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 /* feature_map bits */
98 #define MD_FEATURE_BITMAP_OFFSET 1
99 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
102 #define MD_FEATURE_RESHAPE_ACTIVE 4
104 #define MD_FEATURE_ALL (1|2|4)
107 #define offsetof(t,f) ((int)&(((t*)0)->f))
109 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
111 unsigned int disk_csum
, csum
;
112 unsigned long long newcsum
;
113 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
114 unsigned int *isuper
= (unsigned int*)sb
;
117 /* make sure I can count... */
118 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
119 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
120 sizeof(struct mdp_superblock_1
) != 256) {
121 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
124 disk_csum
= sb
->sb_csum
;
127 for (i
=0; size
>=4; size
-= 4 )
128 newcsum
+= __le32_to_cpu(*isuper
++);
131 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
133 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
134 sb
->sb_csum
= disk_csum
;
139 static void examine_super1(void *sbv
, char *homehost
)
141 struct mdp_superblock_1
*sb
= sbv
;
147 int l
= homehost
? strlen(homehost
) : 0;
149 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
150 printf(" Version : %02d\n", 1);
151 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
152 printf(" Array UUID : ");
153 for (i
=0; i
<16; i
++) {
154 if ((i
&3)==0 && i
!= 0) printf(":");
155 printf("%02x", sb
->set_uuid
[i
]);
158 printf(" Name : %.32s", sb
->set_name
);
159 if (l
> 0 && l
< 32 &&
160 sb
->set_name
[l
] == ':' &&
161 strncmp(sb
->set_name
, homehost
, l
) == 0)
162 printf(" (local to host %s)", homehost
);
164 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
165 printf(" Creation Time : %.24s\n", ctime(&atime
));
166 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
167 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
168 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
170 printf(" Device Size : %llu%s\n", (unsigned long long)sb
->data_size
, human_size(sb
->data_size
<<9));
171 if (__le32_to_cpu(sb
->level
) >= 0) {
173 switch(__le32_to_cpu(sb
->level
)) {
174 case 1: ddsks
=1;break;
176 case 5: ddsks
= sb
->raid_disks
-1; break;
177 case 6: ddsks
= sb
->raid_disks
-2; break;
178 case 10: ddsks
= sb
->raid_disks
/ (sb
->layout
&255) / ((sb
->layout
>>8)&255);
181 printf(" Array Size : %llu%s\n", ddsks
*(unsigned long long)sb
->size
, human_size(ddsks
*sb
->size
<<9));
182 if (sb
->size
!= sb
->data_size
)
183 printf(" Used Size : %llu%s\n", (unsigned long long)sb
->size
, human_size(sb
->size
<<9));
186 printf(" Data Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->data_offset
));
187 if (sb
->super_offset
)
188 printf(" Super Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->super_offset
));
189 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
190 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
191 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
192 printf(" Device UUID : ");
193 for (i
=0; i
<16; i
++) {
194 if ((i
&3)==0 && i
!= 0) printf(":");
195 printf("%02x", sb
->device_uuid
[i
]);
199 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
200 printf("Internal Bitmap : %ld sectors from superblock\n",
201 (long)__le32_to_cpu(sb
->bitmap_offset
));
203 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
204 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
205 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
206 if (__le32_to_cpu(sb
->delta_disks
)) {
207 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
208 if (__le32_to_cpu(sb
->delta_disks
))
209 printf(" (%d->%d)\n",
210 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
211 __le32_to_cpu(sb
->raid_disks
));
213 printf(" (%d->%d)\n", __le32_to_cpu(sb
->raid_disks
),
214 __le32_to_cpu(sb
->raid_disks
)+__le32_to_cpu(sb
->delta_disks
));
216 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
217 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
218 printf(" New Level : %s\n", c
?c
:"-unknown-");
220 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
221 if (__le32_to_cpu(sb
->level
) == 5) {
222 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
223 printf(" New Layout : %s\n", c
?c
:"-unknown-");
225 if (__le32_to_cpu(sb
->level
) == 10) {
226 printf(" New Layout : near=%d, %s=%d\n",
227 __le32_to_cpu(sb
->new_layout
)&255,
228 (__le32_to_cpu(sb
->new_layout
)&0x10000)?"offset":"far",
229 (__le32_to_cpu(sb
->new_layout
)>>8)&255);
232 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
233 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
238 if (sb
->devflags
& WriteMostly1
)
239 printf(" write-mostly");
243 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
244 printf(" Update Time : %.24s\n", ctime(&atime
));
246 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
247 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
249 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
250 __le32_to_cpu(calc_sb_1_csum(sb
)));
251 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
253 if (__le32_to_cpu(sb
->level
) == 5) {
254 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
255 printf(" Layout : %s\n", c
?c
:"-unknown-");
257 if (__le32_to_cpu(sb
->level
) == 10) {
258 int lo
= __le32_to_cpu(sb
->layout
);
259 printf(" Layout : near=%d, %s=%d\n",
261 (lo
&0x10000)?"offset":"far",
264 switch(__le32_to_cpu(sb
->level
)) {
270 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
273 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
278 printf(" Array State : ");
279 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
); d
++) {
283 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
284 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
286 if (i
== __le32_to_cpu(sb
->dev_number
))
291 if (cnt
> 1) printf("?");
292 else if (cnt
== 1 && me
) printf("U");
293 else if (cnt
== 1) printf("u");
297 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
298 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
302 if (faulty
) printf(" %d failed", faulty
);
307 static void brief_examine_super1(void *sbv
)
309 struct mdp_superblock_1
*sb
= sbv
;
312 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
314 printf("ARRAY /dev/?? level=%s metadata=1 num-devices=%d UUID=",
315 c
?c
:"-unknown-", sb
->raid_disks
);
316 for (i
=0; i
<16; i
++) {
317 printf("%02x", sb
->set_uuid
[i
]);
318 if ((i
&3)==0 && i
!= 0) printf(":");
321 printf(" name=%.32s", sb
->set_name
);
325 static void detail_super1(void *sbv
, char *homehost
)
327 struct mdp_superblock_1
*sb
= sbv
;
329 int l
= homehost
? strlen(homehost
) : 0;
331 printf(" Name : %.32s", sb
->set_name
);
332 if (l
> 0 && l
< 32 &&
333 sb
->set_name
[l
] == ':' &&
334 strncmp(sb
->set_name
, homehost
, l
) == 0)
335 printf(" (local to host %s)", homehost
);
336 printf("\n UUID : ");
337 for (i
=0; i
<16; i
++) {
338 if ((i
&3)==0 && i
!= 0) printf(":");
339 printf("%02x", sb
->set_uuid
[i
]);
341 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
344 static void brief_detail_super1(void *sbv
)
346 struct mdp_superblock_1
*sb
= sbv
;
350 printf(" name=%.32s", sb
->set_name
);
352 for (i
=0; i
<16; i
++) {
353 if ((i
&3)==0 && i
!= 0) printf(":");
354 printf("%02x", sb
->set_uuid
[i
]);
360 static int match_home1(void *sbv
, char *homehost
)
362 struct mdp_superblock_1
*sb
= sbv
;
363 int l
= homehost
? strlen(homehost
) : 0;
365 return (l
> 0 && l
< 32 &&
366 sb
->set_name
[l
] == ':' &&
367 strncmp(sb
->set_name
, homehost
, l
) == 0);
370 static void uuid_from_super1(int uuid
[4], void * sbv
)
372 struct mdp_superblock_1
*super
= sbv
;
373 char *cuuid
= (char*)uuid
;
376 cuuid
[i
] = super
->set_uuid
[i
];
379 static void getinfo_super1(struct mdinfo
*info
, void *sbv
)
381 struct mdp_superblock_1
*sb
= sbv
;
386 info
->array
.major_version
= 1;
387 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
388 info
->array
.patch_version
= 0;
389 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
390 info
->array
.level
= __le32_to_cpu(sb
->level
);
391 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
392 info
->array
.md_minor
= -1;
393 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
394 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
395 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
397 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
398 info
->component_size
= __le64_to_cpu(sb
->size
);
400 info
->disk
.major
= 0;
401 info
->disk
.minor
= 0;
402 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
403 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
404 __le32_to_cpu(sb
->max_dev
) > 512)
407 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
409 info
->disk
.raid_disk
= -1;
412 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
415 info
->disk
.state
= 1; /* faulty */
418 info
->disk
.state
= 6; /* active and in sync */
419 info
->disk
.raid_disk
= role
;
421 info
->events
= __le64_to_cpu(sb
->events
);
423 memcpy(info
->uuid
, sb
->set_uuid
, 16);
425 strncpy(info
->name
, sb
->set_name
, 32);
428 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
429 info
->reshape_active
= 1;
430 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
431 info
->new_level
= __le32_to_cpu(sb
->new_level
);
432 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
433 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
434 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
436 info
->reshape_active
= 0;
438 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
439 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
440 if (/*role == 0xFFFF || */role
< info
->array
.raid_disks
)
444 info
->array
.working_disks
= working
;
447 static int update_super1(struct mdinfo
*info
, void *sbv
, char *update
,
448 char *devname
, int verbose
,
449 int uuid_set
, char *homehost
)
451 /* NOTE: for 'assemble' and 'force' we need to return non-zero if any change was made.
452 * For others, the return value is ignored.
455 struct mdp_superblock_1
*sb
= sbv
;
457 if (strcmp(update
, "force")==0) {
458 if (sb
->events
!= __cpu_to_le64(info
->events
))
460 sb
->events
= __cpu_to_le64(info
->events
);
461 switch(__le32_to_cpu(sb
->level
)) {
462 case 5: case 4: case 6:
463 /* need to force clean */
464 if (sb
->resync_offset
!= ~0ULL)
466 sb
->resync_offset
= ~0ULL;
469 if (strcmp(update
, "assemble")==0) {
470 int d
= info
->disk
.number
;
472 if (info
->disk
.state
== 6)
473 want
= __cpu_to_le32(info
->disk
.raid_disk
);
476 if (sb
->dev_roles
[d
] != want
) {
477 sb
->dev_roles
[d
] = want
;
482 if (strcmp(update
, "newdev") == 0) {
483 int d
= info
->disk
.number
;
484 memset(&sb
->disks
[d
], 0, sizeof(sb
->disks
[d
]));
485 sb
->disks
[d
].number
= d
;
486 sb
->disks
[d
].major
= info
->disk
.major
;
487 sb
->disks
[d
].minor
= info
->disk
.minor
;
488 sb
->disks
[d
].raid_disk
= info
->disk
.raid_disk
;
489 sb
->disks
[d
].state
= info
->disk
.state
;
490 sb
->this_disk
= sb
->disks
[d
];
493 if (strcmp(update
, "grow") == 0) {
494 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
497 if (strcmp(update
, "resync") == 0) {
498 /* make sure resync happens */
499 sb
->resync_offset
= ~0ULL;
501 if (strcmp(update
, "uuid") == 0) {
502 memcpy(sb
->set_uuid
, info
->uuid
, 16);
503 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
504 struct bitmap_super_s
*bm
;
505 bm
= (struct bitmap_super_s
*)(sbv
+1024);
506 memcpy(bm
->uuid
, info
->uuid
, 16);
509 if (strcmp(update
, "homehost") == 0 &&
513 c
= strchr(sb
->set_name
, ':');
515 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
517 strncpy(info
->name
, sb
->set_name
, 32);
520 if (strcmp(update
, "name") == 0) {
521 if (info
->name
[0] == 0)
522 sprintf(info
->name
, "%d", info
->array
.md_minor
);
523 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
525 strchr(info
->name
, ':') == NULL
&&
526 strlen(homehost
)+1+strlen(info
->name
) < 32) {
527 strcpy(sb
->set_name
, homehost
);
528 strcat(sb
->set_name
, ":");
529 strcat(sb
->set_name
, info
->name
);
531 strcpy(sb
->set_name
, info
->name
);
533 if (strcmp(update
, "_reshape_progress")==0)
534 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
536 sb
->sb_csum
= calc_sb_1_csum(sb
);
541 static __u64
event_super1(void *sbv
)
543 struct mdp_superblock_1
*sb
= sbv
;
544 return __le64_to_cpu(sb
->events
);
547 static int init_super1(struct supertype
*st
, void **sbp
, mdu_array_info_t
*info
,
548 unsigned long long size
, char *name
, char *homehost
)
550 struct mdp_superblock_1
*sb
= malloc(1024 + sizeof(bitmap_super_t
));
556 if (info
->major_version
== -1) {
557 /* zeroing superblock */
562 spares
= info
->working_disks
- info
->active_disks
;
563 if (info
->raid_disks
+ spares
> 384) {
564 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
565 info
->raid_disks
, spares
, 384);
569 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
570 sb
->major_version
= __cpu_to_le32(1);
574 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
575 read(rfd
, sb
->set_uuid
, 16) != 16) {
576 *(__u32
*)(sb
->set_uuid
) = random();
577 *(__u32
*)(sb
->set_uuid
+4) = random();
578 *(__u32
*)(sb
->set_uuid
+8) = random();
579 *(__u32
*)(sb
->set_uuid
+12) = random();
581 if (rfd
>= 0) close(rfd
);
583 if (name
== NULL
|| *name
== 0) {
584 sprintf(defname
, "%d", info
->md_minor
);
587 memset(sb
->set_name
, 0, 32);
589 strchr(name
, ':')== NULL
&&
590 strlen(homehost
)+1+strlen(name
) < 32) {
591 strcpy(sb
->set_name
, homehost
);
592 strcat(sb
->set_name
, ":");
593 strcat(sb
->set_name
, name
);
595 strcpy(sb
->set_name
, name
);
597 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
598 sb
->level
= __cpu_to_le32(info
->level
);
599 sb
->layout
= __cpu_to_le32(info
->layout
);
600 sb
->size
= __cpu_to_le64(size
*2ULL);
601 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
602 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
604 sb
->data_offset
= __cpu_to_le64(0);
605 sb
->data_size
= __cpu_to_le64(0);
606 sb
->super_offset
= __cpu_to_le64(0);
607 sb
->recovery_offset
= __cpu_to_le64(0);
609 sb
->utime
= sb
->ctime
;
610 sb
->events
= __cpu_to_le64(1);
611 if (info
->state
& (1<<MD_SB_CLEAN
))
612 sb
->resync_offset
= ~0ULL;
614 sb
->resync_offset
= 0;
615 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
616 sizeof(sb
->dev_roles
[0]));
617 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
619 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
625 /* Add a device to the superblock being created */
626 static void add_to_super1(void *sbv
, mdu_disk_info_t
*dk
)
628 struct mdp_superblock_1
*sb
= sbv
;
629 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
630 if ((dk
->state
& 6) == 6) /* active, sync */
631 *rp
= __cpu_to_le16(dk
->raid_disk
);
632 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
638 static void locate_bitmap1(struct supertype
*st
, int fd
, void *sbv
);
640 static int store_super1(struct supertype
*st
, int fd
, void *sbv
)
642 struct mdp_superblock_1
*sb
= sbv
;
643 unsigned long long sb_offset
;
646 unsigned long long dsize
;
649 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
652 if (ioctl(fd
, BLKGETSIZE
, &size
))
655 dsize
= (unsigned long long)size
;
663 * Calculate the position of the superblock.
664 * It is always aligned to a 4K boundary and
665 * depending on minor_version, it can be:
666 * 0: At least 8K, but less than 12K, from end of device
667 * 1: At start of device
668 * 2: 4K from start of device.
670 switch(st
->minor_version
) {
674 sb_offset
&= ~(4*2-1);
688 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
689 0 != __le64_to_cpu(sb
->super_offset
)
691 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
695 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
698 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
700 if (write(fd
, sb
, sbsize
) != sbsize
)
703 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
704 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
706 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
707 locate_bitmap1(st
, fd
, sbv
);
708 write(fd
, bm
, sizeof(*bm
));
715 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
);
717 static int write_init_super1(struct supertype
*st
, void *sbv
,
718 mdu_disk_info_t
*dinfo
, char *devname
)
720 struct mdp_superblock_1
*sb
= sbv
;
722 int fd
= open(devname
, O_RDWR
| O_EXCL
);
726 unsigned long size
, space
;
727 unsigned long long dsize
, array_size
;
732 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
737 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
738 if (dinfo
->state
& (1<<MD_DISK_WRITEMOSTLY
))
739 sb
->devflags
|= WriteMostly1
;
741 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
742 read(rfd
, sb
->device_uuid
, 16) != 16) {
743 *(__u32
*)(sb
->device_uuid
) = random();
744 *(__u32
*)(sb
->device_uuid
+4) = random();
745 *(__u32
*)(sb
->device_uuid
+8) = random();
746 *(__u32
*)(sb
->device_uuid
+12) = random();
748 if (rfd
>= 0) close(rfd
);
751 if (load_super1(st
, fd
, &refsbv
, NULL
)==0) {
752 struct mdp_superblock_1
*refsb
= refsbv
;
754 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
755 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
756 /* same array, so preserve events and dev_number */
757 sb
->events
= refsb
->events
;
758 sb
->dev_number
= refsb
->dev_number
;
764 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
767 if (ioctl(fd
, BLKGETSIZE
, &size
))
781 * Calculate the position of the superblock.
782 * It is always aligned to a 4K boundary and
783 * depending on minor_version, it can be:
784 * 0: At least 8K, but less than 12K, from end of device
785 * 1: At start of device
786 * 2: 4K from start of device.
787 * Depending on the array size, we might leave extra space
790 array_size
= __le64_to_cpu(sb
->size
);
791 switch(st
->minor_version
) {
795 sb_offset
&= ~(4*2-1);
796 sb
->super_offset
= __cpu_to_le64(sb_offset
);
797 sb
->data_offset
= __cpu_to_le64(0);
798 if (sb_offset
-64*2 >= array_size
&& array_size
> 8*1024*1024*2)
799 sb
->data_size
= __cpu_to_le64(sb_offset
-64*2);
801 sb
->data_size
= __cpu_to_le64(sb_offset
);
804 sb
->super_offset
= __cpu_to_le64(0);
805 if (dsize
- 64*2 >= array_size
&& array_size
> 8*1024*1024*2)
809 sb
->data_offset
= __cpu_to_le64(space
); /* leave space for super and bitmap */
810 sb
->data_size
= __cpu_to_le64(dsize
- space
);
814 if (dsize
- 4*2 - 64*2 >= array_size
&& array_size
> 8*1024*1024*2)
818 sb
->super_offset
= __cpu_to_le64(sb_offset
);
819 sb
->data_offset
= __cpu_to_le64(sb_offset
+space
);
820 sb
->data_size
= __cpu_to_le64(dsize
- 4*2 - space
);
827 sb
->sb_csum
= calc_sb_1_csum(sb
);
828 rv
= store_super1(st
, fd
, sb
);
830 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
832 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
833 rv
= st
->ss
->write_bitmap(st
, fd
, sbv
);
838 static int compare_super1(void **firstp
, void *secondv
)
842 * 0 same, or first was empty, and second was copied
843 * 1 second had wrong number
847 struct mdp_superblock_1
*first
= *firstp
;
848 struct mdp_superblock_1
*second
= secondv
;
850 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
852 if (second
->major_version
!= __cpu_to_le32(1))
856 first
= malloc(1024+sizeof(bitmap_super_t
));
857 memcpy(first
, second
, 1024+sizeof(bitmap_super_t
));
861 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
864 if (first
->ctime
!= second
->ctime
||
865 first
->level
!= second
->level
||
866 first
->layout
!= second
->layout
||
867 first
->size
!= second
->size
||
868 first
->chunksize
!= second
->chunksize
||
869 first
->raid_disks
!= second
->raid_disks
)
874 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
)
877 unsigned long long dsize
;
878 unsigned long long sb_offset
;
879 struct mdp_superblock_1
*super
;
881 struct bitmap_super_s
*bsb
;
884 if (st
->ss
== NULL
) {
887 /* guess... choose latest ctime */
889 for (st
->minor_version
= 0; st
->minor_version
<= 2 ; st
->minor_version
++) {
890 switch(load_super1(st
, fd
, sbp
, devname
)) {
891 case 0: super
= *sbp
;
892 if (bestvers
== -1 ||
893 bestctime
< __le64_to_cpu(super
->ctime
)) {
894 bestvers
= st
->minor_version
;
895 bestctime
= __le64_to_cpu(super
->ctime
);
900 case 1: st
->ss
= NULL
; return 1; /*bad device */
901 case 2: break; /* bad, try next */
904 if (bestvers
!= -1) {
906 st
->minor_version
= bestvers
;
909 rv
= load_super1(st
, fd
, sbp
, devname
);
910 if (rv
) st
->ss
= NULL
;
917 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
920 if (ioctl(fd
, BLKGETSIZE
, &size
)) {
922 fprintf(stderr
, Name
": cannot find device size for %s: %s\n",
923 devname
, strerror(errno
));
932 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
938 * Calculate the position of the superblock.
939 * It is always aligned to a 4K boundary and
940 * depending on minor_version, it can be:
941 * 0: At least 8K, but less than 12K, from end of device
942 * 1: At start of device
943 * 2: 4K from start of device.
945 switch(st
->minor_version
) {
949 sb_offset
&= ~(4*2-1);
961 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
964 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
966 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
967 devname
, strerror(errno
));
971 super
= malloc(1024 + sizeof(bitmap_super_t
));
973 if (read(fd
, super
, 1024) != 1024) {
975 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
981 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
983 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
984 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
989 if (__le32_to_cpu(super
->major_version
) != 1) {
991 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
992 devname
, __le32_to_cpu(super
->major_version
));
996 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
998 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
1005 /* Now check on the bitmap superblock */
1006 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1008 /* Read the bitmap superblock and make sure it looks
1009 * valid. If it doesn't clear the bit. An --assemble --force
1010 * should get that written out.
1012 locate_bitmap1(st
, fd
, super
);
1013 if (read(fd
, ((char*)super
)+1024, sizeof(struct bitmap_super_s
))
1014 != sizeof(struct bitmap_super_s
))
1017 uuid_from_super1(uuid
, super
);
1018 bsb
= (struct bitmap_super_s
*)(((char*)super
)+1024);
1019 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1020 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1025 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
) & ~1);
1030 static struct supertype
*match_metadata_desc1(char *arg
)
1032 struct supertype
*st
= malloc(sizeof(*st
));
1037 if (strcmp(arg
, "1") == 0 ||
1038 strcmp(arg
, "1.0") == 0 ||
1039 strcmp(arg
, "default/large") == 0) {
1040 st
->minor_version
= 0;
1043 if (strcmp(arg
, "1.1") == 0) {
1044 st
->minor_version
= 1;
1047 if (strcmp(arg
, "1.2") == 0) {
1048 st
->minor_version
= 2;
1056 /* find available size on device with this devsize, using
1057 * superblock type st, and reserving 'reserve' sectors for
1060 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1065 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1066 * if biffer than 200Gig, save 128k
1068 if (devsize
> 200*1024*1024*2)
1070 else if (devsize
> 8*1024*1024*2)
1073 switch(st
->minor_version
) {
1076 return ((devsize
- 8*2 ) & ~(4*2-1));
1078 /* at start, 4K for superblock and possible bitmap */
1079 return devsize
- 4*2;
1081 /* 4k from start, 4K for superblock and possible bitmap */
1082 return devsize
- (4+4)*2;
1088 add_internal_bitmap1(struct supertype
*st
, void *sbv
,
1089 int chunk
, int delay
, int write_behind
, unsigned long long size
,
1090 int may_change
, int major
)
1093 * If not may_change, then this is a 'Grow', and the bitmap
1094 * must fit after the superblock.
1095 * If may_change, then this is create, and we can put the bitmap
1096 * before the superblock if we like, or may move the start.
1097 * For now, just squeeze the bitmap into 3k and don't change anything.
1099 * size is in sectors, chunk is in bytes !!!
1102 unsigned long long bits
;
1103 unsigned long long max_bits
= (3*512 - sizeof(bitmap_super_t
)) * 8;
1104 unsigned long long min_chunk
;
1105 struct mdp_superblock_1
*sb
= sbv
;
1106 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + 1024);
1108 if (st
->minor_version
&& !may_change
&&
1109 __le64_to_cpu(sb
->data_offset
) - __le64_to_cpu(sb
->super_offset
) < 8)
1110 return 0; /* doesn't fit */
1114 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1115 bits
= (size
*512)/min_chunk
+1;
1116 while (bits
> max_bits
) {
1122 else if (chunk
< min_chunk
)
1123 return 0; /* chunk size too small */
1124 if (chunk
== 0) /* rounding problem */
1127 sb
->bitmap_offset
= __cpu_to_le32(2);
1129 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) | 1);
1130 memset(bms
, 0, sizeof(*bms
));
1131 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1132 bms
->version
= __cpu_to_le32(major
);
1133 uuid_from_super1((int*)bms
->uuid
, sb
);
1134 bms
->chunksize
= __cpu_to_le32(chunk
);
1135 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1136 bms
->sync_size
= __cpu_to_le64(size
);
1137 bms
->write_behind
= __cpu_to_le32(write_behind
);
1143 static void locate_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
1145 unsigned long long offset
;
1146 struct mdp_superblock_1
*sb
;
1150 if (st
->ss
->load_super(st
, fd
, &sbv
, NULL
))
1151 return; /* no error I hope... */
1156 offset
= __le64_to_cpu(sb
->super_offset
);
1157 offset
+= (long) __le32_to_cpu(sb
->bitmap_offset
);
1160 lseek64(fd
, offset
<<9, 0);
1163 static int write_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
1165 struct mdp_superblock_1
*sb
= sbv
;
1166 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+1024);
1172 locate_bitmap1(st
, fd
, sbv
);
1174 if (write(fd
, ((char*)sb
)+1024, sizeof(bitmap_super_t
)) !=
1175 sizeof(bitmap_super_t
))
1177 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1178 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1179 memset(buf
, 0xff, sizeof(buf
));
1180 while (towrite
> 0) {
1182 if (n
> sizeof(buf
))
1184 n
= write(fd
, buf
, n
);
1197 struct superswitch super1
= {
1199 .examine_super
= examine_super1
,
1200 .brief_examine_super
= brief_examine_super1
,
1201 .detail_super
= detail_super1
,
1202 .brief_detail_super
= brief_detail_super1
,
1204 .match_home
= match_home1
,
1205 .uuid_from_super
= uuid_from_super1
,
1206 .getinfo_super
= getinfo_super1
,
1207 .update_super
= update_super1
,
1208 .event_super
= event_super1
,
1209 .init_super
= init_super1
,
1210 .add_to_super
= add_to_super1
,
1211 .store_super
= store_super1
,
1212 .write_init_super
= write_init_super1
,
1213 .compare_super
= compare_super1
,
1214 .load_super
= load_super1
,
1215 .match_metadata_desc
= match_metadata_desc1
,
1216 .avail_size
= avail_size1
,
1217 .add_internal_bitmap
= add_internal_bitmap1
,
1218 .locate_bitmap
= locate_bitmap1
,
1219 .write_bitmap
= write_bitmap1
,
1221 #if __BYTE_ORDER == BIG_ENDIAN