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 /* DEV_NUM_PREF is a subpath to numbered MD devices, e.g. /dev/md1 or directory name.
104 * DEV_NUM_PREF_LEN is a length with Null byte excluded.
107 #define DEV_NUM_PREF "/dev/md"
108 #define DEV_NUM_PREF_LEN (sizeof(DEV_NUM_PREF) - 1)
109 #endif /* DEV_NUM_PREF */
111 /* DEV_MD_DIR points to named MD devices directory.
112 * DEV_MD_DIR_LEN is a length with Null byte excluded.
115 #define DEV_MD_DIR "/dev/md/"
116 #define DEV_MD_DIR_LEN (sizeof(DEV_MD_DIR) - 1)
117 #endif /* DEV_MD_DIR */
119 /* MAP_DIR should be somewhere that persists across the pivotroot
120 * from early boot to late boot.
121 * /run seems to have emerged as the best standard.
124 #define MAP_DIR "/run/mdadm"
126 /* MAP_FILE is what we name the map file we put in MAP_DIR, in case you
127 * want something other than the default of "map"
130 #define MAP_FILE "map"
131 #endif /* MAP_FILE */
132 /* MDMON_DIR is where pid and socket files used for communicating
133 * with mdmon normally live. Best is /var/run/mdadm as
134 * mdmon is needed at early boot then it needs to write there prior
135 * to /var/run being mounted read/write, and it also then needs to
136 * persist beyond when /var/run is mounter read-only. So, to be
137 * safe, the default is somewhere that is read/write early in the
138 * boot process and stays up as long as possible during shutdown.
141 #define MDMON_DIR "/run/mdadm"
142 #endif /* MDMON_DIR */
144 /* FAILED_SLOTS is where to save files storing recent removal of array
145 * member in order to allow future reuse of disk inserted in the same
146 * slot for array recovery
148 #ifndef FAILED_SLOTS_DIR
149 #define FAILED_SLOTS_DIR "/run/mdadm/failed-slots"
150 #endif /* FAILED_SLOTS */
152 #ifndef MDMON_SERVICE
153 #define MDMON_SERVICE "mdmon"
154 #endif /* MDMON_SERVICE */
157 #define GROW_SERVICE "mdadm-grow-continue"
158 #endif /* GROW_SERVICE */
166 /* Redhat don't like to #include <asm/byteorder.h>, and
167 * some time include <linux/byteorder/xxx_endian.h> isn't enough,
168 * and there is no standard conversion function so... */
169 /* And dietlibc doesn't think byteswap is ok, so.. */
170 /* #include <byteswap.h> */
171 #define __mdadm_bswap_16(x) (((x) & 0x00ffU) << 8 | \
172 ((x) & 0xff00U) >> 8)
173 #define __mdadm_bswap_32(x) (((x) & 0x000000ffU) << 24 | \
174 ((x) & 0xff000000U) >> 24 | \
175 ((x) & 0x0000ff00U) << 8 | \
176 ((x) & 0x00ff0000U) >> 8)
177 #define __mdadm_bswap_64(x) (((x) & 0x00000000000000ffULL) << 56 | \
178 ((x) & 0xff00000000000000ULL) >> 56 | \
179 ((x) & 0x000000000000ff00ULL) << 40 | \
180 ((x) & 0x00ff000000000000ULL) >> 40 | \
181 ((x) & 0x0000000000ff0000ULL) << 24 | \
182 ((x) & 0x0000ff0000000000ULL) >> 24 | \
183 ((x) & 0x00000000ff000000ULL) << 8 | \
184 ((x) & 0x000000ff00000000ULL) >> 8)
186 #if !defined(__KLIBC__)
187 #if BYTE_ORDER == LITTLE_ENDIAN
188 #define __cpu_to_le16(_x) (unsigned int)(_x)
189 #define __cpu_to_le32(_x) (unsigned int)(_x)
190 #define __cpu_to_le64(_x) (unsigned long long)(_x)
191 #define __le16_to_cpu(_x) (unsigned int)(_x)
192 #define __le32_to_cpu(_x) (unsigned int)(_x)
193 #define __le64_to_cpu(_x) (unsigned long long)(_x)
195 #define __cpu_to_be16(_x) __mdadm_bswap_16(_x)
196 #define __cpu_to_be32(_x) __mdadm_bswap_32(_x)
197 #define __cpu_to_be64(_x) __mdadm_bswap_64(_x)
198 #define __be16_to_cpu(_x) __mdadm_bswap_16(_x)
199 #define __be32_to_cpu(_x) __mdadm_bswap_32(_x)
200 #define __be64_to_cpu(_x) __mdadm_bswap_64(_x)
201 #elif BYTE_ORDER == BIG_ENDIAN
202 #define __cpu_to_le16(_x) __mdadm_bswap_16(_x)
203 #define __cpu_to_le32(_x) __mdadm_bswap_32(_x)
204 #define __cpu_to_le64(_x) __mdadm_bswap_64(_x)
205 #define __le16_to_cpu(_x) __mdadm_bswap_16(_x)
206 #define __le32_to_cpu(_x) __mdadm_bswap_32(_x)
207 #define __le64_to_cpu(_x) __mdadm_bswap_64(_x)
209 #define __cpu_to_be16(_x) (unsigned int)(_x)
210 #define __cpu_to_be32(_x) (unsigned int)(_x)
211 #define __cpu_to_be64(_x) (unsigned long long)(_x)
212 #define __be16_to_cpu(_x) (unsigned int)(_x)
213 #define __be32_to_cpu(_x) (unsigned int)(_x)
214 #define __be64_to_cpu(_x) (unsigned long long)(_x)
216 # error "unknown endianness."
218 #endif /* __KLIBC__ */
221 * Partially stolen from include/linux/unaligned/packed_struct.h
223 struct __una_u16
{ __u16 x
; } __attribute__ ((packed
));
224 struct __una_u32
{ __u32 x
; } __attribute__ ((packed
));
226 static inline __u16
__get_unaligned16(const void *p
)
228 const struct __una_u16
*ptr
= (const struct __una_u16
*)p
;
232 static inline __u32
__get_unaligned32(const void *p
)
234 const struct __una_u32
*ptr
= (const struct __una_u32
*)p
;
238 static inline void __put_unaligned16(__u16 val
, void *p
)
240 struct __una_u16
*ptr
= (struct __una_u16
*)p
;
244 static inline void __put_unaligned32(__u32 val
, void *p
)
246 struct __una_u32
*ptr
= (struct __una_u32
*)p
;
251 * Check at compile time that something is of a particular type.
252 * Always evaluates to 1 so you may use it easily in comparisons.
255 #define typecheck(type,x) \
257 typeof(x) __dummy2; \
258 (void)(&__dummy == &__dummy2); \
263 * These inlines deal with timer wrapping correctly.
265 * time_after(a,b) returns true if the time a is after time b.
268 #define time_after(a,b) \
269 (typecheck(unsigned int, a) && \
270 typecheck(unsigned int, b) && \
271 ((int)((b) - (a)) < 0))
273 #define time_before(a,b) time_after(b,a)
276 * min()/max()/clamp() macros that also do
277 * strict type-checking.. See the
278 * "unnecessary" pointer comparison.
280 #define min(x, y) ({ \
281 typeof(x) _min1 = (x); \
282 typeof(y) _min2 = (y); \
283 (void) (&_min1 == &_min2); \
284 _min1 < _min2 ? _min1 : _min2; })
286 #define max(x, y) ({ \
287 typeof(x) _max1 = (x); \
288 typeof(y) _max2 = (y); \
289 (void) (&_max1 == &_max2); \
290 _max1 > _max2 ? _max1 : _max2; })
292 #define ARRAY_SIZE(x) (sizeof(x)/sizeof(x[0]))
294 #define KIB_TO_BYTES(x) ((x) << 10)
295 #define SEC_TO_BYTES(x) ((x) << 9)
297 extern const char Name
[];
300 unsigned long long sector
;
307 struct md_bb_entry
*entries
;
310 /* general information that might be extracted from a superblock */
312 mdu_array_info_t array
;
313 mdu_disk_info_t disk
;
317 unsigned long long data_offset
;
318 unsigned long long new_data_offset
;
319 unsigned long long component_size
; /* same as array.size, except in
320 * sectors and up to 64bits.
322 unsigned long long custom_array_size
; /* size for non-default sized
323 * arrays (in sectors)
326 #define VOLUME_RESHAPE 1
327 #define CONTAINER_RESHAPE 2
328 #define RESHAPE_NO_BACKUP 16 /* Mask 'or'ed in */
330 unsigned long long reshape_progress
;
331 int recovery_blocked
; /* for external metadata it
332 * indicates that there is
333 * reshape in progress in
335 * for native metadata it is
336 * reshape_active field mirror
338 int journal_device_required
;
342 CONSISTENCY_POLICY_UNKNOWN
,
343 CONSISTENCY_POLICY_NONE
,
344 CONSISTENCY_POLICY_RESYNC
,
345 CONSISTENCY_POLICY_BITMAP
,
346 CONSISTENCY_POLICY_JOURNAL
,
347 CONSISTENCY_POLICY_PPL
,
348 } consistency_policy
;
350 /* During reshape we can sometimes change the data_offset to avoid
351 * over-writing still-valid data. We need to know if there is space.
352 * So getinfo_super will fill in space_before and space_after in sectors.
353 * data_offset can be increased or decreased by this amount.
355 unsigned long long space_before
, space_after
;
357 unsigned long long resync_start
; /* per-array resync position */
358 unsigned long long recovery_start
; /* per-device rebuild position */
359 #define MaxSector (~0ULL) /* resync/recovery complete position */
361 long bitmap_offset
; /* 0 == none, 1 == a file */
362 unsigned int ppl_size
;
364 unsigned long long ppl_sector
;
365 unsigned long safe_mode_delay
; /* ms delay to mark clean */
366 int new_level
, delta_disks
, new_layout
, new_chunk
;
368 unsigned long cache_size
; /* size of raid456 stripe cache*/
370 char text_version
[50];
372 int container_member
; /* for assembling external-metatdata arrays
373 * This is to be used internally by metadata
375 int container_enough
; /* flag external handlers can set to
376 * indicate that subarrays have not enough (-1),
377 * enough to start (0), or all expected disks (1) */
382 /* Device info for mdmon: */
389 #define DS_WRITE_MOSTLY 4
391 #define DS_BLOCKED 16
392 #define DS_REMOVE 1024
393 #define DS_UNBLOCK 2048
394 int prev_state
, curr_state
, next_state
;
396 /* info read from sysfs */
420 struct supertype
*supertype
;
423 struct spare_criteria
{
424 unsigned long long min_size
;
425 unsigned int sector_size
;
441 extern char short_options
[];
442 extern char short_monitor_options
[];
443 extern char short_bitmap_options
[];
444 extern char short_bitmap_auto_options
[];
445 extern struct option long_options
[];
446 extern char Version
[], Usage
[], Help
[], OptionHelp
[],
448 Help_create
[], Help_build
[], Help_assemble
[], Help_grow
[],
450 Help_manage
[], Help_misc
[], Help_monitor
[], Help_config
[];
452 /* for option that don't have short equivilents, we assign arbitrary
453 * numbers later than any 'short' character option.
455 enum special_options
{
541 UOPT_LAYOUT_ORIGINAL
,
542 UOPT_LAYOUT_ALTERNATE
,
543 UOPT_LAYOUT_UNSPECIFIED
,
548 * Code specific options, cannot be set by the user
551 UOPT_SPEC_FORCE_ARRAY
,
553 UOPT_SPEC_LINEAR_GROW_NEW
,
554 UOPT_SPEC_LINEAR_GROW_UPDATE
,
555 UOPT_SPEC__RESHAPE_PROGRESS
,
556 UOPT_SPEC_WRITEMOSTLY
,
559 UOPT_SPEC_NOFAILFAST
,
560 UOPT_SPEC_REVERT_RESHAPE_NOBACKUP
,
563 extern void fprint_update_options(FILE *outf
, enum update_opt update_mode
);
565 enum prefix_standard
{
577 FlagDefault
, FlagSet
, FlagClear
,
580 /* structures read from config file */
581 /* List of mddevice names and identifiers
582 * Identifiers can be:
584 * super-minor=decimal-minor-number-from-superblock
585 * devices=comma,separated,list,of,device,names,with,wildcards
587 * If multiple fields are present, the intersection of all matching
588 * devices is considered
590 #define UnSet (0xfffe)
600 char *devices
; /* comma separated list of device
601 * names with wild cards
606 struct supertype
*st
;
607 int autof
; /* 1 for normal, 2 for partitioned */
612 char *container
; /* /dev/whatever name of container, or
613 * uuid of container. You would expect
614 * this to be the 'devname' or UUID
615 * of some other entry.
617 char *member
; /* subarray within a container */
619 struct mddev_ident
*next
;
621 /* fields needed by different users of this structure */
622 int assembled
; /* set when assembly succeeds */
634 int require_homehost
;
639 enum update_opt update
;
665 unsigned long long size
;
666 unsigned long long data_offset
;
667 int consistency_policy
;
670 /* List of device names - wildcards expanded */
673 int disposition
; /* 'a' for add, 'r' for remove, 'f' for fail,
675 * Not set for names read from .config
677 enum flag_mode writemostly
;
678 enum flag_mode failfast
;
679 int used
; /* set when used */
680 long long data_offset
;
681 struct mddev_dev
*next
;
684 typedef struct mapping
{
693 char *pattern
; /* U for up, _ for down */
694 int percent
; /* -1 if no resync */
695 int resync
; /* 3 if check, 2 if reshape, 1 if resync, 0 if recovery */
698 char * metadata_version
;
701 struct dev_member
*next
;
703 struct mdstat_ent
*next
;
706 extern struct mdstat_ent
*mdstat_read(int hold
, int start
);
707 extern void mdstat_close(void);
708 extern void free_mdstat(struct mdstat_ent
*ms
);
709 extern int mdstat_wait(int seconds
);
710 extern void mdstat_wait_fd(int fd
, const sigset_t
*sigmask
);
711 extern int mddev_busy(char *devnm
);
712 extern struct mdstat_ent
*mdstat_by_component(char *name
);
713 extern struct mdstat_ent
*mdstat_by_subdev(char *subdev
, char *container
);
716 struct map_ent
*next
;
723 extern int map_update(struct map_ent
**mpp
, char *devnm
, char *metadata
,
724 int uuid
[4], char *path
);
725 extern void map_remove(struct map_ent
**map
, char *devnm
);
726 extern struct map_ent
*map_by_uuid(struct map_ent
**map
, int uuid
[4]);
727 extern struct map_ent
*map_by_devnm(struct map_ent
**map
, char *devnm
);
728 extern void map_free(struct map_ent
*map
);
729 extern struct map_ent
*map_by_name(struct map_ent
**map
, char *name
);
730 extern void map_read(struct map_ent
**melp
);
731 extern int map_write(struct map_ent
*mel
);
732 extern void map_delete(struct map_ent
**mapp
, char *devnm
);
733 extern void map_add(struct map_ent
**melp
,
734 char *devnm
, char *metadata
, int uuid
[4], char *path
);
735 extern int map_lock(struct map_ent
**melp
);
736 extern void map_unlock(struct map_ent
**melp
);
737 extern void map_fork(void);
739 /* various details can be requested */
740 enum sysfs_read_flags
{
741 GET_LEVEL
= (1 << 0),
742 GET_LAYOUT
= (1 << 1),
743 GET_COMPONENT
= (1 << 2),
744 GET_CHUNK
= (1 << 3),
745 GET_CACHE
= (1 << 4),
746 GET_MISMATCH
= (1 << 5),
747 GET_VERSION
= (1 << 6),
748 GET_DISKS
= (1 << 7),
749 GET_SAFEMODE
= (1 << 9),
750 GET_BITMAP_LOCATION
= (1 << 10),
752 GET_DEVS
= (1 << 20), /* gets role, major, minor */
753 GET_OFFSET
= (1 << 21),
754 GET_SIZE
= (1 << 22),
755 GET_STATE
= (1 << 23),
756 GET_ERROR
= (1 << 24),
757 GET_ARRAY_STATE
= (1 << 25),
758 GET_CONSISTENCY_POLICY
= (1 << 26),
759 GET_DEVS_ALL
= (1 << 27),
762 /* If fd >= 0, get the array it is open on,
765 extern int sysfs_open(char *devnm
, char *devname
, char *attr
);
766 extern int sysfs_init(struct mdinfo
*mdi
, int fd
, char *devnm
);
767 extern void sysfs_init_dev(struct mdinfo
*mdi
, dev_t devid
);
768 extern void sysfs_free(struct mdinfo
*sra
);
769 extern struct mdinfo
*sysfs_read(int fd
, char *devnm
, unsigned long options
);
770 extern int sysfs_attr_match(const char *attr
, const char *str
);
771 extern int sysfs_match_word(const char *word
, char **list
);
772 extern int sysfs_set_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
773 char *name
, char *val
);
774 extern int sysfs_set_num(struct mdinfo
*sra
, struct mdinfo
*dev
,
775 char *name
, unsigned long long val
);
776 extern int sysfs_set_num_signed(struct mdinfo
*sra
, struct mdinfo
*dev
,
777 char *name
, long long val
);
778 extern int sysfs_uevent(struct mdinfo
*sra
, char *event
);
779 extern int sysfs_get_fd(struct mdinfo
*sra
, struct mdinfo
*dev
,
781 extern int sysfs_fd_get_ll(int fd
, unsigned long long *val
);
782 extern int sysfs_get_ll(struct mdinfo
*sra
, struct mdinfo
*dev
,
783 char *name
, unsigned long long *val
);
784 extern int sysfs_fd_get_two(int fd
, unsigned long long *v1
, unsigned long long *v2
);
785 extern int sysfs_get_two(struct mdinfo
*sra
, struct mdinfo
*dev
,
786 char *name
, unsigned long long *v1
, unsigned long long *v2
);
787 extern int sysfs_fd_get_str(int fd
, char *val
, int size
);
788 extern int sysfs_attribute_available(struct mdinfo
*sra
, struct mdinfo
*dev
,
790 extern int sysfs_get_str(struct mdinfo
*sra
, struct mdinfo
*dev
,
791 char *name
, char *val
, int size
);
792 extern int sysfs_set_safemode(struct mdinfo
*sra
, unsigned long ms
);
793 extern int sysfs_set_array(struct mdinfo
*info
, int vers
);
794 extern int sysfs_add_disk(struct mdinfo
*sra
, struct mdinfo
*sd
, int resume
);
795 extern int sysfs_disk_to_scsi_id(int fd
, __u32
*id
);
796 extern int sysfs_unique_holder(char *devnm
, long rdev
);
797 extern int sysfs_freeze_array(struct mdinfo
*sra
);
798 extern int sysfs_wait(int fd
, int *msec
);
799 extern int load_sys(char *path
, char *buf
, int len
);
800 extern int zero_disk_range(int fd
, unsigned long long sector
, size_t count
);
801 extern int reshape_prepare_fdlist(char *devname
,
805 unsigned long blocks
,
808 unsigned long long *offsets
);
809 extern void reshape_free_fdlist(int *fdlist
,
810 unsigned long long *offsets
,
812 extern int reshape_open_backup_file(char *backup
,
817 unsigned long long *offsets
,
820 extern unsigned long compute_backup_blocks(int nchunk
, int ochunk
,
821 unsigned int ndata
, unsigned int odata
);
822 extern char *locate_backup(char *name
);
823 extern char *make_backup(char *name
);
825 extern int save_stripes(int *source
, unsigned long long *offsets
,
826 int raid_disks
, int chunk_size
, int level
, int layout
,
827 int nwrites
, int *dest
,
828 unsigned long long start
, unsigned long long length
,
830 extern int restore_stripes(int *dest
, unsigned long long *offsets
,
831 int raid_disks
, int chunk_size
, int level
, int layout
,
832 int source
, unsigned long long read_offset
,
833 unsigned long long start
, unsigned long long length
,
837 #define Sendmail "/usr/lib/sendmail -t"
840 #define SYSLOG_FACILITY LOG_DAEMON
841 extern char *map_num_s(mapping_t
*map
, int num
);
842 extern char *map_num(mapping_t
*map
, int num
);
843 extern int map_name(mapping_t
*map
, char *name
);
844 extern mapping_t r0layout
[], r5layout
[], r6layout
[],
845 pers
[], modes
[], faultylayout
[];
846 extern mapping_t consistency_policies
[], sysfs_array_states
[], update_options
[];
848 extern char *map_dev_preferred(int major
, int minor
, int create
,
850 static inline char *map_dev(int major
, int minor
, int create
)
852 return map_dev_preferred(major
, minor
, create
, NULL
);
856 * is_fd_valid() - check file descriptor.
857 * @fd: file descriptor.
859 * The function checks if @fd is nonnegative integer and shall be used only
860 * to verify open() result.
862 static inline int is_fd_valid(int fd
)
868 * is_level456() - check whether given level is between inclusive 4 and 6.
869 * @level: level to check.
871 * Return: true if condition is met, false otherwise
873 static inline bool is_level456(int level
)
875 return (level
>= 4 && level
<= 6);
879 * close_fd() - verify, close and unset file descriptor.
880 * @fd: pointer to file descriptor.
882 * The function closes and invalidates file descriptor if appropriative. It
883 * ignores incorrect file descriptor quitely to simplify error handling.
885 static inline void close_fd(int *fd
)
887 if (is_fd_valid(*fd
) && close(*fd
) == 0)
892 struct metadata_update
;
894 /* 'struct reshape' records the intermediate states of
896 * The starting geometry is converted to the 'before' geometry
897 * by at most an atomic level change. They could be the same.
898 * Similarly the 'after' geometry is converted to the final
899 * geometry by at most a level change.
900 * Note that 'before' and 'after' must have the same level.
901 * 'blocks' is the minimum number of sectors for a reshape unit.
902 * This will be a multiple of the stripe size in each of the
903 * 'before' and 'after' geometries.
904 * If 'blocks' is 0, no restriping is necessary.
905 * 'min_offset_change' is the minimum change to data_offset to
906 * allow the reshape to happen. It is at least the larger of
907 * the old and new chunk sizes, and typically the same as 'blocks'
908 * divided by number of data disks.
912 int parity
; /* number of parity blocks/devices */
917 unsigned long long backup_blocks
;
918 unsigned long long min_offset_change
;
919 unsigned long long stripes
; /* number of old stripes that comprise 'blocks'*/
920 unsigned long long new_size
; /* New size of array in sectors */
923 /* A superswitch provides entry point to a metadata handler.
925 * The superswitch primarily operates on some "metadata" that
926 * is accessed via the 'supertype'.
927 * This metadata has one of three possible sources.
928 * 1/ It is read from a single device. In this case it may not completely
929 * describe the array or arrays as some information might be on other
931 * 2/ It is read from all devices in a container. In this case all
932 * information is present.
933 * 3/ It is created by ->init_super / ->add_to_super. In this case it will
934 * be complete once enough ->add_to_super calls have completed.
936 * When creating an array inside a container, the metadata will be
937 * formed by a combination of 2 and 3. The metadata or the array is read,
938 * then new information is added.
940 * The metadata must sometimes have a concept of a 'current' array
941 * and a 'current' device.
942 * The 'current' array is set by init_super to be the newly created array,
943 * or is set by super_by_fd when it finds it is looking at an array inside
946 * The 'current' device is either the device that the metadata was read from
947 * in case 1, or the last device added by add_to_super in case 3.
948 * Case 2 does not identify a 'current' device.
950 extern struct superswitch
{
952 /* Used to report details of metadata read from a component
953 * device. ->load_super has been called.
955 void (*examine_super
)(struct supertype
*st
, char *homehost
);
956 void (*brief_examine_super
)(struct supertype
*st
, int verbose
);
957 void (*brief_examine_subarrays
)(struct supertype
*st
, int verbose
);
958 void (*export_examine_super
)(struct supertype
*st
);
959 int (*examine_badblocks
)(struct supertype
*st
, int fd
, char *devname
);
960 int (*copy_metadata
)(struct supertype
*st
, int from
, int to
);
962 /* Used to report details of an active array.
963 * ->load_super was possibly given a 'component' string.
965 void (*detail_super
)(struct supertype
*st
, char *homehost
,
967 void (*brief_detail_super
)(struct supertype
*st
, char *subarray
);
968 void (*export_detail_super
)(struct supertype
*st
);
970 /* Optional: platform hardware / firmware details */
971 int (*detail_platform
)(int verbose
, int enumerate_only
, char *controller_path
);
972 int (*export_detail_platform
)(int verbose
, char *controller_path
);
975 * to get uuid to storing in bitmap metadata
976 * and 'reshape' backup-data metadata
977 * To see if a device is being re-added to an array it was part of.
979 void (*uuid_from_super
)(struct supertype
*st
, int uuid
[4]);
981 /* Extract generic details from metadata. This could be details about
982 * the container, or about an individual array within the container.
983 * The determination is made either by:
984 * load_super being given a 'component' string.
985 * validate_geometry determining what to create.
986 * The info includes both array information and device information.
987 * The particular device should be:
988 * The last device added by add_to_super
989 * The device the metadata was loaded from by load_super
990 * If 'map' is present, then it is an array raid_disks long
991 * (raid_disk must already be set and correct) and it is filled
992 * with 1 for slots that are thought to be active and 0 for slots which
993 * appear to be failed/missing.
994 * *info is zeroed out before data is added.
996 void (*getinfo_super
)(struct supertype
*st
, struct mdinfo
*info
, char *map
);
997 struct mdinfo
*(*getinfo_super_disks
)(struct supertype
*st
);
998 /* Check if the given metadata is flagged as belonging to "this"
999 * host. 0 for 'no', 1 for 'yes', -1 for "Don't record homehost"
1001 int (*match_home
)(struct supertype
*st
, char *homehost
);
1003 /* Make one of several generic modifications to metadata
1004 * prior to assembly (or other times).
1005 * sparc2.2 - first bug in early 0.90 metadata
1006 * super-minor - change name of 0.90 metadata
1007 * summaries - 'correct' any redundant data
1008 * resync - mark array as dirty to trigger a resync.
1009 * uuid - set new uuid - only 0.90 or 1.x
1010 * name - change the name of the array (where supported)
1011 * homehost - change which host this array is tied to.
1012 * devicesize - If metadata is at start of device, change recorded
1013 * device size to match actual device size
1014 * byteorder - swap bytes for 0.90 metadata
1016 * force-one - mark that device as uptodate, not old or failed.
1017 * force-array - mark array as clean if it would not otherwise
1019 * assemble - not sure how this is different from force-one...
1020 * linear-grow-new - add a new device to a linear array, but don't
1021 * change the size: so superblock still matches
1022 * linear-grow-update - now change the size of the array.
1023 * writemostly - set the WriteMostly1 bit in the superblock devflags
1024 * readwrite - clear the WriteMostly1 bit in the superblock devflags
1025 * failfast - set the FailFast1 bit in the superblock
1026 * nofailfast - clear the FailFast1 bit
1027 * no-bitmap - clear any record that a bitmap is present.
1028 * bbl - add a bad-block-log if possible
1029 * no-bbl - remove any bad-block-log is it is empty.
1030 * force-no-bbl - remove any bad-block-log even if empty.
1031 * revert-reshape - If a reshape is in progress, modify metadata so
1032 * it will resume going in the opposite direction.
1034 int (*update_super
)(struct supertype
*st
, struct mdinfo
*info
,
1035 enum update_opt update
,
1036 char *devname
, int verbose
,
1037 int uuid_set
, char *homehost
);
1039 /* Create new metadata for new array as described. This could
1040 * be a new container, or an array in a pre-existing container.
1041 * Also used to zero metadata prior to writing it to invalidate old
1044 int (*init_super
)(struct supertype
*st
, mdu_array_info_t
*info
,
1045 struct shape
*s
, char *name
,
1046 char *homehost
, int *uuid
,
1047 unsigned long long data_offset
);
1049 /* update the metadata to include new device, either at create or
1050 * when hot-adding a spare.
1052 int (*add_to_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
,
1053 int fd
, char *devname
,
1054 unsigned long long data_offset
);
1055 /* update the metadata to delete a device,
1056 * when hot-removing.
1058 int (*remove_from_super
)(struct supertype
*st
, mdu_disk_info_t
*dinfo
);
1060 /* Write metadata to one device when fixing problems or adding
1063 int (*store_super
)(struct supertype
*st
, int fd
);
1065 /* Write all metadata for this array.
1067 int (*write_init_super
)(struct supertype
*st
);
1068 /* Check if metadata read from one device is compatible with an array,
1069 * used when assembling an array, or pseudo-assembling was with
1070 * "--examine --brief"
1071 * If "st" has not yet been loaded the superblock from, "tst" is
1072 * moved in, otherwise the superblock in 'st' is compared with
1075 int (*compare_super
)(struct supertype
*st
, struct supertype
*tst
,
1077 /* Load metadata from a single device. If 'devname' is not NULL
1078 * print error messages as appropriate */
1079 int (*load_super
)(struct supertype
*st
, int fd
, char *devname
);
1080 /* 'fd' is a 'container' md array - load array metadata from the
1083 int (*load_container
)(struct supertype
*st
, int fd
, char *devname
);
1084 /* If 'arg' is a valid name of this metadata type, allocate and
1085 * return a 'supertype' for the particular minor version */
1086 struct supertype
* (*match_metadata_desc
)(char *arg
);
1087 /* If a device has the given size, and the data_offset has been
1088 * requested - work out how much space is available for data.
1089 * This involves adjusting for reserved space (e.g. bitmaps)
1090 * and for any rounding.
1091 * 'mdadm' only calls this for existing arrays where a possible
1092 * spare is being added. However some super-handlers call it
1093 * internally from validate_geometry when creating an array.
1095 __u64 (*avail_size
)(struct supertype
*st
, __u64 size
,
1096 unsigned long long data_offset
);
1098 * Return spare criteria for array:
1099 * - minimum disk size can be used in array;
1100 * - sector size can be used in array.
1101 * Return values: 0 - for success and -EINVAL on error.
1103 int (*get_spare_criteria
)(struct supertype
*st
,
1104 struct spare_criteria
*sc
);
1105 /* Find somewhere to put a bitmap - possibly auto-size it - and
1106 * update the metadata to record this. The array may be newly
1107 * created, in which case data_size may be updated, or it might
1108 * already exist. Metadata handler can know if init_super
1109 * has been called, but not write_init_super.
1113 int (*add_internal_bitmap
)(struct supertype
*st
, int *chunkp
,
1114 int delay
, int write_behind
,
1115 unsigned long long size
, int may_change
, int major
);
1116 /* Perform additional setup required to activate a bitmap.
1118 int (*set_bitmap
)(struct supertype
*st
, struct mdinfo
*info
);
1119 /* Seek 'fd' to start of write-intent-bitmap. Must be an
1120 * md-native format bitmap
1122 int (*locate_bitmap
)(struct supertype
*st
, int fd
, int node_num
);
1123 /* if add_internal_bitmap succeeded for existing array, this
1126 int (*write_bitmap
)(struct supertype
*st
, int fd
, enum bitmap_update update
);
1127 /* Free the superblock and any other allocated data */
1128 void (*free_super
)(struct supertype
*st
);
1130 /* validate_geometry is called with an st returned by
1131 * match_metadata_desc.
1132 * It should check that the geometry described is compatible with
1133 * the metadata type. It will be called repeatedly as devices
1134 * added to validate changing size and new devices. If there are
1135 * inter-device dependencies, it should record sufficient details
1136 * so these can be validated.
1137 * Both 'size' and '*freesize' are in sectors. chunk is KiB.
1139 * 1: everything is OK
1140 * 0: not OK for some reason - if 'verbose', then error was reported.
1141 * -1: st->sb was NULL, 'subdev' is a member of a container of this
1142 * type, but array is not acceptable for some reason
1143 * message was reported even if verbose is 0.
1145 int (*validate_geometry
)(struct supertype
*st
, int level
, int layout
,
1147 int *chunk
, unsigned long long size
,
1148 unsigned long long data_offset
,
1149 char *subdev
, unsigned long long *freesize
,
1150 int consistency_policy
, int verbose
);
1152 /* Return a linked list of 'mdinfo' structures for all arrays
1153 * in the container. For non-containers, it is like
1154 * getinfo_super with an allocated mdinfo.*/
1155 struct mdinfo
*(*container_content
)(struct supertype
*st
, char *subarray
);
1156 /* query the supertype for default geometry */
1157 void (*default_geometry
)(struct supertype
*st
, int *level
, int *layout
, int *chunk
); /* optional */
1158 /* Permit subarray's to be deleted from inactive containers */
1159 int (*kill_subarray
)(struct supertype
*st
,
1160 char *subarray_id
); /* optional */
1162 * update_subarray() - Permit subarray to be modified.
1164 * @subarray: Subarray name.
1165 * @update: Update option.
1166 * @ident: Optional identifiers.
1168 int (*update_subarray
)(struct supertype
*st
, char *subarray
,
1169 enum update_opt update
, struct mddev_ident
*ident
);
1170 /* Check if reshape is supported for this external format.
1171 * st is obtained from super_by_fd() where st->subarray[0] is
1172 * initialized to indicate if reshape is being performed at the
1173 * container or subarray level
1175 #define APPLY_METADATA_CHANGES 1
1176 #define ROLLBACK_METADATA_CHANGES 0
1178 int (*reshape_super
)(struct supertype
*st
,
1179 unsigned long long size
, int level
,
1180 int layout
, int chunksize
, int raid_disks
,
1181 int delta_disks
, char *backup
, char *dev
,
1183 int verbose
); /* optional */
1184 int (*manage_reshape
)( /* optional */
1185 int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1186 struct supertype
*st
, unsigned long blocks
,
1187 int *fds
, unsigned long long *offsets
,
1188 int dests
, int *destfd
, unsigned long long *destoffsets
);
1191 int (*open_new
)(struct supertype
*c
, struct active_array
*a
,
1194 /* Tell the metadata handler the current state of the array.
1195 * This covers whether it is known to be consistent (no pending writes)
1196 * and how far along a resync is known to have progressed
1197 * (in a->resync_start).
1198 * resync status is really irrelevant if the array is not consistent,
1199 * but some metadata (DDF!) have a place to record the distinction.
1200 * If 'consistent' is '2', then the array can mark it dirty if a
1201 * resync/recovery/whatever is required, or leave it clean if not.
1202 * Return value is 0 dirty (not consistent) and 1 if clean.
1203 * it is only really important if consistent is passed in as '2'.
1205 int (*set_array_state
)(struct active_array
*a
, int consistent
);
1207 /* When the state of a device might have changed, we call set_disk to
1208 * tell the metadata what the current state is.
1209 * Typically this happens on spare->in_sync and (spare|in_sync)->faulty
1211 * set_disk might be called when the state of the particular disk has
1212 * not in fact changed.
1214 void (*set_disk
)(struct active_array
*a
, int n
, int state
);
1215 void (*sync_metadata
)(struct supertype
*st
);
1216 void (*process_update
)(struct supertype
*st
,
1217 struct metadata_update
*update
);
1218 /* Prepare updates allocates extra memory that might be
1219 * needed. If the update cannot be understood, return 0.
1221 int (*prepare_update
)(struct supertype
*st
,
1222 struct metadata_update
*update
);
1224 /* activate_spare will check if the array is degraded and, if it
1225 * is, try to find some spare space in the container.
1226 * On success, it add appropriate updates (For process_update) to
1227 * to the 'updates' list and returns a list of 'mdinfo' identifying
1228 * the device, or devices as there might be multiple missing
1229 * devices and multiple spares available.
1231 struct mdinfo
*(*activate_spare
)(struct active_array
*a
,
1232 struct metadata_update
**updates
);
1234 * Return statically allocated string that represents metadata specific
1235 * controller domain of the disk. The domain is used in disk domain
1236 * matching functions. Disks belong to the same domain if the they have
1237 * the same domain from mdadm.conf and belong the same metadata domain.
1238 * Returning NULL or not providing this handler means that metadata
1239 * does not distinguish the differences between disks that belong to
1240 * different controllers. They are in the domain specified by
1241 * configuration file (mdadm.conf).
1242 * In case when the metadata has the notion of domains based on disk
1243 * it shall return NULL for disks that do not belong to the controller
1244 * the supported domains. Such disks will form another domain and won't
1245 * be mixed with supported ones.
1247 const char *(*get_disk_controller_domain
)(const char *path
);
1249 /* for external backup area */
1250 int (*recover_backup
)(struct supertype
*st
, struct mdinfo
*info
);
1252 /* validate container after assemble */
1253 int (*validate_container
)(struct mdinfo
*info
);
1255 /* write initial empty PPL on device */
1256 int (*write_init_ppl
)(struct supertype
*st
, struct mdinfo
*info
, int fd
);
1258 /* validate ppl before assemble */
1259 int (*validate_ppl
)(struct supertype
*st
, struct mdinfo
*info
,
1260 struct mdinfo
*disk
);
1262 /* records new bad block in metadata */
1263 int (*record_bad_block
)(struct active_array
*a
, int n
,
1264 unsigned long long sector
, int length
);
1266 /* clears bad block from metadata */
1267 int (*clear_bad_block
)(struct active_array
*a
, int n
,
1268 unsigned long long sector
, int length
);
1270 /* get list of bad blocks from metadata */
1271 struct md_bb
*(*get_bad_blocks
)(struct active_array
*a
, int n
);
1273 int swapuuid
; /* true if uuid is bigending rather than hostendian */
1275 const char *name
; /* canonical metadata name */
1278 extern struct superswitch super0
, super1
;
1279 extern struct superswitch super_imsm
, super_ddf
;
1280 extern struct superswitch mbr
, gpt
;
1282 void imsm_set_no_platform(int v
);
1284 struct metadata_update
{
1287 void *space
; /* allocated space that monitor will use */
1288 void **space_list
; /* list of allocated spaces that monitor can
1289 * use or that it returned.
1291 struct metadata_update
*next
;
1294 /* A supertype holds a particular collection of metadata.
1295 * It identifies the metadata type by the superswitch, and the particular
1296 * sub-version of that metadata type.
1297 * metadata read in or created is stored in 'sb' and 'info'.
1298 * There are also fields used by mdmon to track containers.
1300 * A supertype may refer to:
1301 * Just an array, possibly in a container
1302 * A container, not identifying any particular array
1303 * Info read from just one device, not yet fully describing the array/container.
1306 * A supertype is created by:
1312 struct superswitch
*ss
;
1315 char container_devnm
[32]; /* devnm of container */
1318 void *other
; /* Hack used to convert v0.90 to v1.0 */
1319 unsigned long long devsize
;
1320 unsigned long long data_offset
; /* used by v1.x only */
1321 int ignore_hw_compat
; /* used to inform metadata handlers that it should ignore
1322 HW/firmware related incompatability to load metadata.
1323 Used when examining metadata to display content of disk
1324 when user has no hw/firmare compatible system.
1326 struct metadata_update
*updates
;
1327 struct metadata_update
**update_tail
;
1329 /* extra stuff used by mdmon */
1330 struct active_array
*arrays
;
1331 int sock
; /* listen to external programs */
1332 char devnm
[32]; /* e.g. md0. This appears in metadata_version:
1340 struct mdinfo
*devs
;
1344 extern struct supertype
*super_by_fd(int fd
, char **subarray
);
1345 enum guess_types
{ guess_any
, guess_array
, guess_partitions
};
1346 extern struct supertype
*guess_super_type(int fd
, enum guess_types guess_type
);
1347 static inline struct supertype
*guess_super(int fd
) {
1348 return guess_super_type(fd
, guess_any
);
1350 extern struct supertype
*dup_super(struct supertype
*st
);
1351 extern int get_dev_size(int fd
, char *dname
, unsigned long long *sizep
);
1352 extern int get_dev_sector_size(int fd
, char *dname
, unsigned int *sectsizep
);
1353 extern int must_be_container(int fd
);
1354 extern int dev_size_from_id(dev_t id
, unsigned long long *size
);
1355 extern int dev_sector_size_from_id(dev_t id
, unsigned int *size
);
1356 void wait_for(char *dev
, int fd
);
1359 * Data structures for policy management.
1360 * Each device can have a policy structure that lists
1361 * various name/value pairs each possibly with a metadata associated.
1362 * The policy list is sorted by name/value/metadata
1365 struct dev_policy
*next
;
1366 char *name
; /* None of these strings are allocated. They are
1367 * all just references to strings which are known
1368 * to exist elsewhere.
1369 * name and metadata can be compared by address equality.
1371 const char *metadata
;
1375 extern char pol_act
[], pol_domain
[], pol_metadata
[], pol_auto
[];
1377 /* iterate over the sublist starting at list, having the same
1378 * 'name' as 'list', and matching the given metadata (Where
1379 * NULL matches anything
1381 #define pol_for_each(item, list, _metadata) \
1383 item && item->name == list->name; \
1384 item = item->next) \
1385 if (!(!_metadata || !item->metadata || _metadata == item->metadata)) \
1389 * policy records read from mdadm are largely just name-value pairs.
1390 * The names are constants, not strdupped
1393 struct pol_rule
*next
;
1394 char *type
; /* rule_policy or rule_part */
1399 char *dups
; /* duplicates of 'value' with a partNN appended */
1403 extern char rule_policy
[], rule_part
[];
1404 extern char rule_path
[], rule_type
[];
1405 extern char type_part
[], type_disk
[];
1407 extern void policyline(char *line
, char *type
);
1408 extern void policy_add(char *type
, ...);
1409 extern void policy_free(void);
1411 extern struct dev_policy
*path_policy(char **paths
, char *type
);
1412 extern struct dev_policy
*disk_policy(struct mdinfo
*disk
);
1413 extern struct dev_policy
*devid_policy(int devid
);
1414 extern void dev_policy_free(struct dev_policy
*p
);
1416 //extern void pol_new(struct dev_policy **pol, char *name, char *val, char *metadata);
1417 extern void pol_add(struct dev_policy
**pol
, char *name
, char *val
, char *metadata
);
1418 extern struct dev_policy
*pol_find(struct dev_policy
*pol
, char *name
);
1420 enum policy_action
{
1424 act_spare
, /* This only applies to bare devices */
1425 act_spare_same_slot
, /* this allows non-bare devices,
1426 * but only if recent removal */
1427 act_force_spare
, /* this allow non-bare devices in any case */
1431 extern int policy_action_allows(struct dev_policy
*plist
, const char *metadata
,
1432 enum policy_action want
);
1433 extern int disk_action_allows(struct mdinfo
*disk
, const char *metadata
,
1434 enum policy_action want
);
1437 struct domainlist
*next
;
1441 extern int domain_test(struct domainlist
*dom
, struct dev_policy
*pol
,
1442 const char *metadata
);
1443 extern struct domainlist
*domain_from_array(struct mdinfo
*mdi
,
1444 const char *metadata
);
1445 extern void domainlist_add_dev(struct domainlist
**dom
, int devid
,
1446 const char *metadata
);
1447 extern void domain_free(struct domainlist
*dl
);
1448 extern void domain_merge(struct domainlist
**domp
, struct dev_policy
*pol
,
1449 const char *metadata
);
1450 void domain_add(struct domainlist
**domp
, char *domain
);
1452 extern void policy_save_path(char *id_path
, struct map_ent
*array
);
1453 extern int policy_check_path(struct mdinfo
*disk
, struct map_ent
*array
);
1455 extern void sysfs_rules_apply(char *devnm
, struct mdinfo
*dev
);
1456 extern void sysfsline(char *line
);
1462 #define HAVE_NFTW we assume
1466 # include <features.h>
1467 # ifndef __UCLIBC_HAS_LFS__
1468 # define lseek64 lseek
1470 # ifndef __UCLIBC_HAS_FTW__
1480 #if defined(__KLIBC__)
1496 extern int add_dev(const char *name
, const struct stat
*stb
, int flag
, struct FTW
*s
);
1498 extern int Manage_ro(char *devname
, int fd
, int readonly
);
1499 extern int Manage_run(char *devname
, int fd
, struct context
*c
);
1500 extern int Manage_stop(char *devname
, int fd
, int quiet
,
1502 extern int Manage_subdevs(char *devname
, int fd
,
1503 struct mddev_dev
*devlist
, int verbose
, int test
,
1504 enum update_opt update
, int force
);
1505 extern int autodetect(void);
1506 extern int Grow_Add_device(char *devname
, int fd
, char *newdev
);
1507 extern int Grow_addbitmap(char *devname
, int fd
,
1508 struct context
*c
, struct shape
*s
);
1509 extern int Grow_reshape(char *devname
, int fd
,
1510 struct mddev_dev
*devlist
,
1511 struct context
*c
, struct shape
*s
);
1512 extern int Grow_restart(struct supertype
*st
, struct mdinfo
*info
,
1513 int *fdlist
, int cnt
, char *backup_file
, int verbose
);
1514 extern int Grow_continue(int mdfd
, struct supertype
*st
,
1515 struct mdinfo
*info
, char *backup_file
,
1516 int forked
, int freeze_reshape
);
1517 extern int Grow_consistency_policy(char *devname
, int fd
,
1518 struct context
*c
, struct shape
*s
);
1520 extern int restore_backup(struct supertype
*st
,
1521 struct mdinfo
*content
,
1524 char **backup_filep
,
1526 extern int Grow_continue_command(char *devname
, int fd
,
1527 char *backup_file
, int verbose
);
1529 extern int Assemble(struct supertype
*st
, char *mddev
,
1530 struct mddev_ident
*ident
,
1531 struct mddev_dev
*devlist
,
1534 extern int Build(char *mddev
, struct mddev_dev
*devlist
,
1535 struct shape
*s
, struct context
*c
);
1537 extern int Create(struct supertype
*st
, char *mddev
,
1538 char *name
, int *uuid
,
1539 int subdevs
, struct mddev_dev
*devlist
,
1543 extern int Detail(char *dev
, struct context
*c
);
1544 extern int Detail_Platform(struct superswitch
*ss
, int scan
, int verbose
, int export
, char *controller_path
);
1545 extern int Query(char *dev
);
1546 extern int ExamineBadblocks(char *devname
, int brief
, struct supertype
*forcest
);
1547 extern int Examine(struct mddev_dev
*devlist
, struct context
*c
,
1548 struct supertype
*forcest
);
1549 extern int Monitor(struct mddev_dev
*devlist
,
1550 char *mailaddr
, char *alert_cmd
,
1552 int daemonise
, int oneshot
,
1553 int dosyslog
, char *pidfile
, int increments
,
1556 extern int Kill(char *dev
, struct supertype
*st
, int force
, int verbose
, int noexcl
);
1557 extern int Kill_subarray(char *dev
, char *subarray
, int verbose
);
1558 extern int Update_subarray(char *dev
, char *subarray
, enum update_opt update
, struct mddev_ident
*ident
, int quiet
);
1559 extern int Wait(char *dev
);
1560 extern int WaitClean(char *dev
, int verbose
);
1561 extern int SetAction(char *dev
, char *action
);
1563 extern int Incremental(struct mddev_dev
*devlist
, struct context
*c
,
1564 struct supertype
*st
);
1565 extern void RebuildMap(void);
1566 extern int IncrementalScan(struct context
*c
, char *devnm
);
1567 extern int IncrementalRemove(char *devname
, char *path
, int verbose
);
1568 extern int CreateBitmap(char *filename
, int force
, char uuid
[16],
1569 unsigned long chunksize
, unsigned long daemon_sleep
,
1570 unsigned long write_behind
,
1571 unsigned long long array_size
,
1573 extern int ExamineBitmap(char *filename
, int brief
, struct supertype
*st
);
1574 extern int IsBitmapDirty(char *filename
);
1575 extern int Write_rules(char *rule_name
);
1576 extern int bitmap_update_uuid(int fd
, int *uuid
, int swap
);
1578 /* calculate the size of the bitmap given the array size and bitmap chunksize */
1579 static inline unsigned long long
1580 bitmap_bits(unsigned long long array_size
, unsigned long chunksize
)
1582 return (array_size
* 512 + chunksize
- 1) / chunksize
;
1585 extern int Dump_metadata(char *dev
, char *dir
, struct context
*c
,
1586 struct supertype
*st
);
1587 extern int Restore_metadata(char *dev
, char *dir
, struct context
*c
,
1588 struct supertype
*st
, int only
);
1590 int md_array_valid(int fd
);
1591 int md_array_active(int fd
);
1592 int md_array_is_active(struct mdinfo
*info
);
1593 int md_get_array_info(int fd
, struct mdu_array_info_s
*array
);
1594 int md_set_array_info(int fd
, struct mdu_array_info_s
*array
);
1595 int md_get_disk_info(int fd
, struct mdu_disk_info_s
*disk
);
1596 extern int get_linux_version(void);
1597 extern int mdadm_version(char *version
);
1598 extern unsigned long long parse_size(char *size
);
1599 extern int parse_uuid(char *str
, int uuid
[4]);
1600 int default_layout(struct supertype
*st
, int level
, int verbose
);
1601 extern int is_near_layout_10(int layout
);
1602 extern int parse_layout_10(char *layout
);
1603 extern int parse_layout_faulty(char *layout
);
1604 extern int parse_num(int *dest
, const char *num
);
1605 extern int parse_cluster_confirm_arg(char *inp
, char **devname
, int *slot
);
1606 extern int check_ext2(int fd
, char *name
);
1607 extern int check_reiser(int fd
, char *name
);
1608 extern int check_raid(int fd
, char *name
);
1609 extern int check_partitions(int fd
, char *dname
,
1610 unsigned long long freesize
,
1611 unsigned long long size
);
1612 extern int fstat_is_blkdev(int fd
, char *devname
, dev_t
*rdev
);
1613 extern int stat_is_blkdev(char *devname
, dev_t
*rdev
);
1615 extern bool is_dev_alive(char *path
);
1616 extern int get_mdp_major(void);
1617 extern int get_maj_min(char *dev
, int *major
, int *minor
);
1618 extern bool is_bit_set(int *val
, unsigned char index
);
1619 extern int dev_open(char *dev
, int flags
);
1620 extern int open_dev(char *devnm
);
1621 extern void reopen_mddev(int mdfd
);
1622 extern int open_dev_flags(char *devnm
, int flags
);
1623 extern int open_dev_excl(char *devnm
);
1624 extern int is_standard(char *dev
, int *nump
);
1625 extern int same_dev(char *one
, char *two
);
1626 extern int compare_paths (char* path1
,char* path2
);
1627 extern void enable_fds(int devices
);
1628 extern void manage_fork_fds(int close_all
);
1629 extern int continue_via_systemd(char *devnm
, char *service_name
, char *prefix
);
1631 extern void ident_init(struct mddev_ident
*ident
);
1633 extern int parse_auto(char *str
, char *msg
, int config
);
1634 extern struct mddev_ident
*conf_get_ident(char *dev
);
1635 extern struct mddev_dev
*conf_get_devs(void);
1636 extern int conf_test_dev(char *devname
);
1637 extern int conf_test_metadata(const char *version
, struct dev_policy
*pol
, int is_homehost
);
1638 extern struct createinfo
*conf_get_create_info(void);
1639 extern void set_conffile(char *file
);
1640 extern char *conf_get_mailaddr(void);
1641 extern char *conf_get_mailfrom(void);
1642 extern char *conf_get_program(void);
1643 extern char *conf_get_homehost(int *require_homehostp
);
1644 extern char *conf_get_homecluster(void);
1645 extern int conf_get_monitor_delay(void);
1646 extern char *conf_line(FILE *file
);
1647 extern char *conf_word(FILE *file
, int allow_key
);
1648 extern void print_quoted(char *str
);
1649 extern void print_escape(char *str
);
1650 extern int use_udev(void);
1651 extern unsigned long GCD(unsigned long a
, unsigned long b
);
1652 extern int conf_name_is_free(char *name
);
1653 extern bool is_devname_ignore(char *devname
);
1654 extern bool is_devname_md_numbered(const char *devname
);
1655 extern bool is_devname_md_d_numbered(const char *devname
);
1656 extern int conf_verify_devnames(struct mddev_ident
*array_list
);
1657 extern int devname_matches(char *name
, char *match
);
1658 extern struct mddev_ident
*conf_match(struct supertype
*st
,
1659 struct mdinfo
*info
,
1661 int verbose
, int *rvp
);
1663 extern void free_line(char *line
);
1664 extern int match_oneof(char *devices
, char *devname
);
1665 extern void uuid_from_super(int uuid
[4], mdp_super_t
*super
);
1666 extern const int uuid_zero
[4];
1667 extern int same_uuid(int a
[4], int b
[4], int swapuuid
);
1668 extern void copy_uuid(void *a
, int b
[4], int swapuuid
);
1669 extern char *__fname_from_uuid(int id
[4], int swap
, char *buf
, char sep
);
1670 extern char *fname_from_uuid(struct supertype
*st
,
1671 struct mdinfo
*info
, char *buf
, char sep
);
1672 extern unsigned long calc_csum(void *super
, int bytes
);
1673 extern int enough(int level
, int raid_disks
, int layout
, int clean
,
1675 extern int ask(char *mesg
);
1676 extern unsigned long long get_component_size(int fd
);
1677 extern void remove_partitions(int fd
);
1678 extern int test_partition(int fd
);
1679 extern int test_partition_from_id(dev_t id
);
1680 extern int get_data_disks(int level
, int layout
, int raid_disks
);
1681 extern unsigned long long calc_array_size(int level
, int raid_disks
, int layout
,
1682 int chunksize
, unsigned long long devsize
);
1683 extern int flush_metadata_updates(struct supertype
*st
);
1684 extern void append_metadata_update(struct supertype
*st
, void *buf
, int len
);
1685 extern int assemble_container_content(struct supertype
*st
, int mdfd
,
1686 struct mdinfo
*content
,
1688 char *chosen_name
, int *result
);
1690 #define INCR_UNSAFE 2
1691 #define INCR_ALREADY 4
1693 extern struct mdinfo
*container_choose_spares(struct supertype
*st
,
1694 struct spare_criteria
*criteria
,
1695 struct domainlist
*domlist
,
1697 const char *metadata
, int get_one
);
1698 extern int move_spare(char *from_devname
, char *to_devname
, dev_t devid
);
1699 extern int add_disk(int mdfd
, struct supertype
*st
,
1700 struct mdinfo
*sra
, struct mdinfo
*info
);
1701 extern int remove_disk(int mdfd
, struct supertype
*st
,
1702 struct mdinfo
*sra
, struct mdinfo
*info
);
1703 extern int hot_remove_disk(int mdfd
, unsigned long dev
, int force
);
1704 extern int sys_hot_remove_disk(int statefd
, int force
);
1705 extern int set_array_info(int mdfd
, struct supertype
*st
, struct mdinfo
*info
);
1706 unsigned long long min_recovery_start(struct mdinfo
*array
);
1708 extern char *human_size(long long bytes
);
1709 extern char *human_size_brief(long long bytes
, int prefix
);
1710 extern void print_r10_layout(int layout
);
1712 extern char *find_free_devnm(int use_partitions
);
1714 extern void put_md_name(char *name
);
1715 extern char *devid2kname(dev_t devid
);
1716 extern char *devid2devnm(dev_t devid
);
1717 extern dev_t
devnm2devid(char *devnm
);
1718 extern char *get_md_name(char *devnm
);
1720 extern char DefaultConfFile
[];
1722 extern int create_mddev(char *dev
, char *name
, int autof
, int trustworthy
,
1723 char *chosen
, int block_udev
);
1724 /* values for 'trustworthy' */
1726 #define LOCAL_ANY 10
1729 extern int open_mddev(char *dev
, int report_errors
);
1730 extern int is_mddev(char *dev
);
1731 extern int open_container(int fd
);
1732 extern int metadata_container_matches(char *metadata
, char *devnm
);
1733 extern int metadata_subdev_matches(char *metadata
, char *devnm
);
1734 extern int is_container_member(struct mdstat_ent
*ent
, char *devname
);
1735 extern int is_subarray_active(char *subarray
, char *devname
);
1736 extern int open_subarray(char *dev
, char *subarray
, struct supertype
*st
, int quiet
);
1737 extern struct superswitch
*version_to_superswitch(char *vers
);
1739 extern int mdmon_running(char *devnm
);
1740 extern int mdmon_pid(char *devnm
);
1741 extern int check_env(char *name
);
1742 extern __u32
random32(void);
1743 extern void random_uuid(__u8
*buf
);
1744 extern int start_mdmon(char *devnm
);
1746 extern int child_monitor(int afd
, struct mdinfo
*sra
, struct reshape
*reshape
,
1747 struct supertype
*st
, unsigned long stripes
,
1748 int *fds
, unsigned long long *offsets
,
1749 int dests
, int *destfd
, unsigned long long *destoffsets
);
1750 void abort_reshape(struct mdinfo
*sra
);
1752 void *super1_make_v0(struct supertype
*st
, struct mdinfo
*info
, mdp_super_t
*sb0
);
1754 extern char *stat2kname(struct stat
*st
);
1755 extern char *fd2kname(int fd
);
1756 extern char *stat2devnm(struct stat
*st
);
1757 bool stat_is_md_dev(struct stat
*st
);
1758 extern char *fd2devnm(int fd
);
1759 extern void udev_block(char *devnm
);
1760 extern void udev_unblock(void);
1762 extern int in_initrd(void);
1765 void *cmap_handle
; /* corosync lib related */
1767 int (*initialize
)(cmap_handle_t
*handle
);
1768 int (*get_string
)(cmap_handle_t handle
,
1771 int (*finalize
)(cmap_handle_t handle
);
1774 extern void set_cmap_hooks(void);
1775 extern void set_hooks(void);
1778 void *dlm_handle
; /* dlm lib related */
1780 dlm_lshandle_t (*create_lockspace
)(const char *name
,
1782 dlm_lshandle_t (*open_lockspace
)(const char *name
);
1783 int (*release_lockspace
)(const char *name
, dlm_lshandle_t ls
,
1785 int (*ls_lock
)(dlm_lshandle_t lockspace
, uint32_t mode
,
1786 struct dlm_lksb
*lksb
, uint32_t flags
,
1787 const void *name
, unsigned int namelen
,
1788 uint32_t parent
, void (*astaddr
) (void *astarg
),
1789 void *astarg
, void (*bastaddr
) (void *astarg
),
1791 int (*ls_unlock_wait
)(dlm_lshandle_t lockspace
, uint32_t lkid
,
1792 uint32_t flags
, struct dlm_lksb
*lksb
);
1793 int (*ls_get_fd
)(dlm_lshandle_t ls
);
1794 int (*dispatch
)(int fd
);
1797 extern int get_cluster_name(char **name
);
1798 extern int dlm_funs_ready(void);
1799 extern int cluster_get_dlmlock(void);
1800 extern int cluster_release_dlmlock(void);
1801 extern void set_dlm_hooks(void);
1803 #define MSEC_TO_NSEC(msec) ((msec) * 1000000)
1804 #define USEC_TO_NSEC(usec) ((usec) * 1000)
1805 extern void sleep_for(unsigned int sec
, long nsec
, bool wake_after_interrupt
);
1806 extern bool is_directory(const char *path
);
1807 extern bool is_file(const char *path
);
1809 #define _ROUND_UP(val, base) (((val) + (base) - 1) & ~(base - 1))
1810 #define ROUND_UP(val, base) _ROUND_UP(val, (typeof(val))(base))
1811 #define ROUND_UP_PTR(ptr, base) ((typeof(ptr)) \
1812 (ROUND_UP((unsigned long)(ptr), base)))
1814 static inline int is_subarray(char *vers
)
1816 /* The version string for a 'subarray' (an array in a container)
1818 * /containername/componentname for normal read-write arrays
1819 * -containername/componentname for arrays which mdmon must not
1820 * reconfigure. They might be read-only
1821 * or might be undergoing reshape etc.
1822 * containername is e.g. md0, md_d1
1823 * componentname is dependant on the metadata. e.g. '1' 'S1' ...
1825 return (*vers
== '/' || *vers
== '-');
1828 static inline char *to_subarray(struct mdstat_ent
*ent
, char *container
)
1830 return &ent
->metadata_version
[10+strlen(container
)+1];
1834 * signal_s() - Wrapper for sigaction() with signal()-like interface.
1835 * @sig: The signal to set the signal handler to.
1836 * @handler: The signal handler.
1838 * Return: previous handler or SIG_ERR on failure.
1840 static inline sighandler_t
signal_s(int sig
, sighandler_t handler
)
1842 struct sigaction new_act
;
1843 struct sigaction old_act
;
1845 new_act
.sa_handler
= handler
;
1846 new_act
.sa_flags
= 0;
1848 if (sigaction(sig
, &new_act
, &old_act
) == 0)
1849 return old_act
.sa_handler
;
1855 #define dprintf(fmt, arg...) \
1856 fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##arg)
1857 #define dprintf_cont(fmt, arg...) \
1858 fprintf(stderr, fmt, ##arg)
1860 #define dprintf(fmt, arg...) \
1861 ({ if (0) fprintf(stderr, "%s: %s: " fmt, Name, __func__, ##arg); 0; })
1862 #define dprintf_cont(fmt, arg...) \
1863 ({ if (0) fprintf(stderr, fmt, ##arg); 0; })
1866 static inline int xasprintf(char **strp
, const char *fmt
, ...) {
1870 ret
= vasprintf(strp
, fmt
, ap
);
1877 #define pr_err(fmt, args...) fprintf(stderr, "%s: %s: "fmt, Name, __func__, ##args)
1879 #define pr_err(fmt, args...) fprintf(stderr, "%s: "fmt, Name, ##args)
1881 #define cont_err(fmt ...) fprintf(stderr, " " fmt)
1883 #define pr_info(fmt, args...) printf("%s: "fmt, Name, ##args)
1885 void *xmalloc(size_t len
);
1886 void *xrealloc(void *ptr
, size_t len
);
1887 void *xcalloc(size_t num
, size_t size
);
1888 char *xstrdup(const char *str
);
1890 #define LEVEL_MULTIPATH (-4)
1891 #define LEVEL_LINEAR (-1)
1892 #define LEVEL_FAULTY (-5)
1894 /* kernel module doesn't know about these */
1895 #define LEVEL_CONTAINER (-100)
1896 #define LEVEL_UNSUPPORTED (-200)
1898 /* the kernel does know about this one ... */
1899 #define LEVEL_NONE (-1000000)
1903 #define WriteTransient 0
1904 #define ReadTransient 1
1905 #define WritePersistent 2
1906 #define ReadPersistent 3
1907 #define WriteAll 4 /* doesn't go to device */
1908 #define ReadFixable 5
1911 #define ClearErrors 31
1912 #define ClearFaults 30
1914 #define AllPersist 100 /* internal use only */
1915 #define NoPersist 101
1917 #define ModeMask 0x1f
1924 #define minor(x) ((x)&0xff)
1925 #define major(x) (((x)>>8)&0xff)
1926 #define makedev(M,m) (((M)<<8) | (m))
1930 RAID0_ORIG_LAYOUT
= 1,
1931 RAID0_ALT_MULTIZONE_LAYOUT
= 2,
1935 #define ALGORITHM_LEFT_ASYMMETRIC 0
1936 #define ALGORITHM_RIGHT_ASYMMETRIC 1
1937 #define ALGORITHM_LEFT_SYMMETRIC 2
1938 #define ALGORITHM_RIGHT_SYMMETRIC 3
1940 /* Define non-rotating (raid4) algorithms. These allow
1941 * conversion of raid4 to raid5.
1943 #define ALGORITHM_PARITY_0 4 /* P or P,Q are initial devices */
1944 #define ALGORITHM_PARITY_N 5 /* P or P,Q are final devices. */
1946 /* DDF RAID6 layouts differ from md/raid6 layouts in two ways.
1947 * Firstly, the exact positioning of the parity block is slightly
1948 * different between the 'LEFT_*' modes of md and the "_N_*" modes
1950 * Secondly, or order of datablocks over which the Q syndrome is computed
1952 * Consequently we have different layouts for DDF/raid6 than md/raid6.
1953 * These layouts are from the DDFv1.2 spec.
1954 * Interestingly DDFv1.2-Errata-A does not specify N_CONTINUE but
1955 * leaves RLQ=3 as 'Vendor Specific'
1958 #define ALGORITHM_ROTATING_ZERO_RESTART 8 /* DDF PRL=6 RLQ=1 */
1959 #define ALGORITHM_ROTATING_N_RESTART 9 /* DDF PRL=6 RLQ=2 */
1960 #define ALGORITHM_ROTATING_N_CONTINUE 10 /*DDF PRL=6 RLQ=3 */
1962 /* For every RAID5 algorithm we define a RAID6 algorithm
1963 * with exactly the same layout for data and parity, and
1964 * with the Q block always on the last device (N-1).
1965 * This allows trivial conversion from RAID5 to RAID6
1967 #define ALGORITHM_LEFT_ASYMMETRIC_6 16
1968 #define ALGORITHM_RIGHT_ASYMMETRIC_6 17
1969 #define ALGORITHM_LEFT_SYMMETRIC_6 18
1970 #define ALGORITHM_RIGHT_SYMMETRIC_6 19
1971 #define ALGORITHM_PARITY_0_6 20
1972 #define ALGORITHM_PARITY_N_6 ALGORITHM_PARITY_N
1974 /* Define PATH_MAX in case we don't use glibc or standard library does
1975 * not have PATH_MAX defined. Assume max path length is 4K characters.
1978 #define PATH_MAX 4096
1981 #define RESYNC_NONE -1
1982 #define RESYNC_DELAYED -2
1983 #define RESYNC_PENDING -3
1984 #define RESYNC_REMOTE -4
1985 #define RESYNC_UNKNOWN -5
1987 /* When using "GET_DISK_INFO" it isn't certain how high
1988 * we need to check. So we impose an absolute limit of
1989 * MAX_DISKS. This needs to be much more than the largest
1990 * number of devices any metadata can support. Currently
1991 * v1.x can support 1920
1993 #define MAX_DISKS 4096
1995 /* Sometimes the 'size' value passed needs to mean "Maximum".
1996 * In those cases with use MAX_SIZE
2000 /* We want to use unsigned numbers for sector counts, but need
2001 * a value for 'invalid'. Use '1'.
2003 #define INVALID_SECTORS 1
2004 /* And another special number needed for --data_offset=variable */
2005 #define VARIABLE_OFFSET 3
2008 * This is true for native and DDF, IMSM allows 16.
2010 #define MD_NAME_MAX 32
2013 * is_container() - check if @level is &LEVEL_CONTAINER
2014 * @level: level value
2017 * 1 if level is equal to &LEVEL_CONTAINER, 0 otherwise.
2019 static inline int is_container(const int level
)
2021 if (level
== LEVEL_CONTAINER
)