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_CHUNK 512
68 #define DEFAULT_BITMAP_CHUNK 4096
69 #define DEFAULT_BITMAP_DELAY 5
70 #define DEFAULT_MAX_WRITE_BEHIND 256
72 /* MAP_DIR should be somewhere that persists across the pivotroot
73 * from early boot to late boot.
74 * Currently /dev seems to be the only option on most distros.
77 #define MAP_DIR "/dev/.mdadm"
79 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
80 * want something other than the default of "map"
83 #define MAP_FILE "map"
85 /* MDMON_DIR is where pid and socket files used for communicating
86 * with mdmon normally live. It *should* be /var/run, but when
87 * mdmon is needed at early boot then it needs to write there prior
88 * to /var/run being mounted read/write, and it also then needs to
89 * persist beyond when /var/run is mounter read-only. So, to be
90 * safe, the default is somewhere that is read/write early in the
91 * boot process and stays up as long as possible during shutdown.
94 #define MDMON_DIR "/dev/.mdadm/"
95 #endif /* MDMON_DIR */
97 /* FAILED_SLOTS is where to save files storing recent removal of array
98 * member in order to allow future reuse of disk inserted in the same
99 * slot for array recovery
101 #ifndef FAILED_SLOTS_DIR
102 #define FAILED_SLOTS_DIR "/dev/.mdadm/failed-slots"
103 #endif /* FAILED_SLOTS */
111 /* Redhat don't like to #include <asm/byteorder.h>, and
112 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
113 * and there is no standard conversion function so... */
114 /* And dietlibc doesn't think byteswap is ok, so.. */
115 /* #include <byteswap.h> */
116 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
117 ((x) & 0xff00U) >> 8)
118 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
119 ((x) & 0xff000000U) >> 24 | \
120 ((x) & 0x0000ff00U) << 8 | \
121 ((x) & 0x00ff0000U) >> 8)
122 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
123 ((x) & 0xff00000000000000ULL) >> 56 | \
124 ((x) & 0x000000000000ff00ULL) << 40 | \
125 ((x) & 0x00ff000000000000ULL) >> 40 | \
126 ((x) & 0x0000000000ff0000ULL) << 24 | \
127 ((x) & 0x0000ff0000000000ULL) >> 24 | \
128 ((x) & 0x00000000ff000000ULL) << 8 | \
129 ((x) & 0x000000ff00000000ULL) >> 8)
131 #if !defined(__KLIBC__)
132 #if BYTE_ORDER == LITTLE_ENDIAN
133 #define __cpu_to_le16(_x) (_x)
134 #define __cpu_to_le32(_x) (_x)
135 #define __cpu_to_le64(_x) (_x)
136 #define __le16_to_cpu(_x) (_x)
137 #define __le32_to_cpu(_x) (_x)
138 #define __le64_to_cpu(_x) (_x)
140 #define __cpu_to_be16(_x) bswap_16(_x)
141 #define __cpu_to_be32(_x) bswap_32(_x)
142 #define __cpu_to_be64(_x) bswap_64(_x)
143 #define __be16_to_cpu(_x) bswap_16(_x)
144 #define __be32_to_cpu(_x) bswap_32(_x)
145 #define __be64_to_cpu(_x) bswap_64(_x)
146 #elif BYTE_ORDER == BIG_ENDIAN
147 #define __cpu_to_le16(_x) bswap_16(_x)
148 #define __cpu_to_le32(_x) bswap_32(_x)
149 #define __cpu_to_le64(_x) bswap_64(_x)
150 #define __le16_to_cpu(_x) bswap_16(_x)
151 #define __le32_to_cpu(_x) bswap_32(_x)
152 #define __le64_to_cpu(_x) bswap_64(_x)
154 #define __cpu_to_be16(_x) (_x)
155 #define __cpu_to_be32(_x) (_x)
156 #define __cpu_to_be64(_x) (_x)
157 #define __be16_to_cpu(_x) (_x)
158 #define __be32_to_cpu(_x) (_x)
159 #define __be64_to_cpu(_x) (_x)
161 # error "unknown endianness."
163 #endif /* __KLIBC__ */
167 * min()/max()/clamp() macros that also do
168 * strict type-checking.. See the
169 * "unnecessary" pointer comparison.
171 #define min(x, y) ({ \
172 typeof(x) _min1 = (x); \
173 typeof(y) _min2 = (y); \
174 (void) (&_min1 == &_min2); \
175 _min1 < _min2 ? _min1 : _min2; })
177 #define max(x, y) ({ \
178 typeof(x) _max1 = (x); \
179 typeof(y) _max2 = (y); \
180 (void) (&_max1 == &_max2); \
181 _max1 > _max2 ? _max1 : _max2; })
183 /* general information that might be extracted from a superblock */
185 mdu_array_info_t array
;
186 mdu_disk_info_t disk
;
190 unsigned long long data_offset
;
191 unsigned long long component_size
; /* same as array.size, except in
192 * sectors and up to 64bits.
194 unsigned long long custom_array_size
; /* size for non-default sized
195 * arrays (in sectors)
198 #define VOLUME_RESHAPE 1
199 #define CONTAINER_RESHAPE 2
201 unsigned long long reshape_progress
;
202 int recovery_blocked
; /* for external metadata it
203 * indicates that there is
204 * reshape in progress in
206 * for native metadata it is
207 * reshape_active field mirror
210 unsigned long long resync_start
; /* per-array resync position */
211 unsigned long long recovery_start
; /* per-device rebuild position */
212 #define MaxSector (~0ULL) /* resync/recovery complete position */
214 long bitmap_offset
; /* 0 == none, 1 == a file */
215 unsigned long safe_mode_delay
; /* ms delay to mark clean */
216 int new_level
, delta_disks
, new_layout
, new_chunk
;
218 unsigned long cache_size
; /* size of raid456 stripe cache*/
220 char text_version
[50];
222 int container_member
; /* for assembling external-metatdata arrays
223 * This is to be used internally by metadata
225 int container_enough
; /* flag external handlers can set to
226 * indicate that subarrays have not enough (-1),
227 * enough to start (0), or all expected disks (1) */
232 /* Device info for mdmon: */
237 #define DS_WRITE_MOSTLY 4
239 #define DS_BLOCKED 16
240 #define DS_REMOVE 1024
241 #define DS_UNBLOCK 2048
242 int prev_state
, curr_state
, next_state
;
252 struct supertype
*supertype
;
269 extern char short_options
[];
270 extern char short_bitmap_options
[];
271 extern char short_bitmap_auto_options
[];
272 extern struct option long_options
[];
273 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
274 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
276 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
278 /* for option that don't have short equivilents, we assign arbitrary
279 * numbers later than any 'short' character option.
281 enum special_options
{
327 /* structures read from config file */
328 /* List of mddevice names and identifiers
329 * Identifiers can be:
331 * super-minor=decimal-minor-number-from-superblock
332 * devices=comma,separated,list,of,device,names,with,wildcards
334 * If multiple fields are present, the intersection of all matching
335 * devices is considered
337 #define UnSet (0xfffe)
347 char *devices
; /* comma separated list of device
348 * names with wild cards
353 struct supertype
*st
;
354 int autof
; /* 1 for normal, 2 for partitioned */
359 char *container
; /* /dev/whatever name of container, or
360 * uuid of container. You would expect
361 * this to be the 'devname' or UUID
362 * of some other entry.
364 char *member
; /* subarray within a container */
366 struct mddev_ident
*next
;
368 /* fields needed by different users of this structure */
369 int assembled
; /* set when assembly succeeds */
373 /* List of device names - wildcards expanded */
376 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail.
377 * Not set for names read from .config
379 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
381 char used
; /* set when used */
382 struct mddev_dev
*next
;
385 typedef struct mapping
{
396 char *pattern
; /* U or up, _ for down */
397 int percent
; /* -1 if no resync */
398 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
401 char * metadata_version
;
404 struct dev_member
*next
;
406 struct mdstat_ent
*next
;
409 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
410 extern void free_mdstat(struct mdstat_ent
*ms
);
411 extern void mdstat_wait(int seconds
);
412 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
413 extern int mddev_busy(int devnum
);
414 extern struct mdstat_ent
*mdstat_by_component(char *name
);
415 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, int container
);
418 struct map_ent
*next
;
425 extern int map_update(struct map_ent
**mpp
, int devnum
, char *metadata
,
426 int uuid
[4], char *path
);
427 extern void map_remove(struct map_ent
**map
, int devnum
);
428 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
429 extern struct map_ent
*map_by_devnum(struct map_ent
**map
, int devnum
);
430 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
431 extern void map_read(struct map_ent
**melp
);
432 extern int map_write(struct map_ent
*mel
);
433 extern void map_delete(struct map_ent
**mapp
, int devnum
);
434 extern void map_free(struct map_ent
*map
);
435 extern void map_add(struct map_ent
**melp
,
436 int devnum
, char *metadata
, int uuid
[4], char *path
);
437 extern int map_lock(struct map_ent
**melp
);
438 extern void map_unlock(struct map_ent
**melp
);
439 extern void map_fork(void);
441 /* various details can be requested */
442 enum sysfs_read_flags
{
443 GET_LEVEL
= (1 << 0),
444 GET_LAYOUT
= (1 << 1),
445 GET_COMPONENT
= (1 << 2),
446 GET_CHUNK
= (1 << 3),
447 GET_CACHE
= (1 << 4),
448 GET_MISMATCH
= (1 << 5),
449 GET_VERSION
= (1 << 6),
450 GET_DISKS
= (1 << 7),
451 GET_DEGRADED
= (1 << 8),
452 GET_SAFEMODE
= (1 << 9),
453 GET_BITMAP_LOCATION
= (1 << 10),
455 GET_DEVS
= (1 << 20), /* gets role, major, minor */
456 GET_OFFSET
= (1 << 21),
457 GET_SIZE
= (1 << 22),
458 GET_STATE
= (1 << 23),
459 GET_ERROR
= (1 << 24),
462 /* If fd >= 0, get the array it is open on,
463 * else use devnum. >=0 -> major9. <0.....
465 extern int sysfs_open(int devnum
, char *devname
, char *attr
);
466 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
);
467 extern void sysfs_free(struct mdinfo
*sra
);
468 extern struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
);
469 extern int sysfs_attr_match(const char *attr
, const char *str
);
470 extern int sysfs_match_word(const char *word
, char **list
);
471 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
472 char *name
, char *val
);
473 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
474 char *name
, unsigned long long val
);
475 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
476 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
478 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
479 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
480 char *name
, unsigned long long *val
);
481 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
482 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
484 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
485 char *name
, char *val
, int size
);
486 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
487 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
488 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
489 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
490 extern int sysfs_unique_holder(int devnum
, long rdev
);
491 extern int sysfs_freeze_array(struct mdinfo
*sra
);
492 extern int load_sys(char *path
, char *buf
);
493 extern int reshape_prepare_fdlist(char *devname
,
497 unsigned long blocks
,
500 unsigned long long *offsets
);
501 extern void reshape_free_fdlist(int *fdlist
,
502 unsigned long long *offsets
,
504 extern int reshape_open_backup_file(char *backup
,
509 unsigned long long *offsets
,
511 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
512 unsigned int ndata
, unsigned int odata
);
514 extern int save_stripes(int *source
, unsigned long long *offsets
,
515 int raid_disks
, int chunk_size
, int level
, int layout
,
516 int nwrites
, int *dest
,
517 unsigned long long start
, unsigned long long length
,
519 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
520 int raid_disks
, int chunk_size
, int level
, int layout
,
521 int source
, unsigned long long read_offset
,
522 unsigned long long start
, unsigned long long length
,
526 #define Sendmail "/usr/lib/sendmail -t"
529 #define SYSLOG_FACILITY LOG_DAEMON
531 extern char *map_num(mapping_t
*map
, int num
);
532 extern int map_name(mapping_t
*map
, char *name
);
533 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
535 extern char *map_dev(int major
, int minor
, int create
);
538 struct metadata_update
;
541 /* 'struct reshape' records the intermediate states
543 * The starting geometry is converted to the 'before' geometry
544 * by at most an atomic level change. They could be the same.
545 * Similarly the 'after' geometry is converted to the final
546 * geometry by at most a level change.
547 * Note that 'before' and 'after' must have the same level.
548 * 'blocks' is the minimum number of sectors for a reshape unit.
549 * This will be a multiple of the stripe size in each of the
550 * 'before' and 'after' geometries.
551 * If 'blocks' is 0, no restriping is necessary.
555 int parity
; /* number of parity blocks/devices */
560 unsigned long long backup_blocks
;
561 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
562 unsigned long long new_size
; /* New size of array in sectors */
565 /* A superswitch provides entry point the a metadata handler.
567 * The super_switch primarily operates on some "metadata" that
568 * is accessed via the 'supertype'.
569 * This metadata has one of three possible sources.
570 * 1/ It is read from a single device. In this case it may not completely
571 * describe the array or arrays as some information might be on other
573 * 2/ It is read from all devices in a container. In this case all
574 * information is present.
575 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
576 * be complete once enough ->add_to_super calls have completed.
578 * When creating an array inside a container, the metadata will be
579 * formed by a combination of 2 and 3. The metadata or the array is read,
580 * then new information is added.
582 * The metadata must sometimes have a concept of a 'current' array
583 * and a 'current' device.
584 * The 'current' array is set by init_super to be the newly created array,
585 * or is set by super_by_fd when it finds it is looking at an array inside
588 * The 'current' device is either the device that the metadata was read from
589 * in case 1, or the last device added by add_to_super in case 3.
590 * Case 2 does not identify a 'current' device.
592 extern struct superswitch
{
594 /* Used to report details of metadata read from a component
595 * device. ->load_super has been called.
597 void (*examine_super
)(struct supertype
*st
, char *homehost
);
598 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
599 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
600 void (*export_examine_super
)(struct supertype
*st
);
602 /* Used to report details of an active array.
603 * ->load_super was possibly given a 'component' string.
605 void (*detail_super
)(struct supertype
*st
, char *homehost
);
606 void (*brief_detail_super
)(struct supertype
*st
);
607 void (*export_detail_super
)(struct supertype
*st
);
609 /* Optional: platform hardware / firmware details */
610 int (*detail_platform
)(int verbose
, int enumerate_only
);
613 * to get uuid to storing in bitmap metadata
614 * and 'reshape' backup-data metadata
615 * To see if a device is being re-added to an array it was part of.
617 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
619 /* Extract generic details from metadata. This could be details about
620 * the container, or about an individual array within the container.
621 * The determination is made either by:
622 * load_super being given a 'component' string.
623 * validate_geometry determining what to create.
624 * The info includes both array information and device information.
625 * The particular device should be:
626 * The last device added by add_to_super
627 * The device the metadata was loaded from by load_super
628 * If 'map' is present, then it is an array raid_disks long
629 * (raid_disk must already be set and correct) and it is filled
630 * with 1 for slots that are thought to be active and 0 for slots which
631 * appear to be failed/missing.
632 * *info is zeroed out before data is added.
634 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
635 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
636 /* Check if the given metadata is flagged as belonging to "this"
637 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
639 int (*match_home
)(struct supertype
*st
, char *homehost
);
641 /* Make one of several generic modifications to metadata
642 * prior to assembly (or other times).
643 * sparc2.2 - first bug in early 0.90 metadata
644 * super-minor - change name of 0.90 metadata
645 * summaries - 'correct' any redundant data
646 * resync - mark array as dirty to trigger a resync.
647 * uuid - set new uuid - only 0.90 or 1.x
648 * name - change the name of the array (where supported)
649 * homehost - change which host this array is tied to.
650 * devicesize - If metadata is at start of device, change recorded
651 * device size to match actual device size
652 * byteorder - swap bytes for 0.90 metadata
654 * force-one - mark that device as uptodate, not old or failed.
655 * force-array - mark array as clean if it would not otherwise
657 * assemble - not sure how this is different from force-one...
658 * linear-grow-new - add a new device to a linear array, but don't
659 * change the size: so superblock still matches
660 * linear-grow-update - now change the size of the array.
661 * writemostly - set the WriteMostly1 bit in the superblock devflags
662 * readwrite - clear the WriteMostly1 bit in the superblock devflags
664 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
666 char *devname
, int verbose
,
667 int uuid_set
, char *homehost
);
669 /* Create new metadata for new array as described. This could
670 * be a new container, or an array in a pre-existing container.
671 * Also used to zero metadata prior to writing it to invalidate old
674 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
675 unsigned long long size
, char *name
,
676 char *homehost
, int *uuid
);
678 /* update the metadata to include new device, either at create or
679 * when hot-adding a spare.
681 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
682 int fd
, char *devname
);
683 /* update the metadata to delete a device,
686 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
688 /* Write metadata to one device when fixing problems or adding
691 int (*store_super
)(struct supertype
*st
, int fd
);
693 /* Write all metadata for this array.
695 int (*write_init_super
)(struct supertype
*st
);
696 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
697 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
698 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
699 struct supertype
* (*match_metadata_desc
)(char *arg
);
700 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
);
701 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
702 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
703 int delay
, int write_behind
,
704 unsigned long long size
, int may_change
, int major
);
705 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
706 int (*write_bitmap
)(struct supertype
*st
, int fd
);
707 void (*free_super
)(struct supertype
*st
);
709 /* validate_geometry is called with an st returned by
710 * match_metadata_desc.
711 * It should check that the geometry described in compatible with
712 * the metadata type. It will be called repeatedly as devices
713 * added to validate changing size and new devices. If there are
714 * inter-device dependencies, it should record sufficient details
715 * so these can be validated.
716 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
718 * 1: everything is OK
719 * 0: not OK for some reason - if 'verbose', then error was reported.
720 * -1: st->sb was NULL, 'subdev' is a member of a container of this
721 * types, but array is not acceptable for some reason
722 * message was reported even if verbose is 0.
724 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
726 int *chunk
, unsigned long long size
,
727 char *subdev
, unsigned long long *freesize
,
730 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
731 /* query the supertype for default geometry */
732 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
733 /* Permit subarray's to be deleted from inactive containers */
734 int (*kill_subarray
)(struct supertype
*st
); /* optional */
735 /* Permit subarray's to be modified */
736 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
737 char *update
, struct mddev_ident
*ident
); /* optional */
738 /* Check if reshape is supported for this external format.
739 * st is obtained from super_by_fd() where st->subarray[0] is
740 * initialized to indicate if reshape is being performed at the
741 * container or subarray level
743 #define APPLY_METADATA_CHANGES 1
744 #define ROLLBACK_METADATA_CHANGES 0
746 int (*reshape_super
)(struct supertype
*st
, long long size
, int level
,
747 int layout
, int chunksize
, int raid_disks
,
748 int delta_disks
, char *backup
, char *dev
,
750 int verbose
); /* optional */
751 int (*manage_reshape
)( /* optional */
752 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
753 struct supertype
*st
, unsigned long blocks
,
754 int *fds
, unsigned long long *offsets
,
755 int dests
, int *destfd
, unsigned long long *destoffsets
);
758 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
761 /* Tell the metadata handler the current state of the array.
762 * This covers whether it is known to be consistent (no pending writes)
763 * and how far along a resync is known to have progressed
764 * (in a->resync_start).
765 * resync status is really irrelevant if the array is not consistent,
766 * but some metadata (DDF!) have a place to record the distinction.
767 * If 'consistent' is '2', then the array can mark it dirty if a
768 * resync/recovery/whatever is required, or leave it clean if not.
769 * Return value is 0 dirty (not consistent) and 1 if clean.
770 * it is only really important if consistent is passed in as '2'.
772 int (*set_array_state
)(struct active_array
*a
, int consistent
);
774 /* When the state of a device might have changed, we call set_disk to
775 * tell the metadata what the current state is.
776 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
778 * set_disk might be called when the state of the particular disk has
779 * not in fact changed.
781 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
782 void (*sync_metadata
)(struct supertype
*st
);
783 void (*process_update
)(struct supertype
*st
,
784 struct metadata_update
*update
);
785 void (*prepare_update
)(struct supertype
*st
,
786 struct metadata_update
*update
);
788 /* activate_spare will check if the array is degraded and, if it
789 * is, try to find some spare space in the container.
790 * On success, it add appropriate updates (For process_update) to
791 * to the 'updates' list and returns a list of 'mdinfo' identifying
792 * the device, or devices as there might be multiple missing
793 * devices and multiple spares available.
795 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
796 struct metadata_update
**updates
);
798 * Return statically allocated string that represents metadata specific
799 * controller domain of the disk. The domain is used in disk domain
800 * matching functions. Disks belong to the same domain if the they have
801 * the same domain from mdadm.conf and belong the same metadata domain.
802 * Returning NULL or not providing this handler means that metadata
803 * does not distinguish the differences between disks that belong to
804 * different controllers. They are in the domain specified by
805 * configuration file (mdadm.conf).
806 * In case when the metadata has the notion of domains based on disk
807 * it shall return NULL for disks that do not belong to the controller
808 * the supported domains. Such disks will form another domain and won't
809 * be mixed with supported ones.
811 const char *(*get_disk_controller_domain
)(const char *path
);
813 /* for external backup area */
814 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
816 int swapuuid
; /* true if uuid is bigending rather than hostendian */
818 const char *name
; /* canonical metadata name */
821 extern struct superswitch super0
, super1
;
822 extern struct superswitch super_imsm
, super_ddf
;
823 extern struct superswitch mbr
, gpt
;
825 struct metadata_update
{
828 void *space
; /* allocated space that monitor will use */
829 void **space_list
; /* list of allocated spaces that monitor can
830 * use or that it returned.
832 struct metadata_update
*next
;
835 /* A supertype holds a particular collection of metadata.
836 * It identifies the metadata type by the superswitch, and the particular
837 * sub-version of that metadata type.
838 * metadata read in or created is stored in 'sb' and 'info'.
839 * There are also fields used by mdmon to track containers.
841 * A supertype may refer to:
842 * Just an array, possibly in a container
843 * A container, not identifying any particular array
844 * Info read from just one device, not yet fully describing the array/container.
847 * A supertype is created by:
853 struct superswitch
*ss
;
856 int container_dev
; /* devnum of container */
859 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
860 HW/firmware related incompatability to load metadata.
861 Used when examining metadata to display content of disk
862 when user has no hw/firmare compatible system.
864 struct metadata_update
*updates
;
865 struct metadata_update
**update_tail
;
867 /* extra stuff used by mdmon */
868 struct active_array
*arrays
;
869 int sock
; /* listen to external programs */
871 char *devname
; /* e.g. md0. This appears in metadata_verison:
881 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
882 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
883 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
884 static inline struct supertype
*guess_super(int fd
) {
885 return guess_super_type(fd
, guess_any
);
887 extern struct supertype
*dup_super(struct supertype
*st
);
888 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
889 extern int must_be_container(int fd
);
890 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
891 extern void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
,
892 mdu_disk_info_t
*disk
);
893 void wait_for(char *dev
, int fd
);
896 * Data structures for policy management.
897 * Each device can have a policy structure that lists
898 * various name/value pairs each possibly with a metadata associated.
899 * The policy list is sorted by name/value/metadata
902 struct dev_policy
*next
;
903 char *name
; /* None of these strings are allocated. They are
904 * all just references to strings which are known
905 * to exist elsewhere.
906 * name and metadata can be compared by address equality.
908 const char *metadata
;
912 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
914 /* iterate over the sublist starting at list, having the same
915 * 'name' as 'list', and matching the given metadata (Where
916 * NULL matches anything
918 #define pol_for_each(item, list, _metadata) \
920 item && item->name == list->name; \
922 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
926 * policy records read from mdadm are largely just name-value pairs.
927 * The names are constants, not strdupped
930 struct pol_rule
*next
;
931 char *type
; /* rule_policy or rule_part */
936 char *dups
; /* duplicates of 'value' with a partNN appended */
940 extern char rule_policy
[], rule_part
[];
941 extern char rule_path
[], rule_type
[];
942 extern char type_part
[], type_disk
[];
944 extern void policyline(char *line
, char *type
);
945 extern void policy_add(char *type
, ...);
946 extern void policy_free(void);
948 extern struct dev_policy
*path_policy(char *path
, char *type
);
949 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
950 extern struct dev_policy
*devnum_policy(int dev
);
951 extern void dev_policy_free(struct dev_policy
*p
);
953 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
954 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
955 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
961 act_spare
, /* This only applies to bare devices */
962 act_spare_same_slot
, /* this allows non-bare devices,
963 * but only if recent removal */
964 act_force_spare
, /* this allow non-bare devices in any case */
968 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
969 enum policy_action want
);
970 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
971 enum policy_action want
);
974 struct domainlist
*next
;
978 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
979 const char *metadata
);
980 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
981 const char *metadata
);
982 extern void domainlist_add_dev(struct domainlist
**dom
, int devnum
,
983 const char *metadata
);
984 extern void domain_free(struct domainlist
*dl
);
985 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
986 const char *metadata
);
987 void domain_add(struct domainlist
**domp
, char *domain
);
989 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
990 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
996 #define HAVE_NFTW we assume
1000 # include <features.h>
1001 # ifndef __UCLIBC_HAS_LFS__
1002 # define lseek64 lseek
1004 # ifndef __UCLIBC_HAS_FTW__
1014 #if defined(__KLIBC__)
1030 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1033 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1034 extern int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
);
1035 extern int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
);
1036 extern int Manage_subdevs(char *devname
, int fd
,
1037 struct mddev_dev
*devlist
, int verbose
, int test
,
1038 char *update
, int force
);
1039 extern int autodetect(void);
1040 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1041 extern int Grow_addbitmap(char *devname
, int fd
, char *file
, int chunk
, int delay
, int write_behind
, int force
);
1042 extern int Grow_reshape(char *devname
, int fd
, int quiet
, char *backup_file
,
1044 int level
, char *layout_str
, int chunksize
, int raid_disks
,
1045 struct mddev_dev
*devlist
,
1046 int assume_clean
, int force
);
1047 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1048 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1049 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1050 struct mdinfo
*info
, char *backup_file
,
1051 int freeze_reshape
);
1053 extern int restore_backup(struct supertype
*st
,
1054 struct mdinfo
*content
,
1059 extern int Grow_continue_command(char *devname
, int fd
,
1060 char *backup_file
, int verbose
);
1062 extern int Assemble(struct supertype
*st
, char *mddev
,
1063 struct mddev_ident
*ident
,
1064 struct mddev_dev
*devlist
,
1065 char *backup_file
, int invalid_backup
,
1066 int readonly
, int runstop
,
1067 char *update
, char *homehost
, int require_homehost
,
1068 int verbose
, int force
, int freeze_reshape
);
1070 extern int Build(char *mddev
, int chunk
, int level
, int layout
,
1071 int raiddisks
, struct mddev_dev
*devlist
, int assume_clean
,
1072 char *bitmap_file
, int bitmap_chunk
, int write_behind
,
1073 int delay
, int verbose
, int autof
, unsigned long long size
);
1076 extern int Create(struct supertype
*st
, char *mddev
,
1077 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
1078 char *name
, char *homehost
, int *uuid
,
1079 int subdevs
, struct mddev_dev
*devlist
,
1080 int runstop
, int verbose
, int force
, int assume_clean
,
1081 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
);
1083 extern int Detail(char *dev
, int brief
, int export
, int test
, char *homehost
);
1084 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
);
1085 extern int Query(char *dev
);
1086 extern int Examine(struct mddev_dev
*devlist
, int brief
, int export
, int scan
,
1087 int SparcAdjust
, struct supertype
*forcest
, char *homehost
);
1088 extern int Monitor(struct mddev_dev
*devlist
,
1089 char *mailaddr
, char *alert_cmd
,
1090 int period
, int daemonise
, int scan
, int oneshot
,
1091 int dosyslog
, int test
, char *pidfile
, int increments
,
1094 extern int Kill(char *dev
, struct supertype
*st
, int force
, int quiet
, int noexcl
);
1095 extern int Kill_subarray(char *dev
, char *subarray
, int quiet
);
1096 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1097 extern int Wait(char *dev
);
1098 extern int WaitClean(char *dev
, int sock
, int verbose
);
1100 extern int Incremental(char *devname
, int verbose
, int runstop
,
1101 struct supertype
*st
, char *homehost
, int require_homehost
,
1102 int autof
, int freeze_reshape
);
1103 extern void RebuildMap(void);
1104 extern int IncrementalScan(int verbose
);
1105 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1106 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1107 unsigned long chunksize
, unsigned long daemon_sleep
,
1108 unsigned long write_behind
,
1109 unsigned long long array_size
,
1111 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1112 extern int Write_rules(char *rule_name
);
1113 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1114 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
1116 extern int md_get_version(int fd
);
1117 extern int get_linux_version(void);
1118 extern int mdadm_version(char *version
);
1119 extern long long parse_size(char *size
);
1120 extern int parse_uuid(char *str
, int uuid
[4]);
1121 extern int parse_layout_10(char *layout
);
1122 extern int parse_layout_faulty(char *layout
);
1123 extern int check_ext2(int fd
, char *name
);
1124 extern int check_reiser(int fd
, char *name
);
1125 extern int check_raid(int fd
, char *name
);
1126 extern int check_partitions(int fd
, char *dname
,
1127 unsigned long long freesize
,
1128 unsigned long long size
);
1130 extern int get_mdp_major(void);
1131 extern int dev_open(char *dev
, int flags
);
1132 extern int open_dev(int devnum
);
1133 extern int open_dev_flags(int devnum
, int flags
);
1134 extern int open_dev_excl(int devnum
);
1135 extern int is_standard(char *dev
, int *nump
);
1136 extern int same_dev(char *one
, char *two
);
1138 extern int parse_auto(char *str
, char *msg
, int config
);
1139 extern struct mddev_ident
*conf_get_ident(char *dev
);
1140 extern struct mddev_dev
*conf_get_devs(void);
1141 extern int conf_test_dev(char *devname
);
1142 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1143 extern struct createinfo
*conf_get_create_info(void);
1144 extern void set_conffile(char *file
);
1145 extern char *conf_get_mailaddr(void);
1146 extern char *conf_get_mailfrom(void);
1147 extern char *conf_get_program(void);
1148 extern char *conf_get_homehost(int *require_homehostp
);
1149 extern char *conf_line(FILE *file
);
1150 extern char *conf_word(FILE *file
, int allow_key
);
1151 extern int conf_name_is_free(char *name
);
1152 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1153 extern int devname_matches(char *name
, char *match
);
1154 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1155 struct mdinfo
*info
,
1157 int verbose
, int *rvp
);
1158 extern int experimental(void);
1160 extern void free_line(char *line
);
1161 extern int match_oneof(char *devices
, char *devname
);
1162 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1163 extern const int uuid_zero
[4];
1164 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1165 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1166 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1167 extern char *fname_from_uuid(struct supertype
*st
,
1168 struct mdinfo
*info
, char *buf
, char sep
);
1169 extern unsigned long calc_csum(void *super
, int bytes
);
1170 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1172 extern int enough_fd(int fd
);
1173 extern int ask(char *mesg
);
1174 extern unsigned long long get_component_size(int fd
);
1175 extern void remove_partitions(int fd
);
1176 extern int test_partition(int fd
);
1177 extern int test_partition_from_id(dev_t id
);
1178 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1179 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1180 int chunksize
, unsigned long long devsize
);
1181 extern int flush_metadata_updates(struct supertype
*st
);
1182 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1183 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1184 struct mdinfo
*content
, int runstop
,
1185 char *chosen_name
, int verbose
,
1186 char *backup_file
, int freeze_reshape
);
1187 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1188 unsigned long long min_size
,
1189 struct domainlist
*domlist
,
1191 const char *metadata
, int get_one
);
1192 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1193 extern int add_disk(int mdfd
, struct supertype
*st
,
1194 struct mdinfo
*sra
, struct mdinfo
*info
);
1195 extern int remove_disk(int mdfd
, struct supertype
*st
,
1196 struct mdinfo
*sra
, struct mdinfo
*info
);
1197 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1198 unsigned long long min_recovery_start(struct mdinfo
*array
);
1200 extern char *human_size(long long bytes
);
1201 extern char *human_size_brief(long long bytes
);
1202 extern void print_r10_layout(int layout
);
1204 #define NoMdDev (1<<23)
1205 extern int find_free_devnum(int use_partitions
);
1207 extern void put_md_name(char *name
);
1208 extern char *get_md_name(int dev
);
1210 extern char DefaultConfFile
[];
1212 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1214 /* values for 'trustworthy' */
1216 #define LOCAL_ANY 10
1219 extern int open_mddev(char *dev
, int report_errors
);
1220 extern int open_container(int fd
);
1221 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1222 extern int is_subarray_active(char *subarray
, char *devname
);
1223 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1224 extern struct superswitch
*version_to_superswitch(char *vers
);
1226 extern int mdmon_running(int devnum
);
1227 extern int mdmon_pid(int devnum
);
1228 extern int check_env(char *name
);
1229 extern __u32
random32(void);
1230 extern int start_mdmon(int devnum
);
1232 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1233 struct supertype
*st
, unsigned long stripes
,
1234 int *fds
, unsigned long long *offsets
,
1235 int dests
, int *destfd
, unsigned long long *destoffsets
);
1236 void abort_reshape(struct mdinfo
*sra
);
1238 extern char *devnum2devname(int num
);
1239 extern void fmt_devname(char *name
, int num
);
1240 extern int devname2devnum(char *name
);
1241 extern int stat2devnum(struct stat
*st
);
1242 extern int fd2devnum(int fd
);
1244 static inline int dev2major(int d
)
1249 return get_mdp_major();
1252 static inline int dev2minor(int d
)
1256 return (-1-d
) << MdpMinorShift
;
1259 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1260 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1261 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1262 (ROUND_UP((unsigned long)(ptr), base)))
1264 static inline int is_subarray(char *vers
)
1266 /* The version string for a 'subarray' (an array in a container)
1268 * /containername/componentname for normal read-write arrays
1269 * -containername/componentname for arrays which mdmon must not
1270 * reconfigure. They might be read-only
1271 * or might be undergoing reshape etc.
1272 * containername is e.g. md0, md_d1
1273 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1275 return (*vers
== '/' || *vers
== '-');
1278 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1280 return &ent
->metadata_version
[10+strlen(container
)+1];
1284 #define dprintf(fmt, arg...) \
1285 fprintf(stderr, fmt, ##arg)
1287 #define dprintf(fmt, arg...) \
1288 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1292 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1296 ret
= vasprintf(strp
, fmt
, ap
);
1302 #define LEVEL_MULTIPATH (-4)
1303 #define LEVEL_LINEAR (-1)
1304 #define LEVEL_FAULTY (-5)
1306 /* kernel module doesn't know about these */
1307 #define LEVEL_CONTAINER (-100)
1308 #define LEVEL_UNSUPPORTED (-200)
1313 #define WriteTransient 0
1314 #define ReadTransient 1
1315 #define WritePersistent 2
1316 #define ReadPersistent 3
1317 #define WriteAll 4 /* doesn't go to device */
1318 #define ReadFixable 5
1321 #define ClearErrors 31
1322 #define ClearFaults 30
1324 #define AllPersist 100 /* internal use only */
1325 #define NoPersist 101
1327 #define ModeMask 0x1f
1335 #define minor(x) ((x)&0xff)
1336 #define major(x) (((x)>>8)&0xff)
1337 #define makedev(M,m) (((M)<<8) | (m))
1341 #define ALGORITHM_LEFT_ASYMMETRIC 0
1342 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1343 #define ALGORITHM_LEFT_SYMMETRIC 2
1344 #define ALGORITHM_RIGHT_SYMMETRIC 3
1346 /* Define non-rotating (raid4) algorithms. These allow
1347 * conversion of raid4 to raid5.
1349 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1350 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1352 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1353 * Firstly, the exact positioning of the parity block is slightly
1354 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1356 * Secondly, or order of datablocks over which the Q syndrome is computed
1358 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1359 * These layouts are from the DDFv1.2 spec.
1360 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1361 * leaves RLQ=3 as 'Vendor Specific'
1364 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1365 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1366 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1369 /* For every RAID5 algorithm we define a RAID6 algorithm
1370 * with exactly the same layout for data and parity, and
1371 * with the Q block always on the last device (N-1).
1372 * This allows trivial conversion from RAID5 to RAID6
1374 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1375 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1376 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1377 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1378 #define ALGORITHM_PARITY_0_6 20
1379 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1381 /* Define PATH_MAX in case we don't use glibc or standard library does
1382 * not have PATH_MAX defined. Assume max path length is 4K characters.
1385 #define PATH_MAX 4096
1388 #define PROCESS_DELAYED -2
1389 #define PROCESS_PENDING -3
1391 extern int __offroot
;