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>
28 * The version-1 superblock :
29 * All numeric fields are little-endian.
31 * total size: 256 bytes plus 2 per device.
32 * 1K allows 384 devices.
34 struct mdp_superblock_1
{
35 /* constant array information - 128 bytes */
36 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
37 __u32 major_version
; /* 1 */
38 __u32 feature_map
; /* 0 for now */
39 __u32 pad0
; /* always set to 0 when writing */
41 __u8 set_uuid
[16]; /* user-space generated. */
42 char set_name
[32]; /* set and interpreted by user-space */
44 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
45 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
46 __u32 layout
; /* only for raid5 currently */
47 __u64 size
; /* used size of component devices, in 512byte sectors */
49 __u32 chunksize
; /* in 512byte sectors */
51 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
52 * NOTE: signed, so bitmap can be before superblock
53 * only meaningful of feature_map[0] is set.
56 /* These are only valid with feature bit '4' */
57 __u32 new_level
; /* new level we are reshaping to */
58 __u64 reshape_position
; /* next address in array-space for reshape */
59 __u32 delta_disks
; /* change in number of raid_disks */
60 __u32 new_layout
; /* new layout */
61 __u32 new_chunk
; /* new chunk size (sectors) */
62 __u32 new_offset
; /* signed number to add to data_offset in new
63 * layout. 0 == no-change. This can be
64 * different on each device in the array.
67 /* constant this-device information - 64 bytes */
68 __u64 data_offset
; /* sector start of data, often 0 */
69 __u64 data_size
; /* sectors in this device that can be used for data */
70 __u64 super_offset
; /* sector start of this superblock */
71 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
72 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
73 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
74 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
75 __u8 devflags
; /* per-device flags. Only one defined...*/
76 #define WriteMostly1 1 /* mask for writemostly flag in above */
77 /* bad block log. If there are any bad blocks the feature flag is set.
78 * if offset and size are non-zero, that space is reserved and available.
80 __u8 bblog_shift
; /* shift from sectors to block size for badblocklist */
81 __u16 bblog_size
; /* number of sectors reserved for badblocklist */
82 __u32 bblog_offset
; /* sector offset from superblock to bblog, signed */
84 /* array state information - 64 bytes */
85 __u64 utime
; /* 40 bits second, 24 btes microseconds */
86 __u64 events
; /* incremented when superblock updated */
87 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
88 __u32 sb_csum
; /* checksum upto dev_roles[max_dev] */
89 __u32 max_dev
; /* size of dev_roles[] array to consider */
90 __u8 pad3
[64-32]; /* set to 0 when writing */
92 /* device state information. Indexed by dev_number.
94 * Note there are no per-device state flags. State information is rolled
95 * into the 'roles' value. If a device is spare or faulty, then it doesn't
96 * have a meaningful role.
98 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
101 #define MAX_SB_SIZE 4096
102 /* bitmap super size is 256, but we round up to a sector for alignment */
103 #define BM_SUPER_SIZE 512
104 #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2)
105 #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \
106 + sizeof(struct misc_dev_info))
108 struct misc_dev_info
{
112 /* feature_map bits */
113 #define MD_FEATURE_BITMAP_OFFSET 1
114 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
117 #define MD_FEATURE_RESHAPE_ACTIVE 4
118 #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */
119 #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an
120 * active device with same 'role'.
121 * 'recovery_offset' is also set.
123 #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number
124 * of devices, but is going
127 #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */
128 #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \
129 |MD_FEATURE_RECOVERY_OFFSET \
130 |MD_FEATURE_RESHAPE_ACTIVE \
131 |MD_FEATURE_BAD_BLOCKS \
132 |MD_FEATURE_REPLACEMENT \
133 |MD_FEATURE_RESHAPE_BACKWARDS \
134 |MD_FEATURE_NEW_OFFSET \
137 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
139 unsigned int disk_csum
, csum
;
140 unsigned long long newcsum
;
141 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
142 unsigned int *isuper
= (unsigned int*)sb
;
144 /* make sure I can count... */
145 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
146 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
147 sizeof(struct mdp_superblock_1
) != 256) {
148 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
151 disk_csum
= sb
->sb_csum
;
154 for (; size
>=4; size
-= 4 ) {
155 newcsum
+= __le32_to_cpu(*isuper
);
160 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
162 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
163 sb
->sb_csum
= disk_csum
;
164 return __cpu_to_le32(csum
);
168 * Information related to file descriptor used for aligned reads/writes.
169 * Cache the block size.
176 static void init_afd(struct align_fd
*afd
, int fd
)
180 if (ioctl(afd
->fd
, BLKSSZGET
, &afd
->blk_sz
) != 0)
184 static char abuf
[4096+4096];
185 static int aread(struct align_fd
*afd
, void *buf
, int len
)
188 * On devices with a 4K sector size, we need to read
189 * the full sector and copy relevant bits into
198 if (!bsize
|| bsize
> 4096 || len
> 4096) {
200 fprintf(stderr
, "WARNING - aread() called with "
201 "invalid block size\n");
204 b
= ROUND_UP_PTR((char *)abuf
, 4096);
206 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
208 n
= read(afd
->fd
, b
, iosize
);
211 lseek(afd
->fd
, len
- n
, 1);
218 static int awrite(struct align_fd
*afd
, void *buf
, int len
)
221 * On devices with a 4K sector size, we need to write
222 * the full sector. We pre-read if the sector is larger
224 * The address must be sector-aligned.
231 if (!bsize
|| bsize
> 4096 || len
> 4096) {
233 fprintf(stderr
, "WARNING - awrite() called with "
234 "invalid block size\n");
237 b
= ROUND_UP_PTR((char *)abuf
, 4096);
239 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
243 n
= read(afd
->fd
, b
, iosize
);
246 lseek(afd
->fd
, -n
, 1);
250 n
= write(afd
->fd
, b
, iosize
);
253 lseek(afd
->fd
, len
- n
, 1);
258 static void examine_super1(struct supertype
*st
, char *homehost
)
260 struct mdp_superblock_1
*sb
= st
->sb
;
267 int l
= homehost
? strlen(homehost
) : 0;
269 unsigned long long sb_offset
;
272 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
273 printf(" Version : 1");
274 sb_offset
= __le64_to_cpu(sb
->super_offset
);
277 else if (sb_offset
<= 8)
281 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
282 printf(" Array UUID : ");
283 for (i
=0; i
<16; i
++) {
284 if ((i
&3)==0 && i
!= 0) printf(":");
285 printf("%02x", sb
->set_uuid
[i
]);
288 printf(" Name : %.32s", sb
->set_name
);
289 if (l
> 0 && l
< 32 &&
290 sb
->set_name
[l
] == ':' &&
291 strncmp(sb
->set_name
, homehost
, l
) == 0)
292 printf(" (local to host %s)", homehost
);
294 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
295 printf(" Creation Time : %.24s\n", ctime(&atime
));
296 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
297 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
298 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
300 printf(" Avail Dev Size : %llu%s\n",
301 (unsigned long long)__le64_to_cpu(sb
->data_size
),
302 human_size(__le64_to_cpu(sb
->data_size
)<<9));
303 if (__le32_to_cpu(sb
->level
) > 0) {
304 int ddsks
= 0, ddsks_denom
= 1;
305 switch(__le32_to_cpu(sb
->level
)) {
306 case 1: ddsks
=1;break;
308 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
309 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
311 layout
= __le32_to_cpu(sb
->layout
);
312 ddsks
= __le32_to_cpu(sb
->raid_disks
);
313 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
316 long long asize
= __le64_to_cpu(sb
->size
);
317 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
318 printf(" Array Size : %llu%s\n",
319 asize
>> 10, human_size(asize
));
321 if (sb
->size
!= sb
->data_size
)
322 printf(" Used Dev Size : %llu%s\n",
323 (unsigned long long)__le64_to_cpu(sb
->size
),
324 human_size(__le64_to_cpu(sb
->size
)<<9));
327 printf(" Data Offset : %llu sectors\n",
328 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
329 if (sb
->new_offset
&&
330 (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_NEW_OFFSET
)) {
331 unsigned long long offset
= __le64_to_cpu(sb
->data_offset
);
332 offset
+= (signed)(int32_t)__le32_to_cpu(sb
->new_offset
);
333 printf(" New Offset : %llu sectors\n", offset
);
335 printf(" Super Offset : %llu sectors\n",
336 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
337 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
338 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
340 st
->ss
->getinfo_super(st
, &info
, NULL
);
341 if (info
.space_after
!= 1 &&
342 !(__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_NEW_OFFSET
))
343 printf(" Unused Space : before=%llu sectors, after=%llu sectors\n",
344 info
.space_before
, info
.space_after
);
346 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
347 printf(" Device UUID : ");
348 for (i
=0; i
<16; i
++) {
349 if ((i
&3)==0 && i
!= 0) printf(":");
350 printf("%02x", sb
->device_uuid
[i
]);
354 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
355 printf("Internal Bitmap : %ld sectors from superblock\n",
356 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
358 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)) {
359 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
360 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
361 if (__le32_to_cpu(sb
->delta_disks
)) {
362 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
363 printf(" (%d->%d)\n",
364 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
365 __le32_to_cpu(sb
->raid_disks
));
366 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
367 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
369 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
370 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
371 printf(" New Level : %s\n", c
?c
:"-unknown-");
373 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
374 if (__le32_to_cpu(sb
->level
) == 5) {
375 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
376 printf(" New Layout : %s\n", c
?c
:"-unknown-");
378 if (__le32_to_cpu(sb
->level
) == 6) {
379 c
= map_num(r6layout
, __le32_to_cpu(sb
->new_layout
));
380 printf(" New Layout : %s\n", c
?c
:"-unknown-");
382 if (__le32_to_cpu(sb
->level
) == 10) {
383 printf(" New Layout :");
384 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
388 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
389 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
394 if (sb
->devflags
& WriteMostly1
)
395 printf(" write-mostly");
399 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
400 printf(" Update Time : %.24s\n", ctime(&atime
));
402 if (sb
->bblog_size
&& sb
->bblog_offset
) {
403 printf(" Bad Block Log : %d entries available at offset %ld sectors",
404 __le16_to_cpu(sb
->bblog_size
)*512/8,
405 (long)(int32_t)__le32_to_cpu(sb
->bblog_offset
));
406 if (sb
->feature_map
&
407 __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
408 printf(" - bad blocks present.");
412 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
413 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
415 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
416 __le32_to_cpu(calc_sb_1_csum(sb
)));
417 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
419 if (__le32_to_cpu(sb
->level
) == 5) {
420 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
421 printf(" Layout : %s\n", c
?c
:"-unknown-");
423 if (__le32_to_cpu(sb
->level
) == 6) {
424 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
425 printf(" Layout : %s\n", c
?c
:"-unknown-");
427 if (__le32_to_cpu(sb
->level
) == 10) {
428 int lo
= __le32_to_cpu(sb
->layout
);
430 print_r10_layout(lo
);
433 switch(__le32_to_cpu(sb
->level
)) {
439 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
442 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
448 /* This turns out to just be confusing */
449 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
450 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
451 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
453 for (d
=0; d
< i
; d
++) {
454 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
456 if (role
== 0xffff) printf("empty");
457 else if(role
== 0xfffe) printf("failed");
458 else printf("%d", role
);
462 printf(" Device Role : ");
463 d
= __le32_to_cpu(sb
->dev_number
);
464 if (d
< __le32_to_cpu(sb
->max_dev
))
465 role
= __le16_to_cpu(sb
->dev_roles
[d
]);
470 else if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_REPLACEMENT
))
471 printf("Replacement device %d\n", role
);
473 printf("Active device %d\n", role
);
475 printf(" Array State : ");
476 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
479 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
480 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
494 /* This is confusing too */
496 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
497 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
501 if (faulty
) printf(" %d failed", faulty
);
503 printf(" ('A' == active, '.' == missing, 'R' == replacing)");
507 static void brief_examine_super1(struct supertype
*st
, int verbose
)
509 struct mdp_superblock_1
*sb
= st
->sb
;
511 unsigned long long sb_offset
;
513 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
515 nm
= strchr(sb
->set_name
, ':');
518 else if (sb
->set_name
[0])
530 printf(" level=%s", c
);
531 sb_offset
= __le64_to_cpu(sb
->super_offset
);
533 printf(" metadata=1.1 ");
534 else if (sb_offset
<= 8)
535 printf(" metadata=1.2 ");
537 printf(" metadata=1.0 ");
539 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
541 for (i
=0; i
<16; i
++) {
542 if ((i
&3)==0 && i
!= 0) printf(":");
543 printf("%02x", sb
->set_uuid
[i
]);
545 if (sb
->set_name
[0]) {
547 print_quoted(sb
->set_name
);
552 static void export_examine_super1(struct supertype
*st
)
554 struct mdp_superblock_1
*sb
= st
->sb
;
559 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
560 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
562 if (sb
->set_name
[i
] == '\n' ||
563 sb
->set_name
[i
] == '\0') {
568 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
569 if (__le32_to_cpu(sb
->level
) > 0) {
570 int ddsks
= 0, ddsks_denom
= 1;
571 switch(__le32_to_cpu(sb
->level
)) {
572 case 1: ddsks
=1;break;
574 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
575 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
577 layout
= __le32_to_cpu(sb
->layout
);
578 ddsks
= __le32_to_cpu(sb
->raid_disks
);
579 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
582 long long asize
= __le64_to_cpu(sb
->size
);
583 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
584 printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize
,JEDEC
));
588 for (i
=0; i
<16; i
++) {
589 if ((i
&3)==0 && i
!= 0) printf(":");
590 printf("%02x", sb
->set_uuid
[i
]);
593 printf("MD_UPDATE_TIME=%llu\n",
594 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
595 printf("MD_DEV_UUID=");
596 for (i
=0; i
<16; i
++) {
597 if ((i
&3)==0 && i
!= 0) printf(":");
598 printf("%02x", sb
->device_uuid
[i
]);
601 printf("MD_EVENTS=%llu\n",
602 (unsigned long long)__le64_to_cpu(sb
->events
));
605 static int copy_metadata1(struct supertype
*st
, int from
, int to
)
607 /* Read superblock. If it looks good, write it out.
608 * Then if a bitmap is present, copy that.
609 * And if a bad-block-list is present, copy that too.
612 unsigned long long dsize
, sb_offset
;
613 const int bufsize
= 4*1024;
614 struct mdp_superblock_1 super
, *sb
;
616 if (posix_memalign(&buf
, 4096, bufsize
) != 0)
619 if (!get_dev_size(from
, NULL
, &dsize
))
625 switch(st
->minor_version
) {
629 sb_offset
&= ~(4*2-1);
641 if (lseek64(from
, sb_offset
<< 9, 0) < 0LL)
643 if (read(from
, buf
, bufsize
) != bufsize
)
647 super
= *sb
; // save most of sb for when we reuse buf
649 if (__le32_to_cpu(super
.magic
) != MD_SB_MAGIC
||
650 __le32_to_cpu(super
.major_version
) != 1 ||
651 __le64_to_cpu(super
.super_offset
) != sb_offset
||
652 calc_sb_1_csum(sb
) != super
.sb_csum
)
655 if (lseek64(to
, sb_offset
<< 9, 0) < 0LL)
657 if (write(to
, buf
, bufsize
) != bufsize
)
660 if (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
661 unsigned long long bitmap_offset
= sb_offset
;
662 int bytes
= 4096; // just an estimate.
664 struct align_fd afrom
, ato
;
666 init_afd(&afrom
, from
);
669 bitmap_offset
+= (int32_t)__le32_to_cpu(super
.bitmap_offset
);
671 if (lseek64(from
, bitmap_offset
<<9, 0) < 0)
673 if (lseek64(to
, bitmap_offset
<<9, 0) < 0)
676 for (written
= 0; written
< bytes
; ) {
677 int n
= bytes
- written
;
680 if (aread(&afrom
, buf
, n
) != n
)
683 /* have the header, can calculate
684 * correct bitmap bytes */
688 bits
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
689 bytes
= (bits
+7) >> 3;
690 bytes
+= sizeof(bitmap_super_t
);
691 bytes
= ROUND_UP(bytes
, 512);
695 if (awrite(&ato
, buf
, n
) != n
)
701 if (super
.bblog_size
!= 0 &&
702 __le32_to_cpu(super
.bblog_size
) <= 100 &&
703 super
.bblog_offset
!= 0 &&
704 (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BAD_BLOCKS
))) {
705 /* There is a bad block log */
706 unsigned long long bb_offset
= sb_offset
;
707 int bytes
= __le32_to_cpu(super
.bblog_size
) * 512;
709 struct align_fd afrom
, ato
;
711 init_afd(&afrom
, from
);
714 bb_offset
+= (int32_t)__le32_to_cpu(super
.bblog_offset
);
716 if (lseek64(from
, bb_offset
<<9, 0) < 0)
718 if (lseek64(to
, bb_offset
<<9, 0) < 0)
721 for (written
= 0; written
< bytes
; ) {
722 int n
= bytes
- written
;
725 if (aread(&afrom
, buf
, n
) != n
)
728 if (awrite(&ato
, buf
, n
) != n
)
742 static void detail_super1(struct supertype
*st
, char *homehost
)
744 struct mdp_superblock_1
*sb
= st
->sb
;
746 int l
= homehost
? strlen(homehost
) : 0;
748 printf(" Name : %.32s", sb
->set_name
);
749 if (l
> 0 && l
< 32 &&
750 sb
->set_name
[l
] == ':' &&
751 strncmp(sb
->set_name
, homehost
, l
) == 0)
752 printf(" (local to host %s)", homehost
);
753 printf("\n UUID : ");
754 for (i
=0; i
<16; i
++) {
755 if ((i
&3)==0 && i
!= 0) printf(":");
756 printf("%02x", sb
->set_uuid
[i
]);
758 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
761 static void brief_detail_super1(struct supertype
*st
)
763 struct mdp_superblock_1
*sb
= st
->sb
;
766 if (sb
->set_name
[0]) {
768 print_quoted(sb
->set_name
);
771 for (i
=0; i
<16; i
++) {
772 if ((i
&3)==0 && i
!= 0) printf(":");
773 printf("%02x", sb
->set_uuid
[i
]);
777 static void export_detail_super1(struct supertype
*st
)
779 struct mdp_superblock_1
*sb
= st
->sb
;
784 if (sb
->set_name
[i
] == '\n' ||
785 sb
->set_name
[i
] == '\0') {
790 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
793 static int examine_badblocks_super1(struct supertype
*st
, int fd
, char *devname
)
795 struct mdp_superblock_1
*sb
= st
->sb
;
796 unsigned long long offset
;
801 if (!sb
->bblog_size
|| __le32_to_cpu(sb
->bblog_size
) > 100
802 || !sb
->bblog_offset
){
803 printf("No bad-blocks list configured on %s\n", devname
);
806 if ((sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
808 printf("Bad-blocks list is empty in %s\n", devname
);
812 size
= __le32_to_cpu(sb
->bblog_size
)* 512;
813 if (posix_memalign((void**)&bbl
, 4096, size
) != 0) {
814 pr_err("%s could not allocate badblocks list\n", __func__
);
817 offset
= __le64_to_cpu(sb
->super_offset
) +
818 (int)__le32_to_cpu(sb
->bblog_offset
);
820 if (lseek64(fd
, offset
, 0) < 0) {
821 pr_err("Cannot seek to bad-blocks list\n");
824 if (read(fd
, bbl
, size
) != size
) {
825 pr_err("Cannot read bad-blocks list\n");
828 /* 64bits per entry. 10 bits is block-count, 54 bits is block
829 * offset. Blocks are sectors unless bblog->shift makes them bigger
832 printf("Bad-blocks on %s:\n", devname
);
833 for (i
= 0; i
< size
/8; i
++, bbp
++) {
834 __u64 bb
= __le64_to_cpu(*bbp
);
835 int count
= bb
& 0x3ff;
836 unsigned long long sector
= bb
>> 10;
841 sector
<<= sb
->bblog_shift
;
842 count
<<= sb
->bblog_shift
;
844 printf("%20llu for %d sectors\n", sector
, count
);
851 static int match_home1(struct supertype
*st
, char *homehost
)
853 struct mdp_superblock_1
*sb
= st
->sb
;
854 int l
= homehost
? strlen(homehost
) : 0;
856 return (l
> 0 && l
< 32 &&
857 sb
->set_name
[l
] == ':' &&
858 strncmp(sb
->set_name
, homehost
, l
) == 0);
861 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
863 struct mdp_superblock_1
*super
= st
->sb
;
864 char *cuuid
= (char*)uuid
;
867 cuuid
[i
] = super
->set_uuid
[i
];
870 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
872 struct mdp_superblock_1
*sb
= st
->sb
;
873 struct bitmap_super_s
*bsb
= (void*)(((char*)sb
)+MAX_SB_SIZE
);
874 struct misc_dev_info
*misc
= (void*)(((char*)sb
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
878 unsigned int map_disks
= info
->array
.raid_disks
;
879 unsigned long long super_offset
;
880 unsigned long long data_size
;
882 memset(info
, 0, sizeof(*info
));
883 info
->array
.major_version
= 1;
884 info
->array
.minor_version
= st
->minor_version
;
885 info
->array
.patch_version
= 0;
886 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
887 info
->array
.level
= __le32_to_cpu(sb
->level
);
888 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
889 info
->array
.md_minor
= -1;
890 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
891 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
892 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
894 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
)
897 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
898 info
->component_size
= __le64_to_cpu(sb
->size
);
899 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
))
900 info
->bitmap_offset
= (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
902 info
->disk
.major
= 0;
903 info
->disk
.minor
= 0;
904 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
905 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
906 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
909 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
911 super_offset
= __le64_to_cpu(sb
->super_offset
);
912 if (info
->array
.level
<= 0)
913 data_size
= __le64_to_cpu(sb
->data_size
);
915 data_size
= __le64_to_cpu(sb
->size
);
916 if (info
->data_offset
< super_offset
) {
917 unsigned long long end
;
918 info
->space_before
= info
->data_offset
;
921 if (sb
->bblog_offset
&& sb
->bblog_size
) {
922 unsigned long long bboffset
= super_offset
;
923 bboffset
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
928 if (super_offset
+ info
->bitmap_offset
< end
)
929 end
= super_offset
+ info
->bitmap_offset
;
931 if (info
->data_offset
+ data_size
< end
)
932 info
->space_after
= end
- data_size
- info
->data_offset
;
934 info
->space_after
= 0;
936 unsigned long long earliest
;
937 earliest
= super_offset
+ (32+4)*2; /* match kernel */
938 if (info
->bitmap_offset
> 0) {
939 unsigned long long bmend
= info
->bitmap_offset
;
940 unsigned long long size
= __le64_to_cpu(bsb
->sync_size
);
941 size
/= __le32_to_cpu(bsb
->chunksize
) >> 9;
942 size
= (size
+ 7) >> 3;
943 size
+= sizeof(bitmap_super_t
);
944 size
= ROUND_UP(size
, 4096);
947 if (bmend
> earliest
)
950 if (sb
->bblog_offset
&& sb
->bblog_size
) {
951 unsigned long long bbend
= super_offset
;
952 bbend
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
953 bbend
+= __le32_to_cpu(sb
->bblog_size
);
954 if (bbend
> earliest
)
957 if (earliest
< info
->data_offset
)
958 info
->space_before
= info
->data_offset
- earliest
;
960 info
->space_before
= 0;
961 info
->space_after
= misc
->device_size
- data_size
- info
->data_offset
;
963 if (info
->space_before
== 0 && info
->space_after
== 0) {
964 /* It will look like we don't support data_offset changes,
965 * be we do - it's just that there is no room.
966 * A change that reduced the number of devices should
967 * still be allowed, so set the otherwise useless value of '1'
969 info
->space_after
= 1;
972 info
->disk
.raid_disk
= -1;
975 info
->disk
.state
= 0; /* spare: not active, not sync, not faulty */
978 info
->disk
.state
= 1; /* faulty */
981 info
->disk
.state
= 6; /* active and in sync */
982 info
->disk
.raid_disk
= role
;
984 if (sb
->devflags
& WriteMostly1
)
985 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
986 info
->events
= __le64_to_cpu(sb
->events
);
987 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
988 info
->safe_mode_delay
= 200;
990 memcpy(info
->uuid
, sb
->set_uuid
, 16);
992 strncpy(info
->name
, sb
->set_name
, 32);
995 if ((__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_REPLACEMENT
)) {
996 info
->disk
.state
&= ~(1 << MD_DISK_SYNC
);
997 info
->disk
.state
|= 1 << MD_DISK_REPLACEMENT
;
1000 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
1001 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
1003 info
->recovery_start
= MaxSector
;
1005 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
1006 info
->reshape_active
= 1;
1007 if ((sb
->feature_map
& __le32_to_cpu(MD_FEATURE_NEW_OFFSET
)) &&
1008 sb
->new_offset
!= 0)
1009 info
->reshape_active
|= RESHAPE_NO_BACKUP
;
1010 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
1011 info
->new_level
= __le32_to_cpu(sb
->new_level
);
1012 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
1013 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
1014 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
1015 if (info
->delta_disks
< 0)
1016 info
->array
.raid_disks
-= info
->delta_disks
;
1018 info
->reshape_active
= 0;
1020 info
->recovery_blocked
= info
->reshape_active
;
1023 for (i
=0; i
<map_disks
; i
++)
1025 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
1026 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
1027 if (/*role == 0xFFFF || */role
< (unsigned) info
->array
.raid_disks
) {
1029 if (map
&& role
< map_disks
)
1034 info
->array
.working_disks
= working
;
1037 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
1039 struct mdinfo
*info
;
1044 info
= xmalloc(sizeof(*info
));
1045 getinfo_super1(st
, info
, NULL
);
1049 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
1051 char *devname
, int verbose
,
1052 int uuid_set
, char *homehost
)
1054 /* NOTE: for 'assemble' and 'force' we need to return non-zero
1055 * if any change was made. For others, the return value is
1059 struct mdp_superblock_1
*sb
= st
->sb
;
1061 if (strcmp(update
, "homehost") == 0 &&
1063 /* Note that 'homehost' is special as it is really
1068 c
= strchr(sb
->set_name
, ':');
1070 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
1072 strncpy(info
->name
, sb
->set_name
, 32);
1076 if (strcmp(update
, "force-one")==0) {
1077 /* Not enough devices for a working array,
1078 * so bring this one up-to-date
1080 if (sb
->events
!= __cpu_to_le64(info
->events
))
1082 sb
->events
= __cpu_to_le64(info
->events
);
1083 } else if (strcmp(update
, "force-array")==0) {
1084 /* Degraded array and 'force' requests to
1085 * maybe need to mark it 'clean'.
1087 switch(__le32_to_cpu(sb
->level
)) {
1088 case 5: case 4: case 6:
1089 /* need to force clean */
1090 if (sb
->resync_offset
!= MaxSector
)
1092 sb
->resync_offset
= MaxSector
;
1094 } else if (strcmp(update
, "assemble")==0) {
1095 int d
= info
->disk
.number
;
1097 if (info
->disk
.state
& (1<<MD_DISK_ACTIVE
))
1098 want
= info
->disk
.raid_disk
;
1101 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
1102 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
1105 if (info
->reshape_active
&&
1106 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1107 info
->delta_disks
>= 0 &&
1108 info
->reshape_progress
< __le64_to_cpu(sb
->reshape_position
)) {
1109 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
1112 if (info
->reshape_active
&&
1113 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1114 info
->delta_disks
< 0 &&
1115 info
->reshape_progress
> __le64_to_cpu(sb
->reshape_position
)) {
1116 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
1119 } else if (strcmp(update
, "linear-grow-new") == 0) {
1122 unsigned int max
= __le32_to_cpu(sb
->max_dev
);
1124 for (i
=0 ; i
< max
; i
++)
1125 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
1127 sb
->dev_number
= __cpu_to_le32(i
);
1128 info
->disk
.number
= i
;
1129 if (max
>= __le32_to_cpu(sb
->max_dev
))
1130 sb
->max_dev
= __cpu_to_le32(max
+1);
1132 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1133 read(rfd
, sb
->device_uuid
, 16) != 16) {
1134 __u32 r
[4] = {random(), random(), random(), random()};
1135 memcpy(sb
->device_uuid
, r
, 16);
1141 __cpu_to_le16(info
->disk
.raid_disk
);
1143 fd
= open(devname
, O_RDONLY
);
1145 unsigned long long ds
;
1146 get_dev_size(fd
, devname
, &ds
);
1149 if (__le64_to_cpu(sb
->super_offset
) <
1150 __le64_to_cpu(sb
->data_offset
)) {
1151 sb
->data_size
= __cpu_to_le64(
1152 ds
- __le64_to_cpu(sb
->data_offset
));
1155 ds
&= ~(unsigned long long)(4*2-1);
1156 sb
->super_offset
= __cpu_to_le64(ds
);
1157 sb
->data_size
= __cpu_to_le64(
1158 ds
- __le64_to_cpu(sb
->data_offset
));
1161 } else if (strcmp(update
, "linear-grow-update") == 0) {
1162 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
1163 sb
->dev_roles
[info
->disk
.number
] =
1164 __cpu_to_le16(info
->disk
.raid_disk
);
1165 } else if (strcmp(update
, "resync") == 0) {
1166 /* make sure resync happens */
1167 sb
->resync_offset
= 0ULL;
1168 } else if (strcmp(update
, "uuid") == 0) {
1169 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
1171 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
1172 struct bitmap_super_s
*bm
;
1173 bm
= (struct bitmap_super_s
*)(st
->sb
+MAX_SB_SIZE
);
1174 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
1176 } else if (strcmp(update
, "no-bitmap") == 0) {
1177 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
1178 } else if (strcmp(update
, "bbl") == 0) {
1179 /* only possible if there is room after the bitmap, or if
1180 * there is no bitmap
1182 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
1183 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
1184 long bitmap_offset
= (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
);
1185 long bm_sectors
= 0;
1189 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1190 struct bitmap_super_s
*bsb
;
1191 bsb
= (struct bitmap_super_s
*)(((char*)sb
)+MAX_SB_SIZE
);
1192 bm_sectors
= bitmap_sectors(bsb
);
1195 if (sb_offset
< data_offset
) {
1196 /* 1.1 or 1.2. Put bbl after bitmap leaving at least 32K
1199 bb_offset
= sb_offset
+ 8;
1200 if (bm_sectors
&& bitmap_offset
> 0)
1201 bb_offset
= bitmap_offset
+ bm_sectors
;
1202 while (bb_offset
< (long)sb_offset
+ 8 + 32*2
1203 && bb_offset
+ 8+8 <= (long)data_offset
)
1204 /* too close to bitmap, and room to grow */
1206 if (bb_offset
+ 8 <= (long)data_offset
) {
1207 sb
->bblog_size
= __cpu_to_le16(8);
1208 sb
->bblog_offset
= __cpu_to_le32(bb_offset
);
1211 /* 1.0 - Put bbl just before super block */
1212 if (bm_sectors
&& bitmap_offset
< 0)
1213 space
= -bitmap_offset
- bm_sectors
;
1215 space
= sb_offset
- data_offset
-
1216 __le64_to_cpu(sb
->data_size
);
1218 sb
->bblog_size
= __cpu_to_le16(8);
1219 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
1222 } else if (strcmp(update
, "no-bbl") == 0) {
1223 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
1224 pr_err("Cannot remove active bbl from %s\n",devname
);
1227 sb
->bblog_shift
= 0;
1228 sb
->bblog_offset
= 0;
1230 } else if (strcmp(update
, "name") == 0) {
1231 if (info
->name
[0] == 0)
1232 sprintf(info
->name
, "%d", info
->array
.md_minor
);
1233 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
1235 strchr(info
->name
, ':') == NULL
&&
1236 strlen(homehost
)+1+strlen(info
->name
) < 32) {
1237 strcpy(sb
->set_name
, homehost
);
1238 strcat(sb
->set_name
, ":");
1239 strcat(sb
->set_name
, info
->name
);
1241 strcpy(sb
->set_name
, info
->name
);
1242 } else if (strcmp(update
, "devicesize") == 0 &&
1243 __le64_to_cpu(sb
->super_offset
) <
1244 __le64_to_cpu(sb
->data_offset
)) {
1245 /* set data_size to device size less data_offset */
1246 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
1247 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
1248 printf("Size was %llu\n", (unsigned long long)
1249 __le64_to_cpu(sb
->data_size
));
1250 sb
->data_size
= __cpu_to_le64(
1251 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
1252 printf("Size is %llu\n", (unsigned long long)
1253 __le64_to_cpu(sb
->data_size
));
1254 } else if (strcmp(update
, "revert-reshape") == 0) {
1256 if (!(sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)))
1257 pr_err("No active reshape to revert on %s\n",
1261 unsigned long long reshape_sectors
;
1264 /* reshape_position is a little messy.
1265 * Its value must be a multiple of the larger
1266 * chunk size, and of the "after" data disks.
1267 * So when reverting we need to change it to
1268 * be a multiple of the new "after" data disks,
1269 * which is the old "before".
1270 * If it isn't already a multiple of 'before',
1271 * the only thing we could do would be
1272 * copy some block around on the disks, which
1273 * is easy to get wrong.
1274 * So we reject a revert-reshape unless the
1275 * alignment is good.
1277 if (__le32_to_cpu(sb
->level
) >= 4 &&
1278 __le32_to_cpu(sb
->level
) <= 6) {
1279 reshape_sectors
= __le64_to_cpu(sb
->reshape_position
);
1280 reshape_chunk
= __le32_to_cpu(sb
->new_chunk
);
1281 reshape_chunk
*= __le32_to_cpu(sb
->raid_disks
) - __le32_to_cpu(sb
->delta_disks
) -
1282 (__le32_to_cpu(sb
->level
)==6 ? 2 : 1);
1283 if (reshape_sectors
% reshape_chunk
) {
1284 pr_err("Reshape position is not suitably aligned.\n");
1285 pr_err("Try normal assembly and stop again\n");
1289 sb
->raid_disks
= __cpu_to_le32(__le32_to_cpu(sb
->raid_disks
) -
1290 __le32_to_cpu(sb
->delta_disks
));
1291 if (sb
->delta_disks
== 0)
1292 sb
->feature_map
^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS
);
1294 sb
->delta_disks
= __cpu_to_le32(-__le32_to_cpu(sb
->delta_disks
));
1296 temp
= sb
->new_layout
;
1297 sb
->new_layout
= sb
->layout
;
1300 temp
= sb
->new_chunk
;
1301 sb
->new_chunk
= sb
->chunksize
;
1302 sb
->chunksize
= temp
;
1304 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_NEW_OFFSET
)) {
1305 long offset_delta
= (int32_t)__le32_to_cpu(sb
->new_offset
);
1306 sb
->data_offset
= __cpu_to_le64(__le64_to_cpu(sb
->data_offset
) + offset_delta
);
1307 sb
->new_offset
= __cpu_to_le32(-offset_delta
);
1308 sb
->data_size
= __cpu_to_le64(__le64_to_cpu(sb
->data_size
) - offset_delta
);
1311 } else if (strcmp(update
, "_reshape_progress")==0)
1312 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
1313 else if (strcmp(update
, "writemostly")==0)
1314 sb
->devflags
|= WriteMostly1
;
1315 else if (strcmp(update
, "readwrite")==0)
1316 sb
->devflags
&= ~WriteMostly1
;
1320 sb
->sb_csum
= calc_sb_1_csum(sb
);
1324 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
1325 unsigned long long size
, char *name
, char *homehost
,
1326 int *uuid
, unsigned long long data_offset
)
1328 struct mdp_superblock_1
*sb
;
1334 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
1335 pr_err("%s could not allocate superblock\n", __func__
);
1338 memset(sb
, 0, SUPER1_SIZE
);
1342 /* zeroing superblock */
1346 spares
= info
->working_disks
- info
->active_disks
;
1347 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
1348 pr_err("too many devices requested: %d+%d > %d\n",
1349 info
->raid_disks
, spares
, MAX_DEVS
);
1353 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
1354 sb
->major_version
= __cpu_to_le32(1);
1355 sb
->feature_map
= 0;
1359 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
1361 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1362 read(rfd
, sb
->set_uuid
, 16) != 16) {
1363 __u32 r
[4] = {random(), random(), random(), random()};
1364 memcpy(sb
->set_uuid
, r
, 16);
1366 if (rfd
>= 0) close(rfd
);
1369 if (name
== NULL
|| *name
== 0) {
1370 sprintf(defname
, "%d", info
->md_minor
);
1374 strchr(name
, ':')== NULL
&&
1375 strlen(homehost
)+1+strlen(name
) < 32) {
1376 strcpy(sb
->set_name
, homehost
);
1377 strcat(sb
->set_name
, ":");
1378 strcat(sb
->set_name
, name
);
1380 strcpy(sb
->set_name
, name
);
1382 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
1383 sb
->level
= __cpu_to_le32(info
->level
);
1384 sb
->layout
= __cpu_to_le32(info
->layout
);
1385 sb
->size
= __cpu_to_le64(size
*2ULL);
1386 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
1387 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
1389 sb
->data_offset
= __cpu_to_le64(data_offset
);
1390 sb
->data_size
= __cpu_to_le64(0);
1391 sb
->super_offset
= __cpu_to_le64(0);
1392 sb
->recovery_offset
= __cpu_to_le64(0);
1394 sb
->utime
= sb
->ctime
;
1395 sb
->events
= __cpu_to_le64(1);
1396 if (info
->state
& (1<<MD_SB_CLEAN
))
1397 sb
->resync_offset
= MaxSector
;
1399 sb
->resync_offset
= 0;
1400 sbsize
= sizeof(struct mdp_superblock_1
) + 2 * (info
->raid_disks
+ spares
);
1401 sbsize
= ROUND_UP(sbsize
, 512);
1402 sb
->max_dev
= __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
1404 memset(sb
->dev_roles
, 0xff, MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
1412 long long data_offset
;
1413 mdu_disk_info_t disk
;
1414 struct devinfo
*next
;
1417 /* Add a device to the superblock being created */
1418 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
1419 int fd
, char *devname
, unsigned long long data_offset
)
1421 struct mdp_superblock_1
*sb
= st
->sb
;
1422 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
1423 struct devinfo
*di
, **dip
;
1425 if ((dk
->state
& 6) == 6) /* active, sync */
1426 *rp
= __cpu_to_le16(dk
->raid_disk
);
1427 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
1432 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
1433 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
1434 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
1436 sb
->dev_number
= __cpu_to_le32(dk
->number
);
1437 sb
->devflags
= 0; /* don't copy another disks flags */
1438 sb
->sb_csum
= calc_sb_1_csum(sb
);
1440 dip
= (struct devinfo
**)&st
->info
;
1442 dip
= &(*dip
)->next
;
1443 di
= xmalloc(sizeof(struct devinfo
));
1445 di
->devname
= devname
;
1447 di
->data_offset
= data_offset
;
1455 static void locate_bitmap1(struct supertype
*st
, int fd
);
1457 static int store_super1(struct supertype
*st
, int fd
)
1459 struct mdp_superblock_1
*sb
= st
->sb
;
1460 unsigned long long sb_offset
;
1461 struct align_fd afd
;
1463 unsigned long long dsize
;
1465 if (!get_dev_size(fd
, NULL
, &dsize
))
1476 * Calculate the position of the superblock.
1477 * It is always aligned to a 4K boundary and
1478 * depending on minor_version, it can be:
1479 * 0: At least 8K, but less than 12K, from end of device
1480 * 1: At start of device
1481 * 2: 4K from start of device.
1483 switch(st
->minor_version
) {
1487 sb_offset
&= ~(4*2-1);
1499 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1500 0 != __le64_to_cpu(sb
->super_offset
)
1502 pr_err("internal error - sb_offset is wrong\n");
1506 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1509 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1511 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1514 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1515 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1516 (((char*)sb
)+MAX_SB_SIZE
);
1517 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1518 locate_bitmap1(st
, fd
);
1519 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1527 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1529 static unsigned long choose_bm_space(unsigned long devsize
)
1531 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1532 * if bigger than 200Gig, save 128k
1533 * NOTE: result must be multiple of 4K else bad things happen
1534 * on 4K-sector devices.
1536 if (devsize
< 64*2) return 0;
1537 if (devsize
- 64*2 >= 200*1024*1024*2)
1539 if (devsize
- 4*2 > 8*1024*1024*2)
1544 static void free_super1(struct supertype
*st
);
1547 static int write_init_super1(struct supertype
*st
)
1549 struct mdp_superblock_1
*sb
= st
->sb
;
1550 struct supertype
*refst
;
1553 unsigned long long bm_space
;
1555 unsigned long long dsize
, array_size
;
1556 unsigned long long sb_offset
;
1557 unsigned long long data_offset
;
1559 for (di
= st
->info
; di
; di
= di
->next
) {
1560 if (di
->disk
.state
& (1 << MD_DISK_FAULTY
))
1565 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
1568 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1569 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1570 sb
->devflags
|= WriteMostly1
;
1572 sb
->devflags
&= ~WriteMostly1
;
1574 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1575 read(rfd
, sb
->device_uuid
, 16) != 16) {
1576 __u32 r
[4] = {random(), random(), random(), random()};
1577 memcpy(sb
->device_uuid
, r
, 16);
1584 refst
= dup_super(st
);
1585 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1586 struct mdp_superblock_1
*refsb
= refst
->sb
;
1588 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1589 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
1590 /* same array, so preserve events and
1592 sb
->events
= refsb
->events
;
1593 /* bugs in 2.6.17 and earlier mean the
1594 * dev_number chosen in Manage must be preserved
1596 if (get_linux_version() >= 2006018)
1597 sb
->dev_number
= refsb
->dev_number
;
1603 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
1616 * Calculate the position of the superblock.
1617 * It is always aligned to a 4K boundary and
1618 * depending on minor_version, it can be:
1619 * 0: At least 8K, but less than 12K, from end of device
1620 * 1: At start of device
1621 * 2: 4K from start of device.
1622 * data_offset has already been set.
1624 array_size
= __le64_to_cpu(sb
->size
);
1625 /* work out how much space we left for a bitmap,
1626 * Add 8 sectors for bad block log */
1627 bm_space
= choose_bm_space(array_size
) + 8;
1629 data_offset
= di
->data_offset
;
1630 if (data_offset
== INVALID_SECTORS
)
1631 data_offset
= st
->data_offset
;
1632 switch(st
->minor_version
) {
1634 if (data_offset
== INVALID_SECTORS
)
1638 sb_offset
&= ~(4*2-1);
1639 sb
->data_offset
= __cpu_to_le64(data_offset
);
1640 sb
->super_offset
= __cpu_to_le64(sb_offset
);
1641 if (sb_offset
< array_size
+ bm_space
)
1642 bm_space
= sb_offset
- array_size
;
1643 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
1644 if (bm_space
>= 8) {
1645 sb
->bblog_size
= __cpu_to_le16(8);
1646 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
1650 sb
->super_offset
= __cpu_to_le64(0);
1651 if (data_offset
== INVALID_SECTORS
)
1654 sb
->data_offset
= __cpu_to_le64(data_offset
);
1655 sb
->data_size
= __cpu_to_le64(dsize
- data_offset
);
1656 if (data_offset
>= 8 + 32*2 + 8) {
1657 sb
->bblog_size
= __cpu_to_le16(8);
1658 sb
->bblog_offset
= __cpu_to_le32(8 + 32*2);
1659 } else if (data_offset
>= 16) {
1660 sb
->bblog_size
= __cpu_to_le16(8);
1661 sb
->bblog_offset
= __cpu_to_le32(data_offset
-8);
1666 sb
->super_offset
= __cpu_to_le64(sb_offset
);
1667 if (data_offset
== INVALID_SECTORS
)
1670 sb
->data_offset
= __cpu_to_le64(data_offset
);
1671 sb
->data_size
= __cpu_to_le64(dsize
- data_offset
);
1672 if (data_offset
>= 16 + 32*2 + 8) {
1673 sb
->bblog_size
= __cpu_to_le16(8);
1674 sb
->bblog_offset
= __cpu_to_le32(8 + 32*2);
1675 } else if (data_offset
>= 16+16) {
1676 sb
->bblog_size
= __cpu_to_le16(8);
1677 /* '8' sectors for the bblog, and another '8'
1678 * because we want offset from superblock, not
1681 sb
->bblog_offset
= __cpu_to_le32(data_offset
-8-8);
1685 pr_err("Failed to write invalid "
1686 "metadata format 1.%i to %s\n",
1687 st
->minor_version
, di
->devname
);
1691 if (conf_get_create_info()->bblist
== 0) {
1693 sb
->bblog_offset
= 0;
1696 sb
->sb_csum
= calc_sb_1_csum(sb
);
1697 rv
= store_super1(st
, di
->fd
);
1698 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
1699 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
1707 pr_err("Failed to write metadata to %s\n",
1714 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
1718 * 0 same, or first was empty, and second was copied
1719 * 1 second had wrong number
1721 * 3 wrong other info
1723 struct mdp_superblock_1
*first
= st
->sb
;
1724 struct mdp_superblock_1
*second
= tst
->sb
;
1726 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
1728 if (second
->major_version
!= __cpu_to_le32(1))
1732 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
1733 pr_err("%s could not allocate superblock\n", __func__
);
1736 memcpy(first
, second
, SUPER1_SIZE
);
1740 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
1743 if (first
->ctime
!= second
->ctime
||
1744 first
->level
!= second
->level
||
1745 first
->layout
!= second
->layout
||
1746 first
->size
!= second
->size
||
1747 first
->chunksize
!= second
->chunksize
||
1748 first
->raid_disks
!= second
->raid_disks
)
1753 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
1755 unsigned long long dsize
;
1756 unsigned long long sb_offset
;
1757 struct mdp_superblock_1
*super
;
1759 struct bitmap_super_s
*bsb
;
1760 struct misc_dev_info
*misc
;
1761 struct align_fd afd
;
1767 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1769 struct supertype tst
;
1770 __u64 bestctime
= 0;
1771 /* guess... choose latest ctime */
1772 memset(&tst
, 0, sizeof(tst
));
1774 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1775 switch(load_super1(&tst
, fd
, devname
)) {
1776 case 0: super
= tst
.sb
;
1777 if (bestvers
== -1 ||
1778 bestctime
< __le64_to_cpu(super
->ctime
)) {
1779 bestvers
= tst
.minor_version
;
1780 bestctime
= __le64_to_cpu(super
->ctime
);
1785 case 1: return 1; /*bad device */
1786 case 2: break; /* bad, try next */
1789 if (bestvers
!= -1) {
1791 tst
.minor_version
= bestvers
;
1793 tst
.max_devs
= MAX_DEVS
;
1794 rv
= load_super1(&tst
, fd
, devname
);
1801 if (!get_dev_size(fd
, devname
, &dsize
))
1807 pr_err("%s is too small for md: size is %llu sectors.\n",
1813 * Calculate the position of the superblock.
1814 * It is always aligned to a 4K boundary and
1815 * depending on minor_version, it can be:
1816 * 0: At least 8K, but less than 12K, from end of device
1817 * 1: At start of device
1818 * 2: 4K from start of device.
1820 switch(st
->minor_version
) {
1824 sb_offset
&= ~(4*2-1);
1836 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1838 pr_err("Cannot seek to superblock on %s: %s\n",
1839 devname
, strerror(errno
));
1843 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
1844 pr_err("%s could not allocate superblock\n",
1849 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
1851 pr_err("Cannot read superblock on %s\n",
1857 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1859 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1860 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1865 if (__le32_to_cpu(super
->major_version
) != 1) {
1867 pr_err("Cannot interpret superblock on %s - version is %d\n",
1868 devname
, __le32_to_cpu(super
->major_version
));
1872 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1874 pr_err("No superblock found on %s (super_offset is wrong)\n",
1881 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1883 misc
= (struct misc_dev_info
*) (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
1884 misc
->device_size
= dsize
;
1885 if (st
->data_offset
== INVALID_SECTORS
)
1886 st
->data_offset
= __le64_to_cpu(super
->data_offset
);
1888 /* Now check on the bitmap superblock */
1889 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1891 /* Read the bitmap superblock and make sure it looks
1892 * valid. If it doesn't clear the bit. An --assemble --force
1893 * should get that written out.
1895 locate_bitmap1(st
, fd
);
1896 if (aread(&afd
, bsb
, 512) != 512)
1899 uuid_from_super1(st
, uuid
);
1900 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1901 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1906 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
1907 & ~MD_FEATURE_BITMAP_OFFSET
);
1911 static struct supertype
*match_metadata_desc1(char *arg
)
1913 struct supertype
*st
= xcalloc(1, sizeof(*st
));
1915 st
->container_devnm
[0] = 0;
1917 st
->max_devs
= MAX_DEVS
;
1919 st
->data_offset
= INVALID_SECTORS
;
1920 /* leading zeros can be safely ignored. --detail generates them. */
1923 if (strcmp(arg
, "1.0") == 0 ||
1924 strcmp(arg
, "1.00") == 0) {
1925 st
->minor_version
= 0;
1928 if (strcmp(arg
, "1.1") == 0 ||
1929 strcmp(arg
, "1.01") == 0
1931 st
->minor_version
= 1;
1934 if (strcmp(arg
, "1.2") == 0 ||
1935 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
1936 strcmp(arg
, "default") == 0 ||
1937 #endif /* DEFAULT_OLD_METADATA */
1938 strcmp(arg
, "1.02") == 0) {
1939 st
->minor_version
= 2;
1942 if (strcmp(arg
, "1") == 0 ||
1943 strcmp(arg
, "default") == 0) {
1944 st
->minor_version
= -1;
1952 /* find available size on device with this devsize, using
1953 * superblock type st, and reserving 'reserve' sectors for
1956 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
,
1957 unsigned long long data_offset
)
1959 struct mdp_superblock_1
*super
= st
->sb
;
1966 if (__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
1967 /* hot-add. allow for actual size of bitmap */
1968 struct bitmap_super_s
*bsb
;
1969 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1970 bmspace
= bitmap_sectors(bsb
);
1973 /* Allow space for bad block log */
1974 if (super
->bblog_size
)
1975 bbspace
= __le16_to_cpu(super
->bblog_size
);
1977 if (st
->minor_version
< 0)
1978 /* not specified, so time to set default */
1979 st
->minor_version
= 2;
1981 if (data_offset
== INVALID_SECTORS
)
1982 data_offset
= st
->data_offset
;
1984 if (data_offset
!= INVALID_SECTORS
)
1985 switch(st
->minor_version
) {
1987 return devsize
- data_offset
- 8*2 - bbspace
;
1990 return devsize
- data_offset
;
1997 switch(st
->minor_version
) {
2000 return ((devsize
- 8*2 - bbspace
) & ~(4*2-1));
2002 /* at start, 4K for superblock and possible bitmap */
2003 return devsize
- 4*2 - bbspace
;
2005 /* 4k from start, 4K for superblock and possible bitmap */
2006 return devsize
- (4+4)*2 - bbspace
;
2012 add_internal_bitmap1(struct supertype
*st
,
2013 int *chunkp
, int delay
, int write_behind
,
2014 unsigned long long size
,
2015 int may_change
, int major
)
2018 * If not may_change, then this is a 'Grow' without sysfs support for
2019 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
2020 * If may_change, then this is create or a Grow with sysfs syupport,
2021 * and we can put the bitmap wherever we like.
2023 * size is in sectors, chunk is in bytes !!!
2026 unsigned long long bits
;
2027 unsigned long long max_bits
;
2028 unsigned long long min_chunk
;
2030 long bbl_offset
, bbl_size
;
2031 unsigned long long chunk
= *chunkp
;
2034 struct mdp_superblock_1
*sb
= st
->sb
;
2035 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
2038 if (__le64_to_cpu(sb
->data_size
) == 0)
2039 /* Must be creating the array, else data_size would be non-zero */
2041 switch(st
->minor_version
) {
2043 /* either 3K after the superblock (when hot-add),
2044 * or some amount of space before.
2047 /* We are creating array, so we *know* how much room has
2052 room
= choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2054 room
= __le64_to_cpu(sb
->super_offset
)
2055 - __le64_to_cpu(sb
->data_offset
)
2056 - __le64_to_cpu(sb
->data_size
);
2057 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2060 bbl_offset
= (__s32
)__le32_to_cpu(sb
->bblog_offset
);
2061 if (bbl_size
< -bbl_offset
)
2062 bbl_size
= -bbl_offset
;
2064 if (!may_change
|| (room
< 3*2 &&
2065 __le32_to_cpu(sb
->max_dev
) <= 384)) {
2070 offset
= 0; /* means movable offset */
2075 case 2: /* between superblock and data */
2079 room
= choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2081 room
= __le64_to_cpu(sb
->data_offset
)
2082 - __le64_to_cpu(sb
->super_offset
);
2083 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2085 room
= __le32_to_cpu(sb
->bblog_offset
) + bbl_size
;
2090 room
-= 2; /* Leave 1K for superblock */
2094 room
-= 4*2; /* leave 4K for superblock */
2104 if (chunk
== UnSet
&& room
> 128*2)
2105 /* Limit to 128K of bitmap when chunk size not requested */
2109 /* No room for a bitmap */
2112 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
2114 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
2115 bits
= (size
*512)/min_chunk
+1;
2116 while (bits
> max_bits
) {
2120 if (chunk
== UnSet
) {
2121 /* For practical purpose, 64Meg is a good
2122 * default chunk size for internal bitmaps.
2125 if (chunk
< 64*1024*1024)
2126 chunk
= 64*1024*1024;
2127 } else if (chunk
< min_chunk
)
2128 return 0; /* chunk size too small */
2129 if (chunk
== 0) /* rounding problem */
2133 /* start bitmap on a 4K boundary with enough space for
2136 bits
= (size
*512) / chunk
+ 1;
2137 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
2138 room
*= 8; /* convert 4K blocks to sectors */
2139 offset
= -room
- bbl_size
;
2142 sb
->bitmap_offset
= (int32_t)__cpu_to_le32(offset
);
2144 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
)
2145 | MD_FEATURE_BITMAP_OFFSET
);
2146 memset(bms
, 0, sizeof(*bms
));
2147 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
2148 bms
->version
= __cpu_to_le32(major
);
2149 uuid_from_super1(st
, uuid
);
2150 memcpy(bms
->uuid
, uuid
, 16);
2151 bms
->chunksize
= __cpu_to_le32(chunk
);
2152 bms
->daemon_sleep
= __cpu_to_le32(delay
);
2153 bms
->sync_size
= __cpu_to_le64(size
);
2154 bms
->write_behind
= __cpu_to_le32(write_behind
);
2160 static void locate_bitmap1(struct supertype
*st
, int fd
)
2162 unsigned long long offset
;
2163 struct mdp_superblock_1
*sb
;
2167 if (st
->ss
->load_super(st
, fd
, NULL
))
2168 return; /* no error I hope... */
2173 offset
= __le64_to_cpu(sb
->super_offset
);
2174 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
2177 lseek64(fd
, offset
<<9, 0);
2180 static int write_bitmap1(struct supertype
*st
, int fd
)
2182 struct mdp_superblock_1
*sb
= st
->sb
;
2183 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
2187 struct align_fd afd
;
2191 locate_bitmap1(st
, fd
);
2193 if (posix_memalign(&buf
, 4096, 4096))
2196 memset(buf
, 0xff, 4096);
2197 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
2199 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
2200 towrite
= (towrite
+7) >> 3; /* bits to bytes */
2201 towrite
+= sizeof(bitmap_super_t
);
2202 towrite
= ROUND_UP(towrite
, 512);
2203 while (towrite
> 0) {
2207 n
= awrite(&afd
, buf
, n
);
2212 memset(buf
, 0xff, 4096);
2222 static void free_super1(struct supertype
*st
)
2227 struct devinfo
*di
= st
->info
;
2228 st
->info
= di
->next
;
2237 static int validate_geometry1(struct supertype
*st
, int level
,
2238 int layout
, int raiddisks
,
2239 int *chunk
, unsigned long long size
,
2240 unsigned long long data_offset
,
2241 char *subdev
, unsigned long long *freesize
,
2244 unsigned long long ldsize
, devsize
;
2246 unsigned long long headroom
;
2249 if (level
== LEVEL_CONTAINER
) {
2251 pr_err("1.x metadata does not support containers\n");
2254 if (*chunk
== UnSet
)
2255 *chunk
= DEFAULT_CHUNK
;
2260 if (st
->minor_version
< 0)
2261 /* not specified, so time to set default */
2262 st
->minor_version
= 2;
2264 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
2267 pr_err("super1.x cannot open %s: %s\n",
2268 subdev
, strerror(errno
));
2272 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
2278 devsize
= ldsize
>> 9;
2284 /* creating: allow suitable space for bitmap */
2285 bmspace
= choose_bm_space(devsize
);
2287 if (data_offset
== INVALID_SECTORS
)
2288 data_offset
= st
->data_offset
;
2289 if (data_offset
== INVALID_SECTORS
)
2290 switch (st
->minor_version
) {
2296 /* Choose data offset appropriate for this device
2297 * and use as default for whole array.
2298 * The data_offset must allow for bitmap space
2299 * and base metadata, should allow for some headroom
2300 * for reshape, and should be rounded to multiple
2302 * Headroom is limited to 128M, but aim for about 0.1%
2304 headroom
= 128*1024*2;
2305 while ((headroom
<< 10) > devsize
&&
2307 headroom
/ 2 >= ((unsigned)(*chunk
)*2)*2))
2309 data_offset
= 12*2 + bmspace
+ headroom
;
2310 #define ONE_MEG (2*1024)
2311 if (data_offset
> ONE_MEG
)
2312 data_offset
= (data_offset
/ ONE_MEG
) * ONE_MEG
;
2315 if (st
->data_offset
== INVALID_SECTORS
)
2316 st
->data_offset
= data_offset
;
2317 switch(st
->minor_version
) {
2318 case 0: /* metadata at end. Round down and subtract space to reserve */
2319 devsize
= (devsize
& ~(4ULL*2-1));
2320 /* space for metadata, bblog, bitmap */
2321 devsize
-= 8*2 + 8 + bmspace
;
2325 devsize
-= data_offset
;
2328 *freesize
= devsize
;
2331 #endif /* MDASSEMBLE */
2333 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
)
2335 /* Create a v1.0 superblock based on 'info'*/
2337 struct mdp_superblock_1
*sb
;
2340 unsigned long long offset
;
2342 if (posix_memalign(&ret
, 4096, 1024) != 0)
2345 memset(ret
, 0, 1024);
2346 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
2347 sb
->major_version
= __cpu_to_le32(1);
2349 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
2350 sprintf(sb
->set_name
, "%d", sb0
->md_minor
);
2351 sb
->ctime
= __cpu_to_le32(info
->array
.ctime
+1);
2352 sb
->level
= __cpu_to_le32(info
->array
.level
);
2353 sb
->layout
= __cpu_to_le32(info
->array
.layout
);
2354 sb
->size
= __cpu_to_le64(info
->component_size
);
2355 sb
->chunksize
= __cpu_to_le32(info
->array
.chunk_size
/512);
2356 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
2357 if (info
->array
.level
> 0)
2358 sb
->data_size
= sb
->size
;
2360 sb
->data_size
= st
->ss
->avail_size(st
, st
->devsize
/512, 0);
2361 sb
->resync_offset
= MaxSector
;
2362 sb
->max_dev
= __cpu_to_le32(MD_SB_DISKS
);
2363 sb
->dev_number
= __cpu_to_le32(info
->disk
.number
);
2364 sb
->utime
= __cpu_to_le64(info
->array
.utime
);
2366 offset
= st
->devsize
/512 - 8*2;
2368 sb
->super_offset
= __cpu_to_le64(offset
);
2369 //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset);
2371 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
2372 read(rfd
, sb
->device_uuid
, 16) != 16) {
2373 __u32 r
[4] = {random(), random(), random(), random()};
2374 memcpy(sb
->device_uuid
, r
, 16);
2379 for (i
= 0; i
< MD_SB_DISKS
; i
++) {
2380 int state
= sb0
->disks
[i
].state
;
2381 sb
->dev_roles
[i
] = 0xFFFF;
2382 if ((state
& (1<<MD_DISK_SYNC
)) &&
2383 !(state
& (1<<MD_DISK_FAULTY
)))
2384 sb
->dev_roles
[i
] = __cpu_to_le16(sb0
->disks
[i
].raid_disk
);
2386 sb
->sb_csum
= calc_sb_1_csum(sb
);
2390 struct superswitch super1
= {
2392 .examine_super
= examine_super1
,
2393 .brief_examine_super
= brief_examine_super1
,
2394 .export_examine_super
= export_examine_super1
,
2395 .detail_super
= detail_super1
,
2396 .brief_detail_super
= brief_detail_super1
,
2397 .export_detail_super
= export_detail_super1
,
2398 .write_init_super
= write_init_super1
,
2399 .validate_geometry
= validate_geometry1
,
2400 .add_to_super
= add_to_super1
,
2401 .examine_badblocks
= examine_badblocks_super1
,
2402 .copy_metadata
= copy_metadata1
,
2404 .match_home
= match_home1
,
2405 .uuid_from_super
= uuid_from_super1
,
2406 .getinfo_super
= getinfo_super1
,
2407 .container_content
= container_content1
,
2408 .update_super
= update_super1
,
2409 .init_super
= init_super1
,
2410 .store_super
= store_super1
,
2411 .compare_super
= compare_super1
,
2412 .load_super
= load_super1
,
2413 .match_metadata_desc
= match_metadata_desc1
,
2414 .avail_size
= avail_size1
,
2415 .add_internal_bitmap
= add_internal_bitmap1
,
2416 .locate_bitmap
= locate_bitmap1
,
2417 .write_bitmap
= write_bitmap1
,
2418 .free_super
= free_super1
,
2419 #if __BYTE_ORDER == BIG_ENDIAN