2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2016 Neil Brown <neilb@suse.com>
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
; /* used for raid5, raid6, raid10, and raid0 */
47 __u64 size
; /* used size of component devices, in 512byte sectors */
49 __u32 chunksize
; /* in 512byte sectors */
52 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
53 * NOTE: signed, so bitmap can be before superblock
54 * only meaningful of feature_map[0] is set.
57 /* only meaningful when feature_map[MD_FEATURE_PPL] is set */
59 __s16 offset
; /* sectors from start of superblock that ppl starts */
60 __u16 size
; /* ppl size in sectors */
64 /* These are only valid with feature bit '4' */
65 __u32 new_level
; /* new level we are reshaping to */
66 __u64 reshape_position
; /* next address in array-space for reshape */
67 __u32 delta_disks
; /* change in number of raid_disks */
68 __u32 new_layout
; /* new layout */
69 __u32 new_chunk
; /* new chunk size (sectors) */
70 __u32 new_offset
; /* signed number to add to data_offset in new
71 * layout. 0 == no-change. This can be
72 * different on each device in the array.
75 /* constant this-device information - 64 bytes */
76 __u64 data_offset
; /* sector start of data, often 0 */
77 __u64 data_size
; /* sectors in this device that can be used for data */
78 __u64 super_offset
; /* sector start of this superblock */
80 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
81 __u64 journal_tail
;/* journal tail of journal device (from data_offset) */
83 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
84 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
85 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
86 __u8 devflags
; /* per-device flags. Only one defined...*/
87 #define WriteMostly1 1 /* mask for writemostly flag in above */
88 #define FailFast1 2 /* Device should get FailFast requests */
89 /* bad block log. If there are any bad blocks the feature flag is set.
90 * if offset and size are non-zero, that space is reserved and available.
92 __u8 bblog_shift
; /* shift from sectors to block size for badblock list */
93 __u16 bblog_size
; /* number of sectors reserved for badblock list */
94 __u32 bblog_offset
; /* sector offset from superblock to bblog, signed */
96 /* array state information - 64 bytes */
97 __u64 utime
; /* 40 bits second, 24 bits microseconds */
98 __u64 events
; /* incremented when superblock updated */
99 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
100 __u32 sb_csum
; /* checksum upto dev_roles[max_dev] */
101 __u32 max_dev
; /* size of dev_roles[] array to consider */
102 __u8 pad3
[64-32]; /* set to 0 when writing */
104 /* device state information. Indexed by dev_number.
106 * Note there are no per-device state flags. State information is rolled
107 * into the 'roles' value. If a device is spare or faulty, then it doesn't
108 * have a meaningful role.
110 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
113 #define MAX_SB_SIZE 4096
114 /* bitmap super size is 256, but we round up to a sector for alignment */
115 #define BM_SUPER_SIZE 512
116 #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2)
117 #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \
118 + sizeof(struct misc_dev_info))
120 struct misc_dev_info
{
124 #define MULTIPLE_PPL_AREA_SIZE_SUPER1 (1024 * 1024) /* Size of the whole
127 /* feature_map bits */
128 #define MD_FEATURE_BITMAP_OFFSET 1
129 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
132 #define MD_FEATURE_RESHAPE_ACTIVE 4
133 #define MD_FEATURE_BAD_BLOCKS 8 /* badblock list is not empty */
134 #define MD_FEATURE_REPLACEMENT 16 /* This device is replacing an
135 * active device with same 'role'.
136 * 'recovery_offset' is also set.
138 #define MD_FEATURE_RESHAPE_BACKWARDS 32 /* Reshape doesn't change number
139 * of devices, but is going
142 #define MD_FEATURE_NEW_OFFSET 64 /* new_offset must be honoured */
143 #define MD_FEATURE_BITMAP_VERSIONED 256 /* bitmap version number checked properly */
144 #define MD_FEATURE_JOURNAL 512 /* support write journal */
145 #define MD_FEATURE_PPL 1024 /* support PPL */
146 #define MD_FEATURE_MUTLIPLE_PPLS 2048 /* support for multiple PPLs */
147 #define MD_FEATURE_RAID0_LAYOUT 4096 /* layout is meaningful in RAID0 */
148 #define MD_FEATURE_ALL (MD_FEATURE_BITMAP_OFFSET \
149 |MD_FEATURE_RECOVERY_OFFSET \
150 |MD_FEATURE_RESHAPE_ACTIVE \
151 |MD_FEATURE_BAD_BLOCKS \
152 |MD_FEATURE_REPLACEMENT \
153 |MD_FEATURE_RESHAPE_BACKWARDS \
154 |MD_FEATURE_NEW_OFFSET \
155 |MD_FEATURE_BITMAP_VERSIONED \
156 |MD_FEATURE_JOURNAL \
158 |MD_FEATURE_MULTIPLE_PPLS \
159 |MD_FEATURE_RAID0_LAYOUT \
162 static int role_from_sb(struct mdp_superblock_1
*sb
)
167 d
= __le32_to_cpu(sb
->dev_number
);
168 if (d
< __le32_to_cpu(sb
->max_dev
))
169 role
= __le16_to_cpu(sb
->dev_roles
[d
]);
171 role
= MD_DISK_ROLE_SPARE
;
175 /* return how many bytes are needed for bitmap, for cluster-md each node
176 * should have it's own bitmap */
177 static unsigned int calc_bitmap_size(bitmap_super_t
*bms
, unsigned int boundary
)
179 unsigned long long bits
, bytes
;
181 bits
= bitmap_bits(__le64_to_cpu(bms
->sync_size
),
182 __le32_to_cpu(bms
->chunksize
));
183 bytes
= (bits
+7) >> 3;
184 bytes
+= sizeof(bitmap_super_t
);
185 bytes
= ROUND_UP(bytes
, boundary
);
190 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
192 unsigned int disk_csum
, csum
;
193 unsigned long long newcsum
;
194 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
195 unsigned int *isuper
= (unsigned int*)sb
;
197 /* make sure I can count... */
198 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
199 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
200 sizeof(struct mdp_superblock_1
) != 256) {
201 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
204 disk_csum
= sb
->sb_csum
;
207 for (; size
>=4; size
-= 4 ) {
208 newcsum
+= __le32_to_cpu(*isuper
);
213 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
215 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
216 sb
->sb_csum
= disk_csum
;
217 return __cpu_to_le32(csum
);
221 * Information related to file descriptor used for aligned reads/writes.
222 * Cache the block size.
229 static void init_afd(struct align_fd
*afd
, int fd
)
232 if (!get_dev_sector_size(afd
->fd
, NULL
, (unsigned int *)&afd
->blk_sz
))
236 static char abuf
[4096+4096];
238 static int aread(struct align_fd
*afd
, void *buf
, int len
)
241 * On devices with a 4K sector size, we need to read
242 * the full sector and copy relevant bits into
251 if (!bsize
|| bsize
> 4096 || len
> 4096) {
253 fprintf(stderr
, "WARNING - aread() called with invalid block size\n");
256 b
= ROUND_UP_PTR((char *)abuf
, 4096);
258 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
260 n
= read(afd
->fd
, b
, iosize
);
263 lseek(afd
->fd
, len
- n
, 1);
270 static int awrite(struct align_fd
*afd
, void *buf
, int len
)
273 * On devices with a 4K sector size, we need to write
274 * the full sector. We pre-read if the sector is larger
276 * The address must be sector-aligned.
283 if (!bsize
|| bsize
> 4096 || len
> 4096) {
285 fprintf(stderr
, "WARNING - awrite() called with invalid block size\n");
288 b
= ROUND_UP_PTR((char *)abuf
, 4096);
290 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
294 n
= read(afd
->fd
, b
, iosize
);
297 lseek(afd
->fd
, -n
, 1);
301 n
= write(afd
->fd
, b
, iosize
);
304 lseek(afd
->fd
, len
- n
, 1);
308 static inline unsigned int md_feature_any_ppl_on(__u32 feature_map
)
310 return ((__cpu_to_le32(feature_map
) &
311 (MD_FEATURE_PPL
| MD_FEATURE_MUTLIPLE_PPLS
)));
314 static inline unsigned int choose_ppl_space(int chunk
)
316 return (PPL_HEADER_SIZE
>> 9) + (chunk
> 128*2 ? chunk
: 128*2);
319 static void examine_super1(struct supertype
*st
, char *homehost
)
321 struct mdp_superblock_1
*sb
= st
->sb
;
322 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
329 int l
= homehost
? strlen(homehost
) : 0;
331 unsigned long long sb_offset
;
333 int inconsistent
= 0;
335 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
336 printf(" Version : 1");
337 sb_offset
= __le64_to_cpu(sb
->super_offset
);
340 else if (sb_offset
<= 8)
344 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
345 printf(" Array UUID : ");
346 for (i
=0; i
<16; i
++) {
347 if ((i
&3)==0 && i
!= 0) printf(":");
348 printf("%02x", sb
->set_uuid
[i
]);
351 printf(" Name : %.32s", sb
->set_name
);
352 if (l
> 0 && l
< 32 &&
353 sb
->set_name
[l
] == ':' &&
354 strncmp(sb
->set_name
, homehost
, l
) == 0)
355 printf(" (local to host %s)", homehost
);
357 if (bms
->nodes
> 0 &&
358 (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
))
359 printf(" Cluster Name : %-64s\n", bms
->cluster_name
);
360 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
361 printf(" Creation Time : %.24s\n", ctime(&atime
));
362 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
363 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
364 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
366 printf(" Avail Dev Size : %llu sectors%s\n",
367 (unsigned long long)__le64_to_cpu(sb
->data_size
),
368 human_size(__le64_to_cpu(sb
->data_size
)<<9));
369 if (__le32_to_cpu(sb
->level
) > 0) {
370 int ddsks
= 0, ddsks_denom
= 1;
371 switch(__le32_to_cpu(sb
->level
)) {
372 case 1: ddsks
=1;break;
374 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
375 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
377 layout
= __le32_to_cpu(sb
->layout
);
378 ddsks
= __le32_to_cpu(sb
->raid_disks
);
379 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
382 long long asize
= __le64_to_cpu(sb
->size
);
383 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
384 printf(" Array Size : %llu KiB%s\n",
385 asize
>> 10, human_size(asize
));
387 if (sb
->size
!= sb
->data_size
)
388 printf(" Used Dev Size : %llu sectors%s\n",
389 (unsigned long long)__le64_to_cpu(sb
->size
),
390 human_size(__le64_to_cpu(sb
->size
)<<9));
393 printf(" Data Offset : %llu sectors\n",
394 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
395 if (sb
->new_offset
&&
396 (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_NEW_OFFSET
)) {
397 unsigned long long offset
= __le64_to_cpu(sb
->data_offset
);
398 offset
+= (signed)(int32_t)__le32_to_cpu(sb
->new_offset
);
399 printf(" New Offset : %llu sectors\n", offset
);
401 printf(" Super Offset : %llu sectors\n",
402 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
403 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
404 printf("Recovery Offset : %llu sectors\n",
405 (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
407 st
->ss
->getinfo_super(st
, &info
, NULL
);
408 if (info
.space_after
!= 1 &&
409 !(__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_NEW_OFFSET
)) {
410 printf(" Unused Space : before=%llu sectors, ",
412 if (info
.space_after
< INT64_MAX
)
413 printf("after=%llu sectors\n", info
.space_after
);
415 printf("after=-%llu sectors DEVICE TOO SMALL\n",
416 UINT64_MAX
- info
.space_after
);
418 printf(" State : %s%s\n",
419 (__le64_to_cpu(sb
->resync_offset
)+1) ? "active":"clean",
420 (info
.space_after
> INT64_MAX
) ? " TRUNCATED DEVICE" : "");
421 printf(" Device UUID : ");
422 for (i
=0; i
<16; i
++) {
423 if ((i
&3)==0 && i
!= 0)
425 printf("%02x", sb
->device_uuid
[i
]);
429 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
430 printf("Internal Bitmap : %ld sectors from superblock\n",
431 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
432 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
433 printf(" PPL : %u sectors at offset %d sectors from superblock\n",
434 __le16_to_cpu(sb
->ppl
.size
),
435 __le16_to_cpu(sb
->ppl
.offset
));
437 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)) {
438 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)
439 __le64_to_cpu(sb
->reshape_position
)/2,
440 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
441 if (__le32_to_cpu(sb
->delta_disks
)) {
442 printf(" Delta Devices : %d",
443 __le32_to_cpu(sb
->delta_disks
));
444 printf(" (%d->%d)\n",
445 __le32_to_cpu(sb
->raid_disks
) -
446 __le32_to_cpu(sb
->delta_disks
),
447 __le32_to_cpu(sb
->raid_disks
));
448 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
449 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
451 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
452 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
453 printf(" New Level : %s\n", c
?c
:"-unknown-");
455 if (__le32_to_cpu(sb
->new_layout
) !=
456 __le32_to_cpu(sb
->layout
)) {
457 if (__le32_to_cpu(sb
->level
) == 5) {
458 c
= map_num(r5layout
,
459 __le32_to_cpu(sb
->new_layout
));
460 printf(" New Layout : %s\n", c
?c
:"-unknown-");
462 if (__le32_to_cpu(sb
->level
) == 6) {
463 c
= map_num(r6layout
,
464 __le32_to_cpu(sb
->new_layout
));
465 printf(" New Layout : %s\n", c
?c
:"-unknown-");
467 if (__le32_to_cpu(sb
->level
) == 10) {
468 printf(" New Layout :");
469 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
473 if (__le32_to_cpu(sb
->new_chunk
) !=
474 __le32_to_cpu(sb
->chunksize
))
475 printf(" New Chunksize : %dK\n",
476 __le32_to_cpu(sb
->new_chunk
)/2);
481 if (sb
->devflags
& WriteMostly1
)
482 printf(" write-mostly");
483 if (sb
->devflags
& FailFast1
)
488 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
489 printf(" Update Time : %.24s\n", ctime(&atime
));
491 if (sb
->bblog_size
&& sb
->bblog_offset
) {
492 printf(" Bad Block Log : %d entries available at offset %ld sectors",
493 __le16_to_cpu(sb
->bblog_size
)*512/8,
494 (long)(int32_t)__le32_to_cpu(sb
->bblog_offset
));
495 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
496 printf(" - bad blocks present.");
500 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
501 printf(" Checksum : %x - correct\n",
502 __le32_to_cpu(sb
->sb_csum
));
504 printf(" Checksum : %x - expected %x\n",
505 __le32_to_cpu(sb
->sb_csum
),
506 __le32_to_cpu(calc_sb_1_csum(sb
)));
507 printf(" Events : %llu\n",
508 (unsigned long long)__le64_to_cpu(sb
->events
));
510 if (__le32_to_cpu(sb
->level
) == 0 &&
511 (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
))) {
512 c
= map_num(r0layout
, __le32_to_cpu(sb
->layout
));
513 printf(" Layout : %s\n", c
?c
:"-unknown-");
515 if (__le32_to_cpu(sb
->level
) == 5) {
516 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
517 printf(" Layout : %s\n", c
?c
:"-unknown-");
519 if (__le32_to_cpu(sb
->level
) == 6) {
520 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
521 printf(" Layout : %s\n", c
?c
:"-unknown-");
523 if (__le32_to_cpu(sb
->level
) == 10) {
524 int lo
= __le32_to_cpu(sb
->layout
);
526 print_r10_layout(lo
);
529 switch(__le32_to_cpu(sb
->level
)) {
535 printf(" Chunk Size : %dK\n",
536 __le32_to_cpu(sb
->chunksize
)/2);
539 printf(" Rounding : %dK\n",
540 __le32_to_cpu(sb
->chunksize
)/2);
547 /* This turns out to just be confusing */
548 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
549 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
550 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != MD_DISK_ROLE_SPARE
)
552 for (d
= 0; d
< i
; d
++) {
553 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
556 if (role
== MD_DISK_ROLE_SPARE
)
559 if(role
== MD_DISK_ROLE_FAULTY
)
566 printf(" Device Role : ");
567 role
= role_from_sb(sb
);
568 if (role
>= MD_DISK_ROLE_FAULTY
)
570 else if (role
== MD_DISK_ROLE_JOURNAL
)
572 else if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_REPLACEMENT
))
573 printf("Replacement device %d\n", role
);
575 printf("Active device %d\n", role
);
577 printf(" Array State : ");
578 for (d
= 0; d
< __le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
581 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
582 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
586 if (cnt
== 2 && __le32_to_cpu(sb
->level
) > 0)
598 /* This is confusing too */
600 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
601 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
602 if (role
== MD_DISK_ROLE_FAULTY
)
606 printf(" %d failed", faulty
);
608 printf(" ('A' == active, '.' == missing, 'R' == replacing)");
610 for (d
= 0; d
< __le32_to_cpu(sb
->max_dev
); d
++) {
611 unsigned int r
= __le16_to_cpu(sb
->dev_roles
[d
]);
612 if (r
<= MD_DISK_ROLE_MAX
&&
613 r
> __le32_to_cpu(sb
->raid_disks
) + delta_extra
)
617 printf("WARNING Array state is inconsistent - each number should appear only once\n");
618 for (d
= 0; d
< __le32_to_cpu(sb
->max_dev
); d
++)
619 if (__le16_to_cpu(sb
->dev_roles
[d
]) >= MD_DISK_ROLE_FAULTY
)
622 printf(" %d:%d", d
, __le16_to_cpu(sb
->dev_roles
[d
]));
627 static void brief_examine_super1(struct supertype
*st
, int verbose
)
629 struct mdp_superblock_1
*sb
= st
->sb
;
631 unsigned long long sb_offset
;
633 char *c
= map_num(pers
, __le32_to_cpu(sb
->level
));
635 nm
= strchr(sb
->set_name
, ':');
638 else if (sb
->set_name
[0])
650 printf(" level=%s", c
);
651 sb_offset
= __le64_to_cpu(sb
->super_offset
);
653 printf(" metadata=1.1 ");
654 else if (sb_offset
<= 8)
655 printf(" metadata=1.2 ");
657 printf(" metadata=1.0 ");
659 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
661 for (i
= 0; i
< 16; i
++) {
662 if ((i
&3)==0 && i
!= 0)
664 printf("%02x", sb
->set_uuid
[i
]);
666 if (sb
->set_name
[0]) {
668 print_quoted(sb
->set_name
);
673 static void export_examine_super1(struct supertype
*st
)
675 struct mdp_superblock_1
*sb
= st
->sb
;
680 printf("MD_LEVEL=%s\n", map_num_s(pers
, __le32_to_cpu(sb
->level
)));
681 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
682 for (i
= 0; i
< 32; i
++)
683 if (sb
->set_name
[i
] == '\n' || sb
->set_name
[i
] == '\0') {
688 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
689 if (__le32_to_cpu(sb
->level
) > 0) {
690 int ddsks
= 0, ddsks_denom
= 1;
691 switch(__le32_to_cpu(sb
->level
)) {
697 ddsks
= __le32_to_cpu(sb
->raid_disks
)-1;
700 ddsks
= __le32_to_cpu(sb
->raid_disks
)-2;
703 layout
= __le32_to_cpu(sb
->layout
);
704 ddsks
= __le32_to_cpu(sb
->raid_disks
);
705 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
708 long long asize
= __le64_to_cpu(sb
->size
);
709 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
710 printf("MD_ARRAY_SIZE=%s\n",
711 human_size_brief(asize
, JEDEC
));
715 for (i
= 0; i
< 16; i
++) {
716 if ((i
&3) == 0 && i
!= 0)
718 printf("%02x", sb
->set_uuid
[i
]);
721 printf("MD_UPDATE_TIME=%llu\n",
722 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
723 printf("MD_DEV_UUID=");
724 for (i
= 0; i
< 16; i
++) {
725 if ((i
&3) == 0 && i
!= 0)
727 printf("%02x", sb
->device_uuid
[i
]);
730 printf("MD_EVENTS=%llu\n",
731 (unsigned long long)__le64_to_cpu(sb
->events
));
734 static int copy_metadata1(struct supertype
*st
, int from
, int to
)
736 /* Read superblock. If it looks good, write it out.
737 * Then if a bitmap is present, copy that.
738 * And if a bad-block-list is present, copy that too.
741 unsigned long long dsize
, sb_offset
;
742 const int bufsize
= 4*1024;
743 struct mdp_superblock_1 super
, *sb
;
745 if (posix_memalign(&buf
, 4096, bufsize
) != 0)
748 if (!get_dev_size(from
, NULL
, &dsize
))
754 switch(st
->minor_version
) {
758 sb_offset
&= ~(4*2-1);
770 if (lseek64(from
, sb_offset
<< 9, 0) < 0LL)
772 if (read(from
, buf
, bufsize
) != bufsize
)
776 super
= *sb
; // save most of sb for when we reuse buf
778 if (__le32_to_cpu(super
.magic
) != MD_SB_MAGIC
||
779 __le32_to_cpu(super
.major_version
) != 1 ||
780 __le64_to_cpu(super
.super_offset
) != sb_offset
||
781 calc_sb_1_csum(sb
) != super
.sb_csum
)
784 if (lseek64(to
, sb_offset
<< 9, 0) < 0LL)
786 if (write(to
, buf
, bufsize
) != bufsize
)
789 if (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
790 unsigned long long bitmap_offset
= sb_offset
;
791 int bytes
= 4096; // just an estimate.
793 struct align_fd afrom
, ato
;
795 init_afd(&afrom
, from
);
798 bitmap_offset
+= (int32_t)__le32_to_cpu(super
.bitmap_offset
);
800 if (lseek64(from
, bitmap_offset
<<9, 0) < 0)
802 if (lseek64(to
, bitmap_offset
<<9, 0) < 0)
805 for (written
= 0; written
< bytes
; ) {
806 int n
= bytes
- written
;
809 if (aread(&afrom
, buf
, n
) != n
)
812 /* have the header, can calculate
813 * correct bitmap bytes */
816 bytes
= calc_bitmap_size(bms
, 512);
820 if (awrite(&ato
, buf
, n
) != n
)
826 if (super
.bblog_size
!= 0 &&
827 __le16_to_cpu(super
.bblog_size
) <= 100 &&
828 super
.bblog_offset
!= 0 &&
829 (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BAD_BLOCKS
))) {
830 /* There is a bad block log */
831 unsigned long long bb_offset
= sb_offset
;
832 int bytes
= __le16_to_cpu(super
.bblog_size
) * 512;
834 struct align_fd afrom
, ato
;
836 init_afd(&afrom
, from
);
839 bb_offset
+= (int32_t)__le32_to_cpu(super
.bblog_offset
);
841 if (lseek64(from
, bb_offset
<<9, 0) < 0)
843 if (lseek64(to
, bb_offset
<<9, 0) < 0)
846 for (written
= 0; written
< bytes
; ) {
847 int n
= bytes
- written
;
850 if (aread(&afrom
, buf
, n
) != n
)
853 if (awrite(&ato
, buf
, n
) != n
)
867 static void detail_super1(struct supertype
*st
, char *homehost
, char *subarray
)
869 struct mdp_superblock_1
*sb
= st
->sb
;
870 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
872 int l
= homehost
? strlen(homehost
) : 0;
874 printf(" Name : %.32s", sb
->set_name
);
875 if (l
> 0 && l
< 32 && sb
->set_name
[l
] == ':' &&
876 strncmp(sb
->set_name
, homehost
, l
) == 0)
877 printf(" (local to host %s)", homehost
);
878 if (bms
->nodes
> 0 &&
879 (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
))
880 printf("\n Cluster Name : %-64s", bms
->cluster_name
);
881 printf("\n UUID : ");
882 for (i
= 0; i
< 16; i
++) {
883 if ((i
&3) == 0 && i
!= 0)
885 printf("%02x", sb
->set_uuid
[i
]);
887 printf("\n Events : %llu\n\n",
888 (unsigned long long)__le64_to_cpu(sb
->events
));
891 static void brief_detail_super1(struct supertype
*st
, char *subarray
)
893 struct mdp_superblock_1
*sb
= st
->sb
;
896 if (sb
->set_name
[0]) {
898 print_quoted(sb
->set_name
);
901 for (i
= 0; i
< 16; i
++) {
902 if ((i
& 3) == 0 && i
!= 0)
904 printf("%02x", sb
->set_uuid
[i
]);
908 static void export_detail_super1(struct supertype
*st
)
910 struct mdp_superblock_1
*sb
= st
->sb
;
914 for (i
= 0; i
< 32; i
++)
915 if (sb
->set_name
[i
] == '\n' || sb
->set_name
[i
] == '\0') {
920 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
923 static int examine_badblocks_super1(struct supertype
*st
, int fd
, char *devname
)
925 struct mdp_superblock_1
*sb
= st
->sb
;
926 unsigned long long offset
;
931 if (!sb
->bblog_size
|| __le16_to_cpu(sb
->bblog_size
) > 100 ||
933 printf("No bad-blocks list configured on %s\n", devname
);
936 if ((sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
)) == 0) {
937 printf("Bad-blocks list is empty in %s\n", devname
);
941 size
= __le16_to_cpu(sb
->bblog_size
)* 512;
942 if (posix_memalign((void**)&bbl
, 4096, size
) != 0) {
943 pr_err("could not allocate badblocks list\n");
946 offset
= __le64_to_cpu(sb
->super_offset
) +
947 (int)__le32_to_cpu(sb
->bblog_offset
);
949 if (lseek64(fd
, offset
, 0) < 0) {
950 pr_err("Cannot seek to bad-blocks list\n");
953 if (read(fd
, bbl
, size
) != size
) {
954 pr_err("Cannot read bad-blocks list\n");
957 /* 64bits per entry. 10 bits is block-count, 54 bits is block
958 * offset. Blocks are sectors unless bblog->shift makes them bigger
961 printf("Bad-blocks on %s:\n", devname
);
962 for (i
= 0; i
< size
/8; i
++, bbp
++) {
963 __u64 bb
= __le64_to_cpu(*bbp
);
964 int count
= bb
& 0x3ff;
965 unsigned long long sector
= bb
>> 10;
970 sector
<<= sb
->bblog_shift
;
971 count
<<= sb
->bblog_shift
;
973 printf("%20llu for %d sectors\n", sector
, count
);
978 static int match_home1(struct supertype
*st
, char *homehost
)
980 struct mdp_superblock_1
*sb
= st
->sb
;
981 int l
= homehost
? strlen(homehost
) : 0;
983 return (l
> 0 && l
< 32 && sb
->set_name
[l
] == ':' &&
984 strncmp(sb
->set_name
, homehost
, l
) == 0);
987 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
989 struct mdp_superblock_1
*super
= st
->sb
;
990 char *cuuid
= (char*)uuid
;
992 for (i
= 0; i
< 16; i
++)
993 cuuid
[i
] = super
->set_uuid
[i
];
996 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
998 struct mdp_superblock_1
*sb
= st
->sb
;
999 struct bitmap_super_s
*bsb
= (void*)(((char*)sb
)+MAX_SB_SIZE
);
1000 struct misc_dev_info
*misc
=
1001 (void*)(((char*)sb
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
1005 unsigned int map_disks
= info
->array
.raid_disks
;
1006 unsigned long long super_offset
;
1007 unsigned long long data_size
;
1009 memset(info
, 0, sizeof(*info
));
1010 info
->array
.major_version
= 1;
1011 info
->array
.minor_version
= st
->minor_version
;
1012 info
->array
.patch_version
= 0;
1013 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
1014 info
->array
.level
= __le32_to_cpu(sb
->level
);
1015 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
1016 info
->array
.md_minor
= -1;
1017 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
1018 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
1019 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
1021 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
) ? 1 : 0;
1023 super_offset
= __le64_to_cpu(sb
->super_offset
);
1024 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
1025 info
->component_size
= __le64_to_cpu(sb
->size
);
1026 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
1027 info
->bitmap_offset
= (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
1028 if (__le32_to_cpu(bsb
->nodes
) > 1)
1029 info
->array
.state
|= (1 << MD_SB_CLUSTERED
);
1030 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1031 info
->ppl_offset
= __le16_to_cpu(sb
->ppl
.offset
);
1032 info
->ppl_size
= __le16_to_cpu(sb
->ppl
.size
);
1033 info
->ppl_sector
= super_offset
+ info
->ppl_offset
;
1036 info
->disk
.major
= 0;
1037 info
->disk
.minor
= 0;
1038 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
1039 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
1040 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
1041 role
= MD_DISK_ROLE_FAULTY
;
1043 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
1045 if (info
->array
.level
<= 0)
1046 data_size
= __le64_to_cpu(sb
->data_size
);
1048 data_size
= __le64_to_cpu(sb
->size
);
1049 if (info
->data_offset
< super_offset
) {
1050 unsigned long long end
;
1051 info
->space_before
= info
->data_offset
;
1054 if (sb
->bblog_offset
&& sb
->bblog_size
) {
1055 unsigned long long bboffset
= super_offset
;
1056 bboffset
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
1061 if (super_offset
+ info
->bitmap_offset
+ info
->ppl_offset
< end
)
1062 end
= super_offset
+ info
->bitmap_offset
+
1065 if (info
->data_offset
+ data_size
< end
)
1066 info
->space_after
= end
- data_size
- info
->data_offset
;
1068 info
->space_after
= 0;
1070 unsigned long long earliest
;
1071 earliest
= super_offset
+ (32+4)*2; /* match kernel */
1072 if (info
->bitmap_offset
> 0) {
1073 unsigned long long bmend
= info
->bitmap_offset
;
1074 unsigned long long size
= calc_bitmap_size(bsb
, 4096);
1077 if (bmend
> earliest
)
1079 } else if (info
->ppl_offset
> 0) {
1080 unsigned long long pplend
;
1082 pplend
= info
->ppl_offset
+ info
->ppl_size
;
1083 if (pplend
> earliest
)
1086 if (sb
->bblog_offset
&& sb
->bblog_size
) {
1087 unsigned long long bbend
= super_offset
;
1088 bbend
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
1089 bbend
+= __le16_to_cpu(sb
->bblog_size
);
1090 if (bbend
> earliest
)
1093 if (earliest
< info
->data_offset
)
1094 info
->space_before
= info
->data_offset
- earliest
;
1096 info
->space_before
= 0;
1097 info
->space_after
= misc
->device_size
- data_size
-
1100 if (info
->space_before
== 0 && info
->space_after
== 0) {
1101 /* It will look like we don't support data_offset changes,
1102 * be we do - it's just that there is no room.
1103 * A change that reduced the number of devices should
1104 * still be allowed, so set the otherwise useless value of '1'
1106 info
->space_after
= 1;
1109 info
->disk
.raid_disk
= -1;
1111 case MD_DISK_ROLE_SPARE
:
1112 /* spare: not active, not sync, not faulty */
1113 info
->disk
.state
= 0;
1115 case MD_DISK_ROLE_FAULTY
:
1116 info
->disk
.state
= (1 << MD_DISK_FAULTY
); /* faulty */
1118 case MD_DISK_ROLE_JOURNAL
:
1119 info
->disk
.state
= (1 << MD_DISK_JOURNAL
);
1120 info
->disk
.raid_disk
= role
;
1121 /* journal uses all 4kB blocks*/
1122 info
->space_after
= (misc
->device_size
- info
->data_offset
) % 8;
1125 info
->disk
.state
= 6; /* active and in sync */
1126 info
->disk
.raid_disk
= role
;
1128 if (sb
->devflags
& WriteMostly1
)
1129 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
1130 if (sb
->devflags
& FailFast1
)
1131 info
->disk
.state
|= (1 << MD_DISK_FAILFAST
);
1132 info
->events
= __le64_to_cpu(sb
->events
);
1133 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
1134 info
->safe_mode_delay
= 200;
1136 memcpy(info
->uuid
, sb
->set_uuid
, 16);
1138 strncpy(info
->name
, sb
->set_name
, 32);
1141 if ((__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_REPLACEMENT
)) {
1142 info
->disk
.state
&= ~(1 << MD_DISK_SYNC
);
1143 info
->disk
.state
|= 1 << MD_DISK_REPLACEMENT
;
1146 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
1147 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
1149 info
->recovery_start
= MaxSector
;
1151 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
1152 info
->reshape_active
= 1;
1153 if ((sb
->feature_map
& __le32_to_cpu(MD_FEATURE_NEW_OFFSET
)) &&
1154 sb
->new_offset
!= 0)
1155 info
->reshape_active
|= RESHAPE_NO_BACKUP
;
1156 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
1157 info
->new_level
= __le32_to_cpu(sb
->new_level
);
1158 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
1159 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
1160 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
1161 if (info
->delta_disks
< 0)
1162 info
->array
.raid_disks
-= info
->delta_disks
;
1164 info
->reshape_active
= 0;
1166 info
->recovery_blocked
= info
->reshape_active
;
1169 for (i
=0; i
<map_disks
; i
++)
1171 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
1172 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
1173 if (/*role == MD_DISK_ROLE_SPARE || */role
< (unsigned) info
->array
.raid_disks
) {
1175 if (map
&& role
< map_disks
)
1180 info
->array
.working_disks
= working
;
1182 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_JOURNAL
)) {
1183 info
->journal_device_required
= 1;
1184 info
->consistency_policy
= CONSISTENCY_POLICY_JOURNAL
;
1185 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1186 info
->consistency_policy
= CONSISTENCY_POLICY_PPL
;
1187 } else if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
1188 info
->consistency_policy
= CONSISTENCY_POLICY_BITMAP
;
1189 } else if (info
->array
.level
<= 0) {
1190 info
->consistency_policy
= CONSISTENCY_POLICY_NONE
;
1192 info
->consistency_policy
= CONSISTENCY_POLICY_RESYNC
;
1195 info
->journal_clean
= 0;
1198 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
1200 struct mdinfo
*info
;
1205 info
= xmalloc(sizeof(*info
));
1206 getinfo_super1(st
, info
, NULL
);
1210 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
1211 char *update
, char *devname
, int verbose
,
1212 int uuid_set
, char *homehost
)
1214 /* NOTE: for 'assemble' and 'force' we need to return non-zero
1215 * if any change was made. For others, the return value is
1219 struct mdp_superblock_1
*sb
= st
->sb
;
1220 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
1222 if (strcmp(update
, "homehost") == 0 &&
1224 /* Note that 'homehost' is special as it is really
1229 c
= strchr(sb
->set_name
, ':');
1231 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
1233 strncpy(info
->name
, sb
->set_name
, 32);
1237 if (strcmp(update
, "force-one")==0) {
1238 /* Not enough devices for a working array,
1239 * so bring this one up-to-date
1241 if (sb
->events
!= __cpu_to_le64(info
->events
))
1243 sb
->events
= __cpu_to_le64(info
->events
);
1244 } else if (strcmp(update
, "force-array")==0) {
1245 /* Degraded array and 'force' requests to
1246 * maybe need to mark it 'clean'.
1248 switch(__le32_to_cpu(sb
->level
)) {
1252 /* need to force clean */
1253 if (sb
->resync_offset
!= MaxSector
)
1255 sb
->resync_offset
= MaxSector
;
1257 } else if (strcmp(update
, "assemble")==0) {
1258 int d
= info
->disk
.number
;
1260 if (info
->disk
.state
& (1<<MD_DISK_ACTIVE
))
1261 want
= info
->disk
.raid_disk
;
1262 else if (info
->disk
.state
& (1<<MD_DISK_JOURNAL
))
1263 want
= MD_DISK_ROLE_JOURNAL
;
1265 want
= MD_DISK_ROLE_SPARE
;
1266 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
1267 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
1270 if (info
->reshape_active
&&
1272 __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1273 info
->delta_disks
>= 0 &&
1274 info
->reshape_progress
<
1275 __le64_to_cpu(sb
->reshape_position
)) {
1276 sb
->reshape_position
=
1277 __cpu_to_le64(info
->reshape_progress
);
1280 if (info
->reshape_active
&&
1282 __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1283 info
->delta_disks
< 0 &&
1284 info
->reshape_progress
>
1285 __le64_to_cpu(sb
->reshape_position
)) {
1286 sb
->reshape_position
=
1287 __cpu_to_le64(info
->reshape_progress
);
1290 } else if (strcmp(update
, "linear-grow-new") == 0) {
1293 int max
= __le32_to_cpu(sb
->max_dev
);
1298 for (i
= 0; i
< max
; i
++)
1299 if (__le16_to_cpu(sb
->dev_roles
[i
]) >=
1300 MD_DISK_ROLE_FAULTY
)
1302 if (i
!= info
->disk
.number
)
1304 sb
->dev_number
= __cpu_to_le32(i
);
1307 sb
->max_dev
= __cpu_to_le32(max
+1);
1311 random_uuid(sb
->device_uuid
);
1313 sb
->dev_roles
[i
] = __cpu_to_le16(info
->disk
.raid_disk
);
1315 fd
= open(devname
, O_RDONLY
);
1317 unsigned long long ds
;
1318 get_dev_size(fd
, devname
, &ds
);
1321 if (__le64_to_cpu(sb
->super_offset
) <
1322 __le64_to_cpu(sb
->data_offset
)) {
1323 sb
->data_size
= __cpu_to_le64(
1324 ds
- __le64_to_cpu(sb
->data_offset
));
1327 ds
&= ~(unsigned long long)(4*2-1);
1328 sb
->super_offset
= __cpu_to_le64(ds
);
1329 sb
->data_size
= __cpu_to_le64(
1330 ds
- __le64_to_cpu(sb
->data_offset
));
1333 } else if (strcmp(update
, "linear-grow-update") == 0) {
1334 int max
= __le32_to_cpu(sb
->max_dev
);
1335 int i
= info
->disk
.number
;
1336 if (max
> MAX_DEVS
|| i
> MAX_DEVS
)
1341 sb
->max_dev
= __cpu_to_le32(max
+1);
1342 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
1343 sb
->dev_roles
[info
->disk
.number
] =
1344 __cpu_to_le16(info
->disk
.raid_disk
);
1345 } else if (strcmp(update
, "resync") == 0) {
1346 /* make sure resync happens */
1347 sb
->resync_offset
= 0ULL;
1348 } else if (strcmp(update
, "uuid") == 0) {
1349 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
1351 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
)
1352 memcpy(bms
->uuid
, sb
->set_uuid
, 16);
1353 } else if (strcmp(update
, "no-bitmap") == 0) {
1354 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
1355 if (bms
->version
== BITMAP_MAJOR_CLUSTERED
&& !IsBitmapDirty(devname
))
1356 sb
->resync_offset
= MaxSector
;
1357 } else if (strcmp(update
, "bbl") == 0) {
1358 /* only possible if there is room after the bitmap, or if
1359 * there is no bitmap
1361 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
1362 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
1363 long bitmap_offset
= 0;
1364 long bm_sectors
= 0;
1367 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1368 bitmap_offset
= (long)__le32_to_cpu(sb
->bitmap_offset
);
1369 bm_sectors
= calc_bitmap_size(bms
, 4096) >> 9;
1370 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1371 bitmap_offset
= (long)__le16_to_cpu(sb
->ppl
.offset
);
1372 bm_sectors
= (long)__le16_to_cpu(sb
->ppl
.size
);
1375 if (sb_offset
< data_offset
) {
1377 * 1.1 or 1.2. Put bbl after bitmap leaving
1381 bb_offset
= sb_offset
+ 8;
1382 if (bm_sectors
&& bitmap_offset
> 0)
1383 bb_offset
= bitmap_offset
+ bm_sectors
;
1384 while (bb_offset
< (long)sb_offset
+ 8 + 32*2 &&
1385 bb_offset
+ 8+8 <= (long)data_offset
)
1386 /* too close to bitmap, and room to grow */
1388 if (bb_offset
+ 8 <= (long)data_offset
) {
1389 sb
->bblog_size
= __cpu_to_le16(8);
1390 sb
->bblog_offset
= __cpu_to_le32(bb_offset
);
1393 /* 1.0 - Put bbl just before super block */
1394 if (bm_sectors
&& bitmap_offset
< 0)
1395 space
= -bitmap_offset
- bm_sectors
;
1397 space
= sb_offset
- data_offset
-
1398 __le64_to_cpu(sb
->data_size
);
1400 sb
->bblog_size
= __cpu_to_le16(8);
1401 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
1404 } else if (strcmp(update
, "no-bbl") == 0) {
1405 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
1406 pr_err("Cannot remove active bbl from %s\n",devname
);
1409 sb
->bblog_shift
= 0;
1410 sb
->bblog_offset
= 0;
1412 } else if (strcmp(update
, "force-no-bbl") == 0) {
1413 sb
->feature_map
&= ~ __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
);
1415 sb
->bblog_shift
= 0;
1416 sb
->bblog_offset
= 0;
1417 } else if (strcmp(update
, "ppl") == 0) {
1418 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
1419 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
1420 unsigned long long data_size
= __le64_to_cpu(sb
->data_size
);
1421 long bb_offset
= __le32_to_cpu(sb
->bblog_offset
);
1425 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1426 pr_err("Cannot add PPL to array with bitmap\n");
1430 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_JOURNAL
)) {
1431 pr_err("Cannot add PPL to array with journal\n");
1435 if (sb_offset
< data_offset
) {
1437 space
= bb_offset
- 8;
1439 space
= data_offset
- sb_offset
- 8;
1442 offset
= -(sb_offset
- data_offset
- data_size
);
1443 if (offset
< INT16_MIN
)
1445 space
= -(offset
- bb_offset
);
1448 if (space
< (PPL_HEADER_SIZE
>> 9) + 8) {
1449 pr_err("Not enough space to add ppl\n");
1453 if (space
>= (MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9)) {
1454 space
= (MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9);
1456 int optimal_space
= choose_ppl_space(
1457 __le32_to_cpu(sb
->chunksize
));
1458 if (space
> optimal_space
)
1459 space
= optimal_space
;
1460 if (space
> UINT16_MAX
)
1464 sb
->ppl
.offset
= __cpu_to_le16(offset
);
1465 sb
->ppl
.size
= __cpu_to_le16(space
);
1466 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_PPL
);
1467 } else if (strcmp(update
, "no-ppl") == 0) {
1468 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_PPL
|
1469 MD_FEATURE_MUTLIPLE_PPLS
);
1470 } else if (strcmp(update
, "name") == 0) {
1471 if (info
->name
[0] == 0)
1472 sprintf(info
->name
, "%d", info
->array
.md_minor
);
1473 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
1475 strchr(info
->name
, ':') == NULL
&&
1476 strlen(homehost
)+1+strlen(info
->name
) < 32) {
1477 strcpy(sb
->set_name
, homehost
);
1478 strcat(sb
->set_name
, ":");
1479 strcat(sb
->set_name
, info
->name
);
1483 namelen
= min((int)strlen(info
->name
),
1484 (int)sizeof(sb
->set_name
) - 1);
1485 memcpy(sb
->set_name
, info
->name
, namelen
);
1486 memset(&sb
->set_name
[namelen
], '\0',
1487 sizeof(sb
->set_name
) - namelen
);
1489 } else if (strcmp(update
, "devicesize") == 0 &&
1490 __le64_to_cpu(sb
->super_offset
) <
1491 __le64_to_cpu(sb
->data_offset
)) {
1492 /* set data_size to device size less data_offset */
1493 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
1494 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
1495 sb
->data_size
= __cpu_to_le64(
1496 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
1497 } else if (strncmp(update
, "revert-reshape", 14) == 0) {
1499 if (!(sb
->feature_map
&
1500 __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)))
1501 pr_err("No active reshape to revert on %s\n",
1505 unsigned long long reshape_sectors
;
1508 /* If the reshape hasn't started, just stop it.
1509 * It is conceivable that a stripe was modified but
1510 * the metadata not updated. In that case the backup
1511 * should have been used to get passed the critical stage.
1512 * If that couldn't happen, the "-nobackup" version
1515 if (strcmp(update
, "revert-reshape-nobackup") == 0 &&
1516 sb
->reshape_position
== 0 &&
1517 (__le32_to_cpu(sb
->delta_disks
) > 0 ||
1518 (__le32_to_cpu(sb
->delta_disks
) == 0 &&
1519 !(sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS
))))) {
1520 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
);
1521 sb
->raid_disks
= __cpu_to_le32(__le32_to_cpu(sb
->raid_disks
) -
1522 __le32_to_cpu(sb
->delta_disks
));
1523 sb
->delta_disks
= 0;
1526 /* reshape_position is a little messy.
1527 * Its value must be a multiple of the larger
1528 * chunk size, and of the "after" data disks.
1529 * So when reverting we need to change it to
1530 * be a multiple of the new "after" data disks,
1531 * which is the old "before".
1532 * If it isn't already a multiple of 'before',
1533 * the only thing we could do would be
1534 * copy some block around on the disks, which
1535 * is easy to get wrong.
1536 * So we reject a revert-reshape unless the
1537 * alignment is good.
1539 if (is_level456(__le32_to_cpu(sb
->level
))) {
1541 __le64_to_cpu(sb
->reshape_position
);
1542 reshape_chunk
= __le32_to_cpu(sb
->new_chunk
);
1543 reshape_chunk
*= __le32_to_cpu(sb
->raid_disks
) -
1544 __le32_to_cpu(sb
->delta_disks
) -
1545 (__le32_to_cpu(sb
->level
)==6 ? 2 : 1);
1546 if (reshape_sectors
% reshape_chunk
) {
1547 pr_err("Reshape position is not suitably aligned.\n");
1548 pr_err("Try normal assembly and stop again\n");
1553 __cpu_to_le32(__le32_to_cpu(sb
->raid_disks
) -
1554 __le32_to_cpu(sb
->delta_disks
));
1555 if (sb
->delta_disks
== 0)
1556 sb
->feature_map
^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS
);
1558 sb
->delta_disks
= __cpu_to_le32(-__le32_to_cpu(sb
->delta_disks
));
1560 temp
= sb
->new_layout
;
1561 sb
->new_layout
= sb
->layout
;
1564 temp
= sb
->new_chunk
;
1565 sb
->new_chunk
= sb
->chunksize
;
1566 sb
->chunksize
= temp
;
1568 if (sb
->feature_map
&
1569 __cpu_to_le32(MD_FEATURE_NEW_OFFSET
)) {
1571 (int32_t)__le32_to_cpu(sb
->new_offset
);
1572 sb
->data_offset
= __cpu_to_le64(__le64_to_cpu(sb
->data_offset
) + offset_delta
);
1573 sb
->new_offset
= __cpu_to_le32(-offset_delta
);
1574 sb
->data_size
= __cpu_to_le64(__le64_to_cpu(sb
->data_size
) - offset_delta
);
1578 } else if (strcmp(update
, "_reshape_progress") == 0)
1579 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
1580 else if (strcmp(update
, "writemostly") == 0)
1581 sb
->devflags
|= WriteMostly1
;
1582 else if (strcmp(update
, "readwrite") == 0)
1583 sb
->devflags
&= ~WriteMostly1
;
1584 else if (strcmp(update
, "failfast") == 0)
1585 sb
->devflags
|= FailFast1
;
1586 else if (strcmp(update
, "nofailfast") == 0)
1587 sb
->devflags
&= ~FailFast1
;
1588 else if (strcmp(update
, "layout-original") == 0 ||
1589 strcmp(update
, "layout-alternate") == 0 ||
1590 strcmp(update
, "layout-unspecified") == 0) {
1591 if (__le32_to_cpu(sb
->level
) != 0) {
1592 pr_err("%s: %s only supported for RAID0\n",
1593 devname
?:"", update
);
1595 } else if (strcmp(update
, "layout-unspecified") == 0) {
1596 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
1599 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
1600 sb
->layout
= __cpu_to_le32(update
[7] == 'o' ? 1 : 2);
1605 sb
->sb_csum
= calc_sb_1_csum(sb
);
1610 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
1611 struct shape
*s
, char *name
, char *homehost
,
1612 int *uuid
, unsigned long long data_offset
)
1614 struct mdp_superblock_1
*sb
;
1619 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
1620 pr_err("could not allocate superblock\n");
1623 memset(sb
, 0, SUPER1_SIZE
);
1627 /* zeroing superblock */
1631 spares
= info
->working_disks
- info
->active_disks
;
1632 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
1633 pr_err("too many devices requested: %d+%d > %d\n",
1634 info
->raid_disks
, spares
, MAX_DEVS
);
1638 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
1639 sb
->major_version
= __cpu_to_le32(1);
1640 sb
->feature_map
= 0;
1644 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
1646 random_uuid(sb
->set_uuid
);;
1648 if (name
== NULL
|| *name
== 0) {
1649 sprintf(defname
, "%d", info
->md_minor
);
1653 strchr(name
, ':')== NULL
&&
1654 strlen(homehost
)+1+strlen(name
) < 32) {
1655 strcpy(sb
->set_name
, homehost
);
1656 strcat(sb
->set_name
, ":");
1657 strcat(sb
->set_name
, name
);
1661 namelen
= min((int)strlen(name
),
1662 (int)sizeof(sb
->set_name
) - 1);
1663 memcpy(sb
->set_name
, name
, namelen
);
1664 memset(&sb
->set_name
[namelen
], '\0',
1665 sizeof(sb
->set_name
) - namelen
);
1668 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
1669 sb
->level
= __cpu_to_le32(info
->level
);
1670 sb
->layout
= __cpu_to_le32(info
->layout
);
1671 sb
->size
= __cpu_to_le64(s
->size
*2ULL);
1672 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
1673 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
1675 sb
->data_offset
= __cpu_to_le64(data_offset
);
1676 sb
->data_size
= __cpu_to_le64(0);
1677 sb
->super_offset
= __cpu_to_le64(0);
1678 sb
->recovery_offset
= __cpu_to_le64(0);
1680 sb
->utime
= sb
->ctime
;
1681 sb
->events
= __cpu_to_le64(1);
1682 if (info
->state
& (1<<MD_SB_CLEAN
))
1683 sb
->resync_offset
= MaxSector
;
1685 sb
->resync_offset
= 0;
1686 sbsize
= sizeof(struct mdp_superblock_1
) +
1687 2 * (info
->raid_disks
+ spares
);
1688 sbsize
= ROUND_UP(sbsize
, 512);
1690 __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
1692 memset(sb
->dev_roles
, 0xff,
1693 MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
1695 if (s
->consistency_policy
== CONSISTENCY_POLICY_PPL
)
1696 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_PPL
);
1704 long long data_offset
;
1705 unsigned long long dev_size
;
1706 mdu_disk_info_t disk
;
1707 struct devinfo
*next
;
1710 /* Add a device to the superblock being created */
1711 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
1712 int fd
, char *devname
, unsigned long long data_offset
)
1714 struct mdp_superblock_1
*sb
= st
->sb
;
1715 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
1716 struct devinfo
*di
, **dip
;
1719 dk_state
= dk
->state
& ~(1<<MD_DISK_FAILFAST
);
1720 if ((dk_state
& (1<<MD_DISK_ACTIVE
)) &&
1721 (dk_state
& (1<<MD_DISK_SYNC
)))/* active, sync */
1722 *rp
= __cpu_to_le16(dk
->raid_disk
);
1723 else if (dk_state
& (1<<MD_DISK_JOURNAL
))
1724 *rp
= MD_DISK_ROLE_JOURNAL
;
1725 else if ((dk_state
& ~(1<<MD_DISK_ACTIVE
)) == 0)
1726 /* active or idle -> spare */
1727 *rp
= MD_DISK_ROLE_SPARE
;
1729 *rp
= MD_DISK_ROLE_FAULTY
;
1731 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
1732 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
1733 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
1735 sb
->dev_number
= __cpu_to_le32(dk
->number
);
1736 sb
->devflags
= 0; /* don't copy another disks flags */
1737 sb
->sb_csum
= calc_sb_1_csum(sb
);
1739 dip
= (struct devinfo
**)&st
->info
;
1741 dip
= &(*dip
)->next
;
1742 di
= xmalloc(sizeof(struct devinfo
));
1744 di
->devname
= devname
;
1746 di
->data_offset
= data_offset
;
1747 get_dev_size(fd
, NULL
, &di
->dev_size
);
1754 static int locate_bitmap1(struct supertype
*st
, int fd
, int node_num
);
1756 static int store_super1(struct supertype
*st
, int fd
)
1758 struct mdp_superblock_1
*sb
= st
->sb
;
1759 unsigned long long sb_offset
;
1760 struct align_fd afd
;
1762 unsigned long long dsize
;
1764 if (!get_dev_size(fd
, NULL
, &dsize
))
1775 * Calculate the position of the superblock.
1776 * It is always aligned to a 4K boundary and
1777 * depending on minor_version, it can be:
1778 * 0: At least 8K, but less than 12K, from end of device
1779 * 1: At start of device
1780 * 2: 4K from start of device.
1782 switch(st
->minor_version
) {
1786 sb_offset
&= ~(4*2-1);
1798 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1799 0 != __le64_to_cpu(sb
->super_offset
)
1801 pr_err("internal error - sb_offset is wrong\n");
1805 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1808 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1810 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1813 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1814 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1815 (((char*)sb
)+MAX_SB_SIZE
);
1816 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1817 locate_bitmap1(st
, fd
, 0);
1818 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1827 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1829 static unsigned long choose_bm_space(unsigned long devsize
)
1831 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1832 * if bigger than 200Gig, save 128k
1833 * NOTE: result must be multiple of 4K else bad things happen
1834 * on 4K-sector devices.
1838 if (devsize
- 64*2 >= 200*1024*1024*2)
1840 if (devsize
- 4*2 > 8*1024*1024*2)
1845 static void free_super1(struct supertype
*st
);
1847 __u32
crc32c_le(__u32 crc
, unsigned char const *p
, size_t len
);
1849 static int write_init_ppl1(struct supertype
*st
, struct mdinfo
*info
, int fd
)
1851 struct mdp_superblock_1
*sb
= st
->sb
;
1853 struct ppl_header
*ppl_hdr
;
1856 /* first clear entire ppl space */
1857 ret
= zero_disk_range(fd
, info
->ppl_sector
, info
->ppl_size
);
1861 ret
= posix_memalign(&buf
, 4096, PPL_HEADER_SIZE
);
1863 pr_err("Failed to allocate PPL header buffer\n");
1867 memset(buf
, 0, PPL_HEADER_SIZE
);
1869 memset(ppl_hdr
->reserved
, 0xff, PPL_HDR_RESERVED
);
1870 ppl_hdr
->signature
= __cpu_to_le32(~crc32c_le(~0, sb
->set_uuid
,
1871 sizeof(sb
->set_uuid
)));
1872 ppl_hdr
->checksum
= __cpu_to_le32(~crc32c_le(~0, buf
, PPL_HEADER_SIZE
));
1874 if (lseek64(fd
, info
->ppl_sector
* 512, SEEK_SET
) < 0) {
1876 perror("Failed to seek to PPL header location");
1879 if (!ret
&& write(fd
, buf
, PPL_HEADER_SIZE
) != PPL_HEADER_SIZE
) {
1881 perror("Write PPL header failed");
1891 #define META_BLOCK_SIZE 4096
1893 static int write_empty_r5l_meta_block(struct supertype
*st
, int fd
)
1895 struct r5l_meta_block
*mb
;
1896 struct mdp_superblock_1
*sb
= st
->sb
;
1897 struct align_fd afd
;
1902 if (posix_memalign((void**)&mb
, 4096, META_BLOCK_SIZE
) != 0) {
1903 pr_err("Could not allocate memory for the meta block.\n");
1907 memset(mb
, 0, META_BLOCK_SIZE
);
1909 mb
->magic
= __cpu_to_le32(R5LOG_MAGIC
);
1910 mb
->version
= R5LOG_VERSION
;
1911 mb
->meta_size
= __cpu_to_le32(sizeof(struct r5l_meta_block
));
1912 mb
->seq
= __cpu_to_le64(random32());
1913 mb
->position
= __cpu_to_le64(0);
1915 crc
= crc32c_le(0xffffffff, sb
->set_uuid
, sizeof(sb
->set_uuid
));
1916 crc
= crc32c_le(crc
, (void *)mb
, META_BLOCK_SIZE
);
1919 if (lseek64(fd
, __le64_to_cpu(sb
->data_offset
) * 512, 0) < 0LL) {
1920 pr_err("cannot seek to offset of the meta block\n");
1924 if (awrite(&afd
, mb
, META_BLOCK_SIZE
) != META_BLOCK_SIZE
) {
1925 pr_err("failed to store write the meta block \n");
1938 static int write_init_super1(struct supertype
*st
)
1940 struct mdp_superblock_1
*sb
= st
->sb
;
1941 struct supertype
*refst
;
1943 unsigned long long bm_space
;
1945 unsigned long long dsize
, array_size
;
1946 unsigned long long sb_offset
;
1947 unsigned long long data_offset
;
1949 int raid0_need_layout
= 0;
1951 for (di
= st
->info
; di
; di
= di
->next
) {
1952 if (di
->disk
.state
& (1 << MD_DISK_JOURNAL
))
1953 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_JOURNAL
);
1954 if (sb
->level
== 0 && sb
->layout
!= 0) {
1955 struct devinfo
*di2
= st
->info
;
1956 unsigned long long s1
, s2
;
1958 if (di
->data_offset
!= INVALID_SECTORS
)
1959 s1
-= di
->data_offset
;
1960 s1
/= __le32_to_cpu(sb
->chunksize
);
1962 if (di2
->data_offset
!= INVALID_SECTORS
)
1963 s2
-= di2
->data_offset
;
1964 s2
/= __le32_to_cpu(sb
->chunksize
);
1966 raid0_need_layout
= 1;
1970 for (di
= st
->info
; di
; di
= di
->next
) {
1971 if (di
->disk
.state
& (1 << MD_DISK_FAULTY
))
1976 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
1979 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1980 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1981 sb
->devflags
|= WriteMostly1
;
1983 sb
->devflags
&= ~WriteMostly1
;
1984 if (di
->disk
.state
& (1<<MD_DISK_FAILFAST
))
1985 sb
->devflags
|= FailFast1
;
1987 sb
->devflags
&= ~FailFast1
;
1989 random_uuid(sb
->device_uuid
);
1991 if (!(di
->disk
.state
& (1<<MD_DISK_JOURNAL
)))
1994 refst
= dup_super(st
);
1995 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1996 struct mdp_superblock_1
*refsb
= refst
->sb
;
1998 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1999 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
2000 /* same array, so preserve events and
2002 sb
->events
= refsb
->events
;
2003 /* bugs in 2.6.17 and earlier mean the
2004 * dev_number chosen in Manage must be preserved
2006 if (get_linux_version() >= 2006018)
2007 sb
->dev_number
= refsb
->dev_number
;
2013 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
2026 * Calculate the position of the superblock.
2027 * It is always aligned to a 4K boundary and
2028 * depending on minor_version, it can be:
2029 * 0: At least 8K, but less than 12K, from end of device
2030 * 1: At start of device
2031 * 2: 4K from start of device.
2032 * data_offset has already been set.
2034 array_size
= __le64_to_cpu(sb
->size
);
2036 /* work out how much space we left for a bitmap */
2037 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
2038 bitmap_super_t
*bms
= (bitmap_super_t
*)
2039 (((char *)sb
) + MAX_SB_SIZE
);
2040 bm_space
= calc_bitmap_size(bms
, 4096) >> 9;
2041 bm_offset
= (long)__le32_to_cpu(sb
->bitmap_offset
);
2042 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
2043 bm_space
= MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9;
2044 if (st
->minor_version
== 0)
2045 bm_offset
= -bm_space
- 8;
2048 sb
->ppl
.offset
= __cpu_to_le16(bm_offset
);
2049 sb
->ppl
.size
= __cpu_to_le16(bm_space
);
2051 bm_space
= choose_bm_space(array_size
);
2055 data_offset
= di
->data_offset
;
2056 if (data_offset
== INVALID_SECTORS
)
2057 data_offset
= st
->data_offset
;
2058 switch(st
->minor_version
) {
2060 /* Add 8 sectors for bad block log */
2062 if (data_offset
== INVALID_SECTORS
)
2066 sb_offset
&= ~(4*2-1);
2067 sb
->data_offset
= __cpu_to_le64(data_offset
);
2068 sb
->super_offset
= __cpu_to_le64(sb_offset
);
2069 if (sb_offset
< array_size
+ bm_space
)
2070 bm_space
= sb_offset
- array_size
;
2071 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
2072 if (bm_space
>= 8) {
2073 sb
->bblog_size
= __cpu_to_le16(8);
2074 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
2079 sb_offset
= st
->minor_version
== 2 ? 8 : 0;
2080 sb
->super_offset
= __cpu_to_le64(sb_offset
);
2081 if (data_offset
== INVALID_SECTORS
)
2082 data_offset
= sb_offset
+ 16;
2084 sb
->data_offset
= __cpu_to_le64(data_offset
);
2085 sb
->data_size
= __cpu_to_le64(dsize
- data_offset
);
2086 if (data_offset
>= sb_offset
+bm_offset
+bm_space
+8) {
2087 sb
->bblog_size
= __cpu_to_le16(8);
2088 sb
->bblog_offset
= __cpu_to_le32(bm_offset
+
2090 } else if (data_offset
>= sb_offset
+ 16) {
2091 sb
->bblog_size
= __cpu_to_le16(8);
2092 /* '8' sectors for the bblog, and 'sb_offset'
2093 * because we want offset from superblock, not
2096 sb
->bblog_offset
= __cpu_to_le32(data_offset
-
2101 pr_err("Failed to write invalid metadata format 1.%i to %s\n",
2102 st
->minor_version
, di
->devname
);
2107 * Disable badblock log on clusters, or when
2108 * explicitly requested
2110 if (st
->nodes
> 0 || conf_get_create_info()->bblist
== 0) {
2112 sb
->bblog_offset
= 0;
2115 /* RAID0 needs a layout if devices aren't all the same size */
2116 if (raid0_need_layout
)
2117 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
2119 sb
->sb_csum
= calc_sb_1_csum(sb
);
2120 rv
= store_super1(st
, di
->fd
);
2122 if (rv
== 0 && (di
->disk
.state
& (1 << MD_DISK_JOURNAL
))) {
2123 rv
= write_empty_r5l_meta_block(st
, di
->fd
);
2129 (__le32_to_cpu(sb
->feature_map
) &
2130 MD_FEATURE_BITMAP_OFFSET
)) {
2131 rv
= st
->ss
->write_bitmap(st
, di
->fd
, NodeNumUpdate
);
2132 } else if (rv
== 0 &&
2133 md_feature_any_ppl_on(sb
->feature_map
)) {
2136 st
->ss
->getinfo_super(st
, &info
, NULL
);
2137 rv
= st
->ss
->write_init_ppl(st
, &info
, di
->fd
);
2147 pr_err("Failed to write metadata to %s\n", di
->devname
);
2152 static int compare_super1(struct supertype
*st
, struct supertype
*tst
,
2157 * 0 same, or first was empty, and second was copied
2158 * 1 second had wrong number
2160 * 3 wrong other info
2162 struct mdp_superblock_1
*first
= st
->sb
;
2163 struct mdp_superblock_1
*second
= tst
->sb
;
2165 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
2167 if (second
->major_version
!= __cpu_to_le32(1))
2171 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
2172 pr_err("could not allocate superblock\n");
2175 memcpy(first
, second
, SUPER1_SIZE
);
2179 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
2182 if (first
->ctime
!= second
->ctime
||
2183 first
->level
!= second
->level
||
2184 first
->layout
!= second
->layout
||
2185 first
->size
!= second
->size
||
2186 first
->chunksize
!= second
->chunksize
||
2187 first
->raid_disks
!= second
->raid_disks
)
2192 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
2194 unsigned long long dsize
;
2195 unsigned long long sb_offset
;
2196 struct mdp_superblock_1
*super
;
2198 struct bitmap_super_s
*bsb
;
2199 struct misc_dev_info
*misc
;
2200 struct align_fd afd
;
2206 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
2208 struct supertype tst
;
2209 __u64 bestctime
= 0;
2210 /* guess... choose latest ctime */
2211 memset(&tst
, 0, sizeof(tst
));
2213 for (tst
.minor_version
= 0; tst
.minor_version
<= 2;
2214 tst
.minor_version
++) {
2215 tst
.ignore_hw_compat
= st
->ignore_hw_compat
;
2216 switch(load_super1(&tst
, fd
, devname
)) {
2217 case 0: super
= tst
.sb
;
2218 if (bestvers
== -1 ||
2219 bestctime
< __le64_to_cpu(super
->ctime
)) {
2220 bestvers
= tst
.minor_version
;
2221 bestctime
= __le64_to_cpu(super
->ctime
);
2226 case 1: return 1; /*bad device */
2227 case 2: break; /* bad, try next */
2230 if (bestvers
!= -1) {
2232 tst
.minor_version
= bestvers
;
2234 tst
.max_devs
= MAX_DEVS
;
2235 rv
= load_super1(&tst
, fd
, devname
);
2242 if (!get_dev_size(fd
, devname
, &dsize
))
2248 pr_err("%s is too small for md: size is %llu sectors.\n",
2254 * Calculate the position of the superblock.
2255 * It is always aligned to a 4K boundary and
2256 * depending on minor_version, it can be:
2257 * 0: At least 8K, but less than 12K, from end of device
2258 * 1: At start of device
2259 * 2: 4K from start of device.
2261 switch(st
->minor_version
) {
2265 sb_offset
&= ~(4*2-1);
2277 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
2279 pr_err("Cannot seek to superblock on %s: %s\n",
2280 devname
, strerror(errno
));
2284 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
2285 pr_err("could not allocate superblock\n");
2289 memset(super
, 0, SUPER1_SIZE
);
2291 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
2293 pr_err("Cannot read superblock on %s\n",
2299 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
2301 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
2302 devname
, MD_SB_MAGIC
,
2303 __le32_to_cpu(super
->magic
));
2308 if (__le32_to_cpu(super
->major_version
) != 1) {
2310 pr_err("Cannot interpret superblock on %s - version is %d\n",
2311 devname
, __le32_to_cpu(super
->major_version
));
2315 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
2317 pr_err("No superblock found on %s (super_offset is wrong)\n",
2323 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
2325 misc
= (struct misc_dev_info
*)
2326 (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
2327 misc
->device_size
= dsize
;
2328 if (st
->data_offset
== INVALID_SECTORS
)
2329 st
->data_offset
= __le64_to_cpu(super
->data_offset
);
2331 if (st
->minor_version
>= 1 &&
2332 st
->ignore_hw_compat
== 0 &&
2333 (dsize
< (__le64_to_cpu(super
->data_offset
) +
2334 __le64_to_cpu(super
->size
))
2336 dsize
< (__le64_to_cpu(super
->data_offset
) +
2337 __le64_to_cpu(super
->data_size
)))) {
2339 pr_err("Device %s is not large enough for data described in superblock\n",
2346 /* Now check on the bitmap superblock */
2347 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
2349 /* Read the bitmap superblock and make sure it looks
2350 * valid. If it doesn't clear the bit. An --assemble --force
2351 * should get that written out.
2353 locate_bitmap1(st
, fd
, 0);
2354 if (aread(&afd
, bsb
, 512) != 512)
2357 uuid_from_super1(st
, uuid
);
2358 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
2359 memcmp(bsb
->uuid
, uuid
, 16) != 0)
2364 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
2365 & ~MD_FEATURE_BITMAP_OFFSET
);
2369 static struct supertype
*match_metadata_desc1(char *arg
)
2371 struct supertype
*st
= xcalloc(1, sizeof(*st
));
2373 st
->container_devnm
[0] = 0;
2375 st
->max_devs
= MAX_DEVS
;
2377 st
->data_offset
= INVALID_SECTORS
;
2378 /* leading zeros can be safely ignored. --detail generates them. */
2381 if (strcmp(arg
, "1.0") == 0 || strcmp(arg
, "1.00") == 0) {
2382 st
->minor_version
= 0;
2385 if (strcmp(arg
, "1.1") == 0 || strcmp(arg
, "1.01") == 0
2387 st
->minor_version
= 1;
2390 if (strcmp(arg
, "1.2") == 0 ||
2391 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
2392 strcmp(arg
, "default") == 0 ||
2393 #endif /* DEFAULT_OLD_METADATA */
2394 strcmp(arg
, "1.02") == 0) {
2395 st
->minor_version
= 2;
2398 if (strcmp(arg
, "1") == 0 || strcmp(arg
, "default") == 0) {
2399 st
->minor_version
= -1;
2407 /* find available size on device with this devsize, using
2408 * superblock type st, and reserving 'reserve' sectors for
2411 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
,
2412 unsigned long long data_offset
)
2414 struct mdp_superblock_1
*super
= st
->sb
;
2420 if (__le32_to_cpu(super
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
) {
2421 /* hot-add. allow for actual size of bitmap */
2422 struct bitmap_super_s
*bsb
;
2423 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
2424 bmspace
= calc_bitmap_size(bsb
, 4096) >> 9;
2425 } else if (md_feature_any_ppl_on(super
->feature_map
)) {
2426 bmspace
= __le16_to_cpu(super
->ppl
.size
);
2429 /* Allow space for bad block log */
2430 if (super
->bblog_size
)
2431 bbspace
= __le16_to_cpu(super
->bblog_size
);
2433 if (st
->minor_version
< 0)
2434 /* not specified, so time to set default */
2435 st
->minor_version
= 2;
2437 if (data_offset
== INVALID_SECTORS
)
2438 data_offset
= st
->data_offset
;
2440 if (data_offset
!= INVALID_SECTORS
)
2441 switch(st
->minor_version
) {
2443 return devsize
- data_offset
- 8*2 - bbspace
;
2446 return devsize
- data_offset
;
2453 switch(st
->minor_version
) {
2456 return ((devsize
- 8*2 - bbspace
) & ~(4*2-1));
2458 /* at start, 4K for superblock and possible bitmap */
2459 return devsize
- 4*2 - bbspace
;
2461 /* 4k from start, 4K for superblock and possible bitmap */
2462 return devsize
- (4+4)*2 - bbspace
;
2468 add_internal_bitmap1(struct supertype
*st
,
2469 int *chunkp
, int delay
, int write_behind
,
2470 unsigned long long size
,
2471 int may_change
, int major
)
2474 * If not may_change, then this is a 'Grow' without sysfs support for
2475 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
2476 * If may_change, then this is create or a Grow with sysfs support,
2477 * and we can put the bitmap wherever we like.
2479 * size is in sectors, chunk is in bytes !!!
2482 unsigned long long bits
;
2483 unsigned long long max_bits
;
2484 unsigned long long min_chunk
;
2486 long bbl_offset
, bbl_size
;
2487 unsigned long long chunk
= *chunkp
;
2491 struct mdp_superblock_1
*sb
= st
->sb
;
2492 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
2495 if (__le64_to_cpu(sb
->data_size
) == 0)
2497 * Must be creating the array, else data_size
2501 switch(st
->minor_version
) {
2504 * either 3K after the superblock (when hot-add),
2505 * or some amount of space before.
2509 * We are creating array, so we *know* how much room has
2515 choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2517 room
= __le64_to_cpu(sb
->super_offset
)
2518 - __le64_to_cpu(sb
->data_offset
)
2519 - __le64_to_cpu(sb
->data_size
);
2520 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2523 bbl_offset
= (__s32
)__le32_to_cpu(sb
->bblog_offset
);
2524 if (bbl_size
< -bbl_offset
)
2525 bbl_size
= -bbl_offset
;
2528 (room
< 3*2 && __le32_to_cpu(sb
->max_dev
) <= 384)) {
2533 offset
= 0; /* means movable offset */
2538 case 2: /* between superblock and data */
2543 choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2545 room
= __le64_to_cpu(sb
->data_offset
)
2546 - __le64_to_cpu(sb
->super_offset
);
2547 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2550 __le32_to_cpu(sb
->bblog_offset
) + bbl_size
;
2555 room
-= 2; /* Leave 1K for superblock */
2559 room
-= 4*2; /* leave 4K for superblock */
2569 if (chunk
== UnSet
&& room
> 128*2)
2570 /* Limit to 128K of bitmap when chunk size not requested */
2574 /* No room for a bitmap */
2577 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
2579 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
2580 bits
= (size
*512)/min_chunk
+1;
2581 while (bits
> max_bits
) {
2585 if (chunk
== UnSet
) {
2586 /* For practical purpose, 64Meg is a good
2587 * default chunk size for internal bitmaps.
2590 if (chunk
< 64*1024*1024)
2591 chunk
= 64*1024*1024;
2592 } else if (chunk
< min_chunk
)
2593 return -EINVAL
; /* chunk size too small */
2594 if (chunk
== 0) /* rounding problem */
2598 /* start bitmap on a 4K boundary with enough space for
2601 bits
= (size
*512) / chunk
+ 1;
2602 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
2603 room
*= 8; /* convert 4K blocks to sectors */
2604 offset
= -room
- bbl_size
;
2607 sb
->bitmap_offset
= (int32_t)__cpu_to_le32(offset
);
2609 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) |
2610 MD_FEATURE_BITMAP_OFFSET
);
2611 memset(bms
, 0, sizeof(*bms
));
2612 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
2613 bms
->version
= __cpu_to_le32(major
);
2614 uuid_from_super1(st
, uuid
);
2615 memcpy(bms
->uuid
, uuid
, 16);
2616 bms
->chunksize
= __cpu_to_le32(chunk
);
2617 bms
->daemon_sleep
= __cpu_to_le32(delay
);
2618 bms
->sync_size
= __cpu_to_le64(size
);
2619 bms
->write_behind
= __cpu_to_le32(write_behind
);
2620 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2622 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) |
2623 MD_FEATURE_BITMAP_VERSIONED
);
2624 if (st
->cluster_name
) {
2625 len
= sizeof(bms
->cluster_name
);
2626 strncpy((char *)bms
->cluster_name
, st
->cluster_name
, len
);
2627 bms
->cluster_name
[len
- 1] = '\0';
2634 static int locate_bitmap1(struct supertype
*st
, int fd
, int node_num
)
2636 unsigned long long offset
, bm_sectors_per_node
;
2637 struct mdp_superblock_1
*sb
;
2638 bitmap_super_t
*bms
;
2643 if (st
->ss
->load_super(st
, fd
, NULL
))
2644 return -1; /* no error I hope... */
2649 if ((__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
))
2654 offset
= __le64_to_cpu(sb
->super_offset
) + (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
2656 bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
2657 bm_sectors_per_node
= calc_bitmap_size(bms
, 4096) >> 9;
2658 offset
+= bm_sectors_per_node
* node_num
;
2662 lseek64(fd
, offset
<<9, 0);
2666 static int write_bitmap1(struct supertype
*st
, int fd
, enum bitmap_update update
)
2668 struct mdp_superblock_1
*sb
= st
->sb
;
2669 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
2672 int towrite
, n
, len
;
2673 struct align_fd afd
;
2675 unsigned long long total_bm_space
, bm_space_per_node
;
2679 /* update cluster name */
2680 if (st
->cluster_name
) {
2681 len
= sizeof(bms
->cluster_name
);
2682 memset((char *)bms
->cluster_name
, 0, len
);
2683 strncpy((char *)bms
->cluster_name
,
2684 st
->cluster_name
, len
);
2685 bms
->cluster_name
[len
- 1] = '\0';
2689 /* cluster md only supports superblock 1.2 now */
2690 if (st
->minor_version
!= 2 &&
2691 bms
->version
== BITMAP_MAJOR_CLUSTERED
) {
2692 pr_err("Warning: cluster md only works with superblock 1.2\n");
2696 if (bms
->version
== BITMAP_MAJOR_CLUSTERED
) {
2697 if (st
->nodes
== 1) {
2698 /* the parameter for nodes is not valid */
2699 pr_err("Warning: cluster-md at least needs two nodes\n");
2701 } else if (st
->nodes
== 0) {
2703 * parameter "--nodes" is not specified, (eg, add a disk to
2707 } else if (__cpu_to_le32(st
->nodes
) < bms
->nodes
) {
2709 * Since the nodes num is not increased, no
2710 * need to check the space enough or not,
2711 * just update bms->nodes
2713 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2718 * no need to change bms->nodes for other
2722 pr_err("Warning: --nodes option is only suitable for clustered bitmap\n");
2727 * Each node has an independent bitmap, it is necessary to
2728 * calculate the space is enough or not, first get how many
2729 * bytes for the total bitmap
2731 bm_space_per_node
= calc_bitmap_size(bms
, 4096);
2733 total_bm_space
= 512 * (__le64_to_cpu(sb
->data_offset
) -
2734 __le64_to_cpu(sb
->super_offset
));
2735 /* leave another 4k for superblock */
2736 total_bm_space
= total_bm_space
- 4096;
2738 if (bm_space_per_node
* st
->nodes
> total_bm_space
) {
2739 pr_err("Warning: The max num of nodes can't exceed %llu\n",
2740 total_bm_space
/ bm_space_per_node
);
2744 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2753 if (locate_bitmap1(st
, fd
, 0) < 0) {
2754 pr_err("Error: Invalid bitmap\n");
2758 if (posix_memalign(&buf
, 4096, 4096))
2762 /* Only the bitmap[0] should resync
2763 * whole device on initial assembly
2766 memset(buf
, 0x00, 4096);
2768 memset(buf
, 0xff, 4096);
2769 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
2772 * use 4096 boundary if bitmap_offset is aligned
2773 * with 8 sectors, then it should compatible with
2776 if (__le32_to_cpu(sb
->bitmap_offset
) & 7)
2777 towrite
= calc_bitmap_size(bms
, 512);
2779 towrite
= calc_bitmap_size(bms
, 4096);
2780 while (towrite
> 0) {
2784 n
= awrite(&afd
, buf
, n
);
2790 memset(buf
, 0x00, 4096);
2792 memset(buf
, 0xff, 4096);
2799 } while (++i
< __le32_to_cpu(bms
->nodes
));
2805 static void free_super1(struct supertype
*st
)
2811 struct devinfo
*di
= st
->info
;
2812 st
->info
= di
->next
;
2820 static int validate_geometry1(struct supertype
*st
, int level
,
2821 int layout
, int raiddisks
,
2822 int *chunk
, unsigned long long size
,
2823 unsigned long long data_offset
,
2824 char *subdev
, unsigned long long *freesize
,
2825 int consistency_policy
, int verbose
)
2827 unsigned long long ldsize
, devsize
;
2829 unsigned long long headroom
;
2830 unsigned long long overhead
;
2833 if (level
== LEVEL_CONTAINER
) {
2835 pr_err("1.x metadata does not support containers\n");
2838 if (*chunk
== UnSet
)
2839 *chunk
= DEFAULT_CHUNK
;
2844 if (st
->minor_version
< 0)
2845 /* not specified, so time to set default */
2846 st
->minor_version
= 2;
2848 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
2851 pr_err("super1.x cannot open %s: %s\n",
2852 subdev
, strerror(errno
));
2856 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
2862 devsize
= ldsize
>> 9;
2864 /* creating: allow suitable space for bitmap or PPL */
2865 if (consistency_policy
== CONSISTENCY_POLICY_PPL
)
2866 bmspace
= MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9;
2868 bmspace
= choose_bm_space(devsize
);
2870 if (data_offset
== INVALID_SECTORS
)
2871 data_offset
= st
->data_offset
;
2872 if (data_offset
== INVALID_SECTORS
)
2873 switch (st
->minor_version
) {
2879 /* Choose data offset appropriate for this device
2880 * and use as default for whole array.
2881 * The data_offset must allow for bitmap space
2882 * and base metadata, should allow for some headroom
2883 * for reshape, and should be rounded to multiple
2885 * Headroom is limited to 128M, but aim for about 0.1%
2887 headroom
= 128*1024*2;
2888 while ((headroom
<< 10) > devsize
&&
2890 headroom
/ 2 >= ((unsigned)(*chunk
)*2)*2))
2892 data_offset
= 12*2 + bmspace
+ headroom
;
2893 #define ONE_MEG (2*1024)
2894 data_offset
= ROUND_UP(data_offset
, ONE_MEG
);
2897 if (st
->data_offset
== INVALID_SECTORS
)
2898 st
->data_offset
= data_offset
;
2899 switch(st
->minor_version
) {
2900 case 0: /* metadata at end. Round down and subtract space to reserve */
2901 devsize
= (devsize
& ~(4ULL*2-1));
2902 /* space for metadata, bblog, bitmap/ppl */
2903 overhead
= 8*2 + 8 + bmspace
;
2904 if (devsize
< overhead
) /* detect underflow */
2905 goto dev_too_small_err
;
2906 devsize
-= overhead
;
2910 if (devsize
< data_offset
) /* detect underflow */
2911 goto dev_too_small_err
;
2912 devsize
-= data_offset
;
2915 *freesize
= devsize
;
2918 /* Error condition, device cannot even hold the overhead. */
2920 fprintf(stderr
, "device %s is too small (%lluK) for "
2921 "required metadata!\n", subdev
, devsize
>>1);
2926 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
)
2928 /* Create a v1.0 superblock based on 'info'*/
2930 struct mdp_superblock_1
*sb
;
2932 unsigned long long offset
;
2934 if (posix_memalign(&ret
, 4096, 1024) != 0)
2937 memset(ret
, 0, 1024);
2938 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
2939 sb
->major_version
= __cpu_to_le32(1);
2941 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
2942 sprintf(sb
->set_name
, "%d", sb0
->md_minor
);
2943 sb
->ctime
= __cpu_to_le32(info
->array
.ctime
+1);
2944 sb
->level
= __cpu_to_le32(info
->array
.level
);
2945 sb
->layout
= __cpu_to_le32(info
->array
.layout
);
2946 sb
->size
= __cpu_to_le64(info
->component_size
);
2947 sb
->chunksize
= __cpu_to_le32(info
->array
.chunk_size
/512);
2948 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
2949 if (info
->array
.level
> 0)
2950 sb
->data_size
= sb
->size
;
2952 sb
->data_size
= st
->ss
->avail_size(st
, st
->devsize
/512, 0);
2953 sb
->resync_offset
= MaxSector
;
2954 sb
->max_dev
= __cpu_to_le32(MD_SB_DISKS
);
2955 sb
->dev_number
= __cpu_to_le32(info
->disk
.number
);
2956 sb
->utime
= __cpu_to_le64(info
->array
.utime
);
2958 offset
= st
->devsize
/512 - 8*2;
2960 sb
->super_offset
= __cpu_to_le64(offset
);
2961 //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset);
2963 random_uuid(sb
->device_uuid
);
2965 for (i
= 0; i
< MD_SB_DISKS
; i
++) {
2966 int state
= sb0
->disks
[i
].state
;
2967 sb
->dev_roles
[i
] = MD_DISK_ROLE_SPARE
;
2968 if ((state
& (1<<MD_DISK_SYNC
)) &&
2969 !(state
& (1<<MD_DISK_FAULTY
)))
2970 sb
->dev_roles
[i
] = __cpu_to_le16(sb0
->disks
[i
].raid_disk
);
2972 sb
->sb_csum
= calc_sb_1_csum(sb
);
2976 struct superswitch super1
= {
2977 .examine_super
= examine_super1
,
2978 .brief_examine_super
= brief_examine_super1
,
2979 .export_examine_super
= export_examine_super1
,
2980 .detail_super
= detail_super1
,
2981 .brief_detail_super
= brief_detail_super1
,
2982 .export_detail_super
= export_detail_super1
,
2983 .write_init_super
= write_init_super1
,
2984 .validate_geometry
= validate_geometry1
,
2985 .add_to_super
= add_to_super1
,
2986 .examine_badblocks
= examine_badblocks_super1
,
2987 .copy_metadata
= copy_metadata1
,
2988 .write_init_ppl
= write_init_ppl1
,
2989 .match_home
= match_home1
,
2990 .uuid_from_super
= uuid_from_super1
,
2991 .getinfo_super
= getinfo_super1
,
2992 .container_content
= container_content1
,
2993 .update_super
= update_super1
,
2994 .init_super
= init_super1
,
2995 .store_super
= store_super1
,
2996 .compare_super
= compare_super1
,
2997 .load_super
= load_super1
,
2998 .match_metadata_desc
= match_metadata_desc1
,
2999 .avail_size
= avail_size1
,
3000 .add_internal_bitmap
= add_internal_bitmap1
,
3001 .locate_bitmap
= locate_bitmap1
,
3002 .write_bitmap
= write_bitmap1
,
3003 .free_super
= free_super1
,
3004 #if __BYTE_ORDER == BIG_ENDIAN