]>
git.ipfire.org Git - thirdparty/mdadm.git/blob - super1.c
2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2004 Neil Brown <neilb@cse.unsw.edu.au>
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@cse.unsw.edu.au>
24 * School of Computer Science and Engineering
25 * The University of New South Wales
32 #include "asm/byteorder.h"
34 * The version-1 superblock :
35 * All numeric fields are little-endian.
37 * total size: 256 bytes plus 2 per device.
38 * 1K allows 384 devices.
40 struct mdp_superblock_1
{
41 /* constant array information - 128 bytes */
42 __u32 magic
; /* MD_SB_MAGIC: 0xa92b4efc - little endian */
43 __u32 major_version
; /* 1 */
44 __u32 feature_map
; /* 0 for now */
45 __u32 pad0
; /* always set to 0 when writing */
47 __u8 set_uuid
[16]; /* user-space generated. */
48 char set_name
[32]; /* set and interpreted by user-space */
50 __u64 ctime
; /* lo 40 bits are seconds, top 24 are microseconds or 0*/
51 __u32 level
; /* -4 (multipath), -1 (linear), 0,1,4,5 */
52 __u32 layout
; /* only for raid5 currently */
53 __u64 size
; /* used size of component devices, in 512byte sectors */
55 __u32 chunksize
; /* in 512byte sectors */
57 __u8 pad1
[128-96]; /* set to 0 when written */
59 /* constant this-device information - 64 bytes */
60 __u64 data_offset
; /* sector start of data, often 0 */
61 __u64 data_size
; /* sectors in this device that can be used for data */
62 __u64 super_offset
; /* sector start of this superblock */
63 __u64 recovery_offset
;/* sectors before this offset (from data_offset) have been recovered */
64 __u32 dev_number
; /* permanent identifier of this device - not role in raid */
65 __u32 cnt_corrected_read
; /* number of read errors that were corrected by re-writing */
66 __u8 device_uuid
[16]; /* user-space setable, ignored by kernel */
67 __u8 pad2
[64-56]; /* set to 0 when writing */
69 /* array state information - 64 bytes */
70 __u64 utime
; /* 40 bits second, 24 btes microseconds */
71 __u64 events
; /* incremented when superblock updated */
72 __u64 resync_offset
; /* data before this offset (from data_offset) known to be in sync */
73 __u32 sb_csum
; /* checksum upto devs[max_dev] */
74 __u32 max_dev
; /* size of devs[] array to consider */
75 __u8 pad3
[64-32]; /* set to 0 when writing */
77 /* device state information. Indexed by dev_number.
79 * Note there are no per-device state flags. State information is rolled
80 * into the 'roles' value. If a device is spare or faulty, then it doesn't
81 * have a meaningful role.
83 __u16 dev_roles
[0]; /* role in array, or 0xffff for a spare, or 0xfffe for faulty */
87 #define offsetof(t,f) ((int)&(((t*)0)->f))
89 static unsigned int calc_sb_1_csum(struct mdp_superblock_1
* sb
)
91 unsigned int disk_csum
, csum
;
92 unsigned long long newcsum
;
93 int size
= sizeof(*sb
) + __le32_to_cpu(sb
->max_dev
)*2;
94 unsigned int *isuper
= (unsigned int*)sb
;
97 /* make sure I can count... */
98 if (offsetof(struct mdp_superblock_1
,data_offset
) != 128 ||
99 offsetof(struct mdp_superblock_1
, utime
) != 192 ||
100 sizeof(struct mdp_superblock_1
) != 256) {
101 fprintf(stderr
, "WARNING - superblock isn't sized correctly\n");
104 disk_csum
= sb
->sb_csum
;
107 for (i
=0; size
>=4; size
-= 4 )
108 newcsum
+= __le32_to_cpu(*isuper
++);
111 newcsum
+= __le16_to_cpu(*(unsigned short*) isuper
);
113 csum
= (newcsum
& 0xffffffff) + (newcsum
>> 32);
114 sb
->sb_csum
= disk_csum
;
119 static void examine_super1(void *sbv
)
121 struct mdp_superblock_1
*sb
= sbv
;
128 printf(" Magic : %08x\n", __le32_to_cpu(sb
->magic
));
129 printf(" Version : %02d.%02d\n", 1, __le32_to_cpu(sb
->feature_map
));
130 printf(" Array UUID : ");
131 for (i
=0; i
<16; i
++) {
132 printf("%02x", sb
->set_uuid
[i
]);
133 if ((i
&3)==0 && i
!= 0) printf(":");
136 printf(" Name : %.32s\n", sb
->set_name
);
138 atime
= __le64_to_cpu(sb
->ctime
) & 0xFFFFFFFFFFULL
;
139 printf(" Creation Time : %.24s\n", ctime(&atime
));
140 c
=map_num(pers
, __le32_to_cpu(sb
->level
));
141 printf(" Raid Level : %s\n", c
?c
:"-unknown-");
142 printf(" Raid Devices : %d\n", __le32_to_cpu(sb
->raid_disks
));
144 printf(" Device Size : %llu%s\n", (unsigned long long)sb
->data_size
, human_size(sb
->data_size
<<9));
146 printf(" Data Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->data_offset
));
147 if (sb
->super_offset
)
148 printf(" Super Offset : %llu sectors\n", (unsigned long long)__le64_to_cpu(sb
->super_offset
));
149 printf(" Device UUID : ");
150 for (i
=0; i
<16; i
++) {
151 printf("%02x", sb
->set_uuid
[i
]);
152 if ((i
&3)==0 && i
!= 0) printf(":");
156 atime
= __le64_to_cpu(sb
->utime
) & 0xFFFFFFFFFFULL
;
157 printf(" Update Time : %.24s\n", ctime(&atime
));
159 if (calc_sb_1_csum(sb
) == sb
->sb_csum
)
160 printf(" Checksum : %x - correct\n", __le32_to_cpu(sb
->sb_csum
));
162 printf(" Checksum : %x - expected %x\n", __le32_to_cpu(sb
->sb_csum
),
163 __le32_to_cpu(calc_sb_1_csum(sb
)));
164 printf(" Events : %llu\n", (unsigned long long)__le64_to_cpu(sb
->events
));
166 if (__le32_to_cpu(sb
->level
) == 5) {
167 c
= map_num(r5layout
, __le32_to_cpu(sb
->layout
));
168 printf(" Layout : %s\n", c
?c
:"-unknown-");
170 switch(__le32_to_cpu(sb
->level
)) {
174 printf(" Chunk Size : %dK\n", __le32_to_cpu(sb
->chunksize
/2));
177 printf(" Rounding : %dK\n", __le32_to_cpu(sb
->chunksize
/2));
182 printf(" Array State : ");
183 for (d
=0; d
<__le32_to_cpu(sb
->raid_disks
); d
++) {
187 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
188 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
190 if (i
== __le32_to_cpu(sb
->dev_number
))
195 if (cnt
> 1) printf("?");
196 else if (cnt
== 1 && me
) printf("U");
197 else if (cnt
== 1) printf("u");
201 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
202 int role
= __le16_to_cpu(sb
->dev_roles
[i
]);
204 case 0xFFFF: spares
++; break;
205 case 0xFFFE: faulty
++;
208 if (spares
) printf(" %d spares", spares
);
209 if (faulty
) printf(" %d failed", faulty
);
214 static void brief_examine_super1(void *sbv
)
216 struct mdp_superblock_1
*sb
= sbv
;
219 char *c
=map_num(pers
, __le32_to_cpu(sb
->level
));
221 printf("ARRAY /dev/?? level=%s metadata=1 num-devices=%d UUID=",
222 c
?c
:"-unknown-", sb
->raid_disks
);
223 for (i
=0; i
<16; i
++) {
224 printf("%02x", sb
->set_uuid
[i
]);
225 if ((i
&3)==0 && i
!= 0) printf(":");
230 static void detail_super1(void *sbv
)
232 struct mdp_superblock_1
*sb
= sbv
;
236 for (i
=0; i
<16; i
++) {
237 printf("%02x", sb
->set_uuid
[i
]);
238 if ((i
&3)==0 && i
!= 0) printf(":");
240 printf("\n Events : %llu\n\n", (unsigned long long)__le64_to_cpu(sb
->events
));
243 static void brief_detail_super1(void *sbv
)
245 struct mdp_superblock_1
*sb
= sbv
;
249 for (i
=0; i
<16; i
++) {
250 printf("%02x", sb
->set_uuid
[i
]);
251 if ((i
&3)==0 && i
!= 0) printf(":");
255 static void uuid_from_super1(int uuid
[4], void * sbv
)
257 struct mdp_superblock_1
*super
= sbv
;
258 char *cuuid
= (char*)uuid
;
261 cuuid
[i
] = super
->set_uuid
[i
];
264 static void getinfo_super1(struct mdinfo
*info
, void *sbv
)
266 struct mdp_superblock_1
*sb
= sbv
;
271 info
->array
.major_version
= 1;
272 info
->array
.minor_version
= __le32_to_cpu(sb
->feature_map
);
273 info
->array
.patch_version
= 0;
274 info
->array
.raid_disks
= __le32_to_cpu(sb
->raid_disks
);
275 info
->array
.level
= __le32_to_cpu(sb
->level
);
276 info
->array
.md_minor
= -1;
277 info
->array
.ctime
= __le64_to_cpu(sb
->ctime
);
279 info
->disk
.major
= 0;
280 info
->disk
.minor
= 0;
282 if (__le32_to_cpu(sb
->dev_number
) >= __le32_to_cpu(sb
->max_dev
) ||
283 __le32_to_cpu(sb
->max_dev
) > 512)
286 role
= __le16_to_cpu(sb
->dev_roles
[__le32_to_cpu(sb
->dev_number
)]);
288 info
->disk
.raid_disk
= -1;
291 info
->disk
.state
= 2; /* spare: ACTIVE, not sync, not faulty */
294 info
->disk
.state
= 1; /* faulty */
297 info
->disk
.state
= 6; /* active and in sync */
298 info
->disk
.raid_disk
= role
;
300 info
->events
= __le64_to_cpu(sb
->events
);
302 memcpy(info
->uuid
, sb
->set_uuid
, 16);
304 for (i
=0; i
< __le32_to_cpu(sb
->max_dev
); i
++) {
305 role
= __le16_to_cpu(sb
->dev_roles
[i
]);
306 if (role
== 0xFFFF || role
< info
->array
.raid_disks
)
310 info
->array
.working_disks
= working
;
313 static int update_super1(struct mdinfo
*info
, void *sbv
, char *update
, char *devname
, int verbose
)
316 struct mdp_superblock_1
*sb
= sbv
;
318 if (strcmp(update
, "force")==0) {
319 sb
->events
= __cpu_to_le32(info
->events
);
320 switch(__le32_to_cpu(sb
->level
)) {
321 case 5: case 4: case 6:
322 /* need to force clean */
323 sb
->resync_offset
= ~0ULL;
326 if (strcmp(update
, "assemble")==0) {
327 int d
= info
->disk
.number
;
329 if (info
->disk
.state
== 6)
330 want
= __cpu_to_le32(info
->disk
.raid_disk
);
333 if (sb
->dev_roles
[d
] != want
) {
334 sb
->dev_roles
[d
] = want
;
339 if (strcmp(update
, "newdev") == 0) {
340 int d
= info
->disk
.number
;
341 memset(&sb
->disks
[d
], 0, sizeof(sb
->disks
[d
]));
342 sb
->disks
[d
].number
= d
;
343 sb
->disks
[d
].major
= info
->disk
.major
;
344 sb
->disks
[d
].minor
= info
->disk
.minor
;
345 sb
->disks
[d
].raid_disk
= info
->disk
.raid_disk
;
346 sb
->disks
[d
].state
= info
->disk
.state
;
347 sb
->this_disk
= sb
->disks
[d
];
350 if (strcmp(update
, "grow") == 0) {
351 sb
->raid_disks
= __cpu_to_le32(info
->array
.raid_disks
);
354 if (strcmp(update
, "resync") == 0) {
355 /* make sure resync happens */
356 sb
->resync_offset
= ~0ULL;
359 sb
->sb_csum
= calc_sb_1_csum(sb
);
364 static __u64
event_super1(void *sbv
)
366 struct mdp_superblock_1
*sb
= sbv
;
367 return __le64_to_cpu(sb
->events
);
370 static int init_super1(void **sbp
, mdu_array_info_t
*info
)
372 struct mdp_superblock_1
*sb
= malloc(1024);
376 if (info
->major_version
== -1)
377 /* zeroing superblock */
380 spares
= info
->working_disks
- info
->active_disks
;
381 if (info
->raid_disks
+ spares
> 384) {
382 fprintf(stderr
, Name
": too many devices requested: %d+%d > %d\n",
383 info
->raid_disks
, spares
, 384);
388 sb
->magic
= __cpu_to_le32(MD_SB_MAGIC
);
389 sb
->major_version
= __cpu_to_le32(1);
392 *(__u32
*)(sb
->set_uuid
) = random();
393 *(__u32
*)(sb
->set_uuid
+4) = random();
394 *(__u32
*)(sb
->set_uuid
+8) = random();
395 *(__u32
*)(sb
->set_uuid
+12) = random();
399 sb
->ctime
= __cpu_to_le64((unsigned long long)time(0));
400 sb
->level
= __cpu_to_le32(info
->level
);
401 sb
->layout
= __cpu_to_le32(info
->level
);
402 sb
->size
= __cpu_to_le64(info
->size
*2ULL);
403 sb
->chunksize
= __cpu_to_le32(info
->chunk_size
>>9);
404 sb
->raid_disks
= __cpu_to_le32(info
->raid_disks
);
406 sb
->data_offset
= __cpu_to_le64(0);
407 sb
->data_size
= __cpu_to_le64(0);
408 sb
->super_offset
= __cpu_to_le64(0);
409 sb
->recovery_offset
= __cpu_to_le64(0);
411 sb
->utime
= sb
->ctime
;
412 sb
->events
= __cpu_to_le64(1);
413 if (info
->state
& MD_SB_CLEAN
)
414 sb
->resync_offset
= ~0ULL;
416 sb
->resync_offset
= 0;
417 sb
->max_dev
= __cpu_to_le32((1024- sizeof(struct mdp_superblock_1
))/
418 sizeof(sb
->dev_roles
[0]));
419 memset(sb
->pad3
, 0, sizeof(sb
->pad3
));
421 memset(sb
->dev_roles
, 0xff, 1024 - sizeof(struct mdp_superblock_1
));
427 /* Add a device to the superblock being created */
428 static void add_to_super1(void *sbv
, mdu_disk_info_t
*dk
)
430 struct mdp_superblock_1
*sb
= sbv
;
431 __u16
*rp
= sb
->dev_roles
+ dk
->number
;
432 if (dk
->state
== 6) /* active, sync */
433 *rp
= __cpu_to_le16(dk
->raid_disk
);
434 else if (dk
->state
== 2) /* active -> spare */
440 static int store_super1(int fd
, void *sbv
)
442 struct mdp_superblock_1
*sb
= sbv
;
447 sb_offset
= __le64_to_cpu(sb
->super_offset
) << 9;
449 if (lseek64(fd
, sb_offset
, 0)< 0LL)
452 sbsize
= sizeof(*sb
) + 2 * __le32_to_cpu(sb
->max_dev
);
454 if (write(fd
, sb
, sbsize
) != sbsize
)
461 static int write_init_super1(struct supertype
*st
, void *sbv
, mdu_disk_info_t
*dinfo
, char *devname
)
463 struct mdp_superblock_1
*sb
= sbv
;
464 int fd
= open(devname
, O_RDWR
, O_EXCL
);
472 fprintf(stderr
, Name
": Failed to open %s to write superblock\n",
477 sb
->dev_number
= __cpu_to_le32(dinfo
->number
);
478 *(__u32
*)(sb
->device_uuid
) = random();
479 *(__u32
*)(sb
->device_uuid
+4) = random();
480 *(__u32
*)(sb
->device_uuid
+8) = random();
481 *(__u32
*)(sb
->device_uuid
+12) = random();
484 if (ioctl(fd
, BLKGETSIZE
, &size
)) {
496 * Calculate the position of the superblock.
497 * It is always aligned to a 4K boundary and
498 * depending on minor_version, it can be:
499 * 0: At least 8K, but less than 12K, from end of device
500 * 1: At start of device
501 * 2: 4K from start of device.
503 switch(st
->minor_version
) {
507 sb_offset
&= ~(4*2-1);
508 sb
->super_offset
= __cpu_to_le64(sb_offset
);
509 sb
->data_offset
= __cpu_to_le64(0);
510 sb
->data_size
= sb
->super_offset
;
513 sb
->super_offset
= __cpu_to_le64(0);
514 sb
->data_offset
= __cpu_to_le64(2);
515 sb
->data_size
= __cpu_to_le64(size
- 2);
519 sb
->super_offset
= __cpu_to_le64(sb_offset
);
520 sb
->data_offset
= __cpu_to_le64(sb_offset
+2);
521 sb
->data_size
= __cpu_to_le64(size
- 4*2 - 2);
528 sb
->sb_csum
= calc_sb_1_csum(sb
);
529 rv
= store_super1(fd
, sb
);
531 fprintf(stderr
, Name
": failed to write superblock to %s\n", devname
);
536 static int compare_super1(void **firstp
, void *secondv
)
540 * 0 same, or first was empty, and second was copied
541 * 1 second had wrong number
545 struct mdp_superblock_1
*first
= *firstp
;
546 struct mdp_superblock_1
*second
= secondv
;
548 if (second
->magic
!= __cpu_to_le32(MD_SB_MAGIC
))
550 if (second
->major_version
!= __cpu_to_le32(1))
554 first
= malloc(1024);
555 memcpy(first
, second
, 1024);
559 if (memcmp(first
->set_uuid
, second
->set_uuid
, 16)!= 0)
562 if (first
->ctime
!= second
->ctime
||
563 first
->level
!= second
->level
||
564 first
->layout
!= second
->layout
||
565 first
->size
!= second
->size
||
566 first
->chunksize
!= second
->chunksize
||
567 first
->raid_disks
!= second
->raid_disks
)
572 static int load_super1(struct supertype
*st
, int fd
, void **sbp
, char *devname
)
575 unsigned long long sb_offset
;
576 struct mdp_superblock_1
*super
;
580 if (st
->ss
== NULL
) {
583 /* guess... choose latest ctime */
585 for (st
->minor_version
= 0; st
->minor_version
<= 2 ; st
->minor_version
++) {
586 switch(load_super1(st
, fd
, sbp
, devname
)) {
587 case 0: super
= *sbp
;
588 if (bestvers
== -1 ||
589 bestctime
< __le64_to_cpu(super
->ctime
)) {
590 bestvers
= st
->minor_version
;
591 bestctime
= __le64_to_cpu(super
->ctime
);
596 case 1: st
->ss
= NULL
; return 1; /*bad device */
597 case 2: break; /* bad, try next */
600 if (bestvers
!= -1) {
602 st
->minor_version
= bestvers
;
604 rv
= load_super1(st
, fd
, sbp
, devname
);
605 if (rv
) st
->ss
= NULL
;
611 if (ioctl(fd
, BLKGETSIZE
, &size
)) {
613 fprintf(stderr
, Name
": cannot find device size for %s: %s\n",
614 devname
, strerror(errno
));
620 fprintf(stderr
, Name
": %s is too small for md: size is %lu sectors.\n",
626 * Calculate the position of the superblock.
627 * It is always aligned to a 4K boundary and
628 * depeding on minor_version, it can be:
629 * 0: At least 8K, but less than 12K, from end of device
630 * 1: At start of device
631 * 2: 4K from start of device.
633 switch(st
->minor_version
) {
637 sb_offset
&= ~(4*2-1);
649 ioctl(fd
, BLKFLSBUF
, 0); /* make sure we read current data */
652 if (lseek64(fd
, sb_offset
<< 9, 0)< 0LL) {
654 fprintf(stderr
, Name
": Cannot seek to superblock on %s: %s\n",
655 devname
, strerror(errno
));
659 super
= malloc(1024);
661 if (read(fd
, super
, 1024) != 1024) {
663 fprintf(stderr
, Name
": Cannot read superblock on %s\n",
669 if (__le32_to_cpu(super
->magic
) != MD_SB_MAGIC
) {
671 fprintf(stderr
, Name
": No super block found on %s (Expected magic %08x, got %08x)\n",
672 devname
, MD_SB_MAGIC
, __le32_to_cpu(super
->magic
));
677 if (__le32_to_cpu(super
->major_version
) != 1) {
679 fprintf(stderr
, Name
": Cannot interpret superblock on %s - version is %d\n",
680 devname
, __le32_to_cpu(super
->major_version
));
684 if (__le64_to_cpu(super
->super_offset
) != sb_offset
) {
686 fprintf(stderr
, Name
": No superblock found on %s (super_offset is wrong)\n",
696 static struct supertype
*match_metadata_desc1(char *arg
)
698 struct supertype
*st
= malloc(sizeof(*st
));
702 if (strcmp(arg
, "1") == 0 ||
703 strcmp(arg
, "1.0") == 0) {
704 st
->minor_version
= 0;
707 if (strcmp(arg
, "1.1") == 0) {
708 st
->minor_version
= 1;
711 if (strcmp(arg
, "1.2") == 0) {
712 st
->minor_version
= 2;
720 static __u64
avail_size1(__u64 devsize
)
725 return (devsize
- 8*2 ) & ~(4*2-1);
728 struct superswitch super1
= {
729 .examine_super
= examine_super1
,
730 .brief_examine_super
= brief_examine_super1
,
731 .detail_super
= detail_super1
,
732 .brief_detail_super
= brief_detail_super1
,
733 .uuid_from_super
= uuid_from_super1
,
734 .getinfo_super
= getinfo_super1
,
735 .update_super
= update_super1
,
736 .event_super
= event_super1
,
737 .init_super
= init_super1
,
738 .add_to_super
= add_to_super1
,
739 .store_super
= store_super1
,
740 .write_init_super
= write_init_super1
,
741 .compare_super
= compare_super1
,
742 .load_super
= load_super1
,
743 .match_metadata_desc
= match_metadata_desc1
,
744 .avail_size
= avail_size1
,