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", MAX_RAID_SERIAL_LEN
, 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 (unsigned long long)units
,
2330 (unsigned long long)blocks_per_unit
,
2331 info
->reshape_progress
);
2333 used_disks
= imsm_num_data_members(dev
, 1);
2334 if (used_disks
> 0) {
2335 array_blocks
= map
->blocks_per_member
*
2337 /* round array size down to closest MB
2339 info
->custom_array_size
= (array_blocks
2340 >> SECT_PER_MB_SHIFT
)
2341 << SECT_PER_MB_SHIFT
;
2345 /* we could emulate the checkpointing of
2346 * 'sync_action=check' migrations, but for now
2347 * we just immediately complete them
2350 /* this is handled by container_content_imsm() */
2351 case MIGR_STATE_CHANGE
:
2352 /* FIXME handle other migrations */
2354 /* we are not dirty, so... */
2355 info
->resync_start
= MaxSector
;
2359 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
2360 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
2362 info
->array
.major_version
= -1;
2363 info
->array
.minor_version
= -2;
2364 devname
= devnum2devname(st
->container_dev
);
2365 *info
->text_version
= '\0';
2367 sprintf(info
->text_version
, "/%s/%d", devname
, info
->container_member
);
2369 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
2370 uuid_from_super_imsm(st
, info
->uuid
);
2374 for (i
=0; i
<map_disks
; i
++) {
2376 if (i
< info
->array
.raid_disks
) {
2377 struct imsm_disk
*dsk
;
2378 j
= get_imsm_disk_idx(dev
, i
, -1);
2379 dsk
= get_imsm_disk(super
, j
);
2380 if (dsk
&& (dsk
->status
& CONFIGURED_DISK
))
2387 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
);
2388 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
);
2390 static struct imsm_disk
*get_imsm_missing(struct intel_super
*super
, __u8 index
)
2394 for (d
= super
->missing
; d
; d
= d
->next
)
2395 if (d
->index
== index
)
2400 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
, char *map
)
2402 struct intel_super
*super
= st
->sb
;
2403 struct imsm_disk
*disk
;
2404 int map_disks
= info
->array
.raid_disks
;
2405 int max_enough
= -1;
2407 struct imsm_super
*mpb
;
2409 if (super
->current_vol
>= 0) {
2410 getinfo_super_imsm_volume(st
, info
, map
);
2413 memset(info
, 0, sizeof(*info
));
2415 /* Set raid_disks to zero so that Assemble will always pull in valid
2418 info
->array
.raid_disks
= 0;
2419 info
->array
.level
= LEVEL_CONTAINER
;
2420 info
->array
.layout
= 0;
2421 info
->array
.md_minor
= -1;
2422 info
->array
.ctime
= 0; /* N/A for imsm */
2423 info
->array
.utime
= 0;
2424 info
->array
.chunk_size
= 0;
2426 info
->disk
.major
= 0;
2427 info
->disk
.minor
= 0;
2428 info
->disk
.raid_disk
= -1;
2429 info
->reshape_active
= 0;
2430 info
->array
.major_version
= -1;
2431 info
->array
.minor_version
= -2;
2432 strcpy(info
->text_version
, "imsm");
2433 info
->safe_mode_delay
= 0;
2434 info
->disk
.number
= -1;
2435 info
->disk
.state
= 0;
2437 info
->recovery_start
= MaxSector
;
2439 /* do we have the all the insync disks that we expect? */
2440 mpb
= super
->anchor
;
2442 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2443 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2444 int failed
, enough
, j
, missing
= 0;
2445 struct imsm_map
*map
;
2448 failed
= imsm_count_failed(super
, dev
);
2449 state
= imsm_check_degraded(super
, dev
, failed
);
2450 map
= get_imsm_map(dev
, dev
->vol
.migr_state
);
2452 /* any newly missing disks?
2453 * (catches single-degraded vs double-degraded)
2455 for (j
= 0; j
< map
->num_members
; j
++) {
2456 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
2457 __u32 idx
= ord_to_idx(ord
);
2459 if (!(ord
& IMSM_ORD_REBUILD
) &&
2460 get_imsm_missing(super
, idx
)) {
2466 if (state
== IMSM_T_STATE_FAILED
)
2468 else if (state
== IMSM_T_STATE_DEGRADED
&&
2469 (state
!= map
->map_state
|| missing
))
2471 else /* we're normal, or already degraded */
2474 /* in the missing/failed disk case check to see
2475 * if at least one array is runnable
2477 max_enough
= max(max_enough
, enough
);
2479 dprintf("%s: enough: %d\n", __func__
, max_enough
);
2480 info
->container_enough
= max_enough
;
2483 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
2485 disk
= &super
->disks
->disk
;
2486 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
2487 info
->component_size
= reserved
;
2488 info
->disk
.state
= is_configured(disk
) ? (1 << MD_DISK_ACTIVE
) : 0;
2489 /* we don't change info->disk.raid_disk here because
2490 * this state will be finalized in mdmon after we have
2491 * found the 'most fresh' version of the metadata
2493 info
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2494 info
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2497 /* only call uuid_from_super_imsm when this disk is part of a populated container,
2498 * ->compare_super may have updated the 'num_raid_devs' field for spares
2500 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
2501 uuid_from_super_imsm(st
, info
->uuid
);
2503 memcpy(info
->uuid
, uuid_zero
, sizeof(uuid_zero
));
2505 /* I don't know how to compute 'map' on imsm, so use safe default */
2508 for (i
= 0; i
< map_disks
; i
++)
2514 /* allocates memory and fills disk in mdinfo structure
2515 * for each disk in array */
2516 struct mdinfo
*getinfo_super_disks_imsm(struct supertype
*st
)
2518 struct mdinfo
*mddev
= NULL
;
2519 struct intel_super
*super
= st
->sb
;
2520 struct imsm_disk
*disk
;
2523 if (!super
|| !super
->disks
)
2526 mddev
= malloc(sizeof(*mddev
));
2528 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2531 memset(mddev
, 0, sizeof(*mddev
));
2535 tmp
= malloc(sizeof(*tmp
));
2537 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2542 memset(tmp
, 0, sizeof(*tmp
));
2544 tmp
->next
= mddev
->devs
;
2546 tmp
->disk
.number
= count
++;
2547 tmp
->disk
.major
= dl
->major
;
2548 tmp
->disk
.minor
= dl
->minor
;
2549 tmp
->disk
.state
= is_configured(disk
) ?
2550 (1 << MD_DISK_ACTIVE
) : 0;
2551 tmp
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2552 tmp
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2553 tmp
->disk
.raid_disk
= -1;
2559 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
2560 char *update
, char *devname
, int verbose
,
2561 int uuid_set
, char *homehost
)
2563 /* For 'assemble' and 'force' we need to return non-zero if any
2564 * change was made. For others, the return value is ignored.
2565 * Update options are:
2566 * force-one : This device looks a bit old but needs to be included,
2567 * update age info appropriately.
2568 * assemble: clear any 'faulty' flag to allow this device to
2570 * force-array: Array is degraded but being forced, mark it clean
2571 * if that will be needed to assemble it.
2573 * newdev: not used ????
2574 * grow: Array has gained a new device - this is currently for
2576 * resync: mark as dirty so a resync will happen.
2577 * name: update the name - preserving the homehost
2578 * uuid: Change the uuid of the array to match watch is given
2580 * Following are not relevant for this imsm:
2581 * sparc2.2 : update from old dodgey metadata
2582 * super-minor: change the preferred_minor number
2583 * summaries: update redundant counters.
2584 * homehost: update the recorded homehost
2585 * _reshape_progress: record new reshape_progress position.
2588 struct intel_super
*super
= st
->sb
;
2589 struct imsm_super
*mpb
;
2591 /* we can only update container info */
2592 if (!super
|| super
->current_vol
>= 0 || !super
->anchor
)
2595 mpb
= super
->anchor
;
2597 if (strcmp(update
, "uuid") == 0 && uuid_set
&& !info
->update_private
)
2599 else if (strcmp(update
, "uuid") == 0 && uuid_set
&& info
->update_private
) {
2600 mpb
->orig_family_num
= *((__u32
*) info
->update_private
);
2602 } else if (strcmp(update
, "uuid") == 0) {
2603 __u32
*new_family
= malloc(sizeof(*new_family
));
2605 /* update orig_family_number with the incoming random
2606 * data, report the new effective uuid, and store the
2607 * new orig_family_num for future updates.
2610 memcpy(&mpb
->orig_family_num
, info
->uuid
, sizeof(__u32
));
2611 uuid_from_super_imsm(st
, info
->uuid
);
2612 *new_family
= mpb
->orig_family_num
;
2613 info
->update_private
= new_family
;
2616 } else if (strcmp(update
, "assemble") == 0)
2621 /* successful update? recompute checksum */
2623 mpb
->check_sum
= __le32_to_cpu(__gen_imsm_checksum(mpb
));
2628 static size_t disks_to_mpb_size(int disks
)
2632 size
= sizeof(struct imsm_super
);
2633 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
2634 size
+= 2 * sizeof(struct imsm_dev
);
2635 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
2636 size
+= (4 - 2) * sizeof(struct imsm_map
);
2637 /* 4 possible disk_ord_tbl's */
2638 size
+= 4 * (disks
- 1) * sizeof(__u32
);
2643 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
2645 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
2648 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
2651 static void free_devlist(struct intel_super
*super
)
2653 struct intel_dev
*dv
;
2655 while (super
->devlist
) {
2656 dv
= super
->devlist
->next
;
2657 free(super
->devlist
->dev
);
2658 free(super
->devlist
);
2659 super
->devlist
= dv
;
2663 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
2665 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
2668 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
2672 * 0 same, or first was empty, and second was copied
2673 * 1 second had wrong number
2675 * 3 wrong other info
2677 struct intel_super
*first
= st
->sb
;
2678 struct intel_super
*sec
= tst
->sb
;
2685 /* in platform dependent environment test if the disks
2686 * use the same Intel hba
2688 if (!check_env("IMSM_NO_PLATFORM")) {
2689 if (!first
->hba
|| !sec
->hba
||
2690 (first
->hba
->type
!= sec
->hba
->type
)) {
2692 "HBAs of devices does not match %s != %s\n",
2693 first
->hba
? get_sys_dev_type(first
->hba
->type
) : NULL
,
2694 sec
->hba
? get_sys_dev_type(sec
->hba
->type
) : NULL
);
2699 /* if an anchor does not have num_raid_devs set then it is a free
2702 if (first
->anchor
->num_raid_devs
> 0 &&
2703 sec
->anchor
->num_raid_devs
> 0) {
2704 /* Determine if these disks might ever have been
2705 * related. Further disambiguation can only take place
2706 * in load_super_imsm_all
2708 __u32 first_family
= first
->anchor
->orig_family_num
;
2709 __u32 sec_family
= sec
->anchor
->orig_family_num
;
2711 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
,
2712 MAX_SIGNATURE_LENGTH
) != 0)
2715 if (first_family
== 0)
2716 first_family
= first
->anchor
->family_num
;
2717 if (sec_family
== 0)
2718 sec_family
= sec
->anchor
->family_num
;
2720 if (first_family
!= sec_family
)
2726 /* if 'first' is a spare promote it to a populated mpb with sec's
2729 if (first
->anchor
->num_raid_devs
== 0 &&
2730 sec
->anchor
->num_raid_devs
> 0) {
2732 struct intel_dev
*dv
;
2733 struct imsm_dev
*dev
;
2735 /* we need to copy raid device info from sec if an allocation
2736 * fails here we don't associate the spare
2738 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
2739 dv
= malloc(sizeof(*dv
));
2742 dev
= malloc(sizeof_imsm_dev(get_imsm_dev(sec
, i
), 1));
2749 dv
->next
= first
->devlist
;
2750 first
->devlist
= dv
;
2752 if (i
< sec
->anchor
->num_raid_devs
) {
2753 /* allocation failure */
2754 free_devlist(first
);
2755 fprintf(stderr
, "imsm: failed to associate spare\n");
2758 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
2759 first
->anchor
->orig_family_num
= sec
->anchor
->orig_family_num
;
2760 first
->anchor
->family_num
= sec
->anchor
->family_num
;
2761 memcpy(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
);
2762 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++)
2763 imsm_copy_dev(get_imsm_dev(first
, i
), get_imsm_dev(sec
, i
));
2769 static void fd2devname(int fd
, char *name
)
2773 char dname
[PATH_MAX
];
2778 if (fstat(fd
, &st
) != 0)
2780 sprintf(path
, "/sys/dev/block/%d:%d",
2781 major(st
.st_rdev
), minor(st
.st_rdev
));
2783 rv
= readlink(path
, dname
, sizeof(dname
));
2788 nm
= strrchr(dname
, '/');
2790 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
2793 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
2795 static int imsm_read_serial(int fd
, char *devname
,
2796 __u8 serial
[MAX_RAID_SERIAL_LEN
])
2798 unsigned char scsi_serial
[255];
2807 memset(scsi_serial
, 0, sizeof(scsi_serial
));
2809 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
2811 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
2812 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2813 fd2devname(fd
, (char *) serial
);
2820 Name
": Failed to retrieve serial for %s\n",
2825 rsp_len
= scsi_serial
[3];
2829 Name
": Failed to retrieve serial for %s\n",
2833 rsp_buf
= (char *) &scsi_serial
[4];
2835 /* trim all whitespace and non-printable characters and convert
2838 for (i
= 0, dest
= rsp_buf
; i
< rsp_len
; i
++) {
2841 /* ':' is reserved for use in placeholder serial
2842 * numbers for missing disks
2850 len
= dest
- rsp_buf
;
2853 /* truncate leading characters */
2854 if (len
> MAX_RAID_SERIAL_LEN
) {
2855 dest
+= len
- MAX_RAID_SERIAL_LEN
;
2856 len
= MAX_RAID_SERIAL_LEN
;
2859 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2860 memcpy(serial
, dest
, len
);
2865 static int serialcmp(__u8
*s1
, __u8
*s2
)
2867 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
2870 static void serialcpy(__u8
*dest
, __u8
*src
)
2872 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
2876 static struct dl
*serial_to_dl(__u8
*serial
, struct intel_super
*super
)
2880 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2881 if (serialcmp(dl
->serial
, serial
) == 0)
2888 static struct imsm_disk
*
2889 __serial_to_disk(__u8
*serial
, struct imsm_super
*mpb
, int *idx
)
2893 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2894 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
2896 if (serialcmp(disk
->serial
, serial
) == 0) {
2907 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
2909 struct imsm_disk
*disk
;
2914 __u8 serial
[MAX_RAID_SERIAL_LEN
];
2916 rv
= imsm_read_serial(fd
, devname
, serial
);
2921 dl
= calloc(1, sizeof(*dl
));
2925 Name
": failed to allocate disk buffer for %s\n",
2931 dl
->major
= major(stb
.st_rdev
);
2932 dl
->minor
= minor(stb
.st_rdev
);
2933 dl
->next
= super
->disks
;
2934 dl
->fd
= keep_fd
? fd
: -1;
2935 assert(super
->disks
== NULL
);
2937 serialcpy(dl
->serial
, serial
);
2940 fd2devname(fd
, name
);
2942 dl
->devname
= strdup(devname
);
2944 dl
->devname
= strdup(name
);
2946 /* look up this disk's index in the current anchor */
2947 disk
= __serial_to_disk(dl
->serial
, super
->anchor
, &dl
->index
);
2950 /* only set index on disks that are a member of a
2951 * populated contianer, i.e. one with raid_devs
2953 if (is_failed(&dl
->disk
))
2955 else if (is_spare(&dl
->disk
))
2963 /* When migrating map0 contains the 'destination' state while map1
2964 * contains the current state. When not migrating map0 contains the
2965 * current state. This routine assumes that map[0].map_state is set to
2966 * the current array state before being called.
2968 * Migration is indicated by one of the following states
2969 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
2970 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
2971 * map1state=unitialized)
2972 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
2974 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
2975 * map1state=degraded)
2976 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
2979 static void migrate(struct imsm_dev
*dev
, struct intel_super
*super
,
2980 __u8 to_state
, int migr_type
)
2982 struct imsm_map
*dest
;
2983 struct imsm_map
*src
= get_imsm_map(dev
, 0);
2985 dev
->vol
.migr_state
= 1;
2986 set_migr_type(dev
, migr_type
);
2987 dev
->vol
.curr_migr_unit
= 0;
2988 dest
= get_imsm_map(dev
, 1);
2990 /* duplicate and then set the target end state in map[0] */
2991 memcpy(dest
, src
, sizeof_imsm_map(src
));
2992 if ((migr_type
== MIGR_REBUILD
) ||
2993 (migr_type
== MIGR_GEN_MIGR
)) {
2997 for (i
= 0; i
< src
->num_members
; i
++) {
2998 ord
= __le32_to_cpu(src
->disk_ord_tbl
[i
]);
2999 set_imsm_ord_tbl_ent(src
, i
, ord_to_idx(ord
));
3003 if (migr_type
== MIGR_GEN_MIGR
)
3004 /* Clear migration record */
3005 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
3007 src
->map_state
= to_state
;
3010 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
3012 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3013 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
3016 /* merge any IMSM_ORD_REBUILD bits that were not successfully
3017 * completed in the last migration.
3019 * FIXME add support for raid-level-migration
3021 for (i
= 0; i
< prev
->num_members
; i
++)
3022 for (j
= 0; j
< map
->num_members
; j
++)
3023 /* during online capacity expansion
3024 * disks position can be changed if takeover is used
3026 if (ord_to_idx(map
->disk_ord_tbl
[j
]) ==
3027 ord_to_idx(prev
->disk_ord_tbl
[i
])) {
3028 map
->disk_ord_tbl
[j
] |= prev
->disk_ord_tbl
[i
];
3032 dev
->vol
.migr_state
= 0;
3033 dev
->vol
.migr_type
= 0;
3034 dev
->vol
.curr_migr_unit
= 0;
3035 map
->map_state
= map_state
;
3039 static int parse_raid_devices(struct intel_super
*super
)
3042 struct imsm_dev
*dev_new
;
3043 size_t len
, len_migr
;
3045 size_t space_needed
= 0;
3046 struct imsm_super
*mpb
= super
->anchor
;
3048 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3049 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3050 struct intel_dev
*dv
;
3052 len
= sizeof_imsm_dev(dev_iter
, 0);
3053 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
3055 space_needed
+= len_migr
- len
;
3057 dv
= malloc(sizeof(*dv
));
3060 if (max_len
< len_migr
)
3062 if (max_len
> len_migr
)
3063 space_needed
+= max_len
- len_migr
;
3064 dev_new
= malloc(max_len
);
3069 imsm_copy_dev(dev_new
, dev_iter
);
3072 dv
->next
= super
->devlist
;
3073 super
->devlist
= dv
;
3076 /* ensure that super->buf is large enough when all raid devices
3079 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
3082 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
3083 if (posix_memalign(&buf
, 512, len
) != 0)
3086 memcpy(buf
, super
->buf
, super
->len
);
3087 memset(buf
+ super
->len
, 0, len
- super
->len
);
3096 /* retrieve a pointer to the bbm log which starts after all raid devices */
3097 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
3101 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
3103 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
3109 /*******************************************************************************
3110 * Function: check_mpb_migr_compatibility
3111 * Description: Function checks for unsupported migration features:
3112 * - migration optimization area (pba_of_lba0)
3113 * - descending reshape (ascending_migr)
3115 * super : imsm metadata information
3117 * 0 : migration is compatible
3118 * -1 : migration is not compatible
3119 ******************************************************************************/
3120 int check_mpb_migr_compatibility(struct intel_super
*super
)
3122 struct imsm_map
*map0
, *map1
;
3123 struct migr_record
*migr_rec
= super
->migr_rec
;
3126 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3127 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3130 dev_iter
->vol
.migr_state
== 1 &&
3131 dev_iter
->vol
.migr_type
== MIGR_GEN_MIGR
) {
3132 /* This device is migrating */
3133 map0
= get_imsm_map(dev_iter
, 0);
3134 map1
= get_imsm_map(dev_iter
, 1);
3135 if (map0
->pba_of_lba0
!= map1
->pba_of_lba0
)
3136 /* migration optimization area was used */
3138 if (migr_rec
->ascending_migr
== 0
3139 && migr_rec
->dest_depth_per_unit
> 0)
3140 /* descending reshape not supported yet */
3147 static void __free_imsm(struct intel_super
*super
, int free_disks
);
3149 /* load_imsm_mpb - read matrix metadata
3150 * allocates super->mpb to be freed by free_imsm
3152 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
3154 unsigned long long dsize
;
3155 unsigned long long sectors
;
3157 struct imsm_super
*anchor
;
3160 get_dev_size(fd
, NULL
, &dsize
);
3164 Name
": %s: device to small for imsm\n",
3169 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
3171 fprintf(stderr
, Name
3172 ": Cannot seek to anchor block on %s: %s\n",
3173 devname
, strerror(errno
));
3177 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
3180 Name
": Failed to allocate imsm anchor buffer"
3181 " on %s\n", devname
);
3184 if (read(fd
, anchor
, 512) != 512) {
3187 Name
": Cannot read anchor block on %s: %s\n",
3188 devname
, strerror(errno
));
3193 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
3196 Name
": no IMSM anchor on %s\n", devname
);
3201 __free_imsm(super
, 0);
3202 /* reload capability and hba */
3204 /* capability and hba must be updated with new super allocation */
3205 find_intel_hba_capability(fd
, super
, devname
);
3206 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
3207 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
3210 Name
": unable to allocate %zu byte mpb buffer\n",
3215 memcpy(super
->buf
, anchor
, 512);
3217 sectors
= mpb_sectors(anchor
) - 1;
3220 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
3221 fprintf(stderr
, Name
3222 ": %s could not allocate migr_rec buffer\n", __func__
);
3228 check_sum
= __gen_imsm_checksum(super
->anchor
);
3229 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3232 Name
": IMSM checksum %x != %x on %s\n",
3234 __le32_to_cpu(super
->anchor
->check_sum
),
3242 /* read the extended mpb */
3243 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
3246 Name
": Cannot seek to extended mpb on %s: %s\n",
3247 devname
, strerror(errno
));
3251 if ((unsigned)read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
3254 Name
": Cannot read extended mpb on %s: %s\n",
3255 devname
, strerror(errno
));
3259 check_sum
= __gen_imsm_checksum(super
->anchor
);
3260 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3263 Name
": IMSM checksum %x != %x on %s\n",
3264 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
3269 /* FIXME the BBM log is disk specific so we cannot use this global
3270 * buffer for all disks. Ok for now since we only look at the global
3271 * bbm_log_size parameter to gate assembly
3273 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
3278 static int read_imsm_migr_rec(int fd
, struct intel_super
*super
);
3281 load_and_parse_mpb(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
3285 err
= load_imsm_mpb(fd
, super
, devname
);
3288 err
= load_imsm_disk(fd
, super
, devname
, keep_fd
);
3291 err
= parse_raid_devices(super
);
3296 static void __free_imsm_disk(struct dl
*d
)
3308 static void free_imsm_disks(struct intel_super
*super
)
3312 while (super
->disks
) {
3314 super
->disks
= d
->next
;
3315 __free_imsm_disk(d
);
3317 while (super
->disk_mgmt_list
) {
3318 d
= super
->disk_mgmt_list
;
3319 super
->disk_mgmt_list
= d
->next
;
3320 __free_imsm_disk(d
);
3322 while (super
->missing
) {
3324 super
->missing
= d
->next
;
3325 __free_imsm_disk(d
);
3330 /* free all the pieces hanging off of a super pointer */
3331 static void __free_imsm(struct intel_super
*super
, int free_disks
)
3333 struct intel_hba
*elem
, *next
;
3339 /* unlink capability description */
3341 if (super
->migr_rec_buf
) {
3342 free(super
->migr_rec_buf
);
3343 super
->migr_rec_buf
= NULL
;
3346 free_imsm_disks(super
);
3347 free_devlist(super
);
3351 free((void *)elem
->path
);
3359 static void free_imsm(struct intel_super
*super
)
3361 __free_imsm(super
, 1);
3365 static void free_super_imsm(struct supertype
*st
)
3367 struct intel_super
*super
= st
->sb
;
3376 static struct intel_super
*alloc_super(void)
3378 struct intel_super
*super
= malloc(sizeof(*super
));
3381 memset(super
, 0, sizeof(*super
));
3382 super
->current_vol
= -1;
3383 super
->create_offset
= ~((__u32
) 0);
3389 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
3391 static int find_intel_hba_capability(int fd
, struct intel_super
*super
, char *devname
)
3393 struct sys_dev
*hba_name
;
3396 if ((fd
< 0) || check_env("IMSM_NO_PLATFORM")) {
3401 hba_name
= find_disk_attached_hba(fd
, NULL
);
3405 Name
": %s is not attached to Intel(R) RAID controller.\n",
3409 rv
= attach_hba_to_super(super
, hba_name
);
3412 struct intel_hba
*hba
= super
->hba
;
3414 fprintf(stderr
, Name
": %s is attached to Intel(R) %s RAID "
3415 "controller (%s),\n"
3416 " but the container is assigned to Intel(R) "
3417 "%s RAID controller (",
3420 hba_name
->pci_id
? : "Err!",
3421 get_sys_dev_type(hba_name
->type
));
3424 fprintf(stderr
, "%s", hba
->pci_id
? : "Err!");
3426 fprintf(stderr
, ", ");
3430 fprintf(stderr
, ").\n"
3431 " Mixing devices attached to different controllers "
3432 "is not allowed.\n");
3434 free_sys_dev(&hba_name
);
3437 super
->orom
= find_imsm_capability(hba_name
->type
);
3438 free_sys_dev(&hba_name
);
3445 /* find_missing - helper routine for load_super_imsm_all that identifies
3446 * disks that have disappeared from the system. This routine relies on
3447 * the mpb being uptodate, which it is at load time.
3449 static int find_missing(struct intel_super
*super
)
3452 struct imsm_super
*mpb
= super
->anchor
;
3454 struct imsm_disk
*disk
;
3456 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3457 disk
= __get_imsm_disk(mpb
, i
);
3458 dl
= serial_to_dl(disk
->serial
, super
);
3462 dl
= malloc(sizeof(*dl
));
3468 dl
->devname
= strdup("missing");
3470 serialcpy(dl
->serial
, disk
->serial
);
3473 dl
->next
= super
->missing
;
3474 super
->missing
= dl
;
3480 static struct intel_disk
*disk_list_get(__u8
*serial
, struct intel_disk
*disk_list
)
3482 struct intel_disk
*idisk
= disk_list
;
3485 if (serialcmp(idisk
->disk
.serial
, serial
) == 0)
3487 idisk
= idisk
->next
;
3493 static int __prep_thunderdome(struct intel_super
**table
, int tbl_size
,
3494 struct intel_super
*super
,
3495 struct intel_disk
**disk_list
)
3497 struct imsm_disk
*d
= &super
->disks
->disk
;
3498 struct imsm_super
*mpb
= super
->anchor
;
3501 for (i
= 0; i
< tbl_size
; i
++) {
3502 struct imsm_super
*tbl_mpb
= table
[i
]->anchor
;
3503 struct imsm_disk
*tbl_d
= &table
[i
]->disks
->disk
;
3505 if (tbl_mpb
->family_num
== mpb
->family_num
) {
3506 if (tbl_mpb
->check_sum
== mpb
->check_sum
) {
3507 dprintf("%s: mpb from %d:%d matches %d:%d\n",
3508 __func__
, super
->disks
->major
,
3509 super
->disks
->minor
,
3510 table
[i
]->disks
->major
,
3511 table
[i
]->disks
->minor
);
3515 if (((is_configured(d
) && !is_configured(tbl_d
)) ||
3516 is_configured(d
) == is_configured(tbl_d
)) &&
3517 tbl_mpb
->generation_num
< mpb
->generation_num
) {
3518 /* current version of the mpb is a
3519 * better candidate than the one in
3520 * super_table, but copy over "cross
3521 * generational" status
3523 struct intel_disk
*idisk
;
3525 dprintf("%s: mpb from %d:%d replaces %d:%d\n",
3526 __func__
, super
->disks
->major
,
3527 super
->disks
->minor
,
3528 table
[i
]->disks
->major
,
3529 table
[i
]->disks
->minor
);
3531 idisk
= disk_list_get(tbl_d
->serial
, *disk_list
);
3532 if (idisk
&& is_failed(&idisk
->disk
))
3533 tbl_d
->status
|= FAILED_DISK
;
3536 struct intel_disk
*idisk
;
3537 struct imsm_disk
*disk
;
3539 /* tbl_mpb is more up to date, but copy
3540 * over cross generational status before
3543 disk
= __serial_to_disk(d
->serial
, mpb
, NULL
);
3544 if (disk
&& is_failed(disk
))
3545 d
->status
|= FAILED_DISK
;
3547 idisk
= disk_list_get(d
->serial
, *disk_list
);
3550 if (disk
&& is_configured(disk
))
3551 idisk
->disk
.status
|= CONFIGURED_DISK
;
3554 dprintf("%s: mpb from %d:%d prefer %d:%d\n",
3555 __func__
, super
->disks
->major
,
3556 super
->disks
->minor
,
3557 table
[i
]->disks
->major
,
3558 table
[i
]->disks
->minor
);
3566 table
[tbl_size
++] = super
;
3570 /* update/extend the merged list of imsm_disk records */
3571 for (j
= 0; j
< mpb
->num_disks
; j
++) {
3572 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, j
);
3573 struct intel_disk
*idisk
;
3575 idisk
= disk_list_get(disk
->serial
, *disk_list
);
3577 idisk
->disk
.status
|= disk
->status
;
3578 if (is_configured(&idisk
->disk
) ||
3579 is_failed(&idisk
->disk
))
3580 idisk
->disk
.status
&= ~(SPARE_DISK
);
3582 idisk
= calloc(1, sizeof(*idisk
));
3585 idisk
->owner
= IMSM_UNKNOWN_OWNER
;
3586 idisk
->disk
= *disk
;
3587 idisk
->next
= *disk_list
;
3591 if (serialcmp(idisk
->disk
.serial
, d
->serial
) == 0)
3598 static struct intel_super
*
3599 validate_members(struct intel_super
*super
, struct intel_disk
*disk_list
,
3602 struct imsm_super
*mpb
= super
->anchor
;
3606 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3607 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
3608 struct intel_disk
*idisk
;
3610 idisk
= disk_list_get(disk
->serial
, disk_list
);
3612 if (idisk
->owner
== owner
||
3613 idisk
->owner
== IMSM_UNKNOWN_OWNER
)
3616 dprintf("%s: '%.16s' owner %d != %d\n",
3617 __func__
, disk
->serial
, idisk
->owner
,
3620 dprintf("%s: unknown disk %x [%d]: %.16s\n",
3621 __func__
, __le32_to_cpu(mpb
->family_num
), i
,
3627 if (ok_count
== mpb
->num_disks
)
3632 static void show_conflicts(__u32 family_num
, struct intel_super
*super_list
)
3634 struct intel_super
*s
;
3636 for (s
= super_list
; s
; s
= s
->next
) {
3637 if (family_num
!= s
->anchor
->family_num
)
3639 fprintf(stderr
, "Conflict, offlining family %#x on '%s'\n",
3640 __le32_to_cpu(family_num
), s
->disks
->devname
);
3644 static struct intel_super
*
3645 imsm_thunderdome(struct intel_super
**super_list
, int len
)
3647 struct intel_super
*super_table
[len
];
3648 struct intel_disk
*disk_list
= NULL
;
3649 struct intel_super
*champion
, *spare
;
3650 struct intel_super
*s
, **del
;
3655 memset(super_table
, 0, sizeof(super_table
));
3656 for (s
= *super_list
; s
; s
= s
->next
)
3657 tbl_size
= __prep_thunderdome(super_table
, tbl_size
, s
, &disk_list
);
3659 for (i
= 0; i
< tbl_size
; i
++) {
3660 struct imsm_disk
*d
;
3661 struct intel_disk
*idisk
;
3662 struct imsm_super
*mpb
= super_table
[i
]->anchor
;
3665 d
= &s
->disks
->disk
;
3667 /* 'd' must appear in merged disk list for its
3668 * configuration to be valid
3670 idisk
= disk_list_get(d
->serial
, disk_list
);
3671 if (idisk
&& idisk
->owner
== i
)
3672 s
= validate_members(s
, disk_list
, i
);
3677 dprintf("%s: marking family: %#x from %d:%d offline\n",
3678 __func__
, mpb
->family_num
,
3679 super_table
[i
]->disks
->major
,
3680 super_table
[i
]->disks
->minor
);
3684 /* This is where the mdadm implementation differs from the Windows
3685 * driver which has no strict concept of a container. We can only
3686 * assemble one family from a container, so when returning a prodigal
3687 * array member to this system the code will not be able to disambiguate
3688 * the container contents that should be assembled ("foreign" versus
3689 * "local"). It requires user intervention to set the orig_family_num
3690 * to a new value to establish a new container. The Windows driver in
3691 * this situation fixes up the volume name in place and manages the
3692 * foreign array as an independent entity.
3697 for (i
= 0; i
< tbl_size
; i
++) {
3698 struct intel_super
*tbl_ent
= super_table
[i
];
3704 if (tbl_ent
->anchor
->num_raid_devs
== 0) {
3709 if (s
&& !is_spare
) {
3710 show_conflicts(tbl_ent
->anchor
->family_num
, *super_list
);
3712 } else if (!s
&& !is_spare
)
3725 fprintf(stderr
, "Chose family %#x on '%s', "
3726 "assemble conflicts to new container with '--update=uuid'\n",
3727 __le32_to_cpu(s
->anchor
->family_num
), s
->disks
->devname
);
3729 /* collect all dl's onto 'champion', and update them to
3730 * champion's version of the status
3732 for (s
= *super_list
; s
; s
= s
->next
) {
3733 struct imsm_super
*mpb
= champion
->anchor
;
3734 struct dl
*dl
= s
->disks
;
3739 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3740 struct imsm_disk
*disk
;
3742 disk
= __serial_to_disk(dl
->serial
, mpb
, &dl
->index
);
3745 /* only set index on disks that are a member of
3746 * a populated contianer, i.e. one with
3749 if (is_failed(&dl
->disk
))
3751 else if (is_spare(&dl
->disk
))
3757 if (i
>= mpb
->num_disks
) {
3758 struct intel_disk
*idisk
;
3760 idisk
= disk_list_get(dl
->serial
, disk_list
);
3761 if (idisk
&& is_spare(&idisk
->disk
) &&
3762 !is_failed(&idisk
->disk
) && !is_configured(&idisk
->disk
))
3770 dl
->next
= champion
->disks
;
3771 champion
->disks
= dl
;
3775 /* delete 'champion' from super_list */
3776 for (del
= super_list
; *del
; ) {
3777 if (*del
== champion
) {
3778 *del
= (*del
)->next
;
3781 del
= &(*del
)->next
;
3783 champion
->next
= NULL
;
3787 struct intel_disk
*idisk
= disk_list
;
3789 disk_list
= disk_list
->next
;
3796 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
3800 struct intel_super
*super_list
= NULL
;
3801 struct intel_super
*super
= NULL
;
3802 int devnum
= fd2devnum(fd
);
3808 /* check if 'fd' an opened container */
3809 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
3813 if (sra
->array
.major_version
!= -1 ||
3814 sra
->array
.minor_version
!= -2 ||
3815 strcmp(sra
->text_version
, "imsm") != 0) {
3820 for (sd
= sra
->devs
, i
= 0; sd
; sd
= sd
->next
, i
++) {
3821 struct intel_super
*s
= alloc_super();
3829 s
->next
= super_list
;
3833 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3834 dfd
= dev_open(nm
, O_RDWR
);
3838 rv
= find_intel_hba_capability(dfd
, s
, devname
);
3839 /* no orom/efi or non-intel hba of the disk */
3843 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3845 /* retry the load if we might have raced against mdmon */
3846 if (err
== 3 && mdmon_running(devnum
))
3847 for (retry
= 0; retry
< 3; retry
++) {
3849 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3857 /* all mpbs enter, maybe one leaves */
3858 super
= imsm_thunderdome(&super_list
, i
);
3864 if (find_missing(super
) != 0) {
3870 /* load migration record */
3871 err
= load_imsm_migr_rec(super
, NULL
);
3877 /* Check migration compatibility */
3878 if (check_mpb_migr_compatibility(super
) != 0) {
3879 fprintf(stderr
, Name
": Unsupported migration detected");
3881 fprintf(stderr
, " on %s\n", devname
);
3883 fprintf(stderr
, " (IMSM).\n");
3892 while (super_list
) {
3893 struct intel_super
*s
= super_list
;
3895 super_list
= super_list
->next
;
3904 st
->container_dev
= devnum
;
3905 if (err
== 0 && st
->ss
== NULL
) {
3906 st
->ss
= &super_imsm
;
3907 st
->minor_version
= 0;
3908 st
->max_devs
= IMSM_MAX_DEVICES
;
3913 static int load_container_imsm(struct supertype
*st
, int fd
, char *devname
)
3915 return load_super_imsm_all(st
, fd
, &st
->sb
, devname
);
3919 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
3921 struct intel_super
*super
;
3924 if (test_partition(fd
))
3925 /* IMSM not allowed on partitions */
3928 free_super_imsm(st
);
3930 super
= alloc_super();
3933 Name
": malloc of %zu failed.\n",
3937 /* Load hba and capabilities if they exist.
3938 * But do not preclude loading metadata in case capabilities or hba are
3939 * non-compliant and ignore_hw_compat is set.
3941 rv
= find_intel_hba_capability(fd
, super
, devname
);
3942 /* no orom/efi or non-intel hba of the disk */
3943 if ((rv
!= 0) && (st
->ignore_hw_compat
== 0)) {
3946 Name
": No OROM/EFI properties for %s\n", devname
);
3950 rv
= load_and_parse_mpb(fd
, super
, devname
, 0);
3955 Name
": Failed to load all information "
3956 "sections on %s\n", devname
);
3962 if (st
->ss
== NULL
) {
3963 st
->ss
= &super_imsm
;
3964 st
->minor_version
= 0;
3965 st
->max_devs
= IMSM_MAX_DEVICES
;
3968 /* load migration record */
3969 if (load_imsm_migr_rec(super
, NULL
) == 0) {
3970 /* Check for unsupported migration features */
3971 if (check_mpb_migr_compatibility(super
) != 0) {
3973 Name
": Unsupported migration detected");
3975 fprintf(stderr
, " on %s\n", devname
);
3977 fprintf(stderr
, " (IMSM).\n");
3985 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
3987 if (info
->level
== 1)
3989 return info
->chunk_size
>> 9;
3992 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
, int num_domains
)
3996 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
3997 num_stripes
/= num_domains
;
4002 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
4004 if (info
->level
== 1)
4005 return info
->size
* 2;
4007 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
4010 static void imsm_update_version_info(struct intel_super
*super
)
4012 /* update the version and attributes */
4013 struct imsm_super
*mpb
= super
->anchor
;
4015 struct imsm_dev
*dev
;
4016 struct imsm_map
*map
;
4019 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4020 dev
= get_imsm_dev(super
, i
);
4021 map
= get_imsm_map(dev
, 0);
4022 if (__le32_to_cpu(dev
->size_high
) > 0)
4023 mpb
->attributes
|= MPB_ATTRIB_2TB
;
4025 /* FIXME detect when an array spans a port multiplier */
4027 mpb
->attributes
|= MPB_ATTRIB_PM
;
4030 if (mpb
->num_raid_devs
> 1 ||
4031 mpb
->attributes
!= MPB_ATTRIB_CHECKSUM_VERIFY
) {
4032 version
= MPB_VERSION_ATTRIBS
;
4033 switch (get_imsm_raid_level(map
)) {
4034 case 0: mpb
->attributes
|= MPB_ATTRIB_RAID0
; break;
4035 case 1: mpb
->attributes
|= MPB_ATTRIB_RAID1
; break;
4036 case 10: mpb
->attributes
|= MPB_ATTRIB_RAID10
; break;
4037 case 5: mpb
->attributes
|= MPB_ATTRIB_RAID5
; break;
4040 if (map
->num_members
>= 5)
4041 version
= MPB_VERSION_5OR6_DISK_ARRAY
;
4042 else if (dev
->status
== DEV_CLONE_N_GO
)
4043 version
= MPB_VERSION_CNG
;
4044 else if (get_imsm_raid_level(map
) == 5)
4045 version
= MPB_VERSION_RAID5
;
4046 else if (map
->num_members
>= 3)
4047 version
= MPB_VERSION_3OR4_DISK_ARRAY
;
4048 else if (get_imsm_raid_level(map
) == 1)
4049 version
= MPB_VERSION_RAID1
;
4051 version
= MPB_VERSION_RAID0
;
4053 strcpy(((char *) mpb
->sig
) + strlen(MPB_SIGNATURE
), version
);
4057 static int check_name(struct intel_super
*super
, char *name
, int quiet
)
4059 struct imsm_super
*mpb
= super
->anchor
;
4060 char *reason
= NULL
;
4063 if (strlen(name
) > MAX_RAID_SERIAL_LEN
)
4064 reason
= "must be 16 characters or less";
4066 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4067 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
4069 if (strncmp((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
) == 0) {
4070 reason
= "already exists";
4075 if (reason
&& !quiet
)
4076 fprintf(stderr
, Name
": imsm volume name %s\n", reason
);
4081 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
4082 unsigned long long size
, char *name
,
4083 char *homehost
, int *uuid
)
4085 /* We are creating a volume inside a pre-existing container.
4086 * so st->sb is already set.
4088 struct intel_super
*super
= st
->sb
;
4089 struct imsm_super
*mpb
= super
->anchor
;
4090 struct intel_dev
*dv
;
4091 struct imsm_dev
*dev
;
4092 struct imsm_vol
*vol
;
4093 struct imsm_map
*map
;
4094 int idx
= mpb
->num_raid_devs
;
4096 unsigned long long array_blocks
;
4097 size_t size_old
, size_new
;
4098 __u32 num_data_stripes
;
4100 if (super
->orom
&& mpb
->num_raid_devs
>= super
->orom
->vpa
) {
4101 fprintf(stderr
, Name
": This imsm-container already has the "
4102 "maximum of %d volumes\n", super
->orom
->vpa
);
4106 /* ensure the mpb is large enough for the new data */
4107 size_old
= __le32_to_cpu(mpb
->mpb_size
);
4108 size_new
= disks_to_mpb_size(info
->nr_disks
);
4109 if (size_new
> size_old
) {
4111 size_t size_round
= ROUND_UP(size_new
, 512);
4113 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
4114 fprintf(stderr
, Name
": could not allocate new mpb\n");
4117 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4118 fprintf(stderr
, Name
4119 ": %s could not allocate migr_rec buffer\n",
4125 memcpy(mpb_new
, mpb
, size_old
);
4128 super
->anchor
= mpb_new
;
4129 mpb
->mpb_size
= __cpu_to_le32(size_new
);
4130 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
4132 super
->current_vol
= idx
;
4133 /* when creating the first raid device in this container set num_disks
4134 * to zero, i.e. delete this spare and add raid member devices in
4135 * add_to_super_imsm_volume()
4137 if (super
->current_vol
== 0)
4140 if (!check_name(super
, name
, 0))
4142 dv
= malloc(sizeof(*dv
));
4144 fprintf(stderr
, Name
": failed to allocate device list entry\n");
4147 dev
= calloc(1, sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
4150 fprintf(stderr
, Name
": could not allocate raid device\n");
4154 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
4155 if (info
->level
== 1)
4156 array_blocks
= info_to_blocks_per_member(info
);
4158 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
4159 info
->layout
, info
->chunk_size
,
4161 /* round array size down to closest MB */
4162 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
4164 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
4165 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
4166 dev
->status
= (DEV_READ_COALESCING
| DEV_WRITE_COALESCING
);
4168 vol
->migr_state
= 0;
4169 set_migr_type(dev
, MIGR_INIT
);
4171 vol
->curr_migr_unit
= 0;
4172 map
= get_imsm_map(dev
, 0);
4173 map
->pba_of_lba0
= __cpu_to_le32(super
->create_offset
);
4174 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
4175 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
4176 map
->failed_disk_num
= ~0;
4177 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
4178 IMSM_T_STATE_NORMAL
;
4181 if (info
->level
== 1 && info
->raid_disks
> 2) {
4184 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
4185 "in a raid1 volume\n");
4189 map
->raid_level
= info
->level
;
4190 if (info
->level
== 10) {
4191 map
->raid_level
= 1;
4192 map
->num_domains
= info
->raid_disks
/ 2;
4193 } else if (info
->level
== 1)
4194 map
->num_domains
= info
->raid_disks
;
4196 map
->num_domains
= 1;
4198 num_data_stripes
= info_to_num_data_stripes(info
, map
->num_domains
);
4199 map
->num_data_stripes
= __cpu_to_le32(num_data_stripes
);
4201 map
->num_members
= info
->raid_disks
;
4202 for (i
= 0; i
< map
->num_members
; i
++) {
4203 /* initialized in add_to_super */
4204 set_imsm_ord_tbl_ent(map
, i
, IMSM_ORD_REBUILD
);
4206 mpb
->num_raid_devs
++;
4209 dv
->index
= super
->current_vol
;
4210 dv
->next
= super
->devlist
;
4211 super
->devlist
= dv
;
4213 imsm_update_version_info(super
);
4218 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
4219 unsigned long long size
, char *name
,
4220 char *homehost
, int *uuid
)
4222 /* This is primarily called by Create when creating a new array.
4223 * We will then get add_to_super called for each component, and then
4224 * write_init_super called to write it out to each device.
4225 * For IMSM, Create can create on fresh devices or on a pre-existing
4227 * To create on a pre-existing array a different method will be called.
4228 * This one is just for fresh drives.
4230 struct intel_super
*super
;
4231 struct imsm_super
*mpb
;
4236 return init_super_imsm_volume(st
, info
, size
, name
, homehost
, uuid
);
4239 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
4243 super
= alloc_super();
4244 if (super
&& posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
4249 fprintf(stderr
, Name
4250 ": %s could not allocate superblock\n", __func__
);
4253 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4254 fprintf(stderr
, Name
4255 ": %s could not allocate migr_rec buffer\n", __func__
);
4260 memset(super
->buf
, 0, mpb_size
);
4262 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4266 /* zeroing superblock */
4270 mpb
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4272 version
= (char *) mpb
->sig
;
4273 strcpy(version
, MPB_SIGNATURE
);
4274 version
+= strlen(MPB_SIGNATURE
);
4275 strcpy(version
, MPB_VERSION_RAID0
);
4281 static int add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
4282 int fd
, char *devname
)
4284 struct intel_super
*super
= st
->sb
;
4285 struct imsm_super
*mpb
= super
->anchor
;
4287 struct imsm_dev
*dev
;
4288 struct imsm_map
*map
;
4291 dev
= get_imsm_dev(super
, super
->current_vol
);
4292 map
= get_imsm_map(dev
, 0);
4294 if (! (dk
->state
& (1<<MD_DISK_SYNC
))) {
4295 fprintf(stderr
, Name
": %s: Cannot add spare devices to IMSM volume\n",
4301 /* we're doing autolayout so grab the pre-marked (in
4302 * validate_geometry) raid_disk
4304 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4305 if (dl
->raiddisk
== dk
->raid_disk
)
4308 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4309 if (dl
->major
== dk
->major
&&
4310 dl
->minor
== dk
->minor
)
4315 fprintf(stderr
, Name
": %s is not a member of the same container\n", devname
);
4319 /* add a pristine spare to the metadata */
4320 if (dl
->index
< 0) {
4321 dl
->index
= super
->anchor
->num_disks
;
4322 super
->anchor
->num_disks
++;
4324 /* Check the device has not already been added */
4325 slot
= get_imsm_disk_slot(map
, dl
->index
);
4327 (get_imsm_ord_tbl_ent(dev
, slot
, -1) & IMSM_ORD_REBUILD
) == 0) {
4328 fprintf(stderr
, Name
": %s has been included in this array twice\n",
4332 set_imsm_ord_tbl_ent(map
, dk
->number
, dl
->index
);
4333 dl
->disk
.status
= CONFIGURED_DISK
;
4335 /* if we are creating the first raid device update the family number */
4336 if (super
->current_vol
== 0) {
4338 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
4339 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
4341 if (!_dev
|| !_disk
) {
4342 fprintf(stderr
, Name
": BUG mpb setup error\n");
4348 sum
+= __gen_imsm_checksum(mpb
);
4349 mpb
->family_num
= __cpu_to_le32(sum
);
4350 mpb
->orig_family_num
= mpb
->family_num
;
4357 static int add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
4358 int fd
, char *devname
)
4360 struct intel_super
*super
= st
->sb
;
4362 unsigned long long size
;
4367 /* If we are on an RAID enabled platform check that the disk is
4368 * attached to the raid controller.
4369 * We do not need to test disks attachment for container based additions,
4370 * they shall be already tested when container was created/assembled.
4372 rv
= find_intel_hba_capability(fd
, super
, devname
);
4373 /* no orom/efi or non-intel hba of the disk */
4375 dprintf("capability: %p fd: %d ret: %d\n",
4376 super
->orom
, fd
, rv
);
4380 if (super
->current_vol
>= 0)
4381 return add_to_super_imsm_volume(st
, dk
, fd
, devname
);
4384 dd
= malloc(sizeof(*dd
));
4387 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4390 memset(dd
, 0, sizeof(*dd
));
4391 dd
->major
= major(stb
.st_rdev
);
4392 dd
->minor
= minor(stb
.st_rdev
);
4394 dd
->devname
= devname
? strdup(devname
) : NULL
;
4397 dd
->action
= DISK_ADD
;
4398 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
4401 Name
": failed to retrieve scsi serial, aborting\n");
4406 get_dev_size(fd
, NULL
, &size
);
4408 serialcpy(dd
->disk
.serial
, dd
->serial
);
4409 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
4410 dd
->disk
.status
= SPARE_DISK
;
4411 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
4412 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
4414 dd
->disk
.scsi_id
= __cpu_to_le32(0);
4416 if (st
->update_tail
) {
4417 dd
->next
= super
->disk_mgmt_list
;
4418 super
->disk_mgmt_list
= dd
;
4420 dd
->next
= super
->disks
;
4422 super
->updates_pending
++;
4429 static int remove_from_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
)
4431 struct intel_super
*super
= st
->sb
;
4434 /* remove from super works only in mdmon - for communication
4435 * manager - monitor. Check if communication memory buffer
4438 if (!st
->update_tail
) {
4440 Name
": %s shall be used in mdmon context only"
4441 "(line %d).\n", __func__
, __LINE__
);
4444 dd
= malloc(sizeof(*dd
));
4447 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4450 memset(dd
, 0, sizeof(*dd
));
4451 dd
->major
= dk
->major
;
4452 dd
->minor
= dk
->minor
;
4455 dd
->disk
.status
= SPARE_DISK
;
4456 dd
->action
= DISK_REMOVE
;
4458 dd
->next
= super
->disk_mgmt_list
;
4459 super
->disk_mgmt_list
= dd
;
4465 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
);
4469 struct imsm_super anchor
;
4470 } spare_record
__attribute__ ((aligned(512)));
4472 /* spare records have their own family number and do not have any defined raid
4475 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
4477 struct imsm_super
*mpb
= super
->anchor
;
4478 struct imsm_super
*spare
= &spare_record
.anchor
;
4482 spare
->mpb_size
= __cpu_to_le32(sizeof(struct imsm_super
)),
4483 spare
->generation_num
= __cpu_to_le32(1UL),
4484 spare
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4485 spare
->num_disks
= 1,
4486 spare
->num_raid_devs
= 0,
4487 spare
->cache_size
= mpb
->cache_size
,
4488 spare
->pwr_cycle_count
= __cpu_to_le32(1),
4490 snprintf((char *) spare
->sig
, MAX_SIGNATURE_LENGTH
,
4491 MPB_SIGNATURE MPB_VERSION_RAID0
);
4493 for (d
= super
->disks
; d
; d
= d
->next
) {
4497 spare
->disk
[0] = d
->disk
;
4498 sum
= __gen_imsm_checksum(spare
);
4499 spare
->family_num
= __cpu_to_le32(sum
);
4500 spare
->orig_family_num
= 0;
4501 sum
= __gen_imsm_checksum(spare
);
4502 spare
->check_sum
= __cpu_to_le32(sum
);
4504 if (store_imsm_mpb(d
->fd
, spare
)) {
4505 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4506 __func__
, d
->major
, d
->minor
, strerror(errno
));
4518 static int write_super_imsm(struct supertype
*st
, int doclose
)
4520 struct intel_super
*super
= st
->sb
;
4521 struct imsm_super
*mpb
= super
->anchor
;
4527 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
4529 int clear_migration_record
= 1;
4531 /* 'generation' is incremented everytime the metadata is written */
4532 generation
= __le32_to_cpu(mpb
->generation_num
);
4534 mpb
->generation_num
= __cpu_to_le32(generation
);
4536 /* fix up cases where previous mdadm releases failed to set
4539 if (mpb
->orig_family_num
== 0)
4540 mpb
->orig_family_num
= mpb
->family_num
;
4542 for (d
= super
->disks
; d
; d
= d
->next
) {
4546 mpb
->disk
[d
->index
] = d
->disk
;
4550 for (d
= super
->missing
; d
; d
= d
->next
) {
4551 mpb
->disk
[d
->index
] = d
->disk
;
4554 mpb
->num_disks
= num_disks
;
4555 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
4557 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4558 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
4559 struct imsm_dev
*dev2
= get_imsm_dev(super
, i
);
4561 imsm_copy_dev(dev
, dev2
);
4562 mpb_size
+= sizeof_imsm_dev(dev
, 0);
4564 if (is_gen_migration(dev2
))
4565 clear_migration_record
= 0;
4567 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
4568 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4570 /* recalculate checksum */
4571 sum
= __gen_imsm_checksum(mpb
);
4572 mpb
->check_sum
= __cpu_to_le32(sum
);
4574 if (clear_migration_record
)
4575 memset(super
->migr_rec_buf
, 0, 512);
4577 /* write the mpb for disks that compose raid devices */
4578 for (d
= super
->disks
; d
; d
= d
->next
) {
4581 if (store_imsm_mpb(d
->fd
, mpb
))
4582 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4583 __func__
, d
->major
, d
->minor
, strerror(errno
));
4584 if (clear_migration_record
) {
4585 unsigned long long dsize
;
4587 get_dev_size(d
->fd
, NULL
, &dsize
);
4588 if (lseek64(d
->fd
, dsize
- 512, SEEK_SET
) >= 0) {
4589 if (write(d
->fd
, super
->migr_rec_buf
, 512) != 512)
4590 perror("Write migr_rec failed");
4600 return write_super_imsm_spares(super
, doclose
);
4606 static int create_array(struct supertype
*st
, int dev_idx
)
4609 struct imsm_update_create_array
*u
;
4610 struct intel_super
*super
= st
->sb
;
4611 struct imsm_dev
*dev
= get_imsm_dev(super
, dev_idx
);
4612 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4613 struct disk_info
*inf
;
4614 struct imsm_disk
*disk
;
4617 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0) +
4618 sizeof(*inf
) * map
->num_members
;
4621 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4626 u
->type
= update_create_array
;
4627 u
->dev_idx
= dev_idx
;
4628 imsm_copy_dev(&u
->dev
, dev
);
4629 inf
= get_disk_info(u
);
4630 for (i
= 0; i
< map
->num_members
; i
++) {
4631 int idx
= get_imsm_disk_idx(dev
, i
, -1);
4633 disk
= get_imsm_disk(super
, idx
);
4634 serialcpy(inf
[i
].serial
, disk
->serial
);
4636 append_metadata_update(st
, u
, len
);
4641 static int mgmt_disk(struct supertype
*st
)
4643 struct intel_super
*super
= st
->sb
;
4645 struct imsm_update_add_remove_disk
*u
;
4647 if (!super
->disk_mgmt_list
)
4653 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4658 u
->type
= update_add_remove_disk
;
4659 append_metadata_update(st
, u
, len
);
4664 static int write_init_super_imsm(struct supertype
*st
)
4666 struct intel_super
*super
= st
->sb
;
4667 int current_vol
= super
->current_vol
;
4669 /* we are done with current_vol reset it to point st at the container */
4670 super
->current_vol
= -1;
4672 if (st
->update_tail
) {
4673 /* queue the recently created array / added disk
4674 * as a metadata update */
4677 /* determine if we are creating a volume or adding a disk */
4678 if (current_vol
< 0) {
4679 /* in the mgmt (add/remove) disk case we are running
4680 * in mdmon context, so don't close fd's
4682 return mgmt_disk(st
);
4684 rv
= create_array(st
, current_vol
);
4689 for (d
= super
->disks
; d
; d
= d
->next
)
4690 Kill(d
->devname
, NULL
, 0, 1, 1);
4691 return write_super_imsm(st
, 1);
4696 static int store_super_imsm(struct supertype
*st
, int fd
)
4698 struct intel_super
*super
= st
->sb
;
4699 struct imsm_super
*mpb
= super
? super
->anchor
: NULL
;
4705 return store_imsm_mpb(fd
, mpb
);
4711 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
4713 return __le32_to_cpu(mpb
->bbm_log_size
);
4717 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
4718 int layout
, int raiddisks
, int chunk
,
4719 unsigned long long size
, char *dev
,
4720 unsigned long long *freesize
,
4724 unsigned long long ldsize
;
4725 struct intel_super
*super
=NULL
;
4728 if (level
!= LEVEL_CONTAINER
)
4733 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
4736 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
4737 dev
, strerror(errno
));
4740 if (!get_dev_size(fd
, dev
, &ldsize
)) {
4745 /* capabilities retrieve could be possible
4746 * note that there is no fd for the disks in array.
4748 super
= alloc_super();
4751 Name
": malloc of %zu failed.\n",
4757 rv
= find_intel_hba_capability(fd
, super
, verbose
? dev
: NULL
);
4761 fd2devname(fd
, str
);
4762 dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n",
4763 fd
, str
, super
->orom
, rv
, raiddisks
);
4765 /* no orom/efi or non-intel hba of the disk */
4771 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4773 fprintf(stderr
, Name
": %d exceeds maximum number of"
4774 " platform supported disks: %d\n",
4775 raiddisks
, super
->orom
->tds
);
4781 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
4787 static unsigned long long find_size(struct extent
*e
, int *idx
, int num_extents
)
4789 const unsigned long long base_start
= e
[*idx
].start
;
4790 unsigned long long end
= base_start
+ e
[*idx
].size
;
4793 if (base_start
== end
)
4797 for (i
= *idx
; i
< num_extents
; i
++) {
4798 /* extend overlapping extents */
4799 if (e
[i
].start
>= base_start
&&
4800 e
[i
].start
<= end
) {
4803 if (e
[i
].start
+ e
[i
].size
> end
)
4804 end
= e
[i
].start
+ e
[i
].size
;
4805 } else if (e
[i
].start
> end
) {
4811 return end
- base_start
;
4814 static unsigned long long merge_extents(struct intel_super
*super
, int sum_extents
)
4816 /* build a composite disk with all known extents and generate a new
4817 * 'maxsize' given the "all disks in an array must share a common start
4818 * offset" constraint
4820 struct extent
*e
= calloc(sum_extents
, sizeof(*e
));
4824 unsigned long long pos
;
4825 unsigned long long start
= 0;
4826 unsigned long long maxsize
;
4827 unsigned long reserve
;
4832 /* coalesce and sort all extents. also, check to see if we need to
4833 * reserve space between member arrays
4836 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
4839 for (i
= 0; i
< dl
->extent_cnt
; i
++)
4842 qsort(e
, sum_extents
, sizeof(*e
), cmp_extent
);
4847 while (i
< sum_extents
) {
4848 e
[j
].start
= e
[i
].start
;
4849 e
[j
].size
= find_size(e
, &i
, sum_extents
);
4851 if (e
[j
-1].size
== 0)
4860 unsigned long long esize
;
4862 esize
= e
[i
].start
- pos
;
4863 if (esize
>= maxsize
) {
4868 pos
= e
[i
].start
+ e
[i
].size
;
4870 } while (e
[i
-1].size
);
4876 /* FIXME assumes volume at offset 0 is the first volume in a
4879 if (start_extent
> 0)
4880 reserve
= IMSM_RESERVED_SECTORS
; /* gap between raid regions */
4884 if (maxsize
< reserve
)
4887 super
->create_offset
= ~((__u32
) 0);
4888 if (start
+ reserve
> super
->create_offset
)
4889 return 0; /* start overflows create_offset */
4890 super
->create_offset
= start
+ reserve
;
4892 return maxsize
- reserve
;
4895 static int is_raid_level_supported(const struct imsm_orom
*orom
, int level
, int raiddisks
)
4897 if (level
< 0 || level
== 6 || level
== 4)
4900 /* if we have an orom prevent invalid raid levels */
4903 case 0: return imsm_orom_has_raid0(orom
);
4906 return imsm_orom_has_raid1e(orom
);
4907 return imsm_orom_has_raid1(orom
) && raiddisks
== 2;
4908 case 10: return imsm_orom_has_raid10(orom
) && raiddisks
== 4;
4909 case 5: return imsm_orom_has_raid5(orom
) && raiddisks
> 2;
4912 return 1; /* not on an Intel RAID platform so anything goes */
4918 #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
4920 * validate volume parameters with OROM/EFI capabilities
4923 validate_geometry_imsm_orom(struct intel_super
*super
, int level
, int layout
,
4924 int raiddisks
, int *chunk
, int verbose
)
4929 /* validate container capabilities */
4930 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4932 fprintf(stderr
, Name
": %d exceeds maximum number of"
4933 " platform supported disks: %d\n",
4934 raiddisks
, super
->orom
->tds
);
4938 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
4939 if (super
->orom
&& (!is_raid_level_supported(super
->orom
, level
,
4941 pr_vrb(": platform does not support raid%d with %d disk%s\n",
4942 level
, raiddisks
, raiddisks
> 1 ? "s" : "");
4945 if (super
->orom
&& level
!= 1) {
4946 if (chunk
&& (*chunk
== 0 || *chunk
== UnSet
))
4947 *chunk
= imsm_orom_default_chunk(super
->orom
);
4948 else if (chunk
&& !imsm_orom_has_chunk(super
->orom
, *chunk
)) {
4949 pr_vrb(": platform does not support a chunk size of: "
4954 if (layout
!= imsm_level_to_layout(level
)) {
4956 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
4957 else if (level
== 10)
4958 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
4960 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
4967 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
4968 * FIX ME add ahci details
4970 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
4971 int layout
, int raiddisks
, int *chunk
,
4972 unsigned long long size
, char *dev
,
4973 unsigned long long *freesize
,
4977 struct intel_super
*super
= st
->sb
;
4978 struct imsm_super
*mpb
= super
->anchor
;
4980 unsigned long long pos
= 0;
4981 unsigned long long maxsize
;
4985 /* We must have the container info already read in. */
4989 if (!validate_geometry_imsm_orom(super
, level
, layout
, raiddisks
, chunk
, verbose
)) {
4990 fprintf(stderr
, Name
": RAID gemetry validation failed. "
4991 "Cannot proceed with the action(s).\n");
4995 /* General test: make sure there is space for
4996 * 'raiddisks' device extents of size 'size' at a given
4999 unsigned long long minsize
= size
;
5000 unsigned long long start_offset
= MaxSector
;
5003 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
5004 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5009 e
= get_extents(super
, dl
);
5012 unsigned long long esize
;
5013 esize
= e
[i
].start
- pos
;
5014 if (esize
>= minsize
)
5016 if (found
&& start_offset
== MaxSector
) {
5019 } else if (found
&& pos
!= start_offset
) {
5023 pos
= e
[i
].start
+ e
[i
].size
;
5025 } while (e
[i
-1].size
);
5030 if (dcnt
< raiddisks
) {
5032 fprintf(stderr
, Name
": imsm: Not enough "
5033 "devices with space for this array "
5041 /* This device must be a member of the set */
5042 if (stat(dev
, &stb
) < 0)
5044 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
5046 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5047 if (dl
->major
== (int)major(stb
.st_rdev
) &&
5048 dl
->minor
== (int)minor(stb
.st_rdev
))
5053 fprintf(stderr
, Name
": %s is not in the "
5054 "same imsm set\n", dev
);
5056 } else if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
) {
5057 /* If a volume is present then the current creation attempt
5058 * cannot incorporate new spares because the orom may not
5059 * understand this configuration (all member disks must be
5060 * members of each array in the container).
5062 fprintf(stderr
, Name
": %s is a spare and a volume"
5063 " is already defined for this container\n", dev
);
5064 fprintf(stderr
, Name
": The option-rom requires all member"
5065 " disks to be a member of all volumes\n");
5069 /* retrieve the largest free space block */
5070 e
= get_extents(super
, dl
);
5075 unsigned long long esize
;
5077 esize
= e
[i
].start
- pos
;
5078 if (esize
>= maxsize
)
5080 pos
= e
[i
].start
+ e
[i
].size
;
5082 } while (e
[i
-1].size
);
5087 fprintf(stderr
, Name
": unable to determine free space for: %s\n",
5091 if (maxsize
< size
) {
5093 fprintf(stderr
, Name
": %s not enough space (%llu < %llu)\n",
5094 dev
, maxsize
, size
);
5098 /* count total number of extents for merge */
5100 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5102 i
+= dl
->extent_cnt
;
5104 maxsize
= merge_extents(super
, i
);
5105 if (maxsize
< size
|| maxsize
== 0) {
5107 fprintf(stderr
, Name
": not enough space after merge (%llu < %llu)\n",
5112 *freesize
= maxsize
;
5117 static int reserve_space(struct supertype
*st
, int raiddisks
,
5118 unsigned long long size
, int chunk
,
5119 unsigned long long *freesize
)
5121 struct intel_super
*super
= st
->sb
;
5122 struct imsm_super
*mpb
= super
->anchor
;
5127 unsigned long long maxsize
;
5128 unsigned long long minsize
;
5132 /* find the largest common start free region of the possible disks */
5136 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5142 /* don't activate new spares if we are orom constrained
5143 * and there is already a volume active in the container
5145 if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
)
5148 e
= get_extents(super
, dl
);
5151 for (i
= 1; e
[i
-1].size
; i
++)
5159 maxsize
= merge_extents(super
, extent_cnt
);
5163 minsize
= chunk
* 2;
5165 if (cnt
< raiddisks
||
5166 (super
->orom
&& used
&& used
!= raiddisks
) ||
5167 maxsize
< minsize
||
5169 fprintf(stderr
, Name
": not enough devices with space to create array.\n");
5170 return 0; /* No enough free spaces large enough */
5182 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5184 dl
->raiddisk
= cnt
++;
5191 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
5192 int raiddisks
, int *chunk
, unsigned long long size
,
5193 char *dev
, unsigned long long *freesize
,
5201 * if given unused devices create a container
5202 * if given given devices in a container create a member volume
5204 if (level
== LEVEL_CONTAINER
) {
5205 /* Must be a fresh device to add to a container */
5206 return validate_geometry_imsm_container(st
, level
, layout
,
5208 chunk
?*chunk
:0, size
,
5214 if (st
->sb
&& freesize
) {
5215 /* we are being asked to automatically layout a
5216 * new volume based on the current contents of
5217 * the container. If the the parameters can be
5218 * satisfied reserve_space will record the disks,
5219 * start offset, and size of the volume to be
5220 * created. add_to_super and getinfo_super
5221 * detect when autolayout is in progress.
5223 if (!validate_geometry_imsm_orom(st
->sb
, level
, layout
,
5227 return reserve_space(st
, raiddisks
, size
,
5228 chunk
?*chunk
:0, freesize
);
5233 /* creating in a given container */
5234 return validate_geometry_imsm_volume(st
, level
, layout
,
5235 raiddisks
, chunk
, size
,
5236 dev
, freesize
, verbose
);
5239 /* This device needs to be a device in an 'imsm' container */
5240 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
5244 Name
": Cannot create this array on device %s\n",
5249 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
5251 fprintf(stderr
, Name
": Cannot open %s: %s\n",
5252 dev
, strerror(errno
));
5255 /* Well, it is in use by someone, maybe an 'imsm' container. */
5256 cfd
= open_container(fd
);
5260 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
5264 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
5265 if (sra
&& sra
->array
.major_version
== -1 &&
5266 strcmp(sra
->text_version
, "imsm") == 0)
5270 /* This is a member of a imsm container. Load the container
5271 * and try to create a volume
5273 struct intel_super
*super
;
5275 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
) == 0) {
5277 st
->container_dev
= fd2devnum(cfd
);
5279 return validate_geometry_imsm_volume(st
, level
, layout
,
5287 fprintf(stderr
, Name
": failed container membership check\n");
5293 static void default_geometry_imsm(struct supertype
*st
, int *level
, int *layout
, int *chunk
)
5295 struct intel_super
*super
= st
->sb
;
5297 if (level
&& *level
== UnSet
)
5298 *level
= LEVEL_CONTAINER
;
5300 if (level
&& layout
&& *layout
== UnSet
)
5301 *layout
= imsm_level_to_layout(*level
);
5303 if (chunk
&& (*chunk
== UnSet
|| *chunk
== 0) &&
5304 super
&& super
->orom
)
5305 *chunk
= imsm_orom_default_chunk(super
->orom
);
5308 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
);
5310 static int kill_subarray_imsm(struct supertype
*st
)
5312 /* remove the subarray currently referenced by ->current_vol */
5314 struct intel_dev
**dp
;
5315 struct intel_super
*super
= st
->sb
;
5316 __u8 current_vol
= super
->current_vol
;
5317 struct imsm_super
*mpb
= super
->anchor
;
5319 if (super
->current_vol
< 0)
5321 super
->current_vol
= -1; /* invalidate subarray cursor */
5323 /* block deletions that would change the uuid of active subarrays
5325 * FIXME when immutable ids are available, but note that we'll
5326 * also need to fixup the invalidated/active subarray indexes in
5329 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5332 if (i
< current_vol
)
5334 sprintf(subarray
, "%u", i
);
5335 if (is_subarray_active(subarray
, st
->devname
)) {
5337 Name
": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
5344 if (st
->update_tail
) {
5345 struct imsm_update_kill_array
*u
= malloc(sizeof(*u
));
5349 u
->type
= update_kill_array
;
5350 u
->dev_idx
= current_vol
;
5351 append_metadata_update(st
, u
, sizeof(*u
));
5356 for (dp
= &super
->devlist
; *dp
;)
5357 if ((*dp
)->index
== current_vol
) {
5360 handle_missing(super
, (*dp
)->dev
);
5361 if ((*dp
)->index
> current_vol
)
5366 /* no more raid devices, all active components are now spares,
5367 * but of course failed are still failed
5369 if (--mpb
->num_raid_devs
== 0) {
5372 for (d
= super
->disks
; d
; d
= d
->next
)
5373 if (d
->index
> -2) {
5375 d
->disk
.status
= SPARE_DISK
;
5379 super
->updates_pending
++;
5384 static int update_subarray_imsm(struct supertype
*st
, char *subarray
,
5385 char *update
, struct mddev_ident
*ident
)
5387 /* update the subarray currently referenced by ->current_vol */
5388 struct intel_super
*super
= st
->sb
;
5389 struct imsm_super
*mpb
= super
->anchor
;
5391 if (strcmp(update
, "name") == 0) {
5392 char *name
= ident
->name
;
5396 if (is_subarray_active(subarray
, st
->devname
)) {
5398 Name
": Unable to update name of active subarray\n");
5402 if (!check_name(super
, name
, 0))
5405 vol
= strtoul(subarray
, &ep
, 10);
5406 if (*ep
!= '\0' || vol
>= super
->anchor
->num_raid_devs
)
5409 if (st
->update_tail
) {
5410 struct imsm_update_rename_array
*u
= malloc(sizeof(*u
));
5414 u
->type
= update_rename_array
;
5416 snprintf((char *) u
->name
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5417 append_metadata_update(st
, u
, sizeof(*u
));
5419 struct imsm_dev
*dev
;
5422 dev
= get_imsm_dev(super
, vol
);
5423 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5424 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5425 dev
= get_imsm_dev(super
, i
);
5426 handle_missing(super
, dev
);
5428 super
->updates_pending
++;
5436 static int is_gen_migration(struct imsm_dev
*dev
)
5441 if (!dev
->vol
.migr_state
)
5444 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5449 #endif /* MDASSEMBLE */
5451 static int is_rebuilding(struct imsm_dev
*dev
)
5453 struct imsm_map
*migr_map
;
5455 if (!dev
->vol
.migr_state
)
5458 if (migr_type(dev
) != MIGR_REBUILD
)
5461 migr_map
= get_imsm_map(dev
, 1);
5463 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
5469 static void update_recovery_start(struct intel_super
*super
,
5470 struct imsm_dev
*dev
,
5471 struct mdinfo
*array
)
5473 struct mdinfo
*rebuild
= NULL
;
5477 if (!is_rebuilding(dev
))
5480 /* Find the rebuild target, but punt on the dual rebuild case */
5481 for (d
= array
->devs
; d
; d
= d
->next
)
5482 if (d
->recovery_start
== 0) {
5489 /* (?) none of the disks are marked with
5490 * IMSM_ORD_REBUILD, so assume they are missing and the
5491 * disk_ord_tbl was not correctly updated
5493 dprintf("%s: failed to locate out-of-sync disk\n", __func__
);
5497 units
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
5498 rebuild
->recovery_start
= units
* blocks_per_migr_unit(super
, dev
);
5502 static int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
);
5505 static struct mdinfo
*container_content_imsm(struct supertype
*st
, char *subarray
)
5507 /* Given a container loaded by load_super_imsm_all,
5508 * extract information about all the arrays into
5510 * If 'subarray' is given, just extract info about that array.
5512 * For each imsm_dev create an mdinfo, fill it in,
5513 * then look for matching devices in super->disks
5514 * and create appropriate device mdinfo.
5516 struct intel_super
*super
= st
->sb
;
5517 struct imsm_super
*mpb
= super
->anchor
;
5518 struct mdinfo
*rest
= NULL
;
5522 int spare_disks
= 0;
5524 /* do not assemble arrays when not all attributes are supported */
5525 if (imsm_check_attributes(mpb
->attributes
) == 0) {
5526 fprintf(stderr
, Name
": IMSM metadata loading not allowed "
5527 "due to attributes incompatibility.\n");
5531 /* check for bad blocks */
5532 if (imsm_bbm_log_size(super
->anchor
))
5535 /* count spare devices, not used in maps
5537 for (d
= super
->disks
; d
; d
= d
->next
)
5541 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5542 struct imsm_dev
*dev
;
5543 struct imsm_map
*map
;
5544 struct imsm_map
*map2
;
5545 struct mdinfo
*this;
5550 (i
!= strtoul(subarray
, &ep
, 10) || *ep
!= '\0'))
5553 dev
= get_imsm_dev(super
, i
);
5554 map
= get_imsm_map(dev
, 0);
5555 map2
= get_imsm_map(dev
, 1);
5557 /* do not publish arrays that are in the middle of an
5558 * unsupported migration
5560 if (dev
->vol
.migr_state
&&
5561 (migr_type(dev
) == MIGR_STATE_CHANGE
)) {
5562 fprintf(stderr
, Name
": cannot assemble volume '%.16s':"
5563 " unsupported migration in progress\n",
5567 /* do not publish arrays that are not support by controller's
5571 chunk
= __le16_to_cpu(map
->blocks_per_strip
) >> 1;
5573 if (!validate_geometry_imsm_orom(super
,
5574 get_imsm_raid_level(map
), /* RAID level */
5575 imsm_level_to_layout(get_imsm_raid_level(map
)),
5576 map
->num_members
, /* raid disks */
5579 fprintf(stderr
, Name
": RAID gemetry validation failed. "
5580 "Cannot proceed with the action(s).\n");
5583 #endif /* MDASSEMBLE */
5584 this = malloc(sizeof(*this));
5586 fprintf(stderr
, Name
": failed to allocate %zu bytes\n",
5591 super
->current_vol
= i
;
5592 getinfo_super_imsm_volume(st
, this, NULL
);
5594 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
5595 unsigned long long recovery_start
;
5596 struct mdinfo
*info_d
;
5603 idx
= get_imsm_disk_idx(dev
, slot
, 0);
5604 ord
= get_imsm_ord_tbl_ent(dev
, slot
, -1);
5605 for (d
= super
->disks
; d
; d
= d
->next
)
5606 if (d
->index
== idx
)
5609 recovery_start
= MaxSector
;
5612 if (d
&& is_failed(&d
->disk
))
5614 if (ord
& IMSM_ORD_REBUILD
)
5618 * if we skip some disks the array will be assmebled degraded;
5619 * reset resync start to avoid a dirty-degraded
5620 * situation when performing the intial sync
5622 * FIXME handle dirty degraded
5624 if ((skip
|| recovery_start
== 0) && !dev
->vol
.dirty
)
5625 this->resync_start
= MaxSector
;
5629 info_d
= calloc(1, sizeof(*info_d
));
5631 fprintf(stderr
, Name
": failed to allocate disk"
5632 " for volume %.16s\n", dev
->volume
);
5633 info_d
= this->devs
;
5635 struct mdinfo
*d
= info_d
->next
;
5644 info_d
->next
= this->devs
;
5645 this->devs
= info_d
;
5647 info_d
->disk
.number
= d
->index
;
5648 info_d
->disk
.major
= d
->major
;
5649 info_d
->disk
.minor
= d
->minor
;
5650 info_d
->disk
.raid_disk
= slot
;
5651 info_d
->recovery_start
= recovery_start
;
5653 if (slot
< map2
->num_members
)
5654 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5656 this->array
.spare_disks
++;
5658 if (slot
< map
->num_members
)
5659 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5661 this->array
.spare_disks
++;
5663 if (info_d
->recovery_start
== MaxSector
)
5664 this->array
.working_disks
++;
5666 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
5667 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
5668 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
5670 /* now that the disk list is up-to-date fixup recovery_start */
5671 update_recovery_start(super
, dev
, this);
5672 this->array
.spare_disks
+= spare_disks
;
5675 /* check for reshape */
5676 if (this->reshape_active
== 1)
5677 recover_backup_imsm(st
, this);
5682 /* if array has bad blocks, set suitable bit in array status */
5684 rest
->array
.state
|= (1<<MD_SB_BBM_ERRORS
);
5690 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
5692 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5695 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
5696 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
5698 switch (get_imsm_raid_level(map
)) {
5700 return IMSM_T_STATE_FAILED
;
5703 if (failed
< map
->num_members
)
5704 return IMSM_T_STATE_DEGRADED
;
5706 return IMSM_T_STATE_FAILED
;
5711 * check to see if any mirrors have failed, otherwise we
5712 * are degraded. Even numbered slots are mirrored on
5716 /* gcc -Os complains that this is unused */
5717 int insync
= insync
;
5719 for (i
= 0; i
< map
->num_members
; i
++) {
5720 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
5721 int idx
= ord_to_idx(ord
);
5722 struct imsm_disk
*disk
;
5724 /* reset the potential in-sync count on even-numbered
5725 * slots. num_copies is always 2 for imsm raid10
5730 disk
= get_imsm_disk(super
, idx
);
5731 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5734 /* no in-sync disks left in this mirror the
5738 return IMSM_T_STATE_FAILED
;
5741 return IMSM_T_STATE_DEGRADED
;
5745 return IMSM_T_STATE_DEGRADED
;
5747 return IMSM_T_STATE_FAILED
;
5753 return map
->map_state
;
5756 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
5760 struct imsm_disk
*disk
;
5761 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5762 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
5766 /* at the beginning of migration we set IMSM_ORD_REBUILD on
5767 * disks that are being rebuilt. New failures are recorded to
5768 * map[0]. So we look through all the disks we started with and
5769 * see if any failures are still present, or if any new ones
5772 * FIXME add support for online capacity expansion and
5773 * raid-level-migration
5775 for (i
= 0; i
< prev
->num_members
; i
++) {
5776 ord
= __le32_to_cpu(prev
->disk_ord_tbl
[i
]);
5777 ord
|= __le32_to_cpu(map
->disk_ord_tbl
[i
]);
5778 idx
= ord_to_idx(ord
);
5780 disk
= get_imsm_disk(super
, idx
);
5781 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5789 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
5792 struct intel_super
*super
= c
->sb
;
5793 struct imsm_super
*mpb
= super
->anchor
;
5795 if (atoi(inst
) >= mpb
->num_raid_devs
) {
5796 fprintf(stderr
, "%s: subarry index %d, out of range\n",
5797 __func__
, atoi(inst
));
5801 dprintf("imsm: open_new %s\n", inst
);
5802 a
->info
.container_member
= atoi(inst
);
5806 static int is_resyncing(struct imsm_dev
*dev
)
5808 struct imsm_map
*migr_map
;
5810 if (!dev
->vol
.migr_state
)
5813 if (migr_type(dev
) == MIGR_INIT
||
5814 migr_type(dev
) == MIGR_REPAIR
)
5817 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5820 migr_map
= get_imsm_map(dev
, 1);
5822 if ((migr_map
->map_state
== IMSM_T_STATE_NORMAL
) &&
5823 (dev
->vol
.migr_type
!= MIGR_GEN_MIGR
))
5829 /* return true if we recorded new information */
5830 static int mark_failure(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5834 struct imsm_map
*map
;
5836 /* new failures are always set in map[0] */
5837 map
= get_imsm_map(dev
, 0);
5839 slot
= get_imsm_disk_slot(map
, idx
);
5843 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
5844 if (is_failed(disk
) && (ord
& IMSM_ORD_REBUILD
))
5847 disk
->status
|= FAILED_DISK
;
5848 set_imsm_ord_tbl_ent(map
, slot
, idx
| IMSM_ORD_REBUILD
);
5849 if (map
->failed_disk_num
== 0xff)
5850 map
->failed_disk_num
= slot
;
5854 static void mark_missing(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5856 mark_failure(dev
, disk
, idx
);
5858 if (disk
->scsi_id
== __cpu_to_le32(~(__u32
)0))
5861 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
5862 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
5865 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
)
5871 if (!super
->missing
)
5873 failed
= imsm_count_failed(super
, dev
);
5874 map_state
= imsm_check_degraded(super
, dev
, failed
);
5876 dprintf("imsm: mark missing\n");
5877 end_migration(dev
, map_state
);
5878 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
5879 mark_missing(dev
, &dl
->disk
, dl
->index
);
5880 super
->updates_pending
++;
5883 static unsigned long long imsm_set_array_size(struct imsm_dev
*dev
)
5885 int used_disks
= imsm_num_data_members(dev
, 0);
5886 unsigned long long array_blocks
;
5887 struct imsm_map
*map
;
5889 if (used_disks
== 0) {
5890 /* when problems occures
5891 * return current array_blocks value
5893 array_blocks
= __le32_to_cpu(dev
->size_high
);
5894 array_blocks
= array_blocks
<< 32;
5895 array_blocks
+= __le32_to_cpu(dev
->size_low
);
5897 return array_blocks
;
5900 /* set array size in metadata
5902 map
= get_imsm_map(dev
, 0);
5903 array_blocks
= map
->blocks_per_member
* used_disks
;
5905 /* round array size down to closest MB
5907 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
5908 dev
->size_low
= __cpu_to_le32((__u32
)array_blocks
);
5909 dev
->size_high
= __cpu_to_le32((__u32
)(array_blocks
>> 32));
5911 return array_blocks
;
5914 static void imsm_set_disk(struct active_array
*a
, int n
, int state
);
5916 static void imsm_progress_container_reshape(struct intel_super
*super
)
5918 /* if no device has a migr_state, but some device has a
5919 * different number of members than the previous device, start
5920 * changing the number of devices in this device to match
5923 struct imsm_super
*mpb
= super
->anchor
;
5924 int prev_disks
= -1;
5928 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5929 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
5930 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5931 struct imsm_map
*map2
;
5932 int prev_num_members
;
5934 if (dev
->vol
.migr_state
)
5937 if (prev_disks
== -1)
5938 prev_disks
= map
->num_members
;
5939 if (prev_disks
== map
->num_members
)
5942 /* OK, this array needs to enter reshape mode.
5943 * i.e it needs a migr_state
5946 copy_map_size
= sizeof_imsm_map(map
);
5947 prev_num_members
= map
->num_members
;
5948 map
->num_members
= prev_disks
;
5949 dev
->vol
.migr_state
= 1;
5950 dev
->vol
.curr_migr_unit
= 0;
5951 dev
->vol
.migr_type
= MIGR_GEN_MIGR
;
5952 for (i
= prev_num_members
;
5953 i
< map
->num_members
; i
++)
5954 set_imsm_ord_tbl_ent(map
, i
, i
);
5955 map2
= get_imsm_map(dev
, 1);
5956 /* Copy the current map */
5957 memcpy(map2
, map
, copy_map_size
);
5958 map2
->num_members
= prev_num_members
;
5960 imsm_set_array_size(dev
);
5961 super
->updates_pending
++;
5965 /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
5966 * states are handled in imsm_set_disk() with one exception, when a
5967 * resync is stopped due to a new failure this routine will set the
5968 * 'degraded' state for the array.
5970 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
5972 int inst
= a
->info
.container_member
;
5973 struct intel_super
*super
= a
->container
->sb
;
5974 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
5975 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5976 int failed
= imsm_count_failed(super
, dev
);
5977 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
5978 __u32 blocks_per_unit
;
5980 if (dev
->vol
.migr_state
&&
5981 dev
->vol
.migr_type
== MIGR_GEN_MIGR
) {
5982 /* array state change is blocked due to reshape action
5984 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
5985 * - finish the reshape (if last_checkpoint is big and action != reshape)
5986 * - update curr_migr_unit
5988 if (a
->curr_action
== reshape
) {
5989 /* still reshaping, maybe update curr_migr_unit */
5990 goto mark_checkpoint
;
5992 if (a
->last_checkpoint
== 0 && a
->prev_action
== reshape
) {
5993 /* for some reason we aborted the reshape.
5995 * disable automatic metadata rollback
5996 * user action is required to recover process
5999 struct imsm_map
*map2
= get_imsm_map(dev
, 1);
6000 dev
->vol
.migr_state
= 0;
6001 dev
->vol
.migr_type
= 0;
6002 dev
->vol
.curr_migr_unit
= 0;
6003 memcpy(map
, map2
, sizeof_imsm_map(map2
));
6004 super
->updates_pending
++;
6007 if (a
->last_checkpoint
>= a
->info
.component_size
) {
6008 unsigned long long array_blocks
;
6012 used_disks
= imsm_num_data_members(dev
, 0);
6013 if (used_disks
> 0) {
6015 map
->blocks_per_member
*
6017 /* round array size down to closest MB
6019 array_blocks
= (array_blocks
6020 >> SECT_PER_MB_SHIFT
)
6021 << SECT_PER_MB_SHIFT
;
6022 a
->info
.custom_array_size
= array_blocks
;
6023 /* encourage manager to update array
6027 a
->check_reshape
= 1;
6029 /* finalize online capacity expansion/reshape */
6030 for (mdi
= a
->info
.devs
; mdi
; mdi
= mdi
->next
)
6032 mdi
->disk
.raid_disk
,
6035 imsm_progress_container_reshape(super
);
6040 /* before we activate this array handle any missing disks */
6041 if (consistent
== 2)
6042 handle_missing(super
, dev
);
6044 if (consistent
== 2 &&
6045 (!is_resync_complete(&a
->info
) ||
6046 map_state
!= IMSM_T_STATE_NORMAL
||
6047 dev
->vol
.migr_state
))
6050 if (is_resync_complete(&a
->info
)) {
6051 /* complete intialization / resync,
6052 * recovery and interrupted recovery is completed in
6055 if (is_resyncing(dev
)) {
6056 dprintf("imsm: mark resync done\n");
6057 end_migration(dev
, map_state
);
6058 super
->updates_pending
++;
6059 a
->last_checkpoint
= 0;
6061 } else if (!is_resyncing(dev
) && !failed
) {
6062 /* mark the start of the init process if nothing is failed */
6063 dprintf("imsm: mark resync start\n");
6064 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
)
6065 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_INIT
);
6067 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_REPAIR
);
6068 super
->updates_pending
++;
6072 /* skip checkpointing for general migration,
6073 * it is controlled in mdadm
6075 if (is_gen_migration(dev
))
6076 goto skip_mark_checkpoint
;
6078 /* check if we can update curr_migr_unit from resync_start, recovery_start */
6079 blocks_per_unit
= blocks_per_migr_unit(super
, dev
);
6080 if (blocks_per_unit
) {
6084 units
= a
->last_checkpoint
/ blocks_per_unit
;
6087 /* check that we did not overflow 32-bits, and that
6088 * curr_migr_unit needs updating
6090 if (units32
== units
&&
6092 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != units32
) {
6093 dprintf("imsm: mark checkpoint (%u)\n", units32
);
6094 dev
->vol
.curr_migr_unit
= __cpu_to_le32(units32
);
6095 super
->updates_pending
++;
6099 skip_mark_checkpoint
:
6100 /* mark dirty / clean */
6101 if (dev
->vol
.dirty
!= !consistent
) {
6102 dprintf("imsm: mark '%s'\n", consistent
? "clean" : "dirty");
6107 super
->updates_pending
++;
6113 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
6115 int inst
= a
->info
.container_member
;
6116 struct intel_super
*super
= a
->container
->sb
;
6117 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6118 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6119 struct imsm_disk
*disk
;
6124 if (n
> map
->num_members
)
6125 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
6126 n
, map
->num_members
- 1);
6131 dprintf("imsm: set_disk %d:%x\n", n
, state
);
6133 ord
= get_imsm_ord_tbl_ent(dev
, n
, -1);
6134 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
6136 /* check for new failures */
6137 if (state
& DS_FAULTY
) {
6138 if (mark_failure(dev
, disk
, ord_to_idx(ord
)))
6139 super
->updates_pending
++;
6142 /* check if in_sync */
6143 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
&& is_rebuilding(dev
)) {
6144 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6146 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
6147 super
->updates_pending
++;
6150 failed
= imsm_count_failed(super
, dev
);
6151 map_state
= imsm_check_degraded(super
, dev
, failed
);
6153 /* check if recovery complete, newly degraded, or failed */
6154 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
6155 end_migration(dev
, map_state
);
6156 map
= get_imsm_map(dev
, 0);
6157 map
->failed_disk_num
= ~0;
6158 super
->updates_pending
++;
6159 a
->last_checkpoint
= 0;
6160 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
6161 map
->map_state
!= map_state
&&
6162 !dev
->vol
.migr_state
) {
6163 dprintf("imsm: mark degraded\n");
6164 map
->map_state
= map_state
;
6165 super
->updates_pending
++;
6166 a
->last_checkpoint
= 0;
6167 } else if (map_state
== IMSM_T_STATE_FAILED
&&
6168 map
->map_state
!= map_state
) {
6169 dprintf("imsm: mark failed\n");
6170 end_migration(dev
, map_state
);
6171 super
->updates_pending
++;
6172 a
->last_checkpoint
= 0;
6173 } else if (is_gen_migration(dev
)) {
6174 dprintf("imsm: Detected General Migration in state: ");
6175 if (map_state
== IMSM_T_STATE_NORMAL
) {
6176 end_migration(dev
, map_state
);
6177 map
= get_imsm_map(dev
, 0);
6178 map
->failed_disk_num
= ~0;
6179 dprintf("normal\n");
6181 if (map_state
== IMSM_T_STATE_DEGRADED
) {
6182 printf("degraded\n");
6183 end_migration(dev
, map_state
);
6185 dprintf("failed\n");
6187 map
->map_state
= map_state
;
6189 super
->updates_pending
++;
6193 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
)
6196 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
6197 unsigned long long dsize
;
6198 unsigned long long sectors
;
6200 get_dev_size(fd
, NULL
, &dsize
);
6202 if (mpb_size
> 512) {
6203 /* -1 to account for anchor */
6204 sectors
= mpb_sectors(mpb
) - 1;
6206 /* write the extended mpb to the sectors preceeding the anchor */
6207 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
6210 if ((unsigned long long)write(fd
, buf
+ 512, 512 * sectors
)
6215 /* first block is stored on second to last sector of the disk */
6216 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
6219 if (write(fd
, buf
, 512) != 512)
6225 static void imsm_sync_metadata(struct supertype
*container
)
6227 struct intel_super
*super
= container
->sb
;
6229 dprintf("sync metadata: %d\n", super
->updates_pending
);
6230 if (!super
->updates_pending
)
6233 write_super_imsm(container
, 0);
6235 super
->updates_pending
= 0;
6238 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
6240 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6241 int i
= get_imsm_disk_idx(dev
, idx
, -1);
6244 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6248 if (dl
&& is_failed(&dl
->disk
))
6252 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
6257 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
,
6258 struct active_array
*a
, int activate_new
,
6259 struct mdinfo
*additional_test_list
)
6261 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6262 int idx
= get_imsm_disk_idx(dev
, slot
, -1);
6263 struct imsm_super
*mpb
= super
->anchor
;
6264 struct imsm_map
*map
;
6265 unsigned long long pos
;
6270 __u32 array_start
= 0;
6271 __u32 array_end
= 0;
6273 struct mdinfo
*test_list
;
6275 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6276 /* If in this array, skip */
6277 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6278 if (d
->state_fd
>= 0 &&
6279 d
->disk
.major
== dl
->major
&&
6280 d
->disk
.minor
== dl
->minor
) {
6281 dprintf("%x:%x already in array\n",
6282 dl
->major
, dl
->minor
);
6287 test_list
= additional_test_list
;
6289 if (test_list
->disk
.major
== dl
->major
&&
6290 test_list
->disk
.minor
== dl
->minor
) {
6291 dprintf("%x:%x already in additional test list\n",
6292 dl
->major
, dl
->minor
);
6295 test_list
= test_list
->next
;
6300 /* skip in use or failed drives */
6301 if (is_failed(&dl
->disk
) || idx
== dl
->index
||
6303 dprintf("%x:%x status (failed: %d index: %d)\n",
6304 dl
->major
, dl
->minor
, is_failed(&dl
->disk
), idx
);
6308 /* skip pure spares when we are looking for partially
6309 * assimilated drives
6311 if (dl
->index
== -1 && !activate_new
)
6314 /* Does this unused device have the requisite free space?
6315 * It needs to be able to cover all member volumes
6317 ex
= get_extents(super
, dl
);
6319 dprintf("cannot get extents\n");
6322 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
6323 dev
= get_imsm_dev(super
, i
);
6324 map
= get_imsm_map(dev
, 0);
6326 /* check if this disk is already a member of
6329 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
6335 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
6336 array_end
= array_start
+
6337 __le32_to_cpu(map
->blocks_per_member
) - 1;
6340 /* check that we can start at pba_of_lba0 with
6341 * blocks_per_member of space
6343 if (array_start
>= pos
&& array_end
< ex
[j
].start
) {
6347 pos
= ex
[j
].start
+ ex
[j
].size
;
6349 } while (ex
[j
-1].size
);
6356 if (i
< mpb
->num_raid_devs
) {
6357 dprintf("%x:%x does not have %u to %u available\n",
6358 dl
->major
, dl
->minor
, array_start
, array_end
);
6369 static int imsm_rebuild_allowed(struct supertype
*cont
, int dev_idx
, int failed
)
6371 struct imsm_dev
*dev2
;
6372 struct imsm_map
*map
;
6378 dev2
= get_imsm_dev(cont
->sb
, dev_idx
);
6380 state
= imsm_check_degraded(cont
->sb
, dev2
, failed
);
6381 if (state
== IMSM_T_STATE_FAILED
) {
6382 map
= get_imsm_map(dev2
, 0);
6385 for (slot
= 0; slot
< map
->num_members
; slot
++) {
6387 * Check if failed disks are deleted from intel
6388 * disk list or are marked to be deleted
6390 idx
= get_imsm_disk_idx(dev2
, slot
, -1);
6391 idisk
= get_imsm_dl_disk(cont
->sb
, idx
);
6393 * Do not rebuild the array if failed disks
6394 * from failed sub-array are not removed from
6398 is_failed(&idisk
->disk
) &&
6399 (idisk
->action
!= DISK_REMOVE
))
6407 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
6408 struct metadata_update
**updates
)
6411 * Find a device with unused free space and use it to replace a
6412 * failed/vacant region in an array. We replace failed regions one a
6413 * array at a time. The result is that a new spare disk will be added
6414 * to the first failed array and after the monitor has finished
6415 * propagating failures the remainder will be consumed.
6417 * FIXME add a capability for mdmon to request spares from another
6421 struct intel_super
*super
= a
->container
->sb
;
6422 int inst
= a
->info
.container_member
;
6423 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6424 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6425 int failed
= a
->info
.array
.raid_disks
;
6426 struct mdinfo
*rv
= NULL
;
6429 struct metadata_update
*mu
;
6431 struct imsm_update_activate_spare
*u
;
6436 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
6437 if ((d
->curr_state
& DS_FAULTY
) &&
6439 /* wait for Removal to happen */
6441 if (d
->state_fd
>= 0)
6445 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
6446 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
6448 if (dev
->vol
.migr_state
&&
6449 dev
->vol
.migr_type
== MIGR_GEN_MIGR
)
6450 /* No repair during migration */
6453 if (a
->info
.array
.level
== 4)
6454 /* No repair for takeovered array
6455 * imsm doesn't support raid4
6459 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
6463 * If there are any failed disks check state of the other volume.
6464 * Block rebuild if the another one is failed until failed disks
6465 * are removed from container.
6468 dprintf("found failed disks in %s, check if there another"
6469 "failed sub-array.\n",
6471 /* check if states of the other volumes allow for rebuild */
6472 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
6474 allowed
= imsm_rebuild_allowed(a
->container
,
6482 /* For each slot, if it is not working, find a spare */
6483 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
6484 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6485 if (d
->disk
.raid_disk
== i
)
6487 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
6488 if (d
&& (d
->state_fd
>= 0))
6492 * OK, this device needs recovery. Try to re-add the
6493 * previous occupant of this slot, if this fails see if
6494 * we can continue the assimilation of a spare that was
6495 * partially assimilated, finally try to activate a new
6498 dl
= imsm_readd(super
, i
, a
);
6500 dl
= imsm_add_spare(super
, i
, a
, 0, NULL
);
6502 dl
= imsm_add_spare(super
, i
, a
, 1, NULL
);
6506 /* found a usable disk with enough space */
6507 di
= malloc(sizeof(*di
));
6510 memset(di
, 0, sizeof(*di
));
6512 /* dl->index will be -1 in the case we are activating a
6513 * pristine spare. imsm_process_update() will create a
6514 * new index in this case. Once a disk is found to be
6515 * failed in all member arrays it is kicked from the
6518 di
->disk
.number
= dl
->index
;
6520 /* (ab)use di->devs to store a pointer to the device
6523 di
->devs
= (struct mdinfo
*) dl
;
6525 di
->disk
.raid_disk
= i
;
6526 di
->disk
.major
= dl
->major
;
6527 di
->disk
.minor
= dl
->minor
;
6529 di
->recovery_start
= 0;
6530 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
6531 di
->component_size
= a
->info
.component_size
;
6532 di
->container_member
= inst
;
6533 super
->random
= random32();
6537 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
6538 i
, di
->data_offset
);
6544 /* No spares found */
6546 /* Now 'rv' has a list of devices to return.
6547 * Create a metadata_update record to update the
6548 * disk_ord_tbl for the array
6550 mu
= malloc(sizeof(*mu
));
6552 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
6553 if (mu
->buf
== NULL
) {
6560 struct mdinfo
*n
= rv
->next
;
6569 mu
->space_list
= NULL
;
6570 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
6571 mu
->next
= *updates
;
6572 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
6574 for (di
= rv
; di
; di
= di
->next
) {
6575 u
->type
= update_activate_spare
;
6576 u
->dl
= (struct dl
*) di
->devs
;
6578 u
->slot
= di
->disk
.raid_disk
;
6589 static int disks_overlap(struct intel_super
*super
, int idx
, struct imsm_update_create_array
*u
)
6591 struct imsm_dev
*dev
= get_imsm_dev(super
, idx
);
6592 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6593 struct imsm_map
*new_map
= get_imsm_map(&u
->dev
, 0);
6594 struct disk_info
*inf
= get_disk_info(u
);
6595 struct imsm_disk
*disk
;
6599 for (i
= 0; i
< map
->num_members
; i
++) {
6600 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
, -1));
6601 for (j
= 0; j
< new_map
->num_members
; j
++)
6602 if (serialcmp(disk
->serial
, inf
[j
].serial
) == 0)
6610 static struct dl
*get_disk_super(struct intel_super
*super
, int major
, int minor
)
6612 struct dl
*dl
= NULL
;
6613 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6614 if ((dl
->major
== major
) && (dl
->minor
== minor
))
6619 static int remove_disk_super(struct intel_super
*super
, int major
, int minor
)
6621 struct dl
*prev
= NULL
;
6625 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6626 if ((dl
->major
== major
) && (dl
->minor
== minor
)) {
6629 prev
->next
= dl
->next
;
6631 super
->disks
= dl
->next
;
6633 __free_imsm_disk(dl
);
6634 dprintf("%s: removed %x:%x\n",
6635 __func__
, major
, minor
);
6643 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
);
6645 static int add_remove_disk_update(struct intel_super
*super
)
6647 int check_degraded
= 0;
6648 struct dl
*disk
= NULL
;
6649 /* add/remove some spares to/from the metadata/contrainer */
6650 while (super
->disk_mgmt_list
) {
6651 struct dl
*disk_cfg
;
6653 disk_cfg
= super
->disk_mgmt_list
;
6654 super
->disk_mgmt_list
= disk_cfg
->next
;
6655 disk_cfg
->next
= NULL
;
6657 if (disk_cfg
->action
== DISK_ADD
) {
6658 disk_cfg
->next
= super
->disks
;
6659 super
->disks
= disk_cfg
;
6661 dprintf("%s: added %x:%x\n",
6662 __func__
, disk_cfg
->major
,
6664 } else if (disk_cfg
->action
== DISK_REMOVE
) {
6665 dprintf("Disk remove action processed: %x.%x\n",
6666 disk_cfg
->major
, disk_cfg
->minor
);
6667 disk
= get_disk_super(super
,
6671 /* store action status */
6672 disk
->action
= DISK_REMOVE
;
6673 /* remove spare disks only */
6674 if (disk
->index
== -1) {
6675 remove_disk_super(super
,
6680 /* release allocate disk structure */
6681 __free_imsm_disk(disk_cfg
);
6684 return check_degraded
;
6688 static int apply_reshape_migration_update(struct imsm_update_reshape_migration
*u
,
6689 struct intel_super
*super
,
6692 struct intel_dev
*id
;
6693 void **tofree
= NULL
;
6696 dprintf("apply_reshape_migration_update()\n");
6697 if ((u
->subdev
< 0) ||
6699 dprintf("imsm: Error: Wrong subdev: %i\n", u
->subdev
);
6702 if ((space_list
== NULL
) || (*space_list
== NULL
)) {
6703 dprintf("imsm: Error: Memory is not allocated\n");
6707 for (id
= super
->devlist
; id
; id
= id
->next
) {
6708 if (id
->index
== (unsigned)u
->subdev
) {
6709 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subdev
);
6710 struct imsm_map
*map
;
6711 struct imsm_dev
*new_dev
=
6712 (struct imsm_dev
*)*space_list
;
6713 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6715 struct dl
*new_disk
;
6717 if (new_dev
== NULL
)
6719 *space_list
= **space_list
;
6720 memcpy(new_dev
, dev
, sizeof_imsm_dev(dev
, 0));
6721 map
= get_imsm_map(new_dev
, 0);
6723 dprintf("imsm: Error: migration in progress");
6727 to_state
= map
->map_state
;
6728 if ((u
->new_level
== 5) && (map
->raid_level
== 0)) {
6730 /* this should not happen */
6731 if (u
->new_disks
[0] < 0) {
6732 map
->failed_disk_num
=
6733 map
->num_members
- 1;
6734 to_state
= IMSM_T_STATE_DEGRADED
;
6736 to_state
= IMSM_T_STATE_NORMAL
;
6738 migrate(new_dev
, super
, to_state
, MIGR_GEN_MIGR
);
6739 if (u
->new_level
> -1)
6740 map
->raid_level
= u
->new_level
;
6741 migr_map
= get_imsm_map(new_dev
, 1);
6742 if ((u
->new_level
== 5) &&
6743 (migr_map
->raid_level
== 0)) {
6744 int ord
= map
->num_members
- 1;
6745 migr_map
->num_members
--;
6746 if (u
->new_disks
[0] < 0)
6747 ord
|= IMSM_ORD_REBUILD
;
6748 set_imsm_ord_tbl_ent(map
,
6749 map
->num_members
- 1,
6753 tofree
= (void **)dev
;
6755 /* update chunk size
6757 if (u
->new_chunksize
> 0)
6758 map
->blocks_per_strip
=
6759 __cpu_to_le16(u
->new_chunksize
* 2);
6763 if ((u
->new_level
!= 5) ||
6764 (migr_map
->raid_level
!= 0) ||
6765 (migr_map
->raid_level
== map
->raid_level
))
6768 if (u
->new_disks
[0] >= 0) {
6771 new_disk
= get_disk_super(super
,
6772 major(u
->new_disks
[0]),
6773 minor(u
->new_disks
[0]));
6774 dprintf("imsm: new disk for reshape is: %i:%i "
6775 "(%p, index = %i)\n",
6776 major(u
->new_disks
[0]),
6777 minor(u
->new_disks
[0]),
6778 new_disk
, new_disk
->index
);
6779 if (new_disk
== NULL
)
6780 goto error_disk_add
;
6782 new_disk
->index
= map
->num_members
- 1;
6783 /* slot to fill in autolayout
6785 new_disk
->raiddisk
= new_disk
->index
;
6786 new_disk
->disk
.status
|= CONFIGURED_DISK
;
6787 new_disk
->disk
.status
&= ~SPARE_DISK
;
6789 goto error_disk_add
;
6792 *tofree
= *space_list
;
6793 /* calculate new size
6795 imsm_set_array_size(new_dev
);
6802 *space_list
= tofree
;
6806 dprintf("Error: imsm: Cannot find disk.\n");
6811 static int apply_reshape_container_disks_update(struct imsm_update_reshape
*u
,
6812 struct intel_super
*super
,
6815 struct dl
*new_disk
;
6816 struct intel_dev
*id
;
6818 int delta_disks
= u
->new_raid_disks
- u
->old_raid_disks
;
6819 int disk_count
= u
->old_raid_disks
;
6820 void **tofree
= NULL
;
6821 int devices_to_reshape
= 1;
6822 struct imsm_super
*mpb
= super
->anchor
;
6824 unsigned int dev_id
;
6826 dprintf("imsm: apply_reshape_container_disks_update()\n");
6828 /* enable spares to use in array */
6829 for (i
= 0; i
< delta_disks
; i
++) {
6830 new_disk
= get_disk_super(super
,
6831 major(u
->new_disks
[i
]),
6832 minor(u
->new_disks
[i
]));
6833 dprintf("imsm: new disk for reshape is: %i:%i "
6834 "(%p, index = %i)\n",
6835 major(u
->new_disks
[i
]), minor(u
->new_disks
[i
]),
6836 new_disk
, new_disk
->index
);
6837 if ((new_disk
== NULL
) ||
6838 ((new_disk
->index
>= 0) &&
6839 (new_disk
->index
< u
->old_raid_disks
)))
6840 goto update_reshape_exit
;
6841 new_disk
->index
= disk_count
++;
6842 /* slot to fill in autolayout
6844 new_disk
->raiddisk
= new_disk
->index
;
6845 new_disk
->disk
.status
|=
6847 new_disk
->disk
.status
&= ~SPARE_DISK
;
6850 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
6851 mpb
->num_raid_devs
);
6852 /* manage changes in volume
6854 for (dev_id
= 0; dev_id
< mpb
->num_raid_devs
; dev_id
++) {
6855 void **sp
= *space_list
;
6856 struct imsm_dev
*newdev
;
6857 struct imsm_map
*newmap
, *oldmap
;
6859 for (id
= super
->devlist
; id
; id
= id
->next
) {
6860 if (id
->index
== dev_id
)
6869 /* Copy the dev, but not (all of) the map */
6870 memcpy(newdev
, id
->dev
, sizeof(*newdev
));
6871 oldmap
= get_imsm_map(id
->dev
, 0);
6872 newmap
= get_imsm_map(newdev
, 0);
6873 /* Copy the current map */
6874 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6875 /* update one device only
6877 if (devices_to_reshape
) {
6878 dprintf("imsm: modifying subdev: %i\n",
6880 devices_to_reshape
--;
6881 newdev
->vol
.migr_state
= 1;
6882 newdev
->vol
.curr_migr_unit
= 0;
6883 newdev
->vol
.migr_type
= MIGR_GEN_MIGR
;
6884 newmap
->num_members
= u
->new_raid_disks
;
6885 for (i
= 0; i
< delta_disks
; i
++) {
6886 set_imsm_ord_tbl_ent(newmap
,
6887 u
->old_raid_disks
+ i
,
6888 u
->old_raid_disks
+ i
);
6890 /* New map is correct, now need to save old map
6892 newmap
= get_imsm_map(newdev
, 1);
6893 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6895 imsm_set_array_size(newdev
);
6898 sp
= (void **)id
->dev
;
6903 /* Clear migration record */
6904 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
6907 *space_list
= tofree
;
6910 update_reshape_exit
:
6915 static int apply_takeover_update(struct imsm_update_takeover
*u
,
6916 struct intel_super
*super
,
6919 struct imsm_dev
*dev
= NULL
;
6920 struct intel_dev
*dv
;
6921 struct imsm_dev
*dev_new
;
6922 struct imsm_map
*map
;
6926 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
6927 if (dv
->index
== (unsigned int)u
->subarray
) {
6935 map
= get_imsm_map(dev
, 0);
6937 if (u
->direction
== R10_TO_R0
) {
6938 /* Number of failed disks must be half of initial disk number */
6939 if (imsm_count_failed(super
, dev
) != (map
->num_members
/ 2))
6942 /* iterate through devices to mark removed disks as spare */
6943 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6944 if (dm
->disk
.status
& FAILED_DISK
) {
6945 int idx
= dm
->index
;
6946 /* update indexes on the disk list */
6947 /* FIXME this loop-with-the-loop looks wrong, I'm not convinced
6948 the index values will end up being correct.... NB */
6949 for (du
= super
->disks
; du
; du
= du
->next
)
6950 if (du
->index
> idx
)
6952 /* mark as spare disk */
6953 dm
->disk
.status
= SPARE_DISK
;
6958 map
->num_members
= map
->num_members
/ 2;
6959 map
->map_state
= IMSM_T_STATE_NORMAL
;
6960 map
->num_domains
= 1;
6961 map
->raid_level
= 0;
6962 map
->failed_disk_num
= -1;
6965 if (u
->direction
== R0_TO_R10
) {
6967 /* update slots in current disk list */
6968 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6972 /* create new *missing* disks */
6973 for (i
= 0; i
< map
->num_members
; i
++) {
6974 space
= *space_list
;
6977 *space_list
= *space
;
6979 memcpy(du
, super
->disks
, sizeof(*du
));
6983 du
->index
= (i
* 2) + 1;
6984 sprintf((char *)du
->disk
.serial
,
6985 " MISSING_%d", du
->index
);
6986 sprintf((char *)du
->serial
,
6987 "MISSING_%d", du
->index
);
6988 du
->next
= super
->missing
;
6989 super
->missing
= du
;
6991 /* create new dev and map */
6992 space
= *space_list
;
6995 *space_list
= *space
;
6996 dev_new
= (void *)space
;
6997 memcpy(dev_new
, dev
, sizeof(*dev
));
6998 /* update new map */
6999 map
= get_imsm_map(dev_new
, 0);
7000 map
->num_members
= map
->num_members
* 2;
7001 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7002 map
->num_domains
= 2;
7003 map
->raid_level
= 1;
7004 /* replace dev<->dev_new */
7007 /* update disk order table */
7008 for (du
= super
->disks
; du
; du
= du
->next
)
7010 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7011 for (du
= super
->missing
; du
; du
= du
->next
)
7012 if (du
->index
>= 0) {
7013 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7014 mark_missing(dv
->dev
, &du
->disk
, du
->index
);
7020 static void imsm_process_update(struct supertype
*st
,
7021 struct metadata_update
*update
)
7024 * crack open the metadata_update envelope to find the update record
7025 * update can be one of:
7026 * update_reshape_container_disks - all the arrays in the container
7027 * are being reshaped to have more devices. We need to mark
7028 * the arrays for general migration and convert selected spares
7029 * into active devices.
7030 * update_activate_spare - a spare device has replaced a failed
7031 * device in an array, update the disk_ord_tbl. If this disk is
7032 * present in all member arrays then also clear the SPARE_DISK
7034 * update_create_array
7036 * update_rename_array
7037 * update_add_remove_disk
7039 struct intel_super
*super
= st
->sb
;
7040 struct imsm_super
*mpb
;
7041 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7043 /* update requires a larger buf but the allocation failed */
7044 if (super
->next_len
&& !super
->next_buf
) {
7045 super
->next_len
= 0;
7049 if (super
->next_buf
) {
7050 memcpy(super
->next_buf
, super
->buf
, super
->len
);
7052 super
->len
= super
->next_len
;
7053 super
->buf
= super
->next_buf
;
7055 super
->next_len
= 0;
7056 super
->next_buf
= NULL
;
7059 mpb
= super
->anchor
;
7062 case update_general_migration_checkpoint
: {
7063 struct intel_dev
*id
;
7064 struct imsm_update_general_migration_checkpoint
*u
=
7065 (void *)update
->buf
;
7067 dprintf("imsm: process_update() "
7068 "for update_general_migration_checkpoint called\n");
7070 /* find device under general migration */
7071 for (id
= super
->devlist
; id
; id
= id
->next
) {
7072 if (is_gen_migration(id
->dev
)) {
7073 id
->dev
->vol
.curr_migr_unit
=
7074 __cpu_to_le32(u
->curr_migr_unit
);
7075 super
->updates_pending
++;
7080 case update_takeover
: {
7081 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7082 if (apply_takeover_update(u
, super
, &update
->space_list
)) {
7083 imsm_update_version_info(super
);
7084 super
->updates_pending
++;
7089 case update_reshape_container_disks
: {
7090 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7091 if (apply_reshape_container_disks_update(
7092 u
, super
, &update
->space_list
))
7093 super
->updates_pending
++;
7096 case update_reshape_migration
: {
7097 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7098 if (apply_reshape_migration_update(
7099 u
, super
, &update
->space_list
))
7100 super
->updates_pending
++;
7103 case update_activate_spare
: {
7104 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
7105 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
7106 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7107 struct imsm_map
*migr_map
;
7108 struct active_array
*a
;
7109 struct imsm_disk
*disk
;
7114 int victim
= get_imsm_disk_idx(dev
, u
->slot
, -1);
7117 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
7122 fprintf(stderr
, "error: imsm_activate_spare passed "
7123 "an unknown disk (index: %d)\n",
7128 super
->updates_pending
++;
7129 /* count failures (excluding rebuilds and the victim)
7130 * to determine map[0] state
7133 for (i
= 0; i
< map
->num_members
; i
++) {
7136 disk
= get_imsm_disk(super
,
7137 get_imsm_disk_idx(dev
, i
, -1));
7138 if (!disk
|| is_failed(disk
))
7142 /* adding a pristine spare, assign a new index */
7143 if (dl
->index
< 0) {
7144 dl
->index
= super
->anchor
->num_disks
;
7145 super
->anchor
->num_disks
++;
7148 disk
->status
|= CONFIGURED_DISK
;
7149 disk
->status
&= ~SPARE_DISK
;
7152 to_state
= imsm_check_degraded(super
, dev
, failed
);
7153 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7154 migrate(dev
, super
, to_state
, MIGR_REBUILD
);
7155 migr_map
= get_imsm_map(dev
, 1);
7156 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
7157 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
7159 /* update the family_num to mark a new container
7160 * generation, being careful to record the existing
7161 * family_num in orig_family_num to clean up after
7162 * earlier mdadm versions that neglected to set it.
7164 if (mpb
->orig_family_num
== 0)
7165 mpb
->orig_family_num
= mpb
->family_num
;
7166 mpb
->family_num
+= super
->random
;
7168 /* count arrays using the victim in the metadata */
7170 for (a
= st
->arrays
; a
; a
= a
->next
) {
7171 dev
= get_imsm_dev(super
, a
->info
.container_member
);
7172 map
= get_imsm_map(dev
, 0);
7174 if (get_imsm_disk_slot(map
, victim
) >= 0)
7178 /* delete the victim if it is no longer being
7184 /* We know that 'manager' isn't touching anything,
7185 * so it is safe to delete
7187 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
7188 if ((*dlp
)->index
== victim
)
7191 /* victim may be on the missing list */
7193 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
7194 if ((*dlp
)->index
== victim
)
7196 imsm_delete(super
, dlp
, victim
);
7200 case update_create_array
: {
7201 /* someone wants to create a new array, we need to be aware of
7202 * a few races/collisions:
7203 * 1/ 'Create' called by two separate instances of mdadm
7204 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
7205 * devices that have since been assimilated via
7207 * In the event this update can not be carried out mdadm will
7208 * (FIX ME) notice that its update did not take hold.
7210 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7211 struct intel_dev
*dv
;
7212 struct imsm_dev
*dev
;
7213 struct imsm_map
*map
, *new_map
;
7214 unsigned long long start
, end
;
7215 unsigned long long new_start
, new_end
;
7217 struct disk_info
*inf
;
7220 /* handle racing creates: first come first serve */
7221 if (u
->dev_idx
< mpb
->num_raid_devs
) {
7222 dprintf("%s: subarray %d already defined\n",
7223 __func__
, u
->dev_idx
);
7227 /* check update is next in sequence */
7228 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
7229 dprintf("%s: can not create array %d expected index %d\n",
7230 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
7234 new_map
= get_imsm_map(&u
->dev
, 0);
7235 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
7236 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
7237 inf
= get_disk_info(u
);
7239 /* handle activate_spare versus create race:
7240 * check to make sure that overlapping arrays do not include
7243 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7244 dev
= get_imsm_dev(super
, i
);
7245 map
= get_imsm_map(dev
, 0);
7246 start
= __le32_to_cpu(map
->pba_of_lba0
);
7247 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
7248 if ((new_start
>= start
&& new_start
<= end
) ||
7249 (start
>= new_start
&& start
<= new_end
))
7254 if (disks_overlap(super
, i
, u
)) {
7255 dprintf("%s: arrays overlap\n", __func__
);
7260 /* check that prepare update was successful */
7261 if (!update
->space
) {
7262 dprintf("%s: prepare update failed\n", __func__
);
7266 /* check that all disks are still active before committing
7267 * changes. FIXME: could we instead handle this by creating a
7268 * degraded array? That's probably not what the user expects,
7269 * so better to drop this update on the floor.
7271 for (i
= 0; i
< new_map
->num_members
; i
++) {
7272 dl
= serial_to_dl(inf
[i
].serial
, super
);
7274 dprintf("%s: disk disappeared\n", __func__
);
7279 super
->updates_pending
++;
7281 /* convert spares to members and fixup ord_tbl */
7282 for (i
= 0; i
< new_map
->num_members
; i
++) {
7283 dl
= serial_to_dl(inf
[i
].serial
, super
);
7284 if (dl
->index
== -1) {
7285 dl
->index
= mpb
->num_disks
;
7287 dl
->disk
.status
|= CONFIGURED_DISK
;
7288 dl
->disk
.status
&= ~SPARE_DISK
;
7290 set_imsm_ord_tbl_ent(new_map
, i
, dl
->index
);
7295 update
->space
= NULL
;
7296 imsm_copy_dev(dev
, &u
->dev
);
7297 dv
->index
= u
->dev_idx
;
7298 dv
->next
= super
->devlist
;
7299 super
->devlist
= dv
;
7300 mpb
->num_raid_devs
++;
7302 imsm_update_version_info(super
);
7305 /* mdmon knows how to release update->space, but not
7306 * ((struct intel_dev *) update->space)->dev
7308 if (update
->space
) {
7314 case update_kill_array
: {
7315 struct imsm_update_kill_array
*u
= (void *) update
->buf
;
7316 int victim
= u
->dev_idx
;
7317 struct active_array
*a
;
7318 struct intel_dev
**dp
;
7319 struct imsm_dev
*dev
;
7321 /* sanity check that we are not affecting the uuid of
7322 * active arrays, or deleting an active array
7324 * FIXME when immutable ids are available, but note that
7325 * we'll also need to fixup the invalidated/active
7326 * subarray indexes in mdstat
7328 for (a
= st
->arrays
; a
; a
= a
->next
)
7329 if (a
->info
.container_member
>= victim
)
7331 /* by definition if mdmon is running at least one array
7332 * is active in the container, so checking
7333 * mpb->num_raid_devs is just extra paranoia
7335 dev
= get_imsm_dev(super
, victim
);
7336 if (a
|| !dev
|| mpb
->num_raid_devs
== 1) {
7337 dprintf("failed to delete subarray-%d\n", victim
);
7341 for (dp
= &super
->devlist
; *dp
;)
7342 if ((*dp
)->index
== (unsigned)super
->current_vol
) {
7345 if ((*dp
)->index
> (unsigned)victim
)
7349 mpb
->num_raid_devs
--;
7350 super
->updates_pending
++;
7353 case update_rename_array
: {
7354 struct imsm_update_rename_array
*u
= (void *) update
->buf
;
7355 char name
[MAX_RAID_SERIAL_LEN
+1];
7356 int target
= u
->dev_idx
;
7357 struct active_array
*a
;
7358 struct imsm_dev
*dev
;
7360 /* sanity check that we are not affecting the uuid of
7363 snprintf(name
, MAX_RAID_SERIAL_LEN
, "%s", (char *) u
->name
);
7364 name
[MAX_RAID_SERIAL_LEN
] = '\0';
7365 for (a
= st
->arrays
; a
; a
= a
->next
)
7366 if (a
->info
.container_member
== target
)
7368 dev
= get_imsm_dev(super
, u
->dev_idx
);
7369 if (a
|| !dev
|| !check_name(super
, name
, 1)) {
7370 dprintf("failed to rename subarray-%d\n", target
);
7374 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
7375 super
->updates_pending
++;
7378 case update_add_remove_disk
: {
7379 /* we may be able to repair some arrays if disks are
7380 * being added, check teh status of add_remove_disk
7381 * if discs has been added.
7383 if (add_remove_disk_update(super
)) {
7384 struct active_array
*a
;
7386 super
->updates_pending
++;
7387 for (a
= st
->arrays
; a
; a
= a
->next
)
7388 a
->check_degraded
= 1;
7393 fprintf(stderr
, "error: unsuported process update type:"
7394 "(type: %d)\n", type
);
7398 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
);
7400 static void imsm_prepare_update(struct supertype
*st
,
7401 struct metadata_update
*update
)
7404 * Allocate space to hold new disk entries, raid-device entries or a new
7405 * mpb if necessary. The manager synchronously waits for updates to
7406 * complete in the monitor, so new mpb buffers allocated here can be
7407 * integrated by the monitor thread without worrying about live pointers
7408 * in the manager thread.
7410 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7411 struct intel_super
*super
= st
->sb
;
7412 struct imsm_super
*mpb
= super
->anchor
;
7417 case update_general_migration_checkpoint
:
7418 dprintf("imsm: prepare_update() "
7419 "for update_general_migration_checkpoint called\n");
7421 case update_takeover
: {
7422 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7423 if (u
->direction
== R0_TO_R10
) {
7424 void **tail
= (void **)&update
->space_list
;
7425 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subarray
);
7426 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7427 int num_members
= map
->num_members
;
7431 /* allocate memory for added disks */
7432 for (i
= 0; i
< num_members
; i
++) {
7433 size
= sizeof(struct dl
);
7434 space
= malloc(size
);
7443 /* allocate memory for new device */
7444 size
= sizeof_imsm_dev(super
->devlist
->dev
, 0) +
7445 (num_members
* sizeof(__u32
));
7446 space
= malloc(size
);
7455 len
= disks_to_mpb_size(num_members
* 2);
7457 /* if allocation didn't success, free buffer */
7458 while (update
->space_list
) {
7459 void **sp
= update
->space_list
;
7460 update
->space_list
= *sp
;
7468 case update_reshape_container_disks
: {
7469 /* Every raid device in the container is about to
7470 * gain some more devices, and we will enter a
7472 * So each 'imsm_map' will be bigger, and the imsm_vol
7473 * will now hold 2 of them.
7474 * Thus we need new 'struct imsm_dev' allocations sized
7475 * as sizeof_imsm_dev but with more devices in both maps.
7477 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7478 struct intel_dev
*dl
;
7479 void **space_tail
= (void**)&update
->space_list
;
7481 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7483 for (dl
= super
->devlist
; dl
; dl
= dl
->next
) {
7484 int size
= sizeof_imsm_dev(dl
->dev
, 1);
7486 if (u
->new_raid_disks
> u
->old_raid_disks
)
7487 size
+= sizeof(__u32
)*2*
7488 (u
->new_raid_disks
- u
->old_raid_disks
);
7497 len
= disks_to_mpb_size(u
->new_raid_disks
);
7498 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7501 case update_reshape_migration
: {
7502 /* for migration level 0->5 we need to add disks
7503 * so the same as for container operation we will copy
7504 * device to the bigger location.
7505 * in memory prepared device and new disk area are prepared
7506 * for usage in process update
7508 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7509 struct intel_dev
*id
;
7510 void **space_tail
= (void **)&update
->space_list
;
7513 int current_level
= -1;
7515 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7517 /* add space for bigger array in update
7519 for (id
= super
->devlist
; id
; id
= id
->next
) {
7520 if (id
->index
== (unsigned)u
->subdev
) {
7521 size
= sizeof_imsm_dev(id
->dev
, 1);
7522 if (u
->new_raid_disks
> u
->old_raid_disks
)
7523 size
+= sizeof(__u32
)*2*
7524 (u
->new_raid_disks
- u
->old_raid_disks
);
7534 if (update
->space_list
== NULL
)
7537 /* add space for disk in update
7539 size
= sizeof(struct dl
);
7542 free(update
->space_list
);
7543 update
->space_list
= NULL
;
7550 /* add spare device to update
7552 for (id
= super
->devlist
; id
; id
= id
->next
)
7553 if (id
->index
== (unsigned)u
->subdev
) {
7554 struct imsm_dev
*dev
;
7555 struct imsm_map
*map
;
7557 dev
= get_imsm_dev(super
, u
->subdev
);
7558 map
= get_imsm_map(dev
, 0);
7559 current_level
= map
->raid_level
;
7562 if ((u
->new_level
== 5) && (u
->new_level
!= current_level
)) {
7563 struct mdinfo
*spares
;
7565 spares
= get_spares_for_grow(st
);
7573 makedev(dev
->disk
.major
,
7575 dl
= get_disk_super(super
,
7578 dl
->index
= u
->old_raid_disks
;
7584 len
= disks_to_mpb_size(u
->new_raid_disks
);
7585 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7588 case update_create_array
: {
7589 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7590 struct intel_dev
*dv
;
7591 struct imsm_dev
*dev
= &u
->dev
;
7592 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7594 struct disk_info
*inf
;
7598 inf
= get_disk_info(u
);
7599 len
= sizeof_imsm_dev(dev
, 1);
7600 /* allocate a new super->devlist entry */
7601 dv
= malloc(sizeof(*dv
));
7603 dv
->dev
= malloc(len
);
7608 update
->space
= NULL
;
7612 /* count how many spares will be converted to members */
7613 for (i
= 0; i
< map
->num_members
; i
++) {
7614 dl
= serial_to_dl(inf
[i
].serial
, super
);
7616 /* hmm maybe it failed?, nothing we can do about
7621 if (count_memberships(dl
, super
) == 0)
7624 len
+= activate
* sizeof(struct imsm_disk
);
7631 /* check if we need a larger metadata buffer */
7632 if (super
->next_buf
)
7633 buf_len
= super
->next_len
;
7635 buf_len
= super
->len
;
7637 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
7638 /* ok we need a larger buf than what is currently allocated
7639 * if this allocation fails process_update will notice that
7640 * ->next_len is set and ->next_buf is NULL
7642 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
7643 if (super
->next_buf
)
7644 free(super
->next_buf
);
7646 super
->next_len
= buf_len
;
7647 if (posix_memalign(&super
->next_buf
, 512, buf_len
) == 0)
7648 memset(super
->next_buf
, 0, buf_len
);
7650 super
->next_buf
= NULL
;
7654 /* must be called while manager is quiesced */
7655 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
)
7657 struct imsm_super
*mpb
= super
->anchor
;
7659 struct imsm_dev
*dev
;
7660 struct imsm_map
*map
;
7661 int i
, j
, num_members
;
7664 dprintf("%s: deleting device[%d] from imsm_super\n",
7667 /* shift all indexes down one */
7668 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
7669 if (iter
->index
> (int)index
)
7671 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
7672 if (iter
->index
> (int)index
)
7675 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7676 dev
= get_imsm_dev(super
, i
);
7677 map
= get_imsm_map(dev
, 0);
7678 num_members
= map
->num_members
;
7679 for (j
= 0; j
< num_members
; j
++) {
7680 /* update ord entries being careful not to propagate
7681 * ord-flags to the first map
7683 ord
= get_imsm_ord_tbl_ent(dev
, j
, -1);
7685 if (ord_to_idx(ord
) <= index
)
7688 map
= get_imsm_map(dev
, 0);
7689 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
7690 map
= get_imsm_map(dev
, 1);
7692 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
7697 super
->updates_pending
++;
7699 struct dl
*dl
= *dlp
;
7701 *dlp
= (*dlp
)->next
;
7702 __free_imsm_disk(dl
);
7705 #endif /* MDASSEMBLE */
7706 /*******************************************************************************
7707 * Function: open_backup_targets
7708 * Description: Function opens file descriptors for all devices given in
7711 * info : general array info
7712 * raid_disks : number of disks
7713 * raid_fds : table of device's file descriptors
7717 ******************************************************************************/
7718 int open_backup_targets(struct mdinfo
*info
, int raid_disks
, int *raid_fds
)
7722 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7725 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
)) {
7726 dprintf("disk is faulty!!\n");
7730 if ((sd
->disk
.raid_disk
>= raid_disks
) ||
7731 (sd
->disk
.raid_disk
< 0))
7734 dn
= map_dev(sd
->disk
.major
,
7736 raid_fds
[sd
->disk
.raid_disk
] = dev_open(dn
, O_RDWR
);
7737 if (raid_fds
[sd
->disk
.raid_disk
] < 0) {
7738 fprintf(stderr
, "cannot open component\n");
7746 /*******************************************************************************
7747 * Function: init_migr_record_imsm
7748 * Description: Function inits imsm migration record
7750 * super : imsm internal array info
7751 * dev : device under migration
7752 * info : general array info to find the smallest device
7755 ******************************************************************************/
7756 void init_migr_record_imsm(struct supertype
*st
, struct imsm_dev
*dev
,
7757 struct mdinfo
*info
)
7759 struct intel_super
*super
= st
->sb
;
7760 struct migr_record
*migr_rec
= super
->migr_rec
;
7762 unsigned long long dsize
, dev_sectors
;
7763 long long unsigned min_dev_sectors
= -1LLU;
7767 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7768 struct imsm_map
*map_src
= get_imsm_map(dev
, 1);
7769 unsigned long long num_migr_units
;
7770 unsigned long long array_blocks
;
7772 memset(migr_rec
, 0, sizeof(struct migr_record
));
7773 migr_rec
->family_num
= __cpu_to_le32(super
->anchor
->family_num
);
7775 /* only ascending reshape supported now */
7776 migr_rec
->ascending_migr
= __cpu_to_le32(1);
7778 migr_rec
->dest_depth_per_unit
= GEN_MIGR_AREA_SIZE
/
7779 max(map_dest
->blocks_per_strip
, map_src
->blocks_per_strip
);
7780 migr_rec
->dest_depth_per_unit
*= map_dest
->blocks_per_strip
;
7781 new_data_disks
= imsm_num_data_members(dev
, 0);
7782 migr_rec
->blocks_per_unit
=
7783 __cpu_to_le32(migr_rec
->dest_depth_per_unit
* new_data_disks
);
7784 migr_rec
->dest_depth_per_unit
=
7785 __cpu_to_le32(migr_rec
->dest_depth_per_unit
);
7786 array_blocks
= info
->component_size
* new_data_disks
;
7788 array_blocks
/ __le32_to_cpu(migr_rec
->blocks_per_unit
);
7790 if (array_blocks
% __le32_to_cpu(migr_rec
->blocks_per_unit
))
7792 migr_rec
->num_migr_units
= __cpu_to_le32(num_migr_units
);
7794 migr_rec
->post_migr_vol_cap
= dev
->size_low
;
7795 migr_rec
->post_migr_vol_cap_hi
= dev
->size_high
;
7798 /* Find the smallest dev */
7799 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7800 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
7801 fd
= dev_open(nm
, O_RDONLY
);
7804 get_dev_size(fd
, NULL
, &dsize
);
7805 dev_sectors
= dsize
/ 512;
7806 if (dev_sectors
< min_dev_sectors
)
7807 min_dev_sectors
= dev_sectors
;
7810 migr_rec
->ckpt_area_pba
= __cpu_to_le32(min_dev_sectors
-
7811 RAID_DISK_RESERVED_BLOCKS_IMSM_HI
);
7813 write_imsm_migr_rec(st
);
7818 /*******************************************************************************
7819 * Function: save_backup_imsm
7820 * Description: Function saves critical data stripes to Migration Copy Area
7821 * and updates the current migration unit status.
7822 * Use restore_stripes() to form a destination stripe,
7823 * and to write it to the Copy Area.
7825 * st : supertype information
7826 * dev : imsm device that backup is saved for
7827 * info : general array info
7828 * buf : input buffer
7829 * length : length of data to backup (blocks_per_unit)
7833 ******************************************************************************/
7834 int save_backup_imsm(struct supertype
*st
,
7835 struct imsm_dev
*dev
,
7836 struct mdinfo
*info
,
7841 struct intel_super
*super
= st
->sb
;
7842 unsigned long long *target_offsets
= NULL
;
7843 int *targets
= NULL
;
7845 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7846 int new_disks
= map_dest
->num_members
;
7847 int dest_layout
= 0;
7849 unsigned long long start
;
7850 int data_disks
= imsm_num_data_members(dev
, 0);
7852 targets
= malloc(new_disks
* sizeof(int));
7856 for (i
= 0; i
< new_disks
; i
++)
7859 target_offsets
= malloc(new_disks
* sizeof(unsigned long long));
7860 if (!target_offsets
)
7863 start
= info
->reshape_progress
* 512;
7864 for (i
= 0; i
< new_disks
; i
++) {
7865 target_offsets
[i
] = (unsigned long long)
7866 __le32_to_cpu(super
->migr_rec
->ckpt_area_pba
) * 512;
7867 /* move back copy area adderss, it will be moved forward
7868 * in restore_stripes() using start input variable
7870 target_offsets
[i
] -= start
/data_disks
;
7873 if (open_backup_targets(info
, new_disks
, targets
))
7876 dest_layout
= imsm_level_to_layout(map_dest
->raid_level
);
7877 dest_chunk
= __le16_to_cpu(map_dest
->blocks_per_strip
) * 512;
7879 if (restore_stripes(targets
, /* list of dest devices */
7880 target_offsets
, /* migration record offsets */
7883 map_dest
->raid_level
,
7885 -1, /* source backup file descriptor */
7886 0, /* input buf offset
7887 * always 0 buf is already offseted */
7891 fprintf(stderr
, Name
": Error restoring stripes\n");
7899 for (i
= 0; i
< new_disks
; i
++)
7900 if (targets
[i
] >= 0)
7904 free(target_offsets
);
7909 /*******************************************************************************
7910 * Function: save_checkpoint_imsm
7911 * Description: Function called for current unit status update
7912 * in the migration record. It writes it to disk.
7914 * super : imsm internal array info
7915 * info : general array info
7919 * 2: failure, means no valid migration record
7920 * / no general migration in progress /
7921 ******************************************************************************/
7922 int save_checkpoint_imsm(struct supertype
*st
, struct mdinfo
*info
, int state
)
7924 struct intel_super
*super
= st
->sb
;
7925 unsigned long long blocks_per_unit
;
7926 unsigned long long curr_migr_unit
;
7928 if (load_imsm_migr_rec(super
, info
) != 0) {
7929 dprintf("imsm: ERROR: Cannot read migration record "
7930 "for checkpoint save.\n");
7934 blocks_per_unit
= __le32_to_cpu(super
->migr_rec
->blocks_per_unit
);
7935 if (blocks_per_unit
== 0) {
7936 dprintf("imsm: no migration in progress.\n");
7939 curr_migr_unit
= info
->reshape_progress
/ blocks_per_unit
;
7940 /* check if array is alligned to copy area
7941 * if it is not alligned, add one to current migration unit value
7942 * this can happend on array reshape finish only
7944 if (info
->reshape_progress
% blocks_per_unit
)
7947 super
->migr_rec
->curr_migr_unit
=
7948 __cpu_to_le32(curr_migr_unit
);
7949 super
->migr_rec
->rec_status
= __cpu_to_le32(state
);
7950 super
->migr_rec
->dest_1st_member_lba
=
7951 __cpu_to_le32(curr_migr_unit
*
7952 __le32_to_cpu(super
->migr_rec
->dest_depth_per_unit
));
7953 if (write_imsm_migr_rec(st
) < 0) {
7954 dprintf("imsm: Cannot write migration record "
7955 "outside backup area\n");
7962 /*******************************************************************************
7963 * Function: recover_backup_imsm
7964 * Description: Function recovers critical data from the Migration Copy Area
7965 * while assembling an array.
7967 * super : imsm internal array info
7968 * info : general array info
7970 * 0 : success (or there is no data to recover)
7972 ******************************************************************************/
7973 int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
)
7975 struct intel_super
*super
= st
->sb
;
7976 struct migr_record
*migr_rec
= super
->migr_rec
;
7977 struct imsm_map
*map_dest
= NULL
;
7978 struct intel_dev
*id
= NULL
;
7979 unsigned long long read_offset
;
7980 unsigned long long write_offset
;
7982 int *targets
= NULL
;
7983 int new_disks
, i
, err
;
7986 unsigned long curr_migr_unit
= __le32_to_cpu(migr_rec
->curr_migr_unit
);
7987 unsigned long num_migr_units
= __le32_to_cpu(migr_rec
->num_migr_units
);
7989 int skipped_disks
= 0;
7990 int max_degradation
;
7992 err
= sysfs_get_str(info
, NULL
, "array_state", (char *)buffer
, 20);
7996 /* recover data only during assemblation */
7997 if (strncmp(buffer
, "inactive", 8) != 0)
7999 /* no data to recover */
8000 if (__le32_to_cpu(migr_rec
->rec_status
) == UNIT_SRC_NORMAL
)
8002 if (curr_migr_unit
>= num_migr_units
)
8005 /* find device during reshape */
8006 for (id
= super
->devlist
; id
; id
= id
->next
)
8007 if (is_gen_migration(id
->dev
))
8012 map_dest
= get_imsm_map(id
->dev
, 0);
8013 new_disks
= map_dest
->num_members
;
8014 max_degradation
= new_disks
- imsm_num_data_members(id
->dev
, 0);
8016 read_offset
= (unsigned long long)
8017 __le32_to_cpu(migr_rec
->ckpt_area_pba
) * 512;
8019 write_offset
= ((unsigned long long)
8020 __le32_to_cpu(migr_rec
->dest_1st_member_lba
) +
8021 __le32_to_cpu(map_dest
->pba_of_lba0
)) * 512;
8023 unit_len
= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8024 if (posix_memalign((void **)&buf
, 512, unit_len
) != 0)
8026 targets
= malloc(new_disks
* sizeof(int));
8030 open_backup_targets(info
, new_disks
, targets
);
8032 for (i
= 0; i
< new_disks
; i
++) {
8033 if (targets
[i
] < 0) {
8037 if (lseek64(targets
[i
], read_offset
, SEEK_SET
) < 0) {
8039 Name
": Cannot seek to block: %s\n",
8043 if ((unsigned)read(targets
[i
], buf
, unit_len
) != unit_len
) {
8045 Name
": Cannot read copy area block: %s\n",
8049 if (lseek64(targets
[i
], write_offset
, SEEK_SET
) < 0) {
8051 Name
": Cannot seek to block: %s\n",
8055 if ((unsigned)write(targets
[i
], buf
, unit_len
) != unit_len
) {
8057 Name
": Cannot restore block: %s\n",
8063 if (skipped_disks
> max_degradation
) {
8065 Name
": Cannot restore data from backup."
8066 " Too many failed disks\n");
8070 if (save_checkpoint_imsm(st
, info
, UNIT_SRC_NORMAL
)) {
8071 /* ignore error == 2, this can mean end of reshape here
8073 dprintf("imsm: Cannot write checkpoint to "
8074 "migration record (UNIT_SRC_NORMAL) during restart\n");
8080 for (i
= 0; i
< new_disks
; i
++)
8089 static char disk_by_path
[] = "/dev/disk/by-path/";
8091 static const char *imsm_get_disk_controller_domain(const char *path
)
8093 char disk_path
[PATH_MAX
];
8097 strncpy(disk_path
, disk_by_path
, PATH_MAX
- 1);
8098 strncat(disk_path
, path
, PATH_MAX
- strlen(disk_path
) - 1);
8099 if (stat(disk_path
, &st
) == 0) {
8100 struct sys_dev
* hba
;
8103 path
= devt_to_devpath(st
.st_rdev
);
8106 hba
= find_disk_attached_hba(-1, path
);
8107 if (hba
&& hba
->type
== SYS_DEV_SAS
)
8109 else if (hba
&& hba
->type
== SYS_DEV_SATA
)
8113 dprintf("path: %s hba: %s attached: %s\n",
8114 path
, (hba
) ? hba
->path
: "NULL", drv
);
8122 static int imsm_find_array_minor_by_subdev(int subdev
, int container
, int *minor
)
8124 char subdev_name
[20];
8125 struct mdstat_ent
*mdstat
;
8127 sprintf(subdev_name
, "%d", subdev
);
8128 mdstat
= mdstat_by_subdev(subdev_name
, container
);
8132 *minor
= mdstat
->devnum
;
8133 free_mdstat(mdstat
);
8137 static int imsm_reshape_is_allowed_on_container(struct supertype
*st
,
8138 struct geo_params
*geo
,
8139 int *old_raid_disks
)
8141 /* currently we only support increasing the number of devices
8142 * for a container. This increases the number of device for each
8143 * member array. They must all be RAID0 or RAID5.
8146 struct mdinfo
*info
, *member
;
8147 int devices_that_can_grow
= 0;
8149 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): "
8150 "st->devnum = (%i)\n",
8153 if (geo
->size
!= -1 ||
8154 geo
->level
!= UnSet
||
8155 geo
->layout
!= UnSet
||
8156 geo
->chunksize
!= 0 ||
8157 geo
->raid_disks
== UnSet
) {
8158 dprintf("imsm: Container operation is allowed for "
8159 "raid disks number change only.\n");
8163 info
= container_content_imsm(st
, NULL
);
8164 for (member
= info
; member
; member
= member
->next
) {
8168 dprintf("imsm: checking device_num: %i\n",
8169 member
->container_member
);
8171 if (geo
->raid_disks
<= member
->array
.raid_disks
) {
8172 /* we work on container for Online Capacity Expansion
8173 * only so raid_disks has to grow
8175 dprintf("imsm: for container operation raid disks "
8176 "increase is required\n");
8180 if ((info
->array
.level
!= 0) &&
8181 (info
->array
.level
!= 5)) {
8182 /* we cannot use this container with other raid level
8184 dprintf("imsm: for container operation wrong"
8185 " raid level (%i) detected\n",
8189 /* check for platform support
8190 * for this raid level configuration
8192 struct intel_super
*super
= st
->sb
;
8193 if (!is_raid_level_supported(super
->orom
,
8194 member
->array
.level
,
8196 dprintf("platform does not support raid%d with"
8200 geo
->raid_disks
> 1 ? "s" : "");
8203 /* check if component size is aligned to chunk size
8205 if (info
->component_size
%
8206 (info
->array
.chunk_size
/512)) {
8207 dprintf("Component size is not aligned to "
8213 if (*old_raid_disks
&&
8214 info
->array
.raid_disks
!= *old_raid_disks
)
8216 *old_raid_disks
= info
->array
.raid_disks
;
8218 /* All raid5 and raid0 volumes in container
8219 * have to be ready for Online Capacity Expansion
8220 * so they need to be assembled. We have already
8221 * checked that no recovery etc is happening.
8223 result
= imsm_find_array_minor_by_subdev(member
->container_member
,
8227 dprintf("imsm: cannot find array\n");
8230 devices_that_can_grow
++;
8233 if (!member
&& devices_that_can_grow
)
8237 dprintf("\tContainer operation allowed\n");
8239 dprintf("\tError: %i\n", ret_val
);
8244 /* Function: get_spares_for_grow
8245 * Description: Allocates memory and creates list of spare devices
8246 * avaliable in container. Checks if spare drive size is acceptable.
8247 * Parameters: Pointer to the supertype structure
8248 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
8251 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
)
8253 unsigned long long min_size
= min_acceptable_spare_size_imsm(st
);
8254 return container_choose_spares(st
, min_size
, NULL
, NULL
, NULL
, 0);
8257 /******************************************************************************
8258 * function: imsm_create_metadata_update_for_reshape
8259 * Function creates update for whole IMSM container.
8261 ******************************************************************************/
8262 static int imsm_create_metadata_update_for_reshape(
8263 struct supertype
*st
,
8264 struct geo_params
*geo
,
8266 struct imsm_update_reshape
**updatep
)
8268 struct intel_super
*super
= st
->sb
;
8269 struct imsm_super
*mpb
= super
->anchor
;
8270 int update_memory_size
= 0;
8271 struct imsm_update_reshape
*u
= NULL
;
8272 struct mdinfo
*spares
= NULL
;
8274 int delta_disks
= 0;
8277 dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n",
8280 delta_disks
= geo
->raid_disks
- old_raid_disks
;
8282 /* size of all update data without anchor */
8283 update_memory_size
= sizeof(struct imsm_update_reshape
);
8285 /* now add space for spare disks that we need to add. */
8286 update_memory_size
+= sizeof(u
->new_disks
[0]) * (delta_disks
- 1);
8288 u
= calloc(1, update_memory_size
);
8291 "cannot get memory for imsm_update_reshape update\n");
8294 u
->type
= update_reshape_container_disks
;
8295 u
->old_raid_disks
= old_raid_disks
;
8296 u
->new_raid_disks
= geo
->raid_disks
;
8298 /* now get spare disks list
8300 spares
= get_spares_for_grow(st
);
8303 || delta_disks
> spares
->array
.spare_disks
) {
8304 fprintf(stderr
, Name
": imsm: ERROR: Cannot get spare devices "
8305 "for %s.\n", geo
->dev_name
);
8310 /* we have got spares
8311 * update disk list in imsm_disk list table in anchor
8313 dprintf("imsm: %i spares are available.\n\n",
8314 spares
->array
.spare_disks
);
8317 for (i
= 0; i
< delta_disks
; i
++) {
8322 u
->new_disks
[i
] = makedev(dev
->disk
.major
,
8324 dl
= get_disk_super(super
, dev
->disk
.major
, dev
->disk
.minor
);
8325 dl
->index
= mpb
->num_disks
;
8335 dprintf("imsm: reshape update preparation :");
8336 if (i
== delta_disks
) {
8339 return update_memory_size
;
8342 dprintf(" Error\n");
8347 /******************************************************************************
8348 * function: imsm_create_metadata_update_for_migration()
8349 * Creates update for IMSM array.
8351 ******************************************************************************/
8352 static int imsm_create_metadata_update_for_migration(
8353 struct supertype
*st
,
8354 struct geo_params
*geo
,
8355 struct imsm_update_reshape_migration
**updatep
)
8357 struct intel_super
*super
= st
->sb
;
8358 int update_memory_size
= 0;
8359 struct imsm_update_reshape_migration
*u
= NULL
;
8360 struct imsm_dev
*dev
;
8361 int previous_level
= -1;
8363 dprintf("imsm_create_metadata_update_for_migration(enter)"
8364 " New Level = %i\n", geo
->level
);
8366 /* size of all update data without anchor */
8367 update_memory_size
= sizeof(struct imsm_update_reshape_migration
);
8369 u
= calloc(1, update_memory_size
);
8371 dprintf("error: cannot get memory for "
8372 "imsm_create_metadata_update_for_migration\n");
8375 u
->type
= update_reshape_migration
;
8376 u
->subdev
= super
->current_vol
;
8377 u
->new_level
= geo
->level
;
8378 u
->new_layout
= geo
->layout
;
8379 u
->new_raid_disks
= u
->old_raid_disks
= geo
->raid_disks
;
8380 u
->new_disks
[0] = -1;
8381 u
->new_chunksize
= -1;
8383 dev
= get_imsm_dev(super
, u
->subdev
);
8385 struct imsm_map
*map
;
8387 map
= get_imsm_map(dev
, 0);
8389 int current_chunk_size
=
8390 __le16_to_cpu(map
->blocks_per_strip
) / 2;
8392 if (geo
->chunksize
!= current_chunk_size
) {
8393 u
->new_chunksize
= geo
->chunksize
/ 1024;
8395 "chunk size change from %i to %i\n",
8396 current_chunk_size
, u
->new_chunksize
);
8398 previous_level
= map
->raid_level
;
8401 if ((geo
->level
== 5) && (previous_level
== 0)) {
8402 struct mdinfo
*spares
= NULL
;
8404 u
->new_raid_disks
++;
8405 spares
= get_spares_for_grow(st
);
8406 if ((spares
== NULL
) || (spares
->array
.spare_disks
< 1)) {
8409 update_memory_size
= 0;
8410 dprintf("error: cannot get spare device "
8411 "for requested migration");
8416 dprintf("imsm: reshape update preparation : OK\n");
8419 return update_memory_size
;
8422 static void imsm_update_metadata_locally(struct supertype
*st
,
8425 struct metadata_update mu
;
8430 mu
.space_list
= NULL
;
8432 imsm_prepare_update(st
, &mu
);
8433 imsm_process_update(st
, &mu
);
8435 while (mu
.space_list
) {
8436 void **space
= mu
.space_list
;
8437 mu
.space_list
= *space
;
8442 /***************************************************************************
8443 * Function: imsm_analyze_change
8444 * Description: Function analyze change for single volume
8445 * and validate if transition is supported
8446 * Parameters: Geometry parameters, supertype structure
8447 * Returns: Operation type code on success, -1 if fail
8448 ****************************************************************************/
8449 enum imsm_reshape_type
imsm_analyze_change(struct supertype
*st
,
8450 struct geo_params
*geo
)
8457 getinfo_super_imsm_volume(st
, &info
, NULL
);
8458 if ((geo
->level
!= info
.array
.level
) &&
8459 (geo
->level
>= 0) &&
8460 (geo
->level
!= UnSet
)) {
8461 switch (info
.array
.level
) {
8463 if (geo
->level
== 5) {
8464 change
= CH_MIGRATION
;
8465 if (geo
->layout
!= ALGORITHM_LEFT_ASYMMETRIC
) {
8467 Name
" Error. Requested Layout "
8468 "not supported (left-asymmetric layout "
8469 "is supported only)!\n");
8471 goto analyse_change_exit
;
8475 if (geo
->level
== 10) {
8476 change
= CH_TAKEOVER
;
8481 if (geo
->level
== 0) {
8482 change
= CH_TAKEOVER
;
8487 if (geo
->level
== 0) {
8488 change
= CH_TAKEOVER
;
8495 Name
" Error. Level Migration from %d to %d "
8497 info
.array
.level
, geo
->level
);
8498 goto analyse_change_exit
;
8501 geo
->level
= info
.array
.level
;
8503 if ((geo
->layout
!= info
.array
.layout
)
8504 && ((geo
->layout
!= UnSet
) && (geo
->layout
!= -1))) {
8505 change
= CH_MIGRATION
;
8506 if ((info
.array
.layout
== 0)
8507 && (info
.array
.level
== 5)
8508 && (geo
->layout
== 5)) {
8509 /* reshape 5 -> 4 */
8510 } else if ((info
.array
.layout
== 5)
8511 && (info
.array
.level
== 5)
8512 && (geo
->layout
== 0)) {
8513 /* reshape 4 -> 5 */
8518 Name
" Error. Layout Migration from %d to %d "
8520 info
.array
.layout
, geo
->layout
);
8522 goto analyse_change_exit
;
8525 geo
->layout
= info
.array
.layout
;
8527 if ((geo
->chunksize
> 0) && (geo
->chunksize
!= UnSet
)
8528 && (geo
->chunksize
!= info
.array
.chunk_size
))
8529 change
= CH_MIGRATION
;
8531 geo
->chunksize
= info
.array
.chunk_size
;
8533 chunk
= geo
->chunksize
/ 1024;
8534 if (!validate_geometry_imsm(st
,
8544 struct intel_super
*super
= st
->sb
;
8545 struct imsm_super
*mpb
= super
->anchor
;
8547 if (mpb
->num_raid_devs
> 1) {
8549 Name
" Error. Cannot perform operation on %s"
8550 "- for this operation it MUST be single "
8551 "array in container\n",
8557 analyse_change_exit
:
8562 int imsm_takeover(struct supertype
*st
, struct geo_params
*geo
)
8564 struct intel_super
*super
= st
->sb
;
8565 struct imsm_update_takeover
*u
;
8567 u
= malloc(sizeof(struct imsm_update_takeover
));
8571 u
->type
= update_takeover
;
8572 u
->subarray
= super
->current_vol
;
8574 /* 10->0 transition */
8575 if (geo
->level
== 0)
8576 u
->direction
= R10_TO_R0
;
8578 /* 0->10 transition */
8579 if (geo
->level
== 10)
8580 u
->direction
= R0_TO_R10
;
8582 /* update metadata locally */
8583 imsm_update_metadata_locally(st
, u
,
8584 sizeof(struct imsm_update_takeover
));
8585 /* and possibly remotely */
8586 if (st
->update_tail
)
8587 append_metadata_update(st
, u
,
8588 sizeof(struct imsm_update_takeover
));
8595 static int imsm_reshape_super(struct supertype
*st
, long long size
, int level
,
8596 int layout
, int chunksize
, int raid_disks
,
8597 int delta_disks
, char *backup
, char *dev
,
8601 struct geo_params geo
;
8603 dprintf("imsm: reshape_super called.\n");
8605 memset(&geo
, 0, sizeof(struct geo_params
));
8608 geo
.dev_id
= st
->devnum
;
8611 geo
.layout
= layout
;
8612 geo
.chunksize
= chunksize
;
8613 geo
.raid_disks
= raid_disks
;
8614 if (delta_disks
!= UnSet
)
8615 geo
.raid_disks
+= delta_disks
;
8617 dprintf("\tfor level : %i\n", geo
.level
);
8618 dprintf("\tfor raid_disks : %i\n", geo
.raid_disks
);
8620 if (experimental() == 0)
8623 if (st
->container_dev
== st
->devnum
) {
8624 /* On container level we can only increase number of devices. */
8625 dprintf("imsm: info: Container operation\n");
8626 int old_raid_disks
= 0;
8628 if (imsm_reshape_is_allowed_on_container(
8629 st
, &geo
, &old_raid_disks
)) {
8630 struct imsm_update_reshape
*u
= NULL
;
8633 len
= imsm_create_metadata_update_for_reshape(
8634 st
, &geo
, old_raid_disks
, &u
);
8637 dprintf("imsm: Cannot prepare update\n");
8638 goto exit_imsm_reshape_super
;
8642 /* update metadata locally */
8643 imsm_update_metadata_locally(st
, u
, len
);
8644 /* and possibly remotely */
8645 if (st
->update_tail
)
8646 append_metadata_update(st
, u
, len
);
8651 fprintf(stderr
, Name
": (imsm) Operation "
8652 "is not allowed on this container\n");
8655 /* On volume level we support following operations
8656 * - takeover: raid10 -> raid0; raid0 -> raid10
8657 * - chunk size migration
8658 * - migration: raid5 -> raid0; raid0 -> raid5
8660 struct intel_super
*super
= st
->sb
;
8661 struct intel_dev
*dev
= super
->devlist
;
8663 dprintf("imsm: info: Volume operation\n");
8664 /* find requested device */
8666 if (imsm_find_array_minor_by_subdev(
8667 dev
->index
, st
->container_dev
, &devnum
) == 0
8668 && devnum
== geo
.dev_id
)
8673 fprintf(stderr
, Name
" Cannot find %s (%i) subarray\n",
8674 geo
.dev_name
, geo
.dev_id
);
8675 goto exit_imsm_reshape_super
;
8677 super
->current_vol
= dev
->index
;
8678 change
= imsm_analyze_change(st
, &geo
);
8681 ret_val
= imsm_takeover(st
, &geo
);
8683 case CH_MIGRATION
: {
8684 struct imsm_update_reshape_migration
*u
= NULL
;
8686 imsm_create_metadata_update_for_migration(
8690 "Cannot prepare update\n");
8694 /* update metadata locally */
8695 imsm_update_metadata_locally(st
, u
, len
);
8696 /* and possibly remotely */
8697 if (st
->update_tail
)
8698 append_metadata_update(st
, u
, len
);
8708 exit_imsm_reshape_super
:
8709 dprintf("imsm: reshape_super Exit code = %i\n", ret_val
);
8713 /*******************************************************************************
8714 * Function: wait_for_reshape_imsm
8715 * Description: Function writes new sync_max value and waits until
8716 * reshape process reach new position
8718 * sra : general array info
8719 * ndata : number of disks in new array's layout
8722 * 1 : there is no reshape in progress,
8724 ******************************************************************************/
8725 int wait_for_reshape_imsm(struct mdinfo
*sra
, int ndata
)
8727 int fd
= sysfs_get_fd(sra
, NULL
, "reshape_position");
8728 unsigned long long completed
;
8729 /* to_complete : new sync_max position */
8730 unsigned long long to_complete
= sra
->reshape_progress
;
8731 unsigned long long position_to_set
= to_complete
/ ndata
;
8734 dprintf("imsm: wait_for_reshape_imsm() "
8735 "cannot open reshape_position\n");
8739 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8740 dprintf("imsm: wait_for_reshape_imsm() "
8741 "cannot read reshape_position (no reshape in progres)\n");
8746 if (completed
> to_complete
) {
8747 dprintf("imsm: wait_for_reshape_imsm() "
8748 "wrong next position to set %llu (%llu)\n",
8749 to_complete
, completed
);
8753 dprintf("Position set: %llu\n", position_to_set
);
8754 if (sysfs_set_num(sra
, NULL
, "sync_max",
8755 position_to_set
) != 0) {
8756 dprintf("imsm: wait_for_reshape_imsm() "
8757 "cannot set reshape position to %llu\n",
8768 select(fd
+1, &rfds
, NULL
, NULL
, NULL
);
8769 if (sysfs_get_str(sra
, NULL
, "sync_action",
8771 strncmp(action
, "reshape", 7) != 0)
8773 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8774 dprintf("imsm: wait_for_reshape_imsm() "
8775 "cannot read reshape_position (in loop)\n");
8779 } while (completed
< to_complete
);
8785 /*******************************************************************************
8786 * Function: check_degradation_change
8787 * Description: Check that array hasn't become failed.
8789 * info : for sysfs access
8790 * sources : source disks descriptors
8791 * degraded: previous degradation level
8794 ******************************************************************************/
8795 int check_degradation_change(struct mdinfo
*info
,
8799 unsigned long long new_degraded
;
8800 sysfs_get_ll(info
, NULL
, "degraded", &new_degraded
);
8801 if (new_degraded
!= (unsigned long long)degraded
) {
8802 /* check each device to ensure it is still working */
8805 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
8806 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
))
8808 if (sd
->disk
.state
& (1<<MD_DISK_SYNC
)) {
8810 if (sysfs_get_str(info
,
8811 sd
, "state", sbuf
, 20) < 0 ||
8812 strstr(sbuf
, "faulty") ||
8813 strstr(sbuf
, "in_sync") == NULL
) {
8814 /* this device is dead */
8815 sd
->disk
.state
= (1<<MD_DISK_FAULTY
);
8816 if (sd
->disk
.raid_disk
>= 0 &&
8817 sources
[sd
->disk
.raid_disk
] >= 0) {
8819 sd
->disk
.raid_disk
]);
8820 sources
[sd
->disk
.raid_disk
] =
8829 return new_degraded
;
8832 /*******************************************************************************
8833 * Function: imsm_manage_reshape
8834 * Description: Function finds array under reshape and it manages reshape
8835 * process. It creates stripes backups (if required) and sets
8838 * afd : Backup handle (nattive) - not used
8839 * sra : general array info
8840 * reshape : reshape parameters - not used
8841 * st : supertype structure
8842 * blocks : size of critical section [blocks]
8843 * fds : table of source device descriptor
8844 * offsets : start of array (offest per devices)
8846 * destfd : table of destination device descriptor
8847 * destoffsets : table of destination offsets (per device)
8849 * 1 : success, reshape is done
8851 ******************************************************************************/
8852 static int imsm_manage_reshape(
8853 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
8854 struct supertype
*st
, unsigned long backup_blocks
,
8855 int *fds
, unsigned long long *offsets
,
8856 int dests
, int *destfd
, unsigned long long *destoffsets
)
8859 struct intel_super
*super
= st
->sb
;
8860 struct intel_dev
*dv
= NULL
;
8861 struct imsm_dev
*dev
= NULL
;
8862 struct imsm_map
*map_src
;
8863 int migr_vol_qan
= 0;
8864 int ndata
, odata
; /* [bytes] */
8865 int chunk
; /* [bytes] */
8866 struct migr_record
*migr_rec
;
8868 unsigned int buf_size
; /* [bytes] */
8869 unsigned long long max_position
; /* array size [bytes] */
8870 unsigned long long next_step
; /* [blocks]/[bytes] */
8871 unsigned long long old_data_stripe_length
;
8872 unsigned long long start_src
; /* [bytes] */
8873 unsigned long long start
; /* [bytes] */
8874 unsigned long long start_buf_shift
; /* [bytes] */
8876 int source_layout
= 0;
8878 if (!fds
|| !offsets
|| !sra
)
8881 /* Find volume during the reshape */
8882 for (dv
= super
->devlist
; dv
; dv
= dv
->next
) {
8883 if (dv
->dev
->vol
.migr_type
== MIGR_GEN_MIGR
8884 && dv
->dev
->vol
.migr_state
== 1) {
8889 /* Only one volume can migrate at the same time */
8890 if (migr_vol_qan
!= 1) {
8891 fprintf(stderr
, Name
" : %s", migr_vol_qan
?
8892 "Number of migrating volumes greater than 1\n" :
8893 "There is no volume during migrationg\n");
8897 map_src
= get_imsm_map(dev
, 1);
8898 if (map_src
== NULL
)
8901 ndata
= imsm_num_data_members(dev
, 0);
8902 odata
= imsm_num_data_members(dev
, 1);
8904 chunk
= __le16_to_cpu(map_src
->blocks_per_strip
) * 512;
8905 old_data_stripe_length
= odata
* chunk
;
8907 migr_rec
= super
->migr_rec
;
8909 /* initialize migration record for start condition */
8910 if (sra
->reshape_progress
== 0)
8911 init_migr_record_imsm(st
, dev
, sra
);
8913 if (__le32_to_cpu(migr_rec
->rec_status
) != UNIT_SRC_NORMAL
) {
8914 dprintf("imsm: cannot restart migration when data "
8915 "are present in copy area.\n");
8921 buf_size
= __le32_to_cpu(migr_rec
->blocks_per_unit
) * 512;
8922 /* extend buffer size for parity disk */
8923 buf_size
+= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8924 /* add space for stripe aligment */
8925 buf_size
+= old_data_stripe_length
;
8926 if (posix_memalign((void **)&buf
, 4096, buf_size
)) {
8927 dprintf("imsm: Cannot allocate checpoint buffer\n");
8931 max_position
= sra
->component_size
* ndata
;
8932 source_layout
= imsm_level_to_layout(map_src
->raid_level
);
8934 while (__le32_to_cpu(migr_rec
->curr_migr_unit
) <
8935 __le32_to_cpu(migr_rec
->num_migr_units
)) {
8936 /* current reshape position [blocks] */
8937 unsigned long long current_position
=
8938 __le32_to_cpu(migr_rec
->blocks_per_unit
)
8939 * __le32_to_cpu(migr_rec
->curr_migr_unit
);
8940 unsigned long long border
;
8942 /* Check that array hasn't become failed.
8944 degraded
= check_degradation_change(sra
, fds
, degraded
);
8946 dprintf("imsm: Abort reshape due to degradation"
8947 " level (%i)\n", degraded
);
8951 next_step
= __le32_to_cpu(migr_rec
->blocks_per_unit
);
8953 if ((current_position
+ next_step
) > max_position
)
8954 next_step
= max_position
- current_position
;
8956 start
= current_position
* 512;
8958 /* allign reading start to old geometry */
8959 start_buf_shift
= start
% old_data_stripe_length
;
8960 start_src
= start
- start_buf_shift
;
8962 border
= (start_src
/ odata
) - (start
/ ndata
);
8964 if (border
<= __le32_to_cpu(migr_rec
->dest_depth_per_unit
)) {
8965 /* save critical stripes to buf
8966 * start - start address of current unit
8968 * start_src - start address of current unit
8969 * to backup alligned to source array
8972 unsigned long long next_step_filler
= 0;
8973 unsigned long long copy_length
= next_step
* 512;
8975 /* allign copy area length to stripe in old geometry */
8976 next_step_filler
= ((copy_length
+ start_buf_shift
)
8977 % old_data_stripe_length
);
8978 if (next_step_filler
)
8979 next_step_filler
= (old_data_stripe_length
8980 - next_step_filler
);
8981 dprintf("save_stripes() parameters: start = %llu,"
8982 "\tstart_src = %llu,\tnext_step*512 = %llu,"
8983 "\tstart_in_buf_shift = %llu,"
8984 "\tnext_step_filler = %llu\n",
8985 start
, start_src
, copy_length
,
8986 start_buf_shift
, next_step_filler
);
8988 if (save_stripes(fds
, offsets
, map_src
->num_members
,
8989 chunk
, map_src
->raid_level
,
8990 source_layout
, 0, NULL
, start_src
,
8992 next_step_filler
+ start_buf_shift
,
8994 dprintf("imsm: Cannot save stripes"
8998 /* Convert data to destination format and store it
8999 * in backup general migration area
9001 if (save_backup_imsm(st
, dev
, sra
,
9002 buf
+ start_buf_shift
, copy_length
)) {
9003 dprintf("imsm: Cannot save stripes to "
9004 "target devices\n");
9007 if (save_checkpoint_imsm(st
, sra
,
9008 UNIT_SRC_IN_CP_AREA
)) {
9009 dprintf("imsm: Cannot write checkpoint to "
9010 "migration record (UNIT_SRC_IN_CP_AREA)\n");
9014 /* set next step to use whole border area */
9015 border
/= next_step
;
9017 next_step
*= border
;
9019 /* When data backed up, checkpoint stored,
9020 * kick the kernel to reshape unit of data
9022 next_step
= next_step
+ sra
->reshape_progress
;
9023 /* limit next step to array max position */
9024 if (next_step
> max_position
)
9025 next_step
= max_position
;
9026 sysfs_set_num(sra
, NULL
, "suspend_lo", sra
->reshape_progress
);
9027 sysfs_set_num(sra
, NULL
, "suspend_hi", next_step
);
9028 sra
->reshape_progress
= next_step
;
9030 /* wait until reshape finish */
9031 if (wait_for_reshape_imsm(sra
, ndata
) < 0) {
9032 dprintf("wait_for_reshape_imsm returned error!\n");
9036 if (save_checkpoint_imsm(st
, sra
, UNIT_SRC_NORMAL
) == 1) {
9037 /* ignore error == 2, this can mean end of reshape here
9039 dprintf("imsm: Cannot write checkpoint to "
9040 "migration record (UNIT_SRC_NORMAL)\n");
9046 /* return '1' if done */
9054 #endif /* MDASSEMBLE */
9056 struct superswitch super_imsm
= {
9058 .examine_super
= examine_super_imsm
,
9059 .brief_examine_super
= brief_examine_super_imsm
,
9060 .brief_examine_subarrays
= brief_examine_subarrays_imsm
,
9061 .export_examine_super
= export_examine_super_imsm
,
9062 .detail_super
= detail_super_imsm
,
9063 .brief_detail_super
= brief_detail_super_imsm
,
9064 .write_init_super
= write_init_super_imsm
,
9065 .validate_geometry
= validate_geometry_imsm
,
9066 .add_to_super
= add_to_super_imsm
,
9067 .remove_from_super
= remove_from_super_imsm
,
9068 .detail_platform
= detail_platform_imsm
,
9069 .kill_subarray
= kill_subarray_imsm
,
9070 .update_subarray
= update_subarray_imsm
,
9071 .load_container
= load_container_imsm
,
9072 .default_geometry
= default_geometry_imsm
,
9073 .get_disk_controller_domain
= imsm_get_disk_controller_domain
,
9074 .reshape_super
= imsm_reshape_super
,
9075 .manage_reshape
= imsm_manage_reshape
,
9076 .recover_backup
= recover_backup_imsm
,
9078 .match_home
= match_home_imsm
,
9079 .uuid_from_super
= uuid_from_super_imsm
,
9080 .getinfo_super
= getinfo_super_imsm
,
9081 .getinfo_super_disks
= getinfo_super_disks_imsm
,
9082 .update_super
= update_super_imsm
,
9084 .avail_size
= avail_size_imsm
,
9085 .min_acceptable_spare_size
= min_acceptable_spare_size_imsm
,
9087 .compare_super
= compare_super_imsm
,
9089 .load_super
= load_super_imsm
,
9090 .init_super
= init_super_imsm
,
9091 .store_super
= store_super_imsm
,
9092 .free_super
= free_super_imsm
,
9093 .match_metadata_desc
= match_metadata_desc_imsm
,
9094 .container_content
= container_content_imsm
,
9102 .open_new
= imsm_open_new
,
9103 .set_array_state
= imsm_set_array_state
,
9104 .set_disk
= imsm_set_disk
,
9105 .sync_metadata
= imsm_sync_metadata
,
9106 .activate_spare
= imsm_activate_spare
,
9107 .process_update
= imsm_process_update
,
9108 .prepare_update
= imsm_prepare_update
,
9109 #endif /* MDASSEMBLE */