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, after=%llu sectors\n",
411 info
.space_before
, info
.space_after
);
413 printf(" State : %s\n",
414 (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
415 printf(" Device UUID : ");
416 for (i
=0; i
<16; i
++) {
417 if ((i
&3)==0 && i
!= 0)
419 printf("%02x", sb
->device_uuid
[i
]);
423 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
424 printf("Internal Bitmap : %ld sectors from superblock\n",
425 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
426 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
427 printf(" PPL : %u sectors at offset %d sectors from superblock\n",
428 __le16_to_cpu(sb
->ppl
.size
),
429 __le16_to_cpu(sb
->ppl
.offset
));
431 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)) {
432 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)
433 __le64_to_cpu(sb
->reshape_position
)/2,
434 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
435 if (__le32_to_cpu(sb
->delta_disks
)) {
436 printf(" Delta Devices : %d",
437 __le32_to_cpu(sb
->delta_disks
));
438 printf(" (%d->%d)\n",
439 __le32_to_cpu(sb
->raid_disks
) -
440 __le32_to_cpu(sb
->delta_disks
),
441 __le32_to_cpu(sb
->raid_disks
));
442 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
443 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
445 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
446 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
447 printf(" New Level : %s\n", c
?c
:"-unknown-");
449 if (__le32_to_cpu(sb
->new_layout
) !=
450 __le32_to_cpu(sb
->layout
)) {
451 if (__le32_to_cpu(sb
->level
) == 5) {
452 c
= map_num(r5layout
,
453 __le32_to_cpu(sb
->new_layout
));
454 printf(" New Layout : %s\n", c
?c
:"-unknown-");
456 if (__le32_to_cpu(sb
->level
) == 6) {
457 c
= map_num(r6layout
,
458 __le32_to_cpu(sb
->new_layout
));
459 printf(" New Layout : %s\n", c
?c
:"-unknown-");
461 if (__le32_to_cpu(sb
->level
) == 10) {
462 printf(" New Layout :");
463 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
467 if (__le32_to_cpu(sb
->new_chunk
) !=
468 __le32_to_cpu(sb
->chunksize
))
469 printf(" New Chunksize : %dK\n",
470 __le32_to_cpu(sb
->new_chunk
)/2);
475 if (sb
->devflags
& WriteMostly1
)
476 printf(" write-mostly");
477 if (sb
->devflags
& FailFast1
)
482 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
483 printf(" Update Time : %.24s\n", ctime(&atime
));
485 if (sb
->bblog_size
&& sb
->bblog_offset
) {
486 printf(" Bad Block Log : %d entries available at offset %ld sectors",
487 __le16_to_cpu(sb
->bblog_size
)*512/8,
488 (long)(int32_t)__le32_to_cpu(sb
->bblog_offset
));
489 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
490 printf(" - bad blocks present.");
494 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
495 printf(" Checksum : %x - correct\n",
496 __le32_to_cpu(sb
->sb_csum
));
498 printf(" Checksum : %x - expected %x\n",
499 __le32_to_cpu(sb
->sb_csum
),
500 __le32_to_cpu(calc_sb_1_csum(sb
)));
501 printf(" Events : %llu\n",
502 (unsigned long long)__le64_to_cpu(sb
->events
));
504 if (__le32_to_cpu(sb
->level
) == 0 &&
505 (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
))) {
506 c
= map_num(r0layout
, __le32_to_cpu(sb
->layout
));
507 printf(" Layout : %s\n", c
?c
:"-unknown-");
509 if (__le32_to_cpu(sb
->level
) == 5) {
510 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
511 printf(" Layout : %s\n", c
?c
:"-unknown-");
513 if (__le32_to_cpu(sb
->level
) == 6) {
514 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
515 printf(" Layout : %s\n", c
?c
:"-unknown-");
517 if (__le32_to_cpu(sb
->level
) == 10) {
518 int lo
= __le32_to_cpu(sb
->layout
);
520 print_r10_layout(lo
);
523 switch(__le32_to_cpu(sb
->level
)) {
529 printf(" Chunk Size : %dK\n",
530 __le32_to_cpu(sb
->chunksize
)/2);
533 printf(" Rounding : %dK\n",
534 __le32_to_cpu(sb
->chunksize
)/2);
541 /* This turns out to just be confusing */
542 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
543 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
544 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != MD_DISK_ROLE_SPARE
)
546 for (d
= 0; d
< i
; d
++) {
547 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
550 if (role
== MD_DISK_ROLE_SPARE
)
553 if(role
== MD_DISK_ROLE_FAULTY
)
560 printf(" Device Role : ");
561 role
= role_from_sb(sb
);
562 if (role
>= MD_DISK_ROLE_FAULTY
)
564 else if (role
== MD_DISK_ROLE_JOURNAL
)
566 else if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_REPLACEMENT
))
567 printf("Replacement device %d\n", role
);
569 printf("Active device %d\n", role
);
571 printf(" Array State : ");
572 for (d
= 0; d
< __le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
575 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
576 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
580 if (cnt
== 2 && __le32_to_cpu(sb
->level
) > 0)
592 /* This is confusing too */
594 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
595 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
596 if (role
== MD_DISK_ROLE_FAULTY
)
600 printf(" %d failed", faulty
);
602 printf(" ('A' == active, '.' == missing, 'R' == replacing)");
604 for (d
= 0; d
< __le32_to_cpu(sb
->max_dev
); d
++) {
605 unsigned int r
= __le16_to_cpu(sb
->dev_roles
[d
]);
606 if (r
<= MD_DISK_ROLE_MAX
&&
607 r
> __le32_to_cpu(sb
->raid_disks
) + delta_extra
)
611 printf("WARNING Array state is inconsistent - each number should appear only once\n");
612 for (d
= 0; d
< __le32_to_cpu(sb
->max_dev
); d
++)
613 if (__le16_to_cpu(sb
->dev_roles
[d
]) >= MD_DISK_ROLE_FAULTY
)
616 printf(" %d:%d", d
, __le16_to_cpu(sb
->dev_roles
[d
]));
621 static void brief_examine_super1(struct supertype
*st
, int verbose
)
623 struct mdp_superblock_1
*sb
= st
->sb
;
625 unsigned long long sb_offset
;
627 char *c
= map_num(pers
, __le32_to_cpu(sb
->level
));
629 nm
= strchr(sb
->set_name
, ':');
632 else if (sb
->set_name
[0])
644 printf(" level=%s", c
);
645 sb_offset
= __le64_to_cpu(sb
->super_offset
);
647 printf(" metadata=1.1 ");
648 else if (sb_offset
<= 8)
649 printf(" metadata=1.2 ");
651 printf(" metadata=1.0 ");
653 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
655 for (i
= 0; i
< 16; i
++) {
656 if ((i
&3)==0 && i
!= 0)
658 printf("%02x", sb
->set_uuid
[i
]);
660 if (sb
->set_name
[0]) {
662 print_quoted(sb
->set_name
);
667 static void export_examine_super1(struct supertype
*st
)
669 struct mdp_superblock_1
*sb
= st
->sb
;
674 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
675 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
676 for (i
= 0; i
< 32; i
++)
677 if (sb
->set_name
[i
] == '\n' || sb
->set_name
[i
] == '\0') {
682 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
683 if (__le32_to_cpu(sb
->level
) > 0) {
684 int ddsks
= 0, ddsks_denom
= 1;
685 switch(__le32_to_cpu(sb
->level
)) {
691 ddsks
= __le32_to_cpu(sb
->raid_disks
)-1;
694 ddsks
= __le32_to_cpu(sb
->raid_disks
)-2;
697 layout
= __le32_to_cpu(sb
->layout
);
698 ddsks
= __le32_to_cpu(sb
->raid_disks
);
699 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
702 long long asize
= __le64_to_cpu(sb
->size
);
703 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
704 printf("MD_ARRAY_SIZE=%s\n",
705 human_size_brief(asize
, JEDEC
));
709 for (i
= 0; i
< 16; i
++) {
710 if ((i
&3) == 0 && i
!= 0)
712 printf("%02x", sb
->set_uuid
[i
]);
715 printf("MD_UPDATE_TIME=%llu\n",
716 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
717 printf("MD_DEV_UUID=");
718 for (i
= 0; i
< 16; i
++) {
719 if ((i
&3) == 0 && i
!= 0)
721 printf("%02x", sb
->device_uuid
[i
]);
724 printf("MD_EVENTS=%llu\n",
725 (unsigned long long)__le64_to_cpu(sb
->events
));
728 static int copy_metadata1(struct supertype
*st
, int from
, int to
)
730 /* Read superblock. If it looks good, write it out.
731 * Then if a bitmap is present, copy that.
732 * And if a bad-block-list is present, copy that too.
735 unsigned long long dsize
, sb_offset
;
736 const int bufsize
= 4*1024;
737 struct mdp_superblock_1 super
, *sb
;
739 if (posix_memalign(&buf
, 4096, bufsize
) != 0)
742 if (!get_dev_size(from
, NULL
, &dsize
))
748 switch(st
->minor_version
) {
752 sb_offset
&= ~(4*2-1);
764 if (lseek64(from
, sb_offset
<< 9, 0) < 0LL)
766 if (read(from
, buf
, bufsize
) != bufsize
)
770 super
= *sb
; // save most of sb for when we reuse buf
772 if (__le32_to_cpu(super
.magic
) != MD_SB_MAGIC
||
773 __le32_to_cpu(super
.major_version
) != 1 ||
774 __le64_to_cpu(super
.super_offset
) != sb_offset
||
775 calc_sb_1_csum(sb
) != super
.sb_csum
)
778 if (lseek64(to
, sb_offset
<< 9, 0) < 0LL)
780 if (write(to
, buf
, bufsize
) != bufsize
)
783 if (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
784 unsigned long long bitmap_offset
= sb_offset
;
785 int bytes
= 4096; // just an estimate.
787 struct align_fd afrom
, ato
;
789 init_afd(&afrom
, from
);
792 bitmap_offset
+= (int32_t)__le32_to_cpu(super
.bitmap_offset
);
794 if (lseek64(from
, bitmap_offset
<<9, 0) < 0)
796 if (lseek64(to
, bitmap_offset
<<9, 0) < 0)
799 for (written
= 0; written
< bytes
; ) {
800 int n
= bytes
- written
;
803 if (aread(&afrom
, buf
, n
) != n
)
806 /* have the header, can calculate
807 * correct bitmap bytes */
810 bytes
= calc_bitmap_size(bms
, 512);
814 if (awrite(&ato
, buf
, n
) != n
)
820 if (super
.bblog_size
!= 0 &&
821 __le16_to_cpu(super
.bblog_size
) <= 100 &&
822 super
.bblog_offset
!= 0 &&
823 (super
.feature_map
& __le32_to_cpu(MD_FEATURE_BAD_BLOCKS
))) {
824 /* There is a bad block log */
825 unsigned long long bb_offset
= sb_offset
;
826 int bytes
= __le16_to_cpu(super
.bblog_size
) * 512;
828 struct align_fd afrom
, ato
;
830 init_afd(&afrom
, from
);
833 bb_offset
+= (int32_t)__le32_to_cpu(super
.bblog_offset
);
835 if (lseek64(from
, bb_offset
<<9, 0) < 0)
837 if (lseek64(to
, bb_offset
<<9, 0) < 0)
840 for (written
= 0; written
< bytes
; ) {
841 int n
= bytes
- written
;
844 if (aread(&afrom
, buf
, n
) != n
)
847 if (awrite(&ato
, buf
, n
) != n
)
861 static void detail_super1(struct supertype
*st
, char *homehost
, char *subarray
)
863 struct mdp_superblock_1
*sb
= st
->sb
;
864 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
866 int l
= homehost
? strlen(homehost
) : 0;
868 printf(" Name : %.32s", sb
->set_name
);
869 if (l
> 0 && l
< 32 && sb
->set_name
[l
] == ':' &&
870 strncmp(sb
->set_name
, homehost
, l
) == 0)
871 printf(" (local to host %s)", homehost
);
872 if (bms
->nodes
> 0 &&
873 (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
))
874 printf("\n Cluster Name : %-64s", bms
->cluster_name
);
875 printf("\n UUID : ");
876 for (i
= 0; i
< 16; i
++) {
877 if ((i
&3) == 0 && i
!= 0)
879 printf("%02x", sb
->set_uuid
[i
]);
881 printf("\n Events : %llu\n\n",
882 (unsigned long long)__le64_to_cpu(sb
->events
));
885 static void brief_detail_super1(struct supertype
*st
, char *subarray
)
887 struct mdp_superblock_1
*sb
= st
->sb
;
890 if (sb
->set_name
[0]) {
892 print_quoted(sb
->set_name
);
895 for (i
= 0; i
< 16; i
++) {
896 if ((i
& 3) == 0 && i
!= 0)
898 printf("%02x", sb
->set_uuid
[i
]);
902 static void export_detail_super1(struct supertype
*st
)
904 struct mdp_superblock_1
*sb
= st
->sb
;
908 for (i
= 0; i
< 32; i
++)
909 if (sb
->set_name
[i
] == '\n' || sb
->set_name
[i
] == '\0') {
914 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
917 static int examine_badblocks_super1(struct supertype
*st
, int fd
, char *devname
)
919 struct mdp_superblock_1
*sb
= st
->sb
;
920 unsigned long long offset
;
925 if (!sb
->bblog_size
|| __le16_to_cpu(sb
->bblog_size
) > 100 ||
927 printf("No bad-blocks list configured on %s\n", devname
);
930 if ((sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
)) == 0) {
931 printf("Bad-blocks list is empty in %s\n", devname
);
935 size
= __le16_to_cpu(sb
->bblog_size
)* 512;
936 if (posix_memalign((void**)&bbl
, 4096, size
) != 0) {
937 pr_err("could not allocate badblocks list\n");
940 offset
= __le64_to_cpu(sb
->super_offset
) +
941 (int)__le32_to_cpu(sb
->bblog_offset
);
943 if (lseek64(fd
, offset
, 0) < 0) {
944 pr_err("Cannot seek to bad-blocks list\n");
947 if (read(fd
, bbl
, size
) != size
) {
948 pr_err("Cannot read bad-blocks list\n");
951 /* 64bits per entry. 10 bits is block-count, 54 bits is block
952 * offset. Blocks are sectors unless bblog->shift makes them bigger
955 printf("Bad-blocks on %s:\n", devname
);
956 for (i
= 0; i
< size
/8; i
++, bbp
++) {
957 __u64 bb
= __le64_to_cpu(*bbp
);
958 int count
= bb
& 0x3ff;
959 unsigned long long sector
= bb
>> 10;
964 sector
<<= sb
->bblog_shift
;
965 count
<<= sb
->bblog_shift
;
967 printf("%20llu for %d sectors\n", sector
, count
);
972 static int match_home1(struct supertype
*st
, char *homehost
)
974 struct mdp_superblock_1
*sb
= st
->sb
;
975 int l
= homehost
? strlen(homehost
) : 0;
977 return (l
> 0 && l
< 32 && sb
->set_name
[l
] == ':' &&
978 strncmp(sb
->set_name
, homehost
, l
) == 0);
981 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
983 struct mdp_superblock_1
*super
= st
->sb
;
984 char *cuuid
= (char*)uuid
;
986 for (i
= 0; i
< 16; i
++)
987 cuuid
[i
] = super
->set_uuid
[i
];
990 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
992 struct mdp_superblock_1
*sb
= st
->sb
;
993 struct bitmap_super_s
*bsb
= (void*)(((char*)sb
)+MAX_SB_SIZE
);
994 struct misc_dev_info
*misc
=
995 (void*)(((char*)sb
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
999 unsigned int map_disks
= info
->array
.raid_disks
;
1000 unsigned long long super_offset
;
1001 unsigned long long data_size
;
1003 memset(info
, 0, sizeof(*info
));
1004 info
->array
.major_version
= 1;
1005 info
->array
.minor_version
= st
->minor_version
;
1006 info
->array
.patch_version
= 0;
1007 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
1008 info
->array
.level
= __le32_to_cpu(sb
->level
);
1009 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
1010 info
->array
.md_minor
= -1;
1011 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
1012 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
1013 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
1015 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
) ? 1 : 0;
1017 super_offset
= __le64_to_cpu(sb
->super_offset
);
1018 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
1019 info
->component_size
= __le64_to_cpu(sb
->size
);
1020 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
1021 info
->bitmap_offset
= (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
1022 if (__le32_to_cpu(bsb
->nodes
) > 1)
1023 info
->array
.state
|= (1 << MD_SB_CLUSTERED
);
1024 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1025 info
->ppl_offset
= __le16_to_cpu(sb
->ppl
.offset
);
1026 info
->ppl_size
= __le16_to_cpu(sb
->ppl
.size
);
1027 info
->ppl_sector
= super_offset
+ info
->ppl_offset
;
1030 info
->disk
.major
= 0;
1031 info
->disk
.minor
= 0;
1032 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
1033 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
1034 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
1035 role
= MD_DISK_ROLE_FAULTY
;
1037 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
1039 if (info
->array
.level
<= 0)
1040 data_size
= __le64_to_cpu(sb
->data_size
);
1042 data_size
= __le64_to_cpu(sb
->size
);
1043 if (info
->data_offset
< super_offset
) {
1044 unsigned long long end
;
1045 info
->space_before
= info
->data_offset
;
1048 if (sb
->bblog_offset
&& sb
->bblog_size
) {
1049 unsigned long long bboffset
= super_offset
;
1050 bboffset
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
1055 if (super_offset
+ info
->bitmap_offset
+ info
->ppl_offset
< end
)
1056 end
= super_offset
+ info
->bitmap_offset
+
1059 if (info
->data_offset
+ data_size
< end
)
1060 info
->space_after
= end
- data_size
- info
->data_offset
;
1062 info
->space_after
= 0;
1064 unsigned long long earliest
;
1065 earliest
= super_offset
+ (32+4)*2; /* match kernel */
1066 if (info
->bitmap_offset
> 0) {
1067 unsigned long long bmend
= info
->bitmap_offset
;
1068 unsigned long long size
= calc_bitmap_size(bsb
, 4096);
1071 if (bmend
> earliest
)
1073 } else if (info
->ppl_offset
> 0) {
1074 unsigned long long pplend
;
1076 pplend
= info
->ppl_offset
+ info
->ppl_size
;
1077 if (pplend
> earliest
)
1080 if (sb
->bblog_offset
&& sb
->bblog_size
) {
1081 unsigned long long bbend
= super_offset
;
1082 bbend
+= (int32_t)__le32_to_cpu(sb
->bblog_offset
);
1083 bbend
+= __le16_to_cpu(sb
->bblog_size
);
1084 if (bbend
> earliest
)
1087 if (earliest
< info
->data_offset
)
1088 info
->space_before
= info
->data_offset
- earliest
;
1090 info
->space_before
= 0;
1091 info
->space_after
= misc
->device_size
- data_size
-
1094 if (info
->space_before
== 0 && info
->space_after
== 0) {
1095 /* It will look like we don't support data_offset changes,
1096 * be we do - it's just that there is no room.
1097 * A change that reduced the number of devices should
1098 * still be allowed, so set the otherwise useless value of '1'
1100 info
->space_after
= 1;
1103 info
->disk
.raid_disk
= -1;
1105 case MD_DISK_ROLE_SPARE
:
1106 /* spare: not active, not sync, not faulty */
1107 info
->disk
.state
= 0;
1109 case MD_DISK_ROLE_FAULTY
:
1110 info
->disk
.state
= (1 << MD_DISK_FAULTY
); /* faulty */
1112 case MD_DISK_ROLE_JOURNAL
:
1113 info
->disk
.state
= (1 << MD_DISK_JOURNAL
);
1114 info
->disk
.raid_disk
= role
;
1115 /* journal uses all 4kB blocks*/
1116 info
->space_after
= (misc
->device_size
- info
->data_offset
) % 8;
1119 info
->disk
.state
= 6; /* active and in sync */
1120 info
->disk
.raid_disk
= role
;
1122 if (sb
->devflags
& WriteMostly1
)
1123 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
1124 if (sb
->devflags
& FailFast1
)
1125 info
->disk
.state
|= (1 << MD_DISK_FAILFAST
);
1126 info
->events
= __le64_to_cpu(sb
->events
);
1127 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
1128 info
->safe_mode_delay
= 200;
1130 memcpy(info
->uuid
, sb
->set_uuid
, 16);
1132 strncpy(info
->name
, sb
->set_name
, 32);
1135 if ((__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_REPLACEMENT
)) {
1136 info
->disk
.state
&= ~(1 << MD_DISK_SYNC
);
1137 info
->disk
.state
|= 1 << MD_DISK_REPLACEMENT
;
1140 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
1141 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
1143 info
->recovery_start
= MaxSector
;
1145 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
1146 info
->reshape_active
= 1;
1147 if ((sb
->feature_map
& __le32_to_cpu(MD_FEATURE_NEW_OFFSET
)) &&
1148 sb
->new_offset
!= 0)
1149 info
->reshape_active
|= RESHAPE_NO_BACKUP
;
1150 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
1151 info
->new_level
= __le32_to_cpu(sb
->new_level
);
1152 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
1153 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
1154 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
1155 if (info
->delta_disks
< 0)
1156 info
->array
.raid_disks
-= info
->delta_disks
;
1158 info
->reshape_active
= 0;
1160 info
->recovery_blocked
= info
->reshape_active
;
1163 for (i
=0; i
<map_disks
; i
++)
1165 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
1166 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
1167 if (/*role == MD_DISK_ROLE_SPARE || */role
< (unsigned) info
->array
.raid_disks
) {
1169 if (map
&& role
< map_disks
)
1174 info
->array
.working_disks
= working
;
1176 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_JOURNAL
)) {
1177 info
->journal_device_required
= 1;
1178 info
->consistency_policy
= CONSISTENCY_POLICY_JOURNAL
;
1179 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1180 info
->consistency_policy
= CONSISTENCY_POLICY_PPL
;
1181 } else if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
)) {
1182 info
->consistency_policy
= CONSISTENCY_POLICY_BITMAP
;
1183 } else if (info
->array
.level
<= 0) {
1184 info
->consistency_policy
= CONSISTENCY_POLICY_NONE
;
1186 info
->consistency_policy
= CONSISTENCY_POLICY_RESYNC
;
1189 info
->journal_clean
= 0;
1192 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
1194 struct mdinfo
*info
;
1199 info
= xmalloc(sizeof(*info
));
1200 getinfo_super1(st
, info
, NULL
);
1204 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
1205 char *update
, char *devname
, int verbose
,
1206 int uuid_set
, char *homehost
)
1208 /* NOTE: for 'assemble' and 'force' we need to return non-zero
1209 * if any change was made. For others, the return value is
1213 struct mdp_superblock_1
*sb
= st
->sb
;
1214 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
1216 if (strcmp(update
, "homehost") == 0 &&
1218 /* Note that 'homehost' is special as it is really
1223 c
= strchr(sb
->set_name
, ':');
1225 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
1227 strncpy(info
->name
, sb
->set_name
, 32);
1231 if (strcmp(update
, "force-one")==0) {
1232 /* Not enough devices for a working array,
1233 * so bring this one up-to-date
1235 if (sb
->events
!= __cpu_to_le64(info
->events
))
1237 sb
->events
= __cpu_to_le64(info
->events
);
1238 } else if (strcmp(update
, "force-array")==0) {
1239 /* Degraded array and 'force' requests to
1240 * maybe need to mark it 'clean'.
1242 switch(__le32_to_cpu(sb
->level
)) {
1246 /* need to force clean */
1247 if (sb
->resync_offset
!= MaxSector
)
1249 sb
->resync_offset
= MaxSector
;
1251 } else if (strcmp(update
, "assemble")==0) {
1252 int d
= info
->disk
.number
;
1254 if (info
->disk
.state
& (1<<MD_DISK_ACTIVE
))
1255 want
= info
->disk
.raid_disk
;
1256 else if (info
->disk
.state
& (1<<MD_DISK_JOURNAL
))
1257 want
= MD_DISK_ROLE_JOURNAL
;
1259 want
= MD_DISK_ROLE_SPARE
;
1260 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
1261 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
1264 if (info
->reshape_active
&&
1266 __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1267 info
->delta_disks
>= 0 &&
1268 info
->reshape_progress
<
1269 __le64_to_cpu(sb
->reshape_position
)) {
1270 sb
->reshape_position
=
1271 __cpu_to_le64(info
->reshape_progress
);
1274 if (info
->reshape_active
&&
1276 __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
1277 info
->delta_disks
< 0 &&
1278 info
->reshape_progress
>
1279 __le64_to_cpu(sb
->reshape_position
)) {
1280 sb
->reshape_position
=
1281 __cpu_to_le64(info
->reshape_progress
);
1284 } else if (strcmp(update
, "linear-grow-new") == 0) {
1287 int max
= __le32_to_cpu(sb
->max_dev
);
1292 for (i
= 0; i
< max
; i
++)
1293 if (__le16_to_cpu(sb
->dev_roles
[i
]) >=
1294 MD_DISK_ROLE_FAULTY
)
1296 if (i
!= info
->disk
.number
)
1298 sb
->dev_number
= __cpu_to_le32(i
);
1301 sb
->max_dev
= __cpu_to_le32(max
+1);
1305 random_uuid(sb
->device_uuid
);
1307 sb
->dev_roles
[i
] = __cpu_to_le16(info
->disk
.raid_disk
);
1309 fd
= open(devname
, O_RDONLY
);
1311 unsigned long long ds
;
1312 get_dev_size(fd
, devname
, &ds
);
1315 if (__le64_to_cpu(sb
->super_offset
) <
1316 __le64_to_cpu(sb
->data_offset
)) {
1317 sb
->data_size
= __cpu_to_le64(
1318 ds
- __le64_to_cpu(sb
->data_offset
));
1321 ds
&= ~(unsigned long long)(4*2-1);
1322 sb
->super_offset
= __cpu_to_le64(ds
);
1323 sb
->data_size
= __cpu_to_le64(
1324 ds
- __le64_to_cpu(sb
->data_offset
));
1327 } else if (strcmp(update
, "linear-grow-update") == 0) {
1328 int max
= __le32_to_cpu(sb
->max_dev
);
1329 int i
= info
->disk
.number
;
1330 if (max
> MAX_DEVS
|| i
> MAX_DEVS
)
1335 sb
->max_dev
= __cpu_to_le32(max
+1);
1336 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
1337 sb
->dev_roles
[info
->disk
.number
] =
1338 __cpu_to_le16(info
->disk
.raid_disk
);
1339 } else if (strcmp(update
, "resync") == 0) {
1340 /* make sure resync happens */
1341 sb
->resync_offset
= 0ULL;
1342 } else if (strcmp(update
, "uuid") == 0) {
1343 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
1345 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
)
1346 memcpy(bms
->uuid
, sb
->set_uuid
, 16);
1347 } else if (strcmp(update
, "no-bitmap") == 0) {
1348 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
1349 if (bms
->version
== BITMAP_MAJOR_CLUSTERED
&& !IsBitmapDirty(devname
))
1350 sb
->resync_offset
= MaxSector
;
1351 } else if (strcmp(update
, "bbl") == 0) {
1352 /* only possible if there is room after the bitmap, or if
1353 * there is no bitmap
1355 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
1356 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
1357 long bitmap_offset
= 0;
1358 long bm_sectors
= 0;
1361 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1362 bitmap_offset
= (long)__le32_to_cpu(sb
->bitmap_offset
);
1363 bm_sectors
= calc_bitmap_size(bms
, 4096) >> 9;
1364 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
1365 bitmap_offset
= (long)__le16_to_cpu(sb
->ppl
.offset
);
1366 bm_sectors
= (long)__le16_to_cpu(sb
->ppl
.size
);
1369 if (sb_offset
< data_offset
) {
1371 * 1.1 or 1.2. Put bbl after bitmap leaving
1375 bb_offset
= sb_offset
+ 8;
1376 if (bm_sectors
&& bitmap_offset
> 0)
1377 bb_offset
= bitmap_offset
+ bm_sectors
;
1378 while (bb_offset
< (long)sb_offset
+ 8 + 32*2 &&
1379 bb_offset
+ 8+8 <= (long)data_offset
)
1380 /* too close to bitmap, and room to grow */
1382 if (bb_offset
+ 8 <= (long)data_offset
) {
1383 sb
->bblog_size
= __cpu_to_le16(8);
1384 sb
->bblog_offset
= __cpu_to_le32(bb_offset
);
1387 /* 1.0 - Put bbl just before super block */
1388 if (bm_sectors
&& bitmap_offset
< 0)
1389 space
= -bitmap_offset
- bm_sectors
;
1391 space
= sb_offset
- data_offset
-
1392 __le64_to_cpu(sb
->data_size
);
1394 sb
->bblog_size
= __cpu_to_le16(8);
1395 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
1398 } else if (strcmp(update
, "no-bbl") == 0) {
1399 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
))
1400 pr_err("Cannot remove active bbl from %s\n",devname
);
1403 sb
->bblog_shift
= 0;
1404 sb
->bblog_offset
= 0;
1406 } else if (strcmp(update
, "force-no-bbl") == 0) {
1407 sb
->feature_map
&= ~ __cpu_to_le32(MD_FEATURE_BAD_BLOCKS
);
1409 sb
->bblog_shift
= 0;
1410 sb
->bblog_offset
= 0;
1411 } else if (strcmp(update
, "ppl") == 0) {
1412 unsigned long long sb_offset
= __le64_to_cpu(sb
->super_offset
);
1413 unsigned long long data_offset
= __le64_to_cpu(sb
->data_offset
);
1414 unsigned long long data_size
= __le64_to_cpu(sb
->data_size
);
1415 long bb_offset
= __le32_to_cpu(sb
->bblog_offset
);
1419 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1420 pr_err("Cannot add PPL to array with bitmap\n");
1424 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_JOURNAL
)) {
1425 pr_err("Cannot add PPL to array with journal\n");
1429 if (sb_offset
< data_offset
) {
1431 space
= bb_offset
- 8;
1433 space
= data_offset
- sb_offset
- 8;
1436 offset
= -(sb_offset
- data_offset
- data_size
);
1437 if (offset
< INT16_MIN
)
1439 space
= -(offset
- bb_offset
);
1442 if (space
< (PPL_HEADER_SIZE
>> 9) + 8) {
1443 pr_err("Not enough space to add ppl\n");
1447 if (space
>= (MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9)) {
1448 space
= (MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9);
1450 int optimal_space
= choose_ppl_space(
1451 __le32_to_cpu(sb
->chunksize
));
1452 if (space
> optimal_space
)
1453 space
= optimal_space
;
1454 if (space
> UINT16_MAX
)
1458 sb
->ppl
.offset
= __cpu_to_le16(offset
);
1459 sb
->ppl
.size
= __cpu_to_le16(space
);
1460 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_PPL
);
1461 } else if (strcmp(update
, "no-ppl") == 0) {
1462 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_PPL
|
1463 MD_FEATURE_MUTLIPLE_PPLS
);
1464 } else if (strcmp(update
, "name") == 0) {
1465 if (info
->name
[0] == 0)
1466 sprintf(info
->name
, "%d", info
->array
.md_minor
);
1467 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
1469 strchr(info
->name
, ':') == NULL
&&
1470 strlen(homehost
)+1+strlen(info
->name
) < 32) {
1471 strcpy(sb
->set_name
, homehost
);
1472 strcat(sb
->set_name
, ":");
1473 strcat(sb
->set_name
, info
->name
);
1477 namelen
= min((int)strlen(info
->name
),
1478 (int)sizeof(sb
->set_name
) - 1);
1479 memcpy(sb
->set_name
, info
->name
, namelen
);
1480 memset(&sb
->set_name
[namelen
], '\0',
1481 sizeof(sb
->set_name
) - namelen
);
1483 } else if (strcmp(update
, "devicesize") == 0 &&
1484 __le64_to_cpu(sb
->super_offset
) <
1485 __le64_to_cpu(sb
->data_offset
)) {
1486 /* set data_size to device size less data_offset */
1487 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
1488 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
1489 sb
->data_size
= __cpu_to_le64(
1490 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
1491 } else if (strncmp(update
, "revert-reshape", 14) == 0) {
1493 if (!(sb
->feature_map
&
1494 __cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
)))
1495 pr_err("No active reshape to revert on %s\n",
1499 unsigned long long reshape_sectors
;
1502 /* If the reshape hasn't started, just stop it.
1503 * It is conceivable that a stripe was modified but
1504 * the metadata not updated. In that case the backup
1505 * should have been used to get passed the critical stage.
1506 * If that couldn't happen, the "-nobackup" version
1509 if (strcmp(update
, "revert-reshape-nobackup") == 0 &&
1510 sb
->reshape_position
== 0 &&
1511 (__le32_to_cpu(sb
->delta_disks
) > 0 ||
1512 (__le32_to_cpu(sb
->delta_disks
) == 0 &&
1513 !(sb
->feature_map
& __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS
))))) {
1514 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_RESHAPE_ACTIVE
);
1515 sb
->raid_disks
= __cpu_to_le32(__le32_to_cpu(sb
->raid_disks
) -
1516 __le32_to_cpu(sb
->delta_disks
));
1517 sb
->delta_disks
= 0;
1520 /* reshape_position is a little messy.
1521 * Its value must be a multiple of the larger
1522 * chunk size, and of the "after" data disks.
1523 * So when reverting we need to change it to
1524 * be a multiple of the new "after" data disks,
1525 * which is the old "before".
1526 * If it isn't already a multiple of 'before',
1527 * the only thing we could do would be
1528 * copy some block around on the disks, which
1529 * is easy to get wrong.
1530 * So we reject a revert-reshape unless the
1531 * alignment is good.
1533 if (__le32_to_cpu(sb
->level
) >= 4 &&
1534 __le32_to_cpu(sb
->level
) <= 6) {
1536 __le64_to_cpu(sb
->reshape_position
);
1537 reshape_chunk
= __le32_to_cpu(sb
->new_chunk
);
1538 reshape_chunk
*= __le32_to_cpu(sb
->raid_disks
) -
1539 __le32_to_cpu(sb
->delta_disks
) -
1540 (__le32_to_cpu(sb
->level
)==6 ? 2 : 1);
1541 if (reshape_sectors
% reshape_chunk
) {
1542 pr_err("Reshape position is not suitably aligned.\n");
1543 pr_err("Try normal assembly and stop again\n");
1548 __cpu_to_le32(__le32_to_cpu(sb
->raid_disks
) -
1549 __le32_to_cpu(sb
->delta_disks
));
1550 if (sb
->delta_disks
== 0)
1551 sb
->feature_map
^= __cpu_to_le32(MD_FEATURE_RESHAPE_BACKWARDS
);
1553 sb
->delta_disks
= __cpu_to_le32(-__le32_to_cpu(sb
->delta_disks
));
1555 temp
= sb
->new_layout
;
1556 sb
->new_layout
= sb
->layout
;
1559 temp
= sb
->new_chunk
;
1560 sb
->new_chunk
= sb
->chunksize
;
1561 sb
->chunksize
= temp
;
1563 if (sb
->feature_map
&
1564 __cpu_to_le32(MD_FEATURE_NEW_OFFSET
)) {
1566 (int32_t)__le32_to_cpu(sb
->new_offset
);
1567 sb
->data_offset
= __cpu_to_le64(__le64_to_cpu(sb
->data_offset
) + offset_delta
);
1568 sb
->new_offset
= __cpu_to_le32(-offset_delta
);
1569 sb
->data_size
= __cpu_to_le64(__le64_to_cpu(sb
->data_size
) - offset_delta
);
1573 } else if (strcmp(update
, "_reshape_progress") == 0)
1574 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
1575 else if (strcmp(update
, "writemostly") == 0)
1576 sb
->devflags
|= WriteMostly1
;
1577 else if (strcmp(update
, "readwrite") == 0)
1578 sb
->devflags
&= ~WriteMostly1
;
1579 else if (strcmp(update
, "failfast") == 0)
1580 sb
->devflags
|= FailFast1
;
1581 else if (strcmp(update
, "nofailfast") == 0)
1582 sb
->devflags
&= ~FailFast1
;
1583 else if (strcmp(update
, "layout-original") == 0 ||
1584 strcmp(update
, "layout-alternate") == 0 ||
1585 strcmp(update
, "layout-unspecified") == 0) {
1586 if (__le32_to_cpu(sb
->level
) != 0) {
1587 pr_err("%s: %s only supported for RAID0\n",
1588 devname
?:"", update
);
1590 } else if (strcmp(update
, "layout-unspecified") == 0) {
1591 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
1594 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
1595 sb
->layout
= __cpu_to_le32(update
[7] == 'o' ? 1 : 2);
1600 sb
->sb_csum
= calc_sb_1_csum(sb
);
1605 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
1606 struct shape
*s
, char *name
, char *homehost
,
1607 int *uuid
, unsigned long long data_offset
)
1609 struct mdp_superblock_1
*sb
;
1614 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
1615 pr_err("could not allocate superblock\n");
1618 memset(sb
, 0, SUPER1_SIZE
);
1622 /* zeroing superblock */
1626 spares
= info
->working_disks
- info
->active_disks
;
1627 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
1628 pr_err("too many devices requested: %d+%d > %d\n",
1629 info
->raid_disks
, spares
, MAX_DEVS
);
1633 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
1634 sb
->major_version
= __cpu_to_le32(1);
1635 sb
->feature_map
= 0;
1639 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
1641 random_uuid(sb
->set_uuid
);;
1643 if (name
== NULL
|| *name
== 0) {
1644 sprintf(defname
, "%d", info
->md_minor
);
1648 strchr(name
, ':')== NULL
&&
1649 strlen(homehost
)+1+strlen(name
) < 32) {
1650 strcpy(sb
->set_name
, homehost
);
1651 strcat(sb
->set_name
, ":");
1652 strcat(sb
->set_name
, name
);
1656 namelen
= min((int)strlen(name
),
1657 (int)sizeof(sb
->set_name
) - 1);
1658 memcpy(sb
->set_name
, name
, namelen
);
1659 memset(&sb
->set_name
[namelen
], '\0',
1660 sizeof(sb
->set_name
) - namelen
);
1663 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
1664 sb
->level
= __cpu_to_le32(info
->level
);
1665 sb
->layout
= __cpu_to_le32(info
->layout
);
1666 sb
->size
= __cpu_to_le64(s
->size
*2ULL);
1667 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
1668 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
1670 sb
->data_offset
= __cpu_to_le64(data_offset
);
1671 sb
->data_size
= __cpu_to_le64(0);
1672 sb
->super_offset
= __cpu_to_le64(0);
1673 sb
->recovery_offset
= __cpu_to_le64(0);
1675 sb
->utime
= sb
->ctime
;
1676 sb
->events
= __cpu_to_le64(1);
1677 if (info
->state
& (1<<MD_SB_CLEAN
))
1678 sb
->resync_offset
= MaxSector
;
1680 sb
->resync_offset
= 0;
1681 sbsize
= sizeof(struct mdp_superblock_1
) +
1682 2 * (info
->raid_disks
+ spares
);
1683 sbsize
= ROUND_UP(sbsize
, 512);
1685 __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
1687 memset(sb
->dev_roles
, 0xff,
1688 MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
1690 if (s
->consistency_policy
== CONSISTENCY_POLICY_PPL
)
1691 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_PPL
);
1699 long long data_offset
;
1700 unsigned long long dev_size
;
1701 mdu_disk_info_t disk
;
1702 struct devinfo
*next
;
1705 /* Add a device to the superblock being created */
1706 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
1707 int fd
, char *devname
, unsigned long long data_offset
)
1709 struct mdp_superblock_1
*sb
= st
->sb
;
1710 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
1711 struct devinfo
*di
, **dip
;
1714 dk_state
= dk
->state
& ~(1<<MD_DISK_FAILFAST
);
1715 if ((dk_state
& (1<<MD_DISK_ACTIVE
)) &&
1716 (dk_state
& (1<<MD_DISK_SYNC
)))/* active, sync */
1717 *rp
= __cpu_to_le16(dk
->raid_disk
);
1718 else if (dk_state
& (1<<MD_DISK_JOURNAL
))
1719 *rp
= MD_DISK_ROLE_JOURNAL
;
1720 else if ((dk_state
& ~(1<<MD_DISK_ACTIVE
)) == 0)
1721 /* active or idle -> spare */
1722 *rp
= MD_DISK_ROLE_SPARE
;
1724 *rp
= MD_DISK_ROLE_FAULTY
;
1726 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
1727 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
1728 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
1730 sb
->dev_number
= __cpu_to_le32(dk
->number
);
1731 sb
->devflags
= 0; /* don't copy another disks flags */
1732 sb
->sb_csum
= calc_sb_1_csum(sb
);
1734 dip
= (struct devinfo
**)&st
->info
;
1736 dip
= &(*dip
)->next
;
1737 di
= xmalloc(sizeof(struct devinfo
));
1739 di
->devname
= devname
;
1741 di
->data_offset
= data_offset
;
1742 get_dev_size(fd
, NULL
, &di
->dev_size
);
1749 static int locate_bitmap1(struct supertype
*st
, int fd
, int node_num
);
1751 static int store_super1(struct supertype
*st
, int fd
)
1753 struct mdp_superblock_1
*sb
= st
->sb
;
1754 unsigned long long sb_offset
;
1755 struct align_fd afd
;
1757 unsigned long long dsize
;
1759 if (!get_dev_size(fd
, NULL
, &dsize
))
1770 * Calculate the position of the superblock.
1771 * It is always aligned to a 4K boundary and
1772 * depending on minor_version, it can be:
1773 * 0: At least 8K, but less than 12K, from end of device
1774 * 1: At start of device
1775 * 2: 4K from start of device.
1777 switch(st
->minor_version
) {
1781 sb_offset
&= ~(4*2-1);
1793 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1794 0 != __le64_to_cpu(sb
->super_offset
)
1796 pr_err("internal error - sb_offset is wrong\n");
1800 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1803 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1805 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1808 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1809 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1810 (((char*)sb
)+MAX_SB_SIZE
);
1811 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1812 locate_bitmap1(st
, fd
, 0);
1813 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1822 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1824 static unsigned long choose_bm_space(unsigned long devsize
)
1826 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1827 * if bigger than 200Gig, save 128k
1828 * NOTE: result must be multiple of 4K else bad things happen
1829 * on 4K-sector devices.
1833 if (devsize
- 64*2 >= 200*1024*1024*2)
1835 if (devsize
- 4*2 > 8*1024*1024*2)
1840 static void free_super1(struct supertype
*st
);
1842 __u32
crc32c_le(__u32 crc
, unsigned char const *p
, size_t len
);
1844 static int write_init_ppl1(struct supertype
*st
, struct mdinfo
*info
, int fd
)
1846 struct mdp_superblock_1
*sb
= st
->sb
;
1848 struct ppl_header
*ppl_hdr
;
1851 /* first clear entire ppl space */
1852 ret
= zero_disk_range(fd
, info
->ppl_sector
, info
->ppl_size
);
1856 ret
= posix_memalign(&buf
, 4096, PPL_HEADER_SIZE
);
1858 pr_err("Failed to allocate PPL header buffer\n");
1862 memset(buf
, 0, PPL_HEADER_SIZE
);
1864 memset(ppl_hdr
->reserved
, 0xff, PPL_HDR_RESERVED
);
1865 ppl_hdr
->signature
= __cpu_to_le32(~crc32c_le(~0, sb
->set_uuid
,
1866 sizeof(sb
->set_uuid
)));
1867 ppl_hdr
->checksum
= __cpu_to_le32(~crc32c_le(~0, buf
, PPL_HEADER_SIZE
));
1869 if (lseek64(fd
, info
->ppl_sector
* 512, SEEK_SET
) < 0) {
1871 perror("Failed to seek to PPL header location");
1874 if (!ret
&& write(fd
, buf
, PPL_HEADER_SIZE
) != PPL_HEADER_SIZE
) {
1876 perror("Write PPL header failed");
1886 #define META_BLOCK_SIZE 4096
1888 static int write_empty_r5l_meta_block(struct supertype
*st
, int fd
)
1890 struct r5l_meta_block
*mb
;
1891 struct mdp_superblock_1
*sb
= st
->sb
;
1892 struct align_fd afd
;
1897 if (posix_memalign((void**)&mb
, 4096, META_BLOCK_SIZE
) != 0) {
1898 pr_err("Could not allocate memory for the meta block.\n");
1902 memset(mb
, 0, META_BLOCK_SIZE
);
1904 mb
->magic
= __cpu_to_le32(R5LOG_MAGIC
);
1905 mb
->version
= R5LOG_VERSION
;
1906 mb
->meta_size
= __cpu_to_le32(sizeof(struct r5l_meta_block
));
1907 mb
->seq
= __cpu_to_le64(random32());
1908 mb
->position
= __cpu_to_le64(0);
1910 crc
= crc32c_le(0xffffffff, sb
->set_uuid
, sizeof(sb
->set_uuid
));
1911 crc
= crc32c_le(crc
, (void *)mb
, META_BLOCK_SIZE
);
1914 if (lseek64(fd
, __le64_to_cpu(sb
->data_offset
) * 512, 0) < 0LL) {
1915 pr_err("cannot seek to offset of the meta block\n");
1919 if (awrite(&afd
, mb
, META_BLOCK_SIZE
) != META_BLOCK_SIZE
) {
1920 pr_err("failed to store write the meta block \n");
1933 static int write_init_super1(struct supertype
*st
)
1935 struct mdp_superblock_1
*sb
= st
->sb
;
1936 struct supertype
*refst
;
1938 unsigned long long bm_space
;
1940 unsigned long long dsize
, array_size
;
1941 unsigned long long sb_offset
;
1942 unsigned long long data_offset
;
1944 int raid0_need_layout
= 0;
1946 for (di
= st
->info
; di
; di
= di
->next
) {
1947 if (di
->disk
.state
& (1 << MD_DISK_JOURNAL
))
1948 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_JOURNAL
);
1949 if (sb
->level
== 0 && sb
->layout
!= 0) {
1950 struct devinfo
*di2
= st
->info
;
1951 unsigned long long s1
, s2
;
1953 if (di
->data_offset
!= INVALID_SECTORS
)
1954 s1
-= di
->data_offset
;
1955 s1
/= __le32_to_cpu(sb
->chunksize
);
1957 if (di2
->data_offset
!= INVALID_SECTORS
)
1958 s2
-= di2
->data_offset
;
1959 s2
/= __le32_to_cpu(sb
->chunksize
);
1961 raid0_need_layout
= 1;
1965 for (di
= st
->info
; di
; di
= di
->next
) {
1966 if (di
->disk
.state
& (1 << MD_DISK_FAULTY
))
1971 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
1974 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1975 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1976 sb
->devflags
|= WriteMostly1
;
1978 sb
->devflags
&= ~WriteMostly1
;
1979 if (di
->disk
.state
& (1<<MD_DISK_FAILFAST
))
1980 sb
->devflags
|= FailFast1
;
1982 sb
->devflags
&= ~FailFast1
;
1984 random_uuid(sb
->device_uuid
);
1986 if (!(di
->disk
.state
& (1<<MD_DISK_JOURNAL
)))
1989 refst
= dup_super(st
);
1990 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1991 struct mdp_superblock_1
*refsb
= refst
->sb
;
1993 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1994 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
1995 /* same array, so preserve events and
1997 sb
->events
= refsb
->events
;
1998 /* bugs in 2.6.17 and earlier mean the
1999 * dev_number chosen in Manage must be preserved
2001 if (get_linux_version() >= 2006018)
2002 sb
->dev_number
= refsb
->dev_number
;
2008 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
2021 * Calculate the position of the superblock.
2022 * It is always aligned to a 4K boundary and
2023 * depending on minor_version, it can be:
2024 * 0: At least 8K, but less than 12K, from end of device
2025 * 1: At start of device
2026 * 2: 4K from start of device.
2027 * data_offset has already been set.
2029 array_size
= __le64_to_cpu(sb
->size
);
2031 /* work out how much space we left for a bitmap */
2032 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
2033 bitmap_super_t
*bms
= (bitmap_super_t
*)
2034 (((char *)sb
) + MAX_SB_SIZE
);
2035 bm_space
= calc_bitmap_size(bms
, 4096) >> 9;
2036 bm_offset
= (long)__le32_to_cpu(sb
->bitmap_offset
);
2037 } else if (md_feature_any_ppl_on(sb
->feature_map
)) {
2038 bm_space
= MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9;
2039 if (st
->minor_version
== 0)
2040 bm_offset
= -bm_space
- 8;
2043 sb
->ppl
.offset
= __cpu_to_le16(bm_offset
);
2044 sb
->ppl
.size
= __cpu_to_le16(bm_space
);
2046 bm_space
= choose_bm_space(array_size
);
2050 data_offset
= di
->data_offset
;
2051 if (data_offset
== INVALID_SECTORS
)
2052 data_offset
= st
->data_offset
;
2053 switch(st
->minor_version
) {
2055 /* Add 8 sectors for bad block log */
2057 if (data_offset
== INVALID_SECTORS
)
2061 sb_offset
&= ~(4*2-1);
2062 sb
->data_offset
= __cpu_to_le64(data_offset
);
2063 sb
->super_offset
= __cpu_to_le64(sb_offset
);
2064 if (sb_offset
< array_size
+ bm_space
)
2065 bm_space
= sb_offset
- array_size
;
2066 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
2067 if (bm_space
>= 8) {
2068 sb
->bblog_size
= __cpu_to_le16(8);
2069 sb
->bblog_offset
= __cpu_to_le32((unsigned)-8);
2074 sb_offset
= st
->minor_version
== 2 ? 8 : 0;
2075 sb
->super_offset
= __cpu_to_le64(sb_offset
);
2076 if (data_offset
== INVALID_SECTORS
)
2077 data_offset
= sb_offset
+ 16;
2079 sb
->data_offset
= __cpu_to_le64(data_offset
);
2080 sb
->data_size
= __cpu_to_le64(dsize
- data_offset
);
2081 if (data_offset
>= sb_offset
+bm_offset
+bm_space
+8) {
2082 sb
->bblog_size
= __cpu_to_le16(8);
2083 sb
->bblog_offset
= __cpu_to_le32(bm_offset
+
2085 } else if (data_offset
>= sb_offset
+ 16) {
2086 sb
->bblog_size
= __cpu_to_le16(8);
2087 /* '8' sectors for the bblog, and 'sb_offset'
2088 * because we want offset from superblock, not
2091 sb
->bblog_offset
= __cpu_to_le32(data_offset
-
2096 pr_err("Failed to write invalid metadata format 1.%i to %s\n",
2097 st
->minor_version
, di
->devname
);
2102 * Disable badblock log on clusters, or when
2103 * explicitly requested
2105 if (st
->nodes
> 0 || conf_get_create_info()->bblist
== 0) {
2107 sb
->bblog_offset
= 0;
2110 /* RAID0 needs a layout if devices aren't all the same size */
2111 if (raid0_need_layout
)
2112 sb
->feature_map
|= __cpu_to_le32(MD_FEATURE_RAID0_LAYOUT
);
2114 sb
->sb_csum
= calc_sb_1_csum(sb
);
2115 rv
= store_super1(st
, di
->fd
);
2117 if (rv
== 0 && (di
->disk
.state
& (1 << MD_DISK_JOURNAL
))) {
2118 rv
= write_empty_r5l_meta_block(st
, di
->fd
);
2124 (__le32_to_cpu(sb
->feature_map
) &
2125 MD_FEATURE_BITMAP_OFFSET
)) {
2126 rv
= st
->ss
->write_bitmap(st
, di
->fd
, NodeNumUpdate
);
2127 } else if (rv
== 0 &&
2128 md_feature_any_ppl_on(sb
->feature_map
)) {
2131 st
->ss
->getinfo_super(st
, &info
, NULL
);
2132 rv
= st
->ss
->write_init_ppl(st
, &info
, di
->fd
);
2142 pr_err("Failed to write metadata to %s\n", di
->devname
);
2147 static int compare_super1(struct supertype
*st
, struct supertype
*tst
,
2152 * 0 same, or first was empty, and second was copied
2153 * 1 second had wrong number
2155 * 3 wrong other info
2157 struct mdp_superblock_1
*first
= st
->sb
;
2158 struct mdp_superblock_1
*second
= tst
->sb
;
2160 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
2162 if (second
->major_version
!= __cpu_to_le32(1))
2166 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
2167 pr_err("could not allocate superblock\n");
2170 memcpy(first
, second
, SUPER1_SIZE
);
2174 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
2177 if (first
->ctime
!= second
->ctime
||
2178 first
->level
!= second
->level
||
2179 first
->layout
!= second
->layout
||
2180 first
->size
!= second
->size
||
2181 first
->chunksize
!= second
->chunksize
||
2182 first
->raid_disks
!= second
->raid_disks
)
2187 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
2189 unsigned long long dsize
;
2190 unsigned long long sb_offset
;
2191 struct mdp_superblock_1
*super
;
2193 struct bitmap_super_s
*bsb
;
2194 struct misc_dev_info
*misc
;
2195 struct align_fd afd
;
2201 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
2203 struct supertype tst
;
2204 __u64 bestctime
= 0;
2205 /* guess... choose latest ctime */
2206 memset(&tst
, 0, sizeof(tst
));
2208 for (tst
.minor_version
= 0; tst
.minor_version
<= 2;
2209 tst
.minor_version
++) {
2210 switch(load_super1(&tst
, fd
, devname
)) {
2211 case 0: super
= tst
.sb
;
2212 if (bestvers
== -1 ||
2213 bestctime
< __le64_to_cpu(super
->ctime
)) {
2214 bestvers
= tst
.minor_version
;
2215 bestctime
= __le64_to_cpu(super
->ctime
);
2220 case 1: return 1; /*bad device */
2221 case 2: break; /* bad, try next */
2224 if (bestvers
!= -1) {
2226 tst
.minor_version
= bestvers
;
2228 tst
.max_devs
= MAX_DEVS
;
2229 rv
= load_super1(&tst
, fd
, devname
);
2236 if (!get_dev_size(fd
, devname
, &dsize
))
2242 pr_err("%s is too small for md: size is %llu sectors.\n",
2248 * Calculate the position of the superblock.
2249 * It is always aligned to a 4K boundary and
2250 * depending on minor_version, it can be:
2251 * 0: At least 8K, but less than 12K, from end of device
2252 * 1: At start of device
2253 * 2: 4K from start of device.
2255 switch(st
->minor_version
) {
2259 sb_offset
&= ~(4*2-1);
2271 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
2273 pr_err("Cannot seek to superblock on %s: %s\n",
2274 devname
, strerror(errno
));
2278 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
2279 pr_err("could not allocate superblock\n");
2283 memset(super
, 0, SUPER1_SIZE
);
2285 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
2287 pr_err("Cannot read superblock on %s\n",
2293 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
2295 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
2296 devname
, MD_SB_MAGIC
,
2297 __le32_to_cpu(super
->magic
));
2302 if (__le32_to_cpu(super
->major_version
) != 1) {
2304 pr_err("Cannot interpret superblock on %s - version is %d\n",
2305 devname
, __le32_to_cpu(super
->major_version
));
2309 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
2311 pr_err("No superblock found on %s (super_offset is wrong)\n",
2318 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
2320 misc
= (struct misc_dev_info
*)
2321 (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
2322 misc
->device_size
= dsize
;
2323 if (st
->data_offset
== INVALID_SECTORS
)
2324 st
->data_offset
= __le64_to_cpu(super
->data_offset
);
2326 /* Now check on the bitmap superblock */
2327 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
2329 /* Read the bitmap superblock and make sure it looks
2330 * valid. If it doesn't clear the bit. An --assemble --force
2331 * should get that written out.
2333 locate_bitmap1(st
, fd
, 0);
2334 if (aread(&afd
, bsb
, 512) != 512)
2337 uuid_from_super1(st
, uuid
);
2338 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
2339 memcmp(bsb
->uuid
, uuid
, 16) != 0)
2344 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
2345 & ~MD_FEATURE_BITMAP_OFFSET
);
2349 static struct supertype
*match_metadata_desc1(char *arg
)
2351 struct supertype
*st
= xcalloc(1, sizeof(*st
));
2353 st
->container_devnm
[0] = 0;
2355 st
->max_devs
= MAX_DEVS
;
2357 st
->data_offset
= INVALID_SECTORS
;
2358 /* leading zeros can be safely ignored. --detail generates them. */
2361 if (strcmp(arg
, "1.0") == 0 || strcmp(arg
, "1.00") == 0) {
2362 st
->minor_version
= 0;
2365 if (strcmp(arg
, "1.1") == 0 || strcmp(arg
, "1.01") == 0
2367 st
->minor_version
= 1;
2370 if (strcmp(arg
, "1.2") == 0 ||
2371 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
2372 strcmp(arg
, "default") == 0 ||
2373 #endif /* DEFAULT_OLD_METADATA */
2374 strcmp(arg
, "1.02") == 0) {
2375 st
->minor_version
= 2;
2378 if (strcmp(arg
, "1") == 0 || strcmp(arg
, "default") == 0) {
2379 st
->minor_version
= -1;
2387 /* find available size on device with this devsize, using
2388 * superblock type st, and reserving 'reserve' sectors for
2391 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
,
2392 unsigned long long data_offset
)
2394 struct mdp_superblock_1
*super
= st
->sb
;
2400 if (__le32_to_cpu(super
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
) {
2401 /* hot-add. allow for actual size of bitmap */
2402 struct bitmap_super_s
*bsb
;
2403 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
2404 bmspace
= calc_bitmap_size(bsb
, 4096) >> 9;
2405 } else if (md_feature_any_ppl_on(super
->feature_map
)) {
2406 bmspace
= __le16_to_cpu(super
->ppl
.size
);
2409 /* Allow space for bad block log */
2410 if (super
->bblog_size
)
2411 bbspace
= __le16_to_cpu(super
->bblog_size
);
2413 if (st
->minor_version
< 0)
2414 /* not specified, so time to set default */
2415 st
->minor_version
= 2;
2417 if (data_offset
== INVALID_SECTORS
)
2418 data_offset
= st
->data_offset
;
2420 if (data_offset
!= INVALID_SECTORS
)
2421 switch(st
->minor_version
) {
2423 return devsize
- data_offset
- 8*2 - bbspace
;
2426 return devsize
- data_offset
;
2433 switch(st
->minor_version
) {
2436 return ((devsize
- 8*2 - bbspace
) & ~(4*2-1));
2438 /* at start, 4K for superblock and possible bitmap */
2439 return devsize
- 4*2 - bbspace
;
2441 /* 4k from start, 4K for superblock and possible bitmap */
2442 return devsize
- (4+4)*2 - bbspace
;
2448 add_internal_bitmap1(struct supertype
*st
,
2449 int *chunkp
, int delay
, int write_behind
,
2450 unsigned long long size
,
2451 int may_change
, int major
)
2454 * If not may_change, then this is a 'Grow' without sysfs support for
2455 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
2456 * If may_change, then this is create or a Grow with sysfs support,
2457 * and we can put the bitmap wherever we like.
2459 * size is in sectors, chunk is in bytes !!!
2462 unsigned long long bits
;
2463 unsigned long long max_bits
;
2464 unsigned long long min_chunk
;
2466 long bbl_offset
, bbl_size
;
2467 unsigned long long chunk
= *chunkp
;
2471 struct mdp_superblock_1
*sb
= st
->sb
;
2472 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
2475 if (__le64_to_cpu(sb
->data_size
) == 0)
2477 * Must be creating the array, else data_size
2481 switch(st
->minor_version
) {
2484 * either 3K after the superblock (when hot-add),
2485 * or some amount of space before.
2489 * We are creating array, so we *know* how much room has
2495 choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2497 room
= __le64_to_cpu(sb
->super_offset
)
2498 - __le64_to_cpu(sb
->data_offset
)
2499 - __le64_to_cpu(sb
->data_size
);
2500 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2503 bbl_offset
= (__s32
)__le32_to_cpu(sb
->bblog_offset
);
2504 if (bbl_size
< -bbl_offset
)
2505 bbl_size
= -bbl_offset
;
2508 (room
< 3*2 && __le32_to_cpu(sb
->max_dev
) <= 384)) {
2513 offset
= 0; /* means movable offset */
2518 case 2: /* between superblock and data */
2523 choose_bm_space(__le64_to_cpu(sb
->size
)) + bbl_size
;
2525 room
= __le64_to_cpu(sb
->data_offset
)
2526 - __le64_to_cpu(sb
->super_offset
);
2527 bbl_size
= __le16_to_cpu(sb
->bblog_size
);
2530 __le32_to_cpu(sb
->bblog_offset
) + bbl_size
;
2535 room
-= 2; /* Leave 1K for superblock */
2539 room
-= 4*2; /* leave 4K for superblock */
2549 if (chunk
== UnSet
&& room
> 128*2)
2550 /* Limit to 128K of bitmap when chunk size not requested */
2554 /* No room for a bitmap */
2557 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
2559 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
2560 bits
= (size
*512)/min_chunk
+1;
2561 while (bits
> max_bits
) {
2565 if (chunk
== UnSet
) {
2566 /* For practical purpose, 64Meg is a good
2567 * default chunk size for internal bitmaps.
2570 if (chunk
< 64*1024*1024)
2571 chunk
= 64*1024*1024;
2572 } else if (chunk
< min_chunk
)
2573 return -EINVAL
; /* chunk size too small */
2574 if (chunk
== 0) /* rounding problem */
2578 /* start bitmap on a 4K boundary with enough space for
2581 bits
= (size
*512) / chunk
+ 1;
2582 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
2583 room
*= 8; /* convert 4K blocks to sectors */
2584 offset
= -room
- bbl_size
;
2587 sb
->bitmap_offset
= (int32_t)__cpu_to_le32(offset
);
2589 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) |
2590 MD_FEATURE_BITMAP_OFFSET
);
2591 memset(bms
, 0, sizeof(*bms
));
2592 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
2593 bms
->version
= __cpu_to_le32(major
);
2594 uuid_from_super1(st
, uuid
);
2595 memcpy(bms
->uuid
, uuid
, 16);
2596 bms
->chunksize
= __cpu_to_le32(chunk
);
2597 bms
->daemon_sleep
= __cpu_to_le32(delay
);
2598 bms
->sync_size
= __cpu_to_le64(size
);
2599 bms
->write_behind
= __cpu_to_le32(write_behind
);
2600 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2602 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
) |
2603 MD_FEATURE_BITMAP_VERSIONED
);
2604 if (st
->cluster_name
) {
2605 len
= sizeof(bms
->cluster_name
);
2606 strncpy((char *)bms
->cluster_name
, st
->cluster_name
, len
);
2607 bms
->cluster_name
[len
- 1] = '\0';
2614 static int locate_bitmap1(struct supertype
*st
, int fd
, int node_num
)
2616 unsigned long long offset
, bm_sectors_per_node
;
2617 struct mdp_superblock_1
*sb
;
2618 bitmap_super_t
*bms
;
2623 if (st
->ss
->load_super(st
, fd
, NULL
))
2624 return -1; /* no error I hope... */
2629 if ((__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_BITMAP_OFFSET
))
2634 offset
= __le64_to_cpu(sb
->super_offset
) + (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
2636 bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
2637 bm_sectors_per_node
= calc_bitmap_size(bms
, 4096) >> 9;
2638 offset
+= bm_sectors_per_node
* node_num
;
2642 lseek64(fd
, offset
<<9, 0);
2646 static int write_bitmap1(struct supertype
*st
, int fd
, enum bitmap_update update
)
2648 struct mdp_superblock_1
*sb
= st
->sb
;
2649 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
2652 int towrite
, n
, len
;
2653 struct align_fd afd
;
2655 unsigned long long total_bm_space
, bm_space_per_node
;
2659 /* update cluster name */
2660 if (st
->cluster_name
) {
2661 len
= sizeof(bms
->cluster_name
);
2662 memset((char *)bms
->cluster_name
, 0, len
);
2663 strncpy((char *)bms
->cluster_name
,
2664 st
->cluster_name
, len
);
2665 bms
->cluster_name
[len
- 1] = '\0';
2669 /* cluster md only supports superblock 1.2 now */
2670 if (st
->minor_version
!= 2 &&
2671 bms
->version
== BITMAP_MAJOR_CLUSTERED
) {
2672 pr_err("Warning: cluster md only works with superblock 1.2\n");
2676 if (bms
->version
== BITMAP_MAJOR_CLUSTERED
) {
2677 if (__cpu_to_le32(st
->nodes
) < bms
->nodes
) {
2679 * Since the nodes num is not increased, no
2680 * need to check the space enough or not,
2681 * just update bms->nodes
2683 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2688 * no need to change bms->nodes for other
2692 pr_err("Warning: --nodes option is only suitable for clustered bitmap\n");
2697 * Each node has an independent bitmap, it is necessary to
2698 * calculate the space is enough or not, first get how many
2699 * bytes for the total bitmap
2701 bm_space_per_node
= calc_bitmap_size(bms
, 4096);
2703 total_bm_space
= 512 * (__le64_to_cpu(sb
->data_offset
) -
2704 __le64_to_cpu(sb
->super_offset
));
2705 /* leave another 4k for superblock */
2706 total_bm_space
= total_bm_space
- 4096;
2708 if (bm_space_per_node
* st
->nodes
> total_bm_space
) {
2709 pr_err("Warning: The max num of nodes can't exceed %llu\n",
2710 total_bm_space
/ bm_space_per_node
);
2714 bms
->nodes
= __cpu_to_le32(st
->nodes
);
2723 if (locate_bitmap1(st
, fd
, 0) < 0) {
2724 pr_err("Error: Invalid bitmap\n");
2728 if (posix_memalign(&buf
, 4096, 4096))
2732 /* Only the bitmap[0] should resync
2733 * whole device on initial assembly
2736 memset(buf
, 0x00, 4096);
2738 memset(buf
, 0xff, 4096);
2739 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
2742 * use 4096 boundary if bitmap_offset is aligned
2743 * with 8 sectors, then it should compatible with
2746 if (__le32_to_cpu(sb
->bitmap_offset
) & 7)
2747 towrite
= calc_bitmap_size(bms
, 512);
2749 towrite
= calc_bitmap_size(bms
, 4096);
2750 while (towrite
> 0) {
2754 n
= awrite(&afd
, buf
, n
);
2760 memset(buf
, 0x00, 4096);
2762 memset(buf
, 0xff, 4096);
2769 } while (++i
< __le32_to_cpu(bms
->nodes
));
2775 static void free_super1(struct supertype
*st
)
2781 struct devinfo
*di
= st
->info
;
2782 st
->info
= di
->next
;
2790 static int validate_geometry1(struct supertype
*st
, int level
,
2791 int layout
, int raiddisks
,
2792 int *chunk
, unsigned long long size
,
2793 unsigned long long data_offset
,
2794 char *subdev
, unsigned long long *freesize
,
2795 int consistency_policy
, int verbose
)
2797 unsigned long long ldsize
, devsize
;
2799 unsigned long long headroom
;
2800 unsigned long long overhead
;
2803 if (level
== LEVEL_CONTAINER
) {
2805 pr_err("1.x metadata does not support containers\n");
2808 if (*chunk
== UnSet
)
2809 *chunk
= DEFAULT_CHUNK
;
2814 if (st
->minor_version
< 0)
2815 /* not specified, so time to set default */
2816 st
->minor_version
= 2;
2818 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
2821 pr_err("super1.x cannot open %s: %s\n",
2822 subdev
, strerror(errno
));
2826 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
2832 devsize
= ldsize
>> 9;
2834 /* creating: allow suitable space for bitmap or PPL */
2835 if (consistency_policy
== CONSISTENCY_POLICY_PPL
)
2836 bmspace
= MULTIPLE_PPL_AREA_SIZE_SUPER1
>> 9;
2838 bmspace
= choose_bm_space(devsize
);
2840 if (data_offset
== INVALID_SECTORS
)
2841 data_offset
= st
->data_offset
;
2842 if (data_offset
== INVALID_SECTORS
)
2843 switch (st
->minor_version
) {
2849 /* Choose data offset appropriate for this device
2850 * and use as default for whole array.
2851 * The data_offset must allow for bitmap space
2852 * and base metadata, should allow for some headroom
2853 * for reshape, and should be rounded to multiple
2855 * Headroom is limited to 128M, but aim for about 0.1%
2857 headroom
= 128*1024*2;
2858 while ((headroom
<< 10) > devsize
&&
2860 headroom
/ 2 >= ((unsigned)(*chunk
)*2)*2))
2862 data_offset
= 12*2 + bmspace
+ headroom
;
2863 #define ONE_MEG (2*1024)
2864 data_offset
= ROUND_UP(data_offset
, ONE_MEG
);
2867 if (st
->data_offset
== INVALID_SECTORS
)
2868 st
->data_offset
= data_offset
;
2869 switch(st
->minor_version
) {
2870 case 0: /* metadata at end. Round down and subtract space to reserve */
2871 devsize
= (devsize
& ~(4ULL*2-1));
2872 /* space for metadata, bblog, bitmap/ppl */
2873 overhead
= 8*2 + 8 + bmspace
;
2874 if (devsize
< overhead
) /* detect underflow */
2875 goto dev_too_small_err
;
2876 devsize
-= overhead
;
2880 if (devsize
< data_offset
) /* detect underflow */
2881 goto dev_too_small_err
;
2882 devsize
-= data_offset
;
2885 *freesize
= devsize
;
2888 /* Error condition, device cannot even hold the overhead. */
2890 fprintf(stderr
, "device %s is too small (%lluK) for "
2891 "required metadata!\n", subdev
, devsize
>>1);
2896 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
)
2898 /* Create a v1.0 superblock based on 'info'*/
2900 struct mdp_superblock_1
*sb
;
2902 unsigned long long offset
;
2904 if (posix_memalign(&ret
, 4096, 1024) != 0)
2907 memset(ret
, 0, 1024);
2908 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
2909 sb
->major_version
= __cpu_to_le32(1);
2911 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
2912 sprintf(sb
->set_name
, "%d", sb0
->md_minor
);
2913 sb
->ctime
= __cpu_to_le32(info
->array
.ctime
+1);
2914 sb
->level
= __cpu_to_le32(info
->array
.level
);
2915 sb
->layout
= __cpu_to_le32(info
->array
.layout
);
2916 sb
->size
= __cpu_to_le64(info
->component_size
);
2917 sb
->chunksize
= __cpu_to_le32(info
->array
.chunk_size
/512);
2918 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
2919 if (info
->array
.level
> 0)
2920 sb
->data_size
= sb
->size
;
2922 sb
->data_size
= st
->ss
->avail_size(st
, st
->devsize
/512, 0);
2923 sb
->resync_offset
= MaxSector
;
2924 sb
->max_dev
= __cpu_to_le32(MD_SB_DISKS
);
2925 sb
->dev_number
= __cpu_to_le32(info
->disk
.number
);
2926 sb
->utime
= __cpu_to_le64(info
->array
.utime
);
2928 offset
= st
->devsize
/512 - 8*2;
2930 sb
->super_offset
= __cpu_to_le64(offset
);
2931 //*(__u64*)(st->other + 128 + 8 + 8) = __cpu_to_le64(offset);
2933 random_uuid(sb
->device_uuid
);
2935 for (i
= 0; i
< MD_SB_DISKS
; i
++) {
2936 int state
= sb0
->disks
[i
].state
;
2937 sb
->dev_roles
[i
] = MD_DISK_ROLE_SPARE
;
2938 if ((state
& (1<<MD_DISK_SYNC
)) &&
2939 !(state
& (1<<MD_DISK_FAULTY
)))
2940 sb
->dev_roles
[i
] = __cpu_to_le16(sb0
->disks
[i
].raid_disk
);
2942 sb
->sb_csum
= calc_sb_1_csum(sb
);
2946 struct superswitch super1
= {
2947 .examine_super
= examine_super1
,
2948 .brief_examine_super
= brief_examine_super1
,
2949 .export_examine_super
= export_examine_super1
,
2950 .detail_super
= detail_super1
,
2951 .brief_detail_super
= brief_detail_super1
,
2952 .export_detail_super
= export_detail_super1
,
2953 .write_init_super
= write_init_super1
,
2954 .validate_geometry
= validate_geometry1
,
2955 .add_to_super
= add_to_super1
,
2956 .examine_badblocks
= examine_badblocks_super1
,
2957 .copy_metadata
= copy_metadata1
,
2958 .write_init_ppl
= write_init_ppl1
,
2959 .match_home
= match_home1
,
2960 .uuid_from_super
= uuid_from_super1
,
2961 .getinfo_super
= getinfo_super1
,
2962 .container_content
= container_content1
,
2963 .update_super
= update_super1
,
2964 .init_super
= init_super1
,
2965 .store_super
= store_super1
,
2966 .compare_super
= compare_super1
,
2967 .load_super
= load_super1
,
2968 .match_metadata_desc
= match_metadata_desc1
,
2969 .avail_size
= avail_size1
,
2970 .add_internal_bitmap
= add_internal_bitmap1
,
2971 .locate_bitmap
= locate_bitmap1
,
2972 .write_bitmap
= write_bitmap1
,
2973 .free_super
= free_super1
,
2974 #if __BYTE_ORDER == BIG_ENDIAN