2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
26 #define _FILE_OFFSET_BITS 64
28 #if !defined(__dietlibc__) && !defined(__KLIBC__)
29 extern __off64_t lseek64
__P ((int __fd
, __off64_t __offset
, int __whence
));
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
36 #include <sys/types.h>
49 /* dietlibc has deprecated random and srandom!! */
55 #include <linux/kdev_t.h>
56 /*#include <linux/fs.h> */
57 #include <sys/mount.h>
58 #include <asm/types.h>
59 #include <sys/ioctl.h>
61 #define MdpMinorShift 6
64 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
67 #define DEFAULT_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
71 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * Currently /dev seems to be the only option on most distros.
76 #define MAP_DIR "/dev/.mdadm"
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
82 #define MAP_FILE "map"
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. It *should* be /var/run, but when
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
93 #define MDMON_DIR "/dev/.mdadm/"
94 #endif /* MDMON_DIR */
96 /* FAILED_SLOTS is where to save files storing recent removal of array
97 * member in order to allow future reuse of disk inserted in the same
98 * slot for array recovery
100 #ifndef FAILED_SLOTS_DIR
101 #define FAILED_SLOTS_DIR "/dev/.mdadm/failed-slots"
102 #endif /* FAILED_SLOTS */
110 /* Redhat don't like to #include <asm/byteorder.h>, and
111 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
112 * and there is no standard conversion function so... */
113 /* And dietlibc doesn't think byteswap is ok, so.. */
114 /* #include <byteswap.h> */
115 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
116 ((x) & 0xff00U) >> 8)
117 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
118 ((x) & 0xff000000U) >> 24 | \
119 ((x) & 0x0000ff00U) << 8 | \
120 ((x) & 0x00ff0000U) >> 8)
121 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
122 ((x) & 0xff00000000000000ULL) >> 56 | \
123 ((x) & 0x000000000000ff00ULL) << 40 | \
124 ((x) & 0x00ff000000000000ULL) >> 40 | \
125 ((x) & 0x0000000000ff0000ULL) << 24 | \
126 ((x) & 0x0000ff0000000000ULL) >> 24 | \
127 ((x) & 0x00000000ff000000ULL) << 8 | \
128 ((x) & 0x000000ff00000000ULL) >> 8)
130 #if !defined(__KLIBC__)
131 #if BYTE_ORDER == LITTLE_ENDIAN
132 #define __cpu_to_le16(_x) (_x)
133 #define __cpu_to_le32(_x) (_x)
134 #define __cpu_to_le64(_x) (_x)
135 #define __le16_to_cpu(_x) (_x)
136 #define __le32_to_cpu(_x) (_x)
137 #define __le64_to_cpu(_x) (_x)
139 #define __cpu_to_be16(_x) bswap_16(_x)
140 #define __cpu_to_be32(_x) bswap_32(_x)
141 #define __cpu_to_be64(_x) bswap_64(_x)
142 #define __be16_to_cpu(_x) bswap_16(_x)
143 #define __be32_to_cpu(_x) bswap_32(_x)
144 #define __be64_to_cpu(_x) bswap_64(_x)
145 #elif BYTE_ORDER == BIG_ENDIAN
146 #define __cpu_to_le16(_x) bswap_16(_x)
147 #define __cpu_to_le32(_x) bswap_32(_x)
148 #define __cpu_to_le64(_x) bswap_64(_x)
149 #define __le16_to_cpu(_x) bswap_16(_x)
150 #define __le32_to_cpu(_x) bswap_32(_x)
151 #define __le64_to_cpu(_x) bswap_64(_x)
153 #define __cpu_to_be16(_x) (_x)
154 #define __cpu_to_be32(_x) (_x)
155 #define __cpu_to_be64(_x) (_x)
156 #define __be16_to_cpu(_x) (_x)
157 #define __be32_to_cpu(_x) (_x)
158 #define __be64_to_cpu(_x) (_x)
160 # error "unknown endianness."
162 #endif /* __KLIBC__ */
166 * min()/max()/clamp() macros that also do
167 * strict type-checking.. See the
168 * "unnecessary" pointer comparison.
170 #define min(x, y) ({ \
171 typeof(x) _min1 = (x); \
172 typeof(y) _min2 = (y); \
173 (void) (&_min1 == &_min2); \
174 _min1 < _min2 ? _min1 : _min2; })
176 #define max(x, y) ({ \
177 typeof(x) _max1 = (x); \
178 typeof(y) _max2 = (y); \
179 (void) (&_max1 == &_max2); \
180 _max1 > _max2 ? _max1 : _max2; })
182 /* general information that might be extracted from a superblock */
184 mdu_array_info_t array
;
185 mdu_disk_info_t disk
;
189 unsigned long long data_offset
;
190 unsigned long long component_size
; /* same as array.size, except in
191 * sectors and up to 64bits.
193 unsigned long long custom_array_size
; /* size for non-default sized
194 * arrays (in sectors)
197 unsigned long long reshape_progress
;
199 unsigned long long resync_start
; /* per-array resync position */
200 unsigned long long recovery_start
; /* per-device rebuild position */
201 #define MaxSector (~0ULL) /* resync/recovery complete position */
203 unsigned long safe_mode_delay
; /* ms delay to mark clean */
204 int new_level
, delta_disks
, new_layout
, new_chunk
;
206 unsigned long cache_size
; /* size of raid456 stripe cache*/
208 char text_version
[50];
209 void *update_private
; /* for passing metadata-format
210 * specific update data
211 * between successive calls to
215 int container_member
; /* for assembling external-metatdata arrays
216 * This is to be used internally by metadata
218 int container_enough
; /* flag external handlers can set to
219 * indicate that subarrays have not enough (-1),
220 * enough to start (0), or all expected disks (1) */
225 /* Device info for mdmon: */
230 #define DS_WRITE_MOSTLY 4
232 #define DS_BLOCKED 16
233 #define DS_REMOVE 1024
234 #define DS_UNBLOCK 2048
235 int prev_state
, curr_state
, next_state
;
245 struct supertype
*supertype
;
262 extern char short_options
[];
263 extern char short_bitmap_options
[];
264 extern char short_bitmap_auto_options
[];
265 extern struct option long_options
[];
266 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
267 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
269 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
271 /* for option that don't have short equivilents, we assign arbitrary
272 * numbers later than any 'short' character option.
274 enum special_options
{
317 /* structures read from config file */
318 /* List of mddevice names and identifiers
319 * Identifiers can be:
321 * super-minor=decimal-minor-number-from-superblock
322 * devices=comma,separated,list,of,device,names,with,wildcards
324 * If multiple fields are present, the intersection of all matching
325 * devices is considered
327 #define UnSet (0xfffe)
337 char *devices
; /* comma separated list of device
338 * names with wild cards
343 struct supertype
*st
;
344 int autof
; /* 1 for normal, 2 for partitioned */
349 char *container
; /* /dev/whatever name of container, or
350 * uuid of container. You would expect
351 * this to be the 'devname' or UUID
352 * of some other entry.
354 char *member
; /* subarray within a container */
356 struct mddev_ident
*next
;
358 /* fields needed by different users of this structure */
359 int assembled
; /* set when assembly succeeds */
363 /* List of device names - wildcards expanded */
366 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail.
367 * Not set for names read from .config
369 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
371 char used
; /* set when used */
372 struct mddev_dev
*next
;
375 typedef struct mapping
{
386 char *pattern
; /* U or up, _ for down */
387 int percent
; /* -1 if no resync */
388 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
391 char * metadata_version
;
394 struct dev_member
*next
;
396 struct mdstat_ent
*next
;
399 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
400 extern void free_mdstat(struct mdstat_ent
*ms
);
401 extern void mdstat_wait(int seconds
);
402 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
403 extern int mddev_busy(int devnum
);
404 extern struct mdstat_ent
*mdstat_by_component(char *name
);
405 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, int container
);
408 struct map_ent
*next
;
415 extern int map_update(struct map_ent
**mpp
, int devnum
, char *metadata
,
416 int uuid
[4], char *path
);
417 extern void map_remove(struct map_ent
**map
, int devnum
);
418 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
419 extern struct map_ent
*map_by_devnum(struct map_ent
**map
, int devnum
);
420 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
421 extern void map_read(struct map_ent
**melp
);
422 extern int map_write(struct map_ent
*mel
);
423 extern void map_delete(struct map_ent
**mapp
, int devnum
);
424 extern void map_free(struct map_ent
*map
);
425 extern void map_add(struct map_ent
**melp
,
426 int devnum
, char *metadata
, int uuid
[4], char *path
);
427 extern int map_lock(struct map_ent
**melp
);
428 extern void map_unlock(struct map_ent
**melp
);
430 /* various details can be requested */
431 enum sysfs_read_flags
{
432 GET_LEVEL
= (1 << 0),
433 GET_LAYOUT
= (1 << 1),
434 GET_COMPONENT
= (1 << 2),
435 GET_CHUNK
= (1 << 3),
436 GET_CACHE
= (1 << 4),
437 GET_MISMATCH
= (1 << 5),
438 GET_VERSION
= (1 << 6),
439 GET_DISKS
= (1 << 7),
440 GET_DEGRADED
= (1 << 8),
441 GET_SAFEMODE
= (1 << 9),
442 GET_DEVS
= (1 << 10), /* gets role, major, minor */
443 GET_OFFSET
= (1 << 11),
444 GET_SIZE
= (1 << 12),
445 GET_STATE
= (1 << 13),
446 GET_ERROR
= (1 << 14),
449 /* If fd >= 0, get the array it is open on,
450 * else use devnum. >=0 -> major9. <0.....
452 extern int sysfs_open(int devnum
, char *devname
, char *attr
);
453 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
);
454 extern void sysfs_free(struct mdinfo
*sra
);
455 extern struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
);
456 extern int sysfs_attr_match(const char *attr
, const char *str
);
457 extern int sysfs_match_word(const char *word
, char **list
);
458 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
459 char *name
, char *val
);
460 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
461 char *name
, unsigned long long val
);
462 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
463 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
465 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
466 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
467 char *name
, unsigned long long *val
);
468 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
469 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
471 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
472 char *name
, char *val
, int size
);
473 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
474 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
475 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
476 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
477 extern int sysfs_unique_holder(int devnum
, long rdev
);
478 extern int sysfs_freeze_array(struct mdinfo
*sra
);
479 extern int load_sys(char *path
, char *buf
);
480 extern int reshape_prepare_fdlist(char *devname
,
484 unsigned long blocks
,
487 unsigned long long *offsets
);
488 extern void reshape_free_fdlist(int *fdlist
,
489 unsigned long long *offsets
,
491 extern int reshape_open_backup_file(char *backup
,
496 unsigned long long *offsets
,
498 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
499 unsigned int ndata
, unsigned int odata
);
501 extern int save_stripes(int *source
, unsigned long long *offsets
,
502 int raid_disks
, int chunk_size
, int level
, int layout
,
503 int nwrites
, int *dest
,
504 unsigned long long start
, unsigned long long length
,
506 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
507 int raid_disks
, int chunk_size
, int level
, int layout
,
508 int source
, unsigned long long read_offset
,
509 unsigned long long start
, unsigned long long length
);
512 #define Sendmail "/usr/lib/sendmail -t"
515 #define SYSLOG_FACILITY LOG_DAEMON
517 extern char *map_num(mapping_t
*map
, int num
);
518 extern int map_name(mapping_t
*map
, char *name
);
519 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
521 extern char *map_dev(int major
, int minor
, int create
);
524 struct metadata_update
;
527 /* 'struct reshape' records the intermediate states
529 * The starting geometry is converted to the 'before' geometry
530 * by at most an atomic level change. They could be the same.
531 * Similarly the 'after' geometry is converted to the final
532 * geometry by at most a level change.
533 * Note that 'before' and 'after' must have the same level.
534 * 'blocks' is the minimum number of sectors for a reshape unit.
535 * This will be a multiple of the stripe size in each of the
536 * 'before' and 'after' geometries.
537 * If 'blocks' is 0, no restriping is necessary.
541 int parity
; /* number of parity blocks/devices */
546 unsigned long long backup_blocks
;
547 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
548 unsigned long long new_size
; /* New size of array in sectors */
551 /* A superswitch provides entry point the a metadata handler.
553 * The super_switch primarily operates on some "metadata" that
554 * is accessed via the 'supertype'.
555 * This metadata has one of three possible sources.
556 * 1/ It is read from a single device. In this case it may not completely
557 * describe the array or arrays as some information might be on other
559 * 2/ It is read from all devices in a container. In this case all
560 * information is present.
561 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
562 * be complete once enough ->add_to_super calls have completed.
564 * When creating an array inside a container, the metadata will be
565 * formed by a combination of 2 and 3. The metadata or the array is read,
566 * then new information is added.
568 * The metadata must sometimes have a concept of a 'current' array
569 * and a 'current' device.
570 * The 'current' array is set by init_super to be the newly created array,
571 * or is set by super_by_fd when it finds it is looking at an array inside
574 * The 'current' device is either the device that the metadata was read from
575 * in case 1, or the last device added by add_to_super in case 3.
576 * Case 2 does not identify a 'current' device.
578 extern struct superswitch
{
580 /* Used to report details of metadata read from a component
581 * device. ->load_super has been called.
583 void (*examine_super
)(struct supertype
*st
, char *homehost
);
584 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
585 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
586 void (*export_examine_super
)(struct supertype
*st
);
588 /* Used to report details of an active array.
589 * ->load_super was possibly given a 'component' string.
591 void (*detail_super
)(struct supertype
*st
, char *homehost
);
592 void (*brief_detail_super
)(struct supertype
*st
);
593 void (*export_detail_super
)(struct supertype
*st
);
595 /* Optional: platform hardware / firmware details */
596 int (*detail_platform
)(int verbose
, int enumerate_only
);
599 * to get uuid to storing in bitmap metadata
600 * and 'reshape' backup-data metadata
601 * To see if a device is being re-added to an array it was part of.
603 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
605 /* Extract generic details from metadata. This could be details about
606 * the container, or about an individual array within the container.
607 * The determination is made either by:
608 * load_super being given a 'component' string.
609 * validate_geometry determining what to create.
610 * The info includes both array information and device information.
611 * The particular device should be:
612 * The last device added by add_to_super
613 * The device the metadata was loaded from by load_super
614 * If 'map' is present, then it is an array raid_disks long
615 * (raid_disk must already be set and correct) and it is filled
616 * with 1 for slots that are thought to be active and 0 for slots which
617 * appear to be failed/missing.
619 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
620 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
621 /* Check if the given metadata is flagged as belonging to "this"
622 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
624 int (*match_home
)(struct supertype
*st
, char *homehost
);
626 /* Make one of several generic modifications to metadata
627 * prior to assembly (or other times).
628 * sparc2.2 - first bug in early 0.90 metadata
629 * super-minor - change name of 0.90 metadata
630 * summaries - 'correct' any redundant data
631 * resync - mark array as dirty to trigger a resync.
632 * uuid - set new uuid - only 0.90 or 1.x
633 * name - change the name of the array (where supported)
634 * homehost - change which host this array is tied to.
635 * devicesize - If metadata is at start of device, change recorded
636 * device size to match actual device size
637 * byteorder - swap bytes for 0.90 metadata
639 * force-one - mark that device as uptodate, not old or failed.
640 * force-array - mark array as clean if it would not otherwise
642 * assemble - not sure how this is different from force-one...
643 * linear-grow-new - add a new device to a linear array, but don't
644 * change the size: so superblock still matches
645 * linear-grow-update - now change the size of the array.
647 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
649 char *devname
, int verbose
,
650 int uuid_set
, char *homehost
);
652 /* Create new metadata for new array as described. This could
653 * be a new container, or an array in a pre-existing container.
654 * Also used to zero metadata prior to writing it to invalidate old
657 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
658 unsigned long long size
, char *name
,
659 char *homehost
, int *uuid
);
661 /* update the metadata to include new device, either at create or
662 * when hot-adding a spare.
664 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
665 int fd
, char *devname
);
666 /* update the metadata to delete a device,
669 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
671 /* Write metadata to one device when fixing problems or adding
674 int (*store_super
)(struct supertype
*st
, int fd
);
676 /* Write all metadata for this array.
678 int (*write_init_super
)(struct supertype
*st
);
679 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
680 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
681 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
682 struct supertype
* (*match_metadata_desc
)(char *arg
);
683 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
);
684 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
685 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
686 int delay
, int write_behind
,
687 unsigned long long size
, int may_change
, int major
);
688 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
689 int (*write_bitmap
)(struct supertype
*st
, int fd
);
690 void (*free_super
)(struct supertype
*st
);
692 /* validate_geometry is called with an st returned by
693 * match_metadata_desc.
694 * It should check that the geometry described in compatible with
695 * the metadata type. It will be called repeatedly as devices
696 * added to validate changing size and new devices. If there are
697 * inter-device dependencies, it should record sufficient details
698 * so these can be validated.
699 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
701 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
703 int chunk
, unsigned long long size
,
704 char *subdev
, unsigned long long *freesize
,
707 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
708 /* query the supertype for default geometry */
709 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
710 /* Permit subarray's to be deleted from inactive containers */
711 int (*kill_subarray
)(struct supertype
*st
); /* optional */
712 /* Permit subarray's to be modified */
713 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
714 char *update
, struct mddev_ident
*ident
); /* optional */
715 /* Check if reshape is supported for this external format.
716 * st is obtained from super_by_fd() where st->subarray[0] is
717 * initialized to indicate if reshape is being performed at the
718 * container or subarray level
720 int (*reshape_super
)(struct supertype
*st
, long long size
, int level
,
721 int layout
, int chunksize
, int raid_disks
,
722 char *backup
, char *dev
, int verbose
); /* optional */
723 int (*manage_reshape
)( /* optional */
724 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
725 struct supertype
*st
, unsigned long blocks
,
726 int *fds
, unsigned long long *offsets
,
727 int dests
, int *destfd
, unsigned long long *destoffsets
);
730 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
733 /* Tell the metadata handler the current state of the array.
734 * This covers whether it is known to be consistent (no pending writes)
735 * and how far along a resync is known to have progressed
736 * (in a->resync_start).
737 * resync status is really irrelevant if the array is not consistent,
738 * but some metadata (DDF!) have a place to record the distinction.
739 * If 'consistent' is '2', then the array can mark it dirty if a
740 * resync/recovery/whatever is required, or leave it clean if not.
741 * Return value is 0 dirty (not consistent) and 1 if clean.
742 * it is only really important if consistent is passed in as '2'.
744 int (*set_array_state
)(struct active_array
*a
, int consistent
);
746 /* When the state of a device might have changed, we call set_disk to
747 * tell the metadata what the current state is.
748 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
750 * set_disk might be called when the state of the particular disk has
751 * not in fact changed.
753 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
754 void (*sync_metadata
)(struct supertype
*st
);
755 void (*process_update
)(struct supertype
*st
,
756 struct metadata_update
*update
);
757 void (*prepare_update
)(struct supertype
*st
,
758 struct metadata_update
*update
);
760 /* activate_spare will check if the array is degraded and, if it
761 * is, try to find some spare space in the container.
762 * On success, it add appropriate updates (For process_update) to
763 * to the 'updates' list and returns a list of 'mdinfo' identifying
764 * the device, or devices as there might be multiple missing
765 * devices and multiple spares available.
767 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
768 struct metadata_update
**updates
);
770 * Return statically allocated string that represents metadata specific
771 * controller domain of the disk. The domain is used in disk domain
772 * matching functions. Disks belong to the same domain if the they have
773 * the same domain from mdadm.conf and belong the same metadata domain.
774 * Returning NULL or not providing this handler means that metadata
775 * does not distinguish the differences between disks that belong to
776 * different controllers. They are in the domain specified by
777 * configuration file (mdadm.conf).
778 * In case when the metadata has the notion of domains based on disk
779 * it shall return NULL for disks that do not belong to the controller
780 * the supported domains. Such disks will form another domain and won't
781 * be mixed with supported ones.
783 const char *(*get_disk_controller_domain
)(const char *path
);
785 int swapuuid
; /* true if uuid is bigending rather than hostendian */
787 const char *name
; /* canonical metadata name */
790 extern struct superswitch super0
, super1
;
791 extern struct superswitch super_imsm
, super_ddf
;
792 extern struct superswitch mbr
, gpt
;
794 struct metadata_update
{
797 void *space
; /* allocated space that monitor will use */
798 void **space_list
; /* list of allocated spaces that monitor can
799 * use or that it returned.
801 struct metadata_update
*next
;
804 /* A supertype holds a particular collection of metadata.
805 * It identifies the metadata type by the superswitch, and the particular
806 * sub-version of that metadata type.
807 * metadata read in or created is stored in 'sb' and 'info'.
808 * There are also fields used by mdmon to track containers.
810 * A supertype may refer to:
811 * Just an array, possibly in a container
812 * A container, not identifying any particular array
813 * Info read from just one device, not yet fully describing the array/container.
816 * A supertype is created by:
822 struct superswitch
*ss
;
825 int container_dev
; /* devnum of container */
829 struct metadata_update
*updates
;
830 struct metadata_update
**update_tail
;
832 /* extra stuff used by mdmon */
833 struct active_array
*arrays
;
834 int sock
; /* listen to external programs */
836 char *devname
; /* e.g. md0. This appears in metadata_verison:
845 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
846 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
847 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
848 static inline struct supertype
*guess_super(int fd
) {
849 return guess_super_type(fd
, guess_any
);
851 extern struct supertype
*dup_super(struct supertype
*st
);
852 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
853 extern int must_be_container(int fd
);
854 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
855 extern void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
,
856 mdu_disk_info_t
*disk
);
857 void wait_for(char *dev
, int fd
);
860 * Data structures for policy management.
861 * Each device can have a policy structure that lists
862 * various name/value pairs each possibly with a metadata associated.
863 * The policy list is sorted by name/value/metadata
866 struct dev_policy
*next
;
867 char *name
; /* None of these strings are allocated. They are
868 * all just references to strings which are known
869 * to exist elsewhere.
870 * name and metadata can be compared by address equality.
872 const char *metadata
;
876 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
878 /* iterate over the sublist starting at list, having the same
879 * 'name' as 'list', and matching the given metadata (Where
880 * NULL matches anything
882 #define pol_for_each(item, list, _metadata) \
884 item && item->name == list->name; \
886 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
890 * policy records read from mdadm are largely just name-value pairs.
891 * The names are constants, not strdupped
894 struct pol_rule
*next
;
895 char *type
; /* rule_policy or rule_part */
900 char *dups
; /* duplicates of 'value' with a partNN appended */
904 extern char rule_policy
[], rule_part
[];
905 extern char rule_path
[], rule_type
[];
906 extern char type_part
[], type_disk
[];
908 extern void policyline(char *line
, char *type
);
909 extern void policy_add(char *type
, ...);
910 extern void policy_free(void);
912 extern struct dev_policy
*path_policy(char *path
, char *type
);
913 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
914 extern struct dev_policy
*devnum_policy(int dev
);
915 extern void dev_policy_free(struct dev_policy
*p
);
917 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
918 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
919 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
925 act_spare
, /* This only applies to bare devices */
926 act_spare_same_slot
, /* this allows non-bare devices,
927 * but only if recent removal */
928 act_force_spare
, /* this allow non-bare devices in any case */
932 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
933 enum policy_action want
);
934 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
935 enum policy_action want
);
938 struct domainlist
*next
;
942 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
943 const char *metadata
);
944 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
945 const char *metadata
);
946 extern void domainlist_add_dev(struct domainlist
**dom
, int devnum
,
947 const char *metadata
);
948 extern void domain_free(struct domainlist
*dl
);
949 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
950 const char *metadata
);
951 void domain_add(struct domainlist
**domp
, char *domain
);
953 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
954 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
960 #define HAVE_NFTW we assume
964 # include <features.h>
965 # ifndef __UCLIBC_HAS_LFS__
966 # define lseek64 lseek
968 # ifndef __UCLIBC_HAS_FTW__
978 #if defined(__KLIBC__)
994 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
997 extern int Manage_ro(char *devname
, int fd
, int readonly
);
998 extern int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
);
999 extern int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
);
1000 extern int Manage_subdevs(char *devname
, int fd
,
1001 struct mddev_dev
*devlist
, int verbose
, int test
,
1003 extern int autodetect(void);
1004 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1005 extern int Grow_addbitmap(char *devname
, int fd
, char *file
, int chunk
, int delay
, int write_behind
, int force
);
1006 extern int Grow_reshape(char *devname
, int fd
, int quiet
, char *backup_file
,
1008 int level
, char *layout_str
, int chunksize
, int raid_disks
,
1010 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1011 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1012 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1013 struct mdinfo
*info
, char *backup_file
);
1015 extern int Assemble(struct supertype
*st
, char *mddev
,
1016 struct mddev_ident
*ident
,
1017 struct mddev_dev
*devlist
,
1018 char *backup_file
, int invalid_backup
,
1019 int readonly
, int runstop
,
1020 char *update
, char *homehost
, int require_homehost
,
1021 int verbose
, int force
);
1023 extern int Build(char *mddev
, int chunk
, int level
, int layout
,
1024 int raiddisks
, struct mddev_dev
*devlist
, int assume_clean
,
1025 char *bitmap_file
, int bitmap_chunk
, int write_behind
,
1026 int delay
, int verbose
, int autof
, unsigned long long size
);
1029 extern int Create(struct supertype
*st
, char *mddev
,
1030 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
1031 char *name
, char *homehost
, int *uuid
,
1032 int subdevs
, struct mddev_dev
*devlist
,
1033 int runstop
, int verbose
, int force
, int assume_clean
,
1034 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
);
1036 extern int Detail(char *dev
, int brief
, int export
, int test
, char *homehost
);
1037 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
);
1038 extern int Query(char *dev
);
1039 extern int Examine(struct mddev_dev
*devlist
, int brief
, int export
, int scan
,
1040 int SparcAdjust
, struct supertype
*forcest
, char *homehost
);
1041 extern int Monitor(struct mddev_dev
*devlist
,
1042 char *mailaddr
, char *alert_cmd
,
1043 int period
, int daemonise
, int scan
, int oneshot
,
1044 int dosyslog
, int test
, char *pidfile
, int increments
,
1047 extern int Kill(char *dev
, struct supertype
*st
, int force
, int quiet
, int noexcl
);
1048 extern int Kill_subarray(char *dev
, char *subarray
, int quiet
);
1049 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1050 extern int Wait(char *dev
);
1051 extern int WaitClean(char *dev
, int sock
, int verbose
);
1053 extern int Incremental(char *devname
, int verbose
, int runstop
,
1054 struct supertype
*st
, char *homehost
, int require_homehost
,
1056 extern void RebuildMap(void);
1057 extern int IncrementalScan(int verbose
);
1058 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1059 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1060 unsigned long chunksize
, unsigned long daemon_sleep
,
1061 unsigned long write_behind
,
1062 unsigned long long array_size
,
1064 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1065 extern int Write_rules(char *rule_name
);
1066 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1067 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
1069 extern int md_get_version(int fd
);
1070 extern int get_linux_version(void);
1071 extern int mdadm_version(char *version
);
1072 extern long long parse_size(char *size
);
1073 extern int parse_uuid(char *str
, int uuid
[4]);
1074 extern int parse_layout_10(char *layout
);
1075 extern int parse_layout_faulty(char *layout
);
1076 extern int check_ext2(int fd
, char *name
);
1077 extern int check_reiser(int fd
, char *name
);
1078 extern int check_raid(int fd
, char *name
);
1079 extern int check_partitions(int fd
, char *dname
, unsigned long long freesize
);
1081 extern int get_mdp_major(void);
1082 extern int dev_open(char *dev
, int flags
);
1083 extern int open_dev(int devnum
);
1084 extern int open_dev_excl(int devnum
);
1085 extern int is_standard(char *dev
, int *nump
);
1086 extern int same_dev(char *one
, char *two
);
1088 extern int parse_auto(char *str
, char *msg
, int config
);
1089 extern struct mddev_ident
*conf_get_ident(char *dev
);
1090 extern struct mddev_dev
*conf_get_devs(void);
1091 extern int conf_test_dev(char *devname
);
1092 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1093 extern struct createinfo
*conf_get_create_info(void);
1094 extern void set_conffile(char *file
);
1095 extern char *conf_get_mailaddr(void);
1096 extern char *conf_get_mailfrom(void);
1097 extern char *conf_get_program(void);
1098 extern char *conf_get_homehost(int *require_homehostp
);
1099 extern char *conf_line(FILE *file
);
1100 extern char *conf_word(FILE *file
, int allow_key
);
1101 extern int conf_name_is_free(char *name
);
1102 extern int devname_matches(char *name
, char *match
);
1103 extern struct mddev_ident
*conf_match(struct mdinfo
*info
, struct supertype
*st
);
1104 extern int experimental(void);
1106 extern void free_line(char *line
);
1107 extern int match_oneof(char *devices
, char *devname
);
1108 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1109 extern const int uuid_zero
[4];
1110 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1111 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1112 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1113 extern char *fname_from_uuid(struct supertype
*st
,
1114 struct mdinfo
*info
, char *buf
, char sep
);
1115 extern unsigned long calc_csum(void *super
, int bytes
);
1116 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1117 char *avail
, int avail_disks
);
1118 extern int enough_fd(int fd
);
1119 extern int ask(char *mesg
);
1120 extern unsigned long long get_component_size(int fd
);
1121 extern void remove_partitions(int fd
);
1122 extern int test_partition(int fd
);
1123 extern int test_partition_from_id(dev_t id
);
1124 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1125 int chunksize
, unsigned long long devsize
);
1126 extern int flush_metadata_updates(struct supertype
*st
);
1127 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1128 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1129 struct mdinfo
*content
, int runstop
,
1130 char *chosen_name
, int verbose
);
1131 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1132 unsigned long long min_size
,
1133 struct domainlist
*domlist
,
1135 const char *metadata
, int get_one
);
1136 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1137 extern int add_disk(int mdfd
, struct supertype
*st
,
1138 struct mdinfo
*sra
, struct mdinfo
*info
);
1139 extern int remove_disk(int mdfd
, struct supertype
*st
,
1140 struct mdinfo
*sra
, struct mdinfo
*info
);
1141 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1142 unsigned long long min_recovery_start(struct mdinfo
*array
);
1144 extern char *human_size(long long bytes
);
1145 extern char *human_size_brief(long long bytes
);
1146 extern void print_r10_layout(int layout
);
1148 #define NoMdDev (1<<23)
1149 extern int find_free_devnum(int use_partitions
);
1151 extern void put_md_name(char *name
);
1152 extern char *get_md_name(int dev
);
1154 extern char DefaultConfFile
[];
1156 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1158 /* values for 'trustworthy' */
1160 #define LOCAL_ANY 10
1163 extern int open_mddev(char *dev
, int report_errors
);
1164 extern int open_container(int fd
);
1165 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1166 extern int is_subarray_active(char *subarray
, char *devname
);
1167 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1168 extern struct superswitch
*version_to_superswitch(char *vers
);
1170 extern int mdmon_running(int devnum
);
1171 extern int mdmon_pid(int devnum
);
1172 extern int check_env(char *name
);
1173 extern __u32
random32(void);
1174 extern int start_mdmon(int devnum
);
1176 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1177 struct supertype
*st
, unsigned long stripes
,
1178 int *fds
, unsigned long long *offsets
,
1179 int dests
, int *destfd
, unsigned long long *destoffsets
);
1181 extern char *devnum2devname(int num
);
1182 extern void fmt_devname(char *name
, int num
);
1183 extern int devname2devnum(char *name
);
1184 extern int stat2devnum(struct stat
*st
);
1185 extern int fd2devnum(int fd
);
1187 static inline int dev2major(int d
)
1192 return get_mdp_major();
1195 static inline int dev2minor(int d
)
1199 return (-1-d
) << MdpMinorShift
;
1202 static inline int ROUND_UP(int a
, int base
)
1204 return ((a
+base
-1)/base
)*base
;
1207 static inline int is_subarray(char *vers
)
1209 /* The version string for a 'subarray' (an array in a container)
1211 * /containername/componentname for normal read-write arrays
1212 * -containername/componentname for arrays which mdmon must not
1213 * reconfigure. They might be read-only
1214 * or might be undergoing reshape etc.
1215 * containername is e.g. md0, md_d1
1216 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1218 return (*vers
== '/' || *vers
== '-');
1221 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1223 return &ent
->metadata_version
[10+strlen(container
)+1];
1227 #define dprintf(fmt, arg...) \
1228 fprintf(stderr, fmt, ##arg)
1230 #define dprintf(fmt, arg...) \
1231 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1235 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1239 ret
= vasprintf(strp
, fmt
, ap
);
1245 #define LEVEL_MULTIPATH (-4)
1246 #define LEVEL_LINEAR (-1)
1247 #define LEVEL_FAULTY (-5)
1249 /* kernel module doesn't know about these */
1250 #define LEVEL_CONTAINER (-100)
1251 #define LEVEL_UNSUPPORTED (-200)
1256 #define WriteTransient 0
1257 #define ReadTransient 1
1258 #define WritePersistent 2
1259 #define ReadPersistent 3
1260 #define WriteAll 4 /* doesn't go to device */
1261 #define ReadFixable 5
1264 #define ClearErrors 31
1265 #define ClearFaults 30
1267 #define AllPersist 100 /* internal use only */
1268 #define NoPersist 101
1270 #define ModeMask 0x1f
1278 #define minor(x) ((x)&0xff)
1279 #define major(x) (((x)>>8)&0xff)
1280 #define makedev(M,m) (((M)<<8) | (m))
1284 #define ALGORITHM_LEFT_ASYMMETRIC 0
1285 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1286 #define ALGORITHM_LEFT_SYMMETRIC 2
1287 #define ALGORITHM_RIGHT_SYMMETRIC 3
1289 /* Define non-rotating (raid4) algorithms. These allow
1290 * conversion of raid4 to raid5.
1292 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1293 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1295 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1296 * Firstly, the exact positioning of the parity block is slightly
1297 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1299 * Secondly, or order of datablocks over which the Q syndrome is computed
1301 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1302 * These layouts are from the DDFv1.2 spec.
1303 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1304 * leaves RLQ=3 as 'Vendor Specific'
1307 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1308 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1309 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1312 /* For every RAID5 algorithm we define a RAID6 algorithm
1313 * with exactly the same layout for data and parity, and
1314 * with the Q block always on the last device (N-1).
1315 * This allows trivial conversion from RAID5 to RAID6
1317 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1318 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1319 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1320 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1321 #define ALGORITHM_PARITY_0_6 20
1322 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1324 /* Define PATH_MAX in case we don't use glibc or standard library does
1325 * not have PATH_MAX defined. Assume max path length is 4K characters.
1328 #define PATH_MAX 4096