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
,
155 struct imsm_update_activate_spare
{
156 enum imsm_update_type type
;
160 struct imsm_update_activate_spare
*next
;
163 static struct supertype
*match_metadata_desc_imsm(char *arg
)
165 struct supertype
*st
;
167 if (strcmp(arg
, "imsm") != 0 &&
168 strcmp(arg
, "default") != 0
172 st
= malloc(sizeof(*st
));
173 memset(st
, 0, sizeof(*st
));
174 st
->ss
= &super_imsm
;
175 st
->max_devs
= IMSM_MAX_DEVICES
;
176 st
->minor_version
= 0;
181 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
183 return &mpb
->sig
[MPB_SIG_LEN
];
186 static struct imsm_disk
*get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
188 if (index
> mpb
->num_disks
- 1)
190 return &mpb
->disk
[index
];
193 static __u32
gen_imsm_checksum(struct imsm_super
*mpb
)
195 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
196 __u32
*p
= (__u32
*) mpb
;
200 sum
+= __le32_to_cpu(*p
++);
202 return sum
- __le32_to_cpu(mpb
->check_sum
);
205 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
)
207 size_t size
= sizeof(*dev
);
209 /* each map has disk_ord_tbl[num_members - 1] additional space */
210 size
+= sizeof(__u32
) * (dev
->vol
.map
[0].num_members
- 1);
212 /* migrating means an additional map */
213 if (dev
->vol
.migr_state
) {
214 size
+= sizeof(struct imsm_map
);
215 size
+= sizeof(__u32
) * (dev
->vol
.map
[1].num_members
- 1);
221 static struct imsm_dev
*get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
227 if (index
> mpb
->num_raid_devs
- 1)
230 /* devices start after all disks */
231 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
233 for (i
= 0; i
<= index
; i
++)
235 return _mpb
+ offset
;
237 offset
+= sizeof_imsm_dev(_mpb
+ offset
);
242 static __u32
get_imsm_disk_idx(struct imsm_map
*map
, int slot
)
244 __u32
*ord_tbl
= &map
->disk_ord_tbl
[slot
];
246 /* top byte is 'special' */
247 return __le32_to_cpu(*ord_tbl
& ~(0xff << 24));
250 static int get_imsm_raid_level(struct imsm_map
*map
)
252 if (map
->raid_level
== 1) {
253 if (map
->num_members
== 2)
259 return map
->raid_level
;
262 static int cmp_extent(const void *av
, const void *bv
)
264 const struct extent
*a
= av
;
265 const struct extent
*b
= bv
;
266 if (a
->start
< b
->start
)
268 if (a
->start
> b
->start
)
273 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
275 /* find a list of used extents on the given physical device */
276 struct imsm_super
*mpb
= super
->mpb
;
277 struct imsm_disk
*disk
;
278 struct extent
*rv
, *e
;
282 disk
= get_imsm_disk(mpb
, dl
->index
);
286 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
287 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
288 struct imsm_map
*map
= dev
->vol
.map
;
290 for (j
= 0; j
< map
->num_members
; j
++) {
291 __u32 index
= get_imsm_disk_idx(map
, j
);
293 if (index
== dl
->index
)
297 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
302 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
303 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
304 struct imsm_map
*map
= dev
->vol
.map
;
306 for (j
= 0; j
< map
->num_members
; j
++) {
307 __u32 index
= get_imsm_disk_idx(map
, j
);
309 if (index
== dl
->index
) {
310 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
311 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
316 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
318 e
->start
= __le32_to_cpu(disk
->total_blocks
) -
319 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
325 static void print_imsm_dev(struct imsm_dev
*dev
, int index
)
329 struct imsm_map
*map
= dev
->vol
.map
;
332 printf("[%s]:\n", dev
->volume
);
333 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
334 printf(" Members : %d\n", map
->num_members
);
335 for (slot
= 0; slot
< map
->num_members
; slot
++)
336 if (index
== get_imsm_disk_idx(map
, slot
))
338 if (slot
< map
->num_members
)
339 printf(" This Slot : %d\n", slot
);
341 printf(" This Slot : ?\n");
342 sz
= __le32_to_cpu(dev
->size_high
);
344 sz
+= __le32_to_cpu(dev
->size_low
);
345 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
346 human_size(sz
* 512));
347 sz
= __le32_to_cpu(map
->blocks_per_member
);
348 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
349 human_size(sz
* 512));
350 printf(" Sector Offset : %u\n",
351 __le32_to_cpu(map
->pba_of_lba0
));
352 printf(" Num Stripes : %u\n",
353 __le32_to_cpu(map
->num_data_stripes
));
354 printf(" Chunk Size : %u KiB\n",
355 __le16_to_cpu(map
->blocks_per_strip
) / 2);
356 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
357 printf(" Migrate State : %s\n", dev
->vol
.migr_state
? "migrating" : "idle");
358 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
359 printf(" Map State : %s\n", map_state_str
[map
->map_state
]);
362 static void print_imsm_disk(struct imsm_super
*mpb
, int index
)
364 struct imsm_disk
*disk
= get_imsm_disk(mpb
, index
);
365 char str
[MAX_RAID_SERIAL_LEN
];
370 snprintf(str
, MAX_RAID_SERIAL_LEN
, "%s", disk
->serial
);
371 printf(" Disk%02d Serial : %s\n", index
, str
);
372 s
= __le32_to_cpu(disk
->status
);
373 printf(" State :%s%s%s%s\n", s
&SPARE_DISK
? " spare" : "",
374 s
&CONFIGURED_DISK
? " active" : "",
375 s
&FAILED_DISK
? " failed" : "",
376 s
&USABLE_DISK
? " usable" : "");
377 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
378 sz
= __le32_to_cpu(disk
->total_blocks
) -
379 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
* mpb
->num_raid_devs
);
380 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
381 human_size(sz
* 512));
384 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
386 struct intel_super
*super
= st
->sb
;
387 struct imsm_super
*mpb
= super
->mpb
;
388 char str
[MAX_SIGNATURE_LENGTH
];
392 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
393 printf(" Magic : %s\n", str
);
394 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
395 printf(" Version : %s\n", get_imsm_version(mpb
));
396 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
397 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
398 sum
= __le32_to_cpu(mpb
->check_sum
);
399 printf(" Checksum : %08x %s\n", sum
,
400 gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
401 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
402 printf(" Disks : %d\n", mpb
->num_disks
);
403 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
404 print_imsm_disk(mpb
, super
->disks
->index
);
405 for (i
= 0; i
< mpb
->num_raid_devs
; i
++)
406 print_imsm_dev(get_imsm_dev(mpb
, i
), super
->disks
->index
);
407 for (i
= 0; i
< mpb
->num_disks
; i
++) {
408 if (i
== super
->disks
->index
)
410 print_imsm_disk(mpb
, i
);
414 static void brief_examine_super_imsm(struct supertype
*st
)
416 struct intel_super
*super
= st
->sb
;
417 struct imsm_super
*mpb
= super
->mpb
;
419 printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
420 __le32_to_cpu(mpb
->family_num
));
423 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
425 printf("%s\n", __FUNCTION__
);
428 static void brief_detail_super_imsm(struct supertype
*st
)
430 printf("%s\n", __FUNCTION__
);
434 static int match_home_imsm(struct supertype
*st
, char *homehost
)
436 printf("%s\n", __FUNCTION__
);
441 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
443 printf("%s\n", __FUNCTION__
);
448 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
450 __u8
*v
= get_imsm_version(mpb
);
451 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
452 char major
[] = { 0, 0, 0 };
453 char minor
[] = { 0 ,0, 0 };
454 char patch
[] = { 0, 0, 0 };
455 char *ver_parse
[] = { major
, minor
, patch
};
459 while (*v
!= '\0' && v
< end
) {
460 if (*v
!= '.' && j
< 2)
461 ver_parse
[i
][j
++] = *v
;
469 *m
= strtol(minor
, NULL
, 0);
470 *p
= strtol(patch
, NULL
, 0);
474 static int imsm_level_to_layout(int level
)
482 return ALGORITHM_LEFT_SYMMETRIC
;
484 return 0x102; //FIXME is this correct?
489 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
491 struct intel_super
*super
= st
->sb
;
492 struct imsm_super
*mpb
= super
->mpb
;
493 struct imsm_dev
*dev
= get_imsm_dev(mpb
, super
->current_vol
);
494 struct imsm_map
*map
= &dev
->vol
.map
[0];
496 info
->container_member
= super
->current_vol
;
497 info
->array
.raid_disks
= map
->num_members
;
498 info
->array
.level
= get_imsm_raid_level(map
);
499 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
500 info
->array
.md_minor
= -1;
501 info
->array
.ctime
= 0;
502 info
->array
.utime
= 0;
503 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
* 512);
505 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
506 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
508 info
->disk
.major
= 0;
509 info
->disk
.minor
= 0;
511 sprintf(info
->text_version
, "/%s/%d",
512 devnum2devname(st
->container_dev
),
513 info
->container_member
);
517 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
519 struct intel_super
*super
= st
->sb
;
520 struct imsm_super
*mpb
= super
->mpb
;
521 struct imsm_disk
*disk
;
524 if (super
->current_vol
>= 0) {
525 getinfo_super_imsm_volume(st
, info
);
528 info
->array
.raid_disks
= mpb
->num_disks
;
529 info
->array
.level
= LEVEL_CONTAINER
;
530 info
->array
.layout
= 0;
531 info
->array
.md_minor
= -1;
532 info
->array
.ctime
= 0; /* N/A for imsm */
533 info
->array
.utime
= 0;
534 info
->array
.chunk_size
= 0;
536 info
->disk
.major
= 0;
537 info
->disk
.minor
= 0;
538 info
->disk
.raid_disk
= -1;
539 info
->reshape_active
= 0;
540 strcpy(info
->text_version
, "imsm");
541 info
->disk
.number
= -1;
542 info
->disk
.state
= 0;
545 disk
= get_imsm_disk(mpb
, super
->disks
->index
);
546 info
->disk
.number
= super
->disks
->index
;
547 info
->disk
.raid_disk
= super
->disks
->index
;
548 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) -
549 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
550 info
->component_size
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
551 s
= __le32_to_cpu(disk
->status
);
552 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
553 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
554 info
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
558 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
559 char *update
, char *devname
, int verbose
,
560 int uuid_set
, char *homehost
)
564 /* For 'assemble' and 'force' we need to return non-zero if any
565 * change was made. For others, the return value is ignored.
566 * Update options are:
567 * force-one : This device looks a bit old but needs to be included,
568 * update age info appropriately.
569 * assemble: clear any 'faulty' flag to allow this device to
571 * force-array: Array is degraded but being forced, mark it clean
572 * if that will be needed to assemble it.
574 * newdev: not used ????
575 * grow: Array has gained a new device - this is currently for
577 * resync: mark as dirty so a resync will happen.
578 * name: update the name - preserving the homehost
580 * Following are not relevant for this imsm:
581 * sparc2.2 : update from old dodgey metadata
582 * super-minor: change the preferred_minor number
583 * summaries: update redundant counters.
584 * uuid: Change the uuid of the array to match watch is given
585 * homehost: update the recorded homehost
586 * _reshape_progress: record new reshape_progress position.
589 //struct intel_super *super = st->sb;
590 //struct imsm_super *mpb = super->mpb;
592 if (strcmp(update
, "grow") == 0) {
594 if (strcmp(update
, "resync") == 0) {
595 /* dev->vol.dirty = 1; */
598 /* IMSM has no concept of UUID or homehost */
603 static size_t disks_to_mpb_size(int disks
)
607 size
= sizeof(struct imsm_super
);
608 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
609 size
+= 2 * sizeof(struct imsm_dev
);
610 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
611 size
+= (4 - 2) * sizeof(struct imsm_map
);
612 /* 4 possible disk_ord_tbl's */
613 size
+= 4 * (disks
- 1) * sizeof(__u32
);
618 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
620 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
623 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
626 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
630 * 0 same, or first was empty, and second was copied
631 * 1 second had wrong number
635 struct intel_super
*first
= st
->sb
;
636 struct intel_super
*sec
= tst
->sb
;
644 if (memcmp(first
->mpb
->sig
, sec
->mpb
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
646 if (first
->mpb
->family_num
!= sec
->mpb
->family_num
)
648 if (first
->mpb
->mpb_size
!= sec
->mpb
->mpb_size
)
650 if (first
->mpb
->check_sum
!= sec
->mpb
->check_sum
)
656 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
658 static int imsm_read_serial(int fd
, char *devname
,
659 __u8 serial
[MAX_RAID_SERIAL_LEN
])
661 unsigned char scsi_serial
[255];
667 memset(scsi_serial
, 0, sizeof(scsi_serial
));
669 sg_fd
= sysfs_disk_to_sg(fd
);
673 Name
": Failed to open sg interface for %s: %s\n",
674 devname
, strerror(errno
));
678 rv
= scsi_get_serial(sg_fd
, scsi_serial
, sizeof(scsi_serial
));
684 Name
": Failed to retrieve serial for %s\n",
689 rsp_len
= scsi_serial
[3];
690 for (i
= 0, cnt
= 0; i
< rsp_len
; i
++) {
691 if (!isspace(scsi_serial
[4 + i
]))
692 serial
[cnt
++] = scsi_serial
[4 + i
];
693 if (cnt
== MAX_RAID_SERIAL_LEN
)
697 serial
[MAX_RAID_SERIAL_LEN
- 1] = '\0';
703 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
705 struct imsm_super
*mpb
= super
->mpb
;
708 struct imsm_disk
*disk
;
712 dl
= malloc(sizeof(*dl
));
716 Name
": failed to allocate disk buffer for %s\n",
720 memset(dl
, 0, sizeof(*dl
));
723 dl
->major
= major(stb
.st_rdev
);
724 dl
->minor
= minor(stb
.st_rdev
);
725 dl
->next
= super
->disks
;
726 dl
->fd
= keep_fd
? fd
: -1;
727 dl
->devname
= devname
? strdup(devname
) : NULL
;
730 rv
= imsm_read_serial(fd
, devname
, dl
->serial
);
735 /* look up this disk's index */
736 for (i
= 0; i
< mpb
->num_disks
; i
++) {
737 disk
= get_imsm_disk(mpb
, i
);
739 if (memcmp(disk
->serial
, dl
->serial
, MAX_RAID_SERIAL_LEN
) == 0)
743 if (i
> mpb
->num_disks
)
751 /* load_imsm_mpb - read matrix metadata
752 * allocates super->mpb to be freed by free_super
754 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
756 unsigned long long dsize
;
757 size_t len
, mpb_size
;
758 unsigned long long sectors
;
760 struct imsm_super
*anchor
;
763 get_dev_size(fd
, NULL
, &dsize
);
765 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
768 Name
": Cannot seek to anchor block on %s: %s\n",
769 devname
, strerror(errno
));
774 if (posix_memalign((void**)&anchor
, 512, len
) != 0) {
777 Name
": Failed to allocate imsm anchor buffer"
778 " on %s\n", devname
);
781 if (read(fd
, anchor
, len
) != len
) {
784 Name
": Cannot read anchor block on %s: %s\n",
785 devname
, strerror(errno
));
790 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
793 Name
": no IMSM anchor on %s\n", devname
);
798 mpb_size
= __le32_to_cpu(anchor
->mpb_size
);
799 mpb_size
= ROUND_UP(mpb_size
, 512);
800 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
803 Name
": unable to allocate %zu byte mpb buffer\n",
808 memcpy(super
->buf
, anchor
, len
);
810 sectors
= mpb_sectors(anchor
) - 1;
813 return load_imsm_disk(fd
, super
, devname
, 0);
815 /* read the extended mpb */
816 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
819 Name
": Cannot seek to extended mpb on %s: %s\n",
820 devname
, strerror(errno
));
824 len
= mpb_size
- 512;
825 if (read(fd
, super
->buf
+ 512, len
) != len
) {
828 Name
": Cannot read extended mpb on %s: %s\n",
829 devname
, strerror(errno
));
833 check_sum
= gen_imsm_checksum(super
->mpb
);
834 if (check_sum
!= __le32_to_cpu(super
->mpb
->check_sum
)) {
837 Name
": IMSM checksum %x != %x on %s\n",
838 check_sum
, __le32_to_cpu(super
->mpb
->check_sum
),
843 return load_imsm_disk(fd
, super
, devname
, 0);
846 static void free_imsm_disks(struct intel_super
*super
)
848 while (super
->disks
) {
849 struct dl
*d
= super
->disks
;
851 super
->disks
= d
->next
;
860 static void free_imsm(struct intel_super
*super
)
864 free_imsm_disks(super
);
869 static void free_super_imsm(struct supertype
*st
)
871 struct intel_super
*super
= st
->sb
;
880 static struct intel_super
*alloc_super(int creating_imsm
)
882 struct intel_super
*super
= malloc(sizeof(*super
));
885 memset(super
, 0, sizeof(*super
));
886 super
->creating_imsm
= creating_imsm
;
887 super
->current_vol
= -1;
894 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
895 char *devname
, int keep_fd
)
898 struct intel_super
*super
;
899 struct mdinfo
*sd
, *best
= NULL
;
906 /* check if this disk is a member of an active array */
907 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
911 if (sra
->array
.major_version
!= -1 ||
912 sra
->array
.minor_version
!= -2 ||
913 strcmp(sra
->text_version
, "imsm") != 0)
916 super
= alloc_super(0);
920 /* find the most up to date disk in this array */
921 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
922 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
923 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
928 rv
= load_imsm_mpb(dfd
, super
, NULL
);
932 gen
= __le32_to_cpu(super
->mpb
->generation_num
);
933 if (!best
|| gen
> bestgen
) {
948 /* load the most up to date anchor */
949 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
950 dfd
= dev_open(nm
, O_RDONLY
);
955 rv
= load_imsm_mpb(dfd
, super
, NULL
);
962 /* reset the disk list */
963 free_imsm_disks(super
);
965 /* populate disk list */
966 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
967 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
968 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
973 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
978 if (st
->subarray
[0]) {
979 if (atoi(st
->subarray
) <= super
->mpb
->num_raid_devs
)
980 super
->current_vol
= atoi(st
->subarray
);
986 if (st
->ss
== NULL
) {
987 st
->ss
= &super_imsm
;
988 st
->minor_version
= 0;
989 st
->max_devs
= IMSM_MAX_DEVICES
;
990 st
->container_dev
= fd2devnum(fd
);
997 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
999 struct intel_super
*super
;
1003 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1006 if (st
->subarray
[0])
1007 return 1; /* FIXME */
1009 super
= alloc_super(0);
1012 Name
": malloc of %zu failed.\n",
1017 rv
= load_imsm_mpb(fd
, super
, devname
);
1022 Name
": Failed to load all information "
1023 "sections on %s\n", devname
);
1029 if (st
->ss
== NULL
) {
1030 st
->ss
= &super_imsm
;
1031 st
->minor_version
= 0;
1032 st
->max_devs
= IMSM_MAX_DEVICES
;
1038 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1040 if (info
->level
== 1)
1042 return info
->chunk_size
>> 9;
1045 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1049 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1050 if (info
->level
== 1)
1056 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1057 unsigned long long size
, char *name
,
1058 char *homehost
, int *uuid
)
1060 /* We are creating a volume inside a pre-existing container.
1061 * so st->sb is already set.
1063 struct intel_super
*super
= st
->sb
;
1064 struct imsm_super
*mpb
= super
->mpb
;
1065 struct imsm_dev
*dev
;
1066 struct imsm_vol
*vol
;
1067 struct imsm_map
*map
;
1068 int idx
= mpb
->num_raid_devs
;
1070 unsigned long long array_blocks
;
1072 size_t size_old
, size_new
;
1074 if (mpb
->num_raid_devs
>= 2) {
1075 fprintf(stderr
, Name
": This imsm-container already has the "
1076 "maximum of 2 volumes\n");
1080 /* ensure the mpb is large enough for the new data */
1081 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1082 size_new
= disks_to_mpb_size(info
->nr_disks
);
1083 if (size_new
> size_old
) {
1085 size_t size_round
= ROUND_UP(size_new
, 512);
1087 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1088 fprintf(stderr
, Name
": could not allocate new mpb\n");
1091 memcpy(mpb_new
, mpb
, size_old
);
1094 super
->mpb
= mpb_new
;
1095 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1096 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1098 super
->current_vol
= idx
;
1099 sprintf(st
->subarray
, "%d", idx
);
1100 mpb
->num_raid_devs
++;
1101 dev
= get_imsm_dev(mpb
, idx
);
1102 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1103 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1104 info
->layout
, info
->chunk_size
,
1106 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1107 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1108 dev
->status
= __cpu_to_le32(0);
1109 dev
->reserved_blocks
= __cpu_to_le32(0);
1111 vol
->migr_state
= 0;
1114 for (i
= 0; i
< idx
; i
++) {
1115 struct imsm_dev
*prev
= get_imsm_dev(mpb
, i
);
1116 struct imsm_map
*pmap
= &prev
->vol
.map
[0];
1118 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1119 offset
+= IMSM_RESERVED_SECTORS
;
1122 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1123 map
->blocks_per_member
= __cpu_to_le32(info
->size
* 2);
1124 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1125 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1126 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1127 IMSM_T_STATE_NORMAL
;
1129 if (info
->level
== 1 && info
->raid_disks
> 2) {
1130 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1131 "in a raid1 volume\n");
1134 if (info
->level
== 10)
1135 map
->raid_level
= 1;
1137 map
->raid_level
= info
->level
;
1139 map
->num_members
= info
->raid_disks
;
1140 for (i
= 0; i
< map
->num_members
; i
++) {
1141 /* initialized in add_to_super */
1142 map
->disk_ord_tbl
[i
] = __cpu_to_le32(0);
1148 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1149 unsigned long long size
, char *name
,
1150 char *homehost
, int *uuid
)
1152 /* This is primarily called by Create when creating a new array.
1153 * We will then get add_to_super called for each component, and then
1154 * write_init_super called to write it out to each device.
1155 * For IMSM, Create can create on fresh devices or on a pre-existing
1157 * To create on a pre-existing array a different method will be called.
1158 * This one is just for fresh drives.
1160 struct intel_super
*super
;
1161 struct imsm_super
*mpb
;
1169 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1172 super
= alloc_super(1);
1175 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1176 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1181 memset(mpb
, 0, mpb_size
);
1183 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1184 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1185 strlen(MPB_VERSION_RAID5
));
1186 mpb
->mpb_size
= mpb_size
;
1192 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1193 int fd
, char *devname
)
1195 struct intel_super
*super
= st
->sb
;
1196 struct imsm_super
*mpb
= super
->mpb
;
1198 struct imsm_dev
*dev
;
1199 struct imsm_map
*map
;
1200 struct imsm_disk
*disk
;
1203 dev
= get_imsm_dev(mpb
, super
->current_vol
);
1204 map
= &dev
->vol
.map
[0];
1206 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1207 if (dl
->major
== dk
->major
&&
1208 dl
->minor
== dk
->minor
)
1210 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1213 map
->disk_ord_tbl
[dk
->number
] = __cpu_to_le32(dl
->index
);
1215 disk
= get_imsm_disk(mpb
, dl
->index
);
1216 status
= CONFIGURED_DISK
| USABLE_DISK
;
1217 disk
->status
= __cpu_to_le32(status
);
1220 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1221 int fd
, char *devname
)
1223 struct intel_super
*super
= st
->sb
;
1224 struct imsm_super
*mpb
= super
->mpb
;
1225 struct imsm_disk
*disk
;
1227 unsigned long long size
;
1232 if (super
->current_vol
>= 0) {
1233 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1238 dd
= malloc(sizeof(*dd
));
1241 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1244 memset(dd
, 0, sizeof(*dd
));
1245 dd
->major
= major(stb
.st_rdev
);
1246 dd
->minor
= minor(stb
.st_rdev
);
1247 dd
->index
= dk
->number
;
1248 dd
->devname
= devname
? strdup(devname
) : NULL
;
1249 dd
->next
= super
->disks
;
1251 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1254 Name
": failed to retrieve scsi serial "
1255 "using \'%s\' instead\n", devname
);
1256 strcpy((char *) dd
->serial
, devname
);
1259 if (mpb
->num_disks
<= dk
->number
)
1260 mpb
->num_disks
= dk
->number
+ 1;
1262 disk
= get_imsm_disk(mpb
, dk
->number
);
1263 get_dev_size(fd
, NULL
, &size
);
1265 status
= USABLE_DISK
| SPARE_DISK
;
1266 strcpy((char *) disk
->serial
, (char *) dd
->serial
);
1267 disk
->total_blocks
= __cpu_to_le32(size
);
1268 disk
->status
= __cpu_to_le32(status
);
1269 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1270 disk
->scsi_id
= __cpu_to_le32(id
);
1272 disk
->scsi_id
= __cpu_to_le32(0);
1274 /* update the family number if we are creating a container */
1275 if (super
->creating_imsm
)
1276 mpb
->family_num
= __cpu_to_le32(gen_imsm_checksum(mpb
));
1281 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1283 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1285 struct imsm_super
*mpb
= super
->mpb
;
1290 /* 'generation' is incremented everytime the metadata is written */
1291 generation
= __le32_to_cpu(mpb
->generation_num
);
1293 mpb
->generation_num
= __cpu_to_le32(generation
);
1295 /* recalculate checksum */
1296 sum
= gen_imsm_checksum(mpb
);
1297 mpb
->check_sum
= __cpu_to_le32(sum
);
1299 for (d
= super
->disks
; d
; d
= d
->next
) {
1300 if (store_imsm_mpb(d
->fd
, super
)) {
1301 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1302 __func__
, d
->major
, d
->minor
, strerror(errno
));
1314 static int write_init_super_imsm(struct supertype
*st
)
1316 return write_super_imsm(st
->sb
, 1);
1319 static int store_zero_imsm(struct supertype
*st
, int fd
)
1321 unsigned long long dsize
;
1324 get_dev_size(fd
, NULL
, &dsize
);
1326 /* first block is stored on second to last sector of the disk */
1327 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
1330 if (posix_memalign(&buf
, 512, 512) != 0)
1333 memset(buf
, 0, sizeof(buf
));
1334 if (write(fd
, buf
, sizeof(buf
)) != sizeof(buf
))
1339 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
1340 int layout
, int raiddisks
, int chunk
,
1341 unsigned long long size
, char *dev
,
1342 unsigned long long *freesize
,
1346 unsigned long long ldsize
;
1348 if (level
!= LEVEL_CONTAINER
)
1353 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1356 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
1357 dev
, strerror(errno
));
1360 if (!get_dev_size(fd
, dev
, &ldsize
)) {
1366 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
1371 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
1372 * FIX ME add ahci details
1374 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
1375 int layout
, int raiddisks
, int chunk
,
1376 unsigned long long size
, char *dev
,
1377 unsigned long long *freesize
,
1381 struct intel_super
*super
= st
->sb
;
1383 unsigned long long pos
= 0;
1384 unsigned long long maxsize
;
1388 if (level
== LEVEL_CONTAINER
)
1391 if (level
== 1 && raiddisks
> 2) {
1393 fprintf(stderr
, Name
": imsm does not support more "
1394 "than 2 in a raid1 configuration\n");
1398 /* We must have the container info already read in. */
1403 /* General test: make sure there is space for
1404 * 'raiddisks' device extents of size 'size' at a given
1407 unsigned long long minsize
= size
*2 /* convert to blocks */;
1408 unsigned long long start_offset
= ~0ULL;
1411 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
1412 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1417 e
= get_extents(super
, dl
);
1420 unsigned long long esize
;
1421 esize
= e
[i
].start
- pos
;
1422 if (esize
>= minsize
)
1424 if (found
&& start_offset
== ~0ULL) {
1427 } else if (found
&& pos
!= start_offset
) {
1431 pos
= e
[i
].start
+ e
[i
].size
;
1433 } while (e
[i
-1].size
);
1438 if (dcnt
< raiddisks
) {
1440 fprintf(stderr
, Name
": imsm: Not enough "
1441 "devices with space for this array "
1448 /* This device must be a member of the set */
1449 if (stat(dev
, &stb
) < 0)
1451 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
1453 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1454 if (dl
->major
== major(stb
.st_rdev
) &&
1455 dl
->minor
== minor(stb
.st_rdev
))
1460 fprintf(stderr
, Name
": %s is not in the "
1461 "same imsm set\n", dev
);
1464 e
= get_extents(super
, dl
);
1468 unsigned long long esize
;
1469 esize
= e
[i
].start
- pos
;
1470 if (esize
>= maxsize
)
1472 pos
= e
[i
].start
+ e
[i
].size
;
1474 } while (e
[i
-1].size
);
1475 *freesize
= maxsize
;
1480 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
1481 int raiddisks
, int chunk
, unsigned long long size
,
1482 char *dev
, unsigned long long *freesize
,
1488 /* if given unused devices create a container
1489 * if given given devices in a container create a member volume
1491 if (level
== LEVEL_CONTAINER
) {
1492 /* Must be a fresh device to add to a container */
1493 return validate_geometry_imsm_container(st
, level
, layout
,
1494 raiddisks
, chunk
, size
,
1500 /* creating in a given container */
1501 return validate_geometry_imsm_volume(st
, level
, layout
,
1502 raiddisks
, chunk
, size
,
1503 dev
, freesize
, verbose
);
1506 /* limit creation to the following levels */
1518 /* This device needs to be a device in an 'imsm' container */
1519 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1523 Name
": Cannot create this array on device %s\n",
1528 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
1530 fprintf(stderr
, Name
": Cannot open %s: %s\n",
1531 dev
, strerror(errno
));
1534 /* Well, it is in use by someone, maybe an 'imsm' container. */
1535 cfd
= open_container(fd
);
1539 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
1543 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
1545 if (sra
&& sra
->array
.major_version
== -1 &&
1546 strcmp(sra
->text_version
, "imsm") == 0) {
1547 /* This is a member of a imsm container. Load the container
1548 * and try to create a volume
1550 struct intel_super
*super
;
1552 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
1554 st
->container_dev
= fd2devnum(cfd
);
1556 return validate_geometry_imsm_volume(st
, level
, layout
,
1562 } else /* may belong to another container */
1568 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
1570 /* Given a container loaded by load_super_imsm_all,
1571 * extract information about all the arrays into
1574 * For each imsm_dev create an mdinfo, fill it in,
1575 * then look for matching devices in super->disks
1576 * and create appropriate device mdinfo.
1578 struct intel_super
*super
= st
->sb
;
1579 struct imsm_super
*mpb
= super
->mpb
;
1580 struct mdinfo
*rest
= NULL
;
1583 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1584 struct imsm_dev
*dev
= get_imsm_dev(mpb
, i
);
1585 struct imsm_vol
*vol
= &dev
->vol
;
1586 struct imsm_map
*map
= vol
->map
;
1587 struct mdinfo
*this;
1591 this = malloc(sizeof(*this));
1592 memset(this, 0, sizeof(*this));
1596 this->array
.level
= get_imsm_raid_level(map
);
1597 this->array
.raid_disks
= map
->num_members
;
1598 this->array
.layout
= imsm_level_to_layout(this->array
.level
);
1599 this->array
.md_minor
= -1;
1600 this->array
.ctime
= 0;
1601 this->array
.utime
= 0;
1602 this->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
1603 this->array
.state
= !vol
->dirty
;
1604 this->container_member
= i
;
1605 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
1606 this->resync_start
= 0;
1608 this->resync_start
= ~0ULL;
1610 strncpy(this->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
1611 this->name
[MAX_RAID_SERIAL_LEN
] = 0;
1613 sprintf(this->text_version
, "/%s/%d",
1614 devnum2devname(st
->container_dev
),
1615 this->container_member
);
1617 memset(this->uuid
, 0, sizeof(this->uuid
));
1619 sz
= __le32_to_cpu(dev
->size_high
);
1621 sz
+= __le32_to_cpu(dev
->size_low
);
1622 this->component_size
= sz
;
1623 this->array
.size
= this->component_size
/ 2;
1625 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
1626 struct imsm_disk
*disk
;
1627 struct mdinfo
*info_d
;
1632 idx
= __le32_to_cpu(map
->disk_ord_tbl
[slot
] & ~(0xff << 24));
1633 for (d
= super
->disks
; d
; d
= d
->next
)
1634 if (d
->index
== idx
)
1638 break; /* shouldn't this be continue ?? */
1640 info_d
= malloc(sizeof(*info_d
));
1642 break; /* ditto ?? */
1643 memset(info_d
, 0, sizeof(*info_d
));
1644 info_d
->next
= this->devs
;
1645 this->devs
= info_d
;
1647 disk
= get_imsm_disk(mpb
, idx
);
1648 s
= __le32_to_cpu(disk
->status
);
1650 info_d
->disk
.number
= d
->index
;
1651 info_d
->disk
.major
= d
->major
;
1652 info_d
->disk
.minor
= d
->minor
;
1653 info_d
->disk
.raid_disk
= slot
;
1654 info_d
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
1655 info_d
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
1656 info_d
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
1658 this->array
.working_disks
++;
1660 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
1661 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
1662 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
1664 strcpy(info_d
->name
, d
->devname
);
1672 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
1675 struct intel_super
*super
= c
->sb
;
1676 struct imsm_super
*mpb
= super
->mpb
;
1678 if (atoi(inst
) + 1 > mpb
->num_raid_devs
) {
1679 fprintf(stderr
, "%s: subarry index %d, out of range\n",
1680 __func__
, atoi(inst
));
1684 dprintf("imsm: open_new %s\n", inst
);
1685 a
->info
.container_member
= atoi(inst
);
1689 static __u8
imsm_check_degraded(struct imsm_super
*mpb
, int n
, int failed
)
1691 struct imsm_dev
*dev
= get_imsm_dev(mpb
, n
);
1692 struct imsm_map
*map
= dev
->vol
.map
;
1695 return map
->map_state
;
1697 switch (get_imsm_raid_level(map
)) {
1699 return IMSM_T_STATE_FAILED
;
1702 if (failed
< map
->num_members
)
1703 return IMSM_T_STATE_DEGRADED
;
1705 return IMSM_T_STATE_FAILED
;
1710 * check to see if any mirrors have failed,
1711 * otherwise we are degraded
1713 int device_per_mirror
= 2; /* FIXME is this always the case?
1714 * and are they always adjacent?
1719 for (i
= 0; i
< map
->num_members
; i
++) {
1720 int idx
= get_imsm_disk_idx(map
, i
);
1721 struct imsm_disk
*disk
= get_imsm_disk(mpb
, idx
);
1723 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
1726 if (failed
>= device_per_mirror
)
1727 return IMSM_T_STATE_FAILED
;
1729 /* reset 'failed' for next mirror set */
1730 if (!((i
+ 1) % device_per_mirror
))
1734 return IMSM_T_STATE_DEGRADED
;
1738 return IMSM_T_STATE_DEGRADED
;
1740 return IMSM_T_STATE_FAILED
;
1746 return map
->map_state
;
1749 static int imsm_count_failed(struct imsm_super
*mpb
, struct imsm_map
*map
)
1753 struct imsm_disk
*disk
;
1755 for (i
= 0; i
< map
->num_members
; i
++) {
1756 int idx
= get_imsm_disk_idx(map
, i
);
1758 disk
= get_imsm_disk(mpb
, idx
);
1759 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
1766 static void imsm_set_array_state(struct active_array
*a
, int consistent
)
1768 int inst
= a
->info
.container_member
;
1769 struct intel_super
*super
= a
->container
->sb
;
1770 struct imsm_dev
*dev
= get_imsm_dev(super
->mpb
, inst
);
1771 struct imsm_map
*map
= &dev
->vol
.map
[0];
1772 int dirty
= !consistent
;
1776 if (a
->resync_start
== ~0ULL) {
1777 failed
= imsm_count_failed(super
->mpb
, map
);
1778 map_state
= imsm_check_degraded(super
->mpb
, inst
, failed
);
1780 map_state
= IMSM_T_STATE_NORMAL
;
1781 if (map
->map_state
!= map_state
) {
1782 dprintf("imsm: map_state %d: %d\n",
1784 map
->map_state
= map_state
;
1785 super
->updates_pending
++;
1789 if (dev
->vol
.dirty
!= dirty
) {
1790 dprintf("imsm: mark '%s' (%llu)\n",
1791 dirty
?"dirty":"clean", a
->resync_start
);
1793 dev
->vol
.dirty
= dirty
;
1794 super
->updates_pending
++;
1798 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
1800 int inst
= a
->info
.container_member
;
1801 struct intel_super
*super
= a
->container
->sb
;
1802 struct imsm_dev
*dev
= get_imsm_dev(super
->mpb
, inst
);
1803 struct imsm_map
*map
= dev
->vol
.map
;
1804 struct imsm_disk
*disk
;
1807 int new_failure
= 0;
1809 if (n
> map
->num_members
)
1810 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
1811 n
, map
->num_members
- 1);
1816 dprintf("imsm: set_disk %d:%x\n", n
, state
);
1818 disk
= get_imsm_disk(super
->mpb
, get_imsm_disk_idx(map
, n
));
1820 /* check if we have seen this failure before */
1821 status
= __le32_to_cpu(disk
->status
);
1822 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
1823 status
|= FAILED_DISK
;
1824 disk
->status
= __cpu_to_le32(status
);
1829 * the number of failures have changed, count up 'failed' to determine
1830 * degraded / failed status
1832 if (new_failure
&& map
->map_state
!= IMSM_T_STATE_FAILED
)
1833 failed
= imsm_count_failed(super
->mpb
, map
);
1836 map
->map_state
= imsm_check_degraded(super
->mpb
, inst
, failed
);
1839 super
->updates_pending
++;
1842 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
1844 struct imsm_super
*mpb
= super
->mpb
;
1845 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
1846 unsigned long long dsize
;
1847 unsigned long long sectors
;
1849 get_dev_size(fd
, NULL
, &dsize
);
1851 if (mpb_size
> 512) {
1852 /* -1 to account for anchor */
1853 sectors
= mpb_sectors(mpb
) - 1;
1855 /* write the extended mpb to the sectors preceeding the anchor */
1856 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
1859 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
1863 /* first block is stored on second to last sector of the disk */
1864 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
1867 if (write(fd
, super
->buf
, 512) != 512)
1873 static void imsm_sync_metadata(struct supertype
*container
)
1875 struct intel_super
*super
= container
->sb
;
1877 if (!super
->updates_pending
)
1880 write_super_imsm(super
, 0);
1882 super
->updates_pending
= 0;
1885 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
1886 struct metadata_update
**updates
)
1889 * Take a device that is marked spare in the metadata and use it to
1890 * replace a failed/vacant slot in an array. There may be a case where
1891 * a device is failed in one array but active in a second.
1892 * imsm_process_update catches this case and does not clear the SPARE_DISK
1893 * flag, allowing the second array to start using the device on failure.
1894 * SPARE_DISK is cleared when all arrays are using a device.
1896 * FIXME: is this a valid use of SPARE_DISK?
1899 struct intel_super
*super
= a
->container
->sb
;
1900 struct imsm_super
*mpb
= super
->mpb
;
1901 int inst
= a
->info
.container_member
;
1902 struct imsm_dev
*dev
= get_imsm_dev(mpb
, inst
);
1903 struct imsm_map
*map
= dev
->vol
.map
;
1904 int failed
= a
->info
.array
.raid_disks
;
1905 struct mdinfo
*rv
= NULL
;
1908 struct metadata_update
*mu
;
1910 struct imsm_update_activate_spare
*u
;
1914 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
1915 if ((d
->curr_state
& DS_FAULTY
) &&
1917 /* wait for Removal to happen */
1919 if (d
->state_fd
>= 0)
1923 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
1924 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
1925 if (imsm_check_degraded(mpb
, inst
, failed
) != IMSM_T_STATE_DEGRADED
)
1928 /* For each slot, if it is not working, find a spare */
1930 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
1931 for (d
= a
->info
.devs
; d
; d
= d
->next
)
1932 if (d
->disk
.raid_disk
== i
)
1934 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
1935 if (d
&& (d
->state_fd
>= 0))
1938 /* OK, this device needs recovery. Find a spare */
1939 for ( ; dl
; dl
= dl
->next
) {
1940 unsigned long long esize
;
1941 unsigned long long pos
;
1944 struct imsm_disk
*disk
;
1949 /* If in this array, skip */
1950 for (d2
= a
->info
.devs
; d2
; d2
= d2
->next
)
1951 if (d2
->disk
.major
== dl
->major
&&
1952 d2
->disk
.minor
== dl
->minor
) {
1953 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
1959 /* is this unused device marked as a spare? */
1960 disk
= get_imsm_disk(mpb
, dl
->index
);
1961 if (!(__le32_to_cpu(disk
->status
) & SPARE_DISK
))
1964 /* We are allowed to use this device - is there space?
1965 * We need a->info.component_size sectors */
1966 ex
= get_extents(super
, dl
);
1968 dprintf("cannot get extents\n");
1974 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
1977 /* check that we can start at pba_of_lba0 with
1978 * a->info.component_size of space
1980 esize
= ex
[j
].start
- pos
;
1981 if (array_start
>= pos
&&
1982 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
1986 pos
= ex
[j
].start
+ ex
[j
].size
;
1989 } while (ex
[j
-1].size
);
1993 dprintf("%x:%x does not have %llu at %d\n",
1994 dl
->major
, dl
->minor
,
1995 a
->info
.component_size
,
1996 __le32_to_cpu(map
->pba_of_lba0
));
2001 /* found a usable disk with enough space */
2002 di
= malloc(sizeof(*di
));
2003 memset(di
, 0, sizeof(*di
));
2004 di
->disk
.number
= dl
->index
;
2005 di
->disk
.raid_disk
= i
;
2006 di
->disk
.major
= dl
->major
;
2007 di
->disk
.minor
= dl
->minor
;
2009 di
->data_offset
= array_start
;
2010 di
->component_size
= a
->info
.component_size
;
2011 di
->container_member
= inst
;
2015 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2023 /* No spares found */
2025 /* Now 'rv' has a list of devices to return.
2026 * Create a metadata_update record to update the
2027 * disk_ord_tbl for the array
2029 mu
= malloc(sizeof(*mu
));
2030 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2032 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
2033 mu
->next
= *updates
;
2034 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
2036 for (di
= rv
; di
; di
= di
->next
) {
2037 u
->type
= update_activate_spare
;
2038 u
->disk_idx
= di
->disk
.number
;
2039 u
->slot
= di
->disk
.raid_disk
;
2050 static int weight(unsigned int field
)
2054 for (weight
= 0; field
; weight
++)
2060 static void imsm_process_update(struct supertype
*st
,
2061 struct metadata_update
*update
)
2064 * crack open the metadata_update envelope to find the update record
2065 * update can be one of:
2066 * update_activate_spare - a spare device has replaced a failed
2067 * device in an array, update the disk_ord_tbl. If this disk is
2068 * present in all member arrays then also clear the SPARE_DISK
2071 struct intel_super
*super
= st
->sb
;
2072 struct imsm_super
*mpb
= super
->mpb
;
2073 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
2076 case update_activate_spare
: {
2077 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
2078 struct imsm_dev
*dev
= get_imsm_dev(mpb
, u
->array
);
2079 struct imsm_map
*map
= &dev
->vol
.map
[0];
2080 struct active_array
*a
;
2081 struct imsm_disk
*disk
;
2085 unsigned int members
;
2090 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2091 if (dl
->index
== u
->disk_idx
)
2095 fprintf(stderr
, "error: imsm_activate_spare passed "
2096 "an unknown disk_idx: %d\n", u
->disk_idx
);
2100 super
->updates_pending
++;
2102 victim
= get_imsm_disk_idx(map
, u
->slot
);
2103 map
->disk_ord_tbl
[u
->slot
] = __cpu_to_le32(u
->disk_idx
);
2104 disk
= get_imsm_disk(mpb
, u
->disk_idx
);
2105 status
= __le32_to_cpu(disk
->status
);
2106 status
|= CONFIGURED_DISK
;
2107 disk
->status
= __cpu_to_le32(status
);
2109 /* map unique/live arrays using the spare */
2112 for (a
= st
->arrays
; a
; a
= a
->next
) {
2113 int inst
= a
->info
.container_member
;
2115 dev
= get_imsm_dev(mpb
, inst
);
2116 map
= &dev
->vol
.map
[0];
2117 if (map
->raid_level
> 0)
2118 members
|= 1 << inst
;
2119 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2120 if (d
->disk
.major
== dl
->major
&&
2121 d
->disk
.minor
== dl
->minor
)
2125 /* until all arrays that can absorb this disk have absorbed
2126 * this disk it can still be considered a spare
2128 if (weight(found
) >= weight(members
)) {
2129 status
= __le32_to_cpu(disk
->status
);
2130 status
&= ~SPARE_DISK
;
2131 disk
->status
= __cpu_to_le32(status
);
2134 /* count arrays using the victim in the metadata */
2136 for (a
= st
->arrays
; a
; a
= a
->next
) {
2137 dev
= get_imsm_dev(mpb
, a
->info
.container_member
);
2138 map
= &dev
->vol
.map
[0];
2139 for (i
= 0; i
< map
->num_members
; i
++)
2140 if (victim
== get_imsm_disk_idx(map
, i
))
2144 /* clear some flags if the victim is no longer being
2147 disk
= get_imsm_disk(mpb
, victim
);
2149 status
= __le32_to_cpu(disk
->status
);
2150 status
&= ~(CONFIGURED_DISK
| USABLE_DISK
);
2151 disk
->status
= __cpu_to_le32(status
);
2157 struct superswitch super_imsm
= {
2159 .examine_super
= examine_super_imsm
,
2160 .brief_examine_super
= brief_examine_super_imsm
,
2161 .detail_super
= detail_super_imsm
,
2162 .brief_detail_super
= brief_detail_super_imsm
,
2163 .write_init_super
= write_init_super_imsm
,
2165 .match_home
= match_home_imsm
,
2166 .uuid_from_super
= uuid_from_super_imsm
,
2167 .getinfo_super
= getinfo_super_imsm
,
2168 .update_super
= update_super_imsm
,
2170 .avail_size
= avail_size_imsm
,
2172 .compare_super
= compare_super_imsm
,
2174 .load_super
= load_super_imsm
,
2175 .init_super
= init_super_imsm
,
2176 .add_to_super
= add_to_super_imsm
,
2177 .store_super
= store_zero_imsm
,
2178 .free_super
= free_super_imsm
,
2179 .match_metadata_desc
= match_metadata_desc_imsm
,
2180 .container_content
= container_content_imsm
,
2182 .validate_geometry
= validate_geometry_imsm
,
2186 .open_new
= imsm_open_new
,
2187 .load_super
= load_super_imsm
,
2188 .set_array_state
= imsm_set_array_state
,
2189 .set_disk
= imsm_set_disk
,
2190 .sync_metadata
= imsm_sync_metadata
,
2191 .activate_spare
= imsm_activate_spare
,
2192 .process_update
= imsm_process_update
,