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 unsigned long safe_mode_delay
; /* ms delay to mark clean */
215 int new_level
, delta_disks
, new_layout
, new_chunk
;
217 unsigned long cache_size
; /* size of raid456 stripe cache*/
219 char text_version
[50];
220 void *update_private
; /* for passing metadata-format
221 * specific update data
222 * between successive calls to
226 int container_member
; /* for assembling external-metatdata arrays
227 * This is to be used internally by metadata
229 int container_enough
; /* flag external handlers can set to
230 * indicate that subarrays have not enough (-1),
231 * enough to start (0), or all expected disks (1) */
236 /* Device info for mdmon: */
241 #define DS_WRITE_MOSTLY 4
243 #define DS_BLOCKED 16
244 #define DS_REMOVE 1024
245 #define DS_UNBLOCK 2048
246 int prev_state
, curr_state
, next_state
;
256 struct supertype
*supertype
;
273 extern char short_options
[];
274 extern char short_bitmap_options
[];
275 extern char short_bitmap_auto_options
[];
276 extern struct option long_options
[];
277 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
278 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
280 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
282 /* for option that don't have short equivilents, we assign arbitrary
283 * numbers later than any 'short' character option.
285 enum special_options
{
330 /* structures read from config file */
331 /* List of mddevice names and identifiers
332 * Identifiers can be:
334 * super-minor=decimal-minor-number-from-superblock
335 * devices=comma,separated,list,of,device,names,with,wildcards
337 * If multiple fields are present, the intersection of all matching
338 * devices is considered
340 #define UnSet (0xfffe)
350 char *devices
; /* comma separated list of device
351 * names with wild cards
356 struct supertype
*st
;
357 int autof
; /* 1 for normal, 2 for partitioned */
362 char *container
; /* /dev/whatever name of container, or
363 * uuid of container. You would expect
364 * this to be the 'devname' or UUID
365 * of some other entry.
367 char *member
; /* subarray within a container */
369 struct mddev_ident
*next
;
371 /* fields needed by different users of this structure */
372 int assembled
; /* set when assembly succeeds */
376 /* List of device names - wildcards expanded */
379 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail.
380 * Not set for names read from .config
382 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
384 char used
; /* set when used */
385 struct mddev_dev
*next
;
388 typedef struct mapping
{
399 char *pattern
; /* U or up, _ for down */
400 int percent
; /* -1 if no resync */
401 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
404 char * metadata_version
;
407 struct dev_member
*next
;
409 struct mdstat_ent
*next
;
412 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
413 extern void free_mdstat(struct mdstat_ent
*ms
);
414 extern void mdstat_wait(int seconds
);
415 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
416 extern int mddev_busy(int devnum
);
417 extern struct mdstat_ent
*mdstat_by_component(char *name
);
418 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, int container
);
421 struct map_ent
*next
;
428 extern int map_update(struct map_ent
**mpp
, int devnum
, char *metadata
,
429 int uuid
[4], char *path
);
430 extern void map_remove(struct map_ent
**map
, int devnum
);
431 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
432 extern struct map_ent
*map_by_devnum(struct map_ent
**map
, int devnum
);
433 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
434 extern void map_read(struct map_ent
**melp
);
435 extern int map_write(struct map_ent
*mel
);
436 extern void map_delete(struct map_ent
**mapp
, int devnum
);
437 extern void map_free(struct map_ent
*map
);
438 extern void map_add(struct map_ent
**melp
,
439 int devnum
, char *metadata
, int uuid
[4], char *path
);
440 extern int map_lock(struct map_ent
**melp
);
441 extern void map_unlock(struct map_ent
**melp
);
442 extern void map_fork(void);
444 /* various details can be requested */
445 enum sysfs_read_flags
{
446 GET_LEVEL
= (1 << 0),
447 GET_LAYOUT
= (1 << 1),
448 GET_COMPONENT
= (1 << 2),
449 GET_CHUNK
= (1 << 3),
450 GET_CACHE
= (1 << 4),
451 GET_MISMATCH
= (1 << 5),
452 GET_VERSION
= (1 << 6),
453 GET_DISKS
= (1 << 7),
454 GET_DEGRADED
= (1 << 8),
455 GET_SAFEMODE
= (1 << 9),
456 GET_DEVS
= (1 << 10), /* gets role, major, minor */
457 GET_OFFSET
= (1 << 11),
458 GET_SIZE
= (1 << 12),
459 GET_STATE
= (1 << 13),
460 GET_ERROR
= (1 << 14),
463 /* If fd >= 0, get the array it is open on,
464 * else use devnum. >=0 -> major9. <0.....
466 extern int sysfs_open(int devnum
, char *devname
, char *attr
);
467 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
);
468 extern void sysfs_free(struct mdinfo
*sra
);
469 extern struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
);
470 extern int sysfs_attr_match(const char *attr
, const char *str
);
471 extern int sysfs_match_word(const char *word
, char **list
);
472 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
473 char *name
, char *val
);
474 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
475 char *name
, unsigned long long val
);
476 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
477 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
479 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
480 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
481 char *name
, unsigned long long *val
);
482 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
483 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
485 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
486 char *name
, char *val
, int size
);
487 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
488 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
489 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
490 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
491 extern int sysfs_unique_holder(int devnum
, long rdev
);
492 extern int sysfs_freeze_array(struct mdinfo
*sra
);
493 extern int load_sys(char *path
, char *buf
);
494 extern int reshape_prepare_fdlist(char *devname
,
498 unsigned long blocks
,
501 unsigned long long *offsets
);
502 extern void reshape_free_fdlist(int *fdlist
,
503 unsigned long long *offsets
,
505 extern int reshape_open_backup_file(char *backup
,
510 unsigned long long *offsets
,
512 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
513 unsigned int ndata
, unsigned int odata
);
515 extern int save_stripes(int *source
, unsigned long long *offsets
,
516 int raid_disks
, int chunk_size
, int level
, int layout
,
517 int nwrites
, int *dest
,
518 unsigned long long start
, unsigned long long length
,
520 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
521 int raid_disks
, int chunk_size
, int level
, int layout
,
522 int source
, unsigned long long read_offset
,
523 unsigned long long start
, unsigned long long length
,
527 #define Sendmail "/usr/lib/sendmail -t"
530 #define SYSLOG_FACILITY LOG_DAEMON
532 extern char *map_num(mapping_t
*map
, int num
);
533 extern int map_name(mapping_t
*map
, char *name
);
534 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
536 extern char *map_dev(int major
, int minor
, int create
);
539 struct metadata_update
;
542 /* 'struct reshape' records the intermediate states
544 * The starting geometry is converted to the 'before' geometry
545 * by at most an atomic level change. They could be the same.
546 * Similarly the 'after' geometry is converted to the final
547 * geometry by at most a level change.
548 * Note that 'before' and 'after' must have the same level.
549 * 'blocks' is the minimum number of sectors for a reshape unit.
550 * This will be a multiple of the stripe size in each of the
551 * 'before' and 'after' geometries.
552 * If 'blocks' is 0, no restriping is necessary.
556 int parity
; /* number of parity blocks/devices */
561 unsigned long long backup_blocks
;
562 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
563 unsigned long long new_size
; /* New size of array in sectors */
566 /* A superswitch provides entry point the a metadata handler.
568 * The super_switch primarily operates on some "metadata" that
569 * is accessed via the 'supertype'.
570 * This metadata has one of three possible sources.
571 * 1/ It is read from a single device. In this case it may not completely
572 * describe the array or arrays as some information might be on other
574 * 2/ It is read from all devices in a container. In this case all
575 * information is present.
576 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
577 * be complete once enough ->add_to_super calls have completed.
579 * When creating an array inside a container, the metadata will be
580 * formed by a combination of 2 and 3. The metadata or the array is read,
581 * then new information is added.
583 * The metadata must sometimes have a concept of a 'current' array
584 * and a 'current' device.
585 * The 'current' array is set by init_super to be the newly created array,
586 * or is set by super_by_fd when it finds it is looking at an array inside
589 * The 'current' device is either the device that the metadata was read from
590 * in case 1, or the last device added by add_to_super in case 3.
591 * Case 2 does not identify a 'current' device.
593 extern struct superswitch
{
595 /* Used to report details of metadata read from a component
596 * device. ->load_super has been called.
598 void (*examine_super
)(struct supertype
*st
, char *homehost
);
599 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
600 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
601 void (*export_examine_super
)(struct supertype
*st
);
603 /* Used to report details of an active array.
604 * ->load_super was possibly given a 'component' string.
606 void (*detail_super
)(struct supertype
*st
, char *homehost
);
607 void (*brief_detail_super
)(struct supertype
*st
);
608 void (*export_detail_super
)(struct supertype
*st
);
610 /* Optional: platform hardware / firmware details */
611 int (*detail_platform
)(int verbose
, int enumerate_only
);
614 * to get uuid to storing in bitmap metadata
615 * and 'reshape' backup-data metadata
616 * To see if a device is being re-added to an array it was part of.
618 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
620 /* Extract generic details from metadata. This could be details about
621 * the container, or about an individual array within the container.
622 * The determination is made either by:
623 * load_super being given a 'component' string.
624 * validate_geometry determining what to create.
625 * The info includes both array information and device information.
626 * The particular device should be:
627 * The last device added by add_to_super
628 * The device the metadata was loaded from by load_super
629 * If 'map' is present, then it is an array raid_disks long
630 * (raid_disk must already be set and correct) and it is filled
631 * with 1 for slots that are thought to be active and 0 for slots which
632 * appear to be failed/missing.
633 * *info is zeroed out before data is added.
635 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
636 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
637 /* Check if the given metadata is flagged as belonging to "this"
638 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
640 int (*match_home
)(struct supertype
*st
, char *homehost
);
642 /* Make one of several generic modifications to metadata
643 * prior to assembly (or other times).
644 * sparc2.2 - first bug in early 0.90 metadata
645 * super-minor - change name of 0.90 metadata
646 * summaries - 'correct' any redundant data
647 * resync - mark array as dirty to trigger a resync.
648 * uuid - set new uuid - only 0.90 or 1.x
649 * name - change the name of the array (where supported)
650 * homehost - change which host this array is tied to.
651 * devicesize - If metadata is at start of device, change recorded
652 * device size to match actual device size
653 * byteorder - swap bytes for 0.90 metadata
655 * force-one - mark that device as uptodate, not old or failed.
656 * force-array - mark array as clean if it would not otherwise
658 * assemble - not sure how this is different from force-one...
659 * linear-grow-new - add a new device to a linear array, but don't
660 * change the size: so superblock still matches
661 * linear-grow-update - now change the size of the array.
662 * writemostly - set the WriteMostly1 bit in the superblock devflags
663 * readwrite - clear the WriteMostly1 bit in the superblock devflags
665 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
667 char *devname
, int verbose
,
668 int uuid_set
, char *homehost
);
670 /* Create new metadata for new array as described. This could
671 * be a new container, or an array in a pre-existing container.
672 * Also used to zero metadata prior to writing it to invalidate old
675 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
676 unsigned long long size
, char *name
,
677 char *homehost
, int *uuid
);
679 /* update the metadata to include new device, either at create or
680 * when hot-adding a spare.
682 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
683 int fd
, char *devname
);
684 /* update the metadata to delete a device,
687 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
689 /* Write metadata to one device when fixing problems or adding
692 int (*store_super
)(struct supertype
*st
, int fd
);
694 /* Write all metadata for this array.
696 int (*write_init_super
)(struct supertype
*st
);
697 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
698 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
699 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
700 struct supertype
* (*match_metadata_desc
)(char *arg
);
701 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
);
702 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
703 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
704 int delay
, int write_behind
,
705 unsigned long long size
, int may_change
, int major
);
706 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
707 int (*write_bitmap
)(struct supertype
*st
, int fd
);
708 void (*free_super
)(struct supertype
*st
);
710 /* validate_geometry is called with an st returned by
711 * match_metadata_desc.
712 * It should check that the geometry described in compatible with
713 * the metadata type. It will be called repeatedly as devices
714 * added to validate changing size and new devices. If there are
715 * inter-device dependencies, it should record sufficient details
716 * so these can be validated.
717 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
719 * 1: everything is OK
720 * 0: not OK for some reason - if 'verbose', then error was reported.
721 * -1: st->sb was NULL, 'subdev' is a member of a container of this
722 * types, but array is not acceptable for some reason
723 * message was reported even if verbose is 0.
725 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
727 int *chunk
, unsigned long long size
,
728 char *subdev
, unsigned long long *freesize
,
731 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
732 /* query the supertype for default geometry */
733 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
734 /* Permit subarray's to be deleted from inactive containers */
735 int (*kill_subarray
)(struct supertype
*st
); /* optional */
736 /* Permit subarray's to be modified */
737 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
738 char *update
, struct mddev_ident
*ident
); /* optional */
739 /* Check if reshape is supported for this external format.
740 * st is obtained from super_by_fd() where st->subarray[0] is
741 * initialized to indicate if reshape is being performed at the
742 * container or subarray level
744 int (*reshape_super
)(struct supertype
*st
, long long size
, int level
,
745 int layout
, int chunksize
, int raid_disks
,
746 int delta_disks
, char *backup
, char *dev
,
747 int verbose
); /* optional */
748 int (*manage_reshape
)( /* optional */
749 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
750 struct supertype
*st
, unsigned long blocks
,
751 int *fds
, unsigned long long *offsets
,
752 int dests
, int *destfd
, unsigned long long *destoffsets
);
755 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
758 /* Tell the metadata handler the current state of the array.
759 * This covers whether it is known to be consistent (no pending writes)
760 * and how far along a resync is known to have progressed
761 * (in a->resync_start).
762 * resync status is really irrelevant if the array is not consistent,
763 * but some metadata (DDF!) have a place to record the distinction.
764 * If 'consistent' is '2', then the array can mark it dirty if a
765 * resync/recovery/whatever is required, or leave it clean if not.
766 * Return value is 0 dirty (not consistent) and 1 if clean.
767 * it is only really important if consistent is passed in as '2'.
769 int (*set_array_state
)(struct active_array
*a
, int consistent
);
771 /* When the state of a device might have changed, we call set_disk to
772 * tell the metadata what the current state is.
773 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
775 * set_disk might be called when the state of the particular disk has
776 * not in fact changed.
778 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
779 void (*sync_metadata
)(struct supertype
*st
);
780 void (*process_update
)(struct supertype
*st
,
781 struct metadata_update
*update
);
782 void (*prepare_update
)(struct supertype
*st
,
783 struct metadata_update
*update
);
785 /* activate_spare will check if the array is degraded and, if it
786 * is, try to find some spare space in the container.
787 * On success, it add appropriate updates (For process_update) to
788 * to the 'updates' list and returns a list of 'mdinfo' identifying
789 * the device, or devices as there might be multiple missing
790 * devices and multiple spares available.
792 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
793 struct metadata_update
**updates
);
795 * Return statically allocated string that represents metadata specific
796 * controller domain of the disk. The domain is used in disk domain
797 * matching functions. Disks belong to the same domain if the they have
798 * the same domain from mdadm.conf and belong the same metadata domain.
799 * Returning NULL or not providing this handler means that metadata
800 * does not distinguish the differences between disks that belong to
801 * different controllers. They are in the domain specified by
802 * configuration file (mdadm.conf).
803 * In case when the metadata has the notion of domains based on disk
804 * it shall return NULL for disks that do not belong to the controller
805 * the supported domains. Such disks will form another domain and won't
806 * be mixed with supported ones.
808 const char *(*get_disk_controller_domain
)(const char *path
);
810 /* for external backup area */
811 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
813 int swapuuid
; /* true if uuid is bigending rather than hostendian */
815 const char *name
; /* canonical metadata name */
818 extern struct superswitch super0
, super1
;
819 extern struct superswitch super_imsm
, super_ddf
;
820 extern struct superswitch mbr
, gpt
;
822 struct metadata_update
{
825 void *space
; /* allocated space that monitor will use */
826 void **space_list
; /* list of allocated spaces that monitor can
827 * use or that it returned.
829 struct metadata_update
*next
;
832 /* A supertype holds a particular collection of metadata.
833 * It identifies the metadata type by the superswitch, and the particular
834 * sub-version of that metadata type.
835 * metadata read in or created is stored in 'sb' and 'info'.
836 * There are also fields used by mdmon to track containers.
838 * A supertype may refer to:
839 * Just an array, possibly in a container
840 * A container, not identifying any particular array
841 * Info read from just one device, not yet fully describing the array/container.
844 * A supertype is created by:
850 struct superswitch
*ss
;
853 int container_dev
; /* devnum of container */
856 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
857 HW/firmware related incompatability to load metadata.
858 Used when examining metadata to display content of disk
859 when user has no hw/firmare compatible system.
861 struct metadata_update
*updates
;
862 struct metadata_update
**update_tail
;
864 /* extra stuff used by mdmon */
865 struct active_array
*arrays
;
866 int sock
; /* listen to external programs */
868 char *devname
; /* e.g. md0. This appears in metadata_verison:
877 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
878 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
879 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
880 static inline struct supertype
*guess_super(int fd
) {
881 return guess_super_type(fd
, guess_any
);
883 extern struct supertype
*dup_super(struct supertype
*st
);
884 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
885 extern int must_be_container(int fd
);
886 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
887 extern void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
,
888 mdu_disk_info_t
*disk
);
889 void wait_for(char *dev
, int fd
);
892 * Data structures for policy management.
893 * Each device can have a policy structure that lists
894 * various name/value pairs each possibly with a metadata associated.
895 * The policy list is sorted by name/value/metadata
898 struct dev_policy
*next
;
899 char *name
; /* None of these strings are allocated. They are
900 * all just references to strings which are known
901 * to exist elsewhere.
902 * name and metadata can be compared by address equality.
904 const char *metadata
;
908 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
910 /* iterate over the sublist starting at list, having the same
911 * 'name' as 'list', and matching the given metadata (Where
912 * NULL matches anything
914 #define pol_for_each(item, list, _metadata) \
916 item && item->name == list->name; \
918 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
922 * policy records read from mdadm are largely just name-value pairs.
923 * The names are constants, not strdupped
926 struct pol_rule
*next
;
927 char *type
; /* rule_policy or rule_part */
932 char *dups
; /* duplicates of 'value' with a partNN appended */
936 extern char rule_policy
[], rule_part
[];
937 extern char rule_path
[], rule_type
[];
938 extern char type_part
[], type_disk
[];
940 extern void policyline(char *line
, char *type
);
941 extern void policy_add(char *type
, ...);
942 extern void policy_free(void);
944 extern struct dev_policy
*path_policy(char *path
, char *type
);
945 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
946 extern struct dev_policy
*devnum_policy(int dev
);
947 extern void dev_policy_free(struct dev_policy
*p
);
949 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
950 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
951 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
957 act_spare
, /* This only applies to bare devices */
958 act_spare_same_slot
, /* this allows non-bare devices,
959 * but only if recent removal */
960 act_force_spare
, /* this allow non-bare devices in any case */
964 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
965 enum policy_action want
);
966 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
967 enum policy_action want
);
970 struct domainlist
*next
;
974 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
975 const char *metadata
);
976 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
977 const char *metadata
);
978 extern void domainlist_add_dev(struct domainlist
**dom
, int devnum
,
979 const char *metadata
);
980 extern void domain_free(struct domainlist
*dl
);
981 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
982 const char *metadata
);
983 void domain_add(struct domainlist
**domp
, char *domain
);
985 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
986 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
992 #define HAVE_NFTW we assume
996 # include <features.h>
997 # ifndef __UCLIBC_HAS_LFS__
998 # define lseek64 lseek
1000 # ifndef __UCLIBC_HAS_FTW__
1010 #if defined(__KLIBC__)
1026 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1029 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1030 extern int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
);
1031 extern int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
);
1032 extern int Manage_subdevs(char *devname
, int fd
,
1033 struct mddev_dev
*devlist
, int verbose
, int test
,
1034 char *update
, int force
);
1035 extern int autodetect(void);
1036 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1037 extern int Grow_addbitmap(char *devname
, int fd
, char *file
, int chunk
, int delay
, int write_behind
, int force
);
1038 extern int Grow_reshape(char *devname
, int fd
, int quiet
, char *backup_file
,
1040 int level
, char *layout_str
, int chunksize
, int raid_disks
,
1041 struct mddev_dev
*devlist
,
1042 int assume_clean
, int force
);
1043 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1044 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1045 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1046 struct mdinfo
*info
, char *backup_file
,
1047 int freeze_reshape
);
1049 extern int restore_backup(struct supertype
*st
,
1050 struct mdinfo
*content
,
1055 extern int Grow_continue_command(char *devname
, int fd
,
1056 char *backup_file
, int verbose
);
1058 extern int Assemble(struct supertype
*st
, char *mddev
,
1059 struct mddev_ident
*ident
,
1060 struct mddev_dev
*devlist
,
1061 char *backup_file
, int invalid_backup
,
1062 int readonly
, int runstop
,
1063 char *update
, char *homehost
, int require_homehost
,
1064 int verbose
, int force
, int freeze_reshape
);
1066 extern int Build(char *mddev
, int chunk
, int level
, int layout
,
1067 int raiddisks
, struct mddev_dev
*devlist
, int assume_clean
,
1068 char *bitmap_file
, int bitmap_chunk
, int write_behind
,
1069 int delay
, int verbose
, int autof
, unsigned long long size
);
1072 extern int Create(struct supertype
*st
, char *mddev
,
1073 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
1074 char *name
, char *homehost
, int *uuid
,
1075 int subdevs
, struct mddev_dev
*devlist
,
1076 int runstop
, int verbose
, int force
, int assume_clean
,
1077 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
);
1079 extern int Detail(char *dev
, int brief
, int export
, int test
, char *homehost
);
1080 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
);
1081 extern int Query(char *dev
);
1082 extern int Examine(struct mddev_dev
*devlist
, int brief
, int export
, int scan
,
1083 int SparcAdjust
, struct supertype
*forcest
, char *homehost
);
1084 extern int Monitor(struct mddev_dev
*devlist
,
1085 char *mailaddr
, char *alert_cmd
,
1086 int period
, int daemonise
, int scan
, int oneshot
,
1087 int dosyslog
, int test
, char *pidfile
, int increments
,
1090 extern int Kill(char *dev
, struct supertype
*st
, int force
, int quiet
, int noexcl
);
1091 extern int Kill_subarray(char *dev
, char *subarray
, int quiet
);
1092 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1093 extern int Wait(char *dev
);
1094 extern int WaitClean(char *dev
, int sock
, int verbose
);
1096 extern int Incremental(char *devname
, int verbose
, int runstop
,
1097 struct supertype
*st
, char *homehost
, int require_homehost
,
1098 int autof
, int freeze_reshape
);
1099 extern void RebuildMap(void);
1100 extern int IncrementalScan(int verbose
);
1101 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1102 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1103 unsigned long chunksize
, unsigned long daemon_sleep
,
1104 unsigned long write_behind
,
1105 unsigned long long array_size
,
1107 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1108 extern int Write_rules(char *rule_name
);
1109 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1110 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
1112 extern int md_get_version(int fd
);
1113 extern int get_linux_version(void);
1114 extern int mdadm_version(char *version
);
1115 extern long long parse_size(char *size
);
1116 extern int parse_uuid(char *str
, int uuid
[4]);
1117 extern int parse_layout_10(char *layout
);
1118 extern int parse_layout_faulty(char *layout
);
1119 extern int check_ext2(int fd
, char *name
);
1120 extern int check_reiser(int fd
, char *name
);
1121 extern int check_raid(int fd
, char *name
);
1122 extern int check_partitions(int fd
, char *dname
,
1123 unsigned long long freesize
,
1124 unsigned long long size
);
1126 extern int get_mdp_major(void);
1127 extern int dev_open(char *dev
, int flags
);
1128 extern int open_dev(int devnum
);
1129 extern int open_dev_flags(int devnum
, int flags
);
1130 extern int open_dev_excl(int devnum
);
1131 extern int is_standard(char *dev
, int *nump
);
1132 extern int same_dev(char *one
, char *two
);
1134 extern int parse_auto(char *str
, char *msg
, int config
);
1135 extern struct mddev_ident
*conf_get_ident(char *dev
);
1136 extern struct mddev_dev
*conf_get_devs(void);
1137 extern int conf_test_dev(char *devname
);
1138 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1139 extern struct createinfo
*conf_get_create_info(void);
1140 extern void set_conffile(char *file
);
1141 extern char *conf_get_mailaddr(void);
1142 extern char *conf_get_mailfrom(void);
1143 extern char *conf_get_program(void);
1144 extern char *conf_get_homehost(int *require_homehostp
);
1145 extern char *conf_line(FILE *file
);
1146 extern char *conf_word(FILE *file
, int allow_key
);
1147 extern int conf_name_is_free(char *name
);
1148 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1149 extern int devname_matches(char *name
, char *match
);
1150 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1151 struct mdinfo
*info
,
1153 int verbose
, int *rvp
);
1154 extern int experimental(void);
1156 extern void free_line(char *line
);
1157 extern int match_oneof(char *devices
, char *devname
);
1158 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1159 extern const int uuid_zero
[4];
1160 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1161 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1162 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1163 extern char *fname_from_uuid(struct supertype
*st
,
1164 struct mdinfo
*info
, char *buf
, char sep
);
1165 extern unsigned long calc_csum(void *super
, int bytes
);
1166 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1167 char *avail
, int avail_disks
);
1168 extern int enough_fd(int fd
);
1169 extern int ask(char *mesg
);
1170 extern unsigned long long get_component_size(int fd
);
1171 extern void remove_partitions(int fd
);
1172 extern int test_partition(int fd
);
1173 extern int test_partition_from_id(dev_t id
);
1174 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1175 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1176 int chunksize
, unsigned long long devsize
);
1177 extern int flush_metadata_updates(struct supertype
*st
);
1178 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1179 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1180 struct mdinfo
*content
, int runstop
,
1181 char *chosen_name
, int verbose
,
1182 char *backup_file
, int freeze_reshape
);
1183 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1184 unsigned long long min_size
,
1185 struct domainlist
*domlist
,
1187 const char *metadata
, int get_one
);
1188 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1189 extern int add_disk(int mdfd
, struct supertype
*st
,
1190 struct mdinfo
*sra
, struct mdinfo
*info
);
1191 extern int remove_disk(int mdfd
, struct supertype
*st
,
1192 struct mdinfo
*sra
, struct mdinfo
*info
);
1193 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1194 unsigned long long min_recovery_start(struct mdinfo
*array
);
1196 extern char *human_size(long long bytes
);
1197 extern char *human_size_brief(long long bytes
);
1198 extern void print_r10_layout(int layout
);
1200 #define NoMdDev (1<<23)
1201 extern int find_free_devnum(int use_partitions
);
1203 extern void put_md_name(char *name
);
1204 extern char *get_md_name(int dev
);
1206 extern char DefaultConfFile
[];
1208 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1210 /* values for 'trustworthy' */
1212 #define LOCAL_ANY 10
1215 extern int open_mddev(char *dev
, int report_errors
);
1216 extern int open_container(int fd
);
1217 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1218 extern int is_subarray_active(char *subarray
, char *devname
);
1219 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1220 extern struct superswitch
*version_to_superswitch(char *vers
);
1222 extern int mdmon_running(int devnum
);
1223 extern int mdmon_pid(int devnum
);
1224 extern int check_env(char *name
);
1225 extern __u32
random32(void);
1226 extern int start_mdmon(int devnum
);
1228 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1229 struct supertype
*st
, unsigned long stripes
,
1230 int *fds
, unsigned long long *offsets
,
1231 int dests
, int *destfd
, unsigned long long *destoffsets
);
1232 void abort_reshape(struct mdinfo
*sra
);
1234 extern char *devnum2devname(int num
);
1235 extern void fmt_devname(char *name
, int num
);
1236 extern int devname2devnum(char *name
);
1237 extern int stat2devnum(struct stat
*st
);
1238 extern int fd2devnum(int fd
);
1240 static inline int dev2major(int d
)
1245 return get_mdp_major();
1248 static inline int dev2minor(int d
)
1252 return (-1-d
) << MdpMinorShift
;
1255 static inline int ROUND_UP(int a
, int base
)
1257 return ((a
+base
-1)/base
)*base
;
1260 static inline int is_subarray(char *vers
)
1262 /* The version string for a 'subarray' (an array in a container)
1264 * /containername/componentname for normal read-write arrays
1265 * -containername/componentname for arrays which mdmon must not
1266 * reconfigure. They might be read-only
1267 * or might be undergoing reshape etc.
1268 * containername is e.g. md0, md_d1
1269 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1271 return (*vers
== '/' || *vers
== '-');
1274 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1276 return &ent
->metadata_version
[10+strlen(container
)+1];
1280 #define dprintf(fmt, arg...) \
1281 fprintf(stderr, fmt, ##arg)
1283 #define dprintf(fmt, arg...) \
1284 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1288 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1292 ret
= vasprintf(strp
, fmt
, ap
);
1298 #define LEVEL_MULTIPATH (-4)
1299 #define LEVEL_LINEAR (-1)
1300 #define LEVEL_FAULTY (-5)
1302 /* kernel module doesn't know about these */
1303 #define LEVEL_CONTAINER (-100)
1304 #define LEVEL_UNSUPPORTED (-200)
1309 #define WriteTransient 0
1310 #define ReadTransient 1
1311 #define WritePersistent 2
1312 #define ReadPersistent 3
1313 #define WriteAll 4 /* doesn't go to device */
1314 #define ReadFixable 5
1317 #define ClearErrors 31
1318 #define ClearFaults 30
1320 #define AllPersist 100 /* internal use only */
1321 #define NoPersist 101
1323 #define ModeMask 0x1f
1331 #define minor(x) ((x)&0xff)
1332 #define major(x) (((x)>>8)&0xff)
1333 #define makedev(M,m) (((M)<<8) | (m))
1337 #define ALGORITHM_LEFT_ASYMMETRIC 0
1338 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1339 #define ALGORITHM_LEFT_SYMMETRIC 2
1340 #define ALGORITHM_RIGHT_SYMMETRIC 3
1342 /* Define non-rotating (raid4) algorithms. These allow
1343 * conversion of raid4 to raid5.
1345 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1346 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1348 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1349 * Firstly, the exact positioning of the parity block is slightly
1350 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1352 * Secondly, or order of datablocks over which the Q syndrome is computed
1354 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1355 * These layouts are from the DDFv1.2 spec.
1356 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1357 * leaves RLQ=3 as 'Vendor Specific'
1360 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1361 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1362 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1365 /* For every RAID5 algorithm we define a RAID6 algorithm
1366 * with exactly the same layout for data and parity, and
1367 * with the Q block always on the last device (N-1).
1368 * This allows trivial conversion from RAID5 to RAID6
1370 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1371 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1372 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1373 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1374 #define ALGORITHM_PARITY_0_6 20
1375 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1377 /* Define PATH_MAX in case we don't use glibc or standard library does
1378 * not have PATH_MAX defined. Assume max path length is 4K characters.
1381 #define PATH_MAX 4096
1384 #define PROCESS_DELAYED -2
1385 #define PROCESS_PENDING -3