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
37 #include <sys/types.h>
52 /* Newer glibc requires sys/sysmacros.h directly for makedev() */
53 #include <sys/sysmacros.h>
56 /* dietlibc has deprecated random and srandom!! */
63 typedef uint64_t cmap_handle_t
;
65 #include <corosync/cmap.h>
72 #define LKF_NOQUEUE 0x00000001
74 #define EUNLOCK 0x10002
76 typedef void *dlm_lshandle_t
;
86 #include <linux/kdev_t.h>
87 /*#include <linux/fs.h> */
88 #include <sys/mount.h>
89 #include <asm/types.h>
90 #include <sys/ioctl.h>
92 #define MdpMinorShift 6
95 #define BLKGETSIZE64 _IOR(0x12,114,size_t) /* return device size in bytes (u64 *arg) */
98 #define DEFAULT_CHUNK 512
99 #define DEFAULT_BITMAP_CHUNK 4096
100 #define DEFAULT_BITMAP_DELAY 5
101 #define DEFAULT_MAX_WRITE_BEHIND 256
103 /* MAP_DIR should be somewhere that persists across the pivotroot
104 * from early boot to late boot.
105 * /run seems to have emerged as the best standard.
108 #define MAP_DIR "/run/mdadm"
110 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
111 * want something other than the default of "map"
114 #define MAP_FILE "map"
115 #endif /* MAP_FILE */
116 /* MDMON_DIR is where pid and socket files used for communicating
117 * with mdmon normally live. Best is /var/run/mdadm as
118 * mdmon is needed at early boot then it needs to write there prior
119 * to /var/run being mounted read/write, and it also then needs to
120 * persist beyond when /var/run is mounter read-only. So, to be
121 * safe, the default is somewhere that is read/write early in the
122 * boot process and stays up as long as possible during shutdown.
125 #define MDMON_DIR "/run/mdadm"
126 #endif /* MDMON_DIR */
128 /* FAILED_SLOTS is where to save files storing recent removal of array
129 * member in order to allow future reuse of disk inserted in the same
130 * slot for array recovery
132 #ifndef FAILED_SLOTS_DIR
133 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
134 #endif /* FAILED_SLOTS */
136 #ifndef MDMON_SERVICE
137 #define MDMON_SERVICE "mdmon"
138 #endif /* MDMON_SERVICE */
141 #define GROW_SERVICE "mdadm-grow-continue"
142 #endif /* GROW_SERVICE */
150 /* Redhat don't like to #include <asm/byteorder.h>, and
151 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
152 * and there is no standard conversion function so... */
153 /* And dietlibc doesn't think byteswap is ok, so.. */
154 /* #include <byteswap.h> */
155 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
156 ((x) & 0xff00U) >> 8)
157 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
158 ((x) & 0xff000000U) >> 24 | \
159 ((x) & 0x0000ff00U) << 8 | \
160 ((x) & 0x00ff0000U) >> 8)
161 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
162 ((x) & 0xff00000000000000ULL) >> 56 | \
163 ((x) & 0x000000000000ff00ULL) << 40 | \
164 ((x) & 0x00ff000000000000ULL) >> 40 | \
165 ((x) & 0x0000000000ff0000ULL) << 24 | \
166 ((x) & 0x0000ff0000000000ULL) >> 24 | \
167 ((x) & 0x00000000ff000000ULL) << 8 | \
168 ((x) & 0x000000ff00000000ULL) >> 8)
170 #if !defined(__KLIBC__)
171 #if BYTE_ORDER == LITTLE_ENDIAN
172 #define __cpu_to_le16(_x) (unsigned int)(_x)
173 #define __cpu_to_le32(_x) (unsigned int)(_x)
174 #define __cpu_to_le64(_x) (unsigned long long)(_x)
175 #define __le16_to_cpu(_x) (unsigned int)(_x)
176 #define __le32_to_cpu(_x) (unsigned int)(_x)
177 #define __le64_to_cpu(_x) (unsigned long long)(_x)
179 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
180 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
181 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
182 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
183 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
184 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
185 #elif BYTE_ORDER == BIG_ENDIAN
186 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
187 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
188 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
189 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
190 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
191 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
193 #define __cpu_to_be16(_x) (unsigned int)(_x)
194 #define __cpu_to_be32(_x) (unsigned int)(_x)
195 #define __cpu_to_be64(_x) (unsigned long long)(_x)
196 #define __be16_to_cpu(_x) (unsigned int)(_x)
197 #define __be32_to_cpu(_x) (unsigned int)(_x)
198 #define __be64_to_cpu(_x) (unsigned long long)(_x)
200 # error "unknown endianness."
202 #endif /* __KLIBC__ */
205 * Partially stolen from include/linux/unaligned/packed_struct.h
207 struct __una_u16
{ __u16 x
; } __attribute__ ((packed
));
208 struct __una_u32
{ __u32 x
; } __attribute__ ((packed
));
210 static inline __u16
__get_unaligned16(const void *p
)
212 const struct __una_u16
*ptr
= (const struct __una_u16
*)p
;
216 static inline __u32
__get_unaligned32(const void *p
)
218 const struct __una_u32
*ptr
= (const struct __una_u32
*)p
;
222 static inline void __put_unaligned16(__u16 val
, void *p
)
224 struct __una_u16
*ptr
= (struct __una_u16
*)p
;
228 static inline void __put_unaligned32(__u32 val
, void *p
)
230 struct __una_u32
*ptr
= (struct __una_u32
*)p
;
235 * Check at compile time that something is of a particular type.
236 * Always evaluates to 1 so you may use it easily in comparisons.
239 #define typecheck(type,x) \
241 typeof(x) __dummy2; \
242 (void)(&__dummy == &__dummy2); \
247 * These inlines deal with timer wrapping correctly.
249 * time_after(a,b) returns true if the time a is after time b.
252 #define time_after(a,b) \
253 (typecheck(unsigned int, a) && \
254 typecheck(unsigned int, b) && \
255 ((int)((b) - (a)) < 0))
257 #define time_before(a,b) time_after(b,a)
260 * min()/max()/clamp() macros that also do
261 * strict type-checking.. See the
262 * "unnecessary" pointer comparison.
264 #define min(x, y) ({ \
265 typeof(x) _min1 = (x); \
266 typeof(y) _min2 = (y); \
267 (void) (&_min1 == &_min2); \
268 _min1 < _min2 ? _min1 : _min2; })
270 #define max(x, y) ({ \
271 typeof(x) _max1 = (x); \
272 typeof(y) _max2 = (y); \
273 (void) (&_max1 == &_max2); \
274 _max1 > _max2 ? _max1 : _max2; })
276 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
278 extern const char Name
[];
281 unsigned long long sector
;
288 struct md_bb_entry
*entries
;
291 /* general information that might be extracted from a superblock */
293 mdu_array_info_t array
;
294 mdu_disk_info_t disk
;
298 unsigned long long data_offset
;
299 unsigned long long new_data_offset
;
300 unsigned long long component_size
; /* same as array.size, except in
301 * sectors and up to 64bits.
303 unsigned long long custom_array_size
; /* size for non-default sized
304 * arrays (in sectors)
307 #define VOLUME_RESHAPE 1
308 #define CONTAINER_RESHAPE 2
309 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
311 unsigned long long reshape_progress
;
312 int recovery_blocked
; /* for external metadata it
313 * indicates that there is
314 * reshape in progress in
316 * for native metadata it is
317 * reshape_active field mirror
319 int journal_device_required
;
323 CONSISTENCY_POLICY_UNKNOWN
,
324 CONSISTENCY_POLICY_NONE
,
325 CONSISTENCY_POLICY_RESYNC
,
326 CONSISTENCY_POLICY_BITMAP
,
327 CONSISTENCY_POLICY_JOURNAL
,
328 CONSISTENCY_POLICY_PPL
,
329 } consistency_policy
;
331 /* During reshape we can sometimes change the data_offset to avoid
332 * over-writing still-valid data. We need to know if there is space.
333 * So getinfo_super will fill in space_before and space_after in sectors.
334 * data_offset can be increased or decreased by this amount.
336 unsigned long long space_before
, space_after
;
338 unsigned long long resync_start
; /* per-array resync position */
339 unsigned long long recovery_start
; /* per-device rebuild position */
340 #define MaxSector (~0ULL) /* resync/recovery complete position */
342 long bitmap_offset
; /* 0 == none, 1 == a file */
343 unsigned int ppl_size
;
345 unsigned long long ppl_sector
;
346 unsigned long safe_mode_delay
; /* ms delay to mark clean */
347 int new_level
, delta_disks
, new_layout
, new_chunk
;
349 unsigned long cache_size
; /* size of raid456 stripe cache*/
351 char text_version
[50];
353 int container_member
; /* for assembling external-metatdata arrays
354 * This is to be used internally by metadata
356 int container_enough
; /* flag external handlers can set to
357 * indicate that subarrays have not enough (-1),
358 * enough to start (0), or all expected disks (1) */
363 /* Device info for mdmon: */
370 #define DS_WRITE_MOSTLY 4
372 #define DS_BLOCKED 16
373 #define DS_REMOVE 1024
374 #define DS_UNBLOCK 2048
375 int prev_state
, curr_state
, next_state
;
377 /* info read from sysfs */
401 struct supertype
*supertype
;
404 struct spare_criteria
{
405 unsigned long long min_size
;
406 unsigned int sector_size
;
422 extern char short_options
[];
423 extern char short_monitor_options
[];
424 extern char short_bitmap_options
[];
425 extern char short_bitmap_auto_options
[];
426 extern struct option long_options
[];
427 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
429 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
431 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
433 /* for option that don't have short equivilents, we assign arbitrary
434 * numbers later than any 'short' character option.
436 enum special_options
{
521 UOPT_LAYOUT_ORIGINAL
,
522 UOPT_LAYOUT_ALTERNATE
,
523 UOPT_LAYOUT_UNSPECIFIED
,
528 * Code specific options, cannot be set by the user
531 UOPT_SPEC_FORCE_ARRAY
,
533 UOPT_SPEC_LINEAR_GROW_NEW
,
534 UOPT_SPEC_LINEAR_GROW_UPDATE
,
535 UOPT_SPEC__RESHAPE_PROGRESS
,
536 UOPT_SPEC_WRITEMOSTLY
,
539 UOPT_SPEC_NOFAILFAST
,
540 UOPT_SPEC_REVERT_RESHAPE_NOBACKUP
,
543 extern void fprint_update_options(FILE *outf
, enum update_opt update_mode
);
545 enum prefix_standard
{
557 FlagDefault
, FlagSet
, FlagClear
,
560 /* structures read from config file */
561 /* List of mddevice names and identifiers
562 * Identifiers can be:
564 * super-minor=decimal-minor-number-from-superblock
565 * devices=comma,separated,list,of,device,names,with,wildcards
567 * If multiple fields are present, the intersection of all matching
568 * devices is considered
570 #define UnSet (0xfffe)
580 char *devices
; /* comma separated list of device
581 * names with wild cards
586 struct supertype
*st
;
587 int autof
; /* 1 for normal, 2 for partitioned */
592 char *container
; /* /dev/whatever name of container, or
593 * uuid of container. You would expect
594 * this to be the 'devname' or UUID
595 * of some other entry.
597 char *member
; /* subarray within a container */
599 struct mddev_ident
*next
;
601 /* fields needed by different users of this structure */
602 int assembled
; /* set when assembly succeeds */
614 int require_homehost
;
644 unsigned long long size
;
645 unsigned long long data_offset
;
646 int consistency_policy
;
649 /* List of device names - wildcards expanded */
652 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
654 * Not set for names read from .config
656 enum flag_mode writemostly
;
657 enum flag_mode failfast
;
658 int used
; /* set when used */
659 long long data_offset
;
660 struct mddev_dev
*next
;
663 typedef struct mapping
{
672 char *pattern
; /* U for up, _ for down */
673 int percent
; /* -1 if no resync */
674 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
677 char * metadata_version
;
680 struct dev_member
*next
;
682 struct mdstat_ent
*next
;
685 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
686 extern void mdstat_close(void);
687 extern void free_mdstat(struct mdstat_ent
*ms
);
688 extern int mdstat_wait(int seconds
);
689 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
690 extern int mddev_busy(char *devnm
);
691 extern struct mdstat_ent
*mdstat_by_component(char *name
);
692 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
695 struct map_ent
*next
;
702 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
703 int uuid
[4], char *path
);
704 extern void map_remove(struct map_ent
**map
, char *devnm
);
705 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
706 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
707 extern void map_free(struct map_ent
*map
);
708 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
709 extern void map_read(struct map_ent
**melp
);
710 extern int map_write(struct map_ent
*mel
);
711 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
712 extern void map_add(struct map_ent
**melp
,
713 char *devnm
, char *metadata
, int uuid
[4], char *path
);
714 extern int map_lock(struct map_ent
**melp
);
715 extern void map_unlock(struct map_ent
**melp
);
716 extern void map_fork(void);
718 /* various details can be requested */
719 enum sysfs_read_flags
{
720 GET_LEVEL
= (1 << 0),
721 GET_LAYOUT
= (1 << 1),
722 GET_COMPONENT
= (1 << 2),
723 GET_CHUNK
= (1 << 3),
724 GET_CACHE
= (1 << 4),
725 GET_MISMATCH
= (1 << 5),
726 GET_VERSION
= (1 << 6),
727 GET_DISKS
= (1 << 7),
728 GET_SAFEMODE
= (1 << 9),
729 GET_BITMAP_LOCATION
= (1 << 10),
731 GET_DEVS
= (1 << 20), /* gets role, major, minor */
732 GET_OFFSET
= (1 << 21),
733 GET_SIZE
= (1 << 22),
734 GET_STATE
= (1 << 23),
735 GET_ERROR
= (1 << 24),
736 GET_ARRAY_STATE
= (1 << 25),
737 GET_CONSISTENCY_POLICY
= (1 << 26),
738 GET_DEVS_ALL
= (1 << 27),
741 /* If fd >= 0, get the array it is open on,
744 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
745 extern int sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
746 extern void sysfs_init_dev(struct mdinfo
*mdi
, dev_t devid
);
747 extern void sysfs_free(struct mdinfo
*sra
);
748 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
749 extern int sysfs_attr_match(const char *attr
, const char *str
);
750 extern int sysfs_match_word(const char *word
, char **list
);
751 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
752 char *name
, char *val
);
753 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
754 char *name
, unsigned long long val
);
755 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
756 char *name
, long long val
);
757 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
758 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
760 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
761 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
762 char *name
, unsigned long long *val
);
763 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
764 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
765 char *name
, unsigned long long *v1
, unsigned long long *v2
);
766 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
767 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
769 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
770 char *name
, char *val
, int size
);
771 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
772 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
773 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
774 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
775 extern int sysfs_unique_holder(char *devnm
, long rdev
);
776 extern int sysfs_freeze_array(struct mdinfo
*sra
);
777 extern int sysfs_wait(int fd
, int *msec
);
778 extern int load_sys(char *path
, char *buf
, int len
);
779 extern int zero_disk_range(int fd
, unsigned long long sector
, size_t count
);
780 extern int reshape_prepare_fdlist(char *devname
,
784 unsigned long blocks
,
787 unsigned long long *offsets
);
788 extern void reshape_free_fdlist(int *fdlist
,
789 unsigned long long *offsets
,
791 extern int reshape_open_backup_file(char *backup
,
796 unsigned long long *offsets
,
799 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
800 unsigned int ndata
, unsigned int odata
);
801 extern char *locate_backup(char *name
);
802 extern char *make_backup(char *name
);
804 extern int save_stripes(int *source
, unsigned long long *offsets
,
805 int raid_disks
, int chunk_size
, int level
, int layout
,
806 int nwrites
, int *dest
,
807 unsigned long long start
, unsigned long long length
,
809 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
810 int raid_disks
, int chunk_size
, int level
, int layout
,
811 int source
, unsigned long long read_offset
,
812 unsigned long long start
, unsigned long long length
,
816 #define Sendmail "/usr/lib/sendmail -t"
819 #define SYSLOG_FACILITY LOG_DAEMON
820 extern char *map_num_s(mapping_t
*map
, int num
);
821 extern char *map_num(mapping_t
*map
, int num
);
822 extern int map_name(mapping_t
*map
, char *name
);
823 extern mapping_t r0layout
[], r5layout
[], r6layout
[],
824 pers
[], modes
[], faultylayout
[];
825 extern mapping_t consistency_policies
[], sysfs_array_states
[], update_options
[];
827 extern char *map_dev_preferred(int major
, int minor
, int create
,
829 static inline char *map_dev(int major
, int minor
, int create
)
831 return map_dev_preferred(major
, minor
, create
, NULL
);
835 * is_fd_valid() - check file descriptor.
836 * @fd: file descriptor.
838 * The function checks if @fd is nonnegative integer and shall be used only
839 * to verify open() result.
841 static inline int is_fd_valid(int fd
)
847 * is_level456() - check whether given level is between inclusive 4 and 6.
848 * @level: level to check.
850 * Return: true if condition is met, false otherwise
852 static inline bool is_level456(int level
)
854 return (level
>= 4 && level
<= 6);
858 * close_fd() - verify, close and unset file descriptor.
859 * @fd: pointer to file descriptor.
861 * The function closes and invalidates file descriptor if appropriative. It
862 * ignores incorrect file descriptor quitely to simplify error handling.
864 static inline void close_fd(int *fd
)
866 if (is_fd_valid(*fd
) && close(*fd
) == 0)
871 struct metadata_update
;
873 /* 'struct reshape' records the intermediate states of
875 * The starting geometry is converted to the 'before' geometry
876 * by at most an atomic level change. They could be the same.
877 * Similarly the 'after' geometry is converted to the final
878 * geometry by at most a level change.
879 * Note that 'before' and 'after' must have the same level.
880 * 'blocks' is the minimum number of sectors for a reshape unit.
881 * This will be a multiple of the stripe size in each of the
882 * 'before' and 'after' geometries.
883 * If 'blocks' is 0, no restriping is necessary.
884 * 'min_offset_change' is the minimum change to data_offset to
885 * allow the reshape to happen. It is at least the larger of
886 * the old and new chunk sizes, and typically the same as 'blocks'
887 * divided by number of data disks.
891 int parity
; /* number of parity blocks/devices */
896 unsigned long long backup_blocks
;
897 unsigned long long min_offset_change
;
898 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
899 unsigned long long new_size
; /* New size of array in sectors */
902 /* A superswitch provides entry point to a metadata handler.
904 * The superswitch primarily operates on some "metadata" that
905 * is accessed via the 'supertype'.
906 * This metadata has one of three possible sources.
907 * 1/ It is read from a single device. In this case it may not completely
908 * describe the array or arrays as some information might be on other
910 * 2/ It is read from all devices in a container. In this case all
911 * information is present.
912 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
913 * be complete once enough ->add_to_super calls have completed.
915 * When creating an array inside a container, the metadata will be
916 * formed by a combination of 2 and 3. The metadata or the array is read,
917 * then new information is added.
919 * The metadata must sometimes have a concept of a 'current' array
920 * and a 'current' device.
921 * The 'current' array is set by init_super to be the newly created array,
922 * or is set by super_by_fd when it finds it is looking at an array inside
925 * The 'current' device is either the device that the metadata was read from
926 * in case 1, or the last device added by add_to_super in case 3.
927 * Case 2 does not identify a 'current' device.
929 extern struct superswitch
{
931 /* Used to report details of metadata read from a component
932 * device. ->load_super has been called.
934 void (*examine_super
)(struct supertype
*st
, char *homehost
);
935 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
936 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
937 void (*export_examine_super
)(struct supertype
*st
);
938 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
939 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
941 /* Used to report details of an active array.
942 * ->load_super was possibly given a 'component' string.
944 void (*detail_super
)(struct supertype
*st
, char *homehost
,
946 void (*brief_detail_super
)(struct supertype
*st
, char *subarray
);
947 void (*export_detail_super
)(struct supertype
*st
);
949 /* Optional: platform hardware / firmware details */
950 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
951 int (*export_detail_platform
)(int verbose
, char *controller_path
);
954 * to get uuid to storing in bitmap metadata
955 * and 'reshape' backup-data metadata
956 * To see if a device is being re-added to an array it was part of.
958 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
960 /* Extract generic details from metadata. This could be details about
961 * the container, or about an individual array within the container.
962 * The determination is made either by:
963 * load_super being given a 'component' string.
964 * validate_geometry determining what to create.
965 * The info includes both array information and device information.
966 * The particular device should be:
967 * The last device added by add_to_super
968 * The device the metadata was loaded from by load_super
969 * If 'map' is present, then it is an array raid_disks long
970 * (raid_disk must already be set and correct) and it is filled
971 * with 1 for slots that are thought to be active and 0 for slots which
972 * appear to be failed/missing.
973 * *info is zeroed out before data is added.
975 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
976 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
977 /* Check if the given metadata is flagged as belonging to "this"
978 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
980 int (*match_home
)(struct supertype
*st
, char *homehost
);
982 /* Make one of several generic modifications to metadata
983 * prior to assembly (or other times).
984 * sparc2.2 - first bug in early 0.90 metadata
985 * super-minor - change name of 0.90 metadata
986 * summaries - 'correct' any redundant data
987 * resync - mark array as dirty to trigger a resync.
988 * uuid - set new uuid - only 0.90 or 1.x
989 * name - change the name of the array (where supported)
990 * homehost - change which host this array is tied to.
991 * devicesize - If metadata is at start of device, change recorded
992 * device size to match actual device size
993 * byteorder - swap bytes for 0.90 metadata
995 * force-one - mark that device as uptodate, not old or failed.
996 * force-array - mark array as clean if it would not otherwise
998 * assemble - not sure how this is different from force-one...
999 * linear-grow-new - add a new device to a linear array, but don't
1000 * change the size: so superblock still matches
1001 * linear-grow-update - now change the size of the array.
1002 * writemostly - set the WriteMostly1 bit in the superblock devflags
1003 * readwrite - clear the WriteMostly1 bit in the superblock devflags
1004 * failfast - set the FailFast1 bit in the superblock
1005 * nofailfast - clear the FailFast1 bit
1006 * no-bitmap - clear any record that a bitmap is present.
1007 * bbl - add a bad-block-log if possible
1008 * no-bbl - remove any bad-block-log is it is empty.
1009 * force-no-bbl - remove any bad-block-log even if empty.
1010 * revert-reshape - If a reshape is in progress, modify metadata so
1011 * it will resume going in the opposite direction.
1013 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
1014 enum update_opt update
,
1015 char *devname
, int verbose
,
1016 int uuid_set
, char *homehost
);
1018 /* Create new metadata for new array as described. This could
1019 * be a new container, or an array in a pre-existing container.
1020 * Also used to zero metadata prior to writing it to invalidate old
1023 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
1024 struct shape
*s
, char *name
,
1025 char *homehost
, int *uuid
,
1026 unsigned long long data_offset
);
1028 /* update the metadata to include new device, either at create or
1029 * when hot-adding a spare.
1031 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
1032 int fd
, char *devname
,
1033 unsigned long long data_offset
);
1034 /* update the metadata to delete a device,
1035 * when hot-removing.
1037 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
1039 /* Write metadata to one device when fixing problems or adding
1042 int (*store_super
)(struct supertype
*st
, int fd
);
1044 /* Write all metadata for this array.
1046 int (*write_init_super
)(struct supertype
*st
);
1047 /* Check if metadata read from one device is compatible with an array,
1048 * used when assembling an array, or pseudo-assembling was with
1049 * "--examine --brief"
1050 * If "st" has not yet been loaded the superblock from, "tst" is
1051 * moved in, otherwise the superblock in 'st' is compared with
1054 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
,
1056 /* Load metadata from a single device. If 'devname' is not NULL
1057 * print error messages as appropriate */
1058 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
1059 /* 'fd' is a 'container' md array - load array metadata from the
1062 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
1063 /* If 'arg' is a valid name of this metadata type, allocate and
1064 * return a 'supertype' for the particular minor version */
1065 struct supertype
* (*match_metadata_desc
)(char *arg
);
1066 /* If a device has the given size, and the data_offset has been
1067 * requested - work out how much space is available for data.
1068 * This involves adjusting for reserved space (e.g. bitmaps)
1069 * and for any rounding.
1070 * 'mdadm' only calls this for existing arrays where a possible
1071 * spare is being added. However some super-handlers call it
1072 * internally from validate_geometry when creating an array.
1074 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
1075 unsigned long long data_offset
);
1077 * Return spare criteria for array:
1078 * - minimum disk size can be used in array;
1079 * - sector size can be used in array.
1080 * Return values: 0 - for success and -EINVAL on error.
1082 int (*get_spare_criteria
)(struct supertype
*st
,
1083 struct spare_criteria
*sc
);
1084 /* Find somewhere to put a bitmap - possibly auto-size it - and
1085 * update the metadata to record this. The array may be newly
1086 * created, in which case data_size may be updated, or it might
1087 * already exist. Metadata handler can know if init_super
1088 * has been called, but not write_init_super.
1092 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
1093 int delay
, int write_behind
,
1094 unsigned long long size
, int may_change
, int major
);
1095 /* Perform additional setup required to activate a bitmap.
1097 int (*set_bitmap
)(struct supertype
*st
, struct mdinfo
*info
);
1098 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1099 * md-native format bitmap
1101 int (*locate_bitmap
)(struct supertype
*st
, int fd
, int node_num
);
1102 /* if add_internal_bitmap succeeded for existing array, this
1105 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
1106 /* Free the superblock and any other allocated data */
1107 void (*free_super
)(struct supertype
*st
);
1109 /* validate_geometry is called with an st returned by
1110 * match_metadata_desc.
1111 * It should check that the geometry described is compatible with
1112 * the metadata type. It will be called repeatedly as devices
1113 * added to validate changing size and new devices. If there are
1114 * inter-device dependencies, it should record sufficient details
1115 * so these can be validated.
1116 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1118 * 1: everything is OK
1119 * 0: not OK for some reason - if 'verbose', then error was reported.
1120 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1121 * type, but array is not acceptable for some reason
1122 * message was reported even if verbose is 0.
1124 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
1126 int *chunk
, unsigned long long size
,
1127 unsigned long long data_offset
,
1128 char *subdev
, unsigned long long *freesize
,
1129 int consistency_policy
, int verbose
);
1131 /* Return a linked list of 'mdinfo' structures for all arrays
1132 * in the container. For non-containers, it is like
1133 * getinfo_super with an allocated mdinfo.*/
1134 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
1135 /* query the supertype for default geometry */
1136 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
1137 /* Permit subarray's to be deleted from inactive containers */
1138 int (*kill_subarray
)(struct supertype
*st
,
1139 char *subarray_id
); /* optional */
1141 * update_subarray() - Permit subarray to be modified.
1143 * @subarray: Subarray name.
1144 * @update: Update option.
1145 * @ident: Optional identifiers.
1147 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
1148 enum update_opt update
, struct mddev_ident
*ident
);
1149 /* Check if reshape is supported for this external format.
1150 * st is obtained from super_by_fd() where st->subarray[0] is
1151 * initialized to indicate if reshape is being performed at the
1152 * container or subarray level
1154 #define APPLY_METADATA_CHANGES 1
1155 #define ROLLBACK_METADATA_CHANGES 0
1157 int (*reshape_super
)(struct supertype
*st
,
1158 unsigned long long size
, int level
,
1159 int layout
, int chunksize
, int raid_disks
,
1160 int delta_disks
, char *backup
, char *dev
,
1162 int verbose
); /* optional */
1163 int (*manage_reshape
)( /* optional */
1164 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1165 struct supertype
*st
, unsigned long blocks
,
1166 int *fds
, unsigned long long *offsets
,
1167 int dests
, int *destfd
, unsigned long long *destoffsets
);
1170 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
1173 /* Tell the metadata handler the current state of the array.
1174 * This covers whether it is known to be consistent (no pending writes)
1175 * and how far along a resync is known to have progressed
1176 * (in a->resync_start).
1177 * resync status is really irrelevant if the array is not consistent,
1178 * but some metadata (DDF!) have a place to record the distinction.
1179 * If 'consistent' is '2', then the array can mark it dirty if a
1180 * resync/recovery/whatever is required, or leave it clean if not.
1181 * Return value is 0 dirty (not consistent) and 1 if clean.
1182 * it is only really important if consistent is passed in as '2'.
1184 int (*set_array_state
)(struct active_array
*a
, int consistent
);
1186 /* When the state of a device might have changed, we call set_disk to
1187 * tell the metadata what the current state is.
1188 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1190 * set_disk might be called when the state of the particular disk has
1191 * not in fact changed.
1193 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
1194 void (*sync_metadata
)(struct supertype
*st
);
1195 void (*process_update
)(struct supertype
*st
,
1196 struct metadata_update
*update
);
1197 /* Prepare updates allocates extra memory that might be
1198 * needed. If the update cannot be understood, return 0.
1200 int (*prepare_update
)(struct supertype
*st
,
1201 struct metadata_update
*update
);
1203 /* activate_spare will check if the array is degraded and, if it
1204 * is, try to find some spare space in the container.
1205 * On success, it add appropriate updates (For process_update) to
1206 * to the 'updates' list and returns a list of 'mdinfo' identifying
1207 * the device, or devices as there might be multiple missing
1208 * devices and multiple spares available.
1210 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
1211 struct metadata_update
**updates
);
1213 * Return statically allocated string that represents metadata specific
1214 * controller domain of the disk. The domain is used in disk domain
1215 * matching functions. Disks belong to the same domain if the they have
1216 * the same domain from mdadm.conf and belong the same metadata domain.
1217 * Returning NULL or not providing this handler means that metadata
1218 * does not distinguish the differences between disks that belong to
1219 * different controllers. They are in the domain specified by
1220 * configuration file (mdadm.conf).
1221 * In case when the metadata has the notion of domains based on disk
1222 * it shall return NULL for disks that do not belong to the controller
1223 * the supported domains. Such disks will form another domain and won't
1224 * be mixed with supported ones.
1226 const char *(*get_disk_controller_domain
)(const char *path
);
1228 /* for external backup area */
1229 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
1231 /* validate container after assemble */
1232 int (*validate_container
)(struct mdinfo
*info
);
1234 /* write initial empty PPL on device */
1235 int (*write_init_ppl
)(struct supertype
*st
, struct mdinfo
*info
, int fd
);
1237 /* validate ppl before assemble */
1238 int (*validate_ppl
)(struct supertype
*st
, struct mdinfo
*info
,
1239 struct mdinfo
*disk
);
1241 /* records new bad block in metadata */
1242 int (*record_bad_block
)(struct active_array
*a
, int n
,
1243 unsigned long long sector
, int length
);
1245 /* clears bad block from metadata */
1246 int (*clear_bad_block
)(struct active_array
*a
, int n
,
1247 unsigned long long sector
, int length
);
1249 /* get list of bad blocks from metadata */
1250 struct md_bb
*(*get_bad_blocks
)(struct active_array
*a
, int n
);
1252 int swapuuid
; /* true if uuid is bigending rather than hostendian */
1254 const char *name
; /* canonical metadata name */
1257 extern struct superswitch super0
, super1
;
1258 extern struct superswitch super_imsm
, super_ddf
;
1259 extern struct superswitch mbr
, gpt
;
1261 struct metadata_update
{
1264 void *space
; /* allocated space that monitor will use */
1265 void **space_list
; /* list of allocated spaces that monitor can
1266 * use or that it returned.
1268 struct metadata_update
*next
;
1271 /* A supertype holds a particular collection of metadata.
1272 * It identifies the metadata type by the superswitch, and the particular
1273 * sub-version of that metadata type.
1274 * metadata read in or created is stored in 'sb' and 'info'.
1275 * There are also fields used by mdmon to track containers.
1277 * A supertype may refer to:
1278 * Just an array, possibly in a container
1279 * A container, not identifying any particular array
1280 * Info read from just one device, not yet fully describing the array/container.
1283 * A supertype is created by:
1289 struct superswitch
*ss
;
1292 char container_devnm
[32]; /* devnm of container */
1295 void *other
; /* Hack used to convert v0.90 to v1.0 */
1296 unsigned long long devsize
;
1297 unsigned long long data_offset
; /* used by v1.x only */
1298 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1299 HW/firmware related incompatability to load metadata.
1300 Used when examining metadata to display content of disk
1301 when user has no hw/firmare compatible system.
1303 struct metadata_update
*updates
;
1304 struct metadata_update
**update_tail
;
1306 /* extra stuff used by mdmon */
1307 struct active_array
*arrays
;
1308 int sock
; /* listen to external programs */
1309 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1317 struct mdinfo
*devs
;
1321 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1322 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1323 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1324 static inline struct supertype
*guess_super(int fd
) {
1325 return guess_super_type(fd
, guess_any
);
1327 extern struct supertype
*dup_super(struct supertype
*st
);
1328 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1329 extern int get_dev_sector_size(int fd
, char *dname
, unsigned int *sectsizep
);
1330 extern int must_be_container(int fd
);
1331 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1332 extern int dev_sector_size_from_id(dev_t id
, unsigned int *size
);
1333 void wait_for(char *dev
, int fd
);
1336 * Data structures for policy management.
1337 * Each device can have a policy structure that lists
1338 * various name/value pairs each possibly with a metadata associated.
1339 * The policy list is sorted by name/value/metadata
1342 struct dev_policy
*next
;
1343 char *name
; /* None of these strings are allocated. They are
1344 * all just references to strings which are known
1345 * to exist elsewhere.
1346 * name and metadata can be compared by address equality.
1348 const char *metadata
;
1352 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1354 /* iterate over the sublist starting at list, having the same
1355 * 'name' as 'list', and matching the given metadata (Where
1356 * NULL matches anything
1358 #define pol_for_each(item, list, _metadata) \
1360 item && item->name == list->name; \
1361 item = item->next) \
1362 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1366 * policy records read from mdadm are largely just name-value pairs.
1367 * The names are constants, not strdupped
1370 struct pol_rule
*next
;
1371 char *type
; /* rule_policy or rule_part */
1376 char *dups
; /* duplicates of 'value' with a partNN appended */
1380 extern char rule_policy
[], rule_part
[];
1381 extern char rule_path
[], rule_type
[];
1382 extern char type_part
[], type_disk
[];
1384 extern void policyline(char *line
, char *type
);
1385 extern void policy_add(char *type
, ...);
1386 extern void policy_free(void);
1388 extern struct dev_policy
*path_policy(char **paths
, char *type
);
1389 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1390 extern struct dev_policy
*devid_policy(int devid
);
1391 extern void dev_policy_free(struct dev_policy
*p
);
1393 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1394 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1395 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1397 enum policy_action
{
1401 act_spare
, /* This only applies to bare devices */
1402 act_spare_same_slot
, /* this allows non-bare devices,
1403 * but only if recent removal */
1404 act_force_spare
, /* this allow non-bare devices in any case */
1408 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1409 enum policy_action want
);
1410 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1411 enum policy_action want
);
1414 struct domainlist
*next
;
1418 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1419 const char *metadata
);
1420 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1421 const char *metadata
);
1422 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1423 const char *metadata
);
1424 extern void domain_free(struct domainlist
*dl
);
1425 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1426 const char *metadata
);
1427 void domain_add(struct domainlist
**domp
, char *domain
);
1429 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1430 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1432 extern void sysfs_rules_apply(char *devnm
, struct mdinfo
*dev
);
1433 extern void sysfsline(char *line
);
1439 #define HAVE_NFTW we assume
1443 # include <features.h>
1444 # ifndef __UCLIBC_HAS_LFS__
1445 # define lseek64 lseek
1447 # ifndef __UCLIBC_HAS_FTW__
1457 #if defined(__KLIBC__)
1473 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1475 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1476 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1477 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1479 extern int Manage_subdevs(char *devname
, int fd
,
1480 struct mddev_dev
*devlist
, int verbose
, int test
,
1481 char *update
, int force
);
1482 extern int autodetect(void);
1483 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1484 extern int Grow_addbitmap(char *devname
, int fd
,
1485 struct context
*c
, struct shape
*s
);
1486 extern int Grow_reshape(char *devname
, int fd
,
1487 struct mddev_dev
*devlist
,
1488 struct context
*c
, struct shape
*s
);
1489 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1490 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1491 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1492 struct mdinfo
*info
, char *backup_file
,
1493 int forked
, int freeze_reshape
);
1494 extern int Grow_consistency_policy(char *devname
, int fd
,
1495 struct context
*c
, struct shape
*s
);
1497 extern int restore_backup(struct supertype
*st
,
1498 struct mdinfo
*content
,
1501 char **backup_filep
,
1503 extern int Grow_continue_command(char *devname
, int fd
,
1504 char *backup_file
, int verbose
);
1506 extern int Assemble(struct supertype
*st
, char *mddev
,
1507 struct mddev_ident
*ident
,
1508 struct mddev_dev
*devlist
,
1511 extern int Build(char *mddev
, struct mddev_dev
*devlist
,
1512 struct shape
*s
, struct context
*c
);
1514 extern int Create(struct supertype
*st
, char *mddev
,
1515 char *name
, int *uuid
,
1516 int subdevs
, struct mddev_dev
*devlist
,
1520 extern int Detail(char *dev
, struct context
*c
);
1521 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1522 extern int Query(char *dev
);
1523 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1524 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1525 struct supertype
*forcest
);
1526 extern int Monitor(struct mddev_dev
*devlist
,
1527 char *mailaddr
, char *alert_cmd
,
1529 int daemonise
, int oneshot
,
1530 int dosyslog
, char *pidfile
, int increments
,
1533 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1534 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1535 extern int Update_subarray(char *dev
, char *subarray
, char *update
, struct mddev_ident
*ident
, int quiet
);
1536 extern int Wait(char *dev
);
1537 extern int WaitClean(char *dev
, int verbose
);
1538 extern int SetAction(char *dev
, char *action
);
1540 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1541 struct supertype
*st
);
1542 extern void RebuildMap(void);
1543 extern int IncrementalScan(struct context
*c
, char *devnm
);
1544 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1545 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1546 unsigned long chunksize
, unsigned long daemon_sleep
,
1547 unsigned long write_behind
,
1548 unsigned long long array_size
,
1550 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1551 extern int IsBitmapDirty(char *filename
);
1552 extern int Write_rules(char *rule_name
);
1553 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1555 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1556 static inline unsigned long long
1557 bitmap_bits(unsigned long long array_size
, unsigned long chunksize
)
1559 return (array_size
* 512 + chunksize
- 1) / chunksize
;
1562 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1563 struct supertype
*st
);
1564 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1565 struct supertype
*st
, int only
);
1567 int md_array_valid(int fd
);
1568 int md_array_active(int fd
);
1569 int md_array_is_active(struct mdinfo
*info
);
1570 int md_get_array_info(int fd
, struct mdu_array_info_s
*array
);
1571 int md_set_array_info(int fd
, struct mdu_array_info_s
*array
);
1572 int md_get_disk_info(int fd
, struct mdu_disk_info_s
*disk
);
1573 extern int get_linux_version(void);
1574 extern int mdadm_version(char *version
);
1575 extern unsigned long long parse_size(char *size
);
1576 extern int parse_uuid(char *str
, int uuid
[4]);
1577 int default_layout(struct supertype
*st
, int level
, int verbose
);
1578 extern int is_near_layout_10(int layout
);
1579 extern int parse_layout_10(char *layout
);
1580 extern int parse_layout_faulty(char *layout
);
1581 extern int parse_num(int *dest
, char *num
);
1582 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1583 extern int check_ext2(int fd
, char *name
);
1584 extern int check_reiser(int fd
, char *name
);
1585 extern int check_raid(int fd
, char *name
);
1586 extern int check_partitions(int fd
, char *dname
,
1587 unsigned long long freesize
,
1588 unsigned long long size
);
1589 extern int fstat_is_blkdev(int fd
, char *devname
, dev_t
*rdev
);
1590 extern int stat_is_blkdev(char *devname
, dev_t
*rdev
);
1592 extern bool is_dev_alive(char *path
);
1593 extern int get_mdp_major(void);
1594 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1595 extern bool is_bit_set(int *val
, unsigned char index
);
1596 extern int dev_open(char *dev
, int flags
);
1597 extern int open_dev(char *devnm
);
1598 extern void reopen_mddev(int mdfd
);
1599 extern int open_dev_flags(char *devnm
, int flags
);
1600 extern int open_dev_excl(char *devnm
);
1601 extern int is_standard(char *dev
, int *nump
);
1602 extern int same_dev(char *one
, char *two
);
1603 extern int compare_paths (char* path1
,char* path2
);
1604 extern void enable_fds(int devices
);
1605 extern void manage_fork_fds(int close_all
);
1606 extern int continue_via_systemd(char *devnm
, char *service_name
);
1608 extern void ident_init(struct mddev_ident
*ident
);
1610 extern int parse_auto(char *str
, char *msg
, int config
);
1611 extern struct mddev_ident
*conf_get_ident(char *dev
);
1612 extern struct mddev_dev
*conf_get_devs(void);
1613 extern int conf_test_dev(char *devname
);
1614 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1615 extern struct createinfo
*conf_get_create_info(void);
1616 extern void set_conffile(char *file
);
1617 extern char *conf_get_mailaddr(void);
1618 extern char *conf_get_mailfrom(void);
1619 extern char *conf_get_program(void);
1620 extern char *conf_get_homehost(int *require_homehostp
);
1621 extern char *conf_get_homecluster(void);
1622 extern int conf_get_monitor_delay(void);
1623 extern char *conf_line(FILE *file
);
1624 extern char *conf_word(FILE *file
, int allow_key
);
1625 extern void print_quoted(char *str
);
1626 extern void print_escape(char *str
);
1627 extern int use_udev(void);
1628 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1629 extern int conf_name_is_free(char *name
);
1630 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1631 extern int devname_matches(char *name
, char *match
);
1632 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1633 struct mdinfo
*info
,
1635 int verbose
, int *rvp
);
1637 extern void free_line(char *line
);
1638 extern int match_oneof(char *devices
, char *devname
);
1639 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1640 extern const int uuid_zero
[4];
1641 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1642 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1643 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1644 extern char *fname_from_uuid(struct supertype
*st
,
1645 struct mdinfo
*info
, char *buf
, char sep
);
1646 extern unsigned long calc_csum(void *super
, int bytes
);
1647 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1649 extern int ask(char *mesg
);
1650 extern unsigned long long get_component_size(int fd
);
1651 extern void remove_partitions(int fd
);
1652 extern int test_partition(int fd
);
1653 extern int test_partition_from_id(dev_t id
);
1654 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1655 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1656 int chunksize
, unsigned long long devsize
);
1657 extern int flush_metadata_updates(struct supertype
*st
);
1658 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1659 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1660 struct mdinfo
*content
,
1662 char *chosen_name
, int *result
);
1664 #define INCR_UNSAFE 2
1665 #define INCR_ALREADY 4
1667 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1668 struct spare_criteria
*criteria
,
1669 struct domainlist
*domlist
,
1671 const char *metadata
, int get_one
);
1672 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1673 extern int add_disk(int mdfd
, struct supertype
*st
,
1674 struct mdinfo
*sra
, struct mdinfo
*info
);
1675 extern int remove_disk(int mdfd
, struct supertype
*st
,
1676 struct mdinfo
*sra
, struct mdinfo
*info
);
1677 extern int hot_remove_disk(int mdfd
, unsigned long dev
, int force
);
1678 extern int sys_hot_remove_disk(int statefd
, int force
);
1679 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1680 unsigned long long min_recovery_start(struct mdinfo
*array
);
1682 extern char *human_size(long long bytes
);
1683 extern char *human_size_brief(long long bytes
, int prefix
);
1684 extern void print_r10_layout(int layout
);
1686 extern char *find_free_devnm(int use_partitions
);
1688 extern void put_md_name(char *name
);
1689 extern char *devid2kname(dev_t devid
);
1690 extern char *devid2devnm(dev_t devid
);
1691 extern dev_t
devnm2devid(char *devnm
);
1692 extern char *get_md_name(char *devnm
);
1694 extern char DefaultConfFile
[];
1696 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1697 char *chosen
, int block_udev
);
1698 /* values for 'trustworthy' */
1700 #define LOCAL_ANY 10
1703 extern int open_mddev(char *dev
, int report_errors
);
1704 extern int is_mddev(char *dev
);
1705 extern int open_container(int fd
);
1706 extern int metadata_container_matches(char *metadata
, char *devnm
);
1707 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1708 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1709 extern int is_subarray_active(char *subarray
, char *devname
);
1710 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1711 extern struct superswitch
*version_to_superswitch(char *vers
);
1713 extern int mdmon_running(char *devnm
);
1714 extern int mdmon_pid(char *devnm
);
1715 extern int check_env(char *name
);
1716 extern __u32
random32(void);
1717 extern void random_uuid(__u8
*buf
);
1718 extern int start_mdmon(char *devnm
);
1720 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1721 struct supertype
*st
, unsigned long stripes
,
1722 int *fds
, unsigned long long *offsets
,
1723 int dests
, int *destfd
, unsigned long long *destoffsets
);
1724 void abort_reshape(struct mdinfo
*sra
);
1726 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1728 extern char *stat2kname(struct stat
*st
);
1729 extern char *fd2kname(int fd
);
1730 extern char *stat2devnm(struct stat
*st
);
1731 bool stat_is_md_dev(struct stat
*st
);
1732 extern char *fd2devnm(int fd
);
1733 extern void udev_block(char *devnm
);
1734 extern void udev_unblock(void);
1736 extern int in_initrd(void);
1739 void *cmap_handle
; /* corosync lib related */
1741 int (*initialize
)(cmap_handle_t
*handle
);
1742 int (*get_string
)(cmap_handle_t handle
,
1745 int (*finalize
)(cmap_handle_t handle
);
1748 extern void set_cmap_hooks(void);
1749 extern void set_hooks(void);
1752 void *dlm_handle
; /* dlm lib related */
1754 dlm_lshandle_t (*create_lockspace
)(const char *name
,
1756 dlm_lshandle_t (*open_lockspace
)(const char *name
);
1757 int (*release_lockspace
)(const char *name
, dlm_lshandle_t ls
,
1759 int (*ls_lock
)(dlm_lshandle_t lockspace
, uint32_t mode
,
1760 struct dlm_lksb
*lksb
, uint32_t flags
,
1761 const void *name
, unsigned int namelen
,
1762 uint32_t parent
, void (*astaddr
) (void *astarg
),
1763 void *astarg
, void (*bastaddr
) (void *astarg
),
1765 int (*ls_unlock_wait
)(dlm_lshandle_t lockspace
, uint32_t lkid
,
1766 uint32_t flags
, struct dlm_lksb
*lksb
);
1767 int (*ls_get_fd
)(dlm_lshandle_t ls
);
1768 int (*dispatch
)(int fd
);
1771 extern int get_cluster_name(char **name
);
1772 extern int dlm_funs_ready(void);
1773 extern int cluster_get_dlmlock(void);
1774 extern int cluster_release_dlmlock(void);
1775 extern void set_dlm_hooks(void);
1777 #define MSEC_TO_NSEC(msec) ((msec) * 1000000)
1778 #define USEC_TO_NSEC(usec) ((usec) * 1000)
1779 extern void sleep_for(unsigned int sec
, long nsec
, bool wake_after_interrupt
);
1781 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1782 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1783 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1784 (ROUND_UP((unsigned long)(ptr), base)))
1786 static inline int is_subarray(char *vers
)
1788 /* The version string for a 'subarray' (an array in a container)
1790 * /containername/componentname for normal read-write arrays
1791 * -containername/componentname for arrays which mdmon must not
1792 * reconfigure. They might be read-only
1793 * or might be undergoing reshape etc.
1794 * containername is e.g. md0, md_d1
1795 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1797 return (*vers
== '/' || *vers
== '-');
1800 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1802 return &ent
->metadata_version
[10+strlen(container
)+1];
1806 * signal_s() - Wrapper for sigaction() with signal()-like interface.
1807 * @sig: The signal to set the signal handler to.
1808 * @handler: The signal handler.
1810 * Return: previous handler or SIG_ERR on failure.
1812 static inline sighandler_t
signal_s(int sig
, sighandler_t handler
)
1814 struct sigaction new_act
;
1815 struct sigaction old_act
;
1817 new_act
.sa_handler
= handler
;
1818 new_act
.sa_flags
= 0;
1820 if (sigaction(sig
, &new_act
, &old_act
) == 0)
1821 return old_act
.sa_handler
;
1827 #define dprintf(fmt, arg...) \
1828 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1829 #define dprintf_cont(fmt, arg...) \
1830 fprintf(stderr, fmt, ##arg)
1832 #define dprintf(fmt, arg...) \
1833 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1834 #define dprintf_cont(fmt, arg...) \
1835 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1838 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1842 ret
= vasprintf(strp
, fmt
, ap
);
1849 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1851 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1853 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1855 void *xmalloc(size_t len
);
1856 void *xrealloc(void *ptr
, size_t len
);
1857 void *xcalloc(size_t num
, size_t size
);
1858 char *xstrdup(const char *str
);
1860 #define LEVEL_MULTIPATH (-4)
1861 #define LEVEL_LINEAR (-1)
1862 #define LEVEL_FAULTY (-5)
1864 /* kernel module doesn't know about these */
1865 #define LEVEL_CONTAINER (-100)
1866 #define LEVEL_UNSUPPORTED (-200)
1868 /* the kernel does know about this one ... */
1869 #define LEVEL_NONE (-1000000)
1873 #define WriteTransient 0
1874 #define ReadTransient 1
1875 #define WritePersistent 2
1876 #define ReadPersistent 3
1877 #define WriteAll 4 /* doesn't go to device */
1878 #define ReadFixable 5
1881 #define ClearErrors 31
1882 #define ClearFaults 30
1884 #define AllPersist 100 /* internal use only */
1885 #define NoPersist 101
1887 #define ModeMask 0x1f
1894 #define minor(x) ((x)&0xff)
1895 #define major(x) (((x)>>8)&0xff)
1896 #define makedev(M,m) (((M)<<8) | (m))
1900 RAID0_ORIG_LAYOUT
= 1,
1901 RAID0_ALT_MULTIZONE_LAYOUT
= 2,
1905 #define ALGORITHM_LEFT_ASYMMETRIC 0
1906 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1907 #define ALGORITHM_LEFT_SYMMETRIC 2
1908 #define ALGORITHM_RIGHT_SYMMETRIC 3
1910 /* Define non-rotating (raid4) algorithms. These allow
1911 * conversion of raid4 to raid5.
1913 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1914 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1916 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1917 * Firstly, the exact positioning of the parity block is slightly
1918 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1920 * Secondly, or order of datablocks over which the Q syndrome is computed
1922 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1923 * These layouts are from the DDFv1.2 spec.
1924 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1925 * leaves RLQ=3 as 'Vendor Specific'
1928 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1929 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1930 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1932 /* For every RAID5 algorithm we define a RAID6 algorithm
1933 * with exactly the same layout for data and parity, and
1934 * with the Q block always on the last device (N-1).
1935 * This allows trivial conversion from RAID5 to RAID6
1937 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1938 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1939 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1940 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1941 #define ALGORITHM_PARITY_0_6 20
1942 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1944 /* Define PATH_MAX in case we don't use glibc or standard library does
1945 * not have PATH_MAX defined. Assume max path length is 4K characters.
1948 #define PATH_MAX 4096
1951 #define RESYNC_NONE -1
1952 #define RESYNC_DELAYED -2
1953 #define RESYNC_PENDING -3
1954 #define RESYNC_REMOTE -4
1955 #define RESYNC_UNKNOWN -5
1957 /* When using "GET_DISK_INFO" it isn't certain how high
1958 * we need to check. So we impose an absolute limit of
1959 * MAX_DISKS. This needs to be much more than the largest
1960 * number of devices any metadata can support. Currently
1961 * v1.x can support 1920
1963 #define MAX_DISKS 4096
1965 /* Sometimes the 'size' value passed needs to mean "Maximum".
1966 * In those cases with use MAX_SIZE
1970 /* We want to use unsigned numbers for sector counts, but need
1971 * a value for 'invalid'. Use '1'.
1973 #define INVALID_SECTORS 1
1974 /* And another special number needed for --data_offset=variable */
1975 #define VARIABLE_OFFSET 3
1978 * This is true for native and DDF, IMSM allows 16.
1980 #define MD_NAME_MAX 32
1983 * is_container() - check if @level is &LEVEL_CONTAINER
1984 * @level: level value
1987 * 1 if level is equal to &LEVEL_CONTAINER, 0 otherwise.
1989 static inline int is_container(const int level
)
1991 if (level
== LEVEL_CONTAINER
)