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
24 #include "platform-intel.h"
30 /* MPB == Metadata Parameter Block */
31 #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
32 #define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
33 #define MPB_VERSION_RAID0 "1.0.00"
34 #define MPB_VERSION_RAID1 "1.1.00"
35 #define MPB_VERSION_MANY_VOLUMES_PER_ARRAY "1.2.00"
36 #define MPB_VERSION_3OR4_DISK_ARRAY "1.2.01"
37 #define MPB_VERSION_RAID5 "1.2.02"
38 #define MPB_VERSION_5OR6_DISK_ARRAY "1.2.04"
39 #define MPB_VERSION_CNG "1.2.06"
40 #define MPB_VERSION_ATTRIBS "1.3.00"
41 #define MAX_SIGNATURE_LENGTH 32
42 #define MAX_RAID_SERIAL_LEN 16
44 #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000)
45 #define MPB_ATTRIB_PM __cpu_to_le32(0x40000000)
46 #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000)
47 #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001)
48 #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002)
49 #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004)
50 #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008)
51 #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010)
52 #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020)
54 #define MPB_SECTOR_CNT 418
55 #define IMSM_RESERVED_SECTORS 4096
56 #define SECT_PER_MB_SHIFT 11
58 /* Disk configuration info. */
59 #define IMSM_MAX_DEVICES 255
61 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
62 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
63 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
64 #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
65 #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
66 #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
67 __u32 status
; /* 0xF0 - 0xF3 */
68 __u32 owner_cfg_num
; /* which config 0,1,2... owns this disk */
69 #define IMSM_DISK_FILLERS 4
70 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
73 /* RAID map configuration infos. */
75 __u32 pba_of_lba0
; /* start address of partition */
76 __u32 blocks_per_member
;/* blocks per member */
77 __u32 num_data_stripes
; /* number of data stripes */
78 __u16 blocks_per_strip
;
79 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
80 #define IMSM_T_STATE_NORMAL 0
81 #define IMSM_T_STATE_UNINITIALIZED 1
82 #define IMSM_T_STATE_DEGRADED 2
83 #define IMSM_T_STATE_FAILED 3
85 #define IMSM_T_RAID0 0
86 #define IMSM_T_RAID1 1
87 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
88 __u8 num_members
; /* number of member disks */
89 __u8 num_domains
; /* number of parity domains */
90 __u8 failed_disk_num
; /* valid only when state is degraded */
92 __u32 filler
[7]; /* expansion area */
93 #define IMSM_ORD_REBUILD (1 << 24)
94 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
95 * top byte contains some flags
97 } __attribute__ ((packed
));
100 __u32 curr_migr_unit
;
101 __u32 checkpoint_id
; /* id to access curr_migr_unit */
102 __u8 migr_state
; /* Normal or Migrating */
104 #define MIGR_REBUILD 1
105 #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
106 #define MIGR_GEN_MIGR 3
107 #define MIGR_STATE_CHANGE 4
108 #define MIGR_REPAIR 5
109 __u8 migr_type
; /* Initializing, Rebuilding, ... */
111 __u8 fs_state
; /* fast-sync state for CnG (0xff == disabled) */
112 __u16 verify_errors
; /* number of mismatches */
113 __u16 bad_blocks
; /* number of bad blocks during verify */
115 struct imsm_map map
[1];
116 /* here comes another one if migr_state */
117 } __attribute__ ((packed
));
120 __u8 volume
[MAX_RAID_SERIAL_LEN
];
123 #define DEV_BOOTABLE __cpu_to_le32(0x01)
124 #define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
125 #define DEV_READ_COALESCING __cpu_to_le32(0x04)
126 #define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
127 #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
128 #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
129 #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
130 #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
131 #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
132 #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
133 #define DEV_CLONE_N_GO __cpu_to_le32(0x400)
134 #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
135 #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
136 __u32 status
; /* Persistent RaidDev status */
137 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
141 __u8 cng_master_disk
;
145 #define IMSM_DEV_FILLERS 10
146 __u32 filler
[IMSM_DEV_FILLERS
];
148 } __attribute__ ((packed
));
151 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
152 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
153 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
154 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
155 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
156 __u32 error_log_size
; /* 0x30 - 0x33 in bytes */
157 __u32 attributes
; /* 0x34 - 0x37 */
158 __u8 num_disks
; /* 0x38 Number of configured disks */
159 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
160 __u8 error_log_pos
; /* 0x3A */
161 __u8 fill
[1]; /* 0x3B */
162 __u32 cache_size
; /* 0x3c - 0x40 in mb */
163 __u32 orig_family_num
; /* 0x40 - 0x43 original family num */
164 __u32 pwr_cycle_count
; /* 0x44 - 0x47 simulated power cycle count for array */
165 __u32 bbm_log_size
; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
166 #define IMSM_FILLERS 35
167 __u32 filler
[IMSM_FILLERS
]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
168 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
169 /* here comes imsm_dev[num_raid_devs] */
170 /* here comes BBM logs */
171 } __attribute__ ((packed
));
173 #define BBM_LOG_MAX_ENTRIES 254
175 struct bbm_log_entry
{
176 __u64 defective_block_start
;
177 #define UNREADABLE 0xFFFFFFFF
178 __u32 spare_block_offset
;
179 __u16 remapped_marked_count
;
181 } __attribute__ ((__packed__
));
184 __u32 signature
; /* 0xABADB10C */
186 __u32 reserved_spare_block_count
; /* 0 */
187 __u32 reserved
; /* 0xFFFF */
188 __u64 first_spare_lba
;
189 struct bbm_log_entry mapped_block_entries
[BBM_LOG_MAX_ENTRIES
];
190 } __attribute__ ((__packed__
));
194 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
197 static __u8
migr_type(struct imsm_dev
*dev
)
199 if (dev
->vol
.migr_type
== MIGR_VERIFY
&&
200 dev
->status
& DEV_VERIFY_AND_FIX
)
203 return dev
->vol
.migr_type
;
206 static void set_migr_type(struct imsm_dev
*dev
, __u8 migr_type
)
208 /* for compatibility with older oroms convert MIGR_REPAIR, into
209 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
211 if (migr_type
== MIGR_REPAIR
) {
212 dev
->vol
.migr_type
= MIGR_VERIFY
;
213 dev
->status
|= DEV_VERIFY_AND_FIX
;
215 dev
->vol
.migr_type
= migr_type
;
216 dev
->status
&= ~DEV_VERIFY_AND_FIX
;
220 static unsigned int sector_count(__u32 bytes
)
222 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
225 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
227 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
231 struct imsm_dev
*dev
;
232 struct intel_dev
*next
;
236 /* internal representation of IMSM metadata */
239 void *buf
; /* O_DIRECT buffer for reading/writing metadata */
240 struct imsm_super
*anchor
; /* immovable parameters */
242 size_t len
; /* size of the 'buf' allocation */
243 void *next_buf
; /* for realloc'ing buf from the manager */
245 int updates_pending
; /* count of pending updates for mdmon */
246 int creating_imsm
; /* flag to indicate container creation */
247 int current_vol
; /* index of raid device undergoing creation */
248 __u32 create_offset
; /* common start for 'current_vol' */
249 __u32 random
; /* random data for seeding new family numbers */
250 struct intel_dev
*devlist
;
254 __u8 serial
[MAX_RAID_SERIAL_LEN
];
257 struct imsm_disk disk
;
260 struct extent
*e
; /* for determining freespace @ create */
261 int raiddisk
; /* slot to fill in autolayout */
263 struct dl
*add
; /* list of disks to add while mdmon active */
264 struct dl
*missing
; /* disks removed while we weren't looking */
265 struct bbm_log
*bbm_log
;
266 const char *hba
; /* device path of the raid controller for this metadata */
267 const struct imsm_orom
*orom
; /* platform firmware support */
268 struct intel_super
*next
; /* (temp) list for disambiguating family_num */
272 struct imsm_disk disk
;
273 #define IMSM_UNKNOWN_OWNER (-1)
275 struct intel_disk
*next
;
279 unsigned long long start
, size
;
282 /* definition of messages passed to imsm_process_update */
283 enum imsm_update_type
{
284 update_activate_spare
,
289 struct imsm_update_activate_spare
{
290 enum imsm_update_type type
;
294 struct imsm_update_activate_spare
*next
;
298 __u8 serial
[MAX_RAID_SERIAL_LEN
];
301 struct imsm_update_create_array
{
302 enum imsm_update_type type
;
307 struct imsm_update_add_disk
{
308 enum imsm_update_type type
;
311 static struct supertype
*match_metadata_desc_imsm(char *arg
)
313 struct supertype
*st
;
315 if (strcmp(arg
, "imsm") != 0 &&
316 strcmp(arg
, "default") != 0
320 st
= malloc(sizeof(*st
));
323 memset(st
, 0, sizeof(*st
));
324 st
->ss
= &super_imsm
;
325 st
->max_devs
= IMSM_MAX_DEVICES
;
326 st
->minor_version
= 0;
332 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
334 return &mpb
->sig
[MPB_SIG_LEN
];
338 /* retrieve a disk directly from the anchor when the anchor is known to be
339 * up-to-date, currently only at load time
341 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
343 if (index
>= mpb
->num_disks
)
345 return &mpb
->disk
[index
];
349 /* retrieve a disk from the parsed metadata */
350 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
354 for (d
= super
->disks
; d
; d
= d
->next
)
355 if (d
->index
== index
)
362 /* generate a checksum directly from the anchor when the anchor is known to be
363 * up-to-date, currently only at load or write_super after coalescing
365 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
367 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
368 __u32
*p
= (__u32
*) mpb
;
372 sum
+= __le32_to_cpu(*p
);
376 return sum
- __le32_to_cpu(mpb
->check_sum
);
379 static size_t sizeof_imsm_map(struct imsm_map
*map
)
381 return sizeof(struct imsm_map
) + sizeof(__u32
) * (map
->num_members
- 1);
384 struct imsm_map
*get_imsm_map(struct imsm_dev
*dev
, int second_map
)
386 struct imsm_map
*map
= &dev
->vol
.map
[0];
388 if (second_map
&& !dev
->vol
.migr_state
)
390 else if (second_map
) {
393 return ptr
+ sizeof_imsm_map(map
);
399 /* return the size of the device.
400 * migr_state increases the returned size if map[0] were to be duplicated
402 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
, int migr_state
)
404 size_t size
= sizeof(*dev
) - sizeof(struct imsm_map
) +
405 sizeof_imsm_map(get_imsm_map(dev
, 0));
407 /* migrating means an additional map */
408 if (dev
->vol
.migr_state
)
409 size
+= sizeof_imsm_map(get_imsm_map(dev
, 1));
411 size
+= sizeof_imsm_map(get_imsm_map(dev
, 0));
417 /* retrieve disk serial number list from a metadata update */
418 static struct disk_info
*get_disk_info(struct imsm_update_create_array
*update
)
421 struct disk_info
*inf
;
423 inf
= u
+ sizeof(*update
) - sizeof(struct imsm_dev
) +
424 sizeof_imsm_dev(&update
->dev
, 0);
430 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
436 if (index
>= mpb
->num_raid_devs
)
439 /* devices start after all disks */
440 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
442 for (i
= 0; i
<= index
; i
++)
444 return _mpb
+ offset
;
446 offset
+= sizeof_imsm_dev(_mpb
+ offset
, 0);
451 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
453 struct intel_dev
*dv
;
455 if (index
>= super
->anchor
->num_raid_devs
)
457 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
458 if (dv
->index
== index
)
463 static __u32
get_imsm_ord_tbl_ent(struct imsm_dev
*dev
, int slot
)
465 struct imsm_map
*map
;
467 if (dev
->vol
.migr_state
)
468 map
= get_imsm_map(dev
, 1);
470 map
= get_imsm_map(dev
, 0);
472 /* top byte identifies disk under rebuild */
473 return __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
476 #define ord_to_idx(ord) (((ord) << 8) >> 8)
477 static __u32
get_imsm_disk_idx(struct imsm_dev
*dev
, int slot
)
479 __u32 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
481 return ord_to_idx(ord
);
484 static void set_imsm_ord_tbl_ent(struct imsm_map
*map
, int slot
, __u32 ord
)
486 map
->disk_ord_tbl
[slot
] = __cpu_to_le32(ord
);
489 static int get_imsm_disk_slot(struct imsm_map
*map
, int idx
)
494 for (slot
= 0; slot
< map
->num_members
; slot
++) {
495 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
496 if (ord_to_idx(ord
) == idx
)
503 static int get_imsm_raid_level(struct imsm_map
*map
)
505 if (map
->raid_level
== 1) {
506 if (map
->num_members
== 2)
512 return map
->raid_level
;
515 static int cmp_extent(const void *av
, const void *bv
)
517 const struct extent
*a
= av
;
518 const struct extent
*b
= bv
;
519 if (a
->start
< b
->start
)
521 if (a
->start
> b
->start
)
526 static int count_memberships(struct dl
*dl
, struct intel_super
*super
)
531 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
532 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
533 struct imsm_map
*map
= get_imsm_map(dev
, 0);
535 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
542 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
544 /* find a list of used extents on the given physical device */
545 struct extent
*rv
, *e
;
547 int memberships
= count_memberships(dl
, super
);
548 __u32 reservation
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
550 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
555 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
556 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
557 struct imsm_map
*map
= get_imsm_map(dev
, 0);
559 if (get_imsm_disk_slot(map
, dl
->index
) >= 0) {
560 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
561 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
565 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
567 /* determine the start of the metadata
568 * when no raid devices are defined use the default
569 * ...otherwise allow the metadata to truncate the value
570 * as is the case with older versions of imsm
573 struct extent
*last
= &rv
[memberships
- 1];
576 remainder
= __le32_to_cpu(dl
->disk
.total_blocks
) -
577 (last
->start
+ last
->size
);
578 /* round down to 1k block to satisfy precision of the kernel
582 /* make sure remainder is still sane */
583 if (remainder
< ROUND_UP(super
->len
, 512) >> 9)
584 remainder
= ROUND_UP(super
->len
, 512) >> 9;
585 if (reservation
> remainder
)
586 reservation
= remainder
;
588 e
->start
= __le32_to_cpu(dl
->disk
.total_blocks
) - reservation
;
593 /* try to determine how much space is reserved for metadata from
594 * the last get_extents() entry, otherwise fallback to the
597 static __u32
imsm_reserved_sectors(struct intel_super
*super
, struct dl
*dl
)
603 /* for spares just return a minimal reservation which will grow
604 * once the spare is picked up by an array
607 return MPB_SECTOR_CNT
;
609 e
= get_extents(super
, dl
);
611 return MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
613 /* scroll to last entry */
614 for (i
= 0; e
[i
].size
; i
++)
617 rv
= __le32_to_cpu(dl
->disk
.total_blocks
) - e
[i
].start
;
624 static int is_spare(struct imsm_disk
*disk
)
626 return (disk
->status
& SPARE_DISK
) == SPARE_DISK
;
629 static int is_configured(struct imsm_disk
*disk
)
631 return (disk
->status
& CONFIGURED_DISK
) == CONFIGURED_DISK
;
634 static int is_failed(struct imsm_disk
*disk
)
636 return (disk
->status
& FAILED_DISK
) == FAILED_DISK
;
640 static void print_imsm_dev(struct imsm_dev
*dev
, char *uuid
, int disk_idx
)
644 struct imsm_map
*map
= get_imsm_map(dev
, 0);
648 printf("[%.16s]:\n", dev
->volume
);
649 printf(" UUID : %s\n", uuid
);
650 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
651 printf(" Members : %d\n", map
->num_members
);
652 slot
= get_imsm_disk_slot(map
, disk_idx
);
654 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
655 printf(" This Slot : %d%s\n", slot
,
656 ord
& IMSM_ORD_REBUILD
? " (out-of-sync)" : "");
658 printf(" This Slot : ?\n");
659 sz
= __le32_to_cpu(dev
->size_high
);
661 sz
+= __le32_to_cpu(dev
->size_low
);
662 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
663 human_size(sz
* 512));
664 sz
= __le32_to_cpu(map
->blocks_per_member
);
665 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
666 human_size(sz
* 512));
667 printf(" Sector Offset : %u\n",
668 __le32_to_cpu(map
->pba_of_lba0
));
669 printf(" Num Stripes : %u\n",
670 __le32_to_cpu(map
->num_data_stripes
));
671 printf(" Chunk Size : %u KiB\n",
672 __le16_to_cpu(map
->blocks_per_strip
) / 2);
673 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
674 printf(" Migrate State : %s", dev
->vol
.migr_state
? "migrating" : "idle\n");
675 if (dev
->vol
.migr_state
) {
676 if (migr_type(dev
) == MIGR_INIT
)
677 printf(": initializing\n");
678 else if (migr_type(dev
) == MIGR_REBUILD
)
679 printf(": rebuilding\n");
680 else if (migr_type(dev
) == MIGR_VERIFY
)
682 else if (migr_type(dev
) == MIGR_GEN_MIGR
)
683 printf(": general migration\n");
684 else if (migr_type(dev
) == MIGR_STATE_CHANGE
)
685 printf(": state change\n");
686 else if (migr_type(dev
) == MIGR_REPAIR
)
687 printf(": repair\n");
689 printf(": <unknown:%d>\n", migr_type(dev
));
691 printf(" Map State : %s", map_state_str
[map
->map_state
]);
692 if (dev
->vol
.migr_state
) {
693 struct imsm_map
*map
= get_imsm_map(dev
, 1);
694 printf(" <-- %s", map_state_str
[map
->map_state
]);
697 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
700 static void print_imsm_disk(struct imsm_super
*mpb
, int index
, __u32 reserved
)
702 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
703 char str
[MAX_RAID_SERIAL_LEN
+ 1];
706 if (index
< 0 || !disk
)
710 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
711 printf(" Disk%02d Serial : %s\n", index
, str
);
712 printf(" State :%s%s%s\n", is_spare(disk
) ? " spare" : "",
713 is_configured(disk
) ? " active" : "",
714 is_failed(disk
) ? " failed" : "");
715 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
716 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
717 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
718 human_size(sz
* 512));
721 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
);
723 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
725 struct intel_super
*super
= st
->sb
;
726 struct imsm_super
*mpb
= super
->anchor
;
727 char str
[MAX_SIGNATURE_LENGTH
];
732 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
735 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
736 printf(" Magic : %s\n", str
);
737 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
738 printf(" Version : %s\n", get_imsm_version(mpb
));
739 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb
->orig_family_num
));
740 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
741 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
742 getinfo_super_imsm(st
, &info
);
743 fname_from_uuid(st
, &info
, nbuf
, ':');
744 printf(" UUID : %s\n", nbuf
+ 5);
745 sum
= __le32_to_cpu(mpb
->check_sum
);
746 printf(" Checksum : %08x %s\n", sum
,
747 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
748 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
749 printf(" Disks : %d\n", mpb
->num_disks
);
750 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
751 print_imsm_disk(mpb
, super
->disks
->index
, reserved
);
752 if (super
->bbm_log
) {
753 struct bbm_log
*log
= super
->bbm_log
;
756 printf("Bad Block Management Log:\n");
757 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
758 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
759 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
760 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
761 printf(" First Spare : %llx\n",
762 (unsigned long long) __le64_to_cpu(log
->first_spare_lba
));
764 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
766 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
768 super
->current_vol
= i
;
769 getinfo_super_imsm(st
, &info
);
770 fname_from_uuid(st
, &info
, nbuf
, ':');
771 print_imsm_dev(dev
, nbuf
+ 5, super
->disks
->index
);
773 for (i
= 0; i
< mpb
->num_disks
; i
++) {
774 if (i
== super
->disks
->index
)
776 print_imsm_disk(mpb
, i
, reserved
);
780 static void brief_examine_super_imsm(struct supertype
*st
, int verbose
)
782 /* We just write a generic IMSM ARRAY entry */
785 struct intel_super
*super
= st
->sb
;
787 if (!super
->anchor
->num_raid_devs
) {
788 printf("ARRAY metadata=imsm\n");
792 getinfo_super_imsm(st
, &info
);
793 fname_from_uuid(st
, &info
, nbuf
, ':');
794 printf("ARRAY metadata=imsm UUID=%s\n", nbuf
+ 5);
797 static void brief_examine_subarrays_imsm(struct supertype
*st
, int verbose
)
799 /* We just write a generic IMSM ARRAY entry */
803 struct intel_super
*super
= st
->sb
;
806 if (!super
->anchor
->num_raid_devs
)
809 getinfo_super_imsm(st
, &info
);
810 fname_from_uuid(st
, &info
, nbuf
, ':');
811 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
812 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
814 super
->current_vol
= i
;
815 getinfo_super_imsm(st
, &info
);
816 fname_from_uuid(st
, &info
, nbuf1
, ':');
817 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
818 dev
->volume
, nbuf
+ 5, i
, nbuf1
+ 5);
822 static void export_examine_super_imsm(struct supertype
*st
)
824 struct intel_super
*super
= st
->sb
;
825 struct imsm_super
*mpb
= super
->anchor
;
829 getinfo_super_imsm(st
, &info
);
830 fname_from_uuid(st
, &info
, nbuf
, ':');
831 printf("MD_METADATA=imsm\n");
832 printf("MD_LEVEL=container\n");
833 printf("MD_UUID=%s\n", nbuf
+5);
834 printf("MD_DEVICES=%u\n", mpb
->num_disks
);
837 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
842 getinfo_super_imsm(st
, &info
);
843 fname_from_uuid(st
, &info
, nbuf
, ':');
844 printf("\n UUID : %s\n", nbuf
+ 5);
847 static void brief_detail_super_imsm(struct supertype
*st
)
851 getinfo_super_imsm(st
, &info
);
852 fname_from_uuid(st
, &info
, nbuf
, ':');
853 printf(" UUID=%s", nbuf
+ 5);
856 static int imsm_read_serial(int fd
, char *devname
, __u8
*serial
);
857 static void fd2devname(int fd
, char *name
);
859 static int imsm_enumerate_ports(const char *hba_path
, int port_count
, int host_base
, int verbose
)
861 /* dump an unsorted list of devices attached to ahci, as well as
862 * non-connected ports
864 int hba_len
= strlen(hba_path
) + 1;
869 unsigned long port_mask
= (1 << port_count
) - 1;
871 if (port_count
> sizeof(port_mask
) * 8) {
873 fprintf(stderr
, Name
": port_count %d out of range\n", port_count
);
877 /* scroll through /sys/dev/block looking for devices attached to
880 dir
= opendir("/sys/dev/block");
881 for (ent
= dir
? readdir(dir
) : NULL
; ent
; ent
= readdir(dir
)) {
892 if (sscanf(ent
->d_name
, "%d:%d", &major
, &minor
) != 2)
894 path
= devt_to_devpath(makedev(major
, minor
));
897 if (!path_attached_to_hba(path
, hba_path
)) {
903 /* retrieve the scsi device type */
904 if (asprintf(&device
, "/sys/dev/block/%d:%d/device/xxxxxxx", major
, minor
) < 0) {
906 fprintf(stderr
, Name
": failed to allocate 'device'\n");
910 sprintf(device
, "/sys/dev/block/%d:%d/device/type", major
, minor
);
911 if (load_sys(device
, buf
) != 0) {
913 fprintf(stderr
, Name
": failed to read device type for %s\n",
919 type
= strtoul(buf
, NULL
, 10);
921 /* if it's not a disk print the vendor and model */
922 if (!(type
== 0 || type
== 7 || type
== 14)) {
925 sprintf(device
, "/sys/dev/block/%d:%d/device/vendor", major
, minor
);
926 if (load_sys(device
, buf
) == 0) {
927 strncpy(vendor
, buf
, sizeof(vendor
));
928 vendor
[sizeof(vendor
) - 1] = '\0';
929 c
= (char *) &vendor
[sizeof(vendor
) - 1];
930 while (isspace(*c
) || *c
== '\0')
934 sprintf(device
, "/sys/dev/block/%d:%d/device/model", major
, minor
);
935 if (load_sys(device
, buf
) == 0) {
936 strncpy(model
, buf
, sizeof(model
));
937 model
[sizeof(model
) - 1] = '\0';
938 c
= (char *) &model
[sizeof(model
) - 1];
939 while (isspace(*c
) || *c
== '\0')
943 if (vendor
[0] && model
[0])
944 sprintf(buf
, "%.64s %.64s", vendor
, model
);
946 switch (type
) { /* numbers from hald/linux/device.c */
947 case 1: sprintf(buf
, "tape"); break;
948 case 2: sprintf(buf
, "printer"); break;
949 case 3: sprintf(buf
, "processor"); break;
951 case 5: sprintf(buf
, "cdrom"); break;
952 case 6: sprintf(buf
, "scanner"); break;
953 case 8: sprintf(buf
, "media_changer"); break;
954 case 9: sprintf(buf
, "comm"); break;
955 case 12: sprintf(buf
, "raid"); break;
956 default: sprintf(buf
, "unknown");
962 /* chop device path to 'host%d' and calculate the port number */
963 c
= strchr(&path
[hba_len
], '/');
966 fprintf(stderr
, Name
": %s - invalid path name\n", path
+ hba_len
);
971 if (sscanf(&path
[hba_len
], "host%d", &port
) == 1)
975 *c
= '/'; /* repair the full string */
976 fprintf(stderr
, Name
": failed to determine port number for %s\n",
983 /* mark this port as used */
984 port_mask
&= ~(1 << port
);
986 /* print out the device information */
988 printf(" Port%d : - non-disk device (%s) -\n", port
, buf
);
992 fd
= dev_open(ent
->d_name
, O_RDONLY
);
994 printf(" Port%d : - disk info unavailable -\n", port
);
997 printf(" Port%d : %s", port
, buf
);
998 if (imsm_read_serial(fd
, NULL
, (__u8
*) buf
) == 0)
999 printf(" (%s)\n", buf
);
1014 for (i
= 0; i
< port_count
; i
++)
1015 if (port_mask
& (1 << i
))
1016 printf(" Port%d : - no device attached -\n", i
);
1022 static int detail_platform_imsm(int verbose
, int enumerate_only
)
1024 /* There are two components to imsm platform support, the ahci SATA
1025 * controller and the option-rom. To find the SATA controller we
1026 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
1027 * controller with the Intel vendor id is present. This approach
1028 * allows mdadm to leverage the kernel's ahci detection logic, with the
1029 * caveat that if ahci.ko is not loaded mdadm will not be able to
1030 * detect platform raid capabilities. The option-rom resides in a
1031 * platform "Adapter ROM". We scan for its signature to retrieve the
1032 * platform capabilities. If raid support is disabled in the BIOS the
1033 * option-rom capability structure will not be available.
1035 const struct imsm_orom
*orom
;
1036 struct sys_dev
*list
, *hba
;
1039 const char *hba_path
;
1043 if (enumerate_only
) {
1044 if (check_env("IMSM_NO_PLATFORM") || find_imsm_orom())
1049 list
= find_driver_devices("pci", "ahci");
1050 for (hba
= list
; hba
; hba
= hba
->next
)
1051 if (devpath_to_vendor(hba
->path
) == 0x8086)
1056 fprintf(stderr
, Name
": unable to find active ahci controller\n");
1057 free_sys_dev(&list
);
1060 fprintf(stderr
, Name
": found Intel SATA AHCI Controller\n");
1061 hba_path
= hba
->path
;
1063 free_sys_dev(&list
);
1065 orom
= find_imsm_orom();
1068 fprintf(stderr
, Name
": imsm option-rom not found\n");
1072 printf(" Platform : Intel(R) Matrix Storage Manager\n");
1073 printf(" Version : %d.%d.%d.%d\n", orom
->major_ver
, orom
->minor_ver
,
1074 orom
->hotfix_ver
, orom
->build
);
1075 printf(" RAID Levels :%s%s%s%s%s\n",
1076 imsm_orom_has_raid0(orom
) ? " raid0" : "",
1077 imsm_orom_has_raid1(orom
) ? " raid1" : "",
1078 imsm_orom_has_raid1e(orom
) ? " raid1e" : "",
1079 imsm_orom_has_raid10(orom
) ? " raid10" : "",
1080 imsm_orom_has_raid5(orom
) ? " raid5" : "");
1081 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
1082 imsm_orom_has_chunk(orom
, 2) ? " 2k" : "",
1083 imsm_orom_has_chunk(orom
, 4) ? " 4k" : "",
1084 imsm_orom_has_chunk(orom
, 8) ? " 8k" : "",
1085 imsm_orom_has_chunk(orom
, 16) ? " 16k" : "",
1086 imsm_orom_has_chunk(orom
, 32) ? " 32k" : "",
1087 imsm_orom_has_chunk(orom
, 64) ? " 64k" : "",
1088 imsm_orom_has_chunk(orom
, 128) ? " 128k" : "",
1089 imsm_orom_has_chunk(orom
, 256) ? " 256k" : "",
1090 imsm_orom_has_chunk(orom
, 512) ? " 512k" : "",
1091 imsm_orom_has_chunk(orom
, 1024*1) ? " 1M" : "",
1092 imsm_orom_has_chunk(orom
, 1024*2) ? " 2M" : "",
1093 imsm_orom_has_chunk(orom
, 1024*4) ? " 4M" : "",
1094 imsm_orom_has_chunk(orom
, 1024*8) ? " 8M" : "",
1095 imsm_orom_has_chunk(orom
, 1024*16) ? " 16M" : "",
1096 imsm_orom_has_chunk(orom
, 1024*32) ? " 32M" : "",
1097 imsm_orom_has_chunk(orom
, 1024*64) ? " 64M" : "");
1098 printf(" Max Disks : %d\n", orom
->tds
);
1099 printf(" Max Volumes : %d\n", orom
->vpa
);
1100 printf(" I/O Controller : %s\n", hba_path
);
1102 /* find the smallest scsi host number to determine a port number base */
1103 dir
= opendir(hba_path
);
1104 for (ent
= dir
? readdir(dir
) : NULL
; ent
; ent
= readdir(dir
)) {
1107 if (sscanf(ent
->d_name
, "host%d", &host
) != 1)
1109 if (port_count
== 0)
1111 else if (host
< host_base
)
1114 if (host
+ 1 > port_count
+ host_base
)
1115 port_count
= host
+ 1 - host_base
;
1121 if (!port_count
|| imsm_enumerate_ports(hba_path
, port_count
,
1122 host_base
, verbose
) != 0) {
1124 fprintf(stderr
, Name
": failed to enumerate ports\n");
1132 static int match_home_imsm(struct supertype
*st
, char *homehost
)
1134 /* the imsm metadata format does not specify any host
1135 * identification information. We return -1 since we can never
1136 * confirm nor deny whether a given array is "meant" for this
1137 * host. We rely on compare_super and the 'family_num' fields to
1138 * exclude member disks that do not belong, and we rely on
1139 * mdadm.conf to specify the arrays that should be assembled.
1140 * Auto-assembly may still pick up "foreign" arrays.
1146 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
1148 /* The uuid returned here is used for:
1149 * uuid to put into bitmap file (Create, Grow)
1150 * uuid for backup header when saving critical section (Grow)
1151 * comparing uuids when re-adding a device into an array
1152 * In these cases the uuid required is that of the data-array,
1153 * not the device-set.
1154 * uuid to recognise same set when adding a missing device back
1155 * to an array. This is a uuid for the device-set.
1157 * For each of these we can make do with a truncated
1158 * or hashed uuid rather than the original, as long as
1160 * In each case the uuid required is that of the data-array,
1161 * not the device-set.
1163 /* imsm does not track uuid's so we synthesis one using sha1 on
1164 * - The signature (Which is constant for all imsm array, but no matter)
1165 * - the orig_family_num of the container
1166 * - the index number of the volume
1167 * - the 'serial' number of the volume.
1168 * Hopefully these are all constant.
1170 struct intel_super
*super
= st
->sb
;
1173 struct sha1_ctx ctx
;
1174 struct imsm_dev
*dev
= NULL
;
1177 /* some mdadm versions failed to set ->orig_family_num, in which
1178 * case fall back to ->family_num. orig_family_num will be
1179 * fixed up with the first metadata update.
1181 family_num
= super
->anchor
->orig_family_num
;
1182 if (family_num
== 0)
1183 family_num
= super
->anchor
->family_num
;
1184 sha1_init_ctx(&ctx
);
1185 sha1_process_bytes(super
->anchor
->sig
, MPB_SIG_LEN
, &ctx
);
1186 sha1_process_bytes(&family_num
, sizeof(__u32
), &ctx
);
1187 if (super
->current_vol
>= 0)
1188 dev
= get_imsm_dev(super
, super
->current_vol
);
1190 __u32 vol
= super
->current_vol
;
1191 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
1192 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
1194 sha1_finish_ctx(&ctx
, buf
);
1195 memcpy(uuid
, buf
, 4*4);
1200 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
1202 __u8
*v
= get_imsm_version(mpb
);
1203 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
1204 char major
[] = { 0, 0, 0 };
1205 char minor
[] = { 0 ,0, 0 };
1206 char patch
[] = { 0, 0, 0 };
1207 char *ver_parse
[] = { major
, minor
, patch
};
1211 while (*v
!= '\0' && v
< end
) {
1212 if (*v
!= '.' && j
< 2)
1213 ver_parse
[i
][j
++] = *v
;
1221 *m
= strtol(minor
, NULL
, 0);
1222 *p
= strtol(patch
, NULL
, 0);
1226 static int imsm_level_to_layout(int level
)
1234 return ALGORITHM_LEFT_ASYMMETRIC
;
1241 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
1243 struct intel_super
*super
= st
->sb
;
1244 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
1245 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1249 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1250 if (dl
->raiddisk
== info
->disk
.raid_disk
)
1252 info
->container_member
= super
->current_vol
;
1253 info
->array
.raid_disks
= map
->num_members
;
1254 info
->array
.level
= get_imsm_raid_level(map
);
1255 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
1256 info
->array
.md_minor
= -1;
1257 info
->array
.ctime
= 0;
1258 info
->array
.utime
= 0;
1259 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
1260 info
->array
.state
= !dev
->vol
.dirty
;
1261 info
->custom_array_size
= __le32_to_cpu(dev
->size_high
);
1262 info
->custom_array_size
<<= 32;
1263 info
->custom_array_size
|= __le32_to_cpu(dev
->size_low
);
1265 info
->disk
.major
= 0;
1266 info
->disk
.minor
= 0;
1268 info
->disk
.major
= dl
->major
;
1269 info
->disk
.minor
= dl
->minor
;
1272 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
1273 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
1274 memset(info
->uuid
, 0, sizeof(info
->uuid
));
1276 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
1277 info
->resync_start
= 0;
1278 else if (dev
->vol
.migr_state
)
1279 /* FIXME add curr_migr_unit to resync_start conversion */
1280 info
->resync_start
= 0;
1282 info
->resync_start
= ~0ULL;
1284 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
1285 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
1287 info
->array
.major_version
= -1;
1288 info
->array
.minor_version
= -2;
1289 devname
= devnum2devname(st
->container_dev
);
1290 *info
->text_version
= '\0';
1292 sprintf(info
->text_version
, "/%s/%d", devname
, info
->container_member
);
1294 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
1295 uuid_from_super_imsm(st
, info
->uuid
);
1298 /* check the config file to see if we can return a real uuid for this spare */
1299 static void fixup_container_spare_uuid(struct mdinfo
*inf
)
1301 struct mddev_ident_s
*array_list
;
1303 if (inf
->array
.level
!= LEVEL_CONTAINER
||
1304 memcmp(inf
->uuid
, uuid_match_any
, sizeof(int[4])) != 0)
1307 array_list
= conf_get_ident(NULL
);
1309 for (; array_list
; array_list
= array_list
->next
) {
1310 if (array_list
->uuid_set
) {
1311 struct supertype
*_sst
; /* spare supertype */
1312 struct supertype
*_cst
; /* container supertype */
1314 _cst
= array_list
->st
;
1316 _sst
= _cst
->ss
->match_metadata_desc(inf
->text_version
);
1321 memcpy(inf
->uuid
, array_list
->uuid
, sizeof(int[4]));
1329 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
1331 struct intel_super
*super
= st
->sb
;
1332 struct imsm_disk
*disk
;
1334 if (super
->current_vol
>= 0) {
1335 getinfo_super_imsm_volume(st
, info
);
1339 /* Set raid_disks to zero so that Assemble will always pull in valid
1342 info
->array
.raid_disks
= 0;
1343 info
->array
.level
= LEVEL_CONTAINER
;
1344 info
->array
.layout
= 0;
1345 info
->array
.md_minor
= -1;
1346 info
->array
.ctime
= 0; /* N/A for imsm */
1347 info
->array
.utime
= 0;
1348 info
->array
.chunk_size
= 0;
1350 info
->disk
.major
= 0;
1351 info
->disk
.minor
= 0;
1352 info
->disk
.raid_disk
= -1;
1353 info
->reshape_active
= 0;
1354 info
->array
.major_version
= -1;
1355 info
->array
.minor_version
= -2;
1356 strcpy(info
->text_version
, "imsm");
1357 info
->safe_mode_delay
= 0;
1358 info
->disk
.number
= -1;
1359 info
->disk
.state
= 0;
1363 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
1365 disk
= &super
->disks
->disk
;
1366 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
1367 info
->component_size
= reserved
;
1368 info
->disk
.state
= is_configured(disk
) ? (1 << MD_DISK_ACTIVE
) : 0;
1369 /* we don't change info->disk.raid_disk here because
1370 * this state will be finalized in mdmon after we have
1371 * found the 'most fresh' version of the metadata
1373 info
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
1374 info
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
1377 /* only call uuid_from_super_imsm when this disk is part of a populated container,
1378 * ->compare_super may have updated the 'num_raid_devs' field for spares
1380 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
1381 uuid_from_super_imsm(st
, info
->uuid
);
1383 memcpy(info
->uuid
, uuid_match_any
, sizeof(int[4]));
1384 fixup_container_spare_uuid(info
);
1388 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
1389 char *update
, char *devname
, int verbose
,
1390 int uuid_set
, char *homehost
)
1392 /* For 'assemble' and 'force' we need to return non-zero if any
1393 * change was made. For others, the return value is ignored.
1394 * Update options are:
1395 * force-one : This device looks a bit old but needs to be included,
1396 * update age info appropriately.
1397 * assemble: clear any 'faulty' flag to allow this device to
1399 * force-array: Array is degraded but being forced, mark it clean
1400 * if that will be needed to assemble it.
1402 * newdev: not used ????
1403 * grow: Array has gained a new device - this is currently for
1405 * resync: mark as dirty so a resync will happen.
1406 * name: update the name - preserving the homehost
1407 * uuid: Change the uuid of the array to match watch is given
1409 * Following are not relevant for this imsm:
1410 * sparc2.2 : update from old dodgey metadata
1411 * super-minor: change the preferred_minor number
1412 * summaries: update redundant counters.
1413 * homehost: update the recorded homehost
1414 * _reshape_progress: record new reshape_progress position.
1417 struct intel_super
*super
= st
->sb
;
1418 struct imsm_super
*mpb
;
1420 /* we can only update container info */
1421 if (!super
|| super
->current_vol
>= 0 || !super
->anchor
)
1424 mpb
= super
->anchor
;
1426 if (strcmp(update
, "uuid") == 0 && uuid_set
&& !info
->update_private
)
1428 Name
": '--uuid' not supported for imsm metadata\n");
1429 else if (strcmp(update
, "uuid") == 0 && uuid_set
&& info
->update_private
) {
1430 mpb
->orig_family_num
= *((__u32
*) info
->update_private
);
1432 } else if (strcmp(update
, "uuid") == 0) {
1433 __u32
*new_family
= malloc(sizeof(*new_family
));
1435 /* update orig_family_number with the incoming random
1436 * data, report the new effective uuid, and store the
1437 * new orig_family_num for future updates.
1440 memcpy(&mpb
->orig_family_num
, info
->uuid
, sizeof(__u32
));
1441 uuid_from_super_imsm(st
, info
->uuid
);
1442 *new_family
= mpb
->orig_family_num
;
1443 info
->update_private
= new_family
;
1446 } else if (strcmp(update
, "assemble") == 0)
1450 Name
": '--update=%s' not supported for imsm metadata\n",
1453 /* successful update? recompute checksum */
1455 mpb
->check_sum
= __le32_to_cpu(__gen_imsm_checksum(mpb
));
1460 static size_t disks_to_mpb_size(int disks
)
1464 size
= sizeof(struct imsm_super
);
1465 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
1466 size
+= 2 * sizeof(struct imsm_dev
);
1467 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
1468 size
+= (4 - 2) * sizeof(struct imsm_map
);
1469 /* 4 possible disk_ord_tbl's */
1470 size
+= 4 * (disks
- 1) * sizeof(__u32
);
1475 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
1477 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
1480 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
1483 static void free_devlist(struct intel_super
*super
)
1485 struct intel_dev
*dv
;
1487 while (super
->devlist
) {
1488 dv
= super
->devlist
->next
;
1489 free(super
->devlist
->dev
);
1490 free(super
->devlist
);
1491 super
->devlist
= dv
;
1495 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
1497 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
1500 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
1504 * 0 same, or first was empty, and second was copied
1505 * 1 second had wrong number
1507 * 3 wrong other info
1509 struct intel_super
*first
= st
->sb
;
1510 struct intel_super
*sec
= tst
->sb
;
1518 /* if an anchor does not have num_raid_devs set then it is a free
1521 if (first
->anchor
->num_raid_devs
> 0 &&
1522 sec
->anchor
->num_raid_devs
> 0) {
1523 /* Determine if these disks might ever have been
1524 * related. Further disambiguation can only take place
1525 * in load_super_imsm_all
1527 __u32 first_family
= first
->anchor
->orig_family_num
;
1528 __u32 sec_family
= sec
->anchor
->orig_family_num
;
1530 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
,
1531 MAX_SIGNATURE_LENGTH
) != 0)
1534 if (first_family
== 0)
1535 first_family
= first
->anchor
->family_num
;
1536 if (sec_family
== 0)
1537 sec_family
= sec
->anchor
->family_num
;
1539 if (first_family
!= sec_family
)
1545 /* if 'first' is a spare promote it to a populated mpb with sec's
1548 if (first
->anchor
->num_raid_devs
== 0 &&
1549 sec
->anchor
->num_raid_devs
> 0) {
1551 struct intel_dev
*dv
;
1552 struct imsm_dev
*dev
;
1554 /* we need to copy raid device info from sec if an allocation
1555 * fails here we don't associate the spare
1557 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
1558 dv
= malloc(sizeof(*dv
));
1561 dev
= malloc(sizeof_imsm_dev(get_imsm_dev(sec
, i
), 1));
1568 dv
->next
= first
->devlist
;
1569 first
->devlist
= dv
;
1571 if (i
< sec
->anchor
->num_raid_devs
) {
1572 /* allocation failure */
1573 free_devlist(first
);
1574 fprintf(stderr
, "imsm: failed to associate spare\n");
1577 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
1578 first
->anchor
->orig_family_num
= sec
->anchor
->orig_family_num
;
1579 first
->anchor
->family_num
= sec
->anchor
->family_num
;
1580 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++)
1581 imsm_copy_dev(get_imsm_dev(first
, i
), get_imsm_dev(sec
, i
));
1587 static void fd2devname(int fd
, char *name
)
1596 if (fstat(fd
, &st
) != 0)
1598 sprintf(path
, "/sys/dev/block/%d:%d",
1599 major(st
.st_rdev
), minor(st
.st_rdev
));
1601 rv
= readlink(path
, dname
, sizeof(dname
));
1606 nm
= strrchr(dname
, '/');
1608 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
1611 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
1613 static int imsm_read_serial(int fd
, char *devname
,
1614 __u8 serial
[MAX_RAID_SERIAL_LEN
])
1616 unsigned char scsi_serial
[255];
1625 memset(scsi_serial
, 0, sizeof(scsi_serial
));
1627 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
1629 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
1630 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1631 fd2devname(fd
, (char *) serial
);
1638 Name
": Failed to retrieve serial for %s\n",
1643 rsp_len
= scsi_serial
[3];
1647 Name
": Failed to retrieve serial for %s\n",
1651 rsp_buf
= (char *) &scsi_serial
[4];
1653 /* trim all whitespace and non-printable characters and convert
1656 for (i
= 0, dest
= rsp_buf
; i
< rsp_len
; i
++) {
1659 /* ':' is reserved for use in placeholder serial
1660 * numbers for missing disks
1668 len
= dest
- rsp_buf
;
1671 /* truncate leading characters */
1672 if (len
> MAX_RAID_SERIAL_LEN
) {
1673 dest
+= len
- MAX_RAID_SERIAL_LEN
;
1674 len
= MAX_RAID_SERIAL_LEN
;
1677 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
1678 memcpy(serial
, dest
, len
);
1683 static int serialcmp(__u8
*s1
, __u8
*s2
)
1685 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
1688 static void serialcpy(__u8
*dest
, __u8
*src
)
1690 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
1694 static struct dl
*serial_to_dl(__u8
*serial
, struct intel_super
*super
)
1698 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1699 if (serialcmp(dl
->serial
, serial
) == 0)
1706 static struct imsm_disk
*
1707 __serial_to_disk(__u8
*serial
, struct imsm_super
*mpb
, int *idx
)
1711 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1712 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
1714 if (serialcmp(disk
->serial
, serial
) == 0) {
1725 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
1727 struct imsm_disk
*disk
;
1732 __u8 serial
[MAX_RAID_SERIAL_LEN
];
1734 rv
= imsm_read_serial(fd
, devname
, serial
);
1739 dl
= calloc(1, sizeof(*dl
));
1743 Name
": failed to allocate disk buffer for %s\n",
1749 dl
->major
= major(stb
.st_rdev
);
1750 dl
->minor
= minor(stb
.st_rdev
);
1751 dl
->next
= super
->disks
;
1752 dl
->fd
= keep_fd
? fd
: -1;
1753 assert(super
->disks
== NULL
);
1755 serialcpy(dl
->serial
, serial
);
1758 fd2devname(fd
, name
);
1760 dl
->devname
= strdup(devname
);
1762 dl
->devname
= strdup(name
);
1764 /* look up this disk's index in the current anchor */
1765 disk
= __serial_to_disk(dl
->serial
, super
->anchor
, &dl
->index
);
1768 /* only set index on disks that are a member of a
1769 * populated contianer, i.e. one with raid_devs
1771 if (is_failed(&dl
->disk
))
1773 else if (is_spare(&dl
->disk
))
1781 /* When migrating map0 contains the 'destination' state while map1
1782 * contains the current state. When not migrating map0 contains the
1783 * current state. This routine assumes that map[0].map_state is set to
1784 * the current array state before being called.
1786 * Migration is indicated by one of the following states
1787 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
1788 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
1789 * map1state=unitialized)
1790 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
1792 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
1793 * map1state=degraded)
1795 static void migrate(struct imsm_dev
*dev
, __u8 to_state
, int migr_type
)
1797 struct imsm_map
*dest
;
1798 struct imsm_map
*src
= get_imsm_map(dev
, 0);
1800 dev
->vol
.migr_state
= 1;
1801 set_migr_type(dev
, migr_type
);
1802 dev
->vol
.curr_migr_unit
= 0;
1803 dest
= get_imsm_map(dev
, 1);
1805 /* duplicate and then set the target end state in map[0] */
1806 memcpy(dest
, src
, sizeof_imsm_map(src
));
1807 if (migr_type
== MIGR_REBUILD
) {
1811 for (i
= 0; i
< src
->num_members
; i
++) {
1812 ord
= __le32_to_cpu(src
->disk_ord_tbl
[i
]);
1813 set_imsm_ord_tbl_ent(src
, i
, ord_to_idx(ord
));
1817 src
->map_state
= to_state
;
1820 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
1822 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1823 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
1826 /* merge any IMSM_ORD_REBUILD bits that were not successfully
1827 * completed in the last migration.
1829 * FIXME add support for online capacity expansion and
1830 * raid-level-migration
1832 for (i
= 0; i
< prev
->num_members
; i
++)
1833 map
->disk_ord_tbl
[i
] |= prev
->disk_ord_tbl
[i
];
1835 dev
->vol
.migr_state
= 0;
1836 dev
->vol
.curr_migr_unit
= 0;
1837 map
->map_state
= map_state
;
1841 static int parse_raid_devices(struct intel_super
*super
)
1844 struct imsm_dev
*dev_new
;
1845 size_t len
, len_migr
;
1846 size_t space_needed
= 0;
1847 struct imsm_super
*mpb
= super
->anchor
;
1849 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1850 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
1851 struct intel_dev
*dv
;
1853 len
= sizeof_imsm_dev(dev_iter
, 0);
1854 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
1856 space_needed
+= len_migr
- len
;
1858 dv
= malloc(sizeof(*dv
));
1861 dev_new
= malloc(len_migr
);
1866 imsm_copy_dev(dev_new
, dev_iter
);
1869 dv
->next
= super
->devlist
;
1870 super
->devlist
= dv
;
1873 /* ensure that super->buf is large enough when all raid devices
1876 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
1879 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
1880 if (posix_memalign(&buf
, 512, len
) != 0)
1883 memcpy(buf
, super
->buf
, super
->len
);
1884 memset(buf
+ super
->len
, 0, len
- super
->len
);
1893 /* retrieve a pointer to the bbm log which starts after all raid devices */
1894 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
1898 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
1900 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
1906 static void __free_imsm(struct intel_super
*super
, int free_disks
);
1908 /* load_imsm_mpb - read matrix metadata
1909 * allocates super->mpb to be freed by free_super
1911 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
1913 unsigned long long dsize
;
1914 unsigned long long sectors
;
1916 struct imsm_super
*anchor
;
1919 get_dev_size(fd
, NULL
, &dsize
);
1921 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
1924 Name
": Cannot seek to anchor block on %s: %s\n",
1925 devname
, strerror(errno
));
1929 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
1932 Name
": Failed to allocate imsm anchor buffer"
1933 " on %s\n", devname
);
1936 if (read(fd
, anchor
, 512) != 512) {
1939 Name
": Cannot read anchor block on %s: %s\n",
1940 devname
, strerror(errno
));
1945 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
1948 Name
": no IMSM anchor on %s\n", devname
);
1953 __free_imsm(super
, 0);
1954 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
1955 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
1958 Name
": unable to allocate %zu byte mpb buffer\n",
1963 memcpy(super
->buf
, anchor
, 512);
1965 sectors
= mpb_sectors(anchor
) - 1;
1968 check_sum
= __gen_imsm_checksum(super
->anchor
);
1969 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
1972 Name
": IMSM checksum %x != %x on %s\n",
1974 __le32_to_cpu(super
->anchor
->check_sum
),
1982 /* read the extended mpb */
1983 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
1986 Name
": Cannot seek to extended mpb on %s: %s\n",
1987 devname
, strerror(errno
));
1991 if (read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
1994 Name
": Cannot read extended mpb on %s: %s\n",
1995 devname
, strerror(errno
));
1999 check_sum
= __gen_imsm_checksum(super
->anchor
);
2000 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
2003 Name
": IMSM checksum %x != %x on %s\n",
2004 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
2009 /* FIXME the BBM log is disk specific so we cannot use this global
2010 * buffer for all disks. Ok for now since we only look at the global
2011 * bbm_log_size parameter to gate assembly
2013 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
2019 load_and_parse_mpb(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
2023 err
= load_imsm_mpb(fd
, super
, devname
);
2026 err
= load_imsm_disk(fd
, super
, devname
, keep_fd
);
2029 err
= parse_raid_devices(super
);
2034 static void __free_imsm_disk(struct dl
*d
)
2045 static void free_imsm_disks(struct intel_super
*super
)
2049 while (super
->disks
) {
2051 super
->disks
= d
->next
;
2052 __free_imsm_disk(d
);
2054 while (super
->missing
) {
2056 super
->missing
= d
->next
;
2057 __free_imsm_disk(d
);
2062 /* free all the pieces hanging off of a super pointer */
2063 static void __free_imsm(struct intel_super
*super
, int free_disks
)
2070 free_imsm_disks(super
);
2071 free_devlist(super
);
2073 free((void *) super
->hba
);
2078 static void free_imsm(struct intel_super
*super
)
2080 __free_imsm(super
, 1);
2084 static void free_super_imsm(struct supertype
*st
)
2086 struct intel_super
*super
= st
->sb
;
2095 static struct intel_super
*alloc_super(int creating_imsm
)
2097 struct intel_super
*super
= malloc(sizeof(*super
));
2100 memset(super
, 0, sizeof(*super
));
2101 super
->creating_imsm
= creating_imsm
;
2102 super
->current_vol
= -1;
2103 super
->create_offset
= ~((__u32
) 0);
2104 if (!check_env("IMSM_NO_PLATFORM"))
2105 super
->orom
= find_imsm_orom();
2106 if (super
->orom
&& !check_env("IMSM_TEST_OROM")) {
2107 struct sys_dev
*list
, *ent
;
2109 /* find the first intel ahci controller */
2110 list
= find_driver_devices("pci", "ahci");
2111 for (ent
= list
; ent
; ent
= ent
->next
)
2112 if (devpath_to_vendor(ent
->path
) == 0x8086)
2115 super
->hba
= ent
->path
;
2118 free_sys_dev(&list
);
2126 /* find_missing - helper routine for load_super_imsm_all that identifies
2127 * disks that have disappeared from the system. This routine relies on
2128 * the mpb being uptodate, which it is at load time.
2130 static int find_missing(struct intel_super
*super
)
2133 struct imsm_super
*mpb
= super
->anchor
;
2135 struct imsm_disk
*disk
;
2137 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2138 disk
= __get_imsm_disk(mpb
, i
);
2139 dl
= serial_to_dl(disk
->serial
, super
);
2143 dl
= malloc(sizeof(*dl
));
2149 dl
->devname
= strdup("missing");
2151 serialcpy(dl
->serial
, disk
->serial
);
2154 dl
->next
= super
->missing
;
2155 super
->missing
= dl
;
2161 static struct intel_disk
*disk_list_get(__u8
*serial
, struct intel_disk
*disk_list
)
2163 struct intel_disk
*idisk
= disk_list
;
2166 if (serialcmp(idisk
->disk
.serial
, serial
) == 0)
2168 idisk
= idisk
->next
;
2174 static int __prep_thunderdome(struct intel_super
**table
, int tbl_size
,
2175 struct intel_super
*super
,
2176 struct intel_disk
**disk_list
)
2178 struct imsm_disk
*d
= &super
->disks
->disk
;
2179 struct imsm_super
*mpb
= super
->anchor
;
2182 for (i
= 0; i
< tbl_size
; i
++) {
2183 struct imsm_super
*tbl_mpb
= table
[i
]->anchor
;
2184 struct imsm_disk
*tbl_d
= &table
[i
]->disks
->disk
;
2186 if (tbl_mpb
->family_num
== mpb
->family_num
) {
2187 if (tbl_mpb
->check_sum
== mpb
->check_sum
) {
2188 dprintf("%s: mpb from %d:%d matches %d:%d\n",
2189 __func__
, super
->disks
->major
,
2190 super
->disks
->minor
,
2191 table
[i
]->disks
->major
,
2192 table
[i
]->disks
->minor
);
2196 if (((is_configured(d
) && !is_configured(tbl_d
)) ||
2197 is_configured(d
) == is_configured(tbl_d
)) &&
2198 tbl_mpb
->generation_num
< mpb
->generation_num
) {
2199 /* current version of the mpb is a
2200 * better candidate than the one in
2201 * super_table, but copy over "cross
2202 * generational" status
2204 struct intel_disk
*idisk
;
2206 dprintf("%s: mpb from %d:%d replaces %d:%d\n",
2207 __func__
, super
->disks
->major
,
2208 super
->disks
->minor
,
2209 table
[i
]->disks
->major
,
2210 table
[i
]->disks
->minor
);
2212 idisk
= disk_list_get(tbl_d
->serial
, *disk_list
);
2213 if (idisk
&& is_failed(&idisk
->disk
))
2214 tbl_d
->status
|= FAILED_DISK
;
2217 struct intel_disk
*idisk
;
2218 struct imsm_disk
*disk
;
2220 /* tbl_mpb is more up to date, but copy
2221 * over cross generational status before
2224 disk
= __serial_to_disk(d
->serial
, mpb
, NULL
);
2225 if (disk
&& is_failed(disk
))
2226 d
->status
|= FAILED_DISK
;
2228 idisk
= disk_list_get(d
->serial
, *disk_list
);
2231 if (disk
&& is_configured(disk
))
2232 idisk
->disk
.status
|= CONFIGURED_DISK
;
2235 dprintf("%s: mpb from %d:%d prefer %d:%d\n",
2236 __func__
, super
->disks
->major
,
2237 super
->disks
->minor
,
2238 table
[i
]->disks
->major
,
2239 table
[i
]->disks
->minor
);
2247 table
[tbl_size
++] = super
;
2251 /* update/extend the merged list of imsm_disk records */
2252 for (j
= 0; j
< mpb
->num_disks
; j
++) {
2253 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, j
);
2254 struct intel_disk
*idisk
;
2256 idisk
= disk_list_get(disk
->serial
, *disk_list
);
2258 idisk
->disk
.status
|= disk
->status
;
2259 if (is_configured(&idisk
->disk
) ||
2260 is_failed(&idisk
->disk
))
2261 idisk
->disk
.status
&= ~(SPARE_DISK
);
2263 idisk
= calloc(1, sizeof(*idisk
));
2266 idisk
->owner
= IMSM_UNKNOWN_OWNER
;
2267 idisk
->disk
= *disk
;
2268 idisk
->next
= *disk_list
;
2272 if (serialcmp(idisk
->disk
.serial
, d
->serial
) == 0)
2279 static struct intel_super
*
2280 validate_members(struct intel_super
*super
, struct intel_disk
*disk_list
,
2283 struct imsm_super
*mpb
= super
->anchor
;
2287 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2288 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
2289 struct intel_disk
*idisk
;
2291 idisk
= disk_list_get(disk
->serial
, disk_list
);
2293 if (idisk
->owner
== owner
||
2294 idisk
->owner
== IMSM_UNKNOWN_OWNER
)
2297 dprintf("%s: '%.16s' owner %d != %d\n",
2298 __func__
, disk
->serial
, idisk
->owner
,
2301 dprintf("%s: unknown disk %x [%d]: %.16s\n",
2302 __func__
, __le32_to_cpu(mpb
->family_num
), i
,
2308 if (ok_count
== mpb
->num_disks
)
2313 static void show_conflicts(__u32 family_num
, struct intel_super
*super_list
)
2315 struct intel_super
*s
;
2317 for (s
= super_list
; s
; s
= s
->next
) {
2318 if (family_num
!= s
->anchor
->family_num
)
2320 fprintf(stderr
, "Conflict, offlining family %#x on '%s'\n",
2321 __le32_to_cpu(family_num
), s
->disks
->devname
);
2325 static struct intel_super
*
2326 imsm_thunderdome(struct intel_super
**super_list
, int len
)
2328 struct intel_super
*super_table
[len
];
2329 struct intel_disk
*disk_list
= NULL
;
2330 struct intel_super
*champion
, *spare
;
2331 struct intel_super
*s
, **del
;
2336 memset(super_table
, 0, sizeof(super_table
));
2337 for (s
= *super_list
; s
; s
= s
->next
)
2338 tbl_size
= __prep_thunderdome(super_table
, tbl_size
, s
, &disk_list
);
2340 for (i
= 0; i
< tbl_size
; i
++) {
2341 struct imsm_disk
*d
;
2342 struct intel_disk
*idisk
;
2343 struct imsm_super
*mpb
= super_table
[i
]->anchor
;
2346 d
= &s
->disks
->disk
;
2348 /* 'd' must appear in merged disk list for its
2349 * configuration to be valid
2351 idisk
= disk_list_get(d
->serial
, disk_list
);
2352 if (idisk
&& idisk
->owner
== i
)
2353 s
= validate_members(s
, disk_list
, i
);
2358 dprintf("%s: marking family: %#x from %d:%d offline\n",
2359 __func__
, mpb
->family_num
,
2360 super_table
[i
]->disks
->major
,
2361 super_table
[i
]->disks
->minor
);
2365 /* This is where the mdadm implementation differs from the Windows
2366 * driver which has no strict concept of a container. We can only
2367 * assemble one family from a container, so when returning a prodigal
2368 * array member to this system the code will not be able to disambiguate
2369 * the container contents that should be assembled ("foreign" versus
2370 * "local"). It requires user intervention to set the orig_family_num
2371 * to a new value to establish a new container. The Windows driver in
2372 * this situation fixes up the volume name in place and manages the
2373 * foreign array as an independent entity.
2378 for (i
= 0; i
< tbl_size
; i
++) {
2379 struct intel_super
*tbl_ent
= super_table
[i
];
2385 if (tbl_ent
->anchor
->num_raid_devs
== 0) {
2390 if (s
&& !is_spare
) {
2391 show_conflicts(tbl_ent
->anchor
->family_num
, *super_list
);
2393 } else if (!s
&& !is_spare
)
2406 fprintf(stderr
, "Chose family %#x on '%s', "
2407 "assemble conflicts to new container with '--update=uuid'\n",
2408 __le32_to_cpu(s
->anchor
->family_num
), s
->disks
->devname
);
2410 /* collect all dl's onto 'champion', and update them to
2411 * champion's version of the status
2413 for (s
= *super_list
; s
; s
= s
->next
) {
2414 struct imsm_super
*mpb
= champion
->anchor
;
2415 struct dl
*dl
= s
->disks
;
2420 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2421 struct imsm_disk
*disk
;
2423 disk
= __serial_to_disk(dl
->serial
, mpb
, &dl
->index
);
2426 /* only set index on disks that are a member of
2427 * a populated contianer, i.e. one with
2430 if (is_failed(&dl
->disk
))
2432 else if (is_spare(&dl
->disk
))
2438 if (i
>= mpb
->num_disks
) {
2439 struct intel_disk
*idisk
;
2441 idisk
= disk_list_get(dl
->serial
, disk_list
);
2442 if (is_spare(&idisk
->disk
) &&
2443 !is_failed(&idisk
->disk
) && !is_configured(&idisk
->disk
))
2451 dl
->next
= champion
->disks
;
2452 champion
->disks
= dl
;
2456 /* delete 'champion' from super_list */
2457 for (del
= super_list
; *del
; ) {
2458 if (*del
== champion
) {
2459 *del
= (*del
)->next
;
2462 del
= &(*del
)->next
;
2464 champion
->next
= NULL
;
2468 struct intel_disk
*idisk
= disk_list
;
2470 disk_list
= disk_list
->next
;
2477 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
2478 char *devname
, int keep_fd
)
2481 struct intel_super
*super_list
= NULL
;
2482 struct intel_super
*super
= NULL
;
2483 int devnum
= fd2devnum(fd
);
2488 enum sysfs_read_flags flags
;
2490 flags
= GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
;
2491 if (mdmon_running(devnum
))
2492 flags
|= SKIP_GONE_DEVS
;
2494 /* check if 'fd' an opened container */
2495 sra
= sysfs_read(fd
, 0, flags
);
2499 if (sra
->array
.major_version
!= -1 ||
2500 sra
->array
.minor_version
!= -2 ||
2501 strcmp(sra
->text_version
, "imsm") != 0) {
2506 for (sd
= sra
->devs
, i
= 0; sd
; sd
= sd
->next
, i
++) {
2507 struct intel_super
*s
= alloc_super(0);
2514 s
->next
= super_list
;
2518 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2519 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
2523 err
= load_and_parse_mpb(dfd
, s
, NULL
, keep_fd
);
2525 /* retry the load if we might have raced against mdmon */
2526 if (err
== 3 && mdmon_running(devnum
))
2527 for (retry
= 0; retry
< 3; retry
++) {
2529 err
= load_and_parse_mpb(dfd
, s
, NULL
, keep_fd
);
2539 /* all mpbs enter, maybe one leaves */
2540 super
= imsm_thunderdome(&super_list
, i
);
2546 if (find_missing(super
) != 0) {
2552 if (st
->subarray
[0]) {
2553 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
2554 super
->current_vol
= atoi(st
->subarray
);
2564 while (super_list
) {
2565 struct intel_super
*s
= super_list
;
2567 super_list
= super_list
->next
;
2576 st
->container_dev
= devnum
;
2577 if (err
== 0 && st
->ss
== NULL
) {
2578 st
->ss
= &super_imsm
;
2579 st
->minor_version
= 0;
2580 st
->max_devs
= IMSM_MAX_DEVICES
;
2582 st
->loaded_container
= 1;
2588 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
2590 struct intel_super
*super
;
2594 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
2598 free_super_imsm(st
);
2600 super
= alloc_super(0);
2603 Name
": malloc of %zu failed.\n",
2608 rv
= load_and_parse_mpb(fd
, super
, devname
, 0);
2613 Name
": Failed to load all information "
2614 "sections on %s\n", devname
);
2619 if (st
->subarray
[0]) {
2620 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
2621 super
->current_vol
= atoi(st
->subarray
);
2629 if (st
->ss
== NULL
) {
2630 st
->ss
= &super_imsm
;
2631 st
->minor_version
= 0;
2632 st
->max_devs
= IMSM_MAX_DEVICES
;
2634 st
->loaded_container
= 0;
2639 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
2641 if (info
->level
== 1)
2643 return info
->chunk_size
>> 9;
2646 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
, int num_domains
)
2650 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
2651 num_stripes
/= num_domains
;
2656 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
2658 if (info
->level
== 1)
2659 return info
->size
* 2;
2661 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
2664 static void imsm_update_version_info(struct intel_super
*super
)
2666 /* update the version and attributes */
2667 struct imsm_super
*mpb
= super
->anchor
;
2669 struct imsm_dev
*dev
;
2670 struct imsm_map
*map
;
2673 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2674 dev
= get_imsm_dev(super
, i
);
2675 map
= get_imsm_map(dev
, 0);
2676 if (__le32_to_cpu(dev
->size_high
) > 0)
2677 mpb
->attributes
|= MPB_ATTRIB_2TB
;
2679 /* FIXME detect when an array spans a port multiplier */
2681 mpb
->attributes
|= MPB_ATTRIB_PM
;
2684 if (mpb
->num_raid_devs
> 1 ||
2685 mpb
->attributes
!= MPB_ATTRIB_CHECKSUM_VERIFY
) {
2686 version
= MPB_VERSION_ATTRIBS
;
2687 switch (get_imsm_raid_level(map
)) {
2688 case 0: mpb
->attributes
|= MPB_ATTRIB_RAID0
; break;
2689 case 1: mpb
->attributes
|= MPB_ATTRIB_RAID1
; break;
2690 case 10: mpb
->attributes
|= MPB_ATTRIB_RAID10
; break;
2691 case 5: mpb
->attributes
|= MPB_ATTRIB_RAID5
; break;
2694 if (map
->num_members
>= 5)
2695 version
= MPB_VERSION_5OR6_DISK_ARRAY
;
2696 else if (dev
->status
== DEV_CLONE_N_GO
)
2697 version
= MPB_VERSION_CNG
;
2698 else if (get_imsm_raid_level(map
) == 5)
2699 version
= MPB_VERSION_RAID5
;
2700 else if (map
->num_members
>= 3)
2701 version
= MPB_VERSION_3OR4_DISK_ARRAY
;
2702 else if (get_imsm_raid_level(map
) == 1)
2703 version
= MPB_VERSION_RAID1
;
2705 version
= MPB_VERSION_RAID0
;
2707 strcpy(((char *) mpb
->sig
) + strlen(MPB_SIGNATURE
), version
);
2711 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
2712 unsigned long long size
, char *name
,
2713 char *homehost
, int *uuid
)
2715 /* We are creating a volume inside a pre-existing container.
2716 * so st->sb is already set.
2718 struct intel_super
*super
= st
->sb
;
2719 struct imsm_super
*mpb
= super
->anchor
;
2720 struct intel_dev
*dv
;
2721 struct imsm_dev
*dev
;
2722 struct imsm_vol
*vol
;
2723 struct imsm_map
*map
;
2724 int idx
= mpb
->num_raid_devs
;
2726 unsigned long long array_blocks
;
2727 size_t size_old
, size_new
;
2728 __u32 num_data_stripes
;
2730 if (super
->orom
&& mpb
->num_raid_devs
>= super
->orom
->vpa
) {
2731 fprintf(stderr
, Name
": This imsm-container already has the "
2732 "maximum of %d volumes\n", super
->orom
->vpa
);
2736 /* ensure the mpb is large enough for the new data */
2737 size_old
= __le32_to_cpu(mpb
->mpb_size
);
2738 size_new
= disks_to_mpb_size(info
->nr_disks
);
2739 if (size_new
> size_old
) {
2741 size_t size_round
= ROUND_UP(size_new
, 512);
2743 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
2744 fprintf(stderr
, Name
": could not allocate new mpb\n");
2747 memcpy(mpb_new
, mpb
, size_old
);
2750 super
->anchor
= mpb_new
;
2751 mpb
->mpb_size
= __cpu_to_le32(size_new
);
2752 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
2754 super
->current_vol
= idx
;
2755 /* when creating the first raid device in this container set num_disks
2756 * to zero, i.e. delete this spare and add raid member devices in
2757 * add_to_super_imsm_volume()
2759 if (super
->current_vol
== 0)
2762 for (i
= 0; i
< super
->current_vol
; i
++) {
2763 dev
= get_imsm_dev(super
, i
);
2764 if (strncmp((char *) dev
->volume
, name
,
2765 MAX_RAID_SERIAL_LEN
) == 0) {
2766 fprintf(stderr
, Name
": '%s' is already defined for this container\n",
2772 sprintf(st
->subarray
, "%d", idx
);
2773 dv
= malloc(sizeof(*dv
));
2775 fprintf(stderr
, Name
": failed to allocate device list entry\n");
2778 dev
= malloc(sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
2781 fprintf(stderr
, Name
": could not allocate raid device\n");
2784 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
2785 if (info
->level
== 1)
2786 array_blocks
= info_to_blocks_per_member(info
);
2788 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
2789 info
->layout
, info
->chunk_size
,
2791 /* round array size down to closest MB */
2792 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
2794 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
2795 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
2796 dev
->status
= __cpu_to_le32(0);
2797 dev
->reserved_blocks
= __cpu_to_le32(0);
2799 vol
->migr_state
= 0;
2800 set_migr_type(dev
, MIGR_INIT
);
2802 vol
->curr_migr_unit
= 0;
2803 map
= get_imsm_map(dev
, 0);
2804 map
->pba_of_lba0
= __cpu_to_le32(super
->create_offset
);
2805 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
2806 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
2807 map
->failed_disk_num
= ~0;
2808 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
2809 IMSM_T_STATE_NORMAL
;
2812 if (info
->level
== 1 && info
->raid_disks
> 2) {
2813 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
2814 "in a raid1 volume\n");
2818 map
->raid_level
= info
->level
;
2819 if (info
->level
== 10) {
2820 map
->raid_level
= 1;
2821 map
->num_domains
= info
->raid_disks
/ 2;
2822 } else if (info
->level
== 1)
2823 map
->num_domains
= info
->raid_disks
;
2825 map
->num_domains
= 1;
2827 num_data_stripes
= info_to_num_data_stripes(info
, map
->num_domains
);
2828 map
->num_data_stripes
= __cpu_to_le32(num_data_stripes
);
2830 map
->num_members
= info
->raid_disks
;
2831 for (i
= 0; i
< map
->num_members
; i
++) {
2832 /* initialized in add_to_super */
2833 set_imsm_ord_tbl_ent(map
, i
, 0);
2835 mpb
->num_raid_devs
++;
2838 dv
->index
= super
->current_vol
;
2839 dv
->next
= super
->devlist
;
2840 super
->devlist
= dv
;
2842 imsm_update_version_info(super
);
2847 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
2848 unsigned long long size
, char *name
,
2849 char *homehost
, int *uuid
)
2851 /* This is primarily called by Create when creating a new array.
2852 * We will then get add_to_super called for each component, and then
2853 * write_init_super called to write it out to each device.
2854 * For IMSM, Create can create on fresh devices or on a pre-existing
2856 * To create on a pre-existing array a different method will be called.
2857 * This one is just for fresh drives.
2859 struct intel_super
*super
;
2860 struct imsm_super
*mpb
;
2865 return init_super_imsm_volume(st
, info
, size
, name
, homehost
, uuid
);
2868 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
2872 super
= alloc_super(1);
2873 if (super
&& posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
2878 fprintf(stderr
, Name
2879 ": %s could not allocate superblock\n", __func__
);
2882 memset(super
->buf
, 0, mpb_size
);
2884 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
2888 /* zeroing superblock */
2892 mpb
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
2894 version
= (char *) mpb
->sig
;
2895 strcpy(version
, MPB_SIGNATURE
);
2896 version
+= strlen(MPB_SIGNATURE
);
2897 strcpy(version
, MPB_VERSION_RAID0
);
2903 static int add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
2904 int fd
, char *devname
)
2906 struct intel_super
*super
= st
->sb
;
2907 struct imsm_super
*mpb
= super
->anchor
;
2909 struct imsm_dev
*dev
;
2910 struct imsm_map
*map
;
2912 dev
= get_imsm_dev(super
, super
->current_vol
);
2913 map
= get_imsm_map(dev
, 0);
2915 if (! (dk
->state
& (1<<MD_DISK_SYNC
))) {
2916 fprintf(stderr
, Name
": %s: Cannot add spare devices to IMSM volume\n",
2922 /* we're doing autolayout so grab the pre-marked (in
2923 * validate_geometry) raid_disk
2925 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2926 if (dl
->raiddisk
== dk
->raid_disk
)
2929 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2930 if (dl
->major
== dk
->major
&&
2931 dl
->minor
== dk
->minor
)
2936 fprintf(stderr
, Name
": %s is not a member of the same container\n", devname
);
2940 /* add a pristine spare to the metadata */
2941 if (dl
->index
< 0) {
2942 dl
->index
= super
->anchor
->num_disks
;
2943 super
->anchor
->num_disks
++;
2945 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
2946 dl
->disk
.status
= CONFIGURED_DISK
;
2948 /* if we are creating the first raid device update the family number */
2949 if (super
->current_vol
== 0) {
2951 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
2952 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
2954 if (!_dev
|| !_disk
) {
2955 fprintf(stderr
, Name
": BUG mpb setup error\n");
2961 sum
+= __gen_imsm_checksum(mpb
);
2962 mpb
->family_num
= __cpu_to_le32(sum
);
2963 mpb
->orig_family_num
= mpb
->family_num
;
2969 static int add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
2970 int fd
, char *devname
)
2972 struct intel_super
*super
= st
->sb
;
2974 unsigned long long size
;
2979 /* if we are on an RAID enabled platform check that the disk is
2980 * attached to the raid controller
2982 if (super
->hba
&& !disk_attached_to_hba(fd
, super
->hba
)) {
2984 Name
": %s is not attached to the raid controller: %s\n",
2985 devname
? : "disk", super
->hba
);
2989 if (super
->current_vol
>= 0)
2990 return add_to_super_imsm_volume(st
, dk
, fd
, devname
);
2993 dd
= malloc(sizeof(*dd
));
2996 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
2999 memset(dd
, 0, sizeof(*dd
));
3000 dd
->major
= major(stb
.st_rdev
);
3001 dd
->minor
= minor(stb
.st_rdev
);
3003 dd
->devname
= devname
? strdup(devname
) : NULL
;
3006 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
3009 Name
": failed to retrieve scsi serial, aborting\n");
3014 get_dev_size(fd
, NULL
, &size
);
3016 serialcpy(dd
->disk
.serial
, dd
->serial
);
3017 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
3018 dd
->disk
.status
= SPARE_DISK
;
3019 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
3020 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
3022 dd
->disk
.scsi_id
= __cpu_to_le32(0);
3024 if (st
->update_tail
) {
3025 dd
->next
= super
->add
;
3028 dd
->next
= super
->disks
;
3035 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
);
3039 struct imsm_super anchor
;
3040 } spare_record
__attribute__ ((aligned(512)));
3042 /* spare records have their own family number and do not have any defined raid
3045 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
3047 struct imsm_super
*mpb
= super
->anchor
;
3048 struct imsm_super
*spare
= &spare_record
.anchor
;
3052 spare
->mpb_size
= __cpu_to_le32(sizeof(struct imsm_super
)),
3053 spare
->generation_num
= __cpu_to_le32(1UL),
3054 spare
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
3055 spare
->num_disks
= 1,
3056 spare
->num_raid_devs
= 0,
3057 spare
->cache_size
= mpb
->cache_size
,
3058 spare
->pwr_cycle_count
= __cpu_to_le32(1),
3060 snprintf((char *) spare
->sig
, MAX_SIGNATURE_LENGTH
,
3061 MPB_SIGNATURE MPB_VERSION_RAID0
);
3063 for (d
= super
->disks
; d
; d
= d
->next
) {
3067 spare
->disk
[0] = d
->disk
;
3068 sum
= __gen_imsm_checksum(spare
);
3069 spare
->family_num
= __cpu_to_le32(sum
);
3070 spare
->orig_family_num
= 0;
3071 sum
= __gen_imsm_checksum(spare
);
3072 spare
->check_sum
= __cpu_to_le32(sum
);
3074 if (store_imsm_mpb(d
->fd
, spare
)) {
3075 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
3076 __func__
, d
->major
, d
->minor
, strerror(errno
));
3088 static int write_super_imsm(struct intel_super
*super
, int doclose
)
3090 struct imsm_super
*mpb
= super
->anchor
;
3096 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
3098 /* 'generation' is incremented everytime the metadata is written */
3099 generation
= __le32_to_cpu(mpb
->generation_num
);
3101 mpb
->generation_num
= __cpu_to_le32(generation
);
3103 /* fix up cases where previous mdadm releases failed to set
3106 if (mpb
->orig_family_num
== 0)
3107 mpb
->orig_family_num
= mpb
->family_num
;
3109 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
3110 for (d
= super
->disks
; d
; d
= d
->next
) {
3114 mpb
->disk
[d
->index
] = d
->disk
;
3116 for (d
= super
->missing
; d
; d
= d
->next
)
3117 mpb
->disk
[d
->index
] = d
->disk
;
3119 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3120 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
3122 imsm_copy_dev(dev
, get_imsm_dev(super
, i
));
3123 mpb_size
+= sizeof_imsm_dev(dev
, 0);
3125 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
3126 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
3128 /* recalculate checksum */
3129 sum
= __gen_imsm_checksum(mpb
);
3130 mpb
->check_sum
= __cpu_to_le32(sum
);
3132 /* write the mpb for disks that compose raid devices */
3133 for (d
= super
->disks
; d
; d
= d
->next
) {
3136 if (store_imsm_mpb(d
->fd
, mpb
))
3137 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
3138 __func__
, d
->major
, d
->minor
, strerror(errno
));
3146 return write_super_imsm_spares(super
, doclose
);
3152 static int create_array(struct supertype
*st
, int dev_idx
)
3155 struct imsm_update_create_array
*u
;
3156 struct intel_super
*super
= st
->sb
;
3157 struct imsm_dev
*dev
= get_imsm_dev(super
, dev_idx
);
3158 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3159 struct disk_info
*inf
;
3160 struct imsm_disk
*disk
;
3163 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0) +
3164 sizeof(*inf
) * map
->num_members
;
3167 fprintf(stderr
, "%s: failed to allocate update buffer\n",
3172 u
->type
= update_create_array
;
3173 u
->dev_idx
= dev_idx
;
3174 imsm_copy_dev(&u
->dev
, dev
);
3175 inf
= get_disk_info(u
);
3176 for (i
= 0; i
< map
->num_members
; i
++) {
3177 int idx
= get_imsm_disk_idx(dev
, i
);
3179 disk
= get_imsm_disk(super
, idx
);
3180 serialcpy(inf
[i
].serial
, disk
->serial
);
3182 append_metadata_update(st
, u
, len
);
3187 static int _add_disk(struct supertype
*st
)
3189 struct intel_super
*super
= st
->sb
;
3191 struct imsm_update_add_disk
*u
;
3199 fprintf(stderr
, "%s: failed to allocate update buffer\n",
3204 u
->type
= update_add_disk
;
3205 append_metadata_update(st
, u
, len
);
3210 static int write_init_super_imsm(struct supertype
*st
)
3212 struct intel_super
*super
= st
->sb
;
3213 int current_vol
= super
->current_vol
;
3215 /* we are done with current_vol reset it to point st at the container */
3216 super
->current_vol
= -1;
3218 if (st
->update_tail
) {
3219 /* queue the recently created array / added disk
3220 * as a metadata update */
3224 /* determine if we are creating a volume or adding a disk */
3225 if (current_vol
< 0) {
3226 /* in the add disk case we are running in mdmon
3227 * context, so don't close fd's
3229 return _add_disk(st
);
3231 rv
= create_array(st
, current_vol
);
3233 for (d
= super
->disks
; d
; d
= d
->next
) {
3240 return write_super_imsm(st
->sb
, 1);
3244 static int store_super_imsm(struct supertype
*st
, int fd
)
3246 struct intel_super
*super
= st
->sb
;
3247 struct imsm_super
*mpb
= super
? super
->anchor
: NULL
;
3253 return store_imsm_mpb(fd
, mpb
);
3259 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
3261 return __le32_to_cpu(mpb
->bbm_log_size
);
3265 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
3266 int layout
, int raiddisks
, int chunk
,
3267 unsigned long long size
, char *dev
,
3268 unsigned long long *freesize
,
3272 unsigned long long ldsize
;
3273 const struct imsm_orom
*orom
;
3275 if (level
!= LEVEL_CONTAINER
)
3280 if (check_env("IMSM_NO_PLATFORM"))
3283 orom
= find_imsm_orom();
3284 if (orom
&& raiddisks
> orom
->tds
) {
3286 fprintf(stderr
, Name
": %d exceeds maximum number of"
3287 " platform supported disks: %d\n",
3288 raiddisks
, orom
->tds
);
3292 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3295 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
3296 dev
, strerror(errno
));
3299 if (!get_dev_size(fd
, dev
, &ldsize
)) {
3305 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
3310 static unsigned long long find_size(struct extent
*e
, int *idx
, int num_extents
)
3312 const unsigned long long base_start
= e
[*idx
].start
;
3313 unsigned long long end
= base_start
+ e
[*idx
].size
;
3316 if (base_start
== end
)
3320 for (i
= *idx
; i
< num_extents
; i
++) {
3321 /* extend overlapping extents */
3322 if (e
[i
].start
>= base_start
&&
3323 e
[i
].start
<= end
) {
3326 if (e
[i
].start
+ e
[i
].size
> end
)
3327 end
= e
[i
].start
+ e
[i
].size
;
3328 } else if (e
[i
].start
> end
) {
3334 return end
- base_start
;
3337 static unsigned long long merge_extents(struct intel_super
*super
, int sum_extents
)
3339 /* build a composite disk with all known extents and generate a new
3340 * 'maxsize' given the "all disks in an array must share a common start
3341 * offset" constraint
3343 struct extent
*e
= calloc(sum_extents
, sizeof(*e
));
3347 unsigned long long pos
;
3348 unsigned long long start
= 0;
3349 unsigned long long maxsize
;
3350 unsigned long reserve
;
3353 return ~0ULL; /* error */
3355 /* coalesce and sort all extents. also, check to see if we need to
3356 * reserve space between member arrays
3359 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
3362 for (i
= 0; i
< dl
->extent_cnt
; i
++)
3365 qsort(e
, sum_extents
, sizeof(*e
), cmp_extent
);
3370 while (i
< sum_extents
) {
3371 e
[j
].start
= e
[i
].start
;
3372 e
[j
].size
= find_size(e
, &i
, sum_extents
);
3374 if (e
[j
-1].size
== 0)
3383 unsigned long long esize
;
3385 esize
= e
[i
].start
- pos
;
3386 if (esize
>= maxsize
) {
3391 pos
= e
[i
].start
+ e
[i
].size
;
3393 } while (e
[i
-1].size
);
3396 if (start_extent
> 0)
3397 reserve
= IMSM_RESERVED_SECTORS
; /* gap between raid regions */
3401 if (maxsize
< reserve
)
3404 super
->create_offset
= ~((__u32
) 0);
3405 if (start
+ reserve
> super
->create_offset
)
3406 return ~0ULL; /* start overflows create_offset */
3407 super
->create_offset
= start
+ reserve
;
3409 return maxsize
- reserve
;
3412 static int is_raid_level_supported(const struct imsm_orom
*orom
, int level
, int raiddisks
)
3414 if (level
< 0 || level
== 6 || level
== 4)
3417 /* if we have an orom prevent invalid raid levels */
3420 case 0: return imsm_orom_has_raid0(orom
);
3423 return imsm_orom_has_raid1e(orom
);
3424 return imsm_orom_has_raid1(orom
) && raiddisks
== 2;
3425 case 10: return imsm_orom_has_raid10(orom
) && raiddisks
== 4;
3426 case 5: return imsm_orom_has_raid5(orom
) && raiddisks
> 2;
3429 return 1; /* not on an Intel RAID platform so anything goes */
3434 #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
3435 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
3436 * FIX ME add ahci details
3438 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
3439 int layout
, int raiddisks
, int chunk
,
3440 unsigned long long size
, char *dev
,
3441 unsigned long long *freesize
,
3445 struct intel_super
*super
= st
->sb
;
3446 struct imsm_super
*mpb
;
3448 unsigned long long pos
= 0;
3449 unsigned long long maxsize
;
3453 /* We must have the container info already read in. */
3456 mpb
= super
->anchor
;
3458 if (!is_raid_level_supported(super
->orom
, level
, raiddisks
)) {
3459 pr_vrb(": platform does not support raid%d with %d disk%s\n",
3460 level
, raiddisks
, raiddisks
> 1 ? "s" : "");
3463 if (super
->orom
&& level
!= 1 &&
3464 !imsm_orom_has_chunk(super
->orom
, chunk
)) {
3465 pr_vrb(": platform does not support a chunk size of: %d\n", chunk
);
3468 if (layout
!= imsm_level_to_layout(level
)) {
3470 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
3471 else if (level
== 10)
3472 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
3474 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
3480 /* General test: make sure there is space for
3481 * 'raiddisks' device extents of size 'size' at a given
3484 unsigned long long minsize
= size
;
3485 unsigned long long start_offset
= ~0ULL;
3488 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
3489 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
3494 e
= get_extents(super
, dl
);
3497 unsigned long long esize
;
3498 esize
= e
[i
].start
- pos
;
3499 if (esize
>= minsize
)
3501 if (found
&& start_offset
== ~0ULL) {
3504 } else if (found
&& pos
!= start_offset
) {
3508 pos
= e
[i
].start
+ e
[i
].size
;
3510 } while (e
[i
-1].size
);
3515 if (dcnt
< raiddisks
) {
3517 fprintf(stderr
, Name
": imsm: Not enough "
3518 "devices with space for this array "
3526 /* This device must be a member of the set */
3527 if (stat(dev
, &stb
) < 0)
3529 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
3531 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
3532 if (dl
->major
== major(stb
.st_rdev
) &&
3533 dl
->minor
== minor(stb
.st_rdev
))
3538 fprintf(stderr
, Name
": %s is not in the "
3539 "same imsm set\n", dev
);
3541 } else if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
) {
3542 /* If a volume is present then the current creation attempt
3543 * cannot incorporate new spares because the orom may not
3544 * understand this configuration (all member disks must be
3545 * members of each array in the container).
3547 fprintf(stderr
, Name
": %s is a spare and a volume"
3548 " is already defined for this container\n", dev
);
3549 fprintf(stderr
, Name
": The option-rom requires all member"
3550 " disks to be a member of all volumes\n");
3554 /* retrieve the largest free space block */
3555 e
= get_extents(super
, dl
);
3560 unsigned long long esize
;
3562 esize
= e
[i
].start
- pos
;
3563 if (esize
>= maxsize
)
3565 pos
= e
[i
].start
+ e
[i
].size
;
3567 } while (e
[i
-1].size
);
3572 fprintf(stderr
, Name
": unable to determine free space for: %s\n",
3576 if (maxsize
< size
) {
3578 fprintf(stderr
, Name
": %s not enough space (%llu < %llu)\n",
3579 dev
, maxsize
, size
);
3583 /* count total number of extents for merge */
3585 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
3587 i
+= dl
->extent_cnt
;
3589 maxsize
= merge_extents(super
, i
);
3590 if (maxsize
< size
) {
3592 fprintf(stderr
, Name
": not enough space after merge (%llu < %llu)\n",
3595 } else if (maxsize
== ~0ULL) {
3597 fprintf(stderr
, Name
": failed to merge %d extents\n", i
);
3601 *freesize
= maxsize
;
3606 static int reserve_space(struct supertype
*st
, int raiddisks
,
3607 unsigned long long size
, int chunk
,
3608 unsigned long long *freesize
)
3610 struct intel_super
*super
= st
->sb
;
3611 struct imsm_super
*mpb
= super
->anchor
;
3616 unsigned long long maxsize
;
3617 unsigned long long minsize
;
3621 /* find the largest common start free region of the possible disks */
3625 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
3631 /* don't activate new spares if we are orom constrained
3632 * and there is already a volume active in the container
3634 if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
)
3637 e
= get_extents(super
, dl
);
3640 for (i
= 1; e
[i
-1].size
; i
++)
3648 maxsize
= merge_extents(super
, extent_cnt
);
3653 if (cnt
< raiddisks
||
3654 (super
->orom
&& used
&& used
!= raiddisks
) ||
3655 maxsize
< minsize
) {
3656 fprintf(stderr
, Name
": not enough devices with space to create array.\n");
3657 return 0; /* No enough free spaces large enough */
3669 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
3671 dl
->raiddisk
= cnt
++;
3678 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
3679 int raiddisks
, int chunk
, unsigned long long size
,
3680 char *dev
, unsigned long long *freesize
,
3687 /* if given unused devices create a container
3688 * if given given devices in a container create a member volume
3690 if (level
== LEVEL_CONTAINER
) {
3691 /* Must be a fresh device to add to a container */
3692 return validate_geometry_imsm_container(st
, level
, layout
,
3693 raiddisks
, chunk
, size
,
3699 if (st
->sb
&& freesize
) {
3700 /* we are being asked to automatically layout a
3701 * new volume based on the current contents of
3702 * the container. If the the parameters can be
3703 * satisfied reserve_space will record the disks,
3704 * start offset, and size of the volume to be
3705 * created. add_to_super and getinfo_super
3706 * detect when autolayout is in progress.
3708 return reserve_space(st
, raiddisks
, size
, chunk
, freesize
);
3713 /* creating in a given container */
3714 return validate_geometry_imsm_volume(st
, level
, layout
,
3715 raiddisks
, chunk
, size
,
3716 dev
, freesize
, verbose
);
3719 /* limit creation to the following levels */
3729 fprintf(stderr
, Name
3730 ": IMSM only supports levels 0,1,5,10\n");
3734 /* This device needs to be a device in an 'imsm' container */
3735 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
3739 Name
": Cannot create this array on device %s\n",
3744 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
3746 fprintf(stderr
, Name
": Cannot open %s: %s\n",
3747 dev
, strerror(errno
));
3750 /* Well, it is in use by someone, maybe an 'imsm' container. */
3751 cfd
= open_container(fd
);
3755 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
3759 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
3760 if (sra
&& sra
->array
.major_version
== -1 &&
3761 strcmp(sra
->text_version
, "imsm") == 0)
3765 /* This is a member of a imsm container. Load the container
3766 * and try to create a volume
3768 struct intel_super
*super
;
3770 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
3772 st
->container_dev
= fd2devnum(cfd
);
3774 return validate_geometry_imsm_volume(st
, level
, layout
,
3782 fprintf(stderr
, Name
": failed container membership check\n");
3787 #endif /* MDASSEMBLE */
3789 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
3791 /* Given a container loaded by load_super_imsm_all,
3792 * extract information about all the arrays into
3795 * For each imsm_dev create an mdinfo, fill it in,
3796 * then look for matching devices in super->disks
3797 * and create appropriate device mdinfo.
3799 struct intel_super
*super
= st
->sb
;
3800 struct imsm_super
*mpb
= super
->anchor
;
3801 struct mdinfo
*rest
= NULL
;
3804 /* do not assemble arrays that might have bad blocks */
3805 if (imsm_bbm_log_size(super
->anchor
)) {
3806 fprintf(stderr
, Name
": BBM log found in metadata. "
3807 "Cannot activate array(s).\n");
3811 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
3812 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
3813 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3814 struct mdinfo
*this;
3817 /* do not publish arrays that are in the middle of an
3818 * unsupported migration
3820 if (dev
->vol
.migr_state
&&
3821 (migr_type(dev
) == MIGR_GEN_MIGR
||
3822 migr_type(dev
) == MIGR_STATE_CHANGE
)) {
3823 fprintf(stderr
, Name
": cannot assemble volume '%.16s':"
3824 " unsupported migration in progress\n",
3829 this = malloc(sizeof(*this));
3831 fprintf(stderr
, Name
": failed to allocate %lu bytes\n",
3835 memset(this, 0, sizeof(*this));
3838 super
->current_vol
= i
;
3839 getinfo_super_imsm_volume(st
, this);
3840 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
3841 struct mdinfo
*info_d
;
3848 idx
= get_imsm_disk_idx(dev
, slot
);
3849 ord
= get_imsm_ord_tbl_ent(dev
, slot
);
3850 for (d
= super
->disks
; d
; d
= d
->next
)
3851 if (d
->index
== idx
)
3856 if (d
&& is_failed(&d
->disk
))
3858 if (ord
& IMSM_ORD_REBUILD
)
3862 * if we skip some disks the array will be assmebled degraded;
3863 * reset resync start to avoid a dirty-degraded situation
3865 * FIXME handle dirty degraded
3867 if (skip
&& !dev
->vol
.dirty
)
3868 this->resync_start
= ~0ULL;
3872 info_d
= malloc(sizeof(*info_d
));
3874 fprintf(stderr
, Name
": failed to allocate disk"
3875 " for volume %.16s\n", dev
->volume
);
3880 memset(info_d
, 0, sizeof(*info_d
));
3881 info_d
->next
= this->devs
;
3882 this->devs
= info_d
;
3884 info_d
->disk
.number
= d
->index
;
3885 info_d
->disk
.major
= d
->major
;
3886 info_d
->disk
.minor
= d
->minor
;
3887 info_d
->disk
.raid_disk
= slot
;
3889 this->array
.working_disks
++;
3891 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
3892 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
3893 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
3903 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
3906 struct intel_super
*super
= c
->sb
;
3907 struct imsm_super
*mpb
= super
->anchor
;
3909 if (atoi(inst
) >= mpb
->num_raid_devs
) {
3910 fprintf(stderr
, "%s: subarry index %d, out of range\n",
3911 __func__
, atoi(inst
));
3915 dprintf("imsm: open_new %s\n", inst
);
3916 a
->info
.container_member
= atoi(inst
);
3920 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
3922 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3925 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
3926 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
3928 switch (get_imsm_raid_level(map
)) {
3930 return IMSM_T_STATE_FAILED
;
3933 if (failed
< map
->num_members
)
3934 return IMSM_T_STATE_DEGRADED
;
3936 return IMSM_T_STATE_FAILED
;
3941 * check to see if any mirrors have failed, otherwise we
3942 * are degraded. Even numbered slots are mirrored on
3946 /* gcc -Os complains that this is unused */
3947 int insync
= insync
;
3949 for (i
= 0; i
< map
->num_members
; i
++) {
3950 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
);
3951 int idx
= ord_to_idx(ord
);
3952 struct imsm_disk
*disk
;
3954 /* reset the potential in-sync count on even-numbered
3955 * slots. num_copies is always 2 for imsm raid10
3960 disk
= get_imsm_disk(super
, idx
);
3961 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
3964 /* no in-sync disks left in this mirror the
3968 return IMSM_T_STATE_FAILED
;
3971 return IMSM_T_STATE_DEGRADED
;
3975 return IMSM_T_STATE_DEGRADED
;
3977 return IMSM_T_STATE_FAILED
;
3983 return map
->map_state
;
3986 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
3990 struct imsm_disk
*disk
;
3991 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3992 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
3996 /* at the beginning of migration we set IMSM_ORD_REBUILD on
3997 * disks that are being rebuilt. New failures are recorded to
3998 * map[0]. So we look through all the disks we started with and
3999 * see if any failures are still present, or if any new ones
4002 * FIXME add support for online capacity expansion and
4003 * raid-level-migration
4005 for (i
= 0; i
< prev
->num_members
; i
++) {
4006 ord
= __le32_to_cpu(prev
->disk_ord_tbl
[i
]);
4007 ord
|= __le32_to_cpu(map
->disk_ord_tbl
[i
]);
4008 idx
= ord_to_idx(ord
);
4010 disk
= get_imsm_disk(super
, idx
);
4011 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
4018 static int is_resyncing(struct imsm_dev
*dev
)
4020 struct imsm_map
*migr_map
;
4022 if (!dev
->vol
.migr_state
)
4025 if (migr_type(dev
) == MIGR_INIT
||
4026 migr_type(dev
) == MIGR_REPAIR
)
4029 migr_map
= get_imsm_map(dev
, 1);
4031 if (migr_map
->map_state
== IMSM_T_STATE_NORMAL
)
4037 static int is_rebuilding(struct imsm_dev
*dev
)
4039 struct imsm_map
*migr_map
;
4041 if (!dev
->vol
.migr_state
)
4044 if (migr_type(dev
) != MIGR_REBUILD
)
4047 migr_map
= get_imsm_map(dev
, 1);
4049 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
4055 /* return true if we recorded new information */
4056 static int mark_failure(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
4060 struct imsm_map
*map
;
4062 /* new failures are always set in map[0] */
4063 map
= get_imsm_map(dev
, 0);
4065 slot
= get_imsm_disk_slot(map
, idx
);
4069 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
4070 if (is_failed(disk
) && (ord
& IMSM_ORD_REBUILD
))
4073 disk
->status
|= FAILED_DISK
;
4074 disk
->status
&= ~CONFIGURED_DISK
;
4075 set_imsm_ord_tbl_ent(map
, slot
, idx
| IMSM_ORD_REBUILD
);
4076 if (~map
->failed_disk_num
== 0)
4077 map
->failed_disk_num
= slot
;
4081 static void mark_missing(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
4083 mark_failure(dev
, disk
, idx
);
4085 if (disk
->scsi_id
== __cpu_to_le32(~(__u32
)0))
4088 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
4089 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
4092 /* Handle dirty -> clean transititions and resync. Degraded and rebuild
4093 * states are handled in imsm_set_disk() with one exception, when a
4094 * resync is stopped due to a new failure this routine will set the
4095 * 'degraded' state for the array.
4097 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
4099 int inst
= a
->info
.container_member
;
4100 struct intel_super
*super
= a
->container
->sb
;
4101 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
4102 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4103 int failed
= imsm_count_failed(super
, dev
);
4104 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
4106 /* before we activate this array handle any missing disks */
4107 if (consistent
== 2 && super
->missing
) {
4110 dprintf("imsm: mark missing\n");
4111 end_migration(dev
, map_state
);
4112 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
4113 mark_missing(dev
, &dl
->disk
, dl
->index
);
4114 super
->updates_pending
++;
4117 if (consistent
== 2 &&
4118 (!is_resync_complete(a
) ||
4119 map_state
!= IMSM_T_STATE_NORMAL
||
4120 dev
->vol
.migr_state
))
4123 if (is_resync_complete(a
)) {
4124 /* complete intialization / resync,
4125 * recovery and interrupted recovery is completed in
4128 if (is_resyncing(dev
)) {
4129 dprintf("imsm: mark resync done\n");
4130 end_migration(dev
, map_state
);
4131 super
->updates_pending
++;
4133 } else if (!is_resyncing(dev
) && !failed
) {
4134 /* mark the start of the init process if nothing is failed */
4135 dprintf("imsm: mark resync start (%llu)\n", a
->resync_start
);
4136 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
)
4137 migrate(dev
, IMSM_T_STATE_NORMAL
, MIGR_INIT
);
4139 migrate(dev
, IMSM_T_STATE_NORMAL
, MIGR_REPAIR
);
4140 super
->updates_pending
++;
4143 /* FIXME check if we can update curr_migr_unit from resync_start */
4145 /* mark dirty / clean */
4146 if (dev
->vol
.dirty
!= !consistent
) {
4147 dprintf("imsm: mark '%s' (%llu)\n",
4148 consistent
? "clean" : "dirty", a
->resync_start
);
4153 super
->updates_pending
++;
4158 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
4160 int inst
= a
->info
.container_member
;
4161 struct intel_super
*super
= a
->container
->sb
;
4162 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
4163 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4164 struct imsm_disk
*disk
;
4169 if (n
> map
->num_members
)
4170 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
4171 n
, map
->num_members
- 1);
4176 dprintf("imsm: set_disk %d:%x\n", n
, state
);
4178 ord
= get_imsm_ord_tbl_ent(dev
, n
);
4179 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
4181 /* check for new failures */
4182 if (state
& DS_FAULTY
) {
4183 if (mark_failure(dev
, disk
, ord_to_idx(ord
)))
4184 super
->updates_pending
++;
4187 /* check if in_sync */
4188 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
&& is_rebuilding(dev
)) {
4189 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
4191 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
4192 super
->updates_pending
++;
4195 failed
= imsm_count_failed(super
, dev
);
4196 map_state
= imsm_check_degraded(super
, dev
, failed
);
4198 /* check if recovery complete, newly degraded, or failed */
4199 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
4200 end_migration(dev
, map_state
);
4201 map
= get_imsm_map(dev
, 0);
4202 map
->failed_disk_num
= ~0;
4203 super
->updates_pending
++;
4204 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
4205 map
->map_state
!= map_state
&&
4206 !dev
->vol
.migr_state
) {
4207 dprintf("imsm: mark degraded\n");
4208 map
->map_state
= map_state
;
4209 super
->updates_pending
++;
4210 } else if (map_state
== IMSM_T_STATE_FAILED
&&
4211 map
->map_state
!= map_state
) {
4212 dprintf("imsm: mark failed\n");
4213 end_migration(dev
, map_state
);
4214 super
->updates_pending
++;
4218 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
)
4221 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
4222 unsigned long long dsize
;
4223 unsigned long long sectors
;
4225 get_dev_size(fd
, NULL
, &dsize
);
4227 if (mpb_size
> 512) {
4228 /* -1 to account for anchor */
4229 sectors
= mpb_sectors(mpb
) - 1;
4231 /* write the extended mpb to the sectors preceeding the anchor */
4232 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
4235 if (write(fd
, buf
+ 512, 512 * sectors
) != 512 * sectors
)
4239 /* first block is stored on second to last sector of the disk */
4240 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
4243 if (write(fd
, buf
, 512) != 512)
4249 static void imsm_sync_metadata(struct supertype
*container
)
4251 struct intel_super
*super
= container
->sb
;
4253 if (!super
->updates_pending
)
4256 write_super_imsm(super
, 0);
4258 super
->updates_pending
= 0;
4261 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
4263 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
4264 int i
= get_imsm_disk_idx(dev
, idx
);
4267 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4271 if (dl
&& is_failed(&dl
->disk
))
4275 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
4280 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
,
4281 struct active_array
*a
, int activate_new
)
4283 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
4284 int idx
= get_imsm_disk_idx(dev
, slot
);
4285 struct imsm_super
*mpb
= super
->anchor
;
4286 struct imsm_map
*map
;
4287 unsigned long long pos
;
4296 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
4297 /* If in this array, skip */
4298 for (d
= a
->info
.devs
; d
; d
= d
->next
)
4299 if (d
->state_fd
>= 0 &&
4300 d
->disk
.major
== dl
->major
&&
4301 d
->disk
.minor
== dl
->minor
) {
4302 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
4308 /* skip in use or failed drives */
4309 if (is_failed(&dl
->disk
) || idx
== dl
->index
||
4311 dprintf("%x:%x status (failed: %d index: %d)\n",
4312 dl
->major
, dl
->minor
, is_failed(&dl
->disk
), idx
);
4316 /* skip pure spares when we are looking for partially
4317 * assimilated drives
4319 if (dl
->index
== -1 && !activate_new
)
4322 /* Does this unused device have the requisite free space?
4323 * It needs to be able to cover all member volumes
4325 ex
= get_extents(super
, dl
);
4327 dprintf("cannot get extents\n");
4330 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4331 dev
= get_imsm_dev(super
, i
);
4332 map
= get_imsm_map(dev
, 0);
4334 /* check if this disk is already a member of
4337 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
4343 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
4344 array_end
= array_start
+
4345 __le32_to_cpu(map
->blocks_per_member
) - 1;
4348 /* check that we can start at pba_of_lba0 with
4349 * blocks_per_member of space
4351 if (array_start
>= pos
&& array_end
< ex
[j
].start
) {
4355 pos
= ex
[j
].start
+ ex
[j
].size
;
4357 } while (ex
[j
-1].size
);
4364 if (i
< mpb
->num_raid_devs
) {
4365 dprintf("%x:%x does not have %u to %u available\n",
4366 dl
->major
, dl
->minor
, array_start
, array_end
);
4376 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
4377 struct metadata_update
**updates
)
4380 * Find a device with unused free space and use it to replace a
4381 * failed/vacant region in an array. We replace failed regions one a
4382 * array at a time. The result is that a new spare disk will be added
4383 * to the first failed array and after the monitor has finished
4384 * propagating failures the remainder will be consumed.
4386 * FIXME add a capability for mdmon to request spares from another
4390 struct intel_super
*super
= a
->container
->sb
;
4391 int inst
= a
->info
.container_member
;
4392 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
4393 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4394 int failed
= a
->info
.array
.raid_disks
;
4395 struct mdinfo
*rv
= NULL
;
4398 struct metadata_update
*mu
;
4400 struct imsm_update_activate_spare
*u
;
4404 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
4405 if ((d
->curr_state
& DS_FAULTY
) &&
4407 /* wait for Removal to happen */
4409 if (d
->state_fd
>= 0)
4413 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
4414 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
4415 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
4418 /* For each slot, if it is not working, find a spare */
4419 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
4420 for (d
= a
->info
.devs
; d
; d
= d
->next
)
4421 if (d
->disk
.raid_disk
== i
)
4423 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
4424 if (d
&& (d
->state_fd
>= 0))
4428 * OK, this device needs recovery. Try to re-add the
4429 * previous occupant of this slot, if this fails see if
4430 * we can continue the assimilation of a spare that was
4431 * partially assimilated, finally try to activate a new
4434 dl
= imsm_readd(super
, i
, a
);
4436 dl
= imsm_add_spare(super
, i
, a
, 0);
4438 dl
= imsm_add_spare(super
, i
, a
, 1);
4442 /* found a usable disk with enough space */
4443 di
= malloc(sizeof(*di
));
4446 memset(di
, 0, sizeof(*di
));
4448 /* dl->index will be -1 in the case we are activating a
4449 * pristine spare. imsm_process_update() will create a
4450 * new index in this case. Once a disk is found to be
4451 * failed in all member arrays it is kicked from the
4454 di
->disk
.number
= dl
->index
;
4456 /* (ab)use di->devs to store a pointer to the device
4459 di
->devs
= (struct mdinfo
*) dl
;
4461 di
->disk
.raid_disk
= i
;
4462 di
->disk
.major
= dl
->major
;
4463 di
->disk
.minor
= dl
->minor
;
4465 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
4466 di
->component_size
= a
->info
.component_size
;
4467 di
->container_member
= inst
;
4468 super
->random
= random32();
4472 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
4473 i
, di
->data_offset
);
4479 /* No spares found */
4481 /* Now 'rv' has a list of devices to return.
4482 * Create a metadata_update record to update the
4483 * disk_ord_tbl for the array
4485 mu
= malloc(sizeof(*mu
));
4487 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
4488 if (mu
->buf
== NULL
) {
4495 struct mdinfo
*n
= rv
->next
;
4504 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
4505 mu
->next
= *updates
;
4506 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
4508 for (di
= rv
; di
; di
= di
->next
) {
4509 u
->type
= update_activate_spare
;
4510 u
->dl
= (struct dl
*) di
->devs
;
4512 u
->slot
= di
->disk
.raid_disk
;
4523 static int disks_overlap(struct intel_super
*super
, int idx
, struct imsm_update_create_array
*u
)
4525 struct imsm_dev
*dev
= get_imsm_dev(super
, idx
);
4526 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4527 struct imsm_map
*new_map
= get_imsm_map(&u
->dev
, 0);
4528 struct disk_info
*inf
= get_disk_info(u
);
4529 struct imsm_disk
*disk
;
4533 for (i
= 0; i
< map
->num_members
; i
++) {
4534 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
4535 for (j
= 0; j
< new_map
->num_members
; j
++)
4536 if (serialcmp(disk
->serial
, inf
[j
].serial
) == 0)
4543 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
);
4545 static void imsm_process_update(struct supertype
*st
,
4546 struct metadata_update
*update
)
4549 * crack open the metadata_update envelope to find the update record
4550 * update can be one of:
4551 * update_activate_spare - a spare device has replaced a failed
4552 * device in an array, update the disk_ord_tbl. If this disk is
4553 * present in all member arrays then also clear the SPARE_DISK
4556 struct intel_super
*super
= st
->sb
;
4557 struct imsm_super
*mpb
;
4558 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
4560 /* update requires a larger buf but the allocation failed */
4561 if (super
->next_len
&& !super
->next_buf
) {
4562 super
->next_len
= 0;
4566 if (super
->next_buf
) {
4567 memcpy(super
->next_buf
, super
->buf
, super
->len
);
4569 super
->len
= super
->next_len
;
4570 super
->buf
= super
->next_buf
;
4572 super
->next_len
= 0;
4573 super
->next_buf
= NULL
;
4576 mpb
= super
->anchor
;
4579 case update_activate_spare
: {
4580 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
4581 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
4582 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4583 struct imsm_map
*migr_map
;
4584 struct active_array
*a
;
4585 struct imsm_disk
*disk
;
4590 int victim
= get_imsm_disk_idx(dev
, u
->slot
);
4593 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4598 fprintf(stderr
, "error: imsm_activate_spare passed "
4599 "an unknown disk (index: %d)\n",
4604 super
->updates_pending
++;
4606 /* count failures (excluding rebuilds and the victim)
4607 * to determine map[0] state
4610 for (i
= 0; i
< map
->num_members
; i
++) {
4613 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
));
4614 if (!disk
|| is_failed(disk
))
4618 /* adding a pristine spare, assign a new index */
4619 if (dl
->index
< 0) {
4620 dl
->index
= super
->anchor
->num_disks
;
4621 super
->anchor
->num_disks
++;
4624 disk
->status
|= CONFIGURED_DISK
;
4625 disk
->status
&= ~SPARE_DISK
;
4628 to_state
= imsm_check_degraded(super
, dev
, failed
);
4629 map
->map_state
= IMSM_T_STATE_DEGRADED
;
4630 migrate(dev
, to_state
, MIGR_REBUILD
);
4631 migr_map
= get_imsm_map(dev
, 1);
4632 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
4633 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
4635 /* update the family_num to mark a new container
4636 * generation, being careful to record the existing
4637 * family_num in orig_family_num to clean up after
4638 * earlier mdadm versions that neglected to set it.
4640 if (mpb
->orig_family_num
== 0)
4641 mpb
->orig_family_num
= mpb
->family_num
;
4642 mpb
->family_num
+= super
->random
;
4644 /* count arrays using the victim in the metadata */
4646 for (a
= st
->arrays
; a
; a
= a
->next
) {
4647 dev
= get_imsm_dev(super
, a
->info
.container_member
);
4648 map
= get_imsm_map(dev
, 0);
4650 if (get_imsm_disk_slot(map
, victim
) >= 0)
4654 /* delete the victim if it is no longer being
4660 /* We know that 'manager' isn't touching anything,
4661 * so it is safe to delete
4663 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
4664 if ((*dlp
)->index
== victim
)
4667 /* victim may be on the missing list */
4669 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
4670 if ((*dlp
)->index
== victim
)
4672 imsm_delete(super
, dlp
, victim
);
4676 case update_create_array
: {
4677 /* someone wants to create a new array, we need to be aware of
4678 * a few races/collisions:
4679 * 1/ 'Create' called by two separate instances of mdadm
4680 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
4681 * devices that have since been assimilated via
4683 * In the event this update can not be carried out mdadm will
4684 * (FIX ME) notice that its update did not take hold.
4686 struct imsm_update_create_array
*u
= (void *) update
->buf
;
4687 struct intel_dev
*dv
;
4688 struct imsm_dev
*dev
;
4689 struct imsm_map
*map
, *new_map
;
4690 unsigned long long start
, end
;
4691 unsigned long long new_start
, new_end
;
4693 struct disk_info
*inf
;
4696 /* handle racing creates: first come first serve */
4697 if (u
->dev_idx
< mpb
->num_raid_devs
) {
4698 dprintf("%s: subarray %d already defined\n",
4699 __func__
, u
->dev_idx
);
4703 /* check update is next in sequence */
4704 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
4705 dprintf("%s: can not create array %d expected index %d\n",
4706 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
4710 new_map
= get_imsm_map(&u
->dev
, 0);
4711 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
4712 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
4713 inf
= get_disk_info(u
);
4715 /* handle activate_spare versus create race:
4716 * check to make sure that overlapping arrays do not include
4719 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4720 dev
= get_imsm_dev(super
, i
);
4721 map
= get_imsm_map(dev
, 0);
4722 start
= __le32_to_cpu(map
->pba_of_lba0
);
4723 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
4724 if ((new_start
>= start
&& new_start
<= end
) ||
4725 (start
>= new_start
&& start
<= new_end
))
4730 if (disks_overlap(super
, i
, u
)) {
4731 dprintf("%s: arrays overlap\n", __func__
);
4736 /* check that prepare update was successful */
4737 if (!update
->space
) {
4738 dprintf("%s: prepare update failed\n", __func__
);
4742 /* check that all disks are still active before committing
4743 * changes. FIXME: could we instead handle this by creating a
4744 * degraded array? That's probably not what the user expects,
4745 * so better to drop this update on the floor.
4747 for (i
= 0; i
< new_map
->num_members
; i
++) {
4748 dl
= serial_to_dl(inf
[i
].serial
, super
);
4750 dprintf("%s: disk disappeared\n", __func__
);
4755 super
->updates_pending
++;
4757 /* convert spares to members and fixup ord_tbl */
4758 for (i
= 0; i
< new_map
->num_members
; i
++) {
4759 dl
= serial_to_dl(inf
[i
].serial
, super
);
4760 if (dl
->index
== -1) {
4761 dl
->index
= mpb
->num_disks
;
4763 dl
->disk
.status
|= CONFIGURED_DISK
;
4764 dl
->disk
.status
&= ~SPARE_DISK
;
4766 set_imsm_ord_tbl_ent(new_map
, i
, dl
->index
);
4771 update
->space
= NULL
;
4772 imsm_copy_dev(dev
, &u
->dev
);
4773 dv
->index
= u
->dev_idx
;
4774 dv
->next
= super
->devlist
;
4775 super
->devlist
= dv
;
4776 mpb
->num_raid_devs
++;
4778 imsm_update_version_info(super
);
4781 /* mdmon knows how to release update->space, but not
4782 * ((struct intel_dev *) update->space)->dev
4784 if (update
->space
) {
4790 case update_add_disk
:
4792 /* we may be able to repair some arrays if disks are
4795 struct active_array
*a
;
4797 super
->updates_pending
++;
4798 for (a
= st
->arrays
; a
; a
= a
->next
)
4799 a
->check_degraded
= 1;
4801 /* add some spares to the metadata */
4802 while (super
->add
) {
4806 super
->add
= al
->next
;
4807 al
->next
= super
->disks
;
4809 dprintf("%s: added %x:%x\n",
4810 __func__
, al
->major
, al
->minor
);
4817 static void imsm_prepare_update(struct supertype
*st
,
4818 struct metadata_update
*update
)
4821 * Allocate space to hold new disk entries, raid-device entries or a new
4822 * mpb if necessary. The manager synchronously waits for updates to
4823 * complete in the monitor, so new mpb buffers allocated here can be
4824 * integrated by the monitor thread without worrying about live pointers
4825 * in the manager thread.
4827 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
4828 struct intel_super
*super
= st
->sb
;
4829 struct imsm_super
*mpb
= super
->anchor
;
4834 case update_create_array
: {
4835 struct imsm_update_create_array
*u
= (void *) update
->buf
;
4836 struct intel_dev
*dv
;
4837 struct imsm_dev
*dev
= &u
->dev
;
4838 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4840 struct disk_info
*inf
;
4844 inf
= get_disk_info(u
);
4845 len
= sizeof_imsm_dev(dev
, 1);
4846 /* allocate a new super->devlist entry */
4847 dv
= malloc(sizeof(*dv
));
4849 dv
->dev
= malloc(len
);
4854 update
->space
= NULL
;
4858 /* count how many spares will be converted to members */
4859 for (i
= 0; i
< map
->num_members
; i
++) {
4860 dl
= serial_to_dl(inf
[i
].serial
, super
);
4862 /* hmm maybe it failed?, nothing we can do about
4867 if (count_memberships(dl
, super
) == 0)
4870 len
+= activate
* sizeof(struct imsm_disk
);
4877 /* check if we need a larger metadata buffer */
4878 if (super
->next_buf
)
4879 buf_len
= super
->next_len
;
4881 buf_len
= super
->len
;
4883 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
4884 /* ok we need a larger buf than what is currently allocated
4885 * if this allocation fails process_update will notice that
4886 * ->next_len is set and ->next_buf is NULL
4888 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
4889 if (super
->next_buf
)
4890 free(super
->next_buf
);
4892 super
->next_len
= buf_len
;
4893 if (posix_memalign(&super
->next_buf
, 512, buf_len
) == 0)
4894 memset(super
->next_buf
, 0, buf_len
);
4896 super
->next_buf
= NULL
;
4900 /* must be called while manager is quiesced */
4901 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, int index
)
4903 struct imsm_super
*mpb
= super
->anchor
;
4905 struct imsm_dev
*dev
;
4906 struct imsm_map
*map
;
4907 int i
, j
, num_members
;
4910 dprintf("%s: deleting device[%d] from imsm_super\n",
4913 /* shift all indexes down one */
4914 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
4915 if (iter
->index
> index
)
4917 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
4918 if (iter
->index
> index
)
4921 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4922 dev
= get_imsm_dev(super
, i
);
4923 map
= get_imsm_map(dev
, 0);
4924 num_members
= map
->num_members
;
4925 for (j
= 0; j
< num_members
; j
++) {
4926 /* update ord entries being careful not to propagate
4927 * ord-flags to the first map
4929 ord
= get_imsm_ord_tbl_ent(dev
, j
);
4931 if (ord_to_idx(ord
) <= index
)
4934 map
= get_imsm_map(dev
, 0);
4935 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
4936 map
= get_imsm_map(dev
, 1);
4938 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
4943 super
->updates_pending
++;
4945 struct dl
*dl
= *dlp
;
4947 *dlp
= (*dlp
)->next
;
4948 __free_imsm_disk(dl
);
4951 #endif /* MDASSEMBLE */
4953 struct superswitch super_imsm
= {
4955 .examine_super
= examine_super_imsm
,
4956 .brief_examine_super
= brief_examine_super_imsm
,
4957 .brief_examine_subarrays
= brief_examine_subarrays_imsm
,
4958 .export_examine_super
= export_examine_super_imsm
,
4959 .detail_super
= detail_super_imsm
,
4960 .brief_detail_super
= brief_detail_super_imsm
,
4961 .write_init_super
= write_init_super_imsm
,
4962 .validate_geometry
= validate_geometry_imsm
,
4963 .add_to_super
= add_to_super_imsm
,
4964 .detail_platform
= detail_platform_imsm
,
4966 .match_home
= match_home_imsm
,
4967 .uuid_from_super
= uuid_from_super_imsm
,
4968 .getinfo_super
= getinfo_super_imsm
,
4969 .update_super
= update_super_imsm
,
4971 .avail_size
= avail_size_imsm
,
4973 .compare_super
= compare_super_imsm
,
4975 .load_super
= load_super_imsm
,
4976 .init_super
= init_super_imsm
,
4977 .store_super
= store_super_imsm
,
4978 .free_super
= free_super_imsm
,
4979 .match_metadata_desc
= match_metadata_desc_imsm
,
4980 .container_content
= container_content_imsm
,
4981 .default_layout
= imsm_level_to_layout
,
4988 .open_new
= imsm_open_new
,
4989 .load_super
= load_super_imsm
,
4990 .set_array_state
= imsm_set_array_state
,
4991 .set_disk
= imsm_set_disk
,
4992 .sync_metadata
= imsm_sync_metadata
,
4993 .activate_spare
= imsm_activate_spare
,
4994 .process_update
= imsm_process_update
,
4995 .prepare_update
= imsm_prepare_update
,
4996 #endif /* MDASSEMBLE */