2 * mdadm - manage Linux "md" devices aka RAID arrays.
4 * Copyright (C) 2001-2009 Neil Brown <neilb@suse.de>
7 * This program is free software; you can redistribute it and/or modify
8 * it under the terms of the GNU General Public License as published by
9 * the Free Software Foundation; either version 2 of the License, or
10 * (at your option) any later version.
12 * This program is distributed in the hope that it will be useful,
13 * but WITHOUT ANY WARRANTY; without even the implied warranty of
14 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
15 * GNU General Public License for more details.
17 * You should have received a copy of the GNU General Public License
18 * along with this program; if not, write to the Free Software
19 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA
22 * Email: <neilb@suse.de>
26 #define _FILE_OFFSET_BITS 64
28 #if !defined(__dietlibc__) && !defined(__KLIBC__)
29 extern __off64_t lseek64
__P ((int __fd
, __off64_t __offset
, int __whence
));
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
36 #include <sys/types.h>
49 /* dietlibc has deprecated random and srandom!! */
55 #include <linux/kdev_t.h>
56 /*#include <linux/fs.h> */
57 #include <sys/mount.h>
58 #include <asm/types.h>
59 #include <sys/ioctl.h>
61 #define MdpMinorShift 6
64 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
67 #define DEFAULT_BITMAP_CHUNK 4096
68 #define DEFAULT_BITMAP_DELAY 5
69 #define DEFAULT_MAX_WRITE_BEHIND 256
71 /* MAP_DIR should be somewhere that persists across the pivotroot
72 * from early boot to late boot.
73 * Currently /dev seems to be the only option on most distros.
76 #define MAP_DIR "/dev/.mdadm"
78 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
79 * want something other than the default of "map"
82 #define MAP_FILE "map"
84 /* MDMON_DIR is where pid and socket files used for communicating
85 * with mdmon normally live. It *should* be /var/run, but when
86 * mdmon is needed at early boot then it needs to write there prior
87 * to /var/run being mounted read/write, and it also then needs to
88 * persist beyond when /var/run is mounter read-only. So, to be
89 * safe, the default is somewhere that is read/write early in the
90 * boot process and stays up as long as possible during shutdown.
93 #define MDMON_DIR "/dev/.mdadm/"
94 #endif /* MDMON_DIR */
102 /* Redhat don't like to #include <asm/byteorder.h>, and
103 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
104 * and there is no standard conversion function so... */
105 /* And dietlibc doesn't think byteswap is ok, so.. */
106 /* #include <byteswap.h> */
107 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
108 ((x) & 0xff00U) >> 8)
109 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
110 ((x) & 0xff000000U) >> 24 | \
111 ((x) & 0x0000ff00U) << 8 | \
112 ((x) & 0x00ff0000U) >> 8)
113 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
114 ((x) & 0xff00000000000000ULL) >> 56 | \
115 ((x) & 0x000000000000ff00ULL) << 40 | \
116 ((x) & 0x00ff000000000000ULL) >> 40 | \
117 ((x) & 0x0000000000ff0000ULL) << 24 | \
118 ((x) & 0x0000ff0000000000ULL) >> 24 | \
119 ((x) & 0x00000000ff000000ULL) << 8 | \
120 ((x) & 0x000000ff00000000ULL) >> 8)
122 #if !defined(__KLIBC__)
123 #if BYTE_ORDER == LITTLE_ENDIAN
124 #define __cpu_to_le16(_x) (_x)
125 #define __cpu_to_le32(_x) (_x)
126 #define __cpu_to_le64(_x) (_x)
127 #define __le16_to_cpu(_x) (_x)
128 #define __le32_to_cpu(_x) (_x)
129 #define __le64_to_cpu(_x) (_x)
131 #define __cpu_to_be16(_x) bswap_16(_x)
132 #define __cpu_to_be32(_x) bswap_32(_x)
133 #define __cpu_to_be64(_x) bswap_64(_x)
134 #define __be16_to_cpu(_x) bswap_16(_x)
135 #define __be32_to_cpu(_x) bswap_32(_x)
136 #define __be64_to_cpu(_x) bswap_64(_x)
137 #elif BYTE_ORDER == BIG_ENDIAN
138 #define __cpu_to_le16(_x) bswap_16(_x)
139 #define __cpu_to_le32(_x) bswap_32(_x)
140 #define __cpu_to_le64(_x) bswap_64(_x)
141 #define __le16_to_cpu(_x) bswap_16(_x)
142 #define __le32_to_cpu(_x) bswap_32(_x)
143 #define __le64_to_cpu(_x) bswap_64(_x)
145 #define __cpu_to_be16(_x) (_x)
146 #define __cpu_to_be32(_x) (_x)
147 #define __cpu_to_be64(_x) (_x)
148 #define __be16_to_cpu(_x) (_x)
149 #define __be32_to_cpu(_x) (_x)
150 #define __be64_to_cpu(_x) (_x)
152 # error "unknown endianness."
154 #endif /* __KLIBC__ */
158 * min()/max()/clamp() macros that also do
159 * strict type-checking.. See the
160 * "unnecessary" pointer comparison.
162 #define min(x, y) ({ \
163 typeof(x) _min1 = (x); \
164 typeof(y) _min2 = (y); \
165 (void) (&_min1 == &_min2); \
166 _min1 < _min2 ? _min1 : _min2; })
168 #define max(x, y) ({ \
169 typeof(x) _max1 = (x); \
170 typeof(y) _max2 = (y); \
171 (void) (&_max1 == &_max2); \
172 _max1 > _max2 ? _max1 : _max2; })
174 /* general information that might be extracted from a superblock */
176 mdu_array_info_t array
;
177 mdu_disk_info_t disk
;
181 unsigned long long data_offset
;
182 unsigned long long component_size
; /* same as array.size, except in
183 * sectors and up to 64bits.
185 unsigned long long custom_array_size
; /* size for non-default sized
186 * arrays (in sectors)
189 unsigned long long reshape_progress
;
191 unsigned long long resync_start
; /* per-array resync position */
192 unsigned long long recovery_start
; /* per-device rebuild position */
193 #define MaxSector (~0ULL) /* resync/recovery complete position */
195 unsigned long safe_mode_delay
; /* ms delay to mark clean */
196 int new_level
, delta_disks
, new_layout
, new_chunk
;
198 unsigned long cache_size
; /* size of raid456 stripe cache*/
200 char text_version
[50];
201 void *update_private
; /* for passing metadata-format
202 * specific update data
203 * between successive calls to
207 int container_member
; /* for assembling external-metatdata arrays
208 * This is to be used internally by metadata
210 int container_enough
; /* flag external handlers can set to
211 * indicate that subarrays have not enough (-1),
212 * enough to start (0), or all expected disks (1) */
217 /* Device info for mdmon: */
222 #define DS_WRITE_MOSTLY 4
224 #define DS_BLOCKED 16
225 #define DS_REMOVE 1024
226 #define DS_UNBLOCK 2048
227 int prev_state
, curr_state
, next_state
;
237 struct supertype
*supertype
;
254 extern char short_options
[];
255 extern char short_bitmap_options
[];
256 extern char short_bitmap_auto_options
[];
257 extern struct option long_options
[];
258 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
259 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
261 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
263 /* for option that don't have short equivilents, we assign arbitrary
264 * small numbers. '1' means an undecorated option, so we start at '2'.
265 * (note we must stop before we get to 65 i.e. 'A')
267 enum special_options
{
282 UpdateSubarray
, /* 16 */
286 /* structures read from config file */
287 /* List of mddevice names and identifiers
288 * Identifiers can be:
290 * super-minor=decimal-minor-number-from-superblock
291 * devices=comma,separated,list,of,device,names,with,wildcards
293 * If multiple fields are present, the intersection of all matching
294 * devices is considered
296 #define UnSet (0xfffe)
306 char *devices
; /* comma separated list of device
307 * names with wild cards
312 struct supertype
*st
;
313 int autof
; /* 1 for normal, 2 for partitioned */
318 char *container
; /* /dev/whatever name of container, or
319 * uuid of container. You would expect
320 * this to be the 'devname' or UUID
321 * of some other entry.
323 char *member
; /* subarray within a container */
325 struct mddev_ident
*next
;
327 /* fields needed by different users of this structure */
328 int assembled
; /* set when assembly succeeds */
332 /* List of device names - wildcards expanded */
335 char disposition
; /* 'a' for add, 'r' for remove, 'f' for fail.
336 * Not set for names read from .config
338 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
340 char used
; /* set when used */
341 struct mddev_dev
*next
;
344 typedef struct mapping
{
355 char *pattern
; /* U or up, _ for down */
356 int percent
; /* -1 if no resync */
357 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
361 char * metadata_version
;
364 struct dev_member
*next
;
366 struct mdstat_ent
*next
;
369 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
370 extern void free_mdstat(struct mdstat_ent
*ms
);
371 extern void mdstat_wait(int seconds
);
372 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
373 extern int mddev_busy(int devnum
);
374 extern struct mdstat_ent
*mdstat_by_component(char *name
);
377 struct map_ent
*next
;
384 extern int map_update(struct map_ent
**mpp
, int devnum
, char *metadata
,
385 int uuid
[4], char *path
);
386 extern void map_remove(struct map_ent
**map
, int devnum
);
387 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
388 extern struct map_ent
*map_by_devnum(struct map_ent
**map
, int devnum
);
389 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
390 extern void map_read(struct map_ent
**melp
);
391 extern int map_write(struct map_ent
*mel
);
392 extern void map_delete(struct map_ent
**mapp
, int devnum
);
393 extern void map_free(struct map_ent
*map
);
394 extern void map_add(struct map_ent
**melp
,
395 int devnum
, char *metadata
, int uuid
[4], char *path
);
396 extern int map_lock(struct map_ent
**melp
);
397 extern void map_unlock(struct map_ent
**melp
);
399 /* various details can be requested */
400 enum sysfs_read_flags
{
401 GET_LEVEL
= (1 << 0),
402 GET_LAYOUT
= (1 << 1),
403 GET_COMPONENT
= (1 << 2),
404 GET_CHUNK
= (1 << 3),
405 GET_CACHE
= (1 << 4),
406 GET_MISMATCH
= (1 << 5),
407 GET_VERSION
= (1 << 6),
408 GET_DISKS
= (1 << 7),
409 GET_DEGRADED
= (1 << 8),
410 GET_SAFEMODE
= (1 << 9),
411 GET_DEVS
= (1 << 10), /* gets role, major, minor */
412 GET_OFFSET
= (1 << 11),
413 GET_SIZE
= (1 << 12),
414 GET_STATE
= (1 << 13),
415 GET_ERROR
= (1 << 14),
418 /* If fd >= 0, get the array it is open on,
419 * else use devnum. >=0 -> major9. <0.....
421 extern int sysfs_open(int devnum
, char *devname
, char *attr
);
422 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, int devnum
);
423 extern void sysfs_free(struct mdinfo
*sra
);
424 extern struct mdinfo
*sysfs_read(int fd
, int devnum
, unsigned long options
);
425 extern int sysfs_attr_match(const char *attr
, const char *str
);
426 extern int sysfs_match_word(const char *word
, char **list
);
427 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
428 char *name
, char *val
);
429 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
430 char *name
, unsigned long long val
);
431 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
432 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
434 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
435 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
436 char *name
, unsigned long long *val
);
437 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
438 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
439 char *name
, char *val
, int size
);
440 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
441 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
442 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
443 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
444 extern int sysfs_unique_holder(int devnum
, long rdev
);
445 extern int load_sys(char *path
, char *buf
);
448 extern int save_stripes(int *source
, unsigned long long *offsets
,
449 int raid_disks
, int chunk_size
, int level
, int layout
,
450 int nwrites
, int *dest
,
451 unsigned long long start
, unsigned long long length
,
453 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
454 int raid_disks
, int chunk_size
, int level
, int layout
,
455 int source
, unsigned long long read_offset
,
456 unsigned long long start
, unsigned long long length
);
459 #define Sendmail "/usr/lib/sendmail -t"
462 #define SYSLOG_FACILITY LOG_DAEMON
464 extern char *map_num(mapping_t
*map
, int num
);
465 extern int map_name(mapping_t
*map
, char *name
);
466 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
468 extern char *map_dev(int major
, int minor
, int create
);
471 struct metadata_update
;
473 /* A superswitch provides entry point the a metadata handler.
475 * The super_switch primarily operates on some "metadata" that
476 * is accessed via the 'supertype'.
477 * This metadata has one of three possible sources.
478 * 1/ It is read from a single device. In this case it may not completely
479 * describe the array or arrays as some information might be on other
481 * 2/ It is read from all devices in a container. In this case all
482 * information is present.
483 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
484 * be complete once enough ->add_to_super calls have completed.
486 * When creating an array inside a container, the metadata will be
487 * formed by a combination of 2 and 3. The metadata or the array is read,
488 * then new information is added.
490 * The metadata must sometimes have a concept of a 'current' array
491 * and a 'current' device.
492 * The 'current' array is set by init_super to be the newly created array,
493 * or is set by super_by_fd when it finds it is looking at an array inside
496 * The 'current' device is either the device that the metadata was read from
497 * in case 1, or the last device added by add_to_super in case 3.
498 * Case 2 does not identify a 'current' device.
500 extern struct superswitch
{
502 /* Used to report details of metadata read from a component
503 * device. ->load_super has been called.
505 void (*examine_super
)(struct supertype
*st
, char *homehost
);
506 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
507 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
508 void (*export_examine_super
)(struct supertype
*st
);
510 /* Used to report details of an active array.
511 * ->load_super was possibly given a 'component' string.
513 void (*detail_super
)(struct supertype
*st
, char *homehost
);
514 void (*brief_detail_super
)(struct supertype
*st
);
515 void (*export_detail_super
)(struct supertype
*st
);
517 /* Optional: platform hardware / firmware details */
518 int (*detail_platform
)(int verbose
, int enumerate_only
);
521 * to get uuid to storing in bitmap metadata
522 * and 'reshape' backup-data metadata
523 * To see if a device is being re-added to an array it was part of.
525 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
527 /* Extract generic details from metadata. This could be details about
528 * the container, or about an individual array within the container.
529 * The determination is made either by:
530 * load_super being given a 'component' string.
531 * validate_geometry determining what to create.
532 * The info includes both array information and device information.
533 * The particular device should be:
534 * The last device added by add_to_super
535 * The device the metadata was loaded from by load_super
536 * If 'map' is present, then it is an array raid_disks long
537 * (raid_disk must already be set and correct) and it is filled
538 * with 1 for slots that are thought to be active and 0 for slots which
539 * appear to be failed/missing.
541 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
543 /* Check if the given metadata is flagged as belonging to "this"
544 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
546 int (*match_home
)(struct supertype
*st
, char *homehost
);
548 /* Make one of several generic modifications to metadata
549 * prior to assembly (or other times).
550 * sparc2.2 - first bug in early 0.90 metadata
551 * super-minor - change name of 0.90 metadata
552 * summaries - 'correct' any redundant data
553 * resync - mark array as dirty to trigger a resync.
554 * uuid - set new uuid - only 0.90 or 1.x
555 * name - change the name of the array (where supported)
556 * homehost - change which host this array is tied to.
557 * devicesize - If metadata is at start of device, change recorded
558 * device size to match actual device size
559 * byteorder - swap bytes for 0.90 metadata
561 * force-one - mark that device as uptodate, not old or failed.
562 * force-array - mark array as clean if it would not otherwise
564 * assemble - not sure how this is different from force-one...
565 * linear-grow-new - add a new device to a linear array, but don't
566 * change the size: so superblock still matches
567 * linear-grow-update - now change the size of the array.
569 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
571 char *devname
, int verbose
,
572 int uuid_set
, char *homehost
);
574 /* Create new metadata for new array as described. This could
575 * be a new container, or an array in a pre-existing container.
576 * Also used to zero metadata prior to writing it to invalidate old
579 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
580 unsigned long long size
, char *name
,
581 char *homehost
, int *uuid
);
583 /* update the metadata to include new device, either at create or
584 * when hot-adding a spare.
586 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
587 int fd
, char *devname
);
589 /* Write metadata to one device when fixing problems or adding
592 int (*store_super
)(struct supertype
*st
, int fd
);
594 /* Write all metadata for this array.
596 int (*write_init_super
)(struct supertype
*st
);
597 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
598 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
599 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
600 struct supertype
* (*match_metadata_desc
)(char *arg
);
601 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
);
602 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
603 int delay
, int write_behind
,
604 unsigned long long size
, int may_change
, int major
);
605 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
606 int (*write_bitmap
)(struct supertype
*st
, int fd
);
607 void (*free_super
)(struct supertype
*st
);
609 /* validate_geometry is called with an st returned by
610 * match_metadata_desc.
611 * It should check that the geometry described in compatible with
612 * the metadata type. It will be called repeatedly as devices
613 * added to validate changing size and new devices. If there are
614 * inter-device dependencies, it should record sufficient details
615 * so these can be validated.
616 * Both 'size' and '*freesize' are in sectors. chunk is bytes.
618 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
620 int chunk
, unsigned long long size
,
621 char *subdev
, unsigned long long *freesize
,
624 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
625 /* Allow a metadata handler to override mdadm's default layouts */
626 int (*default_layout
)(int level
); /* optional */
627 /* query the supertype for default chunk size */
628 int (*default_chunk
)(struct supertype
*st
); /* optional */
629 /* Permit subarray's to be deleted from inactive containers */
630 int (*kill_subarray
)(struct supertype
*st
); /* optional */
631 /* Permit subarray's to be modified */
632 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
633 char *update
, struct mddev_ident
*ident
); /* optional */
636 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
639 /* Tell the metadata handler the current state of the array.
640 * This covers whether it is known to be consistent (no pending writes)
641 * and how far along a resync is known to have progressed
642 * (in a->resync_start).
643 * resync status is really irrelevant if the array is not consistent,
644 * but some metadata (DDF!) have a place to record the distinction.
645 * If 'consistent' is '2', then the array can mark it dirty if a
646 * resync/recovery/whatever is required, or leave it clean if not.
647 * Return value is 0 dirty (not consistent) and 1 if clean.
648 * it is only really important if consistent is passed in as '2'.
650 int (*set_array_state
)(struct active_array
*a
, int consistent
);
652 /* When the state of a device might have changed, we call set_disk to
653 * tell the metadata what the current state is.
654 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
656 * set_disk might be called when the state of the particular disk has
657 * not in fact changed.
659 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
660 void (*sync_metadata
)(struct supertype
*st
);
661 void (*process_update
)(struct supertype
*st
,
662 struct metadata_update
*update
);
663 void (*prepare_update
)(struct supertype
*st
,
664 struct metadata_update
*update
);
666 /* activate_spare will check if the array is degraded and, if it
667 * is, try to find some spare space in the container.
668 * On success, it add appropriate updates (For process_update) to
669 * to the 'updates' list and returns a list of 'mdinfo' identifying
670 * the device, or devices as there might be multiple missing
671 * devices and multiple spares available.
673 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
674 struct metadata_update
**updates
);
676 int swapuuid
; /* true if uuid is bigending rather than hostendian */
678 const char *name
; /* canonical metadata name */
681 extern struct superswitch super0
, super1
;
682 extern struct superswitch super_imsm
, super_ddf
;
683 extern struct superswitch mbr
, gpt
;
685 struct metadata_update
{
688 void *space
; /* allocated space that monitor will use */
689 struct metadata_update
*next
;
692 /* A supertype holds a particular collection of metadata.
693 * It identifies the metadata type by the superswitch, and the particular
694 * sub-version of that metadata type.
695 * metadata read in or created is stored in 'sb' and 'info'.
696 * There are also fields used by mdmon to track containers.
698 * A supertype may refer to:
699 * Just an array, possibly in a container
700 * A container, not identifying any particular array
701 * Info read from just one device, not yet fully describing the array/container.
704 * A supertype is created by:
710 struct superswitch
*ss
;
713 int container_dev
; /* devnum of container */
717 struct metadata_update
*updates
;
718 struct metadata_update
**update_tail
;
720 /* extra stuff used by mdmon */
721 struct active_array
*arrays
;
722 int sock
; /* listen to external programs */
724 char *devname
; /* e.g. md0. This appears in metadata_verison:
733 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
734 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
735 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
736 static inline struct supertype
*guess_super(int fd
) {
737 return guess_super_type(fd
, guess_any
);
739 extern struct supertype
*dup_super(struct supertype
*st
);
740 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
741 extern int must_be_container(int fd
);
742 extern void get_one_disk(int mdfd
, mdu_array_info_t
*ainf
,
743 mdu_disk_info_t
*disk
);
744 void wait_for(char *dev
, int fd
);
747 * Data structures for policy management.
748 * Each device can have a policy structure that lists
749 * various name/value pairs each possibly with a metadata associated.
750 * The policy list is sorted by name/value/metadata
753 struct dev_policy
*next
;
754 char *name
; /* None of these strings are allocated. They are
755 * all just references to strings which are known
756 * to exist elsewhere.
757 * name and metadata can be compared by address equality.
759 const char *metadata
;
763 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
765 /* iterate over the sublist starting at list, having the same
766 * 'name' as 'list', and matching the given metadata (Where
767 * NULL matches anything
769 #define pol_for_each(item, list, _metadata) \
771 item && item->name == list->name; \
773 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
777 * policy records read from mdadm are largely just name-value pairs.
778 * The names are constants, not strdupped
781 struct pol_rule
*next
;
782 char *type
; /* rule_policy or rule_part */
787 char *dups
; /* duplicates of 'value' with a partNN appended */
791 extern char rule_policy
[], rule_part
[];
792 extern char rule_path
[], rule_type
[];
793 extern char type_part
[], type_disk
[];
795 extern void policyline(char *line
, char *type
);
796 extern void policy_add(char *type
, ...);
797 extern void policy_free(void);
799 extern struct dev_policy
*path_policy(char *path
, char *type
);
800 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
801 extern struct dev_policy
*devnum_policy(int dev
);
802 extern void dev_policy_free(struct dev_policy
*p
);
804 extern void pol_new(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
805 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
816 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
817 enum policy_action want
);
818 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
819 enum policy_action want
);
822 struct domainlist
*next
;
826 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
827 const char *metadata
);
828 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
829 const char *metadata
);
830 extern void domain_free(struct domainlist
*dl
);
831 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
832 const char *metadata
);
838 #define HAVE_NFTW we assume
842 # include <features.h>
843 # ifndef __UCLIBC_HAS_LFS__
844 # define lseek64 lseek
846 # ifndef __UCLIBC_HAS_FTW__
856 #if defined(__KLIBC__)
872 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
875 extern int Manage_ro(char *devname
, int fd
, int readonly
);
876 extern int Manage_runstop(char *devname
, int fd
, int runstop
, int quiet
);
877 extern int Manage_resize(char *devname
, int fd
, long long size
, int raid_disks
);
878 extern int Manage_subdevs(char *devname
, int fd
,
879 struct mddev_dev
*devlist
, int verbose
, int test
);
880 extern int autodetect(void);
881 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
882 extern int Grow_addbitmap(char *devname
, int fd
, char *file
, int chunk
, int delay
, int write_behind
, int force
);
883 extern int Grow_reshape(char *devname
, int fd
, int quiet
, char *backup_file
,
885 int level
, char *layout_str
, int chunksize
, int raid_disks
);
886 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
887 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
888 extern int Grow_continue(int mdfd
, struct supertype
*st
,
889 struct mdinfo
*info
, char *backup_file
);
891 extern int Assemble(struct supertype
*st
, char *mddev
,
892 struct mddev_ident
*ident
,
893 struct mddev_dev
*devlist
, char *backup_file
,
894 int readonly
, int runstop
,
895 char *update
, char *homehost
, int require_homehost
,
896 int verbose
, int force
);
898 extern int Build(char *mddev
, int chunk
, int level
, int layout
,
899 int raiddisks
, struct mddev_dev
*devlist
, int assume_clean
,
900 char *bitmap_file
, int bitmap_chunk
, int write_behind
,
901 int delay
, int verbose
, int autof
, unsigned long long size
);
904 extern int Create(struct supertype
*st
, char *mddev
,
905 int chunk
, int level
, int layout
, unsigned long long size
, int raiddisks
, int sparedisks
,
906 char *name
, char *homehost
, int *uuid
,
907 int subdevs
, struct mddev_dev
*devlist
,
908 int runstop
, int verbose
, int force
, int assume_clean
,
909 char *bitmap_file
, int bitmap_chunk
, int write_behind
, int delay
, int autof
);
911 extern int Detail(char *dev
, int brief
, int export
, int test
, char *homehost
);
912 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
);
913 extern int Query(char *dev
);
914 extern int Examine(struct mddev_dev
*devlist
, int brief
, int export
, int scan
,
915 int SparcAdjust
, struct supertype
*forcest
, char *homehost
);
916 extern int Monitor(struct mddev_dev
*devlist
,
917 char *mailaddr
, char *alert_cmd
,
918 int period
, int daemonise
, int scan
, int oneshot
,
919 int dosyslog
, int test
, char *pidfile
, int increments
);
921 extern int Kill(char *dev
, struct supertype
*st
, int force
, int quiet
, int noexcl
);
922 extern int Kill_subarray(char *dev
, char *subarray
, int quiet
);
923 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
924 extern int Wait(char *dev
);
925 extern int WaitClean(char *dev
, int sock
, int verbose
);
927 extern int Incremental(char *devname
, int verbose
, int runstop
,
928 struct supertype
*st
, char *homehost
, int require_homehost
,
930 extern void RebuildMap(void);
931 extern int IncrementalScan(int verbose
);
932 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
933 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
934 unsigned long chunksize
, unsigned long daemon_sleep
,
935 unsigned long write_behind
,
936 unsigned long long array_size
,
938 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
939 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
940 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
942 extern int md_get_version(int fd
);
943 extern int get_linux_version(void);
944 extern long long parse_size(char *size
);
945 extern int parse_uuid(char *str
, int uuid
[4]);
946 extern int parse_layout_10(char *layout
);
947 extern int parse_layout_faulty(char *layout
);
948 extern int check_ext2(int fd
, char *name
);
949 extern int check_reiser(int fd
, char *name
);
950 extern int check_raid(int fd
, char *name
);
951 extern int check_partitions(int fd
, char *dname
, unsigned long long freesize
);
953 extern int get_mdp_major(void);
954 extern int dev_open(char *dev
, int flags
);
955 extern int open_dev(int devnum
);
956 extern int open_dev_excl(int devnum
);
957 extern int is_standard(char *dev
, int *nump
);
958 extern int same_dev(char *one
, char *two
);
960 extern int parse_auto(char *str
, char *msg
, int config
);
961 extern struct mddev_ident
*conf_get_ident(char *dev
);
962 extern struct mddev_dev
*conf_get_devs(void);
963 extern int conf_test_dev(char *devname
);
964 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
965 extern struct createinfo
*conf_get_create_info(void);
966 extern void set_conffile(char *file
);
967 extern char *conf_get_mailaddr(void);
968 extern char *conf_get_mailfrom(void);
969 extern char *conf_get_program(void);
970 extern char *conf_get_homehost(int *require_homehostp
);
971 extern char *conf_line(FILE *file
);
972 extern char *conf_word(FILE *file
, int allow_key
);
973 extern int conf_name_is_free(char *name
);
974 extern int devname_matches(char *name
, char *match
);
975 extern struct mddev_ident
*conf_match(struct mdinfo
*info
, struct supertype
*st
);
977 extern void free_line(char *line
);
978 extern int match_oneof(char *devices
, char *devname
);
979 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
980 extern const int uuid_match_any
[4];
981 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
982 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
983 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
984 extern char *fname_from_uuid(struct supertype
*st
,
985 struct mdinfo
*info
, char *buf
, char sep
);
986 extern unsigned long calc_csum(void *super
, int bytes
);
987 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
988 char *avail
, int avail_disks
);
989 extern int enough_fd(int fd
);
990 extern int ask(char *mesg
);
991 extern unsigned long long get_component_size(int fd
);
992 extern void remove_partitions(int fd
);
993 extern int test_partition(int fd
);
994 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
995 int chunksize
, unsigned long long devsize
);
996 extern int flush_metadata_updates(struct supertype
*st
);
997 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
998 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
999 struct mdinfo
*content
, int runstop
,
1000 char *chosen_name
, int verbose
);
1002 extern int add_disk(int mdfd
, struct supertype
*st
,
1003 struct mdinfo
*sra
, struct mdinfo
*info
);
1004 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1005 unsigned long long min_recovery_start(struct mdinfo
*array
);
1007 extern char *human_size(long long bytes
);
1008 extern char *human_size_brief(long long bytes
);
1009 extern void print_r10_layout(int layout
);
1011 #define NoMdDev (1<<23)
1012 extern int find_free_devnum(int use_partitions
);
1014 extern void put_md_name(char *name
);
1015 extern char *get_md_name(int dev
);
1017 extern char DefaultConfFile
[];
1019 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1021 /* values for 'trustworthy' */
1023 #define LOCAL_ANY 10
1026 extern int open_mddev(char *dev
, int report_errors
);
1027 extern int open_container(int fd
);
1028 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1029 extern int is_subarray_active(char *subarray
, char *devname
);
1030 int is_container_active(char *devname
);
1031 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1032 extern struct superswitch
*version_to_superswitch(char *vers
);
1034 extern int mdmon_running(int devnum
);
1035 extern int mdmon_pid(int devnum
);
1036 extern int check_env(char *name
);
1037 extern __u32
random32(void);
1038 extern int start_mdmon(int devnum
);
1040 extern char *devnum2devname(int num
);
1041 extern int devname2devnum(char *name
);
1042 extern int stat2devnum(struct stat
*st
);
1043 extern int fd2devnum(int fd
);
1045 static inline int dev2major(int d
)
1050 return get_mdp_major();
1053 static inline int dev2minor(int d
)
1057 return (-1-d
) << MdpMinorShift
;
1060 static inline int ROUND_UP(int a
, int base
)
1062 return ((a
+base
-1)/base
)*base
;
1065 static inline int is_subarray(char *vers
)
1067 /* The version string for a 'subarray' (an array in a container)
1069 * /containername/componentname for normal read-write arrays
1070 * -containername/componentname for read-only arrays.
1071 * containername is e.g. md0, md_d1
1072 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1074 return (*vers
== '/' || *vers
== '-');
1078 #define dprintf(fmt, arg...) \
1079 fprintf(stderr, fmt, ##arg)
1081 #define dprintf(fmt, arg...) \
1082 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1086 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1090 ret
= vasprintf(strp
, fmt
, ap
);
1096 #define LEVEL_MULTIPATH (-4)
1097 #define LEVEL_LINEAR (-1)
1098 #define LEVEL_FAULTY (-5)
1100 /* kernel module doesn't know about these */
1101 #define LEVEL_CONTAINER (-100)
1102 #define LEVEL_UNSUPPORTED (-200)
1107 #define WriteTransient 0
1108 #define ReadTransient 1
1109 #define WritePersistent 2
1110 #define ReadPersistent 3
1111 #define WriteAll 4 /* doesn't go to device */
1112 #define ReadFixable 5
1115 #define ClearErrors 31
1116 #define ClearFaults 30
1118 #define AllPersist 100 /* internal use only */
1119 #define NoPersist 101
1121 #define ModeMask 0x1f
1129 #define minor(x) ((x)&0xff)
1130 #define major(x) (((x)>>8)&0xff)
1131 #define makedev(M,m) (((M)<<8) | (m))
1135 #define ALGORITHM_LEFT_ASYMMETRIC 0
1136 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1137 #define ALGORITHM_LEFT_SYMMETRIC 2
1138 #define ALGORITHM_RIGHT_SYMMETRIC 3
1140 /* Define non-rotating (raid4) algorithms. These allow
1141 * conversion of raid4 to raid5.
1143 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1144 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1146 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1147 * Firstly, the exact positioning of the parity block is slightly
1148 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1150 * Secondly, or order of datablocks over which the Q syndrome is computed
1152 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1153 * These layouts are from the DDFv1.2 spec.
1154 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1155 * leaves RLQ=3 as 'Vendor Specific'
1158 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1159 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1160 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1163 /* For every RAID5 algorithm we define a RAID6 algorithm
1164 * with exactly the same layout for data and parity, and
1165 * with the Q block always on the last device (N-1).
1166 * This allows trivial conversion from RAID5 to RAID6
1168 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1169 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1170 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1171 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1172 #define ALGORITHM_PARITY_0_6 20
1173 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1175 /* Define PATH_MAX in case we don't use glibc or standard library does
1176 * not have PATH_MAX defined. Assume max path length is 4K characters.
1179 #define PATH_MAX 4096