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
45 #define MPB_ATTRIB_RAID0 __cpu_to_le32(0x00000001)
47 #define MPB_ATTRIB_RAID1 __cpu_to_le32(0x00000002)
49 #define MPB_ATTRIB_RAID10 __cpu_to_le32(0x00000004)
51 #define MPB_ATTRIB_RAID1E __cpu_to_le32(0x00000008)
53 #define MPB_ATTRIB_RAID5 __cpu_to_le32(0x00000010)
54 /* supports RAID CNG */
55 #define MPB_ATTRIB_RAIDCNG __cpu_to_le32(0x00000020)
56 /* supports expanded stripe sizes of 256K, 512K and 1MB */
57 #define MPB_ATTRIB_EXP_STRIPE_SIZE __cpu_to_le32(0x00000040)
59 /* The OROM Support RST Caching of Volumes */
60 #define MPB_ATTRIB_NVM __cpu_to_le32(0x02000000)
61 /* The OROM supports creating disks greater than 2TB */
62 #define MPB_ATTRIB_2TB_DISK __cpu_to_le32(0x04000000)
63 /* The OROM supports Bad Block Management */
64 #define MPB_ATTRIB_BBM __cpu_to_le32(0x08000000)
66 /* THe OROM Supports NVM Caching of Volumes */
67 #define MPB_ATTRIB_NEVER_USE2 __cpu_to_le32(0x10000000)
68 /* The OROM supports creating volumes greater than 2TB */
69 #define MPB_ATTRIB_2TB __cpu_to_le32(0x20000000)
70 /* originally for PMP, now it's wasted b/c. Never use this bit! */
71 #define MPB_ATTRIB_NEVER_USE __cpu_to_le32(0x40000000)
72 /* Verify MPB contents against checksum after reading MPB */
73 #define MPB_ATTRIB_CHECKSUM_VERIFY __cpu_to_le32(0x80000000)
75 /* Define all supported attributes that have to be accepted by mdadm
77 #define MPB_ATTRIB_SUPPORTED MPB_ATTRIB_CHECKSUM_VERIFY | \
79 MPB_ATTRIB_2TB_DISK | \
84 MPB_ATTRIB_EXP_STRIPE_SIZE
86 #define MPB_SECTOR_CNT 2210
87 #define IMSM_RESERVED_SECTORS 4096
88 #define SECT_PER_MB_SHIFT 11
90 /* Disk configuration info. */
91 #define IMSM_MAX_DEVICES 255
93 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
94 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
95 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
96 #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
97 #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
98 #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
99 __u32 status
; /* 0xF0 - 0xF3 */
100 __u32 owner_cfg_num
; /* which config 0,1,2... owns this disk */
101 #define IMSM_DISK_FILLERS 4
102 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
105 /* RAID map configuration infos. */
107 __u32 pba_of_lba0
; /* start address of partition */
108 __u32 blocks_per_member
;/* blocks per member */
109 __u32 num_data_stripes
; /* number of data stripes */
110 __u16 blocks_per_strip
;
111 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
112 #define IMSM_T_STATE_NORMAL 0
113 #define IMSM_T_STATE_UNINITIALIZED 1
114 #define IMSM_T_STATE_DEGRADED 2
115 #define IMSM_T_STATE_FAILED 3
117 #define IMSM_T_RAID0 0
118 #define IMSM_T_RAID1 1
119 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
120 __u8 num_members
; /* number of member disks */
121 __u8 num_domains
; /* number of parity domains */
122 __u8 failed_disk_num
; /* valid only when state is degraded */
124 __u32 filler
[7]; /* expansion area */
125 #define IMSM_ORD_REBUILD (1 << 24)
126 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
127 * top byte contains some flags
129 } __attribute__ ((packed
));
132 __u32 curr_migr_unit
;
133 __u32 checkpoint_id
; /* id to access curr_migr_unit */
134 __u8 migr_state
; /* Normal or Migrating */
136 #define MIGR_REBUILD 1
137 #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
138 #define MIGR_GEN_MIGR 3
139 #define MIGR_STATE_CHANGE 4
140 #define MIGR_REPAIR 5
141 __u8 migr_type
; /* Initializing, Rebuilding, ... */
143 __u8 fs_state
; /* fast-sync state for CnG (0xff == disabled) */
144 __u16 verify_errors
; /* number of mismatches */
145 __u16 bad_blocks
; /* number of bad blocks during verify */
147 struct imsm_map map
[1];
148 /* here comes another one if migr_state */
149 } __attribute__ ((packed
));
152 __u8 volume
[MAX_RAID_SERIAL_LEN
];
155 #define DEV_BOOTABLE __cpu_to_le32(0x01)
156 #define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
157 #define DEV_READ_COALESCING __cpu_to_le32(0x04)
158 #define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
159 #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
160 #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
161 #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
162 #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
163 #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
164 #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
165 #define DEV_CLONE_N_GO __cpu_to_le32(0x400)
166 #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
167 #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
168 __u32 status
; /* Persistent RaidDev status */
169 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
173 __u8 cng_master_disk
;
177 #define IMSM_DEV_FILLERS 10
178 __u32 filler
[IMSM_DEV_FILLERS
];
180 } __attribute__ ((packed
));
183 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
184 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
185 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
186 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
187 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
188 __u32 error_log_size
; /* 0x30 - 0x33 in bytes */
189 __u32 attributes
; /* 0x34 - 0x37 */
190 __u8 num_disks
; /* 0x38 Number of configured disks */
191 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
192 __u8 error_log_pos
; /* 0x3A */
193 __u8 fill
[1]; /* 0x3B */
194 __u32 cache_size
; /* 0x3c - 0x40 in mb */
195 __u32 orig_family_num
; /* 0x40 - 0x43 original family num */
196 __u32 pwr_cycle_count
; /* 0x44 - 0x47 simulated power cycle count for array */
197 __u32 bbm_log_size
; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
198 #define IMSM_FILLERS 35
199 __u32 filler
[IMSM_FILLERS
]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
200 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
201 /* here comes imsm_dev[num_raid_devs] */
202 /* here comes BBM logs */
203 } __attribute__ ((packed
));
205 #define BBM_LOG_MAX_ENTRIES 254
207 struct bbm_log_entry
{
208 __u64 defective_block_start
;
209 #define UNREADABLE 0xFFFFFFFF
210 __u32 spare_block_offset
;
211 __u16 remapped_marked_count
;
213 } __attribute__ ((__packed__
));
216 __u32 signature
; /* 0xABADB10C */
218 __u32 reserved_spare_block_count
; /* 0 */
219 __u32 reserved
; /* 0xFFFF */
220 __u64 first_spare_lba
;
221 struct bbm_log_entry mapped_block_entries
[BBM_LOG_MAX_ENTRIES
];
222 } __attribute__ ((__packed__
));
226 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
229 #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
231 #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
233 #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
234 * be recovered using srcMap */
235 #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
236 * already been migrated and must
237 * be recovered from checkpoint area */
239 __u32 rec_status
; /* Status used to determine how to restart
240 * migration in case it aborts
242 __u32 curr_migr_unit
; /* 0..numMigrUnits-1 */
243 __u32 family_num
; /* Family number of MPB
244 * containing the RaidDev
245 * that is migrating */
246 __u32 ascending_migr
; /* True if migrating in increasing
248 __u32 blocks_per_unit
; /* Num disk blocks per unit of operation */
249 __u32 dest_depth_per_unit
; /* Num member blocks each destMap
251 * advances per unit-of-operation */
252 __u32 ckpt_area_pba
; /* Pba of first block of ckpt copy area */
253 __u32 dest_1st_member_lba
; /* First member lba on first
254 * stripe of destination */
255 __u32 num_migr_units
; /* Total num migration units-of-op */
256 __u32 post_migr_vol_cap
; /* Size of volume after
257 * migration completes */
258 __u32 post_migr_vol_cap_hi
; /* Expansion space for LBA64 */
259 __u32 ckpt_read_disk_num
; /* Which member disk in destSubMap[0] the
260 * migration ckpt record was read from
261 * (for recovered migrations) */
262 } __attribute__ ((__packed__
));
264 static __u8
migr_type(struct imsm_dev
*dev
)
266 if (dev
->vol
.migr_type
== MIGR_VERIFY
&&
267 dev
->status
& DEV_VERIFY_AND_FIX
)
270 return dev
->vol
.migr_type
;
273 static void set_migr_type(struct imsm_dev
*dev
, __u8 migr_type
)
275 /* for compatibility with older oroms convert MIGR_REPAIR, into
276 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
278 if (migr_type
== MIGR_REPAIR
) {
279 dev
->vol
.migr_type
= MIGR_VERIFY
;
280 dev
->status
|= DEV_VERIFY_AND_FIX
;
282 dev
->vol
.migr_type
= migr_type
;
283 dev
->status
&= ~DEV_VERIFY_AND_FIX
;
287 static unsigned int sector_count(__u32 bytes
)
289 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
292 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
294 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
298 struct imsm_dev
*dev
;
299 struct intel_dev
*next
;
304 enum sys_dev_type type
;
307 struct intel_hba
*next
;
314 /* internal representation of IMSM metadata */
317 void *buf
; /* O_DIRECT buffer for reading/writing metadata */
318 struct imsm_super
*anchor
; /* immovable parameters */
321 void *migr_rec_buf
; /* buffer for I/O operations */
322 struct migr_record
*migr_rec
; /* migration record */
324 size_t len
; /* size of the 'buf' allocation */
325 void *next_buf
; /* for realloc'ing buf from the manager */
327 int updates_pending
; /* count of pending updates for mdmon */
328 int current_vol
; /* index of raid device undergoing creation */
329 __u32 create_offset
; /* common start for 'current_vol' */
330 __u32 random
; /* random data for seeding new family numbers */
331 struct intel_dev
*devlist
;
335 __u8 serial
[MAX_RAID_SERIAL_LEN
];
338 struct imsm_disk disk
;
341 struct extent
*e
; /* for determining freespace @ create */
342 int raiddisk
; /* slot to fill in autolayout */
345 struct dl
*disk_mgmt_list
; /* list of disks to add/remove while mdmon
347 struct dl
*missing
; /* disks removed while we weren't looking */
348 struct bbm_log
*bbm_log
;
349 struct intel_hba
*hba
; /* device path of the raid controller for this metadata */
350 const struct imsm_orom
*orom
; /* platform firmware support */
351 struct intel_super
*next
; /* (temp) list for disambiguating family_num */
355 struct imsm_disk disk
;
356 #define IMSM_UNKNOWN_OWNER (-1)
358 struct intel_disk
*next
;
362 unsigned long long start
, size
;
365 /* definitions of reshape process types */
366 enum imsm_reshape_type
{
371 /* definition of messages passed to imsm_process_update */
372 enum imsm_update_type
{
373 update_activate_spare
,
377 update_add_remove_disk
,
378 update_reshape_container_disks
,
379 update_reshape_migration
,
381 update_general_migration_checkpoint
,
384 struct imsm_update_activate_spare
{
385 enum imsm_update_type type
;
389 struct imsm_update_activate_spare
*next
;
402 enum takeover_direction
{
406 struct imsm_update_takeover
{
407 enum imsm_update_type type
;
409 enum takeover_direction direction
;
412 struct imsm_update_reshape
{
413 enum imsm_update_type type
;
417 int new_disks
[1]; /* new_raid_disks - old_raid_disks makedev number */
420 struct imsm_update_reshape_migration
{
421 enum imsm_update_type type
;
424 /* fields for array migration changes
431 int new_disks
[1]; /* new_raid_disks - old_raid_disks makedev number */
434 struct imsm_update_general_migration_checkpoint
{
435 enum imsm_update_type type
;
436 __u32 curr_migr_unit
;
440 __u8 serial
[MAX_RAID_SERIAL_LEN
];
443 struct imsm_update_create_array
{
444 enum imsm_update_type type
;
449 struct imsm_update_kill_array
{
450 enum imsm_update_type type
;
454 struct imsm_update_rename_array
{
455 enum imsm_update_type type
;
456 __u8 name
[MAX_RAID_SERIAL_LEN
];
460 struct imsm_update_add_remove_disk
{
461 enum imsm_update_type type
;
465 static const char *_sys_dev_type
[] = {
466 [SYS_DEV_UNKNOWN
] = "Unknown",
467 [SYS_DEV_SAS
] = "SAS",
468 [SYS_DEV_SATA
] = "SATA"
471 const char *get_sys_dev_type(enum sys_dev_type type
)
473 if (type
>= SYS_DEV_MAX
)
474 type
= SYS_DEV_UNKNOWN
;
476 return _sys_dev_type
[type
];
479 static struct intel_hba
* alloc_intel_hba(struct sys_dev
*device
)
481 struct intel_hba
*result
= malloc(sizeof(*result
));
483 result
->type
= device
->type
;
484 result
->path
= strdup(device
->path
);
486 if (result
->path
&& (result
->pci_id
= strrchr(result
->path
, '/')) != NULL
)
492 static struct intel_hba
* find_intel_hba(struct intel_hba
*hba
, struct sys_dev
*device
)
494 struct intel_hba
*result
=NULL
;
495 for (result
= hba
; result
; result
= result
->next
) {
496 if (result
->type
== device
->type
&& strcmp(result
->path
, device
->path
) == 0)
502 static int attach_hba_to_super(struct intel_super
*super
, struct sys_dev
*device
)
504 struct intel_hba
*hba
;
506 /* check if disk attached to Intel HBA */
507 hba
= find_intel_hba(super
->hba
, device
);
510 /* Check if HBA is already attached to super */
511 if (super
->hba
== NULL
) {
512 super
->hba
= alloc_intel_hba(device
);
517 /* Intel metadata allows for all disks attached to the same type HBA.
518 * Do not sypport odf HBA types mixing
520 if (device
->type
!= hba
->type
)
526 hba
->next
= alloc_intel_hba(device
);
530 static struct sys_dev
* find_disk_attached_hba(int fd
, const char *devname
)
532 struct sys_dev
*list
, *elem
, *prev
;
535 if ((list
= find_intel_devices()) == NULL
)
539 disk_path
= (char *) devname
;
541 disk_path
= diskfd_to_devpath(fd
);
548 for (prev
= NULL
, elem
= list
; elem
; prev
= elem
, elem
= elem
->next
) {
549 if (path_attached_to_hba(disk_path
, elem
->path
)) {
553 prev
->next
= elem
->next
;
555 if (disk_path
!= devname
)
561 if (disk_path
!= devname
)
569 static int find_intel_hba_capability(int fd
, struct intel_super
*super
,
572 static struct supertype
*match_metadata_desc_imsm(char *arg
)
574 struct supertype
*st
;
576 if (strcmp(arg
, "imsm") != 0 &&
577 strcmp(arg
, "default") != 0
581 st
= malloc(sizeof(*st
));
584 memset(st
, 0, sizeof(*st
));
585 st
->container_dev
= NoMdDev
;
586 st
->ss
= &super_imsm
;
587 st
->max_devs
= IMSM_MAX_DEVICES
;
588 st
->minor_version
= 0;
594 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
596 return &mpb
->sig
[MPB_SIG_LEN
];
600 /* retrieve a disk directly from the anchor when the anchor is known to be
601 * up-to-date, currently only at load time
603 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
605 if (index
>= mpb
->num_disks
)
607 return &mpb
->disk
[index
];
610 /* retrieve the disk description based on a index of the disk
613 static struct dl
*get_imsm_dl_disk(struct intel_super
*super
, __u8 index
)
617 for (d
= super
->disks
; d
; d
= d
->next
)
618 if (d
->index
== index
)
623 /* retrieve a disk from the parsed metadata */
624 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
628 dl
= get_imsm_dl_disk(super
, index
);
635 /* generate a checksum directly from the anchor when the anchor is known to be
636 * up-to-date, currently only at load or write_super after coalescing
638 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
640 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
641 __u32
*p
= (__u32
*) mpb
;
645 sum
+= __le32_to_cpu(*p
);
649 return sum
- __le32_to_cpu(mpb
->check_sum
);
652 static size_t sizeof_imsm_map(struct imsm_map
*map
)
654 return sizeof(struct imsm_map
) + sizeof(__u32
) * (map
->num_members
- 1);
657 struct imsm_map
*get_imsm_map(struct imsm_dev
*dev
, int second_map
)
659 /* A device can have 2 maps if it is in the middle of a migration.
661 * 0 - we return the first map
662 * 1 - we return the second map if it exists, else NULL
663 * -1 - we return the second map if it exists, else the first
665 struct imsm_map
*map
= &dev
->vol
.map
[0];
667 if (second_map
== 1 && !dev
->vol
.migr_state
)
669 else if (second_map
== 1 ||
670 (second_map
< 0 && dev
->vol
.migr_state
)) {
673 return ptr
+ sizeof_imsm_map(map
);
679 /* return the size of the device.
680 * migr_state increases the returned size if map[0] were to be duplicated
682 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
, int migr_state
)
684 size_t size
= sizeof(*dev
) - sizeof(struct imsm_map
) +
685 sizeof_imsm_map(get_imsm_map(dev
, 0));
687 /* migrating means an additional map */
688 if (dev
->vol
.migr_state
)
689 size
+= sizeof_imsm_map(get_imsm_map(dev
, 1));
691 size
+= sizeof_imsm_map(get_imsm_map(dev
, 0));
697 /* retrieve disk serial number list from a metadata update */
698 static struct disk_info
*get_disk_info(struct imsm_update_create_array
*update
)
701 struct disk_info
*inf
;
703 inf
= u
+ sizeof(*update
) - sizeof(struct imsm_dev
) +
704 sizeof_imsm_dev(&update
->dev
, 0);
710 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
716 if (index
>= mpb
->num_raid_devs
)
719 /* devices start after all disks */
720 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
722 for (i
= 0; i
<= index
; i
++)
724 return _mpb
+ offset
;
726 offset
+= sizeof_imsm_dev(_mpb
+ offset
, 0);
731 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
733 struct intel_dev
*dv
;
735 if (index
>= super
->anchor
->num_raid_devs
)
737 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
738 if (dv
->index
== index
)
746 * == 1 get second map
747 * == -1 than get map according to the current migr_state
749 static __u32
get_imsm_ord_tbl_ent(struct imsm_dev
*dev
,
753 struct imsm_map
*map
;
755 map
= get_imsm_map(dev
, second_map
);
757 /* top byte identifies disk under rebuild */
758 return __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
761 #define ord_to_idx(ord) (((ord) << 8) >> 8)
762 static __u32
get_imsm_disk_idx(struct imsm_dev
*dev
, int slot
, int second_map
)
764 __u32 ord
= get_imsm_ord_tbl_ent(dev
, slot
, second_map
);
766 return ord_to_idx(ord
);
769 static void set_imsm_ord_tbl_ent(struct imsm_map
*map
, int slot
, __u32 ord
)
771 map
->disk_ord_tbl
[slot
] = __cpu_to_le32(ord
);
774 static int get_imsm_disk_slot(struct imsm_map
*map
, unsigned idx
)
779 for (slot
= 0; slot
< map
->num_members
; slot
++) {
780 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
781 if (ord_to_idx(ord
) == idx
)
788 static int get_imsm_raid_level(struct imsm_map
*map
)
790 if (map
->raid_level
== 1) {
791 if (map
->num_members
== 2)
797 return map
->raid_level
;
800 static int cmp_extent(const void *av
, const void *bv
)
802 const struct extent
*a
= av
;
803 const struct extent
*b
= bv
;
804 if (a
->start
< b
->start
)
806 if (a
->start
> b
->start
)
811 static int count_memberships(struct dl
*dl
, struct intel_super
*super
)
816 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
817 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
818 struct imsm_map
*map
= get_imsm_map(dev
, 0);
820 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
827 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
829 /* find a list of used extents on the given physical device */
830 struct extent
*rv
, *e
;
832 int memberships
= count_memberships(dl
, super
);
833 __u32 reservation
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
835 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
840 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
841 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
842 struct imsm_map
*map
= get_imsm_map(dev
, 0);
844 if (get_imsm_disk_slot(map
, dl
->index
) >= 0) {
845 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
846 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
850 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
852 /* determine the start of the metadata
853 * when no raid devices are defined use the default
854 * ...otherwise allow the metadata to truncate the value
855 * as is the case with older versions of imsm
858 struct extent
*last
= &rv
[memberships
- 1];
861 remainder
= __le32_to_cpu(dl
->disk
.total_blocks
) -
862 (last
->start
+ last
->size
);
863 /* round down to 1k block to satisfy precision of the kernel
867 /* make sure remainder is still sane */
868 if (remainder
< (unsigned)ROUND_UP(super
->len
, 512) >> 9)
869 remainder
= ROUND_UP(super
->len
, 512) >> 9;
870 if (reservation
> remainder
)
871 reservation
= remainder
;
873 e
->start
= __le32_to_cpu(dl
->disk
.total_blocks
) - reservation
;
878 /* try to determine how much space is reserved for metadata from
879 * the last get_extents() entry, otherwise fallback to the
882 static __u32
imsm_reserved_sectors(struct intel_super
*super
, struct dl
*dl
)
888 /* for spares just return a minimal reservation which will grow
889 * once the spare is picked up by an array
892 return MPB_SECTOR_CNT
;
894 e
= get_extents(super
, dl
);
896 return MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
898 /* scroll to last entry */
899 for (i
= 0; e
[i
].size
; i
++)
902 rv
= __le32_to_cpu(dl
->disk
.total_blocks
) - e
[i
].start
;
909 static int is_spare(struct imsm_disk
*disk
)
911 return (disk
->status
& SPARE_DISK
) == SPARE_DISK
;
914 static int is_configured(struct imsm_disk
*disk
)
916 return (disk
->status
& CONFIGURED_DISK
) == CONFIGURED_DISK
;
919 static int is_failed(struct imsm_disk
*disk
)
921 return (disk
->status
& FAILED_DISK
) == FAILED_DISK
;
924 /* Return minimum size of a spare that can be used in this array*/
925 static unsigned long long min_acceptable_spare_size_imsm(struct supertype
*st
)
927 struct intel_super
*super
= st
->sb
;
931 unsigned long long rv
= 0;
935 /* find first active disk in array */
937 while (dl
&& (is_failed(&dl
->disk
) || dl
->index
== -1))
941 /* find last lba used by subarrays */
942 e
= get_extents(super
, dl
);
945 for (i
= 0; e
[i
].size
; i
++)
948 rv
= e
[i
-1].start
+ e
[i
-1].size
;
950 /* add the amount of space needed for metadata */
951 rv
= rv
+ MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
956 static __u64
blocks_per_migr_unit(struct intel_super
*super
,
957 struct imsm_dev
*dev
);
959 static void print_imsm_dev(struct intel_super
*super
,
960 struct imsm_dev
*dev
,
966 struct imsm_map
*map
= get_imsm_map(dev
, 0);
967 struct imsm_map
*map2
= get_imsm_map(dev
, 1);
971 printf("[%.16s]:\n", dev
->volume
);
972 printf(" UUID : %s\n", uuid
);
973 printf(" RAID Level : %d", get_imsm_raid_level(map
));
975 printf(" <-- %d", get_imsm_raid_level(map2
));
977 printf(" Members : %d", map
->num_members
);
979 printf(" <-- %d", map2
->num_members
);
981 printf(" Slots : [");
982 for (i
= 0; i
< map
->num_members
; i
++) {
983 ord
= get_imsm_ord_tbl_ent(dev
, i
, 0);
984 printf("%s", ord
& IMSM_ORD_REBUILD
? "_" : "U");
989 for (i
= 0; i
< map2
->num_members
; i
++) {
990 ord
= get_imsm_ord_tbl_ent(dev
, i
, 1);
991 printf("%s", ord
& IMSM_ORD_REBUILD
? "_" : "U");
996 printf(" Failed disk : ");
997 if (map
->failed_disk_num
== 0xff)
1000 printf("%i", map
->failed_disk_num
);
1002 slot
= get_imsm_disk_slot(map
, disk_idx
);
1004 ord
= get_imsm_ord_tbl_ent(dev
, slot
, -1);
1005 printf(" This Slot : %d%s\n", slot
,
1006 ord
& IMSM_ORD_REBUILD
? " (out-of-sync)" : "");
1008 printf(" This Slot : ?\n");
1009 sz
= __le32_to_cpu(dev
->size_high
);
1011 sz
+= __le32_to_cpu(dev
->size_low
);
1012 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
1013 human_size(sz
* 512));
1014 sz
= __le32_to_cpu(map
->blocks_per_member
);
1015 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
1016 human_size(sz
* 512));
1017 printf(" Sector Offset : %u\n",
1018 __le32_to_cpu(map
->pba_of_lba0
));
1019 printf(" Num Stripes : %u\n",
1020 __le32_to_cpu(map
->num_data_stripes
));
1021 printf(" Chunk Size : %u KiB",
1022 __le16_to_cpu(map
->blocks_per_strip
) / 2);
1024 printf(" <-- %u KiB",
1025 __le16_to_cpu(map2
->blocks_per_strip
) / 2);
1027 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
1028 printf(" Migrate State : ");
1029 if (dev
->vol
.migr_state
) {
1030 if (migr_type(dev
) == MIGR_INIT
)
1031 printf("initialize\n");
1032 else if (migr_type(dev
) == MIGR_REBUILD
)
1033 printf("rebuild\n");
1034 else if (migr_type(dev
) == MIGR_VERIFY
)
1036 else if (migr_type(dev
) == MIGR_GEN_MIGR
)
1037 printf("general migration\n");
1038 else if (migr_type(dev
) == MIGR_STATE_CHANGE
)
1039 printf("state change\n");
1040 else if (migr_type(dev
) == MIGR_REPAIR
)
1043 printf("<unknown:%d>\n", migr_type(dev
));
1046 printf(" Map State : %s", map_state_str
[map
->map_state
]);
1047 if (dev
->vol
.migr_state
) {
1048 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1050 printf(" <-- %s", map_state_str
[map
->map_state
]);
1051 printf("\n Checkpoint : %u (%llu)",
1052 __le32_to_cpu(dev
->vol
.curr_migr_unit
),
1053 (unsigned long long)blocks_per_migr_unit(super
, dev
));
1056 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
1059 static void print_imsm_disk(struct imsm_super
*mpb
, int index
, __u32 reserved
)
1061 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
1062 char str
[MAX_RAID_SERIAL_LEN
+ 1];
1065 if (index
< 0 || !disk
)
1069 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
1070 printf(" Disk%02d Serial : %s\n", index
, str
);
1071 printf(" State :%s%s%s\n", is_spare(disk
) ? " spare" : "",
1072 is_configured(disk
) ? " active" : "",
1073 is_failed(disk
) ? " failed" : "");
1074 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
1075 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
1076 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
1077 human_size(sz
* 512));
1080 static int is_gen_migration(struct imsm_dev
*dev
);
1082 void examine_migr_rec_imsm(struct intel_super
*super
)
1084 struct migr_record
*migr_rec
= super
->migr_rec
;
1085 struct imsm_super
*mpb
= super
->anchor
;
1088 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1089 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1090 if (is_gen_migration(dev
) == 0)
1093 printf("\nMigration Record Information:");
1094 if (super
->disks
->index
> 1) {
1095 printf(" Empty\n ");
1096 printf("Examine one of first two disks in array\n");
1099 printf("\n Status : ");
1100 if (__le32_to_cpu(migr_rec
->rec_status
) == UNIT_SRC_NORMAL
)
1103 printf("Contains Data\n");
1104 printf(" Current Unit : %u\n",
1105 __le32_to_cpu(migr_rec
->curr_migr_unit
));
1106 printf(" Family : %u\n",
1107 __le32_to_cpu(migr_rec
->family_num
));
1108 printf(" Ascending : %u\n",
1109 __le32_to_cpu(migr_rec
->ascending_migr
));
1110 printf(" Blocks Per Unit : %u\n",
1111 __le32_to_cpu(migr_rec
->blocks_per_unit
));
1112 printf(" Dest. Depth Per Unit : %u\n",
1113 __le32_to_cpu(migr_rec
->dest_depth_per_unit
));
1114 printf(" Checkpoint Area pba : %u\n",
1115 __le32_to_cpu(migr_rec
->ckpt_area_pba
));
1116 printf(" First member lba : %u\n",
1117 __le32_to_cpu(migr_rec
->dest_1st_member_lba
));
1118 printf(" Total Number of Units : %u\n",
1119 __le32_to_cpu(migr_rec
->num_migr_units
));
1120 printf(" Size of volume : %u\n",
1121 __le32_to_cpu(migr_rec
->post_migr_vol_cap
));
1122 printf(" Expansion space for LBA64 : %u\n",
1123 __le32_to_cpu(migr_rec
->post_migr_vol_cap_hi
));
1124 printf(" Record was read from : %u\n",
1125 __le32_to_cpu(migr_rec
->ckpt_read_disk_num
));
1130 #endif /* MDASSEMBLE */
1131 /*******************************************************************************
1132 * function: imsm_check_attributes
1133 * Description: Function checks if features represented by attributes flags
1134 * are supported by mdadm.
1136 * attributes - Attributes read from metadata
1138 * 0 - passed attributes contains unsupported features flags
1139 * 1 - all features are supported
1140 ******************************************************************************/
1141 static int imsm_check_attributes(__u32 attributes
)
1144 __u32 not_supported
= (MPB_ATTRIB_SUPPORTED
)^0xffffffff;
1146 not_supported
&= attributes
;
1147 if (not_supported
) {
1148 fprintf(stderr
, Name
"(IMSM): Unsupported attributes : %x\n", not_supported
);
1149 if (not_supported
& MPB_ATTRIB_CHECKSUM_VERIFY
) {
1150 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1151 not_supported
^= MPB_ATTRIB_CHECKSUM_VERIFY
;
1153 if (not_supported
& MPB_ATTRIB_2TB
) {
1154 dprintf("\t\tMPB_ATTRIB_2TB\n");
1155 not_supported
^= MPB_ATTRIB_2TB
;
1157 if (not_supported
& MPB_ATTRIB_RAID0
) {
1158 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1159 not_supported
^= MPB_ATTRIB_RAID0
;
1161 if (not_supported
& MPB_ATTRIB_RAID1
) {
1162 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1163 not_supported
^= MPB_ATTRIB_RAID1
;
1165 if (not_supported
& MPB_ATTRIB_RAID10
) {
1166 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1167 not_supported
^= MPB_ATTRIB_RAID10
;
1169 if (not_supported
& MPB_ATTRIB_RAID1E
) {
1170 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1171 not_supported
^= MPB_ATTRIB_RAID1E
;
1173 if (not_supported
& MPB_ATTRIB_RAID5
) {
1174 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1175 not_supported
^= MPB_ATTRIB_RAID5
;
1177 if (not_supported
& MPB_ATTRIB_RAIDCNG
) {
1178 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1179 not_supported
^= MPB_ATTRIB_RAIDCNG
;
1181 if (not_supported
& MPB_ATTRIB_BBM
) {
1182 dprintf("\t\tMPB_ATTRIB_BBM\n");
1183 not_supported
^= MPB_ATTRIB_BBM
;
1185 if (not_supported
& MPB_ATTRIB_CHECKSUM_VERIFY
) {
1186 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1187 not_supported
^= MPB_ATTRIB_CHECKSUM_VERIFY
;
1189 if (not_supported
& MPB_ATTRIB_EXP_STRIPE_SIZE
) {
1190 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1191 not_supported
^= MPB_ATTRIB_EXP_STRIPE_SIZE
;
1193 if (not_supported
& MPB_ATTRIB_2TB_DISK
) {
1194 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1195 not_supported
^= MPB_ATTRIB_2TB_DISK
;
1197 if (not_supported
& MPB_ATTRIB_NEVER_USE2
) {
1198 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1199 not_supported
^= MPB_ATTRIB_NEVER_USE2
;
1201 if (not_supported
& MPB_ATTRIB_NEVER_USE
) {
1202 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1203 not_supported
^= MPB_ATTRIB_NEVER_USE
;
1207 dprintf(Name
"(IMSM): Unknown attributes : %x\n", not_supported
);
1216 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
, char *map
);
1218 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
1220 struct intel_super
*super
= st
->sb
;
1221 struct imsm_super
*mpb
= super
->anchor
;
1222 char str
[MAX_SIGNATURE_LENGTH
];
1227 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
1230 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
1231 printf(" Magic : %s\n", str
);
1232 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
1233 printf(" Version : %s\n", get_imsm_version(mpb
));
1234 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb
->orig_family_num
));
1235 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
1236 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
1237 printf(" Attributes : ");
1238 if (imsm_check_attributes(mpb
->attributes
))
1239 printf("All supported\n");
1241 printf("not supported\n");
1242 getinfo_super_imsm(st
, &info
, NULL
);
1243 fname_from_uuid(st
, &info
, nbuf
, ':');
1244 printf(" UUID : %s\n", nbuf
+ 5);
1245 sum
= __le32_to_cpu(mpb
->check_sum
);
1246 printf(" Checksum : %08x %s\n", sum
,
1247 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
1248 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
1249 printf(" Disks : %d\n", mpb
->num_disks
);
1250 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
1251 print_imsm_disk(mpb
, super
->disks
->index
, reserved
);
1252 if (super
->bbm_log
) {
1253 struct bbm_log
*log
= super
->bbm_log
;
1256 printf("Bad Block Management Log:\n");
1257 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
1258 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
1259 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
1260 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
1261 printf(" First Spare : %llx\n",
1262 (unsigned long long) __le64_to_cpu(log
->first_spare_lba
));
1264 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1266 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1268 super
->current_vol
= i
;
1269 getinfo_super_imsm(st
, &info
, NULL
);
1270 fname_from_uuid(st
, &info
, nbuf
, ':');
1271 print_imsm_dev(super
, dev
, nbuf
+ 5, super
->disks
->index
);
1273 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1274 if (i
== super
->disks
->index
)
1276 print_imsm_disk(mpb
, i
, reserved
);
1278 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1279 struct imsm_disk
*disk
;
1280 char str
[MAX_RAID_SERIAL_LEN
+ 1];
1288 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
1289 printf(" Disk Serial : %s\n", str
);
1290 printf(" State :%s%s%s\n", is_spare(disk
) ? " spare" : "",
1291 is_configured(disk
) ? " active" : "",
1292 is_failed(disk
) ? " failed" : "");
1293 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
1294 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
1295 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
1296 human_size(sz
* 512));
1299 examine_migr_rec_imsm(super
);
1302 static void brief_examine_super_imsm(struct supertype
*st
, int verbose
)
1304 /* We just write a generic IMSM ARRAY entry */
1307 struct intel_super
*super
= st
->sb
;
1309 if (!super
->anchor
->num_raid_devs
) {
1310 printf("ARRAY metadata=imsm\n");
1314 getinfo_super_imsm(st
, &info
, NULL
);
1315 fname_from_uuid(st
, &info
, nbuf
, ':');
1316 printf("ARRAY metadata=imsm UUID=%s\n", nbuf
+ 5);
1319 static void brief_examine_subarrays_imsm(struct supertype
*st
, int verbose
)
1321 /* We just write a generic IMSM ARRAY entry */
1325 struct intel_super
*super
= st
->sb
;
1328 if (!super
->anchor
->num_raid_devs
)
1331 getinfo_super_imsm(st
, &info
, NULL
);
1332 fname_from_uuid(st
, &info
, nbuf
, ':');
1333 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1334 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
1336 super
->current_vol
= i
;
1337 getinfo_super_imsm(st
, &info
, NULL
);
1338 fname_from_uuid(st
, &info
, nbuf1
, ':');
1339 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
1340 dev
->volume
, nbuf
+ 5, i
, nbuf1
+ 5);
1344 static void export_examine_super_imsm(struct supertype
*st
)
1346 struct intel_super
*super
= st
->sb
;
1347 struct imsm_super
*mpb
= super
->anchor
;
1351 getinfo_super_imsm(st
, &info
, NULL
);
1352 fname_from_uuid(st
, &info
, nbuf
, ':');
1353 printf("MD_METADATA=imsm\n");
1354 printf("MD_LEVEL=container\n");
1355 printf("MD_UUID=%s\n", nbuf
+5);
1356 printf("MD_DEVICES=%u\n", mpb
->num_disks
);
1359 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
1364 getinfo_super_imsm(st
, &info
, NULL
);
1365 fname_from_uuid(st
, &info
, nbuf
, ':');
1366 printf("\n UUID : %s\n", nbuf
+ 5);
1369 static void brief_detail_super_imsm(struct supertype
*st
)
1373 getinfo_super_imsm(st
, &info
, NULL
);
1374 fname_from_uuid(st
, &info
, nbuf
, ':');
1375 printf(" UUID=%s", nbuf
+ 5);
1378 static int imsm_read_serial(int fd
, char *devname
, __u8
*serial
);
1379 static void fd2devname(int fd
, char *name
);
1381 static int ahci_enumerate_ports(const char *hba_path
, int port_count
, int host_base
, int verbose
)
1383 /* dump an unsorted list of devices attached to AHCI Intel storage
1384 * controller, as well as non-connected ports
1386 int hba_len
= strlen(hba_path
) + 1;
1391 unsigned long port_mask
= (1 << port_count
) - 1;
1393 if (port_count
> (int)sizeof(port_mask
) * 8) {
1395 fprintf(stderr
, Name
": port_count %d out of range\n", port_count
);
1399 /* scroll through /sys/dev/block looking for devices attached to
1402 dir
= opendir("/sys/dev/block");
1403 for (ent
= dir
? readdir(dir
) : NULL
; ent
; ent
= readdir(dir
)) {
1414 if (sscanf(ent
->d_name
, "%d:%d", &major
, &minor
) != 2)
1416 path
= devt_to_devpath(makedev(major
, minor
));
1419 if (!path_attached_to_hba(path
, hba_path
)) {
1425 /* retrieve the scsi device type */
1426 if (asprintf(&device
, "/sys/dev/block/%d:%d/device/xxxxxxx", major
, minor
) < 0) {
1428 fprintf(stderr
, Name
": failed to allocate 'device'\n");
1432 sprintf(device
, "/sys/dev/block/%d:%d/device/type", major
, minor
);
1433 if (load_sys(device
, buf
) != 0) {
1435 fprintf(stderr
, Name
": failed to read device type for %s\n",
1441 type
= strtoul(buf
, NULL
, 10);
1443 /* if it's not a disk print the vendor and model */
1444 if (!(type
== 0 || type
== 7 || type
== 14)) {
1447 sprintf(device
, "/sys/dev/block/%d:%d/device/vendor", major
, minor
);
1448 if (load_sys(device
, buf
) == 0) {
1449 strncpy(vendor
, buf
, sizeof(vendor
));
1450 vendor
[sizeof(vendor
) - 1] = '\0';
1451 c
= (char *) &vendor
[sizeof(vendor
) - 1];
1452 while (isspace(*c
) || *c
== '\0')
1456 sprintf(device
, "/sys/dev/block/%d:%d/device/model", major
, minor
);
1457 if (load_sys(device
, buf
) == 0) {
1458 strncpy(model
, buf
, sizeof(model
));
1459 model
[sizeof(model
) - 1] = '\0';
1460 c
= (char *) &model
[sizeof(model
) - 1];
1461 while (isspace(*c
) || *c
== '\0')
1465 if (vendor
[0] && model
[0])
1466 sprintf(buf
, "%.64s %.64s", vendor
, model
);
1468 switch (type
) { /* numbers from hald/linux/device.c */
1469 case 1: sprintf(buf
, "tape"); break;
1470 case 2: sprintf(buf
, "printer"); break;
1471 case 3: sprintf(buf
, "processor"); break;
1473 case 5: sprintf(buf
, "cdrom"); break;
1474 case 6: sprintf(buf
, "scanner"); break;
1475 case 8: sprintf(buf
, "media_changer"); break;
1476 case 9: sprintf(buf
, "comm"); break;
1477 case 12: sprintf(buf
, "raid"); break;
1478 default: sprintf(buf
, "unknown");
1484 /* chop device path to 'host%d' and calculate the port number */
1485 c
= strchr(&path
[hba_len
], '/');
1488 fprintf(stderr
, Name
": %s - invalid path name\n", path
+ hba_len
);
1493 if (sscanf(&path
[hba_len
], "host%d", &port
) == 1)
1497 *c
= '/'; /* repair the full string */
1498 fprintf(stderr
, Name
": failed to determine port number for %s\n",
1505 /* mark this port as used */
1506 port_mask
&= ~(1 << port
);
1508 /* print out the device information */
1510 printf(" Port%d : - non-disk device (%s) -\n", port
, buf
);
1514 fd
= dev_open(ent
->d_name
, O_RDONLY
);
1516 printf(" Port%d : - disk info unavailable -\n", port
);
1518 fd2devname(fd
, buf
);
1519 printf(" Port%d : %s", port
, buf
);
1520 if (imsm_read_serial(fd
, NULL
, (__u8
*) buf
) == 0)
1521 printf(" (%s)\n", buf
);
1536 for (i
= 0; i
< port_count
; i
++)
1537 if (port_mask
& (1 << i
))
1538 printf(" Port%d : - no device attached -\n", i
);
1544 static void print_found_intel_controllers(struct sys_dev
*elem
)
1546 for (; elem
; elem
= elem
->next
) {
1547 fprintf(stderr
, Name
": found Intel(R) ");
1548 if (elem
->type
== SYS_DEV_SATA
)
1549 fprintf(stderr
, "SATA ");
1550 else if (elem
->type
== SYS_DEV_SAS
)
1551 fprintf(stderr
, "SAS ");
1552 fprintf(stderr
, "RAID controller");
1554 fprintf(stderr
, " at %s", elem
->pci_id
);
1555 fprintf(stderr
, ".\n");
1560 static int ahci_get_port_count(const char *hba_path
, int *port_count
)
1567 if ((dir
= opendir(hba_path
)) == NULL
)
1570 for (ent
= readdir(dir
); ent
; ent
= readdir(dir
)) {
1573 if (sscanf(ent
->d_name
, "host%d", &host
) != 1)
1575 if (*port_count
== 0)
1577 else if (host
< host_base
)
1580 if (host
+ 1 > *port_count
+ host_base
)
1581 *port_count
= host
+ 1 - host_base
;
1587 static void print_imsm_capability(const struct imsm_orom
*orom
)
1589 printf(" Platform : Intel(R) Matrix Storage Manager\n");
1590 printf(" Version : %d.%d.%d.%d\n", orom
->major_ver
, orom
->minor_ver
,
1591 orom
->hotfix_ver
, orom
->build
);
1592 printf(" RAID Levels :%s%s%s%s%s\n",
1593 imsm_orom_has_raid0(orom
) ? " raid0" : "",
1594 imsm_orom_has_raid1(orom
) ? " raid1" : "",
1595 imsm_orom_has_raid1e(orom
) ? " raid1e" : "",
1596 imsm_orom_has_raid10(orom
) ? " raid10" : "",
1597 imsm_orom_has_raid5(orom
) ? " raid5" : "");
1598 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
1599 imsm_orom_has_chunk(orom
, 2) ? " 2k" : "",
1600 imsm_orom_has_chunk(orom
, 4) ? " 4k" : "",
1601 imsm_orom_has_chunk(orom
, 8) ? " 8k" : "",
1602 imsm_orom_has_chunk(orom
, 16) ? " 16k" : "",
1603 imsm_orom_has_chunk(orom
, 32) ? " 32k" : "",
1604 imsm_orom_has_chunk(orom
, 64) ? " 64k" : "",
1605 imsm_orom_has_chunk(orom
, 128) ? " 128k" : "",
1606 imsm_orom_has_chunk(orom
, 256) ? " 256k" : "",
1607 imsm_orom_has_chunk(orom
, 512) ? " 512k" : "",
1608 imsm_orom_has_chunk(orom
, 1024*1) ? " 1M" : "",
1609 imsm_orom_has_chunk(orom
, 1024*2) ? " 2M" : "",
1610 imsm_orom_has_chunk(orom
, 1024*4) ? " 4M" : "",
1611 imsm_orom_has_chunk(orom
, 1024*8) ? " 8M" : "",
1612 imsm_orom_has_chunk(orom
, 1024*16) ? " 16M" : "",
1613 imsm_orom_has_chunk(orom
, 1024*32) ? " 32M" : "",
1614 imsm_orom_has_chunk(orom
, 1024*64) ? " 64M" : "");
1615 printf(" Max Disks : %d\n", orom
->tds
);
1616 printf(" Max Volumes : %d\n", orom
->vpa
);
1620 static int detail_platform_imsm(int verbose
, int enumerate_only
)
1622 /* There are two components to imsm platform support, the ahci SATA
1623 * controller and the option-rom. To find the SATA controller we
1624 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
1625 * controller with the Intel vendor id is present. This approach
1626 * allows mdadm to leverage the kernel's ahci detection logic, with the
1627 * caveat that if ahci.ko is not loaded mdadm will not be able to
1628 * detect platform raid capabilities. The option-rom resides in a
1629 * platform "Adapter ROM". We scan for its signature to retrieve the
1630 * platform capabilities. If raid support is disabled in the BIOS the
1631 * option-rom capability structure will not be available.
1633 const struct imsm_orom
*orom
;
1634 struct sys_dev
*list
, *hba
;
1639 if (enumerate_only
) {
1640 if (check_env("IMSM_NO_PLATFORM"))
1642 list
= find_intel_devices();
1645 for (hba
= list
; hba
; hba
= hba
->next
) {
1646 orom
= find_imsm_capability(hba
->type
);
1652 free_sys_dev(&list
);
1656 list
= find_intel_devices();
1659 fprintf(stderr
, Name
": no active Intel(R) RAID "
1660 "controller found.\n");
1661 free_sys_dev(&list
);
1664 print_found_intel_controllers(list
);
1666 for (hba
= list
; hba
; hba
= hba
->next
) {
1667 orom
= find_imsm_capability(hba
->type
);
1669 fprintf(stderr
, Name
": imsm capabilities not found for controller: %s (type %s)\n",
1670 hba
->path
, get_sys_dev_type(hba
->type
));
1672 print_imsm_capability(orom
);
1675 for (hba
= list
; hba
; hba
= hba
->next
) {
1676 printf(" I/O Controller : %s (%s)\n",
1677 hba
->path
, get_sys_dev_type(hba
->type
));
1679 if (hba
->type
== SYS_DEV_SATA
) {
1680 host_base
= ahci_get_port_count(hba
->path
, &port_count
);
1681 if (ahci_enumerate_ports(hba
->path
, port_count
, host_base
, verbose
)) {
1683 fprintf(stderr
, Name
": failed to enumerate "
1684 "ports on SATA controller at %s.", hba
->pci_id
);
1690 free_sys_dev(&list
);
1695 static int match_home_imsm(struct supertype
*st
, char *homehost
)
1697 /* the imsm metadata format does not specify any host
1698 * identification information. We return -1 since we can never
1699 * confirm nor deny whether a given array is "meant" for this
1700 * host. We rely on compare_super and the 'family_num' fields to
1701 * exclude member disks that do not belong, and we rely on
1702 * mdadm.conf to specify the arrays that should be assembled.
1703 * Auto-assembly may still pick up "foreign" arrays.
1709 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
1711 /* The uuid returned here is used for:
1712 * uuid to put into bitmap file (Create, Grow)
1713 * uuid for backup header when saving critical section (Grow)
1714 * comparing uuids when re-adding a device into an array
1715 * In these cases the uuid required is that of the data-array,
1716 * not the device-set.
1717 * uuid to recognise same set when adding a missing device back
1718 * to an array. This is a uuid for the device-set.
1720 * For each of these we can make do with a truncated
1721 * or hashed uuid rather than the original, as long as
1723 * In each case the uuid required is that of the data-array,
1724 * not the device-set.
1726 /* imsm does not track uuid's so we synthesis one using sha1 on
1727 * - The signature (Which is constant for all imsm array, but no matter)
1728 * - the orig_family_num of the container
1729 * - the index number of the volume
1730 * - the 'serial' number of the volume.
1731 * Hopefully these are all constant.
1733 struct intel_super
*super
= st
->sb
;
1736 struct sha1_ctx ctx
;
1737 struct imsm_dev
*dev
= NULL
;
1740 /* some mdadm versions failed to set ->orig_family_num, in which
1741 * case fall back to ->family_num. orig_family_num will be
1742 * fixed up with the first metadata update.
1744 family_num
= super
->anchor
->orig_family_num
;
1745 if (family_num
== 0)
1746 family_num
= super
->anchor
->family_num
;
1747 sha1_init_ctx(&ctx
);
1748 sha1_process_bytes(super
->anchor
->sig
, MPB_SIG_LEN
, &ctx
);
1749 sha1_process_bytes(&family_num
, sizeof(__u32
), &ctx
);
1750 if (super
->current_vol
>= 0)
1751 dev
= get_imsm_dev(super
, super
->current_vol
);
1753 __u32 vol
= super
->current_vol
;
1754 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
1755 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
1757 sha1_finish_ctx(&ctx
, buf
);
1758 memcpy(uuid
, buf
, 4*4);
1763 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
1765 __u8
*v
= get_imsm_version(mpb
);
1766 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
1767 char major
[] = { 0, 0, 0 };
1768 char minor
[] = { 0 ,0, 0 };
1769 char patch
[] = { 0, 0, 0 };
1770 char *ver_parse
[] = { major
, minor
, patch
};
1774 while (*v
!= '\0' && v
< end
) {
1775 if (*v
!= '.' && j
< 2)
1776 ver_parse
[i
][j
++] = *v
;
1784 *m
= strtol(minor
, NULL
, 0);
1785 *p
= strtol(patch
, NULL
, 0);
1789 static __u32
migr_strip_blocks_resync(struct imsm_dev
*dev
)
1791 /* migr_strip_size when repairing or initializing parity */
1792 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1793 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1795 switch (get_imsm_raid_level(map
)) {
1800 return 128*1024 >> 9;
1804 static __u32
migr_strip_blocks_rebuild(struct imsm_dev
*dev
)
1806 /* migr_strip_size when rebuilding a degraded disk, no idea why
1807 * this is different than migr_strip_size_resync(), but it's good
1810 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1811 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1813 switch (get_imsm_raid_level(map
)) {
1816 if (map
->num_members
% map
->num_domains
== 0)
1817 return 128*1024 >> 9;
1821 return max((__u32
) 64*1024 >> 9, chunk
);
1823 return 128*1024 >> 9;
1827 static __u32
num_stripes_per_unit_resync(struct imsm_dev
*dev
)
1829 struct imsm_map
*lo
= get_imsm_map(dev
, 0);
1830 struct imsm_map
*hi
= get_imsm_map(dev
, 1);
1831 __u32 lo_chunk
= __le32_to_cpu(lo
->blocks_per_strip
);
1832 __u32 hi_chunk
= __le32_to_cpu(hi
->blocks_per_strip
);
1834 return max((__u32
) 1, hi_chunk
/ lo_chunk
);
1837 static __u32
num_stripes_per_unit_rebuild(struct imsm_dev
*dev
)
1839 struct imsm_map
*lo
= get_imsm_map(dev
, 0);
1840 int level
= get_imsm_raid_level(lo
);
1842 if (level
== 1 || level
== 10) {
1843 struct imsm_map
*hi
= get_imsm_map(dev
, 1);
1845 return hi
->num_domains
;
1847 return num_stripes_per_unit_resync(dev
);
1850 static __u8
imsm_num_data_members(struct imsm_dev
*dev
, int second_map
)
1852 /* named 'imsm_' because raid0, raid1 and raid10
1853 * counter-intuitively have the same number of data disks
1855 struct imsm_map
*map
= get_imsm_map(dev
, second_map
);
1857 switch (get_imsm_raid_level(map
)) {
1861 return map
->num_members
;
1863 return map
->num_members
- 1;
1865 dprintf("%s: unsupported raid level\n", __func__
);
1870 static __u32
parity_segment_depth(struct imsm_dev
*dev
)
1872 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1873 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1875 switch(get_imsm_raid_level(map
)) {
1878 return chunk
* map
->num_domains
;
1880 return chunk
* map
->num_members
;
1886 static __u32
map_migr_block(struct imsm_dev
*dev
, __u32 block
)
1888 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1889 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1890 __u32 strip
= block
/ chunk
;
1892 switch (get_imsm_raid_level(map
)) {
1895 __u32 vol_strip
= (strip
* map
->num_domains
) + 1;
1896 __u32 vol_stripe
= vol_strip
/ map
->num_members
;
1898 return vol_stripe
* chunk
+ block
% chunk
;
1900 __u32 stripe
= strip
/ (map
->num_members
- 1);
1902 return stripe
* chunk
+ block
% chunk
;
1909 static __u64
blocks_per_migr_unit(struct intel_super
*super
,
1910 struct imsm_dev
*dev
)
1912 /* calculate the conversion factor between per member 'blocks'
1913 * (md/{resync,rebuild}_start) and imsm migration units, return
1914 * 0 for the 'not migrating' and 'unsupported migration' cases
1916 if (!dev
->vol
.migr_state
)
1919 switch (migr_type(dev
)) {
1920 case MIGR_GEN_MIGR
: {
1921 struct migr_record
*migr_rec
= super
->migr_rec
;
1922 return __le32_to_cpu(migr_rec
->blocks_per_unit
);
1927 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1928 __u32 stripes_per_unit
;
1929 __u32 blocks_per_unit
;
1938 /* yes, this is really the translation of migr_units to
1939 * per-member blocks in the 'resync' case
1941 stripes_per_unit
= num_stripes_per_unit_resync(dev
);
1942 migr_chunk
= migr_strip_blocks_resync(dev
);
1943 disks
= imsm_num_data_members(dev
, 0);
1944 blocks_per_unit
= stripes_per_unit
* migr_chunk
* disks
;
1945 stripe
= __le16_to_cpu(map
->blocks_per_strip
) * disks
;
1946 segment
= blocks_per_unit
/ stripe
;
1947 block_rel
= blocks_per_unit
- segment
* stripe
;
1948 parity_depth
= parity_segment_depth(dev
);
1949 block_map
= map_migr_block(dev
, block_rel
);
1950 return block_map
+ parity_depth
* segment
;
1952 case MIGR_REBUILD
: {
1953 __u32 stripes_per_unit
;
1956 stripes_per_unit
= num_stripes_per_unit_rebuild(dev
);
1957 migr_chunk
= migr_strip_blocks_rebuild(dev
);
1958 return migr_chunk
* stripes_per_unit
;
1960 case MIGR_STATE_CHANGE
:
1966 static int imsm_level_to_layout(int level
)
1974 return ALGORITHM_LEFT_ASYMMETRIC
;
1981 /*******************************************************************************
1982 * Function: read_imsm_migr_rec
1983 * Description: Function reads imsm migration record from last sector of disk
1985 * fd : disk descriptor
1986 * super : metadata info
1990 ******************************************************************************/
1991 static int read_imsm_migr_rec(int fd
, struct intel_super
*super
)
1994 unsigned long long dsize
;
1996 get_dev_size(fd
, NULL
, &dsize
);
1997 if (lseek64(fd
, dsize
- 512, SEEK_SET
) < 0) {
1999 Name
": Cannot seek to anchor block: %s\n",
2003 if (read(fd
, super
->migr_rec_buf
, 512) != 512) {
2005 Name
": Cannot read migr record block: %s\n",
2015 /*******************************************************************************
2016 * Function: load_imsm_migr_rec
2017 * Description: Function reads imsm migration record (it is stored at the last
2020 * super : imsm internal array info
2021 * info : general array info
2025 ******************************************************************************/
2026 static int load_imsm_migr_rec(struct intel_super
*super
, struct mdinfo
*info
)
2029 struct dl
*dl
= NULL
;
2035 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
2036 /* read only from one of the first two slots */
2037 if ((sd
->disk
.raid_disk
> 1) ||
2038 (sd
->disk
.raid_disk
< 0))
2040 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2041 fd
= dev_open(nm
, O_RDONLY
);
2047 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2048 /* read only from one of the first two slots */
2051 sprintf(nm
, "%d:%d", dl
->major
, dl
->minor
);
2052 fd
= dev_open(nm
, O_RDONLY
);
2059 retval
= read_imsm_migr_rec(fd
, super
);
2068 /*******************************************************************************
2069 * function: imsm_create_metadata_checkpoint_update
2070 * Description: It creates update for checkpoint change.
2072 * super : imsm internal array info
2073 * u : pointer to prepared update
2076 * If length is equal to 0, input pointer u contains no update
2077 ******************************************************************************/
2078 static int imsm_create_metadata_checkpoint_update(
2079 struct intel_super
*super
,
2080 struct imsm_update_general_migration_checkpoint
**u
)
2083 int update_memory_size
= 0;
2085 dprintf("imsm_create_metadata_checkpoint_update(enter)\n");
2091 /* size of all update data without anchor */
2092 update_memory_size
=
2093 sizeof(struct imsm_update_general_migration_checkpoint
);
2095 *u
= calloc(1, update_memory_size
);
2097 dprintf("error: cannot get memory for "
2098 "imsm_create_metadata_checkpoint_update update\n");
2101 (*u
)->type
= update_general_migration_checkpoint
;
2102 (*u
)->curr_migr_unit
= __le32_to_cpu(super
->migr_rec
->curr_migr_unit
);
2103 dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n",
2104 (*u
)->curr_migr_unit
);
2106 return update_memory_size
;
2110 static void imsm_update_metadata_locally(struct supertype
*st
,
2111 void *buf
, int len
);
2113 /*******************************************************************************
2114 * Function: write_imsm_migr_rec
2115 * Description: Function writes imsm migration record
2116 * (at the last sector of disk)
2118 * super : imsm internal array info
2122 ******************************************************************************/
2123 static int write_imsm_migr_rec(struct supertype
*st
)
2125 struct intel_super
*super
= st
->sb
;
2126 unsigned long long dsize
;
2132 struct imsm_update_general_migration_checkpoint
*u
;
2134 for (sd
= super
->disks
; sd
; sd
= sd
->next
) {
2135 /* write to 2 first slots only */
2136 if ((sd
->index
< 0) || (sd
->index
> 1))
2138 sprintf(nm
, "%d:%d", sd
->major
, sd
->minor
);
2139 fd
= dev_open(nm
, O_RDWR
);
2142 get_dev_size(fd
, NULL
, &dsize
);
2143 if (lseek64(fd
, dsize
- 512, SEEK_SET
) < 0) {
2145 Name
": Cannot seek to anchor block: %s\n",
2149 if (write(fd
, super
->migr_rec_buf
, 512) != 512) {
2151 Name
": Cannot write migr record block: %s\n",
2158 /* update checkpoint information in metadata */
2159 len
= imsm_create_metadata_checkpoint_update(super
, &u
);
2162 dprintf("imsm: Cannot prepare update\n");
2165 /* update metadata locally */
2166 imsm_update_metadata_locally(st
, u
, len
);
2167 /* and possibly remotely */
2168 if (st
->update_tail
) {
2169 append_metadata_update(st
, u
, len
);
2170 /* during reshape we do all work inside metadata handler
2171 * manage_reshape(), so metadata update has to be triggered
2174 flush_metadata_updates(st
);
2175 st
->update_tail
= &st
->updates
;
2185 #endif /* MDASSEMBLE */
2187 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
, char *dmap
)
2189 struct intel_super
*super
= st
->sb
;
2190 struct migr_record
*migr_rec
= super
->migr_rec
;
2191 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
2192 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2193 struct imsm_map
*prev_map
= get_imsm_map(dev
, 1);
2194 struct imsm_map
*map_to_analyse
= map
;
2197 unsigned int component_size_alligment
;
2198 int map_disks
= info
->array
.raid_disks
;
2200 memset(info
, 0, sizeof(*info
));
2202 map_to_analyse
= prev_map
;
2206 info
->container_member
= super
->current_vol
;
2207 info
->array
.raid_disks
= map
->num_members
;
2208 info
->array
.level
= get_imsm_raid_level(map_to_analyse
);
2209 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
2210 info
->array
.md_minor
= -1;
2211 info
->array
.ctime
= 0;
2212 info
->array
.utime
= 0;
2213 info
->array
.chunk_size
=
2214 __le16_to_cpu(map_to_analyse
->blocks_per_strip
) << 9;
2215 info
->array
.state
= !dev
->vol
.dirty
;
2216 info
->custom_array_size
= __le32_to_cpu(dev
->size_high
);
2217 info
->custom_array_size
<<= 32;
2218 info
->custom_array_size
|= __le32_to_cpu(dev
->size_low
);
2219 if (prev_map
&& map
->map_state
== prev_map
->map_state
) {
2220 info
->reshape_active
= 1;
2221 info
->new_level
= get_imsm_raid_level(map
);
2222 info
->new_layout
= imsm_level_to_layout(info
->new_level
);
2223 info
->new_chunk
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
2224 info
->delta_disks
= map
->num_members
- prev_map
->num_members
;
2225 if (info
->delta_disks
) {
2226 /* this needs to be applied to every array
2229 info
->reshape_active
= 2;
2231 /* We shape information that we give to md might have to be
2232 * modify to cope with md's requirement for reshaping arrays.
2233 * For example, when reshaping a RAID0, md requires it to be
2234 * presented as a degraded RAID4.
2235 * Also if a RAID0 is migrating to a RAID5 we need to specify
2236 * the array as already being RAID5, but the 'before' layout
2237 * is a RAID4-like layout.
2239 switch (info
->array
.level
) {
2241 switch(info
->new_level
) {
2243 /* conversion is happening as RAID4 */
2244 info
->array
.level
= 4;
2245 info
->array
.raid_disks
+= 1;
2248 /* conversion is happening as RAID5 */
2249 info
->array
.level
= 5;
2250 info
->array
.layout
= ALGORITHM_PARITY_N
;
2251 info
->delta_disks
-= 1;
2254 /* FIXME error message */
2255 info
->array
.level
= UnSet
;
2261 info
->new_level
= UnSet
;
2262 info
->new_layout
= UnSet
;
2263 info
->new_chunk
= info
->array
.chunk_size
;
2264 info
->delta_disks
= 0;
2266 info
->disk
.major
= 0;
2267 info
->disk
.minor
= 0;
2269 info
->disk
.major
= dl
->major
;
2270 info
->disk
.minor
= dl
->minor
;
2273 info
->data_offset
= __le32_to_cpu(map_to_analyse
->pba_of_lba0
);
2274 info
->component_size
=
2275 __le32_to_cpu(map_to_analyse
->blocks_per_member
);
2277 /* check component size aligment
2279 component_size_alligment
=
2280 info
->component_size
% (info
->array
.chunk_size
/512);
2282 if (component_size_alligment
&&
2283 (info
->array
.level
!= 1) && (info
->array
.level
!= UnSet
)) {
2284 dprintf("imsm: reported component size alligned from %llu ",
2285 info
->component_size
);
2286 info
->component_size
-= component_size_alligment
;
2287 dprintf("to %llu (%i).\n",
2288 info
->component_size
, component_size_alligment
);
2291 memset(info
->uuid
, 0, sizeof(info
->uuid
));
2292 info
->recovery_start
= MaxSector
;
2294 info
->reshape_progress
= 0;
2295 info
->resync_start
= MaxSector
;
2296 if (map_to_analyse
->map_state
== IMSM_T_STATE_UNINITIALIZED
||
2298 info
->resync_start
= 0;
2300 if (dev
->vol
.migr_state
) {
2301 switch (migr_type(dev
)) {
2304 __u64 blocks_per_unit
= blocks_per_migr_unit(super
,
2306 __u64 units
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
2308 info
->resync_start
= blocks_per_unit
* units
;
2311 case MIGR_GEN_MIGR
: {
2312 __u64 blocks_per_unit
= blocks_per_migr_unit(super
,
2314 __u64 units
= __le32_to_cpu(migr_rec
->curr_migr_unit
);
2315 unsigned long long array_blocks
;
2318 if (__le32_to_cpu(migr_rec
->ascending_migr
) &&
2320 (__le32_to_cpu(migr_rec
->num_migr_units
)-1)) &&
2321 (super
->migr_rec
->rec_status
==
2322 __cpu_to_le32(UNIT_SRC_IN_CP_AREA
)))
2325 info
->reshape_progress
= blocks_per_unit
* units
;
2327 dprintf("IMSM: General Migration checkpoint : %llu "
2328 "(%llu) -> read reshape progress : %llu\n",
2329 units
, blocks_per_unit
, info
->reshape_progress
);
2331 used_disks
= imsm_num_data_members(dev
, 1);
2332 if (used_disks
> 0) {
2333 array_blocks
= map
->blocks_per_member
*
2335 /* round array size down to closest MB
2337 info
->custom_array_size
= (array_blocks
2338 >> SECT_PER_MB_SHIFT
)
2339 << SECT_PER_MB_SHIFT
;
2343 /* we could emulate the checkpointing of
2344 * 'sync_action=check' migrations, but for now
2345 * we just immediately complete them
2348 /* this is handled by container_content_imsm() */
2349 case MIGR_STATE_CHANGE
:
2350 /* FIXME handle other migrations */
2352 /* we are not dirty, so... */
2353 info
->resync_start
= MaxSector
;
2357 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
2358 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
2360 info
->array
.major_version
= -1;
2361 info
->array
.minor_version
= -2;
2362 devname
= devnum2devname(st
->container_dev
);
2363 *info
->text_version
= '\0';
2365 sprintf(info
->text_version
, "/%s/%d", devname
, info
->container_member
);
2367 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
2368 uuid_from_super_imsm(st
, info
->uuid
);
2372 for (i
=0; i
<map_disks
; i
++) {
2374 if (i
< info
->array
.raid_disks
) {
2375 struct imsm_disk
*dsk
;
2376 j
= get_imsm_disk_idx(dev
, i
, -1);
2377 dsk
= get_imsm_disk(super
, j
);
2378 if (dsk
&& (dsk
->status
& CONFIGURED_DISK
))
2385 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
);
2386 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
);
2388 static struct imsm_disk
*get_imsm_missing(struct intel_super
*super
, __u8 index
)
2392 for (d
= super
->missing
; d
; d
= d
->next
)
2393 if (d
->index
== index
)
2398 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
, char *map
)
2400 struct intel_super
*super
= st
->sb
;
2401 struct imsm_disk
*disk
;
2402 int map_disks
= info
->array
.raid_disks
;
2403 int max_enough
= -1;
2405 struct imsm_super
*mpb
;
2407 if (super
->current_vol
>= 0) {
2408 getinfo_super_imsm_volume(st
, info
, map
);
2411 memset(info
, 0, sizeof(*info
));
2413 /* Set raid_disks to zero so that Assemble will always pull in valid
2416 info
->array
.raid_disks
= 0;
2417 info
->array
.level
= LEVEL_CONTAINER
;
2418 info
->array
.layout
= 0;
2419 info
->array
.md_minor
= -1;
2420 info
->array
.ctime
= 0; /* N/A for imsm */
2421 info
->array
.utime
= 0;
2422 info
->array
.chunk_size
= 0;
2424 info
->disk
.major
= 0;
2425 info
->disk
.minor
= 0;
2426 info
->disk
.raid_disk
= -1;
2427 info
->reshape_active
= 0;
2428 info
->array
.major_version
= -1;
2429 info
->array
.minor_version
= -2;
2430 strcpy(info
->text_version
, "imsm");
2431 info
->safe_mode_delay
= 0;
2432 info
->disk
.number
= -1;
2433 info
->disk
.state
= 0;
2435 info
->recovery_start
= MaxSector
;
2437 /* do we have the all the insync disks that we expect? */
2438 mpb
= super
->anchor
;
2440 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2441 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2442 int failed
, enough
, j
, missing
= 0;
2443 struct imsm_map
*map
;
2446 failed
= imsm_count_failed(super
, dev
);
2447 state
= imsm_check_degraded(super
, dev
, failed
);
2448 map
= get_imsm_map(dev
, dev
->vol
.migr_state
);
2450 /* any newly missing disks?
2451 * (catches single-degraded vs double-degraded)
2453 for (j
= 0; j
< map
->num_members
; j
++) {
2454 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
2455 __u32 idx
= ord_to_idx(ord
);
2457 if (!(ord
& IMSM_ORD_REBUILD
) &&
2458 get_imsm_missing(super
, idx
)) {
2464 if (state
== IMSM_T_STATE_FAILED
)
2466 else if (state
== IMSM_T_STATE_DEGRADED
&&
2467 (state
!= map
->map_state
|| missing
))
2469 else /* we're normal, or already degraded */
2472 /* in the missing/failed disk case check to see
2473 * if at least one array is runnable
2475 max_enough
= max(max_enough
, enough
);
2477 dprintf("%s: enough: %d\n", __func__
, max_enough
);
2478 info
->container_enough
= max_enough
;
2481 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
2483 disk
= &super
->disks
->disk
;
2484 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
2485 info
->component_size
= reserved
;
2486 info
->disk
.state
= is_configured(disk
) ? (1 << MD_DISK_ACTIVE
) : 0;
2487 /* we don't change info->disk.raid_disk here because
2488 * this state will be finalized in mdmon after we have
2489 * found the 'most fresh' version of the metadata
2491 info
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2492 info
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2495 /* only call uuid_from_super_imsm when this disk is part of a populated container,
2496 * ->compare_super may have updated the 'num_raid_devs' field for spares
2498 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
2499 uuid_from_super_imsm(st
, info
->uuid
);
2501 memcpy(info
->uuid
, uuid_zero
, sizeof(uuid_zero
));
2503 /* I don't know how to compute 'map' on imsm, so use safe default */
2506 for (i
= 0; i
< map_disks
; i
++)
2512 /* allocates memory and fills disk in mdinfo structure
2513 * for each disk in array */
2514 struct mdinfo
*getinfo_super_disks_imsm(struct supertype
*st
)
2516 struct mdinfo
*mddev
= NULL
;
2517 struct intel_super
*super
= st
->sb
;
2518 struct imsm_disk
*disk
;
2521 if (!super
|| !super
->disks
)
2524 mddev
= malloc(sizeof(*mddev
));
2526 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2529 memset(mddev
, 0, sizeof(*mddev
));
2533 tmp
= malloc(sizeof(*tmp
));
2535 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2540 memset(tmp
, 0, sizeof(*tmp
));
2542 tmp
->next
= mddev
->devs
;
2544 tmp
->disk
.number
= count
++;
2545 tmp
->disk
.major
= dl
->major
;
2546 tmp
->disk
.minor
= dl
->minor
;
2547 tmp
->disk
.state
= is_configured(disk
) ?
2548 (1 << MD_DISK_ACTIVE
) : 0;
2549 tmp
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2550 tmp
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2551 tmp
->disk
.raid_disk
= -1;
2557 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
2558 char *update
, char *devname
, int verbose
,
2559 int uuid_set
, char *homehost
)
2561 /* For 'assemble' and 'force' we need to return non-zero if any
2562 * change was made. For others, the return value is ignored.
2563 * Update options are:
2564 * force-one : This device looks a bit old but needs to be included,
2565 * update age info appropriately.
2566 * assemble: clear any 'faulty' flag to allow this device to
2568 * force-array: Array is degraded but being forced, mark it clean
2569 * if that will be needed to assemble it.
2571 * newdev: not used ????
2572 * grow: Array has gained a new device - this is currently for
2574 * resync: mark as dirty so a resync will happen.
2575 * name: update the name - preserving the homehost
2576 * uuid: Change the uuid of the array to match watch is given
2578 * Following are not relevant for this imsm:
2579 * sparc2.2 : update from old dodgey metadata
2580 * super-minor: change the preferred_minor number
2581 * summaries: update redundant counters.
2582 * homehost: update the recorded homehost
2583 * _reshape_progress: record new reshape_progress position.
2586 struct intel_super
*super
= st
->sb
;
2587 struct imsm_super
*mpb
;
2589 /* we can only update container info */
2590 if (!super
|| super
->current_vol
>= 0 || !super
->anchor
)
2593 mpb
= super
->anchor
;
2595 if (strcmp(update
, "uuid") == 0 && uuid_set
&& !info
->update_private
)
2597 else if (strcmp(update
, "uuid") == 0 && uuid_set
&& info
->update_private
) {
2598 mpb
->orig_family_num
= *((__u32
*) info
->update_private
);
2600 } else if (strcmp(update
, "uuid") == 0) {
2601 __u32
*new_family
= malloc(sizeof(*new_family
));
2603 /* update orig_family_number with the incoming random
2604 * data, report the new effective uuid, and store the
2605 * new orig_family_num for future updates.
2608 memcpy(&mpb
->orig_family_num
, info
->uuid
, sizeof(__u32
));
2609 uuid_from_super_imsm(st
, info
->uuid
);
2610 *new_family
= mpb
->orig_family_num
;
2611 info
->update_private
= new_family
;
2614 } else if (strcmp(update
, "assemble") == 0)
2619 /* successful update? recompute checksum */
2621 mpb
->check_sum
= __le32_to_cpu(__gen_imsm_checksum(mpb
));
2626 static size_t disks_to_mpb_size(int disks
)
2630 size
= sizeof(struct imsm_super
);
2631 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
2632 size
+= 2 * sizeof(struct imsm_dev
);
2633 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
2634 size
+= (4 - 2) * sizeof(struct imsm_map
);
2635 /* 4 possible disk_ord_tbl's */
2636 size
+= 4 * (disks
- 1) * sizeof(__u32
);
2641 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
2643 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
2646 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
2649 static void free_devlist(struct intel_super
*super
)
2651 struct intel_dev
*dv
;
2653 while (super
->devlist
) {
2654 dv
= super
->devlist
->next
;
2655 free(super
->devlist
->dev
);
2656 free(super
->devlist
);
2657 super
->devlist
= dv
;
2661 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
2663 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
2666 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
2670 * 0 same, or first was empty, and second was copied
2671 * 1 second had wrong number
2673 * 3 wrong other info
2675 struct intel_super
*first
= st
->sb
;
2676 struct intel_super
*sec
= tst
->sb
;
2683 /* in platform dependent environment test if the disks
2684 * use the same Intel hba
2686 if (!check_env("IMSM_NO_PLATFORM")) {
2687 if (!first
->hba
|| !sec
->hba
||
2688 (first
->hba
->type
!= sec
->hba
->type
)) {
2690 "HBAs of devices does not match %s != %s\n",
2691 first
->hba
? get_sys_dev_type(first
->hba
->type
) : NULL
,
2692 sec
->hba
? get_sys_dev_type(sec
->hba
->type
) : NULL
);
2697 /* if an anchor does not have num_raid_devs set then it is a free
2700 if (first
->anchor
->num_raid_devs
> 0 &&
2701 sec
->anchor
->num_raid_devs
> 0) {
2702 /* Determine if these disks might ever have been
2703 * related. Further disambiguation can only take place
2704 * in load_super_imsm_all
2706 __u32 first_family
= first
->anchor
->orig_family_num
;
2707 __u32 sec_family
= sec
->anchor
->orig_family_num
;
2709 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
,
2710 MAX_SIGNATURE_LENGTH
) != 0)
2713 if (first_family
== 0)
2714 first_family
= first
->anchor
->family_num
;
2715 if (sec_family
== 0)
2716 sec_family
= sec
->anchor
->family_num
;
2718 if (first_family
!= sec_family
)
2724 /* if 'first' is a spare promote it to a populated mpb with sec's
2727 if (first
->anchor
->num_raid_devs
== 0 &&
2728 sec
->anchor
->num_raid_devs
> 0) {
2730 struct intel_dev
*dv
;
2731 struct imsm_dev
*dev
;
2733 /* we need to copy raid device info from sec if an allocation
2734 * fails here we don't associate the spare
2736 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
2737 dv
= malloc(sizeof(*dv
));
2740 dev
= malloc(sizeof_imsm_dev(get_imsm_dev(sec
, i
), 1));
2747 dv
->next
= first
->devlist
;
2748 first
->devlist
= dv
;
2750 if (i
< sec
->anchor
->num_raid_devs
) {
2751 /* allocation failure */
2752 free_devlist(first
);
2753 fprintf(stderr
, "imsm: failed to associate spare\n");
2756 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
2757 first
->anchor
->orig_family_num
= sec
->anchor
->orig_family_num
;
2758 first
->anchor
->family_num
= sec
->anchor
->family_num
;
2759 memcpy(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
);
2760 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++)
2761 imsm_copy_dev(get_imsm_dev(first
, i
), get_imsm_dev(sec
, i
));
2767 static void fd2devname(int fd
, char *name
)
2771 char dname
[PATH_MAX
];
2776 if (fstat(fd
, &st
) != 0)
2778 sprintf(path
, "/sys/dev/block/%d:%d",
2779 major(st
.st_rdev
), minor(st
.st_rdev
));
2781 rv
= readlink(path
, dname
, sizeof(dname
));
2786 nm
= strrchr(dname
, '/');
2788 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
2791 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
2793 static int imsm_read_serial(int fd
, char *devname
,
2794 __u8 serial
[MAX_RAID_SERIAL_LEN
])
2796 unsigned char scsi_serial
[255];
2805 memset(scsi_serial
, 0, sizeof(scsi_serial
));
2807 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
2809 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
2810 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2811 fd2devname(fd
, (char *) serial
);
2818 Name
": Failed to retrieve serial for %s\n",
2823 rsp_len
= scsi_serial
[3];
2827 Name
": Failed to retrieve serial for %s\n",
2831 rsp_buf
= (char *) &scsi_serial
[4];
2833 /* trim all whitespace and non-printable characters and convert
2836 for (i
= 0, dest
= rsp_buf
; i
< rsp_len
; i
++) {
2839 /* ':' is reserved for use in placeholder serial
2840 * numbers for missing disks
2848 len
= dest
- rsp_buf
;
2851 /* truncate leading characters */
2852 if (len
> MAX_RAID_SERIAL_LEN
) {
2853 dest
+= len
- MAX_RAID_SERIAL_LEN
;
2854 len
= MAX_RAID_SERIAL_LEN
;
2857 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2858 memcpy(serial
, dest
, len
);
2863 static int serialcmp(__u8
*s1
, __u8
*s2
)
2865 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
2868 static void serialcpy(__u8
*dest
, __u8
*src
)
2870 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
2874 static struct dl
*serial_to_dl(__u8
*serial
, struct intel_super
*super
)
2878 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2879 if (serialcmp(dl
->serial
, serial
) == 0)
2886 static struct imsm_disk
*
2887 __serial_to_disk(__u8
*serial
, struct imsm_super
*mpb
, int *idx
)
2891 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2892 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
2894 if (serialcmp(disk
->serial
, serial
) == 0) {
2905 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
2907 struct imsm_disk
*disk
;
2912 __u8 serial
[MAX_RAID_SERIAL_LEN
];
2914 rv
= imsm_read_serial(fd
, devname
, serial
);
2919 dl
= calloc(1, sizeof(*dl
));
2923 Name
": failed to allocate disk buffer for %s\n",
2929 dl
->major
= major(stb
.st_rdev
);
2930 dl
->minor
= minor(stb
.st_rdev
);
2931 dl
->next
= super
->disks
;
2932 dl
->fd
= keep_fd
? fd
: -1;
2933 assert(super
->disks
== NULL
);
2935 serialcpy(dl
->serial
, serial
);
2938 fd2devname(fd
, name
);
2940 dl
->devname
= strdup(devname
);
2942 dl
->devname
= strdup(name
);
2944 /* look up this disk's index in the current anchor */
2945 disk
= __serial_to_disk(dl
->serial
, super
->anchor
, &dl
->index
);
2948 /* only set index on disks that are a member of a
2949 * populated contianer, i.e. one with raid_devs
2951 if (is_failed(&dl
->disk
))
2953 else if (is_spare(&dl
->disk
))
2961 /* When migrating map0 contains the 'destination' state while map1
2962 * contains the current state. When not migrating map0 contains the
2963 * current state. This routine assumes that map[0].map_state is set to
2964 * the current array state before being called.
2966 * Migration is indicated by one of the following states
2967 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
2968 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
2969 * map1state=unitialized)
2970 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
2972 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
2973 * map1state=degraded)
2974 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
2977 static void migrate(struct imsm_dev
*dev
, struct intel_super
*super
,
2978 __u8 to_state
, int migr_type
)
2980 struct imsm_map
*dest
;
2981 struct imsm_map
*src
= get_imsm_map(dev
, 0);
2983 dev
->vol
.migr_state
= 1;
2984 set_migr_type(dev
, migr_type
);
2985 dev
->vol
.curr_migr_unit
= 0;
2986 dest
= get_imsm_map(dev
, 1);
2988 /* duplicate and then set the target end state in map[0] */
2989 memcpy(dest
, src
, sizeof_imsm_map(src
));
2990 if ((migr_type
== MIGR_REBUILD
) ||
2991 (migr_type
== MIGR_GEN_MIGR
)) {
2995 for (i
= 0; i
< src
->num_members
; i
++) {
2996 ord
= __le32_to_cpu(src
->disk_ord_tbl
[i
]);
2997 set_imsm_ord_tbl_ent(src
, i
, ord_to_idx(ord
));
3001 if (migr_type
== MIGR_GEN_MIGR
)
3002 /* Clear migration record */
3003 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
3005 src
->map_state
= to_state
;
3008 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
3010 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3011 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
3014 /* merge any IMSM_ORD_REBUILD bits that were not successfully
3015 * completed in the last migration.
3017 * FIXME add support for raid-level-migration
3019 for (i
= 0; i
< prev
->num_members
; i
++)
3020 for (j
= 0; j
< map
->num_members
; j
++)
3021 /* during online capacity expansion
3022 * disks position can be changed if takeover is used
3024 if (ord_to_idx(map
->disk_ord_tbl
[j
]) ==
3025 ord_to_idx(prev
->disk_ord_tbl
[i
])) {
3026 map
->disk_ord_tbl
[j
] |= prev
->disk_ord_tbl
[i
];
3030 dev
->vol
.migr_state
= 0;
3031 dev
->vol
.migr_type
= 0;
3032 dev
->vol
.curr_migr_unit
= 0;
3033 map
->map_state
= map_state
;
3037 static int parse_raid_devices(struct intel_super
*super
)
3040 struct imsm_dev
*dev_new
;
3041 size_t len
, len_migr
;
3043 size_t space_needed
= 0;
3044 struct imsm_super
*mpb
= super
->anchor
;
3046 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3047 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3048 struct intel_dev
*dv
;
3050 len
= sizeof_imsm_dev(dev_iter
, 0);
3051 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
3053 space_needed
+= len_migr
- len
;
3055 dv
= malloc(sizeof(*dv
));
3058 if (max_len
< len_migr
)
3060 if (max_len
> len_migr
)
3061 space_needed
+= max_len
- len_migr
;
3062 dev_new
= malloc(max_len
);
3067 imsm_copy_dev(dev_new
, dev_iter
);
3070 dv
->next
= super
->devlist
;
3071 super
->devlist
= dv
;
3074 /* ensure that super->buf is large enough when all raid devices
3077 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
3080 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
3081 if (posix_memalign(&buf
, 512, len
) != 0)
3084 memcpy(buf
, super
->buf
, super
->len
);
3085 memset(buf
+ super
->len
, 0, len
- super
->len
);
3094 /* retrieve a pointer to the bbm log which starts after all raid devices */
3095 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
3099 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
3101 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
3107 /*******************************************************************************
3108 * Function: check_mpb_migr_compatibility
3109 * Description: Function checks for unsupported migration features:
3110 * - migration optimization area (pba_of_lba0)
3111 * - descending reshape (ascending_migr)
3113 * super : imsm metadata information
3115 * 0 : migration is compatible
3116 * -1 : migration is not compatible
3117 ******************************************************************************/
3118 int check_mpb_migr_compatibility(struct intel_super
*super
)
3120 struct imsm_map
*map0
, *map1
;
3121 struct migr_record
*migr_rec
= super
->migr_rec
;
3124 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3125 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3128 dev_iter
->vol
.migr_state
== 1 &&
3129 dev_iter
->vol
.migr_type
== MIGR_GEN_MIGR
) {
3130 /* This device is migrating */
3131 map0
= get_imsm_map(dev_iter
, 0);
3132 map1
= get_imsm_map(dev_iter
, 1);
3133 if (map0
->pba_of_lba0
!= map1
->pba_of_lba0
)
3134 /* migration optimization area was used */
3136 if (migr_rec
->ascending_migr
== 0
3137 && migr_rec
->dest_depth_per_unit
> 0)
3138 /* descending reshape not supported yet */
3145 static void __free_imsm(struct intel_super
*super
, int free_disks
);
3147 /* load_imsm_mpb - read matrix metadata
3148 * allocates super->mpb to be freed by free_imsm
3150 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
3152 unsigned long long dsize
;
3153 unsigned long long sectors
;
3155 struct imsm_super
*anchor
;
3158 get_dev_size(fd
, NULL
, &dsize
);
3162 Name
": %s: device to small for imsm\n",
3167 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
3169 fprintf(stderr
, Name
3170 ": Cannot seek to anchor block on %s: %s\n",
3171 devname
, strerror(errno
));
3175 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
3178 Name
": Failed to allocate imsm anchor buffer"
3179 " on %s\n", devname
);
3182 if (read(fd
, anchor
, 512) != 512) {
3185 Name
": Cannot read anchor block on %s: %s\n",
3186 devname
, strerror(errno
));
3191 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
3194 Name
": no IMSM anchor on %s\n", devname
);
3199 __free_imsm(super
, 0);
3200 /* reload capability and hba */
3202 /* capability and hba must be updated with new super allocation */
3203 find_intel_hba_capability(fd
, super
, devname
);
3204 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
3205 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
3208 Name
": unable to allocate %zu byte mpb buffer\n",
3213 memcpy(super
->buf
, anchor
, 512);
3215 sectors
= mpb_sectors(anchor
) - 1;
3218 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
3219 fprintf(stderr
, Name
3220 ": %s could not allocate migr_rec buffer\n", __func__
);
3226 check_sum
= __gen_imsm_checksum(super
->anchor
);
3227 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3230 Name
": IMSM checksum %x != %x on %s\n",
3232 __le32_to_cpu(super
->anchor
->check_sum
),
3240 /* read the extended mpb */
3241 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
3244 Name
": Cannot seek to extended mpb on %s: %s\n",
3245 devname
, strerror(errno
));
3249 if ((unsigned)read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
3252 Name
": Cannot read extended mpb on %s: %s\n",
3253 devname
, strerror(errno
));
3257 check_sum
= __gen_imsm_checksum(super
->anchor
);
3258 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3261 Name
": IMSM checksum %x != %x on %s\n",
3262 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
3267 /* FIXME the BBM log is disk specific so we cannot use this global
3268 * buffer for all disks. Ok for now since we only look at the global
3269 * bbm_log_size parameter to gate assembly
3271 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
3276 static int read_imsm_migr_rec(int fd
, struct intel_super
*super
);
3279 load_and_parse_mpb(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
3283 err
= load_imsm_mpb(fd
, super
, devname
);
3286 err
= load_imsm_disk(fd
, super
, devname
, keep_fd
);
3289 err
= parse_raid_devices(super
);
3294 static void __free_imsm_disk(struct dl
*d
)
3306 static void free_imsm_disks(struct intel_super
*super
)
3310 while (super
->disks
) {
3312 super
->disks
= d
->next
;
3313 __free_imsm_disk(d
);
3315 while (super
->disk_mgmt_list
) {
3316 d
= super
->disk_mgmt_list
;
3317 super
->disk_mgmt_list
= d
->next
;
3318 __free_imsm_disk(d
);
3320 while (super
->missing
) {
3322 super
->missing
= d
->next
;
3323 __free_imsm_disk(d
);
3328 /* free all the pieces hanging off of a super pointer */
3329 static void __free_imsm(struct intel_super
*super
, int free_disks
)
3331 struct intel_hba
*elem
, *next
;
3337 /* unlink capability description */
3339 if (super
->migr_rec_buf
) {
3340 free(super
->migr_rec_buf
);
3341 super
->migr_rec_buf
= NULL
;
3344 free_imsm_disks(super
);
3345 free_devlist(super
);
3349 free((void *)elem
->path
);
3357 static void free_imsm(struct intel_super
*super
)
3359 __free_imsm(super
, 1);
3363 static void free_super_imsm(struct supertype
*st
)
3365 struct intel_super
*super
= st
->sb
;
3374 static struct intel_super
*alloc_super(void)
3376 struct intel_super
*super
= malloc(sizeof(*super
));
3379 memset(super
, 0, sizeof(*super
));
3380 super
->current_vol
= -1;
3381 super
->create_offset
= ~((__u32
) 0);
3387 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
3389 static int find_intel_hba_capability(int fd
, struct intel_super
*super
, char *devname
)
3391 struct sys_dev
*hba_name
;
3394 if ((fd
< 0) || check_env("IMSM_NO_PLATFORM")) {
3399 hba_name
= find_disk_attached_hba(fd
, NULL
);
3403 Name
": %s is not attached to Intel(R) RAID controller.\n",
3407 rv
= attach_hba_to_super(super
, hba_name
);
3410 struct intel_hba
*hba
= super
->hba
;
3412 fprintf(stderr
, Name
": %s is attached to Intel(R) %s RAID "
3413 "controller (%s),\n"
3414 " but the container is assigned to Intel(R) "
3415 "%s RAID controller (",
3418 hba_name
->pci_id
? : "Err!",
3419 get_sys_dev_type(hba_name
->type
));
3422 fprintf(stderr
, "%s", hba
->pci_id
? : "Err!");
3424 fprintf(stderr
, ", ");
3428 fprintf(stderr
, ").\n"
3429 " Mixing devices attached to different controllers "
3430 "is not allowed.\n");
3432 free_sys_dev(&hba_name
);
3435 super
->orom
= find_imsm_capability(hba_name
->type
);
3436 free_sys_dev(&hba_name
);
3443 /* find_missing - helper routine for load_super_imsm_all that identifies
3444 * disks that have disappeared from the system. This routine relies on
3445 * the mpb being uptodate, which it is at load time.
3447 static int find_missing(struct intel_super
*super
)
3450 struct imsm_super
*mpb
= super
->anchor
;
3452 struct imsm_disk
*disk
;
3454 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3455 disk
= __get_imsm_disk(mpb
, i
);
3456 dl
= serial_to_dl(disk
->serial
, super
);
3460 dl
= malloc(sizeof(*dl
));
3466 dl
->devname
= strdup("missing");
3468 serialcpy(dl
->serial
, disk
->serial
);
3471 dl
->next
= super
->missing
;
3472 super
->missing
= dl
;
3478 static struct intel_disk
*disk_list_get(__u8
*serial
, struct intel_disk
*disk_list
)
3480 struct intel_disk
*idisk
= disk_list
;
3483 if (serialcmp(idisk
->disk
.serial
, serial
) == 0)
3485 idisk
= idisk
->next
;
3491 static int __prep_thunderdome(struct intel_super
**table
, int tbl_size
,
3492 struct intel_super
*super
,
3493 struct intel_disk
**disk_list
)
3495 struct imsm_disk
*d
= &super
->disks
->disk
;
3496 struct imsm_super
*mpb
= super
->anchor
;
3499 for (i
= 0; i
< tbl_size
; i
++) {
3500 struct imsm_super
*tbl_mpb
= table
[i
]->anchor
;
3501 struct imsm_disk
*tbl_d
= &table
[i
]->disks
->disk
;
3503 if (tbl_mpb
->family_num
== mpb
->family_num
) {
3504 if (tbl_mpb
->check_sum
== mpb
->check_sum
) {
3505 dprintf("%s: mpb from %d:%d matches %d:%d\n",
3506 __func__
, super
->disks
->major
,
3507 super
->disks
->minor
,
3508 table
[i
]->disks
->major
,
3509 table
[i
]->disks
->minor
);
3513 if (((is_configured(d
) && !is_configured(tbl_d
)) ||
3514 is_configured(d
) == is_configured(tbl_d
)) &&
3515 tbl_mpb
->generation_num
< mpb
->generation_num
) {
3516 /* current version of the mpb is a
3517 * better candidate than the one in
3518 * super_table, but copy over "cross
3519 * generational" status
3521 struct intel_disk
*idisk
;
3523 dprintf("%s: mpb from %d:%d replaces %d:%d\n",
3524 __func__
, super
->disks
->major
,
3525 super
->disks
->minor
,
3526 table
[i
]->disks
->major
,
3527 table
[i
]->disks
->minor
);
3529 idisk
= disk_list_get(tbl_d
->serial
, *disk_list
);
3530 if (idisk
&& is_failed(&idisk
->disk
))
3531 tbl_d
->status
|= FAILED_DISK
;
3534 struct intel_disk
*idisk
;
3535 struct imsm_disk
*disk
;
3537 /* tbl_mpb is more up to date, but copy
3538 * over cross generational status before
3541 disk
= __serial_to_disk(d
->serial
, mpb
, NULL
);
3542 if (disk
&& is_failed(disk
))
3543 d
->status
|= FAILED_DISK
;
3545 idisk
= disk_list_get(d
->serial
, *disk_list
);
3548 if (disk
&& is_configured(disk
))
3549 idisk
->disk
.status
|= CONFIGURED_DISK
;
3552 dprintf("%s: mpb from %d:%d prefer %d:%d\n",
3553 __func__
, super
->disks
->major
,
3554 super
->disks
->minor
,
3555 table
[i
]->disks
->major
,
3556 table
[i
]->disks
->minor
);
3564 table
[tbl_size
++] = super
;
3568 /* update/extend the merged list of imsm_disk records */
3569 for (j
= 0; j
< mpb
->num_disks
; j
++) {
3570 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, j
);
3571 struct intel_disk
*idisk
;
3573 idisk
= disk_list_get(disk
->serial
, *disk_list
);
3575 idisk
->disk
.status
|= disk
->status
;
3576 if (is_configured(&idisk
->disk
) ||
3577 is_failed(&idisk
->disk
))
3578 idisk
->disk
.status
&= ~(SPARE_DISK
);
3580 idisk
= calloc(1, sizeof(*idisk
));
3583 idisk
->owner
= IMSM_UNKNOWN_OWNER
;
3584 idisk
->disk
= *disk
;
3585 idisk
->next
= *disk_list
;
3589 if (serialcmp(idisk
->disk
.serial
, d
->serial
) == 0)
3596 static struct intel_super
*
3597 validate_members(struct intel_super
*super
, struct intel_disk
*disk_list
,
3600 struct imsm_super
*mpb
= super
->anchor
;
3604 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3605 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
3606 struct intel_disk
*idisk
;
3608 idisk
= disk_list_get(disk
->serial
, disk_list
);
3610 if (idisk
->owner
== owner
||
3611 idisk
->owner
== IMSM_UNKNOWN_OWNER
)
3614 dprintf("%s: '%.16s' owner %d != %d\n",
3615 __func__
, disk
->serial
, idisk
->owner
,
3618 dprintf("%s: unknown disk %x [%d]: %.16s\n",
3619 __func__
, __le32_to_cpu(mpb
->family_num
), i
,
3625 if (ok_count
== mpb
->num_disks
)
3630 static void show_conflicts(__u32 family_num
, struct intel_super
*super_list
)
3632 struct intel_super
*s
;
3634 for (s
= super_list
; s
; s
= s
->next
) {
3635 if (family_num
!= s
->anchor
->family_num
)
3637 fprintf(stderr
, "Conflict, offlining family %#x on '%s'\n",
3638 __le32_to_cpu(family_num
), s
->disks
->devname
);
3642 static struct intel_super
*
3643 imsm_thunderdome(struct intel_super
**super_list
, int len
)
3645 struct intel_super
*super_table
[len
];
3646 struct intel_disk
*disk_list
= NULL
;
3647 struct intel_super
*champion
, *spare
;
3648 struct intel_super
*s
, **del
;
3653 memset(super_table
, 0, sizeof(super_table
));
3654 for (s
= *super_list
; s
; s
= s
->next
)
3655 tbl_size
= __prep_thunderdome(super_table
, tbl_size
, s
, &disk_list
);
3657 for (i
= 0; i
< tbl_size
; i
++) {
3658 struct imsm_disk
*d
;
3659 struct intel_disk
*idisk
;
3660 struct imsm_super
*mpb
= super_table
[i
]->anchor
;
3663 d
= &s
->disks
->disk
;
3665 /* 'd' must appear in merged disk list for its
3666 * configuration to be valid
3668 idisk
= disk_list_get(d
->serial
, disk_list
);
3669 if (idisk
&& idisk
->owner
== i
)
3670 s
= validate_members(s
, disk_list
, i
);
3675 dprintf("%s: marking family: %#x from %d:%d offline\n",
3676 __func__
, mpb
->family_num
,
3677 super_table
[i
]->disks
->major
,
3678 super_table
[i
]->disks
->minor
);
3682 /* This is where the mdadm implementation differs from the Windows
3683 * driver which has no strict concept of a container. We can only
3684 * assemble one family from a container, so when returning a prodigal
3685 * array member to this system the code will not be able to disambiguate
3686 * the container contents that should be assembled ("foreign" versus
3687 * "local"). It requires user intervention to set the orig_family_num
3688 * to a new value to establish a new container. The Windows driver in
3689 * this situation fixes up the volume name in place and manages the
3690 * foreign array as an independent entity.
3695 for (i
= 0; i
< tbl_size
; i
++) {
3696 struct intel_super
*tbl_ent
= super_table
[i
];
3702 if (tbl_ent
->anchor
->num_raid_devs
== 0) {
3707 if (s
&& !is_spare
) {
3708 show_conflicts(tbl_ent
->anchor
->family_num
, *super_list
);
3710 } else if (!s
&& !is_spare
)
3723 fprintf(stderr
, "Chose family %#x on '%s', "
3724 "assemble conflicts to new container with '--update=uuid'\n",
3725 __le32_to_cpu(s
->anchor
->family_num
), s
->disks
->devname
);
3727 /* collect all dl's onto 'champion', and update them to
3728 * champion's version of the status
3730 for (s
= *super_list
; s
; s
= s
->next
) {
3731 struct imsm_super
*mpb
= champion
->anchor
;
3732 struct dl
*dl
= s
->disks
;
3737 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3738 struct imsm_disk
*disk
;
3740 disk
= __serial_to_disk(dl
->serial
, mpb
, &dl
->index
);
3743 /* only set index on disks that are a member of
3744 * a populated contianer, i.e. one with
3747 if (is_failed(&dl
->disk
))
3749 else if (is_spare(&dl
->disk
))
3755 if (i
>= mpb
->num_disks
) {
3756 struct intel_disk
*idisk
;
3758 idisk
= disk_list_get(dl
->serial
, disk_list
);
3759 if (idisk
&& is_spare(&idisk
->disk
) &&
3760 !is_failed(&idisk
->disk
) && !is_configured(&idisk
->disk
))
3768 dl
->next
= champion
->disks
;
3769 champion
->disks
= dl
;
3773 /* delete 'champion' from super_list */
3774 for (del
= super_list
; *del
; ) {
3775 if (*del
== champion
) {
3776 *del
= (*del
)->next
;
3779 del
= &(*del
)->next
;
3781 champion
->next
= NULL
;
3785 struct intel_disk
*idisk
= disk_list
;
3787 disk_list
= disk_list
->next
;
3794 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
3798 struct intel_super
*super_list
= NULL
;
3799 struct intel_super
*super
= NULL
;
3800 int devnum
= fd2devnum(fd
);
3806 /* check if 'fd' an opened container */
3807 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
3811 if (sra
->array
.major_version
!= -1 ||
3812 sra
->array
.minor_version
!= -2 ||
3813 strcmp(sra
->text_version
, "imsm") != 0) {
3818 for (sd
= sra
->devs
, i
= 0; sd
; sd
= sd
->next
, i
++) {
3819 struct intel_super
*s
= alloc_super();
3827 s
->next
= super_list
;
3831 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3832 dfd
= dev_open(nm
, O_RDWR
);
3836 rv
= find_intel_hba_capability(dfd
, s
, devname
);
3837 /* no orom/efi or non-intel hba of the disk */
3841 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3843 /* retry the load if we might have raced against mdmon */
3844 if (err
== 3 && mdmon_running(devnum
))
3845 for (retry
= 0; retry
< 3; retry
++) {
3847 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3855 /* all mpbs enter, maybe one leaves */
3856 super
= imsm_thunderdome(&super_list
, i
);
3862 if (find_missing(super
) != 0) {
3868 /* load migration record */
3869 err
= load_imsm_migr_rec(super
, NULL
);
3875 /* Check migration compatibility */
3876 if (check_mpb_migr_compatibility(super
) != 0) {
3877 fprintf(stderr
, Name
": Unsupported migration detected");
3879 fprintf(stderr
, " on %s\n", devname
);
3881 fprintf(stderr
, " (IMSM).\n");
3890 while (super_list
) {
3891 struct intel_super
*s
= super_list
;
3893 super_list
= super_list
->next
;
3902 st
->container_dev
= devnum
;
3903 if (err
== 0 && st
->ss
== NULL
) {
3904 st
->ss
= &super_imsm
;
3905 st
->minor_version
= 0;
3906 st
->max_devs
= IMSM_MAX_DEVICES
;
3911 static int load_container_imsm(struct supertype
*st
, int fd
, char *devname
)
3913 return load_super_imsm_all(st
, fd
, &st
->sb
, devname
);
3917 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
3919 struct intel_super
*super
;
3922 if (test_partition(fd
))
3923 /* IMSM not allowed on partitions */
3926 free_super_imsm(st
);
3928 super
= alloc_super();
3931 Name
": malloc of %zu failed.\n",
3935 /* Load hba and capabilities if they exist.
3936 * But do not preclude loading metadata in case capabilities or hba are
3937 * non-compliant and ignore_hw_compat is set.
3939 rv
= find_intel_hba_capability(fd
, super
, devname
);
3940 /* no orom/efi or non-intel hba of the disk */
3941 if ((rv
!= 0) && (st
->ignore_hw_compat
== 0)) {
3944 Name
": No OROM/EFI properties for %s\n", devname
);
3948 rv
= load_and_parse_mpb(fd
, super
, devname
, 0);
3953 Name
": Failed to load all information "
3954 "sections on %s\n", devname
);
3960 if (st
->ss
== NULL
) {
3961 st
->ss
= &super_imsm
;
3962 st
->minor_version
= 0;
3963 st
->max_devs
= IMSM_MAX_DEVICES
;
3966 /* load migration record */
3967 if (load_imsm_migr_rec(super
, NULL
) == 0) {
3968 /* Check for unsupported migration features */
3969 if (check_mpb_migr_compatibility(super
) != 0) {
3971 Name
": Unsupported migration detected");
3973 fprintf(stderr
, " on %s\n", devname
);
3975 fprintf(stderr
, " (IMSM).\n");
3983 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
3985 if (info
->level
== 1)
3987 return info
->chunk_size
>> 9;
3990 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
, int num_domains
)
3994 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
3995 num_stripes
/= num_domains
;
4000 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
4002 if (info
->level
== 1)
4003 return info
->size
* 2;
4005 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
4008 static void imsm_update_version_info(struct intel_super
*super
)
4010 /* update the version and attributes */
4011 struct imsm_super
*mpb
= super
->anchor
;
4013 struct imsm_dev
*dev
;
4014 struct imsm_map
*map
;
4017 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4018 dev
= get_imsm_dev(super
, i
);
4019 map
= get_imsm_map(dev
, 0);
4020 if (__le32_to_cpu(dev
->size_high
) > 0)
4021 mpb
->attributes
|= MPB_ATTRIB_2TB
;
4023 /* FIXME detect when an array spans a port multiplier */
4025 mpb
->attributes
|= MPB_ATTRIB_PM
;
4028 if (mpb
->num_raid_devs
> 1 ||
4029 mpb
->attributes
!= MPB_ATTRIB_CHECKSUM_VERIFY
) {
4030 version
= MPB_VERSION_ATTRIBS
;
4031 switch (get_imsm_raid_level(map
)) {
4032 case 0: mpb
->attributes
|= MPB_ATTRIB_RAID0
; break;
4033 case 1: mpb
->attributes
|= MPB_ATTRIB_RAID1
; break;
4034 case 10: mpb
->attributes
|= MPB_ATTRIB_RAID10
; break;
4035 case 5: mpb
->attributes
|= MPB_ATTRIB_RAID5
; break;
4038 if (map
->num_members
>= 5)
4039 version
= MPB_VERSION_5OR6_DISK_ARRAY
;
4040 else if (dev
->status
== DEV_CLONE_N_GO
)
4041 version
= MPB_VERSION_CNG
;
4042 else if (get_imsm_raid_level(map
) == 5)
4043 version
= MPB_VERSION_RAID5
;
4044 else if (map
->num_members
>= 3)
4045 version
= MPB_VERSION_3OR4_DISK_ARRAY
;
4046 else if (get_imsm_raid_level(map
) == 1)
4047 version
= MPB_VERSION_RAID1
;
4049 version
= MPB_VERSION_RAID0
;
4051 strcpy(((char *) mpb
->sig
) + strlen(MPB_SIGNATURE
), version
);
4055 static int check_name(struct intel_super
*super
, char *name
, int quiet
)
4057 struct imsm_super
*mpb
= super
->anchor
;
4058 char *reason
= NULL
;
4061 if (strlen(name
) > MAX_RAID_SERIAL_LEN
)
4062 reason
= "must be 16 characters or less";
4064 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4065 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
4067 if (strncmp((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
) == 0) {
4068 reason
= "already exists";
4073 if (reason
&& !quiet
)
4074 fprintf(stderr
, Name
": imsm volume name %s\n", reason
);
4079 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
4080 unsigned long long size
, char *name
,
4081 char *homehost
, int *uuid
)
4083 /* We are creating a volume inside a pre-existing container.
4084 * so st->sb is already set.
4086 struct intel_super
*super
= st
->sb
;
4087 struct imsm_super
*mpb
= super
->anchor
;
4088 struct intel_dev
*dv
;
4089 struct imsm_dev
*dev
;
4090 struct imsm_vol
*vol
;
4091 struct imsm_map
*map
;
4092 int idx
= mpb
->num_raid_devs
;
4094 unsigned long long array_blocks
;
4095 size_t size_old
, size_new
;
4096 __u32 num_data_stripes
;
4098 if (super
->orom
&& mpb
->num_raid_devs
>= super
->orom
->vpa
) {
4099 fprintf(stderr
, Name
": This imsm-container already has the "
4100 "maximum of %d volumes\n", super
->orom
->vpa
);
4104 /* ensure the mpb is large enough for the new data */
4105 size_old
= __le32_to_cpu(mpb
->mpb_size
);
4106 size_new
= disks_to_mpb_size(info
->nr_disks
);
4107 if (size_new
> size_old
) {
4109 size_t size_round
= ROUND_UP(size_new
, 512);
4111 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
4112 fprintf(stderr
, Name
": could not allocate new mpb\n");
4115 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4116 fprintf(stderr
, Name
4117 ": %s could not allocate migr_rec buffer\n",
4123 memcpy(mpb_new
, mpb
, size_old
);
4126 super
->anchor
= mpb_new
;
4127 mpb
->mpb_size
= __cpu_to_le32(size_new
);
4128 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
4130 super
->current_vol
= idx
;
4131 /* when creating the first raid device in this container set num_disks
4132 * to zero, i.e. delete this spare and add raid member devices in
4133 * add_to_super_imsm_volume()
4135 if (super
->current_vol
== 0)
4138 if (!check_name(super
, name
, 0))
4140 dv
= malloc(sizeof(*dv
));
4142 fprintf(stderr
, Name
": failed to allocate device list entry\n");
4145 dev
= calloc(1, sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
4148 fprintf(stderr
, Name
": could not allocate raid device\n");
4152 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
4153 if (info
->level
== 1)
4154 array_blocks
= info_to_blocks_per_member(info
);
4156 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
4157 info
->layout
, info
->chunk_size
,
4159 /* round array size down to closest MB */
4160 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
4162 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
4163 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
4164 dev
->status
= (DEV_READ_COALESCING
| DEV_WRITE_COALESCING
);
4166 vol
->migr_state
= 0;
4167 set_migr_type(dev
, MIGR_INIT
);
4169 vol
->curr_migr_unit
= 0;
4170 map
= get_imsm_map(dev
, 0);
4171 map
->pba_of_lba0
= __cpu_to_le32(super
->create_offset
);
4172 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
4173 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
4174 map
->failed_disk_num
= ~0;
4175 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
4176 IMSM_T_STATE_NORMAL
;
4179 if (info
->level
== 1 && info
->raid_disks
> 2) {
4182 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
4183 "in a raid1 volume\n");
4187 map
->raid_level
= info
->level
;
4188 if (info
->level
== 10) {
4189 map
->raid_level
= 1;
4190 map
->num_domains
= info
->raid_disks
/ 2;
4191 } else if (info
->level
== 1)
4192 map
->num_domains
= info
->raid_disks
;
4194 map
->num_domains
= 1;
4196 num_data_stripes
= info_to_num_data_stripes(info
, map
->num_domains
);
4197 map
->num_data_stripes
= __cpu_to_le32(num_data_stripes
);
4199 map
->num_members
= info
->raid_disks
;
4200 for (i
= 0; i
< map
->num_members
; i
++) {
4201 /* initialized in add_to_super */
4202 set_imsm_ord_tbl_ent(map
, i
, IMSM_ORD_REBUILD
);
4204 mpb
->num_raid_devs
++;
4207 dv
->index
= super
->current_vol
;
4208 dv
->next
= super
->devlist
;
4209 super
->devlist
= dv
;
4211 imsm_update_version_info(super
);
4216 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
4217 unsigned long long size
, char *name
,
4218 char *homehost
, int *uuid
)
4220 /* This is primarily called by Create when creating a new array.
4221 * We will then get add_to_super called for each component, and then
4222 * write_init_super called to write it out to each device.
4223 * For IMSM, Create can create on fresh devices or on a pre-existing
4225 * To create on a pre-existing array a different method will be called.
4226 * This one is just for fresh drives.
4228 struct intel_super
*super
;
4229 struct imsm_super
*mpb
;
4234 return init_super_imsm_volume(st
, info
, size
, name
, homehost
, uuid
);
4237 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
4241 super
= alloc_super();
4242 if (super
&& posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
4247 fprintf(stderr
, Name
4248 ": %s could not allocate superblock\n", __func__
);
4251 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4252 fprintf(stderr
, Name
4253 ": %s could not allocate migr_rec buffer\n", __func__
);
4258 memset(super
->buf
, 0, mpb_size
);
4260 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4264 /* zeroing superblock */
4268 mpb
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4270 version
= (char *) mpb
->sig
;
4271 strcpy(version
, MPB_SIGNATURE
);
4272 version
+= strlen(MPB_SIGNATURE
);
4273 strcpy(version
, MPB_VERSION_RAID0
);
4279 static int add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
4280 int fd
, char *devname
)
4282 struct intel_super
*super
= st
->sb
;
4283 struct imsm_super
*mpb
= super
->anchor
;
4285 struct imsm_dev
*dev
;
4286 struct imsm_map
*map
;
4289 dev
= get_imsm_dev(super
, super
->current_vol
);
4290 map
= get_imsm_map(dev
, 0);
4292 if (! (dk
->state
& (1<<MD_DISK_SYNC
))) {
4293 fprintf(stderr
, Name
": %s: Cannot add spare devices to IMSM volume\n",
4299 /* we're doing autolayout so grab the pre-marked (in
4300 * validate_geometry) raid_disk
4302 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4303 if (dl
->raiddisk
== dk
->raid_disk
)
4306 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4307 if (dl
->major
== dk
->major
&&
4308 dl
->minor
== dk
->minor
)
4313 fprintf(stderr
, Name
": %s is not a member of the same container\n", devname
);
4317 /* add a pristine spare to the metadata */
4318 if (dl
->index
< 0) {
4319 dl
->index
= super
->anchor
->num_disks
;
4320 super
->anchor
->num_disks
++;
4322 /* Check the device has not already been added */
4323 slot
= get_imsm_disk_slot(map
, dl
->index
);
4325 (get_imsm_ord_tbl_ent(dev
, slot
, -1) & IMSM_ORD_REBUILD
) == 0) {
4326 fprintf(stderr
, Name
": %s has been included in this array twice\n",
4330 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
4331 dl
->disk
.status
= CONFIGURED_DISK
;
4333 /* if we are creating the first raid device update the family number */
4334 if (super
->current_vol
== 0) {
4336 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
4337 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
4339 if (!_dev
|| !_disk
) {
4340 fprintf(stderr
, Name
": BUG mpb setup error\n");
4346 sum
+= __gen_imsm_checksum(mpb
);
4347 mpb
->family_num
= __cpu_to_le32(sum
);
4348 mpb
->orig_family_num
= mpb
->family_num
;
4355 static int add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
4356 int fd
, char *devname
)
4358 struct intel_super
*super
= st
->sb
;
4360 unsigned long long size
;
4365 /* If we are on an RAID enabled platform check that the disk is
4366 * attached to the raid controller.
4367 * We do not need to test disks attachment for container based additions,
4368 * they shall be already tested when container was created/assembled.
4370 rv
= find_intel_hba_capability(fd
, super
, devname
);
4371 /* no orom/efi or non-intel hba of the disk */
4373 dprintf("capability: %p fd: %d ret: %d\n",
4374 super
->orom
, fd
, rv
);
4378 if (super
->current_vol
>= 0)
4379 return add_to_super_imsm_volume(st
, dk
, fd
, devname
);
4382 dd
= malloc(sizeof(*dd
));
4385 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4388 memset(dd
, 0, sizeof(*dd
));
4389 dd
->major
= major(stb
.st_rdev
);
4390 dd
->minor
= minor(stb
.st_rdev
);
4392 dd
->devname
= devname
? strdup(devname
) : NULL
;
4395 dd
->action
= DISK_ADD
;
4396 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
4399 Name
": failed to retrieve scsi serial, aborting\n");
4404 get_dev_size(fd
, NULL
, &size
);
4406 serialcpy(dd
->disk
.serial
, dd
->serial
);
4407 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
4408 dd
->disk
.status
= SPARE_DISK
;
4409 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
4410 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
4412 dd
->disk
.scsi_id
= __cpu_to_le32(0);
4414 if (st
->update_tail
) {
4415 dd
->next
= super
->disk_mgmt_list
;
4416 super
->disk_mgmt_list
= dd
;
4418 dd
->next
= super
->disks
;
4420 super
->updates_pending
++;
4427 static int remove_from_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
)
4429 struct intel_super
*super
= st
->sb
;
4432 /* remove from super works only in mdmon - for communication
4433 * manager - monitor. Check if communication memory buffer
4436 if (!st
->update_tail
) {
4438 Name
": %s shall be used in mdmon context only"
4439 "(line %d).\n", __func__
, __LINE__
);
4442 dd
= malloc(sizeof(*dd
));
4445 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4448 memset(dd
, 0, sizeof(*dd
));
4449 dd
->major
= dk
->major
;
4450 dd
->minor
= dk
->minor
;
4453 dd
->disk
.status
= SPARE_DISK
;
4454 dd
->action
= DISK_REMOVE
;
4456 dd
->next
= super
->disk_mgmt_list
;
4457 super
->disk_mgmt_list
= dd
;
4463 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
);
4467 struct imsm_super anchor
;
4468 } spare_record
__attribute__ ((aligned(512)));
4470 /* spare records have their own family number and do not have any defined raid
4473 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
4475 struct imsm_super
*mpb
= super
->anchor
;
4476 struct imsm_super
*spare
= &spare_record
.anchor
;
4480 spare
->mpb_size
= __cpu_to_le32(sizeof(struct imsm_super
)),
4481 spare
->generation_num
= __cpu_to_le32(1UL),
4482 spare
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4483 spare
->num_disks
= 1,
4484 spare
->num_raid_devs
= 0,
4485 spare
->cache_size
= mpb
->cache_size
,
4486 spare
->pwr_cycle_count
= __cpu_to_le32(1),
4488 snprintf((char *) spare
->sig
, MAX_SIGNATURE_LENGTH
,
4489 MPB_SIGNATURE MPB_VERSION_RAID0
);
4491 for (d
= super
->disks
; d
; d
= d
->next
) {
4495 spare
->disk
[0] = d
->disk
;
4496 sum
= __gen_imsm_checksum(spare
);
4497 spare
->family_num
= __cpu_to_le32(sum
);
4498 spare
->orig_family_num
= 0;
4499 sum
= __gen_imsm_checksum(spare
);
4500 spare
->check_sum
= __cpu_to_le32(sum
);
4502 if (store_imsm_mpb(d
->fd
, spare
)) {
4503 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4504 __func__
, d
->major
, d
->minor
, strerror(errno
));
4516 static int write_super_imsm(struct supertype
*st
, int doclose
)
4518 struct intel_super
*super
= st
->sb
;
4519 struct imsm_super
*mpb
= super
->anchor
;
4525 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
4527 int clear_migration_record
= 1;
4529 /* 'generation' is incremented everytime the metadata is written */
4530 generation
= __le32_to_cpu(mpb
->generation_num
);
4532 mpb
->generation_num
= __cpu_to_le32(generation
);
4534 /* fix up cases where previous mdadm releases failed to set
4537 if (mpb
->orig_family_num
== 0)
4538 mpb
->orig_family_num
= mpb
->family_num
;
4540 for (d
= super
->disks
; d
; d
= d
->next
) {
4544 mpb
->disk
[d
->index
] = d
->disk
;
4548 for (d
= super
->missing
; d
; d
= d
->next
) {
4549 mpb
->disk
[d
->index
] = d
->disk
;
4552 mpb
->num_disks
= num_disks
;
4553 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
4555 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4556 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
4557 struct imsm_dev
*dev2
= get_imsm_dev(super
, i
);
4559 imsm_copy_dev(dev
, dev2
);
4560 mpb_size
+= sizeof_imsm_dev(dev
, 0);
4562 if (is_gen_migration(dev2
))
4563 clear_migration_record
= 0;
4565 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
4566 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4568 /* recalculate checksum */
4569 sum
= __gen_imsm_checksum(mpb
);
4570 mpb
->check_sum
= __cpu_to_le32(sum
);
4572 if (clear_migration_record
)
4573 memset(super
->migr_rec_buf
, 0, 512);
4575 /* write the mpb for disks that compose raid devices */
4576 for (d
= super
->disks
; d
; d
= d
->next
) {
4579 if (store_imsm_mpb(d
->fd
, mpb
))
4580 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4581 __func__
, d
->major
, d
->minor
, strerror(errno
));
4582 if (clear_migration_record
) {
4583 unsigned long long dsize
;
4585 get_dev_size(d
->fd
, NULL
, &dsize
);
4586 if (lseek64(d
->fd
, dsize
- 512, SEEK_SET
) >= 0) {
4587 if (write(d
->fd
, super
->migr_rec_buf
, 512) != 512)
4588 perror("Write migr_rec failed");
4598 return write_super_imsm_spares(super
, doclose
);
4604 static int create_array(struct supertype
*st
, int dev_idx
)
4607 struct imsm_update_create_array
*u
;
4608 struct intel_super
*super
= st
->sb
;
4609 struct imsm_dev
*dev
= get_imsm_dev(super
, dev_idx
);
4610 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4611 struct disk_info
*inf
;
4612 struct imsm_disk
*disk
;
4615 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0) +
4616 sizeof(*inf
) * map
->num_members
;
4619 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4624 u
->type
= update_create_array
;
4625 u
->dev_idx
= dev_idx
;
4626 imsm_copy_dev(&u
->dev
, dev
);
4627 inf
= get_disk_info(u
);
4628 for (i
= 0; i
< map
->num_members
; i
++) {
4629 int idx
= get_imsm_disk_idx(dev
, i
, -1);
4631 disk
= get_imsm_disk(super
, idx
);
4632 serialcpy(inf
[i
].serial
, disk
->serial
);
4634 append_metadata_update(st
, u
, len
);
4639 static int mgmt_disk(struct supertype
*st
)
4641 struct intel_super
*super
= st
->sb
;
4643 struct imsm_update_add_remove_disk
*u
;
4645 if (!super
->disk_mgmt_list
)
4651 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4656 u
->type
= update_add_remove_disk
;
4657 append_metadata_update(st
, u
, len
);
4662 static int write_init_super_imsm(struct supertype
*st
)
4664 struct intel_super
*super
= st
->sb
;
4665 int current_vol
= super
->current_vol
;
4667 /* we are done with current_vol reset it to point st at the container */
4668 super
->current_vol
= -1;
4670 if (st
->update_tail
) {
4671 /* queue the recently created array / added disk
4672 * as a metadata update */
4675 /* determine if we are creating a volume or adding a disk */
4676 if (current_vol
< 0) {
4677 /* in the mgmt (add/remove) disk case we are running
4678 * in mdmon context, so don't close fd's
4680 return mgmt_disk(st
);
4682 rv
= create_array(st
, current_vol
);
4687 for (d
= super
->disks
; d
; d
= d
->next
)
4688 Kill(d
->devname
, NULL
, 0, 1, 1);
4689 return write_super_imsm(st
, 1);
4694 static int store_super_imsm(struct supertype
*st
, int fd
)
4696 struct intel_super
*super
= st
->sb
;
4697 struct imsm_super
*mpb
= super
? super
->anchor
: NULL
;
4703 return store_imsm_mpb(fd
, mpb
);
4709 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
4711 return __le32_to_cpu(mpb
->bbm_log_size
);
4715 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
4716 int layout
, int raiddisks
, int chunk
,
4717 unsigned long long size
, char *dev
,
4718 unsigned long long *freesize
,
4722 unsigned long long ldsize
;
4723 struct intel_super
*super
=NULL
;
4726 if (level
!= LEVEL_CONTAINER
)
4731 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
4734 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
4735 dev
, strerror(errno
));
4738 if (!get_dev_size(fd
, dev
, &ldsize
)) {
4743 /* capabilities retrieve could be possible
4744 * note that there is no fd for the disks in array.
4746 super
= alloc_super();
4749 Name
": malloc of %zu failed.\n",
4755 rv
= find_intel_hba_capability(fd
, super
, verbose
? dev
: NULL
);
4759 fd2devname(fd
, str
);
4760 dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n",
4761 fd
, str
, super
->orom
, rv
, raiddisks
);
4763 /* no orom/efi or non-intel hba of the disk */
4769 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4771 fprintf(stderr
, Name
": %d exceeds maximum number of"
4772 " platform supported disks: %d\n",
4773 raiddisks
, super
->orom
->tds
);
4779 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
4785 static unsigned long long find_size(struct extent
*e
, int *idx
, int num_extents
)
4787 const unsigned long long base_start
= e
[*idx
].start
;
4788 unsigned long long end
= base_start
+ e
[*idx
].size
;
4791 if (base_start
== end
)
4795 for (i
= *idx
; i
< num_extents
; i
++) {
4796 /* extend overlapping extents */
4797 if (e
[i
].start
>= base_start
&&
4798 e
[i
].start
<= end
) {
4801 if (e
[i
].start
+ e
[i
].size
> end
)
4802 end
= e
[i
].start
+ e
[i
].size
;
4803 } else if (e
[i
].start
> end
) {
4809 return end
- base_start
;
4812 static unsigned long long merge_extents(struct intel_super
*super
, int sum_extents
)
4814 /* build a composite disk with all known extents and generate a new
4815 * 'maxsize' given the "all disks in an array must share a common start
4816 * offset" constraint
4818 struct extent
*e
= calloc(sum_extents
, sizeof(*e
));
4822 unsigned long long pos
;
4823 unsigned long long start
= 0;
4824 unsigned long long maxsize
;
4825 unsigned long reserve
;
4830 /* coalesce and sort all extents. also, check to see if we need to
4831 * reserve space between member arrays
4834 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
4837 for (i
= 0; i
< dl
->extent_cnt
; i
++)
4840 qsort(e
, sum_extents
, sizeof(*e
), cmp_extent
);
4845 while (i
< sum_extents
) {
4846 e
[j
].start
= e
[i
].start
;
4847 e
[j
].size
= find_size(e
, &i
, sum_extents
);
4849 if (e
[j
-1].size
== 0)
4858 unsigned long long esize
;
4860 esize
= e
[i
].start
- pos
;
4861 if (esize
>= maxsize
) {
4866 pos
= e
[i
].start
+ e
[i
].size
;
4868 } while (e
[i
-1].size
);
4874 /* FIXME assumes volume at offset 0 is the first volume in a
4877 if (start_extent
> 0)
4878 reserve
= IMSM_RESERVED_SECTORS
; /* gap between raid regions */
4882 if (maxsize
< reserve
)
4885 super
->create_offset
= ~((__u32
) 0);
4886 if (start
+ reserve
> super
->create_offset
)
4887 return 0; /* start overflows create_offset */
4888 super
->create_offset
= start
+ reserve
;
4890 return maxsize
- reserve
;
4893 static int is_raid_level_supported(const struct imsm_orom
*orom
, int level
, int raiddisks
)
4895 if (level
< 0 || level
== 6 || level
== 4)
4898 /* if we have an orom prevent invalid raid levels */
4901 case 0: return imsm_orom_has_raid0(orom
);
4904 return imsm_orom_has_raid1e(orom
);
4905 return imsm_orom_has_raid1(orom
) && raiddisks
== 2;
4906 case 10: return imsm_orom_has_raid10(orom
) && raiddisks
== 4;
4907 case 5: return imsm_orom_has_raid5(orom
) && raiddisks
> 2;
4910 return 1; /* not on an Intel RAID platform so anything goes */
4916 #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
4918 * validate volume parameters with OROM/EFI capabilities
4921 validate_geometry_imsm_orom(struct intel_super
*super
, int level
, int layout
,
4922 int raiddisks
, int *chunk
, int verbose
)
4927 /* validate container capabilities */
4928 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4930 fprintf(stderr
, Name
": %d exceeds maximum number of"
4931 " platform supported disks: %d\n",
4932 raiddisks
, super
->orom
->tds
);
4936 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
4937 if (super
->orom
&& (!is_raid_level_supported(super
->orom
, level
,
4939 pr_vrb(": platform does not support raid%d with %d disk%s\n",
4940 level
, raiddisks
, raiddisks
> 1 ? "s" : "");
4943 if (super
->orom
&& level
!= 1) {
4944 if (chunk
&& (*chunk
== 0 || *chunk
== UnSet
))
4945 *chunk
= imsm_orom_default_chunk(super
->orom
);
4946 else if (chunk
&& !imsm_orom_has_chunk(super
->orom
, *chunk
)) {
4947 pr_vrb(": platform does not support a chunk size of: "
4952 if (layout
!= imsm_level_to_layout(level
)) {
4954 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
4955 else if (level
== 10)
4956 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
4958 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
4965 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
4966 * FIX ME add ahci details
4968 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
4969 int layout
, int raiddisks
, int *chunk
,
4970 unsigned long long size
, char *dev
,
4971 unsigned long long *freesize
,
4975 struct intel_super
*super
= st
->sb
;
4976 struct imsm_super
*mpb
= super
->anchor
;
4978 unsigned long long pos
= 0;
4979 unsigned long long maxsize
;
4983 /* We must have the container info already read in. */
4987 if (!validate_geometry_imsm_orom(super
, level
, layout
, raiddisks
, chunk
, verbose
)) {
4988 fprintf(stderr
, Name
": RAID gemetry validation failed. "
4989 "Cannot proceed with the action(s).\n");
4993 /* General test: make sure there is space for
4994 * 'raiddisks' device extents of size 'size' at a given
4997 unsigned long long minsize
= size
;
4998 unsigned long long start_offset
= MaxSector
;
5001 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
5002 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5007 e
= get_extents(super
, dl
);
5010 unsigned long long esize
;
5011 esize
= e
[i
].start
- pos
;
5012 if (esize
>= minsize
)
5014 if (found
&& start_offset
== MaxSector
) {
5017 } else if (found
&& pos
!= start_offset
) {
5021 pos
= e
[i
].start
+ e
[i
].size
;
5023 } while (e
[i
-1].size
);
5028 if (dcnt
< raiddisks
) {
5030 fprintf(stderr
, Name
": imsm: Not enough "
5031 "devices with space for this array "
5039 /* This device must be a member of the set */
5040 if (stat(dev
, &stb
) < 0)
5042 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
5044 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5045 if (dl
->major
== (int)major(stb
.st_rdev
) &&
5046 dl
->minor
== (int)minor(stb
.st_rdev
))
5051 fprintf(stderr
, Name
": %s is not in the "
5052 "same imsm set\n", dev
);
5054 } else if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
) {
5055 /* If a volume is present then the current creation attempt
5056 * cannot incorporate new spares because the orom may not
5057 * understand this configuration (all member disks must be
5058 * members of each array in the container).
5060 fprintf(stderr
, Name
": %s is a spare and a volume"
5061 " is already defined for this container\n", dev
);
5062 fprintf(stderr
, Name
": The option-rom requires all member"
5063 " disks to be a member of all volumes\n");
5067 /* retrieve the largest free space block */
5068 e
= get_extents(super
, dl
);
5073 unsigned long long esize
;
5075 esize
= e
[i
].start
- pos
;
5076 if (esize
>= maxsize
)
5078 pos
= e
[i
].start
+ e
[i
].size
;
5080 } while (e
[i
-1].size
);
5085 fprintf(stderr
, Name
": unable to determine free space for: %s\n",
5089 if (maxsize
< size
) {
5091 fprintf(stderr
, Name
": %s not enough space (%llu < %llu)\n",
5092 dev
, maxsize
, size
);
5096 /* count total number of extents for merge */
5098 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5100 i
+= dl
->extent_cnt
;
5102 maxsize
= merge_extents(super
, i
);
5103 if (maxsize
< size
|| maxsize
== 0) {
5105 fprintf(stderr
, Name
": not enough space after merge (%llu < %llu)\n",
5110 *freesize
= maxsize
;
5115 static int reserve_space(struct supertype
*st
, int raiddisks
,
5116 unsigned long long size
, int chunk
,
5117 unsigned long long *freesize
)
5119 struct intel_super
*super
= st
->sb
;
5120 struct imsm_super
*mpb
= super
->anchor
;
5125 unsigned long long maxsize
;
5126 unsigned long long minsize
;
5130 /* find the largest common start free region of the possible disks */
5134 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5140 /* don't activate new spares if we are orom constrained
5141 * and there is already a volume active in the container
5143 if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
)
5146 e
= get_extents(super
, dl
);
5149 for (i
= 1; e
[i
-1].size
; i
++)
5157 maxsize
= merge_extents(super
, extent_cnt
);
5161 minsize
= chunk
* 2;
5163 if (cnt
< raiddisks
||
5164 (super
->orom
&& used
&& used
!= raiddisks
) ||
5165 maxsize
< minsize
||
5167 fprintf(stderr
, Name
": not enough devices with space to create array.\n");
5168 return 0; /* No enough free spaces large enough */
5180 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5182 dl
->raiddisk
= cnt
++;
5189 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
5190 int raiddisks
, int *chunk
, unsigned long long size
,
5191 char *dev
, unsigned long long *freesize
,
5199 * if given unused devices create a container
5200 * if given given devices in a container create a member volume
5202 if (level
== LEVEL_CONTAINER
) {
5203 /* Must be a fresh device to add to a container */
5204 return validate_geometry_imsm_container(st
, level
, layout
,
5206 chunk
?*chunk
:0, size
,
5212 if (st
->sb
&& freesize
) {
5213 /* we are being asked to automatically layout a
5214 * new volume based on the current contents of
5215 * the container. If the the parameters can be
5216 * satisfied reserve_space will record the disks,
5217 * start offset, and size of the volume to be
5218 * created. add_to_super and getinfo_super
5219 * detect when autolayout is in progress.
5221 if (!validate_geometry_imsm_orom(st
->sb
, level
, layout
,
5225 return reserve_space(st
, raiddisks
, size
,
5226 chunk
?*chunk
:0, freesize
);
5231 /* creating in a given container */
5232 return validate_geometry_imsm_volume(st
, level
, layout
,
5233 raiddisks
, chunk
, size
,
5234 dev
, freesize
, verbose
);
5237 /* This device needs to be a device in an 'imsm' container */
5238 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
5242 Name
": Cannot create this array on device %s\n",
5247 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
5249 fprintf(stderr
, Name
": Cannot open %s: %s\n",
5250 dev
, strerror(errno
));
5253 /* Well, it is in use by someone, maybe an 'imsm' container. */
5254 cfd
= open_container(fd
);
5258 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
5262 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
5263 if (sra
&& sra
->array
.major_version
== -1 &&
5264 strcmp(sra
->text_version
, "imsm") == 0)
5268 /* This is a member of a imsm container. Load the container
5269 * and try to create a volume
5271 struct intel_super
*super
;
5273 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
) == 0) {
5275 st
->container_dev
= fd2devnum(cfd
);
5277 return validate_geometry_imsm_volume(st
, level
, layout
,
5285 fprintf(stderr
, Name
": failed container membership check\n");
5291 static void default_geometry_imsm(struct supertype
*st
, int *level
, int *layout
, int *chunk
)
5293 struct intel_super
*super
= st
->sb
;
5295 if (level
&& *level
== UnSet
)
5296 *level
= LEVEL_CONTAINER
;
5298 if (level
&& layout
&& *layout
== UnSet
)
5299 *layout
= imsm_level_to_layout(*level
);
5301 if (chunk
&& (*chunk
== UnSet
|| *chunk
== 0) &&
5302 super
&& super
->orom
)
5303 *chunk
= imsm_orom_default_chunk(super
->orom
);
5306 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
);
5308 static int kill_subarray_imsm(struct supertype
*st
)
5310 /* remove the subarray currently referenced by ->current_vol */
5312 struct intel_dev
**dp
;
5313 struct intel_super
*super
= st
->sb
;
5314 __u8 current_vol
= super
->current_vol
;
5315 struct imsm_super
*mpb
= super
->anchor
;
5317 if (super
->current_vol
< 0)
5319 super
->current_vol
= -1; /* invalidate subarray cursor */
5321 /* block deletions that would change the uuid of active subarrays
5323 * FIXME when immutable ids are available, but note that we'll
5324 * also need to fixup the invalidated/active subarray indexes in
5327 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5330 if (i
< current_vol
)
5332 sprintf(subarray
, "%u", i
);
5333 if (is_subarray_active(subarray
, st
->devname
)) {
5335 Name
": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
5342 if (st
->update_tail
) {
5343 struct imsm_update_kill_array
*u
= malloc(sizeof(*u
));
5347 u
->type
= update_kill_array
;
5348 u
->dev_idx
= current_vol
;
5349 append_metadata_update(st
, u
, sizeof(*u
));
5354 for (dp
= &super
->devlist
; *dp
;)
5355 if ((*dp
)->index
== current_vol
) {
5358 handle_missing(super
, (*dp
)->dev
);
5359 if ((*dp
)->index
> current_vol
)
5364 /* no more raid devices, all active components are now spares,
5365 * but of course failed are still failed
5367 if (--mpb
->num_raid_devs
== 0) {
5370 for (d
= super
->disks
; d
; d
= d
->next
)
5371 if (d
->index
> -2) {
5373 d
->disk
.status
= SPARE_DISK
;
5377 super
->updates_pending
++;
5382 static int update_subarray_imsm(struct supertype
*st
, char *subarray
,
5383 char *update
, struct mddev_ident
*ident
)
5385 /* update the subarray currently referenced by ->current_vol */
5386 struct intel_super
*super
= st
->sb
;
5387 struct imsm_super
*mpb
= super
->anchor
;
5389 if (strcmp(update
, "name") == 0) {
5390 char *name
= ident
->name
;
5394 if (is_subarray_active(subarray
, st
->devname
)) {
5396 Name
": Unable to update name of active subarray\n");
5400 if (!check_name(super
, name
, 0))
5403 vol
= strtoul(subarray
, &ep
, 10);
5404 if (*ep
!= '\0' || vol
>= super
->anchor
->num_raid_devs
)
5407 if (st
->update_tail
) {
5408 struct imsm_update_rename_array
*u
= malloc(sizeof(*u
));
5412 u
->type
= update_rename_array
;
5414 snprintf((char *) u
->name
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5415 append_metadata_update(st
, u
, sizeof(*u
));
5417 struct imsm_dev
*dev
;
5420 dev
= get_imsm_dev(super
, vol
);
5421 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5422 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5423 dev
= get_imsm_dev(super
, i
);
5424 handle_missing(super
, dev
);
5426 super
->updates_pending
++;
5434 static int is_gen_migration(struct imsm_dev
*dev
)
5439 if (!dev
->vol
.migr_state
)
5442 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5447 #endif /* MDASSEMBLE */
5449 static int is_rebuilding(struct imsm_dev
*dev
)
5451 struct imsm_map
*migr_map
;
5453 if (!dev
->vol
.migr_state
)
5456 if (migr_type(dev
) != MIGR_REBUILD
)
5459 migr_map
= get_imsm_map(dev
, 1);
5461 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
5467 static void update_recovery_start(struct intel_super
*super
,
5468 struct imsm_dev
*dev
,
5469 struct mdinfo
*array
)
5471 struct mdinfo
*rebuild
= NULL
;
5475 if (!is_rebuilding(dev
))
5478 /* Find the rebuild target, but punt on the dual rebuild case */
5479 for (d
= array
->devs
; d
; d
= d
->next
)
5480 if (d
->recovery_start
== 0) {
5487 /* (?) none of the disks are marked with
5488 * IMSM_ORD_REBUILD, so assume they are missing and the
5489 * disk_ord_tbl was not correctly updated
5491 dprintf("%s: failed to locate out-of-sync disk\n", __func__
);
5495 units
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
5496 rebuild
->recovery_start
= units
* blocks_per_migr_unit(super
, dev
);
5500 static int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
);
5503 static struct mdinfo
*container_content_imsm(struct supertype
*st
, char *subarray
)
5505 /* Given a container loaded by load_super_imsm_all,
5506 * extract information about all the arrays into
5508 * If 'subarray' is given, just extract info about that array.
5510 * For each imsm_dev create an mdinfo, fill it in,
5511 * then look for matching devices in super->disks
5512 * and create appropriate device mdinfo.
5514 struct intel_super
*super
= st
->sb
;
5515 struct imsm_super
*mpb
= super
->anchor
;
5516 struct mdinfo
*rest
= NULL
;
5520 int spare_disks
= 0;
5522 /* do not assemble arrays when not all attributes are supported */
5523 if (imsm_check_attributes(mpb
->attributes
) == 0) {
5524 fprintf(stderr
, Name
": IMSM metadata loading not allowed "
5525 "due to attributes incompatibility.\n");
5529 /* check for bad blocks */
5530 if (imsm_bbm_log_size(super
->anchor
))
5533 /* count spare devices, not used in maps
5535 for (d
= super
->disks
; d
; d
= d
->next
)
5539 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5540 struct imsm_dev
*dev
;
5541 struct imsm_map
*map
;
5542 struct imsm_map
*map2
;
5543 struct mdinfo
*this;
5548 (i
!= strtoul(subarray
, &ep
, 10) || *ep
!= '\0'))
5551 dev
= get_imsm_dev(super
, i
);
5552 map
= get_imsm_map(dev
, 0);
5553 map2
= get_imsm_map(dev
, 1);
5555 /* do not publish arrays that are in the middle of an
5556 * unsupported migration
5558 if (dev
->vol
.migr_state
&&
5559 (migr_type(dev
) == MIGR_STATE_CHANGE
)) {
5560 fprintf(stderr
, Name
": cannot assemble volume '%.16s':"
5561 " unsupported migration in progress\n",
5565 /* do not publish arrays that are not support by controller's
5569 chunk
= __le16_to_cpu(map
->blocks_per_strip
) >> 1;
5571 if (!validate_geometry_imsm_orom(super
,
5572 get_imsm_raid_level(map
), /* RAID level */
5573 imsm_level_to_layout(get_imsm_raid_level(map
)),
5574 map
->num_members
, /* raid disks */
5577 fprintf(stderr
, Name
": RAID gemetry validation failed. "
5578 "Cannot proceed with the action(s).\n");
5581 #endif /* MDASSEMBLE */
5582 this = malloc(sizeof(*this));
5584 fprintf(stderr
, Name
": failed to allocate %zu bytes\n",
5589 super
->current_vol
= i
;
5590 getinfo_super_imsm_volume(st
, this, NULL
);
5592 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
5593 unsigned long long recovery_start
;
5594 struct mdinfo
*info_d
;
5601 idx
= get_imsm_disk_idx(dev
, slot
, 0);
5602 ord
= get_imsm_ord_tbl_ent(dev
, slot
, -1);
5603 for (d
= super
->disks
; d
; d
= d
->next
)
5604 if (d
->index
== idx
)
5607 recovery_start
= MaxSector
;
5610 if (d
&& is_failed(&d
->disk
))
5612 if (ord
& IMSM_ORD_REBUILD
)
5616 * if we skip some disks the array will be assmebled degraded;
5617 * reset resync start to avoid a dirty-degraded
5618 * situation when performing the intial sync
5620 * FIXME handle dirty degraded
5622 if ((skip
|| recovery_start
== 0) && !dev
->vol
.dirty
)
5623 this->resync_start
= MaxSector
;
5627 info_d
= calloc(1, sizeof(*info_d
));
5629 fprintf(stderr
, Name
": failed to allocate disk"
5630 " for volume %.16s\n", dev
->volume
);
5631 info_d
= this->devs
;
5633 struct mdinfo
*d
= info_d
->next
;
5642 info_d
->next
= this->devs
;
5643 this->devs
= info_d
;
5645 info_d
->disk
.number
= d
->index
;
5646 info_d
->disk
.major
= d
->major
;
5647 info_d
->disk
.minor
= d
->minor
;
5648 info_d
->disk
.raid_disk
= slot
;
5649 info_d
->recovery_start
= recovery_start
;
5651 if (slot
< map2
->num_members
)
5652 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5654 this->array
.spare_disks
++;
5656 if (slot
< map
->num_members
)
5657 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5659 this->array
.spare_disks
++;
5661 if (info_d
->recovery_start
== MaxSector
)
5662 this->array
.working_disks
++;
5664 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
5665 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
5666 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
5668 /* now that the disk list is up-to-date fixup recovery_start */
5669 update_recovery_start(super
, dev
, this);
5670 this->array
.spare_disks
+= spare_disks
;
5673 /* check for reshape */
5674 if (this->reshape_active
== 1)
5675 recover_backup_imsm(st
, this);
5680 /* if array has bad blocks, set suitable bit in array status */
5682 rest
->array
.state
|= (1<<MD_SB_BBM_ERRORS
);
5688 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
5690 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5693 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
5694 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
5696 switch (get_imsm_raid_level(map
)) {
5698 return IMSM_T_STATE_FAILED
;
5701 if (failed
< map
->num_members
)
5702 return IMSM_T_STATE_DEGRADED
;
5704 return IMSM_T_STATE_FAILED
;
5709 * check to see if any mirrors have failed, otherwise we
5710 * are degraded. Even numbered slots are mirrored on
5714 /* gcc -Os complains that this is unused */
5715 int insync
= insync
;
5717 for (i
= 0; i
< map
->num_members
; i
++) {
5718 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
5719 int idx
= ord_to_idx(ord
);
5720 struct imsm_disk
*disk
;
5722 /* reset the potential in-sync count on even-numbered
5723 * slots. num_copies is always 2 for imsm raid10
5728 disk
= get_imsm_disk(super
, idx
);
5729 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5732 /* no in-sync disks left in this mirror the
5736 return IMSM_T_STATE_FAILED
;
5739 return IMSM_T_STATE_DEGRADED
;
5743 return IMSM_T_STATE_DEGRADED
;
5745 return IMSM_T_STATE_FAILED
;
5751 return map
->map_state
;
5754 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
5758 struct imsm_disk
*disk
;
5759 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5760 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
5764 /* at the beginning of migration we set IMSM_ORD_REBUILD on
5765 * disks that are being rebuilt. New failures are recorded to
5766 * map[0]. So we look through all the disks we started with and
5767 * see if any failures are still present, or if any new ones
5770 * FIXME add support for online capacity expansion and
5771 * raid-level-migration
5773 for (i
= 0; i
< prev
->num_members
; i
++) {
5774 ord
= __le32_to_cpu(prev
->disk_ord_tbl
[i
]);
5775 ord
|= __le32_to_cpu(map
->disk_ord_tbl
[i
]);
5776 idx
= ord_to_idx(ord
);
5778 disk
= get_imsm_disk(super
, idx
);
5779 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5787 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
5790 struct intel_super
*super
= c
->sb
;
5791 struct imsm_super
*mpb
= super
->anchor
;
5793 if (atoi(inst
) >= mpb
->num_raid_devs
) {
5794 fprintf(stderr
, "%s: subarry index %d, out of range\n",
5795 __func__
, atoi(inst
));
5799 dprintf("imsm: open_new %s\n", inst
);
5800 a
->info
.container_member
= atoi(inst
);
5804 static int is_resyncing(struct imsm_dev
*dev
)
5806 struct imsm_map
*migr_map
;
5808 if (!dev
->vol
.migr_state
)
5811 if (migr_type(dev
) == MIGR_INIT
||
5812 migr_type(dev
) == MIGR_REPAIR
)
5815 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5818 migr_map
= get_imsm_map(dev
, 1);
5820 if ((migr_map
->map_state
== IMSM_T_STATE_NORMAL
) &&
5821 (dev
->vol
.migr_type
!= MIGR_GEN_MIGR
))
5827 /* return true if we recorded new information */
5828 static int mark_failure(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5832 struct imsm_map
*map
;
5834 /* new failures are always set in map[0] */
5835 map
= get_imsm_map(dev
, 0);
5837 slot
= get_imsm_disk_slot(map
, idx
);
5841 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
5842 if (is_failed(disk
) && (ord
& IMSM_ORD_REBUILD
))
5845 disk
->status
|= FAILED_DISK
;
5846 set_imsm_ord_tbl_ent(map
, slot
, idx
| IMSM_ORD_REBUILD
);
5847 if (map
->failed_disk_num
== 0xff)
5848 map
->failed_disk_num
= slot
;
5852 static void mark_missing(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5854 mark_failure(dev
, disk
, idx
);
5856 if (disk
->scsi_id
== __cpu_to_le32(~(__u32
)0))
5859 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
5860 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
5863 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
)
5869 if (!super
->missing
)
5871 failed
= imsm_count_failed(super
, dev
);
5872 map_state
= imsm_check_degraded(super
, dev
, failed
);
5874 dprintf("imsm: mark missing\n");
5875 end_migration(dev
, map_state
);
5876 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
5877 mark_missing(dev
, &dl
->disk
, dl
->index
);
5878 super
->updates_pending
++;
5881 static unsigned long long imsm_set_array_size(struct imsm_dev
*dev
)
5883 int used_disks
= imsm_num_data_members(dev
, 0);
5884 unsigned long long array_blocks
;
5885 struct imsm_map
*map
;
5887 if (used_disks
== 0) {
5888 /* when problems occures
5889 * return current array_blocks value
5891 array_blocks
= __le32_to_cpu(dev
->size_high
);
5892 array_blocks
= array_blocks
<< 32;
5893 array_blocks
+= __le32_to_cpu(dev
->size_low
);
5895 return array_blocks
;
5898 /* set array size in metadata
5900 map
= get_imsm_map(dev
, 0);
5901 array_blocks
= map
->blocks_per_member
* used_disks
;
5903 /* round array size down to closest MB
5905 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
5906 dev
->size_low
= __cpu_to_le32((__u32
)array_blocks
);
5907 dev
->size_high
= __cpu_to_le32((__u32
)(array_blocks
>> 32));
5909 return array_blocks
;
5912 static void imsm_set_disk(struct active_array
*a
, int n
, int state
);
5914 static void imsm_progress_container_reshape(struct intel_super
*super
)
5916 /* if no device has a migr_state, but some device has a
5917 * different number of members than the previous device, start
5918 * changing the number of devices in this device to match
5921 struct imsm_super
*mpb
= super
->anchor
;
5922 int prev_disks
= -1;
5926 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5927 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
5928 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5929 struct imsm_map
*map2
;
5930 int prev_num_members
;
5932 if (dev
->vol
.migr_state
)
5935 if (prev_disks
== -1)
5936 prev_disks
= map
->num_members
;
5937 if (prev_disks
== map
->num_members
)
5940 /* OK, this array needs to enter reshape mode.
5941 * i.e it needs a migr_state
5944 copy_map_size
= sizeof_imsm_map(map
);
5945 prev_num_members
= map
->num_members
;
5946 map
->num_members
= prev_disks
;
5947 dev
->vol
.migr_state
= 1;
5948 dev
->vol
.curr_migr_unit
= 0;
5949 dev
->vol
.migr_type
= MIGR_GEN_MIGR
;
5950 for (i
= prev_num_members
;
5951 i
< map
->num_members
; i
++)
5952 set_imsm_ord_tbl_ent(map
, i
, i
);
5953 map2
= get_imsm_map(dev
, 1);
5954 /* Copy the current map */
5955 memcpy(map2
, map
, copy_map_size
);
5956 map2
->num_members
= prev_num_members
;
5958 imsm_set_array_size(dev
);
5959 super
->updates_pending
++;
5963 /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
5964 * states are handled in imsm_set_disk() with one exception, when a
5965 * resync is stopped due to a new failure this routine will set the
5966 * 'degraded' state for the array.
5968 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
5970 int inst
= a
->info
.container_member
;
5971 struct intel_super
*super
= a
->container
->sb
;
5972 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
5973 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5974 int failed
= imsm_count_failed(super
, dev
);
5975 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
5976 __u32 blocks_per_unit
;
5978 if (dev
->vol
.migr_state
&&
5979 dev
->vol
.migr_type
== MIGR_GEN_MIGR
) {
5980 /* array state change is blocked due to reshape action
5982 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
5983 * - finish the reshape (if last_checkpoint is big and action != reshape)
5984 * - update curr_migr_unit
5986 if (a
->curr_action
== reshape
) {
5987 /* still reshaping, maybe update curr_migr_unit */
5988 goto mark_checkpoint
;
5990 if (a
->last_checkpoint
== 0 && a
->prev_action
== reshape
) {
5991 /* for some reason we aborted the reshape.
5993 * disable automatic metadata rollback
5994 * user action is required to recover process
5997 struct imsm_map
*map2
= get_imsm_map(dev
, 1);
5998 dev
->vol
.migr_state
= 0;
5999 dev
->vol
.migr_type
= 0;
6000 dev
->vol
.curr_migr_unit
= 0;
6001 memcpy(map
, map2
, sizeof_imsm_map(map2
));
6002 super
->updates_pending
++;
6005 if (a
->last_checkpoint
>= a
->info
.component_size
) {
6006 unsigned long long array_blocks
;
6010 used_disks
= imsm_num_data_members(dev
, 0);
6011 if (used_disks
> 0) {
6013 map
->blocks_per_member
*
6015 /* round array size down to closest MB
6017 array_blocks
= (array_blocks
6018 >> SECT_PER_MB_SHIFT
)
6019 << SECT_PER_MB_SHIFT
;
6020 a
->info
.custom_array_size
= array_blocks
;
6021 /* encourage manager to update array
6025 a
->check_reshape
= 1;
6027 /* finalize online capacity expansion/reshape */
6028 for (mdi
= a
->info
.devs
; mdi
; mdi
= mdi
->next
)
6030 mdi
->disk
.raid_disk
,
6033 imsm_progress_container_reshape(super
);
6038 /* before we activate this array handle any missing disks */
6039 if (consistent
== 2)
6040 handle_missing(super
, dev
);
6042 if (consistent
== 2 &&
6043 (!is_resync_complete(&a
->info
) ||
6044 map_state
!= IMSM_T_STATE_NORMAL
||
6045 dev
->vol
.migr_state
))
6048 if (is_resync_complete(&a
->info
)) {
6049 /* complete intialization / resync,
6050 * recovery and interrupted recovery is completed in
6053 if (is_resyncing(dev
)) {
6054 dprintf("imsm: mark resync done\n");
6055 end_migration(dev
, map_state
);
6056 super
->updates_pending
++;
6057 a
->last_checkpoint
= 0;
6059 } else if (!is_resyncing(dev
) && !failed
) {
6060 /* mark the start of the init process if nothing is failed */
6061 dprintf("imsm: mark resync start\n");
6062 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
)
6063 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_INIT
);
6065 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_REPAIR
);
6066 super
->updates_pending
++;
6070 /* skip checkpointing for general migration,
6071 * it is controlled in mdadm
6073 if (is_gen_migration(dev
))
6074 goto skip_mark_checkpoint
;
6076 /* check if we can update curr_migr_unit from resync_start, recovery_start */
6077 blocks_per_unit
= blocks_per_migr_unit(super
, dev
);
6078 if (blocks_per_unit
) {
6082 units
= a
->last_checkpoint
/ blocks_per_unit
;
6085 /* check that we did not overflow 32-bits, and that
6086 * curr_migr_unit needs updating
6088 if (units32
== units
&&
6090 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != units32
) {
6091 dprintf("imsm: mark checkpoint (%u)\n", units32
);
6092 dev
->vol
.curr_migr_unit
= __cpu_to_le32(units32
);
6093 super
->updates_pending
++;
6097 skip_mark_checkpoint
:
6098 /* mark dirty / clean */
6099 if (dev
->vol
.dirty
!= !consistent
) {
6100 dprintf("imsm: mark '%s'\n", consistent
? "clean" : "dirty");
6105 super
->updates_pending
++;
6111 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
6113 int inst
= a
->info
.container_member
;
6114 struct intel_super
*super
= a
->container
->sb
;
6115 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6116 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6117 struct imsm_disk
*disk
;
6122 if (n
> map
->num_members
)
6123 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
6124 n
, map
->num_members
- 1);
6129 dprintf("imsm: set_disk %d:%x\n", n
, state
);
6131 ord
= get_imsm_ord_tbl_ent(dev
, n
, -1);
6132 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
6134 /* check for new failures */
6135 if (state
& DS_FAULTY
) {
6136 if (mark_failure(dev
, disk
, ord_to_idx(ord
)))
6137 super
->updates_pending
++;
6140 /* check if in_sync */
6141 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
&& is_rebuilding(dev
)) {
6142 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6144 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
6145 super
->updates_pending
++;
6148 failed
= imsm_count_failed(super
, dev
);
6149 map_state
= imsm_check_degraded(super
, dev
, failed
);
6151 /* check if recovery complete, newly degraded, or failed */
6152 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
6153 end_migration(dev
, map_state
);
6154 map
= get_imsm_map(dev
, 0);
6155 map
->failed_disk_num
= ~0;
6156 super
->updates_pending
++;
6157 a
->last_checkpoint
= 0;
6158 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
6159 map
->map_state
!= map_state
&&
6160 !dev
->vol
.migr_state
) {
6161 dprintf("imsm: mark degraded\n");
6162 map
->map_state
= map_state
;
6163 super
->updates_pending
++;
6164 a
->last_checkpoint
= 0;
6165 } else if (map_state
== IMSM_T_STATE_FAILED
&&
6166 map
->map_state
!= map_state
) {
6167 dprintf("imsm: mark failed\n");
6168 end_migration(dev
, map_state
);
6169 super
->updates_pending
++;
6170 a
->last_checkpoint
= 0;
6171 } else if (is_gen_migration(dev
)) {
6172 dprintf("imsm: Detected General Migration in state: ");
6173 if (map_state
== IMSM_T_STATE_NORMAL
) {
6174 end_migration(dev
, map_state
);
6175 map
= get_imsm_map(dev
, 0);
6176 map
->failed_disk_num
= ~0;
6177 dprintf("normal\n");
6179 if (map_state
== IMSM_T_STATE_DEGRADED
) {
6180 printf("degraded\n");
6181 end_migration(dev
, map_state
);
6183 dprintf("failed\n");
6185 map
->map_state
= map_state
;
6187 super
->updates_pending
++;
6191 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
)
6194 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
6195 unsigned long long dsize
;
6196 unsigned long long sectors
;
6198 get_dev_size(fd
, NULL
, &dsize
);
6200 if (mpb_size
> 512) {
6201 /* -1 to account for anchor */
6202 sectors
= mpb_sectors(mpb
) - 1;
6204 /* write the extended mpb to the sectors preceeding the anchor */
6205 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
6208 if ((unsigned long long)write(fd
, buf
+ 512, 512 * sectors
)
6213 /* first block is stored on second to last sector of the disk */
6214 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
6217 if (write(fd
, buf
, 512) != 512)
6223 static void imsm_sync_metadata(struct supertype
*container
)
6225 struct intel_super
*super
= container
->sb
;
6227 dprintf("sync metadata: %d\n", super
->updates_pending
);
6228 if (!super
->updates_pending
)
6231 write_super_imsm(container
, 0);
6233 super
->updates_pending
= 0;
6236 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
6238 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6239 int i
= get_imsm_disk_idx(dev
, idx
, -1);
6242 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6246 if (dl
&& is_failed(&dl
->disk
))
6250 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
6255 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
,
6256 struct active_array
*a
, int activate_new
,
6257 struct mdinfo
*additional_test_list
)
6259 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6260 int idx
= get_imsm_disk_idx(dev
, slot
, -1);
6261 struct imsm_super
*mpb
= super
->anchor
;
6262 struct imsm_map
*map
;
6263 unsigned long long pos
;
6268 __u32 array_start
= 0;
6269 __u32 array_end
= 0;
6271 struct mdinfo
*test_list
;
6273 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6274 /* If in this array, skip */
6275 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6276 if (d
->state_fd
>= 0 &&
6277 d
->disk
.major
== dl
->major
&&
6278 d
->disk
.minor
== dl
->minor
) {
6279 dprintf("%x:%x already in array\n",
6280 dl
->major
, dl
->minor
);
6285 test_list
= additional_test_list
;
6287 if (test_list
->disk
.major
== dl
->major
&&
6288 test_list
->disk
.minor
== dl
->minor
) {
6289 dprintf("%x:%x already in additional test list\n",
6290 dl
->major
, dl
->minor
);
6293 test_list
= test_list
->next
;
6298 /* skip in use or failed drives */
6299 if (is_failed(&dl
->disk
) || idx
== dl
->index
||
6301 dprintf("%x:%x status (failed: %d index: %d)\n",
6302 dl
->major
, dl
->minor
, is_failed(&dl
->disk
), idx
);
6306 /* skip pure spares when we are looking for partially
6307 * assimilated drives
6309 if (dl
->index
== -1 && !activate_new
)
6312 /* Does this unused device have the requisite free space?
6313 * It needs to be able to cover all member volumes
6315 ex
= get_extents(super
, dl
);
6317 dprintf("cannot get extents\n");
6320 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
6321 dev
= get_imsm_dev(super
, i
);
6322 map
= get_imsm_map(dev
, 0);
6324 /* check if this disk is already a member of
6327 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
6333 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
6334 array_end
= array_start
+
6335 __le32_to_cpu(map
->blocks_per_member
) - 1;
6338 /* check that we can start at pba_of_lba0 with
6339 * blocks_per_member of space
6341 if (array_start
>= pos
&& array_end
< ex
[j
].start
) {
6345 pos
= ex
[j
].start
+ ex
[j
].size
;
6347 } while (ex
[j
-1].size
);
6354 if (i
< mpb
->num_raid_devs
) {
6355 dprintf("%x:%x does not have %u to %u available\n",
6356 dl
->major
, dl
->minor
, array_start
, array_end
);
6367 static int imsm_rebuild_allowed(struct supertype
*cont
, int dev_idx
, int failed
)
6369 struct imsm_dev
*dev2
;
6370 struct imsm_map
*map
;
6376 dev2
= get_imsm_dev(cont
->sb
, dev_idx
);
6378 state
= imsm_check_degraded(cont
->sb
, dev2
, failed
);
6379 if (state
== IMSM_T_STATE_FAILED
) {
6380 map
= get_imsm_map(dev2
, 0);
6383 for (slot
= 0; slot
< map
->num_members
; slot
++) {
6385 * Check if failed disks are deleted from intel
6386 * disk list or are marked to be deleted
6388 idx
= get_imsm_disk_idx(dev2
, slot
, -1);
6389 idisk
= get_imsm_dl_disk(cont
->sb
, idx
);
6391 * Do not rebuild the array if failed disks
6392 * from failed sub-array are not removed from
6396 is_failed(&idisk
->disk
) &&
6397 (idisk
->action
!= DISK_REMOVE
))
6405 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
6406 struct metadata_update
**updates
)
6409 * Find a device with unused free space and use it to replace a
6410 * failed/vacant region in an array. We replace failed regions one a
6411 * array at a time. The result is that a new spare disk will be added
6412 * to the first failed array and after the monitor has finished
6413 * propagating failures the remainder will be consumed.
6415 * FIXME add a capability for mdmon to request spares from another
6419 struct intel_super
*super
= a
->container
->sb
;
6420 int inst
= a
->info
.container_member
;
6421 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6422 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6423 int failed
= a
->info
.array
.raid_disks
;
6424 struct mdinfo
*rv
= NULL
;
6427 struct metadata_update
*mu
;
6429 struct imsm_update_activate_spare
*u
;
6434 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
6435 if ((d
->curr_state
& DS_FAULTY
) &&
6437 /* wait for Removal to happen */
6439 if (d
->state_fd
>= 0)
6443 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
6444 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
6446 if (dev
->vol
.migr_state
&&
6447 dev
->vol
.migr_type
== MIGR_GEN_MIGR
)
6448 /* No repair during migration */
6451 if (a
->info
.array
.level
== 4)
6452 /* No repair for takeovered array
6453 * imsm doesn't support raid4
6457 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
6461 * If there are any failed disks check state of the other volume.
6462 * Block rebuild if the another one is failed until failed disks
6463 * are removed from container.
6466 dprintf("found failed disks in %s, check if there another"
6467 "failed sub-array.\n",
6469 /* check if states of the other volumes allow for rebuild */
6470 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
6472 allowed
= imsm_rebuild_allowed(a
->container
,
6480 /* For each slot, if it is not working, find a spare */
6481 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
6482 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6483 if (d
->disk
.raid_disk
== i
)
6485 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
6486 if (d
&& (d
->state_fd
>= 0))
6490 * OK, this device needs recovery. Try to re-add the
6491 * previous occupant of this slot, if this fails see if
6492 * we can continue the assimilation of a spare that was
6493 * partially assimilated, finally try to activate a new
6496 dl
= imsm_readd(super
, i
, a
);
6498 dl
= imsm_add_spare(super
, i
, a
, 0, NULL
);
6500 dl
= imsm_add_spare(super
, i
, a
, 1, NULL
);
6504 /* found a usable disk with enough space */
6505 di
= malloc(sizeof(*di
));
6508 memset(di
, 0, sizeof(*di
));
6510 /* dl->index will be -1 in the case we are activating a
6511 * pristine spare. imsm_process_update() will create a
6512 * new index in this case. Once a disk is found to be
6513 * failed in all member arrays it is kicked from the
6516 di
->disk
.number
= dl
->index
;
6518 /* (ab)use di->devs to store a pointer to the device
6521 di
->devs
= (struct mdinfo
*) dl
;
6523 di
->disk
.raid_disk
= i
;
6524 di
->disk
.major
= dl
->major
;
6525 di
->disk
.minor
= dl
->minor
;
6527 di
->recovery_start
= 0;
6528 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
6529 di
->component_size
= a
->info
.component_size
;
6530 di
->container_member
= inst
;
6531 super
->random
= random32();
6535 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
6536 i
, di
->data_offset
);
6542 /* No spares found */
6544 /* Now 'rv' has a list of devices to return.
6545 * Create a metadata_update record to update the
6546 * disk_ord_tbl for the array
6548 mu
= malloc(sizeof(*mu
));
6550 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
6551 if (mu
->buf
== NULL
) {
6558 struct mdinfo
*n
= rv
->next
;
6567 mu
->space_list
= NULL
;
6568 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
6569 mu
->next
= *updates
;
6570 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
6572 for (di
= rv
; di
; di
= di
->next
) {
6573 u
->type
= update_activate_spare
;
6574 u
->dl
= (struct dl
*) di
->devs
;
6576 u
->slot
= di
->disk
.raid_disk
;
6587 static int disks_overlap(struct intel_super
*super
, int idx
, struct imsm_update_create_array
*u
)
6589 struct imsm_dev
*dev
= get_imsm_dev(super
, idx
);
6590 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6591 struct imsm_map
*new_map
= get_imsm_map(&u
->dev
, 0);
6592 struct disk_info
*inf
= get_disk_info(u
);
6593 struct imsm_disk
*disk
;
6597 for (i
= 0; i
< map
->num_members
; i
++) {
6598 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
, -1));
6599 for (j
= 0; j
< new_map
->num_members
; j
++)
6600 if (serialcmp(disk
->serial
, inf
[j
].serial
) == 0)
6608 static struct dl
*get_disk_super(struct intel_super
*super
, int major
, int minor
)
6610 struct dl
*dl
= NULL
;
6611 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6612 if ((dl
->major
== major
) && (dl
->minor
== minor
))
6617 static int remove_disk_super(struct intel_super
*super
, int major
, int minor
)
6619 struct dl
*prev
= NULL
;
6623 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6624 if ((dl
->major
== major
) && (dl
->minor
== minor
)) {
6627 prev
->next
= dl
->next
;
6629 super
->disks
= dl
->next
;
6631 __free_imsm_disk(dl
);
6632 dprintf("%s: removed %x:%x\n",
6633 __func__
, major
, minor
);
6641 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
);
6643 static int add_remove_disk_update(struct intel_super
*super
)
6645 int check_degraded
= 0;
6646 struct dl
*disk
= NULL
;
6647 /* add/remove some spares to/from the metadata/contrainer */
6648 while (super
->disk_mgmt_list
) {
6649 struct dl
*disk_cfg
;
6651 disk_cfg
= super
->disk_mgmt_list
;
6652 super
->disk_mgmt_list
= disk_cfg
->next
;
6653 disk_cfg
->next
= NULL
;
6655 if (disk_cfg
->action
== DISK_ADD
) {
6656 disk_cfg
->next
= super
->disks
;
6657 super
->disks
= disk_cfg
;
6659 dprintf("%s: added %x:%x\n",
6660 __func__
, disk_cfg
->major
,
6662 } else if (disk_cfg
->action
== DISK_REMOVE
) {
6663 dprintf("Disk remove action processed: %x.%x\n",
6664 disk_cfg
->major
, disk_cfg
->minor
);
6665 disk
= get_disk_super(super
,
6669 /* store action status */
6670 disk
->action
= DISK_REMOVE
;
6671 /* remove spare disks only */
6672 if (disk
->index
== -1) {
6673 remove_disk_super(super
,
6678 /* release allocate disk structure */
6679 __free_imsm_disk(disk_cfg
);
6682 return check_degraded
;
6686 static int apply_reshape_migration_update(struct imsm_update_reshape_migration
*u
,
6687 struct intel_super
*super
,
6690 struct intel_dev
*id
;
6691 void **tofree
= NULL
;
6694 dprintf("apply_reshape_migration_update()\n");
6695 if ((u
->subdev
< 0) ||
6697 dprintf("imsm: Error: Wrong subdev: %i\n", u
->subdev
);
6700 if ((space_list
== NULL
) || (*space_list
== NULL
)) {
6701 dprintf("imsm: Error: Memory is not allocated\n");
6705 for (id
= super
->devlist
; id
; id
= id
->next
) {
6706 if (id
->index
== (unsigned)u
->subdev
) {
6707 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subdev
);
6708 struct imsm_map
*map
;
6709 struct imsm_dev
*new_dev
=
6710 (struct imsm_dev
*)*space_list
;
6711 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6713 struct dl
*new_disk
;
6715 if (new_dev
== NULL
)
6717 *space_list
= **space_list
;
6718 memcpy(new_dev
, dev
, sizeof_imsm_dev(dev
, 0));
6719 map
= get_imsm_map(new_dev
, 0);
6721 dprintf("imsm: Error: migration in progress");
6725 to_state
= map
->map_state
;
6726 if ((u
->new_level
== 5) && (map
->raid_level
== 0)) {
6728 /* this should not happen */
6729 if (u
->new_disks
[0] < 0) {
6730 map
->failed_disk_num
=
6731 map
->num_members
- 1;
6732 to_state
= IMSM_T_STATE_DEGRADED
;
6734 to_state
= IMSM_T_STATE_NORMAL
;
6736 migrate(new_dev
, super
, to_state
, MIGR_GEN_MIGR
);
6737 if (u
->new_level
> -1)
6738 map
->raid_level
= u
->new_level
;
6739 migr_map
= get_imsm_map(new_dev
, 1);
6740 if ((u
->new_level
== 5) &&
6741 (migr_map
->raid_level
== 0)) {
6742 int ord
= map
->num_members
- 1;
6743 migr_map
->num_members
--;
6744 if (u
->new_disks
[0] < 0)
6745 ord
|= IMSM_ORD_REBUILD
;
6746 set_imsm_ord_tbl_ent(map
,
6747 map
->num_members
- 1,
6751 tofree
= (void **)dev
;
6753 /* update chunk size
6755 if (u
->new_chunksize
> 0)
6756 map
->blocks_per_strip
=
6757 __cpu_to_le16(u
->new_chunksize
* 2);
6761 if ((u
->new_level
!= 5) ||
6762 (migr_map
->raid_level
!= 0) ||
6763 (migr_map
->raid_level
== map
->raid_level
))
6766 if (u
->new_disks
[0] >= 0) {
6769 new_disk
= get_disk_super(super
,
6770 major(u
->new_disks
[0]),
6771 minor(u
->new_disks
[0]));
6772 dprintf("imsm: new disk for reshape is: %i:%i "
6773 "(%p, index = %i)\n",
6774 major(u
->new_disks
[0]),
6775 minor(u
->new_disks
[0]),
6776 new_disk
, new_disk
->index
);
6777 if (new_disk
== NULL
)
6778 goto error_disk_add
;
6780 new_disk
->index
= map
->num_members
- 1;
6781 /* slot to fill in autolayout
6783 new_disk
->raiddisk
= new_disk
->index
;
6784 new_disk
->disk
.status
|= CONFIGURED_DISK
;
6785 new_disk
->disk
.status
&= ~SPARE_DISK
;
6787 goto error_disk_add
;
6790 *tofree
= *space_list
;
6791 /* calculate new size
6793 imsm_set_array_size(new_dev
);
6800 *space_list
= tofree
;
6804 dprintf("Error: imsm: Cannot find disk.\n");
6809 static int apply_reshape_container_disks_update(struct imsm_update_reshape
*u
,
6810 struct intel_super
*super
,
6813 struct dl
*new_disk
;
6814 struct intel_dev
*id
;
6816 int delta_disks
= u
->new_raid_disks
- u
->old_raid_disks
;
6817 int disk_count
= u
->old_raid_disks
;
6818 void **tofree
= NULL
;
6819 int devices_to_reshape
= 1;
6820 struct imsm_super
*mpb
= super
->anchor
;
6822 unsigned int dev_id
;
6824 dprintf("imsm: apply_reshape_container_disks_update()\n");
6826 /* enable spares to use in array */
6827 for (i
= 0; i
< delta_disks
; i
++) {
6828 new_disk
= get_disk_super(super
,
6829 major(u
->new_disks
[i
]),
6830 minor(u
->new_disks
[i
]));
6831 dprintf("imsm: new disk for reshape is: %i:%i "
6832 "(%p, index = %i)\n",
6833 major(u
->new_disks
[i
]), minor(u
->new_disks
[i
]),
6834 new_disk
, new_disk
->index
);
6835 if ((new_disk
== NULL
) ||
6836 ((new_disk
->index
>= 0) &&
6837 (new_disk
->index
< u
->old_raid_disks
)))
6838 goto update_reshape_exit
;
6839 new_disk
->index
= disk_count
++;
6840 /* slot to fill in autolayout
6842 new_disk
->raiddisk
= new_disk
->index
;
6843 new_disk
->disk
.status
|=
6845 new_disk
->disk
.status
&= ~SPARE_DISK
;
6848 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
6849 mpb
->num_raid_devs
);
6850 /* manage changes in volume
6852 for (dev_id
= 0; dev_id
< mpb
->num_raid_devs
; dev_id
++) {
6853 void **sp
= *space_list
;
6854 struct imsm_dev
*newdev
;
6855 struct imsm_map
*newmap
, *oldmap
;
6857 for (id
= super
->devlist
; id
; id
= id
->next
) {
6858 if (id
->index
== dev_id
)
6867 /* Copy the dev, but not (all of) the map */
6868 memcpy(newdev
, id
->dev
, sizeof(*newdev
));
6869 oldmap
= get_imsm_map(id
->dev
, 0);
6870 newmap
= get_imsm_map(newdev
, 0);
6871 /* Copy the current map */
6872 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6873 /* update one device only
6875 if (devices_to_reshape
) {
6876 dprintf("imsm: modifying subdev: %i\n",
6878 devices_to_reshape
--;
6879 newdev
->vol
.migr_state
= 1;
6880 newdev
->vol
.curr_migr_unit
= 0;
6881 newdev
->vol
.migr_type
= MIGR_GEN_MIGR
;
6882 newmap
->num_members
= u
->new_raid_disks
;
6883 for (i
= 0; i
< delta_disks
; i
++) {
6884 set_imsm_ord_tbl_ent(newmap
,
6885 u
->old_raid_disks
+ i
,
6886 u
->old_raid_disks
+ i
);
6888 /* New map is correct, now need to save old map
6890 newmap
= get_imsm_map(newdev
, 1);
6891 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6893 imsm_set_array_size(newdev
);
6896 sp
= (void **)id
->dev
;
6901 /* Clear migration record */
6902 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
6905 *space_list
= tofree
;
6908 update_reshape_exit
:
6913 static int apply_takeover_update(struct imsm_update_takeover
*u
,
6914 struct intel_super
*super
,
6917 struct imsm_dev
*dev
= NULL
;
6918 struct intel_dev
*dv
;
6919 struct imsm_dev
*dev_new
;
6920 struct imsm_map
*map
;
6924 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
6925 if (dv
->index
== (unsigned int)u
->subarray
) {
6933 map
= get_imsm_map(dev
, 0);
6935 if (u
->direction
== R10_TO_R0
) {
6936 /* Number of failed disks must be half of initial disk number */
6937 if (imsm_count_failed(super
, dev
) != (map
->num_members
/ 2))
6940 /* iterate through devices to mark removed disks as spare */
6941 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6942 if (dm
->disk
.status
& FAILED_DISK
) {
6943 int idx
= dm
->index
;
6944 /* update indexes on the disk list */
6945 /* FIXME this loop-with-the-loop looks wrong, I'm not convinced
6946 the index values will end up being correct.... NB */
6947 for (du
= super
->disks
; du
; du
= du
->next
)
6948 if (du
->index
> idx
)
6950 /* mark as spare disk */
6951 dm
->disk
.status
= SPARE_DISK
;
6956 map
->num_members
= map
->num_members
/ 2;
6957 map
->map_state
= IMSM_T_STATE_NORMAL
;
6958 map
->num_domains
= 1;
6959 map
->raid_level
= 0;
6960 map
->failed_disk_num
= -1;
6963 if (u
->direction
== R0_TO_R10
) {
6965 /* update slots in current disk list */
6966 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6970 /* create new *missing* disks */
6971 for (i
= 0; i
< map
->num_members
; i
++) {
6972 space
= *space_list
;
6975 *space_list
= *space
;
6977 memcpy(du
, super
->disks
, sizeof(*du
));
6981 du
->index
= (i
* 2) + 1;
6982 sprintf((char *)du
->disk
.serial
,
6983 " MISSING_%d", du
->index
);
6984 sprintf((char *)du
->serial
,
6985 "MISSING_%d", du
->index
);
6986 du
->next
= super
->missing
;
6987 super
->missing
= du
;
6989 /* create new dev and map */
6990 space
= *space_list
;
6993 *space_list
= *space
;
6994 dev_new
= (void *)space
;
6995 memcpy(dev_new
, dev
, sizeof(*dev
));
6996 /* update new map */
6997 map
= get_imsm_map(dev_new
, 0);
6998 map
->num_members
= map
->num_members
* 2;
6999 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7000 map
->num_domains
= 2;
7001 map
->raid_level
= 1;
7002 /* replace dev<->dev_new */
7005 /* update disk order table */
7006 for (du
= super
->disks
; du
; du
= du
->next
)
7008 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7009 for (du
= super
->missing
; du
; du
= du
->next
)
7010 if (du
->index
>= 0) {
7011 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7012 mark_missing(dv
->dev
, &du
->disk
, du
->index
);
7018 static void imsm_process_update(struct supertype
*st
,
7019 struct metadata_update
*update
)
7022 * crack open the metadata_update envelope to find the update record
7023 * update can be one of:
7024 * update_reshape_container_disks - all the arrays in the container
7025 * are being reshaped to have more devices. We need to mark
7026 * the arrays for general migration and convert selected spares
7027 * into active devices.
7028 * update_activate_spare - a spare device has replaced a failed
7029 * device in an array, update the disk_ord_tbl. If this disk is
7030 * present in all member arrays then also clear the SPARE_DISK
7032 * update_create_array
7034 * update_rename_array
7035 * update_add_remove_disk
7037 struct intel_super
*super
= st
->sb
;
7038 struct imsm_super
*mpb
;
7039 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7041 /* update requires a larger buf but the allocation failed */
7042 if (super
->next_len
&& !super
->next_buf
) {
7043 super
->next_len
= 0;
7047 if (super
->next_buf
) {
7048 memcpy(super
->next_buf
, super
->buf
, super
->len
);
7050 super
->len
= super
->next_len
;
7051 super
->buf
= super
->next_buf
;
7053 super
->next_len
= 0;
7054 super
->next_buf
= NULL
;
7057 mpb
= super
->anchor
;
7060 case update_general_migration_checkpoint
: {
7061 struct intel_dev
*id
;
7062 struct imsm_update_general_migration_checkpoint
*u
=
7063 (void *)update
->buf
;
7065 dprintf("imsm: process_update() "
7066 "for update_general_migration_checkpoint called\n");
7068 /* find device under general migration */
7069 for (id
= super
->devlist
; id
; id
= id
->next
) {
7070 if (is_gen_migration(id
->dev
)) {
7071 id
->dev
->vol
.curr_migr_unit
=
7072 __cpu_to_le32(u
->curr_migr_unit
);
7073 super
->updates_pending
++;
7078 case update_takeover
: {
7079 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7080 if (apply_takeover_update(u
, super
, &update
->space_list
)) {
7081 imsm_update_version_info(super
);
7082 super
->updates_pending
++;
7087 case update_reshape_container_disks
: {
7088 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7089 if (apply_reshape_container_disks_update(
7090 u
, super
, &update
->space_list
))
7091 super
->updates_pending
++;
7094 case update_reshape_migration
: {
7095 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7096 if (apply_reshape_migration_update(
7097 u
, super
, &update
->space_list
))
7098 super
->updates_pending
++;
7101 case update_activate_spare
: {
7102 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
7103 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
7104 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7105 struct imsm_map
*migr_map
;
7106 struct active_array
*a
;
7107 struct imsm_disk
*disk
;
7112 int victim
= get_imsm_disk_idx(dev
, u
->slot
, -1);
7115 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
7120 fprintf(stderr
, "error: imsm_activate_spare passed "
7121 "an unknown disk (index: %d)\n",
7126 super
->updates_pending
++;
7127 /* count failures (excluding rebuilds and the victim)
7128 * to determine map[0] state
7131 for (i
= 0; i
< map
->num_members
; i
++) {
7134 disk
= get_imsm_disk(super
,
7135 get_imsm_disk_idx(dev
, i
, -1));
7136 if (!disk
|| is_failed(disk
))
7140 /* adding a pristine spare, assign a new index */
7141 if (dl
->index
< 0) {
7142 dl
->index
= super
->anchor
->num_disks
;
7143 super
->anchor
->num_disks
++;
7146 disk
->status
|= CONFIGURED_DISK
;
7147 disk
->status
&= ~SPARE_DISK
;
7150 to_state
= imsm_check_degraded(super
, dev
, failed
);
7151 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7152 migrate(dev
, super
, to_state
, MIGR_REBUILD
);
7153 migr_map
= get_imsm_map(dev
, 1);
7154 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
7155 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
7157 /* update the family_num to mark a new container
7158 * generation, being careful to record the existing
7159 * family_num in orig_family_num to clean up after
7160 * earlier mdadm versions that neglected to set it.
7162 if (mpb
->orig_family_num
== 0)
7163 mpb
->orig_family_num
= mpb
->family_num
;
7164 mpb
->family_num
+= super
->random
;
7166 /* count arrays using the victim in the metadata */
7168 for (a
= st
->arrays
; a
; a
= a
->next
) {
7169 dev
= get_imsm_dev(super
, a
->info
.container_member
);
7170 map
= get_imsm_map(dev
, 0);
7172 if (get_imsm_disk_slot(map
, victim
) >= 0)
7176 /* delete the victim if it is no longer being
7182 /* We know that 'manager' isn't touching anything,
7183 * so it is safe to delete
7185 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
7186 if ((*dlp
)->index
== victim
)
7189 /* victim may be on the missing list */
7191 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
7192 if ((*dlp
)->index
== victim
)
7194 imsm_delete(super
, dlp
, victim
);
7198 case update_create_array
: {
7199 /* someone wants to create a new array, we need to be aware of
7200 * a few races/collisions:
7201 * 1/ 'Create' called by two separate instances of mdadm
7202 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
7203 * devices that have since been assimilated via
7205 * In the event this update can not be carried out mdadm will
7206 * (FIX ME) notice that its update did not take hold.
7208 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7209 struct intel_dev
*dv
;
7210 struct imsm_dev
*dev
;
7211 struct imsm_map
*map
, *new_map
;
7212 unsigned long long start
, end
;
7213 unsigned long long new_start
, new_end
;
7215 struct disk_info
*inf
;
7218 /* handle racing creates: first come first serve */
7219 if (u
->dev_idx
< mpb
->num_raid_devs
) {
7220 dprintf("%s: subarray %d already defined\n",
7221 __func__
, u
->dev_idx
);
7225 /* check update is next in sequence */
7226 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
7227 dprintf("%s: can not create array %d expected index %d\n",
7228 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
7232 new_map
= get_imsm_map(&u
->dev
, 0);
7233 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
7234 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
7235 inf
= get_disk_info(u
);
7237 /* handle activate_spare versus create race:
7238 * check to make sure that overlapping arrays do not include
7241 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7242 dev
= get_imsm_dev(super
, i
);
7243 map
= get_imsm_map(dev
, 0);
7244 start
= __le32_to_cpu(map
->pba_of_lba0
);
7245 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
7246 if ((new_start
>= start
&& new_start
<= end
) ||
7247 (start
>= new_start
&& start
<= new_end
))
7252 if (disks_overlap(super
, i
, u
)) {
7253 dprintf("%s: arrays overlap\n", __func__
);
7258 /* check that prepare update was successful */
7259 if (!update
->space
) {
7260 dprintf("%s: prepare update failed\n", __func__
);
7264 /* check that all disks are still active before committing
7265 * changes. FIXME: could we instead handle this by creating a
7266 * degraded array? That's probably not what the user expects,
7267 * so better to drop this update on the floor.
7269 for (i
= 0; i
< new_map
->num_members
; i
++) {
7270 dl
= serial_to_dl(inf
[i
].serial
, super
);
7272 dprintf("%s: disk disappeared\n", __func__
);
7277 super
->updates_pending
++;
7279 /* convert spares to members and fixup ord_tbl */
7280 for (i
= 0; i
< new_map
->num_members
; i
++) {
7281 dl
= serial_to_dl(inf
[i
].serial
, super
);
7282 if (dl
->index
== -1) {
7283 dl
->index
= mpb
->num_disks
;
7285 dl
->disk
.status
|= CONFIGURED_DISK
;
7286 dl
->disk
.status
&= ~SPARE_DISK
;
7288 set_imsm_ord_tbl_ent(new_map
, i
, dl
->index
);
7293 update
->space
= NULL
;
7294 imsm_copy_dev(dev
, &u
->dev
);
7295 dv
->index
= u
->dev_idx
;
7296 dv
->next
= super
->devlist
;
7297 super
->devlist
= dv
;
7298 mpb
->num_raid_devs
++;
7300 imsm_update_version_info(super
);
7303 /* mdmon knows how to release update->space, but not
7304 * ((struct intel_dev *) update->space)->dev
7306 if (update
->space
) {
7312 case update_kill_array
: {
7313 struct imsm_update_kill_array
*u
= (void *) update
->buf
;
7314 int victim
= u
->dev_idx
;
7315 struct active_array
*a
;
7316 struct intel_dev
**dp
;
7317 struct imsm_dev
*dev
;
7319 /* sanity check that we are not affecting the uuid of
7320 * active arrays, or deleting an active array
7322 * FIXME when immutable ids are available, but note that
7323 * we'll also need to fixup the invalidated/active
7324 * subarray indexes in mdstat
7326 for (a
= st
->arrays
; a
; a
= a
->next
)
7327 if (a
->info
.container_member
>= victim
)
7329 /* by definition if mdmon is running at least one array
7330 * is active in the container, so checking
7331 * mpb->num_raid_devs is just extra paranoia
7333 dev
= get_imsm_dev(super
, victim
);
7334 if (a
|| !dev
|| mpb
->num_raid_devs
== 1) {
7335 dprintf("failed to delete subarray-%d\n", victim
);
7339 for (dp
= &super
->devlist
; *dp
;)
7340 if ((*dp
)->index
== (unsigned)super
->current_vol
) {
7343 if ((*dp
)->index
> (unsigned)victim
)
7347 mpb
->num_raid_devs
--;
7348 super
->updates_pending
++;
7351 case update_rename_array
: {
7352 struct imsm_update_rename_array
*u
= (void *) update
->buf
;
7353 char name
[MAX_RAID_SERIAL_LEN
+1];
7354 int target
= u
->dev_idx
;
7355 struct active_array
*a
;
7356 struct imsm_dev
*dev
;
7358 /* sanity check that we are not affecting the uuid of
7361 snprintf(name
, MAX_RAID_SERIAL_LEN
, "%s", (char *) u
->name
);
7362 name
[MAX_RAID_SERIAL_LEN
] = '\0';
7363 for (a
= st
->arrays
; a
; a
= a
->next
)
7364 if (a
->info
.container_member
== target
)
7366 dev
= get_imsm_dev(super
, u
->dev_idx
);
7367 if (a
|| !dev
|| !check_name(super
, name
, 1)) {
7368 dprintf("failed to rename subarray-%d\n", target
);
7372 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
7373 super
->updates_pending
++;
7376 case update_add_remove_disk
: {
7377 /* we may be able to repair some arrays if disks are
7378 * being added, check teh status of add_remove_disk
7379 * if discs has been added.
7381 if (add_remove_disk_update(super
)) {
7382 struct active_array
*a
;
7384 super
->updates_pending
++;
7385 for (a
= st
->arrays
; a
; a
= a
->next
)
7386 a
->check_degraded
= 1;
7391 fprintf(stderr
, "error: unsuported process update type:"
7392 "(type: %d)\n", type
);
7396 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
);
7398 static void imsm_prepare_update(struct supertype
*st
,
7399 struct metadata_update
*update
)
7402 * Allocate space to hold new disk entries, raid-device entries or a new
7403 * mpb if necessary. The manager synchronously waits for updates to
7404 * complete in the monitor, so new mpb buffers allocated here can be
7405 * integrated by the monitor thread without worrying about live pointers
7406 * in the manager thread.
7408 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7409 struct intel_super
*super
= st
->sb
;
7410 struct imsm_super
*mpb
= super
->anchor
;
7415 case update_general_migration_checkpoint
:
7416 dprintf("imsm: prepare_update() "
7417 "for update_general_migration_checkpoint called\n");
7419 case update_takeover
: {
7420 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7421 if (u
->direction
== R0_TO_R10
) {
7422 void **tail
= (void **)&update
->space_list
;
7423 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subarray
);
7424 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7425 int num_members
= map
->num_members
;
7429 /* allocate memory for added disks */
7430 for (i
= 0; i
< num_members
; i
++) {
7431 size
= sizeof(struct dl
);
7432 space
= malloc(size
);
7441 /* allocate memory for new device */
7442 size
= sizeof_imsm_dev(super
->devlist
->dev
, 0) +
7443 (num_members
* sizeof(__u32
));
7444 space
= malloc(size
);
7453 len
= disks_to_mpb_size(num_members
* 2);
7455 /* if allocation didn't success, free buffer */
7456 while (update
->space_list
) {
7457 void **sp
= update
->space_list
;
7458 update
->space_list
= *sp
;
7466 case update_reshape_container_disks
: {
7467 /* Every raid device in the container is about to
7468 * gain some more devices, and we will enter a
7470 * So each 'imsm_map' will be bigger, and the imsm_vol
7471 * will now hold 2 of them.
7472 * Thus we need new 'struct imsm_dev' allocations sized
7473 * as sizeof_imsm_dev but with more devices in both maps.
7475 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7476 struct intel_dev
*dl
;
7477 void **space_tail
= (void**)&update
->space_list
;
7479 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7481 for (dl
= super
->devlist
; dl
; dl
= dl
->next
) {
7482 int size
= sizeof_imsm_dev(dl
->dev
, 1);
7484 if (u
->new_raid_disks
> u
->old_raid_disks
)
7485 size
+= sizeof(__u32
)*2*
7486 (u
->new_raid_disks
- u
->old_raid_disks
);
7495 len
= disks_to_mpb_size(u
->new_raid_disks
);
7496 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7499 case update_reshape_migration
: {
7500 /* for migration level 0->5 we need to add disks
7501 * so the same as for container operation we will copy
7502 * device to the bigger location.
7503 * in memory prepared device and new disk area are prepared
7504 * for usage in process update
7506 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7507 struct intel_dev
*id
;
7508 void **space_tail
= (void **)&update
->space_list
;
7511 int current_level
= -1;
7513 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7515 /* add space for bigger array in update
7517 for (id
= super
->devlist
; id
; id
= id
->next
) {
7518 if (id
->index
== (unsigned)u
->subdev
) {
7519 size
= sizeof_imsm_dev(id
->dev
, 1);
7520 if (u
->new_raid_disks
> u
->old_raid_disks
)
7521 size
+= sizeof(__u32
)*2*
7522 (u
->new_raid_disks
- u
->old_raid_disks
);
7532 if (update
->space_list
== NULL
)
7535 /* add space for disk in update
7537 size
= sizeof(struct dl
);
7540 free(update
->space_list
);
7541 update
->space_list
= NULL
;
7548 /* add spare device to update
7550 for (id
= super
->devlist
; id
; id
= id
->next
)
7551 if (id
->index
== (unsigned)u
->subdev
) {
7552 struct imsm_dev
*dev
;
7553 struct imsm_map
*map
;
7555 dev
= get_imsm_dev(super
, u
->subdev
);
7556 map
= get_imsm_map(dev
, 0);
7557 current_level
= map
->raid_level
;
7560 if ((u
->new_level
== 5) && (u
->new_level
!= current_level
)) {
7561 struct mdinfo
*spares
;
7563 spares
= get_spares_for_grow(st
);
7571 makedev(dev
->disk
.major
,
7573 dl
= get_disk_super(super
,
7576 dl
->index
= u
->old_raid_disks
;
7582 len
= disks_to_mpb_size(u
->new_raid_disks
);
7583 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7586 case update_create_array
: {
7587 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7588 struct intel_dev
*dv
;
7589 struct imsm_dev
*dev
= &u
->dev
;
7590 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7592 struct disk_info
*inf
;
7596 inf
= get_disk_info(u
);
7597 len
= sizeof_imsm_dev(dev
, 1);
7598 /* allocate a new super->devlist entry */
7599 dv
= malloc(sizeof(*dv
));
7601 dv
->dev
= malloc(len
);
7606 update
->space
= NULL
;
7610 /* count how many spares will be converted to members */
7611 for (i
= 0; i
< map
->num_members
; i
++) {
7612 dl
= serial_to_dl(inf
[i
].serial
, super
);
7614 /* hmm maybe it failed?, nothing we can do about
7619 if (count_memberships(dl
, super
) == 0)
7622 len
+= activate
* sizeof(struct imsm_disk
);
7629 /* check if we need a larger metadata buffer */
7630 if (super
->next_buf
)
7631 buf_len
= super
->next_len
;
7633 buf_len
= super
->len
;
7635 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
7636 /* ok we need a larger buf than what is currently allocated
7637 * if this allocation fails process_update will notice that
7638 * ->next_len is set and ->next_buf is NULL
7640 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
7641 if (super
->next_buf
)
7642 free(super
->next_buf
);
7644 super
->next_len
= buf_len
;
7645 if (posix_memalign(&super
->next_buf
, 512, buf_len
) == 0)
7646 memset(super
->next_buf
, 0, buf_len
);
7648 super
->next_buf
= NULL
;
7652 /* must be called while manager is quiesced */
7653 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
)
7655 struct imsm_super
*mpb
= super
->anchor
;
7657 struct imsm_dev
*dev
;
7658 struct imsm_map
*map
;
7659 int i
, j
, num_members
;
7662 dprintf("%s: deleting device[%d] from imsm_super\n",
7665 /* shift all indexes down one */
7666 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
7667 if (iter
->index
> (int)index
)
7669 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
7670 if (iter
->index
> (int)index
)
7673 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7674 dev
= get_imsm_dev(super
, i
);
7675 map
= get_imsm_map(dev
, 0);
7676 num_members
= map
->num_members
;
7677 for (j
= 0; j
< num_members
; j
++) {
7678 /* update ord entries being careful not to propagate
7679 * ord-flags to the first map
7681 ord
= get_imsm_ord_tbl_ent(dev
, j
, -1);
7683 if (ord_to_idx(ord
) <= index
)
7686 map
= get_imsm_map(dev
, 0);
7687 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
7688 map
= get_imsm_map(dev
, 1);
7690 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
7695 super
->updates_pending
++;
7697 struct dl
*dl
= *dlp
;
7699 *dlp
= (*dlp
)->next
;
7700 __free_imsm_disk(dl
);
7703 #endif /* MDASSEMBLE */
7704 /*******************************************************************************
7705 * Function: open_backup_targets
7706 * Description: Function opens file descriptors for all devices given in
7709 * info : general array info
7710 * raid_disks : number of disks
7711 * raid_fds : table of device's file descriptors
7715 ******************************************************************************/
7716 int open_backup_targets(struct mdinfo
*info
, int raid_disks
, int *raid_fds
)
7720 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7723 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
)) {
7724 dprintf("disk is faulty!!\n");
7728 if ((sd
->disk
.raid_disk
>= raid_disks
) ||
7729 (sd
->disk
.raid_disk
< 0))
7732 dn
= map_dev(sd
->disk
.major
,
7734 raid_fds
[sd
->disk
.raid_disk
] = dev_open(dn
, O_RDWR
);
7735 if (raid_fds
[sd
->disk
.raid_disk
] < 0) {
7736 fprintf(stderr
, "cannot open component\n");
7744 /*******************************************************************************
7745 * Function: init_migr_record_imsm
7746 * Description: Function inits imsm migration record
7748 * super : imsm internal array info
7749 * dev : device under migration
7750 * info : general array info to find the smallest device
7753 ******************************************************************************/
7754 void init_migr_record_imsm(struct supertype
*st
, struct imsm_dev
*dev
,
7755 struct mdinfo
*info
)
7757 struct intel_super
*super
= st
->sb
;
7758 struct migr_record
*migr_rec
= super
->migr_rec
;
7760 unsigned long long dsize
, dev_sectors
;
7761 long long unsigned min_dev_sectors
= -1LLU;
7765 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7766 struct imsm_map
*map_src
= get_imsm_map(dev
, 1);
7767 unsigned long long num_migr_units
;
7768 unsigned long long array_blocks
;
7770 memset(migr_rec
, 0, sizeof(struct migr_record
));
7771 migr_rec
->family_num
= __cpu_to_le32(super
->anchor
->family_num
);
7773 /* only ascending reshape supported now */
7774 migr_rec
->ascending_migr
= __cpu_to_le32(1);
7776 migr_rec
->dest_depth_per_unit
= GEN_MIGR_AREA_SIZE
/
7777 max(map_dest
->blocks_per_strip
, map_src
->blocks_per_strip
);
7778 migr_rec
->dest_depth_per_unit
*= map_dest
->blocks_per_strip
;
7779 new_data_disks
= imsm_num_data_members(dev
, 0);
7780 migr_rec
->blocks_per_unit
=
7781 __cpu_to_le32(migr_rec
->dest_depth_per_unit
* new_data_disks
);
7782 migr_rec
->dest_depth_per_unit
=
7783 __cpu_to_le32(migr_rec
->dest_depth_per_unit
);
7784 array_blocks
= info
->component_size
* new_data_disks
;
7786 array_blocks
/ __le32_to_cpu(migr_rec
->blocks_per_unit
);
7788 if (array_blocks
% __le32_to_cpu(migr_rec
->blocks_per_unit
))
7790 migr_rec
->num_migr_units
= __cpu_to_le32(num_migr_units
);
7792 migr_rec
->post_migr_vol_cap
= dev
->size_low
;
7793 migr_rec
->post_migr_vol_cap_hi
= dev
->size_high
;
7796 /* Find the smallest dev */
7797 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7798 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
7799 fd
= dev_open(nm
, O_RDONLY
);
7802 get_dev_size(fd
, NULL
, &dsize
);
7803 dev_sectors
= dsize
/ 512;
7804 if (dev_sectors
< min_dev_sectors
)
7805 min_dev_sectors
= dev_sectors
;
7808 migr_rec
->ckpt_area_pba
= __cpu_to_le32(min_dev_sectors
-
7809 RAID_DISK_RESERVED_BLOCKS_IMSM_HI
);
7811 write_imsm_migr_rec(st
);
7816 /*******************************************************************************
7817 * Function: save_backup_imsm
7818 * Description: Function saves critical data stripes to Migration Copy Area
7819 * and updates the current migration unit status.
7820 * Use restore_stripes() to form a destination stripe,
7821 * and to write it to the Copy Area.
7823 * st : supertype information
7824 * dev : imsm device that backup is saved for
7825 * info : general array info
7826 * buf : input buffer
7827 * length : length of data to backup (blocks_per_unit)
7831 ******************************************************************************/
7832 int save_backup_imsm(struct supertype
*st
,
7833 struct imsm_dev
*dev
,
7834 struct mdinfo
*info
,
7839 struct intel_super
*super
= st
->sb
;
7840 unsigned long long *target_offsets
= NULL
;
7841 int *targets
= NULL
;
7843 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7844 int new_disks
= map_dest
->num_members
;
7845 int dest_layout
= 0;
7847 unsigned long long start
;
7848 int data_disks
= imsm_num_data_members(dev
, 0);
7850 targets
= malloc(new_disks
* sizeof(int));
7854 for (i
= 0; i
< new_disks
; i
++)
7857 target_offsets
= malloc(new_disks
* sizeof(unsigned long long));
7858 if (!target_offsets
)
7861 start
= info
->reshape_progress
* 512;
7862 for (i
= 0; i
< new_disks
; i
++) {
7863 target_offsets
[i
] = (unsigned long long)
7864 __le32_to_cpu(super
->migr_rec
->ckpt_area_pba
) * 512;
7865 /* move back copy area adderss, it will be moved forward
7866 * in restore_stripes() using start input variable
7868 target_offsets
[i
] -= start
/data_disks
;
7871 if (open_backup_targets(info
, new_disks
, targets
))
7874 dest_layout
= imsm_level_to_layout(map_dest
->raid_level
);
7875 dest_chunk
= __le16_to_cpu(map_dest
->blocks_per_strip
) * 512;
7877 if (restore_stripes(targets
, /* list of dest devices */
7878 target_offsets
, /* migration record offsets */
7881 map_dest
->raid_level
,
7883 -1, /* source backup file descriptor */
7884 0, /* input buf offset
7885 * always 0 buf is already offseted */
7889 fprintf(stderr
, Name
": Error restoring stripes\n");
7897 for (i
= 0; i
< new_disks
; i
++)
7898 if (targets
[i
] >= 0)
7902 free(target_offsets
);
7907 /*******************************************************************************
7908 * Function: save_checkpoint_imsm
7909 * Description: Function called for current unit status update
7910 * in the migration record. It writes it to disk.
7912 * super : imsm internal array info
7913 * info : general array info
7917 * 2: failure, means no valid migration record
7918 * / no general migration in progress /
7919 ******************************************************************************/
7920 int save_checkpoint_imsm(struct supertype
*st
, struct mdinfo
*info
, int state
)
7922 struct intel_super
*super
= st
->sb
;
7923 unsigned long long blocks_per_unit
;
7924 unsigned long long curr_migr_unit
;
7926 if (load_imsm_migr_rec(super
, info
) != 0) {
7927 dprintf("imsm: ERROR: Cannot read migration record "
7928 "for checkpoint save.\n");
7932 blocks_per_unit
= __le32_to_cpu(super
->migr_rec
->blocks_per_unit
);
7933 if (blocks_per_unit
== 0) {
7934 dprintf("imsm: no migration in progress.\n");
7937 curr_migr_unit
= info
->reshape_progress
/ blocks_per_unit
;
7938 /* check if array is alligned to copy area
7939 * if it is not alligned, add one to current migration unit value
7940 * this can happend on array reshape finish only
7942 if (info
->reshape_progress
% blocks_per_unit
)
7945 super
->migr_rec
->curr_migr_unit
=
7946 __cpu_to_le32(curr_migr_unit
);
7947 super
->migr_rec
->rec_status
= __cpu_to_le32(state
);
7948 super
->migr_rec
->dest_1st_member_lba
=
7949 __cpu_to_le32(curr_migr_unit
*
7950 __le32_to_cpu(super
->migr_rec
->dest_depth_per_unit
));
7951 if (write_imsm_migr_rec(st
) < 0) {
7952 dprintf("imsm: Cannot write migration record "
7953 "outside backup area\n");
7960 /*******************************************************************************
7961 * Function: recover_backup_imsm
7962 * Description: Function recovers critical data from the Migration Copy Area
7963 * while assembling an array.
7965 * super : imsm internal array info
7966 * info : general array info
7968 * 0 : success (or there is no data to recover)
7970 ******************************************************************************/
7971 int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
)
7973 struct intel_super
*super
= st
->sb
;
7974 struct migr_record
*migr_rec
= super
->migr_rec
;
7975 struct imsm_map
*map_dest
= NULL
;
7976 struct intel_dev
*id
= NULL
;
7977 unsigned long long read_offset
;
7978 unsigned long long write_offset
;
7980 int *targets
= NULL
;
7981 int new_disks
, i
, err
;
7984 unsigned long curr_migr_unit
= __le32_to_cpu(migr_rec
->curr_migr_unit
);
7985 unsigned long num_migr_units
= __le32_to_cpu(migr_rec
->num_migr_units
);
7987 int skipped_disks
= 0;
7988 int max_degradation
;
7990 err
= sysfs_get_str(info
, NULL
, "array_state", (char *)buffer
, 20);
7994 /* recover data only during assemblation */
7995 if (strncmp(buffer
, "inactive", 8) != 0)
7997 /* no data to recover */
7998 if (__le32_to_cpu(migr_rec
->rec_status
) == UNIT_SRC_NORMAL
)
8000 if (curr_migr_unit
>= num_migr_units
)
8003 /* find device during reshape */
8004 for (id
= super
->devlist
; id
; id
= id
->next
)
8005 if (is_gen_migration(id
->dev
))
8010 map_dest
= get_imsm_map(id
->dev
, 0);
8011 new_disks
= map_dest
->num_members
;
8012 max_degradation
= new_disks
- imsm_num_data_members(id
->dev
, 0);
8014 read_offset
= (unsigned long long)
8015 __le32_to_cpu(migr_rec
->ckpt_area_pba
) * 512;
8017 write_offset
= ((unsigned long long)
8018 __le32_to_cpu(migr_rec
->dest_1st_member_lba
) +
8019 __le32_to_cpu(map_dest
->pba_of_lba0
)) * 512;
8021 unit_len
= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8022 if (posix_memalign((void **)&buf
, 512, unit_len
) != 0)
8024 targets
= malloc(new_disks
* sizeof(int));
8028 open_backup_targets(info
, new_disks
, targets
);
8030 for (i
= 0; i
< new_disks
; i
++) {
8031 if (targets
[i
] < 0) {
8035 if (lseek64(targets
[i
], read_offset
, SEEK_SET
) < 0) {
8037 Name
": Cannot seek to block: %s\n",
8041 if ((unsigned)read(targets
[i
], buf
, unit_len
) != unit_len
) {
8043 Name
": Cannot read copy area block: %s\n",
8047 if (lseek64(targets
[i
], write_offset
, SEEK_SET
) < 0) {
8049 Name
": Cannot seek to block: %s\n",
8053 if ((unsigned)write(targets
[i
], buf
, unit_len
) != unit_len
) {
8055 Name
": Cannot restore block: %s\n",
8061 if (skipped_disks
> max_degradation
) {
8063 Name
": Cannot restore data from backup."
8064 " Too many failed disks\n");
8068 if (save_checkpoint_imsm(st
, info
, UNIT_SRC_NORMAL
)) {
8069 /* ignore error == 2, this can mean end of reshape here
8071 dprintf("imsm: Cannot write checkpoint to "
8072 "migration record (UNIT_SRC_NORMAL) during restart\n");
8078 for (i
= 0; i
< new_disks
; i
++)
8087 static char disk_by_path
[] = "/dev/disk/by-path/";
8089 static const char *imsm_get_disk_controller_domain(const char *path
)
8091 char disk_path
[PATH_MAX
];
8095 strncpy(disk_path
, disk_by_path
, PATH_MAX
- 1);
8096 strncat(disk_path
, path
, PATH_MAX
- strlen(disk_path
) - 1);
8097 if (stat(disk_path
, &st
) == 0) {
8098 struct sys_dev
* hba
;
8101 path
= devt_to_devpath(st
.st_rdev
);
8104 hba
= find_disk_attached_hba(-1, path
);
8105 if (hba
&& hba
->type
== SYS_DEV_SAS
)
8107 else if (hba
&& hba
->type
== SYS_DEV_SATA
)
8111 dprintf("path: %s hba: %s attached: %s\n",
8112 path
, (hba
) ? hba
->path
: "NULL", drv
);
8120 static int imsm_find_array_minor_by_subdev(int subdev
, int container
, int *minor
)
8122 char subdev_name
[20];
8123 struct mdstat_ent
*mdstat
;
8125 sprintf(subdev_name
, "%d", subdev
);
8126 mdstat
= mdstat_by_subdev(subdev_name
, container
);
8130 *minor
= mdstat
->devnum
;
8131 free_mdstat(mdstat
);
8135 static int imsm_reshape_is_allowed_on_container(struct supertype
*st
,
8136 struct geo_params
*geo
,
8137 int *old_raid_disks
)
8139 /* currently we only support increasing the number of devices
8140 * for a container. This increases the number of device for each
8141 * member array. They must all be RAID0 or RAID5.
8144 struct mdinfo
*info
, *member
;
8145 int devices_that_can_grow
= 0;
8147 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): "
8148 "st->devnum = (%i)\n",
8151 if (geo
->size
!= -1 ||
8152 geo
->level
!= UnSet
||
8153 geo
->layout
!= UnSet
||
8154 geo
->chunksize
!= 0 ||
8155 geo
->raid_disks
== UnSet
) {
8156 dprintf("imsm: Container operation is allowed for "
8157 "raid disks number change only.\n");
8161 info
= container_content_imsm(st
, NULL
);
8162 for (member
= info
; member
; member
= member
->next
) {
8166 dprintf("imsm: checking device_num: %i\n",
8167 member
->container_member
);
8169 if (geo
->raid_disks
<= member
->array
.raid_disks
) {
8170 /* we work on container for Online Capacity Expansion
8171 * only so raid_disks has to grow
8173 dprintf("imsm: for container operation raid disks "
8174 "increase is required\n");
8178 if ((info
->array
.level
!= 0) &&
8179 (info
->array
.level
!= 5)) {
8180 /* we cannot use this container with other raid level
8182 dprintf("imsm: for container operation wrong"
8183 " raid level (%i) detected\n",
8187 /* check for platform support
8188 * for this raid level configuration
8190 struct intel_super
*super
= st
->sb
;
8191 if (!is_raid_level_supported(super
->orom
,
8192 member
->array
.level
,
8194 dprintf("platform does not support raid%d with"
8198 geo
->raid_disks
> 1 ? "s" : "");
8201 /* check if component size is aligned to chunk size
8203 if (info
->component_size
%
8204 (info
->array
.chunk_size
/512)) {
8205 dprintf("Component size is not aligned to "
8211 if (*old_raid_disks
&&
8212 info
->array
.raid_disks
!= *old_raid_disks
)
8214 *old_raid_disks
= info
->array
.raid_disks
;
8216 /* All raid5 and raid0 volumes in container
8217 * have to be ready for Online Capacity Expansion
8218 * so they need to be assembled. We have already
8219 * checked that no recovery etc is happening.
8221 result
= imsm_find_array_minor_by_subdev(member
->container_member
,
8225 dprintf("imsm: cannot find array\n");
8228 devices_that_can_grow
++;
8231 if (!member
&& devices_that_can_grow
)
8235 dprintf("\tContainer operation allowed\n");
8237 dprintf("\tError: %i\n", ret_val
);
8242 /* Function: get_spares_for_grow
8243 * Description: Allocates memory and creates list of spare devices
8244 * avaliable in container. Checks if spare drive size is acceptable.
8245 * Parameters: Pointer to the supertype structure
8246 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
8249 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
)
8251 unsigned long long min_size
= min_acceptable_spare_size_imsm(st
);
8252 return container_choose_spares(st
, min_size
, NULL
, NULL
, NULL
, 0);
8255 /******************************************************************************
8256 * function: imsm_create_metadata_update_for_reshape
8257 * Function creates update for whole IMSM container.
8259 ******************************************************************************/
8260 static int imsm_create_metadata_update_for_reshape(
8261 struct supertype
*st
,
8262 struct geo_params
*geo
,
8264 struct imsm_update_reshape
**updatep
)
8266 struct intel_super
*super
= st
->sb
;
8267 struct imsm_super
*mpb
= super
->anchor
;
8268 int update_memory_size
= 0;
8269 struct imsm_update_reshape
*u
= NULL
;
8270 struct mdinfo
*spares
= NULL
;
8272 int delta_disks
= 0;
8275 dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n",
8278 delta_disks
= geo
->raid_disks
- old_raid_disks
;
8280 /* size of all update data without anchor */
8281 update_memory_size
= sizeof(struct imsm_update_reshape
);
8283 /* now add space for spare disks that we need to add. */
8284 update_memory_size
+= sizeof(u
->new_disks
[0]) * (delta_disks
- 1);
8286 u
= calloc(1, update_memory_size
);
8289 "cannot get memory for imsm_update_reshape update\n");
8292 u
->type
= update_reshape_container_disks
;
8293 u
->old_raid_disks
= old_raid_disks
;
8294 u
->new_raid_disks
= geo
->raid_disks
;
8296 /* now get spare disks list
8298 spares
= get_spares_for_grow(st
);
8301 || delta_disks
> spares
->array
.spare_disks
) {
8302 fprintf(stderr
, Name
": imsm: ERROR: Cannot get spare devices "
8303 "for %s.\n", geo
->dev_name
);
8308 /* we have got spares
8309 * update disk list in imsm_disk list table in anchor
8311 dprintf("imsm: %i spares are available.\n\n",
8312 spares
->array
.spare_disks
);
8315 for (i
= 0; i
< delta_disks
; i
++) {
8320 u
->new_disks
[i
] = makedev(dev
->disk
.major
,
8322 dl
= get_disk_super(super
, dev
->disk
.major
, dev
->disk
.minor
);
8323 dl
->index
= mpb
->num_disks
;
8333 dprintf("imsm: reshape update preparation :");
8334 if (i
== delta_disks
) {
8337 return update_memory_size
;
8340 dprintf(" Error\n");
8345 /******************************************************************************
8346 * function: imsm_create_metadata_update_for_migration()
8347 * Creates update for IMSM array.
8349 ******************************************************************************/
8350 static int imsm_create_metadata_update_for_migration(
8351 struct supertype
*st
,
8352 struct geo_params
*geo
,
8353 struct imsm_update_reshape_migration
**updatep
)
8355 struct intel_super
*super
= st
->sb
;
8356 int update_memory_size
= 0;
8357 struct imsm_update_reshape_migration
*u
= NULL
;
8358 struct imsm_dev
*dev
;
8359 int previous_level
= -1;
8361 dprintf("imsm_create_metadata_update_for_migration(enter)"
8362 " New Level = %i\n", geo
->level
);
8364 /* size of all update data without anchor */
8365 update_memory_size
= sizeof(struct imsm_update_reshape_migration
);
8367 u
= calloc(1, update_memory_size
);
8369 dprintf("error: cannot get memory for "
8370 "imsm_create_metadata_update_for_migration\n");
8373 u
->type
= update_reshape_migration
;
8374 u
->subdev
= super
->current_vol
;
8375 u
->new_level
= geo
->level
;
8376 u
->new_layout
= geo
->layout
;
8377 u
->new_raid_disks
= u
->old_raid_disks
= geo
->raid_disks
;
8378 u
->new_disks
[0] = -1;
8379 u
->new_chunksize
= -1;
8381 dev
= get_imsm_dev(super
, u
->subdev
);
8383 struct imsm_map
*map
;
8385 map
= get_imsm_map(dev
, 0);
8387 int current_chunk_size
=
8388 __le16_to_cpu(map
->blocks_per_strip
) / 2;
8390 if (geo
->chunksize
!= current_chunk_size
) {
8391 u
->new_chunksize
= geo
->chunksize
/ 1024;
8393 "chunk size change from %i to %i\n",
8394 current_chunk_size
, u
->new_chunksize
);
8396 previous_level
= map
->raid_level
;
8399 if ((geo
->level
== 5) && (previous_level
== 0)) {
8400 struct mdinfo
*spares
= NULL
;
8402 u
->new_raid_disks
++;
8403 spares
= get_spares_for_grow(st
);
8404 if ((spares
== NULL
) || (spares
->array
.spare_disks
< 1)) {
8407 update_memory_size
= 0;
8408 dprintf("error: cannot get spare device "
8409 "for requested migration");
8414 dprintf("imsm: reshape update preparation : OK\n");
8417 return update_memory_size
;
8420 static void imsm_update_metadata_locally(struct supertype
*st
,
8423 struct metadata_update mu
;
8428 mu
.space_list
= NULL
;
8430 imsm_prepare_update(st
, &mu
);
8431 imsm_process_update(st
, &mu
);
8433 while (mu
.space_list
) {
8434 void **space
= mu
.space_list
;
8435 mu
.space_list
= *space
;
8440 /***************************************************************************
8441 * Function: imsm_analyze_change
8442 * Description: Function analyze change for single volume
8443 * and validate if transition is supported
8444 * Parameters: Geometry parameters, supertype structure
8445 * Returns: Operation type code on success, -1 if fail
8446 ****************************************************************************/
8447 enum imsm_reshape_type
imsm_analyze_change(struct supertype
*st
,
8448 struct geo_params
*geo
)
8455 getinfo_super_imsm_volume(st
, &info
, NULL
);
8456 if ((geo
->level
!= info
.array
.level
) &&
8457 (geo
->level
>= 0) &&
8458 (geo
->level
!= UnSet
)) {
8459 switch (info
.array
.level
) {
8461 if (geo
->level
== 5) {
8462 change
= CH_MIGRATION
;
8463 if (geo
->layout
!= ALGORITHM_LEFT_ASYMMETRIC
) {
8465 Name
" Error. Requested Layout "
8466 "not supported (left-asymmetric layout "
8467 "is supported only)!\n");
8469 goto analyse_change_exit
;
8473 if (geo
->level
== 10) {
8474 change
= CH_TAKEOVER
;
8479 if (geo
->level
== 0) {
8480 change
= CH_TAKEOVER
;
8485 if (geo
->level
== 0) {
8486 change
= CH_TAKEOVER
;
8493 Name
" Error. Level Migration from %d to %d "
8495 info
.array
.level
, geo
->level
);
8496 goto analyse_change_exit
;
8499 geo
->level
= info
.array
.level
;
8501 if ((geo
->layout
!= info
.array
.layout
)
8502 && ((geo
->layout
!= UnSet
) && (geo
->layout
!= -1))) {
8503 change
= CH_MIGRATION
;
8504 if ((info
.array
.layout
== 0)
8505 && (info
.array
.level
== 5)
8506 && (geo
->layout
== 5)) {
8507 /* reshape 5 -> 4 */
8508 } else if ((info
.array
.layout
== 5)
8509 && (info
.array
.level
== 5)
8510 && (geo
->layout
== 0)) {
8511 /* reshape 4 -> 5 */
8516 Name
" Error. Layout Migration from %d to %d "
8518 info
.array
.layout
, geo
->layout
);
8520 goto analyse_change_exit
;
8523 geo
->layout
= info
.array
.layout
;
8525 if ((geo
->chunksize
> 0) && (geo
->chunksize
!= UnSet
)
8526 && (geo
->chunksize
!= info
.array
.chunk_size
))
8527 change
= CH_MIGRATION
;
8529 geo
->chunksize
= info
.array
.chunk_size
;
8531 chunk
= geo
->chunksize
/ 1024;
8532 if (!validate_geometry_imsm(st
,
8542 struct intel_super
*super
= st
->sb
;
8543 struct imsm_super
*mpb
= super
->anchor
;
8545 if (mpb
->num_raid_devs
> 1) {
8547 Name
" Error. Cannot perform operation on %s"
8548 "- for this operation it MUST be single "
8549 "array in container\n",
8555 analyse_change_exit
:
8560 int imsm_takeover(struct supertype
*st
, struct geo_params
*geo
)
8562 struct intel_super
*super
= st
->sb
;
8563 struct imsm_update_takeover
*u
;
8565 u
= malloc(sizeof(struct imsm_update_takeover
));
8569 u
->type
= update_takeover
;
8570 u
->subarray
= super
->current_vol
;
8572 /* 10->0 transition */
8573 if (geo
->level
== 0)
8574 u
->direction
= R10_TO_R0
;
8576 /* 0->10 transition */
8577 if (geo
->level
== 10)
8578 u
->direction
= R0_TO_R10
;
8580 /* update metadata locally */
8581 imsm_update_metadata_locally(st
, u
,
8582 sizeof(struct imsm_update_takeover
));
8583 /* and possibly remotely */
8584 if (st
->update_tail
)
8585 append_metadata_update(st
, u
,
8586 sizeof(struct imsm_update_takeover
));
8593 static int imsm_reshape_super(struct supertype
*st
, long long size
, int level
,
8594 int layout
, int chunksize
, int raid_disks
,
8595 int delta_disks
, char *backup
, char *dev
,
8599 struct geo_params geo
;
8601 dprintf("imsm: reshape_super called.\n");
8603 memset(&geo
, 0, sizeof(struct geo_params
));
8606 geo
.dev_id
= st
->devnum
;
8609 geo
.layout
= layout
;
8610 geo
.chunksize
= chunksize
;
8611 geo
.raid_disks
= raid_disks
;
8612 if (delta_disks
!= UnSet
)
8613 geo
.raid_disks
+= delta_disks
;
8615 dprintf("\tfor level : %i\n", geo
.level
);
8616 dprintf("\tfor raid_disks : %i\n", geo
.raid_disks
);
8618 if (experimental() == 0)
8621 if (st
->container_dev
== st
->devnum
) {
8622 /* On container level we can only increase number of devices. */
8623 dprintf("imsm: info: Container operation\n");
8624 int old_raid_disks
= 0;
8626 if (imsm_reshape_is_allowed_on_container(
8627 st
, &geo
, &old_raid_disks
)) {
8628 struct imsm_update_reshape
*u
= NULL
;
8631 len
= imsm_create_metadata_update_for_reshape(
8632 st
, &geo
, old_raid_disks
, &u
);
8635 dprintf("imsm: Cannot prepare update\n");
8636 goto exit_imsm_reshape_super
;
8640 /* update metadata locally */
8641 imsm_update_metadata_locally(st
, u
, len
);
8642 /* and possibly remotely */
8643 if (st
->update_tail
)
8644 append_metadata_update(st
, u
, len
);
8649 fprintf(stderr
, Name
": (imsm) Operation "
8650 "is not allowed on this container\n");
8653 /* On volume level we support following operations
8654 * - takeover: raid10 -> raid0; raid0 -> raid10
8655 * - chunk size migration
8656 * - migration: raid5 -> raid0; raid0 -> raid5
8658 struct intel_super
*super
= st
->sb
;
8659 struct intel_dev
*dev
= super
->devlist
;
8661 dprintf("imsm: info: Volume operation\n");
8662 /* find requested device */
8664 imsm_find_array_minor_by_subdev(dev
->index
, st
->container_dev
, &devnum
);
8665 if (devnum
== geo
.dev_id
)
8670 fprintf(stderr
, Name
" Cannot find %s (%i) subarray\n",
8671 geo
.dev_name
, geo
.dev_id
);
8672 goto exit_imsm_reshape_super
;
8674 super
->current_vol
= dev
->index
;
8675 change
= imsm_analyze_change(st
, &geo
);
8678 ret_val
= imsm_takeover(st
, &geo
);
8680 case CH_MIGRATION
: {
8681 struct imsm_update_reshape_migration
*u
= NULL
;
8683 imsm_create_metadata_update_for_migration(
8687 "Cannot prepare update\n");
8691 /* update metadata locally */
8692 imsm_update_metadata_locally(st
, u
, len
);
8693 /* and possibly remotely */
8694 if (st
->update_tail
)
8695 append_metadata_update(st
, u
, len
);
8705 exit_imsm_reshape_super
:
8706 dprintf("imsm: reshape_super Exit code = %i\n", ret_val
);
8710 /*******************************************************************************
8711 * Function: wait_for_reshape_imsm
8712 * Description: Function writes new sync_max value and waits until
8713 * reshape process reach new position
8715 * sra : general array info
8716 * ndata : number of disks in new array's layout
8719 * 1 : there is no reshape in progress,
8721 ******************************************************************************/
8722 int wait_for_reshape_imsm(struct mdinfo
*sra
, int ndata
)
8724 int fd
= sysfs_get_fd(sra
, NULL
, "reshape_position");
8725 unsigned long long completed
;
8726 /* to_complete : new sync_max position */
8727 unsigned long long to_complete
= sra
->reshape_progress
;
8728 unsigned long long position_to_set
= to_complete
/ ndata
;
8731 dprintf("imsm: wait_for_reshape_imsm() "
8732 "cannot open reshape_position\n");
8736 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8737 dprintf("imsm: wait_for_reshape_imsm() "
8738 "cannot read reshape_position (no reshape in progres)\n");
8743 if (completed
> to_complete
) {
8744 dprintf("imsm: wait_for_reshape_imsm() "
8745 "wrong next position to set %llu (%llu)\n",
8746 to_complete
, completed
);
8750 dprintf("Position set: %llu\n", position_to_set
);
8751 if (sysfs_set_num(sra
, NULL
, "sync_max",
8752 position_to_set
) != 0) {
8753 dprintf("imsm: wait_for_reshape_imsm() "
8754 "cannot set reshape position to %llu\n",
8765 select(fd
+1, &rfds
, NULL
, NULL
, NULL
);
8766 if (sysfs_get_str(sra
, NULL
, "sync_action",
8768 strncmp(action
, "reshape", 7) != 0)
8770 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8771 dprintf("imsm: wait_for_reshape_imsm() "
8772 "cannot read reshape_position (in loop)\n");
8776 } while (completed
< to_complete
);
8782 /*******************************************************************************
8783 * Function: check_degradation_change
8784 * Description: Check that array hasn't become failed.
8786 * info : for sysfs access
8787 * sources : source disks descriptors
8788 * degraded: previous degradation level
8791 ******************************************************************************/
8792 int check_degradation_change(struct mdinfo
*info
,
8796 unsigned long long new_degraded
;
8797 sysfs_get_ll(info
, NULL
, "degraded", &new_degraded
);
8798 if (new_degraded
!= (unsigned long long)degraded
) {
8799 /* check each device to ensure it is still working */
8802 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
8803 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
))
8805 if (sd
->disk
.state
& (1<<MD_DISK_SYNC
)) {
8807 if (sysfs_get_str(info
,
8808 sd
, "state", sbuf
, 20) < 0 ||
8809 strstr(sbuf
, "faulty") ||
8810 strstr(sbuf
, "in_sync") == NULL
) {
8811 /* this device is dead */
8812 sd
->disk
.state
= (1<<MD_DISK_FAULTY
);
8813 if (sd
->disk
.raid_disk
>= 0 &&
8814 sources
[sd
->disk
.raid_disk
] >= 0) {
8816 sd
->disk
.raid_disk
]);
8817 sources
[sd
->disk
.raid_disk
] =
8826 return new_degraded
;
8829 /*******************************************************************************
8830 * Function: imsm_manage_reshape
8831 * Description: Function finds array under reshape and it manages reshape
8832 * process. It creates stripes backups (if required) and sets
8835 * afd : Backup handle (nattive) - not used
8836 * sra : general array info
8837 * reshape : reshape parameters - not used
8838 * st : supertype structure
8839 * blocks : size of critical section [blocks]
8840 * fds : table of source device descriptor
8841 * offsets : start of array (offest per devices)
8843 * destfd : table of destination device descriptor
8844 * destoffsets : table of destination offsets (per device)
8846 * 1 : success, reshape is done
8848 ******************************************************************************/
8849 static int imsm_manage_reshape(
8850 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
8851 struct supertype
*st
, unsigned long backup_blocks
,
8852 int *fds
, unsigned long long *offsets
,
8853 int dests
, int *destfd
, unsigned long long *destoffsets
)
8856 struct intel_super
*super
= st
->sb
;
8857 struct intel_dev
*dv
= NULL
;
8858 struct imsm_dev
*dev
= NULL
;
8859 struct imsm_map
*map_src
;
8860 int migr_vol_qan
= 0;
8861 int ndata
, odata
; /* [bytes] */
8862 int chunk
; /* [bytes] */
8863 struct migr_record
*migr_rec
;
8865 unsigned int buf_size
; /* [bytes] */
8866 unsigned long long max_position
; /* array size [bytes] */
8867 unsigned long long next_step
; /* [blocks]/[bytes] */
8868 unsigned long long old_data_stripe_length
;
8869 unsigned long long start_src
; /* [bytes] */
8870 unsigned long long start
; /* [bytes] */
8871 unsigned long long start_buf_shift
; /* [bytes] */
8873 int source_layout
= 0;
8875 if (!fds
|| !offsets
|| !sra
)
8878 /* Find volume during the reshape */
8879 for (dv
= super
->devlist
; dv
; dv
= dv
->next
) {
8880 if (dv
->dev
->vol
.migr_type
== MIGR_GEN_MIGR
8881 && dv
->dev
->vol
.migr_state
== 1) {
8886 /* Only one volume can migrate at the same time */
8887 if (migr_vol_qan
!= 1) {
8888 fprintf(stderr
, Name
" : %s", migr_vol_qan
?
8889 "Number of migrating volumes greater than 1\n" :
8890 "There is no volume during migrationg\n");
8894 map_src
= get_imsm_map(dev
, 1);
8895 if (map_src
== NULL
)
8898 ndata
= imsm_num_data_members(dev
, 0);
8899 odata
= imsm_num_data_members(dev
, 1);
8901 chunk
= __le16_to_cpu(map_src
->blocks_per_strip
) * 512;
8902 old_data_stripe_length
= odata
* chunk
;
8904 migr_rec
= super
->migr_rec
;
8906 /* initialize migration record for start condition */
8907 if (sra
->reshape_progress
== 0)
8908 init_migr_record_imsm(st
, dev
, sra
);
8910 if (__le32_to_cpu(migr_rec
->rec_status
) != UNIT_SRC_NORMAL
) {
8911 dprintf("imsm: cannot restart migration when data "
8912 "are present in copy area.\n");
8918 buf_size
= __le32_to_cpu(migr_rec
->blocks_per_unit
) * 512;
8919 /* extend buffer size for parity disk */
8920 buf_size
+= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8921 /* add space for stripe aligment */
8922 buf_size
+= old_data_stripe_length
;
8923 if (posix_memalign((void **)&buf
, 4096, buf_size
)) {
8924 dprintf("imsm: Cannot allocate checpoint buffer\n");
8928 max_position
= sra
->component_size
* ndata
;
8929 source_layout
= imsm_level_to_layout(map_src
->raid_level
);
8931 while (__le32_to_cpu(migr_rec
->curr_migr_unit
) <
8932 __le32_to_cpu(migr_rec
->num_migr_units
)) {
8933 /* current reshape position [blocks] */
8934 unsigned long long current_position
=
8935 __le32_to_cpu(migr_rec
->blocks_per_unit
)
8936 * __le32_to_cpu(migr_rec
->curr_migr_unit
);
8937 unsigned long long border
;
8939 /* Check that array hasn't become failed.
8941 degraded
= check_degradation_change(sra
, fds
, degraded
);
8943 dprintf("imsm: Abort reshape due to degradation"
8944 " level (%i)\n", degraded
);
8948 next_step
= __le32_to_cpu(migr_rec
->blocks_per_unit
);
8950 if ((current_position
+ next_step
) > max_position
)
8951 next_step
= max_position
- current_position
;
8953 start
= current_position
* 512;
8955 /* allign reading start to old geometry */
8956 start_buf_shift
= start
% old_data_stripe_length
;
8957 start_src
= start
- start_buf_shift
;
8959 border
= (start_src
/ odata
) - (start
/ ndata
);
8961 if (border
<= __le32_to_cpu(migr_rec
->dest_depth_per_unit
)) {
8962 /* save critical stripes to buf
8963 * start - start address of current unit
8965 * start_src - start address of current unit
8966 * to backup alligned to source array
8969 unsigned long long next_step_filler
= 0;
8970 unsigned long long copy_length
= next_step
* 512;
8972 /* allign copy area length to stripe in old geometry */
8973 next_step_filler
= ((copy_length
+ start_buf_shift
)
8974 % old_data_stripe_length
);
8975 if (next_step_filler
)
8976 next_step_filler
= (old_data_stripe_length
8977 - next_step_filler
);
8978 dprintf("save_stripes() parameters: start = %llu,"
8979 "\tstart_src = %llu,\tnext_step*512 = %llu,"
8980 "\tstart_in_buf_shift = %llu,"
8981 "\tnext_step_filler = %llu\n",
8982 start
, start_src
, copy_length
,
8983 start_buf_shift
, next_step_filler
);
8985 if (save_stripes(fds
, offsets
, map_src
->num_members
,
8986 chunk
, map_src
->raid_level
,
8987 source_layout
, 0, NULL
, start_src
,
8989 next_step_filler
+ start_buf_shift
,
8991 dprintf("imsm: Cannot save stripes"
8995 /* Convert data to destination format and store it
8996 * in backup general migration area
8998 if (save_backup_imsm(st
, dev
, sra
,
8999 buf
+ start_buf_shift
, copy_length
)) {
9000 dprintf("imsm: Cannot save stripes to "
9001 "target devices\n");
9004 if (save_checkpoint_imsm(st
, sra
,
9005 UNIT_SRC_IN_CP_AREA
)) {
9006 dprintf("imsm: Cannot write checkpoint to "
9007 "migration record (UNIT_SRC_IN_CP_AREA)\n");
9011 /* set next step to use whole border area */
9012 border
/= next_step
;
9014 next_step
*= border
;
9016 /* When data backed up, checkpoint stored,
9017 * kick the kernel to reshape unit of data
9019 next_step
= next_step
+ sra
->reshape_progress
;
9020 /* limit next step to array max position */
9021 if (next_step
> max_position
)
9022 next_step
= max_position
;
9023 sysfs_set_num(sra
, NULL
, "suspend_lo", sra
->reshape_progress
);
9024 sysfs_set_num(sra
, NULL
, "suspend_hi", next_step
);
9025 sra
->reshape_progress
= next_step
;
9027 /* wait until reshape finish */
9028 if (wait_for_reshape_imsm(sra
, ndata
) < 0) {
9029 dprintf("wait_for_reshape_imsm returned error!\n");
9033 if (save_checkpoint_imsm(st
, sra
, UNIT_SRC_NORMAL
) == 1) {
9034 /* ignore error == 2, this can mean end of reshape here
9036 dprintf("imsm: Cannot write checkpoint to "
9037 "migration record (UNIT_SRC_NORMAL)\n");
9043 /* return '1' if done */
9051 #endif /* MDASSEMBLE */
9053 struct superswitch super_imsm
= {
9055 .examine_super
= examine_super_imsm
,
9056 .brief_examine_super
= brief_examine_super_imsm
,
9057 .brief_examine_subarrays
= brief_examine_subarrays_imsm
,
9058 .export_examine_super
= export_examine_super_imsm
,
9059 .detail_super
= detail_super_imsm
,
9060 .brief_detail_super
= brief_detail_super_imsm
,
9061 .write_init_super
= write_init_super_imsm
,
9062 .validate_geometry
= validate_geometry_imsm
,
9063 .add_to_super
= add_to_super_imsm
,
9064 .remove_from_super
= remove_from_super_imsm
,
9065 .detail_platform
= detail_platform_imsm
,
9066 .kill_subarray
= kill_subarray_imsm
,
9067 .update_subarray
= update_subarray_imsm
,
9068 .load_container
= load_container_imsm
,
9069 .default_geometry
= default_geometry_imsm
,
9070 .get_disk_controller_domain
= imsm_get_disk_controller_domain
,
9071 .reshape_super
= imsm_reshape_super
,
9072 .manage_reshape
= imsm_manage_reshape
,
9073 .recover_backup
= recover_backup_imsm
,
9075 .match_home
= match_home_imsm
,
9076 .uuid_from_super
= uuid_from_super_imsm
,
9077 .getinfo_super
= getinfo_super_imsm
,
9078 .getinfo_super_disks
= getinfo_super_disks_imsm
,
9079 .update_super
= update_super_imsm
,
9081 .avail_size
= avail_size_imsm
,
9082 .min_acceptable_spare_size
= min_acceptable_spare_size_imsm
,
9084 .compare_super
= compare_super_imsm
,
9086 .load_super
= load_super_imsm
,
9087 .init_super
= init_super_imsm
,
9088 .store_super
= store_super_imsm
,
9089 .free_super
= free_super_imsm
,
9090 .match_metadata_desc
= match_metadata_desc_imsm
,
9091 .container_content
= container_content_imsm
,
9099 .open_new
= imsm_open_new
,
9100 .set_array_state
= imsm_set_array_state
,
9101 .set_disk
= imsm_set_disk
,
9102 .sync_metadata
= imsm_sync_metadata
,
9103 .activate_spare
= imsm_activate_spare
,
9104 .process_update
= imsm_process_update
,
9105 .prepare_update
= imsm_prepare_update
,
9106 #endif /* MDASSEMBLE */