2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
27 * The version-1 superblock :
28 * All numeric fields are little-endian.
30 * total size: 256 bytes plus 2 per device.
31 * 1K allows 384 devices.
33 struct mdp_superblock_1
{
34 /* constant array information - 128 bytes */
35 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
36 __u32 major_version
; /* 1 */
37 __u32 feature_map
; /* 0 for now */
38 __u32 pad0
; /* always set to 0 when writing */
40 __u8 set_uuid
[16]; /* user-space generated. */
41 char set_name
[32]; /* set and interpreted by user-space */
43 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
44 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
45 __u32 layout
; /* only for raid5 currently */
46 __u64 size
; /* used size of component devices, in 512byte sectors */
48 __u32 chunksize
; /* in 512byte sectors */
50 __u32 bitmap_offset
; /* sectors after start of superblock that bitmap starts
51 * NOTE: signed, so bitmap can be before superblock
52 * only meaningful of feature_map[0] is set.
55 /* These are only valid with feature bit '4' */
56 __u32 new_level
; /* new level we are reshaping to */
57 __u64 reshape_position
; /* next address in array-space for reshape */
58 __u32 delta_disks
; /* change in number of raid_disks */
59 __u32 new_layout
; /* new layout */
60 __u32 new_chunk
; /* new chunk size (bytes) */
61 __u8 pad1
[128-124]; /* set to 0 when written */
63 /* constant this-device information - 64 bytes */
64 __u64 data_offset
; /* sector start of data, often 0 */
65 __u64 data_size
; /* sectors in this device that can be used for data */
66 __u64 super_offset
; /* sector start of this superblock */
67 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
68 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
69 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
70 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
71 __u8 devflags
; /* per-device flags. Only one defined...*/
72 #define WriteMostly1 1 /* mask for writemostly flag in above */
73 __u8 pad2
[64-57]; /* set to 0 when writing */
75 /* array state information - 64 bytes */
76 __u64 utime
; /* 40 bits second, 24 btes microseconds */
77 __u64 events
; /* incremented when superblock updated */
78 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
79 __u32 sb_csum
; /* checksum upto dev_roles[max_dev] */
80 __u32 max_dev
; /* size of dev_roles[] array to consider */
81 __u8 pad3
[64-32]; /* set to 0 when writing */
83 /* device state information. Indexed by dev_number.
85 * Note there are no per-device state flags. State information is rolled
86 * into the 'roles' value. If a device is spare or faulty, then it doesn't
87 * have a meaningful role.
89 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
92 #define MAX_SB_SIZE 4096
93 /* bitmap super size is 256, but we round up to a sector for alignment */
94 #define BM_SUPER_SIZE 512
95 #define MAX_DEVS ((int)(MAX_SB_SIZE - sizeof(struct mdp_superblock_1)) / 2)
96 #define SUPER1_SIZE (MAX_SB_SIZE + BM_SUPER_SIZE \
97 + sizeof(struct misc_dev_info))
99 struct misc_dev_info
{
103 /* feature_map bits */
104 #define MD_FEATURE_BITMAP_OFFSET 1
105 #define MD_FEATURE_RECOVERY_OFFSET 2 /* recovery_offset is present and
108 #define MD_FEATURE_RESHAPE_ACTIVE 4
110 #define MD_FEATURE_ALL (1|2|4)
113 #define offsetof(t,f) ((size_t)&(((t*)0)->f))
115 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
117 unsigned int disk_csum
, csum
;
118 unsigned long long newcsum
;
119 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
120 unsigned int *isuper
= (unsigned int*)sb
;
122 /* make sure I can count... */
123 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
124 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
125 sizeof(struct mdp_superblock_1
) != 256) {
126 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
129 disk_csum
= sb
->sb_csum
;
132 for (; size
>=4; size
-= 4 ) {
133 newcsum
+= __le32_to_cpu(*isuper
);
138 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
140 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
141 sb
->sb_csum
= disk_csum
;
142 return __cpu_to_le32(csum
);
146 * Information related to file descriptor used for aligned reads/writes.
147 * Cache the block size.
154 static void init_afd(struct align_fd
*afd
, int fd
)
158 if (ioctl(afd
->fd
, BLKSSZGET
, &afd
->blk_sz
) != 0)
162 static char abuf
[4096+4096];
163 static int aread(struct align_fd
*afd
, void *buf
, int len
)
166 * On devices with a 4K sector size, we need to read
167 * the full sector and copy relevant bits into
176 if (!bsize
|| bsize
> 4096 || len
> 4096) {
178 fprintf(stderr
, "WARNING - aread() called with "
179 "invalid block size\n");
182 b
= ROUND_UP_PTR((char *)abuf
, 4096);
184 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
186 n
= read(afd
->fd
, b
, iosize
);
189 lseek(afd
->fd
, len
- n
, 1);
196 static int awrite(struct align_fd
*afd
, void *buf
, int len
)
199 * On devices with a 4K sector size, we need to write
200 * the full sector. We pre-read if the sector is larger
202 * The address must be sector-aligned.
209 if (!bsize
|| bsize
> 4096 || len
> 4096) {
211 fprintf(stderr
, "WARNING - awrite() called with "
212 "invalid block size\n");
215 b
= ROUND_UP_PTR((char *)abuf
, 4096);
217 for (iosize
= 0; iosize
< len
; iosize
+= bsize
)
221 n
= read(afd
->fd
, b
, iosize
);
224 lseek(afd
->fd
, -n
, 1);
228 n
= write(afd
->fd
, b
, iosize
);
231 lseek(afd
->fd
, len
- n
, 1);
236 static void examine_super1(struct supertype
*st
, char *homehost
)
238 struct mdp_superblock_1
*sb
= st
->sb
;
245 int l
= homehost
? strlen(homehost
) : 0;
247 unsigned long long sb_offset
;
249 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
250 printf(" Version : 1");
251 sb_offset
= __le64_to_cpu(sb
->super_offset
);
254 else if (sb_offset
<= 8)
258 printf(" Feature Map : 0x%x\n", __le32_to_cpu(sb
->feature_map
));
259 printf(" Array UUID : ");
260 for (i
=0; i
<16; i
++) {
261 if ((i
&3)==0 && i
!= 0) printf(":");
262 printf("%02x", sb
->set_uuid
[i
]);
265 printf(" Name : %.32s", sb
->set_name
);
266 if (l
> 0 && l
< 32 &&
267 sb
->set_name
[l
] == ':' &&
268 strncmp(sb
->set_name
, homehost
, l
) == 0)
269 printf(" (local to host %s)", homehost
);
271 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
272 printf(" Creation Time : %.24s\n", ctime(&atime
));
273 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
274 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
275 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
277 printf(" Avail Dev Size : %llu%s\n",
278 (unsigned long long)__le64_to_cpu(sb
->data_size
),
279 human_size(__le64_to_cpu(sb
->data_size
)<<9));
280 if (__le32_to_cpu(sb
->level
) > 0) {
281 int ddsks
= 0, ddsks_denom
= 1;
282 switch(__le32_to_cpu(sb
->level
)) {
283 case 1: ddsks
=1;break;
285 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
286 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
288 layout
= __le32_to_cpu(sb
->layout
);
289 ddsks
= __le32_to_cpu(sb
->raid_disks
);
290 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
293 long long asize
= __le64_to_cpu(sb
->size
);
294 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
295 printf(" Array Size : %llu%s\n",
296 asize
>> 10, human_size(asize
));
298 if (sb
->size
!= sb
->data_size
)
299 printf(" Used Dev Size : %llu%s\n",
300 (unsigned long long)__le64_to_cpu(sb
->size
),
301 human_size(__le64_to_cpu(sb
->size
)<<9));
304 printf(" Data Offset : %llu sectors\n",
305 (unsigned long long)__le64_to_cpu(sb
->data_offset
));
306 printf(" Super Offset : %llu sectors\n",
307 (unsigned long long)__le64_to_cpu(sb
->super_offset
));
308 if (__le32_to_cpu(sb
->feature_map
) & MD_FEATURE_RECOVERY_OFFSET
)
309 printf("Recovery Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->recovery_offset
));
310 printf(" State : %s\n", (__le64_to_cpu(sb
->resync_offset
)+1)? "active":"clean");
311 printf(" Device UUID : ");
312 for (i
=0; i
<16; i
++) {
313 if ((i
&3)==0 && i
!= 0) printf(":");
314 printf("%02x", sb
->device_uuid
[i
]);
318 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
319 printf("Internal Bitmap : %ld sectors from superblock\n",
320 (long)(int32_t)__le32_to_cpu(sb
->bitmap_offset
));
322 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
323 printf(" Reshape pos'n : %llu%s\n", (unsigned long long)__le64_to_cpu(sb
->reshape_position
)/2,
324 human_size(__le64_to_cpu(sb
->reshape_position
)<<9));
325 if (__le32_to_cpu(sb
->delta_disks
)) {
326 printf(" Delta Devices : %d", __le32_to_cpu(sb
->delta_disks
));
327 printf(" (%d->%d)\n",
328 __le32_to_cpu(sb
->raid_disks
)-__le32_to_cpu(sb
->delta_disks
),
329 __le32_to_cpu(sb
->raid_disks
));
330 if ((int)__le32_to_cpu(sb
->delta_disks
) < 0)
331 delta_extra
= -__le32_to_cpu(sb
->delta_disks
);
333 if (__le32_to_cpu(sb
->new_level
) != __le32_to_cpu(sb
->level
)) {
334 c
= map_num(pers
, __le32_to_cpu(sb
->new_level
));
335 printf(" New Level : %s\n", c
?c
:"-unknown-");
337 if (__le32_to_cpu(sb
->new_layout
) != __le32_to_cpu(sb
->layout
)) {
338 if (__le32_to_cpu(sb
->level
) == 5) {
339 c
= map_num(r5layout
, __le32_to_cpu(sb
->new_layout
));
340 printf(" New Layout : %s\n", c
?c
:"-unknown-");
342 if (__le32_to_cpu(sb
->level
) == 6) {
343 c
= map_num(r6layout
, __le32_to_cpu(sb
->new_layout
));
344 printf(" New Layout : %s\n", c
?c
:"-unknown-");
346 if (__le32_to_cpu(sb
->level
) == 10) {
347 printf(" New Layout :");
348 print_r10_layout(__le32_to_cpu(sb
->new_layout
));
352 if (__le32_to_cpu(sb
->new_chunk
) != __le32_to_cpu(sb
->chunksize
))
353 printf(" New Chunksize : %dK\n", __le32_to_cpu(sb
->new_chunk
)/2);
358 if (sb
->devflags
& WriteMostly1
)
359 printf(" write-mostly");
363 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
364 printf(" Update Time : %.24s\n", ctime(&atime
));
366 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
367 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
369 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
370 __le32_to_cpu(calc_sb_1_csum(sb
)));
371 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
373 if (__le32_to_cpu(sb
->level
) == 5) {
374 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
375 printf(" Layout : %s\n", c
?c
:"-unknown-");
377 if (__le32_to_cpu(sb
->level
) == 6) {
378 c
= map_num(r6layout
, __le32_to_cpu(sb
->layout
));
379 printf(" Layout : %s\n", c
?c
:"-unknown-");
381 if (__le32_to_cpu(sb
->level
) == 10) {
382 int lo
= __le32_to_cpu(sb
->layout
);
384 print_r10_layout(lo
);
387 switch(__le32_to_cpu(sb
->level
)) {
393 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
396 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
)/2);
402 /* This turns out to just be confusing */
403 printf(" Array Slot : %d (", __le32_to_cpu(sb
->dev_number
));
404 for (i
= __le32_to_cpu(sb
->max_dev
); i
> 0 ; i
--)
405 if (__le16_to_cpu(sb
->dev_roles
[i
-1]) != 0xffff)
407 for (d
=0; d
< i
; d
++) {
408 int role
= __le16_to_cpu(sb
->dev_roles
[d
]);
410 if (role
== 0xffff) printf("empty");
411 else if(role
== 0xfffe) printf("failed");
412 else printf("%d", role
);
416 printf(" Device Role : ");
417 d
= __le32_to_cpu(sb
->dev_number
);
418 if (d
< __le32_to_cpu(sb
->max_dev
))
419 role
= __le16_to_cpu(sb
->dev_roles
[d
]);
425 printf("Active device %d\n", role
);
427 printf(" Array State : ");
428 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
) + delta_extra
; d
++) {
431 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
432 unsigned int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
436 if (cnt
> 1) printf("?");
437 else if (cnt
== 1) printf("A");
441 /* This is confusing too */
443 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
444 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
448 if (faulty
) printf(" %d failed", faulty
);
450 printf(" ('A' == active, '.' == missing)");
455 static void brief_examine_super1(struct supertype
*st
, int verbose
)
457 struct mdp_superblock_1
*sb
= st
->sb
;
459 unsigned long long sb_offset
;
461 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
463 nm
= strchr(sb
->set_name
, ':');
466 else if (sb
->set_name
[0])
471 printf("ARRAY%s%s", nm
? " /dev/md/":"", nm
);
473 printf(" level=%s", c
);
474 sb_offset
= __le64_to_cpu(sb
->super_offset
);
476 printf(" metadata=1.1 ");
477 else if (sb_offset
<= 8)
478 printf(" metadata=1.2 ");
480 printf(" metadata=1.0 ");
482 printf("num-devices=%d ", __le32_to_cpu(sb
->raid_disks
));
484 for (i
=0; i
<16; i
++) {
485 if ((i
&3)==0 && i
!= 0) printf(":");
486 printf("%02x", sb
->set_uuid
[i
]);
489 printf(" name=%.32s", sb
->set_name
);
493 static void export_examine_super1(struct supertype
*st
)
495 struct mdp_superblock_1
*sb
= st
->sb
;
500 printf("MD_LEVEL=%s\n", map_num(pers
, __le32_to_cpu(sb
->level
)));
501 printf("MD_DEVICES=%d\n", __le32_to_cpu(sb
->raid_disks
));
503 if (sb
->set_name
[i
] == '\n' ||
504 sb
->set_name
[i
] == '\0') {
509 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
510 if (__le32_to_cpu(sb
->level
) > 0) {
511 int ddsks
= 0, ddsks_denom
= 1;
512 switch(__le32_to_cpu(sb
->level
)) {
513 case 1: ddsks
=1;break;
515 case 5: ddsks
= __le32_to_cpu(sb
->raid_disks
)-1; break;
516 case 6: ddsks
= __le32_to_cpu(sb
->raid_disks
)-2; break;
518 layout
= __le32_to_cpu(sb
->layout
);
519 ddsks
= __le32_to_cpu(sb
->raid_disks
);
520 ddsks_denom
= (layout
&255) * ((layout
>>8)&255);
523 long long asize
= __le64_to_cpu(sb
->size
);
524 asize
= (asize
<< 9) * ddsks
/ ddsks_denom
;
525 printf("MD_ARRAY_SIZE=%s\n",human_size_brief(asize
,JEDEC
));
529 for (i
=0; i
<16; i
++) {
530 if ((i
&3)==0 && i
!= 0) printf(":");
531 printf("%02x", sb
->set_uuid
[i
]);
534 printf("MD_UPDATE_TIME=%llu\n",
535 __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
);
536 printf("MD_DEV_UUID=");
537 for (i
=0; i
<16; i
++) {
538 if ((i
&3)==0 && i
!= 0) printf(":");
539 printf("%02x", sb
->device_uuid
[i
]);
542 printf("MD_EVENTS=%llu\n",
543 (unsigned long long)__le64_to_cpu(sb
->events
));
546 static void detail_super1(struct supertype
*st
, char *homehost
)
548 struct mdp_superblock_1
*sb
= st
->sb
;
550 int l
= homehost
? strlen(homehost
) : 0;
552 printf(" Name : %.32s", sb
->set_name
);
553 if (l
> 0 && l
< 32 &&
554 sb
->set_name
[l
] == ':' &&
555 strncmp(sb
->set_name
, homehost
, l
) == 0)
556 printf(" (local to host %s)", homehost
);
557 printf("\n UUID : ");
558 for (i
=0; i
<16; i
++) {
559 if ((i
&3)==0 && i
!= 0) printf(":");
560 printf("%02x", sb
->set_uuid
[i
]);
562 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
565 static void brief_detail_super1(struct supertype
*st
)
567 struct mdp_superblock_1
*sb
= st
->sb
;
571 printf(" name=%.32s", sb
->set_name
);
573 for (i
=0; i
<16; i
++) {
574 if ((i
&3)==0 && i
!= 0) printf(":");
575 printf("%02x", sb
->set_uuid
[i
]);
579 static void export_detail_super1(struct supertype
*st
)
581 struct mdp_superblock_1
*sb
= st
->sb
;
586 if (sb
->set_name
[i
] == '\n' ||
587 sb
->set_name
[i
] == '\0') {
592 printf("MD_NAME=%.*s\n", len
, sb
->set_name
);
597 static int match_home1(struct supertype
*st
, char *homehost
)
599 struct mdp_superblock_1
*sb
= st
->sb
;
600 int l
= homehost
? strlen(homehost
) : 0;
602 return (l
> 0 && l
< 32 &&
603 sb
->set_name
[l
] == ':' &&
604 strncmp(sb
->set_name
, homehost
, l
) == 0);
607 static void uuid_from_super1(struct supertype
*st
, int uuid
[4])
609 struct mdp_superblock_1
*super
= st
->sb
;
610 char *cuuid
= (char*)uuid
;
613 cuuid
[i
] = super
->set_uuid
[i
];
616 static void getinfo_super1(struct supertype
*st
, struct mdinfo
*info
, char *map
)
618 struct mdp_superblock_1
*sb
= st
->sb
;
622 unsigned int map_disks
= info
->array
.raid_disks
;
624 memset(info
, 0, sizeof(*info
));
625 info
->array
.major_version
= 1;
626 info
->array
.minor_version
= st
->minor_version
;
627 info
->array
.patch_version
= 0;
628 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
629 info
->array
.level
= __le32_to_cpu(sb
->level
);
630 info
->array
.layout
= __le32_to_cpu(sb
->layout
);
631 info
->array
.md_minor
= -1;
632 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
633 info
->array
.utime
= __le64_to_cpu(sb
->utime
);
634 info
->array
.chunk_size
= __le32_to_cpu(sb
->chunksize
)*512;
636 (__le64_to_cpu(sb
->resync_offset
) == MaxSector
)
639 info
->data_offset
= __le64_to_cpu(sb
->data_offset
);
640 info
->component_size
= __le64_to_cpu(sb
->size
);
641 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_BITMAP_OFFSET
))
642 info
->bitmap_offset
= (int32_t)__le32_to_cpu(sb
->bitmap_offset
);
644 info
->disk
.major
= 0;
645 info
->disk
.minor
= 0;
646 info
->disk
.number
= __le32_to_cpu(sb
->dev_number
);
647 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
648 __le32_to_cpu(sb
->dev_number
) >= MAX_DEVS
)
651 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
653 info
->disk
.raid_disk
= -1;
656 info
->disk
.state
= 0; /* spare: not active, not sync, not faulty */
659 info
->disk
.state
= 1; /* faulty */
662 info
->disk
.state
= 6; /* active and in sync */
663 info
->disk
.raid_disk
= role
;
665 if (sb
->devflags
& WriteMostly1
)
666 info
->disk
.state
|= (1 << MD_DISK_WRITEMOSTLY
);
667 info
->events
= __le64_to_cpu(sb
->events
);
668 sprintf(info
->text_version
, "1.%d", st
->minor_version
);
669 info
->safe_mode_delay
= 200;
671 memcpy(info
->uuid
, sb
->set_uuid
, 16);
673 strncpy(info
->name
, sb
->set_name
, 32);
676 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RECOVERY_OFFSET
))
677 info
->recovery_start
= __le32_to_cpu(sb
->recovery_offset
);
679 info
->recovery_start
= MaxSector
;
681 if (sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
)) {
682 info
->reshape_active
= 1;
683 info
->reshape_progress
= __le64_to_cpu(sb
->reshape_position
);
684 info
->new_level
= __le32_to_cpu(sb
->new_level
);
685 info
->delta_disks
= __le32_to_cpu(sb
->delta_disks
);
686 info
->new_layout
= __le32_to_cpu(sb
->new_layout
);
687 info
->new_chunk
= __le32_to_cpu(sb
->new_chunk
)<<9;
688 if (info
->delta_disks
< 0)
689 info
->array
.raid_disks
-= info
->delta_disks
;
691 info
->reshape_active
= 0;
693 info
->recovery_blocked
= info
->reshape_active
;
696 for (i
=0; i
<map_disks
; i
++)
698 for (i
= 0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
699 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
700 if (/*role == 0xFFFF || */role
< (unsigned) info
->array
.raid_disks
) {
702 if (map
&& role
< map_disks
)
707 info
->array
.working_disks
= working
;
710 static struct mdinfo
*container_content1(struct supertype
*st
, char *subarray
)
717 info
= xmalloc(sizeof(*info
));
718 getinfo_super1(st
, info
, NULL
);
722 static int update_super1(struct supertype
*st
, struct mdinfo
*info
,
724 char *devname
, int verbose
,
725 int uuid_set
, char *homehost
)
727 /* NOTE: for 'assemble' and 'force' we need to return non-zero
728 * if any change was made. For others, the return value is
732 struct mdp_superblock_1
*sb
= st
->sb
;
734 if (strcmp(update
, "force-one")==0) {
735 /* Not enough devices for a working array,
736 * so bring this one up-to-date
738 if (sb
->events
!= __cpu_to_le64(info
->events
))
740 sb
->events
= __cpu_to_le64(info
->events
);
741 } else if (strcmp(update
, "force-array")==0) {
742 /* Degraded array and 'force' requests to
743 * maybe need to mark it 'clean'.
745 switch(__le32_to_cpu(sb
->level
)) {
746 case 5: case 4: case 6:
747 /* need to force clean */
748 if (sb
->resync_offset
!= MaxSector
)
750 sb
->resync_offset
= MaxSector
;
752 } else if (strcmp(update
, "assemble")==0) {
753 int d
= info
->disk
.number
;
755 if (info
->disk
.state
== 6)
756 want
= info
->disk
.raid_disk
;
759 if (sb
->dev_roles
[d
] != __cpu_to_le16(want
)) {
760 sb
->dev_roles
[d
] = __cpu_to_le16(want
);
763 if (info
->reshape_active
&&
764 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
765 info
->delta_disks
>= 0 &&
766 info
->reshape_progress
< __le64_to_cpu(sb
->reshape_position
)) {
767 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
770 if (info
->reshape_active
&&
771 sb
->feature_map
& __le32_to_cpu(MD_FEATURE_RESHAPE_ACTIVE
) &&
772 info
->delta_disks
< 0 &&
773 info
->reshape_progress
> __le64_to_cpu(sb
->reshape_position
)) {
774 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
777 } else if (strcmp(update
, "linear-grow-new") == 0) {
780 unsigned int max
= __le32_to_cpu(sb
->max_dev
);
782 for (i
=0 ; i
< max
; i
++)
783 if (__le16_to_cpu(sb
->dev_roles
[i
]) >= 0xfffe)
785 sb
->dev_number
= __cpu_to_le32(i
);
786 info
->disk
.number
= i
;
787 if (max
>= __le32_to_cpu(sb
->max_dev
))
788 sb
->max_dev
= __cpu_to_le32(max
+1);
790 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
791 read(rfd
, sb
->device_uuid
, 16) != 16) {
792 __u32 r
[4] = {random(), random(), random(), random()};
793 memcpy(sb
->device_uuid
, r
, 16);
799 __cpu_to_le16(info
->disk
.raid_disk
);
801 fd
= open(devname
, O_RDONLY
);
803 unsigned long long ds
;
804 get_dev_size(fd
, devname
, &ds
);
807 if (__le64_to_cpu(sb
->super_offset
) <
808 __le64_to_cpu(sb
->data_offset
)) {
809 sb
->data_size
= __cpu_to_le64(
810 ds
- __le64_to_cpu(sb
->data_offset
));
813 ds
&= ~(unsigned long long)(4*2-1);
814 sb
->super_offset
= __cpu_to_le64(ds
);
815 sb
->data_size
= __cpu_to_le64(
816 ds
- __le64_to_cpu(sb
->data_offset
));
819 } else if (strcmp(update
, "linear-grow-update") == 0) {
820 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
821 sb
->dev_roles
[info
->disk
.number
] =
822 __cpu_to_le16(info
->disk
.raid_disk
);
823 } else if (strcmp(update
, "resync") == 0) {
824 /* make sure resync happens */
825 sb
->resync_offset
= 0ULL;
826 } else if (strcmp(update
, "uuid") == 0) {
827 copy_uuid(sb
->set_uuid
, info
->uuid
, super1
.swapuuid
);
829 if (__le32_to_cpu(sb
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
830 struct bitmap_super_s
*bm
;
831 bm
= (struct bitmap_super_s
*)(st
->sb
+MAX_SB_SIZE
);
832 memcpy(bm
->uuid
, sb
->set_uuid
, 16);
834 } else if (strcmp(update
, "no-bitmap") == 0) {
835 sb
->feature_map
&= ~__cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
);
836 } else if (strcmp(update
, "homehost") == 0 &&
840 c
= strchr(sb
->set_name
, ':');
842 strncpy(info
->name
, c
+1, 31 - (c
-sb
->set_name
));
844 strncpy(info
->name
, sb
->set_name
, 32);
846 } else if (strcmp(update
, "name") == 0) {
847 if (info
->name
[0] == 0)
848 sprintf(info
->name
, "%d", info
->array
.md_minor
);
849 memset(sb
->set_name
, 0, sizeof(sb
->set_name
));
851 strchr(info
->name
, ':') == NULL
&&
852 strlen(homehost
)+1+strlen(info
->name
) < 32) {
853 strcpy(sb
->set_name
, homehost
);
854 strcat(sb
->set_name
, ":");
855 strcat(sb
->set_name
, info
->name
);
857 strcpy(sb
->set_name
, info
->name
);
858 } else if (strcmp(update
, "devicesize") == 0 &&
859 __le64_to_cpu(sb
->super_offset
) <
860 __le64_to_cpu(sb
->data_offset
)) {
861 /* set data_size to device size less data_offset */
862 struct misc_dev_info
*misc
= (struct misc_dev_info
*)
863 (st
->sb
+ MAX_SB_SIZE
+ BM_SUPER_SIZE
);
864 printf("Size was %llu\n", (unsigned long long)
865 __le64_to_cpu(sb
->data_size
));
866 sb
->data_size
= __cpu_to_le64(
867 misc
->device_size
- __le64_to_cpu(sb
->data_offset
));
868 printf("Size is %llu\n", (unsigned long long)
869 __le64_to_cpu(sb
->data_size
));
870 } else if (strcmp(update
, "_reshape_progress")==0)
871 sb
->reshape_position
= __cpu_to_le64(info
->reshape_progress
);
872 else if (strcmp(update
, "writemostly")==0)
873 sb
->devflags
|= WriteMostly1
;
874 else if (strcmp(update
, "readwrite")==0)
875 sb
->devflags
&= ~WriteMostly1
;
879 sb
->sb_csum
= calc_sb_1_csum(sb
);
883 static int init_super1(struct supertype
*st
, mdu_array_info_t
*info
,
884 unsigned long long size
, char *name
, char *homehost
, int *uuid
)
886 struct mdp_superblock_1
*sb
;
892 if (posix_memalign((void**)&sb
, 4096, SUPER1_SIZE
) != 0) {
893 pr_err("%s could not allocate superblock\n", __func__
);
896 memset(sb
, 0, SUPER1_SIZE
);
900 /* zeroing superblock */
904 spares
= info
->working_disks
- info
->active_disks
;
905 if (info
->raid_disks
+ spares
> MAX_DEVS
) {
906 pr_err("too many devices requested: %d+%d > %d\n",
907 info
->raid_disks
, spares
, MAX_DEVS
);
911 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
912 sb
->major_version
= __cpu_to_le32(1);
917 copy_uuid(sb
->set_uuid
, uuid
, super1
.swapuuid
);
919 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
920 read(rfd
, sb
->set_uuid
, 16) != 16) {
921 __u32 r
[4] = {random(), random(), random(), random()};
922 memcpy(sb
->set_uuid
, r
, 16);
924 if (rfd
>= 0) close(rfd
);
927 if (name
== NULL
|| *name
== 0) {
928 sprintf(defname
, "%d", info
->md_minor
);
932 strchr(name
, ':')== NULL
&&
933 strlen(homehost
)+1+strlen(name
) < 32) {
934 strcpy(sb
->set_name
, homehost
);
935 strcat(sb
->set_name
, ":");
936 strcat(sb
->set_name
, name
);
938 strcpy(sb
->set_name
, name
);
940 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
941 sb
->level
= __cpu_to_le32(info
->level
);
942 sb
->layout
= __cpu_to_le32(info
->layout
);
943 sb
->size
= __cpu_to_le64(size
*2ULL);
944 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
945 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
947 sb
->data_offset
= __cpu_to_le64(0);
948 sb
->data_size
= __cpu_to_le64(0);
949 sb
->super_offset
= __cpu_to_le64(0);
950 sb
->recovery_offset
= __cpu_to_le64(0);
952 sb
->utime
= sb
->ctime
;
953 sb
->events
= __cpu_to_le64(1);
954 if (info
->state
& (1<<MD_SB_CLEAN
))
955 sb
->resync_offset
= MaxSector
;
957 sb
->resync_offset
= 0;
958 sbsize
= sizeof(struct mdp_superblock_1
) + 2 * (info
->raid_disks
+ spares
);
959 sbsize
= ROUND_UP(sbsize
, 512);
960 sb
->max_dev
= __cpu_to_le32((sbsize
- sizeof(struct mdp_superblock_1
)) / 2);
962 memset(sb
->dev_roles
, 0xff, MAX_SB_SIZE
- sizeof(struct mdp_superblock_1
));
970 mdu_disk_info_t disk
;
971 struct devinfo
*next
;
974 /* Add a device to the superblock being created */
975 static int add_to_super1(struct supertype
*st
, mdu_disk_info_t
*dk
,
976 int fd
, char *devname
)
978 struct mdp_superblock_1
*sb
= st
->sb
;
979 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
980 struct devinfo
*di
, **dip
;
982 if ((dk
->state
& 6) == 6) /* active, sync */
983 *rp
= __cpu_to_le16(dk
->raid_disk
);
984 else if ((dk
->state
& ~2) == 0) /* active or idle -> spare */
989 if (dk
->number
>= (int)__le32_to_cpu(sb
->max_dev
) &&
990 __le32_to_cpu(sb
->max_dev
) < MAX_DEVS
)
991 sb
->max_dev
= __cpu_to_le32(dk
->number
+1);
993 sb
->dev_number
= __cpu_to_le32(dk
->number
);
994 sb
->devflags
= 0; /* don't copy another disks flags */
995 sb
->sb_csum
= calc_sb_1_csum(sb
);
997 dip
= (struct devinfo
**)&st
->info
;
1000 di
= xmalloc(sizeof(struct devinfo
));
1002 di
->devname
= devname
;
1011 static void locate_bitmap1(struct supertype
*st
, int fd
);
1013 static int store_super1(struct supertype
*st
, int fd
)
1015 struct mdp_superblock_1
*sb
= st
->sb
;
1016 unsigned long long sb_offset
;
1017 struct align_fd afd
;
1019 unsigned long long dsize
;
1021 if (!get_dev_size(fd
, NULL
, &dsize
))
1032 * Calculate the position of the superblock.
1033 * It is always aligned to a 4K boundary and
1034 * depending on minor_version, it can be:
1035 * 0: At least 8K, but less than 12K, from end of device
1036 * 1: At start of device
1037 * 2: 4K from start of device.
1039 switch(st
->minor_version
) {
1043 sb_offset
&= ~(4*2-1);
1057 if (sb_offset
!= __le64_to_cpu(sb
->super_offset
) &&
1058 0 != __le64_to_cpu(sb
->super_offset
)
1060 pr_err("internal error - sb_offset is wrong\n");
1064 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL)
1067 sbsize
= ROUND_UP(sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
), 512);
1069 if (awrite(&afd
, sb
, sbsize
) != sbsize
)
1072 if (sb
->feature_map
& __cpu_to_le32(MD_FEATURE_BITMAP_OFFSET
)) {
1073 struct bitmap_super_s
*bm
= (struct bitmap_super_s
*)
1074 (((char*)sb
)+MAX_SB_SIZE
);
1075 if (__le32_to_cpu(bm
->magic
) == BITMAP_MAGIC
) {
1076 locate_bitmap1(st
, fd
);
1077 if (awrite(&afd
, bm
, sizeof(*bm
)) != sizeof(*bm
))
1085 static int load_super1(struct supertype
*st
, int fd
, char *devname
);
1087 static unsigned long choose_bm_space(unsigned long devsize
)
1089 /* if the device is bigger than 8Gig, save 64k for bitmap usage,
1090 * if bigger than 200Gig, save 128k
1091 * NOTE: result must be multiple of 4K else bad things happen
1092 * on 4K-sector devices.
1094 if (devsize
< 64*2) return 0;
1095 if (devsize
- 64*2 >= 200*1024*1024*2)
1097 if (devsize
- 4*2 > 8*1024*1024*2)
1102 static void free_super1(struct supertype
*st
);
1105 static int write_init_super1(struct supertype
*st
)
1107 struct mdp_superblock_1
*sb
= st
->sb
;
1108 struct supertype
*refst
;
1111 unsigned long long bm_space
;
1112 unsigned long long reserved
;
1114 unsigned long long dsize
, array_size
;
1115 unsigned long long sb_offset
, headroom
;
1117 for (di
= st
->info
; di
; di
= di
->next
) {
1118 if (di
->disk
.state
== 1)
1123 while (Kill(di
->devname
, NULL
, 0, -1, 1) == 0)
1126 sb
->dev_number
= __cpu_to_le32(di
->disk
.number
);
1127 if (di
->disk
.state
& (1<<MD_DISK_WRITEMOSTLY
))
1128 sb
->devflags
|= WriteMostly1
;
1130 sb
->devflags
&= ~WriteMostly1
;
1132 if ((rfd
= open("/dev/urandom", O_RDONLY
)) < 0 ||
1133 read(rfd
, sb
->device_uuid
, 16) != 16) {
1134 __u32 r
[4] = {random(), random(), random(), random()};
1135 memcpy(sb
->device_uuid
, r
, 16);
1142 refst
= dup_super(st
);
1143 if (load_super1(refst
, di
->fd
, NULL
)==0) {
1144 struct mdp_superblock_1
*refsb
= refst
->sb
;
1146 memcpy(sb
->device_uuid
, refsb
->device_uuid
, 16);
1147 if (memcmp(sb
->set_uuid
, refsb
->set_uuid
, 16)==0) {
1148 /* same array, so preserve events and
1150 sb
->events
= refsb
->events
;
1151 /* bugs in 2.6.17 and earlier mean the
1152 * dev_number chosen in Manage must be preserved
1154 if (get_linux_version() >= 2006018)
1155 sb
->dev_number
= refsb
->dev_number
;
1161 if (!get_dev_size(di
->fd
, NULL
, &dsize
)) {
1175 * Calculate the position of the superblock.
1176 * It is always aligned to a 4K boundary and
1177 * depending on minor_version, it can be:
1178 * 0: At least 8K, but less than 12K, from end of device
1179 * 1: At start of device
1180 * 2: 4K from start of device.
1181 * Depending on the array size, we might leave extra space
1184 array_size
= __le64_to_cpu(sb
->size
);
1185 /* work out how much space we left for a bitmap */
1186 bm_space
= choose_bm_space(array_size
);
1188 /* We try to leave 0.1% at the start for reshape
1189 * operations, but limit this to 128Meg (0.1% of 10Gig)
1190 * which is plenty for efficient reshapes
1192 headroom
= 128 * 1024 * 2;
1193 while (headroom
<< 10 > array_size
)
1196 switch(st
->minor_version
) {
1200 sb_offset
&= ~(4*2-1);
1201 sb
->super_offset
= __cpu_to_le64(sb_offset
);
1202 sb
->data_offset
= __cpu_to_le64(0);
1203 if (sb_offset
< array_size
+ bm_space
)
1204 bm_space
= sb_offset
- array_size
;
1205 sb
->data_size
= __cpu_to_le64(sb_offset
- bm_space
);
1208 sb
->super_offset
= __cpu_to_le64(0);
1209 reserved
= bm_space
+ 4*2;
1210 if (reserved
< headroom
)
1211 reserved
= headroom
;
1212 if (reserved
+ array_size
> dsize
)
1213 reserved
= dsize
- array_size
;
1214 /* Try for multiple of 1Meg so it is nicely aligned */
1215 #define ONE_MEG (2*1024)
1216 if (reserved
> ONE_MEG
)
1217 reserved
= (reserved
/ONE_MEG
) * ONE_MEG
;
1219 /* force 4K alignment */
1222 sb
->data_offset
= __cpu_to_le64(reserved
);
1223 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1227 sb
->super_offset
= __cpu_to_le64(4*2);
1228 if (4*2 + 4*2 + bm_space
+ array_size
1230 bm_space
= dsize
- array_size
1233 reserved
= bm_space
+ 4*2 + 4*2;
1234 if (reserved
< headroom
)
1235 reserved
= headroom
;
1236 if (reserved
+ array_size
> dsize
)
1237 reserved
= dsize
- array_size
;
1238 /* Try for multiple of 1Meg so it is nicely aligned */
1239 #define ONE_MEG (2*1024)
1240 if (reserved
> ONE_MEG
)
1241 reserved
= (reserved
/ONE_MEG
) * ONE_MEG
;
1243 /* force 4K alignment */
1246 sb
->data_offset
= __cpu_to_le64(reserved
);
1247 sb
->data_size
= __cpu_to_le64(dsize
- reserved
);
1250 pr_err("Failed to write invalid "
1251 "metadata format 1.%i to %s\n",
1252 st
->minor_version
, di
->devname
);
1257 sb
->sb_csum
= calc_sb_1_csum(sb
);
1258 rv
= store_super1(st
, di
->fd
);
1259 if (rv
== 0 && (__le32_to_cpu(sb
->feature_map
) & 1))
1260 rv
= st
->ss
->write_bitmap(st
, di
->fd
);
1268 pr_err("Failed to write metadata to %s\n",
1275 static int compare_super1(struct supertype
*st
, struct supertype
*tst
)
1279 * 0 same, or first was empty, and second was copied
1280 * 1 second had wrong number
1282 * 3 wrong other info
1284 struct mdp_superblock_1
*first
= st
->sb
;
1285 struct mdp_superblock_1
*second
= tst
->sb
;
1287 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
1289 if (second
->major_version
!= __cpu_to_le32(1))
1293 if (posix_memalign((void**)&first
, 4096, SUPER1_SIZE
) != 0) {
1294 pr_err("%s could not allocate superblock\n", __func__
);
1297 memcpy(first
, second
, SUPER1_SIZE
);
1301 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
1304 if (first
->ctime
!= second
->ctime
||
1305 first
->level
!= second
->level
||
1306 first
->layout
!= second
->layout
||
1307 first
->size
!= second
->size
||
1308 first
->chunksize
!= second
->chunksize
||
1309 first
->raid_disks
!= second
->raid_disks
)
1314 static int load_super1(struct supertype
*st
, int fd
, char *devname
)
1316 unsigned long long dsize
;
1317 unsigned long long sb_offset
;
1318 struct mdp_superblock_1
*super
;
1320 struct bitmap_super_s
*bsb
;
1321 struct misc_dev_info
*misc
;
1322 struct align_fd afd
;
1328 if (st
->ss
== NULL
|| st
->minor_version
== -1) {
1330 struct supertype tst
;
1331 __u64 bestctime
= 0;
1332 /* guess... choose latest ctime */
1333 memset(&tst
, 0, sizeof(tst
));
1335 for (tst
.minor_version
= 0; tst
.minor_version
<= 2 ; tst
.minor_version
++) {
1336 switch(load_super1(&tst
, fd
, devname
)) {
1337 case 0: super
= tst
.sb
;
1338 if (bestvers
== -1 ||
1339 bestctime
< __le64_to_cpu(super
->ctime
)) {
1340 bestvers
= tst
.minor_version
;
1341 bestctime
= __le64_to_cpu(super
->ctime
);
1346 case 1: return 1; /*bad device */
1347 case 2: break; /* bad, try next */
1350 if (bestvers
!= -1) {
1352 tst
.minor_version
= bestvers
;
1354 tst
.max_devs
= MAX_DEVS
;
1355 rv
= load_super1(&tst
, fd
, devname
);
1362 if (!get_dev_size(fd
, devname
, &dsize
))
1368 pr_err("%s is too small for md: size is %llu sectors.\n",
1374 * Calculate the position of the superblock.
1375 * It is always aligned to a 4K boundary and
1376 * depending on minor_version, it can be:
1377 * 0: At least 8K, but less than 12K, from end of device
1378 * 1: At start of device
1379 * 2: 4K from start of device.
1381 switch(st
->minor_version
) {
1385 sb_offset
&= ~(4*2-1);
1397 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
1400 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
1402 pr_err("Cannot seek to superblock on %s: %s\n",
1403 devname
, strerror(errno
));
1407 if (posix_memalign((void**)&super
, 4096, SUPER1_SIZE
) != 0) {
1408 pr_err("%s could not allocate superblock\n",
1413 if (aread(&afd
, super
, MAX_SB_SIZE
) != MAX_SB_SIZE
) {
1415 pr_err("Cannot read superblock on %s\n",
1421 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
1423 pr_err("No super block found on %s (Expected magic %08x, got %08x)\n",
1424 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
1429 if (__le32_to_cpu(super
->major_version
) != 1) {
1431 pr_err("Cannot interpret superblock on %s - version is %d\n",
1432 devname
, __le32_to_cpu(super
->major_version
));
1436 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
1438 pr_err("No superblock found on %s (super_offset is wrong)\n",
1445 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1447 misc
= (struct misc_dev_info
*) (((char*)super
)+MAX_SB_SIZE
+BM_SUPER_SIZE
);
1448 misc
->device_size
= dsize
;
1450 /* Now check on the bitmap superblock */
1451 if ((__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) == 0)
1453 /* Read the bitmap superblock and make sure it looks
1454 * valid. If it doesn't clear the bit. An --assemble --force
1455 * should get that written out.
1457 locate_bitmap1(st
, fd
);
1458 if (aread(&afd
, bsb
, 512) != 512)
1461 uuid_from_super1(st
, uuid
);
1462 if (__le32_to_cpu(bsb
->magic
) != BITMAP_MAGIC
||
1463 memcmp(bsb
->uuid
, uuid
, 16) != 0)
1468 super
->feature_map
= __cpu_to_le32(__le32_to_cpu(super
->feature_map
)
1469 & ~MD_FEATURE_BITMAP_OFFSET
);
1474 static struct supertype
*match_metadata_desc1(char *arg
)
1476 struct supertype
*st
= xcalloc(1, sizeof(*st
));
1478 st
->container_dev
= NoMdDev
;
1480 st
->max_devs
= MAX_DEVS
;
1482 /* leading zeros can be safely ignored. --detail generates them. */
1485 if (strcmp(arg
, "1.0") == 0 ||
1486 strcmp(arg
, "1.00") == 0) {
1487 st
->minor_version
= 0;
1490 if (strcmp(arg
, "1.1") == 0 ||
1491 strcmp(arg
, "1.01") == 0
1493 st
->minor_version
= 1;
1496 if (strcmp(arg
, "1.2") == 0 ||
1497 #ifndef DEFAULT_OLD_METADATA /* ifdef in super0.c */
1498 strcmp(arg
, "default") == 0 ||
1499 #endif /* DEFAULT_OLD_METADATA */
1500 strcmp(arg
, "1.02") == 0) {
1501 st
->minor_version
= 2;
1504 if (strcmp(arg
, "1") == 0 ||
1505 strcmp(arg
, "default") == 0) {
1506 st
->minor_version
= -1;
1514 /* find available size on device with this devsize, using
1515 * superblock type st, and reserving 'reserve' sectors for
1518 static __u64
avail_size1(struct supertype
*st
, __u64 devsize
)
1520 struct mdp_superblock_1
*super
= st
->sb
;
1525 /* creating: allow suitable space for bitmap */
1526 devsize
-= choose_bm_space(devsize
);
1528 else if (__le32_to_cpu(super
->feature_map
)&MD_FEATURE_BITMAP_OFFSET
) {
1529 /* hot-add. allow for actual size of bitmap */
1530 struct bitmap_super_s
*bsb
;
1531 bsb
= (struct bitmap_super_s
*)(((char*)super
)+MAX_SB_SIZE
);
1532 devsize
-= bitmap_sectors(bsb
);
1536 if (st
->minor_version
< 0)
1537 /* not specified, so time to set default */
1538 st
->minor_version
= 2;
1539 if (super
== NULL
&& st
->minor_version
> 0) {
1540 /* haven't committed to a size yet, so allow some
1541 * slack for space for reshape.
1542 * Limit slack to 128M, but aim for about 0.1%
1544 unsigned long long headroom
= 128*1024*2;
1545 while ((headroom
<< 10) > devsize
)
1547 devsize
-= headroom
;
1549 switch(st
->minor_version
) {
1552 return ((devsize
- 8*2 ) & ~(4*2-1));
1554 /* at start, 4K for superblock and possible bitmap */
1555 return devsize
- 4*2;
1557 /* 4k from start, 4K for superblock and possible bitmap */
1558 return devsize
- (4+4)*2;
1564 add_internal_bitmap1(struct supertype
*st
,
1565 int *chunkp
, int delay
, int write_behind
,
1566 unsigned long long size
,
1567 int may_change
, int major
)
1570 * If not may_change, then this is a 'Grow' without sysfs support for
1571 * bitmaps, and the bitmap must fit after the superblock at 1K offset.
1572 * If may_change, then this is create or a Grow with sysfs syupport,
1573 * and we can put the bitmap wherever we like.
1575 * size is in sectors, chunk is in bytes !!!
1578 unsigned long long bits
;
1579 unsigned long long max_bits
;
1580 unsigned long long min_chunk
;
1582 unsigned long long chunk
= *chunkp
;
1585 struct mdp_superblock_1
*sb
= st
->sb
;
1586 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
) + MAX_SB_SIZE
);
1589 if (__le64_to_cpu(sb
->data_size
) == 0)
1590 /* Must be creating the array, else data_size would be non-zero */
1592 switch(st
->minor_version
) {
1594 /* either 3K after the superblock (when hot-add),
1595 * or some amount of space before.
1598 /* We are creating array, so we *know* how much room has
1602 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1604 room
= __le64_to_cpu(sb
->super_offset
)
1605 - __le64_to_cpu(sb
->data_offset
)
1606 - __le64_to_cpu(sb
->data_size
);
1608 if (!may_change
|| (room
< 3*2 &&
1609 __le32_to_cpu(sb
->max_dev
) <= 384)) {
1613 offset
= 0; /* means movable offset */
1618 case 2: /* between superblock and data */
1621 room
= choose_bm_space(__le64_to_cpu(sb
->size
));
1623 room
= __le64_to_cpu(sb
->data_offset
)
1624 - __le64_to_cpu(sb
->super_offset
);
1626 room
-= 2; /* Leave 1K for superblock */
1629 room
-= 4*2; /* leave 4K for superblock */
1638 if (chunk
== UnSet
&& room
> 128*2)
1639 /* Limit to 128K of bitmap when chunk size not requested */
1642 max_bits
= (room
* 512 - sizeof(bitmap_super_t
)) * 8;
1644 min_chunk
= 4096; /* sub-page chunks don't work yet.. */
1645 bits
= (size
*512)/min_chunk
+1;
1646 while (bits
> max_bits
) {
1650 if (chunk
== UnSet
) {
1651 /* For practical purpose, 64Meg is a good
1652 * default chunk size for internal bitmaps.
1655 if (chunk
< 64*1024*1024)
1656 chunk
= 64*1024*1024;
1657 } else if (chunk
< min_chunk
)
1658 return 0; /* chunk size too small */
1659 if (chunk
== 0) /* rounding problem */
1663 /* start bitmap on a 4K boundary with enough space for
1666 bits
= (size
*512) / chunk
+ 1;
1667 room
= ((bits
+7)/8 + sizeof(bitmap_super_t
) +4095)/4096;
1668 room
*= 8; /* convert 4K blocks to sectors */
1672 sb
->bitmap_offset
= (int32_t)__cpu_to_le32(offset
);
1674 sb
->feature_map
= __cpu_to_le32(__le32_to_cpu(sb
->feature_map
)
1675 | MD_FEATURE_BITMAP_OFFSET
);
1676 memset(bms
, 0, sizeof(*bms
));
1677 bms
->magic
= __cpu_to_le32(BITMAP_MAGIC
);
1678 bms
->version
= __cpu_to_le32(major
);
1679 uuid_from_super1(st
, uuid
);
1680 memcpy(bms
->uuid
, uuid
, 16);
1681 bms
->chunksize
= __cpu_to_le32(chunk
);
1682 bms
->daemon_sleep
= __cpu_to_le32(delay
);
1683 bms
->sync_size
= __cpu_to_le64(size
);
1684 bms
->write_behind
= __cpu_to_le32(write_behind
);
1690 static void locate_bitmap1(struct supertype
*st
, int fd
)
1692 unsigned long long offset
;
1693 struct mdp_superblock_1
*sb
;
1697 if (st
->ss
->load_super(st
, fd
, NULL
))
1698 return; /* no error I hope... */
1703 offset
= __le64_to_cpu(sb
->super_offset
);
1704 offset
+= (int32_t) __le32_to_cpu(sb
->bitmap_offset
);
1707 lseek64(fd
, offset
<<9, 0);
1710 static int write_bitmap1(struct supertype
*st
, int fd
)
1712 struct mdp_superblock_1
*sb
= st
->sb
;
1713 bitmap_super_t
*bms
= (bitmap_super_t
*)(((char*)sb
)+MAX_SB_SIZE
);
1717 struct align_fd afd
;
1721 locate_bitmap1(st
, fd
);
1723 if (posix_memalign(&buf
, 4096, 4096))
1726 memset(buf
, 0xff, 4096);
1727 memcpy(buf
, (char *)bms
, sizeof(bitmap_super_t
));
1729 towrite
= __le64_to_cpu(bms
->sync_size
) / (__le32_to_cpu(bms
->chunksize
)>>9);
1730 towrite
= (towrite
+7) >> 3; /* bits to bytes */
1731 towrite
+= sizeof(bitmap_super_t
);
1732 towrite
= ROUND_UP(towrite
, 512);
1733 while (towrite
> 0) {
1737 n
= awrite(&afd
, buf
, n
);
1742 memset(buf
, 0xff, 4096);
1752 static void free_super1(struct supertype
*st
)
1757 struct devinfo
*di
= st
->info
;
1758 st
->info
= di
->next
;
1767 static int validate_geometry1(struct supertype
*st
, int level
,
1768 int layout
, int raiddisks
,
1769 int *chunk
, unsigned long long size
,
1770 char *subdev
, unsigned long long *freesize
,
1773 unsigned long long ldsize
;
1776 if (level
== LEVEL_CONTAINER
) {
1778 pr_err("1.x metadata does not support containers\n");
1781 if (chunk
&& *chunk
== UnSet
)
1782 *chunk
= DEFAULT_CHUNK
;
1787 fd
= open(subdev
, O_RDONLY
|O_EXCL
, 0);
1790 pr_err("super1.x cannot open %s: %s\n",
1791 subdev
, strerror(errno
));
1795 if (!get_dev_size(fd
, subdev
, &ldsize
)) {
1801 *freesize
= avail_size1(st
, ldsize
>> 9);
1804 #endif /* MDASSEMBLE */
1806 struct superswitch super1
= {
1808 .examine_super
= examine_super1
,
1809 .brief_examine_super
= brief_examine_super1
,
1810 .export_examine_super
= export_examine_super1
,
1811 .detail_super
= detail_super1
,
1812 .brief_detail_super
= brief_detail_super1
,
1813 .export_detail_super
= export_detail_super1
,
1814 .write_init_super
= write_init_super1
,
1815 .validate_geometry
= validate_geometry1
,
1816 .add_to_super
= add_to_super1
,
1818 .match_home
= match_home1
,
1819 .uuid_from_super
= uuid_from_super1
,
1820 .getinfo_super
= getinfo_super1
,
1821 .container_content
= container_content1
,
1822 .update_super
= update_super1
,
1823 .init_super
= init_super1
,
1824 .store_super
= store_super1
,
1825 .compare_super
= compare_super1
,
1826 .load_super
= load_super1
,
1827 .match_metadata_desc
= match_metadata_desc1
,
1828 .avail_size
= avail_size1
,
1829 .add_internal_bitmap
= add_internal_bitmap1
,
1830 .locate_bitmap
= locate_bitmap1
,
1831 .write_bitmap
= write_bitmap1
,
1832 .free_super
= free_super1
,
1833 #if __BYTE_ORDER == BIG_ENDIAN