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
29 extern __off64_t lseek64
__P ((int __fd
, __off64_t __offset
, int __whence
));
30 #elif !defined(lseek64)
31 # if defined(__NO_STAT64) || __WORDSIZE != 32
32 # define lseek64 lseek
36 #include <sys/types.h>
49 /* dietlibc has deprecated random and srandom!! */
54 #include <linux/kdev_t.h>
55 /*#include <linux/fs.h> */
56 #include <sys/mount.h>
57 #include <asm/types.h>
58 #include <sys/ioctl.h>
60 #define MdpMinorShift 6
63 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
66 #define DEFAULT_CHUNK 512
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 * /run seems to have emerged as the best standard.
76 #define MAP_DIR "/run/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. Best is /var/run/mdadm as
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 "/run/mdadm"
94 #endif /* MDMON_DIR */
96 /* FAILED_SLOTS is where to save files storing recent removal of array
97 * member in order to allow future reuse of disk inserted in the same
98 * slot for array recovery
100 #ifndef FAILED_SLOTS_DIR
101 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
102 #endif /* FAILED_SLOTS */
110 /* Redhat don't like to #include <asm/byteorder.h>, and
111 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
112 * and there is no standard conversion function so... */
113 /* And dietlibc doesn't think byteswap is ok, so.. */
114 /* #include <byteswap.h> */
115 #define bswap_16(x) (((x) & 0x00ffU) << 8 | \
116 ((x) & 0xff00U) >> 8)
117 #define bswap_32(x) (((x) & 0x000000ffU) << 24 | \
118 ((x) & 0xff000000U) >> 24 | \
119 ((x) & 0x0000ff00U) << 8 | \
120 ((x) & 0x00ff0000U) >> 8)
121 #define bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
122 ((x) & 0xff00000000000000ULL) >> 56 | \
123 ((x) & 0x000000000000ff00ULL) << 40 | \
124 ((x) & 0x00ff000000000000ULL) >> 40 | \
125 ((x) & 0x0000000000ff0000ULL) << 24 | \
126 ((x) & 0x0000ff0000000000ULL) >> 24 | \
127 ((x) & 0x00000000ff000000ULL) << 8 | \
128 ((x) & 0x000000ff00000000ULL) >> 8)
130 #if !defined(__KLIBC__)
131 #if BYTE_ORDER == LITTLE_ENDIAN
132 #define __cpu_to_le16(_x) (unsigned int)(_x)
133 #define __cpu_to_le32(_x) (unsigned int)(_x)
134 #define __cpu_to_le64(_x) (unsigned long long)(_x)
135 #define __le16_to_cpu(_x) (unsigned int)(_x)
136 #define __le32_to_cpu(_x) (unsigned int)(_x)
137 #define __le64_to_cpu(_x) (unsigned long long)(_x)
139 #define __cpu_to_be16(_x) bswap_16(_x)
140 #define __cpu_to_be32(_x) bswap_32(_x)
141 #define __cpu_to_be64(_x) bswap_64(_x)
142 #define __be16_to_cpu(_x) bswap_16(_x)
143 #define __be32_to_cpu(_x) bswap_32(_x)
144 #define __be64_to_cpu(_x) bswap_64(_x)
145 #elif BYTE_ORDER == BIG_ENDIAN
146 #define __cpu_to_le16(_x) bswap_16(_x)
147 #define __cpu_to_le32(_x) bswap_32(_x)
148 #define __cpu_to_le64(_x) bswap_64(_x)
149 #define __le16_to_cpu(_x) bswap_16(_x)
150 #define __le32_to_cpu(_x) bswap_32(_x)
151 #define __le64_to_cpu(_x) bswap_64(_x)
153 #define __cpu_to_be16(_x) (unsigned int)(_x)
154 #define __cpu_to_be32(_x) (unsigned int)(_x)
155 #define __cpu_to_be64(_x) (unsigned long long)(_x)
156 #define __be16_to_cpu(_x) (unsigned int)(_x)
157 #define __be32_to_cpu(_x) (unsigned int)(_x)
158 #define __be64_to_cpu(_x) (unsigned long long)(_x)
160 # error "unknown endianness."
162 #endif /* __KLIBC__ */
165 * min()/max()/clamp() macros that also do
166 * strict type-checking.. See the
167 * "unnecessary" pointer comparison.
169 #define min(x, y) ({ \
170 typeof(x) _min1 = (x); \
171 typeof(y) _min2 = (y); \
172 (void) (&_min1 == &_min2); \
173 _min1 < _min2 ? _min1 : _min2; })
175 #define max(x, y) ({ \
176 typeof(x) _max1 = (x); \
177 typeof(y) _max2 = (y); \
178 (void) (&_max1 == &_max2); \
179 _max1 > _max2 ? _max1 : _max2; })
181 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
183 extern const char Name
[];
185 /* general information that might be extracted from a superblock */
187 mdu_array_info_t array
;
188 mdu_disk_info_t disk
;
192 unsigned long long data_offset
;
193 unsigned long long new_data_offset
;
194 unsigned long long component_size
; /* same as array.size, except in
195 * sectors and up to 64bits.
197 unsigned long long custom_array_size
; /* size for non-default sized
198 * arrays (in sectors)
201 #define VOLUME_RESHAPE 1
202 #define CONTAINER_RESHAPE 2
203 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
205 unsigned long long reshape_progress
;
206 int recovery_blocked
; /* for external metadata it
207 * indicates that there is
208 * reshape in progress in
210 * for native metadata it is
211 * reshape_active field mirror
213 /* During reshape we can sometimes change the data_offset to avoid
214 * over-writing still-valid data. We need to know if there is space.
215 * So getinfo_super will fill in space_before and space_after in sectors.
216 * data_offset can be increased or decreased by this amount.
218 unsigned long long space_before
, space_after
;
220 unsigned long long resync_start
; /* per-array resync position */
221 unsigned long long recovery_start
; /* per-device rebuild position */
222 #define MaxSector (~0ULL) /* resync/recovery complete position */
224 long bitmap_offset
; /* 0 == none, 1 == a file */
225 unsigned long safe_mode_delay
; /* ms delay to mark clean */
226 int new_level
, delta_disks
, new_layout
, new_chunk
;
228 unsigned long cache_size
; /* size of raid456 stripe cache*/
230 char text_version
[50];
232 int container_member
; /* for assembling external-metatdata arrays
233 * This is to be used internally by metadata
235 int container_enough
; /* flag external handlers can set to
236 * indicate that subarrays have not enough (-1),
237 * enough to start (0), or all expected disks (1) */
242 /* Device info for mdmon: */
247 #define DS_WRITE_MOSTLY 4
249 #define DS_BLOCKED 16
250 #define DS_REMOVE 1024
251 #define DS_UNBLOCK 2048
252 int prev_state
, curr_state
, next_state
;
264 struct supertype
*supertype
;
280 extern char short_options
[];
281 extern char short_bitmap_options
[];
282 extern char short_bitmap_auto_options
[];
283 extern struct option long_options
[];
284 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
286 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
288 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
290 /* for option that don't have short equivilents, we assign arbitrary
291 * numbers later than any 'short' character option.
293 enum special_options
{
352 enum prefix_standard
{
363 /* structures read from config file */
364 /* List of mddevice names and identifiers
365 * Identifiers can be:
367 * super-minor=decimal-minor-number-from-superblock
368 * devices=comma,separated,list,of,device,names,with,wildcards
370 * If multiple fields are present, the intersection of all matching
371 * devices is considered
373 #define UnSet (0xfffe)
383 char *devices
; /* comma separated list of device
384 * names with wild cards
389 struct supertype
*st
;
390 int autof
; /* 1 for normal, 2 for partitioned */
395 char *container
; /* /dev/whatever name of container, or
396 * uuid of container. You would expect
397 * this to be the 'devname' or UUID
398 * of some other entry.
400 char *member
; /* subarray within a container */
402 struct mddev_ident
*next
;
404 /* fields needed by different users of this structure */
405 int assembled
; /* set when assembly succeeds */
416 int require_homehost
;
445 unsigned long long size
;
448 /* List of device names - wildcards expanded */
451 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
453 * Not set for names read from .config
455 char writemostly
; /* 1 for 'set writemostly', 2 for 'clear writemostly' */
456 int used
; /* set when used */
457 long long data_offset
;
458 struct mddev_dev
*next
;
461 typedef struct mapping
{
471 char *pattern
; /* U or up, _ for down */
472 int percent
; /* -1 if no resync */
473 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
476 char * metadata_version
;
479 struct dev_member
*next
;
481 struct mdstat_ent
*next
;
484 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
485 extern void mdstat_close(void);
486 extern void free_mdstat(struct mdstat_ent
*ms
);
487 extern void mdstat_wait(int seconds
);
488 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
489 extern int mddev_busy(char *devnm
);
490 extern struct mdstat_ent
*mdstat_by_component(char *name
);
491 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
494 struct map_ent
*next
;
501 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
502 int uuid
[4], char *path
);
503 extern void map_remove(struct map_ent
**map
, char *devnm
);
504 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
506 static inline struct map_ent
*map_by_devnm(struct map_ent
**map
, char *name
)
510 static inline void map_free(struct map_ent
*map
)
514 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
515 extern void map_free(struct map_ent
*map
);
517 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
518 extern void map_read(struct map_ent
**melp
);
519 extern int map_write(struct map_ent
*mel
);
520 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
521 extern void map_add(struct map_ent
**melp
,
522 char *devnm
, char *metadata
, int uuid
[4], char *path
);
523 extern int map_lock(struct map_ent
**melp
);
524 extern void map_unlock(struct map_ent
**melp
);
525 extern void map_fork(void);
527 /* various details can be requested */
528 enum sysfs_read_flags
{
529 GET_LEVEL
= (1 << 0),
530 GET_LAYOUT
= (1 << 1),
531 GET_COMPONENT
= (1 << 2),
532 GET_CHUNK
= (1 << 3),
533 GET_CACHE
= (1 << 4),
534 GET_MISMATCH
= (1 << 5),
535 GET_VERSION
= (1 << 6),
536 GET_DISKS
= (1 << 7),
537 GET_DEGRADED
= (1 << 8),
538 GET_SAFEMODE
= (1 << 9),
539 GET_BITMAP_LOCATION
= (1 << 10),
541 GET_DEVS
= (1 << 20), /* gets role, major, minor */
542 GET_OFFSET
= (1 << 21),
543 GET_SIZE
= (1 << 22),
544 GET_STATE
= (1 << 23),
545 GET_ERROR
= (1 << 24),
548 /* If fd >= 0, get the array it is open on,
551 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
552 extern void sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
553 extern void sysfs_free(struct mdinfo
*sra
);
554 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
555 extern int sysfs_attr_match(const char *attr
, const char *str
);
556 extern int sysfs_match_word(const char *word
, char **list
);
557 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
558 char *name
, char *val
);
559 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
560 char *name
, unsigned long long val
);
561 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
562 char *name
, long long val
);
563 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
564 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
566 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
567 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
568 char *name
, unsigned long long *val
);
569 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
570 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
571 char *name
, unsigned long long *v1
, unsigned long long *v2
);
572 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
573 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
575 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
576 char *name
, char *val
, int size
);
577 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
578 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
579 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
580 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
581 extern int sysfs_unique_holder(char *devnm
, long rdev
);
582 extern int sysfs_freeze_array(struct mdinfo
*sra
);
583 extern int sysfs_wait(int fd
, int *msec
);
584 extern int load_sys(char *path
, char *buf
);
585 extern int reshape_prepare_fdlist(char *devname
,
589 unsigned long blocks
,
592 unsigned long long *offsets
);
593 extern void reshape_free_fdlist(int *fdlist
,
594 unsigned long long *offsets
,
596 extern int reshape_open_backup_file(char *backup
,
601 unsigned long long *offsets
,
604 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
605 unsigned int ndata
, unsigned int odata
);
606 extern char *locate_backup(char *name
);
607 extern char *make_backup(char *name
);
609 extern int save_stripes(int *source
, unsigned long long *offsets
,
610 int raid_disks
, int chunk_size
, int level
, int layout
,
611 int nwrites
, int *dest
,
612 unsigned long long start
, unsigned long long length
,
614 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
615 int raid_disks
, int chunk_size
, int level
, int layout
,
616 int source
, unsigned long long read_offset
,
617 unsigned long long start
, unsigned long long length
,
621 #define Sendmail "/usr/lib/sendmail -t"
624 #define SYSLOG_FACILITY LOG_DAEMON
626 extern char *map_num(mapping_t
*map
, int num
);
627 extern int map_name(mapping_t
*map
, char *name
);
628 extern mapping_t r5layout
[], r6layout
[], pers
[], modes
[], faultylayout
[];
630 extern char *map_dev_preferred(int major
, int minor
, int create
,
632 static inline char *map_dev(int major
, int minor
, int create
)
634 return map_dev_preferred(major
, minor
, create
, NULL
);
638 struct metadata_update
;
640 /* 'struct reshape' records the intermediate states of
642 * The starting geometry is converted to the 'before' geometry
643 * by at most an atomic level change. They could be the same.
644 * Similarly the 'after' geometry is converted to the final
645 * geometry by at most a level change.
646 * Note that 'before' and 'after' must have the same level.
647 * 'blocks' is the minimum number of sectors for a reshape unit.
648 * This will be a multiple of the stripe size in each of the
649 * 'before' and 'after' geometries.
650 * If 'blocks' is 0, no restriping is necessary.
651 * 'min_offset_change' is the minimum change to data_offset to
652 * allow the reshape to happen. It is at least the larger of
653 * the old and new chunk sizes, and typically the same as 'blocks'
654 * divided by number of data disks.
658 int parity
; /* number of parity blocks/devices */
663 unsigned long long backup_blocks
;
664 unsigned long long min_offset_change
;
665 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
666 unsigned long long new_size
; /* New size of array in sectors */
669 /* A superswitch provides entry point the a metadata handler.
671 * The superswitch primarily operates on some "metadata" that
672 * is accessed via the 'supertype'.
673 * This metadata has one of three possible sources.
674 * 1/ It is read from a single device. In this case it may not completely
675 * describe the array or arrays as some information might be on other
677 * 2/ It is read from all devices in a container. In this case all
678 * information is present.
679 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
680 * be complete once enough ->add_to_super calls have completed.
682 * When creating an array inside a container, the metadata will be
683 * formed by a combination of 2 and 3. The metadata or the array is read,
684 * then new information is added.
686 * The metadata must sometimes have a concept of a 'current' array
687 * and a 'current' device.
688 * The 'current' array is set by init_super to be the newly created array,
689 * or is set by super_by_fd when it finds it is looking at an array inside
692 * The 'current' device is either the device that the metadata was read from
693 * in case 1, or the last device added by add_to_super in case 3.
694 * Case 2 does not identify a 'current' device.
696 extern struct superswitch
{
698 /* Used to report details of metadata read from a component
699 * device. ->load_super has been called.
701 void (*examine_super
)(struct supertype
*st
, char *homehost
);
702 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
703 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
704 void (*export_examine_super
)(struct supertype
*st
);
705 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
706 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
708 /* Used to report details of an active array.
709 * ->load_super was possibly given a 'component' string.
711 void (*detail_super
)(struct supertype
*st
, char *homehost
);
712 void (*brief_detail_super
)(struct supertype
*st
);
713 void (*export_detail_super
)(struct supertype
*st
);
715 /* Optional: platform hardware / firmware details */
716 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
717 int (*export_detail_platform
)(int verbose
, char *controller_path
);
720 * to get uuid to storing in bitmap metadata
721 * and 'reshape' backup-data metadata
722 * To see if a device is being re-added to an array it was part of.
724 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
726 /* Extract generic details from metadata. This could be details about
727 * the container, or about an individual array within the container.
728 * The determination is made either by:
729 * load_super being given a 'component' string.
730 * validate_geometry determining what to create.
731 * The info includes both array information and device information.
732 * The particular device should be:
733 * The last device added by add_to_super
734 * The device the metadata was loaded from by load_super
735 * If 'map' is present, then it is an array raid_disks long
736 * (raid_disk must already be set and correct) and it is filled
737 * with 1 for slots that are thought to be active and 0 for slots which
738 * appear to be failed/missing.
739 * *info is zeroed out before data is added.
741 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
742 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
743 /* Check if the given metadata is flagged as belonging to "this"
744 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
746 int (*match_home
)(struct supertype
*st
, char *homehost
);
748 /* Make one of several generic modifications to metadata
749 * prior to assembly (or other times).
750 * sparc2.2 - first bug in early 0.90 metadata
751 * super-minor - change name of 0.90 metadata
752 * summaries - 'correct' any redundant data
753 * resync - mark array as dirty to trigger a resync.
754 * uuid - set new uuid - only 0.90 or 1.x
755 * name - change the name of the array (where supported)
756 * homehost - change which host this array is tied to.
757 * devicesize - If metadata is at start of device, change recorded
758 * device size to match actual device size
759 * byteorder - swap bytes for 0.90 metadata
761 * force-one - mark that device as uptodate, not old or failed.
762 * force-array - mark array as clean if it would not otherwise
764 * assemble - not sure how this is different from force-one...
765 * linear-grow-new - add a new device to a linear array, but don't
766 * change the size: so superblock still matches
767 * linear-grow-update - now change the size of the array.
768 * writemostly - set the WriteMostly1 bit in the superblock devflags
769 * readwrite - clear the WriteMostly1 bit in the superblock devflags
770 * no-bitmap - clear any record that a bitmap is present.
771 * bbl - add a bad-block-log if possible
772 * no-bbl - remove and bad-block-log is it is empty.
773 * revert-reshape - If a reshape is in progress, modify metadata so
774 * it will resume going in the opposite direction.
776 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
778 char *devname
, int verbose
,
779 int uuid_set
, char *homehost
);
781 /* Create new metadata for new array as described. This could
782 * be a new container, or an array in a pre-existing container.
783 * Also used to zero metadata prior to writing it to invalidate old
786 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
787 unsigned long long size
, char *name
,
788 char *homehost
, int *uuid
,
789 unsigned long long data_offset
);
791 /* update the metadata to include new device, either at create or
792 * when hot-adding a spare.
794 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
795 int fd
, char *devname
,
796 unsigned long long data_offset
);
797 /* update the metadata to delete a device,
800 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
802 /* Write metadata to one device when fixing problems or adding
805 int (*store_super
)(struct supertype
*st
, int fd
);
807 /* Write all metadata for this array.
809 int (*write_init_super
)(struct supertype
*st
);
810 /* Check if metadata read from one device is compatible with an array,
811 * used when assembling an array, or pseudo-assembling was with
812 * "--examine --brief"
813 * If "st" has not yet been loaded the superblock from, "tst" is
814 * moved in, otherwise the superblock in 'st' is compared with
817 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
);
818 /* Load metadata from a single device. If 'devname' is not NULL
819 * print error messages as appropriate */
820 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
821 /* 'fd' is a 'container' md array - load array metadata from the
824 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
825 /* If 'arg' is a valid name of this metadata type, allocate and
826 * return a 'supertype' for the particular minor version */
827 struct supertype
* (*match_metadata_desc
)(char *arg
);
828 /* If a device has the given size, and the data_offset has been
829 * requested - work out how much space is available for data.
830 * This involves adjusting for reserved space (e.g. bitmaps)
831 * and for any rounding.
832 * 'mdadm' only calls this for existing arrays where a possible
833 * spare is being added. However some super-handlers call it
834 * internally from validate_geometry when creating an array.
836 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
837 unsigned long long data_offset
);
838 /* This is similar to 'avail_size' in purpose, but is used for
839 * containers for which there is no 'component size' to compare.
840 * This reports that whole-device size which is a minimum
842 unsigned long long (*min_acceptable_spare_size
)(struct supertype
*st
);
843 /* Find somewhere to put a bitmap - possibly auto-size it - and
844 * update the metadata to record this. The array may be newly
845 * created, in which case data_size may be updated, or it might
846 * already exist. Metadata handler can know if init_super
847 * has been called, but not write_init_super.
849 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
850 int delay
, int write_behind
,
851 unsigned long long size
, int may_change
, int major
);
852 /* Seek 'fd' to start of write-intent-bitmap. Must be an
853 * md-native format bitmap
855 void (*locate_bitmap
)(struct supertype
*st
, int fd
);
856 /* if add_internal_bitmap succeeded for existing array, this
859 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
860 /* Free the superblock and any other allocated data */
861 void (*free_super
)(struct supertype
*st
);
863 /* validate_geometry is called with an st returned by
864 * match_metadata_desc.
865 * It should check that the geometry described is compatible with
866 * the metadata type. It will be called repeatedly as devices
867 * added to validate changing size and new devices. If there are
868 * inter-device dependencies, it should record sufficient details
869 * so these can be validated.
870 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
872 * 1: everything is OK
873 * 0: not OK for some reason - if 'verbose', then error was reported.
874 * -1: st->sb was NULL, 'subdev' is a member of a container of this
875 * type, but array is not acceptable for some reason
876 * message was reported even if verbose is 0.
878 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
880 int *chunk
, unsigned long long size
,
881 unsigned long long data_offset
,
882 char *subdev
, unsigned long long *freesize
,
885 /* Return a linked list of 'mdinfo' structures for all arrays
886 * in the container. For non-containers, it is like
887 * getinfo_super with an allocated mdinfo.*/
888 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
889 /* query the supertype for default geometry */
890 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
891 /* Permit subarray's to be deleted from inactive containers */
892 int (*kill_subarray
)(struct supertype
*st
); /* optional */
893 /* Permit subarray's to be modified */
894 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
895 char *update
, struct mddev_ident
*ident
); /* optional */
896 /* Check if reshape is supported for this external format.
897 * st is obtained from super_by_fd() where st->subarray[0] is
898 * initialized to indicate if reshape is being performed at the
899 * container or subarray level
901 #define APPLY_METADATA_CHANGES 1
902 #define ROLLBACK_METADATA_CHANGES 0
904 int (*reshape_super
)(struct supertype
*st
,
905 unsigned long long size
, int level
,
906 int layout
, int chunksize
, int raid_disks
,
907 int delta_disks
, char *backup
, char *dev
,
909 int verbose
); /* optional */
910 int (*manage_reshape
)( /* optional */
911 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
912 struct supertype
*st
, unsigned long blocks
,
913 int *fds
, unsigned long long *offsets
,
914 int dests
, int *destfd
, unsigned long long *destoffsets
);
917 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
920 /* Tell the metadata handler the current state of the array.
921 * This covers whether it is known to be consistent (no pending writes)
922 * and how far along a resync is known to have progressed
923 * (in a->resync_start).
924 * resync status is really irrelevant if the array is not consistent,
925 * but some metadata (DDF!) have a place to record the distinction.
926 * If 'consistent' is '2', then the array can mark it dirty if a
927 * resync/recovery/whatever is required, or leave it clean if not.
928 * Return value is 0 dirty (not consistent) and 1 if clean.
929 * it is only really important if consistent is passed in as '2'.
931 int (*set_array_state
)(struct active_array
*a
, int consistent
);
933 /* When the state of a device might have changed, we call set_disk to
934 * tell the metadata what the current state is.
935 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
937 * set_disk might be called when the state of the particular disk has
938 * not in fact changed.
940 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
941 void (*sync_metadata
)(struct supertype
*st
);
942 void (*process_update
)(struct supertype
*st
,
943 struct metadata_update
*update
);
944 /* Prepare updates allocates extra memory that might be
945 * needed. If the update cannot be understood, return 0.
947 int (*prepare_update
)(struct supertype
*st
,
948 struct metadata_update
*update
);
950 /* activate_spare will check if the array is degraded and, if it
951 * is, try to find some spare space in the container.
952 * On success, it add appropriate updates (For process_update) to
953 * to the 'updates' list and returns a list of 'mdinfo' identifying
954 * the device, or devices as there might be multiple missing
955 * devices and multiple spares available.
957 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
958 struct metadata_update
**updates
);
960 * Return statically allocated string that represents metadata specific
961 * controller domain of the disk. The domain is used in disk domain
962 * matching functions. Disks belong to the same domain if the they have
963 * the same domain from mdadm.conf and belong the same metadata domain.
964 * Returning NULL or not providing this handler means that metadata
965 * does not distinguish the differences between disks that belong to
966 * different controllers. They are in the domain specified by
967 * configuration file (mdadm.conf).
968 * In case when the metadata has the notion of domains based on disk
969 * it shall return NULL for disks that do not belong to the controller
970 * the supported domains. Such disks will form another domain and won't
971 * be mixed with supported ones.
973 const char *(*get_disk_controller_domain
)(const char *path
);
975 /* for external backup area */
976 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
978 /* validate container after assemble */
979 int (*validate_container
)(struct mdinfo
*info
);
981 int swapuuid
; /* true if uuid is bigending rather than hostendian */
983 const char *name
; /* canonical metadata name */
986 extern struct superswitch super0
, super1
;
987 extern struct superswitch super_imsm
, super_ddf
;
988 extern struct superswitch mbr
, gpt
;
990 struct metadata_update
{
993 void *space
; /* allocated space that monitor will use */
994 void **space_list
; /* list of allocated spaces that monitor can
995 * use or that it returned.
997 struct metadata_update
*next
;
1000 /* A supertype holds a particular collection of metadata.
1001 * It identifies the metadata type by the superswitch, and the particular
1002 * sub-version of that metadata type.
1003 * metadata read in or created is stored in 'sb' and 'info'.
1004 * There are also fields used by mdmon to track containers.
1006 * A supertype may refer to:
1007 * Just an array, possibly in a container
1008 * A container, not identifying any particular array
1009 * Info read from just one device, not yet fully describing the array/container.
1012 * A supertype is created by:
1018 struct superswitch
*ss
;
1021 char container_devnm
[32]; /* devnm of container */
1024 void *other
; /* Hack used to convert v0.90 to v1.0 */
1025 unsigned long long devsize
;
1026 unsigned long long data_offset
; /* used by v1.x only */
1027 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1028 HW/firmware related incompatability to load metadata.
1029 Used when examining metadata to display content of disk
1030 when user has no hw/firmare compatible system.
1032 struct metadata_update
*updates
;
1033 struct metadata_update
**update_tail
;
1035 /* extra stuff used by mdmon */
1036 struct active_array
*arrays
;
1037 int sock
; /* listen to external programs */
1038 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1046 struct mdinfo
*devs
;
1050 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1051 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1052 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1053 static inline struct supertype
*guess_super(int fd
) {
1054 return guess_super_type(fd
, guess_any
);
1056 extern struct supertype
*dup_super(struct supertype
*st
);
1057 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1058 extern int must_be_container(int fd
);
1059 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1060 void wait_for(char *dev
, int fd
);
1063 * Data structures for policy management.
1064 * Each device can have a policy structure that lists
1065 * various name/value pairs each possibly with a metadata associated.
1066 * The policy list is sorted by name/value/metadata
1069 struct dev_policy
*next
;
1070 char *name
; /* None of these strings are allocated. They are
1071 * all just references to strings which are known
1072 * to exist elsewhere.
1073 * name and metadata can be compared by address equality.
1075 const char *metadata
;
1079 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1081 /* iterate over the sublist starting at list, having the same
1082 * 'name' as 'list', and matching the given metadata (Where
1083 * NULL matches anything
1085 #define pol_for_each(item, list, _metadata) \
1087 item && item->name == list->name; \
1088 item = item->next) \
1089 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1093 * policy records read from mdadm are largely just name-value pairs.
1094 * The names are constants, not strdupped
1097 struct pol_rule
*next
;
1098 char *type
; /* rule_policy or rule_part */
1103 char *dups
; /* duplicates of 'value' with a partNN appended */
1107 extern char rule_policy
[], rule_part
[];
1108 extern char rule_path
[], rule_type
[];
1109 extern char type_part
[], type_disk
[];
1111 extern void policyline(char *line
, char *type
);
1112 extern void policy_add(char *type
, ...);
1113 extern void policy_free(void);
1115 extern struct dev_policy
*path_policy(char *path
, char *type
);
1116 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1117 extern struct dev_policy
*devid_policy(int devid
);
1118 extern void dev_policy_free(struct dev_policy
*p
);
1120 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1121 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1122 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1124 enum policy_action
{
1128 act_spare
, /* This only applies to bare devices */
1129 act_spare_same_slot
, /* this allows non-bare devices,
1130 * but only if recent removal */
1131 act_force_spare
, /* this allow non-bare devices in any case */
1135 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1136 enum policy_action want
);
1137 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1138 enum policy_action want
);
1141 struct domainlist
*next
;
1145 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1146 const char *metadata
);
1147 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1148 const char *metadata
);
1149 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1150 const char *metadata
);
1151 extern void domain_free(struct domainlist
*dl
);
1152 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1153 const char *metadata
);
1154 void domain_add(struct domainlist
**domp
, char *domain
);
1156 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1157 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1163 #define HAVE_NFTW we assume
1167 # include <features.h>
1168 # ifndef __UCLIBC_HAS_LFS__
1169 # define lseek64 lseek
1171 # ifndef __UCLIBC_HAS_FTW__
1181 #if defined(__KLIBC__)
1197 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1199 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1200 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1201 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1203 extern int Manage_subdevs(char *devname
, int fd
,
1204 struct mddev_dev
*devlist
, int verbose
, int test
,
1205 char *update
, int force
);
1206 extern int autodetect(void);
1207 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1208 extern int Grow_addbitmap(char *devname
, int fd
,
1209 struct context
*c
, struct shape
*s
);
1210 extern int Grow_reshape(char *devname
, int fd
,
1211 struct mddev_dev
*devlist
,
1212 unsigned long long data_offset
,
1213 struct context
*c
, struct shape
*s
);
1214 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1215 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1216 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1217 struct mdinfo
*info
, char *backup_file
,
1218 int forked
, int freeze_reshape
);
1220 extern int restore_backup(struct supertype
*st
,
1221 struct mdinfo
*content
,
1224 char **backup_filep
,
1226 extern int Grow_continue_command(char *devname
, int fd
,
1227 char *backup_file
, int verbose
);
1229 extern int Assemble(struct supertype
*st
, char *mddev
,
1230 struct mddev_ident
*ident
,
1231 struct mddev_dev
*devlist
,
1234 extern int Build(char *mddev
, struct mddev_dev
*devlist
,
1235 struct shape
*s
, struct context
*c
);
1237 extern int Create(struct supertype
*st
, char *mddev
,
1238 char *name
, int *uuid
,
1239 int subdevs
, struct mddev_dev
*devlist
,
1242 unsigned long long data_offset
);
1244 extern int Detail(char *dev
, struct context
*c
);
1245 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1246 extern int Query(char *dev
);
1247 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1248 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1249 struct supertype
*forcest
);
1250 extern int Monitor(struct mddev_dev
*devlist
,
1251 char *mailaddr
, char *alert_cmd
,
1253 int daemonise
, int oneshot
,
1254 int dosyslog
, char *pidfile
, int increments
,
1257 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1258 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1259 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1260 extern int Wait(char *dev
);
1261 extern int WaitClean(char *dev
, int sock
, int verbose
);
1262 extern int SetAction(char *dev
, char *action
);
1264 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1265 struct supertype
*st
);
1266 extern void RebuildMap(void);
1267 extern int IncrementalScan(struct context
*c
, char *devnm
);
1268 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1269 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1270 unsigned long chunksize
, unsigned long daemon_sleep
,
1271 unsigned long write_behind
,
1272 unsigned long long array_size
,
1274 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1275 extern int Write_rules(char *rule_name
);
1276 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1277 extern unsigned long bitmap_sectors(struct bitmap_super_s
*bsb
);
1278 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1279 struct supertype
*st
);
1280 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1281 struct supertype
*st
, int only
);
1283 extern int md_get_version(int fd
);
1284 extern int get_linux_version(void);
1285 extern int mdadm_version(char *version
);
1286 extern unsigned long long parse_size(char *size
);
1287 extern int parse_uuid(char *str
, int uuid
[4]);
1288 extern int parse_layout_10(char *layout
);
1289 extern int parse_layout_faulty(char *layout
);
1290 extern long parse_num(char *num
);
1291 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1292 extern int check_ext2(int fd
, char *name
);
1293 extern int check_reiser(int fd
, char *name
);
1294 extern int check_raid(int fd
, char *name
);
1295 extern int check_partitions(int fd
, char *dname
,
1296 unsigned long long freesize
,
1297 unsigned long long size
);
1299 extern int get_mdp_major(void);
1300 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1301 extern int dev_open(char *dev
, int flags
);
1302 extern int open_dev(char *devnm
);
1303 extern void reopen_mddev(int mdfd
);
1304 extern int open_dev_flags(char *devnm
, int flags
);
1305 extern int open_dev_excl(char *devnm
);
1306 extern int is_standard(char *dev
, int *nump
);
1307 extern int same_dev(char *one
, char *two
);
1308 extern int compare_paths (char* path1
,char* path2
);
1309 extern void enable_fds(int devices
);
1311 extern int parse_auto(char *str
, char *msg
, int config
);
1312 extern struct mddev_ident
*conf_get_ident(char *dev
);
1313 extern struct mddev_dev
*conf_get_devs(void);
1314 extern int conf_test_dev(char *devname
);
1315 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1316 extern struct createinfo
*conf_get_create_info(void);
1317 extern void set_conffile(char *file
);
1318 extern char *conf_get_mailaddr(void);
1319 extern char *conf_get_mailfrom(void);
1320 extern char *conf_get_program(void);
1321 extern char *conf_get_homehost(int *require_homehostp
);
1322 extern char *conf_get_homecluster(void);
1323 extern char *conf_line(FILE *file
);
1324 extern char *conf_word(FILE *file
, int allow_key
);
1325 extern void print_quoted(char *str
);
1326 extern void print_escape(char *str
);
1327 extern int use_udev(void);
1328 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1329 extern int conf_name_is_free(char *name
);
1330 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1331 extern int devname_matches(char *name
, char *match
);
1332 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1333 struct mdinfo
*info
,
1335 int verbose
, int *rvp
);
1336 extern int experimental(void);
1338 extern void free_line(char *line
);
1339 extern int match_oneof(char *devices
, char *devname
);
1340 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1341 extern const int uuid_zero
[4];
1342 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1343 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1344 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1345 extern char *fname_from_uuid(struct supertype
*st
,
1346 struct mdinfo
*info
, char *buf
, char sep
);
1347 extern unsigned long calc_csum(void *super
, int bytes
);
1348 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1350 extern int enough_fd(int fd
);
1351 extern int ask(char *mesg
);
1352 extern unsigned long long get_component_size(int fd
);
1353 extern void remove_partitions(int fd
);
1354 extern int test_partition(int fd
);
1355 extern int test_partition_from_id(dev_t id
);
1356 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1357 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1358 int chunksize
, unsigned long long devsize
);
1359 extern int flush_metadata_updates(struct supertype
*st
);
1360 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1361 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1362 struct mdinfo
*content
,
1364 char *chosen_name
, int *result
);
1366 #define INCR_UNSAFE 2
1367 #define INCR_ALREADY 4
1369 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1370 unsigned long long min_size
,
1371 struct domainlist
*domlist
,
1373 const char *metadata
, int get_one
);
1374 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1375 extern int add_disk(int mdfd
, struct supertype
*st
,
1376 struct mdinfo
*sra
, struct mdinfo
*info
);
1377 extern int remove_disk(int mdfd
, struct supertype
*st
,
1378 struct mdinfo
*sra
, struct mdinfo
*info
);
1379 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1380 unsigned long long min_recovery_start(struct mdinfo
*array
);
1382 extern char *human_size(long long bytes
);
1383 extern char *human_size_brief(long long bytes
, int prefix
);
1384 extern void print_r10_layout(int layout
);
1386 extern char *find_free_devnm(int use_partitions
);
1388 extern void put_md_name(char *name
);
1389 extern char *devid2kname(int devid
);
1390 extern char *devid2devnm(int devid
);
1391 extern int devnm2devid(char *devnm
);
1392 extern char *get_md_name(char *devnm
);
1394 extern char DefaultConfFile
[];
1396 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1398 /* values for 'trustworthy' */
1400 #define LOCAL_ANY 10
1403 extern int open_mddev(char *dev
, int report_errors
);
1404 extern int open_container(int fd
);
1405 extern int metadata_container_matches(char *metadata
, char *devnm
);
1406 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1407 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1408 extern int is_subarray_active(char *subarray
, char *devname
);
1409 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1410 extern struct superswitch
*version_to_superswitch(char *vers
);
1412 extern int mdmon_running(char *devnm
);
1413 extern int mdmon_pid(char *devnm
);
1414 extern int check_env(char *name
);
1415 extern __u32
random32(void);
1416 extern int start_mdmon(char *devnm
);
1418 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1419 struct supertype
*st
, unsigned long stripes
,
1420 int *fds
, unsigned long long *offsets
,
1421 int dests
, int *destfd
, unsigned long long *destoffsets
);
1422 void abort_reshape(struct mdinfo
*sra
);
1424 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1426 extern void fmt_devname(char *name
, int num
);
1427 extern char *stat2devnm(struct stat
*st
);
1428 extern char *fd2devnm(int fd
);
1430 extern int in_initrd(void);
1431 extern int get_cluster_name(char **name
);
1433 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1434 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1435 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1436 (ROUND_UP((unsigned long)(ptr), base)))
1438 static inline int is_subarray(char *vers
)
1440 /* The version string for a 'subarray' (an array in a container)
1442 * /containername/componentname for normal read-write arrays
1443 * -containername/componentname for arrays which mdmon must not
1444 * reconfigure. They might be read-only
1445 * or might be undergoing reshape etc.
1446 * containername is e.g. md0, md_d1
1447 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1449 return (*vers
== '/' || *vers
== '-');
1452 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1454 return &ent
->metadata_version
[10+strlen(container
)+1];
1458 #define dprintf(fmt, arg...) \
1459 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1460 #define dprintf_cont(fmt, arg...) \
1461 fprintf(stderr, fmt, ##arg)
1463 #define dprintf(fmt, arg...) \
1464 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1465 #define dprintf_cont(fmt, arg...) \
1466 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1470 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1474 ret
= vasprintf(strp
, fmt
, ap
);
1481 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1483 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1485 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1487 void *xmalloc(size_t len
);
1488 void *xrealloc(void *ptr
, size_t len
);
1489 void *xcalloc(size_t num
, size_t size
);
1490 char *xstrdup(const char *str
);
1492 #define LEVEL_MULTIPATH (-4)
1493 #define LEVEL_LINEAR (-1)
1494 #define LEVEL_FAULTY (-5)
1496 /* kernel module doesn't know about these */
1497 #define LEVEL_CONTAINER (-100)
1498 #define LEVEL_UNSUPPORTED (-200)
1500 /* the kernel does know about this one ... */
1501 #define LEVEL_NONE (-1000000)
1505 #define WriteTransient 0
1506 #define ReadTransient 1
1507 #define WritePersistent 2
1508 #define ReadPersistent 3
1509 #define WriteAll 4 /* doesn't go to device */
1510 #define ReadFixable 5
1513 #define ClearErrors 31
1514 #define ClearFaults 30
1516 #define AllPersist 100 /* internal use only */
1517 #define NoPersist 101
1519 #define ModeMask 0x1f
1526 #define minor(x) ((x)&0xff)
1527 #define major(x) (((x)>>8)&0xff)
1528 #define makedev(M,m) (((M)<<8) | (m))
1532 #define ALGORITHM_LEFT_ASYMMETRIC 0
1533 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1534 #define ALGORITHM_LEFT_SYMMETRIC 2
1535 #define ALGORITHM_RIGHT_SYMMETRIC 3
1537 /* Define non-rotating (raid4) algorithms. These allow
1538 * conversion of raid4 to raid5.
1540 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1541 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1543 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1544 * Firstly, the exact positioning of the parity block is slightly
1545 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1547 * Secondly, or order of datablocks over which the Q syndrome is computed
1549 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1550 * These layouts are from the DDFv1.2 spec.
1551 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1552 * leaves RLQ=3 as 'Vendor Specific'
1555 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1556 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1557 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1559 /* For every RAID5 algorithm we define a RAID6 algorithm
1560 * with exactly the same layout for data and parity, and
1561 * with the Q block always on the last device (N-1).
1562 * This allows trivial conversion from RAID5 to RAID6
1564 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1565 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1566 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1567 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1568 #define ALGORITHM_PARITY_0_6 20
1569 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1571 /* Define PATH_MAX in case we don't use glibc or standard library does
1572 * not have PATH_MAX defined. Assume max path length is 4K characters.
1575 #define PATH_MAX 4096
1578 #define RESYNC_NONE -1
1579 #define RESYNC_DELAYED -2
1580 #define RESYNC_PENDING -3
1581 #define RESYNC_UNKNOWN -4
1583 /* When using "GET_DISK_INFO" it isn't certain how high
1584 * we need to check. So we impose an absolute limit of
1585 * MAX_DISKS. This needs to be much more than the largest
1586 * number of devices any metadata can support. Currently
1587 * v1.x can support 1920
1589 #define MAX_DISKS 4096
1591 /* Sometimes the 'size' value passed needs to mean "Maximum".
1592 * In those cases with use MAX_SIZE
1596 /* We want to use unsigned numbers for sector counts, but need
1597 * a value for 'invalid'. Use '1'.
1599 #define INVALID_SECTORS 1
1600 /* And another special number needed for --data_offset=variable */
1601 #define VARIABLE_OFFSET 3