2 * mdadm - Intel(R) Matrix Storage Manager Support
4 * Copyright (C) 2002-2007 Intel Corporation
6 * This program is free software; you can redistribute it and/or modify it
7 * under the terms and conditions of the GNU General Public License,
8 * version 2, as published by the Free Software Foundation.
10 * This program is distributed in the hope it will be useful, but WITHOUT
11 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
12 * FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for
15 * You should have received a copy of the GNU General Public License along with
16 * this program; if not, write to the Free Software Foundation, Inc.,
17 * 51 Franklin St - Fifth Floor, Boston, MA 02110-1301 USA.
26 /* MPB == Metadata Parameter Block */
27 #define MPB_SIGNATURE "Intel Raid ISM Cfg Sig. "
28 #define MPB_SIG_LEN (strlen(MPB_SIGNATURE))
29 #define MPB_VERSION_RAID0 "1.0.00"
30 #define MPB_VERSION_RAID1 "1.1.00"
31 #define MPB_VERSION_RAID5 "1.2.02"
32 #define MAX_SIGNATURE_LENGTH 32
33 #define MAX_RAID_SERIAL_LEN 16
34 #define MPB_SECTOR_CNT 418
35 #define IMSM_RESERVED_SECTORS 4096
37 /* Disk configuration info. */
38 #define IMSM_MAX_DEVICES 255
40 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
41 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
42 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
43 __u32 status
; /* 0xF0 - 0xF3 */
44 #define SPARE_DISK 0x01 /* Spare */
45 #define CONFIGURED_DISK 0x02 /* Member of some RaidDev */
46 #define FAILED_DISK 0x04 /* Permanent failure */
47 #define USABLE_DISK 0x08 /* Fully usable unless FAILED_DISK is set */
49 #define IMSM_DISK_FILLERS 5
50 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
53 /* RAID map configuration infos. */
55 __u32 pba_of_lba0
; /* start address of partition */
56 __u32 blocks_per_member
;/* blocks per member */
57 __u32 num_data_stripes
; /* number of data stripes */
58 __u16 blocks_per_strip
;
59 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
60 #define IMSM_T_STATE_NORMAL 0
61 #define IMSM_T_STATE_UNINITIALIZED 1
62 #define IMSM_T_STATE_DEGRADED 2 /* FIXME: is this correct? */
63 #define IMSM_T_STATE_FAILED 3 /* FIXME: is this correct? */
65 #define IMSM_T_RAID0 0
66 #define IMSM_T_RAID1 1
67 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
68 __u8 num_members
; /* number of member disks */
70 __u32 filler
[7]; /* expansion area */
71 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
73 } __attribute__ ((packed
));
77 __u8 migr_state
; /* Normal or Migrating */
78 __u8 migr_type
; /* Initializing, Rebuilding, ... */
82 struct imsm_map map
[1];
83 /* here comes another one if migr_state */
84 } __attribute__ ((packed
));
87 __u8 volume
[MAX_RAID_SERIAL_LEN
];
90 __u32 status
; /* Persistent RaidDev status */
91 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
92 #define IMSM_DEV_FILLERS 12
93 __u32 filler
[IMSM_DEV_FILLERS
];
95 } __attribute__ ((packed
));
98 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
99 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
100 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
101 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
102 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
103 __u32 error_log_size
; /* 0x30 - 0x33 in bytes */
104 __u32 attributes
; /* 0x34 - 0x37 */
105 __u8 num_disks
; /* 0x38 Number of configured disks */
106 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
107 __u8 error_log_pos
; /* 0x3A */
108 __u8 fill
[1]; /* 0x3B */
109 __u32 cache_size
; /* 0x3c - 0x40 in mb */
110 __u32 orig_family_num
; /* 0x40 - 0x43 original family num */
111 __u32 pwr_cycle_count
; /* 0x44 - 0x47 simulated power cycle count for array */
112 __u32 bbm_log_size
; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
113 #define IMSM_FILLERS 35
114 __u32 filler
[IMSM_FILLERS
]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
115 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
116 /* here comes imsm_dev[num_raid_devs] */
117 /* here comes BBM logs */
118 } __attribute__ ((packed
));
120 #define BBM_LOG_MAX_ENTRIES 254
122 struct bbm_log_entry
{
123 __u64 defective_block_start
;
124 #define UNREADABLE 0xFFFFFFFF
125 __u32 spare_block_offset
;
126 __u16 remapped_marked_count
;
128 } __attribute__ ((__packed__
));
131 __u32 signature
; /* 0xABADB10C */
133 __u32 reserved_spare_block_count
; /* 0 */
134 __u32 reserved
; /* 0xFFFF */
135 __u64 first_spare_lba
;
136 struct bbm_log_entry mapped_block_entries
[BBM_LOG_MAX_ENTRIES
];
137 } __attribute__ ((__packed__
));
141 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
144 static unsigned int sector_count(__u32 bytes
)
146 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
149 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
151 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
154 /* internal representation of IMSM metadata */
157 void *buf
; /* O_DIRECT buffer for reading/writing metadata */
158 struct imsm_super
*anchor
; /* immovable parameters */
160 size_t len
; /* size of the 'buf' allocation */
161 int updates_pending
; /* count of pending updates for mdmon */
162 int creating_imsm
; /* flag to indicate container creation */
163 int current_vol
; /* index of raid device undergoing creation */
164 #define IMSM_MAX_DISKS 6
165 struct imsm_disk
*disk_tbl
[IMSM_MAX_DISKS
];
166 #define IMSM_MAX_RAID_DEVS 2
167 struct imsm_dev
*dev_tbl
[IMSM_MAX_RAID_DEVS
];
171 __u8 serial
[MAX_RAID_SERIAL_LEN
];
176 struct bbm_log
*bbm_log
;
180 unsigned long long start
, size
;
183 /* definition of messages passed to imsm_process_update */
184 enum imsm_update_type
{
185 update_activate_spare
,
189 struct imsm_update_activate_spare
{
190 enum imsm_update_type type
;
194 struct imsm_update_activate_spare
*next
;
197 struct imsm_update_create_array
{
198 enum imsm_update_type type
;
203 static int imsm_env_devname_as_serial(void)
205 char *val
= getenv("IMSM_DEVNAME_AS_SERIAL");
207 if (val
&& atoi(val
) == 1)
214 static struct supertype
*match_metadata_desc_imsm(char *arg
)
216 struct supertype
*st
;
218 if (strcmp(arg
, "imsm") != 0 &&
219 strcmp(arg
, "default") != 0
223 st
= malloc(sizeof(*st
));
224 memset(st
, 0, sizeof(*st
));
225 st
->ss
= &super_imsm
;
226 st
->max_devs
= IMSM_MAX_DEVICES
;
227 st
->minor_version
= 0;
232 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
234 return &mpb
->sig
[MPB_SIG_LEN
];
237 /* retrieve a disk directly from the anchor when the anchor is known to be
238 * up-to-date, currently only at load time
240 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
242 if (index
>= mpb
->num_disks
)
244 return &mpb
->disk
[index
];
247 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
249 if (index
>= super
->anchor
->num_disks
)
251 return super
->disk_tbl
[index
];
254 /* generate a checksum directly from the anchor when the anchor is known to be
255 * up-to-date, currently only at load or write_super after coalescing
257 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
259 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
260 __u32
*p
= (__u32
*) mpb
;
264 sum
+= __le32_to_cpu(*p
++);
266 return sum
- __le32_to_cpu(mpb
->check_sum
);
269 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
)
271 size_t size
= sizeof(*dev
);
273 /* each map has disk_ord_tbl[num_members - 1] additional space */
274 size
+= sizeof(__u32
) * (dev
->vol
.map
[0].num_members
- 1);
276 /* migrating means an additional map */
277 if (dev
->vol
.migr_state
) {
278 size
+= sizeof(struct imsm_map
);
279 size
+= sizeof(__u32
) * (dev
->vol
.map
[1].num_members
- 1);
285 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
291 if (index
>= mpb
->num_raid_devs
)
294 /* devices start after all disks */
295 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
297 for (i
= 0; i
<= index
; i
++)
299 return _mpb
+ offset
;
301 offset
+= sizeof_imsm_dev(_mpb
+ offset
);
306 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
308 if (index
>= super
->anchor
->num_raid_devs
)
310 return super
->dev_tbl
[index
];
313 static __u32
get_imsm_disk_idx(struct imsm_map
*map
, int slot
)
315 __u32
*ord_tbl
= &map
->disk_ord_tbl
[slot
];
317 /* top byte is 'special' */
318 return __le32_to_cpu(*ord_tbl
& ~(0xff << 24));
321 static int get_imsm_raid_level(struct imsm_map
*map
)
323 if (map
->raid_level
== 1) {
324 if (map
->num_members
== 2)
330 return map
->raid_level
;
333 static int cmp_extent(const void *av
, const void *bv
)
335 const struct extent
*a
= av
;
336 const struct extent
*b
= bv
;
337 if (a
->start
< b
->start
)
339 if (a
->start
> b
->start
)
344 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
346 /* find a list of used extents on the given physical device */
347 struct imsm_disk
*disk
;
348 struct extent
*rv
, *e
;
352 disk
= get_imsm_disk(super
, dl
->index
);
356 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
357 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
358 struct imsm_map
*map
= dev
->vol
.map
;
360 for (j
= 0; j
< map
->num_members
; j
++) {
361 __u32 index
= get_imsm_disk_idx(map
, j
);
363 if (index
== dl
->index
)
367 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
372 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
373 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
374 struct imsm_map
*map
= dev
->vol
.map
;
376 for (j
= 0; j
< map
->num_members
; j
++) {
377 __u32 index
= get_imsm_disk_idx(map
, j
);
379 if (index
== dl
->index
) {
380 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
381 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
386 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
388 e
->start
= __le32_to_cpu(disk
->total_blocks
) -
389 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
395 static void print_imsm_dev(struct imsm_dev
*dev
, int index
)
399 struct imsm_map
*map
= dev
->vol
.map
;
402 printf("[%s]:\n", dev
->volume
);
403 printf(" RAID Level : %d\n", get_imsm_raid_level(map
));
404 printf(" Members : %d\n", map
->num_members
);
405 for (slot
= 0; slot
< map
->num_members
; slot
++)
406 if (index
== get_imsm_disk_idx(map
, slot
))
408 if (slot
< map
->num_members
)
409 printf(" This Slot : %d\n", slot
);
411 printf(" This Slot : ?\n");
412 sz
= __le32_to_cpu(dev
->size_high
);
414 sz
+= __le32_to_cpu(dev
->size_low
);
415 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
416 human_size(sz
* 512));
417 sz
= __le32_to_cpu(map
->blocks_per_member
);
418 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
419 human_size(sz
* 512));
420 printf(" Sector Offset : %u\n",
421 __le32_to_cpu(map
->pba_of_lba0
));
422 printf(" Num Stripes : %u\n",
423 __le32_to_cpu(map
->num_data_stripes
));
424 printf(" Chunk Size : %u KiB\n",
425 __le16_to_cpu(map
->blocks_per_strip
) / 2);
426 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
427 printf(" Migrate State : %s\n", dev
->vol
.migr_state
? "migrating" : "idle");
428 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
429 printf(" Map State : %s\n", map_state_str
[map
->map_state
]);
432 static void print_imsm_disk(struct imsm_super
*mpb
, int index
)
434 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
435 char str
[MAX_RAID_SERIAL_LEN
];
443 snprintf(str
, MAX_RAID_SERIAL_LEN
, "%s", disk
->serial
);
444 printf(" Disk%02d Serial : %s\n", index
, str
);
445 s
= __le32_to_cpu(disk
->status
);
446 printf(" State :%s%s%s%s\n", s
&SPARE_DISK
? " spare" : "",
447 s
&CONFIGURED_DISK
? " active" : "",
448 s
&FAILED_DISK
? " failed" : "",
449 s
&USABLE_DISK
? " usable" : "");
450 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
451 sz
= __le32_to_cpu(disk
->total_blocks
) -
452 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
* mpb
->num_raid_devs
);
453 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
454 human_size(sz
* 512));
457 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
459 struct intel_super
*super
= st
->sb
;
460 struct imsm_super
*mpb
= super
->anchor
;
461 char str
[MAX_SIGNATURE_LENGTH
];
465 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
466 printf(" Magic : %s\n", str
);
467 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
468 printf(" Version : %s\n", get_imsm_version(mpb
));
469 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
470 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
471 sum
= __le32_to_cpu(mpb
->check_sum
);
472 printf(" Checksum : %08x %s\n", sum
,
473 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
474 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
475 printf(" Disks : %d\n", mpb
->num_disks
);
476 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
477 print_imsm_disk(mpb
, super
->disks
->index
);
478 if (super
->bbm_log
) {
479 struct bbm_log
*log
= super
->bbm_log
;
482 printf("Bad Block Management Log:\n");
483 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
484 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
485 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
486 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
487 printf(" First Spare : %llx\n", __le64_to_cpu(log
->first_spare_lba
));
489 for (i
= 0; i
< mpb
->num_raid_devs
; i
++)
490 print_imsm_dev(__get_imsm_dev(mpb
, i
), super
->disks
->index
);
491 for (i
= 0; i
< mpb
->num_disks
; i
++) {
492 if (i
== super
->disks
->index
)
494 print_imsm_disk(mpb
, i
);
498 static void brief_examine_super_imsm(struct supertype
*st
)
500 struct intel_super
*super
= st
->sb
;
502 printf("ARRAY /dev/imsm family=%08x metadata=external:imsm\n",
503 __le32_to_cpu(super
->anchor
->family_num
));
506 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
508 printf("%s\n", __FUNCTION__
);
511 static void brief_detail_super_imsm(struct supertype
*st
)
513 printf("%s\n", __FUNCTION__
);
517 static int match_home_imsm(struct supertype
*st
, char *homehost
)
519 printf("%s\n", __FUNCTION__
);
524 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
526 printf("%s\n", __FUNCTION__
);
531 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
533 __u8
*v
= get_imsm_version(mpb
);
534 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
535 char major
[] = { 0, 0, 0 };
536 char minor
[] = { 0 ,0, 0 };
537 char patch
[] = { 0, 0, 0 };
538 char *ver_parse
[] = { major
, minor
, patch
};
542 while (*v
!= '\0' && v
< end
) {
543 if (*v
!= '.' && j
< 2)
544 ver_parse
[i
][j
++] = *v
;
552 *m
= strtol(minor
, NULL
, 0);
553 *p
= strtol(patch
, NULL
, 0);
557 static int imsm_level_to_layout(int level
)
565 return ALGORITHM_LEFT_ASYMMETRIC
;
567 return 0x102; //FIXME is this correct?
572 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
)
574 struct intel_super
*super
= st
->sb
;
575 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
576 struct imsm_map
*map
= &dev
->vol
.map
[0];
578 info
->container_member
= super
->current_vol
;
579 info
->array
.raid_disks
= map
->num_members
;
580 info
->array
.level
= get_imsm_raid_level(map
);
581 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
582 info
->array
.md_minor
= -1;
583 info
->array
.ctime
= 0;
584 info
->array
.utime
= 0;
585 info
->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
* 512);
587 info
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
588 info
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
590 info
->disk
.major
= 0;
591 info
->disk
.minor
= 0;
593 sprintf(info
->text_version
, "/%s/%d",
594 devnum2devname(st
->container_dev
),
595 info
->container_member
);
599 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
)
601 struct intel_super
*super
= st
->sb
;
602 struct imsm_disk
*disk
;
605 if (super
->current_vol
>= 0) {
606 getinfo_super_imsm_volume(st
, info
);
609 info
->array
.raid_disks
= super
->anchor
->num_disks
;
610 info
->array
.level
= LEVEL_CONTAINER
;
611 info
->array
.layout
= 0;
612 info
->array
.md_minor
= -1;
613 info
->array
.ctime
= 0; /* N/A for imsm */
614 info
->array
.utime
= 0;
615 info
->array
.chunk_size
= 0;
617 info
->disk
.major
= 0;
618 info
->disk
.minor
= 0;
619 info
->disk
.raid_disk
= -1;
620 info
->reshape_active
= 0;
621 strcpy(info
->text_version
, "imsm");
622 info
->disk
.number
= -1;
623 info
->disk
.state
= 0;
626 disk
= get_imsm_disk(super
, super
->disks
->index
);
628 info
->disk
.number
= -1;
629 info
->disk
.raid_disk
= -1;
632 info
->disk
.number
= super
->disks
->index
;
633 info
->disk
.raid_disk
= super
->disks
->index
;
634 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) -
635 (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
636 info
->component_size
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
637 s
= __le32_to_cpu(disk
->status
);
638 info
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
639 info
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
640 info
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
644 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
645 char *update
, char *devname
, int verbose
,
646 int uuid_set
, char *homehost
)
650 /* For 'assemble' and 'force' we need to return non-zero if any
651 * change was made. For others, the return value is ignored.
652 * Update options are:
653 * force-one : This device looks a bit old but needs to be included,
654 * update age info appropriately.
655 * assemble: clear any 'faulty' flag to allow this device to
657 * force-array: Array is degraded but being forced, mark it clean
658 * if that will be needed to assemble it.
660 * newdev: not used ????
661 * grow: Array has gained a new device - this is currently for
663 * resync: mark as dirty so a resync will happen.
664 * name: update the name - preserving the homehost
666 * Following are not relevant for this imsm:
667 * sparc2.2 : update from old dodgey metadata
668 * super-minor: change the preferred_minor number
669 * summaries: update redundant counters.
670 * uuid: Change the uuid of the array to match watch is given
671 * homehost: update the recorded homehost
672 * _reshape_progress: record new reshape_progress position.
675 //struct intel_super *super = st->sb;
676 //struct imsm_super *mpb = super->mpb;
678 if (strcmp(update
, "grow") == 0) {
680 if (strcmp(update
, "resync") == 0) {
681 /* dev->vol.dirty = 1; */
684 /* IMSM has no concept of UUID or homehost */
689 static size_t disks_to_mpb_size(int disks
)
693 size
= sizeof(struct imsm_super
);
694 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
695 size
+= 2 * sizeof(struct imsm_dev
);
696 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
697 size
+= (4 - 2) * sizeof(struct imsm_map
);
698 /* 4 possible disk_ord_tbl's */
699 size
+= 4 * (disks
- 1) * sizeof(__u32
);
704 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
706 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
709 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
712 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
716 * 0 same, or first was empty, and second was copied
717 * 1 second had wrong number
721 struct intel_super
*first
= st
->sb
;
722 struct intel_super
*sec
= tst
->sb
;
730 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
) != 0)
732 if (first
->anchor
->family_num
!= sec
->anchor
->family_num
)
734 if (first
->anchor
->mpb_size
!= sec
->anchor
->mpb_size
)
736 if (first
->anchor
->check_sum
!= sec
->anchor
->check_sum
)
742 static void fd2devname(int fd
, char *name
)
751 if (fstat(fd
, &st
) != 0)
753 sprintf(path
, "/sys/dev/block/%d:%d",
754 major(st
.st_rdev
), minor(st
.st_rdev
));
756 rv
= readlink(path
, dname
, sizeof(dname
));
761 nm
= strrchr(dname
, '/');
763 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
767 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
769 static int imsm_read_serial(int fd
, char *devname
,
770 __u8 serial
[MAX_RAID_SERIAL_LEN
])
772 unsigned char scsi_serial
[255];
777 memset(scsi_serial
, 0, sizeof(scsi_serial
));
779 if (imsm_env_devname_as_serial()) {
780 char name
[MAX_RAID_SERIAL_LEN
];
782 fd2devname(fd
, name
);
783 strcpy((char *) serial
, name
);
787 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
792 Name
": Failed to retrieve serial for %s\n",
797 rsp_len
= scsi_serial
[3];
798 for (i
= 0, cnt
= 0; i
< rsp_len
; i
++) {
799 if (!isspace(scsi_serial
[4 + i
]))
800 serial
[cnt
++] = scsi_serial
[4 + i
];
801 if (cnt
== MAX_RAID_SERIAL_LEN
)
805 serial
[MAX_RAID_SERIAL_LEN
- 1] = '\0';
811 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
815 struct imsm_disk
*disk
;
819 dl
= malloc(sizeof(*dl
));
820 disk
= malloc(sizeof(*disk
));
824 Name
": failed to allocate disk buffer for %s\n",
832 memset(dl
, 0, sizeof(*dl
));
833 memset(disk
, 0, sizeof(*disk
));
836 dl
->major
= major(stb
.st_rdev
);
837 dl
->minor
= minor(stb
.st_rdev
);
838 dl
->next
= super
->disks
;
839 dl
->fd
= keep_fd
? fd
: -1;
840 dl
->devname
= devname
? strdup(devname
) : NULL
;
843 rv
= imsm_read_serial(fd
, devname
, dl
->serial
);
848 /* look up this disk's index */
849 for (i
= 0; i
< super
->anchor
->num_disks
; i
++) {
850 struct imsm_disk
*disk_iter
;
852 disk_iter
= __get_imsm_disk(super
->anchor
, i
);
854 if (memcmp(disk_iter
->serial
, dl
->serial
,
855 MAX_RAID_SERIAL_LEN
) == 0) {
857 super
->disk_tbl
[i
] = disk
;
863 if (i
== super
->anchor
->num_disks
) {
866 Name
": failed to match serial \'%s\' for %s\n",
867 dl
->serial
, devname
);
875 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
881 for (i
= 0; i
< src
->vol
.map
[0].num_members
; i
++)
882 dest
->vol
.map
[0].disk_ord_tbl
[i
] = src
->vol
.map
[0].disk_ord_tbl
[i
];
884 if (!src
->vol
.migr_state
)
887 dest
->vol
.map
[1] = src
->vol
.map
[1];
888 for (i
= 0; i
< src
->vol
.map
[1].num_members
; i
++)
889 dest
->vol
.map
[1].disk_ord_tbl
[i
] = src
->vol
.map
[1].disk_ord_tbl
[i
];
892 static int parse_raid_devices(struct intel_super
*super
)
895 struct imsm_dev
*dev_new
;
898 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
899 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
901 len
= sizeof_imsm_dev(dev_iter
);
902 dev_new
= malloc(len
);
905 imsm_copy_dev(dev_new
, dev_iter
);
906 super
->dev_tbl
[i
] = dev_new
;
912 /* retrieve a pointer to the bbm log which starts after all raid devices */
913 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
917 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
919 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
925 static void __free_imsm(struct intel_super
*super
);
927 /* load_imsm_mpb - read matrix metadata
928 * allocates super->mpb to be freed by free_super
930 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
932 unsigned long long dsize
;
933 unsigned long long sectors
;
935 struct imsm_super
*anchor
;
939 get_dev_size(fd
, NULL
, &dsize
);
941 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
944 Name
": Cannot seek to anchor block on %s: %s\n",
945 devname
, strerror(errno
));
949 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
952 Name
": Failed to allocate imsm anchor buffer"
953 " on %s\n", devname
);
956 if (read(fd
, anchor
, 512) != 512) {
959 Name
": Cannot read anchor block on %s: %s\n",
960 devname
, strerror(errno
));
965 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
968 Name
": no IMSM anchor on %s\n", devname
);
974 super
->len
= __le32_to_cpu(anchor
->mpb_size
);
975 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
976 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
979 Name
": unable to allocate %zu byte mpb buffer\n",
984 memcpy(super
->buf
, anchor
, 512);
986 sectors
= mpb_sectors(anchor
) - 1;
989 rc
= load_imsm_disk(fd
, super
, devname
, 0);
991 rc
= parse_raid_devices(super
);
995 /* read the extended mpb */
996 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
999 Name
": Cannot seek to extended mpb on %s: %s\n",
1000 devname
, strerror(errno
));
1004 if (read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
1007 Name
": Cannot read extended mpb on %s: %s\n",
1008 devname
, strerror(errno
));
1012 check_sum
= __gen_imsm_checksum(super
->anchor
);
1013 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
1016 Name
": IMSM checksum %x != %x on %s\n",
1017 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
1022 /* FIXME the BBM log is disk specific so we cannot use this global
1023 * buffer for all disks. Ok for now since we only look at the global
1024 * bbm_log_size parameter to gate assembly
1026 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
1028 rc
= load_imsm_disk(fd
, super
, devname
, 0);
1030 rc
= parse_raid_devices(super
);
1034 static void free_imsm_disks(struct intel_super
*super
)
1038 while (super
->disks
) {
1039 struct dl
*d
= super
->disks
;
1041 super
->disks
= d
->next
;
1048 for (i
= 0; i
< IMSM_MAX_DISKS
; i
++)
1049 if (super
->disk_tbl
[i
]) {
1050 free(super
->disk_tbl
[i
]);
1051 super
->disk_tbl
[i
] = NULL
;
1055 /* free all the pieces hanging off of a super pointer */
1056 static void __free_imsm(struct intel_super
*super
)
1064 free_imsm_disks(super
);
1065 for (i
= 0; i
< IMSM_MAX_RAID_DEVS
; i
++)
1066 if (super
->dev_tbl
[i
]) {
1067 free(super
->dev_tbl
[i
]);
1068 super
->dev_tbl
[i
] = NULL
;
1072 static void free_imsm(struct intel_super
*super
)
1078 static void free_super_imsm(struct supertype
*st
)
1080 struct intel_super
*super
= st
->sb
;
1089 static struct intel_super
*alloc_super(int creating_imsm
)
1091 struct intel_super
*super
= malloc(sizeof(*super
));
1094 memset(super
, 0, sizeof(*super
));
1095 super
->creating_imsm
= creating_imsm
;
1096 super
->current_vol
= -1;
1103 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
1104 char *devname
, int keep_fd
)
1107 struct intel_super
*super
;
1108 struct mdinfo
*sd
, *best
= NULL
;
1115 /* check if this disk is a member of an active array */
1116 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
1120 if (sra
->array
.major_version
!= -1 ||
1121 sra
->array
.minor_version
!= -2 ||
1122 strcmp(sra
->text_version
, "imsm") != 0)
1125 super
= alloc_super(0);
1129 /* find the most up to date disk in this array */
1130 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1131 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1132 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1137 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1141 gen
= __le32_to_cpu(super
->anchor
->generation_num
);
1142 if (!best
|| gen
> bestgen
) {
1157 /* load the most up to date anchor */
1158 sprintf(nm
, "%d:%d", best
->disk
.major
, best
->disk
.minor
);
1159 dfd
= dev_open(nm
, O_RDONLY
);
1164 rv
= load_imsm_mpb(dfd
, super
, NULL
);
1171 /* reset the disk list */
1172 free_imsm_disks(super
);
1174 /* populate disk list */
1175 for (sd
= sra
->devs
; sd
; sd
= sd
->next
) {
1176 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
1177 dfd
= dev_open(nm
, keep_fd
? O_RDWR
: O_RDONLY
);
1182 load_imsm_disk(dfd
, super
, NULL
, keep_fd
);
1187 if (st
->subarray
[0]) {
1188 if (atoi(st
->subarray
) <= super
->anchor
->num_raid_devs
)
1189 super
->current_vol
= atoi(st
->subarray
);
1195 if (st
->ss
== NULL
) {
1196 st
->ss
= &super_imsm
;
1197 st
->minor_version
= 0;
1198 st
->max_devs
= IMSM_MAX_DEVICES
;
1199 st
->container_dev
= fd2devnum(fd
);
1206 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
1208 struct intel_super
*super
;
1212 if (load_super_imsm_all(st
, fd
, &st
->sb
, devname
, 1) == 0)
1215 if (st
->subarray
[0])
1216 return 1; /* FIXME */
1218 super
= alloc_super(0);
1221 Name
": malloc of %zu failed.\n",
1226 rv
= load_imsm_mpb(fd
, super
, devname
);
1231 Name
": Failed to load all information "
1232 "sections on %s\n", devname
);
1238 if (st
->ss
== NULL
) {
1239 st
->ss
= &super_imsm
;
1240 st
->minor_version
= 0;
1241 st
->max_devs
= IMSM_MAX_DEVICES
;
1247 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
1249 if (info
->level
== 1)
1251 return info
->chunk_size
>> 9;
1254 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
)
1258 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
1259 if (info
->level
== 1)
1265 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
1267 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
1270 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
1271 unsigned long long size
, char *name
,
1272 char *homehost
, int *uuid
)
1274 /* We are creating a volume inside a pre-existing container.
1275 * so st->sb is already set.
1277 struct intel_super
*super
= st
->sb
;
1278 struct imsm_super
*mpb
= super
->anchor
;
1279 struct imsm_dev
*dev
;
1280 struct imsm_vol
*vol
;
1281 struct imsm_map
*map
;
1282 int idx
= mpb
->num_raid_devs
;
1284 unsigned long long array_blocks
;
1286 size_t size_old
, size_new
;
1288 if (mpb
->num_raid_devs
>= 2) {
1289 fprintf(stderr
, Name
": This imsm-container already has the "
1290 "maximum of 2 volumes\n");
1294 /* ensure the mpb is large enough for the new data */
1295 size_old
= __le32_to_cpu(mpb
->mpb_size
);
1296 size_new
= disks_to_mpb_size(info
->nr_disks
);
1297 if (size_new
> size_old
) {
1299 size_t size_round
= ROUND_UP(size_new
, 512);
1301 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
1302 fprintf(stderr
, Name
": could not allocate new mpb\n");
1305 memcpy(mpb_new
, mpb
, size_old
);
1308 super
->anchor
= mpb_new
;
1309 mpb
->mpb_size
= __cpu_to_le32(size_new
);
1310 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
1312 super
->current_vol
= idx
;
1313 sprintf(st
->subarray
, "%d", idx
);
1314 dev
= malloc(sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
1316 fprintf(stderr
, Name
": could not allocate raid device\n");
1319 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
1320 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
1321 info
->layout
, info
->chunk_size
,
1323 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
1324 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
1325 dev
->status
= __cpu_to_le32(0);
1326 dev
->reserved_blocks
= __cpu_to_le32(0);
1328 vol
->migr_state
= 0;
1331 for (i
= 0; i
< idx
; i
++) {
1332 struct imsm_dev
*prev
= get_imsm_dev(super
, i
);
1333 struct imsm_map
*pmap
= &prev
->vol
.map
[0];
1335 offset
+= __le32_to_cpu(pmap
->blocks_per_member
);
1336 offset
+= IMSM_RESERVED_SECTORS
;
1339 map
->pba_of_lba0
= __cpu_to_le32(offset
);
1340 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
1341 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
1342 map
->num_data_stripes
= __cpu_to_le32(info_to_num_data_stripes(info
));
1343 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
1344 IMSM_T_STATE_NORMAL
;
1346 if (info
->level
== 1 && info
->raid_disks
> 2) {
1347 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
1348 "in a raid1 volume\n");
1351 if (info
->level
== 10)
1352 map
->raid_level
= 1;
1354 map
->raid_level
= info
->level
;
1356 map
->num_members
= info
->raid_disks
;
1357 for (i
= 0; i
< map
->num_members
; i
++) {
1358 /* initialized in add_to_super */
1359 map
->disk_ord_tbl
[i
] = __cpu_to_le32(0);
1361 mpb
->num_raid_devs
++;
1362 super
->dev_tbl
[super
->current_vol
] = dev
;
1367 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
1368 unsigned long long size
, char *name
,
1369 char *homehost
, int *uuid
)
1371 /* This is primarily called by Create when creating a new array.
1372 * We will then get add_to_super called for each component, and then
1373 * write_init_super called to write it out to each device.
1374 * For IMSM, Create can create on fresh devices or on a pre-existing
1376 * To create on a pre-existing array a different method will be called.
1377 * This one is just for fresh drives.
1379 struct intel_super
*super
;
1380 struct imsm_super
*mpb
;
1388 return init_super_imsm_volume(st
, info
, size
, name
, homehost
,
1391 super
= alloc_super(1);
1394 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
1395 if (posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
1400 memset(mpb
, 0, mpb_size
);
1402 memcpy(mpb
->sig
, MPB_SIGNATURE
, strlen(MPB_SIGNATURE
));
1403 memcpy(mpb
->sig
+ strlen(MPB_SIGNATURE
), MPB_VERSION_RAID5
,
1404 strlen(MPB_VERSION_RAID5
));
1405 mpb
->mpb_size
= mpb_size
;
1411 static void add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
1412 int fd
, char *devname
)
1414 struct intel_super
*super
= st
->sb
;
1416 struct imsm_dev
*dev
;
1417 struct imsm_map
*map
;
1418 struct imsm_disk
*disk
;
1421 dev
= get_imsm_dev(super
, super
->current_vol
);
1422 map
= &dev
->vol
.map
[0];
1424 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
1425 if (dl
->major
== dk
->major
&&
1426 dl
->minor
== dk
->minor
)
1428 if (!dl
|| ! (dk
->state
& (1<<MD_DISK_SYNC
)))
1431 map
->disk_ord_tbl
[dk
->number
] = __cpu_to_le32(dl
->index
);
1433 disk
= get_imsm_disk(super
, dl
->index
);
1434 status
= CONFIGURED_DISK
| USABLE_DISK
;
1435 disk
->status
= __cpu_to_le32(status
);
1438 static void add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
1439 int fd
, char *devname
)
1441 struct intel_super
*super
= st
->sb
;
1442 struct imsm_super
*mpb
= super
->anchor
;
1443 struct imsm_disk
*disk
;
1445 unsigned long long size
;
1450 if (super
->current_vol
>= 0) {
1451 add_to_super_imsm_volume(st
, dk
, fd
, devname
);
1456 dd
= malloc(sizeof(*dd
));
1457 disk
= malloc(sizeof(*disk
));
1460 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
1467 memset(dd
, 0, sizeof(*dd
));
1468 memset(disk
, 0, sizeof(*disk
));
1469 dd
->major
= major(stb
.st_rdev
);
1470 dd
->minor
= minor(stb
.st_rdev
);
1471 dd
->index
= dk
->number
;
1472 dd
->devname
= devname
? strdup(devname
) : NULL
;
1473 dd
->next
= super
->disks
;
1475 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
1478 Name
": failed to retrieve scsi serial, aborting\n");
1484 if (mpb
->num_disks
<= dk
->number
)
1485 mpb
->num_disks
= dk
->number
+ 1;
1487 get_dev_size(fd
, NULL
, &size
);
1489 status
= USABLE_DISK
| SPARE_DISK
;
1490 strcpy((char *) disk
->serial
, (char *) dd
->serial
);
1491 disk
->total_blocks
= __cpu_to_le32(size
);
1492 disk
->status
= __cpu_to_le32(status
);
1493 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
1494 disk
->scsi_id
= __cpu_to_le32(id
);
1496 disk
->scsi_id
= __cpu_to_le32(0);
1497 super
->disk_tbl
[dd
->index
] = disk
;
1499 /* update the family number if we are creating a container */
1500 if (super
->creating_imsm
) {
1501 disk
= __get_imsm_disk(mpb
, dd
->index
);
1502 *disk
= *super
->disk_tbl
[dd
->index
]; /* copy in new disk */
1503 mpb
->family_num
= __cpu_to_le32(__gen_imsm_checksum(mpb
));
1509 static int store_imsm_mpb(int fd
, struct intel_super
*super
);
1511 static int write_super_imsm(struct intel_super
*super
, int doclose
)
1513 struct imsm_super
*mpb
= super
->anchor
;
1519 /* 'generation' is incremented everytime the metadata is written */
1520 generation
= __le32_to_cpu(mpb
->generation_num
);
1522 mpb
->generation_num
= __cpu_to_le32(generation
);
1524 for (i
= 0; i
< mpb
->num_disks
; i
++)
1525 mpb
->disk
[i
] = *super
->disk_tbl
[i
];
1526 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1527 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1529 imsm_copy_dev(dev
, super
->dev_tbl
[i
]);
1532 /* recalculate checksum */
1533 sum
= __gen_imsm_checksum(mpb
);
1534 mpb
->check_sum
= __cpu_to_le32(sum
);
1536 for (d
= super
->disks
; d
; d
= d
->next
) {
1537 if (store_imsm_mpb(d
->fd
, super
)) {
1538 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
1539 __func__
, d
->major
, d
->minor
, strerror(errno
));
1551 static int write_init_super_imsm(struct supertype
*st
)
1553 if (st
->update_tail
) {
1554 /* queue the recently created array as a metadata update */
1556 struct imsm_update_create_array
*u
;
1557 struct intel_super
*super
= st
->sb
;
1558 struct imsm_dev
*dev
;
1561 if (super
->current_vol
< 0 ||
1562 !(dev
= get_imsm_dev(super
, super
->current_vol
))) {
1563 fprintf(stderr
, "%s: could not determine sub-array\n",
1569 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
);
1572 fprintf(stderr
, "%s: failed to allocate update buffer\n",
1577 u
->type
= update_create_array
;
1578 u
->dev_idx
= super
->current_vol
;
1579 imsm_copy_dev(&u
->dev
, dev
);
1580 append_metadata_update(st
, u
, len
);
1582 for (d
= super
->disks
; d
; d
= d
->next
) {
1589 return write_super_imsm(st
->sb
, 1);
1592 static int store_zero_imsm(struct supertype
*st
, int fd
)
1594 unsigned long long dsize
;
1597 get_dev_size(fd
, NULL
, &dsize
);
1599 /* first block is stored on second to last sector of the disk */
1600 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
1603 if (posix_memalign(&buf
, 512, 512) != 0)
1606 memset(buf
, 0, 512);
1607 if (write(fd
, buf
, 512) != 512)
1612 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
1613 int layout
, int raiddisks
, int chunk
,
1614 unsigned long long size
, char *dev
,
1615 unsigned long long *freesize
,
1619 unsigned long long ldsize
;
1621 if (level
!= LEVEL_CONTAINER
)
1626 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1629 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
1630 dev
, strerror(errno
));
1633 if (!get_dev_size(fd
, dev
, &ldsize
)) {
1639 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
1644 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
1645 * FIX ME add ahci details
1647 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
1648 int layout
, int raiddisks
, int chunk
,
1649 unsigned long long size
, char *dev
,
1650 unsigned long long *freesize
,
1654 struct intel_super
*super
= st
->sb
;
1656 unsigned long long pos
= 0;
1657 unsigned long long maxsize
;
1661 if (level
== LEVEL_CONTAINER
)
1664 if (level
== 1 && raiddisks
> 2) {
1666 fprintf(stderr
, Name
": imsm does not support more "
1667 "than 2 in a raid1 configuration\n");
1671 /* We must have the container info already read in. */
1676 /* General test: make sure there is space for
1677 * 'raiddisks' device extents of size 'size' at a given
1680 unsigned long long minsize
= size
*2 /* convert to blocks */;
1681 unsigned long long start_offset
= ~0ULL;
1684 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
1685 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1690 e
= get_extents(super
, dl
);
1693 unsigned long long esize
;
1694 esize
= e
[i
].start
- pos
;
1695 if (esize
>= minsize
)
1697 if (found
&& start_offset
== ~0ULL) {
1700 } else if (found
&& pos
!= start_offset
) {
1704 pos
= e
[i
].start
+ e
[i
].size
;
1706 } while (e
[i
-1].size
);
1711 if (dcnt
< raiddisks
) {
1713 fprintf(stderr
, Name
": imsm: Not enough "
1714 "devices with space for this array "
1721 /* This device must be a member of the set */
1722 if (stat(dev
, &stb
) < 0)
1724 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
1726 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1727 if (dl
->major
== major(stb
.st_rdev
) &&
1728 dl
->minor
== minor(stb
.st_rdev
))
1733 fprintf(stderr
, Name
": %s is not in the "
1734 "same imsm set\n", dev
);
1737 e
= get_extents(super
, dl
);
1741 unsigned long long esize
;
1742 esize
= e
[i
].start
- pos
;
1743 if (esize
>= maxsize
)
1745 pos
= e
[i
].start
+ e
[i
].size
;
1747 } while (e
[i
-1].size
);
1748 *freesize
= maxsize
;
1753 int imsm_bbm_log_size(struct imsm_super
*mpb
)
1755 return __le32_to_cpu(mpb
->bbm_log_size
);
1758 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
1759 int raiddisks
, int chunk
, unsigned long long size
,
1760 char *dev
, unsigned long long *freesize
,
1766 /* if given unused devices create a container
1767 * if given given devices in a container create a member volume
1769 if (level
== LEVEL_CONTAINER
) {
1770 /* Must be a fresh device to add to a container */
1771 return validate_geometry_imsm_container(st
, level
, layout
,
1772 raiddisks
, chunk
, size
,
1778 /* creating in a given container */
1779 return validate_geometry_imsm_volume(st
, level
, layout
,
1780 raiddisks
, chunk
, size
,
1781 dev
, freesize
, verbose
);
1784 /* limit creation to the following levels */
1796 /* This device needs to be a device in an 'imsm' container */
1797 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
1801 Name
": Cannot create this array on device %s\n",
1806 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
1808 fprintf(stderr
, Name
": Cannot open %s: %s\n",
1809 dev
, strerror(errno
));
1812 /* Well, it is in use by someone, maybe an 'imsm' container. */
1813 cfd
= open_container(fd
);
1817 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
1821 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
1823 if (sra
&& sra
->array
.major_version
== -1 &&
1824 strcmp(sra
->text_version
, "imsm") == 0) {
1825 /* This is a member of a imsm container. Load the container
1826 * and try to create a volume
1828 struct intel_super
*super
;
1830 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
, 1) == 0) {
1832 st
->container_dev
= fd2devnum(cfd
);
1834 return validate_geometry_imsm_volume(st
, level
, layout
,
1840 } else /* may belong to another container */
1846 static struct mdinfo
*container_content_imsm(struct supertype
*st
)
1848 /* Given a container loaded by load_super_imsm_all,
1849 * extract information about all the arrays into
1852 * For each imsm_dev create an mdinfo, fill it in,
1853 * then look for matching devices in super->disks
1854 * and create appropriate device mdinfo.
1856 struct intel_super
*super
= st
->sb
;
1857 struct imsm_super
*mpb
= super
->anchor
;
1858 struct mdinfo
*rest
= NULL
;
1861 /* do not assemble arrays that might have bad blocks */
1862 if (imsm_bbm_log_size(super
->anchor
)) {
1863 fprintf(stderr
, Name
": BBM log found in metadata. "
1864 "Cannot activate array(s).\n");
1868 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1869 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
1870 struct imsm_vol
*vol
= &dev
->vol
;
1871 struct imsm_map
*map
= vol
->map
;
1872 struct mdinfo
*this;
1875 this = malloc(sizeof(*this));
1876 memset(this, 0, sizeof(*this));
1880 this->array
.level
= get_imsm_raid_level(map
);
1881 this->array
.raid_disks
= map
->num_members
;
1882 this->array
.layout
= imsm_level_to_layout(this->array
.level
);
1883 this->array
.md_minor
= -1;
1884 this->array
.ctime
= 0;
1885 this->array
.utime
= 0;
1886 this->array
.chunk_size
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
1887 this->array
.state
= !vol
->dirty
;
1888 this->container_member
= i
;
1889 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
|| dev
->vol
.dirty
)
1890 this->resync_start
= 0;
1892 this->resync_start
= ~0ULL;
1894 strncpy(this->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
1895 this->name
[MAX_RAID_SERIAL_LEN
] = 0;
1897 sprintf(this->text_version
, "/%s/%d",
1898 devnum2devname(st
->container_dev
),
1899 this->container_member
);
1901 memset(this->uuid
, 0, sizeof(this->uuid
));
1903 this->component_size
= __le32_to_cpu(map
->blocks_per_member
);
1905 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
1906 struct imsm_disk
*disk
;
1907 struct mdinfo
*info_d
;
1912 idx
= __le32_to_cpu(map
->disk_ord_tbl
[slot
] & ~(0xff << 24));
1913 for (d
= super
->disks
; d
; d
= d
->next
)
1914 if (d
->index
== idx
)
1918 break; /* shouldn't this be continue ?? */
1920 info_d
= malloc(sizeof(*info_d
));
1922 break; /* ditto ?? */
1923 memset(info_d
, 0, sizeof(*info_d
));
1924 info_d
->next
= this->devs
;
1925 this->devs
= info_d
;
1927 disk
= get_imsm_disk(super
, idx
);
1928 s
= __le32_to_cpu(disk
->status
);
1930 info_d
->disk
.number
= d
->index
;
1931 info_d
->disk
.major
= d
->major
;
1932 info_d
->disk
.minor
= d
->minor
;
1933 info_d
->disk
.raid_disk
= slot
;
1934 info_d
->disk
.state
= s
& CONFIGURED_DISK
? (1 << MD_DISK_ACTIVE
) : 0;
1935 info_d
->disk
.state
|= s
& FAILED_DISK
? (1 << MD_DISK_FAULTY
) : 0;
1936 info_d
->disk
.state
|= s
& USABLE_DISK
? (1 << MD_DISK_SYNC
) : 0;
1938 this->array
.working_disks
++;
1940 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
1941 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
1942 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
1944 strcpy(info_d
->name
, d
->devname
);
1952 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
1955 struct intel_super
*super
= c
->sb
;
1956 struct imsm_super
*mpb
= super
->anchor
;
1958 if (atoi(inst
) >= mpb
->num_raid_devs
) {
1959 fprintf(stderr
, "%s: subarry index %d, out of range\n",
1960 __func__
, atoi(inst
));
1964 dprintf("imsm: open_new %s\n", inst
);
1965 a
->info
.container_member
= atoi(inst
);
1969 static __u8
imsm_check_degraded(struct intel_super
*super
, int n
, int failed
)
1971 struct imsm_dev
*dev
= get_imsm_dev(super
, n
);
1972 struct imsm_map
*map
= dev
->vol
.map
;
1975 return map
->map_state
;
1977 switch (get_imsm_raid_level(map
)) {
1979 return IMSM_T_STATE_FAILED
;
1982 if (failed
< map
->num_members
)
1983 return IMSM_T_STATE_DEGRADED
;
1985 return IMSM_T_STATE_FAILED
;
1990 * check to see if any mirrors have failed,
1991 * otherwise we are degraded
1993 int device_per_mirror
= 2; /* FIXME is this always the case?
1994 * and are they always adjacent?
1999 for (i
= 0; i
< map
->num_members
; i
++) {
2000 int idx
= get_imsm_disk_idx(map
, i
);
2001 struct imsm_disk
*disk
= get_imsm_disk(super
, idx
);
2003 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
2006 if (failed
>= device_per_mirror
)
2007 return IMSM_T_STATE_FAILED
;
2009 /* reset 'failed' for next mirror set */
2010 if (!((i
+ 1) % device_per_mirror
))
2014 return IMSM_T_STATE_DEGRADED
;
2018 return IMSM_T_STATE_DEGRADED
;
2020 return IMSM_T_STATE_FAILED
;
2026 return map
->map_state
;
2029 static int imsm_count_failed(struct intel_super
*super
, struct imsm_map
*map
)
2033 struct imsm_disk
*disk
;
2035 for (i
= 0; i
< map
->num_members
; i
++) {
2036 int idx
= get_imsm_disk_idx(map
, i
);
2038 disk
= get_imsm_disk(super
, idx
);
2039 if (__le32_to_cpu(disk
->status
) & FAILED_DISK
)
2046 static void imsm_set_array_state(struct active_array
*a
, int consistent
)
2048 int inst
= a
->info
.container_member
;
2049 struct intel_super
*super
= a
->container
->sb
;
2050 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2051 struct imsm_map
*map
= &dev
->vol
.map
[0];
2052 int dirty
= !consistent
;
2056 if (a
->resync_start
== ~0ULL) {
2057 failed
= imsm_count_failed(super
, map
);
2058 map_state
= imsm_check_degraded(super
, inst
, failed
);
2059 /* complete recovery or initial resync */
2061 map_state
= IMSM_T_STATE_NORMAL
;
2062 if (map
->map_state
!= map_state
) {
2063 dprintf("imsm: map_state %d: %d\n",
2065 map
->map_state
= map_state
;
2066 super
->updates_pending
++;
2069 /* complete resync */
2070 if (!dirty
&& dev
->vol
.dirty
) {
2071 dprintf("imsm: mark 'clean'\n");
2073 super
->updates_pending
++;
2079 if (dirty
&& !dev
->vol
.dirty
) {
2080 dprintf("imsm: mark 'dirty' (%llu)\n", a
->resync_start
);
2082 super
->updates_pending
++;
2086 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
2088 int inst
= a
->info
.container_member
;
2089 struct intel_super
*super
= a
->container
->sb
;
2090 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2091 struct imsm_map
*map
= dev
->vol
.map
;
2092 struct imsm_disk
*disk
;
2095 int new_failure
= 0;
2097 if (n
> map
->num_members
)
2098 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
2099 n
, map
->num_members
- 1);
2104 dprintf("imsm: set_disk %d:%x\n", n
, state
);
2106 disk
= get_imsm_disk(super
, get_imsm_disk_idx(map
, n
));
2108 /* check for new failures */
2109 status
= __le32_to_cpu(disk
->status
);
2110 if ((state
& DS_FAULTY
) && !(status
& FAILED_DISK
)) {
2111 status
|= FAILED_DISK
;
2112 disk
->status
= __cpu_to_le32(status
);
2114 super
->updates_pending
++;
2117 /* the number of failures have changed, count up 'failed' to determine
2118 * degraded / failed status
2120 if (new_failure
&& map
->map_state
!= IMSM_T_STATE_FAILED
)
2121 failed
= imsm_count_failed(super
, map
);
2123 /* determine map_state based on failed or in_sync count */
2125 map
->map_state
= imsm_check_degraded(super
, inst
, failed
);
2126 else if (map
->map_state
== IMSM_T_STATE_DEGRADED
) {
2130 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2131 if (d
->curr_state
& DS_INSYNC
)
2134 if (working
== a
->info
.array
.raid_disks
) {
2135 map
->map_state
= IMSM_T_STATE_NORMAL
;
2136 super
->updates_pending
++;
2141 static int store_imsm_mpb(int fd
, struct intel_super
*super
)
2143 struct imsm_super
*mpb
= super
->anchor
;
2144 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
2145 unsigned long long dsize
;
2146 unsigned long long sectors
;
2148 get_dev_size(fd
, NULL
, &dsize
);
2150 if (mpb_size
> 512) {
2151 /* -1 to account for anchor */
2152 sectors
= mpb_sectors(mpb
) - 1;
2154 /* write the extended mpb to the sectors preceeding the anchor */
2155 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
2158 if (write(fd
, super
->buf
+ 512, 512 * sectors
) != 512 * sectors
)
2162 /* first block is stored on second to last sector of the disk */
2163 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
2166 if (write(fd
, super
->buf
, 512) != 512)
2172 static void imsm_sync_metadata(struct supertype
*container
)
2174 struct intel_super
*super
= container
->sb
;
2176 if (!super
->updates_pending
)
2179 write_super_imsm(super
, 0);
2181 super
->updates_pending
= 0;
2184 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
2185 struct metadata_update
**updates
)
2188 * Take a device that is marked spare in the metadata and use it to
2189 * replace a failed/vacant slot in an array. There may be a case where
2190 * a device is failed in one array but active in a second.
2191 * imsm_process_update catches this case and does not clear the SPARE_DISK
2192 * flag, allowing the second array to start using the device on failure.
2193 * SPARE_DISK is cleared when all arrays are using a device.
2195 * FIXME: is this a valid use of SPARE_DISK?
2198 struct intel_super
*super
= a
->container
->sb
;
2199 int inst
= a
->info
.container_member
;
2200 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
2201 struct imsm_map
*map
= dev
->vol
.map
;
2202 int failed
= a
->info
.array
.raid_disks
;
2203 struct mdinfo
*rv
= NULL
;
2206 struct metadata_update
*mu
;
2208 struct imsm_update_activate_spare
*u
;
2212 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
2213 if ((d
->curr_state
& DS_FAULTY
) &&
2215 /* wait for Removal to happen */
2217 if (d
->state_fd
>= 0)
2221 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
2222 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
2223 if (imsm_check_degraded(super
, inst
, failed
) != IMSM_T_STATE_DEGRADED
)
2226 /* For each slot, if it is not working, find a spare */
2228 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
2229 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2230 if (d
->disk
.raid_disk
== i
)
2232 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
2233 if (d
&& (d
->state_fd
>= 0))
2236 /* OK, this device needs recovery. Find a spare */
2237 for ( ; dl
; dl
= dl
->next
) {
2238 unsigned long long esize
;
2239 unsigned long long pos
;
2242 struct imsm_disk
*disk
;
2247 /* If in this array, skip */
2248 for (d2
= a
->info
.devs
; d2
; d2
= d2
->next
)
2249 if (d2
->disk
.major
== dl
->major
&&
2250 d2
->disk
.minor
== dl
->minor
) {
2251 dprintf("%x:%x already in array\n", dl
->major
, dl
->minor
);
2257 /* is this unused device marked as a spare? */
2258 disk
= get_imsm_disk(super
, dl
->index
);
2259 if (!(__le32_to_cpu(disk
->status
) & SPARE_DISK
))
2262 /* We are allowed to use this device - is there space?
2263 * We need a->info.component_size sectors */
2264 ex
= get_extents(super
, dl
);
2266 dprintf("cannot get extents\n");
2272 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
2275 /* check that we can start at pba_of_lba0 with
2276 * a->info.component_size of space
2278 esize
= ex
[j
].start
- pos
;
2279 if (array_start
>= pos
&&
2280 array_start
+ a
->info
.component_size
< ex
[j
].start
) {
2284 pos
= ex
[j
].start
+ ex
[j
].size
;
2287 } while (ex
[j
-1].size
);
2291 dprintf("%x:%x does not have %llu at %d\n",
2292 dl
->major
, dl
->minor
,
2293 a
->info
.component_size
,
2294 __le32_to_cpu(map
->pba_of_lba0
));
2299 /* found a usable disk with enough space */
2300 di
= malloc(sizeof(*di
));
2301 memset(di
, 0, sizeof(*di
));
2302 di
->disk
.number
= dl
->index
;
2303 di
->disk
.raid_disk
= i
;
2304 di
->disk
.major
= dl
->major
;
2305 di
->disk
.minor
= dl
->minor
;
2307 di
->data_offset
= array_start
;
2308 di
->component_size
= a
->info
.component_size
;
2309 di
->container_member
= inst
;
2313 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
2321 /* No spares found */
2323 /* Now 'rv' has a list of devices to return.
2324 * Create a metadata_update record to update the
2325 * disk_ord_tbl for the array
2327 mu
= malloc(sizeof(*mu
));
2328 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
2330 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
2331 mu
->next
= *updates
;
2332 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
2334 for (di
= rv
; di
; di
= di
->next
) {
2335 u
->type
= update_activate_spare
;
2336 u
->disk_idx
= di
->disk
.number
;
2337 u
->slot
= di
->disk
.raid_disk
;
2348 static int weight(unsigned int field
)
2352 for (weight
= 0; field
; weight
++)
2358 static int disks_overlap(struct imsm_map
*m1
, struct imsm_map
*m2
)
2364 for (i
= 0; i
< m1
->num_members
; i
++) {
2365 idx
= get_imsm_disk_idx(m1
, i
);
2366 for (j
= 0; j
< m2
->num_members
; j
++)
2367 if (idx
== get_imsm_disk_idx(m2
, j
))
2374 static void imsm_process_update(struct supertype
*st
,
2375 struct metadata_update
*update
)
2378 * crack open the metadata_update envelope to find the update record
2379 * update can be one of:
2380 * update_activate_spare - a spare device has replaced a failed
2381 * device in an array, update the disk_ord_tbl. If this disk is
2382 * present in all member arrays then also clear the SPARE_DISK
2385 struct intel_super
*super
= st
->sb
;
2386 struct imsm_super
*mpb
= super
->anchor
;
2387 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
2390 case update_activate_spare
: {
2391 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
2392 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
2393 struct imsm_map
*map
= &dev
->vol
.map
[0];
2394 struct active_array
*a
;
2395 struct imsm_disk
*disk
;
2399 unsigned int members
;
2404 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2405 if (dl
->index
== u
->disk_idx
)
2409 fprintf(stderr
, "error: imsm_activate_spare passed "
2410 "an unknown disk_idx: %d\n", u
->disk_idx
);
2414 super
->updates_pending
++;
2416 victim
= get_imsm_disk_idx(map
, u
->slot
);
2417 map
->disk_ord_tbl
[u
->slot
] = __cpu_to_le32(u
->disk_idx
);
2418 disk
= get_imsm_disk(super
, u
->disk_idx
);
2419 status
= __le32_to_cpu(disk
->status
);
2420 status
|= CONFIGURED_DISK
;
2421 disk
->status
= __cpu_to_le32(status
);
2423 /* map unique/live arrays using the spare */
2426 for (a
= st
->arrays
; a
; a
= a
->next
) {
2427 int inst
= a
->info
.container_member
;
2429 dev
= get_imsm_dev(super
, inst
);
2430 map
= &dev
->vol
.map
[0];
2431 if (map
->raid_level
> 0)
2432 members
|= 1 << inst
;
2433 for (d
= a
->info
.devs
; d
; d
= d
->next
)
2434 if (d
->disk
.major
== dl
->major
&&
2435 d
->disk
.minor
== dl
->minor
)
2439 /* until all arrays that can absorb this disk have absorbed
2440 * this disk it can still be considered a spare
2442 if (weight(found
) >= weight(members
)) {
2443 status
= __le32_to_cpu(disk
->status
);
2444 status
&= ~SPARE_DISK
;
2445 disk
->status
= __cpu_to_le32(status
);
2448 /* count arrays using the victim in the metadata */
2450 for (a
= st
->arrays
; a
; a
= a
->next
) {
2451 dev
= get_imsm_dev(super
, a
->info
.container_member
);
2452 map
= &dev
->vol
.map
[0];
2453 for (i
= 0; i
< map
->num_members
; i
++)
2454 if (victim
== get_imsm_disk_idx(map
, i
))
2458 /* clear some flags if the victim is no longer being
2461 disk
= get_imsm_disk(super
, victim
);
2463 status
= __le32_to_cpu(disk
->status
);
2464 status
&= ~(CONFIGURED_DISK
| USABLE_DISK
);
2465 disk
->status
= __cpu_to_le32(status
);
2469 case update_create_array
: {
2470 /* someone wants to create a new array, we need to be aware of
2471 * a few races/collisions:
2472 * 1/ 'Create' called by two separate instances of mdadm
2473 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
2474 * devices that have since been assimilated via
2476 * In the event this update can not be carried out mdadm will
2477 * (FIX ME) notice that its update did not take hold.
2479 struct imsm_update_create_array
*u
= (void *) update
->buf
;
2480 struct imsm_dev
*dev
;
2481 struct imsm_map
*map
, *new_map
;
2482 unsigned long long start
, end
;
2483 unsigned long long new_start
, new_end
;
2487 /* handle racing creates: first come first serve */
2488 if (u
->dev_idx
< mpb
->num_raid_devs
) {
2489 dprintf("%s: subarray %d already defined\n",
2490 __func__
, u
->dev_idx
);
2494 /* check update is next in sequence */
2495 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
2496 dprintf("%s: can not create arrays out of sequence\n",
2501 new_map
= &u
->dev
.vol
.map
[0];
2502 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
2503 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
2505 /* handle activate_spare versus create race:
2506 * check to make sure that overlapping arrays do not include
2509 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2510 dev
= get_imsm_dev(super
, i
);
2511 map
= &dev
->vol
.map
[0];
2512 start
= __le32_to_cpu(map
->pba_of_lba0
);
2513 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
2514 if ((new_start
>= start
&& new_start
<= end
) ||
2515 (start
>= new_start
&& start
<= new_end
))
2517 if (overlap
&& disks_overlap(map
, new_map
)) {
2518 dprintf("%s: arrays overlap\n", __func__
);
2522 /* check num_members sanity */
2523 if (new_map
->num_members
> mpb
->num_disks
) {
2524 dprintf("%s: num_disks out of range\n", __func__
);
2528 /* check that prepare update was successful */
2529 if (!update
->space
) {
2530 dprintf("%s: prepare update failed\n", __func__
);
2534 super
->updates_pending
++;
2535 dev
= update
->space
;
2536 update
->space
= NULL
;
2537 imsm_copy_dev(dev
, &u
->dev
);
2538 super
->dev_tbl
[u
->dev_idx
] = dev
;
2539 mpb
->num_raid_devs
++;
2541 /* fix up flags, if arrays overlap then the drives can not be
2544 for (i
= 0; i
< map
->num_members
; i
++) {
2545 struct imsm_disk
*disk
;
2548 disk
= get_imsm_disk(super
, get_imsm_disk_idx(map
, i
));
2549 status
= __le32_to_cpu(disk
->status
);
2550 status
|= CONFIGURED_DISK
;
2552 status
&= ~SPARE_DISK
;
2553 disk
->status
= __cpu_to_le32(status
);
2560 static void imsm_prepare_update(struct supertype
*st
,
2561 struct metadata_update
*update
)
2564 * Allocate space to hold new disk entries, raid-device entries or a
2565 * new mpb if necessary. We currently maintain an mpb large enough to
2566 * hold 2 subarrays for the given number of disks. This may not be
2567 * sufficient when reshaping.
2569 * FIX ME handle the reshape case.
2571 * The monitor will be able to safely change super->mpb by arranging
2572 * for it to be freed in check_update_queue(). I.e. the monitor thread
2573 * will start using the new pointer and the manager can continue to use
2574 * the old value until check_update_queue() runs.
2576 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
2579 case update_create_array
: {
2580 struct imsm_update_create_array
*u
= (void *) update
->buf
;
2581 size_t len
= sizeof_imsm_dev(&u
->dev
);
2583 update
->space
= malloc(len
);
2593 struct superswitch super_imsm
= {
2595 .examine_super
= examine_super_imsm
,
2596 .brief_examine_super
= brief_examine_super_imsm
,
2597 .detail_super
= detail_super_imsm
,
2598 .brief_detail_super
= brief_detail_super_imsm
,
2599 .write_init_super
= write_init_super_imsm
,
2601 .match_home
= match_home_imsm
,
2602 .uuid_from_super
= uuid_from_super_imsm
,
2603 .getinfo_super
= getinfo_super_imsm
,
2604 .update_super
= update_super_imsm
,
2606 .avail_size
= avail_size_imsm
,
2608 .compare_super
= compare_super_imsm
,
2610 .load_super
= load_super_imsm
,
2611 .init_super
= init_super_imsm
,
2612 .add_to_super
= add_to_super_imsm
,
2613 .store_super
= store_zero_imsm
,
2614 .free_super
= free_super_imsm
,
2615 .match_metadata_desc
= match_metadata_desc_imsm
,
2616 .container_content
= container_content_imsm
,
2618 .validate_geometry
= validate_geometry_imsm
,
2622 .open_new
= imsm_open_new
,
2623 .load_super
= load_super_imsm
,
2624 .set_array_state
= imsm_set_array_state
,
2625 .set_disk
= imsm_set_disk
,
2626 .sync_metadata
= imsm_sync_metadata
,
2627 .activate_spare
= imsm_activate_spare
,
2628 .process_update
= imsm_process_update
,
2629 .prepare_update
= imsm_prepare_update
,