2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 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@suse.de>
27 * The version-1 superblock :
28 * All numeric fields are little-endian.
30 * total size: 256 bytes plus 2 per device.
31 * 1K allows 384 devices.
33 struct mdp_superblock_1
{
34 /* constant array information - 128 bytes */
35 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
36 __u32 major_version
; /* 1 */
37 __u32 feature_map
; /* 0 for now */
38 __u32 pad0
; /* always set to 0 when writing */
40 __u8 set_uuid
[16]; /* user-space generated. */
41 char set_name
[32]; /* set and interpreted by user-space */
43 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
44 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
45 __u32 layout
; /* only for raid5 currently */
46 __u64 size
; /* used size of component devices, in 512byte sectors */
48 __u32 chunksize
; /* in 512byte sectors */
50 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
51 * NOTE: signed, so bitmap can be before superblock
52 * only meaningful of feature_map[0] is set.
55 /* These are only valid with feature bit '4' */
56 __u32 new_level
; /* new level we are reshaping to */
57 __u64 reshape_position
; /* next address in array-space for reshape */
58 __u32 delta_disks
; /* change in number of raid_disks */
59 __u32 new_layout
; /* new layout */
60 __u32 new_chunk
; /* new chunk size (bytes) */
61 __u8 pad1
[128-124]; /* set to 0 when written */
63 /* constant this-device information - 64 bytes */
64 __u64 data_offset
; /* sector start of data, often 0 */
65 __u64 data_size
; /* sectors in this device that can be used for data */
66 __u64 super_offset
; /* sector start of this superblock */
67 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
68 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
69 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
70 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
71 __u8 devflags
; /* per-device flags. Only one defined...*/
72 #define WriteMostly1 1 /* mask for writemostly flag in above */
73 __u8 pad2
[64-57]; /* set to 0 when writing */
75 /* array state information - 64 bytes */
76 __u64 utime
; /* 40 bits second, 24 btes microseconds */
77 __u64 events
; /* incremented when superblock updated */
78 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
79 __u32 sb_csum
; /* checksum upto dev_roles[max_dev] */
80 __u32 max_dev
; /* size of dev_roles[] array to consider */
81 __u8 pad3
[64-32]; /* set to 0 when writing */
83 /* device state information. Indexed by dev_number.
85 * Note there are no per-device state flags. State information is rolled
86 * into the 'roles' value. If a device is spare or faulty, then it doesn't
87 * have a meaningful role.
89 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
92 #define MAX_SB_SIZE 4096
93 /* bitmap super size is 256, but we round up to a sector for alignment */
94 #define BM_SUPER_SIZE 512
95 #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2)
96 #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \
97 + sizeof(struct misc_dev_info))
99 struct misc_dev_info
{
103 /* feature_map bits */
104 #define MD_FEATURE_BITMAP_OFFSET 1
105 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
108 #define MD_FEATURE_RESHAPE_ACTIVE 4
110 #define MD_FEATURE_ALL (1|2|4)
113 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
115 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
117 unsigned int disk_csum
, csum
;
118 unsigned long long newcsum
;
119 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
120 unsigned int *isuper
= (unsigned int*)sb
;
122 /* make sure I can count... */
123 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
124 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
125 sizeof(struct mdp_superblock_1
) != 256) {
126 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
129 disk_csum
= sb
->sb_csum
;
132 for (; size
>=4; size
-= 4 ) {
133 newcsum
+= __le32_to_cpu(*isuper
);
138 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
140 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
141 sb
->sb_csum
= disk_csum
;
142 return __cpu_to_le32(csum
);
146 * Information related to file descriptor used for aligned reads/writes.
147 * Cache the block size.
154 static void init_afd(struct align_fd
*afd
, int fd
)
158 if (ioctl(afd
->fd
, BLKSSZGET
, &afd
->blk_sz
) != 0)
162 static char abuf
[4096+4096];
163 static int aread(struct align_fd
*afd
, void *buf
, int len
)
166 * On devices with a 4K sector size, we need to read
167 * the full sector and copy relevant bits into
176 if (!bsize
|| bsize
> 4096 || len
> 4096) {
178 fprintf(stderr
, "WARNING - aread() called with "
179 "invalid block size\n");
182 b
= ROUND_UP_PTR((char *)abuf
, 4096);
184 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
186 n
= read(afd
->fd
, b
, iosize
);
189 lseek(afd
->fd
, len
- n
, 1);
196 static int awrite(struct align_fd
*afd
, void *buf
, int len
)
199 * On devices with a 4K sector size, we need to write
200 * the full sector. We pre-read if the sector is larger
202 * The address must be sector-aligned.
209 if (!bsize
|| bsize
> 4096 || len
> 4096) {
211 fprintf(stderr
, "WARNING - awrite() called with "
212 "invalid block size\n");
215 b
= ROUND_UP_PTR((char *)abuf
, 4096);
217 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
221 n
= read(afd
->fd
, b
, iosize
);
224 lseek(afd
->fd
, -n
, 1);
228 n
= write(afd
->fd
, b
, iosize
);
231 lseek(afd
->fd
, len
- n
, 1);
236 static void examine_super1(struct supertype
*st
, char *homehost
)
238 struct mdp_superblock_1
*sb
= st
->sb
;
245 int l
= homehost
? strlen(homehost
) : 0;
247 unsigned long long sb_offset
;
249 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
250 printf(" Version : 1");
251 sb_offset
= __le64_to_cpu(sb
->super_offset
);
254 else if (sb_offset
<= 8)
258 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
259 printf(" Array UUID : ");
260 for (i
=0; i
<16; i
++) {
261 if ((i
&3)==0 && i
!= 0) printf(":");
262 printf("%02x", sb
->set_uuid
[i
]);
265 printf(" Name : %.32s", sb
->set_name
);
266 if (l
> 0 && l
< 32 &&
267 sb
->set_name
[l
] == ':' &&
268 strncmp(sb
->set_name
, homehost
, l
) == 0)
269 printf(" (local to host %s)", homehost
);
271 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
272 printf(" Creation Time : %.24s\n", ctime(&atime
));
273 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
274 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
275 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
277 printf(" Avail Dev Size : %llu%s\n",
278 (unsigned long long)__le64_to_cpu(sb
->data_size
),
279 human_size(__le64_to_cpu(sb
->data_size
)<<9));
280 if (__le32_to_cpu(sb
->level
) > 0) {
282 switch(__le32_to_cpu(sb
->level
)) {
283 case 1: ddsks
=1;break;
285 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
286 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
288 layout
= __le32_to_cpu(sb
->layout
);
289 ddsks
= __le32_to_cpu(sb
->raid_disks
)
290 / (layout
&255) / ((layout
>>8)&255);
293 printf(" Array Size : %llu%s\n",
294 ddsks
*(unsigned long long)__le64_to_cpu(sb
->size
),
295 human_size(ddsks
*__le64_to_cpu(sb
->size
)<<9));
296 if (sb
->size
!= sb
->data_size
)
297 printf(" Used Dev Size : %llu%s\n",
298 (unsigned long long)__le64_to_cpu(sb
->size
),
299 human_size(__le64_to_cpu(sb
->size
)<<9));
302 printf(" Data Offset : %llu sectors\n",
303 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
304 printf(" Super Offset : %llu sectors\n",
305 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
306 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
307 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
308 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
309 printf(" Device UUID : ");
310 for (i
=0; i
<16; i
++) {
311 if ((i
&3)==0 && i
!= 0) printf(":");
312 printf("%02x", sb
->device_uuid
[i
]);
316 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
317 printf("Internal Bitmap : %ld sectors from superblock\n",
318 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
320 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
321 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
322 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
323 if (__le32_to_cpu(sb
->delta_disks
)) {
324 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
325 printf(" (%d->%d)\n",
326 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
327 __le32_to_cpu(sb
->raid_disks
));
328 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
329 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
331 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
332 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
333 printf(" New Level : %s\n", c
?c
:"-unknown-");
335 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
336 if (__le32_to_cpu(sb
->level
) == 5) {
337 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
338 printf(" New Layout : %s\n", c
?c
:"-unknown-");
340 if (__le32_to_cpu(sb
->level
) == 6) {
341 c
= map_num(r6layout
, __le32_to_cpu(sb
->new_layout
));
342 printf(" New Layout : %s\n", c
?c
:"-unknown-");
344 if (__le32_to_cpu(sb
->level
) == 10) {
345 printf(" New Layout :");
346 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
350 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
351 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
356 if (sb
->devflags
& WriteMostly1
)
357 printf(" write-mostly");
361 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
362 printf(" Update Time : %.24s\n", ctime(&atime
));
364 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
365 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
367 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
368 __le32_to_cpu(calc_sb_1_csum(sb
)));
369 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
371 if (__le32_to_cpu(sb
->level
) == 5) {
372 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
373 printf(" Layout : %s\n", c
?c
:"-unknown-");
375 if (__le32_to_cpu(sb
->level
) == 6) {
376 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
377 printf(" Layout : %s\n", c
?c
:"-unknown-");
379 if (__le32_to_cpu(sb
->level
) == 10) {
380 int lo
= __le32_to_cpu(sb
->layout
);
382 print_r10_layout(lo
);
385 switch(__le32_to_cpu(sb
->level
)) {
391 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
394 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
400 /* This turns out to just be confusing */
401 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
402 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
403 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
405 for (d
=0; d
< i
; d
++) {
406 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
408 if (role
== 0xffff) printf("empty");
409 else if(role
== 0xfffe) printf("failed");
410 else printf("%d", role
);
414 printf(" Device Role : ");
415 d
= __le32_to_cpu(sb
->dev_number
);
416 if (d
< __le32_to_cpu(sb
->max_dev
))
417 role
= __le16_to_cpu(sb
->dev_roles
[d
]);
423 printf("Active device %d\n", role
);
425 printf(" Array State : ");
426 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
429 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
430 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
434 if (cnt
> 1) printf("?");
435 else if (cnt
== 1) printf("A");
439 /* This is confusing too */
441 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
442 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
446 if (faulty
) printf(" %d failed", faulty
);
448 printf(" ('A' == active, '.' == missing)");
453 static void brief_examine_super1(struct supertype
*st
, int verbose
)
455 struct mdp_superblock_1
*sb
= st
->sb
;
457 unsigned long long sb_offset
;
459 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
461 nm
= strchr(sb
->set_name
, ':');
464 else if (sb
->set_name
[0])
469 printf("ARRAY%s%s", nm
? " /dev/md/":"", nm
);
471 printf(" level=%s", c
);
472 sb_offset
= __le64_to_cpu(sb
->super_offset
);
474 printf(" metadata=1.1 ");
475 else if (sb_offset
<= 8)
476 printf(" metadata=1.2 ");
478 printf(" metadata=1.0 ");
480 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
482 for (i
=0; i
<16; i
++) {
483 if ((i
&3)==0 && i
!= 0) printf(":");
484 printf("%02x", sb
->set_uuid
[i
]);
487 printf(" name=%.32s", sb
->set_name
);
491 static void export_examine_super1(struct supertype
*st
)
493 struct mdp_superblock_1
*sb
= st
->sb
;
497 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
498 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
500 if (sb
->set_name
[i
] == '\n' ||
501 sb
->set_name
[i
] == '\0') {
506 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
508 for (i
=0; i
<16; i
++) {
509 if ((i
&3)==0 && i
!= 0) printf(":");
510 printf("%02x", sb
->set_uuid
[i
]);
513 printf("MD_UPDATE_TIME=%llu\n",
514 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
515 printf("MD_DEV_UUID=");
516 for (i
=0; i
<16; i
++) {
517 if ((i
&3)==0 && i
!= 0) printf(":");
518 printf("%02x", sb
->device_uuid
[i
]);
521 printf("MD_EVENTS=%llu\n",
522 (unsigned long long)__le64_to_cpu(sb
->events
));
525 static void detail_super1(struct supertype
*st
, char *homehost
)
527 struct mdp_superblock_1
*sb
= st
->sb
;
529 int l
= homehost
? strlen(homehost
) : 0;
531 printf(" Name : %.32s", sb
->set_name
);
532 if (l
> 0 && l
< 32 &&
533 sb
->set_name
[l
] == ':' &&
534 strncmp(sb
->set_name
, homehost
, l
) == 0)
535 printf(" (local to host %s)", homehost
);
536 printf("\n UUID : ");
537 for (i
=0; i
<16; i
++) {
538 if ((i
&3)==0 && i
!= 0) printf(":");
539 printf("%02x", sb
->set_uuid
[i
]);
541 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
544 static void brief_detail_super1(struct supertype
*st
)
546 struct mdp_superblock_1
*sb
= st
->sb
;
550 printf(" name=%.32s", sb
->set_name
);
552 for (i
=0; i
<16; i
++) {
553 if ((i
&3)==0 && i
!= 0) printf(":");
554 printf("%02x", sb
->set_uuid
[i
]);
558 static void export_detail_super1(struct supertype
*st
)
560 struct mdp_superblock_1
*sb
= st
->sb
;
565 if (sb
->set_name
[i
] == '\n' ||
566 sb
->set_name
[i
] == '\0') {
571 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
576 static int match_home1(struct supertype
*st
, char *homehost
)
578 struct mdp_superblock_1
*sb
= st
->sb
;
579 int l
= homehost
? strlen(homehost
) : 0;
581 return (l
> 0 && l
< 32 &&
582 sb
->set_name
[l
] == ':' &&
583 strncmp(sb
->set_name
, homehost
, l
) == 0);
586 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
588 struct mdp_superblock_1
*super
= st
->sb
;
589 char *cuuid
= (char*)uuid
;
592 cuuid
[i
] = super
->set_uuid
[i
];
595 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
597 struct mdp_superblock_1
*sb
= st
->sb
;
601 unsigned int map_disks
= info
->array
.raid_disks
;
603 memset(info
, 0, sizeof(*info
));
604 info
->array
.major_version
= 1;
605 info
->array
.minor_version
= st
->minor_version
;
606 info
->array
.patch_version
= 0;
607 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
608 info
->array
.level
= __le32_to_cpu(sb
->level
);
609 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
610 info
->array
.md_minor
= -1;
611 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
612 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
613 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
615 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
)
618 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
619 info
->component_size
= __le64_to_cpu(sb
->size
);
620 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
))
621 info
->bitmap_offset
= __le32_to_cpu(sb
->bitmap_offset
);
623 info
->disk
.major
= 0;
624 info
->disk
.minor
= 0;
625 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
626 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
627 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
630 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
632 info
->disk
.raid_disk
= -1;
635 info
->disk
.state
= 0; /* spare: not active, not sync, not faulty */
638 info
->disk
.state
= 1; /* faulty */
641 info
->disk
.state
= 6; /* active and in sync */
642 info
->disk
.raid_disk
= role
;
644 if (sb
->devflags
& WriteMostly1
)
645 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
646 info
->events
= __le64_to_cpu(sb
->events
);
647 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
648 info
->safe_mode_delay
= 200;
650 memcpy(info
->uuid
, sb
->set_uuid
, 16);
652 strncpy(info
->name
, sb
->set_name
, 32);
655 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
656 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
658 info
->recovery_start
= MaxSector
;
660 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
661 info
->reshape_active
= 1;
662 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
663 info
->new_level
= __le32_to_cpu(sb
->new_level
);
664 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
665 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
666 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
667 if (info
->delta_disks
< 0)
668 info
->array
.raid_disks
-= info
->delta_disks
;
670 info
->reshape_active
= 0;
672 info
->recovery_blocked
= info
->reshape_active
;
675 for (i
=0; i
<map_disks
; i
++)
677 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
678 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
679 if (/*role == 0xFFFF || */role
< (unsigned) info
->array
.raid_disks
) {
681 if (map
&& role
< map_disks
)
686 info
->array
.working_disks
= working
;
689 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
696 info
= malloc(sizeof(*info
));
697 getinfo_super1(st
, info
, NULL
);
701 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
703 char *devname
, int verbose
,
704 int uuid_set
, char *homehost
)
706 /* NOTE: for 'assemble' and 'force' we need to return non-zero
707 * if any change was made. For others, the return value is
711 struct mdp_superblock_1
*sb
= st
->sb
;
713 if (strcmp(update
, "force-one")==0) {
714 /* Not enough devices for a working array,
715 * so bring this one up-to-date
717 if (sb
->events
!= __cpu_to_le64(info
->events
))
719 sb
->events
= __cpu_to_le64(info
->events
);
720 } else if (strcmp(update
, "force-array")==0) {
721 /* Degraded array and 'force' requests to
722 * maybe need to mark it 'clean'.
724 switch(__le32_to_cpu(sb
->level
)) {
725 case 5: case 4: case 6:
726 /* need to force clean */
727 if (sb
->resync_offset
!= MaxSector
)
729 sb
->resync_offset
= MaxSector
;
731 } else if (strcmp(update
, "assemble")==0) {
732 int d
= info
->disk
.number
;
734 if (info
->disk
.state
== 6)
735 want
= info
->disk
.raid_disk
;
738 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
739 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
742 if (info
->reshape_active
&&
743 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
744 info
->delta_disks
>= 0 &&
745 info
->reshape_progress
< __le64_to_cpu(sb
->reshape_position
)) {
746 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
749 if (info
->reshape_active
&&
750 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
751 info
->delta_disks
< 0 &&
752 info
->reshape_progress
> __le64_to_cpu(sb
->reshape_position
)) {
753 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
756 } else if (strcmp(update
, "linear-grow-new") == 0) {
759 unsigned int max
= __le32_to_cpu(sb
->max_dev
);
761 for (i
=0 ; i
< max
; i
++)
762 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
764 sb
->dev_number
= __cpu_to_le32(i
);
765 info
->disk
.number
= i
;
766 if (max
>= __le32_to_cpu(sb
->max_dev
))
767 sb
->max_dev
= __cpu_to_le32(max
+1);
769 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
770 read(rfd
, sb
->device_uuid
, 16) != 16) {
771 __u32 r
[4] = {random(), random(), random(), random()};
772 memcpy(sb
->device_uuid
, r
, 16);
778 __cpu_to_le16(info
->disk
.raid_disk
);
780 fd
= open(devname
, O_RDONLY
);
782 unsigned long long ds
;
783 get_dev_size(fd
, devname
, &ds
);
786 if (__le64_to_cpu(sb
->super_offset
) <
787 __le64_to_cpu(sb
->data_offset
)) {
788 sb
->data_size
= __cpu_to_le64(
789 ds
- __le64_to_cpu(sb
->data_offset
));
792 ds
&= ~(unsigned long long)(4*2-1);
793 sb
->super_offset
= __cpu_to_le64(ds
);
794 sb
->data_size
= __cpu_to_le64(
795 ds
- __le64_to_cpu(sb
->data_offset
));
798 } else if (strcmp(update
, "linear-grow-update") == 0) {
799 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
800 sb
->dev_roles
[info
->disk
.number
] =
801 __cpu_to_le16(info
->disk
.raid_disk
);
802 } else if (strcmp(update
, "resync") == 0) {
803 /* make sure resync happens */
804 sb
->resync_offset
= 0ULL;
805 } else if (strcmp(update
, "uuid") == 0) {
806 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
808 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
809 struct bitmap_super_s
*bm
;
810 bm
= (struct bitmap_super_s
*)(st
->sb
+MAX_SB_SIZE
);
811 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
813 } else if (strcmp(update
, "no-bitmap") == 0) {
814 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
815 } else if (strcmp(update
, "homehost") == 0 &&
819 c
= strchr(sb
->set_name
, ':');
821 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
823 strncpy(info
->name
, sb
->set_name
, 32);
825 } else if (strcmp(update
, "name") == 0) {
826 if (info
->name
[0] == 0)
827 sprintf(info
->name
, "%d", info
->array
.md_minor
);
828 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
830 strchr(info
->name
, ':') == NULL
&&
831 strlen(homehost
)+1+strlen(info
->name
) < 32) {
832 strcpy(sb
->set_name
, homehost
);
833 strcat(sb
->set_name
, ":");
834 strcat(sb
->set_name
, info
->name
);
836 strcpy(sb
->set_name
, info
->name
);
837 } else if (strcmp(update
, "devicesize") == 0 &&
838 __le64_to_cpu(sb
->super_offset
) <
839 __le64_to_cpu(sb
->data_offset
)) {
840 /* set data_size to device size less data_offset */
841 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
842 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
843 printf("Size was %llu\n", (unsigned long long)
844 __le64_to_cpu(sb
->data_size
));
845 sb
->data_size
= __cpu_to_le64(
846 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
847 printf("Size is %llu\n", (unsigned long long)
848 __le64_to_cpu(sb
->data_size
));
849 } else if (strcmp(update
, "_reshape_progress")==0)
850 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
851 else if (strcmp(update
, "writemostly")==0)
852 sb
->devflags
|= WriteMostly1
;
853 else if (strcmp(update
, "readwrite")==0)
854 sb
->devflags
&= ~WriteMostly1
;
858 sb
->sb_csum
= calc_sb_1_csum(sb
);
862 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
863 unsigned long long size
, char *name
, char *homehost
, int *uuid
)
865 struct mdp_superblock_1
*sb
;
871 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
873 ": %s could not allocate superblock\n", __func__
);
876 memset(sb
, 0, SUPER1_SIZE
);
880 /* zeroing superblock */
884 spares
= info
->working_disks
- info
->active_disks
;
885 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
886 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
887 info
->raid_disks
, spares
, MAX_DEVS
);
891 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
892 sb
->major_version
= __cpu_to_le32(1);
897 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
899 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
900 read(rfd
, sb
->set_uuid
, 16) != 16) {
901 __u32 r
[4] = {random(), random(), random(), random()};
902 memcpy(sb
->set_uuid
, r
, 16);
904 if (rfd
>= 0) close(rfd
);
907 if (name
== NULL
|| *name
== 0) {
908 sprintf(defname
, "%d", info
->md_minor
);
912 strchr(name
, ':')== NULL
&&
913 strlen(homehost
)+1+strlen(name
) < 32) {
914 strcpy(sb
->set_name
, homehost
);
915 strcat(sb
->set_name
, ":");
916 strcat(sb
->set_name
, name
);
918 strcpy(sb
->set_name
, name
);
920 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
921 sb
->level
= __cpu_to_le32(info
->level
);
922 sb
->layout
= __cpu_to_le32(info
->layout
);
923 sb
->size
= __cpu_to_le64(size
*2ULL);
924 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
925 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
927 sb
->data_offset
= __cpu_to_le64(0);
928 sb
->data_size
= __cpu_to_le64(0);
929 sb
->super_offset
= __cpu_to_le64(0);
930 sb
->recovery_offset
= __cpu_to_le64(0);
932 sb
->utime
= sb
->ctime
;
933 sb
->events
= __cpu_to_le64(1);
934 if (info
->state
& (1<<MD_SB_CLEAN
))
935 sb
->resync_offset
= MaxSector
;
937 sb
->resync_offset
= 0;
938 sbsize
= sizeof(struct mdp_superblock_1
) + 2 * (info
->raid_disks
+ spares
);
939 sbsize
= ROUND_UP(sbsize
, 512);
940 sb
->max_dev
= __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
942 memset(sb
->dev_roles
, 0xff, MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
950 mdu_disk_info_t disk
;
951 struct devinfo
*next
;
954 /* Add a device to the superblock being created */
955 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
956 int fd
, char *devname
)
958 struct mdp_superblock_1
*sb
= st
->sb
;
959 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
960 struct devinfo
*di
, **dip
;
962 if ((dk
->state
& 6) == 6) /* active, sync */
963 *rp
= __cpu_to_le16(dk
->raid_disk
);
964 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
969 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
970 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
971 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
973 sb
->dev_number
= __cpu_to_le32(dk
->number
);
974 sb
->devflags
= 0; /* don't copy another disks flags */
975 sb
->sb_csum
= calc_sb_1_csum(sb
);
977 dip
= (struct devinfo
**)&st
->info
;
980 di
= malloc(sizeof(struct devinfo
));
982 di
->devname
= devname
;
991 static void locate_bitmap1(struct supertype
*st
, int fd
);
993 static int store_super1(struct supertype
*st
, int fd
)
995 struct mdp_superblock_1
*sb
= st
->sb
;
996 unsigned long long sb_offset
;
999 unsigned long long dsize
;
1001 if (!get_dev_size(fd
, NULL
, &dsize
))
1012 * Calculate the position of the superblock.
1013 * It is always aligned to a 4K boundary and
1014 * depending on minor_version, it can be:
1015 * 0: At least 8K, but less than 12K, from end of device
1016 * 1: At start of device
1017 * 2: 4K from start of device.
1019 switch(st
->minor_version
) {
1023 sb_offset
&= ~(4*2-1);
1037 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1038 0 != __le64_to_cpu(sb
->super_offset
)
1040 fprintf(stderr
, Name
": internal error - sb_offset is wrong\n");
1044 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1047 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1049 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1052 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1053 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1054 (((char*)sb
)+MAX_SB_SIZE
);
1055 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1056 locate_bitmap1(st
, fd
);
1057 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1065 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1067 static unsigned long choose_bm_space(unsigned long devsize
)
1069 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1070 * if bigger than 200Gig, save 128k
1071 * NOTE: result must be multiple of 4K else bad things happen
1072 * on 4K-sector devices.
1074 if (devsize
< 64*2) return 0;
1075 if (devsize
- 64*2 >= 200*1024*1024*2)
1077 if (devsize
- 4*2 > 8*1024*1024*2)
1082 static void free_super1(struct supertype
*st
);
1085 static int write_init_super1(struct supertype
*st
)
1087 struct mdp_superblock_1
*sb
= st
->sb
;
1088 struct supertype
*refst
;
1091 unsigned long long bm_space
;
1092 unsigned long long reserved
;
1094 unsigned long long dsize
, array_size
;
1095 unsigned long long sb_offset
;
1097 for (di
= st
->info
; di
&& ! rv
; di
= di
->next
) {
1098 if (di
->disk
.state
== 1)
1103 while (Kill(di
->devname
, NULL
, 0, 1, 1) == 0)
1106 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1107 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1108 sb
->devflags
|= WriteMostly1
;
1110 sb
->devflags
&= ~WriteMostly1
;
1112 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1113 read(rfd
, sb
->device_uuid
, 16) != 16) {
1114 __u32 r
[4] = {random(), random(), random(), random()};
1115 memcpy(sb
->device_uuid
, r
, 16);
1122 refst
= dup_super(st
);
1123 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1124 struct mdp_superblock_1
*refsb
= refst
->sb
;
1126 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1127 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
1128 /* same array, so preserve events and
1130 sb
->events
= refsb
->events
;
1131 /* bugs in 2.6.17 and earlier mean the
1132 * dev_number chosen in Manage must be preserved
1134 if (get_linux_version() >= 2006018)
1135 sb
->dev_number
= refsb
->dev_number
;
1141 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
1155 * Calculate the position of the superblock.
1156 * It is always aligned to a 4K boundary and
1157 * depending on minor_version, it can be:
1158 * 0: At least 8K, but less than 12K, from end of device
1159 * 1: At start of device
1160 * 2: 4K from start of device.
1161 * Depending on the array size, we might leave extra space
1164 array_size
= __le64_to_cpu(sb
->size
);
1165 /* work out how much space we left for a bitmap */
1166 bm_space
= choose_bm_space(array_size
);
1168 switch(st
->minor_version
) {
1172 sb_offset
&= ~(4*2-1);
1173 sb
->super_offset
= __cpu_to_le64(sb_offset
);
1174 sb
->data_offset
= __cpu_to_le64(0);
1175 if (sb_offset
< array_size
+ bm_space
)
1176 bm_space
= sb_offset
- array_size
;
1177 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
1180 sb
->super_offset
= __cpu_to_le64(0);
1181 reserved
= bm_space
+ 4*2;
1182 /* Try for multiple of 1Meg so it is nicely aligned */
1183 #define ONE_MEG (2*1024)
1184 reserved
= ((reserved
+ ONE_MEG
-1)/ONE_MEG
) * ONE_MEG
;
1185 if (reserved
+ __le64_to_cpu(sb
->size
) > dsize
)
1186 reserved
= dsize
- __le64_to_cpu(sb
->size
);
1187 /* force 4K alignment */
1190 sb
->data_offset
= __cpu_to_le64(reserved
);
1191 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1195 sb
->super_offset
= __cpu_to_le64(4*2);
1196 if (4*2 + 4*2 + bm_space
+ __le64_to_cpu(sb
->size
)
1198 bm_space
= dsize
- __le64_to_cpu(sb
->size
)
1201 reserved
= bm_space
+ 4*2 + 4*2;
1202 /* Try for multiple of 1Meg so it is nicely aligned */
1203 #define ONE_MEG (2*1024)
1204 reserved
= ((reserved
+ ONE_MEG
-1)/ONE_MEG
) * ONE_MEG
;
1205 if (reserved
+ __le64_to_cpu(sb
->size
) > dsize
)
1206 reserved
= dsize
- __le64_to_cpu(sb
->size
);
1207 /* force 4K alignment */
1210 sb
->data_offset
= __cpu_to_le64(reserved
);
1211 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1214 fprintf(stderr
, Name
": Failed to write invalid "
1215 "metadata format 1.%i to %s\n",
1216 st
->minor_version
, di
->devname
);
1222 sb
->sb_csum
= calc_sb_1_csum(sb
);
1223 rv
= store_super1(st
, di
->fd
);
1224 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
1225 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
1231 fprintf(stderr
, Name
": Failed to write metadata to %s\n",
1238 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
1242 * 0 same, or first was empty, and second was copied
1243 * 1 second had wrong number
1245 * 3 wrong other info
1247 struct mdp_superblock_1
*first
= st
->sb
;
1248 struct mdp_superblock_1
*second
= tst
->sb
;
1250 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
1252 if (second
->major_version
!= __cpu_to_le32(1))
1256 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
1257 fprintf(stderr
, Name
1258 ": %s could not allocate superblock\n", __func__
);
1261 memcpy(first
, second
, SUPER1_SIZE
);
1265 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
1268 if (first
->ctime
!= second
->ctime
||
1269 first
->level
!= second
->level
||
1270 first
->layout
!= second
->layout
||
1271 first
->size
!= second
->size
||
1272 first
->chunksize
!= second
->chunksize
||
1273 first
->raid_disks
!= second
->raid_disks
)
1278 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
1280 unsigned long long dsize
;
1281 unsigned long long sb_offset
;
1282 struct mdp_superblock_1
*super
;
1284 struct bitmap_super_s
*bsb
;
1285 struct misc_dev_info
*misc
;
1286 struct align_fd afd
;
1292 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1294 struct supertype tst
;
1295 __u64 bestctime
= 0;
1296 /* guess... choose latest ctime */
1297 memset(&tst
, 0, sizeof(tst
));
1299 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1300 switch(load_super1(&tst
, fd
, devname
)) {
1301 case 0: super
= tst
.sb
;
1302 if (bestvers
== -1 ||
1303 bestctime
< __le64_to_cpu(super
->ctime
)) {
1304 bestvers
= tst
.minor_version
;
1305 bestctime
= __le64_to_cpu(super
->ctime
);
1310 case 1: return 1; /*bad device */
1311 case 2: break; /* bad, try next */
1314 if (bestvers
!= -1) {
1316 tst
.minor_version
= bestvers
;
1318 tst
.max_devs
= MAX_DEVS
;
1319 rv
= load_super1(&tst
, fd
, devname
);
1326 if (!get_dev_size(fd
, devname
, &dsize
))
1332 fprintf(stderr
, Name
": %s is too small for md: size is %llu sectors.\n",
1338 * Calculate the position of the superblock.
1339 * It is always aligned to a 4K boundary and
1340 * depending on minor_version, it can be:
1341 * 0: At least 8K, but less than 12K, from end of device
1342 * 1: At start of device
1343 * 2: 4K from start of device.
1345 switch(st
->minor_version
) {
1349 sb_offset
&= ~(4*2-1);
1361 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
1364 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1366 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
1367 devname
, strerror(errno
));
1371 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
1372 fprintf(stderr
, Name
": %s could not allocate superblock\n",
1377 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
1379 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
1385 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1387 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
1388 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1393 if (__le32_to_cpu(super
->major_version
) != 1) {
1395 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
1396 devname
, __le32_to_cpu(super
->major_version
));
1400 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1402 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
1409 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1411 misc
= (struct misc_dev_info
*) (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
1412 misc
->device_size
= dsize
;
1414 /* Now check on the bitmap superblock */
1415 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1417 /* Read the bitmap superblock and make sure it looks
1418 * valid. If it doesn't clear the bit. An --assemble --force
1419 * should get that written out.
1421 locate_bitmap1(st
, fd
);
1422 if (aread(&afd
, bsb
, 512) != 512)
1425 uuid_from_super1(st
, uuid
);
1426 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1427 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1432 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
1433 & ~MD_FEATURE_BITMAP_OFFSET
);
1438 static struct supertype
*match_metadata_desc1(char *arg
)
1440 struct supertype
*st
= calloc(1, sizeof(*st
));
1444 st
->container_dev
= NoMdDev
;
1446 st
->max_devs
= MAX_DEVS
;
1448 /* leading zeros can be safely ignored. --detail generates them. */
1451 if (strcmp(arg
, "1.0") == 0 ||
1452 strcmp(arg
, "1.00") == 0) {
1453 st
->minor_version
= 0;
1456 if (strcmp(arg
, "1.1") == 0 ||
1457 strcmp(arg
, "1.01") == 0
1459 st
->minor_version
= 1;
1462 if (strcmp(arg
, "1.2") == 0 ||
1463 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
1464 strcmp(arg
, "default") == 0 ||
1465 #endif /* DEFAULT_OLD_METADATA */
1466 strcmp(arg
, "1.02") == 0) {
1467 st
->minor_version
= 2;
1470 if (strcmp(arg
, "1") == 0 ||
1471 strcmp(arg
, "default") == 0) {
1472 st
->minor_version
= -1;
1480 /* find available size on device with this devsize, using
1481 * superblock type st, and reserving 'reserve' sectors for
1484 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1486 struct mdp_superblock_1
*super
= st
->sb
;
1491 /* creating: allow suitable space for bitmap */
1492 devsize
-= choose_bm_space(devsize
);
1494 else if (__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
1495 /* hot-add. allow for actual size of bitmap */
1496 struct bitmap_super_s
*bsb
;
1497 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1498 devsize
-= bitmap_sectors(bsb
);
1502 if (st
->minor_version
< 0)
1503 /* not specified, so time to set default */
1504 st
->minor_version
= 2;
1505 if (super
== NULL
&& st
->minor_version
> 0) {
1506 /* haven't committed to a size yet, so allow some
1507 * slack for alignment of data_offset.
1508 * We haven't access to device details so allow
1509 * 1 Meg if bigger than 1Gig
1511 if (devsize
> 1024*1024*2)
1514 switch(st
->minor_version
) {
1517 return ((devsize
- 8*2 ) & ~(4*2-1));
1519 /* at start, 4K for superblock and possible bitmap */
1520 return devsize
- 4*2;
1522 /* 4k from start, 4K for superblock and possible bitmap */
1523 return devsize
- (4+4)*2;
1529 add_internal_bitmap1(struct supertype
*st
,
1530 int *chunkp
, int delay
, int write_behind
,
1531 unsigned long long size
,
1532 int may_change
, int major
)
1535 * If not may_change, then this is a 'Grow' without sysfs support for
1536 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
1537 * If may_change, then this is create or a Grow with sysfs syupport,
1538 * and we can put the bitmap wherever we like.
1540 * size is in sectors, chunk is in bytes !!!
1543 unsigned long long bits
;
1544 unsigned long long max_bits
;
1545 unsigned long long min_chunk
;
1547 unsigned long long chunk
= *chunkp
;
1550 struct mdp_superblock_1
*sb
= st
->sb
;
1551 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
1554 if (__le64_to_cpu(sb
->data_size
) == 0)
1555 /* Must be creating the array, else data_size would be non-zero */
1557 switch(st
->minor_version
) {
1559 /* either 3K after the superblock (when hot-add),
1560 * or some amount of space before.
1563 /* We are creating array, so we *know* how much room has
1567 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1569 room
= __le64_to_cpu(sb
->super_offset
)
1570 - __le64_to_cpu(sb
->data_offset
)
1571 - __le64_to_cpu(sb
->data_size
);
1573 if (!may_change
|| (room
< 3*2 &&
1574 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1578 offset
= 0; /* means movable offset */
1583 case 2: /* between superblock and data */
1586 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1588 room
= __le64_to_cpu(sb
->data_offset
)
1589 - __le64_to_cpu(sb
->super_offset
);
1591 room
-= 2; /* Leave 1K for superblock */
1594 room
-= 4*2; /* leave 4K for superblock */
1603 if (chunk
== UnSet
&& room
> 128*2)
1604 /* Limit to 128K of bitmap when chunk size not requested */
1607 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1609 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1610 bits
= (size
*512)/min_chunk
+1;
1611 while (bits
> max_bits
) {
1615 if (chunk
== UnSet
) {
1616 /* For practical purpose, 64Meg is a good
1617 * default chunk size for internal bitmaps.
1620 if (chunk
< 64*1024*1024)
1621 chunk
= 64*1024*1024;
1622 } else if (chunk
< min_chunk
)
1623 return 0; /* chunk size too small */
1624 if (chunk
== 0) /* rounding problem */
1628 /* start bitmap on a 4K boundary with enough space for
1631 bits
= (size
*512) / chunk
+ 1;
1632 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
1633 room
*= 8; /* convert 4K blocks to sectors */
1637 sb
->bitmap_offset
= __cpu_to_le32(offset
);
1639 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
)
1640 | MD_FEATURE_BITMAP_OFFSET
);
1641 memset(bms
, 0, sizeof(*bms
));
1642 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1643 bms
->version
= __cpu_to_le32(major
);
1644 uuid_from_super1(st
, uuid
);
1645 memcpy(bms
->uuid
, uuid
, 16);
1646 bms
->chunksize
= __cpu_to_le32(chunk
);
1647 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1648 bms
->sync_size
= __cpu_to_le64(size
);
1649 bms
->write_behind
= __cpu_to_le32(write_behind
);
1655 static void locate_bitmap1(struct supertype
*st
, int fd
)
1657 unsigned long long offset
;
1658 struct mdp_superblock_1
*sb
;
1662 if (st
->ss
->load_super(st
, fd
, NULL
))
1663 return; /* no error I hope... */
1668 offset
= __le64_to_cpu(sb
->super_offset
);
1669 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
1672 lseek64(fd
, offset
<<9, 0);
1675 static int write_bitmap1(struct supertype
*st
, int fd
)
1677 struct mdp_superblock_1
*sb
= st
->sb
;
1678 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
1682 struct align_fd afd
;
1686 locate_bitmap1(st
, fd
);
1688 if (posix_memalign(&buf
, 4096, 4096))
1691 memset(buf
, 0xff, 4096);
1692 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
1694 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1695 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1696 towrite
+= sizeof(bitmap_super_t
);
1697 towrite
= ROUND_UP(towrite
, 512);
1698 while (towrite
> 0) {
1702 n
= awrite(&afd
, buf
, n
);
1707 memset(buf
, 0xff, 4096);
1717 static void free_super1(struct supertype
*st
)
1722 struct devinfo
*di
= st
->info
;
1723 st
->info
= di
->next
;
1732 static int validate_geometry1(struct supertype
*st
, int level
,
1733 int layout
, int raiddisks
,
1734 int *chunk
, unsigned long long size
,
1735 char *subdev
, unsigned long long *freesize
,
1738 unsigned long long ldsize
;
1741 if (level
== LEVEL_CONTAINER
) {
1743 fprintf(stderr
, Name
": 1.x metadata does not support containers\n");
1746 if (chunk
&& *chunk
== UnSet
)
1747 *chunk
= DEFAULT_CHUNK
;
1752 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1755 fprintf(stderr
, Name
": super1.x cannot open %s: %s\n",
1756 subdev
, strerror(errno
));
1760 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1766 *freesize
= avail_size1(st
, ldsize
>> 9);
1769 #endif /* MDASSEMBLE */
1771 struct superswitch super1
= {
1773 .examine_super
= examine_super1
,
1774 .brief_examine_super
= brief_examine_super1
,
1775 .export_examine_super
= export_examine_super1
,
1776 .detail_super
= detail_super1
,
1777 .brief_detail_super
= brief_detail_super1
,
1778 .export_detail_super
= export_detail_super1
,
1779 .write_init_super
= write_init_super1
,
1780 .validate_geometry
= validate_geometry1
,
1781 .add_to_super
= add_to_super1
,
1783 .match_home
= match_home1
,
1784 .uuid_from_super
= uuid_from_super1
,
1785 .getinfo_super
= getinfo_super1
,
1786 .container_content
= container_content1
,
1787 .update_super
= update_super1
,
1788 .init_super
= init_super1
,
1789 .store_super
= store_super1
,
1790 .compare_super
= compare_super1
,
1791 .load_super
= load_super1
,
1792 .match_metadata_desc
= match_metadata_desc1
,
1793 .avail_size
= avail_size1
,
1794 .add_internal_bitmap
= add_internal_bitmap1
,
1795 .locate_bitmap
= locate_bitmap1
,
1796 .write_bitmap
= write_bitmap1
,
1797 .free_super
= free_super1
,
1798 #if __BYTE_ORDER == BIG_ENDIAN