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 /* bad block log. If there are any bad blocks the feature flag is set.
74 * if offset and size are non-zero, that space is reserved and available.
76 __u8 bblog_shift
; /* shift from sectors to block size for badblocklist */
77 __u16 bblog_size
; /* number of sectors reserved for badblocklist */
78 __u32 bblog_offset
; /* sector offset from superblock to bblog, signed */
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 dev_roles[max_dev] */
85 __u32 max_dev
; /* size of dev_roles[] 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 #define MAX_SB_SIZE 4096
98 /* bitmap super size is 256, but we round up to a sector for alignment */
99 #define BM_SUPER_SIZE 512
100 #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2)
101 #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \
102 + sizeof(struct misc_dev_info))
104 struct misc_dev_info
{
108 /* feature_map bits */
109 #define MD_FEATURE_BITMAP_OFFSET 1
110 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
113 #define MD_FEATURE_RESHAPE_ACTIVE 4
114 #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */
116 #define MD_FEATURE_ALL (1|2|4|8)
119 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
121 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
123 unsigned int disk_csum
, csum
;
124 unsigned long long newcsum
;
125 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
126 unsigned int *isuper
= (unsigned int*)sb
;
128 /* make sure I can count... */
129 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
130 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
131 sizeof(struct mdp_superblock_1
) != 256) {
132 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
135 disk_csum
= sb
->sb_csum
;
138 for (; size
>=4; size
-= 4 ) {
139 newcsum
+= __le32_to_cpu(*isuper
);
144 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
146 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
147 sb
->sb_csum
= disk_csum
;
148 return __cpu_to_le32(csum
);
152 * Information related to file descriptor used for aligned reads/writes.
153 * Cache the block size.
160 static void init_afd(struct align_fd
*afd
, int fd
)
164 if (ioctl(afd
->fd
, BLKSSZGET
, &afd
->blk_sz
) != 0)
168 static char abuf
[4096+4096];
169 static int aread(struct align_fd
*afd
, void *buf
, int len
)
172 * On devices with a 4K sector size, we need to read
173 * the full sector and copy relevant bits into
182 if (!bsize
|| bsize
> 4096 || len
> 4096) {
184 fprintf(stderr
, "WARNING - aread() called with "
185 "invalid block size\n");
188 b
= ROUND_UP_PTR((char *)abuf
, 4096);
190 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
192 n
= read(afd
->fd
, b
, iosize
);
195 lseek(afd
->fd
, len
- n
, 1);
202 static int awrite(struct align_fd
*afd
, void *buf
, int len
)
205 * On devices with a 4K sector size, we need to write
206 * the full sector. We pre-read if the sector is larger
208 * The address must be sector-aligned.
215 if (!bsize
|| bsize
> 4096 || len
> 4096) {
217 fprintf(stderr
, "WARNING - awrite() called with "
218 "invalid block size\n");
221 b
= ROUND_UP_PTR((char *)abuf
, 4096);
223 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
227 n
= read(afd
->fd
, b
, iosize
);
230 lseek(afd
->fd
, -n
, 1);
234 n
= write(afd
->fd
, b
, iosize
);
237 lseek(afd
->fd
, len
- n
, 1);
242 static void examine_super1(struct supertype
*st
, char *homehost
)
244 struct mdp_superblock_1
*sb
= st
->sb
;
251 int l
= homehost
? strlen(homehost
) : 0;
253 unsigned long long sb_offset
;
255 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
256 printf(" Version : 1");
257 sb_offset
= __le64_to_cpu(sb
->super_offset
);
260 else if (sb_offset
<= 8)
264 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
265 printf(" Array UUID : ");
266 for (i
=0; i
<16; i
++) {
267 if ((i
&3)==0 && i
!= 0) printf(":");
268 printf("%02x", sb
->set_uuid
[i
]);
271 printf(" Name : %.32s", sb
->set_name
);
272 if (l
> 0 && l
< 32 &&
273 sb
->set_name
[l
] == ':' &&
274 strncmp(sb
->set_name
, homehost
, l
) == 0)
275 printf(" (local to host %s)", homehost
);
277 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
278 printf(" Creation Time : %.24s\n", ctime(&atime
));
279 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
280 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
281 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
283 printf(" Avail Dev Size : %llu%s\n",
284 (unsigned long long)__le64_to_cpu(sb
->data_size
),
285 human_size(__le64_to_cpu(sb
->data_size
)<<9));
286 if (__le32_to_cpu(sb
->level
) > 0) {
287 int ddsks
= 0, ddsks_denom
= 1;
288 switch(__le32_to_cpu(sb
->level
)) {
289 case 1: ddsks
=1;break;
291 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
292 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
294 layout
= __le32_to_cpu(sb
->layout
);
295 ddsks
= __le32_to_cpu(sb
->raid_disks
);
296 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
299 long long asize
= __le64_to_cpu(sb
->size
);
300 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
301 printf(" Array Size : %llu%s\n",
302 asize
>> 10, human_size(asize
));
304 if (sb
->size
!= sb
->data_size
)
305 printf(" Used Dev Size : %llu%s\n",
306 (unsigned long long)__le64_to_cpu(sb
->size
),
307 human_size(__le64_to_cpu(sb
->size
)<<9));
310 printf(" Data Offset : %llu sectors\n",
311 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
312 printf(" Super Offset : %llu sectors\n",
313 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
314 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
315 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
316 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
317 printf(" Device UUID : ");
318 for (i
=0; i
<16; i
++) {
319 if ((i
&3)==0 && i
!= 0) printf(":");
320 printf("%02x", sb
->device_uuid
[i
]);
324 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
325 printf("Internal Bitmap : %ld sectors from superblock\n",
326 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
328 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)) {
329 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
330 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
331 if (__le32_to_cpu(sb
->delta_disks
)) {
332 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
333 printf(" (%d->%d)\n",
334 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
335 __le32_to_cpu(sb
->raid_disks
));
336 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
337 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
339 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
340 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
341 printf(" New Level : %s\n", c
?c
:"-unknown-");
343 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
344 if (__le32_to_cpu(sb
->level
) == 5) {
345 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
346 printf(" New Layout : %s\n", c
?c
:"-unknown-");
348 if (__le32_to_cpu(sb
->level
) == 6) {
349 c
= map_num(r6layout
, __le32_to_cpu(sb
->new_layout
));
350 printf(" New Layout : %s\n", c
?c
:"-unknown-");
352 if (__le32_to_cpu(sb
->level
) == 10) {
353 printf(" New Layout :");
354 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
358 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
359 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
364 if (sb
->devflags
& WriteMostly1
)
365 printf(" write-mostly");
369 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
370 printf(" Update Time : %.24s\n", ctime(&atime
));
372 if (sb
->bblog_size
&& sb
->bblog_offset
) {
373 printf(" Bad Block Log : %d entries available at offset %ld sectors",
374 __le16_to_cpu(sb
->bblog_size
)*512/8,
375 (long)__le32_to_cpu(sb
->bblog_offset
));
376 if (sb
->feature_map
&
377 __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
378 printf(" - bad blocks present.");
383 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
384 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
386 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
387 __le32_to_cpu(calc_sb_1_csum(sb
)));
388 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
390 if (__le32_to_cpu(sb
->level
) == 5) {
391 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
392 printf(" Layout : %s\n", c
?c
:"-unknown-");
394 if (__le32_to_cpu(sb
->level
) == 6) {
395 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
396 printf(" Layout : %s\n", c
?c
:"-unknown-");
398 if (__le32_to_cpu(sb
->level
) == 10) {
399 int lo
= __le32_to_cpu(sb
->layout
);
401 print_r10_layout(lo
);
404 switch(__le32_to_cpu(sb
->level
)) {
410 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
413 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
419 /* This turns out to just be confusing */
420 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
421 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
422 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
424 for (d
=0; d
< i
; d
++) {
425 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
427 if (role
== 0xffff) printf("empty");
428 else if(role
== 0xfffe) printf("failed");
429 else printf("%d", role
);
433 printf(" Device Role : ");
434 d
= __le32_to_cpu(sb
->dev_number
);
435 if (d
< __le32_to_cpu(sb
->max_dev
))
436 role
= __le16_to_cpu(sb
->dev_roles
[d
]);
442 printf("Active device %d\n", role
);
444 printf(" Array State : ");
445 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
448 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
449 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
453 if (cnt
> 1) printf("?");
454 else if (cnt
== 1) printf("A");
458 /* This is confusing too */
460 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
461 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
465 if (faulty
) printf(" %d failed", faulty
);
467 printf(" ('A' == active, '.' == missing)");
472 static void brief_examine_super1(struct supertype
*st
, int verbose
)
474 struct mdp_superblock_1
*sb
= st
->sb
;
476 unsigned long long sb_offset
;
478 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
480 nm
= strchr(sb
->set_name
, ':');
483 else if (sb
->set_name
[0])
495 printf(" level=%s", c
);
496 sb_offset
= __le64_to_cpu(sb
->super_offset
);
498 printf(" metadata=1.1 ");
499 else if (sb_offset
<= 8)
500 printf(" metadata=1.2 ");
502 printf(" metadata=1.0 ");
504 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
506 for (i
=0; i
<16; i
++) {
507 if ((i
&3)==0 && i
!= 0) printf(":");
508 printf("%02x", sb
->set_uuid
[i
]);
510 if (sb
->set_name
[0]) {
512 print_quoted(sb
->set_name
);
517 static void export_examine_super1(struct supertype
*st
)
519 struct mdp_superblock_1
*sb
= st
->sb
;
524 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
525 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
527 if (sb
->set_name
[i
] == '\n' ||
528 sb
->set_name
[i
] == '\0') {
533 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
534 if (__le32_to_cpu(sb
->level
) > 0) {
535 int ddsks
= 0, ddsks_denom
= 1;
536 switch(__le32_to_cpu(sb
->level
)) {
537 case 1: ddsks
=1;break;
539 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
540 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
542 layout
= __le32_to_cpu(sb
->layout
);
543 ddsks
= __le32_to_cpu(sb
->raid_disks
);
544 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
547 long long asize
= __le64_to_cpu(sb
->size
);
548 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
549 printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize
,JEDEC
));
553 for (i
=0; i
<16; i
++) {
554 if ((i
&3)==0 && i
!= 0) printf(":");
555 printf("%02x", sb
->set_uuid
[i
]);
558 printf("MD_UPDATE_TIME=%llu\n",
559 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
560 printf("MD_DEV_UUID=");
561 for (i
=0; i
<16; i
++) {
562 if ((i
&3)==0 && i
!= 0) printf(":");
563 printf("%02x", sb
->device_uuid
[i
]);
566 printf("MD_EVENTS=%llu\n",
567 (unsigned long long)__le64_to_cpu(sb
->events
));
570 static void detail_super1(struct supertype
*st
, char *homehost
)
572 struct mdp_superblock_1
*sb
= st
->sb
;
574 int l
= homehost
? strlen(homehost
) : 0;
576 printf(" Name : %.32s", sb
->set_name
);
577 if (l
> 0 && l
< 32 &&
578 sb
->set_name
[l
] == ':' &&
579 strncmp(sb
->set_name
, homehost
, l
) == 0)
580 printf(" (local to host %s)", homehost
);
581 printf("\n UUID : ");
582 for (i
=0; i
<16; i
++) {
583 if ((i
&3)==0 && i
!= 0) printf(":");
584 printf("%02x", sb
->set_uuid
[i
]);
586 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
589 static void brief_detail_super1(struct supertype
*st
)
591 struct mdp_superblock_1
*sb
= st
->sb
;
594 if (sb
->set_name
[0]) {
596 print_quoted(sb
->set_name
);
599 for (i
=0; i
<16; i
++) {
600 if ((i
&3)==0 && i
!= 0) printf(":");
601 printf("%02x", sb
->set_uuid
[i
]);
605 static void export_detail_super1(struct supertype
*st
)
607 struct mdp_superblock_1
*sb
= st
->sb
;
612 if (sb
->set_name
[i
] == '\n' ||
613 sb
->set_name
[i
] == '\0') {
618 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
623 static int match_home1(struct supertype
*st
, char *homehost
)
625 struct mdp_superblock_1
*sb
= st
->sb
;
626 int l
= homehost
? strlen(homehost
) : 0;
628 return (l
> 0 && l
< 32 &&
629 sb
->set_name
[l
] == ':' &&
630 strncmp(sb
->set_name
, homehost
, l
) == 0);
633 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
635 struct mdp_superblock_1
*super
= st
->sb
;
636 char *cuuid
= (char*)uuid
;
639 cuuid
[i
] = super
->set_uuid
[i
];
642 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
644 struct mdp_superblock_1
*sb
= st
->sb
;
648 unsigned int map_disks
= info
->array
.raid_disks
;
650 memset(info
, 0, sizeof(*info
));
651 info
->array
.major_version
= 1;
652 info
->array
.minor_version
= st
->minor_version
;
653 info
->array
.patch_version
= 0;
654 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
655 info
->array
.level
= __le32_to_cpu(sb
->level
);
656 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
657 info
->array
.md_minor
= -1;
658 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
659 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
660 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
662 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
)
665 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
666 info
->component_size
= __le64_to_cpu(sb
->size
);
667 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
))
668 info
->bitmap_offset
= (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
670 info
->disk
.major
= 0;
671 info
->disk
.minor
= 0;
672 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
673 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
674 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
677 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
679 info
->disk
.raid_disk
= -1;
682 info
->disk
.state
= 0; /* spare: not active, not sync, not faulty */
685 info
->disk
.state
= 1; /* faulty */
688 info
->disk
.state
= 6; /* active and in sync */
689 info
->disk
.raid_disk
= role
;
691 if (sb
->devflags
& WriteMostly1
)
692 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
693 info
->events
= __le64_to_cpu(sb
->events
);
694 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
695 info
->safe_mode_delay
= 200;
697 memcpy(info
->uuid
, sb
->set_uuid
, 16);
699 strncpy(info
->name
, sb
->set_name
, 32);
702 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
703 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
705 info
->recovery_start
= MaxSector
;
707 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
708 info
->reshape_active
= 1;
709 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
710 info
->new_level
= __le32_to_cpu(sb
->new_level
);
711 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
712 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
713 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
714 if (info
->delta_disks
< 0)
715 info
->array
.raid_disks
-= info
->delta_disks
;
717 info
->reshape_active
= 0;
719 info
->recovery_blocked
= info
->reshape_active
;
722 for (i
=0; i
<map_disks
; i
++)
724 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
725 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
726 if (/*role == 0xFFFF || */role
< (unsigned) info
->array
.raid_disks
) {
728 if (map
&& role
< map_disks
)
733 info
->array
.working_disks
= working
;
736 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
743 info
= xmalloc(sizeof(*info
));
744 getinfo_super1(st
, info
, NULL
);
748 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
750 char *devname
, int verbose
,
751 int uuid_set
, char *homehost
)
753 /* NOTE: for 'assemble' and 'force' we need to return non-zero
754 * if any change was made. For others, the return value is
758 struct mdp_superblock_1
*sb
= st
->sb
;
760 if (strcmp(update
, "force-one")==0) {
761 /* Not enough devices for a working array,
762 * so bring this one up-to-date
764 if (sb
->events
!= __cpu_to_le64(info
->events
))
766 sb
->events
= __cpu_to_le64(info
->events
);
767 } else if (strcmp(update
, "force-array")==0) {
768 /* Degraded array and 'force' requests to
769 * maybe need to mark it 'clean'.
771 switch(__le32_to_cpu(sb
->level
)) {
772 case 5: case 4: case 6:
773 /* need to force clean */
774 if (sb
->resync_offset
!= MaxSector
)
776 sb
->resync_offset
= MaxSector
;
778 } else if (strcmp(update
, "assemble")==0) {
779 int d
= info
->disk
.number
;
781 if (info
->disk
.state
== 6)
782 want
= info
->disk
.raid_disk
;
785 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
786 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
789 if (info
->reshape_active
&&
790 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
791 info
->delta_disks
>= 0 &&
792 info
->reshape_progress
< __le64_to_cpu(sb
->reshape_position
)) {
793 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
796 if (info
->reshape_active
&&
797 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
798 info
->delta_disks
< 0 &&
799 info
->reshape_progress
> __le64_to_cpu(sb
->reshape_position
)) {
800 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
803 } else if (strcmp(update
, "linear-grow-new") == 0) {
806 unsigned int max
= __le32_to_cpu(sb
->max_dev
);
808 for (i
=0 ; i
< max
; i
++)
809 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
811 sb
->dev_number
= __cpu_to_le32(i
);
812 info
->disk
.number
= i
;
813 if (max
>= __le32_to_cpu(sb
->max_dev
))
814 sb
->max_dev
= __cpu_to_le32(max
+1);
816 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
817 read(rfd
, sb
->device_uuid
, 16) != 16) {
818 __u32 r
[4] = {random(), random(), random(), random()};
819 memcpy(sb
->device_uuid
, r
, 16);
825 __cpu_to_le16(info
->disk
.raid_disk
);
827 fd
= open(devname
, O_RDONLY
);
829 unsigned long long ds
;
830 get_dev_size(fd
, devname
, &ds
);
833 if (__le64_to_cpu(sb
->super_offset
) <
834 __le64_to_cpu(sb
->data_offset
)) {
835 sb
->data_size
= __cpu_to_le64(
836 ds
- __le64_to_cpu(sb
->data_offset
));
839 ds
&= ~(unsigned long long)(4*2-1);
840 sb
->super_offset
= __cpu_to_le64(ds
);
841 sb
->data_size
= __cpu_to_le64(
842 ds
- __le64_to_cpu(sb
->data_offset
));
845 } else if (strcmp(update
, "linear-grow-update") == 0) {
846 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
847 sb
->dev_roles
[info
->disk
.number
] =
848 __cpu_to_le16(info
->disk
.raid_disk
);
849 } else if (strcmp(update
, "resync") == 0) {
850 /* make sure resync happens */
851 sb
->resync_offset
= 0ULL;
852 } else if (strcmp(update
, "uuid") == 0) {
853 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
855 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
856 struct bitmap_super_s
*bm
;
857 bm
= (struct bitmap_super_s
*)(st
->sb
+MAX_SB_SIZE
);
858 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
860 } else if (strcmp(update
, "no-bitmap") == 0) {
861 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
862 } else if (strcmp(update
, "bbl") == 0) {
863 /* only possible if there is room after the bitmap, or if
866 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
867 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
868 long bitmap_offset
= (long)__le64_to_cpu(sb
->bitmap_offset
);
872 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
873 struct bitmap_super_s
*bsb
;
874 bsb
= (struct bitmap_super_s
*)(((char*)sb
)+MAX_SB_SIZE
);
875 bm_sectors
= bitmap_sectors(bsb
);
878 if (sb_offset
< data_offset
) {
879 /* 1.1 or 1.2. Put bbl just before data
882 space
= data_offset
- sb_offset
;
883 bb_offset
= space
- 8;
884 if (bm_sectors
&& bitmap_offset
> 0)
885 space
-= (bitmap_offset
+ bm_sectors
);
887 space
-= 8; /* The superblock */
889 sb
->bblog_size
= __cpu_to_le16(8);
890 sb
->bblog_offset
= __cpu_to_le32(bb_offset
);
893 /* 1.0 - Put bbl just before super block */
894 if (bm_sectors
&& bitmap_offset
< 0)
895 space
= -bitmap_offset
- bm_sectors
;
897 space
= sb_offset
- data_offset
-
898 __le64_to_cpu(sb
->data_size
);
900 sb
->bblog_size
= __cpu_to_le16(8);
901 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
904 } else if (strcmp(update
, "no-bbl") == 0) {
905 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
906 pr_err("Cannot remove active bbl from %s\n",devname
);
910 sb
->bblog_offset
= 0;
912 } else if (strcmp(update
, "homehost") == 0 &&
916 c
= strchr(sb
->set_name
, ':');
918 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
920 strncpy(info
->name
, sb
->set_name
, 32);
922 } else if (strcmp(update
, "name") == 0) {
923 if (info
->name
[0] == 0)
924 sprintf(info
->name
, "%d", info
->array
.md_minor
);
925 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
927 strchr(info
->name
, ':') == NULL
&&
928 strlen(homehost
)+1+strlen(info
->name
) < 32) {
929 strcpy(sb
->set_name
, homehost
);
930 strcat(sb
->set_name
, ":");
931 strcat(sb
->set_name
, info
->name
);
933 strcpy(sb
->set_name
, info
->name
);
934 } else if (strcmp(update
, "devicesize") == 0 &&
935 __le64_to_cpu(sb
->super_offset
) <
936 __le64_to_cpu(sb
->data_offset
)) {
937 /* set data_size to device size less data_offset */
938 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
939 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
940 printf("Size was %llu\n", (unsigned long long)
941 __le64_to_cpu(sb
->data_size
));
942 sb
->data_size
= __cpu_to_le64(
943 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
944 printf("Size is %llu\n", (unsigned long long)
945 __le64_to_cpu(sb
->data_size
));
946 } else if (strcmp(update
, "_reshape_progress")==0)
947 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
948 else if (strcmp(update
, "writemostly")==0)
949 sb
->devflags
|= WriteMostly1
;
950 else if (strcmp(update
, "readwrite")==0)
951 sb
->devflags
&= ~WriteMostly1
;
955 sb
->sb_csum
= calc_sb_1_csum(sb
);
959 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
960 unsigned long long size
, char *name
, char *homehost
,
961 int *uuid
, unsigned long long data_offset
)
963 struct mdp_superblock_1
*sb
;
969 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
970 pr_err("%s could not allocate superblock\n", __func__
);
973 memset(sb
, 0, SUPER1_SIZE
);
977 /* zeroing superblock */
981 spares
= info
->working_disks
- info
->active_disks
;
982 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
983 pr_err("too many devices requested: %d+%d > %d\n",
984 info
->raid_disks
, spares
, MAX_DEVS
);
988 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
989 sb
->major_version
= __cpu_to_le32(1);
994 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
996 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
997 read(rfd
, sb
->set_uuid
, 16) != 16) {
998 __u32 r
[4] = {random(), random(), random(), random()};
999 memcpy(sb
->set_uuid
, r
, 16);
1001 if (rfd
>= 0) close(rfd
);
1004 if (name
== NULL
|| *name
== 0) {
1005 sprintf(defname
, "%d", info
->md_minor
);
1009 strchr(name
, ':')== NULL
&&
1010 strlen(homehost
)+1+strlen(name
) < 32) {
1011 strcpy(sb
->set_name
, homehost
);
1012 strcat(sb
->set_name
, ":");
1013 strcat(sb
->set_name
, name
);
1015 strcpy(sb
->set_name
, name
);
1017 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
1018 sb
->level
= __cpu_to_le32(info
->level
);
1019 sb
->layout
= __cpu_to_le32(info
->layout
);
1020 sb
->size
= __cpu_to_le64(size
*2ULL);
1021 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
1022 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
1024 sb
->data_offset
= __cpu_to_le64(data_offset
);
1025 sb
->data_size
= __cpu_to_le64(0);
1026 sb
->super_offset
= __cpu_to_le64(0);
1027 sb
->recovery_offset
= __cpu_to_le64(0);
1029 sb
->utime
= sb
->ctime
;
1030 sb
->events
= __cpu_to_le64(1);
1031 if (info
->state
& (1<<MD_SB_CLEAN
))
1032 sb
->resync_offset
= MaxSector
;
1034 sb
->resync_offset
= 0;
1035 sbsize
= sizeof(struct mdp_superblock_1
) + 2 * (info
->raid_disks
+ spares
);
1036 sbsize
= ROUND_UP(sbsize
, 512);
1037 sb
->max_dev
= __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
1039 memset(sb
->dev_roles
, 0xff, MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
1047 mdu_disk_info_t disk
;
1048 struct devinfo
*next
;
1051 /* Add a device to the superblock being created */
1052 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
1053 int fd
, char *devname
)
1055 struct mdp_superblock_1
*sb
= st
->sb
;
1056 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
1057 struct devinfo
*di
, **dip
;
1059 if ((dk
->state
& 6) == 6) /* active, sync */
1060 *rp
= __cpu_to_le16(dk
->raid_disk
);
1061 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
1066 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
1067 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
1068 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
1070 sb
->dev_number
= __cpu_to_le32(dk
->number
);
1071 sb
->devflags
= 0; /* don't copy another disks flags */
1072 sb
->sb_csum
= calc_sb_1_csum(sb
);
1074 dip
= (struct devinfo
**)&st
->info
;
1076 dip
= &(*dip
)->next
;
1077 di
= xmalloc(sizeof(struct devinfo
));
1079 di
->devname
= devname
;
1088 static void locate_bitmap1(struct supertype
*st
, int fd
);
1090 static int store_super1(struct supertype
*st
, int fd
)
1092 struct mdp_superblock_1
*sb
= st
->sb
;
1093 unsigned long long sb_offset
;
1094 struct align_fd afd
;
1096 unsigned long long dsize
;
1098 if (!get_dev_size(fd
, NULL
, &dsize
))
1109 * Calculate the position of the superblock.
1110 * It is always aligned to a 4K boundary and
1111 * depending on minor_version, it can be:
1112 * 0: At least 8K, but less than 12K, from end of device
1113 * 1: At start of device
1114 * 2: 4K from start of device.
1116 switch(st
->minor_version
) {
1120 sb_offset
&= ~(4*2-1);
1134 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1135 0 != __le64_to_cpu(sb
->super_offset
)
1137 pr_err("internal error - sb_offset is wrong\n");
1141 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1144 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1146 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1149 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1150 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1151 (((char*)sb
)+MAX_SB_SIZE
);
1152 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1153 locate_bitmap1(st
, fd
);
1154 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1162 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1164 static unsigned long choose_bm_space(unsigned long devsize
)
1166 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1167 * if bigger than 200Gig, save 128k
1168 * NOTE: result must be multiple of 4K else bad things happen
1169 * on 4K-sector devices.
1171 if (devsize
< 64*2) return 0;
1172 if (devsize
- 64*2 >= 200*1024*1024*2)
1174 if (devsize
- 4*2 > 8*1024*1024*2)
1179 static void free_super1(struct supertype
*st
);
1182 static int write_init_super1(struct supertype
*st
)
1184 struct mdp_superblock_1
*sb
= st
->sb
;
1185 struct supertype
*refst
;
1188 unsigned long long bm_space
;
1189 unsigned long long reserved
;
1191 unsigned long long dsize
, array_size
;
1192 unsigned long long sb_offset
, headroom
;
1193 unsigned long long data_offset
;
1195 for (di
= st
->info
; di
; di
= di
->next
) {
1196 if (di
->disk
.state
== 1)
1201 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
1204 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1205 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1206 sb
->devflags
|= WriteMostly1
;
1208 sb
->devflags
&= ~WriteMostly1
;
1210 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1211 read(rfd
, sb
->device_uuid
, 16) != 16) {
1212 __u32 r
[4] = {random(), random(), random(), random()};
1213 memcpy(sb
->device_uuid
, r
, 16);
1220 refst
= dup_super(st
);
1221 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1222 struct mdp_superblock_1
*refsb
= refst
->sb
;
1224 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1225 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
1226 /* same array, so preserve events and
1228 sb
->events
= refsb
->events
;
1229 /* bugs in 2.6.17 and earlier mean the
1230 * dev_number chosen in Manage must be preserved
1232 if (get_linux_version() >= 2006018)
1233 sb
->dev_number
= refsb
->dev_number
;
1239 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
1253 * Calculate the position of the superblock.
1254 * It is always aligned to a 4K boundary and
1255 * depending on minor_version, it can be:
1256 * 0: At least 8K, but less than 12K, from end of device
1257 * 1: At start of device
1258 * 2: 4K from start of device.
1259 * Depending on the array size, we might leave extra space
1261 * Also leave 4K for bad-block log.
1263 array_size
= __le64_to_cpu(sb
->size
);
1264 /* work out how much space we left for a bitmap,
1265 * Add 8 sectors for bad block log */
1266 bm_space
= choose_bm_space(array_size
) + 8;
1268 /* We try to leave 0.1% at the start for reshape
1269 * operations, but limit this to 128Meg (0.1% of 10Gig)
1270 * which is plenty for efficient reshapes
1272 headroom
= 128 * 1024 * 2;
1273 while (headroom
<< 10 > array_size
)
1276 switch(st
->minor_version
) {
1280 sb_offset
&= ~(4*2-1);
1281 sb
->super_offset
= __cpu_to_le64(sb_offset
);
1282 data_offset
= __le64_to_cpu(sb
->data_offset
);
1283 if (data_offset
== INVALID_SECTORS
)
1284 sb
->data_offset
= 0;
1285 if (sb_offset
< array_size
+ bm_space
)
1286 bm_space
= sb_offset
- array_size
;
1287 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
1288 if (bm_space
>= 8) {
1289 sb
->bblog_size
= __cpu_to_le16(8);
1290 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
1294 sb
->super_offset
= __cpu_to_le64(0);
1295 data_offset
= __le64_to_cpu(sb
->data_offset
);
1296 if (data_offset
== INVALID_SECTORS
) {
1297 reserved
= bm_space
+ 4*2;
1298 if (reserved
< headroom
)
1299 reserved
= headroom
;
1300 if (reserved
+ array_size
> dsize
)
1301 reserved
= dsize
- array_size
;
1302 /* Try for multiple of 1Meg so it is nicely aligned */
1303 #define ONE_MEG (2*1024)
1304 if (reserved
> ONE_MEG
)
1305 reserved
= (reserved
/ONE_MEG
) * ONE_MEG
;
1307 /* force 4K alignment */
1311 reserved
= data_offset
;
1313 sb
->data_offset
= __cpu_to_le64(reserved
);
1314 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1315 if (reserved
>= 16) {
1316 sb
->bblog_size
= __cpu_to_le16(8);
1317 sb
->bblog_offset
= __cpu_to_le32(reserved
-8);
1322 sb
->super_offset
= __cpu_to_le64(4*2);
1323 data_offset
= __le64_to_cpu(sb
->data_offset
);
1324 if (data_offset
== INVALID_SECTORS
) {
1325 if (4*2 + 4*2 + bm_space
+ array_size
1327 bm_space
= dsize
- array_size
1330 reserved
= bm_space
+ 4*2 + 4*2;
1331 if (reserved
< headroom
)
1332 reserved
= headroom
;
1333 if (reserved
+ array_size
> dsize
)
1334 reserved
= dsize
- array_size
;
1335 /* Try for multiple of 1Meg so it is nicely aligned */
1336 #define ONE_MEG (2*1024)
1337 if (reserved
> ONE_MEG
)
1338 reserved
= (reserved
/ONE_MEG
) * ONE_MEG
;
1340 /* force 4K alignment */
1344 reserved
= data_offset
;
1346 sb
->data_offset
= __cpu_to_le64(reserved
);
1347 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1348 if (reserved
>= 16+16) {
1349 sb
->bblog_size
= __cpu_to_le16(8);
1350 /* '8' sectors for the bblog, and another '8'
1351 * because we want offset from superblock, not
1354 sb
->bblog_offset
= __cpu_to_le32(reserved
-8-8);
1358 pr_err("Failed to write invalid "
1359 "metadata format 1.%i to %s\n",
1360 st
->minor_version
, di
->devname
);
1365 sb
->sb_csum
= calc_sb_1_csum(sb
);
1366 rv
= store_super1(st
, di
->fd
);
1367 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
1368 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
1376 pr_err("Failed to write metadata to %s\n",
1383 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
1387 * 0 same, or first was empty, and second was copied
1388 * 1 second had wrong number
1390 * 3 wrong other info
1392 struct mdp_superblock_1
*first
= st
->sb
;
1393 struct mdp_superblock_1
*second
= tst
->sb
;
1395 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
1397 if (second
->major_version
!= __cpu_to_le32(1))
1401 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
1402 pr_err("%s could not allocate superblock\n", __func__
);
1405 memcpy(first
, second
, SUPER1_SIZE
);
1409 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
1412 if (first
->ctime
!= second
->ctime
||
1413 first
->level
!= second
->level
||
1414 first
->layout
!= second
->layout
||
1415 first
->size
!= second
->size
||
1416 first
->chunksize
!= second
->chunksize
||
1417 first
->raid_disks
!= second
->raid_disks
)
1422 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
1424 unsigned long long dsize
;
1425 unsigned long long sb_offset
;
1426 struct mdp_superblock_1
*super
;
1428 struct bitmap_super_s
*bsb
;
1429 struct misc_dev_info
*misc
;
1430 struct align_fd afd
;
1436 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1438 struct supertype tst
;
1439 __u64 bestctime
= 0;
1440 /* guess... choose latest ctime */
1441 memset(&tst
, 0, sizeof(tst
));
1443 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1444 switch(load_super1(&tst
, fd
, devname
)) {
1445 case 0: super
= tst
.sb
;
1446 if (bestvers
== -1 ||
1447 bestctime
< __le64_to_cpu(super
->ctime
)) {
1448 bestvers
= tst
.minor_version
;
1449 bestctime
= __le64_to_cpu(super
->ctime
);
1454 case 1: return 1; /*bad device */
1455 case 2: break; /* bad, try next */
1458 if (bestvers
!= -1) {
1460 tst
.minor_version
= bestvers
;
1462 tst
.max_devs
= MAX_DEVS
;
1463 rv
= load_super1(&tst
, fd
, devname
);
1470 if (!get_dev_size(fd
, devname
, &dsize
))
1476 pr_err("%s is too small for md: size is %llu sectors.\n",
1482 * Calculate the position of the superblock.
1483 * It is always aligned to a 4K boundary and
1484 * depending on minor_version, it can be:
1485 * 0: At least 8K, but less than 12K, from end of device
1486 * 1: At start of device
1487 * 2: 4K from start of device.
1489 switch(st
->minor_version
) {
1493 sb_offset
&= ~(4*2-1);
1505 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
1508 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1510 pr_err("Cannot seek to superblock on %s: %s\n",
1511 devname
, strerror(errno
));
1515 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
1516 pr_err("%s could not allocate superblock\n",
1521 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
1523 pr_err("Cannot read superblock on %s\n",
1529 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1531 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1532 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1537 if (__le32_to_cpu(super
->major_version
) != 1) {
1539 pr_err("Cannot interpret superblock on %s - version is %d\n",
1540 devname
, __le32_to_cpu(super
->major_version
));
1544 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1546 pr_err("No superblock found on %s (super_offset is wrong)\n",
1553 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1555 misc
= (struct misc_dev_info
*) (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
1556 misc
->device_size
= dsize
;
1558 /* Now check on the bitmap superblock */
1559 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1561 /* Read the bitmap superblock and make sure it looks
1562 * valid. If it doesn't clear the bit. An --assemble --force
1563 * should get that written out.
1565 locate_bitmap1(st
, fd
);
1566 if (aread(&afd
, bsb
, 512) != 512)
1569 uuid_from_super1(st
, uuid
);
1570 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1571 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1576 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
1577 & ~MD_FEATURE_BITMAP_OFFSET
);
1582 static struct supertype
*match_metadata_desc1(char *arg
)
1584 struct supertype
*st
= xcalloc(1, sizeof(*st
));
1586 st
->container_dev
= NoMdDev
;
1588 st
->max_devs
= MAX_DEVS
;
1590 /* leading zeros can be safely ignored. --detail generates them. */
1593 if (strcmp(arg
, "1.0") == 0 ||
1594 strcmp(arg
, "1.00") == 0) {
1595 st
->minor_version
= 0;
1598 if (strcmp(arg
, "1.1") == 0 ||
1599 strcmp(arg
, "1.01") == 0
1601 st
->minor_version
= 1;
1604 if (strcmp(arg
, "1.2") == 0 ||
1605 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
1606 strcmp(arg
, "default") == 0 ||
1607 #endif /* DEFAULT_OLD_METADATA */
1608 strcmp(arg
, "1.02") == 0) {
1609 st
->minor_version
= 2;
1612 if (strcmp(arg
, "1") == 0 ||
1613 strcmp(arg
, "default") == 0) {
1614 st
->minor_version
= -1;
1622 /* find available size on device with this devsize, using
1623 * superblock type st, and reserving 'reserve' sectors for
1626 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
,
1627 unsigned long long data_offset
)
1629 struct mdp_superblock_1
*super
= st
->sb
;
1635 /* creating: allow suitable space for bitmap */
1636 bmspace
= choose_bm_space(devsize
);
1638 else if (__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
1639 /* hot-add. allow for actual size of bitmap */
1640 struct bitmap_super_s
*bsb
;
1641 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1642 bmspace
= bitmap_sectors(bsb
);
1645 /* Allow space for bad block log */
1646 if (super
&& super
->bblog_size
)
1647 devsize
-= __le16_to_cpu(super
->bblog_size
);
1652 if (st
->minor_version
< 0)
1653 /* not specified, so time to set default */
1654 st
->minor_version
= 2;
1656 if (data_offset
!= INVALID_SECTORS
)
1657 switch(st
->minor_version
) {
1659 return devsize
- data_offset
- 8*2;
1662 return devsize
- data_offset
;
1669 if (super
== NULL
&& st
->minor_version
> 0) {
1670 /* haven't committed to a size yet, so allow some
1671 * slack for space for reshape.
1672 * Limit slack to 128M, but aim for about 0.1%
1674 unsigned long long headroom
= 128*1024*2;
1675 while ((headroom
<< 10) > devsize
)
1677 devsize
-= headroom
;
1679 switch(st
->minor_version
) {
1682 return ((devsize
- 8*2 ) & ~(4*2-1));
1684 /* at start, 4K for superblock and possible bitmap */
1685 return devsize
- 4*2;
1687 /* 4k from start, 4K for superblock and possible bitmap */
1688 return devsize
- (4+4)*2;
1694 add_internal_bitmap1(struct supertype
*st
,
1695 int *chunkp
, int delay
, int write_behind
,
1696 unsigned long long size
,
1697 int may_change
, int major
)
1700 * If not may_change, then this is a 'Grow' without sysfs support for
1701 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
1702 * If may_change, then this is create or a Grow with sysfs syupport,
1703 * and we can put the bitmap wherever we like.
1705 * size is in sectors, chunk is in bytes !!!
1708 unsigned long long bits
;
1709 unsigned long long max_bits
;
1710 unsigned long long min_chunk
;
1712 long bbl_offset
, bbl_size
;
1713 unsigned long long chunk
= *chunkp
;
1716 struct mdp_superblock_1
*sb
= st
->sb
;
1717 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
1721 if (__le64_to_cpu(sb
->data_size
) == 0)
1722 /* Must be creating the array, else data_size would be non-zero */
1724 switch(st
->minor_version
) {
1726 /* either 3K after the superblock (when hot-add),
1727 * or some amount of space before.
1730 /* We are creating array, so we *know* how much room has
1734 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1737 room
= __le64_to_cpu(sb
->super_offset
)
1738 - __le64_to_cpu(sb
->data_offset
)
1739 - __le64_to_cpu(sb
->data_size
);
1740 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
1743 bbl_offset
= (__s32
)__le32_to_cpu(sb
->bblog_offset
);
1744 if (bbl_size
< -bbl_offset
)
1745 bbl_size
= -bbl_offset
;
1747 if (!may_change
|| (room
< 3*2 &&
1748 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1753 offset
= 0; /* means movable offset */
1758 case 2: /* between superblock and data */
1761 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1764 room
= __le64_to_cpu(sb
->data_offset
)
1765 - __le64_to_cpu(sb
->super_offset
);
1766 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
1768 room
= __le32_to_cpu(sb
->bblog_offset
) + bbl_size
;
1773 room
-= 2; /* Leave 1K for superblock */
1777 room
-= 4*2; /* leave 4K for superblock */
1787 if (chunk
== UnSet
&& room
> 128*2)
1788 /* Limit to 128K of bitmap when chunk size not requested */
1791 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1793 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1794 bits
= (size
*512)/min_chunk
+1;
1795 while (bits
> max_bits
) {
1799 if (chunk
== UnSet
) {
1800 /* For practical purpose, 64Meg is a good
1801 * default chunk size for internal bitmaps.
1804 if (chunk
< 64*1024*1024)
1805 chunk
= 64*1024*1024;
1806 } else if (chunk
< min_chunk
)
1807 return 0; /* chunk size too small */
1808 if (chunk
== 0) /* rounding problem */
1812 /* start bitmap on a 4K boundary with enough space for
1815 bits
= (size
*512) / chunk
+ 1;
1816 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
1817 room
*= 8; /* convert 4K blocks to sectors */
1818 offset
= -room
- bbl_size
;
1821 sb
->bitmap_offset
= (int32_t)__cpu_to_le32(offset
);
1823 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
)
1824 | MD_FEATURE_BITMAP_OFFSET
);
1825 memset(bms
, 0, sizeof(*bms
));
1826 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1827 bms
->version
= __cpu_to_le32(major
);
1828 uuid_from_super1(st
, uuid
);
1829 memcpy(bms
->uuid
, uuid
, 16);
1830 bms
->chunksize
= __cpu_to_le32(chunk
);
1831 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1832 bms
->sync_size
= __cpu_to_le64(size
);
1833 bms
->write_behind
= __cpu_to_le32(write_behind
);
1839 static void locate_bitmap1(struct supertype
*st
, int fd
)
1841 unsigned long long offset
;
1842 struct mdp_superblock_1
*sb
;
1846 if (st
->ss
->load_super(st
, fd
, NULL
))
1847 return; /* no error I hope... */
1852 offset
= __le64_to_cpu(sb
->super_offset
);
1853 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
1856 lseek64(fd
, offset
<<9, 0);
1859 static int write_bitmap1(struct supertype
*st
, int fd
)
1861 struct mdp_superblock_1
*sb
= st
->sb
;
1862 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
1866 struct align_fd afd
;
1870 locate_bitmap1(st
, fd
);
1872 if (posix_memalign(&buf
, 4096, 4096))
1875 memset(buf
, 0xff, 4096);
1876 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
1878 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1879 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1880 towrite
+= sizeof(bitmap_super_t
);
1881 towrite
= ROUND_UP(towrite
, 512);
1882 while (towrite
> 0) {
1886 n
= awrite(&afd
, buf
, n
);
1891 memset(buf
, 0xff, 4096);
1901 static void free_super1(struct supertype
*st
)
1906 struct devinfo
*di
= st
->info
;
1907 st
->info
= di
->next
;
1916 static int validate_geometry1(struct supertype
*st
, int level
,
1917 int layout
, int raiddisks
,
1918 int *chunk
, unsigned long long size
,
1919 unsigned long long data_offset
,
1920 char *subdev
, unsigned long long *freesize
,
1923 unsigned long long ldsize
;
1926 if (level
== LEVEL_CONTAINER
) {
1928 pr_err("1.x metadata does not support containers\n");
1931 if (chunk
&& *chunk
== UnSet
)
1932 *chunk
= DEFAULT_CHUNK
;
1937 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1940 pr_err("super1.x cannot open %s: %s\n",
1941 subdev
, strerror(errno
));
1945 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1951 *freesize
= avail_size1(st
, ldsize
>> 9, data_offset
);
1954 #endif /* MDASSEMBLE */
1956 struct superswitch super1
= {
1958 .examine_super
= examine_super1
,
1959 .brief_examine_super
= brief_examine_super1
,
1960 .export_examine_super
= export_examine_super1
,
1961 .detail_super
= detail_super1
,
1962 .brief_detail_super
= brief_detail_super1
,
1963 .export_detail_super
= export_detail_super1
,
1964 .write_init_super
= write_init_super1
,
1965 .validate_geometry
= validate_geometry1
,
1966 .add_to_super
= add_to_super1
,
1968 .match_home
= match_home1
,
1969 .uuid_from_super
= uuid_from_super1
,
1970 .getinfo_super
= getinfo_super1
,
1971 .container_content
= container_content1
,
1972 .update_super
= update_super1
,
1973 .init_super
= init_super1
,
1974 .store_super
= store_super1
,
1975 .compare_super
= compare_super1
,
1976 .load_super
= load_super1
,
1977 .match_metadata_desc
= match_metadata_desc1
,
1978 .avail_size
= avail_size1
,
1979 .add_internal_bitmap
= add_internal_bitmap1
,
1980 .locate_bitmap
= locate_bitmap1
,
1981 .write_bitmap
= write_bitmap1
,
1982 .free_super
= free_super1
,
1983 #if __BYTE_ORDER == BIG_ENDIAN