2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2004 Neil Brown <neilb@cse.unsw.edu.au>
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 #include "asm/byteorder.h"
34 * The version-1 superblock :
35 * All numeric fields are little-endian.
37 * total size: 256 bytes plus 2 per device.
38 * 1K allows 384 devices.
40 struct mdp_superblock_1
{
41 /* constant array information - 128 bytes */
42 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
43 __u32 major_version
; /* 1 */
44 __u32 feature_map
; /* 0 for now */
45 __u32 pad0
; /* always set to 0 when writing */
47 __u8 set_uuid
[16]; /* user-space generated. */
48 char set_name
[32]; /* set and interpreted by user-space */
50 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
51 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
52 __u32 layout
; /* only for raid5 currently */
53 __u64 size
; /* used size of component devices, in 512byte sectors */
55 __u32 chunksize
; /* in 512byte sectors */
57 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
58 * NOTE: signed, so bitmap can be before superblock
59 * only meaningful of feature_map[0] is set.
62 /* These are only valid with feature bit '4' */
63 __u64 reshape_position
; /* next address in array-space for reshape */
64 __u32 new_level
; /* new level we are reshaping to */
65 __u32 delta_disks
; /* change in number of raid_disks */
66 __u32 new_layout
; /* new layout */
67 __u32 new_chunk
; /* new chunk size (bytes) */
68 __u8 pad1
[128-124]; /* set to 0 when written */
70 /* constant this-device information - 64 bytes */
71 __u64 data_offset
; /* sector start of data, often 0 */
72 __u64 data_size
; /* sectors in this device that can be used for data */
73 __u64 super_offset
; /* sector start of this superblock */
74 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
75 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
76 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
77 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
78 __u8 devflags
; /* per-device flags. Only one defined...*/
79 #define WriteMostly1 1 /* mask for writemostly flag in above */
80 __u8 pad2
[64-57]; /* set to 0 when writing */
82 /* array state information - 64 bytes */
83 __u64 utime
; /* 40 bits second, 24 btes microseconds */
84 __u64 events
; /* incremented when superblock updated */
85 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
86 __u32 sb_csum
; /* checksum upto devs[max_dev] */
87 __u32 max_dev
; /* size of devs[] array to consider */
88 __u8 pad3
[64-32]; /* set to 0 when writing */
90 /* device state information. Indexed by dev_number.
92 * Note there are no per-device state flags. State information is rolled
93 * into the 'roles' value. If a device is spare or faulty, then it doesn't
94 * have a meaningful role.
96 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
99 /* feature_map bits */
100 #define MD_FEATURE_BITMAP_OFFSET 1
101 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
104 #define MD_FEATURE_RESHAPE_ACTIVE 4
106 #define MD_FEATURE_ALL (1|2|4)
109 #define offsetof(t,f) ((int)&(((t*)0)->f))
111 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
113 unsigned int disk_csum
, csum
;
114 unsigned long long newcsum
;
115 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
116 unsigned int *isuper
= (unsigned int*)sb
;
119 /* make sure I can count... */
120 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
121 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
122 sizeof(struct mdp_superblock_1
) != 256) {
123 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
126 disk_csum
= sb
->sb_csum
;
129 for (i
=0; size
>=4; size
-= 4 )
130 newcsum
+= __le32_to_cpu(*isuper
++);
133 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
135 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
136 sb
->sb_csum
= disk_csum
;
141 static void examine_super1(void *sbv
)
143 struct mdp_superblock_1
*sb
= sbv
;
150 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
151 printf(" Version : %02d.%02d\n", 1, __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\n", sb
->set_name
);
160 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
161 printf(" Creation Time : %.24s\n", ctime(&atime
));
162 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
163 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
164 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
166 printf(" Device Size : %llu%s\n", (unsigned long long)sb
->data_size
, human_size(sb
->data_size
<<9));
167 if (__le32_to_cpu(sb
->level
) >= 0) {
169 switch(__le32_to_cpu(sb
->level
)) {
170 case 1: ddsks
=1;break;
172 case 5: ddsks
= sb
->raid_disks
-1; break;
173 case 6: ddsks
= sb
->raid_disks
-2; break;
174 case 10: ddsks
= sb
->raid_disks
/ (sb
->layout
&255) / ((sb
->layout
>>8)&255);
177 printf(" Array Size : %llu%s\n", ddsks
*(unsigned long long)sb
->size
, human_size(ddsks
*sb
->size
<<9));
178 if (sb
->size
!= sb
->data_size
)
179 printf(" Used Size : %llu%s\n", (unsigned long long)sb
->size
, human_size(sb
->size
<<9));
182 printf(" Data Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->data_offset
));
183 if (sb
->super_offset
)
184 printf(" Super Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->super_offset
));
185 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
186 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
187 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
188 printf(" Device UUID : ");
189 for (i
=0; i
<16; i
++) {
190 if ((i
&3)==0 && i
!= 0) printf(":");
191 printf("%02x", sb
->device_uuid
[i
]);
195 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
196 printf(" Reshape pos'n : %llu%s\n", __le64_to_cpu(sb
->reshape_position
)/2,
197 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
198 if (__le32_to_cpu(sb
->delta_disks
)) {
199 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
200 if (__le32_to_cpu(sb
->delta_disks
))
201 printf(" (%d->%d)\n",
202 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
203 __le32_to_cpu(sb
->raid_disks
));
205 printf(" (%d->%d)\n", __le32_to_cpu(sb
->raid_disks
),
206 __le32_to_cpu(sb
->raid_disks
)+__le32_to_cpu(sb
->delta_disks
));
208 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
209 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
210 printf(" New Level : %s\n", c
?c
:"-unknown-");
212 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
213 if (__le32_to_cpu(sb
->level
) == 5) {
214 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
215 printf(" New Layout : %s\n", c
?c
:"-unknown-");
217 if (__le32_to_cpu(sb
->level
) == 10) {
218 printf(" New Layout : near=%d, far=%d\n",
219 __le32_to_cpu(sb
->new_layout
)&255,
220 (__le32_to_cpu(sb
->new_layout
)>>8)&255);
223 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
224 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
229 if (sb
->devflags
& WriteMostly1
)
230 printf(" write-mostly");
234 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
235 printf(" Update Time : %.24s\n", ctime(&atime
));
237 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
238 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
240 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
241 __le32_to_cpu(calc_sb_1_csum(sb
)));
242 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
244 if (__le32_to_cpu(sb
->level
) == 5) {
245 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
246 printf(" Layout : %s\n", c
?c
:"-unknown-");
248 if (__le32_to_cpu(sb
->level
) == 10) {
249 int lo
= __le32_to_cpu(sb
->layout
);
250 printf(" Layout : near=%d, far=%d\n",
251 lo
&255, (lo
>>8)&255);
253 switch(__le32_to_cpu(sb
->level
)) {
259 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
262 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
267 printf(" Array State : ");
268 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
); d
++) {
272 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
273 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
275 if (i
== __le32_to_cpu(sb
->dev_number
))
280 if (cnt
> 1) printf("?");
281 else if (cnt
== 1 && me
) printf("U");
282 else if (cnt
== 1) printf("u");
286 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
287 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
291 if (faulty
) printf(" %d failed", faulty
);
296 static void brief_examine_super1(void *sbv
)
298 struct mdp_superblock_1
*sb
= sbv
;
301 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
303 printf("ARRAY /dev/?? level=%s metadata=1 num-devices=%d UUID=",
304 c
?c
:"-unknown-", sb
->raid_disks
);
305 for (i
=0; i
<16; i
++) {
306 printf("%02x", sb
->set_uuid
[i
]);
307 if ((i
&3)==0 && i
!= 0) printf(":");
310 printf(" name=%.32s", sb
->set_name
);
314 static void detail_super1(void *sbv
)
316 struct mdp_superblock_1
*sb
= sbv
;
319 printf(" Name : %.32s\n", sb
->set_name
);
321 for (i
=0; i
<16; i
++) {
322 if ((i
&3)==0 && i
!= 0) printf(":");
323 printf("%02x", sb
->set_uuid
[i
]);
325 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
328 static void brief_detail_super1(void *sbv
)
330 struct mdp_superblock_1
*sb
= sbv
;
334 printf(" name=%.32s", sb
->set_name
);
336 for (i
=0; i
<16; i
++) {
337 if ((i
&3)==0 && i
!= 0) printf(":");
338 printf("%02x", sb
->set_uuid
[i
]);
344 static void uuid_from_super1(int uuid
[4], void * sbv
)
346 struct mdp_superblock_1
*super
= sbv
;
347 char *cuuid
= (char*)uuid
;
350 cuuid
[i
] = super
->set_uuid
[i
];
353 static void getinfo_super1(struct mdinfo
*info
, mddev_ident_t ident
, void *sbv
)
355 struct mdp_superblock_1
*sb
= sbv
;
360 info
->array
.major_version
= 1;
361 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
362 info
->array
.patch_version
= 0;
363 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
364 info
->array
.level
= __le32_to_cpu(sb
->level
);
365 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
366 info
->array
.md_minor
= -1;
367 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
369 info
->disk
.major
= 0;
370 info
->disk
.minor
= 0;
371 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
372 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
373 __le32_to_cpu(sb
->max_dev
) > 512)
376 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
378 info
->disk
.raid_disk
= -1;
381 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
384 info
->disk
.state
= 1; /* faulty */
387 info
->disk
.state
= 6; /* active and in sync */
388 info
->disk
.raid_disk
= role
;
390 info
->events
= __le64_to_cpu(sb
->events
);
392 memcpy(info
->uuid
, sb
->set_uuid
, 16);
394 strncpy(ident
->name
, sb
->set_name
, 32);
397 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
398 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
399 if (/*role == 0xFFFF || */role
< info
->array
.raid_disks
)
403 info
->array
.working_disks
= working
;
406 static int update_super1(struct mdinfo
*info
, void *sbv
, char *update
, char *devname
, int verbose
)
409 struct mdp_superblock_1
*sb
= sbv
;
411 if (strcmp(update
, "force")==0) {
412 sb
->events
= __cpu_to_le64(info
->events
);
413 switch(__le32_to_cpu(sb
->level
)) {
414 case 5: case 4: case 6:
415 /* need to force clean */
416 sb
->resync_offset
= ~0ULL;
419 if (strcmp(update
, "assemble")==0) {
420 int d
= info
->disk
.number
;
422 if (info
->disk
.state
== 6)
423 want
= __cpu_to_le32(info
->disk
.raid_disk
);
426 if (sb
->dev_roles
[d
] != want
) {
427 sb
->dev_roles
[d
] = want
;
432 if (strcmp(update
, "newdev") == 0) {
433 int d
= info
->disk
.number
;
434 memset(&sb
->disks
[d
], 0, sizeof(sb
->disks
[d
]));
435 sb
->disks
[d
].number
= d
;
436 sb
->disks
[d
].major
= info
->disk
.major
;
437 sb
->disks
[d
].minor
= info
->disk
.minor
;
438 sb
->disks
[d
].raid_disk
= info
->disk
.raid_disk
;
439 sb
->disks
[d
].state
= info
->disk
.state
;
440 sb
->this_disk
= sb
->disks
[d
];
443 if (strcmp(update
, "grow") == 0) {
444 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
447 if (strcmp(update
, "resync") == 0) {
448 /* make sure resync happens */
449 sb
->resync_offset
= ~0ULL;
451 if (strcmp(update
, "uuid") == 0)
452 memcpy(sb
->set_uuid
, info
->uuid
, 16);
454 sb
->sb_csum
= calc_sb_1_csum(sb
);
459 static __u64
event_super1(void *sbv
)
461 struct mdp_superblock_1
*sb
= sbv
;
462 return __le64_to_cpu(sb
->events
);
465 static int init_super1(struct supertype
*st
, void **sbp
, mdu_array_info_t
*info
, unsigned long long size
, char *name
)
467 struct mdp_superblock_1
*sb
= malloc(1024 + sizeof(bitmap_super_t
));
472 if (info
->major_version
== -1)
473 /* zeroing superblock */
476 spares
= info
->working_disks
- info
->active_disks
;
477 if (info
->raid_disks
+ spares
> 384) {
478 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
479 info
->raid_disks
, spares
, 384);
484 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
485 sb
->major_version
= __cpu_to_le32(1);
489 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
490 read(rfd
, sb
->set_uuid
, 16) != 16) {
491 *(__u32
*)(sb
->set_uuid
) = random();
492 *(__u32
*)(sb
->set_uuid
+4) = random();
493 *(__u32
*)(sb
->set_uuid
+8) = random();
494 *(__u32
*)(sb
->set_uuid
+12) = random();
496 if (rfd
>= 0) close(rfd
);
498 memset(sb
->set_name
, 0, 32);
499 strcpy(sb
->set_name
, name
);
501 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
502 sb
->level
= __cpu_to_le32(info
->level
);
503 sb
->layout
= __cpu_to_le32(info
->layout
);
504 sb
->size
= __cpu_to_le64(size
*2ULL);
505 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
506 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
508 sb
->data_offset
= __cpu_to_le64(0);
509 sb
->data_size
= __cpu_to_le64(0);
510 sb
->super_offset
= __cpu_to_le64(0);
511 sb
->recovery_offset
= __cpu_to_le64(0);
513 sb
->utime
= sb
->ctime
;
514 sb
->events
= __cpu_to_le64(1);
515 if (info
->state
& (1<<MD_SB_CLEAN
))
516 sb
->resync_offset
= ~0ULL;
518 sb
->resync_offset
= 0;
519 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
520 sizeof(sb
->dev_roles
[0]));
521 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
523 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
529 /* Add a device to the superblock being created */
530 static void add_to_super1(void *sbv
, mdu_disk_info_t
*dk
)
532 struct mdp_superblock_1
*sb
= sbv
;
533 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
534 if ((dk
->state
& 6) == 6) /* active, sync */
535 *rp
= __cpu_to_le16(dk
->raid_disk
);
536 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
542 static int store_super1(struct supertype
*st
, int fd
, void *sbv
)
544 struct mdp_superblock_1
*sb
= sbv
;
545 unsigned long long sb_offset
;
548 unsigned long long dsize
;
551 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
554 if (ioctl(fd
, BLKGETSIZE
, &size
))
557 dsize
= (unsigned long long)size
;
565 * Calculate the position of the superblock.
566 * It is always aligned to a 4K boundary and
567 * depending on minor_version, it can be:
568 * 0: At least 8K, but less than 12K, from end of device
569 * 1: At start of device
570 * 2: 4K from start of device.
572 switch(st
->minor_version
) {
576 sb_offset
&= ~(4*2-1);
590 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
591 0 != __le64_to_cpu(sb
->super_offset
)
593 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
597 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
600 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
602 if (write(fd
, sb
, sbsize
) != sbsize
)
609 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
);
611 static int write_init_super1(struct supertype
*st
, void *sbv
,
612 mdu_disk_info_t
*dinfo
, char *devname
)
614 struct mdp_superblock_1
*sb
= sbv
;
616 int fd
= open(devname
, O_RDWR
| O_EXCL
);
621 unsigned long long dsize
;
626 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
631 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
632 if (dinfo
->state
& (1<<MD_DISK_WRITEMOSTLY
))
633 sb
->devflags
|= WriteMostly1
;
635 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
636 read(rfd
, sb
->device_uuid
, 16) != 16) {
637 *(__u32
*)(sb
->device_uuid
) = random();
638 *(__u32
*)(sb
->device_uuid
+4) = random();
639 *(__u32
*)(sb
->device_uuid
+8) = random();
640 *(__u32
*)(sb
->device_uuid
+12) = random();
642 if (rfd
>= 0) close(rfd
);
645 if (load_super1(st
, fd
, &refsbv
, NULL
)==0) {
646 struct mdp_superblock_1
*refsb
= refsbv
;
648 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
649 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
650 /* same array, so preserve events and dev_number */
651 sb
->events
= refsb
->events
;
652 sb
->dev_number
= refsb
->dev_number
;
658 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
661 if (ioctl(fd
, BLKGETSIZE
, &size
))
675 * Calculate the position of the superblock.
676 * It is always aligned to a 4K boundary and
677 * depending on minor_version, it can be:
678 * 0: At least 8K, but less than 12K, from end of device
679 * 1: At start of device
680 * 2: 4K from start of device.
682 switch(st
->minor_version
) {
686 sb_offset
&= ~(4*2-1);
687 sb
->super_offset
= __cpu_to_le64(sb_offset
);
688 sb
->data_offset
= __cpu_to_le64(0);
689 sb
->data_size
= __cpu_to_le64(sb_offset
);
692 sb
->super_offset
= __cpu_to_le64(0);
693 sb
->data_offset
= __cpu_to_le64(4*2); /* leave 4k for super and bitmap */
694 sb
->data_size
= __cpu_to_le64(dsize
- 4*2);
698 sb
->super_offset
= __cpu_to_le64(sb_offset
);
699 sb
->data_offset
= __cpu_to_le64(sb_offset
+4*2);
700 sb
->data_size
= __cpu_to_le64(dsize
- 4*2 - 4*2);
707 sb
->sb_csum
= calc_sb_1_csum(sb
);
708 rv
= store_super1(st
, fd
, sb
);
710 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
712 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
713 rv
= st
->ss
->write_bitmap(st
, fd
, sbv
);
718 static int compare_super1(void **firstp
, void *secondv
)
722 * 0 same, or first was empty, and second was copied
723 * 1 second had wrong number
727 struct mdp_superblock_1
*first
= *firstp
;
728 struct mdp_superblock_1
*second
= secondv
;
730 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
732 if (second
->major_version
!= __cpu_to_le32(1))
736 first
= malloc(1024);
737 memcpy(first
, second
, 1024);
741 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
744 if (first
->ctime
!= second
->ctime
||
745 first
->level
!= second
->level
||
746 first
->layout
!= second
->layout
||
747 first
->size
!= second
->size
||
748 first
->chunksize
!= second
->chunksize
||
749 first
->raid_disks
!= second
->raid_disks
)
754 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
)
757 unsigned long long dsize
;
758 unsigned long long sb_offset
;
759 struct mdp_superblock_1
*super
;
763 if (st
->ss
== NULL
) {
766 /* guess... choose latest ctime */
768 for (st
->minor_version
= 0; st
->minor_version
<= 2 ; st
->minor_version
++) {
769 switch(load_super1(st
, fd
, sbp
, devname
)) {
770 case 0: super
= *sbp
;
771 if (bestvers
== -1 ||
772 bestctime
< __le64_to_cpu(super
->ctime
)) {
773 bestvers
= st
->minor_version
;
774 bestctime
= __le64_to_cpu(super
->ctime
);
779 case 1: st
->ss
= NULL
; return 1; /*bad device */
780 case 2: break; /* bad, try next */
783 if (bestvers
!= -1) {
785 st
->minor_version
= bestvers
;
788 rv
= load_super1(st
, fd
, sbp
, devname
);
789 if (rv
) st
->ss
= NULL
;
796 if (ioctl(fd
, BLKGETSIZE64
, &dsize
) != 0)
799 if (ioctl(fd
, BLKGETSIZE
, &size
)) {
801 fprintf(stderr
, Name
": cannot find device size for %s: %s\n",
802 devname
, strerror(errno
));
811 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
817 * Calculate the position of the superblock.
818 * It is always aligned to a 4K boundary and
819 * depeding on minor_version, it can be:
820 * 0: At least 8K, but less than 12K, from end of device
821 * 1: At start of device
822 * 2: 4K from start of device.
824 switch(st
->minor_version
) {
828 sb_offset
&= ~(4*2-1);
840 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
843 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
845 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
846 devname
, strerror(errno
));
850 super
= malloc(1024 + sizeof(bitmap_super_t
));
852 if (read(fd
, super
, 1024) != 1024) {
854 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
860 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
862 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
863 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
868 if (__le32_to_cpu(super
->major_version
) != 1) {
870 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
871 devname
, __le32_to_cpu(super
->major_version
));
875 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
877 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
887 static struct supertype
*match_metadata_desc1(char *arg
)
889 struct supertype
*st
= malloc(sizeof(*st
));
894 if (strcmp(arg
, "1") == 0 ||
895 strcmp(arg
, "1.0") == 0) {
896 st
->minor_version
= 0;
899 if (strcmp(arg
, "1.1") == 0) {
900 st
->minor_version
= 1;
903 if (strcmp(arg
, "1.2") == 0) {
904 st
->minor_version
= 2;
912 /* find available size on device with this devsize, using
913 * superblock type st, and reserving 'reserve' sectors for
916 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
921 switch(st
->minor_version
) {
924 return ((devsize
- 8*2 ) & ~(4*2-1));
926 /* at start, 4K for superblock and possible bitmap */
927 return devsize
- 4*2;
929 /* 4k from start, 4K for superblock and possible bitmap */
930 return devsize
- (4+4)*2;
936 add_internal_bitmap1(struct supertype
*st
, void *sbv
,
937 int chunk
, int delay
, int write_behind
, unsigned long long size
, int may_change
, int major
)
940 * If not may_change, then this is a 'Grow', and the bitmap
941 * must fit after the superblock.
942 * If may_change, then this is create, and we can put the bitmap
943 * before the superblock if we like, or may move the start.
944 * For now, just squeeze the bitmap into 3k and don't change anything.
946 * size is in sectors, chunk is in bytes !!!
949 unsigned long long bits
;
950 unsigned long long max_bits
= (3*512 - sizeof(bitmap_super_t
)) * 8;
951 unsigned long long min_chunk
;
952 struct mdp_superblock_1
*sb
= sbv
;
953 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + 1024);
955 if (st
->minor_version
&& !may_change
&&
956 __le64_to_cpu(sb
->data_offset
) - __le64_to_cpu(sb
->super_offset
) < 8)
957 return 0; /* doesn't fit */
961 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
962 bits
= (size
*512)/min_chunk
+1;
963 while (bits
> max_bits
) {
969 else if (chunk
< min_chunk
)
970 return 0; /* chunk size too small */
972 sb
->bitmap_offset
= __cpu_to_le32(2);
974 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) | 1);
975 memset(bms
, 0, sizeof(*bms
));
976 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
977 bms
->version
= __cpu_to_le32(major
);
978 uuid_from_super1((int*)bms
->uuid
, sb
);
979 bms
->chunksize
= __cpu_to_le32(chunk
);
980 bms
->daemon_sleep
= __cpu_to_le32(delay
);
981 bms
->sync_size
= __cpu_to_le64(size
);
982 bms
->write_behind
= __cpu_to_le32(write_behind
);
988 void locate_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
990 unsigned long long offset
;
991 struct mdp_superblock_1
*sb
;
994 if (st
->ss
->load_super(st
, fd
, sbv
, NULL
))
995 return; /* no error I hope... */
999 offset
= __le64_to_cpu(sb
->super_offset
);
1000 offset
+= (long) __le32_to_cpu(sb
->bitmap_offset
);
1003 lseek64(fd
, offset
<<9, 0);
1006 int write_bitmap1(struct supertype
*st
, int fd
, void *sbv
)
1008 struct mdp_superblock_1
*sb
= sbv
;
1009 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+1024);
1015 locate_bitmap1(st
, fd
, sbv
);
1017 write(fd
, ((char*)sb
)+1024, sizeof(bitmap_super_t
));
1018 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1019 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1020 memset(buf
, 0xff, sizeof(buf
));
1021 while (towrite
> 0) {
1023 if (n
> sizeof(buf
))
1025 n
= write(fd
, buf
, n
);
1038 struct superswitch super1
= {
1040 .examine_super
= examine_super1
,
1041 .brief_examine_super
= brief_examine_super1
,
1042 .detail_super
= detail_super1
,
1043 .brief_detail_super
= brief_detail_super1
,
1045 .uuid_from_super
= uuid_from_super1
,
1046 .getinfo_super
= getinfo_super1
,
1047 .update_super
= update_super1
,
1048 .event_super
= event_super1
,
1049 .init_super
= init_super1
,
1050 .add_to_super
= add_to_super1
,
1051 .store_super
= store_super1
,
1052 .write_init_super
= write_init_super1
,
1053 .compare_super
= compare_super1
,
1054 .load_super
= load_super1
,
1055 .match_metadata_desc
= match_metadata_desc1
,
1056 .avail_size
= avail_size1
,
1057 .add_internal_bitmap
= add_internal_bitmap1
,
1058 .locate_bitmap
= locate_bitmap1
,
1059 .write_bitmap
= write_bitmap1
,
1061 #if __BYTE_ORDER == BIG_ENDIAN