2 * mdadm - Intel(R) Matrix Storage Manager Support
4 * Copyright (C) 2002-2007 Intel Corporation
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
26 /* MPB == Metadata Parameter Block */
27 #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
28 #define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
29 #define MPB_VERSION_RAID0 "1.0.00"
30 #define MPB_VERSION_RAID1 "1.1.00"
31 #define MPB_VERSION_RAID5 "1.2.02"
32 #define MAX_SIGNATURE_LENGTH 32
33 #define MAX_RAID_SERIAL_LEN 16
34 #define MPB_SECTOR_CNT 418
35 #define IMSM_RESERVED_SECTORS 4096
37 /* Disk configuration info. */
38 #define IMSM_MAX_DEVICES 255
40 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
41 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
42 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
43 __u32 status
; /* 0xF0 - 0xF3 */
44 #define SPARE_DISK 0x01 /* Spare */
45 #define CONFIGURED_DISK 0x02 /* Member of some RaidDev */
46 #define FAILED_DISK 0x04 /* Permanent failure */
47 #define USABLE_DISK 0x08 /* Fully usable unless FAILED_DISK is set */
49 #define IMSM_DISK_FILLERS 5
50 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
53 /* RAID map configuration infos. */
55 __u32 pba_of_lba0
; /* start address of partition */
56 __u32 blocks_per_member
;/* blocks per member */
57 __u32 num_data_stripes
; /* number of data stripes */
58 __u16 blocks_per_strip
;
59 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
60 #define IMSM_T_STATE_NORMAL 0
61 #define IMSM_T_STATE_UNINITIALIZED 1
62 #define IMSM_T_STATE_DEGRADED 2 /* FIXME: is this correct? */
63 #define IMSM_T_STATE_FAILED 3 /* FIXME: is this correct? */
65 #define IMSM_T_RAID0 0
66 #define IMSM_T_RAID1 1
67 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
68 __u8 num_members
; /* number of member disks */
70 __u32 filler
[7]; /* expansion area */
71 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
73 } __attribute__ ((packed
));
77 __u8 migr_state
; /* Normal or Migrating */
78 __u8 migr_type
; /* Initializing, Rebuilding, ... */
82 struct imsm_map map
[1];
83 /* here comes another one if migr_state */
84 } __attribute__ ((packed
));
87 __u8 volume
[MAX_RAID_SERIAL_LEN
];
90 __u32 status
; /* Persistent RaidDev status */
91 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
92 #define IMSM_DEV_FILLERS 12
93 __u32 filler
[IMSM_DEV_FILLERS
];
95 } __attribute__ ((packed
));
98 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
99 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
100 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
101 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
102 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
103 __u32 reserved
[2]; /* 0x30 - 0x37 */
104 __u8 num_disks
; /* 0x38 Number of configured disks */
105 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
106 __u8 fill
[2]; /* 0x3A - 0x3B */
107 #define IMSM_FILLERS 39
108 __u32 filler
[IMSM_FILLERS
]; /* 0x3C - 0xD7 RAID_MPB_FILLERS */
109 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
110 /* here comes imsm_dev[num_raid_devs] */
111 } __attribute__ ((packed
));
114 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
117 static unsigned int sector_count(__u32 bytes
)
119 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
122 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
124 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
127 /* internal representation of IMSM metadata */
130 struct imsm_super
*mpb
;
133 int updates_pending
; /* count of pending updates for mdmon */
134 int creating_imsm
; /* flag to indicate container creation */
135 int current_vol
; /* index of raid device undergoing creation */
139 __u8 serial
[MAX_RAID_SERIAL_LEN
];
147 unsigned long long start
, size
;
150 /* definition of messages passed to imsm_process_update */
151 enum imsm_update_type
{
152 update_activate_spare
,
156 struct imsm_update_activate_spare
{
157 enum imsm_update_type type
;
161 struct imsm_update_activate_spare
*next
;
164 struct imsm_update_create_array
{
165 enum imsm_update_type type
;
170 static int imsm_env_devname_as_serial(void)
172 char *val
= getenv("IMSM_DEVNAME_AS_SERIAL");
174 if (val
&& atoi(val
) == 1)
181 static struct supertype
*match_metadata_desc_imsm(char *arg
)
183 struct supertype
*st
;
185 if (strcmp(arg
, "imsm") != 0 &&
186 strcmp(arg
, "default") != 0
190 st
= malloc(sizeof(*st
));
191 memset(st
, 0, sizeof(*st
));
192 st
->ss
= &super_imsm
;
193 st
->max_devs
= IMSM_MAX_DEVICES
;
194 st
->minor_version
= 0;
199 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
201 return &mpb
->sig
[MPB_SIG_LEN
];
204 static struct imsm_disk
*get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
206 if (index
> mpb
->num_disks
- 1)
208 return &mpb
->disk
[index
];
211 static __u32
gen_imsm_checksum(struct imsm_super
*mpb
)
213 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
214 __u32
*p
= (__u32
*) mpb
;
218 sum
+= __le32_to_cpu(*p
++);
220 return sum
- __le32_to_cpu(mpb
->check_sum
);
223 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
)
225 size_t size
= sizeof(*dev
);
227 /* each map has disk_ord_tbl[num_members - 1] additional space */
228 size
+= sizeof(__u32
) * (dev
->vol
.map
[0].num_members
- 1);
230 /* migrating means an additional map */
231 if (dev
->vol
.migr_state
) {
232 size
+= sizeof(struct imsm_map
);
233 size
+= sizeof(__u32
) * (dev
->vol
.map
[1].num_members
- 1);
239 static struct imsm_dev
*get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
245 if (index
> mpb
->num_raid_devs
- 1)
248 /* devices start after all disks */
249 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
251 for (i
= 0; i
<= index
; i
++)
253 return _mpb
+ offset
;
255 offset
+= sizeof_imsm_dev(_mpb
+ offset
);
260 static __u32
get_imsm_disk_idx(struct imsm_map
*map
, int slot
)
262 __u32
*ord_tbl
= &map
->disk_ord_tbl
[slot
];
264 /* top byte is 'special' */
265 return __le32_to_cpu(*ord_tbl
& ~(0xff << 24));
268 static int get_imsm_raid_level(struct imsm_map
*map
)
270 if (map
->raid_level
== 1) {
271 if (map
->num_members
== 2)
277 return map
->raid_level
;
280 static int cmp_extent(const void *av
, const void *bv
)
282 const struct extent
*a
= av
;
283 const struct extent
*b
= bv
;
284 if (a
->start
< b
->start
)
286 if (a
->start
> b
->start
)
291 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
293 /* find a list of used extents on the given physical device */
294 struct imsm_super
*mpb
= super
->mpb
;
295 struct imsm_disk
*disk
;
296 struct extent
*rv
, *e
;
300 disk
= get_imsm_disk(mpb
, dl
->index
);
304 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
305 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
306 struct imsm_map
*map
= dev
->vol
.map
;
308 for (j
= 0; j
< map
->num_members
; j
++) {
309 __u32 index
= get_imsm_disk_idx(map
, j
);
311 if (index
== dl
->index
)
315 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
320 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
321 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
322 struct imsm_map
*map
= dev
->vol
.map
;
324 for (j
= 0; j
< map
->num_members
; j
++) {
325 __u32 index
= get_imsm_disk_idx(map
, j
);
327 if (index
== dl
->index
) {
328 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
329 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
334 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
336 e
->start
= __le32_to_cpu(disk
->total_blocks
) -
337 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
343 static void print_imsm_dev(struct imsm_dev
*dev
, int index
)
347 struct imsm_map
*map
= dev
->vol
.map
;
350 printf("[%s]:\n", dev
->volume
);
351 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
352 printf(" Members : %d\n", map
->num_members
);
353 for (slot
= 0; slot
< map
->num_members
; slot
++)
354 if (index
== get_imsm_disk_idx(map
, slot
))
356 if (slot
< map
->num_members
)
357 printf(" This Slot : %d\n", slot
);
359 printf(" This Slot : ?\n");
360 sz
= __le32_to_cpu(dev
->size_high
);
362 sz
+= __le32_to_cpu(dev
->size_low
);
363 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
364 human_size(sz
* 512));
365 sz
= __le32_to_cpu(map
->blocks_per_member
);
366 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
367 human_size(sz
* 512));
368 printf(" Sector Offset : %u\n",
369 __le32_to_cpu(map
->pba_of_lba0
));
370 printf(" Num Stripes : %u\n",
371 __le32_to_cpu(map
->num_data_stripes
));
372 printf(" Chunk Size : %u KiB\n",
373 __le16_to_cpu(map
->blocks_per_strip
) / 2);
374 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
375 printf(" Migrate State : %s\n", dev
->vol
.migr_state
? "migrating" : "idle");
376 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
377 printf(" Map State : %s\n", map_state_str
[map
->map_state
]);
380 static void print_imsm_disk(struct imsm_super
*mpb
, int index
)
382 struct imsm_disk
*disk
= get_imsm_disk(mpb
, index
);
383 char str
[MAX_RAID_SERIAL_LEN
];
391 snprintf(str
, MAX_RAID_SERIAL_LEN
, "%s", disk
->serial
);
392 printf(" Disk%02d Serial : %s\n", index
, str
);
393 s
= __le32_to_cpu(disk
->status
);
394 printf(" State :%s%s%s%s\n", s
&SPARE_DISK
? " spare" : "",
395 s
&CONFIGURED_DISK
? " active" : "",
396 s
&FAILED_DISK
? " failed" : "",
397 s
&USABLE_DISK
? " usable" : "");
398 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
399 sz
= __le32_to_cpu(disk
->total_blocks
) -
400 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
* mpb
->num_raid_devs
);
401 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
402 human_size(sz
* 512));
405 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
407 struct intel_super
*super
= st
->sb
;
408 struct imsm_super
*mpb
= super
->mpb
;
409 char str
[MAX_SIGNATURE_LENGTH
];
413 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
414 printf(" Magic : %s\n", str
);
415 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
416 printf(" Version : %s\n", get_imsm_version(mpb
));
417 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
418 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
419 sum
= __le32_to_cpu(mpb
->check_sum
);
420 printf(" Checksum : %08x %s\n", sum
,
421 gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
422 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
423 printf(" Disks : %d\n", mpb
->num_disks
);
424 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
425 print_imsm_disk(mpb
, super
->disks
->index
);
426 for (i
= 0; i
< mpb
->num_raid_devs
; i
++)
427 print_imsm_dev(get_imsm_dev(mpb
, i
), super
->disks
->index
);
428 for (i
= 0; i
< mpb
->num_disks
; i
++) {
429 if (i
== super
->disks
->index
)
431 print_imsm_disk(mpb
, i
);
435 static void brief_examine_super_imsm(struct supertype
*st
)
437 struct intel_super
*super
= st
->sb
;
438 struct imsm_super
*mpb
= super
->mpb
;
440 printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
441 __le32_to_cpu(mpb
->family_num
));
444 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
446 printf("%s\n", __FUNCTION__
);
449 static void brief_detail_super_imsm(struct supertype
*st
)
451 printf("%s\n", __FUNCTION__
);
455 static int match_home_imsm(struct supertype
*st
, char *homehost
)
457 printf("%s\n", __FUNCTION__
);
462 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
464 printf("%s\n", __FUNCTION__
);
469 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
471 __u8
*v
= get_imsm_version(mpb
);
472 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
473 char major
[] = { 0, 0, 0 };
474 char minor
[] = { 0 ,0, 0 };
475 char patch
[] = { 0, 0, 0 };
476 char *ver_parse
[] = { major
, minor
, patch
};
480 while (*v
!= '\0' && v
< end
) {
481 if (*v
!= '.' && j
< 2)
482 ver_parse
[i
][j
++] = *v
;
490 *m
= strtol(minor
, NULL
, 0);
491 *p
= strtol(patch
, NULL
, 0);
495 static int imsm_level_to_layout(int level
)
503 return ALGORITHM_LEFT_SYMMETRIC
;
505 return 0x102; //FIXME is this correct?
510 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
512 struct intel_super
*super
= st
->sb
;
513 struct imsm_super
*mpb
= super
->mpb
;
514 struct imsm_dev
*dev
= get_imsm_dev(mpb
, super
->current_vol
);
515 struct imsm_map
*map
= &dev
->vol
.map
[0];
517 info
->container_member
= super
->current_vol
;
518 info
->array
.raid_disks
= map
->num_members
;
519 info
->array
.level
= get_imsm_raid_level(map
);
520 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
521 info
->array
.md_minor
= -1;
522 info
->array
.ctime
= 0;
523 info
->array
.utime
= 0;
524 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
* 512);
526 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
527 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
529 info
->disk
.major
= 0;
530 info
->disk
.minor
= 0;
532 sprintf(info
->text_version
, "/%s/%d",
533 devnum2devname(st
->container_dev
),
534 info
->container_member
);
538 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
540 struct intel_super
*super
= st
->sb
;
541 struct imsm_super
*mpb
= super
->mpb
;
542 struct imsm_disk
*disk
;
545 if (super
->current_vol
>= 0) {
546 getinfo_super_imsm_volume(st
, info
);
549 info
->array
.raid_disks
= mpb
->num_disks
;
550 info
->array
.level
= LEVEL_CONTAINER
;
551 info
->array
.layout
= 0;
552 info
->array
.md_minor
= -1;
553 info
->array
.ctime
= 0; /* N/A for imsm */
554 info
->array
.utime
= 0;
555 info
->array
.chunk_size
= 0;
557 info
->disk
.major
= 0;
558 info
->disk
.minor
= 0;
559 info
->disk
.raid_disk
= -1;
560 info
->reshape_active
= 0;
561 strcpy(info
->text_version
, "imsm");
562 info
->disk
.number
= -1;
563 info
->disk
.state
= 0;
566 disk
= get_imsm_disk(mpb
, super
->disks
->index
);
568 info
->disk
.number
= -1;
569 info
->disk
.raid_disk
= -1;
572 info
->disk
.number
= super
->disks
->index
;
573 info
->disk
.raid_disk
= super
->disks
->index
;
574 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) -
575 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
576 info
->component_size
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
577 s
= __le32_to_cpu(disk
->status
);
578 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
579 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
580 info
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
584 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
585 char *update
, char *devname
, int verbose
,
586 int uuid_set
, char *homehost
)
590 /* For 'assemble' and 'force' we need to return non-zero if any
591 * change was made. For others, the return value is ignored.
592 * Update options are:
593 * force-one : This device looks a bit old but needs to be included,
594 * update age info appropriately.
595 * assemble: clear any 'faulty' flag to allow this device to
597 * force-array: Array is degraded but being forced, mark it clean
598 * if that will be needed to assemble it.
600 * newdev: not used ????
601 * grow: Array has gained a new device - this is currently for
603 * resync: mark as dirty so a resync will happen.
604 * name: update the name - preserving the homehost
606 * Following are not relevant for this imsm:
607 * sparc2.2 : update from old dodgey metadata
608 * super-minor: change the preferred_minor number
609 * summaries: update redundant counters.
610 * uuid: Change the uuid of the array to match watch is given
611 * homehost: update the recorded homehost
612 * _reshape_progress: record new reshape_progress position.
615 //struct intel_super *super = st->sb;
616 //struct imsm_super *mpb = super->mpb;
618 if (strcmp(update
, "grow") == 0) {
620 if (strcmp(update
, "resync") == 0) {
621 /* dev->vol.dirty = 1; */
624 /* IMSM has no concept of UUID or homehost */
629 static size_t disks_to_mpb_size(int disks
)
633 size
= sizeof(struct imsm_super
);
634 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
635 size
+= 2 * sizeof(struct imsm_dev
);
636 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
637 size
+= (4 - 2) * sizeof(struct imsm_map
);
638 /* 4 possible disk_ord_tbl's */
639 size
+= 4 * (disks
- 1) * sizeof(__u32
);
644 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
646 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
649 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
652 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
656 * 0 same, or first was empty, and second was copied
657 * 1 second had wrong number
661 struct intel_super
*first
= st
->sb
;
662 struct intel_super
*sec
= tst
->sb
;
670 if (memcmp(first
->mpb
->sig
, sec
->mpb
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
672 if (first
->mpb
->family_num
!= sec
->mpb
->family_num
)
674 if (first
->mpb
->mpb_size
!= sec
->mpb
->mpb_size
)
676 if (first
->mpb
->check_sum
!= sec
->mpb
->check_sum
)
682 static void fd2devname(int fd
, char *name
)
691 if (fstat(fd
, &st
) != 0)
693 sprintf(path
, "/sys/dev/block/%d:%d",
694 major(st
.st_rdev
), minor(st
.st_rdev
));
696 rv
= readlink(path
, dname
, sizeof(dname
));
701 nm
= strrchr(dname
, '/');
703 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
707 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
709 static int imsm_read_serial(int fd
, char *devname
,
710 __u8 serial
[MAX_RAID_SERIAL_LEN
])
712 unsigned char scsi_serial
[255];
718 memset(scsi_serial
, 0, sizeof(scsi_serial
));
720 sg_fd
= sysfs_disk_to_sg(fd
);
724 Name
": Failed to open sg interface for %s: %s\n",
725 devname
, strerror(errno
));
729 if (imsm_env_devname_as_serial()) {
730 char name
[MAX_RAID_SERIAL_LEN
];
732 fd2devname(fd
, name
);
733 strcpy((char *) serial
, name
);
737 rv
= scsi_get_serial(sg_fd
, scsi_serial
, sizeof(scsi_serial
));
743 Name
": Failed to retrieve serial for %s\n",
748 rsp_len
= scsi_serial
[3];
749 for (i
= 0, cnt
= 0; i
< rsp_len
; i
++) {
750 if (!isspace(scsi_serial
[4 + i
]))
751 serial
[cnt
++] = scsi_serial
[4 + i
];
752 if (cnt
== MAX_RAID_SERIAL_LEN
)
756 serial
[MAX_RAID_SERIAL_LEN
- 1] = '\0';
762 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
764 struct imsm_super
*mpb
= super
->mpb
;
767 struct imsm_disk
*disk
;
771 dl
= malloc(sizeof(*dl
));
775 Name
": failed to allocate disk buffer for %s\n",
779 memset(dl
, 0, sizeof(*dl
));
782 dl
->major
= major(stb
.st_rdev
);
783 dl
->minor
= minor(stb
.st_rdev
);
784 dl
->next
= super
->disks
;
785 dl
->fd
= keep_fd
? fd
: -1;
786 dl
->devname
= devname
? strdup(devname
) : NULL
;
789 rv
= imsm_read_serial(fd
, devname
, dl
->serial
);
794 /* look up this disk's index */
795 for (i
= 0; i
< mpb
->num_disks
; i
++) {
796 disk
= get_imsm_disk(mpb
, i
);
798 if (memcmp(disk
->serial
, dl
->serial
, MAX_RAID_SERIAL_LEN
) == 0)
802 if (i
> mpb
->num_disks
- 1) {
805 Name
": failed to match serial \'%s\' for %s\n",
806 dl
->serial
, devname
);
815 /* load_imsm_mpb - read matrix metadata
816 * allocates super->mpb to be freed by free_super
818 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
820 unsigned long long dsize
;
821 size_t len
, mpb_size
;
822 unsigned long long sectors
;
824 struct imsm_super
*anchor
;
827 get_dev_size(fd
, NULL
, &dsize
);
829 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
832 Name
": Cannot seek to anchor block on %s: %s\n",
833 devname
, strerror(errno
));
838 if (posix_memalign((void**)&anchor
, 512, len
) != 0) {
841 Name
": Failed to allocate imsm anchor buffer"
842 " on %s\n", devname
);
845 if (read(fd
, anchor
, len
) != len
) {
848 Name
": Cannot read anchor block on %s: %s\n",
849 devname
, strerror(errno
));
854 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
857 Name
": no IMSM anchor on %s\n", devname
);
862 mpb_size
= __le32_to_cpu(anchor
->mpb_size
);
863 mpb_size
= ROUND_UP(mpb_size
, 512);
864 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
867 Name
": unable to allocate %zu byte mpb buffer\n",
872 memcpy(super
->buf
, anchor
, len
);
874 sectors
= mpb_sectors(anchor
) - 1;
877 return load_imsm_disk(fd
, super
, devname
, 0);
879 /* read the extended mpb */
880 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
883 Name
": Cannot seek to extended mpb on %s: %s\n",
884 devname
, strerror(errno
));
888 len
= mpb_size
- 512;
889 if (read(fd
, super
->buf
+ 512, len
) != len
) {
892 Name
": Cannot read extended mpb on %s: %s\n",
893 devname
, strerror(errno
));
897 check_sum
= gen_imsm_checksum(super
->mpb
);
898 if (check_sum
!= __le32_to_cpu(super
->mpb
->check_sum
)) {
901 Name
": IMSM checksum %x != %x on %s\n",
902 check_sum
, __le32_to_cpu(super
->mpb
->check_sum
),
907 return load_imsm_disk(fd
, super
, devname
, 0);
910 static void free_imsm_disks(struct intel_super
*super
)
912 while (super
->disks
) {
913 struct dl
*d
= super
->disks
;
915 super
->disks
= d
->next
;
924 static void free_imsm(struct intel_super
*super
)
928 free_imsm_disks(super
);
933 static void free_super_imsm(struct supertype
*st
)
935 struct intel_super
*super
= st
->sb
;
944 static struct intel_super
*alloc_super(int creating_imsm
)
946 struct intel_super
*super
= malloc(sizeof(*super
));
949 memset(super
, 0, sizeof(*super
));
950 super
->creating_imsm
= creating_imsm
;
951 super
->current_vol
= -1;
958 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
959 char *devname
, int keep_fd
)
962 struct intel_super
*super
;
963 struct mdinfo
*sd
, *best
= NULL
;
970 /* check if this disk is a member of an active array */
971 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
975 if (sra
->array
.major_version
!= -1 ||
976 sra
->array
.minor_version
!= -2 ||
977 strcmp(sra
->text_version
, "imsm") != 0)
980 super
= alloc_super(0);
984 /* find the most up to date disk in this array */
985 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
986 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
987 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
992 rv
= load_imsm_mpb(dfd
, super
, NULL
);
996 gen
= __le32_to_cpu(super
->mpb
->generation_num
);
997 if (!best
|| gen
> bestgen
) {
1012 /* load the most up to date anchor */
1013 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
1014 dfd
= dev_open(nm
, O_RDONLY
);
1019 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1026 /* reset the disk list */
1027 free_imsm_disks(super
);
1029 /* populate disk list */
1030 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1031 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1032 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1037 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
1042 if (st
->subarray
[0]) {
1043 if (atoi(st
->subarray
) <= super
->mpb
->num_raid_devs
)
1044 super
->current_vol
= atoi(st
->subarray
);
1050 if (st
->ss
== NULL
) {
1051 st
->ss
= &super_imsm
;
1052 st
->minor_version
= 0;
1053 st
->max_devs
= IMSM_MAX_DEVICES
;
1054 st
->container_dev
= fd2devnum(fd
);
1061 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
1063 struct intel_super
*super
;
1067 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1070 if (st
->subarray
[0])
1071 return 1; /* FIXME */
1073 super
= alloc_super(0);
1076 Name
": malloc of %zu failed.\n",
1081 rv
= load_imsm_mpb(fd
, super
, devname
);
1086 Name
": Failed to load all information "
1087 "sections on %s\n", devname
);
1093 if (st
->ss
== NULL
) {
1094 st
->ss
= &super_imsm
;
1095 st
->minor_version
= 0;
1096 st
->max_devs
= IMSM_MAX_DEVICES
;
1102 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1104 if (info
->level
== 1)
1106 return info
->chunk_size
>> 9;
1109 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1113 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1114 if (info
->level
== 1)
1120 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1121 unsigned long long size
, char *name
,
1122 char *homehost
, int *uuid
)
1124 /* We are creating a volume inside a pre-existing container.
1125 * so st->sb is already set.
1127 struct intel_super
*super
= st
->sb
;
1128 struct imsm_super
*mpb
= super
->mpb
;
1129 struct imsm_dev
*dev
;
1130 struct imsm_vol
*vol
;
1131 struct imsm_map
*map
;
1132 int idx
= mpb
->num_raid_devs
;
1134 unsigned long long array_blocks
;
1136 size_t size_old
, size_new
;
1138 if (mpb
->num_raid_devs
>= 2) {
1139 fprintf(stderr
, Name
": This imsm-container already has the "
1140 "maximum of 2 volumes\n");
1144 /* ensure the mpb is large enough for the new data */
1145 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1146 size_new
= disks_to_mpb_size(info
->nr_disks
);
1147 if (size_new
> size_old
) {
1149 size_t size_round
= ROUND_UP(size_new
, 512);
1151 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1152 fprintf(stderr
, Name
": could not allocate new mpb\n");
1155 memcpy(mpb_new
, mpb
, size_old
);
1158 super
->mpb
= mpb_new
;
1159 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1160 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1162 super
->current_vol
= idx
;
1163 sprintf(st
->subarray
, "%d", idx
);
1164 mpb
->num_raid_devs
++;
1165 dev
= get_imsm_dev(mpb
, idx
);
1166 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1167 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1168 info
->layout
, info
->chunk_size
,
1170 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1171 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1172 dev
->status
= __cpu_to_le32(0);
1173 dev
->reserved_blocks
= __cpu_to_le32(0);
1175 vol
->migr_state
= 0;
1178 for (i
= 0; i
< idx
; i
++) {
1179 struct imsm_dev
*prev
= get_imsm_dev(mpb
, i
);
1180 struct imsm_map
*pmap
= &prev
->vol
.map
[0];
1182 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1183 offset
+= IMSM_RESERVED_SECTORS
;
1186 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1187 map
->blocks_per_member
= __cpu_to_le32(info
->size
* 2);
1188 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1189 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1190 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1191 IMSM_T_STATE_NORMAL
;
1193 if (info
->level
== 1 && info
->raid_disks
> 2) {
1194 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1195 "in a raid1 volume\n");
1198 if (info
->level
== 10)
1199 map
->raid_level
= 1;
1201 map
->raid_level
= info
->level
;
1203 map
->num_members
= info
->raid_disks
;
1204 for (i
= 0; i
< map
->num_members
; i
++) {
1205 /* initialized in add_to_super */
1206 map
->disk_ord_tbl
[i
] = __cpu_to_le32(0);
1212 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1213 unsigned long long size
, char *name
,
1214 char *homehost
, int *uuid
)
1216 /* This is primarily called by Create when creating a new array.
1217 * We will then get add_to_super called for each component, and then
1218 * write_init_super called to write it out to each device.
1219 * For IMSM, Create can create on fresh devices or on a pre-existing
1221 * To create on a pre-existing array a different method will be called.
1222 * This one is just for fresh drives.
1224 struct intel_super
*super
;
1225 struct imsm_super
*mpb
;
1233 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1236 super
= alloc_super(1);
1239 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1240 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1245 memset(mpb
, 0, mpb_size
);
1247 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1248 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1249 strlen(MPB_VERSION_RAID5
));
1250 mpb
->mpb_size
= mpb_size
;
1256 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1257 int fd
, char *devname
)
1259 struct intel_super
*super
= st
->sb
;
1260 struct imsm_super
*mpb
= super
->mpb
;
1262 struct imsm_dev
*dev
;
1263 struct imsm_map
*map
;
1264 struct imsm_disk
*disk
;
1267 dev
= get_imsm_dev(mpb
, super
->current_vol
);
1268 map
= &dev
->vol
.map
[0];
1270 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1271 if (dl
->major
== dk
->major
&&
1272 dl
->minor
== dk
->minor
)
1274 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1277 map
->disk_ord_tbl
[dk
->number
] = __cpu_to_le32(dl
->index
);
1279 disk
= get_imsm_disk(mpb
, dl
->index
);
1280 status
= CONFIGURED_DISK
| USABLE_DISK
;
1281 disk
->status
= __cpu_to_le32(status
);
1284 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1285 int fd
, char *devname
)
1287 struct intel_super
*super
= st
->sb
;
1288 struct imsm_super
*mpb
= super
->mpb
;
1289 struct imsm_disk
*disk
;
1291 unsigned long long size
;
1296 if (super
->current_vol
>= 0) {
1297 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1302 dd
= malloc(sizeof(*dd
));
1305 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1308 memset(dd
, 0, sizeof(*dd
));
1309 dd
->major
= major(stb
.st_rdev
);
1310 dd
->minor
= minor(stb
.st_rdev
);
1311 dd
->index
= dk
->number
;
1312 dd
->devname
= devname
? strdup(devname
) : NULL
;
1313 dd
->next
= super
->disks
;
1315 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1318 Name
": failed to retrieve scsi serial, aborting\n");
1322 if (mpb
->num_disks
<= dk
->number
)
1323 mpb
->num_disks
= dk
->number
+ 1;
1325 disk
= get_imsm_disk(mpb
, dk
->number
);
1326 get_dev_size(fd
, NULL
, &size
);
1328 status
= USABLE_DISK
| SPARE_DISK
;
1329 strcpy((char *) disk
->serial
, (char *) dd
->serial
);
1330 disk
->total_blocks
= __cpu_to_le32(size
);
1331 disk
->status
= __cpu_to_le32(status
);
1332 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1333 disk
->scsi_id
= __cpu_to_le32(id
);
1335 disk
->scsi_id
= __cpu_to_le32(0);
1337 /* update the family number if we are creating a container */
1338 if (super
->creating_imsm
)
1339 mpb
->family_num
= __cpu_to_le32(gen_imsm_checksum(mpb
));
1344 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1346 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1348 struct imsm_super
*mpb
= super
->mpb
;
1353 /* 'generation' is incremented everytime the metadata is written */
1354 generation
= __le32_to_cpu(mpb
->generation_num
);
1356 mpb
->generation_num
= __cpu_to_le32(generation
);
1358 /* recalculate checksum */
1359 sum
= gen_imsm_checksum(mpb
);
1360 mpb
->check_sum
= __cpu_to_le32(sum
);
1362 for (d
= super
->disks
; d
; d
= d
->next
) {
1363 if (store_imsm_mpb(d
->fd
, super
)) {
1364 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1365 __func__
, d
->major
, d
->minor
, strerror(errno
));
1377 static int write_init_super_imsm(struct supertype
*st
)
1379 if (st
->update_tail
) {
1380 /* queue the recently created array as a metadata update */
1382 struct imsm_update_create_array
*u
;
1383 struct intel_super
*super
= st
->sb
;
1384 struct imsm_super
*mpb
= super
->mpb
;
1385 struct imsm_dev
*dev
;
1386 struct imsm_map
*map
;
1389 if (super
->current_vol
< 0 ||
1390 !(dev
= get_imsm_dev(mpb
, super
->current_vol
))) {
1391 fprintf(stderr
, "%s: could not determine sub-array\n",
1397 map
= &dev
->vol
.map
[0];
1398 len
= sizeof(*u
) + sizeof(__u32
) * (map
->num_members
- 1);
1401 fprintf(stderr
, "%s: failed to allocate update buffer\n",
1406 u
->type
= update_create_array
;
1407 u
->dev_idx
= super
->current_vol
;
1408 memcpy(&u
->dev
, dev
, sizeof(*dev
));
1409 memcpy(u
->dev
.vol
.map
[0].disk_ord_tbl
, map
->disk_ord_tbl
,
1410 sizeof(__u32
) * map
->num_members
);
1411 append_metadata_update(st
, u
, len
);
1413 for (d
= super
->disks
; d
; d
= d
->next
) {
1420 return write_super_imsm(st
->sb
, 1);
1423 static int store_zero_imsm(struct supertype
*st
, int fd
)
1425 unsigned long long dsize
;
1428 get_dev_size(fd
, NULL
, &dsize
);
1430 /* first block is stored on second to last sector of the disk */
1431 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
1434 if (posix_memalign(&buf
, 512, 512) != 0)
1437 memset(buf
, 0, 512);
1438 if (write(fd
, buf
, 512) != 512)
1443 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
1444 int layout
, int raiddisks
, int chunk
,
1445 unsigned long long size
, char *dev
,
1446 unsigned long long *freesize
,
1450 unsigned long long ldsize
;
1452 if (level
!= LEVEL_CONTAINER
)
1457 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1460 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
1461 dev
, strerror(errno
));
1464 if (!get_dev_size(fd
, dev
, &ldsize
)) {
1470 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
1475 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
1476 * FIX ME add ahci details
1478 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
1479 int layout
, int raiddisks
, int chunk
,
1480 unsigned long long size
, char *dev
,
1481 unsigned long long *freesize
,
1485 struct intel_super
*super
= st
->sb
;
1487 unsigned long long pos
= 0;
1488 unsigned long long maxsize
;
1492 if (level
== LEVEL_CONTAINER
)
1495 if (level
== 1 && raiddisks
> 2) {
1497 fprintf(stderr
, Name
": imsm does not support more "
1498 "than 2 in a raid1 configuration\n");
1502 /* We must have the container info already read in. */
1507 /* General test: make sure there is space for
1508 * 'raiddisks' device extents of size 'size' at a given
1511 unsigned long long minsize
= size
*2 /* convert to blocks */;
1512 unsigned long long start_offset
= ~0ULL;
1515 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
1516 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1521 e
= get_extents(super
, dl
);
1524 unsigned long long esize
;
1525 esize
= e
[i
].start
- pos
;
1526 if (esize
>= minsize
)
1528 if (found
&& start_offset
== ~0ULL) {
1531 } else if (found
&& pos
!= start_offset
) {
1535 pos
= e
[i
].start
+ e
[i
].size
;
1537 } while (e
[i
-1].size
);
1542 if (dcnt
< raiddisks
) {
1544 fprintf(stderr
, Name
": imsm: Not enough "
1545 "devices with space for this array "
1552 /* This device must be a member of the set */
1553 if (stat(dev
, &stb
) < 0)
1555 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
1557 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1558 if (dl
->major
== major(stb
.st_rdev
) &&
1559 dl
->minor
== minor(stb
.st_rdev
))
1564 fprintf(stderr
, Name
": %s is not in the "
1565 "same imsm set\n", dev
);
1568 e
= get_extents(super
, dl
);
1572 unsigned long long esize
;
1573 esize
= e
[i
].start
- pos
;
1574 if (esize
>= maxsize
)
1576 pos
= e
[i
].start
+ e
[i
].size
;
1578 } while (e
[i
-1].size
);
1579 *freesize
= maxsize
;
1584 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
1585 int raiddisks
, int chunk
, unsigned long long size
,
1586 char *dev
, unsigned long long *freesize
,
1592 /* if given unused devices create a container
1593 * if given given devices in a container create a member volume
1595 if (level
== LEVEL_CONTAINER
) {
1596 /* Must be a fresh device to add to a container */
1597 return validate_geometry_imsm_container(st
, level
, layout
,
1598 raiddisks
, chunk
, size
,
1604 /* creating in a given container */
1605 return validate_geometry_imsm_volume(st
, level
, layout
,
1606 raiddisks
, chunk
, size
,
1607 dev
, freesize
, verbose
);
1610 /* limit creation to the following levels */
1622 /* This device needs to be a device in an 'imsm' container */
1623 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1627 Name
": Cannot create this array on device %s\n",
1632 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
1634 fprintf(stderr
, Name
": Cannot open %s: %s\n",
1635 dev
, strerror(errno
));
1638 /* Well, it is in use by someone, maybe an 'imsm' container. */
1639 cfd
= open_container(fd
);
1643 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
1647 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
1649 if (sra
&& sra
->array
.major_version
== -1 &&
1650 strcmp(sra
->text_version
, "imsm") == 0) {
1651 /* This is a member of a imsm container. Load the container
1652 * and try to create a volume
1654 struct intel_super
*super
;
1656 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
1658 st
->container_dev
= fd2devnum(cfd
);
1660 return validate_geometry_imsm_volume(st
, level
, layout
,
1666 } else /* may belong to another container */
1672 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
1674 /* Given a container loaded by load_super_imsm_all,
1675 * extract information about all the arrays into
1678 * For each imsm_dev create an mdinfo, fill it in,
1679 * then look for matching devices in super->disks
1680 * and create appropriate device mdinfo.
1682 struct intel_super
*super
= st
->sb
;
1683 struct imsm_super
*mpb
= super
->mpb
;
1684 struct mdinfo
*rest
= NULL
;
1687 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1688 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
1689 struct imsm_vol
*vol
= &dev
->vol
;
1690 struct imsm_map
*map
= vol
->map
;
1691 struct mdinfo
*this;
1695 this = malloc(sizeof(*this));
1696 memset(this, 0, sizeof(*this));
1700 this->array
.level
= get_imsm_raid_level(map
);
1701 this->array
.raid_disks
= map
->num_members
;
1702 this->array
.layout
= imsm_level_to_layout(this->array
.level
);
1703 this->array
.md_minor
= -1;
1704 this->array
.ctime
= 0;
1705 this->array
.utime
= 0;
1706 this->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
1707 this->array
.state
= !vol
->dirty
;
1708 this->container_member
= i
;
1709 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
1710 this->resync_start
= 0;
1712 this->resync_start
= ~0ULL;
1714 strncpy(this->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
1715 this->name
[MAX_RAID_SERIAL_LEN
] = 0;
1717 sprintf(this->text_version
, "/%s/%d",
1718 devnum2devname(st
->container_dev
),
1719 this->container_member
);
1721 memset(this->uuid
, 0, sizeof(this->uuid
));
1723 sz
= __le32_to_cpu(dev
->size_high
);
1725 sz
+= __le32_to_cpu(dev
->size_low
);
1726 this->component_size
= sz
;
1727 this->array
.size
= this->component_size
/ 2;
1729 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
1730 struct imsm_disk
*disk
;
1731 struct mdinfo
*info_d
;
1736 idx
= __le32_to_cpu(map
->disk_ord_tbl
[slot
] & ~(0xff << 24));
1737 for (d
= super
->disks
; d
; d
= d
->next
)
1738 if (d
->index
== idx
)
1742 break; /* shouldn't this be continue ?? */
1744 info_d
= malloc(sizeof(*info_d
));
1746 break; /* ditto ?? */
1747 memset(info_d
, 0, sizeof(*info_d
));
1748 info_d
->next
= this->devs
;
1749 this->devs
= info_d
;
1751 disk
= get_imsm_disk(mpb
, idx
);
1752 s
= __le32_to_cpu(disk
->status
);
1754 info_d
->disk
.number
= d
->index
;
1755 info_d
->disk
.major
= d
->major
;
1756 info_d
->disk
.minor
= d
->minor
;
1757 info_d
->disk
.raid_disk
= slot
;
1758 info_d
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
1759 info_d
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
1760 info_d
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
1762 this->array
.working_disks
++;
1764 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
1765 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
1766 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
1768 strcpy(info_d
->name
, d
->devname
);
1776 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
1779 struct intel_super
*super
= c
->sb
;
1780 struct imsm_super
*mpb
= super
->mpb
;
1782 if (atoi(inst
) + 1 > mpb
->num_raid_devs
) {
1783 fprintf(stderr
, "%s: subarry index %d, out of range\n",
1784 __func__
, atoi(inst
));
1788 dprintf("imsm: open_new %s\n", inst
);
1789 a
->info
.container_member
= atoi(inst
);
1793 static __u8
imsm_check_degraded(struct imsm_super
*mpb
, int n
, int failed
)
1795 struct imsm_dev
*dev
= get_imsm_dev(mpb
, n
);
1796 struct imsm_map
*map
= dev
->vol
.map
;
1799 return map
->map_state
;
1801 switch (get_imsm_raid_level(map
)) {
1803 return IMSM_T_STATE_FAILED
;
1806 if (failed
< map
->num_members
)
1807 return IMSM_T_STATE_DEGRADED
;
1809 return IMSM_T_STATE_FAILED
;
1814 * check to see if any mirrors have failed,
1815 * otherwise we are degraded
1817 int device_per_mirror
= 2; /* FIXME is this always the case?
1818 * and are they always adjacent?
1823 for (i
= 0; i
< map
->num_members
; i
++) {
1824 int idx
= get_imsm_disk_idx(map
, i
);
1825 struct imsm_disk
*disk
= get_imsm_disk(mpb
, idx
);
1827 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
1830 if (failed
>= device_per_mirror
)
1831 return IMSM_T_STATE_FAILED
;
1833 /* reset 'failed' for next mirror set */
1834 if (!((i
+ 1) % device_per_mirror
))
1838 return IMSM_T_STATE_DEGRADED
;
1842 return IMSM_T_STATE_DEGRADED
;
1844 return IMSM_T_STATE_FAILED
;
1850 return map
->map_state
;
1853 static int imsm_count_failed(struct imsm_super
*mpb
, struct imsm_map
*map
)
1857 struct imsm_disk
*disk
;
1859 for (i
= 0; i
< map
->num_members
; i
++) {
1860 int idx
= get_imsm_disk_idx(map
, i
);
1862 disk
= get_imsm_disk(mpb
, idx
);
1863 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
1870 static void imsm_set_array_state(struct active_array
*a
, int consistent
)
1872 int inst
= a
->info
.container_member
;
1873 struct intel_super
*super
= a
->container
->sb
;
1874 struct imsm_dev
*dev
= get_imsm_dev(super
->mpb
, inst
);
1875 struct imsm_map
*map
= &dev
->vol
.map
[0];
1876 int dirty
= !consistent
;
1880 if (a
->resync_start
== ~0ULL) {
1881 failed
= imsm_count_failed(super
->mpb
, map
);
1882 map_state
= imsm_check_degraded(super
->mpb
, inst
, failed
);
1884 map_state
= IMSM_T_STATE_NORMAL
;
1885 if (map
->map_state
!= map_state
) {
1886 dprintf("imsm: map_state %d: %d\n",
1888 map
->map_state
= map_state
;
1889 super
->updates_pending
++;
1893 if (dev
->vol
.dirty
!= dirty
) {
1894 dprintf("imsm: mark '%s' (%llu)\n",
1895 dirty
?"dirty":"clean", a
->resync_start
);
1897 dev
->vol
.dirty
= dirty
;
1898 super
->updates_pending
++;
1902 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
1904 int inst
= a
->info
.container_member
;
1905 struct intel_super
*super
= a
->container
->sb
;
1906 struct imsm_dev
*dev
= get_imsm_dev(super
->mpb
, inst
);
1907 struct imsm_map
*map
= dev
->vol
.map
;
1908 struct imsm_disk
*disk
;
1911 int new_failure
= 0;
1913 if (n
> map
->num_members
)
1914 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
1915 n
, map
->num_members
- 1);
1920 dprintf("imsm: set_disk %d:%x\n", n
, state
);
1922 disk
= get_imsm_disk(super
->mpb
, get_imsm_disk_idx(map
, n
));
1924 /* check for new failures */
1925 status
= __le32_to_cpu(disk
->status
);
1926 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
1927 status
|= FAILED_DISK
;
1928 disk
->status
= __cpu_to_le32(status
);
1930 super
->updates_pending
++;
1933 /* the number of failures have changed, count up 'failed' to determine
1934 * degraded / failed status
1936 if (new_failure
&& map
->map_state
!= IMSM_T_STATE_FAILED
)
1937 failed
= imsm_count_failed(super
->mpb
, map
);
1939 /* determine map_state based on failed or in_sync count */
1941 map
->map_state
= imsm_check_degraded(super
->mpb
, inst
, failed
);
1942 else if (map
->map_state
== IMSM_T_STATE_DEGRADED
) {
1946 for (d
= a
->info
.devs
; d
; d
= d
->next
)
1947 if (d
->curr_state
& DS_INSYNC
)
1950 if (working
== a
->info
.array
.raid_disks
) {
1951 map
->map_state
= IMSM_T_STATE_NORMAL
;
1952 super
->updates_pending
++;
1957 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
1959 struct imsm_super
*mpb
= super
->mpb
;
1960 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
1961 unsigned long long dsize
;
1962 unsigned long long sectors
;
1964 get_dev_size(fd
, NULL
, &dsize
);
1966 if (mpb_size
> 512) {
1967 /* -1 to account for anchor */
1968 sectors
= mpb_sectors(mpb
) - 1;
1970 /* write the extended mpb to the sectors preceeding the anchor */
1971 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
1974 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
1978 /* first block is stored on second to last sector of the disk */
1979 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
1982 if (write(fd
, super
->buf
, 512) != 512)
1988 static void imsm_sync_metadata(struct supertype
*container
)
1990 struct intel_super
*super
= container
->sb
;
1992 if (!super
->updates_pending
)
1995 write_super_imsm(super
, 0);
1997 super
->updates_pending
= 0;
2000 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
2001 struct metadata_update
**updates
)
2004 * Take a device that is marked spare in the metadata and use it to
2005 * replace a failed/vacant slot in an array. There may be a case where
2006 * a device is failed in one array but active in a second.
2007 * imsm_process_update catches this case and does not clear the SPARE_DISK
2008 * flag, allowing the second array to start using the device on failure.
2009 * SPARE_DISK is cleared when all arrays are using a device.
2011 * FIXME: is this a valid use of SPARE_DISK?
2014 struct intel_super
*super
= a
->container
->sb
;
2015 struct imsm_super
*mpb
= super
->mpb
;
2016 int inst
= a
->info
.container_member
;
2017 struct imsm_dev
*dev
= get_imsm_dev(mpb
, inst
);
2018 struct imsm_map
*map
= dev
->vol
.map
;
2019 int failed
= a
->info
.array
.raid_disks
;
2020 struct mdinfo
*rv
= NULL
;
2023 struct metadata_update
*mu
;
2025 struct imsm_update_activate_spare
*u
;
2029 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
2030 if ((d
->curr_state
& DS_FAULTY
) &&
2032 /* wait for Removal to happen */
2034 if (d
->state_fd
>= 0)
2038 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
2039 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
2040 if (imsm_check_degraded(mpb
, inst
, failed
) != IMSM_T_STATE_DEGRADED
)
2043 /* For each slot, if it is not working, find a spare */
2045 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
2046 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2047 if (d
->disk
.raid_disk
== i
)
2049 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
2050 if (d
&& (d
->state_fd
>= 0))
2053 /* OK, this device needs recovery. Find a spare */
2054 for ( ; dl
; dl
= dl
->next
) {
2055 unsigned long long esize
;
2056 unsigned long long pos
;
2059 struct imsm_disk
*disk
;
2064 /* If in this array, skip */
2065 for (d2
= a
->info
.devs
; d2
; d2
= d2
->next
)
2066 if (d2
->disk
.major
== dl
->major
&&
2067 d2
->disk
.minor
== dl
->minor
) {
2068 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
2074 /* is this unused device marked as a spare? */
2075 disk
= get_imsm_disk(mpb
, dl
->index
);
2076 if (!(__le32_to_cpu(disk
->status
) & SPARE_DISK
))
2079 /* We are allowed to use this device - is there space?
2080 * We need a->info.component_size sectors */
2081 ex
= get_extents(super
, dl
);
2083 dprintf("cannot get extents\n");
2089 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
2092 /* check that we can start at pba_of_lba0 with
2093 * a->info.component_size of space
2095 esize
= ex
[j
].start
- pos
;
2096 if (array_start
>= pos
&&
2097 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
2101 pos
= ex
[j
].start
+ ex
[j
].size
;
2104 } while (ex
[j
-1].size
);
2108 dprintf("%x:%x does not have %llu at %d\n",
2109 dl
->major
, dl
->minor
,
2110 a
->info
.component_size
,
2111 __le32_to_cpu(map
->pba_of_lba0
));
2116 /* found a usable disk with enough space */
2117 di
= malloc(sizeof(*di
));
2118 memset(di
, 0, sizeof(*di
));
2119 di
->disk
.number
= dl
->index
;
2120 di
->disk
.raid_disk
= i
;
2121 di
->disk
.major
= dl
->major
;
2122 di
->disk
.minor
= dl
->minor
;
2124 di
->data_offset
= array_start
;
2125 di
->component_size
= a
->info
.component_size
;
2126 di
->container_member
= inst
;
2130 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2138 /* No spares found */
2140 /* Now 'rv' has a list of devices to return.
2141 * Create a metadata_update record to update the
2142 * disk_ord_tbl for the array
2144 mu
= malloc(sizeof(*mu
));
2145 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2147 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
2148 mu
->next
= *updates
;
2149 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
2151 for (di
= rv
; di
; di
= di
->next
) {
2152 u
->type
= update_activate_spare
;
2153 u
->disk_idx
= di
->disk
.number
;
2154 u
->slot
= di
->disk
.raid_disk
;
2165 static int weight(unsigned int field
)
2169 for (weight
= 0; field
; weight
++)
2175 static int disks_overlap(struct imsm_map
*m1
, struct imsm_map
*m2
)
2181 for (i
= 0; i
< m1
->num_members
; i
++) {
2182 idx
= get_imsm_disk_idx(m1
, i
);
2183 for (j
= 0; j
< m2
->num_members
; j
++)
2184 if (idx
== get_imsm_disk_idx(m2
, j
))
2191 static void imsm_process_update(struct supertype
*st
,
2192 struct metadata_update
*update
)
2195 * crack open the metadata_update envelope to find the update record
2196 * update can be one of:
2197 * update_activate_spare - a spare device has replaced a failed
2198 * device in an array, update the disk_ord_tbl. If this disk is
2199 * present in all member arrays then also clear the SPARE_DISK
2202 struct intel_super
*super
= st
->sb
;
2203 struct imsm_super
*mpb
= super
->mpb
;
2204 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
2207 case update_activate_spare
: {
2208 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
2209 struct imsm_dev
*dev
= get_imsm_dev(mpb
, u
->array
);
2210 struct imsm_map
*map
= &dev
->vol
.map
[0];
2211 struct active_array
*a
;
2212 struct imsm_disk
*disk
;
2216 unsigned int members
;
2221 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2222 if (dl
->index
== u
->disk_idx
)
2226 fprintf(stderr
, "error: imsm_activate_spare passed "
2227 "an unknown disk_idx: %d\n", u
->disk_idx
);
2231 super
->updates_pending
++;
2233 victim
= get_imsm_disk_idx(map
, u
->slot
);
2234 map
->disk_ord_tbl
[u
->slot
] = __cpu_to_le32(u
->disk_idx
);
2235 disk
= get_imsm_disk(mpb
, u
->disk_idx
);
2236 status
= __le32_to_cpu(disk
->status
);
2237 status
|= CONFIGURED_DISK
;
2238 disk
->status
= __cpu_to_le32(status
);
2240 /* map unique/live arrays using the spare */
2243 for (a
= st
->arrays
; a
; a
= a
->next
) {
2244 int inst
= a
->info
.container_member
;
2246 dev
= get_imsm_dev(mpb
, inst
);
2247 map
= &dev
->vol
.map
[0];
2248 if (map
->raid_level
> 0)
2249 members
|= 1 << inst
;
2250 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2251 if (d
->disk
.major
== dl
->major
&&
2252 d
->disk
.minor
== dl
->minor
)
2256 /* until all arrays that can absorb this disk have absorbed
2257 * this disk it can still be considered a spare
2259 if (weight(found
) >= weight(members
)) {
2260 status
= __le32_to_cpu(disk
->status
);
2261 status
&= ~SPARE_DISK
;
2262 disk
->status
= __cpu_to_le32(status
);
2265 /* count arrays using the victim in the metadata */
2267 for (a
= st
->arrays
; a
; a
= a
->next
) {
2268 dev
= get_imsm_dev(mpb
, a
->info
.container_member
);
2269 map
= &dev
->vol
.map
[0];
2270 for (i
= 0; i
< map
->num_members
; i
++)
2271 if (victim
== get_imsm_disk_idx(map
, i
))
2275 /* clear some flags if the victim is no longer being
2278 disk
= get_imsm_disk(mpb
, victim
);
2280 status
= __le32_to_cpu(disk
->status
);
2281 status
&= ~(CONFIGURED_DISK
| USABLE_DISK
);
2282 disk
->status
= __cpu_to_le32(status
);
2286 case update_create_array
: {
2287 /* someone wants to create a new array, we need to be aware of
2288 * a few races/collisions:
2289 * 1/ 'Create' called by two separate instances of mdadm
2290 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
2291 * devices that have since been assimilated via
2293 * In the event this update can not be carried out mdadm will
2294 * (FIX ME) notice that its update did not take hold.
2296 struct imsm_update_create_array
*u
= (void *) update
->buf
;
2297 struct imsm_dev
*dev
;
2298 struct imsm_map
*map
, *new_map
;
2299 unsigned long long start
, end
;
2300 unsigned long long new_start
, new_end
;
2304 /* handle racing creates: first come first serve */
2305 if (u
->dev_idx
< mpb
->num_raid_devs
) {
2306 dprintf("%s: subarray %d already defined\n",
2307 __func__
, u
->dev_idx
);
2311 /* check update is next in sequence */
2312 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
2313 dprintf("%s: can not create arrays out of sequence\n",
2318 new_map
= &u
->dev
.vol
.map
[0];
2319 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
2320 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
2322 /* handle activate_spare versus create race:
2323 * check to make sure that overlapping arrays do not include
2326 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2327 dev
= get_imsm_dev(mpb
, i
);
2328 map
= &dev
->vol
.map
[0];
2329 start
= __le32_to_cpu(map
->pba_of_lba0
);
2330 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
2331 if ((new_start
>= start
&& new_start
<= end
) ||
2332 (start
>= new_start
&& start
<= new_end
))
2334 if (overlap
&& disks_overlap(map
, new_map
)) {
2335 dprintf("%s: arrays overlap\n", __func__
);
2339 /* check num_members sanity */
2340 if (new_map
->num_members
> mpb
->num_disks
) {
2341 dprintf("%s: num_disks out of range\n", __func__
);
2345 super
->updates_pending
++;
2346 mpb
->num_raid_devs
++;
2347 dev
= get_imsm_dev(mpb
, u
->dev_idx
);
2348 memcpy(dev
, &u
->dev
, sizeof(*dev
));
2349 map
= &dev
->vol
.map
[0];
2350 memcpy(map
->disk_ord_tbl
, new_map
->disk_ord_tbl
,
2351 sizeof(__u32
) * new_map
->num_members
);
2353 /* fix up flags, if arrays overlap then the drives can not be
2356 for (i
= 0; i
< map
->num_members
; i
++) {
2357 struct imsm_disk
*disk
;
2360 disk
= get_imsm_disk(mpb
, get_imsm_disk_idx(map
, i
));
2361 status
= __le32_to_cpu(disk
->status
);
2362 status
|= CONFIGURED_DISK
;
2364 status
&= ~SPARE_DISK
;
2365 disk
->status
= __cpu_to_le32(status
);
2372 static void imsm_prepare_update(struct supertype
*st
,
2373 struct metadata_update
*update
)
2375 /* Allocate space to hold a new mpb if necessary. We currently
2376 * allocate enough to hold 2 subarrays for the given number of disks.
2377 * This may not be sufficient iff reshaping.
2379 * FIX ME handle the reshape case.
2381 * The monitor will be able to safely change super->mpb by arranging
2382 * for it to be freed in check_update_queue(). I.e. the monitor thread
2383 * will start using the new pointer and the manager can continue to use
2384 * the old value until check_update_queue() runs.
2390 struct superswitch super_imsm
= {
2392 .examine_super
= examine_super_imsm
,
2393 .brief_examine_super
= brief_examine_super_imsm
,
2394 .detail_super
= detail_super_imsm
,
2395 .brief_detail_super
= brief_detail_super_imsm
,
2396 .write_init_super
= write_init_super_imsm
,
2398 .match_home
= match_home_imsm
,
2399 .uuid_from_super
= uuid_from_super_imsm
,
2400 .getinfo_super
= getinfo_super_imsm
,
2401 .update_super
= update_super_imsm
,
2403 .avail_size
= avail_size_imsm
,
2405 .compare_super
= compare_super_imsm
,
2407 .load_super
= load_super_imsm
,
2408 .init_super
= init_super_imsm
,
2409 .add_to_super
= add_to_super_imsm
,
2410 .store_super
= store_zero_imsm
,
2411 .free_super
= free_super_imsm
,
2412 .match_metadata_desc
= match_metadata_desc_imsm
,
2413 .container_content
= container_content_imsm
,
2415 .validate_geometry
= validate_geometry_imsm
,
2419 .open_new
= imsm_open_new
,
2420 .load_super
= load_super_imsm
,
2421 .set_array_state
= imsm_set_array_state
,
2422 .set_disk
= imsm_set_disk
,
2423 .sync_metadata
= imsm_sync_metadata
,
2424 .activate_spare
= imsm_activate_spare
,
2425 .process_update
= imsm_process_update
,
2426 .prepare_update
= imsm_prepare_update
,