
375 for (i
=0; i
<16; i
++) {
376 if ((i
&3)==0 && i
!= 0) printf(":");
377 printf("%02x", sb
->set_uuid
[i
]);
379 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
382 static void brief_detail_super1(void *sbv
)
384 struct mdp_superblock_1
*sb
= sbv
;
388 printf(" name=%.32s", sb
->set_name
);
390 for (i
=0; i
<16; i
++) {
391 if ((i
&3)==0 && i
!= 0) printf(":");
392 printf("%02x", sb
->set_uuid
[i
]);
398 static int match_home1(void *sbv
, char *homehost
)
400 struct mdp_superblock_1
*sb
= sbv
;
401 int l
= homehost
? strlen(homehost
) : 0;
403 return (l
> 0 && l
< 32 &&
404 sb
->set_name
[l
] == ':' &&
405 strncmp(sb
->set_name
, homehost
, l
) == 0);
408 static void uuid_from_super1(int uuid
[4], void * sbv
)
410 struct mdp_superblock_1
*super
= sbv
;
411 char *cuuid
= (char*)uuid
;
414 cuuid
[i
] = super
->set_uuid
[i
];
417 static void getinfo_super1(struct mdinfo
*info
, void *sbv
)
419 struct mdp_superblock_1
*sb
= sbv
;
424 info
->array
.major_version
= 1;
425 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
426 info
->array
.patch_version
= 0;
427 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
428 info
->array
.level
= __le32_to_cpu(sb
->level
);
429 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
430 info
->array
.md_minor
= -1;
431 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
432 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
433 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
435 (__le64_to_cpu(sb
->resync_offset
) >= __le64_to_cpu(sb
->size
))
438 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
439 info
->component_size
= __le64_to_cpu(sb
->size
);
441 info
->disk
.major
= 0;
442 info
->disk
.minor
= 0;
443 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
444 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
445 __le32_to_cpu(sb
->max_dev
) > 512)
448 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
450 info
->disk
.raid_disk
= -1;
453 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
456 info
->disk
.state
= 1; /* faulty */
459 info
->disk
.state
= 6; /* active and in sync */
460 info
->disk
.raid_disk
= role
;
462 info
->events
= __le64_to_cpu(sb
->events
);
464 memcpy(info
->uuid
, sb
->set_uuid
, 16);
466 strncpy(info
->name
, sb
->set_name
, 32);
469 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
470 info
->reshape_active
= 1;
471 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
472 info
->new_level
= __le32_to_cpu(sb
->new_level
);
473 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
474 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
475 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
477 info
->reshape_active
= 0;
479 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
480 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
481 if (/*role == 0xFFFF || */role
< info
->array
.raid_disks
)
485 info
->array
.working_disks
= working
;
488 static int update_super1(struct mdinfo
*info
, void *sbv
, char *update
,
489 char *devname
, int verbose
,
490 int uuid_set
, char *homehost
)
492 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
493 * For others, the return value is ignored.
496 struct mdp_superblock_1
*sb
= sbv
;
498 if (strcmp(update
, "force-one")==0) {
499 /* Not enough devices for a working array,
500 * so bring this one up-to-date
502 if (sb
->events
!= __cpu_to_le64(info
->events
))
504 sb
->events
= __cpu_to_le64(info
->events
);
506 if (strcmp(update
, "force-array")==0) {
507 /* Degraded array and 'force' requests to
508 * maybe need to mark it 'clean'.
510 switch(__le32_to_cpu(sb
->level
)) {
511 case 5: case 4: case 6:
512 /* need to force clean */
513 if (sb
->resync_offset
!= ~0ULL)
515 sb
->resync_offset
= ~0ULL;
518 if (strcmp(update
, "assemble")==0) {
519 int d
= info
->disk
.number
;
521 if (info
->disk
.state
== 6)
522 want
= __cpu_to_le32(info
->disk
.raid_disk
);
525 if (sb
->dev_roles
[d
] != want
) {
526 sb
->dev_roles
[d
] = want
;
530 if (strcmp(update
, "grow") == 0) {
531 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
532 /* As we are just adding a spare, there is no need to
533 * make any change to the dev_roles array
536 if (strcmp(update
, "resync") == 0) {
537 /* make sure resync happens */
538 sb
->resync_offset
= 0ULL;
540 if (strcmp(update
, "uuid") == 0) {
541 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
543 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
544 struct bitmap_super_s
*bm
;
545 bm
= (struct bitmap_super_s
*)(sbv
+1024);
546 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
549 if (strcmp(update
, "homehost") == 0 &&
553 c
= strchr(sb
->set_name
, ':');
555 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
557 strncpy(info
->name
, sb
->set_name
, 32);
560 if (strcmp(update
, "name") == 0) {
561 if (info
->name
[0] == 0)
562 sprintf(info
->name
, "%d", info
->array
.md_minor
);
563 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
565 strchr(info
->name
, ':') == NULL
&&
566 strlen(homehost
)+1+strlen(info
->name
) < 32) {
567 strcpy(sb
->set_name
, homehost
);
568 strcat(sb
->set_name
, ":");
569 strcat(sb
->set_name
, info
->name
);
571 strcpy(sb
->set_name
, info
->name
);
573 if (strcmp(update
, "devicesize") == 0 &&
574 __le64_to_cpu(sb
->super_offset
) <
575 __le64_to_cpu(sb
->data_offset
)) {
576 /* set data_size to device size less data_offset */
577 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
578 (sbv
+ 1024 + sizeof(struct bitmap_super_s
));
579 printf("Size was %llu\n", __le64_to_cpu(sb
->data_size
));
580 sb
->data_size
= __cpu_to_le64(
581 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
582 printf("Size is %llu\n", __le64_to_cpu(sb
->data_size
));
584 if (strcmp(update
, "_reshape_progress")==0)
585 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
587 sb
->sb_csum
= calc_sb_1_csum(sb
);
591 static int init_super1(struct supertype
*st
, void **sbp
, mdu_array_info_t
*info
,
592 unsigned long long size
, char *name
, char *homehost
, int *uuid
)
594 struct mdp_superblock_1
*sb
= malloc(1024 + sizeof(bitmap_super_t
) +
595 sizeof(struct misc_dev_info
));
601 if (info
->major_version
== -1) {
602 /* zeroing superblock */
607 spares
= info
->working_disks
- info
->active_disks
;
608 if (info
->raid_disks
+ spares
> 384) {
609 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
610 info
->raid_disks
, spares
, 384);
614 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
615 sb
->major_version
= __cpu_to_le32(1);
620 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
622 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
623 read(rfd
, sb
->set_uuid
, 16) != 16) {
624 *(__u32
*)(sb
->set_uuid
) = random();
625 *(__u32
*)(sb
->set_uuid
+4) = random();
626 *(__u32
*)(sb
->set_uuid
+8) = random();
627 *(__u32
*)(sb
->set_uuid
+12) = random();
629 if (rfd
>= 0) close(rfd
);
632 if (name
== NULL
|| *name
== 0) {
633 sprintf(defname
, "%d", info
->md_minor
);
636 memset(sb
->set_name
, 0, 32);
638 strchr(name
, ':')== NULL
&&
639 strlen(homehost
)+1+strlen(name
) < 32) {
640 strcpy(sb
->set_name
, homehost
);
641 strcat(sb
->set_name
, ":");
642 strcat(sb
->set_name
, name
);
644 strcpy(sb
->set_name
, name
);
646 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
647 sb
->level
= __cpu_to_le32(info
->level
);
648 sb
->layout
= __cpu_to_le32(info
->layout
);
649 sb
->size
= __cpu_to_le64(size
*2ULL);
650 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
651 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
653 sb
->data_offset
= __cpu_to_le64(0);
654 sb
->data_size
= __cpu_to_le64(0);
655 sb
->super_offset
= __cpu_to_le64(0);
656 sb
->recovery_offset
= __cpu_to_le64(0);
658 sb
->utime
= sb
->ctime
;
659 sb
->events
= __cpu_to_le64(1);
660 if (info
->state
& (1<<MD_SB_CLEAN
))
661 sb
->resync_offset
= ~0ULL;
663 sb
->resync_offset
= 0;
664 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
665 sizeof(sb
->dev_roles
[0]));
666 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
668 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
674 /* Add a device to the superblock being created */
675 static void add_to_super1(void *sbv
, mdu_disk_info_t
*dk
)
677 struct mdp_superblock_1
*sb
= sbv
;
678 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
679 if ((dk
->state
& 6) == 6) /* active, sync */
680 *rp
= __cpu_to_le16(dk
->raid_disk
);
681 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
687 static void locate_bitmap1(struct supertype
*st
, int fd
, void *sbv
);
689 static int store_super1(struct supertype
*st
, int fd
, void *sbv
)
691 struct mdp_superblock_1
*sb
= sbv
;
692 unsigned long long sb_offset
;
694 unsigned long long dsize
;
696 if (!get_dev_size(fd
, NULL
, &dsize
))
705 * Calculate the position of the superblock.
706 * It is always aligned to a 4K boundary and
707 * depending on minor_version, it can be:
708 * 0: At least 8K, but less than 12K, from end of device
709 * 1: At start of device
710 * 2: 4K from start of device.
712 switch(st
->minor_version
) {
716 sb_offset
&= ~(4*2-1);
730 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
731 0 != __le64_to_cpu(sb
->super_offset
)
733 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
737 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
740 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
742 if (write(fd
, sb
, sbsize
) != sbsize
)
745 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
746 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
748 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
749 locate_bitmap1(st
, fd
, sbv
);
750 if (write(fd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
758 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
);
760 static int write_init_super1(struct supertype
*st
, void *sbv
,
761 mdu_disk_info_t
*dinfo
, char *devname
)
763 struct mdp_superblock_1
*sb
= sbv
;
765 int fd
= open(devname
, O_RDWR
| O_EXCL
);
771 unsigned long long dsize
, array_size
;
776 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
781 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
782 if (dinfo
->state
& (1<<MD_DISK_WRITEMOSTLY
))
783 sb
->devflags
|= __cpu_to_le32(WriteMostly1
);
785 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
786 read(rfd
, sb
->device_uuid
, 16) != 16) {
787 *(__u32
*)(sb
->device_uuid
) = random();
788 *(__u32
*)(sb
->device_uuid
+4) = random();
789 *(__u32
*)(sb
->device_uuid
+8) = random();
790 *(__u32
*)(sb
->device_uuid
+12) = random();
792 if (rfd
>= 0) close(rfd
);
795 if (load_super1(st
, fd
, &refsbv
, NULL
)==0) {
796 struct mdp_superblock_1
*refsb
= refsbv
;
798 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
799 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
800 /* same array, so preserve events and dev_number */
801 sb
->events
= refsb
->events
;
802 /* bugs in 2.6.17 and earlier mean the dev_number
803 * chosen in Manage must be preserved
805 if (get_linux_version() >= 2006018)
806 sb
->dev_number
= refsb
->dev_number
;
811 if (!get_dev_size(fd
, NULL
, &dsize
))
822 * Calculate the position of the superblock.
823 * It is always aligned to a 4K boundary and
824 * depending on minor_version, it can be:
825 * 0: At least 8K, but less than 12K, from end of device
826 * 1: At start of device
827 * 2: 4K from start of device.
828 * Depending on the array size, we might leave extra space
831 array_size
= __le64_to_cpu(sb
->size
);
832 /* work out how much space we left of a bitmap */
833 if (array_size
>= 200*1024*1024*2)
835 else if (array_size
> 8*1024*1024*2)
840 switch(st
->minor_version
) {
844 sb_offset
&= ~(4*2-1);
845 sb
->super_offset
= __cpu_to_le64(sb_offset
);
846 sb
->data_offset
= __cpu_to_le64(0);
847 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
850 sb
->super_offset
= __cpu_to_le64(0);
851 sb
->data_offset
= __cpu_to_le64(bm_space
+ 4*2);
852 sb
->data_size
= __cpu_to_le64(dsize
- bm_space
- 4*2);
856 if (dsize
- 4*2 - 64*2 >= array_size
&& array_size
> 8*1024*1024*2)
860 sb
->super_offset
= __cpu_to_le64(4*2);
861 sb
->data_offset
= __cpu_to_le64(4*2 + 4*2 + bm_space
);
862 sb
->data_size
= __cpu_to_le64(dsize
- 4*2 - 4*2 - bm_space
);
869 sb
->sb_csum
= calc_sb_1_csum(sb
);
870 rv
= store_super1(st
, fd
, sb
);
872 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
874 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
875 rv
= st
->ss
->write_bitmap(st
, fd
, sbv
);
880 static int compare_super1(void **firstp
, void *secondv
)
884 * 0 same, or first was empty, and second was copied
885 * 1 second had wrong number
889 struct mdp_superblock_1
*first
= *firstp
;
890 struct mdp_superblock_1
*second
= secondv
;
892 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
894 if (second
->major_version
!= __cpu_to_le32(1))
898 first
= malloc(1024+sizeof(bitmap_super_t
) +
899 sizeof(struct misc_dev_info
));
900 memcpy(first
, second
, 1024+sizeof(bitmap_super_t
) +
901 sizeof(struct misc_dev_info
));
905 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
908 if (first
->ctime
!= second
->ctime
||
909 first
->level
!= second
->level
||
910 first
->layout
!= second
->layout
||
911 first
->size
!= second
->size
||
912 first
->chunksize
!= second
->chunksize
||
913 first
->raid_disks
!= second
->raid_disks
)
918 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
)
920 unsigned long long dsize
;
921 unsigned long long sb_offset
;
922 struct mdp_superblock_1
*super
;
924 struct bitmap_super_s
*bsb
;
925 struct misc_dev_info
*misc
;
928 if (st
->ss
== NULL
) {
931 /* guess... choose latest ctime */
933 for (st
->minor_version
= 0; st
->minor_version
<= 2 ; st
->minor_version
++) {
934 switch(load_super1(st
, fd
, sbp
, devname
)) {
935 case 0: super
= *sbp
;
936 if (bestvers
== -1 ||
937 bestctime
< __le64_to_cpu(super
->ctime
)) {
938 bestvers
= st
->minor_version
;
939 bestctime
= __le64_to_cpu(super
->ctime
);
944 case 1: st
->ss
= NULL
; return 1; /*bad device */
945 case 2: break; /* bad, try next */
948 if (bestvers
!= -1) {
950 st
->minor_version
= bestvers
;
953 rv
= load_super1(st
, fd
, sbp
, devname
);
954 if (rv
) st
->ss
= NULL
;
960 if (!get_dev_size(fd
, devname
, &dsize
))
966 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
972 * Calculate the position of the superblock.
973 * It is always aligned to a 4K boundary and
974 * depending on minor_version, it can be:
975 * 0: At least 8K, but less than 12K, from end of device
976 * 1: At start of device
977 * 2: 4K from start of device.
979 switch(st
->minor_version
) {
983 sb_offset
&= ~(4*2-1);
995 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
998 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1000 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
1001 devname
, strerror(errno
));
1005 super
= malloc(1024 + sizeof(bitmap_super_t
) +
1006 sizeof(struct misc_dev_info
));
1008 if (read(fd
, super
, 1024) != 1024) {
1010 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
1016 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1018 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
1019 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1024 if (__le32_to_cpu(super
->major_version
) != 1) {
1026 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
1027 devname
, __le32_to_cpu(super
->major_version
));
1031 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1033 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
1040 bsb
= (struct bitmap_super_s
*)(((char*)super
)+1024);
1042 misc
= (struct misc_dev_info
*) (bsb
+1);
1043 misc
->device_size
= dsize
;
1045 /* Now check on the bitmap superblock */
1046 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1048 /* Read the bitmap superblock and make sure it looks
1049 * valid. If it doesn't clear the bit. An --assemble --force
1050 * should get that written out.
1052 locate_bitmap1(st
, fd
, super
);
1053 if (read(fd
, ((char*)super
)+1024, sizeof(struct bitmap_super_s
))
1054 != sizeof(struct bitmap_super_s
))
1057 uuid_from_super1(uuid
, super
);
1058 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1059 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1064 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
) & ~1);
1069 static struct supertype
*match_metadata_desc1(char *arg
)
1071 struct supertype
*st
= malloc(sizeof(*st
));
1076 if (strcmp(arg
, "1") == 0 ||
1077 strcmp(arg
, "1.0") == 0 ||
1078 strcmp(arg
, "default/large") == 0) {
1079 st
->minor_version
= 0;
1082 if (strcmp(arg
, "1.1") == 0) {
1083 st
->minor_version
= 1;
1086 if (strcmp(arg
, "1.2") == 0) {
1087 st
->minor_version
= 2;
1095 /* find available size on device with this devsize, using
1096 * superblock type st, and reserving 'reserve' sectors for
1099 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1104 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1105 * if bigger than 200Gig, save 128k
1107 if (devsize
-64*2 >= 200*1024*1024*2)
1109 else if (devsize
>= 8*1024*1024*2)
1112 switch(st
->minor_version
) {
1115 return ((devsize
- 8*2 ) & ~(4*2-1));
1117 /* at start, 4K for superblock and possible bitmap */
1118 return devsize
- 4*2;
1120 /* 4k from start, 4K for superblock and possible bitmap */
1121 return devsize
- (4+4)*2;
1127 add_internal_bitmap1(struct supertype
*st
, void *sbv
,
1128 int *chunkp
, int delay
, int write_behind
,
1129 unsigned long long size
,
1130 int may_change
, int major
)
1133 * If not may_change, then this is a 'Grow', and the bitmap
1134 * must fit after the superblock.
1135 * If may_change, then this is create, and we can put the bitmap
1136 * before the superblock if we like, or may move the start.
1137 * If !may_change, the bitmap MUST live at offset of 1K, until
1138 * we get a sysfs interface.
1140 * size is in sectors, chunk is in bytes !!!
1143 unsigned long long bits
;
1144 unsigned long long max_bits
;
1145 unsigned long long min_chunk
;
1147 int chunk
= *chunkp
;
1149 struct mdp_superblock_1
*sb
= sbv
;
1150 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + 1024);
1152 switch(st
->minor_version
) {
1154 /* either 3K after the superblock, or some amount of space
1158 /* We are creating array, so we *know* how much room has
1162 if (__le64_to_cpu(sb
->size
) >= 200*1024*1024*2)
1164 else if (__le64_to_cpu(sb
->size
) > 8*1024*1024*2)
1171 room
= __le64_to_cpu(sb
->super_offset
)
1172 - __le64_to_cpu(sb
->data_offset
)
1173 - __le64_to_cpu(sb
->data_size
);
1174 /* remove '1 ||' when we can set offset via sysfs */
1175 if (1 || (room
< 3*2 &&
1176 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1180 offset
= 0; /* means movable offset */
1185 case 2: /* between superblock and data */
1188 if (__le64_to_cpu(sb
->size
) >= 200*1024*1024*2)
1190 else if (__le64_to_cpu(sb
->size
) > 8*1024*1024*2)
1195 room
= __le64_to_cpu(sb
->data_offset
)
1196 - __le64_to_cpu(sb
->super_offset
);
1197 if (1 || __le32_to_cpu(sb
->max_dev
) <= 384) {
1210 if (chunk
== UnSet
&& room
> 128*2)
1211 /* Limit to 128K of bitmap when chunk size not requested */
1214 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1216 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1217 bits
= (size
*512)/min_chunk
+1;
1218 while (bits
> max_bits
) {
1224 else if (chunk
< min_chunk
)
1225 return 0; /* chunk size too small */
1226 if (chunk
== 0) /* rounding problem */
1230 bits
= (size
*512) / chunk
+ 1;
1231 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +511)/512;
1235 sb
->bitmap_offset
= __cpu_to_le32(offset
);
1237 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) | 1);
1238 memset(bms
, 0, sizeof(*bms
));
1239 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1240 bms
->version
= __cpu_to_le32(major
);
1241 uuid_from_super1((int*)bms
->uuid
, sb
);
1242 bms
->chunksize
= __cpu_to_le32(chunk
);
1243 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1244 bms
->sync_size
= __cpu_to_le64(size
);
1245 bms
->write_behind
= __cpu_to_le32(write_behind
);
1252 static void locate_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
1254 unsigned long long offset
;
1255 struct mdp_superblock_1
*sb
;
1259 if (st
->ss
->load_super(st
, fd
, &sbv
, NULL
))
1260 return; /* no error I hope... */
1265 offset
= __le64_to_cpu(sb
->super_offset
);
1266 offset
+= (long) __le32_to_cpu(sb
->bitmap_offset
);
1269 lseek64(fd
, offset
<<9, 0);
1272 static int write_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
1274 struct mdp_superblock_1
*sb
= sbv
;
1275 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+1024);
1281 locate_bitmap1(st
, fd
, sbv
);
1283 if (write(fd
, ((char*)sb
)+1024, sizeof(bitmap_super_t
)) !=
1284 sizeof(bitmap_super_t
))
1286 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1287 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1288 memset(buf
, 0xff, sizeof(buf
));
1289 while (towrite
> 0) {
1291 if (n
> sizeof(buf
))
1293 n
= write(fd
, buf
, n
);
1306 struct superswitch super1
= {
1308 .examine_super
= examine_super1
,
1309 .brief_examine_super
= brief_examine_super1
,
1310 .detail_super
= detail_super1
,
1311 .brief_detail_super
= brief_detail_super1
,
1313 .match_home
= match_home1
,
1314 .uuid_from_super
= uuid_from_super1
,
1315 .getinfo_super
= getinfo_super1
,
1316 .update_super
= update_super1
,
1317 .init_super
= init_super1
,
1318 .add_to_super
= add_to_super1
,
1319 .store_super
= store_super1
,
1320 .write_init_super
= write_init_super1
,
1321 .compare_super
= compare_super1
,
1322 .load_super
= load_super1
,
1323 .match_metadata_desc
= match_metadata_desc1
,
1324 .avail_size
= avail_size1
,
1325 .add_internal_bitmap
= add_internal_bitmap1
,
1326 .locate_bitmap
= locate_bitmap1
,
1327 .write_bitmap
= write_bitmap1
,
1329 #if __BYTE_ORDER == BIG_ENDIAN