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 struct supertype
*match_metadata_desc_imsm(char *arg
)
218 struct supertype
*st
;
220 if (strcmp(arg
, "imsm") != 0 &&
221 strcmp(arg
, "default") != 0
225 st
= malloc(sizeof(*st
));
226 memset(st
, 0, sizeof(*st
));
227 st
->ss
= &super_imsm
;
228 st
->max_devs
= IMSM_MAX_DEVICES
;
229 st
->minor_version
= 0;
235 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
237 return &mpb
->sig
[MPB_SIG_LEN
];
241 /* retrieve a disk directly from the anchor when the anchor is known to be
242 * up-to-date, currently only at load time
244 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
246 if (index
>= mpb
->num_disks
)
248 return &mpb
->disk
[index
];
252 /* retrieve a disk from the parsed metadata */
253 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
257 for (d
= super
->disks
; d
; d
= d
->next
)
258 if (d
->index
== index
)
265 /* generate a checksum directly from the anchor when the anchor is known to be
266 * up-to-date, currently only at load or write_super after coalescing
268 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
270 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
271 __u32
*p
= (__u32
*) mpb
;
275 sum
+= __le32_to_cpu(*p
);
279 return sum
- __le32_to_cpu(mpb
->check_sum
);
282 static size_t sizeof_imsm_map(struct imsm_map
*map
)
284 return sizeof(struct imsm_map
) + sizeof(__u32
) * (map
->num_members
- 1);
287 struct imsm_map
*get_imsm_map(struct imsm_dev
*dev
, int second_map
)
289 struct imsm_map
*map
= &dev
->vol
.map
[0];
291 if (second_map
&& !dev
->vol
.migr_state
)
293 else if (second_map
) {
296 return ptr
+ sizeof_imsm_map(map
);
302 /* return the size of the device.
303 * migr_state increases the returned size if map[0] were to be duplicated
305 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
, int migr_state
)
307 size_t size
= sizeof(*dev
) - sizeof(struct imsm_map
) +
308 sizeof_imsm_map(get_imsm_map(dev
, 0));
310 /* migrating means an additional map */
311 if (dev
->vol
.migr_state
)
312 size
+= sizeof_imsm_map(get_imsm_map(dev
, 1));
314 size
+= sizeof_imsm_map(get_imsm_map(dev
, 0));
319 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
325 if (index
>= mpb
->num_raid_devs
)
328 /* devices start after all disks */
329 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
331 for (i
= 0; i
<= index
; i
++)
333 return _mpb
+ offset
;
335 offset
+= sizeof_imsm_dev(_mpb
+ offset
, 0);
340 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
342 if (index
>= super
->anchor
->num_raid_devs
)
344 return super
->dev_tbl
[index
];
347 static __u32
get_imsm_ord_tbl_ent(struct imsm_dev
*dev
, int slot
)
349 struct imsm_map
*map
;
351 if (dev
->vol
.migr_state
)
352 map
= get_imsm_map(dev
, 1);
354 map
= get_imsm_map(dev
, 0);
356 /* top byte identifies disk under rebuild */
357 return __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
360 #define ord_to_idx(ord) (((ord) << 8) >> 8)
361 static __u32
get_imsm_disk_idx(struct imsm_dev
*dev
, int slot
)
363 __u32 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
365 return ord_to_idx(ord
);
368 static void set_imsm_ord_tbl_ent(struct imsm_map
*map
, int slot
, __u32 ord
)
370 map
->disk_ord_tbl
[slot
] = __cpu_to_le32(ord
);
373 static int get_imsm_raid_level(struct imsm_map
*map
)
375 if (map
->raid_level
== 1) {
376 if (map
->num_members
== 2)
382 return map
->raid_level
;
385 static int cmp_extent(const void *av
, const void *bv
)
387 const struct extent
*a
= av
;
388 const struct extent
*b
= bv
;
389 if (a
->start
< b
->start
)
391 if (a
->start
> b
->start
)
396 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
398 /* find a list of used extents on the given physical device */
399 struct extent
*rv
, *e
;
402 __u32 reservation
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
404 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
405 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
406 struct imsm_map
*map
= get_imsm_map(dev
, 0);
408 for (j
= 0; j
< map
->num_members
; j
++) {
409 __u32 index
= get_imsm_disk_idx(dev
, j
);
411 if (index
== dl
->index
)
415 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
420 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
421 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
422 struct imsm_map
*map
= get_imsm_map(dev
, 0);
424 for (j
= 0; j
< map
->num_members
; j
++) {
425 __u32 index
= get_imsm_disk_idx(dev
, j
);
427 if (index
== dl
->index
) {
428 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
429 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
434 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
436 /* determine the start of the metadata
437 * when no raid devices are defined use the default
438 * ...otherwise allow the metadata to truncate the value
439 * as is the case with older versions of imsm
442 struct extent
*last
= &rv
[memberships
- 1];
445 remainder
= __le32_to_cpu(dl
->disk
.total_blocks
) -
446 (last
->start
+ last
->size
);
447 if (reservation
> remainder
)
448 reservation
= remainder
;
450 e
->start
= __le32_to_cpu(dl
->disk
.total_blocks
) - reservation
;
455 /* try to determine how much space is reserved for metadata from
456 * the last get_extents() entry, otherwise fallback to the
459 static __u32
imsm_reserved_sectors(struct intel_super
*super
, struct dl
*dl
)
465 /* for spares just return a minimal reservation which will grow
466 * once the spare is picked up by an array
469 return MPB_SECTOR_CNT
;
471 e
= get_extents(super
, dl
);
473 return MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
475 /* scroll to last entry */
476 for (i
= 0; e
[i
].size
; i
++)
479 rv
= __le32_to_cpu(dl
->disk
.total_blocks
) - e
[i
].start
;
487 static void print_imsm_dev(struct imsm_dev
*dev
, char *uuid
, int disk_idx
)
491 struct imsm_map
*map
= get_imsm_map(dev
, 0);
495 printf("[%.16s]:\n", dev
->volume
);
496 printf(" UUID : %s\n", uuid
);
497 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
498 printf(" Members : %d\n", map
->num_members
);
499 for (slot
= 0; slot
< map
->num_members
; slot
++)
500 if (disk_idx
== get_imsm_disk_idx(dev
, slot
))
502 if (slot
< map
->num_members
) {
503 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
504 printf(" This Slot : %d%s\n", slot
,
505 ord
& IMSM_ORD_REBUILD
? " (out-of-sync)" : "");
507 printf(" This Slot : ?\n");
508 sz
= __le32_to_cpu(dev
->size_high
);
510 sz
+= __le32_to_cpu(dev
->size_low
);
511 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
512 human_size(sz
* 512));
513 sz
= __le32_to_cpu(map
->blocks_per_member
);
514 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
515 human_size(sz
* 512));
516 printf(" Sector Offset : %u\n",
517 __le32_to_cpu(map
->pba_of_lba0
));
518 printf(" Num Stripes : %u\n",
519 __le32_to_cpu(map
->num_data_stripes
));
520 printf(" Chunk Size : %u KiB\n",
521 __le16_to_cpu(map
->blocks_per_strip
) / 2);
522 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
523 printf(" Migrate State : %s", dev
->vol
.migr_state
? "migrating" : "idle");
524 if (dev
->vol
.migr_state
)
525 printf(": %s", dev
->vol
.migr_type
? "rebuilding" : "initializing");
527 printf(" Map State : %s", map_state_str
[map
->map_state
]);
528 if (dev
->vol
.migr_state
) {
529 struct imsm_map
*map
= get_imsm_map(dev
, 1);
530 printf(" <-- %s", map_state_str
[map
->map_state
]);
533 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
536 static void print_imsm_disk(struct imsm_super
*mpb
, int index
, __u32 reserved
)
538 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
539 char str
[MAX_RAID_SERIAL_LEN
+ 1];
547 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
548 printf(" Disk%02d Serial : %s\n", index
, str
);
549 s
= __le32_to_cpu(disk
->status
);
550 printf(" State :%s%s%s%s\n", s
&SPARE_DISK
? " spare" : "",
551 s
&CONFIGURED_DISK
? " active" : "",
552 s
&FAILED_DISK
? " failed" : "",
553 s
&USABLE_DISK
? " usable" : "");
554 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
555 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
556 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
557 human_size(sz
* 512));
560 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
);
562 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
564 struct intel_super
*super
= st
->sb
;
565 struct imsm_super
*mpb
= super
->anchor
;
566 char str
[MAX_SIGNATURE_LENGTH
];
571 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
574 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
575 printf(" Magic : %s\n", str
);
576 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
577 printf(" Version : %s\n", get_imsm_version(mpb
));
578 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
579 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
580 getinfo_super_imsm(st
, &info
);
581 fname_from_uuid(st
, &info
, nbuf
,'-');
582 printf(" UUID : %s\n", nbuf
+ 5);
583 sum
= __le32_to_cpu(mpb
->check_sum
);
584 printf(" Checksum : %08x %s\n", sum
,
585 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
586 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
587 printf(" Disks : %d\n", mpb
->num_disks
);
588 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
589 print_imsm_disk(mpb
, super
->disks
->index
, reserved
);
590 if (super
->bbm_log
) {
591 struct bbm_log
*log
= super
->bbm_log
;
594 printf("Bad Block Management Log:\n");
595 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
596 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
597 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
598 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
599 printf(" First Spare : %llx\n", __le64_to_cpu(log
->first_spare_lba
));
601 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
603 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
605 super
->current_vol
= i
;
606 getinfo_super_imsm(st
, &info
);
607 fname_from_uuid(st
, &info
, nbuf
, '-');
608 print_imsm_dev(dev
, nbuf
+ 5, super
->disks
->index
);
610 for (i
= 0; i
< mpb
->num_disks
; i
++) {
611 if (i
== super
->disks
->index
)
613 print_imsm_disk(mpb
, i
, reserved
);
617 static void brief_examine_super_imsm(struct supertype
*st
)
619 /* We just write a generic IMSM ARRAY entry */
623 struct intel_super
*super
= st
->sb
;
626 if (!super
->anchor
->num_raid_devs
)
629 getinfo_super_imsm(st
, &info
);
630 fname_from_uuid(st
, &info
, nbuf
,'-');
631 printf("ARRAY metadata=imsm auto=md UUID=%s\n", nbuf
+ 5);
632 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
633 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
635 super
->current_vol
= i
;
636 getinfo_super_imsm(st
, &info
);
637 fname_from_uuid(st
, &info
, nbuf1
,'-');
638 printf("ARRAY /dev/md/%.16s container=%s\n"
639 " member=%d auto=mdp UUID=%s\n",
640 dev
->volume
, nbuf
+ 5, i
, nbuf1
+ 5);
644 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
649 getinfo_super_imsm(st
, &info
);
650 fname_from_uuid(st
, &info
, nbuf
,'-');
651 printf("\n UUID : %s\n", nbuf
+ 5);
654 static void brief_detail_super_imsm(struct supertype
*st
)
658 getinfo_super_imsm(st
, &info
);
659 fname_from_uuid(st
, &info
, nbuf
,'-');
660 printf(" UUID=%s", nbuf
+ 5);
664 static int match_home_imsm(struct supertype
*st
, char *homehost
)
666 printf("%s\n", __FUNCTION__
);
671 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
673 /* The uuid returned here is used for:
674 * uuid to put into bitmap file (Create, Grow)
675 * uuid for backup header when saving critical section (Grow)
676 * comparing uuids when re-adding a device into an array
677 * In these cases the uuid required is that of the data-array,
678 * not the device-set.
679 * uuid to recognise same set when adding a missing device back
680 * to an array. This is a uuid for the device-set.
682 * For each of these we can make do with a truncated
683 * or hashed uuid rather than the original, as long as
685 * In each case the uuid required is that of the data-array,
686 * not the device-set.
688 /* imsm does not track uuid's so we synthesis one using sha1 on
689 * - The signature (Which is constant for all imsm array, but no matter)
690 * - the family_num of the container
691 * - the index number of the volume
692 * - the 'serial' number of the volume.
693 * Hopefully these are all constant.
695 struct intel_super
*super
= st
->sb
;
699 struct imsm_dev
*dev
= NULL
;
702 sha1_process_bytes(super
->anchor
->sig
, MAX_SIGNATURE_LENGTH
, &ctx
);
703 sha1_process_bytes(&super
->anchor
->family_num
, sizeof(__u32
), &ctx
);
704 if (super
->current_vol
>= 0)
705 dev
= get_imsm_dev(super
, super
->current_vol
);
707 __u32 vol
= super
->current_vol
;
708 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
709 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
711 sha1_finish_ctx(&ctx
, buf
);
712 memcpy(uuid
, buf
, 4*4);
717 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
719 __u8
*v
= get_imsm_version(mpb
);
720 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
721 char major
[] = { 0, 0, 0 };
722 char minor
[] = { 0 ,0, 0 };
723 char patch
[] = { 0, 0, 0 };
724 char *ver_parse
[] = { major
, minor
, patch
};
728 while (*v
!= '\0' && v
< end
) {
729 if (*v
!= '.' && j
< 2)
730 ver_parse
[i
][j
++] = *v
;
738 *m
= strtol(minor
, NULL
, 0);
739 *p
= strtol(patch
, NULL
, 0);
743 static int imsm_level_to_layout(int level
)
751 return ALGORITHM_LEFT_ASYMMETRIC
;
758 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
760 struct intel_super
*super
= st
->sb
;
761 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
762 struct imsm_map
*map
= get_imsm_map(dev
, 0);
764 info
->container_member
= super
->current_vol
;
765 info
->array
.raid_disks
= map
->num_members
;
766 info
->array
.level
= get_imsm_raid_level(map
);
767 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
768 info
->array
.md_minor
= -1;
769 info
->array
.ctime
= 0;
770 info
->array
.utime
= 0;
771 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
772 info
->array
.state
= !dev
->vol
.dirty
;
774 info
->disk
.major
= 0;
775 info
->disk
.minor
= 0;
777 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
778 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
779 memset(info
->uuid
, 0, sizeof(info
->uuid
));
781 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
782 info
->resync_start
= 0;
783 else if (dev
->vol
.migr_state
)
784 info
->resync_start
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
786 info
->resync_start
= ~0ULL;
788 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
789 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
791 info
->array
.major_version
= -1;
792 info
->array
.minor_version
= -2;
793 sprintf(info
->text_version
, "/%s/%d",
794 devnum2devname(st
->container_dev
),
795 info
->container_member
);
796 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
797 uuid_from_super_imsm(st
, info
->uuid
);
801 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
803 struct intel_super
*super
= st
->sb
;
804 struct imsm_disk
*disk
;
807 if (super
->current_vol
>= 0) {
808 getinfo_super_imsm_volume(st
, info
);
812 /* Set raid_disks to zero so that Assemble will always pull in valid
815 info
->array
.raid_disks
= 0;
816 info
->array
.level
= LEVEL_CONTAINER
;
817 info
->array
.layout
= 0;
818 info
->array
.md_minor
= -1;
819 info
->array
.ctime
= 0; /* N/A for imsm */
820 info
->array
.utime
= 0;
821 info
->array
.chunk_size
= 0;
823 info
->disk
.major
= 0;
824 info
->disk
.minor
= 0;
825 info
->disk
.raid_disk
= -1;
826 info
->reshape_active
= 0;
827 info
->array
.major_version
= -1;
828 info
->array
.minor_version
= -2;
829 strcpy(info
->text_version
, "imsm");
830 info
->safe_mode_delay
= 0;
831 info
->disk
.number
= -1;
832 info
->disk
.state
= 0;
836 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
838 disk
= &super
->disks
->disk
;
839 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
840 info
->component_size
= reserved
;
841 s
= __le32_to_cpu(disk
->status
);
842 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
843 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
844 info
->disk
.state
|= s
& SPARE_DISK
? 0 : (1 << MD_DISK_SYNC
);
847 /* only call uuid_from_super_imsm when this disk is part of a populated container,
848 * ->compare_super may have updated the 'num_raid_devs' field for spares
850 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
851 uuid_from_super_imsm(st
, info
->uuid
);
853 memcpy(info
->uuid
, uuid_match_any
, sizeof(int[4]));
856 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
857 char *update
, char *devname
, int verbose
,
858 int uuid_set
, char *homehost
)
862 /* For 'assemble' and 'force' we need to return non-zero if any
863 * change was made. For others, the return value is ignored.
864 * Update options are:
865 * force-one : This device looks a bit old but needs to be included,
866 * update age info appropriately.
867 * assemble: clear any 'faulty' flag to allow this device to
869 * force-array: Array is degraded but being forced, mark it clean
870 * if that will be needed to assemble it.
872 * newdev: not used ????
873 * grow: Array has gained a new device - this is currently for
875 * resync: mark as dirty so a resync will happen.
876 * name: update the name - preserving the homehost
878 * Following are not relevant for this imsm:
879 * sparc2.2 : update from old dodgey metadata
880 * super-minor: change the preferred_minor number
881 * summaries: update redundant counters.
882 * uuid: Change the uuid of the array to match watch is given
883 * homehost: update the recorded homehost
884 * _reshape_progress: record new reshape_progress position.
887 //struct intel_super *super = st->sb;
888 //struct imsm_super *mpb = super->mpb;
890 if (strcmp(update
, "grow") == 0) {
892 if (strcmp(update
, "resync") == 0) {
893 /* dev->vol.dirty = 1; */
896 /* IMSM has no concept of UUID or homehost */
901 static size_t disks_to_mpb_size(int disks
)
905 size
= sizeof(struct imsm_super
);
906 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
907 size
+= 2 * sizeof(struct imsm_dev
);
908 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
909 size
+= (4 - 2) * sizeof(struct imsm_map
);
910 /* 4 possible disk_ord_tbl's */
911 size
+= 4 * (disks
- 1) * sizeof(__u32
);
916 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
918 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
921 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
924 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
928 * 0 same, or first was empty, and second was copied
929 * 1 second had wrong number
933 struct intel_super
*first
= st
->sb
;
934 struct intel_super
*sec
= tst
->sb
;
942 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
945 /* if an anchor does not have num_raid_devs set then it is a free
948 if (first
->anchor
->num_raid_devs
> 0 &&
949 sec
->anchor
->num_raid_devs
> 0) {
950 if (first
->anchor
->family_num
!= sec
->anchor
->family_num
)
954 /* if 'first' is a spare promote it to a populated mpb with sec's
957 if (first
->anchor
->num_raid_devs
== 0 &&
958 sec
->anchor
->num_raid_devs
> 0) {
961 /* we need to copy raid device info from sec if an allocation
962 * fails here we don't associate the spare
964 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
965 first
->dev_tbl
[i
] = malloc(sizeof(struct imsm_dev
));
966 if (!first
->dev_tbl
) {
968 free(first
->dev_tbl
[i
]);
969 first
->dev_tbl
[i
] = NULL
;
971 fprintf(stderr
, "imsm: failed to associate spare\n");
974 *first
->dev_tbl
[i
] = *sec
->dev_tbl
[i
];
977 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
978 first
->anchor
->family_num
= sec
->anchor
->family_num
;
984 static void fd2devname(int fd
, char *name
)
993 if (fstat(fd
, &st
) != 0)
995 sprintf(path
, "/sys/dev/block/%d:%d",
996 major(st
.st_rdev
), minor(st
.st_rdev
));
998 rv
= readlink(path
, dname
, sizeof(dname
));
1003 nm
= strrchr(dname
, '/');
1005 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
1009 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
1011 static int imsm_read_serial(int fd
, char *devname
,
1012 __u8 serial
[MAX_RAID_SERIAL_LEN
])
1014 unsigned char scsi_serial
[255];
1020 memset(scsi_serial
, 0, sizeof(scsi_serial
));
1022 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
1024 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
1025 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1026 fd2devname(fd
, (char *) serial
);
1033 Name
": Failed to retrieve serial for %s\n",
1038 /* trim leading whitespace */
1039 rsp_len
= scsi_serial
[3];
1040 rsp_buf
= (char *) &scsi_serial
[4];
1045 /* truncate len to the end of rsp_buf if necessary */
1046 if (c
+ MAX_RAID_SERIAL_LEN
> rsp_buf
+ rsp_len
)
1047 len
= rsp_len
- (c
- rsp_buf
);
1049 len
= MAX_RAID_SERIAL_LEN
;
1051 /* initialize the buffer and copy rsp_buf characters */
1052 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1053 memcpy(serial
, c
, len
);
1055 /* trim trailing whitespace starting with the last character copied */
1056 c
= (char *) &serial
[len
- 1];
1057 while (isspace(*c
) || *c
== '\0')
1063 static int serialcmp(__u8
*s1
, __u8
*s2
)
1065 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
1068 static void serialcpy(__u8
*dest
, __u8
*src
)
1070 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
1074 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
1081 __u8 serial
[MAX_RAID_SERIAL_LEN
];
1083 rv
= imsm_read_serial(fd
, devname
, serial
);
1088 /* check if this is a disk we have seen before. it may be a spare in
1089 * super->disks while the current anchor believes it is a raid member,
1090 * check if we need to update dl->index
1092 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1093 if (serialcmp(dl
->serial
, serial
) == 0)
1097 dl
= malloc(sizeof(*dl
));
1104 Name
": failed to allocate disk buffer for %s\n",
1111 dl
->major
= major(stb
.st_rdev
);
1112 dl
->minor
= minor(stb
.st_rdev
);
1113 dl
->next
= super
->disks
;
1114 dl
->fd
= keep_fd
? fd
: -1;
1115 dl
->devname
= devname
? strdup(devname
) : NULL
;
1116 serialcpy(dl
->serial
, serial
);
1118 } else if (keep_fd
) {
1123 /* look up this disk's index in the current anchor */
1124 for (i
= 0; i
< super
->anchor
->num_disks
; i
++) {
1125 struct imsm_disk
*disk_iter
;
1127 disk_iter
= __get_imsm_disk(super
->anchor
, i
);
1129 if (serialcmp(disk_iter
->serial
, dl
->serial
) == 0) {
1132 dl
->disk
= *disk_iter
;
1133 status
= __le32_to_cpu(dl
->disk
.status
);
1134 /* only set index on disks that are a member of a
1135 * populated contianer, i.e. one with raid_devs
1137 if (status
& FAILED_DISK
)
1139 else if (status
& SPARE_DISK
)
1148 /* no match, maybe a stale failed drive */
1149 if (i
== super
->anchor
->num_disks
&& dl
->index
>= 0) {
1150 dl
->disk
= *__get_imsm_disk(super
->anchor
, dl
->index
);
1151 if (__le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
1161 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
1163 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
1167 /* When migrating map0 contains the 'destination' state while map1
1168 * contains the current state. When not migrating map0 contains the
1169 * current state. This routine assumes that map[0].map_state is set to
1170 * the current array state before being called.
1172 * Migration is indicated by one of the following states
1173 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
1174 * 2/ Initialize (migr_state=1 migr_type=0 map0state=normal
1175 * map1state=unitialized)
1176 * 3/ Verify (Resync) (migr_state=1 migr_type=1 map0state=normal
1178 * 4/ Rebuild (migr_state=1 migr_type=1 map0state=normal
1179 * map1state=degraded)
1181 static void migrate(struct imsm_dev
*dev
, __u8 to_state
, int rebuild_resync
)
1183 struct imsm_map
*dest
;
1184 struct imsm_map
*src
= get_imsm_map(dev
, 0);
1186 dev
->vol
.migr_state
= 1;
1187 dev
->vol
.migr_type
= rebuild_resync
;
1188 dev
->vol
.curr_migr_unit
= 0;
1189 dest
= get_imsm_map(dev
, 1);
1191 memcpy(dest
, src
, sizeof_imsm_map(src
));
1192 src
->map_state
= to_state
;
1195 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
1197 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1199 dev
->vol
.migr_state
= 0;
1200 dev
->vol
.curr_migr_unit
= 0;
1201 map
->map_state
= map_state
;
1205 static int parse_raid_devices(struct intel_super
*super
)
1208 struct imsm_dev
*dev_new
;
1209 size_t len
, len_migr
;
1210 size_t space_needed
= 0;
1211 struct imsm_super
*mpb
= super
->anchor
;
1213 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1214 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
1216 len
= sizeof_imsm_dev(dev_iter
, 0);
1217 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
1219 space_needed
+= len_migr
- len
;
1221 dev_new
= malloc(len_migr
);
1224 imsm_copy_dev(dev_new
, dev_iter
);
1225 super
->dev_tbl
[i
] = dev_new
;
1228 /* ensure that super->buf is large enough when all raid devices
1231 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
1234 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
1235 if (posix_memalign(&buf
, 512, len
) != 0)
1238 memcpy(buf
, super
->buf
, len
);
1247 /* retrieve a pointer to the bbm log which starts after all raid devices */
1248 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
1252 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
1254 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
1260 static void __free_imsm(struct intel_super
*super
, int free_disks
);
1262 /* load_imsm_mpb - read matrix metadata
1263 * allocates super->mpb to be freed by free_super
1265 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
1267 unsigned long long dsize
;
1268 unsigned long long sectors
;
1270 struct imsm_super
*anchor
;
1274 get_dev_size(fd
, NULL
, &dsize
);
1276 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
1279 Name
": Cannot seek to anchor block on %s: %s\n",
1280 devname
, strerror(errno
));
1284 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
1287 Name
": Failed to allocate imsm anchor buffer"
1288 " on %s\n", devname
);
1291 if (read(fd
, anchor
, 512) != 512) {
1294 Name
": Cannot read anchor block on %s: %s\n",
1295 devname
, strerror(errno
));
1300 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
1303 Name
": no IMSM anchor on %s\n", devname
);
1308 __free_imsm(super
, 0);
1309 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
1310 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
1313 Name
": unable to allocate %zu byte mpb buffer\n",
1318 memcpy(super
->buf
, anchor
, 512);
1320 sectors
= mpb_sectors(anchor
) - 1;
1323 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1325 rc
= parse_raid_devices(super
);
1329 /* read the extended mpb */
1330 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
1333 Name
": Cannot seek to extended mpb on %s: %s\n",
1334 devname
, strerror(errno
));
1338 if (read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
1341 Name
": Cannot read extended mpb on %s: %s\n",
1342 devname
, strerror(errno
));
1346 check_sum
= __gen_imsm_checksum(super
->anchor
);
1347 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
1350 Name
": IMSM checksum %x != %x on %s\n",
1351 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
1356 /* FIXME the BBM log is disk specific so we cannot use this global
1357 * buffer for all disks. Ok for now since we only look at the global
1358 * bbm_log_size parameter to gate assembly
1360 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
1362 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1364 rc
= parse_raid_devices(super
);
1369 static void __free_imsm_disk(struct dl
*d
)
1378 static void free_imsm_disks(struct intel_super
*super
)
1382 while (super
->disks
) {
1384 super
->disks
= d
->next
;
1385 __free_imsm_disk(d
);
1387 while (super
->missing
) {
1389 super
->missing
= d
->next
;
1390 __free_imsm_disk(d
);
1395 /* free all the pieces hanging off of a super pointer */
1396 static void __free_imsm(struct intel_super
*super
, int free_disks
)
1405 free_imsm_disks(super
);
1406 for (i
= 0; i
< IMSM_MAX_RAID_DEVS
; i
++)
1407 if (super
->dev_tbl
[i
]) {
1408 free(super
->dev_tbl
[i
]);
1409 super
->dev_tbl
[i
] = NULL
;
1413 static void free_imsm(struct intel_super
*super
)
1415 __free_imsm(super
, 1);
1419 static void free_super_imsm(struct supertype
*st
)
1421 struct intel_super
*super
= st
->sb
;
1430 static struct intel_super
*alloc_super(int creating_imsm
)
1432 struct intel_super
*super
= malloc(sizeof(*super
));
1435 memset(super
, 0, sizeof(*super
));
1436 super
->creating_imsm
= creating_imsm
;
1437 super
->current_vol
= -1;
1444 /* find_missing - helper routine for load_super_imsm_all that identifies
1445 * disks that have disappeared from the system. This routine relies on
1446 * the mpb being uptodate, which it is at load time.
1448 static int find_missing(struct intel_super
*super
)
1451 struct imsm_super
*mpb
= super
->anchor
;
1453 struct imsm_disk
*disk
;
1456 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1457 disk
= __get_imsm_disk(mpb
, i
);
1458 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1459 if (serialcmp(dl
->disk
.serial
, disk
->serial
) == 0)
1463 /* ok we have a 'disk' without a live entry in
1466 status
= __le32_to_cpu(disk
->status
);
1467 if (status
& FAILED_DISK
|| !(status
& USABLE_DISK
))
1468 continue; /* never mind, already marked */
1470 dl
= malloc(sizeof(*dl
));
1476 dl
->devname
= strdup("missing");
1478 serialcpy(dl
->serial
, disk
->serial
);
1480 dl
->next
= super
->missing
;
1481 super
->missing
= dl
;
1487 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
1488 char *devname
, int keep_fd
)
1491 struct intel_super
*super
;
1492 struct mdinfo
*sd
, *best
= NULL
;
1499 /* check if this disk is a member of an active array */
1500 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
1504 if (sra
->array
.major_version
!= -1 ||
1505 sra
->array
.minor_version
!= -2 ||
1506 strcmp(sra
->text_version
, "imsm") != 0)
1509 super
= alloc_super(0);
1513 /* find the most up to date disk in this array, skipping spares */
1514 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1515 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1516 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1521 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1525 if (super
->anchor
->num_raid_devs
== 0)
1528 gen
= __le32_to_cpu(super
->anchor
->generation_num
);
1529 if (!best
|| gen
> bestgen
) {
1544 /* load the most up to date anchor */
1545 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
1546 dfd
= dev_open(nm
, O_RDONLY
);
1551 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1558 /* re-parse the disk list with the current anchor */
1559 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1560 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1561 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1566 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
1572 if (find_missing(super
) != 0) {
1577 if (st
->subarray
[0]) {
1578 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
1579 super
->current_vol
= atoi(st
->subarray
);
1585 st
->container_dev
= fd2devnum(fd
);
1586 if (st
->ss
== NULL
) {
1587 st
->ss
= &super_imsm
;
1588 st
->minor_version
= 0;
1589 st
->max_devs
= IMSM_MAX_DEVICES
;
1591 st
->loaded_container
= 1;
1597 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
1599 struct intel_super
*super
;
1603 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1606 if (st
->subarray
[0])
1607 return 1; /* FIXME */
1609 super
= alloc_super(0);
1612 Name
": malloc of %zu failed.\n",
1617 rv
= load_imsm_mpb(fd
, super
, devname
);
1622 Name
": Failed to load all information "
1623 "sections on %s\n", devname
);
1629 if (st
->ss
== NULL
) {
1630 st
->ss
= &super_imsm
;
1631 st
->minor_version
= 0;
1632 st
->max_devs
= IMSM_MAX_DEVICES
;
1634 st
->loaded_container
= 0;
1639 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1641 if (info
->level
== 1)
1643 return info
->chunk_size
>> 9;
1646 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1650 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1651 if (info
->level
== 1)
1657 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
1659 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
1662 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1663 unsigned long long size
, char *name
,
1664 char *homehost
, int *uuid
)
1666 /* We are creating a volume inside a pre-existing container.
1667 * so st->sb is already set.
1669 struct intel_super
*super
= st
->sb
;
1670 struct imsm_super
*mpb
= super
->anchor
;
1671 struct imsm_dev
*dev
;
1672 struct imsm_vol
*vol
;
1673 struct imsm_map
*map
;
1674 int idx
= mpb
->num_raid_devs
;
1676 unsigned long long array_blocks
;
1678 size_t size_old
, size_new
;
1680 if (mpb
->num_raid_devs
>= 2) {
1681 fprintf(stderr
, Name
": This imsm-container already has the "
1682 "maximum of 2 volumes\n");
1686 /* ensure the mpb is large enough for the new data */
1687 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1688 size_new
= disks_to_mpb_size(info
->nr_disks
);
1689 if (size_new
> size_old
) {
1691 size_t size_round
= ROUND_UP(size_new
, 512);
1693 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1694 fprintf(stderr
, Name
": could not allocate new mpb\n");
1697 memcpy(mpb_new
, mpb
, size_old
);
1700 super
->anchor
= mpb_new
;
1701 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1702 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1704 super
->current_vol
= idx
;
1705 /* when creating the first raid device in this container set num_disks
1706 * to zero, i.e. delete this spare and add raid member devices in
1707 * add_to_super_imsm_volume()
1709 if (super
->current_vol
== 0)
1711 sprintf(st
->subarray
, "%d", idx
);
1712 dev
= malloc(sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
1714 fprintf(stderr
, Name
": could not allocate raid device\n");
1717 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1718 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1719 info
->layout
, info
->chunk_size
,
1721 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1722 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1723 dev
->status
= __cpu_to_le32(0);
1724 dev
->reserved_blocks
= __cpu_to_le32(0);
1726 vol
->migr_state
= 0;
1729 vol
->curr_migr_unit
= 0;
1730 for (i
= 0; i
< idx
; i
++) {
1731 struct imsm_dev
*prev
= get_imsm_dev(super
, i
);
1732 struct imsm_map
*pmap
= get_imsm_map(prev
, 0);
1734 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1735 offset
+= IMSM_RESERVED_SECTORS
;
1737 map
= get_imsm_map(dev
, 0);
1738 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1739 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
1740 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1741 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1742 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1743 IMSM_T_STATE_NORMAL
;
1745 if (info
->level
== 1 && info
->raid_disks
> 2) {
1746 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1747 "in a raid1 volume\n");
1750 if (info
->level
== 10)
1751 map
->raid_level
= 1;
1753 map
->raid_level
= info
->level
;
1755 map
->num_members
= info
->raid_disks
;
1756 for (i
= 0; i
< map
->num_members
; i
++) {
1757 /* initialized in add_to_super */
1758 set_imsm_ord_tbl_ent(map
, i
, 0);
1760 mpb
->num_raid_devs
++;
1761 super
->dev_tbl
[super
->current_vol
] = dev
;
1766 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1767 unsigned long long size
, char *name
,
1768 char *homehost
, int *uuid
)
1770 /* This is primarily called by Create when creating a new array.
1771 * We will then get add_to_super called for each component, and then
1772 * write_init_super called to write it out to each device.
1773 * For IMSM, Create can create on fresh devices or on a pre-existing
1775 * To create on a pre-existing array a different method will be called.
1776 * This one is just for fresh drives.
1778 struct intel_super
*super
;
1779 struct imsm_super
*mpb
;
1787 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1790 super
= alloc_super(1);
1793 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1794 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1799 memset(mpb
, 0, mpb_size
);
1801 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1802 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1803 strlen(MPB_VERSION_RAID5
));
1804 mpb
->mpb_size
= mpb_size
;
1811 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1812 int fd
, char *devname
)
1814 struct intel_super
*super
= st
->sb
;
1815 struct imsm_super
*mpb
= super
->anchor
;
1817 struct imsm_dev
*dev
;
1818 struct imsm_map
*map
;
1821 dev
= get_imsm_dev(super
, super
->current_vol
);
1822 map
= get_imsm_map(dev
, 0);
1824 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1825 if (dl
->major
== dk
->major
&&
1826 dl
->minor
== dk
->minor
)
1829 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1832 /* add a pristine spare to the metadata */
1833 if (dl
->index
< 0) {
1834 dl
->index
= super
->anchor
->num_disks
;
1835 super
->anchor
->num_disks
++;
1837 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
1838 status
= CONFIGURED_DISK
| USABLE_DISK
;
1839 dl
->disk
.status
= __cpu_to_le32(status
);
1841 /* if we are creating the first raid device update the family number */
1842 if (super
->current_vol
== 0) {
1844 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
1845 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
1849 sum
= __gen_imsm_checksum(mpb
);
1850 mpb
->family_num
= __cpu_to_le32(sum
);
1854 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1855 int fd
, char *devname
)
1857 struct intel_super
*super
= st
->sb
;
1859 unsigned long long size
;
1864 if (super
->current_vol
>= 0) {
1865 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1870 dd
= malloc(sizeof(*dd
));
1873 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1876 memset(dd
, 0, sizeof(*dd
));
1877 dd
->major
= major(stb
.st_rdev
);
1878 dd
->minor
= minor(stb
.st_rdev
);
1880 dd
->devname
= devname
? strdup(devname
) : NULL
;
1882 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1885 Name
": failed to retrieve scsi serial, aborting\n");
1890 get_dev_size(fd
, NULL
, &size
);
1892 status
= USABLE_DISK
| SPARE_DISK
;
1893 serialcpy(dd
->disk
.serial
, dd
->serial
);
1894 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
1895 dd
->disk
.status
= __cpu_to_le32(status
);
1896 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1897 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
1899 dd
->disk
.scsi_id
= __cpu_to_le32(0);
1901 if (st
->update_tail
) {
1902 dd
->next
= super
->add
;
1905 dd
->next
= super
->disks
;
1910 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1912 /* spare records have their own family number and do not have any defined raid
1915 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
1917 struct imsm_super mpb_save
;
1918 struct imsm_super
*mpb
= super
->anchor
;
1923 mpb
->num_raid_devs
= 0;
1925 mpb
->mpb_size
= sizeof(struct imsm_super
);
1926 mpb
->generation_num
= __cpu_to_le32(1UL);
1928 for (d
= super
->disks
; d
; d
= d
->next
) {
1932 mpb
->disk
[0] = d
->disk
;
1933 sum
= __gen_imsm_checksum(mpb
);
1934 mpb
->family_num
= __cpu_to_le32(sum
);
1935 sum
= __gen_imsm_checksum(mpb
);
1936 mpb
->check_sum
= __cpu_to_le32(sum
);
1938 if (store_imsm_mpb(d
->fd
, super
)) {
1939 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1940 __func__
, d
->major
, d
->minor
, strerror(errno
));
1954 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1956 struct imsm_super
*mpb
= super
->anchor
;
1962 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
1964 /* 'generation' is incremented everytime the metadata is written */
1965 generation
= __le32_to_cpu(mpb
->generation_num
);
1967 mpb
->generation_num
= __cpu_to_le32(generation
);
1969 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
1970 for (d
= super
->disks
; d
; d
= d
->next
) {
1974 mpb
->disk
[d
->index
] = d
->disk
;
1976 for (d
= super
->missing
; d
; d
= d
->next
)
1977 mpb
->disk
[d
->index
] = d
->disk
;
1979 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1980 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1982 imsm_copy_dev(dev
, super
->dev_tbl
[i
]);
1983 mpb_size
+= sizeof_imsm_dev(dev
, 0);
1985 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
1986 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
1988 /* recalculate checksum */
1989 sum
= __gen_imsm_checksum(mpb
);
1990 mpb
->check_sum
= __cpu_to_le32(sum
);
1992 /* write the mpb for disks that compose raid devices */
1993 for (d
= super
->disks
; d
; d
= d
->next
) {
1996 if (store_imsm_mpb(d
->fd
, super
))
1997 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1998 __func__
, d
->major
, d
->minor
, strerror(errno
));
2006 return write_super_imsm_spares(super
, doclose
);
2012 static int create_array(struct supertype
*st
)
2015 struct imsm_update_create_array
*u
;
2016 struct intel_super
*super
= st
->sb
;
2017 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
2019 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0);
2022 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2027 u
->type
= update_create_array
;
2028 u
->dev_idx
= super
->current_vol
;
2029 imsm_copy_dev(&u
->dev
, dev
);
2030 append_metadata_update(st
, u
, len
);
2035 static int _add_disk(struct supertype
*st
)
2037 struct intel_super
*super
= st
->sb
;
2039 struct imsm_update_add_disk
*u
;
2047 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2052 u
->type
= update_add_disk
;
2053 append_metadata_update(st
, u
, len
);
2058 static int write_init_super_imsm(struct supertype
*st
)
2060 if (st
->update_tail
) {
2061 /* queue the recently created array / added disk
2062 * as a metadata update */
2063 struct intel_super
*super
= st
->sb
;
2067 /* determine if we are creating a volume or adding a disk */
2068 if (super
->current_vol
< 0) {
2069 /* in the add disk case we are running in mdmon
2070 * context, so don't close fd's
2072 return _add_disk(st
);
2074 rv
= create_array(st
);
2076 for (d
= super
->disks
; d
; d
= d
->next
) {
2083 return write_super_imsm(st
->sb
, 1);
2087 static int store_zero_imsm(struct supertype
*st
, int fd
)
2089 unsigned long long dsize
;
2092 get_dev_size(fd
, NULL
, &dsize
);
2094 /* first block is stored on second to last sector of the disk */
2095 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2098 if (posix_memalign(&buf
, 512, 512) != 0)
2101 memset(buf
, 0, 512);
2102 if (write(fd
, buf
, 512) != 512)
2107 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
2109 return __le32_to_cpu(mpb
->bbm_log_size
);
2113 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
2114 int layout
, int raiddisks
, int chunk
,
2115 unsigned long long size
, char *dev
,
2116 unsigned long long *freesize
,
2120 unsigned long long ldsize
;
2122 if (level
!= LEVEL_CONTAINER
)
2127 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2130 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
2131 dev
, strerror(errno
));
2134 if (!get_dev_size(fd
, dev
, &ldsize
)) {
2140 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
2145 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
2146 * FIX ME add ahci details
2148 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
2149 int layout
, int raiddisks
, int chunk
,
2150 unsigned long long size
, char *dev
,
2151 unsigned long long *freesize
,
2155 struct intel_super
*super
= st
->sb
;
2157 unsigned long long pos
= 0;
2158 unsigned long long maxsize
;
2162 if (level
== LEVEL_CONTAINER
)
2165 if (level
== 1 && raiddisks
> 2) {
2167 fprintf(stderr
, Name
": imsm does not support more "
2168 "than 2 in a raid1 configuration\n");
2172 /* We must have the container info already read in. */
2177 /* General test: make sure there is space for
2178 * 'raiddisks' device extents of size 'size' at a given
2181 unsigned long long minsize
= size
*2 /* convert to blocks */;
2182 unsigned long long start_offset
= ~0ULL;
2185 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
2186 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2191 e
= get_extents(super
, dl
);
2194 unsigned long long esize
;
2195 esize
= e
[i
].start
- pos
;
2196 if (esize
>= minsize
)
2198 if (found
&& start_offset
== ~0ULL) {
2201 } else if (found
&& pos
!= start_offset
) {
2205 pos
= e
[i
].start
+ e
[i
].size
;
2207 } while (e
[i
-1].size
);
2212 if (dcnt
< raiddisks
) {
2214 fprintf(stderr
, Name
": imsm: Not enough "
2215 "devices with space for this array "
2222 /* This device must be a member of the set */
2223 if (stat(dev
, &stb
) < 0)
2225 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
2227 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2228 if (dl
->major
== major(stb
.st_rdev
) &&
2229 dl
->minor
== minor(stb
.st_rdev
))
2234 fprintf(stderr
, Name
": %s is not in the "
2235 "same imsm set\n", dev
);
2238 e
= get_extents(super
, dl
);
2242 unsigned long long esize
;
2243 esize
= e
[i
].start
- pos
;
2244 if (esize
>= maxsize
)
2246 pos
= e
[i
].start
+ e
[i
].size
;
2248 } while (e
[i
-1].size
);
2249 *freesize
= maxsize
;
2254 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
2255 int raiddisks
, int chunk
, unsigned long long size
,
2256 char *dev
, unsigned long long *freesize
,
2262 /* if given unused devices create a container
2263 * if given given devices in a container create a member volume
2265 if (level
== LEVEL_CONTAINER
) {
2266 /* Must be a fresh device to add to a container */
2267 return validate_geometry_imsm_container(st
, level
, layout
,
2268 raiddisks
, chunk
, size
,
2274 /* creating in a given container */
2275 return validate_geometry_imsm_volume(st
, level
, layout
,
2276 raiddisks
, chunk
, size
,
2277 dev
, freesize
, verbose
);
2280 /* limit creation to the following levels */
2292 /* This device needs to be a device in an 'imsm' container */
2293 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2297 Name
": Cannot create this array on device %s\n",
2302 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
2304 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2305 dev
, strerror(errno
));
2308 /* Well, it is in use by someone, maybe an 'imsm' container. */
2309 cfd
= open_container(fd
);
2313 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
2317 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
2319 if (sra
&& sra
->array
.major_version
== -1 &&
2320 strcmp(sra
->text_version
, "imsm") == 0) {
2321 /* This is a member of a imsm container. Load the container
2322 * and try to create a volume
2324 struct intel_super
*super
;
2326 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
2328 st
->container_dev
= fd2devnum(cfd
);
2330 return validate_geometry_imsm_volume(st
, level
, layout
,
2336 } else /* may belong to another container */
2341 #endif /* MDASSEMBLE */
2343 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
2345 /* Given a container loaded by load_super_imsm_all,
2346 * extract information about all the arrays into
2349 * For each imsm_dev create an mdinfo, fill it in,
2350 * then look for matching devices in super->disks
2351 * and create appropriate device mdinfo.
2353 struct intel_super
*super
= st
->sb
;
2354 struct imsm_super
*mpb
= super
->anchor
;
2355 struct mdinfo
*rest
= NULL
;
2358 /* do not assemble arrays that might have bad blocks */
2359 if (imsm_bbm_log_size(super
->anchor
)) {
2360 fprintf(stderr
, Name
": BBM log found in metadata. "
2361 "Cannot activate array(s).\n");
2365 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2366 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2367 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2368 struct mdinfo
*this;
2371 this = malloc(sizeof(*this));
2372 memset(this, 0, sizeof(*this));
2375 super
->current_vol
= i
;
2376 getinfo_super_imsm_volume(st
, this);
2377 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
2378 struct mdinfo
*info_d
;
2386 idx
= get_imsm_disk_idx(dev
, slot
);
2387 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
2388 for (d
= super
->disks
; d
; d
= d
->next
)
2389 if (d
->index
== idx
)
2395 s
= d
? __le32_to_cpu(d
->disk
.status
) : 0;
2396 if (s
& FAILED_DISK
)
2398 if (!(s
& USABLE_DISK
))
2400 if (ord
& IMSM_ORD_REBUILD
)
2404 * if we skip some disks the array will be assmebled degraded;
2405 * reset resync start to avoid a dirty-degraded situation
2407 * FIXME handle dirty degraded
2409 if (skip
&& !dev
->vol
.dirty
)
2410 this->resync_start
= ~0ULL;
2414 info_d
= malloc(sizeof(*info_d
));
2416 fprintf(stderr
, Name
": failed to allocate disk"
2417 " for volume %s\n", (char *) dev
->volume
);
2422 memset(info_d
, 0, sizeof(*info_d
));
2423 info_d
->next
= this->devs
;
2424 this->devs
= info_d
;
2426 info_d
->disk
.number
= d
->index
;
2427 info_d
->disk
.major
= d
->major
;
2428 info_d
->disk
.minor
= d
->minor
;
2429 info_d
->disk
.raid_disk
= slot
;
2431 this->array
.working_disks
++;
2433 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
2434 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2435 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
2437 strcpy(info_d
->name
, d
->devname
);
2447 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
2450 struct intel_super
*super
= c
->sb
;
2451 struct imsm_super
*mpb
= super
->anchor
;
2453 if (atoi(inst
) >= mpb
->num_raid_devs
) {
2454 fprintf(stderr
, "%s: subarry index %d, out of range\n",
2455 __func__
, atoi(inst
));
2459 dprintf("imsm: open_new %s\n", inst
);
2460 a
->info
.container_member
= atoi(inst
);
2464 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
2466 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2469 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
2470 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
2472 switch (get_imsm_raid_level(map
)) {
2474 return IMSM_T_STATE_FAILED
;
2477 if (failed
< map
->num_members
)
2478 return IMSM_T_STATE_DEGRADED
;
2480 return IMSM_T_STATE_FAILED
;
2485 * check to see if any mirrors have failed, otherwise we
2486 * are degraded. Even numbered slots are mirrored on
2490 /* gcc -Os complains that this is unused */
2491 int insync
= insync
;
2493 for (i
= 0; i
< map
->num_members
; i
++) {
2494 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2495 int idx
= ord_to_idx(ord
);
2496 struct imsm_disk
*disk
;
2498 /* reset the potential in-sync count on even-numbered
2499 * slots. num_copies is always 2 for imsm raid10
2504 disk
= get_imsm_disk(super
, idx
);
2506 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2507 ord
& IMSM_ORD_REBUILD
)
2510 /* no in-sync disks left in this mirror the
2514 return IMSM_T_STATE_FAILED
;
2517 return IMSM_T_STATE_DEGRADED
;
2521 return IMSM_T_STATE_DEGRADED
;
2523 return IMSM_T_STATE_FAILED
;
2529 return map
->map_state
;
2532 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
2536 struct imsm_disk
*disk
;
2537 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2539 for (i
= 0; i
< map
->num_members
; i
++) {
2540 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2541 int idx
= ord_to_idx(ord
);
2543 disk
= get_imsm_disk(super
, idx
);
2545 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2546 ord
& IMSM_ORD_REBUILD
)
2553 static int is_resyncing(struct imsm_dev
*dev
)
2555 struct imsm_map
*migr_map
;
2557 if (!dev
->vol
.migr_state
)
2560 if (dev
->vol
.migr_type
== 0)
2563 migr_map
= get_imsm_map(dev
, 1);
2565 if (migr_map
->map_state
== IMSM_T_STATE_NORMAL
)
2571 static int is_rebuilding(struct imsm_dev
*dev
)
2573 struct imsm_map
*migr_map
;
2575 if (!dev
->vol
.migr_state
)
2578 if (dev
->vol
.migr_type
== 0)
2581 migr_map
= get_imsm_map(dev
, 1);
2583 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
2589 static void mark_failure(struct imsm_disk
*disk
)
2591 __u32 status
= __le32_to_cpu(disk
->status
);
2593 if (status
& FAILED_DISK
)
2595 status
|= FAILED_DISK
;
2596 disk
->status
= __cpu_to_le32(status
);
2597 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
2598 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
2601 /* Handle dirty -> clean transititions and resync. Degraded and rebuild
2602 * states are handled in imsm_set_disk() with one exception, when a
2603 * resync is stopped due to a new failure this routine will set the
2604 * 'degraded' state for the array.
2606 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
2608 int inst
= a
->info
.container_member
;
2609 struct intel_super
*super
= a
->container
->sb
;
2610 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2611 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2612 int failed
= imsm_count_failed(super
, dev
);
2613 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
2615 /* before we activate this array handle any missing disks */
2616 if (consistent
== 2 && super
->missing
) {
2619 dprintf("imsm: mark missing\n");
2620 end_migration(dev
, map_state
);
2621 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
2622 mark_failure(&dl
->disk
);
2623 super
->updates_pending
++;
2626 if (consistent
== 2 &&
2627 (!is_resync_complete(a
) ||
2628 map_state
!= IMSM_T_STATE_NORMAL
||
2629 dev
->vol
.migr_state
))
2632 if (is_resync_complete(a
)) {
2633 /* complete intialization / resync,
2634 * recovery is completed in ->set_disk
2636 if (is_resyncing(dev
)) {
2637 dprintf("imsm: mark resync done\n");
2638 end_migration(dev
, map_state
);
2639 super
->updates_pending
++;
2641 } else if (!is_resyncing(dev
) && !failed
) {
2642 /* mark the start of the init process if nothing is failed */
2643 dprintf("imsm: mark resync start (%llu)\n", a
->resync_start
);
2644 map
->map_state
= map_state
;
2645 migrate(dev
, IMSM_T_STATE_NORMAL
,
2646 map
->map_state
== IMSM_T_STATE_NORMAL
);
2647 super
->updates_pending
++;
2650 /* check if we can update the migration checkpoint */
2651 if (dev
->vol
.migr_state
&&
2652 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != a
->resync_start
) {
2653 dprintf("imsm: checkpoint migration (%llu)\n", a
->resync_start
);
2654 dev
->vol
.curr_migr_unit
= __cpu_to_le32(a
->resync_start
);
2655 super
->updates_pending
++;
2658 /* mark dirty / clean */
2659 if (dev
->vol
.dirty
!= !consistent
) {
2660 dprintf("imsm: mark '%s' (%llu)\n",
2661 consistent
? "clean" : "dirty", a
->resync_start
);
2666 super
->updates_pending
++;
2671 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
2673 int inst
= a
->info
.container_member
;
2674 struct intel_super
*super
= a
->container
->sb
;
2675 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2676 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2677 struct imsm_disk
*disk
;
2683 if (n
> map
->num_members
)
2684 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
2685 n
, map
->num_members
- 1);
2690 dprintf("imsm: set_disk %d:%x\n", n
, state
);
2692 ord
= get_imsm_ord_tbl_ent(dev
, n
);
2693 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
2695 /* check for new failures */
2696 status
= __le32_to_cpu(disk
->status
);
2697 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
2699 super
->updates_pending
++;
2702 /* check if in_sync */
2703 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
) {
2704 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
2706 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
2707 super
->updates_pending
++;
2710 failed
= imsm_count_failed(super
, dev
);
2711 map_state
= imsm_check_degraded(super
, dev
, failed
);
2713 /* check if recovery complete, newly degraded, or failed */
2714 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
2715 end_migration(dev
, map_state
);
2716 super
->updates_pending
++;
2717 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
2718 map
->map_state
!= map_state
&&
2719 !dev
->vol
.migr_state
) {
2720 dprintf("imsm: mark degraded\n");
2721 map
->map_state
= map_state
;
2722 super
->updates_pending
++;
2723 } else if (map_state
== IMSM_T_STATE_FAILED
&&
2724 map
->map_state
!= map_state
) {
2725 dprintf("imsm: mark failed\n");
2726 end_migration(dev
, map_state
);
2727 super
->updates_pending
++;
2731 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
2733 struct imsm_super
*mpb
= super
->anchor
;
2734 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
2735 unsigned long long dsize
;
2736 unsigned long long sectors
;
2738 get_dev_size(fd
, NULL
, &dsize
);
2740 if (mpb_size
> 512) {
2741 /* -1 to account for anchor */
2742 sectors
= mpb_sectors(mpb
) - 1;
2744 /* write the extended mpb to the sectors preceeding the anchor */
2745 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
2748 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
2752 /* first block is stored on second to last sector of the disk */
2753 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2756 if (write(fd
, super
->buf
, 512) != 512)
2762 static void imsm_sync_metadata(struct supertype
*container
)
2764 struct intel_super
*super
= container
->sb
;
2766 if (!super
->updates_pending
)
2769 write_super_imsm(super
, 0);
2771 super
->updates_pending
= 0;
2774 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
2776 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2777 int i
= get_imsm_disk_idx(dev
, idx
);
2780 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2784 if (dl
&& __le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
2788 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
2793 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
, struct active_array
*a
)
2795 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2796 int idx
= get_imsm_disk_idx(dev
, slot
);
2797 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2798 unsigned long long esize
;
2799 unsigned long long pos
;
2808 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2809 /* If in this array, skip */
2810 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2811 if (d
->state_fd
>= 0 &&
2812 d
->disk
.major
== dl
->major
&&
2813 d
->disk
.minor
== dl
->minor
) {
2814 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
2820 /* skip in use or failed drives */
2821 status
= __le32_to_cpu(dl
->disk
.status
);
2822 if (status
& FAILED_DISK
|| idx
== dl
->index
) {
2823 dprintf("%x:%x status ( %s%s)\n",
2824 dl
->major
, dl
->minor
,
2825 status
& FAILED_DISK
? "failed " : "",
2826 idx
== dl
->index
? "in use " : "");
2830 /* Does this unused device have the requisite free space?
2831 * We need a->info.component_size sectors
2833 ex
= get_extents(super
, dl
);
2835 dprintf("cannot get extents\n");
2841 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
2844 /* check that we can start at pba_of_lba0 with
2845 * a->info.component_size of space
2847 esize
= ex
[j
].start
- pos
;
2848 if (array_start
>= pos
&&
2849 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
2853 pos
= ex
[j
].start
+ ex
[j
].size
;
2856 } while (ex
[j
-1].size
);
2860 dprintf("%x:%x does not have %llu at %d\n",
2861 dl
->major
, dl
->minor
,
2862 a
->info
.component_size
,
2863 __le32_to_cpu(map
->pba_of_lba0
));
2873 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
2874 struct metadata_update
**updates
)
2877 * Find a device with unused free space and use it to replace a
2878 * failed/vacant region in an array. We replace failed regions one a
2879 * array at a time. The result is that a new spare disk will be added
2880 * to the first failed array and after the monitor has finished
2881 * propagating failures the remainder will be consumed.
2883 * FIXME add a capability for mdmon to request spares from another
2887 struct intel_super
*super
= a
->container
->sb
;
2888 int inst
= a
->info
.container_member
;
2889 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2890 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2891 int failed
= a
->info
.array
.raid_disks
;
2892 struct mdinfo
*rv
= NULL
;
2895 struct metadata_update
*mu
;
2897 struct imsm_update_activate_spare
*u
;
2901 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
2902 if ((d
->curr_state
& DS_FAULTY
) &&
2904 /* wait for Removal to happen */
2906 if (d
->state_fd
>= 0)
2910 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
2911 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
2912 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
2915 /* For each slot, if it is not working, find a spare */
2916 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
2917 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2918 if (d
->disk
.raid_disk
== i
)
2920 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
2921 if (d
&& (d
->state_fd
>= 0))
2925 * OK, this device needs recovery. Try to re-add the previous
2926 * occupant of this slot, if this fails add a new spare
2928 dl
= imsm_readd(super
, i
, a
);
2930 dl
= imsm_add_spare(super
, i
, a
);
2934 /* found a usable disk with enough space */
2935 di
= malloc(sizeof(*di
));
2938 memset(di
, 0, sizeof(*di
));
2940 /* dl->index will be -1 in the case we are activating a
2941 * pristine spare. imsm_process_update() will create a
2942 * new index in this case. Once a disk is found to be
2943 * failed in all member arrays it is kicked from the
2946 di
->disk
.number
= dl
->index
;
2948 /* (ab)use di->devs to store a pointer to the device
2951 di
->devs
= (struct mdinfo
*) dl
;
2953 di
->disk
.raid_disk
= i
;
2954 di
->disk
.major
= dl
->major
;
2955 di
->disk
.minor
= dl
->minor
;
2957 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2958 di
->component_size
= a
->info
.component_size
;
2959 di
->container_member
= inst
;
2963 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2964 i
, di
->data_offset
);
2970 /* No spares found */
2972 /* Now 'rv' has a list of devices to return.
2973 * Create a metadata_update record to update the
2974 * disk_ord_tbl for the array
2976 mu
= malloc(sizeof(*mu
));
2978 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2979 if (mu
->buf
== NULL
) {
2986 struct mdinfo
*n
= rv
->next
;
2995 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
2996 mu
->next
= *updates
;
2997 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
2999 for (di
= rv
; di
; di
= di
->next
) {
3000 u
->type
= update_activate_spare
;
3001 u
->dl
= (struct dl
*) di
->devs
;
3003 u
->slot
= di
->disk
.raid_disk
;
3014 static int disks_overlap(struct imsm_dev
*d1
, struct imsm_dev
*d2
)
3016 struct imsm_map
*m1
= get_imsm_map(d1
, 0);
3017 struct imsm_map
*m2
= get_imsm_map(d2
, 0);
3022 for (i
= 0; i
< m1
->num_members
; i
++) {
3023 idx
= get_imsm_disk_idx(d1
, i
);
3024 for (j
= 0; j
< m2
->num_members
; j
++)
3025 if (idx
== get_imsm_disk_idx(d2
, j
))
3032 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
);
3034 static void imsm_process_update(struct supertype
*st
,
3035 struct metadata_update
*update
)
3038 * crack open the metadata_update envelope to find the update record
3039 * update can be one of:
3040 * update_activate_spare - a spare device has replaced a failed
3041 * device in an array, update the disk_ord_tbl. If this disk is
3042 * present in all member arrays then also clear the SPARE_DISK
3045 struct intel_super
*super
= st
->sb
;
3046 struct imsm_super
*mpb
;
3047 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3049 /* update requires a larger buf but the allocation failed */
3050 if (super
->next_len
&& !super
->next_buf
) {
3051 super
->next_len
= 0;
3055 if (super
->next_buf
) {
3056 memcpy(super
->next_buf
, super
->buf
, super
->len
);
3058 super
->len
= super
->next_len
;
3059 super
->buf
= super
->next_buf
;
3061 super
->next_len
= 0;
3062 super
->next_buf
= NULL
;
3065 mpb
= super
->anchor
;
3068 case update_activate_spare
: {
3069 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
3070 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
3071 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3072 struct imsm_map
*migr_map
;
3073 struct active_array
*a
;
3074 struct imsm_disk
*disk
;
3080 int victim
= get_imsm_disk_idx(dev
, u
->slot
);
3083 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
3088 fprintf(stderr
, "error: imsm_activate_spare passed "
3089 "an unknown disk (index: %d)\n",
3094 super
->updates_pending
++;
3096 /* count failures (excluding rebuilds and the victim)
3097 * to determine map[0] state
3100 for (i
= 0; i
< map
->num_members
; i
++) {
3103 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3105 __le32_to_cpu(disk
->status
) & FAILED_DISK
)
3109 /* adding a pristine spare, assign a new index */
3110 if (dl
->index
< 0) {
3111 dl
->index
= super
->anchor
->num_disks
;
3112 super
->anchor
->num_disks
++;
3115 status
= __le32_to_cpu(disk
->status
);
3116 status
|= CONFIGURED_DISK
;
3117 status
&= ~SPARE_DISK
;
3118 disk
->status
= __cpu_to_le32(status
);
3121 to_state
= imsm_check_degraded(super
, dev
, failed
);
3122 map
->map_state
= IMSM_T_STATE_DEGRADED
;
3123 migrate(dev
, to_state
, 1);
3124 migr_map
= get_imsm_map(dev
, 1);
3125 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
3126 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
3128 /* count arrays using the victim in the metadata */
3130 for (a
= st
->arrays
; a
; a
= a
->next
) {
3131 dev
= get_imsm_dev(super
, a
->info
.container_member
);
3132 for (i
= 0; i
< map
->num_members
; i
++)
3133 if (victim
== get_imsm_disk_idx(dev
, i
))
3137 /* delete the victim if it is no longer being
3143 /* We know that 'manager' isn't touching anything,
3144 * so it is safe to delete
3146 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
3147 if ((*dlp
)->index
== victim
)
3150 /* victim may be on the missing list */
3152 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
3153 if ((*dlp
)->index
== victim
)
3155 imsm_delete(super
, dlp
, victim
);
3159 case update_create_array
: {
3160 /* someone wants to create a new array, we need to be aware of
3161 * a few races/collisions:
3162 * 1/ 'Create' called by two separate instances of mdadm
3163 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
3164 * devices that have since been assimilated via
3166 * In the event this update can not be carried out mdadm will
3167 * (FIX ME) notice that its update did not take hold.
3169 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3170 struct imsm_dev
*dev
;
3171 struct imsm_map
*map
, *new_map
;
3172 unsigned long long start
, end
;
3173 unsigned long long new_start
, new_end
;
3177 /* handle racing creates: first come first serve */
3178 if (u
->dev_idx
< mpb
->num_raid_devs
) {
3179 dprintf("%s: subarray %d already defined\n",
3180 __func__
, u
->dev_idx
);
3184 /* check update is next in sequence */
3185 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
3186 dprintf("%s: can not create array %d expected index %d\n",
3187 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
3191 new_map
= get_imsm_map(&u
->dev
, 0);
3192 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
3193 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
3195 /* handle activate_spare versus create race:
3196 * check to make sure that overlapping arrays do not include
3199 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3200 dev
= get_imsm_dev(super
, i
);
3201 map
= get_imsm_map(dev
, 0);
3202 start
= __le32_to_cpu(map
->pba_of_lba0
);
3203 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
3204 if ((new_start
>= start
&& new_start
<= end
) ||
3205 (start
>= new_start
&& start
<= new_end
))
3207 if (overlap
&& disks_overlap(dev
, &u
->dev
)) {
3208 dprintf("%s: arrays overlap\n", __func__
);
3212 /* check num_members sanity */
3213 if (new_map
->num_members
> mpb
->num_disks
) {
3214 dprintf("%s: num_disks out of range\n", __func__
);
3218 /* check that prepare update was successful */
3219 if (!update
->space
) {
3220 dprintf("%s: prepare update failed\n", __func__
);
3224 super
->updates_pending
++;
3225 dev
= update
->space
;
3226 map
= get_imsm_map(dev
, 0);
3227 update
->space
= NULL
;
3228 imsm_copy_dev(dev
, &u
->dev
);
3229 map
= get_imsm_map(dev
, 0);
3230 super
->dev_tbl
[u
->dev_idx
] = dev
;
3231 mpb
->num_raid_devs
++;
3234 for (i
= 0; i
< map
->num_members
; i
++) {
3235 struct imsm_disk
*disk
;
3238 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3239 status
= __le32_to_cpu(disk
->status
);
3240 status
|= CONFIGURED_DISK
;
3241 status
&= ~SPARE_DISK
;
3242 disk
->status
= __cpu_to_le32(status
);
3246 case update_add_disk
:
3248 /* we may be able to repair some arrays if disks are
3251 struct active_array
*a
;
3253 super
->updates_pending
++;
3254 for (a
= st
->arrays
; a
; a
= a
->next
)
3255 a
->check_degraded
= 1;
3257 /* add some spares to the metadata */
3258 while (super
->add
) {
3262 super
->add
= al
->next
;
3263 al
->next
= super
->disks
;
3265 dprintf("%s: added %x:%x\n",
3266 __func__
, al
->major
, al
->minor
);
3273 static void imsm_prepare_update(struct supertype
*st
,
3274 struct metadata_update
*update
)
3277 * Allocate space to hold new disk entries, raid-device entries or a new
3278 * mpb if necessary. The manager synchronously waits for updates to
3279 * complete in the monitor, so new mpb buffers allocated here can be
3280 * integrated by the monitor thread without worrying about live pointers
3281 * in the manager thread.
3283 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3284 struct intel_super
*super
= st
->sb
;
3285 struct imsm_super
*mpb
= super
->anchor
;
3290 case update_create_array
: {
3291 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3293 len
= sizeof_imsm_dev(&u
->dev
, 1);
3294 update
->space
= malloc(len
);
3301 /* check if we need a larger metadata buffer */
3302 if (super
->next_buf
)
3303 buf_len
= super
->next_len
;
3305 buf_len
= super
->len
;
3307 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
3308 /* ok we need a larger buf than what is currently allocated
3309 * if this allocation fails process_update will notice that
3310 * ->next_len is set and ->next_buf is NULL
3312 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
3313 if (super
->next_buf
)
3314 free(super
->next_buf
);
3316 super
->next_len
= buf_len
;
3317 if (posix_memalign(&super
->next_buf
, buf_len
, 512) != 0)
3318 super
->next_buf
= NULL
;
3322 /* must be called while manager is quiesced */
3323 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
)
3325 struct imsm_super
*mpb
= super
->anchor
;
3327 struct imsm_dev
*dev
;
3328 struct imsm_map
*map
;
3329 int i
, j
, num_members
;
3332 dprintf("%s: deleting device[%d] from imsm_super\n",
3335 /* shift all indexes down one */
3336 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
3337 if (iter
->index
> index
)
3339 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
3340 if (iter
->index
> index
)
3343 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3344 dev
= get_imsm_dev(super
, i
);
3345 map
= get_imsm_map(dev
, 0);
3346 num_members
= map
->num_members
;
3347 for (j
= 0; j
< num_members
; j
++) {
3348 /* update ord entries being careful not to propagate
3349 * ord-flags to the first map
3351 ord
= get_imsm_ord_tbl_ent(dev
, j
);
3353 if (ord_to_idx(ord
) <= index
)
3356 map
= get_imsm_map(dev
, 0);
3357 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
3358 map
= get_imsm_map(dev
, 1);
3360 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
3365 super
->updates_pending
++;
3367 struct dl
*dl
= *dlp
;
3369 *dlp
= (*dlp
)->next
;
3370 __free_imsm_disk(dl
);
3373 #endif /* MDASSEMBLE */
3375 struct superswitch super_imsm
= {
3377 .examine_super
= examine_super_imsm
,
3378 .brief_examine_super
= brief_examine_super_imsm
,
3379 .detail_super
= detail_super_imsm
,
3380 .brief_detail_super
= brief_detail_super_imsm
,
3381 .write_init_super
= write_init_super_imsm
,
3382 .validate_geometry
= validate_geometry_imsm
,
3383 .add_to_super
= add_to_super_imsm
,
3385 .match_home
= match_home_imsm
,
3386 .uuid_from_super
= uuid_from_super_imsm
,
3387 .getinfo_super
= getinfo_super_imsm
,
3388 .update_super
= update_super_imsm
,
3390 .avail_size
= avail_size_imsm
,
3392 .compare_super
= compare_super_imsm
,
3394 .load_super
= load_super_imsm
,
3395 .init_super
= init_super_imsm
,
3396 .store_super
= store_zero_imsm
,
3397 .free_super
= free_super_imsm
,
3398 .match_metadata_desc
= match_metadata_desc_imsm
,
3399 .container_content
= container_content_imsm
,
3405 .open_new
= imsm_open_new
,
3406 .load_super
= load_super_imsm
,
3407 .set_array_state
= imsm_set_array_state
,
3408 .set_disk
= imsm_set_disk
,
3409 .sync_metadata
= imsm_sync_metadata
,
3410 .activate_spare
= imsm_activate_spare
,
3411 .process_update
= imsm_process_update
,
3412 .prepare_update
= imsm_prepare_update
,
3413 #endif /* MDASSEMBLE */