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 attributes that are unused but not harmful */
87 #define MPB_ATTRIB_IGNORED (MPB_ATTRIB_NEVER_USE)
89 #define MPB_SECTOR_CNT 2210
90 #define IMSM_RESERVED_SECTORS 4096
91 #define SECT_PER_MB_SHIFT 11
93 /* Disk configuration info. */
94 #define IMSM_MAX_DEVICES 255
96 __u8 serial
[MAX_RAID_SERIAL_LEN
];/* 0xD8 - 0xE7 ascii serial number */
97 __u32 total_blocks
; /* 0xE8 - 0xEB total blocks */
98 __u32 scsi_id
; /* 0xEC - 0xEF scsi ID */
99 #define SPARE_DISK __cpu_to_le32(0x01) /* Spare */
100 #define CONFIGURED_DISK __cpu_to_le32(0x02) /* Member of some RaidDev */
101 #define FAILED_DISK __cpu_to_le32(0x04) /* Permanent failure */
102 __u32 status
; /* 0xF0 - 0xF3 */
103 __u32 owner_cfg_num
; /* which config 0,1,2... owns this disk */
104 #define IMSM_DISK_FILLERS 4
105 __u32 filler
[IMSM_DISK_FILLERS
]; /* 0xF4 - 0x107 MPB_DISK_FILLERS for future expansion */
108 /* RAID map configuration infos. */
110 __u32 pba_of_lba0
; /* start address of partition */
111 __u32 blocks_per_member
;/* blocks per member */
112 __u32 num_data_stripes
; /* number of data stripes */
113 __u16 blocks_per_strip
;
114 __u8 map_state
; /* Normal, Uninitialized, Degraded, Failed */
115 #define IMSM_T_STATE_NORMAL 0
116 #define IMSM_T_STATE_UNINITIALIZED 1
117 #define IMSM_T_STATE_DEGRADED 2
118 #define IMSM_T_STATE_FAILED 3
120 #define IMSM_T_RAID0 0
121 #define IMSM_T_RAID1 1
122 #define IMSM_T_RAID5 5 /* since metadata version 1.2.02 ? */
123 __u8 num_members
; /* number of member disks */
124 __u8 num_domains
; /* number of parity domains */
125 __u8 failed_disk_num
; /* valid only when state is degraded */
127 __u32 filler
[7]; /* expansion area */
128 #define IMSM_ORD_REBUILD (1 << 24)
129 __u32 disk_ord_tbl
[1]; /* disk_ord_tbl[num_members],
130 * top byte contains some flags
132 } __attribute__ ((packed
));
135 __u32 curr_migr_unit
;
136 __u32 checkpoint_id
; /* id to access curr_migr_unit */
137 __u8 migr_state
; /* Normal or Migrating */
139 #define MIGR_REBUILD 1
140 #define MIGR_VERIFY 2 /* analagous to echo check > sync_action */
141 #define MIGR_GEN_MIGR 3
142 #define MIGR_STATE_CHANGE 4
143 #define MIGR_REPAIR 5
144 __u8 migr_type
; /* Initializing, Rebuilding, ... */
146 __u8 fs_state
; /* fast-sync state for CnG (0xff == disabled) */
147 __u16 verify_errors
; /* number of mismatches */
148 __u16 bad_blocks
; /* number of bad blocks during verify */
150 struct imsm_map map
[1];
151 /* here comes another one if migr_state */
152 } __attribute__ ((packed
));
155 __u8 volume
[MAX_RAID_SERIAL_LEN
];
158 #define DEV_BOOTABLE __cpu_to_le32(0x01)
159 #define DEV_BOOT_DEVICE __cpu_to_le32(0x02)
160 #define DEV_READ_COALESCING __cpu_to_le32(0x04)
161 #define DEV_WRITE_COALESCING __cpu_to_le32(0x08)
162 #define DEV_LAST_SHUTDOWN_DIRTY __cpu_to_le32(0x10)
163 #define DEV_HIDDEN_AT_BOOT __cpu_to_le32(0x20)
164 #define DEV_CURRENTLY_HIDDEN __cpu_to_le32(0x40)
165 #define DEV_VERIFY_AND_FIX __cpu_to_le32(0x80)
166 #define DEV_MAP_STATE_UNINIT __cpu_to_le32(0x100)
167 #define DEV_NO_AUTO_RECOVERY __cpu_to_le32(0x200)
168 #define DEV_CLONE_N_GO __cpu_to_le32(0x400)
169 #define DEV_CLONE_MAN_SYNC __cpu_to_le32(0x800)
170 #define DEV_CNG_MASTER_DISK_NUM __cpu_to_le32(0x1000)
171 __u32 status
; /* Persistent RaidDev status */
172 __u32 reserved_blocks
; /* Reserved blocks at beginning of volume */
176 __u8 cng_master_disk
;
180 #define IMSM_DEV_FILLERS 10
181 __u32 filler
[IMSM_DEV_FILLERS
];
183 } __attribute__ ((packed
));
186 __u8 sig
[MAX_SIGNATURE_LENGTH
]; /* 0x00 - 0x1F */
187 __u32 check_sum
; /* 0x20 - 0x23 MPB Checksum */
188 __u32 mpb_size
; /* 0x24 - 0x27 Size of MPB */
189 __u32 family_num
; /* 0x28 - 0x2B Checksum from first time this config was written */
190 __u32 generation_num
; /* 0x2C - 0x2F Incremented each time this array's MPB is written */
191 __u32 error_log_size
; /* 0x30 - 0x33 in bytes */
192 __u32 attributes
; /* 0x34 - 0x37 */
193 __u8 num_disks
; /* 0x38 Number of configured disks */
194 __u8 num_raid_devs
; /* 0x39 Number of configured volumes */
195 __u8 error_log_pos
; /* 0x3A */
196 __u8 fill
[1]; /* 0x3B */
197 __u32 cache_size
; /* 0x3c - 0x40 in mb */
198 __u32 orig_family_num
; /* 0x40 - 0x43 original family num */
199 __u32 pwr_cycle_count
; /* 0x44 - 0x47 simulated power cycle count for array */
200 __u32 bbm_log_size
; /* 0x48 - 0x4B - size of bad Block Mgmt Log in bytes */
201 #define IMSM_FILLERS 35
202 __u32 filler
[IMSM_FILLERS
]; /* 0x4C - 0xD7 RAID_MPB_FILLERS */
203 struct imsm_disk disk
[1]; /* 0xD8 diskTbl[numDisks] */
204 /* here comes imsm_dev[num_raid_devs] */
205 /* here comes BBM logs */
206 } __attribute__ ((packed
));
208 #define BBM_LOG_MAX_ENTRIES 254
210 struct bbm_log_entry
{
211 __u64 defective_block_start
;
212 #define UNREADABLE 0xFFFFFFFF
213 __u32 spare_block_offset
;
214 __u16 remapped_marked_count
;
216 } __attribute__ ((__packed__
));
219 __u32 signature
; /* 0xABADB10C */
221 __u32 reserved_spare_block_count
; /* 0 */
222 __u32 reserved
; /* 0xFFFF */
223 __u64 first_spare_lba
;
224 struct bbm_log_entry mapped_block_entries
[BBM_LOG_MAX_ENTRIES
];
225 } __attribute__ ((__packed__
));
229 static char *map_state_str
[] = { "normal", "uninitialized", "degraded", "failed" };
232 #define RAID_DISK_RESERVED_BLOCKS_IMSM_HI 2209
234 #define GEN_MIGR_AREA_SIZE 2048 /* General Migration Copy Area size in blocks */
236 #define UNIT_SRC_NORMAL 0 /* Source data for curr_migr_unit must
237 * be recovered using srcMap */
238 #define UNIT_SRC_IN_CP_AREA 1 /* Source data for curr_migr_unit has
239 * already been migrated and must
240 * be recovered from checkpoint area */
242 __u32 rec_status
; /* Status used to determine how to restart
243 * migration in case it aborts
245 __u32 curr_migr_unit
; /* 0..numMigrUnits-1 */
246 __u32 family_num
; /* Family number of MPB
247 * containing the RaidDev
248 * that is migrating */
249 __u32 ascending_migr
; /* True if migrating in increasing
251 __u32 blocks_per_unit
; /* Num disk blocks per unit of operation */
252 __u32 dest_depth_per_unit
; /* Num member blocks each destMap
254 * advances per unit-of-operation */
255 __u32 ckpt_area_pba
; /* Pba of first block of ckpt copy area */
256 __u32 dest_1st_member_lba
; /* First member lba on first
257 * stripe of destination */
258 __u32 num_migr_units
; /* Total num migration units-of-op */
259 __u32 post_migr_vol_cap
; /* Size of volume after
260 * migration completes */
261 __u32 post_migr_vol_cap_hi
; /* Expansion space for LBA64 */
262 __u32 ckpt_read_disk_num
; /* Which member disk in destSubMap[0] the
263 * migration ckpt record was read from
264 * (for recovered migrations) */
265 } __attribute__ ((__packed__
));
267 static __u8
migr_type(struct imsm_dev
*dev
)
269 if (dev
->vol
.migr_type
== MIGR_VERIFY
&&
270 dev
->status
& DEV_VERIFY_AND_FIX
)
273 return dev
->vol
.migr_type
;
276 static void set_migr_type(struct imsm_dev
*dev
, __u8 migr_type
)
278 /* for compatibility with older oroms convert MIGR_REPAIR, into
279 * MIGR_VERIFY w/ DEV_VERIFY_AND_FIX status
281 if (migr_type
== MIGR_REPAIR
) {
282 dev
->vol
.migr_type
= MIGR_VERIFY
;
283 dev
->status
|= DEV_VERIFY_AND_FIX
;
285 dev
->vol
.migr_type
= migr_type
;
286 dev
->status
&= ~DEV_VERIFY_AND_FIX
;
290 static unsigned int sector_count(__u32 bytes
)
292 return ((bytes
+ (512-1)) & (~(512-1))) / 512;
295 static unsigned int mpb_sectors(struct imsm_super
*mpb
)
297 return sector_count(__le32_to_cpu(mpb
->mpb_size
));
301 struct imsm_dev
*dev
;
302 struct intel_dev
*next
;
307 enum sys_dev_type type
;
310 struct intel_hba
*next
;
317 /* internal representation of IMSM metadata */
320 void *buf
; /* O_DIRECT buffer for reading/writing metadata */
321 struct imsm_super
*anchor
; /* immovable parameters */
324 void *migr_rec_buf
; /* buffer for I/O operations */
325 struct migr_record
*migr_rec
; /* migration record */
327 size_t len
; /* size of the 'buf' allocation */
328 void *next_buf
; /* for realloc'ing buf from the manager */
330 int updates_pending
; /* count of pending updates for mdmon */
331 int current_vol
; /* index of raid device undergoing creation */
332 __u32 create_offset
; /* common start for 'current_vol' */
333 __u32 random
; /* random data for seeding new family numbers */
334 struct intel_dev
*devlist
;
338 __u8 serial
[MAX_RAID_SERIAL_LEN
];
341 struct imsm_disk disk
;
344 struct extent
*e
; /* for determining freespace @ create */
345 int raiddisk
; /* slot to fill in autolayout */
347 } *disks
, *current_disk
;
348 struct dl
*disk_mgmt_list
; /* list of disks to add/remove while mdmon
350 struct dl
*missing
; /* disks removed while we weren't looking */
351 struct bbm_log
*bbm_log
;
352 struct intel_hba
*hba
; /* device path of the raid controller for this metadata */
353 const struct imsm_orom
*orom
; /* platform firmware support */
354 struct intel_super
*next
; /* (temp) list for disambiguating family_num */
358 struct imsm_disk disk
;
359 #define IMSM_UNKNOWN_OWNER (-1)
361 struct intel_disk
*next
;
365 unsigned long long start
, size
;
368 /* definitions of reshape process types */
369 enum imsm_reshape_type
{
374 /* definition of messages passed to imsm_process_update */
375 enum imsm_update_type
{
376 update_activate_spare
,
380 update_add_remove_disk
,
381 update_reshape_container_disks
,
382 update_reshape_migration
,
384 update_general_migration_checkpoint
,
387 struct imsm_update_activate_spare
{
388 enum imsm_update_type type
;
392 struct imsm_update_activate_spare
*next
;
405 enum takeover_direction
{
409 struct imsm_update_takeover
{
410 enum imsm_update_type type
;
412 enum takeover_direction direction
;
415 struct imsm_update_reshape
{
416 enum imsm_update_type type
;
420 int new_disks
[1]; /* new_raid_disks - old_raid_disks makedev number */
423 struct imsm_update_reshape_migration
{
424 enum imsm_update_type type
;
427 /* fields for array migration changes
434 int new_disks
[1]; /* new_raid_disks - old_raid_disks makedev number */
437 struct imsm_update_general_migration_checkpoint
{
438 enum imsm_update_type type
;
439 __u32 curr_migr_unit
;
443 __u8 serial
[MAX_RAID_SERIAL_LEN
];
446 struct imsm_update_create_array
{
447 enum imsm_update_type type
;
452 struct imsm_update_kill_array
{
453 enum imsm_update_type type
;
457 struct imsm_update_rename_array
{
458 enum imsm_update_type type
;
459 __u8 name
[MAX_RAID_SERIAL_LEN
];
463 struct imsm_update_add_remove_disk
{
464 enum imsm_update_type type
;
468 static const char *_sys_dev_type
[] = {
469 [SYS_DEV_UNKNOWN
] = "Unknown",
470 [SYS_DEV_SAS
] = "SAS",
471 [SYS_DEV_SATA
] = "SATA"
474 const char *get_sys_dev_type(enum sys_dev_type type
)
476 if (type
>= SYS_DEV_MAX
)
477 type
= SYS_DEV_UNKNOWN
;
479 return _sys_dev_type
[type
];
482 static struct intel_hba
* alloc_intel_hba(struct sys_dev
*device
)
484 struct intel_hba
*result
= malloc(sizeof(*result
));
486 result
->type
= device
->type
;
487 result
->path
= strdup(device
->path
);
489 if (result
->path
&& (result
->pci_id
= strrchr(result
->path
, '/')) != NULL
)
495 static struct intel_hba
* find_intel_hba(struct intel_hba
*hba
, struct sys_dev
*device
)
497 struct intel_hba
*result
=NULL
;
498 for (result
= hba
; result
; result
= result
->next
) {
499 if (result
->type
== device
->type
&& strcmp(result
->path
, device
->path
) == 0)
505 static int attach_hba_to_super(struct intel_super
*super
, struct sys_dev
*device
)
507 struct intel_hba
*hba
;
509 /* check if disk attached to Intel HBA */
510 hba
= find_intel_hba(super
->hba
, device
);
513 /* Check if HBA is already attached to super */
514 if (super
->hba
== NULL
) {
515 super
->hba
= alloc_intel_hba(device
);
520 /* Intel metadata allows for all disks attached to the same type HBA.
521 * Do not sypport odf HBA types mixing
523 if (device
->type
!= hba
->type
)
529 hba
->next
= alloc_intel_hba(device
);
533 static struct sys_dev
* find_disk_attached_hba(int fd
, const char *devname
)
535 struct sys_dev
*list
, *elem
, *prev
;
538 if ((list
= find_intel_devices()) == NULL
)
542 disk_path
= (char *) devname
;
544 disk_path
= diskfd_to_devpath(fd
);
551 for (prev
= NULL
, elem
= list
; elem
; prev
= elem
, elem
= elem
->next
) {
552 if (path_attached_to_hba(disk_path
, elem
->path
)) {
556 prev
->next
= elem
->next
;
558 if (disk_path
!= devname
)
564 if (disk_path
!= devname
)
572 static int find_intel_hba_capability(int fd
, struct intel_super
*super
,
575 static struct supertype
*match_metadata_desc_imsm(char *arg
)
577 struct supertype
*st
;
579 if (strcmp(arg
, "imsm") != 0 &&
580 strcmp(arg
, "default") != 0
584 st
= malloc(sizeof(*st
));
587 memset(st
, 0, sizeof(*st
));
588 st
->container_dev
= NoMdDev
;
589 st
->ss
= &super_imsm
;
590 st
->max_devs
= IMSM_MAX_DEVICES
;
591 st
->minor_version
= 0;
597 static __u8
*get_imsm_version(struct imsm_super
*mpb
)
599 return &mpb
->sig
[MPB_SIG_LEN
];
603 /* retrieve a disk directly from the anchor when the anchor is known to be
604 * up-to-date, currently only at load time
606 static struct imsm_disk
*__get_imsm_disk(struct imsm_super
*mpb
, __u8 index
)
608 if (index
>= mpb
->num_disks
)
610 return &mpb
->disk
[index
];
613 /* retrieve the disk description based on a index of the disk
616 static struct dl
*get_imsm_dl_disk(struct intel_super
*super
, __u8 index
)
620 for (d
= super
->disks
; d
; d
= d
->next
)
621 if (d
->index
== index
)
626 /* retrieve a disk from the parsed metadata */
627 static struct imsm_disk
*get_imsm_disk(struct intel_super
*super
, __u8 index
)
631 dl
= get_imsm_dl_disk(super
, index
);
638 /* generate a checksum directly from the anchor when the anchor is known to be
639 * up-to-date, currently only at load or write_super after coalescing
641 static __u32
__gen_imsm_checksum(struct imsm_super
*mpb
)
643 __u32 end
= mpb
->mpb_size
/ sizeof(end
);
644 __u32
*p
= (__u32
*) mpb
;
648 sum
+= __le32_to_cpu(*p
);
652 return sum
- __le32_to_cpu(mpb
->check_sum
);
655 static size_t sizeof_imsm_map(struct imsm_map
*map
)
657 return sizeof(struct imsm_map
) + sizeof(__u32
) * (map
->num_members
- 1);
660 struct imsm_map
*get_imsm_map(struct imsm_dev
*dev
, int second_map
)
662 /* A device can have 2 maps if it is in the middle of a migration.
664 * 0 - we return the first map
665 * 1 - we return the second map if it exists, else NULL
666 * -1 - we return the second map if it exists, else the first
668 struct imsm_map
*map
= &dev
->vol
.map
[0];
670 if (second_map
== 1 && !dev
->vol
.migr_state
)
672 else if (second_map
== 1 ||
673 (second_map
< 0 && dev
->vol
.migr_state
)) {
676 return ptr
+ sizeof_imsm_map(map
);
682 /* return the size of the device.
683 * migr_state increases the returned size if map[0] were to be duplicated
685 static size_t sizeof_imsm_dev(struct imsm_dev
*dev
, int migr_state
)
687 size_t size
= sizeof(*dev
) - sizeof(struct imsm_map
) +
688 sizeof_imsm_map(get_imsm_map(dev
, 0));
690 /* migrating means an additional map */
691 if (dev
->vol
.migr_state
)
692 size
+= sizeof_imsm_map(get_imsm_map(dev
, 1));
694 size
+= sizeof_imsm_map(get_imsm_map(dev
, 0));
700 /* retrieve disk serial number list from a metadata update */
701 static struct disk_info
*get_disk_info(struct imsm_update_create_array
*update
)
704 struct disk_info
*inf
;
706 inf
= u
+ sizeof(*update
) - sizeof(struct imsm_dev
) +
707 sizeof_imsm_dev(&update
->dev
, 0);
713 static struct imsm_dev
*__get_imsm_dev(struct imsm_super
*mpb
, __u8 index
)
719 if (index
>= mpb
->num_raid_devs
)
722 /* devices start after all disks */
723 offset
= ((void *) &mpb
->disk
[mpb
->num_disks
]) - _mpb
;
725 for (i
= 0; i
<= index
; i
++)
727 return _mpb
+ offset
;
729 offset
+= sizeof_imsm_dev(_mpb
+ offset
, 0);
734 static struct imsm_dev
*get_imsm_dev(struct intel_super
*super
, __u8 index
)
736 struct intel_dev
*dv
;
738 if (index
>= super
->anchor
->num_raid_devs
)
740 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
741 if (dv
->index
== index
)
749 * == 1 get second map
750 * == -1 than get map according to the current migr_state
752 static __u32
get_imsm_ord_tbl_ent(struct imsm_dev
*dev
,
756 struct imsm_map
*map
;
758 map
= get_imsm_map(dev
, second_map
);
760 /* top byte identifies disk under rebuild */
761 return __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
764 #define ord_to_idx(ord) (((ord) << 8) >> 8)
765 static __u32
get_imsm_disk_idx(struct imsm_dev
*dev
, int slot
, int second_map
)
767 __u32 ord
= get_imsm_ord_tbl_ent(dev
, slot
, second_map
);
769 return ord_to_idx(ord
);
772 static void set_imsm_ord_tbl_ent(struct imsm_map
*map
, int slot
, __u32 ord
)
774 map
->disk_ord_tbl
[slot
] = __cpu_to_le32(ord
);
777 static int get_imsm_disk_slot(struct imsm_map
*map
, unsigned idx
)
782 for (slot
= 0; slot
< map
->num_members
; slot
++) {
783 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
784 if (ord_to_idx(ord
) == idx
)
791 static int get_imsm_raid_level(struct imsm_map
*map
)
793 if (map
->raid_level
== 1) {
794 if (map
->num_members
== 2)
800 return map
->raid_level
;
803 static int cmp_extent(const void *av
, const void *bv
)
805 const struct extent
*a
= av
;
806 const struct extent
*b
= bv
;
807 if (a
->start
< b
->start
)
809 if (a
->start
> b
->start
)
814 static int count_memberships(struct dl
*dl
, struct intel_super
*super
)
819 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
820 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
821 struct imsm_map
*map
= get_imsm_map(dev
, 0);
823 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
830 static struct extent
*get_extents(struct intel_super
*super
, struct dl
*dl
)
832 /* find a list of used extents on the given physical device */
833 struct extent
*rv
, *e
;
835 int memberships
= count_memberships(dl
, super
);
836 __u32 reservation
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
838 rv
= malloc(sizeof(struct extent
) * (memberships
+ 1));
843 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
844 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
845 struct imsm_map
*map
= get_imsm_map(dev
, 0);
847 if (get_imsm_disk_slot(map
, dl
->index
) >= 0) {
848 e
->start
= __le32_to_cpu(map
->pba_of_lba0
);
849 e
->size
= __le32_to_cpu(map
->blocks_per_member
);
853 qsort(rv
, memberships
, sizeof(*rv
), cmp_extent
);
855 /* determine the start of the metadata
856 * when no raid devices are defined use the default
857 * ...otherwise allow the metadata to truncate the value
858 * as is the case with older versions of imsm
861 struct extent
*last
= &rv
[memberships
- 1];
864 remainder
= __le32_to_cpu(dl
->disk
.total_blocks
) -
865 (last
->start
+ last
->size
);
866 /* round down to 1k block to satisfy precision of the kernel
870 /* make sure remainder is still sane */
871 if (remainder
< (unsigned)ROUND_UP(super
->len
, 512) >> 9)
872 remainder
= ROUND_UP(super
->len
, 512) >> 9;
873 if (reservation
> remainder
)
874 reservation
= remainder
;
876 e
->start
= __le32_to_cpu(dl
->disk
.total_blocks
) - reservation
;
881 /* try to determine how much space is reserved for metadata from
882 * the last get_extents() entry, otherwise fallback to the
885 static __u32
imsm_reserved_sectors(struct intel_super
*super
, struct dl
*dl
)
891 /* for spares just return a minimal reservation which will grow
892 * once the spare is picked up by an array
895 return MPB_SECTOR_CNT
;
897 e
= get_extents(super
, dl
);
899 return MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
901 /* scroll to last entry */
902 for (i
= 0; e
[i
].size
; i
++)
905 rv
= __le32_to_cpu(dl
->disk
.total_blocks
) - e
[i
].start
;
912 static int is_spare(struct imsm_disk
*disk
)
914 return (disk
->status
& SPARE_DISK
) == SPARE_DISK
;
917 static int is_configured(struct imsm_disk
*disk
)
919 return (disk
->status
& CONFIGURED_DISK
) == CONFIGURED_DISK
;
922 static int is_failed(struct imsm_disk
*disk
)
924 return (disk
->status
& FAILED_DISK
) == FAILED_DISK
;
927 /* Return minimum size of a spare that can be used in this array*/
928 static unsigned long long min_acceptable_spare_size_imsm(struct supertype
*st
)
930 struct intel_super
*super
= st
->sb
;
934 unsigned long long rv
= 0;
938 /* find first active disk in array */
940 while (dl
&& (is_failed(&dl
->disk
) || dl
->index
== -1))
944 /* find last lba used by subarrays */
945 e
= get_extents(super
, dl
);
948 for (i
= 0; e
[i
].size
; i
++)
951 rv
= e
[i
-1].start
+ e
[i
-1].size
;
953 /* add the amount of space needed for metadata */
954 rv
= rv
+ MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
959 static __u64
blocks_per_migr_unit(struct intel_super
*super
,
960 struct imsm_dev
*dev
);
962 static void print_imsm_dev(struct intel_super
*super
,
963 struct imsm_dev
*dev
,
969 struct imsm_map
*map
= get_imsm_map(dev
, 0);
970 struct imsm_map
*map2
= get_imsm_map(dev
, 1);
974 printf("[%.16s]:\n", dev
->volume
);
975 printf(" UUID : %s\n", uuid
);
976 printf(" RAID Level : %d", get_imsm_raid_level(map
));
978 printf(" <-- %d", get_imsm_raid_level(map2
));
980 printf(" Members : %d", map
->num_members
);
982 printf(" <-- %d", map2
->num_members
);
984 printf(" Slots : [");
985 for (i
= 0; i
< map
->num_members
; i
++) {
986 ord
= get_imsm_ord_tbl_ent(dev
, i
, 0);
987 printf("%s", ord
& IMSM_ORD_REBUILD
? "_" : "U");
992 for (i
= 0; i
< map2
->num_members
; i
++) {
993 ord
= get_imsm_ord_tbl_ent(dev
, i
, 1);
994 printf("%s", ord
& IMSM_ORD_REBUILD
? "_" : "U");
999 printf(" Failed disk : ");
1000 if (map
->failed_disk_num
== 0xff)
1003 printf("%i", map
->failed_disk_num
);
1005 slot
= get_imsm_disk_slot(map
, disk_idx
);
1007 ord
= get_imsm_ord_tbl_ent(dev
, slot
, -1);
1008 printf(" This Slot : %d%s\n", slot
,
1009 ord
& IMSM_ORD_REBUILD
? " (out-of-sync)" : "");
1011 printf(" This Slot : ?\n");
1012 sz
= __le32_to_cpu(dev
->size_high
);
1014 sz
+= __le32_to_cpu(dev
->size_low
);
1015 printf(" Array Size : %llu%s\n", (unsigned long long)sz
,
1016 human_size(sz
* 512));
1017 sz
= __le32_to_cpu(map
->blocks_per_member
);
1018 printf(" Per Dev Size : %llu%s\n", (unsigned long long)sz
,
1019 human_size(sz
* 512));
1020 printf(" Sector Offset : %u\n",
1021 __le32_to_cpu(map
->pba_of_lba0
));
1022 printf(" Num Stripes : %u\n",
1023 __le32_to_cpu(map
->num_data_stripes
));
1024 printf(" Chunk Size : %u KiB",
1025 __le16_to_cpu(map
->blocks_per_strip
) / 2);
1027 printf(" <-- %u KiB",
1028 __le16_to_cpu(map2
->blocks_per_strip
) / 2);
1030 printf(" Reserved : %d\n", __le32_to_cpu(dev
->reserved_blocks
));
1031 printf(" Migrate State : ");
1032 if (dev
->vol
.migr_state
) {
1033 if (migr_type(dev
) == MIGR_INIT
)
1034 printf("initialize\n");
1035 else if (migr_type(dev
) == MIGR_REBUILD
)
1036 printf("rebuild\n");
1037 else if (migr_type(dev
) == MIGR_VERIFY
)
1039 else if (migr_type(dev
) == MIGR_GEN_MIGR
)
1040 printf("general migration\n");
1041 else if (migr_type(dev
) == MIGR_STATE_CHANGE
)
1042 printf("state change\n");
1043 else if (migr_type(dev
) == MIGR_REPAIR
)
1046 printf("<unknown:%d>\n", migr_type(dev
));
1049 printf(" Map State : %s", map_state_str
[map
->map_state
]);
1050 if (dev
->vol
.migr_state
) {
1051 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1053 printf(" <-- %s", map_state_str
[map
->map_state
]);
1054 printf("\n Checkpoint : %u (%llu)",
1055 __le32_to_cpu(dev
->vol
.curr_migr_unit
),
1056 (unsigned long long)blocks_per_migr_unit(super
, dev
));
1059 printf(" Dirty State : %s\n", dev
->vol
.dirty
? "dirty" : "clean");
1062 static void print_imsm_disk(struct imsm_super
*mpb
, int index
, __u32 reserved
)
1064 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, index
);
1065 char str
[MAX_RAID_SERIAL_LEN
+ 1];
1068 if (index
< 0 || !disk
)
1072 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
1073 printf(" Disk%02d Serial : %s\n", index
, str
);
1074 printf(" State :%s%s%s\n", is_spare(disk
) ? " spare" : "",
1075 is_configured(disk
) ? " active" : "",
1076 is_failed(disk
) ? " failed" : "");
1077 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
1078 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
1079 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
1080 human_size(sz
* 512));
1083 static int is_gen_migration(struct imsm_dev
*dev
);
1085 void examine_migr_rec_imsm(struct intel_super
*super
)
1087 struct migr_record
*migr_rec
= super
->migr_rec
;
1088 struct imsm_super
*mpb
= super
->anchor
;
1091 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1092 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1093 if (is_gen_migration(dev
) == 0)
1096 printf("\nMigration Record Information:");
1097 if (super
->disks
->index
> 1) {
1098 printf(" Empty\n ");
1099 printf("Examine one of first two disks in array\n");
1102 printf("\n Status : ");
1103 if (__le32_to_cpu(migr_rec
->rec_status
) == UNIT_SRC_NORMAL
)
1106 printf("Contains Data\n");
1107 printf(" Current Unit : %u\n",
1108 __le32_to_cpu(migr_rec
->curr_migr_unit
));
1109 printf(" Family : %u\n",
1110 __le32_to_cpu(migr_rec
->family_num
));
1111 printf(" Ascending : %u\n",
1112 __le32_to_cpu(migr_rec
->ascending_migr
));
1113 printf(" Blocks Per Unit : %u\n",
1114 __le32_to_cpu(migr_rec
->blocks_per_unit
));
1115 printf(" Dest. Depth Per Unit : %u\n",
1116 __le32_to_cpu(migr_rec
->dest_depth_per_unit
));
1117 printf(" Checkpoint Area pba : %u\n",
1118 __le32_to_cpu(migr_rec
->ckpt_area_pba
));
1119 printf(" First member lba : %u\n",
1120 __le32_to_cpu(migr_rec
->dest_1st_member_lba
));
1121 printf(" Total Number of Units : %u\n",
1122 __le32_to_cpu(migr_rec
->num_migr_units
));
1123 printf(" Size of volume : %u\n",
1124 __le32_to_cpu(migr_rec
->post_migr_vol_cap
));
1125 printf(" Expansion space for LBA64 : %u\n",
1126 __le32_to_cpu(migr_rec
->post_migr_vol_cap_hi
));
1127 printf(" Record was read from : %u\n",
1128 __le32_to_cpu(migr_rec
->ckpt_read_disk_num
));
1133 #endif /* MDASSEMBLE */
1134 /*******************************************************************************
1135 * function: imsm_check_attributes
1136 * Description: Function checks if features represented by attributes flags
1137 * are supported by mdadm.
1139 * attributes - Attributes read from metadata
1141 * 0 - passed attributes contains unsupported features flags
1142 * 1 - all features are supported
1143 ******************************************************************************/
1144 static int imsm_check_attributes(__u32 attributes
)
1147 __u32 not_supported
= MPB_ATTRIB_SUPPORTED
^0xffffffff;
1149 not_supported
&= ~MPB_ATTRIB_IGNORED
;
1151 not_supported
&= attributes
;
1152 if (not_supported
) {
1153 fprintf(stderr
, Name
"(IMSM): Unsupported attributes : %x\n",
1154 (unsigned)__le32_to_cpu(not_supported
));
1155 if (not_supported
& MPB_ATTRIB_CHECKSUM_VERIFY
) {
1156 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY \n");
1157 not_supported
^= MPB_ATTRIB_CHECKSUM_VERIFY
;
1159 if (not_supported
& MPB_ATTRIB_2TB
) {
1160 dprintf("\t\tMPB_ATTRIB_2TB\n");
1161 not_supported
^= MPB_ATTRIB_2TB
;
1163 if (not_supported
& MPB_ATTRIB_RAID0
) {
1164 dprintf("\t\tMPB_ATTRIB_RAID0\n");
1165 not_supported
^= MPB_ATTRIB_RAID0
;
1167 if (not_supported
& MPB_ATTRIB_RAID1
) {
1168 dprintf("\t\tMPB_ATTRIB_RAID1\n");
1169 not_supported
^= MPB_ATTRIB_RAID1
;
1171 if (not_supported
& MPB_ATTRIB_RAID10
) {
1172 dprintf("\t\tMPB_ATTRIB_RAID10\n");
1173 not_supported
^= MPB_ATTRIB_RAID10
;
1175 if (not_supported
& MPB_ATTRIB_RAID1E
) {
1176 dprintf("\t\tMPB_ATTRIB_RAID1E\n");
1177 not_supported
^= MPB_ATTRIB_RAID1E
;
1179 if (not_supported
& MPB_ATTRIB_RAID5
) {
1180 dprintf("\t\tMPB_ATTRIB_RAID5\n");
1181 not_supported
^= MPB_ATTRIB_RAID5
;
1183 if (not_supported
& MPB_ATTRIB_RAIDCNG
) {
1184 dprintf("\t\tMPB_ATTRIB_RAIDCNG\n");
1185 not_supported
^= MPB_ATTRIB_RAIDCNG
;
1187 if (not_supported
& MPB_ATTRIB_BBM
) {
1188 dprintf("\t\tMPB_ATTRIB_BBM\n");
1189 not_supported
^= MPB_ATTRIB_BBM
;
1191 if (not_supported
& MPB_ATTRIB_CHECKSUM_VERIFY
) {
1192 dprintf("\t\tMPB_ATTRIB_CHECKSUM_VERIFY (== MPB_ATTRIB_LEGACY)\n");
1193 not_supported
^= MPB_ATTRIB_CHECKSUM_VERIFY
;
1195 if (not_supported
& MPB_ATTRIB_EXP_STRIPE_SIZE
) {
1196 dprintf("\t\tMPB_ATTRIB_EXP_STRIP_SIZE\n");
1197 not_supported
^= MPB_ATTRIB_EXP_STRIPE_SIZE
;
1199 if (not_supported
& MPB_ATTRIB_2TB_DISK
) {
1200 dprintf("\t\tMPB_ATTRIB_2TB_DISK\n");
1201 not_supported
^= MPB_ATTRIB_2TB_DISK
;
1203 if (not_supported
& MPB_ATTRIB_NEVER_USE2
) {
1204 dprintf("\t\tMPB_ATTRIB_NEVER_USE2\n");
1205 not_supported
^= MPB_ATTRIB_NEVER_USE2
;
1207 if (not_supported
& MPB_ATTRIB_NEVER_USE
) {
1208 dprintf("\t\tMPB_ATTRIB_NEVER_USE\n");
1209 not_supported
^= MPB_ATTRIB_NEVER_USE
;
1213 dprintf(Name
"(IMSM): Unknown attributes : %x\n", not_supported
);
1222 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
, char *map
);
1224 static void examine_super_imsm(struct supertype
*st
, char *homehost
)
1226 struct intel_super
*super
= st
->sb
;
1227 struct imsm_super
*mpb
= super
->anchor
;
1228 char str
[MAX_SIGNATURE_LENGTH
];
1233 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
1236 snprintf(str
, MPB_SIG_LEN
, "%s", mpb
->sig
);
1237 printf(" Magic : %s\n", str
);
1238 snprintf(str
, strlen(MPB_VERSION_RAID0
), "%s", get_imsm_version(mpb
));
1239 printf(" Version : %s\n", get_imsm_version(mpb
));
1240 printf(" Orig Family : %08x\n", __le32_to_cpu(mpb
->orig_family_num
));
1241 printf(" Family : %08x\n", __le32_to_cpu(mpb
->family_num
));
1242 printf(" Generation : %08x\n", __le32_to_cpu(mpb
->generation_num
));
1243 printf(" Attributes : ");
1244 if (imsm_check_attributes(mpb
->attributes
))
1245 printf("All supported\n");
1247 printf("not supported\n");
1248 getinfo_super_imsm(st
, &info
, NULL
);
1249 fname_from_uuid(st
, &info
, nbuf
, ':');
1250 printf(" UUID : %s\n", nbuf
+ 5);
1251 sum
= __le32_to_cpu(mpb
->check_sum
);
1252 printf(" Checksum : %08x %s\n", sum
,
1253 __gen_imsm_checksum(mpb
) == sum
? "correct" : "incorrect");
1254 printf(" MPB Sectors : %d\n", mpb_sectors(mpb
));
1255 printf(" Disks : %d\n", mpb
->num_disks
);
1256 printf(" RAID Devices : %d\n", mpb
->num_raid_devs
);
1257 print_imsm_disk(mpb
, super
->disks
->index
, reserved
);
1258 if (super
->bbm_log
) {
1259 struct bbm_log
*log
= super
->bbm_log
;
1262 printf("Bad Block Management Log:\n");
1263 printf(" Log Size : %d\n", __le32_to_cpu(mpb
->bbm_log_size
));
1264 printf(" Signature : %x\n", __le32_to_cpu(log
->signature
));
1265 printf(" Entry Count : %d\n", __le32_to_cpu(log
->entry_count
));
1266 printf(" Spare Blocks : %d\n", __le32_to_cpu(log
->reserved_spare_block_count
));
1267 printf(" First Spare : %llx\n",
1268 (unsigned long long) __le64_to_cpu(log
->first_spare_lba
));
1270 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
1272 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
1274 super
->current_vol
= i
;
1275 getinfo_super_imsm(st
, &info
, NULL
);
1276 fname_from_uuid(st
, &info
, nbuf
, ':');
1277 print_imsm_dev(super
, dev
, nbuf
+ 5, super
->disks
->index
);
1279 for (i
= 0; i
< mpb
->num_disks
; i
++) {
1280 if (i
== super
->disks
->index
)
1282 print_imsm_disk(mpb
, i
, reserved
);
1284 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
1285 struct imsm_disk
*disk
;
1286 char str
[MAX_RAID_SERIAL_LEN
+ 1];
1294 snprintf(str
, MAX_RAID_SERIAL_LEN
+ 1, "%s", disk
->serial
);
1295 printf(" Disk Serial : %s\n", str
);
1296 printf(" State :%s%s%s\n", is_spare(disk
) ? " spare" : "",
1297 is_configured(disk
) ? " active" : "",
1298 is_failed(disk
) ? " failed" : "");
1299 printf(" Id : %08x\n", __le32_to_cpu(disk
->scsi_id
));
1300 sz
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
1301 printf(" Usable Size : %llu%s\n", (unsigned long long)sz
,
1302 human_size(sz
* 512));
1305 examine_migr_rec_imsm(super
);
1308 static void brief_examine_super_imsm(struct supertype
*st
, int verbose
)
1310 /* We just write a generic IMSM ARRAY entry */
1313 struct intel_super
*super
= st
->sb
;
1315 if (!super
->anchor
->num_raid_devs
) {
1316 printf("ARRAY metadata=imsm\n");
1320 getinfo_super_imsm(st
, &info
, NULL
);
1321 fname_from_uuid(st
, &info
, nbuf
, ':');
1322 printf("ARRAY metadata=imsm UUID=%s\n", nbuf
+ 5);
1325 static void brief_examine_subarrays_imsm(struct supertype
*st
, int verbose
)
1327 /* We just write a generic IMSM ARRAY entry */
1331 struct intel_super
*super
= st
->sb
;
1334 if (!super
->anchor
->num_raid_devs
)
1337 getinfo_super_imsm(st
, &info
, NULL
);
1338 fname_from_uuid(st
, &info
, nbuf
, ':');
1339 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
1340 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
1342 super
->current_vol
= i
;
1343 getinfo_super_imsm(st
, &info
, NULL
);
1344 fname_from_uuid(st
, &info
, nbuf1
, ':');
1345 printf("ARRAY /dev/md/%.16s container=%s member=%d UUID=%s\n",
1346 dev
->volume
, nbuf
+ 5, i
, nbuf1
+ 5);
1350 static void export_examine_super_imsm(struct supertype
*st
)
1352 struct intel_super
*super
= st
->sb
;
1353 struct imsm_super
*mpb
= super
->anchor
;
1357 getinfo_super_imsm(st
, &info
, NULL
);
1358 fname_from_uuid(st
, &info
, nbuf
, ':');
1359 printf("MD_METADATA=imsm\n");
1360 printf("MD_LEVEL=container\n");
1361 printf("MD_UUID=%s\n", nbuf
+5);
1362 printf("MD_DEVICES=%u\n", mpb
->num_disks
);
1365 static void detail_super_imsm(struct supertype
*st
, char *homehost
)
1370 getinfo_super_imsm(st
, &info
, NULL
);
1371 fname_from_uuid(st
, &info
, nbuf
, ':');
1372 printf("\n UUID : %s\n", nbuf
+ 5);
1375 static void brief_detail_super_imsm(struct supertype
*st
)
1379 getinfo_super_imsm(st
, &info
, NULL
);
1380 fname_from_uuid(st
, &info
, nbuf
, ':');
1381 printf(" UUID=%s", nbuf
+ 5);
1384 static int imsm_read_serial(int fd
, char *devname
, __u8
*serial
);
1385 static void fd2devname(int fd
, char *name
);
1387 static int ahci_enumerate_ports(const char *hba_path
, int port_count
, int host_base
, int verbose
)
1389 /* dump an unsorted list of devices attached to AHCI Intel storage
1390 * controller, as well as non-connected ports
1392 int hba_len
= strlen(hba_path
) + 1;
1397 unsigned long port_mask
= (1 << port_count
) - 1;
1399 if (port_count
> (int)sizeof(port_mask
) * 8) {
1401 fprintf(stderr
, Name
": port_count %d out of range\n", port_count
);
1405 /* scroll through /sys/dev/block looking for devices attached to
1408 dir
= opendir("/sys/dev/block");
1409 for (ent
= dir
? readdir(dir
) : NULL
; ent
; ent
= readdir(dir
)) {
1420 if (sscanf(ent
->d_name
, "%d:%d", &major
, &minor
) != 2)
1422 path
= devt_to_devpath(makedev(major
, minor
));
1425 if (!path_attached_to_hba(path
, hba_path
)) {
1431 /* retrieve the scsi device type */
1432 if (asprintf(&device
, "/sys/dev/block/%d:%d/device/xxxxxxx", major
, minor
) < 0) {
1434 fprintf(stderr
, Name
": failed to allocate 'device'\n");
1438 sprintf(device
, "/sys/dev/block/%d:%d/device/type", major
, minor
);
1439 if (load_sys(device
, buf
) != 0) {
1441 fprintf(stderr
, Name
": failed to read device type for %s\n",
1447 type
= strtoul(buf
, NULL
, 10);
1449 /* if it's not a disk print the vendor and model */
1450 if (!(type
== 0 || type
== 7 || type
== 14)) {
1453 sprintf(device
, "/sys/dev/block/%d:%d/device/vendor", major
, minor
);
1454 if (load_sys(device
, buf
) == 0) {
1455 strncpy(vendor
, buf
, sizeof(vendor
));
1456 vendor
[sizeof(vendor
) - 1] = '\0';
1457 c
= (char *) &vendor
[sizeof(vendor
) - 1];
1458 while (isspace(*c
) || *c
== '\0')
1462 sprintf(device
, "/sys/dev/block/%d:%d/device/model", major
, minor
);
1463 if (load_sys(device
, buf
) == 0) {
1464 strncpy(model
, buf
, sizeof(model
));
1465 model
[sizeof(model
) - 1] = '\0';
1466 c
= (char *) &model
[sizeof(model
) - 1];
1467 while (isspace(*c
) || *c
== '\0')
1471 if (vendor
[0] && model
[0])
1472 sprintf(buf
, "%.64s %.64s", vendor
, model
);
1474 switch (type
) { /* numbers from hald/linux/device.c */
1475 case 1: sprintf(buf
, "tape"); break;
1476 case 2: sprintf(buf
, "printer"); break;
1477 case 3: sprintf(buf
, "processor"); break;
1479 case 5: sprintf(buf
, "cdrom"); break;
1480 case 6: sprintf(buf
, "scanner"); break;
1481 case 8: sprintf(buf
, "media_changer"); break;
1482 case 9: sprintf(buf
, "comm"); break;
1483 case 12: sprintf(buf
, "raid"); break;
1484 default: sprintf(buf
, "unknown");
1490 /* chop device path to 'host%d' and calculate the port number */
1491 c
= strchr(&path
[hba_len
], '/');
1494 fprintf(stderr
, Name
": %s - invalid path name\n", path
+ hba_len
);
1499 if (sscanf(&path
[hba_len
], "host%d", &port
) == 1)
1503 *c
= '/'; /* repair the full string */
1504 fprintf(stderr
, Name
": failed to determine port number for %s\n",
1511 /* mark this port as used */
1512 port_mask
&= ~(1 << port
);
1514 /* print out the device information */
1516 printf(" Port%d : - non-disk device (%s) -\n", port
, buf
);
1520 fd
= dev_open(ent
->d_name
, O_RDONLY
);
1522 printf(" Port%d : - disk info unavailable -\n", port
);
1524 fd2devname(fd
, buf
);
1525 printf(" Port%d : %s", port
, buf
);
1526 if (imsm_read_serial(fd
, NULL
, (__u8
*) buf
) == 0)
1527 printf(" (%.*s)\n", MAX_RAID_SERIAL_LEN
, buf
);
1542 for (i
= 0; i
< port_count
; i
++)
1543 if (port_mask
& (1 << i
))
1544 printf(" Port%d : - no device attached -\n", i
);
1550 static void print_found_intel_controllers(struct sys_dev
*elem
)
1552 for (; elem
; elem
= elem
->next
) {
1553 fprintf(stderr
, Name
": found Intel(R) ");
1554 if (elem
->type
== SYS_DEV_SATA
)
1555 fprintf(stderr
, "SATA ");
1556 else if (elem
->type
== SYS_DEV_SAS
)
1557 fprintf(stderr
, "SAS ");
1558 fprintf(stderr
, "RAID controller");
1560 fprintf(stderr
, " at %s", elem
->pci_id
);
1561 fprintf(stderr
, ".\n");
1566 static int ahci_get_port_count(const char *hba_path
, int *port_count
)
1573 if ((dir
= opendir(hba_path
)) == NULL
)
1576 for (ent
= readdir(dir
); ent
; ent
= readdir(dir
)) {
1579 if (sscanf(ent
->d_name
, "host%d", &host
) != 1)
1581 if (*port_count
== 0)
1583 else if (host
< host_base
)
1586 if (host
+ 1 > *port_count
+ host_base
)
1587 *port_count
= host
+ 1 - host_base
;
1593 static void print_imsm_capability(const struct imsm_orom
*orom
)
1595 printf(" Platform : Intel(R) Matrix Storage Manager\n");
1596 printf(" Version : %d.%d.%d.%d\n", orom
->major_ver
, orom
->minor_ver
,
1597 orom
->hotfix_ver
, orom
->build
);
1598 printf(" RAID Levels :%s%s%s%s%s\n",
1599 imsm_orom_has_raid0(orom
) ? " raid0" : "",
1600 imsm_orom_has_raid1(orom
) ? " raid1" : "",
1601 imsm_orom_has_raid1e(orom
) ? " raid1e" : "",
1602 imsm_orom_has_raid10(orom
) ? " raid10" : "",
1603 imsm_orom_has_raid5(orom
) ? " raid5" : "");
1604 printf(" Chunk Sizes :%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s%s\n",
1605 imsm_orom_has_chunk(orom
, 2) ? " 2k" : "",
1606 imsm_orom_has_chunk(orom
, 4) ? " 4k" : "",
1607 imsm_orom_has_chunk(orom
, 8) ? " 8k" : "",
1608 imsm_orom_has_chunk(orom
, 16) ? " 16k" : "",
1609 imsm_orom_has_chunk(orom
, 32) ? " 32k" : "",
1610 imsm_orom_has_chunk(orom
, 64) ? " 64k" : "",
1611 imsm_orom_has_chunk(orom
, 128) ? " 128k" : "",
1612 imsm_orom_has_chunk(orom
, 256) ? " 256k" : "",
1613 imsm_orom_has_chunk(orom
, 512) ? " 512k" : "",
1614 imsm_orom_has_chunk(orom
, 1024*1) ? " 1M" : "",
1615 imsm_orom_has_chunk(orom
, 1024*2) ? " 2M" : "",
1616 imsm_orom_has_chunk(orom
, 1024*4) ? " 4M" : "",
1617 imsm_orom_has_chunk(orom
, 1024*8) ? " 8M" : "",
1618 imsm_orom_has_chunk(orom
, 1024*16) ? " 16M" : "",
1619 imsm_orom_has_chunk(orom
, 1024*32) ? " 32M" : "",
1620 imsm_orom_has_chunk(orom
, 1024*64) ? " 64M" : "");
1621 printf(" Max Disks : %d\n", orom
->tds
);
1622 printf(" Max Volumes : %d\n", orom
->vpa
);
1626 static int detail_platform_imsm(int verbose
, int enumerate_only
)
1628 /* There are two components to imsm platform support, the ahci SATA
1629 * controller and the option-rom. To find the SATA controller we
1630 * simply look in /sys/bus/pci/drivers/ahci to see if an ahci
1631 * controller with the Intel vendor id is present. This approach
1632 * allows mdadm to leverage the kernel's ahci detection logic, with the
1633 * caveat that if ahci.ko is not loaded mdadm will not be able to
1634 * detect platform raid capabilities. The option-rom resides in a
1635 * platform "Adapter ROM". We scan for its signature to retrieve the
1636 * platform capabilities. If raid support is disabled in the BIOS the
1637 * option-rom capability structure will not be available.
1639 const struct imsm_orom
*orom
;
1640 struct sys_dev
*list
, *hba
;
1645 if (enumerate_only
) {
1646 if (check_env("IMSM_NO_PLATFORM"))
1648 list
= find_intel_devices();
1651 for (hba
= list
; hba
; hba
= hba
->next
) {
1652 orom
= find_imsm_capability(hba
->type
);
1658 free_sys_dev(&list
);
1662 list
= find_intel_devices();
1665 fprintf(stderr
, Name
": no active Intel(R) RAID "
1666 "controller found.\n");
1667 free_sys_dev(&list
);
1670 print_found_intel_controllers(list
);
1672 for (hba
= list
; hba
; hba
= hba
->next
) {
1673 orom
= find_imsm_capability(hba
->type
);
1675 fprintf(stderr
, Name
": imsm capabilities not found for controller: %s (type %s)\n",
1676 hba
->path
, get_sys_dev_type(hba
->type
));
1678 print_imsm_capability(orom
);
1681 for (hba
= list
; hba
; hba
= hba
->next
) {
1682 printf(" I/O Controller : %s (%s)\n",
1683 hba
->path
, get_sys_dev_type(hba
->type
));
1685 if (hba
->type
== SYS_DEV_SATA
) {
1686 host_base
= ahci_get_port_count(hba
->path
, &port_count
);
1687 if (ahci_enumerate_ports(hba
->path
, port_count
, host_base
, verbose
)) {
1689 fprintf(stderr
, Name
": failed to enumerate "
1690 "ports on SATA controller at %s.", hba
->pci_id
);
1696 free_sys_dev(&list
);
1701 static int match_home_imsm(struct supertype
*st
, char *homehost
)
1703 /* the imsm metadata format does not specify any host
1704 * identification information. We return -1 since we can never
1705 * confirm nor deny whether a given array is "meant" for this
1706 * host. We rely on compare_super and the 'family_num' fields to
1707 * exclude member disks that do not belong, and we rely on
1708 * mdadm.conf to specify the arrays that should be assembled.
1709 * Auto-assembly may still pick up "foreign" arrays.
1715 static void uuid_from_super_imsm(struct supertype
*st
, int uuid
[4])
1717 /* The uuid returned here is used for:
1718 * uuid to put into bitmap file (Create, Grow)
1719 * uuid for backup header when saving critical section (Grow)
1720 * comparing uuids when re-adding a device into an array
1721 * In these cases the uuid required is that of the data-array,
1722 * not the device-set.
1723 * uuid to recognise same set when adding a missing device back
1724 * to an array. This is a uuid for the device-set.
1726 * For each of these we can make do with a truncated
1727 * or hashed uuid rather than the original, as long as
1729 * In each case the uuid required is that of the data-array,
1730 * not the device-set.
1732 /* imsm does not track uuid's so we synthesis one using sha1 on
1733 * - The signature (Which is constant for all imsm array, but no matter)
1734 * - the orig_family_num of the container
1735 * - the index number of the volume
1736 * - the 'serial' number of the volume.
1737 * Hopefully these are all constant.
1739 struct intel_super
*super
= st
->sb
;
1742 struct sha1_ctx ctx
;
1743 struct imsm_dev
*dev
= NULL
;
1746 /* some mdadm versions failed to set ->orig_family_num, in which
1747 * case fall back to ->family_num. orig_family_num will be
1748 * fixed up with the first metadata update.
1750 family_num
= super
->anchor
->orig_family_num
;
1751 if (family_num
== 0)
1752 family_num
= super
->anchor
->family_num
;
1753 sha1_init_ctx(&ctx
);
1754 sha1_process_bytes(super
->anchor
->sig
, MPB_SIG_LEN
, &ctx
);
1755 sha1_process_bytes(&family_num
, sizeof(__u32
), &ctx
);
1756 if (super
->current_vol
>= 0)
1757 dev
= get_imsm_dev(super
, super
->current_vol
);
1759 __u32 vol
= super
->current_vol
;
1760 sha1_process_bytes(&vol
, sizeof(vol
), &ctx
);
1761 sha1_process_bytes(dev
->volume
, MAX_RAID_SERIAL_LEN
, &ctx
);
1763 sha1_finish_ctx(&ctx
, buf
);
1764 memcpy(uuid
, buf
, 4*4);
1769 get_imsm_numerical_version(struct imsm_super
*mpb
, int *m
, int *p
)
1771 __u8
*v
= get_imsm_version(mpb
);
1772 __u8
*end
= mpb
->sig
+ MAX_SIGNATURE_LENGTH
;
1773 char major
[] = { 0, 0, 0 };
1774 char minor
[] = { 0 ,0, 0 };
1775 char patch
[] = { 0, 0, 0 };
1776 char *ver_parse
[] = { major
, minor
, patch
};
1780 while (*v
!= '\0' && v
< end
) {
1781 if (*v
!= '.' && j
< 2)
1782 ver_parse
[i
][j
++] = *v
;
1790 *m
= strtol(minor
, NULL
, 0);
1791 *p
= strtol(patch
, NULL
, 0);
1795 static __u32
migr_strip_blocks_resync(struct imsm_dev
*dev
)
1797 /* migr_strip_size when repairing or initializing parity */
1798 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1799 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1801 switch (get_imsm_raid_level(map
)) {
1806 return 128*1024 >> 9;
1810 static __u32
migr_strip_blocks_rebuild(struct imsm_dev
*dev
)
1812 /* migr_strip_size when rebuilding a degraded disk, no idea why
1813 * this is different than migr_strip_size_resync(), but it's good
1816 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1817 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1819 switch (get_imsm_raid_level(map
)) {
1822 if (map
->num_members
% map
->num_domains
== 0)
1823 return 128*1024 >> 9;
1827 return max((__u32
) 64*1024 >> 9, chunk
);
1829 return 128*1024 >> 9;
1833 static __u32
num_stripes_per_unit_resync(struct imsm_dev
*dev
)
1835 struct imsm_map
*lo
= get_imsm_map(dev
, 0);
1836 struct imsm_map
*hi
= get_imsm_map(dev
, 1);
1837 __u32 lo_chunk
= __le32_to_cpu(lo
->blocks_per_strip
);
1838 __u32 hi_chunk
= __le32_to_cpu(hi
->blocks_per_strip
);
1840 return max((__u32
) 1, hi_chunk
/ lo_chunk
);
1843 static __u32
num_stripes_per_unit_rebuild(struct imsm_dev
*dev
)
1845 struct imsm_map
*lo
= get_imsm_map(dev
, 0);
1846 int level
= get_imsm_raid_level(lo
);
1848 if (level
== 1 || level
== 10) {
1849 struct imsm_map
*hi
= get_imsm_map(dev
, 1);
1851 return hi
->num_domains
;
1853 return num_stripes_per_unit_resync(dev
);
1856 static __u8
imsm_num_data_members(struct imsm_dev
*dev
, int second_map
)
1858 /* named 'imsm_' because raid0, raid1 and raid10
1859 * counter-intuitively have the same number of data disks
1861 struct imsm_map
*map
= get_imsm_map(dev
, second_map
);
1863 switch (get_imsm_raid_level(map
)) {
1867 return map
->num_members
;
1869 return map
->num_members
- 1;
1871 dprintf("%s: unsupported raid level\n", __func__
);
1876 static __u32
parity_segment_depth(struct imsm_dev
*dev
)
1878 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1879 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1881 switch(get_imsm_raid_level(map
)) {
1884 return chunk
* map
->num_domains
;
1886 return chunk
* map
->num_members
;
1892 static __u32
map_migr_block(struct imsm_dev
*dev
, __u32 block
)
1894 struct imsm_map
*map
= get_imsm_map(dev
, 1);
1895 __u32 chunk
= __le32_to_cpu(map
->blocks_per_strip
);
1896 __u32 strip
= block
/ chunk
;
1898 switch (get_imsm_raid_level(map
)) {
1901 __u32 vol_strip
= (strip
* map
->num_domains
) + 1;
1902 __u32 vol_stripe
= vol_strip
/ map
->num_members
;
1904 return vol_stripe
* chunk
+ block
% chunk
;
1906 __u32 stripe
= strip
/ (map
->num_members
- 1);
1908 return stripe
* chunk
+ block
% chunk
;
1915 static __u64
blocks_per_migr_unit(struct intel_super
*super
,
1916 struct imsm_dev
*dev
)
1918 /* calculate the conversion factor between per member 'blocks'
1919 * (md/{resync,rebuild}_start) and imsm migration units, return
1920 * 0 for the 'not migrating' and 'unsupported migration' cases
1922 if (!dev
->vol
.migr_state
)
1925 switch (migr_type(dev
)) {
1926 case MIGR_GEN_MIGR
: {
1927 struct migr_record
*migr_rec
= super
->migr_rec
;
1928 return __le32_to_cpu(migr_rec
->blocks_per_unit
);
1933 struct imsm_map
*map
= get_imsm_map(dev
, 0);
1934 __u32 stripes_per_unit
;
1935 __u32 blocks_per_unit
;
1944 /* yes, this is really the translation of migr_units to
1945 * per-member blocks in the 'resync' case
1947 stripes_per_unit
= num_stripes_per_unit_resync(dev
);
1948 migr_chunk
= migr_strip_blocks_resync(dev
);
1949 disks
= imsm_num_data_members(dev
, 0);
1950 blocks_per_unit
= stripes_per_unit
* migr_chunk
* disks
;
1951 stripe
= __le16_to_cpu(map
->blocks_per_strip
) * disks
;
1952 segment
= blocks_per_unit
/ stripe
;
1953 block_rel
= blocks_per_unit
- segment
* stripe
;
1954 parity_depth
= parity_segment_depth(dev
);
1955 block_map
= map_migr_block(dev
, block_rel
);
1956 return block_map
+ parity_depth
* segment
;
1958 case MIGR_REBUILD
: {
1959 __u32 stripes_per_unit
;
1962 stripes_per_unit
= num_stripes_per_unit_rebuild(dev
);
1963 migr_chunk
= migr_strip_blocks_rebuild(dev
);
1964 return migr_chunk
* stripes_per_unit
;
1966 case MIGR_STATE_CHANGE
:
1972 static int imsm_level_to_layout(int level
)
1980 return ALGORITHM_LEFT_ASYMMETRIC
;
1987 /*******************************************************************************
1988 * Function: read_imsm_migr_rec
1989 * Description: Function reads imsm migration record from last sector of disk
1991 * fd : disk descriptor
1992 * super : metadata info
1996 ******************************************************************************/
1997 static int read_imsm_migr_rec(int fd
, struct intel_super
*super
)
2000 unsigned long long dsize
;
2002 get_dev_size(fd
, NULL
, &dsize
);
2003 if (lseek64(fd
, dsize
- 512, SEEK_SET
) < 0) {
2005 Name
": Cannot seek to anchor block: %s\n",
2009 if (read(fd
, super
->migr_rec_buf
, 512) != 512) {
2011 Name
": Cannot read migr record block: %s\n",
2021 /*******************************************************************************
2022 * Function: load_imsm_migr_rec
2023 * Description: Function reads imsm migration record (it is stored at the last
2026 * super : imsm internal array info
2027 * info : general array info
2031 ******************************************************************************/
2032 static int load_imsm_migr_rec(struct intel_super
*super
, struct mdinfo
*info
)
2035 struct dl
*dl
= NULL
;
2041 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
2042 /* read only from one of the first two slots */
2043 if ((sd
->disk
.raid_disk
> 1) ||
2044 (sd
->disk
.raid_disk
< 0))
2046 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
2047 fd
= dev_open(nm
, O_RDONLY
);
2053 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
2054 /* read only from one of the first two slots */
2057 sprintf(nm
, "%d:%d", dl
->major
, dl
->minor
);
2058 fd
= dev_open(nm
, O_RDONLY
);
2065 retval
= read_imsm_migr_rec(fd
, super
);
2074 /*******************************************************************************
2075 * function: imsm_create_metadata_checkpoint_update
2076 * Description: It creates update for checkpoint change.
2078 * super : imsm internal array info
2079 * u : pointer to prepared update
2082 * If length is equal to 0, input pointer u contains no update
2083 ******************************************************************************/
2084 static int imsm_create_metadata_checkpoint_update(
2085 struct intel_super
*super
,
2086 struct imsm_update_general_migration_checkpoint
**u
)
2089 int update_memory_size
= 0;
2091 dprintf("imsm_create_metadata_checkpoint_update(enter)\n");
2097 /* size of all update data without anchor */
2098 update_memory_size
=
2099 sizeof(struct imsm_update_general_migration_checkpoint
);
2101 *u
= calloc(1, update_memory_size
);
2103 dprintf("error: cannot get memory for "
2104 "imsm_create_metadata_checkpoint_update update\n");
2107 (*u
)->type
= update_general_migration_checkpoint
;
2108 (*u
)->curr_migr_unit
= __le32_to_cpu(super
->migr_rec
->curr_migr_unit
);
2109 dprintf("imsm_create_metadata_checkpoint_update: prepared for %u\n",
2110 (*u
)->curr_migr_unit
);
2112 return update_memory_size
;
2116 static void imsm_update_metadata_locally(struct supertype
*st
,
2117 void *buf
, int len
);
2119 /*******************************************************************************
2120 * Function: write_imsm_migr_rec
2121 * Description: Function writes imsm migration record
2122 * (at the last sector of disk)
2124 * super : imsm internal array info
2128 ******************************************************************************/
2129 static int write_imsm_migr_rec(struct supertype
*st
)
2131 struct intel_super
*super
= st
->sb
;
2132 unsigned long long dsize
;
2138 struct imsm_update_general_migration_checkpoint
*u
;
2140 for (sd
= super
->disks
; sd
; sd
= sd
->next
) {
2141 /* write to 2 first slots only */
2142 if ((sd
->index
< 0) || (sd
->index
> 1))
2144 sprintf(nm
, "%d:%d", sd
->major
, sd
->minor
);
2145 fd
= dev_open(nm
, O_RDWR
);
2148 get_dev_size(fd
, NULL
, &dsize
);
2149 if (lseek64(fd
, dsize
- 512, SEEK_SET
) < 0) {
2151 Name
": Cannot seek to anchor block: %s\n",
2155 if (write(fd
, super
->migr_rec_buf
, 512) != 512) {
2157 Name
": Cannot write migr record block: %s\n",
2164 /* update checkpoint information in metadata */
2165 len
= imsm_create_metadata_checkpoint_update(super
, &u
);
2168 dprintf("imsm: Cannot prepare update\n");
2171 /* update metadata locally */
2172 imsm_update_metadata_locally(st
, u
, len
);
2173 /* and possibly remotely */
2174 if (st
->update_tail
) {
2175 append_metadata_update(st
, u
, len
);
2176 /* during reshape we do all work inside metadata handler
2177 * manage_reshape(), so metadata update has to be triggered
2180 flush_metadata_updates(st
);
2181 st
->update_tail
= &st
->updates
;
2191 #endif /* MDASSEMBLE */
2193 static void getinfo_super_imsm_volume(struct supertype
*st
, struct mdinfo
*info
, char *dmap
)
2195 struct intel_super
*super
= st
->sb
;
2196 struct migr_record
*migr_rec
= super
->migr_rec
;
2197 struct imsm_dev
*dev
= get_imsm_dev(super
, super
->current_vol
);
2198 struct imsm_map
*map
= get_imsm_map(dev
, 0);
2199 struct imsm_map
*prev_map
= get_imsm_map(dev
, 1);
2200 struct imsm_map
*map_to_analyse
= map
;
2203 unsigned int component_size_alligment
;
2204 int map_disks
= info
->array
.raid_disks
;
2206 memset(info
, 0, sizeof(*info
));
2208 map_to_analyse
= prev_map
;
2210 dl
= super
->current_disk
;
2212 info
->container_member
= super
->current_vol
;
2213 info
->array
.raid_disks
= map
->num_members
;
2214 info
->array
.level
= get_imsm_raid_level(map_to_analyse
);
2215 info
->array
.layout
= imsm_level_to_layout(info
->array
.level
);
2216 info
->array
.md_minor
= -1;
2217 info
->array
.ctime
= 0;
2218 info
->array
.utime
= 0;
2219 info
->array
.chunk_size
=
2220 __le16_to_cpu(map_to_analyse
->blocks_per_strip
) << 9;
2221 info
->array
.state
= !dev
->vol
.dirty
;
2222 info
->custom_array_size
= __le32_to_cpu(dev
->size_high
);
2223 info
->custom_array_size
<<= 32;
2224 info
->custom_array_size
|= __le32_to_cpu(dev
->size_low
);
2225 if (prev_map
&& map
->map_state
== prev_map
->map_state
) {
2226 info
->reshape_active
= 1;
2227 info
->new_level
= get_imsm_raid_level(map
);
2228 info
->new_layout
= imsm_level_to_layout(info
->new_level
);
2229 info
->new_chunk
= __le16_to_cpu(map
->blocks_per_strip
) << 9;
2230 info
->delta_disks
= map
->num_members
- prev_map
->num_members
;
2231 if (info
->delta_disks
) {
2232 /* this needs to be applied to every array
2235 info
->reshape_active
= 2;
2237 /* We shape information that we give to md might have to be
2238 * modify to cope with md's requirement for reshaping arrays.
2239 * For example, when reshaping a RAID0, md requires it to be
2240 * presented as a degraded RAID4.
2241 * Also if a RAID0 is migrating to a RAID5 we need to specify
2242 * the array as already being RAID5, but the 'before' layout
2243 * is a RAID4-like layout.
2245 switch (info
->array
.level
) {
2247 switch(info
->new_level
) {
2249 /* conversion is happening as RAID4 */
2250 info
->array
.level
= 4;
2251 info
->array
.raid_disks
+= 1;
2254 /* conversion is happening as RAID5 */
2255 info
->array
.level
= 5;
2256 info
->array
.layout
= ALGORITHM_PARITY_N
;
2257 info
->delta_disks
-= 1;
2260 /* FIXME error message */
2261 info
->array
.level
= UnSet
;
2267 info
->new_level
= UnSet
;
2268 info
->new_layout
= UnSet
;
2269 info
->new_chunk
= info
->array
.chunk_size
;
2270 info
->delta_disks
= 0;
2274 info
->disk
.major
= dl
->major
;
2275 info
->disk
.minor
= dl
->minor
;
2276 info
->disk
.number
= dl
->index
;
2277 info
->disk
.raid_disk
= get_imsm_disk_slot(map_to_analyse
,
2281 info
->data_offset
= __le32_to_cpu(map_to_analyse
->pba_of_lba0
);
2282 info
->component_size
=
2283 __le32_to_cpu(map_to_analyse
->blocks_per_member
);
2285 /* check component size aligment
2287 component_size_alligment
=
2288 info
->component_size
% (info
->array
.chunk_size
/512);
2290 if (component_size_alligment
&&
2291 (info
->array
.level
!= 1) && (info
->array
.level
!= UnSet
)) {
2292 dprintf("imsm: reported component size alligned from %llu ",
2293 info
->component_size
);
2294 info
->component_size
-= component_size_alligment
;
2295 dprintf("to %llu (%i).\n",
2296 info
->component_size
, component_size_alligment
);
2299 memset(info
->uuid
, 0, sizeof(info
->uuid
));
2300 info
->recovery_start
= MaxSector
;
2302 info
->reshape_progress
= 0;
2303 info
->resync_start
= MaxSector
;
2304 if (map_to_analyse
->map_state
== IMSM_T_STATE_UNINITIALIZED
||
2306 info
->resync_start
= 0;
2308 if (dev
->vol
.migr_state
) {
2309 switch (migr_type(dev
)) {
2312 __u64 blocks_per_unit
= blocks_per_migr_unit(super
,
2314 __u64 units
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
2316 info
->resync_start
= blocks_per_unit
* units
;
2319 case MIGR_GEN_MIGR
: {
2320 __u64 blocks_per_unit
= blocks_per_migr_unit(super
,
2322 __u64 units
= __le32_to_cpu(migr_rec
->curr_migr_unit
);
2323 unsigned long long array_blocks
;
2326 if (__le32_to_cpu(migr_rec
->ascending_migr
) &&
2328 (__le32_to_cpu(migr_rec
->num_migr_units
)-1)) &&
2329 (super
->migr_rec
->rec_status
==
2330 __cpu_to_le32(UNIT_SRC_IN_CP_AREA
)))
2333 info
->reshape_progress
= blocks_per_unit
* units
;
2335 dprintf("IMSM: General Migration checkpoint : %llu "
2336 "(%llu) -> read reshape progress : %llu\n",
2337 (unsigned long long)units
,
2338 (unsigned long long)blocks_per_unit
,
2339 info
->reshape_progress
);
2341 used_disks
= imsm_num_data_members(dev
, 1);
2342 if (used_disks
> 0) {
2343 array_blocks
= map
->blocks_per_member
*
2345 /* round array size down to closest MB
2347 info
->custom_array_size
= (array_blocks
2348 >> SECT_PER_MB_SHIFT
)
2349 << SECT_PER_MB_SHIFT
;
2353 /* we could emulate the checkpointing of
2354 * 'sync_action=check' migrations, but for now
2355 * we just immediately complete them
2358 /* this is handled by container_content_imsm() */
2359 case MIGR_STATE_CHANGE
:
2360 /* FIXME handle other migrations */
2362 /* we are not dirty, so... */
2363 info
->resync_start
= MaxSector
;
2367 strncpy(info
->name
, (char *) dev
->volume
, MAX_RAID_SERIAL_LEN
);
2368 info
->name
[MAX_RAID_SERIAL_LEN
] = 0;
2370 info
->array
.major_version
= -1;
2371 info
->array
.minor_version
= -2;
2372 devname
= devnum2devname(st
->container_dev
);
2373 *info
->text_version
= '\0';
2375 sprintf(info
->text_version
, "/%s/%d", devname
, info
->container_member
);
2377 info
->safe_mode_delay
= 4000; /* 4 secs like the Matrix driver */
2378 uuid_from_super_imsm(st
, info
->uuid
);
2382 for (i
=0; i
<map_disks
; i
++) {
2384 if (i
< info
->array
.raid_disks
) {
2385 struct imsm_disk
*dsk
;
2386 j
= get_imsm_disk_idx(dev
, i
, -1);
2387 dsk
= get_imsm_disk(super
, j
);
2388 if (dsk
&& (dsk
->status
& CONFIGURED_DISK
))
2395 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
);
2396 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
);
2398 static struct imsm_disk
*get_imsm_missing(struct intel_super
*super
, __u8 index
)
2402 for (d
= super
->missing
; d
; d
= d
->next
)
2403 if (d
->index
== index
)
2408 static void getinfo_super_imsm(struct supertype
*st
, struct mdinfo
*info
, char *map
)
2410 struct intel_super
*super
= st
->sb
;
2411 struct imsm_disk
*disk
;
2412 int map_disks
= info
->array
.raid_disks
;
2413 int max_enough
= -1;
2415 struct imsm_super
*mpb
;
2417 if (super
->current_vol
>= 0) {
2418 getinfo_super_imsm_volume(st
, info
, map
);
2421 memset(info
, 0, sizeof(*info
));
2423 /* Set raid_disks to zero so that Assemble will always pull in valid
2426 info
->array
.raid_disks
= 0;
2427 info
->array
.level
= LEVEL_CONTAINER
;
2428 info
->array
.layout
= 0;
2429 info
->array
.md_minor
= -1;
2430 info
->array
.ctime
= 0; /* N/A for imsm */
2431 info
->array
.utime
= 0;
2432 info
->array
.chunk_size
= 0;
2434 info
->disk
.major
= 0;
2435 info
->disk
.minor
= 0;
2436 info
->disk
.raid_disk
= -1;
2437 info
->reshape_active
= 0;
2438 info
->array
.major_version
= -1;
2439 info
->array
.minor_version
= -2;
2440 strcpy(info
->text_version
, "imsm");
2441 info
->safe_mode_delay
= 0;
2442 info
->disk
.number
= -1;
2443 info
->disk
.state
= 0;
2445 info
->recovery_start
= MaxSector
;
2447 /* do we have the all the insync disks that we expect? */
2448 mpb
= super
->anchor
;
2450 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
2451 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
2452 int failed
, enough
, j
, missing
= 0;
2453 struct imsm_map
*map
;
2456 failed
= imsm_count_failed(super
, dev
);
2457 state
= imsm_check_degraded(super
, dev
, failed
);
2458 map
= get_imsm_map(dev
, dev
->vol
.migr_state
);
2460 /* any newly missing disks?
2461 * (catches single-degraded vs double-degraded)
2463 for (j
= 0; j
< map
->num_members
; j
++) {
2464 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
2465 __u32 idx
= ord_to_idx(ord
);
2467 if (!(ord
& IMSM_ORD_REBUILD
) &&
2468 get_imsm_missing(super
, idx
)) {
2474 if (state
== IMSM_T_STATE_FAILED
)
2476 else if (state
== IMSM_T_STATE_DEGRADED
&&
2477 (state
!= map
->map_state
|| missing
))
2479 else /* we're normal, or already degraded */
2482 /* in the missing/failed disk case check to see
2483 * if at least one array is runnable
2485 max_enough
= max(max_enough
, enough
);
2487 dprintf("%s: enough: %d\n", __func__
, max_enough
);
2488 info
->container_enough
= max_enough
;
2491 __u32 reserved
= imsm_reserved_sectors(super
, super
->disks
);
2493 disk
= &super
->disks
->disk
;
2494 info
->data_offset
= __le32_to_cpu(disk
->total_blocks
) - reserved
;
2495 info
->component_size
= reserved
;
2496 info
->disk
.state
= is_configured(disk
) ? (1 << MD_DISK_ACTIVE
) : 0;
2497 /* we don't change info->disk.raid_disk here because
2498 * this state will be finalized in mdmon after we have
2499 * found the 'most fresh' version of the metadata
2501 info
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2502 info
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2505 /* only call uuid_from_super_imsm when this disk is part of a populated container,
2506 * ->compare_super may have updated the 'num_raid_devs' field for spares
2508 if (info
->disk
.state
& (1 << MD_DISK_SYNC
) || super
->anchor
->num_raid_devs
)
2509 uuid_from_super_imsm(st
, info
->uuid
);
2511 memcpy(info
->uuid
, uuid_zero
, sizeof(uuid_zero
));
2513 /* I don't know how to compute 'map' on imsm, so use safe default */
2516 for (i
= 0; i
< map_disks
; i
++)
2522 /* allocates memory and fills disk in mdinfo structure
2523 * for each disk in array */
2524 struct mdinfo
*getinfo_super_disks_imsm(struct supertype
*st
)
2526 struct mdinfo
*mddev
= NULL
;
2527 struct intel_super
*super
= st
->sb
;
2528 struct imsm_disk
*disk
;
2531 if (!super
|| !super
->disks
)
2534 mddev
= malloc(sizeof(*mddev
));
2536 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2539 memset(mddev
, 0, sizeof(*mddev
));
2543 tmp
= malloc(sizeof(*tmp
));
2545 fprintf(stderr
, Name
": Failed to allocate memory.\n");
2550 memset(tmp
, 0, sizeof(*tmp
));
2552 tmp
->next
= mddev
->devs
;
2554 tmp
->disk
.number
= count
++;
2555 tmp
->disk
.major
= dl
->major
;
2556 tmp
->disk
.minor
= dl
->minor
;
2557 tmp
->disk
.state
= is_configured(disk
) ?
2558 (1 << MD_DISK_ACTIVE
) : 0;
2559 tmp
->disk
.state
|= is_failed(disk
) ? (1 << MD_DISK_FAULTY
) : 0;
2560 tmp
->disk
.state
|= is_spare(disk
) ? 0 : (1 << MD_DISK_SYNC
);
2561 tmp
->disk
.raid_disk
= -1;
2567 static int update_super_imsm(struct supertype
*st
, struct mdinfo
*info
,
2568 char *update
, char *devname
, int verbose
,
2569 int uuid_set
, char *homehost
)
2571 /* For 'assemble' and 'force' we need to return non-zero if any
2572 * change was made. For others, the return value is ignored.
2573 * Update options are:
2574 * force-one : This device looks a bit old but needs to be included,
2575 * update age info appropriately.
2576 * assemble: clear any 'faulty' flag to allow this device to
2578 * force-array: Array is degraded but being forced, mark it clean
2579 * if that will be needed to assemble it.
2581 * newdev: not used ????
2582 * grow: Array has gained a new device - this is currently for
2584 * resync: mark as dirty so a resync will happen.
2585 * name: update the name - preserving the homehost
2586 * uuid: Change the uuid of the array to match watch is given
2588 * Following are not relevant for this imsm:
2589 * sparc2.2 : update from old dodgey metadata
2590 * super-minor: change the preferred_minor number
2591 * summaries: update redundant counters.
2592 * homehost: update the recorded homehost
2593 * _reshape_progress: record new reshape_progress position.
2596 struct intel_super
*super
= st
->sb
;
2597 struct imsm_super
*mpb
;
2599 /* we can only update container info */
2600 if (!super
|| super
->current_vol
>= 0 || !super
->anchor
)
2603 mpb
= super
->anchor
;
2605 if (strcmp(update
, "uuid") == 0 && uuid_set
&& !info
->update_private
)
2607 else if (strcmp(update
, "uuid") == 0 && uuid_set
&& info
->update_private
) {
2608 mpb
->orig_family_num
= *((__u32
*) info
->update_private
);
2610 } else if (strcmp(update
, "uuid") == 0) {
2611 __u32
*new_family
= malloc(sizeof(*new_family
));
2613 /* update orig_family_number with the incoming random
2614 * data, report the new effective uuid, and store the
2615 * new orig_family_num for future updates.
2618 memcpy(&mpb
->orig_family_num
, info
->uuid
, sizeof(__u32
));
2619 uuid_from_super_imsm(st
, info
->uuid
);
2620 *new_family
= mpb
->orig_family_num
;
2621 info
->update_private
= new_family
;
2624 } else if (strcmp(update
, "assemble") == 0)
2629 /* successful update? recompute checksum */
2631 mpb
->check_sum
= __le32_to_cpu(__gen_imsm_checksum(mpb
));
2636 static size_t disks_to_mpb_size(int disks
)
2640 size
= sizeof(struct imsm_super
);
2641 size
+= (disks
- 1) * sizeof(struct imsm_disk
);
2642 size
+= 2 * sizeof(struct imsm_dev
);
2643 /* up to 2 maps per raid device (-2 for imsm_maps in imsm_dev */
2644 size
+= (4 - 2) * sizeof(struct imsm_map
);
2645 /* 4 possible disk_ord_tbl's */
2646 size
+= 4 * (disks
- 1) * sizeof(__u32
);
2651 static __u64
avail_size_imsm(struct supertype
*st
, __u64 devsize
)
2653 if (devsize
< (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
))
2656 return devsize
- (MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
);
2659 static void free_devlist(struct intel_super
*super
)
2661 struct intel_dev
*dv
;
2663 while (super
->devlist
) {
2664 dv
= super
->devlist
->next
;
2665 free(super
->devlist
->dev
);
2666 free(super
->devlist
);
2667 super
->devlist
= dv
;
2671 static void imsm_copy_dev(struct imsm_dev
*dest
, struct imsm_dev
*src
)
2673 memcpy(dest
, src
, sizeof_imsm_dev(src
, 0));
2676 static int compare_super_imsm(struct supertype
*st
, struct supertype
*tst
)
2680 * 0 same, or first was empty, and second was copied
2681 * 1 second had wrong number
2683 * 3 wrong other info
2685 struct intel_super
*first
= st
->sb
;
2686 struct intel_super
*sec
= tst
->sb
;
2693 /* in platform dependent environment test if the disks
2694 * use the same Intel hba
2696 if (!check_env("IMSM_NO_PLATFORM")) {
2697 if (!first
->hba
|| !sec
->hba
||
2698 (first
->hba
->type
!= sec
->hba
->type
)) {
2700 "HBAs of devices does not match %s != %s\n",
2701 first
->hba
? get_sys_dev_type(first
->hba
->type
) : NULL
,
2702 sec
->hba
? get_sys_dev_type(sec
->hba
->type
) : NULL
);
2707 /* if an anchor does not have num_raid_devs set then it is a free
2710 if (first
->anchor
->num_raid_devs
> 0 &&
2711 sec
->anchor
->num_raid_devs
> 0) {
2712 /* Determine if these disks might ever have been
2713 * related. Further disambiguation can only take place
2714 * in load_super_imsm_all
2716 __u32 first_family
= first
->anchor
->orig_family_num
;
2717 __u32 sec_family
= sec
->anchor
->orig_family_num
;
2719 if (memcmp(first
->anchor
->sig
, sec
->anchor
->sig
,
2720 MAX_SIGNATURE_LENGTH
) != 0)
2723 if (first_family
== 0)
2724 first_family
= first
->anchor
->family_num
;
2725 if (sec_family
== 0)
2726 sec_family
= sec
->anchor
->family_num
;
2728 if (first_family
!= sec_family
)
2734 /* if 'first' is a spare promote it to a populated mpb with sec's
2737 if (first
->anchor
->num_raid_devs
== 0 &&
2738 sec
->anchor
->num_raid_devs
> 0) {
2740 struct intel_dev
*dv
;
2741 struct imsm_dev
*dev
;
2743 /* we need to copy raid device info from sec if an allocation
2744 * fails here we don't associate the spare
2746 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++) {
2747 dv
= malloc(sizeof(*dv
));
2750 dev
= malloc(sizeof_imsm_dev(get_imsm_dev(sec
, i
), 1));
2757 dv
->next
= first
->devlist
;
2758 first
->devlist
= dv
;
2760 if (i
< sec
->anchor
->num_raid_devs
) {
2761 /* allocation failure */
2762 free_devlist(first
);
2763 fprintf(stderr
, "imsm: failed to associate spare\n");
2766 first
->anchor
->num_raid_devs
= sec
->anchor
->num_raid_devs
;
2767 first
->anchor
->orig_family_num
= sec
->anchor
->orig_family_num
;
2768 first
->anchor
->family_num
= sec
->anchor
->family_num
;
2769 memcpy(first
->anchor
->sig
, sec
->anchor
->sig
, MAX_SIGNATURE_LENGTH
);
2770 for (i
= 0; i
< sec
->anchor
->num_raid_devs
; i
++)
2771 imsm_copy_dev(get_imsm_dev(first
, i
), get_imsm_dev(sec
, i
));
2777 static void fd2devname(int fd
, char *name
)
2781 char dname
[PATH_MAX
];
2786 if (fstat(fd
, &st
) != 0)
2788 sprintf(path
, "/sys/dev/block/%d:%d",
2789 major(st
.st_rdev
), minor(st
.st_rdev
));
2791 rv
= readlink(path
, dname
, sizeof(dname
));
2796 nm
= strrchr(dname
, '/');
2798 snprintf(name
, MAX_RAID_SERIAL_LEN
, "/dev/%s", nm
);
2801 extern int scsi_get_serial(int fd
, void *buf
, size_t buf_len
);
2803 static int imsm_read_serial(int fd
, char *devname
,
2804 __u8 serial
[MAX_RAID_SERIAL_LEN
])
2806 unsigned char scsi_serial
[255];
2815 memset(scsi_serial
, 0, sizeof(scsi_serial
));
2817 rv
= scsi_get_serial(fd
, scsi_serial
, sizeof(scsi_serial
));
2819 if (rv
&& check_env("IMSM_DEVNAME_AS_SERIAL")) {
2820 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2821 fd2devname(fd
, (char *) serial
);
2828 Name
": Failed to retrieve serial for %s\n",
2833 rsp_len
= scsi_serial
[3];
2837 Name
": Failed to retrieve serial for %s\n",
2841 rsp_buf
= (char *) &scsi_serial
[4];
2843 /* trim all whitespace and non-printable characters and convert
2846 for (i
= 0, dest
= rsp_buf
; i
< rsp_len
; i
++) {
2849 /* ':' is reserved for use in placeholder serial
2850 * numbers for missing disks
2858 len
= dest
- rsp_buf
;
2861 /* truncate leading characters */
2862 if (len
> MAX_RAID_SERIAL_LEN
) {
2863 dest
+= len
- MAX_RAID_SERIAL_LEN
;
2864 len
= MAX_RAID_SERIAL_LEN
;
2867 memset(serial
, 0, MAX_RAID_SERIAL_LEN
);
2868 memcpy(serial
, dest
, len
);
2873 static int serialcmp(__u8
*s1
, __u8
*s2
)
2875 return strncmp((char *) s1
, (char *) s2
, MAX_RAID_SERIAL_LEN
);
2878 static void serialcpy(__u8
*dest
, __u8
*src
)
2880 strncpy((char *) dest
, (char *) src
, MAX_RAID_SERIAL_LEN
);
2884 static struct dl
*serial_to_dl(__u8
*serial
, struct intel_super
*super
)
2888 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
2889 if (serialcmp(dl
->serial
, serial
) == 0)
2896 static struct imsm_disk
*
2897 __serial_to_disk(__u8
*serial
, struct imsm_super
*mpb
, int *idx
)
2901 for (i
= 0; i
< mpb
->num_disks
; i
++) {
2902 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
2904 if (serialcmp(disk
->serial
, serial
) == 0) {
2915 load_imsm_disk(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
2917 struct imsm_disk
*disk
;
2922 __u8 serial
[MAX_RAID_SERIAL_LEN
];
2924 rv
= imsm_read_serial(fd
, devname
, serial
);
2929 dl
= calloc(1, sizeof(*dl
));
2933 Name
": failed to allocate disk buffer for %s\n",
2939 dl
->major
= major(stb
.st_rdev
);
2940 dl
->minor
= minor(stb
.st_rdev
);
2941 dl
->next
= super
->disks
;
2942 dl
->fd
= keep_fd
? fd
: -1;
2943 assert(super
->disks
== NULL
);
2945 serialcpy(dl
->serial
, serial
);
2948 fd2devname(fd
, name
);
2950 dl
->devname
= strdup(devname
);
2952 dl
->devname
= strdup(name
);
2954 /* look up this disk's index in the current anchor */
2955 disk
= __serial_to_disk(dl
->serial
, super
->anchor
, &dl
->index
);
2958 /* only set index on disks that are a member of a
2959 * populated contianer, i.e. one with raid_devs
2961 if (is_failed(&dl
->disk
))
2963 else if (is_spare(&dl
->disk
))
2971 /* When migrating map0 contains the 'destination' state while map1
2972 * contains the current state. When not migrating map0 contains the
2973 * current state. This routine assumes that map[0].map_state is set to
2974 * the current array state before being called.
2976 * Migration is indicated by one of the following states
2977 * 1/ Idle (migr_state=0 map0state=normal||unitialized||degraded||failed)
2978 * 2/ Initialize (migr_state=1 migr_type=MIGR_INIT map0state=normal
2979 * map1state=unitialized)
2980 * 3/ Repair (Resync) (migr_state=1 migr_type=MIGR_REPAIR map0state=normal
2982 * 4/ Rebuild (migr_state=1 migr_type=MIGR_REBUILD map0state=normal
2983 * map1state=degraded)
2984 * 5/ Migration (mig_state=1 migr_type=MIGR_GEN_MIGR map0state=normal
2987 static void migrate(struct imsm_dev
*dev
, struct intel_super
*super
,
2988 __u8 to_state
, int migr_type
)
2990 struct imsm_map
*dest
;
2991 struct imsm_map
*src
= get_imsm_map(dev
, 0);
2993 dev
->vol
.migr_state
= 1;
2994 set_migr_type(dev
, migr_type
);
2995 dev
->vol
.curr_migr_unit
= 0;
2996 dest
= get_imsm_map(dev
, 1);
2998 /* duplicate and then set the target end state in map[0] */
2999 memcpy(dest
, src
, sizeof_imsm_map(src
));
3000 if ((migr_type
== MIGR_REBUILD
) ||
3001 (migr_type
== MIGR_GEN_MIGR
)) {
3005 for (i
= 0; i
< src
->num_members
; i
++) {
3006 ord
= __le32_to_cpu(src
->disk_ord_tbl
[i
]);
3007 set_imsm_ord_tbl_ent(src
, i
, ord_to_idx(ord
));
3011 if (migr_type
== MIGR_GEN_MIGR
)
3012 /* Clear migration record */
3013 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
3015 src
->map_state
= to_state
;
3018 static void end_migration(struct imsm_dev
*dev
, __u8 map_state
)
3020 struct imsm_map
*map
= get_imsm_map(dev
, 0);
3021 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
3024 /* merge any IMSM_ORD_REBUILD bits that were not successfully
3025 * completed in the last migration.
3027 * FIXME add support for raid-level-migration
3029 for (i
= 0; i
< prev
->num_members
; i
++)
3030 for (j
= 0; j
< map
->num_members
; j
++)
3031 /* during online capacity expansion
3032 * disks position can be changed if takeover is used
3034 if (ord_to_idx(map
->disk_ord_tbl
[j
]) ==
3035 ord_to_idx(prev
->disk_ord_tbl
[i
])) {
3036 map
->disk_ord_tbl
[j
] |= prev
->disk_ord_tbl
[i
];
3040 dev
->vol
.migr_state
= 0;
3041 dev
->vol
.migr_type
= 0;
3042 dev
->vol
.curr_migr_unit
= 0;
3043 map
->map_state
= map_state
;
3047 static int parse_raid_devices(struct intel_super
*super
)
3050 struct imsm_dev
*dev_new
;
3051 size_t len
, len_migr
;
3053 size_t space_needed
= 0;
3054 struct imsm_super
*mpb
= super
->anchor
;
3056 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3057 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3058 struct intel_dev
*dv
;
3060 len
= sizeof_imsm_dev(dev_iter
, 0);
3061 len_migr
= sizeof_imsm_dev(dev_iter
, 1);
3063 space_needed
+= len_migr
- len
;
3065 dv
= malloc(sizeof(*dv
));
3068 if (max_len
< len_migr
)
3070 if (max_len
> len_migr
)
3071 space_needed
+= max_len
- len_migr
;
3072 dev_new
= malloc(max_len
);
3077 imsm_copy_dev(dev_new
, dev_iter
);
3080 dv
->next
= super
->devlist
;
3081 super
->devlist
= dv
;
3084 /* ensure that super->buf is large enough when all raid devices
3087 if (__le32_to_cpu(mpb
->mpb_size
) + space_needed
> super
->len
) {
3090 len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + space_needed
, 512);
3091 if (posix_memalign(&buf
, 512, len
) != 0)
3094 memcpy(buf
, super
->buf
, super
->len
);
3095 memset(buf
+ super
->len
, 0, len
- super
->len
);
3104 /* retrieve a pointer to the bbm log which starts after all raid devices */
3105 struct bbm_log
*__get_imsm_bbm_log(struct imsm_super
*mpb
)
3109 if (__le32_to_cpu(mpb
->bbm_log_size
)) {
3111 ptr
+= mpb
->mpb_size
- __le32_to_cpu(mpb
->bbm_log_size
);
3117 /*******************************************************************************
3118 * Function: check_mpb_migr_compatibility
3119 * Description: Function checks for unsupported migration features:
3120 * - migration optimization area (pba_of_lba0)
3121 * - descending reshape (ascending_migr)
3123 * super : imsm metadata information
3125 * 0 : migration is compatible
3126 * -1 : migration is not compatible
3127 ******************************************************************************/
3128 int check_mpb_migr_compatibility(struct intel_super
*super
)
3130 struct imsm_map
*map0
, *map1
;
3131 struct migr_record
*migr_rec
= super
->migr_rec
;
3134 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
3135 struct imsm_dev
*dev_iter
= __get_imsm_dev(super
->anchor
, i
);
3138 dev_iter
->vol
.migr_state
== 1 &&
3139 dev_iter
->vol
.migr_type
== MIGR_GEN_MIGR
) {
3140 /* This device is migrating */
3141 map0
= get_imsm_map(dev_iter
, 0);
3142 map1
= get_imsm_map(dev_iter
, 1);
3143 if (map0
->pba_of_lba0
!= map1
->pba_of_lba0
)
3144 /* migration optimization area was used */
3146 if (migr_rec
->ascending_migr
== 0
3147 && migr_rec
->dest_depth_per_unit
> 0)
3148 /* descending reshape not supported yet */
3155 static void __free_imsm(struct intel_super
*super
, int free_disks
);
3157 /* load_imsm_mpb - read matrix metadata
3158 * allocates super->mpb to be freed by free_imsm
3160 static int load_imsm_mpb(int fd
, struct intel_super
*super
, char *devname
)
3162 unsigned long long dsize
;
3163 unsigned long long sectors
;
3165 struct imsm_super
*anchor
;
3168 get_dev_size(fd
, NULL
, &dsize
);
3172 Name
": %s: device to small for imsm\n",
3177 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0) {
3179 fprintf(stderr
, Name
3180 ": Cannot seek to anchor block on %s: %s\n",
3181 devname
, strerror(errno
));
3185 if (posix_memalign((void**)&anchor
, 512, 512) != 0) {
3188 Name
": Failed to allocate imsm anchor buffer"
3189 " on %s\n", devname
);
3192 if (read(fd
, anchor
, 512) != 512) {
3195 Name
": Cannot read anchor block on %s: %s\n",
3196 devname
, strerror(errno
));
3201 if (strncmp((char *) anchor
->sig
, MPB_SIGNATURE
, MPB_SIG_LEN
) != 0) {
3204 Name
": no IMSM anchor on %s\n", devname
);
3209 __free_imsm(super
, 0);
3210 /* reload capability and hba */
3212 /* capability and hba must be updated with new super allocation */
3213 find_intel_hba_capability(fd
, super
, devname
);
3214 super
->len
= ROUND_UP(anchor
->mpb_size
, 512);
3215 if (posix_memalign(&super
->buf
, 512, super
->len
) != 0) {
3218 Name
": unable to allocate %zu byte mpb buffer\n",
3223 memcpy(super
->buf
, anchor
, 512);
3225 sectors
= mpb_sectors(anchor
) - 1;
3228 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
3229 fprintf(stderr
, Name
3230 ": %s could not allocate migr_rec buffer\n", __func__
);
3236 check_sum
= __gen_imsm_checksum(super
->anchor
);
3237 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3240 Name
": IMSM checksum %x != %x on %s\n",
3242 __le32_to_cpu(super
->anchor
->check_sum
),
3250 /* read the extended mpb */
3251 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0) {
3254 Name
": Cannot seek to extended mpb on %s: %s\n",
3255 devname
, strerror(errno
));
3259 if ((unsigned)read(fd
, super
->buf
+ 512, super
->len
- 512) != super
->len
- 512) {
3262 Name
": Cannot read extended mpb on %s: %s\n",
3263 devname
, strerror(errno
));
3267 check_sum
= __gen_imsm_checksum(super
->anchor
);
3268 if (check_sum
!= __le32_to_cpu(super
->anchor
->check_sum
)) {
3271 Name
": IMSM checksum %x != %x on %s\n",
3272 check_sum
, __le32_to_cpu(super
->anchor
->check_sum
),
3277 /* FIXME the BBM log is disk specific so we cannot use this global
3278 * buffer for all disks. Ok for now since we only look at the global
3279 * bbm_log_size parameter to gate assembly
3281 super
->bbm_log
= __get_imsm_bbm_log(super
->anchor
);
3286 static int read_imsm_migr_rec(int fd
, struct intel_super
*super
);
3289 load_and_parse_mpb(int fd
, struct intel_super
*super
, char *devname
, int keep_fd
)
3293 err
= load_imsm_mpb(fd
, super
, devname
);
3296 err
= load_imsm_disk(fd
, super
, devname
, keep_fd
);
3299 err
= parse_raid_devices(super
);
3304 static void __free_imsm_disk(struct dl
*d
)
3316 static void free_imsm_disks(struct intel_super
*super
)
3320 while (super
->disks
) {
3322 super
->disks
= d
->next
;
3323 __free_imsm_disk(d
);
3325 while (super
->disk_mgmt_list
) {
3326 d
= super
->disk_mgmt_list
;
3327 super
->disk_mgmt_list
= d
->next
;
3328 __free_imsm_disk(d
);
3330 while (super
->missing
) {
3332 super
->missing
= d
->next
;
3333 __free_imsm_disk(d
);
3338 /* free all the pieces hanging off of a super pointer */
3339 static void __free_imsm(struct intel_super
*super
, int free_disks
)
3341 struct intel_hba
*elem
, *next
;
3347 /* unlink capability description */
3349 if (super
->migr_rec_buf
) {
3350 free(super
->migr_rec_buf
);
3351 super
->migr_rec_buf
= NULL
;
3354 free_imsm_disks(super
);
3355 free_devlist(super
);
3359 free((void *)elem
->path
);
3367 static void free_imsm(struct intel_super
*super
)
3369 __free_imsm(super
, 1);
3373 static void free_super_imsm(struct supertype
*st
)
3375 struct intel_super
*super
= st
->sb
;
3384 static struct intel_super
*alloc_super(void)
3386 struct intel_super
*super
= malloc(sizeof(*super
));
3389 memset(super
, 0, sizeof(*super
));
3390 super
->current_vol
= -1;
3391 super
->create_offset
= ~((__u32
) 0);
3397 * find and allocate hba and OROM/EFI based on valid fd of RAID component device
3399 static int find_intel_hba_capability(int fd
, struct intel_super
*super
, char *devname
)
3401 struct sys_dev
*hba_name
;
3404 if ((fd
< 0) || check_env("IMSM_NO_PLATFORM")) {
3409 hba_name
= find_disk_attached_hba(fd
, NULL
);
3413 Name
": %s is not attached to Intel(R) RAID controller.\n",
3417 rv
= attach_hba_to_super(super
, hba_name
);
3420 struct intel_hba
*hba
= super
->hba
;
3422 fprintf(stderr
, Name
": %s is attached to Intel(R) %s RAID "
3423 "controller (%s),\n"
3424 " but the container is assigned to Intel(R) "
3425 "%s RAID controller (",
3428 hba_name
->pci_id
? : "Err!",
3429 get_sys_dev_type(hba_name
->type
));
3432 fprintf(stderr
, "%s", hba
->pci_id
? : "Err!");
3434 fprintf(stderr
, ", ");
3438 fprintf(stderr
, ").\n"
3439 " Mixing devices attached to different controllers "
3440 "is not allowed.\n");
3442 free_sys_dev(&hba_name
);
3445 super
->orom
= find_imsm_capability(hba_name
->type
);
3446 free_sys_dev(&hba_name
);
3453 /* find_missing - helper routine for load_super_imsm_all that identifies
3454 * disks that have disappeared from the system. This routine relies on
3455 * the mpb being uptodate, which it is at load time.
3457 static int find_missing(struct intel_super
*super
)
3460 struct imsm_super
*mpb
= super
->anchor
;
3462 struct imsm_disk
*disk
;
3464 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3465 disk
= __get_imsm_disk(mpb
, i
);
3466 dl
= serial_to_dl(disk
->serial
, super
);
3470 dl
= malloc(sizeof(*dl
));
3476 dl
->devname
= strdup("missing");
3478 serialcpy(dl
->serial
, disk
->serial
);
3481 dl
->next
= super
->missing
;
3482 super
->missing
= dl
;
3488 static struct intel_disk
*disk_list_get(__u8
*serial
, struct intel_disk
*disk_list
)
3490 struct intel_disk
*idisk
= disk_list
;
3493 if (serialcmp(idisk
->disk
.serial
, serial
) == 0)
3495 idisk
= idisk
->next
;
3501 static int __prep_thunderdome(struct intel_super
**table
, int tbl_size
,
3502 struct intel_super
*super
,
3503 struct intel_disk
**disk_list
)
3505 struct imsm_disk
*d
= &super
->disks
->disk
;
3506 struct imsm_super
*mpb
= super
->anchor
;
3509 for (i
= 0; i
< tbl_size
; i
++) {
3510 struct imsm_super
*tbl_mpb
= table
[i
]->anchor
;
3511 struct imsm_disk
*tbl_d
= &table
[i
]->disks
->disk
;
3513 if (tbl_mpb
->family_num
== mpb
->family_num
) {
3514 if (tbl_mpb
->check_sum
== mpb
->check_sum
) {
3515 dprintf("%s: mpb from %d:%d matches %d:%d\n",
3516 __func__
, super
->disks
->major
,
3517 super
->disks
->minor
,
3518 table
[i
]->disks
->major
,
3519 table
[i
]->disks
->minor
);
3523 if (((is_configured(d
) && !is_configured(tbl_d
)) ||
3524 is_configured(d
) == is_configured(tbl_d
)) &&
3525 tbl_mpb
->generation_num
< mpb
->generation_num
) {
3526 /* current version of the mpb is a
3527 * better candidate than the one in
3528 * super_table, but copy over "cross
3529 * generational" status
3531 struct intel_disk
*idisk
;
3533 dprintf("%s: mpb from %d:%d replaces %d:%d\n",
3534 __func__
, super
->disks
->major
,
3535 super
->disks
->minor
,
3536 table
[i
]->disks
->major
,
3537 table
[i
]->disks
->minor
);
3539 idisk
= disk_list_get(tbl_d
->serial
, *disk_list
);
3540 if (idisk
&& is_failed(&idisk
->disk
))
3541 tbl_d
->status
|= FAILED_DISK
;
3544 struct intel_disk
*idisk
;
3545 struct imsm_disk
*disk
;
3547 /* tbl_mpb is more up to date, but copy
3548 * over cross generational status before
3551 disk
= __serial_to_disk(d
->serial
, mpb
, NULL
);
3552 if (disk
&& is_failed(disk
))
3553 d
->status
|= FAILED_DISK
;
3555 idisk
= disk_list_get(d
->serial
, *disk_list
);
3558 if (disk
&& is_configured(disk
))
3559 idisk
->disk
.status
|= CONFIGURED_DISK
;
3562 dprintf("%s: mpb from %d:%d prefer %d:%d\n",
3563 __func__
, super
->disks
->major
,
3564 super
->disks
->minor
,
3565 table
[i
]->disks
->major
,
3566 table
[i
]->disks
->minor
);
3574 table
[tbl_size
++] = super
;
3578 /* update/extend the merged list of imsm_disk records */
3579 for (j
= 0; j
< mpb
->num_disks
; j
++) {
3580 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, j
);
3581 struct intel_disk
*idisk
;
3583 idisk
= disk_list_get(disk
->serial
, *disk_list
);
3585 idisk
->disk
.status
|= disk
->status
;
3586 if (is_configured(&idisk
->disk
) ||
3587 is_failed(&idisk
->disk
))
3588 idisk
->disk
.status
&= ~(SPARE_DISK
);
3590 idisk
= calloc(1, sizeof(*idisk
));
3593 idisk
->owner
= IMSM_UNKNOWN_OWNER
;
3594 idisk
->disk
= *disk
;
3595 idisk
->next
= *disk_list
;
3599 if (serialcmp(idisk
->disk
.serial
, d
->serial
) == 0)
3606 static struct intel_super
*
3607 validate_members(struct intel_super
*super
, struct intel_disk
*disk_list
,
3610 struct imsm_super
*mpb
= super
->anchor
;
3614 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3615 struct imsm_disk
*disk
= __get_imsm_disk(mpb
, i
);
3616 struct intel_disk
*idisk
;
3618 idisk
= disk_list_get(disk
->serial
, disk_list
);
3620 if (idisk
->owner
== owner
||
3621 idisk
->owner
== IMSM_UNKNOWN_OWNER
)
3624 dprintf("%s: '%.16s' owner %d != %d\n",
3625 __func__
, disk
->serial
, idisk
->owner
,
3628 dprintf("%s: unknown disk %x [%d]: %.16s\n",
3629 __func__
, __le32_to_cpu(mpb
->family_num
), i
,
3635 if (ok_count
== mpb
->num_disks
)
3640 static void show_conflicts(__u32 family_num
, struct intel_super
*super_list
)
3642 struct intel_super
*s
;
3644 for (s
= super_list
; s
; s
= s
->next
) {
3645 if (family_num
!= s
->anchor
->family_num
)
3647 fprintf(stderr
, "Conflict, offlining family %#x on '%s'\n",
3648 __le32_to_cpu(family_num
), s
->disks
->devname
);
3652 static struct intel_super
*
3653 imsm_thunderdome(struct intel_super
**super_list
, int len
)
3655 struct intel_super
*super_table
[len
];
3656 struct intel_disk
*disk_list
= NULL
;
3657 struct intel_super
*champion
, *spare
;
3658 struct intel_super
*s
, **del
;
3663 memset(super_table
, 0, sizeof(super_table
));
3664 for (s
= *super_list
; s
; s
= s
->next
)
3665 tbl_size
= __prep_thunderdome(super_table
, tbl_size
, s
, &disk_list
);
3667 for (i
= 0; i
< tbl_size
; i
++) {
3668 struct imsm_disk
*d
;
3669 struct intel_disk
*idisk
;
3670 struct imsm_super
*mpb
= super_table
[i
]->anchor
;
3673 d
= &s
->disks
->disk
;
3675 /* 'd' must appear in merged disk list for its
3676 * configuration to be valid
3678 idisk
= disk_list_get(d
->serial
, disk_list
);
3679 if (idisk
&& idisk
->owner
== i
)
3680 s
= validate_members(s
, disk_list
, i
);
3685 dprintf("%s: marking family: %#x from %d:%d offline\n",
3686 __func__
, mpb
->family_num
,
3687 super_table
[i
]->disks
->major
,
3688 super_table
[i
]->disks
->minor
);
3692 /* This is where the mdadm implementation differs from the Windows
3693 * driver which has no strict concept of a container. We can only
3694 * assemble one family from a container, so when returning a prodigal
3695 * array member to this system the code will not be able to disambiguate
3696 * the container contents that should be assembled ("foreign" versus
3697 * "local"). It requires user intervention to set the orig_family_num
3698 * to a new value to establish a new container. The Windows driver in
3699 * this situation fixes up the volume name in place and manages the
3700 * foreign array as an independent entity.
3705 for (i
= 0; i
< tbl_size
; i
++) {
3706 struct intel_super
*tbl_ent
= super_table
[i
];
3712 if (tbl_ent
->anchor
->num_raid_devs
== 0) {
3717 if (s
&& !is_spare
) {
3718 show_conflicts(tbl_ent
->anchor
->family_num
, *super_list
);
3720 } else if (!s
&& !is_spare
)
3733 fprintf(stderr
, "Chose family %#x on '%s', "
3734 "assemble conflicts to new container with '--update=uuid'\n",
3735 __le32_to_cpu(s
->anchor
->family_num
), s
->disks
->devname
);
3737 /* collect all dl's onto 'champion', and update them to
3738 * champion's version of the status
3740 for (s
= *super_list
; s
; s
= s
->next
) {
3741 struct imsm_super
*mpb
= champion
->anchor
;
3742 struct dl
*dl
= s
->disks
;
3747 for (i
= 0; i
< mpb
->num_disks
; i
++) {
3748 struct imsm_disk
*disk
;
3750 disk
= __serial_to_disk(dl
->serial
, mpb
, &dl
->index
);
3753 /* only set index on disks that are a member of
3754 * a populated contianer, i.e. one with
3757 if (is_failed(&dl
->disk
))
3759 else if (is_spare(&dl
->disk
))
3765 if (i
>= mpb
->num_disks
) {
3766 struct intel_disk
*idisk
;
3768 idisk
= disk_list_get(dl
->serial
, disk_list
);
3769 if (idisk
&& is_spare(&idisk
->disk
) &&
3770 !is_failed(&idisk
->disk
) && !is_configured(&idisk
->disk
))
3778 dl
->next
= champion
->disks
;
3779 champion
->disks
= dl
;
3783 /* delete 'champion' from super_list */
3784 for (del
= super_list
; *del
; ) {
3785 if (*del
== champion
) {
3786 *del
= (*del
)->next
;
3789 del
= &(*del
)->next
;
3791 champion
->next
= NULL
;
3795 struct intel_disk
*idisk
= disk_list
;
3797 disk_list
= disk_list
->next
;
3804 static int load_super_imsm_all(struct supertype
*st
, int fd
, void **sbp
,
3808 struct intel_super
*super_list
= NULL
;
3809 struct intel_super
*super
= NULL
;
3810 int devnum
= fd2devnum(fd
);
3816 /* check if 'fd' an opened container */
3817 sra
= sysfs_read(fd
, 0, GET_LEVEL
|GET_VERSION
|GET_DEVS
|GET_STATE
);
3821 if (sra
->array
.major_version
!= -1 ||
3822 sra
->array
.minor_version
!= -2 ||
3823 strcmp(sra
->text_version
, "imsm") != 0) {
3828 for (sd
= sra
->devs
, i
= 0; sd
; sd
= sd
->next
, i
++) {
3829 struct intel_super
*s
= alloc_super();
3837 s
->next
= super_list
;
3841 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
3842 dfd
= dev_open(nm
, O_RDWR
);
3846 rv
= find_intel_hba_capability(dfd
, s
, devname
);
3847 /* no orom/efi or non-intel hba of the disk */
3851 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3853 /* retry the load if we might have raced against mdmon */
3854 if (err
== 3 && mdmon_running(devnum
))
3855 for (retry
= 0; retry
< 3; retry
++) {
3857 err
= load_and_parse_mpb(dfd
, s
, NULL
, 1);
3865 /* all mpbs enter, maybe one leaves */
3866 super
= imsm_thunderdome(&super_list
, i
);
3872 if (find_missing(super
) != 0) {
3878 /* load migration record */
3879 err
= load_imsm_migr_rec(super
, NULL
);
3885 /* Check migration compatibility */
3886 if (check_mpb_migr_compatibility(super
) != 0) {
3887 fprintf(stderr
, Name
": Unsupported migration detected");
3889 fprintf(stderr
, " on %s\n", devname
);
3891 fprintf(stderr
, " (IMSM).\n");
3900 while (super_list
) {
3901 struct intel_super
*s
= super_list
;
3903 super_list
= super_list
->next
;
3912 st
->container_dev
= devnum
;
3913 if (err
== 0 && st
->ss
== NULL
) {
3914 st
->ss
= &super_imsm
;
3915 st
->minor_version
= 0;
3916 st
->max_devs
= IMSM_MAX_DEVICES
;
3921 static int load_container_imsm(struct supertype
*st
, int fd
, char *devname
)
3923 return load_super_imsm_all(st
, fd
, &st
->sb
, devname
);
3927 static int load_super_imsm(struct supertype
*st
, int fd
, char *devname
)
3929 struct intel_super
*super
;
3932 if (test_partition(fd
))
3933 /* IMSM not allowed on partitions */
3936 free_super_imsm(st
);
3938 super
= alloc_super();
3941 Name
": malloc of %zu failed.\n",
3945 /* Load hba and capabilities if they exist.
3946 * But do not preclude loading metadata in case capabilities or hba are
3947 * non-compliant and ignore_hw_compat is set.
3949 rv
= find_intel_hba_capability(fd
, super
, devname
);
3950 /* no orom/efi or non-intel hba of the disk */
3951 if ((rv
!= 0) && (st
->ignore_hw_compat
== 0)) {
3954 Name
": No OROM/EFI properties for %s\n", devname
);
3958 rv
= load_and_parse_mpb(fd
, super
, devname
, 0);
3963 Name
": Failed to load all information "
3964 "sections on %s\n", devname
);
3970 if (st
->ss
== NULL
) {
3971 st
->ss
= &super_imsm
;
3972 st
->minor_version
= 0;
3973 st
->max_devs
= IMSM_MAX_DEVICES
;
3976 /* load migration record */
3977 if (load_imsm_migr_rec(super
, NULL
) == 0) {
3978 /* Check for unsupported migration features */
3979 if (check_mpb_migr_compatibility(super
) != 0) {
3981 Name
": Unsupported migration detected");
3983 fprintf(stderr
, " on %s\n", devname
);
3985 fprintf(stderr
, " (IMSM).\n");
3993 static __u16
info_to_blocks_per_strip(mdu_array_info_t
*info
)
3995 if (info
->level
== 1)
3997 return info
->chunk_size
>> 9;
4000 static __u32
info_to_num_data_stripes(mdu_array_info_t
*info
, int num_domains
)
4004 num_stripes
= (info
->size
* 2) / info_to_blocks_per_strip(info
);
4005 num_stripes
/= num_domains
;
4010 static __u32
info_to_blocks_per_member(mdu_array_info_t
*info
)
4012 if (info
->level
== 1)
4013 return info
->size
* 2;
4015 return (info
->size
* 2) & ~(info_to_blocks_per_strip(info
) - 1);
4018 static void imsm_update_version_info(struct intel_super
*super
)
4020 /* update the version and attributes */
4021 struct imsm_super
*mpb
= super
->anchor
;
4023 struct imsm_dev
*dev
;
4024 struct imsm_map
*map
;
4027 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4028 dev
= get_imsm_dev(super
, i
);
4029 map
= get_imsm_map(dev
, 0);
4030 if (__le32_to_cpu(dev
->size_high
) > 0)
4031 mpb
->attributes
|= MPB_ATTRIB_2TB
;
4033 /* FIXME detect when an array spans a port multiplier */
4035 mpb
->attributes
|= MPB_ATTRIB_PM
;
4038 if (mpb
->num_raid_devs
> 1 ||
4039 mpb
->attributes
!= MPB_ATTRIB_CHECKSUM_VERIFY
) {
4040 version
= MPB_VERSION_ATTRIBS
;
4041 switch (get_imsm_raid_level(map
)) {
4042 case 0: mpb
->attributes
|= MPB_ATTRIB_RAID0
; break;
4043 case 1: mpb
->attributes
|= MPB_ATTRIB_RAID1
; break;
4044 case 10: mpb
->attributes
|= MPB_ATTRIB_RAID10
; break;
4045 case 5: mpb
->attributes
|= MPB_ATTRIB_RAID5
; break;
4048 if (map
->num_members
>= 5)
4049 version
= MPB_VERSION_5OR6_DISK_ARRAY
;
4050 else if (dev
->status
== DEV_CLONE_N_GO
)
4051 version
= MPB_VERSION_CNG
;
4052 else if (get_imsm_raid_level(map
) == 5)
4053 version
= MPB_VERSION_RAID5
;
4054 else if (map
->num_members
>= 3)
4055 version
= MPB_VERSION_3OR4_DISK_ARRAY
;
4056 else if (get_imsm_raid_level(map
) == 1)
4057 version
= MPB_VERSION_RAID1
;
4059 version
= MPB_VERSION_RAID0
;
4061 strcpy(((char *) mpb
->sig
) + strlen(MPB_SIGNATURE
), version
);
4065 static int check_name(struct intel_super
*super
, char *name
, int quiet
)
4067 struct imsm_super
*mpb
= super
->anchor
;
4068 char *reason
= NULL
;
4071 if (strlen(name
) > MAX_RAID_SERIAL_LEN
)
4072 reason
= "must be 16 characters or less";
4074 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4075 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
4077 if (strncmp((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
) == 0) {
4078 reason
= "already exists";
4083 if (reason
&& !quiet
)
4084 fprintf(stderr
, Name
": imsm volume name %s\n", reason
);
4089 static int init_super_imsm_volume(struct supertype
*st
, mdu_array_info_t
*info
,
4090 unsigned long long size
, char *name
,
4091 char *homehost
, int *uuid
)
4093 /* We are creating a volume inside a pre-existing container.
4094 * so st->sb is already set.
4096 struct intel_super
*super
= st
->sb
;
4097 struct imsm_super
*mpb
= super
->anchor
;
4098 struct intel_dev
*dv
;
4099 struct imsm_dev
*dev
;
4100 struct imsm_vol
*vol
;
4101 struct imsm_map
*map
;
4102 int idx
= mpb
->num_raid_devs
;
4104 unsigned long long array_blocks
;
4105 size_t size_old
, size_new
;
4106 __u32 num_data_stripes
;
4108 if (super
->orom
&& mpb
->num_raid_devs
>= super
->orom
->vpa
) {
4109 fprintf(stderr
, Name
": This imsm-container already has the "
4110 "maximum of %d volumes\n", super
->orom
->vpa
);
4114 /* ensure the mpb is large enough for the new data */
4115 size_old
= __le32_to_cpu(mpb
->mpb_size
);
4116 size_new
= disks_to_mpb_size(info
->nr_disks
);
4117 if (size_new
> size_old
) {
4119 size_t size_round
= ROUND_UP(size_new
, 512);
4121 if (posix_memalign(&mpb_new
, 512, size_round
) != 0) {
4122 fprintf(stderr
, Name
": could not allocate new mpb\n");
4125 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4126 fprintf(stderr
, Name
4127 ": %s could not allocate migr_rec buffer\n",
4133 memcpy(mpb_new
, mpb
, size_old
);
4136 super
->anchor
= mpb_new
;
4137 mpb
->mpb_size
= __cpu_to_le32(size_new
);
4138 memset(mpb_new
+ size_old
, 0, size_round
- size_old
);
4140 super
->current_vol
= idx
;
4141 /* when creating the first raid device in this container set num_disks
4142 * to zero, i.e. delete this spare and add raid member devices in
4143 * add_to_super_imsm_volume()
4145 if (super
->current_vol
== 0)
4148 if (!check_name(super
, name
, 0))
4150 dv
= malloc(sizeof(*dv
));
4152 fprintf(stderr
, Name
": failed to allocate device list entry\n");
4155 dev
= calloc(1, sizeof(*dev
) + sizeof(__u32
) * (info
->raid_disks
- 1));
4158 fprintf(stderr
, Name
": could not allocate raid device\n");
4162 strncpy((char *) dev
->volume
, name
, MAX_RAID_SERIAL_LEN
);
4163 if (info
->level
== 1)
4164 array_blocks
= info_to_blocks_per_member(info
);
4166 array_blocks
= calc_array_size(info
->level
, info
->raid_disks
,
4167 info
->layout
, info
->chunk_size
,
4169 /* round array size down to closest MB */
4170 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
4172 dev
->size_low
= __cpu_to_le32((__u32
) array_blocks
);
4173 dev
->size_high
= __cpu_to_le32((__u32
) (array_blocks
>> 32));
4174 dev
->status
= (DEV_READ_COALESCING
| DEV_WRITE_COALESCING
);
4176 vol
->migr_state
= 0;
4177 set_migr_type(dev
, MIGR_INIT
);
4179 vol
->curr_migr_unit
= 0;
4180 map
= get_imsm_map(dev
, 0);
4181 map
->pba_of_lba0
= __cpu_to_le32(super
->create_offset
);
4182 map
->blocks_per_member
= __cpu_to_le32(info_to_blocks_per_member(info
));
4183 map
->blocks_per_strip
= __cpu_to_le16(info_to_blocks_per_strip(info
));
4184 map
->failed_disk_num
= ~0;
4185 map
->map_state
= info
->level
? IMSM_T_STATE_UNINITIALIZED
:
4186 IMSM_T_STATE_NORMAL
;
4189 if (info
->level
== 1 && info
->raid_disks
> 2) {
4192 fprintf(stderr
, Name
": imsm does not support more than 2 disks"
4193 "in a raid1 volume\n");
4197 map
->raid_level
= info
->level
;
4198 if (info
->level
== 10) {
4199 map
->raid_level
= 1;
4200 map
->num_domains
= info
->raid_disks
/ 2;
4201 } else if (info
->level
== 1)
4202 map
->num_domains
= info
->raid_disks
;
4204 map
->num_domains
= 1;
4206 num_data_stripes
= info_to_num_data_stripes(info
, map
->num_domains
);
4207 map
->num_data_stripes
= __cpu_to_le32(num_data_stripes
);
4209 map
->num_members
= info
->raid_disks
;
4210 for (i
= 0; i
< map
->num_members
; i
++) {
4211 /* initialized in add_to_super */
4212 set_imsm_ord_tbl_ent(map
, i
, IMSM_ORD_REBUILD
);
4214 mpb
->num_raid_devs
++;
4217 dv
->index
= super
->current_vol
;
4218 dv
->next
= super
->devlist
;
4219 super
->devlist
= dv
;
4221 imsm_update_version_info(super
);
4226 static int init_super_imsm(struct supertype
*st
, mdu_array_info_t
*info
,
4227 unsigned long long size
, char *name
,
4228 char *homehost
, int *uuid
)
4230 /* This is primarily called by Create when creating a new array.
4231 * We will then get add_to_super called for each component, and then
4232 * write_init_super called to write it out to each device.
4233 * For IMSM, Create can create on fresh devices or on a pre-existing
4235 * To create on a pre-existing array a different method will be called.
4236 * This one is just for fresh drives.
4238 struct intel_super
*super
;
4239 struct imsm_super
*mpb
;
4244 return init_super_imsm_volume(st
, info
, size
, name
, homehost
, uuid
);
4247 mpb_size
= disks_to_mpb_size(info
->nr_disks
);
4251 super
= alloc_super();
4252 if (super
&& posix_memalign(&super
->buf
, 512, mpb_size
) != 0) {
4257 fprintf(stderr
, Name
4258 ": %s could not allocate superblock\n", __func__
);
4261 if (posix_memalign(&super
->migr_rec_buf
, 512, 512) != 0) {
4262 fprintf(stderr
, Name
4263 ": %s could not allocate migr_rec buffer\n", __func__
);
4268 memset(super
->buf
, 0, mpb_size
);
4270 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4274 /* zeroing superblock */
4278 mpb
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4280 version
= (char *) mpb
->sig
;
4281 strcpy(version
, MPB_SIGNATURE
);
4282 version
+= strlen(MPB_SIGNATURE
);
4283 strcpy(version
, MPB_VERSION_RAID0
);
4289 static int add_to_super_imsm_volume(struct supertype
*st
, mdu_disk_info_t
*dk
,
4290 int fd
, char *devname
)
4292 struct intel_super
*super
= st
->sb
;
4293 struct imsm_super
*mpb
= super
->anchor
;
4295 struct imsm_dev
*dev
;
4296 struct imsm_map
*map
;
4299 dev
= get_imsm_dev(super
, super
->current_vol
);
4300 map
= get_imsm_map(dev
, 0);
4302 if (! (dk
->state
& (1<<MD_DISK_SYNC
))) {
4303 fprintf(stderr
, Name
": %s: Cannot add spare devices to IMSM volume\n",
4309 /* we're doing autolayout so grab the pre-marked (in
4310 * validate_geometry) raid_disk
4312 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4313 if (dl
->raiddisk
== dk
->raid_disk
)
4316 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
4317 if (dl
->major
== dk
->major
&&
4318 dl
->minor
== dk
->minor
)
4323 fprintf(stderr
, Name
": %s is not a member of the same container\n", devname
);
4327 /* add a pristine spare to the metadata */
4328 if (dl
->index
< 0) {
4329 dl
->index
= super
->anchor
->num_disks
;
4330 super
->anchor
->num_disks
++;
4332 /* Check the device has not already been added */
4333 slot
= get_imsm_disk_slot(map
, dl
->index
);
4335 (get_imsm_ord_tbl_ent(dev
, slot
, -1) & IMSM_ORD_REBUILD
) == 0) {
4336 fprintf(stderr
, Name
": %s has been included in this array twice\n",
4340 set_imsm_ord_tbl_ent(map
, dk
->raid_disk
, dl
->index
);
4341 dl
->disk
.status
= CONFIGURED_DISK
;
4343 /* if we are creating the first raid device update the family number */
4344 if (super
->current_vol
== 0) {
4346 struct imsm_dev
*_dev
= __get_imsm_dev(mpb
, 0);
4347 struct imsm_disk
*_disk
= __get_imsm_disk(mpb
, dl
->index
);
4349 if (!_dev
|| !_disk
) {
4350 fprintf(stderr
, Name
": BUG mpb setup error\n");
4356 sum
+= __gen_imsm_checksum(mpb
);
4357 mpb
->family_num
= __cpu_to_le32(sum
);
4358 mpb
->orig_family_num
= mpb
->family_num
;
4360 super
->current_disk
= dl
;
4365 static int add_to_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
,
4366 int fd
, char *devname
)
4368 struct intel_super
*super
= st
->sb
;
4370 unsigned long long size
;
4375 /* If we are on an RAID enabled platform check that the disk is
4376 * attached to the raid controller.
4377 * We do not need to test disks attachment for container based additions,
4378 * they shall be already tested when container was created/assembled.
4380 rv
= find_intel_hba_capability(fd
, super
, devname
);
4381 /* no orom/efi or non-intel hba of the disk */
4383 dprintf("capability: %p fd: %d ret: %d\n",
4384 super
->orom
, fd
, rv
);
4388 if (super
->current_vol
>= 0)
4389 return add_to_super_imsm_volume(st
, dk
, fd
, devname
);
4392 dd
= malloc(sizeof(*dd
));
4395 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4398 memset(dd
, 0, sizeof(*dd
));
4399 dd
->major
= major(stb
.st_rdev
);
4400 dd
->minor
= minor(stb
.st_rdev
);
4402 dd
->devname
= devname
? strdup(devname
) : NULL
;
4405 dd
->action
= DISK_ADD
;
4406 rv
= imsm_read_serial(fd
, devname
, dd
->serial
);
4409 Name
": failed to retrieve scsi serial, aborting\n");
4414 get_dev_size(fd
, NULL
, &size
);
4416 serialcpy(dd
->disk
.serial
, dd
->serial
);
4417 dd
->disk
.total_blocks
= __cpu_to_le32(size
);
4418 dd
->disk
.status
= SPARE_DISK
;
4419 if (sysfs_disk_to_scsi_id(fd
, &id
) == 0)
4420 dd
->disk
.scsi_id
= __cpu_to_le32(id
);
4422 dd
->disk
.scsi_id
= __cpu_to_le32(0);
4424 if (st
->update_tail
) {
4425 dd
->next
= super
->disk_mgmt_list
;
4426 super
->disk_mgmt_list
= dd
;
4428 dd
->next
= super
->disks
;
4430 super
->updates_pending
++;
4437 static int remove_from_super_imsm(struct supertype
*st
, mdu_disk_info_t
*dk
)
4439 struct intel_super
*super
= st
->sb
;
4442 /* remove from super works only in mdmon - for communication
4443 * manager - monitor. Check if communication memory buffer
4446 if (!st
->update_tail
) {
4448 Name
": %s shall be used in mdmon context only"
4449 "(line %d).\n", __func__
, __LINE__
);
4452 dd
= malloc(sizeof(*dd
));
4455 Name
": malloc failed %s:%d.\n", __func__
, __LINE__
);
4458 memset(dd
, 0, sizeof(*dd
));
4459 dd
->major
= dk
->major
;
4460 dd
->minor
= dk
->minor
;
4463 dd
->disk
.status
= SPARE_DISK
;
4464 dd
->action
= DISK_REMOVE
;
4466 dd
->next
= super
->disk_mgmt_list
;
4467 super
->disk_mgmt_list
= dd
;
4473 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
);
4477 struct imsm_super anchor
;
4478 } spare_record
__attribute__ ((aligned(512)));
4480 /* spare records have their own family number and do not have any defined raid
4483 static int write_super_imsm_spares(struct intel_super
*super
, int doclose
)
4485 struct imsm_super
*mpb
= super
->anchor
;
4486 struct imsm_super
*spare
= &spare_record
.anchor
;
4490 spare
->mpb_size
= __cpu_to_le32(sizeof(struct imsm_super
)),
4491 spare
->generation_num
= __cpu_to_le32(1UL),
4492 spare
->attributes
= MPB_ATTRIB_CHECKSUM_VERIFY
;
4493 spare
->num_disks
= 1,
4494 spare
->num_raid_devs
= 0,
4495 spare
->cache_size
= mpb
->cache_size
,
4496 spare
->pwr_cycle_count
= __cpu_to_le32(1),
4498 snprintf((char *) spare
->sig
, MAX_SIGNATURE_LENGTH
,
4499 MPB_SIGNATURE MPB_VERSION_RAID0
);
4501 for (d
= super
->disks
; d
; d
= d
->next
) {
4505 spare
->disk
[0] = d
->disk
;
4506 sum
= __gen_imsm_checksum(spare
);
4507 spare
->family_num
= __cpu_to_le32(sum
);
4508 spare
->orig_family_num
= 0;
4509 sum
= __gen_imsm_checksum(spare
);
4510 spare
->check_sum
= __cpu_to_le32(sum
);
4512 if (store_imsm_mpb(d
->fd
, spare
)) {
4513 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4514 __func__
, d
->major
, d
->minor
, strerror(errno
));
4526 static int write_super_imsm(struct supertype
*st
, int doclose
)
4528 struct intel_super
*super
= st
->sb
;
4529 struct imsm_super
*mpb
= super
->anchor
;
4535 __u32 mpb_size
= sizeof(struct imsm_super
) - sizeof(struct imsm_disk
);
4537 int clear_migration_record
= 1;
4539 /* 'generation' is incremented everytime the metadata is written */
4540 generation
= __le32_to_cpu(mpb
->generation_num
);
4542 mpb
->generation_num
= __cpu_to_le32(generation
);
4544 /* fix up cases where previous mdadm releases failed to set
4547 if (mpb
->orig_family_num
== 0)
4548 mpb
->orig_family_num
= mpb
->family_num
;
4550 for (d
= super
->disks
; d
; d
= d
->next
) {
4554 mpb
->disk
[d
->index
] = d
->disk
;
4558 for (d
= super
->missing
; d
; d
= d
->next
) {
4559 mpb
->disk
[d
->index
] = d
->disk
;
4562 mpb
->num_disks
= num_disks
;
4563 mpb_size
+= sizeof(struct imsm_disk
) * mpb
->num_disks
;
4565 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
4566 struct imsm_dev
*dev
= __get_imsm_dev(mpb
, i
);
4567 struct imsm_dev
*dev2
= get_imsm_dev(super
, i
);
4569 imsm_copy_dev(dev
, dev2
);
4570 mpb_size
+= sizeof_imsm_dev(dev
, 0);
4572 if (is_gen_migration(dev2
))
4573 clear_migration_record
= 0;
4575 mpb_size
+= __le32_to_cpu(mpb
->bbm_log_size
);
4576 mpb
->mpb_size
= __cpu_to_le32(mpb_size
);
4578 /* recalculate checksum */
4579 sum
= __gen_imsm_checksum(mpb
);
4580 mpb
->check_sum
= __cpu_to_le32(sum
);
4582 if (clear_migration_record
)
4583 memset(super
->migr_rec_buf
, 0, 512);
4585 /* write the mpb for disks that compose raid devices */
4586 for (d
= super
->disks
; d
; d
= d
->next
) {
4589 if (store_imsm_mpb(d
->fd
, mpb
))
4590 fprintf(stderr
, "%s: failed for device %d:%d %s\n",
4591 __func__
, d
->major
, d
->minor
, strerror(errno
));
4592 if (clear_migration_record
) {
4593 unsigned long long dsize
;
4595 get_dev_size(d
->fd
, NULL
, &dsize
);
4596 if (lseek64(d
->fd
, dsize
- 512, SEEK_SET
) >= 0) {
4597 if (write(d
->fd
, super
->migr_rec_buf
, 512) != 512)
4598 perror("Write migr_rec failed");
4608 return write_super_imsm_spares(super
, doclose
);
4614 static int create_array(struct supertype
*st
, int dev_idx
)
4617 struct imsm_update_create_array
*u
;
4618 struct intel_super
*super
= st
->sb
;
4619 struct imsm_dev
*dev
= get_imsm_dev(super
, dev_idx
);
4620 struct imsm_map
*map
= get_imsm_map(dev
, 0);
4621 struct disk_info
*inf
;
4622 struct imsm_disk
*disk
;
4625 len
= sizeof(*u
) - sizeof(*dev
) + sizeof_imsm_dev(dev
, 0) +
4626 sizeof(*inf
) * map
->num_members
;
4629 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4634 u
->type
= update_create_array
;
4635 u
->dev_idx
= dev_idx
;
4636 imsm_copy_dev(&u
->dev
, dev
);
4637 inf
= get_disk_info(u
);
4638 for (i
= 0; i
< map
->num_members
; i
++) {
4639 int idx
= get_imsm_disk_idx(dev
, i
, -1);
4641 disk
= get_imsm_disk(super
, idx
);
4642 serialcpy(inf
[i
].serial
, disk
->serial
);
4644 append_metadata_update(st
, u
, len
);
4649 static int mgmt_disk(struct supertype
*st
)
4651 struct intel_super
*super
= st
->sb
;
4653 struct imsm_update_add_remove_disk
*u
;
4655 if (!super
->disk_mgmt_list
)
4661 fprintf(stderr
, "%s: failed to allocate update buffer\n",
4666 u
->type
= update_add_remove_disk
;
4667 append_metadata_update(st
, u
, len
);
4672 static int write_init_super_imsm(struct supertype
*st
)
4674 struct intel_super
*super
= st
->sb
;
4675 int current_vol
= super
->current_vol
;
4677 /* we are done with current_vol reset it to point st at the container */
4678 super
->current_vol
= -1;
4680 if (st
->update_tail
) {
4681 /* queue the recently created array / added disk
4682 * as a metadata update */
4685 /* determine if we are creating a volume or adding a disk */
4686 if (current_vol
< 0) {
4687 /* in the mgmt (add/remove) disk case we are running
4688 * in mdmon context, so don't close fd's
4690 return mgmt_disk(st
);
4692 rv
= create_array(st
, current_vol
);
4697 for (d
= super
->disks
; d
; d
= d
->next
)
4698 Kill(d
->devname
, NULL
, 0, 1, 1);
4699 return write_super_imsm(st
, 1);
4704 static int store_super_imsm(struct supertype
*st
, int fd
)
4706 struct intel_super
*super
= st
->sb
;
4707 struct imsm_super
*mpb
= super
? super
->anchor
: NULL
;
4713 return store_imsm_mpb(fd
, mpb
);
4719 static int imsm_bbm_log_size(struct imsm_super
*mpb
)
4721 return __le32_to_cpu(mpb
->bbm_log_size
);
4725 static int validate_geometry_imsm_container(struct supertype
*st
, int level
,
4726 int layout
, int raiddisks
, int chunk
,
4727 unsigned long long size
, char *dev
,
4728 unsigned long long *freesize
,
4732 unsigned long long ldsize
;
4733 struct intel_super
*super
=NULL
;
4736 if (level
!= LEVEL_CONTAINER
)
4741 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
4744 fprintf(stderr
, Name
": imsm: Cannot open %s: %s\n",
4745 dev
, strerror(errno
));
4748 if (!get_dev_size(fd
, dev
, &ldsize
)) {
4753 /* capabilities retrieve could be possible
4754 * note that there is no fd for the disks in array.
4756 super
= alloc_super();
4759 Name
": malloc of %zu failed.\n",
4765 rv
= find_intel_hba_capability(fd
, super
, verbose
? dev
: NULL
);
4769 fd2devname(fd
, str
);
4770 dprintf("validate_geometry_imsm_container: fd: %d %s orom: %p rv: %d raiddisk: %d\n",
4771 fd
, str
, super
->orom
, rv
, raiddisks
);
4773 /* no orom/efi or non-intel hba of the disk */
4779 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4781 fprintf(stderr
, Name
": %d exceeds maximum number of"
4782 " platform supported disks: %d\n",
4783 raiddisks
, super
->orom
->tds
);
4789 *freesize
= avail_size_imsm(st
, ldsize
>> 9);
4795 static unsigned long long find_size(struct extent
*e
, int *idx
, int num_extents
)
4797 const unsigned long long base_start
= e
[*idx
].start
;
4798 unsigned long long end
= base_start
+ e
[*idx
].size
;
4801 if (base_start
== end
)
4805 for (i
= *idx
; i
< num_extents
; i
++) {
4806 /* extend overlapping extents */
4807 if (e
[i
].start
>= base_start
&&
4808 e
[i
].start
<= end
) {
4811 if (e
[i
].start
+ e
[i
].size
> end
)
4812 end
= e
[i
].start
+ e
[i
].size
;
4813 } else if (e
[i
].start
> end
) {
4819 return end
- base_start
;
4822 static unsigned long long merge_extents(struct intel_super
*super
, int sum_extents
)
4824 /* build a composite disk with all known extents and generate a new
4825 * 'maxsize' given the "all disks in an array must share a common start
4826 * offset" constraint
4828 struct extent
*e
= calloc(sum_extents
, sizeof(*e
));
4832 unsigned long long pos
;
4833 unsigned long long start
= 0;
4834 unsigned long long maxsize
;
4835 unsigned long reserve
;
4840 /* coalesce and sort all extents. also, check to see if we need to
4841 * reserve space between member arrays
4844 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
4847 for (i
= 0; i
< dl
->extent_cnt
; i
++)
4850 qsort(e
, sum_extents
, sizeof(*e
), cmp_extent
);
4855 while (i
< sum_extents
) {
4856 e
[j
].start
= e
[i
].start
;
4857 e
[j
].size
= find_size(e
, &i
, sum_extents
);
4859 if (e
[j
-1].size
== 0)
4868 unsigned long long esize
;
4870 esize
= e
[i
].start
- pos
;
4871 if (esize
>= maxsize
) {
4876 pos
= e
[i
].start
+ e
[i
].size
;
4878 } while (e
[i
-1].size
);
4884 /* FIXME assumes volume at offset 0 is the first volume in a
4887 if (start_extent
> 0)
4888 reserve
= IMSM_RESERVED_SECTORS
; /* gap between raid regions */
4892 if (maxsize
< reserve
)
4895 super
->create_offset
= ~((__u32
) 0);
4896 if (start
+ reserve
> super
->create_offset
)
4897 return 0; /* start overflows create_offset */
4898 super
->create_offset
= start
+ reserve
;
4900 return maxsize
- reserve
;
4903 static int is_raid_level_supported(const struct imsm_orom
*orom
, int level
, int raiddisks
)
4905 if (level
< 0 || level
== 6 || level
== 4)
4908 /* if we have an orom prevent invalid raid levels */
4911 case 0: return imsm_orom_has_raid0(orom
);
4914 return imsm_orom_has_raid1e(orom
);
4915 return imsm_orom_has_raid1(orom
) && raiddisks
== 2;
4916 case 10: return imsm_orom_has_raid10(orom
) && raiddisks
== 4;
4917 case 5: return imsm_orom_has_raid5(orom
) && raiddisks
> 2;
4920 return 1; /* not on an Intel RAID platform so anything goes */
4925 static int imsm_default_chunk(const struct imsm_orom
*orom
)
4927 /* up to 512 if the plaform supports it, otherwise the platform max.
4928 * 128 if no platform detected
4930 int fs
= max(7, orom
? fls(orom
->sss
) : 0);
4932 return min(512, (1 << fs
));
4935 #define pr_vrb(fmt, arg...) (void) (verbose && fprintf(stderr, Name fmt, ##arg))
4937 * validate volume parameters with OROM/EFI capabilities
4940 validate_geometry_imsm_orom(struct intel_super
*super
, int level
, int layout
,
4941 int raiddisks
, int *chunk
, int verbose
)
4946 /* validate container capabilities */
4947 if (super
->orom
&& raiddisks
> super
->orom
->tds
) {
4949 fprintf(stderr
, Name
": %d exceeds maximum number of"
4950 " platform supported disks: %d\n",
4951 raiddisks
, super
->orom
->tds
);
4955 /* capabilities of OROM tested - copied from validate_geometry_imsm_volume */
4956 if (super
->orom
&& (!is_raid_level_supported(super
->orom
, level
,
4958 pr_vrb(": platform does not support raid%d with %d disk%s\n",
4959 level
, raiddisks
, raiddisks
> 1 ? "s" : "");
4963 if (chunk
&& (*chunk
== 0 || *chunk
== UnSet
))
4964 *chunk
= imsm_default_chunk(super
->orom
);
4966 if (super
->orom
&& chunk
&& !imsm_orom_has_chunk(super
->orom
, *chunk
)) {
4967 pr_vrb(": platform does not support a chunk size of: "
4972 if (layout
!= imsm_level_to_layout(level
)) {
4974 pr_vrb(": imsm raid 5 only supports the left-asymmetric layout\n");
4975 else if (level
== 10)
4976 pr_vrb(": imsm raid 10 only supports the n2 layout\n");
4978 pr_vrb(": imsm unknown layout %#x for this raid level %d\n",
4985 /* validate_geometry_imsm_volume - lifted from validate_geometry_ddf_bvd
4986 * FIX ME add ahci details
4988 static int validate_geometry_imsm_volume(struct supertype
*st
, int level
,
4989 int layout
, int raiddisks
, int *chunk
,
4990 unsigned long long size
, char *dev
,
4991 unsigned long long *freesize
,
4995 struct intel_super
*super
= st
->sb
;
4996 struct imsm_super
*mpb
= super
->anchor
;
4998 unsigned long long pos
= 0;
4999 unsigned long long maxsize
;
5003 /* We must have the container info already read in. */
5007 if (!validate_geometry_imsm_orom(super
, level
, layout
, raiddisks
, chunk
, verbose
)) {
5008 fprintf(stderr
, Name
": RAID gemetry validation failed. "
5009 "Cannot proceed with the action(s).\n");
5013 /* General test: make sure there is space for
5014 * 'raiddisks' device extents of size 'size' at a given
5017 unsigned long long minsize
= size
;
5018 unsigned long long start_offset
= MaxSector
;
5021 minsize
= MPB_SECTOR_CNT
+ IMSM_RESERVED_SECTORS
;
5022 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5027 e
= get_extents(super
, dl
);
5030 unsigned long long esize
;
5031 esize
= e
[i
].start
- pos
;
5032 if (esize
>= minsize
)
5034 if (found
&& start_offset
== MaxSector
) {
5037 } else if (found
&& pos
!= start_offset
) {
5041 pos
= e
[i
].start
+ e
[i
].size
;
5043 } while (e
[i
-1].size
);
5048 if (dcnt
< raiddisks
) {
5050 fprintf(stderr
, Name
": imsm: Not enough "
5051 "devices with space for this array "
5059 /* This device must be a member of the set */
5060 if (stat(dev
, &stb
) < 0)
5062 if ((S_IFMT
& stb
.st_mode
) != S_IFBLK
)
5064 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5065 if (dl
->major
== (int)major(stb
.st_rdev
) &&
5066 dl
->minor
== (int)minor(stb
.st_rdev
))
5071 fprintf(stderr
, Name
": %s is not in the "
5072 "same imsm set\n", dev
);
5074 } else if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
) {
5075 /* If a volume is present then the current creation attempt
5076 * cannot incorporate new spares because the orom may not
5077 * understand this configuration (all member disks must be
5078 * members of each array in the container).
5080 fprintf(stderr
, Name
": %s is a spare and a volume"
5081 " is already defined for this container\n", dev
);
5082 fprintf(stderr
, Name
": The option-rom requires all member"
5083 " disks to be a member of all volumes\n");
5087 /* retrieve the largest free space block */
5088 e
= get_extents(super
, dl
);
5093 unsigned long long esize
;
5095 esize
= e
[i
].start
- pos
;
5096 if (esize
>= maxsize
)
5098 pos
= e
[i
].start
+ e
[i
].size
;
5100 } while (e
[i
-1].size
);
5105 fprintf(stderr
, Name
": unable to determine free space for: %s\n",
5109 if (maxsize
< size
) {
5111 fprintf(stderr
, Name
": %s not enough space (%llu < %llu)\n",
5112 dev
, maxsize
, size
);
5116 /* count total number of extents for merge */
5118 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5120 i
+= dl
->extent_cnt
;
5122 maxsize
= merge_extents(super
, i
);
5123 if (maxsize
< size
|| maxsize
== 0) {
5125 fprintf(stderr
, Name
": not enough space after merge (%llu < %llu)\n",
5130 *freesize
= maxsize
;
5135 static int reserve_space(struct supertype
*st
, int raiddisks
,
5136 unsigned long long size
, int chunk
,
5137 unsigned long long *freesize
)
5139 struct intel_super
*super
= st
->sb
;
5140 struct imsm_super
*mpb
= super
->anchor
;
5145 unsigned long long maxsize
;
5146 unsigned long long minsize
;
5150 /* find the largest common start free region of the possible disks */
5154 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
5160 /* don't activate new spares if we are orom constrained
5161 * and there is already a volume active in the container
5163 if (super
->orom
&& dl
->index
< 0 && mpb
->num_raid_devs
)
5166 e
= get_extents(super
, dl
);
5169 for (i
= 1; e
[i
-1].size
; i
++)
5177 maxsize
= merge_extents(super
, extent_cnt
);
5181 minsize
= chunk
* 2;
5183 if (cnt
< raiddisks
||
5184 (super
->orom
&& used
&& used
!= raiddisks
) ||
5185 maxsize
< minsize
||
5187 fprintf(stderr
, Name
": not enough devices with space to create array.\n");
5188 return 0; /* No enough free spaces large enough */
5200 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
5202 dl
->raiddisk
= cnt
++;
5209 static int validate_geometry_imsm(struct supertype
*st
, int level
, int layout
,
5210 int raiddisks
, int *chunk
, unsigned long long size
,
5211 char *dev
, unsigned long long *freesize
,
5219 * if given unused devices create a container
5220 * if given given devices in a container create a member volume
5222 if (level
== LEVEL_CONTAINER
) {
5223 /* Must be a fresh device to add to a container */
5224 return validate_geometry_imsm_container(st
, level
, layout
,
5226 chunk
?*chunk
:0, size
,
5232 if (st
->sb
&& freesize
) {
5233 /* we are being asked to automatically layout a
5234 * new volume based on the current contents of
5235 * the container. If the the parameters can be
5236 * satisfied reserve_space will record the disks,
5237 * start offset, and size of the volume to be
5238 * created. add_to_super and getinfo_super
5239 * detect when autolayout is in progress.
5241 if (!validate_geometry_imsm_orom(st
->sb
, level
, layout
,
5245 return reserve_space(st
, raiddisks
, size
,
5246 chunk
?*chunk
:0, freesize
);
5251 /* creating in a given container */
5252 return validate_geometry_imsm_volume(st
, level
, layout
,
5253 raiddisks
, chunk
, size
,
5254 dev
, freesize
, verbose
);
5257 /* This device needs to be a device in an 'imsm' container */
5258 fd
= open(dev
, O_RDONLY
|O_EXCL
, 0);
5262 Name
": Cannot create this array on device %s\n",
5267 if (errno
!= EBUSY
|| (fd
= open(dev
, O_RDONLY
, 0)) < 0) {
5269 fprintf(stderr
, Name
": Cannot open %s: %s\n",
5270 dev
, strerror(errno
));
5273 /* Well, it is in use by someone, maybe an 'imsm' container. */
5274 cfd
= open_container(fd
);
5278 fprintf(stderr
, Name
": Cannot use %s: It is busy\n",
5282 sra
= sysfs_read(cfd
, 0, GET_VERSION
);
5283 if (sra
&& sra
->array
.major_version
== -1 &&
5284 strcmp(sra
->text_version
, "imsm") == 0)
5288 /* This is a member of a imsm container. Load the container
5289 * and try to create a volume
5291 struct intel_super
*super
;
5293 if (load_super_imsm_all(st
, cfd
, (void **) &super
, NULL
) == 0) {
5295 st
->container_dev
= fd2devnum(cfd
);
5297 return validate_geometry_imsm_volume(st
, level
, layout
,
5305 fprintf(stderr
, Name
": failed container membership check\n");
5311 static void default_geometry_imsm(struct supertype
*st
, int *level
, int *layout
, int *chunk
)
5313 struct intel_super
*super
= st
->sb
;
5315 if (level
&& *level
== UnSet
)
5316 *level
= LEVEL_CONTAINER
;
5318 if (level
&& layout
&& *layout
== UnSet
)
5319 *layout
= imsm_level_to_layout(*level
);
5321 if (chunk
&& (*chunk
== UnSet
|| *chunk
== 0))
5322 *chunk
= imsm_default_chunk(super
->orom
);
5325 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
);
5327 static int kill_subarray_imsm(struct supertype
*st
)
5329 /* remove the subarray currently referenced by ->current_vol */
5331 struct intel_dev
**dp
;
5332 struct intel_super
*super
= st
->sb
;
5333 __u8 current_vol
= super
->current_vol
;
5334 struct imsm_super
*mpb
= super
->anchor
;
5336 if (super
->current_vol
< 0)
5338 super
->current_vol
= -1; /* invalidate subarray cursor */
5340 /* block deletions that would change the uuid of active subarrays
5342 * FIXME when immutable ids are available, but note that we'll
5343 * also need to fixup the invalidated/active subarray indexes in
5346 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5349 if (i
< current_vol
)
5351 sprintf(subarray
, "%u", i
);
5352 if (is_subarray_active(subarray
, st
->devname
)) {
5354 Name
": deleting subarray-%d would change the UUID of active subarray-%d, aborting\n",
5361 if (st
->update_tail
) {
5362 struct imsm_update_kill_array
*u
= malloc(sizeof(*u
));
5366 u
->type
= update_kill_array
;
5367 u
->dev_idx
= current_vol
;
5368 append_metadata_update(st
, u
, sizeof(*u
));
5373 for (dp
= &super
->devlist
; *dp
;)
5374 if ((*dp
)->index
== current_vol
) {
5377 handle_missing(super
, (*dp
)->dev
);
5378 if ((*dp
)->index
> current_vol
)
5383 /* no more raid devices, all active components are now spares,
5384 * but of course failed are still failed
5386 if (--mpb
->num_raid_devs
== 0) {
5389 for (d
= super
->disks
; d
; d
= d
->next
)
5390 if (d
->index
> -2) {
5392 d
->disk
.status
= SPARE_DISK
;
5396 super
->updates_pending
++;
5401 static int update_subarray_imsm(struct supertype
*st
, char *subarray
,
5402 char *update
, struct mddev_ident
*ident
)
5404 /* update the subarray currently referenced by ->current_vol */
5405 struct intel_super
*super
= st
->sb
;
5406 struct imsm_super
*mpb
= super
->anchor
;
5408 if (strcmp(update
, "name") == 0) {
5409 char *name
= ident
->name
;
5413 if (is_subarray_active(subarray
, st
->devname
)) {
5415 Name
": Unable to update name of active subarray\n");
5419 if (!check_name(super
, name
, 0))
5422 vol
= strtoul(subarray
, &ep
, 10);
5423 if (*ep
!= '\0' || vol
>= super
->anchor
->num_raid_devs
)
5426 if (st
->update_tail
) {
5427 struct imsm_update_rename_array
*u
= malloc(sizeof(*u
));
5431 u
->type
= update_rename_array
;
5433 snprintf((char *) u
->name
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5434 append_metadata_update(st
, u
, sizeof(*u
));
5436 struct imsm_dev
*dev
;
5439 dev
= get_imsm_dev(super
, vol
);
5440 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
5441 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5442 dev
= get_imsm_dev(super
, i
);
5443 handle_missing(super
, dev
);
5445 super
->updates_pending
++;
5453 static int is_gen_migration(struct imsm_dev
*dev
)
5458 if (!dev
->vol
.migr_state
)
5461 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5466 #endif /* MDASSEMBLE */
5468 static int is_rebuilding(struct imsm_dev
*dev
)
5470 struct imsm_map
*migr_map
;
5472 if (!dev
->vol
.migr_state
)
5475 if (migr_type(dev
) != MIGR_REBUILD
)
5478 migr_map
= get_imsm_map(dev
, 1);
5480 if (migr_map
->map_state
== IMSM_T_STATE_DEGRADED
)
5486 static void update_recovery_start(struct intel_super
*super
,
5487 struct imsm_dev
*dev
,
5488 struct mdinfo
*array
)
5490 struct mdinfo
*rebuild
= NULL
;
5494 if (!is_rebuilding(dev
))
5497 /* Find the rebuild target, but punt on the dual rebuild case */
5498 for (d
= array
->devs
; d
; d
= d
->next
)
5499 if (d
->recovery_start
== 0) {
5506 /* (?) none of the disks are marked with
5507 * IMSM_ORD_REBUILD, so assume they are missing and the
5508 * disk_ord_tbl was not correctly updated
5510 dprintf("%s: failed to locate out-of-sync disk\n", __func__
);
5514 units
= __le32_to_cpu(dev
->vol
.curr_migr_unit
);
5515 rebuild
->recovery_start
= units
* blocks_per_migr_unit(super
, dev
);
5519 static int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
);
5522 static struct mdinfo
*container_content_imsm(struct supertype
*st
, char *subarray
)
5524 /* Given a container loaded by load_super_imsm_all,
5525 * extract information about all the arrays into
5527 * If 'subarray' is given, just extract info about that array.
5529 * For each imsm_dev create an mdinfo, fill it in,
5530 * then look for matching devices in super->disks
5531 * and create appropriate device mdinfo.
5533 struct intel_super
*super
= st
->sb
;
5534 struct imsm_super
*mpb
= super
->anchor
;
5535 struct mdinfo
*rest
= NULL
;
5539 int spare_disks
= 0;
5541 /* do not assemble arrays when not all attributes are supported */
5542 if (imsm_check_attributes(mpb
->attributes
) == 0) {
5543 fprintf(stderr
, Name
": IMSM metadata loading not allowed "
5544 "due to attributes incompatibility.\n");
5548 /* check for bad blocks */
5549 if (imsm_bbm_log_size(super
->anchor
))
5552 /* count spare devices, not used in maps
5554 for (d
= super
->disks
; d
; d
= d
->next
)
5558 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5559 struct imsm_dev
*dev
;
5560 struct imsm_map
*map
;
5561 struct imsm_map
*map2
;
5562 struct mdinfo
*this;
5567 (i
!= strtoul(subarray
, &ep
, 10) || *ep
!= '\0'))
5570 dev
= get_imsm_dev(super
, i
);
5571 map
= get_imsm_map(dev
, 0);
5572 map2
= get_imsm_map(dev
, 1);
5574 /* do not publish arrays that are in the middle of an
5575 * unsupported migration
5577 if (dev
->vol
.migr_state
&&
5578 (migr_type(dev
) == MIGR_STATE_CHANGE
)) {
5579 fprintf(stderr
, Name
": cannot assemble volume '%.16s':"
5580 " unsupported migration in progress\n",
5584 /* do not publish arrays that are not support by controller's
5588 chunk
= __le16_to_cpu(map
->blocks_per_strip
) >> 1;
5590 if (!validate_geometry_imsm_orom(super
,
5591 get_imsm_raid_level(map
), /* RAID level */
5592 imsm_level_to_layout(get_imsm_raid_level(map
)),
5593 map
->num_members
, /* raid disks */
5596 fprintf(stderr
, Name
": RAID gemetry validation failed. "
5597 "Cannot proceed with the action(s).\n");
5600 #endif /* MDASSEMBLE */
5601 this = malloc(sizeof(*this));
5603 fprintf(stderr
, Name
": failed to allocate %zu bytes\n",
5608 super
->current_vol
= i
;
5609 getinfo_super_imsm_volume(st
, this, NULL
);
5611 for (slot
= 0 ; slot
< map
->num_members
; slot
++) {
5612 unsigned long long recovery_start
;
5613 struct mdinfo
*info_d
;
5620 idx
= get_imsm_disk_idx(dev
, slot
, 0);
5621 ord
= get_imsm_ord_tbl_ent(dev
, slot
, -1);
5622 for (d
= super
->disks
; d
; d
= d
->next
)
5623 if (d
->index
== idx
)
5626 recovery_start
= MaxSector
;
5629 if (d
&& is_failed(&d
->disk
))
5631 if (ord
& IMSM_ORD_REBUILD
)
5635 * if we skip some disks the array will be assmebled degraded;
5636 * reset resync start to avoid a dirty-degraded
5637 * situation when performing the intial sync
5639 * FIXME handle dirty degraded
5641 if ((skip
|| recovery_start
== 0) && !dev
->vol
.dirty
)
5642 this->resync_start
= MaxSector
;
5646 info_d
= calloc(1, sizeof(*info_d
));
5648 fprintf(stderr
, Name
": failed to allocate disk"
5649 " for volume %.16s\n", dev
->volume
);
5650 info_d
= this->devs
;
5652 struct mdinfo
*d
= info_d
->next
;
5661 info_d
->next
= this->devs
;
5662 this->devs
= info_d
;
5664 info_d
->disk
.number
= d
->index
;
5665 info_d
->disk
.major
= d
->major
;
5666 info_d
->disk
.minor
= d
->minor
;
5667 info_d
->disk
.raid_disk
= slot
;
5668 info_d
->recovery_start
= recovery_start
;
5670 if (slot
< map2
->num_members
)
5671 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5673 this->array
.spare_disks
++;
5675 if (slot
< map
->num_members
)
5676 info_d
->disk
.state
= (1 << MD_DISK_ACTIVE
);
5678 this->array
.spare_disks
++;
5680 if (info_d
->recovery_start
== MaxSector
)
5681 this->array
.working_disks
++;
5683 info_d
->events
= __le32_to_cpu(mpb
->generation_num
);
5684 info_d
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
5685 info_d
->component_size
= __le32_to_cpu(map
->blocks_per_member
);
5687 /* now that the disk list is up-to-date fixup recovery_start */
5688 update_recovery_start(super
, dev
, this);
5689 this->array
.spare_disks
+= spare_disks
;
5692 /* check for reshape */
5693 if (this->reshape_active
== 1)
5694 recover_backup_imsm(st
, this);
5699 /* if array has bad blocks, set suitable bit in array status */
5701 rest
->array
.state
|= (1<<MD_SB_BBM_ERRORS
);
5707 static __u8
imsm_check_degraded(struct intel_super
*super
, struct imsm_dev
*dev
, int failed
)
5709 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5712 return map
->map_state
== IMSM_T_STATE_UNINITIALIZED
?
5713 IMSM_T_STATE_UNINITIALIZED
: IMSM_T_STATE_NORMAL
;
5715 switch (get_imsm_raid_level(map
)) {
5717 return IMSM_T_STATE_FAILED
;
5720 if (failed
< map
->num_members
)
5721 return IMSM_T_STATE_DEGRADED
;
5723 return IMSM_T_STATE_FAILED
;
5728 * check to see if any mirrors have failed, otherwise we
5729 * are degraded. Even numbered slots are mirrored on
5733 /* gcc -Os complains that this is unused */
5734 int insync
= insync
;
5736 for (i
= 0; i
< map
->num_members
; i
++) {
5737 __u32 ord
= get_imsm_ord_tbl_ent(dev
, i
, -1);
5738 int idx
= ord_to_idx(ord
);
5739 struct imsm_disk
*disk
;
5741 /* reset the potential in-sync count on even-numbered
5742 * slots. num_copies is always 2 for imsm raid10
5747 disk
= get_imsm_disk(super
, idx
);
5748 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5751 /* no in-sync disks left in this mirror the
5755 return IMSM_T_STATE_FAILED
;
5758 return IMSM_T_STATE_DEGRADED
;
5762 return IMSM_T_STATE_DEGRADED
;
5764 return IMSM_T_STATE_FAILED
;
5770 return map
->map_state
;
5773 static int imsm_count_failed(struct intel_super
*super
, struct imsm_dev
*dev
)
5777 struct imsm_disk
*disk
;
5778 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5779 struct imsm_map
*prev
= get_imsm_map(dev
, dev
->vol
.migr_state
);
5783 /* at the beginning of migration we set IMSM_ORD_REBUILD on
5784 * disks that are being rebuilt. New failures are recorded to
5785 * map[0]. So we look through all the disks we started with and
5786 * see if any failures are still present, or if any new ones
5789 * FIXME add support for online capacity expansion and
5790 * raid-level-migration
5792 for (i
= 0; i
< prev
->num_members
; i
++) {
5793 ord
= __le32_to_cpu(prev
->disk_ord_tbl
[i
]);
5794 ord
|= __le32_to_cpu(map
->disk_ord_tbl
[i
]);
5795 idx
= ord_to_idx(ord
);
5797 disk
= get_imsm_disk(super
, idx
);
5798 if (!disk
|| is_failed(disk
) || ord
& IMSM_ORD_REBUILD
)
5806 static int imsm_open_new(struct supertype
*c
, struct active_array
*a
,
5809 struct intel_super
*super
= c
->sb
;
5810 struct imsm_super
*mpb
= super
->anchor
;
5812 if (atoi(inst
) >= mpb
->num_raid_devs
) {
5813 fprintf(stderr
, "%s: subarry index %d, out of range\n",
5814 __func__
, atoi(inst
));
5818 dprintf("imsm: open_new %s\n", inst
);
5819 a
->info
.container_member
= atoi(inst
);
5823 static int is_resyncing(struct imsm_dev
*dev
)
5825 struct imsm_map
*migr_map
;
5827 if (!dev
->vol
.migr_state
)
5830 if (migr_type(dev
) == MIGR_INIT
||
5831 migr_type(dev
) == MIGR_REPAIR
)
5834 if (migr_type(dev
) == MIGR_GEN_MIGR
)
5837 migr_map
= get_imsm_map(dev
, 1);
5839 if ((migr_map
->map_state
== IMSM_T_STATE_NORMAL
) &&
5840 (dev
->vol
.migr_type
!= MIGR_GEN_MIGR
))
5846 /* return true if we recorded new information */
5847 static int mark_failure(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5851 struct imsm_map
*map
;
5853 /* new failures are always set in map[0] */
5854 map
= get_imsm_map(dev
, 0);
5856 slot
= get_imsm_disk_slot(map
, idx
);
5860 ord
= __le32_to_cpu(map
->disk_ord_tbl
[slot
]);
5861 if (is_failed(disk
) && (ord
& IMSM_ORD_REBUILD
))
5864 disk
->status
|= FAILED_DISK
;
5865 set_imsm_ord_tbl_ent(map
, slot
, idx
| IMSM_ORD_REBUILD
);
5866 if (map
->failed_disk_num
== 0xff)
5867 map
->failed_disk_num
= slot
;
5871 static void mark_missing(struct imsm_dev
*dev
, struct imsm_disk
*disk
, int idx
)
5873 mark_failure(dev
, disk
, idx
);
5875 if (disk
->scsi_id
== __cpu_to_le32(~(__u32
)0))
5878 disk
->scsi_id
= __cpu_to_le32(~(__u32
)0);
5879 memmove(&disk
->serial
[0], &disk
->serial
[1], MAX_RAID_SERIAL_LEN
- 1);
5882 static void handle_missing(struct intel_super
*super
, struct imsm_dev
*dev
)
5888 if (!super
->missing
)
5890 failed
= imsm_count_failed(super
, dev
);
5891 map_state
= imsm_check_degraded(super
, dev
, failed
);
5893 dprintf("imsm: mark missing\n");
5894 end_migration(dev
, map_state
);
5895 for (dl
= super
->missing
; dl
; dl
= dl
->next
)
5896 mark_missing(dev
, &dl
->disk
, dl
->index
);
5897 super
->updates_pending
++;
5900 static unsigned long long imsm_set_array_size(struct imsm_dev
*dev
)
5902 int used_disks
= imsm_num_data_members(dev
, 0);
5903 unsigned long long array_blocks
;
5904 struct imsm_map
*map
;
5906 if (used_disks
== 0) {
5907 /* when problems occures
5908 * return current array_blocks value
5910 array_blocks
= __le32_to_cpu(dev
->size_high
);
5911 array_blocks
= array_blocks
<< 32;
5912 array_blocks
+= __le32_to_cpu(dev
->size_low
);
5914 return array_blocks
;
5917 /* set array size in metadata
5919 map
= get_imsm_map(dev
, 0);
5920 array_blocks
= map
->blocks_per_member
* used_disks
;
5922 /* round array size down to closest MB
5924 array_blocks
= (array_blocks
>> SECT_PER_MB_SHIFT
) << SECT_PER_MB_SHIFT
;
5925 dev
->size_low
= __cpu_to_le32((__u32
)array_blocks
);
5926 dev
->size_high
= __cpu_to_le32((__u32
)(array_blocks
>> 32));
5928 return array_blocks
;
5931 static void imsm_set_disk(struct active_array
*a
, int n
, int state
);
5933 static void imsm_progress_container_reshape(struct intel_super
*super
)
5935 /* if no device has a migr_state, but some device has a
5936 * different number of members than the previous device, start
5937 * changing the number of devices in this device to match
5940 struct imsm_super
*mpb
= super
->anchor
;
5941 int prev_disks
= -1;
5945 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
5946 struct imsm_dev
*dev
= get_imsm_dev(super
, i
);
5947 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5948 struct imsm_map
*map2
;
5949 int prev_num_members
;
5951 if (dev
->vol
.migr_state
)
5954 if (prev_disks
== -1)
5955 prev_disks
= map
->num_members
;
5956 if (prev_disks
== map
->num_members
)
5959 /* OK, this array needs to enter reshape mode.
5960 * i.e it needs a migr_state
5963 copy_map_size
= sizeof_imsm_map(map
);
5964 prev_num_members
= map
->num_members
;
5965 map
->num_members
= prev_disks
;
5966 dev
->vol
.migr_state
= 1;
5967 dev
->vol
.curr_migr_unit
= 0;
5968 dev
->vol
.migr_type
= MIGR_GEN_MIGR
;
5969 for (i
= prev_num_members
;
5970 i
< map
->num_members
; i
++)
5971 set_imsm_ord_tbl_ent(map
, i
, i
);
5972 map2
= get_imsm_map(dev
, 1);
5973 /* Copy the current map */
5974 memcpy(map2
, map
, copy_map_size
);
5975 map2
->num_members
= prev_num_members
;
5977 imsm_set_array_size(dev
);
5978 super
->updates_pending
++;
5982 /* Handle dirty -> clean transititions, resync and reshape. Degraded and rebuild
5983 * states are handled in imsm_set_disk() with one exception, when a
5984 * resync is stopped due to a new failure this routine will set the
5985 * 'degraded' state for the array.
5987 static int imsm_set_array_state(struct active_array
*a
, int consistent
)
5989 int inst
= a
->info
.container_member
;
5990 struct intel_super
*super
= a
->container
->sb
;
5991 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
5992 struct imsm_map
*map
= get_imsm_map(dev
, 0);
5993 int failed
= imsm_count_failed(super
, dev
);
5994 __u8 map_state
= imsm_check_degraded(super
, dev
, failed
);
5995 __u32 blocks_per_unit
;
5997 if (dev
->vol
.migr_state
&&
5998 dev
->vol
.migr_type
== MIGR_GEN_MIGR
) {
5999 /* array state change is blocked due to reshape action
6001 * - abort the reshape (if last_checkpoint is 0 and action!= reshape)
6002 * - finish the reshape (if last_checkpoint is big and action != reshape)
6003 * - update curr_migr_unit
6005 if (a
->curr_action
== reshape
) {
6006 /* still reshaping, maybe update curr_migr_unit */
6007 goto mark_checkpoint
;
6009 if (a
->last_checkpoint
== 0 && a
->prev_action
== reshape
) {
6010 /* for some reason we aborted the reshape.
6012 * disable automatic metadata rollback
6013 * user action is required to recover process
6016 struct imsm_map
*map2
= get_imsm_map(dev
, 1);
6017 dev
->vol
.migr_state
= 0;
6018 dev
->vol
.migr_type
= 0;
6019 dev
->vol
.curr_migr_unit
= 0;
6020 memcpy(map
, map2
, sizeof_imsm_map(map2
));
6021 super
->updates_pending
++;
6024 if (a
->last_checkpoint
>= a
->info
.component_size
) {
6025 unsigned long long array_blocks
;
6029 used_disks
= imsm_num_data_members(dev
, 0);
6030 if (used_disks
> 0) {
6032 map
->blocks_per_member
*
6034 /* round array size down to closest MB
6036 array_blocks
= (array_blocks
6037 >> SECT_PER_MB_SHIFT
)
6038 << SECT_PER_MB_SHIFT
;
6039 a
->info
.custom_array_size
= array_blocks
;
6040 /* encourage manager to update array
6044 a
->check_reshape
= 1;
6046 /* finalize online capacity expansion/reshape */
6047 for (mdi
= a
->info
.devs
; mdi
; mdi
= mdi
->next
)
6049 mdi
->disk
.raid_disk
,
6052 imsm_progress_container_reshape(super
);
6057 /* before we activate this array handle any missing disks */
6058 if (consistent
== 2)
6059 handle_missing(super
, dev
);
6061 if (consistent
== 2 &&
6062 (!is_resync_complete(&a
->info
) ||
6063 map_state
!= IMSM_T_STATE_NORMAL
||
6064 dev
->vol
.migr_state
))
6067 if (is_resync_complete(&a
->info
)) {
6068 /* complete intialization / resync,
6069 * recovery and interrupted recovery is completed in
6072 if (is_resyncing(dev
)) {
6073 dprintf("imsm: mark resync done\n");
6074 end_migration(dev
, map_state
);
6075 super
->updates_pending
++;
6076 a
->last_checkpoint
= 0;
6078 } else if (!is_resyncing(dev
) && !failed
) {
6079 /* mark the start of the init process if nothing is failed */
6080 dprintf("imsm: mark resync start\n");
6081 if (map
->map_state
== IMSM_T_STATE_UNINITIALIZED
)
6082 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_INIT
);
6084 migrate(dev
, super
, IMSM_T_STATE_NORMAL
, MIGR_REPAIR
);
6085 super
->updates_pending
++;
6089 /* skip checkpointing for general migration,
6090 * it is controlled in mdadm
6092 if (is_gen_migration(dev
))
6093 goto skip_mark_checkpoint
;
6095 /* check if we can update curr_migr_unit from resync_start, recovery_start */
6096 blocks_per_unit
= blocks_per_migr_unit(super
, dev
);
6097 if (blocks_per_unit
) {
6101 units
= a
->last_checkpoint
/ blocks_per_unit
;
6104 /* check that we did not overflow 32-bits, and that
6105 * curr_migr_unit needs updating
6107 if (units32
== units
&&
6109 __le32_to_cpu(dev
->vol
.curr_migr_unit
) != units32
) {
6110 dprintf("imsm: mark checkpoint (%u)\n", units32
);
6111 dev
->vol
.curr_migr_unit
= __cpu_to_le32(units32
);
6112 super
->updates_pending
++;
6116 skip_mark_checkpoint
:
6117 /* mark dirty / clean */
6118 if (dev
->vol
.dirty
!= !consistent
) {
6119 dprintf("imsm: mark '%s'\n", consistent
? "clean" : "dirty");
6124 super
->updates_pending
++;
6130 static void imsm_set_disk(struct active_array
*a
, int n
, int state
)
6132 int inst
= a
->info
.container_member
;
6133 struct intel_super
*super
= a
->container
->sb
;
6134 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6135 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6136 struct imsm_disk
*disk
;
6141 if (n
> map
->num_members
)
6142 fprintf(stderr
, "imsm: set_disk %d out of range 0..%d\n",
6143 n
, map
->num_members
- 1);
6148 dprintf("imsm: set_disk %d:%x\n", n
, state
);
6150 ord
= get_imsm_ord_tbl_ent(dev
, n
, -1);
6151 disk
= get_imsm_disk(super
, ord_to_idx(ord
));
6153 /* check for new failures */
6154 if (state
& DS_FAULTY
) {
6155 if (mark_failure(dev
, disk
, ord_to_idx(ord
)))
6156 super
->updates_pending
++;
6159 /* check if in_sync */
6160 if (state
& DS_INSYNC
&& ord
& IMSM_ORD_REBUILD
&& is_rebuilding(dev
)) {
6161 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6163 set_imsm_ord_tbl_ent(migr_map
, n
, ord_to_idx(ord
));
6164 super
->updates_pending
++;
6167 failed
= imsm_count_failed(super
, dev
);
6168 map_state
= imsm_check_degraded(super
, dev
, failed
);
6170 /* check if recovery complete, newly degraded, or failed */
6171 if (map_state
== IMSM_T_STATE_NORMAL
&& is_rebuilding(dev
)) {
6172 end_migration(dev
, map_state
);
6173 map
= get_imsm_map(dev
, 0);
6174 map
->failed_disk_num
= ~0;
6175 super
->updates_pending
++;
6176 a
->last_checkpoint
= 0;
6177 } else if (map_state
== IMSM_T_STATE_DEGRADED
&&
6178 map
->map_state
!= map_state
&&
6179 !dev
->vol
.migr_state
) {
6180 dprintf("imsm: mark degraded\n");
6181 map
->map_state
= map_state
;
6182 super
->updates_pending
++;
6183 a
->last_checkpoint
= 0;
6184 } else if (map_state
== IMSM_T_STATE_FAILED
&&
6185 map
->map_state
!= map_state
) {
6186 dprintf("imsm: mark failed\n");
6187 end_migration(dev
, map_state
);
6188 super
->updates_pending
++;
6189 a
->last_checkpoint
= 0;
6190 } else if (is_gen_migration(dev
)) {
6191 dprintf("imsm: Detected General Migration in state: ");
6192 if (map_state
== IMSM_T_STATE_NORMAL
) {
6193 end_migration(dev
, map_state
);
6194 map
= get_imsm_map(dev
, 0);
6195 map
->failed_disk_num
= ~0;
6196 dprintf("normal\n");
6198 if (map_state
== IMSM_T_STATE_DEGRADED
) {
6199 printf("degraded\n");
6200 end_migration(dev
, map_state
);
6202 dprintf("failed\n");
6204 map
->map_state
= map_state
;
6206 super
->updates_pending
++;
6210 static int store_imsm_mpb(int fd
, struct imsm_super
*mpb
)
6213 __u32 mpb_size
= __le32_to_cpu(mpb
->mpb_size
);
6214 unsigned long long dsize
;
6215 unsigned long long sectors
;
6217 get_dev_size(fd
, NULL
, &dsize
);
6219 if (mpb_size
> 512) {
6220 /* -1 to account for anchor */
6221 sectors
= mpb_sectors(mpb
) - 1;
6223 /* write the extended mpb to the sectors preceeding the anchor */
6224 if (lseek64(fd
, dsize
- (512 * (2 + sectors
)), SEEK_SET
) < 0)
6227 if ((unsigned long long)write(fd
, buf
+ 512, 512 * sectors
)
6232 /* first block is stored on second to last sector of the disk */
6233 if (lseek64(fd
, dsize
- (512 * 2), SEEK_SET
) < 0)
6236 if (write(fd
, buf
, 512) != 512)
6242 static void imsm_sync_metadata(struct supertype
*container
)
6244 struct intel_super
*super
= container
->sb
;
6246 dprintf("sync metadata: %d\n", super
->updates_pending
);
6247 if (!super
->updates_pending
)
6250 write_super_imsm(container
, 0);
6252 super
->updates_pending
= 0;
6255 static struct dl
*imsm_readd(struct intel_super
*super
, int idx
, struct active_array
*a
)
6257 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6258 int i
= get_imsm_disk_idx(dev
, idx
, -1);
6261 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6265 if (dl
&& is_failed(&dl
->disk
))
6269 dprintf("%s: found %x:%x\n", __func__
, dl
->major
, dl
->minor
);
6274 static struct dl
*imsm_add_spare(struct intel_super
*super
, int slot
,
6275 struct active_array
*a
, int activate_new
,
6276 struct mdinfo
*additional_test_list
)
6278 struct imsm_dev
*dev
= get_imsm_dev(super
, a
->info
.container_member
);
6279 int idx
= get_imsm_disk_idx(dev
, slot
, -1);
6280 struct imsm_super
*mpb
= super
->anchor
;
6281 struct imsm_map
*map
;
6282 unsigned long long pos
;
6287 __u32 array_start
= 0;
6288 __u32 array_end
= 0;
6290 struct mdinfo
*test_list
;
6292 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6293 /* If in this array, skip */
6294 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6295 if (d
->state_fd
>= 0 &&
6296 d
->disk
.major
== dl
->major
&&
6297 d
->disk
.minor
== dl
->minor
) {
6298 dprintf("%x:%x already in array\n",
6299 dl
->major
, dl
->minor
);
6304 test_list
= additional_test_list
;
6306 if (test_list
->disk
.major
== dl
->major
&&
6307 test_list
->disk
.minor
== dl
->minor
) {
6308 dprintf("%x:%x already in additional test list\n",
6309 dl
->major
, dl
->minor
);
6312 test_list
= test_list
->next
;
6317 /* skip in use or failed drives */
6318 if (is_failed(&dl
->disk
) || idx
== dl
->index
||
6320 dprintf("%x:%x status (failed: %d index: %d)\n",
6321 dl
->major
, dl
->minor
, is_failed(&dl
->disk
), idx
);
6325 /* skip pure spares when we are looking for partially
6326 * assimilated drives
6328 if (dl
->index
== -1 && !activate_new
)
6331 /* Does this unused device have the requisite free space?
6332 * It needs to be able to cover all member volumes
6334 ex
= get_extents(super
, dl
);
6336 dprintf("cannot get extents\n");
6339 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
6340 dev
= get_imsm_dev(super
, i
);
6341 map
= get_imsm_map(dev
, 0);
6343 /* check if this disk is already a member of
6346 if (get_imsm_disk_slot(map
, dl
->index
) >= 0)
6352 array_start
= __le32_to_cpu(map
->pba_of_lba0
);
6353 array_end
= array_start
+
6354 __le32_to_cpu(map
->blocks_per_member
) - 1;
6357 /* check that we can start at pba_of_lba0 with
6358 * blocks_per_member of space
6360 if (array_start
>= pos
&& array_end
< ex
[j
].start
) {
6364 pos
= ex
[j
].start
+ ex
[j
].size
;
6366 } while (ex
[j
-1].size
);
6373 if (i
< mpb
->num_raid_devs
) {
6374 dprintf("%x:%x does not have %u to %u available\n",
6375 dl
->major
, dl
->minor
, array_start
, array_end
);
6386 static int imsm_rebuild_allowed(struct supertype
*cont
, int dev_idx
, int failed
)
6388 struct imsm_dev
*dev2
;
6389 struct imsm_map
*map
;
6395 dev2
= get_imsm_dev(cont
->sb
, dev_idx
);
6397 state
= imsm_check_degraded(cont
->sb
, dev2
, failed
);
6398 if (state
== IMSM_T_STATE_FAILED
) {
6399 map
= get_imsm_map(dev2
, 0);
6402 for (slot
= 0; slot
< map
->num_members
; slot
++) {
6404 * Check if failed disks are deleted from intel
6405 * disk list or are marked to be deleted
6407 idx
= get_imsm_disk_idx(dev2
, slot
, -1);
6408 idisk
= get_imsm_dl_disk(cont
->sb
, idx
);
6410 * Do not rebuild the array if failed disks
6411 * from failed sub-array are not removed from
6415 is_failed(&idisk
->disk
) &&
6416 (idisk
->action
!= DISK_REMOVE
))
6424 static struct mdinfo
*imsm_activate_spare(struct active_array
*a
,
6425 struct metadata_update
**updates
)
6428 * Find a device with unused free space and use it to replace a
6429 * failed/vacant region in an array. We replace failed regions one a
6430 * array at a time. The result is that a new spare disk will be added
6431 * to the first failed array and after the monitor has finished
6432 * propagating failures the remainder will be consumed.
6434 * FIXME add a capability for mdmon to request spares from another
6438 struct intel_super
*super
= a
->container
->sb
;
6439 int inst
= a
->info
.container_member
;
6440 struct imsm_dev
*dev
= get_imsm_dev(super
, inst
);
6441 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6442 int failed
= a
->info
.array
.raid_disks
;
6443 struct mdinfo
*rv
= NULL
;
6446 struct metadata_update
*mu
;
6448 struct imsm_update_activate_spare
*u
;
6453 for (d
= a
->info
.devs
; d
; d
= d
->next
) {
6454 if ((d
->curr_state
& DS_FAULTY
) &&
6456 /* wait for Removal to happen */
6458 if (d
->state_fd
>= 0)
6462 dprintf("imsm: activate spare: inst=%d failed=%d (%d) level=%d\n",
6463 inst
, failed
, a
->info
.array
.raid_disks
, a
->info
.array
.level
);
6465 if (dev
->vol
.migr_state
&&
6466 dev
->vol
.migr_type
== MIGR_GEN_MIGR
)
6467 /* No repair during migration */
6470 if (a
->info
.array
.level
== 4)
6471 /* No repair for takeovered array
6472 * imsm doesn't support raid4
6476 if (imsm_check_degraded(super
, dev
, failed
) != IMSM_T_STATE_DEGRADED
)
6480 * If there are any failed disks check state of the other volume.
6481 * Block rebuild if the another one is failed until failed disks
6482 * are removed from container.
6485 dprintf("found failed disks in %s, check if there another"
6486 "failed sub-array.\n",
6488 /* check if states of the other volumes allow for rebuild */
6489 for (i
= 0; i
< super
->anchor
->num_raid_devs
; i
++) {
6491 allowed
= imsm_rebuild_allowed(a
->container
,
6499 /* For each slot, if it is not working, find a spare */
6500 for (i
= 0; i
< a
->info
.array
.raid_disks
; i
++) {
6501 for (d
= a
->info
.devs
; d
; d
= d
->next
)
6502 if (d
->disk
.raid_disk
== i
)
6504 dprintf("found %d: %p %x\n", i
, d
, d
?d
->curr_state
:0);
6505 if (d
&& (d
->state_fd
>= 0))
6509 * OK, this device needs recovery. Try to re-add the
6510 * previous occupant of this slot, if this fails see if
6511 * we can continue the assimilation of a spare that was
6512 * partially assimilated, finally try to activate a new
6515 dl
= imsm_readd(super
, i
, a
);
6517 dl
= imsm_add_spare(super
, i
, a
, 0, NULL
);
6519 dl
= imsm_add_spare(super
, i
, a
, 1, NULL
);
6523 /* found a usable disk with enough space */
6524 di
= malloc(sizeof(*di
));
6527 memset(di
, 0, sizeof(*di
));
6529 /* dl->index will be -1 in the case we are activating a
6530 * pristine spare. imsm_process_update() will create a
6531 * new index in this case. Once a disk is found to be
6532 * failed in all member arrays it is kicked from the
6535 di
->disk
.number
= dl
->index
;
6537 /* (ab)use di->devs to store a pointer to the device
6540 di
->devs
= (struct mdinfo
*) dl
;
6542 di
->disk
.raid_disk
= i
;
6543 di
->disk
.major
= dl
->major
;
6544 di
->disk
.minor
= dl
->minor
;
6546 di
->recovery_start
= 0;
6547 di
->data_offset
= __le32_to_cpu(map
->pba_of_lba0
);
6548 di
->component_size
= a
->info
.component_size
;
6549 di
->container_member
= inst
;
6550 super
->random
= random32();
6554 dprintf("%x:%x to be %d at %llu\n", dl
->major
, dl
->minor
,
6555 i
, di
->data_offset
);
6561 /* No spares found */
6563 /* Now 'rv' has a list of devices to return.
6564 * Create a metadata_update record to update the
6565 * disk_ord_tbl for the array
6567 mu
= malloc(sizeof(*mu
));
6569 mu
->buf
= malloc(sizeof(struct imsm_update_activate_spare
) * num_spares
);
6570 if (mu
->buf
== NULL
) {
6577 struct mdinfo
*n
= rv
->next
;
6586 mu
->space_list
= NULL
;
6587 mu
->len
= sizeof(struct imsm_update_activate_spare
) * num_spares
;
6588 mu
->next
= *updates
;
6589 u
= (struct imsm_update_activate_spare
*) mu
->buf
;
6591 for (di
= rv
; di
; di
= di
->next
) {
6592 u
->type
= update_activate_spare
;
6593 u
->dl
= (struct dl
*) di
->devs
;
6595 u
->slot
= di
->disk
.raid_disk
;
6606 static int disks_overlap(struct intel_super
*super
, int idx
, struct imsm_update_create_array
*u
)
6608 struct imsm_dev
*dev
= get_imsm_dev(super
, idx
);
6609 struct imsm_map
*map
= get_imsm_map(dev
, 0);
6610 struct imsm_map
*new_map
= get_imsm_map(&u
->dev
, 0);
6611 struct disk_info
*inf
= get_disk_info(u
);
6612 struct imsm_disk
*disk
;
6616 for (i
= 0; i
< map
->num_members
; i
++) {
6617 disk
= get_imsm_disk(super
, get_imsm_disk_idx(dev
, i
, -1));
6618 for (j
= 0; j
< new_map
->num_members
; j
++)
6619 if (serialcmp(disk
->serial
, inf
[j
].serial
) == 0)
6627 static struct dl
*get_disk_super(struct intel_super
*super
, int major
, int minor
)
6629 struct dl
*dl
= NULL
;
6630 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
6631 if ((dl
->major
== major
) && (dl
->minor
== minor
))
6636 static int remove_disk_super(struct intel_super
*super
, int major
, int minor
)
6638 struct dl
*prev
= NULL
;
6642 for (dl
= super
->disks
; dl
; dl
= dl
->next
) {
6643 if ((dl
->major
== major
) && (dl
->minor
== minor
)) {
6646 prev
->next
= dl
->next
;
6648 super
->disks
= dl
->next
;
6650 __free_imsm_disk(dl
);
6651 dprintf("%s: removed %x:%x\n",
6652 __func__
, major
, minor
);
6660 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
);
6662 static int add_remove_disk_update(struct intel_super
*super
)
6664 int check_degraded
= 0;
6665 struct dl
*disk
= NULL
;
6666 /* add/remove some spares to/from the metadata/contrainer */
6667 while (super
->disk_mgmt_list
) {
6668 struct dl
*disk_cfg
;
6670 disk_cfg
= super
->disk_mgmt_list
;
6671 super
->disk_mgmt_list
= disk_cfg
->next
;
6672 disk_cfg
->next
= NULL
;
6674 if (disk_cfg
->action
== DISK_ADD
) {
6675 disk_cfg
->next
= super
->disks
;
6676 super
->disks
= disk_cfg
;
6678 dprintf("%s: added %x:%x\n",
6679 __func__
, disk_cfg
->major
,
6681 } else if (disk_cfg
->action
== DISK_REMOVE
) {
6682 dprintf("Disk remove action processed: %x.%x\n",
6683 disk_cfg
->major
, disk_cfg
->minor
);
6684 disk
= get_disk_super(super
,
6688 /* store action status */
6689 disk
->action
= DISK_REMOVE
;
6690 /* remove spare disks only */
6691 if (disk
->index
== -1) {
6692 remove_disk_super(super
,
6697 /* release allocate disk structure */
6698 __free_imsm_disk(disk_cfg
);
6701 return check_degraded
;
6705 static int apply_reshape_migration_update(struct imsm_update_reshape_migration
*u
,
6706 struct intel_super
*super
,
6709 struct intel_dev
*id
;
6710 void **tofree
= NULL
;
6713 dprintf("apply_reshape_migration_update()\n");
6714 if ((u
->subdev
< 0) ||
6716 dprintf("imsm: Error: Wrong subdev: %i\n", u
->subdev
);
6719 if ((space_list
== NULL
) || (*space_list
== NULL
)) {
6720 dprintf("imsm: Error: Memory is not allocated\n");
6724 for (id
= super
->devlist
; id
; id
= id
->next
) {
6725 if (id
->index
== (unsigned)u
->subdev
) {
6726 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subdev
);
6727 struct imsm_map
*map
;
6728 struct imsm_dev
*new_dev
=
6729 (struct imsm_dev
*)*space_list
;
6730 struct imsm_map
*migr_map
= get_imsm_map(dev
, 1);
6732 struct dl
*new_disk
;
6734 if (new_dev
== NULL
)
6736 *space_list
= **space_list
;
6737 memcpy(new_dev
, dev
, sizeof_imsm_dev(dev
, 0));
6738 map
= get_imsm_map(new_dev
, 0);
6740 dprintf("imsm: Error: migration in progress");
6744 to_state
= map
->map_state
;
6745 if ((u
->new_level
== 5) && (map
->raid_level
== 0)) {
6747 /* this should not happen */
6748 if (u
->new_disks
[0] < 0) {
6749 map
->failed_disk_num
=
6750 map
->num_members
- 1;
6751 to_state
= IMSM_T_STATE_DEGRADED
;
6753 to_state
= IMSM_T_STATE_NORMAL
;
6755 migrate(new_dev
, super
, to_state
, MIGR_GEN_MIGR
);
6756 if (u
->new_level
> -1)
6757 map
->raid_level
= u
->new_level
;
6758 migr_map
= get_imsm_map(new_dev
, 1);
6759 if ((u
->new_level
== 5) &&
6760 (migr_map
->raid_level
== 0)) {
6761 int ord
= map
->num_members
- 1;
6762 migr_map
->num_members
--;
6763 if (u
->new_disks
[0] < 0)
6764 ord
|= IMSM_ORD_REBUILD
;
6765 set_imsm_ord_tbl_ent(map
,
6766 map
->num_members
- 1,
6770 tofree
= (void **)dev
;
6772 /* update chunk size
6774 if (u
->new_chunksize
> 0)
6775 map
->blocks_per_strip
=
6776 __cpu_to_le16(u
->new_chunksize
* 2);
6780 if ((u
->new_level
!= 5) ||
6781 (migr_map
->raid_level
!= 0) ||
6782 (migr_map
->raid_level
== map
->raid_level
))
6785 if (u
->new_disks
[0] >= 0) {
6788 new_disk
= get_disk_super(super
,
6789 major(u
->new_disks
[0]),
6790 minor(u
->new_disks
[0]));
6791 dprintf("imsm: new disk for reshape is: %i:%i "
6792 "(%p, index = %i)\n",
6793 major(u
->new_disks
[0]),
6794 minor(u
->new_disks
[0]),
6795 new_disk
, new_disk
->index
);
6796 if (new_disk
== NULL
)
6797 goto error_disk_add
;
6799 new_disk
->index
= map
->num_members
- 1;
6800 /* slot to fill in autolayout
6802 new_disk
->raiddisk
= new_disk
->index
;
6803 new_disk
->disk
.status
|= CONFIGURED_DISK
;
6804 new_disk
->disk
.status
&= ~SPARE_DISK
;
6806 goto error_disk_add
;
6809 *tofree
= *space_list
;
6810 /* calculate new size
6812 imsm_set_array_size(new_dev
);
6819 *space_list
= tofree
;
6823 dprintf("Error: imsm: Cannot find disk.\n");
6828 static int apply_reshape_container_disks_update(struct imsm_update_reshape
*u
,
6829 struct intel_super
*super
,
6832 struct dl
*new_disk
;
6833 struct intel_dev
*id
;
6835 int delta_disks
= u
->new_raid_disks
- u
->old_raid_disks
;
6836 int disk_count
= u
->old_raid_disks
;
6837 void **tofree
= NULL
;
6838 int devices_to_reshape
= 1;
6839 struct imsm_super
*mpb
= super
->anchor
;
6841 unsigned int dev_id
;
6843 dprintf("imsm: apply_reshape_container_disks_update()\n");
6845 /* enable spares to use in array */
6846 for (i
= 0; i
< delta_disks
; i
++) {
6847 new_disk
= get_disk_super(super
,
6848 major(u
->new_disks
[i
]),
6849 minor(u
->new_disks
[i
]));
6850 dprintf("imsm: new disk for reshape is: %i:%i "
6851 "(%p, index = %i)\n",
6852 major(u
->new_disks
[i
]), minor(u
->new_disks
[i
]),
6853 new_disk
, new_disk
->index
);
6854 if ((new_disk
== NULL
) ||
6855 ((new_disk
->index
>= 0) &&
6856 (new_disk
->index
< u
->old_raid_disks
)))
6857 goto update_reshape_exit
;
6858 new_disk
->index
= disk_count
++;
6859 /* slot to fill in autolayout
6861 new_disk
->raiddisk
= new_disk
->index
;
6862 new_disk
->disk
.status
|=
6864 new_disk
->disk
.status
&= ~SPARE_DISK
;
6867 dprintf("imsm: volume set mpb->num_raid_devs = %i\n",
6868 mpb
->num_raid_devs
);
6869 /* manage changes in volume
6871 for (dev_id
= 0; dev_id
< mpb
->num_raid_devs
; dev_id
++) {
6872 void **sp
= *space_list
;
6873 struct imsm_dev
*newdev
;
6874 struct imsm_map
*newmap
, *oldmap
;
6876 for (id
= super
->devlist
; id
; id
= id
->next
) {
6877 if (id
->index
== dev_id
)
6886 /* Copy the dev, but not (all of) the map */
6887 memcpy(newdev
, id
->dev
, sizeof(*newdev
));
6888 oldmap
= get_imsm_map(id
->dev
, 0);
6889 newmap
= get_imsm_map(newdev
, 0);
6890 /* Copy the current map */
6891 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6892 /* update one device only
6894 if (devices_to_reshape
) {
6895 dprintf("imsm: modifying subdev: %i\n",
6897 devices_to_reshape
--;
6898 newdev
->vol
.migr_state
= 1;
6899 newdev
->vol
.curr_migr_unit
= 0;
6900 newdev
->vol
.migr_type
= MIGR_GEN_MIGR
;
6901 newmap
->num_members
= u
->new_raid_disks
;
6902 for (i
= 0; i
< delta_disks
; i
++) {
6903 set_imsm_ord_tbl_ent(newmap
,
6904 u
->old_raid_disks
+ i
,
6905 u
->old_raid_disks
+ i
);
6907 /* New map is correct, now need to save old map
6909 newmap
= get_imsm_map(newdev
, 1);
6910 memcpy(newmap
, oldmap
, sizeof_imsm_map(oldmap
));
6912 imsm_set_array_size(newdev
);
6915 sp
= (void **)id
->dev
;
6920 /* Clear migration record */
6921 memset(super
->migr_rec
, 0, sizeof(struct migr_record
));
6924 *space_list
= tofree
;
6927 update_reshape_exit
:
6932 static int apply_takeover_update(struct imsm_update_takeover
*u
,
6933 struct intel_super
*super
,
6936 struct imsm_dev
*dev
= NULL
;
6937 struct intel_dev
*dv
;
6938 struct imsm_dev
*dev_new
;
6939 struct imsm_map
*map
;
6943 for (dv
= super
->devlist
; dv
; dv
= dv
->next
)
6944 if (dv
->index
== (unsigned int)u
->subarray
) {
6952 map
= get_imsm_map(dev
, 0);
6954 if (u
->direction
== R10_TO_R0
) {
6955 /* Number of failed disks must be half of initial disk number */
6956 if (imsm_count_failed(super
, dev
) != (map
->num_members
/ 2))
6959 /* iterate through devices to mark removed disks as spare */
6960 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6961 if (dm
->disk
.status
& FAILED_DISK
) {
6962 int idx
= dm
->index
;
6963 /* update indexes on the disk list */
6964 /* FIXME this loop-with-the-loop looks wrong, I'm not convinced
6965 the index values will end up being correct.... NB */
6966 for (du
= super
->disks
; du
; du
= du
->next
)
6967 if (du
->index
> idx
)
6969 /* mark as spare disk */
6970 dm
->disk
.status
= SPARE_DISK
;
6975 map
->num_members
= map
->num_members
/ 2;
6976 map
->map_state
= IMSM_T_STATE_NORMAL
;
6977 map
->num_domains
= 1;
6978 map
->raid_level
= 0;
6979 map
->failed_disk_num
= -1;
6982 if (u
->direction
== R0_TO_R10
) {
6984 /* update slots in current disk list */
6985 for (dm
= super
->disks
; dm
; dm
= dm
->next
) {
6989 /* create new *missing* disks */
6990 for (i
= 0; i
< map
->num_members
; i
++) {
6991 space
= *space_list
;
6994 *space_list
= *space
;
6996 memcpy(du
, super
->disks
, sizeof(*du
));
7000 du
->index
= (i
* 2) + 1;
7001 sprintf((char *)du
->disk
.serial
,
7002 " MISSING_%d", du
->index
);
7003 sprintf((char *)du
->serial
,
7004 "MISSING_%d", du
->index
);
7005 du
->next
= super
->missing
;
7006 super
->missing
= du
;
7008 /* create new dev and map */
7009 space
= *space_list
;
7012 *space_list
= *space
;
7013 dev_new
= (void *)space
;
7014 memcpy(dev_new
, dev
, sizeof(*dev
));
7015 /* update new map */
7016 map
= get_imsm_map(dev_new
, 0);
7017 map
->num_members
= map
->num_members
* 2;
7018 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7019 map
->num_domains
= 2;
7020 map
->raid_level
= 1;
7021 /* replace dev<->dev_new */
7024 /* update disk order table */
7025 for (du
= super
->disks
; du
; du
= du
->next
)
7027 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7028 for (du
= super
->missing
; du
; du
= du
->next
)
7029 if (du
->index
>= 0) {
7030 set_imsm_ord_tbl_ent(map
, du
->index
, du
->index
);
7031 mark_missing(dv
->dev
, &du
->disk
, du
->index
);
7037 static void imsm_process_update(struct supertype
*st
,
7038 struct metadata_update
*update
)
7041 * crack open the metadata_update envelope to find the update record
7042 * update can be one of:
7043 * update_reshape_container_disks - all the arrays in the container
7044 * are being reshaped to have more devices. We need to mark
7045 * the arrays for general migration and convert selected spares
7046 * into active devices.
7047 * update_activate_spare - a spare device has replaced a failed
7048 * device in an array, update the disk_ord_tbl. If this disk is
7049 * present in all member arrays then also clear the SPARE_DISK
7051 * update_create_array
7053 * update_rename_array
7054 * update_add_remove_disk
7056 struct intel_super
*super
= st
->sb
;
7057 struct imsm_super
*mpb
;
7058 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7060 /* update requires a larger buf but the allocation failed */
7061 if (super
->next_len
&& !super
->next_buf
) {
7062 super
->next_len
= 0;
7066 if (super
->next_buf
) {
7067 memcpy(super
->next_buf
, super
->buf
, super
->len
);
7069 super
->len
= super
->next_len
;
7070 super
->buf
= super
->next_buf
;
7072 super
->next_len
= 0;
7073 super
->next_buf
= NULL
;
7076 mpb
= super
->anchor
;
7079 case update_general_migration_checkpoint
: {
7080 struct intel_dev
*id
;
7081 struct imsm_update_general_migration_checkpoint
*u
=
7082 (void *)update
->buf
;
7084 dprintf("imsm: process_update() "
7085 "for update_general_migration_checkpoint called\n");
7087 /* find device under general migration */
7088 for (id
= super
->devlist
; id
; id
= id
->next
) {
7089 if (is_gen_migration(id
->dev
)) {
7090 id
->dev
->vol
.curr_migr_unit
=
7091 __cpu_to_le32(u
->curr_migr_unit
);
7092 super
->updates_pending
++;
7097 case update_takeover
: {
7098 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7099 if (apply_takeover_update(u
, super
, &update
->space_list
)) {
7100 imsm_update_version_info(super
);
7101 super
->updates_pending
++;
7106 case update_reshape_container_disks
: {
7107 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7108 if (apply_reshape_container_disks_update(
7109 u
, super
, &update
->space_list
))
7110 super
->updates_pending
++;
7113 case update_reshape_migration
: {
7114 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7115 if (apply_reshape_migration_update(
7116 u
, super
, &update
->space_list
))
7117 super
->updates_pending
++;
7120 case update_activate_spare
: {
7121 struct imsm_update_activate_spare
*u
= (void *) update
->buf
;
7122 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->array
);
7123 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7124 struct imsm_map
*migr_map
;
7125 struct active_array
*a
;
7126 struct imsm_disk
*disk
;
7131 int victim
= get_imsm_disk_idx(dev
, u
->slot
, -1);
7134 for (dl
= super
->disks
; dl
; dl
= dl
->next
)
7139 fprintf(stderr
, "error: imsm_activate_spare passed "
7140 "an unknown disk (index: %d)\n",
7145 super
->updates_pending
++;
7146 /* count failures (excluding rebuilds and the victim)
7147 * to determine map[0] state
7150 for (i
= 0; i
< map
->num_members
; i
++) {
7153 disk
= get_imsm_disk(super
,
7154 get_imsm_disk_idx(dev
, i
, -1));
7155 if (!disk
|| is_failed(disk
))
7159 /* adding a pristine spare, assign a new index */
7160 if (dl
->index
< 0) {
7161 dl
->index
= super
->anchor
->num_disks
;
7162 super
->anchor
->num_disks
++;
7165 disk
->status
|= CONFIGURED_DISK
;
7166 disk
->status
&= ~SPARE_DISK
;
7169 to_state
= imsm_check_degraded(super
, dev
, failed
);
7170 map
->map_state
= IMSM_T_STATE_DEGRADED
;
7171 migrate(dev
, super
, to_state
, MIGR_REBUILD
);
7172 migr_map
= get_imsm_map(dev
, 1);
7173 set_imsm_ord_tbl_ent(map
, u
->slot
, dl
->index
);
7174 set_imsm_ord_tbl_ent(migr_map
, u
->slot
, dl
->index
| IMSM_ORD_REBUILD
);
7176 /* update the family_num to mark a new container
7177 * generation, being careful to record the existing
7178 * family_num in orig_family_num to clean up after
7179 * earlier mdadm versions that neglected to set it.
7181 if (mpb
->orig_family_num
== 0)
7182 mpb
->orig_family_num
= mpb
->family_num
;
7183 mpb
->family_num
+= super
->random
;
7185 /* count arrays using the victim in the metadata */
7187 for (a
= st
->arrays
; a
; a
= a
->next
) {
7188 dev
= get_imsm_dev(super
, a
->info
.container_member
);
7189 map
= get_imsm_map(dev
, 0);
7191 if (get_imsm_disk_slot(map
, victim
) >= 0)
7195 /* delete the victim if it is no longer being
7201 /* We know that 'manager' isn't touching anything,
7202 * so it is safe to delete
7204 for (dlp
= &super
->disks
; *dlp
; dlp
= &(*dlp
)->next
)
7205 if ((*dlp
)->index
== victim
)
7208 /* victim may be on the missing list */
7210 for (dlp
= &super
->missing
; *dlp
; dlp
= &(*dlp
)->next
)
7211 if ((*dlp
)->index
== victim
)
7213 imsm_delete(super
, dlp
, victim
);
7217 case update_create_array
: {
7218 /* someone wants to create a new array, we need to be aware of
7219 * a few races/collisions:
7220 * 1/ 'Create' called by two separate instances of mdadm
7221 * 2/ 'Create' versus 'activate_spare': mdadm has chosen
7222 * devices that have since been assimilated via
7224 * In the event this update can not be carried out mdadm will
7225 * (FIX ME) notice that its update did not take hold.
7227 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7228 struct intel_dev
*dv
;
7229 struct imsm_dev
*dev
;
7230 struct imsm_map
*map
, *new_map
;
7231 unsigned long long start
, end
;
7232 unsigned long long new_start
, new_end
;
7234 struct disk_info
*inf
;
7237 /* handle racing creates: first come first serve */
7238 if (u
->dev_idx
< mpb
->num_raid_devs
) {
7239 dprintf("%s: subarray %d already defined\n",
7240 __func__
, u
->dev_idx
);
7244 /* check update is next in sequence */
7245 if (u
->dev_idx
!= mpb
->num_raid_devs
) {
7246 dprintf("%s: can not create array %d expected index %d\n",
7247 __func__
, u
->dev_idx
, mpb
->num_raid_devs
);
7251 new_map
= get_imsm_map(&u
->dev
, 0);
7252 new_start
= __le32_to_cpu(new_map
->pba_of_lba0
);
7253 new_end
= new_start
+ __le32_to_cpu(new_map
->blocks_per_member
);
7254 inf
= get_disk_info(u
);
7256 /* handle activate_spare versus create race:
7257 * check to make sure that overlapping arrays do not include
7260 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7261 dev
= get_imsm_dev(super
, i
);
7262 map
= get_imsm_map(dev
, 0);
7263 start
= __le32_to_cpu(map
->pba_of_lba0
);
7264 end
= start
+ __le32_to_cpu(map
->blocks_per_member
);
7265 if ((new_start
>= start
&& new_start
<= end
) ||
7266 (start
>= new_start
&& start
<= new_end
))
7271 if (disks_overlap(super
, i
, u
)) {
7272 dprintf("%s: arrays overlap\n", __func__
);
7277 /* check that prepare update was successful */
7278 if (!update
->space
) {
7279 dprintf("%s: prepare update failed\n", __func__
);
7283 /* check that all disks are still active before committing
7284 * changes. FIXME: could we instead handle this by creating a
7285 * degraded array? That's probably not what the user expects,
7286 * so better to drop this update on the floor.
7288 for (i
= 0; i
< new_map
->num_members
; i
++) {
7289 dl
= serial_to_dl(inf
[i
].serial
, super
);
7291 dprintf("%s: disk disappeared\n", __func__
);
7296 super
->updates_pending
++;
7298 /* convert spares to members and fixup ord_tbl */
7299 for (i
= 0; i
< new_map
->num_members
; i
++) {
7300 dl
= serial_to_dl(inf
[i
].serial
, super
);
7301 if (dl
->index
== -1) {
7302 dl
->index
= mpb
->num_disks
;
7304 dl
->disk
.status
|= CONFIGURED_DISK
;
7305 dl
->disk
.status
&= ~SPARE_DISK
;
7307 set_imsm_ord_tbl_ent(new_map
, i
, dl
->index
);
7312 update
->space
= NULL
;
7313 imsm_copy_dev(dev
, &u
->dev
);
7314 dv
->index
= u
->dev_idx
;
7315 dv
->next
= super
->devlist
;
7316 super
->devlist
= dv
;
7317 mpb
->num_raid_devs
++;
7319 imsm_update_version_info(super
);
7322 /* mdmon knows how to release update->space, but not
7323 * ((struct intel_dev *) update->space)->dev
7325 if (update
->space
) {
7331 case update_kill_array
: {
7332 struct imsm_update_kill_array
*u
= (void *) update
->buf
;
7333 int victim
= u
->dev_idx
;
7334 struct active_array
*a
;
7335 struct intel_dev
**dp
;
7336 struct imsm_dev
*dev
;
7338 /* sanity check that we are not affecting the uuid of
7339 * active arrays, or deleting an active array
7341 * FIXME when immutable ids are available, but note that
7342 * we'll also need to fixup the invalidated/active
7343 * subarray indexes in mdstat
7345 for (a
= st
->arrays
; a
; a
= a
->next
)
7346 if (a
->info
.container_member
>= victim
)
7348 /* by definition if mdmon is running at least one array
7349 * is active in the container, so checking
7350 * mpb->num_raid_devs is just extra paranoia
7352 dev
= get_imsm_dev(super
, victim
);
7353 if (a
|| !dev
|| mpb
->num_raid_devs
== 1) {
7354 dprintf("failed to delete subarray-%d\n", victim
);
7358 for (dp
= &super
->devlist
; *dp
;)
7359 if ((*dp
)->index
== (unsigned)super
->current_vol
) {
7362 if ((*dp
)->index
> (unsigned)victim
)
7366 mpb
->num_raid_devs
--;
7367 super
->updates_pending
++;
7370 case update_rename_array
: {
7371 struct imsm_update_rename_array
*u
= (void *) update
->buf
;
7372 char name
[MAX_RAID_SERIAL_LEN
+1];
7373 int target
= u
->dev_idx
;
7374 struct active_array
*a
;
7375 struct imsm_dev
*dev
;
7377 /* sanity check that we are not affecting the uuid of
7380 snprintf(name
, MAX_RAID_SERIAL_LEN
, "%s", (char *) u
->name
);
7381 name
[MAX_RAID_SERIAL_LEN
] = '\0';
7382 for (a
= st
->arrays
; a
; a
= a
->next
)
7383 if (a
->info
.container_member
== target
)
7385 dev
= get_imsm_dev(super
, u
->dev_idx
);
7386 if (a
|| !dev
|| !check_name(super
, name
, 1)) {
7387 dprintf("failed to rename subarray-%d\n", target
);
7391 snprintf((char *) dev
->volume
, MAX_RAID_SERIAL_LEN
, "%s", name
);
7392 super
->updates_pending
++;
7395 case update_add_remove_disk
: {
7396 /* we may be able to repair some arrays if disks are
7397 * being added, check teh status of add_remove_disk
7398 * if discs has been added.
7400 if (add_remove_disk_update(super
)) {
7401 struct active_array
*a
;
7403 super
->updates_pending
++;
7404 for (a
= st
->arrays
; a
; a
= a
->next
)
7405 a
->check_degraded
= 1;
7410 fprintf(stderr
, "error: unsuported process update type:"
7411 "(type: %d)\n", type
);
7415 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
);
7417 static void imsm_prepare_update(struct supertype
*st
,
7418 struct metadata_update
*update
)
7421 * Allocate space to hold new disk entries, raid-device entries or a new
7422 * mpb if necessary. The manager synchronously waits for updates to
7423 * complete in the monitor, so new mpb buffers allocated here can be
7424 * integrated by the monitor thread without worrying about live pointers
7425 * in the manager thread.
7427 enum imsm_update_type type
= *(enum imsm_update_type
*) update
->buf
;
7428 struct intel_super
*super
= st
->sb
;
7429 struct imsm_super
*mpb
= super
->anchor
;
7434 case update_general_migration_checkpoint
:
7435 dprintf("imsm: prepare_update() "
7436 "for update_general_migration_checkpoint called\n");
7438 case update_takeover
: {
7439 struct imsm_update_takeover
*u
= (void *)update
->buf
;
7440 if (u
->direction
== R0_TO_R10
) {
7441 void **tail
= (void **)&update
->space_list
;
7442 struct imsm_dev
*dev
= get_imsm_dev(super
, u
->subarray
);
7443 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7444 int num_members
= map
->num_members
;
7448 /* allocate memory for added disks */
7449 for (i
= 0; i
< num_members
; i
++) {
7450 size
= sizeof(struct dl
);
7451 space
= malloc(size
);
7460 /* allocate memory for new device */
7461 size
= sizeof_imsm_dev(super
->devlist
->dev
, 0) +
7462 (num_members
* sizeof(__u32
));
7463 space
= malloc(size
);
7472 len
= disks_to_mpb_size(num_members
* 2);
7474 /* if allocation didn't success, free buffer */
7475 while (update
->space_list
) {
7476 void **sp
= update
->space_list
;
7477 update
->space_list
= *sp
;
7485 case update_reshape_container_disks
: {
7486 /* Every raid device in the container is about to
7487 * gain some more devices, and we will enter a
7489 * So each 'imsm_map' will be bigger, and the imsm_vol
7490 * will now hold 2 of them.
7491 * Thus we need new 'struct imsm_dev' allocations sized
7492 * as sizeof_imsm_dev but with more devices in both maps.
7494 struct imsm_update_reshape
*u
= (void *)update
->buf
;
7495 struct intel_dev
*dl
;
7496 void **space_tail
= (void**)&update
->space_list
;
7498 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7500 for (dl
= super
->devlist
; dl
; dl
= dl
->next
) {
7501 int size
= sizeof_imsm_dev(dl
->dev
, 1);
7503 if (u
->new_raid_disks
> u
->old_raid_disks
)
7504 size
+= sizeof(__u32
)*2*
7505 (u
->new_raid_disks
- u
->old_raid_disks
);
7514 len
= disks_to_mpb_size(u
->new_raid_disks
);
7515 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7518 case update_reshape_migration
: {
7519 /* for migration level 0->5 we need to add disks
7520 * so the same as for container operation we will copy
7521 * device to the bigger location.
7522 * in memory prepared device and new disk area are prepared
7523 * for usage in process update
7525 struct imsm_update_reshape_migration
*u
= (void *)update
->buf
;
7526 struct intel_dev
*id
;
7527 void **space_tail
= (void **)&update
->space_list
;
7530 int current_level
= -1;
7532 dprintf("imsm: imsm_prepare_update() for update_reshape\n");
7534 /* add space for bigger array in update
7536 for (id
= super
->devlist
; id
; id
= id
->next
) {
7537 if (id
->index
== (unsigned)u
->subdev
) {
7538 size
= sizeof_imsm_dev(id
->dev
, 1);
7539 if (u
->new_raid_disks
> u
->old_raid_disks
)
7540 size
+= sizeof(__u32
)*2*
7541 (u
->new_raid_disks
- u
->old_raid_disks
);
7551 if (update
->space_list
== NULL
)
7554 /* add space for disk in update
7556 size
= sizeof(struct dl
);
7559 free(update
->space_list
);
7560 update
->space_list
= NULL
;
7567 /* add spare device to update
7569 for (id
= super
->devlist
; id
; id
= id
->next
)
7570 if (id
->index
== (unsigned)u
->subdev
) {
7571 struct imsm_dev
*dev
;
7572 struct imsm_map
*map
;
7574 dev
= get_imsm_dev(super
, u
->subdev
);
7575 map
= get_imsm_map(dev
, 0);
7576 current_level
= map
->raid_level
;
7579 if ((u
->new_level
== 5) && (u
->new_level
!= current_level
)) {
7580 struct mdinfo
*spares
;
7582 spares
= get_spares_for_grow(st
);
7590 makedev(dev
->disk
.major
,
7592 dl
= get_disk_super(super
,
7595 dl
->index
= u
->old_raid_disks
;
7601 len
= disks_to_mpb_size(u
->new_raid_disks
);
7602 dprintf("New anchor length is %llu\n", (unsigned long long)len
);
7605 case update_create_array
: {
7606 struct imsm_update_create_array
*u
= (void *) update
->buf
;
7607 struct intel_dev
*dv
;
7608 struct imsm_dev
*dev
= &u
->dev
;
7609 struct imsm_map
*map
= get_imsm_map(dev
, 0);
7611 struct disk_info
*inf
;
7615 inf
= get_disk_info(u
);
7616 len
= sizeof_imsm_dev(dev
, 1);
7617 /* allocate a new super->devlist entry */
7618 dv
= malloc(sizeof(*dv
));
7620 dv
->dev
= malloc(len
);
7625 update
->space
= NULL
;
7629 /* count how many spares will be converted to members */
7630 for (i
= 0; i
< map
->num_members
; i
++) {
7631 dl
= serial_to_dl(inf
[i
].serial
, super
);
7633 /* hmm maybe it failed?, nothing we can do about
7638 if (count_memberships(dl
, super
) == 0)
7641 len
+= activate
* sizeof(struct imsm_disk
);
7648 /* check if we need a larger metadata buffer */
7649 if (super
->next_buf
)
7650 buf_len
= super
->next_len
;
7652 buf_len
= super
->len
;
7654 if (__le32_to_cpu(mpb
->mpb_size
) + len
> buf_len
) {
7655 /* ok we need a larger buf than what is currently allocated
7656 * if this allocation fails process_update will notice that
7657 * ->next_len is set and ->next_buf is NULL
7659 buf_len
= ROUND_UP(__le32_to_cpu(mpb
->mpb_size
) + len
, 512);
7660 if (super
->next_buf
)
7661 free(super
->next_buf
);
7663 super
->next_len
= buf_len
;
7664 if (posix_memalign(&super
->next_buf
, 512, buf_len
) == 0)
7665 memset(super
->next_buf
, 0, buf_len
);
7667 super
->next_buf
= NULL
;
7671 /* must be called while manager is quiesced */
7672 static void imsm_delete(struct intel_super
*super
, struct dl
**dlp
, unsigned index
)
7674 struct imsm_super
*mpb
= super
->anchor
;
7676 struct imsm_dev
*dev
;
7677 struct imsm_map
*map
;
7678 int i
, j
, num_members
;
7681 dprintf("%s: deleting device[%d] from imsm_super\n",
7684 /* shift all indexes down one */
7685 for (iter
= super
->disks
; iter
; iter
= iter
->next
)
7686 if (iter
->index
> (int)index
)
7688 for (iter
= super
->missing
; iter
; iter
= iter
->next
)
7689 if (iter
->index
> (int)index
)
7692 for (i
= 0; i
< mpb
->num_raid_devs
; i
++) {
7693 dev
= get_imsm_dev(super
, i
);
7694 map
= get_imsm_map(dev
, 0);
7695 num_members
= map
->num_members
;
7696 for (j
= 0; j
< num_members
; j
++) {
7697 /* update ord entries being careful not to propagate
7698 * ord-flags to the first map
7700 ord
= get_imsm_ord_tbl_ent(dev
, j
, -1);
7702 if (ord_to_idx(ord
) <= index
)
7705 map
= get_imsm_map(dev
, 0);
7706 set_imsm_ord_tbl_ent(map
, j
, ord_to_idx(ord
- 1));
7707 map
= get_imsm_map(dev
, 1);
7709 set_imsm_ord_tbl_ent(map
, j
, ord
- 1);
7714 super
->updates_pending
++;
7716 struct dl
*dl
= *dlp
;
7718 *dlp
= (*dlp
)->next
;
7719 __free_imsm_disk(dl
);
7722 #endif /* MDASSEMBLE */
7723 /*******************************************************************************
7724 * Function: open_backup_targets
7725 * Description: Function opens file descriptors for all devices given in
7728 * info : general array info
7729 * raid_disks : number of disks
7730 * raid_fds : table of device's file descriptors
7734 ******************************************************************************/
7735 int open_backup_targets(struct mdinfo
*info
, int raid_disks
, int *raid_fds
)
7739 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7742 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
)) {
7743 dprintf("disk is faulty!!\n");
7747 if ((sd
->disk
.raid_disk
>= raid_disks
) ||
7748 (sd
->disk
.raid_disk
< 0))
7751 dn
= map_dev(sd
->disk
.major
,
7753 raid_fds
[sd
->disk
.raid_disk
] = dev_open(dn
, O_RDWR
);
7754 if (raid_fds
[sd
->disk
.raid_disk
] < 0) {
7755 fprintf(stderr
, "cannot open component\n");
7763 /*******************************************************************************
7764 * Function: init_migr_record_imsm
7765 * Description: Function inits imsm migration record
7767 * super : imsm internal array info
7768 * dev : device under migration
7769 * info : general array info to find the smallest device
7772 ******************************************************************************/
7773 void init_migr_record_imsm(struct supertype
*st
, struct imsm_dev
*dev
,
7774 struct mdinfo
*info
)
7776 struct intel_super
*super
= st
->sb
;
7777 struct migr_record
*migr_rec
= super
->migr_rec
;
7779 unsigned long long dsize
, dev_sectors
;
7780 long long unsigned min_dev_sectors
= -1LLU;
7784 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7785 struct imsm_map
*map_src
= get_imsm_map(dev
, 1);
7786 unsigned long long num_migr_units
;
7787 unsigned long long array_blocks
;
7789 memset(migr_rec
, 0, sizeof(struct migr_record
));
7790 migr_rec
->family_num
= __cpu_to_le32(super
->anchor
->family_num
);
7792 /* only ascending reshape supported now */
7793 migr_rec
->ascending_migr
= __cpu_to_le32(1);
7795 migr_rec
->dest_depth_per_unit
= GEN_MIGR_AREA_SIZE
/
7796 max(map_dest
->blocks_per_strip
, map_src
->blocks_per_strip
);
7797 migr_rec
->dest_depth_per_unit
*= map_dest
->blocks_per_strip
;
7798 new_data_disks
= imsm_num_data_members(dev
, 0);
7799 migr_rec
->blocks_per_unit
=
7800 __cpu_to_le32(migr_rec
->dest_depth_per_unit
* new_data_disks
);
7801 migr_rec
->dest_depth_per_unit
=
7802 __cpu_to_le32(migr_rec
->dest_depth_per_unit
);
7803 array_blocks
= info
->component_size
* new_data_disks
;
7805 array_blocks
/ __le32_to_cpu(migr_rec
->blocks_per_unit
);
7807 if (array_blocks
% __le32_to_cpu(migr_rec
->blocks_per_unit
))
7809 migr_rec
->num_migr_units
= __cpu_to_le32(num_migr_units
);
7811 migr_rec
->post_migr_vol_cap
= dev
->size_low
;
7812 migr_rec
->post_migr_vol_cap_hi
= dev
->size_high
;
7815 /* Find the smallest dev */
7816 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
7817 sprintf(nm
, "%d:%d", sd
->disk
.major
, sd
->disk
.minor
);
7818 fd
= dev_open(nm
, O_RDONLY
);
7821 get_dev_size(fd
, NULL
, &dsize
);
7822 dev_sectors
= dsize
/ 512;
7823 if (dev_sectors
< min_dev_sectors
)
7824 min_dev_sectors
= dev_sectors
;
7827 migr_rec
->ckpt_area_pba
= __cpu_to_le32(min_dev_sectors
-
7828 RAID_DISK_RESERVED_BLOCKS_IMSM_HI
);
7830 write_imsm_migr_rec(st
);
7835 /*******************************************************************************
7836 * Function: save_backup_imsm
7837 * Description: Function saves critical data stripes to Migration Copy Area
7838 * and updates the current migration unit status.
7839 * Use restore_stripes() to form a destination stripe,
7840 * and to write it to the Copy Area.
7842 * st : supertype information
7843 * dev : imsm device that backup is saved for
7844 * info : general array info
7845 * buf : input buffer
7846 * length : length of data to backup (blocks_per_unit)
7850 ******************************************************************************/
7851 int save_backup_imsm(struct supertype
*st
,
7852 struct imsm_dev
*dev
,
7853 struct mdinfo
*info
,
7858 struct intel_super
*super
= st
->sb
;
7859 unsigned long long *target_offsets
= NULL
;
7860 int *targets
= NULL
;
7862 struct imsm_map
*map_dest
= get_imsm_map(dev
, 0);
7863 int new_disks
= map_dest
->num_members
;
7864 int dest_layout
= 0;
7866 unsigned long long start
;
7867 int data_disks
= imsm_num_data_members(dev
, 0);
7869 targets
= malloc(new_disks
* sizeof(int));
7873 for (i
= 0; i
< new_disks
; i
++)
7876 target_offsets
= malloc(new_disks
* sizeof(unsigned long long));
7877 if (!target_offsets
)
7880 start
= info
->reshape_progress
* 512;
7881 for (i
= 0; i
< new_disks
; i
++) {
7882 target_offsets
[i
] = (unsigned long long)
7883 __le32_to_cpu(super
->migr_rec
->ckpt_area_pba
) * 512;
7884 /* move back copy area adderss, it will be moved forward
7885 * in restore_stripes() using start input variable
7887 target_offsets
[i
] -= start
/data_disks
;
7890 if (open_backup_targets(info
, new_disks
, targets
))
7893 dest_layout
= imsm_level_to_layout(map_dest
->raid_level
);
7894 dest_chunk
= __le16_to_cpu(map_dest
->blocks_per_strip
) * 512;
7896 if (restore_stripes(targets
, /* list of dest devices */
7897 target_offsets
, /* migration record offsets */
7900 map_dest
->raid_level
,
7902 -1, /* source backup file descriptor */
7903 0, /* input buf offset
7904 * always 0 buf is already offseted */
7908 fprintf(stderr
, Name
": Error restoring stripes\n");
7916 for (i
= 0; i
< new_disks
; i
++)
7917 if (targets
[i
] >= 0)
7921 free(target_offsets
);
7926 /*******************************************************************************
7927 * Function: save_checkpoint_imsm
7928 * Description: Function called for current unit status update
7929 * in the migration record. It writes it to disk.
7931 * super : imsm internal array info
7932 * info : general array info
7936 * 2: failure, means no valid migration record
7937 * / no general migration in progress /
7938 ******************************************************************************/
7939 int save_checkpoint_imsm(struct supertype
*st
, struct mdinfo
*info
, int state
)
7941 struct intel_super
*super
= st
->sb
;
7942 unsigned long long blocks_per_unit
;
7943 unsigned long long curr_migr_unit
;
7945 if (load_imsm_migr_rec(super
, info
) != 0) {
7946 dprintf("imsm: ERROR: Cannot read migration record "
7947 "for checkpoint save.\n");
7951 blocks_per_unit
= __le32_to_cpu(super
->migr_rec
->blocks_per_unit
);
7952 if (blocks_per_unit
== 0) {
7953 dprintf("imsm: no migration in progress.\n");
7956 curr_migr_unit
= info
->reshape_progress
/ blocks_per_unit
;
7957 /* check if array is alligned to copy area
7958 * if it is not alligned, add one to current migration unit value
7959 * this can happend on array reshape finish only
7961 if (info
->reshape_progress
% blocks_per_unit
)
7964 super
->migr_rec
->curr_migr_unit
=
7965 __cpu_to_le32(curr_migr_unit
);
7966 super
->migr_rec
->rec_status
= __cpu_to_le32(state
);
7967 super
->migr_rec
->dest_1st_member_lba
=
7968 __cpu_to_le32(curr_migr_unit
*
7969 __le32_to_cpu(super
->migr_rec
->dest_depth_per_unit
));
7970 if (write_imsm_migr_rec(st
) < 0) {
7971 dprintf("imsm: Cannot write migration record "
7972 "outside backup area\n");
7979 /*******************************************************************************
7980 * Function: recover_backup_imsm
7981 * Description: Function recovers critical data from the Migration Copy Area
7982 * while assembling an array.
7984 * super : imsm internal array info
7985 * info : general array info
7987 * 0 : success (or there is no data to recover)
7989 ******************************************************************************/
7990 int recover_backup_imsm(struct supertype
*st
, struct mdinfo
*info
)
7992 struct intel_super
*super
= st
->sb
;
7993 struct migr_record
*migr_rec
= super
->migr_rec
;
7994 struct imsm_map
*map_dest
= NULL
;
7995 struct intel_dev
*id
= NULL
;
7996 unsigned long long read_offset
;
7997 unsigned long long write_offset
;
7999 int *targets
= NULL
;
8000 int new_disks
, i
, err
;
8003 unsigned long curr_migr_unit
= __le32_to_cpu(migr_rec
->curr_migr_unit
);
8004 unsigned long num_migr_units
= __le32_to_cpu(migr_rec
->num_migr_units
);
8006 int skipped_disks
= 0;
8007 int max_degradation
;
8009 err
= sysfs_get_str(info
, NULL
, "array_state", (char *)buffer
, 20);
8013 /* recover data only during assemblation */
8014 if (strncmp(buffer
, "inactive", 8) != 0)
8016 /* no data to recover */
8017 if (__le32_to_cpu(migr_rec
->rec_status
) == UNIT_SRC_NORMAL
)
8019 if (curr_migr_unit
>= num_migr_units
)
8022 /* find device during reshape */
8023 for (id
= super
->devlist
; id
; id
= id
->next
)
8024 if (is_gen_migration(id
->dev
))
8029 map_dest
= get_imsm_map(id
->dev
, 0);
8030 new_disks
= map_dest
->num_members
;
8031 max_degradation
= new_disks
- imsm_num_data_members(id
->dev
, 0);
8033 read_offset
= (unsigned long long)
8034 __le32_to_cpu(migr_rec
->ckpt_area_pba
) * 512;
8036 write_offset
= ((unsigned long long)
8037 __le32_to_cpu(migr_rec
->dest_1st_member_lba
) +
8038 __le32_to_cpu(map_dest
->pba_of_lba0
)) * 512;
8040 unit_len
= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8041 if (posix_memalign((void **)&buf
, 512, unit_len
) != 0)
8043 targets
= malloc(new_disks
* sizeof(int));
8047 open_backup_targets(info
, new_disks
, targets
);
8049 for (i
= 0; i
< new_disks
; i
++) {
8050 if (targets
[i
] < 0) {
8054 if (lseek64(targets
[i
], read_offset
, SEEK_SET
) < 0) {
8056 Name
": Cannot seek to block: %s\n",
8060 if ((unsigned)read(targets
[i
], buf
, unit_len
) != unit_len
) {
8062 Name
": Cannot read copy area block: %s\n",
8066 if (lseek64(targets
[i
], write_offset
, SEEK_SET
) < 0) {
8068 Name
": Cannot seek to block: %s\n",
8072 if ((unsigned)write(targets
[i
], buf
, unit_len
) != unit_len
) {
8074 Name
": Cannot restore block: %s\n",
8080 if (skipped_disks
> max_degradation
) {
8082 Name
": Cannot restore data from backup."
8083 " Too many failed disks\n");
8087 if (save_checkpoint_imsm(st
, info
, UNIT_SRC_NORMAL
)) {
8088 /* ignore error == 2, this can mean end of reshape here
8090 dprintf("imsm: Cannot write checkpoint to "
8091 "migration record (UNIT_SRC_NORMAL) during restart\n");
8097 for (i
= 0; i
< new_disks
; i
++)
8106 static char disk_by_path
[] = "/dev/disk/by-path/";
8108 static const char *imsm_get_disk_controller_domain(const char *path
)
8110 char disk_path
[PATH_MAX
];
8114 strncpy(disk_path
, disk_by_path
, PATH_MAX
- 1);
8115 strncat(disk_path
, path
, PATH_MAX
- strlen(disk_path
) - 1);
8116 if (stat(disk_path
, &st
) == 0) {
8117 struct sys_dev
* hba
;
8120 path
= devt_to_devpath(st
.st_rdev
);
8123 hba
= find_disk_attached_hba(-1, path
);
8124 if (hba
&& hba
->type
== SYS_DEV_SAS
)
8126 else if (hba
&& hba
->type
== SYS_DEV_SATA
)
8130 dprintf("path: %s hba: %s attached: %s\n",
8131 path
, (hba
) ? hba
->path
: "NULL", drv
);
8139 static int imsm_find_array_minor_by_subdev(int subdev
, int container
, int *minor
)
8141 char subdev_name
[20];
8142 struct mdstat_ent
*mdstat
;
8144 sprintf(subdev_name
, "%d", subdev
);
8145 mdstat
= mdstat_by_subdev(subdev_name
, container
);
8149 *minor
= mdstat
->devnum
;
8150 free_mdstat(mdstat
);
8154 static int imsm_reshape_is_allowed_on_container(struct supertype
*st
,
8155 struct geo_params
*geo
,
8156 int *old_raid_disks
)
8158 /* currently we only support increasing the number of devices
8159 * for a container. This increases the number of device for each
8160 * member array. They must all be RAID0 or RAID5.
8163 struct mdinfo
*info
, *member
;
8164 int devices_that_can_grow
= 0;
8166 dprintf("imsm: imsm_reshape_is_allowed_on_container(ENTER): "
8167 "st->devnum = (%i)\n",
8170 if (geo
->size
!= -1 ||
8171 geo
->level
!= UnSet
||
8172 geo
->layout
!= UnSet
||
8173 geo
->chunksize
!= 0 ||
8174 geo
->raid_disks
== UnSet
) {
8175 dprintf("imsm: Container operation is allowed for "
8176 "raid disks number change only.\n");
8180 info
= container_content_imsm(st
, NULL
);
8181 for (member
= info
; member
; member
= member
->next
) {
8185 dprintf("imsm: checking device_num: %i\n",
8186 member
->container_member
);
8188 if (geo
->raid_disks
<= member
->array
.raid_disks
) {
8189 /* we work on container for Online Capacity Expansion
8190 * only so raid_disks has to grow
8192 dprintf("imsm: for container operation raid disks "
8193 "increase is required\n");
8197 if ((info
->array
.level
!= 0) &&
8198 (info
->array
.level
!= 5)) {
8199 /* we cannot use this container with other raid level
8201 dprintf("imsm: for container operation wrong"
8202 " raid level (%i) detected\n",
8206 /* check for platform support
8207 * for this raid level configuration
8209 struct intel_super
*super
= st
->sb
;
8210 if (!is_raid_level_supported(super
->orom
,
8211 member
->array
.level
,
8213 dprintf("platform does not support raid%d with"
8217 geo
->raid_disks
> 1 ? "s" : "");
8220 /* check if component size is aligned to chunk size
8222 if (info
->component_size
%
8223 (info
->array
.chunk_size
/512)) {
8224 dprintf("Component size is not aligned to "
8230 if (*old_raid_disks
&&
8231 info
->array
.raid_disks
!= *old_raid_disks
)
8233 *old_raid_disks
= info
->array
.raid_disks
;
8235 /* All raid5 and raid0 volumes in container
8236 * have to be ready for Online Capacity Expansion
8237 * so they need to be assembled. We have already
8238 * checked that no recovery etc is happening.
8240 result
= imsm_find_array_minor_by_subdev(member
->container_member
,
8244 dprintf("imsm: cannot find array\n");
8247 devices_that_can_grow
++;
8250 if (!member
&& devices_that_can_grow
)
8254 dprintf("\tContainer operation allowed\n");
8256 dprintf("\tError: %i\n", ret_val
);
8261 /* Function: get_spares_for_grow
8262 * Description: Allocates memory and creates list of spare devices
8263 * avaliable in container. Checks if spare drive size is acceptable.
8264 * Parameters: Pointer to the supertype structure
8265 * Returns: Pointer to the list of spare devices (mdinfo structure) on success,
8268 static struct mdinfo
*get_spares_for_grow(struct supertype
*st
)
8270 unsigned long long min_size
= min_acceptable_spare_size_imsm(st
);
8271 return container_choose_spares(st
, min_size
, NULL
, NULL
, NULL
, 0);
8274 /******************************************************************************
8275 * function: imsm_create_metadata_update_for_reshape
8276 * Function creates update for whole IMSM container.
8278 ******************************************************************************/
8279 static int imsm_create_metadata_update_for_reshape(
8280 struct supertype
*st
,
8281 struct geo_params
*geo
,
8283 struct imsm_update_reshape
**updatep
)
8285 struct intel_super
*super
= st
->sb
;
8286 struct imsm_super
*mpb
= super
->anchor
;
8287 int update_memory_size
= 0;
8288 struct imsm_update_reshape
*u
= NULL
;
8289 struct mdinfo
*spares
= NULL
;
8291 int delta_disks
= 0;
8294 dprintf("imsm_update_metadata_for_reshape(enter) raid_disks = %i\n",
8297 delta_disks
= geo
->raid_disks
- old_raid_disks
;
8299 /* size of all update data without anchor */
8300 update_memory_size
= sizeof(struct imsm_update_reshape
);
8302 /* now add space for spare disks that we need to add. */
8303 update_memory_size
+= sizeof(u
->new_disks
[0]) * (delta_disks
- 1);
8305 u
= calloc(1, update_memory_size
);
8308 "cannot get memory for imsm_update_reshape update\n");
8311 u
->type
= update_reshape_container_disks
;
8312 u
->old_raid_disks
= old_raid_disks
;
8313 u
->new_raid_disks
= geo
->raid_disks
;
8315 /* now get spare disks list
8317 spares
= get_spares_for_grow(st
);
8320 || delta_disks
> spares
->array
.spare_disks
) {
8321 fprintf(stderr
, Name
": imsm: ERROR: Cannot get spare devices "
8322 "for %s.\n", geo
->dev_name
);
8327 /* we have got spares
8328 * update disk list in imsm_disk list table in anchor
8330 dprintf("imsm: %i spares are available.\n\n",
8331 spares
->array
.spare_disks
);
8334 for (i
= 0; i
< delta_disks
; i
++) {
8339 u
->new_disks
[i
] = makedev(dev
->disk
.major
,
8341 dl
= get_disk_super(super
, dev
->disk
.major
, dev
->disk
.minor
);
8342 dl
->index
= mpb
->num_disks
;
8352 dprintf("imsm: reshape update preparation :");
8353 if (i
== delta_disks
) {
8356 return update_memory_size
;
8359 dprintf(" Error\n");
8364 /******************************************************************************
8365 * function: imsm_create_metadata_update_for_migration()
8366 * Creates update for IMSM array.
8368 ******************************************************************************/
8369 static int imsm_create_metadata_update_for_migration(
8370 struct supertype
*st
,
8371 struct geo_params
*geo
,
8372 struct imsm_update_reshape_migration
**updatep
)
8374 struct intel_super
*super
= st
->sb
;
8375 int update_memory_size
= 0;
8376 struct imsm_update_reshape_migration
*u
= NULL
;
8377 struct imsm_dev
*dev
;
8378 int previous_level
= -1;
8380 dprintf("imsm_create_metadata_update_for_migration(enter)"
8381 " New Level = %i\n", geo
->level
);
8383 /* size of all update data without anchor */
8384 update_memory_size
= sizeof(struct imsm_update_reshape_migration
);
8386 u
= calloc(1, update_memory_size
);
8388 dprintf("error: cannot get memory for "
8389 "imsm_create_metadata_update_for_migration\n");
8392 u
->type
= update_reshape_migration
;
8393 u
->subdev
= super
->current_vol
;
8394 u
->new_level
= geo
->level
;
8395 u
->new_layout
= geo
->layout
;
8396 u
->new_raid_disks
= u
->old_raid_disks
= geo
->raid_disks
;
8397 u
->new_disks
[0] = -1;
8398 u
->new_chunksize
= -1;
8400 dev
= get_imsm_dev(super
, u
->subdev
);
8402 struct imsm_map
*map
;
8404 map
= get_imsm_map(dev
, 0);
8406 int current_chunk_size
=
8407 __le16_to_cpu(map
->blocks_per_strip
) / 2;
8409 if (geo
->chunksize
!= current_chunk_size
) {
8410 u
->new_chunksize
= geo
->chunksize
/ 1024;
8412 "chunk size change from %i to %i\n",
8413 current_chunk_size
, u
->new_chunksize
);
8415 previous_level
= map
->raid_level
;
8418 if ((geo
->level
== 5) && (previous_level
== 0)) {
8419 struct mdinfo
*spares
= NULL
;
8421 u
->new_raid_disks
++;
8422 spares
= get_spares_for_grow(st
);
8423 if ((spares
== NULL
) || (spares
->array
.spare_disks
< 1)) {
8426 update_memory_size
= 0;
8427 dprintf("error: cannot get spare device "
8428 "for requested migration");
8433 dprintf("imsm: reshape update preparation : OK\n");
8436 return update_memory_size
;
8439 static void imsm_update_metadata_locally(struct supertype
*st
,
8442 struct metadata_update mu
;
8447 mu
.space_list
= NULL
;
8449 imsm_prepare_update(st
, &mu
);
8450 imsm_process_update(st
, &mu
);
8452 while (mu
.space_list
) {
8453 void **space
= mu
.space_list
;
8454 mu
.space_list
= *space
;
8459 /***************************************************************************
8460 * Function: imsm_analyze_change
8461 * Description: Function analyze change for single volume
8462 * and validate if transition is supported
8463 * Parameters: Geometry parameters, supertype structure
8464 * Returns: Operation type code on success, -1 if fail
8465 ****************************************************************************/
8466 enum imsm_reshape_type
imsm_analyze_change(struct supertype
*st
,
8467 struct geo_params
*geo
)
8474 getinfo_super_imsm_volume(st
, &info
, NULL
);
8475 if ((geo
->level
!= info
.array
.level
) &&
8476 (geo
->level
>= 0) &&
8477 (geo
->level
!= UnSet
)) {
8478 switch (info
.array
.level
) {
8480 if (geo
->level
== 5) {
8481 change
= CH_MIGRATION
;
8482 if (geo
->layout
!= ALGORITHM_LEFT_ASYMMETRIC
) {
8484 Name
" Error. Requested Layout "
8485 "not supported (left-asymmetric layout "
8486 "is supported only)!\n");
8488 goto analyse_change_exit
;
8492 if (geo
->level
== 10) {
8493 change
= CH_TAKEOVER
;
8498 if (geo
->level
== 0) {
8499 change
= CH_TAKEOVER
;
8504 if (geo
->level
== 0) {
8505 change
= CH_TAKEOVER
;
8512 Name
" Error. Level Migration from %d to %d "
8514 info
.array
.level
, geo
->level
);
8515 goto analyse_change_exit
;
8518 geo
->level
= info
.array
.level
;
8520 if ((geo
->layout
!= info
.array
.layout
)
8521 && ((geo
->layout
!= UnSet
) && (geo
->layout
!= -1))) {
8522 change
= CH_MIGRATION
;
8523 if ((info
.array
.layout
== 0)
8524 && (info
.array
.level
== 5)
8525 && (geo
->layout
== 5)) {
8526 /* reshape 5 -> 4 */
8527 } else if ((info
.array
.layout
== 5)
8528 && (info
.array
.level
== 5)
8529 && (geo
->layout
== 0)) {
8530 /* reshape 4 -> 5 */
8535 Name
" Error. Layout Migration from %d to %d "
8537 info
.array
.layout
, geo
->layout
);
8539 goto analyse_change_exit
;
8542 geo
->layout
= info
.array
.layout
;
8544 if ((geo
->chunksize
> 0) && (geo
->chunksize
!= UnSet
)
8545 && (geo
->chunksize
!= info
.array
.chunk_size
))
8546 change
= CH_MIGRATION
;
8548 geo
->chunksize
= info
.array
.chunk_size
;
8550 chunk
= geo
->chunksize
/ 1024;
8551 if (!validate_geometry_imsm(st
,
8561 struct intel_super
*super
= st
->sb
;
8562 struct imsm_super
*mpb
= super
->anchor
;
8564 if (mpb
->num_raid_devs
> 1) {
8566 Name
" Error. Cannot perform operation on %s"
8567 "- for this operation it MUST be single "
8568 "array in container\n",
8574 analyse_change_exit
:
8579 int imsm_takeover(struct supertype
*st
, struct geo_params
*geo
)
8581 struct intel_super
*super
= st
->sb
;
8582 struct imsm_update_takeover
*u
;
8584 u
= malloc(sizeof(struct imsm_update_takeover
));
8588 u
->type
= update_takeover
;
8589 u
->subarray
= super
->current_vol
;
8591 /* 10->0 transition */
8592 if (geo
->level
== 0)
8593 u
->direction
= R10_TO_R0
;
8595 /* 0->10 transition */
8596 if (geo
->level
== 10)
8597 u
->direction
= R0_TO_R10
;
8599 /* update metadata locally */
8600 imsm_update_metadata_locally(st
, u
,
8601 sizeof(struct imsm_update_takeover
));
8602 /* and possibly remotely */
8603 if (st
->update_tail
)
8604 append_metadata_update(st
, u
,
8605 sizeof(struct imsm_update_takeover
));
8612 static int imsm_reshape_super(struct supertype
*st
, long long size
, int level
,
8613 int layout
, int chunksize
, int raid_disks
,
8614 int delta_disks
, char *backup
, char *dev
,
8618 struct geo_params geo
;
8620 dprintf("imsm: reshape_super called.\n");
8622 memset(&geo
, 0, sizeof(struct geo_params
));
8625 geo
.dev_id
= st
->devnum
;
8628 geo
.layout
= layout
;
8629 geo
.chunksize
= chunksize
;
8630 geo
.raid_disks
= raid_disks
;
8631 if (delta_disks
!= UnSet
)
8632 geo
.raid_disks
+= delta_disks
;
8634 dprintf("\tfor level : %i\n", geo
.level
);
8635 dprintf("\tfor raid_disks : %i\n", geo
.raid_disks
);
8637 if (experimental() == 0)
8640 if (st
->container_dev
== st
->devnum
) {
8641 /* On container level we can only increase number of devices. */
8642 dprintf("imsm: info: Container operation\n");
8643 int old_raid_disks
= 0;
8645 if (imsm_reshape_is_allowed_on_container(
8646 st
, &geo
, &old_raid_disks
)) {
8647 struct imsm_update_reshape
*u
= NULL
;
8650 len
= imsm_create_metadata_update_for_reshape(
8651 st
, &geo
, old_raid_disks
, &u
);
8654 dprintf("imsm: Cannot prepare update\n");
8655 goto exit_imsm_reshape_super
;
8659 /* update metadata locally */
8660 imsm_update_metadata_locally(st
, u
, len
);
8661 /* and possibly remotely */
8662 if (st
->update_tail
)
8663 append_metadata_update(st
, u
, len
);
8668 fprintf(stderr
, Name
": (imsm) Operation "
8669 "is not allowed on this container\n");
8672 /* On volume level we support following operations
8673 * - takeover: raid10 -> raid0; raid0 -> raid10
8674 * - chunk size migration
8675 * - migration: raid5 -> raid0; raid0 -> raid5
8677 struct intel_super
*super
= st
->sb
;
8678 struct intel_dev
*dev
= super
->devlist
;
8680 dprintf("imsm: info: Volume operation\n");
8681 /* find requested device */
8683 if (imsm_find_array_minor_by_subdev(
8684 dev
->index
, st
->container_dev
, &devnum
) == 0
8685 && devnum
== geo
.dev_id
)
8690 fprintf(stderr
, Name
" Cannot find %s (%i) subarray\n",
8691 geo
.dev_name
, geo
.dev_id
);
8692 goto exit_imsm_reshape_super
;
8694 super
->current_vol
= dev
->index
;
8695 change
= imsm_analyze_change(st
, &geo
);
8698 ret_val
= imsm_takeover(st
, &geo
);
8700 case CH_MIGRATION
: {
8701 struct imsm_update_reshape_migration
*u
= NULL
;
8703 imsm_create_metadata_update_for_migration(
8707 "Cannot prepare update\n");
8711 /* update metadata locally */
8712 imsm_update_metadata_locally(st
, u
, len
);
8713 /* and possibly remotely */
8714 if (st
->update_tail
)
8715 append_metadata_update(st
, u
, len
);
8725 exit_imsm_reshape_super
:
8726 dprintf("imsm: reshape_super Exit code = %i\n", ret_val
);
8730 /*******************************************************************************
8731 * Function: wait_for_reshape_imsm
8732 * Description: Function writes new sync_max value and waits until
8733 * reshape process reach new position
8735 * sra : general array info
8736 * ndata : number of disks in new array's layout
8739 * 1 : there is no reshape in progress,
8741 ******************************************************************************/
8742 int wait_for_reshape_imsm(struct mdinfo
*sra
, int ndata
)
8744 int fd
= sysfs_get_fd(sra
, NULL
, "reshape_position");
8745 unsigned long long completed
;
8746 /* to_complete : new sync_max position */
8747 unsigned long long to_complete
= sra
->reshape_progress
;
8748 unsigned long long position_to_set
= to_complete
/ ndata
;
8751 dprintf("imsm: wait_for_reshape_imsm() "
8752 "cannot open reshape_position\n");
8756 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8757 dprintf("imsm: wait_for_reshape_imsm() "
8758 "cannot read reshape_position (no reshape in progres)\n");
8763 if (completed
> to_complete
) {
8764 dprintf("imsm: wait_for_reshape_imsm() "
8765 "wrong next position to set %llu (%llu)\n",
8766 to_complete
, completed
);
8770 dprintf("Position set: %llu\n", position_to_set
);
8771 if (sysfs_set_num(sra
, NULL
, "sync_max",
8772 position_to_set
) != 0) {
8773 dprintf("imsm: wait_for_reshape_imsm() "
8774 "cannot set reshape position to %llu\n",
8785 select(fd
+1, &rfds
, NULL
, NULL
, NULL
);
8786 if (sysfs_get_str(sra
, NULL
, "sync_action",
8788 strncmp(action
, "reshape", 7) != 0)
8790 if (sysfs_fd_get_ll(fd
, &completed
) < 0) {
8791 dprintf("imsm: wait_for_reshape_imsm() "
8792 "cannot read reshape_position (in loop)\n");
8796 } while (completed
< to_complete
);
8802 /*******************************************************************************
8803 * Function: check_degradation_change
8804 * Description: Check that array hasn't become failed.
8806 * info : for sysfs access
8807 * sources : source disks descriptors
8808 * degraded: previous degradation level
8811 ******************************************************************************/
8812 int check_degradation_change(struct mdinfo
*info
,
8816 unsigned long long new_degraded
;
8817 sysfs_get_ll(info
, NULL
, "degraded", &new_degraded
);
8818 if (new_degraded
!= (unsigned long long)degraded
) {
8819 /* check each device to ensure it is still working */
8822 for (sd
= info
->devs
; sd
; sd
= sd
->next
) {
8823 if (sd
->disk
.state
& (1<<MD_DISK_FAULTY
))
8825 if (sd
->disk
.state
& (1<<MD_DISK_SYNC
)) {
8827 if (sysfs_get_str(info
,
8828 sd
, "state", sbuf
, 20) < 0 ||
8829 strstr(sbuf
, "faulty") ||
8830 strstr(sbuf
, "in_sync") == NULL
) {
8831 /* this device is dead */
8832 sd
->disk
.state
= (1<<MD_DISK_FAULTY
);
8833 if (sd
->disk
.raid_disk
>= 0 &&
8834 sources
[sd
->disk
.raid_disk
] >= 0) {
8836 sd
->disk
.raid_disk
]);
8837 sources
[sd
->disk
.raid_disk
] =
8846 return new_degraded
;
8849 /*******************************************************************************
8850 * Function: imsm_manage_reshape
8851 * Description: Function finds array under reshape and it manages reshape
8852 * process. It creates stripes backups (if required) and sets
8855 * afd : Backup handle (nattive) - not used
8856 * sra : general array info
8857 * reshape : reshape parameters - not used
8858 * st : supertype structure
8859 * blocks : size of critical section [blocks]
8860 * fds : table of source device descriptor
8861 * offsets : start of array (offest per devices)
8863 * destfd : table of destination device descriptor
8864 * destoffsets : table of destination offsets (per device)
8866 * 1 : success, reshape is done
8868 ******************************************************************************/
8869 static int imsm_manage_reshape(
8870 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
8871 struct supertype
*st
, unsigned long backup_blocks
,
8872 int *fds
, unsigned long long *offsets
,
8873 int dests
, int *destfd
, unsigned long long *destoffsets
)
8876 struct intel_super
*super
= st
->sb
;
8877 struct intel_dev
*dv
= NULL
;
8878 struct imsm_dev
*dev
= NULL
;
8879 struct imsm_map
*map_src
;
8880 int migr_vol_qan
= 0;
8881 int ndata
, odata
; /* [bytes] */
8882 int chunk
; /* [bytes] */
8883 struct migr_record
*migr_rec
;
8885 unsigned int buf_size
; /* [bytes] */
8886 unsigned long long max_position
; /* array size [bytes] */
8887 unsigned long long next_step
; /* [blocks]/[bytes] */
8888 unsigned long long old_data_stripe_length
;
8889 unsigned long long start_src
; /* [bytes] */
8890 unsigned long long start
; /* [bytes] */
8891 unsigned long long start_buf_shift
; /* [bytes] */
8893 int source_layout
= 0;
8895 if (!fds
|| !offsets
|| !sra
)
8898 /* Find volume during the reshape */
8899 for (dv
= super
->devlist
; dv
; dv
= dv
->next
) {
8900 if (dv
->dev
->vol
.migr_type
== MIGR_GEN_MIGR
8901 && dv
->dev
->vol
.migr_state
== 1) {
8906 /* Only one volume can migrate at the same time */
8907 if (migr_vol_qan
!= 1) {
8908 fprintf(stderr
, Name
" : %s", migr_vol_qan
?
8909 "Number of migrating volumes greater than 1\n" :
8910 "There is no volume during migrationg\n");
8914 map_src
= get_imsm_map(dev
, 1);
8915 if (map_src
== NULL
)
8918 ndata
= imsm_num_data_members(dev
, 0);
8919 odata
= imsm_num_data_members(dev
, 1);
8921 chunk
= __le16_to_cpu(map_src
->blocks_per_strip
) * 512;
8922 old_data_stripe_length
= odata
* chunk
;
8924 migr_rec
= super
->migr_rec
;
8926 /* initialize migration record for start condition */
8927 if (sra
->reshape_progress
== 0)
8928 init_migr_record_imsm(st
, dev
, sra
);
8930 if (__le32_to_cpu(migr_rec
->rec_status
) != UNIT_SRC_NORMAL
) {
8931 dprintf("imsm: cannot restart migration when data "
8932 "are present in copy area.\n");
8938 buf_size
= __le32_to_cpu(migr_rec
->blocks_per_unit
) * 512;
8939 /* extend buffer size for parity disk */
8940 buf_size
+= __le32_to_cpu(migr_rec
->dest_depth_per_unit
) * 512;
8941 /* add space for stripe aligment */
8942 buf_size
+= old_data_stripe_length
;
8943 if (posix_memalign((void **)&buf
, 4096, buf_size
)) {
8944 dprintf("imsm: Cannot allocate checpoint buffer\n");
8948 max_position
= sra
->component_size
* ndata
;
8949 source_layout
= imsm_level_to_layout(map_src
->raid_level
);
8951 while (__le32_to_cpu(migr_rec
->curr_migr_unit
) <
8952 __le32_to_cpu(migr_rec
->num_migr_units
)) {
8953 /* current reshape position [blocks] */
8954 unsigned long long current_position
=
8955 __le32_to_cpu(migr_rec
->blocks_per_unit
)
8956 * __le32_to_cpu(migr_rec
->curr_migr_unit
);
8957 unsigned long long border
;
8959 /* Check that array hasn't become failed.
8961 degraded
= check_degradation_change(sra
, fds
, degraded
);
8963 dprintf("imsm: Abort reshape due to degradation"
8964 " level (%i)\n", degraded
);
8968 next_step
= __le32_to_cpu(migr_rec
->blocks_per_unit
);
8970 if ((current_position
+ next_step
) > max_position
)
8971 next_step
= max_position
- current_position
;
8973 start
= current_position
* 512;
8975 /* allign reading start to old geometry */
8976 start_buf_shift
= start
% old_data_stripe_length
;
8977 start_src
= start
- start_buf_shift
;
8979 border
= (start_src
/ odata
) - (start
/ ndata
);
8981 if (border
<= __le32_to_cpu(migr_rec
->dest_depth_per_unit
)) {
8982 /* save critical stripes to buf
8983 * start - start address of current unit
8985 * start_src - start address of current unit
8986 * to backup alligned to source array
8989 unsigned long long next_step_filler
= 0;
8990 unsigned long long copy_length
= next_step
* 512;
8992 /* allign copy area length to stripe in old geometry */
8993 next_step_filler
= ((copy_length
+ start_buf_shift
)
8994 % old_data_stripe_length
);
8995 if (next_step_filler
)
8996 next_step_filler
= (old_data_stripe_length
8997 - next_step_filler
);
8998 dprintf("save_stripes() parameters: start = %llu,"
8999 "\tstart_src = %llu,\tnext_step*512 = %llu,"
9000 "\tstart_in_buf_shift = %llu,"
9001 "\tnext_step_filler = %llu\n",
9002 start
, start_src
, copy_length
,
9003 start_buf_shift
, next_step_filler
);
9005 if (save_stripes(fds
, offsets
, map_src
->num_members
,
9006 chunk
, map_src
->raid_level
,
9007 source_layout
, 0, NULL
, start_src
,
9009 next_step_filler
+ start_buf_shift
,
9011 dprintf("imsm: Cannot save stripes"
9015 /* Convert data to destination format and store it
9016 * in backup general migration area
9018 if (save_backup_imsm(st
, dev
, sra
,
9019 buf
+ start_buf_shift
, copy_length
)) {
9020 dprintf("imsm: Cannot save stripes to "
9021 "target devices\n");
9024 if (save_checkpoint_imsm(st
, sra
,
9025 UNIT_SRC_IN_CP_AREA
)) {
9026 dprintf("imsm: Cannot write checkpoint to "
9027 "migration record (UNIT_SRC_IN_CP_AREA)\n");
9031 /* set next step to use whole border area */
9032 border
/= next_step
;
9034 next_step
*= border
;
9036 /* When data backed up, checkpoint stored,
9037 * kick the kernel to reshape unit of data
9039 next_step
= next_step
+ sra
->reshape_progress
;
9040 /* limit next step to array max position */
9041 if (next_step
> max_position
)
9042 next_step
= max_position
;
9043 sysfs_set_num(sra
, NULL
, "suspend_lo", sra
->reshape_progress
);
9044 sysfs_set_num(sra
, NULL
, "suspend_hi", next_step
);
9045 sra
->reshape_progress
= next_step
;
9047 /* wait until reshape finish */
9048 if (wait_for_reshape_imsm(sra
, ndata
) < 0) {
9049 dprintf("wait_for_reshape_imsm returned error!\n");
9053 if (save_checkpoint_imsm(st
, sra
, UNIT_SRC_NORMAL
) == 1) {
9054 /* ignore error == 2, this can mean end of reshape here
9056 dprintf("imsm: Cannot write checkpoint to "
9057 "migration record (UNIT_SRC_NORMAL)\n");
9063 /* return '1' if done */
9071 #endif /* MDASSEMBLE */
9073 struct superswitch super_imsm
= {
9075 .examine_super
= examine_super_imsm
,
9076 .brief_examine_super
= brief_examine_super_imsm
,
9077 .brief_examine_subarrays
= brief_examine_subarrays_imsm
,
9078 .export_examine_super
= export_examine_super_imsm
,
9079 .detail_super
= detail_super_imsm
,
9080 .brief_detail_super
= brief_detail_super_imsm
,
9081 .write_init_super
= write_init_super_imsm
,
9082 .validate_geometry
= validate_geometry_imsm
,
9083 .add_to_super
= add_to_super_imsm
,
9084 .remove_from_super
= remove_from_super_imsm
,
9085 .detail_platform
= detail_platform_imsm
,
9086 .kill_subarray
= kill_subarray_imsm
,
9087 .update_subarray
= update_subarray_imsm
,
9088 .load_container
= load_container_imsm
,
9089 .default_geometry
= default_geometry_imsm
,
9090 .get_disk_controller_domain
= imsm_get_disk_controller_domain
,
9091 .reshape_super
= imsm_reshape_super
,
9092 .manage_reshape
= imsm_manage_reshape
,
9093 .recover_backup
= recover_backup_imsm
,
9095 .match_home
= match_home_imsm
,
9096 .uuid_from_super
= uuid_from_super_imsm
,
9097 .getinfo_super
= getinfo_super_imsm
,
9098 .getinfo_super_disks
= getinfo_super_disks_imsm
,
9099 .update_super
= update_super_imsm
,
9101 .avail_size
= avail_size_imsm
,
9102 .min_acceptable_spare_size
= min_acceptable_spare_size_imsm
,
9104 .compare_super
= compare_super_imsm
,
9106 .load_super
= load_super_imsm
,
9107 .init_super
= init_super_imsm
,
9108 .store_super
= store_super_imsm
,
9109 .free_super
= free_super_imsm
,
9110 .match_metadata_desc
= match_metadata_desc_imsm
,
9111 .container_content
= container_content_imsm
,
9119 .open_new
= imsm_open_new
,
9120 .set_array_state
= imsm_set_array_state
,
9121 .set_disk
= imsm_set_disk
,
9122 .sync_metadata
= imsm_sync_metadata
,
9123 .activate_spare
= imsm_activate_spare
,
9124 .process_update
= imsm_process_update
,
9125 .prepare_update
= imsm_prepare_update
,
9126 #endif /* MDASSEMBLE */