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 /* the imsm metadata format does not specify any host
667 * identification information. We return -1 since we can never
668 * confirm nor deny whether a given array is "meant" for this
669 * host. We rely on compare_super and the 'family_num' field to
670 * exclude member disks that do not belong, and we rely on
671 * mdadm.conf to specify the arrays that should be assembled.
672 * Auto-assembly may still pick up "foreign" arrays.
678 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
680 /* The uuid returned here is used for:
681 * uuid to put into bitmap file (Create, Grow)
682 * uuid for backup header when saving critical section (Grow)
683 * comparing uuids when re-adding a device into an array
684 * In these cases the uuid required is that of the data-array,
685 * not the device-set.
686 * uuid to recognise same set when adding a missing device back
687 * to an array. This is a uuid for the device-set.
689 * For each of these we can make do with a truncated
690 * or hashed uuid rather than the original, as long as
692 * In each case the uuid required is that of the data-array,
693 * not the device-set.
695 /* imsm does not track uuid's so we synthesis one using sha1 on
696 * - The signature (Which is constant for all imsm array, but no matter)
697 * - the family_num of the container
698 * - the index number of the volume
699 * - the 'serial' number of the volume.
700 * Hopefully these are all constant.
702 struct intel_super
*super
= st
->sb
;
706 struct imsm_dev
*dev
= NULL
;
709 sha1_process_bytes(super
->anchor
->sig
, MAX_SIGNATURE_LENGTH
, &ctx
);
710 sha1_process_bytes(&super
->anchor
->family_num
, sizeof(__u32
), &ctx
);
711 if (super
->current_vol
>= 0)
712 dev
= get_imsm_dev(super
, super
->current_vol
);
714 __u32 vol
= super
->current_vol
;
715 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
716 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
718 sha1_finish_ctx(&ctx
, buf
);
719 memcpy(uuid
, buf
, 4*4);
724 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
726 __u8
*v
= get_imsm_version(mpb
);
727 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
728 char major
[] = { 0, 0, 0 };
729 char minor
[] = { 0 ,0, 0 };
730 char patch
[] = { 0, 0, 0 };
731 char *ver_parse
[] = { major
, minor
, patch
};
735 while (*v
!= '\0' && v
< end
) {
736 if (*v
!= '.' && j
< 2)
737 ver_parse
[i
][j
++] = *v
;
745 *m
= strtol(minor
, NULL
, 0);
746 *p
= strtol(patch
, NULL
, 0);
750 static int imsm_level_to_layout(int level
)
758 return ALGORITHM_LEFT_ASYMMETRIC
;
765 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
767 struct intel_super
*super
= st
->sb
;
768 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
769 struct imsm_map
*map
= get_imsm_map(dev
, 0);
771 info
->container_member
= super
->current_vol
;
772 info
->array
.raid_disks
= map
->num_members
;
773 info
->array
.level
= get_imsm_raid_level(map
);
774 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
775 info
->array
.md_minor
= -1;
776 info
->array
.ctime
= 0;
777 info
->array
.utime
= 0;
778 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
779 info
->array
.state
= !dev
->vol
.dirty
;
781 info
->disk
.major
= 0;
782 info
->disk
.minor
= 0;
784 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
785 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
786 memset(info
->uuid
, 0, sizeof(info
->uuid
));
788 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
789 info
->resync_start
= 0;
790 else if (dev
->vol
.migr_state
)
791 info
->resync_start
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
793 info
->resync_start
= ~0ULL;
795 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
796 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
798 info
->array
.major_version
= -1;
799 info
->array
.minor_version
= -2;
800 sprintf(info
->text_version
, "/%s/%d",
801 devnum2devname(st
->container_dev
),
802 info
->container_member
);
803 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
804 uuid_from_super_imsm(st
, info
->uuid
);
808 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
810 struct intel_super
*super
= st
->sb
;
811 struct imsm_disk
*disk
;
814 if (super
->current_vol
>= 0) {
815 getinfo_super_imsm_volume(st
, info
);
819 /* Set raid_disks to zero so that Assemble will always pull in valid
822 info
->array
.raid_disks
= 0;
823 info
->array
.level
= LEVEL_CONTAINER
;
824 info
->array
.layout
= 0;
825 info
->array
.md_minor
= -1;
826 info
->array
.ctime
= 0; /* N/A for imsm */
827 info
->array
.utime
= 0;
828 info
->array
.chunk_size
= 0;
830 info
->disk
.major
= 0;
831 info
->disk
.minor
= 0;
832 info
->disk
.raid_disk
= -1;
833 info
->reshape_active
= 0;
834 info
->array
.major_version
= -1;
835 info
->array
.minor_version
= -2;
836 strcpy(info
->text_version
, "imsm");
837 info
->safe_mode_delay
= 0;
838 info
->disk
.number
= -1;
839 info
->disk
.state
= 0;
843 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
845 disk
= &super
->disks
->disk
;
846 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
847 info
->component_size
= reserved
;
848 s
= __le32_to_cpu(disk
->status
);
849 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
850 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
851 info
->disk
.state
|= s
& SPARE_DISK
? 0 : (1 << MD_DISK_SYNC
);
854 /* only call uuid_from_super_imsm when this disk is part of a populated container,
855 * ->compare_super may have updated the 'num_raid_devs' field for spares
857 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
858 uuid_from_super_imsm(st
, info
->uuid
);
860 memcpy(info
->uuid
, uuid_match_any
, sizeof(int[4]));
863 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
864 char *update
, char *devname
, int verbose
,
865 int uuid_set
, char *homehost
)
869 /* For 'assemble' and 'force' we need to return non-zero if any
870 * change was made. For others, the return value is ignored.
871 * Update options are:
872 * force-one : This device looks a bit old but needs to be included,
873 * update age info appropriately.
874 * assemble: clear any 'faulty' flag to allow this device to
876 * force-array: Array is degraded but being forced, mark it clean
877 * if that will be needed to assemble it.
879 * newdev: not used ????
880 * grow: Array has gained a new device - this is currently for
882 * resync: mark as dirty so a resync will happen.
883 * name: update the name - preserving the homehost
885 * Following are not relevant for this imsm:
886 * sparc2.2 : update from old dodgey metadata
887 * super-minor: change the preferred_minor number
888 * summaries: update redundant counters.
889 * uuid: Change the uuid of the array to match watch is given
890 * homehost: update the recorded homehost
891 * _reshape_progress: record new reshape_progress position.
894 //struct intel_super *super = st->sb;
895 //struct imsm_super *mpb = super->mpb;
897 if (strcmp(update
, "grow") == 0) {
899 if (strcmp(update
, "resync") == 0) {
900 /* dev->vol.dirty = 1; */
903 /* IMSM has no concept of UUID or homehost */
908 static size_t disks_to_mpb_size(int disks
)
912 size
= sizeof(struct imsm_super
);
913 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
914 size
+= 2 * sizeof(struct imsm_dev
);
915 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
916 size
+= (4 - 2) * sizeof(struct imsm_map
);
917 /* 4 possible disk_ord_tbl's */
918 size
+= 4 * (disks
- 1) * sizeof(__u32
);
923 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
925 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
928 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
931 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
935 * 0 same, or first was empty, and second was copied
936 * 1 second had wrong number
940 struct intel_super
*first
= st
->sb
;
941 struct intel_super
*sec
= tst
->sb
;
949 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
952 /* if an anchor does not have num_raid_devs set then it is a free
955 if (first
->anchor
->num_raid_devs
> 0 &&
956 sec
->anchor
->num_raid_devs
> 0) {
957 if (first
->anchor
->family_num
!= sec
->anchor
->family_num
)
961 /* if 'first' is a spare promote it to a populated mpb with sec's
964 if (first
->anchor
->num_raid_devs
== 0 &&
965 sec
->anchor
->num_raid_devs
> 0) {
968 /* we need to copy raid device info from sec if an allocation
969 * fails here we don't associate the spare
971 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
972 first
->dev_tbl
[i
] = malloc(sizeof(struct imsm_dev
));
973 if (!first
->dev_tbl
) {
975 free(first
->dev_tbl
[i
]);
976 first
->dev_tbl
[i
] = NULL
;
978 fprintf(stderr
, "imsm: failed to associate spare\n");
981 *first
->dev_tbl
[i
] = *sec
->dev_tbl
[i
];
984 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
985 first
->anchor
->family_num
= sec
->anchor
->family_num
;
991 static void fd2devname(int fd
, char *name
)
1000 if (fstat(fd
, &st
) != 0)
1002 sprintf(path
, "/sys/dev/block/%d:%d",
1003 major(st
.st_rdev
), minor(st
.st_rdev
));
1005 rv
= readlink(path
, dname
, sizeof(dname
));
1010 nm
= strrchr(dname
, '/');
1012 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
1016 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
1018 static int imsm_read_serial(int fd
, char *devname
,
1019 __u8 serial
[MAX_RAID_SERIAL_LEN
])
1021 unsigned char scsi_serial
[255];
1027 memset(scsi_serial
, 0, sizeof(scsi_serial
));
1029 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
1031 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
1032 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1033 fd2devname(fd
, (char *) serial
);
1040 Name
": Failed to retrieve serial for %s\n",
1045 /* trim leading whitespace */
1046 rsp_len
= scsi_serial
[3];
1047 rsp_buf
= (char *) &scsi_serial
[4];
1052 /* truncate len to the end of rsp_buf if necessary */
1053 if (c
+ MAX_RAID_SERIAL_LEN
> rsp_buf
+ rsp_len
)
1054 len
= rsp_len
- (c
- rsp_buf
);
1056 len
= MAX_RAID_SERIAL_LEN
;
1058 /* initialize the buffer and copy rsp_buf characters */
1059 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1060 memcpy(serial
, c
, len
);
1062 /* trim trailing whitespace starting with the last character copied */
1063 c
= (char *) &serial
[len
- 1];
1064 while (isspace(*c
) || *c
== '\0')
1070 static int serialcmp(__u8
*s1
, __u8
*s2
)
1072 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
1075 static void serialcpy(__u8
*dest
, __u8
*src
)
1077 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
1081 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
1088 __u8 serial
[MAX_RAID_SERIAL_LEN
];
1090 rv
= imsm_read_serial(fd
, devname
, serial
);
1095 /* check if this is a disk we have seen before. it may be a spare in
1096 * super->disks while the current anchor believes it is a raid member,
1097 * check if we need to update dl->index
1099 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1100 if (serialcmp(dl
->serial
, serial
) == 0)
1104 dl
= malloc(sizeof(*dl
));
1111 Name
": failed to allocate disk buffer for %s\n",
1118 dl
->major
= major(stb
.st_rdev
);
1119 dl
->minor
= minor(stb
.st_rdev
);
1120 dl
->next
= super
->disks
;
1121 dl
->fd
= keep_fd
? fd
: -1;
1122 dl
->devname
= devname
? strdup(devname
) : NULL
;
1123 serialcpy(dl
->serial
, serial
);
1125 } else if (keep_fd
) {
1130 /* look up this disk's index in the current anchor */
1131 for (i
= 0; i
< super
->anchor
->num_disks
; i
++) {
1132 struct imsm_disk
*disk_iter
;
1134 disk_iter
= __get_imsm_disk(super
->anchor
, i
);
1136 if (serialcmp(disk_iter
->serial
, dl
->serial
) == 0) {
1139 dl
->disk
= *disk_iter
;
1140 status
= __le32_to_cpu(dl
->disk
.status
);
1141 /* only set index on disks that are a member of a
1142 * populated contianer, i.e. one with raid_devs
1144 if (status
& FAILED_DISK
)
1146 else if (status
& SPARE_DISK
)
1155 /* no match, maybe a stale failed drive */
1156 if (i
== super
->anchor
->num_disks
&& dl
->index
>= 0) {
1157 dl
->disk
= *__get_imsm_disk(super
->anchor
, dl
->index
);
1158 if (__le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
1168 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
1170 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
1174 /* When migrating map0 contains the 'destination' state while map1
1175 * contains the current state. When not migrating map0 contains the
1176 * current state. This routine assumes that map[0].map_state is set to
1177 * the current array state before being called.
1179 * Migration is indicated by one of the following states
1180 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
1181 * 2/ Initialize (migr_state=1 migr_type=0 map0state=normal
1182 * map1state=unitialized)
1183 * 3/ Verify (Resync) (migr_state=1 migr_type=1 map0state=normal
1185 * 4/ Rebuild (migr_state=1 migr_type=1 map0state=normal
1186 * map1state=degraded)
1188 static void migrate(struct imsm_dev
*dev
, __u8 to_state
, int rebuild_resync
)
1190 struct imsm_map
*dest
;
1191 struct imsm_map
*src
= get_imsm_map(dev
, 0);
1193 dev
->vol
.migr_state
= 1;
1194 dev
->vol
.migr_type
= rebuild_resync
;
1195 dev
->vol
.curr_migr_unit
= 0;
1196 dest
= get_imsm_map(dev
, 1);
1198 memcpy(dest
, src
, sizeof_imsm_map(src
));
1199 src
->map_state
= to_state
;
1202 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
1204 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1206 dev
->vol
.migr_state
= 0;
1207 dev
->vol
.curr_migr_unit
= 0;
1208 map
->map_state
= map_state
;
1212 static int parse_raid_devices(struct intel_super
*super
)
1215 struct imsm_dev
*dev_new
;
1216 size_t len
, len_migr
;
1217 size_t space_needed
= 0;
1218 struct imsm_super
*mpb
= super
->anchor
;
1220 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1221 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
1223 len
= sizeof_imsm_dev(dev_iter
, 0);
1224 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
1226 space_needed
+= len_migr
- len
;
1228 dev_new
= malloc(len_migr
);
1231 imsm_copy_dev(dev_new
, dev_iter
);
1232 super
->dev_tbl
[i
] = dev_new
;
1235 /* ensure that super->buf is large enough when all raid devices
1238 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
1241 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
1242 if (posix_memalign(&buf
, 512, len
) != 0)
1245 memcpy(buf
, super
->buf
, len
);
1254 /* retrieve a pointer to the bbm log which starts after all raid devices */
1255 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
1259 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
1261 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
1267 static void __free_imsm(struct intel_super
*super
, int free_disks
);
1269 /* load_imsm_mpb - read matrix metadata
1270 * allocates super->mpb to be freed by free_super
1272 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
1274 unsigned long long dsize
;
1275 unsigned long long sectors
;
1277 struct imsm_super
*anchor
;
1281 get_dev_size(fd
, NULL
, &dsize
);
1283 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
1286 Name
": Cannot seek to anchor block on %s: %s\n",
1287 devname
, strerror(errno
));
1291 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
1294 Name
": Failed to allocate imsm anchor buffer"
1295 " on %s\n", devname
);
1298 if (read(fd
, anchor
, 512) != 512) {
1301 Name
": Cannot read anchor block on %s: %s\n",
1302 devname
, strerror(errno
));
1307 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
1310 Name
": no IMSM anchor on %s\n", devname
);
1315 __free_imsm(super
, 0);
1316 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
1317 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
1320 Name
": unable to allocate %zu byte mpb buffer\n",
1325 memcpy(super
->buf
, anchor
, 512);
1327 sectors
= mpb_sectors(anchor
) - 1;
1330 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1332 rc
= parse_raid_devices(super
);
1336 /* read the extended mpb */
1337 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
1340 Name
": Cannot seek to extended mpb on %s: %s\n",
1341 devname
, strerror(errno
));
1345 if (read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
1348 Name
": Cannot read extended mpb on %s: %s\n",
1349 devname
, strerror(errno
));
1353 check_sum
= __gen_imsm_checksum(super
->anchor
);
1354 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
1357 Name
": IMSM checksum %x != %x on %s\n",
1358 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
1363 /* FIXME the BBM log is disk specific so we cannot use this global
1364 * buffer for all disks. Ok for now since we only look at the global
1365 * bbm_log_size parameter to gate assembly
1367 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
1369 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1371 rc
= parse_raid_devices(super
);
1376 static void __free_imsm_disk(struct dl
*d
)
1385 static void free_imsm_disks(struct intel_super
*super
)
1389 while (super
->disks
) {
1391 super
->disks
= d
->next
;
1392 __free_imsm_disk(d
);
1394 while (super
->missing
) {
1396 super
->missing
= d
->next
;
1397 __free_imsm_disk(d
);
1402 /* free all the pieces hanging off of a super pointer */
1403 static void __free_imsm(struct intel_super
*super
, int free_disks
)
1412 free_imsm_disks(super
);
1413 for (i
= 0; i
< IMSM_MAX_RAID_DEVS
; i
++)
1414 if (super
->dev_tbl
[i
]) {
1415 free(super
->dev_tbl
[i
]);
1416 super
->dev_tbl
[i
] = NULL
;
1420 static void free_imsm(struct intel_super
*super
)
1422 __free_imsm(super
, 1);
1426 static void free_super_imsm(struct supertype
*st
)
1428 struct intel_super
*super
= st
->sb
;
1437 static struct intel_super
*alloc_super(int creating_imsm
)
1439 struct intel_super
*super
= malloc(sizeof(*super
));
1442 memset(super
, 0, sizeof(*super
));
1443 super
->creating_imsm
= creating_imsm
;
1444 super
->current_vol
= -1;
1451 /* find_missing - helper routine for load_super_imsm_all that identifies
1452 * disks that have disappeared from the system. This routine relies on
1453 * the mpb being uptodate, which it is at load time.
1455 static int find_missing(struct intel_super
*super
)
1458 struct imsm_super
*mpb
= super
->anchor
;
1460 struct imsm_disk
*disk
;
1463 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1464 disk
= __get_imsm_disk(mpb
, i
);
1465 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1466 if (serialcmp(dl
->disk
.serial
, disk
->serial
) == 0)
1470 /* ok we have a 'disk' without a live entry in
1473 status
= __le32_to_cpu(disk
->status
);
1474 if (status
& FAILED_DISK
|| !(status
& USABLE_DISK
))
1475 continue; /* never mind, already marked */
1477 dl
= malloc(sizeof(*dl
));
1483 dl
->devname
= strdup("missing");
1485 serialcpy(dl
->serial
, disk
->serial
);
1487 dl
->next
= super
->missing
;
1488 super
->missing
= dl
;
1494 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
1495 char *devname
, int keep_fd
)
1498 struct intel_super
*super
;
1499 struct mdinfo
*sd
, *best
= NULL
;
1506 /* check if this disk is a member of an active array */
1507 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
1511 if (sra
->array
.major_version
!= -1 ||
1512 sra
->array
.minor_version
!= -2 ||
1513 strcmp(sra
->text_version
, "imsm") != 0)
1516 super
= alloc_super(0);
1520 /* find the most up to date disk in this array, skipping spares */
1521 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1522 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1523 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1528 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1532 if (super
->anchor
->num_raid_devs
== 0)
1535 gen
= __le32_to_cpu(super
->anchor
->generation_num
);
1536 if (!best
|| gen
> bestgen
) {
1551 /* load the most up to date anchor */
1552 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
1553 dfd
= dev_open(nm
, O_RDONLY
);
1558 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1565 /* re-parse the disk list with the current anchor */
1566 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1567 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1568 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1573 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
1579 if (find_missing(super
) != 0) {
1584 if (st
->subarray
[0]) {
1585 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
1586 super
->current_vol
= atoi(st
->subarray
);
1592 st
->container_dev
= fd2devnum(fd
);
1593 if (st
->ss
== NULL
) {
1594 st
->ss
= &super_imsm
;
1595 st
->minor_version
= 0;
1596 st
->max_devs
= IMSM_MAX_DEVICES
;
1598 st
->loaded_container
= 1;
1604 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
1606 struct intel_super
*super
;
1610 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1613 if (st
->subarray
[0])
1614 return 1; /* FIXME */
1616 super
= alloc_super(0);
1619 Name
": malloc of %zu failed.\n",
1624 rv
= load_imsm_mpb(fd
, super
, devname
);
1629 Name
": Failed to load all information "
1630 "sections on %s\n", devname
);
1636 if (st
->ss
== NULL
) {
1637 st
->ss
= &super_imsm
;
1638 st
->minor_version
= 0;
1639 st
->max_devs
= IMSM_MAX_DEVICES
;
1641 st
->loaded_container
= 0;
1646 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1648 if (info
->level
== 1)
1650 return info
->chunk_size
>> 9;
1653 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1657 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1658 if (info
->level
== 1)
1664 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
1666 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
1669 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1670 unsigned long long size
, char *name
,
1671 char *homehost
, int *uuid
)
1673 /* We are creating a volume inside a pre-existing container.
1674 * so st->sb is already set.
1676 struct intel_super
*super
= st
->sb
;
1677 struct imsm_super
*mpb
= super
->anchor
;
1678 struct imsm_dev
*dev
;
1679 struct imsm_vol
*vol
;
1680 struct imsm_map
*map
;
1681 int idx
= mpb
->num_raid_devs
;
1683 unsigned long long array_blocks
;
1685 size_t size_old
, size_new
;
1687 if (mpb
->num_raid_devs
>= 2) {
1688 fprintf(stderr
, Name
": This imsm-container already has the "
1689 "maximum of 2 volumes\n");
1693 /* ensure the mpb is large enough for the new data */
1694 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1695 size_new
= disks_to_mpb_size(info
->nr_disks
);
1696 if (size_new
> size_old
) {
1698 size_t size_round
= ROUND_UP(size_new
, 512);
1700 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1701 fprintf(stderr
, Name
": could not allocate new mpb\n");
1704 memcpy(mpb_new
, mpb
, size_old
);
1707 super
->anchor
= mpb_new
;
1708 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1709 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1711 super
->current_vol
= idx
;
1712 /* when creating the first raid device in this container set num_disks
1713 * to zero, i.e. delete this spare and add raid member devices in
1714 * add_to_super_imsm_volume()
1716 if (super
->current_vol
== 0)
1718 sprintf(st
->subarray
, "%d", idx
);
1719 dev
= malloc(sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
1721 fprintf(stderr
, Name
": could not allocate raid device\n");
1724 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1725 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1726 info
->layout
, info
->chunk_size
,
1728 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1729 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1730 dev
->status
= __cpu_to_le32(0);
1731 dev
->reserved_blocks
= __cpu_to_le32(0);
1733 vol
->migr_state
= 0;
1736 vol
->curr_migr_unit
= 0;
1737 for (i
= 0; i
< idx
; i
++) {
1738 struct imsm_dev
*prev
= get_imsm_dev(super
, i
);
1739 struct imsm_map
*pmap
= get_imsm_map(prev
, 0);
1741 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1742 offset
+= IMSM_RESERVED_SECTORS
;
1744 map
= get_imsm_map(dev
, 0);
1745 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1746 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
1747 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1748 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1749 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1750 IMSM_T_STATE_NORMAL
;
1752 if (info
->level
== 1 && info
->raid_disks
> 2) {
1753 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1754 "in a raid1 volume\n");
1757 if (info
->level
== 10)
1758 map
->raid_level
= 1;
1760 map
->raid_level
= info
->level
;
1762 map
->num_members
= info
->raid_disks
;
1763 for (i
= 0; i
< map
->num_members
; i
++) {
1764 /* initialized in add_to_super */
1765 set_imsm_ord_tbl_ent(map
, i
, 0);
1767 mpb
->num_raid_devs
++;
1768 super
->dev_tbl
[super
->current_vol
] = dev
;
1773 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1774 unsigned long long size
, char *name
,
1775 char *homehost
, int *uuid
)
1777 /* This is primarily called by Create when creating a new array.
1778 * We will then get add_to_super called for each component, and then
1779 * write_init_super called to write it out to each device.
1780 * For IMSM, Create can create on fresh devices or on a pre-existing
1782 * To create on a pre-existing array a different method will be called.
1783 * This one is just for fresh drives.
1785 struct intel_super
*super
;
1786 struct imsm_super
*mpb
;
1794 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1797 super
= alloc_super(1);
1800 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1801 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1806 memset(mpb
, 0, mpb_size
);
1808 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1809 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1810 strlen(MPB_VERSION_RAID5
));
1811 mpb
->mpb_size
= mpb_size
;
1818 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1819 int fd
, char *devname
)
1821 struct intel_super
*super
= st
->sb
;
1822 struct imsm_super
*mpb
= super
->anchor
;
1824 struct imsm_dev
*dev
;
1825 struct imsm_map
*map
;
1828 dev
= get_imsm_dev(super
, super
->current_vol
);
1829 map
= get_imsm_map(dev
, 0);
1831 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1832 if (dl
->major
== dk
->major
&&
1833 dl
->minor
== dk
->minor
)
1836 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1839 /* add a pristine spare to the metadata */
1840 if (dl
->index
< 0) {
1841 dl
->index
= super
->anchor
->num_disks
;
1842 super
->anchor
->num_disks
++;
1844 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
1845 status
= CONFIGURED_DISK
| USABLE_DISK
;
1846 dl
->disk
.status
= __cpu_to_le32(status
);
1848 /* if we are creating the first raid device update the family number */
1849 if (super
->current_vol
== 0) {
1851 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
1852 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
1856 sum
= __gen_imsm_checksum(mpb
);
1857 mpb
->family_num
= __cpu_to_le32(sum
);
1861 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1862 int fd
, char *devname
)
1864 struct intel_super
*super
= st
->sb
;
1866 unsigned long long size
;
1871 if (super
->current_vol
>= 0) {
1872 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1877 dd
= malloc(sizeof(*dd
));
1880 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1883 memset(dd
, 0, sizeof(*dd
));
1884 dd
->major
= major(stb
.st_rdev
);
1885 dd
->minor
= minor(stb
.st_rdev
);
1887 dd
->devname
= devname
? strdup(devname
) : NULL
;
1889 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1892 Name
": failed to retrieve scsi serial, aborting\n");
1897 get_dev_size(fd
, NULL
, &size
);
1899 status
= USABLE_DISK
| SPARE_DISK
;
1900 serialcpy(dd
->disk
.serial
, dd
->serial
);
1901 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
1902 dd
->disk
.status
= __cpu_to_le32(status
);
1903 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1904 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
1906 dd
->disk
.scsi_id
= __cpu_to_le32(0);
1908 if (st
->update_tail
) {
1909 dd
->next
= super
->add
;
1912 dd
->next
= super
->disks
;
1917 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1919 /* spare records have their own family number and do not have any defined raid
1922 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
1924 struct imsm_super mpb_save
;
1925 struct imsm_super
*mpb
= super
->anchor
;
1930 mpb
->num_raid_devs
= 0;
1932 mpb
->mpb_size
= sizeof(struct imsm_super
);
1933 mpb
->generation_num
= __cpu_to_le32(1UL);
1935 for (d
= super
->disks
; d
; d
= d
->next
) {
1939 mpb
->disk
[0] = d
->disk
;
1940 sum
= __gen_imsm_checksum(mpb
);
1941 mpb
->family_num
= __cpu_to_le32(sum
);
1942 sum
= __gen_imsm_checksum(mpb
);
1943 mpb
->check_sum
= __cpu_to_le32(sum
);
1945 if (store_imsm_mpb(d
->fd
, super
)) {
1946 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1947 __func__
, d
->major
, d
->minor
, strerror(errno
));
1961 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1963 struct imsm_super
*mpb
= super
->anchor
;
1969 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
1971 /* 'generation' is incremented everytime the metadata is written */
1972 generation
= __le32_to_cpu(mpb
->generation_num
);
1974 mpb
->generation_num
= __cpu_to_le32(generation
);
1976 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
1977 for (d
= super
->disks
; d
; d
= d
->next
) {
1981 mpb
->disk
[d
->index
] = d
->disk
;
1983 for (d
= super
->missing
; d
; d
= d
->next
)
1984 mpb
->disk
[d
->index
] = d
->disk
;
1986 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1987 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1989 imsm_copy_dev(dev
, super
->dev_tbl
[i
]);
1990 mpb_size
+= sizeof_imsm_dev(dev
, 0);
1992 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
1993 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
1995 /* recalculate checksum */
1996 sum
= __gen_imsm_checksum(mpb
);
1997 mpb
->check_sum
= __cpu_to_le32(sum
);
1999 /* write the mpb for disks that compose raid devices */
2000 for (d
= super
->disks
; d
; d
= d
->next
) {
2003 if (store_imsm_mpb(d
->fd
, super
))
2004 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
2005 __func__
, d
->major
, d
->minor
, strerror(errno
));
2013 return write_super_imsm_spares(super
, doclose
);
2019 static int create_array(struct supertype
*st
)
2022 struct imsm_update_create_array
*u
;
2023 struct intel_super
*super
= st
->sb
;
2024 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
2026 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0);
2029 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2034 u
->type
= update_create_array
;
2035 u
->dev_idx
= super
->current_vol
;
2036 imsm_copy_dev(&u
->dev
, dev
);
2037 append_metadata_update(st
, u
, len
);
2042 static int _add_disk(struct supertype
*st
)
2044 struct intel_super
*super
= st
->sb
;
2046 struct imsm_update_add_disk
*u
;
2054 fprintf(stderr
, "%s: failed to allocate update buffer\n",
2059 u
->type
= update_add_disk
;
2060 append_metadata_update(st
, u
, len
);
2065 static int write_init_super_imsm(struct supertype
*st
)
2067 if (st
->update_tail
) {
2068 /* queue the recently created array / added disk
2069 * as a metadata update */
2070 struct intel_super
*super
= st
->sb
;
2074 /* determine if we are creating a volume or adding a disk */
2075 if (super
->current_vol
< 0) {
2076 /* in the add disk case we are running in mdmon
2077 * context, so don't close fd's
2079 return _add_disk(st
);
2081 rv
= create_array(st
);
2083 for (d
= super
->disks
; d
; d
= d
->next
) {
2090 return write_super_imsm(st
->sb
, 1);
2094 static int store_zero_imsm(struct supertype
*st
, int fd
)
2096 unsigned long long dsize
;
2099 get_dev_size(fd
, NULL
, &dsize
);
2101 /* first block is stored on second to last sector of the disk */
2102 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2105 if (posix_memalign(&buf
, 512, 512) != 0)
2108 memset(buf
, 0, 512);
2109 if (write(fd
, buf
, 512) != 512)
2114 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
2116 return __le32_to_cpu(mpb
->bbm_log_size
);
2120 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
2121 int layout
, int raiddisks
, int chunk
,
2122 unsigned long long size
, char *dev
,
2123 unsigned long long *freesize
,
2127 unsigned long long ldsize
;
2129 if (level
!= LEVEL_CONTAINER
)
2134 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2137 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
2138 dev
, strerror(errno
));
2141 if (!get_dev_size(fd
, dev
, &ldsize
)) {
2147 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
2152 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
2153 * FIX ME add ahci details
2155 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
2156 int layout
, int raiddisks
, int chunk
,
2157 unsigned long long size
, char *dev
,
2158 unsigned long long *freesize
,
2162 struct intel_super
*super
= st
->sb
;
2164 unsigned long long pos
= 0;
2165 unsigned long long maxsize
;
2169 if (level
== LEVEL_CONTAINER
)
2172 if (level
== 1 && raiddisks
> 2) {
2174 fprintf(stderr
, Name
": imsm does not support more "
2175 "than 2 in a raid1 configuration\n");
2179 /* We must have the container info already read in. */
2184 /* General test: make sure there is space for
2185 * 'raiddisks' device extents of size 'size' at a given
2188 unsigned long long minsize
= size
*2 /* convert to blocks */;
2189 unsigned long long start_offset
= ~0ULL;
2192 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
2193 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2198 e
= get_extents(super
, dl
);
2201 unsigned long long esize
;
2202 esize
= e
[i
].start
- pos
;
2203 if (esize
>= minsize
)
2205 if (found
&& start_offset
== ~0ULL) {
2208 } else if (found
&& pos
!= start_offset
) {
2212 pos
= e
[i
].start
+ e
[i
].size
;
2214 } while (e
[i
-1].size
);
2219 if (dcnt
< raiddisks
) {
2221 fprintf(stderr
, Name
": imsm: Not enough "
2222 "devices with space for this array "
2229 /* This device must be a member of the set */
2230 if (stat(dev
, &stb
) < 0)
2232 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
2234 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2235 if (dl
->major
== major(stb
.st_rdev
) &&
2236 dl
->minor
== minor(stb
.st_rdev
))
2241 fprintf(stderr
, Name
": %s is not in the "
2242 "same imsm set\n", dev
);
2245 e
= get_extents(super
, dl
);
2249 unsigned long long esize
;
2250 esize
= e
[i
].start
- pos
;
2251 if (esize
>= maxsize
)
2253 pos
= e
[i
].start
+ e
[i
].size
;
2255 } while (e
[i
-1].size
);
2256 *freesize
= maxsize
;
2261 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
2262 int raiddisks
, int chunk
, unsigned long long size
,
2263 char *dev
, unsigned long long *freesize
,
2269 /* if given unused devices create a container
2270 * if given given devices in a container create a member volume
2272 if (level
== LEVEL_CONTAINER
) {
2273 /* Must be a fresh device to add to a container */
2274 return validate_geometry_imsm_container(st
, level
, layout
,
2275 raiddisks
, chunk
, size
,
2281 /* creating in a given container */
2282 return validate_geometry_imsm_volume(st
, level
, layout
,
2283 raiddisks
, chunk
, size
,
2284 dev
, freesize
, verbose
);
2287 /* limit creation to the following levels */
2299 /* This device needs to be a device in an 'imsm' container */
2300 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
2304 Name
": Cannot create this array on device %s\n",
2309 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
2311 fprintf(stderr
, Name
": Cannot open %s: %s\n",
2312 dev
, strerror(errno
));
2315 /* Well, it is in use by someone, maybe an 'imsm' container. */
2316 cfd
= open_container(fd
);
2320 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
2324 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
2326 if (sra
&& sra
->array
.major_version
== -1 &&
2327 strcmp(sra
->text_version
, "imsm") == 0) {
2328 /* This is a member of a imsm container. Load the container
2329 * and try to create a volume
2331 struct intel_super
*super
;
2333 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
2335 st
->container_dev
= fd2devnum(cfd
);
2337 return validate_geometry_imsm_volume(st
, level
, layout
,
2343 } else /* may belong to another container */
2348 #endif /* MDASSEMBLE */
2350 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
2352 /* Given a container loaded by load_super_imsm_all,
2353 * extract information about all the arrays into
2356 * For each imsm_dev create an mdinfo, fill it in,
2357 * then look for matching devices in super->disks
2358 * and create appropriate device mdinfo.
2360 struct intel_super
*super
= st
->sb
;
2361 struct imsm_super
*mpb
= super
->anchor
;
2362 struct mdinfo
*rest
= NULL
;
2365 /* do not assemble arrays that might have bad blocks */
2366 if (imsm_bbm_log_size(super
->anchor
)) {
2367 fprintf(stderr
, Name
": BBM log found in metadata. "
2368 "Cannot activate array(s).\n");
2372 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2373 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2374 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2375 struct mdinfo
*this;
2378 this = malloc(sizeof(*this));
2379 memset(this, 0, sizeof(*this));
2382 super
->current_vol
= i
;
2383 getinfo_super_imsm_volume(st
, this);
2384 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
2385 struct mdinfo
*info_d
;
2393 idx
= get_imsm_disk_idx(dev
, slot
);
2394 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
2395 for (d
= super
->disks
; d
; d
= d
->next
)
2396 if (d
->index
== idx
)
2402 s
= d
? __le32_to_cpu(d
->disk
.status
) : 0;
2403 if (s
& FAILED_DISK
)
2405 if (!(s
& USABLE_DISK
))
2407 if (ord
& IMSM_ORD_REBUILD
)
2411 * if we skip some disks the array will be assmebled degraded;
2412 * reset resync start to avoid a dirty-degraded situation
2414 * FIXME handle dirty degraded
2416 if (skip
&& !dev
->vol
.dirty
)
2417 this->resync_start
= ~0ULL;
2421 info_d
= malloc(sizeof(*info_d
));
2423 fprintf(stderr
, Name
": failed to allocate disk"
2424 " for volume %s\n", (char *) dev
->volume
);
2429 memset(info_d
, 0, sizeof(*info_d
));
2430 info_d
->next
= this->devs
;
2431 this->devs
= info_d
;
2433 info_d
->disk
.number
= d
->index
;
2434 info_d
->disk
.major
= d
->major
;
2435 info_d
->disk
.minor
= d
->minor
;
2436 info_d
->disk
.raid_disk
= slot
;
2438 this->array
.working_disks
++;
2440 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
2441 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2442 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
2444 strcpy(info_d
->name
, d
->devname
);
2454 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
2457 struct intel_super
*super
= c
->sb
;
2458 struct imsm_super
*mpb
= super
->anchor
;
2460 if (atoi(inst
) >= mpb
->num_raid_devs
) {
2461 fprintf(stderr
, "%s: subarry index %d, out of range\n",
2462 __func__
, atoi(inst
));
2466 dprintf("imsm: open_new %s\n", inst
);
2467 a
->info
.container_member
= atoi(inst
);
2471 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
2473 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2476 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
2477 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
2479 switch (get_imsm_raid_level(map
)) {
2481 return IMSM_T_STATE_FAILED
;
2484 if (failed
< map
->num_members
)
2485 return IMSM_T_STATE_DEGRADED
;
2487 return IMSM_T_STATE_FAILED
;
2492 * check to see if any mirrors have failed, otherwise we
2493 * are degraded. Even numbered slots are mirrored on
2497 /* gcc -Os complains that this is unused */
2498 int insync
= insync
;
2500 for (i
= 0; i
< map
->num_members
; i
++) {
2501 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2502 int idx
= ord_to_idx(ord
);
2503 struct imsm_disk
*disk
;
2505 /* reset the potential in-sync count on even-numbered
2506 * slots. num_copies is always 2 for imsm raid10
2511 disk
= get_imsm_disk(super
, idx
);
2513 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2514 ord
& IMSM_ORD_REBUILD
)
2517 /* no in-sync disks left in this mirror the
2521 return IMSM_T_STATE_FAILED
;
2524 return IMSM_T_STATE_DEGRADED
;
2528 return IMSM_T_STATE_DEGRADED
;
2530 return IMSM_T_STATE_FAILED
;
2536 return map
->map_state
;
2539 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
2543 struct imsm_disk
*disk
;
2544 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2546 for (i
= 0; i
< map
->num_members
; i
++) {
2547 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
2548 int idx
= ord_to_idx(ord
);
2550 disk
= get_imsm_disk(super
, idx
);
2552 __le32_to_cpu(disk
->status
) & FAILED_DISK
||
2553 ord
& IMSM_ORD_REBUILD
)
2560 static int is_resyncing(struct imsm_dev
*dev
)
2562 struct imsm_map
*migr_map
;
2564 if (!dev
->vol
.migr_state
)
2567 if (dev
->vol
.migr_type
== 0)
2570 migr_map
= get_imsm_map(dev
, 1);
2572 if (migr_map
->map_state
== IMSM_T_STATE_NORMAL
)
2578 static int is_rebuilding(struct imsm_dev
*dev
)
2580 struct imsm_map
*migr_map
;
2582 if (!dev
->vol
.migr_state
)
2585 if (dev
->vol
.migr_type
== 0)
2588 migr_map
= get_imsm_map(dev
, 1);
2590 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
2596 static void mark_failure(struct imsm_disk
*disk
)
2598 __u32 status
= __le32_to_cpu(disk
->status
);
2600 if (status
& FAILED_DISK
)
2602 status
|= FAILED_DISK
;
2603 disk
->status
= __cpu_to_le32(status
);
2604 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
2605 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
2608 /* Handle dirty -> clean transititions and resync. Degraded and rebuild
2609 * states are handled in imsm_set_disk() with one exception, when a
2610 * resync is stopped due to a new failure this routine will set the
2611 * 'degraded' state for the array.
2613 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
2615 int inst
= a
->info
.container_member
;
2616 struct intel_super
*super
= a
->container
->sb
;
2617 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2618 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2619 int failed
= imsm_count_failed(super
, dev
);
2620 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
2622 /* before we activate this array handle any missing disks */
2623 if (consistent
== 2 && super
->missing
) {
2626 dprintf("imsm: mark missing\n");
2627 end_migration(dev
, map_state
);
2628 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
2629 mark_failure(&dl
->disk
);
2630 super
->updates_pending
++;
2633 if (consistent
== 2 &&
2634 (!is_resync_complete(a
) ||
2635 map_state
!= IMSM_T_STATE_NORMAL
||
2636 dev
->vol
.migr_state
))
2639 if (is_resync_complete(a
)) {
2640 /* complete intialization / resync,
2641 * recovery is completed in ->set_disk
2643 if (is_resyncing(dev
)) {
2644 dprintf("imsm: mark resync done\n");
2645 end_migration(dev
, map_state
);
2646 super
->updates_pending
++;
2648 } else if (!is_resyncing(dev
) && !failed
) {
2649 /* mark the start of the init process if nothing is failed */
2650 dprintf("imsm: mark resync start (%llu)\n", a
->resync_start
);
2651 map
->map_state
= map_state
;
2652 migrate(dev
, IMSM_T_STATE_NORMAL
,
2653 map
->map_state
== IMSM_T_STATE_NORMAL
);
2654 super
->updates_pending
++;
2657 /* check if we can update the migration checkpoint */
2658 if (dev
->vol
.migr_state
&&
2659 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != a
->resync_start
) {
2660 dprintf("imsm: checkpoint migration (%llu)\n", a
->resync_start
);
2661 dev
->vol
.curr_migr_unit
= __cpu_to_le32(a
->resync_start
);
2662 super
->updates_pending
++;
2665 /* mark dirty / clean */
2666 if (dev
->vol
.dirty
!= !consistent
) {
2667 dprintf("imsm: mark '%s' (%llu)\n",
2668 consistent
? "clean" : "dirty", a
->resync_start
);
2673 super
->updates_pending
++;
2678 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
2680 int inst
= a
->info
.container_member
;
2681 struct intel_super
*super
= a
->container
->sb
;
2682 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2683 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2684 struct imsm_disk
*disk
;
2690 if (n
> map
->num_members
)
2691 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
2692 n
, map
->num_members
- 1);
2697 dprintf("imsm: set_disk %d:%x\n", n
, state
);
2699 ord
= get_imsm_ord_tbl_ent(dev
, n
);
2700 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
2702 /* check for new failures */
2703 status
= __le32_to_cpu(disk
->status
);
2704 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
2706 super
->updates_pending
++;
2709 /* check if in_sync */
2710 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
) {
2711 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
2713 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
2714 super
->updates_pending
++;
2717 failed
= imsm_count_failed(super
, dev
);
2718 map_state
= imsm_check_degraded(super
, dev
, failed
);
2720 /* check if recovery complete, newly degraded, or failed */
2721 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
2722 end_migration(dev
, map_state
);
2723 super
->updates_pending
++;
2724 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
2725 map
->map_state
!= map_state
&&
2726 !dev
->vol
.migr_state
) {
2727 dprintf("imsm: mark degraded\n");
2728 map
->map_state
= map_state
;
2729 super
->updates_pending
++;
2730 } else if (map_state
== IMSM_T_STATE_FAILED
&&
2731 map
->map_state
!= map_state
) {
2732 dprintf("imsm: mark failed\n");
2733 end_migration(dev
, map_state
);
2734 super
->updates_pending
++;
2738 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
2740 struct imsm_super
*mpb
= super
->anchor
;
2741 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
2742 unsigned long long dsize
;
2743 unsigned long long sectors
;
2745 get_dev_size(fd
, NULL
, &dsize
);
2747 if (mpb_size
> 512) {
2748 /* -1 to account for anchor */
2749 sectors
= mpb_sectors(mpb
) - 1;
2751 /* write the extended mpb to the sectors preceeding the anchor */
2752 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
2755 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
2759 /* first block is stored on second to last sector of the disk */
2760 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2763 if (write(fd
, super
->buf
, 512) != 512)
2769 static void imsm_sync_metadata(struct supertype
*container
)
2771 struct intel_super
*super
= container
->sb
;
2773 if (!super
->updates_pending
)
2776 write_super_imsm(super
, 0);
2778 super
->updates_pending
= 0;
2781 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
2783 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2784 int i
= get_imsm_disk_idx(dev
, idx
);
2787 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2791 if (dl
&& __le32_to_cpu(dl
->disk
.status
) & FAILED_DISK
)
2795 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
2800 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
, struct active_array
*a
)
2802 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
2803 int idx
= get_imsm_disk_idx(dev
, slot
);
2804 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2805 unsigned long long esize
;
2806 unsigned long long pos
;
2815 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2816 /* If in this array, skip */
2817 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2818 if (d
->state_fd
>= 0 &&
2819 d
->disk
.major
== dl
->major
&&
2820 d
->disk
.minor
== dl
->minor
) {
2821 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
2827 /* skip in use or failed drives */
2828 status
= __le32_to_cpu(dl
->disk
.status
);
2829 if (status
& FAILED_DISK
|| idx
== dl
->index
) {
2830 dprintf("%x:%x status ( %s%s)\n",
2831 dl
->major
, dl
->minor
,
2832 status
& FAILED_DISK
? "failed " : "",
2833 idx
== dl
->index
? "in use " : "");
2837 /* Does this unused device have the requisite free space?
2838 * We need a->info.component_size sectors
2840 ex
= get_extents(super
, dl
);
2842 dprintf("cannot get extents\n");
2848 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
2851 /* check that we can start at pba_of_lba0 with
2852 * a->info.component_size of space
2854 esize
= ex
[j
].start
- pos
;
2855 if (array_start
>= pos
&&
2856 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
2860 pos
= ex
[j
].start
+ ex
[j
].size
;
2863 } while (ex
[j
-1].size
);
2867 dprintf("%x:%x does not have %llu at %d\n",
2868 dl
->major
, dl
->minor
,
2869 a
->info
.component_size
,
2870 __le32_to_cpu(map
->pba_of_lba0
));
2880 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
2881 struct metadata_update
**updates
)
2884 * Find a device with unused free space and use it to replace a
2885 * failed/vacant region in an array. We replace failed regions one a
2886 * array at a time. The result is that a new spare disk will be added
2887 * to the first failed array and after the monitor has finished
2888 * propagating failures the remainder will be consumed.
2890 * FIXME add a capability for mdmon to request spares from another
2894 struct intel_super
*super
= a
->container
->sb
;
2895 int inst
= a
->info
.container_member
;
2896 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2897 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2898 int failed
= a
->info
.array
.raid_disks
;
2899 struct mdinfo
*rv
= NULL
;
2902 struct metadata_update
*mu
;
2904 struct imsm_update_activate_spare
*u
;
2908 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
2909 if ((d
->curr_state
& DS_FAULTY
) &&
2911 /* wait for Removal to happen */
2913 if (d
->state_fd
>= 0)
2917 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
2918 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
2919 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
2922 /* For each slot, if it is not working, find a spare */
2923 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
2924 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2925 if (d
->disk
.raid_disk
== i
)
2927 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
2928 if (d
&& (d
->state_fd
>= 0))
2932 * OK, this device needs recovery. Try to re-add the previous
2933 * occupant of this slot, if this fails add a new spare
2935 dl
= imsm_readd(super
, i
, a
);
2937 dl
= imsm_add_spare(super
, i
, a
);
2941 /* found a usable disk with enough space */
2942 di
= malloc(sizeof(*di
));
2945 memset(di
, 0, sizeof(*di
));
2947 /* dl->index will be -1 in the case we are activating a
2948 * pristine spare. imsm_process_update() will create a
2949 * new index in this case. Once a disk is found to be
2950 * failed in all member arrays it is kicked from the
2953 di
->disk
.number
= dl
->index
;
2955 /* (ab)use di->devs to store a pointer to the device
2958 di
->devs
= (struct mdinfo
*) dl
;
2960 di
->disk
.raid_disk
= i
;
2961 di
->disk
.major
= dl
->major
;
2962 di
->disk
.minor
= dl
->minor
;
2964 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
2965 di
->component_size
= a
->info
.component_size
;
2966 di
->container_member
= inst
;
2970 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2971 i
, di
->data_offset
);
2977 /* No spares found */
2979 /* Now 'rv' has a list of devices to return.
2980 * Create a metadata_update record to update the
2981 * disk_ord_tbl for the array
2983 mu
= malloc(sizeof(*mu
));
2985 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2986 if (mu
->buf
== NULL
) {
2993 struct mdinfo
*n
= rv
->next
;
3002 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
3003 mu
->next
= *updates
;
3004 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
3006 for (di
= rv
; di
; di
= di
->next
) {
3007 u
->type
= update_activate_spare
;
3008 u
->dl
= (struct dl
*) di
->devs
;
3010 u
->slot
= di
->disk
.raid_disk
;
3021 static int disks_overlap(struct imsm_dev
*d1
, struct imsm_dev
*d2
)
3023 struct imsm_map
*m1
= get_imsm_map(d1
, 0);
3024 struct imsm_map
*m2
= get_imsm_map(d2
, 0);
3029 for (i
= 0; i
< m1
->num_members
; i
++) {
3030 idx
= get_imsm_disk_idx(d1
, i
);
3031 for (j
= 0; j
< m2
->num_members
; j
++)
3032 if (idx
== get_imsm_disk_idx(d2
, j
))
3039 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
);
3041 static void imsm_process_update(struct supertype
*st
,
3042 struct metadata_update
*update
)
3045 * crack open the metadata_update envelope to find the update record
3046 * update can be one of:
3047 * update_activate_spare - a spare device has replaced a failed
3048 * device in an array, update the disk_ord_tbl. If this disk is
3049 * present in all member arrays then also clear the SPARE_DISK
3052 struct intel_super
*super
= st
->sb
;
3053 struct imsm_super
*mpb
;
3054 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3056 /* update requires a larger buf but the allocation failed */
3057 if (super
->next_len
&& !super
->next_buf
) {
3058 super
->next_len
= 0;
3062 if (super
->next_buf
) {
3063 memcpy(super
->next_buf
, super
->buf
, super
->len
);
3065 super
->len
= super
->next_len
;
3066 super
->buf
= super
->next_buf
;
3068 super
->next_len
= 0;
3069 super
->next_buf
= NULL
;
3072 mpb
= super
->anchor
;
3075 case update_activate_spare
: {
3076 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
3077 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
3078 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3079 struct imsm_map
*migr_map
;
3080 struct active_array
*a
;
3081 struct imsm_disk
*disk
;
3087 int victim
= get_imsm_disk_idx(dev
, u
->slot
);
3090 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
3095 fprintf(stderr
, "error: imsm_activate_spare passed "
3096 "an unknown disk (index: %d)\n",
3101 super
->updates_pending
++;
3103 /* count failures (excluding rebuilds and the victim)
3104 * to determine map[0] state
3107 for (i
= 0; i
< map
->num_members
; i
++) {
3110 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3112 __le32_to_cpu(disk
->status
) & FAILED_DISK
)
3116 /* adding a pristine spare, assign a new index */
3117 if (dl
->index
< 0) {
3118 dl
->index
= super
->anchor
->num_disks
;
3119 super
->anchor
->num_disks
++;
3122 status
= __le32_to_cpu(disk
->status
);
3123 status
|= CONFIGURED_DISK
;
3124 status
&= ~SPARE_DISK
;
3125 disk
->status
= __cpu_to_le32(status
);
3128 to_state
= imsm_check_degraded(super
, dev
, failed
);
3129 map
->map_state
= IMSM_T_STATE_DEGRADED
;
3130 migrate(dev
, to_state
, 1);
3131 migr_map
= get_imsm_map(dev
, 1);
3132 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
3133 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
3135 /* count arrays using the victim in the metadata */
3137 for (a
= st
->arrays
; a
; a
= a
->next
) {
3138 dev
= get_imsm_dev(super
, a
->info
.container_member
);
3139 for (i
= 0; i
< map
->num_members
; i
++)
3140 if (victim
== get_imsm_disk_idx(dev
, i
))
3144 /* delete the victim if it is no longer being
3150 /* We know that 'manager' isn't touching anything,
3151 * so it is safe to delete
3153 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
3154 if ((*dlp
)->index
== victim
)
3157 /* victim may be on the missing list */
3159 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
3160 if ((*dlp
)->index
== victim
)
3162 imsm_delete(super
, dlp
, victim
);
3166 case update_create_array
: {
3167 /* someone wants to create a new array, we need to be aware of
3168 * a few races/collisions:
3169 * 1/ 'Create' called by two separate instances of mdadm
3170 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
3171 * devices that have since been assimilated via
3173 * In the event this update can not be carried out mdadm will
3174 * (FIX ME) notice that its update did not take hold.
3176 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3177 struct imsm_dev
*dev
;
3178 struct imsm_map
*map
, *new_map
;
3179 unsigned long long start
, end
;
3180 unsigned long long new_start
, new_end
;
3184 /* handle racing creates: first come first serve */
3185 if (u
->dev_idx
< mpb
->num_raid_devs
) {
3186 dprintf("%s: subarray %d already defined\n",
3187 __func__
, u
->dev_idx
);
3191 /* check update is next in sequence */
3192 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
3193 dprintf("%s: can not create array %d expected index %d\n",
3194 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
3198 new_map
= get_imsm_map(&u
->dev
, 0);
3199 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
3200 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
3202 /* handle activate_spare versus create race:
3203 * check to make sure that overlapping arrays do not include
3206 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3207 dev
= get_imsm_dev(super
, i
);
3208 map
= get_imsm_map(dev
, 0);
3209 start
= __le32_to_cpu(map
->pba_of_lba0
);
3210 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
3211 if ((new_start
>= start
&& new_start
<= end
) ||
3212 (start
>= new_start
&& start
<= new_end
))
3214 if (overlap
&& disks_overlap(dev
, &u
->dev
)) {
3215 dprintf("%s: arrays overlap\n", __func__
);
3219 /* check num_members sanity */
3220 if (new_map
->num_members
> mpb
->num_disks
) {
3221 dprintf("%s: num_disks out of range\n", __func__
);
3225 /* check that prepare update was successful */
3226 if (!update
->space
) {
3227 dprintf("%s: prepare update failed\n", __func__
);
3231 super
->updates_pending
++;
3232 dev
= update
->space
;
3233 map
= get_imsm_map(dev
, 0);
3234 update
->space
= NULL
;
3235 imsm_copy_dev(dev
, &u
->dev
);
3236 map
= get_imsm_map(dev
, 0);
3237 super
->dev_tbl
[u
->dev_idx
] = dev
;
3238 mpb
->num_raid_devs
++;
3241 for (i
= 0; i
< map
->num_members
; i
++) {
3242 struct imsm_disk
*disk
;
3245 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
3246 status
= __le32_to_cpu(disk
->status
);
3247 status
|= CONFIGURED_DISK
;
3248 status
&= ~SPARE_DISK
;
3249 disk
->status
= __cpu_to_le32(status
);
3253 case update_add_disk
:
3255 /* we may be able to repair some arrays if disks are
3258 struct active_array
*a
;
3260 super
->updates_pending
++;
3261 for (a
= st
->arrays
; a
; a
= a
->next
)
3262 a
->check_degraded
= 1;
3264 /* add some spares to the metadata */
3265 while (super
->add
) {
3269 super
->add
= al
->next
;
3270 al
->next
= super
->disks
;
3272 dprintf("%s: added %x:%x\n",
3273 __func__
, al
->major
, al
->minor
);
3280 static void imsm_prepare_update(struct supertype
*st
,
3281 struct metadata_update
*update
)
3284 * Allocate space to hold new disk entries, raid-device entries or a new
3285 * mpb if necessary. The manager synchronously waits for updates to
3286 * complete in the monitor, so new mpb buffers allocated here can be
3287 * integrated by the monitor thread without worrying about live pointers
3288 * in the manager thread.
3290 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
3291 struct intel_super
*super
= st
->sb
;
3292 struct imsm_super
*mpb
= super
->anchor
;
3297 case update_create_array
: {
3298 struct imsm_update_create_array
*u
= (void *) update
->buf
;
3300 len
= sizeof_imsm_dev(&u
->dev
, 1);
3301 update
->space
= malloc(len
);
3308 /* check if we need a larger metadata buffer */
3309 if (super
->next_buf
)
3310 buf_len
= super
->next_len
;
3312 buf_len
= super
->len
;
3314 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
3315 /* ok we need a larger buf than what is currently allocated
3316 * if this allocation fails process_update will notice that
3317 * ->next_len is set and ->next_buf is NULL
3319 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
3320 if (super
->next_buf
)
3321 free(super
->next_buf
);
3323 super
->next_len
= buf_len
;
3324 if (posix_memalign(&super
->next_buf
, buf_len
, 512) != 0)
3325 super
->next_buf
= NULL
;
3329 /* must be called while manager is quiesced */
3330 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
)
3332 struct imsm_super
*mpb
= super
->anchor
;
3334 struct imsm_dev
*dev
;
3335 struct imsm_map
*map
;
3336 int i
, j
, num_members
;
3339 dprintf("%s: deleting device[%d] from imsm_super\n",
3342 /* shift all indexes down one */
3343 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
3344 if (iter
->index
> index
)
3346 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
3347 if (iter
->index
> index
)
3350 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3351 dev
= get_imsm_dev(super
, i
);
3352 map
= get_imsm_map(dev
, 0);
3353 num_members
= map
->num_members
;
3354 for (j
= 0; j
< num_members
; j
++) {
3355 /* update ord entries being careful not to propagate
3356 * ord-flags to the first map
3358 ord
= get_imsm_ord_tbl_ent(dev
, j
);
3360 if (ord_to_idx(ord
) <= index
)
3363 map
= get_imsm_map(dev
, 0);
3364 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
3365 map
= get_imsm_map(dev
, 1);
3367 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
3372 super
->updates_pending
++;
3374 struct dl
*dl
= *dlp
;
3376 *dlp
= (*dlp
)->next
;
3377 __free_imsm_disk(dl
);
3380 #endif /* MDASSEMBLE */
3382 struct superswitch super_imsm
= {
3384 .examine_super
= examine_super_imsm
,
3385 .brief_examine_super
= brief_examine_super_imsm
,
3386 .detail_super
= detail_super_imsm
,
3387 .brief_detail_super
= brief_detail_super_imsm
,
3388 .write_init_super
= write_init_super_imsm
,
3389 .validate_geometry
= validate_geometry_imsm
,
3390 .add_to_super
= add_to_super_imsm
,
3392 .match_home
= match_home_imsm
,
3393 .uuid_from_super
= uuid_from_super_imsm
,
3394 .getinfo_super
= getinfo_super_imsm
,
3395 .update_super
= update_super_imsm
,
3397 .avail_size
= avail_size_imsm
,
3399 .compare_super
= compare_super_imsm
,
3401 .load_super
= load_super_imsm
,
3402 .init_super
= init_super_imsm
,
3403 .store_super
= store_zero_imsm
,
3404 .free_super
= free_super_imsm
,
3405 .match_metadata_desc
= match_metadata_desc_imsm
,
3406 .container_content
= container_content_imsm
,
3412 .open_new
= imsm_open_new
,
3413 .load_super
= load_super_imsm
,
3414 .set_array_state
= imsm_set_array_state
,
3415 .set_disk
= imsm_set_disk
,
3416 .sync_metadata
= imsm_sync_metadata
,
3417 .activate_spare
= imsm_activate_spare
,
3418 .process_update
= imsm_process_update
,
3419 .prepare_update
= imsm_prepare_update
,
3420 #endif /* MDASSEMBLE */