2 * mdadm - Intel(R) Matrix Storage Manager Support
4 * Copyright (C) 2002-2008 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.
20 #define HAVE_STDINT_H 1
28 /* MPB == Metadata Parameter Block */
29 #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
30 #define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
31 #define MPB_VERSION_RAID0 "1.0.00"
32 #define MPB_VERSION_RAID1 "1.1.00"
33 #define MPB_VERSION_RAID5 "1.2.02"
34 #define MAX_SIGNATURE_LENGTH 32
35 #define MAX_RAID_SERIAL_LEN 16
36 #define MPB_SECTOR_CNT 418
37 #define IMSM_RESERVED_SECTORS 4096
39 /* Disk configuration info. */
40 #define IMSM_MAX_DEVICES 255
42 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
43 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
44 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
45 __u32 status
; /* 0xF0 - 0xF3 */
46 #define SPARE_DISK 0x01 /* Spare */
47 #define CONFIGURED_DISK 0x02 /* Member of some RaidDev */
48 #define FAILED_DISK 0x04 /* Permanent failure */
49 #define USABLE_DISK 0x08 /* Fully usable unless FAILED_DISK is set */
51 #define IMSM_DISK_FILLERS 5
52 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
55 /* RAID map configuration infos. */
57 __u32 pba_of_lba0
; /* start address of partition */
58 __u32 blocks_per_member
;/* blocks per member */
59 __u32 num_data_stripes
; /* number of data stripes */
60 __u16 blocks_per_strip
;
61 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
62 #define IMSM_T_STATE_NORMAL 0
63 #define IMSM_T_STATE_UNINITIALIZED 1
64 #define IMSM_T_STATE_DEGRADED 2 /* FIXME: is this correct? */
65 #define IMSM_T_STATE_FAILED 3 /* FIXME: is this correct? */
67 #define IMSM_T_RAID0 0
68 #define IMSM_T_RAID1 1
69 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
70 __u8 num_members
; /* number of member disks */
72 __u32 filler
[7]; /* expansion area */
73 #define IMSM_ORD_REBUILD (1 << 24)
74 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
75 * top byte contains some flags
77 } __attribute__ ((packed
));
82 __u8 migr_state
; /* Normal or Migrating */
83 __u8 migr_type
; /* Initializing, Rebuilding, ... */
87 struct imsm_map map
[1];
88 /* here comes another one if migr_state */
89 } __attribute__ ((packed
));
92 __u8 volume
[MAX_RAID_SERIAL_LEN
];
95 __u32 status
; /* Persistent RaidDev status */
96 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
97 #define IMSM_DEV_FILLERS 12
98 __u32 filler
[IMSM_DEV_FILLERS
];
100 } __attribute__ ((packed
));
103 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
104 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
105 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
106 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
107 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
108 __u32 error_log_size
; /* 0x30 - 0x33 in bytes */
109 __u32 attributes
; /* 0x34 - 0x37 */
110 __u8 num_disks
; /* 0x38 Number of configured disks */
111 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
112 __u8 error_log_pos
; /* 0x3A */
113 __u8 fill
[1]; /* 0x3B */
114 __u32 cache_size
; /* 0x3c - 0x40 in mb */
115 __u32 orig_family_num
; /* 0x40 - 0x43 original family num */
116 __u32 pwr_cycle_count
; /* 0x44 - 0x47 simulated power cycle count for array */
117 __u32 bbm_log_size
; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
118 #define IMSM_FILLERS 35
119 __u32 filler
[IMSM_FILLERS
]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
120 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
121 /* here comes imsm_dev[num_raid_devs] */
122 /* here comes BBM logs */
123 } __attribute__ ((packed
));
125 #define BBM_LOG_MAX_ENTRIES 254
127 struct bbm_log_entry
{
128 __u64 defective_block_start
;
129 #define UNREADABLE 0xFFFFFFFF
130 __u32 spare_block_offset
;
131 __u16 remapped_marked_count
;
133 } __attribute__ ((__packed__
));
136 __u32 signature
; /* 0xABADB10C */
138 __u32 reserved_spare_block_count
; /* 0 */
139 __u32 reserved
; /* 0xFFFF */
140 __u64 first_spare_lba
;
141 struct bbm_log_entry mapped_block_entries
[BBM_LOG_MAX_ENTRIES
];
142 } __attribute__ ((__packed__
));
146 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
149 static unsigned int sector_count(__u32 bytes
)
151 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
154 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
156 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
159 /* internal representation of IMSM metadata */
162 void *buf
; /* O_DIRECT buffer for reading/writing metadata */
163 struct imsm_super
*anchor
; /* immovable parameters */
165 size_t len
; /* size of the 'buf' allocation */
166 void *next_buf
; /* for realloc'ing buf from the manager */
168 int updates_pending
; /* count of pending updates for mdmon */
169 int creating_imsm
; /* flag to indicate container creation */
170 int current_vol
; /* index of raid device undergoing creation */
171 #define IMSM_MAX_RAID_DEVS 2
172 struct imsm_dev
*dev_tbl
[IMSM_MAX_RAID_DEVS
];
176 __u8 serial
[MAX_RAID_SERIAL_LEN
];
179 struct imsm_disk disk
;
182 struct dl
*add
; /* list of disks to add while mdmon active */
183 struct dl
*missing
; /* disks removed while we weren't looking */
184 struct bbm_log
*bbm_log
;
188 unsigned long long start
, size
;
191 /* definition of messages passed to imsm_process_update */
192 enum imsm_update_type
{
193 update_activate_spare
,
198 struct imsm_update_activate_spare
{
199 enum imsm_update_type type
;
203 struct imsm_update_activate_spare
*next
;
206 struct imsm_update_create_array
{
207 enum imsm_update_type type
;
212 struct imsm_update_add_disk
{
213 enum imsm_update_type type
;
216 static int imsm_env_devname_as_serial(void)
218 char *val
= getenv("IMSM_DEVNAME_AS_SERIAL");
220 if (val
&& atoi(val
) == 1)
227 static struct supertype
*match_metadata_desc_imsm(char *arg
)
229 struct supertype
*st
;
231 if (strcmp(arg
, "imsm") != 0 &&
232 strcmp(arg
, "default") != 0
236 st
= malloc(sizeof(*st
));
237 memset(st
, 0, sizeof(*st
));
238 st
->ss
= &super_imsm
;
239 st
->max_devs
= IMSM_MAX_DEVICES
;
240 st
->minor_version
= 0;
246 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
248 return &mpb
->sig
[MPB_SIG_LEN
];
252 /* retrieve a disk directly from the anchor when the anchor is known to be
253 * up-to-date, currently only at load time
255 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
257 if (index
>= mpb
->num_disks
)
259 return &mpb
->disk
[index
];
263 /* retrieve a disk from the parsed metadata */
264 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
268 for (d
= super
->disks
; d
; d
= d
->next
)
269 if (d
->index
== index
)
276 /* generate a checksum directly from the anchor when the anchor is known to be
277 * up-to-date, currently only at load or write_super after coalescing
279 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
281 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
282 __u32
*p
= (__u32
*) mpb
;
286 sum
+= __le32_to_cpu(*p
++);
288 return sum
- __le32_to_cpu(mpb
->check_sum
);
291 static size_t sizeof_imsm_map(struct imsm_map
*map
)
293 return sizeof(struct imsm_map
) + sizeof(__u32
) * (map
->num_members
- 1);
296 struct imsm_map
*get_imsm_map(struct imsm_dev
*dev
, int second_map
)
298 struct imsm_map
*map
= &dev
->vol
.map
[0];
300 if (second_map
&& !dev
->vol
.migr_state
)
302 else if (second_map
) {
305 return ptr
+ sizeof_imsm_map(map
);
311 /* return the size of the device.
312 * migr_state increases the returned size if map[0] were to be duplicated
314 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
, int migr_state
)
316 size_t size
= sizeof(*dev
) - sizeof(struct imsm_map
) +
317 sizeof_imsm_map(get_imsm_map(dev
, 0));
319 /* migrating means an additional map */
320 if (dev
->vol
.migr_state
)
321 size
+= sizeof_imsm_map(get_imsm_map(dev
, 1));
323 size
+= sizeof_imsm_map(get_imsm_map(dev
, 0));
328 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
334 if (index
>= mpb
->num_raid_devs
)
337 /* devices start after all disks */
338 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
340 for (i
= 0; i
<= index
; i
++)
342 return _mpb
+ offset
;
344 offset
+= sizeof_imsm_dev(_mpb
+ offset
, 0);
349 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
351 if (index
>= super
->anchor
->num_raid_devs
)
353 return super
->dev_tbl
[index
];
356 static __u32
get_imsm_ord_tbl_ent(struct imsm_dev
*dev
, int slot
)
358 struct imsm_map
*map
;
360 if (dev
->vol
.migr_state
)
361 map
= get_imsm_map(dev
, 1);
363 map
= get_imsm_map(dev
, 0);
365 /* top byte identifies disk under rebuild */
366 return __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
369 #define ord_to_idx(ord) (((ord) << 8) >> 8)
370 static __u32
get_imsm_disk_idx(struct imsm_dev
*dev
, int slot
)
372 __u32 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
374 return ord_to_idx(ord
);
377 static void set_imsm_ord_tbl_ent(struct imsm_map
*map
, int slot
, __u32 ord
)
379 map
->disk_ord_tbl
[slot
] = __cpu_to_le32(ord
);
382 static int get_imsm_raid_level(struct imsm_map
*map
)
384 if (map
->raid_level
== 1) {
385 if (map
->num_members
== 2)
391 return map
->raid_level
;
394 static int cmp_extent(const void *av
, const void *bv
)
396 const struct extent
*a
= av
;
397 const struct extent
*b
= bv
;
398 if (a
->start
< b
->start
)
400 if (a
->start
> b
->start
)
405 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
407 /* find a list of used extents on the given physical device */
408 struct extent
*rv
, *e
;
411 __u32 reservation
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
413 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
414 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
415 struct imsm_map
*map
= get_imsm_map(dev
, 0);
417 for (j
= 0; j
< map
->num_members
; j
++) {
418 __u32 index
= get_imsm_disk_idx(dev
, j
);
420 if (index
== dl
->index
)
424 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
429 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
430 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
431 struct imsm_map
*map
= get_imsm_map(dev
, 0);
433 for (j
= 0; j
< map
->num_members
; j
++) {
434 __u32 index
= get_imsm_disk_idx(dev
, j
);
436 if (index
== dl
->index
) {
437 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
438 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
443 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
445 /* determine the start of the metadata
446 * when no raid devices are defined use the default
447 * ...otherwise allow the metadata to truncate the value
448 * as is the case with older versions of imsm
451 struct extent
*last
= &rv
[memberships
- 1];
454 remainder
= __le32_to_cpu(dl
->disk
.total_blocks
) -
455 (last
->start
+ last
->size
);
456 if (reservation
> remainder
)
457 reservation
= remainder
;
459 e
->start
= __le32_to_cpu(dl
->disk
.total_blocks
) - reservation
;
464 /* try to determine how much space is reserved for metadata from
465 * the last get_extents() entry, otherwise fallback to the
468 static __u32
imsm_reserved_sectors(struct intel_super
*super
, struct dl
*dl
)
474 /* for spares just return a minimal reservation which will grow
475 * once the spare is picked up by an array
478 return MPB_SECTOR_CNT
;
480 e
= get_extents(super
, dl
);
482 return MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
484 /* scroll to last entry */
485 for (i
= 0; e
[i
].size
; i
++)
488 rv
= __le32_to_cpu(dl
->disk
.total_blocks
) - e
[i
].start
;
496 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
);
498 static void print_imsm_dev(struct imsm_dev
*dev
, int index
)
502 struct imsm_map
*map
= get_imsm_map(dev
, 0);
506 printf("[%.16s]:\n", dev
->volume
);
507 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
508 printf(" Members : %d\n", map
->num_members
);
509 for (slot
= 0; slot
< map
->num_members
; slot
++)
510 if (index
== get_imsm_disk_idx(dev
, slot
))
512 if (slot
< map
->num_members
) {
513 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
514 printf(" This Slot : %d%s\n", slot
,
515 ord
& IMSM_ORD_REBUILD
? " (out-of-sync)" : "");
517 printf(" This Slot : ?\n");
518 sz
= __le32_to_cpu(dev
->size_high
);
520 sz
+= __le32_to_cpu(dev
->size_low
);
521 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
522 human_size(sz
* 512));
523 sz
= __le32_to_cpu(map
->blocks_per_member
);
524 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
525 human_size(sz
* 512));
526 printf(" Sector Offset : %u\n",
527 __le32_to_cpu(map
->pba_of_lba0
));
528 printf(" Num Stripes : %u\n",
529 __le32_to_cpu(map
->num_data_stripes
));
530 printf(" Chunk Size : %u KiB\n",
531 __le16_to_cpu(map
->blocks_per_strip
) / 2);
532 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
533 printf(" Migrate State : %s", dev
->vol
.migr_state
? "migrating" : "idle");
534 if (dev
->vol
.migr_state
)
535 printf(": %s", dev
->vol
.migr_type
? "rebuilding" : "initializing");
537 printf(" Map State : %s", map_state_str
[map
->map_state
]);
538 if (dev
->vol
.migr_state
) {
539 struct imsm_map
*map
= get_imsm_map(dev
, 1);
540 printf(" <-- %s", map_state_str
[map
->map_state
]);
543 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
546 static void print_imsm_disk(struct imsm_super
*mpb
, int index
, __u32 reserved
)
548 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
549 char str
[MAX_RAID_SERIAL_LEN
+ 1];
557 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
558 printf(" Disk%02d Serial : %s\n", index
, str
);
559 s
= __le32_to_cpu(disk
->status
);
560 printf(" State :%s%s%s%s\n", s
&SPARE_DISK
? " spare" : "",
561 s
&CONFIGURED_DISK
? " active" : "",
562 s
&FAILED_DISK
? " failed" : "",
563 s
&USABLE_DISK
? " usable" : "");
564 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
565 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
566 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
567 human_size(sz
* 512));
570 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
572 struct intel_super
*super
= st
->sb
;
573 struct imsm_super
*mpb
= super
->anchor
;
574 char str
[MAX_SIGNATURE_LENGTH
];
579 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
582 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
583 printf(" Magic : %s\n", str
);
584 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
585 printf(" Version : %s\n", get_imsm_version(mpb
));
586 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
587 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
588 getinfo_super_imsm(st
, &info
);
589 fname_from_uuid(st
, &info
, nbuf
,'-');
590 printf(" UUID : %s\n", nbuf
+ 5);
591 sum
= __le32_to_cpu(mpb
->check_sum
);
592 printf(" Checksum : %08x %s\n", sum
,
593 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
594 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
595 printf(" Disks : %d\n", mpb
->num_disks
);
596 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
597 print_imsm_disk(mpb
, super
->disks
->index
, reserved
);
598 if (super
->bbm_log
) {
599 struct bbm_log
*log
= super
->bbm_log
;
602 printf("Bad Block Management Log:\n");
603 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
604 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
605 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
606 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
607 printf(" First Spare : %llx\n", __le64_to_cpu(log
->first_spare_lba
));
609 for (i
= 0; i
< mpb
->num_raid_devs
; i
++)
610 print_imsm_dev(__get_imsm_dev(mpb
, i
), super
->disks
->index
);
611 for (i
= 0; i
< mpb
->num_disks
; i
++) {
612 if (i
== super
->disks
->index
)
614 print_imsm_disk(mpb
, i
, reserved
);
618 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
);
620 static void brief_examine_super_imsm(struct supertype
*st
)
622 /* We just write a generic IMSM ARRAY entry */
625 struct intel_super
*super
= st
->sb
;
628 if (!super
->anchor
->num_raid_devs
)
631 getinfo_super_imsm(st
, &info
);
632 fname_from_uuid(st
, &info
, nbuf
,'-');
633 printf("ARRAY /dev/imsm metadata=imsm auto=md UUID=%s\n", nbuf
+ 5);
634 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
635 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
637 super
->current_vol
= i
;
638 getinfo_super_imsm(st
, &info
);
639 fname_from_uuid(st
, &info
, nbuf
,'-');
640 printf("ARRAY /dev/md/%.16s container=/dev/imsm member=%d auto=mdp UUID=%s\n",
641 dev
->volume
, i
, nbuf
+ 5);
645 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
647 printf("%s\n", __FUNCTION__
);
650 static void brief_detail_super_imsm(struct supertype
*st
)
654 getinfo_super_imsm(st
, &info
);
655 fname_from_uuid(st
, &info
, nbuf
,'-');
656 printf(" UUID=%s", nbuf
+ 5);
660 static int match_home_imsm(struct supertype
*st
, char *homehost
)
662 printf("%s\n", __FUNCTION__
);
667 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
669 /* The uuid returned here is used for:
670 * uuid to put into bitmap file (Create, Grow)
671 * uuid for backup header when saving critical section (Grow)
672 * comparing uuids when re-adding a device into an array
673 * In these cases the uuid required is that of the data-array,
674 * not the device-set.
675 * uuid to recognise same set when adding a missing device back
676 * to an array. This is a uuid for the device-set.
678 * For each of these we can make do with a truncated
679 * or hashed uuid rather than the original, as long as
681 * In each case the uuid required is that of the data-array,
682 * not the device-set.
684 /* imsm does not track uuid's so we synthesis one using sha1 on
685 * - The signature (Which is constant for all imsm array, but no matter)
686 * - the family_num of the container
687 * - the index number of the volume
688 * - the 'serial' number of the volume.
689 * Hopefully these are all constant.
691 struct intel_super
*super
= st
->sb
;
695 struct imsm_dev
*dev
= NULL
;
698 sha1_process_bytes(super
->anchor
->sig
, MAX_SIGNATURE_LENGTH
, &ctx
);
699 sha1_process_bytes(&super
->anchor
->family_num
, sizeof(__u32
), &ctx
);
700 if (super
->current_vol
>= 0)
701 dev
= get_imsm_dev(super
, super
->current_vol
);
703 __u32 vol
= super
->current_vol
;
704 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
705 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
707 sha1_finish_ctx(&ctx
, buf
);
708 memcpy(uuid
, buf
, 4*4);
713 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
715 __u8
*v
= get_imsm_version(mpb
);
716 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
717 char major
[] = { 0, 0, 0 };
718 char minor
[] = { 0 ,0, 0 };
719 char patch
[] = { 0, 0, 0 };
720 char *ver_parse
[] = { major
, minor
, patch
};
724 while (*v
!= '\0' && v
< end
) {
725 if (*v
!= '.' && j
< 2)
726 ver_parse
[i
][j
++] = *v
;
734 *m
= strtol(minor
, NULL
, 0);
735 *p
= strtol(patch
, NULL
, 0);
739 static int imsm_level_to_layout(int level
)
747 return ALGORITHM_LEFT_ASYMMETRIC
;
754 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
756 struct intel_super
*super
= st
->sb
;
757 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
758 struct imsm_map
*map
= get_imsm_map(dev
, 0);
760 info
->container_member
= super
->current_vol
;
761 info
->array
.raid_disks
= map
->num_members
;
762 info
->array
.level
= get_imsm_raid_level(map
);
763 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
764 info
->array
.md_minor
= -1;
765 info
->array
.ctime
= 0;
766 info
->array
.utime
= 0;
767 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
768 info
->array
.state
= !dev
->vol
.dirty
;
770 info
->disk
.major
= 0;
771 info
->disk
.minor
= 0;
773 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
774 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
775 memset(info
->uuid
, 0, sizeof(info
->uuid
));
777 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
778 info
->resync_start
= 0;
779 else if (dev
->vol
.migr_state
)
780 info
->resync_start
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
782 info
->resync_start
= ~0ULL;
784 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
785 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
787 info
->array
.major_version
= -1;
788 info
->array
.minor_version
= -2;
789 sprintf(info
->text_version
, "/%s/%d",
790 devnum2devname(st
->container_dev
),
791 info
->container_member
);
792 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
793 uuid_from_super_imsm(st
, info
->uuid
);
797 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
799 struct intel_super
*super
= st
->sb
;
800 struct imsm_disk
*disk
;
803 if (super
->current_vol
>= 0) {
804 getinfo_super_imsm_volume(st
, info
);
808 /* Set raid_disks to zero so that Assemble will always pull in valid
811 info
->array
.raid_disks
= 0;
812 info
->array
.level
= LEVEL_CONTAINER
;
813 info
->array
.layout
= 0;
814 info
->array
.md_minor
= -1;
815 info
->array
.ctime
= 0; /* N/A for imsm */
816 info
->array
.utime
= 0;
817 info
->array
.chunk_size
= 0;
819 info
->disk
.major
= 0;
820 info
->disk
.minor
= 0;
821 info
->disk
.raid_disk
= -1;
822 info
->reshape_active
= 0;
823 info
->array
.major_version
= -1;
824 info
->array
.minor_version
= -2;
825 strcpy(info
->text_version
, "imsm");
826 info
->safe_mode_delay
= 0;
827 info
->disk
.number
= -1;
828 info
->disk
.state
= 0;
832 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
834 disk
= &super
->disks
->disk
;
835 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
836 info
->component_size
= reserved
;
837 s
= __le32_to_cpu(disk
->status
);
838 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
839 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
840 info
->disk
.state
|= s
& SPARE_DISK
? 0 : (1 << MD_DISK_SYNC
);
843 /* only call uuid_from_super_imsm when this disk is part of a populated container,
844 * ->compare_super may have updated the 'num_raid_devs' field for spares
846 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
847 uuid_from_super_imsm(st
, info
->uuid
);
849 memcpy(info
->uuid
, uuid_match_any
, sizeof(int[4]));
852 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
853 char *update
, char *devname
, int verbose
,
854 int uuid_set
, char *homehost
)
858 /* For 'assemble' and 'force' we need to return non-zero if any
859 * change was made. For others, the return value is ignored.
860 * Update options are:
861 * force-one : This device looks a bit old but needs to be included,
862 * update age info appropriately.
863 * assemble: clear any 'faulty' flag to allow this device to
865 * force-array: Array is degraded but being forced, mark it clean
866 * if that will be needed to assemble it.
868 * newdev: not used ????
869 * grow: Array has gained a new device - this is currently for
871 * resync: mark as dirty so a resync will happen.
872 * name: update the name - preserving the homehost
874 * Following are not relevant for this imsm:
875 * sparc2.2 : update from old dodgey metadata
876 * super-minor: change the preferred_minor number
877 * summaries: update redundant counters.
878 * uuid: Change the uuid of the array to match watch is given
879 * homehost: update the recorded homehost
880 * _reshape_progress: record new reshape_progress position.
883 //struct intel_super *super = st->sb;
884 //struct imsm_super *mpb = super->mpb;
886 if (strcmp(update
, "grow") == 0) {
888 if (strcmp(update
, "resync") == 0) {
889 /* dev->vol.dirty = 1; */
892 /* IMSM has no concept of UUID or homehost */
897 static size_t disks_to_mpb_size(int disks
)
901 size
= sizeof(struct imsm_super
);
902 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
903 size
+= 2 * sizeof(struct imsm_dev
);
904 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
905 size
+= (4 - 2) * sizeof(struct imsm_map
);
906 /* 4 possible disk_ord_tbl's */
907 size
+= 4 * (disks
- 1) * sizeof(__u32
);
912 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
914 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
917 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
920 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
924 * 0 same, or first was empty, and second was copied
925 * 1 second had wrong number
929 struct intel_super
*first
= st
->sb
;
930 struct intel_super
*sec
= tst
->sb
;
938 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
941 /* if an anchor does not have num_raid_devs set then it is a free
944 if (first
->anchor
->num_raid_devs
> 0 &&
945 sec
->anchor
->num_raid_devs
> 0) {
946 if (first
->anchor
->family_num
!= sec
->anchor
->family_num
)
950 /* if 'first' is a spare promote it to a populated mpb with sec's
953 if (first
->anchor
->num_raid_devs
== 0 &&
954 sec
->anchor
->num_raid_devs
> 0) {
957 /* we need to copy raid device info from sec if an allocation
958 * fails here we don't associate the spare
960 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
961 first
->dev_tbl
[i
] = malloc(sizeof(struct imsm_dev
));
962 if (!first
->dev_tbl
) {
964 free(first
->dev_tbl
[i
]);
965 first
->dev_tbl
[i
] = NULL
;
967 fprintf(stderr
, "imsm: failed to associate spare\n");
970 *first
->dev_tbl
[i
] = *sec
->dev_tbl
[i
];
973 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
974 first
->anchor
->family_num
= sec
->anchor
->family_num
;
980 static void fd2devname(int fd
, char *name
)
989 if (fstat(fd
, &st
) != 0)
991 sprintf(path
, "/sys/dev/block/%d:%d",
992 major(st
.st_rdev
), minor(st
.st_rdev
));
994 rv
= readlink(path
, dname
, sizeof(dname
));
999 nm
= strrchr(dname
, '/');
1001 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
1005 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
1007 static int imsm_read_serial(int fd
, char *devname
,
1008 __u8 serial
[MAX_RAID_SERIAL_LEN
])
1010 unsigned char scsi_serial
[255];
1016 memset(scsi_serial
, 0, sizeof(scsi_serial
));
1018 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
1020 if (rv
&& imsm_env_devname_as_serial()) {
1021 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1022 fd2devname(fd
, (char *) serial
);
1029 Name
": Failed to retrieve serial for %s\n",
1034 /* trim leading whitespace */
1035 rsp_len
= scsi_serial
[3];
1036 rsp_buf
= (char *) &scsi_serial
[4];
1041 /* truncate len to the end of rsp_buf if necessary */
1042 if (c
+ MAX_RAID_SERIAL_LEN
> rsp_buf
+ rsp_len
)
1043 len
= rsp_len
- (c
- rsp_buf
);
1045 len
= MAX_RAID_SERIAL_LEN
;
1047 /* initialize the buffer and copy rsp_buf characters */
1048 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1049 memcpy(serial
, c
, len
);
1051 /* trim trailing whitespace starting with the last character copied */
1052 c
= (char *) &serial
[len
- 1];
1053 while (isspace(*c
) || *c
== '\0')
1059 static int serialcmp(__u8
*s1
, __u8
*s2
)
1061 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
1064 static void serialcpy(__u8
*dest
, __u8
*src
)
1066 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
1070 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
1077 __u8 serial
[MAX_RAID_SERIAL_LEN
];
1079 rv
= imsm_read_serial(fd
, devname
, serial
);
1084 /* check if this is a disk we have seen before. it may be a spare in
1085 * super->disks while the current anchor believes it is a raid member,
1086 * check if we need to update dl->index
1088 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1089 if (serialcmp(dl
->serial
, serial
) == 0)
1093 dl
= malloc(sizeof(*dl
));
1100 Name
": failed to allocate disk buffer for %s\n",
1107 dl
->major
= major(stb
.st_rdev
);
1108 dl
->minor
= minor(stb
.st_rdev
);
1109 dl
->next
= super
->disks
;
1110 dl
->fd
= keep_fd
? fd
: -1;
1111 dl
->devname
= devname
? strdup(devname
) : NULL
;
1112 serialcpy(dl
->serial
, serial
);
1114 } else if (keep_fd
) {
1119 /* look up this disk's index in the current anchor */
1120 for (i
= 0; i
< super
->anchor
->num_disks
; i
++) {
1121 struct imsm_disk
*disk_iter
;
1123 disk_iter
= __get_imsm_disk(super
->anchor
, i
);
1125 if (serialcmp(disk_iter
->serial
, dl
->serial
) == 0) {
1128 dl
->disk
= *disk_iter
;
1129 status
= __le32_to_cpu(dl
->disk
.status
);
1130 /* only set index on disks that are a member of a
1131 * populated contianer, i.e. one with raid_devs
1133 if (status
& FAILED_DISK
)
1135 else if (status
& SPARE_DISK
)
1144 /* no match, maybe a stale failed drive */
1145 if (i
== super
->anchor
->num_disks
&& dl
->index
>= 0) {
1146 dl
->disk
= *__get_imsm_disk(super
->anchor
, dl
->index
);
1147 if (__le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
1157 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
1159 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
1163 /* When migrating map0 contains the 'destination' state while map1
1164 * contains the current state. When not migrating map0 contains the
1165 * current state. This routine assumes that map[0].map_state is set to
1166 * the current array state before being called.
1168 * Migration is indicated by one of the following states
1169 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
1170 * 2/ Initialize (migr_state=1 migr_type=0 map0state=normal
1171 * map1state=unitialized)
1172 * 3/ Verify (Resync) (migr_state=1 migr_type=1 map0state=normal
1174 * 4/ Rebuild (migr_state=1 migr_type=1 map0state=normal
1175 * map1state=degraded)
1177 static void migrate(struct imsm_dev
*dev
, __u8 to_state
, int rebuild_resync
)
1179 struct imsm_map
*dest
;
1180 struct imsm_map
*src
= get_imsm_map(dev
, 0);
1182 dev
->vol
.migr_state
= 1;
1183 dev
->vol
.migr_type
= rebuild_resync
;
1184 dev
->vol
.curr_migr_unit
= 0;
1185 dest
= get_imsm_map(dev
, 1);
1187 memcpy(dest
, src
, sizeof_imsm_map(src
));
1188 src
->map_state
= to_state
;
1191 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
1193 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1195 dev
->vol
.migr_state
= 0;
1196 dev
->vol
.curr_migr_unit
= 0;
1197 map
->map_state
= map_state
;
1201 static int parse_raid_devices(struct intel_super
*super
)
1204 struct imsm_dev
*dev_new
;
1205 size_t len
, len_migr
;
1206 size_t space_needed
= 0;
1207 struct imsm_super
*mpb
= super
->anchor
;
1209 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1210 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
1212 len
= sizeof_imsm_dev(dev_iter
, 0);
1213 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
1215 space_needed
+= len_migr
- len
;
1217 dev_new
= malloc(len_migr
);
1220 imsm_copy_dev(dev_new
, dev_iter
);
1221 super
->dev_tbl
[i
] = dev_new
;
1224 /* ensure that super->buf is large enough when all raid devices
1227 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
1230 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
1231 if (posix_memalign(&buf
, 512, len
) != 0)
1234 memcpy(buf
, super
->buf
, len
);
1243 /* retrieve a pointer to the bbm log which starts after all raid devices */
1244 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
1248 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
1250 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
1256 static void __free_imsm(struct intel_super
*super
, int free_disks
);
1258 /* load_imsm_mpb - read matrix metadata
1259 * allocates super->mpb to be freed by free_super
1261 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
1263 unsigned long long dsize
;
1264 unsigned long long sectors
;
1266 struct imsm_super
*anchor
;
1270 get_dev_size(fd
, NULL
, &dsize
);
1272 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
1275 Name
": Cannot seek to anchor block on %s: %s\n",
1276 devname
, strerror(errno
));
1280 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
1283 Name
": Failed to allocate imsm anchor buffer"
1284 " on %s\n", devname
);
1287 if (read(fd
, anchor
, 512) != 512) {
1290 Name
": Cannot read anchor block on %s: %s\n",
1291 devname
, strerror(errno
));
1296 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
1299 Name
": no IMSM anchor on %s\n", devname
);
1304 __free_imsm(super
, 0);
1305 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
1306 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
1309 Name
": unable to allocate %zu byte mpb buffer\n",
1314 memcpy(super
->buf
, anchor
, 512);
1316 sectors
= mpb_sectors(anchor
) - 1;
1319 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1321 rc
= parse_raid_devices(super
);
1325 /* read the extended mpb */
1326 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
1329 Name
": Cannot seek to extended mpb on %s: %s\n",
1330 devname
, strerror(errno
));
1334 if (read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
1337 Name
": Cannot read extended mpb on %s: %s\n",
1338 devname
, strerror(errno
));
1342 check_sum
= __gen_imsm_checksum(super
->anchor
);
1343 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
1346 Name
": IMSM checksum %x != %x on %s\n",
1347 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
1352 /* FIXME the BBM log is disk specific so we cannot use this global
1353 * buffer for all disks. Ok for now since we only look at the global
1354 * bbm_log_size parameter to gate assembly
1356 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
1358 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1360 rc
= parse_raid_devices(super
);
1365 static void __free_imsm_disk(struct dl
*d
)
1374 static void free_imsm_disks(struct intel_super
*super
)
1378 while (super
->disks
) {
1380 super
->disks
= d
->next
;
1381 __free_imsm_disk(d
);
1383 while (super
->missing
) {
1385 super
->missing
= d
->next
;
1386 __free_imsm_disk(d
);
1391 /* free all the pieces hanging off of a super pointer */
1392 static void __free_imsm(struct intel_super
*super
, int free_disks
)
1401 free_imsm_disks(super
);
1402 for (i
= 0; i
< IMSM_MAX_RAID_DEVS
; i
++)
1403 if (super
->dev_tbl
[i
]) {
1404 free(super
->dev_tbl
[i
]);
1405 super
->dev_tbl
[i
] = NULL
;
1409 static void free_imsm(struct intel_super
*super
)
1411 __free_imsm(super
, 1);
1415 static void free_super_imsm(struct supertype
*st
)
1417 struct intel_super
*super
= st
->sb
;
1426 static struct intel_super
*alloc_super(int creating_imsm
)
1428 struct intel_super
*super
= malloc(sizeof(*super
));
1431 memset(super
, 0, sizeof(*super
));
1432 super
->creating_imsm
= creating_imsm
;
1433 super
->current_vol
= -1;
1440 /* find_missing - helper routine for load_super_imsm_all that identifies
1441 * disks that have disappeared from the system. This routine relies on
1442 * the mpb being uptodate, which it is at load time.
1444 static int find_missing(struct intel_super
*super
)
1447 struct imsm_super
*mpb
= super
->anchor
;
1449 struct imsm_disk
*disk
;
1452 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1453 disk
= __get_imsm_disk(mpb
, i
);
1454 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1455 if (serialcmp(dl
->disk
.serial
, disk
->serial
) == 0)
1459 /* ok we have a 'disk' without a live entry in
1462 status
= __le32_to_cpu(disk
->status
);
1463 if (status
& FAILED_DISK
|| !(status
& USABLE_DISK
))
1464 continue; /* never mind, already marked */
1466 dl
= malloc(sizeof(*dl
));
1472 dl
->devname
= strdup("missing");
1474 serialcpy(dl
->serial
, disk
->serial
);
1476 dl
->next
= super
->missing
;
1477 super
->missing
= dl
;
1483 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
1484 char *devname
, int keep_fd
)
1487 struct intel_super
*super
;
1488 struct mdinfo
*sd
, *best
= NULL
;
1495 /* check if this disk is a member of an active array */
1496 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
1500 if (sra
->array
.major_version
!= -1 ||
1501 sra
->array
.minor_version
!= -2 ||
1502 strcmp(sra
->text_version
, "imsm") != 0)
1505 super
= alloc_super(0);
1509 /* find the most up to date disk in this array, skipping spares */
1510 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1511 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1512 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1517 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1521 if (super
->anchor
->num_raid_devs
== 0)
1524 gen
= __le32_to_cpu(super
->anchor
->generation_num
);
1525 if (!best
|| gen
> bestgen
) {
1540 /* load the most up to date anchor */
1541 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
1542 dfd
= dev_open(nm
, O_RDONLY
);
1547 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1554 /* re-parse the disk list with the current anchor */
1555 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1556 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1557 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1562 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
1568 if (find_missing(super
) != 0) {
1573 if (st
->subarray
[0]) {
1574 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
1575 super
->current_vol
= atoi(st
->subarray
);
1581 st
->container_dev
= fd2devnum(fd
);
1582 if (st
->ss
== NULL
) {
1583 st
->ss
= &super_imsm
;
1584 st
->minor_version
= 0;
1585 st
->max_devs
= IMSM_MAX_DEVICES
;
1587 st
->loaded_container
= 1;
1593 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
1595 struct intel_super
*super
;
1599 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1602 if (st
->subarray
[0])
1603 return 1; /* FIXME */
1605 super
= alloc_super(0);
1608 Name
": malloc of %zu failed.\n",
1613 rv
= load_imsm_mpb(fd
, super
, devname
);
1618 Name
": Failed to load all information "
1619 "sections on %s\n", devname
);
1625 if (st
->ss
== NULL
) {
1626 st
->ss
= &super_imsm
;
1627 st
->minor_version
= 0;
1628 st
->max_devs
= IMSM_MAX_DEVICES
;
1630 st
->loaded_container
= 0;
1635 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1637 if (info
->level
== 1)
1639 return info
->chunk_size
>> 9;
1642 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1646 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1647 if (info
->level
== 1)
1653 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
1655 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
1658 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1659 unsigned long long size
, char *name
,
1660 char *homehost
, int *uuid
)
1662 /* We are creating a volume inside a pre-existing container.
1663 * so st->sb is already set.
1665 struct intel_super
*super
= st
->sb
;
1666 struct imsm_super
*mpb
= super
->anchor
;
1667 struct imsm_dev
*dev
;
1668 struct imsm_vol
*vol
;
1669 struct imsm_map
*map
;
1670 int idx
= mpb
->num_raid_devs
;
1672 unsigned long long array_blocks
;
1674 size_t size_old
, size_new
;
1676 if (mpb
->num_raid_devs
>= 2) {
1677 fprintf(stderr
, Name
": This imsm-container already has the "
1678 "maximum of 2 volumes\n");
1682 /* ensure the mpb is large enough for the new data */
1683 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1684 size_new
= disks_to_mpb_size(info
->nr_disks
);
1685 if (size_new
> size_old
) {
1687 size_t size_round
= ROUND_UP(size_new
, 512);
1689 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1690 fprintf(stderr
, Name
": could not allocate new mpb\n");
1693 memcpy(mpb_new
, mpb
, size_old
);
1696 super
->anchor
= mpb_new
;
1697 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1698 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1700 super
->current_vol
= idx
;
1701 /* when creating the first raid device in this container set num_disks
1702 * to zero, i.e. delete this spare and add raid member devices in
1703 * add_to_super_imsm_volume()
1705 if (super
->current_vol
== 0)
1707 sprintf(st
->subarray
, "%d", idx
);
1708 dev
= malloc(sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
1710 fprintf(stderr
, Name
": could not allocate raid device\n");
1713 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1714 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1715 info
->layout
, info
->chunk_size
,
1717 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1718 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1719 dev
->status
= __cpu_to_le32(0);
1720 dev
->reserved_blocks
= __cpu_to_le32(0);
1722 vol
->migr_state
= 0;
1725 vol
->curr_migr_unit
= 0;
1726 for (i
= 0; i
< idx
; i
++) {
1727 struct imsm_dev
*prev
= get_imsm_dev(super
, i
);
1728 struct imsm_map
*pmap
= get_imsm_map(prev
, 0);
1730 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1731 offset
+= IMSM_RESERVED_SECTORS
;
1733 map
= get_imsm_map(dev
, 0);
1734 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1735 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
1736 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1737 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1738 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1739 IMSM_T_STATE_NORMAL
;
1741 if (info
->level
== 1 && info
->raid_disks
> 2) {
1742 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1743 "in a raid1 volume\n");
1746 if (info
->level
== 10)
1747 map
->raid_level
= 1;
1749 map
->raid_level
= info
->level
;
1751 map
->num_members
= info
->raid_disks
;
1752 for (i
= 0; i
< map
->num_members
; i
++) {
1753 /* initialized in add_to_super */
1754 set_imsm_ord_tbl_ent(map
, i
, 0);
1756 mpb
->num_raid_devs
++;
1757 super
->dev_tbl
[super
->current_vol
] = dev
;
1762 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1763 unsigned long long size
, char *name
,
1764 char *homehost
, int *uuid
)
1766 /* This is primarily called by Create when creating a new array.
1767 * We will then get add_to_super called for each component, and then
1768 * write_init_super called to write it out to each device.
1769 * For IMSM, Create can create on fresh devices or on a pre-existing
1771 * To create on a pre-existing array a different method will be called.
1772 * This one is just for fresh drives.
1774 struct intel_super
*super
;
1775 struct imsm_super
*mpb
;
1783 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1786 super
= alloc_super(1);
1789 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1790 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1795 memset(mpb
, 0, mpb_size
);
1797 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1798 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1799 strlen(MPB_VERSION_RAID5
));
1800 mpb
->mpb_size
= mpb_size
;
1807 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1808 int fd
, char *devname
)
1810 struct intel_super
*super
= st
->sb
;
1811 struct imsm_super
*mpb
= super
->anchor
;
1813 struct imsm_dev
*dev
;
1814 struct imsm_map
*map
;
1817 dev
= get_imsm_dev(super
, super
->current_vol
);
1818 map
= get_imsm_map(dev
, 0);
1820 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1821 if (dl
->major
== dk
->major
&&
1822 dl
->minor
== dk
->minor
)
1825 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1828 /* add a pristine spare to the metadata */
1829 if (dl
->index
< 0) {
1830 dl
->index
= super
->anchor
->num_disks
;
1831 super
->anchor
->num_disks
++;
1833 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
1834 status
= CONFIGURED_DISK
| USABLE_DISK
;
1835 dl
->disk
.status
= __cpu_to_le32(status
);
1837 /* if we are creating the first raid device update the family number */
1838 if (super
->current_vol
== 0) {
1840 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
1841 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
1845 sum
= __gen_imsm_checksum(mpb
);
1846 mpb
->family_num
= __cpu_to_le32(sum
);
1850 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1851 int fd
, char *devname
)
1853 struct intel_super
*super
= st
->sb
;
1855 unsigned long long size
;
1860 if (super
->current_vol
>= 0) {
1861 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1866 dd
= malloc(sizeof(*dd
));
1869 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1872 memset(dd
, 0, sizeof(*dd
));
1873 dd
->major
= major(stb
.st_rdev
);
1874 dd
->minor
= minor(stb
.st_rdev
);
1876 dd
->devname
= devname
? strdup(devname
) : NULL
;
1878 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1881 Name
": failed to retrieve scsi serial, aborting\n");
1886 get_dev_size(fd
, NULL
, &size
);
1888 status
= USABLE_DISK
| SPARE_DISK
;
1889 serialcpy(dd
->disk
.serial
, dd
->serial
);
1890 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
1891 dd
->disk
.status
= __cpu_to_le32(status
);
1892 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1893 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
1895 dd
->disk
.scsi_id
= __cpu_to_le32(0);
1897 if (st
->update_tail
) {
1898 dd
->next
= super
->add
;
1901 dd
->next
= super
->disks
;
1906 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1908 /* spare records have their own family number and do not have any defined raid
1911 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
1913 struct imsm_super mpb_save
;
1914 struct imsm_super
*mpb
= super
->anchor
;
1919 mpb
->num_raid_devs
= 0;
1921 mpb
->mpb_size
= sizeof(struct imsm_super
);
1922 mpb
->generation_num
= __cpu_to_le32(1UL);
1924 for (d
= super
->disks
; d
; d
= d
->next
) {
1928 mpb
->disk
[0] = d
->disk
;
1929 sum
= __gen_imsm_checksum(mpb
);
1930 mpb
->family_num
= __cpu_to_le32(sum
);
1931 sum
= __gen_imsm_checksum(mpb
);
1932 mpb
->check_sum
= __cpu_to_le32(sum
);
1934 if (store_imsm_mpb(d
->fd
, super
)) {
1935 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1936 __func__
, d
->major
, d
->minor
, strerror(errno
));
1950 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1952 struct imsm_super
*mpb
= super
->anchor
;
1958 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
1960 /* 'generation' is incremented everytime the metadata is written */
1961 generation
= __le32_to_cpu(mpb
->generation_num
);
1963 mpb
->generation_num
= __cpu_to_le32(generation
);
1965 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
1966 for (d
= super
->disks
; d
; d
= d
->next
) {
1970 mpb
->disk
[d
->index
] = d
->disk
;
1972 for (d
= super
->missing
; d
; d
= d
->next
)
1973 mpb
->disk
[d
->index
] = d
->disk
;
1975 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1976 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1978 imsm_copy_dev(dev
, super
->dev_tbl
[i
]);
1979 mpb_size
+= sizeof_imsm_dev(dev
, 0);
1981 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
1982 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
1984 /* recalculate checksum */
1985 sum
= __gen_imsm_checksum(mpb
);
1986 mpb
->check_sum
= __cpu_to_le32(sum
);
1988 /* write the mpb for disks that compose raid devices */
1989 for (d
= super
->disks
; d
; d
= d
->next
) {
1992 if (store_imsm_mpb(d
->fd
, super
))
1993 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1994 __func__
, d
->major
, d
->minor
, strerror(errno
));
2002 return write_super_imsm_spares(super
, doclose
);
2008 static int create_array(struct supertype
*st
)
2011 struct imsm_update_create_array
*u
;
2012 struct intel_super
*super
= st
->sb
;
2013 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
2015 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0);
2018 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2023 u
->type
= update_create_array
;
2024 u
->dev_idx
= super
->current_vol
;
2025 imsm_copy_dev(&u
->dev
, dev
);
2026 append_metadata_update(st
, u
, len
);
2031 static int _add_disk(struct supertype
*st
)
2033 struct intel_super
*super
= st
->sb
;
2035 struct imsm_update_add_disk
*u
;
2043 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2048 u
->type
= update_add_disk
;
2049 append_metadata_update(st
, u
, len
);
2054 static int write_init_super_imsm(struct supertype
*st
)
2056 if (st
->update_tail
) {
2057 /* queue the recently created array / added disk
2058 * as a metadata update */
2059 struct intel_super
*super
= st
->sb
;
2063 /* determine if we are creating a volume or adding a disk */
2064 if (super
->current_vol
< 0) {
2065 /* in the add disk case we are running in mdmon
2066 * context, so don't close fd's
2068 return _add_disk(st
);
2070 rv
= create_array(st
);
2072 for (d
= super
->disks
; d
; d
= d
->next
) {
2079 return write_super_imsm(st
->sb
, 1);
2083 static int store_zero_imsm(struct supertype
*st
, int fd
)
2085 unsigned long long dsize
;
2088 get_dev_size(fd
, NULL
, &dsize
);
2090 /* first block is stored on second to last sector of the disk */
2091 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2094 if (posix_memalign(&buf
, 512, 512) != 0)
2097 memset(buf
, 0, 512);
2098 if (write(fd
, buf
, 512) != 512)
2103 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
2105 return __le32_to_cpu(mpb
->bbm_log_size
);
2109 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
2110 int layout
, int raiddisks
, int chunk
,
2111 unsigned long long size
, char *dev
,
2112 unsigned long long *freesize
,
2116 unsigned long long ldsize
;
2118 if (level
!= LEVEL_CONTAINER
)
2123 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2126 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
2127 dev
, strerror(errno
));
2130 if (!get_dev_size(fd
, dev
, &ldsize
)) {
2136 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
2141 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
2142 * FIX ME add ahci details
2144 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
2145 int layout
, int raiddisks
, int chunk
,
2146 unsigned long long size
, char *dev
,
2147 unsigned long long *freesize
,
2151 struct intel_super
*super
= st
->sb
;
2153 unsigned long long pos
= 0;
2154 unsigned long long maxsize
;
2158 if (level
== LEVEL_CONTAINER
)
2161 if (level
== 1 && raiddisks
> 2) {
2163 fprintf(stderr
, Name
": imsm does not support more "
2164 "than 2 in a raid1 configuration\n");
2168 /* We must have the container info already read in. */
2173 /* General test: make sure there is space for
2174 * 'raiddisks' device extents of size 'size' at a given
2177 unsigned long long minsize
= size
*2 /* convert to blocks */;
2178 unsigned long long start_offset
= ~0ULL;
2181 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
2182 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2187 e
= get_extents(super
, dl
);
2190 unsigned long long esize
;
2191 esize
= e
[i
].start
- pos
;
2192 if (esize
>= minsize
)
2194 if (found
&& start_offset
== ~0ULL) {
2197 } else if (found
&& pos
!= start_offset
) {
2201 pos
= e
[i
].start
+ e
[i
].size
;
2203 } while (e
[i
-1].size
);
2208 if (dcnt
< raiddisks
) {
2210 fprintf(stderr
, Name
": imsm: Not enough "
2211 "devices with space for this array "
2218 /* This device must be a member of the set */
2219 if (stat(dev
, &stb
) < 0)
2221 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
2223 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2224 if (dl
->major
== major(stb
.st_rdev
) &&
2225 dl
->minor
== minor(stb
.st_rdev
))
2230 fprintf(stderr
, Name
": %s is not in the "
2231 "same imsm set\n", dev
);
2234 e
= get_extents(super
, dl
);
2238 unsigned long long esize
;
2239 esize
= e
[i
].start
- pos
;
2240 if (esize
>= maxsize
)
2242 pos
= e
[i
].start
+ e
[i
].size
;
2244 } while (e
[i
-1].size
);
2245 *freesize
= maxsize
;
2250 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
2251 int raiddisks
, int chunk
, unsigned long long size
,
2252 char *dev
, unsigned long long *freesize
,
2258 /* if given unused devices create a container
2259 * if given given devices in a container create a member volume
2261 if (level
== LEVEL_CONTAINER
) {
2262 /* Must be a fresh device to add to a container */
2263 return validate_geometry_imsm_container(st
, level
, layout
,
2264 raiddisks
, chunk
, size
,
2270 /* creating in a given container */
2271 return validate_geometry_imsm_volume(st
, level
, layout
,
2272 raiddisks
, chunk
, size
,
2273 dev
, freesize
, verbose
);
2276 /* limit creation to the following levels */
2288 /* This device needs to be a device in an 'imsm' container */
2289 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2293 Name
": Cannot create this array on device %s\n",
2298 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
2300 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2301 dev
, strerror(errno
));
2304 /* Well, it is in use by someone, maybe an 'imsm' container. */
2305 cfd
= open_container(fd
);
2309 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
2313 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
2315 if (sra
&& sra
->array
.major_version
== -1 &&
2316 strcmp(sra
->text_version
, "imsm") == 0) {
2317 /* This is a member of a imsm container. Load the container
2318 * and try to create a volume
2320 struct intel_super
*super
;
2322 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
2324 st
->container_dev
= fd2devnum(cfd
);
2326 return validate_geometry_imsm_volume(st
, level
, layout
,
2332 } else /* may belong to another container */
2337 #endif /* MDASSEMBLE */
2339 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
2341 /* Given a container loaded by load_super_imsm_all,
2342 * extract information about all the arrays into
2345 * For each imsm_dev create an mdinfo, fill it in,
2346 * then look for matching devices in super->disks
2347 * and create appropriate device mdinfo.
2349 struct intel_super
*super
= st
->sb
;
2350 struct imsm_super
*mpb
= super
->anchor
;
2351 struct mdinfo
*rest
= NULL
;
2354 /* do not assemble arrays that might have bad blocks */
2355 if (imsm_bbm_log_size(super
->anchor
)) {
2356 fprintf(stderr
, Name
": BBM log found in metadata. "
2357 "Cannot activate array(s).\n");
2361 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2362 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2363 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2364 struct mdinfo
*this;
2367 this = malloc(sizeof(*this));
2368 memset(this, 0, sizeof(*this));
2371 super
->current_vol
= i
;
2372 getinfo_super_imsm_volume(st
, this);
2373 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
2374 struct mdinfo
*info_d
;
2382 idx
= get_imsm_disk_idx(dev
, slot
);
2383 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
2384 for (d
= super
->disks
; d
; d
= d
->next
)
2385 if (d
->index
== idx
)
2391 s
= d
? __le32_to_cpu(d
->disk
.status
) : 0;
2392 if (s
& FAILED_DISK
)
2394 if (!(s
& USABLE_DISK
))
2396 if (ord
& IMSM_ORD_REBUILD
)
2400 * if we skip some disks the array will be assmebled degraded;
2401 * reset resync start to avoid a dirty-degraded situation
2403 * FIXME handle dirty degraded
2405 if (skip
&& !dev
->vol
.dirty
)
2406 this->resync_start
= ~0ULL;
2410 info_d
= malloc(sizeof(*info_d
));
2412 fprintf(stderr
, Name
": failed to allocate disk"
2413 " for volume %s\n", (char *) dev
->volume
);
2418 memset(info_d
, 0, sizeof(*info_d
));
2419 info_d
->next
= this->devs
;
2420 this->devs
= info_d
;
2422 info_d
->disk
.number
= d
->index
;
2423 info_d
->disk
.major
= d
->major
;
2424 info_d
->disk
.minor
= d
->minor
;
2425 info_d
->disk
.raid_disk
= slot
;
2427 this->array
.working_disks
++;
2429 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
2430 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2431 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
2433 strcpy(info_d
->name
, d
->devname
);
2443 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
2446 struct intel_super
*super
= c
->sb
;
2447 struct imsm_super
*mpb
= super
->anchor
;
2449 if (atoi(inst
) >= mpb
->num_raid_devs
) {
2450 fprintf(stderr
, "%s: subarry index %d, out of range\n",
2451 __func__
, atoi(inst
));
2455 dprintf("imsm: open_new %s\n", inst
);
2456 a
->info
.container_member
= atoi(inst
);
2460 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
2462 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2465 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
2466 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
2468 switch (get_imsm_raid_level(map
)) {
2470 return IMSM_T_STATE_FAILED
;
2473 if (failed
< map
->num_members
)
2474 return IMSM_T_STATE_DEGRADED
;
2476 return IMSM_T_STATE_FAILED
;
2481 * check to see if any mirrors have failed, otherwise we
2482 * are degraded. Even numbered slots are mirrored on
2488 for (i
= 0; i
< map
->num_members
; i
++) {
2489 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2490 int idx
= ord_to_idx(ord
);
2491 struct imsm_disk
*disk
;
2493 /* reset the potential in-sync count on even-numbered
2494 * slots. num_copies is always 2 for imsm raid10
2499 disk
= get_imsm_disk(super
, idx
);
2501 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2502 ord
& IMSM_ORD_REBUILD
)
2505 /* no in-sync disks left in this mirror the
2509 return IMSM_T_STATE_FAILED
;
2512 return IMSM_T_STATE_DEGRADED
;
2516 return IMSM_T_STATE_DEGRADED
;
2518 return IMSM_T_STATE_FAILED
;
2524 return map
->map_state
;
2527 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
2531 struct imsm_disk
*disk
;
2532 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2534 for (i
= 0; i
< map
->num_members
; i
++) {
2535 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2536 int idx
= ord_to_idx(ord
);
2538 disk
= get_imsm_disk(super
, idx
);
2540 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2541 ord
& IMSM_ORD_REBUILD
)
2548 static int is_resyncing(struct imsm_dev
*dev
)
2550 struct imsm_map
*migr_map
;
2552 if (!dev
->vol
.migr_state
)
2555 if (dev
->vol
.migr_type
== 0)
2558 migr_map
= get_imsm_map(dev
, 1);
2560 if (migr_map
->map_state
== IMSM_T_STATE_NORMAL
)
2566 static int is_rebuilding(struct imsm_dev
*dev
)
2568 struct imsm_map
*migr_map
;
2570 if (!dev
->vol
.migr_state
)
2573 if (dev
->vol
.migr_type
== 0)
2576 migr_map
= get_imsm_map(dev
, 1);
2578 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
2584 static void mark_failure(struct imsm_disk
*disk
)
2586 __u32 status
= __le32_to_cpu(disk
->status
);
2588 if (status
& FAILED_DISK
)
2590 status
|= FAILED_DISK
;
2591 disk
->status
= __cpu_to_le32(status
);
2592 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
2593 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
2596 /* Handle dirty -> clean transititions and resync. Degraded and rebuild
2597 * states are handled in imsm_set_disk() with one exception, when a
2598 * resync is stopped due to a new failure this routine will set the
2599 * 'degraded' state for the array.
2601 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
2603 int inst
= a
->info
.container_member
;
2604 struct intel_super
*super
= a
->container
->sb
;
2605 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2606 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2607 int failed
= imsm_count_failed(super
, dev
);
2608 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
2610 /* before we activate this array handle any missing disks */
2611 if (consistent
== 2 && super
->missing
) {
2614 dprintf("imsm: mark missing\n");
2615 end_migration(dev
, map_state
);
2616 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
2617 mark_failure(&dl
->disk
);
2618 super
->updates_pending
++;
2621 if (consistent
== 2 &&
2622 (!is_resync_complete(a
) ||
2623 map_state
!= IMSM_T_STATE_NORMAL
||
2624 dev
->vol
.migr_state
))
2627 if (is_resync_complete(a
)) {
2628 /* complete intialization / resync,
2629 * recovery is completed in ->set_disk
2631 if (is_resyncing(dev
)) {
2632 dprintf("imsm: mark resync done\n");
2633 end_migration(dev
, map_state
);
2634 super
->updates_pending
++;
2636 } else if (!is_resyncing(dev
) && !failed
) {
2637 /* mark the start of the init process if nothing is failed */
2638 dprintf("imsm: mark resync start (%llu)\n", a
->resync_start
);
2639 map
->map_state
= map_state
;
2640 migrate(dev
, IMSM_T_STATE_NORMAL
,
2641 map
->map_state
== IMSM_T_STATE_NORMAL
);
2642 super
->updates_pending
++;
2645 /* check if we can update the migration checkpoint */
2646 if (dev
->vol
.migr_state
&&
2647 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != a
->resync_start
) {
2648 dprintf("imsm: checkpoint migration (%llu)\n", a
->resync_start
);
2649 dev
->vol
.curr_migr_unit
= __cpu_to_le32(a
->resync_start
);
2650 super
->updates_pending
++;
2653 /* mark dirty / clean */
2654 if (dev
->vol
.dirty
!= !consistent
) {
2655 dprintf("imsm: mark '%s' (%llu)\n",
2656 consistent
? "clean" : "dirty", a
->resync_start
);
2661 super
->updates_pending
++;
2666 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
2668 int inst
= a
->info
.container_member
;
2669 struct intel_super
*super
= a
->container
->sb
;
2670 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2671 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2672 struct imsm_disk
*disk
;
2678 if (n
> map
->num_members
)
2679 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
2680 n
, map
->num_members
- 1);
2685 dprintf("imsm: set_disk %d:%x\n", n
, state
);
2687 ord
= get_imsm_ord_tbl_ent(dev
, n
);
2688 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
2690 /* check for new failures */
2691 status
= __le32_to_cpu(disk
->status
);
2692 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
2694 super
->updates_pending
++;
2697 /* check if in_sync */
2698 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
) {
2699 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
2701 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
2702 super
->updates_pending
++;
2705 failed
= imsm_count_failed(super
, dev
);
2706 map_state
= imsm_check_degraded(super
, dev
, failed
);
2708 /* check if recovery complete, newly degraded, or failed */
2709 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
2710 end_migration(dev
, map_state
);
2711 super
->updates_pending
++;
2712 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
2713 map
->map_state
!= map_state
&&
2714 !dev
->vol
.migr_state
) {
2715 dprintf("imsm: mark degraded\n");
2716 map
->map_state
= map_state
;
2717 super
->updates_pending
++;
2718 } else if (map_state
== IMSM_T_STATE_FAILED
&&
2719 map
->map_state
!= map_state
) {
2720 dprintf("imsm: mark failed\n");
2721 end_migration(dev
, map_state
);
2722 super
->updates_pending
++;
2726 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
2728 struct imsm_super
*mpb
= super
->anchor
;
2729 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
2730 unsigned long long dsize
;
2731 unsigned long long sectors
;
2733 get_dev_size(fd
, NULL
, &dsize
);
2735 if (mpb_size
> 512) {
2736 /* -1 to account for anchor */
2737 sectors
= mpb_sectors(mpb
) - 1;
2739 /* write the extended mpb to the sectors preceeding the anchor */
2740 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
2743 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
2747 /* first block is stored on second to last sector of the disk */
2748 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2751 if (write(fd
, super
->buf
, 512) != 512)
2757 static void imsm_sync_metadata(struct supertype
*container
)
2759 struct intel_super
*super
= container
->sb
;
2761 if (!super
->updates_pending
)
2764 write_super_imsm(super
, 0);
2766 super
->updates_pending
= 0;
2769 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
2771 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2772 int i
= get_imsm_disk_idx(dev
, idx
);
2775 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2779 if (dl
&& __le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
2783 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
2788 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
, struct active_array
*a
)
2790 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2791 int idx
= get_imsm_disk_idx(dev
, slot
);
2792 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2793 unsigned long long esize
;
2794 unsigned long long pos
;
2803 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2804 /* If in this array, skip */
2805 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2806 if (d
->state_fd
>= 0 &&
2807 d
->disk
.major
== dl
->major
&&
2808 d
->disk
.minor
== dl
->minor
) {
2809 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
2815 /* skip in use or failed drives */
2816 status
= __le32_to_cpu(dl
->disk
.status
);
2817 if (status
& FAILED_DISK
|| idx
== dl
->index
) {
2818 dprintf("%x:%x status ( %s%s)\n",
2819 dl
->major
, dl
->minor
,
2820 status
& FAILED_DISK
? "failed " : "",
2821 idx
== dl
->index
? "in use " : "");
2825 /* Does this unused device have the requisite free space?
2826 * We need a->info.component_size sectors
2828 ex
= get_extents(super
, dl
);
2830 dprintf("cannot get extents\n");
2836 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
2839 /* check that we can start at pba_of_lba0 with
2840 * a->info.component_size of space
2842 esize
= ex
[j
].start
- pos
;
2843 if (array_start
>= pos
&&
2844 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
2848 pos
= ex
[j
].start
+ ex
[j
].size
;
2851 } while (ex
[j
-1].size
);
2855 dprintf("%x:%x does not have %llu at %d\n",
2856 dl
->major
, dl
->minor
,
2857 a
->info
.component_size
,
2858 __le32_to_cpu(map
->pba_of_lba0
));
2868 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
2869 struct metadata_update
**updates
)
2872 * Find a device with unused free space and use it to replace a
2873 * failed/vacant region in an array. We replace failed regions one a
2874 * array at a time. The result is that a new spare disk will be added
2875 * to the first failed array and after the monitor has finished
2876 * propagating failures the remainder will be consumed.
2878 * FIXME add a capability for mdmon to request spares from another
2882 struct intel_super
*super
= a
->container
->sb
;
2883 int inst
= a
->info
.container_member
;
2884 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2885 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2886 int failed
= a
->info
.array
.raid_disks
;
2887 struct mdinfo
*rv
= NULL
;
2890 struct metadata_update
*mu
;
2892 struct imsm_update_activate_spare
*u
;
2896 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
2897 if ((d
->curr_state
& DS_FAULTY
) &&
2899 /* wait for Removal to happen */
2901 if (d
->state_fd
>= 0)
2905 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
2906 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
2907 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
2910 /* For each slot, if it is not working, find a spare */
2911 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
2912 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2913 if (d
->disk
.raid_disk
== i
)
2915 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
2916 if (d
&& (d
->state_fd
>= 0))
2920 * OK, this device needs recovery. Try to re-add the previous
2921 * occupant of this slot, if this fails add a new spare
2923 dl
= imsm_readd(super
, i
, a
);
2925 dl
= imsm_add_spare(super
, i
, a
);
2929 /* found a usable disk with enough space */
2930 di
= malloc(sizeof(*di
));
2933 memset(di
, 0, sizeof(*di
));
2935 /* dl->index will be -1 in the case we are activating a
2936 * pristine spare. imsm_process_update() will create a
2937 * new index in this case. Once a disk is found to be
2938 * failed in all member arrays it is kicked from the
2941 di
->disk
.number
= dl
->index
;
2943 /* (ab)use di->devs to store a pointer to the device
2946 di
->devs
= (struct mdinfo
*) dl
;
2948 di
->disk
.raid_disk
= i
;
2949 di
->disk
.major
= dl
->major
;
2950 di
->disk
.minor
= dl
->minor
;
2952 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2953 di
->component_size
= a
->info
.component_size
;
2954 di
->container_member
= inst
;
2958 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2959 i
, di
->data_offset
);
2965 /* No spares found */
2967 /* Now 'rv' has a list of devices to return.
2968 * Create a metadata_update record to update the
2969 * disk_ord_tbl for the array
2971 mu
= malloc(sizeof(*mu
));
2973 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2974 if (mu
->buf
== NULL
) {
2981 struct mdinfo
*n
= rv
->next
;
2990 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
2991 mu
->next
= *updates
;
2992 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
2994 for (di
= rv
; di
; di
= di
->next
) {
2995 u
->type
= update_activate_spare
;
2996 u
->dl
= (struct dl
*) di
->devs
;
2998 u
->slot
= di
->disk
.raid_disk
;
3009 static int disks_overlap(struct imsm_dev
*d1
, struct imsm_dev
*d2
)
3011 struct imsm_map
*m1
= get_imsm_map(d1
, 0);
3012 struct imsm_map
*m2
= get_imsm_map(d2
, 0);
3017 for (i
= 0; i
< m1
->num_members
; i
++) {
3018 idx
= get_imsm_disk_idx(d1
, i
);
3019 for (j
= 0; j
< m2
->num_members
; j
++)
3020 if (idx
== get_imsm_disk_idx(d2
, j
))
3027 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
);
3029 static void imsm_process_update(struct supertype
*st
,
3030 struct metadata_update
*update
)
3033 * crack open the metadata_update envelope to find the update record
3034 * update can be one of:
3035 * update_activate_spare - a spare device has replaced a failed
3036 * device in an array, update the disk_ord_tbl. If this disk is
3037 * present in all member arrays then also clear the SPARE_DISK
3040 struct intel_super
*super
= st
->sb
;
3041 struct imsm_super
*mpb
;
3042 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3044 /* update requires a larger buf but the allocation failed */
3045 if (super
->next_len
&& !super
->next_buf
) {
3046 super
->next_len
= 0;
3050 if (super
->next_buf
) {
3051 memcpy(super
->next_buf
, super
->buf
, super
->len
);
3053 super
->len
= super
->next_len
;
3054 super
->buf
= super
->next_buf
;
3056 super
->next_len
= 0;
3057 super
->next_buf
= NULL
;
3060 mpb
= super
->anchor
;
3063 case update_activate_spare
: {
3064 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
3065 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
3066 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3067 struct imsm_map
*migr_map
;
3068 struct active_array
*a
;
3069 struct imsm_disk
*disk
;
3075 int victim
= get_imsm_disk_idx(dev
, u
->slot
);
3078 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
3083 fprintf(stderr
, "error: imsm_activate_spare passed "
3084 "an unknown disk (index: %d)\n",
3089 super
->updates_pending
++;
3091 /* count failures (excluding rebuilds and the victim)
3092 * to determine map[0] state
3095 for (i
= 0; i
< map
->num_members
; i
++) {
3098 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3100 __le32_to_cpu(disk
->status
) & FAILED_DISK
)
3104 /* adding a pristine spare, assign a new index */
3105 if (dl
->index
< 0) {
3106 dl
->index
= super
->anchor
->num_disks
;
3107 super
->anchor
->num_disks
++;
3110 status
= __le32_to_cpu(disk
->status
);
3111 status
|= CONFIGURED_DISK
;
3112 status
&= ~SPARE_DISK
;
3113 disk
->status
= __cpu_to_le32(status
);
3116 to_state
= imsm_check_degraded(super
, dev
, failed
);
3117 map
->map_state
= IMSM_T_STATE_DEGRADED
;
3118 migrate(dev
, to_state
, 1);
3119 migr_map
= get_imsm_map(dev
, 1);
3120 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
3121 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
3123 /* count arrays using the victim in the metadata */
3125 for (a
= st
->arrays
; a
; a
= a
->next
) {
3126 dev
= get_imsm_dev(super
, a
->info
.container_member
);
3127 for (i
= 0; i
< map
->num_members
; i
++)
3128 if (victim
== get_imsm_disk_idx(dev
, i
))
3132 /* delete the victim if it is no longer being
3138 /* We know that 'manager' isn't touching anything,
3139 * so it is safe to delete
3141 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
3142 if ((*dlp
)->index
== victim
)
3145 /* victim may be on the missing list */
3147 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
3148 if ((*dlp
)->index
== victim
)
3150 imsm_delete(super
, dlp
, victim
);
3154 case update_create_array
: {
3155 /* someone wants to create a new array, we need to be aware of
3156 * a few races/collisions:
3157 * 1/ 'Create' called by two separate instances of mdadm
3158 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
3159 * devices that have since been assimilated via
3161 * In the event this update can not be carried out mdadm will
3162 * (FIX ME) notice that its update did not take hold.
3164 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3165 struct imsm_dev
*dev
;
3166 struct imsm_map
*map
, *new_map
;
3167 unsigned long long start
, end
;
3168 unsigned long long new_start
, new_end
;
3172 /* handle racing creates: first come first serve */
3173 if (u
->dev_idx
< mpb
->num_raid_devs
) {
3174 dprintf("%s: subarray %d already defined\n",
3175 __func__
, u
->dev_idx
);
3179 /* check update is next in sequence */
3180 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
3181 dprintf("%s: can not create array %d expected index %d\n",
3182 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
3186 new_map
= get_imsm_map(&u
->dev
, 0);
3187 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
3188 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
3190 /* handle activate_spare versus create race:
3191 * check to make sure that overlapping arrays do not include
3194 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3195 dev
= get_imsm_dev(super
, i
);
3196 map
= get_imsm_map(dev
, 0);
3197 start
= __le32_to_cpu(map
->pba_of_lba0
);
3198 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
3199 if ((new_start
>= start
&& new_start
<= end
) ||
3200 (start
>= new_start
&& start
<= new_end
))
3202 if (overlap
&& disks_overlap(dev
, &u
->dev
)) {
3203 dprintf("%s: arrays overlap\n", __func__
);
3207 /* check num_members sanity */
3208 if (new_map
->num_members
> mpb
->num_disks
) {
3209 dprintf("%s: num_disks out of range\n", __func__
);
3213 /* check that prepare update was successful */
3214 if (!update
->space
) {
3215 dprintf("%s: prepare update failed\n", __func__
);
3219 super
->updates_pending
++;
3220 dev
= update
->space
;
3221 map
= get_imsm_map(dev
, 0);
3222 update
->space
= NULL
;
3223 imsm_copy_dev(dev
, &u
->dev
);
3224 map
= get_imsm_map(dev
, 0);
3225 super
->dev_tbl
[u
->dev_idx
] = dev
;
3226 mpb
->num_raid_devs
++;
3229 for (i
= 0; i
< map
->num_members
; i
++) {
3230 struct imsm_disk
*disk
;
3233 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3234 status
= __le32_to_cpu(disk
->status
);
3235 status
|= CONFIGURED_DISK
;
3236 status
&= ~SPARE_DISK
;
3237 disk
->status
= __cpu_to_le32(status
);
3241 case update_add_disk
:
3243 /* we may be able to repair some arrays if disks are
3246 struct active_array
*a
;
3248 super
->updates_pending
++;
3249 for (a
= st
->arrays
; a
; a
= a
->next
)
3250 a
->check_degraded
= 1;
3252 /* add some spares to the metadata */
3253 while (super
->add
) {
3257 super
->add
= al
->next
;
3258 al
->next
= super
->disks
;
3260 dprintf("%s: added %x:%x\n",
3261 __func__
, al
->major
, al
->minor
);
3268 static void imsm_prepare_update(struct supertype
*st
,
3269 struct metadata_update
*update
)
3272 * Allocate space to hold new disk entries, raid-device entries or a new
3273 * mpb if necessary. The manager synchronously waits for updates to
3274 * complete in the monitor, so new mpb buffers allocated here can be
3275 * integrated by the monitor thread without worrying about live pointers
3276 * in the manager thread.
3278 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3279 struct intel_super
*super
= st
->sb
;
3280 struct imsm_super
*mpb
= super
->anchor
;
3285 case update_create_array
: {
3286 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3288 len
= sizeof_imsm_dev(&u
->dev
, 1);
3289 update
->space
= malloc(len
);
3296 /* check if we need a larger metadata buffer */
3297 if (super
->next_buf
)
3298 buf_len
= super
->next_len
;
3300 buf_len
= super
->len
;
3302 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
3303 /* ok we need a larger buf than what is currently allocated
3304 * if this allocation fails process_update will notice that
3305 * ->next_len is set and ->next_buf is NULL
3307 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
3308 if (super
->next_buf
)
3309 free(super
->next_buf
);
3311 super
->next_len
= buf_len
;
3312 if (posix_memalign(&super
->next_buf
, buf_len
, 512) != 0)
3313 super
->next_buf
= NULL
;
3317 /* must be called while manager is quiesced */
3318 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
)
3320 struct imsm_super
*mpb
= super
->anchor
;
3322 struct imsm_dev
*dev
;
3323 struct imsm_map
*map
;
3324 int i
, j
, num_members
;
3327 dprintf("%s: deleting device[%d] from imsm_super\n",
3330 /* shift all indexes down one */
3331 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
3332 if (iter
->index
> index
)
3334 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
3335 if (iter
->index
> index
)
3338 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3339 dev
= get_imsm_dev(super
, i
);
3340 map
= get_imsm_map(dev
, 0);
3341 num_members
= map
->num_members
;
3342 for (j
= 0; j
< num_members
; j
++) {
3343 /* update ord entries being careful not to propagate
3344 * ord-flags to the first map
3346 ord
= get_imsm_ord_tbl_ent(dev
, j
);
3348 if (ord_to_idx(ord
) <= index
)
3351 map
= get_imsm_map(dev
, 0);
3352 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
3353 map
= get_imsm_map(dev
, 1);
3355 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
3360 super
->updates_pending
++;
3362 struct dl
*dl
= *dlp
;
3364 *dlp
= (*dlp
)->next
;
3365 __free_imsm_disk(dl
);
3368 #endif /* MDASSEMBLE */
3370 struct superswitch super_imsm
= {
3372 .examine_super
= examine_super_imsm
,
3373 .brief_examine_super
= brief_examine_super_imsm
,
3374 .detail_super
= detail_super_imsm
,
3375 .brief_detail_super
= brief_detail_super_imsm
,
3376 .write_init_super
= write_init_super_imsm
,
3377 .validate_geometry
= validate_geometry_imsm
,
3378 .add_to_super
= add_to_super_imsm
,
3380 .match_home
= match_home_imsm
,
3381 .uuid_from_super
= uuid_from_super_imsm
,
3382 .getinfo_super
= getinfo_super_imsm
,
3383 .update_super
= update_super_imsm
,
3385 .avail_size
= avail_size_imsm
,
3387 .compare_super
= compare_super_imsm
,
3389 .load_super
= load_super_imsm
,
3390 .init_super
= init_super_imsm
,
3391 .store_super
= store_zero_imsm
,
3392 .free_super
= free_super_imsm
,
3393 .match_metadata_desc
= match_metadata_desc_imsm
,
3394 .container_content
= container_content_imsm
,
3400 .open_new
= imsm_open_new
,
3401 .load_super
= load_super_imsm
,
3402 .set_array_state
= imsm_set_array_state
,
3403 .set_disk
= imsm_set_disk
,
3404 .sync_metadata
= imsm_sync_metadata
,
3405 .activate_spare
= imsm_activate_spare
,
3406 .process_update
= imsm_process_update
,
3407 .prepare_update
= imsm_prepare_update
,
3408 #endif /* MDASSEMBLE */