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 unsigned long long reshape_progress
;
200 unsigned long long resync_start
; /* per-array resync position */
201 unsigned long long recovery_start
; /* per-device rebuild position */
202 #define MaxSector (~0ULL) /* resync/recovery complete position */
204 unsigned long safe_mode_delay
; /* ms delay to mark clean */
205 int new_level
, delta_disks
, new_layout
, new_chunk
;
207 unsigned long cache_size
; /* size of raid456 stripe cache*/
209 char text_version
[50];
210 void *update_private
; /* for passing metadata-format
211 * specific update data
212 * between successive calls to
216 int container_member
; /* for assembling external-metatdata arrays
217 * This is to be used internally by metadata
219 int container_enough
; /* flag external handlers can set to
220 * indicate that subarrays have not enough (-1),
221 * enough to start (0), or all expected disks (1) */
226 /* Device info for mdmon: */
231 #define DS_WRITE_MOSTLY 4
233 #define DS_BLOCKED 16
234 #define DS_REMOVE 1024
235 #define DS_UNBLOCK 2048
236 int prev_state
, curr_state
, next_state
;
246 struct supertype
*supertype
;
263 extern char short_options
[];
264 extern char short_bitmap_options
[];
265 extern char short_bitmap_auto_options
[];
266 extern struct option long_options
[];
267 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
268 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
270 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
272 /* for option that don't have short equivilents, we assign arbitrary
273 * numbers later than any 'short' character option.
275 enum special_options
{
318 /* structures read from config file */
319 /* List of mddevice names and identifiers
320 * Identifiers can be:
322 * super-minor=decimal-minor-number-from-superblock
323 * devices=comma,separated,list,of,device,names,with,wildcards
325 * If multiple fields are present, the intersection of all matching
326 * devices is considered
328 #define UnSet (0xfffe)
338 char *devices
; /* comma separated list of device
339 * names with wild cards
344 struct supertype
*st
;
345 int autof
; /* 1 for normal, 2 for partitioned */
350 char *container
; /* /dev/whatever name of container, or
351 * uuid of container. You would expect
352 * this to be the 'devname' or UUID
353 * of some other entry.
355 char *member
; /* subarray within a container */
357 struct mddev_ident
*next
;
359 /* fields needed by different users of this structure */
360 int assembled
; /* set when assembly succeeds */
364 /* List of device names - wildcards expanded */
367 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail.
368 * Not set for names read from .config
370 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
372 char used
; /* set when used */
373 struct mddev_dev
*next
;
376 typedef struct mapping
{
387 char *pattern
; /* U or up, _ for down */
388 int percent
; /* -1 if no resync */
389 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
392 char * metadata_version
;
395 struct dev_member
*next
;
397 struct mdstat_ent
*next
;
400 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
401 extern void free_mdstat(struct mdstat_ent
*ms
);
402 extern void mdstat_wait(int seconds
);
403 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
404 extern int mddev_busy(int devnum
);
405 extern struct mdstat_ent
*mdstat_by_component(char *name
);
406 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, int container
);
409 struct map_ent
*next
;
416 extern int map_update(struct map_ent
**mpp
, int devnum
, char *metadata
,
417 int uuid
[4], char *path
);
418 extern void map_remove(struct map_ent
**map
, int devnum
);
419 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
420 extern struct map_ent
*map_by_devnum(struct map_ent
**map
, int devnum
);
421 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
422 extern void map_read(struct map_ent
**melp
);
423 extern int map_write(struct map_ent
*mel
);
424 extern void map_delete(struct map_ent
**mapp
, int devnum
);
425 extern void map_free(struct map_ent
*map
);
426 extern void map_add(struct map_ent
**melp
,
427 int devnum
, char *metadata
, int uuid
[4], char *path
);
428 extern int map_lock(struct map_ent
**melp
);
429 extern void map_unlock(struct map_ent
**melp
);
431 /* various details can be requested */
432 enum sysfs_read_flags
{
433 GET_LEVEL
= (1 << 0),
434 GET_LAYOUT
= (1 << 1),
435 GET_COMPONENT
= (1 << 2),
436 GET_CHUNK
= (1 << 3),
437 GET_CACHE
= (1 << 4),
438 GET_MISMATCH
= (1 << 5),
439 GET_VERSION
= (1 << 6),
440 GET_DISKS
= (1 << 7),
441 GET_DEGRADED
= (1 << 8),
442 GET_SAFEMODE
= (1 << 9),
443 GET_DEVS
= (1 << 10), /* gets role, major, minor */
444 GET_OFFSET
= (1 << 11),
445 GET_SIZE
= (1 << 12),
446 GET_STATE
= (1 << 13),
447 GET_ERROR
= (1 << 14),
450 /* If fd >= 0, get the array it is open on,
451 * else use devnum. >=0 -> major9. <0.....
453 extern int sysfs_open(int devnum
, char *devname
, char *attr
);
454 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
);
455 extern void sysfs_free(struct mdinfo
*sra
);
456 extern struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
);
457 extern int sysfs_attr_match(const char *attr
, const char *str
);
458 extern int sysfs_match_word(const char *word
, char **list
);
459 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
460 char *name
, char *val
);
461 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
462 char *name
, unsigned long long val
);
463 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
464 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
466 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
467 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
468 char *name
, unsigned long long *val
);
469 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
470 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
472 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
473 char *name
, char *val
, int size
);
474 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
475 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
476 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
477 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
478 extern int sysfs_unique_holder(int devnum
, long rdev
);
479 extern int sysfs_freeze_array(struct mdinfo
*sra
);
480 extern int load_sys(char *path
, char *buf
);
481 extern int reshape_prepare_fdlist(char *devname
,
485 unsigned long blocks
,
488 unsigned long long *offsets
);
489 extern void reshape_free_fdlist(int *fdlist
,
490 unsigned long long *offsets
,
492 extern int reshape_open_backup_file(char *backup
,
497 unsigned long long *offsets
,
499 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
500 unsigned int ndata
, unsigned int odata
);
502 extern int save_stripes(int *source
, unsigned long long *offsets
,
503 int raid_disks
, int chunk_size
, int level
, int layout
,
504 int nwrites
, int *dest
,
505 unsigned long long start
, unsigned long long length
,
507 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
508 int raid_disks
, int chunk_size
, int level
, int layout
,
509 int source
, unsigned long long read_offset
,
510 unsigned long long start
, unsigned long long length
,
514 #define Sendmail "/usr/lib/sendmail -t"
517 #define SYSLOG_FACILITY LOG_DAEMON
519 extern char *map_num(mapping_t
*map
, int num
);
520 extern int map_name(mapping_t
*map
, char *name
);
521 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
523 extern char *map_dev(int major
, int minor
, int create
);
526 struct metadata_update
;
529 /* 'struct reshape' records the intermediate states
531 * The starting geometry is converted to the 'before' geometry
532 * by at most an atomic level change. They could be the same.
533 * Similarly the 'after' geometry is converted to the final
534 * geometry by at most a level change.
535 * Note that 'before' and 'after' must have the same level.
536 * 'blocks' is the minimum number of sectors for a reshape unit.
537 * This will be a multiple of the stripe size in each of the
538 * 'before' and 'after' geometries.
539 * If 'blocks' is 0, no restriping is necessary.
543 int parity
; /* number of parity blocks/devices */
548 unsigned long long backup_blocks
;
549 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
550 unsigned long long new_size
; /* New size of array in sectors */
553 /* A superswitch provides entry point the a metadata handler.
555 * The super_switch primarily operates on some "metadata" that
556 * is accessed via the 'supertype'.
557 * This metadata has one of three possible sources.
558 * 1/ It is read from a single device. In this case it may not completely
559 * describe the array or arrays as some information might be on other
561 * 2/ It is read from all devices in a container. In this case all
562 * information is present.
563 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
564 * be complete once enough ->add_to_super calls have completed.
566 * When creating an array inside a container, the metadata will be
567 * formed by a combination of 2 and 3. The metadata or the array is read,
568 * then new information is added.
570 * The metadata must sometimes have a concept of a 'current' array
571 * and a 'current' device.
572 * The 'current' array is set by init_super to be the newly created array,
573 * or is set by super_by_fd when it finds it is looking at an array inside
576 * The 'current' device is either the device that the metadata was read from
577 * in case 1, or the last device added by add_to_super in case 3.
578 * Case 2 does not identify a 'current' device.
580 extern struct superswitch
{
582 /* Used to report details of metadata read from a component
583 * device. ->load_super has been called.
585 void (*examine_super
)(struct supertype
*st
, char *homehost
);
586 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
587 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
588 void (*export_examine_super
)(struct supertype
*st
);
590 /* Used to report details of an active array.
591 * ->load_super was possibly given a 'component' string.
593 void (*detail_super
)(struct supertype
*st
, char *homehost
);
594 void (*brief_detail_super
)(struct supertype
*st
);
595 void (*export_detail_super
)(struct supertype
*st
);
597 /* Optional: platform hardware / firmware details */
598 int (*detail_platform
)(int verbose
, int enumerate_only
);
601 * to get uuid to storing in bitmap metadata
602 * and 'reshape' backup-data metadata
603 * To see if a device is being re-added to an array it was part of.
605 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
607 /* Extract generic details from metadata. This could be details about
608 * the container, or about an individual array within the container.
609 * The determination is made either by:
610 * load_super being given a 'component' string.
611 * validate_geometry determining what to create.
612 * The info includes both array information and device information.
613 * The particular device should be:
614 * The last device added by add_to_super
615 * The device the metadata was loaded from by load_super
616 * If 'map' is present, then it is an array raid_disks long
617 * (raid_disk must already be set and correct) and it is filled
618 * with 1 for slots that are thought to be active and 0 for slots which
619 * appear to be failed/missing.
620 * *info is zeroed out before data is added.
622 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
623 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
624 /* Check if the given metadata is flagged as belonging to "this"
625 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
627 int (*match_home
)(struct supertype
*st
, char *homehost
);
629 /* Make one of several generic modifications to metadata
630 * prior to assembly (or other times).
631 * sparc2.2 - first bug in early 0.90 metadata
632 * super-minor - change name of 0.90 metadata
633 * summaries - 'correct' any redundant data
634 * resync - mark array as dirty to trigger a resync.
635 * uuid - set new uuid - only 0.90 or 1.x
636 * name - change the name of the array (where supported)
637 * homehost - change which host this array is tied to.
638 * devicesize - If metadata is at start of device, change recorded
639 * device size to match actual device size
640 * byteorder - swap bytes for 0.90 metadata
642 * force-one - mark that device as uptodate, not old or failed.
643 * force-array - mark array as clean if it would not otherwise
645 * assemble - not sure how this is different from force-one...
646 * linear-grow-new - add a new device to a linear array, but don't
647 * change the size: so superblock still matches
648 * linear-grow-update - now change the size of the array.
649 * writemostly - set the WriteMostly1 bit in the superblock devflags
650 * readwrite - clear the WriteMostly1 bit in the superblock devflags
652 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
654 char *devname
, int verbose
,
655 int uuid_set
, char *homehost
);
657 /* Create new metadata for new array as described. This could
658 * be a new container, or an array in a pre-existing container.
659 * Also used to zero metadata prior to writing it to invalidate old
662 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
663 unsigned long long size
, char *name
,
664 char *homehost
, int *uuid
);
666 /* update the metadata to include new device, either at create or
667 * when hot-adding a spare.
669 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
670 int fd
, char *devname
);
671 /* update the metadata to delete a device,
674 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
676 /* Write metadata to one device when fixing problems or adding
679 int (*store_super
)(struct supertype
*st
, int fd
);
681 /* Write all metadata for this array.
683 int (*write_init_super
)(struct supertype
*st
);
684 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
685 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
686 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
687 struct supertype
* (*match_metadata_desc
)(char *arg
);
688 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
);
689 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
690 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
691 int delay
, int write_behind
,
692 unsigned long long size
, int may_change
, int major
);
693 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
694 int (*write_bitmap
)(struct supertype
*st
, int fd
);
695 void (*free_super
)(struct supertype
*st
);
697 /* validate_geometry is called with an st returned by
698 * match_metadata_desc.
699 * It should check that the geometry described in compatible with
700 * the metadata type. It will be called repeatedly as devices
701 * added to validate changing size and new devices. If there are
702 * inter-device dependencies, it should record sufficient details
703 * so these can be validated.
704 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
706 * 1: everything is OK
707 * 0: not OK for some reason - if 'verbose', then error was reported.
708 * -1: st->sb was NULL, 'subdev' is a member of a container of this
709 * types, but array is not acceptable for some reason
710 * message was reported even if verbose is 0.
712 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
714 int *chunk
, unsigned long long size
,
715 char *subdev
, unsigned long long *freesize
,
718 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
719 /* query the supertype for default geometry */
720 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
721 /* Permit subarray's to be deleted from inactive containers */
722 int (*kill_subarray
)(struct supertype
*st
); /* optional */
723 /* Permit subarray's to be modified */
724 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
725 char *update
, struct mddev_ident
*ident
); /* optional */
726 /* Check if reshape is supported for this external format.
727 * st is obtained from super_by_fd() where st->subarray[0] is
728 * initialized to indicate if reshape is being performed at the
729 * container or subarray level
731 int (*reshape_super
)(struct supertype
*st
, long long size
, int level
,
732 int layout
, int chunksize
, int raid_disks
,
733 int delta_disks
, char *backup
, char *dev
,
734 int verbose
); /* optional */
735 int (*manage_reshape
)( /* optional */
736 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
737 struct supertype
*st
, unsigned long blocks
,
738 int *fds
, unsigned long long *offsets
,
739 int dests
, int *destfd
, unsigned long long *destoffsets
);
742 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
745 /* Tell the metadata handler the current state of the array.
746 * This covers whether it is known to be consistent (no pending writes)
747 * and how far along a resync is known to have progressed
748 * (in a->resync_start).
749 * resync status is really irrelevant if the array is not consistent,
750 * but some metadata (DDF!) have a place to record the distinction.
751 * If 'consistent' is '2', then the array can mark it dirty if a
752 * resync/recovery/whatever is required, or leave it clean if not.
753 * Return value is 0 dirty (not consistent) and 1 if clean.
754 * it is only really important if consistent is passed in as '2'.
756 int (*set_array_state
)(struct active_array
*a
, int consistent
);
758 /* When the state of a device might have changed, we call set_disk to
759 * tell the metadata what the current state is.
760 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
762 * set_disk might be called when the state of the particular disk has
763 * not in fact changed.
765 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
766 void (*sync_metadata
)(struct supertype
*st
);
767 void (*process_update
)(struct supertype
*st
,
768 struct metadata_update
*update
);
769 void (*prepare_update
)(struct supertype
*st
,
770 struct metadata_update
*update
);
772 /* activate_spare will check if the array is degraded and, if it
773 * is, try to find some spare space in the container.
774 * On success, it add appropriate updates (For process_update) to
775 * to the 'updates' list and returns a list of 'mdinfo' identifying
776 * the device, or devices as there might be multiple missing
777 * devices and multiple spares available.
779 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
780 struct metadata_update
**updates
);
782 * Return statically allocated string that represents metadata specific
783 * controller domain of the disk. The domain is used in disk domain
784 * matching functions. Disks belong to the same domain if the they have
785 * the same domain from mdadm.conf and belong the same metadata domain.
786 * Returning NULL or not providing this handler means that metadata
787 * does not distinguish the differences between disks that belong to
788 * different controllers. They are in the domain specified by
789 * configuration file (mdadm.conf).
790 * In case when the metadata has the notion of domains based on disk
791 * it shall return NULL for disks that do not belong to the controller
792 * the supported domains. Such disks will form another domain and won't
793 * be mixed with supported ones.
795 const char *(*get_disk_controller_domain
)(const char *path
);
797 /* for external backup area */
798 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
800 int swapuuid
; /* true if uuid is bigending rather than hostendian */
802 const char *name
; /* canonical metadata name */
805 extern struct superswitch super0
, super1
;
806 extern struct superswitch super_imsm
, super_ddf
;
807 extern struct superswitch mbr
, gpt
;
809 struct metadata_update
{
812 void *space
; /* allocated space that monitor will use */
813 void **space_list
; /* list of allocated spaces that monitor can
814 * use or that it returned.
816 struct metadata_update
*next
;
819 /* A supertype holds a particular collection of metadata.
820 * It identifies the metadata type by the superswitch, and the particular
821 * sub-version of that metadata type.
822 * metadata read in or created is stored in 'sb' and 'info'.
823 * There are also fields used by mdmon to track containers.
825 * A supertype may refer to:
826 * Just an array, possibly in a container
827 * A container, not identifying any particular array
828 * Info read from just one device, not yet fully describing the array/container.
831 * A supertype is created by:
837 struct superswitch
*ss
;
840 int container_dev
; /* devnum of container */
843 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
844 HW/firmware related incompatability to load metadata.
845 Used when examining metadata to display content of disk
846 when user has no hw/firmare compatible system.
848 struct metadata_update
*updates
;
849 struct metadata_update
**update_tail
;
851 /* extra stuff used by mdmon */
852 struct active_array
*arrays
;
853 int sock
; /* listen to external programs */
855 char *devname
; /* e.g. md0. This appears in metadata_verison:
864 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
865 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
866 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
867 static inline struct supertype
*guess_super(int fd
) {
868 return guess_super_type(fd
, guess_any
);
870 extern struct supertype
*dup_super(struct supertype
*st
);
871 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
872 extern int must_be_container(int fd
);
873 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
874 extern void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
,
875 mdu_disk_info_t
*disk
);
876 void wait_for(char *dev
, int fd
);
879 * Data structures for policy management.
880 * Each device can have a policy structure that lists
881 * various name/value pairs each possibly with a metadata associated.
882 * The policy list is sorted by name/value/metadata
885 struct dev_policy
*next
;
886 char *name
; /* None of these strings are allocated. They are
887 * all just references to strings which are known
888 * to exist elsewhere.
889 * name and metadata can be compared by address equality.
891 const char *metadata
;
895 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
897 /* iterate over the sublist starting at list, having the same
898 * 'name' as 'list', and matching the given metadata (Where
899 * NULL matches anything
901 #define pol_for_each(item, list, _metadata) \
903 item && item->name == list->name; \
905 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
909 * policy records read from mdadm are largely just name-value pairs.
910 * The names are constants, not strdupped
913 struct pol_rule
*next
;
914 char *type
; /* rule_policy or rule_part */
919 char *dups
; /* duplicates of 'value' with a partNN appended */
923 extern char rule_policy
[], rule_part
[];
924 extern char rule_path
[], rule_type
[];
925 extern char type_part
[], type_disk
[];
927 extern void policyline(char *line
, char *type
);
928 extern void policy_add(char *type
, ...);
929 extern void policy_free(void);
931 extern struct dev_policy
*path_policy(char *path
, char *type
);
932 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
933 extern struct dev_policy
*devnum_policy(int dev
);
934 extern void dev_policy_free(struct dev_policy
*p
);
936 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
937 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
938 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
944 act_spare
, /* This only applies to bare devices */
945 act_spare_same_slot
, /* this allows non-bare devices,
946 * but only if recent removal */
947 act_force_spare
, /* this allow non-bare devices in any case */
951 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
952 enum policy_action want
);
953 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
954 enum policy_action want
);
957 struct domainlist
*next
;
961 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
962 const char *metadata
);
963 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
964 const char *metadata
);
965 extern void domainlist_add_dev(struct domainlist
**dom
, int devnum
,
966 const char *metadata
);
967 extern void domain_free(struct domainlist
*dl
);
968 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
969 const char *metadata
);
970 void domain_add(struct domainlist
**domp
, char *domain
);
972 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
973 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
979 #define HAVE_NFTW we assume
983 # include <features.h>
984 # ifndef __UCLIBC_HAS_LFS__
985 # define lseek64 lseek
987 # ifndef __UCLIBC_HAS_FTW__
997 #if defined(__KLIBC__)
1013 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1016 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1017 extern int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
);
1018 extern int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
);
1019 extern int Manage_subdevs(char *devname
, int fd
,
1020 struct mddev_dev
*devlist
, int verbose
, int test
,
1021 char *update
, int force
);
1022 extern int autodetect(void);
1023 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1024 extern int Grow_addbitmap(char *devname
, int fd
, char *file
, int chunk
, int delay
, int write_behind
, int force
);
1025 extern int Grow_reshape(char *devname
, int fd
, int quiet
, char *backup_file
,
1027 int level
, char *layout_str
, int chunksize
, int raid_disks
,
1028 struct mddev_dev
*devlist
,
1029 int assume_clean
, int force
);
1030 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1031 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1032 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1033 struct mdinfo
*info
, char *backup_file
);
1034 extern int restore_backup(struct supertype
*st
,
1035 struct mdinfo
*content
,
1041 extern int Assemble(struct supertype
*st
, char *mddev
,
1042 struct mddev_ident
*ident
,
1043 struct mddev_dev
*devlist
,
1044 char *backup_file
, int invalid_backup
,
1045 int readonly
, int runstop
,
1046 char *update
, char *homehost
, int require_homehost
,
1047 int verbose
, int force
);
1049 extern int Build(char *mddev
, int chunk
, int level
, int layout
,
1050 int raiddisks
, struct mddev_dev
*devlist
, int assume_clean
,
1051 char *bitmap_file
, int bitmap_chunk
, int write_behind
,
1052 int delay
, int verbose
, int autof
, unsigned long long size
);
1055 extern int Create(struct supertype
*st
, char *mddev
,
1056 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
1057 char *name
, char *homehost
, int *uuid
,
1058 int subdevs
, struct mddev_dev
*devlist
,
1059 int runstop
, int verbose
, int force
, int assume_clean
,
1060 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
);
1062 extern int Detail(char *dev
, int brief
, int export
, int test
, char *homehost
);
1063 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
);
1064 extern int Query(char *dev
);
1065 extern int Examine(struct mddev_dev
*devlist
, int brief
, int export
, int scan
,
1066 int SparcAdjust
, struct supertype
*forcest
, char *homehost
);
1067 extern int Monitor(struct mddev_dev
*devlist
,
1068 char *mailaddr
, char *alert_cmd
,
1069 int period
, int daemonise
, int scan
, int oneshot
,
1070 int dosyslog
, int test
, char *pidfile
, int increments
,
1073 extern int Kill(char *dev
, struct supertype
*st
, int force
, int quiet
, int noexcl
);
1074 extern int Kill_subarray(char *dev
, char *subarray
, int quiet
);
1075 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1076 extern int Wait(char *dev
);
1077 extern int WaitClean(char *dev
, int sock
, int verbose
);
1079 extern int Incremental(char *devname
, int verbose
, int runstop
,
1080 struct supertype
*st
, char *homehost
, int require_homehost
,
1082 extern void RebuildMap(void);
1083 extern int IncrementalScan(int verbose
);
1084 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1085 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1086 unsigned long chunksize
, unsigned long daemon_sleep
,
1087 unsigned long write_behind
,
1088 unsigned long long array_size
,
1090 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1091 extern int Write_rules(char *rule_name
);
1092 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1093 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
1095 extern int md_get_version(int fd
);
1096 extern int get_linux_version(void);
1097 extern int mdadm_version(char *version
);
1098 extern long long parse_size(char *size
);
1099 extern int parse_uuid(char *str
, int uuid
[4]);
1100 extern int parse_layout_10(char *layout
);
1101 extern int parse_layout_faulty(char *layout
);
1102 extern int check_ext2(int fd
, char *name
);
1103 extern int check_reiser(int fd
, char *name
);
1104 extern int check_raid(int fd
, char *name
);
1105 extern int check_partitions(int fd
, char *dname
,
1106 unsigned long long freesize
,
1107 unsigned long long size
);
1109 extern int get_mdp_major(void);
1110 extern int dev_open(char *dev
, int flags
);
1111 extern int open_dev(int devnum
);
1112 extern int open_dev_flags(int devnum
, int flags
);
1113 extern int open_dev_excl(int devnum
);
1114 extern int is_standard(char *dev
, int *nump
);
1115 extern int same_dev(char *one
, char *two
);
1117 extern int parse_auto(char *str
, char *msg
, int config
);
1118 extern struct mddev_ident
*conf_get_ident(char *dev
);
1119 extern struct mddev_dev
*conf_get_devs(void);
1120 extern int conf_test_dev(char *devname
);
1121 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1122 extern struct createinfo
*conf_get_create_info(void);
1123 extern void set_conffile(char *file
);
1124 extern char *conf_get_mailaddr(void);
1125 extern char *conf_get_mailfrom(void);
1126 extern char *conf_get_program(void);
1127 extern char *conf_get_homehost(int *require_homehostp
);
1128 extern char *conf_line(FILE *file
);
1129 extern char *conf_word(FILE *file
, int allow_key
);
1130 extern int conf_name_is_free(char *name
);
1131 extern int devname_matches(char *name
, char *match
);
1132 extern struct mddev_ident
*conf_match(struct mdinfo
*info
, struct supertype
*st
);
1133 extern int experimental(void);
1135 extern void free_line(char *line
);
1136 extern int match_oneof(char *devices
, char *devname
);
1137 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1138 extern const int uuid_zero
[4];
1139 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1140 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1141 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1142 extern char *fname_from_uuid(struct supertype
*st
,
1143 struct mdinfo
*info
, char *buf
, char sep
);
1144 extern unsigned long calc_csum(void *super
, int bytes
);
1145 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1146 char *avail
, int avail_disks
);
1147 extern int enough_fd(int fd
);
1148 extern int ask(char *mesg
);
1149 extern unsigned long long get_component_size(int fd
);
1150 extern void remove_partitions(int fd
);
1151 extern int test_partition(int fd
);
1152 extern int test_partition_from_id(dev_t id
);
1153 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1154 int chunksize
, unsigned long long devsize
);
1155 extern int flush_metadata_updates(struct supertype
*st
);
1156 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1157 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1158 struct mdinfo
*content
, int runstop
,
1159 char *chosen_name
, int verbose
,
1161 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1162 unsigned long long min_size
,
1163 struct domainlist
*domlist
,
1165 const char *metadata
, int get_one
);
1166 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1167 extern int add_disk(int mdfd
, struct supertype
*st
,
1168 struct mdinfo
*sra
, struct mdinfo
*info
);
1169 extern int remove_disk(int mdfd
, struct supertype
*st
,
1170 struct mdinfo
*sra
, struct mdinfo
*info
);
1171 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1172 unsigned long long min_recovery_start(struct mdinfo
*array
);
1174 extern char *human_size(long long bytes
);
1175 extern char *human_size_brief(long long bytes
);
1176 extern void print_r10_layout(int layout
);
1178 #define NoMdDev (1<<23)
1179 extern int find_free_devnum(int use_partitions
);
1181 extern void put_md_name(char *name
);
1182 extern char *get_md_name(int dev
);
1184 extern char DefaultConfFile
[];
1186 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1188 /* values for 'trustworthy' */
1190 #define LOCAL_ANY 10
1193 extern int open_mddev(char *dev
, int report_errors
);
1194 extern int open_container(int fd
);
1195 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1196 extern int is_subarray_active(char *subarray
, char *devname
);
1197 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1198 extern struct superswitch
*version_to_superswitch(char *vers
);
1200 extern int mdmon_running(int devnum
);
1201 extern int mdmon_pid(int devnum
);
1202 extern int check_env(char *name
);
1203 extern __u32
random32(void);
1204 extern int start_mdmon(int devnum
);
1206 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1207 struct supertype
*st
, unsigned long stripes
,
1208 int *fds
, unsigned long long *offsets
,
1209 int dests
, int *destfd
, unsigned long long *destoffsets
);
1210 void abort_reshape(struct mdinfo
*sra
);
1212 extern char *devnum2devname(int num
);
1213 extern void fmt_devname(char *name
, int num
);
1214 extern int devname2devnum(char *name
);
1215 extern int stat2devnum(struct stat
*st
);
1216 extern int fd2devnum(int fd
);
1218 static inline int dev2major(int d
)
1223 return get_mdp_major();
1226 static inline int dev2minor(int d
)
1230 return (-1-d
) << MdpMinorShift
;
1233 static inline int ROUND_UP(int a
, int base
)
1235 return ((a
+base
-1)/base
)*base
;
1238 static inline int is_subarray(char *vers
)
1240 /* The version string for a 'subarray' (an array in a container)
1242 * /containername/componentname for normal read-write arrays
1243 * -containername/componentname for arrays which mdmon must not
1244 * reconfigure. They might be read-only
1245 * or might be undergoing reshape etc.
1246 * containername is e.g. md0, md_d1
1247 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1249 return (*vers
== '/' || *vers
== '-');
1252 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1254 return &ent
->metadata_version
[10+strlen(container
)+1];
1258 #define dprintf(fmt, arg...) \
1259 fprintf(stderr, fmt, ##arg)
1261 #define dprintf(fmt, arg...) \
1262 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1266 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1270 ret
= vasprintf(strp
, fmt
, ap
);
1276 #define LEVEL_MULTIPATH (-4)
1277 #define LEVEL_LINEAR (-1)
1278 #define LEVEL_FAULTY (-5)
1280 /* kernel module doesn't know about these */
1281 #define LEVEL_CONTAINER (-100)
1282 #define LEVEL_UNSUPPORTED (-200)
1287 #define WriteTransient 0
1288 #define ReadTransient 1
1289 #define WritePersistent 2
1290 #define ReadPersistent 3
1291 #define WriteAll 4 /* doesn't go to device */
1292 #define ReadFixable 5
1295 #define ClearErrors 31
1296 #define ClearFaults 30
1298 #define AllPersist 100 /* internal use only */
1299 #define NoPersist 101
1301 #define ModeMask 0x1f
1309 #define minor(x) ((x)&0xff)
1310 #define major(x) (((x)>>8)&0xff)
1311 #define makedev(M,m) (((M)<<8) | (m))
1315 #define ALGORITHM_LEFT_ASYMMETRIC 0
1316 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1317 #define ALGORITHM_LEFT_SYMMETRIC 2
1318 #define ALGORITHM_RIGHT_SYMMETRIC 3
1320 /* Define non-rotating (raid4) algorithms. These allow
1321 * conversion of raid4 to raid5.
1323 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1324 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1326 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1327 * Firstly, the exact positioning of the parity block is slightly
1328 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1330 * Secondly, or order of datablocks over which the Q syndrome is computed
1332 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1333 * These layouts are from the DDFv1.2 spec.
1334 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1335 * leaves RLQ=3 as 'Vendor Specific'
1338 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1339 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1340 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1343 /* For every RAID5 algorithm we define a RAID6 algorithm
1344 * with exactly the same layout for data and parity, and
1345 * with the Q block always on the last device (N-1).
1346 * This allows trivial conversion from RAID5 to RAID6
1348 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1349 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1350 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1351 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1352 #define ALGORITHM_PARITY_0_6 20
1353 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1355 /* Define PATH_MAX in case we don't use glibc or standard library does
1356 * not have PATH_MAX defined. Assume max path length is 4K characters.
1359 #define PATH_MAX 4096
1362 #define PROCESS_DELAYED -2
1363 #define PROCESS_PENDING -3